text
stringlengths 254
1.16M
|
|---|
---
title: 'Associations between mortality from COVID-19 and other causes: A state-level
analysis'
authors:
- Anneliese N. Luck
- Andrew C. Stokes
- Katherine Hempstead
- Eugenio Paglino
- Samuel H. Preston
journal: PLOS ONE
year: 2023
pmcid: PMC9987806
doi: 10.1371/journal.pone.0281683
license: CC BY 4.0
---
# Associations between mortality from COVID-19 and other causes: A state-level analysis
## Abstract
### Background
During the COVID-19 pandemic, the high death toll from COVID-19 was accompanied by a rise in mortality from other causes of death. The objective of this study was to identify the relationship between mortality from COVID-19 and changes in mortality from specific causes of death by exploiting spatial variation in these relationships across US states.
### Methods
We use cause-specific mortality data from CDC Wonder and population estimates from the US Census Bureau to examine relationships at the state level between mortality from COVID-19 and changes in mortality from other causes of death. We calculate age-standardized death rates (ASDR) for three age groups, nine underlying causes of death, and all 50 states and the District of Columbia between the first full year of the pandemic (March 2020-February 2021) and the year prior (March 2019-February 2020). We then estimate the relationship between changes in cause-specific ASDR and COVID-19 ASDR using linear regression analysis weighted by the size of the state’s population.
### Results
We estimate that causes of death other than COVID-19 represent $19.6\%$ of the total mortality burden associated with COVID-19 during the first year of the COVID-19 pandemic. At ages 25+, circulatory disease accounted for $51.3\%$ of this burden while dementia ($16.4\%$), other respiratory diseases ($12.4\%$), influenza/pneumonia ($8.7\%$) and diabetes ($8.6\%$) also contribute. In contrast, there was an inverse association across states between COVID-19 death rates and changes in death rates from cancer. We found no state-level association between COVID-19 mortality and rising mortality from external causes.
### Conclusions
States with unusually high death rates from COVID-19 experienced an even larger mortality burden than implied by those rates alone. Circulatory disease served as the most important route through which COVID-19 mortality affected death rates from other causes of death. Dementia and other respiratory diseases made the second and third largest contributions. In contrast, mortality from neoplasms tended to decline in states with the highest death rates from COVID-19. Such information may help to inform state-level responses aimed at easing the full mortality burden of the COVID-19 pandemic.
## Introduction
Public health crises, such as the current COVID-19 pandemic, can pose threats to health and mortality beyond those directly attributable to the disease causing the crisis itself. For example, during the 1918 flu epidemic in the United States, the country faced higher levels of mortality not only from influenza, but from other respiratory diseases and circulatory diseases [1–6]. There are many reasons to anticipate that the COVID-19 pandemic may have a similar impact on mortality from other causes of death. These include disruptions in healthcare and treatment, worsened economic conditions and financial hardship, as well as synergies between COVID-19 and other diseases. Increases in causes of death other than COVID-19 may also partially reflect misdiagnosis of COVID-19 deaths due to inconsistencies in diagnostic and death certification practices [7].
Research on the COVID-19 pandemic has primarily focused on deaths directly attributed to COVID-19 as the underlying cause of death on the death certificate [8]. Less work examines increases in mortality from causes of death other than COVID-19, despite indications that this increase made up a consequential share of the increase in all-cause mortality experienced during the pandemic. Recent estimates suggest that increases in non-COVID-19 mortality accounted for between $12\%$ to $28\%$ of the increase in all-cause mortality experienced during the first year of the pandemic [9–11], with heart disease, diabetes, dementia, and external causes of death implicated as the key drivers of this increase [8, 12–15].
However, most of this emerging work on increases in non-COVID-19 mortality has focused on the national level, despite the great deal of spatial variation in health and mortality across the United States [16–18]. In particular, mortality differences between states have become increasingly salient over the past several decades [19–21]. Indeed, emerging research has begun to document state-level variation in COVID-19 infection and mortality, as well as in responses to the pandemic [22, 23]. Yet the literature focused on mortality during the pandemic for causes other than COVID-19 has generally neglected the spatial patterning of the pandemic’s impact, while other studies highlight salient spatial patterns in pandemic mortality without paying attention to causes of death [9, 10].
The present paper merges these two approaches, with the objective of examining the spatial patterns in mortality change from major causes of death between March 1, 2019 and February 28, 2021. To capture a more complete picture of the association between the COVID-19 pandemic and mortality across the United States, we ask whether states with the highest burden of mortality directly attributable to COVID-19 also had higher mortality from other causes of death. We examine state-level associations between COVID-19 age-standardized death rates and changes in age-standardized death rates for selected causes of death. In doing so, this study captures spatial patterns in cause of death relationships observed during the first year of the pandemic, enabling a better understanding of the full mortality impact of the COVID-19 crisis.
## Data sources
We obtained provisional data on all-cause and cause-specific mortality from CDC Wonder [24]. Data were queried from the provisional mortality statistics tool by five-year age group, underlying cause of death, and state of residence of the decedent for the period March 2019 through February 2021 as reported by February 6, 2022. We incorporated an approximate 12-month lag to account for delays in mortality reporting, especially for external causes of death [25]. Additionally, mid-year population estimates for 2019 and 2020 by age and state were obtained from the Census Bureau’s Vintage 2020 state-level population estimates [26].
## Cause of death categorization
We assigned deaths to COVID-19 if the International Statistical Classification of Diseases and Related Health Problems, Revision Ten, (ICD-10) code U07.1 appeared as the underlying cause of death on the death certificate [27]. Deaths from causes other than COVID-19 were classified into categories to facilitate analyses, which included circulatory disease, diabetes, Alzheimer’s disease and related dementias (ADRD), malignant neoplasms, influenza and pneumonia, other respiratory diseases, external causes, other natural causes, as well as a separate category for signs and symptoms not elsewhere classified.
Building on our previous work on cause-specific mortality during the pandemic [27], this list of selected causes was largely based on the original cause-specific mortality data published by the CDC [28]. In selecting causes of death for analysis, we chose to focus on broad categories rather than individual causes of death to reduce risk of data suppression in states with low cause-specific death counts. These causes of death, excluding the catch-all other natural cause category, account for approximately $85\%$ of all-cause mortality in both periods (Table 1). Many of the included causes overlap with the CDC’s list of leading causes of death in the United States [29] or appear on the CDC’s recently published list of COVID-19 co-morbid conditions [30], including influenza and pneumonia, other respiratory diseases, circulatory diseases, malignant neoplasms, dementia, and diabetes. In addition, we include external causes and signs and symptoms not classified to build off previous work that has examined these causes of death in their analyses [8, 12, 32, 56]. We also offer a supplementary analysis investigated a more detailed list of select external causes commonly examined in the literature, including drug overdose, homicide, suicide, and transport accidents. A full table of ICD-10 codes for each cause of death category can be found in the S1 Table in S1 Appendix.
**Table 1**
| Unnamed: 0 | 3/2019-2/2020 | 3/2019-2/2020.1 | 3/2020-2/2021 | 3/2020-2/2021.1 |
| --- | --- | --- | --- | --- |
| | ASDR | % of all-cause | ASDR | % of all-cause |
| All-Cause | 619.1 | 100.0% | 749.7 | 100.0% |
| COVID-19 | 0.0 | 0.0% | 105.7 | 14.1% |
| Non-COVID-19 | 619.1 | 100.0% | 643.9 | 85.9% |
| Dementia | 60.0 | 9.7% | 66.3 | 8.8% |
| Diabetes | 19.2 | 3.1% | 22.5 | 3.0% |
| Circulatory Disease | 192.9 | 31.2% | 201.8 | 26.9% |
| Influenza & Pneumonia | 10.9 | 1.8% | 10.2 | 1.4% |
| Other Respiratory | 48.8 | 7.9% | 45.1 | 6.0% |
| Malignant Neoplasms | 132.3 | 21.4% | 128.9 | 17.2% |
| Signs Not Classified | 6.1 | 1.0% | 6.5 | 0.9% |
| External Causes | 49.1 | 7.9% | 55.0 | 7.3% |
| Other | 98.8 | 16.0% | 106.7 | 14.2% |
Cause-specific mortality data is suppressed where the death count for the period was less than 10. However, our use of US states as the unit of analysis and focus primarily on leading causes of deaths resulted in marginal suppression in our data. Across the set of nine inclusive causes of death included in the study, less than $1.4\%$ of all-cause deaths were unaccounted for, driven primarily by suppression at younger age groups and in less populous states. Given the limited suppression in our data, we assume no deaths occurred when data was suppressed for a given age group, state of residence, and cause of death combination. We instead address suppression through the use of weighted regressions, which reduces the influence of less populous states where data suppression was most likely.
## Age-standardized death rates
The study compared age-standardized death rates for age groups 25+, 65+, and 25–64 observed in the first full year of the pandemic (March 2020 to February 2021) to baseline pre-pandemic rates during the prior corresponding period (March 2019 to February 2020). Although our time period spanned February to March, we used July 1 population estimates as the mid-period estimate of the population given the unavailability of monthly-level data. Age-specific death rates were then age-standardized using the mid-year 2020 national age distribution for population aged 25+ as well as for ages 25–64 and 65+ [31].
## Analytic approach
Our units of observation are US states. States are a convenient vehicle for investigating disease interrelations since many programs and policies addressed to harness the pandemic were implemented at the state level. Further, states are generally large enough that data suppression due to small numbers in NCHS data releases is relatively uncommon, in contrast to county-level data. However, data suppression may appear in relation to mortality from less prevalent causes of death in our study, such as the more detailed categories of external causes.
All death rates in this analysis were age-standardized using the age distribution of the United States in 2020 in 5-year age groups. The principal focus was on age-standardized death rates (per 100,000) for ages 25+. We also examined age-standardized rates at ages 65+, where chronic diseases dominate, and at working ages 25–64, where external causes of death are more common.
We examine relations between changes in mortality for a particular cause of death and COVID-19 mortality using weighted linear regression analysis. Each state’s observation is weighted by the corresponding size of the state’s population in the appropriate age span. Weighted regression is used so that results better reflect the population distribution of the United States.
The basic regression equation is: ΔMik=αi+βi⋅Mck+ε, where ΔMik = change in age-standardized death rate from cause i in state k Mck = age-standardized death rate from COVID-19 in state k αi = constant term expressing change in mortality from cause i that is unrelated to COVID-19 mortality
βi = increase in mortality from cause i per unit increase in mortality from COVID-19 *We focus* on absolute changes in age-standardized death rates as the most direct measure of the change in frequency of death per person. In doing so, the sum of βi’s across a set of mutually exclusive and exhaustive causes of death, excluding COVID-19, will add to the βi coefficient when all non-COVID-19 mortality is regressed on COVID-19 mortality, Mck. In this fashion, the relation between mortality from non-COVID-19 causes and COVID-19 mortality can be uniquely decomposed into relations for various causes of death. The contribution of a particular cause of death to the relation between COVID-19 and all-cause mortality is assessed by the ratio of the cause-specific βi to the all-cause βi. Additional regressions were also implemented in which age-standardized death rates were standardized (mean = 0 and standard deviation = 1) to further investigate the magnitude of relationship between increases in cause-specific and COVID-19 mortality. All models were implemented using the stats package in R.
## Cause-specific mortality by state
Table 2 presents the estimated all-cause, COVID-19, and non-COVID-19 age-standardized death rates (ASDR) during the year prior to the pandemic (March 1, 2019 to February 28, 2020) and during the first full year of the pandemic (March 1, 2020 to February 28, 2021) across all 50 US states and the District of Columbia. Fig 1 displays the state distribution of age-standardized death rates for COVID-19 and non-COVID-19, as well as across all nine selected causes of death across US states in the first year of the pandemic.
**Fig 1:** *State distribution of cause-specific ASDR for ages 25+ from $\frac{3}{2020}$-$\frac{2}{2021.}$Note: ASDR = age-standardized death rates. Units are deaths per 100,000.* TABLE_PLACEHOLDER:Table 2 As seen in Table 2, the five states with the highest levels of all-cause mortality pre-pandemic were located in the South. This geographic patterning remained consistent during the first year of the pandemic, with these Southern states consistently facing the highest levels of both all-cause and non-COVID-19 mortality [Fig 1]. However, more geographic variation emerged among states with the highest levels of COVID-19 mortality. Though most of the highest COVID-19 mortality states were located in the South, New York and New Jersey topped the list as leaders in COVID-19 mortality. These two states faced dramatic shifts in mortality from 2019 to 2020, having the highest death rates from COVID-19 in the country during the first year of the pandemic yet the being in among the ten states with the lowest mortality in the pre-pandemic year.
The levels of COVID-19 mortality during the first year of the pandemic also proved much more variable across states than non-COVID-19 or all-cause mortality [Table 1]. Across both years, the state with the highest levels of all-cause and non-COVID-19 mortality (Mississippi) had death rates 1.9–2.3 times higher than the state with the lowest all-cause mortality rate (Hawaii). From March 2020 to February 2021, however, the states with the highest levels of COVID-19 mortality (New York and New Jersey) had death rates from COVID-19 ranging between 9.4 to 10.7 times higher than those with the lowest levels of COVID-19 mortality (Hawaii and Vermont).
Fig 1 shows that similar geographic patterns generally hold across the select causes of death examined in this study. In 2020, across nearly every cause of death, the five states with the highest levels of mortality are located in the Southern region of the United States, with the only exceptions being influenza and pneumonia, external cause deaths, other signs and symptoms, and all other causes.
## COVID-19 and changes in cause-specific mortality
Next, we directly examine the relationships between age-standardized death rates from COVID-19 during the first full year of the pandemic (March 1, 2020 to February 28, 2021) and changes in age-standardized death rates from other causes of death between this period and the prior year (March 1, 2019 to February 28, 2020). The relationships are plotted in Fig 2. The line in the Figure was fitted by weighted linear regression, the parameters of which are shown in Table 3.
**Fig 2:** *Association between state-level COVID-19 mortality and change in cause-specific mortality, ages 25+.Note: ASDR = age-standardized death rates. Change is measured as the age-standardized death rate in first full year of the pandemic (March 2020 to February 2021) relative to the same period a year prior (March 2019 to February 2020). Line reflects weighted linear regression where weights are the corresponding size of population in state i. Points above the dotted line indicates states with increases in cause-specific age ASDR, while below indicates decreases. β represents the change in ASDR for a particular cause of death associated with a 1 per 100,000 unit increase in COVID-19 ASDR. States are identified in the non-COVID-19 panel, with the associated abbreviations found in Table 1.* TABLE_PLACEHOLDER:Table 3 The coefficient of the regression of the change in all-cause mortality on COVID-19 mortality is 1.24, meaning that for each one-unit increase in COVID-19 mortality, all-cause mortality rose by 1.24 units [Table 3]. This result implies that each increase of one unit in COVID-19 mortality was accompanied by an increase of 0.24 units in mortality from all other causes of death combined, so that other causes represent $\frac{0.24}{1.24}$ or $19.6\%$ of the total impact of COVID-19 on all-cause mortality.
The causes of death that were associated with an increase in COVID-19 mortality can be gauged by the magnitude of the beta coefficients in cause-specific regressions, which represent the change in mortality for a particular cause of death per unit change in COVID-19 mortality. Across causes of death, these coefficients sum to the beta of 1.24 pertaining to all-cause mortality. The marginal difference between the direct regression of all-cause mortality (1.243) and sum of betas across causes of death (1.240) reflect rounding error and/or suppression of data with small sample sizes. The highest coefficient, 0.125, pertained to circulatory disease. This result suggests that for every 1 unit (1 per 100,000) increase in COVID-19 mortality, there was an associated increase of 0.125 per 100,000 in circulatory mortality. This coefficient was followed by that of dementia (0.040), other respiratory diseases (0.030) and diabetes (0.021). These four causes account for nearly $89\%$ of the 0.243 increase in non-COVID-19 causes of death per unit increase in COVID-19 mortality. Table 3 shows that the relationship between mortality from dementia and from COVID-19 is essentially limited to ages 65+. As a proportion of the increase of $19.6\%$ in all-cause mortality attributable to non-Covid causes of death ages 25+, circulatory disease accounts for $52.9\%$ of this burden while dementia ($21.1\%$), other respiratory diseases ($11.6\%$), influenza/pneumonia ($9.7\%$) and diabetes ($7.1\%$) also contribute.
Malignant neoplasms are the only cause of death showing a significant negative association with COVID-19 mortality at ages 25+. This relationship was concentrated at ages 65+, where a one-unit increase in COVID-19 mortality was found to be associated with a 0.037 unit decrease in cancer mortality. Thus, on average, individuals living in states with higher burdens of COVID-19 mortality also faced declining mortality from cancer.
We observe no association at the state level between COVID-19 mortality and changes in mortality from external causes [Table 3]. To investigate whether the aggregate of external causes is obscuring important relationships involving more detailed external causes of death, we conducted a supplementary analysis that examined mortality from drug overdose, homicide, suicide, transport accidents, and all other external causes of death presented in the S1 Appendix. As shown in S2 Table in S1 Appendix, none of the changes in mortality from external causes was significantly related to mortality from COVID-19.
The associations between COVID-19 mortality and mortality from other causes of death can also be expressed in standard deviation units, i.e., using standardized beta coefficients, which introduce comparable scales for examining the strength of relationships across causes of death. These coefficients are presented in Table 4 and Fig 3. The ordering of causes of death in terms of the strength of associations is fairly similar to that in Table 3, with circulatory disease showing the highest coefficient in both cases. However, influenza/pneumonia, a cause of death with relatively low mortality, emerges as having the second highest beta coefficient when measured in standard deviation units. In fact, among the working-age population (25–64), the beta coefficient on influenza/pneumonia appears to have an even a stronger association with COVID-19 mortality than circulatory disease, a cause of death with relatively high mortality. Thus, once a comparable scaling of their death rates is imposed, death rates from the two infectious diseases are seen to be closely related across states.
**Fig 3:** *Standardized association between state-level COVID-19 mortality and change in cause-specific mortality, ages 25+.Note: ASDR = age-standardized death rates. Change is measured as the age-standardized death rate in first full year of the pandemic (March 2020 to February 2021) relative to the same period a year prior (March 2019 to February 2020), with coefficients standardized (mean = 0 and standard deviation = 1). Line reflects weighted linear regression where weights are the corresponding size of population in state i. Points above the dotted line indicates states with increases in cause-specific age ASDR, while below indicates decreases. β represents the change in ASDR for a particular cause of death in SD units associated with a 1 SD unit increase in COVID-19 ASDR.* TABLE_PLACEHOLDER:Table 4
## Regional trends
Further, the state-level analyses also suggestively reveal regional differences between actual and predicted non-COVID-19 mortality, informing our understanding of the spatial patterning of excess mortality across the United States. The state-level relationship illustrated in the results highlights clear regional patterns in the states that either experienced unexpectedly high or low death rates from COVID-19 mortality relative to their change in non-COVID-19 death rates, with a clear divergence between the states located in the South and Northeast. As evident from Fig 2, Southern states appear to dominate the positive residuals, with all states in the region with one exception lying above the fitted regression line of non-COVID-19 mortality on COVID-19 mortality. This analysis suggests that Southern states generally had more non-COVID-19 mortality than would be predicted based on their COVID-19 mortality during the first year of the pandemic. For example, Mississippi (MS), which faced some of the highest rates of COVID-19 mortality in the country, experienced increases in non-COVID-19 mortality approximately double the level predicted by our analysis [Fig 2].
Conversely, all but one of the states located in the Northeast lie below the regression line, suggesting these states saw smaller increases in non-COVID-19 mortality than anticipated based on the levels of COVID-19 mortality experienced during the first year of the pandemic. Massachusetts (MA) serves as a clear example of this dynamic. Despite facing COVID-19 death rates of nearly 120 per 100,000 in the first year of the pandemic, which would correspond in our analysis to an increase in non-COVID-19 mortality of over 25 per 100,000, the state actually experienced a decrease in non-COVID-19 mortality during the time period. Additionally, this regional patterning appears to be largely consistent across the non-COVID-19 causes of death examined in the study, but particularly so for dementia, circulatory diseases, and external causes [Fig 2].
## Discussion
This study estimates that changes in mortality from causes of death other than COVID-19 represent $19.6\%$ of the total change in mortality between March 2019-February 2020 and March 2020-February 2021 [Table 1]. This estimate of the effects of the COVID-19 pandemic on other causes of death is in line with estimates by the Center for Disease Control that suggest that the contribution of non-COVID-19 deaths range from $12\%$ to $25\%$ of the total change in mortality. The causes of death that contribute the most to the increase in all-cause mortality associated with mortality from COVID-19 are circulatory diseases and dementia. Other causes of death making large contributions are other respiratory diseases (including chronic obstructive pulmonary disease), influenza/pneumonia, and diabetes. Circulatory disease, dementia, and diabetes were also the three causes of death showing the largest mortality increase in the nation between 2019 and 2020, apart from external causes and COVID-19 itself [35]. Below we consider several possible explanations for a positive association between COVID-19 mortality and mortality attributed to the natural causes of death.
## Changes in health care utilization
First, positive associations may relate to the fact that states with high COVID-19 death rates may have had larger declines in health care utilization than states with low COVID-19 death rates due to interruption and delays in the provision of health care services and greater hospital avoidance [33–35]. Chronic conditions for which management requires frequent medical monitoring, such as diabetes and circulatory disease, are likely to be most affected by these reductions [36]. These interruptions may have been exacerbated by increased housing and food insecurity brought on by the pandemic, an effect that may be particularly salient among those living with chronic illnesses or who face acute health emergencies and cannot afford medicines or medical supplies [37–39]. Further, some similar infectious diseases that spread by respiratory routes may be highly correlated across states because they respond to the same set of environmental influences, including shelter-in-place and masking policies. For example, influenza/pneumonia shows the second closest association with COVID-19 mortality when associations are measured in standard deviation units.
## Cause-of-death coding practices
Additionally, diagnostic and coding practices may have resulted in COVID-19 deaths being inappropriately assigned to another underlying cause. CDC guidelines suggest that, if COVID-19 were listed on a death certificate, it should be classified as the underlying cause of death, with pre-existing conditions listed as contributing causes of death [40]. Instead, during the first year of the pandemic, $13\%$ of death certificates that included COVID-19 had it placed as a contributing cause of death [41]. If the proportion of deaths that should have been assigned to COVID-19 but are instead assigned to another underlying cause is roughly constant from state to state, it would create a positive association between COVID-19 mortality and mortality from that other cause. Such a pattern could be obscured or even reversed if there were sufficient interstate variability in diagnostic and coding practices.
Coding confusions are particularly likely when there are synergies between COVID-19 infection and another disease. By synergies, we refer to physiological relations between COVID-19 and a medical condition such that the mortality risk from joint exposure to the two conditions exceeds the sum of risks from the individual exposures. Synergies may also be present by virtue of damage that COVID-19 infection does to various organ systems among people with no prior disease [42–44].
Some prior studies help to cast light on the presence or absence of synergies for certain causes of death. Using data on English cohorts, one such study compared the odds ratio of death rates for people with a particular condition in 2020, when COVID-19 was present, to ratios in 2015–2019, when it was not [45]. Conditions that raised the relative odds of death during the pandemic were dementia, diabetes, hypertension, stroke, and coronary heart disease (insignificantly), while cancer and chronic obstructive pulmonary disease (COPD) had lower odds ratios in 2020 than in 2019. Similarly, Tarazi et al. [ 2021] use data from Medicare beneficiaries that permit the construction of rate ratios in the two periods, finding enrollees with dementia, diabetes, hypertension, cardiac disorders, and COPD were at greater relative risk in 2021 than in 2015–19 [46]. However, as in the English study, Medicare enrollees with cancer were at lower relative risk in 2020 than in the prior period (2015–19), possibly implying an absence of a cancer synergy with COVID-19. A handful of studies have begun to provide additional evidence of synergies between COVID-19 and particular causes of death, including coronary heart disease or cardiovascular disease [47, 48].
## Malignant neoplasms: An anomaly
Interestingly, we find that malignant neoplasms do not behave like other chronic diseases in their relation to COVID-19. States with higher death rates from COVID-19 instead had smaller increases, or larger declines, in mortality from cancer than states with lower mortality from COVID-19. Although somewhat surprising, this finding is not inconsistent with emerging evidence on this topic. Mortality from cancer at the national level has also proved anomalous, showing no change between 2019 and 2020 at ages 25+ while mortality from other major causes was rising [49]. These findings align with the studies mentioned above [45, 46], which found that individuals with cancer did not suffer from unusually high relative risks of death during 2020, the first year of the pandemic, compared to 2015–2019. Interestingly, this experience is also consistent with that of the 1918 influenza epidemic in the United States, when the outbreak was found to be correlated with higher mortality from respiratory and heart disease, but not with that from cancer [45].
However, an absence of synergy between cancer and COVID-19 would not explain why cancer mortality fell in the states with the highest burden of COVID-19 mortality. One hypothesized mechanism that would create this negative association between mortality from COVID-19 and other causes of death, such as cancer mortality, can be termed frailty selection [50]. It is plausible that frailer cancer patients would be more likely to die from COVID-19 than less frail patients when infected, resulting in a remaining cancer population that is somewhat less frail as a result of the pandemic and thus less likely to die from cancer. Although such a process of selection should have been operating with respect to the other diseases as well, its importance may have been obscured by the power of disease synergies, perhaps expressed through greater degrees of diagnostic and coding error. It has also been suggested that, in view of the seriousness of the condition, cancer patients may have been more likely to shield or be compliant with social distancing measures [49]. In states with higher COVID-19 transmission and more COVID-19 deaths, cancer patients may have sheltered-in-place at higher rates than cancer patients in places with less COVID-19 transmission.
Nonetheless, the relationship between cancer and COVID-19 during the pandemic remains surprisingly unclear. Emerging work has shown that heterogeneity in cancer types, as well as variation in cancer epidemiology and practice across countries, have vastly complicated our empirical understanding of this relationship [51, 52]. This study thus adds to the growing body of literature that calls for more research on this topic so that we may better understand the link between cancer and COVID-19.
## External causes of death
Another anomalous cause of death observed in our study is external causes. Research has consistently documented a sharp increase in death rates from external causes during 2020 [8, 12, 32, 53]. This work has pointed to pandemic-related disruptions and hardships as responsible for increases in stress, depression, and substance use, contributing to increased mortality from drug overdose and alcohol abuse [54, 55], as well as traffic accidents [56]. Our results, however, show no association between mortality from COVID-19 and that from external causes across states [Table 2].
It has been pointed out that mortality from drug overdose, a major component of external causes, was rising before the pandemic and that some of the increase in 2020 and 2021 from external causes may reflect a continuation of that trend [17, 19, 20, 54]. Our results suggestively support this interpretation: there is no observed “dose/response” relationship between COVID-19 mortality and mortality from external causes. Instead, there is a very large positive intercept in the external cause regression, the largest for any cause of death at both ages 25+ and 25–64 [Table 2]. This is particularly true for drug overdose deaths among working aged populations [Table 3]. The intercept is an indicator of what changes in mortality from external causes would have been for populations in a state where the death rate from COVID-19 were set at zero, i.e., where there was no COVID-19 mortality experienced during the first year of pandemic. The implication is that external cause mortality would have risen substantially over the period even in the absence of the COVID-19 pandemic.
However, the fact that external cause mortality appears unresponsive to the level of COVID-19 mortality in a state does not imply that there was no impact of the pandemic on external causes of death, only that the relationship between COVID-19 and external cause mortality did not have a prominent spatial component. There are likely indirect effects of COVID-19 that were “universal” across the country that would not be captured in the spatial relationships in our regressions. This may be particularly true for external cause mortality, where increases in certain causes of death–such as drug overdose and homicide–may be partly in response to the pandemic-related economic hardship and associated stress experienced on a national scale [57].
## Other explanatory factors
This discussion does not exhaust the possible reasons why states with higher levels of mortality from COVID-19 would have greater increases in mortality from other causes of death. Over the past several decades, a large body of research has emerged to examine the vast array of sociodemographic, behavioral, environmental, and institutional factors that drive differences in health and mortality across states in the United States [17, 19, 20, 58]. Within the context of the COVID-19 pandemic, many of the factors operating at a state level that raised or lowered death rates from COVID-19 may also be expected to have raised or lowered death rates from another cause.
These could include state-level variation in the extent of co-morbidities, such as obesity, diabetes, and heart disease [59]. For example, the well-documented “stroke belt’ of the Central South, where cardiovascular diseases are highly concentrated, may underlie the spatial patterns that emerge between circulatory diseases and COVID-19 mortality observed in the study [60, 61]. Additionally, prior work has pointed to other pandemic-related factors, such as a state’s efforts to control the pandemic through public health measures [59, 60] or the ability of its healthcare systems to react to the pandemic [62]. These factors, coupled with broader state-level variation in socioeconomic status, racial and ethnic composition, and population density are likely to collectively shape the unequal impact of the pandemic across states [63, 64]. Although conducting a full-scale multivariate analysis of state-specific circumstances that fashioned the mortality response to the pandemic is beyond the scope of this paper, examining the factors that drive the spatial patterns in mortality for major causes of death observed in this study is a promising avenue for future research.
## Limitations
We recognize several limitations in this analysis. First, the full impact of the COVID-19 pandemic on health and mortality may be experienced with a lag, while this analysis solely examines the relationships that emerged in the first full year of the COVID-19 pandemic. For example, delays in diagnoses driven by the pandemic’s disruption of health services may serve to impact mortality primarily in the long-term, which could be missed in the present study given the focus on the first full year of the pandemic [65, 66].
Second, interstate variation in diagnostic and coding practices may work to obscure the true relationship between mortality from COVID-19 and another cause of death. If an absence of diagnostic testing in some places caused certifiers to systematically list a co-morbid condition, rather than COVID-19, as the underlying cause, death rates from comorbid conditions would be inflated while those from COVID-19 would be deflated relative to places where certifiers had full information. Such variation in coding practices is capable of creating an inverse association between COVID-19 and another cause of death. For example, Massachusetts may represent a concrete instance of unusual coding practices. We document unexpectedly high death rates from COVID-19 in Massachusetts relative to the change in non-COVID-19 death rates (Fig 1). This result raises the possibility that deaths from COVID-19 are over-recorded in Massachusetts, a possibility first suggested by Ackley et al [10]. Subsequent developments confirm that Massachusetts was using criteria for a COVID-19 death that were much more liberal than those which were used in other states [67]. Similarly, the fact that we observe Southern states consistently above the regression line predicting non-COVID-19 mortality raises the possibility that non-COVID-19 mortality was systematically higher than recorded in the South, suggesting that many COVID-19 deaths went uncounted there. Despite these apparent regional disparities in coding, the fact that relations between COVID-19 and mortality from most other cause of death are positive is consistent with the idea that states’ classification practices are not dissimilar enough to create inverse relationships. More research is needed to fully understand how inconsistencies in diagnostic and coding practices across states may shape the observed associations between COVID-19 and other causes of death.
Additionally, since mortality data was extracted by age group, state of residence, and cause of death, there is likely some degree of data suppression for less common causes of death in less populous states. We mitigate this concern through the use of weighted regressions, where each state is weighted by the corresponding size of the state’s population, thus reducing the influence of smaller states where suppression is more likely to occur and better reflecting the population distribution of the United States.
Finally, it is possible that the relationship between mortality from COVID-19 and other causes of death changed over the course of the pandemic, particularly across the various peaks in COVID-19 mortality. This may be particularly true as testing became more widespread and coding practices more established. However, given the already granular extraction of mortality data by age group, state, and cause of death, the study focused on mortality changes experienced in the first full year of the pandemic. Future research should investigate the extent to which these findings hold across distinct COVID-19 mortality peaks experienced during the pandemic.
Despite these limitations, this study takes a novel approach to understanding the full impact of the COVID-19 pandemic on mortality in the United States and is among the first to examine spatial patterns in mortality for major causes of death during the pandemic. In theorizing the possible mechanisms that underlie the pandemic’s impact on mortality from non-COVID-19 causes, this study contributes to a more complete understanding of how the COVID-19 pandemic may have directly and indirectly shaped the landscape of mortality in the United States.
## References
1. Scragg R.. **Effect of influenza epidemics on Australian mortality**. *Med J Aust* (1985.0) **142** 98-102. PMID: 3965919
2. Collins SD. **Excess Mortality from Causes Other than Influenza and Pneumonia during Influenza Epidemics**. *Public Health Rep* (1932.0) **47** 2159-79. PMID: 19315373
3. 3Pearl R. Influenza Studies. On Certain General Statistical Aspects of the 1918 Epidemic In American Cities. 1919 Aug 8 [cited 2022 Apr 15]; Available from: https://quod.lib.umich.edu/f/flu/7780flu.0016.877/1/—influenza-studies-on-certain-general-statistical-aspects?view=image
4. Nguyen JL, Yang W, Ito K, Matte TD, Shaman J, Kinney PL. **Seasonal Influenza Infections and Cardiovascular Disease Mortality**. *JAMA Cardiol* (2016.0) **1** 274-81. DOI: 10.1001/jamacardio.2016.0433
5. Chow EJ, Rolfes MA, O’Halloran A, Anderson EJ, Bennett NM, Billing L. **Acute Cardiovascular Events Associated With Influenza in Hospitalized Adults: A Cross-sectional Study**. *Ann Intern Med* (2020.0) **173** 605-13. DOI: 10.7326/M20-1509
6. Almond D, Mazumder B. **The 1918 Influenza Pandemic and Subsequent Health Outcomes: An Analysis of SIPP Data**. *Am Econ Rev* (2005.0) **95** 258-62. DOI: 10.1257/000282805774669943
7. Stokes AC, Lundberg DJ, Bor J, Bibbins-Domingo K. **Excess Deaths During the COVID-19 Pandemic: Implications for US Death Investigation Systems**. *Am J Public Health* (2021.0) **111** S53-4. DOI: 10.2105/AJPH.2021.306331
8. Luck AN, Preston SH, Elo IT, Stokes AC. **The unequal burden of the Covid-19 pandemic: Capturing racial/ethnic disparities in US cause-specific mortality**. *SSM Popul Health* (2022.0) **17** 101012. DOI: 10.1016/j.ssmph.2021.101012
9. Stokes AC, Lundberg DJ, Elo IT, Hempstead K, Bor J, Preston SH. **COVID-19 and excess mortality in the United States: A county-level analysis**. *PLoS Med* (2021.0) **18** e1003571. DOI: 10.1371/journal.pmed.1003571
10. Ackley CA, Lundberg DJ, Ma L, Elo IT, Preston SH, Stokes AC. **County-level estimates of excess mortality associated with COVID-19 in the United States**. *SSM-Population Health* (2022.0) **17**. DOI: 10.1016/j.ssmph.2021.101021
11. Rossen LM, Branum AM, Ahmad FB, Sutton PD, Anderson RN. **Notes from the Field: Update on Excess Deaths Associated with the COVID-19 Pandemic—United States, January 26, 2020–February 27, 2021 | MMWR**. *MMWR Morb Mortal Wkly Rep* (2021.0) **70** 570-1. PMID: 33857065
12. Glei DA. **The US Midlife Mortality Crisis Continues: Excess Cause-Specific Mortality During 2020**. *Am J Epidemiol* (2022.0). DOI: 10.1093/aje/kwac055
13. 13Ruhm CJ. Excess Deaths in the United States During the First Year of COVID-19 [Internet]. NBER Working Paper. National Bureau of Economic Research; 2021. (Working Paper Series). Available from: http://www.nber.org/papers/w29503
14. Shiels MS, Haque AT, Haozous EA, Albert PS, Almeida JS, García-Closas M. **Racial and Ethnic Disparities in Excess Deaths During the COVID-19 Pandemic, March to December 2020**. *Ann Intern Med* (2021.0) **174** 1693-9. DOI: 10.7326/M21-2134
15. Chen Y-H, Stokes AC, Aschmann HE, Chen R, DeVost S, Kiang MV. **Excess natural-cause deaths in California by cause and setting: March 2020 through February 2021**. *PNAS Nexus* (2022.0) **1**. DOI: 10.1093/pnasnexus/pgac079
16. Woolf SH, Schoomaker H. **Life Expectancy and Mortality Rates in the United States, 1959–2017**. *JAMA* (2019.0) **322** 1996-2016. DOI: 10.1001/jama.2019.16932
17. Fenelon A.. **Geographic Divergence in Mortality in the United States**. *Popul Dev Rev* (2013.0) **39** 611-34. DOI: 10.1111/j.1728-4457.2013.00630.x
18. Chetty R, Stepner M, Abraham S, Lin S, Scuderi B, Turner N. **The Association Between Income and Life Expectancy in the United States, 2001–2014**. *JAMA* (2016.0) **315** 1750-66. DOI: 10.1001/jama.2016.4226
19. Montez JK, Beckfield J, Cooney JK, Grumbach JM, Hayward MD, Koytak HZ. **US State Policies, Politics, and Life Expectancy**. *Milbank Q* (2020.0) **98** 668-99. DOI: 10.1111/1468-0009.12469
20. Montez JK, Zajacova A, Hayward MD. **Explaining Inequalities in Women’s Mortality between U.S. States**. *SSM Popul Health* (2016.0) **2** 561-71. DOI: 10.1016/j.ssmph.2016.07.004
21. Miller CE, Vasan RS. **The southern rural health and mortality penalty: A review of regional health inequities in the United States**. *Soc Sci Med* (2021.0) **268** 113443. DOI: 10.1016/j.socscimed.2020.113443
22. Loomba RS, Aggarwal G, Aggarwal S, Flores S, Villarreal EG, Farias JS. **Disparities in case frequency and mortality of coronavirus disease 2019 (COVID-19) among various states in the United States**. *Ann Med* (2021.0) **53** 151-9. DOI: 10.1080/07853890.2020.1840620
23. 23Kerpen P, Moore S, Mulligan CB. A Final Report Card on the States’ Response to COVID-19 [Internet]. National Bureau of Economic Research; 2022. (Working Paper Series). Available from: http://www.nber.org/papers/w29928
24. 24National Center for Health Statistics. Provisional Multiple Cause of Death on CDC WONDER Online Database, released 2021. Data are from the final Multiple cause of Death Files, 2018–2019, and from provisional data for years 2020–2021, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program [Internet]. 2022 [cited 2022 Feb 6]. Available from: http://wonder.cdc.gov/mcd-icd10-provisional.html
25. Ahmad FB, Anderson RN, Knight K, Rossen LM, Sutton PD. **Advancements in the National Vital Statistics System to Meet the Real-Time Data Needs of a Pandemic**. *Am J Public Health* (2021.0) **111** 2133-40. DOI: 10.2105/AJPH.2021.306519
26. 26US Census Bureau. National Population by Characteristics: 2010–2020. 2022.
27. 27National Vital Statistics System. Guidance for Certifying Deaths Due to Coronavirus Disease 2019 (COVID–19). Vital Statistics Reporting Guidance [Internet]. 2020 Apr;Report No. 3. Available from: https://www.cdc.gov/nchs/data/nvss/vsrg/vsrg03-508.pdf
28. 28NCHS/DVS. AH Monthly Provisional Counts of Deaths for Select Causes of Death by Sex, Age, and Race and Hispanic Origin [Internet]. 2020 [cited 2022 Nov 2]. Available from: https://data.cdc.gov/NCHS/AH-Monthly-Provisional-Counts-of-Deaths-for-Select/65mz-jvh5
29. 29Leading Causes of Death [Internet]. 2022 [cited 2022 Nov 1]. Available from: https://www.cdc.gov/nchs/fastats/leading-causes-of-death.htm
30. 30National Center for Health Statistics. Excess Deaths Associated with COVID-19 [Internet]. 2022 [cited 2022 Apr 1]. Available from: https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm
31. Preston Samuel H., Heuveline Patrick, Guillot Michel. *Demography: Measuring and Modeling Population Processes* (2000.0)
32. Faust JS, Du C, Mayes KD, Li S-X, Lin Z, Barnett ML. **Mortality From Drug Overdoses, Homicides, Unintentional Injuries, Motor Vehicle Crashes, and Suicides During the Pandemic, March-August 2020**. *JAMA* (2021.0) **326** 84-6. DOI: 10.1001/jama.2021.8012
33. Anderson KE, McGinty EE, Presskreischer R, Barry CL. **Reports of Forgone Medical Care Among US Adults During the Initial Phase of the COVID-19 Pandemic**. *JAMA Network Open* (2021.0) **4** e2034882. DOI: 10.1001/jamanetworkopen.2020.34882
34. Zhang J.. **Hospital Avoidance and Unintended Deaths during the COVID-19 Pandemic**. *American Journal of Health Economics* (2021.0) **7** 405-26
35. Friedman AB, Barfield D, David G, Diller T, Gunnarson C, Liu M. **Delayed emergencies: The composition and magnitude of non-respiratory emergency department visits during the COVID-19 pandemic**. *J Am Coll Emerg Physicians Open* (2021.0) **2** e12349. DOI: 10.1002/emp2.12349
36. Vamos EP, Khunti K. **Indirect effects of the COVID-19 pandemic on people with type 2 diabetes: time to urgently move into a recovery phase**. *BMJ Qual Saf* (2021.0). DOI: 10.1136/bmjqs-2021-014079
37. Wolfson JA, Leung CW. **Food Insecurity and COVID-19: Disparities in Early Effects for US Adults**. *Nutrients* (2020.0) **12**. DOI: 10.3390/nu12061648
38. Mann FD, Krueger RF, Vohs KD. **Personal economic anxiety in response to COVID-19**. *Pers Individ Dif* (2020.0) **167** 110233. DOI: 10.1016/j.paid.2020.110233
39. Prime H, Wade M, Browne DT. **Risk and resilience in family well-being during the COVID-19 pandemic**. *Am Psychol* (2020.0) **75** 631-43. DOI: 10.1037/amp0000660
40. 40Centers for Disease Control and Prevention. National Center for Health Statistics. COVID-19 Coding and Reporting Guidance—National Vital Statistics System [Internet]. 2021 [cited 2022 May 31]. Available from: https://www.cdc.gov/nchs/covid19/coding-and-reporting.htm
41. Gundlapalli AV, Lavery AM, Boehmer TK, Beach MJ, Walke HT, Sutton PD. **Death Certificate–Based ICD-10 Diagnosis Codes for COVID-19 Mortality**. *Morbidity and Mortality Weekly Report* (2021.0)
42. Xie Y, Al-Aly Z. **Risks and burdens of incident diabetes in long COVID: a cohort study**. *Lancet Diabetes Endocrinol* (2022.0) **10** 311-21. DOI: 10.1016/S2213-8587(22)00044-4
43. Abbasi J.. **The COVID Heart-One Year After SARS-CoV-2 Infection, Patients Have an Array of Increased Cardiovascular Risks**. *JAMA* (2022.0) **327** 1113-4. DOI: 10.1001/jama.2022.2411
44. Douaud G, Lee S, Alfaro-Almagro F, Arthofer C, Wang C, McCarthy P. **SARS-CoV-2 is associated with changes in brain structure in UK Biobank**. *Nature* (2022.0). DOI: 10.1038/s41586-022-04569-5
45. Carey IM, Cook DG, Harris T, DeWilde S, Chaudhry UAR, Strachan DP. **Risk factors for excess all-cause mortality during the first wave of the COVID-19 pandemic in England: A retrospective cohort study of primary care data**. *PLoS One* (2021.0) **16** e0260381. DOI: 10.1371/journal.pone.0260381
46. Tarazi WW, Finegold K, Sheingold SH, Wong Samson L, Zuckerman R, Bosworth A. **COVID-19-Related Deaths And Excess Deaths Among Medicare Fee-For-Service Beneficiaries**. *Health Aff* (2021.0) **40** 879-85. DOI: 10.1377/hlthaff.2020.02521
47. Gu T, Chu Q, Yu Z, Fa B, Li A, Xu L. **History of coronary heart disease increased the mortality rate of patients with COVID-19: a nested case-control study**. *BMJ Open* (2020.0) **10** e038976. DOI: 10.1136/bmjopen-2020-038976
48. Sahni S, Gupta G, Sarda R, Pandey S, Pandey RM, Sinha S. **Impact of metabolic and cardiovascular disease on COVID-19 mortality: A systematic review and meta-analysis**. *Diabetes Metab Syndr* (2021.0) **15** 102308. DOI: 10.1016/j.dsx.2021.102308
49. Bhaskaran K, Bacon S, Evans SJ, Bates CJ, Rentsch CT, MacKenna B. **Factors associated with deaths due to COVID-19 versus other causes: population-based cohort analysis of UK primary care data and linked national death registrations within the OpenSAFELY platform**. *Lancet Reg Health Eur* (2021.0) **6** 100109. DOI: 10.1016/j.lanepe.2021.100109
50. Andrasfay T, Goldman N. **Reductions in 2020 US life expectancy due to COVID-19 and the disproportionate impact on the Black and Latino populations**. *Proc Natl Acad Sci U S A* (2021.0) **118**. DOI: 10.1073/pnas.2014746118
51. Xia Y, Jin R, Zhao J, Li W, Shen H. **Risk of COVID-19 for patients with cancer**. *Lancet Oncol* (2020.0) **21** e180. DOI: 10.1016/S1470-2045(20)30150-9
52. Miyashita H, Mikami T, Chopra N, Yamada T, Chernyavsky S, Rizk D. **Do patients with cancer have a poorer prognosis of COVID-19? An experience in New York City**. *Ann Oncol* (2020.0) **31** 1088-9. DOI: 10.1016/j.annonc.2020.04.006
53. Chen R, Aschmann HE, Chen Y-H, Glymour MM, Bibbins-Domingo K, Stokes AC. **Racial and Ethnic Disparities in Estimated Excess Mortality From External Causes in the US, March to December 2020**. *JAMA Intern Med* (2022.0) **182** 776-8. DOI: 10.1001/jamainternmed.2022.1461
54. Dubey MJ, Ghosh R, Chatterjee S, Biswas P, Chatterjee S, Dubey S. **COVID-19 and addiction**. *Diabetes Metab Syndr* (2020.0) **14** 817-23. DOI: 10.1016/j.dsx.2020.06.008
55. Witteveen D, Velthorst E. **Economic hardship and mental health complaints during COVID-19**. *Proc Natl Acad Sci U S A* (2020.0) **117** 27277-84. DOI: 10.1073/pnas.2009609117
56. 56USDOT. National Highway Traffic Safety Administration Early Estimate of Motor Vehicle Traffic Fatalities for the First 9 months (Jan-Sep) of 2021. Feb 2022. DOT [Internet]. 2022 Feb;HS 813(240). Available from: https://www.nhtsa.gov/sites/nhtsa.gov/files/2021-09/Early-Estimate-Motor-Vehicle-Traffic-Fatalities-Q1-2021.pdf
57. Lee W-E, Park SW, Weinberger DM, Olson D, Simonsen L, Grenfell BT. **Direct and indirect mortality impacts of the COVID-19 pandemic in the US, March 2020-April 2021**. *medRxiv* (2022.0). DOI: 10.1101/2022.02.10.22270721
58. Montez JK, Zajacova A, Hayward MD, Woolf SH, Chapman D, Beckfield J. **Educational Disparities in Adult Mortality Across U.S. States: How Do They Differ, and Have They Changed Since the Mid-1980s?**. *Demography* (2019.0) **56** 621-44. DOI: 10.1007/s13524-018-0750-z
59. Mokdad AH, Ballestros K, Echko M, Glenn S, Olsen HE. **The State of US Health, 1990–2016: Burden of Diseases, Injuries, and Risk Factors Among US States**. *JAMA* (2018.0) **319** 1444-72. DOI: 10.1001/jama.2018.0158
60. Howard G.. **Why do we have a stroke belt in the southeastern United States? A review of unlikely and uninvestigated potential causes**. *Am J Med Sci* (1999.0) **317** 160-7. DOI: 10.1097/00000441-199903000-00005
61. Lanska DJ, Kuller LH. **The geography of stroke mortality in the United States and the concept of a stroke belt**. *Stroke* (1995.0) **26** 1145-9. DOI: 10.1161/01.str.26.7.1145
62. French G, Hulse M, Nguyen D, Sobotka K, Webster K, Corman J. **Impact of Hospital Strain on Excess Deaths During the COVID-19 Pandemic—United States, July 2020-July 2021**. *MMWR Morb Mortal Wkly Rep* (2021.0) **70** 1613-6. DOI: 10.15585/mmwr.mm7046a5
63. Sehra ST, Fundin S, Lavery C, Baker JF. **Differences in race and other state-level characteristics and associations with mortality from COVID-19 infection**. *J Med Virol* (2020.0) **92** 2406-8. DOI: 10.1002/jmv.26095
64. Oronce CIA, Scannell CA, Kawachi I, Tsugawa Y. **Association Between State-Level Income Inequality and COVID-19 Cases and Mortality in the USA**. *J Gen Intern Med* (2020.0) **35** 2791-3. DOI: 10.1007/s11606-020-05971-3
65. Alagoz O, Lowry KP, Kurian AW, Mandelblatt JS, Ergun MA, Huang H. **Impact of the COVID-19 Pandemic on Breast Cancer Mortality in the US: Estimates From Collaborative Simulation Modeling**. *J Natl Cancer Inst* (2021.0) **113** 1484-94. DOI: 10.1093/jnci/djab097
66. Duffy SW, Seedat F, Kearins O, Press M, Walton J, Myles J. **The projected impact of the COVID-19 lockdown on breast cancer deaths in England due to the cessation of population screening: a national estimation**. *Br J Cancer* (2022.0) **126** 1355-61. DOI: 10.1038/s41416-022-01714-9
67. 67Chris Lisinski, State House News Service. Mass. public health department reports “significant overcount” of COVID deaths [Internet]. WBUR. 2022 [cited 2022 Apr 3]. Available from: https://www.wbur.org/news/2022/03/10/new-covid-death-definition
|
---
title: 'Prevalence and associated factors of delayed sputum smear conversion in patients
treated for smear positive pulmonary tuberculosis: A retrospective follow up study
in Sabah, Malaysia'
authors:
- Linghui Amanda Khor
- Ulfa Nur Izzati A. Wahid
- Lee Lee Ling
- Sarah Michael S. Liansim
- Jush’n Oon
- Mahendran Naidu Balakrishnan
- Wei Leik Ng
- Ai Theng Cheong
journal: PLOS ONE
year: 2023
pmcid: PMC9987811
doi: 10.1371/journal.pone.0282733
license: CC BY 4.0
---
# Prevalence and associated factors of delayed sputum smear conversion in patients treated for smear positive pulmonary tuberculosis: A retrospective follow up study in Sabah, Malaysia
## Abstract
### Introduction
Tuberculosis remains a major health problem globally and in Malaysia, particularly in the state of Sabah. Delayed sputum conversion is associated with treatment failure, drug-resistant tuberculosis and mortality. We aimed to determine the prevalence of delayed sputum conversion among smear positive pulmonary tuberculosis (PTB) patients and its associated factors in Sabah, Malaysia.
### Methods
A retrospective follow up study on all patients newly diagnosed with smear positive pulmonary tuberculosis from 2017 to 2019 was conducted at three government health clinics in Sabah, utilizing data from a national electronic tuberculosis database and medical records. Descriptive statistics and binary logistic regression were applied for data analysis. The outcome of the study was the sputum conversion status at the end of the two-month intensive treatment phase with either successful conversion to smear negative or non-conversion.
### Results
374 patients were included in the analysis. Our patients were generally younger than 60 years old with no medical illness and varying proportions of tuberculosis severity as judged by radiographic appearance and sputum bacillary load upon diagnosis. Foreigners constituted $27.8\%$ of our sample. $8.8\%$ (confidence interval: 6.2–12.2) did not convert to smear negative at the end of the intensive phase. Binary logistic regression showed that older patients ≥60 years old (adjusted odds ratio, AOR = 4.303), foreigners (AOR = 3.184) and patients with higher sputum bacillary load at diagnosis [2+ (AOR = 5.061) and 3+ (AOR = 4.992)] were more likely to have delayed sputum smear conversion.
### Conclusion
The prevalence of delayed sputum conversion in our study was considerably low at $8.8\%$ with age ≥60 years old, foreigners and higher pre-treatment sputum bacillary load associated with delayed conversion. Healthcare providers should take note of these factors and ensure the patients receive proper follow up treatment.
## Introduction
Tuberculosis (TB) remains a significant health problem globally. It is estimated there were 1.6 million deaths due to tuberculosis worldwide in 2021, following an upward trend from 1.4 million in 2019 and 1.5 million in 2020 [1]. TB is expected to rank second only to COVID-19 as the cause of death from a single infectious agent in 2020 and 2021 [1]. The Southeast Asia region bears the highest TB burden. In 2021, $45\%$ of new TB cases were reported in Southeast Asia, followed by Africa ($23\%$), Western Pacific ($18\%$) and smaller proportions in other regions [1].
The main source of transmission for TB is from smear positive pulmonary TB (PTB) patients via infective droplets from their lungs and throat [2]. Direct microscopic observation of sputum smear for acid-fast bacilli (AFB) plays an important role in treatment monitoring and this method is widely and easily available in developing countries such as Malaysia [3]. As recommended by World Health Organization (WHO), sputum conversion rate (SCR) to smear negative from a smear positive patient at the end of the intensive phase of anti-tuberculosis therapy (ATT) (at the end of the second month of treatment duration), is an operational indicator of the national TB control programs’ capacity and an essentially important clinical indicator of treatment response and disease prognosis [4]. Delayed sputum conversion, defined as non-conversion to smear negative PTB at the end of intensive phase, is associated with poorer outcomes, specifically treatment failure, and increased risk of drug resistance and higher mortality [5–7]. Delayed sputum conversion also contributes to higher treatment cost and additional burden to healthcare services, In Malaysia, treatment success rate for TB was $78\%$ in 2020, lower than the global success rate of $86\%$ in the same year [1]. Tackling delayed sputum conversion is one important strategy to improve the TB treatment success rate.
In Sabah, one of the states in Borneo Malaysia, TB notification constituted $20\%$ of all TB notification nationwide between 2012 and 2018, despite representing only about $10\%$ of the Malaysian population [8]. The TB incidence rate in Sabah during that period was reported as 128 per 100,000 population, which was higher than the national incidence rate (97 per 100,000 population) [9]. Sabah also has a unique and diverse sociodemographic composition compared to other states in Malaysia, comprising 42 ethnic groups, almost 200 sub-ethnic groups and a large proportion of immigrants, both legal and illegal, from neighbouring countries such as Philippines and Indonesia.
In this study, we aimed to determine the prevalence of delayed sputum conversion in patients with smear positive PTB and its associated factors on the west coast of Sabah. It is useful to obtain insight into the extent of delayed sputum conversion in this region with high TB burden and unique demographic profile. Understanding the factors associated with delayed sputum conversion could guide the development of public health policy to tackle this issue.
## Study design and setting
A retrospective follow up study was conducted involving all new cases of patients with smear positive PTB at three health clinics on the central west coast of Sabah, namely Tamparuli Health Clinic in Tuaran, Penampang Health Clinic in Penampang and Luyang Health Clinic in Kota Kinabalu, from 1st January 2017 to 31st December 2019.
These three clinics were purposely selected because they were the main clinic that treat tuberculosis in three different districts in Sabah, namely Tuaran, Penampang and Kota Kinabalu. These three clinics acted as the main treating centre for tuberculosis in their respective district. Each TB unit in the respective clinic consisted of medical officer, medical assistant and nurses. All patients suspected of having PTB would be required to submit at least two sputum specimens for microscopic examination with at least one early morning specimen when possible. Sputum specimens were sent for acid-fast bacilli (AFB) smear and culture. AFB smears were performed routinely in the clinics using Ziehl-Nielsen stain and were examined under direct microscopy by medical laboratory technologists (MLT). MLT in health clinics are trained in analyzing the AFB smear and routinely receive refresher courses. Patients with smear positive PTB could be started on treatment in any of these clinics. Cases like smear negative PTB, extrapulmonary TB and drug-resistant TB such as multidrug-resistant TB (MDR-TB) would be referred to the hospital for treatment initiation. Once diagnosed and started on treatment, they could continue treatment in these clinics. Medication was provided on a daily basis for intensive phase and weekly basis for maintenance phase. All relevant information from case note was entered into MyTB system, an electronic TB clinical registry system operated by the Ministry of Health Malaysia.
## Study population
We included all patients newly diagnosed with smear positive PTB who were 18 years old and above, and under regular follow up at the study sites until the end of the two-month intensive phase. We included patients with drug-resistant TB as well. Patients who defaulted treatment before the end of the two-month intensive phase were excluded from the study because sputum conversion would only be monitored at the end of intensive phase. We also excluded those with disseminated tuberculosis because they were being followed up in hospitals instead of health clinics.
## Data collection
We used universal sampling and extracted all cases that fulfilled our study criteria from the MyTB database. Information that was readily available from MyTB database was the age at diagnosis, ethnicity, nationality, gender, education level, smoking status at diagnosis, diabetes and human immunodeficiency virus (HIV) status at diagnosis, sputum AFB load at diagnosis, chest X-ray (CXR) severity at diagnosis, presence of MDR-TB and status of sputum conversion at the end of the two-month intensive phase. Some information that was not available from the MyTB database such as presence of other co-morbidities, alcohol dependence status at diagnosis, duration of symptoms before diagnosis and number of days missing directly observed therapy (DOT) were obtained from the patients’ manual medical records that were kept in the respective health clinics. All data were recorded using data collection form.
## Variables and operational definition
Smear-positive PTB was diagnosed in either one of the following ways [10]: (i) two or more positive sputum AFB smears, or (ii) one positive sputum AFB smear accompanied by abnormalities in CXR suggestive of PTB (as determined by the physician), or (iii) one positive sputum AFB smear and one positive sputum culture Mycobacterium tuberculosis.
The dependent variable was the delayed sputum conversion, defined as smear-positive PTB with a sputum sample that remains AFB smear-positive at the end of the two-month intensive phase of anti-tuberculous treatment.
The independent variables were the age at diagnosis, ethnicity, nationality, gender, education level, smoking status at diagnosis, alcohol dependence status at diagnosis, presence of diabetes mellitus at diagnosis, co-morbidities, HIV status at diagnosis, sputum AFB load at diagnosis, CXR severity at diagnosis, duration of symptoms before diagnosis, presence of MDR-TB and number of days missing DOT. Alcohol dependence was based on the fulfilment of three or more DSM-IV dependence criteria of substance use disorders within 12 months [10]. Presence of diabetes mellitus at diagnosis was defined as patients who were diagnosed with diabetes mellitus at the point of diagnosis of TB or were previously diagnosed by physician with two abnormal glucose results (if asymptomatic), one abnormal glucose result (if symptomatic), or HbA1c >$6.3\%$ based on Malaysian guideline [11]. Co-morbidities included medical conditions that were either recorded in the MyTB database or medical records in clinic, such as hypertension, dyslipidemia, ischaemic heart disease, hepatitis B, chronic kidney disease, stroke and peptic ulcer disease. HIV status at diagnosis referred either to patients with pre-existing HIV, or who were newly diagnosed with upon diagnosis of TB. Sputum AFB load was graded as scanty, 1+, 2+, and 3+ based on the number of AFB seen microscopically before initiation of treatment (negative: no bacilli per 100 fields of observation; scanty: 1–9 bacilli per 100 fields; 1+: 10 to 99 bacilli per 100 fields; 2+: 1–10 per field; and 3+: >10 per field of observation). CXR severity at diagnosis was graded by treating physicians into minimal, moderate or far advanced based on Malaysian guideline [12]. Duration of symptoms before diagnosis was defined by the duration of symptoms suggestive of PTB from the time of onset to the point of diagnosis, based on the history documented by physician in the medical records. MDR-TB was defined as strains of TB that are resistant to at least two main first-line antituberculous drugs (i.e. isoniazid and rifampicin), as demonstrated in the sputum culture and sensitivity result [10]. Number of days missing DOT was defined as the number of days patient missed the antituberculous medication during the intensive phase as documented in the medical records.
To maintain the quality of data, validation rules were implemented in the Microsoft Excel sheet for data entry. Standardized vocabularies and units were used with regular sessions among researchers to discuss any irregularities. Completed data entries were screened to detect any irregularities.
## Data analysis
Data was entered and analyzed using IBM Statistical Program for Social Sciences (SPSS) software version 26. The categorical data were presented as frequency and percentage. For the age variable, 60 years old was chosen as the cut-off point for analysis because Malaysia public healthcare system defined people aged 60 years and above as older persons or senior citizens with healthcare policies for older persons designed around this cut-off point of age. The outcome of this study (dependent variable) was sputum conversion at the end of the 2-month intensive phase, with either successful conversion to smear negative or non-conversion (delayed sputum conversion). The association between the dependent and independent variables was examined by Chi-square test. The assumptions for Chi-square test were checked; all expected frequencies were greater than 1 and at least $80\%$ are greater than 5. The level of significance was set at p-value of less than 0.05. Significant variables in Chi-square test and variables which might have clinical importance to predict delayed sputum conversion, based on literature review, researchers’ experiences and observations, were included in the bivariable logistic regression. Bivariable and multivariable binary logistic regression were performed to determine the association between the independent variables with delayed sputum conversion. Variables with p-value <0.25 from bivariable logistic regression were included in the multivariable logistic regression model. Literature supports the inclusion of p-value < 0.25 into the multivariable regression analysis [13]. Adjusted odds ratio was calculated with p-value of less than 0.05 considered statistically significant. Model fit was checked with Hosmer and Lemeshow test and classification table.
## Ethics
Ethical approval for this study was obtained from the Medical Research & Ethics Committee, Ministry of Health Malaysia (NMRR-20-1581-53331). The data extracted from the MyTB database were de-identified to protect patients’ confidentiality. Patients’ consent was not required by the ethics committee as the study only analyzed secondary data from the database and medical records.
## Results
A total of 374 new cases of smear positive pulmonary tuberculosis that fulfilled the inclusion criteria were included in this study. The sociodemographic and clinical characteristics of the cohort were shown in Table 1. The majority of the patients were male in the age group of 18 to 59 years old. About one-third of our cohort ($27.8\%$, $$n = 104$$) was foreigner.
**Table 1**
| Sociodemographic characteristics | Frequency, n (%) |
| --- | --- |
| Age • 18 to 59 years old • 60 years old and above | 319 (85.3%)55 (14.7%) |
| Gender • Male • Female | 250 (66.8%)124 (33.2%) |
| Nationality • Malaysian • Foreigner | 270 (72.2%)104 (27.8%) |
| Educational level • Primary/ No formal education • Secondary/ Tertiary • Unknown | 98 (26.2%)114 (30.5%)162 (43.3%) |
| Clinical characteristics | Frequency, n (%) |
| Comorbidities • No known medical illness • 1 comorbid • 2 comorbid and more | 267 (71.4%)70 (18.7%)37 (9.9%) |
| HIV status • Non-reactive • Reactive | 369 (98.7%)5 (1.3%) |
| Presence of diabetes mellitus • Yes • No | 43 (11.5%)331 (88.5%) |
| Smoking status • Non-smoker • Active smoker • Ex-smoker | 206 (55.1%)105 (28.1%)63 (16.8%) |
| Alcohol dependence status • No • Yes | 301 (80.5%)73 (19.5%) |
| Severity of CXR • Mild/ No lesion • Moderate • Severe | 116 (31.0%)165 (44.1%)93 (24.9%) |
| Duration of symptoms before diagnosis • 1 month or less • 2 to 6 months • 7 months and above | 206 (55.0%)130 (34.8%)38 (10.2%) |
| Sputum AFB load at diagnosis • Scanty • 1+ • 2+ • 3+ | 84 (22.5%)139 (37.2%)85 (22.7%)66 (17.6%) |
| Presence of missed DOT • No • Yes | 312 (83.4%)62 (16.6%) |
| Presence of MDR-TB • No • Yes | 370 (98.9%)4 (1.1%) |
The majority of our cohort had no known medical illness ($71.4\%$, $$n = 267$$) with only $11.8\%$ ($$n = 43$$) had diabetes mellitus and $1.3\%$ ($$n = 5$$) had HIV. Most patients did not smoke ($55.1\%$, $$n = 206$$) and consume alcohol ($80.5\%$, $$n = 301$$). More patients had moderate CXR severity ($44.1\%$, $$n = 165$$), 1+ AFB load at diagnosis ($37.2\%$, $$n = 139$$) and symptoms for a month or less at diagnosis ($55.0\%$, $$n = 206$$). Majority of the patients adhered to DOT ($83.4\%$, $$n = 312$$) and only $1.1\%$ ($$n = 4$$) were diagnosed with drug-resistant TB, which were all MDR-TB.
Out of these 374 patients with smear positive pulmonary tuberculosis, 33 ($8.8\%$, $95\%$ confidence interval, CI: 6.2–12.2) were non-converters at the end of the intensive phase which was at the second month of anti-tuberculous therapy, thus classified as having delayed sputum smear conversion.
For the bivariate analysis using Chi-square (Table 2) and simple logistic regression (Table 3), age group and sputum AFB load at diagnosis were significantly associated with delayed sputum smear conversion ($p \leq 0.05$). Bivariate analysis and simple logistic regression were not performed for the presence of HIV and MDR-TB as the frequency in those variables were too small for meaningful analysis (5 or less).
Variables with significance of $p \leq 0.250$ in univariate analysis were included in the multivariable logistic regression such as age, nationality, education level, severity of CXR, duration of symptoms and sputum AFB load. After controlling for all these factors, only age, nationality and sputum AFB load at diagnosis were found to be significantly associated with delayed sputum smear conversion. Older patients ≥60 years old (adjusted odds ratio, AOR = 4.303), foreigners (AOR = 3.184) and patients with higher sputum AFB load at diagnosis [2+ (AOR = 5.061) and 3+ (AOR = 4.992)] were more likely to have delayed sputum smear conversion. This model fit was based on the Hosmer and Lemeshow test which showed non-significant result (p-value = 0.735) and the percentage from the classification table ($91.2\%$ correctly classified).
## Discussion
Key findings from our study are: 1) the prevalence of delayed sputum conversion from 2017 to 2019 in our cohort is $8.8\%$ (CI 6.2–12.2); 2) Older patients ≥ 60 years old, foreigners and patients with higher sputum AFB load at diagnosis are associated with delayed sputum conversion.
Table 4 outlines the comparison of our prevalence data with a few recent studies in the past five years. Our finding is consistent with another Malaysian study which was also conducted in Sabah in different research sites, where they identified $7.2\%$ of patients having delayed sputum conversion from 2013 to 2018 [14]. In comparison, our prevalence of delayed sputum conversion was not too high. Ibrahim [2022] and Bhatti [2021] reported a higher prevalence of $19.1\%$ and $30.5\%$ respectively in Malaysia [15, 16]. Similar large variation of prevalence data was observed globally as observed in Table 4 ($8.3\%$ to $35\%$), and as reported in earlier studies, from $8\%$ in Cameroon up to $30\%$ in Cleveland [14–21]. This variation is observed despite the relatively standardized, effective first-line regime for antituberculous therapy worldwide. The different sociodemographic profiles may be one explanation. For example, Ibrahim [2022] focused on aboriginal groups in Peninsular Malaysia (separated from Borneo Malaysia) where the prevalence was much higher at $19.1\%$ with no overlapping confidence limit with ours and Mokti’s [2021] study [14, 16]. Poor health accessibility may be a factor here in this group, leading to poor adherence and delayed sputum conversion. Likewise, health accessibility is an issue observed globally and is an important one to tackle as tuberculosis typically affects lesser developed countries [1].
**Table 4**
| No | Author (Year) | Country | Prevalence, % (Confidence interval) | Associated factors |
| --- | --- | --- | --- | --- |
| 1 | Our study | Sabah, Malaysia | 8.8 (6.2–12.2) | Age≥60ForeignerHigher sputum bacillary load |
| 2 | Ibrahim MN (2022) | Peninsular Malaysia (Aborigine group) | 19.1 (15.7–22.9) | SmokingDiabetes mellitusHIV infection |
| 3 | Mokti K (2021) | Sabah, Malaysia | 7.2 (6.2–8.2) | Moderate to advanced CXRAge> 60SmokingNo DOTS supervisorNon-MalaysianSuburban residence |
| 4 | Gunda (2017) | Rural Tanzania | 8.3 (4.5–13.8) | MaleAge>50Sputum bacillary load of 3+ |
| 5 | Azza (2019) | Tunisia | 35 (31–40) | Diabetes mellitusSmokingHaemoptysisHigher sputum bacillary load |
| 6 | Bhatti (2021) | Penang, Malaysia (hospital patients) | 30.5 (27.8–33.3) | Age≥50Blue-collar jobsSmokingHigher sputum bacillary load Relapse and interruption in treatment |
| 7 | Asemahagn (2021) | Ethiopia | 15.0 (11.0–19.6) | Higher BMIHigher sputum bacillary load HIV infectionDiabetes mellitusSmokingSocietal stigma, delay in TB service |
We found that the following variables were significantly associated with delayed sputum conversion in our cohort: age ≥ 60 years old, foreigner status and higher sputum AFB load at diagnosis. Older age and higher sputum AFB load have consistently been shown to be associated with delayed conversion in previous studies [2, 14, 15, 17, 18]. Older age can lead to poorer immune response, causing ineffective clearance of the bacilli. Delay in timely health-seeking behaviour was also observed in older person, leading to poor progress in treatment for tuberculosis where regular, close follow up in intensive phase is important [22]. The higher sputum AFB load at diagnosis an indication of a heavier mycobacterial burden. More time may be needed to clear the heavier load, be it lived or dead bacilli [23]. Higher bacillary load was associated with poorer treatment outcome and higher mortality rate in tuberculosis [24]. Another possibility is treatment failure due to multidrug-resistant tuberculosis (MDR-TB) but we did not find many MDR-TB when we traced the sputum culture result. Higher sputum AFB load is something identifiable at diagnosis and reduction in bacillary load can be observed as early as three days with effective treatment [25]. It is perhaps worth considering an earlier and more rigorous follow up with sputum AFB smear for patients with higher bacillary load on diagnosis. The current practice in primary care clinics in *Malaysia is* to follow up after two to four weeks of treatment initiation, with sputum AFB smear being repeated then and after two months of intensive therapy. We would be able to detect poor response to treatment early and act accordingly.
In our study, we also included foreigners in our analysis. We found that foreigners in the state of Sabah were more likely to have delayed sputum conversion, similar to Mokti et al. [ 14]. The social demography in the state of *Sabah is* peculiar and different from the rest of Malaysia as they are known to have more foreigners. It has been estimated that non-citizen made up $30\%$ of population in the state of Sabah, mainly from Indonesia and Philippines [26]. Foreign nationality was associated with poorer treatment outcome for tuberculosis in another local study [27]. It is not a practice among healthcare facilities to identify and deport any illegal immigrants who came to seek treatment for infectious diseases such as tuberculosis in Malaysia although they may be encouraged to continue treatment in their home countries. It is known that health-seeking behaviour is different among foreigners compared to locals, owing to various factors such as fear of deportation, financial constraints and language barriers [28]. Foreigners may tend to present late with higher severity of disease. That in turn may contribute to poorer treatment outcomes such as delayed sputum conversion. We also take note that this problem may be the tip of the iceberg as our study only captured foreigners who presented to the clinics, not those who avoided seeking medical treatment in government healthcare facilities.
We did not find a significant association with other variables that are significant in previous studies, namely the presence of diabetes mellitus, CXR severity, education level, smoking status, alcohol status, gap in treatment and duration of symptoms [2, 15, 19]. Of these factors, diabetes mellitus and smoking were more consistently captured in other studies as shown in Table 4 [14–16, 20, 21]. For diabetes mellitus, while many studies showed that its presence was associated with delayed sputum conversion, there were also studies which presented conflicting findings where presence of diabetes mellitus was not associated with poorer treatment outcome for TB such as delayed sputum conversion [29, 30]. Mahishale V [2017] reported a more specific finding where poor glycemic control upon diagnosis of TB predicted poorer outcomes such as delayed sputum conversion [31]. Shewade [2017] showed inadequate high-quality data in their systematic review to delineate the effect of glycaemic control on TB treatment outcome [32]. We hypothesized that our cohort of diabetic patients had better glycaemic control and thus, not significantly associated with delayed sputum conversion.
Smoking had also been associated with more extensive lung disease, lung cavitation, delayed sputum conversion at two months, higher default rates, treatment failures and relapses [33]. Ex-smokers also showed poorer treatment outcomes for tuberculosis [33]. The negative effect of smoking on TB treatment outcome would increase further if coupled with alcohol drinking [34]. Conflicting findings were also seen for smoking, where some studies found that smoking was not associated with negative outcome for tuberculosis [35, 36]. Bay JG [2022] posed an interesting hypothesis where they deduced that their cohort of smoker did not have poorer outcome due to better socioeconomic standing (employed, able to buy cigarettes, have better nutrition, better educated) [37]. In our study, we did not find any association of smoking with delayed sputum conversion. It may be possible that our cohort were mainly light or early smokers with no significant lung damage yet. More data would be needed in our cohort to explore the hypothesis of better socioeconomic profile as a protective factor against delayed sputum conversion.
An interesting finding for us is that the prolonged duration of symptoms before diagnosis of PTB was not significantly associated with delayed sputum conversion in our study. Some studies demonstrated that patients with longer duration of symptom more than 2–3 months had delayed conversion [38, 39]. This counter the argument that prolonged duration of symptoms may reflect higher mycobacterial burden due to progress of disease or prolonged exposure to source [2]. Objective measurement of disease burdens such as initial sputum AFB load and severity of CXR would be more useful in the initial assessment of patients with tuberculosis.
## Strength and limitation
Our study was a multicentre study where we included three main health clinics which were the treatment centre for tuberculosis, spanned across three different districts in Sabah. Another strength of our study was we extracted and verified data from patients’ medical records as well instead of just relying on the MyTB database. Certain variables were either not available or incomplete in MyTB database such as presence of other co-morbidities, alcohol status at diagnosis, duration of symptoms before diagnosis, sputum AFB load and number of days missing DOTS.
There were several limitations of our study. One, the sample size in our study may not be large enough to capture any significant difference. This was more evident for variables such as HIV status and presence of MDR-TB as their numbers were too small in our study for meaningful analysis. Two, our data on certain variables may not be that robust as well. While we did investigate the co-morbidities of our cohort, the co-morbidities varied a lot in the types of disease, rendering the numbers inadequate for analysis. A more detailed look into the characteristics of our patients with diabetes such as duration of disease, metabolic control including glycaemic, lipid and obesity profile, and smoking habits might be helpful to explain the insignificant results [40, 41]. Three, our outcome of sputum conversion was based on the result of sputum AFB smear. It was recognized that standard AFB smears could not differentiate dead bacilli from viable ones but performing routine sputum culture to demonstrate successful conversion was not feasible in resource-limited settings such as ours.
## Implications to practice and recommendations
Patients with risk factors for delayed sputum smear conversion at the completion of intensive phase of anti-tuberculous therapy (ATT) after two months should be identified earlier to avoid poorer disease outcomes, disease complications and possible progression to MDRTB. In this study, older age and high initial bacillary load at diagnosis were shown to be significantly associated with delayed sputum conversion. Patients with these identified risk factors should be properly and closely monitored and treated, with consideration of a more frequent follow up regime as opposed to current local practice, to ensure sputum conversion by the end of the intensive treatment phase.
It is also pertinent to recognize that most foreigners with symptoms are likely to present initially to a private primary healthcare provider. Malaysia health care system consists of a dual system in which the primary health care services are delivered by government health clinics and private general practices (GPs). In and around the capital city of the state of Sabah, Kota Kinabalu, there are approximately 300 GPs which act as a conduit for the initial recognition of and diagnosis of PTB in the population. The early identification of patients with risk factors for delayed sputum conversion, will help GPs to expedite the identification and referral of patients to an appropriate tertiary care centre for treatment to reduce the risk of spread of TB in the community.
Older person and foreigners are considered the vulnerable groups in this context. Future health policy should be designed to support these populations in terms of their follow up and treatment for tuberculosis to improve the rate of sputum conversion and ultimately treatment outcome for tuberculosis.
We also recommend that the MyTB database includes more robust data on clinical profile of patients such as the glycemic control upon diagnosis and the concurrent co-morbidities besides diabetes mellitus. The data would be useful for further research to elucidate the risk factors for poorer TB outcomes. Further research on the more effective follow up and treatment regime for patients at higher risk of delayed sputum conversion is recommended.
## Conclusion
The prevalence of delayed sputum conversion found in this study was considerably low at $8.8\%$ (CI: 6.2–12.2). Foreigners, older person ≥ 60 years old and patients with high sputum AFB load (AFB 2+ and more) at diagnosis were at higher risk for delayed sputum conversion. Healthcare providers should take note of these factors and ensure the patients receive proper follow up treatment.
## References
1. 1Global tuberculosis report
2022 [Internet]. 2022 [cited 15 Dec 2022]. Available from: https://www.who.int/teams/global-tuberculosis-programme/tb-reports/global-tuberculosis-report-2022.. *Global tuberculosis report* (2022.0)
2. Mohd Anwar S A, SMS, S M, Saliluddin Y. **Factors delaying sputum conversion in smear positive pulmonary tuberculosis: a systematic review**. *International Journal of Public Health and Clinical Sciences* (2018.0) **5** 56-61
3. Abd Rahman N, Mokhtar K. **Challenges of National TB Control Program Implementation: The Malaysian Experience**. *Procedia—Social and Behavioral Sciences.* (2015.0) **172** 578-84
4. **WHO guidelines on tuberculosis infection prevention and control: 2019 update. WHO Guidelines Approved by the Guidelines Review Committee**. *Geneva2019*
5. Djouma FN, Noubom M, Ateudjieu J, Donfack H. **Delay in sputum smear conversion and outcomes of smear-positive tuberculosis patients: a retrospective cohort study in Bafoussam, Cameroon**. *BMC Infect Dis* (2015.0) **15** 139. DOI: 10.1186/s12879-015-0876-1
6. Pefura-Yone EW, Kengne AP, Kuaban C. **Non-conversion of sputum culture among patients with smear positive pulmonary tuberculosis in Cameroon: a prospective cohort study**. *BMC Infectious Diseases* (2014.0) **14** 138. DOI: 10.1186/1471-2334-14-138
7. Ukwaja KN, Oshi DC, Oshi SN, Alobu I. **Profile and treatment outcome of smear-positive TB patients who failed to smear convert after 2 months of treatment in Nigeria**. *Trans R Soc Trop Med Hyg* (2014.0) **108** 431-8. DOI: 10.1093/trstmh/tru070
8. Goroh MMD, Rajahram GS, Avoi R, Van Den Boogaard CHA, William T, Ralph AP. **Epidemiology of tuberculosis in Sabah, Malaysia, 2012–2018**. *Infect Dis Poverty* (2020.0) **9** 119. DOI: 10.1186/s40249-020-00739-7
9. Avoi R, Liaw YC. **Tuberculosis Death Epidemiology and Its Associated Risk Factors in Sabah**. *Malaysia. Int J Environ Res Public Health* (2021.0) **18**
10. 10American Psychiatric Association (APA). Diagnostic and statistical manual of mental disorders. 4th ed. Washington, DC: APA; 2000.. *Diagnostic and statistical manual of mental disorders* (2000.0)
11. 11Ministry of Health Malaysia. Clinical practice guidelines on management of type 2 diabetes mellitus. 6th ed. Putrajaya, Malaysia: Malaysian Health Technology Assessment Section; 2020.. *Clinical practice guidelines on management of type 2 diabetes mellitus* (2020.0)
12. 12Ministry of Health Malaysia. Clinical practice guidelines on management of tuberculosis. 4th ed. Putrajaya, Malaysia: Malaysian Health Technology Assessment Section; 2021.. *Clinical practice guidelines on management of tuberculosis* (2021.0)
13. Zhang Z.. **Model building strategy for logistic regression: purposeful selection**. *Annals of Translational Medicine* (2016.0) **4** 111. DOI: 10.21037/atm.2016.02.15
14. Mokti K, Md Isa Z, Sharip J, Abu Bakar SN, Atil A, Hayati F. **Predictors of delayed sputum smear conversion among pulmonary tuberculosis patients in Kota Kinabalu, Malaysia: A retrospective cohort study**. *Medicine (Baltimore)* (2021.0) **100** e26841. DOI: 10.1097/MD.0000000000026841
15. Bhatti Z, Khan AH, Sulaiman SAS, Laghari M, Ali I. **Determining the risk factors associated with delayed sputum conversion at the end of the intensive phase among tuberculosis patients**. *East Mediterr Health J* (2021.0) **27** 755-63. DOI: 10.26719/2021.27.8.755
16. Ibrahim MN, Nik Husain NR, Daud A, Chinnayah T. **Epidemiology and Risk Factors of Delayed Sputum Smear Conversion in Malaysian Aborigines with Smear-Positive Pulmonary Tuberculosis**. *Int J Environ Res Public Health* (2022.0) **19**. DOI: 10.3390/ijerph19042365
17. Caetano Mota P, Carvalho A, Valente I, Braga R, Duarte R. **Predictors of delayed sputum smear and culture conversion among a Portuguese population with pulmonary tuberculosis**. *Rev Port Pneumol* (2012.0) **18** 72-9. DOI: 10.1016/j.rppneu.2011.12.005
18. Gunda DW, Nkandala I, Kavishe GA, Kilonzo SB, Kabangila R, Mpondo BC. **Prevalence and Risk Factors of Delayed Sputum Conversion among Patients Treated for Smear Positive PTB in Northwestern Rural Tanzania: A Retrospective Cohort Study**. *J Trop Med* (2017.0) **2017** 5352906. DOI: 10.1155/2017/5352906
19. Parikh R, Nataraj G, Kanade S, Khatri V, Mehta P. **Time to sputum conversion in smear positive pulmonary TB patients on category I DOTS and factors delaying it**. *J Assoc Physicians India* (2012.0) **60** 22-6
20. Asemahagn MA. **Sputum smear conversion and associated factors among smear-positive pulmonary tuberculosis patients in East Gojjam Zone, Northwest Ethiopia: a longitudinal study**. *BMC Pulmonary Medicine* (2021.0) **21** 118. DOI: 10.1186/s12890-021-01483-w
21. Slim A, Daghfous H, Ben Mansour A, Kchouk H, Ezzaouia A, Ben Saad S. **Predictive factors of delayed sputum conversion in pulmonary tuberculosis**. *European Respiratory Journal* (2019.0) **54** 4616
22. Arora VK, Singla N, Sarin R. **Profile of geriatric patients under DOTS in Revised National Tuberculosis Control Programme**. *Indian J Chest Dis Allied Sci* (2003.0) **45** 231-5. PMID: 12962456
23. Kang HK, Jeong BH, Lee H, Park HY, Jeon K, Huh HJ. **Clinical significance of smear positivity for acid-fast bacilli after >/ = 5 months of treatment in patients with drug-susceptible pulmonary tuberculosis**. *Medicine (Baltimore)* (2016.0) **95** e4540. PMID: 27495111
24. Brahmapurkar KP, Brahmapurkar VK, Zodpey SP. **Sputum smear grading and treatment outcome among directly observed treatment-short course patients of tuberculosis unit, Jagdalpur, Bastar**. *J Family Med Prim Care* (2017.0) **6** 293-6. DOI: 10.4103/jfmpc.jfmpc_24_16
25. Honeyborne I, McHugh TD, Phillips PP, Bannoo S, Bateson A, Carroll N. **Molecular bacterial load assay, a culture-free biomarker for rapid and accurate quantification of sputum Mycobacterium tuberculosis bacillary load during treatment**. *J Clin Microbiol* (2011.0) **49** 3905-11. DOI: 10.1128/JCM.00547-11
26. Hassan K.. *The dependency of economy of Sabah on foreign workers: a study using labour force survery* (2017.0)
27. Atif M, Sulaiman SA, Shafie AA, Ali I, Asif M, Babar ZU. **Treatment outcome of new smear positive pulmonary tuberculosis patients in Penang, Malaysia**. *BMC Infect Dis* (2014.0) **14** 399. DOI: 10.1186/1471-2334-14-399
28. Osman AF, Abdul Mutalib M, Tafran K, Tumin M, Chong CS. **Demographic and Socioeconomic Variables Associated With Health Care-Seeking Behavior Among Foreign Workers in Malaysia**. *Asia Pac J Public Health* (2020.0) **32** 42-8. DOI: 10.1177/1010539519893801
29. Singla R, Khan N, Al-Sharif N, Ai-Sayegh MO, Shaikh MA, Osman MM. **Influence of diabetes on manifestations and treatment outcome of pulmonary TB patients**. *Int J Tuberc Lung Dis* (2006.0) **10** 74-9. PMID: 16466041
30. Singla R, Osman MM, Khan N, Al-Sharif N, Al-Sayegh MO, Shaikh MA. **Factors predicting persistent sputum smear positivity among pulmonary tuberculosis patients 2 months after treatment**. *Int J Tuberc Lung Dis* (2003.0) **7** 58-64. PMID: 12701836
31. Mahishale V, Avuthu S, Patil B, Lolly M, Eti A, Khan S. **Effect of Poor Glycemic Control in Newly Diagnosed Patients with Smear-Positive Pulmonary Tuberculosis and Type-2 Diabetes Mellitus**. *Iran J Med Sci* (2017.0) **42** 144-51. DOI: 10.4103/0974-5009.115470
32. Shewade HD, Jeyashree K, Mahajan P, Shah AN, Kirubakaran R, Rao R. **Effect of glycemic control and type of diabetes treatment on unsuccessful TB treatment outcomes among people with TB-Diabetes: A systematic review**. *PLoS One* (2017.0) **12** e0186697. DOI: 10.1371/journal.pone.0186697
33. Mahishale V, Patil B, Lolly M, Eti A, Khan S. **Prevalence of Smoking and Its Impact on Treatment Outcomes in Newly Diagnosed Pulmonary Tuberculosis Patients: A Hospital-Based Prospective Study**. *Chonnam Med J* (2015.0) **51** 86-90. DOI: 10.4068/cmj.2015.51.2.86
34. Ma Y, Che NY, Liu YH, Shu W, Du J, Xie SH. **The joint impact of smoking plus alcohol drinking on treatment of pulmonary tuberculosis**. *Eur J Clin Microbiol Infect Dis* (2019.0) **38** 651-7. DOI: 10.1007/s10096-019-03489-z
35. Pai M, Mohan A, Dheda K, Leung CC, Yew WW, Christopher DJ. **Lethal interaction: the colliding epidemics of tobacco and tuberculosis**. *Expert Rev Anti Infect Ther* (2007.0) **5** 385-91. DOI: 10.1586/14787210.5.3.385
36. Wang JY, Hsueh PR, Jan IS, Lee LN, Liaw YS, Yang PC. **The effect of smoking on tuberculosis: different patterns and poorer outcomes**. *Int J Tuberc Lung Dis* (2007.0) **11** 143-9. PMID: 17263283
37. Bay JG, Patsche CB, Svendsen NM, Gomes VF, Rudolf F, Wejse C. **Tobacco smoking impact on tuberculosis treatment outcome: an observational study from West Africa**. *Int J Infect Dis* (2022.0) **124** S50-S5. DOI: 10.1016/j.ijid.2022.07.067
38. Fortun J, Martin-Davila P, Molina A, Navas E, Hermida JM, Cobo J. **Sputum conversion among patients with pulmonary tuberculosis: are there implications for removal of respiratory isolation**. *J Antimicrob Chemother* (2007.0) **59** 794-8. DOI: 10.1093/jac/dkm025
39. Singla R, Bharty SK, Gupta UA, Khayyam KU, Vohra V, Singla N. **Sputum smear positivity at two months in previously untreated pulmonary tuberculosis patients**. *Int J Mycobacteriol* (2013.0) **2** 199-205. DOI: 10.1016/j.ijmyco.2013.08.002
40. Magee MJ, Salindri AD, Kyaw NTT, Auld SC, Haw JS, Umpierrez GE. **Stress Hyperglycemia in Patients with Tuberculosis Disease: Epidemiology and Clinical Implications**. *Curr Diab Rep* (2018.0) **18** 71. DOI: 10.1007/s11892-018-1036-y
41. Ngo MD, Bartlett S, Ronacher K. **Diabetes-Associated Susceptibility to Tuberculosis: Contribution of Hyperglycemia vs**. *Dyslipidemia. Microorganisms* (2021.0) **9**
|
---
title: Human ACE2 expression, a major tropism determinant for SARS-CoV-2, is regulated
by upstream and intragenic elements
authors:
- John N. Snouwaert
- Leigh A. Jania
- Trang Nguyen
- David R. Martinez
- Alexandra Schäfer
- Nicholas J. Catanzaro
- Kendra L. Gully
- Ralph S. Baric
- Mark Heise
- Martin T. Ferris
- Elizabeth Anderson
- Katia Pressey
- Jacob A. Dillard
- Sharon Taft-Benz
- Victoria K. Baxter
- Jenny P-Y Ting
- Beverly H. Koller
journal: PLOS Pathogens
year: 2023
pmcid: PMC9987828
doi: 10.1371/journal.ppat.1011168
license: CC BY 4.0
---
# Human ACE2 expression, a major tropism determinant for SARS-CoV-2, is regulated by upstream and intragenic elements
## Abstract
Angiotensin-converting enzyme 2 (ACE2), part of the renin-angiotensin system (RAS), serves as an entry point for SARS-CoV-2, leading to viral proliferation in permissive cell types. Using mouse lines in which the Ace2 locus has been humanized by syntenic replacement, we show that regulation of basal and interferon induced ACE2 expression, relative expression levels of different ACE2 transcripts, and sexual dimorphism in ACE2 expression are unique to each species, differ between tissues, and are determined by both intragenic and upstream promoter elements. Our results indicate that the higher levels of expression of ACE2 observed in the lungs of mice relative to humans may reflect the fact that the mouse promoter drives expression of ACE2 in populous airway club cells while the human promoter drives expression in alveolar type 2 (AT2) cells. In contrast to transgenic mice in which human ACE2 is expressed in ciliated cells under the control of the human FOXJ1 promoter, mice expressing ACE2 in club cells under the control of the endogenous Ace2 promoter show a robust immune response after infection with SARS-CoV-2, leading to rapid clearance of the virus. This supports a model in which differential expression of ACE2 determines which cell types in the lung are infected, and this in turn modulates the host response and outcome of COVID-19.
## Author summary
SARS-CoV-2, the virus responsible for COVID-19, infects the human respiratory tract by binding to angiotensin-converting enzyme 2 (ACE2), a protein on the outer surface of cells that is exposed to the air we inhale. Differences in the structure of ACE2 between mouse and man prevent the virus from infecting cells lining the airways of the mouse, limiting the usefulness of wildtype mice as a model system for studying COVID-19. To circumvent this problem, we have created two mouse lines in which the chromosomal segment that encodes the mouse ACE2 protein has been replaced with the equivalent segment of human DNA. In the first of these lines, expression of human ACE2 is regulated by the mouse upstream promoter, while in the second line it is regulated by the human upstream promoter. Using these mice, we show both qualitative and quantitative differences between ACE2 expression driven by the mouse and human promoters that are likely to impact disease progression in the two species. Mice expressing the full-length human ACE2 gene under the control of its own promoter should provide a useful model system for understanding the complex pathological processes and immune responses associated with COVID-19 in humans.
## Introduction
SARS-CoV-2 is capable of entering cells by two distinct mechanisms: through endocytosis into endosomes and through the direct fusion of the viral particle with the cell membrane [1,2]. Non-specific endocytosis followed by endosome lysosome maturation and activation of the spike protein by cathepsin L can results in viral production by highly endocytic cells [3]. On the other hand, direct infection of cells with coronaviruses by delivery of the SARS-CoV-2 RNA across the plasma membrane requires binding of the viral spike protein to a cell surface receptor. To date, ACE2 has been shown to serve as this receptor for three coronaviruses: NL63, SARS-CoV, and SARS-CoV-2 [4–6]. Fusion of the viral and host membranes is dependent both on activation of the spike protein by host proteases such as the type II transmembrane serine protease, TMPRSS2, and on juxtamembranous cleavage of ACE2 by TMPRSS2 [7,8]. While the relative role of these two mechanisms of viral entry into host cells at various stages of COVID pathogenesis has not been defined in vivo, it is generally agreed that initial infection is dependent on the expression of ACE2 by airway epithelial cells.
ACE2 is a type-1 transmembrane protein that includes extracellular monocarboxypeptidase and collectrin domains as a result of the fusion of an angiotensin-converting enzyme-like gene with a collectrin-like gene early in evolution [9,10]. Consistent with the presence of these very different extracellular domains, ACE2 has been assigned two different functions. In both humans and mice, ACE2 is most highly expressed in the gastrointestinal tract, where ACE2 functions as an accessory protein. In addition to a proposed role for ACE2 in digestion of peptides present in chyme in preparation for their transport, its collectrin domain ensures proper trafficking of the amino acid transporter, B0AT1 (SLC6A19), to the plasma membrane of enterocytes [11]. Deficiency in B0AT1 transport leads to reductions in tryptophan and glycine in the blood and to an inflammatory bowel disease [12–14]. However, it is the mono-carboxypeptidase activity of ACE2 that has been the focus of the vast majority of studies of ACE2 function.
The carboxypeptidase activity of ACE2 degrades the ACE-derived vasoconstriction peptide, angiotensin II (Ang II), in an organ specific manner, yielding the vasodilator peptide, Ang1-7 [15,16]. ACE2 is thus considered a key regulator of the renin-angiotensin system (RAS) [17]. A body of evidence suggests that this ACE2 enzymatic activity serves to counteract effects mediated by angiotensin converting enzyme (ACE), and it has been proposed that imbalances in ACE/ACE2 activity contribute to diseases such as hypertension, progressive renal disease, and diabetes, all of which have been associated with risk for severe COVID-19 [18]. In addition, a number of studies using Ace2-/- mouse lines support a role for ACE2 in protection from lung injury, perhaps independent of its contribution to Ang II homeostasis [19,20]. ACE2 displays activity towards a number of peptides in addition to Ang II that could contribute to a potential protective function in the lung and other organ systems. For instance, ACE2, along with ACE, has been assigned a regulatory role in the kallikrien-kinin pathway [21,22]. The prominent role of ACE2 in many of the physiological systems altered in COVID-19, including the multifocal tissue damage in the microcirculatory environment of many organs, supports the argument that SARS-CoV-2 activity provokes a transient molecular disease resulting in part from reduced ACE2 activity [23,24]. Changes in ACE2 activity after SARS-CoV-2 infection may be particularly problematic in individuals in whom ACE2 expression is already altered secondary to obesity, diabetes and hypertension [25,26].
A number of model systems have been developed for identifying the mechanisms underlying the regulation of ACE2 expression as well as for elucidating the role that ACE2 plays in various physiological processes, including SARS-CoV-2 infection and COVID-19-associated pathology [27–31]. However, because these systems have been developed with the primary goal of obtaining sufficiently high ACE2 expression in the airways to allow SARS-CoV-2 infection, they may display developmental and tissue specific patterns of ACE2 expression that differ in important ways from those observed in humans. These differences in expression necessarily limit the usefulness of these systems in identifying and understanding the more subtle roles played by ACE2 in the development and resolution of COVID-19. To overcome these limitations, we have developed two mouse lines in which the ACE2 locus has been humanized by syntenic replacement. Characterization of ACE2 expression in these lines reveals that the mechanisms underlying the observed patterns of expression are unique to each species, differ between tissues, and are determined by both intragenic and upstream promoter elements. Furthermore, we show that differences in expression impact susceptibility to SARS-CoV-2 infection and immune responses.
## Generation of the humanized ACE2 lines
The mouse and human Ace2/ACE2 genes are located on the X chromosome in the region neighboring the pseudoautosomal region [32]. The mouse and human genes are remarkably similar, both in their structure and in the protein domains they encode (Fig 1A and 1B). The ectodomain of both proteins can be cleaved by ADAM17 to generate soluble ACE2 (sACE2) [8]. Differences have been noted in the regulatory elements that control expression of the mouse and human genes, with the most notable being the presence of an intragenic promoter upstream of a primate-specific exon within intron 8 of the human gene. Initiation of transcription from this internal promoter gives rise to a transcript encoding a truncated isoform of ACE2, designated dACE2 or short ACE2 [33]. The encoded protein lacks both the enzymatic activity of the full-length protein and its ability to bind SARS-Cov-2. Both human and mouse ACE2 can catalyze the formation of Ang1-7 from Ang II [34], although some differences have been noted in their enzymatic activity.
**Fig 1:** *Schematic showing the protein domains and exon intron organization of the mouse and human ACE2 locus and the structure of the loci in the two humanized mouse lines.A. Protein structural and functional domains of the ACE2 protein. The ACE2 protein consists of a signal peptide (SP), an extracellular domain, a transmembrane domain (TM) and a cytoplasmic domain (CP). B. Exonic structure of human ACE2 gene (red) and flanking genes (pink) and mouse Ace2 gene (blue) and flanking genes (light blue). Regions of the human protein identified as interacting with the spike proteins of SARS-Co-V and SARS-CoV-2 are indicated by pink boxes above the human gene. Mouse ACE2 does not bind the spike protein in these two coronaviruses. The cleavage site for ADAM17 and TMPRSS11D/TMPRSS2 are shown. Tall boxes represent coding exons while shorter boxes represent 5’ and 3’ UTRs. In both species the start codon is located in the second exon, labeled here as 1b. Both mouse and human also express a transcript (not shown) which, although it lacks exon 1a, still gives rise to an ACE2 protein with the identical amino acid sequence. The distal promoter region (DP) and proximal promoter region (PP) are present in both mouse and human. The human gene contains a third transcriptional start site between exons 8 and 9, which gives rise to a shorter transcript encoding a protein expected to lack enzymatic function and the ability to bind SARS-CoV-2. Four interferon response elements (ISREs) are located in the promoter region upstream of the sequence that encodes the first exon of the shorter transcript (exon 1c). Exon 1c is only present in primates. C. Structure of humanized Ace2 locus. The locus in which only the coding sequence is humanized, MP-ACE2, is shown above. In this locus the human sequence extends from the start codon to approximately 4 kb downstream of the 3’ end of the final exon. In the locus in which humanization includes the promoter, hACE2, the human sequence extends from approximately 3.5 kb upstream of the start codon to approximately 4 kb downstream of the 3’ end of the final exon. The orange arrow in each humanized locus represents the floxed neomycin resistance marker, which is removed by transient expression of Cre.*
We used syntenic replacement of the endogenous mouse *Ace2* gene to generate two mouse lines that are expected to express human ACE2 (Fig 1C). The two lines differ only in the location of the 5’ crossover event, with MP-ACE2 leaving the mouse promoter and 5’ UTR in place, while hACE2 replaces the entire *Ace2* gene, including 5’ regulatory regions, with a 51 kb syntenic segment of the human locus. This includes 3.5 kb of sequence upstream of the start codon. An ideogram of the structure of the ACE2 locus present in the mouse lines included in this study along with nomenclature used in referring to them is shown in Fig 2A.
**Fig 2:** *Comparison of ACE2 expression in intestinal tissue of the humanized mouse lines.A. Ideogram of the structure promoter/exon intron origin, and nomenclature/color scheme used in referring to the three mouse lines compared in these studies. mAce2 has the mouse promoter and coding sequence, MP-ACE2 has the mouse promoter but human coding sequence, and hACE2 has the human promoter and coding sequence. B. Droplet digital PCR (ddPCR) evaluation of ACE2 expression in cDNA prepared from male (♂) and female (♀) mice of the indicated genotype. Mean value for each is shown below. C. Western analysis of lysates prepared from mice of the indicted genotype/sex with the human-specific MAB933 antibody and ab15348, which recognizes both human and mouse ACE2. D. Comparison of the proximal to distal expression in the intestinal tract of ACE2/Ace2 in female mice heterozygous for the mouse locus (mACE2) and humanized locus (hACE2). Expression observed in the jejunum for the human and mouse gene was assigned a value of 1. ** p<0.01. For statistical comparison between all groups in B, see S1 Table.*
In defining the transition sites from mouse to human DNA in the two lines, we considered information available concerning the promoter of the mouse and human genes and presence of transcription factor binding motifs [35]. For example, two evolutionarily conserved regions, referred to as the distal and proximal promoter regions, have been identified just upstream of the ACE2 translational start codon in both the mouse and human genes [36]. Mice and humans both express ACE2 transcripts that include an untranslated first exon (exon 1a) [37,38]. In the MP-ACE2 mouse line, the mouse/human transition occurs at the translational start site in exon1b, thus maximizing the length of the upstream mouse regulatory region and ensuring that splicing events involved in generation of the 5’UTR occur between mouse-derived sequences.
## Expression of ACE2 in intestinal tract of humanized mice
ACE2/Ace2 is highly expressed in both the human and mouse intestinal tract, where it is located at the luminal membrane of small intestine enterocytes. Expression is also observed, albeit at lower levels, in intestinal crypt cells and in the colon [12]. We initiated characterization of the humanized mouse lines by examination of transcript levels in the jejunum (Fig 2 and S1 Fig). ACE2 is located on the X chromosome in both humans and mice. If X inactivation of the locus is random, approximately $50\%$ of the cells in tissues from female offspring will express the human gene and approximately $50\%$ will express the mouse gene. To simplify comparison of ACE2 and Ace2 expression between female and male mice throughout this report, the number of copies of Ace2 and ACE2 transcripts present in cDNA prepared from females heterozygous for the humanized locus (mAce2 x MP-ACE2 or mAce2 x hACE2) were multiplied by two.
Analysis of mRNA from the jejunum showed expression of Ace2 and ACE2 in the mAce2 and MP-ACE female mice, with mean values of 4415 and 3319 copies per ng of cDNA respectively (Fig 2B). Expression of ACE2 in the two MP-ACE2 males was slightly lower than Ace2 expression in mAce2 males (Fig 2B). This was surprising as, while sexual dimorphism for ACE2 has been reported, for the majority of tissues examined ACE2 is reported to be expressed at higher levels in males [39,40]. The expression of ACE2 in the mouse line carrying the fully humanized locus (hACE2) was also reduced by about $50\%$ compared to mACE2 (Fig 2B). However, in this case no sexual dimorphism was observed. The human promoter drove reduced gene expression of ACE2 in both hACE2 males and females compared to their mAce2 count parts.
We determined if the difference we noted in ACE2 mRNA levels between lines correlated with differences in protein levels detectable by western analysis (Fig 2C). ACE2 levels were measured in lysates prepared from the jejunum of mAce2, MP-ACE2 and hACE2 mice using two different anti-ACE2 antibodies, MAB933 and ab15348. MAB933 is a species specific monoclonal antibody that recognizes the extracellular region of human ACE2, and ab15348 is a rabbit serum raised against epitopes present in the intracellular/carboxyl terminal domain of mouse and human ACE2. The anti-ACE2 monoclonal, MAB933, detected a 120 kDa protein in lysates prepared from male and female MP-ACE2 and hACE2 animals. As expected, the MAB933 did not identify any band corresponding to ACE2 in lysates prepared from the mAce2 mice. Consistent with the lower mRNA levels of ACE2 in female hACE2 mice compared to MP-ACE2 females, a noticeably less intense 120 kDa band was observed in the females carrying the fully humanized line.
Expression of ACE2 was then compared in lysates from the jejunum of mAce2, MP-ACE2 and hACE2 male mice. Consistent with the similar mRNA levels, the human specific antibody, MAB933, detected a 120 kDa band of similar intensity in the lysates from the MP-ACE2 and hACE2 mice, while no signal was observed in the lane corresponding to the sample from mAce2 animals (Fig 2C).
Lysates were also analyzed with anti-ACE2 serum, ab15348, which recognizes both human and mouse ACE2 proteins (Fig 2C). Similar to previous reports, we noted the differing mobility of mouse ACE2 and human ACE2 on western analysis [41,42]. Although human and mouse ACE2 are 805 a.a. in length (92,463 Da) and the predicted molecular weight of ACE2 from the two species is virtually identical, mouse ACE2 consistently displayed a lower molecular weight on western analysis than the human protein, perhaps reflecting the absence of two of the seven N-glycosylation sites present in human ACE2 [43]. Also apparent on analysis of the lysates with MAB933 and ab15348 were differences between these antibodies in there cross reactivity with other proteins present in the lysates. Despite this limitation, the ab15348 anti-ACE2 antibody identified ACE2 protein in lysates from both species, with the highest intensity corresponding to lysates prepared from the jejunum of the mAce2 mice. This is consistent with the 2 fold higher Ace2 cDNA copies in male mAce2 mice when compared to the two humanized lines (Fig 2B). Despite differences in the level of expression, human and mouse ACE2/Ace2 showed a remarkably similar pattern of expression, with the lowest levels observed in the stomach and colon (Fig 2D).
## ACE2 expression in kidney and heart
Both human and mouse express high levels of ACE2/Ace2 in the kidney, and approximately $50\%$ of ACE2 expression is associated with proximal tubule cells, where it is localized to the apical (luminal) brush border of the epithelial cells [44,45]. We examined ACE2/Ace2 mRNA levels in mAce2, MP-ACE2, and hACE2 mice by ddPCR. As predicted by previous studies, expression of Ace2 in the kidney of the mAce2 mice was lower than that observed in the intestinal tract, with approximately 500 copies/ng of RNA compared to 4,000 copies/ng in the intestinal tract (Compare Figs 2B and 3A). Consistent with previous reports, expression of Ace2 was slightly higher in males [46], but in our evaluation by ddPCR, this difference failed to reach significance. Robust expression of ACE2 was observed in the MP-ACE2 females; however, levels were significantly lower than those measured in mAce2 animals. This decrease in ACE2 expression was not observed in males. In contrast, although only two MP-ACE2 males were included in the study, expression of ACE2 in these animals was similar to that observed in mAce2 males, suggesting that regulatory elements contributing to sexual dimorphism in ACE2 expression may reside within the human coding segment of the ACE2 locus. However, a prominent role for the promoter in driving sexual dimorphism was supported by our examination of expression of ACE2 in mice with the fully humanized locus, as a four-fold increase in ACE2 transcripts was observed in male hACE2 kidneys compared to those from females (Fig 3A).
**Fig 3:** *Comparison of expression of ACE2 in kidney of the humanized mouse lines.A. Droplet digital PCR (ddPCR) evaluation of cDNA prepared from male (♂) and female (♀) mice of the indicated genotype. Mean value for each is shown below the bar graph. B&C. Western analysis of lysates prepared from mice of the indicted genotype and sex with the human specific MAB933 antibody (B) and ab15348 (C), which recognizes both human and mouse ACE2. D. kidney ACE2 carboxypeptidase activity measured at one hour after initiation of the reaction. E. ACE2 carboxypeptidase activity in urine collected from mice of indicated sex and genotype determined one hour after initiation of the reaction. ** p<0.01. For statistical comparison between all groups in A see S1 Table. For statistical comparison between all groups in D and E see S1 Table.*
We determined whether these differences in mRNA levels translated into differences in ACE2 protein levels in the kidney. Western blot analysis using the human specific anti-ACE2 antibody MAB933 of lysates prepared from MP-ACE2 and hACE2 mice showed an increase in protein levels in the kidney of the mice expressing the fully humanized locus. Consistent with the small increase in mRNA levels in the female hACE2 mice compared to their MP-ACE2 counterparts, an increase in ACE2 was observed in lysates from the females carrying the fully humanized locus (Fig 3B).
We next compared ACE2 protein levels in kidney lysates prepared from mAce2, MP-ACE2 and hACE2 males using the rabbit polyclonal antibody ab15348, which recognizes both human and mouse ACE2 (Fig 3C). The hierarchy of expression of ACE2/Ace2 mRNA in the three mouse lines was recapitulated on western blot analysis, with the lowest levels of expression observed in mAce2 mice, a slight increase observed in MP-ACE2 line, and a dramatic increase in the hACE2 kidney.
Together, our results suggest that elements within the human promoter amplify sexual dimorphism in ACE2 expression in the kidney. However, minor contribution by elements within the coding regions of the gene is also supported by the less dramatic difference we observed between the sexes in ACE2 expression in the MP-ACE2 mice compared to hACE2 mice.
Similar studies of expression of ACE2 in the mouse heart were carried out. Again, a very complex pattern of expression was observed, with the mouse promoter failing to drive levels of ACE2 expression observed in wild type mice. The lower levels of expression in the MP-ACE2 mice compared to the hACE2 mice suggested interaction between intergenic and upstream regulatory regions (S2 Fig).
## Evaluation of ACE2 carboxypeptidase activity in humanized mouse lines
ACE2 encodes a carboxypeptidase, and thus enzymatic activity is commonly used as a surrogate measure of ACE2 protein levels [47]. Activity towards a known ACE2 substrate is assessed by carrying out the reaction in the presence and absence of an ACE2 inhibitor [41]. An important advantage of this method is the ability to measure soluble ACE2 (sACE2). The ectodomain of ACE2 is shed from the cell membrane [48], both constitutively and in response to physiological changes; however, sACE2 retains both its carboxypeptidase activity and ability to bind SARS-CoV-2 [49]. We therefore examined ACE2 activity in both kidney lysates and urine (Fig 3D and 3E). sACE2 was detectable in urine [50], likely reflecting shedding from epithelia of the proximal tubules [51], as its size makes it unlikely that it is excluded from the glomerular filtrate. ACE2 activity in both the kidney and the urine from the mAce2, MP-ACE2 and hACE2 mouse lines generally aligned with the number of Ace2/ACE2 transcripts observed by ddPCR. Sexual dimorphism was observed in all three lines, but the magnitude of the difference between male and females was increased with humanization of the locus.
## ACE2 expression in the lung and airways
The central role of the lungs in COVID-19 related morbidity and mortality is surprising given the extremely low level of expression of this gene in the human lung, with a normalized expression level of 0.8 compared to 122.0 for the small intestine, 23.0 for the kidney, and 10.5 for the heart [52,53]. In contrast, a comparison of ACE2 activity between mouse organs reported that ACE2 activity in the mouse lung was relatively high, being approximately $25\%$ that of the kidney, and no significant sex bias was observed [46]. We therefore determined whether this species difference in ACE2/Ace2 expression in the airways was reiterated in the two humanized mouse lines.
Lung, trachea, nasal epithelia were collected from mAce2, MP-ACE2, and hACE2 mice and evaluated for expression of ACE2/Ace2 by ddPCR (Fig 4A, 4C and 4E). In collection of the nasal epithelia, the respiratory and olfactory epithelia were isolated individually from the three mouse lines for mRNA expression analysis (Fig 4E). The enrichment for olfactory and respiratory epithelium was evaluated by examining expression of genes characteristic of these two tissues: Muc5a for respiratory epithelium and Ugt2a$\frac{1}{2}$ for olfactory epithelium (S3 Fig) [54,55]. As an additional indicator of levels of ACE2 protein levels, carboxypeptidase activity was determined in tissue homogenates for all tissues except olfactory epithelium (Fig 4B, 4D and 4F).
**Fig 4:** *Comparison of ACE2 expression in lung and airways of humanized mouse lines.A,C,E. ddPCR evaluation of cDNA prepared from male (♂) and female (♀): A, lung; C, trachea; E, nasal epithelium of mice of the indicated genotype. Mean number of copies per nanogram of cDNA for each group is shown in the table below the bar graphs for each tissue. B, D, F. ACE2 carboxypeptidase activity in samples evaluated for expression: B, activity in lung; D, activity in trachea; F, activity in nasal epithelium (respiratory). Activity in each sample was measured 1 hour after initiation of the reaction. For group sizes and statistical comparison between all groups in A, C, and E, see S1 Table. For group sizes and statistical comparison between all groups in B, D, and F, see S2 Table. OE, olfactory epithelium; RE, respiratory epithelium.*
Robust expression of endogenous murine *Ace2* gene was seen in the lung, with over 100 copies/ng cDNA, only four-fold lower than that observed in the kidney (~400 copies/ng). ACE2 carboxypeptidase activity paralleled this high expression, with activity approximately $25\%$ of that observed in the mAce2 kidney (Figs 3D vs 4B). In contrast, expression of the humanized ACE2 locus in the hACE2 lung was approximately 100 fold lower than in the kidney, in agreement with the relative expression of ACE2 in these two organs in humans [52]. Slightly higher ACE2 expression was seen in the lung, trachea and nasal respiratory epithelia of male hACE2 mice relative to females, although this difference did not reach the level of statistical difference. In contrast, ACE2 expression in the olfactory epithelium of female hACE2 mice was approximately eight-fold higher than that measured in male hACE2 mice and more than 15-fold higher than that measured in other parts of the airway for either sex.
Expression in the airways of the MP-ACE2 mice was generally higher than that from the endogenous mouse Ace2 locus, except in the lungs, where expression from the MP-ACE2 locus was reduced by approximately $80\%$ in female and $60\%$ in male lungs relative to expression from the Ace2 locus (Fig 4A). These results were somewhat unexpected, since expression from both loci is driven by the same endogenous mouse upstream promoter elements. ACE2 specific carboxypeptidase activity was easily measurable in the airway epithelia of the MP-ACE2 and hACE2 mice, verifying translation of the human transcripts (Fig 4B, 4D and 4F). The decrease in MP-ACE2 expression in the lung was paralleled by an approximate fourfold decrease in human ACE2 protein levels (Fig 4B).
## ACE2 expression in lung club and AT2 cells
A possible explanation for the low expression of ACE2 in the lung of the MP-ACE2 line is that the human gene is not expressed in club cells, which have been shown to be vulnerable to infection by the SARS-CoV-2 MA10 virus [56]. The sensitivity of mouse club cells to naphthalene because of their unique expression of Cyp2f2 provided a means of addressing this question [57]. Furthermore, carrying out the study with heterozygous female mice that carry one copy of Ace2 and one copy of the MP-ACE2 allowed assessment of expression from the human and mouse genes in the same tissue sample. Mice were treated for either 24 or 48 hours with naphthalene after which mRNA prepared from the lungs was evaluated for selective loss of club cells and expression of Ace2 and ACE2 was determined (Fig 5A) [58]. Within 48 hours, expression of two club cell mRNAs, Cyp2f2 (Cytochrome P450 2F2) and Scgb1a1 (Uteroglobin), could no longer be detected. Expression of Foxj1, which is expressed primarily in lung ciliated cells, was significantly increased. This is consistent with previous demonstration of the reparative role of ciliated cells, characterized by an increase in Foxj1 expression, as these cells lose their cilia and undergo squamous cell metaplasia in response to club death and detachment from the basement membrane [59]. Expression of the ciliated cell-specific Dnah6 (dynein axonemal heavy chain 6), essential in producing force for ciliary beating, is marginally reduced and then restored as the ciliated cells reestablish normal morphology and function. Expression of Ace2 could be detected 48 hour. after treatment, albeit at approximately $30\%$ of the level observed in vehicle treated animals. In comparison, all MP-ACE2 expression was lost in the naphthalene treated MP-ACE2 animals, suggesting that expression of ACE2 in the MP-ACE2 mice is largely localized to the Cyp2f2-expressing, naphthalene sensitive club cell population. Differences in ACE2 expression between the MP-ACE2 and mAce2 mice could reflect the expression of Ace2 in one or more additional cell populations in the mouse.
**Fig 5:** *Expression of ACE2 in club and alveolar type II (AT2) cells.A. Expression of ACE2 and Ace2 in female mice heterozygous for mAce2/MP-ACE2 24 and 48 hours after exposure to naphthalene (Naph) normalized to expression of ACE2 and Ace2 in the vehicle treated animals. Club and ciliated cell resistance to naphthalene was evaluated by measuring change in expression of cilia and club cell specific marker genes in samples during this same time interval. B. ACE2 expression is unaltered in heterozygous mAce2/hACE2 treated with naphthalene. C. Expression of Sftpc verifies enrichment of AT2 population from lung of mAce2/hACE2 heterozygous females. D. ACE2 expression is increased six fold by in airway epithelium enriched for AT2 cells. E. Comparison of ACE2 carboxypeptidase activity in lung lysates and BALF of mAce2 and MP-ACE2 mice. For group sizes and statistical comparison between all groups in E see S2 Table. ** p<0.01.*
Using a similar strategy, we measured ACE2 expression in female mice heterozygous for endogenous mouse Ace2 and hACE2 (Fig 5B). While a substantial decrease in the expression of Ace2 was observed 48 hours after naphthalene treatment, no change in the expression of the ACE2 was detected. These results indicate that the human and mouse promoters differ markedly in their ability to direct expression of Ace2/ACE2 to specific cell populations in the lung, with the mouse promoter driving high expression in club cells.
To determine the possible source of ACE2 expression in the hACE2 mice, populations enriched for alveolar type II (AT2) cells were isolated from heterozygous female mice. Enrichment was verified by assessing changes in expression of the AT2 specific transcript, Sftpc (Fig 5C). A six fold increase in expression of ACE2 was observed in the AT2 enriched population (Fig 5D). This is consistent with a model in which AT2 cells are the major cell type in which the human promoter drives ACE2 expression in the mouse lung.
## Shedding of ACE2 into the airway lumen
ACE2 ectodomain shedding has been reported for human airway epithelia [49] and mouse lungs, where it is reported to play a role in regulation of neutrophil influx [60]. More recently, it has been suggested that the shed ectodomain (soluble ACE2) can mediate entry of SARS-CoV-2 into cells lacking ACE2 expression [61]. Certainly, loss of ACE2 from the cell membrane will impact it’s vulnerability to viral entry [62]. To determine the sensitivity of the human protein to shedding, we measured human ACE2 in the BALF of the humanized mouse line (Fig 5E). Whole lung lavage was carried out on mAce2 and MP-ACE2 mice prior to tissue collection. Carboxypeptidase activity was determined in both the bronchoalveolar lavage fluid (BALF) and lung homogenates. ACE2 enzyme activity was observed in the lavage and lung samples of both animals, indicating that human protein is actively shed by mouse epithelium.
## Expression of the ACE2 isoforms in the humanized mouse line
The ACE2/Ace2 promoter in both human and mice is bipartite, with full length ACE2/Ace2 transcripts beginning with one of two initial exons, 1a or 1b [35]. Human RNA-seq studies indicate low levels of exon1a transcripts in most tissue examined, with the highest expression levels reaching approximately $10\%$ of those of exon1b transcripts in the small intestine [63]. Primers specific for transcripts for ACE2/Ace2 that include exon 1a were generated and used to examine expression of 1a containing transcripts in tissues from the mAce2, MP-ACE2, and hACE2 mice. In mAce2 and MP-ACE2 mice, we found that exon1a containing transcripts comprised all of the transcripts present in the mouse lung and trachea (Fig 6A). In the fully humanized hACE2 mice (Fig 6B), similar to the MP-ACE and mAce2 mice, all the ACE2 transcripts in the lung included exon1a. However, in contrast to the MP-ACE2 and mAce2 mice, no exon 1a transcripts were observed in the tracheal samples from the hACE2 mice. Thus the human promoter only supports 1b transcripts. In the kidney and intestinal tract, exon1a containing transcripts were absent in all three mouse lines (Fig 6C). This is congruent with the human RNAseq evaluation of exon1a transcripts in that we found about $10\%$ of the transcripts in the intestinal tract of the hACE2 mice included exon 1a [63]. These results indicate that inclusion of exon 1a in transcripts encoding ACE2 varies not only between species but also between tissues. This further supports a model in which expression of ACE2/Ace2 is regulated in a species- and tissue-specific manner.
**Fig 6:** *Expression of ACE2 exon1a and exon1b mRNA transcripts in wild type and humanized mouse lines and induction of exon 1c transcripts after exposure to tilorone.A-C Comparison of total ACE2/Ace2 transcripts measured by ddPCR (open bars) to those that include exon 1a in mRNA (shaded bars) isolated from the tissues and mouse lines indicated. D-E shows induction of exon1c transcripts in lung of humanized mice after tilorone (+ til) induced increase in interferon. ddPCR evaluation of ACE2 lung transcripts using PCR primers that recognized all ACE2 transcripts (open bars), primers that recognize either the short “Δ” transcript, or primers specific for the “long”, full length ACE2 mRNA (shaded bars). F. Induction of interferon responsive genes (IRG) was verified by demonstration of increased expression of Nlrc5 gene in the tilorone treated samples by qPCR. Expression is normalized to that measured in vehicle treated animals. For group sizes and statistical comparison between all groups see S1 Table. *p<0.05; ** p<0.01; *** p<0.005.*
## In vivo induction of dACE2/short ACE2 by interferon
Evaluation of changes in mRNA expression in airway epithelial cells from naïve and inflamed lungs has identified ACE2 as an interferon-stimulated gene (ISG) [64,65]. More recently, the effect of interferon on expression of both full length and variant transcripts has been extensively explored in primary epithelium and epithelial cell lines [33,63]. These studies report that interferon mediates the increase in an ACE2 mRNA transcript initiated in intron 8 from an intronic promoter, which results in the inclusion of an exon unique to primates, exon1c (see Fig 1). The transcript is predicted to encode a truncated ACE2 isoform (short or delta (d)ACE2) that lacks both the carboxypeptidase and SARS-CoV-2 binding domains of the longer protein. However, direct in vivo evidence supporting the induction of dACE2 expression in response to interferon is lacking, as is evidence for the dependence of its expression on an internal promoter and its specificity for airway epithelium. To this end, the three mouse lines, mAce2, MP-ACE2 and hACE2, were treated with tilorone, a potent interferon inducing drug [66], and expression was evaluated by ddPCR using three sets of primers: primers that recognize all human ACE2 transcripts, primers that detect only the full length transcript, and primers that detect only the dACE2 (short ACE2) transcript (Fig 6). The induction of ISG in all lines was verified by examination of induction of Nlrc5 (Fig 6F) [67]. No increase in the expression of Ace2 was observed in the tilorone treated animals. In contrast, we observed similar fold increases in the copy number of total ACE2 transcripts in both the MP-ACE2 and the hACE2 lines. The increase in expression of ACE2 in the fully humanized mouse was accompanied by increases in the dACE2 but not long ACE2 transcript. In contrast, an increase in the full length transcript was observed in the MP-ACE2 mice. Thus, while mouse Ace2 was not induced by interferon, the combined presence of the mouse promoter and the human exon/intron sequences in MP-ACE2 mice rendered the mouse promoter interferon sensitive.
## SARS-CoV-2 infection of MP-ACE2 mice
To determine the susceptibility of the humanized lines to SARS-CoV-2 infections, seven MP-ACE2 (N2 BALB/cBy) male mice, 16 weeks of age, and two mAce2 littermates were exposed intranasally (i.n.) to 105 pfu of virus (Fig 7A). No difference in the overall health of the mice was observed over the 5-day period, including no significant change in weight. At the end of this time, animals were euthanized, and homogenates were generated from the lung for evaluation of viral load and for isolation of mRNA. Virus was detected in five of the seven MP-ACE2 mice, although titers were relatively low compared to those achieved with the mouse-adapted virus [56] in all but one of the animals (Fig 7A). As expected, viral titers were below detection in the mAce2 littermate controls. As MP-ACE2 mice are generated by syntenic replacement, they do not express the mouse ACE2 receptor, and thus cell entry is mediated by human ACE2.
To verify that expression of ACE2 in the MP-ACE2 mouse line was sufficient to support viral proliferation, an additional experiment was carried out using MP-ACE2 mice on the C57BL/6N genetic background. In this case, C57BL/6N mice carrying the FOXJ1-ACE2 transgene were included as a positive control for infection and replication of SARS-CoV-2 in the lung. This mouse line, which has been used extensively for study of SARS-CoV and SARS-CoV-2 [27,29,68] carries a transgene in which human ACE2 expression is driven by the promoter of FOXJ1, a gene which in the lung is expressed by ciliated cells [69]. We verified expression of ACE2 in this line, comparing expression to that in MP-ACE2 mice and to Ace2 in wild type mice (Fig 7B). Robust expression was observed in the FOXJ1-ACE2 lung, intermediate between that of Ace2 in mAce2 mice and ACE2 in MP-ACE2 mice. Because of the rapid resolution of infection in the MP-ACE2 BALB/cBy mice, the lung SARS-CoV-2 load was examined 2 and 5 days after intranasal delivery of 1 x 105 PFU. Mock infected FOXJ1-ACE2 and MP-ACE2 mice were included as controls. Again, no loss of weight was observed in either mock or viral infected MP-ACE2 mice. Despite the lower expression of ACE2 in the MP-ACE2 mice compared to the FOXJ1-ACE2 animals, the viral loads measured in the lung of the MP-ACE2 were approximately 20-fold higher than those of the FOXJ1-ACE2 mice 2 days post infection. However, because of the small group size, the difference did not achieve significance (Fig 7C). Interestingly, despite the observed higher titers of virus in the MP-ACE2 mice compared to the FOXJ1-ACE2 line 2 days after inoculation, by day 5 the MP-ACE2 mice had cleared the virus while measurable titers remained in the FOXJ1-ACE2 animals.
**Fig 7:** *SARS-CoV2 exposure of MP-ACE2, and mAce2 mice.~16 week old mice of the indicated genotype were infected with 105 PFUs of SAR-CoV-2 or vehicle (mock infected) and evaluated at either 2 days (DPI 2) or 5 days (DPI 5) for: A,C viral titer; B, abundance of ACE2 mRNA, determined by ddPCR in FOXJ1-ACE2 transgenic mouse line; C,D expression of SARS-CoV-2 nuclear capsid genes; E, expression of lung cell-type-specific genes; D-H, expression of the indicated pro-inflammatory genes. MP-ACE2 mice were N3/N4 generation C57BL/6N, N2 generation BALB/cBy, or ~N5 generation FOXJ1-ACE2 C57BL/6 (fuchsia). mACE2 (black) are littermates of the BALB/cBy MP-ACE2 animals. For group sizes and statistical comparison between all groups see S1 Table. * p<0.05; ** p<0.01; *** p<0.005; **** p<0.001.*
mRNA was prepared from the lungs of the infected and control animals, and the levels of SARS-CoV-2 subgenomic and genomic nucleocapsid (N) RNA were determined (Fig 7D and 7E). Two days after infection of MP-ACE2 mice, high levels of N-RNA were observed, and, consistent with differences in viral load, these were 10-fold higher in the MP-ACE2 mice compared to the FOXJ1-ACE2 animals. At day 5, despite clearance of virus, high levels of N-RNA were still easily detected. N-RNA was also detected in all 7 MP-ACE2 BALB/cBy mice. As expected, given the species specificity of SARS-CoV-2, nuclear capsid RNA was not detected in RNA prepared from the lungs of the two mAce2 littermates exposed to virus.
## Host response to viral infection
We determined whether SARS-CoV-2 infection led to measurable changes in major lung epithelial cell populations by qPCR analysis of mRNA prepared from mock and SARS-CoV-2 infected FOXJ1-ACE2 and MP-ACE2 mice using a panel of probes specific for alveolar type I (AT1), AT2, ciliated, or club cells. No changes were observed in the expression of AT1 or AT2 cell markers. The pattern of change in the expression of the two club cell specific genes, Cyp2f2 and Scgb1a1 in MP-ACE2 mice followed that observed with naphthalene mediated damage to club cells. Expression was decreased two days post infection, with rapid normalization by day 5. Similarly, changes in Foxj1 expression followed those observed in mice after damage to club cells. In contrast, no significant change in the expression of these genes could be measured in the FOXJ1-ACE2 mice. However, we cannot rule out the possibility that this reflects the smaller number of these mice included in the experiment.
An analysis of the expression of cytokines and chemokines in the infected MP-ACE2 mice revealed a robust immune response to SARS-CoV-2 infection, after which homeostasis was rapidly restored. A significant increase in Il1b was observed at day 2, with normalization by day 5 (Fig 7G). Tnf and Ccl2 followed a similar pattern, with elevated expression on day 2 that normalized with reduction of the viral load by day 5 (Fig 7H). A dramatic increase in Il6, which was found to be increased six-fold relative to mock infected animals, was detected at day 2 (Fig 7G). While transcripts for Infg were more abundant in the SARS-CoV-2 exposed animals at day 2, the increase in expression did not achieve significance. However, a robust and significant increase in the INF-γ sensitive *Nlrc5* gene was observed (Fig 7H). As NLRC5 plays a key role in increased MHC class I gene expression secondary to interferon induction, we examined the expression of H2-K1 in the lung. A significant increase was observed in expression of this MHC antigen, an important component of both natural killer and cytotoxic T cell mediated removal of virally infected cells [70,71].
Histological evaluation of the lungs of the MP-ACE2 (Fig 8) mice supports a model in which viral replication in club cells leads to cell death and exfoliation of this airway epithelial population. The distal airways, the alveoli, and alveolar duct remained relatively unaffected. In contrast, loss of club cells as well as morphological changes in club cells consistent with both apoptosis and necrosis were seen throughout the length of the airway (Fig 8D–8F). However, repair was rapid, with little cell death, and repopulation of the airways was observed by 5 days post infection. In contrast, death and exfoliation of airway epithelial cells was less apparent in the Foxj-ACE2 mice, suggesting perhaps an increased tolerance to viral replication. Loss of ciliated cells was not apparent on examination of the larger airways where ciliated cells can be easily identified (Fig 8I). Surprisingly and perhaps in response to failure to clear the virus, histological changes were apparent in the distal lung, with septal thickening, increased cellularity and changes in the architecture of the alveolar unit.
**Fig 8:** *Histological analysis of Foxj1 and MP-ACE2 SARS-CoV-2 infected lungs.Tissue sections from left lobe of lungs from mock-infected and SARS-CoV-2 infected animals at 2 days post infection, fixed and stained with hematoxylin and eosin. (A-C) Mock infected MP-ACE2. (D-F) MP-ACE2 infected animals. (G-I) Foxj1-ACE2 infected animals. (A) Distal lung of a mock infected MP-ACE2 mouse showing a terminal bronchiole (tb) and the associated alveolar duct (ad) with club cells dominating the epithelium. (B) High magnification of the epithelium showing healthy club cells (arrowhead), which are easily identified by eosinophilic “club” shaped cytoplasmic protrusions into the airway. (C) Bronchi of mock-infected animal shows an epithelium dominated by ciliated cells (arrow) and submucosal smooth muscle (sm) and connective tissue (ct) characteristic of larger conducting airways. (D-F) Similar regions of lungs from MP-ACE2 SARS-CoV-2 infected mice show limited changes in architecture of the alveolar/capillary units. However, extensive damage to epithelium (particularly apoptosis/necrosis of club cells), is apparent extending from the terminal bronchiole to the bronchus. (D) Detached cells form a loosely adherent layer of cellular debris at the apical surface of the terminal bronchiole (red dashed line). Higher magnification of the epithelium throughout the airway (E,F) shows an abundance of detached epithelial cells with pyknotic nuclei and karyorrhexis, as well as cells with condensed eosinophilic cytoplasm consistent with apoptotic death (red star). (G-I) Limited damage to airway epithelium, club cells and ciliated cells was observed in the Foxj1-ACE2 mice. However, changes in the architecture of the distal lung is apparent (G, H) with loss of alveolar/capillary units, increased cellularity and septal thickening. (A,D,G) Bars represent 150 μm. Bars in remaining panels represent 50 μm.*
## Differential immune response of the Foxj1-ACE2 and MP-ACE2 mice to viral infection
An additional experiment was carried out to verify the robust proliferation of SARS-CoV-2 in the MP-ACE2 mice. A cohort of Foxj1-ACE2 mice was included to further examine the initial observation of differences between the two lines in the clearance of virus over the 5-day period following viral exposure (Fig 9). Mice were infected with 1 x 105 PFUs of virus and weighed daily. A small drop in body weight was observed during the course of the experiment but failed to reach significance for either mouse line relative to mock infected animals (Fig 9A and 9B). No change in body condition score was noted in the mice over the course of the experiment, and the gross appearance of SARS-CoV-2 infected lungs did not differ from those of mock infected animals. Consistent with studies of the smaller Foxj1-ACE2 cohort shown in Fig 7, ACE2 levels in the Foxj1-ACE2 mice were over 2 fold higher than those measured in the MP-ACE2 animals.
**Fig 9:** *Comparison of SARS-CoV-2 clearance and immune response after exposure of Foxj1-ACE2 and MP-ACE2 mice to SARS-CoV-2.(A and B) Weight loss of mice over the 5 day duration of the experiment did not differ (2 way ANOVA). (C) Expression of ACE2 determined by ddPCR in mice of the indicated genotypes. (D) Copies of SARS-CoV-2 sgRNA (nucleocapsid transcript) present in RNA prepared from the right inferior lobe. (E-H) Expression of the indicated proinflammatory gene. The level of expression in the control animals (mock or mACE2) is assigned a value of one. For MP-ACE2 mice, mAce2 littermates served as control animals, with both receiving an inoculum 105 PFU SARS-CoV-2. For the Foxj1-ACE2 mice, mock (PBS) treated mice served as controls. For group sizes and statistical comparison between all groups, see S1 Table. * p<0.05; ** p<0.01; *** p<0.005; **** p<0.001.*
Cohorts of animals were euthanized at two days and five days post infection, and ACE2 expression and viral load were assessed by measurement of nucelocapsid sgRNA transcripts present in mRNA prepared from the lung. Surprisingly but consistent with the results described above, a two log drop in sgRNA SARS-CoV-2 levels was observed in the MP-ACE2 mice, while sgRNA levels remained unaltered in the Foxj1-ACE2 animals five days post infection. A major difference in the pathogenesis of disease in the two ACE2 mouse lines was also apparent on evaluation of cytokine levels in the lung. As shown in the previous experiment, the clearance of virus in the MP-ACE2 mice was paralleled by a robust increase in transcripts for Ccl2, TNFα, and IL-6 (Fig 9E–9G). Elevated levels of Nlrc5 transcripts were observed in the MP-ACE2 mice, indicative of an interferon-mediated increase in expression of MHC antigens. In contrast, no significant increase in the levels of these markers was observed in the Foxj1-ACE2 mice. These changes in gene expression as well as the histological changes we observed on analysis of the MP-ACE2 infected mice are consistent with a model in which SARS-CoV-2 infection of lung ACE2 expressing club cells in the MP-ACE2 mice induces a rapid and protective immune response that limits viral spread. Together, the immune response and death of the infected club cells mediate clearance of virus from the lung and initiation of epithelial repair by 5 days post infection.
## Discussion
To begin to address the complexity of human and mouse ACE2 expression in vivo, with a particular focus on the expression pattern in the airways, and to determine how this complexity impacts viral disease mediated by ACE2 dependent coronaviruses, we have generated two mouse lines in which the endogenous Ace2 locus is humanized by syntenic replacement. These lines differ only in the upstream promoter driving expression of the human gene, with the MP-ACE2 line including the upstream region from the mouse gene and the hACE2 line including the upstream region from the human gene. While no one model can recapitulate all aspects of COVID, we believe that the mouse lines we have generated will provide a unique tool for studying disease pathogenesis and immune responses. In particular, they will be invaluable in understanding how variation in SARS-CoV-2 impacts its ability to circumvent immune surveillance. We therefore believe that these lines represent an improvement relative to previously reported approaches for the study of SARS-CoV-2 pathogenesis in the mouse.
The mouse ACE2 models generated to date fall into two general categories. The first category consists of transgenic models in which expression of a human ACE2 transgene is driven by an exogenous promoter. In these models, the expression pattern is determined largely by the promoter chosen, although it can be modified by the copy number of the transgene and the genomic location or locations into which the transgene has inserted. It is not uncommon for the transgene to form rolling circles prior to insertion, resulting in integration of potentially hundreds of copies into the genome. It is also important to note that, in transgenic models of this type, the endogenous mouse *Ace2* gene is still expressed. While this is not a complicating factor for coronaviruses that bind specifically to the human ACE2 protein, some viruses such as Omicron have been reported also to bind mouse ACE2.
The K18-ACE2 model, in which expression of human ACE2 is driven by the promoter of the human KRT18 gene, is the most widely used of the transgenic models [28]. It expresses high levels of human ACE2 receptors on many epithelial cell populations, both in the lung and other organs. The pattern of ACE2 expression observed in this model does not recapitulate that observed in human or in mouse. Infection of mice carrying the K18-ACE2 transgene with SARS-CoV or SARS-CoV2 results in devastating destruction of the lung parenchyma, loss of lung barrier function, and viremia [72]. The severity of the disease observed in this model has made it extremely valuable for the study of vaccines, antibodies that prevent infection, and drugs that limit proliferation of the virus. However, this model cannot recapitulate the pathophysiology and immunopathology of COVID, an infection limited to the upper and conducting airways.
We have included in our results a comparison of our new models to a second transgenic model in which expression of human ACE2 is driven by the promoter from the human FOXJ1 gene [29]. This model differs from the K18-ACE2 transgenic model in that expression of human ACE2 in the lung is primarily limited to ciliated cells, a population of epithelial cells limited to upper airways and conducting airways. Expression of ACE2 in human ciliated cells is well documented.
The two newly developed mouse lines that we describe here fall into a second category of models generated by modification of the endogenous mouse Ace2 locus. Several previously described lines also fall into this category [30,73]. The best described of these models are lines in which a human cDNA has been knocked into the second exon of the mouse locus, so that expression of human ACE2 is regulated by the mouse Ace2 promoter. These lines differ from our MP-ACE2 line in several important ways. First, although normal expression of the mouse *Ace2* gene is disrupted in these lines by insertion of the human cDNA, virtually the entire mouse *Ace2* gene, including all regulatory elements and coding sequences is still present in the endogenous locus. Second, because the introduced human coding sequence consists of a cDNA, it lacks any intronic regulatory elements present in the full-length human gene and is incapable of giving rise to the short form of ACE2. And third, the introduced human gene includes an exogenous polyadenylation signal which may influence its expression pattern in ways that are difficult to predict. Infection of the ACE2 knock in lines with SARS-CoV results in a milder, non-fatal disease phenotype. Expression of ACE2 in the airways of one of these models has been shown to occur predominantly in club cells [76]. A direct comparison of these knock in lines with our MP-ACE2 line should be helpful in further defining the regulatory impact of intragenic regulatory elements on ACE2 expression in both healthy animals and various disease states.
The previously described ACE2 knock in lines differ from our hACE2 line in that expression of human ACE2 in the hACE2 line is driven by the human ACE2 promoter. Of particular significance, similar to the expression pattern reported for human lung, we did not observe ACE2 expression in the club cells of hACE2 mice. This indicates that expression of ACE2 in this cell type is specific to mouse. Our hACE2 mice will thus provide a more authentic system than the previously described cDNA knock in models for examining the impact of environmental and genetic factors on the activity of the human gene. For example, it should be possible to determine the impact of COVID comorbidities on the activity of the human ACE2 promoter in the conducting airways and lung parenchyma.
A complementary approach that has been taken to allow the use of the mouse as a model system for SARS-CoV-2 research has been the adaptation of the human virus to the mouse. Several mouse-adapted viruses have been generated [56,74]. A major advantage of this approach is that it allows the plethora of mouse knockout lines and mouse disease models to be applied immediately to the study of this respiratory coronavirus. However, compared to the use of mice humanized by syntenic replacement, there are several limitations inherent in this strategy.
First, in addition to changes engineered into the coding sequence of the S protein to facilitate binding of the modified virus to mouse ACE2, creation of an adapted virus capable of causing severe disease required experimental evolution in vivo via serial passage of SARS-CoV-2 MA in the lungs of young adult mice. This process resulted in the introduction of non-synonymous mutations in four other viral genes including in the replicase (NSP4, NSP7, NSP8) and in accessory ORF6 [56], in addition to the changes introduced through reverse genetics into the gene encoding the S protein [68]. It is not unlikely that these changes impact immune detection, innate immune modulation, and aspects of disease pathogenesis. Defining such differences will be important for establishing the translational value of studies using mouse adapted viral strains. As SARS-CoV-2 MA10 still binds human ACE2, this could be accomplished using the lines we describe here.
A second limitation in the use of mouse adapted SARS-CoV-2 is that the approach is of limited usefulness in the study of disease and immune responses to the SARS-CoV-2 variants that will undoubtedly continue to emerge. It is reasonable to assume that these will show variable ability to bind ACE2 of different species. Some of these, such as Omicron, may display some ability to engage the mouse ACE2 receptor [75]. Direct comparison of the SARS-CoV-2 variants will require a model that lacks mouse ACE2 and expresses human ACE2 in relevant cells in the airways, such as those described here.
A final drawback to the use of mouse adapted SARS-CoV-2 is that although the expression pattern of ACE2 is generally similar in mouse and human, as indicated by the present study as well as previous reports, major species differences are also apparent. The lung stands out among those organs displaying the most remarkable quantitative and qualitative differences in ACE2 expression. In contrast to the human airways, in which ACE2 expression diminishes in the proximal to distal direction, extremely high levels of ACE2 are found in the distal mouse lung. As discussed further below, it is not unlikely that the generally higher expression of ACE2 in the mouse lung, as well as differences in the distribution of expression among cell types, impacts the progression of disease and the role played in it by different arms of the immune response. For example, most natural immunity to SARS-CoV-2 in the human population evolved without recruitment of immune populations in response to catastrophic lung damage and loss of barrier function subsequent to infection of the distal lung. Therefore, the usefulness of any mouse adapted virus will be limited by the fact that its ability to bind host cells will be determined by the murine rather than human pattern of ACE2 expression.
Overall, patterns of ACE2 expression driven by the hACE2 locus corresponded relatively well to expectations based on previous reports, including recent single-cell RNA-Seq studies. Notable differences were observed between the expression of the human and the mouse genes. Interestingly these differences in expression pattern could not be assigned entirely to the 5’ regulatory region of the gene, as the expression pattern from the MP-ACE2 locus differed from that of the endogenous mouse Ace2 locus. This supports a role for motifs within the coding region of the gene in determining gene expression. This model is consistent with the mild SARS-CoV-2 induced disease observed in mice in which the human cDNA coding segment was “knocked in” to the mouse locus, placing expression under the control of the mouse promoter [30,76]. Our studies show that cooperation between regionally unique regulatory elements in the coding region of the gene extend to a number of aspects of gene expression, including the initiation of transcription from exon 1a, 1b, or 1c; the ability of interferon to induce expression of the full length versus the dACE2 (short) transcript; and the extent of sexual dimorphism in ACE2 expression. Furthermore, in the absence of intronic murine regulatory elements, upstream murine regulatory elements alone were insufficient to direct the mouse pattern of expression in the lung.
Our studies also address the extensively discussed vulnerability of the lung to SARS-CoV-2 in light of extremely low expression of the viral receptor in this organ. The lack of a clear consensus on ACE2 expression in the respiratory tract in part reflects the lack of well characterized and validated anti-ACE2 antibodies suitable for immunohistochemistry, which has hampered identification of ACE2-expressing cell populations in the respiratory tract [77]. We chose ab15348 anti-ACE2 rabbit serum for our studies, not only because it recognizes mouse and human ACE2, but also because it has been extensively used to assess ACE2 levels in human lung and airways by immunohistochemistry [78]. Although we were able to visualize human ACE2 protein by western analysis of intestine and kidney using the ab15348 serum, attempts to visualize either mouse or human ACE2 protein in the lung using this approach were unsuccessful. Furthermore, we observed high levels of non-specific hybridization of this antibody with other proteins in both kidney and intestinal lysates. While this lack of specificity may not be observed in all lots of this antibody, it provides an explanation for inconsistencies between reports on the expression of ACE2 in human tissue samples, and thus may explain some of the difficulties in assigning expression to epithelial cell populations by immunohistochemistry. Antibodies directed to SARS-CoV-2 nucleocapsid protein have also been used as a marker for expression of ACE2 [56] in infected mice, in human lung tissue, and in primary cells. Although this approach is convenient because of the high specificity of the antibodies, these reagents provide only surrogate markers for ACE2 and therefore fall short of providing direct evidence of protein expression. In quantifying ACE2 expression in our humanized lines, we therefore chose to rely on ddPCR/qPCR analysis and, when possible, on measurement of ACE2 enzymatic activity.
In our analysis we found that, not only was ACE2 expressed at higher levels in the wildtype mouse lung than in the lungs of the humanized lines, but the pattern of expression differed substantially between wildtype and humanized mice. The high expression of Ace2 in mouse lung is consistent with the severe disease observed on infection of mice with the mouse adapted SARS-CoV-2 MA10. In contrast, severe lung disease is observed in only a small percentage of humans and is often associated with co- morbidities, including obesity, hypertension, and type II diabetes. Based on single-cell RNA-seq analysis in the human lung and airways, it has been reported previously that ACE2 is primarily expressed by ciliated cells, although expression by AT2 cells has also been well documented in the healthy lung, where $1\%$ to $3\%$ of these cells express measurable amounts of ACE2 mRNA. Studies of air-liquid interface cultures established from human lung airways have raised the possibility of an expansion of “tropism” to basal and club cells after initial infection of ciliated cells [79]. Whether this occurs during COVID-19 is not known. In contrast, our results indicate that club cells are the primary contributor to mouse lung Ace2 mRNA, as removal of these cells results in loss of over $75\%$ of the Ace2 transcripts.
On the other hand, expression of ACE2 in human club cells has not been consistently observed, and, consistent with this observation, depletion of club cells from the lungs of our hACE2 mice did not result in a measurable change in the levels of human ACE2 mRNA. This species difference in ACE2 expression by club cells can be attributed primarily to the mouse upstream regulatory elements, as expression of human ACE2 in the MP-ACE2 mice was eliminated by club cell depletion. The fact that expression of ACE2 from the MP-ACE2 locus was decreased relative to that from the endogenous mouse locus suggests that, although the expression of ACE2 in club cells requires the mouse promoter, intronic elements within the mouse *Ace2* gene can further amplify expression in club cells. Our results indicate that the low expression of ACE2 in the hACE2 mice largely reflects the lack of expression in club cells. Our studies indicate that ACE2 expression in this line can largely be assigned to AT2 cells, as expression levels increased six-fold in samples enriched for this population. In contrast, enrichment for AT2 cells did not increase expression of the mouse *Ace2* gene.
Despite the fact that mACE2 mice express ACE2 at higher levels than mACE2 mice, only the latter were susceptible to infection with SARS-CoV-2. These results reiterate the finding that even relatively low levels of expression of human ACE2 are sufficient to render cells permissive to infection by SARS-CoV-2 in vivo [68]. The increased severity of disease caused by SARS-CoV-2 MA10 in wildtype mice relative to that caused by SARS-CoV-2 in our MP-ACE2 line may be the result of expression of the endogenous ACE2 receptor in AT2 cells of wildtype mice. Club cells may represent a more dispensable cell population in the mouse lung than AT1 or AT2 cells, as suggested by the rapid recovery of the club cell population after naphthalene exposure. In comparison, treatment with agents such as paraquat, which damage AT2 and AT1 cells, results in long term damage to the lung characterized by fibrosis and loss of alveolar units.
The decrease in expression of two club cell specific mRNAs suggests that the expression profile of club cells is radically altered during viral infection and/or that, similar to their fate after naphthalene exposure, these cells are shed into the airway lumen as part of a well-defined repair process [59]. In contrast, we observed no decrease in expression of ciliated cell markers (Foxj1/Dnah6) in the lungs of the infected FOXJ1-ACE2 transgenic animals, suggesting that infection of these cells with SARS-CoV-2 may result in a very different sequence of events. Interestingly, although ACE2 expression, measured both by ddPCR and ACE2 activity, was higher in the FOXJ1-ACE2 mice than in the MP-ACE2 mice, the viral titers achieved were approximately ten-fold lower. There are a number of possible explanations for this, including differences in expression of TMRPSS on ciliated and club cells and differences in expression of other proteins required for successful co-opting of the cellular machinery by SARS-CoV-2 while avoiding triggering of immune related pathways. Despite the higher viral load after club cell infection, the lung of the MP-ACE2 mice quickly cleared the virus, while significant titers remained in the ciliated cells of infected FOXJ1-ACE2 mice. This suggests that the cell types infected by SARS-CoV-2 may influence the duration of viral production, epithelial cell death, and the types of cytokines released, thus shaping the nature of the immune response provoked by viral invasion of the lung. Examination of mRNA from the infected MP-ACE2 lungs identified effector molecules that may have a central role in what may be viewed as a protective immune response. Support for the early protective role of some cytokines has been described. For example, IL-6 has been shown to contribute to resolution of RSV infection in mice [80].
SARS-CoV viruses are distinguished from a number of related coronaviruses that cause mild disease by their use of ACE2 as the receptor by which they gain entry into cells. The expression of ACE2 therefore is an important factor in defining the tropism of the virus, and this, together with other characteristics of ACE2 expressing cells, directs disease pathogenesis. Our studies indicate that the regulation of ACE2/Ace2 expression is complex, with expression regulated differently in humans and mice as well as within individual tissues within each species. This is particularly true of ACE2 expression in the lung. An understanding of the regulatory mechanisms underlying these differences as well as the impact these differences have on various lung epithelial cell populations will be essential in improving our understanding of the factors underlying COVID-19 pathogenesis.
## Ethics and biosafety
All animal work meets the standards of the Institutional Animal Care and Use Committee at University of North Carolina at Chapel Hill as set out in guidelines outlined by the U.S. Department of Agriculture as well as by the Association for the Assessment and Accreditation of Laboratory Animal Care. Safety conditions and approved standard operating procedures have been followed for all experiments involving SARS-CoV-2. BSL3 facilities at University of North Carolina at Chapel Hill were designed to meet safety requirements recommended by the U.S. Department of Health and Human Services, Biosafety in Microbiological and Biomedical Laboratories (BMBL), the National Institutes of Health (NIH), the Centers for Disease Control and Prevention (CDC), and the Public Health Service. We have submitted laboratory safety plans, and the CDC and the University of North Carolina at Chapel Hill Department of Environmental Health and Safety (EHS) have approved the facility for use.
## Generation of mouse lines
The Ace2 displacer constructs were assembled using a standard recombineering approach. The mouse arms of homology were derived from the 129s7/AB2.2 bMQ BAC library clone bMQ306a09 (Source BioScience). The segment of human genomic DNA containing the ACE2 gene was derived from the human tile path BAC RP11-478H11 (BACPAC Resources). The resistance marker gene used for selection of embryonic stem (ES) cells in which the Displacer construct underwent genomic integration consisted of a PGK-neo cassette flanked by mutant loxP sites. Excision of the marker gene with transient expression of Cre leaves a nonfunctional lox site in its place. The recombination events for generation of both lines were carried out in the 129 ES cell line, Phnx43. ES cells used were derived from 129SvEv mice. ES cells carrying the correctly modified locus were used to generate chimeric mice, which were bred to 129S6/SvEv/Tac mice (Taconic Bioscience) to maintain the mutation on this genetic background. Chimeras were also bred to C57BL/6NCrl (Charles River Laboratories) and BALC/cBy, strain 000650 (The Jackson Laboratory) mice, and offspring were used for examination of susceptibility to SARS-CoV-2. The MP-ACE2 mouse line is designated by the MMRC as 069708-UNC, 069709-UNC and 069713-UNC; the hACE2 line is designated as 069710-UNC, 069714-UNC, and 069715-UNC; and the FOXJ1-ACE2 line is designated as 066719-UNC.
## Enrichment for alveolar type II cells
Alveolar type II cells were isolated as described previously [81–83]. Briefly, lungs were perfused and airways lavaged to remove blood and airways cells, respectively. The lungs were instilled with one mL of 10 U/mL of Dispase II (Roche) in PBS, followed by 0.5 mL of $1\%$ low melting agarose, removed and incubated in 0.5 mL of Dispase II solution for 45 minutes at room temperature. At the end of the incubation period, digested tissue was transferred to 7 mL of DMEM supplemented with $10\%$ FBS and $0.01\%$ DNase I, teased free of bronchi and bronchioles and further incubated. The resulting cell suspension was filtered through a 70 μm centrifuged at 300 x g for 10 minutes at 4°C. The cell pellet was resuspended in 1 mL of $4\%$ isotonic Percoll (Sigma, St. Louis, MO) and the suspension overlaid on a Percoll step and centrifuged at 400 x g for 20 minutes at 4°C with no brake. AT2 enriched cells were collected at the 10–$30\%$ interface and washed once with PBS. To further remove contaminants, cells were stained with biotinylated antibodies against lineage markers (anti-CD45, anti-CD$\frac{16}{32}$, anti-CD31, and anti-integrin B4), labeled with streptavidin microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany), and then subjected to magnetic separation according to manufacturer’s instructions. The enriched type II cells were collected and used for isolation of RNA.
## RNA preparation and expression analysis
Tissues were collected from male and female mice between 10 and 16 weeks of age. Total RNA was isolated using an RNA isolation solvent (Stat-60; Tel-Test, Friendswood, TX, USA) according to the manufacturer’s protocol. For qRT-PCR, RNA was reverse transcribed to cDNA using a high-capacity cDNA archive kit (Applied Biosystems) following the manufacturer’s recommended protocol. All probes and primers were purchased from a commercial vendor (Applied Biosystems). Amplification of DNA was carried out on the Applied Biosystems 7900 HT Fast RT-PCR System using qBio Blue (Genesee Scientific). Each sample was run in duplicate, and relative expression was determined by normalizing samples to 18S RNA (ΔΔCT). Expression of SARS-CoV-2 nucleocapsid in the lung was determines using the 2019-nCOV_N1 primers, and probes were as follows: Forward primer: GACCCC AAA ATC AGC GAA AT, Reverse primer: TCT GGT TAC TGC CAG TTG AAT CTG, Probe: AC CCC GCA TT ACG TTT GGT GGA CC (CDC N1 qRT-PCR assay) [84]. Comparison was to a CoV-2 standard curve RNA produced by PCR amplification of SARS-CoV-2 nucleocapsid by which a 5’ T7 polymerase promoter was introduced. This amplicon was used as template to generate in vitro transcribed RNA, which was then quantified and serially diluted (108–101 copies/μl).
## Droplet digital PCR (ddPCR)
RNA and cDNA were prepared as described above. ddPCR reactions were performed with QX200 Droplet Digital PCR System (Bio-Rad, Hercules, CA) according to the manufacturer’s instructions. Each FAM reaction mixture (20 μL) contained 10 μL ddPCR Supermix for Probes (No dUTP) (Bio-Rad), 9 μL of template cDNA (up 100 ng), and 1 μL of Primers/Probes (900 nM per primer and 250 nM probe). And SYBR Green 20 μL reaction mixtures contained 10 μL iTaq Universal SYBR Green Supermix (Bio-Rad), 1 μL of primers (250 nM each primer), and 9 μL of template cDNA (up to 100 ng). The reaction mixtures were loaded into a disposable droplet generator cartridge (Bio-Rad), and droplets were formed using Bio-Rad QX-100 emulsification device. The contents were transferred to a 96-well Eppendorf reaction plate (Genesee Scientific; Morrisville, NC) and sealed with foil using Eppendorf 96-well heat sealer. PCR amplification of the droplets was performed using a C1000 Touch Thermal Cycler (Bio-Rad) with the following parameters: 95°C for 10 min, followed by 40 cycles of 94°C for 30 sec and 60°C for 1 min, and a final 98°C for 10 min. After PCR amplification, the plate was scanned using a QX200 Droplet Reader (Bio-Rad). QX Manager Software (Bio-Rad) was used to analyze the data by calculating the absolute copy number of the target DNA (units of copies/μL) using Poisson distribution analysis. All TaqMan gene expression probes were purchased from Applied Biosystems. Primers for SYBR reactions were designed using the NCBI Primer Designing Tool.
## ACE2 activity in tissues, urine and bronchial lavage fluid
ACE2 activity in tissue lysates was measured using specific fluorogenic ACE2 substrate (Mca-APK-(Dnp) (AnaSpec, San Jose, CA) in the presence or absence of the ACE2 inhibitor (MLN-4760) (Sigma-Aldrich, St. Louis, MO) as previously described [85]. Tissue samples were homogenized in lysis buffer (75 mM Tris-HCl, pH 7.5, 1 M NaCl, 0.5 mM ZnCl2, 0.01 mM Captopril, 0.1 mM Z-Pro-Prolinal, 1mM PMSF, EDTA-free inhibitor cocktail tablet from Roche, and $0.5\%$ Triton X-100) and centrifuged at 14,000 x g for 10 minutes at 4°C. Protein concentration in tissue lysates was measured using the Bradford method. Tissue lysates (10 μg of protein for kidney extracts and 40 μg of protein for heart, lung, trachea, and sinus extracts) were pre incubated with 70 μL of assay buffer (75 mM Tris-HCl, pH 7.5, 1 M NaCl, 0.5 mM ZnCl2, 0.01 mM Captopril, 0.1 mM Z-Pro-Prolinal, and EDTA-free inhibitor cocktail tablet from Roche) with or without ACE inhibitor MLN-4760 (10 μM final) for 30 minutes at room temperature. After the incubation with ACE2 inhibitor, 30 μL of ACE2 substrate buffer (75 mM Tris-HCl, pH 7.5, 1 M NaCl, 0.5 mM ZnCl2, 0.01 mM Captopril, 0.1 mM Z-Pro-Prolinal, and 0.167 mM Mca-APK-Dpn) was added to each well to initiate the reaction. Samples were incubated in the dark for 1 hour at room temperature, and fluorescence values were measured at an excitation wavelength of 320 nm and emission wavelength of 420 nm using a BioTek Cytation5 plate reader (BioTek instruments, Winooski, VT). Results were expressed as ΔRFU (Relative Fluorescence Unit) after subtraction of RFU values obtained in the presence of MLN-4760. BALF was collected from mice as previously described, and urine was collected during necropsy.
## Naphthalene treatment
Mice received a single intraperitoneal injection of naphthalene (Sigma-Aldrich, St. Louis, MO) solution (200 mg/kg) freshly prepared in corn oil. As controls, mice were injected with vehicle only. Mice were euthanized at 24 or 48 hours after the injections. Lungs were then surgically removed, snap frozen, and stored at -80°C for subsequent RNA extraction.
## Virus strains
The virus strains icSARS-CoV-2 WT [86] and SARS-CoV-2 WT stocks were grown using Vero E6 cells and titered via plaque assay. Vero E6 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, GIBCO), $5\%$ Fetal Clone II serum (Hyclone), and 1X antibiotic/antimycotic (GIBCO). Briefly, serially diluted virus was added to a monolayer of Vero E6 cells and overlayed with media containing $0.8\%$ agarose. Plaques were counted after three days after visualization by staining with Neutral Red dye. All viral infections were carried out under biosafety level 3 (BSL-3) conditions under negative pressure. All personnel conducting viral experiments wore Tyvek suits equipped with personal powered air-purifying respirators.
## In Vivo infection
All mice were bred at UNC at Chapel Hill. Anesthetized (ketamine/xylazine) mice were intranasally infected with 1 x 105 PFU of SARS-CoV-2, whereas mock infected mice received only PBS. Mice were monitored daily for any weight loss or decrease in lung function. For determination of viral load and RNA analysis, samples were collected at indicated time points after euthanizing mice by isoflurane overdose.
## Histology
After euthanasia, the left lung lobe was harvested inflated with $10\%$ phosphate buffered formalin and then further fixed by submersion in fixative for 7 days. Lung lobes were embedded in paraffin and sectioned at 3 μm thickness. Sequential sections were stained with hematoxylin and eosin.
## Quantification and statistical analysis
Build-in functions of GraphPad Prism were used for data analyses and visualization. Specific statistical tests as well as numbers of animals are included in respective figure legends and detailed in S1–S3 Tables. These include the number of mice in each group and comparisons between all mice of each sex and genotype.
## References
1. Murgolo N, Therien AG, Howell B, Klein D, Koeplinger K. **SARS-CoV-2 tropism, entry, replication, and propagation: Considerations for drug discovery and development**. *PLoS Pathog* (2021) **17** e1009225. DOI: 10.1371/journal.ppat.1009225
2. Jackson CB, Farzan M, Chen B, Choe H. **Mechanisms of SARS-CoV-2 entry into cells**. *Nature Reviews Molecular Cell Biology* (2022) **23** 3-20. DOI: 10.1038/s41580-021-00418-x
3. Puray-Chavez M, LaPak KM, Schrank TP, Elliott JL, Bhatt DP. **Systematic analysis of SARS-CoV-2 infection of an ACE2-negative human airway cell**. *Cell Rep* (2021) **36** 109364. DOI: 10.1016/j.celrep.2021.109364
4. Hoffmann M, Kleine-Weber H, Schroeder S, Kruger N, Herrler T. **SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor**. *Cell* (2020) **181** 271-280 e278. DOI: 10.1016/j.cell.2020.02.052
5. Li W, Moore MJ, Vasilieva N, Sui J, Wong SK. **Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus**. *Nature* (2003) **426** 450-454. DOI: 10.1038/nature02145
6. Hofmann H, Pyrc K, van der Hoek L, Geier M, Berkhout B. **Human coronavirus NL63 employs the severe acute respiratory syndrome coronavirus receptor for cellular entry**. *Proc Natl Acad Sci U S A* (2005) **102** 7988-7993. DOI: 10.1073/pnas.0409465102
7. Shulla A, Heald-Sargent T, Subramanya G, Zhao J, Perlman S. **A transmembrane serine protease is linked to the severe acute respiratory syndrome coronavirus receptor and activates virus entry**. *J Virol* (2011) **85** 873-882. DOI: 10.1128/JVI.02062-10
8. Heurich A, Hofmann-Winkler H, Gierer S, Liepold T, Jahn O. **TMPRSS2 and ADAM17 cleave ACE2 differentially and only proteolysis by TMPRSS2 augments entry driven by the severe acute respiratory syndrome coronavirus spike protein**. *J Virol* (2014) **88** 1293-1307. DOI: 10.1128/JVI.02202-13
9. Donoghue M, Hsieh F, Baronas E, Godbout K, Gosselin M. **A novel angiotensin-converting enzyme-related carboxypeptidase (ACE2) converts angiotensin I to angiotensin 1–9**. *Circ Res* (2000) **87** E1-9. DOI: 10.1161/01.res.87.5.e1
10. Turner AJ, Hooper NM. **The angiotensin-converting enzyme gene family: genomics and pharmacology**. *Trends Pharmacol Sci* (2002) **23** 177-183. DOI: 10.1016/s0165-6147(00)01994-5
11. Camargo SM, Singer D, Makrides V, Huggel K, Pos KM. **Tissue-specific amino acid transporter partners ACE2 and collectrin differentially interact with hartnup mutations**. *Gastroenterology* (2009) **136** 872-882. DOI: 10.1053/j.gastro.2008.10.055
12. Hashimoto T, Perlot T, Rehman A, Trichereau J, Ishiguro H. **ACE2 links amino acid malnutrition to microbial ecology and intestinal inflammation**. *Nature* (2012) **487** 477-481. DOI: 10.1038/nature11228
13. Singer D, Camargo SM, Ramadan T, Schafer M, Mariotta L. **Defective intestinal amino acid absorption in Ace2 null mice**. *Am J Physiol Gastrointest Liver Physiol* (2012) **303** G686-695. DOI: 10.1152/ajpgi.00140.2012
14. Alenina N, Bader M. **ACE2 in Brain Physiology and Pathophysiology: Evidence from Transgenic Animal Models**. *Neurochem Res* (2019) **44** 1323-1329. DOI: 10.1007/s11064-018-2679-4
15. Guy JL, Jackson RM, Acharya KR, Sturrock ED, Hooper NM. **Angiotensin-converting enzyme-2 (ACE2): comparative modeling of the active site, specificity requirements, and chloride dependence**. *Biochemistry* (2003) **42** 13185-13192. DOI: 10.1021/bi035268s
16. Serfozo P, Wysocki J, Gulua G, Schulze A, Ye M. **Ang II (Angiotensin II) Conversion to Angiotensin-(1–7) in the Circulation Is POP (Prolyloligopeptidase)-Dependent and ACE2 (Angiotensin-Converting Enzyme 2)-Independent**. *Hypertension* (2020) **75** 173-182. DOI: 10.1161/HYPERTENSIONAHA.119.14071
17. Iwai M, Horiuchi M. **Devil and angel in the renin-angiotensin system: ACE-angiotensin II-AT1 receptor axis vs. ACE2-angiotensin-(1–7)-Mas receptor axis**. *Hypertens Res* (2009) **32** 533-536. DOI: 10.1038/hr.2009.74
18. Kragstrup TW, Singh HS, Grundberg I, Nielsen AL, Rivellese F. **Plasma ACE2 predicts outcome of COVID-19 in hospitalized patients**. *PLoS One* (2021) **16** e0252799. DOI: 10.1371/journal.pone.0252799
19. Imai Y, Kuba K, Rao S, Huan Y, Guo F. **Angiotensin-converting enzyme 2 protects from severe acute lung failure**. *Nature* (2005) **436** 112-116. DOI: 10.1038/nature03712
20. Zhang H, Baker A. **Recombinant human ACE2: acing out angiotensin II in ARDS therapy**. *Crit Care* (2017) **21** 305. DOI: 10.1186/s13054-017-1882-z
21. Vickers C, Hales P, Kaushik V, Dick L, Gavin J. **Hydrolysis of biological peptides by human angiotensin-converting enzyme-related carboxypeptidase**. *J Biol Chem* (2002) **277** 14838-14843. DOI: 10.1074/jbc.M200581200
22. Sodhi CP, Wohlford-Lenane C, Yamaguchi Y, Prindle T, Fulton WB. **Attenuation of pulmonary ACE2 activity impairs inactivation of des-Arg(9) bradykinin/BKB1R axis and facilitates LPS-induced neutrophil infiltration**. *Am J Physiol Lung Cell Mol Physiol* (2018) **314** L17-L31. DOI: 10.1152/ajplung.00498.2016
23. Ramos SG, Rattis B, Ottaviani G, Celes MRN, Dias EP. **ACE2 Down-Regulation May Act as a Transient Molecular Disease Causing RAAS Dysregulation and Tissue Damage in the Microcirculatory Environment Among COVID-19 Patients**. *Am J Pathol* (2021) **191** 1154-1164. DOI: 10.1016/j.ajpath.2021.04.010
24. Cousin VL, Giraud R, Bendjelid K. **Pathophysiology of COVID-19: Everywhere You Look You Will See ACE2!**. *Front Med (Lausanne)* (2021) **8** 694029. DOI: 10.3389/fmed.2021.694029
25. Tikellis C, Bialkowski K, Pete J, Sheehy K, Su Q. **ACE2 deficiency modifies renoprotection afforded by ACE inhibition in experimental diabetes**. *Diabetes* (2008) **57** 1018-1025. DOI: 10.2337/db07-1212
26. Patel VB, Mori J, McLean BA, Basu R, Das SK. **ACE2 Deficiency Worsens Epicardial Adipose Tissue Inflammation and Cardiac Dysfunction in Response to Diet-Induced Obesity**. *Diabetes* (2016) **65** 85-95. DOI: 10.2337/db15-0399
27. Jiang RD, Liu MQ, Chen Y, Shan C, Zhou YW. **Pathogenesis of SARS-CoV-2 in Transgenic Mice Expressing Human Angiotensin-Converting Enzyme 2**. *Cell* (2020) **182** 50-58 e58. DOI: 10.1016/j.cell.2020.05.027
28. McCray PB, Pewe L, Wohlford-Lenane C, Hickey M, Manzel L. **Lethal infection of K18-hACE2 mice infected with severe acute respiratory syndrome coronavirus**. *J Virol* (2007) **81** 813-821. DOI: 10.1128/JVI.02012-06
29. Menachery VD, Yount BL, Sims AC, Debbink K, Agnihothram SS. **SARS-like WIV1-CoV poised for human emergence**. *Proc Natl Acad Sci U S A* (2016) **113** 3048-3053. DOI: 10.1073/pnas.1517719113
30. Sun SH, Chen Q, Gu HJ, Yang G, Wang YX. **A Mouse Model of SARS-CoV-2 Infection and Pathogenesis**. *Cell Host Microbe* (2020) **28** 124-133 e124. DOI: 10.1016/j.chom.2020.05.020
31. Li Y, Cao L, Li G, Cong F, Li Y. **Remdesivir Metabolite GS-441524 Effectively Inhibits SARS-CoV-2 Infection in Mouse Models**. *J Med Chem* (2021). DOI: 10.1021/acs.jmedchem.0c01929
32. Blaschke RJ, Rappold GA. **Man to mouse—lessons learned from the distal end of the human X chromosome**. *Genome Res* (1997) **7** 1114-1117. DOI: 10.1101/gr.7.12.1114
33. Blume C, Jackson CL, Spalluto CM, Legebeke J, Nazlamova L. **A novel ACE2 isoform is expressed in human respiratory epithelia and is upregulated in response to interferons and RNA respiratory virus infection**. *Nat Genet* (2021) **53** 205-214. DOI: 10.1038/s41588-020-00759-x
34. Ye M, Wysocki J, Gonzalez-Pacheco FR, Salem M, Evora K. **Murine recombinant angiotensin-converting enzyme 2: effect on angiotensin II-dependent hypertension and distinctive angiotensin-converting enzyme 2 inhibitor characteristics on rodent and human angiotensin-converting enzyme 2**. *Hypertension* (2012) **60** 730-740. DOI: 10.1161/HYPERTENSIONAHA.112.198622
35. Pedersen KB, Chhabra KH, Nguyen VK, Xia H, Lazartigues E. **The transcription factor HNF1alpha induces expression of angiotensin-converting enzyme 2 (ACE2) in pancreatic islets from evolutionarily conserved promoter motifs**. *Biochim Biophys Acta* (2013) **1829** 1225-1235. PMID: 24100303
36. Komatsu T, Suzuki Y, Imai J, Sugano S, Hida M. **Molecular cloning, mRNA expression and chromosomal localization of mouse angiotensin-converting enzyme-related carboxypeptidase (mACE2)**. *DNA Seq* (2002) **13** 217-220. DOI: 10.1080/1042517021000021608
37. Wiener RS, Cao YX, Hinds A, Ramirez MI, Williams MC. **Angiotensin converting enzyme 2 is primarily epithelial and is developmentally regulated in the mouse lung**. *J Cell Biochem* (2007) **101** 1278-1291. DOI: 10.1002/jcb.21248
38. Itoyama S, Keicho N, Hijikata M, Quy T, Phi NC. **Identification of an alternative 5’-untranslated exon and new polymorphisms of angiotensin-converting enzyme 2 gene: lack of association with SARS in the Vietnamese population**. *Am J Med Genet A* (2005) **136** 52-57. DOI: 10.1002/ajmg.a.30779
39. Tukiainen T, Villani AC, Yen A, Rivas MA, Marshall JL. **Landscape of X chromosome inactivation across human tissues**. *Nature* (2017) **550** 244-248. DOI: 10.1038/nature24265
40. Uri K, Fagyas M, Kertesz A, Borbely A, Jenei C. **Circulating ACE2 activity correlates with cardiovascular disease development**. *J Renin Angiotensin Aldosterone Syst* (2016) **17**. DOI: 10.1177/1470320316668435
41. Pedersen KB, Sriramula S, Chhabra KH, Xia H, Lazartigues E. **Species-specific inhibitor sensitivity of angiotensin-converting enzyme 2 (ACE2) and its implication for ACE2 activity assays**. *Am J Physiol Regul Integr Comp Physiol* (2011) **301** R1293-1299. DOI: 10.1152/ajpregu.00339.2011
42. Nadarajah R, Milagres R, Dilauro M, Gutsol A, Xiao F. **Podocyte-specific overexpression of human angiotensin-converting enzyme 2 attenuates diabetic nephropathy in mice**. *Kidney Int* (2012) **82** 292-303. DOI: 10.1038/ki.2012.83
43. Chou CF, Loh CB, Foo YK, Shen S, Fielding BC. **ACE2 orthologues in non-mammalian vertebrates (Danio, Gallus, Fugu, Tetraodon and Xenopus)**. *Gene* (2006) **377** 46-55. DOI: 10.1016/j.gene.2006.03.010
44. Mizuiri S, Hemmi H, Arita M, Ohashi Y, Tanaka Y. **Expression of ACE and ACE2 in individuals with diabetic kidney disease and healthy controls**. *Am J Kidney Dis* (2008) **51** 613-623. DOI: 10.1053/j.ajkd.2007.11.022
45. Lely AT, Hamming I, van Goor H, Navis GJ. **Renal ACE2 expression in human kidney disease**. *J Pathol* (2004) **204** 587-593. DOI: 10.1002/path.1670
46. Liu J, Ji H, Zheng W, Wu X, Zhu JJ. **Sex differences in renal angiotensin converting enzyme 2 (ACE2) activity are 17beta-oestradiol-dependent and sex chromosome-independent**. *Biol Sex Differ* (2010) **1** 6. PMID: 21208466
47. Wysocki J, Ye M, Soler MJ, Gurley SB, Xiao HD. **ACE and ACE2 activity in diabetic mice**. *Diabetes* (2006) **55** 2132-2139. DOI: 10.2337/db06-0033
48. Lambert DW, Yarski M, Warner FJ, Thornhill P, Parkin ET. **Tumor necrosis factor-alpha convertase (ADAM17) mediates regulated ectodomain shedding of the severe-acute respiratory syndrome-coronavirus (SARS-CoV) receptor, angiotensin-converting enzyme-2 (ACE2)**. *J Biol Chem* (2005) **280** 30113-30119. DOI: 10.1074/jbc.M505111200
49. Jia HP, Look DC, Tan P, Shi L, Hickey M. **Ectodomain shedding of angiotensin converting enzyme 2 in human airway epithelia**. *Am J Physiol Lung Cell Mol Physiol* (2009) **297** L84-96. DOI: 10.1152/ajplung.00071.2009
50. Warner FJ, Lew RA, Smith AI, Lambert DW, Hooper NM. **Angiotensin-converting enzyme 2 (ACE2), but not ACE, is preferentially localized to the apical surface of polarized kidney cells**. *J Biol Chem* (2005) **280** 39353-39362. DOI: 10.1074/jbc.M508914200
51. Shaltout HA, Westwood BM, Averill DB, Ferrario CM, Figueroa JP. **Angiotensin metabolism in renal proximal tubules, urine, and serum of sheep: evidence for ACE2-dependent processing of angiotensin II**. *Am J Physiol Renal Physiol* (2007) **292** F82-91. DOI: 10.1152/ajprenal.00139.2006
52. Hikmet F, Mear L, Edvinsson A, Micke P, Uhlen M. **The protein expression profile of ACE2 in human tissues**. *Mol Syst Biol* (2020) **16** e9610. DOI: 10.15252/msb.20209610
53. Nawijn MC, Timens W. **Can ACE2 expression explain SARS-CoV-2 infection of the respiratory epithelia in COVID-19?**. *Mol Syst Biol* (2020) **16** e9841. DOI: 10.15252/msb.20209841
54. Jedlitschky G, Cassidy AJ, Sales M, Pratt N, Burchell B. **Cloning and characterization of a novel human olfactory UDP-glucuronosyltransferase**. *Biochem J* (1999) **340** 837-843. PMID: 10359671
55. Amini SE, Gouyer V, Portal C, Gottrand F, Desseyn JL. **Muc5b is mainly expressed and sialylated in the nasal olfactory epithelium whereas Muc5ac is exclusively expressed and fucosylated in the nasal respiratory epithelium**. *Histochem Cell Biol* (2019) **152** 167-174. DOI: 10.1007/s00418-019-01785-5
56. Leist SR, Dinnon KH, Schafer A, Tse LV, Okuda K. **A Mouse-Adapted SARS-CoV-2 Induces Acute Lung Injury and Mortality in Standard Laboratory Mice**. *Cell* (2020) **183** 1070-1085 e1012. DOI: 10.1016/j.cell.2020.09.050
57. Reid WD, Ilett KF, Glick JM, Krishna G. **Metabolism and binding of aromatic hydrocarbons in the lung. Relationship to experimental bronchiolar necrosis**. *Am Rev Respir Dis* (1973) **107** 539-551. DOI: 10.1164/arrd.1973.107.4.539
58. Mahvi D, Bank H, Harley R. **Morphology of a naphthalene-induced bronchiolar lesion**. *Am J Pathol* (1977) **86** 558-572. PMID: 842612
59. Park KS, Wells JM, Zorn AM, Wert SE, Laubach VE. **Transdifferentiation of ciliated cells during repair of the respiratory epithelium**. *Am J Respir Cell Mol Biol* (2006) **34** 151-157. DOI: 10.1165/rcmb.2005-0332OC
60. Sodhi CP, Nguyen J, Yamaguchi Y, Werts AD, Lu P. **A Dynamic Variation of Pulmonary ACE2 Is Required to Modulate Neutrophilic Inflammation in Response to Pseudomonas aeruginosa Lung Infection in Mice**. *J Immunol* (2019) **203** 3000-3012. DOI: 10.4049/jimmunol.1900579
61. Yeung ML, Teng JLL, Jia L, Zhang C, Huang C. **Soluble ACE2-mediated cell entry of SARS-CoV-2 via interaction with proteins related to the renin-angiotensin system**. *Cell* (2021) **184** 2212-2228 e2212. DOI: 10.1016/j.cell.2021.02.053
62. Kazemi S, Lopez-Munoz AD, Holly J, Jin L, Yewdell JW. **Variations in cell-surface ACE2 levels alter direct binding of SARS-CoV-2 Spike protein and viral infectivity: Implications for measuring Spike protein interactions with animal ACE2 orthologs**. *bioRxiv* (2021). DOI: 10.1101/2021.10.21.465386
63. Onabajo OO, Banday AR, Stanifer ML, Yan W, Obajemu A. **Interferons and viruses induce a novel truncated ACE2 isoform and not the full-length SARS-CoV-2 receptor**. *Nat Genet* (2020) **52** 1283-1293. DOI: 10.1038/s41588-020-00731-9
64. Ziegler CGK, Allon SJ, Nyquist SK, Mbano IM, Miao VN. **SARS-CoV-2 Receptor ACE2 Is an Interferon-Stimulated Gene in Human Airway Epithelial Cells and Is Detected in Specific Cell Subsets across Tissues**. *Cell* (2020) **181** 1016-1035 e1019. DOI: 10.1016/j.cell.2020.04.035
65. Chua RL, Lukassen S, Trump S, Hennig BP, Wendisch D. **COVID-19 severity correlates with airway epithelium-immune cell interactions identified by single-cell analysis**. *Nat Biotechnol* (2020) **38** 970-979. DOI: 10.1038/s41587-020-0602-4
66. Stringfellow DA, Glasgow LA. **Tilorone hydrochloride: an oral interferon-inducing agent**. *Antimicrob Agents Chemother* (1972) **2** 73-78. DOI: 10.1128/AAC.2.2.73
67. Meissner TB, Li A, Biswas A, Lee KH, Liu YJ. **NLR family member NLRC5 is a transcriptional regulator of MHC class I genes**. *Proc Natl Acad Sci U S A* (2010) **107** 13794-13799. DOI: 10.1073/pnas.1008684107
68. Dinnon KH, Leist SR, Schafer A, Edwards CE, Martinez DR. **A mouse-adapted model of SARS-CoV-2 to test COVID-19 countermeasures**. *Nature* (2020) **586** 560-566. DOI: 10.1038/s41586-020-2708-8
69. Ostrowski LE, Hutchins JR, Zakel K, O’Neal WK. **Targeting expression of a transgene to the airway surface epithelium using a ciliated cell-specific promoter**. *Mol Ther* (2003) **8** 637-645. DOI: 10.1016/s1525-0016(03)00221-1
70. Hoglund P, Brodin P. **Current perspectives of natural killer cell education by MHC class I molecules**. *Nat Rev Immunol* (2010) **10** 724-734. DOI: 10.1038/nri2835
71. van de Sandt CE, Barcena M, Koster AJ, Kasper J, Kirkpatrick CJ. **Human CD8(+) T Cells Damage Noninfected Epithelial Cells during Influenza Virus Infection In Vitro**. *Am J Respir Cell Mol Biol* (2017) **57** 536-546. DOI: 10.1165/rcmb.2016-0377OC
72. Winkler ES, Bailey AL, Kafai NM, Nair S, McCune BT. **SARS-CoV-2 infection of human ACE2-transgenic mice causes severe lung inflammation and impaired function**. *Nat Immunol* (2020) **21** 1327-1335. DOI: 10.1038/s41590-020-0778-2
73. Zhou B, Thao TTN, Hoffmann D, Taddeo A, Ebert N. **SARS-CoV-2 spike D614G change enhances replication and transmission**. *Nature* (2021) **592** 122-127. DOI: 10.1038/s41586-021-03361-1
74. Muruato A, Vu MN, Johnson BA, Davis-Gardner ME, Vanderheiden A. **Mouse-adapted SARS-CoV-2 protects animals from lethal SARS-CoV challenge**. *PLoS Biol* (2021) **19** e3001284. DOI: 10.1371/journal.pbio.3001284
75. Halfmann PJ, Iida S, Iwatsuki-Horimoto K, Maemura T, Kiso M. **SARS-CoV-2 Omicron virus causes attenuated disease in mice and hamsters**. *Nature* (2022). DOI: 10.1038/s41586-022-04441-6
76. Winkler ES, Chen RE, Alam F, Yildiz S, Case JB. **SARS-CoV-2 Causes Lung Infection without Severe Disease in Human ACE2 Knock-In Mice**. *J Virol* (2022) **96** e0151121. DOI: 10.1128/JVI.01511-21
77. Zamorano Cuervo N, Grandvaux N. **ACE2: Evidence of role as entry receptor for SARS-CoV-2 and implications in comorbidities**. *Elife* (2020) **9**. DOI: 10.7554/eLife.61390
78. Lee IT, Nakayama T, Wu CT, Goltsev Y, Jiang S. **ACE2 localizes to the respiratory cilia and is not increased by ACE inhibitors or ARBs**. *Nat Commun* (2020) **11** 5453. DOI: 10.1038/s41467-020-19145-6
79. Ravindra NG, Alfajaro MM, Gasque V, Huston NC, Wan H. **Single-cell longitudinal analysis of SARS-CoV-2 infection in human airway epithelium identifies target cells, alterations in gene expression, and cell state changes**. *PLoS Biol* (2021) **19** e3001143. DOI: 10.1371/journal.pbio.3001143
80. Pyle CJ, Uwadiae FI, Swieboda DP, Harker JA. **Early IL-6 signalling promotes IL-27 dependent maturation of regulatory T cells in the lungs and resolution of viral immunopathology**. *PLoS Pathog* (2017) **13** e1006640. DOI: 10.1371/journal.ppat.1006640
81. Corti M, Brody AR, Harrison JH. **Isolation and primary culture of murine alveolar type II cells**. *Am J Respir Cell Mol Biol* (1996) **14** 309-315. DOI: 10.1165/ajrcmb.14.4.8600933
82. Sinha M, Lowell CA. **Isolation of Highly Pure Primary Mouse Alveolar Epithelial Type II Cells by Flow Cytometric Cell Sorting**. *Bio Protoc* (2016) **6**. DOI: 10.21769/BioProtoc.2013
83. Weller NK, Karnovsky MJ. **Improved isolation of rat lung alveolar type II cells. More representative recovery and retention of cell polarity**. *Am J Pathol* (1986) **122** 92-100. PMID: 2934989
84. Lu X, Wang L, Sakthivel SK, Whitaker B, Murray J. **US CDC Real-Time Reverse Transcription PCR Panel for Detection of Severe Acute Respiratory Syndrome Coronavirus 2**. *Emerg Infect Dis* (2020) **26**. DOI: 10.3201/eid2608.201246
85. Wang Y, Cassis LA, Thatcher SE. **Use of a Fluorescent Substrate to Measure ACE2 Activity in the Mouse Abdominal Aorta**. *Methods Mol Biol* (2017) **1614** 61-67. DOI: 10.1007/978-1-4939-7030-8_5
86. Hou YJ, Okuda K, Edwards CE, Martinez DR, Asakura T. **SARS-CoV-2 Reverse Genetics Reveals a Variable Infection Gradient in the Respiratory Tract**. *Cell* (2020) **182** 429-446 e414. DOI: 10.1016/j.cell.2020.05.042
|
---
title: Neighborhood social organization exposures and racial/ethnic disparities in
hypertension risk in Los Angeles
authors:
- Gregory Sharp
- Richard M. Carpiano
journal: PLOS ONE
year: 2023
pmcid: PMC9987829
doi: 10.1371/journal.pone.0282648
license: CC BY 4.0
---
# Neighborhood social organization exposures and racial/ethnic disparities in hypertension risk in Los Angeles
## Abstract
Despite a growing evidence base documenting associations between neighborhood characteristics and the risk of developing high blood pressure, little work has established the role played by neighborhood social organization exposures in racial/ethnic disparities in hypertension risk. There is also ambiguity around prior estimates of neighborhood effects on hypertension prevalence, given the lack of attention paid to individuals’ exposures to both residential and nonresidential spaces. This study contributes to the neighborhoods and hypertension literature by using novel longitudinal data from the Los Angeles Family and Neighborhood Survey to construct exposure-weighted measures of neighborhood social organization characteristics—organizational participation and collective efficacy—and examine their associations with hypertension risk, as well as their relative contributions to racial/ethnic differences in hypertension. We also assess whether the hypertension effects of neighborhood social organization vary across our sample of Black, Latino, and White adults. Results from random effects logistic regression models indicate that adults living in neighborhoods where people are highly active in informal and formal organizations have a lower probability of being hypertensive. This protective effect of exposure to neighborhood organizational participation is also significantly stronger for Black adults than Latino and White adults, such that, at high levels of neighborhood organizational participation, the observed Black-White and Black-Latino hypertension differences are substantially reduced to nonsignificance. Nonlinear decomposition results also indicate that almost one-fifth of the Black-White hypertension gap can be explained by differential exposures to neighborhood social organization.
## Introduction
A prevailing feature across North American and European countries is that ethnic minority adults continue to have significantly higher rates of hypertension than their White adult counterparts [1, 2]. In the United States, recent estimates indicate that Black adults not only have a higher age-adjusted hypertension prevalence than White and Latino adults, but they also have lower rates of high blood pressure treatment and control [2–5]. Racial disparities are even wider when focusing on Los Angeles County, the most populous county in the U.S., where Black hypertension prevalence is more than twice that of White and Latino populations [6, 7]. These racially uneven patterns of hypertension are particularly alarming when considering that high blood pressure continues to be a prominent risk factor for stroke, heart failure, coronary heart disease, and all-cause mortality in the U.S. [2]. And while the contributions of individual risk factors, such as health behaviors [8], socioeconomic status (SES) [9], and psychosocial stressors [10–12] to racial differences in hypertension are well documented, little work has examined how differential exposures to neighborhood social conditions contribute to these inequalities. A better understanding of these community-wide social processes can help inform public health policies and interventions geared toward mitigating the risks of developing high blood pressure, particularly for communities of color in urban areas.
To date, research has found associations between neighborhood structural factors and high blood pressure independent of individual-level characteristics. Specifically, adults are at a heightened risk of having or developing hypertension when living in neighborhoods that are socioeconomically disadvantaged [13–15] or residentially segregated [16, 17], and devoid of healthy lifestyle resources (e.g., healthy food availability, recreational opportunities) within the built environment [18, 19]. There is also evidence, however, that sharing neighborhoods with people of similar racial/ethnic backgrounds (i.e., co-ethnics), particularly for Latinos, is protective of poor health, presumably through the diffusion of healthy behaviors, norms, and information [20, 21]. Additional research has demonstrated that chronic exposures to neighborhood stressors, such as crime and disorder [22–25] and residential segregation [17, 26], explain portions of the racial/ethnic gap in hypertension risk.
Despite prior work on neighborhood stressors, scant evidence exists on the role of neighborhood social organization and hypertension risk and disparities. In particular, collective efficacy—the extent to which neighborhood residents trust and support one another and are willing to intervene on behalf of the collective good [27]—has been linked to health and well-being, including self-rated health [28], obesity risk [29, 30], asthma [31], health behaviors [32, 33], and mental health [34]. The neighborhood social cohesion and informal social control that embody collective efficacy may be associated with hypertension risk through such mechanisms as proliferating pro-health social norms and behaviors, attracting local resources that facilitate physical activity, and ameliorating fears of crime and disorder. A related but distinct construct, neighborhood organizational participation may also associate with better health by connecting residents to integral health-promoting resources both inside and outside the local neighborhood, and by fostering a sense of community that could benefit even those who do not affiliate with organizations [32, 35–37].
There is also a lack of research investigating whether the hypertension effects of neighborhood social organization vary across racial and ethnic populations. And given that low-income, minority-concentrated neighborhoods have the capacity to build collective efficacy and mobilize resources that buffer stressful contextual conditions [35, 38], Black and Latino adults may especially benefit from living in and being exposed to tight-knit, organized communities in Los Angeles. Indeed, some qualitative studies on African Americans, for example, document healthier behaviors (e.g., dietary intake) [32] and outcomes (obesity, high blood pressure) [39, 40] in high collective efficacy communities. Taken together, we hypothesize that exposure to higher levels of neighborhood organizational participation and collective efficacy will be associated with a lower likelihood of being hypertensive and will contribute to racial/ethnic disparities in hypertension. We further contend that living in neighborhoods with greater social organization will matter more for Black and Latino adults than White adults.
Another limitation of existing studies is an overreliance on the residential neighborhood as the only consequential space for hypertension risk. For instance, people tend to spend much of their time outside of their local neighborhood performing routine activities, such as working, shopping, and exercising [41], and daily exposure to these various activity spaces could have consequences for triggering stress and developing high blood pressure. Yet, existing hypertension research conceptualizes neighborhoods as only encompassing the residential context and does not consider the amount of time spent in individuals’ nonresidential spaces. This may be particularly salient in Los Angeles where compared with White individuals, African Americans and Latinos are more likely to live in socioeconomically disadvantaged areas, as well as conduct their daily activities in disadvantaged, under-resourced, and racially isolated neighborhoods [42, 43]. Studies also show that daily mobility is facilitated or restricted by adults’ race/ethnicity, SES, and the characteristics of their activity spaces, which further conditions the duration of exposure to home and away neighborhoods [44]. As a result, not accounting for nonresidential exposures may be a source of confounding that leads to misestimated or biased residential effects on health [45, 46].
To address these research gaps, we use novel longitudinal data from the Los Angeles Family and Neighborhood Survey (L.A.FANS) to examine the role of neighborhood social organizational exposures in adults’ hypertension risk. Our study extends prior work on neighborhoods and hypertension in three important ways. First, we assess social organizational stress-buffering mechanisms (organizational participation, collective efficacy) underexplored in hypertension studies. Second, we employ a counterfactual decomposition technique to estimate the relative contributions of these neighborhood exposure characteristics to racial/ethnic hypertension gaps. Our results indicate that neighborhood social organization is not only associated with a lower risk of being hypertensive, but also accounts for roughly one-fifth of the racial disparity. Finally, compared with studies relying solely on the residential neighborhood, our measures of neighborhood context are more complete by adjusting for the amount of time people spend in their residential neighborhoods, in addition to the neighborhoods in which they conduct their routine activities (i.e., activity spaces).
## Data sources
This paper uses longitudinal data from the Los Angeles Family and Neighborhood Survey (L.A.FANS). Administered in two waves (Wave 1 in 2000–2002 and Wave 2 in 2006–2008), L.A.FANS is a stratified random sample of 65 census tracts in Los Angeles County, California sampled from three tract poverty strata: very poor (tracts in the 90th or above percentile); poor (tracts in the 60-89th percentiles); and nonpoor (tracts below the 60th percentile). In Wave 1, L.A.FANS randomly selected and interviewed adults and children from over 3,000 households across the 65 sampled tracts [47]. In Wave 2, an attempt was made to re-interview all respondents in the original sample, while also interviewing a sample of newcomers to each neighborhood, but standard in-person interviews with health-related questions were only administered to those who remained in L.A. County [48]. The State University of New York at Buffalo Institutional Review Board approved all study protocols and the use of L.A.FANS restricted data. Consent was waived because L.A.FANS is a secondary data source.
Of the roughly 2,600 originally sampled adults (age 18 and over), 1,187 were interviewed in Wave 2. Once 34 respondents who did not report an activity space location are excluded, there are 1,153 panel respondents. Due to insufficient sample sizes of other ethnic groups, we limit our study to Latino, non-Latino White, and non-Latino Black adults, resulting in 1,065 respondents. An additional 19 respondents were removed for having missing data on any of the analysis variables, yielding a final sample of 1,046 adult respondents. With a negligible portion of the sample with missing data ($1.8\%$), we employ listwise deletion, rather than multiple imputation. We structure our data longitudinally such that each observation represents one person-period, resulting in a total analytic sample of 2,092 person-periods.
L.A.FANS has respondent attrition between Waves 1 and 2. To address this issue, L.A.FANS provides panel weights to be used in all longitudinal analyses, which are a combination of the Wave 1 design weight and a Wave 2 attrition adjustment. Panel weights are designed to account for the oversampling of census tracts in the poorest strata of L.A. County, the oversampling of households with children, and the attrition of eligible Wave 1 panel members due to non-response [48]. These panel weights are also designed to make the sample representative of the L.A. County adult population at Wave 1 who reside in the county at Wave 2. L.A.FANS staff derived the attrition factor by executing logistic regression models using Wave 1 variables to predict non-response among panel respondents who at Wave 2 were not known to be ineligible (e.g., deceased, incarcerated). The inverse of the predicted probability of non-response obtained from the logistic regression models was used as the attrition weight [48]. A comparison of Wave 1 baseline characteristics indicates that panel respondents typically have more children, education, and income, as well as higher rates of employment and homeownership than respondents who left the panel.
L.A.FANS is an ideal source of data for studying how neighborhood exposures matter for individual health and well-being in Los Angeles [e.g., 42, 43, 49–55]. A key advantage of the L.A.FANS is the availability of census tract identifiers based on where respondents live, in addition to several locations respondents frequent and spend time (i.e., activity spaces). More specifically, L.A.FANS interviewers asked respondents to report the locations of five major activities: their current workplace (for all jobs), where they typically shop for groceries, where they worship, and where they obtain healthcare for illnesses and preventative care. Respondents were permitted to report up to three locations per activity in Wave 1 and up to four in Wave 2. For each activity location, respondents reported the addresses or cross-streets, from which geocodes were generated by L.A.FANS staff [48]. Another unique feature of L.A.FANS data is the ability to create neighborhood-level measures of social organization from survey items (described below). Incorporating the amount of time into our contextual exposures is a final novel benefit of using L.A.FANS.
Using this tract-level detail, we append census tract information from Census 2000 and the 2005–2009 American Community Survey to Waves 1 and 2 respondent-level data, respectively, and construct measures of adults’ neighborhood and activity space racial/ethnic and socioeconomic exposures. Despite the limitations of using census tracts as proxies for neighborhoods they are designed to be standardized in terms of their demographic, social, and economic characteristics, as well as being demarcated by discernible physical boundaries, such as bodies of water and bridges. All census tracts have been normalized to 2000 boundaries.
## Hypertension
The dependent variable is a dichotomous self-reported measure indicating whether the respondent has hypertension. Specifically, respondents were asked “*Has a* doctor ever told you that you have high blood pressure or hypertension?”
## Neighborhood measures
We include two measures of neighborhood social organization. Neighborhood organizational participation captures whether the respondent participated in a local voluntary association during the past year across nine types of groups (e.g., neighborhood block meeting). Second, collective efficacy reflects residents’ perceptions of social cohesion and informal social control at the neighborhood level [27]. Social cohesion is measured with five L.A.FANS questions capturing whether respondents perceive their neighborhood as close-knit, trustworthy, helpful, friendly, and sharing common values. Informal social control is based on four survey questions addressing the likelihood that neighbors would intervene if children in the neighborhood were disrespecting adults, skipping school, or vandalizing property, and whether adults are watchful of the neighborhood. The specific survey questions that comprise neighborhood organizational participation and collective efficacy are presented in S1 Table.
To derive our neighborhood social organization measures, we follow a well-documented “ecometric” approach to creating aggregates of survey responses pertaining to respondents’ neighborhood perceptions and behaviors [19, 27, 36, 56]. To this end, we execute three-level item response models (items nested within individuals nested within census tracts) and use empirical Bayes estimates (EB residuals) to arrive at each neighborhood’s organizational participation and collective efficacy scores, the details of which have been described elsewhere [36].
There are two neighborhood structural measures: socioeconomic disadvantage and co-ethnic density. Socioeconomic disadvantage is a widely used composite measure of neighborhood SES [57] comprised of five variables (all percentages): individuals living below the poverty line, individuals in the labor force unemployed, households on public assistance, female-headed households with children, and individuals 25 and over who did not graduate from high school. Having neighbors of the same race/ethnicity may improve individual health through the diffusion of healthy behaviors and information [20]. Co-ethnic density is the percentage of the neighborhood population that matches the race/ethnicity as the respondent based on three groups: Latino, non-Latino White, and non-Latino Black.
We also use L.A.FANS data to estimate respondents’ average time per week spent in the following activity locations: workplace, grocery store, place of worship, and healthcare. Following prior work described elsewhere [54, 55], we derive exposure weights for each respondent and then apply these weights to their home and activity space measures to arrive at new exposure-weighted scores (e.g., neighborhood collective efficacy exposure). Activity space exposure measures represent a weighted average across all respondents’ activity space contexts that reflects individuals’ overall activity space exposures rather than separate measures for each activity location [54, 55]. The global activity space measure is preferred here because the separate activity space measures (e.g., workplace, grocery store) are highly correlated with one another, whereas correlations between residential neighborhood exposure measures and overall activity space exposure measures are weak to moderate. By weighting these contextual variables by exposure, they now reflect personal contextual exposure measures at the individual level [58]. Note that collective efficacy and organizational participation have only residential exposure versions because they are based on L.A.FANS survey questions pertaining to the respondent’s current neighborhood of residence, and sample sizes across activity space neighborhoods were insufficient to create activity space social organization measures.
## Individual covariates
Beyond binary indicators for our three racial/ethnic groups (Latino, non-Latino Black, and non-Latino White), our models adjust for a range of individual-level covariates implicated in past research on neighborhoods and chronic conditions [e.g., 13, 17, 19, 55, 57]. Demographic characteristics include age (years) and binary indicators for whether the respondent is female, foreign born, married, and has children under 18 in the household. Socioeconomic characteristics are family income—the sum of earned and transfer household income in 2007 dollars and transformed using the inverse hyperbolic sine (IHS) function to account for zeros; education (years); and whether the respondent is employed and has health insurance. Additional individual-level controls are length of residence (IHS-transformed years lived in the current neighborhood) and survey wave. Table 1 presents survey-weighted descriptive statistics for all analysis variables.
**Table 1**
| Unnamed: 0 | Total Sample | Black | White | Latino |
| --- | --- | --- | --- | --- |
| Variables | Mean (SD) / % | Mean (SD) / % | Mean (SD) / % | Mean (SD) / % |
| Self-reported hypertension | 21.75 | 41.77 | 21.48 | 18.05 |
| Race/ethnicity | | | | |
| Black | 9.10 | | | |
| White | 45.05 | | | |
| Latino | 45.85 | | | |
| Residential neighborhood characteristics | | | | |
| Organizational participation | 0.34 (1.16) | 0.32 (1.10) | 0.64 (1.35) | 0.04 (0.86) |
| Collective efficacy | 0.30 (0.99) | 0.20 (0.88) | 0.63 (1.07) | 0.01 (0.81) |
| Socioeconomic disadvantage | 0.07 (0.78) | 0.41 (0.88) | -0.41 (0.60) | 0.48 (0.64) |
| Co-ethnic density | 35.56 (21.32) | 12.77 (10.54) | 38.94 (20.52) | 36.75 (21.00) |
| Activity space characteristics | | | | |
| Socioeconomic disadvantage | -0.02 (0.17) | 0.02 (0.17) | -0.06 (0.17) | 0.02 (0.15) |
| Co-ethnic density | 6.00 (7.22) | 1.90 (2.90) | 7.37 (8.09) | 5.46 (6.50) |
| Individual-level covariates | | | | |
| Age (years) | 44.05 (15.52) | 43.27 (16.20) | 48.52 (16.15) | 39.80 (13.39) |
| Female | 48.00 | 59.86 | 43.67 | 49.89 |
| Foreign born | 40.84 | 3.55 | 11.31 | 77.24 |
| Married | 49.36 | 28.47 | 53.42 | 49.51 |
| Presence of children | 46.80 | 50.51 | 30.65 | 61.94 |
| Family income (1000s) | 60.77 (73.62) | 52.52 (37.21) | 83.05 (96.79) | 40.52 (37.99) |
| Education (years) | 13.29 (4.29) | 14.52 (2.29) | 15.75 (2.75) | 10.62 (4.27) |
| Employed | 69.77 | 63.48 | 69.39 | 71.38 |
| Uninsured | 25.75 | 11.79 | 13.05 | 40.99 |
| Length of residence (years) | 9.85 (10.59) | 8.79 (8.97) | 11.96 (12.38) | 7.99 (8.38) |
| N (person-periods) | 2092 | 216 | 622 | 1254 |
## Statistical analysis
To examine the study’s first objective, we estimate associations between neighborhood social organization exposure and hypertension risk by executing a series of random effects logistic models. We choose a random effects model because of the longitudinal and multilevel structure of the data. Recall that our neighborhood exposure measures are at the respondent level resulting in a two-level model—time (survey wave) nested within individuals. Here, we prefer the random effects model to the fixed effects model because of its ability to examine both time-invariant and time-varying variables, as well as our substantive interest in between-effects (i.e., racial/ethnic disparities) [59]. Our modeling strategy proceeds as follows: Model 1 presents the baseline racial/ethnic gap in hypertension risk, with Black adults as the reference group. Model 2 enters our neighborhood social organization measures (organizational participation and collective efficacy). Model 3 adjusts for Model 2 variables and includes residential socioeconomic disadvantage and co-ethnic density, while Model 4 controls for activity space versions of socioeconomic disadvantage and co-ethnic density. The full model (Model 5) adjusts for our individual-level controls. For ease of interpretation, we convert logistic regression coefficients to average marginal effects (AMEs) with $95\%$ confidence intervals (CIs) derived from robust standard errors clustered at the individual level. We also report the intraclass correlation (ICC) for each model.
To address our second aim, we examine whether the hypertension effects of neighborhood social organization vary across racial/ethnic groups. Specifically, both neighborhood organizational participation and collective efficacy are interacted with race/ethnicity in separate fully adjusted random effects logistic models. The results are illustrated by plotting predicted probabilities of having hypertension by levels of neighborhood social organization with $95\%$ CIs.
For our final objective, we explore sources of the racial/ethnic gaps in hypertension risk—between Black and White adults and Black and Latino adults—using Fairlie’s extension of the Blinder-Oaxaca decomposition technique for nonlinear models [60]. A common approach to assessing the contributing factors to racial/ethnic disparities in high blood pressure and other chronic diseases [61, 62], decomposition methods construct a counterfactual reflecting how the Black-White gap in hypertension would change, for example, if Black adults had the same neighborhood and individual characteristics as White adults. To do so, we use estimates from group-specific logistic models and partition Black-White (and Black-Latino) differences into the part explained by observed characteristics and an unexplained part, which reflects group differences in unobserved characteristics. See S1 Appendix for a detailed description of our application of the nonlinear decomposition of hypertension disparities. We apply L.A.FANS panel survey weights to all analyses, which were executed using Stata 16 [63].
## Multivariable model results
Table 2 presents results from a series of random effects logistic regression models predicting hypertension risk. Recall that our contextual exposure (residential and activity space) measures have been adjusted for the average amount of time adults spend per week in each context. In Model 1, we see that a large statistically significant disparity in the unadjusted probability of having hypertension exists between Black adults and both White and Latino adults. For example, White and Latino adults have a 19.0 and 20.5 percentage-point lower probability of having hypertension, respectively, than Black adults. Model 2 provides some evidence that living in neighborhoods with higher levels of organizational participation reduces hypertension risk (AME = -0.020, $95\%$ CI = -0.044 to 0.004, $$p \leq 0.09$$), while collective efficacy does not reach statistical significance. Introducing residential social structural characteristics in model 3 considerably diminishes the Black-White hypertension gap (AME = -0.134, $95\%$ CI = -0.217 to -0.052). In addition, residential socioeconomic disadvantage exposure is associated with a significantly higher probability of being hypertensive (AME = 0.051, $95\%$ CI = 0.014 to 0.088).
**Table 2**
| Variables | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 |
| --- | --- | --- | --- | --- | --- |
| Race/ethnicity (ref = Black) | | | | | |
| White | -.190*** | -.175*** | -.134*** | -.125** | -.135*** |
| | (-.260, -.120) | (-.247, -.103) | (-.217, -.052) | (-.210, -.041) | (-.210, -.061) |
| Latino | -.205*** | -.209*** | -.201*** | -.195*** | -.145*** |
| | (-.273, -.137) | (-.277, -.141) | (-.273, -.130) | (-.268, -.123) | (-.221, -.069) |
| Residential neighborhood characteristics | | | | | |
| Organizational participation | | -.020 | -.016 | -.018 | -.027* |
| | | (-.044, .004) | (-.040, .008) | (-.042, .006) | (-.050, -.004) |
| Collective efficacy | | -.009 | .003 | .002 | -.003 |
| | | (-.035, .016) | (-.026, .032) | (-.027, .031) | (-.031, .025) |
| Socioeconomic disadvantage | | | .051** | .050** | .049** |
| | | | (.014, .088) | (.012, .087) | (.013, .085) |
| Co-ethnic density | | | -.001 | .000 | -.010 |
| | | | (-.011, .010) | (-.011, .010) | (-.022, .001) |
| Activity space characteristics | | | | | |
| Socioeconomic disadvantage | | | | -.038 | -.044 |
| | | | | (-.190, .115) | (-.195, .107) |
| Co-ethnic density | | | | -.029 | -.009 |
| | | | | (-.066, .008) | (-.050, .032) |
| ICC | 0.674 | 0.678 | 0.682 | 0.678 | 0.589 |
| Individual controls | No | No | No | No | Yes |
Model 4 of Table 2 adjusts for activity space socioeconomic disadvantage and co-ethnic density exposures. Doing so slightly reduces the magnitude of the Black-White and Black-Latino gaps in hypertension, and these activity space social structural exposures are not significantly associated with hypertension risk. In the fully adjusted model (Model 5), neighborhood organizational participation is significantly associated with a lower probability of having high blood pressure (AME = -.027, CI = -.050 to -.004). We also see that accounting for the complete battery of contextual and individual characteristics reduces the Black-White and Black-Latino hypertension disparities to 13.5 and 14.5 percentage points, respectively, but the gaps remain statistically significant (see S2 Table for full model results).
## Do the effects of neighborhood social organization vary across racial/ethnic groups?
We also examine whether the hypertension effects of exposure to neighborhood social organizational characteristics vary across Black, Latino, and White adults. To do so, we interact our neighborhood organizational participation and collective efficacy measures with race/ethnicity in separate fully adjusted random effects logistic models—i.e., including all the variables from Model 5 in Table 2 (for interaction model results, see Models 6a and 6b in S2 Table). In Fig 1, we present predicted probabilities and $95\%$ CIs of hypertension risk based on high (one standard deviation above the mean) and low (one standard deviation below the mean) levels of residential organizational participation for Black, White, and Latino adults. The figure illustrates a strong protective effect for Black individuals living in neighborhoods where their neighbors are actively involved in local organizations and associations. For example, the probability of being hypertensive is over $63\%$ lower when Black adults reside in neighborhoods with high (compared with low) levels of organizational participation (.58 vs..21, $p \leq .05$). By contrast, neighborhood participation plays a minimal role on hypertension risk among White and Latino adults. Also noteworthy, the racial gap in high blood pressure is effectively eliminated at high levels of organizational participation, as evidenced by the overlapping confidence intervals.
**Fig 1:** *Predicted probabilities of having hypertension by levels of neighborhood organizational participation and racial/ethnic group.Data from L.A.FANS Waves 1 and 2. Estimates from fully adjusted random effects logistic model.*
In Fig 2, a similar pattern exists where the Black disadvantage in hypertension is significant at low levels of neighborhood collective efficacy, whereas the Black-White and Black-Latino gaps are not significantly different when considering highly efficacious residential neighborhoods. Though imprecise, there is some evidence that the probability that Black adults have high blood pressure is reduced by $34\%$ when living in high (versus low) collective efficacy neighborhoods (.50 vs..33, p = n.s.).
**Fig 2:** *Predicted probabilities of having hypertension by levels of neighborhood collective efficacy and racial/ethnic group.Data from L.A.FANS Waves 1 and 2. Estimates from fully adjusted random effects logistic model.*
## Decomposition of racial/ethnic disparities in hypertension risk
Next, we employ a counterfactual nonlinear decomposition [60] to examine how racial/ethnic disparities in hypertension risk would change if Black adults had the same neighborhood social environment characteristics, as well as individual characteristics, as White and Latino adults, respectively. In Fig 3, each bar represents the percentage of the Black-White gap (20.3 percentage points, see Table 1) explained by social environment characteristics. Beginning with neighborhood social organizational features, almost $10\%$ of the Black-White hypertension disparity is due to different levels of residential organizational participation in Black and White adults’ home neighborhoods. By contrast, the negative percentage explained for neighborhood collective efficacy indicates that the Black-White gap would widen by $8.2\%$ if Black and White residents lived in neighborhoods with comparable levels of collective efficacy. Fig 3 also shows that racial differences in levels of residential co-ethnic density explain nearly $25\%$ of the Black-White hypertension gap. An additional $5.3\%$ of the gap is explained by Black-White disparities in levels of residential socioeconomic disadvantage. Finally, equalizing Black and White levels of activity space social structural environments would increase the hypertension disparity by roughly $4\%$.
**Fig 3:** *Decomposition of Black-White differences in hypertension risk by neighborhood exposure characteristics.Whites are used as the reference group. The y-axis represents the percentage of the hypertension gap explained by each variable or group of variables.*
With respect to the observed Black-Latino hypertension disparity, Fig 4 illustrates that a minimal portion of the Black-Latino hypertension gap can be attributed to neighborhood social organizational and structural characteristics. Our observed variables are not substantial sources of the Black-Latino gap, which is unsurprising considering that in Los Angeles, Black and Latino adults face comparable risk factors in terms of their place exposures and individual characteristics (see Table 1).
**Fig 4:** *Decomposition of Black-Latino differences in hypertension risk by neighborhood exposure characteristics.Latinos are used as the reference group. The y-axis represents the percentage of the hypertension gap explained by each variable or group of variables.*
## Discussion
Drawing on recent advances in place and health effects research and analyzing unique multilevel data on Los Angeles County residents, we investigate whether and how neighborhood social organizational characteristics matter for hypertension risk and contribute to racial/ethnic hypertension disparities. Our findings reveal that neighborhood organizational participation is associated with a lower probability of being hypertensive. This suggests that living in neighborhoods where people are involved in informal and formal organizations and associations (e.g., neighborhood watch, civic groups, ethnic pride organizations) may protect residents against excessive exposure to area stressors that can elevate the risks of developing high blood pressure.
Comparatively, we find that neighborhood collective efficacy (social cohesion, expectations for informal social control) is not significantly associated with being hypertensive. This is consistent with one study reporting a null effect of neighborhood social cohesion on hypertension [18], but contrary to another study finding that social cohesion is associated with a lower risk of hypertension [19].
More compelling, we find that the protective effect of neighborhood organizational participation is significantly stronger for Black adults than Latino and White adults, respectively. Black residents who live in highly organized communities have a lower risk of being hypertensive by over $60\%$ compared with living in neighborhoods with low levels of organizational involvement where the Black disadvantage in hypertension risk is at its widest. Moreover, at high levels of neighborhood organizational participation, these observed hypertension differences between Black adults and Latino and White adults are substantially narrowed and no longer statistically significant (see Fig 1). This is perhaps unsurprising, given that prior research on neighborhood collective action in Los Angeles indicates that Black residents organize, affiliate, and mobilize to solve local neighborhood problems (e.g., crime, disorder), thereby reducing stress and blood pressure levels [35]. Though not as striking as organizational participation, we also show that the hypertension effects of neighborhood collective efficacy vary across racial/ethnic groups, such that the hypertension disparity between Black adults and Latino and White adults is no longer significant at high levels of collective efficacy. Echoing extant work, this highlights the salience of neighborhood social connectedness, mutual trust, and a willingness to act on behalf of fellow neighbors as a potential stress-buffering mechanism for African Americans [32].
Results from our decomposition of the racial/ethnic gap in hypertension shows that almost one-fifth of the Black-White hypertension disparity can be explained by differences in the neighborhood social organization exposures of Black and White adults’ residential communities. In addition, almost a quarter of the Black-White gap in high blood pressure can be attributed to differential exposures to neighborhood co-ethnic density, which is in line with studies reporting protective ethnic density effects for racial/ethnic minority health outcomes and behaviors [20, 64]. To better understand how neighborhood social exposures contribute to hypertension risk among Black and Latino adults, neighborhood organizational participation and other stress-buffering mechanisms should be prioritized in future research. One ethnographic study of Black adults with hypertension, for example, points to excessive contextual stressors, such as unsafe local surroundings and a lack of access to health-promoting resources (adequate healthcare, healthy food options) as exacerbating high blood pressure [65]. Another potential mechanism we encourage researchers to investigate is the evolving built environment in Los Angeles, particularly the role of community organizations, such as nonprofits [66], libraries [67], and other routine organizations geared toward improving the health, safety, and overall well-being of low-income, communities of color [68]. Gentrification processes should also be explored, as recent work suggests that living in gentrifying neighborhoods is equally beneficial for Black, White, and Latino residents [69, 70].
Aligning with the burgeoning activity space and health literature, we consider both spatial and temporal dimensions of exposure by the amount of time adults spend in their residential and nonresidential neighborhoods. Doing so provides more conservative estimates of neighborhood effects on hypertension risk and avoids common pitfalls associated with conventional neighborhood studies that do not account for individuals’ daily mobility over time and space [71, 72]. Chaix and colleagues [45], for example, refer to this type of bias as the “residential effect fallacy,” which results from not accounting for individuals’ nonresidential exposures and the subsequent confounding with residential exposures. This is corroborated in other studies noting that people’s daily mobility exposures to nonresidential places confounds or attenuates residential neighborhood effects on health [46, 50]. In our study, we find that our exposure-weighted contextual results are substantively similar to unweighted results, but we consider these estimates more conservative than those using measures based on the residential neighborhood not accounting for durations of exposure.
Our study has some limitations. Due to our use of L.A.FANS data, our results may not be generalizable beyond Los Angeles County. In addition, our dependent variable is based on a self-reported measure of hypertension, rather than resting seated blood pressure measurements. Prior research has shown that individuals may not be aware that they have hypertension, and while self-reports may underreport the prevalence of high blood pressure, there is general consistency with physician diagnoses [73–75]. Yet, our prevalence estimates for Black, Latino, and White adults are in line with both self-reported and measured hypertension from nationally representative surveys [76–78]. Even more important, our survey-weighted results are on par with measured hypertension prevalence estimates from Los Angeles County during our study timeframe [7]. Thus, we consider any underreporting of hypertension in the L.A.FANS to be minimal and not induce substantial bias into our estimates, and that our conclusions regarding the role of neighborhood social organization exposures and hypertension risk in Los Angeles would hold for measured hypertension.
Another limitation is that L.A.FANS does not ascertain an exhaustive list of respondents’ routine activities and their locations. If some routine activities occur outside the residential neighborhood that are not captured by L.A.FANS (e.g., visits with family and friends) then exposure weights for our activity space measures will be underestimated and our contextual estimates biased toward zero. Linking individual-focused epidemiological data to creative, theoretically grounded area-level measures should be a top priority in future research. These include use of GPS tracking [79], ecological momentary assessments (EMAs) [80], and qualitative interviews to better capture spatial and temporal dynamics of adult neighborhood exposures. On this front, researchers should pursue spatially fluid indicators of community social organization that can add insights into how different ethnic groups engage with and perceive their home and away contexts. Finally, given that our neighborhood social exposure measures are based on residential compositions, these may or may not be indicative of the actual social environments to which people are exposed at different times of the day. The racial/ethnic and socioeconomic daily trajectories of neighborhoods may evolve throughout the day, and thus real-time estimates of exposure should be collected to gain more precise neighborhood exposure effects on health behaviors and outcomes.
In conclusion, that racial/ethnic disparities in hypertension persist even when accounting for multiple neighborhood and individual factors suggests that researchers are presented with challenges for thinking about the myriad mechanisms through which social and behavioral conditions impact biological states and conditions. A logical starting point for conceptualization and measurement is longstanding structural racism, which, in the U.S., has profoundly shaped the residential and broader geographical circumstances of not only Black Americans, but immigrants and other people of color [81, 82]. Extending this more broadly to international contexts (for which research on contextual determinants of blood pressure has predominantly focused on residential neighborhoods versus activity spaces), future research must consider the potential impacts of urban and rural policies that shape the daily health-related circumstances, opportunities, and risks of residential neighborhoods as well as the locations that different populations occupy or inhabit throughout their daily routines.
## References
1. Agyemang C, Kunst A, Bhopal R, Zaninotto P, Unwin N, Nazroo J. **A cross-national comparative study of blood pressure and hypertension between English and Dutch South-Asian–and African-origin populations: The role of national context**. *Am J Hypertens* (2010) **23** 639-48. DOI: 10.1038/ajh.2010.39
2. Tsao CW, Aday AW, Almarzooq ZI, Alonso A, Beaton AZ, Bittencourt MS. **Heart disease and stroke statistics—2022 update: A report from the American Heart Association**. *Circulation* (2022) **145**. DOI: 10.1161/CIR.0000000000001052
3. Aggarwal R, Chiu N, Wadhera RK, Moran AE, Raber I, Shen C. **Racial/ethnic disparities in hypertension prevalence, awareness, treatment, and control in the United States, 2013 to 2018**. *Hypertension* (2021) **78** 1719-26. DOI: 10.1161/HYPERTENSIONAHA.121.17570
4. Hardy ST, Chen L, Cherrington AL, Moise N, Jaeger BC, Foti K. **Racial and ethnic differences in blood pressure among US adults, 1999–2018**. *Hypertension* (2021) **78** 1730-41. DOI: 10.1161/HYPERTENSIONAHA.121.18086
5. Thomas SJ, Booth JN, Dai C, Li X, Allen N, Calhoun D. **Cumulative incidence of hypertension by 55 years of age in Blacks and Whites: The CARDIA study**. *J Am Heart Assoc* (2018) **7** e007988. PMID: 29997132
6. Eidem E, Nagano S, Steinberg L, Johnson E, Lightstone AS, Cui Y. *Los Angeles County Department of Public Health Office of Women’s Health* (2017) 28
7. Hales CM, Carroll MD, Simon PA, Kuo T, Ogden CL. **Hypertension prevalence, awareness, treatment, and control among adults aged ≥18 years—Los Angeles County, 1999–2006 and 2007–2014**. *MMWR Morb Mortal Wkly Rep* (2017) **66** 846-9. PMID: 28817553
8. Bassett DR, Fitzhugh EC, Crespo CJ, King GA, McLaughlin JE. **Physical activity and ethnic differences in hypertension prevalence in the United States**. *Prev Med* (2002) **34** 179-86. DOI: 10.1006/pmed.2001.0969
9. Williams DR, Mohammed SA, Leavell J, Collins C. **Race, socioeconomic status and health: Complexities, ongoing challenges and research opportunities**. *Ann N Y Acad Sci* (2010) **1186** 69-101. DOI: 10.1111/j.1749-6632.2009.05339.x
10. Cuevas AG, Williams DR, Albert MA. **Psychosocial factors and hypertension**. *Cardiol Clin* (2017) **35** 223-30. PMID: 28411896
11. Brondolo E, Love EE, Pencille M, Schoenthaler A, Ogedegbe G. **Racism and hypertension: A review of the empirical evidence and implications for clinical practice**. *Am J Hypertens* (2011) **24** 518-29. DOI: 10.1038/ajh.2011.9
12. Hicken MT, Lee H, Morenoff J, House JS, Williams DR. **Racial/ethnic disparities in hypertension prevalence: Reconsidering the role of chronic stress**. *Am J Public Health* (2014) **104** 117-23. DOI: 10.2105/AJPH.2013.301395
13. Claudel SE, Adu-Brimpong J, Banks A, Ayers C, Albert MA, Das SR. **Association between neighborhood-level socioeconomic deprivation and incident hypertension: A longitudinal analysis of data from the Dallas heart study**. *Am Heart J* (2018) **204** 109-18. DOI: 10.1016/j.ahj.2018.07.005
14. Cubbin C, Hadden WC, Winkleby MA. **Neighborhood context and cardiovascular disease risk factors: The contribution of material deprivation**. *Ethn Dis* (2001) **11** 687-700. PMID: 11763293
15. Wagner KJP, Boing AF, Subramanian S, Höfelmann DA, D’Orsi E. **Effects of neighborhood socioeconomic status on blood pressure in older adults**. *Rev Saúde Pública* (2016) **50** 78. DOI: 10.1590/S1518-8787.2016050006595
16. Kershaw KN, Robinson WR, Gordon-Larsen P, Hicken MT, Goff DC, Carnethon MR. **Association of changes in neighborhood-level racial residential segregation with changes in blood pressure among black adults: The CARDIA study**. *JAMA Intern Med* (2017) **177** 996. DOI: 10.1001/jamainternmed.2017.1226
17. Gao X, Kershaw KN, Barber S, Schreiner PJ, Do DP, Diez Roux AV. **Associations between residential segregation and incident hypertension: The Multi‐Ethnic Study of Atherosclerosis**. *J Am Heart Assoc* (2022) **11** e023084. DOI: 10.1161/JAHA.121.023084
18. Kaiser P, Diez Roux AV, Mujahid M, Carnethon M, Bertoni A, Adar SD. **Neighborhood environments and incident hypertension in the Multi-Ethnic Study of Atherosclerosis**. *Am J Epidemiol* (2016) **183** 988-97. DOI: 10.1093/aje/kwv296
19. Mujahid MS, Diez Roux AV, Morenoff JD, Raghunathan TE, Cooper RS, Ni H. **Neighborhood characteristics and hypertension**. *Epidemiology* (2008) **19** 590-8. DOI: 10.1097/EDE.0b013e3181772cb2
20. Bécares L, Shaw R, Nazroo J, Stafford M, Albor C, Atkin K. **Ethnic density effects on physical morbidity, mortality, and health behaviors: A systematic review of the literature**. *Am J Public Health* (2012) **102** e33-66. DOI: 10.2105/AJPH.2012.300832
21. Viruell-Fuentes EA, Ponce NA, Alegría M. **Neighborhood context and hypertension outcomes among Latinos in Chicago**. *J Immigr Minor Health* (2012) **14** 959-67. DOI: 10.1007/s10903-012-9608-4
22. Agyemang C, van Hooijdonk C, Wendel-Vos W, Ujcic-Voortman JK, Lindeman E, Stronks K. **Ethnic differences in the effect of environmental stressors on blood pressure and hypertension in the Netherlands**. *BMC Public Health* (2007) **7** 118. DOI: 10.1186/1471-2458-7-118
23. Mujahid MS, Roux AVD, Cooper RC, Shea S, Williams DR. **Neighborhood stressors and race/ethnic differences in hypertension prevalence (The Multi-Ethnic Study of Atherosclerosis)**. *Am J Hypertens* (2011) **24** 187-93. DOI: 10.1038/ajh.2010.200
24. Mujahid MS, Moore LV, Petito LC, Kershaw KN, Watson K, Diez Roux AV. **Neighborhoods and racial/ethnic differences in ideal cardiovascular health (the Multi-Ethnic Study of Atherosclerosis)**. *Health Place* (2017) **44** 61-9. DOI: 10.1016/j.healthplace.2017.01.005
25. Mayne SL, Moore KA, Powell-Wiley TM, Evenson KR, Block R, Kershaw KN. **Longitudinal associations of neighborhood crime and perceived safety with blood pressure: The Multi-Ethnic Study of Atherosclerosis (MESA)**. *Am J Hypertens* (2018) **31** 1024-32. DOI: 10.1093/ajh/hpy066
26. Kershaw KN, Diez Roux AV, Burgard SA, Lisabeth LD, Mujahid MS, Schulz AJ. **Metropolitan-level racial residential segregation and black-white disparities in hypertension**. *Am J Epidemiol* (2011) **174** 537-45. DOI: 10.1093/aje/kwr116
27. Sampson RJ, Raudenbush SW, Earls F. **Neighborhoods and violent crime: A multilevel study of collective efficacy**. *Science* (1997) **277** 918-24. DOI: 10.1126/science.277.5328.918
28. Browning CR, Cagney KA. **Neighborhood structural disadvantage, collective efficacy, and self-rated physical health in an urban setting**. *J Health Soc Behav* (2002) **43** 383-99. PMID: 12664672
29. Cohen DA, Finch BK, Bower A, Sastry N. **Collective efficacy and obesity: The potential influence of social factors on health**. *Soc Sci Med* (2006) **62** 769-78. DOI: 10.1016/j.socscimed.2005.06.033
30. Ullmann SH, Goldman N, Pebley AR. **Contextual factors and weight change over time: A comparison between U.S. Hispanics and other population sub-groups**. *Soc Sci Med* (2013) **90** 40-8. DOI: 10.1016/j.socscimed.2013.04.024
31. Cagney KA, Browning CR. **Exploring neighborhood-level variation in asthma and other respiratory diseases: The contribution of neighborhood social context**. *J Gen Intern Med* (2004) **19** 229-36. DOI: 10.1111/j.1525-1497.2004.30359.x
32. Hughes-Halbert C, Bellamy S, Briggs V, Bowman M, Delmoor E, Kumanyika S. **Collective efficacy and obesity-related health behaviors in a community sample of African Americans**. *J Community Health* (2014) **39** 124-31. DOI: 10.1007/s10900-013-9748-z
33. Jackson N, Denny S, Sheridan J, Zhao J, Ameratunga S. **The role of neighborhood disadvantage, physical disorder, and collective efficacy in adolescent alcohol use: A multilevel path analysis**. *Health Place* (2016) **41** 24-33. DOI: 10.1016/j.healthplace.2016.07.005
34. Ahern J, Galea S. **Collective efficacy and major depression in urban neighborhoods**. *Am J Epidemiol* (2011) **173** 1453-62. DOI: 10.1093/aje/kwr030
35. Altschuler A, Somkin CP, Adler NE. **Local services and amenities, neighborhood social capital, and health**. *Soc Sci Med* (2004) **59** 1219-29. DOI: 10.1016/j.socscimed.2004.01.008
36. Carpiano RM. **Neighborhood social capital and adult health: An empirical test of a Bourdieu-based model**. *Health Place* (2007) **13** 639-55. DOI: 10.1016/j.healthplace.2006.09.001
37. Stockdale SE, Wells KB, Tang L, Belin TR, Zhang L, Sherbourne CD. **The importance of social context: neighborhood stressors, stress-buffering mechanisms, and alcohol, drug, and mental health disorders**. *Soc Sci Med* (2007) **65** 1867-81. DOI: 10.1016/j.socscimed.2007.05.045
38. Swaroop S, Morenoff JD. **Building community: The neighborhood context of social organization**. *Soc Forces* (2006) **84** 1665-95
39. Al-Bayan M, Islam N, Edwards S, Duncan DT. **Neighborhood perceptions and hypertension among low-income black women: A qualitative study**. *BMC Public Health* (2016) **16** 1075. DOI: 10.1186/s12889-016-3741-2
40. Coulon SM, Wilson DK, Alia KA, Van Horn ML. **Multilevel associations of neighborhood poverty, crime, and satisfaction with blood pressure in African-American adults**. *Am J Hypertens* (2016) **29** 90-5. DOI: 10.1093/ajh/hpv060
41. Cagney KA, York Cornwell E, Goldman AW, Cai L. **Urban mobility and activity space**. *Annu Rev Sociol* (2020) **46** 623-48
42. Browning CR, Calder CA, Krivo LJ, Smith AL, Boettner B. **Socioeconomic segregation of activity spaces in urban neighborhoods: Does shared residence mean shared routines?**. *RSF Russell Sage Found J Soc Sci* (2017) **3** 210-31. DOI: 10.7758/RSF.2017.3.2.09
43. Krivo LJ, Washington HM, Peterson RD, Browning CR, Calder CA, Kwan MP. **Social isolation of disadvantage and advantage: The reproduction of inequality in urban space**. *Soc Forces* (2013) **92** 141-64
44. Shareck M, Frohlich KL, Kestens Y. **Considering daily mobility for a more comprehensive understanding of contextual effects on social inequalities in health: A conceptual proposal**. *Health Place* (2014) **29** 154-60. DOI: 10.1016/j.healthplace.2014.07.007
45. Chaix B, Duncan D, Vallée J, Vernez-Moudon A, Benmarhnia T, Kestens Y. **The “residential” effect fallacy in neighborhood and health studies: Frmal definition, empirical identification, and correction**. *Epidemiology* (2017) **28** 789-97. PMID: 28767516
46. Kwan MP. **The neighborhood effect averaging problem (neap): An elusive confounder of the neighborhood effect**. *Int J Environ Res Public Health* (2018) **15** 1841. DOI: 10.3390/ijerph15091841
47. Peterson CE, Sastry N, Pebley AR, Ghosh-Dastidar B, Williamson S, Lara-Cinisomo S. *The Los Angeles Family and Neighborhood Survey: Codebook* (2004)
48. Peterson CE, Pebley AR, Sastry N, Yuhas K, Ghosh-Dastidar B, Haas AC. *The Los Angeles Family and Neighborhood Survey, Wave 2: User’s Guide and Codebook* (2011)
49. Inagami S, Cohen DA, Finch BK. **Non-residential neighborhood exposures suppress neighborhood effects on self-rated health**. *Soc Sci Med* (2007) **65** 1779-91. DOI: 10.1016/j.socscimed.2007.05.051
50. Sharp G, Denney JT, Kimbro RT. **Multiple contexts of exposure: Activity spaces, residential neighborhoods, and self-rated health**. *Soc Sci Med* (2015) **146** 204-13. DOI: 10.1016/j.socscimed.2015.10.040
51. Browning CR, Calder CA, Boettner B, Tarrence J, Khan K, Soller B. **Neighborhoods, activity spaces, and the span of adolescent exposures**. *Am Sociol Rev* (2021) **86** 201-33. DOI: 10.1177/0003122421994219
52. Browning CR, Calder CA, Soller B, Jackson AL, Dirlam J. **Ecological networks and neighborhood social organization**. *Am J Sociol* (2017) **122** 1939-88. DOI: 10.1086/691261
53. Jones M, Pebley AR. **Redefining neighborhoods using common destinations: Social characteristics of activity spaces and home census tracts compared**. *Demography* (2014) **51** 727-52. DOI: 10.1007/s13524-014-0283-z
54. Kimbro RT, Sharp G, Denney JT. **Home and away: Area socioeconomic disadvantage and obesity risk. Health Place**. (2017) **44** 94-102
55. Sharp G, Kimbro RT. **Neighborhood social environments, healthy resources, and adult diabetes: Accounting for activity space exposures**. *Health Place* (2021) **67** 102473. DOI: 10.1016/j.healthplace.2020.102473
56. Mujahid MS, Diez Roux AV, Morenoff JD, Raghunathan T. **Assessing the measurement properties of neighborhood scales: From psychometrics to ecometrics**. *Am J Epidemiol* (2007) **165** 858-67. DOI: 10.1093/aje/kwm040
57. Morenoff JD, House JS, Hansen BB, Williams DR, Kaplan GA, Hunte HE. **Understanding social disparities in hypertension prevalence, awareness, treatment, and control: The role of neighborhood context**. *Soc Sci Med* (2007) **65** 1853-66. DOI: 10.1016/j.socscimed.2007.05.038
58. Kwan MP. **From place-based to people-based exposure measures**. *Soc Sci Med* (2009) **69** 1311-3. DOI: 10.1016/j.socscimed.2009.07.013
59. Bell A, Fairbrother M, Jones K. **Fixed and random effects models: Making an informed choice**. *Qual Quant* (2019) **53** 1051-74
60. Fairlie RW. **An extension of the Blinder-Oaxaca decomposition technique to logit and probit models**. *J Econ Soc Meas* (2005) **30** 305-16
61. Basu S, Hong A, Siddiqi A. **Using decomposition analysis to identify modifiable racial disparities in the distribution of blood pressure in the United States**. *Am J Epidemiol* (2015) **182** 345-53. DOI: 10.1093/aje/kwv079
62. Gaskin DJ, Zare H, Jackson JW, Ibe C, Slocum J. **Decomposing race and ethnic differences in CVD risk factors for mid-life women**. *J Racial Ethn Health Disparities* (2021) **8** 174-85. DOI: 10.1007/s40615-020-00769-9
63. 63StataCorp. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC; 2019.. *Stata Statistical Software: Release 16* (2019)
64. Yang TC, Lei L, Kurtulus A. **Neighborhood ethnic density and self-rated health: Investigating the mechanisms through social capital and health behaviors**. *Health Place* (2018) **53** 193-202. DOI: 10.1016/j.healthplace.2018.08.011
65. Koehler K, Lewis L. **F. Cronholm P. Neighborhood and social influences on blood pressure: An exploration of causation in the explanatory models of hypertension among African Americans**. *J Community Med* (2018) **1** 1002
66. Sharkey P, Torrats-Espinosa G, Takyar D. **Community and the crime decline: The causal effect of local nonprofits on violent crime**. *Am Sociol Rev* (2017) **82** 1214-40
67. Klinenberg E.. *Palaces for the People: How Social Infrastructure Can Help Fight Inequality, Polarization, and the Decline of Civic Life* (2018)
68. Small ML, Gose LE. **How do low-income people form survival networks? Routine organizations as brokers**. *Ann Am Acad Pol Soc Sci* (2020) **689** 89-109
69. Agbai CO. **Shifting neighborhoods, shifting health: A longitudinal analysis of gentrification and health in Los Angeles County**. *Soc Sci Res* (2021) **100** 102603. DOI: 10.1016/j.ssresearch.2021.102603
70. Smith GS, McCleary RR, Thorpe RJ. **Racial disparities in hypertension prevalence within US gentrifying neighborhoods**. *Int J Environ Res Public Health* (2020) **17** 7889. DOI: 10.3390/ijerph17217889
71. Matthews SA, Yang TC. **Spatial polygamy and contextual exposures (spaces): Promoting activity space approaches in research on place and health**. *Am Behav Sci* (2013) **57** 1057-81. DOI: 10.1177/0002764213487345
72. Kwan MP. **The uncertain geographic context problem**. *Ann Assoc Am Geogr* (2012) **102** 958-68
73. Wellman JL, Holmes B, Hill SY. **Accuracy of self‐reported hypertension: Effect of age, gender, and history of alcohol dependence**. *J Clin Hypertens* (2020) **22** 842-9. DOI: 10.1111/jch.13854
74. Yoon SSS, Ostchega Y, Louis T. **Recent trends in the prevalence of high blood pressure and its treatment and control, 1999–2008**. *NCHS Data Brief* (2010) 1-8. PMID: 21050532
75. Brown AF, Ang A, Pebley AR. **The relationship between neighborhood characteristics and self-rated health for adults with chronic conditions**. *Am J Public Health* (2007) **97** 926-32. DOI: 10.2105/AJPH.2005.069443
76. Borrell LN, Crawford ND. **Disparities in self-reported hypertension in Hispanic subgroups, non-Hispanic black and non-Hispanic white adults: The National Health Interview Survey**. *Ann Epidemiol* (2008) **18** 803-12. DOI: 10.1016/j.annepidem.2008.07.008
77. Fang J, Yang Q, Ayala C, Loustalot F. **Disparities in access to care among US adults with self-reported hypertension**. *Am J Hypertens* (2014) **27** 1377-86. DOI: 10.1093/ajh/hpu061
78. Ong KL, Cheung BMY, Man YB, Lau CP, Lam KSL. **Prevalence, awareness, treatment, and control of hypertension among United States adults 1999–2004**. *Hypertension* (2007) **49** 69-75. DOI: 10.1161/01.HYP.0000252676.46043.18
79. Zenk SN, Schulz AJ, Matthews SA, Odoms-Young A, Wilbur J, Wegrzyn L. **Activity space environment and dietary and physical activity behaviors: A pilot study**. *Health Place* (2011) **17** 1150-61. DOI: 10.1016/j.healthplace.2011.05.001
80. York Cornwell E, Goldman AW. **Neighborhood disorder and distress in real time: Evidence from a smartphone-based study of older adults**. *J Health Soc Behav* (2020) **61** 523-41. DOI: 10.1177/0022146520967660
81. Hicken MT, Kravitz-Wirtz N, Durkee M, Jackson JS. **Racial inequalities in health: Framing future research**. *Soc Sci Med* (2018) **199** 11-8. DOI: 10.1016/j.socscimed.2017.12.027
82. Williams DR, Lawrence JA, Davis BA. **Racism and health: Evidence and needed research**. *Annu Rev Public Health* (2019) **40** 105-25. DOI: 10.1146/annurev-publhealth-040218-043750
|
---
title: 'Morpho-radiological and brain endocast analysis in the study of Hyperostosis
Frontalis Interna (HFI): A combined approach'
authors:
- Elena Varotto
- Francesco Pio Cafarelli
- Francesca Maglietta
- Cícero Moraes
- Pietrantonio Ricci
- Francesco Maria Galassi
journal: PLOS ONE
year: 2023
pmcid: PMC9987830
doi: 10.1371/journal.pone.0281727
license: CC BY 4.0
---
# Morpho-radiological and brain endocast analysis in the study of Hyperostosis Frontalis Interna (HFI): A combined approach
## Abstract
The purpose of this study is to anatomically evaluate the impact on the patient intra vitam of an endocranial condition on a late 20th century skull housed in the Section of Legal Medicine of the University of Foggia (Foggia, Apulia, Italy). After performing a retrospective diagnosis, the condition is framed in the broader context of studies on this pathology. An anthropological and radiological analysis (X-ray and CT scan imaging) made it possible to confirm the preliminary information and to detail the osteological diagnosis of HFI. In order to assess the impact on the cerebral surface of the endocranial growth a 3D endocast was obtained using the Software OrtogOnBlender. The skull is demonstrated to have belonged to a female senile individual known, from limited documentary evidence, to have suffered from a psychiatric condition during her life. The final diagnosis is hyperostosis frontalis interna (HFI), Type D. Although a direct correlation between the demonstrated intracranial bony growth and the onset of the patient’s psychiatric condition is difficult to retrospectively ascertain, the pressure exerted on this female individual’s frontal lobe may have contributed to further degenerative behavioural changes in the last years of her life. This case adds to previous knowledge, especially from the palaeopathological literature, on this condition and, for the first time, presents a neuroanatomical approach to assess the global impact of the disease.
## Introduction
The Section of Legal Medicine of the University of Foggia (Foggia, Apulia, Italy) has hosted for decades in its archives the fragments of a skull which had been poorly restored with excessive layers of an unspecified type of glue (Fig 1A) and in a state of preservation which, given the grayish-brown colour of the cranial surfaces and the presence of some pebbles obstructing some natural cranial foramina, suggested that it may have been buried underground for some time.
**Fig 1:** *a. The skull at the moment of its discovery; b. The skull during the correct restoration; c. The final stage of the restoration process; d. The completely restored skull.*
This skull has been preserved in the university hospital for research purposes, after being examined during a legal case. Now it belongs to the hospital and is being made available for scientific purposes since no living relatives are known to exist. In the box in which the skull was found there were some notes on the case to which the skull pertained dating back to the 1990, probably left by the medical examiner who first analysed it; on January 1999 this coroner carried out the exhumation of the woman to whom the skull had belonged. The woman, who died about 80 years old and about 7 years before her exhumation took place, had allegedly been affected by a psychiatric illness, although no documental evidence certifying a more precise diagnosis has been found so far.
From the morphological perspective, having noted a peculiar morphology of the endocranial surface of the frontal bone, a multidisciplinary research group analysed this specimen through a full anthropological and radiological study in order to clarify the nature of that visually detected aberrant anatomy.
## Materials and methods
The fragmented skull was washed in order to remove the glue and then newly restored with water-based glue (polyvinyl acetate) with the aim of preserving the bone tissue better and properly reconstruct as much of the skull as possible (Fig 1B–1D).
Standard methods used in bioarchaeological and forensic anthropological studies were utilised in order to assess the biological profile from the skull. The sex of the individual to whom the skull belonged was determined using the Ferembach et al. method [1], while age at death was estimated from ectocranial sutures using the Meindl and Lovejoy method [2]. A general visual pathological assessment of the entire skull was performed and compared to similar morphologies described in the relevant literature.
The restored skull was then subjected to a radiological investigation and a 3D virtual reconstruction: The CT scans were used to generate a multiplanar reconstruction (MPR) of the skull and a video (vd. S1 Video) using the OsiriX Lite version 11.0.3 (Pixmeo Software).
Subsequently, the software OrtogOnBlender [3] was used in order to reconstruct the skull from the CT scans and extract the endocast representing the brain sheathed by the encephalic meninges.
OrtogOnBlender is an add-on for surgical planning that expands Blender’s capabilities so that it performs tasks that were not originally developed, such as reading, viewing, editing, reconstructing and exporting CT scans. Both of the aforementioned tools are free and open source.
Finally, as typically performed in the multidisciplinary study in human remains [4–7], after a morphological description of the frontal endocranial lesion and the formulation of a retrospective diagnosis, a contextualisation of this case in the broader field of biological anthropological studies was provided.
The tomography was reconstructed in a 3D mesh [8] using the threshold with the value of -300 (Fig 2A, right side). A female mature adult skull of European ancestry from a virtual donor (Fig 2A, left side) was imported to serve as a reference for the tighter fitting of the pieces, as the skull was broken and slightly deformed by analogical collage. The base skull was segmented into pieces corresponding to the original cracks (Fig 2C and 2D) and the pieces were coupled in the best possible fit (Fig 2F) respecting the crack structure and the reference of the virtual donor (Fig 2E). Once the studied skull was adapted onto the virtual model, the pieces were joined in a Boolean mesh [9]. Through a 3D mesh editing process, the inner shell of the reconstructed skull was used to generate the endocranium [10] (Fig 2G and 2H). Once the skull and endocranium model were closed (without holes), an aggregated tomography was created resulting in a DICOM file with different densities for the skull and endocranium volumes (Fig 2I–2L).
**Fig 2:** *a-l. Steps of the 3D generation of the endocranium.*
The skull, as shown above, has some regions missing (Fig 3A) and this reflects the generation of an incomplete endocranium (Fig 3A and 3B). By way of comparison, an endocranium was created from a complete skull (Fig 3D), it was aligned with the incomplete endocranium and as several regions lacked structure, the authors chose part of the frontal surface where in both the structure was complete (Fig 3E and 3F). Alignment was performed manually in a digital scene, in order to make the most objective adjustments without any mesh overlap, since it made it possible to make objects transparent, create clipping borders (cuts as in floor plans) and other commands that helped to observe the interlocking of the pieces (links to more information on this technique: https://docs.blender.org/manual/en/latest/render/shader_nodes/shader/transparent.html; https://docs.blender.org/manual/en/2.79/editors/3dview/navigate/clip.html).
**Fig 3:** *a-g. Brain endocast reconstruction and volume of the brain including the meninges. The area of the frontal lobe affected by the pathological process is indicated by a rectangle.*
Once the region with Boolean calculations was isolated, two objects were generated (Fig 3G) whose volumes were calculated.
In order to have more meaningful results, a larger sample was used for morphological comparison: 18 skulls available for scientific research and anonymised in author C. M.’s tomographic collection were selected, 10 skulls of individuals with European ancestry (Brazilian Europeans) and 8 skulls of Europeans, all belonging to female adults. The skulls, already digitised (Fig 4A), were aligned with that with the endocranial alterations (Fig 4C), and received the same volumetric segmentation with Boolean calculations applied to the skull with the investigated alteration (Fig 4B and 4D).
**Fig 4:** *a-d. Phases of the alignment process between the digitised donors’ skulls with the one with the anatomical alteration studied in this article.*
## Results
Considering the morphological features of the skull, it was determined to have belonged to a female individual, whereas her age category was estimated to be that of a senile (score for sites 1–7 ‘vault’ sutural ages: 21), based on the total closure of the cranial sutures that could be observed.
The ectocranial surface of the skull showed no alterations. However, on the endocranial surface of the skull, an aberrant morphology was detected: the frontal bone, except for the midline (insertion line of the falx cerebri), is affected by an irregurlarly elevated surface (more than $50\%$ of the whole area) with clearly demarcated borders having the shape of nodular continuous bony overgrowths. Some smaller nodules can also be found on the parietal surfaces near the coronal suture (Fig 5) sparing the grooves for the ascending branches of the middle meningeal arteries: the left one, nonetheless, is deeper than normally found.
**Fig 5:** *Detail of the bony overgrowth on the endocranial surface of the frontal and parietal bones (black arrows).On the left half of the frontal bone, very close to the coronal suture, it is possible to see the deep groove of the ascending branch of the middle meningeal artery (red arrow).*
As early as during the restoration process a marked thickening of the cranial vault could be appreciated, as it would be confirmed by the radiological examination. This showed that the frontal sinuses are almost entirely obliterated by the bony growth, which filled their natural pneumatic spaces: this condition can be noticed more clearly in an antero-posterior projection radiograph (Fig 6) as well as on coronal, axial and sagittal CT-scans (Fig 7A–7C). Moreover, the entire skull is affected by thickening of the cranial tables, including at the occipital level (Fig 7C). The thickness of the frontal bone (inner table + diploe + outer table), measured at four different sites, yields an average of 11.75 mm. The thickness of the right parietal bone is 6.32 mm, and that of the contralateral is 6.47 mm.
**Fig 6:** *Postero-anterior X-ray with measurement of the thickness of the frontal and parietal bones indicating the hyperostosis.The obliteration of the frontal sinuses can be observed.* **Fig 7:** *CT-scans in all projections: a. coronal view, b. axial view, c. sagittal view. The thickening and extent of the bony growth in the form of nodules protruding into the cranial cavity can be appreciated. A general thickening of the parietal and occipital bones can also be seen.*
Specifically in the frontal bone, the ectocranial plate retained its normal appearance, while the diploic space and endocranial plate were thickened.
The nodular appearance of the endocranial surface can also be properly evaluated on 3D virtual reconstructions of the skull (Fig 8A and 8B). The calculated endocranial volume results in 70.54 cm³ in the specimen with the endocranial alteration and 98.55 cm³ in the non-pathological skull adopted as a comparison.
**Fig 8:** *3D virtual reconstruction: a. superior view of the restored skull; b. coronal section of the frontal bone with the software OsiriX Lite.*
Furthermore, even when more skulls are considered in comparative terms for statistical reasons, the reduction in endocranial volume is confirmed. Indeed, the volumetry of the donor’s skull used above was closer to 100 cm³ than the higher concentrations of volumes from the 18 skulls (average volume = 85.56 cm³ with SD ± 6.92). Additionally, it can be underlined that the skull with HFI was larger than the average of the donors, so that, even in the case of a larger skull, the endocranial volumetry was smaller than that of all the donors (Fig 9).
**Fig 9:** *Graph showing the distribution of endocranial volumes (expressed in cm3) in the skull with HFI vs in the skulls used for comparisons.*
## Discussion and conclusion
Hyperostosis frontalis interna (henceforth HFI), whose aetiology is currently unknown–hormonal influence on bone growth being postulated–is often an incidental finding, with a prevalence in the general population between $5\%$ and $12\%$ and a predilection for the female sex [11]. HFI is defined as a remodelling of the inner plate of the frontal bone, typically presenting at the level of the frontal eminence of the endocranial plate, into a more cancellous morpho-phenotype, with midline, ex-suture, sparing, was described as being at the border between normal and pathological anatomy for the first time by Santorini and Morgagni in the 18th century [12]. As Hawkins and Martin put it in 1965, ‘[d]espite a voluminous literature very little is known about hyperostosis frontalis interna. Thus its mode of formation and growth is unknown, the incidence among the general population is uncertain, and the question as to whether it is merely an anatomical anomaly or a pathological lesion remains un-answered.’ Much of that scientific uncertainty still holds true to this day [13].
From the anatomical point of view, at gross examination HFI is characterised by an increased volume and porosity of the inner plate and diploe of the frontal bone with nodular or sessile benign overgrowths in the inner aspect of the osseous surfaces; it may affect the bone in a focal or diffuse fashion, and normally it is present bilaterally and symmetrically. This condition can also extend to the parietal bones: if their involvement is complete, this presentation is named hyperostosis frontoparietalis [14]. However, as demonstrated by Hershkovitz and colleagues through electron microscopy, in HFI it is only the endocranial layer that changes in its morphology [15]. It should, nonetheless, be underlined that in the case presented here a generalised diploic thickening can be seen, which could point to a combination of HFI and cranial thickness as described by May and colleagues [16].
HFI has been rarely reported in bioarchaeological studies, most of the existing communications on this condition are in the form of single reports [15, 17].
The absence of involvement of the outer plate of the frontal bone is essential to make a differential diagnosis with some common metabolic bone diseases, such as rickets, acromegaly and Paget’s disease [15]. In addition, regarding neoplasms, meningioma, endosteal osteoma or osteosarcoma were excluded because they are rarely multifocal [12, 15]. The typically thickened inner table of the frontal bone is radiographically described as dense and spherical structures, as a sort of ‘bullets’, tending to protrude into the diploe and can also protrude into the cranial cavity like ‘string of pearls’ [12]. HFI can be associated with rare syndromes such as Morgagni’s syndrome (HFI, obesity, virilism), Stewart-Morel syndrome (HFI, obesity, mental disturbances), and Troell-Junet syndrome (HFI, acromegaly, toxic goitre, and diabetes mellitus) [12, 18–20].
Although these case reports often state that several clinical conditions seem to be associated with HFI, such as testosterone/estrogen dysmetabolism, psychiatric disease, obesity, diabetes, although in hazard/risk studies no significant differences were found between HFI and control groups, so that it should be considered an independent entity. Currently, there is not a univocal link between signs and symptoms and HFI, and the only evidence is about its reported association with elderly post-menopausal women [15, 21].
Very important in the study of this condition are also the anthropological examinations by Hershkovitz and colleagues, who investigated over 2,000 skulls from different geographical sites and historical periods, down to the 19th century AD and belonging to various ethnic groups, which did not have signs of HFI, while HFI was identified in $24\%$ (female) and $5\%$ (male) of 1,700 skulls dated to the 20th century [15]. This discrepancy was explained by the fact that high HFI prevalence is due to greater longevity, above all in women as a result of a long estrogen stimulation. In addition, the more pronounced longevity, as a result of improved life conditions could have been influenced by the environment [15]. To explain this phenomenon, some researchers have hypothesised that food availability during human evolution determines a metabolic rate increase, that leads to leptin level rise; leptin, for its part, controlling hypothalamic metabolic pathways, influence BMI, energy expenditure and adrenal tone [22].
The data derived from this multidisciplinary analysis allow us to confirm that the skull may indeed have belonged to a female senile individual, hence confirming the scant information found together with the skull fragments. From the pathological perspective and based on the literature summarised above, a confident retrospective diagnosis of HFI can be formulated, particularly in the light of the similarity with other cases from the scientific literarure with the endocranial layer described as showing a presentation compatible with the radiological findings already described by Hershkovitz and colleagues: ‘a hyperdense layer, a ballooned, vascularised area, and a thin cortical shell encapsulating it’ [15] (Fig 4A–4C). The condition appears advanced and severe considering that the thickness of the frontal bone is much greater than the width usually recorded in non-pathological skulls [23].
As far as a more precise classification of HFI is concerned, Hershkovitz and colleagues proposed four types of HFI, based on the quantity of bone involvement and extension, appearance, shape and location of the lesions [15]. Nikolic analysed the aspects and occurrence of different types of HFI, including the study 248 of deceased females, with HFI found in 45 of them ($18.4\%$), and demonstrated that HFI has no correlation with age [24].
Based on these considerations, the case we presented can be reasonably catalogued as a severe Type D HFI (i.e. continuous nodular bony formations involving more than $50\%$ of the endocranium of the frontal bone) of the classification by Hershkovitz and colleagues [15].
Compared to other morphological studies on HIF, particularly those on skeletal remains from anthropological and forensic contexts, in the present study a neuroanatomical approach, mediated by the above-seen generation of a brain endocast, allows for additional considerations on the presentation of this condition and its neuro-psychiatric correlations. The brain endocast shows a conspicuous atrophy of the woman’s brain at the level of the frontal pole of the cerebrum, with a particular involvement of the frontal lobe. Particularly affected by the inward growth of the endocranium is the prefrontal cortex which appears atrophic as a result of a mass-effect exerted by the bone tissue pushing onto the subjacent cerebral structure. Brodmann areas 9 and 10 appear to be the ones mostly affected by the process, with lesser, yet detectable involvement of areas 8 and 46. In the occipital lobe the HFI appears to be causing a limited reduction in volume of areas 17 and 18. Especially, *Brodmann area* 9 is involved in several functions including short-term memory, inductive reasoning, attributing intention, auditory verbal attention, empathy, etc [25]. It was reported to be affected in bipolar disorder [26]. In 1953, Notkin considered several types of interplay between cranial changes and reported neuropsychiatric symptoms, one of which was the possibility that the psychosis he had diagnosed in some patients ‘could be the result of interference with the function of the brain by direct action of the structural bone changes’, particularly as he had observed in one case showing ‘definite neurologic signs’, hence suggesting a mechanism of direct action onto the brain, although he also considered more possibilities such as the fact that both psychotic and bone changes could result from common unknown aetiologic factors [27]. Notkin noted that in favour of this last possibility was the evidence that some of the cases analysed in his study ‘started with various functional types of psychosis’ and that they ‘finally changed into organic reactions’, whereas other cases ‘showed signs of organic deterioration practically from the beginning’ [27]. A similar causative correlation was more recently proposed by Gilbert et al. in 2012 [28], who, just like in the skeletal case presented here, studied a type-D presentation of HFI, which had even reached a much more advanced stage affecting not only the prefrontal cortex but also the frontal one. The major advantage of Gilbert’s study was that the scientific team could perform an MRI in an intra vitam patient and could access this postmenopausal woman’s clinical file, which included a diagnosis of psychiatric disorders inclusive of a ‘tendency towards paranoia, emotional indifference and aggressiveness’. Interestingly Gilbert and colleagues highlighted ‘a progressive worsening in behaviour’, which can be postulated to be compatible with the contextual growth of the inner cranial table [28]. A similar anatomo-clinical correlation can be reasonably postulated since in the case discussed here an initial alteration of the endocranial surface was noted in the areas of the frontal bone close to the coronal suture and in the more frontal part of the parietal bones, hence being the harbinger of further inward expansion and involvement of the frontal cortex.
Unfortunately, from the purely clinical perspective, the detrimental loss of information on the woman whose skull has been the object of the present investigation does not allow us to make any confident attempt at precisely correlating anatomical alterations with behavioural changes and neurological symptoms, yet this approach shows how in forensic and anthropological cases the study of the brain endocast can help scientists retrospectively understand the impact of cerebral changes induced by describable endocranial alterations. In addition, reinforcing some previous clinical case reports, this our analysis suggests that a larger neuroradiological study ought to be implemented in living patients with known neuropsychiatric diagnoses in order to monitor and stage the progression of this complex endocrine-osteological-neurological entity. This could be matched by more post-mortem assessments in both forensic and anatomical contexts. Once more, the study of the dead could help shed light on the nature of pathological processes, and ultimately help the living.
## Ethical statement
This study initially started after being authorised by the then director of the Section of Legal Medicine of the University of Foggia in 2018, co-author of this study Prof. Pietrantonio Ricci. This case falls under the umbrella of the Italian Police Mortuary Rules (DPR 09.10.1990 n° 285, art. 83). For the skull with HFI no consent was applicable. The present study also used a bank of anonymised CT scans of individuals of different ancestry (virtual donors), being Brazilians, Moldovans and Malaysians, under the protocol USMKK/PPP/JEPeM [259.3[2]], which received ethical approval from the Human Research Ethics Committee, Universiti Sains Malaysia. Co-author of this article Cicero Moraes confirms that he was authorized to used data from their dataset for his own research, including the present study.
## References
1. Ferembach D, Schwidetzky I, Stloukal M. **Recommandations pour déterminer l’âge et le sexe sur le squelette**. *Bull Mem Soc Anthropol Paris* (1979.0) **6** 7-45
2. Meindl RS, Lovejoy CO. **Ectocranial suture closure: a revised method for the determination of age at death based on the lateral-anterior sutures**. *Am J Phys Anthropol* (1985.0) **68** 57-66. PMID: 4061602
3. 3Moraes C, Dornelles R, da Rosa E. OrtogOnBlender—O que é e Aspectos Técnicos. Figshare, 2020. 10.6084/m9.figshare.12923729.v1. DOI: 10.6084/m9.figshare.12923729.v1
4. Varotto E, Magro MT, Brancato R, Lubritto C, Memeo L, Galassi FM. **Unique Osteoid Osteoma of the Frontal Sinus From the Late Roman Empire**. *J Craniofac Surg* (2019.0) **30** 965-966. DOI: 10.1097/SCS.0000000000005312
5. Galassi FM, Varotto E, Angelici D, Picchi D. **Further Paleoradiological Evidence of Frontal Sinus Osteoma in Ancient Egypt**. *J Craniofac Surg* (2020.0) **31** 604-605. DOI: 10.1097/SCS.0000000000006240
6. Habicht ME, Bianucci R, Buckley SA, Fletcher J, Bouwman AS, Öhrström LM. **Queen Nefertari, the Royal Spouse of Pharaoh Ramses II: A Multidisciplinary Investigation of the Mummified Remains Found in Her Tomb (QV66)**. *PloS One* (2016.0) **11** e0166571. DOI: 10.1371/journal.pone.0166571
7. Seiler R, Habicht ME, Rühli FJ, Galassi FM. **First-time complete visualization of a preserved meningeal artery in the mummy of Nakht-ta-Netjeret (ca. 950 BC)**. *Neurol Sci* (2019.0) **40** 409-411. DOI: 10.1007/s10072-018-3565-1
8. 8Moraes C, Dornelles R, da Rosa E. Sistema de Reconstrução de Tomografia Computadorizada Baseado no Slicer 3D e no DicomToMesh. Figshare, 2021. 10.6084/m9.figshare.13513890.v1.. DOI: 10.6084/m9.figshare.13513890.v1
9. 9https://docs.blender.org/manual/en/latest/modeling/modifiers/generate/booleans.html (last accessed on 27th February 2021).
10. Abdullah JY, Saidin M, Rajion ZA, Hadi H, Mohamad N, Moraes C, Abdullah JM. **Using 21st-Century Technologies to Determine the Cognitive Capabilities of a 11,000-Year-Old Perak Man Who Had Brachymesophalangia Type A2**. *Malays J Med Sci* (2021.0) **28** 1-8. DOI: 10.21315/mjms2021.28.1.1
11. Western AG, Bekvalac JJ. **Hyperostosis frontalis interna in female historic skeletal populations: Age, sex hormones and the impact of industrialization**. *Am J Phys Anthropol* (2017.0) **162** 501-515. DOI: 10.1002/ajpa.23133
12. She R, Szakacs J. **Hyperostosis Frontalis Interna: Case Report and Review of Literature**. *Ann Clin Lab Sci* (2004.0) **34** 206-208. PMID: 15228235
13. Hawkins TD, Martin L. **Incidence of Hyperostosis Frontalis Interna in Patients at a General Hospital and at a Mental Hospital**. *J Neurol Neurosurg Psychiatry* (1965.0) **28** 171-174. DOI: 10.1136/jnnp.28.2.171
14. Tripathi M, Bal C, Damle NA, Singhal A. **Hyperostosis fronto-parietalis mimicking metastasis to the skull: Unveiled on SPECT/CT**. *Indian J Nucl Med* (2012.0) **27** 272-273. DOI: 10.4103/0972-3919.115406
15. Hershkovitz I, Greenwald C, Rothschild BM, Latimer B, Dutour O, Jellema JM. **Hyperostosis frontalis interna: an anthropological perspective**. *Am J Phys Anthropol* (1999.0) **109** 303-325. DOI: 10.1002/(SICI)1096-8644(199907)109:3<303::AID-AJPA3>3.0.CO;2-I
16. May H, Mali Y, Dar G, Abbas J, Hershkovitz I, Peled N. **Intracranial volume, cranial thickness, and hyperostosis frontalis interna in the elderly**. *Am J Hum Biol* (2012.0) **24** 812-819. DOI: 10.1002/ajhb.22325
17. Ruhli FJ, Henneberg M. **Are hyperostosis frontalis interna and leptin linked? A hypothetical approach about hormonal influence on human microevolution**. *Med Hypotheses* (2002.0) **58** 378-381. DOI: 10.1054/mehy.2001.1481
18. Chaljub G, Johnson RF, Johnson RF, Sitton CW. **Unusually exuberant hyperostosis frontalis interna: MRI**. *Neuroradiology* (1999.0) **41** 44-45. DOI: 10.1007/s002340050703
19. Schneeberg NG, Woolhandler G, Levine R. **The clinical significance of hyperostosis frontalis interna**. *J Clin Endocrinol* (1947.0) **7** 624-635. DOI: 10.1210/jcem-7-9-624
20. Dann S.. **Metabolic craniopathy: a review of the literature with report of a case with diabetes mellitus**. *Ann Intern Med* (1951.0) **34** 163-202. PMID: 14790546
21. Fakoya A, Heymans J, McCrary A, Rodriguez O, Cardona A, Afolabi A. **Hyperostosis Frontalis Interna: A Case Report**. *J Health Sci* (2020.0) **10** 170-172
22. Ruhli FJ, Boni T, Henneberg M. **Hyperostosis frontalis interna: archaeological evidence of possible microevolution of human sex steroid?**. *HOMO* (2004.0) **55** 91-99. PMID: 15553271
23. Ruan J, Prasad P. **The effects of skull thickness variations on human head dynamic impact responses**. *Stapp Car Crash J* (2001.0) **45** 395-414. DOI: 10.4271/2001-22-0018
24. Nikolic S, Djonić D, Zivković V, Babić D, Juković F, Djurić M. **Rate of Occurrence, Gross Appearance, and Age Relation of Hyperostosis Frontalis Interna in Females. A Prospective Autopsy Study**. *Am J Forensic Med Pathol* (2010.0) **31** 205-207. PMID: 20177366
25. Goel V, Gold B, Kapur S, Houle S. **The seats of reason? An imaging study of deductive and inductive reasoning**. *Neuroreport* (1997.0) **8** 1305-1310. DOI: 10.1097/00001756-199703240-00049
26. Brooks JO 3rd, Bearden CE, Hoblyn JC, Woodard SA, Ketter TA. **Prefrontal and paralimbic metabolic dysregulation related to sustained attention in euthymic older adults with bipolar disorder**. *Bipolar Disord* (2010.0) **12** 866-874. DOI: 10.1111/j.1399-5618.2010.00881.x
27. Notkin J.. **Frontal bone hyperostosis in psychoses; a clinical study**. *Am J Psychiatry* (1953.0) **109** 929-935. DOI: 10.1176/ajp.109.12.929
28. Gilbert T, Ait S, Delphin F, Raharisondraibe E, Bonnefoy M. **Frontal cortex dysfunction due to extensive hyperostosis frontalis interna**. *BMJ Case Rep* (2012.0) **2012**. DOI: 10.1136/bcr.07.2011.4439
|
---
title: Fear of weight gain during cognitive behavioral therapy for binge-spectrum
eating disorders
authors:
- Rachel M. Butler
- Elizabeth Lampe
- Claire Trainor
- Stephanie M. Manasse
journal: Eating and Weight Disorders
year: 2023
pmcid: PMC9988191
doi: 10.1007/s40519-023-01541-8
license: CC BY 4.0
---
# Fear of weight gain during cognitive behavioral therapy for binge-spectrum eating disorders
## Abstract
### Purpose
Fear of weight gain may play a central role in maintaining eating disorders (EDs), but research on the role of fear of weight gain during cognitive behavioral therapy (CBT-E) for binge-spectrum EDs is sparse. We examined changes in fear of weight gain during CBT-E for binge-spectrum EDs. We investigated whether fear of weight gain predicted loss of control (LOC) eating or weight change.
### Methods
Participants ($$n = 63$$) were adults of any gender recruited as part of a larger trial. Participants received 12 sessions of CBT-E, completed diagnostic assessments at pre-, mid-, and post-treatment, and completed brief surveys before sessions.
### Results
Fear of weight gain decreased across treatment, moderated by diagnosis. Those with bulimia nervosa spectrum EDs (BN-spectrum), compared to binge eating disorder, reported higher fear of weight gain at baseline and experienced a larger decrease in fear across treatment. Those reporting higher fear of weight gain at a given session experienced more frequent LOC episodes the following week. Fear of weight gain was not associated with session-by-session changes in BMI.
### Conclusion
CBT-E results in decreases in fear of weight gain, but levels remain high at post-treatment, especially for those with BN-spectrum EDs. Future interventions should consider targeting fear of weight gain as a maintaining factor for LOC episodes
### Trial registration
NCT04076553.
### Level of evidence
Level II controlled trial without randomization.
## Introduction
Eating disorders (EDs) are severe psychiatric illnesses characterized by disrupted consumption of food (e.g., restrictive or loss of control [LOC] eating) and often involve overvaluation of one’s shape and weight, a drive for thinness, and body image disturbances [1]. Specifically, bulimia nervosa (BN) and binge eating disorder (BED) involve binge eating episodes in which an individual consumes an objectively large amount of food in a relatively short amount of time and experiences a sense of loss of control [1, 2]. Those with BN also engage in compensatory behaviors (e.g., purging, laxative use, excessive exercise) to counteract the effects of a binge episode [1]. Cognitive behavioral therapy for EDs (CBT, including an enhanced version, CBT-E) is a front-line treatment for BN and BED; however, only about 30–$50\%$ of individuals achieve remission of ED symptoms following treatment [3, 4]. Considering that nearly half of patients have suboptimal treatment outcomes from the gold-standard treatment, further investigation of maintenance factors in EDs, and the effects of CBT-E on these specific factors is warranted.
A clearer understanding of maintaining factors in binge-spectrum EDs would allow for development of more targeted, specific interventions. Fear-based maintaining factors in EDs, such as fear of weight gain or “fatness” and its consequences,1 are potential intervention points. Fear of weight gain is a central symptom for many individuals with EDs [5] and has been found to predict worse ED pathology in a sample with anorexia nervosa [6]. In fact, fear of weight gain is theorized to be a core ED symptom transdiagnostically, including for those with BED [7, 8]. Additionally, in a sample seeking weight loss, those with BED had significantly higher fear of weight gain than those without BED, regardless of weight status [9].
Fear of weight gain may originate from an overvaluation of one’s shape and weight—a core symptom across binge-spectrum EDs [5, 8]. As a person prioritizes having an “ideal” shape or weight over other aspects of their identity, fear of weight gain may motivate disordered eating behaviors [10]. Further, learning theory suggests that fears may arise in EDs as unwanted outcomes (e.g., social rejection) are paired with a stimulus (e.g., weight gain), thus reinforcing the learned relationship between weight gain and fear [11, 12]. Those with EDs fear negative outcomes of weight gain, including social consequences (e.g., being rejected), personal consequences (e.g., being “lazy”, losing control over life), physical sensations, and social eating [13, 14].
In addition to being a central symptom for EDs, fear of weight gain may also be motivating disordered eating behaviors through a cycle of avoidance, as is seen in anxiety disorders [10]. Individuals with a fear of weight gain tend to avoid foods they believe will lead to weight gain [12, 15], develop ritualized eating behaviors, and engage in body checking or avoidance in an attempt to minimize risk of weight gain. As in other anxiety-based disorders, these avoidance behaviors actually strengthen fears and perpetuate the fear and avoidance cycle [16]. For example, restrained eating patterns and food and body avoidance contribute to binge eating episodes in binge-spectrum EDs. Additionally, for those with BN, compensatory behaviors (e.g., purging, excessive exercise, laxatives) serve as an effort to reduce fear of weight gain following binge episodes and maintain the binge/purge cycle [2, 10]. As individuals experience fear about changes in their weight, they are driven to engage in behaviors to avoid weight gain and associate these avoidance techniques with weight stability—they fear that if they were to abandon those behaviors, weight gain would occur and, with it, feared social and personal consequences [10, 17]. Although many of these behaviors occur to varying degrees for those with BED (e.g., body avoidance, dysregulated and restrained eating patterns), further exploration of the relevance of fear of weight gain in maintaining binge eating for those with BED is needed.
Theory supports conceptualization of EDs using an anxiety-based model, but little research exists on the temporal relationship between fear and disordered eating behaviors. For example, fear of weight gain may lead to restrictive eating, thus resulting in LOC episodes. Similarly, fear of weight gain may increase attention to threat (e.g., foods high in fat content), leading to an attempt to avoid these foods followed by a binge episode containing these foods. A better understanding of this link would clarify whether targeting fear of weight gain directly in treatment should result in decreases in disordered eating behaviors. Additionally, many patients with EDs place value on fear of weight gain, as they believe it will lead to achievement of the weight loss goal and/or protect them from the feared outcome of weight gain [18]. Individuals may believe that a reduction in fear would result in subsequent weight gain, and they are disinclined to challenge its validity. A better understanding of the relationship between fear of weight gain and actual weight changes over the course of treatment may help clinicians garner buy-in for confronting fear of weight gain in treatment.
CBT-E does not specifically or explicitly target fear of weight gain. Although CBT-E is not an exposure-based treatment—such as the first-line interventions for targeting fears in other populations (e.g., anxiety, OCD)—there are a number of components to CBT-E that may inherently require confrontation of fears [19]. CBT-E involves regularizing eating patterns, which may force some to confront fears that eating differently will result in weight gain. A complement to regular eating is weekly open weighing, which allows the individual to gain disconfirming evidence that, in fact, feared weight gain as a result of changing eating patterns does not occur [20]. Further along in CBT-E, individuals engage in dietary rule breaking experiments, which allow them to break rules surrounding eating and observe outcomes. These processes violate expectancies about feared outcomes and likely assist in decreasing fear of weight gain. In a study of individuals with anorexia nervosa, decreases in fear of weight gain occurred during CBT and were predictive of improvements in dietary restraint [21]. Research has yet to examine whether CBT-E results in changes in fear of weight gain for adults with BN or BED. Additionally, given that those with BN engage in significantly more behaviors aimed at avoiding feared weight gain (e.g., purging, excessive exercise, laxative use), it will be important to understand whether those with BN experience differential changes in fear of weight gain compared to those with BED across treatment.
The current study sought to clarify whether CBT-E produces changes in fear of weight gain and to better understand whether fear of weight gain is predictive of disordered eating behaviors and treatment outcomes during 12 sessions of modified CBT-E for binge-spectrum EDs. Modified CBT-E included self-monitoring, regular eating, overvaluation of weight/shape, reducing restriction/restraint, and addressing mood-related changes to eating. We predicted that [1] fear of weight gain would decrease over the course of treatment given the inherent confrontation of fears involved in CBT-E, [2] changes in fear of weight gain over treatment would be moderated by diagnosis (i.e., BN-spectrum, BED), [3] within-person increases in fear of weight gain would be associated with higher frequencies of LOC eating episodes in the following week, and [4] between-person, those with a higher fear of weight gain would experience a higher frequency of LOC episodes. Additionally, many individuals with EDs hold the metacognitive belief that their fear of weight gain is critical to evading actual weight gain. To understand whether fear of weight gain was actually associated with weight change, we explored the bidirectional relationship between fear of weight gain and BMI within- and between-person across treatment.
## Participants
We recruited adults with clinically significant binge-spectrum EDs ($$n = 63$$), including BN-spectrum and BED, from the community for participation in a parent trial of CBT-E augmented by inhibitory control training (clinicaltrials.gov identifier: NCT04076553). Participants were included in the parent trial if they were between 18 and 55 years old, experienced an average of at least one objective binge eating episode per week over the previous 12 weeks, had stable psychiatric medication for the past 3 months (if applicable), had a reliable Internet connection, and were located in the USA and willing and able to participate in remote intervention and assessments. Participants were excluded if they were not fluent in English, were below a BMI of 18.5, were planning to begin (in the next 6 months) or currently participating in another weight loss treatment or psychotherapy for binge eating and/or weight loss. Participants were not eligible if they had a diagnosis of autism spectrum disorder or intellectual disability, were currently experiencing other severe psychopathology that would limit their ability to engage in the treatment program (e.g., severe depression, substance dependence, active psychotic disorder), or demonstrated high levels of inhibitory control (and thus would not benefit from the inhibitory control training portion of the treatment). Participants were also excluded from the parent trial if at least half of their binge episodes were composed nearly entirely of fruit/vegetables (i.e., $80\%$ or more of the total food consumed during binges were raw fruits and vegetables) because of the intention to test inhibitory control training toward more traditional binge foods (e.g., pizza, ice cream). Participants were not eligible if they had experienced a recent head trauma, neurological condition, or brain condition that would interfere with completion of daily computer trainings.
The current study represents a secondary analysis of data from the parent trial. For the current study’s analyses, participants were included if they completed at least one session of treatment. Of these 63 individuals, 11 ($17.5\%$ of sample) dropped out of treatment prior to completing all 12 sessions. Attrition rates in our study are comparable to rates in other trials of CBT for EDs [3]. Participants completed 10.48 treatment sessions on average (SD = 3.51). In the current sample, 29 participants were randomized to the “sham” training condition, and 34 to the inhibitory control training condition. Table 1 depicts participant demographic information, diagnoses, and BMI.Table 1Participant demographics and baseline characteristics by diagnosis ($$n = 63$$)*Bulimia nervosa* (incl. subthreshold; $$n = 27$$)Binge eating disorder ($$n = 36$$)Mean or nSD or %Mean or nSD or %Age37.212.742.99.6Gender Male$13.7\%$616.7 Female$2696.3\%$3083.3Race White$2177.8\%$$3186.1\%$ African American$27.4\%$$411.1\%$ Asian$27.4\%$$00\%$ Multiracial$13.7\%$$12.8\%$ Unknown/prefer not to say$13.7\%$$00\%$Ethnicity Hispanic/Latinx$414.8\%$$38.3\%$ Non-Hispanic$2385.2\%$$3391.7\%$Disordered eating behaviors Binge episodes past 3 months91.458.991.348.6 Compensatory behaviors 3 months71.777.30.51.4BMI31.38.336.212.6Fear of weight gain GFFS 10-item scores30.95.723.87.4BMI body mass index, GFFS Goldfarb Fear of Fat Scale
## Recruitment and assessments
Participants were recruited to participate in a larger randomized controlled trial (RCT) of inhibitory control training adjunct to CBT-E for binge-spectrum EDs. The current sample includes all participants recruited for the RCT. Recruitment methods included radio and social media advertising. Interested individuals completed a phone screen to assess initial eligibility before being invited for a baseline assessment to determine final eligibility. Assessments were conducted by independent trained evaluators at pre-treatment, after session 4 (“mid-treatment”), after session 12 (“post-treatment”), and at 3-month follow-up. The current study utilizes data from pre-treatment, mid-treatment, and post-treatment assessments. Participants also completed short online surveys before each CBT-E session. Participants received free treatment and were compensated $100 for completing all assessments and up to $100 for inhibitory control trainings. Procedures were approved by the Institutional Review Board and informed consent was obtained from all participants.
## Treatment
All participants received modified 12-session CBT-E. We modified the original CBT-E manual (focused version; the default version focused exclusively on eating disorder psychopathology) based on Fairburn [19] to be delivered in 12 weekly individual sessions. Participants completed a 120-min intake (Session 1), and all following sessions were 60 min. CBT-E consisted of self-monitoring, regular eating (focused on reduction of dietary restraint), urge management strategies, discussion of overvaluation of weight and shape, dietary rule breaking experiments, and addressing event and mood-related triggers to binge eating. Elements of the complex broad version of CBT-E such as reducing perfectionism and improving low self-esteem were omitted due to the time constraints of only 12 sessions. Otherwise, there was close correspondence between Fairburn’s focused CBT-E and the modified version delivered in this trial, it was simply compressed into fewer sessions. Participants were given homework (e.g., self-monitoring, regular eating goals, reducing shape checking, etc.) between sessions. Each session consisted of a review of homework, in-session weighing, content addressing one of the above treatment targets (e.g., overvaluation of weight/shape), and assignment of homework. The treatment was delivered by graduate student clinicians who were supervised weekly by licensed clinical psychologists. Due to the COVID-19 pandemic, CBT-E sessions were conducted in-person weekly prior to March 2020 and remotely via videoconference after March 2020.
Participants were also randomized to complete a 10-min computerized inhibitory control training or a sham training daily for the first four weeks, then once weekly for the duration of the treatment. These trainings were an adjunct to CBT-E and did not impact the CBT-E intervention. The inhibitory control training aimed to increase inhibitory control toward food items via a Go/No Go Task in which participants were presented visual food or non-food stimuli and instructed to respond as quickly as possible except when a “no go signal” (e.g., a blue circle) appeared, which was always paired with stimuli representing the participant’s self-reported binge foods [22]. The sham condition contained the same stimuli and instructions, but there were no “no go” signals (i.e., participants responded as quickly as possible to every stimuli), which was meant to serve as an attention control for the inhibitory control training. Participants completed these trainings online at home on their personal computers.
## Eating pathology
The Eating Disorders Examination 17.0 (EDE) [23] was used to assess disordered eating symptoms over the previous 3 months. The EDE is a well-validated, semi-structured diagnostic interview. Based on EDE interviews, individuals were assigned a diagnosis of BN, “low-frequency BN” (i.e., other specified feeding and eating disorder), or BED using behavioral criteria based on DSM-5 frequencies for diagnoses: BN ($$n = 22$$) was defined by 12 or more objective binge episodes and 12 or more compensatory behaviors in the past 3 months, low-frequency BN ($$n = 5$$) was defined by 12 or more objective binge episodes and between 6 and 11 compensatory behaviors in the past 3 months, and BED ($$n = 36$$) was defined as having had at least 12 objective binge episodes and fewer than 6 compensatory behaviors in the past 3 months. Research has demonstrated that there is limited clinical utility in the distinction between sub- and full-threshold BN Johnson et al., [ 24]; thus, for our analyses, we merged individuals with BN and “low-frequency BN” into one group: BN-spectrum.
## Fear of weight gain
The Goldfarb Fear of Fat Scale (GFFS; [25] is a 10-item measure that was used to assess fear of weight gain or “fatness” at pre-treatment, mid-treatment, and post-treatment assessments. Items are rated on a 4-point scale from 1 (“very untrue”) to 4 (“very true”), with scores ranging from 10 to 40 on the full scale. To reduce participant burden, we selected four items from the GFFS to administer prior to each therapy session, with possible scores ranging from 4 to 16. These items were: “My biggest fear is becoming fat,” “I am afraid to gain even a little weight,” “Becoming fat would be the worst thing that could happen to me” and “If I eat even a little, I may lose control and not stop eating”. The GFFS is a reliable and well-validated measure of fear of weight gain among individuals with EDs and has been shown to differentiate weight-related fears in clinical and non-clinical samples [25]. Internal consistency of the ten-item measure in the current sample at pre-treatment was good (α = 0.88). The four-item measure demonstrated acceptable internal consistency at session 1 (α = 0.75).
## Loss of control episodes
Prior to each session, participants were asked to report frequency of binge eating or LOC episodes over the past 7 days. Participants were asked “How many times have you felt a sense of loss of control over your eating?” The items did not distinguish between objective and subjective binge episodes, and thus the weekly measure of LOC episodes likely included a range of sizes of LOC episodes.
## Body mass index (BMI)
Session-level BMI was calculated using participants’ self-reported height (given at baseline) and a weight obtained (in pounds). Weight was collected by the therapist during in-person sessions and was reported by the participant during sessions that occurred via videoconference. Baseline BMI was obtained using participant-reported height and a weight obtained (in pounds) by the assessor at the pre-treatment assessment.
## Data analytic plan
Data were analyzed using SPSS Version 26 (IBM [26] and R [27]. We ran linear models using multilevel modeling (MLM) due to the nested nature of our longitudinal data (observations within person). In all analyses, we included fixed predictor variables and the random intercept of person. We used restricted maximum likelihood estimation to handle missing data, which constituted $8.78\%$ of session data and $7.3\%$ of assessment data.
Time-varying predictors were separated into between-person effects which were grand mean centered (mean aggregate across the sessions) and within-person effects which were person-mean centered (deviation from participant’s mean at each session; raw—mean). Time (session number or assessment point) was included in each model as a predictor.
We examined whether fear of weight gain changed throughout treatment using session-by-session data. We conducted a MLM with session number (i.e., 1–12) as a fixed effect predicting four-item GFFS scores. Additionally, we examined whether fear of weight gain using the full scale (10-item GFFS) changed across the pre-treatment, mid-treatment, and post-treatment assessment points. We conducted an MLM with assessment time point (i.e., 1, 2, 3) as the fixed effect predicting GFFS scores.
We examined whether the change in fear of weight gain (4-item GFFS) session by session was moderated by diagnosis of BN-spectrum compared to BED. We conducted an MLM with session (i.e., 1–12), diagnosis (BN-spectrum, BED), and the interaction between session and diagnosis as fixed effects predicting session GFFS scores. We also examined whether the change in fear of weight gain (10-item GFFS) during treatment was moderated by diagnosis. We conducted an MLM with assessment point (i.e., 1, 2, 3), diagnosis (BN-spectrum, BED), and the interaction between assessment point and diagnosis as fixed effects predicting GFFS scores.
We conducted an MLM to examine whether GFFS scores were predictive of LOC episode frequency. We used a lagging procedure such that GFFS scores at a session “t” predicted LOC episode frequency (patient-reported) at session “t + 1”. This lagging procedure allows for testing the direction of effects, in other words, fear of weight gain at a given week predicting the following week’s LOC episode frequency. This MLM included the fixed effect of GFFS scores at session “t” predicting LOC episodes at session “t + 1”.
We conducted MLMs to investigate whether BMI predicted fear of weight gain. We lagged GFFS scores in order to test whether BMI at session “t” (fixed effect) predicted GFFS scores at session “t + 1”. Conversely, we examined whether GFFS scores at session “t” (fixed effect) predicted BMI at session “t + 1”.
As a sensitivity analysis, we included computerized training condition (i.e., sham, inhibitory control training) in all models as a control, but findings remained consistent. Thus, we reported results of analyses without controlling for training condition. Cohen’s d effect sizes were calculated by transforming t-statistics.
## Baseline associations
We examined whether demographic characteristics (age, gender, ethnicity, race, and BMI) were associated with GFFS scores at baseline. Age was not correlated with GFFS scores, r = − 0.09, $$p \leq 0.50.$$ Female participants had significantly higher GFFS scores at baseline, t [60] = 2.09, $$p \leq 0.04.$$ Those who identified as Latinx or Hispanic did not have significantly different GFFS scores than those who did not, t [60] = 0.36, $$p \leq 0.72.$$ GFFS scores were not significantly different between racial identities, F [4, 57] = 0.91, $$p \leq 0.46.$$ BMI and GFFS scores were not correlated at baseline, r = − 0.06, $$p \leq 0.67.$$ Those with BN-spectrum disorders had higher GFFS scores at baseline than those with BED, t [60] = 4.08, $p \leq 0.001$, $d = 1.05.$ Past month binge episode frequency and BMI were not significantly different between those with BN-spectrum and BED, ps > 0.09.
## Change in fear of weight gain during treatment
GFFS scores decreased significantly over the course of treatment from session 1 ($M = 11.32$ SD = 2.97) to session 12 ($M = 8.64$, SD = 3.54), B = − 0.24, SE_B = 0.02, β = − 0.07, t (591.92) = − 14.05, $p \leq 0.001$, $d = 1.16.$
GFFS scores decreased significantly across treatment at assessment points, B = − 2.47, SE_B = 0.37, β = − 0.30, t (106.56) = − 6.61, $p \leq 0.001$, $d = 0.93.$ Pairwise comparisons revealed that GFFS scores did not decrease pre-treatment ($M = 26.89$, SD = 7.53) to mid-treatment ($M = 25.52$, SD = 8.14), B = − 1.31, SE_B = 0.71, β = − 0.16, t (104.89) = − 1.84, $$p \leq 0.07$$, $d = 0.26$, but decreased significantly from pre-treatment to post-treatment ($M = 21.90$, SD = 8.31), B = − 4.99, SE_B = 0.74, β = − 0.61, t (105.33) = − 6.76, $p \leq 0.001$, $d = 0.94.$ See Table 2 for descriptive statistics for outcome variables across treatment.
**Table 2**
| Unnamed: 0 | BMI | LOC episodes past week | GFFS 10-item | GFFS 4-item |
| --- | --- | --- | --- | --- |
| | Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) |
| Assessments | Assessments | Assessments | Assessments | Assessments |
| Baseline | 34.14 (11.18) | – | 26.89 (7.53) | 10.82 (3.30) |
| Mid-treatment | 34.36 (11.69) | – | 25.52 (8.14) | 10.09 (3.37) |
| Post-treatment | 34.00 (11.78) | – | 21.90 (8.31) | 8.72 (3.41) |
| Sessions | Sessions | Sessions | Sessions | Sessions |
| 1 | – | 5.07 (4.69) | – | 11.32 (2.97) |
| 2 | – | 4.65 (3.85) | – | 11.44 (3.34) |
| 3 | 35.71 (11.95) | 2.89 (2.62) | – | 10.80 (3.56) |
| 4 | 34.61 (11.64) | 2.42 (2.20) | – | 10.37 (3.29) |
| 5 | 34.13 (11.85) | 1.74 (2.23) | – | 10.06 (3.47) |
| 6 | 34.32 (11.78) | 1.85 (2.78) | – | 10.16 (3.22) |
| 7 | 34.77 (11.98) | 1.35 (2.04) | – | 9.87 (3.44) |
| 8 | 34.44 (11.90) | 1.35 (2.02) | – | 9.78 (3.54) |
| 9 | 34.87 (12.01) | 1.10 (1.39) | – | 8.94 (3.30) |
| 10 | 34.14 (11.42) | 1.17 (1.96) | – | 9.44 (3.35) |
| 11 | 33.73 (11.99) | 0.43 (0.76) | – | 8.86 (3.56) |
| 12 | 35.73 (12.63) | 0.67 (1.02) | – | 8.64 (3.54) |
## Change in fear of weight gain moderated by diagnosis
Those with BN-spectrum tended to have higher GFFS scores, $B = 2.17$, SE_B = 0.77, β = 0.63, t (69.99) = 2.81, $p \leq 0.01$, $d = 0.39.$ Diagnosis did not moderate the change in GFFS scores over time, $B = 0.03$, SE_B = 0.03, β = 0.01, t (591.29) = 1.02, $$p \leq 0.31$$, $d = 0.14.$
Again, those with BN-spectrum tended to have higher GFFS scores, $B = 8.71$, SE_B = 2.26, β = − 1.06, t (124.14) = 3.85, $p \leq 0.001$, $d = 0.54.$ Diagnosis moderated the change in GFFS scores over time such that those with BN-spectrum had greater decreases in GFFS scores over the course of treatment than those with BED, $B = 2.07$, SE_B = 0.72, β = − 0.25, t (105.50) = 2.85, $p \leq 0.01$, $d = 0.40$ (see Fig. 1).Fig. 1Change in fear of weight gain at assessments moderated by diagnosis. GFFS = Goldfarb Fear of Fat Scale, 10-item. Possible scores range from 10 to 40
## Fear of weight gain predicting LOC eating
We found between-person effects of GFFS scores on the patient-reported LOC frequency, $B = 0.27$, SE_B = 0.07, β = 0.30, t (51.40) = 3.99, $p \leq 0.001$, $d = 0.54$, such that those who had higher fears of weight gain (compared to other participants) at a given session reported more LOC episodes at the next session. We found no within-person effect of GFFS scores on patient-reported LOC episode frequency, B = − 0.01, SE_B = 0.08, β = − 0.01, t (581.34) = − 0.07, $$p \leq 0.96$$, $d = 0.02$, indicating that higher fear of weight gain (compared to one’s own norms) was not predictive of LOC episode frequency.
## Session BMI and fear of weight gain
We found no within-person effects of BMI on GFFS scores, $B = 0.05$, SE_B = 0.07, β = 0.17, t (402.30) = 0.71, $$p \leq 0.48$$, $d = 0.10$, and no between-person effects, $B = 0.03$, SE_B = 0.03, β = 0.09, t (54.03) = 0.80, $$p \leq 0.43$$, $d = 0.14.$ These findings suggest that having a higher BMI (compared to the group mean or compared to one’s own mean) at a given session did not predict fear of weight gain at the following session.
We found no within-person effects of GFFS scores on BMI, $B = 0.02$, SE_B = 0.05, β = 0.01, t (400.06) = 0.42, $$p \leq 0.67$$, $d = 0.06.$ There were also no between-person effects of GFFS scores on BMI, $B = 0.53$, SE_B = 0.51, β = 0.16, t (52.95) = 1.04, $$p \leq 0.30$$, $d = 0.15.$ These findings suggest that having higher GFFS scores (compared to the group mean or compared to one’s own mean) did not predict BMI at the following session.
## Discussion
The current study was the first to our knowledge to examine fear of weight gain during CBT-E for the treatment of binge-spectrum EDs. Notably, we found that fear of weight gain decreased over the course of CBT-E, both in session-by-session ratings and from pre- to post-treatment assessment. These findings are encouraging, as CBT-E does not explicitly target this fear even though it is posited to be a maintaining factor of disordered eating [10]. We also found that those with BN-spectrum EDs began treatment with higher fear of weight gain and experienced greater reduction in fear across CBT-E. Our findings suggest that those with higher fear of weight gain at a given session experience more frequent LOC episodes in the following week, indicating a temporal link between this fear and LOC episodes. Finally, we noted that BMI and fear of weight gain were not associated during treatment, suggesting that changes in an individual’s weight do not correlate with fear of weight gain.
Despite decreases in fear of weight gain, fear remained somewhat elevated at the end of treatment (in particular, for those with BN-spectrum EDs) compared to scores reported in prior research with non-ED samples [28]. Exploration of which components of CBT-E contribute to decreases in fear of weight gain would allow for clinicians to emphasize certain components of treatment for individuals with particularly salient fears. Additionally, continued research into exposure-based treatments—the first-line intervention for addressing fears in other populations (e.g., anxiety, OCD)—may be a worthwhile approach for targeting fear of weight gain. These approaches include targeting food avoidance using exposures to feared foods and imaginal exposure to address the feared social and personal consequences (i.e., loss of identity, disgust, discomfort in one’s body) as a result of weight gain [29–31]. These exposure-based interventions have demonstrated marked improvements in disordered eating-related fears including fears of weight gain [30, 31]. Further investigation into exposure-based approaches to the treatment of binge-spectrum disorders is warranted if we hope to produce tangible changes in the fear of weight gain through treatment.
Both those with BN-spectrum EDs and BED started treatment with elevated levels of fear of weight gain compared to non-ED samples in prior research [28, 32]. GFFS scores for those with BN in the current sample were similar to previously published BN samples ($M = 33.2$; [28]. Individuals with BN (including low frequency) had higher fear of weight gain and experienced greater decreases in that fear over the course of treatment than those with BED. This suggests that those regularly engaging in compensatory behaviors experience higher fears of weight gain, possibly leading individuals to seek out compensatory behaviors as a method of mitigating the fear. Additionally, compensatory behaviors reinforce the fear, because individuals are prevented from learning that feared outcomes do not occur, and lack of weight change is associated with the use of the compensatory behavior [10]. CBT-E specifically assists patients in stopping compensatory behaviors [19], which may account for the finding that those with BN-spectrum EDs experienced a greater decrease in fear over the course of treatment as they relinquished use of these behaviors. Unfortunately, we were unable to effectively examine the relationship between fear of weight gain and compensatory behaviors due to the diagnostically mixed nature of our sample (i.e., high variability and a positively skewed distribution of compensatory behaviors). Future research should investigate the causal link between fear of weight gain and compensatory behaviors for those with BN.
Those with a higher fear of weight gain experienced more LOC episodes in the following week. Fear of weight gain appears to be a maintaining factor for LOC eating, even during treatment, such that individuals who are experiencing a higher level of fear are more at risk for a LOC episode shortly thereafter. This may occur through a pathway of restrictive or restrained eating. On the other hand, experiencing less intense fear of weight gain may be somewhat protective in that it puts individuals at lower risk for LOC episodes. This corroborates the theory suggesting that fears of weight gain are a core maintaining factor in EDs, and that targeting the fear may reduce symptoms [5, 10]. Alternatively, those with higher fear of weight gain may perceive more eating episodes as loss of control episodes than those who experience less intense fear of weight gain. We did not find within-person effects, which may be a result of lower within-person variability in fears of weight gain session-by-session. Altogether, our preliminary findings suggest that fear of weight gain may play a causal role in LOC eating, and future research should continue to explore this as a maintaining factor.
Interestingly, we found that BMI was not associated with fear of weight gain at pre-treatment, which adds to previous findings from a sample with anorexia nervosa [21]. Further, BMI was not predictive of fear of weight gain, nor was fear of weight gain predictive of BMI. This finding is highly clinically relevant, as many individuals with EDs believe fear of weight gain to be protective of actual weight gain, such that they are often ambivalent or reluctant to challenge the fear. The fact that BMI does not change as fear of weight gain decreases suggest that the fear is, in fact, not protective. Clinicians may consider using this information as a form of psychoeducation for patients when discussing the importance of confronting their fear. It is important to consider that our sample included individuals with binge-spectrum EDs who were not at a significantly low weight, so weight gain was typically not an explicit treatment goal. Future research should examine the association between BMI and fear of weight gain in a sample of low weight individuals who are undergoing weight restoration during CBT-E.
## Strengths and limits
Examining fear of weight gain at each session allowed us to predict the following week LOC episodes to establish a temporal association between fear of weight gain and LOC. Limitations include a somewhat narrow sample, as participants were ineligible to participate if they had high inhibitory control due to the intention to test inhibitory control training in the broader RCT. Additionally, although it would have been optimal to calculate clinically significant change in fear of weight gain, we were unable to do so given lack of norms for the GFFS and high variability in GFFS scores at pre-treatment (SD = 7.5, ranging 10–40). It will be critical to examine whether the changes in fear of weight gain during CBT-E are clinically significant, or whether additional methods of targeting fear of weight gain during treatment (i.e., exposure therapy; [17]) must be incorporated for those with more intense fears to attain significant improvements. Four items from the GFFS were administered at pre-session surveys, which were selected from a well-validated measure and intended to reduce participant burden. Additionally, prior research on fear of weight gain has tended to measure the construct using single items pulled from the EDE-Q due to lack of multi-item measures evaluating the construct (e.g., [6, 31]. Given that the four-item measure has not been validated, future research should explore methods of assessing fear of weight gain. The course of CBT-E implemented in our trial was also relatively short (12 sessions, perhaps a longer duration of treatment would have resulted in greater decreases in fear of weight gain.
CBT-E appears to implicitly target fear of weight gain during treatment of binge-spectrum EDs likely through regularizing eating patterns, reducing compensatory behaviors, and breaking dietary rules. We also observed that fear of weight gain predicted more frequent LOC episodes, suggesting that fear of weight gain is a worthy target for future interventions for BN and BED. Clinicians may find it helpful to explicitly discuss fear of weight gain with patients and to make reducing this fear a direct aim of treatment, especially for those experiencing more intense fear. Additional research is needed to understand whether fear of weight gain predicts use of compensatory behaviors in those with BN. Furthermore, future research should investigate whether an intervention explicitly designed to target fear of weight gain (e.g., exposure therapy) would produce even greater changes in fear than CBT-E, and whether reductions in fear result in decreases in LOC episodes and compensatory behaviors. Finally, longitudinal studies must examine whether decreases in fear of weight gain during treatment have an effect on long-term remission from EDs.
## What is already known on this subject?
Research has established that fear of weight gain is a core symptom in eating disorders, including for those with bulimia nervosa and binge eating disorder [5]. Little is known about whether the current gold-standard treatment for eating disorders, CBT-E, targets fear of weight gain. Additionally, research has yet to examine whether fear of weight gain is directly associated with binge episodes.
## What does this study add?
Our study demonstrates that CBT-E does result in decreases in fear of weight gain across treatment. Those with higher fear of weight gain experience more LOC episodes in the following week, suggesting there may be a causal link between fear of weight gain and LOC eating.
## References
1. 1.American Psychiatric AssociationDiagnostic and statistical manual of mental disorders20135WashingtonAmerican Psychiatric Association. *Diagnostic and statistical manual of mental disorders* (2013.0)
2. Fairburn CG, Marcus MD, Wilson GT, Fairburn CG, Wilson GT. **Cognitive-behavioral therapy for binge eating and bulimia nervosa: a comprehensive treatment manual**. *Binge eating: nature, assessment, and treatment* (1993.0) 361-404
3. Atwood ME, Friedman A. **A systematic review of enhanced cognitive behavioral therapy (CBT-E) for eating disorders**. *Int J Eat Disord* (2020.0) **53** 311-330. DOI: 10.1002/eat.23206
4. Hay P. **A systematic review of evidence for psychological treatments in eating disorders: 2005–2012**. *Int J Eat Disord* (2013.0) **46** 462-469. DOI: 10.1002/eat.22103
5. Levinson CA, Williams BM. **Eating disorder fear networks: Identification of central eating disorder fears**. *Int J Eat Disord* (2020.0) **53** 1960-1973. DOI: 10.1002/eat.23382
6. Linardon J, Phillipou A, Castle D, Newton R, Harrison P, Cistullo LL, Brennan L. **The relative associations of shape and weight over-evaluation, preoccupation, dissatisfaction, and fear of weight gain with measures of psychopathology: an extension study in individuals with anorexia nervosa**. *Eat Behav* (2018.0) **29** 54-58. DOI: 10.1016/j.eatbeh.2018.03.002
7. Goldschmidt AB, Crosby RD, Cao L, Moessner M, Forbush KT, Accurso EC, Le Grange D. **Network analysis of pediatric eating disorder symptoms in a treatment-seeking, transdiagnostic sample**. *J Abnorm Psychol* (2018.0) **127** 251-264. DOI: 10.1037/abn0000327
8. Wang SB, Jones PJ, Dreier M, Elliott H, Grilo CM. **Core psychopathology of treatment-seeking patients with binge-eating disorder: a network analysis investigation**. *Psychol Med* (2019.0) **49** 1923-1928. DOI: 10.1017/S0033291718002702
9. Bullock AJ, Barber J, Barnes RD. **Characterizing fear of weight gain and sensitivity to weight gain in individuals seeking weight loss treatment**. *Eat Weight Dis Stud Anorexia Bulimia Obes* (2021.0) **26** 385-393. DOI: 10.1007/s40519-020-00862-2
10. Schaumberg K, Reilly EE, Gorrell S, Levinson CA, Farrell NR, Brown TA, Anderson LM. **Conceptualizing eating disorder psychopathology using an anxiety disorders framework: evidence and implications for exposure-based clinical research**. *Clin Psychol Rev* (2021.0) **83** 101952. DOI: 10.1016/j.cpr.2020.101952
11. Murray SB, Treanor M, Liao B, Loeb KL, Griffiths S, Le Grange D. **Extinction theory & anorexia nervosa: deepening therapeutic mechanisms**. *Behav Res Ther* (2016.0) **87** 1-10. DOI: 10.1016/j.brat.2016.08.017
12. Strober M. **Pathologic fear conditioning and anorexia nervosa: On the search for novel paradigms**. *Int J Eat Disord* (2004.0) **35** 504-508. DOI: 10.1002/eat.20029
13. Brown ML, Levinson CA. **Core eating disorder fears: prevalence and differences in eating disorder fears across eating disorder diagnoses**. *Int J Eat Disord* (2022.0) **55** 956-965. DOI: 10.1002/eat.23728
14. Levinson CA, Vanzhula IA, Christian C. **Development and validation of the eating disorder fear questionnaire and interview: preliminary investigation of eating disorder fears**. *EaT Behav* (2019.0) **35** 101320. DOI: 10.1016/j.eatbeh.2019.101320
15. Steinglass JE, Eisen JL, Attia E, Mayer L, Walsh BT. **Is anorexia nervosa a delusional disorder? An assessment of eating beliefs in anorexia nervosa**. *J Psychiatr Pract* (2007.0) **13** 65-71. DOI: 10.1097/01.pra.0000265762.79753.88
16. Christian C, Levinson CA. **An integrated review of fear and avoidance learning in anxiety disorders and application to eating disorders**. *New Ideas Psychol* (2022.0) **67** 100964. DOI: 10.1016/j.newideapsych.2022.100964
17. Reilly EE, Anderson LM, Gorrell S, Schaumberg K, Anderson DA. **Expanding exposure-based interventions for eating disorders**. *Int J Eat Disord* (2017.0) **50** 1137-1141. DOI: 10.1002/eat.22761
18. Essayli JH, Vitousek KM. **Cognitive Behavioral therapy with eating disordered youth**. *Cognitive behavioral therapy in youth: tradition and innovation* (2020.0) 163-187
19. Fairburn CG. *Cognitive behavior therapy and eating disorders* (2008.0)
20. Waller G, Mountford VA. **Weighing patients within cognitive-behavioural therapy for eating disorders: how, when and why**. *Behav Res Ther* (2015.0) **70** 1-10. DOI: 10.1016/j.brat.2015.04.004
21. Calugi S, El Ghoch M, Conti M, Dalle Grave R. **Preoccupation with shape or weight, fear of weight gain, feeling fat and treatment outcomes in patients with anorexia nervosa: a longitudinal study**. *Behav Res Ther* (2018.0) **105** 63-68. DOI: 10.1016/j.brat.2018.04.001
22. Manasse SM, Lampe EW, Gillikin L, Payne-Reichert A, Zhang F, Juarascio AS, Forman EM. **The project REBOOT protocol: evaluating a personalized inhibitory control training as an adjunct to cognitive behavioral therapy for bulimia nervosa and binge-eating disorder**. *Int J Eat Disord* (2020.0) **53** 1007-1013. DOI: 10.1002/eat.23225
23. 23.Fairburn CG, Cooper Z, O’Connor M. (2008). Eating
Disorder Examination (16.0D). In Fairburn CG. Cognitive Behavior Therapy and Eating Disorders. Guilford Press, New York.
24. 24.Johnson SN, Forbush KT, Swanson TJ, Christensen KA (2021) An empirical evaluation of the
diagnostic threshold between full-threshold and sub-threshold bulimia nervosa. Eat Behav 42:101540 10.1016/j.eatbeh.2021.101540
25. Goldfarb LA, Dykens EM, Gerrard M. **The Goldfarb fear of fat scale**. *J Pers Assess* (1985.0) **49** 329-332. DOI: 10.1207/s15327752jpa4903_21
26. Corporation IBM. *IBM SPSS statistics for windows version 26.0* (2019.0)
27. 27.R Core Team (2013). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Retrieved Sep 1, 2022. URL http://www.R-project.org/.
28. Chernyak Y, Lowe MR. **Motivations for dieting: drive for thinness is different from drive for objective thinness**. *J Abnorm Psychol* (2010.0) **119** 276-281. DOI: 10.1037/a0018398
29. Butler RM, Heimberg RG. **Exposure therapy for eating disorders: a systematic review**. *Clin Psychol Rev* (2020.0) **78** 101851. DOI: 10.1016/j.cpr.2020.101851
30. Levinson CA, Christian C, Ram SS, Vanzhula I, Brosof LC, Michelson LP, Williams BM. **Eating disorder symptoms and core eating disorder fears decrease during online imaginal exposure therapy for eating disorders**. *J Affect Disord* (2020.0) **276** 585-591. DOI: 10.1016/j.jad.2020.07.075
31. Levinson CA, Williams BM, Christian C. **What are the emotions underlying feeling fat and fear of weight gain?**. *J Affect Disord* (2020.0) **277** 146-152. DOI: 10.1016/j.jad.2020.08.012
32. McKenzie SJ, Williamson DA, Cubic BA. **Stable and reactive body image disturbances in bulimia nervosa**. *Behav Ther* (1993.0) **24** 195-207. DOI: 10.1016/S0005-7894(05)80263-1
|
---
title: Comparison of 30-day planned and unplanned readmissions in a tertiary teaching
hospital in China
authors:
- Mengjiao Zhang
- Siru Liu
- Yongdong Bi
- Jialin Liu
journal: BMC Health Services Research
year: 2023
pmcid: PMC9988192
doi: 10.1186/s12913-023-09193-1
license: CC BY 4.0
---
# Comparison of 30-day planned and unplanned readmissions in a tertiary teaching hospital in China
## Abstract
### Purpose
The purpose of this study was to analyze and compare the clinical characteristics of patients with 30-day planned and unplanned readmissions and to identify patients at high risk for unplanned readmissions. This will facilitate a better understanding of these readmissions and improve and optimize resource utilization for this patient population.
### Methods
A retrospective cohort descriptive study was conducted at the West China Hospital (WCH), Sichuan University from January 1, 2015, to December 31, 2020. Discharged patients (≥ 18 years old) were divided into unplanned readmission and planned readmission groups according to 30-day readmission status. Demographic and related information was collected for each patient. Logistic regression analysis was used to assess the association between unplanned patient characteristics and the risk of readmission.
### Results
We identified 1,118,437 patients from 1,242,496 discharged patients, including 74,494 ($6.7\%$) 30-day planned readmissions and 9,895 ($0.9\%$) unplanned readmissions. The most common diseases of planned readmissions were antineoplastic chemotherapy (62,$\frac{756}{177}$,749; $35.3\%$), radiotherapy sessions for malignancy ($\frac{919}{8}$,229; $11.2\%$), and systemic lupus erythematosus ($\frac{607}{4}$,620; $13.1\%$). The most common diseases of unplanned readmissions were antineoplastic chemotherapy ($\frac{2038}{177}$,747; $1.1\%$), age-related cataract ($\frac{1061}{21}$,255; $5.0\%$), and unspecified disorder of refraction ($\frac{544}{5}$,134; $10.6\%$). There were statistically significant differences between planned and unplanned readmissions in terms of patient sex, marital status, age, length of initial stay, the time between discharge, ICU stay, surgery, and health insurance.
### Conclusion
Accurate information on 30-day planned and unplanned readmissions facilitates effective planning of healthcare resource allocation. Identifying risk factors for 30-day unplanned readmissions can help develop interventions to reduce readmission rates.
## Introduction
Hospital readmission is a serious, common, and costly adverse patient outcome. Unplanned readmission is not only an indicator of the critical quality of care for patients but also a significant factor in rising healthcare costs [1, 2]. Readmissions account for billions of dollars in annual Medicare expenditures [3]. There is growing recognition that readmission is an outcome measure for quality of healthcare, cost reduction, and transitions of care [4].
Unplanned readmission within 30 days of discharge is an important indicator of the cost and quality of healthcare service and is strongly related to clinical and sociodemographic characteristics [5, 6]. Reducing readmissions is a priority for hospitals and clinicians to improve the quality of healthcare and reduce costs. To address readmissions, the Centers for Medicare and Medicaid Services developed the Hospital Readmissions Reduction Program (HRRP) in 2010 and implemented it in 2012 [7, 8]. Its purpose is to encourage hospitals to improve the quality and transition of healthcare to a better plan of discharge, thereby effectively reducing avoidable 30-day readmissions [8]. The first important step in reducing readmissions is to determine the incidence, risk factors, and causes of readmission. This information can help identify patients at high risk of readmission and target interventions to reduce avoidable readmissions.
Research on the relationship between healthcare quality and readmission needs to distinguish between planned and unplanned readmissions, as only unplanned readmissions can reflect the healthcare quality at the first hospitalization. Planned readmissions may be associated with the utilization of hospital resources (multiple admissions for reimbursement purposes or therapeutic procedures), but not with the healthcare quality process [9]. One study showed that patients with hematological and oncological diseases, renal disease, heart failure, and chronic obstructive pulmonary disease had the highest odds of unplanned readmission across all age groups [10]. However, another study found no risk factors for readmission, except that readmitted patients were significantly older than those who were not readmitted [11]. Despite a large number of readmission studies, it is unclear whether planned and unplanned 30-day readmissions differ between hospitals. The reasons for unplanned readmissions are also not fully elucidated [10, 11]. The relative contribution of patient-level risk factors and structural hospital characteristics to the variation in unplanned readmissions is not fully understood. As reported by the OECD, identifying truly unplanned readmissions is complex [12].
This study aimed to describe the incidence of planned and unplanned 30-day readmissions and to investigate the incidence of time. We sought to analyze the characteristics of readmitted patients and to identify risk factors associated with unplanned readmission. This will provide a basis for improving the quality of healthcare and optimizing the discharge process.
## Patients and setting
A retrospective descriptive cohort study was conducted at West China Hospital (WCH), Sichuan University. WCH is a 4300-bed teaching hospital in Sichuan, one of the best and largest hospitals in China. In 2021, more than 7.75 million patients visited the outpatient and emergency departments, 279,000 patients were discharged from inpatient departments, and more than 196,000 operations were performed (http://www.wchscu.cn/Home.html). The study cohort includes patients discharged between 1 and 2015 and 31 December 2020. Only patients aged 18 years or older at the time of the index admission were included. All patient data were obtained from the hospital’s electronic health record (EHR).
## Study variables
In this study, we used six years (2015–2020) of discharge EHR with relevant information on patient characteristics and hospital admissions (e.g., date of admission and discharge, principal diagnosis). This included information on patient demographic characteristics (e.g., age, sex, marital status, type of health insurance) and clinical characteristics (e.g., length of stay, surgery or not, ICU stay or not). A patient was readmitted if a new admission occurred within 30 days of the first discharge and was related to the index admission. Day patients and outpatients were excluded. Transfers between units within the same hospital and between hospitals were not considered as readmissions.
The patient ID was used to identify all patients who were readmitted within 30 days. These patients created a 30-day readmission group and a 30-day non-readmission group. Planned versus unplanned readmissions were identified by revisiting the medical records of all patients readmitted within 30 days. Planned readmission was defined as an intentionally planned readmission during the index admission, and patients without a planned readmission were defined as unplanned readmissions. Diagnosis was determined using the International Classification of Diseases, 10th Revision (ICD-10). Patient demographic and clinical characteristics were obtained from the EHR.
The main outcome measure of this research was unplanned and planned 30-day readmission to the hospital. Second, the risk factors for unplanned 30-day readmission were analyzed.
## Statistical analysis
All variables were reported before analysis using frequencies and percentages or means, medians, and standard deviations. The distributions of continuous variables were assessed using histograms. Univariate analysis and bivariate logistic regression analyses were performed for unplanned readmission within 30 days. For univariate analysis, we used the Student t-test for continuous variables and the chi-square test for categorical variables. We performed all statistical analyses using IBM SPSS Statistics 20. Statistical significance was determined by $p \leq 0.05.$
## Ethics statement
The study was approved by the Bioethics Committee of the West China Hospital Sichuan University (2022 − 174). Only information that was routinely collected during hospitalization was used. We used anonymous electronic medical records, so we did not seek written consent from participants.
## Results
In January 2015 and December 2020, data on 1,242,496 discharged patients were available for analysis. We excluded 124,059 patients due to death [7,496] and patients aged < 18 years [116,563], leaving a total of 1,118,437 patients for the analysis dataset. Of the 1,118,437 patients, 84,389 were readmitted within 30 days of discharge. This included 74,494 ($6.7\%$) planned readmissions and 9,895 ($0.9\%$) unplanned readmissions. There were significantly more women with planned readmissions than unplanned readmissions, $67.0\%$ and $51.3\%$ respectively. The age groups with the highest number of planned and unplanned readmissions were 40≤-<49 years (21,653; $29.1\%$) and 50≤-<59 years (2,074; $21.0\%$). The number of days of initial hospitalization for patients with planned and unplanned readmissions was predominantly in the 1≤-<4-day group, $67.0\%$ [49,936] and $46.5\%$ [4,600], respectively (Table 1). The age group with the lowest number of patients with both planned and unplanned readmissions was ≥ 80 years ($0.9\%$ vs. $4.2\%$). Table 1 shows the demographics and associated risk factors for planned and unplanned readmissions.
Bivariate logistic regression was used to identify variables independently associated with an increased risk of 30-day unplanned readmission. In a bivariate logistic regression model, factors significantly associated with 30-day unplanned readmission were age, LOS, sex (male), marital status (separated/divorced, single, widowed/other), ICU stay, and surgery (Table 2).
Table 1Characteristics of patient readmissionsPlanned readmissionn = 74494Unplanned readmissionn = 9895x2/tp Gender female49,938($67.0\%$)5,073($51.3\%$)956.98P<0.001 male24,556($33.0\%$)4,822($48.7\%$) *Marital status* Married/partner68,879($92.4\%$)8,236($83.2\%$) Separated/divorced1,395($1.9\%$)224($2.3\%$)1,117.38P<0.001 single2,959($4.0\%$)1,055($10.7\%$) Widowed/other1,261($1.7\%$)380($3.8\%$)Age years 18≤-<303,513($0.5\%$)1,243($12.5\%$) 30≤-<397,631($10.2\%$)1,213($12.3\%$) 40≤-<4921,653($29.1\%$)1,929($19.5\%$) 50≤ -<5921,346($28.7\%$)2,074($21.0\%$)2584.87P<0.001 60≤-<6915,463($20.8\%$)1,941($19.6\%$)5 70≤ -<794,213($5.7\%$)1,079($10.9\%$) ≥ 80675($0.9\%$)416($4.2\%$)Mean(SD)51.59(12.41)51.71(16.83)-0.64p<0.52LOS (days) 1≤-<449,936($67.0\%$)4,600($46.5\%$) 4≤-<711,397($15.3\%$)1,374($13.9\%$) 7≤-<149,560($12.8\%$)2,173($22.0\%$)3382.70P<0.001 14≤-<201,612($2.2\%$)737($7.4\%$)8 20≤-<301,297($1.7\%$)644($6.5\%$) ≥ 30692($0.9\%$)367($3.7\%$)Mean(SD)4.09(6.78)8.04(12.19)31.57P<0.001IDR (days) 1 ≤-<31,262($1.7\%$)1,093($11.0\%$) 3 ≤-<5936($1.3\%$)364($3.7\%$) 5≤<73,485($4.7\%$)1,165($11.8\%$) 7 ≤ <1010,159($13.6\%$)790($8.0\%$4285.42P<0.001 10≤ <2016,980($22.8\%$)1,964($19.8\%$)5 20≤ <3041,672($55.9\%$)4,519($45.7\%$)Mean(SD)18.54(7.96)16.18(10.01)22.54P<0.001No stay in ICU74,394($99.9\%$)9,704($98.1\%$)819.92P<0.001Stay in ICU100($0.1\%$)191($1.9\%$)Surgery No47,694($64.0\%$)4,131($41.7\%$)1,828.96P<0.001 Yes26,800($36.0\%$)5,764($58.3\%$)Health insurance Yes67,765($91.1\%$)8,395($84.8\%$)372.67P<0.001 No2322($3.1\%$)525($5.3\%$) Missing4407($5.8\%$)975($9.9\%$)LOS: Length of initial stay; IDR: Interval from discharge to readmission Table 2Bivariate logistic regression analysis of risk factors for 30-day unplanned readmissionBS.E.P valueExp(B)$95\%$ C.I.Age years0.0010.0010.4231.0010.9991.002LOS (days)0.05<0.001<0.0011.051.051.05IDR (days)-0.03<0.001<0.0010.970.960.97Gender0.660.02<0.0011.931.852.02Marital<0.001Separated/divorced0.300.07<0.0011.341.161.55single1.090.04<0.0012.982.773.21Widowed/other0.920.06<0.0012.522.242.83ICU2.680.12<0.00114.6411.4918.67Surgery0.910.02<0.0012.482.382.59Health insurance<0.001No-0.580.04<0.0010.560.520.60Missing0.020.060.721.020.911.15
## Comparison of diseases with planned and unplanned readmissions
The most common disease for planned readmissions was antineoplastic chemotherapy (62,$\frac{756}{177}$,749; $35.3\%$), followed by radiotherapy sessions for malignancy ($\frac{919}{8}$,229; $11.2\%$), and systemic lupus erythematosus ($\frac{607}{4}$,620; $13.1\%$). The most common diseases of unplanned readmission was antineoplastic chemotherapy ($\frac{2038}{177}$,747; $1.1\%$), followed by age-related cataract ($\frac{1061}{21}$,255; $5.0\%$), and unspecified disorder of refraction ($\frac{544}{5}$,134; $10.6\%$) (Table 3).
Table 3Disease readmission numbers and ratesPlanned readmissionUnplanned readmissionDiseaseNDischarged(%)DiseaseNDischarged(%)Encounter for antineoplastic chemotherapy62,756177,74935.3Encounter for antineoplastic chemotherapy2,038177,7491.1Z51.11Z51.11Radiotherapy session9198,22911.2Age-related cataract1,06121,2555.0Z51.0H25.900Systemic lupus erythematosus6074,62013.1Unspecified disorder of refraction5445,13410.6M32.0H52.7Systemic lupus erythematosus with:kidney involvement2961,58518.7Radiotherapy session32328,1081.1M32.1 + N085Z51.0Multiple myeloma29061647.1Systemic lupus erythematosus1884,6204.1C90.0M32.0Unspecified disorder of refraction2145,1344.2Iridocyclitis697,4920.9H52.701H20.9Age-related cataract21321,2551.0Chronic obstructive pulmonary disease with acute exacerbation674,4571.5H25.9J44.1Malignant neoplasm of bronchus and lung2027,7632.6Systemic lupus erythematosus with:kidney involvement641,5854.0C34M32.1 + N085Encounter for other specified aftercare1112,7734.0Hemiplegia, unspecified512,0632.5Z51.89G81.9Malignant neoplasm of rectum894,5881.9Acute pancreatitis479,3500.5C20.0K85.9
## Comparison of departments with planned and unplanned readmissions
Planned readmissions were most common in the department of head and neck oncology both in terms of the total number [38,461] and readmission rate ($45.0\%$) followed by the department of hematology (9,851; $29.9\%$). Head and neck oncology had the highest number of unplanned readmissions (1,172; $1.4\%$), followed by nephrology (595; $1.5\%$), followed by the department of nephrology (595, $1.5\%$) and hematology (547; $1.7\%$). The highest rate of unplanned readmission was in the department of rheumatology ($2.2\%$), followed by the department of ophthalmology ($2.1\%$) and hematology ($1.7\%$). The number and rate of planned and unplanned discharges in the department of internal medicine were much higher than the number and rate of planned and unplanned admissions in the department of surgery (Table 4).
Table 4Characteristics of hospital departmentsDepartmentPlannedNDischargedN(%)DepartmentUnplannedNDischargedN(%)Internal medicineHead & Neck Oncology38,46185,46445.0Head & Neck Oncology1,17285,4641.4Hematology9,85132,96829.9Nephrology59541,1301.5Abdominal Oncology9,15457,50415.9Hematology54732,9681.7Thoracic Oncology6,21437,19216.7Thoracic Oncology41737,1921.1Lung Cancer Center1,98616,47012.1Rheumatology41618,7212.2SurgeryOphthalmology65989,1550.7Ophthalmology1,74984,7422.1Gastrointestinal Surgery46539,4801.2Urology24745,2870.6Urology15846,9660.3Liver Surgery22330,4130.7Orthopedics8463,4870.1Orthopedics17363,4870.3Breast Surgery7328,2280.3Gastrointestinal Surgery17239,4800.4
## Annual distribution of planned and unplanned readmissions
The number of planned readmissions and planned readmission rates increased progressively with the number of admissions except in 2020. The number of planned readmissions and the readmission rate (15,996; $7.4\%$) were the highest in 2019. The number and readmission rate of unplanned readmissions (2,451; $1.3\%$) were highest in 2020 (Table 5).
Table 5Chronological distribution of planned and unplanned readmissionsYearDischargedNPlanned(N, %)Unplanned(N %)2015163,5158,461 ($5.2\%$)1,245 ($0.8\%$)2016170,56710,901 ($6.4\%$)1,432 ($0.8\%$)2017185,49912,377 ($6.7\%$)1,352 ($0.7\%$)2018201,60414,845 ($7.4\%$)1,471 ($0.7\%$)2019215,28915,996 ($7.4\%$)1,944 ($0.9\%$)2020181,96311,914 ($6.5\%$)2,451 ($1.3\%$)Total1,118,43774,494 ($6.7\%$)9,895 ($0.9\%$)
## Monthly distribution of planned and unplanned readmissions
Among the 12 months of the year, the highest number of planned readmissions was recorded in November (7,283; $7.2\%$), followed by September (7,201; $7.2\%$) and July (7,108; $7.2\%$). The highest number of unplanned admissions was recorded in September (1,020; $1.0\%$), followed by November (1,008; $1.0\%$) and December (984; $0.9\%$). The month with the lowest number of both planned and unplanned admissions was February (4,372; $7.0\%$ compared with 435; $0.7\%$) (Table 6).
Table 6Monthly distribution of planned and unplanned readmissionsMonthDischargedPlannedUnplannedJan91,8615,139($5.6\%$)713($0.8\%$)Feb62,4984,372($7.0\%$)435($0.7\%$)Mar92,2505,455($5.9\%$)603($0.7\%$)Apr94,2575,783($6.1\%$)736($0.8\%$)May95,0826,378($6.7\%$)787($0.8\%$)Jun95,3046,263($6.6\%$)806($0.8\%$)Jul98,9687,108($7.2\%$)982($1.0\%$)Aug95,8556,549($6.8\%$)868($0.9\%$)Sept99,5737,201($7.2\%$)1,020($1.0\%$)Oct85,8466,765($7.9\%$)953($1.1\%$)Nov101,1907,283($7.2\%$)1,008($1.0\%$)Dec105,7536,198($5.9\%$)984($0.9\%$)Total1,118,43774,494($6.7\%$)9,895($0.9\%$)
## Discussion
In this study, we analyzed planned and unplanned 30-day readmission rates and associated characteristics at a large general university hospital in China. It is the only hospital in the region with the highest referral rate across all medical specialties and it treats the most complex and difficult cases, which are more likely to be readmitted.
We found some significant differences between 30-day planned and unplanned readmission patients. According to our study, women are twice as likely as men to have plan readmission. However, there were $15.7\%$ points more female patients with planned readmissions than female patients with unplanned readmissions. There were significantly more female patients with planned readmission than those with unplanned readmission ($P \leq 0.001$). Some studies identified men as a risk factor for 30-day readmission [13]. However, most studies that included sex-based readmission showed no difference between sex and readmission rate [14, 15]. Although gender may be a risk factor for readmission in some diseases [14], large prospective studies of gender-related readmission are needed.
The proportion of patients living with a spouse was significantly higher for planned readmissions than for unplanned readmissions, and the proportion of divorced, single and widowed patients was significantly higher for unplanned readmissions than for planned readmissions ($p \leq 0.001$). This finding is consistent with previous studies showing that marriage has a protective effect on unplanned readmission [16–18].
In this study, the interval from discharge to readmission was mainly concentrated in 20≤-≤30 days for both planned and unplanned readmission patients, $55.9\%$ and $45.7\%$, respectively, the interval between planned readmission was significantly higher than that of unplanned readmission (18.54±7.96 vs. 16.18±10.01 days). The results of this study showed that 5,683 patients ($7.6\%$) had planned readmission within 7 days (1≤-<7), of which 1,262 ($1.7\%$) were readmitted within 3 days (1 ≤-<3). From a clinical perspective, this may be a misclassification of planned readmissions. This requires additional validation work (review of the medical records) to examine in more detail planned readmissions that may have been misclassified [19]. Most studies suggested that a 7-day cut-off is an effective intervention point for early and preventable readmissions [19]. Readmissions within the first seven days after hospital discharge were more likely to be preventable than those occurring in a late period of 8–30 days [19–21]. Some studies have shown that early readmissions (≦ 7 days) within 30 days of discharge are twice as likely to be preventable as late readmissions, with adjusted preventability rates decreasing significantly after day 7 post-discharge. Readmissions within one week of discharge were more likely to be preventable [20, 22, 23]. Other studies considered readmissions that occurred within 0 to 10 days were judged to be preventable [24].
In this study, oncology patients had the highest number of chemotherapy treatments in both planned and unplanned readmissions. This is related to the highest number of oncology patients hospitalized in this study. The number and proportion of planned readmissions in internal medicine were much higher than in surgery. With the exception of ophthalmology, the number and proportion of unplanned readmissions were much higher in internal medicine than in surgery. Planned readmissions were primarily related to the specific nature of a disease, which is considered unavoidable because it results from a typical clinical pathway [25]. However, insight into planned readmissions can facilitate the efficient allocation and optimization of healthcare resources.
Based on the results of this study, there was a correlation between planned readmissions of patients, with the number of hospital admissions increasing each year from 2015 to 2019, as did the number of planned readmissions. A decrease in both inpatient admissions and planned readmissions in 2020 due to the COVID-19 pandemic. However, we found no correlation between the number of unplanned readmissions and hospital admissions. In particular, although the number of hospitalizations was lower in 2020 than in 2017, the number and proportion of unplanned readmissions were the highest. This may be related to COVID-19 affecting the health status of patients or the quality of care. This reason needs to be investigated further.
Throughout the year, planned and unplanned readmissions showed a monthly distribution over the last 6 years. The highest numbers of planned and unplanned readmissions were recorded in November (7,283; $7.2\%$) and September (1,020; $1.0\%$). However, the number of planned and unplanned readmissions was lowest in February (4,372; $7.0\%$ vs. 435; $0.7\%$). February is usually the Chinese Lunar New Year, and due to traditional Chinese culture, hospital visits are not usually made during the New Year [26]. This factor is often specific to the context of Asian countries and reflects the social and cultural context. Therefore, we believe that social and cultural factors are also the influencing factors of planned and unplanned readmission.
## Limitation
There are several limitations to this study. First, the study used data from a university hospital, and our findings may only apply to similar providers, so generalizing these findings to other types of hospitals may be risky. Second, the study was retrospective and included a limited number of variables, so it is subject to residual confounding and may differ from the true causal effect. Third, we validated the ICD-10 codes; there may be inaccuracies in the coding that could introduce imprecision into our estimates.
## Conclusion
This study found that social and cultural factors may also influence planned and unplanned readmissions. Planned readmissions of less than 7 days may be misclassified and should be reviewed as unplanned readmissions. The study of planned readmissions can help to optimize and allocate healthcare resources. Analysis of risk factors for unplanned readmissions (LOS, male, separated/divorced, single, widowed/other, ICU stay and surgery) will help identify key combinations of interventions that are effective in preventing readmissions.
## References
1. Spiva L, Hand M, VanBrackle L, McVay F. **Validation of a predictive model to identify patients at high risk for Hospital Readmission**. *J Healthc Qual* (2016.0) **38** 34-41. DOI: 10.1111/jhq.12070
2. Hasan O, Meltzer DO, Shaykevich SA. **Hospital readmission in general medicine patients: a prediction model**. *J Gen Intern Med* (2010.0) **25** 211-9. DOI: 10.1007/s11606-009-1196-1
3. 3.CMS Office of Minority HealthImpact of Hospital Readmissions reduction initiatives on vulnerable populations2020Baltimore, MDCenters for Medicare & Medicaid Services; September. *Impact of Hospital Readmissions reduction initiatives on vulnerable populations* (2020.0)
4. Ryu B, Yoo S, Kim S, Choi J. **Thirty-day hospital readmission prediction model based on common data model with weather and air quality data**. *Sci Rep* (2021.0) **11** 23313. DOI: 10.1038/s41598-021-02395-9
5. Sharma Y, Miller M, Kaambwa B, Shahi R, Hakendorf P, Horwood C. **Factors influencing early and late readmissions in australian hospitalised patients and investigating role of admission nutrition status as a predictor of hospital readmissions: a cohort study**. *BMJ Open* (2018.0) **8** 972-8. DOI: 10.1136/bmjopen-2018-022246
6. Zamir D, Zamir M, Reitblat T, Zeev W, Polishchuk I. **Readmissions to hospital within 30 days of discharge from the internal medicine wards in southern Israel**. *Eur J Intern Med* (2006.0) **17** 20-3. DOI: 10.1016/j.ejim.2005.10.004
7. 7.Centers for Medicare & Medicaid Services (CMS). Hospital Readmissions Reduction Program (HRRP). Available from: https://www.cms.gov/Medicare/Medicare-Fee-for Service-Payment/AcuteInpatientPPS/Readmissions-Reduction-Program. Accesed Jan 2022.
8. Wadhera RK, Yeh RW, Joynt Maddox KE. **The Hospital Readmissions Reduction Program - Time for a Reboot**. *N Engl J Med* (2019.0) **380** 2289-91. DOI: 10.1056/NEJMp1901225
9. Kossovsky MP, Perneger TV, Sarasin FP, Bolla F, Borst F, Gaspoz JM. **Comparison between planned and unplanned readmissions to a department of internal medicine**. *J Clin Epidemiol* (1999.0) **52** 151-6. DOI: 10.1016/S0895-4356(98)00142-5
10. 10.Roshanghalb A, Mazzali C, Lettieri E, Paganoni AM, Bottle A. Stability over time of the “hospital effect” on 30-day unplanned readmissions: evidence from administrative data. Health Policy. 2021 Oct;125(10):1393–7. 10.1016/j.healthpol.2021.07.009.
11. 11.Tahhan G, Farber A, Shah NK, Krafcik BM, Sachs TE, Kalish JA, et al. Characterization of Planned and unplanned 30-Day readmissions following vascular Surgical Procedures. Vasc Endovascular Surg. 2017 Jan;51(1):17–22. 10.1177/1538574416682176.
12. 12.OECD. Unplanned hospital re-admissions for mental disorders, in Health at a Glance 2011. 10.1787/health_glance-2011-46-en
13. Roger C, Debuyzer E, Dehl M, Bulaïd Y, Lamrani A, Havet E, Mertl P. **Factors associated with hospital stay length, discharge destination, and 30-day readmission rate after primary hip or knee arthroplasty: Retrospective Cohort Study**. *Orthop Traumatol Surg Res* (2019.0) **105** 949-55. DOI: 10.1016/j.otsr.2019.04.012
14. Hoang-Kim A, Parpia C, Freitas C, Austin PC, Ross HJ, Wijeysundera HC, Tu K, Mak S, Farkouh ME, Sun LY, Schull MJ, Mason R, Lee DS, Rochon PA. **Readmission rates following heart failure: a scoping review of sex and gender based considerations**. *BMC Cardiovasc Disord* (2020.0) **20** 223. DOI: 10.1186/s12872-020-01422-3
15. Kansagara D, Englander H, Salanitro A, Kagen D, Theobald C, Freeman M, Kripalani S. **Risk prediction models for hospital readmission: a systematic review**. *JAMA* (2011.0) **306** 1688-98. DOI: 10.1001/jama.2011.1515
16. Heaton PC, Desai VC, Kelton CM, Rajpathak SN. **Sulfonylurea use and the risk of hospital readmission in patients with type 2 diabetes**. *BMC Endocr Disord* (2016.0) **16** 4. DOI: 10.1186/s12902-016-0084-z
17. Mor A, Ulrichsen SP, Svensson E, Berencsi K, Thomsen RW. **Does marriage protect against hospitalization with pneumonia? A population-based case-control study**. *Clin Epidemiol* (2013.0) **5** 397-405. DOI: 10.2147/CLEP.S50505
18. 18.Shah LM, Ding J, Spaulding EM, Yang WE, Lee MA, Demo R, et al. Sociodemographic characteristics Predicting Digital Health intervention use after Acute myocardial infarction. J Cardiovasc Transl Res. 2021 Oct;14(5):951–61.
19. 19.Ellimoottil C, Khouri RK, Dhir A, Hou H, Miller DC, Dupree JM. An opportunity to improve Medicare’s Planned Readmissions measure. J Hosp Med. 2017 Oct;12(10):840–2.
20. Graham KL, Auerbach AD, Schnipper JL, Flanders SA, Kim CS, Robinson EJ. **Preventability of early Versus Late Hospital Readmissions in a National Cohort of General Medicine Patients**. *Ann Intern Med* (2018.0) **168** 766-74. DOI: 10.7326/M17-1724
21. Graham KL, Wilker EH, Howell MD, Davis RB, Marcantonio ER. **Differences between early and late readmissions among patients: a cohort study**. *Ann Intern Med* (2015.0) **162** 741-9. DOI: 10.7326/M14-2159
22. Meurs EAIM, Siegert CEH, Uitvlugt E, Morabet NE, Stoffels RJ, Schölvinck DW. **Clinical characteristics and risk factors of preventable hospital readmissions within 30 days**. *Sci Rep* (2021.0) **11** 20172. DOI: 10.1038/s41598-021-99250-8
23. Kumar V, Chaudhary N, Achebe MM. **Epidemiology and predictors of all-cause 30-Day readmission in patients with sickle cell crisis**. *Sci Rep* (2020.0) **10** 2082. DOI: 10.1038/s41598-020-58934-3
24. Bianco A, Molè A, Nobile CG, Di Giuseppe G, Pileggi C, Angelillo IF. **Hospital readmission prevalence and analysis of those potentially avoidable in southern Italy**. *PLoS ONE* (2012.0) **7** e48263. DOI: 10.1371/journal.pone.0048263
25. Kryś J, Łyszczarz B, Wyszkowska Z, Kędziora-Kornatowska K. **Prevalence, reasons, and predisposing factors Associated with 30-day hospital readmissions in Poland**. *Int J Environ Res Public Health* (2019.0) **16** 2339. DOI: 10.3390/ijerph16132339
26. Chiu SL, Gee MJ, Muo CH, Chu CL, Lan SJ, Chen CL. **The sociocultural effects on orthopedic surgeries in Taiwan**. *PLoS ONE* (2018.0) **13** e0195183. DOI: 10.1371/journal.pone.0195183
|
---
title: Decision tree learning in Neo4j on homogeneous and unconnected graph nodes
from biological and clinical datasets
authors:
- Rahul Mondal
- Minh Dung Do
- Nasim Uddin Ahmed
- Daniel Walke
- Daniel Micheel
- David Broneske
- Gunter Saake
- Robert Heyer
journal: BMC Medical Informatics and Decision Making
year: 2023
pmcid: PMC9988195
doi: 10.1186/s12911-023-02112-8
license: CC BY 4.0
---
# Decision tree learning in Neo4j on homogeneous and unconnected graph nodes from biological and clinical datasets
## Abstract
### Background
Graph databases enable efficient storage of heterogeneous, highly-interlinked data, such as clinical data. Subsequently, researchers can extract relevant features from these datasets and apply machine learning for diagnosis, biomarker discovery, or understanding pathogenesis.
### Methods
To facilitate machine learning and save time for extracting data from the graph database, we developed and optimized Decision Tree Plug-in (DTP) containing 24 procedures to generate and evaluate decision trees directly in the graph database Neo4j on homogeneous and unconnected nodes.
### Results
Creation of the decision tree for three clinical datasets directly in the graph database from the nodes required between 0.059 and 0.099 s, while calculating the decision tree with the same algorithm in Java from CSV files took 0.085–0.112 s. Furthermore, our approach was faster than the standard decision tree implementations in R (0.62 s) and equal to Python (0.08 s), also using CSV files as input for small datasets. In addition, we have explored the strengths of DTP by evaluating a large dataset (approx. 250,000 instances) to predict patients with diabetes and compared the performance against algorithms generated by state-of-the-art packages in R and Python. By doing so, we have been able to show competitive results on the performance of Neo4j, in terms of quality of predictions as well as time efficiency. Furthermore, we could show that high body-mass index and high blood pressure are the main risk factors for diabetes.
### Conclusion
Overall, our work shows that integrating machine learning into graph databases saves time for additional processes as well as external memory, and could be applied to a variety of use cases, including clinical applications. This provides user with the advantages of high scalability, visualization and complex querying.
## Background
Graph databases enable efficient storage of heterogeneous and highly interlinked data, such as clinical datasets [1]. Usually, clinical data sets comprise the patient information, diagnoses, metadata, and results of different examinations (for instance, simple blood pressure measurements, the latest CT and MRT scans, or high-resolution omics data) that are often graph shaped. Subsequently, researchers can extract relevant features from these datasets and apply machine learning for diagnosis, biomarker discovery, or understanding pathogenesis.
However, data extraction and subsequent machine learning using a standard machine learning toolbox have the additional process of storing data in memory external to the database. Hence, a better workflow would be to apply machine learning directly to the data stored in the graph database.
To show the feasibility of this approach, we apply decision tree learning directly in Neo4j and persist the final tree in Neo4j. Therefore, we have created an open-source Neo4j plugin (Decision Tree Plugin (DTP))1, which exposes procedures for decision tree creation and execution on data stored in Neo4j. Thus, the final created decision trees can also be visualized in the Neo4j Browser.
While building DTP, we used three clinical datasets to realize common trends of such data, such as missing values, feature handling and evaluation metrics while generating decision tree algorithms in Neo4j. To assess its efficiency, we have evaluated the accuracy, Matthews Correlation Coefficient2 and the computational time of our plugin compared to decision tree functions from Python and R on the datasets. Furthermore, we applied our procedures to a fourth two-log fold larger dataset about diabetes to assess big data performance and evaluate its clinical applicability.
In our research on learning algorithms in Neo4j, we contribute the following:DTP comprises 24 procedures, which can read CSV files, map nodes, split data, generate decision tree using three different splitting criteria (Fig. 1), perform k-fold cross validation, validate the classifier and visualize the decision tree. Moreover, it contains procedures to analyze the features in a dataset, with their respective values of gini index, information gain and/or gain ratio. The novelty of integrating machine learning algorithms is specific to Neo4j and not graph databases in general. An extensive comparison of our plugin with state-of-the-art machine learning libraries in Python and R shows that our plugin achieves similar quality of predictions due to the integration of machine learning in the Neo4j graph database. Further, there is only a negligible increase in generation time of algorithms when compared to Python. Finally, we have used our plug-in to generate decision tree algorithms on a very large dataset, comprising demographic and pathogenic information of diabetic, borderline diabetic and non-diabetic patients. We used the generated algorithms to gain basic clinical insights on the disease, and prove that this tool could be used for such use cases. In the following, we will present our research questions, background as well as related work of our study. Thereafter, in the section Methods, we describe the used methodology, our implemented plugin and the used datasets. In the section Results, we evaluate DTP against Java, R and Python decision tree algorithms. Finally in the section Discussion we discuss key aspects of our research and conclude our research by proposing future work in the section Conclusion.
## Research questions
To evaluate the importance and performance of our plug-in, we defined three research questions, which we will answer through the evaluation of our plugin:
## RQ 1
What is the difference in the quality of predictions for algorithms generated in DTP, when compared against algorithms generated by standard libraries from Python and R?
## RQ 2
What is the difference in generation time of decision tree algorithms, generated by DTP, compared to standard libraries from Python and R?
## RQ 3
Could applying decision tree learning on homogeneous and unconnected nodes created from large clinical datasets provide basic clinical insights?
Now we will provide a brief overview on graph databases, decision tree algorithms and a short introduction about diabetes to interpret the clinical background. Afterwards, we will discuss related work on the integration of machine learning in databases. Table 1Comparison of Database Management Systems [3, 4]. RDBMS, OODBMS and Graph refers to Relational, Object-Oriented and Graph Database Management SystemsRDBMSOODBMSGraph (Neo4j)Flexibility (lack of schema)LowMediumHighQuery languageSQLRarely implementedCypherQuery performanceHighHighHighIntegrity constraintsYesYesYesLevel of supportHighLowHighEase of programmingHighHighHighSecurityHighLowHighScalabilityLowHighHigh Fig. 1Data flow in the decision tree plug-in (DTP) Fig. 2Available procedures in DTP
## Graph database management system: Neo4j
Graph databases like Neo4j (Table 1), equipped with its query language Cypher, are NoSQL databases that store data within a graph structure, enabling flexible queries through interlinked data. The main advantages over relational and object-oriented databases are its flexible data models and schema. Furthermore, graph database are easily scalable, making it ideal to store large datasets and perform read or write operations [5].
In addition to the advantages intrinsic to a graph database, Neo4j offers procedures that handle complex operations [6]. They are implemented in Java [7] and are compiled into a Jar file, which can later be deployed to the database by adding the Jar files into the plugin directory on each individual or clustered server. A procedure can be invoked either as a stand-alone procedure from the application, the command-line or as a part of Cypher statements.
Cypher is a declarative query language, optimized for graphs, and therefore, can be used for describing visual patterns in graphs using ASCII-Art syntax. This makes Cypher queries much simpler and easier to write compared to SQL [8].
There could be three main arguments for using a graph database like Neo4j, to analyse clinical data of any scale:Scalability Graph databases are scalable [9].Node-edge Structure This helps in forming and analysing complex relationships. [ 10]Query Performance Querying can be versatile in graph databases which helps in analysing complex data structures [9].
## Decision tree algorithms
Table 2Overview about decision tree algorithms grouped by the splitting criteriaData typesMissing valuesData splittingGini indexCategorical and numericalCan handleNo restrictionsInformation gainNumericalCan not handleNo restrictionsGain ratioCategorical and numericalCan handleBinary Decision tree algorithms are machine learning algorithms predicting attributes based on tree-like decision rules. In the literature, several implementations of decision tree algorithms are known (Patel et al. [ 11]), including CART (Breiman et al. [ 12]), ID3 (Quinlan [13]), C4.5 (Quinlan [14]). These algorithms can be separated based on their splitting criteria (Table 2) into algorithms using either Gini Index (GI), Information Gain (IG), or Gain Ratio (GR).
In our plug-in, we provide procedures for all three splitting criteria—Gini Index, Information Gain, and Gain Ratio. *The* generated trees are binary, implying a binary split at a decision node, with one path agreeing to a certain threshold and the other one disagreeing.
We have also provided a parameter to perform pre-pruning on the generated decision trees through our procedures and reduce the size of tree as per user choice. This is useful as it can help to overcome over-fitting of decision trees algorithms and remove noise/outliers from training data [15].
## Diabetes
Diabetes is a group of diseases characterized by hyperglycemia. It results either from defects in insulin secretion, insulin action, or both [16]. *In* general, diabetes has two etiopathogenetic categories. The first, diabetes type I, is also called insulin-dependent diabetes. This diabetes type is caused by the autoimmune destruction of beta cells and is associated with multiple genetic predispositions and environmental factors that are poorly defined. Diabetes type II, called non-insulin-dependent diabetes, is the more frequent type (90–95 % of all diabetes cases).
The risk of developing type II diabetes increases with age, obesity, hypertension, smoking, and lack of physical activity. It is also more frequent in individuals with hypertension or dyslipidemia (i.e., low HDL cholesterol concentration and high LDL-cholesterol concentration) and is associated with a strong genetic predisposition. This form is often undiagnosed for years because of the gradually developing hyperglycemia. At early stages, hyperglycemia is not severe enough to cause diabetes symptoms [16], however, chronic hyperglycemia can cause dysfunction and failure of different organs, e.g., eyes, kidneys, nerves, and heart. Therefore, diabetes is a potential risk factor for stroke and cardiovascular diseases [17]. Hence, our investigations of the diabetes dataset represent an important real-world use case.
## Related work
While the integration of machine learning algorithms into a graph database (i.e., Neo4j) is novel, both supervised and unsupervised learning algorithms were already applied on the following relational database management systems [18] allowing users to implement learning algorithms directly on the database:Amazon RedshiftBlazing SQLGoogle Cloud Big QueryIBM DB2 WarehouseKineticaMicrosoft SQL Server Machine Learning ServicesOracle Cloud Infrastructure (OCI) Data ScienceVertica Analytics PlatformFocusing on Neo4j, Max De Marzi et al. [ 19] created custom procedures in Java to predict for a particular data set whether a student passes an exam. However, these procedures are strictly confined to that dataset and cannot be used for other data sets. Analogously, Michael Hunger et al. [ 20] as well as Anjana and K. Lavanya et al. [ 21] showed some steps towards machine learning in Neo4j but provided no universal method.
## Methods
This section presents our applied methods to build and optimize the DTP. Furthermore, we describe the datasets used to evaluate our algorithms. For our study, we selected the graph database Neo4j due to its flexibility and the possibility to extend its functionalities using Java procedures, which are the backbone of the whole DTP infrastructure (Figs. 1,2). Java is one of the most versatile languages and can be currently executed on most operating systems. Therefore, *Java is* a great language choice to equip a graph database like Neo4j, with several tools for machine learning.
Figure 1 highlights the flow of data between Java and Neo4j when the procedures in DTP are executed, either from CSV files or from homogeneous and unconnected nodes. The DTP data flow (Fig. 1) starts with uploading the input data (CSV files or nodes in Neo4j) and splitting them into train and test data.
The next step is to select among the different splitting criteria (see Section Decision tree algorithms) and run the decision tree implementation to generate a classifier. The learned classifier could be evaluated based on a confusion matrix, accuracy, generation time, and Matthews correlation coefficient and could be applied to new data sets. The resulting nodes and edges of the tree are stored in Neo4j, allowing visualization of the resulting decision tree.
We will now elaborate about the used splitting criteria, the implemented stored procedures and the possibility to visualize the resulting decision tree.
## Procedures of the decision tree plugin
To assess the performance of the different algorithms, splitting criteria and Neo4j, we implemented DTP as a set of Java-based procedures in Neo4j (see Fig. 2). The whole package of DTP is saved as a Jar file which should be copied inside the Neo4j plugins directory of a database. Afterwards, the database should be restarted to make these procedures available, through Cypher queries to generate a decision tree.
These procedures build the tree in Java, returns node and edge buckets to Neo4j for tree visualization and afterwards tests the decision tree with the test data instances. The test data is recursively passed through the tree until it reaches a leaf node, in which case it returns the found class label – the final prediction. To create the confusion matrix (with or without cross validation), the actual labels are compared against the predicted labels to calculate the accuracy, Matthews correlation coefficient, or output the confusion matrix and the time needed to generate the algorithms. Table 3Computational complexity analysisComputational complexityTimeSpaceTrain complexityO(dim*nlogn)O(n)Test complexityO(depth)O(n)n = number of instances/nodes dim = number of dimensions/variables depth = depth of generated decision tree In Table 3 we provide a summary of the computational complexity of the Java procedures contained in DTP, using big O notation. For any decision tree procedure, there are two processes involved—training and testing the algorithm. Training complexity is naturally higher as during testing the task is to just traverse the tree, generated during training, where several calculations of splitting criteria are required.
The procedures that can visualize the tree also contain a parameter—“max_depth”, which will limit the depth of the tree to the specified value. It is a pre-pruning mechanism which will stop decision tree generation when the mentioned depth is reached and majority class label is assigned to impure nodes.
Note that with each tree generated from procedures, our plug-in DTP will generate and display the confusion matrix, accuracy, Matthews Correlation Coefficient and generation time of algorithms. The decision tree procedures in DTP can be categorized as follows:
## Cross validation (without tree visualization)
Six procedures are provided to perform k-fold cross-validation on a single CSV file or a set of nodes. Users can specify the class label and number of folds for every iteration along with CSV file path, exclusively for tree generation from CSV files. It is important to note that there are no tree visualization for cross validation.
## Decision tree from CSV files (with tree visualization)
These three procedures are used to create a decision tree from plain CSV files, and the resulting tree is stored in Neo4j. To this end, DTP reads 2 CSV files (of train and test data) and detects whether a variable is categorical or numerical. Furthermore, a user-defined class label can be chosen. DTP then recursively calculates the best splits based on the splitting criteria, depending on the chosen algorithm.
The result of the previous step is a bucket of nodes and relationships which represent the tree, stored as a graph in Neo4j for subsequent inference or visualization. The user can specify file paths for training and testing, and the class label. For pre-pruning, user must also set prune = “True” to set the max_depth value of decision tree.
## Decision tree from graph-shaped data in Neo4j (with tree visualization)
To create a decision tree from homogeneous and unconnected nodes in Neo4j, a user can map two distinct sets of nodes into training and testing, through the procedures. Alternatively, we have created a procedure that allows the user to automatically split the data into train and test sets.
In order to allow generating the decision tree, we implemented three different functions according to the favored splitting criterion. These functions then use the labeled training data and persists the tree nodes and relationships in Neo4j just as the procedures on the CSV files do. The user can specify the class and for pre-pruning, user must also set prune = “True” to set the max_depth value of decision tree.
## Decision tree without tree visualization (confusion matrix)
A user is also allowed to generate and assess decision tree algorithms without tree visualization, we have implemented 6 procedures – 3 to generate the confusion matrix from CSV files and 3 to generate it from nodes. This is useful in cases where the dataset is large, and the node visualization creates a significant delay in the tree visualization. The user can specify file paths for training and testing and the class label. For pre-pruning, user must also set prune = “True” to set the max_depth value of decision tree.
## Feature analysis
For further in depth analysis of each feature in datasets, DTP contains procedures to obtain values for gini index, information gain or gain ratio as calculated at every level while the tree is being generated. This will allow user to have an elaborate overview on how each variable affects the generation of tree at each level. The user can specify pruning of tree to a max_depth for these procedures and can be generated from both nodes and/or csv files.
## Data
Table 4Comparison of datasets used in the experimentsInstancesVariablesTarget categoriesClass imbalanceDataset 1299132MediumDataset 248503LowDataset 31485122HighDataset 4253,680223High To evaluate our decision tree algorithms, we searched for clinical datasets in Kaggle, GitHub, and in research papers. After surveying and investigating several datasets, we selected four datasets about the prediction of heart failures (Dataset 1), inflammatory bowel disease (Dataset 2), classification between flu and COVID-19 (Dataset 3) and prediction of diabetes among patients (Dataset 4):
## Dataset 1: heart failure prediction
The heart failure dataset by Davide Chicco and Giuseppe Jurman [22] from Kaggle contains 299 patients’ data with 13 demographics variables and has been used to predict survival of patients using machine learning algorithm. Of the 13 variables, 7 are continuous numeric and the rest are categorical, including the class variable. The class variable is the property DEATH_EVENT, where 1 represents death of a patient (96 instances) and 0 (203 instances) their survival.
## Dataset 2: prediction of inflammatory bowel disease from microbiome
The inflammatory bowel disease dataset by T. Lehmann [23] consists of 2,969 meta-proteins whose presence has been measured among a group of 48 patients of 3 separate groups. All the variables in this dataset are numeric variables. We have considered only the 50 most abundant meta-proteins while training and evaluating the decision tree algorithms. The class variable is Patient Type – C, CD or UC, where C identifies control patients (20 instances), CD for patients with Crohn’s Disease (13 instances) and UC for patients with Ulcerative Colitis (15 instances).
## Dataset 3: H1N1/COVID-19 classification
The H1N1/COVID-19 dataset was taken from a research article by Li et al. [ 24] that applied machine learning on a dataset of 1,485 patients with 50 demographic variables. The class variable is Diagnosis – H1N1 (1072 instances) or COVID-19 (413 instances). Since a lot of variables were plainly null, we have reduced the data to 12 variables out of the 50 available to sharpen the results. There are two numeric variables which are continuous and the rest are categorical variables, including the class variable.
## Dataset 4: diabetes (type II) health indicators
This dataset was taken from Kaggle and was uploaded by Alex Teboul [25] who has cleaned and consolidated the data from the original BRFSS 2015 [26] dataset consisting of data from a survey of patients concerning diabetes. The original dataset was compiled by the Center for Disease Control and Prevention which is the national public health agency of the United States. The cleaned data consists of 253,680 patients with 22 demographic and clinical variables. The class variable is Diabetes_012 consisting of 3 labels – 0 (213,703 instances) indicating patient being non-diabetic, 1 (4631 instances) indicating prediabetes or 3 (35,346 instances) indicating patient being diagnosed with type 2 diabetes. There are two numeric variables which are continuous, and the rest are categorical variables, including the class label.
## Experimental setup
Table 5Decision tree configurations for experiments 1, 2 and 3: three datasets for each combination of language and splitting criteria42 algorithms generatedGini indexInformation gainGain ratioR (rpart, RWeka)333Python (sklearn)NA33Java333DTP (csv)333DTP (nodes)333 For our evaluation, we have compared the performance of the custom java procedures in our plug-in to algorithms generated by the following standard packages in R and python:R package “rpart” for gini index and information gain and “JWeka” for gain ratiopython package “sklearn” for gini index and information gain. No standard package implementation was found for gain ratio during evaluation. All the experiments were carried out on a desktop PC with the following specifications:Processor AMD Ryzen5 3600, 6 cores (3.6 Ghz)Memory 16 GB of RAMGraphic NVIDIA GeForce RTX2070 (8 GB)
## Experiments 1, 2 and 3: building and optimizing DTP
For our experiments, we applied k-fold cross-validation for all the three datasets and all four approaches (Python, R, Java, Neo4j) and evaluated the algorithms. Each decision tree algorithm was regenerated thirty times with the accuracy, Matthews correlation coefficient and generation time averaged out for all the iterations. The number of folds for cross-validation was varied across the datasets, due to the difference in their instance size (see Table 3).
In total, we generated 42 (see Table 5) cross-validated decision tree algorithms for this evaluation in R, Python, Java3, DTP (CSV)4 and DTP (nodes)5 and 2 (Gini Index and Info Gain) in Python. This was due to the unavailability of a generic package for Gain Ratio in Python.
## Experiment 4: evaluating DTP on a large dataset
*To* generate and evaluate decision tree algorithms on Dataset 4, we have performed 5-fold cross validation (80 percent data for training and 20 percent for testing) on the whole dataset in R, Python, Java and DTP (csv) and DTP (nodes), totaling to 14 decision tree algorithms, across the mentioned tools.
## Results
Fig. 3Box Plots—Accuracy and Matthews Correlation Coefficient of the algorithms: A, D for different tools including DTP, B, E for different splitting criteria in DTP, and C, F for the datasets 1-3 in DTP Fig. 4Box Plots—Generation Time of the Decision Tree Algorithms: A for different tools including DTP, B for different splitting criteria in DTP, and C for the datasets 1-3 in DTP Fig. 5Box Plots—Evaluation of the diabetes dataset(Dataset 4), across different tools: A accuracy, B precision, C Matthews Correlation Coefficient, and D generation time In the following, we examine the performance of DTP in Neo4j compared to standard machine learning algorithms in R and Python at first for the small datasets 1-3 and afterwards for the big dataset 4.
## Experiment 1: prediction performance comparison
In our first experiment, we assessed whether the different implementations and thus data accesses impact the accuracy and Matthews correlation coefficient (MCC). To compare the accuracy distribution visually, we visualized box plots in Fig. 3. The results show that all algorithms act in the same accuracy and MCC range.
While our DTP trained on data in Neo4j and from CSV files is on par with the Java implementation, it is clearly more stable than the Python decision tree. Considering the different splitting algorithms, gain ratio is the best metric providing the best median value. Low MCC for Python (Fig. 3A) was caused by overfitted decision trees along with class imbalance in Dataset 3 (Table 4).
## Experiment 2: computational time comparison
Next, we evaluate the generation time of DTP-Neo4j-Plugin for the three test datasets against the implementations from R and Python in Fig. 4. The Neo4j integration had a positive impact on the generation time. The decision tree learning from data inside Neo4j is the fastest and most stable approach. In contrast, R had a big deviation in generation time when loading data from CSV files. A more in-depth investigation has shown that this is due to a lot of data shuffling inside R, which creates a considerable overhead. The investi- gation of the splitting algorithms shows that gain ratio provides the fastest generation times.
The high performance of python, in terms of low generation time, could be attributed to the package sklearn, part of which was written in C and C++, which are extremely fast at compiling. [ 27]Fig. 6Scatter Plots with Line of Regression—To interpolate the effect of instance size (rows/nodes) on generation time and accuracy of algorithms generated by DTP for all the 4 datasets (Dataset 1, 2, 3 and 4)
## Experiment 3: impact of dataset characteristics
Fig. 7Dataset 4 uploaded as homogeneous and unconnected nodes in Neo4j Figure 4C (with reference to Table 4) shows that the generation time is directly proportional to the number of instances used while training the algorithm. It might seem that the generation time is directly proportional to accuracy as well which can be explained through a causal link – a higher number of training instances takes up a higher generation time and provides higher accuracy as well, which is quite understandable, since a well-trained algorithm would provide better accuracy.
A regression plot for all algorithms generated on the 4 datasets in DTP is shown in Fig. 6. This is discussed briefly in the next section.
## Experiment 4: evaluation of algorithms generated on dataset 4 to predict diabetes in patients
In this experiment (Fig. 5), we evaluated the different tools while running on the large Dataset 4. Note that the differences in values are quite insignificant with respect to quality of predictions, while there are noticeable differences in the generation time of the algorithms. Python’s package sklearn provides consistently fast performance for Gini Index and Information Gain, regardless of the size of dataset.
To calculate the precision of predictions from Dataset 4, we assumed binary classification while assigning “True Positive” and “True Negative” labels to values in the 3*3 confusion matrix. Hence, the precision values highlight the relevance of the predictions in distinguishing between people with no diabetes (class label = 0) vs. people with borderline and confirmed diabetes (class label = 1 and 2). There was insignificant difference between precision among the different tools.
## Performance of DTP procedures
With slight differences in the experimental results, it is safe to say that DTP performs quite similarly with packages in Python and R, in terms of prediction quality as well as generation time of algorithms. The primary barrier in predictive modelling on large data, such as decision tree classification, is a significant drop in the quality of predictions. Without pruning mechanisms and hyper-parameter tuning, large trees have led to over-fitting and randomness in classification.
Having algorithms generated on 4 datasets of varying size has shown that DTP reflects the scalability of a graph database. We can see in Fig. 6 a slight increase in generation time in seconds while maintaining accuracy over a high range of instance size.
## Neo4j Visualization
All the data used in this research were uploaded as homogeneous and unconnected nodes in Neo4j database for decision tree generation, as shown in Fig. 7, which is a visualization of the patients’ data in the dataset 4. All the variables were mapped as node features, with each node representing a patient. Fig. 8Decision Tree for Dataset 4 (split = gain ratio) The red nodes represent the leaf nodes indicating diagnosis of diabetes [2], borderline [1] or no diabetes [0] in a patient, while the blue nodes are the decision nodes. Note that, this tree was generated on a subset of dataset 4 after the class imbalance was handled. There were 13893 instances (4631 for each class label) and 22 variables. The tree has been pruned to max_depth = 2 Apart from the associated convenient methods for data splitting and validation of the classifier, Neo4j allows for an intuitive visualization of the decision tree as shown in Fig. 8. It stores the final decision tree as a set of nodes and its relationship, which can be stored in Neo4J, queried and reused for classification of further data. The visual interface of Neo4j is quite interactive as a user can move the nodes around, focus on any specific node and also pass a single instance for classification through cypher queries.
## Risk factors for developing diabetes
*After* generating several decision tree algorithm in DTP and using feature analysis procedures to compare values of gini index, information gain and gain ratio, we could confidently say that the main risk factors for developing pre-Diabetes and Diabetes were high body mass index and high blood pressure (Fig. 8). This fits to the general statements of the center of disease and control and prevention6 and confirms the potential of using DTP for mining clinical datasets.
## Discussion of research questions
To conclude the evaluation, we will answer the research questions proposed in section Research Questions.
RQ 1—*What is* the difference in the quality of predictions for algorithms generated in DTP, when compared against algorithms generated by standard libraries from Python and R?
Answer—DTP revealed slightly less accuracy as R and Python, but better MCC coefficients. The differences were due to overfitting and could be further diminished by pruning and hyper-parameter tuning.
RQ 2—*What is* the difference in generation time of decision tree algorithms, generated by DTP, compared to standard libraries from Python and R?
Answer—DTP required less computational time than R or python for small datasets, but more time than python for large data sets.
RQ 3—Could applying decision tree learning on homogeneous and unconnected nodes created from large clinical datasets provide basic clinical insights?
Answer—*This is* clearly possible as we have been able to analyze Dataset 4 by generating a decision tree in DTP and state the following—“High body-mass index and blood pressure are primary risk factors for developing diabetes.”
## Conclusion
In this paper, we investigated the feasibility of integrating decision tree learning, applied directly on graph-shaped clinical data in Neo4j. To this end, we implemented a plugin for Neo4j as a set of stored Java procedures that allow to train and persist a decision tree in Neo4j. When it comes to incorporating cross-platform tools, Java packages, though time-consuming to create and refine, can outperform other platforms in accuracy and computational efficiency. As such, also our Neo4j plugin DTP has reached similar performance compared to Python and R.
Being written in Java, DTP could be easily extended and further optimized. However, at the time of compiling this manuscript, python drivers for Neo4j has been released, and we believe integrating a high performance programming language like python might possibly increase the performance of learning algorithms in Neo4j.
The node-edge view of data and the classifier will facilitate the data analysis. To further improve the data mining, researchers can enrich the graph by further biological and clinical metadata. In numerous ways, researchers are trying to incorporate the strengths of graph database into predictive modelling. With the additional advantage of interactive visualization, researchers are turning to graph data for their research to create novel implementations of traditional statistical and machine learning algorithms.
Neo4j is also helpful in the visualization and analyses of clinical data. The node-edge structure is quite effective to visualize patients with several variables which shows promise for further research. This work motivated us to continue research to incorporate all forms of learning algorithms in graph databases—unsupervised, supervised, semi-supervised and representation learning. One such approach to perform link prediction on scholarly data, in Neo4j, has been performed by Sobhgol et al. [ 28] which has provided promising results in accuracy, even more so in the computational efficiency, similar to our results in DTP.
Proposed scope of future research could be integration of learning algorithms using python drivers, post-pruning mechanisms on DTP, implementation of decision tree classification on homogeneous and connected nodes, and/or heterogeneous nodes (for both connected and unconnected) in Neo4j.
## About this supplement
This article has been published as part of BMC Medical Informatics and Decision Making Volume 22 Supplement 6, 2022 Selected articles from the 17th International Conference on Computational Intelligence Methods for Bioinformatics and Biostatistics (CIBB 2021). The full contents of the supplement are available online at https://bmcmedinformdecismak.biomedcentral.com/articles/supplements/volume-22-supplement-6.
## References
1. 1.Santos A, Colaço AR, Nielsen AB, Niu L, Geyer PE, Coscia F, Albrechtsen NJW, Mundt F, Jensen LJ, Mann M. Clinical knowledge graph integrates proteomics data into clinical decision-making. bioRxiv 2020; 10.1101/2020.05.09.084897.
2. Chicco D, Jurman G. **The advantages of the matthews correlation coefficient (mcc) over f1 score and accuracy in binary classification evaluation**. *BMC Genomics* (2020.0). DOI: 10.1186/s12864-019-6413-7
3. Aziz T, Haq E-U, Muhammad D. **Performance based comparison between RDBMS and OODBMS**. *Int J Comput Appl* (2018.0) **180** 42-46. DOI: 10.5120/ijca2018916410
4. 4.Vicknair C, Macias M, Zhao Z, Nan X, Chen Y, Wilkins D. A comparison of a graph database and a relational database. ACM Press, 2010; 10.1145/1900008.1900067.
5. 5.Pokorn J. Graph databases: their power and limitations 2015.
6. 6.Marzi M.D. Dynamic rule based decision trees in Neo4j 2018.
7. 7.Neo4j: User-defined Procedures. https://neo4j.com/docs/java-reference/current/extending-neo4j/procedures-and-functions/procedures/.
8. 8.Michael Hunger R.B, Lyon W. RDBMS and Graphs: SQL vs. Cypher Query Languages 2016.
9. 9.Fernandes D, Bernardino J. Graph databases comparison: Allegrograph, arangodb, infinitegraph, neo4j, and orientdb. In: Proceedings of the 7th international conference on data science, technology and applications. DATA 2018, pp. 373–380. SCITEPRESS—Science and Technology Publications, Lda, 2018; 10.5220/0006910203730380.
10. Kalamaras I, Glykos K, Megalooikonomou V, Votis K, Tzovaras D. **Graph-based visualization of sensitive medical data**. *Multimedia Tools Appl* (2022.0) **81** 209-236. DOI: 10.1007/s11042-021-10990-1
11. Patel H, Prajapati P. **Study and analysis of decision tree based classification algorithms**. *Int J Comput Sci Eng* (2018.0) **6** 74-78
12. 12.Breiman L, Friedman J, Olshen R, Stone C. Cart: classification and regression trees (1984). Belmont, CA: Wadsworth; 1993.
13. 13.Quinlan JR. Induction of decision trees. Machine Learning. 1986;1.
14. 14.Quinlan J.R. Programs for machine learning, 1993.
15. 15.Bramer M. Pre-pruning classification trees to reduce overfitting in noisy domains. In: Yin H, Allinson N, Freeman R, Keane J, Hubbard S editors Intelligent data engineering and automated learning—IDEAL 2002, pp. 7–12. Springer, 2002.
16. Association AD. **Diagnosis and classification of diabetes mellitus**. *Diabetes Care* (2013.0) **37** 81-90. DOI: 10.2337/dc14-S081
17. Chen R, Ovbiagele B, Feng W. **Diabetes and stroke: epidemiology, pathophysiology, pharmaceuticals and outcomes**. *Am J Med Sci* (2016.0) **351** 380-386. DOI: 10.1016/j.amjms.2016.01.011
18. 18.8 Databases supporting in-database machine learning. https://www.infoworld.com/article/3607762/8-databases-supporting-in-database-machine-learning.html.
19. 19.Dynamic Rule Based Decision Trees in Neo4j. https://maxdemarzi.com/2018/01/14/dynamic-rule-based-decision-trees-in-neo4j.
20. 20.Neo4j Machine Learning Procedures. https://github.com/neo4j-contrib/neo4j-ml-procedures.
21. 21.Anjana S, Lavanya K. An application of cypher query-based dynamic rule-based decision tree over suicide statistics dataset with neo4j. In: Intelligent IoT systems in personalized health care, pp. 293–313 2021.
22. Chicco D, Jurman G. **Machine learning can predict survival of patients with heart failure from serum creatinine and ejection fraction alone**. *BMC Med Inf Decis Mak* (2020.0) **20** 1-16
23. Lehmann T, Schallert K, Vilchez-Vargas R, Benndorf D. **Metaproteomics of fecal samples of crohn’s disease and ulcerative colitis**. *J Proteomics* (2019.0) **201** 93-103. DOI: 10.1016/j.jprot.2019.04.009
24. Li W, Ma J, Shende N. **Using machine learning of clinical data to diagnose covid-19: a systematic review and meta-analysis**. *BMC Med Inf Decis Mak* (2020.0). DOI: 10.1186/s12911-020-01266-z
25. 25.Diabetes Health Indicators Dataset. https://www.kaggle.com/alexteboul/diabetes-health-indicators-dataset.
26. 26.Behavioral Risk Factor Surveillance System. https://www.kaggle.com/cdc/behavioral-risk-factor-surveillance-system.
27. Prechelt L. **An empirical comparison of seven programming languages**. *Computer* (2000.0) **33** 23-29. DOI: 10.1109/2.876288
28. 28.Sobhgol S, Durand G, L, R, Saake G. Machine learning within a graph database: A case study on link prediction for scholarly data. In: International conference on enterprise information systems, pp. 159–166 2021.
|
---
title: Estimating immunity with mathematical models for SARS-CoV-2 after COVID-19
vaccination
authors:
- Yoshifumi Uwamino
- Kengo Nagashima
- Ayumi Yoshifuji
- Shigeru Suga
- Mizuho Nagao
- Takao Fujisawa
- Munekazu Ryuzaki
- Yoshiaki Takemoto
- Ho Namkoong
- Masatoshi Wakui
- Hiromichi Matsushita
- Naoki Hasegawa
- Yasunori Sato
- Mitsuru Murata
journal: NPJ Vaccines
year: 2023
pmcid: PMC9988198
doi: 10.1038/s41541-023-00626-w
license: CC BY 4.0
---
# Estimating immunity with mathematical models for SARS-CoV-2 after COVID-19 vaccination
## Abstract
Tools that can be used to estimate antibody waning following COVID-19 vaccinations can facilitate an understanding of the current immune status of the population. In this study, a two-compartment-based mathematical model is formulated to describe the dynamics of the anti-SARS-CoV-2 antibody in healthy adults using serially measured waning antibody concentration data obtained in a prospective cohort study of 673 healthcare providers vaccinated with two doses of BNT162b2 vaccine. The datasets of 165 healthcare providers and 292 elderly patients with or without hemodialysis were used for external validation. Internal validation of the model demonstrated $97.0\%$ accuracy, and external validation of the datasets of healthcare workers, hemodialysis patients, and nondialysis patients demonstrated $98.2\%$, $83.3\%$, and $83.8\%$ accuracy, respectively. The internal and external validations demonstrated that this model also fits the data of various populations with or without underlying illnesses. Furthermore, using this model, we developed a smart device application that can rapidly calculate the timing of negative seroconversion.
## Introduction
Two mRNA vaccines for SARS-CoV-2, BNT162b2 (Pfizer-BioNTech) and mRNA-1273 (Moderna), have demonstrated significant effectiveness after only two doses of vaccine1,2. Since the waning of immunogenicity was reported during the time course3,4, the administration of booster doses was accelerated worldwide. For example, the Center for Disease Control and Prevention (USA) recommended that people who have received two doses of BNT162b2 receive a third dose at least 5 months after receiving the second dose5. The recommended timing of booster dose administration was determined primarily based on the results of studies assessing vaccine effectiveness during the time course6–8. However, conducting population-based vaccine effectiveness studies might be practically difficult in certain countries because the researchers must track the occurrences of infection within a large population of vaccinated people for a long time. Antibody concentrations (titers) and antispike protein immunoglobulin G (IgG) are related to protection against the infection9,10; therefore, measuring the concentration of antibodies in vaccinated people would be helpful in developing public health policies about vaccination and infection control. However, repetitive testing is expensive. Therefore, models for estimating the future dynamics of antibodies are required. In addition, it is difficult for the general population to perceive the waning of immunity following vaccination, which might be one of the reasons for unwillingness to take the booster dose11. The development of digital tools for estimating individual antibody dynamics might lead to a better understanding of waning immunity following vaccination, implying the need for an increased rate of booster vaccination, which remains low in several countries12.
In this study, using the vaccinated cohort data, we develop and validate a mathematical model for estimating antibody waning following vaccination. Hence, we establish a prototype smart phone application to estimate the future waning of antibodies based on a single measurement of a SARS-CoV-2 antibody titer.
## Existing Data
Figure 1 depicts semi-logarithmic plots of antibody titers for healthy medical workers at Keio University Hospital ($$n = 657$$) after two doses of the BNT162b2 vaccine. Figure 1a indicates that the antibody titer increased after two doses and subsequently decreased from 3 to 26 weeks. The rate of decrease was similar from 3 to 13 weeks and slightly lower after 13 weeks. Figure 1b–d indicate the average antibody titer stratified by groups defined by antibody titer at week 3, age, and sex. The characteristics differed for each group, but the rates of decrease were not different. The differences in the antibody titers were the largest when the data were grouped by antibody titer at week 3.Fig. 1Semi-logarithmic plots of SARS-CoV-2 spike antibodies after two doses of BNT162b2 vaccine.a Box plots of all available cases ($$n = 657$$; cases for which data were available only at week zero were excluded); logarithmic means stratified by b antibodies at week 3, c age group, and d sex. Error bars indicate $95\%$ confidence intervals. b The cutoffs for antibodies at week 3 were determined based on the 25th, 50th, and 75th percentiles. All the demonstrated antibody titers were anti-receptor-binding domain IgG.
## Model Selection
We constructed three candidate models with the objective of developing a mathematical model to describe and predict individual antibody titers following two doses: [1] one-compartment model, [2] two-compartment model, and [3] double exponential model. We selected [1] and [2] as the candidate models to represent the elimination of antibodies generated in the body as they transition between compartments. We selected [3] as a nonlinear model to describe the change in the antibody titer (that is, the responses) over time. Further details on the model structures and hyperparameters are provided in the Methods section. We fitted the three candidate models to data obtained from participants at Keio University Hospital ($$n = 657$$) and found that model [2] demonstrated the smallest leave-one-out cross-validation information criterion (LOOIC) and best-fit (see Table 1). This result was consistent with the trend of the rate of decrease, as depicted in Fig. 1 (a constant rate of decrease from 3 to 13 weeks, with a slightly lower rate after 13 weeks). For the best-fit model, the convergence of the Markov chain Monte Carlo samplers was achieved and sufficient quality was demonstrated (see Supplementary Figs. 1–4 and Supplementary Table 2).Table 1Results of model comparison. LOOIC[1] One-compartment model44350.3[2] Two-compartment model38810.6[3] Double exponential model43491.4LOOIC (leave-one-out cross-validation information criterion) is a measure of the goodness of fit of a model, with smaller values indicating a better fit. The two-compartment model LOOIC was the smallest and emphasized, therefore, the best fit. Bold means the best fit with the minimum value.
## Model validation
We applied a further test of the prediction performance of the two-compartment model through 10-fold cross-validation on the dataset from the participants at Keio University Hospital and obtained an accuracy of $97.0\%$ ($95\%$ confidence interval (CI): $95.2\%$–$98.1\%$), RMSE of 0.430, and Pearson’s correlation coefficient of 0.841 (Table 2). We defined accuracy as follows: accuracy = ((no. of patients whose $95\%$ prediction interval included the actual value)/(no. patients)) × 100 [%].Table 2Results of internal and external validations. DataAccuracy ($95\%$ CI)RMSEPearson’s correlation coefficientInternal validationCross-validation: healthy medical workers ($$n = 657$$)$97.0\%$ ($95.2\%$, $98.1\%$)0.4300.841External validationMie National Hospital: healthy medical workers (input data for 3.5 weeks; $$n = 165$$)$98.2\%$ ($94.4\%$, $99.5\%$)0.5080.709External validationMie National Hospital: healthy medical workers (input data for 12 weeks; $$n = 165$$)$94.5\%$ ($89.9\%$, $97.5\%$)0.4840.819External validationJSDT: hemodialysis patients ($$n = 192$$)$83.3\%$ ($77.1\%$, $88.2\%$)0.9240.721External validationJSDT: nondialysis elderly patients ($$n = 100$$)$83.8\%$ ($75.0\%$, $90.3\%$)0.9790.553Accuracy = ((no. of patients whose $95\%$ prediction interval included the actual value)/(no. of patients)) × 100 [%], and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathrm{RMSE}}}}_t = \sqrt {{\textstyle{1 \over n}}\mathop {\sum}\nolimits_{$i = 1$}^n {\left[{\log Y_{it} - \log \left\{ {{{{\mathrm{median}}}}\left({\hat Y_{it}} \right)} \right\}} \right]^2} }$$\end{document}RMSEt=1n∑$i = 1$nlogYit−logmedianY^it2. The RMSE and Pearson’s correlation coefficient between log Yit and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\log \left\{ {{{{\mathrm{median}}}}\left({\hat Y_{it}} \right)} \right\}$$\end{document}logmedianY^it at 26 weeks were indicated for cross-validation, and RMSE and Pearson’s correlation coefficient at 24 weeks were indicated for Mie National Hospital and JSDT.JSDT The Japanese Society for Dialysis Therapy.
Furthermore, we verified the prediction performance of the model using external data. We found that, at 24 weeks, the model demonstrated a prediction accuracy of $98.2\%$ ($95\%$ CI: $94.4\%$–$99.5\%$), RMSE of 0.508, and Pearson’s correlation coefficient of 0.709 for healthcare workers from the Mie National Hospital ($$n = 165$$) with 3.5 weeks of input data; prediction accuracy of $94.5\%$ ($95\%$ CI: $89.9\%$–$97.5\%$), RMSE of 0.484, and Pearson’s correlation coefficient of 0.819 for healthcare workers from the Mie National Hospital ($$n = 165$$) with 12 weeks of input data; prediction accuracy of $83.3\%$ ($95\%$ CI: $77.1\%$–$88.2\%$), RMSE of 0.924, and Pearson’s correlation coefficient of 0.721 for dialysis patients from the Infection Control Committee of the Japanese Society for Dialysis Therapy (JSDT; $$n = 192$$); and prediction accuracy of $83.3\%$ ($95\%$ CI: $75.0\%$–$90.3\%$), RMSE of 0.979, and Pearson’s correlation coefficient of 0.553 for nondialysis elderly patients from JSDT ($$n = 100$$) (Table 2). For the participants in the JSDT cohort whose predictions failed, most of the predicted values were higher than the actual values.
## Prediction results
Figure 2 presents the prediction results from fitting the two-compartment model to three selected participants from Keio University Hospital. We randomly selected participants whose antibody titer increased to ~2500 BAU/mL (Fig. 2a, c) and 3500 BAU/mL (Fig. 2b) after two vaccinations. The predicted results for the participants with higher antibody titers at 3 weeks suggested that they maintained higher antibody titers (Fig. 2a, b). The two-compartment model fitted well, as the change in the rate of decrease in the antibody titers from 3 and 13 weeks and after 13 weeks was accurately modeled (Fig. 2a, b).Fig. 2Prediction results. Prediction results from fitting the best-fit model to three selected participants from Keio University Hospital. The area to the right of the vertical reference line displays the predicted results. Data and predicted results of antibody titers (anti-receptor-binding domain IgG) for a participant whose antibody titer increased to approximately 2500 BAU/mL (a) and 3500 BAU/mL (b) after two vaccination doses. c Data and predicted results for a participant with only one measurement available (week 3) owing to missing data. Time instants are shown when the lower limit of the $95\%$ prediction interval and the median value falls below the 154 BAU/mL threshold, the horizontal reference line, for classification as a protective antibody titer. The vertical axis represents the logarithmic scale.
Because we used a hierarchical Bayes model, we could obtain prediction results for the participants with only one measurement point owing to missing data or a new participant (see Fig. 2c and Supplementary Figs. 5–7). The time instants are indicated when the lower limit of the $95\%$ prediction interval and the median value fall below the 154 BAU/mL threshold for classification as a protective antibody titer (Fig. 2a–c). The subjects with higher antibody titers in the third week crossed the classification threshold later than those with lower antibody titers at the third week. As the participant indicated in Fig. 2c had only one point of data, the width of the prediction interval was wide, reflecting the small amount of information available.
Figure 3 depicts the distribution of the time points at which the lower limit of the $95\%$ prediction interval for each participant fell below the protective classification threshold. Participants with high antibody titers at week 3 had a delayed fall below the protective threshold. This result was consistent with the trend in the group differences indicated in Fig. 1b. Comparing the groups with the lowest and highest antibody titers at week 3, we see that the distribution peak differs by ~25 weeks. Fig. 3Empirical distribution functions of predicted results of time instants at which antibody titers fall below the positive threshold. Empirical distribution functions of the lower limit of the $95\%$ prediction interval for each participant falling below the 154 BAU/mL threshold for classification as a protective antibody titer, stratified by antibodies (anti-receptor-binding domain IgG) at week 3 ($$n = 657$$). The cutoffs for antibodies at week 3 were determined based on the 25th, 50th, and 75th percentiles.
## Development of smart device application
The prototype of an iOS-based smart device application was built using the proposed model. The user was required to input the date of their second BNT162b2 vaccine dose, the date of their antibody titer measurement, the type of reagent used for the antibody titer measurement, and the antibody titer. Subsequently, the antibody dynamic was simulated by the model built into the application, and the estimated date on which the antibody titers would become lower than 154 BAU/mL, which was the protective antibody threshold proposed by Goldblatt et al.13, was displayed (Fig. 4 and Supplementary video). The protective antibody threshold was still to be fixed; it could be variable based on the epidemic variants. Therefore, the administrator could freely adjust the threshold based on advice from public health authorities. Fig. 4Prototype of antibody simulation application using model. Screen image of the iOS application named “COVID Vaccine Navi.” Information about the date of vaccination singly measured antibody titer, and estimated date of negative seroconversion is presented. A demonstration video is available in the supplementary material.
## Discussion
Although mathematical models are commonly used for estimating drug concentration in pharmacodynamics, their application in estimating antibody dynamics following vaccination is rare. Our model successfully simulated the individual waning curves of antibody titers following the administration of mRNA COVID-19 vaccine, which was verified by various datasets. We applied the pharmacodynamic model to understand the dynamics of antibody-titer-induced humoral immunity. This model was originally designed to estimate the concentration of an administered compound or its metabolites. Therefore, it is interesting that it can be used for estimating the antibody titer produced by vaccination, which is neither the administered compound (the vaccine itself) nor its metabolites (RNA). In the generation of antibodies following vaccination, far more complicated processes such as transcription, antigen presentation, and antigen-specific immune cell inductions are included than in the kinetics of other medications, such as antibiotics.
Regarding the application of pharmacodynamic models to antibody dynamics, Favresse et al.14 discussed the suitability of the one-compartment model for antibody waning over 3 months after BNT162b2 vaccination. However, the present study demonstrated better predictability using the two-compartment model as compared with the one-compartment one.
Two-compartment models are often used as pharmacodynamic models that have two different distribution areas, such as plasma and target organs, for medication with various metabolizations. Although the reason that the two-compartment model demonstrated better predictability is unclear, we hypothesized that two different types of IgG production kinetics mimic two different distribution areas, which we named “two immunological compartments.” The first “compartment” refers to IgG production by immature memory B cells stimulated by the second dose of the vaccine. Immature memory B cells specific to SARS-CoV-2 spike proteins are differentiated after the first dose of the vaccine. Following the second dose, the SARS-CoV-2 spike-protein-specific immature memory B cells rapidly produce IgG for SARS-CoV-2 spike proteins with low avidity. These immature memory B cells are not long-lasting; therefore, the antibody titer in this compartment declines rapidly. The second “compartment” is IgG production by plasma cells. Some of the immature memory B cells specific to SARS-CoV-2 spike proteins are selected and matured in the germinal center through the stimulation of antigen-peptide-translated mRNA from the second dose and differentiated into long-term memory and plasma cells. The plasma cells can continuously produce high-avidity IgG specific to SARS-CoV-2 spike proteins without antigen stimulation. The IgG produced by the plasma cells is long-lasting; therefore, the antibody titer in this compartment declines slowly. Although further investigation is essential for validating this hypothesis, the “two immunological compartments” approach, consisting of memory B cells and long-lived plasma cells, might contribute to research on the suitability of two-compartment models15–17. Although simpler antibody kinetics were demonstrated through the mathematical modeling study of antibody kinetics following vaccinia virus vaccine administration when compared with the proposed model18, it is difficult to compare humoral immunity generated by mRNA-based SARS-CoV-2 vaccines and live-attenuated vaccinia virus vaccines, which are considered to induce life-long immunity. In addition, the observation periods in our study were much longer than in the study of the vaccinia virus.
The data obtained in this study suggest that antibody titers differ more considerably between individuals rather than based on factors such as age and gender. We confirmed through internal and external validations that a hierarchical Bayesian model can account for individual differences and be used for predictions with a high accuracy rate. The hierarchical Bayesian model can also predict changes in future antibody titers based on a single point of measurement data.
The datasets used for building the models and for external validation provided by Mie National Hospital consisted mostly of data from healthy young or middle-aged people. However, the JSDT cohort consisted of 192 elderly dialysis patients receiving hemodialysis (HD) and 100 elderly patients with some underlying diseases such as diabetes and hypertension. Although insufficient antibody production was reported among HD, diabetic, and hypertensive patients19–22, our external validation demonstrated as high as $80\%$ accuracy, suggesting that the proposed model is effective, even in groups with underlying diseases.
Our study had three limitations: First, all the participants in the study were administered the BNT162b2 vaccine. Therefore, it is uncertain whether our model can be used for people who were administered other types of COVID-19 vaccines and were younger than the participants in the selected cohorts, that is, children and juveniles. Second, the proposed model could only be validated for up to 26 weeks based on the available data. Therefore, it is unclear how accurate the estimates will be in later time instances. Finally, the model cannot estimate the antibody titer after the administration of the booster dose. As a large number of people are being administered booster doses worldwide, the modification and validation of the proposed model using a dataset of antibodies from people who have received various types of COVID-19 vaccines and those who have received booster doses are warranted.
In conclusion, the anti-SARS-CoV-2 antibody dynamics of healthy adults vaccinated with two doses of BNT162b2 vaccine were described accurately using a mathematical model based on the two-compartment model.
## Antibody-titer datasets for model construction
To construct the mathematical models, we obtained consecutively measured antibody titers from a prospective cohort study of BNT162b2-vaccinated healthcare providers. The study included 673 participants who had received two doses of BNT162b2 vaccine at Keio University Hospital (Tokyo, Japan) (Supplementary Table 1). Five serum samples were collected from each participant before vaccination and then 3 weeks, 8 weeks, 3 months, and 6 mo after the administration of the two doses. The IgG antibody titers for the receptor-binding domain of the SARS-CoV-2 spike proteins were measured using Alinity SARS-CoV-2 IgG II reagents and an Alinity analyzer (Abbott; IL, USA). The study protocol was approved by the Ethics Committee of Keio University School of Medicine (approval No. 20210301), and written informed consent was obtained from all the participants. The measured antibody titers in the original AU/mL units were converted into BAU/mL using the conversion formula 1 BAU/mL = 0.142 AU/mL, following the manufacturer’s instructions.
## Antibody datasets for external validation
To externally validate the model, antibody titer data were obtained from 165 healthy healthcare providers who had received two doses of BNT162b2 vaccine at the National Health Organization Mie National Hospital (Mie, Japan). The IgG antibody titer for SARS-CoV-2 spike proteins of consecutively obtained serum samples was measured using an enzyme-linked immunoassay-based kit (Denka Co. Ltd, Tokyo, Japan), which has been certified by WHO international standard serum samples. Four serum samples were collected from each subject before vaccination and then 3.5 weeks, 3 months, and six 6 months after the administration of the two doses. The study protocol was approved by the Ethics Committee of the National Hospital Organization Mie National Hospital (Approval No. 2021-141), and written informed consent was obtained from all the participants.
In addition, the antibody titers of nondialysis elderly patients ($$n = 100$$) and HD patients ($$n = 192$$) obtained from the JSDT were measured using the Ortho-Clinical Diagnostics VITROS® Anti-SARS-CoV-2 IgG Chemiluminescent Immunoassay, which has been certified by WHO international standard serum samples. Four serum samples were collected from each subject before vaccination and then 2 weeks, 3 months, and 6 months after the administration of the two doses. This study was approved by the Ethics Committee of the JSDT (Approval Nos 1–10), and written informed consent was obtained from all the participants (Supplementary Table 1).
## Model development
To describe and predict individual antibody titers after two doses of BNT162b2 vaccine, we constructed three hierarchical Bayes models: [1] one-compartment model, [2] two-compartment model, and [3] double exponential model. A compartmental model, such as [1] and [2], is a type of differential equation model used to describe how materials transition among the compartments of a system. To represent the elimination of antibodies, we selected [1] and [2] as candidate models of the antibodies transitioning among compartments. Model [3] is a Weibull-type model used to model dose and time responses, and it was selected as a nonlinear model to describe the change in antibody titers (that is, response) over time. We also considered other models, including other Weibull-type models, and age and/or sex as covariates, but these models did not converge as well. We subsequently determined the best-fit model among the three candidate models and evaluated the internal and external validities of the best model.
Let Yit denote the observation vector for the ith subject ($i = 1$, 2, …, n) at time t, where t is the number of weeks following the second vaccine dose ($t = 3$, 8, 13, 26). We used the following hierarchical Bayes models:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\log Y_{it} \sim {{{\mathcal{N}}}}\left({\log f_{\left(j \right)}\left({t\left| {{{{\mathbf{\uptheta }}}}_{ij}} \right.} \right),\sigma _Y^2} \right),$$\end{document}logYit~Nlogfjt𝛉ij,σY2,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta _{ij} \sim {{{\mathcal{N}}}}\left({{{{\mathbf{\upmu }}}}_j,{{{\mathbf{{\Sigma}}}}}_j} \right),$$\end{document}θij~N𝛍j,Σj,where f(j) denotes a nonlinear regression function and θij represents the parameter vector of the ith subject, which specifies the nonlinear regression function. We define f(j) and θij for each model as follows:One-compartment model:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f_{\left(1 \right)}\left({t\left| {{{{\mathbf{\uptheta }}}}_{i1}} \right.} \right) = \exp \left({a_{i1}} \right)\exp \left({ - b_{i1}t} \right),$$\end{document}f1t𝛉i1=expai1exp−bi1t,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{\uptheta }}}}_{i1} = \left({a_{i1},b_{i1}} \right),$$\end{document}𝛉i1=ai1,bi1,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{\upmu }}}}_1 = \left({\mu _a,\mu _b} \right)^T,$$\end{document}𝛍1=μa,μbT,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{{\Sigma}}}}}_1 = \left({\begin{array}{*{20}{c}} {\sigma _a^2} & {\rho \sigma _a\sigma _b} \\ {\rho \sigma _a\sigma _b} & {\sigma _b^2} \end{array}} \right);$$\end{document}Σ1=σa2ρσaσbρσaσbσb2;Two-compartment model:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f_{\left(2 \right)}\left({t\left| {{{{\mathbf{\uptheta }}}}_{i2}} \right.} \right) = \exp \left({a_{i2}} \right)\exp \left({ - b_{i2}t} \right) + \exp \left({c_{i2}} \right)\exp \left({ - d_{i2}t} \right),$$\end{document}f2t𝛉i2=expai2exp−bi2t+expci2exp−di2t,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{\uptheta }}}}_{i2} = \left({a_{i2},b_{i2},c_{i2},d_{i2}} \right),$$\end{document}𝛉i2=ai2,bi2,ci2,di2,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{\upmu }}}}_2 = \left({\mu _a,\mu _b,\mu _c,\mu _d} \right)^T,$$\end{document}𝛍2=μa,μb,μc,μdT,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{{\Sigma}}}}}_2 = \left({\begin{array}{*{20}{c}} {{{{\mathbf{{\Sigma}}}}}_{21}} & {{{\mathbf{0}}}} \\ {{{\mathbf{0}}}} & {{{{\mathbf{{\Sigma}}}}}_{22}} \end{array}} \right) = \left({\begin{array}{*{20}{c}} {\sigma _a^2} & {\rho _1\sigma _a\sigma _b} & 0 & 0 \\ {\rho _1\sigma _a\sigma _b} & {\sigma _b^2} & 0 & 0 \\ 0 & 0 & {\sigma _c^2} & {\rho _2\sigma _c\sigma _d} \\ 0 & 0 & {\rho _2\sigma _c\sigma _d} & {\sigma _d^2} \end{array}} \right);$$\end{document}Σ2=Σ2100Σ22=σa2ρ1σaσb00ρ1σaσbσb20000σc2ρ2σcσd00ρ2σcσdσd2;Double exponential model:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f_{\left(3 \right)}\left({t\left| {{{{\mathbf{\uptheta }}}}_{i3}} \right.} \right) = \exp \left\{ {\exp \left({a_{i3}} \right)\exp \left({ - b_{i3}t} \right)} \right\} - 1,$$\end{document}f3t𝛉i3=expexpai3exp−bi3t−1,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{\uptheta }}}}_{i3} = \left({a_{i3},b_{i3}} \right),$$\end{document}𝛉i3=ai3,bi3,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{\upmu }}}}_3 = \left({\mu _a,\mu _b} \right)^T,$$\end{document}𝛍3=μa,μbT,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{{\Sigma}}}}}_3 = \left({\begin{array}{*{20}{c}} {\sigma _a^2} & {\rho \sigma _a\sigma _b} \\ {\rho \sigma _a\sigma _b} & {\sigma _b^2} \end{array}} \right);$$\end{document}Σ3=σa2ρσaσbρσaσbσb2; We used weak informative priors as follows: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mu _a,\mu _b,\mu _c,\mu _d \sim {{{\mathcal{N}}}}\left({0,100^2} \right)$$\end{document}μa,μb,μc,μd~N0,1002, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sigma _a,\sigma _b,\sigma _c,\sigma _d \sim {{{\mathrm{HalfCauchy}}}}\left({0,50} \right)$$\end{document}σa,σb,σc,σd~HalfCauchy0,50, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathbf{{\Sigma}}}}}_1,{{{\mathbf{{\Sigma}}}}}_{21},{{{\mathbf{{\Sigma}}}}}_{22},{{{\mathbf{{\Sigma}}}}}_3 \sim {{{\mathrm{LJKCorr}}}}\left(1 \right)$$\end{document}Σ1,Σ21,Σ22,Σ3~LJKCorr1. We assumed that there was no correlation between the various compartments.
We used models that consider individual differences through a hierarchical structure for describing and predicting individual profiles.
The four candidate models were fitted using Stan in R, software version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria)23. We used a Hamiltonian Monte Carlo algorithm to generate samples from the posterior distributions of the parameters. We then evaluated the sampling convergence using trace plots and the Gelman–Rubin statistic, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat R$$\end{document}R^24, which was confirmed to be >1.01 for all parameters. The predictive performances of the models were compared using the LOOIC25. Fig. 5 depicts the best-fit model, that is, the two-compartment model. We estimated the posterior predictive distributions for the prediction of antibody titers with $95\%$ prediction intervals for each time instant. We derived the time instants at which each participant’s lower limit of the $95\%$ prediction interval falls below the 154 BAU/mL threshold for classification as a positive sample. Fig. 5Diagram of the finally fitted two-compartment model. Each parameter is defined as follows: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k_{i21} = {\textstyle{{A_{i2}d_{i2} + C_{i2}b_{i2}} \over {A_{i2} + C_{i2}}}}$$\end{document}ki21=Ai2di2+Ci2bi2Ai2+Ci2, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A_{i2} = \exp \left({a_{i2}} \right)$$\end{document}Ai2=expai2, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C_{i2} = \exp \left({c_{i2}} \right)$$\end{document}Ci2=expci2, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k_{i10} = {\textstyle{{b_{i2}d_{i2}} \over {k_{i21}}}}$$\end{document}ki10=bi2di2ki21, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k_{i12} = b_{i2} + d_{i2} - k_{i21} - k_{i10}$$\end{document}ki12=bi2+di2−ki21−ki10.
## Internal validation
We used grouped 10-fold cross-validation for prediction to assess the internal validity of the best-fit model described above. First, we divided the dataset into 10 equal parts and used nine to train and one to test. Thereafter, we randomly sampled one of the 3-, 8-, and 13-week measurements from each subject in the test data and used these as inputs to obtain $95\%$ prediction intervals for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat Y_{it}$$\end{document}Y^it at 26 weeks. We obtained prediction intervals by fitting a best-fit model using the nine parts of the training data. We then assessed whether these $95\%$ prediction intervals included the actual values of Yit, root mean squared errors (RMSEs), and Pearson’s correlation coefficient between log Yit and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\log \left\{ {{{{\mathrm{median}}}}\left({\hat Y_{it}} \right)} \right\}$$\end{document}logmedianY^it at 26 weeks. We repeated this process, in which each of the ten parts participated in the test once. We defined the accuracy and RMSE at a time t as follows: accuracy = ((no. of patients whose $95\%$ prediction interval included the actual value)/(no. patients)) × 100 [%], and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathrm{RMSE}}}}_t = \sqrt {{\textstyle{1 \over n}}\mathop {\sum}\nolimits_{$i = 1$}^n {\left[{\log Y_{it} - \log \left\{ {{{{\mathrm{median}}}}\left({\hat Y_{it}} \right)} \right\}} \right]^2} }$$\end{document}RMSEt=1n∑$i = 1$nlogYit−logmedianY^it2.
## External validation
We assessed the external validity using three datasets (Mie National Hospital workers, $$n = 165$$; JSDT dialysis patients, $$n = 192$$; JSDT nondialysis elderly patients, $$n = 100$$), which were different from those used in the development of the best-fit model. We used different assay kits for different target populations and evaluated these to validate the generalizability of the best-fit model. Mie National *Hospital is* located in western Japan, where the SARS-CoV-2 infection rates were low. The participants from Mie National Hospital were healthy medical workers. Participants from JSDT were dialysis and nondialysis elderly patients; therefore, a different assay kit from that used for the Mie National Hospital participants was used.
For the Mie National Hospital data, we used measurements for each subject at 3.5 or 12 weeks as inputs to obtain $95\%$ prediction intervals for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat Y_{it}$$\end{document}Y^it at 24 weeks. For JSDT, we used the measurements for each subject at two weeks as inputs to obtain $95\%$ prediction intervals for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat Y_{it}$$\end{document}Y^it at 24 weeks. We performed external validation on three populations, that is, Mie Hospital workers, dialysis patients at JSDT, and nondialysis elderly patients at JSDT. We assessed whether the $95\%$ prediction intervals included the actual values of Yit, RMSEs, and Pearson’s correlation coefficient at 24 weeks.
## Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
## Supplementary information
Supplement Information Video REPORTING SUMMARY The online version contains supplementary material available at 10.1038/s41541-023-00626-w.
## References
1. Dagan N. **BNT162b2 mRNA Covid-19 vaccine in a nationwide mass vaccination setting**. *N. Engl. J. Med.* (2021.0) **384** 1412-1423. DOI: 10.1056/NEJMoa2101765
2. Baden LR. **Efficacy and safety of the mRNA-1273 SARS-CoV-2 vaccine**. *N. Engl. J. Med.* (2021.0) **384** 403-416. DOI: 10.1056/NEJMoa2035389
3. Tré-Hardy M. **Immunogenicity of mRNA-1273 COVID vaccine after 6 months surveillance in health care workers; a third dose is necessary**. *J. Infect.* (2021.0) **83** 559-564. DOI: 10.1016/j.jinf.2021.08.031
4. Levin EG. **Waning immune humoral response to BNT162b2 Covid-19 vaccine over 6 months**. *N. Engl. J. Med.* (2021.0) **385** e84. DOI: 10.1056/NEJMoa2114583
5. 5.Interim CDC. COVID-19 immunization schedule for ages 5 years and older. https://www.cdc.gov/vaccines/covid-19/downloads/COVID-19-immunization-schedule-ages-5yrs-older.pdf (2022).
6. Polack FP. **Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine**. *N. Engl. J. Med.* (2020.0) **383** 2603-2615. DOI: 10.1056/NEJMoa2034577
7. Chemaitelly H. **Waning of BNT162b2 vaccine protection against SARS-CoV-2 infection in Qatar**. *N. Engl. J. Med.* (2021.0) **385** e83. DOI: 10.1056/NEJMoa2114114
8. Goldberg Y. **Waning immunity after the BNT162b2 vaccine in Israel**. *N. Engl. J. Med.* (2021.0) **385** e85. DOI: 10.1056/NEJMoa2114228
9. Dimeglio C, Herin F, Martin-Blondel G, Miedougé M, Izopet J. **Antibody titers and protection against a SARS-CoV-2 infection**. *J. Infect.* (2022.0) **84** 248-288. DOI: 10.1016/j.jinf.2021.09.013
10. Bergwerk M. **Covid-19 breakthrough infections in vaccinated health care workers**. *N. Engl. J. Med.* (2021.0) **385** 1474-1484. DOI: 10.1056/NEJMoa2109072
11. Al-Qerem W, Al Bawab AQ, Hammad A, Ling J, Alasmari F. **Willingness of the Jordanian population to receive a COVID-19 booster dose: a cross-sectional study**. *Vaccines (Basel)* (2022.0) **10** 410. DOI: 10.3390/vaccines10030410
12. 12.Our World_in_Data. COVID-19 Vaccine Boosters Administered per 100 People. https://ourworldindata.org/ (2022).
13. Goldblatt D. **Towards a population-based threshold of protection for COVID-19 vaccines**. *Vaccine* (2022.0) **40** 306-315. DOI: 10.1016/j.vaccine.2021.12.006
14. Favresse J. **Antibody titres decline 3-month post-vaccination with BNT162b2**. *Emerg. Microbes Infect.* (2021.0) **10** 1495-1498. DOI: 10.1080/22221751.2021.1953403
15. Dhenni R, Phan TG. **The geography of memory B cell reactivation in vaccine-induced immunity and in autoimmune disease relapses**. *Immunol. Rev.* (2020.0) **296** 62-86. DOI: 10.1111/imr.12862
16. Turner JS. **SARS-CoV-2 mRNA vaccines induce persistent human germinal centre responses**. *Nature* (2021.0) **596** 109-113. DOI: 10.1038/s41586-021-03738-2
17. Pradenas E. **Stable neutralizing antibody levels 6 months after mild and severe COVID-19 episodes**. *Med (NY)* (2021.0) **2** 313-320.e4
18. Le D, Miller JD, Ganusov VV. **Mathematical modeling provides kinetic details of the human immune response to vaccination**. *Front. Cell Infect. Microbiol* (2015.0) **4** 177. DOI: 10.3389/fcimb.2014.00177
19. Jahn M. **Humoral response to SARS-CoV-2-Vaccination with BNT162b2 (Pfizer-BioNTech) in patients on hemodialysis**. *Vaccines (Basel)* (2021.0) **9** 360. DOI: 10.3390/vaccines9040360
20. Kato S. **Aspects of immune dysfunction in end-stage renal disease**. *Clin. J. Am. Soc. Nephrol.* (2008.0) **3** 1526-1533. DOI: 10.2215/CJN.00950208
21. Ali H. **Robust antibody levels in both diabetic and non-diabetic individuals after BNT162b2 mRNA COVID-19 vaccination**. *Front. Immunol.* (2021.0) **12** 752233. DOI: 10.3389/fimmu.2021.752233
22. Watanabe M. **Central obesity, smoking habit, and hypertension are associated with lower antibody titres in response to COVID-19 mRNA vaccine**. *Diabetes Metab. Res. Rev.* (2022.0) **38** e3465. DOI: 10.1002/dmrr.3465
23. Carpenter B. **Stan: a probabilistic programming language**. *J. Stat. Softw.* (2017.0) **76** 1-32. DOI: 10.18637/jss.v076.i01
24. Gelman A, Rubin DB. **Inference from iterative simulation using multiple sequences**. *Stat. Sci.* (1992.0) **7** 457-472. DOI: 10.1214/ss/1177011136
25. Vehtari A, Gelman A, Gabry J. **Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC**. *Stat. Comput.* (2017.0) **27** 1413-1432. DOI: 10.1007/s11222-016-9696-4
|
---
title: Anxiety and Stress Related to COVID-19 Among the Community Dwelling Older Adults
Residing in the Largest Refugee Camp of the World
authors:
- Afsana Anwar
- Uday Narayan Yadav
- Md. Nazmul Huda
- Sukanta Das
- Simon Rosenbaum
- A. R. M. Mehrab Ali
- Probal Kumar Mondal
- Abu Ansar Md. Rizwan
- Syed Far Abid Hossain
- Suvasish Das Shuvo
- Sabuj Kanti Mistry
journal: Community Mental Health Journal
year: 2023
pmcid: PMC9988202
doi: 10.1007/s10597-023-01101-5
license: CC BY 4.0
---
# Anxiety and Stress Related to COVID-19 Among the Community Dwelling Older Adults Residing in the Largest Refugee Camp of the World
## Abstract
The current cross-sectional study was conducted among 864 older adults aged ≥ 60 years residing in Rohingya refugee camp through face-to-face interviews during November–December 2021. COVID-19-related anxiety was measured using the five-point Coronavirus Anxiety Scale (CAS) and perceived stress using the 10-point Perceived Stress Scale (PSS). The linear regression model identified the factors associated with COVID-19-related anxiety and perceived stress. The prevalence of COVID-19-related anxiety and perceived stress was $68\%$ and $93\%$, respectively. The average COVID-19-related anxiety score expected to be significantly higher among those who were physically inactive, concerned about COVID-19, had a close friend/family member diagnosed with COVID-19, and had some difficulty in getting food and routine medical care during the COVID-19 pandemic. Meanwhile, the average perceived stress score was expected to be significantly higher among those without partners, who were feeling overwhelmed by COVID-19, and who experienced COVID-19-related anxiety during the pandemic. The findings suggest providing immediate psychosocial support to older Rohingya adults.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s10597-023-01101-5.
## Introduction
The coronavirus disease 2019 (COVID-19) pandemic and accompanying control measures created major interruptions to social norms and public health practices worldwide (Chadi et al., 2022). According to the World Health Organization (WHO), as of 17 November 2022, over 632 million confirmed cases of COVID-19 had been identified, and over 6.5 million deaths had occurred due to the pandemic (World Health Organization, 2022c). Evidence shows that COVID-19 pandemic not only hampered the physical health of the diseased, but also affected the mental well-being of the patients and the community as a whole (Knolle et al., 2021; Pfefferbaum & North, 2020). This has been a particular concern for the vulnerable populations such as older adults, refugees and people with long term conditions (Yadav et al., 2020).
Emerging evidence during the COVID-19 pandemic shows that mental health outcomes, including anxiety, stress and depression, affected almost everybody globally (Mistry et al., 2022; Robinson et al., 2022; Xiong et al., 2020). The COVID-19 pandemic exerted a detrimental psychological impact at both individual and community levels (Salari et al., 2020). The WHO has stated that the global prevalence of anxiety has increased by $25\%$ during the COVID-19 pandemic (World Health Organization, 2022b). Recent systematic reviews also reported that the pooled prevalence of anxiety and stress to be $31.9\%$ and $29.6\%$ respectively in general population amid COVID-19 pandemic (Mahmud et al., 2022; Salari et al., 2020). Unplanned lockdowns and difficulties in accessing essential services, including food, health care and medication, as well as social distancing and isolation as preventive measures have contributed to increased anxiety and stress during the pandemic (Cheruvu & Chiyaka, 2019; Tausch et al., 2022). Notably, higher COVID-19 related anxiety also found to be positively associated with increased stress during the COVID-19 pandemic (Hu et al., 2021). While everyone is at risk of adverse mental health conditions during the pandemic, older adults are particularly vulnerable (Vahia et al., 2020; Webb & Chen, 2022). Pre-existing medical health conditions (e.g. respiratory diseases, hypertension, chronic kidney diseases, and obesity), limited financial and social support, limited access to resources and emergency services make older people more vulnerable to developing mental illnesses during the COVID-19 pandemic (Lee et al., 2020).
Worsening of mental health conditions during the COVID-19 pandemic can be a particular concern among the displaced and refugee population, partly because of their unsuitable living condition, lack of access to health services, material deprivation, isolation, and uncertainty in their life (Anwar et al., 2022; Júnior et al., 2020; Singh et al., 2018; Spiritus-Beerden et al., 2021). Evidence showed that the prevalence of mental health problems among refugee population varied from $20\%$ to $80\%$ globally (Júnior et al., 2020). The Rohingyas are a Muslim minority from Myanmar, mainly seek refuge in Bangladesh since the seventies due to denial of citizenship from the concerned authority from Myanmar, but have also fled to other countries such as Saudi Arabia, Pakistan and Malaysia (Alam 2019). A total of 943,529 people are currently living in Rohingya camp in Cox’s Bazar, a Southern district of Bangladesh, and is the largest refugee camp in the world (United Nations High Commissioner for Refugees, 2022). Older adults comprise a significant portion of the Rohingya community and recent records suggest that older people comprise $3.6\%$ of the total population residing in the camp (United Nations High Commissioner for Refugees, 2022). Rohingya camp is densely populated, with 40,000 people living per square kilometer and lack access to basic amenities such as safe water, satiation and health facilities (Kamal et al., 2020). COVID-19 has also been a particular concern in the Rohingya camp as there were 6,793 confirmed COVID-19 cases and 45 deaths till 17 November 2022 (World Health Organization, 2022a).
Age being a crucial factor for emotional distress, the impact of COVID-19 on aggravating anxiety and stress among the older adults residing in the camps can be serious. A longitudinal study conducted on the Rohingya population showed that the stress level was significantly increased during the COVID-19 pandemic compared to that of the pre-pandemic rate (Palit et al., 2022). However, the participants of this study were younger adults. Another study conducted among older adults aged 60 years and above residing in the Rohingya refugee camp demonstrated a high prevalence of depressive symptoms during the COVID-19 pandemic (Mistry et al., 2021b). However, to the best of our knowledge, no study has explored the prevalence of anxiety and stress during the pandemic among the older adults living in the Rohingya camp in Bangladesh. Therefore, the current study aimed to (i) examine the prevalence of COVID-19-related anxiety and perceived stress among the community dwelling older Rohingya adults and (ii) identify the factors associated with COVID-19-related anxiety and perceived stress among them.
## Study Design and Participants
This study followed a cross-sectional design conducted between November and December 2021. Participants were older adults aged 60 years and above residing in five purposively selected sub-camps of Rohingya refugee camp in Bangladesh where Social Assistance and Rehabilitation for the Physically Vulnerable (SARPV) is currently working.
Considering an unknown prevalence of anxiety/stress with a $5\%$ margin of error, $95\%$ level of confidence, $80\%$ power of the test and $25\%$ non-response rate, the required sample size was calculated as 973. Finally, a total of 864 participants consented to take part in the study from the approached 973 participants (response rate $88.8\%$). The SARPV had a list of all participants residing in the five selected sub-camps (of the 34 sub-camps in the Rohingya camp) which were used as the sampling frame for the study. A simple random sampling technique (computer generated numbers) was used to select the required number of participants. Age of the beneficiaries were verified using SMART card provided by UNHCR containing all relevant information. The inclusion criteria were aged 60 years or above and residents of Rohingya camps. The exclusion criteria were presence of any adverse mental conditions (clinically diagnosed schizophrenia, bipolar mood disorder, dementia/cognitive impairment), a hearing disability, or an inability to communicate.
## Outcome Measure
COVID-19-related anxiety was measured using the Bengali version of the five-point Coronavirus Anxiety Scale (CAS) (Ahmed et al., 2020), which was translated to Rakhine language (commonly used by Rohingya people). Participants were asked about their level of COVID-19-related anxiety, experienced in the last two weeks before the survey using the CAS and their agreement/disagreement with five CAS items were assessed using a five-point Likert Scale. Therefore, the cumulative score ranged from 0 to 20, where the higher the scores, the greater the anxiety of COVID-19. We further classified the participants as having COVID-19-related anxiety (if they reported having anxiety in any one of the CAS items) or not having COVID-19-related anxiety (if they reported they have had no anxiety in any CAS item). We found the reliability of the scale among the participants acceptable (Cronbach’s α = 0.84).
Similarly, perceived stress was measured using the 10-items Perceived Stress Scale (PSS), which was previously validated among Bangladeshi population (Islam, 2020). This tool was translated to the Rakhine language and participants were asked if they perceived stress in the last month preceding the survey using the PSS-10 and their agreement/disagreement with the ten items were assessed using a five-point Likert Scale. Therefore, the cumulative score ranged from 0 to 40, where higher scores indicated greater levels of perceived stress. We further classified the participants as having perceived stress (if they reported having stress in any one of the PSS items) or not having perceived stress (if they reported they have had no stress in any PSS item). Reliability of the scale was acceptable (Cronbach’s α = 0.79).
## Explanatory Variables
Explanatory variables considered in this study were selected based on extensive literature review (Huda et al., 2021; Lou et al., 2012; Mistry et al., 2021b; Perez et al., 2001; Renner et al., 2021; Stubbs et al., 2017; Tinghög et al., 2017). We considered age (categorized as 60–69, and ≥ 70), sex (male/female), marital status (married/without partner), formal education (yes/no), household size (≤ 4 or > 4), monthly family income in Bangladeshi Taka (BDT) where 1 USD ~ 90 BDT (Living on aid alone, have some additional income, current occupation (employed/unemployed or retired), living arrangements (living alone or with family), walking distance to the nearest health centre (< 30 min/ ≥ 30 min), currently suffering from any non-communicable chronic diseases (NCDs) (yes/no), level of physical activity (regular at least 2–4 h per week/none or sedentary), feeling concerned about COVID-19 (hardly, sometimes/often), feeling overwhelmed by COVID-19 (hardly, sometimes/often), close friend or family member previously diagnosed with COVID-19 (no or not sure/yes), frequency of communication with friends and family during COVID-19 (less than previous/same as previous), difficulty in obtaining food, earning money and getting routine medical care during COVID-19 (no/yes) as explanatory variables in the study.
The median household size of the Rohingya population residing in the camp is 4.0 (Bhatia et al., 2018). Therefore, we categorized the household size as ≤ 4 or > 4. Self-reported information on sufferings from chronic conditions (e.g., arthritis, hypertension, heart diseases, stroke, hypercholesterolemia, diabetes, chronic respiratory diseases, chronic kidney disease, and cancer) was collected. Thereafter, a new variable was created, “currently suffering from any non-communicable chronic diseases (NCDs)” which was categorized as “No” if they did not have any of these diseases and “Yes” if they had at least one of these diseases. In line with the previous literature, we categorised the walking distance to the nearest health centre as less than 30 min and equal to or more than 30 min (Mistry et al., 2021b).
## Data Collection Tools and Techniques
A pre-tested semi-structured questionnaire in the Rakhine language was used to collect the information through face-to-face interviews. Data were electronically recorded in SurveyCTO mobile app (https://www.surveycto.com/) by two enumerators, who were fluent in Rakhine dialects. They had previous experience in administering health surveys using electronic data collection platforms. The enumerators were trained for three days before commencing data collection, including on procedures of maintaining COVID-19 safe behaviours during the data collection.
The Bengali version of the questionnaire was first translated to Rakhine dialects and then back-translated to Bengali by two staff of SARPV who understand both Bengali and Rakhine language. The Rakhine version of the tool was piloted among a small sample ($$n = 10$$) of older Rohingya adults from the selected camps to refine the language in the final version. The participants approved the tool translated into Rakhine language without any corrections or modifications. Data collection was accomplished using this final tool through face-to-face interviews of participants. Each interview took approximately 30 min.
## Statistical Analysis
The distribution of the variables was assessed through descriptive statistics. Variables were checked for missing values, and none of the outcome and explanatory variables had any missing values. Two separate multiple linear regression models were performed to explore the factors associated with COVID-19-related anxiety and perceived stress among the participants. All independent variables were examined for multicollinearity before entering them in the regression analysis, and no significant multicollinearity between any independent variables was identified. We have also checked whether the dependent variables met the assumptions for linear regression and found that they met the assumptions including that they are normally distributed. Thereafter, a backward elimination criterion with the Akaike Information Criterion (AIC) was employed to select the final model. Briefly, the backward elimination algorithm starts with a full model (model with all variables) and drops one by one variable from the model based on the statistical significance of that variable. In this case, the adjusted beta coefficient, p-value, and $95\%$ confidence interval ($95\%$ CI) for the final multiple linear regression analysis model are reported in the main table, and the multicollinearity diagnostics results are presented in a separate table. All analyses were performed using the statistical software package Stata (Version 17.0).
## Characteristics of the Participants
A total of 864 participants aged 60 years and over participated in this study. Table 1 shows the socio-demographic characteristics of the participants as well as their perceived opinion on COVID-19-related information. More than half of the participants ($57\%$) were males. The majority of the participants were aged 60–69 years ($72\%$), currently married ($79\%$), had no formal schooling ($89\%$), lived with family members or others ($91\%$), was living on aid alone ($67\%$), and were currently unemployed or retired ($89\%$). Around half of the participants had large household size with more than 4 members ($57\%$) and were currently suffering from any non-communicable chronic diseases (NCDs) ($50\%$). Nearly one-third of the participants ($31\%$) resided more than 30 min walking distance from the nearest health center, and more than two-thirds ($66\%$) did not engage in regular physical activity. In terms of perceived opinion on COVID-19-related information, most participants were somewhat to very concerned ($80\%$) and felt overwhelmed ($78\%$) by COVID-19. About $8\%$ of the participants’ close friends or family members were previously diagnosed with COVID-19. More than a quarter of the participants ($29\%$) reported that the frequency of communication during COVID-19 was less often than previous. Most participants reported that they were facing some difficulty in accessing food ($81\%$), earning money ($90\%$), and getting routine medical care ($73\%$) during the COVID-19 pandemic. Table 1Characteristics of the participants ($$n = 864$$)Characteristicsn%Age (year) 60–6962572 > = 7023928Sex Male48656 Female37844Marital status Currently Married68379 Without partner18121Formal schooling No formal schooling76989 Having formal schooling9511Household size ≤ 437243 > 449257Living arrangement Living with family members/others78290 Living alone8210Family monthly income (BDT) Living on aid58067 Have some additional income*22033Current occupation Employed9411 Unemployed/retired77089Walking distance to the nearest health centre < 30 min59269 ≥ 30 min27231Currently suffering from any non-communicable chronic diseases (NCDs) No43150 Yes43350Level of physical activity Regular at least 2–4 h in a week29534 None/Sedentary56966Feeling concerned about COVID-19 Not concerned16920 Somewhat to very concern69580Feeling overwhelmed by COVID-19 No19022 Yes67478Close friend or family member previously diagnosed with COVID-19 No/Not sure79592 Yes698Frequency of communication during COVID-19 More than or same as previous61371 Less often than previous25129Difficulty of obtaining food during COVID-19 No difficulty16319 Some difficulty70181Difficulty in earning money during COVID-19 No difficulty839 Some difficulty78191Difficulty of getting routine medical care during COVID-19 No difficulty23427 Some difficulty63073*Those who had some additional income other than aid, all of them earned lower than the World Bank defined lower poverty line of USD 3.20 a day
## Prevalence of COVID-19-Related Anxiety and Perceived Stress
Data from the Coronavirus Anxiety Scale (CAS) are presented in Table 2, and the prevalence of perceived stress on Perceived Stress Scale (PSS) is presented in Table 3. The overall prevalence of COVID-19-related anxiety was found to be $68\%$ among the participants. The individual percentage of people who were affirmative to the different items of the CAS was around $50\%$ in each item except that of the last item which was a bit lower ($42\%$). Meanwhile, the overall prevalence of perceived stress was found to be $93\%$ and individual percentage of people who were affirmative to different items of the PSS was more than $70\%$ in each item (Table 3). However, no significant difference was observed in percentage of participants having perceived stress in terms of having COVID-19 anxiety or not ($94\%$ versus $92\%$) (data not shown).Table 2Prevalence of COVID-19-related anxiety ($$n = 864$$)n%I felt dizzy, lightheaded, or faint, when I read or listened to news about the coronavirus43550I had trouble falling or staying asleep because I was thinking about the coronavirus47755I felt paralyzed or frozen when I thought about or was exposed to information about the coronavirus43851I lost interest in eating when I thought about or was exposed to information about the coronavirus44552I felt nauseous or had stomach problems when I thought about or was exposed to information about the coronavirus35942Overall prevalence of anxiety59168Table 3Prevalence of perceived stress ($$n = 864$$)n%In the last month, how often have you been upset because of something that happed unexpectedly?62272In the last month, how often have you felt that you were unable to control the important events in your life?63574In the last month, how often have you felt nervous and stressed?62973In the last month, how often have you felt confident about your ability to handle personal problems?65976In the last month, how often have you felt that things were going your way?61371In the last month, how often have you found that you could not cope with all the things that you should do?65576In the last month, how often have you been able to control irritations in your life?65776In the last month, how often have you felt that you were on top of things?64675In the last month, how often have you been angered because of things that happen and were uncontrolled?60970In the last month, how often have you felt difficulties were piling up so high that you cannot overcome?64875Overall prevalence of stress80493
## Factors Associated with COVID-19-related Anxiety and Perceived Stress
We performed two separate multiple linear regression models to explore the factors associated with COVID-19-related anxiety and perceived stress and results are presented in Table 4 and Table 5. The model multicollinearity diagnosis results for COVID-19-related anxiety and perceived stress are presented separately in the supplementary Tables 1, 2, respectively. VIF values less than 10 for each variable indicate the absence of multicollinearity. Table 4Factors associated with COVID-19-related anxiety among the participants ($$n = 864$$)Characteristicsβ1P$95\%$CISex MaleRef Female0.480.139-0.16, 1.11Marital status Currently marriedRef Without partner0.790.0370.05, 1.53Formal schooling No formal schoolingRef Having formal schooling0.840.140-0.28, 1.95Household size ≤ 4Ref > 4-0.550.071-1.15, 0.05Current occupation EmployedRef Unemployed/retired-2.36 < 0.001-3.47, -1.24Level of physical activity Regular at least 2–4 h per weekRef None/Sedentary1.31 < 0.0010.65, 1.97Feeling concerned about COVID-19 HardlyRef Sometimes/often1.89 < 0.0011.22, 2.57Close friend or family member previously diagnosed with COVID-19 No/Not sureRef Yes2.44 < 0.0011.17, 3.72Difficulty in getting food during COVID-19 No difficultyRef Some difficulty1.56 < 0.0010.74, 2.39Difficulty of getting routine medical care during COVID-19 No difficultyRef Some difficulty2.66 < 0.0011.88, 3.441We performed a multiple linear regression model to explore the factors associated with COVID-19-related anxiety; adjusted models included all the variables listed in Table 4R2: 0.29Table 5Factors associated with perceived stress among the participants ($$n = 864$$)Characteristicsβ1P$95\%$CIMarital status Currently marriedRef Without partner1.640.0050.50, 2.78Household size ≤ 4Ref > 4− 0.720.127− 1.65, 0.21Current occupation EmployedRef Unemployed/retired− 2.380.001− 3,83, -0.92Feeling overwhelmed by COVID-19 HardlyRef Sometimes/often2.20 < 0.0011.05, 3.36Frequency of communication during COVID-19 More than or same as previousRef Less often than previous− 2.81 < 0.001− 3.88, − 1.75Difficulty in getting food during COVID-19 No difficultyRef Some difficulty1.120.158− 0.43, 2.66Experiencing COVID-19-related anxiety NoRef Yes2.150.0010.85, 3.451We performed a multiple linear regression model to explore the factors associated with perceived stress; adjusted models included all the variables listed in Table 5R2: 0.33 The factors associated with COVID-19-related anxiety reveled in the adjusted regression model are presented in Table 4. The average COVID-19-related anxiety score was expected to be significantly higher (p-value < 0.001) among participants who did not engage in regular physical activity, compared to those who were engaged in at least 2–4 h per week of regular physical activity. Similarly, the average COVID-19-related anxiety score was expected to be significantly higher among the participants who were sometimes or very often concerned about COVID-19 (p-value < 0.001), whose close friends or family members were previously diagnosed with COVID-19 (p-value < 0.001), who had some difficulty in getting food during COVID-19 (p-value < 0.001), and who felt some difficulty of getting routine medical care during COVID-19 pandemic (p-value < 0.001) compared to their counterparts. On the other hand, the average COVID-19-related anxiety score was expected to be significantly lower among those who were without a partner (p-value = 0.037) and who were currently unemployed or retired (p-value < 0.001) compared to their respective counterparts.
The factors associated with perceived stress revealed in the adjusted regression model are presented in Table 5. The average perceived stress score was expected to be significantly higher among participants who were without a partner (p-value = 0.005), who were sometimes to very often feeling overwhelmed by COVID-19 (p-value < 0.001) and who experienced COVID-19-related anxiety (p-value = 0.001) compared to their counterparts. On the other hand, the average perceived stress score was expected to be significantly lower among participants who were currently unemployed or retired (p-value = 0.001), and whose frequency of communication during COVID-19 was less than previous (p-value < 0.001) compared to their respective counterparts.
## Discussion
The current study found that the prevalence of COVID-19-related anxiety and perceived stress was $68\%$ and $93\%$, respectively. This study also revealed that the average COVID-19-related anxiety score was expected to be significantly higher among those who were physically inactive, was concerned about COVID-19, had close friend or family members diagnosed with COVID-19, and had some difficulty in getting food and routine medical care during the pandemic, compared to their respective counterparts. Meanwhile, the participants without partners, who were feeling overwhelmed by COVID-19, and who experienced COVID-19-related anxiety during the pandemic expected to have significantly higher average perceived stress score.
This study reported a very high prevalence of COVID-19-related anxiety and perceived stress ($68\%$ and $93\%$ respectively) among the older adults residing in the Rohingya refugee camp. Poor-socioeconomic conditions, previous traumatic experiences, ongoing risks in their life, uncertainties, and poor access to health care services during the COVID-19 pandemic might have resulted higher level of COVID-19-related anxiety and perceived stress among the participants (Limon et al., 2020; Mistry et al., 2021b, c). We did not find any study conducted in the refugee settings exploring COVID-19 related anxiety and stress using the scales used in the current study to compare with. However, a few studies in other refugee setting used different scales and reported a relatively lower level of COVID-19 related anxiety and stress. For example, a cross-sectional study conducted on migrant returnees (mean age 26 years) in quarantine in Ethiopia found that $48.9\%$ had anxiety symptoms, and more than one-third of the participants ($35.6\%$) had encountered stress (Habtamu et al., 2021). Another study conducted during the COVID-19 pandemic on Bhutanese and Burmese refugees (aged 30–50 years) showed that $68.8\%$ of the participants had stress (Zhang et al., 2022). While these studies were focused on adult population, an Ethiopian study showed a $68.7\%$ prevalence of anxiety among older adults during the COVID-19 pandemic (Jemal et al., 2021). Several factors may explain the differences in anxiety and stress levels among participants in the above studies and those in the current study. They include the usage of different measurement tools, the age range of the study population, pandemic and pre-pandemic condition, socio-cultural variations, refugee settings and other factors affecting anxiety and stress levels.
We found that both COVID-19-related anxiety and perceived stress was higher among the participants who lost their partners. Immediate partner could play a crucial role providing mental support to the fellow partner, particularly during an emergency like COVID-19 pandemic when the support from outside is limited (Jiang et al., 2022; Vowels et al., 2021). In absence of the partner, the participants would have found it difficult to deal with their emotions related to the overwhelming fear associated with COIVD-19 pandemic (Mistry et al., 2021a, b, c, d; Quadros et al., 2021), resulting higher level of COVID-19-related anxiety and perceived stress. We also found that COVID-19-related anxiety and perceived stress was significantly higher among the participants who were feeling concerned/overwhelmed by COVID-19 pandemic. This is also mediated through COVID-19-related fear which resulted a feeling of concern or overwhelm of the pandemic among the participants (Mistry et al., 2021d; Yadav et al., 2021). Previous research conducted among the older adults also documented that worriness related to COVID-19 pandemic resulted in adverse mental health conditions (Khalaf et al., 2022; Webb & Chen, 2022). In line with this, it was revealed that COVID-19-related anxiety was higher among those whose close friends or family members were previously diagnosed with COVID-19. Naturally, a known incident of COVID-19 case within the close connection might have made the participants more fearful of the disease resulting a higher level of anxiety associated with it. Previous research also documented adverse mental health conditions among the people who had their family members diagnosed with COVID-19 (Heesakkers et al., 2022; Tanoue et al., 2020).
Interestingly, we found that COVID-19-related anxiety and perceived stress was significantly lower among the participants who were unemployed or retired. This is probably because the unemployed or retired participants received higher financial assistance from the humanitarian agencies working in the camp during this pandemic (Khan et al., 2020). However, studies conducted among adults in other refugee setting reported that unemployed people faced a higher level of anxiety during this COVID-19 pandemic (Kira et al., 2021; Spiritus-Beerden et al., 2021).
In our study, COVID-19-related anxiety score was higher among the participants who faced difficulties getting food and routine medical care during the pandemic. It is evident that the onset of COVID-19 pandemic resulted in widespread food insecurity and hunger, particularly among the refugee population (Manirambona et al., 2021; UNCHR, 2021) which have caused higher level of anxiety related to the disease. Previous research conducted in the refugee setting also documented that food insecurity has resulted in psychosocial stress (Kaur et al., 2020; Turrini et al., 2017). A study conducted among older Rohingya adults reported that medical services were limited during the COVID-19 pandemic ((Mistry et al., 2021c). Limited health care services, limited access to medicine, difficulties accessing health facilities and fear of getting limited health services from humanitarian organizations might have resulted in higher COVID-19-related anxiety among the participants (Barua & Karia, 2020; Mistry et al., 2021c). This findings are similar to a study conducted among older adults from the Rohingya refugee camps, which showed that difficulties getting food and health care services were associated with older adults’ increased mental health issues during the pandemic (Mistry et al., 2021b).
In our study we found that COVID-19-related anxiety was an independent risk factor of higher perceived stress among the participants during the COVID-19 pandemic. However, in the bi-variate analysis no significant difference was noted in the percentage of participants having perceived stress between the groups having COVID-19-related anxiety or not. This is probably because unlike bi-variate analysis, the regression analysis presents the association after controlling for all the confounding factors. Previous research also highlighted the association between COVID-19-related anxiety and perceived stress (Gallagher et al., 2020; Hu et al., 2021). People often get fearful of COVID-19 pandemic considering its potentials of resulting significant morbidities and mortalities (Chalhoub et al., 2022; Mistry et al., 2021d). Older adults can be particularly fearful thinking that COVID-19 is most lethal among the older adults (Singhal et al., 2021). High level of COVID-19 fear could result in significant anxiety related to COVID-19 which in turn could instigate high level of stress.
Interestingly, findings of the present study also revealed that perceived stress was higher among those who had lesser communication with their family members and friends during the COVID-19 pandemic compared to that of before. This is probably because when people meet their close ones, they tend to discuss more on an important issue like COVID-19 pandemic including its lethality and global reach making them more fearful of it (Mertens et al., 2020) which in turn results higher stress of it. Previous research conducted among the older adults from the Rohingya refugee camp of Bangladesh also revealed that COVID-19 fear was high among those who had lesser communication with their friends and family members during the pandemic (Mistry et al., 2021a).
## Implications for Policy and Practice
Several organizations (International Organization for Migration (IOM), Save the Children International, Action Aid Bangladesh, Caritas Bangladesh, Terre des Hommes) are currently providing mental health and psychosocial support to Rohingya refugees living in Bangladesh (Refugee, 2019). The supports these organizations provided includes community awareness session, need-based individual counseling, podcasting of positive and coping messages in native dialect of Rohingya refugees. However, none of these initiatives particularly addresses the concerns and circumstances of older population and even if available, the acceptability and effectiveness of those initiatives among older adults have not been evaluated. Therefore, our findings suggest the need of evaluation of ongoing interventions and application of those findings in co-design of a people centered interventions that can address the comprehensive psychological determinants of refugee’s health in Bangladesh. Our study’s findings may guide the policymakers, concerned authorities, national and international agencies, and different stakeholders to take appropriate action to address anxiety and stress problems before it poses long term impact on physical and psychological health.
## Strengths and Limitations of the Study
To our knowledge, this is the first study examining COVID-19-related anxiety and perceived stress among older Rohingya refugees during the COVID-19 pandemic. Our study contributed to the limited international literature on anxiety and stress among older Rohingya refugees living in Bangladesh and beyond. Moreover, the study population and study area are unique because the Rohingya camp in *Bangladesh is* the largest refugee setting in the world. The findings of the study will add important information to the limited existing literature on anxiety and stress among displaced, migrated, and refugee people. Despite these strengths, there are some limitations to mention. First, this study was cross-sectional in design therefore temporal relationship could not be established. Secondly, a limited number of camps and sites were selected for data collection due to restrictions on data collection capacity, which could affect the generalizability of the findings for the entire camp population. Third, this study considered only a few explanatory variables and therefore provides a partial overview of factors that might affect COIVD-19-related anxiety and perceived stress among the participants. Moreover, COVID-19-related anxiety and perceived stress reported in the study was based on self-reported data and not on clinical diagnosis.
## Conclusion
This study found that the prevalence of COVID-19-related anxiety and perceived stress were substantial among the older adults residing in the Rohingya refugee camp in Bangladesh. Various sociodemographic and COVID-19-related factors were associated with this high prevalence of COVID-19-related anxiety and perceived stress that need to be addressed holistically to improve the wellbeing of older adults. Our findings recommend strengthening the existing mental health programs to address growing unmet needs related to anxiety and stress for older adults residing in the Refugee camps. We also highlight the necessity of more comprehensive, integrated, and focused services that would meet the unmet and unattended needs of older adults. Finally, it is critical to provide immediate psychosocial and mental health support for the older Rohingya adults during this COVID-19 pandemic and beyond.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 26 KB)
## References
1. Ahmed O, Faisal RA, Sharker T, Lee SA, Jobe MC. **Adaptation of the Bangla version of the COVID-19 Anxiety Scale**. *International Journal of Mental Health and Addiction* (2020.0) **20** 284-295. DOI: 10.1007/s11469-020-00357-2
2. Alam, S. (2019). INFOGRAPHIC - Top Rohingya-hosting countries. https://www.aa.com.tr/en/asia-pacific/infographic-top-rohingya-hosting-countries/1563674. Accessed 13 Nov 2022.
3. Anwar A, Mondal PK, Yadav UN, Shamim AA, Rizwan AAM. **Implications of updated protocol for classification of childhood malnutrition and service delivery in world’s largest refugee camp amid this COVID-19 pandemic**. *Public Health Nutrition* (2022.0) **25** 538-542. DOI: 10.1017/S1368980022000052
4. Barua A, Karia RH. **Challenges faced by Rohingya refugees in the COVID-19 pandemic**. *Annals of Global Health* (2020.0). DOI: 10.5334/aogh.3052
5. Bhatia A, Mahmud A, Fuller A, Shin R, Rahman A, Shatil T, Sultana M, Morshed KAM, Leaning J, Balsari S. **The rohingya in Cox’s bazar: When the stateless seek refuge**. *Health and Human Rights* (2018.0) **20** 105. PMID: 30568406
6. Chadi N, Ryan NC, Geoffroy M-C. **COVID-19 and the impacts on youth mental health: Emerging evidence from longitudinal studies**. *Canadian Journal of Public Health* (2022.0) **113** 44-52. DOI: 10.17269/s41997-021-00567-8
7. Chalhoub Z, Koubeissy H, Fares Y, Abou-Abbas L. **Fear and death anxiety in the shadow of COVID-19 among the Lebanese population: A cross-sectional study**. *PLoS ONE* (2022.0) **17** e0270567. DOI: 10.1371/journal.pone.0270567
8. Cheruvu VK, Chiyaka ET. **Prevalence of depressive symptoms among older adults who reported medical cost as a barrier to seeking health care: Findings from a nationally representative sample**. *BMC Geriatrics* (2019.0) **19** 1-10. DOI: 10.1186/s12877-019-1203-2
9. Gallagher MW, Zvolensky MJ, Long LJ, Rogers AH, Garey L. **The impact of Covid-19 experiences and associated stress on anxiety, depression, and functional impairment in American adults**. *Cognitive Therapy and Research* (2020.0) **44** 1043-1051. DOI: 10.1007/s10608-020-10143-y
10. Habtamu K, Desie Y, Asnake M, Lera EG. **Psychological distress among Ethiopian migrant returnees who were in quarantine in the context of COVID-19: institution-based cross-sectional study**. *BMC Psychiatry* (2021.0) **21** 1-15. DOI: 10.1186/s12888-021-03429-2
11. Heesakkers H, van der Hoeven JG, Corsten S, Janssen I, Ewalds E, Burgers-Bonthuis D, Rettig TCD, Jacobs C, van Santen S, Slooter AJC, van der Woude MCE, Zegers M, van den Boogaard M. **Mental health symptoms in family members of COVID-19 ICU survivors 3 and 12 months after ICU admission: A multicentre prospective cohort study**. *Intensive Care Medicine* (2022.0) **48** 322-331. DOI: 10.1007/s00134-021-06615-8
12. Hu Y, Ye B, Tan J. **Stress of COVID-19, Anxiety, economic insecurity, and mental health literacy: A structural equation modeling approach**. *Frontiers in Psychology* (2021.0) **12** 707079. DOI: 10.3389/fpsyg.2021.707079
13. Huda M, Billah M, Sharmin S, Amanullah ASM, Hossin MZ. **Associations between family social circumstances and psychological distress among the university students of Bangladesh: To what extent do the lifestyle factors mediate?**. *BMC Psychology* (2021.0) **9** 1-11. DOI: 10.1186/s40359-021-00587-6
14. Islam MN. **Psychometric properties of the Bangla version of PSS-10: Is it a single-factor measure or not?**. *Hellenic Journal of Psychology* (2020.0) **17** 15-34
15. Jemal K, Geleta TA, Deriba BS, Awol M. **Anxiety and depression symptoms in older adults during coronavirus disease 2019 pandemic: A community-based cross-sectional study**. *SAGE Open Medicine* (2021.0) **9** 20503121211040050. DOI: 10.1177/20503121211040050
16. Jiang D, Chiu MM, Liu S. **Daily positive support and perceived stress during COVID-19 outbreak: The role of daily gratitude within couples**. *Journal of Happiness Studies* (2022.0) **23** 65-79. DOI: 10.1007/s10902-021-00387-0
17. Júnior JG, de Sales JP, Moreira MM, Pinheiro WR, Lima CKT, Neto MLR. **A crisis within the crisis: The mental health situation of refugees in the world during the 2019 coronavirus (2019-nCoV) outbreak**. *Psychiatry Research* (2020.0). DOI: 10.1016/j.psychres.2020.113000
18. Kamal A-HM, Huda DMN, Dell DCA, Hossain DSZ, Ahmed SS. **Translational strategies to control and prevent spread of COVID-19 in the rohiynga refugee camps in Bangladesh**. *Global Biosecurity* (2020.0). DOI: 10.31646/gbio.77
19. Kaur K, Sulaiman AH, Yoon CK, Hashim AH, Kaur M, Hui KO, Sabki ZA, Francis B, Singh S, Gill JS. **Elucidating mental health disorders among Rohingya refugees: A Malaysian perspective**. *International Journal of Environmental Research and Public Health* (2020.0) **17** 6730. DOI: 10.3390/ijerph17186730
20. Khalaf OO, Abdalgeleel SA, Mostafa N. **Fear of COVID-19 infection and its relation to depressive and anxiety symptoms among elderly population: Online survey**. *Middle East Current Psychiatry* (2022.0) **29** 1-8. DOI: 10.1186/s43045-022-00177-1
21. Khan HT, Rahman MA, Molla MH, Shahjahan M, Abdullah RBJAI. **Humanitarian emergencies of Rohingya older people in Bangladesh: a qualitative study on hopes and reality**. *Ageing International* (2020.0) **47** 20-37. DOI: 10.1007/s12126-020-09400-y
22. Kira IA, Alpay EH, Ayna YE, Shuwiekh HAM, Ashby JS, Turkeli A. **The effects of COVID-19 continuous traumatic stressors on mental health and cognitive functioning: A case example from Turkey**. *Current Psychology* (2021.0). DOI: 10.1007/s12144-021-01743-2
23. Knolle F, Ronan L, Murray GK. **The impact of the COVID-19 pandemic on mental health in the general population: A comparison between Germany and the UK**. *BMC Psychology* (2021.0) **9** 1-17. DOI: 10.1186/s40359-021-00565-y
24. Lee SW, Yang JM, Moon SY, Yoo IK, Ha EK, Kim SY. **Association between mental illness and COVID-19 susceptibility and clinical outcomes in South Korea: a nationwide cohort study**. *The Lancet Psychiatry* (2020.0) **7** 1025-1031. DOI: 10.1016/S2215-0366(20)30421-1
25. Limon MTI, Jubayer MF, Ahmed MU, Rahman H, Kayshar MS. **Rohingya refugees and coronavirus disease-2019: Addressing possible jeopardy from the perspective of Bangladesh**. *Asia Pacific Journal of Public Health* (2020.0) **32** 529-530. DOI: 10.1177/1010539520947887
26. Lou P, Zhu Y, Chen P, Zhang P, Yu J, Zhang N. **Prevalence and correlations with depression, anxiety, and other features in outpatients with chronic obstructive pulmonary disease in China: A cross-sectional case control study**. *BMC Pulmonary Medicine* (2012.0) **12** 1-9. DOI: 10.1186/1471-2466-12-53
27. Mahmud S, Mohsin M, Dewan M, Muyeed A. **The global prevalence of depression, anxiety, stress, and insomnia among general population during COVID-19 pandemic: A systematic review and meta-analysis**. *Trends in Psychology* (2022.0) **31** 143-170. DOI: 10.1007/s43076-021-00116-9
28. Manirambona E, Uwizeyimana T, Uwiringiyimana E, Reddy H. **Impact of the COVID-19 pandemic on the food rations of refugees in Rwanda**. *International Journal for Equity in Health* (2021.0) **20** 1-4. DOI: 10.1186/s12939-021-01450-1
29. Mertens G, Gerritsen L, Duijndam S, Salemink E, Engelhard IM. **Fear of the coronavirus (COVID-19): Predictors in an online study conducted in March 2020**. *Journal of Anxiety Disorders* (2020.0) **74** 102258. DOI: 10.1016/j.janxdis.2020.102258
30. Mistry SK, Mehrab Ali ARM, Akther F, Peprah P, Reza S, Prova S, Yadav UN. **Are older adults of Rohingya community (Forcibly Displaced Myanmar Nationals or FDMNs) in Bangladesh fearful of COVID-19? Findings from a cross-sectional study**. *PLoS ONE* (2021.0) **16** e0253648. DOI: 10.1371/journal.pone.0253648
31. Mistry SK, Ali ARM, Akther F, Yadav UN, Harris MF. **Exploring fear of COVID-19 and its correlates among older adults in Bangladesh**. *Globalization and Health* (2021.0) **17** 1-9. DOI: 10.1186/s12992-021-00698-0
32. Mistry SK, Mehrab Ali ARM, Irfan NM, Yadav UN, Siddique RF, Peprah P, Reza S, Rahman Z, Casanelia L, O'Callaghan C. **Prevalence and correlates of depressive symptoms among Rohingya (forcibly displaced Myanmar nationals or FDMNs) older adults in Bangladesh amid the COVID-19 pandemic**. *Global Mental Health* (2021.0) **8** e23. DOI: 10.1017/gmh.2021.24
33. Mistry SK, Ali ARMM, Yadav UN, Das S, Akter N, Huda MN. **COVID-19 related anxiety and its associated factors: A cross-sectional study on older adults in Bangladesh**. *BMC Psychiatry* (2022.0) **22** 737. DOI: 10.1186/s12888-022-04403-2
34. Mistry SK, Ali AM, Yadav UN, Huda MN, Ghimire S, Bestman A. **Difficulties faced by older Rohingya (forcibly displaced Myanmar nationals) adults in accessing medical services amid the COVID-19 pandemic in Bangladesh**. *BMJ Global Health* (2021.0) **6** e007051. DOI: 10.1136/bmjgh-2021-007051
35. Palit S, Yang H, Li J, Khan M, Saeed A, Hasan MJJC. **The impact of the COVID-19 pandemic on the mental health of Rohingya refugees with pre-existing health problems in Bangladesh**. *Conflict and Health* (2022.0) **16** 1-9. DOI: 10.1186/s13031-022-00443-3
36. Perez LH, Gutierrez LA, Vioque J, Torres Y. **Relation between overweight, diabetes, stress and hypertension: A case–control study in Yarumal-Antioquia Colombia**. *European Journal of Epidemiology* (2001.0) **17** 275-280. DOI: 10.1023/A:1017975925554
37. Pfefferbaum B, North CS. **Mental health and the Covid-19 pandemic**. *New England Journal of Medicine* (2020.0) **383** 510-512. DOI: 10.1056/NEJMp2008017
38. Quadros S, Garg S, Ranjan R, Vijayasarathi G, Mamun MA. **Fear of COVID 19 infection across different cohorts: a scoping review**. *Frontiers in Psychiatry* (2021.0). DOI: 10.3389/fpsyt.2021.708430
39. Refugee, U.N.H.C.f. (2019). Bangladesh refugee emergency
40. Renner A, Jäckle D, Nagl M, Hoffmann R, Röhr S, Jung F, Grochtdreis T, Dams J, König HH, Riedel-Heller S, Kersting A. **Predictors of psychological distress in Syrian refugees with posttraumatic stress in Germany**. *PLoS ONE* (2021.0) **16** e0254406. DOI: 10.1371/journal.pone.0254406
41. Robinson E, Sutin AR, Daly M, Jones A. **A systematic review and meta-analysis of longitudinal cohort studies comparing mental health before versus during the COVID-19 pandemic in 2020**. *Journal of Affective Disorders* (2022.0) **296** 567-576. DOI: 10.1016/j.jad.2021.09.098
42. Salari N, Hosseinian-Far A, Jalali R, Vaisi-Raygani A, Rasoulpoor S, Mohammadi M, Rasoulpoor S, Khaledi-Paveh B. **Prevalence of stress, anxiety, depression among the general population during the COVID-19 pandemic: a systematic review and meta-analysis**. *Globalization and Health* (2020.0) **16** 1-11. PMID: 31898532
43. Singh NS, Bass J, Sumbadze N, Rebok G, Perrin P, Paichadze N. **Identifying mental health problems and Idioms of distress among older adult internally displaced persons in Georgia**. *Social Science & Medicine* (2018.0) **211** 39-47. DOI: 10.1016/j.socscimed.2018.05.007
44. Singhal S, Kumar P, Singh S, Saha S, Dey AB. **Clinical features and outcomes of COVID-19 in older adults: A systematic review and meta-analysis**. *BMC Geriatrics* (2021.0) **21** 1-9. DOI: 10.1186/s12877-021-02261-3
45. Spiritus-Beerden E, Verelst A, Devlieger I, Langer Primdahl N, Botelho Guedes F, Chiarenza A, De Maesschalck S, Durbeej N, Garrido R, Gaspar de Matos M, Ioannidi E, Murphy R, Murphy R, Osman F, Padilla B, Paloma V, Shehadeh A, Sturm G, van den Muijsenbergh M, Vasilikou K, Watters C, Willems S, Skovdal M, Derluyn I. **Mental health of refugees and migrants during the COVID-19 pandemic: The role of experienced discrimination and daily stressors**. *International Journal of Environmental Research and Public Health* (2021.0) **18** 6354. DOI: 10.3390/ijerph18126354
46. Stubbs B, Koyanagi A, Hallgren M, Firth J, Richards J, Schuch F, Rosenbaum S, Rosenbaum S, Veronese N, Lahti J, Vancampfort D. **Physical activity and anxiety: A perspective from the World Health Survey**. *Journal of Affective Disorders* (2017.0) **208** 545-552. DOI: 10.1016/j.jad.2016.10.028
47. Tanoue Y, Nomura S, Yoneoka D, Kawashima T, Eguchi A, Shi S. **Mental health of family, friends, and co-workers of COVID-19 patients in Japan**. *Psychiatry Research* (2020.0) **291** 113067. DOI: 10.1016/j.psychres.2020.113067
48. Tausch A, Souza RO, Viciana CM, Cayetano C, Barbosa J, Hennis AJM. **Strengthening mental health responses to COVID-19 in the Americas: A health policy analysis and recommendations**. *The Lancet Regional Health—Americas* (2022.0) **5** 100118. DOI: 10.1016/j.lana.2021.100118
49. Tinghög P, Malm A, Arwidson C, Sigvardsdotter E, Lundin A, Saboonchi F. **Prevalence of mental ill health, traumas and postmigration stress among refugees from Syria resettled in Sweden after 2011: A population-based survey**. *British Medical Journal Open* (2017.0) **7** e018899
50. Turrini G, Purgato M, Ballette F, Nosè M, Ostuzzi G, Barbui C. **Common mental disorders in asylum seekers and refugees: Umbrella review of prevalence and intervention studies**. *International Journal of Mental Health Systems* (2017.0) **11** 1-14. DOI: 10.1186/s13033-017-0156-0
51. UNCHR. (2021). Pandemic deepens hunger for displaced people the world over. https://www.unhcr.org/en-au/news/stories/2021/3/6062fe334/pandemic-deepens-hunger-displaced-people-world.html. Accessed 29 Nov 2022.
52. United Nations High Commissioner for Refugees. (2022). Operational data portal: Refugee situatios. https://data.unhcr.org/en/situations. Accessed 17 Nov 2022.
53. Vahia IV, Jeste DV, Reynolds CF. **Older adults and the mental health effects of COVID-19**. *JAMA* (2020.0) **324** 2253-2254. DOI: 10.1001/jama.2020.21753
54. Vowels LM, Carnelley KB, Francois-Walcott RRR. **Partner support and goal outcomes during COVID-19: A mixed methods study**. *European Journal of Social Psychology* (2021.0) **51** 393-408. DOI: 10.1002/ejsp.2745
55. Webb LM, Chen CY. **The COVID-19 pandemic's impact on older adults' mental health: Contributing factors, coping strategies, and opportunities for improvement**. *International Journal of Geriatric Psychiatry* (2022.0). DOI: 10.1002/gps.5647
56. World Health Organization. (2022c). WHO Coronavirus (COVID-19) Dashboard. https://covid19.who.int/. Accessed 17 Nov 2022c.
57. World Health Organization. (2022b). COVID-19 pandemic triggers 25% increase in prevalence of anxiety and depression worldwide. https://www.who.int/news/item/02-03-2022b-covid-19-pandemic-triggers-25-increase-in-prevalence-of-anxiety-and-depression-worldwide. Accessed 17 Nov 2022b.
58. World Health Organization. (2022a). Banglaedsh: Cox's Bazar FDMN/Rohingya Refugee & Host Community COVID19 Situation Update. https://app.powerbi.com/view?r=eyJrIjoiOWVkZGU2NGMtM2I3Ny00MDQyLWIwMjEtY2Q0OTM2MTE0ZWJlIiwidCI6ImY2MTBjMGI3LWJkMjQtNGIzOS04MTBiLTNkYzI4MGFmYjU5MCIsImMiOjh9&pageName=ReportSection39710eedc77570aadd8c. Accessed 14 July 2022a.
59. Xiong J, Lipsitz O, Nasri F, Lui LMW, Gill H, Phan L, Chen-Li D, Iacobucci M, Ho R, Majeed A, Majeed A. **Impact of COVID-19 pandemic on mental health in the general population: A systematic review**. *Journal of Affective Disorders* (2020.0) **277** 55-64. DOI: 10.1016/j.jad.2020.08.001
60. Yadav UN, Rayamajhee B, Mistry SK, Parsekar SS, Mishra SK. **A syndemic perspective on the management of non-communicable diseases amid the COVID-19 pandemic in low-and middle-income countries**. *Frontiers in Public Health* (2020.0) **8** 508. DOI: 10.3389/fpubh.2020.00508
61. Yadav UN, Yadav OP, Singh DR, Ghimire S, Rayamajhee B, Mistry SK, Rawal LB, Ali AM, Kumar Tamang M, Mehta S. **Perceived fear of COVID-19 and its associated factors among Nepalese older adults in eastern Nepal: A cross-sectional study**. *PLoS ONE* (2021.0) **16** e0254825. DOI: 10.1371/journal.pone.0254825
62. Zhang M, Gurung A, Anglewicz P, Baniya K, Yun K, Disparities EH. **Discrimination and stress among Asian refugee populations during the COVID-19 pandemic: Evidence from Bhutanese and Burmese refugees in the USA**. *Journal of Racial and Ethnic Health Disparities* (2022.0) **9** 589-597. DOI: 10.1007/s40615-021-00992-y
|
---
title: Lifetime Prevalence of Nonspecific Low Back Pain in Adolescents
authors:
- Stefano Masiero
- Fabio Sarto
- Manuela Cattelan
- Diego Sarto
- Alessandra Del Felice
- Francesco Agostini
- Anna Scanu
journal: American Journal of Physical Medicine & Rehabilitation
year: 2021
pmcid: PMC9988216
doi: 10.1097/PHM.0000000000001720
license: CC BY 4.0
---
# Lifetime Prevalence of Nonspecific Low Back Pain in Adolescents
## Body
Nonspecific low back pain (LBP) is defined as pain and discomfort localized between the costal margin and the lower gluteus folds, with or without radiation to the lower limbs, not attributed to specific and/or known diseases.1,2 *The diagnosis* of this condition in adolescence is of exclusion, for example, in the absence of infections, tumors, spondylolysis, spondylolisthesis, juvenile osteochondrosis of the spine (Scheuermann disease), and rheumatic diseases.3 This aspect must be more emphasized in adult patients (age >20 yrs), owing to the lower frequency of nonspecific LBP in this age group.4 *Epidemiologic data* show that most LBP cases in adolescents are nonspecific. A recent systematic review indicates that the lifetime prevalence of nonspecific LBP in children and adolescents varies between $11.60\%$ and $83.56\%$.5 This wide range is likely because of the heterogeneity and the different cultural and social norms of individuals included.6–10 Indeed, no consensus exists on a sex difference in LBP prevalence,6,11,12 whereas data on the association of height, weight, body mass index (BMI), and anthropometric factors are still inconclusive.13,14
Among the factors associated with the onset, progression, and outcome of this condition, lifestyle factors such as smoking, hours of sleep per night, or long hours sitting (computer, school) and psychosocial factors such as depression, stress, poor academic performance, and perceived weight of backpacks are reported.9,15–17 Physical inactivity is supposed to be associated with higher risk for recurrent LBP, but there are contradictory results reported regarding the association of LBP with physical activity and physical fitness level.18,19 *On this* ground, the link between LBP and physical activity has been described as a U-shaped relationship, where increased risk was found for both subjects with a sedentary lifestyle and those practicing strenuous activities.20 Previous studies reported the prevalence of LBP and associated risk factors in 7542 teenagers aged 13–15 yrs, with a definition of prevalence as the presence of LBP over a 1-yr period.21,22 *In this* study’s cohort, $20.5\%$ of teenagers reported one or more episodes of LBP. Nine hundred ($76.3\%$) had consulted a healthcare professional; a significant association with sex (female), family history, and physical inactivity emerged, whereas anthropometrics or lifestyle items did not correlate.
Nonspecific LBP in adolescents, associated with the risk of developing chronic pain,8,15 has a high impact on the individual as well as on society, with important economic consequences. Therefore, research to highlight LBP causes and develop preventive measures is of utmost importance.
The aim of this study was to investigate the lifetime prevalence and associated factors of nonspecific adolescent LBP to improve knowledge on causative factors and allow measures to prevent chronicity. The aim was to determine the impact of perceived pain on the daily lives and activities of adolescents.
## Abstract
Supplemental digital content is available in the text.
## Background
Many nonconclusive studies have been conducted on low back pain (LBP) in adolescents and associated factors.
### Objective
The aim was to assess the lifetime prevalence and associated factors of LBP in adolescents.
### Materials and Methods
A questionnaire was administered in high school students (14–19-yr-old participants) in Veneto region (Italy). The self-administered, structured questionnaire included anthropometric data; psychologic factors and lifestyle; presence, intensity, and family history of LBP; referral to professional health care for LBP; and a short version of the International Physical Activity Questionnaire.
### Results
A total of 6281 adolescents were recruited; 5204 questionnaires were included in the final analysis. A total of 2549 ($48.98\%$) students reported one or more LBP episodes and 723 ($13.89\%$) reported nonspecific disabling lumbar pain (i.e., no underlying pathology); 1040 ($41.11\%$) subjects with LBP consulted a healthcare professional. A significant association emerged for LBP with sex (female), positive family history, time spent sitting or using electronic devices, sleep deprivation (<5 hrs/night), and low level of physical activity.
### Conclusion
In a large sample of adolescents, LBP lifetime prevalence is high and often associated with disabling pain and sedentary lifestyle, requiring professional care. These findings may support the development of prevention and treatment strategies of LBP in adolescents, reducing the risk of developing chronic pain.
## Participants and Study Design
This is a cross-sectional epidemiologic survey, conducted between February and May 2018 (2017–2018 academic year). Inclusion criteria were students, residing both in urban and rural areas, between the age of 14 and 19 yrs attending high schools in Veneto Region (Italy) who agreed to participate in the survey. Schools were selected on the basis of their zip codes (odd numbers included). Exclusion criteria were already diagnosed spinal pathologies that might cause LBP (Scheuermann disease, spondylolysis, spondylolisthesis, facet arthropathy, sacroiliac joint pain, spondylitic stenosis, compression fracture, and rheumatic diseases) or previous back surgery and back pain areas different from the lumbar region.
## Questionnaire and Data Collection
The study was based on a structured, self-administered questionnaire, ad hoc designed for this epidemiologic survey, consisting of multiple-choice questions.21 Students completed the questionnaires using a laptop, a tablet, or a smartphone. On the day of data collection, the questionnaires were presented by a member of the research team to students in each class during teaching hours: items were illustrated and the students were explained how to fill it in. A temporary password was provided for each class to access the online questionnaire. The questionnaire was anonymous. A pilot study on 78 schools was conducted to test the questionnaire for ease of access, nonequivocal items, and time needed to fill it out. The time required to complete it was on average 20 mins.
The first section consisted of questions regarding demographic items (age, height, weight, BMI, sex). A short version of the International Physical Activity Questionnaire (IPAQ-SF) was included in this section to measure physical activity levels.23 This seven-item questionnaire was developed as a tool for monitoring physical activity and inactivity over the last 7 days. It is divided into four categories: vigorous intensity, moderate intensity, walking, and sitting. For each of these categories, students had to declare for how many days and how many minutes they spent in a specific category of activity. Four subscores expressed in metabolic equivalent of task–minutes per week were obtained by multiplying these data by the intensity coefficients, according to the IPAQ protocol (ipaq.ki.se). Furthermore, a total score was calculated by adding the three subscores related to vigorous and moderate-intensity activity and walking.
According to IPAQ guidelines, individuals who did not answer to the minutes of daily activity or reported more than 960 mins of daily activity were discarded.
The second section collected information regarding type of sporting activity (soccer, volleyball, basketball, athletics, swimming, fitness, rugby, other) and frequency of training sessions (number of weekly hours). Other items investigated lifestyle, such as the daily number of hours of sleep and daily hours with electronic devices (laptop, tablet, or smartphone).
This section ended with items investigating the presence of LBP (at least one episode of LBP in their life), that is, any nonoccasional pain that in some way limited the student in daily activities. The definition of nonspecific LBP followed the European Guidelines for prevention of LBP1: nonspecific LBP is pain and discomfort localized below the costal margin and above the inferior gluteal folds, with or without leg pain, with no other associated back pathology. The final section consisted of questions on the maximum and average level of perceived pain (measured with a numerical rating scale, 0 = no pain, 10 = worst pain) and the need of medical examination. In addition, students were asked whether they ever had to give up social activities because of LBP; those who did were assigned to the disabling LBP group (Dis).
## Ethical Issues
The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of the University Hospital of Padova (n. HEC-DSB/02-19).
Legal guardians signed an informed consent. All procedures were performed according to the Declaration of Helsinki. The STROBE cross-sectional checklist was used for reporting (see Supplemental Checklist, Supplemental Digital Content 1, http://links.lww.com/PHM/B243).24
## Study Size
This study was developed with an explorative aim, with no previous hypothesis about the prevalence of LBP in the population. Thus, computation of the sample size was not performed. However, the number of questionnaires completed ensures a high statistical power for each association test performed, all above $90\%$.
## Statistical Analysis
Descriptive statistics are reported in terms of absolute values and percentages. Univariate analyses on the association between the presence of pain and other categorical variables were performed using chi-square tests. A multivariate analysis, which allows simultaneous evaluation of the association of the different variables with pain, was performed through logistic regression. Evaluation of the significance of covariates in the logistic model was based on the likelihood ratio statistic. In case of missing data, the analyses were performed on individuals with complete answers. Association between the type of lower back pain and the other variables was assessed using the chi-square test. Analyses were run using the statistical software R.25 Statistical significance was set at $P \leq 0.05.$
## RESULTS
Twenty-four schools participated in this survey; 6281 questionnaires were completed, and response rate was $100\%$. Supplementary Figure 1 (Supplemental Digital Content 2, http://links.lww.com/PHM/B244) summarizes the inclusion process. Participants had an average age of 16.93 ± 1.92 yrs (range, 14–19 yrs). Incomplete questionnaires or those with data in the category of exclusion criteria ($$n = 300$$, $4.78\%$) were not included. The resulting data set consisted of 5981 observations (3709 [$62\%$] female students). Supplementary Figure 2 (Supplemental Digital Content 3, http://links.lww.com/PHM/B245) shows the distribution according to age and sex.
Fifty-five percent ($55.6\%$, $$n = 3326$$) of students reported having suffered from back pain (Table 1). These were then divided according to the area of pain: neck pain or LBP. A total of 729 students were excluded because they suffered from neck pain, whereas 48 students were further excluded because they reported having suffered from back pain but did not specify the area; final analyses were based on 5204 responses. There were 2549 ($48.98\%$) participants who reported one or more LBP episodes. The test on the association between sex and back pain distribution showed a significant association ($P \leq 0.001$), with female students suffering more than male students (Table 1). No significant association between BMI and LBP ($$P \leq 0.63$$) emerged.
**TABLE 1**
| Variable | Unnamed: 1 | Total | F, n | M, n | F, % | M, % | P |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Pain | No | 2655 (44.39%) | 1469.0 | 1186.0 | 55.33 | 44.67 | <0.001 |
| | Yes | 3326 (55.61%) | 2240.0 | 1086.0 | 67.35 | 32.65 | <0.001 |
| Area | | | | | | | |
| | No pain | 2655 | 1469.0 | 1186.0 | 55.33 | 44.67 | <0.001 |
| | NP | 729 | 546.0 | 183.0 | 74.9 | 25.1 | <0.001 |
| | LBP | 2549 | 1673.0 | 876.0 | 65.63 | 34.37 | <0.001 |
Table 2 shows that LBP frequency was higher in students who did not practice sports regularly ($51.83\%$) ($P \leq 0.001$). There was no significant association between IPAQ scores and back pain scores ($$P \leq 0.73$$) (Table 2).
**TABLE 2**
| Unnamed: 0 | No Pain, n | LBP, n | No Pain, % | LBP, % | P |
| --- | --- | --- | --- | --- | --- |
| Sport | | | | | |
| No | 737.0 | 793.0 | 48.17 | 51.83 | <0.001 |
| Yes | 1789.0 | 1642.0 | 52.14 | 47.86 | <0.001 |
| | 129.0 | 114.0 | 53.09 | 46.91 | <0.001 |
| Sport played | | | | | |
| None/NA | 976.0 | 1031.0 | 48.63 | 51.37 | |
| Other | 551.0 | 518.0 | 51.54 | 48.46 | |
| Athletics | 69.0 | 78.0 | 46.94 | 53.06 | |
| Basketball | 127.0 | 91.0 | 58.26 | 41.74 | |
| Soccer | 282.0 | 210.0 | 57.32 | 42.68 | |
| Swimming | 159.0 | 119.0 | 57.19 | 42.81 | |
| Body building | 306.0 | 271.0 | 53.03 | 46.97 | |
| Volleyball | 160.0 | 212.0 | 43.01 | 56.99 | |
| Rugby | 25.0 | 19.0 | 56.82 | 43.18 | |
| MET levels | | | | | |
| Low | 440.0 | 429.0 | 43.78 | 42.69 | 0.730 |
| Moderate | 851.0 | 791.0 | 45.12 | 41.94 | 0.730 |
| High | 1364.0 | 1329.0 | 44.84 | 43.69 | 0.730 |
It was observed that the percentage of students with LBP decreased with hours of sleep ($P \leq 0.001$), whereas increased with the number of hours spent sitting ($P \leq 0.001$), time spent using electronic devices ($P \leq 0.001$), and family history ($P \leq 0.001$) (Table 3).
**TABLE 3**
| Unnamed: 0 | No Pain, n | LBP, n | No Pain, % | LBP, % | P |
| --- | --- | --- | --- | --- | --- |
| Hours of sleep | | | | | |
| <5 | 72 | 121 | 37.31 | 62.69 | <0.001 |
| 5–7 | 1155 | 1201 | 49.02 | 50.98 | <0.001 |
| 7–9 | 1357 | 1175 | 53.59 | 46.41 | <0.001 |
| >9 | 71 | 52 | 57.72 | 42.28 | <0.001 |
| Hours sitting | | | | | |
| <5 | 165 | 142 | 53.75 | 46.25 | <0.001 |
| 5–8 | 1777 | 1601 | 52.61 | 47.39 | <0.001 |
| >8 | 713 | 806 | 46.94 | 53.06 | <0.001 |
| Hours spent using tablets/PC/phones | Hours spent using tablets/PC/phones | Hours spent using tablets/PC/phones | Hours spent using tablets/PC/phones | Hours spent using tablets/PC/phones | Hours spent using tablets/PC/phones |
| <2 | 463 | 395 | 53.96 | 46.04 | <0.001 |
| 2–5 | 1515 | 1379 | 52.35 | 47.65 | <0.001 |
| 5–7 | 471 | 529 | 47.10 | 52.90 | <0.001 |
| >7 | 206 | 246 | 45.58 | 54.42 | <0.001 |
| Family history | | | | | |
| No | 1484 | 971 | 60.45 | 39.55 | <0.001 |
| Yes | 1171 | 1578 | 42.60 | 57.40 | <0.001 |
Multivariate analysis on the association of different variables with LBP showed a significant effect of sex, age, sport, hours of sleep, and family history. Given the other covariates, the ratio of probabilities of having had LBP and not having had it for a male student is 0.693 times the same ratio for female students ($P \leq 0.001$). Students who practice sports were less likely to suffer from back pain ($$P \leq 0.002$$). Students who sleep more than 5 hrs per night had a lower chance of reporting LBP ($$P \leq 0.008$$). Lastly, the ratio of probabilities of having LBP and not having it for students with positive family history was 1.87 times the same ratio for those without family history ($P \leq 0.001$).
A total of 1048 ($41.11\%$) students sought medical advice (714 female students), of whom 399 had a disabling LBP. Of the 2549 subjects reporting LBP, 723 ($28.36\%$) had a disabling LBP (Table 4).
**TABLE 4**
| Unnamed: 0 | Non-Dis, n | Dis, n | Non-Dis, % | Dis, % | P |
| --- | --- | --- | --- | --- | --- |
| Sex | | | | | |
| F | 1108.0 | 491.0 | 64.95 | 67.91 | 0.170 |
| M | 598.0 | 232.0 | 35.05 | 32.09 | 0.170 |
| Hours of sleep | | | | | |
| <5 | 64.0 | 50.0 | 56.14 | 43.86 | <0.001 |
| 5–7 | 801.0 | 342.0 | 70.08 | 29.92 | <0.001 |
| 7–9 | 816.0 | 309.0 | 72.53 | 27.47 | <0.001 |
| >9 | 25.0 | 22.0 | 53.19 | 46.81 | <0.001 |
The only significant association with disabling LBP was hours of sleep less than 5 hrs or more than 9 hrs per night ($P \leq 0.001$) (Table 4).
The distribution of maximum pain intensity showed higher numerical rating scale scores for students suffering from a disabling LBP than those suffering from nondisabling LBP ($P \leq 0.001$) (Fig. 1A). A significant association was also present between disabling LBP and the mean pain intensity ($P \leq 0.001$) (Fig. 1B).
**FIGURE 1:** *Distribution of disabling (dis) LBP or non-dis LBP and (A) maximum pain intensity and (B) mean pain intensity.*
## DISCUSSION
This study demonstrated a high lifetime prevalence of nonspecific LBP and associated factors in adolescents in Veneto Region (Italy). In addition, the lifetime prevalence of nonspecific disabling LBP in adolescents is reported, that is, which limited and/or hampered daily life activities and requested medical consultation.
There is general agreement that LBP in adolescents is a health problem requiring much more attention and resources than those devoted at the moment of this writing. In light of lifestyle changes in new generations, studies analyzing LBP risk factors are crucial. The results can be used in the preventive or educational field, which today represents one of the most effective therapeutic approaches in LBP treatment to avoid pain chronicity and the subsequent economic consequences.8,15 The results support the evidence that nonspecific LBP is common in adolescence.5 Indeed, $55.61\%$ (3326 subjects) of students reported having suffered from back pain at least once in their life and $42.62\%$ (2549 subjects) reported one or more episodes of LBP.
LBP lifetime prevalence is a suggestive measure in adolescents: they are more likely to remember pain episodes that occurred also many years earlier, probably because of their emotional, psychologic, and relational life impact.26 In agreement with previous studies, the highest prevalence of LBP was found in the female sex, probably because of a different pain threshold and pain symptom perception.10,27 Other possible related factors are the greater flexibility of the spine compared with males and the possible changes induced by hormonal changes on the appearance and perception of pain.17,28 An association between LBP and BMI was not observed. This is in agreement with previous reports suggesting that nonspecific LBP in adolescents is more related to an incorrect lifestyle.5,8,21 Conversely, it has been demonstrated that in adults, the risk of LBP increases in parallel with BMI and may be modulated by physical activity.29 A clear-cut relationship between physical activity levels (investigated through IPAQ) and LBP was not identified; in fact, it emerged that students who regularly practice sports (at least 2–3 hrs a week) were less likely to suffer of LBP. These data confirm that physical activity, improving muscle elasticity, strength, and likely increasing pain threshold,30 can prevent the onset of LBP.31
The relationship between physical activity levels and LBP is controversial and widely discussed. In fact, it has been observed that both an insufficient as well as an excessive motor activity predisposes to the development of LBP with a U-shaped relationship.32 Specific skills required in different sports expose the vertebral discs to considerable pressures; sports in general increase the risk of injuries, which may lead to LBP.33 An association between LBP and hours of sleep was found. It was observed that sleeping more than 5 hrs a night was associated with a lower probability of suffering from LBP, suggesting that sleep may be a protective factor.
An association between LBP and use of electronic devices (laptop, tablet, or a smartphone) and hours spent sitting was also found,34 suggesting that the two factors may be related. An inadequate prolonged static posture, adopted using these devices, might generate musculoskeletal overload, activating pain receptors.34 The results of this study confirm a predisposition to LBP in subjects with positive family history, likely owing to the genetic factors involved.21,22,28,35 However, the family environment may also play a role because it has been observed that parents may impact on pain threshold level and symptoms complain, heightening the prevalence of disabling LBP in this subsample.21,22,28,35 Apprehension and anxiety by one or both parents can prompt health-seeking behavior, especially when pain presents a chronic course.36 Of 2549 students with LBP, 723 ($28.36\%$) reported at least one episode of nonspecific disabling LBP. However, age and sex did not seem to influence the type of LBP (disabling or not).
An association between disabling LBP and hours of sleep was found. Although sleep may be a protective factor for LBP, it has also been reported sufferers of disabling LBP have a poor sleep quality, negatively affecting both the perception of pain and the quality of life.37 The presence of a disabling pain also increased seeking for healthcare consultation ($55.19\%$). These data are in contrast with previous studies reporting that only $2\%$–$15\%$ of children and adolescents with episodes of LBP require a medical and/or instrumental evaluation.11,28 This difference could be a result of the definition of LBP that was used. Most included studies used structured or semistructured questionnaires with only a partial definition of nonspecific LBP; hence, it could be misleading to draw comparisons with other studies. In fact, according to the definition in this study, the LBP was nonspecific and had to limit the adolescent’s daily activities.
Of note, this study recruited more than 5000 adolescents, whereas surveys in this area of research usually consider less than 1500 participants, and most of these do not reach 500 individuals.5 *The analysis* of such a large sample allows valid and reliable results.
To sum up, these findings stress the need to focus therapeutic efforts with adequate prevention and education programs targeted both to adolescent and relevant adults (e.g., parents, teachers, sports trainers).
## Limitations
This study has some limitations. The main one is the use of an ad hoc questionnaire, which allowed obtaining data from a large sample but is difficult to compare with other studies.
Because different types of schools were included, some minimal bias may be introduced in the study, such as the request for longer autonomous hours in equivalent of grammar schools—thus more time spent sitting rather than exercising. Another limitation is that, despite the fact that adolescents can suffer from pain in different segments of the spine, even simultaneously, the formulation of the questionnaire in this study allowed investigating only one location. Furthermore, the impact of passive and active smoking was not considered. Data on the number and the duration of pain episodes were not included. An investigation of such issues is underway.
Another aspect that was not considered is the perception of the weight of backpacks by adolescents. It has been reported that, rather than the actual and objective weight of the backpack, it is the student’s perception of weight that is associated with LBP.17 Another limitation may have been the imperfect recall of events, which is inherently related to this methodologic approach and cannot be otherwise corrected.
A potential inclusion bias may have been introduced by the exclusion of subjects with a known diagnosis underlying back pain: although this subgroup does experience LBP, the focus of the present study was LBP not related to spinal diseases. The authors reckon that a small sample of individuals in which a diagnosis was not already made might have been included, but they are confident that this may not have substantially impacted on final results.
Another limitation to be considered is the incomplete population sample: for convenience, schools were included on an alternating basis (odd zip codes). Although the randomization method is robust, it needs to be acknowledged that not the whole population aged 14–19 yrs was included.
Lastly, other psychosocial aspects such as depression, anxiety, distress, and exposure to stressful life events were not considered.
## CONCLUSION
The results of this study support the evidence that nonspecific LBP is relatively common among adolescents (mostly in females), especially if they are sedentary and heavy users of electronic devices. A positive family history of LBP is associated with disabling LBP, and family environment (apprehension/anxiety/coping skills) might also play a substantial role. Sleeping more than 5 hrs a night is associated with a lower probability of having LBP. Frequently, adolescents with LBP, particularly those with disabling one, consult a healthcare professional. Practicing sport regularly seems to be associated with a lower probability of having LBP. Further studies are needed to identify those at risk and to define more clearly the role of sports activities in this age group, to promote prevention interventions and plan a personalized rehabilitation program.
## References
1. Burton AK, Balague F, Cardon G. **European guidelines for prevention in LBP**. (2006.0) **15** 136-68
2. Yuan W, Shen J, Chen L. **Differences in nonspecific low back pain between young adult females with and without lumbar scoliosis**. (2019.0) **2019** 9758273. PMID: 30944687
3. Ferraro C, Fraschini P, Masiero S. **Trattamento riabilitativo del paziente in eta` evolutiva affetto da patologie del Rachide**. (2003.0) 5-47
4. DePalma MJ, Ketchum JM, Saullo T. **What is the source of chronic low back pain and does age play a role?**. (2011.0) **12** 224-33. PMID: 21266006
5. Calvo-Muñoz I, Kovacs FM, Roqué M. **Risk factors for low Back pain in childhood and adolescence: A systematic review**. (2018.0) **34** 468-84. PMID: 28915154
6. Salminen JJ. **The adolescent back: A field survey of 370 Finnish school children**. (1984.0) **315** 8-122
7. Skoffer B, Foldspang A. **Physical activity and low-back pain in schoolchildren**. (2008.0) **17** 373-9. PMID: 18180961
8. Kovacs FM, Gestoso M, Gil del Real MT. **Risk factors for non-specific low back pain in schoolchildren and their parents: A population based study**. (2003.0) **103** 259-68. PMID: 12791432
9. Angarita-Fonseca A, Boneth-Collante M, Ariza-Garcia CL. **Factors associated with non-specific low back pain in children aged 10-12 from Bucaramanga, Colombia: A cross-sectional study**. (2019.0) **32** 739-47. PMID: 30814343
10. Minghelli B, Oliveira R, Nunes C. **Non-specific low back pain in adolescents from the south of Portugal: Prevalence and associated factors**. (2014.0) **19** 883-92. PMID: 25145999
11. Burton AK, Clarke RD, McClune TD. **The natural history of low back pain in adolescents**. (1996.0) **21** 2323-8. PMID: 8915066
12. Cakmak A, Yücel B, Ozyalçn SN. **The frequency and associated factors of low back pain among a younger population in Turkey**. (2004.0) **29** 1567-72. PMID: 15247580
13. Hashem LE, Roffey DM, Alfasi AM. **Exploration of the inter-relationships between obesity, physical inactivity, inflammation, and low back pain**. (2018.0) **43** 1218-24. PMID: 29419713
14. Sribastav SS, Long J, He P. **Risk factors associated with pain severity in patients with non-specific low back pain in southern China**. (2018.0) **12** 533-43. PMID: 29879782
15. Harreby M, Neergaard K, Hesselsoe G. **Are radiologic changes in the thoracic and lumbar spine of adolescents risk factors for low back pain in adults? A 25-year prospective cohort study of 640 school children**. (1995.0) **20** 2298-302. PMID: 8553117
16. Zadro JR, Shirley D, Duncan GE. **Familial factors predicting recovery and maintenance of physical activity in people with low back pain: Insights from a population-based twin study**. (2019.0) **23** 367-77. PMID: 30176096
17. Negrini S, Carabalona R. **Backpacks on schoolchildren’s perceptions of load, associations with back pain and factors determining the load**. (2002.0) **27** 187-95. PMID: 11805666
18. Moroder P, Runer A, Resch H. **Low back pain among medical students**. (2011.0) **77** 88-92. PMID: 21473452
19. Deyo RA, Weinstein JN. **Low back pain**. (2001.0) **344** 363-70. PMID: 11172169
20. Heneweer H, Vanhees L, Picavet HS. **Physical activity and low back pain: A U-shaped relation?**. (2009.0) **143** 21-5. PMID: 19217208
21. Masiero S, Carraro E, Celia A. **Prevalence of nonspecific low back pain in schoolchildren aged between 13 and 15 years**. (2008.0) **97** 212-6. PMID: 18177442
22. Masiero S, Carraro E, Sarto D. **Healthcare service use in adolescents with non-specific musculoskeletal pain**. (2010.0) **99** 1224-8. PMID: 20219047
23. Craig CL, Marshall AL, Sjöström M. **International Physical Activity Questionnaire: 12-country reliability and validity**. (2003.0) **35** 1381-95. PMID: 12900694
24. von Elm E, Altman DG, Egger M. **The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies**. (2008.0) **61** 344-9. PMID: 18313558
25. 25R Core Team: R: A Language and Environment for Statistical Computing. Vienna, Austria, R Foundation for Statistical Computing, 2019. Available at: https://www.R-project.org/. Accessed September 1, 2018
26. Jaaniste T, Noel M, von Baeyer CL. **Young children’s ability to report on past, future, and hypothetical pain states: A cognitive-developmental perspective**. (2016.0) **157** 2399-409. PMID: 27429175
27. Keogh E, Eccleston C. **Sex differences in adolescent chronic pain and pain-related coping**. (2006.0) **123** 275-84. PMID: 16644131
28. Balagué F, Damidot P, Nordin M. **Cross-sectional study of the isokinetic muscle trunk strength among school children**. (1993.0) **18** 1199-205. PMID: 8362327
29. Smuck M, Kao MC, Brar N. **Does physical activity influence the relationship between low back pain and obesity?**. (2014.0) **14** 209-16. PMID: 24239800
30. Roth-Isigkeit A, Thyen U, Stöven H. **Pain among children and adolescents: Restrictions in daily living and triggering factors**. (2005.0) **115** e152-62. PMID: 15687423
31. Muntaner-Mas A, Palou P, Ortega FB. **Sports participation and low back pain in schoolchildren**. (2018.0) **31** 811-9. PMID: 29865031
32. Sjolie AN. **Associations between activities and low back pain in adolescents**. (2004.0) **14** 352-9. PMID: 15546330
33. Trompeter K, Fett D, Platen P. **Prevalence of Back pain in sports: A systematic review of the literature**. (2017.0) **47** 1183-207. PMID: 28035587
34. Bento TPF, Cornelio GP, Perrucini PO. **Low back pain in adolescents and association with sociodemographic factors, electronic devices, physical activity and mental health**. (2020.0) **96** 717-24. PMID: 31580844
35. Galozzi P, Maghini I, Bakdounes L. **Prevalence of low back pain and its effect on health-related quality of life in 409 scholar adolescents from the Veneto region**. (2019.0) **71** 132-40. PMID: 31649379
36. Eccleston C, Crombez G, Scotford A. **Adolescent chronic pain: Patterns and predictors of emotional distress in adolescents with chronic pain and their parents**. (2004.0) **108** 221-9. PMID: 15030941
37. You DS, Albu S, Lisenbardt H. **Cumulative childhood adversity as a risk factor for common chronic pain conditions in young adults**. (2019.0) **20** 486-94. PMID: 30011037
|
---
title: Serum Levels of PCSK9 Are Increased in Patients With Active Ulcerative Colitis
Representing a Potential Biomarker of Disease Activity
authors:
- Carla Marinelli
- Fabiana Zingone
- Maria Giovanna Lupo
- Raffaella Marin
- Renata D’Incà
- Alessandro Gubbiotti
- Davide Massimi
- Cesare Casadei
- Brigida Barberio
- Nicola Ferri
- Edoardo Savarino
journal: Journal of Clinical Gastroenterology
year: 2021
pmcid: PMC9988229
doi: 10.1097/MCG.0000000000001607
license: CC BY 4.0
---
# Serum Levels of PCSK9 Are Increased in Patients With Active Ulcerative Colitis Representing a Potential Biomarker of Disease Activity
## Body
Ulcerative colitis (UC) is a bowel disease characterized by a chronic inflammation of the gut, localized exclusively to the colon.1 The etiology is unknown, with both genetic and environmental factors involved. UC has a progressive course with cumulative intestinal damage and potential development of complications, including extraintestinal manifestations (EIMs).1 New evidence suggests that inflammatory bowel diseases (IBDs), particularly UC, are associated with a significant increase of myocardial infarction, stroke, and cardiovascular mortality especially during periods of active disease, although the prevalence of traditional risk factors for cardiovascular disease, such as body mass index (BMI), hypertension, diabetes mellitus, and dyslipidemia is relatively lower in IBD patients than in general population.2–6 Chronic systemic inflammation plays a crucial role in the progressive course of UC and its complications. Several inflammatory molecules [such as interleukin-1β, interleukin-6, C-reactive protein (CRP)] have been investigated, for their possible pathophysiological role.7,8 Importantly, the existence of a link between inflammation and hyperlipidemic status has always been recognized, although a common molecular mediator still needs to be identified.9 *In this* context, recent findings have highlighted the association between proprotein convertase subtilisin/kexin type 9 (PCSK9) levels and chronic low-grade inflammation, suggesting their potential role as markers of inflammation and cardiovascular disease. Indeed, PCSK9 is involved in cholesterol homeostasis by posttranscriptional regulating hepatic low-density lipoprotein (LDL) receptor, and for this reason in atherosclerosis. Beyond cholesterol metabolism, PCSK9 has been investigated for its potential pleiotropic effects, regulating several genes involved in apoptosis, proliferation, immune response, and inflammation.10 To note, 2 fully human monoclonal antibodies targeting PCSK9 (evolocumab and alirocumab) have been recently approved as PCSK9 inhibitors and released on the market to reduce levels of cholesterol and, therefore, cardiovascular risk.11 Given the lack of data on the role of PCSK9 in patients with IBD and the well-known presence of chronic low-grade inflammation in them, the aim of our study was to evaluate PCSK9 serum levels in patients with UC stratified according to disease activity by objective markers of inflammation.
## Background/Goal:
Ulcerative colitis (UC) is characterized by chronic inflammation and progressive course, with potential extraintestinal complications including cardiovascular mortality. Serum proprotein convertase subtilisin/kexin type 9 (PCSK9) levels have been recently recognized as biomarkers of low-grade inflammation and cardiovascular disease. The aim of our study was to evaluate PCSK9 levels in patients with UC and different degrees of disease activity.
### Methods:
We prospectively recruited consecutive patients with UC attending our center at the University Hospital of Padua. Demographics, clinical characteristics, and biochemical data, including PCSK9, high sensitivity C-reactive protein, and fecal calprotectin, were recorded. Moreover, endoscopic procedures were performed in all subjects.
### Results:
We included 112 patients with UC (mean age=52.62±12.84 y; $52.62\%$ males). Patients with UC and abnormal fecal calprotectin (≥250 µg/g) and/or C-reactive protein (≥3 mg/L) had greater levels of PCSK9 compared with UC patients with normal fecal calprotectin and high sensitivity C-reactive protein ($$P \leq 0.03$$ and 0.005, respectively). Higher endoscopic scores in UC were characterized by greater levels of PCSK9 ($$P \leq 0.03$$). Furthermore, we found a positive correlation between PCSK9 levels and fecal calprotectin ($r = 0.18$, $$P \leq 0.04$$), endoscopic Mayo Score ($r = 0.25$, $$P \leq 0.007$$), and UC-Riley Index ($r = 0.22$, $$P \leq 0.01$$). We also found a positive correlation between PCSK9 levels and both total and low-density lipoprotein cholesterol values ($P \leq 0.05$).
### Conclusions:
Serum PCSK9 levels are increased in patients with biochemical and endoscopic evidence of active disease in UC. Further longitudinal studies are necessary to evaluate the role of PCSK9 as a potential biomarker of disease activity and cardiovascular risk in UC.
## Study Population
Ethics Committee of Padua approved this prospective cross-sectional study in May 2019 (protocol number 3312/AO/14). All consecutive patients with UC who presented to our endoscopic service for scheduled activity (ie, surveillance, endoscopic assessment of therapeutic response, etc.) were contacted few days before the colonoscopy to explain the study characteristics and to ask to take part in it. In case of acceptance, they were requested to collect stool specimens immediately before the initiation of bowel cleansing preparation. The nature, duration, and purpose of the study were accurately explained. Before performing study-specific procedures, written informed consent was obtained from each patient enrolled. Inclusion criteria included being older than 18 years and a certain diagnosis of UC, according to international criteria, from at least 6 months. Patients were excluded in case of known pregnancy, diagnosis of acute or chronic liver disease, concomitant or past diagnosis of a chronic immune-mediated inflammatory disease other than IBD, history of prior colectomy, evidence of a concurrent diagnosis of another currently active erosive gastrointestinal mucosal disease, ongoing therapy with any cholesterol medication (eg, statins, protein kinase C inhibitors Repatha/evolocumab), or refuse to sign the informed consent form.
## Disease Activity Evaluation
The day of the endoscopic examination, always performed by the same endoscopist (E.S.), all patients who agreed to be enrolled underwent clinical assessment. Demographics and clinical information were drawn from outpatient medical records and/or in collaboration with the patient. The severity of symptoms reported by the patients was recorded according to partial Mayo Score (pMS). Furthermore, Mayo endoscopic subscore was determined during colonoscopy (eMS). Total Mayo Score (MS) was calculated considering the sum between clinical and endoscopic score. The disease was classified as inactive (MS≤2), mild (3≤MS≤5), moderate (6≤MS≤10), or severe (MS>10).12 Moreover, all patients provided a stool specimen, collected immediately before the start of bowel preparation, for biochemical activity assessment (ie, fecal calprotectin), according to clinical practice. A value of fecal calprotectin ≥250 μg/g was considered abnormal.13 Finally, disease activity was histologically determined on biopsies collected, using UC-Riley Index. This score incorporates 6 histologic features, including: acute and chronic inflammatory cell infiltrate, crypt abscesses and architectural irregularities, mucin depletion, and surface epithelial integrity. A 4-point scale was used to classify each feature as none, mild, moderate, or severe.14
## PCSK9, High-sensitivity C-reactive Protein (hsCRP), and Lipidic Profile Determination
Blood samples were collected immediately before colonoscopy for PCSK9, hsCRP, and lipidic profile determination. After centrifuge (15 min at room temperature, 1300g), serum was transferred to cryovials and stored at −20°C for further analysis.
Total cholesterol and triglycerides were determined with the enzymatic colorimetric methods using cholesterol oxidase/peroxidase aminophenazone and glycerol phosphate oxidase/peroxidase aminophenazone reagents, respectively (Horiba ABX, cat NN° A11A01634, and A11A01640, respectively), and analyzed with Cobas Mira Plus S (ABX Italy). For high-density lipoprotein cholesterol (HDL-C) determination, serum samples were treated with dextran sulfate 500/magnesium chloride for apolipoprotein B–containing lipoproteins precipitation (very LDL, intermediate-density lipoprotein, and LDL, respectively). Then, HDL-C and total cholesterol were quantified. The *Friedewald formula* was applied for the determination of low-density lipoprotein cholesterol (LDL-C): LDL-C=total cholesterol−HDL-C−(triglycerides/5).
Serum PCSK9 concentrations were measured using a commercial enzyme-linked immunosorbent assay kit (R&D Systems, MN; cat. N° SPC900) able to recognize free PCSK9. Briefly, serum samples were diluted 1:20, according to manufacturer’s instructions, and incubated into a microplate precoated with a monoclonal antibody specific for human PCSK9. A 4-parameter logistic curve-fit was generated to obtain sample concentrations, using GraphPad Prism 5. The minimum detectable concentration was 0.219 ng/mL. Intra-assay and interassay coefficients of variation were 5.4±$1.2\%$ and 4.8±$1.9\%$, respectively.
Serum hsCRP concentrations were measured using a commercial enzyme-linked immunosorbent assay kit (apDia, Belgium; cat. N° 740011) able to recognize circulating CRP. Serum samples were diluted 1:1000, according to manufacturer’s instruction, and incubated into a microplate precoated with a monoclonal antibody specific for human CRP. As suggested, sample concentrations were retrieved by generating a linear curve-fit using GraphPad Prism 5. The minimum detectable concentration was ∼0.02 µg/mL. Intra-assay and interassay coefficients of variation ranged $4.1\%$ to $6.9\%$ and $5.8\%$ to $6.3\%$, respectively. A value of hsCRP ≥3 mg/L was considered abnormal.15
## Statistical Analysis
To the best of our knowledge, no studies evaluated the PCSK9 levels in humans according to the grade of intestinal inflammation or more in general in the IBD population. Thus, to estimate the sample size needed for our study, we empirically aimed to detect an increase of at least $25\%$ of the mean PCSK9 levels found in the general population (65.35±30.15) in subjects with active disease as compared with patients in remission.16 Accordingly, we calculated that 47 subjects in remission state and 47 in active state (considering a cutoff of 250 μg/g for calprotectin) were necessary to observe such difference, with a power of $80\%$ and a P-value of 0.05.
Continuous variables were indicated as mean with SD or median with 25th to 75th percentiles if normally distributed or not respectively, while categorical variables were indicated as frequency. Possible differences between 2 groups were assessed with the independent-samples t test or Mann-Whitney test for parametric and nonparametric variables, respectively. Correlation between PCSK9 and the following continuous variables: eMS, pMS, UC-Riley Index, fecal calprotectin, BMI (underweight BMI<18.5, normal weight BMI=18.5-24.9, overweight BMI 25-29.9, obese BMI≥30), total cholesterol, HDL-C and LDL-C, and triglycerides, were conducted using the Pearson correlation coefficient r or Spearman correlation coefficient (rs) for parametric and nonparametric variables, respectively. Linear regression models adjusted for known cardiovascular risk factors (age, sex, BMI, cholesterol, and smoke)2 were used to assess the independent correlation between PCSK9 and markers of disease activity. The receiver operating characteristic curve analysis was used to set the most sensitive and specific serum PCKS9 cutoff in detecting disease activity, CRP, fecal calprotectin, Mayo endoscopic scores, measures of histologic inflammation. Using STATA 11 software for data analysis, P-value was considered statistically significant when <0.05.
## Study Population Characteristics
Among 145 eligible patients contacted, 112 consecutive patients with UC agreed to participate and were enrolled in the present study. The demographic and clinical characteristics of our population have been summarized in Table 1. Fifty-nine ($52.6\%$) patients were males, and the mean age was 52.6±12.8 years. Nine ($8\%$) patients were current smokers, whereas 39 ($34.8\%$) were former smokers. Moreover, we found that about half of our patients ($58\%$) had normal weight, whereas 44 ($39.3\%$) were overweight or obese. Finally, as illustrated in Table 1, comorbidities were quite uncommon likely due to the relatively young age of our population.
**TABLE 1**
| Variables | Study Population (N=112) |
| --- | --- |
| Age [mean±SD (range)] (y) | 52.6±12.8 (21.1-81.9) |
| Sex distribution, n (%) | Sex distribution, n (%) |
| Male | 59 (52.6) |
| Female | 53 (47.4) |
| Smoking stratus, n (%) | Smoking stratus, n (%) |
| Never | 64 (57.2) |
| Smoker | 9 (8.0) |
| Ex-smoker | 39 (34.8) |
| BMI, n (%) | BMI, n (%) |
| Underweight (BMI<18.5) | 3 (2.7) |
| Normal weight (18.5<BMI<24.9) | 65 (58) |
| Overweight: BMI (25<BMI<29.9) | 38 (33.9) |
| Obese (BMI≥30) | 6 (5.4) |
| Comorbidities, n (%) | Comorbidities, n (%) |
| Prior/current cardiovascular disease | 22 (19.6) |
| Diabetes | 4 (3.6) |
| Prior history of peripheral vascular thrombosis | 6 (5.4) |
| Henoch-Schonlein purpura | 1 (0.9) |
| Localization UC, n (%) | Localization UC, n (%) |
| E1 | 13 (11.6) |
| E2 | 36 (32.1) |
| E3 | 63 (56.3) |
| Immunosupressant therapy ongoing, n (%) | 15 (13.4) |
| 5-ASA therapy ongoing, n (%) | 98 (87.5) |
| Biological therapy ongoing, n (%) | Biological therapy ongoing, n (%) |
| Infliximab | 15 (14.3) |
| Humira | 6 (5.4) |
| Vedolizumab | 2 (1.8) |
| Golimumab | 5 (4.5) |
| History of IBD-related surgery, n (%) | 3 (2.7) |
| Calprotectin [median (25th-75th percentiles)] | 139 (60.5-418) |
| Patients with values <250 μg/g | 64 (57.4) |
| Patients with values ≥250 μg/g | 48 (42.6) |
| hsCRP [median (25th-75th percentiles)] | 1.2 (0.2-2.5) |
| Patients with values <3 mg/L | 87 (77.7) |
| Patients with values ≥3 mg/L | 25 (22.3) |
| Partial Mayo Score, n (%) | Partial Mayo Score, n (%) |
| Remission | 88 (78.6) |
| Mild disease | 17 (15.2) |
| Moderate-severe disease | 7 (6.2) |
| Endoscopic Mayo Score, n (%) | Endoscopic Mayo Score, n (%) |
| Remission | 55 (49.1) |
| Mild disease | 30 (26.8) |
| Moderate-severe disease | 27 (24.1) |
| UC-Riley Index (mean±SD) | 4.6±4.4 |
Extension of intestinal disease distribution was as follows: 13 ($11.62\%$) patients had proctitis (E1), 36 ($32.1\%$) patients had left-sided disease (E2), and 63 ($56.2\%$) patients had pancolitis (E3). The majority ($87.5\%$) of patients were taking mesalamine therapy, and 28 ($26\%$) a biologic drug. None of the patients was taking tofacitinib. As to disease activity, 88 ($78.6\%$) patients were considered clinically in remission (pMS<2), but only $50\%$ of them had an eMS compatible with a remission status. The median fecal calprotectin value was 139 µg/g, while the median hsCRP was 1.2 mg/L. Therefore, 48 ($42.6\%$) patients showed biochemical active disease (fecal calprotectin ≥250 μg/g),13 while 25 ($22.3\%$) patients showed abnormal hsCRP values (≥ 3 mg/L).15 As detailed in Table 2, triglycerides determination resulted pathologic (>150 mg/dL, according to clinical practice) in 13 ($11.8\%$) patients, with a mean value of 101.9 mg/dL. Mean total cholesterol was 211 mg/dL ($58.5\%$ patients had >200 mg/dL), mean LDL-C was 134.4 mg/dL ($76.6\%$ patients had >100 mg/dL) and mean HDL-C was 56.2 mg/dL ($58.9\%$ patients had <60 mg/dL). PCSK9 mean value was 165±57.4 ng/mL.
**TABLE 2**
| Features | Mean±SD (Range) |
| --- | --- |
| Triglycerides | 101.9±40.9 (39-274) |
| Patients with values >150 mg/L [n (%)] | 13 (11.8) |
| Total cholesterol | 211±42.5 (119-331) |
| Patients with values >200 mg/L [n (%)] | 65 (58.5) |
| LDL cholesterol | 134.4±40.6 (56.4-253.4) |
| Patients with values >100 mg/L [n (%)] | 86 (76.6) |
| HDL cholesterol | 56.2±15.9 (13-110.4) |
| Patients with values <60 mg/L [n (%)] | 66 (58.9) |
| PCSK9 (ng/mL) | 165±57.4 (71.3-377.07) |
| Median (25th-75th percentiles) | 151.01 (126.1-186.1) |
## PCSK9 Determination in Study Population
Table 3 reports PCSK9 levels according to demographics, disease characteristics, treatments, and lipid profiles in our study population. Patients with UC and abnormal fecal calprotectin (≥250 µg/g) and hsCRP (≥3 mg/L) had statistically significant higher levels of PCSK9 compared with UC patients with normal fecal calprotectin and hsCRP ($$P \leq 0.03$$ and 0.005, respectively). Moreover, higher endoscopic scores in UC were characterized by statistically significant greater serum levels of PCSK9 ($$P \leq 0.03$$), although PCSK9 values did not differ in our population based on clinical activity ($$P \leq 0.27$$) (Table 3). Figure 1 shows graphically PCSK9 distribution based on pathologic calprotectin and hsCPR, clinical, endoscopic, and histologic scores.
Furthermore, we found a statistically significant positive correlation between PCSK9 levels and fecal calprotectin ($r = 0.18$, $$P \leq 0.04$$), hsCRP ($r = 0.26$, $$P \leq 0.006$$), eMS ($r = 0.25$, $$P \leq 0.007$$), and UC-Riley Index ($r = 0.22$, $$P \leq 0.01$$) in UC patients, as detailed in Table 4. As expected, PCSK9 correlated with total cholesterol and LDL-C levels ($r = 0.28$, $$P \leq 0.003$$ and $r = 0.31$, $$P \leq 0.007$$, respectively). In addition, PCSK9 and HDL-C were negatively correlated (r=−0.19, $$P \leq 0.04$$) (Table 4). Receiver operating characteristic curve analyses for serum PCKS9 in detecting disease activity, hsCRP, fecal calprotectin, eMS were also performed observing limited rates of sensitivity and specificity (Fig. 2).
Linear regression models adjusted for cardiovascular risks factors (age, sex, smoke, BMI, and total cholesterol) confirmed the correlation between PCSK9 and fecal calprotectin (Fig. 3), eMS and UC-Riley Index but not for hsCRP (Supplementary Digital Content 1, http://links.lww.com/JCG/A765).
**FIGURE 3:** *Linear regression model between PCSK9 and calprotectin. CI indicates confidence interval; PCSK9, proprotein convertase subtilisin/kexin type 9.*
## DISCUSSION
UC is a chronic condition characterized by periods of recurrence and remission. Indeed, a chronic systemic inflammation despite the lack of symptoms has been demonstrated in previous studies, potentially leading to disease progression and physical as well as psychological disability.17–19 In addition, patients with UC may experience EIMs that can further modify the natural course of their disease and its morbidity.20 PCSK9 has been associated to chronic low-grade inflammation, besides its role in cholesterol metabolism regulation,10 and therefore has been recently recognized as a marker of inflammation and cardiovascular risk, with the potential of playing a role of a target for novel therapies.11 As well recognized, inflammatory processes are complex and involve several molecules. Patients with UC show increased levels of proinflammatory cytokines, including tumor necrosis factor-α. It has been suggested that tumor necrosis factor-α upregulates PCSK9 mRNA and protein synthesis, determining its circulating levels, as previously demonstrated by Ruscica et al.21 Various studies investigated the clinical value of PCSK9 in different conditions, but data in IBD patients are lacking. Thus, we decided to measure the PCSK9 serum levels in patients with UC and to evaluate their correlation with different degrees of disease activity established by clinical, endoscopic, histologic, and biochemical data. We found that serum PCSK9 levels were higher in patients with active UC, and their values were not influenced by confounding factors, including older age, male gender, BMI, and smoking, further supporting the concept that their increasing was related to disease activity.3,22 Future longitudinal studies are mandatory to confirm the role of PCSK9 in the clinical assessment of UC patients and the value of it as a biomarker of disease activity together with its potential role in the cardiovascular evaluation of these subjects.
Chronic systemic inflammation plays a crucial role in influencing the natural course of inflammatory conditions and the development of EIMs, including the occurrence of cardiovascular events. In order to further explore this relationship, several inflammatory molecules have been investigated in various interventional clinical trials, with however unclear results.7 Thus, additional proatherosclerotic pathways have been explored.23 In particular, recent findings highlighted the role of PCSK9 levels,10 so that novel drugs able to inhibit this protein have been developed and launched on the market after having demonstrated that they had benefit with respect to major adverse cardiovascular events in the trial involving high-risk patients.24 Given the increased incidence of cardiovascular events in patients with UC and the similarities with other chronic inflammatory conditions, we investigated the association between PCSK9 and UC. We observed a correlation between serum PCSK9 and fecal calprotectin determination ($r = 0.18$, $$P \leq 0.04$$), a specific marker of intestinal inflammation in IBD patients.25 Furthermore, in line with the well-known interaction between inflammation and PCSK9,26 we found an important correlation also between PCSK9 and both the eMS and the UC-Riley Histology Index ($r = 0.25$, $$P \leq 0.007$$ and $r = 0.22$, $$P \leq 0.01$$, respectively). To note, our additional analysis (regression analysis) emphasized that PCSK9 values were not influenced by traditional cardiovascular risk factors, and therefore they appeared strictly connected to the inflammatory activity of UC.
We found that PCSK9 levels tended to be higher in IBD female population than in males. Even if this result was not statistically significant, it is in line with previous investigations which observed that hormonal regulation induce PCSK9 overexpression in the female general population.16,27 Similarly, PCSK9 levels tended to be higher in subjects with higher BMI, as reported in the literature.28 We reported PCSK9 values in patients with normal weight (160.9±53.2), in those with clinical and endoscopic remission (161.9±54.6 and 150.5±45.9, respectively) and in those with normal fecal calprotectin and hsCRP (155.4±51.5and 157.1±54.3, respectively) are similar to that recently observed in a normal population with normal weight (156±43 ng/dL).28 In contrast, Peng et al29 reported a median levels of PCSK9 in 1225 patients with stable cardiovascular disease of 234.52 ng/mL (interquartile levels ranged from 194.79 to 276.13 ng/mL) which is higher compared with those found in our population (median=151.01, interquartile levels ranged from 126.1 to 186.01).
According to medical literature, total cholesterol and LDL-C showed a positive correlation with PCSK9 serum levels, in line with their physiological role.10 In addition, increased PCSK9 levels correlated with lower HDL-C levels. According to our results, in a previous meta-analysis encompassing 24 randomized controlled trials including >10,000 patients treated with PCSK9 inhibitors, an HDL-C increase, and an LDL-C reduction compared with the placebo group was demonstrated.30 Various studies in medical literature observed that patients with UC present a higher risk of cardiovascular events,31–33 due to both early atherosclerotic processes and hypercoagulable status likely due to the chronic inflammatory condition.34 Indeed, cardiovascular events are more frequently reported in case of disease recurrence or in case of persistent activity.35 *In a* population-based study performed in Olmsted County, Minnesota, from 1980 through 2010, including 736 IBD subjects, Aniwan and colleagues showed, after adjustments for traditional cardiovascular risk factors, that IBD is associated independently with increased risk of acute myocardial infarction [adjusted hazard ratio (aHR), 2.82; $95\%$ confidence interval (CI), 1.98-4.04] and heart failure (aHR, 2.03; $95\%$ CI, 1.36-3.03). Moreover, the relative risk of acute myocardial infarction was significantly increased in patients with Crohn’s disease (CD) (aHR vs. controls, 2.89; $95\%$ CI, 1.65-5.13) or UC (aHR vs. controls, 2.70; $95\%$ CI, 1.69-4.35), whereas the relative risk of heart failure was significantly increased among patients with UC only (aHR, 2.06; $95\%$ CI, 1.18-3.65).36 These findings emphasize the need for monitoring cardiovascular risk factors in IBD and their aggressive reduction.5 However, to date, no biomarkers of cardiovascular risk have been identified, and future longitudinal studies are mandatory to estimate whether PCSK9 could be adopted for the evaluation of cardiovascular risk also in patients with IBD, as it occurs in patients with cardiovascular diseases.
The strength of this study is represented by its prospective design that permitted us to obtain from each patient clear data regarding clinical and endoscopic disease activity to correlate these features with PCSK9 serum levels. Nonetheless, it is also necessary to highlight an important limitation of the present study design. We carried out a cross-sectional study, so far, follow-up data were not available. A longitudinal study would be more appropriate to evaluate serum PCSK9 fluctuations in parallel with the disease state modifications and the possible cardiovascular complications over time. Of note, we decided to exclude patients with CD because of the marked heterogeneity within this condition, the different molecular characteristics of the 2 IBDs and, the lower risk for cardiovascular events observed in CD patients. However, future studies are needed to evaluate the value of PCSK9 levels measurement also in patients with CD. Third, we have to acknowledge that although we observed a significant correlation between PCSK9 and almost all the variables of disease activity measured, the effect size of the Pearson correlation coefficients were rather small. This could be due to the small sample size or the limited number of patients with a high inflammatory burden. Nevertheless, larger prospective studies are necessary to verify the strength of the association between PCSK9 and disease activity in UC. Finally, although we observed a clear correlation between PCSK9 and fecal calprotectin levels, the spread was wide, thus limiting the value of this observation.
In conclusion, PCSK9 represents an interesting mediator of many cellular processes and inflammatory mechanisms. The present research is the first one in the literature indicating a significant association between elevated PCKS9 serum levels and currently recognized markers of inflammation in UC. Data presented here suggest that PCKS9 serum levels could represent a complementary biomarker of disease activity in UC to be used in parallel with CRP and fecal calprotectin. However, the history of PCSK9 is still evolving, and this study could represent a road map for further investigations in the IBD population aimed to evaluate the role of PCSK9 not only as a biomarker of disease activity but also for cardiovascular risk, together with its therapeutic application for cardiovascular risk reduction.
## References
1. Magro F, Gionchetti P, Eliakim R. **Third European evidence-based consensus on diagnosis and management of ulcerative colitis. Part 1: definitions, diagnosis, extra-intestinal manifestations, pregnancy, cancer surveillance, surgery, and ileo-anal pouch disorders**. *J Crohns Colitis* (2017) **11** 649-670. PMID: 28158501
2. Yarur AJ, Deshpande AR, Pechman DM. **Inflammatory bowel disease is associated with an increased incidence of cardiovascular events**. *Am J Gastroenterol* (2011) **106** 741-747. PMID: 21386828
3. Geerling BJ, Badart-Smook A, Stockbrügger RW. **Comprehensive nutritional status in recently diagnosed patients with inflammatory bowel disease compared with population controls**. *Eur J Clin Nutr* (2000) **54** 514-521. PMID: 10878655
4. Kirchgesner J, Beaugerie L, Carrat F. **Increased risk of acute arterial events in young patients and severely active IBD: a nationwide French cohort study**. *Gut* (2018) **67** 1261-1268. PMID: 28647686
5. Panhwar MS, Mansoor E, Al-Kindi SG. **Risk of myocardial infarction in inflammatory bowel disease: a population-based national study**. *Inflamm Bowel Dis* (2019) **25** 1080-1087. PMID: 30500938
6. Singh S, Kullo IJ, Pardi DS, Loftus EV. **Epidemiology, risk factors and management of cardiovascular diseases in IBD**. *Nat Rev Gastroenterol Hepatol* (2015) **12** 26-35. PMID: 25446727
7. Ridker PM, Lüscher TF. **Anti-inflammatory therapies for cardiovascular disease**. *Eur Heart J* (2014) **35** 1782-1791. PMID: 24864079
8. Noutsias M, Pankuweit S, Maisch B. **Biomarkers in inflammatory and noninflammatory cardiomyopathy**. *Herz* (2009) **34** 614-623. PMID: 20024641
9. Ruscica M, Ferri N, Macchi C. **Lipid lowering drugs and inflammatory changes: an impact on cardiovascular outcomes?**. *Ann Med* (2018) **50** 461-484. PMID: 29976096
10. Cesaro A, Bianconi V, Gragnano F. **Beyond cholesterol metabolism: the pleiotropic effects of proprotein convertase subtilisin/kexin type 9 (PCSK9). Genetics, mutations, expression, and perspective for long-term inhibition**. *Biofactors* (2020) **46** 367-380. PMID: 31999032
11. Schmidt AF, Pearce LS, Wilkins JT. **PCSK9 monoclonal antibodies for the primary and secondary prevention of cardiovascular disease**. *Cochrane Database Syst Rev* (2017) **4** CD011748. PMID: 28453187
12. Schroeder KW, Tremaine WJ, Ilstrup DM. **Coated oral 5-aminosalicylic acid therapy for mildly to moderately active ulcerative colitis. A randomized study**. *N Engl J Med* (1987) **317** 1625-1629. PMID: 3317057
13. Lin J-F, Chen J-M, Zuo J-H. **Meta-analysis**. *Inflamm Bowel Dis* (2014) **20** 1407-1415. PMID: 24983982
14. Riley SA, Mani V, Goodman MJ. **Comparison of delayed-release 5-aminosalicylic acid (mesalazine) and sulfasalazine as maintenance treatment for patients with ulcerative colitis**. *Gastroenterology* (1988) **94** 1383-1389. PMID: 2896139
15. Pearson TA, Mensah GA, Alexander RW. **Markers of inflammation and cardiovascular disease: application to clinical and public health practice: a statement for healthcare professionals from the centers for disease control and prevention and the American Heart Association**. *Circulation* (2003) **107** 499-511. PMID: 12551878
16. Cui Q, Ju X, Yang T. **Serum PCSK9 is associated with multiple metabolic factors in a large Han Chinese population**. *Atherosclerosis* (2010) **213** 632-636. PMID: 21040917
17. Marinelli C, Zingone F, Inferrera M. **Factors associated with disability in patients with ulcerative colitis: a cross-sectional study**. *J Dig Dis* (2020) **21** 81-87. PMID: 31859432
18. Barberio B, Zamani M, Black CJ. **Prevalence of anxiety and depression in inflammatory bowel disease: systematic review and meta-analysis**. *Lancet Gastroenterol Hepatol* (2021) **6** 359-370. PMID: 33721557
19. Barberio B, Zingone F, Savarino EV. **Inflammatory bowel disease and sleep disturbance: as usual, quality matters**. *Dig Dis Sci* (2021) **66** 3-4. PMID: 32323074
20. Isene R, Bernklev T, Hoie O. **Extraintestinal manifestations in Crohn’s disease and ulcerative colitis: results from a prospective, population-based European inception cohort**. *Scand J Gastroenterol* (2014) **50** 300-305. PMID: 25535653
21. Ruscica M, Ricci C, Macchi C. **Suppressor of cytokine signaling-3 (SOCS-3) induces proprotein convertase subtilisin kexin type 9 (PCSK9) expression in hepatic HepG2 cell line**. *J Biol Chem* (2016) **291** 3508-3519. PMID: 26668321
22. Jahnsen J, Falch JA, Mowinckel P. **Body composition in patients with inflammatory bowel disease: a population-based study**. *Am J Gastroenterol* (2003) **98** 1556-1562. PMID: 12873577
23. Montecucco F, Liberale L, Bonaventura A. **The role of inflammation in cardiovascular outcome**. *Curr Atheroscler Rep* (2017) **19** 11. PMID: 28194569
24. Ridker PM, Revkin J, Amarenco P. **Cardiovascular efficacy and safety of bococizumab in high-risk patients**. *N Engl J Med* (2017) **376** 1527-1539. PMID: 28304242
25. Yamaguchi S, Takeuchi Y, Arai K. **Fecal calprotectin is a clinically relevant biomarker of mucosal healing in patients with quiescent ulcerative colitis**. *J Gastroenterol Hepatol* (2016) **31** 93-98. PMID: 26212346
26. Feingold KR, Moser AH, Shigenaga JK. **Inflammation stimulates the expression of PCSK9**. *Biochem Biophys Res Commun* (2008) **374** 341-344. PMID: 18638454
27. Lakoski SG, Lagace TA, Cohen JC. **Genetic and metabolic determinants of plasma PCSK9 levels**. *J Clin Endocrinol Metab* (2009) **94** 2537-2543. PMID: 19351729
28. Filippatos TD, Liberopoulos E, Georgoula M. **Effects of increased body weight and short-term weight loss on serum PCSK9 levels—a prospective pilot study**. *Arch Med Sci Atheroscler Dis* (2017) **2** 46-51
29. Peng J, Liu MM, Jin JL. **Association of circulating PCSK9 concentration with cardiovascular metabolic markers and outcomes in stable coronary artery disease patients with or without diabetes: a prospective, observational cohort study**. *Cardiovasc Diabetol* (2020) **19** 167. PMID: 33023603
30. Navarese EP, Kołodziejczak M, Schulze V. **Effects of proprotein convertase subtilisin/kexin type 9 antibodies in adults with hypercholesterolemia: a systematic review and meta-analysis**. *AnnIntern Med* (2015) **163** 40-51
31. Kristensen SL, Ahlehoff O, Lindhardsen J. **Disease activity in inflammatory bowel disease is associated with increased risk of myocardial infarction, stroke and cardiovascular death—a Danish Nationwide Cohort Study**. *PLoS One* (2013) **8** e56944. PMID: 23457642
32. Wu G-C, Leng R-X, Lu Q. **Subclinical atherosclerosis in patients with inflammatory bowel diseases: a systematic review and meta-analysis**. *Angiology* (2017) **68** 447-461. PMID: 27252243
33. Buno DM, Timofte CE, Ciocoiu M. **Cardiovascular manifestations of inflammatory bowel disease: pathogenesis, diagnosis, and preventive strategies**. *Gastroenterol Res Pract* (2019) **2019** 3012509. PMID: 30733802
34. Karmiris K, Avgerinos A, Tavernaraki A. **Prevalence and characteristics of extra-intestinal manifestations in a large cohort of Greek patients with inflammatory bowel disease**. *J Crohns Colitis* (2016) **10** 429-436. PMID: 26721936
35. Hansson GK, Robertson A-KL, Söderberg-Nauclér C. **Inflammation and atherosclerosis**. *Annu Rev Pathol* (2006) **1** 297-329. PMID: 18039117
36. Aniwan S, Pardi DS, Tremaine WJ. **Increased risk of acute myocardial infarction and heart failure in patients with inflammatory bowel diseases**. *Clin Gastroenterol Hepatol* (2018) **16** 1607.e1-1615.e1. PMID: 29702298
|
---
title: What Personal and Work-Related Characteristics of Dutch Construction Workers
With Knee Osteoarthritis Are Associated With Future Work Ability?
authors:
- Britte L. De Kock
- Jack Van der Gragt
- Henk F. Van der Molen
- P. Paul F.M. Kuijer
- Nina Zipfel
journal: Journal of Occupational and Environmental Medicine
year: 2022
pmcid: PMC9988233
doi: 10.1097/JOM.0000000000002730
license: CC BY 4.0
---
# What Personal and Work-Related Characteristics of Dutch Construction Workers With Knee Osteoarthritis Are Associated With Future Work Ability?
## Body
The construction industry is characterized by high physical work demands and prevalent work-related musculoskeletal disorders.1–3 A review including a meta-analysis showed that knee complaints are the most common complaint in the construction industry second only to low-back pain, with a 1-year prevalence of $37\%$ ($95\%$ confidence interval [CI], $22\%$ to $52\%$).4 These knee complaints are in large part attributable to knee osteoarthritis (KO), given the high prevalence of this work-related disease compared with other diagnoses such as meniscal tears or bursitis.5–7 In the upcoming years, a steep rise in KO is forecasted, especially among workers aged between 40 and 65 years.8 Specific construction professions with a high risk of KO are floor layers, asphalt workers, sheet-metal workers, rock workers, plumbers, bricklayers, wood workers, and concrete workers.9 Given the high prevalence of KO and the high physical work demands, these construction workers deserve timely and effective occupational health care to maintain their work ability despite the debilitating effect of KO. Doctor-diagnosed KO has an almost twofold increased risk of sick leave and about $40\%$ to $50\%$ increased risk of disability pension compared with the general population.10 In addition, a recent Dutch study showed that the annual sick leave costs due to KO in the Dutch workforce are substantial for workers consulting an occupational physician, with estimated costs of $43 million. The average costs per sick leave episode are also considerable—$16,846 for KO with an average sick leave duration of 186 days.11 Unfortunately, no occupational health guidelines are available that address how work ability among workers with KO should be managed. Present clinical guidelines for nonoperative treatment are primarily focused on pain and self-reported function of activities of daily life.12 These outcomes are not always a good proxy for being able to work.13 To prevent sick leave and early retirement, and to contribute to a life span approach for KO, occupational health care should target prognostic factors that enhance the construction workers' future work ability.14 Therefore, the aim of this study is to assess the personal and work-related characteristics of construction workers diagnosed with KO associated with their ability to perform their current profession in the following 2 years.
## Abstract
In the upcoming decade, the prevalence of knee osteoarthritis will rise steeply, especially among workers. Therefore, a better insight in prognostic factors to maintain future work ability is of importance, especially in physically demanding jobs. This study shows that both occupational and leisure-time physical activity needs to be addressed.
## Objective
To assess personal and work-related characteristics of construction workers with knee osteoarthritis (KO) associated with their ability to perform their current profession in the following 2 years.
### Methods
A cross-sectional study was performed among Dutch construction workers diagnosed with KO using data from the Worker Health Surveillance. Logistic regression was used to assess the characteristics associated with future work ability.
### Results
On the basis of 344 construction workers with KO, being able to perform their current profession in 2 years' time was associated with working weekly 36 to 45 hours (odds ratio [OR], 3.0 to 6.3), performing high-intensity exercises 1 to 2 times weekly (OR, 2.0 to 2.6), being younger than 56 years (OR, 0.2 to 0.3), and not performing strenuous work activities such as lifting and kneeling (OR, 0.4 to 0.5).
### Conclusions
To keep construction workers with KO at work, intervention studies should evaluate the effects of reducing strenuous work activities and promote leisure-time exercise.
## Study Design
A cross-sectional study was conducted to determine the association between personal and work-related characteristics and the self-reported future ability of construction workers with KO to practice their current profession in 2 years' time. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)checklist was used to describe this cross-sectional study (Appendix 1, http://links.lww.com/JOM/B219).15
## Participants and Measurement
For this study, male construction workers aged between 20 and 68 years were selected. Participants were included when they received a KO diagnosis by an occupational physician. Data were used from the Worker Health Surveillance (WHS) for construction workers performed between 2011 and 2021 in the Netherlands. The WHS consists of two parts, consisting of [1] a standardized physical examination and [2] a self-administered questionnaire. Construction workers were invited to participate in the WHS starting at the age of 20 years, followed by periodic checkups every 5 years and more regularly after the age of 50 years.
## Variables
For the purpose of this study, specific questions from the WHS were selected focusing on knee complaints and future work ability (Appendix 2, http://links.lww.com/JOM/B220). Questions were selected based on risk factors from the literature for KO and work ability.9,10,16–33
## Study Size
For this study, all construction workers were included if doctor-diagnosed KO was established. Data from the WHS between 2011 and 2021 were used, which resulted in a total sample size of 344 male construction workers with clinically diagnosed KO (Fig. 1).
**FIGURE 1:** *Flowchart: inclusion and exclusion criteria of study participants.*
## Primary Outcomes
The selected primary outcome was the self-reported prognosis of work ability to practice the current profession in 2 years' time measured with a self-administered questionnaire taken from the Work Ability Index questionnaire.34,35 The answering categories were unlikely, maybe, and very likely. For this study, the population was divided into two groups: [1] construction workers who reported being “very likely to be practicing their current profession in 2 years' time” (answering category: “very likely”) and [2] construction workers who reported being “not very likely to be practicing their current profession in 2 years' time” (answering categories: “maybe” and “unlikely”). The two groups were created to form comparable group sizes in the number of participants. Independent variables were age, body mass index (BMI; body height and body weight were measured during the physical examination), number of years in the current profession, average number of working hours per week, number of times per week of high-intensity exercise, performance of strenuous work activities, and the worker being of the opinion that work caused the complaints. Continuous variables were recoded into categories. Age was recoded into four categories (<40 to 49 years, 50 to 55 years, 56 to 60 years, >60 years). Given the increased risk of KO with older age, in our opinion, this categorization best represented the included participant group. Body mass index was categorized as normal weight (BMI of 18 to 25 kg/m2), overweight (BMI of 25 to 30 kg/m2), and obesity (BMI > 30 kg/m2). Number of years in the current profession was categorized as 0 to 10 years, 11 to 20 years, 21 to 30 years, 31 to 40 years, and 41 to 50 years. Average number of working hours per week was categorized as 3 to 35 hours, 36 to 40 hours, 41 to 45 hours, and >45 hours. Number of times per week of high-intensity exercise was categorized as 0 times, 1 time, 2 times, and >3 times. Performing regularly strenuous work activities was defined as [1] bending over, [2] kneeling or squatting, and/or [3] lifting, pushing, pulling, or carrying heavy loads during work, with the answering categories of no strenuous work activity, one of these strenuous work activities, two of these activities, and three of these activities. Finally, work ability on a scale from 0 to 10 was categorized as 0 to 5 being poor, 6 to 7 being moderate, 8 to 9 being good, and 10 being excellent.
## Statistical Methods
To get insight into the association between personal and work-related characteristics and the future ability to practice the current profession in 2 years' time, a binary logistic regression was conducted. First, a univariable analysis was conducted. Second, a multivariable binary logistic regression model was performed based on the significance of variables set at P ≤ 0.05 from the univariable analysis. For the multivariable regression model, the forward stepwise procedure was used. Before performing the regression analysis, the included variables were tested on collinearity and power (Appendix 3, http://links.lww.com/JOM/B221). The quality of the model was assessed with the −2-log likelihood ratio test and the Hosmer-Lemeshow test (Appendix 4, http://links.lww.com/JOM/B222, and 5, http://links.lww.com/JOM/B223). A Hosmer-Lemeshow chi-squared with P ≥ 0.05 showed goodness of fit. For the results, categorical variables were presented as numbers and percentages. Numerical variables were additionally presented as mean and standard deviation. The results of the logistic regression analysis were presented using odds ratio (OR) and their corresponding $95\%$ CI, and the P value was set at P ≤ 0.05. Data were analyzed using IBM SPSS Statistics version 26.
## Participants and Descriptive Data
As already stated, a total of 344 Dutch construction workers diagnosed with KO were included in this study (Fig. 1). The personal and work-related characteristics of the participants are presented in Table 1. Most of the participants were above the age of 56 years, felt pain or stiffness in their knee, were overweight, and performed high-intensity exercise zero times per week. Moreover, they worked 36 to 40 hours per week and had worked in their current profession for 21 to 30 years. Participants reported having good physical and mental work ability and reported regularly bending over, kneeling or squatting, and lifting, pushing, pulling, or carrying heavy loads during work.
**TABLE 1**
| Unnamed: 0 | n (%) | Mean (SD) | Unnamed: 3 |
| --- | --- | --- | --- |
| Personal characteristics | | | |
| Age, yrs | | 56.5 (6.1) | |
| <40–49 | 47 (14) | | |
| 50–55 | 74 (22) | | |
| 56–60 | 119 (25) | | |
| >60 | 104 (30) | | |
| Pain or stiffness in knee the previous year | | | |
| No | 42 (12) | | |
| Yes | 301 (88) | | |
| No. times per week of high-intensity exercise (eg, soccer, running, indoor sports, physically demanding work) | | 2.3 (1.2) | |
| 0 | 120 (40) | | |
| 1 | 40 (13) | | |
| 2 | 85 (28) | | |
| >3 | 59 (19) | | |
| BMI | | 28.3 (3.7) | |
| Normal weight (BMI of 18–25) | 62 (18) | | |
| Overweight (BMI of 25–30) | 186 (54) | | |
| Obesity (BMI > 30) | 96 (28) | | |
| Work-related characteristics | | | |
| Average no. workhours per week | | 39.9 (7.2) | |
| 3–35 | 53 (15) | | |
| 36–40 | 207 (60) | | |
| 41–45 | 43 (13) | | |
| >45 | 40 (12) | | |
| No. years working in the current profession | | 24.8 (13.1) | |
| 0–10 | 64 (19) | | |
| 11–20 | 68 (20) | | |
| 21–30 | 95 (28) | | |
| 31–40 | 65 (19) | | |
| 41–50 | 51 (15) | | |
| Work ability at the moment (physical and mental) | | 7.3 (1.4) | |
| 0–5, poor | 30 (9) | | |
| 6–7, moderate | 143 (42) | | |
| 8–9, good | 150 (44) | | |
| 10, excellent | 17 (5) | | |
| Strenuous work activities: (1) bending over, (2) kneeling or squatting, and (3) lifting, pushing, pulling, or carrying heavy loads during work | | 1.3 (1.1) | 1.3 (1.1) |
| No | 99 (30) | | |
| One work activity | 86 (26) | | |
| Two work activities | 77 (24) | | |
| Three work activities | 64 (20) | | |
## Personal and Work-Related Characteristics of Work Ability in 2 Years' Time
The univariable binary logistic regression was performed first for the seven selected variables separately (Table 2). The two variables of BMI and whether work had caused complaints were not significantly associated with work ability in 2 years' time ($$P \leq 0.42$$ and $$P \leq 0.55$$, respectively) and were thus not included in the multivariable analysis.
**TABLE 2**
| Unnamed: 0 | Odds Ratio | 95% Confidence Interval | P Value |
| --- | --- | --- | --- |
| Personal characteristics | Personal characteristics | Personal characteristics | Personal characteristics |
| Age, yrs | | | |
| (Ref: <40–49) | 1.00 | | 0.000 |
| 50–55 | 0.85 | 0.73–1.94 | 0.699 |
| 56–60 | 0.34 | 0.16–0.71 | 0.004 |
| >60 | 0.25 | 0.12–0.54 | 0.000 |
| No. times per week of high-intensity exercise (eg, soccer, running, indoor sports, physically demanding work) | | | |
| (Ref: 0) | 1.00 | | 0.009 |
| 1 | 2.54 | 1.20–5.39 | 0.02 |
| 2 | 2.24 | 1.27–3.97 | 0.006 |
| >3 | 1.10 | 0.59–2.06 | 0.756 |
| BMI | | | |
| (Ref: normal weight [BMI of 18–25]) | 1.00 | | 0.42 |
| Overweight (BMI of 25–30) | 0.68 | 0.38–1.23 | 0.199 |
| Obesity (BMI > 30) | 0.697 | 0.36–1.34 | 0.28 |
| Work-related characteristics | Work-related characteristics | Work-related characteristics | Work-related characteristics |
| I can do my work, but it does cause some complaints | | | |
| (Ref: No) | 1.00 | | 0.55 |
| Yes | 0.88 | 0.57–1.35 | 0.55 |
| Average no. work hours per week | | | |
| (Ref: 3–35) | 1.00 | | 0.001 |
| 36–40 | 2.58 | 1.37–4.85 | 0.003 |
| 41–45 | 6.42 | 2.59–15.90 | 0.000 |
| >45 | 2.15 | 0.93–4.90 | 0.08 |
| No. years working in the current profession | | | |
| (Ref: 0–10) | 1.00 | | 0.003 |
| 11–20 | 1.53 | 0.75–3.10 | 0.24 |
| 21–30 | 0.88 | 0.47–1.67 | 0.70 |
| 31–40 | 1.095 | 0.54–2.21 | 0.80 |
| 41–50 | 0.34 | 0.15–0.72 | 0.005 |
| Strenuous work activities: (1) bending over, (2) kneeling or squatting, and (3) lifting, pushing, pulling, or carrying heavy loads during work | | | |
| (Ref: no) | 1.00 | | 0.02 |
| One work activity | 1.04 | 0.58–1.89 | 0.89 |
| Two work activities | 0.63 | 0.35–1.16 | 0.14 |
| Three work activities | 0.42 | 0.22–0.79 | 0.008 |
The final model for the analysis on the association between personal and work-related characteristics on work ability in 2 years' time is presented in Table 3 and included four variables. Personal and work-related characteristics associated with the ability to practice the current profession in 2 years' time were 1-time high-intensity exercise per week (OR, 2.62; $95\%$ CI, 1.12 to 6.15), 2 times per week of high-intensity exercise (OR, 2.00; $95\%$ CI, 1.06 to 3.80), working for 36 to 40 hours per week (OR, 3.01; $95\%$ CI, 1.36 to 6.67), working 41 to 45 hours per week (OR, 6.25; $95\%$ CI, 2.06 to 18.99), performing two strenuous work activities (OR, 0.45; $95\%$ CI, 0.22 to 0.95), performing three strenuous work activities (OR, 0.40; $95\%$ CI, 0.18 to 0.88), age between 56 and 60 years (OR, 0.22; $95\%$ CI, 0.09 to 0.54), and age > 60 years (OR, 0.28; $95\%$ CI, 0.11 to 0.69). The variable numbers of years working in the current profession was not significant and was thus not included in the final model. The Hosmer-Lemeshow test showed a goodness of fit (χ2 = 4.538, $$P \leq 0.806$$).
**TABLE 3**
| Unnamed: 0 | N Total (N = 343) | Construction Workers Who Are Very Likely to Still Practice Their Current Function in 2 Years' Time (N = 190) | Construction Workers Who Are Unlikely or Maybe Able to Practice Their Current Function in 2 Years' Time (N = 153) | Odds Ratio | 95% Confidence Interval | P Value |
| --- | --- | --- | --- | --- | --- | --- |
| Personal characteristics | Personal characteristics | Personal characteristics | Personal characteristics | Personal characteristics | Personal characteristics | Personal characteristics |
| Age, yrs | | | | | | |
| (Ref: <40–49) | 47 | 35 | 12 | 1.00 | | |
| 50–55 | 74 | 53 | 21 | 0.74 | 0.28–1.93 | 0.53 |
| 56–60 | 119 | 59 | 60 | 0.22 | 0.09–0.54 | 0.001 |
| >60 | 104 | 44 | 60 | 0.28 | 0.11–0.69 | 0.006 |
| No. times per week of high-intensity exercise (eg, soccer, running, indoor sports, physically demanding work) | | | | | | |
| (Ref: 0) | 120 | 54 | 66 | 1.00 | | |
| 1 | 40 | 27 | 13 | 2.62 | 1.12–6.15 | 0.03 |
| 2 | 85 | 55 | 30 | 2.00 | 1.06–3.80 | 0.03 |
| >3 | 59 | 28 | 31 | 1.10 | 0.50–2.31 | 0.87 |
| Work-related characteristics | Work-related characteristics | Work-related characteristics | Work-related characteristics | Work-related characteristics | Work-related characteristics | Work-related characteristics |
| Average no. work hours per week | | | | | | |
| (Ref: 3–35) | 53 | 18 | 35 | 1.00 | | |
| 36–40 | 207 | 118 | 89 | 3.01 | 1.36–6.67 | 0.007 |
| 41–45 | 43 | 33 | 10 | 6.25 | 2.06–18.99 | 0.001 |
| >45 | 40 | 21 | 19 | 1.61 | 0.58–4.45 | 0.36 |
| Strenuous work activities: (1) bending over, (2) kneeling or squatting, and (3) lifting, pushing, pulling, or carrying heavy loads during work | | | | | | |
| (Ref: no) | 102 | 62 | 40 | 1.00 | | |
| One work activity | 87 | 53 | 34 | 1.12 | 0.56–2.23 | 0.76 |
| Two work activities | 72 | 34 | 38 | 0.45 | 0.22–0.95 | 0.04 |
| Three work activities | 65 | 27 | 38 | 0.40 | 0.18–0.88 | 0.02 |
## DISCUSSION
Our study found that the ability to practice the current profession in 2 years' time is associated with high-intensity exercises once or twice a week, working for 36 to 45 hours per week, performing less than two strenuous work activities, and being younger than 56 years.
The following findings seem in line with previous studies on the topic of having KO and enhancing work participation. For instance, being physically active outside work seems especially important for workers in physically strenuous jobs. Despite that the guidelines of the World Health Organization on physical activity and sedentary behavior make no distinction between occupational physical activity and leisure-time physical activity, the present results support the assumption that the so-called Goldilocks Principle also applies for workers with KO.36 *This is* in line with the results of a recent evaluation of this principle for construction and health care workers, reporting that workers who spent more time on physical activity during leisure reported less musculoskeletal pain.37 Of course, being physically active in leisure time is easier said than done given that construction workers do not have a fixed workplace and therefore often have to travel back and forth from work, leaving little time and energy to exercise before or after work. The positive work ability results for remaining physically active in leisure time are also in line with the nonoperative treatment guidelines for KO, supporting a physically active lifestyle to enhance participation.38 Moreover, the Good Life with osteoArthritis in Denmark (GLA:D™) (Research Unit for Musculoskeletal Function and Physiotherapy at the University of Southern Denmark) showed that regular intense exercise (twice weekly for 6 weeks) seemed to improve work participation, reducing sickness absence in 1 year from $24\%$ ($95\%$ CI, 21 to 28) to $15\%$ ($95\%$ CI, 12 to 18).39 Not surprisingly, a high physical workload due to frequent bending over, kneeling or squatting, and lifting, pushing, pulling, or carrying heavy loads is not beneficial for the work ability of construction workers with KO. A high physical workload is an established risk factor for the onset or worsening of KO.14,33 Moreover, KO limits the ability to perform these activities, and therefore, the negative association found with the future work ability of construction workers with KO is in line with the literature.29 The same line of reasoning applies for older age. The prevalence of KO is of course strongly related to an increase of age.14,17 In our study, we decided to create a composite outcome variable of physical workload activities as we expect a high correlation between these activities.30 However, we also conducted an additional analysis of the physical workload activities separately (Appendix 6, http://links.lww.com/JOM/B224). The results of the additional analysis showed that the work activity of bending over of the trunk in the univariate analysis was not found to be significant. This might also be expected given that KO is less likely to limit trunk activity compared with the other two more strongly associated activities that are more demanding for the lower extremities, namely, “kneeling or squatting” and “lifting, pushing, pulling, or carrying heavy loads during work.” Not all findings seem in line with previous studies, however. The fact that working less than 36 hours per week does not support future work ability seems contradictory to current findings.40 Working 36 to 45 hours seems optimal. However, continuing to work a high number of hours as a construction worker does not seem a wise decision, despite construction workers not being able to resist it. On the other hand, the result in this study can possibly be explained due to a healthy worker's survival effect, which means that, in this study, construction workers with KO who were absent from work or who retired early did not participate in the surveillance.41 However, we believe that the risk of selection bias is minimal in our study.42 Both the two oldest age groups (56 to 60 years and >60 years) are less likely to do their current profession in 2 years' time compared with their younger colleagues (see Table 3). This is probably explained by the fact that KO is a progressive disease for which no cure is available.16,17 *It is* interesting to note that BMI was not found to be associated with a less favorable prognosis of work ability in 2 years' time.43 *This is* counterintuitive given that BMI is associated with an increased risk of KO14 and having a higher BMI as a KO patient results in a higher pain score.44 A possible explanation is that workers with KO and being overweight or obese have already structured their job in such a manner that they no longer need to perform the most demanding activities, given their body weight. Further research should be conducted to properly determine the influence of BMI on the prospects of the further career of the construction workers with KO. Moreover, the use of BMI has been debated in the literature as controversial.45 Although BMI might not be the most accurate and suitable measure for determining overweight or obesity, it is one of the most used measurements to screen overweight or obesity risks in various population groups and is well suitable for epidemiological settings such as the correlation between health outcomes and BMI.46 Therefore, measurement of BMI was deemed appropriate for the purpose of the current study, but future studies should investigate the influence of body composition on future work ability especially in physically demanding jobs.
To offer tailor-made prevention and interventions for male construction workers with KO, future research should focus on reducing physically demanding work activities and promoting high-intensity exercise for at least once a week. Previous research among construction workers showed that health promotion worksite interventions improved physical activity and may, thus, contribute to a higher ability to practice their current profession in the future.47 In addition, a Total Worker Health® intervention among construction workers showed that, at 6 months, the intervention group had reduced physically demanding activities and increased recreational physical activity.48 Unfortunately, limited evidence is available to support the suggestion that the use of preventive ergonomic measures to reduce the workload actually reduces the risk of associated musculoskeletal complaints.49 However, health impact assessments based on worksite measurements potentially show positive gains of ergonomics measures at the worksite among construction workers at risk of KO.50
## Limitations and Strengths
A major strength of our study is the presence of a high number of construction workers diagnosed with KO by an occupational physician. Moreover, investigating the combination of personal and work-related characteristics can also be seen as a strength, because both factors influence the ability to continue working as a construction worker with KO. Finally, work ability is measured using a validated method.24 Nevertheless, we faced three limitations in our study. First, this study had a cross-sectional design that does not allow investigating the causal relationship between the outcome and variables.51 Second, the method of diagnosing KO used by the various occupational physicians might have differed. This could have resulted in a more heterogeneous selection of workers who did not have KO as the primary cause of their knee complaints. Moreover, participation in the WHS is voluntary, which might have resulted in a healthy worker selection effect.41 Reasons for not participating or having a lower commitment could be that construction workers were afraid that a negative outcome regarding their future career opportunities might result, or they did not see the personal relevance to participate.41 A third limitation was that the WHS questionnaire made no explicit distinction between occupational physical activity and leisure-time physical activity.52 This may have introduced classification bias as it is challenging to distinguish whether exercise concerns occupational physical activity or leisure-time physical activity.52 An earlier study recommends physical activity assessment based on objective measures including, for example, accelerometers instead of self-reported questionnaire-based assessment.52 However, on the basis of the results of the current study, we assume that participants considered the question on exercise as leisure-time physical activity as most participants reported zero times per week of high-intensity exercise.
Finally, for the primary outcome measure of the self-reported prognosis of work ability to practice the current profession in 2 years' time, the categories “maybe” and “unlikely” were combined to form a dichotomous outcome measure for the analysis. A combination of those categories might be conservative, but we aimed to study the prognostic factors that enhance the construction workers' future work ability to prevent sick leave and early retirement targeting the group that may be at risk of being unable to do their current profession in 2 years' time and that believes to be certain to be unable to do their current profession in 2 years' time. Thus, the current study gives insight into the needs for tailoring future interventions to the prognostic factors of both the at-risk group and the group unlikely to do their current profession in the future.
## CONCLUSION
Dutch construction workers with clinically diagnosed KO who reported that, in 2 years' time, they would be able to practice their current job were working 36 to 45 hours per week, performing high-intensity exercises once or twice a week, were younger than 56 years, and performed less than two strenuous physical activities in their work compared with their counterparts who reported not being able to practice their current job in 2 years' time. To keep construction workers with KO at work, intervention studies should evaluate the effects of reducing strenuous work activities and promote leisure-time exercises.
## References
1. Colin R, Wild P, Paris C, Boini S. **Effect of joint exposure to psychosocial and physical work factors on the incidence of workplace injuries: results from a longitudinal survey**. (2021.0) **63** 921-930. PMID: 34238905
2. **What are the risks to minors who work in the construction industry?**. (2021.0) **63** e462-e463. PMID: 34029298
3. Dale AM, Rohlman DS, Hayibor L, Evanoff BA. **Work organization factors associated with health and work outcomes among apprentice construction workers: comparison between the residential and commercial sectors**. (2021.0) **18** 8899. PMID: 34501489
4. Umer W, Antwi-Afari MF, Li H, Szeto GPY, Wong AYL. **The prevalence of musculoskeletal symptoms in the construction industry: a systematic review and meta-analysis**. (2018.0) **91** 125-144. PMID: 29090335
5. Hulshof CTJ, Pega F, Neupane S. **The effect of occupational exposure to ergonomic risk factors on osteoarthritis of hip or knee and selected other musculoskeletal diseases: a systematic review and meta-analysis from the WHO/ILO joint estimates of the work-related burden of disease and injury**. (2021.0) **150** 106349. PMID: 33546919
6. Bahns C, Bolm-Audorff U, Seidler A, Romero Starke K, Ochsmann E. **Occupational risk factors for meniscal lesions: a systematic review and meta-analysis**. (2021.0) **22** 1042. PMID: 34911509
7. Le Manac'h AP, Ha C, Descatha A, Imbernon E, Roquelaure Y. **Prevalence of knee bursitis in the workforce**. (2012.0) **62** 658-660
8. Kuijer P, Burdorf A. **Prevention at work needed to curb the worldwide strong increase in knee replacement surgery for working-age osteoarthritis patients**. (2020.0) **46** 457-460. PMID: 32780145
9. Jarvholm B, From C, Lewold S, Malchau H, Vingard E. **Incidence of surgically treated osteoarthritis in the hip and knee in male construction workers**. (2008.0) **65** 275-278
10. Hubertsson J, Petersson IF, Thorstensson CA, Englund M. **Risk of sick leave and disability pension in working-age women and men with knee osteoarthritis**. (2013.0) **72** 401-405. PMID: 22679305
11. Hardenberg M, Speklé EM, Coenen P, Brus IM, Kuijer PPFM. **The economic burden of knee and hip osteoarthritis: absenteeism and costs in the Dutch workforce**. (2022.0) **23** 364. PMID: 35436874
12. Bannuru RR, Osani MC, Vaysbrot EE. **OARSI guidelines for the non-surgical management of knee, hip, and polyarticular osteoarthritis**. (2019.0) **27** 1578-1589
13. Van Zaanen Y, Hoorntje A, Koenraadt KLM. **Non-surgical treatment before hip and knee arthroplasty remains underutilized with low satisfaction regarding performance of work, sports, and leisure activities**. (2020.0) **91** 717-723. PMID: 32878525
14. Whittaker JL, Runhaar J, Bierma-Zeinstra S, Roos EM. **A lifespan approach to osteoarthritis prevention**. (2021.0) **29** 1638-1653
15. **STROBE checklists**
16. Katz JN, Arant KR, Loeser RF. **Diagnosis and treatment of hip and knee osteoarthritis: a review**. (2021.0) **325** 568-578. PMID: 33560326
17. Shane Anderson A, Loeser RF. **Why is osteoarthritis an age-related disease?**. (2010.0) **24** 15-26. PMID: 20129196
18. Altman R, Asch E, Bloch D. **Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association**. (1986.0) **29** 1039-1049. PMID: 3741515
19. Stubbs B, Aluko Y, Myint PK, Smith TO. **Prevalence of depressive symptoms and anxiety in osteoarthritis: a systematic review and meta-analysis**. (2016.0) **45** 228-235. PMID: 26795974
20. Toivanen AT, Heliövaara M, Impivaara O. **Obesity, physically demanding work and traumatic knee injury are major risk factors for knee osteoarthritis—a population-based study with a follow-up of 22 years**. (2009.0) **49** 308-314. PMID: 19946021
21. Wilkie R, Blagojevic-Bucknall M, Jordan KP, Lacey R, McBeth J. **Reasons why multimorbidity increases the risk of participation restriction in older adults with lower extremity osteoarthritis: a prospective cohort study in primary care**. (2013.0) **65** 910-919. PMID: 23225783
22. Connelly AE, Tucker AJ, Kott LS, Wright AJ, Duncan AM. **Modifiable lifestyle factors are associated with lower pain levels in adults with knee osteoarthritis**. (2015.0) **20** 241-248
23. Zheng H, Chen C. **Body mass index and risk of knee osteoarthritis: systematic review and meta-analysis of prospective studies**. (2015.0) **5** e007568
24. Łastowiecka E, Bugajska J, Najmiec A, Rell-Bakalarska M, Bownik I, Jędryka-Góral A. **Occupational work and quality of life in osteoarthritis patients**. (2006.0) **27** 131-139. PMID: 17094005
25. Kujala UM, Kettunen J, Paananen H. **Knee osteoarthritis in former runners, soccer players, weight lifters, and shooters**. (1995.0) **38** 539-546. PMID: 7718008
26. Jarvholm B, Lewold S, Malchau H, Vingard E. **Age, bodyweight, smoking habits and the risk of severe osteoarthritis in the hip and knee in men**. (2005.0) **20** 537-542. PMID: 16121763
27. Holmberg S, Thelin A, Thelin N. **Is there an increased risk of knee osteoarthritis among farmers? A population-based case-control study**. (2004.0) **77** 345-350. PMID: 15127209
28. Andersen S, Thygesen LC, Davidsen M, Helweg-Larsen K. **Cumulative years in occupation and the risk of hip or knee osteoarthritis in men and women: a register-based follow-up study**. (2012.0) **69** 325-330
29. McWilliams DF, Leeb BF, Muthuri SG, Doherty M, Zhang W. **Occupational risk factors for osteoarthritis of the knee: a meta-analysis**. (2011.0) **19** 829-839. PMID: 21382500
30. Manninen P, Heliovaara M, Riihimaki H, Suoma-Iainen O. **Physical workload and the risk of severe knee osteoarthritis**. (2002.0) **28** 25-32. PMID: 11871849
31. Jones GT, Harkness EF, Nahit ES, McBeth J, Silman AJ, Macfarlane GJ. **Predicting the onset of knee pain: results from a 2-year prospective study of new workers**. (2007.0) **66** 400-406. PMID: 16935910
32. Kievit AJ, van Geenen RC, Kuijer PP, Pahlplatz TM, Blankevoort L, Schafroth MU. **Total knee arthroplasty and the unforeseen impact on return to work: a cross-sectional multicenter survey**. (2014.0) **29** 1163-1168. PMID: 24524779
33. Verbeek J, Mischke C, Robinson R. **Occupational exposure to knee loading and the risk of osteoarthritis of the knee: a systematic review and a dose-response meta-analysis**. (2017.0) **8** 130-142. PMID: 28593068
34. Ilmarinen J. **The work ability index (WAI)**. (2007.0) **57** 160
35. de Zwart BC, Frings-Dresen MH, van Duivenbooden JC. **Test-retest reliability of the work ability index questionnaire**. (2002.0) **52** 177-181. PMID: 12091582
36. 36World Health Organization. WHO Guidelines on Physical Activity and Sedentary Behaviour. Geneva, Switzerland: World Health Organization; 2020.. *World Health Organization.* (2020.0)
37. Merkus SL, Coenen P, Forsman M, Knardahl S, Veiersted KB, Mathiassen SE. **An exploratory study on the physical activity health paradox—musculoskeletal pain and cardiovascular load during work and leisure in construction and healthcare workers**. (2022.0) **19** 2751. PMID: 35270444
38. Jayabalan P, Ihm J. **Rehabilitation strategies for the athletic individual with early knee osteoarthritis**. (2016.0) **15** 177-183. PMID: 27172082
39. Skou ST, Roos EM. **Good life With osteoArthritis in Denmark (GLA:D™): evidence-based education and supervised neuromuscular exercise delivered by certified physiotherapists nationwide**. (2017.0) **18** 72. PMID: 28173795
40. Iliades C. (2009.0)
41. Siebert U, Rothenbacher D, Daniel U, Brenner H. **Demonstration of the healthy worker survivor effect in a cohort of workers in the construction industry**. (2001.0) **58** 774-779
42. Chowdhury R, Shah D, Payal AR. **Healthy worker effect phenomenon: revisited with emphasis on statistical methods—a review**. (2017.0) **21** 2-8. PMID: 29391741
43. Robroek SJW, Järvholm B, van der Beek AJ, Proper KI, Wahlström J, Burdorf A. **Influence of obesity and physical workload on disability benefits among construction workers followed up for 37 years**. (2017.0) **74** 621-627
44. Raud B, Gay C, Guiguet-Auclair C. **Level of obesity is directly associated with the clinical and functional consequences of knee osteoarthritis**. (2020.0) **10** 3601. PMID: 32107449
45. Frankenfield DC, Rowe WA, Cooney RN, Smith JS, Becker D. **Limits of body mass index to detect obesity and predict body composition**. (2001.0) **17** 26-30. PMID: 11165884
46. Claessen H, Arndt V, Drath C, Brenner H. **Overweight, obesity and risk of work disability: a cohort study of construction workers in Germany**. (2009.0) **66** 402-409. PMID: 19196736
47. Viester L, Verhagen EALM, Bongers PM, van der Beek AJ. **The effect of a health promotion intervention for construction workers on work-related outcomes: results from a randomized controlled trial**. (2015.0) **88** 789-798. PMID: 25481382
48. Peters SE, Grant MP, Rodgers J, Manjourides J, Okechukwu CA, Dennerlein JT. **A cluster randomized controlled trial of a Total Worker Health® intervention on commercial construction sites**. (2018.0) **15** 2354. PMID: 30366387
49. van der Molen HF, Sluiter JK, Frings-Dresen MH. **The use of ergonomic measures and musculoskeletal complaints among carpenters and pavers in a 4.5-year follow-up study**. (2009.0) **52** 954-963. PMID: 19629810
50. Visser S, Van der Molen H, Kuijer P. **A health impact assessment of a preventive measure to reduce the risk of work-related low back pain, lumbosacral radiculopathy and knee osteoarthritis among construction workers in the Netherlands**. (2022.0) **13** S145
51. Kestenbaum B. (2019.0) 9-11
52. Coenen P, Huysmans MA, Holtermann A. **Towards a better understanding of the ‘physical activity paradox’: the need for a research agenda**. (2020.0) **54** 1055-1057. PMID: 32265218
|
---
title: 'The Impact of Education and Insurance Status on Past-Year Dental Visits Among
Older Mexican Adults: Results From the 2001 and 2012 Mexican Health and Aging Study'
authors:
- Jennifer Archuleta
- Hiram Beltrán-Sánchez
journal: Journal of Aging and Health
year: 2022
pmcid: PMC9988238
doi: 10.1177/08982643221086586
license: CC BY 4.0
---
# The Impact of Education and Insurance Status on Past-Year Dental Visits Among Older Mexican Adults: Results From the 2001 and 2012 Mexican Health and Aging Study
## Abstract
Objective: This study assessed past-year dental visits among older Mexican adults from the Mexican Health and Aging Study (MHAS). MHAS is a nationally representative cohort study of adults 50 years and older from Mexico. Methods: *Baseline data* from 2001 were compared with 2012 data. Binary logistic regression identified significant predictors of past-year dental visits. Decomposition techniques examined factors that contributed to changes in dental visits between 2001 and 2012. Results: Education and insurance status were positively associated with past-year dental visits, while decomposition results showed that population composition (more adults receiving insurance and higher education over time) contributed to the increased prevalence of dental visits between 2001 and 2012. Discussion: Education and insurance are critical factors that govern access to oral healthcare. After the provision of universal dental coverage by Mexico’s Seguro Popular in 2003, our results may reflect promising effects of such programs, which can inform future policies in Mexico and other settings.
## Introduction
Equitable access to dental care is necessary for supporting oral health throughout the life course. This is particularly important among older adults as their access to health care in general, and oral health in particular, is typically dependent on retirement benefits that may or may not include coverage for oral health (e.g., Medicare in the United States does not cover dental care). There are many health consequences derived from poor oral health. For example, poor oral health is the leading global cause of disability-adjusted life years among adults aged 65 and older (Marcenes et al., 2013). Oral diseases such as permanent tooth loss, untreated dental caries, and gum infections impact more than $77\%$ of older adults worldwide (Kassebaum et al., 2017). Without proper treatment, oral diseases can increase pain and lead to difficulty with eating, speaking, and swallowing (Furuta & Yamashita, 2013; Naka et al., 2014). Good oral hygiene practices such as brushing, flossing, and routine dental visits are important for the detection and treatment of oral health issues (Coll et al., 2020).
Some evidence suggests that older adults are less likely to have access to dental care which in turn impacts their oral health. For example, an assessment of 194 countries showed that compared to younger age groups, older adults, especially those from low- to middle-income countries, were most affected by oral diseases and had the lowest access to dental care (Peres et al., 2019). Similar trends in Mexico illustrated that only about half of Social Security beneficiaries aged 60 and older had received any oral health service within the past year in the early 2000s (Sánchez-García et al., 2007). Moreover, among older Mexican adults, the likelihood of permanent tooth loss and dental caries increased with each additional year of age (Sánchez-García et al., 2014). In addition, poor oral health status was found to be associated with frailty and lower quality of life among Mexican adults over age 70 (Castrejón-Pérez et al., 2017; Ortíz-Barrios et al., 2019). Despite the need for consistent preventive and emergency dental health services within this population, evidence from the early 2000s suggested that many Mexican older adults lacked access to any type of dental care (Sánchez-García et al., 2007).
The use of oral health services is disproportionately impacted by inequities in socioeconomic status (SES). Studies in 23 countries including Brazil, China, Mexico, and the United States found that older adults with lower levels of income, educational attainment, and dental insurance were less likely to utilize dental care services and were simultaneously more likely to have worse oral health outcomes (Almeida et al., 2017; Andrade et al., 2020; Hernández-Palacios et al., 2015; Sánchez-García et al., 2007). Prior research suggests a positive relationship between high educational attainment and dental visitations among older adults. In many middle- to high-income countries such as the United States, the use of dental care by older adults was more frequent among individuals with higher levels of education than those with fewer years of education (Almeida et al., 2017; Andrade et al., 2020; Ramírez et al., 2011). Among Mexican older adults, access to dental care was also found to be associated with greater educational attainment (Almeida et al., 2017; Andrade et al., 2020).
Additional factors such as comorbidities, gender, and age are also known to influence differential access to dental care. Previous research among older adults from Germany has shown that dental visitations declined with rising age but at a faster rate among women (Spinler et al., 2019). In the United States, females above age 65 were more likely to have a dental visit compared to older males (Gironda et al., 2013; Marchini et al., 2020). Studies in Mexico also found that women aged 60 and older were more likely to utilize dental health services compared to older adult men (Sánchez-García et al., 2007).
Comorbidities in older age were found to have an inverse association with dental care utilization. Older U.S. adults with mobility limitations and chronic conditions such as diabetes, heart problems, and hypertension had low compliance with seeking dental care in the past 12 months (Luo et al., 2018; Ramírez et al., 2011). Furthermore, dental visits and services were found to be less frequent among U.S. older adults with depression and anxiety and those exhibiting symptoms of loneliness and low social support (Burr & Lee, 2013; Okoro et al., 2012). Symptoms of frailty such as exhaustion, weight loss, weakness, and low physical activity were associated with fewer dental visits and poorer self-rated oral health status among older adults in Mexico (Castrejón-Pérez et al., 2012). These trends suggest that comorbidities can discourage older adult populations from seeking oral health treatment due to additional healthcare costs, time constraints, and physical burdens.
In the case of Mexico, important health care policy changes implemented in the early 2000s can also have an impact on access to dental care among older adults. Seguro *Popular is* a public health insurance program that was enacted in Mexico beginning in 2002. This program expanded healthcare services to over 50 million individuals who previously did not receive formal sector healthcare insurance from employers, private insurance, or social security. Unlike U.S. Medicare, Seguro Popular offers access to seven different categories of oral healthcare services for older adults including emergency and preventive dental visits (Comisión Nacional de Protección Social en Salud, 2012; Gutiérrez, 2014). Some evidence suggests that Seguro Popular has increased overall healthcare utilization and the use of diagnostic tests among older adults (Parker et al., 2018).
Along with health insurance, educational attainment among older adults also steadily increased in Mexico between 2001 and 2012. On average, educational attainment was found to be higher among the 2012 wave of older adults than the wave that was interviewed a decade earlier (Díaz-Venegas et al., 2019). The difference in educational attainment between both waves of older adults is an important demographic shift. Changes in the population composition might influence the extent to which an increasingly educated population accounts for an increase in prevalence of past-year dental visits. For example, individuals with more schooling may be more aware of the importance of seeking dental health care and also more likely to utilize the newly created Seguro popular which may lead to a higher prevalence of past-year dental visits in 2012 relative to 2001. Thus, if the proportion of past-year dental visits is greater in 2012 than in 2001, we must discern how much of this change in the prevalence is driven by the population composition (i.e., a larger proportion of the population who attain higher schooling in 2012 vs. 2001). Similarly, changes in this prevalence can be explained by how much the magnitude of this association changed over this period. For example, whether education became a stronger predictor of past-year dental visits in 2012 than in 2001. Decomposition analysis allows one to quantify the extent to which educational attainment (and other population characteristics) contributes to the observed changes in the prevalence of past-year dental visits. Improving trends within Mexico’s educational landscape warrant better understanding of how educational attainment impacts the ability of Mexican older adults to receive dental care.
While evidence suggests that education and insurance can lead to greater oral health utilization among older adults, few studies have explored how access to those resources and the prevalence of dental visits have shifted over time. This is particularly relevant in the case of Mexico as the Mexican government provided universal dental care benefits through Seguro Popular. One decade following the implementation of Seguro Popular, older adults in 2012 were more likely to have accessed any form of healthcare while also having acquired more education than a baseline wave of older adults in 2001. Thus, we might expect to find that changes in oral health access are a result of these changes in the population composition. In addition, oral health access might have also improved due to stronger statistical associations between education and dental visits, and insurance and dental visits over that same period. Together, these conditions enhance access to oral health services for incoming cohorts of older adults. Given that education and access to health insurance are the main factors associated with oral health, we hypothesize that past-year dental visits are positively associated with [1] higher educational attainment and [2] health insurance. We also speculate that changes in the population composition and covariate effects contributed to an increased prevalence of dental care access between 2001 and 2012.
## Data
The Mexican Health and Aging Study (MHAS) is a prospective panel study of adults aged 50 years and older from rural and urban areas across all 32 states in Mexico. MHAS collected data on sociodemographic characteristics, health conditions, aging, migration, and family networks. All questionnaires were administered in-person by trained interviewers from the Instituto Nacional de Estadística y Geografía (INEGI) of Mexico. The first three waves of the study were conducted in 2001, 2003, and 2012 (Wong et al., 2017). The 2001 baseline survey gathered information from a nationally representative sample of Mexican adults who were born before 1952. Households with at least one resident 50 years or older were randomly selected to participate in the study. Six states accounting for $40\%$ of all migrants to the United States were over-sampled. The 2001 sample size consisted of direct interviews ($$n = 14$$,154) and proxy interviews with other household members ($$n = 1032$$) with a response rate of $91.8\%$. Follow-up interviews for this same cohort were conducted in 2003, 2012, 2015, and 2018. A $93.3\%$ response rate resulted in a sample size of 14,250 individuals who participated in the 2003 survey. Last, the 2012 MHAS wave consisted of eligible participants from the 2003 follow-up sample ($$n = 14$$,283), new participant spouses ($$n = 385$$), and an added cohort of 6259 participants within the 50–59 age range (those born between 1952 and 1962). The response rate for the third wave was $88.1\%$ with a total sample size of 18,465 respondents. This study focused on data from baseline and 2012.
## Exclusion Criteria
Analytic data were restricted to direct interviews with participants aged 50 and older at the time of data collection and focused on outcomes from the 2001 and 2012 samples ($$n = 12$$,432 for the 2001 wave and $$n = 13$$,636 for the 2012 wave). In the 2001 analytic sample, 55 observations ($0.4\%$) were excluded due to missing data on the outcome variable and key predictors, resulting in a sample size of 12,377 participants. The same exclusion criteria were applied to the 2012 sample. Out of 13,636 eligible respondents in 2012, 200 observations ($1.5\%$) were dropped for having incomplete responses on key predictors and the dependent variable. The final analytic sample size for the 2012 analysis was 13,436 participants.
## Measures
Having at least one past-year oral health visit was the outcome variable in this analysis. This variable was captured from both 2001 and 2012 questionnaires with the item, “In the last year, about how many times have you seen a dentist?” This variable was consolidated into a binary variable (0 indicates no dental visits and 1 represents at least one oral visit) since only about $6\%$ of each sample had exactly two dental visits in a year, fewer than $5\%$ had three dental visits in a year, and fewer than $5\%$ of participants in each cohort had four or more past-year dental visits.
The following predictors were used in the 2001 and 2012 analyses: age, gender, years of education, health insurance status, net worth, urban residence, current smoking status, number of chronic conditions, mobility limitations, and having five or more depressive symptoms (Castrejón-Pérez et al., 2012; Ramírez et al., 2011; Sánchez-García et al., 2007). Age in years was a continuous variable that was gathered from the demographic section of the survey. Gender was captured as a dichotomous categorical variable from the demographic section of both 2001 and 2012 questionnaires. Response options were “Male” or “Female.” Educational attainment was operationalized as the number of years of completed education by respondents in both survey waves. For this analysis, education was recoded from a continuous variable into a 5-item categorical variable with the following groupings: “0 years of education,” “1–5 years,” “6 years,” “7–9 years,” and “10 or more years.” Next, health insurance included the following options of insurance programs in Mexico: Social Security/Instituto Mexicano del Seguro Social, Institute for Social Security and Services for State Workers/ISSSTE, Pemex/Defensa/Marina, private insurance, or other types of insurance. Respondents without any health insurance were coded as “0” and respondents with any form of public or private insurance were coded as “1.” We used an indicator of wealth constructed by the MHAS team (see Wong et al., 2007). Net worth is based on the monetary value of all assets (including businesses, land, housing, stocks and bonds, and savings) minus debts for individuals (or for the couple if the respondent was married/cohabiting). The MHAS project estimated wealth values while imputing missing values in the components of wealth with the method of sequence regressions (see Wong et al., 2007). The imputation method has several advantages: allowing variable values to be zero, accounting for other imputed variables, and incorporating responses based on categorical responses (“unfolding brackets”). We use four categories of wealth based on quartiles of the distribution. For stratified analysis, the wealth categories were created by gender and age. Next, urban place of residence was a dichotomous variable that defined urbanicity as a primary locality with a population of 15,000 people or more (Salinas et al., 2010). Lifestyle risk factors such as current smoking were defined as “currently smokes cigarettes” (Wong et al., 2008).
The next domain of predictors represents health variables that have an inverse association with past-year dental visits. The chronic conditions variable was a composite of seven self-reported items that indicated “yes” to having been diagnosed with any of the following conditions: Hypertension, diabetes, cancer, respiratory problems, heart problems, stroke, and arthritis. The number of chronic conditions was summed for each respondent, then recoded into a 3-item categorical variable (Díaz-Venegas et al., 2017). Response categories were “None” for individuals with no diagnosis of chronic disease, “One” for listing one chronic disease, and “Two or more” for individuals with at least two chronic conditions.
Next, the mobility limitations variable was collected as a series of nine self-reported items related to activities of daily living. The mobility limitation index was constructed according to Long and Pavalko [2004]. The questionnaire item asked, “Please tell me if you have any difficulty in doing each of the daily activities that I am going to read. Do not include difficulties that you believe will last less than 3 months.” Response options were “yes,” “no,” “can’t do,” doesn’t do,” “refused,” or “don’t know” for the following activities: Walking several blocks, walking one block, sitting for 2 hours, climbing several flights of stairs without resting, stooping/kneeling/crouching, extending arms above shoulder level, pushing or pulling large objects like a chair, lifting or carrying objects that weigh over 5 kg, and picking up a 1-peso coin from a table. In our analysis, the mobility variable was recoded as a binary variable (yes or no). Individuals that answered “no” to any of these items were coded as having no mobility limitations, and those who responded “yes,” “can’t do,” or “doesn’t do” were coded as having one or more mobility limitations.
Last, depressive symptoms were based on the mental health index used by Torres and Wong [2013]. The number of depressive symptoms was a composite score calculated from nine questions with a “yes” or “no” response. Each item asked whether a majority of the time respondents felt the following during the past week: Depressed, everything was an effort, restless sleep, happy, lonely, enjoyed life, sad, tired, and had a lot of energy. Items with a “yes” response for “depressed,” “effort,” “restless sleep,” “lonely,” “sad,” and “tired” were coded as “1.” “ Happy,” “enjoyed life,” and “had a lot of energy” were reversed coded so that “no” responses were coded as “1.” All nine items were summed into a composite score with a range of 0–9. The depression score was then converted into a binary variable by using a value of one for those with a score of five or more or zero otherwise (Torres & Wong, 2013).
## Statistical Analyses
Univariate analyses were conducted for each wave to assess the distributions of each predictor (age, education, net worth, health insurance, urban residence, smoking status, chronic conditions, mobility limitations, and depressive symptoms) and the outcome variable (having at least one past-year dental visit) (Table 1). As shown in Table 1, the population composition by education significantly differs by age and sex between 2001 and 2012. For example, there is a higher fraction of people with 7+ years of education in 2012 than in 2001 among people aged 50–54 for both women and men, but a non-significant difference in 7+ years of education over time among women aged 65+. Similarly, there is a statistically significantly higher fraction of insured (and lower fraction of non-insured) population in 2012. We thus conducted stratified analysis by gender (female and male) and age group (50–64 years and 65 years and above) to assess the differential impact of educational attainment in these population subgroups. Binary logistic regressions evaluated the association between past-year dental visits and the selected predictors in each wave. Variables were incrementally added to each statistical model. Goodness-of-fit was assessed across the logistic regression models using log likelihood-ratio tests, Akaike’s Information Criteria (AIC), and Bayesian Information Criteria (BIC). The final logistic regression analyses included all predictors (referred to as the full model).Table 1.Baseline characteristics from 2001 ($$n = 12$$,377) and 2012 ($$n = 13$$,436) MHAS samples*.Aged 50–64Aged 65+20012012Change 2001–201220012012Change 2001–2012Variable% [1]% [2][2−1]p-value% [1]% [2][2−1]p-value Female One or more dental visits in the past year (dependent variable)31.039.88.8<.00121.229.58.3<.001 Age** mean years (SD)56.3 (4.3)57.4 (4.2)1.1<.00172.7 (6.4)73.2 (6.6)0.5.015Years of education 0 years18.110.9−7.2<.00132.425.0−7.4<.001 1–5 years36.026.6−9.4<.00137.838.10.3.854 6 years20.624.74.1<.00114.818.13.3.003 7–9 years17.323.96.6<.00110.912.71.8.063 10 or more years8.014.06.0<.0014.16.12.0.003Insurance type No insurance32.210.6−21.6<.00130.58.9−21.6<.001 Private or other insurance4.84.6−0.2.6713.63.70.1.070 Seguro popular (in 2012 only)29.825.6 ISSSTE/PEMEX employers17.417.80.4.61122.521.2−1.3.289 IMSS/social security51.346.9−4.4<.00151.452.61.2.430Net worth 1st quartile22.223.51.3.15332.226.9−5.3<.001 2nd quartile25.826.40.6.51222.223.00.8.486 3rd quartile26.225.7−0.5.66323.925.11.2.354 4th quartile25.924.3−1.6.11121.724.93.2.012 Urban residence71.172.61.5.15271.173.82.7.039 Current smoker10.89.1−1.7.0127.34.4−2.9<.001Number of chronic conditions None34.936.61.7.10926.623.6−3.0.017 One38.136.1−2.0.07137.738.60.9.508 Two or more27.027.30.3.80535.737.82.1.151 Has 5 or more depressive symptoms40.237.9−2.3.03147.642.1−5.5<.001 Has at least one mobility limitation65.066.81.8.09879.883.53.7.001 Male One or more dental visits in the past year (dependent variable)26.732.35.6<.00119.728.08.3<.001 Age** mean years (SD)56.2 (4.2)57.9 (4.3)1.7<.00172.7 (6.4)73.1 (6.7)0.4.054Years of education 0 years14.17.5−6.6<.00129.420.4−9.0<.001 1–5 years30.821.7−9.1<.00140.735.9−4.8.001 6 years22.722.3−0.4.72714.320.66.3<.001 7–9 years14.221.06.8<.0016.910.33.4<.001 10 or more years18.227.59.3<.0018.812.94.1<.001Insurance type No insurance38.115.6−22.5<.00135.110.9−24.2<.001 Private or other insurance5.05.40.4.5694.04.00.0.942 Seguro popular (in 2012 only)28.025.6 ISSSTE/PEMEX employers15.015.80.8.42516.818.71.9.098 IMSS/social security46.044.7−1.3.31050.951.40.5.767Net worth 1st quartile19.723.74.0<.00123.721.5−2.2.083 2nd quartile25.024.9−0.1.93125.725.2−0.5.729 3rd quartile26.026.00.0.97125.325.30.0.949 4th quartile29.425.4−4.0.00125.328.12.8.048 Urban residence69.570.81.3.27063.167.84.7.001 Current smoker29.522.6−6.9<.00120.614.2−6.4<.001Number of chronic conditions None55.653.2−2.4.07241.239.2−2.0.176 One29.830.40.6.64834.934.8−0.1.920 Two or more14.616.41.8.05923.926.02.1.107 Has 5 or more depressive symptoms20.217.9−2.3.02030.224.6−5.6<.001 Has at least one mobility limitation42.642.1−0.5.67467.265.2−2.0.160**p-values are estimated from logistic regressions, except for age for which OLS is used, taking into account the complex survey design (sampling weights).Source: Mexican Health and Aging Study (MHAS), 2001–2012.*Percentages are weighted to reflect the national Mexican older adult population in 2001 and 2012, respectively: weighted N in 2001= 778,055; N in 2012= 905,069.
## Decomposition Analysis (Using the Oaxaca–Blinder Procedure)
The Oaxaca–Blinder procedure is a regression-based decomposition analysis that allow us to explore differences in the prevalence of past-year dental visits between the 2001 and 2012 MHAS. This procedure assessed how changes in the prevalence of past-year dental visits between 2001 and 2012 were explained by [1] changes in the composition or endowment of the predictors within each cohort and [2] changes in the impact of each coefficient from each cohort (Jann, 2008). The following equation displays how the population endowment and impact of each covariate contributed to the difference in the prevalence of past-year dental visits between cohort years: [1] D= ΔAgeβAt0+ ΔEducationβEt0+ ΔInsuranceβIt0+ ΔHealthHt0+ Aget0ΔβA+ Educationt0ΔβE+ Insurancet0ΔβI+ Healtht0ΔβH+Constant+Unexplained In equation [1], D represents the difference in the prevalence of past-year dental visits from 2001 to 2012, t 0 represents the 2001 baseline wave, and Δ is the change between 2001 and 2012. The decomposition leads to additive coefficients for each variable. To simplify the presentation of results, we collapse) health variables into a single “health” domain (i.e., adding up coefficients from equation [1]). Full results from equation [1] are shown in Appendix 3. The decomposition shown in equation [1] partitions changes in the prevalence of past-year dental visits between 2001 and 2012 into two counterfactual scenarios. First, it holds the beta coefficients constant at their 2001 level (i.e., beta coefficients from the regression in 2001) while the population composition changes over time (first line of equation [1]), and second, it holds the population composition constant at its 2001 level while the beta coefficients change over time (second line of equation [1]). As it is always the case in regression approaches, there is a residual term which in this case corresponds to differences in the prevalence of past-year dental visits that are not explained by the predictors (third line of equation [1]). All analyses used sampling weights and were conducted using STATA version SE/15.1 statistical software.
There are different ways to interpret decompositions results, here we offer a general description and further elaborate it in the results section. In counterfactual analysis, we are interested in answering the hypothetical question “what if X had not changed, how much of the observed change in the prevalence of past-year dental visits is due to changes in Z.” Thus, in the case of educational attainment, one can interpret results from line 1 in equation [1] as follows: how much of the increase in the prevalence in past-year dental visits is due to changes in the fraction of people with an educational attainment between 2001 and 2012 if the impact of education on said prevalence remains constant over the time period. Similarly, line 2 from equation [1] would indicate: how much of the increase in the prevalence in past-year dental visits is due to changes in the impact of education between 2001 and 2012 if there had not been an increasing fraction of people with more education over time (i.e., the fraction of people by education remains constant at 2001-level over the time period).
## Descriptive Statistics
Table 1 displays baseline characteristics for the 2001 ($$n = 12$$,377) and 2012 ($$n = 13$$,436) MHAS data. From these results, we observed that past-year dental visits significantly increased between 2001 and 2012 across all gender and age groups. The prevalence of past-year dental visits was lowest among males aged 65 and older ($19.7\%$ in 2001 and $28.0\%$ in 2012) and largest among females aged 50 to 64 ($31.0\%$ in 2001 and $39.8\%$ in 2012). *Males* generally had a lower prevalence of past-year dental visits compared to females, but with a declining gap among the 65+ age group.
Regarding educational attainment, more participants attained higher levels of education in 2012 compared to 2001, particularly among younger older adults (aged 50–64). The largest increase was observed among males aged 50 to 64: the proportion of individuals in this group who attained 10 or more years of education increased from $18.2\%$ in 2001 to $27.5\%$ in 2012. Higher levels of education were consistently lowest among females versus males and individuals aged 65 and older compared to the 50 to 64 age group.
Trends in insurance status showed that a significantly larger proportion of the population was insured in 2012 compared to 2001, with a larger increase among people aged 65+. Across this period, insurance rates among 50- to 64-year-old females rose from $67.8\%$ to $89.4\%$, and among 65+ females insurance status increased from $69.5\%$ to $91.1\%$. Among males aged 50- to 64- and 65+, insurance status between cohorts increased from $62.0\%$ to $84.4\%$ and $64.9\%$ to $89.2\%$, respectively. Importantly, the distribution of net worth in 2012 moved to the left among people aged 50–64 and to the right among those aged 65+ relative to 2001. This implied that there was a smaller fraction of people at the higher quartiles of net worth in 2012 among younger adults (fourth quartile: $25.9\%$ in 2001 vs. $24.3\%$ in 2012), but the opposite is true among older adults (fourth quartile: $21.7\%$ in 2001 vs. $24.9\%$ in 2012). Similarly, the prevalence of health conditions significantly increased over time. Depressive symptoms, having two or more chronic conditions, and having mobility limitations all had the highest prevalence among the 65+ age groups versus the lower age category and females compared to males.
## Multiple Logistic Regression
Odds ratios from binary multiple logistic regressions by age group, period and sex are displayed in Table 2. These findings show that education has the strongest association with increased odds of having a past-year dental visit across all subgroups. These trends occurred along a gradient, illustrating that the likelihood of having a past-year dental visit increased incrementally between each educational category from 0 to 10+ years of education. Similar results were found with higher levels of net worth. We also found significant and marginally significant (p-values below.06) associations between having insurance and receiving a past-year dental visit though results were less significant in 2012. Receiving past-year dental care was also more likely among non-smokers versus current smokers (except among 50–64 year-old females in 2001, in which the opposite result was observed) and more prevalent among those who live in urban versus rural areas. Table 2.Weighted logistic regression results: Odds ratios of having a past-year dental visit among Mexican adults aged 50 and older by age group and survey year. Aged 50–64Aged 65+Variable2001201220012012 Female Age0.99.98*.96*.97**Years of education 0 years (reference) 1–5 years1.23+1.35+1.281.06 6 years1.63**1.47*1.83*1.56* 7–9 years2.61**1.95**2.75**2.26** 10 or more years3.43**3.92**2.90**3.67**Net worth 1st quartile (reference) 2nd quartile1.27+1.131.011.15 3rd quartile1.30+1.111.021.19 4th quartile1.71**1.36*1.291.25Insurance status No insurance (reference) Has insurance1.24+1.46*1.69*1.34Urban residence No (reference) Yes1.50**1.53**1.341.30+Current smoker No (reference) Yes1.28+1.060.770.65+Number of chronic conditions None (reference) One1.081.031.091.00 Two or more1.29+1.031.191.23Has at least one mobility limitation No (reference) Yes.961.131.100.83Has 5 or more depressive symptoms No (reference) Yes1.17+.901.16.98 Constant1.062.62+3.676.01* Analytic sample size4406429822993396 Population sample size271,174287,810119,123202,724 Male Age.991.01.98+.98*Years of education 0 years (reference) 1–5 years1.87**1.471.58*1.64** 6 years1.91**1.73+2.13**1.80** 7–9 years2.64**2.44**2.57**2.61** 10 or more years3.95**3.29**3.05**3.51**Net worth 1st quartile (reference) 2nd quartile1.151.181.001.03 3rd quartile1.251.141.051.11 4th quartile1.89**1.80**1.65*1.38+*Insurance status* No insurance (reference) Has insurance1.31*1.261.38+1.30Urban residence No (reference) Yes1.24+1.50**1.321.32+Current smoker No (reference) Yes.96.95.70+.73+Number of chronic conditions None (reference) One1.33*1.41**0.991.39* Two or more1.58**1.35+1.321.39*Has at least one mobility limitation No (reference) Yes.991.081.34+1.11Has 5 or more depressive symptoms No (reference) Yes1.25+.801.40+.91 Constant.82.421.462.53 Analytic sample size3587285220852890 Population sample size254,913214,122132,845200,413Source: Mexican Health and Aging Study (MHAS), 2001–2012. ** $p \leq .001$, * $p \leq .01$, + $p \leq .05.$
Results for the number of chronic conditions, mobility limitations and depressive symptoms show no statistically significant associations with past-year dental visits, except for a few cases. For example, having more chronic conditions increases the odds of past-year dental visit among males but not among females. Additionally, in contrast to the 2012 wave, groups from 2001 had greater odds (between $17\%$ and $40\%$) of visiting a dentist if they had five or more depressive symptoms.
## Decomposition Results: Differences in Prevalence of Past-Year Dental Visits Between 2001 and 2012
The Oaxaca–Blinder decomposition results by sex and age group are shown in Table 3 (results for each individual covariate are shown in the Appendix1,2,3). Panel (A) in the table shows the prevalence of past-year dental visits in each year and the total change over time, which is shown to have increased over time. Panels (B) and (C) show how much of the increase in this prevalence is attributed to changes in the composition of the covariates while holding the impact of the covariates constant over time, changes in the effects of the covariates while holding the population composition constant at 2001-levels, respectively. The sum of the individual components in panels (B) and (C) equals the total change in the prevalence of oral health visits shown in panel (A).Table 3.Weighted Oaxaca–Blinder decomposition results: Assessing the composition and effect of each predictor in the prevalence of past-year dental visits between the 2001 and 2012 in MHAS waves. Aged 50–64Aged 65+ Female A) prevalence% *With a* past-year dental visit% *With a* past-year dental visit 200131.0221.17 201239.7829.52 Total change [2012–2001]8.778.35B) Effects of changes in the composition of covariates in 2012 versus 2001 (%) Age−0.51−0.36 Education2.431.07 Net worth−0.100.20 Insurance1.771.20 Current smoking−0.020.24 Urban0.130.13 Health domain0.09−0.03C) Effects of the changes in the effect of the covariates in 2012 versus 2001 (%) Age−13.4611.37 Education−1.11−2.01 Net worth−2.240.97 Insurance3.13−3.79 Current smoking−0.37−0.14 Urban0.41−0.40 Health domain1−1.40−5.70 Constant20.495.50 Unexplained−0.470.10 Male A) prevalence% *With a* past-year dental visit% *With a* past-year dental visit 200126.7419.68 201232.3327.97 Total change (2012–2001)5.598.28B) Effects of changes in the composition of covariates in 2012 versus 2001 (%) Age0.16−0.20 Education2.541.70 Net worth−0.440.15 Insurance0.991.10 Current smoking0.070.36 Urban0.100.22 Health domain10.230.17C) Effects of the changes in the effect of the covariates in 2012 versus 2001 (%) Age14.82−1.73 Education−2.97−0.02 Net worth−0.52−0.24 Insurance−0.67−0.94 Current smoking−0.070.11 Urban2.86−0.03 Health domain−1.20−1.67 Constant−9.589.88 Unexplained−0.73−0.58Source: Mexican Health and Aging Study (MHAS), 2001–2012.a Health domain includes adding decomposition coefficients from equation [1] for the number of chronic conditions, mobility limitations, and depressive symptoms.
## Population Composition
For both females and males, changes in the fraction of the population with education and insurance had the largest positive impact in explaining the increasing prevalence of a past-year dental visit over time. The magnitude of this impact is generally larger among younger cohorts (aged 50–64). To simplify the description of results, we focus first among women aged 50–64 and then generalize the findings. In this age group, there was a higher fraction of people with more education in 2012 which accounted for about 2.43 percentage points of the $8.77\%$ increase in the prevalence of having had a past-year dental visit between 2001 and 2012. Similarly, a higher fraction of people with health insurance in 2012 among women aged 50–64 accounted for an additional 1.77 percentage points of the $8.77\%$ change in the prevalence past-year dental visits. Thus, the fact that health insurance became more available in recent years led to an increase in the fraction of people who had a past-year dental visit in recent times. Interestingly, results also indicate there was a negligible change in the prevalence of health conditions and in the fraction of people living in urban areas among women aged 50–64 between 2001 and 2012; thus, these covariates accounted for a very small magnitude of the increase in the prevalence of a past-year dental visits, 0.13 and 0.09 percentage points, respectively. In contrast, changes in the age structure of the population and in the net worth distribution contributed to a reduction (negative sign) in the prevalence. This suggests that had the population not become older, or with a lower population fraction in the upper quartiles of the net worth distribution, the observed increase in the prevalence of a past-year dental visit would have been slightly higher by about 0.51 and 0.10 percentage points, respectively, than the observed change of 8.77 % (assuming the effect of the other covariates remained stable between 2001 and 2012).
Results for women aged 65+ are similar to those aged 50–64. For example, had the population not become older in 2012 than 2001, the increase of $8.35\%$ in the prevalence would have been higher by about 0.36 percentage points. Nonetheless, due to an increasing fraction of older adult women with higher education, with access to health insurance, or in the upper quartiles of the net worth distribution, the prevalence went up by about 1.07, 0.20, and 1.2 percentage points, respectively. Results for men are similar to those among women. For example, changes in educational attainment had the largest contribution to the increasing prevalence of past-year dental visits for both younger and older adult males, followed by positive contributions from all other covariates, and negative contribution from net worth among those aged 50–64 but positive contribution to those aged 65+ due to the changing net worth distribution over time.
## Effect of Covariates
Although compositional changes had some relevance in explaining changes in the prevalence of a past-year dental visit, the largest magnitudes in the contributions came from increases in the effect of the covariates. These results can be interpreted as how much of the increase in the prevalence in past-year dental visits is due to changes in the impact of the covariates between 2001 and 2012 if the population composition remained as it was in 2001. That is, a negative sign in the contribution of effect of covariates indicates that the association of a covariate with past-year dental visits is less strong in 2012 than in 2001; in contrast, a positive sign in the contribution suggests a stronger association in 2012 than in 2001. In other words, a given covariate is more (less) predictive of past-year dental visits in 2012 if its contribution is positive (negative). For instance, the association of age with past-year dental visits varies between women and men and between younger and older adults. This association is less strong in 2012 than in 2001 among women aged 50–64 (negative sign), but the opposite is true among those aged 65+. In contrast, among men, age has a stronger association with past-year dental visits in 2012 in younger adults and a lesser one among those aged 65+. These results indicate that a stronger age association with past-year dental visits among older women and younger adult men contributed to about 11.37 and $14.82\%$ points to their corresponding prevalence increase. In contrast, had the association with age being stronger in 2012 among women aged 50–64 and men aged 65+, the prevalence of past-year dental visits would have increased by about 13.46 and $1.73\%$ points, respectively.
The impact of education, net worth, insurance, and health conditions is consistently less strong in 2012 among men, but among women these impacts vary by age group. For example, for women aged 65+, net worth is more predictive of past-year dental visits in 2012 contributing to an increase in the prevalence of about $0.97\%$ points. Nonetheless, all other covariates show a less strong association in 2012 suggesting that had their link with past-year dental visits be more predictive in 2012, we would have observed an increase of about $11.58\%$ points in past-year dental visits from education (2.01), insurance (3.79), and health conditions (5.79), respectively (11.58 = 2.01+ 3.79 + 5.79). Interestingly, access to insurance led to a stronger association with past-year dental visits in 2012 only among women aged 50–64 (with a $3.13\%$ point contribution). For the other age groups, insurance had a lower impact on the prevalence in 2012 than in 2001.
## Summary
Ultimately, a major demographic shift occurred between 2001 and 2012 in which Mexican older adults from 2012 attained higher levels of education and had higher insurance enrollment than the baseline wave. This change in population structure contributed to an increase in the prevalence of dental visits from 2001 to 2012. Nonetheless, population aging and the net worth distribution dragged down some of the progress and if it had not been for these population changes, the prevalence of past-year dental visits could have been higher. The role of the covariate effects by education and insurance were less consistent. In most age and gender groups, the covariate effects of insurance and education became less strong in 2012, except for women aged 50–64 for which insurance was a stronger predictor of past-year dental visits in 2012. Age also play a major role with stronger impact on the prevalence among women aged 65+ and among men aged 50–64.
## Discussion
In this paper, we study the association between socioeconomic and health factors on the prevalence of past-year dental visits among older Mexican adults in 2001 and 2012. As a result of major changes in health policies in the country during the period of study and the continuing aging of the Mexican population, we hypothesize that past-year dental visits are positively associated with higher educational attainment and health insurance. We report three main findings. First, in both 2001 and 2012, there were significant associations between higher educational attainment and past-year dental visits. Second, insurance status was also positively associated with past-year dental visits in both years. Third, the prevalence of past-year dental visits increased between 2001 and 2012 and was partly explained by changes in the population composition and also be changes in the association with the covariates, most notably by the covariate effect of age. For example, changes in the population composition, in which more individuals had access to insurance and higher education in 2012 compared to 2001, contributed to the shift in the prevalence of past-year dental visits.
Our results illustrate a significant positive association between educational attainment and past-year dental visits. This is consistent with previous literature, for example, results from 2001 and 2012 show that older Mexican adults with higher educational attainment were more likely to use any type of dental service in the past year compared to those with fewer years of education (Almeida et al., 2017; Andrade et al., 2020). Our findings further elucidate this link by showing that the increasing prevalence of having a past-year dental visit is in part due to the underlying changes in the population composition of Mexican adults who are achieving older age having attained higher levels of education than previous generations. Moreover, our decomposition results further elucidate that the association between education and past-year dental visits attenuated over time so the impact of education was less important in explaining the increasing prevalence of dental visits. Thus, our findings may suggest that dental care utilization is promoted by the advantages in knowledge and resources that are made more accessible through higher levels of education.
Dental care utilization had differential effects by gender and age group. Compared to adults aged 50–64 years, adults aged 65+ years had an overall lower proportion of dental visits. Males above age 64 had the lowest proportion of dental visits. Similar to findings from a longitudinal study of older adults in Germany (Spinler et al., 2019), our results revealed that female participants were consistently more likely than males to have a past-year dental visit, but the prevalence of dental visits among females decreased at a faster rate than males with older age. These data suggest that although females may have previously had more dental care access or better oral health-seeking behaviors compared to males, other unmeasured factors such as changes in employment status, widowhood, or low social support may lead to increasing disparities in dental care access among females later in life.
From our decomposition results, it was noted that the prevalence of past-year dental visits across all age and gender subgroups increased between 2001 and 2012. We found that changes in the population composition, specifically greater proportions of educational attainment and those with insurance, within this period partly explained this increase in the prevalence of dental visits. As expected, a larger proportion of older adults had insurance following the enactment of Seguro Popular in 2002. These findings suggest that the expansion of resources such as education and public health programs became more accessible in later generations and ultimately, contributed to improvements in dental care access and utilization. However, the covariate effects of education and insurance did not increase between 2001 and 2012. These effects declined because of the dominating contributions of the impact of age, which was found to be associated with a reduction in the prevalence of a dental visit. Nonetheless, the demographic shift towards higher education and insurance status still had an important role in increasing the prevalence of past-year dental visits over that decade.
This study adds to the limited body of knowledge on the impact that insurance and education have on oral health care access in aging populations. No other studies have assessed this topic using decomposition techniques to compare the impact of population composition from the associations with known risk factors between different cohorts of older adults. In particular, the enactment of Seguro Popular between each cohort wave was a notable event in Mexico’s history that transformed access to health services in general, and dental services in particular, across the country. Despite the comprehensive provision of oral healthcare in Mexico, our results showed that still, fewer than $40\%$ of the Mexican population over age 50 received any dental care in the past 12 months (Table 1). Disparities in oral healthcare by age and gender also persist. Tailoring public health programs to bridge these gaps could lead to immediate and long-term benefits that effectively, improve rates of access to oral health care among older adults.
This study has some limitations. First, all measures relied on self-reported data. Recall bias and stigma may have influenced responses about sensitive health topics. To minimize bias, trained researchers followed a uniform methodology to gather in-person information at each respondent’s household. Second, the variable we constructed to assess chronic conditions captures only respondents who are aware of having a condition and may, therefore, undercount undiagnosed cases who have not received any recent care. Third, the operationalization of past-year dental visits entails any form of oral health treatment. Differentiating between preventive and emergency oral health treatment in future analyses would better illustrate the need for specific oral health services. Last, the MHAS questionnaire did not have any indicators for oral health morbidities such as tooth decay, gum disease, and tooth loss or whether a past-year dental visit was for a preventive or reactive care. We are thus unable to further analyze dental care more broadly. Although important, inclusion of oral health condition variables in addition to oral health utilization requires extending an already lengthy survey questionnaire.
Poor oral health is a global disease burden that could be mitigated by routine utilization of preventive or emergency dental visits. Disparities in access to dental care are readily seen among Mexican older adults with lower educational attainment and lack of health insurance. Especially in the midst of the COVID-19 pandemic, the disruption of dental services has burdened Mexican oral health providers and patients (Casillas Santana et al., 2021). Although new sterilization procedures prevent risk of COVID-19 infection, many dental providers have less time and fewer resources to cater to their patient populations than before the pandemic (Casillas Santana et al., 2021). Among the $20\%$ of Mexicans that reside in rural communities, issues of dental care access were already exacerbated by provider shortages (The World Bank, 2018). In 2013, the provider–patient ratio for the general Mexican population was only 12 dentists per 100,000 population (Malmö University & World Health Organization, n.d.). Although this study did not specifically examine provider shortages as a form of access to dental care, future work can pursue ways to support the oral health workforce as they respond to both existing barriers to access and new pandemic-related challenges.
Suggestions for future research include an assessment of universal dental care programs in Mexico and other countries. The impact that disabilities and social support have on oral health-seeking behaviors later in life is another critical area of study. In addition, updated surveillance instruments could collect information not only on oral health utilization, but also on key oral health outcomes such as dental caries, permanent tooth loss, and oral hygiene. Moreover, as the proportion of older Mexican adults residing in the United *States is* increasing, further research can explore the impact of migration on oral health utilization within aging migrant populations (Gonzalez-Barrera & Lopez, 2013). Oral health requires proper maintenance throughout the life course. Providing universal access to dental treatment and removing barriers to care are necessary for meeting the oral health needs of underserved older adults.
## ORCID iD
Jennifer Archuleta https://orcid.org/0000-0003-2044-3688
## References
1. Almeida A. P. S. C., Nunes B. P., Duro S. M. S., Facchini L. A.. **Socioeconomic determinants of access to health services among older adults: A systematic review**. *Revista de Saúde Pública* (2017.0) **51** 50. DOI: 10.1590/s1518-8787.2017051006661
2. Andrade F. B. de, Antunes J. L. F., Andrade F. C. D., Lima-Costa M. F. F., Macinko J.. **Education-related inequalities in dental services use among older adults in 23 countries**. *Journal of Dental Research* (2020.0) **99** 1341-1347. DOI: 10.1177/0022034520935854
3. Burr J. A., Lee H. J.. **Social relationships and dental care service utilization among older adults**. *Journal of Aging and Health* (2013.0) **25** 191-220. DOI: 10.1177/0898264312464497
4. Casillas Santana M. Á., Martínez Zumarán A., Patiño Marín N., Castillo Silva B. E., Sámano Valencia C., Salas Orozco M. F.. **How dentists face the COVID-19 in Mexico: A nationwide cross-sectional study**. *International Journal of Environmental Research and Public Health* (2021.0) **18** 1750. DOI: 10.3390/ijerph18041750
5. Castrejón-Pérez R. C., Borges-Yáñez S. A., Gutiérrez-Robledo L. M., Ávila-Funes J. A.. **Oral health conditions and frailty in Mexican community-dwelling elderly: A cross sectional analysis**. *BMC Public Health* (2012.0) **12** 773. DOI: 10.1186/1471-2458-12-773
6. Castrejón-Pérez R. C., Jiménez-Corona A., Bernabé E., Villa-Romero A. R., Arrivé E., Dartigues J.-F., Gutiérrez-Robledo L. M., Borges-Yáñez S. A.. **Oral disease and 3-year incidence of frailty in Mexican older adults**. *The Journals of Gerontology: Series A* (2017.0) **72** 951-957. DOI: 10.1093/gerona/glw201
7. Coll P. P., Lindsay A., Meng J., Gopalakrishna A., Raghavendra S., Bysani P., O’Brien D.. **The prevention of infections in older adults: Oral health**. *Journal of the American Geriatrics Society* (2020.0) **68** 411-416. DOI: 10.1111/jgs.16154
8. Comisión Nacional de Protección Social en Salud
. (2012). Catálogo universal de servicios de Salud 2012 (CAUSES). http://data.salud.cdmx.gob.mx/portal/seguro_popular/index/pdf/causes2012.pdf. *Catálogo universal de servicios de Salud 2012 (CAUSES)* (2012.0)
9. Díaz-Venegas C., Sáenz J. L., Wong R.. **Family size and old-age wellbeing: Effects of the fertility transition in Mexico**. *Ageing and Society* (2017.0) **37** 495-516. DOI: 10.1017/S0144686X15001221
10. Díaz-Venegas C., Samper-Ternent R., Michaels-Obregón A., Wong R.. **The effect of educational attainment on cognition of older adults: Results from the Mexican Health and Aging Study 2001 and 2012**. *Aging & Mental Health* (2019.0) **23** 1586-1594. DOI: 10.1080/13607863.2018.1501663
11. Furuta M., Yamashita Y.. **Oral health and swallowing problems**. *Current Physical Medicine and Rehabilitation Reports* (2013.0) **1** 216-222. DOI: 10.1007/s40141-013-0026-x
12. Gironda M. W., Maida C., Marcus M., Wang Y., Liu H.. **Social support and dental visits**. *The Journal of the American Dental Association* (2013.0) **144** 188-194. DOI: 10.14219/jada.archive.2013.0098
13. Gonzalez-Barrera A., Lopez M. H.. *A demographic portrait of Mexican-origin Hispanics in the United States* (2013.0)
14. Gutiérrez N. C.. *Mexico: Availability and cost of health care–legal aspects* (2014.0)
15. Hernández-Palacios R. D., Ramírez-Amador V., Jarillo-Soto E. C., Irigoyen-Camacho M. E., Mendoza-Núñez V. M.. **Relationship between gender, income and education and self-perceived oral health among elderly Mexicans. An exploratory study**. *Ciência & Saúde Coletiva* (2015.0) **20** 997-1004. DOI: 10.1590/1413-81232015204.00702014
16. Jann B.. **The Blinder–Oaxaca decomposition for linear regression models**. *The Stata Journal* (2008.0) **8** 453-479. DOI: 10.1177/1536867X0800800401
17. Kassebaum N. J., Smith A. G. C., Bernabé E., Fleming T. D., Reynolds A. E., Vos T., Murray C. J. L., Marcenes W.. **Global, regional, and national prevalence, incidence, and disability-adjusted life years for oral conditions for 195 countries, 1990–2015: A systematic analysis for the global burden of diseases, injuries, and risk factors**. *Journal of Dental Research* (2017.0) **96** 380-387. DOI: 10.1177/0022034517693566
18. Long J. S., Pavalko E. K.. **The life course of activity limitations: Exploring indicators of functional limitations over time**. *Journal of Aging and Health* (2004.0) **16** 490-516. DOI: 10.1177/0898264304265776
19. Luo H., Bell R. A., Wright W., Wu Q., Wu B.. **Trends in annual dental visits among US dentate adults with and without self-reported diabetes and prediabetes, 2004-2014**. *The Journal of the American Dental Association* (2018.0) **149** 460-469. DOI: 10.1016/j.adaj.2018.01.008
20. Malmö University & World Health Organization
. (n.d.). Mexico—Oral health country/area profile project. Retrieved January 13, 2022, from
https://capp.mau.se/country-areas/mexico/. *Mexico—Oral health country/area profile project*
21. Marcenes W., Kassebaum N. J., Bernabé E., Flaxman A., Naghavi M., Lopez A., Murray C. J. L.. **Global burden of oral conditions in 1990-2010: A systematic analysis**. *Journal of Dental Research* (2013.0) **92** 592-597. DOI: 10.1177/0022034513490168
22. Marchini L., Reynolds J. C., Caplan D. J., Sasser S., Russell C.. **Predictors of having a dentist among older adults in Iowa**. *Community Dentistry and Oral Epidemiology* (2020.0) **48** 240-247. DOI: 10.1111/cdoe.12521
23. Naka O., Anastassiadou V., Pissiotis A.. **Association between functional tooth units and chewing ability in older adults: A systematic review**. *Gerodontology* (2014.0) **31** 166-177. DOI: 10.1111/ger.12016
24. Okoro C. A., Strine T. W., Eke P. I., Dhingra S. S., Balluz L. S.. **The association between depression and anxiety and use of oral health services and tooth loss**. *Community Dentistry and Oral Epidemiology* (2012.0) **40** 134-144. DOI: 10.1111/j.1600-0528.2011.00637.x
25. Ortíz-Barrios L. B., Granados-García V., Cruz-Hervert P., Moreno-Tamayo K., Heredia-Ponce E., Sánchez-García S.. **The impact of poor oral health on the oral health-related quality of life (OHRQoL) in older adults: The oral health status through a latent class analysis**. *BMC Oral Health* (2019.0) **19** 141. DOI: 10.1186/s12903-019-0840-3
26. Parker SW, Saenz J, Wong R. **Health insurance and the aging: Evidence from the Seguro popular program in Mexico**. *Demography* (2018.0) **55** 361-386. DOI: 10.1007/s13524-017-0645-4
27. Peres M. A., Macpherson L. M. D., Weyant R. J., Daly B., Venturelli R., Mathur M. R., Listl S., Celeste R. K., Guarnizo-Herreño C. C., Kearns C., Benzian H., Allison P., Watt R. G.. **Oral diseases: A global public health challenge**. *Lancet* (2019.0) **394** 249-260. DOI: 10.1016/S0140-6736(19)31146-8
28. Ramírez M., Ahluwalia K. P., Teresi J. A.. **Correlates of dental visits among community-residing Latino elders: A public health alert**. *Gerodontology* (2011.0) **28** 12-18. DOI: 10.1111/j.1741-2358.2009.00335.x
29. Salinas J. J., Al Snih S., Markides K., Ray L. A., Angel R. J.. **The rural – urban divide: Health services utilization among older Mexicans in Mexico**. *The Journal of Rural Health : Official Journal of the American Rural Health Association and the National Rural Health Care Association* (2010.0) **26** 333-341. DOI: 10.1111/j.1748-0361.2010.00297.x
30. Sánchez-García S., de la Fuente-Hernández J., Juárez-Cedillo T., Mendoza J. M. O., Reyes-Morales H., Solórzano-Santos F., García-Peña C.. **Oral health service utilization by elderly beneficiaries of the Mexican institute of social security in México city**. *BMC Health Services Research* (2007.0) **7** 211. DOI: 10.1186/1472-6963-7-211
31. Sánchez-García S., Heredia-Ponce E., Cruz-Hervert P., Juárez-Cedillo T., Cárdenas-Bahena Á., García-Peña C.. **Oral health status in older adults with social security in Mexico City: Latent class analysis**. *Journal of Clinical and Experimental Dentistry* (2014.0) **6** e29-e35. DOI: 10.4317/jced.51224
32. Spinler K., Aarabi G., Valdez R., Kofahl C., Heydecke G., König H.-H., Hajek A.. **Prevalence and determinants of dental visits among older adults: Findings of a nationally representative longitudinal study**. *BMC Health Services Research* (2019.0) **19** 590. DOI: 10.1186/s12913-019-4427-0
33. The World Bank
(2018). Rural population (% of total population)-Mexico. The World Bank-Data. https://data.worldbank.org/indicator/SP.RUR.TOTL.ZS?locations=MX. *Rural population (% of total population)-Mexico* (2018.0)
34. Torres J. M., Wong R.. **Childhood poverty and depressive symptoms for older adults in Mexico: A life-course analysis**. *Journal of Cross-Cultural Gerontology* (2013.0) **28** 317-337. DOI: 10.1007/s10823-013-9198-1
35. Wong R., Espinoza M., Palloni A.. **Adultos mayores Mexicanos en contexto socioeconómico amplio: Salud y envejecimiento**. *Salud pública de México* (2007.0) **49** S436-S447. DOI: 10.1590/S0036-36342007001000002
36. Wong R., Michaels-Obregon A., Palloni A.. **Cohort profile: The Mexican Health and Aging Study (MHAS)**. *International Journal of Epidemiology* (2017.0) **46** e2-e2. DOI: 10.1093/ije/dyu263
37. Wong R., Ofstedal M. B., Yount K., Agree E. M.. **Unhealthy lifestyles among older adults: Exploring transitions in Mexico and the US**. *European Journal of Ageing* (2008.0) **5** 311-326. DOI: 10.1007/s10433-008-0098-0
|
---
title: Functional hierarchy among different Rab27 effectors involved in secretory
granule exocytosis
authors:
- Kunli Zhao
- Kohichi Matsunaga
- Kouichi Mizuno
- Hao Wang
- Katsuhide Okunishi
- Tetsuro Izumi
journal: eLife
year: 2023
pmcid: PMC9988257
doi: 10.7554/eLife.82821
license: CC BY 4.0
---
# Functional hierarchy among different Rab27 effectors involved in secretory granule exocytosis
## Abstract
The Rab27 effectors are known to play versatile roles in regulated exocytosis. In pancreatic beta cells, exophilin-8 anchors granules in the peripheral actin cortex, whereas granuphilin and melanophilin mediate granule fusion with and without stable docking to the plasma membrane, respectively. However, it is unknown whether these coexisting effectors function in parallel or in sequence to support the whole insulin secretory process. Here, we investigate their functional relationships by comparing the exocytic phenotypes in mouse beta cells simultaneously lacking two effectors with those lacking just one of them. Analyses of prefusion profiles by total internal reflection fluorescence microscopy suggest that melanophilin exclusively functions downstream of exophilin-8 to mobilize granules for fusion from the actin network to the plasma membrane after stimulation. The two effectors are physically linked via the exocyst complex. Downregulation of the exocyst component affects granule exocytosis only in the presence of exophilin-8. The exocyst and exophilin-8 also promote fusion of granules residing beneath the plasma membrane prior to stimulation, although they differentially act on freely diffusible granules and those stably docked to the plasma membrane by granuphilin, respectively. This is the first study to diagram the multiple intracellular pathways of granule exocytosis and the functional hierarchy among different Rab27 effectors within the same cell.
## Introduction
Synaptic vesicles must be docked and primed on the plasma membrane prior to stimulation for successful neuronal transmission to occur within 1 ms after electrical stimulation (Südhof, 2013). In contrast, secretory granules in exocrine and endocrine cells initiate exocytosis at a timepoint that is 1000-fold later than this, even when the Ca2+ concentration is abruptly upregulated by caged-Ca2+ compounds (Kasai, 1999). Their physiological release of bioactive substances likewise occurs much later. For example, digestive enzymes and insulin are secreted from pancreas cells continuously over a range of minutes to hours during food intake and subsequent hyperglycemia. To support such slow and consecutive exocytosis, the exocytic pathway can be neither single nor linear, and the requisite steps for recruiting granules from the cell interior to the cell limits are therefore critical. In fact, total internal reflection fluorescence (TIRF) microscopy monitoring of prefusion behavior in living cells has revealed that insulin granules residing beneath the plasma membrane prior to stimulation, and those recruited from the cell interior after stimulation, fuse in parallel, with some variability in their ratios, during physiological glucose-stimulated insulin secretion (GSIS) (Kasai et al., 2008; Ohara-Imaizumi et al., 2004; Shibasaki et al., 2007). In many secretory cells, a greater number of granules are clustered in the actin cortex at the cell periphery and/or along the plasma membrane compared with other cytoplasmic areas. It has traditionally been considered that granules docked to the plasma membrane form a readily releasable pool, whereas those accumulated within the actin cortex form a reserve pool. However, this may be an oversimplification, considering that there is a gating system to prevent spontaneous or unlimited vesicle fusion in regulated exocytosis. In fact, multiple Rab27 effectors that are involved in intracellular granule trafficking show complex and differential effects on exocytosis (Izumi, 2021). For example, granuphilin (also known as exophilin-2 and Slp4) mediates stable granule docking to the plasma membrane but simultaneously prevents their spontaneous fusion by interacting with a fusion-incompetent, closed form of syntaxins (Gomi et al., 2005; Torii et al., 2002). Another effector, exophilin-8 (also known as MyRIP and Slac2-c), anchors secretory granules within the actin cortex (Bierings et al., 2012; Desnos et al., 2003; Fan et al., 2017; Huet et al., 2012; Mizuno et al., 2011; Nightingale et al., 2009), which is considered to have dual roles in accumulating granules at the cell periphery and in preventing their access to the plasma membrane. Although another effector, melanophilin (also known as exophilin-3 and Slac2-a), similarly captures melanosomes within the peripheral actin network in skin melanocytes (Hammer and Sellers, 2012), it mediates stimulus-induced granule mobilization and immediate fusion to the plasma membrane in pancreatic beta cells (Wang et al., 2020). It is unknown, however, just how the different Rab27 effectors coexisting in the same cell function in a coordinated manner to support the entire exocytic processes. Neither is it known whether multiple secretory pathways and/or rate-limiting steps exist in the final fusion stages of regulated granule exocytosis. To answer these questions, we must first seek to determine whether each effector functions in sequence or in parallel. In this study, we compared the exocytic profiles in beta cells lacking the two effectors with those in cells deficient in each single effector and in wild-type (WT) cells. This yielded valuable insights into the functional hierarchy and relationship among the exocytic steps in which individual effectors are involved. We also found that the exocyst, which universally functions in constitutive exocytosis (Wu and Guo, 2015), plays roles in the physical and functional connections between different Rab27 effectors.
## Melanophilin exclusively functions downstream of exophilin-8 to mediate the exocytosis of granules recruited from the actin cortex to the plasma membrane after stimulation
In monolayer mouse pancreatic beta cells, insulin granules were unevenly distributed with accumulation in the actin cortex (Figure 1—figure supplement 1A). Because the Rab27 effectors, melanophilin and exophilin-8, are known to show affinities to the actin motors, myosin-Va and/or -VIIa (Hammer and Sellers, 2012, Izumi, 2021), these effectors are presumed to function on granules within this peripheral actin network. In fact, they showed a similar uneven distribution and were colocalized especially at the cell periphery (Figure 1—figure supplement 1B). To examine the functional relationship between these effectors, we generated melanophilin/exophilin-8 double-knockout (ME8DKO) mice by crossing exophilin-8-knockout (Exo8KO) mice (Fan et al., 2017) with melanophilin-knockout (MlphKO) mice (see ‘Materials and methods’). The doubly deficient beta cells displayed even distribution of insulin granules in the cytoplasm without accumulation at the cell periphery, in contrast to WT and MlphKO cells, but in a manner similar to that in Exo8KO cells (Figure 1A). Expression of exophilin-8, but not that of melanophilin, in ME8DKO cells at the endogenous level in WT cells redistributed them to the cell periphery (Figure 1B). This is consistent with the previous findings obtained from cells deficient in each single effector, which revealed that exophilin-8 is essential for granule accumulation within the actin cortex, whereas melanophilin is not (Fan et al., 2017; Wang et al., 2020).
**Figure 1.:** *Exophilin-8, but not melanophilin, accumulates insulin granules in the action cortex.(A) WT, MlphKO, Exo8KO, and ME8DKO beta cells were immunostained with anti-insulin antibody. A peripheral accumulation of insulin immunosignals was quantified under confocal microscopy: clustering of insulin immunosignals at least in one corner (upper) or along the plasma membrane (lower) were counted as positive. More than 100 cells were inspected in each of three independent experiments. Note that Exo8KO and ME8DKO cells do not show the peripheral accumulation of insulin granules in contrast to WT and MlphKO cells. (B) ME8DKO cells were infected by adenovirus expressing either HA-exophilin-8 or HA-melanophilin at the endogenous protein levels found in WT cells. LacZ was expressed in WT and ME8DKO cells as controls. The cell extracts were immunoblotted with the indicated antibodies (left). The ME8DKO cells expressing LacZ (upper), HA-exophilin-8 (middle), and HA-melanophilin (lower) were immunostained with anti-insulin and anti-HA antibodies (center), and a peripheral accumulation of insulin was quantified as in (A) (right). Insets represent higher magnification photomicrographs of a cell within the region outlined by frames. Note that expression of HA-exophilin-8, but not HA-melanophilin, rescues the peripheral granule accumulation in ME8DKO cells. Bars, 10 μm. ###p<0.001 by one-way ANOVA.
Figure 1—source data 1.Uncropped blot images of Figure 1B.*
ME8DKO mice show glucose intolerance without changes in body weight or insulin sensitivity, as found in each singly deficient mice (Figure 2—figure supplement 1A). The cells deficient in melanophilin and/or exophilin-8 all showed decreases in GSIS compared with WT cells in both islet batch and perifusion assays, although the decreases in ME8DKO cells tended to be larger than those in MlphKO cells (Figure 2A and B, Figure 2—figure supplement 1B and C). We next monitored insulin granule exocytosis directly by TIRF microscopy in living cells expressing insulin fused with enhanced green fluorescent protein (EGFP). The numbers of visible granules under TIRF microscopy were not different among the four genotypes of cells (Figure 2C). Lack of the change in Exo8KO cells suggests that granules residing beneath the plasma membrane are not necessarily derived or supplied from granules anchored within the actin cortex. The amount of insulin released in the medium correlated well with a reduction in the total number of fusion events detected as a flash followed by diffusion of insulin-EGFP fluorescence during glucose stimulation in each of the mutant cells (Figure 2D, Figure 2—figure supplement 1D). We previously categorized fused insulin granules into three types depending on their prefusion behaviors (Kasai et al., 2008): those having been visible prior to stimulation, ‘residents’; those becoming visible during stimulation, ‘visitors’; and those invisible until fusion, ‘passengers.’ MlphKO cells with the genetic background of C57BL/6N mice showed a specific decrease in the passenger-type exocytosis, particularly in a later phase, compared with WT cells (Figure 2D and E), which is consistent with previous findings in parental leaden cells with the genetic background of C57BR/cdJ mice (Wang et al., 2020). In contrast, both Exo8KO and ME8DKO cells with the same C57BL/6N genetic background displayed decreases in both the resident and passenger types of exocytosis, while changes in the markedly less frequent visitor type were difficult to compare among the cells. Because there are no significant differences in exocytic phenotypes between Exo8KO and ME8DKO cells, melanophilin is thought to function downstream of exophilin-8. Considering the previously identified role of each effector in beta cells (Fan et al., 2017; Wang et al., 2020), the passenger-type exocytosis mediated by melanophilin via interactions with myosin-Va and syntaxin-4 appears to be derived from granules anchored in the actin cortex by exophilin-8.
**Figure 2.:** *Insulin secretory defects of beta cells doubly deficient in melanophilin and exophilin-8 are indistinguishable from those singly deficient in exophilin-8.(A) Islets isolated from WT, MlphKO, Exo8KO, and ME8DKO mice at 12–18 weeks of age were preincubated in 2.8 mmol/L low glucose (LG)-containing KRB buffer at 37℃ for 1 hr. They were then incubated in new LG buffer for 1 hr followed by 25 mmol/L high glucose (HG) buffer for 1 hr. The ratios of insulin secreted in the media to that left in the cell lysates (Figure 2—figure supplement 1B) are shown (left; n = 9 from three mice each). (B) Islets from age-matched mice (19- to 25-week-old) were perfused with 16.7 mmol/L glucose buffer for 30 min, and the ratios of insulin secreted in the media to that left in the cell lysates are plotted as in Figure 2—figure supplement 1C. The area under the curve (AUC) is shown (n = 9 from three mice each). (C–E) A monolayer of islet cells from the above four kinds of mice were infected with adenovirus encoding insulin-EGFP and were observed by total internal reflection fluorescence (TIRF) microscopy (C, left). The numbers of visible granules were manually counted for WT (n = 10 cells from three mice), MlphKO (n = 9 from three mice), Exo8KO (n = 5 cells from three mice), and ME8DKO (n = 8 cells from three mice) cells (C, right). Insulin granule fusion events in response to 25 mmol/L glucose for 20 min were counted as in Figure 2—figure supplement 1D for WT (n = 21 cells from five mice), MlphKO (n = 10 from three mice), Exo8KO (n = 12 cells from three mice), and ME8DKO (n = 12 cells from three mice) cells. The observed fusion events were categorized into three types: residents, visitors, and passengers (D). Summary of the three modes of fusion events (E) in the first phase (left, 0–5 min) and the second phase (right, 5–20 min). Note that the decrease in insulin exocytosis in ME8DKO cells is greater than that in MlphKO cells, specifically due to the decrease in the resident-type exocytosis. Bar, 10 μm. #p<0.05, ##p<0.01, ###p<0.001 by one-way ANOVA.*
## Exophilin-8 and melanophilin form a complex via the exocyst
To investigate the molecular basis by which exophilin-8 and melanophilin sequentially promote the passenger-type exocytosis, we first examined the expression level of each effector in cells lacking the other effector. Although the expression of exophilin-8 was not affected in MlphKO cells, that of melanophilin was decreased to 59.8 ± $9.0\%$ ($$n = 4$$) in Exo8KO cells compared with WT cells (Figure 3A). However, this decrease should not affect GSIS in Exo8KO cells because a similar level of expression (65.4 ± $4.7\%$) in heterozygous MlphKO cells did not reduce it (Figure 3—figure supplement 1). We further found that the exophilin-8 and melanophilin expressed in the pancreatic beta cell line MIN6 form a complex (Figure 3B). These findings suggest that the protein stability of melanophilin partially depends on its interaction with exophilin-8, which is consistent with the model that melanophilin functions downstream of exophilin-8. However, they do not interact when expressed in HEK293A cells (Figure 3—figure supplement 2A), suggesting that they do not interact directly. The two effectors have been shown to interact with different proteins, except for Rab27a, in beta cells (Fan et al., 2017; Wang et al., 2020). The exophilin-8 mutant that loses its binding activity to RIM-BP2, and the melanophilin mutants that lose their binding activity to Rab27a, myosin-Va, or actin, all preserved their binding activity to the other WT effector in MIN6 cells (Figure 3—figure supplement 2B). To identify unknown intermediate proteins, we individually expressed Myc-TEV-FLAG (MEF)-tagged exophilin-8 and melanophilin in MIN6 cells, and we analyzed the proteins in each of the anti-FLAG immunoprecipitates using a liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Figure 3—figure supplement 3). As a result, we identified the exocyst complex components, SEC8, SEC10, and EXO70, in both immunoprecipitates, which is consistent with the previous finding that exophilin-8 interacts with SEC6 and SEC8 in INS-1 $\frac{832}{13}$ cells (Goehring et al., 2007). We confirmed the endogenous interactions among melanophilin, exophilin-8, SEC6, and SEC10 in MIN6 cells (Figure 3C). The exocyst forms an evolutionarily conserved heterooctameric protein complex (Wu and Guo, 2015). To identify the components responsible for the interaction with two Rab27 effectors, we performed a visible immunoprecipitation (VIP) assay, which permits examination of large numbers of protein combinations and complicated one-to-many or many-to-many protein interactions simultaneously (Katoh et al., 2015). As a result, we found that exophilin-8 and melanophilin immediately, if not directly, interact with SEC8 and EXO70, respectively (Figure 3D). Although the exocyst components were found in both immunoprecipitates of melanophilin and exophilin-8, other interacting proteins such as RIM-BP2, RIM2, myosin-VIIa, and myosin-Va were identified in only one of the immunoprecipitates (Figure 3E), suggesting that the majority of each effector form a distinct complex in cells. In addition, although we previously reported that exophilin-8 primarily interacts with myosin-Vlla displaying a molecular mass of ~170 kDa in INS-1 $\frac{832}{13}$ rat cells (Fan et al., 2017), it interacted with mysosin-VIIa with an authentic mass of ~260 kDa in MIN6 mouse cells, which may reflect a species difference between these beta cell lines.
**Figure 3.:** *Exophilin-8 and melanophilin interact through the exocyst in beta cells.(A) Islet extracts (40 μg) from WT, MlphKO, Exo8KO, and ME8DKO mice were immunoblotted with antibodies toward the indicated proteins (left). Protein expression levels were quantified by densitometric analyses from 3 to 5 independent preparations (right). (B) MIN6 cells were infected with adenoviruses encoding control LacZ, mCherry-exophilin-8, and/or FLAG-melanophilin. After 2 days, the cell extracts underwent immunoprecipitation with mCherry nanobody. The immunoprecipitates (IP), as well as the 1% extracts (Input), were immunoblotted with anti-FLAG and anti-red fluorescent protein (RFP) antibodies. (C) MIN6 cell extracts were immunoprecipitated with rabbit anti-melanophilin antibody or control IgG, and the immunoprecipitates were immunoblotted with antibodies toward the indicated proteins. (D) HEK293A cells cultured in 10 cm dishes were transfected with mCherry-fused, exophilin-8 (left) or melanophilin (right) with the indicated EGFP-fused exocyst components. After 48 hr, the cell lysates were subjected to immunoprecipitation with mCherry-nanobody-bound glutathione-Sepharose beads. EGFP and mCherry fluorescence on the precipitated beads was observed by confocal microscopy. (E) MIN6 cells expressing FLAG-tagged, melanophilin or exophlin-8 were immunoprecipitated with anti-FLAG antibody, and the immunoprecipitates were immunoblotted with antibodies toward the indicated proteins. Note that the exocyst complex components exist in both immunoprecipitates, whereas RIM-BP2, RIM2, myosin-VIIa, and myosin-Va largely exist only in one of them. Bars, 100 μm. ###p<0.001 by one-way ANOVA.
Figure 3—source data 1.Uncropped blot images of Figure 3A, B, C and E.*
## The exocyst functions only in the presence of exophilin-8
In mammalian cells, the exocyst complex is assembled after separate formation of the subcomplex 1 (SEC3, SEC5, SEC6, SEC8) and the subcomplex 2 (SEC10, SEC15, EXO70, EXO84) (Ahmed et al., 2018). Because exophilin-8 and melanophilin immediately bind SEC8 and EXO70, respectively (Figure 3D), it is conceivable that the holo-exocyst links the two effectors. Previous VIP assays detected interactions between SEC8 or SEC3 in the subcomplex 1 and SEC10 in the subcomplex 2 (Katoh et al., 2015). Therefore, silencing of SEC10 is expected to efficiently disrupt the formation of the holo-exocyst. However, SEC10 knockdown by specific siRNAs did not affect the granule localization of exophilin-8 or melanophilin (Figure 4—figure supplement 1). Furthermore, there was no effect on the peripheral accumulation of insulin granules in contrast to the case in exophilin-8-deficient cells (Figure 1A). However, SEC10 knockdown significantly decreased the colocalization ratio of melanophilin, but not that of exophilin-8, with SEC6, another exocyst component that is colocalized with insulin granules in MIN6 cells (Tsuboi et al., 2005; Figure 4A). Considering that the two effectors specifically interact with the component of different subcomplexes (Figure 3D), it is reasonable that silencing of SEC10 that disrupts the subcomplex 2 dissociates melanophilin from exophilin-8 associated with the subcomplex 1. To obtain further evidence that the two effectors associate through the exocyst, we downregulated SEC8 that is thought to interact with exophilin-8 (Figure 4—figure supplement 1) and found that SEC8 knockdown markedly disrupted the complex formation (Figure 4B). These findings indicate that the exocyst physically connects the two effectors on the same granule.
**Figure 4.:** *Knockdown of the exocyst component disrupts the interaction between melanophilin and exophilin-8.(A) HA-exophilin-8 and HA-melanophilin were expressed in Exo8KO and MlphKO monolayer beta cells, respectively, at the endogenous levels in WT cells as described in Figure 1B. They were then transfected with control siRNA or siRNA against SEC10 #11 or #12, as shown in Figure 4—figure supplement 1A. After fixation, the cells were coimmunostained with anti-HA and anti-SEC6 antibodies and were observed by confocal microscopy (left). Fluorescent intensity profiles along the indicated line of SEC6 and either HA-exophilin-8 or HA-melanophilin are shown (center). Colocalization was quantified by Pearson’s correlation coefficient (right, n = 18–26 cells from two mice each). Note that SEC10 knockdown (KD) induces dissociation of melanophilin, but not exophilin-8, from SEC6 (black arrowheads). (B) MIN6 cells were transfected with control siRNA or siRNA against SEC8 #12, as shown in Figure 4—figure supplement 1A, and the cell extracts were subjected to immunoprecipitation with control IgG or anti-melanophilin antibody. The immunoprecipitates (IP), as well as the 1% extracts (Input), were immunoblotted with anti-exophilin-8 and anti-melanophilin antibodies. Bars, 10 μm. ###p<0.001 by one-way ANOVA.
Figure 4—source data 1.Uncropped blot images of Figure 4B.*
SEC10 knockdown markedly decreased GSIS in WT cells, but induced no further decrease of GSIS in Exo8KO cells, which was already decreased by half compared with WT cells (Figure 5A, Figure 5—figure supplement 1). These findings suggest that the exocyst functions only in the presence of exophilin-8. As shown in Figure 2C, TIRF microscopy of Exo8KO cells expressing insulin-EGFP revealed no decrease in the number of visible granules (Figure 5B). Similarly, SEC10 knockdown does not affect the number of visible granules in either WT or Exo8KO cells. Consistent with insulin secretion assays (Figure 5A), it induces a marked decrease in glucose-stimulated fusion events in WT cells, but no additional decrease in Exo8KO cells (Figure 5C). Categorization by prefusion behavior revealed decreases in both resident and passenger types of exocytosis in WT cells, which phenocopied Exo8KO cells without SEC10 knockdown (Figure 2D). Again, SEC10 knockdown had no effects on either type of exocytosis in Exo8KO cells. The holo-exocyst appears to function with exophilin-8 in the same pathway because SEC10 knockdown is expected to preserve the interaction between exophilin-8 and the subcomplex 1, as shown in Figure 4A. Taken together, it seems that exophilin-8 and the exocyst mediate the passenger-type exocytosis when they interact with melanophilin on the same granule.
**Figure 5.:** *The exocyst affects insulin granule exocytosis only in the presence of exophilin-8.(A–C) WT or Exo8KO mouse islet cells were twice transfected with control siRNA or siRNA against SEC10 #11 or #12, as shown in Figure 4—figure supplement 1A, and were plated in a 24-well plate (A) or glass base dish (B, C). (A) The transfected monolayer cells were incubated for 1 hr in KRB buffer containing 2.8 mmol/L glucose and were then stimulated for 1 hr in the same buffer or in buffer containing 25 mmol/L glucose. Insulin levels secreted in the media and left in the cell lysates were measured (Figure 5—figure supplement 1), and their ratios are shown (n = 3 from three mice each). (B) The control or SEC10 knockdown (KD) cells (control siRNA-treated WT cells, n = 10 from three mice; SEC10 siRNA#11 or #12-treated WT cells, n = 11 from three mice; control siRNA-treated Exo8KO cells, n = 12 from three mice; SEC10 siRNA#11 or #12-treated Exo8KO cells, n = 12 from three mice) were infected with adenovirus encoding insulin-EGFP and were observed by total internal reflection fluorescence (TIRF) microscopy (left). The numbers of visible granules were manually counted (right). (C) The transfected cells (control siRNA-treated WT cells, n = 22 from three mice; SEC10 siRNA#11-treated WT cells, n = 8 from three mice; SEC10 siRNA#12-treated WT cells, n = 10 from three mice; control siRNA-treated Exo8KO cells, n = 15 from three mice; SEC10 siRNA#11-treated Exo8KO cells, n = 9 from three mice; SEC10 siRNA#12-treated Exo8KO cells, n = 7 from three mice) were infected with adenovirus encoding insulin-EGFP, and fusion events in response to 25 mmol/L glucose for 20 min were counted and categorized under TIRF microscopy as described in Figure 2D. Bar, 10 μm. #p<0.05, ##p<0.01, ###p<0.001 by one-way ANOVA versus control siRNA-treated WT cells.*
## Exophilin-8 promotes the exocytosis of granules residing beneath the plasma membrane only in the presence of granuphilin
As shown in Figure 2D, exophilin-8 deficiency also affects the resident-type exocytosis. Another Rab27 effector, granuphilin, is thought to be deeply involved in this type of exocytosis because beta cells lacking granuphilin display very few granules directly attached to the plasma membrane under electron microscopy (Gomi et al., 2005). Despite this docking defect, these cells display a marked increase in granule exocytosis, possibly because granuphilin interacts with and stabilizes syntaxins in a fusion-incompetent, closed form (Torii et al., 2002). To explore the functional relationship between the two effectors in this type of exocytosis, we generated granuphilin/exophilin-8 double-knockout (GE8DKO) mice. The GE8DKO cells also exhibited an increase in GSIS compared with Exo8KO cells (Figure 6A, Figure 6—figure supplement 1A), which indicates that a significant number of granules fuse efficiently without exophilin-8 and granuphilin, thus without prior capture in the actin cortex and stable docking to the plasma membrane. TIRF microscopy of these cells expressing insulin-EGFP revealed that, although the number of visible granules was markedly decreased in granuphilin-knockout (GrphKO) cells compared with WT cells, as expected, it was not further decreased in GE8DKO cells compared with GrphKO cells (Figure 6B). The number of fusion events during glucose stimulation under TIRF microscopy was correlated with the amount of insulin released in the medium in each mutant cell (Figure 6C). All the types of exocytosis were increased in GrphKO cells, suggesting that the absence of granuphilin facilitates granule easier access to the fusion-competent machinery on the plasma membrane. Simultaneous absence of exophilin-8 induced strikingly differential effects on each type of exocytosis in GE8DKO cells: the increases in the passenger and visitor types were reduced to the level found in Exo8KO cells, whereas the increase in the resident type was completely unaffected. The former finding is consistent with the view that the passenger and visitor types of exocytosis are derived from granules captured in the actin cortex by exophilin-8. However, the latter finding indicates that, at least in the absence of granuphilin, exophilin-8 is dispensable for the exocytosis of granules already residing beneath the plasma membrane prior to stimulation.
**Figure 6.:** *Exophilin-8 deficiency strongly inhibits the resident-type exocytosis from granuphilin-positive, stably docked granules.(A) Islets isolated from WT, Exo8KO, GrphKO, and GE8DKO mice at 12–17 weeks of ages were preincubated in low glucose (LG) KRB buffer for 1 hr. They were then incubated in another LG buffer for 30 min followed by high glucose (HG) buffer for 30 min. The ratios of insulin secreted in the media of that left in the cell lysates (Figure 6—figure supplement 1A) are shown as in Figure 2A (n = 12 from four mice each). (B) A monolayer of islet cells from the above four kinds of mice were infected with adenovirus encoding insulin-EGFP and were observed by total internal reflection fluorescence (TIRF) microscopy (left). The numbers of visible granules were manually counted for WT (n = 15 cells from four mice), Exo8KO (n = 11 cells from three mice), GrphKO (n = 14 cells from four mice), and GE8DKO (n = 17 cells from four mice) cells (right). (C) Fusion events in response to 25 mmol/L glucose for 20 min were counted and categorized as described in Figure 2D (WT: n = 14 cells from three mice; Exo8KO: n = 15 cells from three mice; GrphKO: n = 18 cells from three mice; GE8DKO: n = 20 cells from three mice). Note that the increases in the passenger and visitor types in GrphKO cells were eliminated by the simultaneous absence of exophilin-8 in GE8DKO cells, whereas the increase in the resident type was completely unaffected at all. (D, E) A monolayer of GrphKO (n = 13) and GE8DKO (n = 12) islet cells from three mice each were infected by adenoviruses encoding insulin-EGFP and KuO-granuphilin to mimic WT and Exo8KO cells, respectively, as described in Figure 6—figure supplement 1B. Insulin granule fusion events in response to 25 mmol/L glucose for 20 min were counted and categorized under TIRF microscopy as in Figure 6C, except that the granuphilin-positive and -negative granules were distinguished (D). There were no granuphilin-positive granules showing either passenger or visitor-type exocytosis. The fusion probability of granuphilin-positive and -negative granules is shown as the percentage of those granules displaying the resident-type exocytosis (E). Bar, 10 μm. #p<0.05, ###p<0.001 by one-way ANOVA. *p<0.05, *** p<0.001 by Student’s t test.*
However, the above findings were unexpexted because exophilin-8 deficiency markedly affects the resident-type exocytosis in the presence of granuphilin (Figure 2D). We previously showed in WT cells that the resident-type exocytosis is heterogeneously derived from granuphilin-positive, immobile granules and granuphilin-negative, mobile granules (Mizuno et al., 2016). To directly assess the influence of exophilin-8 deficiency on granuphilin-mediated, docked granule exocytosis, we expressed Kusabira Orange-1 (KuO)-fused granuphilin in GrphKO and GE8DKO cells (Figure 6—figure supplement 1B) because exogenous granuphilin expressed in the presence of endogenous granuphilin abnormally accumulates insulin granules beneath the plasma membrane and severely impairs their exocytosis (Mizuno et al., 2016; Torii et al., 2004; Torii et al., 2002). Under TIRF microscopy, the numbers of granuphilin-positive, -negative, and total granules were not significantly different between the two cell types. We confirmed similar numbers of these visible granules in WT and Exo8KO cells by immunostaining endogenous granuphilin and insulin (Figure 6—figure supplement 1C), suggesting that the expression levels of KuO-granuphilin in GrphKO and GE8DKO cells are properly adjusted to mimic WT and Exo8KO cells, respectively. Consistently, the number of total fusion events as well as the level of passenger-type exocytosis in GE8DKO cells expressing KuO-granuphilin (mimic Exo8KO cells) were markedly decreased compared with that in GrphKO cells expressing KuO-granuphilin (mimic WT cells; Figure 6D). Furthermore, although the levels of resident-type exocytosis from both granuphilin-positive and -negative granules were significantly inhibited in mimic Exo8KO cells compared with mimic WT cells, the suppression was more complete toward granuphilin-positive granules. The fusion probability of granuphilin-positive granules, in which the number of fusion events is normalized by the number of visible granules, was also more strongly decreased (Figure 6E). These findings indicate that exophilin-8 primarily promotes the exocytosis of granules molecularly and stably tethered to the plasma membrane by granuphilin.
## The exocyst promotes the exocytosis of granules residing beneath the plasma membrane in the absence of granuphilin
Because the exocyst is also involved in the resident-type exocytosis (Figure 5C), we next investigated its functional relationship with granuphilin. We first examined the effects of SEC10 knockdown in GrphKO cells expressing insulin-EGFP under TIRF microscopy. Although it did not further decrease the number of visible granules in GrphKO cells (Figure 7A), as was the case in GE8DKO cells (Figure 6B), it markedly decreased all the types of exocytosis, including the resident-type exocytosis (Figure 7B), in contrast to the case in GE8DKO cells (Figure 6C). We then investigated the effect of SEC10 knockdown in GrphKO expressing both insulin-EGFP and KuO-granuphilin. Again, SEC10 knockdown did not change the numbers of granuphilin-positive, -negative, and total granules (Figure 7—figure supplement 1). However, these cells, including those having received control siRNA, did not respond to glucose stimulation well, possibly because two times of siRNA transfection and two kinds of adenovirus infection are too much intervention for vulnerable primary beta cells. Nevertheless, these cells did respond to KCl-induced depolarization. Although SEC10 knockdown significantly reduced the numbers of fusion from both granuphilin-positive and -negative granules (Figure 7C), it specifically decreased the fusion probability of granuphilin-negative granules (Figure 7D). Therefore, the exocyst deficiency primarily affects the resident-type exocytosis from granuphilin-negative granules, in contrast to exophilin-8 deficiency (Figure 6E).
**Figure 7.:** *Exocyst deficiency strongly inhibits the resident-type exocytosis from granuphilin-negative, untethered granules.(A) GrphKO beta cells were transfected with control siRNA (n = 10 from three mice) or siRNA against SEC10 #11 or #12 (n = 19 from three mice) were infected with adenovirus encoding insulin-EGFP as described in Figure 5B. They were observed by total internal reflection fluorescence (TIRF) microscopy (left), and numbers of visible granules were manually counted (right). (B) The control (n = 10 from three mice) or SEC10 knockdown (KD) cells (n = 19 from three mice) were infected with adenovirus encoding insulin-EGFP, and fusion events in response to 25 mmol/L glucose for 20 min were counted and categorized under TIRF microscopy as described in Figure 6C. Note that in contrast to exophilin-8 knockout (Figure 6C), SEC10 knockdown suppresses the resident-type exocytosis in GrphKO cells. (C, D) GrphKO islet cells from three mice were transfected with control (n = 14) or SEC10 siRNAs (n = 15) twice, and were infected with adenoviruses encoding insulin-EGFP and KuO-granuphilin on the next day. Because these cells that had undergone two times siRNA transfection and two kinds of adenovirus infection failed to respond to 25 mmol/L glucose stimulation well, they were depolarized by 60 mmol/L potassium. Fusion events during 6 min were counted and categorized (C) as in Figure 6D. The fusion probability of granuphilin-positive and -negative granules is shown (D) as in Figure 6E. Note that SEC10 knockdown selectively suppresses it from granuphilin-negative granules. Bar, 10 μm. *p<0.05, **p<0.01, ***p<0.001 by Student t test.*
## Discussion
In this study, using distinct prefusion behaviors observed by TIRF microscopy as markers of differential exocytic routes, we investigated the functional hierarchy among different Rab27 effectors in pancreatic beta cells lacking one or two of these effectors. We first present evidence that exophilin-8 functions upstream of melanophilin to drive the passenger-type exocytosis. This type of exocytosis appears to be derived from granules within the actin network because exophilin-8 is essential for granule accumulation in the actin cortex and both effectors function via interactions with myosin-VIIa and myosin-Va, respectively, in beta cells (Fan et al., 2017; Wang et al., 2020). We also show that these two effectors are linked by the exocyst protein complex. Although this evolutionarily conserved protein complex is known to function in constitutive exocytosis, it has not been established whether it plays a universal role in regulated exocytosis. For example, Drosophila with mutation in SEC5 displays a defect in neurite outgrowth but not in neurotransmitter secretion (Murthy et al., 2003). In yeast, the exocyst interacts with Sec4, a member of the Rab protein on secretory vesicles, via exocyst component Sec15 (Guo et al., 1999). Further, Sec4 and Sec15 directly bind Myo2, the yeast myosin-V, for secretory vesicle transport (Jin et al., 2011). The exocyst components also interact with the SNARE fusion machinery: Sec6 with Snc2 (VAMP/synaptobrevin family in mammals) (Shen et al., 2013) and Sec9 (SNAP-25 family in mammals) (Dubuke et al., 2015), and Sec3 with Sso2 (syntaxin family in mammals) (Yue et al., 2017). Although proteins corresponding to Rab27 effectors are absent in yeast, they appear to be involved in similar interactions in mammalian cells. In fact, melanophilin interacts with Rab27a, myosin-Va, and syntaxin-4 in beta cells (Wang et al., 2020). In mammalian cells, the exocyst complex is assembled after the formation of two separate subcomplexes (Ahmed et al., 2018). The holo-exocyst appears to connect exophilin-8 and melanophilin because both subcomplex components commonly exist in the immunoprecipitates of two effectors in beta cells. We further show that exophilin-8 and melanophilin associate via different subcomplex components, SEC8 and EXO70, respectively, and that disruption of one subcomplex results in dissociation between the two effectors. To our knowledge, this is the first example showing that different effectors toward the same Rab form a complex in cells, which corroborates the previous suggestion that multiple effectors on the same granules may smooth the transition between consecutive intermediate exocytic processes (Izumi, 2021).
We then show that exophilin-8 knockout and SEC10 knockdown exhibit very similar defects in granule exocytosis in WT cells, indicating that they function together. Furthermore, neither affects the number of granules visualized by TIRF microscopy. The sole difference is that, although exophilin-8 deficiency erases the granule accumulation in the actin cortex, exocyst deficiency does not, which suggests that the exocyst functions downstream of exophilin-8. Consistent with this view, we could not find any effects of SEC10 knockdown in Exo8KO cells. In the pheochromocytoma cell line, PC12, the nascent granules generated are transported in a microtubule-dependent manner to the cell periphery within a few seconds (Rudolf et al., 2001). In skin melanocytes, melanosomes are dispersed throughout the cytoplasm using the myosin-Va motor along dynamic actin tracks assembled by the SPIRE actin nucleator (Alzahofi et al., 2020). Thus, even without prior capture in the actin cortex by exophilin-8, insulin granules may also be eventually transported close to the plasma membrane using these routes.
We next show that, although exophilin-8 deficiency inhibits the resident-type exocytosis in WT cells, it does not affect it despite its increase in GrphKO cells. However, in GrphKO cells expressing KuO-granuphilin, exophilin-8 deficiency almost completely inhibits the exocytosis of granuphilin-positive granules. Granuphilin-mediated, stably docked granules are thought to require priming machinery for fusion, such as Munc13, that converts a granuphilin-associated, closed form of syntaxin to a fusion-competent, open form (Mizuno and Izumi, 2022). Exophilin-8 can contribute to this process because it is associated with RIM-BP2, RIM, and Munc13 (Fan et al., 2017), which are known to have such a priming role in synaptic vesicle exocytosis (Brockmann et al., 2019; Brockmann et al., 2020). These priming factors could nonspecifically convert even granuphilin-free syntaxins into the open form because exophilin-8 deficiency also significantly inhibits the resident-type exocytosis from granuphilin-negative granules. In fact, the exophilin-8 mutant that loses binding activity to RIM-BP2 has no effect on the decreased insulin secretion in exophilin-8-deficient cells (Fan et al., 2017).
In contrast to exophilin-8 deficiency, SEC10 deficiency suppresses the resident-type exocytosis even in the absence of granuphilin. Furthermore, in GrphKO cells expressing KuO-granuphilin, it specifically decreases the fusion probability from granuphilin-negative granules showing this type of exocytosis. We previously showed that granuphilin-negative granules are more mobile and fusogenic compared with granuphilin-positive granules under TIRF microscopy (Mizuno et al., 2016). However, these diffusible granules must somehow be directionally mobilized to the plasma membrane for fusion. The exocyst likely functions in this process, considering that it generally tethers secretory vesicles to the plasma membrane prior to membrane fusion in a broad range of cells (Wu and Guo, 2015). This view also explains why SEC10 deficiency does not affect the fusion probability of granuphilin-positive granules that have already stably docked to the plasma membrane. However, this tethering step should occur within 100–200 nm distance from the plasma membrane because SEC10 knockdown has no effects on the numbers of granules visualized by TIRF microscopy whether granuphilin is present or not.
In summary, our findings are the first to determine the functional hierarchy among different Rab27 effectors expressed in the same cell, although the exact mechanism remains unknown how these different steps are regulated in a coordinated manner. Some granules trapped in the actin cortex by the action of exophilin-8 stimulus-dependently fuse either immediately by the action of melanophilin (the passenger-type exocytosis) or after staying beneath the plasma membrane for a while (the visitor-type exocytosis) (Figure 8A). The exocyst may help tether granules to the plasma membrane in these types of exocytosis. Other granules somehow transported to the cell periphery without capture in the actin cortex are stably tethered to the plasma membrane by the action of granuphilin or remain untethered beneath the plasma membrane (Figure 8B). The former granules stimulus-dependently fuse after the granuphilin-associated, closed form of syntaxins are converted into the fusion-competent, open form, possibly by the action of exophilin-8-associated priming factors, such as RIM-BP2. The latter granules fuse stochastically after directed to the plasma membrane by the action of the exocyst. It should be noted, however, that granule exocytosis is not abrogated in beta cells lacking any of the single or double Rab27 effectors. This may suggest the existence of other secretory pathways. For example, other Rab27 effectors, such as exophilin-7 (also known as JFC1 or Slp1), can also promote some types of exocytosis (Wang et al., 2013). However, given that granules can arrive close to the plasma membrane and fuse efficiently without prior capture in the actin cortex by exophilin-8 and/or stable docking to the plasma membrane by granuphilin, Rab27 effectors are thought to be evolved to play regulatory roles that prevent spontaneous or inappropriate fusion in regulated secretory pathways, as suggested previously (Izumi, 2021). In any case, it is now evident that there are multiple redundant paths and rate-limiting processes toward the final fusion step in granule exocytosis, which evolved to ensure the robust performance of the mechanisms governing the secretion of vital molecules, such as insulin.
**Figure 8.:** *A schematic model for the functional relationship among different Rab27 effectors and the exocyst in insulin granule exocytosis.At least, four different types of insulin granule exocytosis are discriminated by total internal reflection fluorescence (TIRF) microscopic analyses: the passenger type and the visitor type derived from granules anchored in the actin cortex (A), and the resident type from granules docked to the plasma membrane or from granules that remain untethered beneath the plasma membrane (B). Exophilin-8, melanophilin, granuphilin, and the exocyst differentially regulate each type of exocytosis. See details in ‘Discussion.’*
## Materials and methods
**Key resources table**
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
| --- | --- | --- | --- | --- |
| Gene (Mus musculus) | MyRIP(Exophilin-8) | GenBank | Gene ID: 245049 | |
| Gene (M. musculus) | Mlph(Melanophilin) | GenBank | Gene ID: 171531 | |
| Gene (M. musculus) | Sytl4(Granuphilin) | GenBank | Gene ID: 27359 | |
| Cell line (Homo sapiens) | HEK293A | Invitrogen | Cat#: R70507;RRID:CVCL_6910 | The cell line has been authenticated and tested negative for mycoplasma |
| Cell line (M. musculus) | Insulinoma | Miyazaki et al., 1990 | RRID:CVCL_0431 | The cell line has been authenticated and tested negative for mycoplasma |
| Transfected construct (M. musculus) | On-Target plus non-targeting pool siRNA | Horizon Discovery Ltd | Cat#: D-001810-10-05 | Silencer Select |
| Transfected construct (M. musculus) | siRNA to Exoc5 (SEC10) | Horizon Discovery Ltd | Cat#: 105504 | On-Target plus Set of 4 siRNA (J-047583-11 and -12) |
| Transfected construct (M. musculus) | siRNA to Exoc4 (SEC8) | Horizon Discovery Ltd | Cat#: 20336 | On-Target plus Set of 4 siRNA (J-051541-12-0050) |
| Biological sample (M. musculus) | Primary pancreatic beta cells | This paper | | Freshly isolated islets from male mouse pancreas |
| Antibody | Anti-Rab27a/b (rabbit polyclonal) | IBL | Cat#: 18975; RRID:AB_494635 | WB (1:2000) |
| Antibody | Anti-Rab27a (mouse monoclonal) | BD Biosciences | Cat#: 558532; RRID:AB_647327 | IF (1:100) |
| Antibody | Anti-Exophilin-8 (goat polyclonal) | Abcam | Cat#: ab10149; RRID:AB_296882 | IF (1:100) |
| Antibody | Anti-Melanophilin (goat polyclonal) | Everest Biotech | Cat#: EB05444; RRID:AB_2146092 | WB (1:2000) |
| Antibody | Anti-Melanophilin (rabbit polyclonal) | Proteintech | Cat#: 10338-1-AP;RRID:AB_2146104 | WB (1:2000) |
| Antibody | Anti-Myosin-VIIa (rabbit polyclonal) | Abcam | Cat#: ab3481;RRID:AB_303841 | WB (1:2000) |
| Antibody | Anti-Myosin-Va (rabbit polyclonal) | Cell Signaling Technology | Cat#: 3402;RRID:AB_2148475 | WB (1:2000) |
| Antibody | Anti-RIM-BP2 (rabbit polyclonal) | Proteintech | Cat#: 15716-1-AP;RRID:AB_2878173 | WB (1:2000) |
| Antibody | Anti-RIM1/2 (rabbit polyclonal) | Synaptic Systems | Cat#: 140213;RRID:AB_2832237 | WB (1:2000) |
| Antibody | Anti-SEC6 (mouse monoclonal) | Assay Designs | Cat#: ADI-VAM-SV021;RRID:AB_10618264 | IF (1:100), WB (1:2000) |
| Antibody | Anti-SEC8 (rabbit polyclonal) | Proteintech | Cat#: 11913-1-AP;RRID:AB_2101565 | WB (1:2000) |
| Antibody | Anti-SEC10 (rabbit polyclonal) | Proteintech | Cat#: 17593-1-AP;RRID:AB_2101582 | WB (1:2000) |
| Antibody | Anti-EXO70 (rabbit polyclonal) | Proteintech | Cat#: 12014-1-AP;RRID:AB_2101698 | WB (1:2000) |
| Antibody | Anti-FLAG (rabbit polyclonal) | Sigma-Aldrich | Cat#: F7425;RRID:AB_439687 | WB (1:2000) |
| Antibody | Anti-HA (rabbit polyclonal) | MBL | Cat#: 561;RRID:AB_591839 | WB (1:2000) |
| Antibody | Anti-HA (rat monoclonal) | Roche | Cat#: 11867423001;RRID:AB_390918 | IF (1:100) |
| Antibody | Anti-β-actin (mouse monoclonal) | Sigma-Aldrich | Cat#: A5316;RRID:AB_476743 | WB (1:10,000) |
| Antibody | Anti-red fluorescent protein (RFP) (rabbit polyclonal) | MBL | Cat#: PM005;RRID:AB_591279 | WB (1:10,000) |
| Commercial assay or kit | AlphaLISA insulin kit | PerkinElmer | Cat#: AL350HV/C/F | Insulin detection |
| Commercial assay or kit | Insulin high range kit | Cisbio | Cat#:62IN1PEG | Insulin detection |
| Commercial assay or kit | Insulin ultra sensitive kit | Cisbio | Cat#:62IN2PEH | Insulin detection |
| Recombinant DNA reagent | pEGFP-C3 (plasmid) | Addgene | Cat#: 53755-53762;RRID:Addgene_53755–53762 | GFP version of SEC3, SEC5, SEC6, SEC8, SEC10, SEC15, EXO70, and EXO84 |
| Software, algorithm | ImageQuant TL software | Cytiva | RRID:SCR_014246 | Quantify immunoblotting |
| Software, algorithm | NIS Element Viewer | Nikon | RRID:SCR_014329 | Quantify colocalization and analyze fusion events |
| Software, algorithm | Protein pilot software | SCIEX | RRID:SCR_018681 | Mass spectrometry |
| Other | Rhodamine-conjugated phalloidin | Thermo Fisher Scientific | Cat#: R415 | IF (1:100) |
| Other | Anti-FLAG M2Affinity Gel | Sigma-Aldrich | Cat#: A2220 | IP |
| Other | Protein GSepharose 4Fast Flow | GE Healthcare Biosciences | Cat#: GE17-0618-01 | IP |
| Other | Lipofectamine 3000 | Invitrogen | Cat#: L3000001 | Transfection |
| Other | Lipofectamine RNAiMax | Invitrogen | Cat#: 13778075 | Transfection |
## Mice and pancreatic islet cell preparation
Animal experiments were performed according to the rules and regulations of the Animal Care and Experimental Committees of Gunma University (permit number: 22-010; Maebashi, Japan). Only male mice and their tissues and cells were phenotypically characterized in this study. Leaden (C57J/L) mice with nonfunctional mutation of the gene encoding melanophilin (Mlph) (Matesic et al., 2001) were purchased from The Jackson Laboratory (Strain #:000668, RRID:IMSR_JAX:000668), and were backcrossed with C57BL/6N mice 10 times to generate MlphKO mice. Exo8KO mice in the genetic background of C57BL/6N mice were described previously (Fan et al., 2017). ME8DKO mice were obtained by mating Exo8KO mice with the MlphKO mice described above. GrphKO mice in the genetic background of C3H/He mice were described previously (Gomi et al., 2005). The male Exo8KO mice were mated with the female GrphKO mice. Because the granuphilin and exophilin-8 genes are on the mouse X and 9 chromosomes, respectively, the resultant F1 generation is either male (Grph-/Y, Exo8+/-) or female (Grph+/-, Exo8+/-). By crossing these F1 mice, GE8DKO mice, as well as the WT, GrphKO, and Exo8KO mice, were generated in the F2 generation and used for experiments. Although the resultant F2 mice have a mixture of C57BL/6N and C3H/He genomes, we expected that significant phenotypic changes due to the loss of Rab27 effectors may be preserved despite any influences due to randomly distributed differences in the genome. In fact, we found similar differences in exocytic profiles between WT and Exo8KO cells both in the C57BL/6N background (Figure 2) and in the mixture of the C57BL/6N and C3H/He backgrounds (Figure 6). Furthermore, GrphKO cells in the mixture of the C57BL/6N and C3H/He backgrounds showed changes in granule localization and exocytosis (Figure 6), consistent with the reported phenotypes of GrphKO cells in the C3H/He background (Gomi et al., 2005). Pancreatic islet isolation and dissociation into monolayer cells and insulin secretion assays were performed as described previously (Gomi et al., 2005; Wang et al., 2020). Briefly, islets were isolated from cervically dislocated mice by pancreatic duct injection of collagenase solution, and size-matched five islets were cultured overnight in a 24-well plate. Monolayer islet cells were prepared by incubation with trypsin-EDTA solution and were cultured for further 2 days. Insulin released from isolated islets or monolayer cells was measured by an AlphaLISA insulin kit (PerkinElmer) or insulin high range and ultra-sensitive kits (Cisbio).
## DNA manipulation
Mouse cDNAs of granuphilin (Wang et al., 1999), exophilin-8 (Mizuno et al., 2011), and melanophilin (Wang et al., 2020) were cloned previously. Human cDNAs of SEC3, SEC5, SEC6, SEC8, SEC10, SEC15, EXO70, and EXO84 in the pEGFP-C3 vector were gifts from Dr. Channing J. Der (Addgene plasmid # 53755-53762; http://n2t.net/addgene:53755-53762; RRID:Addgene_53755-53762; Martin et al., 2014). Adenoviruses encoding insulin-EGFP and KuO-granuphilin were described previously (Kasai et al., 2008; Mizuno et al., 2016). Hemagglutinin (HA)-, FLAG-, MEF-, One-STrEP-FLAG (OSF), and mCherry-tagged exophilin-8 and melanophilin were made previously (Fan et al., 2017; Wang et al., 2020). To express exogenous protein, HEK293A cells were transfected with the plasmids using Lipofectamine 3000 reagent (Invitrogen), whereas MIN6 and primary pancreatic beta cells were infected with adenoviruses.
## Cell lines, antibodies, and immunoprocedures
MIN6 cells (RRID:CVCL_0431) were originally provided by Dr. Jun-ichi Miyazaki (Osaka University; Miyazaki et al., 1990). HEK293A cells (RRID:CVCL_6910) were purchased from Invitrogen (Cat# R70507). Guinea pig anti-insulin serum was a gift from H. Kobayashi (Gunma University). Rabbit polyclonal anti-exophilin-8 (αExo8N) and anti-granuphilin (αGrp-N) antibodies are described previously (Fan et al., 2017; Yi et al., 2002). mCherry nanobody was a gift from Drs. Y. Katoh and K. Nakayama (Kyoto University) (Katoh et al., 2016). The sources of commercially available antibodies and their concentrations used are listed in Key Resources Table. Cells were lysed by lysis buffer consisting of 50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, $10\%$ (w/v) glycerol, 100 mmol/L NaF, 10 mmol/L ethylene glycol tetraacetic acid, 1 mmol/L Na3VO4, $1\%$ Triton X-100, 5 μmol/L ZnCl2, 1 mmol/L phenylmethylsulfonyl fluoride, and cOmplete Protease Inhibitor Cocktail (Roche). Immunoblotting and immunoprecipitation were performed as described previously (Matsunaga et al., 2017; Wang et al., 2020). VIP assay using the nanobody was performed as described previously (Katoh et al., 2015). Briefly, HEK293A cells on 10 cm dish were transfected with pEGFP-Sec and either pmCherry-Melanophilin or pmCherry-Exophilin-8 by Lipofectamine 3000. After 48 hr, cells were lysed by 1 mL of lysis buffer, and the cell lysate was centrifuged at 10,000 × g for 10 min, and the supernatant was subjected to immunoprecipitation with a 5 μL gel volume of mCherry nanobody-bound glutathione Sepharose. The beads were washed three times with lysis buffer and transferred to 35 mm glass base dishes (Glass φ12, IWAKI). Green and red fluorescence of beads was observed by confocal laser scanning microscopy. Acquisition of images was performed under fixed conditions. For immunofluorescence procedures, monolayer primary beta cells seeded at 3–5 × 104 cells on a poly-L-lysine-coated glass base dish were cultured overnight in RPMI-1640 medium (11.1 mmol/L glucose) supplemented with $10\%$ fetal calf serum and were fixed by $4\%$ paraformaldehyde for 30 min at room temperature. The cells were washed by phosphate buffered saline (PBS) and permeabilized by PBS containing $0.1\%$ Triton X-100 and 50 mmol/L NH4Cl for 30 min. After blocking with $1\%$ bovine serum albumin in PBS for 30 min, the cells were immunostained and observed under confocal microscopy with a ×100 oil immersion objective lens (1.49 NA). Quantification of immunoblot signals was performed using ImageQuant TL software (Cytiva). Quantification of colocalization between two proteins was performed using NIS Element Viewer software (Nikon).
## TIRF microscopy
TIRF microscopy was performed as described previously (Wang et al., 2020). Briefly, monolayer islet cells on glass base dish were infected with adenovirus encoding insulin-EGFP. Two days thereafter, the cells were preincubated for 30 min in 2.8 mmol/L low glucose (LG)-containing Krebs-Ringer bicarbonate (KRB) buffer at 37℃. They were then incubated in 25 mmol/L high glucose (HG)-containing buffer for 20 min or 60 mmol/L high potassium buffer for 6 min. TIRF microscopy was performed on a ×100 oil immersion objective lens (1.49 NA). The penetration depth of the evanescent field was 100 nm. Images were acquired at 103 ms intervals. Fusion events with a flash followed by diffusion of EGFP signals were manually selected and assigned to one of three types: residents, which were visible for more than 10 s prior to fusion; visitors, which became visible within 10 s of fusion; and passengers, which were not visible prior to fusion. In case of coinfection with adenovirus encoding KuO-granuphilin, sequential multi-color TIRF microscopy was performed as described previously (Mizuno et al., 2016). Briefly, EGFP was excited using a 488 nm solid-state laser, whereas KuO was excited using a 561 nm laser. Excitation illumination was synchronously delivered from an acousto-optic tunable filter-controlled laser launch. A dual-band filter set (LF$\frac{488}{561}$A; Semrock) was applied on a light path.
## Mass spectrometry
MIN6 cells (2 × 107 cells in ten 15 cm dishes) were infected with adenovirus encoding FLAG, FLAG-melanophilin, or MEF-exophilin-8. The anti-FLAG immunoprecipitates were subjected to gel electrophoresis and visualized by Oriole fluorescent gel staining (Bio-Rad). Specific bands were excised and digested in gels with trypsin, and the resulting peptide mixtures were analyzed by a LC-MS/MS system, as described previously (Matsunaga et al., 2017). All MS/MS spectra were separated against the *Mus musculus* (mouse) proteome data set (UP000000589) at the Uniplot using Protein pilot software (SCIEX).
## Silencing of the exocyst component in mouse pancreatic islet cells
The On-*Target plus* Set of four siRNA against mouse Exoc5 (J-047583-11 and -12; Cat# 105504) and mouse Exoc4 (J-051541-12-0050; Cat# 20336), and the On-*Target plus* non-targeting pool siRNA were purchased from Horizon Discovery Ltd. Mouse pancreatic islet cells suspended in 1 × 105 cells/280 μL were transfected with 100 nmol/L siRNA using Lipofectamine RNAiMAX reagent (Invitrogen). After plating on a 24-well plate or glass base dish for 48 hr, the cells were transfected with the same siRNA for the second time. After another 48 hr, the cells were subjected to immunoblotting analyses, immunofluorescent staining, insulin secretion assays, or TIRF microscopy after infection with adenovirus encoding insulin-EGFP.
## Statistical analysis
All quantitative data were assessed as the mean ± SEM. p-Values were calculated using Student’s t test or a one-way ANOVA with a Tukey multiple-comparison using GraphPad Prism software.
## Data and resource availability
All data generated or analyzed during this study are included in the manuscript and supporting files. All noncommercially available resources generated and/or analyzed during this study are available from the corresponding author upon reasonable request.
## Funding Information
This paper was supported by the following grants:
## Data availability
All data generated or analyzed during this study are included in the manuscript and supporting files.
## References
1. Ahmed SM, Nishida-Fukuda H, Li Y, McDonald WH, Gradinaru CC, Macara IG. **Exocyst dynamics during vesicle tethering and fusion**. *Nature Communications* (2018) **9**. DOI: 10.1038/s41467-018-07467-5
2. Alzahofi N, Welz T, Robinson CL, Page EL, Briggs DA, Stainthorp AK, Reekes J, Elbe DA, Straub F, Kallemeijn WW, Tate EW, Goff PS, Sviderskaya EV, Cantero M, Montoliu L, Nedelec F, Miles AK, Bailly M, Kerkhoff E, Hume AN. **Rab27A co-ordinates actin-dependent transport by controlling organelle-associated motors and track assembly proteins**. *Nature Communications* (2020) **11**. DOI: 10.1038/s41467-020-17212-6
3. Bierings R, Hellen N, Kiskin N, Knipe L, Fonseca AV, Patel B, Meli A, Rose M, Hannah MJ, Carter T. **The interplay between the Rab27A effectors slp4-a and myrip controls hormone-evoked Weibel-Palade body exocytosis**. *Blood* (2012) **120** 2757-2767. DOI: 10.1182/blood-2012-05-429936
4. Brockmann MM, Maglione M, Willmes CG, Stumpf A, Bouazza BA, Velasquez LM, Grauel MK, Beed P, Lehmann M, Gimber N, Schmoranzer J, Sigrist SJ, Rosenmund C, Schmitz D. **Rim-Bp2 primes synaptic vesicles via recruitment of Munc13-1 at hippocampal mossy fiber synapses**. *eLife* (2019) **8**. DOI: 10.7554/eLife.43243
5. Brockmann MM, Zarebidaki F, Camacho M, Grauel MK, Trimbuch T, Südhof TC, Rosenmund C. **A trio of active zone proteins comprised of RIM-bps, rims, and munc13s governs neurotransmitter release**. *Cell Reports* (2020) **32**. DOI: 10.1016/j.celrep.2020.107960
6. Desnos C, Schonn J-S, Huet S, Tran VS, El-Amraoui A, Raposo G, Fanget I, Chapuis C, Ménasché G, de Saint Basile G, Petit C, Cribier S, Henry J-P, Darchen F. **Rab27A and its effector myrip link secretory granules to F-actin and control their motion towards release sites**. *The Journal of Cell Biology* (2003) **163** 559-570. DOI: 10.1083/jcb.200302157
7. Dubuke ML, Maniatis S, Shaffer SA, Munson M. **The exocyst subunit sec6 interacts with assembled exocytic SNARE complexes**. *The Journal of Biological Chemistry* (2015) **290** 28245-28256. DOI: 10.1074/jbc.M115.673806
8. Fan F, Matsunaga K, Wang H, Ishizaki R, Kobayashi E, Kiyonari H, Mukumoto Y, Okunishi K, Izumi T. **Exophilin-8 assembles secretory granules for exocytosis in the actin cortex via interaction with RIM-BP2 and myosin-VIIa**. *eLife* (2017) **6**. DOI: 10.7554/eLife.26174
9. Goehring AS, Pedroja BS, Hinke SA, Langeberg LK, Scott JD. **MyRIP anchors protein kinase A to the exocyst complex**. *The Journal of Biological Chemistry* (2007) **282** 33155-33167. DOI: 10.1074/jbc.M705167200
10. Gomi H, Mizutani S, Kasai K, Itohara S, Izumi T. **Granuphilin molecularly docks insulin granules to the fusion machinery**. *The Journal of Cell Biology* (2005) **171** 99-109. DOI: 10.1083/jcb.200505179
11. Guo W, Roth D, Walch-Solimena C, Novick P. **The exocyst is an effector for Sec4p, targeting secretory vesicles to sites of exocytosis**. *The EMBO Journal* (1999) **18** 1071-1080. DOI: 10.1093/emboj/18.4.1071
12. Hammer JA, Sellers JR. **Walking to work: roles for class V myosins as cargo transporters**. *Nature Reviews. Molecular Cell Biology* (2012) **13** 13-26. DOI: 10.1038/nrm3248
13. Huet S, Fanget I, Jouannot O, Meireles P, Zeiske T, Larochette N, Darchen F, Desnos C. **Myrip couples the capture of secretory granules by the actin-rich cell cortex and their attachment to the plasma membrane**. *The Journal of Neuroscience* (2012) **32** 2564-2577. DOI: 10.1523/JNEUROSCI.2724-11.2012
14. Izumi T. **In vivo roles of Rab27 and its effectors in exocytosis**. *Cell Structure and Function* (2021) **46** 79-94. DOI: 10.1247/csf.21043
15. Jin Y, Sultana A, Gandhi P, Franklin E, Hamamoto S, Khan AR, Munson M, Schekman R, Weisman LS. **Myosin V transports secretory vesicles via a Rab GTPase cascade and interaction with the exocyst complex**. *Developmental Cell* (2011) **21** 1156-1170. DOI: 10.1016/j.devcel.2011.10.009
16. Kasai H. **Comparative biology of Ca2+-dependent exocytosis: implications of kinetic diversity for secretory function**. *Trends in Neurosciences* (1999) **22** 88-93. DOI: 10.1016/s0166-2236(98)01293-4
17. Kasai K, Fujita T, Gomi H, Izumi T. **Docking is not a prerequisite but a temporal constraint for fusion of secretory granules**. *Traffic* (2008) **9** 1191-1203. DOI: 10.1111/j.1600-0854.2008.00744.x
18. Katoh Y, Nozaki S, Hartanto D, Miyano R, Nakayama K. **Architectures of multisubunit complexes revealed by a visible immunoprecipitation assay using fluorescent fusion proteins**. *Journal of Cell Science* (2015) **128** 2351-2362. DOI: 10.1242/jcs.168740
19. Katoh Y, Terada M, Nishijima Y, Takei R, Nozaki S, Hamada H, Nakayama K. **Overall architecture of the intraflagellar transport (IFT) -B complex containing cluap1/IFT38 as an essential component of the IFT-B peripheral subcomplex**. *The Journal of Biological Chemistry* (2016) **291** 10962-10975. DOI: 10.1074/jbc.M116.713883
20. Martin TD, Chen X-W, Kaplan REW, Saltiel AR, Walker CL, Reiner DJ, Der CJ. **Ral and Rheb GTPase activating proteins integrate mTOR and GTPase signaling in aging, autophagy, and tumor cell invasion**. *Molecular Cell* (2014) **53** 209-220. DOI: 10.1016/j.molcel.2013.12.004
21. Matesic LE, Yip R, Reuss AE, Swing DA, O’Sullivan TN, Fletcher CF, Copeland NG, Jenkins NA. **Mutations in mlph, encoding a member of the rab effector family, cause the melanosome transport defects observed in leaden mice**. *PNAS* (2001) **98** 10238-10243. DOI: 10.1073/pnas.181336698
22. Matsunaga K, Taoka M, Isobe T, Izumi T. **Rab2a and Rab27A cooperatively regulate the transition from granule maturation to exocytosis through the dual effector Noc2**. *Journal of Cell Science* (2017) **130** 541-550. DOI: 10.1242/jcs.195479
23. Miyazaki J, Araki K, Yamato E, Ikegami H, Asano T, Shibasaki Y, Oka Y, Yamamura K. **Establishment of a pancreatic beta cell line that retains glucose-inducible insulin secretion: special reference to expression of glucose transporter isoforms**. *Endocrinology* (1990) **127** 126-132. DOI: 10.1210/endo-127-1-126
24. Mizuno K, Ramalho JS, Izumi T. **Exophilin8 transiently clusters insulin granules at the actin-rich cell cortex prior to exocytosis**. *Molecular Biology of the Cell* (2011) **22** 1716-1726. DOI: 10.1091/mbc.E10-05-0404
25. Mizuno K, Fujita T, Gomi H, Izumi T. **Granuphilin exclusively mediates functional granule docking to the plasma membrane**. *Scientific Reports* (2016) **6**. DOI: 10.1038/srep23909
26. Mizuno K, Izumi T. **Munc13b stimulus-dependently accumulates on granuphilin-mediated, docked granules prior to fusion**. *Cell Structure and Function* (2022) **47** 31-41. DOI: 10.1247/csf.22005
27. Murthy M, Garza D, Scheller RH, Schwarz TL. **Mutations in the exocyst component Sec5 disrupt neuronal membrane traffic, but neurotransmitter release persists**. *Neuron* (2003) **37** 433-447. DOI: 10.1016/s0896-6273(03)00031-x
28. Nightingale TD, Pattni K, Hume AN, Seabra MC, Cutler DF. **Rab27A and myrip regulate the amount and multimeric state of vWF released from endothelial cells**. *Blood* (2009) **113** 5010-5018. DOI: 10.1182/blood-2008-09-181206
29. Ohara-Imaizumi M, Nishiwaki C, Kikuta T, Nagai S, Nakamichi Y, Nagamatsu S. **Tirf imaging of docking and fusion of single insulin granule motion in primary rat pancreatic beta-cells: different behaviour of granule motion between normal and Goto-Kakizaki diabetic rat beta-cells**. *The Biochemical Journal* (2004) **381** 13-18. DOI: 10.1042/BJ20040434
30. Rudolf R, Salm T, Rustom A, Gerdes HH. **Dynamics of immature secretory granules: role of cytoskeletal elements during transport, cortical restriction, and F-actin-dependent tethering**. *Molecular Biology of the Cell* (2001) **12** 1353-1365. DOI: 10.1091/mbc.12.5.1353
31. Shen D, Yuan H, Hutagalung A, Verma A, Kümmel D, Wu X, Reinisch K, McNew JA, Novick P. **The synaptobrevin homologue snc2p recruits the exocyst to secretory vesicles by binding to Sec6p**. *The Journal of Cell Biology* (2013) **202** 509-526. DOI: 10.1083/jcb.201211148
32. Shibasaki T, Takahashi H, Miki T, Sunaga Y, Matsumura K, Yamanaka M, Zhang C, Tamamoto A, Satoh T, Miyazaki JI, Seino S. **Essential role of epac2/rap1 signaling in regulation of insulin granule dynamics by camp**. *PNAS* (2007) **104** 19333-19338. DOI: 10.1073/pnas.0707054104
33. Südhof TC. **Neurotransmitter release: the last millisecond in the life of a synaptic vesicle**. *Neuron* (2013) **80** 675-690. DOI: 10.1016/j.neuron.2013.10.022
34. Torii S, Zhao S, Yi Z, Takeuchi T, Izumi T. **Granuphilin modulates the exocytosis of secretory granules through interaction with syntaxin 1A**. *Molecular and Cellular Biology* (2002) **22** 5518-5526. DOI: 10.1128/MCB.22.15.5518-5526.2002
35. Torii S, Takeuchi T, Nagamatsu S, Izumi T. **Rab27 effector granuphilin promotes the plasma membrane targeting of insulin granules via interaction with syntaxin 1A**. *The Journal of Biological Chemistry* (2004) **279** 22532-22538. DOI: 10.1074/jbc.M400600200
36. Tsuboi T, Ravier MA, Xie H, Ewart MA, Gould GW, Baldwin SA, Rutter GA. **Mammalian exocyst complex is required for the docking step of insulin vesicle exocytosis**. *The Journal of Biological Chemistry* (2005) **280** 25565-25570. DOI: 10.1074/jbc.M501674200
37. Wang J, Takeuchi T, Yokota H, Izumi T. **Novel rabphilin-3-like protein associates with insulin-containing granules in pancreatic beta cells**. *The Journal of Biological Chemistry* (1999) **274** 28542-28548. DOI: 10.1074/jbc.274.40.28542
38. Wang H, Ishizaki R, Xu J, Kasai K, Kobayashi E, Gomi H, Izumi T. **The Rab27A effector exophilin7 promotes fusion of secretory granules that have not been docked to the plasma membrane**. *Molecular Biology of the Cell* (2013) **24** 319-330. DOI: 10.1091/mbc.E12-04-0265
39. Wang H, Mizuno K, Takahashi N, Kobayashi E, Shirakawa J, Terauchi Y, Kasai H, Okunishi K, Izumi T. **Melanophilin accelerates insulin granule fusion without predocking to the plasma membrane**. *Diabetes* (2020) **69** 2655-2666. DOI: 10.2337/db20-0069
40. Wu B, Guo W. **The exocyst at a glance**. *Journal of Cell Science* (2015) **128** 2957-2964. DOI: 10.1242/jcs.156398
41. Yi Z, Yokota H, Torii S, Aoki T, Hosaka M, Zhao S, Takata K, Takeuchi T, Izumi T. **The Rab27a/granuphilin complex regulates the exocytosis of insulin-containing dense-core granules**. *Molecular and Cellular Biology* (2002) **22** 1858-1867. DOI: 10.1128/MCB.22.6.1858-1867.2002
42. Yue P, Zhang Y, Mei K, Wang S, Lesigang J, Zhu Y, Dong G, Guo W. **Sec3 promotes the initial binary t-SNARE complex assembly and membrane fusion**. *Nature Communications* (2017) **8**. DOI: 10.1038/ncomms14236
|
---
title: 'Multimodal brain age estimates relate to Alzheimer disease biomarkers and
cognition in early stages: a cross-sectional observational study'
authors:
- Peter R Millar
- Brian A Gordon
- Patrick H Luckett
- Tammie LS Benzinger
- Carlos Cruchaga
- Anne M Fagan
- Jason J Hassenstab
- Richard J Perrin
- Suzanne E Schindler
- Ricardo F Allegri
- Gregory S Day
- Martin R Farlow
- Hiroshi Mori
- Georg Nübling
- Adam Sarah
- Adam Sarah
- Allegri Ricardo
- Araki Aki
- Barthelemy Nicolas
- Bateman Randall
- Bechara Jacob
- Benzinger Tammie
- Berman Sarah
- Bodge Courtney
- Brandon Susan
- Brooks William Bill
- Brosch Jared
- Buck Jill
- Buckles Virginia
- Carter Kathleen
- Cash Lisa
- Chen Charlie
- Chhatwal Jasmeer
- Mendez Patricio C
- Chua Jasmin
- Chui Helena
- Courtney Laura
- Cruchaga Carlos
- Day Gregory S
- DeLaCruz Chrismary
- Denner Darcy
- Diffenbacher Anna
- Dincer Aylin
- Donahue Tamara
- Douglas Jane
- Duong Duc
- Egido Noelia
- Esposito Bianca
- Fagan Anne
- Farlow Marty
- Feldman Becca
- Fitzpatrick Colleen
- Flores Shaney
- Fox Nick
- Franklin Erin
- Joseph-Mathurin Nelly
- Fujii Hisako
- Gardener Samantha
- Ghetti Bernardino
- Goate Alison
- Goldberg Sarah
- Goldman Jill
- Gonzalez Alyssa
- Gordon Brian
- Gräber-Sultan Susanne
- Graff-Radford Neill
- Graham Morgan
- Gray Julia
- Gremminger Emily
- Grilo Miguel
- Groves Alex
- Haass Christian
- Häsler Lisa
- Hassenstab Jason
- Hellm Cortaiga
- Herries Elizabeth
- Hoechst-Swisher Laura
- Hofmann Anna
- Holtzman David
- Hornbeck Russ
- Igor Yakushev
- Ihara Ryoko
- Ikeuchi Takeshi
- Ikonomovic Snezana
- Ishii Kenji
- Jack Clifford
- Jerome Gina
- Johnson Erik
- Jucker Mathias
- Karch Celeste
- Käser Stephan
- Kasuga Kensaku
- Keefe Sarah
- Klunk William
- Koeppe Robert
- Koudelis Deb
- Kuder-Buletta Elke
- Laske Christoph
- Levey Allan
- Levin Johannes
- Li Yan
- Lopez MD Oscar
- Marsh Jacob
- Martins Ralph
- Mason Neal S
- Masters Colin
- Mawuenyega Kwasi
- McCullough Austin
- McDade Eric
- Mejia Arlene
- Morenas-Rodriguez Estrella
- Morris John
- Mountz James
- Mummery Cath
- Nadkarni Neelesh
- Nagamatsu Akemi
- Neimeyer Katie
- Niimi Yoshiki
- Noble James
- Norton Joanne
- Nuscher Brigitte
- Obermüller Ulricke
- O'Connor Antoinette
- Patira Riddhi
- Perrin Richard
- Ping Lingyan
- Preische Oliver
- Renton Alan
- Ringman John
- Salloway Stephen
- Schofield Peter
- Senda Michio
- Seyfried Nicholas T
- Shady Kristine
- Shimada Hiroyuki
- Sigurdson Wendy
- Smith Jennifer
- Smith Lori
- Snitz Beth
- Sohrabi Hamid
- Stephens Sochenda
- Taddei Kevin
- Thompson Sarah
- Vöglein Jonathan
- Wang Peter
- Wang Qing
- Weamer Elise
- Xiong Chengjie
- Xu Jinbin
- Xu Xiong
- Randall J Bateman
- John C Morris
- Beau M Ances
journal: eLife
year: 2023
pmcid: PMC9988262
doi: 10.7554/eLife.81869
license: CC BY 4.0
---
# Multimodal brain age estimates relate to Alzheimer disease biomarkers and cognition in early stages: a cross-sectional observational study
## Abstract
### Background:
Estimates of ‘brain-predicted age’ quantify apparent brain age compared to normative trajectories of neuroimaging features. The brain age gap (BAG) between predicted and chronological age is elevated in symptomatic Alzheimer disease (AD) but has not been well explored in presymptomatic AD. Prior studies have typically modeled BAG with structural MRI, but more recently other modalities, including functional connectivity (FC) and multimodal MRI, have been explored.
### Methods:
We trained three models to predict age from FC, structural (S), or multimodal MRI (S+FC) in 390 amyloid-negative cognitively normal (CN/A−) participants (18–89 years old). In independent samples of 144 CN/A−, 154 CN/A+, and 154 cognitively impaired (CI; CDR > 0) participants, we tested relationships between BAG and AD biomarkers of amyloid and tau, as well as a global cognitive composite.
### Results:
All models predicted age in the control training set, with the multimodal model outperforming the unimodal models. All three BAG estimates were significantly elevated in CI compared to controls. FC-BAG was significantly reduced in CN/A+ participants compared to CN/A−. In CI participants only, elevated S-BAG and S+FC BAG were associated with more advanced AD pathology and lower cognitive performance.
### Conclusions:
Both FC-BAG and S-BAG are elevated in CI participants. However, FC and structural MRI also capture complementary signals. Specifically, FC-BAG may capture a unique biphasic response to presymptomatic AD pathology, while S-BAG may capture pathological progression and cognitive decline in the symptomatic stage. A multimodal age-prediction model improves sensitivity to healthy age differences.
### Funding:
This work was supported by the National Institutes of Health (P01-AG026276, P01- AG03991, P30-AG066444, 5-R01-AG052550, 5-R01-AG057680, 1-R01-AG067505, 1S10RR022984-01A1, and U19-AG032438), the BrightFocus Foundation (A2022014F), and the Alzheimer’s Association (SG-20-690363-DIAN).
## eLife digest
The brains of people with advanced Alzheimer’s disease often look older than expected based on the patients’ actual age. This ‘brain age gap’ (how old a brain appears compared to the person’s chronological age) can be calculated thanks to machine learning algorithms which analyse images of the organ to detect changes related to aging. Traditionally, these models have relied on images of the brain structure, such as the size and thickness of various brain areas; more recent models have started to use activity data, such as how different brain regions work together to form functional networks.
While the brain age gap is a useful measure for researchers who investigate aging and disease, it is not yet helpful for clinicians. For example, it is unclear whether the machine learning algorithm could detect changes in the brains of individuals in the initial stages of Alzheimer’s disease, before they start to manifest cognitive symptoms.
Millar et al. explored this question by testing whether models which incorporate structural and activity data could be more sensitive to these early changes. Three machine learning algorithms (relying on either structural data, activity data, or combination of both) were used to predict the brain ages of participants with no sign of disease; with biological markers of Alzheimer’s disease but preserved cognitive functions; and with marked cognitive symptoms of the condition.
Overall, the combined model was slightly better at predicting the brain age of healthy volunteers, and all three models indicated that patients with dementia had a brain which looked older than normal. For this group, the model based on structural data was also able to make predictions which reflected the severity of cognitive decline. Crucially, the algorithm which used activity data predicted that, in individuals with biological markers of Alzheimer’s disease but no cognitive impairment, the brain looked in fact younger than chronological age. Exactly why this is the case remains unclear, but this signal may be driven by neural processes which unfold in the early stages of the disease. While more research is needed, the work by Millar et al. helps to explore how various types of machine learning models could one day be used to assess and predict brain health.
## Introduction
Alzheimer disease (AD) is marked by structural and functional disruptions in the brain, some of which can be observed through multimodal magnetic resonance imaging (MRI) in preclinical and symptomatic stages of the disease (Frisoni et al., 2010; Brier et al., 2014a). More recently, the ‘brain-predicted age’ framework has emerged as a promising tool for neuroimaging analyses, leveraging recent developments and accessibility of machine-learning techniques, as well as large-scale, publicly available neuroimaging datasets (Cole and Franke, 2017b; Franke and Gaser, 2019). These models are trained to quantify how ‘old’ a brain appears, as compared to a normative sample of training data - typically consisting of cognitively normal participants across the adult lifespan (e.g., Cole et al., 2015). Thus, the framework allows for a residual-based interpretation of the brain age gap (BAG), defined as the difference between model-predicted age and chronological age, as an index of vulnerability and/or resistance to underlying disease pathology. Indeed, several studies have demonstrated that BAG is elevated (i.e. the brain ‘appears older’ than expected) in a host of neurological and psychiatric disorders, including symptomatic AD (Franke et al., 2010; Franke and Gaser, 2012; Gaser et al., 2013), as well as schizophrenia (e.g., Koutsouleris et al., 2014), HIV (e.g., Cole et al., 2017c), and type-2 diabetes (e.g., Franke et al., 2013), and moreover, predicts mortality (Cole et al., 2018). Conversely, lower BAG is associated with lower risk of disease progression (Gaser et al., 2013; Wang et al., 2019; Bocancea et al., 2021). Critically, at least one comparison suggests that BAG exceeds other established MRI (hippocampal volume) and CSF (pTau and Aβ42) biomarkers in sensitivity to AD progression (Gaser et al., 2013). Thus, by summarizing complex, non-linear, highly multivariate patterns of neuroimaging features into a simple, interpretable summary metric, BAG may reflect a comprehensive biomarker of brain health.
Several studies have established that symptomatic AD and mild cognitive impairment (MCI) are associated with elevated BAG (Cole and Franke, 2017b; Franke and Gaser, 2019). However, the sensitivity of these model estimates to AD in the presymptmatic stage (i.e. present amyloid pathology in the absence of cognitive decline [Sperling et al., 2011]) is less clear. The development of sensitive, reliable, non-invasive biomarkers of preclinical AD pathology is critical for the assessment of individual AD risk, as well as the evaluation of AD clinical prevention trials. Recent studies have demonstrated that greater BAG is associated with greater amyloid PET burden in a Down syndrome cohort (Cole et al., 2017a) and with greater tau PET burden in sporadic MCI and symptomatic AD (Lee et al., 2022). One approach to maximize sensitivity of BAG to presymptomatic AD pathology may be to train brain age models exclusively on amyloid-negative participants. As undetected AD pathology might influence MRI measures, and thus confound effects otherwise attributed to ‘healthy aging’ (Brier et al., 2014b), including the patterns learned by a traditional brain age model, an alternative model trained on amyloid-negative participants only might be more sensitive to detect presymptomatic AD pathology as deviations in BAG. Indeed, one recent study demonstrated that an amyloid-negative trained brain age model (Ly et al., 2020) is more sensitive to progressive stages of AD than a typical amyloid-insensitive model (Cole et al., 2015). However, this comparison included amyloid-negative and amyloid-positive test samples from two separate cohorts and thus may be driven by cohort, scanner, and/or site differences. To validate the applicability of the brain-predicted age approach to presymptomatic AD, it is important to test a model’s sensitivity to amyloid status, as well as continuous relationships with AD biomarkers, within a single cohort. Another recent comparison demonstrated that both traditional and amyloid-negative trained brain age models were similarly related to molecular AD biomarkers, but that further attempts to ‘disentangle’ AD from brain age by including more advanced AD continuum participants in the training sample significantly reduced relationships between brain age and AD markers (Hwang et al., 2022). Thus, in this study, we will apply the amyloid-negative training approach to a multimodal MRI dataset in order to maximize sensitivity to AD pathology in the presymptomatic stage.
Most of the brain-predicted age reports described above focused primarily on structural MRI. However, other studies have successfully modeled brain age using a variety of other modalities, including metabolic PET (Goyal et al., 2019; Lee et al., 2022), diffusion MRI (Cherubini et al., 2016; Petersen et al., 2022), and functional connectivity (FC) (Dosenbach et al., 2010; Liem et al., 2017; Eavani et al., 2018; Nielsen et al., 2019). Integration of multiple neuroimaging modalities may maximize sensitivity of BAG estimates to preclinical AD. Indeed, recent multimodal comparisons suggest that structural MRI and FC capture complementary age-related signals (Eavani et al., 2018; Dunås et al., 2021) and that age prediction may be improved by incorporating multiple modalities (Liem et al., 2017; Engemann et al., 2020). One recent study has shown that BAG estimates from an FC graph theory-based model are significantly elevated in autosomal dominant AD mutation carriers and are positively associated with amyloid PET (Gonneaud et al., 2021). Furthermore, we have recently demonstrated that FC correlation-based BAG estimates are surprisingly reduced in cognitively normal participants with evidence of amyloid pathology and elevated pTau, as well as in cognitively normal APOE ε4 carriers at genetic risk of AD (Millar et al., 2022). Thus, incorporating FC into BAG models may improve sensitivity to early AD.
This project aimed to develop multimodal models of brain-predicted age, incorporating both FC and structural MRI. Participants with presymptomatic AD pathology were excluded from the training set to maximize sensitivity. We hypothesized that BAG estimates would be sensitive to the presence of AD biomarkers and early cognitive impairment. We further considered whether estimates were continuously associated with AD biomarkers of amyloid and tau, as well as cognition. We hypothesized that FC and structural MRI would capture complementary signals related to age and AD. Thus, we systematically compared models trained on unimodal FC, structural MRI, and combined modalities to test the added utility of multimodal integration in accurately predicting age and whether each modality captures unique relationships with AD biomarkers and cognition.
## Participants
We formed a training sample of healthy controls spanning the adult lifespan by combining structural and FC-MRI data from three sources, as described previously (Millar et al., 2022): the Charles F. and Joanne Knight AD Research Center (ADRC) at Washington University in St. Louis (WUSTL), healthy controls from studies in the Ances lab at WUSTL (Thomas et al., 2013; Petersen et al., 2021), and mutation-negative controls from the Dominantly Inherited Alzheimer Network (DIAN) study of autosomal dominant AD at multiple international sites including WUSTL (McKay et al., 2022). To minimize the likelihood of undetected AD pathology in our training set, participants over the age of 50 were only included in the training set if they were cognitively normal, as assessed by the Clinical Dementia Rating (CDR 0; Morris, 1993), and had at least one biomarker indicating the absence of amyloid pathology (CN/A−, see below). We excluded 59 participants who did not have available CDR or biomarker measures (see Figure 1—figure supplement 1). As CDR and amyloid biomarkers were not available in the Ances lab controls, we included only participants at or below age 50 from this cohort in the training set. These healthy control participants were randomly divided into a training set (~$80\%$; $$n = 390$$) and a held-out test set (~$20\%$; $$n = 97$$), which did not significantly differ in age, sex, education, or race, see Table 1.
**Table 1.**
| Measure | Training sets (total N=390) | Training sets (total N=390).1 | Training sets (total N=390).2 | Test sets (total N=97) § | Test sets (total N=97) §.1 | Test sets (total N=97) §.2 | Analysis sets (total N=452) | Analysis sets (total N=452).1 | Analysis sets (total N=452).2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Measure | Ances Controls(CN/<50) | DIAN Controls(CN/A−) | Knight ADRC Controls(CN/A−) | Ances Controls(CN/<50) | DIAN Controls(CN/A−) | Knight ADRC Controls(CN/A−) | CN/A− | CN/A+ | CI |
| N | 136 | 120 | 134 | 38 | 26 | 33 | 144 | 154 | 154 |
| Age (mean, SD) | 29.92 (9.92) | 40.02 (10.26) | 64.97 (10.57) | 26.68 (7.11) | 41.46 (12.34) | 64.73 (10.57) | 66.93 (8.53) | 72.56 (7.15)‡ | 75.67 (6.86) ‡ |
| CDR (N 0 / N 0.5 / N 1.0 / N 2.0) | | 120 / 0 / 0 / 0 | 134 / 0 / 0 / 0 | | 26 / 0 / 0 / 0 | 33 / 0 / 0 / 0 | 144 / 0 / 0 / 0 | 154 / 0 / 0 / 0 | 0 / 119 / 35 / 2 |
| Amyloid status (N + / N -) | | 120 / 0 | 134 / 0 | | 26 / 0 | 33 / 0 | 144 / 0 | 0 / 154 | 0 / 57 |
| Biomarkers available (N PET / CSF / both) | | 30 / 6 / 79 | 11 / 22 / 91 | | 3 / 1 / 21 | 5 / 0 / 28 | 24 / 0 / 120 | 17 / 0 / 137 | 14 / 0 / 43 |
| APOE ε4 carrier status (N + / N -) | | 76 / 44 | 99 / 34 | | 19 / 7 | 28 / 5 | 115 / 29 | 71 / 83 ‡ | 55 / 98 ‡ |
| MMSE (mean, SD) | | | 29.26 (1.05) | | | 29.45 (0.94) | 29.13 (1.17) | 28.97 (1.33) | 25.37 (3.55) ‡ |
| Sex (N female / N male) | 70 / 64 | 85 / 35 | 84 / 50 | 19 / 18 | 16 / 10 | 22 / 11 | 89 / 55 | 91 / 63 | 68 / 86† |
| Years of education (mean, SD) | 13.68 (2.16) | 14.78 (3.04) | 16.16 (2.43) | 13.95 (1.99) | 14.92 (2.83) | 16.48 (2.43) | 15.71 (2.65) | 15.90 (2.64) | 15.05 (2.97)* |
| Race (N American Indian or Alaska Native) | 1 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 |
| Race (N Asian) | 1 | 1 | 2 | 0 | 0 | 0 | 0 | 1 | 0 |
| Race (N Black) | 67 | 0 | 20 | 17 | 0 | 7 | 17 | 16 | 20 |
| Race (N Native Hawaiian or Other Pacifc Islander) | 2 | 0 | 0 | 2 | 0 | 0 | 0 | 0 | 0 |
| Race (N White) | 57 | 118 | 112 | 17 | 26 | 26 | 127 | 137 | 134 |
| Site | WUSTL | Multiple sites | WUSTL | WUSTL | Multiple sites | WUSTL | WUSTL | WUSTL | WUSTL |
| Scanner | Siemens Trio | Siemens Trio / Verio | Siemens Trio / Biograph | Siemens Trio | Siemens Trio / Verio | Siemens Trio / Biograph | Siemens Trio / Biograph | Siemens Trio / Biograph | Siemens Trio / Biograph |
| Field strength | 3T | 3T | 3T | 3T | 3T | 3T | 3T | 3T | 3T |
Finally, independent samples for hypothesis testing included three groups from the Knight ADRC: a randomly selected sample of 144 CN/A− controls who did not overlap with the training or testing sets, 154 CN/A+ participants, and 154 cognitively impaired (CI) participants (CDR > 0 with a biomarker measure consistent with amyloid pathology [see below] and/or a primary diagnosis of AD or uncertain dementia [McKhann et al., 2011]). See Table 1 for demographic details of each sample. All participants provided written informed consent in accordance with the Declaration of Helsinki and their local institutional review board. All procedures were approved by the Human Research Protection Office at WUSTL (IRB ID # 201204041).
## PET and CSF biomarkers
Amyloid burden was imaged with PET using (11 C)-Pittsburgh Compound B (PIB; Klunk et al., 2004) or (18 F)-Florbetapir (AV45; Wong et al., 2010). Regional standard uptake ratios (SUVRs) were modeled from 30 to 60 min after injection for PIB and from 50 to 70 min for AV45, using cerebellar gray as the reference region (Su et al., 2013). Regions of interest were segmented automatically using FreeSurfer 5.3 (Fischl, 2012). Global amyloid burden was defined as the mean of partial-volume-corrected (PVC) SUVRs from bilateral precuneus, superior and rostral middle frontal, lateral and medial orbitofrontal, and superior and middle temporal regions (Su et al., 2013). Amyloid summary SUVRs were harmonized across tracers using a centiloid conversion (Su et al., 2018).
Tau deposition was imaged with PET using (18 F)-Flortaucipir (AV-1451; Chien et al., 2013). Regional SUVRs were modeled from 80 to 100 min after injection, using cerebellar gray as the reference region. A tau summary measure was defined in the mean PVC SUVRs from bilateral amygdala, entorhinal, inferior temporal, and lateral occipital regions (Mishra et al., 2017).
CSF was collected via lumbar puncture using methods described previously (Fagan et al., 2006). After overnight fasting, 20–30 mL samples of CSF were collected, centrifuged, then aliquoted (500 µL) in polypropylene tubes, and stored at –80°C. CSF amyloid β peptide 42 (Aβ42), Aβ40, and phosphorylated tau-181 (pTau) were measured with automated Lumipulse immunoassays (Fujirebio, Malvern, PA, USA) using a single lot of assays for each analyte. Aβ42 and pTau estimates were each normalized for individual differences in CSF production rates by forming a ratio with Aβ40 as the denominator (Hansson et al., 2019; Guo et al., 2020). As pTau/Aβ40 was highly skewed, we applied a log transformation to these estimates before statistical analysis.
Amyloid positivity was defined using previously published cutoffs for PIB (SUVR > 1.42; Vlassenko et al., 2016) or AV45 (SUVR > 1.19; Su et al., 2019). Additionally, the CSF Aβ42/Aβ40 ratio has been shown to be highly concordant with amyloid PET (positivity cutoff < 0.0673; Schindler et al., 2018; Volluz et al., 2021). Thus, participants were defined as amyloid-positive (for CN/A+ and CI groups) if they had either a PIB, AV45, or CSF Aβ42/Aβ40 ratio measure in the positive range. Participants with discordant positivity between PET and CSF estimates were defined as amyloid-positive.
## Cognitive battery
Knight ADRC participants completed a 2 hr battery of cognitive tests. We examined global cognition by forming a composite of tasks across cognitive domains, including processing speed (Trail Making A; Schindler et al., 2018), executive function (Trail Making B; Schindler et al., 2018), semantic fluency (Animal Naming; Armitage, 1946), and episodic memory (Free and Cued Selective Reminding Test free recall score; Goodglass and Kaplan, 1983; Grober et al., 1988). This composite has recently been used to study individual differences in cognition in relation the preclinical AD biomarkers and structural MRI (Aschenbrenner et al., 2018), as well as functional MRI measures (Millar et al., 2021).
## MRI acquisition
All MRI data were obtained using a Siemens 3T scanner, although there was a variety of specific models within and across studies. As described previously (Millar et al., 2022), participants in the Knight ADRC and Ances lab studies completed one of two comparable structural MRI protocols, varying by scanner (sagittal T1-weighted magnetization-prepared rapid gradient echo sequence [MPRAGE] with repetition time [TR] = 2400 or 2300 ms, echo time [TE] = 3.16 or 2.95 ms, flip angle = 8 or 9°, frames = 176, field of view = sagittal 256×256 or 240×256 mm, 1 mm isotropic or 1×1×1.2 mm voxels; oblique T2-weighted fast spin echo sequence [FSE] with TR = 3200 ms, TE = 455 ms, 256×256 acquisition matrix, 1 mm isotropic voxels) and an identical resting-state fMRI protocol (interleaved whole-brain echo planar imaging sequence [EPI] with TR = 2200 ms, TE = 27 ms, flip angle = 90°, field of view = 256 mm, 4 mm isotropic voxels for two 6 min runs [164 volumes each] of eyes open fixation). DIAN participants completed a similar MPRAGE protocol (TR = 2300ms, TE = 2.95ms, flip angle = 9°, field of view = 270 mm, 1.1×1.1×1.2 mm voxels; McKay et al., 2022). Resting-state EPI sequence parameters for the DIAN participants differed across sites and scanners with the most notable difference being shorter resting-state runs (one 5 min run of 120 volumes; see Supplementary file 1 for summary of structural and functional MRI parameters; McKay et al., 2022).
## FC preprocessing and features
All MRI data were processed using common pipelines. Initial fMRI preprocessing followed conventional methods, as described previously (Shulman et al., 2010; Millar et al., 2022), including frame alignment, debanding, rigid body transformation, bias field correction, and normalization of within-run intensity values to a whole-brain mode of 1000 (Power et al., 2012). Transformation to an age-appropriate in-house atlas template (based on independent samples of either younger adults or CN older adults) was performed using a composition of affine transforms connecting the functional volumes with the T2-weighted and MPRAGE images. Frame alignment was included in a single resampling that generated a volumetric time series of the concatenated runs in isotropic 3 mm atlas space.
As described previously (Fox et al., 2009; Millar et al., 2022), additional processing was performed to allow for nuisance variable regression. Data underwent framewise censoring based on motion estimates (framewise displacement [FD] > 0.3 mm and/or derivative of variance [DVARS] > 2.5 above participant’s mean). To further minimize the confounding influence of head motion on FC estimates (Power et al., 2012) in all samples, we only included scans with low head motion (mean FD < 0.30 mm and > $50\%$ frames retained after motion censoring). BOLD data underwent a temporal band-pass filter (0.005 Hz < f < 0.1 Hz) and nuisance variable regression, including motion parameters, timeseries from FreeSurfer 5.3-defined (Fischl, 2012) whole brain (global signal), CSF, ventricle, and white matter masks, as well as the derivatives of these signals. Finally, BOLD data were spatially blurred (6 mm full width at half maximum).
Final BOLD time series data were averaged across voxels within a set of 300 spherical regions of interest (ROIs) in cortical, subcortical, and cerebellar areas (Seitzman et al., 2020). For each scan, we calculated the 300×300 Fisher-transformed Pearson correlation matrix of the final averaged BOLD time series between all ROIs. We then used the vectorized upper triangle of each correlation matrix (excluding auto-correlations; 44,850 total correlations) as input features for predicting age. Since site and/or scanner differences between samples might confound neuroimaging estimates, we harmonized FC matrices using an empirical Bayes modeling approach (ComBat; Johnson et al., 2007; Fortin et al., 2017), which has previously been applied to FC data (Yu et al., 2018).
## Structural MRI processing and features
All T1-weighted images underwent cortical reconstruction and structural segmentation through a common pipeline with FreeSurfer 5.3 (Fischl et al., 2002; Fischl, 2012). Structural processing included segmentation of subcortical white matter and deep gray matter, intensity normalization, registration to a spherical atlas, and parcellation of the cerebral cortex based on the Desikan atlas (Desikan et al., 2006). Inclusion and exclusion errors of parcellation and segmentation were identified and edited by a centralized team of trained research technicians according to standardized criteria (Su et al., 2013). We then used the FreeSurfer-defined thickness estimates from 68 cortical regions (Desikan et al., 2006), along with volume estimates from 33 subcortical regions (Fischl et al., 2002) as input features for predicting age. We harmonized structural features across sites and scanners using the same ComBat approach (Johnson et al., 2007; Fortin et al., 2017), which has also been applied to structural MRI data (Fortin et al., 2018).
## Gaussian process regression
As described previously (Millar et al., 2022), machine-learning analyses were conducted using the Regression Learner application in Matlab (MathWorks, 2021). We trained two Gaussian process regression (GPR; Rasmussen et al., 2004) models, each with a rational quadratic kernel function to predict chronological age using fully-processed, harmonized MRI features (FC or structural) in the training set. The σ hyperparameter was tuned within each model by searching a range of values from 10–4 to 10*SDage using Bayesian optimization across 100 training evaluations. The optimal value of σ for each model was found (see Figure 1—figure supplement 2) and was applied for all subsequent applications of that model. All other hyperparameters were set to default values (basis function = constant and standardize = true).
Model performance in the training set was assessed using 10-fold cross validation via the Pearson correlation coefficient (r), the proportion of variance explained (R2), the mean absolute error (MAE), and root-mean-square error (RMSE) between true chronological age and the cross-validated age predictions merged across the 10 folds. We then evaluated generalizability of the models to predict age in unseen data by applying the trained models to the held-out test set of healthy controls. Finally, we applied the same fully-trained GPR models to separate analysis sets of 154 CI, 154 CN/A+, and 144 CN/A− controls to test our hypotheses regarding AD-related group effects and individual difference relationships. Unimodal models were each constructed with a single GPR model. The multimodal model was constructed by taking the ‘stacked’ predictions from each first-level unimodal model as features for training a second-level GPR model (Liem et al., 2017; Engemann et al., 2020; Dunås et al., 2021).
For each participant, we calculated model-specific BAG estimates as the difference between chronological age and age predictions from the unimodal FC model (FC-BAG), structural model (S-BAG), and multimodal model (S+FC BAG). To correct for regression dilution commonly observed in similar models (Le et al., 2018; Smith et al., 2019; Liang et al., 2019), we included chronological age as a covariate in all statistical tests of BAG (Cole et al., 2017a; Le et al., 2018). However, to avoid inflating estimates of prediction accuracy (Butler et al., 2021), only uncorrected age prediction values were used for evaluating model performance in the training and test sets.
## Statistical analysis
All statistical analyses were conducted in R 4.0.2 (R Development Core Team, 2020). Demographic differences in the AD samples were tested with independent-samples t tests for continuous variables and χ2 tests for categorical variables, using CN/A− controls as a reference group. Differences in brain age model performance were tested using Williams’s test of difference between dependent correlations sharing one variable, i.e., Pearson’s r between age and each model prediction of age. To correct for age-related bias in BAG (Le et al., 2018; as previously mentioned), we controlled for age as a covariate during all statistical tests. Group differences in each BAG estimate were tested using an omnibus ANOVA test with follow-up pairwise t tests on age-residualized BAG estimates, using a false discovery rate (FDR) correction for multiple comparisons. Assumptions of normality were tested by visual inspection of quantile-quantile plots. Assumptions of equality of variance were tested with Levene’s test. Linear regression models tested the effects of cognitive impairment (CDR > 0 vs. CDR 0) and amyloid positivity (A− vs. A+) on BAG estimates from each model, controlling for true age (as noted above), sex, and years of education. Given the potential confounding influence of head motion on FC-derived measures (Power et al., 2012; Van Dijk et al., 2012; Satterthwaite et al., 2012), we also included mean FD as an additional covariate of non-interest in the FC and S+FC models. We tested continuous relationships with AD biomarkers and cognitive estimates using linear regression models, including the same demographic and motion covariates. Since the range of amyloid biomarkers was drastically reduced in the CN/A− sample, we excluded these participants from models testing continuous amyloid relationships. Effect sizes were computed as partial η2 (ηp2).
## Sample description and demographics
Demographic characteristics of the training sets, test sets, and analysis sets are reported in Table 1. CN/A+ participants were older ($t = 6.15$, $p \leq 0.001$) and more likely to be APOE ε4 carriers (χ2 = 34.73, $p \leq 0.001$) than amyloid-negative controls. Furthermore, CI participants were older ($t = 9.71$, $p \leq 0.001$), more likely male (χ2 = 8.60, $$p \leq 0.003$$), more likely to be APOE ε4 carriers (χ2 = 56.67, $p \leq 0.001$), and had fewer years of education ($t = 2.03$, $p \leq 0.043$), and lower MMSE scores ($t = 12.46$, $p \leq 0.001$) than amyloid-negative controls.
## Comparison of model performance
All models accurately predicted chronological age in the training sets, as assessed using 10-fold cross validation, as well as in the held-out test sets. Overall, prediction accuracy was lowest in the FC model (MAEFC/Train = 8.67 years, R2FC/Train = 0.68, MAEFC/Test = 8.25 years, R2FC/Test = 0.73; see Figure 1A & B). The structural MRI model (MAES/Train = 5.97 years, R2S/Train = 0.81, MAES/Test = 6.26 years, R2S/Test = 0.82; see Figure 1C & D) significantly outperformed the FC model in age prediction accuracy, Williams’s tS vs. FC = 5.39, $p \leq 0.001.$ There was a significant, but modestly sized, positive correlation between FC-BAG and S-BAG in the adult lifespan CN/A− training and testing sets ($r = 0.095$, $$p \leq 0.036$$; see Figure 1—figure supplement 3A), as well as the AD analysis sets ($r = 0.134$, $$p \leq 0.004$$; see Figure 1—figure supplement 3B).
**Figure 1.:** *Performance of the brain age models in the training (left column) and test sets (right column) for each modality: functional connectivity (FC; A and B), structural MRI (S; C and D) and multimodal models (S+FC; E and F).Age predicted by each model (y axis) is plotted against true age (x axis). Colored lines and shaded areas represent regression lines and 95% confidence regions. Dashed black lines represent perfect prediction. Model performance is evaluated by Pearson’s r, proportion of variance explained (R2), mean absolute error (MAE), and root-mean-square error (RMSE).*
Finally, the multimodal model (MAES+FC/Train = 5.34 years, R2S+FC/Train = 0.86, MAES+FC/Test = 5.25 years, R2S+FC/Test = 0.87; see Figure 1E & F) significantly outperformed both the FC model (Williams’s tS+FC vs. FC = 11.20, $p \leq 0.001$) and the structural MRI model (Williams’s tS+FC vs. $S = 5.67$, $p \leq 0.001$). It is possible that the modest increase in the multimodal model was due to capitalizing on noise, simply by adding more features to the structural model. Hence, we also compared the observed R2S+FC to a bootstrapped distribution of R2 performance estimates from 1000 resamples using a model in which the original structural MRI model was stacked with a model trained on randomly reshuffled FC features. Thus, this distribution represents the expected improvements in model performance from simply adding new features to the structural MRI model with the stacked approach. The observed R2S+FC outperformed all R2 estimates from this bootstrapped distribution ($p \leq 0.001$; see Figure 1—figure supplement 4), suggesting that the modest increase in model performance observed in the stacked multimodal (S+FC) model over the unimodal structural model is due to meaningful age-related FC signal, rather than capitalizing on noise in a larger feature set.
## BAG differences in cognitive impairment and amyloid positivity
Residual FC-BAG was normally distributed (see Figure 2—figure supplement 1), and variance in FC-BAG did not significantly differ between the analysis sets, Levene’s statistic = 0.01, $$p \leq 0.988.$$ *An omnibus* ANOVA revealed significant differences in residual FC-BAG across the three groups, F[2,449] = 9.80, $p \leq 0.001.$ FC-BAG was 2.17 years older in CI participants compared to CN controls (β = 2.17, $$p \leq 0.030$$, ηp2 = 0.01; see Figure 2A&B, Table 2A). Follow-up t tests revealed that residual FC-BAG was significantly elevated in CI relative to CN/A+participants (pFDR < 0.001). FC-BAG was also 1.64 years lower in A+ participants compared to A− (β = –1.64, $$p \leq 0.035$$, ηp2 = 0.01), controlling for global CDR and the other covariates. Follow-up t tests revealed that residual FC-BAG was significantly lower in CN/A+ participants compared to CN/A− controls (pFDR = 0.002).
**Figure 2.:** *Group differences in functional connectivity (FC; A and B), structural (S; C and D), and multimodal (S+FC; E and F) brain age in the analysis sets.Comparisons are presented between cognitively normal (Clinical Dementia Rating [CDR] = 0) biomarker-negative controls (CN/A−; blue) vs. CN/A+ (green) vs. cognitively impaired participants (CI, red). Scatterplots (A, C, and E) show predicted vs. true age for each group. Colored lines and shaded areas represent group-specific regression lines and $95\%$ confidence regions. Dashed black lines represent perfect prediction. Violin plots (B, D, and F) show residual FC-brain age gap (BAG; controlling for true age) in each group. p values are reported from pairwise independent-samples t tests.* TABLE_PLACEHOLDER:Table 2.
Residual S-BAG was also normally distributed (see Figure 2—figure supplement 1), and variance in S-BAG did not significantly differ between the analysis sets, Levene’s statistic = 0.10, $$p \leq 0.902.$$ *An omnibus* ANOVA revealed significant differences in residual S-BAG across the three groups, F[2,449] = 20.64, $p \leq 0.001.$ S-BAG was 5.10 years older in CI participants compared to CN controls (β = 5.10, $p \leq 0.001$, ηp2 = 0.04; see Figure 2C&D, Table 2B). Follow-up t tests revealed that residual S-BAG was significantly elevated in CI participants relative to CN/A− and CN/A+ participants (pFDR’s < 0.001). S-BAG did not significantly differ as a function of amyloid positivity, controlling for CDR and the other covariates.
Residual S+FC-BAG was also normally distributed (see Figure 2—figure supplement 1), and variance in S+FC-BAG did not significantly differ between the analysis sets, Levene’s statistic = 0.89, $$p \leq 0.412.$$ *An omnibus* ANOVA revealed significant differences in residual S+FC-BAG across the three groups, F[2,449] = 21.84, $p \leq 0.001.$ S+FC-BAG was 4.31 years older in CI participants compared to CN controls (β = 4.1, $p \leq 0.001$, ηp2 = 0.04; see Figure 2E, F, Table 2C). Follow-up t tests revealed that residual FC-BAG was significantly elevated in CI participants relative to CN/A− and CN/A+ participants (pFDR’s < 0.001). S+FC-BAG did not significantly differ as a function of amyloid positivity, controlling for CDR and the other covariates.
## Relationships with amyloid markers
355 participants (144 CN/A−, 154 CN/A+, 57 CI) had an available amyloid PET scan, and 300 (120 CN/A−, 137 CN/A+, 43 CI) had an available CSF estimate of Aβ$\frac{42}{40.}$ In the FC model, FC-BAG was not significantly related with amyloid PET nor was there an interactive relationship with amyloid PET between groups (see Figure 3A). There were also no significant main effects or interactions between FC-BAG, S-BAG, or S+FC BAG and CSF Aβ$\frac{42}{40}$ (See Figure 3B, D and F).
**Figure 3.:** *Continuous relationships between amyloid biomarkers and functional connectivity (FC-brain age gap [BAG]; A and B), structural (S-BAG; C and D), and multimodal (S+FC BAG; E and F) BAG in the analysis sets.Scatterplots show amyloid PET (A, C, and E) and CSF AB42/40 (B, D, and F) as a function of residual BAG (controlling for true age) in each group. Colored lines and shaded areas represent group-specific regression lines and 95% confidence regions. Dashed black lines represent main effect regression lines across all groups.*
In the structural and multimodal models, there were significant main effects, such that greater S-BAG (β = 0.79, $$p \leq 0.004$$, ηp2 = 0.041; see Figure 3C) and greater S+FC BAG (β = 0.81, $$p \leq 0.015$$, ηp2 = 0.029; see Figure 3E) were both associated with greater amyloid PET. In the multimodal model only, this relationship was further characterized by a non-significant interaction (β = 1.16, $$p \leq 0.087$$, ηp2 = 0.014), such that the association was significantly positive in CI participants interaction (β = 1.53, $$p \leq 0.029$$, ηp2 = 0.092) but not in CN/A+ (β = –0.05, $$p \leq 0.881$$, ηp2 = 0.001).
## Relationships with tau markers
99 participants (42 CN/A–, 40 CN/A+, 17 CI) had an available tau PET scan, and 300 (120 CN/A–, 137 CN/A+, 43 CI) had an available CSF estimate of pTau-181/Aβ40. In the FC model, FC-BAG was not significantly related with tau PET or CSF pTau-181/Aβ40 (see Figure 4A and B). However, there was a non-significant interaction, suggesting a more positive association between CSF pTau-181/Aβ40 and FC-BAG in CI participants but not in CN controls (β = 0.02, $$p \leq 0.059$$, ηp2 = 0.016).
**Figure 4.:** *Continuous relationships between tau biomarkers and functional connectivity (FC-brain age gap [BAG]; A and B), structural (S-BAG; C and D), and multimodal (S+FC BAG; E and F) BAG in the analysis sets.Scatterplots show Tau PET summary (A, C, and E) and log-transformed CSF pTau/Aβ40 (B, D, and F) as a function of residual BAG (controlling for true age) in each group. Colored lines and shaded areas represent group-specific regression lines and 95% confidence regions. Dashed black lines represent main effect regression lines across all groups.*
In the structural and multimodal models, there were significant main effects, such that greater S-BAG (β = 0.02, $p \leq 0.001$, ηp2 = 0.141; see Figure 4C) and greater S+FC BAG (β = 0.02, $$p \leq 0.001$$, ηp2 = 0.110; see Figure 4E) were both associated with greater tau PET. These main effects were further characterized by significant interactions (S-BAG: β = 0.04, $p \leq 0.001$, ηp2 = 0.176; S+FC-BAG: β = 0.07, $p \leq 0.001$, ηp2 = 0.250), such that the positive association was only observed in CI participants, but not in the other groups.
Consistent with tau PET, CSF pTau/Aβ40 demonstrated similar interactive relationships, such that greater S-BAG (β = 0.02, $p \leq 0.001$, ηp2 = 0.052; see Figure 4D) and greater S+FC BAG (β = 0.04, $p \leq 0.001$, ηp2 = 0.075; see Figure 4F) were both associated with greater CSF pTau/Aβ40 in the CI participants, but not in the other groups.
## Relationships with cognition
445 participants (144 CN/A−, 153 CN/A+, 148 CI) had available performance measures from the cognitive composite tasks. In the FC model, there was a significant main effect, such that across all groups, greater FC-BAG was associated with lower cognitive composite score (β = –0.01, $$p \leq 0.006$$, ηp2 = 0.017; see Figure 5A). However, this effect was driven by group differences in both variables, as there were neither relationships between FC-BAG and cognition within any of the groups nor were there any significant interactions.
**Figure 5.:** *Continuous relationships between global cognition and functional connectivity (FC-brain age gap [BAG]; A), structural (S-BAG; B), and multimodal (S+FC BAG; C) in the analysis sets.Scatterplots show global cognition as a function of residual BAG (controlling for true age) in each group. Colored lines and shaded areas represent group-specific regression lines and 95% confidence regions. Dashed black lines represent main effect regression lines across all groups.*
In the structural model and multimodal models, there were significant main effects, such that greater S-BAG (β = –0.03, $p \leq 0.001$, ηp2 = 0.104; see Figure 5B) and greater S+FC BAG (β = –0.03, $p \leq 0.001$, ηp2 = 0.096; see Figure 5C) were both associated with lower cognitive composite scores. Both effects were further characterized by significant interactions such that the negative associations were observed in the CI participants, but not in the other groups (S-BAG: β = –0.03, $p \leq 0.001$, ηp2 = 0.045; S+FC-BAG: β = –0.04, $p \leq 0.001$, ηp2 = 0.047).
## Discussion
We first found that machine-learning models successfully predicted age when trained on FC, structural MRI, and multimodal datasets. As expected, the structural model predicted age with greater accuracy than the FC model, but the multimodal model outperformed both unimodal models. Second, BAG estimates from all models were significantly elevated in CI participants compared to CN controls. BAG estimates in the FC model were significantly reduced in cognitively normal participants with elevated amyloid, but no structural group differences were observed in presymptomatic stages. Third, interactive relationships were observed, such that greater BAG was associated with greater continuous AD biomarker load in CI, but not in CN, participants. Specifically, in the FC model, such a pattern only appeared in a non-significant interaction predicting CSF pTau/Aβ40. However, in the structural model, these interactions were significantly observed in relation to CSF pTau/Aβ40 and tau PET. In the multimodal model, these same interactions were also observed in addition to a non-significant interaction with amyloid PET. Finally, regarding cognitive relationships, similar interactive patterns were observed, such that in CI participants, greater BAG estimates from structural and multimodal models were associated with lower cognitive performance; however, this relationship was not observed in the FC model.
## Predicting brain age with multiple modalities
We found that a GPR model trained on structural MRI features predicted chronological age in a cognitively normal, amyloid-negative adult sample with an R2 of 0.81. This level of performance is comparable to other structural models, which have reported R2s ranging from 0.80 to 0.95 (Cole and Franke, 2017b; Liem et al., 2017; Eavani et al., 2018; Wang et al., 2019; Bashyam et al., 2020; Ly et al., 2020; Gong et al., 2021; Lee et al., 2022). As previously reported (Millar et al., 2022), the FC-trained model predicted age with an R2 of 0.68, again consistent with previous FC models, which have achieved R2s from 0.53 to 0.80 (Liem et al., 2017; Eavani et al., 2018; Gonneaud et al., 2021). Our observation that structural MRI outperformed FC in age prediction is also consistent with previous direct comparisons between modalities (Liem et al., 2017; Eavani et al., 2018; Dunås et al., 2021).
Importantly, however, there was only a modest positive correlation between FC and structural BAG estimates, after correcting for age-related biases, suggesting that functional and structural MRI capture distinct age-related signals. Indeed, the multimodal model outperformed both unimodal models by integrating these complementary signals. These observations, again, are consistent with other recent reports of multimodal age prediction models (Liem et al., 2017; Eavani et al., 2018; Engemann et al., 2020; Dunås et al., 2021). Future models may improve age prediction accuracy by combining data from structural, FC, and/or other neuroimaging modalities, several of which may be available in typical MRI sessions of multiple sequences.
## BAG as a marker of cognitive impairment
Structural BAG was elevated by 5.10 years in CI participants compared to CN controls. This effect is comparable to previous structural age prediction models, demonstrating elevations in AD and MCI samples between 5 and 10 years (Cole and Franke, 2017b; Franke and Gaser, 2019). As previously reported, FC BAG was also elevated in CI participants, but to a relatively smaller extent, i.e., 2.17 years (Millar et al., 2022). The multimodal BAG was similarly elevated in CI participants by 5.10 years. Thus, each model is clearly sensitive to group differences in AD status at the symptomatic stage.
Consistent with one previous report (Lee et al., 2022), we demonstrated that within the CI participants, BAG estimates were related to individual differences in AD biomarkers and cognitive function. These effects were most pronounced in the structural model, which showed relationships with tau biomarkers and cognition in the CI participants, and the multimodal model, which showed relationships with tau, cognition, and amyloid PET. Thus, age prediction models that include structural MRI (including unimodal and multimodal approaches) may be useful in tracking AD pathological progression and cognitive decline within the symptomatic stage of the disease.
## BAG as a marker of presymptomatic AD
We found that structural and multimodal BAG did not differ between cognitively normal participants with and without amyloid pathology. In cognitively normal participants, structural BAG estimates did not significantly associate with individual differences in any AD biomarkers. Overall, although structural and multimodal BAG estimates track well with some biomarkers of AD pathophysiology, as previously reported (Lee et al., 2022), our novel results suggest that these relationships are not observed until the symptomatic stage of the disease, at which point structural changes become more apparent.
As we have previously reported (Millar et al., 2022), FC-BAG was lower in presymptomatic AD participants compared to amyloid-negative controls. Extending beyond this group difference, we now also note that FC-BAG was negatively associated with amyloid PET in CN/A+ participants. The combined reduction of FC-BAG in the presymptomatic stage and increase in the symptomatic stage suggest a biphasic functional response to AD progression, which is partially consistent with some prior suggestions (Jagust and Mormino, 2011; Jones et al., 2016; Jones et al., 2017; Schultz et al., 2017; Wales and Leung, 2021; see Millar et al., 2022 for a more detailed discussion).
Interpretation of this biphasic pattern is still unclear, although the present results provide at least one novel insight. Specifically, one potential interpretation is that the ‘younger’ appearing FC pattern in the presymptomatic stage may reflect a compensatory response to early AD pathology (Cabeza et al., 2018). This interpretation leads to the prediction that reduced FC-BAG should be associated with better cognitive performance in the preclinical stage. However, this interpretation is not supported by the current results, as FC-BAG did not correlate with cognition in any of the analysis samples.
Alternatively, pathological AD-related FC disruptions may be orthogonal to healthy age-related FC differences, as supported by our previous observation that age and AD are predicted by mostly non-overlapping FC networks (Millar et al., 2022). For instance, the ‘younger’ FC pattern in CN/A+ participants may be driven by hyper-excitability in the preclinical stage (Harris et al., 2020; Ranasinghe et al., 2022). It is also worth considering that patterns of younger FC-BAG in CN/A+ participants may somehow correspond to a recent observation that patterns of youthful-appearing aerobic glycolysis are relatively preserved in the presymptomatic stage of AD (Goyal et al., 2022). Finally, this effect may simply be spuriously driven by poor performance of the FC brain age model, sample-specific noise, and/or statistical artifacts related to regression dilution and its correction (Butler et al., 2021). Hence, future studies should attempt to replicate these results in independent samples and further test potential theoretical interpretations.
## BAG as a marker of cognition
Although FC-BAG was not associated with individual differences in a global cognitive composite within any of our analysis samples, greater structural and multimodal BAG estimates were associated with lower cognitive performance within the CI participants. Hence, these estimates may be sensitive markers of cognitive decline in the symptomatic stage. This finding is consistent with previous reports that other structural brain age estimates are associated with cognitive performance in AD (Eavani et al., 2018), Down syndrome (Cole et al., 2017a), HIV (Petersen et al., 2021; Petersen et al., 2022), as well as cognitively normal controls (Richard et al., 2018).
## Limitations and future directions
The training sets included MRI scans from a range of sites, scanners, and acquisition sequence parameters, which may introduce noise and/or confounding variance into MRI features. We attempted to mitigate this problem by: [1] including only data from Siemens 3T scanners with similar protocols; [2] processing all MRI data through common pipelines and quality assessments; and [3] harmonizing across sites and scanners with ComBat (Fortin et al., 2017).
Additionally, the training set ($$n = 390$$) was relatively small compared to prior models, which have included training samples over 1,000 (e.g., Cole et al., 2015; Bashyam et al., 2020). Future studies may further improve model performance by including larger samples of well-characterized participants in the training set.
Although we took appropriate steps to detect and control for AD-related pathology in the CN/A− training sets, we were unable to control for other non-AD pathologies, e.g., Lewy body disease, TDP-43, etc., which may be present.
Structural MRI was quantified using the Desikan atlas (Desikan et al., 2006), which, although widely used, provides a relatively coarse parcellation of structural anatomy and, moreover, does not align with the parcellation used to define FC regions (Seitzman et al., 2020). Although the structural MRI data still outperformed FC in predicting age, future brain age models may further improve performance by using more refined and harmonized anatomical parcellations to define brain regions.
The sample size of continuous biomarker and cognitive analyses differed across the measures, depending on the availability, and was particularly low for analyses of tau PET. Future studies might improve upon this approach by a larger and more complete biomarker sample.
Moreover, estimates of BAG likely capture variance in early-life factors, which may obscure associations with AD and cognition, especially in cross-sectional designs (Vidal-Piñeiro et al., 2021). Future studies may improve the sensitivity of BAG estimates to disease-related markers by testing associations with longitudinal change.
Finally, although the Ances lab controls were relatively diverse, participants in other samples were mostly white and highly educated. Hence, these models may not be generalizable to broader samples. Future models would benefit by using more representative training samples.
## Conclusions
We compared three MRI-based machine-learning models in their ability to predict age, as well as their sensitivity to early-stage AD, AD biomarkers, and cognition. Although FC and structural MRI models were both successful in detecting differences related to healthy aging and cognitive impairment, we note clear evidence that these modalities capture complementary signals. Specifically, FC-BAG was uniquely reduced in cognitively normal participants with elevated amyloid, although the interpretation of this finding still warrants further investigation. In contrast, structural BAG was uniquely associated with biomarkers of AD pathology and cognitive function within the CI participants. Finally, the multimodal age prediction model, which combined FC and structural MRI, further improved the prediction of healthy age differences and also was related to biomarkers and cognition in CI participants. Thus, multimodal brain age models may be useful maximizing sensitivity to AD across the spectrum of disease progression.
## Funding Information
This paper was supported by the following grants:
## Data availability
This project utilized datasets obtained from the Knight ADRC and DIAN. The Knight ADRC and DIAN encourage and facilitate research by current and new investigators, and thus, the data and code are available to all qualified researchers after appropriate review. Requests for access to the data used in this study may be placed to the Knight ADRC Leadership Committee (https://knightadrc.wustl.edu/professionals-clinicians/request-center-resources/) and the DIAN Steering Committee (https://dian.wustl.edu/our-research/for-investigators/dian-observational-study-investigator-resources/data-request-form/). Requests for access to the Ances lab data may be placed to the corresponding author. Code used in this study is available at https://github.com/peterrmillar/MultimodalBrainAge (copy archived at swh:1:rev:de233b8fe813f5fcca317ce0a6353047f0dfbb92).
## References
1. Armitage SG. **An analysis of certain psychological tests used for the evaluation of brain injury**. *Psychological Monographs* (1946) **60** i1-i48. DOI: 10.1037/h0093567
2. Aschenbrenner AJ, Gordon BA, Benzinger TLS, Morris JC, Hassenstab JJ. **Influence of tau PET, amyloid PET, and hippocampal volume on cognition in Alzheimer disease**. *Neurology* (2018) **91** e859-e866. DOI: 10.1212/WNL.0000000000006075
3. Bashyam VM, Erus G, Doshi J, Habes M, Nasrallah IM, Truelove-Hill M, Srinivasan D, Mamourian L, Pomponio R, Fan Y, Launer LJ, Masters CL, Maruff P, Zhuo C, Völzke H, Johnson SC, Fripp J, Koutsouleris N, Satterthwaite TD, Wolf D, Gur RE, Gur RC, Morris J, Albert MS, Grabe HJ, Resnick S, Bryan RN, Wolk DA, Shou H, Davatzikos C. **Mri signatures of brain age and disease over the lifespan based on a deep brain network and 14 468 individuals worldwide**. *Brain* (2020) **143** 2312-2324. DOI: 10.1093/brain/awaa160
4. Bocancea DI, van Loenhoud AC, Groot C, Barkhof F, van der Flier WM, Ossenkoppele R. **Measuring resilience and resistance in aging and Alzheimer disease using residual methods: a systematic review and meta-analysis**. *Neurology* (2021) **97** 474-488. DOI: 10.1212/WNL.0000000000012499
5. Brier MR, Thomas JB, Ances BM. **Network dysfunction in alzheimer’s disease: refining the disconnection hypothesis**. *Brain Connectivity* (2014a) **4** 299-311. DOI: 10.1089/brain.2014.0236
6. Brier MR, Thomas JB, Snyder AZ, Wang L, Fagan AM, Benzinger T, Morris JC, Ances BM. **Unrecognized preclinical alzheimer disease confounds rs-fcmri studies of normal aging**. *Neurology* (2014b) **83** 1613-1619. DOI: 10.1212/WNL.0000000000000939
7. Butler ER, Chen A, Ramadan R, Le TT, Ruparel K, Moore TM, Satterthwaite TD, Zhang F, Shou H, Gur RC, Nichols TE, Shinohara RT. **Pitfalls in brain age analyses**. *Human Brain Mapping* (2021) **42** 4092-4101. DOI: 10.1002/hbm.25533
8. Cabeza R, Albert M, Belleville S, Craik FIM, Duarte A, Grady CL, Lindenberger U, Nyberg L, Park DC, Reuter-Lorenz PA, Rugg MD, Steffener J, Rajah MN. **Maintenance, reserve and compensation: the cognitive neuroscience of healthy ageing**. *Nature Reviews. Neuroscience* (2018) **19** 701-710. DOI: 10.1038/s41583-018-0068-2
9. Cherubini A, Caligiuri ME, Peran P, Sabatini U, Cosentino C, Amato F. **Importance of multimodal MRI in characterizing brain tissue and its potential application for individual age prediction**. *IEEE Journal of Biomedical and Health Informatics* (2016) **20** 1232-1239. DOI: 10.1109/JBHI.2016.2559938
10. Chien DT, Bahri S, Szardenings AK, Walsh JC, Mu F, Su M-Y, Shankle WR, Elizarov A, Kolb HC. **Early clinical PET imaging results with the novel PHF-tau radioligand [ F-18 ] -t807**. *Journal of Alzheimer’s Disease* (2013) **34** 457-468. DOI: 10.3233/JAD-122059
11. Cole JH, Leech R, Sharp DJ. **Prediction of brain age suggests accelerated atrophy after traumatic brain injury**. *Annals of Neurology* (2015) **77** 571-581. DOI: 10.1002/ana.24367
12. Cole JH, Annus T, Wilson LR, Remtulla R, Hong YT, Fryer TD, Acosta-Cabronero J, Cardenas-Blanco A, Smith R, Menon DK, Zaman SH, Nestor PJ, Holland AJ. **Brain-predicted age in down syndrome is associated with beta amyloid deposition and cognitive decline**. *Neurobiology of Aging* (2017a) **56** 41-49. DOI: 10.1016/j.neurobiolaging.2017.04.006
13. Cole JH, Franke K. **Predicting age using neuroimaging: innovative brain ageing biomarkers**. *Trends in Neurosciences* (2017b) **40** 681-690. DOI: 10.1016/j.tins.2017.10.001
14. Cole JH, Underwood J, Caan MWA, De Francesco D, van Zoest RA, Leech R, Wit F, Portegies P, Geurtsen GJ, Schmand BA, Schim van der Loeff MF, Franceschi C, Sabin CA, Majoie C, Winston A, Reiss P, Sharp DJ. **Increased brain-predicted aging in treated HIV disease**. *Neurology* (2017c) **88** 1349-1357. DOI: 10.1212/WNL.0000000000003790
15. Cole JH, Ritchie SJ, Bastin ME, Valdés Hernández MC, Muñoz Maniega S, Royle N, Corley J, Pattie A, Harris SE, Zhang Q, Wray NR, Redmond P, Marioni RE, Starr JM, Cox SR, Wardlaw JM, Sharp DJ, Deary IJ. **Brain age predicts mortality**. *Molecular Psychiatry* (2018) **23** 1385-1392. DOI: 10.1038/mp.2017.62
16. Desikan RS, Ségonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D, Buckner RL, Dale AM, Maguire RP, Hyman BT, Albert MS, Killiany RJ. **An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest**. *NeuroImage* (2006) **31** 968-980. DOI: 10.1016/j.neuroimage.2006.01.021
17. Dosenbach NUF, Nardos B, Cohen AL, Fair DA, Power JD, Church JA, Nelson SM, Wig GS, Vogel AC, Lessov-Schlaggar CN, Barnes KA, Dubis JW, Feczko E, Coalson RS, Pruett JR, Barch DM, Petersen SE, Schlaggar BL. **Prediction of individual brain maturity using fmri**. *Science* (2010) **329** 1358-1361. DOI: 10.1126/science.1194144
18. Dunås T, Wåhlin A, Nyberg L, Boraxbekk CJ. **Multimodal image analysis of apparent brain age identifies physical fitness as predictor of brain maintenance**. *Cerebral Cortex* (2021) **31** 3393-3407. DOI: 10.1093/cercor/bhab019
19. Eavani H, Habes M, Satterthwaite TD, An Y, Hsieh MK, Honnorat N, Erus G, Doshi J, Ferrucci L, Beason-Held LL, Resnick SM, Davatzikos C. **Heterogeneity of structural and functional imaging patterns of advanced brain aging revealed via machine learning methods**. *Neurobiology of Aging* (2018) **71** 41-50. DOI: 10.1016/j.neurobiolaging.2018.06.013
20. Engemann DA, Kozynets O, Sabbagh D, Lemaître G, Varoquaux G, Liem F, Gramfort A. **Combining magnetoencephalography with magnetic resonance imaging enhances learning of surrogate-biomarkers**. *eLife* (2020) **9**. DOI: 10.7554/eLife.54055
21. Fagan AM, Mintun MA, Mach RH, Lee S-Y, Dence CS, Shah AR, LaRossa GN, Spinner ML, Klunk WE, Mathis CA, DeKosky ST, Morris JC, Holtzman DM. **Inverse relation between in vivo amyloid imaging load and cerebrospinal fluid Abeta42 in humans**. *Annals of Neurology* (2006) **59** 512-519. DOI: 10.1002/ana.20730
22. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, van der Kouwe A, Killiany R, Kennedy D, Klaveness S, Montillo A, Makris N, Rosen B, Dale AM. **Whole brain segmentation**. *Neuron* (2002) **33** 341-355. DOI: 10.1016/S0896-6273(02)00569-X
23. Fischl B. **FreeSurfer**. *NeuroImage* (2012) **62** 774-781. DOI: 10.1016/j.neuroimage.2012.01.021
24. Fortin JP, Parker D, Tunç B, Watanabe T, Elliott MA, Ruparel K, Roalf DR, Satterthwaite TD, Gur RC, Gur RE, Schultz RT, Verma R, Shinohara RT. **Harmonization of multi-site diffusion tensor imaging data**. *NeuroImage* (2017) **161** 149-170. DOI: 10.1016/j.neuroimage.2017.08.047
25. Fortin JP, Cullen N, Sheline YI, Taylor WD, Aselcioglu I, Cook PA, Adams P, Cooper C, Fava M, McGrath PJ, McInnis M, Phillips ML, Trivedi MH, Weissman MM, Shinohara RT. **Harmonization of cortical thickness measurements across scanners and sites**. *NeuroImage* (2018) **167** 104-120. DOI: 10.1016/j.neuroimage.2017.11.024
26. Fox MD, Zhang D, Snyder AZ, Raichle ME. **The global signal and observed anticorrelated resting state brain networks**. *J Neurophysiol* (2009) **101** 3270-3283. DOI: 10.1152/jn.90777.2008
27. Franke K, Ziegler G, Klöppel S, Gaser C. **Estimating the age of healthy subjects from T1-weighted MRI scans using kernel methods: exploring the influence of various parameters**. *NeuroImage* (2010) **50** 883-892. DOI: 10.1016/j.neuroimage.2010.01.005
28. Franke K, Gaser C. **Longitudinal changes in individual brainage in healthy aging, mild cognitive impairment, and alzheimer’s disease**. *GeroPsych* (2012) **25** 235-245. DOI: 10.1024/1662-9647/a000074
29. Franke K, Gaser C, Manor B, Novak V. **Advanced brainage in older adults with type 2 diabetes mellitus**. *Frontiers in Aging Neuroscience* (2013) **5**. DOI: 10.3389/fnagi.2013.00090
30. Franke K, Gaser C. **Ten years of brainage as a neuroimaging biomarker of brain aging: what insights have we gained?**. *Frontiers in Neurology* (2019) **10**. DOI: 10.3389/fneur.2019.00789
31. Frisoni GB, Fox NC, Jack CR, Scheltens P, Thompson PM. **The clinical use of structural MRI in Alzheimer disease**. *Nature Reviews. Neurology* (2010) **6** 67-77. DOI: 10.1038/nrneurol.2009.215
32. Gaser C, Franke K, Klöppel S, Koutsouleris N, Sauer H. **BrainAGE in mild cognitive impaired patients: predicting the conversion to alzheimer’s disease**. *PLOS ONE* (2013) **8**. DOI: 10.1371/journal.pone.0067346
33. Gong W, Beckmann CF, Vedaldi A, Smith SM, Peng H. **Optimising a simple fully convolutional network for accurate brain age prediction in the PAC 2019 challenge**. *Frontiers in Psychiatry* (2021) **12**. DOI: 10.3389/fpsyt.2021.627996
34. Gonneaud J, Baria AT, Binette AP, Gordon BA, Chhatwal JP, Cruchaga C. **Accelerated functional brain aging in pre-clinical familial alzheimer’s disease**. *Nat Commun* (2021) **12**. DOI: 10.1038/s41467-021-25492-9
35. Goodglass H, Kaplan E. *Boston Diagnostic Aphasia Examination Booklet, III: Oral Expression: Animal Naming Fluency in Controlled Association* (1983)
36. Goyal MS, Blazey TM, Su Y, Couture LE, Durbin TJ, Bateman RJ, Benzinger TLS, Morris JC, Raichle ME, Vlassenko AG. **Persistent metabolic youth in the aging female brain**. *PNAS* (2019) **116** 3251-3255. DOI: 10.1073/pnas.1815917116
37. Goyal MS, Blazey T, Metcalf NV, McAvoy MP, Strain J, Rahmani M, Durbin TJ, Xiong C, Benzinger TLS, Morris JC, Raichle ME, Vlassenko AG. **Brain Aerobic Glycolysis and Resilience in Alzheimer Disease**. *bioRxiv* (2022). DOI: 10.1101/2022.06.21.497006
38. Grober E, Buschke H, Crystal H, Bang S, Dresner R. **Screening for dementia by memory testing**. *Neurology* (1988) **38** 900-903. DOI: 10.1212/wnl.38.6.900
39. Guo T, Korman D, La Joie R, Shaw LM, Trojanowski JQ, Jagust WJ, Landau SM. **Normalization of CSF ptau measurement by aβ40 improves its performance as a biomarker of alzheimer’s disease**. *Alzheimer’s Research & Therapy* (2020) **12**. DOI: 10.1186/s13195-020-00665-8
40. Hansson O, Lehmann S, Otto M, Zetterberg H, Lewczuk P. **Advantages and disadvantages of the use of the CSF amyloid β (Aβ) 42/40 ratio in the diagnosis of alzheimer’s disease**. *Alzheimer’s Research & Therapy* (2019) **11**. DOI: 10.1186/s13195-019-0485-0
41. Harris SS, Wolf F, De Strooper B, Busche MA. **Tipping the scales: peptide-dependent dysregulation of neural circuit dynamics in alzheimer’s disease**. *Neuron* (2020) **107** 417-435. DOI: 10.1016/j.neuron.2020.06.005
42. Hwang G, Abdulkadir A, Erus G, Habes M, Pomponio R, Shou H, Doshi J, Mamourian E, Rashid T, Bilgel M, Fan Y, Sotiras A, Srinivasan D, Morris JC, Albert MS, Bryan NR, Resnick SM, Nasrallah IM, Davatzikos C, Wolk DA. **Disentangling Alzheimer’s disease neurodegeneration from typical brain ageing using machine learning**. *Brain Communications* (2022) **4**. DOI: 10.1093/braincomms/fcac117
43. Jagust WJ, Mormino EC. **Lifespan brain activity, β-amyloid, and alzheimer’s disease**. *Trends in Cognitive Sciences* (2011) **15** 520-526. DOI: 10.1016/j.tics.2011.09.004
44. Johnson WE, Li C, Rabinovic A. **Adjusting batch effects in microarray expression data using empirical Bayes methods**. *Biostatistics* (2007) **8** 118-127. DOI: 10.1093/biostatistics/kxj037
45. Jones DT, Knopman DS, Gunter JL, Graff-Radford J, Vemuri P, Boeve BF, Petersen RC, Weiner MW, Jack CR. **Cascading network failure across the alzheimer’s disease spectrum**. *Brain* (2016) **139** 547-562. DOI: 10.1093/brain/awv338
46. Jones DT, Graff-Radford J, Lowe VJ, Wiste HJ, Gunter JL, Senjem ML, Botha H, Kantarci K, Boeve BF, Knopman DS, Petersen RC, Jack CR. **Tau, amyloid, and cascading network failure across the alzheimer’s disease spectrum**. *Cortex; a Journal Devoted to the Study of the Nervous System and Behavior* (2017) **97** 143-159. DOI: 10.1016/j.cortex.2017.09.018
47. Klunk WE, Engler H, Nordberg A, Wang Y, Blomqvist G, Holt DP, Bergström M, Savitcheva I, Huang G, Estrada S, Ausén B, Debnath ML, Barletta J, Price JC, Sandell J, Lopresti BJ, Wall A, Koivisto P, Antoni G, Mathis CA, Långström B. **Imaging brain amyloid in Alzheimer’s disease with Pittsburgh compound-B**. *Annals of Neurology* (2004) **55** 306-319. DOI: 10.1002/ana.20009
48. Koutsouleris N, Davatzikos C, Borgwardt S, Gaser C, Bottlender R, Frodl T, Falkai P, Riecher-Rössler A, Möller H-J, Reiser M, Pantelis C, Meisenzahl E. **Accelerated brain aging in schizophrenia and beyond: a neuroanatomical marker of psychiatric disorders**. *Schizophrenia Bulletin* (2014) **40** 1140-1153. DOI: 10.1093/schbul/sbt142
49. Le TT, Kuplicki RT, McKinney BA, Yeh HW, Thompson WK, Paulus MP. **A nonlinear simulation framework supports adjusting for age when analyzing brainage**. *Frontiers in Aging Neuroscience* (2018) **10**. DOI: 10.3389/fnagi.2018.00317
50. Lee J, Burkett BJ, Min HK, Senjem ML, Lundt ES, Botha H, Graff-Radford J, Barnard LR, Gunter JL, Schwarz CG, Kantarci K, Knopman DS, Boeve BF, Lowe VJ, Petersen RC, Jack CR, Jones DT. **Deep learning-based brain age prediction in normal aging and dementia**. *Nature Aging* (2022) **2** 412-424. DOI: 10.1038/s43587-022-00219-7
51. Liang H, Zhang F, Niu X. **Investigating systematic bias in brain age estimation with application to post-traumatic stress disorders**. *Human Brain Mapping* (2019) **40** 3143-3152. DOI: 10.1002/hbm.24588
52. Liem F, Varoquaux G, Kynast J, Beyer F, Kharabian Masouleh S, Huntenburg JM, Lampe L, Rahim M, Abraham A, Craddock RC, Riedel-Heller S, Luck T, Loeffler M, Schroeter ML, Witte AV, Villringer A, Margulies DS. **Predicting brain-age from multimodal imaging data captures cognitive impairment**. *NeuroImage* (2017) **148** 179-188. DOI: 10.1016/j.neuroimage.2016.11.005
53. Ly M, Yu GZ, Karim HT, Muppidi NR, Mizuno A, Klunk WE, Aizenstein HJ. **Improving brain age prediction models: incorporation of amyloid status in alzheimer’s disease**. *Neurobiology of Aging* (2020) **87** 44-48. DOI: 10.1016/j.neurobiolaging.2019.11.005
54. MathWorks
2021Regression learner app1.2.1Internethttps://www.mathworks.com/help/stats/regression-learner-app.html. *Internet* (2021)
55. McKay NS, Gordon BA, Hornbeck RC, Jack CR, Koeppe R, Flores S, Keefe S, Hobbs DA, Joseph-Mathurin N, Wang Q, Rahmani F, Chen CD, McCullough A, Koudelis D, Chua J, Ances BM, Millar PR, Nickels M, Perrin RJ, Allegri RF, Berman SB, Brooks WS, Cash DM, Chhatwal JP, Farlow MR, Fox NC, Fulham M, Ghetti B, Graff-Radford N, Ikeuchi T, Day G, Klunk W, Levin J, Lee JH, Martins R, Masters CL, McConathy J, Mori H, Noble JM, Rowe C, Salloway S, Sanchez-Valle R, Schofield PR, Shimada H, Shoji M, Su Y, Suzuki K, Vöglein J, Yakushev I, Swisher L, Cruchaga C, Hassenstab J, Karch C, McDade E, Xiong C, Morris JC, Bateman RJ, Benzinger TLS. **Neuroimaging within the Dominantly Inherited Alzheimer’s Network (DIAN): PET and MRI**. *bioRxiv* (2022). DOI: 10.1101/2022.03.25.485799
56. McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Kawas CH, Klunk WE, Koroshetz WJ, Manly JJ, Mayeux R, Mohs RC, Morris JC, Rossor MN, Scheltens P, Carrillo MC, Thies B, Weintraub S, Phelps CH. **The diagnosis of dementia due to alzheimer’s disease: recommendations from the national institute on aging-alzheimer’s association workgroups on diagnostic guidelines for alzheimer’s disease**. *Alzheimer’s & Dementia* (2011) **7** 263-269. DOI: 10.1016/j.jalz.2011.03.005
57. Millar PR, Ances BM, Gordon BA, Benzinger TLS, Morris JC, Balota DA. **Evaluating cognitive relationships with resting-state and task-driven blood oxygen level-dependent variability**. *Journal of Cognitive Neuroscience* (2021) **33** 279-302. DOI: 10.1162/jocn_a_01645
58. Millar PR, Luckett PH, Gordon BA, Benzinger TLS, Schindler SE, Fagan AM, Cruchaga C, Bateman RJ, Allegri R, Jucker M, Lee JH, Mori H, Salloway SP, Yakushev I, Morris JC, Ances BM. **Predicting brain age from functional connectivity in symptomatic and preclinical alzheimer disease**. *NeuroImage* (2022) **256**. DOI: 10.1016/j.neuroimage.2022.119228
59. Mishra S, Gordon BA, Su Y, Christensen J, Friedrichsen K, Jackson K, Hornbeck R, Balota DA, Cairns NJ, Morris JC, Ances BM, Benzinger TLS. **AV-1451 PET imaging of tau pathology in preclinical Alzheimer disease: defining a summary measure**. *NeuroImage* (2017) **161** 171-178. DOI: 10.1016/j.neuroimage.2017.07.050
60. Morris JC. **The clinical dementia rating (CDR): current version and scoring rules**. *Neurology* (1993) **43** 2412-2414. DOI: 10.1212/wnl.43.11.2412-a
61. Nielsen AN, Greene DJ, Gratton C, Dosenbach NUF, Petersen SE, Schlaggar BL. **Evaluating the prediction of brain maturity from functional connectivity after motion artifact denoising**. *Cerebral Cortex* (2019) **29** 2455-2469. DOI: 10.1093/cercor/bhy117
62. Petersen KJ, Metcalf N, Cooley S, Tomov D, Vaida F, Paul R, Ances BM. **Accelerated brain aging and cerebral blood flow reduction in persons with human immunodeficiency virus**. *Clinical Infectious Diseases* (2021) **73** 1813-1821. DOI: 10.1093/cid/ciab169
63. Petersen KJ, Strain JF, Cooley SA, Vaida FF, Ances BM. **Machine learning quantifies accelerated white-matter aging in persons with HIV**. *The Journal of Infectious Diseases* (2022) **226** 49-58. DOI: 10.1093/infdis/jiac156
64. Power JD, Barnes KA, Snyder AZ, Schlaggar BL, Petersen SE. **Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion**. *NeuroImage* (2012) **59** 2142-2154. DOI: 10.1016/j.neuroimage.2011.10.018
65. Ranasinghe KG, Verma P, Cai C, Xie X, Kudo K, Gao X, Lerner H, Mizuiri D, Strom A, Iaccarino L, La Joie R, Miller BL, Gorno-Tempini ML, Rankin KP, Jagust WJ, Vossel K, Rabinovici GD, Raj A, Nagarajan SS. **Altered excitatory and inhibitory neuronal subpopulation parameters are distinctly associated with tau and amyloid in Alzheimer’s disease**. *eLife* (2022) **11**. DOI: 10.7554/eLife.77850
66. Rasmussen CE, von Luxburg U, Rätsch G, Carbonell JG, Siekmann J. *Advanced Lectures on Machine Learning* (2004) 63-71. DOI: 10.1007/b100712
67. R Development Core Team
2020R: A language and environment for statistical computing1.2.1Vienna, AustriaR Foundation for Statistical Computinghttps://www.r-project.org/. (2020)
68. Richard G, Kolskår K, Sanders A-M, Kaufmann T, Petersen A, Doan NT, Monereo Sánchez J, Alnæs D, Ulrichsen KM, Dørum ES, Andreassen OA, Nordvik JE, Westlye LT. **Assessing distinct patterns of cognitive aging using tissue-specific brain age prediction based on diffusion tensor imaging and brain morphometry**. *PeerJ* (2018) **6**. DOI: 10.7717/peerj.5908
69. Satterthwaite TD, Wolf DH, Loughead J, Ruparel K, Elliott MA, Hakonarson H, Gur RC, Gur RE. **Impact of in-scanner head motion on multiple measures of functional connectivity: relevance for studies of neurodevelopment in youth**. *NeuroImage* (2012) **60** 623-632. DOI: 10.1016/j.neuroimage.2011.12.063
70. Schindler SE, Gray JD, Gordon BA, Xiong C, Batrla-Utermann R, Quan M, Wahl S, Benzinger TLS, Holtzman DM, Morris JC, Fagan AM. **Cerebrospinal fluid biomarkers measured by elecsys assays compared to amyloid imaging**. *Alzheimer’s & Dementia* (2018) **14** 1460-1469. DOI: 10.1016/j.jalz.2018.01.013
71. Schultz AP, Chhatwal JP, Hedden T, Mormino EC, Hanseeuw BJ, Sepulcre J, Huijbers W, LaPoint M, Buckley RF, Johnson KA, Sperling RA. **Phases of hyperconnectivity and hypoconnectivity in the default mode and salience networks track with amyloid and tau in clinically normal individuals**. *The Journal of Neuroscience* (2017) **37** 4323-4331. DOI: 10.1523/JNEUROSCI.3263-16.2017
72. Seitzman BA, Gratton C, Marek S, Raut RV, Dosenbach NUF, Schlaggar BL, Petersen SE, Greene DJ. **A set of functionally-defined brain regions with improved representation of the subcortex and cerebellum**. *NeuroImage* (2020) **206**. DOI: 10.1016/j.neuroimage.2019.116290
73. Shulman GL, Pope DLW, Astafiev SV, McAvoy MP, Snyder AZ, Corbetta M. **Right hemisphere dominance during spatial selective attention and target detection occurs outside the dorsal frontoparietal network**. *J Neurosci* (2010) **30** 3640-3651. DOI: 10.1523/JNEUROSCI.4085-09.2010
74. Smith SM, Vidaurre D, Alfaro-Almagro F, Nichols TE, Miller KL. **Estimation of brain age delta from brain imaging**. *NeuroImage* (2019) **200** 528-539. DOI: 10.1016/j.neuroimage.2019.06.017
75. Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, Fagan AM. **Toward defining the preclinical stages of alzheimer’s disease: recommendations from the national institute on aging and the alzheimer’s association workgroup**. *Alzheimers Dement* (2011) **7** 280-292. DOI: 10.1016/j.jalz.2011.03.003
76. Su Y, D’Angelo GM, Vlassenko AG, Zhou G, Snyder AZ, Marcus DS, Blazey TM, Christensen JJ, Vora S, Morris JC, Mintun MA, Benzinger TLS. **Quantitative analysis of PIB-PET with freesurfer rois**. *PLOS ONE* (2013) **8**. DOI: 10.1371/journal.pone.0073377
77. Su Y, Flores S, Hornbeck RC, Speidel B, Vlassenko AG, Gordon BA, Koeppe RA, Klunk WE, Xiong C, Morris JC, Benzinger TLS. **Utilizing the centiloid scale in cross-sectional and longitudinal PIB PET studies**. *NeuroImage. Clinical* (2018) **19** 406-416. DOI: 10.1016/j.nicl.2018.04.022
78. Su Y, Flores S, Wang G, Hornbeck RC, Speidel B, Joseph-Mathurin N, Vlassenko AG, Gordon BA, Koeppe RA, Klunk WE, Jack CR, Farlow MR, Salloway S, Snider BJ, Berman SB, Roberson ED, Brosch J, Jimenez-Velazques I, van Dyck CH, Galasko D, Yuan SH, Jayadev S, Honig LS, Gauthier S, Hsiung GYR, Masellis M, Brooks WS, Fulham M, Clarnette R, Masters CL, Wallon D, Hannequin D, Dubois B, Pariente J, Sanchez-Valle R, Mummery C, Ringman JM, Bottlaender M, Klein G, Milosavljevic-Ristic S, McDade E, Xiong C, Morris JC, Bateman RJ, Benzinger TLS. **Comparison of pittsburgh compound B and florbetapir in cross-sectional and longitudinal studies**. *Alzheimer’s & Dementia* (2019) **11** 180-190. DOI: 10.1016/j.dadm.2018.12.008
79. Thomas JB, Brier MR, Snyder AZ, Vaida FF, Ances BM. **Pathways to neurodegeneration: effects of HIV and aging on resting-state functional connectivity**. *Neurology* (2013) **80** 1186-1193. DOI: 10.1212/WNL.0b013e318288792b
80. Van Dijk KRA, Sabuncu MR, Buckner RL. **The influence of head motion on intrinsic functional connectivity MRI**. *NeuroImage* (2012) **59** 431-438. DOI: 10.1016/j.neuroimage.2011.07.044
81. Vidal-Piñeiro D, Wang Y, Krogsrud SK, Amlien IK, Baaré WFC, Bartrés-Faz D, Bertram L, Brandmaier AM, Drevon CA, Düzel S, Ebmeier KP, Henson RN, Junque C, Kievit RA, Kühn S, Leonardsen E, Lindenberger U, Madsen KS, Magnussen F, Mowinckel AM, Nyberg L, Roe JM, Segura B, Smith SM, Sørensen Ø, Suri S, Westerhausen R, Zalesky A, Zsoldos E, Walhovd KB, Fjell AM. **Individual Variations in “Brain Age” Relate to Early Life Factors More than to Longitudinal Brain Change**. *bioRxiv* (2021). DOI: 10.1101/2021.02.08.428915
82. Vlassenko AG, McCue L, Jasielec MS, Su Y, Gordon BA, Xiong C, Holtzman DM, Benzinger TLS, Morris JC, Fagan AM. **Imaging and cerebrospinal fluid biomarkers in early preclinical Alzheimer disease**. *Annals of Neurology* (2016) **80** 379-387. DOI: 10.1002/ana.24719
83. Volluz KE, Schindler SE, Henson RL, Xiong C, Gordon BA, Benzinger TLS, Holtzman DM, Morris JC, Fagan AM. **Correspondence of CSF biomarkers measured by lumipulse assays with amyloid PET**. *Alzheimer’s & Dementia* (2021) **17**. DOI: 10.1002/alz.051085
84. Wales RM, Leung HC. **The effects of amyloid and tau on functional network connectivity in older populations**. *Brain Connectivity* (2021) **11** 599-612. DOI: 10.1089/brain.2020.0902
85. Wang J, Knol MJ, Tiulpin A, Dubost F, de Bruijne M, Vernooij MW, Adams HHH, Ikram MA, Niessen WJ, Roshchupkin GV. **Gray matter age prediction as a biomarker for risk of dementia**. *PNAS* (2019) **116** 21213-21218. DOI: 10.1073/pnas.1902376116
86. Wong DF, Rosenberg PB, Zhou Y, Kumar A, Raymont V, Ravert HT, Dannals RF, Nandi A, Brasić JR, Ye W, Hilton J, Lyketsos C, Kung HF, Joshi AD, Skovronsky DM, Pontecorvo MJ. **In vivo imaging of amyloid deposition in Alzheimer disease using the radioligand 18F-AV-45 (florbetapir [ corrected ] F 18)**. *Journal of Nuclear Medicine* (2010) **51** 913-920. DOI: 10.2967/jnumed.109.069088
87. Yu M, Linn KA, Cook PA, Phillips ML, McInnis M, Fava M, Trivedi MH, Weissman MM, Shinohara RT, Sheline YI. **Statistical harmonization corrects site effects in functional connectivity measurements from multi-site fMRI data**. *Human Brain Mapping* (2018) **39** 4213-4227. DOI: 10.1002/hbm.24241
|
---
title: Evidence for dopaminergic involvement in endogenous modulation of pain relief
authors:
- Simon Desch
- Petra Schweinhardt
- Ben Seymour
- Herta Flor
- Susanne Becker
journal: eLife
year: 2023
pmcid: PMC9988263
doi: 10.7554/eLife.81436
license: CC BY 4.0
---
# Evidence for dopaminergic involvement in endogenous modulation of pain relief
## Abstract
Relief of ongoing pain is a potent motivator of behavior, directing actions to escape from or reduce potentially harmful stimuli. Whereas endogenous modulation of pain events is well characterized, relatively little is known about the modulation of pain relief and its corresponding neurochemical basis. Here, we studied pain modulation during a probabilistic relief-seeking task (a ‘wheel of fortune’ gambling task), in which people actively or passively received reduction of a tonic thermal pain stimulus. We found that relief perception was enhanced by active decisions and unpredictability, and greater in high novelty-seeking trait individuals, consistent with a model in which relief is tuned by its informational content. We then probed the roles of dopaminergic and opioidergic signaling, both of which are implicated in relief processing, by embedding the task in a double-blinded cross-over design with administration of the dopamine precursor levodopa and the opioid receptor antagonist naltrexone. We found that levodopa enhanced each of these information-specific aspects of relief modulation but no significant effects of the opioidergic manipulation. These results show that dopaminergic signaling has a key role in modulating the perception of pain relief to optimize motivation and behavior.
## Introduction
One of the most powerful and universally appreciated aspects of being in pain is the desire for relief, and the positive sensation of pleasure once achieved. However, in contrast to pain itself, much less is known about how the perception of relief is modulated by various behavioral and motivational factors (Becker et al., 2015; Leknes et al., 2008). Theories of pain have suggested that one reason pain itself is modulated is to help optimize the way in which it directs protective behavior (Walters and Williams, 2019). For instance, if pain is increased when learning and responding to it, but reduced in situations in which pain might actually interfere with optimal behavior will have a greater long-term benefit (Fields, 2018; Seymour, 2019). Whether this principle extends to relief has not been tested.
Endogenous modulation of pain involves a number of different processes mediated by distinct descending signaling pathways, and involves at least two critical neurochemical systems: opioidergic and dopaminergic (Bannister, 2019). For instance, opioid signaling has been shown to play a key role in behavioral relief motivation (in rodents, Navratilova et al., 2015b), placebo analgesia (Benedetti, 1996; Eippert et al., 2009), conditioned pain modulation (King et al., 2013), and relief perception (Sirucek et al., 2021). Dopaminergic signaling has clearly been shown to play a role in conditioned place preference induced by pain relief (through activity in midbrain dopaminergic neurons, Navratilova et al., 2012; Navratilova et al., 2015a; Xie et al., 2014), suggesting similar mechanisms as in the well-studied role of dopamine in food rewards (dopaminergic ‘wanting’ versus opioidergic ‘liking’; Barbano and Cador, 2006; Barbano and Cador, 2007; Berridge et al., 2009; Smith et al., 2011). Dopaminergic signaling is also implicated in inhibition of pain by extrinsic rewards, that is rewards external to the pain system (Becker et al., 2013). Hence, whilst it is clear that both opioidergic and dopaminergic signaling play core roles in relief motivation, it isn’t known which system primarily shapes perceived relief as a function of motivation.
The aim of the present study was therefore first to better characterize information processing aspects of relief motivation, and second to investigate the roles of dopaminergic and opioidergic signaling in pain relief perception and modulation. We expected that pain relief would be modulated by the value of information it carries, as hence enhanced by (i) active vs passive reception and thus controllability, since this reflects potential to exploit relief information; (ii) unpredictability, since this reflects the extra information carried by surprising events, and (iii) trait novelty-seeking, since this reflects individual information sensitivity. At the same time, we aimed to identify the potential role of dopamine and opioids for each of these factors, in particular to explore whether increased dopamine availability would enhance endogenous pain relief under these conditions, and whether modulation could be reduced by blocking opioid receptors. We expected increased dopamine availability to enhance phasic release of dopamine in response to rewards, and hence, to increase the effect of active compared to passive reception of pain relief. In contrast, we expected the inhibition of endogenous opioid signaling to decrease the effect of active controllability on pain relief. The latter is based on the observation that blocking of opioid receptors attenuates other types of endogenous pain inhibition such as placebo analgesia (Benedetti, 1996; Eippert et al., 2009) or conditioned pain modulation (King et al., 2013). Finally, we aimed to identify whether an increase or decrease in the modulation of perceived relief by dopamine and opioids is reflected in corresponding increases or decreases in the selection of a more advantageous option, that is decision-making during probabilistic learning.
To test these hypotheses, we employed a previously developed wheel of fortune task utilizing relief of a tonic capsaicin-sensitive thermal pain stimulus as ‘wins’, and allowing to quantify endogenous pain inhibition induced by gaining pain relief in active versus passive conditions (Becker et al., 2015). To test the roles of dopamine and opioids, we analyzed and report data of $$n = 28$$ healthy volunteers who ingested either a single dose of the dopamine precursor levodopa (150 mg), the opioid antagonist naltrexone (50 mg), or placebo in separate testing sessions (double-blinded, placebo controlled cross-over, i.e. within-subjects, design). To allow also the assessment of reinforcement learning, a probabilistic reward schedule associated with the participants’ choices in the wheel of fortune was implemented.
## Endogenous modulation of active pain relief seeking under placebo
To test whether playing the wheel of fortune game induced endogenous pain inhibition by gaining pain relief during active (controllable) decision-making, a test condition in which participants actively engaged in the game and ‘won’ relief of a tonic thermal pain stimulus in the game was compared to a control condition with passive receipt of the same outcomes (Figure 1). As a further comparator the game included an opposite condition in which participants received increases of the thermal stimulation as punishment. This active loss condition was also matched by a passive condition involving receipt of the same course of nociceptive input. We implemented two outcome measures, an explicit rating of perceived intensity after pain relief or increase, and a behavioral measure of perceptual sensitization or habituation to the underlying tonic stimulation within each trial of the game (see Figure 1). The effects of controllability on pain perception were tested in separate linear mixed effects models predicting the outcome measures by the outcome condition, the trial type (active test trials vs. passive control trials), and their interaction in each drug condition. Comparing the effects of active versus passive trials between the pain relief and the pain increase condition (interaction ‘outcome × trial type’) allowed also to test for unspecific effects of arousal and/or distraction: if the effects seen in the active compared to the passive condition were due to such unspecific effects then actively engaging in the game should equally affect pain in both, win and lose trials. In contrast, if the effects were due to increased controllability then we expected to see pain inhibition in win trials but equal or increased pain perception in lose trials. We used post-hoc comparisons to test direction and significance of differences in either outcome condition and report standardized effect sizes (d) for these differences. Note that all reported effect sizes account for random variation within the sample, providing an estimate for the underlying population; due to considerable variance between participants in the present study, this resulted in comparatively small effect sizes.
**Figure 1.:** *Time line of one trial with active decision-making (test trials) of the wheel of fortune game.Experimental pain was implemented using contact heat stimulation on capsaicin sensitized skin on the forearm. In each trial, the temperature increased from a baseline of 30 °C to a predetermined moderately painful stimulation intensity perceived as moderately painful. In each testing session, one of the two colors (pink and blue) of the wheel was associated with a higher chance to win pain relief (counterbalanced across subjects and drug conditions). Pain relief (win) as outcome of the wheel of fortune game (depicted in green) and, pain increase (loss; depicted in red) were implemented as phasic changes in stimulation intensity offsetting from the tonic painful stimulation. Based on a probabilistic reward schedule for these outcomes, participants could learn which color was associated with a better chance to win pain relief. In passive control trials and neutral trials participants did not play the game but had to press a black button after which the wheel started spinning and landed on a random position with no pointer on the wheel. Trials with active decision-making were matched by passive control trials without decision making but the same nociceptive input (control trials), resulting in the same number of pain increase and pain decrease trials as in the active condition. In neutral trials the temperature did not change during the outcome interval of the wheel. Two outcome measures were implemented in all trial types: (i) after the phasic changes during the outcome phase participants rated the perceived momentary intensity of the stimulation on a visual analogue scale (‘VAS intensity’); (ii) after this rating, participants had to adjust the temperature to match the sensation they had memorized at the beginning of the trial, i.e. the initial perception of the tonic stimulation intensity (‘self-adjustment of temperature’). This perceptual discrimination task served as a behavioral assessment of pain sensitization and habituation across the course of one trial. One trial lasted approximately 30 s, phasic offsets occurred after approximately 10 s of tonic pain stimulation. Adapted from Becker et al., 2015. Figure 1 is reproduced from Figure 1 in Becker et al., 2015.*
## Ratings of perceived pain
Replicating previous results, in the placebo (i.e. non-drug) condition participants rated the thermal stimulation as less intense after actively winning pain relief compared to the passive control condition, as rated on visual analogue scales (VAS) from ‘no sensation’ [0] over ‘just painful’ [100] to ‘most intense pain tolerable’ [200]. Furthermore, participants also rated the stimulation as more intense after actively losing compared to the passive control condition (Figure 2A; interaction ‘outcome × trial type’, F[1,1040]=64.14, $p \leq 0.001$; pairwise comparisons: win: test vs. control $p \leq 0.001$, standardized effect size $d = 0.16$; lose: test vs. control, $p \leq 0.001$, $d = 0.27$). This shows that perception of both relief and pain are enhanced by active (instrumental) controllability, as hypothesized.
**Figure 2.:** *Effects of active versus passive condition after pain relief and pain increase in each drug condition.Means (bars) and 95% confidence intervals of means (error bars) for VAS pain intensity ratings (A, B, C) and behaviorally assessed pain perception (D, E, F; within-trial sensitization in pain perception in °C) for each drug session (placebo: n=28, levodopa: n=27, naltrexone: n=28). d indicates the standardized effect size after controlling for random effects and residual variance. ** p<0.01, *** p<0.001, for post-hoc comparisons of test versus control trials.*
As in the placebo condition, participants rated the thermal stimulation as significantly less intense after active relief winning in the wheel of fortune task, and as significantly more intense after receiving phasic pain increases (‘losing’) compared to the respective passive control condition under levodopa (pairwise comparisons: win: test vs. control $p \leq 0.001$, $d = 0.31$; lose: test vs. control, $p \leq 0.001$, $d = 0.32$; Figure 2B) as well as naltrexone (pairwise comparisons: win: test vs. control $p \leq 0.001$, $d = 0.22$; lose: test vs. control, $p \leq 0.001$, $d = 0.27$; Figure 2C).
Moreover, the effect of active relief or increases on pain modulation was differentially modulated by the drugs (Table 1; interaction ‘drug × outcome’, F(2, 1587.30)=4.52, $$p \leq 0.011$$). Specifically, the effect of active relief on perception was significantly larger in the levodopa condition compared to the placebo condition (post-hoc comparison $$p \leq 0.007$$, $d = 0.23$; Figure 3A). No significant difference was found for the naltrexone compared to the placebo condition ($$p \leq 0.252$$, $d = 0.12$). Endogenous modulation did not significantly differ between the levodopa and the naltrexone condition ($$p \leq 0.368$$, $d = 0.11$). Endogenous pain facilitation induced by actively receiving pain increases assessed with VAS ratings did not significantly differ between any drug conditions (all post-hoc comparisons p’s>0.591).
**Table 1.**
| Unnamed: 0 | Pain modulation in VAS ratings of pain intensity | Pain modulation in VAS ratings of pain intensity.1 | Pain modulation in VAS ratings of pain intensity.2 | Pain modulation in VAS ratings of pain intensity.3 | Pain modulation in VAS ratings of pain intensity.4 | Pain modulation in VAS ratings of pain intensity.5 | Pain modulation in behavioral measure (°C) | Pain modulation in behavioral measure (°C).1 | Pain modulation in behavioral measure (°C).2 | Pain modulation in behavioral measure (°C).3 | Pain modulation in behavioral measure (°C).4 | Pain modulation in behavioral measure (°C).5 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | placebo | placebo | levodopa | levodopa | naltrexone | naltrexone | placebo | placebo | levodopa | levodopa | naltrexone | naltrexone |
| | n=28 | n=28 | n=27 | n=27 | n=28 | n=28 | n=28 | n=28 | n=27 | n=27 | n=28 | n=28 |
| Outcome | M | SD | M | SD | M | SD | M | SD | M | SD | M | SD |
| win | –7.31 | 21.51 | –12.98 | 23.54 | –10.09 | 23.79 | –0.09 | 0.64 | –0.14 | 0.66 | –0.05 | 0.74 |
| lose | 12.21 | 21.12 | 13.29 | 20.48 | 12.26 | 22.27 | 0.03 | 0.59 | 0.03 | 0.54 | 0.06 | 0.68 |
## Behaviorally assessed pain perception
In addition to the VAS ratings, participants performed a validated perceptual task (Becker et al., 2011; Kleinböhl et al., 1999) allowing to assess perception of the underlying tonic pain stimulus, which is specifically sensitive to perceptual sensitization and habituation. In this procedure, participants re-adjust the stimulation temperature themselves after the outcome of the wheel of fortune to match their perception at the beginning of each trial. Positive values (i.e. lower self-adjusted temperatures compared to the stimulation intensity at the beginning of the trial) indicate perceptual sensitization across the course of one trial of the game, negative values indicate habituation. For tonic stimulation at intensities that are perceived as painful, perceptual sensitization is expected to occur (Kleinböhl et al., 1999). Differences between the outcome conditions (win, lose) reflect the effect of the phasic changes on the perception of the underlying tonic stimulus. Differences between active and passive trials reflect the effect of controllability on the perception within each outcome condition. In contrast to the VAS ratings, behaviorally assessed pain perception did not significantly differ between test and control trials after winning as well as after losing in the placebo condition (Figure 2D; interaction ‘outcome × trial type’, F[1, 1040]=2.53, $$p \leq 0.112$$).
In contrast to the placebo condition, participants showed significantly less behaviorally assessed sensitization in active compared to passive trials when obtaining pain relief under levodopa (pairwise comparison test vs. control: $$p \leq 0.020$$, $d = 0.11$; Figure 2E) consistent with an extension of pain-inhibitory effects of winning pain relief through to the underlying tonic pain stimulus. Under naltrexone, test and control trials did not significantly differ in the behaviorally assessed pain perception (Figure 2F) as for the placebo condition. Across drugs, behaviorally assessed pain modulation did not significantly differ between placebo, levodopa, and naltrexone (interaction ‘drug × outcome’: F(2, 1592.73)=1. 87, $$p \leq 0.154$$; Figure 3B).
**Figure 3.:** *Effects of drug manipulation on endogenous pain modulation.Effects of drug manipulation on endogenous pain modulation assessed by VAS ratings of pain intensity (A) and behaviorally assessed pain perception (B) after winning and losing in the wheel of fortune game, respectively (placebo: n=28, levodopa: n=27, naltrexone: n=28). Bars show group level means and error bars show 95% confidence interval of the group level mean. d indicates the standardized effect-size after controlling for random effects and residual variance. While the temporal order of sessions did affect pain modulation (Figure 3—figure supplement 1), measures of pain sensitivity, that were not experimentally manipulated (Figure 3—figure supplement 2), and measures of mood (Figure 3—figure supplement 3) did not significantly differ between drug conditions. For individual effects of the drug manipulations on endogenous pain modulation see Figure 3—figure supplement 4.*
## Levodopa increases endogenous pain modulation by active relief with no significant effects of naltrexone on the modulation
We next examined whether endogenous modulation of pain perception within the wheel of fortune game was affected by a levodopa and naltrexone.
## Manipulation check: successful blinding of drug conditions
After the intake of levodopa, one participant reported a weak feeling of nausea and headaches at the end of the experimental session. In 32 out of 83 experimental sessions subjects reported tiredness at the end of the session. However, the frequency did not significantly differ between the three drug conditions (χ2 [2]=2.17, $$p \leq 0.337$$) or between the placebo condition compared to the levodopa and naltrexone condition (χ2 [1]=1.06, $$p \leq 0.304$$). No other side effects were reported. To ensure that participants were kept blinded throughout the testing, they were asked to report at the end of each testing session whether they thought they received levodopa, naltrexone, placebo, or did not know. In 43 out of 83 sessions that were included in the analysis ($52\%$), participants reported that they did not know which drug they received. In 12 out of 28 sessions ($43\%$), participants were correct in assuming that they had ingested the placebo, in 6 out of 27 sessions ($22\%$) levodopa, and in 2 out of 28 sessions ($7\%$) naltrexone. The amount of correct assumptions differed between the drug conditions (χ2 [2]=7.70, $$p \leq 0.021$$). However, post-hoc tests revealed that neither in the levodopa nor in the naltrexone condition participants guessed the correct pharmacological manipulation significantly above chance level (p’s>0.997) and the amount of correct assumptions did not differ significantly between placebo compared to levodopa and naltrexone sessions on the other hand (χ2 [1]=0.11, $$p \leq 0.737$$), suggesting that the blinding was successful.
## Effects of pharmacological manipulations on endogenous modulation by active controllability
We next examined whether endogenous modulation of pain perception within the wheel of fortune game was affected by a levodopa and naltrexone. In addition to the models testing effects of controllability on pain perception within each drug condition, we calculated pain modulation in test trials for both outcome measures as the difference of each test trial to the mean of control trials for the respective outcome condition for each participant. We fitted linear mixed effects models to predict pain modulation by drug condition, outcome, and their interaction. As an additional covariate of no interest we included session order (and respective interaction effects) in these models, as the temporal order of sessions (independent of the order of the application of the drugs) was found to have a differential effect on win and lose outcomes (see Figure 3—figure supplement 1).
## Levodopa and naltrexone influence relief reinforcement learning in the wheel of fortune task
To investigate whether pain relief gained in active relief seeking was associated with an impact on choice related to reinforcement learning, one of the 2 choices in the wheel of fortune was associated with a fixed $75\%$ chance of winning pain relief (choicehigh prob) while the other choice only had a $25\%$ chance to win pain relief (choicelow prob). Participants were not informed of these probabilities in advance. We tested if the proportion of choices of the more rewarding option was higher in the last two out of five blocks of four test trials each of the game, when the subjects already had the chance to explore and learn the different outcome probabilities.
Participants selected the color of the wheel of fortune associated with a higher likelihood for winning relief in $64\%$ (SD = $28\%$) of trials in the placebo condition, consistent with a reinforcement learning effect. Thus, participants chose the color associated with the higher likelihood for winning above chance (χ2[1]=6.64, $$p \leq 0.010$$) on a group level, indicating successful learning.
However, participants’ performance significantly differed between the placebo and the drug conditions (main effect of ‘drug’: χ22 = 11.89, $$p \leq 0.003$$). In contrast to the placebo condition (post-hoc comparison $p \leq 0.001$), under levodopa and under naltrexone participants’ choices did not significantly differ from chance (post-hoc comparisons p’s>0.759). Correspondingly, post-hoc comparisons show that choice behavior significantly differed in the placebo compared to the levodopa condition ($$p \leq 0.015$$) and compared to the naltrexone condition (post-hoc comparison $$p \leq 0.004$$), while choices did not significantly differ between levodopa and naltrexone (post-hoc comparison $$p \leq 0.915$$; Figure 4). This shows that both, dopamine and opioids, may have an influence on relief-related learning and choice.
**Figure 4.:** *Proportion of choices of the color associated with a higher chance of winning pain relief.Bars show group level means and error bars show 95% confidence interval of the group level mean (placebo: n=28, levodopa: n=27, naltrexone: n=28). OR indicates odds ratios as effect size of estimated effects between drugs. *p<0.05, ** p<0.01.*
In an additional exit interview at the end of each session, participants were asked whether they believed that one color of the wheel was associated with a higher chance of winning pain relief. The proportion of participants who reported this color correctly was not above chance (binomial test: p’s>0.5; placebo: $50\%$, levodopa: $37\%$, naltrexone: $39.3\%$). Nevertheless, participants’ belief whether one color of the wheel of fortune task was associated with a higher chance of winning or not significantly influenced their choices ($p \leq 0.001$) and this influence on choices, and thus on learning, depended on the drug condition (interaction ‘drug × belief’: F[2] = 6.91, $$p \leq 0.032$$). Group effects of successful learning, i.e. selecting the color with a higher chance of winning, were driven by participants who were able to report this association (p(choicehigh prob|correct belief) = 0.737, p(choicehigh prob|false or no belief) = 0.545; post-hoc comparison: $$p \leq 0.007$$) under placebo and naltrexone (p’s<0.001) but not under levodopa ($$p \leq 0.922$$). This suggests that successful decision-making was associated with contingency awareness in our task. However, the current data does not allow a conclusion on whether this contingency awareness was a prerequisite for or a consequence of successful learning here.
## Unpredictability and endogenous pain modulation
We next tested whether outcome unpredictability, indicated by reward prediction errors of winning (pain relief) and losing (pain increase) in the game, was associated with endogenous pain modulation, and whether this association differed between drugs. Using hierarchical Bayesian modeling, we fit reinforcement learning models that captured the update of expected values for choice through outcomes of the wheel of fortune (Glimcher, 2011), with a drift diffusion process as the choice rule to participants’ choice and reaction time data. The best predictive accuracy was found for model 4 that used an individually scaled outcome sensitivity, and a sigmoid function to map expected values for the two choices to the drift rate of the diffusion process (Table 2; see Materials nd methods, section Estimation of prediction errors and their role in endogenous pain modulation for details on parametrization of reward learning models).
**Table 2.**
| Model | ELPD | ELPDdiff | se(ELPDdiff) |
| --- | --- | --- | --- |
| Model 4 | –837.71 | 0 | 0.0 |
| Model 3 | –845.44 | –7.73 | 1.51 |
| Model 2 | –997.33 | –159.62 | 15.77 |
| Model 1 | –998.33 | –160.62 | 15.95 |
Posterior predictive simulations from the best-fitting model appropriately describe the observed choices (Figure 5). However, none of the model parameters could exclusively explain the differences between levodopa and naltrexone compared to placebo: the $95\%$ highest density intervals (HDI) for the difference between all group level parameters of the drug effect enclosed zero (see Figure 5—figure supplement 1).
**Figure 5.:** *Posterior distributions of the proportion of choices in favor of choicehigh
prob.Colored areas show 95% highest density interval (HDI95). Dashed lines indicate observed proportion of choices in favor of choicehigh
prob . Placebo (n=28): pchoicehigh
prob = 0.641, HDI95 = [0.614,0.655], posterior (p-value (pp)=0.320); levodopa (n=27): pchoicehigh
prob = 0.507, HDI95 = [0.491,0.530], p=0.679; naltrexone (n=28): pchoicehigh
prob = 0.467, HDI95 = [0.443,0.494], p=0.611. Figure 5—figure supplement 1 shows comparison of drug conditions for each parameter of winning model 4.*
Prediction errors estimated by using subject level parameters of the model showed a significant main effect for the prediction of endogenous pain modulation indicated by VAS ratings (F(1, 1600.3)=452.9, $p \leq 0.001$). A negative estimate of the prediction error (βPE = –0.36) indicates that outcomes that are better than expected (positive prediction errors, which occur when receiving relief) were related to increased relief perception (pain inhibition). Conversely outcomes that are worse than expected (negative prediction errors, occurring with pain increases) were associated with increased pain facilitation (Figure 6). In other words, the more unexpected the relief, the greater the perception of that relief; and the more unexpected the pain increase, the greater the perception of that pain.
**Figure 6.:** *Pain modulation in VAS ratings predicted by prediction errors for each condition.Regression lines indicate prediction from the mixed effects model with predictors ‘PE’, ‘drug’, and their interaction (placebo: n=28, levodopa: n=27, naltrexone: n=28).*
The effect of prediction errors on pain modulation showed a significant interaction with the drug condition (F(2, 1599.5)=7.529, $p \leq 0.001$). Post-hoc analysis confirmed that the negative linear relationship significantly differed from zero for all conditions (p’s<0.001), but this relationship was significantly stronger for levodopa compared to placebo ($p \leq 0.001$) with no significant differences for naltrexone compared to placebo ($$p \leq 0.083$$). Overall, this shows that relief is enhanced to unpredictability, and this effect is sensitive to dopamine.
Estimated prediction errors also showed a significant main effect for the prediction of behaviorally assessed pain modulation (F(1, 1602.1)=9.00, $$p \leq 0.003$$), with a negative estimate (βPE = –0.06) suggesting that sensitization decreased with smaller prediction errors. No significant interaction with of prediction error with drug conditions was found for behaviorally assessed pain perception (interaction ‘PE × drug’: F(2, 1600.1)=0.96, $$p \leq 0.384$$).
## Novelty seeking is linearly associated with increased endogenous pain modulation by pain relief under levodopa
Previous data suggest that endogenous pain inhibition induced by actively winning pain relief is associated with a novelty seeking personality trait: greater individual novelty seeking is associated with greater relief perception (pain inhibition) induced by winning pain relief (Becker et al., 2015). Similar to these results, we found here that endogenous pain modulation, assessed using self-reported pain intensity, induced by winning was associated with participants’ scores on novelty seeking in the NISS questionnaire (Need Inventory of Sensation Seeking; Roth and Hammelstein, 2012; subscale ‘need for stimulation’ (NS)), although this correlation failed to reach statistical significance after correction for multiple comparisons using Bonferroni-Holm method (r=–0.412, $$p \leq 0.073$$). A similar association between novelty seeking and endogenous pain modulation was found in the levodopa condition (r=–0.551, $$p \leq 0.013$$). More importantly, the higher a participants’ novelty seeking score in the NISS questionnaire, the greater the levodopa-related endogenous pain modulation when winning compared to placebo (NISS NS: r=–0.483, $$p \leq 0.034$$, Figure 7). In contrast, higher novelty seeking scores were not correlated with stronger pain modulation induced by winning in the naltrexone condition ($r = 0.153$, $$p \leq 0.381$$) and the naltrexone induced change in pain modulation showed no significant association with novelty seeking ($r = 0.239$, $$p \leq 0.499$$). Pain modulation after losing was not associated with novelty seeking in placebo ($r = 0.083$, $$p \leq 0.866$$), levodopa (r=–0.164, $$p \leq 0.783$$), or naltrexone ($r = 0.405$, $$p \leq 0.133$$).
**Figure 7.:** *Correlation of changes in endogenous pain modulation induced by winning pain relief under levodopa compared to placebo with individuals’ scores on the ‘need for stimulation’ subscale of the NISS questionnaire, n=24.*
No significant correlations with NISS novelty seeking score were found for behaviorally assessed pain modulation in the placebo, levodopa and naltrexone conditions during pain relief or pain increase (|r|’s<0.35, p’s>0.238). Similarly, the difference in pain modulation during pain relief or pain increase between the levodopa and the placebo condition and between the naltrexone and the placebo condition did also not correlate with novelty seeking (|r|’s<0.22, p’s>0.576).
## Discussion
The results show that (i) the perception of relief is sensitive to endogenous modulation during motivated behavior, (ii) this modulation scales with the informational content of the relief, being enhanced when relief is actively controllable, more unexpected, and especially in high trait novelty seeking individuals, (iii) this information-specific modulation is sensitive to manipulation of dopamine signaling, with no significant effects of the manipulation of opioidergic signaling on endogenous modulation; (iv) however, both dopaminergic and opioidergic signaling have an influence on relief-seeking, which may be at least in part dissociable from relief perception. Overall, this shows that dopaminergic signaling is involved in a fundamental component of the endogenous modulation of pain relief.
Theories of the endogenous modulation of pain propose that one of the reasons that pain is modulated is to optimize motivational behavior, in terms of responding, learning, and decision-making (Fields, 2018; Seymour, 2019). That is, pain is increased in situations in which it has a more important role in shaping behavior – for instance when it directs a change in behavior (instrumentally controllable), when it is partly unpredictable (i.e. contains new information), and in otherwise dangerous contexts. This theory centralizes the functional role of pain as a signal for behavioral control that is concerned with the prospective control of behavior. In principle, this can be extended as a potential account for the modulation of relief, because the offset of pain is also important as a control signal for guiding behavior, one which occurs in the context of an ongoing noxious event, such as an injury of some sort (Zhang et al., 2018). We have previously found preliminary evidence of this, by showing that relief perception is enhanced by active controllability (Becker et al., 2015). Here, we intended to test this more precisely, by looking at the role of controllability as well as unpredictability, and also compare to the modulation of phasic increases in tonic pain.
We also set an additional prediction, in that we expected to find that modulation by information content would be greater in novelty-seeking individuals (Becker et al., 2015). This is because novelty seeking describes an explicit information-seeking tendency, in which new information is explored with the potential to lead to knowledge of better outcomes that can be exploited in the future (Wittmann et al., 2008). This illustrates the common basis for intrinsic motivation for novelty and information-seeking for exploitable benefit, and hence we can predict that high trait novelty seekers might be more sensitive to information that occurs through relief outcomes.
Overall, all three predictions were largely borne out by the data: relief perception as measured by VAS ratings was enhanced by controllability, unpredictability and showed a medium sized – although not significant – association with the individual novelty-seeking tendency, consistent with the hypothesis that relief is sensitive to the exploitable information it carries. This provides the first clear formal framework for understanding a key component of relief perception. The principles for controllability and unpredictability also extended to increases in pain, consistent with the notion that increases in tonic pain act in a similar way to phasic pain operating from a pain-free baseline.
## Effects of pharmacological manipulations on endogenous perceptual modulation
Both dopamine and opioids are implicated in relief processing, although their precise roles remain unclear. We found endogenous relief modulation here was modulated by enhanced dopamine availability induced by the intake of levodopa. Importantly, all three core aspects of informational-sensitivity were modulated by levodopa: active controllability, unpredictability, and association with novelty seeking. These findings also illustrate potential parallels with the previous observation of endogenous pain inhibition by extrinsic monetary reward co-occurring with experimental pain (Becker et al., 2013). In this context, monetary reward represents an independent and potentially competing incentive, and when this co-occurs with pain, it means that optimal responding may require suppression of pain responses, especially innate responses that could interfere with reward acquisition. In both cases, the common principle may be the active ‘decision’ by the pain system to tune incoming pain signals to optimize behavior. Levodopa significantly strengthened the association of endogenous modulation of the perception with the extent to which nociceptive input was unpredictable or surprising, suggesting that dopamine is involved in a fine-grained perceptual modulation that may help to control immediate reactions to pain related cues as well as to optimize prospective behavior. Moreover, our results support the view that effects of dopamine on pain perception are not unidirectional but depend on the motivational and the informational value of nociceptive signals. Note that the personality trait of novelty seeking has also been associated with enhanced dopaminergic activity due to lower midbrain (auto)receptor availability (Leyton et al., 2002; Savage et al., 2014; Zald et al., 2008), which further supports a general role for dopamine in information-sensitive behavior (Kakade and Dayan, 2002; Vellani et al., 2020).
The role of dopamine in pain relief in the context of reinforcement is supported by findings of increased dopamine release induced by pain relief in the Nucleus accumbens of rats (Navratilova et al., 2012; Xie et al., 2014). Dopamine release was related to the development of conditioned place preference that could be blocked by dopamine antagonists (Navratilova et al., 2012). Further, Navratilova et al., 2015b showed that dopamine release in the Nucleus accumbens and conditioned place preference in response to pain relief depend on opioidergic signaling: both were blocked by opioid antagonism in the anterior cingulate cortex, an area encoding pain aversiveness.
In contrast to our hypothesis, pharmacologically blocking opioid receptors using naltrexone did not reduce modulation endogenous pain inhibition in this task. The doses and methods used here are comparable to those used in other contexts which have identified opioidergic effects. Using positron emission tomography Weerts et al., 2013 found a blockage of μ-opioid receptors of more than $90\%$ by 50 mg of naltrexone (p.o.) in humans given repeatedly over 4 days. In addition, effects on behavioral functions have been reported with comparable doses that support the efficacy of the opioidergic manipulation. Chelnokova et al., 2014 found attenuating effects of 50 mg naltrexone (p.o.) on wanting as well as liking of social rewards, implicating the involvement of endogenous opioids in the processing of rewarding stimuli. The same dose was also found to attenuate reward directed effort exerted in a value-based decision-making task (Eikemo et al., 2017). Moreover, 50 mg of naltrexone (p.o.) have been shown to reduce endogenous pain inhibition induced by conditioned pain modulation (King et al., 2013). Thus, based on the literature we assume that the opioidergic manipulation was effective in this study, although we do not have a direct manipulation check of this pharmacological manipulation. Despite its effectiveness in blocking endogenous opioid receptors, the effect of naltrexone on reward responses was found to be small (Rabiner et al., 2011). Hence, a lack of power may have limited our chances to find such effects in the present study.
In humans, Sirucek et al., 2021 showed that perception of passively received pain relief was reduced by blocking of opioid receptors. However, in that task, received pain relief did not carry behaviorally relevant information, in contrast to the present task. Further, Sirucek et al., 2021 asked participants to rate the perceived pain relief and its pleasantness, while in the current study participants were asked to rate perceived intensity of the stimulation at certain time points in each trial. Perceived pain relief was therefore estimated indirectly by differences in these ratings. Increased opioid activity in the anterior cingulate cortex has been shown to be associated with selectively decreased pain aversiveness with unaltered sensory pain components (Gomtsian et al., 2019; Maruyama et al., 2018; Navratilova et al., 2015b). In contrast, the present study aimed at quantifying the effect of controllability on the relief perception, with these methods possibly not capturing the effects of opioid blockade on positive affective quality components of the relief experience. Future studies need to address the question whether the affective dimension of enhanced relief perception by rewarding pain relief may be reduced by opioid blockade with the sensory component possibly being unaffected.
Although we did not assess the affective component of the relief experience, we implemented two outcome measures that are assumed to capture independent aspects of the pain experience: VAS ratings indicate perception of phasic changes (outcomes), while the behavioral measure indicates perceptual within-trial sensitization or habituation in response to the tonic stimulation within each trial. We found enhanced endogenous modulation by controllability and unpredictability in VAS ratings, in line with the view that endogenous modulation enhances behaviorally relevant information. In contrast, the within-trial sensitization did not differ between the active and passive conditions under placebo. In a previous study using a similar experimental paradigm Becker et al., 2015 found a reduction of within-trial sensitization after pain relief outcomes by active controllability. Compared to this study we implemented smaller changes in stimulation intensity as outcomes in the wheel of fortune (–3 °C vs –7 °C for pain relief), potentially explaining the differential results.
## Effects of pharmacological manipulations on relief-seeking behavior
One key difference in the current version of the wheel of fortune task, compared to the previous version described in Becker et al., 2015, is that participants’ choices had a non-random association with outcomes that is this was a true instrumental (operant) contingency between actions and outcomes. This allowed us to assess a basic measure of learning – whether subjects are able to learn to select more frequently the option with the better ($75\%$ chance of relief) over worse ($25\%$ chance of relief) outcome. That both levodopa and naltrexone conditions were associated with a reduction of the frequency of choosing the better option, indicates that signals mediated by both neurotransmitters may be involved in choice. However, the data argue against a simple transposition of experienced relief (measured by VAS) into decision value, which for a stationary task such as this, should lead to more deterministic actions in the levodopa condition but no effect under naltrexone compared to placebo. The association of explicit contingency awareness and choice in our task illustrates the fact that multiple decision systems (‘model-based’ and ‘model-free’) might be involved in even simple instrumental tasks, and hence that more sophisticated task manipulations are needed to decompose these different components (Langdon et al., 2018). However, our key finding is that there is at least a simple dissociation between the drug effects on experienced relief and decision-making.
Such dissociation may be due to differential involvement of dopamine and endogenous opioids in different yet interacting aspects of reward and punishment processing. Dopamine has been related to instrumental learning due to its prominent role in mediating reward and aversive prediction errors (Glimcher, 2011; Matsumoto and Hikosaka, 2009; Schultz, 2007; Schultz, 2016). Correspondingly, effects of dopaminergic modulation on value-based decision making and brain activity related to reward prediction errors in the Nucleus accumbens have been reported (Pessiglione et al., 2006). On the other hand, impaired learning functions under dopaminergic medication are known from research in Parkinson’s disease (Breitenstein et al., 2006; Pizzagalli et al., 2008; Santesso et al., 2009; Vo et al., 2016) and have been attributed to dopamine overstimulation (Cools et al., 2001; Vaillancourt et al., 2013). Others argued that dopamine overstimulation does not impair learning of associations or reward expectations, but only the transfer to overt actions (choice behavior; Beeler, 2012; Beeler et al., 2010). Accordingly, Kroemer et al., 2019 found reduced model-free control of choice behavior under levodopa (i.e. a decrease in direct reinforcement of actions by rewards) while both, neural reward prediction error signals and also model-based learning remained unaffected. Given the involvement of multiple decision systems in our task increased dopamine availability might have led to increased explorative behavior. When exploration is favored over exploitation choice behavior is less driven by values learned from the prior reward history. At the same time, dopamine has also been implicated in motivational aspects (incentive salience) of reward processing (Berridge et al., 2009; Smith et al., 2011; Tindell et al., 2005). Hence, dopamine may have increased motivational drive and related facilitation of pain modulation in the present task, while at the same time increased dopamine availability may have reduced the expression of prior reward learning (Beeler, 2012). Opioids have been related to both, incentive salience and the hedonic value of rewards (Berridge et al., 2009; Meier et al., 2021). In humans, bidirectional manipulations have shown that opioid agonism increases while opioid antagonism decreases “wanting” (i.e. incentive salience) as well as “liking” of attractive faces (Chelnokova et al., 2014). The same mechanism was also shown for the effort to work for and the response bias for higher monetary rewards indicating that opioid manipulations affect motivation but also choice behavior (Eikemo et al., 2017). These findings suggest that inhibition of opioidergic activity by blocking endogenous opioid receptors could impair reward processing (independent of effects of endogenous opioids on pain modulation), and hence, explain why participants did not develop a preference for the choice option that was associated with a higher chance to win pain relief under naltrexone in the present task.
## Implication and perspectives
Because the mechanisms underlying learning from pain and pain relief and their recursive influence on pain perception may contribute to the development and maintenance of chronic pain, it is crucial to better understand the roles of dopamine and endogenous opioids in these mechanisms. Accordingly, bidirectional manipulations of both transmitter systems should be used in future studies to better characterize their respective roles in shaping behavior and perception.
The data may have clinical implications. Reward learning has recently been shown to play a role in the transition of acute to chronic pain with a specific pattern of Nucleus accumbens activity in response to a cue predicting pain relief being predictive for chronification (Löffler et al., 2022). This makes pain relief processing a potential leverage point for prevention strategies. Although levodopa or dopamine agonists are not generally used as analgesics in the clinical management of chronic pain, it may be that they could have a potential adjuvant role in management programs, for example when used in the context of rehabilitation strategies that aim to harness endogenous control mechanisms. It is also worth noting that Parkinson’s disease has a well-recognized association with chronic pain, beyond that which can be explained by motor effects, and in keeping with a potential core role for dopamine in the pathogenesis of chronic pain in some contexts (Beiske et al., 2009).
In summary, our study shows that dopamine has a core role in pain relief information processing, by which it modulates the way in which information tunes the modulation of pain to meet motivational demands.
## Participants
Thirty healthy volunteers (16 female, 14 male; age: mean = 27.1 years; SD = 7.9 years) participated in this study. Exclusion criteria were present pain or pain conditions in the last 12 months, mental disorders, excessive gambling, substance abuse behaviors, alcohol consumption of 100 ml or more of alcohol per week, regular night shifts, or sleep disorders. The power estimation was based on the design and the finding of a medium effect size in a previous study using a comparable version of the wheel of fortune game without pharmacological interventions (Becker et al., 2015). The a priori sample size calculation for an $80\%$ chance to detect such an effect at a significance level of yielded a sample size of 28 participants (estimation performed using GPower [Faul et al., 2007; version 3.1] for a repeated-measures ANOVA with a three-level within-subject factor). The study was approved by the Ethics Committee of the Medical Faculty Mannheim, Heidelberg University, and written informed consent was obtained from all participants prior to participation according to the revised Declaration of Helsinki (World Medical Association, 2013).
## Testing sessions
Each participant performed three testing sessions on separate days. Each session comprised a pharmacological intervention and a wheel of fortune game to assess modulation of reward-induced endogenous pain modulation by the interventions. Participants received in one session levodopa to transiently increase the availability of dopamine, in one session the opioid receptor antagonist naltrexone to block opioid receptors, and in one session a placebo for control. To ensure complete washout of the drugs, the testing sessions were separated by at least 2 days (plasma half-life for levodopa: 1.4 hrs Nyholm et al., 2012; plasma half-life for naltrexone: 8 hrs [Wall et al., 1981]). After obtaining written consent in the first testing session, participants were familiarized with the thermal stimuli, the rating scale, and the wheel of fortune game to decrease unspecific effects of novelty and saliency. In each testing session, the thresholding and scaling procedures for individual adjustments of the stimulation intensities started approximately 60 min after drug intake and were performed immediately prior to playing the wheel of fortune game.
## Thermal stimulation
All heat stimuli were applied using a 25x50 mm contact thermode (SENSELab—MSA Thermotest, SOMEDIC Sales AB, Sweden). The baseline temperature was set to 30 °C. Rise and fall rates of the temperature were set to 5 °C/s. All thermal stimuli were applied to the inner forearm of participants’ non-dominant hand after sensitization of the skin using $0.075\%$ topical capsaicin cream to allow for potent pain relief as reward and pain increase as punishment without the risk of skin damage (Becker et al., 2015; Gandhi et al., 2013). By activating temperature-dependent TRPV1 (vanilloid transient receptor potential 1) ion channels capsaicin as the active ingredient of chili pepper induces heat sensitization (Holzer, 1991). To ensure that the entire area of thermal stimulation during the wheel of fortune game was sensitized the cream was applied to an area on the forearm exceeding the area of stimulation by about 1 cm on each side. After 20 min, the capsaicin cream was removed (Dirks et al., 2003; Gandhi et al., 2013) and the thermode was applied. If participants reported the baseline temperature of the thermode (30 °C) as painful because of the preceding sensitization this temperature was lowered until it was perceived as non-painful, which was needed in 8 out of 83 sessions (3 placebo sessions, 1 levodopa session, and 4 naltrexone sessions) that were finally entered into the analysis (see below). The temperature was decreased to 28 °C (1 placebo session, 4 naltrexone sessions) or 26 °C (1 placebo session, 1 levodopa session). The need to lower the baseline temperature was not significantly different between drug conditions (Fisher’s exact test, $$P \leq 0.52$$).
## Determination of stimulation intensities
Participants’ heat pain threshold and heat pain tolerance were assessed using the method of limits three times prior to the wheel of fortune game. The temperature of the thermode increased from baseline with 1 °C/s. Participants were instructed to press the left button of a three-button computer mouse when the pain threshold was reached. The respective temperature was recorded while the temperature further increased. Participants were instructed to press the button again when the pain tolerance threshold was reached. The respective temperature was recorded and the temperature immediately returned to baseline. The arithmetic mean of the temperatures corresponding to the recorded pain threshold and tolerance in the three trials was used as an estimate of the individual heat pain threshold and heat pain tolerance, respectively.
After this threshold and tolerance assessment, an adjustment procedure resembling a staircase method was implemented to determine the stimulation intensities in the wheel of fortune game. Participants received heat stimuli of 20 s duration and continuously rated the perceived intensity of these stimuli on a computerized visual analogue scale (VAS) ranging from ‘no sensation’ [0] over ‘just painful’ [100] to ‘most intense pain tolerable’ [200] (Becker et al., 2013; Villemure et al., 2003) while the stimuli where presented. The VAS scale was presented on the screen with a red marker that could be moved along the scale. Participants adjusted the position of the marker by pressing the right or left mouse button. The marker moved into the respective direction until participants released the button. The participants were instructed to adjust their rating whenever a change in their perception occurred. Participants could adjust their ratings as long as the VAS scale was presented on the screen. The temperature of the first trial was set to the mean of the previously determined pain threshold and tolerance. If the rating at the end of the stimulation was outside a range of 150±10 on the VAS, the temperature for the next trial was adjusted according to the difference to a target rating of 150. This adjustment was determined by multiplying the difference (150 – current rating) by 0.02 and adding the result in °C to the previous temperature. Further, temperature increases between trials were limited to a maximum of 0.5 °C to avoid overshooting of ratings. The procedure was repeated until a rating between 140 and 160 on the VAS was achieved, aiming at a temperature perceived as moderately painful. The corresponding temperature was used as the stimulation intensity in the wheel of fortune game.
## Wheel of fortune game
A wheel of fortune game, adapted from a previously established version (Becker et al., 2015), was used to provide participants with the possibility of winning pain relief. The game comprised three types of trials: test trials, in which participants played the wheel of fortune game and received pain relief or pain increases according to the outcome of the game; control trials, in which participants did not play the game, but received pain relief or pain increases as in the test trials; and neutral trials, in which participant did not play the game and no pain relief or pain increases were implemented. A trial always started with an increase of the temperature to the previously determined tonic pain stimulation intensity. When the stimulation intensity was reached, participants were instructed to memorize the temperature perceived at this moment (Figure 1). After this memorization interval, participants were presented with a wheel of fortune display that was divided into three sections of equal size but different color.
In the test trials, participants were asked to select one of two colors (pink or blue) of the wheel by pressing a corresponding button (left or right) on the mouse. This started the wheel spinning (4.3 s) until it stopped on either the blue or pink section. When the wheel came to a stop and the pointer of the wheel indicated the color the participant had chosen, the stimulation temperature decreased with the aim to induce pain relief (win condition). If the pointer indicated the color the participant had not chosen, the temperature was increased (lose condition). In the control trials, participants had to press a black button unrelated to the sections of the wheel of fortune using the middle mouse button, after which the wheel started spinning as in the test trials. In contrast to the test trials, no pointer was displayed in the control trials and the wheel stopped at a random position. After the wheel came to a stop, the stimulation temperature decreased or increased, resembling the course of stimulation in the test trials, but without winning or losing. By this procedure, nociceptive input in test and control trials was kept the same, allowing to test specifically for endogenous pain modulation induced by winning and losing in the wheel of fortune game.
In neutral trials, participants had to press a black button, as in the control trials, after which the wheel also started spinning. In these neutral trials, the pointer of the wheel always landed the third color of the wheel (white), which could not be selected in test trials, and the stimulation temperature did not change. Neutral trials were used to estimate changes in pain perception occurring over the course of the experiment due to habituation or sensitization independent of the outcomes of the wheel of fortune game.
The participants were instructed that there were two types of trials: trials in which they could choose a color to bet on the outcome of the wheel of fortune and trials in which they had no choice. Specifically, they were told that in the first type of trials they could use the left and right mouse button, respectively, to choose between the pink and blue section of the wheel of fortune. Participants were further instructed that if the wheel lands on the color they had chosen they will win, that is that the stimulation temperature will decrease, while if the wheel lands on the other color, they will lose, that is that the stimulation temperature will increase. For the second type of trials, participants were instructed that they could not choose a color, but were to press a black button, and that after the wheel stopped spinning the temperature would by chance either increase, decrease, or remain constant.
After the interval of the temperature change (in the test trials: outcome of the wheel), participants rated the perceived intensity of the current temperature using the same VAS as described above (Figure 1). After this rating, participants had to adjust the stimulation temperature themselves to match the temperature they had memorized at the beginning of the trial. This self-adjustment operationalizes a behavioral assessment of perceptual sensitization and habituation within one trial (Becker et al., 2011; Becker et al., 2015; Kleinböhl et al., 1999). Participants adjusted the temperature using the left and right button of the mouse to increase and decrease the stimulation temperature. The behavioral measure was calculated as the difference in temperatures in the memorization interval at the beginning of each trial minus this self-adjusted temperature at the end of each the trial. Positive values, that is self-adjusted temperatures lower than the stimulation intensity at the beginning of each trial, indicate perceptual sensitization, while negative values indicate habituation. After this behavioral assessment, the stimulation temperature went back to baseline and after a short break (5 s) the next trial started.
In total, the wheel of fortune game comprised of 45 trials, split into five blocks. Each block consisted of 4 test and 4 control trials followed by one neutral trial, resulting in 20 test trials in which participants chose a color, 20 control trials, and 5 neutral trials. Test and control trials were presented in a predefined, pseudorandomized sequence. In contrast to the previous version of the wheel of fortune (Becker et al., 2015), the outcome of the wheel occurred with certain likelihood to allow for learning to optimize the outcomes of the wheel of fortune. One of the colors (pink or blue) was associated with a $75\%$ chance of winning, while the other was associated with a $25\%$ chance of winning (counterbalanced across participants and testing sessions). If participants did not select a color in the test trials, the neutral outcome (white) of the wheel was displayed and the temperature did not change. The temperature changes in the control trials (pain relief or increase) were matched to the outcomes of the test trials to ensure that the same number of pain relief and pain increase trials were presented in test and control trials.
Pain relief was implemented by a reduction of the stimulation intensity of 3 °C and pain increase was implemented by a rise of 1 °C. The magnitude of these temperature steps was determined and optimized in pilot experiments with the aim of inducing potent pain relief and pain increase without inducing ceiling and floor effects.
Although the main focus of the study was to test different effects on pain relief as implemented in win trials and their corresponding control trials with a decrease in nociceptive input, lose trials and their complementing control trials were crucial to the experimental design. First, for playing the game lose trials were an integral part because of the implemented likelihood for winning which necessarily needs to be accompanied by the chance of losing. Additionally, the risk of losing was thought to increase the participants’ engagement in the game, which in turn was expected to enhance the motivated state induced by playing the wheel of fortune game. Second, they allowed for testing whether pain modulation was driven by controllability or unspecific effects such as arousal and distraction in test compared to control trials of the wheel of fortune game (Becker et al., 2015).
All experimental procedures involving thermal stimulation were controlled by custom-programmed Presentation scripts (Presentation software, Version 17.0, http://www.neurobs.com) providing instructions and other visual cues on a computer screen in front of the participants (code available at https://osf.io/5xjt9).
## Pharmacological manipulations
Participants ingested in one testing session levodopa, in another naltrexone, and in another a placebo (microcrystalline cellulose), following a double-blind, cross-over design with counterbalanced order. Levodopa is an amino acid precursor of dopamine leading to a transient systemic increase of dopamine availability. To inhibit peripheral synthesis of dopamine from levodopa, the single dose of 150 mg levodopa (p.o.) was combined with 62.5 mg of a benserazide to prevent peripheral side effects such as nausea (Rinne et al., 1975). Naltrexone is an opioid receptor antagonist with predominant receptor binding affinity at µ-opioid receptors together with a lower binding affinity at κ-opioid receptors and a much lower affinity at δ-opioid receptors (Raynor et al., 1994). Participants received a single dose of 50 mg naltrexone (p.o.) that has been shown to induce more than $90\%$ receptor blockade (Weerts et al., 2013).
After drug intake, a waiting period of one hour started. This waiting time was chosen based on peak plasma concentrations of levodopa and naltrexone at approximately 1 hr to 1.5 hr after ingestion (Nyholm et al., 2012; Wall et al., 1981). At the end of each testing session, participants indicated whether they thought that they had received the placebo or one of the drugs (response alternatives: ‘placebo’, ‘levodopa’, ‘naltrexone’, or ‘don’t know’) to test for potential unblinding.
## Questionnaire and exit interview
Novelty seeking as personality trait was assessed using the Need Inventory of Sensation Seeking (NISS; Roth and Hammelstein, 2012). The NISS consists of the sub-scales Need for Stimulation (NS) and Avoidance of Rest (AR). We used the NS subscale as a measure for novelty seeking as it reflects the ‘need for novelty and intensity’ (Roth and Hammelstein, 2012). Before playing the wheel of fortune game the affective state of subjects was assessed using computerized versions of the Self-Assessment Manikin (SAM; Bradley and Lang, 1994; Lang, 1980) and a German version (Krohne et al., 1996) of the Positive And Negative Affect Scale (PANAS; Watson et al., 1988). At the of each session, an exit interview was performed, asking for the following information: [1] which drug participants believed to have ingested; [2] if participants believed that choosing one of the two colors was associated with a higher chance to win pain relief; [3] whether participants perceived a difference between test and control trials; [4] whether participants had the impression that the stimulation temperature at the beginning of each trial varied across trials; [5] whether participants had problems indicating their perception on the VAS scale; and [6] whether participants had problems readjusting the initial temperature. Participants gave first yes/no answers and then were asked to specify their answers using open-ended questions.
## Statistical analysis
For the statistical analysis, two participants were excluded, one participant due to the failure to comply with experimental procedures and one due to technical failure of the equipment. For one additional participant, data of one session (levodopa) are missing due to drop-out. Thirty-two out of 3735 single trials of all the remaining sessions were not recorded due to technical failures. In 42 trials, participants did not press a button within the respective interval in the wheel of fortune game. These trials were excluded from the analyses. Note that the NISS questionnaire was missing for two additional subjects due to initial issues at the beginning of the data collection.
To test if blinding was successful we fit a mixed effects logistic regression with the subjects’ assumption on the ingested drug (as reported in the exit interview, see above) being correct as dependent variable. We used ‘drug’ as a fixed factor and to account for repeated measures we modeled a random intercept for each subject. Post-hoc general linear hypothesis tests were used to compare estimated proportions of correct assumptions against chance.
To confirm that the manipulation of the motivated state (test vs. control trials) of the participants by playing the wheel of fortune game did induce pain modulation as intended in each session, we analyzed the VAS ratings and the behavioral pain measure as outcome measures separately for each session with ‘trial type’ and ‘outcome’ as well as their interaction as fixed effects. Here, the interaction effect indicates that active controllability in test trials has different effects for pain relief and pain increase. We used post-hoc tests to confirm that pain intensity was lower in the test compared to the control condition in win trials and equal or higher in the test compared to control condition in lose trials. We used Bonferroni-Holm method for family-wise error correction of these post-hoc tests across drug conditions. To account for the repeated measures design, we modeled a random intercept for each participant and a random slope for outcome of the wheel within each participant.
To obtain an estimate for endogenous pain modulation in each test trial, we subtracted the mean value of all control trials of either the pain relief or the pain increase trials from the value of the winning or losing test trials separately for each session for both the VAS ratings and the behavioral pain measure. Using these differences, negative values indicate pain inhibition and positive values indicate pain facilitation. Estimates for pain modulation were analyzed using linear mixed model procedures with the fixed factors ‘drug’ (levodopa, naltrexone, placebo), ‘outcome’ (win, lose), ‘order’ of sessions [1, 2, 3], and their interaction separately for ratings and behaviorally assessed pain perception as dependent variables. The factor ‘session number’ was added to control for effects of temporal order independent of the drug manipulation that was found to influence pain modulation (see Figure 3—figure supplement 1). Other factors such as baseline pain perception or mood did not affect pain modulation and were not included in further analysis. To account for the repeated measures design we modeled a random intercept for each participant and a random slope for outcome within each participant.
Unbeknown to the subjects, one of the colors in the wheel of fortune was associated with a higher chance to win pain relief. To test whether participants learned to select this color from the implemented reward contingencies we looked at choice behavior in the last 2 blocks of trials only. In this latter phase of the task subjects already had the chance to explore differences in outcomes associated with their choices and were thought show exploitation if they had learned about the contingency. We fitted a mixed effects logistic regression with the subjects’ choices as dependent variable. For a single session, we fit an intercept only model where the intercept represents the group level estimate for the probability to choose the color associated with a higher chance of winning pain relief (choicehigh prob). Drug was used as an additional within-subject factor when testing for differences among levodopa, naltrexone, and placebo. To account for repeated measures, we modeled a random intercept for each subject. To assess the effect of the subjects’ belief about which color was associated with a higher chance to win pain relief (as reported in the exit interview, see above) we added the factor ‘belief’ (either ‘correct belief’ or ‘false or no belief’) to this model.
To test whether endogenous pain modulation due to winning pain relief was related to participants’ personality trait of novelty seeking, pain modulation represented by the differences between test and control trials in the wheel of fortune in VAS ratings and the behaviorally assessed pain perception of the placebo and the levodopa condition were correlated with the NISS NS scores. To test further whether increases in pain modulation induced by levodopa were associated with novelty seeking, differences in pain modulation between the levodopa and placebo session were also correlated with the NISS NS scores. Before calculating these correlations, multivariate outliers were tested using a chi-square test on the squared Mahalanobis distance using an of 0.025 (Filzmoser, 2016), leading to the exclusion of one value for the correlation with the difference of pain modulation between the levodopa and placebo session.
The significance level was set to $5\%$ for all analyses. All statistical analyses were performed using statistical computing software R version 3.5.3 (R Development Core Team, 2019). Mixed model analyses were performed using the lme4 package (Bates et al., 2015). All linear mixed models were estimated using restricted maximum likelihood. Kenward-Roger correction as implemented in the lmerTest package (Kuznetsova et al., 2017) was used to calculate test statistics and degrees of freedom to account for the sample size. *For* general linear mixed effects models Wald χ2 was calculated using car package (Fox and Weisberg, 2011). Post-hoc tests and effect sizes were calculated on estimated marginal means using the emmeans package (Lenth, 2020) where appropriate. Effects sizes were calculated by dividing the difference in marginal means by the pooled standard deviation of the random effects and the residuals providing an estimate for the underlying population (Hedges, 2007). Tukey adjustment was used to account for multiple comparisons in post-hoc tests.
## Estimation of prediction errors and their role in endogenous pain modulation
To analyze how mechanisms of instrumental learning contribute to the observed choice behavior and how this related to reward-induced pain modulation, we fitted reinforcement learning (RL) models to participants’ choices in test trials of the wheel of fortune game. Such models were initially formulated for associative learning (Bush and Mosteller, 1951; Rescorla and Wagner, 1972) and adapted for instrumental learning (Sutton and Barto, 1998). RL models assume that actions are chosen based on the expected outcome. Learning is described as the adaptation of expectations based on experiences. Thus, learning is driven by the discrepancy between a present expectation and the obtained outcome, namely the prediction error. The speed of adaption of the expectation is described by the learning rate, which defines the exponential decay of the influence of previous outcomes on the currently present expectation. For trial-by-trial instrumental learning paradigms, the update of the expectation of an outcome related to a given action (in the present study: choice in the wheel of fortune) is operationalized by calculating the expected value Q of a choice as follows:[1]Qchoice,t+1=Qchoice,t+η×δt where *Qchoice is* the reward expectation for a given choice, t denotes the trial, η is the learning rate, and δt is the prediction error in trial t. The learning rate η determines the speed of adaption; the higher η the more is the expectation influenced by recent compared to former experiences. Since updating of expectations has been shown to differ dependent on the sign of the prediction error (Fontanesi et al., 2019; Gershman, 2015; Pedersen et al., 2017), we modeled independent learning rates for positive (η+) and negative (η-) prediction errors:[2]Qchoice,t+1=Qchoice,t+η+×δt,if δt>0[3]Qchoice,t+1=Qchoice,t+η−×δt,if δt≤0 The prediction error as the difference between the actual and the expected outcome in trial t is formulated as follows:[4]δt=Rt−Qchoice,t with Rt as the outcome of the choice in trial t.
In the wheel of fortune game, outcomes were implemented as changes in stimulation intensities. Accordingly, Rt was positive (+1) for temperature decreases in win trials or negative (–1) for temperature increases in lose trials. The formula shown above assumes a constant outcome sensitivity. To capture potential modulation of the outcome sensitivity, we implemented a scaled outcome sensitivity so that the reward in trial t was multiplied by an individual scaling factor ρ yielding a scaled prediction error:[5]δt=(ρ×Rt)−Qchoice,t Q values were initiated to zero and calculated separately for choices of the color associated with a higher chance to win pain relief (Qhigh prob) and choices of the color associated with a lower chance to win pain relief (Qlow prob).
While RL models traditionally used a softmax choice rule (Daw and Doya, 2006; Luce, 1959), recent studies on value-based decision making have implemented variants of the drift diffusion model (Ratcliff, 1978; Ratcliff and Rouder, 1998) to map expected values to choices (Fontanesi et al., 2019; Pedersen et al., 2017; Peters and D’Esposito, 2020). The drift diffusion model describes decisions as accumulation of noisy evidence for two choice options until a predefined threshold, representing either of the two options, is reached. Such drift diffusion models take response times (RT) of decisions into account and model mathematically cognitive processes underlying the decision process. Figure 8 depicts such a decision process. The range between the decision boundaries is represented by the boundary separation parameter α. Higher values of α lead to slower but more accurate decisions, that is, α represents the speed vs. accuracy tradeoff. The position of the starting point z between the boundaries is determined by a priori biases β toward one of the two options. This parameter β represents the relative distance of z between the boundaries. It can range from 0 to 1 where a value of 0.5 indicates no bias, values below indicate a bias for the lower choice and values above 0.5 a bias for the upper choice. The non-decision time τ describes time needed for processes that are unrelated to the decision process (e.g. stimulus processing). Correspondingly, the reaction time is defined as RT=τ+decision time. Acquisition of evidence starts from the starting point z at time τ as a random walk. The slope of this random walk is determined by the drift rate ν and a decision is made when either the upper or lower boundary is reached. Higher drift rates result in faster and more accurate decisions. The probability of the RT when choosing option x can then be calculated using the Wiener first-passage time distribution (Ratcliff, 1978):[6]RT(x)=Wiener(α,τ,β,ν) where Wiener() returns the probability that is chosen with the observed RT.
**Figure 8.:** *Schematic depiction of the drift diffusion process.Accumulation of evidence starts at point z which is defined by the a-priori bias β and the boundary separation α. Noisy evidence is integrated over time (represented by sample paths in blue and orange, for upper and lower boundary choices, respectively).*
Most variants of reward learning models that use the drift diffusion process as a choice rule replace the constant drift rate by an individually scaled difference of expected values for the both options (Fontanesi et al., 2019; Pedersen et al., 2017; Peters and D’Esposito, 2020). Thus, the drift rate νt, varies across trials as a function of the difference between expected values of the two choice options that in the wheel of fortune corresponded to Qhighprob and Qlowprob, respectively. We implemented a linear mapping of the difference in expected values like Pedersen et al., 2017 where this difference is multiplied by the scaling factor ν:[7]νt=(Qhighprob−Qlowprob)×ν As an alternative scaling method, we implemented a non-linear function as suggested by Fontanesi et al., 2019 in which the scaled difference in expected values is mapped to the drift rate using a sigmoid function, which more closely resembles the non-linear mapping of the softmax function:[8]νt=S((Qhighprob,t−Qlowprob,t)×ν) where S(x) is defined as:[9]S(x)=2 × νmax1 + e−x−νmax With that, ± νmax defines the upper and lower limit of the drift rate, respectively, while the shape or slope of the sigmoid function depends on the scaled difference of expected values.
In summary, we combined different parameterizations of the outcome sensitivity (static or scaled) and the mapping of expected values to the drift rate (linear or sigmoidal) into different models (Table 3).
**Table 3.**
| Model | Outcome sensitivity | Drift rate mapping |
| --- | --- | --- |
| Model 1 | static | linear |
| Model 2 | scaled | linear |
| Model 3 | static | sigmoid |
| Model 4 | scaled | sigmoid |
We used hierarchical Bayesian modeling to fit the reward learning models to the choices of the participants in the test trials. Hierarchical models estimate group and individual parameters simultaneously to mutually inform and constrain each other, which yields reliable estimates for both, individual and group level parameters (Gelman et al., 2013; Kruschke, 2014). Posterior distributions of the parameters were estimated using Hamiltonian Monte Carlo sampling with a No-U-Turn sampler as implemented in the probabilistic language Stan (Carpenter et al., 2017) via its R interface rstan (Stan Development Team, 2020). For each model parameter, we included a global intercept and the main effect of drug (levodopa, naltrexone, placebo). Both, intercept and main effect were allowed to vary for each participant and we modeled a correlation of individual terms for the drug effect across participants to account for repeated measures. We used a non-centered parameterization to reduce dependency between group and individual level parameters (Betancourt and Girolami, 2015). Therefore, both intercept and drug effect were defined by their location (group level effect), scale, and error (individual effects) distributions. A logistic transformation was applied to the learning rate (η+, η-) and a priori bias (β) parameters to restrict values to the range of 0, 1. The location parameters for the intercept of the learning rate were given standard normal priors (N[0,1]) and the scale of these parameters were given half-normal priors (HN[0,1]). The location of the drug effect on learning rate parameters were also given standard normal priors while the scale was given a half-normal prior of HN(0,0.1) to prevent allocation of high prior density at the edges of the range after logistic transformation, resulting in an almost flat prior. The location parameter for the intercept of the a-priori bias was given a normal prior of N(0,0.5) and the scale was given a half-normal prior HN(0,0.1). The location parameter for the drug effect was given a normal prior of N(0,0.5) and the scale was given a half-normal prior of HN(0,0.1). To ensure that the non-decision time (τ) was bounded to be lower than the reaction time the parameter was equivalently transformed to the range [0,1] and multiplied with each subject’s individual minimum reaction time in a given session. Priors were the same as for the learning rate, that is yielding a flat prior after transformation. We used an exponential transformation to constrain the reward sensitivity parameter (ρ), the boundary separation (α), drift rate scaling factor (ν), and the boundary of the drift rate (νmax) to be greater than 0. The location of the global intercept was given a normal prior of N(0.1,0.1) for the reward sensitivity, a normal prior of N(0,0.1) for the boundary separation, a normal prior of N(0.2,0.2) for the drift rate, and a normal prior of N(0.5,0.2) for the drift rate boundary. For the exponentially transformed parameters, the scale of the global intercept was given a half-normal prior of HN(0,0.1), the location of the drug effect was given a normal prior of N(0,0.5), and the scale of the drug effect was given a half-normal prior of N(0,0.1). Individual effects for the intercept as well as for the drug effect were all given standard normal priors. The correlation matrix of individual drug-level effects for each parameter was given a LKJ prior (Lewandowski et al., 2009) of LKJcorr[1]. All models were run on four chains with 4000 samples each. The first 1000 iterations were discarded as warm-up samples for each chain. The convergence of chains was confirmed by the potential scale reduction factor R^.
The fitted models were compared for their best predictive accuracy using K-fold cross-validation (Vehtari et al., 2017). For the cross-validation, we split data into $k = 10$ subsets with each subset containing data of 2–3 participants and calculated the expected log pointwise predictive density (ELPD) based on simulations for each hold-out set yk using parameters estimated from re-fitting the model to the training data set y-k. We calculated ELPDs, their differences, and the standard error of the differences using the R package loo (Vehtari et al., 2020). A higher ELPD indicates a better predictive accuracy. Such a better predictive accuracy was assumed if the difference in ELPD (ELPDdiff) for two models was at least two times the standard error of that difference (se(ELPDdiff)).
For the best fitting model, we performed posterior predictive checks by simulating replicated data sets from posterior draws. As the test statistic for the posterior predictive check, we examined the proportion of choices in favor of the option associated with a higher chance to win pain relief (choicehigh prob) in the last 2 blocks of the wheel of fortune game and compared the proportions observed in this data to the distribution of proportions found in the simulated data sets.
From the best fitting model, we used group level estimates for the main effect of ‘drug’ to compare model parameters between drug conditions using the $95\%$ highest density interval (HDI) of the difference of their posterior distributions.
The means of individual parameter posterior distributions were used to estimate prediction errors for single trials. To test whether these prediction errors predict endogenous pain modulation induced by the wheel of fortune task, we used linear mixed models with the fixed factors ‘prediction error’ and ‘drug’, and their interaction. A random intercept for each subject was included to account for repeated measures. Separate models for VAS ratings and behaviorally assessed pain perception as dependent variables were calculated.
## Funding Information
This paper was supported by the following grants:
## Data availability
Behavioral and questionnaire data is available as csv file at the project's Open Science Framework page (osf.io/5xjt9).
The following dataset was generated: DeschS SchweinhardtP SeymourB FlorH BeckerS 2022Endogenous modulation of pain reliefOpen Science Framework5xjt9
## References
1. Bannister K. **Descending pain modulation: influence and impact**. *Current Opinion in Physiology* (2019) **11** 62-66. DOI: 10.1016/j.cophys.2019.06.004
2. Barbano MF, Cador M. **Differential regulation of the consummatory, motivational and anticipatory aspects of feeding behavior by dopaminergic and opioidergic drugs**. *Neuropsychopharmacology* (2006) **31** 1371-1381. DOI: 10.1038/sj.npp.1300908
3. Barbano MF, Cador M. **Opioids for hedonic experience and dopamine to get ready for it**. *Psychopharmacology* (2007) **191** 497-506. DOI: 10.1007/s00213-006-0521-1
4. Bates D, Mächler M, Bolker BM, Walker SC. **Fitting linear mixed-effects models using lme4**. *Journal of Statistical Software* (2015) **67** 1-48. DOI: 10.18637/jss.v067.i01
5. Becker S, Kleinböhl D, Baus D, Hölzl R. **Operant learning of perceptual sensitization and habituation is impaired in fibromyalgia patients with and without irritable bowel syndrome**. *Pain* (2011) **152** 1408-1417. DOI: 10.1016/j.pain.2011.02.027
6. Becker S, Gandhi W, Elfassy NM, Schweinhardt P. **The role of dopamine in the perceptual modulation of nociceptive stimuli by monetary wins or losses**. *The European Journal of Neuroscience* (2013) **38** 3080-3088. DOI: 10.1111/ejn.12303
7. Becker S, Gandhi W, Kwan S, Ahmed AK, Schweinhardt P. **Doubling your payoff: winning pain relief engages endogenous pain inhibition**. *ENeuro* (2015) **2** 1-11. DOI: 10.1523/ENEURO.0029-15.2015
8. Beeler JA, Daw N, Frazier CRM, Zhuang X. **Tonic dopamine modulates exploitation of reward learning**. *Frontiers in Behavioral Neuroscience* (2010) **4**. DOI: 10.3389/fnbeh.2010.00170
9. Beeler JA. **Thorndike’s law 2.0: dopamine and the regulation of thrift**. *Frontiers in Neuroscience* (2012) **6**. DOI: 10.3389/fnins.2012.00116
10. Beiske AG, Loge JH, Rønningen A, Svensson E. **Pain in Parkinson’s disease: prevalence and characteristics**. *PAIN* (2009) **141** 173-177. DOI: 10.1016/j.pain.2008.12.004
11. Benedetti F. **The opposite effects of the opiate antagonist naloxone and the cholecystokinin antagonist proglumide on placebo analgesia**. *Pain* (1996) **64** 535-543. DOI: 10.1016/0304-3959(95)00179-4
12. Berridge KC, Robinson TE, Aldridge JW. **Dissecting components of reward: “liking”, “wanting”, and learning**. *Current Opinion in Pharmacology* (2009) **9** 65-73. DOI: 10.1016/j.coph.2008.12.014
13. Betancourt M, Girolami M. **Hamiltonian monte carlo for hierarchical models**. *Current Trends in Bayesian Methodology with Applications* (2015) **1** 79-101. DOI: 10.1201/b18502-5
14. Bradley MM, Lang PJ. **Measuring emotion: the self-assessment manikin and the semantic differential**. *Journal of Behavior Therapy and Experimental Psychiatry* (1994) **25** 49-59. DOI: 10.1016/0005-7916(94)90063-9
15. Breitenstein C, Korsukewitz C, Flöel A, Kretzschmar T, Diederich K, Knecht S. **Tonic dopaminergic stimulation impairs associative learning in healthy subjects**. *Neuropsychopharmacology* (2006) **31** 2552-2564. DOI: 10.1038/sj.npp.1301167
16. Bush RR, Mosteller F. **A mathematical model for simple learning**. *Psychological Review* (1951) **58** 313-323. DOI: 10.1037/h0054388
17. Carpenter B, Gelman A, Hoffman MD, Lee D, Goodrich B, Betancourt M, Brubaker MA, Guo J, Li P, Riddell A. **Stan: a probabilistic programming language**. *Journal of Statistical Software* (2017) **76**. DOI: 10.18637/jss.v076.i01
18. Chelnokova O, Laeng B, Eikemo M, Riegels J, Løseth G, Maurud H, Willoch F, Leknes S. **Rewards of beauty: the opioid system mediates social motivation in humans**. *Molecular Psychiatry* (2014) **19** 746-747. DOI: 10.1038/mp.2014.1
19. Cools R, Barker RA, Sahakian BJ, Robbins TW. **Enhanced or impaired cognitive function in parkinson’s disease as a function of dopaminergic medication and task demands**. *Cerebral Cortex* (2001) **11** 1136-1143. DOI: 10.1093/cercor/11.12.1136
20. Daw ND, Doya K. **The computational neurobiology of learning and reward**. *Current Opinion in Neurobiology* (2006) **16** 199-204. DOI: 10.1016/j.conb.2006.03.006
21. Dirks J, Petersen KL, Dahl JB. **The heat/capsaicin sensitization model: a methodologic study**. *The Journal of Pain* (2003) **4** 122-128. DOI: 10.1054/jpai.2003.10
22. Eikemo M, Biele G, Willoch F, Thomsen L, Leknes S. **Opioid modulation of value-based decision-making in healthy humans**. *Neuropsychopharmacology* (2017) **42** 1833-1840. DOI: 10.1038/npp.2017.58
23. Eippert F, Bingel U, Schoell ED, Yacubian J, Klinger R, Lorenz J, Büchel C. **Activation of the opioidergic descending pain control system underlies placebo analgesia**. *Neuron* (2009) **63** 533-543. DOI: 10.1016/j.neuron.2009.07.014
24. Faul F, Erdfelder E, Lang AG, Buchner A. **G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences**. *Behavior Research Methods* (2007) **39** 175-191. DOI: 10.3758/bf03193146
25. Fields HL. **How expectations influence pain**. *PAIN* (2018) **159 Suppl 1** S3-S10. DOI: 10.1097/j.pain.0000000000001272
26. Filzmoser P. **Identification of multivariate outliers: a performance study**. *Austrian Journal of Statistics* (2016) **34** 127-138. DOI: 10.17713/ajs.v34i2.406
27. Fontanesi L, Gluth S, Spektor MS, Rieskamp J. **A reinforcement learning diffusion decision model for value-based decisions**. *Psychonomic Bulletin & Review* (2019) **26** 1099-1121. DOI: 10.3758/s13423-018-1554-2
28. Fox J, Weisberg S. *An R Companion to Applied Regression* (2011)
29. Gandhi W, Becker S, Schweinhardt P. **Pain increases motivational drive to obtain reward, but does not affect associated hedonic responses: a behavioural study in healthy volunteers**. *European Journal of Pain* (2013) **17** 1093-1103. DOI: 10.1002/j.1532-2149.2012.00281.x
30. Gelman A, Carlin JB, Stern HS, Dunson DB, Vehtari A, Rubin DB. *Bayesian Data Analysis* (2013). DOI: 10.1201/b16018
31. Gershman SJ. **Do learning rates adapt to the distribution of rewards?**. *Psychonomic Bulletin & Review* (2015) **22** 1320-1327. DOI: 10.3758/s13423-014-0790-3
32. Glimcher PW. **Understanding dopamine and reinforcement learning: the dopamine reward prediction error hypothesis**. *PNAS* (2011) **108 Suppl 3** 15647-15654. DOI: 10.1073/pnas.1014269108
33. Gomtsian L, Bannister K, Eyde N, Roble D, Dickenson AH, Porreca F, Navratilova E. **Morphine effects within the rodent anterior cingulate cortex and rostral ventromedial medulla reveal separable modulation of affective and sensory qualities of acute or chronic pain**. *Physiology & Behavior* (2019) **176** 139-148. DOI: 10.1016/j.physbeh.2017.03.040
34. Hedges LV. **Effect sizes in cluster-randomized designs**. *Journal of Educational and Behavioral Statistics* (2007) **32** 341-370. DOI: 10.3102/1076998606298043
35. Holzer P. **Capsaicin: cellular targets, mechanisms of action, and selectivity for thin sensory neurons**. *Pharmacological Reviews* (1991) **43** 143-201. PMID: 1852779
36. Kakade S, Dayan P. **Dopamine: generalization and bonuses**. *Neural Networks* (2002) **15** 549-559. DOI: 10.1016/s0893-6080(02)00048-5
37. King CD, Goodin B, Kindler LL, Caudle RM, Edwards RR, Gravenstein N, Riley JL, Fillingim RB. **Reduction of conditioned pain modulation in humans by naltrexone: an exploratory study of the effects of pain catastrophizing**. *Journal of Behavioral Medicine* (2013) **36** 315-327. DOI: 10.1007/s10865-012-9424-2
38. Kleinböhl D, Hölzl R, Möltner A, Rommel C, Weber C, Osswald PM. **Psychophysical measures of sensitization to tonic heat discriminate chronic pain patients**. *Pain* (1999) **81** 35-43. DOI: 10.1016/s0304-3959(98)00266-8
39. Kroemer NB, Lee Y, Pooseh S, Eppinger B, Goschke T, Smolka MN. **L-dopa reduces model-free control of behavior by attenuating the transfer of value to action**. *NeuroImage* (2019) **186** 113-125. DOI: 10.1016/j.neuroimage.2018.10.075
40. Krohne HW, Egloff B, Kohlmann CW, Tausch A. **Untersuchungen mit einer deutschen version der “positive and negative affect schedule.”**. *Diagnostica* (1996) **42** 139-156
41. Kruschke JK. *Doing Bayesian Data Analysis: A Tutorial with R, JAGS, and Stan* (2014). DOI: 10.1016/B978-0-12-405888-0.09999-2
42. Kuznetsova A, Brockhoff PB, Christensen RHB. **Lmertest package: tests in linear mixed effects models**. *Journal of Statistical Software* (2017) **82** 1-26. DOI: 10.18637/jss.v082.i13
43. Lang PJ. *Self-Assessment Manikin* (1980)
44. Langdon AJ, Sharpe MJ, Schoenbaum G, Niv Y. **Model-based predictions for dopamine**. *Current Opinion in Neurobiology* (2018) **49** 1-7. DOI: 10.1016/j.conb.2017.10.006
45. Leknes S, Brooks JCW, Wiech K, Tracey I. **Pain relief as an opponent process: a psychophysical investigation**. *The European Journal of Neuroscience* (2008) **28** 794-801. DOI: 10.1111/j.1460-9568.2008.06380.x
46. Lenth R. *R Package* (2020)
47. Lewandowski D, Kurowicka D, Joe H. **Generating random correlation matrices based on vines and extended onion method**. *Journal of Multivariate Analysis* (2009) **100** 1989-2001. DOI: 10.1016/j.jmva.2009.04.008
48. Leyton M, Boileau I, Benkelfat C, Diksic M, Baker G, Dagher A. **Amphetamine-induced increases in extracellular dopamine, drug wanting, and novelty seeking: a PET/ [ 11C ] raclopride study in healthy men**. *Neuropsychopharmacology* (2002) **27** 1027-1035. DOI: 10.1016/S0893-133X(02)00366-4
49. Löffler M, Levine SM, Usai K, Desch S, Kandić M, Nees F, Flor H. **Corticostriatal circuits in the transition to chronic back pain: the predictive role of reward learning**. *Cell Reports. Medicine* (2022) **3**. DOI: 10.1016/j.xcrm.2022.100677
50. Luce RD. **On the possible psychophysical laws**. *Psychological Review* (1959) **66** 81-95. DOI: 10.1037/h0043178
51. Maruyama C, Deyama S, Nagano Y, Ide S, Kaneda K, Yoshioka M, Minami M. **Suppressive effects of morphine injected into the ventral bed nucleus of the stria terminalis on the affective, but not sensory, component of pain in rats**. *The European Journal of Neuroscience* (2018) **47** 40-47. DOI: 10.1111/ejn.13776
52. Matsumoto M, Hikosaka O. **Two types of dopamine neuron distinctly convey positive and negative motivational signals**. *Nature* (2009) **459** 837-841. DOI: 10.1038/nature08028
53. Meier IM, Eikemo M, Leknes S. **The role of mu-opioids for reward and threat processing in humans: bridging the gap from preclinical to clinical opioid drug studies**. *Current Addiction Reports* (2021) **8** 306-318. DOI: 10.1007/s40429-021-00366-8
54. Navratilova E, Xie JY, Okun A, Qu C, Eyde N, Ci S, Ossipov MH, King T, Fields HL, Porreca F. **Pain relief produces negative reinforcement through activation of mesolimbic reward-valuation circuitry**. *PNAS* (2012) **109** 20709-20713. DOI: 10.1073/pnas.1214605109
55. Navratilova E, Atcherley CW, Porreca F. **Brain circuits encoding reward from pain relief**. *Trends in Neurosciences* (2015a) **38** 741-750. DOI: 10.1016/j.tins.2015.09.003
56. Navratilova E, Xie JY, Meske D, Qu C, Morimura K, Okun A, Arakawa N, Ossipov M, Fields HL, Porreca F. **Endogenous opioid activity in the anterior cingulate cortex is required for relief of pain**. *The Journal of Neuroscience* (2015b) **35** 7264-7271. DOI: 10.1523/JNEUROSCI.3862-14.2015
57. Nyholm D, Lewander T, Gomes-Trolin C, Bäckström T, Panagiotidis G, Ehrnebo M, Nyström C, Aquilonius SM. **Pharmacokinetics of levodopa/carbidopa microtablets versus levodopa/benserazide and levodopa/carbidopa in healthy volunteers**. *Clinical Neuropharmacology* (2012) **35** 111-117. DOI: 10.1097/WNF.0b013e31825645d1
58. Pedersen ML, Frank MJ, Biele G. **The drift diffusion model as the choice rule in reinforcement learning**. *Psychonomic Bulletin & Review* (2017) **24** 1234-1251. DOI: 10.3758/s13423-016-1199-y
59. Pessiglione M, Seymour B, Flandin G, Dolan RJ, Frith CD. **Dopamine-dependent prediction errors underpin reward-seeking behaviour in humans**. *Nature* (2006) **442** 1042-1045. DOI: 10.1038/nature05051
60. Peters J, D’Esposito M. **The drift diffusion model as the choice rule in inter-temporal and risky choice: a case study in medial orbitofrontal cortex lesion patients and controls**. *PLOS Computational Biology* (2020) **16**. DOI: 10.1371/journal.pcbi.1007615
61. Pizzagalli DA, Evins AE, Schetter EC, Frank MJ, Pajtas PE, Santesso DL, Culhane M. **Single dose of a dopamine agonist impairs reinforcement learning in humans: behavioral evidence from a laboratory-based measure of reward responsiveness**. *Psychopharmacology* (2008) **196** 221-232. DOI: 10.1007/s00213-007-0957-y
62. Rabiner EA, Beaver J, Makwana A, Searle G, Long C, Nathan PJ, Newbould RD, Howard J, Miller SR, Bush MA, Hill S, Reiley R, Passchier J, Gunn RN, Matthews PM, Bullmore ET. **Pharmacological differentiation of opioid receptor antagonists by molecular and functional imaging of target occupancy and food reward-related brain activation in humans**. *Molecular Psychiatry* (2011) **16** 826-835. DOI: 10.1038/mp.2011.29
63. Ratcliff R. **A theory of memory retrieval**. *Psychological Review* (1978) **85** 59-108. DOI: 10.1037/0033-295X.85.2.59
64. Ratcliff R, Rouder JN. **Modeling response times for two-choice decisions**. *Psychological Science* (1998) **9** 347-356. DOI: 10.1111/1467-9280.00067
65. Raynor K, Kong H, Chen Y, Yasuda K, Yu L, Bell GI, Reisine T. **Pharmacological characterization of the cloned kappa-, delta-, and mu-opioid receptors**. *Molecular Pharmacology* (1994) **45** 330-334. PMID: 8114680
66. R Development Core Team
2019R: A language and environment for statistical computingVienna, AustriaR Foundation for Statistical Computinghttps://www.r-project.org/index.html. (2019)
67. Rescorla RA, Wagner AR, Black AH, Prokasy WF. *Classical Conditioning. 2. Current Research and Theory* (1972) 64-69
68. Rinne UK, Birket-Smith E, Dupont E, Hansen E, Hyyppä M, Marttila R, Mikkelsen B, Pakkenberg H, Presthus J. **Levodopa alone and in combination with a peripheral decarboxylase inhibitor benserazide (madopar) in the treatment of parkinson’s disease: a controlled clinical trial**. *Journal of Neurology* (1975) **211** 1-9. DOI: 10.1007/BF00312459
69. Roth M, Hammelstein P. **The need inventory of sensation seeking (NISS)**. *European Journal of Psychological Assessment* (2012) **28** 11-18. DOI: 10.1027/1015-5759/a000085
70. Santesso DL, Evins AE, Frank MJ, Schetter EC, Bogdan R, Pizzagalli DA. **Single dose of a dopamine agonist impairs reinforcement learning in humans: evidence from event-related potentials and computational modeling of striatal-cortical function**. *Human Brain Mapping* (2009) **30** 1963-1976. DOI: 10.1002/hbm.20642
71. Savage SW, Zald DH, Cowan RL, Volkow ND, Marks-Shulman PA, Kessler RM, Abumrad NN, Dunn JP. **Regulation of novelty seeking by midbrain dopamine D2/D3 signaling and ghrelin is altered in obesity**. *Obesity* (2014) **22** 1452-1457. DOI: 10.1002/oby.20690
72. Schultz W. **Multiple dopamine functions at different time courses**. *Annual Review of Neuroscience* (2007) **30** 259-288. DOI: 10.1146/annurev.neuro.28.061604.135722
73. Schultz W. **Dopamine reward prediction error coding**. *Dialogues in Clinical Neuroscience* (2016) **18** 23-32. DOI: 10.31887/DCNS.2016.18.1/wschultz
74. Seymour B. **Pain: a precision signal for reinforcement learning and control**. *Neuron* (2019) **101** 1029-1041. DOI: 10.1016/j.neuron.2019.01.055
75. Sirucek L, Price RC, Gandhi W, Hoeppli ME, Fahey E, Qu A, Becker S, Schweinhardt P. **Endogenous opioids contribute to the feeling of pain relief in humans**. *Pain* (2021) **162** 2821-2831. DOI: 10.1097/j.pain.0000000000002285
76. Smith KS, Berridge KC, Aldridge JW. **Disentangling pleasure from incentive salience and learning signals in brain reward circuitry**. *PNAS* (2011) **108** E255-E264. DOI: 10.1073/pnas.1101920108
77. Stan Development Team
2020RStan: the R interface to stanRStanhttp://mc-stan.org/. *RStan* (2020)
78. Sutton RS, Barto AG. **Reinforcement learning: an introduction**. *IEEE Transactions on Neural Networks* (1998) **9**. DOI: 10.1109/TNN.1998.712192
79. Tindell AJ, Berridge KC, Zhang J, Peciña S, Aldridge JW. **Ventral pallidal neurons code incentive motivation: amplification by mesolimbic sensitization and amphetamine**. *The European Journal of Neuroscience* (2005) **22** 2617-2634. DOI: 10.1111/j.1460-9568.2005.04411.x
80. Vaillancourt DE, Schonfeld D, Kwak Y, Bohnen NI, Seidler R. **Dopamine overdose hypothesis: evidence and clinical implications**. *Movement Disorders* (2013) **28** 1920-1929. DOI: 10.1002/mds.25687
81. Vehtari A, Gelman A, Gabry J. **Practical bayesian model evaluation using leave-one-out cross-validation and WAIC**. *Statistics and Computing* (2017) **27** 1413-1432. DOI: 10.1007/s11222-016-9696-4
82. Vehtari A, Gabry J, Magnusson M, Yao Y, Bürkner PC, Paananen T, Gelman A. *Package ‘Loo.’* (2020)
83. Vellani V, de Vries LP, Gaule A, Sharot T. **A selective effect of dopamine on information-seeking**. *eLife* (2020) **9**. DOI: 10.7554/eLife.59152
84. Villemure C, Slotnick BM, Bushnell MC. **Effects of odors on pain perception: deciphering the roles of emotion and attention**. *Pain* (2003) **106** 101-108. DOI: 10.1016/s0304-3959(03)00297-5
85. Vo A, Seergobin KN, Morrow SA, MacDonald PA. **Levodopa impairs probabilistic reversal learning in healthy young adults**. *Psychopharmacology* (2016) **233** 2753-2763. DOI: 10.1007/s00213-016-4322-x
86. Wall ME, Brine DR, Perez-Reyes M. **The metabolism of naltrexone in man**. *NIDA Research Monograph* (1981) **28** 105-131. PMID: 6790999
87. Walters ET, Williams A. **Evolution of mechanisms and behaviour important for pain**. *Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences* (2019) **374**. DOI: 10.1098/rstb.2019.0275
88. Watson D, Clark LA, Tellegen A. **Development and validation of brief measures of positive and negative affect: the PANAS scales**. *Journal of Personality and Social Psychology* (1988) **54** 1063-1070. DOI: 10.1037//0022-3514.54.6.1063
89. Weerts EM, McCaul ME, Kuwabara H, Yang X, Xu X, Dannals RF, Frost JJ, Wong DF, Wand GS. **Influence of OPRM1 asn40asp variant (A118G) on [ 11C ] carfentanil binding potential: preliminary findings in human subjects**. *The International Journal of Neuropsychopharmacology* (2013) **16** 47-53. DOI: 10.1017/S146114571200017X
90. Wittmann BC, Daw ND, Seymour B, Dolan RJ. **Striatal activity underlies novelty-based choice in humans**. *Neuron* (2008) **58** 967-973. DOI: 10.1016/j.neuron.2008.04.027
91. **World medical association declaration of helsinki: ethical principles for medical research involving human subjects**. *JAMA* (2013) **310** 2191-2194. DOI: 10.1001/jama.2013.281053
92. Xie JY, Qu C, Patwardhan A, Ossipov MH, Navratilova E, Becerra L, Borsook D, Porreca F. **Activation of mesocorticolimbic reward circuits for assessment of relief of ongoing pain: a potential biomarker of efficacy**. *Pain* (2014) **155** 1659-1666. DOI: 10.1016/j.pain.2014.05.018
93. Zald DH, Cowan RL, Riccardi P, Baldwin RM, Ansari MS, Li R, Shelby ES, Smith CE, McHugo M, Kessler RM. **Midbrain dopamine receptor availability is inversely associated with novelty-seeking traits in humans**. *The Journal of Neuroscience* (2008) **28** 14372-14378. DOI: 10.1523/JNEUROSCI.2423-08.2008
94. Zhang S, Mano H, Lee M, Yoshida W, Kawato M, Robbins TW, Seymour B. **The control of tonic pain by active relief learning**. *eLife* (2018) **7**. DOI: 10.7554/eLife.31949
|
---
title: 'Incorporating dietary fiber from fruit and vegetable waste in meat products:
a systematic approach for sustainable meat processing and improving the functional,
nutritional and health attributes'
authors:
- Abdul Haque
- Saghir Ahmad
- Z. R. A. A. Azad
- Mohd Adnan
- Syed Amir Ashraf
journal: PeerJ
year: 2023
pmcid: PMC9988266
doi: 10.7717/peerj.14977
license: CC BY 4.0
---
# Incorporating dietary fiber from fruit and vegetable waste in meat products: a systematic approach for sustainable meat processing and improving the functional, nutritional and health attributes
## Abstract
### Background
Every year, the food business produces a sizeable amount of waste, including the portions of fruits and vegetables that are inedible, and those that have reached a stage where they are no longer suitable for human consumption. These by-products comprise of components such as natural antioxidants (polyphenols, carotenoid etc.), dietary fiber, and other trace elements, which can provide functionality to food. Due to changing lifestyles, there is an increased demand for ready-to-eat products like sausages, salami, and meat patties. In this line, meat products like buffalo meat sausages and patties are also gaining the interest of consumers because of their rich taste. Meat, however, has a high percentage of fat and is totally deprived of dietary fiber, which poses severe health problems like cardiovascular (CV) and gastrointestinal diseases. The health-conscious consumer is becoming increasingly aware of the importance of balancing flavor and nutrition. Therefore, to overcome this problem, several fruit and vegetable wastes from their respective industries can be successfully incorporated into meat products that provide dietary fiber and play the role of natural antioxidants; this will slow down lipid oxidation and increase the shelf-life of meat products.
### Methodology
Extensive literature searches have been performed using various scientific search engines. We collected relevant and informative data from subject-specific and recent literature on sustainable food processing of wasted food products. We also looked into the various applications of waste fruit and vegetable products, including cereals, when they are incorporated into meat and meat products. All relevant searches meeting the criteria were included in this review, and exclusion criteria were also set.
### Results
The pomace and peels of fruits like grapes, pomegranates, cauliflower, sweet lime, and other citrus are some of the most commonly used fruit and vegetable by-products. These vegetable by-products help inhibit oxidation (of both lipids and proteins) and the growth of pathogenic and spoilage bacteria, all without altering the consumer’s acceptability of the product on a sensory level. When included in meat products, these by-products have the potential to improve the overall product quality and lengthen its shelf-life under certain circumstances.
### Conclusion
Cost-effective and easily accessible by-products from the fruit and vegetable processing industries can be used in meat products to enhance their quality features (physicochemical, microbial, sensory, and textural aspects) and health benefits. Additionally, this will provides environmental food sustainability by lowering waste disposal and improving the food’s functional efficacy.
## Introduction
Meat is a fundamental component of any balanced diet. Compared to other food sources, meat has a greater protein concentration and a higher degree of bioavailability (Chan, 2004; Biesalski, 2005). In addition, meat is an excellent source of omega-3 fatty acid, cobalamin and iron, containing a high level of all three nutrients (Sharma, Sheehy & Kolonel, 2013). There is a growing trend and popularity among consumers of processed meat products such as sausages, meat patties, kebabs, bologna, meat batters, frankfurters, meatballs, fermented sausage, burgers, etc. ( Daniel et al., 2011; Ritchie, Rosado & Roser, 2017; Shan et al., 2017). Processed meat products are becoming increasingly popular in the developed world due to rising customer demand for their convenience and perceived high quality in terms of taste, flavor, texture, and nutritional profile (Shan et al., 2017; Yeung & Huang, 2017).
However, the perception of meat fat as high in saturated fatty acids (SFA) has led to the belief that meat, particularly red meat, should be avoided. Also, the lack of dietary fiber (DF) in meat and processed meat is responsible for the unfavorable effects associated with these foods. Many health issues, including hypertension, coronary heart disease, obesity, and cardiovascular illnesses, have been linked to regular meat consumption (Micha, Michas & Mozaffarian, 2012). Specific components in diet have been linked to neurological disorders, which has been supported by scientific research (Gómez-Pinilla, 2008). The trend toward healthier food among consumers has led to an uptick in the market for functional meat products in recent years. As a direct consequence, products made from meat are now incorporating plenty of valuable bioactive components (Rashwan et al., 2020). Incorporating ingredients sourced from other natural sources that have been shown to enhance physiological functions of the body, such as increased immunity and anti-aging, qualifies a food product as a functional food (Chappalwar et al., 2020; Elkhalifa et al., 2021).
There is growing evidence that plant meals, especially whole grains and vegetables, are a great way to get the fiber your body needs (Stephen et al., 2017). DF is a type of carbohydrate found in plants that cannot be broken down by the digestive system and absorption by the body in the small intestine (Soliman, 2019). Furthermore, it’s been noticed that around the worldwide, human beings usually eat less than 20 g of DF daily (Stephen et al., 2017) whereas using the energy guideline of 2,000 kcal/day for women and 2,600 kcal/day for men, the recommended daily dietary fiber intake is 28 g/day for adult women and 36 g/day for adult men (USDA, 2005). Because of this, it is essential to produce a variety of DF containing food items. DF can be incorporated to meat products for various health benefits, since meat does not naturally contain any DF. It is anticipated that the need of meat products will expand dramatically in developing nations, and to some extent, it will also increase in developed nations countries (Future Market Insights Global and Consulting Pvt. Ltd., 2022, www.futuremarketinsights.com). As a result, the various meat products have the potential to be enrich with beneficial compounds, while the concentration of compounds that are disadvantageous can be reduced. Therefore, many components can be added to meat products without changing their fundamental qualities (Bhat, 2011). In addition to the health benefits, which customers enjoy, the DF-enriched meat products have enhanced functional features that make them more desirable to buyers (Aleson-Carbonell et al., 2005). Only a few researchers looked at the feasibility of various components. Still, they were able to come up with a functional food that is both safe and effective due to the addition of phytochemicals (Banerjee et al., 2020; Santhi, Kalaikannan & Natarajan, 2020).
Meat and meat products are highly perishable because of their complex composition and high water content, which promotes the growth and action of bacteria associated with decomposition and causes sensory and nutritional modifications due to lipids and proteins oxidation, among other things (Falowo, Fayemi & Muchenje, 2014; Amaral, Da Silva & Da Silva Lannes, 2018). There are various bioactive compounds such as antioxidant, antimicrobials, and novel processing technologies that have been used to extend the storage life of meat products. Natural compounds or substances and methods that employ few or no thermal reactions have been the focus of much attention (Jayasena & Jo, 2013; Sánchez-Ortega et al., 2014; Hugo & Hugo, 2015; Ur Rahman et al., 2018). Moreover, waste materials obtained from agro-industrial industries have acquired importance in the context of technologies that utilize natural compounds, because they represent a source of underutilized, but value-adding compounds or substances such as polyphenols, DF, antioxidants and terpenes (Mamma & Christakopoulos, 2014). The production of wine, juice, fruit and vegetable products like jam, marmalade, and oleoresins, are just some of the businesses responsible for producing a significant portion of these wastes, among others. It has been observed that the residues from these industries contains polyphenols and fiber (Devi et al., 2009; Quezada & Cherian, 2012). *In* general, agro-industrial wastes consist of a considerable percentage of inedible parts of fruits. These inedible parts can include seeds, shells, petals, and roots in some instances, and their weight can account for as much as 10 to 30 percent of the fruit’s overall mass. Utilizing these low-cost by-products from the food processing sectors can lead to new forms of indirect revenue and environmental sustainability. The potential pollution caused by discarding vegetable and fruit wastes is mitigated by their use as a source of nutritional fiber and natural antioxidants. As a result, not only does it benefit the environment by reducing the burden of their disposal, but it also aids the economy by reducing waste. Consumers are attracted to meat products that use plant by-products because they enhance the product’s overall quality, functionality, and shelf life (Pintado et al., 2016; Calderón-Oliver & López-Hernández, 2022). This also develops a sign of satisfaction among the non-vegetarian consumers regarding their diet.
Production of food waste materials in food processing industries is huge, and as a result, a valuable source of potential revenues is getting lost, and the cost of disposing these products is also increasing day by day. In addition to economic losses, unused waste material causes serious environmental concerns and possesses a threat for human health hazards. Hence, improved utilization of food waste materials could help to combat the food insecurity as well as provide potential health benefits as food waste possess several important bioactive compounds. Therefore, this review brings up a current update on incorporation of food waste material into meat products and its significance for improving the nutritional as well as functional characteristics. In addition, this review takes along all the possible food waste materials incorporation enriched with bioactive components to advance the efficacy of food product. Furthermore, the presented review will encourage the food scientific communities, food industries and ready to eat food product produces to bring up more sustainable way of producing food product for consumers with better functional characteristics. Subsequently, such innovative and sustainable way of meat fortification will increase the confidence of consumers while eating meat products.
## Survey methodology
Various scientific search engines like Science Direct, PubMed, Scopus, etc. were searched and approximate number of articles published in last 20 years were retrieved from different search engines of scientific literature. Keywords/phrases used to search relevant data were meat and meat products, sustainable food processing of wasted food, food and vegetable waste, improving the food functional efficacy, by-products from the fruit and vegetable processing industries, health benefits of dietary fiber etc., which reflected only subject-related literature. Exclusion criteria were also set. Studies that do not meet the current inclusion criteria, irrelevant to the topic, abstracts, conference proceedings, editorials and commentary with insufficient data were excluded.
## Nutritional value of meat
The prevalent levels of saturated fat in meat, especially red meat, give it a bad reputation regarding people’s health. Therefore, it is recommended to limit one’s consumption of meat, particularly red meat, in order to reduce an individual’s chance of developing certain disorders and diseases including cardiovascular (CV) diseases (Giromini & Givens, 2022). However, this perspective takes into account the fact that meat is a rich source of several micronutrients, including vitamins and trace minerals that are either absent from meals derived from plants or have a low bioavailability in those foods (Biesalski, 2005). In addition, because it is high in protein and low in carbohydrates, meat has a lesser glycemic index (GI), which is thought to be advantageous in preventing obesity, diabetes, and cancer. Meat is a product that is rich in protein and low in carbohydrates (Biesalski, 2005).
## Protein content of meat
Protein makes up about twenty percent of the average muscle’s composition. Lean meat’s dry matter is primarily made up of proteins, the most abundant component (Biesalski, 2005). Because the human body is unable to produce nine of the amino acids that are found in proteins from other substances, the body must obtain these amino acids from the food it consumes. These amino acids are considered essential or semi-essential. Meat typically has its protein in its protein at relatively high concentrations of the four most important amino acids-sulfur-containing amino acids, lysine, threonine, and tryptophan. It should not come as a surprise that the protein quality of animal proteins, such as those found in meat and milk, tends to be higher than that of plant proteins. When compared to plant proteins, the proteins found in animals are easier to digest. The fact that the majority of plant proteins are wrapped in polysaccharide matrix, inaccessible to proteolytic enzymes, helps to explain this phenomenon to some extent. The protein content of a meal can be improved by the complementing effect of combining plant and animal sources (Colmenero, Ayo & Carballo, 2005).
Proteins included in meat are essential for human nutrition. Since animal proteins must be digested into amino acids or tiny peptides before they can pass through the small intestinal wall and into the bloodstream, their nutritional value is directly related to how easily they can be broken down (Sun, Zhou & Zhao, 2015). Changes in meat protein aggregation and oxidation can affect its digestion by digestive enzymes (Sante-Lhoutellier, Aubry & Gatellier, 2007). Incorporation of plant-based waste in meat products generally does not harm protein digestibility and quality (Lin et al., 2019). Also in another study, Jucara (Euterpe edulis) fruit waste extract was able to stabilize protein oxidation in conventional broiler meat (da Silva Frasao et al., 2021). So, incorporating fruit and vegetable in meat products is safer regarding protein digestibility and helps in stabilizing the protein oxidation.
## Fat content of meat
Animal fat is primarily stored in fatty tissue, which can be further subdivided into adipose fat, subcutaneous fat, intermuscular fat, and marbling fat. The fat found within the muscles is known as marbling, and it helps produce a favorable texture (Lawrie & Ledward, 2014). Depending on the animal’s fat excretion and the preparation method, a given piece of meat may include varying amounts of intermuscular and depot fat. Saturated fatty acids (SFA) are commonly thought to make up the bulk of animal fat, whereas over $50\%$ of the fatty acids in meat are unsaturated (Lawrie & Ledward, 2014). Lipids in meat typically comprise less than half saturated fatty acids (beef 50–$52\%$) and as much as $70\%$ unsaturated fatty acids (Valsta, Tapanainen & Männistö, 2005).
The grinding, cooking, and storing steps in the processing of meat products expose lipids to the air, which causes them to oxidize quickly and irreversibly. Meat and meat products lose their desirable flavor and texture because of rancidity, turn brown, and create hazardous substances including malondialdehyde and cholesterol oxidation products due to lipid oxidation (Choe et al., 2014). The addition of various fruit and waste extract may help in retarding lipid oxidation and reducing the fat content of the meat products. In one study, to a more significant extent, persimmon peel extracts prevented lipid oxidation of pork patties while they were being refrigerated (Choe, Kim & Kim, 2017). In another study, inulin from chicory root was able to reduce the fat at a significant level in pork and chicken meatball (Montoya et al., 2022). Thus incorporation of fruit and vegetable waste and their extract in meat products may help in slowing down lipid oxidation and rancidity.
## Vitamin content of meat
Most fat-soluble vitamins are not easily eliminated from the body and are instead retained in the liver and adipose tissues alongside fat. On the other hand, the body tends to keep a far smaller quantity of water-soluble vitamins than it does fat-soluble vitamins. The majority of vitamins found in livestock and human diets are produced either by plants or microorganisms. Some animal cells can synthesize vitamins like vitamin D, niacin, and ascorbate, and convert pro-vitamins to the active form (McDowell, 2008). Additionally, commensal microbes in the digestive tracts of both ruminants and non-ruminants have the potential to operate as a source of some vitamins, including vitamin K and the water-soluble B-complex vitamins (McDowell, 2008). Regarding human nutrition, meat has long been acknowledged as a reliable source of B vitamins (Obeid et al., 2019).
## Mineral content of meat
The zinc and iron in meat are two of its most valuable nutrients. It is well established that meals rich in animal protein produce more excellent zinc absorption than meals rich in wholegrain cereal (Latunde-Dada & Neale, 2007). Selenium (Se) is an essential trace element to maintain good health. Processes like as antioxidative defense mechanism, inflammation lessening, thyroid hormone synthesis, fertility, reproduction, and DNA synthesis rely on this selenoprotein component (Antonyak et al., 2018). The mineral composition of frankfurter-style sausages was drastically altered when buckwheat husk was added. The addition of this non-meat item resulted in a greater concentration of several trace elements, including manganese, calcium, potassium, and magnesium; the first of these increased by nearly six times, and the last, by more than 40 percent (Salejda et al., 2022). The research conducted on the quality of pork loaves with the addition of hemp seeds, de-hulled hemp seeds, hemp protein, and hemp flour confirmed that there was a significant increase in magnesium and manganese after the addition of these non-meat ingredients to a meat product. This increase occurred after the addition of hemp seeds, hemp protein and hemp flour (Zając et al., 2019). Manganese, potassium, and magnesium levels in meat emulsion were all boosted by the addition of house cricket flour in place of lean meat/fat portions (Kim et al., 2017).
Selenium (Se) is an essential trace element to maintain good health. Processes like as antioxidative defense mechanism, inflammation lessening, thyroid hormone synthesis, fertility, reproduction, and DNA synthesis rely on this selenoprotein component (Hariharan & Dharmaraj, 2020).
## Dietary fiber incorporation in meat products and its effect in various metabolic disorders
Dietary fiber is derived from plant sources, especially agro-waste. These dietary fibers are an excellent source of natural antioxidants and may serve to improve the shelf life of meat products when incorporated into them. Antioxidants are an additive used in meat products that could be substituted with natural antioxidants obtained from DF. These substances prevent the oxidation of lipids, proteins, and pigments in foods, extending their shelf life and maintaining the food’s original color, texture, aroma, flavor, and overall quality (Damodaran & Parkin, 2017). Many fruit and vegetable wastes are the attractive source of DF. When these wastes included in different meat products, the DF content was noticed as in one of the studies, Mosambi (sweet lime) peel powder was incorporated in sausages and patties at various levels. After incorporation, $7.33\%$ and $6.24\%$ of total DF were found in sausage and patties, respectively (Younis, Ahmad & Malik, 2021). Similarly, in other study, $3\%$ DF was found by incorporating the upper stem of white cauliflower ($7.5\%$) in beef sausage (Abul-Fadl, 2012). While, in another study, $2.94\%$ DF was found when $40\%$ oyster mushroom was added in chicken sausage (Ahmad et al., 2020) as shown in Fig. 1. The inclusion of dietary fiber in meat products also helps in health management. Waste from fruits and vegetables includes phytic acid (Jayadev, 2017; Ani & Abel, 2018), which acts as a cation exchange component and a mineral chelator for positively charged ions like cadmium, calcium, zinc, and copper (Ekholm et al., 2003). Chelating agents can influence metal toxicity by mobilizing the toxic metal mainly into the urine (Flora & Pachauri, 2010).
**Figure 1:** *Incorporation of food waste material enriched with dietary fiber to enhance the meat product dietary fiber content.*
Dietary phytochemicals are substances derived from plants that are not considered nutrients but have been connected with a lower risk of developing certain chronic diseases (Liu, 2004). When these phytochemicals are ingested consistently through food, they even have the potential to protect against some malignancies and cardiovascular illnesses (Okarter & Liu, 2010). Soluble dietary fiber, such as barley and oats, protects against heart-related disorders and certain malignancies, and lowers total cholesterol and low-density lipoprotein cholesterol (Pins, 2006). Insoluble dietary fiber, such that found in wheat bran, has been demonstrated to reduce the risk of colon cancer, as well as the risk of developing other cancer, obesity, and gastro-intestinal issues (Stevenson et al., 2012).
## Role of dietary fiber in human health
Roughage-rich diets have been shown to lessen the risk of obesity, CV disease, and several malignancies. DF shows the chances of reduction of several diseases that occurs due to lack of fiber in meat products after incorporating DF from different fruit and vegetable sourced wastes. Foods high in DF have a bulking effect. Over the past few years, an increasing body of research has focused on how to reap the health benefits of DF, mainly by adding fiber to a diet that is otherwise deficient in recent diet due to highly processed food product. DF acts as a laxative, reduces blood cholesterol, and provides antioxidant activity; these are just three of the many benefits that have been tallied during the last half-century. Some DF can’t do what they are credited with doing, while others can’t do it as well or as often.
## Dietary fiber lowers cholesterol and prevents cardiovascular diseases
Because of its ability to lower blood cholesterol and triglyceride levels, DF has gained attention as a potential nutraceutical agent for preventing CV diseases (McRae, 2017). They found that β-glucan or psyllium fibers reduces significantly total serum and LDL cholesterol levels. Diets high in dietary fiber could dramatically lower total cholesterol, on consuming fruits and vegetables in abundance (Soliman, 2019). A few essential mechanisms include the inhibition of lipoprotein formation and cholesterol synthesis, increased insulin sensitivity due to delayed macronutrient absorption (Nweze, Nebechukwu & Bawa, 2021). As a result, consuming DF plays a significant role in reducing the potential of CV diseases.
## Dietary fiber and prevention of gastrointestinal diseases
It is widely known that increasing one’s consumption of fiber results in decreased transit periods, increased feces volume, and enhanced fluid concentration (Soliman, 2019). Compared to the other types of fibers typically taken in through diet, cellulose tends to produce a larger volume of feces, which in turn decreases the transit period (Wichert et al., 2002). It has been demonstrated that DF can absorb various mutagens like Trp-P-1, Trp-P-2, AαC and MeAαC (processed induced mutagenic heterocyclic amines), hence decreasing the amount of time the colon is exposed to these substances (Raman et al., 2013). In short, DF reduces transit time, dilutes and binds toxic colon chemicals, and reduces mucosal exposure to many of these substances. Dietary fiber’s role against colon cancer has been outlined in detail. Anti-carcinogenic activity can be achieved in two ways: (a) by decreasing the synthesis of carcinogenic compounds in the colon and (b) by increasing fecal volume, which decreases the contact of cancer risk agents present in feces with intestinal mucosa (Sharma, Yadav & Ritika, 2008). DF helps lower cancer risk, especially for colorectal cancer, which begins in the large intestine, by promoting fermentation that produces short-chain fatty acids (Xiong et al., 2022).
## Role of dietary fiber in satiety regulation and weight control
Due to the intake of DF in food, the bulk is created, which gives a full stomach feeling and stops a person from eating further for a longer time. Satiety regulation and gastrointestinal emptying are aided by the hormones produced in the gut, which are increased by DF (Costabile et al., 2018). Researchers from all around the world have looked at the weight-control benefits of DF (Puupponen-Pimiä et al., 2002). A total weight loss of 2.494 kg was found in a meta-analysis of studies involving different fiber types (Thompson et al., 2017). Another meta-analysis found that using chitosan supplements decreased nearly 1.814 pounds in total body weight loss (Mhurchu et al., 2005).
## Dietary fiber in the prevention of constipation
Water-insoluble, non-fermentable fibers immediately increase the luminal size, which in turn results in shorter gut transit time, which promotes laxation. On the other hand, water-soluble fibers have a high water-holding capacity, which results in bulky, soft stools that are easier to pass (Slavin, 2013). Two different meta-analyses indicate that fiber supplementation significantly increases stool frequency compared to placebo (Yang et al., 2012; Christodoulides et al., 2016). The risk of getting hemorrhoids may be reduced by taking a fiber supplement; if doing so reduces, the symptoms of constipation and the straining that accompany it.
## Dietary fiber in the prevention of inflammatory bowel disease
There was a $56\%$ decrease in the incidence of Crohn’s disease and a $20\%$ decrease in the incidence of ulcerative colitis, when comparing the highest and lowest categories of DF intake. However, the Crohn’s disease meta-analysis revealed substantial heterogeneity. The incidence of Crohn’s disease was observed to reduce by $13\%$ for every 10 g/d increase in dietary fiber, suggesting a linear dose-response association between the two (Liu et al., 2015). The short-chain fatty acid component of fermentable fiber, butyrate, has been hypothesized to have an anti-inflammatory impact via downregulating the transcription factor NF-kB. According to a meta-analysis supporting this anti-inflammatory effect, supplementing with DF led to a small, but statistically significant decrease in C-reactive protein levels in people who were overweight or obese (Jiao et al., 2015).
## Effect of dietary fiber incorporation on physico-chemical properties of meat products
Fiber addition to meat products has changed their overall composition, which has led to the development of new fiber sources and opened up exciting prospects for their application across several industries. The use of dried pumpkin pulp reduced the patties’ moisture content and raised their ash content. Raw and cooked patties both had a higher pH after incorporation. The capacity to retain moisture improved, as the amount of dry pumpkin pulp added rose. The cooking yield and diameter change, when pumpkin pulp was added were not significantly different (Serdaroğlu et al., 2018). This increase in ash content of meat product was probably due to higher mineral content of pumpkin matrix. As meat ash content rises, it could help end specific mineral deficiency. Trace minerals (iron, zinc, selenium, iodine, copper, chromium, manganese and molybdenum) perform vital functions within the body including thyroid metabolism, antioxidant activity and immune function (Vural et al., 2020). For example, zinc (Zn) is the second most abundant trace element in human, which can’t be stored in the body, thus regular dietary intake is required Zn microelement is very essential for male fertility. It could be considered as a nutrient marker with many potentials in prevention, diagnosis, and treatment of male infertility (Fallah, Mohammad-Hasani & Colagar, 2018). In one study, protein content, mineral content, and crude fiber content were increased by the incorporation of white cauliflower by-product flour in beef sausage; pH was found to be increased by the incorporation of flour made from the upper stem of white cauliflower. The cooking yield was also increased gradually with the incorporation of white cauliflower stem powder, because of its water yielding property (Abul-Fadl, 2012). The quantity of protein and fat in emulsified pork meatballs reduced as rice bran was added. In contrast, the amount of carbohydrates in the meatballs grew dramatically as rice bran was added (Choi et al., 2011). The protein level of frankfurter beef sausages was raised by one percentage point, and the ash content was raised dramatically when $7\%$ of the residue was added. The benefits of increased ash content has already been discussed in previous example. Furthermore, overall lipid levels dropped, and this reduction in lipid provided longer shelf life by cutting down the chance of lipid oxidation. The question of what sort of quality this residue adds to the protein remains unanswered (Savadkoohi et al., 2014). A few other examples are showing the changes in different characteristics of meat products by incorporating various plant-based materials, especially waste, have been listed in Table 1.
**Table 1**
| Product | Incorporation | Characteristics before incorporationat 0 day of study | Characteristics after incorporationat 0 day of study | References |
| --- | --- | --- | --- | --- |
| Sausage | Pomegranate peel (3%) | Moisture–61.89%Protein–15.85%Ash–2.82%pH–7.12 | Moisture–58.82%Protein–16.32%Ash–3.09%pH–7.15 | El-Nashi et al. (2015) |
| Beef sausage | White cauliflower upperstem flour (7.5%) | Moisture–60.81%Fat–37.75%Fiber–0.92% | Moisture–63.40%Fat–28.40%Fiber–3% | Abul-Fadl (2012) |
| Frankfurter | Buckwheat by-product (3%) | Storage loss–1.93%Production yield–85.8%Gumminess–18.2 NChewiness–12.2 Nm | Storage loss–1.36%Production yield–85.5%Gumminess–16.9 NChewiness–11.5 Nm | Salejda et al. (2022) |
| Beef patty | Dried pumpkin pulp and seed (5%) | Moisture–59.71%Ash–2.76%pH–5.89WHC–75% | Moisture–55.83%Ash–2.9%pH–5.92WHC–79.8% | Serdaroğlu et al. (2018) |
| Pork chorizo | Oregano essential oil (0.1%) | Shear force–244.35 gChewiness–180.2 g-mmpH–5.34Browning index–97.16 | Shear force–216.99 gChewiness–95.43 g-mmpH–5.27Browning index–103.42 | Perales-Jasso et al. (2018) |
## Effect of dietary fiber incorporation on functional properties of meat products
Many fruits and vegetable waste acts as DF. This waste, at the same time, contains such compounds, which show antioxidant properties. In addition to protecting cells from damage caused by free radicals, chemicals like polyphenols can also treat diseases and their symptoms by inhibiting inflammatory responses and halting the progression of infection (Shay et al., 2015). As a result, incorporating such substances into meat products may improve their functionality and, thus, their healthfulness. The remnants of fresh dates are the source of polyphenols; when they were included in the formulation of bologna sausages ($15\%$ of the total), the finished product had a polyphenol level of $1.02\%$ (Sánchez-Zapata et al., 2011). This finding suggests that adding extracts rich in polyphenolic chemicals to meat products can serve as an antioxidant and provide health benefits to the end user. Lycopene, a carotenoid present in 80-$90\%$ of ripe tomatoes, has been linked to various health benefits, including a reduced risk of prostate cancer and CV disease (Friedman, 2013). After 21 days in storage, 0.58 mg of lycopene per 100 g of product was discovered in concentrations of up to $1.2\%$ of the tomato peel in sausage. Lycopene was detected in beef burgers cooked at 180 °C for 2 min (Luisa García, Calvo & Dolores Selgas, 2009). Furthermore, the leftovers from tomatoes can be a source of amino acids and trace elements. By incorporating just $7\%$ of the residue, the protein level was raised by $1\%$, while also boosting the ash content from $2.18\%$ to $2.45\%$ in frankfurter beef sausages. The percentage of total lipids dropped from 20.07 to 19.4 as a result (Savadkoohi et al., 2014).
Prebiotics are a potential health benefit of fruit and vegetable waste. Prebiotics are elements that can survive stomach acid, mammalian enzymatic hydrolysis, absorption in GI tract; they are fermentable by intestinal flora and hence foster the expansion of beneficial bacteria like probiotics (Gibson et al., 2004). Prebiotics come in many forms, but some common ones include cellulose and fiber. Fiber from nopal flour ($2\%$) and pineapple peel flour ($3\%$) added to cooked sausages helped inoculated thermos-tolerant (probiotic) lactic acid bacteria thrive over 20 days in storage (Díaz-Vela, Totosaus & Pérez-Chabela, 2015). It is important to note that the amount of bacteria in a formulation such as this one, which contains both a probiotic and a prebiotic, needs to be closely controlled because the bacteria have the potential to degrade the overall quality of the product. It has also been observed in the above-discussed case that the inclusion of agro-industrial waste raises the mineral content of the meat products, which could lead to a rise in mineral consumption and help meet dietary guidelines.
## Effect of dietary fiber incorporation on sensory, textural, and color characteristics of meat products
Important sensory requirements for customer acceptance of meat products include aroma, flavor, color, appearance, tenderness, and juiciness. Adding pea cotyledon fiber to low-fat beef patties ($10\%$ and $14\%$ fat) improved tenderness without adversely affecting the juiciness or intensity of the meat flavor (Luisa García, Cáceres & Dolores Selgas, 2006). While adding lemon albedo to fermented sausages did not change the sausages’ odor, granularity, or salty or acidic taste, it did make them juicier. On the other hand, incorporating fresh lemon albedo enhanced the red color’s perceptual impact (Aleson-Carbonell et al., 2003). Tomato peel extract added to ground beef at concentrations of $1.5\%$, $3\%$, and $4.5\%$resulted in highly acceptable grades for the final product; however, the orange appearance and flavor linked with a $6\%$ addition reduced the acceptance rating due to high concentration of this ingredient (Savadkoohi et al., 2014).
The acceptability and desire to purchase a food product and the quality are all reflected in its aesthetic (texture and color), especially for meat products. Concentrations of fiber from citrus in meat products below $1.5\%$ ($0.5\%$ and $1\%$) increased hardness (Fernandez-Gines et al., 2003). In comparison, concentrations of fiber from citrus in meat products at $2\%$ reduced it, possibly as a result of the effect on the free water retention capacity; the final properties of the texture can be modified by the amount of fiber incorporation in the meat product (Fernandez-Gines et al., 2003). When raw albedo was added to fermented sausages, the sausages had lower chewiness values compared to the control, but when cooked albedo was added, the chewiness was greater than the control. Therefore, the addition of the type of lemon albedo (raw or cooked) in fermented sausages was capable of modifying the chewiness of the product. So to remove the bad effect associated with albedo incorporation in sausage was removed by cooking albedo prior to its incorporation in meat product. Similarly, in another study, pretreatment of sodium chloride was given to peel to remove/reduce the bitterness of the sausage arising from the incorporation of mosambi peel (Younis et al., 2019).
The carotenoid content and antioxidant action of citrus fruit fiber altered the brightness of bologna sausages throughout a wide concentration range ($0.5\%$, $1\%$, $1.5\%$, and $2\%$). This included changes in the intensity of the red and yellow colors (Fernandez-Gines et al., 2003). Irradiated hamburgers (2 and 4 kGy) retained their red color (a*) when dried tomato peel ($3\%$ to $6\%$) was added to the meat (Luisa García, Calvo & Dolores Selgas, 2009). Sausage with $0.6\%$, $0.9\%$, or $1.2\%$ dehydrated tomato peel kept its red hue for 21 days, but the brightness was dimmed and the yellow intensity (b*) was boosted, and there was an improvement in hardness and cutting force, but a decline in cohesiveness (Calvo, García & Selgas, 2008). Here, tomato peel’s high lycopene and beta-carotene concentration proved superior to pulp’s in preserving the red color of final meat products (Joseph et al., 2014). A few other examples showed the changes in different characteristics of meat products by incorporating various plant-based materials, especially waste, and have been listed in Table 1.
## Effect of dietary fiber incorporation on shelf-life of meat products
It is of the utmost importance that the quality and shelf stability of a meat product is maintained during its storage. It has been discovered that the addition of different kinds of fiber sources to meat products can alter the preservation quality in a variety of ways. The inclusion of chia seeds in (camburger) camel burger, too has recorded a reduced TBA (thiobarbituric acid) value, which implies lesser lipid oxidation in comparison to the control, and these burgers were determined to be organoleptically acceptable after storage for a period of 12 days (Zaki, 2018). Because it better inhibits oxymyoglobin oxidation, the addition of oat flour to chicken kofta has resulted in a product that is microbiologically safe and sensorial acceptable over the entire period of 15 days of storage (Lin & Lin, 2006). Grape seed, which includes flavonoids including catechin, epicatechin, procyanidins, and other chemicals with antimicrobial properties, is the most widely utilized and documented industrial waste for this purpose (Friedman, 2014). Uncertain processes, including disruption of the cytoplasmic membrane, blockage of specific metabolic pathways and enzymes, and chelation of critical metals for growth like zinc and iron, underlie the antibacterial effects of polyphenols and other natural chemicals (Daglia, 2012). Incorporation of grape seed in raw pork in aerobic packaging at 20 °C, when inoculated with 105 CFU of Listeria monocytogenes, Staphylococcus aureus, and Salmonella enterica. Decreased growth of L. monocytogenes by $17.5\%$, S. aureus by $14\%$, and S. entericaby $20\%$ (Shan et al., 2009). A 4-log decrease in the growth of L. monocytogenes at 4 °C, 1 log at 7 °C, and no changes at 12 °C was noticed in meat paté incorporated with pomegranate peel. The inoculation was done with 4 log CFU/g of L. monocytogenes at 4 °C, 7 °C, and 12 °C for 46 days (Hayrapetyan, Hazeleger & Beumer, 2012). A few other examples have been listed in Table 2, indicating the shelf-life through changes in microbiological characteristics and antioxidant capacity of different meat products as influenced by various incorporations.
**Table 2**
| Product | Incorporation | Characteristics before incorporation at 0 day of study | Characteristics after incorporation at 0 day of study | References |
| --- | --- | --- | --- | --- |
| Pork chorizo | Oregano essential oil (0.1%) | Antioxidant capacity (DPPH)–26.48%Mesophilic aerobes–4.19 log cfu/g | Antioxidant capacity (DPPH)–27.42%Mesophilic aerobes–4.19 log cfu/g | Perales-Jasso et al. (2018) |
| Beef patty | Blueberry pomace extract (50 g/500ml) | TBA number–0.619Protein carbonyl content–1.219 nmol carbonyl/mg proteinColiform bacteria–3.62 log cfu/g | TBA number–0.390Protein carbonyl content–0.463 nmol carbonyl/mg proteinColiform bacteria–3.50 log cfu/g | Babaoğlu et al. (2022) |
| Beef patty | Blackberry pomace extract (50 g/500 ml) | TBA number–0.619Protein carbonyl content–1.219 nmol carbonyl/mg proteinColiform bacteria–3.62 log cfu/g | TBA number–0.355Protein carbonyl content–0.621 nmol carbonyl/mg proteinColiform bacteria–3.60 log cfu/g | Babaoğlu et al. (2022) |
| Beef sausage | White cauliflower leaf midribs powder (7.5%) | TBA value–0.234 mg/kg sampleTVB–N–4.88 mg/kg sample | TBA value–0.132 mg/kg sampleTVB–N–5.72 mg/kg sample | Abul-Fadl (2012) |
| Sausage | Pomegranate peel (3%) | TBA–0.237 mg of malonaldhyde/kgTVN–8.38 mg nitrogen/100 g sampleYeast and mold count–2 log cfu/gColiform count–1.86 log cfu/g | TBA–0.235 mg of malonaldhyde/kgTVN–8.00 mg nitrogen/100 g sampleYeast and mold count–2.8 log cfu/gColiform count–2 log cfu/g | El-Nashi et al. (2015) |
## Health benefits by interaction of dietary fiber in meat products
Fortifying processed meat products with bioactive components that may help or mitigate the negative health consequences of eating processed meat is one strategy for combating the potential detrimental effects of eating processed meat. Studies reported a correlation between eating meat products incorporated with dietary fiber improves digestive health system, improve digestion, reduces the risk of coronary heart diseases and many more. Fortifying pork sausages with inulin from chicory root has been demonstrated to have considerable effects on the metabolites created in the gastrointestinal tract by the gut microbiota, as shown in a recent study using a rat model (Thøgersen et al., 2018). The short-chain fatty acids acetate, propionate, and butyrate have been identified as pivotal in the positive effects linked with dietary fiber consumption (Zhao et al., 2018; Gill et al., 2018), and their production was increased by fortification of pork sausage with inulin. Butyrate production was also increased by the inclusion of dietary fiber in salami (a fermented meat product) in a human intervention study conducted by Pérez-Burillo et al. [ 2020]. Additionally, dietary butyrylated starch has showed to improve gut short-chain fatty acid content and reduce the production of harmful and carcinogenic O6-methyl-2-deoxyguanosine adducts, the latter of which has been linked to a diet high in red meat (Le Leu et al., 2015). Therefore, recent data suggests that compounds containing fermentable dietary fiber and short-chain fatty acids can mitigate the negative effects of eating processed meat products in the colon. While research into the cancer-fighting potential of unfermentable fiber is still limited, preliminary results from animal model studies are promising in cancer prevention (Corpet & Pierre, 2003).
Colon health benefits from high calcium consumption have been suggested by cohort studies (Huncharek, Muscat & Kupelnick, 2009; Meng et al., 2019). Recent research by Thøgersen et al., [ 2018] in a rat model examined the impact of supplementing processed meat with calcium and inulin. Interestingly, consuming processed meat that has been fortified with calcium-rich milk minerals decreases the formation of undesirable N-nitroso compounds in the gastrointestinal tract and increases the formation of short-chain fatty acids in the colon, compared to eating processed meat that has not been fortified (Thøgersen et al., 2018). In light of these encouraging findings, it appears that potentially harmful effects associated with the consumption of meat can, in fact, be mitigated through the modification of the meat product matrix and the fortification of meat products or through the strategic design of meals that include components like dietary fiber and calcium which counteract unintended effects in the intestinal system that are associated with the consumption of meat.
Once it was discovered that obesity was linked to mortality-inducing conditions like cardiovascular disease, it became clear that this issue needed urgent attention. In meat product formulations, particularly emulsified products that are known to have a high energy value, the addition of dietary fiber can result in a reduction in total energy consumption (Blackwood et al., 2016). This is because dietary fiber acts as fat replacers, in addition to delaying gastric emptying and increasing the distension of the stomach, both of which contribute to a greater perception of satiety (Hoad et al., 2004). Perception of satiety avoids the intake of excess calories and helps in reducing the obesity. In a research, appetite study and in-vitro digestibility of meat product (bologna sausage) incorporated with chia mucilage was conducted (Câmara et al., 2020). It was found that chia mucilage was proven as effective fat replacer in bologna sausage with increased retention time in stomach and consequently resulted in same satiety (Câmara et al., 2020). This will prevent excess eating, and one may avoid obesity by consuming such soluble fiber-based formulations of meat products.
In another study, a meat product (sausage) was formulated using cereal (rice) flour. It was claimed that a daily serving size of 100 g of sausage may correct a potassium and iron deficiency, nearly restore magnesium, calcium, and vitamin A, and cut a dietary fiber deficiency in half (Sadovoy et al., 2021). Also, laboratory mice whose staple diet included the formed sausage showed a significantly increased biological value and safety, as measured by cytological examinations of their blood (Sadovoy et al., 2021). Thus, meat products with dietary fiber from various fruits and vegetables may have health benefits for several disorders and chronic diseases. Eating meat products enriched with dietary fiber and plant-based materials may remove multiple deficiencies.
## Conclusion
A sustainable food system aims to deliver food security and nutrition for all, keeping future generations in view. In recent years it has been noted that almost one-third of the total food produced is lost or wasted worldwide during food production or in the food processing system. Moreover, utilization of these food waste materials can contribute to the alleviation of food insecurity. Since, these food waste materials possess important bioactive compounds and incorporation of these into meat products could enhance the nutritional and functional characteristics of newly developed meat product. In this review, we found that incorporating several fruit and vegetable wastes in their extract or powder form into processed meat products had a very significant enrichment. Incorporating these agro-industrial wastes enhances the products’ sensory and textural qualities and shelf-life, as they are a rich source of DF, several antioxidants, and health-promoting agents. Since, meat products contain a high percentage of fat especially saturated fat and devoid of DF, which poses several health problems like cardiovascular and gastrointestinal diseases. The health conscious consumer is becoming increasingly aware of the importance of balancing flavor and nutrition. Therefore, to overcome this problem, utilization of these food waste materials could pave the way to improve the food insecurity as well as functional attributes. In the future, the incorporation of transformed low-cost food waste-derived products or diet may provide the possibility to lower food costs and this could be a strong incentive for the stakeholders in the business industry to get involved in the utilization of food waste in meat processing industries given that quality and safety are guaranteed. Furthermore, since there are wide ranges of food waste materials around the globe and its scope of utilization, food communities are needed to carry such more investigation concerning the physicochemical characteristics to improve food sustainable system. Finally, food waste recycling in processed meat for human consumption may contribute to the reduction of the environmental impact, improve environmental footprints, and help meet the requirement of a sustainable food system.
## References
1. Abul-Fadl MM. **Nutritional and chemical evaluation of white cauliflower by-products flour and the effect of its addition on beef sausage quality**. *Journal of Applied Sciences Research* (2012) **8** 693-704
2. Ahmad S, Jafarzadeh S, Ariffin F, Zainul Abidin S. **Evaluation of physicochemical, antioxidant and antimicrobial properties of chicken sausage incorporated with different vegetables**. *Italian Journal of Food Science* (2020) **32** 76-90. DOI: 10.14674/IJFS-1574
3. Aleson-Carbonell L, Fernandez-Lopez J, Sayas-Barbera E, Sendra E, Perez-Alvarez JA. **Utilization of lemon albedo in dry-cured sausages**. *Journal of Food Science* (2003) **68** 1826-1830. DOI: 10.1111/j.1365-2621.2003.tb12337.x
4. Aleson-Carbonell L, Fernández-López J, Pérez-Alvarez JA, Kuri V. **Functional and sensory effects of fibre-rich ingredients on breakfast fresh sausages manufacture**. *Food Science and Technology International* (2005) **11** 89-97. DOI: 10.1177/1082013205052003
5. Amaral AB, Da Silva MV, Da Silva Lannes SCS. **Lipid oxidation in meat: mechanisms and protective factors—a review**. *Food Science and Technology* (2018) **38** 1-15. DOI: 10.1590/fst.32518
6. Ani PN, Abel HC. **Nutrient, phytochemical, and antinutrient composition of**. *Food Science & Nutrition* (2018) **6** 653-658. DOI: 10.1002/fsn3.604
7. Antonyak H, Iskra R, Panas N, Lysiuk R, Malavolta M, Mocchegiani E. **Selenium**. *Trace Elements and Minerals in Health and Longevity. Healthy Ageing and Longevity* (2018) 63-98
8. Babaoğlu AS, Unal K, Dilek NM, Poçan HB, Karakaya M. **Antioxidant and antimicrobial effects of blackberry, black chokeberry, blueberry, and red currant pomace extracts on beef patties subject to refrigerated storage**. *Meat Science* (2022) **187** 108765. DOI: 10.1016/j.meatsci.2022.108765
9. Banerjee DK, Das AK, Banerjee R, Pateiro M, Nanda PK, Gadekar YP, Biswas S, McClements DJ, Lorenzo JM. **Application of Enoki Mushroom (**. *Foods* (2020) **9** 432. DOI: 10.3390/foods9040432
10. Bhat Z. **Functional meat products: a review**. *International Journal of Meat Science* (2011) **1** 1-14. DOI: 10.3923/ijmeat.2011.1.14
11. Biesalski H-K. **Meat as a component of a healthy diet—are there any risks or benefits if meat is avoided in the diet?**. *Meat Science* (2005) **70** 509-524. DOI: 10.1016/j.meatsci.2004.07.017
12. Blackwood AD, Salter J, Dettmar PW, Chaplin MF. **Dietary fibre, physicochemical properties and their relationship to health**. *Journal of the Royal Society for the Promotion of Health* (2016) **120** 242-247. DOI: 10.1177/146642400012000412
13. Calderón-Oliver M, López-Hernández LH. **Food vegetable and fruit waste used in meat products**. *Food Reviews International* (2022) **38** 628-654. DOI: 10.1080/87559129.2020.1740732
14. Calvo MM, García ML, Selgas MD. **Dry fermented sausages enriched with lycopene from tomato peel**. *Meat Science* (2008) **80** 167-172. DOI: 10.1016/j.meatsci.2007.11.016
15. Câmara AKFI, Geraldi MV, Okuro PK, Maróstica MR, da Cunha RL, Pollonio MAR. **Satiety and in vitro digestibility of low saturated fat Bologna sausages added of chia mucilage powder and chia mucilage-based emulsion gel**. *Journal of Functional Foods* (2020) **65** 103753. DOI: 10.1016/j.jff.2019.103753
16. Chan W. **Human nutrition | Macronutrients in meat**. *Encyclopedia of Meat Sciences* (2004) 614-618
17. Chappalwar AM, Pathak V, Goswami M, Verma AK. **Development of functional chicken patties with incorporation of mango peel powder as fat replacer**. *Nutrition & Food Science* (2020) **50** 1063-1073. DOI: 10.1108/NFS-07-2019-0230
18. Choe JH, Kim HY, Kim CJ. **Effect of persimmon peel (Diospyros kaki Thumb.) extracts on lipid and protein oxidation of raw ground pork during refrigerated storage**. *Korean Journal for Food Science of Animal Resources* (2017) **37** 254-263. DOI: 10.5851/kosfa.2017.37.2.254
19. Choe JH, Kim HY, Kim YJ, Yeo EJ, Kim CJ. **Antioxidant activity and phenolic content of persimmon peel extracted with different levels of ethanol**. *International Journal of Food Properties* (2014) **17** 1779-1790. DOI: 10.1080/10942912.2012.731460
20. Choi YS, Choi JH, Han DJ, Kim HY, Lee MA, Kim HW, Jeong JY, Kim CJ. **Effects of rice bran fiber on heat-induced gel prepared with pork salt-soluble meat proteins in model system**. *Meat Science* (2011) **88** 59-66. DOI: 10.1016/j.meatsci.2010.12.003
21. Christodoulides S, Dimidi E, Fragkos KC, Farmer AD, Whelan K, Scott SM. **Systematic review with meta-analysis: effect of fibre supplementation on chronic idiopathic constipation in adults**. *Alimentary Pharmacology & Therapeutics* (2016) **44** 103-116. DOI: 10.1111/apt.13662
22. Colmenero FJ, Ayo MJ, Carballo J. **Physicochemical properties of low sodium frankfurter with added walnut: effect of transglutaminase combined with caseinate, KCl and dietary fibre as salt replacers**. *Meat Science* (2005) **69** 781-788. DOI: 10.1016/j.meatsci.2004.11.011
23. Corpet DE, Pierre F. **Point: from animal models to prevention of colon cancer. Systematic review of chemoprevention in min mice and choice of the model system**. *Cancer Epidemiology, Biomarkers & Prevention: A Publication of the American Association for Cancer Research, Cosponsored by the American Society of Preventive Oncology* (2003) **12** 391-400
24. Costabile G, Griffo E, Cipriano P, Vetrani C, Vitale M, Mamone G, Rivellese AA, Riccardi G, Giacco R. **Subjective satiety and plasma PYY concentration after wholemeal pasta**. *Appetite* (2018) **125** 172-181. DOI: 10.1016/j.appet.2018.02.004
25. da Silva Frasao B, Lima Dos Santos Rosario AI, Leal Rodrigues B, Abreu Bitti H, Diogo Baltar J, Nogueira RI, Pereira da Costa M, Conte-Junior CA. **Impact of juçara (Euterpe edulis) fruit waste extracts on the quality of conventional and antibiotic-free broiler meat**. *Poultry Science* (2021) **100** 101232. DOI: 10.1016/j.psj.2021.101232
26. Daglia M. **Polyphenols as antimicrobial agents**. *Current Opinion in Biotechnology* (2012) **23** 174-181. DOI: 10.1016/j.copbio.2011.08.007
27. Damodaran S, Parkin KL. **Fennema’s Food Chemistry, Fifth Edition. CRC Press**. (2017). DOI: 10.1201/9781315372914
28. Daniel CR, Cross AJ, Koebnick C, Sinha R. **Trends in meat consumption in the USA**. *Public Health Nutrition* (2011) **14** 575-583. DOI: 10.1017/S1368980010002077
29. Devi MKA, Gondi M, Sakthivelu G, Giridhar P, Rajasekaran T, Ravishankar GA. **Functional attributes of soybean seeds and products, with reference to isoflavone content and antioxidant activity**. *Food Chemistry* (2009) **114** 771-776. DOI: 10.1016/j.foodchem.2008.10.011
30. Díaz-Vela J, Totosaus A, Pérez-Chabela ML. **Integration of agroindustrial co-products as functional food ingredients: cactus pear (**. *Journal of Food Processing and Preservation* (2015) **39** 2630-2638. DOI: 10.1111/jfpp.12513
31. Ekholm P, Virkki L, Ylinen M, Johansson L. **The effect of phytic acid and some natural chelating agents on the solubility of mineral elements in oat bran**. *Food Chemistry* (2003) **80** 165-170. DOI: 10.1016/S0308-8146(02)00249-2
32. Elkhalifa AEO, Al-Shammari E, Adnan M, Alcantara JC, Mehmood K, Eltoum NE, Awadelkareem AM, Khan MA, Ashraf SA. **Development and characterization of novel biopolymer derived from**. *Molecules* (2021) **26** 3609. DOI: 10.3390/molecules26123609
33. El-Nashi HB, Abdel Fattah AFAK, Abdel Rahman NR, Abd El-Razik MM. **Quality characteristics of beef sausage containing pomegranate peels during refrigerated storage**. *Annals of Agricultural Sciences* (2015) **60** 403-412. DOI: 10.1016/j.aoas.2015.10.002
34. Fallah A, Mohammad-Hasani A, Colagar AH. **Zinc is an essential element for male fertility: a review of Zn roles in men’s health, germination, sperm quality, and fertilization**. *Journal of Reproduction & Infertility* (2018) **19** 69-81. PMID: 30009140
35. Falowo AB, Fayemi PO, Muchenje V. **Natural antioxidants against lipid-protein oxidative deterioration in meat and meat products: a review**. *Food Research International* (2014) **64** 171-181. DOI: 10.1016/j.foodres.2014.06.022
36. Fernandez-Gines JM, Fernandez-Lopez J, Sayas-Barbera E, Sendra E, Perez-Alvarez JA. **Effect of storage conditions on quality characteristics of Bologna sausages made with citrus fiber**. *Journal of Food Science* (2003) **68** 710-714. DOI: 10.1111/j.1365-2621.2003.tb05737.x
37. Flora SJS, Pachauri V. **Chelation in metal intoxication**. *International Journal of Environmental Research and Public Health* (2010) **7** 2745-2788. DOI: 10.3390/ijerph7072745
38. Friedman M. **Anticarcinogenic, cardioprotective, and other health benefits of tomato compounds lycopene, α-tomatine, and tomatidine in pure form and in fresh and processed tomatoes**. *Journal of Agricultural and Food Chemistry* (2013) **61** 9534-9550. DOI: 10.1021/jf402654e
39. Friedman M. **Antibacterial, antiviral, and antifungal properties of wines and winery byproducts in relation to their flavonoid content**. *Journal of Agricultural and Food Chemistry* (2014) **62** 6025-6042. DOI: 10.1021/jf501266s
40. **Processed meat market by type, packaging, meat type & region—forecast 2022–2032. Delaware, United States**. (2022)
41. Gibson GR, Probert HM, Van Loo J, Rastall RA, Roberfroid MB. **Dietary modulation of the human colonic microbiota: updating the concept of prebiotics**. *Nutrition Research Reviews* (2004) **17** 259-275. DOI: 10.1079/NRR200479
42. Gill PA, van Zelm MC, Muir JG, Gibson PR. **Review article: short chain fatty acids as potential therapeutic agents in human gastrointestinal and inflammatory disorders**. *Alimentary Pharmacology & Therapeutics* (2018) **48** 15-34. DOI: 10.1111/apt.14689
43. Giromini C, Givens DI. **Benefits and risks associated with meat consumption during key life processes and in relation to the risk of chronic diseases**. *Foods* (2022) **11** 2063. DOI: 10.3390/foods11142063
44. Gómez-Pinilla F. **Brain foods: the effects of nutrients on brain function**. *Nature Reviews Neuroscience* (2008) **9** 568-578. DOI: 10.1038/nrn2421
45. Hariharan S, Dharmaraj S. **Selenium and selenoproteins: it’s role in regulation of inflammation**. *Inflammopharmacology* (2020) **28** 667-695. DOI: 10.1007/s10787-020-00690-x
46. Hayrapetyan H, Hazeleger WC, Beumer RR. **Inhibition of**. *Food Control* (2012) **23** 66-72. DOI: 10.1016/j.foodcont.2011.06.012
47. Hoad CL, Rayment P, Spiller RC, Marciani L, De Celis Alonso B, Traynor C, Mela DJ, Peters HPF, Gowland PA. *The Journal of Nutrition* (2004) **134** 2293-2300. DOI: 10.1093/jn/134.9.2293
48. Hugo CJ, Hugo A. **Current trends in natural preservatives for fresh sausage products**. *Trends in Food Science & Technology* (2015) **45** 12-23. DOI: 10.1016/j.tifs.2015.05.003
49. Huncharek M, Muscat J, Kupelnick B. **Colorectal cancer risk and dietary intake of calcium, vitamin D, and dairy products: a meta-analysis of 26,335 cases from 60 observational studies**. *Nutrition and Cancer* (2009) **61** 47-69. DOI: 10.1080/01635580802395733
50. Jayadev A. **Comparative analysis of nutritional and anti nutritional components of selected citrus fruit species**. *International Journal for Research in Applied Science and Engineering Technology* (2017) **V** 309-312. DOI: 10.22214/ijraset.2017.10047
51. Jayasena DD, Jo C. **Essential oils as potential antimicrobial agents in meat and meat products: a review**. *Trends in Food Science & Technology* (2013) **34** 96-108. DOI: 10.1016/j.tifs.2013.09.002
52. Jiao J, Xu J-Y, Zhang W, Han S, Qin L-Q. **Effect of dietary fiber on circulating C-reactive protein in overweight and obese adults: a meta-analysis of randomized controlled trials**. *International Journal of Food Sciences and Nutrition* (2015) **66** 114-119. DOI: 10.3109/09637486.2014.959898
53. Joseph S, Chatli MK, Biswas AK, Sahoo J. **Oxidative stability of pork emulsion containing tomato products and pink guava pulp during refrigerated aerobic storage**. *Journal of Food Science and Technology* (2014) **51** 3208-3216. DOI: 10.1007/s13197-012-0820-y
54. Kim HW, Setyabrata D, Lee Y, Jones OG, Kim YHB. **Effect of house cricket (**. *Journal of Food Science* (2017) **82** 2787-2793. DOI: 10.1111/1750-3841.13960
55. Latunde-Dada GO, Neale RJ. **Review: availability of iron from foods**. *International Journal of Food Science & Technology* (2007) **21** 255-268. DOI: 10.1111/j.1365-2621.1986.tb00405.x
56. Lawrie RA, Ledward DA. *Lawrie’s Meat Science* (2014)
57. Le Leu RK, Winter JM, Christophersen CT, Young GP, Humphreys KJ, Hu Y, Gratz SW, Miller RB, Topping DL, Bird AR, Conlon MA. **Butyrylated starch intake can prevent red meat-induced O**. *British Journal of Nutrition* (2015) **114** 220-230. DOI: 10.1017/S0007114515001750
58. Lin Y, Chen K, Tu D, Yu X, Dai Z, Shen Q. **Characterization of dietary fiber from wheat bran (**. *LWT* (2019) **102** 106-112. DOI: 10.1016/j.lwt.2018.12.024
59. Lin KW, Lin HY. **Quality characteristics of Chinese-style meatball containing bacterial cellulose (Nata)**. *Journal of Food Science* (2006) **69** SNQ107-SNQ111. DOI: 10.1111/j.1365-2621.2004.tb13378.x
60. Liu RH. **Potential synergy of phytochemicals in cancer prevention: mechanism of action**. *The Journal of Nutrition* (2004) **134** 3479S-3485S. DOI: 10.1093/jn/134.12.3479S
61. Liu X, Wu Y, Li F, Zhang D. **Dietary fiber intake reduces risk of inflammatory bowel disease: result from a meta-analysis**. *Nutrition Research* (2015) **35** 753-758. DOI: 10.1016/j.nutres.2015.05.021
62. Luisa García M, Calvo MM, Dolores Selgas M. **Beef hamburgers enriched in lycopene using dry tomato peel as an ingredient**. *Meat Science* (2009) **83** 45-49. DOI: 10.1016/j.meatsci.2009.03.009
63. Luisa García M, Cáceres E, Dolores Selgas M. **Effect of inulin on the textural and sensory properties of mortadella, a Spanish cooked meat product**. *International Journal of Food Science and Technology* (2006) **41** 1207-1215. DOI: 10.1111/j.1365-2621.2006.01186.x
64. Mamma D, Christakopoulos P. **Biotransformation of citrus by-products into value added products**. *Waste and Biomass Valorization* (2014) **5** 529-549. DOI: 10.1007/s12649-013-9250-y
65. McDowell LR. *Vitamins in animal and human nutrition* (2008)
66. McRae MP. **Dietary fiber is beneficial for the prevention of cardiovascular disease: an umbrella review of meta-analyses**. *Journal of Chiropractic Medicine* (2017) **16** 289-299. DOI: 10.1016/j.jcm.2017.05.005
67. Meng Y, Sun J, Yu J, Wang C, Su J. **Dietary intakes of calcium, iron, magnesium, and potassium elements and the risk of colorectal cancer: a meta-analysis**. *Biological Trace Element Research* (2019) **189** 325-335. DOI: 10.1007/s12011-018-1474-z
68. Mhurchu CN, Dunshea-Mooij C, Bennett D, Rodgers A. **Effect of chitosan on weight loss in overweight and obese individuals: a systematic review of randomized controlled trials**. *Obesity Reviews* (2005) **6** 35-42. DOI: 10.1111/j.1467-789X.2005.00158.x
69. Micha R, Michas G, Mozaffarian D. **Unprocessed red and processed meats and risk of coronary artery disease and type 2 diabetes—an updated review of the evidence**. *Current Atherosclerosis Reports* (2012) **14** 515-524. DOI: 10.1007/s11883-012-0282-8
70. Montoya L, Quintero N, Ortiz S, Lopera J, Millán P, Rodríguez-Stouvenel A. **Inulin as a fat-reduction ingredient in pork and chicken meatballs: its effects on physicochemical characteristics and consumer perceptions**. *Foods* (2022) **11** 1066. DOI: 10.3390/foods11081066
71. Nweze CC, Nebechukwu EW, Bawa MY. **Dietary fiber and risk of coronary heart diseases**. *GSC Advanced Research and Reviews* (2021) **9** 1-9. DOI: 10.30574/gscarr.2021.9.3.0280
72. Obeid R, Heil SG, Verhoeven MMA, van den Heuvel EGHM, de Groot LCPGM, Eussen SJPM. **Vitamin B12 intake from animal foods, biomarkers, and health aspects**. *Frontiers in Nutrition* (2019) **6** 93. DOI: 10.3389/fnut.2019.00093
73. Okarter N, Liu RH. **Health benefits of whole grain phytochemicals**. *Critical Reviews in Food Science and Nutrition* (2010) **50** 193-208. DOI: 10.1080/10408390802248734
74. Perales-Jasso YJ, Gamez-Noyola SA, Aranda-Ruiz J, Hernandez-Martinez CA, Gutierrez-Soto G, Luna-Maldonado AI, Silva-Vazquez R, Hume ME, Mendez-Zamora G. **Oregano powder substitution and shelf life in pork chorizo using Mexican oregano essential oil**. *Food Science & Nutrition* (2018) **6** 1254-1260. DOI: 10.1002/fsn3.668
75. Pérez-Burillo S, Pastoriza S, Gironés A, Avellaneda A, Pilar Francino M, Rufián-Henares JA. **Potential probiotic salami with dietary fiber modulates metabolism and gut microbiota in a human intervention study**. *Journal of Functional Foods* (2020) **66** 103790. DOI: 10.1016/j.jff.2020.103790
76. **A review of the effects of barley beta-glucan on cardiovascular and diabetic risk**. *Cereal Foods World* (2006) **51** 8-11. DOI: 10.1094/CFW-51-0008
77. Pintado T, Herrero AM, Jiménez-Colmenero F, Ruiz-Capillas C. **Strategies for incorporation of chia (**. *Meat Science* (2016) **114** 75-84. DOI: 10.1016/j.meatsci.2015.12.009
78. Puupponen-Pimiä R, Aura A-M, Oksman-Caldentey KM, Myllärinen P, Saarela M, Mattila-Sandholm T, Poutanen K. **Development of functional ingredients for gut health**. *Trends in Food Science & Technology* (2002) **13** 3-11. DOI: 10.1016/S0924-2244(02)00020-1
79. Quezada N, Cherian G. **Lipid characterization and antioxidant status of the seeds and meals of**. *European Journal of Lipid Science and Technology* (2012) **114** 974-982. DOI: 10.1002/ejlt.201100298
80. Raman M, Nilsson U, Skog K, Lawther M, Nair B, Nyman M. **Physicochemical characterisation of dietary fibre components and their ability to bind some process-induced mutagenic heterocyclic amines, Trp-P-1, Trp-P-2, AαC and MeAαC**. *Food Chemistry* (2013) **138** 2219-2224. DOI: 10.1016/j.foodchem.2012.11.111
81. Rashwan AK, Karim N, Shishir MRI, Bao T, Lu Y, Chen W. **Jujube fruit: a potential nutritious fruit for the development of functional food products**. *Journal of Functional Foods* (2020) **75** 104205. DOI: 10.1016/j.jff.2020.104205
82. Ritchie H, Rosado P, Roser M. **Meat and dairy production**. *Our World in Data* (2017)
83. Sadovoy V, Shchedrina T, Trubina I, Morgunova A, Franko E. **Cooked sausage enriched with essential nutrients for the gastrointestinal diet**. *Foods and Raw Materials* (2021) **9** 345-353. DOI: 10.21603/2308-4057-2021-2-345-353
84. Salejda AM, Olender K, Zielińska-Dawidziak M, Mazur M, Szperlik J, Miedzianka J, Zawiślak I, Kolniak-Ostek J, Szmaja A. **Frankfurter-type sausage enriched with buckwheat by-product as a source of bioactive compounds**. *Foods* (2022) **11** 674. DOI: 10.3390/foods11050674
85. Sánchez-Ortega I, García-Almendárez BE, Santos-López EM, Amaro-Reyes A, Barboza-Corona JE, Regalado C. **Antimicrobial edible films and coatings for meat and meat products preservation**. *The Scientific World Journal* (2014) **2014** 248935. DOI: 10.1155/2014/248935
86. Sánchez-Zapata E, Fernández-López J, Peñaranda M, Fuentes-Zaragoza E, Sendra E, Sayas E, Pérez-Alvarez JA. **Technological properties of date paste obtained from date by-products and its effect on the quality of a cooked meat product**. *Food Research International* (2011) **44** 2401-2407. DOI: 10.1016/j.foodres.2010.04.034
87. Sante-Lhoutellier V, Aubry L, Gatellier P. **Effect of oxidation on**. *Journal of Agricultural and Food Chemistry* (2007) **55** 5343-5348. DOI: 10.1021/jf070252k
88. Santhi D, Kalaikannan A, Natarajan A. **Characteristics and composition of emulsion-based functional low-fat chicken meat balls fortified with dietary fiber sources**. *Journal of Food Process Engineering* (2020) **43** 4. DOI: 10.1111/jfpe.13333
89. Savadkoohi S, Hoogenkamp H, Shamsi K, Farahnaky A. **Color, sensory and textural attributes of beef frankfurter, beef ham and meat-free sausage containing tomato pomace**. *Meat Science* (2014) **97** 410-418. DOI: 10.1016/j.meatsci.2014.03.017
90. Serdaroğlu M, Kavuşan HS, İpek G, Öztürk B. **Evaluation of the quality of beef patties formulated with dried pumpkin pulp and seed**. *Korean Journal for Food Science of Animal Resources* (2018) **38** 1-13. DOI: 10.5851/kosfa.2018.38.1.001
91. Shan B, Cai YZ, Brooks JD, Corke H. **Antibacterial and antioxidant effects of five spice and herb extracts as natural preservatives of raw pork**. *Journal of the Science of Food and Agriculture* (2009) **89** 1879-1885. DOI: 10.1002/jsfa.3667
92. Shan LC, De Brún A, Henchion M, Li C, Murrin C, Wall PG, Monahan FJ. **Consumer evaluations of processed meat products reformulated to be healthier—a conjoint analysis study**. *Meat Science* (2017) **131** 82-89. DOI: 10.1016/j.meatsci.2017.04.239
93. Sharma S, Sheehy T, Kolonel LN. **Contribution of meat to vitamin B**. *Journal of Human Nutrition and Dietetics: The Official Journal of the British Dietetic Association* (2013) **26** 156-168. DOI: 10.1111/jhn.12035
94. Sharma A, Yadav B, Ritika BY. **Resistant starch: physiological roles and food applications**. *Food Reviews International* (2008) **24** 193-234. DOI: 10.1080/87559120801926237
95. Shay J, Elbaz HA, Lee I, Zielske SP, Malek MH, Hüttemann M. **Molecular mechanisms and therapeutic effects of (−)-Epicatechin and other polyphenols in cancer, inflammation, diabetes, and neurodegeneration**. *Oxidative Medicine and Cellular Longevity* (2015) **2015** 1-13. DOI: 10.1155/2015/181260
96. Slavin J. **Fiber and prebiotics: mechanisms and health benefits**. *Nutrients* (2013) **5** 1417-1435. DOI: 10.3390/nu5041417
97. Soliman GA. **Dietary fiber, atherosclerosis, and cardiovascular disease**. *Nutrients* (2019) **11** 1155. DOI: 10.3390/nu11051155
98. Stephen AM, Champ MMJ, Cloran SJ, Fleith M, van Lieshout L, Mejborn H, Burley VJ. **Dietary fibre in Europe: current state of knowledge on definitions, sources, recommendations, intakes and relationships to health**. *Nutrition Research Reviews* (2017) **30** 149-190. DOI: 10.1017/S095442241700004X
99. Stevenson L, Phillips F, O’sullivan K, Walton J. **Wheat bran: its composition and benefits to health, a European perspective**. *International Journal of Food Sciences and Nutrition* (2012) **63** 1001-1013. DOI: 10.3109/09637486.2012.687366
100. Sun W, Zhou F, Zhao M. **Cantonese sausage, processing, storage and composition**. *Processing and Impact on Active Components in Food* (2015) 293-300
101. Thøgersen R, Castro-Mejía JL, Sundekilde UK, Hansen LH, Hansen AK, Nielsen DS, Bertram HC. **Ingestion of an inulin-enriched pork sausage product positively modulates the gut microbiome and metabolome of healthy rats**. *Molecular Nutrition & Food Research* (2018) **62** 1800608. DOI: 10.1002/mnfr.201800608
102. Thompson SV, Hannon BA, An R, Holscher HD. **Effects of isolated soluble fiber supplementation on body weight, glycemia, and insulinemia in adults with overweight and obesity: a systematic review and meta-analysis of randomized controlled trials**. *The American Journal of Clinical Nutrition* (2017) **106** 1514-1528. DOI: 10.3945/ajcn.117.163246
103. Ur Rahman U, Sahar A, Ishaq A, Aadil RM, Zahoor T, Ahmad MH. **Advanced meat preservation methods: a mini review**. *Journal of Food Safety* (2018) **38**. DOI: 10.1111/jfs.12467
104. **US department of health and human services. Dietary guidelines for Americans**. (2005)
105. Valsta LM, Tapanainen H, Männistö S. **Meat fats in nutrition**. *Meat Science* (2005) **70** 525-530. DOI: 10.1016/j.meatsci.2004.12.016
106. Vural Z, Avery A, Kalogiros DI, Coneyworth LJ, Welham SJM. **Trace mineral intake and deficiencies in older adults living in the community and institutions: a systematic review**. *Nutrients* (2020) **12** 1072. DOI: 10.3390/nu12041072
107. Wichert B, Schuster S, Hofmann M, Dobenecker B, Kienzle E. **Influence of different cellulose types on feces quality of dogs**. *The Journal of Nutrition* (2002) **132** 1728S-1729S. DOI: 10.1093/jn/132.6.1728S
108. Xiong RG, Zhou DD, Wu SX, Huang SY, Saimaiti A, Yang ZJ, Shang A, Zhao CN, Gan RY, Li HB. **Health benefits and side effects of short-chain fatty acids**. *Foods* (2022) **11** 2863. DOI: 10.3390/foods11182863
109. Yang J, Wang HP, Zhou L, Xu CF. **Effect of dietary fiber on constipation: a meta analysis**. *World Journal of Gastroenterology* (2012) **18** 7378-7383. DOI: 10.3748/wjg.v18.i48.7378
110. Yeung CK, Huang SC. **Effects of food proteins on sensory and physico-chemical properties of emulsified pork meatballs**. *Journal of Food and Nutrition Research* (2017) **6** 8-12. DOI: 10.12691/jfnr-6-1-2
111. Younis K, Ahmad S, Malik MA. **Mosambi peel powder incorporation in meat products: effect on physicochemical properties and shelf life stability**. *Applied Food Research* (2021) **1** 100015. DOI: 10.1016/j.afres.2021.100015
112. Younis K, Ahmad S, Osama K, Malik MA. **Optimization of de-bittering process of mosambi (**. *Journal of Food Process Engineering* (2019) **42** 133. DOI: 10.1111/jfpe.13185
113. Zając M, Guzik P, Kulawik P, Tkaczewska J, Florkiewicz A, Migdał W. **The quality of pork loaves with the addition of hemp seeds, de-hulled hemp seeds, hemp protein and hemp flour**. *LWT* (2019) **105** 190-199. DOI: 10.1016/j.lwt.2019.02.013
114. Zaki E. **Impact of adding chia seeds (Salvia hispanica) on the quality properties of camel burger “Camburger” during cold storage**. *International Journal of Current Microbiology and Applied Sciences* (2018) **7** 1356-1363. DOI: 10.20546/ijcmas.2018.703.162
115. Zhao L, Zhang F, Ding X, Wu G, Lam YY, Wang X, Fu H, Xue X, Lu C, Ma J, Yu L, Xu C, Ren Z, Xu Y, Xu S, Shen H, Zhu X, Shi Y, Shen Q, Dong W, Liu R, Ling Y, Zeng Y, Wang X, Zhang Q, Wang J, Wang L, Wu Y, Zeng B, Wei H, Zhang M, Peng Y, Zhang C. **Gut bacteria selectively promoted by dietary fibers alleviate type 2 diabetes**. *Science* (2018) **359** 1151-1156. DOI: 10.1126/science.aao5774
|
---
title: Roles of RNA m6A modification in nonalcoholic fatty liver disease
authors:
- Jian Tan
- Yue-fan Wang
- Zhi-hui Dai
- Hao-zan Yin
- Chen-yang Mu
- Si-jie Wang
- Fu Yang
journal: Hepatology Communications
year: 2023
pmcid: PMC9988276
doi: 10.1097/HC9.0000000000000046
license: CC BY 4.0
---
# Roles of RNA m6A modification in nonalcoholic fatty liver disease
## Abstract
NAFLD is a series of liver disorders, and it has become the most prevalent hepatic disease to date. However, there are no approved and effective pharmaceuticals for NAFLD owing to a poor understanding of its pathological mechanisms. While emerging studies have demonstrated that m6A modification is highly associated with NAFLD. In this review, we summarize the general profile of NAFLD and m6A modification, and the role of m6A regulators including erasers, writers, and readers in NAFLD. Finally, we also highlight the clinical significance of m6A in NAFLD.
## INTRODUCTION
NAFLD, a spectrum of liver disorders extending from liver steatosis to NASH, is characterized by liver insulin resistance (IR) and abnormal glucose and lipid metabolism.1–3 Presently, NAFLD has become the most prevalent hepatic disease which affects ∼$25\%$ of people worldwide.4 However, there are no approved pharmacotherapies for NAFLD as the pathogenic mechanisms underlying NAFLD are still poorly understood so far.5 Therefore, it is imperative to further explore the mechanisms of NAFLD and develop novel therapeutic targets for it.
N6-methyladenosine (m6A) is one of the most abundant and essential posttranscriptional modifications in eukaryotic cells.6 Increasing evidence has demonstrated that m6A is involved in a variety of human diseases, comprising NAFLD,7 azoospermia,8 and heart failure,9 especially in human cancers.10–12 Numerous studies have recently shown that m6A plays an important role in the onset and progression of NAFLD by regulating glycolipid metabolism, IR, and chronic inflammation, implying that m6A modification may be a potential therapeutic target for NAFLD.13,14 Consequently, it is critical to investigate the mechanisms as well as the aberrant m6A modification of NAFLD to develop novel therapeutic targets and prognostic markers for NAFLD. Hence, we systematically summarized the general profile of NAFLD and m6A modification, and recent progress in understanding the roles of m6A modulators (writers, erasers, and readers) in NAFLD in this review. Finally, we also highlighted the clinical implications of m6A modification in NAFLD.
## AN OVERVIEW OF NAFLD
NAFLD is generally defined as steatosis of >$5\%$ of hepatocytes that is not caused by alcohol consumption and other specific liver damage factors.15 Factually, NAFLD covers a wide range of pathologies from a benign fatty liver phenotype (steatosis or excessive lipid deposition in hepatocytes) to a severe form called NASH. During the progression of NAFLD, there are a series of pathological evolutions including sustained liver inflammation, hepatocyte death, liver fibrosis, liver cirrhosis, and even liver cancer occur.16 *It is* well known that NAFLD is firmly associated with metabolic dysregulation involving de novo lipogenesis, fatty acid (FA) uptake, FA oxidation, and triglycerides (TGs) export.17–19 Given its close connection to metabolic disorders, some experts even argue that NAFLD should be replaced by metabolic-associated fatty liver disease.20,21 Nonetheless, because the definition of NAFLD cannot be completely covered by that of metabolic-associated fatty liver disease, NAFLD will be used to avoid unnecessary divergences.22 According to recent research, the prevalence of NAFLD is $25\%$ to $30\%$ in the general population and can reach $60\%$ in obese people.23,24 *It is* estimated that by 2030, >300 million people in China, >100 million in the US, and 15 million to 20 million in the major European countries will have NAFLD by 2030.25 In the near future, NAFLD will cause enormous economic losses, but there are currently no effective targeted therapies due to a lack of understanding of its mechanisms.26 *As a* result, it is critical to investigate the mechanisms and specific pathogenesis of NAFLD to develop novel treatment strategies and improve prognosis.27
## AN OVERVIEW OF m6A MODIFICATION
m6A modification was first identified in 1974 in poly(A) RNA fractions and has been found in a variety of eukaryotic RNAs, including mRNAs, transfer RNAs, ribosomal RNAs, circular RNAs, microRNAs, and long noncoding RNAs.28,29 In nature, the term “m6A” refers to the methyl group transfer to the N6 position of adenine, which frequently takes place at the conserved sequence DRACH (D = G/A/U, R = G/A, H = A/U/C) and enriches in stop codons, 3′ untranslated regions, and long introns.30–32 It has been reported that each mRNA in mammals contains 3 to 5 m6A modifications.33 The m6A modification process is dynamic and reversible, with methyltransferases (also known as “writers”) assembling, demethylases (also known as “erasers”) removing, and m6A-binding proteins (also known as “readers”) recognizing and binding (Figure 1).34,35
**Figure 1:** *The biological process of m6A modification. m6A modification is catalyzed by methyltransferases containing METTL3, METTL5, METTL14, METTL16, WTAP, VIRMA, RBM15/15B, ZC3H13, ZC3HC4, and HAKAI. Whereas m6A modification is eliminated by demethylases containing FTO, ALKBH5, ALKBH3, and ALKBH1. In addition, m6A modification can be recognized and bound by m6A readers containing YTHDF1–3, YTHDC1–2, IGF2BP1–3, eIF3, PRRC2A, hnRNPG, hnRNPC, and hnRNPA2B1. IGF2BP1/2/3, PRRC2A, and FMRP all play a role in controlling the stability of m6A-modified RNAs. Abbreviations: ALKBH1/3/5, alkB homolog 1/3/5; eIF3, eukaryotic translation initiation factor 3; FMRP, fragile-X mental retardation protein; FTO, fat mass and obesity–associated protein; hnRNPs, heterogeneous nuclear ribonucleoproteins; IGF2BP, insulin-like growth factor 2 mRNA binding protein; m6A, N6-methyladenosine; METTL3/5/14/16, methyltransferase like-3/5/14/16; MTTP, microsomal triglyceride transfer protein; PRRC2A, proline-rich coiled-coil 2A; RBM15/15B, RNA binding motif protein 15/15B; VIRMA, Vir-like m6A methyltransferase associated; WTAP, WT1-associated protein; YTHDC1–2, YTHDF1–3,YTH N6-methyladenosine RNA binding protein 1–3; ZC3H13, zinc finger CCCH-type containing 13.*
Increasing evidence suggests that m6A regulates the expression of target genes to influence various eukaryotic physiological and pathological processes such as self-renewal, invasion, and proliferation. It has also been well established that m6A modification influences gene expression by interfering with RNA metabolism processes such as splicing, translocation, exporting, translation, stability, and decay.36–38 Congruously, a growing amount of research indicate that dysregulated m6A modification causes aberrant metabolism of RNA related to NAFLD,39 which greatly promotes the development of NAFLD. Therefore, it is essential to investigate associated m6A modification to comprehend the mechanisms of NAFLD.
## ROLES OF m6A ERASERS IN NAFLD
Fat mass and obesity–associated protein (FTO), alkB homolog 5 (ALKBH5), ALKBH3, and ALKBH1 are examples of m6A erasers that are also known as demethylases.40–42 All of the demethylases listed above are members of the ALKB dioxygenase family, which consists of nonheme Fe(II)/-ketoglutarate-dependent dioxygenases.43 The first known demethylase, FTO, relates to NAFLD most, localizes in nuclear speckles and catalyzes the demethylation of RNA.44 There are 3 steps during the FTO-catalyzed demethylation process: [1] oxidase m6A to form N6-hydroxymethyladenosine (hm6A); [2] convert hm6A to N6-formadenosine (f6A); [3] convert f6A to adenosine.45 In addition, a recent study showed that FTO can mediate the demethylation of N6-2’O-dimethyladenosine (m6Am), which is located close to the 5′-cap of small nuclear RNA.46 Numerous studies have recently revealed that FTO is significantly overexpressed in the livers of NAFLD patients and animal models, indicating that FTO may play an important role in NAFLD.47 It has been well-documented that NAFLD is characterized by excessive lipid accumulation in the liver.23 Accordingly, a recent study found that overexpression of FTO reduces the m6A modification of mRNAs involved in lipid metabolism, such as fatty acid synthase, stearoyl-CoA desaturase (SCD), and monoacylglycerol O-acyltransferase 1, and thus promotes their expression, aggravating lipid accumulation in the liver.48 Similarly, Chen et al.49 found that FTO facilitates lipid accumulation in the liver by increasing nuclear translocation and maturation of sterol regulatory element-binding protein-1c (SREBP1c) and promoting the transcription of the cell death-inducing DFFA-like effector C (CIDEC). Furthermore, Hu et al.50 demonstrated that FTO transactivation and m6A demethylation on mRNA of lipogenic genes induced lipogenic gene activation and lipid accumulation during NAFLD and were mediated by glucocorticoid receptor. In addition, Wei et al.4 discovered that FTO facilitates lipid accumulation in NAFLD by suppressing the expression of the peroxisome proliferator-activated receptor α (PPARα). In addition to lipid accumulation, FTO can cause IR in the liver by increasing the expression of gluconeogenic genes such as phosphoenolpyruvate carboxykinase 1 and glucose-6-phosphatase, thereby promoting the progression of NAFLD.51 Further to that, overexpression of FTO reduces TG transport in the liver by decreasing the expression levels of microsomal triglyceride transfer protein, apolipoprotein B, and hepatic lipase C. This results in intracellular TG accumulation by promoting FA synthesis and inhibiting TG hydrolysis during NAFLD.52 *To sum* up, FTO promotes the development of NAFLD by modulating the m6A modification of multiple metabolism-related genes and may represent a potential therapeutic target for NAFLD.
## ROLES OF m6A WRITERS IN NAFLD
m6A writers, also referred to as m6A methyltransferases, catalyze the transfer of methyl groups from S-adenosylmethionine to the nitrogen atom (N) at the sixth position of adenine in RNA. These enzymes include methyltransferase like-3 (METTL3), METTL5, METTL14, METTL16, WT1-associated protein (WTAP), Vir-like m6A methyltransferase associated (also known as KIAA1429), RNA binding motif protein $\frac{15}{15}$B (RBM$\frac{15}{15}$B), zinc finger CCCH-type containing 13 (ZC3H13), and HAKAI (also known as CBLL1, a RING-finger type E3 ubiquitin ligase).45,53–56 Among these m6A methyltransferases, METTL3 and METTL14 had the most pronounced effect on NAFLD. The core subunit of the m6A-METTL complex, METTL3, contains the sole methyltransferase catalytic domain and interacts with METTL14 to form the m6A-METTL complex–named heterodimer METTL3/METTL14. METTL14 provides functional support for METTL3 and associates with it to form m6A-METTL complex.54,57 It has been observed that overexpression of METTL3 inhibits autophagosome-lysosome fusion and hinders autophagosome degradation by targeting Rubicon mRNA to enhance its expression and, as a result, reduces hepatic lipid clearance in NAFLD.58 Nonetheless, Qin et al.59 discovered that myeloid METTL3 loss blunts the pathogenesis of NAFLD by lowering the m6A alteration of DNA damage inducible transcript 4 (DDIT4), which plays a crucial role in the regulation of macrophage activation via the reduction of mTOR and NF-κB signaling pathway activity, and raising its expression, therefore, they summarized that METTL3 played an important role in accelerating NAFLD via increasing the m6A level of DDIT4. In addition, Xie et al.14 demonstrated that elevated METTL3 boosted hepatic IR and stimulated lipid synthesis via N6-methylating fatty acid synthase mRNA and increasing its total mRNA level. In contrast, Li and colleagues found that hepatocyte-specific deletion of METTL3 promoted NAFL-to-NASH progression by enhancing CD36-mediated hepatic-free FA uptake and CCL2-induced inflammation, whereas hepatic upregulation of METTL3 protected against NASH progression by suppressing the expression of CD36 and CCL2. Mechanistically, METTL3 directly bound to the promoters of the CD36 and CCL2 genes and recruited HDAC$\frac{1}{2}$, which caused deacetylation of H3K9 and H3K27 in their promoters, consequently the transcription of CD36 and CCL2 was suppressed. Thus, METTL3 was identified as a previously unrecognized suppressor of the NAFL-to-NASH transition.60 This study revealed that the function of METTL3 in NAFLD not only in enhancing the m6A modification pathway but also in inducing histone deacetylation. Together, these findings show that METTL3’s effects on NAFLD are still controversial, and additional research is needed to understand the biological functions and mechanisms of METTL3 in NAFLD. In addition to METTL3, Yang et al.61 discovered that the expression of METTL14 is raised in NAFLD mice, and that METTL14 enhances de novo FA synthesis and lipid accumulation via raising the protein level of ATP citrate lyase and SCD1 by stabilizing m6A modification on their mRNA, consequently promoting the progression of NAFLD. What’s more, Qiu et al.62 have identified that arsenic-induced NOD-like receptor protein 3 (NLRP3) inflammasome activation contributes to hepatic IR induction during arsenic-induced NAFLD, while NLRP3 mRNA stability was strengthened by METTL14-mediated m6A modification.
In conclusion, these data indicate that METTL3 and METTL14 play a significant role in NAFLD via influencing m6A modification. However, as there are disagreements and ambiguous aspects, additional research is required to understand the processes and therapeutical potentials of m6A methyltransferases in NAFLD.
## ROLES OF m6A READERS IN NAFLD
m6A readers are indispensable m6A-binding proteins that recognize and bind to specific RNA sequences and regulate numerous RNA life cycle activities.63 These m6A readers can be categorized into 3 groups based on the mechanism of m6A recognition: direct readers, m6A switch readers, and indirect readers.35 Direct readers are composed of eukaryotic translation initiation factor 3 (eIF3), YTH domain-containing proteins, and proline-rich coiled-coil 2A. m6A switch readers consist of heterogeneous nuclear ribonucleoproteins (hnRNPs) comprising hnRNPG, hnRNPC, and hnRNPA2B1 and insulin-like growth factor 2 mRNA binding proteins (IGF2BPs) including IGF2BP1, IGF2BP2, and IGF2BP3. And indirect reader involves fragile-X mental retardation protein.48,64,65 Among m6A readers, both YTH domain-containing proteins and IGF2BP2 have been shown to have an effect on NAFLD to varied degrees. YTH domain-containing proteins possess a conserved YT521-B homology domain at the C-terminus, which consists of YTHDF1–3 and YTHDC1–2.66 The YTH domain may identify m6A methylation and bind to m6A-modified RNA in RRACH common sequences.67 YTHDF1, YTHDF2, and YTHDF3 are mostly localized to the cytoplasm, whereas YTHDC1 is primarily localized to the nucleus and YTHDC2 is present in both the nucleus and cytoplasm.48,65,68 Functionally, YTHDF1 can attach to the m6A site near the terminal sequences and cooperate with eIF3 to increase the effectiveness of RNA translation. YTHDF2, meanwhile, regulates the degradation of m6A-dependent RNA, while YTHDF3 supports YTHDF1 and YTHDF2 in a synergistic manner.64,68,69 YTHDC1 can mediate the export of mRNA from the nucleus and recruit the mRNA splicing factors SRSF3 and SRSF10 to regulate RNA splicing,70 while YTHDC2 enhances the translational efficiency of target mRNA by binding to particular m6A sequences.71 YTHDC2, which binds to the mRNA of the lipogenic genes SREBP1c, fatty acid synthase, SCD1, and acetyl-CoA carboxylase 1 (ACC1) and thus decreases the stability and expression of these genes, has recently been shown by Zhou et al.7 to be strikingly downregulated in the livers of obese mice and NAFLD patients. As a result, they concluded that downregulation of YTHDC2 results in overexpression of lipogenic genes and accumulation of excessive triglycerides (TGs) in the liver in an m6A-dependent manner, which facilitates the advancement of NAFLD. In contrast, Yang et al.61 claimed that they had discovered an increase of eIF3G and YTHDC2 in NAFLD mice, but they did not go into detail about the underlying mechanisms and outcomes. In addition, Zhong et al.72 discovered that YTHDF2 binds to PPARα mRNA to increase its mRNA stability and expression, and then promotes lipid accumulation during hepatic steatosis caused by circadian rhythm disruption. In addition, Peng et al.58 demonstrated that overexpression of YTHDF1 can decrease autophagic flux and enhance lipid droplet accumulation in NAFLD by binding to Rubicon mRNA to boost its stability, hence accelerating the progression of NAFLD. While IGF2BP2 has been reported to drive the progression of NASH through elevating hepatic iron deposition and increasing production of hepatic-free cholesterol.73,74 In contrast, global IGF2BP2 deficiency prevents mice from NAFLD by causing resistance to obesity and fatty liver, whereas hepatocyte-specific deletion of IGF2BP2 promotes moderate diet-induced fatty liver through impairing FA oxidation by increasing mRNA degradation of PPARα and carnitine palmitoyltransferase 1A which were supposed to be stabilized and bound by IGF2BP2.75 In conclusion, these results indicate that m6A readers play a crucial and intricate role in NAFLD, although the hidden mechanisms remain poorly known and require additional research.
## CLINICAL SIGNIFICANCE OF m6A IN NAFLD
Mounting evidence confirms that m6A regulators play a vital role in the development of NAFLD; hence, it is reasonable to target these regulators to develop effective therapeutic strategies for NAFLD. Its interesting to note that a recent study found that miR-627-5p could bind to the 3′ untranslated region of FTO and may reduce IR and abnormalities of glucose and lipid metabolism in NAFLD by suppressing the production of FTO.76 Similarly, Lim and colleagues showed that by restoring mitochondrial function and reducing endoplasmic reticulum stress, FTO knockdown can lessen palmitic acid–induced hepatocyte lipotoxicity.47,77 Furthermore, entacapone, a possible FTO inhibitor, has been shown to be beneficial in treating metabolic disorders such as NAFLD.78 Meclofenamic acid, another highly selective FTO inhibitor, has also been observed to diminish the accumulation of TG in liver cells, and may help to relieve NAFLD.79,80 In addition, Chen et al.81 have confirmed that the methyl donor betaine can be utilized to protect against NAFLD in an FTO-dependent manner. In particular, Lu et al.13 demonstrated that curcumin reduces lipopolysaccharide-induced liver damage and hepatic lipid metabolism disruption in NAFLD by boosting m6A RNA methylation, and Li et al.82 demonstrated that exenatide ameliorates hepatic steatosis by decreasing FTO gene expression and fat mass via the PI3K signaling pathway in NAFLD. Moreover, Xie and colleagues found that METTL3 knockdown decreased FASN mRNA levels, which in turn inhibited lipogenesis, leading them to believe that METTL3 would be a possible therapeutic target for treating NAFLD.14,59 However, there are no effective therapies that target METTL3 to treat NAFLD currently; therefore, additional research is required to create innovative medications that target it as well as other m6A modulators. Overall, it is possible to conclude that m6A regulators are prospective therapeutic targets for NAFLD.
## CONCLUSION AND FUTURE DIRECTIONS
NAFLD has gained increasing attention due to its high prevalence, lack of effective treatments, and poor clinical outcomes.3 *While numerous* studies have indicated that m6A modification modulators produce critical and complex effects in the development of NAFLD (Figure 2). Consequently, it is essential to understand these mechanisms and investigate their potential therapeutic value for NAFLD. Our present review comprehensively summarizes the profile of NAFLD and the latest understanding of the roles, mechanisms, and potential clinical significance of m6A in NAFLD. However, the specific functions and mechanisms of m6A regulators in NAFLD are complicated and still elusive. Notably, even the same regulator can exhibit opposite effects in NAFLD in different animal models. Take METTL3 as an example, Qin et al.59 delivered that myeloid METTL3 deficiency impedes the pathological progress of NAFLD via reducing leukocyte infiltration and hepatic damage. In contrast, Li et al.60 reported that hepatocyte-specific METTL3 deletion drives NAFLD-to-NASH progression by enhancing CD36-mediated hepatic-free FA uptake and CCL2-induced inflammation. Furthermore, m6A regulators may function in cooperation in NAFLD. For instance, Inhibiting autophagic flux and aggravating lipid droplet buildup in NAFLD are caused by METTL3-mediated m6A alteration, which is dependent on YTHDF1 increasing Rubicon mRNA stability.58 Collectively, the involvement of m6A regulators in NAFLD has shown complexity and diversity (Table 1), therefore, more research is needed to completely understand their functional mechanisms and clinical significance in NAFLD.
**Figure 2:** *The role of m6A regulators in the development of NAFLD. In NAFLD, m6A regulators (FTO, METTL$\frac{3}{14}$, YTHDF$\frac{1}{2}$, YTHDC2) increase lipid accumulation, IR, and inflammation and decrease mitochondrial content by modulating the expression level of targeted RNAs associated with lipid metabolism including FASN, SCD, MOGAT1, SREBP1, CIDEC, PPARα, ACLY, SCD1, RUBICON, MTTP, APOB, and LIPC, and inflammation including DDIT4, CD36, CCL2, and NLRP3. In NASH, sustained lipid accumulation, inflammation, and IR caused by m6A regulators induce lipotoxicity, ER stress, mitochondrial injury, and oxidative stress, leading to steatohepatitis, cell death, and fibrosis. Abbreviations: ACC1, acetyl-CoA carboxylase 1; ACLY, ATP citrate lyase; APOB, apolipoprotein B; CIDEC, cell death-inducing DFFA-like effector; DDIT4, DNA damage inducible transcript 4; ER, endoplasmic reticulum; FASN, fatty acid synthase; FTO, fat mass and obesity–associated protein; G6PC, glucose-6-phosphatase; IGFBP2, insulin-like growth factor 2 mRNA binding protein; IR, insulin resistance; LD, lipid droplet; LIPC, hepatic lipase C; m6A, N6-methyladenosine; METTL3, methyltransferase like-3; MOGAT1, monoacylglycerol O-acyltransferase 1; MTTP, microsomal triglyceride transfer protein; NLRP3, NOD-like receptor protein 3; PCK1, phosphoenolpyruvate carboxykinase; PPARα, peroxisome proliferator-activated receptor α; SCD, stearoyl-CoA desaturase; SREBP1c, sterol regulatory element-binding protein1c; TG, triglyceride; YTHDC1–2, YTH domain-containing 1–2; YTHDF1–3,YTH N6-methyladenosine RNA binding protein 1–3.* TABLE_PLACEHOLDER:Table 1 Fortunately, the crucial roles of m6A regulators found in NAFLD indicate that they may be promising therapeutic targets. Presently, various FTO-targeting drugs, including entacapone, meclofenamic acid, betaine, exenatide, and curcumin, have been found to be promising in the treatment of NAFLD. However, the available m6A-targeted therapies for NAFLD only target FTO; as a result, innovative therapeutics targeting additional m6A modulators should be investigated in the future. FTO is also overexpressed in NAFLD patients, which could be a sign of a poor clinical outcome, although univariate Cox regression must be done to formally recognize FTO as an independent prognostic factor for NAFLD patients.83 Importantly, the gene expression profile and expression level of m6A regulators, as well as clinical prognostic information for NAFLD patients, should be acquired to determine the prognostic significance of m6A regulators in NAFLD.
## FUNDING INFORMATION
This work was supported by the National Natural Science Foundation of China (Grant number 81972657 and 81672345). The figures of this review was created by Figdraw.
## CONFLICT OF INTEREST
Nothing to report.
## References
1. Watt MJ, Miotto PM, De Nardo W, Montgomery MK. **The liver as an endocrine organ-linking NAFLD and insulin resistance**. *Endocr Rev* (2019) **40** 1367-1393. PMID: 31098621
2. Chen L, Chen XW, Huang X, Song BL, Wang Y, Wang Y. **Regulation of glucose and lipid metabolism in health and disease**. *Sci China Life Sci* (2019) **62** 1420-1458. PMID: 31686320
3. Zhou J, Zhou F, Wang W, Zhang XJ, Ji YX, Zhang P. **Epidemiological features of NAFLD From 1999 to 2018 in China**. *Hepatology* (2020) **71** 1851-1864. PMID: 32012320
4. Wei X, Zhang J, Tang M, Wang X, Fan N, Peng Y. **Fat mass and obesity-associated protein promotes liver steatosis by targeting PPARalpha**. *Lipids Health Dis* (2022) **21** 29. PMID: 35282837
5. Stefan N, Häring HU, Cusi K. **Non-alcoholic fatty liver disease: causes, diagnosis, cardiometabolic consequences, and treatment strategies**. *Lancet Diabetes Endocrinol* (2019) **7** 313-324. PMID: 30174213
6. Deng L, Deng W, Fan S, Chen M, Qi M, Lyu W. **m6A modification: recent advances, anticancer targeted drug discovery and beyond**. *Mol Cancer* (2022) **21** 52. PMID: 35164788
7. Zhou B, Liu C, Xu L, Yuan Y, Zhao J, Zhao W. **N(6)-methyladenosine reader protein YT521-B homology domain-containing 2 suppresses liver steatosis by regulation of mRNA stability of lipogenic genes**. *Hepatology* (2021) **73** 91-103. PMID: 32150756
8. Chen Y, Wang J, Xu D, Xiang Z, Ding J, Yang X. **mA mRNA methylation regulates testosterone synthesis through modulating autophagy in Leydig cells**. *Autophagy* (2021) **17** 457-475. PMID: 31983283
9. Berulava T, Buchholz E, Elerdashvili V, Pena T, Islam MR, Lbik D. **Changes in m6A RNA methylation contribute to heart failure progression by modulating translation**. *Eur J Heart Fail* (2020) **22** 54-66. PMID: 31849158
10. Liu L, Li H, Hu D, Wang Y, Shao W, Zhong J. **Insights into N6-methyladenosine and programmed cell death in cancer**. *Mol Cancer* (2022) **21** 32. PMID: 35090469
11. Liu T, Wei Q, Jin J, Luo Q, Liu Y, Yang Y. **The m6A reader YTHDF1 promotes ovarian cancer progression via augmenting EIF3C translation**. *Nucleic Acids Res* (2020) **48** 3816-3831. PMID: 31996915
12. Ni W, Yao S, Zhou Y, Liu Y, Huang P, Zhou A. **Long noncoding RNA GAS5 inhibits progression of colorectal cancer by interacting with and triggering YAP phosphorylation and degradation and is negatively regulated by the m(6)A reader YTHDF3**. *Mol Cancer* (2019) **18** 143. PMID: 31619268
13. Lu N, Li X, Yu J, Li Y, Wang C, Zhang L. **Curcumin attenuates lipopolysaccharide-induced hepatic lipid metabolism disorder by modification of m(6) A RNA methylation in piglets**. *Lipids* (2018) **53** 53-63. PMID: 29488640
14. Xie W, Ma LL, Xu YQ, Wang BH, Li SM. **METTL3 inhibits hepatic insulin sensitivity via N6-methyladenosine modification of FASN mRNA and promoting fatty acid metabolism**. *Biochem Biophys Res Commun* (2019) **518** 120-126. PMID: 31405565
15. Huang T, Yu L, Pan H, Ma Z, Wu T, Zhang L. **Integrated transcriptomic and translatomic inquiry of the role of betaine on lipid metabolic dysregulation induced by a high-fat diet**. *Front Nutr* (2021) **8** 751436. PMID: 34708066
16. Ma J, Tan X, Kwon Y, Delgado ER, Zarnegar A, DeFrances MC. **A novel humanized model of NASH and its treatment with META4, a potent agonist of MET**. *Cell Mol Gastroenterol Hepatol* (2022) **13** 565-582. PMID: 34756982
17. Roeb E. **Diagnostic and therapy of nonalcoholic fatty liver disease: a narrative review**. *Visc Med* (2022) **38** 126-132. PMID: 35614896
18. Kulathunga K, Wakimoto A, Hiraishi Y, Yadav MK, Gentleman K, Warabi E. **Albino mice with the point mutation at the tyrosinase locus show high cholesterol diet-induced NASH susceptibility**. *Sci Rep* (2021) **11** 21827. PMID: 34750345
19. Cobbina E, Akhlaghi F. **Non-alcoholic fatty liver disease (NAFLD)—pathogenesis, classification, and effect on drug metabolizing enzymes and transporters**. *Drug Metab Rev* (2017) **49** 197-211. PMID: 28303724
20. Qu W, Ma T, Cai J, Zhang X, Zhang P, She Z. **Liver fibrosis and MAFLD: from molecular aspects to novel pharmacological strategies**. *Front Med* (2021) **8** 761538
21. Eslam M, Alkhouri N, Vajro P, Baumann U, Weiss R, Socha P. **Defining paediatric metabolic (dysfunction)-associated fatty liver disease: an international expert consensus statement**. *Lancet Gastroenterol Hepatol* (2021) **6** 864-873. PMID: 34364544
22. Huby T, Gautier E. **Immune cell-mediated features of non-alcoholic steatohepatitis**. *Nat Rev Immunol* (2021) **22** 429-443. PMID: 34741169
23. El-Kassas M, Cabezas J, Iruzubieta P, Zheng MH, Arab JP, Awad A. **Non-alcoholic fatty liver disease: current global burden**. *Semin Liver Dis* (2022) **42** 401-412. PMID: 35617968
24. Li L, Xu M, He C, Wang H, Hu Q. **Polystyrene nanoplastics potentiate the development of hepatic fibrosis in high fat diet fed mice**. *Environ Toxicol* (2022) **37** 362-372. PMID: 34755918
25. O’hara J, Finnegan A, Dhillon H, Ruiz-Casas L, Pedra G, Franks B. **Cost of non-alcoholic steatohepatitis in Europe and the USA: The GAIN Study**. *JHEP Rep* (2020) **2** 100142. PMID: 32775976
26. Schattenberg JM, Lazarus JV, Newsome PN, Serfaty L, Aghemo A, Augustin S. **Disease burden and economic impact of diagnosed non-alcoholic steatohepatitis in five European countries in 2018: a cost-of-illness analysis**. *Liver Int* (2021) **41** 1227-1242. PMID: 33590598
27. Daemen S, Gainullina A, Kalugotla G, He L, Chan MM, Beals JW. **Dynamic shifts in the composition of resident and recruited macrophages influence tissue remodeling in NASH**. *Cell Rep* (2021) **34** 108626. PMID: 33440159
28. Liu N, Pan T. **N6-methyladenosine–encoded epitranscriptomics**. *Nat Struct Mol Biol* (2016) **23** 98-102. PMID: 26840897
29. Shi H, Wei J, He C. **Where, when, and how: context-dependent functions of RNA methylation writers, readers, and erasers**. *Mol Cell* (2019) **74** 640-650. PMID: 31100245
30. Meyer KD, Saletore Y, Zumbo P, Elemento O, Mason CE, Jaffrey SR. **Comprehensive analysis of mRNA methylation reveals enrichment in 3′ UTRs and near stop codons**. *Cell* (2012) **149** 1635-1646. PMID: 22608085
31. Fu Y, Dominissini D, Rechavi G, He C. **Gene expression regulation mediated through reversible m6A RNA methylation**. *Nat Rev Genet* (2014) **15** 293-306. PMID: 24662220
32. Wang Y, Ge C, Yin H, Dai Z, Dong J, Ji M. **Dysregulated N6-methyladenosine (m(6)A) processing in hepatocellular carcinoma**. *Ann Hepatol* (2021) **25** 100538. PMID: 34555511
33. An Y, Duan H. **The role of m6A RNA methylation in cancer metabolism**. *Mol Cancer* (2022) **21** 14. PMID: 35022030
34. He L, Li H, Wu A, Peng Y, Shu G, Yin G. **Functions of N6-methyladenosine and its role in cancer**. *Mol Cancer* (2019) **18** 176. PMID: 31801551
35. Zaccara S, Ries RJ, Jaffrey SR. **Reading, writing and erasing mRNA methylation**. *Nat Rev Mol Cell Biol* (2019) **20** 608-624. PMID: 31520073
36. Liu Q, Gregory RI. **RNAmod: an integrated system for the annotation of mRNA modifications**. *Nucleic Acids Res* (2019) **47** W548-W555. PMID: 31147718
37. Jiang X, Liu B, Nie Z, Duan L, Xiong Q, Jin Z. **The role of m6A modification in the biological functions and diseases**. *Signal Transduct Target Ther* (2021) **6** 74. PMID: 33611339
38. Roignant JY, Soller M. **m(6)A in mRNA: an ancient mechanism for fine-tuning gene expression**. *Trends Genet* (2017) **33** 380-390. PMID: 28499622
39. Luo Z, Zhang Z, Tai L, Zhang L, Sun Z, Zhou L. **Comprehensive analysis of differences of N-methyladenosine RNA methylomes between high-fat-fed and normal mouse livers**. *Epigenomics* (2019) **11** 1267-1282. PMID: 31290331
40. Ueda Y, Ooshio I, Fusamae Y, Kitae K, Kawaguchi M, Jingushi K. **AlkB homolog 3-mediated tRNA demethylation promotes protein synthesis in cancer cells**. *Sci Rep* (2017) **7** 42271. PMID: 28205560
41. Li Y, Su R, Deng X, Chen Y, Chen J. **FTO in cancer: functions, molecular mechanisms, and therapeutic implications**. *Trends Cancer* (2022) **8** 598-614. PMID: 35346615
42. Kaur S, Tam NY, Mcdonough MA, Schofield CJ, Aik WS. **Mechanisms of substrate recognition and N6-methyladenosine demethylation revealed by crystal structures of ALKBH5-RNA complexes**. *Nucleic Acids Res* (2022) **50** 4148-4160. PMID: 35333330
43. Zheng G, Dahl JA, Niu Y, Fedorcsak P, Huang CM, Li CJ. **ALKBH5 is a mammalian RNA demethylase that impacts RNA metabolism and mouse fertility**. *Mol Cell* (2013) **49** 18-29. PMID: 23177736
44. Jia G, Fu Y, Zhao X, Dai Q, Zheng G, Yang Y. **N6-methyladenosine in nuclear RNA is a major substrate of the obesity-associated FTO**. *Nat Chem Biol* (2011) **7** 885-887. PMID: 22002720
45. Zhang Y, Chen W, Zheng X, Guo Y, Cao J, Zhang Y. **Regulatory role and mechanism of m(6)A RNA modification in human metabolic diseases**. *Mol Ther Oncolytics* (2021) **22** 52-63. PMID: 34485686
46. Mauer J, Sindelar M, Despic V, Guez T, Hawley BR, Vasseur JJ. **FTO controls reversible m(6)Am RNA methylation during snRNA biogenesis**. *Nat Chem Biol* (2019) **15** 340-347. PMID: 30778204
47. Lim A, Zhou J, Sinha RA, Singh BK, Ghosh S, Lim KH. **Hepatic FTO expression is increased in NASH and its silencing attenuates palmitic acid-induced lipotoxicity**. *Biochem Biophys Res Commun* (2016) **479** 476-481. PMID: 27651333
48. Li Y, Wang J, Huang C, Shen M, Zhan H, Xu K. **RNA N6-methyladenosine: a promising molecular target in metabolic diseases**. *Cell Biosci* (2020) **10** 19. PMID: 32110378
49. Chen A, Chen X, Cheng S, Shu L, Yan M, Yao L. **FTO promotes SREBP1c maturation and enhances CIDEC transcription during lipid accumulation in HepG2 cells**. *Biochim Biophys Acta Mol Cell Biol Lipids* (2018) **1863** 538-548. PMID: 29486327
50. Hu Y, Feng Y, Zhang L, Jia Y, Cai D, Qian SB. **GR-mediated FTO transactivation induces lipid accumulation in hepatocytes via demethylation of m6A on lipogenic mRNAs**. *RNA Biol* (2020) **17** 930-942. PMID: 32116145
51. Mizuno T. **Fat mass and obesity associated (FTO) gene and hepatic glucose and lipid metabolism**. *Nutrients* (2018) **10** 1600. PMID: 30388740
52. Kang H, Zhang Z, Yu L, Li Y, Liang M, Zhou L. **FTO reduces mitochondria and promotes hepatic fat accumulation through RNA demethylation**. *J Cell Biochem* (2018) **119** 5676-5685. PMID: 29384213
53. Pan X, Huang C, Li J. **The emerging roles of m(6)A modification in liver carcinogenesis**. *Int J Biol Sci* (2021) **17** 271-284. PMID: 33390849
54. Liu J, Yue Y, Han D, Wang X, Fu Y, Zhang L. **A METTL3-METTL14 complex mediates mammalian nuclear RNA N6-adenosine methylation**. *Nat Chem Biol* (2014) **10** 93-95. PMID: 24316715
55. Knuckles P, Lence T, Haussmann IU, Jacob D, Kreim N, Carl SH. **Zc3h13/Flacc is required for adenosine methylation by bridging the mRNA-binding factor Rbm15/Spenito to the m(6)A machinery component Wtap/Fl(2)d**. *Genes Dev* (2018) **32** 415-429. PMID: 29535189
56. Bawankar P, Lence T, Paolantoni C, Haussmann IU, Kazlauskiene M, Jacob D. **HAKAI is required for stabilization of core components of the m(6)A mRNA methylation machinery**. *Nat Commun* (2021) **12** 3778. PMID: 34145251
57. Wang P, Doxtader KA, Nam Y. **Structural basis for cooperative function of Mettl3 and Mettl14 methyltransferases**. *Mol Cell* (2016) **63** 306-317. PMID: 27373337
58. Peng Z, Gong Y, Wang X, He W, Wu L, Zhang L. **METTL3-m(6)A-Rubicon axis inhibits autophagy in nonalcoholic fatty liver disease**. *Mol Ther* (2022) **30** 932-946. PMID: 34547464
59. Qin Y, Li B, Arumugam S, Lu Q, Mankash SM, Li J. **m(6)A mRNA methylation-directed myeloid cell activation controls progression of NAFLD and obesity**. *Cell Rep* (2021) **37** 109968. PMID: 34758326
60. Li X, Yuan B, Lu M, Wang Y, Ding N, Liu C. **The methyltransferase METTL3 negatively regulates nonalcoholic steatohepatitis (NASH) progression**. *Nat Commun* (2021) **12** 7213. PMID: 34893641
61. Yang Y, Cai J, Yang X, Wang K, Sun K, Yang Z. **Dysregulated m6A modification promotes lipogenesis and development of non-alcoholic fatty liver disease and hepatocellular carcinoma**. *Mol Ther* (2022) **30** 2342-2353. PMID: 35192934
62. Qiu T, Wu C, Yao X, Han Q, Wang N, Yuan W. **AS3MT facilitates NLRP3 inflammasome activation by m(6)A modification during arsenic-induced hepatic insulin resistance**. *Cell Biol Toxicol* (2022) 1-17. DOI: 10.1007/s10565-022-09703-7
63. Meyer KD, Jaffrey SR. **Rethinking mA readers, writers, and erasers**. *Annu Rev Cell Dev Biol* (2017) **33** 319-342. PMID: 28759256
64. Wang X, Lu Z, Gomez A, Hon GC, Yue Y, Han D. **N6-methyladenosine-dependent regulation of messenger RNA stability**. *Nature* (2014) **505** 117-120. PMID: 24284625
65. Wang Y, Zhang Y, Du Y, Zhou M, Hu Y, Zhang S. **Emerging roles of N6-methyladenosine (m(6)A) modification in breast cancer**. *Cell Biosci* (2020) **10** 136. PMID: 33292526
66. Zaccara S, Jaffrey SR. **A unified model for the function of YTHDF proteins in regulating m(6)A-modified mRNA**. *Cell* (2020) **181** 1582-95.e18. PMID: 32492408
67. Theler D, Dominguez C, Blatter M, Boudet J, Allain FHT. **Solution structure of the YTH domain in complex with N6-methyladenosine RNA: a reader of methylated RNA**. *Nucleic Acids Res* (2014) **42** 13911-13919. PMID: 25389274
68. Wang X, Zhao BS, Roundtree IA, Lu Z, Han D, Ma H. **N(6)-methyladenosine modulates messenger RNA translation efficiency**. *Cell* (2015) **161** 1388-1399. PMID: 26046440
69. Shi H, Wang X, Lu Z, Zhao BS, Ma H, Hsu PJ. **YTHDF3 facilitates translation and decay of N(6)-methyladenosine-modified RNA**. *Cell Res* (2017) **27** 315-328. PMID: 28106072
70. Kasowitz SD, Ma J, Anderson SJ, Leu NA, Xu Y, Gregory BD. **Nuclear m6A reader YTHDC1 regulates alternative polyadenylation and splicing during mouse oocyte development**. *PLoS Genet* (2018) **14** e1007412. PMID: 29799838
71. Wojtas MN, Pandey RR, Mendel M, Homolka D, Sachidanandam R, Pillai RS. **Regulation of m(6)A Transcripts by the 3’→5’ RNA Helicase YTHDC2 Is Essential for a Successful Meiotic Program in the Mammalian Germline**. *Mol Cell* (2017) **68** 374-387e12. PMID: 29033321
72. Zhong X, Yu J, Frazier K, Weng X, Li Y, Cham CM. **Circadian clock regulation of hepatic lipid metabolism by modulation of m(6)A mRNA methylation**. *Cell Rep* (2018) **25** 1816-1828.e4. PMID: 30428350
73. Simon Y, Kessler SM, Bohle RM, Haybaeck J, Kiemer AK. **The insulin-like growth factor 2 (IGF2) mRNA-binding protein p62/IGF2BP2-2 as a promoter of NAFLD and HCC?**. *Gut* (2014) **63** 861-863. PMID: 24173291
74. Simon Y, Kessler SM, Gemperlein K, Bohle RM, Müller R, Haybaeck J. **Elevated free cholesterol in a p62 overexpression model of non-alcoholic steatohepatitis**. *World J Gastroenterol* (2014) **20** 17839-17850. PMID: 25548482
75. Regué L, Minichiello L, Avruch J, Dai N. **Liver-specific deletion of IGF2 mRNA binding protein-2/IMP2 reduces hepatic fatty acid oxidation and increases hepatic triglyceride accumulation**. *J Biol Chem* (2019) **294** 11944-11951. PMID: 31209109
76. Cheng L, Yu P, Li F, Jiang X, Jiao X, Shen Y. **Human umbilical cord-derived mesenchymal stem cell-exosomal miR-627-5p ameliorates non-alcoholic fatty liver disease by repressing FTO expression**. *Hum Cell* (2021) **34** 1697-1708. PMID: 34410623
77. Cusi K. **Role of obesity and lipotoxicity in the development of nonalcoholic steatohepatitis: pathophysiology and clinical implications**. *Gastroenterology* (2012) **142** 711-725.e6. PMID: 22326434
78. Peng S, Xiao W, Ju D, Sun B, Hou N, Liu Q. **Identification of entacapone as a chemical inhibitor of FTO mediating metabolic regulation through FOXO1**. *Sci Transl Med* (2019) **11** eaau7116. PMID: 30996080
79. Zhao Z, Meng J, Su R, Zhang J, Chen J, Ma X. **Epitranscriptomics in liver disease: basic concepts and therapeutic potential**. *J Hepatol* (2020) **73** 664-679. PMID: 32330603
80. Huang Y, Yan J, Li Q, Li J, Gong S, Zhou H. **Meclofenamic acid selectively inhibits FTO demethylation of m6A over ALKBH5**. *Nucleic Acids Res* (2015) **43** 373-384. PMID: 25452335
81. Chen J, Zhou X, Wu W, Wang X, Wang Y. **FTO-dependent function of N6-methyladenosine is involved in the hepatoprotective effects of betaine on adolescent mice**. *J Physiol Biochem* (2015) **71** 405-413. PMID: 26078098
82. Li S, Wang X, Zhang J, Li J, Liu X, Ma Y. **Exenatide ameliorates hepatic steatosis and attenuates fat mass and FTO gene expression through PI3K signaling pathway in nonalcoholic fatty liver disease**. *Braz J Med Biol Res* (2018) **51** e7299. PMID: 29924135
83. Wu X, Zhang X, Tao L, Dai X, Chen P. **Prognostic value of an m6A RNA methylation regulator-based signature in patients with hepatocellular carcinoma**. *Biomed Res Int* (2020) **2020** 2053902. PMID: 32733931
|
---
title: Sodium cholate ameliorates nonalcoholic steatohepatitis by activation of FXR
signaling
authors:
- Linyu Pan
- Ze Yu
- Xiaolin Liang
- Jiyou Yao
- Yanfang Fu
- Xu He
- Xiaoling Ren
- Jiajia Chen
- Xuejuan Li
- Minqiang Lu
- Tian Lan
journal: Hepatology Communications
year: 2023
pmcid: PMC9988322
doi: 10.1097/HC9.0000000000000039
license: CC BY 4.0
---
# Sodium cholate ameliorates nonalcoholic steatohepatitis by activation of FXR signaling
## Abstract
Non-alcoholic steatohepatitis (NASH) has become a major cause of liver transplantation and liver-associated death. The gut-liver axis is a potential therapy for NASH. Sodium cholate (SC) is a choleretic drug whose main component is bile acids and has anti-inflammatory, antifibrotic, and hepatoprotective effects. This study aimed to investigate whether SC exerts anti-NASH effects by the gut-liver axis. Mice were fed with an high-fat and high-cholesterol (HFHC) diet for 20 weeks to induce NASH. Mice were daily intragastric administrated with SC since the 11th week after initiation of HFHC feeding. The toxic effects of SC on normal hepatocytes were determined by CCK8 assay. The lipid accumulation in hepatocytes was virtualized by Oil Red O staining. The mRNA levels of genes were determined by real-time quantitative PCR assay. SC alleviated hepatic injury, abnormal cholesterol synthesis, and hepatic steatosis and improved serum lipid profile in NASH mice. In addition, SC decreased HFHC–induced hepatic inflammatory cell infiltration and collagen deposition. The target protein-protein interaction network was established through Cytoscape software, and NR1H4 [farnesoid x receptor (FXR)] was identified as a potential target gene for SC treatment in NASH mice. SC-activated hepatic FXR and inhibited CYP7A1 expression to reduce the levels of bile acid. In addition, high-dose SC attenuated the abnormal expression of cancer markers in NASH mouse liver. Finally, SC significantly increased the expression of FXR and FGF15 in NASH mouse intestine. Taken together, SC ameliorates steatosis, inflammation, and fibrosis in NASH mice by activating hepatic and intestinal FXR signaling so as to suppress the levels of bile acid in NASH mouse liver and intestine.
## INTRODUCTION
Nonalcoholic fatty liver disease (NAFLD), a prevailing chronic liver disease, is often related to metabolic diseases including overweight, diabetes, and hyperlipidemia.1 It mainly encompasses the non-progressive nonalcoholic fatty liver (NAFL) and the progressive NASH. NAFL connotes the presence of $5\%$ hepatic steatosis without hepatocyte injury; however, NASH means NAFL is complicated by hepatocellular injury, including with or without fibrosis.2 However, to date, no drug has been officially approved for the treatment of NASH.3 Therefore, there is an urgent need to discover effective drugs for the treatment of NASH.
Sodium cholate (SC), a cholagogue, is a mixture of sodium taurocholate and sodium glycocholate extracted from bovine bile.4 Owing to its multiple pharmacological activities, including anti-inflammatory, antifibrotic, and hepatoprotective effects, SC has been widely used to treat biliary insufficiency, cholecystitis, and cardiovascular disease.4 Recent studies have demonstrated that sodium taurocholate reduces serum total cholesterol, low-density lipoprotein cholesterol (LDL-C), and triglyceride (TG) levels in NAFLD rats.5 Furthermore, mixing sodium glycocholate with sodium taurocholate attenuates the absorption of bile acids from the portal vein into the liver and blocks their absorption in the enterohepatic circulation, resulting in a decrease in bile acid synthesis.6 However, it is not clear whether SC prevents NASH and its underlying mechanism.
The aim of this study was to determine whether SC exerts anti-NASH effects through the gut-liver axis. We used a murine model established by high-fat and high-cholesterol (HFHC) diet feeding, which is characterized by lipid accumulation, inflammation, and fibrosis, and to corroborate the direct effects of SC attenuates hepatocyte injury and lipid accumulation in a palmitic acid (PA)–induced cellular model. We found that SC attenuates bile acid metabolism disorder in NASH mice by activating FXR signaling in the liver and intestine, contributing to the amelioration of steatosis, inflammation, and fibrosis in NASH mice.
## Materials and reagents
Reagents and kits were purchased/obtained as follows: *Atorvastatin calcium* tablets (Pfizer Inc., Dalian, China); Hematoxylin and eosin (H&E) (Biosharp Life Sciences, Hefei, China); Oil Red O (Sigma-Aldrich, St. Louis, USA); Alanine aminotransferase (ALT), Aspartate aminotransferase (AST), Total cholesterol (TC) and TG assay kits (Jiancheng Institute of Biotechnology, Nanjing, China); SYBR Green supermix (Bio-Rad, CA, Berkeley, USA); Anti-rabbit-HRP antibodies (Promega, Madison, WI, USA); Anti-CD68 antibodies (Boster Biological Technology Co, Ltd., Wuhan, China); Anti–toll-like receptor (TLR) (Proteintech Group, Inc., Rosemont, PA, USA); anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibodies (Boster Biological Technology Co, Ltd., Wuhan, China); and anti-FXR antibodies (Proteintech Group, Inc., Rosemont, PA).
## Preparation of SC
The SC was obtained from the Shanghai Zhonghua Pharmaceutical Co. (Shanghai, China), which belongs to the State Drug Certificate H19999122. SC was dissolved in CMC-Na solution to make a turbid solution of SC and frozen at −20 °C. It was defrosted and injected i.p. at the doses indicated (90 and 180 mg/kg) before daily administration.
## Cell culture and treatment
Human hepatocyte line L02 cell (L02) was obtained from the Chinese Academy of Sciences (Shanghai Institute of Biochemistry and Cell Biology, Shanghai, China). It was cultured in $10\%$ fetal bovine serum (FBS) (Bioind, Israel) and sodium pyruvate solution (Gibco, USA) in RPMI-1640 medium (Thermo, USA) at 37 °C with $5\%$ CO2. All cell lines were confirmed to be free of mycoplasma contamination. To emulate the NAFLD model, L02 cells were treated with 0.5 mM PA (Sigma, USA) containing cell culture medium for 24 hours. Bovine serum albumin ($0.5\%$) (BioFroxx, Germany) was used as a control.
## Animal models
Mice were housed in specific pathogen free (SFP) conditions with an ambient temperature of 22–26 °C and humidity of 50–$65\%$, providing $\frac{12}{12}$ hours of alternating day and night light and darkness. The NASH was modeled by feeding mice an HFHC (Nantong Trophic; fat, $40\%$; cholesterol, $2\%$) for 20 weeks and rendering SC (90 and 180 mg/kg) by gavage after 10 weeks. Mice feeding normal chow (NC) were taken as control. All animal protocols were approved by the Animal Ethics Committee of Guangdong Pharmaceutical University.
## Cell-counting kit-8 assay
The L02 cells viability was measured by cell-counting kit-8 (CCK-8) colorimetric assay (BioSharp, China), according to the manufacturer’s instructions. L02 cells were subcultured in 96-well plates with complete medium and treated with different concentrations of SC for 24 hours. After treatments, CCK8 reagent was added into each well, followed by incubating for another 1.5 hours in a 37 °C incubator with $5\%$ CO2. Then, the absorbance was measured by a microplate reader at 450 nm (Bio-Rad, Hercules, CA, USA), and cell viability was expressed as percentage values, as compared with the control group.
## Lactate dehydrogenase
An lactate dehydrogenase (LDH) cytotoxicity assay kit (Jiancheng Institute of Biotechnology, Nanjing, China) was used to examine the release level of LDH according to the instructions, and the optical density of the samples was measured by a microplate spectrophotometer (Bio-Rad, Hercules, CA, USA) at 450 nm. The level of cytokines was measured using ELISA-based kits, according to the manufacturer’s instructions.
## Cellular Oil Red O staining
L02 cells were treated with 0.5 mM PA, rinsed twice with PBS, fixed with $4\%$ paraformaldehyde for 15 minutes, and then stained with $60\%$ Oil Red O working solution for 5 minutes. After washing three times with deionized water, images were observed under a microscope (Olympus, Tokyo, Japan).
## Glucose and insulin tolerance tests
The glucose tolerance test (GTT) was performed on mice fed with HFHC diet for 16 weeks. One week later, the same mice were performed by the insulin tolerance test (ITT). For GTT, mice were fasted for 16 hours. After measuring the baseline blood glucose level by means of a tail nick using a glucometer, 2 g/kg glucose was administered by means of intragastric injection, and glucose levels were measured 15, 30, 60, and 120 minutes after glucose injection. For ITT, mice fasted for 6 hours were injected i.p. with insulin at 0.5 U/kg and their blood glucose concentrations were determined 15, 30, 45, and 60 minutes after insulin injection.
## Serum assays
Serum TG, TC, high-density lipoprotein (HDL), low-density lipoprotein (LDL), ALT and AST levels were measured by a commercial kit, according to the manufacturers’ instruction (Jiancheng Bioengineering Institute, Nanjing, China).
## Quantitative analysis of hepatic TG and total cholesterol
Hepatic TG and TC were extracted from liver tissues with a mixture of chloroform and methanol. The contents of hepatic TG and TC were measured by a commercial kit (Jiancheng Bioengineering Institute, Nanjing, China) and normalized by liver wet weight.
## Quantitative analysis of hepatic free fatty acids and free cholesterol
Hepatic free fatty acid (FFA) and free cholesterol (FC) were extracted from liver tissues with saline. The contents of hepatic FFA and FC were measured by a commercial kit (Jiancheng Bioengineering Institute, Nanjing, China) and normalized by liver wet weight.
## Histopathology
Liver tissue was fixed overnight in $4\%$ paraformaldehyde solution (4 °C), embedded in paraffin, and then sectioned (4 μm) for hematoxylin and eosin (H&E) staining to visualize liver ballooning, steatosis, and inflammatory cell infiltration. Picrosirius red (PSR, 26357-02; Hede Biotechnology Co., Ltd., Beijing, China) staining was performed to visualize the degree of liver fibrosis. The positive areas were quantified using the Image J software. Histologic images of section tissues were captured with a light microscope (Olympus, Tokyo, Japan).
## NAFLD activity score scores
The NAFLD activity score (NAS) score is a semiquantitative scoring system. The main contents of NAS scores are hepatocyte steatosis, hepatic lobular inflammation, and hepatocyte ballooning. The total NAS score is the sum of these 3 scores: [1] hepatocyte steatosis: 0 (<$5\%$), 1 ($5\%$–$33\%$), 2 ($34\%$–$66\%$), 3 (>$66\%$); [2] lobular inflammation 0 (none), 1 (<2), 2 (2–4), 3 (>4); [3] hepatocyte ballooning: 0 (none); 1 (rare); 2 (common).
## Immunohistochemistry
Liver specimens fixed in $4\%$ paraformaldehyde solution were embedded in paraffin blocks. Liver sections (4 μm thick) were processed using a standard immunostaining protocol. For immunohistochemical analysis, liver sections were separated, rehydrated, and incubated sequentially with primary and secondary antibodies. The area of positive staining was measured in high magnification fields on each slide and quantified using Image J.
## Real-time quantitative PCR
Total RNA of liver tissues was extracted using TRIzol reagent (Thermo, USA), followed by reverse transcription and quantitative real-time PCR (q-PCR). From the extracted mRNA, cDNA was synthesized using the PrimeScript™ RT reagent kit with gDNA Eraser (Takara, Beijing, China). All the primer sequence information was shown in Table 1. Q-PCR assay was performed using the SYBR Green Supermix (Bio-Rad, Hercules, CA, USA). The relative amount of each mRNA was calculated by using the comparative threshold cycle method. Comparative threshold values were normalized to GAPDH.
**TABLE 1**
| Primer name | Forward primer sequence | Reverse primer sequence |
| --- | --- | --- |
| Human | Human | Human |
| GAPDH | GGAGCGAGATCCCTCCAAAAT | GGCTGTTGTCATACTTCTCATG |
| CHREBP | GCGTTTTGACCAGATGCGAGAC | CGTTGAAGGACTCAAACAGAGGC |
| SREBP-1C | ACTTCTGGAGGCATCGCAAGCA | AGGTTCCAGAGGAGGCTACAAG |
| DGAT1 | GCTTCAGCAACTACCGTGGCAT | CCTTCAGGAACAGAGAAACCACC |
| CD36 | GTGTGGTGATGTTTGTTGCTTT | CTGGATAAGCAGGTCTCCAACT |
| PPAR-α | CAAGAAGACGGAGTCGGATG | CGTAAAGCCAAAGCTTCCAG |
| ATGL | CCCACTTCAACTCCAAGGACGA | GCAGGTTGTCTGAAATGCCACC |
| HSL | AGCCTTCTGGAACATCACCGAG | TCGGCAGTCAGTGGCATCTCAA |
| MGL | GGCATGGTACTCATTTCGCCTC | GTTTGGCAGCACAAGGTTGAGC |
| Mouse | Mouse | Mouse |
| Gapdh | GTCAAGGCCGAGAATGGGAA | CTCGTGGTTCACACCCATCA |
| Scap | CCGAGCATTCCAACTGGTG | CCATGTTCGGGAAGTAGGCT |
| Srebp2 | GCAGCAACGGGACCATTCT | CCCCATGACTAAGTCCTTCAACT |
| Hmgcr | TGTTCACCGGCAACAACAAGA | CCGCGTTATCGTCAGGATGA |
| Srebp-1c | CGACTACATCCGCTTCTTGCAG | CCTCCATAGACACATCTGTGCC |
| Dgat1 | CCGTGTTTGCTCTGGCATC | TGACCTTCTTCCCTGTAGAG |
| Cd36 | TGAGACTGGGACCATTGGTGAT | CCCAAGTAAGGCCATCTCTACCAT |
| Ppar-α | TATTCGGCTGAAGCTGGTGTAC | CTGGCATTTGTTCCGGTTCT |
| Atgl | GAGGAATGGCCTACTGAACCA | GGCTGCAATTGATCCTCCTCT |
| Il-1β | CCGTGGACCTTCCAGGATGA | GGGAACGTCACACACCAGCA |
| Ccl2 | TACAAGAGGATCACCAGCAGC | ACCTTAGGGCAGATGCAGTT |
| Ccl5 | TGCTGCTTTGCCTACCTCTC | TCTTCTCTGGGTTGGCACAC |
| Fxr | GCTTGATGTGCTACAAAAGCTG | CGTGGTGATGGTTGAATGTCC |
| Shp | TGGGTCCCAAGGAGTATGC | GCTCCAAGACTTCACACAGTG |
| Cyp7a1 | GGGATTGCTGTGGTAGTGAGC | GGTATGGAATCAACCCGTTGTC |
| Cyp8b1 | TCCTCAGGGTGGTACAGGAG | GATAGGGGAAGAGAGCCACC |
| Tnf-α | GACGTGGAACTGGCAGAAGAG | TTGGTGGTTTGTGAGTGTGAG |
| Afp | TCACATCCACGAGGAGTGTTG | GCGTGAATTATGCAGAAGCCTA |
| Ki67 | CAAGGCGAGCCTCAAGAGATA | TGTGCTGTTCTACATGCCCTG |
| Fgf15 | ATGGCGAGAAAGTGGAACGG | GGACCAGCGGAGTACAGGT |
| Asbt | GTCTGTCCCCCAAATGCAACT | CACCCCATAGAAAACATCACCA |
| Ostβ | AGATGCGGCTCCTTGGAATTA | TGGCTGCTTCTTTCGATTTCTG |
| Tlr4 | CTGCAATGGATCAAGGACCA | TTATCTGAAGGTGTTGCACATTC |
| Tlr2 | CCCATTGCTCTTTCACTGCT | CTTCCTTGGAGAGGCTGATG |
| Il6 | AGGAGTGGCTAAGGACCAAGACC | TGCCGAGTAGACCTCATAGTGA |
## Western blot analysis
The L02 cells were lysed with ice-cold RIPA lysis buffer (65 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, $1\%$ Nonidet P-40, $0.5\%$ sodium deoxycholate, and $0.1\%$ SDS) supplemented with a protease inhibitor (Roche, Basel, Switzerland) and a phosphatase inhibitor (Roche, Basel, Switzerland) and centrifuged at 12,000 rpm for 30 minutes at 4 °C. Total proteins (20–40 µg) were electrophoresed on SDS-PAGE gels and transferred to polyvinylidene fluoride membranes (Millipore). Afterwards, the membranes were blocked with $10\%$ nonfat dry milk, followed by incubation with primary and secondary antibodies. Membranes were detected by Clarity Western electrochemiluminescence (ECL) Substrate (Bio-Rad, USA) in conjunction with a chemiluminescence system (New Life Science Products, USA).
## Statistical analysis
All data were statistically analyzed by using GraphPad Prism 8.3.0 (San Diego, California, USA), and the results were expressed as the mean±SD. Differences between the means of the 2 groups were analyzed by a 2-tailed unpaired Student’s t-test and were considered statistically significant when $p \leq 0.05$, while for comparisons between more than 2 groups a one-way ANOVA was performed.
## SC ameliorates PA-induced hepatocyte injury and lipid accumulation in L02 cells
We examined the effect of SC on hepatocyte viability using the CCK8 assay, and the results showed that when the SC concentration was higher than 80 μM, the viability of hepatocytes was decreased (Figure 1A). Next, L02 cells were treated with 0.5 mM PA for 24 hours to establish a cell model of NAFLD and simultaneously administered with various concentrations of SC. The results showed that the cell survival rate gradually increased as the SC concentration increased from 5 to 20 μM, whereas the cell survival rate gradually decreased when the SC concentration increased from 40 to 400 μM (Figure 1B). Thus, 5, 10, and 20 μM of SC were selected for the subsequent cell experiments. LDH level in the supernatant of L02 cells was elevated by PA induction, suggesting hepatocyte was damaged by lipid toxicity. However, LDH level in hepatocyte was reduced by SC treatment (Figure 1C). The TG content in L02 cells was induced by PA, whereas reduced by SC treatment (Figure 1D). This finding was further confirmed by Oil Red O staining (Figure 1E), suggesting that SC treatment attenuated TG accumulation in L02 cells induced by PA. Moreover, q-PCR assays showed that SC-diminished PA induced the mRNA levels of lipid synthesis such as carbohydrate response element binding protein (CHREBP), sterol regulatory element-binding protein 1C (SREBP-1C), and DGAT-1 and lipid uptake genes such as fatty acid translocase CD36 (CD36), whereas did not affect other lipid metabolism-related genes such as peroxisome proliferator-activated receptor-α (PPAR-α), adipose triglyceride lipase (ATGL), monoacylglycerol lipase (MGL), and hormone-sensitive lipase (HSL) (Figure 1F), suggesting that SC reduced lipid synthesis and uptake in L02 cells induced by PA. Collectively, these data demonstrated that SC attenuated hepatocyte injury and lipid accumulation induced by PA in hepatocytes.
**FIGURE 1:** *SC reduces hepatocyte injury and lipid accumulation in L02 cells treated with PA. (A) The effect of SC on L02 cell viability. (B) The effect of SC on L02 cell cytotoxicity. (C) Lactate dehydrogenase contents in L02 cells in the indicated groups treated with PA (0.5 mM) for 24 hours. (D) Triglyceride contents in L02 cells in the indicated groups stimulated with PA (0.5 mM) for 24 hours. (E) Oil Red O staining showing the degrees of lipid accumulation in L02 cells treated with DMSO, 5,10, and 20 μM SC in response to PA (0.5 mM) stimulation for 24 hours. Scale bar, 100 μm. (F) Relative mRNA levels of the indicated lipid synthesis (CHREBP, SREBP-1C, and DGAT-1), lipid uptake genes (CD36), and lipid metabolism-related genes (PPAR-α, ATGL, MGL, and HSL) in L02 cells treated with PA (0.5 mM). n=3 per group. The Data are presented as the mean±SD. # indicates a significant difference between the DMSO group and the PA (0.5 mM) group (t test); *indicates a significant difference between the SC (5 μM)/SC (10 μM)/SC (20 μM) group and the PA (0.5 mM) group (one-way ANOVA). ##
p<0.01, ###
p<0.001 versus DMSO group; *p<0.05, **p<0.01, ***p<0.001 versus cells induced by PA. Abbreviations: ATGL, adipose triglyceride lipase; CD36, fatty acid translocase CD36; CHREBP, carbohydrate response element binding protein; DGAT1, diacylglycerol acyltransferase 1; HSL, hormone-sensitive lipase; MGL, monoacylglycerol lipase; NS, no significance; PA, palmitic acid; PPAR-α, peroxisome proliferator-activated receptor-α; SC, sodium cholate; SREBP-1C, sterol regulatory element-binding protein 1C.*
## SC attenuates HFHC-induced obesity and insulin resistance in mice
To investigate the effects of SC on NASH, we established a NASH mouse model induced by an HFHC diet. Mice were fed with HFHC diet for 20 weeks and treated with SC (90 and 180 mg/kg/d) since the 11th week (Figure 2A). In the 10th week, there were no significant differences in the body weights between the NASH mice and the control mice. At the end of the 20th week, the body weights of the NASH mice were more than those of normal mice, whereas the body weights of the NASH mice treated with atorvastatin (ART) and SC were lower than those of the NASH mice without any treatment (Figure 2B). To assess the effect of SC on insulin resistance in NASH mice, we performed GTT and ITT. After 20 weeks of administration of a high-calorie diet, GTT assay showed that after i.p. injection of glucose, the levels of blood glucose of the mice rapidly reached the peak after 15 minutes, whereas decreased to the baseline after 120 minutes (Figure 2C). The AUC of GTT in NASH mice was larger than that of the normal chow mice; the AUC of GTT in NASH mice with SC treatment was smaller than that of NASH mice without any treatment (Figure 2C). Furthermore, ITT assay showed that after the i.p. injection of insulin, the levels of blood glucose in mice continued to drop in the first 30 minutes, the blood glucose of control mice firstly rebounded after 30 minutes, and the blood glucose of NASH mice started to rebound after 45 minutes (Figure 2D). Moreover, the AUC of ITT in NASH mice was larger than that of the normal chow mice; the AUC of ITT in NASH mice with SC treatment was smaller than that of NASH mice without any treatment (Figure 2D). Taken together, these data suggested that SC treatment improved the rate of glucose metabolism and insulin resistance in NASH mice.
**FIGURE 2:** *SC alleviates insulin resistance and glucose intolerance in mice fed an HFHC diet. (A) A mouse model of NASH was established schematically to investigate the effects of SC on NASH; we fed mice with this high-calorie diet for 20 weeks and treated them with SC (90 and 180 mg/kg/d) starting at week 11. (B) Body weight in the first 10 weeks and the last 10 weeks of the mice. n=10 per group. (C, D) The GTT and ITT assays were performed to evaluate the insulin sensitivity of the indicated groups of mice treated with NC, HFHC, or SC. n=6—10 per group. Data are represented as means±SD. # indicates a significant difference between the NC group and the HFHC group (t test); *indicates a significant difference between the L-SC (Low dose-Sodium Cholate: 90 mg/kg)/H-SC (High dose-Sodium Cholate: 180 mg/kg)/ART group and the HFHC group (one-way ANOVA). ###
p<0.001 versus NC mice; **p<0.01, ***p<0.001 versus mice fed by HFHC. Abbreviations: ART, atorvastatin; HFHC, high-fat and high-cholesterol; ITT, insulin tolerance test; GTT, glucose tolerance test; NC, normal chow.*
## Effects of SC on hepatic injury and serum lipid profile in HFHC-fed mice
The above in vivo experiments in mice demonstrated the comprehensive protective effect of SC in improving the systemic metabolic stress status in NASH mice. Therefore, we next assayed the levels of serum glutamate-pyruvate ALT and AST, the classical indicators of clinical liver function, and the results showed that SC and ART treatment significantly reduced the abnormal rise of serum transaminases induced by HFHC (Figure 3A). In addition, consistent with the previous results, SC and ART treatment decreased the abnormal elevation of serum TG and TC induced by HFHC, suggesting that SC restored lipid metabolism homeostasis in NASH mice (Figure 3B). Abnormal changes in HDL and LDL are also key serum biochemical indicators in the context of NASH, reflecting the systemic metabolic stress state and the extent of vascular endothelial damage. SC and ART treatment improved the abnormal changes of lipoproteins mentioned above (Figure 3C). These data suggested that SC attenuated hepatic injury and serum lipid profile in NASH mice.
**FIGURE 3:** *Effects of sodium cholate on serum biochemical parameters in NASH mice. (A) Levels of serum ALT and AST were measured in mice after 20 weeks. n=7–9 per group. (B) Serum lipid (TG and TC) levels. n=7–9 per group. (C) The serum contents of HDL and LDL of mice in the indicated groups. n=7–9 per group. Data are represented as means±SD. #indicates a significant difference between the NC group and the HFHC group (t test); *indicates a significant difference between the L-SC (Low dose-Sodium Cholate: 90 mg/kg) H-SC (High dose-Sodium Cholate: 180 mg/kg) group and the HFHC group (one-way ANOVA). ###
p<0.001 versus NC mice; *p<0.05, **p<0.01, ***p<0.001 versus mice fed by HFHC. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; ART, atorvastatin; HFHC, high-fat and high-cholesterol; TC, total cholesterol; TG, triglyceride.*
## SC ameliorates hepatic steatosis and injury in HFHC-fed mice
H&E staining exhibited significant steatosis and hepatocyte ballooning in the liver of NASH mice. Nevertheless, treatment with SC dramatically reduced these 2 lesions and its effect is similar to that of ART (Figure 4A). On the basis the above pathological staining results, we assessed the liver injury in mice by 3 histological features, namely the degree of steatosis, the number of inflammatory lesions in the liver lobules, and the degree of ballooning of hepatocytes, by NAS score. The results showed that SC treatment significantly improved hepatic steatosis and injury in HFHC-induced NASH mice (Figure 4B). At week 20, the liver weight and the liver/body weight ratio were increased in NASH mice. However, both were decreased in NASH mice treated with ART and SC compared with NASH mice without any treatment (Figure 4C). Hepatic TG and TC accumulation in HFHC-fed mice was significantly decreased by SC treatment (Figure 4D). However, when the liver is chronically exposed to high levels of FC and FFA, it increases the body’s insulin resistance,7 causing lipotoxicity and promoting the secretion of inflammatory cytokines, leading to hepatocyte damage and progressive fibrosis during NASH. Therefore, we further assayed these 2 key indicators, FFA and FC, and our data suggest that in a mouse model of NASH, SC and ART treatment significantly reduces the levels of FFA and FC (Figure 4E). To confirm the molecular mechanism of SC for restoring hepatic lipid metabolism homeostasis, we next assayed a representative series of genes for cholesterol and lipid metabolism, including genes for cholesterol and lipid biosynthesis, lipolysis, and uptake, using q-PCR assays. The results showed that the mRNA levels of cholesterol synthesis and lipogenesis genes such as SREBP cleavage-activating protein (Scap), Srebp2, HMG-CoA Reductase (Hmgcr), Srebp-1c, and Dgat1 were increased in HFHC-fed mice, whereas decreased by SC and ART treatment (Figure 4F). In addition, SC and ART treatment decreased the mRNA levels of hepatic lipid uptake genes such as Cd36 in HFHC-fed mice (Figure 4F). Conversely, the hepatic mRNA levels of Ppar-α and Atgl were decreased in HFHC-fed mice, whereas increased by SC and ART treatment (Figure 4F). These data suggested that SC attenuated HFHC-induced hepatic steatosis and disorder of hepatic lipid metabolism in mice.
**FIGURE 4:** *Sodium cholate attenuates hepatic steatosis in mice fed the HFHC diet. (A) Hepatic steatosis was measured by H&E staining. Scale bar, 50 μm. (B) NAFLD activity score in the indicated groups. n=7–10 per group. (C) Liver weight and Liver/body weight ratios of the mice. n=8–10 per group. (D) Hepatic lipid (TG and TC) levels. n=7–8 per group. (E) Hepatic lipid (FFA and FC) levels. n=7–8 per group. (F) Quantitative real-time PCR analysis of the transcript levels of genes related to cholesterol synthesis gene (Scap, Srebp2 and Hmgcr), lipogenic genes (Srebp-1c and Dgat1), lipid uptake genes (Cd36), oxidative phosphorylation gene (Ppar-α), and lipolytic gene (Atgl). n=7–9 per group. The Data are presented as the mean±SD. #indicates a significant difference between the NC group and the HFHC group (t test); *indicates a significant difference between the L-SC (Low dose-Sodium Cholate: 90 mg/kg)/H-SC (High dose-Sodium Cholate: 180 mg/kg)/ART group and the HFHC group (one-way ANOVA). ##
p<0.01, ###
p<0.001 versus NC mice; *p<0.05, **p<0.01, ***p<0.001 versus mice fed by HFHC. ns indicates no significance. Abbreviations: ART, atorvastatin; ATGL, adipose triglyceride lipase; CD36, fatty acid translocase CD36; DGAT1, diacylglycerol acyltransferase 1; HMGCR, HMG-CoA reductase; FFA, free fatty acid; FC, free cholesterol; H&E, hematoxylin-eosin; HFHC, high-fat and high-cholesterol; NC, normal chow; PPAR-α, peroxisome proliferator-activated receptor-α; SREBP cleavage-activating protein; SREBP-1C, sterol regulatory element-binding protein 1C; TC, total cholesterol; TG, triglyceride.*
## SC attenuated liver inflammation and progressive fibrosis in HFHC-fed mice
To further investigate the protective effects of SC on NASH in vivo, we evaluated the effects of SC on liver inflammation and fibrosis in NASH mice. Immunohistochemical staining of CD68 showed that NASH mice treated with SC-exhibited decreased inflammatory cell infiltration compared with NASH mice without any treatment (Figure 5A). Furthermore, SC treatment significantly decreased the hepatic mRNA levels of inflammatory cytokines such as Il-1β, Ccl2, and Ccl5 in NASH mice (Figure 5B). In addition, PSR staining showed that SC treatment significantly reduced the hepatic collagen deposition in NASH mice, thus slowing down the progression of liver fibrosis (Figure 5C). Together, these data suggested that SC attenuated hepatic inflammation and fibrosis in NASH mice under a metabolic stress condition.
**FIGURE 5:** *Sodium cholate suppresses hepatic inflammation and fibrosis in mice fed an HFHC diet. (A) Immunohistochemical staining of CD68 in the livers of the indicated mice fed the normal chow or HFHC diet for 20 weeks. n=5 per group. Scale bar, 50 μm. (B) Quantitative real-time PCR analysis of the transcript levels of genes related to inflammation (Il-1β, Ccl2, and Ccl5). n=5 per group. (C) Representative images showing PSR staining in the livers of the indicated mice fed the normal chow or HFHC diet for 20 weeks. n=5 per group. Scale bar, 50 μm. The data are presented as the mean±SD. #indicates a significant difference between the NC group and the HFHC group (t test); *indicates a significant difference between the L-SC (Low dose-Sodium Cholate: 90 mg/kg)/(High dose-Sodium Cholate: 180 mg/kg)/ART group and the HFHC group (one-way ANOVA). ###
p<0.001 versus NC mice; *p<0.05, **p<0.01, ***p<0.001 versus mice fed by HFHC. Abbreviations: ART, atorvastatin; CCL2, C-C motif chemokine ligand 2; CCL5, C-C motif chemokine ligand 5; CD68, fatty acid translocase CD68; HFHC, high-fat and high-cholesterol; NC, normal chow; NS, no significance; PSR, picrosirius red.*
## SC activates FXR signaling in NASH mice
To disclose the key target proteins of SC against NASH, protein-protein interaction network consisted of 97 nodes and 295 edges (Figure 6A) was acquired from the gene expression STRING database.8 Among the target proteins with degree >5, TP53, TNF-α, PPARA, and NR1H4 (FXR) were reported to involve in multiple biological processes contributed to the NASH progression.9 Among the protein-protein interaction network, the cluster containing NR1H4 (FXR) had 25 targets with an average score of 2.48; the cluster containing TNF-α had 20 targets with an average score of 2.05; and the cluster containing PPARA had 23 targets with an average score of 1.62 (the cluster score represents the core density of nodes and topologically adjacent nodes, with higher scores representing more concentrated clusters), suggesting that NR1H4 (FXR) might be a potential direct target gene for SC treatment of NASH (Figure 6A). GO enrichment: GO Ontology enrichment10 showed that there were 101 shared targets were enriched with a p-value threshold of 0.01. Among the top 20 enrichment results, Response to hormone, Steroid metabolic process, Positive regulation of cytosolic calcium ion concentration and nuclear receptor activity were enriched to a relatively high degree and decreased p-values. Thus, it is suggested that nuclear receptor activity might be associated with FXR (Figure 6B). Among 101 outstandingly enriched *Kyoto encyclopedia* of genes and genomes (KEGG) pathway were obtained from the metascape database,11 top 20 ranking KEGG pathway by p-value were descripted in Figure 6C, of which the most target proteins of SC were enriched in PPAR signaling pathway, cell cycle, and bile secretion. Considering the important role of bile acids in the process of NASH, we further investigated the genes related to bile acid synthesis in the liver tissue of NASH mice. FXR, as a key gene, mediates bile acid metabolism.12 Therefore, to verify the role of SC in bile acid metabolism, we examined genes involved in hepatic bile acid synthesis-related genes. The hepatic mRNA levels of Shp, the binding target of Fxr, was reduced in HFHC-fed mice, whereas significantly increased by SC and ART treatment (Figure 6D). Conversely, the hepatic mRNA levels of key enzymes of the major bile acid synthesis pathway in liver such as Cyp7a1 and Cyp8b1 were increased in HFHC-fed mice, whereas decreased by SC and ART treatment (Figure 6D). To further determine at the protein level whether SC ameliorates NASH by the activation of FXR signaling, we verified the levels of FXR in hepatocytes. SC-activated FXR signaling in hepatocytes subjected to PA stimulation (Figure 6E). The KEGG analysis showed that the cancer pathway was the second most enriched pathway, suggesting high-dose SC treatment significantly decreased the mRNA levels of cancer markers such as Tnf-α, α-fetoprotein (Afp), and Ki67 in NASH mice (Figure 6F). Collectively, these data suggested that SC-activated FXR and bile acid synthesis to blunt the pathogenesis of NASH in mice.
**FIGURE 6:** *Hepatic farnesoid x receptor (FXR) is required by sodium cholate (SC) treatment in NASH mice. (A) Protein-protein interaction network (PPI) network of SC on NASH. The network contains 97 nodes and 295 edges. Nodes represent proteins and edges represent protein-protein associations. The darker color nodes represent greater degree of freedom (DOF). (B) GO functional annotation of SC on NASH. Bar plot shows the top 20 GO enrichment terms. (C) Kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis of SC on NASH. Bar plot shows the top 20 KEGG pathway terms and corresponding targets. Different colors of the graph represent different signal pathways. The darker color bars in the pathway, the more targets are enriched. (D) Quantitative real-time PCR analysis of the transcript levels of genes related to bile acid metabolism (Fxr, Shp, Cyp7a1, and Cyp8b1). n=7 per group. (E) Western blotting of proteins involved in the FXR-mediated signaling in cells. Glyceraldehyde-3-phosphate dehydrogenase served as a loading control. (F) Quantitative real-time PCR analysis of the transcript levels of genes related to cancer markers (Tnf-α, Afp, and Ki67). n=6-7 per group. Data are represented as means±SD. #indicates a significant difference between the NC group and the HFHC group (t test); *indicates a significant difference between the L-SC (Low dose-Sodium Cholate: 90 mg/kg)/H-SC (High dose-Sodium Cholate: 180 mg/kg)/ART group and the HFHC group (one-way ANOVA). #
p<0.05, ##
p<0.01, ###
p<0.001 versus normal chow mice; *p<0.05, **p<0.01, ***p<0.001 versus mice fed by HFHC. Abbreviations: ns, no significance; PPAR, peroxisome proliferator-activated receptor.*
## The FXR-FGF15 pathway is required for SC to ameliorate HFHC-induced intestinal inflammation in mice
Next, q-PCR assays showed that the mRNA levels of Fxr and Fgf15 and bile acid efflux transporter protein [organic solute transporter β, Ostβ] in the ileum were significantly decreased in NASH mice, whereas reversed by SC treatment (Figure 7A). In contrast, the mRNA levels of apical sodium-dependent bile acid transporter (Asbt) were increased in the ileal tissues in HFHC-fed mice, whereas decreased after SC and ART treatment (Figure 7A). We next assayed the mRNA levels of intestinal inflammatory cytokines in NASH mice. As expected, SC and ART treatment significantly decreased the intestinal mRNA levels of inflammatory cytokines such as Tlr4, Tlr2, and Il-6 in NASH mice (Figure 7B). Collectively, these data suggested that SC treatment protected mice against intestinal inflammation by HFHC induction, suggesting that SC exerts its anti-NASH effects as least in part through intestinal FXR-mediated anti-inflammatory mechanisms.
**FIGURE 7:** *Activation of FXR-FGF15 pathway is responsible for the effect of SC on intestinal inflammation in NASH mice. (A) Quantitative real-time PCR analysis of the transcript levels of genes related to FXR-FGF15 pathway (Fxr, Fgf15, Asbt, and Ostβ ). n=8 per group. (B) Quantitative real-time PCR analysis of the transcript levels of genes related to inflammation of the intestinal tract (Tlr4, Tlr2, and Il6). n=8 per group. Data are represented as means±SD. #indicates a significant difference between the NC group and the HFHC group (t test); *indicates a significant difference between the L-SC (Low dose-Sodium Cholate: 90 mg/kg)/H-SC (High dose-Sodium Cholate: 180 mg/kg)/ART group and the HFHC group (one-way ANOVA). ##
p<0.01, ###
p<0.001 versus NC mice; *p<0.05, **p<0.01, ***p<0.001 versus mice fed by HFHC. Abbreviations: ART, atorvastatin; HFHC, high-fat and high-cholesterol; NC, normal chow; ns, no significance.*
## DISCUSSION
NAFLD has been the most common chronic liver disease worldwide. There are no approved drugs because of the complexity of the disease and the safety of the drugs.5 Thus, it is urgent to explore new drugs to treat NAFLD. SC is a mixture of sodium taurocholate and sodium glycocholate extracted from bovine bile.13 In the current study, we demonstrated that SC attenuated lipid accumulation and hepatocyte injury in a PA-induced hepatocytes and inhibited steatosis, inflammation, and fibrosis in NASH mice.
Sterol regulatory element–binding protein (SREBPs) is an essential group of transcription factors regulating lipid synthesis.14,15 SREBP-1c is involved in hepatic lipid synthesis, and overexpression of SREBP-1c causes hepatic lipid accumulation and insulin resistance.16 SREBP2 regulates the expression of genes related to cholesterol synthesis and uptake.17 It was found that abnormal expression of SREBP2 would cause disorders of lipid metabolism, especially cholesterol metabolism, which would lead to excessive deposition in adipose tissue and induce NAFLD.17 The downstream target gene of SREBP2 is HMG-CoA Reductase (HMGCR).18 In NAFLD, an abnormal increase in SREBP2 will promote the synthesis of (HMGCR).18 In addition, when the SREBP protein precursor binds to SREBP cleavage-activating protein (SCAP), leading to endoplasmic reticulum stress and exacerbating the imbalance of TG metabolism, which in turn leads to further intracellular lipid accumulation and induces the development of NAFLD.19 TG deposition in hepatocytes caused by the disorders of lipid metabolism is the cornerstone of the development of NAFLD.20 When the liver is unable to regulate the accumulation of lipids through β-oxidation, FC and FFA form lipotoxic substances, leading to endoplasmic reticulum stress, oxidative stress, and inflammatory vesicle activation.21 PPAR-α regulates β-oxidation and the transport of FFA, which plays a crucial role in lipid metabolism.22 Liver CHREBP-specific knockdown exacerbates obesity, hepatic steatosis, and insulin resistance in mice and vice versa.23 Our results showed that SC treatment significantly decreased the hepatic levels of FFA and FC in NASH mice; meanwhile, the hepatic cholesterol synthesis genes such as Scap, Srebp2, and Hmgcr in NASH mice treated with the mRNA levels of high-dose SC were also reduced; in addition, the mRNA levels of lipogenic genes such as Srebp-1c was reduced, whereas the hepatic mRNA levels of Ppar-α and Atgl were significantly increased in NASH mice treated with SC and ART. These results suggested that SC treatment attenuates hepatic steatosis and lipid metabolism disorders in NASH mice.
The inflammatory process, a hallmark of NASH pathogenesis, is associated with hepatocyte injury and the release of multiple proinflammatory cytokines. The levels of proinflammatory cytokines in the liver, including IL-6, IL-1β, CCL2, and CCL5, are correlated with the severity of NASH.24 Macrophages serve a significant role in the pathogenesis of NAFLD. In terms of liver inflammation, the number of CD68-positive cells reflected macrophage recruitment. Macrophage-mediated immune responses are an essential cause of hepatocyte injury during the development of NAFLD.25 The accumulation of large amounts of lipids exposes macrophages to prolonged “antigenic” stimuli, inducing the production of IL-6 and causing a sustained inflammatory response.26 In addition, inflammation of adipose tissue causes liver injury. In obese mice with severe lipid accumulation, the secretion of IL-1β by adipose tissue macrophages (ATMs) is elevated, which increases the rate of lipolysis of adipocytes and promotes hepatic steatosis.27 CCL2 and CCL5 overexpression recruited macrophages that secrete large amounts of inflammatory cytokines and facilitate hepatic steatosis and vice versa, suggesting that the recruitment of macrophages in adipose tissue causes lipid accumulation in mice.27 Our results showed that SC suppressed liver and intestinal inflammation in NASH mice by decreasing the mRNA levels of inflammatory cytokines such as Il-1β, Ccl2, Ccl5, Tlr4, Tlr2, and Il6. Meanwhile, immunohistochemical staining of CD68 demonstrated that treatment with SC reduced inflammatory cell infiltration in livers of NASH mice. These results suggested that SC exerts an anti-inflammatory effect on liver and intestine in NASH mice.
Severe stages of NASH might be accompanied by fibrosis that manifests in the form of excessive deposition of insoluble collagen and extracellular matrix.28 HSC is the main source of myofibroblasts, and bone marrow–derived fibroblasts are a potential source of myofibroblasts, which are associated with the pathogenesis of liver fibrosis.29 Our results showed that mice fed by HFHC diet developed into liver fibrosis. However, SC reduced the severity of liver fibrosis in the NASH mouse model, thereby hindered the progression of liver fibrosis, suggesting SC exerts an antifibrotic effect on NASH-associated fibrosis.
Bile acids are synthesized from cholesterol in the liver and then transported to the intestine.30 They promote the absorption of lipids and fat-soluble vitamins and are important signaling molecules that regulate glucolipid and energy metabolism in the body. Recent studies have shown that disorders of bile acid metabolism are an important cause of the development of NAFLD/NASH.30 Dysregulation of bile acid metabolism affects hepatic lipid metabolism, immune microenvironment, and intestinal bacteria. The composition of BAs in the liver and intestine is regulated by BA-metabolizing enzymes and BA transporters, the majority of which are encoded by the FXR target gene that is predominantly expressed in hepatocytes and lower small intestinal epithelial cells, and by the intestinal microbiota.31 FXR is a bile acid receptor, which regulates bile acid metabolism by regulating key target genes in all aspects of bile acid metabolism, including bile acid synthesis, metabolism, reabsorption, and transport.32 Recent studies have shown that bile acids in the enterohepatic circulation maintain a dynamic balance of bile acids in the body by regulating FXR activity.33 When bile acid pooling occurs to the liver, excess bile acids activate FXR and induce increased expression of small hetero dimer partner (SHP).34 In the intestine, FXR induced the binding of FGF$\frac{15}{19}$ to the FGFR4, inhibiting CYP7A1 expression and bile acid synthesis.35 In the intestine, conjugated bile acids are first actively reabsorbed into the small intestinal mucosal cells through ASBT in the ileum, and subsequently enter the portal vein by the bile acid efflux transporter OSTα/OSTβ in the basolateral layer.36 In addition, many studies have shown that FXR agonists are potential therapeutic drugs for metabolic and inflammatory diseases.37,38 Obeticholic acid (OCA) is a selective FXR agonist with anticholestatic and hepatoprotective properties,39 participating in BA anabolism and enterohepatic circulation, and modulate immune inflammatory and fibrotic.40 Since the FXR agonist OCA currently developed for the treatment of NASH was denied marketing approval by the FDA because of its safety and side effects, it is sufficient to demonstrate the critical role that bile acids play in the NASH disease process. Our results showed that SC activates hepatic FXR signaling and induces the expression of downstream SHP, inhibiting the rate-limiting enzyme for bile acid synthesis (CYP7A1), leading to reduced bile acid synthesis in the liver. In addition, SC significantly increased the expression of FXR and FGF15 in NASH mouse intestine. Finally, SC activation of intestinal FXR reduced intestinal bile acid levels by inhibiting the expression of ASBT transporter to bile acid reabsorption, and inducing OSTβ transporter expression to promote bile acid efflux. KEGG analysis showed that the cancer pathway was the second most enriched pathway, so we assayed several cancer markers. Our results suggested that high-dose SC attenuates the abnormal expression of cancer markers such as Tnf-α, Afp, and Ki67 in NASH mouse liver. Finally, SC activated FXR signaling in hepatocytes subjected to PA stimulation. Taken together, the above results indicated that SC might activate both hepatic and intestinal FXR expression and regulate bile acid synthesis to impede the pathogenesis of NASH.
Collectively, our current results demonstrated that SC attenuates bile acid metabolism disorder in NASH mice by activating FXR signaling in the liver and intestine, contributing to the amelioration of steatosis, inflammation, and fibrosis in NASH mice. Therefore, our findings will provide insight into the development of clinical treatment for NASH (Figure 8).
**FIGURE 8:** *Potential mechanism by which SC attenuated hepatic steatosis, inflammation, fibrosis, and intestinal inflammation in NASH mice. SC reduced the expression of Srebp-1c and Dgat1 in NASH mouse liver, thereby alleviating hepatic steatosis and lipid metabolism disorders. SC suppressed liver and intestinal inflammation in NASH mice by decreasing the mRNA levels of inflammatory cytokines such as Il-1β, Ccl2, Ccl5, Tlr4, Tlr2 and Il6. The protective mechanism of SC not only attributes to the downregulation of Scap, Srebp2 and Hmgcr by SC, inhibiting the synthesis of cholesterol in the NASH mouse liver. More importantly, SC activates hepatic FXR signaling and induces the expression of downstream SHP, inhibiting CYP7A1 expression, leading to reduced bile acid synthesis in the liver. Meanwhile, SC activates intestinal FXR and induces the expression of FGF15, followed by the secretion of FGF15 into the liver and inhibition of CYP7A1 expression in liver and decreased hepatic bile acid synthesis. Furthermore, SC activation of intestinal FXR reduced intestinal bile acid levels by inhibiting the expression of ASBT transporter to bile acid reabsorption, and inducing OSTβ transporter expression to promote bile acid efflux. In conclusion, SC attenuates bile acid metabolism disorder in NASH mice by activating FXR signaling in the liver and intestine, contributing to the amelioration of steatosis, inflammation, and fibrosis in NASH mice. Abbreviations: ASBT, apical sodium-dependent bile acid transporter; CHREBP, carbohydrate response element binding protein; FXR, farnesoid x receptor; HMGCR, HMG-CoA reductase; SREBP cleavage-activating protein; SREBP2, sterol regulatory element-binding protein 2; TLR, toll-like receptor.*
## AUTHOR CONTRIBUTIONS
Tian Lan conceived and designed the experiments. Linyu Pan, Ze Yu, Xiaolin Liang, Jiyou Yao, and Yanfang Fu carried out the experiments and wrote the manuscript. Xu He, Xiaoling Ren, Jiajia Chen, Xuejuan Li, and Minqiang Lu took part in the discussion and proofreading the manuscript.
## CONFLICT OF INTEREST
The authors declare no conflict of interests for this article.
## ETHICS STATEMENT
All animal experiments were performed following the Guide for the Care and Use of Laboratory Animals, and the procedures were approved by the Research Ethical Committee of Guangdong Pharmaceutical University.
## References
1. Loomba R, Friedman SL, Shulman GI. **Mechanisms and disease consequences of nonalcoholic fatty liver disease**. *Cell* (2021) **184** 2537-64. PMID: 33989548
2. Turkish AR. **Nonalcoholic fatty liver disease: emerging mechanisms and consequences**. *Curr Opin Clin Nutr Metab Care* (2008) **11** 128-33. PMID: 18301087
3. Chalasani N, Younossi Z, Lavine JE, Charlton M, Cusi K, Rinella M. **The diagnosis and management of nonalcoholic fatty liver disease: practice guidance from the American Association for the Study of Liver Diseases**. *Hepatology* (2018) **67** 328-57. PMID: 28714183
4. Li Y, Zhu C. **Mechanism of hepatic targeting via oral administration of DSPE-PEG-cholic acid-modified nanoliposomes**. *Int J Nanomedicine* (2017) **12** 1673-84. PMID: 28280334
5. Liu XJ, Liu C, Zhu LY, Fan CL, Niu C, Liu XP. **Hepalatide ameliorated progression of nonalcoholic steatohepatitis in mice**. *Biomed Pharmacother* (2020) **126** 110053. PMID: 32200254
6. Fan QY, Zhang YT, Hou XF, Li Z, Zhang K, Shao Q. **Improved oral bioavailability of notoginsenoside R1 with sodium glycocholate-mediated liposomes: Preparation by supercritical fluid technology and evaluation in vitro and in vivo**. *Int J Pharm* (2018) **552** 360-70. PMID: 30292894
7. Cusi K. **Role of obesity and lipotoxicity in the development of nonalcoholic steatohepatitis: pathophysiology and clinical implications**. *Gastroenterology* (2012) **142** 711-25 e716. PMID: 22326434
8. Subuddhi U, Mishra AK. **Micellization of bile salts in aqueous medium: a fluorescence study**. *Colloids Surf B Biointerfaces* (2007) **57** 102-7. PMID: 17336505
9. Hill EC, O’Donnell L. **Factors associated with low bone density in a juvenile mortality sample**. *FASEB J* (2022) **36**. DOI: 10.1096/fasebj.2022.36.S1.R2403
10. Jiang ZY, Zhou Y, Zhou L, Li SW, Wang BM. **Identification ofkey genes and immune infiltrate in nonalcoholic steatohepatitis: a bioinformatic analysis**. *Biomed Res Int* (2021) 7561645. DOI: 10.1155/2021/7561645
11. Kanehisa M, Furumichi M, Tanabe M, Sato Y, Morishima K. **KEGG: new perspectives on genomes, pathways, diseases and drugs**. *Nucleic Acids Res* (2017) **45** D353-D361. PMID: 27899662
12. Jiang L, Zhang H, Xiao D, Wei H, Chen Y. **Farnesoid X receptor (FXR): structures and ligands**. *Comput Struct Biotechnol J* (2021) **19** 2148-59. PMID: 33995909
13. Kwak K, Yu B, Mouli SK, Larson AC, Kim DH. **Sodium cholate bile acid-stabilized ferumoxytol-doxorubicin-lipiodol emulsion for transcatheter arterial chemoembolization of hepatocellular carcinoma**. *J Vasc Interv Radiol* (2020) **31** 1697-705 e1693. PMID: 32773247
14. DeBose-Boyd RA, Ye J. **SREBPs in lipid metabolism, insulin signaling, and beyond**. *Trends Biochem Sci* (2018) **43** 358-68. PMID: 29500098
15. Zhang X, Coker OO, Chu ES, Fu K, Lau HCH, Wang YX. **Dietary cholesterol drives fatty liver-associated liver cancer by modulating gut microbiota and metabolites**. *Gut* (2021) **70** 761-74. PMID: 32694178
16. Horton JD, Shimomura I, Ikemoto S, Bashmakov Y, Hammer RE. **Overexpression of sterol regulatory element-binding protein-1a in mouse adipose tissue produces adipocyte hypertrophy, increased fatty acid secretion, and fatty liver**. *J Biol Chem* (2003) **278** 36652-60. PMID: 12855691
17. Xue L, Qi H, Zhang H, Ding L, Huang Q, Zhao D. **Targeting SREBP-2-regulated mevalonate metabolism for cancer therapy**. *Front Oncol* (2020) **10** 1510. PMID: 32974183
18. Zhang YY, Fu ZY, Wei J, Qi W, Baituola G, Luo J. **A LIMA1 variant promotes low plasma LDL cholesterol and decreases intestinal cholesterol absorption**. *Science* (2018) **360** 1087-92. PMID: 29880681
19. Chu BB, Liao YC, Qi W, Xie C, Du X, Wang J. **Cholesterol transport through lysosome-peroxisome membrane contacts**. *Cell* (2021) **184** 289. PMID: 33417863
20. Friedman SL, Neuschwander-Tetri BA, Rinella M, Sanyal AJ. **Mechanisms of NAFLD development and therapeutic strategies**. *Nat Med* (2018) **24** 908-22. PMID: 29967350
21. Kwon EY, Jung UJ, Park T, Yun JW, Choi MS. **Luteolin attenuates hepatic steatosis and insulin resistance through the interplay between the liver and adipose tissue in mice with diet-induced obesity**. *Diabetes* (2015) **64** 1658-69. PMID: 25524918
22. Xu J, Xiao G, Trujillo C, Chang V, Blanco L, Joseph SB. **Peroxisome proliferator-activated receptor alpha (PPARalpha) influences substrate utilization for hepatic glucose production**. *J Biol Chem* (2002) **277** 50237-44. PMID: 12176975
23. Iizuka K, Takao K, Yabe D. **ChREBP-mediated regulation of lipid metabolism: involvement of the gut microbiota, liver, and adipose tissue**. *Front Endocrinol (Lausanne)* (2020) **11** 587189. PMID: 33343508
24. Doege H, Baillie RA, Ortegon AM, Tsang B, Wu Q, Punreddy S. **Targeted deletion of FATP5 reveals multiple functions in liver metabolism: alterations in hepatic lipid homeostasis**. *Gastroenterology* (2006) **130** 1245-58. PMID: 16618416
25. Zhan YT, An W. **Roles of liver innate immune cells in nonalcoholic fatty liver disease**. *World J Gastroenterol* (2010) **16** 4652-60. PMID: 20872965
26. Pan X, Chen B, Liu W, Li Y, Hu Z, Lin X. **Circulating iron levels interaction with central obesity on the risk of nonalcoholic fatty liver disease: a case-control study in Southeast China**. *Ann Nutr Metab* (2019) **74** 207-14. PMID: 30870854
27. Raghu H, Lepus CM, Wang Q, Wong HH, Lingampalli N, Oliviero F. **CCL2/CCR2, but not CCL5/CCR5, mediates monocyte recruitment, inflammation and cartilage destruction in osteoarthritis**. *Ann Rheum Dis* (2017) **76** 914-22. PMID: 27965260
28. Kisseleva T, Uchinami H, Feirt N, Quintana-Bustamante O, Segovia JC, Schwabe RF. **Bone marrow-derived fibrocytes participate in pathogenesis of liver fibrosis**. *J Hepatol* (2006) **45** 429-38. PMID: 16846660
29. Sysa P, Potter JJ, Liu X, Mezey E. **Transforming growth factor-beta1 up-regulation of human alpha(1)(I) collagen is mediated by Sp1 and Smad2 transacting factors**. *DNA Cell Biol* (2009) **28** 425-34. PMID: 19558215
30. Russell DW. **Fifty years of advances in bile acid synthesis and metabolism**. *J Lipid Res* (2009) **50** S120-125. PMID: 18815433
31. Caussy C, Hsu C, Singh S, Bassirian S, Kolar J, Faulkner C. **Serum bile acid patterns are associated with the presence of NAFLD in twins, and dose-dependent changes with increase in fibrosis stage in patients with biopsy-proven NAFLD**. *Aliment Pharmacol Ther* (2019) **49** 183-93. PMID: 30506692
32. Dawson PA. **Hepatic bile acid uptake in humans and mice: multiple pathways and expanding potential role for gut-liver signaling**. *Hepatology* (2017) **66** 1384-6. PMID: 28646543
33. Sun L, Cai J, Gonzalez FJ. **The role of farnesoid X receptor in metabolic diseases, and gastrointestinal and liver cancer**. *Nat Rev Gastroenterol Hepatol* (2021) **18** 335-47. PMID: 33568795
34. del Castillo-Olivares A, Campos JA, Pandak WM, Gil G. **The role of alpha1-fetoprotein transcription factor/LRH-1 in bile acid biosynthesis: a known nuclear receptor activator that can act as a suppressor of bile acid biosynthesis**. *J Biol Chem* (2004) **279** 16813-21. PMID: 14766742
35. Ding L, Yang L, Wang Z, Huang W. **Bile acid nuclear receptor FXR and digestive system diseases**. *Acta Pharm Sin B* (2015) **5** 135-44. PMID: 26579439
36. Lin BC, Wang M, Blackmore C, Desnoyers LR. **Liver-specific activities of FGF19 require Klotho beta**. *J Biol Chem* (2007) **282** 27277-84. PMID: 17627937
37. Balasubramaniyan N, Luo Y, Sun AQ, Suchy FJ. **SUMOylation of the farnesoid X receptor (FXR) regulates the expression of FXR target genes**. *J Biol Chem* (2013) **288** 13850-62. PMID: 23546875
38. Wang M, Liu F, Yao Y, Zhang Q, Lu Z, Zhang R. **Network pharmacology-based mechanism prediction and pharmacological validation of Xiaoyan Lidan formula on attenuating alpha-naphthylisothiocyanate induced cholestatic hepatic injury in rats**. *J Ethnopharmacol* (2021) **270** 113816. PMID: 33444723
39. Pellicciari R, Fiorucci S, Camaioni E, Clerici C, Costantino G, Maloney PR. **6alpha-ethyl-chenodeoxycholic acid (6-ECDCA), a potent and selective FXR agonist endowed with anticholestatic activity**. *J Med Chem* (2002) **45** 3569-72. PMID: 12166927
40. Sun L, Cai J, Gonzalez FJ. **The role of farnesoid X receptor in metabolic diseases, and gastrointestinal and liver cancer**. *Nat Rev Gastroenterol Hepatol* (2021) **18** 335-47. PMID: 33568795
|
---
title: Coordinated alternation of DNA methylation and alternative splicing of PBRM1
affect bovine sperm structure and motility
authors:
- Chunhong Yang
- Yao Xiao
- Xiuge Wang
- Xiaochao Wei
- Jinpeng Wang
- Yaping Gao
- Qiang Jiang
- Zhihua Ju
- Yaran Zhang
- Wenhao Liu
- Ning Huang
- Yanqin Li
- Yundong Gao
- Lingling Wang
- Jinming Huang
journal: Epigenetics
year: 2023
pmcid: PMC9988346
doi: 10.1080/15592294.2023.2183339
license: CC BY 4.0
---
# Coordinated alternation of DNA methylation and alternative splicing of PBRM1 affect bovine sperm structure and motility
## ABSTRACT
DNA methylation and gene alternative splicing drive spermatogenesis. In screening DNA methylation markers and transcripts related to sperm motility, semen from three pairs of full-sibling Holstein bulls with high and low motility was subjected to reduced representation bisulphite sequencing. A total of 948 DMRs were found in 874 genes (gDMRs). Approximately $89\%$ of gDMR-related genes harboured alternative splicing events, including SMAD2, KIF17, and PBRM1. One DMR in exon 29 of PBRM1 with the highest 5mC ratio was found, and hypermethylation in this region was related to bull sperm motility. Furthermore, alternative splicing events at exon 29 of PBRM1 were found in bull testis, including PBRM1-complete, PBRM1-SV1 (exon 28 deletion), and PBRM1-SV2 (exons 28–29 deletion). PBRM1-SV2 exhibited significantly higher expression in adult bull testes than in newborn bull testes. In addition, PBRM1 was localized to the redundant nuclear membrane of bull sperm, which might be related to sperm motility caused by sperm tail breakage. Therefore, the hypermethylation of exon 29 may be associated with the production of PBRM1-SV2 in spermatogenesis. These findings indicated that DNA methylation alteration at specific loci could regulate gene splicing and expression and synergistically alter sperm structure and motility.
## Background
Semen quality is an important factor in the evaluation of bull fertility. The common parameters for evaluating semen quality include the semen volume per ejaculate, sperm concentration, sperm motility, post-thaw cryopreserved sperm motility, and sperm deformity rate [1]. Sperm motility is the main index used to assess the semen quality of a bull and determine whether semen can be frozen, affecting bull semen production and economic benefit. Sperm motility is also influenced by many factors, including genetic and epigenetic influences [2–5]. Therefore, the individual differences in semen quality should be identified precisely.
Although various methods are being used to assess bull semen quality from phenotypic and functional parameters, the accurate prediction of semen quality remains a major challenge. Recently, molecular mechanisms underlying the difference in semen quality have received considerable attention. One possible mechanism is DNA methylation, which could explain the phenotypic differences in sibling bulls. In the bull breeding programme, many full-sibling Holstein bulls were obtained via super-ovulation and embryo transfer. Despite being genetically similar, full-sibling bulls usually exhibit different semen quality phenotypes, including volume per ejaculate, sperm concentration, percentage of motility, progressive motility, mean path velocity, percentage of abnormal sperm, intracellular calcium and P25b levels [3,6,7]. Epigenetic regulation may play an irreplaceable role in semen quality. DNA methylation is a stable epigenetic modification of a genome and a transcriptional factor that regulates gene expression, which plays a role in cell reprogramming, tissue differentiation, and diseases [8,9]. The experimental design of twins plays an essential role in estimating the contribution to the inherent genomic sequence versus the epigenetic effects induced by environmental conditions in the expression of complex traits [10,11]. In the case of full siblings, we cannot completely rule out the influence of genetic components because they may have different genetic combinations. Nonetheless, the full-sibling Holstein bulls are an ideal model for investigating the relationship between DNA methylation and semen quality.
High-throughput genomic analyses, such as reduced representation bisulphite sequencing (RRBS), methylated DNA immunoprecipitation sequencing, or whole-genome bisulphite sequencing, allow the study of sperm DNA methylation from a genome-wide perspective. The relationship between DNA methylation patterns and sperm quality, including motility, morphology, and DNA fragmentation, was investigated in human [12–14] and cattle sperm [15]. The abnormality of sperm DNA methylation is closely related to bull semen quality (sperm morphology and motility) and fertility [16–18]. Several studies have revealed the hypomethylation pattern of the bull spermatozoa genome [19,20]. Furthermore, recent studies have shown that the sperm DNA methylation of the bull is closely related to sperm motility [21–23]. However, no functional evidence of the mechanism of DNA methylation in influencing spermatogenesis and sperm motility has been found.
Emerging evidence suggests that DNA methylation not only affects transcription, but also regulates alternative splicing (AS) [24–26]. The DNA methylation level of exons, particularly splicing sites, is higher than that of introns on both sides. About $22\%$ of alternative exons are subjected to DNA methylation [24]. Although DNA methylation and AS regulation are common in spermatogenesis, to our knowledge, there is no evidence that the synergistic effects of DNA methylation and AS play a role in spermatogenesis.
We observed that three pairs of full-sibling Holstein bulls had similar genomic breeding values, but showed different sperm motility phenotypes. Therefore, we speculate that epigenetic mechanisms may play an important role. At the same time, considering the relationship between DNA methylation and AS, we further hypothesize that the synergistic regulation of DNA methylation and AS may be an important mechanism. The purpose of this study was to identify the genomic methylation sites that differ between full-sibling bulls with high and low sperm motility by RRBS, and to clarify the relationship between DNA methylation and AS in combination with previously obtained transcriptome data, and finally to find examples of methylation affecting gene splicing and expression, and thus influencing sperm structure and motility. Here, we provided evidence of the mechanism by which coordinated alterations of epigenetic regulation and RNA splicing of PBRM1 promote spermatogenesis and semen quality.
## Animals and sample collection
Experiment 1: The same batch of fresh semen samples from three pairs of full-sibling Holstein bulls were collected to analyse the genome-wide sperm DNA methylation and identify differentially expressed mRNAs and lncRNAs using RNA-seq as described in our previous study [7]. Genomic breeding values (gEBV) were computed by the Animal Genetics Department of INRA (France National Institute for Agronomic Research; $\frac{14}{3}$ release; https://idele.fr/detail-article/ibl-2014-8-evaluation-internationale-daout-2014). The gEBVs of the total merit index (called index de synthèse global in France or ISU) of each full-sibling Holstein bull pair obtained from Shandong OX Livestock Breeding Co., Ltd., were almost the same (Table S1), indicating that their genome information is similar, which can rule out the false detection of genomic DNA methylation caused by the genetic variation of C > T. The age of full-sibling Holstein bulls ranged from 4 to 6 years, with complete semen collection records, generally two times a week. Semen quality was evaluated in accordance with our national standards (NY-1234-2018) [27]. Among them, sperm motility assessment method is as follows. The semen and diluent were raised 5–10 μL on the slide, covered the slide, and the motility was examined by using an Olympus phase-contrast microscope at 37°C constant temperature to obtain the subjective assessment data. Based on the average sperm motility calculated by counting the semen quality records of the last two years, three pairs of full-sibling Holstein bulls, two of which had different sperm motility, were obtained and allocated into the high motility group (hereafter referred to as H1, H2, and H3; average sperm motility = $68.7\%$; average ISU = 141) and the low motility group (referred to as L1, L2, and L3; average sperm motility = $60\%$; average ISU = 140.7; Table S1). Bulls with sperm motility above $68\%$ were defined as the H group, whereas those with sperm motility below $62\%$ were defined as the L group, and the difference in sperm motility between the paired individuals was at least 8 percentage points. Three millilitres of fresh semen sample of each bull was collected. After evaluating semen quality, each semen was washed with PBS three times, and the pellet was resuspended in 1 mL of PBS and mixed gently. Each sperm suspension was divided into two parts, one for DNA extraction for RRBS and bisulphite sequencing PCR (BSP) analyses and the other for RNA extraction for RNA-seq [7].
Experiment 2: Fresh semen samples collected from about 3-year-old adult bulls ($$n = 6$$) were used to prepare the semen smear. The fresh semen was diluted, and 10 μL of fresh semen was placed and evenly spread onto a slide, dried at room temperature for 2–3 h, and used for immunofluorescence (IF) staining or stored at −80°C for future experiments.
Experiment 3: Testes were collected from 12 Holstein bulls at varying ages, including 2 days old (abbreviation 2 D, $$n = 3$$), 12 months old (12 M, $$n = 3$$), 2 years old (2 Y, $$n = 3$$), and 4 years old (4Y, $$n = 3$$), and each age stage included three bulls. The collected testes were quickly stored in liquid nitrogen to extract total RNA and protein for qRT-PCR and Western blot, respectively.
All protocols for collecting samples from bulls were reviewed and approved by the Animal Care and Use Committee of Shandong Academy of Agricultural Sciences. The experiment was conducted under the regulations and guidelines established by this committee.
## RRBS
Bull sperm DNA was extracted using a high-concentration salt protocol [28]. The quantity and quality of purified DNA were assessed using Nanodrop ND-2000 (Thermo Fisher Scientific, Waltham, USA). Genomic DNA methylation was sequenced using RRBS [29,30]. In brief, an appropriate amount of sperm genomic DNA with A$\frac{260}{280}$ ratio of 1.8:2.0 was digested using the restriction enzyme MspI (New England Biolabs, MA, USA), which specifically recognized shear C|CGG sequences. In addition, this enzyme was not sensitive to methylation modification. The T4 DNA polymerase and Klenow enzyme were used to flatten the hanging structures with different sizes into flat ends. The 300–400 bp DNA fragment was screened and purified using Agencourt AMPure XP beads (Beckman, USA). A base was added to the flat end of the sequence. The DNA fragments with a sticky end were attached to the methylated junction sequence using the Paired-End DNA Sample Prep kit (Illumina) and purified. Therefore, 40–220 bp enzymatic fragments were cut for recovery by using gel electrophoresis and treated with bisulphite to change the unmethylated C into T, whereas the methylated C remained unchanged. These fragments were amplified and purified using the MinElute PCR Purification kit (Qiagen, Germany). Finally, paired-end 2 × 150 bp sequencing was performed using the Illumina Hiseq 4000 platform (Hangzhou Lianchuan Biotechnology Co., Ltd.).
## RRBS data analysis
The raw data of RRBS has been uploaded to NCBI SRA, and the accession number was PRJNA818321. Raw sequencing data were filtered and assembled, and low-quality reads (Q ≤ 10 and N > $5\%$) were filtered out from raw reads. Original and effective sequencing quantities were counted, and Q20, Q30, and GC contents were comprehensively evaluated. High-quality and clean reads were mapped onto the cattle reference genome (Bos_Taurus UMD3.1) using the Bowtie2.1.0 program in the BSgenome software package to ensure accurate and reliable analysis results [31].
The methylation level of 5mC (CpG, CHG, and CHH; H = A/C/T) was classified on the basis of its location, including the promoter, exon, intron, and intergenetic regions. DMRs were determined using the R package-Methyl Kit with default parameters (1000 bp slide windows, 500 bp overlap, P value ≤ 0.05) [32]. The average ratio of 5mC in every DMR of each group was calculated. DMR with |log2FC| ≥ 1 and P ≤ 0.05 was considered significant. DMR-related genes (DMGs) were clustered on the basis of the Gene Ontology (GO) annotation database and KEGG pathway analysis through DAVID (https://david.ncifcrf.gov/) and Pathview (https://pathview.uncc.edu/) [33,34].
## BSP
In addition, 500 ng of genomic DNA from each fresh semen sample was treated using the BisulFlash DNA modification kit (Epigentek, USA) under the manufacturer’s protocol [35]. The BSP primers of DMGs were designed using Methyl Primer Express Software v1.0 (https://www.urogene.org/methprimer/, Table S2). PCR was performed using the TaKaRa LA Taq® (Code No. RR02MA, TAKARA, Japan), and the following components were mixed to prepare the PCR reaction: 0.25 μL of TaKaRa LA Taq (5 U/μL), 2.5 μL of 10× LA Taq Buffer II (Mg2+ Plus), 4 μL of dNTP Mixture (2.5 mM each), 1 μL of BS conversed genomic DNA, 1 μL of PBRM1-BSP-F/R, and 15.25 μL of ddH2O. The PCR reaction was set as follows: 94°C 1 min, 35 cycles of (98°C, 10s; 68°C, 15s and 72°C, 15s), 72°C 10 min, 4°C hold. PCR products were purified using the Gel/PCR extraction kit (Tiangen, Beijing, China), cloned into the pEASY-T3 vector (TransGen, Beijing, China), and transformed into *Escherichia coli* DH5α cells for clone sequencing. Only the sequences derived from clones with > $95\%$ cytosine conversion were analysed. The percentage of DNA methylation was calculated by counting the number of methylated CpGs from the total number of CpG sites in individual clones. The reference sequence of a specific gene was used as a reference for methylation status analysis by using BiQ Analyser v0.7.
## qRT-PCR
Total RNA was extracted using the TaKaRa MiniBEST Universal RNA Extraction Kit (TaKaRa, Dalian, China). RNA was quantified and quality assessed using Nanodrop ND-2000. Complementary DNA was synthesized using the PrimeScriptTM RT reagent Kit with gDNA Eraser (TaKaRa, Dalian, China). The qRT-PCR primers (Table S2) were designed using Primer-blast (NCBI) and performed using TB Green Premix Ex Taq (RR820A, TaKaRa, Dalian, China). The reaction system contains 10 μL of TB Green Premix Ex Taq II (Tli RNaseH Plus, 2×), 0.8 μL of PCR Forward Primer (10 μM), 0.8 μL of PCR Reverse Primer (10 μM), 2 μL of cDNA (<100 ng), and 6.4 μL of DEPC water, reaching 20 μL. qPCR was performed on LightCycler 480 II, and the PCR amplification procedure was as follows: 95°C for 30s to activate the reaction and 95°C for 5 s and 60°C for 30s for 40 cycles. The collection of the melting curve was performed in accordance with the program of the instrument. Raw qRT-PCR Ct values were normalized against the geometric mean of the β-actin gene [6,7]. The 2−ΔΔCt method was applied for calculating the relative gene expression level [36]. Each experiment was performed in triplicate.
## Identification of AS events associated with sperm motility
The RNA-seq data of six semen samples of Holstein bulls were obtained from our previous study (NCBI SRA database, accession number SRP158901), whose semen samples were of the same batch as those used in RRBS [7]. Qualitative analysis and statistics of AS events of the transcriptome of each sample were performed using ASprofile (https://ccb.jhu.edu/software/ASprofile/) based on the gene model predicted by Cufflinks. AS events were primarily divided into five classical types: including or excluding single (SKIP_ON/OFF) and multiple exons (MSKIP_ON/OFF), including or excluding single (IR_ON/OFF) and multiple (MIR_ON/OFF) introns, alternative exon (AE) ends, alternative TSS, and alternative transcription termination site (TTS, Fig. S1). The other five types belong to fuzzy boundary shear types, including approximate SKIP (XSKIP), approximate MSKIP (XMKIP), approximate IR (XIR), approximate MIR (XMIR), and approximate AE (XAE). The difference in AS transcript expression between the H and L groups with |log2FC| ≥ 1 and $P \leq 0.05$ was considered significant.
## Western blot
The total protein of the bull testis was extracted using RIPA lysis buffer (including $1\%$ PIC) and separated using $10\%$ SDS–PAGE. Immunoblot analyses were performed using rabbit anti-PBRM1 (1:1000, ab196022, Abcam), β-actin rabbit (1:10,000, AC026, ABclonal), or anti-α-tubulin mouse mAb (1:5000, AC012, ABclonal) with horseradish peroxidase-conjugated secondary antibodies (goat anti-rabbit IgG, ZB-2301 ZSGB-BIO; goat anti-mouse IgG, ZB-2305, ZSGB-BIO). Immobilon® Western Chemiluminescent HRP Substrate was used as the chemiluminescence detection reagent (P90719, Millipore, USA).
## Immunofluorescence
For IF staining of PBRM1, the fresh bull semen spread onto polylysine-coated slides was air dried, fixed using $4\%$ paraformaldehyde for 5 min, washed three times in PBS, permeated in $0.2\%$ Triton X-100 for 15 min, and blocked using $5\%$ BSA for 1 h at room temperature. After washing with PBS three times, the samples were incubated with primary antibodies diluted with PBS (including anti-rabbit PBRM1 [1:100] (AB196022, Abcam, Cambridge, UK) or mAb 414 antibody (1:100, 902,097, BioLegend, San Diego, USA)) at 4°C overnight. Then, samples were incubated with secondary antibodies, including Alexa Flour® 488-conjugated goat anti-rabbit IgG (1:150; ZF-0511, ZSGB-BIO, Beijing, China) for 1.5 h at 37°C. For co-localization analysis of the nuclear pore complex (NPC) and PBRM1, the mAb 414 antibody was visualized by using a TRITC-conjugated secondary antibody, and PBRM1 labelling was visualized by using a FITC-conjugated secondary antibody. The samples were incubated in DAPI staining solution (C1005, Beyotime, China) for 10 min at room temperature and washed three times with PBS. The samples were added with an anti-fluorescence quencher and placed in a cover glass, and the edge was sealed with nail oil. Immunofluorescence staining was imaged at 200× magnification using an inverted fluorescent microscope (IX73, OLYMPUS, Japan).
## Transmission electron microscopy and immunoelectron microscopy
The ultrastructure of sperm was detected by TEM. Based on the density of fresh bull semen, approximately 5 × 106 sperm was transferred into a 1.5 mL tube and then centrifuged at 4000 × g for 2 min to obtain a pellet. The pellet was fixed in $2.5\%$ glutaraldehyde with 0.1 M cacodylate buffer (pH 7.4), post-fixed with $1\%$ OsO4, dehydrated through a graded series of ethanol solutions to $100\%$ ethanol, and embedded in London Resin (LR) White. Thin sections (50–70 nm) were cut on Leica EM UC7 ultramicrotome (Leica, Germany) and stained with uranyl acetate and lead citrate. All specimens were examined using a transmission electron microscope (HT7800, HITACHI, Japan) at 80 kV.
For electron microscopic immunolocalization, the bull sperm pellet was fixed with $4\%$ paraformaldehyde and $0.01\%$ glutaraldehyde for 12–14 h, washed with PBS five times at 4°C (15 min/time), and then dehydrated through a graded series of ethanol solutions to $100\%$ ethanol. The sample was infiltrated with and embedded in LR white, which was polymerized by UV light for 24 h at −20°C. Thin sections (50–70 nm) were cut using a Leica EM UC7 ultramicrotome (Leica, Germany) and scooped up with nickel mesh. Then, slices were blocked with $1\%$ BSA in PBS and probed with an anti-PBRM1 monoclonal antibody overnight at 4°C. After washing with PBS, sections were incubated for 1 h at room temperature using a 10 nm gold-conjugated secondary antibody, namely, goat anti-rabbit IgG. The sections were washed with distilled water and stained with uranyl acetate and lead citrate before the examination.
## Statistical analysis
Statistical significance of semen quality was measured using Student’s paired t-test. The statistical significance of the gene expression of qRT-PCR and the methylation level of BSP were tested using the Linear Model in the SAS v9.0, and comparisons were performed using Duncan’s multiple-range test. In addition, the P value was adjusted using the Bonferroni method. Chi-square test was used to check the association between loss expression of PBRM1 and sperm tail breakage. A P-value less than 0.05 was considered statistically significant.
## Genome-wide DNA methylation was different between high- and low-sperm-motility groups of bull
Three pairs of full-sibling Holstein bulls with high and low sperm motility were selected to analyse the relationship between DNA methylation and sperm motility. RRBS produced approximately 303 million clean reads and 24.2 Gb of high-quality data (Table S3). The high-quality clean reads were mapped onto the cattle reference genome (*Bos taurus* UMD3.1, Table S4). PCA analysis showed that PC2 could clearly distinguish H from the L group, although differences were observed among individuals (Figure 1a). The distribution of 5mC in chromosomes of bulls is shown in Figure 1b. CpG methylation modification was dominant in DNA methylation types, but CHG and CHH methylation also existed on the bull sperm genome (Figures 1B–C). Figure 1.Overview of sperm DNA methylation and differential DNA methylation regions between the H and L groups. ( a) PCA analysis on RRBS data of the H and L groups. ( b) Distribution of 5mC in chromosomes. The colour from inside to outside the circle represents the sample order: H1, H2, H3, L1, L2, and L3. ( c) Methylation ratio of mCpG, mCHG, and mCHH in gene elements. ( d) *Statistical analysis* of gDMRs in the promoter, exon, and intron of genes. The percentage indicates the proportion of each kind of DMRs in the total number of DMRs. ( e) Heatmap of gDMRs. Pink and blue bars represent the upregulated and downregulated methylation levels, respectively.
A total of 2308 DMRs that differentially methylated regions in the sperm DNA between the H and L groups were obtained ($P \leq 0.05$, |log2(FC)|≥1), including 948 DMRs located in 873 genes (referred to as gDMRs), 1354 DMRs located in repeat elements, and six DMRs located in microRNA regions. About $28\%$ of gDMRs were found in the gene promoter, $41\%$ in exon, and $31\%$ in intron (Figure 1d, Table S5). The heatmap based on gDMRs showed that the DNA methylation patterns of three individuals within the same group were similar, showing good repeatability, whereas the overall gene methylation pattern of the H group was significantly different from that of the L group (Figure 1e).
Gene Ontology enrichment analysis revealed that gDMR-related genes (DMGs) were involved in several molecular functions, including RNA transcription (GO:0006351), cell adhesion (GO:0007155), cell differentiation (GO:0030154), and calmodulin binding (GO:0005516; $P \leq 0.05$, Fig. S2A). Focal adhesion, MAPK, and calcium signalling pathways were enriched by KEGG pathway analysis or Pathview (https://pathview.uncc.edu/home; Fig. S2B and Fig. S3).
## Sperm-motility-related gDMRs may be associated with as events
Alternative splicing events of sperm transcriptome between the H and L group were compared. In total, 1876 AS events in 874 genes were obtained, and the resulting transcripts were significantly and differentially expressed between the H and L groups ($P \leq 0.05$, |log2(FC)|≥1). Alternative TSS and alternative TTS accounted for $39.5\%$ and $32.3\%$, respectively, followed by exon skipping (SKIP, $14.4\%$) and AE ends ($8.9\%$, Fig. S4). This result indicated the critical role of abundant AS events during bull sperm maturation and motility.
Interestingly, AS events were observed in gDMRs, and 781 ($89\%$) gDMRs exhibited AS events. Of which, 123 AS transcripts within 57 genes were significantly and differentially expressed between the H and L groups ($P \leq 0.05$, |log2(FC)|≥1), including spermatogenesis-related genes, such as SMAD2, KIF17, RAB22A, CCDN1, and PBRM1 (Figure 2, Table S6). SMAD2 can be activated by Nodal signalling, and it can promote the proliferation of mouse spermatogonial stem/progenitor cells [37]. One DMR in exon 2 of SMAD2 had a lower methylation level in H than in the L group (Figures 2A and 2B, Table S6). Furthermore, SMAD2-complete was more highly expressed in H than in the L group ($P \leq 0.05$), whereas SMAD2-SV1 (exon 2 deletion) was more highly expressed in L than in the H group, but this difference was not significant (Figures 2A and 2C). These results indicated that the DNA methylation level of exon 2 of SMAD2 may affect its AS (Figures 2A–C). Similarly, KIF17 was specifically expressed in mouse testis and localized primarily in the principal piece of the sperm tail [38]. The 5mC ratio in exon 15 of KIF17 was higher in H than in the L group (Figures 2D and 2E). Furthermore, KIF17-complete was more highly expressed in H than that in the L group ($P \leq 0.05$), but the expression of KIF17-SV1 (18 bp deletion of exon 15) had no difference between the H and L groups (figure 2f). These results indicated that sperm DNA methylation plays an important role in spermatogenesis by modulating gene AS. Figure 2.gDMR in SMAD2 and KIF17 is related to AS events. ( a) Schematic diagram of the relationship between gDMR and alternative splicing of exon 2 of SMAD2. ( b) DNA methylation level in exon 2 of SMAD2 in the H and L groups. ( c) Expression of SMAD2-complete and SMAD2-SV1 of bull sperm from RNA-seq. ( d) Schematic diagram of the relationship between gDMR and alternative splicing of exon 15 of KIF17. ( e) DNA methylation level in exon 15 of KIF17 in the H and L groups. ( f) Expression of KIF17-complete and KIF17-SV1 of bull sperm from RNA-seq. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ ‘ ns’ represents $P \leq 0.05.$
## High methylation level in exon 29 of PBRM1 may promote the generation of alternatively spliced transcripts
PBRM1, encoding a subunit of ATP-dependent chromatin-remodelling complexes, is required to stabilize the SWI/SNF chromatin remodelling complex, which is involved in mouse spermatogenesis [39–41]. Therefore, PBRM1 might play an essential role in bovine spermatogenesis. In our study, one DMR (chr22: 48,830,591–48,830,747) located in exon 29 of PBRM1 should be further studied because the DMR was hypermethylated in the sperm of bulls from the H group, and the 5mC ratio was $74\%$, which was the highest among all gDMRs (Table S5). The BSP results indicated that the H group had a higher methylation level in a partial region of exon 29 than the L group ($P \leq 0.05$, Figures 3A and 3B), which were consistent with the results obtained from the RRBS. Furthermore, three alternative transcripts of PBRM1 were obtained from sperm transcriptome between the H and L groups (Table S6). Figure 3.Differential methylation in exon 29 of PBRM1. ( a) PBRM1 gene structure and validation of the DNA methylation level in exon 29 using BSP (six samples). ( b) DNA methylation ratio of PBRM1 between the H and L groups. Means with different lowercase letters within the same column are different ($P \leq 0.05$).
The bovine PBRM1 gene includes 32 exons and 31 introns. Here, exon skipping events near exon 29 of the bovine PBRM1 gene was observed in the NCBI database (Figure 4a). Therefore, we hypothesized that DNA methylation of exon 29 would affect AS of PBRM1. Consequently, three transcripts, namely, PBRM1-complete, PBRM1-SV1 (exon 28 deletion), and PBRM1-SV2 (exons 28–29 deletion), were identified in bull testes at all stages of sexual maturity (Figure 4b and Supplemental material_original images). PBRM1-SV1 was the primary transcript in different development stages of testes, and the expression level of PBRM1-SV2 gradually increased with age (Figure 4c). PBRM1-complete was rarely and stably expressed across all stages of bull testes (Figure 4c). Compared with newborn bulls, the mRNA expression level of PBRM1 in the testes of sexually mature bulls was significantly increased (Figure 4d). These results indicate that the expression of PBRM1-spliced transcripts may be related to testicular development and spermatogenesis in bulls. Figure 4.Validation of alternative splicing in exon 29 of PBRM1. ( a) Schematic diagram of three spliced transcripts (PBRM1-complete, PBRM1-SV1, and PBRM1-SV2) around exon 28 and 29 of PBRM1 (Bos_Taurus UMD3.1). ( b) Three spliced transcripts of PBRM1 in bull testis were verified by RT-PCR using PBRM1-AS-F/R primers. ( c) qRT-PCR results of the relative mRNA expression of PBRM1-complete, PBRM1-SV1, and PBRM1-SV2 in bull testes. ( d) qRT-PCR results of the total mRNA expression of PBRM1 at different stages of sexual maturity. ( e) Western blot results of PBRM1 protein expression in bull testes and sperm, and β-actin and α-tubulin used as the control. ( f) Expression of PBRM1 isoforms in bull testes at different stages of development. ( g) Expression of PBRM1 isoforms in bull sperm. Means with different lowercase letters within the same column represent significant differences ($P \leq 0.05$). 2 D represents the testes of 2-day-old bulls; 10 M represents the testes of 10-month-old bulls; 2 Y represents the testes of 2-year-old bulls; 4 Y represents the testes of 4-year-old bulls.
Two spliced isoforms of PBRM1, namely, PBRM1-isoform1 and PBRM1-isoform2, were detected in bull testes and sperm by Western blot (Figure 4e and Supplemental material_original images). PBRM1-isoform1 and PBRM1-isoform2 corresponding to PBRM1-SV1 and PBRM1-SV2 were predicted with a molecular weight of approximately 194 and 185 kD, respectively, via SMS2 (https://www.detaibio.com/sms2/protein_mw.html). Furthermore, the expression level of PBRM1-isoform1 was higher than PBRM1-isoform2 in the testes and sperm; however, the expression level of PBRM1-isoform2 in the testes significantly increased with age (figure 4f). No significant difference was observed between PBRM1-isoform1 and PBRM1-isoform2 in sperm (Figure 4g). This finding indicates that the generation of PBRM1-SV2 plays an important role in spermatogenesis. Therefore, we speculate that DNA methylation of exon 29 of bovine PBRM1 may affect the AS of exon 29 during spermatogenesis.
## PBRM1 is located in the redundant nuclear envelope (RNE) of bull sperm, and the absence of expression affects sperm motility caused by sperm tail breakage
Immunofluorescence staining with anti-PBRM1 antibodies revealed two prominent signals in the head-to-tail connection of all normal bull sperm, which were symmetrical on both sides (Figure 5a). Abnormal sperm without a tail (ASWT) was verified as normal heads (NH) or abnormal head (AH) by DAPI (4´,6-diamidino-2-phenylindole) staining and bright-field photography. These abnormal sperm were divided into four categories based on whether the sperm head morphology was normal and whether PBRM1 was expressed: ASWT_NH_PBRM1_positive, ASWT_NH_PBRM1_negative, ASWT_AH_PBRM1_positive, and ASWT_AH_PBRM1_negative (Figure 5b). The statistical results showed that ASWT_NH_PBRM1_negative sperms ($$n = 343$$) accounted for $49\%$ of the total ASWT_NH sperms ($$n = 679$$), whereas ASWT_AH_PBRM1_negative sperms ($$n = 338$$) accounted for $94\%$ of the total ASWT_AH sperms ($$n = 361$$, Figure 5c). Most abnormal spermatozoa had no PBRM1 expression, and sperm tails were broken. The loss of PBRM1 in the head-to-tail connection of sperm is significantly associated with sperm tail breakage ($P \leq 0.05$), thereby affecting sperm motility. Figure 5.Expression and localization of PBRM1 in sperm of adult bull. ( a) PBRM1 is located in the head-to-tail connection of normal sperm of bull. The sperm head was stained with DAPI (blue). PBRM1 was detected by anti-PBRM1 antibody (Green). BF, bright field. Scale bar: 20 µm. ( b) The immunofluorescence test showed four types of sperm, including ASWT_NH_PBRM1_positive, ASWT_NH_PBRM1_negative, ASWT_AH_PBRM1_positive, and ASWT_AH_PBRM1_negative. ASWT: abnormal sperm without tail. NH: normal heads. AH: abnormal head. Scale bar: 20 µm. ( c) Comparison of the percentage of the four types of sperm. The percentage of AH_PBRM1_Negative sperm was higher than AH_PBRM1_Positive. Chi-square test, $P \leq 0.05.$
Immunofluorescence staining showed that PBRM1 is significantly colocalized with the NPC protein, which serves as a marker of the RNE in bull sperm (Figure 6a). Transmission electron microscopy showed two symmetrical redundant nuclear membranes in the head-to-tail connection of sperm (Figure 6b). Furthermore, immunoelectron microscopy showed that the colloidal gold signals to anti-PBRM1 antibody were relatively concentrated and located in the head-to-tail connection of bull sperm (Figure 6c). The results indicate that PBRM1 is expressed in the RNE of bull sperm. Figure 6.PBRM1 proteins are enriched in the region of the redundant nuclear envelope in bovine sperm. ( a) PBRM1 (green) is significantly colocalized with NPC (red) in bull sperm. The scale bar is 20 μm. ( b) Ultrastructure of the head-to-tail connection region of bull sperm by transmission electron microscopy. The RNE structure of bull sperm was highlighted by the black arrow. ( c) Immunogold labelling of bull sperm with an anti-PBRM1 antibody. Positive staining in the head-to-tail connection area was highlighted by the black arrow. N, nuclei; C, centriole.
## Discussion
DNA methylation and RNA transcriptome analyses were used to screen the differential DNA methylation regions of full-sibling Holstein bulls with high and low sperm motilities and to identify the pathways and genes related to sperm motility. Here, 2308 DMRs (including 948 gDMRs, 1354 repDMRs, and six miDMRs), which are involved in RNA transcription, cell adhesion, cell differentiation, and calcium ion binding, are identified. Furthermore, we highlighted that the alteration in DNA methylation and AS of PBRM1 were associated with the structure and motility of bovine sperm. These results provide valuable data for future biomedical research and genomic and epigenomic studies of semen quality, which can be used to uncover the molecular basis underlying the economic value of sperm quality traits in bulls.
The sperm DNA methylation profile is particularly important because it is one of the factors that control the expression of imprinted genes, which are essential for foetal development and foetal growth [42]. One of the biggest concerns is that abnormal sperm epigenetic defects may be passed onto offspring and may affect their susceptibility to disease [43]. Considerable evidence has pointed out the effects of DNA methylation on male fertility and semen quality. For example, abnormalities of DNA methylation are associated with human sperm parameters, including sperm concentration, motility, and morphology [44]. Global sperm DNA methylation is associated with sperm concentration and sperm motility, but it is not associated with sperm vitality or morphology [45]. Sperm cells of low-fertility bulls have a less dense chromatin structure, higher levels of DNA damage, and higher methylation levels than those of high-fertility bulls [46]. In our study, a strong association was observed between global sperm DNA methylation and sperm motility in full-sib Holstein bulls.
Twin models are more advantageous in epigenetic studies because they share almost all genetic variations and many environmental factors, and they form natural matching controls [47]. Several studies have reported the relationship between genome-wide DNA methylation profiles and sperm quality and identified several DMRs implicated in motility by using the twins as a study model. For example, by using methyl-sensitive enrichment and microarray analyses, 580 differentially methylated loci, including fertility-related QTLs, have been identified in four pairs of monozygotic twin bulls, which have daughters with incongruous diverging performances [18]. The analyses of interindividual variations of 28 bull sperm DNA methylation and whole-genome BSP data indicate that the variably methylated regions are related to sperm motility and reproduction [23]. A total of 528 DMRs associated with embryonic development, organ development, reproduction, and the nervous system were obtained in two monozygotic twin bulls with moderately different sperm qualities [48]. In the present study, 948 gDMRs encompassing 837 genes were enriched in the RNA transcription, cell adhesion, cell differentiation, and calmodulin-binding GO terms. KEGG pathway analysis has identified pathways involved in spermatogenesis and sperm motility. For example, the focal adhesion pathway (24 DMGs) contributes to the acrosome integrity and remodelling of the cytoskeleton during sperm capacitation and spermatid adhesion [49,50]. The MAPK signalling pathway (26 DMGs) regulates the dynamics of tight and adhesion junctions and the proliferation and meiosis of germ cells [51]. The calcium signalling pathway (21 DMGs) regulates the sperm motility of mammals by controlling the calcium ion concentration and activity of calcium-dependent proteins [52]. These findings indicate that the divergence of the sperm motility of full-sibling bulls may be due to differential methylation of multiple genes participating in the critical signalling pathways related to spermatogenesis and sperm maturation.
Spermatogenesis is a highly coordinated process that requires tightly regulated gene expression programmed by epigenetic modifiers, including DNA methylation and chromatin remodelling [53]. DNA methylation plays a crucial role in correct sperm functionality. Methylation variation in genes is functionally related to sperm DNA organization and maintenance [15]. In this study, we found that the significant DMR harbours exon 29 of PBRM1, which is related to sperm motility. Furthermore, PBRM1 encodes a subunit of ATP-dependent chromatin-remodelling complexes, which are required to stabilize the SWI/SNF chromatin remodelling complex [40,54,55]. The SWI/SNF chromatin remodelling complex is involved in the spermatogenesis and deficiency of the catalytic subunit of the SWI/SNF chromatin remodelling complex, which results in a meiotic arrest during mouse spermatogenesis [39,41].
Furthermore, we found that the mRNA expression level of PBRM1 is higher in adult bull testis than in newborn bulls. The expression level of PBRM1-SV1 and PBRM1-SV2 was also significantly higher than that of PBRM1-complete in the testis. DNA methylation in promoter regions is a well-characterized epigenetic marker negatively related to gene expression regulation. However, PBRM1 expression is positive with DNA methylation in exon. Interestingly, emerging evidence has shown that DNA methylation also regulates gene AS. Intragenic DNA methylation operates in exon to modulate alternative RNA splicing [24,56]. A recent study has demonstrated that changes in the methylation pattern of alternatively spliced exons, but not constitutively spliced exons or introns, altered inclusion levels using deactivated endonuclease Cas9 fused with enzymes that methylate or demethylate [57]. In the present study, two splice variants, namely, PBRM1-SV1 and PBRM1-SV2, are considered primary transcripts, and they are highly expressed in adult bull testes. This finding indicates that DNA methylation of exon 29 in PBRM1 results in AS of PBRM1 and generates the high expression of splice variants, which warrants further investigation. In addition, we first found that PBRM1 was located in the RNE of bull sperm. The RNE contains an intracellular Ca2+ store, which can provide Ca2+ for the axoneme to enhance sperm motility during hyperactivation [58]. We observed that the lack of PBRM1 in sperm is associated with sperm tail breakage, indicating that PBRM1 is related to sperm motility. We successfully identify an example that DNA methylation alteration at specific loci regulates gene splicing and expression, thereby altering sperm structure and motility.
It should be pointed out that one limitation of this study is that we used only motility and did not measure other phenotypic parameters, such as membrane and DNA integrity which are important factors in sperm quality. Using these parameters can be more informative and potentially reveal more interesting results.
## Conclusions
Our findings indicate that alternations of DNA methylation and RNA AS affect sperm motility in Holstein bull; thus, investigating epigenetic changes is important to understand the genetic mechanism during spermatogenesis. We also found that DNA methylation affects AS of PBRM1, which is necessary for spermatogenesis and maturation, thereby generating phenotypic diversity.
## Disclosure statement
No potential conflict of interest was reported by the author(s)
## Ethical approval statement:
All protocols for collecting samples from bulls were reviewed and approved by the Animal Care and Use Committee of Shandong Academy of Agricultural Sciences. The experiment was conducted under the regulations and guidelines established by this committee.
## Availability of data and materials
The raw data of RRBS has been uploaded to the NCBI SRA, and the accession number was PRJNA818321.
## Author contributions
J.H. conceived and designed the experiments. C.Y., X.Y., X.W., X.C.W., J.W., Z.J., and N.H. performed experiments and analyzed data. Q.J., Y.Z., W.L., L.W., Y.L., and Y.G. collected bull samples and prepared materials. C.Y. and J.H. wrote the manuscript. All the authors read and approved the final manuscript.
## Supplementary material
Supplemental data for this article can be accessed online at https://doi.org/$\frac{10.1080}{15592294.2023.2183339}$
## References
1. Druet T, Fritz S, Sellem E. **Estimation of genetic parameters and genome scan for 15 semen characteristics traits of Holstein bulls**. *J Anim Breed Genet* (2009.0) **126** 269-16. PMID: 19630877
2. Mathevon M, Buhr MM, Dekkers JC.. **Environmental, management, and genetic factors affecting semen production in Holstein bulls**. *J Dairy Sci* (1998.0) **81** 3321-3330. PMID: 9891279
3. Lessard C, Masseau I, Bilodeau JF. **Semen characteristics of genetically identical quadruplet bulls**. *Theriogenology* (2003.0) **59** 1865-1877. PMID: 12566158
4. Snoj T, Kobal S, Majdic G. **Effects of season, age, and breed on semen characteristics in different**. *Theriogenology* (2013.0) **79** 847-852. PMID: 23380262
5. Lonergan P, Fair S. **Influence of bull age, ejaculate number, and season of collection on semen production and sperm motility parameters in Holstein Friesian bulls in a commercial artificial insemination centre**. *J Anim Sci* (2018.0) **96** 2408-2418. PMID: 29767722
6. Guo F, Yang B, Ju ZH. **Alternative splicing, promoter methylation, and functional SNPs of sperm flagella 2 gene in testis and mature spermatozoa of Holstein bulls**. *Reproduction* (2014.0) **147** 241-252. PMID: 24277870
7. Wang X, Yang C, Guo F. **Integrated analysis of mRNAs and long noncoding RNAs in the semen from Holstein bulls with high and low sperm motility**. *Sci Rep* (2019.0) **9** 2092. PMID: 30765858
8. Egger G, Liang G, Aparicio A. **Epigenetics in human disease and prospects for epigenetic therapy**. *Nature* (2004.0) **429** 457-463. PMID: 15164071
9. Suzuki MM, Bird A. **DNA methylation landscapes: provocative insights from epigenomics**. *Nat Rev Genet* (2008.0) **9** 465-476. PMID: 18463664
10. Bell JT, Spector TD. **A twin approach to unraveling epigenetics**. *Trends Genet* (2011.0) **27** 116-125. PMID: 21257220
11. Tan Q, Christiansen L, von Bornemann Hjelmborg J. **Twin methodology in epigenetic studies**. *J Exp Biol* (2015.0) **218** 134-139. PMID: 25568460
12. Du Y, Li M, Chen J. **Promoter targeted bisulfite sequencing reveals DNA methylation profiles associated with low sperm motility in asthenozoospermia**. *Hum Reprod* (2016.0) **31** 24-33. PMID: 26628640
13. Boissonnas CC, Abdalaoui HE, Haelewyn V. **Specific epigenetic alterations of IGF2-H19 locus in spermatozoa from infertile men**. *Eur J Hum Genet* (2010.0) **18** 73-80. PMID: 19584898
14. Laqqan M, Tierling S, Alkhaled Y. **Alterations in sperm DNA methylation patterns of oligospermic males**. *Reprod Biol* (2017.0) **17** 396-400. PMID: 29108863
15. Capra E, Lazzari B, Turri F. **Epigenetic analysis of high and low motile sperm populations reveals methylation variation in satellite regions within the pericentromeric position and in genes functionally related to sperm DNA organization and maintenance in**. *BMC Genomics* (2019.0) **20** 940. PMID: 31810461
16. Verma A, Rajput S, De S. **Genome-wide profiling of sperm DNA methylation in relation to Buffalo (Bubalus bubalis) bull fertility**. *Theriogenology* (2014.0) **82** 750-9.e1. PMID: 25023295
17. Kropp J, Carrillo JA, Namous H. **Male fertility status is associated with DNA methylation signatures in sperm and transcriptomic profiles of bovine preimplantation embryos**. *BMC Genomics* (2017.0) **18** 280. PMID: 28381255
18. Shojaei Saadi HA, É F, Vigneault C. **Genome-wide analysis of sperm DNA methylation from monozygotic twin bulls**. *Reprod Fertil Dev* (2017.0) **29** 838-843. PMID: 26751019
19. Perrier J, Sellem E, Prézelin A. **A multi-scale analysis of bull sperm methylome revealed both species peculiarities and conserved tissue-specific features**. *BMC Genomics* (2018.0) **19** 404. PMID: 29843609
20. Zhou Y, Connor EE, Bickhart DM. **Comparative whole genome DNA methylation profiling of cattle sperm and somatic tissues reveals striking hypomethylated patterns in sperm**. *Gigascience* (2018.0) **7** giy039. PMID: 29635292
21. Ahlawat S, Sharma R, Arora R. **Promoter methylation and expression analysis of Bvh gene in bulls with varying semen motility parameters**. *Theriogenology* (2019.0) **125** 152-156. PMID: 30447494
22. Liu Y, Zhang Y, Yin J. **Distinct H3K9me3 and DNA methylation modifications during mouse spermatogenesis**. *J Biol Chem* (2019.0) **294** 18714-18725. PMID: 31662436
23. Liu S, Fang L, Zhou Y. **Analyses of inter-individual variations of sperm DNA methylation and their potential implications in cattle**. *BMC Genomics* (2019.0) **20** 888. PMID: 31752687
24. Lev Maor G, Yearim A, Ast G. **The alternative role of DNA methylation in splicing regulation**. *Trends Genet* (2015.0) **31** 274-280. PMID: 25837375
25. Linker SM, Urban L, Clark SJ. **Combined single-cell profiling of expression and DNA methylation reveals splicing regulation and heterogeneity**. *Genome Biol* (2019.0) **20** 30. PMID: 30744673
26. Yoshimi A, Lin KT, Wiseman DH. **Coordinated alterations in RNA splicing and epigenetic regulation drive leukaemogenesis**. *Nature* (2019.0) **574** 273-277. PMID: 31578525
27. Lu H, Sun F, Wang G. **Technical code of practice of bovine frozen semen production, NY/T 1234-2018**
28. Herthnek D, Englund S, Willemsen P. **Sensitive detection of Mycobacterium avium subspparatuberculosis in bovine semen by real-time PCR**. *J Appl Microbiol* (2006.0) **100** 1095-1102. PMID: 16630010
29. Gu H, Bock C, Mikkelsen TS. **Genome-scale DNA methylation mapping of clinical samples at single-nucleotide resolution**. *Nat Methods* (2010.0) **7** 133-136. PMID: 20062050
30. Gu H, Smith ZD, Bock C. **Preparation of reduced representation bisulfte sequencing libraries for genome-scale DNA methylation profling**. *Nat Protoc* (2011.0) **6** 468-481. PMID: 21412275
31. Langmead B. *Aligning short sequencing reads with Bowtie. Curr. Protoc. Bioinformatics* (2010.0)
32. Akalin A, Kormaksson M, Li S. **Methylkit: a comprehensive R package for the analysis of genome-wide DNA methylation profiles**. *Genome Biol* (2012.0) **13** R87. PMID: 23034086
33. Luo W, Brouwer C. **Pathview: an R/Biocondutor package for pathway-based data integration and visualization**. *Bioinformatics* (2013.0) **29** 1830-1831. PMID: 23740750
34. Luo W, Pant G, Bhavnasi YK. **Pathview Web: user friendly pathway visualization and data integration**. *Nucleic Acids Res* (2017.0) **45** W501-W508. PMID: 28482075
35. Dupont JM, Tost J, Jammes H. **De novo quantitative bisulfite sequencing using the pyrosequencing technology**. *Anal Biochem* (2004.0) **333** 119-127. PMID: 15351288
36. Livak KJ, Schmittgen TD. **Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method**. *Methods* (2001.0) **25** 402-408. PMID: 11846609
37. He Z, Jiang J, Kokkinaki M. **Nodal signaling via an autocrine pathway promotes proliferation of mouse spermatogonial stem/progenitor cells through Smad2/3 and Oct- 4activation**. *Stem Cells* (2009.0) **27** 2580-2590. PMID: 19688838
38. Kotaja N, Macho B, Sassone-Corsi P. **Microtubule-independent and protein kinase A-mediated function of kinesin KIF17b controls the intracellular transport of activator of CREM in testis (ACT)**. *J Biol Chem* (2005.0) **280** 31739-31745. PMID: 16002395
39. Kim Y, Fedoriw AM, Magnuson T. **An essential role for a mammalian SWI/SNF chromatin-remodeling complex during male meiosis**. *Development* (2012.0) **139** 1133-1140. PMID: 22318225
40. Kadoch C, Crabtree GR. **Mammalian SWI/SNF chromatin remodeling complexes and cancer: mechanistic insights gained from human genomics**. *Sci Adv* (2015.0) **1** e1500447. PMID: 26601204
41. Menon DU, Shibata Y, Mu W. **Mammalian SWI/SNF collaborates with a polycomb-associated protein to regulate male germline transcription in the mouse**. *Development* (2019.0) **146** dev174094. PMID: 31043422
42. Reik W, Walter J. **Genomic imprinting: parental influence on the genome**. *Nat Rev Genet* (2001.0) **2** 21-32. PMID: 11253064
43. Wei Y, Yang CR, Wei YP. **Paternally induced transgenerational inheritance of susceptibility to diabetes in mammals**. *Proc Natl Acad Sci USA* (2014.0) **111** 1873-1878. PMID: 24449870
44. Houshdaran S, Cortessis VK, Siegmund K. **Widespread epigenetic abnormalities suggest a broad DNA methylation erasure defect in abnormal human sperm**. *PLoS One* (2007.0) **2** e1289. PMID: 18074014
45. Montjean D, Zini A, Ravel C. **Sperm global DNA methylation level: association with semen parameters and genome integrity**. *Andrology* (2015.0) **3** 235-240. PMID: 25755112
46. Narud B, Khezri A, Zeremichael TT. **Sperm chromatin integrity and DNA methylation in Norwegian Red bulls of contrasting fertility**. *Mol Reprod Dev* (2021.0) **88** 187-200. PMID: 33634579
47. Bell JT, Spector TD. **DNA methylation studies using twins: what are they telling us?**. *Genome Biol* (2012.0) **13** 172. PMID: 23078798
48. Liu S, Chen S, Cai W. **Divergence analyses of sperm DNA methylomes between monozygotic twin AI bulls**. *Epigenomes* (2019.0) **3** 21. PMID: 34968253
49. González-Fernández L, Macías-García B, Loux SC. **Focal adhesion kinases and calcium/calmodulin-dependent protein kinases regulate protein tyrosine phosphorylation in stallion sperm**. *Biol Reprod* (2013.0) **88** 138. PMID: 23595906
50. Wong EW, Lee WM, Cheng CY. **Secreted Frizzled-related protein 1 (sFRP1) regulates spermatid adhesion in the testis via dephosphorylation of focal adhesion kinase and the nectin-3 adhesion protein complex**. *FASEB J* (2013.0) **27** 464-477. PMID: 23073828
51. Ni F, Hao S, Yang W. **Multiple signaling pathways in Sertoli cells: recent findings in spermatogenesis**. *Cell Death Dis* (2019.0) **10** 541. PMID: 31316051
52. Chung J, Shim S, Everley R. **Structurally distinct Ca(2+) signaling domains of sperm flagella orchestrate tyrosine phosphorylation and motility**. *Cell* (2014.0) **157** 808-822. PMID: 24813608
53. Wang J, Tang C, Wang Q. **NRF1 coordinates with DNA methylation to regulate spermatogenesis**. *FASEB J* (2017.0) **31** 4959-4970. PMID: 28754714
54. Varela I, Tarpey P, Raine K. **Exome sequencing identifies frequent mutation of the SWI/SNF complex gene PBRM1 in renal carcinoma**. *Nature* (2011.0) **469** 539-542. PMID: 21248752
55. Euskirchen G, Auerbach RK, Snyder M. **SWI/SNF chromatin-remodeling factors: multiscale analyses and diverse functions**. *J Biol Chem* (2012.0) **287** 30897-30905. PMID: 22952240
56. Maunakea AK, Chepelev I, Cui K. **Intragenic DNA methylation modulates alternative splicing by recruiting MeCP2 to promote exon recognition**. *Cell Res* (2013.0) **23** 1256-1269. PMID: 23938295
57. Shayevitch R, Askayo D, Keydar I. **The importance of DNA methylation of exons on alternative splicing**. *RNA* (2018.0) **24** 1351-1362. PMID: 30002084
58. Ho HC, Suarez SS. **Characterization of the intracellular calcium store at the base of the sperm flagellum that regulates hyperactivated motility**. *Biol Reprod* (2003.0) **68** 1590-1596. PMID: 12606347
|
---
title: 'The Associations of Platelet Activation and Coagulation Parameters with Obstructive
Sleep Apnoea: A Large-Scale Observational Study'
authors:
- Jundong Yang
- Wenjun Xue
- Zhicheng Wei
- Caiqiong Hou
- Huaming Zhu
- Huajun Xu
- Xiaolin Wu
- Yunhai Feng
- Xinyi Li
journal: International Journal of Clinical Practice
year: 2023
pmcid: PMC9988364
doi: 10.1155/2023/5817644
license: CC BY 4.0
---
# The Associations of Platelet Activation and Coagulation Parameters with Obstructive Sleep Apnoea: A Large-Scale Observational Study
## Abstract
### Objectives
Obstructive sleep apnoea (OSA) is associated with an increased risk of cardiovascular disease, with alterations in coagulability suspected as the mediating factor. This study explored blood coagulability and breathing-related parameters during sleep in patients with OSA.
### Design
Cross-sectional observational study. Setting. Shanghai Sixth People's Hospital. Participants. 903 patients diagnosed by standard polysomnography. Main Outcome and Measures. The relationships between coagulation markers and OSA were evaluated using Pearson's correlation, binary logistic regression, and restricted cubic spline (RCS) analyses.
### Results
The platelet distribution width (PDW) and activated partial thromboplastin time (APTT) decreased significantly with increasing OSA severity (both $p \leq 0.001$). PDW was positively associated with the apnoea-hypopnea index (AHI), oxygen desaturation index (ODI), and microarousal index (MAI) (ß = 0.136, $p \leq 0.001$; ß = 0.155, $p \leq 0.001$; and ß = 0.091, $$p \leq 0.008$$, respectively). APTT was negatively correlated with AHI (ß = −0.128, $p \leq 0.001$) and ODI (ß = −0.123, $$p \leq 0.001$$). PDW was negatively correlated with percentage of sleep time with oxygen saturation below $90\%$(CT90) (ß = −0.092, $$p \leq 0.009$$). The minimum arterial oxygen saturation (SaO2) correlated with PDW (ß = −0.098, $$p \leq 0.004$$), APTT (ß = 0.088, $$p \leq 0.013$$), and prothrombin time (PT) (ß = 0.106, $$p \leq 0.0003$$). ODI was risk factors for PDW abnormalities (odds ratio (OR) = 1.009, $$p \leq 0.009$$) after model adjustment. In the RCS, a nonlinear dose-effect relationship was demonstrated between OSA and the risk of PDW and APTT abnormalities.
### Conclusion
Our study revealed nonlinear relationships between PDW and APTT, and AHI and ODI, in OSA, with AHI and ODI increasing the risk of an abnormal PDW and thus also the cardiovascular risk. This trial is registered with ChiCTR1900025714.
## 1. Introduction
Obstructive sleep apnoea (OSA) is a chronic disease characterised by recurrent partial or complete upper airway collapse, resulting in a temporary pause of breathing during sleep and thus intermittent oxygen desaturation [1]. It is an independent risk factor for a hypercoagulable state and arterial thrombosis, both of which contribute to the development of cardiovascular disease (CVD) [2, 3].
Platelet activation and coagulation, blood viscosity, haematocrit, plasma fibrinogen, and other potential thrombosis markers are increased in patients with OSA [4]. A marker of thrombosis, and thus of platelet activation, is the platelet distribution width (PDW), which reflects the variance in the size of circulating platelets and is better standardized than the mean platelet volume (MPV) [5]. The prothrombin time (PT) and activated partial thromboplastin time (APTT) are used to evaluate intrinsic and extrinsic pathways of coagulation [6]. Changes in either pathway indicate fluctuations in blood coagulability [7]. An increased PDW and decreased APTT are characteristic of OSA [8], and the haematological parameters of patients with OSA improved significantly after treatment [9]. However, the small sample size and the use of nonstandard PSG in most studies have produced incomplete and inaccurate findings.
We explored the relationship between coagulation parameters and OSA in a large cohort. Our study was limited to male participants, as previous studies investigating sex-based differences in OSA [1], venous thromboembolism [10], and CVD [11] identified a higher prevalence of OSA in males than females, as well as an increased risk of coagulation disorders, and thus a higher risk of CVD and thromboembolism.
## 2.1. Subjects
Participants with suspected OSA seen between 2013 and 2018 were enrolled in the study. The inclusion criteria were whole-night PSG and a haematological examination; age ≥18 years; and male gender. The 1,352 initially enrolled patients were then screened according to the following exclusion criteria: previously diagnosed with OSA and treated with continuous positive airway pressure (CPAP), oral orthotics, upper airway surgery, etc.; systemic diseases, such as chronic liver disease, renal insufficiency, hyperthyroidism, hypothyroidism, or tumour; mental or neurological disorders; alcoholism; blood or platelet donation in the last 6 months; regular use of drugs affecting coagulation, such as aspirin, clopidogrel hydrogen sulphate tablets, and low-molecular heparin; and other sleep disorders, such as restless legs syndrome and narcolepsy. Ultimately, 903 male patients were included in this cross-sectional observational study.
The study was conducted in accordance with the Declaration of Helsinki. The Ethics Committee approved this study, and the trial was registered (ChiCTR1900025714) prior to commencement. Informed consent was obtained from all participants.
## 2.2. Polysomnographic Evaluation
Parameters related to breathing during sleep were assessed by overnight PSG (Alice 4 or 5; Respironics, Pittsburgh, PA, USA) and manually scored according to the guideline of the American Academy of Sleep Medicine 2012 [12]. Apnoea was defined as complete cessation or at least $90\%$ deduction of airflow lasting for ≥10 s, and hypopnea as a ≥ $30\%$ reduction accompanied by either a ≥ $3\%$ decrease in oxyhaemoglobin saturation or an associated arousal for ≥10 s. The apnoea-hypopnea index (AHI) was defined as the number of apnoea and hypopnea events per hour during the total sleep time. The oxygen desaturation index (ODI) was calculated as the total number of episodes of ≥$4\%$ oxyhaemoglobin desaturation during sleep. The microarousal index (MAI) was defined as the mean number of arousals per hour of sleep. The arterial oxygen saturation (SaO2, %) was monitored by pulse oximetry during sleep, then the percentage of sleep time with oxygen saturation below $90\%$ (CT90) was calculated according to the data of monitor time, and SaO2. OSA severity was classified based on the AHI, as follows: non-OSA, AHI < 5; mild OSA, 5 ≤ AHI < 15; moderate OSA, 15 ≤ AHI < 30; and severe OSA, AHI ≥ 30.
## 2.3. Anthropometric Measurements and Coagulation Tests
Height and weight were measured using a digital scale, with patients in a standing position and dressed in light clothing with bare feet. Neck circumference (NC) was measured at the middle of the cricothyroid membrane, waist circumference (WC) midway between the lowest rib and iliac crest, and hip circumference (HC) at the largest gluteal circumference. The mean values of the measurements were used in the analysis. BMI was calculated as weight in kilograms divided by height in meters squared (kg/m2). The weight-to-height ratio (WHR) was calculated as WC/HC. Smoking or drinking was defined as subjects who were self-reported to smoke or drink.
Venous blood samples were collected from patients in the morning following the overnight PSG. Ethylenediaminetetraacetic acid (EDTA) and sodium citrate were used as anticoagulants for platelet and coagulation tests, respectively. The platelet count (PLT), MPV and PDW, were measured using an XN-3000 analyser (SYSMEX, Hyogo, Japan) at optimal measurement time. The coagulation tests included the APTT, thrombin time (TT), and PT, all measured using a CS-5100 analyser (SYSMEX). An abnormal PDW was defined as <$9.8\%$ or >$16.2\%$, an abnormal APTT as <20 or >40 seconds, and an abnormal TT as <13 or >21 seconds according to the diagnostic criteria of the Guidelines on the Laboratory Aspects of Assays used in hemostasis and thrombosis [13].
## 2.4. Statistical Analysis
All statistical analyses were performed using SPSS (version. 21.0; IBM Corp., Armonk, NY, USA) and MATLAB 8.0 (MathWorks Corp., Natick, MA, USA) software. Data on the general characteristics of the participants are presented as the median and interquartile range (25–$75\%$). P-values for linear trends across different groups were calculated using the polynomial linear trend test for continuous variables. Bivariate correlation analyses were used to explore the relationships between sleep-breath parameters (AHI, ODI, MAI, CT90, and the minimum SaO2) and the coagulation indicators (PLT, PDW, MPV, APTT, PT, and TT), while binary logistic regression analyses were performed to evaluate the relationships of AHI, ODI, MAI, CT90, and minimum SaO2 with the risk of coagulation. Covariates including age, BMI, WHR, smoking, drinking, and hypertension were adjusted in different models: model 1 (age, BMI); model 2 (variables included in model 1 and hypertension); model 3 (age, WHR, and hypertension); model 4 (smoking and drinking based on model 2); and model 5 (variables in model 4 and additional WHR). Nonlinear relationships between OSA and coagulation indicators were evaluated in a restricted cubic spline (RCS) analysis. Knots for the AHI, ODI, and minimum SaO2 were identified in the RCS analysis using R software (R Development Core Team, Vienna, Austria). Two-sidedp-values <0.05 were considered to indicate statistical significance.
## 3.1. Baseline Characteristics
The final study group consisted of 101 non-OSA, 120 mild OSA, 147 moderate OSA, and 535 severe patients with OSA. The basic anthropometric and haematological characteristics are reported in Table 1. BMI, NC, WC, HC, WHR, smoking, and drinking differed significantly according to OSA severity, tending to show an increase with more severe OSA (p for linear trend <0.001). PDW increased, and APTT decreased significantly, with increasing OSA severity (both $p \leq 0.001$). The associations of MPV, TT, and PT and OSA severity were not significant ($p \leq 0.05$).
## 3.2. Associations between Breathing during Sleep and Coagulation Parameters
Pearson correlation analysis showed that AHI, ODI, and MAI were significantly positively associated with PDW (ß = 0.136, $p \leq 0.001$; ß = 0.155, $p \leq 0.001$; and ß = 0.091, $$p \leq 0.008$$, respectively; Table 2). The minimum SaO2 and CT90 were negatively associated with PDW (ß = −0.098, $$p \leq 0.004$$; and ß = −0.092, $$p \leq 0.009$$, respectively). APTT (ß = 0.088, $$p \leq 0.013$$) and PT (ß = 0.106, $$p \leq 0.003$$) were positively associated with minimum SaO2. APTT was negatively associated with AHI and ODI (ß = −0.128, $p \leq 0.001$; and ß = −0.123, $$p \leq 0.001$$, respectively). There were no associations between TT and OSA-related parameters (Table 2).
AHI, ODI, and minimum SaO2 were associated significantly with the risk of coagulation while ODI, MAI, and CT90 were not (Table 3). AHI and ODI increased the risk of an abnormal PDW after adjustments in model 1 (OR = 1.008, $95\%$ CI: 1.002–1.015, $$p \leq 0.015$$; OR = 1.010, $95\%$ CI: 1.003–1.016, $$p \leq 0.003$$, respectively), and the significance still existed after further adjusting for hypertension, smoking, and drinking (model 4 (OR = 1.008, $95\%$ CI: 1.001–1.015, $$p \leq 0.022$$; OR = 1.009, $95\%$ CI: 1.003–1.016, $$p \leq 0.005$$, respectively)). In model 5 adjusting for all potential confounders, ODI increased the risk of an abnormal PDW (odds ratio (OR) = 1.009, $95\%$ confidence interval ($95\%$ CI): 1.002–1.016, $$p \leq 0.009$$), and minimum SaO2 increased the risk of an abnormal TT (OR = 1.050, $95\%$ CI: 1.007–1.094, $$p \leq 0.021$$).
## 3.3. Nonlinear RCS Analysis
To assess the dose-effect relationship between OSA and the risk of hypercoagulability/hypocoagulability (abnormal coagulation), AHI, ODI, and the minimum SaO2 were analysed as continuous variables (Figure 1, x-axis) and the log odds of PDW, APTT, and TT as categorical variables (Figure 1, y-axis). The risk of abnormal coagulation did not always increase with increasing OSA severity. The relationships of AHI with abnormal PDW (Figure 1(a)) and APTT (Figure 1(b)) were not linear. In the RCS, the knots for AHI and log odds of an abnormal PDW for patients with OSA with an AHI ≤84.26 were 2.235, 17.4, 39.1, and 84.26, respectively, and reached a plateau. Knots for AHI of 2.4, 18.165, and 40.5 indicated a decreased risk of an abnormal APTT, as reflected in the slopes of the curves (Figure 1(a)). However, the risk of an abnormal APTT increased in patients with an AHI >60.57, and a plateau was reached at an AHI of 84.5 (Figure 1(b)). The same trend was observed for AHI and ODI; for patients at the risk of an abnormal PDW, the ODI increased, followed by a plateau (Figure 1(d)). For patients at the risk of an abnormal APTT, the ODI declined, followed by a rise and then a plateau (Figure 1(e)). The relationship between the risk of an abnormal TT and OSA was close to linear (Figures 1(c), 1(f), and 1(i)).
## 4. Discussion
This study explored the associations between breathing-related variables and haematological parameters related to coagulation. The results showed that PDW and APTT were associated with OSA-related traits, especially the AHI and ODI, which are risk factors for an abnormal PDW. Both nonlinear and nonmonotonic relationships between OSA and coagulation were determined.
The size distribution of platelets in the peripheral circulation is expressed as the PDW, which is considered a marker of thromboembolic diseases [14]. Fan et al. [ 15] reported a significantly higher PDW in patients with severe OSA. A meta-analysis also demonstrated that OSA was associated with a high PDW [8]. The significant correlations of the PDW and AHI with the minimum SaO2 in that study were consistent with our results [16]. Children with OSA also have a higher platelet count and higher PDW [17]. In another study, the PDW improved significantly in patients with OSA after CPAP treatment [9]. Our results also indicated a correlation of the PDW with ODI, CT90, and MAI, with ODI independently increasing the risk of an abnormal PDW. This observation suggested that PDW is affected by hypoxia and thus serves as a meaningful blood marker in patients with OSA. Elevated surface marker expression in significantly hypoxemic OSA individuals is consistent with the platelet activation seen in this group.
In addition to platelet activation, thrombosis is modulated by the coagulation system and thus related to CVD [18]. PT, TT, and APTT are indicators of extrinsic, common, and intrinsic coagulation pathways, respectively [6, 7]. A previous study showed that APTT correlated with ODI in children with OSA, which suggested that OSA-related traits enhance coagulability in paediatric OSA [19]. PT was also shown to decrease with increased OSA severity and correlated with AHI [20]. Although in our study, there was no significant difference in PT as a function of OSA severity, we did find that PT was positively associated with the minimum SaO2. Previous findings of an improvement in PT, TT, and APTT in patients with OSA after CPAP treatment [9] and upper airway surgery thus indicate that improved ventilation and the correction of hypoxia improve coagulation parameters in OSA.
In our study, several potential confounders that might affect coagulation were adjusted, and a significant association of hypoxia-related parameters with the risk of coagulation still existed, revealing the associations between OSA and platelet activation and coagulation system. Chronic intermittent hypoxia is the hallmark of OSA, but the underlying mechanism of platelet activation and disturbed coagulation in response to hypoxic conditions remains to be determined. Krieger et al. [ 21] reported that intermittent hypoxemia in OSA suppressed platelet responses to epinephrine and thrombin and decreased the level of the surface marker CD40L, thus signifying increased platelet activation, and it was also demonstrated that platelet activation in hypoxic mice can contribute to hypoxia-induced inflammation [22]. CAPNS1-dependent calpain, activated during the platelet activation cascade, is associated with hypoxia-induced thrombogenesis [23]. Hypoxia acts on the platelet purinergic signaling pathway by increasing P2Y1-receptor and ADP pathway activities, which is of potential therapeutic interest [24]. OSA was shown to elevate circulating levels of platelet-derived microparticles released by platelet activation, thus increasing the risk of CVD [25]. Moreover, such platelet-derived microparticles with CD41+ and annexin V+ variated diurnally, increasing from morning and reaching their peak levels in the afternoon [26], suggesting a more complicated link between platelet activation and the risk of CVD in patients with OSA. Few studies have focused on the intrinsic coagulation pathway in OSA and hypoxia. Consequently, the molecular mechanisms underlying platelet activation and coagulation in OSA and its complications are not fully understood, and circadian rhythm could be considered as a potential aspect in future studies.
Our study also revealed nonlinear relationships of OSA-related traits with PDW and APTT, highlighting the complex role of the coagulation system in OSA. The reciprocal interactions of platelets and the coagulation system, and of plasma coagulants and blood cells, are crucial aspects of thrombus formation and are also keys to elucidating the pathophysiology of thrombosis in many diseases, including OSA [27]. Since adherence to CPAP therapy is limited, further studies of the clinical utility of coagulation indices for monitoring patients with OSA are needed.
Our study had several limitations. Firstly, due to its cross-sectional design, causal relationships could not be determined. Secondly, some haematology markers, such as blood viscosity and clotting factors, were not evaluated. Thirdly, the roles played by environmental factors, such as diet, exercise frequency, and economic conditions, were not considered. Further research is therefore needed to understand the effects of intermittent hypoxia on blood viscosity, fibrinolysis system, and other factors contributing to the hypercoagulation state of patients.
## 5. Conclusion
This study demonstrated associations of PDW and APTT with OSA-related traits and showed that AHI and ODI are risk factors for an abnormal PDW. The nonlinear relationship between platelet activation and coagulation parameters with OSA provided evidence of the complex interactions among platelets, coagulation, and hypoxia, which may contribute to the cardiovascular complications seen in patients with OSA.
## Data Availability
The original clinical data used to support the findings of this study are available from the corresponding author upon request.
## Additional Points
(i) PDW increased, and APTT decreased with the increase in OSA severity. ( ii) PDW and APTT correlate with OSA-related traits. ( iii) OSA increased the risk of an abnormal PDW. ( iv) Nonlinear relationships of OSA-related traits with PDW and APTT were discovered. ( v) The abnormal alteration of platelets and the coagulation system in OSA might contribute to the cardiovascular complications.
## Ethical Approval
The ethics committee of Shanghai Jiao Tong University Affiliated Sixth People's Hospital approved this study according to Helsinki Declaration II, and the trial was registered (ChiCTR1900025714) prior to commencement. All the participants have given the informed consent before taking part in the study.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## Authors' Contributions
JD Y and XY L designed the work; WJ X, ZC W, CQ H, HM Z, HJ Z acquired and analysed data; JD Y, XY L, YH F, XL W drafted, revised, and approved the manuscript; XL W and XY L agree to be accountable for all aspects of the work. Jundong Yang and Wenjun Xue are authors who contributed equally to this paper.
## References
1. Jordan A. S., McSharry D. G., Malhotra A.. **Adult obstructive sleep apnoea**. (2014) **383** 736-747. DOI: 10.1016/s0140-6736(13)60734-5
2. Toraldo D. M., Peverini F., De Benedetto M., De Nuccio F.. **Obstructive sleep apnea syndrome: blood viscosity, blood coagulation abnormalities, and early atherosclerosis**. (2013) **191** 1-7. DOI: 10.1007/s00408-012-9427-3
3. Bikov A., Meszaros M., Schwarz E. I.. **Coagulation and fibrinolysis in obstructive sleep apnoea**. (2021) **22** p. 2834. DOI: 10.3390/ijms22062834
4. Vrints H., Shivalkar B., Heuten H.. **Cardiovascular mechanisms and consequences of obstructive sleep apnoea**. (2013) **68** 169-178. DOI: 10.2143/acb.2981
5. Gabryelska A., Łukasik Z. M., Makowska J. S., Białasiewicz P.. **Obstructive sleep apnea: from intermittent hypoxia to cardiovascular complications via blood platelets**. (2018) **9** p. 635. DOI: 10.3389/fneur.2018.00635
6. Dahlbäck B.. **Blood coagulation**. (2000) **355** 1627-1632. DOI: 10.1016/s0140-6736(00)02225-x
7. Winter W. E., Flax S. D., Harris N. S.. **Coagulation testing in the Core laboratory**. (2017) **48** 295-313. DOI: 10.1093/labmed/lmx050
8. Wu M., Zhou L., Zhu D., Lai T., Chen Z., Shen H.. **Hematological indices as simple, inexpensive and practical severity markers of obstructive sleep apnea syndrome: a meta-analysis**. (2018) **10** 6509-6521. DOI: 10.21037/jtd.2018.10.105
9. Sökücü S. N., Ozdemir C., Dalar L., Karasulu L., Aydın S., Altın S.. **Complete blood count alterations after six months of continuous positive airway pressure treatment in patients with severe obstructive sleep apnea**. (2014) **10** 873-878. DOI: 10.5664/jcsm.3958
10. Roach R. E., Cannegieter S. C., Lijfering W. M.. **Differential risks in men and women for first and recurrent venous thrombosis: the role of genes and environment**. (2014) **12** 1593-1600. DOI: 10.1111/jth.12678
11. Regitz-Zagrosek V., Kararigas G.. **Mechanistic pathways of sex differences in cardiovascular disease**. (2017) **97** 1-37. DOI: 10.1152/physrev.00021.2015
12. Berry R. B., Budhiraja R., Gottlieb D. J.. **Rules for scoring respiratory events in sleep: update of the 2007 AASM manual for the scoring of sleep and associated events. Deliberations of the sleep apnea definitions task force of the American Academy of sleep medicine**. (2012) **8** 597-619. DOI: 10.5664/jcsm.2172
13. Mackie I., Cooper P., Lawrie A.. **Guidelines on the laboratory aspects of assays used in haemostasis and thrombosis**. (2013) **35** 1-13. DOI: 10.1111/ijlh.12004
14. Vagdatli E., Gounari E., Lazaridou E., Katsibourlia E., Tsikopoulou F., Labrianou I.. **Platelet distribution width: a simple, practical and specific marker of activation of coagulation**. (2010) **14** 28-32. PMID: 20411056
15. Fan Z., Lu X., Long H., Li T., Zhang Y.. **The association of hemocyte profile and obstructive sleep apnea**. (2019) **33**. DOI: 10.1002/jcla.22680
16. Song Y. J., Kwon J. H., Kim J. Y., Kim B. Y., Cho K. I.. **The platelet-to-lymphocyte ratio reflects the severity of obstructive sleep apnea syndrome and concurrent hypertension**. (2015) **22** p. 1. DOI: 10.1186/s40885-015-0036-3
17. Barcelo A., Morell-Garcia D., Sanchis P.. **Prothrombotic state in children with obstructive sleep apnea**. (2019) **53** 101-105. DOI: 10.1016/j.sleep.2018.09.022
18. Herzberg M. C.. **Coagulation and thrombosis in cardiovascular disease: plausible contributions of infectious agents**. (2001) **6** 16-19. DOI: 10.1902/annals.2001.6.1.16
19. Shen T., Wang J., Yang W.. **Hematological parameters characteristics in children with obstructive sleep apnea with obesity**. (2021) **14** 1015-1023. DOI: 10.2147/rmhp.s297341
20. Hong S. N., Yun H. C., Yoo J. H., Lee S. H.. **Association between hypercoagulability and severe obstructive sleep apnea**. (2017) **143** 996-1002. DOI: 10.1001/jamaoto.2017.1367
21. Krieger A. C., Anand R., Hernandez-Rosa E.. **Increased platelet activation in sleep apnea subjects with intermittent hypoxemia**. (2020) **24** 1537-1547. DOI: 10.1007/s11325-020-02021-4
22. Delaney C., Davizon-Castillo P., Allawzi A.. **Platelet activation contributes to hypoxia-induced inflammation**. (2021) **320** L413-l421. DOI: 10.1152/ajplung.00519.2020
23. Tyagi T., Ahmad S., Gupta N.. **Altered expression of platelet proteins and calpain activity mediate hypoxia-induced prothrombotic phenotype**. (2014) **123** 1250-1260. DOI: 10.1182/blood-2013-05-501924
24. Paterson G. G., Young J. M., Willson J. A.. **Hypoxia modulates platelet purinergic signalling pathways**. (2020) **120** 253-261. DOI: 10.1055/s-0039-3400305
25. Maruyama K., Morishita E., Sekiya A.. **Plasma levels of platelet-derived microparticles in patients with obstructive sleep apnea syndrome**. (2012) **19** 98-104. DOI: 10.5551/jat.8565
26. Bikov A., Kunos L., Pallinger E.. **Diurnal variation of circulating microvesicles is associated with the severity of obstructive sleep apnoea**. (2017) **21** 595-600. DOI: 10.1007/s11325-017-1464-y
27. Sang Y., Roest M., de Laat B., de Groot P. G., Huskens D.. **Interplay between platelets and coagulation**. (2021) **46**. DOI: 10.1016/j.blre.2020.100733
|
---
title: Mass Cytometry Reveals the Imbalanced Immune State in the Peripheral Blood
of Patients with Essential Hypertension
authors:
- Rui Yang
- Yuhong He
- Honggang Zhang
- Qiuju Zhang
- Bingwei Li
- Changming Xiong
- Yubao Zou
- Bingyang Liu
journal: Cardiovascular Therapeutics
year: 2023
pmcid: PMC9988372
doi: 10.1155/2023/9915178
license: CC BY 4.0
---
# Mass Cytometry Reveals the Imbalanced Immune State in the Peripheral Blood of Patients with Essential Hypertension
## Abstract
Mounting evidence has confirmed that essential hypertension (EH) is closely related to low-grade inflammation, but there is still a lack of in-depth understanding of the state of immune cells in the circulating blood of patients with EH. We analyzed whether hypertensive peripheral blood immune cell balance was destroyed. The peripheral blood mononuclear cells (PBMCs) of all subjects were analyzed using time-of-flight cytometry (CyTOF) based on 42 kinds of metal-binding antibodies. CD45+ cells were categorized into 32 kinds of subsets. Compared with the health control (HC) group, the percentage of total dendritic cells, two kinds of myeloid dendritic cell subsets, one intermediate/nonclassical monocyte subset and one CD4+ central memory T cell subset in the EH group, was significantly higher; the percentage of low-density neutrophils, four kinds of classical monocyte subsets, one CD14lowCD16− monocyte subset, one naive CD4+ and one naive CD8+ T cell subsets, one CD4+ effector and one CD4+ central memory T cell subsets, one CD8+ effector memory T cell subset, and one terminally differentiated γδ T cell subset, decreased significantly in EH. What is more, the expression of many important antigens was enhanced in CD45+ immune cells, granulocytes, and B cells in patients with EH. In conclusion, the altered number and antigen expression of immune cells reflect the imbalanced immune state of the peripheral blood in patients with EH.
## 1. Introduction
According to statistics, the number of people aged 30-79 years with hypertension doubled to 1.28 billion worldwide in the 30 years from 1990 to 2019, while the treatment rate was less than half [1]. We know that essential hypertension (EH) of unknown etiology accounts for the majority of hypertension, even up to $95\%$ [2]. Hypertension and cardiovascular diseases (CVD) are actually low-grade inflammatory diseases [3]. In hypertension, the self-perpetuating cycle of inflammation and oxidative stress is an important cause of vascular pathology and renal damage [4]. Hypertension induces dendritic cells (DCs) to produce reactive oxygen species (ROS) [5], which in turn stimulates DCs to produce IL-1β, IL-6, and IL-23, thereby promoting the polarization of T cells and producing IL-17A [6, 7]. An investigation showed that increased endothelial mechanical stretch promoted the conversion of monocytes to intermediate monocytes and stimulated monocytes to express IL-6, IL-1β, IL-23, and TNF-α [8]. Curiously, researchers found that resting neutrophils from patients with EH had increased CD11b and CD18 fluorescence intensity compared to healthy individuals, but the opposite results appeared in neutrophils activated with formyl-methionyl-leucyl-phenylalanine (fMLP) [9]. T cells have been reported to be reduced in the blood of hypertensive patients [10], and animal experiments have shown increased B cells in the spleen and kidney of hypertensive mice and IgG accumulation in the aortic membrane [11, 12]. Inflammation and abnormal activation of the immune system are now thought to be the important mechanisms in the progression of hypertension, which could be used to develop more effective therapies to alleviate end-organ damage in patients with hypertension [13].
What is the immune state of peripheral blood in EH patients? *There is* not a fully description yet. In this study, time-of-flight cytometry (CyTOF) was used to analyze various types of peripheral blood mononuclear cells (PBMCs) to more comprehensively assess the peripheral blood immune cell state in patients with EH.
## 2.1. Grouping
There were 5 patients with EH and 5 healthy controls (HCs). All subjects were male, and recruited from Fuwai Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College (CAMS & PUMC; Beijing, China). The inclusion criteria for EH were in accordance with the 2018 Chinese guidelines for the management of hypertension, i.e., in the absence of antihypertensive drugs, the mean of three blood pressures measured on the nonsame day, systolic blood pressure (SBP) ≥ 140 mmHg and (or) diastolic blood pressure (DBP) ≥ 90 mmHg, was taken as the basis for hypertension. Diabetes, coronary disease, abnormal liver/renal function, classic chronic inflammatory diseases, secondary hypertension, and other serious diseases were excluded. The study protocol was approved by the Ethics Committee at the Institute of Microcirculation at the CAMS & PUMC and adhered to the tenets of the Declaration of Helsinki as well as applicable Chinese laws. Subjects were measured for blood pressure and risk factors associated with CVD; the basic information of the two groups is shown in Table 1.
## 2.2. Collection of PBMCs and Cell Staining
Detailed steps are shown in supplementary materials and methods, and 42 staining antibodies are shown in Table S1.
## 2.3. CyTOF Detection
Start the mass cytometry (Helios Mass Cytometry, FLUIDIGM), debug instruments and control quality standards (Tuning Solution, FLUIDIGM), the same number of cells in each sample were mixed together, add calibration beads (EQ™ Four Element Calibration Beads, FLUIDIGM), set channel name, data name, collection speed, and total cell number, and collect the raw data.
## 2.4. Data Analysis
The data was processed using the bead normalization method. Sample-specific barcodes were used to distinguish data for each individual in multiple sample data. Gates were used to distinguish live cells and immune cells and exclude debris, cell clumps, and beads (FlowJo v10.0.7). The x-shift clustering algorithm was used to process the date of markers in the panel. Then, the data was downsampled. Data-driven clustering analysis was used to obtain different cell clusters. 300,000 cells in each group were analyzed.
The specified cells were selected, and the dimensionality of the data was reduced by the t-distributed stochastic neighbor embedding (T-SNE) data visualization method. The viSNE map and heat map were generated according to the T-SNE coordinates, and the cell cluster numbers above, the percentages of cell clusters, and the expression intensities of metal-connected antibodies were shown. The sample information was grouped, box plots were made using the cluster-percentage matrix, and the average expression levels of the markers were displayed in the violin plots. The normality of the data was assessed using the Shapiro-Wilk test firstly, then unpaired two-tailed t-test, or Welch's t-test, or Mann–Whitney test was used to compare the difference between the two groups (α = 0.05). Pearson or Spearman correlation analysis was used to calculate the correlation between the index with significant difference between the two groups and blood pressure (GraphPad Prism 9.4.0).
## 3.1. Basic Information of Subjects
The basic situation of the HC and EH groups is shown in Table 1. In our study, SBP, DBP, and risk factors for CVD such as body mass index (BMI), total cholesterol (TC), and low-density lipoprotein (LDL) values were significantly increased in EH compared to HC. Therefore, we also analyzed the direct correlation between some important indicators and blood pressure (the main diagnostic standard of hypertension). In addition, the influence of higher age in the EH group was analyzed and discussed later.
## 3.2. Analysis of PBMCs Using CyTOF
The experimental flow chart is shown in Figure 1(a). The merge sample data is processed by the T-SNE data visualization method, and 32 cell clusters are automatically formed in heat map and viSNE graph; all clusters are classified into different cell types (Figures 1(b)–1(d) and Table 2). The expression of a part of the cell lineage markers is shown in Figure 1(e), and they are clearly distributed among cells of different lineages.
## 3.3. Comparison of Cell Percentage and Marker Expression in CD45+ Cells between the HC and EH Groups
The quantity of different immune cell subsets was compared between the HC and EH groups (Figure 2(a)). Compared with the HC group, the percentage of DCs was significantly increased in EH ($P \leq 0.05$), and the percentage of monocytes was increased but not significantly different ($$P \leq 0.25$$). Interestingly, granulocytes in the PBMC layer of EH were decreased ($P \leq 0.01$). The results showed no change in the percentage of natural killer cells (NKs) between the HC and EH groups ($$P \leq 0.64$$). In our study, there were no differences in the total number of CD4+ T cells ($$P \leq 0.42$$), CD8+ T cells ($$P \leq 0.93$$), and γδ T cells ($$P \leq 0.22$$) between the HC and EH groups. The percentage of B cells also showed no significant difference between the two groups ($$P \leq 0.50$$).
The average expression level of all markers in CD45+ cells of the HC and EH groups is shown in Figure S1, and CD45, CD95, CD45RO, CCR5, Ki67, TLR2, Foxp3, CD38, CD69, PD-1 were increased in EH, while CD11b was decreased. The functions of CD45, CD45RO, and CD11b are described in Table S2.
## 3.4. Characteristics of DCs in EH
DCs are generally divided into myeloid DCs (mDC, expressing CD11c and CD33) and plasmacytoid DCs (pDC, expressing CD123) [14]. In this study, DCs were subdivided into three cell subsets by clustering analysis (Figure 1(b)). Cluster 4 cells were defined as pDCs because of their high expression of CD123 [15], and compared with the HC group, there was no difference in the number of them in the EH group (Figure 2(b)). CX3CR1 (CX3C chemokine receptor 1) is the marker of mDCs [16], so CX3CR1+CD38−CD11c+HLA-DR+ DCs (cluster 5) were defined as mDCs here; we also defined Ki67+CD11c+HLA-DR+CD123− DCs (cluster 6) as proliferative mDCs [14, 15, 17], it is also characterized as CCR5+, which is usually expressed on immature DCs [18]. As the results showed, the two kinds of mDCs were both increased in EH ($P \leq 0.05$, $P \leq 0.05$, respectively; Figure 2(b)).
## 3.5. Characteristics of Monocytes in EH
Monocytes in peripheral blood are usually divided into three subsets: CD14++CD16− (classical), CD14++CD16+ (intermediate), and CD14+CD16++ (nonclassical), which are present in 80-$95\%$, 2-$8\%$, and 2-$11\%$ of circulating monocytes, respectively [19–21]. Intermediate monocytes and nonclassical monocytes expand under inflammatory condition, and the phosphorylation level of STAT3 was higher in intermediate monocytes than in the other two types of monocytes [8]. Typically, CX3CR1 is highly expressed in intermediate and nonclassical monocytes, and HLA-DR is highly expressed on intermediate monocytes [20].
In this research, the only cluster of monocytes (cluster 7; Figure 1(b)) was further divided into 13 cell subsets (Figures 3(a)–3(c)). The results showed an increased percentage of CD45hiCXCR3+CX3CR1+HLA-DRhiCD11blowCD14+CD16+ monocytes (cluster 1), which expresses lower CD14 and higher CD16 like intermediate/nonclassical monocytes, decreased percentage of CD14lowCD16− monocytes (cluster 3), and decreased percentage of CD45lowCXCR3−CD14+CD16-/low monocytes (clusters 8, 9, 10, and 12) similar to classical monocytes in EH ($P \leq 0.05$, $P \leq 0.01$, $P \leq 0.05$, $P \leq 0.01$, $P \leq 0.01$, and $P \leq 0.001$, respectively; Figure 3(d)).
Moreover, two markers, CD14 and CD16, were selected for the pseudotime analysis of monocytes in 13 clusters (Figures S2a and S2b). The results showed that the part of monocytes from patients with EH differentiated toward state 6 and state 7 (Figure 3(e)). And the pseudotime plots showed that the two states were CD14hi, and the expression of CD16 gradually increased from state 7 to state 6.
## 3.6. Characteristics of Granulocytes in EH
Neutrophils account for $70\%$ of the peripheral blood circulation, which are first recruited to the site of inflammation [22]. In this study, interestingly, the results showed that PBMCs contained a small number of cells characterized as CD11c+CD66b+CD31+CD16+CD11b+ (cluster 3) (Figure 1(b)), and the number of these cells was significantly reduced in EH ($P \leq 0.01$; Figure 2(b)). Antigens CD66b, CD16, and CD11b were reported as markers of activated neutrophils [23]. Usually, granulocytes do not appear in the PBMCs but in the upper layer of the red blood cells (normal-density neutrophils, NDNs) when separating by Ficoll density gradient centrifugation method. Neutrophils in the PBMC layer, known as low-density neutrophils (LDNs), have been identified in systemic lupus erythematosus, rheumatoid arthritis, sepsis, asthma, natural pregnancy, tumors [22], COVID-19 [24], and arterial hypertension [25], but the function of LDNs in EH is not yet well defined.
In addition, the average expression of antigens in the only granulocyte cluster of the two groups was compared, and most of the proteins in the EH group showed an increasing trend except CD11b (Figure 4(a)). The expression level of CD45, Ki67, Foxp3, CD24, and CD31 was significantly increased in the EH group, and their functions are shown in Table S2.
## 3.7. Characteristics of NKs in EH
By clustering analysis, four NK cell clusters were obtained (Figure 1(b)). Compared to HC, there was no significant change in the number of NKs in EH (Figures 2(a) and 2(b)).
## 3.8. Characteristics of T Cells in EH
CD4+ T cells, CD8+ T cells, and γδ T cells are closely associated with the development of hypertension [26]. Usually, CD45RO is used as a marker of activated/memory T cells [27], CD45RO−/CD45RA+ is used to distinguish naive T cells [28, 29]. Memory T cells can be further divided into three types based on the lymph node homing receptors CD62L and CCR7 [30, 31]. Central memory T cells (TCM, CD62LhighCCR7+) are mainly stored in secondary lymphoid organs and have a high-proliferative capacity when reactivated [32]. Effector memory T cells (TEM, CD62LlowCCR7−) exist in the periphery and can be rapidly recruited to inflammatory sites and produce effector factors [32, 33]. And resident memory T cells (TRM, CD62LlowCCR7−CD69+) remain at the site of infection in order to provide more rapid and direct protection [32].
## 3.8.1. CD4+ T Cells
In this study, nine CD4+ T cell clusters were identified (Figure 1(b)), and four of them showed significant difference in quantity between the HC and EH groups (Figure 2(b)). Cluster 17 was CD45RO−CCR7+CD95− naive CD4+ T cell subset [28], and the percentage of it was significantly reduced in EH ($P \leq 0.01$). Cluster 21 cells characterized by CD45RO+CCR7−CD27−CD28+CD4+ were defined as effector memory T cells, which were decreased in EH ($P \leq 0.01$). Furthermore, two kinds of central memory CD4+ T cells, CD45lowCD45RO+CCR7+CD27+CD28+CD4+ T cells (cluster 22) and CD45hiCD45RO+CCR7+CD27+CD28+CD4+ T cells (cluster 24), expressed similar antigens. However, the results showed that the number of cluster 22 cells with CD45low decreased in EH, while the number of cluster 24 cells with CD45hi increased ($P \leq 0.05$, $P \leq 0.001$, respectively).
## 3.8.2. CD8+ T Cells
CD8+ T cells were clustered into eight clusters here (Figure 1(b)); the percentage of clusters 25 and 30 was significantly different between the two groups, while other clusters were not (Figure 2(b)). The results showed that CD45RO−CCR7+CD95− naive CD8+ T cells (cluster 25) and CD45lowCD45ROlowCD161+CCR7−CCR6lowCCR5+CD8+ effector memory T cells (cluster 30) [29], were both decreased in EH ($P \leq 0.01$, $P \leq 0.01$, respectively).
## 3.8.3. γδ T Cells
As the results showed, γδ T cells were divided into four cell subsets (Figure 1(b)). Cluster 12 cells characterized by CD45lowCD27−CD45ROlowCD57hiGranzyme B+ were defined as terminally differentiated γδ T cells, and their proportion was reduced in EH ($P \leq 0.05$, Figure 2(b)).
## 3.9. Characteristics of B Cells in EH
In this study, B cells were divided into two subsets (Figure 1(b)), and the number of B cells was not altered in EH. However, compared to HCs, the expression of most markers in B cells from EH patients was significantly increased, including CD45, IgM, CD19, CD45RO, Ki67, CXCR3, CCR7, CD24, CD38, CD31, CXCR5, and HLA-DR (Figure 4(b) and Table S2).
## 3.10. The Correlation between Blood Pressure/Age and Immune Cells
After obtaining the indicators with significant differences between the two groups, we then analyzed their correlation with blood pressure (SBP, DBP, and pulse pressure difference (PP)) and other influencing factor (age), as shown in Table S3.
## 4.1. The Increased Total DCs and mDCs in EH
*In* general, mDCs mainly produce the cytokines IL-1β, IL-6, IL-12, IL-23, and present soluble antigens; pDCs mainly produce type I and type III interferons to combat the virus [14]. CX3CL1, the ligand of CX3CR1, is expressed by smooth muscle cells, activated endothelial cells, epithelial cells, and neurons [34]. CX3CL1/CX3CR1 axis has both chemotaxis and adhesion capacity [35]. It was reported that CX3CR1 on DCs is a kidney-specific “homing receptor” [36]. CD38 is a nicotinamide dinucleotide (NAD+) catabolic enzyme that catabolizes NAD+ into adenosine diphosphate ribose (ADPR) and cyclic ADPR, which leads to calcium mobilization [37, 38]. It was shown that the migration of Cd38−/−DCs to lymph nodes was blocked and that the inhibition of CD38 attenuated the chemotaxis of DC cells towards CCL21 (a ligand of CCR7) [39, 40]. Therefore, myeloid CX3CR1+CD38−CD11c+HLA-DR+ DCs may have the ability to be recruited to the kidney rather than the lymph nodes. Under the physiological condition, immature DCs are mainly responsible for capturing autoantigens to induce immune tolerance; when inflammation occurs, they mature and activate T cells into effector cells [41]. Immature DCs reach the site of inflammation via CCR5, recognize the antigen, and become mature, subsequently triggering a T cell response [42]. The increased Ki67+CD11c+HLA-DR+CD123− mDCs may represent an overactivated immune response in hypertension.
In addition, we found a small number of Foxp3+ DCs in peripheral blood, which may have immunosuppressive effects [43–45], but currently, there are few relevant studies, which deserve more attention.
## 4.2. The Imbalanced State of Monocytes in EH
CD14+CD16+ monocytes have a strong capacity to present antigen due to HLA-DR and promote TNF secretion [46]. As mentioned above, CD45hiCXCR3+CX3CR1+HLA-DRhiCD11blowCD14lowCD16+ monocytes, which are similar to intermediate/nonclassical monocytes, increased in EH, are proinflammatory, and the expression of chemokines may make them easier to migrate to tissues [34, 47, 48]. On the contrary, the four subsets of classical monocytes in the EH group almost disappeared. In conclusion, the changes of monocytes may all indicate enhanced immune activity in EH.
## 4.3. Abnormal State of LDNs in EH
Neutrophils exert their antimicrobial action through phagocytosis, degranulation, and extracellular traps [22]. When inflammation occurs, endothelial cells are activated to express adhesion molecules that promote neutrophil adhesion [22]. Curiously, in our study, LDNs from hypertensive patients were reduced but with higher expressed antigens (except CD11b) compared to controls. Coincidentally, a recent study showed that patients with hypertension have a higher proportion of NDNs compared to healthy individuals [25]. Is there a correlation between LDNs and NDNs? And how? We are not sure whether LDNs are developing immature neutrophils or degranulated neutrophils, and their pathophysiological mechanism in hypertension deserves more exploration.
## 4.4. Varied T Cell Subsets in EH
Total CD4+ T cells and total CD8+ T cells have been reported to be reduced in hypertension [10]. And in our study, the results showed no significant change in their respective total number, but the subsets of CD4+, CD8+, and γδ T cells were altered in EH. We found that CD4 and CD8 were not only expressed on CD4+ T cells and CD8+ T cells, but also on γδ T cells, NKs and DCs. So, different classification standards may be the reasons for the different results.
When inflammation occurs, naive T cells are activated by antigen presented by antigen-presenting cells (APCs), beginning to proliferate and produce effector T cells. When inflammation disappears, small number of effector cells survive and become antigen-specific memory cells [30, 49]. Normally, naive T cells in the blood enter secondary lymphoid organs for patrol and can proliferate rapidly when activated [49]. Our results showed that the percentage of CD45RO−CCR7+CD95− naive CD4+ T cell subset and CD45RO−CCR7+CD95− naive CD8+ T cell subset is reduced in the EH group. DCs can activate naive T cells into effector cells [50]; in our results, increased DCs may also explain the reduced number of naive T cells. It has been reported that naive T cells decrease and memory T cells increase with age [51, 52], but the difference between 22 and 40 years old was not emphasized in those studies. And in this study, the results showed that the decrease of naive CD4+ and naive CD8+ T cells was also associated with increasing age.
In our study, CD4+ and CD8+ memory T cells were divided into multiple subsets and had different changes. CD27 and CCR7 can mediate immune cells homing to lymph nodes; CD4+ and CD8+ T cells expressing CCR7 usually also express CD27 and CD28 (a costimulatory molecule for T cell activation), but not vice versa [29, 53]. And CCR7−CD45RA−CD27−CD28+CD4+ T cells are effector memory T cells that secrete IFN-γ, IL-4, and IL-2 [54]. Meanwhile, the investigators defined CCR7+CD45RA−CD27+CD28+CD4+ T cells as central memory T cells with a weaker ability to produce cytokines [54]. CD161+CD8+ T cells have the ability to secrete IL-17 and cytotoxicity [55, 56]. In addition, the researchers found increased infiltration of CD161+ immune cells in the spleen, kidney, and aorta of spontaneously hypertensive rats (SHR) and that IL-17 secreted by CD161+ cells mediated endothelial injury and hypertension [57]. CCR6 mediates cell migration through the ligand CCL20, which is usually highly expressed in tissues such as intestinal mucosa, lung mucosa, liver, and skin [58]. Above all, the reduction of CD45lowCD45ROlowCD161+CCR7−CCR6lowCCR5+CD8+ effector memory T cells in hypertension may be due to tissue infiltration or cell transformation.
*In* general, γδ T cells are CD4−CD8−CD3+ cells that can bind antigens without the major histocompatibility complex (MHC) [59]. They can produce IFN-γ and IL-17 in infection [60, 61] and have a strong proinflammatory effect [62]. And CD45lowCD27−CD45ROlowCD57hiGranzyme B+γδ T cells were defined as terminally differentiated cells with low proliferative and high effector capacity [59, 63, 64].
## 4.5. Enhanced B Cell Antigens in EH
B cells play an important role in humoral immunity by secreting antibodies and forming memory cells [65]. But the role of B cells in hypertension is still not very clear [10, 66]. Our results showed no significant difference in the number of B cells between the HC and EH groups, which was consistent with the previous study [10]. But the expression of most B cell antigens, CD45, IgM, CD19, CD45RO, Ki67, CXCR3, CCR7, CD24, CD38, CD31, CXCR5 and HLA-DR which affect the function of B cells, is significantly increased in EH. In summary, it is suggested that the activation, migration, and immune effects of B cells may be correspondingly augmented. Since we did not detect cytokines and soluble antibodies here, we were unable to further analyze the role of humoral immunity of B cells in hypertension.
## 5. Conclusion
The summary of this study is shown in Figure 5. Our results showed that the expression of many important antigens was significantly different in CD45+ PBMCs of hypertensive patients, and many characteristic antigens in granulocytes and B cells were enhanced; they may serve as characteristic antigens for the development of EH. The increased percentage of total DCs and two mDC subsets emphasized the distinct role of DCs in hypertension. The number of proinflammatory monocytes and normal monocytes in EH is unbalanced. LDNs are reduced in EH, and more research is needed to determine their origin and function. The percentage of the one naive CD4+ and one naive CD8+ T cell subsets was decreased in EH. However, the changes of memory T cells are diverse. In conclusion, our findings provide a more comprehensive description for the imbalanced immune state and inflammatory environment in peripheral blood of patients with EH, and larger sample size is needed in the future. The detailed mechanisms of altered immune cell state in EH require more research.
## Data Availability
The statistical data that support the findings of this study are available from the corresponding authors upon reasonable request.
## Ethical Approval
The study design was approved by the Ethics Committee at the Institute of Microcirculation at the CAMS & PUMC and adhered to the tenets of the Declaration of Helsinki.
## Conflicts of Interest
The authors declare that there is no conflict of interest regarding the publication of this paper.
## Authors' Contributions
Honggang Zhang and Qiuju Zhang provided the topic ideas and led the project. Changming Xiong, Yubao Zou, and Bingyang Liu assisted in the design of the clinical study and provided clinical samples. Bingwei Li contributed to the formation of the graphs and tables. Rui Yang and Yuhong He analyzed the data and wrote the manuscript. Rui Yang and Yuhong He contributed equally to this work and co-first author.
## References
1. Zhou B., Carrillo-Larco R. M., Danaei G., Riley L. M., Paciorek C. J., Stevens G. A.. **Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants**. (2021) **398** 957-980. DOI: 10.1016/S0140-6736(21)01330-1
2. Zhang J. R., Hu W. N., Li C. Y.. **A review of the epidemiological evidence for Adducin family gene polymorphisms and hypertension**. (2019) **2019** 6. DOI: 10.1155/2019/7135604
3. Mouton A. J., Li X., Hall M. E., Hall J. E.. **Obesity, hypertension, and cardiac dysfunction: novel roles of immunometabolism in macrophage activation and inflammation**. (2020) **126** 789-806. DOI: 10.1161/CIRCRESAHA.119.312321
4. Small H. Y., Migliarino S., Czesnikiewicz-Guzik M., Guzik T. J.. **Hypertension: focus on autoimmunity and oxidative stress**. (2018) **125** 104-115. DOI: 10.1016/j.freeradbiomed.2018.05.085
5. Drummond G. R., Vinh A., Guzik T. J., Sobey C. G.. **Immune mechanisms of hypertension**. (2019) **19** 517-532. DOI: 10.1038/s41577-019-0160-5
6. McMaster W. G., Kirabo A., Madhur M. S., Harrison D. G.. **Inflammation, immunity, and hypertensive end-organ damage**. (2015) **116** 1022-1033. DOI: 10.1161/CIRCRESAHA.116.303697
7. Kirabo A., Fontana V., de Faria A. P.. **DC isoketal-modified proteins activate T cells and promote hypertension**. (2014) **124** 4642-4656. DOI: 10.1172/JCI74084
8. Loperena R., Van Beusecum J. P., Itani H. A.. **Hypertension and increased endothelial mechanical stretch promote monocyte differentiation and activation: roles of STAT3, interleukin 6 and hydrogen peroxide**. (2018) **114** 1547-1563. DOI: 10.1093/cvr/cvy112
9. Reus-Chavarria E., Martinez-Vieyra I., Salinas-Nolasco C.. **Enhanced expression of the Epithelial Sodium Channel in neutrophils from hypertensive patients**. (2019) **1861** 387-402. DOI: 10.1016/j.bbamem.2018.11.003
10. Sereti E., Stamatelopoulos K. S., Zakopoulos N. A., Evangelopoulou A., Mavragani C. P., Evangelopoulos M. E.. **Hypertension: an immune related disorder?**. (2020) **212, article 108247**. DOI: 10.1016/j.clim.2019.108247
11. Chan C. T., Sobey C. G., Lieu M.. **Obligatory role for B cells in the development of angiotensin II-dependent hypertension**. (2015) **66** 1023-1033. DOI: 10.1161/HYPERTENSIONAHA.115.05779
12. Taylor E. B., Barati M. T., Powell D. W., Turbeville H. R., Ryan M. J.. **Plasma cell depletion attenuates hypertension in an experimental model of autoimmune disease**. (2018) **71** 719-728. DOI: 10.1161/HYPERTENSIONAHA.117.10473
13. Balasubbramanian D., Lopez Gelston C. A., Rutkowski J. M., Mitchell B. M.. **Immune cell trafficking, lymphatics and hypertension**. (2019) **176** 1978-1988. DOI: 10.1111/bph.14370
14. Kubiszewska I., Gackowska L., Obrycki L.. **Distribution and maturation state of peripheral blood dendritic cells in children with primary hypertension**. (2022) **45** 401-413. DOI: 10.1038/s41440-021-00809-9
15. Bras A. E., de Haas V., van Stigt A.. **CD123 expression levels in 846 acute leukemia patients based on standardized immunophenotyping**. (2019) **96** 134-142. DOI: 10.1002/cyto.b.21745
16. Lyszkiewicz M., Witzlau K., Pommerencke J., Krueger A.. **Chemokine receptor CX3CR1 promotes dendritic cell development under steady- state conditions**. (2011) **41** 1256-1265. DOI: 10.1002/eji.201040977
17. Sun X., Kaufman P. D.. **Ki-67: more than a proliferation marker**. (2018) **127** 175-186. DOI: 10.1007/s00412-018-0659-8
18. Oppermann M.. **Chemokine receptor CCR5: insights into structure, function, and regulation**. (2004) **16** 1201-1210. DOI: 10.1016/j.cellsig.2004.04.007
19. Ganjali S., Gotto A. M., Ruscica M.. **Monocyte-to-HDL-cholesterol ratio as a prognostic marker in cardiovascular diseases**. (2018) **233** 9237-9246. DOI: 10.1002/jcp.27028
20. Ozanska A., Szymczak D., Rybka J.. **Pattern of human monocyte subpopulations in health and disease**. (2020) **92**. DOI: 10.1111/sji.12883
21. Sampath P., Moideen K., Ranganathan U. D., Bethunaickan R.. **Monocyte subsets: phenotypes and function in tuberculosis infection**. (2018) **9**. DOI: 10.3389/fimmu.2018.01726
22. Rosales C.. **Neutrophil: a cell with many roles in inflammation or several cell types?**. (2018) **9** p. 113. DOI: 10.3389/fphys.2018.00113
23. Silvestre-Roig C., Fridlender Z. G., Glogauer M., Scapini P.. **Neutrophil diversity in health and disease**. (2019) **40** 565-583. DOI: 10.1016/j.it.2019.04.012
24. Morrissey S. M., Geller A. E., Hu X.. **A specific low-density neutrophil population correlates with hypercoagulation and disease severity in hospitalized COVID-19 patients**. (2021) **6**. DOI: 10.1172/jci.insight.148435
25. Cerecedo D., Martinez-Vieyra I., Lopez-Villegas E. O., Hernandez-Cruz A., Loza-Huerta A. D. C.. **Heterogeneity of neutrophils in arterial hypertension**. (2021) **402**. DOI: 10.1016/j.yexcr.2021.112577
26. Ren J., Crowley S. D.. **Role of T-cell activation in salt-sensitive hypertension**. (2019) **316** H1345-H1353. DOI: 10.1152/ajpheart.00096.2019
27. Al Barashdi M. A., Ali A., McMullin M. F., Mills K.. **Protein tyrosine phosphatase receptor type C (PTPRC or CD45)**. (2021) **74** 548-552. DOI: 10.1136/jclinpath-2020-206927
28. Liechti T., Roederer M.. **OMIP-060: 30-parameter flow cytometry panel to assess T cell effector functions and regulatory T cells**. (2019) **95** 1129-1134. DOI: 10.1002/cyto.a.23853
29. Mahnke Y. D., Brodie T. M., Sallusto F., Roederer M., Lugli E.. **The who’s who of T-cell differentiation: human memory T-cell subsets**. (2013) **43** 2797-2809. DOI: 10.1002/eji.201343751
30. Kawabe T., Yi J., Sprent J.. **Homeostasis of naive and memory T lymphocytes**. (2021) **13**. DOI: 10.1101/cshperspect.a037879
31. Itani H. A., Xiao L., Saleh M. A.. **CD70 exacerbates blood pressure elevation and renal damage in response to repeated hypertensive stimuli**. (2016) **118** 1233-1243. DOI: 10.1161/CIRCRESAHA.115.308111
32. Gray J. I., Westerhof L. M., Mac Leod M. K. L.. **The roles of resident, central and effector memory CD4 T-cells in protective immunity following infection or vaccination**. (2018) **154** 574-581. DOI: 10.1111/imm.12929
33. Mueller S. N., Gebhardt T., Carbone F. R., Heath W. R.. **Memory T cell subsets, migration patterns, and tissue residence**. (2013) **31** 137-161. DOI: 10.1146/annurev-immunol-032712-095954
34. Landsman L., Bar-On L., Zernecke A.. **CX3CR1 is required for monocyte homeostasis and atherogenesis by promoting cell survival**. (2009) **113** 963-972. DOI: 10.1182/blood-2008-07-170787
35. Li J., Zhou H., Fu X., Zhang M., Sun F., Fan H.. **Dynamic role of macrophage CX3CR1 expression in inflammatory bowel disease**. (2021) **232** 39-44. DOI: 10.1016/j.imlet.2021.02.001
36. Hochheiser K., Kurts C.. **Selective dependence of kidney dendritic cells on CX3CR1--implications for glomerulonephritis therapy**. (2015) **850** 55-71. DOI: 10.1007/978-3-319-15774-0_5
37. Hogan K. A., Chini C. C. S., Chini E. N.. **The multi-faceted Ecto-enzyme CD38: roles in immunomodulation, cancer, aging, and metabolic diseases**. (2019) **10** p. 1187. DOI: 10.3389/fimmu.2019.01187
38. Glaria E., Valledor A. F.. **Roles of CD38 in the immune response to infection**. (2020) **9** p. 228. DOI: 10.3390/cells9010228
39. Partida-Sanchez S., Goodrich S., Kusser K., Oppenheimer N., Randall T. D., Lund F. E.. **Regulation of dendritic cell trafficking by the ADP-ribosyl cyclase CD38: impact on the development of humoral immunity**. (2004) **20** 279-291. DOI: 10.1016/s1074-7613(04)00048-2
40. Frasca L., Fedele G., Deaglio S.. **CD38 orchestrates migration, survival, and Th1 immune response of human mature dendritic cells**. (2006) **107** 2392-2399. DOI: 10.1182/blood-2005-07-2913
41. Mahnke K., Schmitt E., Bonifaz L., Enk A. H., Jonuleit H.. **Immature, but not inactive: the tolerogenic function of immature dendritic cells**. (2002) **80** 477-483. DOI: 10.1046/j.1440-1711.2002.01115.x
42. Varani S., Frascaroli G., Homman-Loudiyi M., Feld S., Landini M. P., Soderberg-Naucler C.. **Human cytomegalovirus inhibits the migration of immature dendritic cells by down-regulating cell-surface CCR1 and CCR5**. (2004) **77** 219-228. DOI: 10.1189/jlb.0504301
43. Lipscomb M. W., Taylor J. L., Goldbach C. J., Watkins S. C., Wesa A. K., Storkus W. J.. **DC expressing transgene Foxp3 are regulatory APC**. (2010) **40** 480-493. DOI: 10.1002/eji.200939667
44. Gong Y. B., Hu L. N., Liu Y.. **Effects of Foxp3 gene modified dendritic cells on mouse corneal allograft rejection**. (2015) **8** 3965-3973. PMID: 26064298
45. Liu X. Y., Xu L. Z., Luo X. Q.. **Forkhead box protein-3 (Foxp3)-producing dendritic cells suppress allergic response**. (2017) **72** 908-917. DOI: 10.1111/all.13088
46. Ziegler-Heitbrock L.. **The CD14+ CD16+ blood monocytes: their role in infection and inflammation**. (2007) **81** 584-592. DOI: 10.1189/jlb.0806510
47. Tokunaga R., Zhang W., Naseem M.. **CXCL9, CXCL10, CXCL11/CXCR3 axis for immune activation - a target for novel cancer therapy**. (2018) **63** 40-47. DOI: 10.1016/j.ctrv.2017.11.007
48. von Vietinghoff S., Kurts C.. **Regulation and function of CX3CR1 and its ligand CX3CL1 in kidney disease**. (2021) **385** 335-344. DOI: 10.1007/s00441-021-03473-0
49. Gerritsen B., Pandit A.. **The memory of a killer T cell: models of CD8+ T cell differentiation**. (2016) **94** 236-241. DOI: 10.1038/icb.2015.118
50. Morante-Palacios O., Fondelli F., Ballestar E., Martinez-Caceres E. M.. **Tolerogenic dendritic cells in autoimmunity and inflammatory diseases**. (2021) **42** 59-75. DOI: 10.1016/j.it.2020.11.001
51. Xia Y., Liu A., Li W.. **Reference range of naïve T and T memory lymphocyte subsets in peripheral blood of healthy adult**. (2022) **207** 208-217. DOI: 10.1093/cei/uxab038
52. Falcao R. P., De-Santis G. C.. **Age-associated changes of memory (CD45RO+) and naive (CD45R+) T cells**. (1991) **24** 275-279. PMID: 1840425
53. Esensten J. H., Helou Y. A., Chopra G., Weiss A., Bluestone J. A.. **CD28 costimulation: from mechanism to therapy**. (2016) **44** 973-988. DOI: 10.1016/j.immuni.2016.04.020
54. Okada R., Kondo T., Matsuki F., Takata H., Takiguchi M.. **Phenotypic classification of human CD4+ T cell subsets and their differentiation**. (2008) **20** 1189-1199. DOI: 10.1093/intimm/dxn075
55. Fergusson J. R., Fleming V. M., Klenerman P.. **CD161-expressing human T cells**. (2011) **2** p. 36. DOI: 10.3389/fimmu.2011.00036
56. Konduri V., Oyewole-Said D., Vazquez-Perez J.. **CD8(+)CD161(+) T-cells: cytotoxic memory cells with high therapeutic potential**. (2020) **11, article 613204**. DOI: 10.3389/fimmu.2020.613204
57. Singh M. V., Cicha M. Z., Kumar S.. **Abnormal CD161**. (2017) **140** 809-821.e3. DOI: 10.1016/j.jaci.2016.11.039
58. Lee A. Y. S., Korner H.. **The CCR6-CCL20 axis in humoral immunity and T-B cell immunobiology**. (2019) **224** 449-454. DOI: 10.1016/j.imbio.2019.01.005
59. Xu W., Lau Z. W. X., Fulop T., Larbi A.. **The aging of**. (2020) **9**. DOI: 10.3390/cells9051181
60. Caillon A., Mian M. O. R., Fraulob-Aquino J. C.. (2017) **135** 2155-2162. DOI: 10.1161/CIRCULATIONAHA.116.027058
61. Muro R., Takayanagi H., Nitta T.. **T cell receptor signaling for**. (2019) **39** p. 6. DOI: 10.1186/s41232-019-0095-z
62. Li Y., Wu Y., Zhang C.. (2014) **64** 305-314. DOI: 10.1161/HYPERTENSIONAHA.113.02604
63. Caccamo N., Todaro M., Sireci G., Meraviglia S., Stassi G., Dieli F.. **Mechanisms underlying lineage commitment and plasticity of human**. (2013) **10** 30-34. DOI: 10.1038/cmi.2012.42
64. Li Y., Li G., Zhang J., Wu X., Chen X.. **The dual roles of human**. (2020) **11, article 619954**. DOI: 10.3389/fimmu.2020.619954
65. Yuseff M. I., Pierobon P., Reversat A., Lennon-Dumenil A. M.. **How B cells capture, process and present antigens: a crucial role for cell polarity**. (2013) **13** 475-486. DOI: 10.1038/nri3469
66. Guzik T. J., Hoch N. E., Brown K. A.. **Role of the T cell in the genesis of angiotensin II induced hypertension and vascular dysfunction**. (2007) **204** 2449-2460. DOI: 10.1084/jem.20070657
|
---
title: The Impact of Serum Parameters Associated with Kidney Function on the Short-Term
Outcomes and Prognosis of Colorectal Cancer Patients Undergoing Radical Surgery
authors:
- Bin Zhang
- Xu-Rui Liu
- Xiao-Yu Liu
- Bing Kang
- Chao Yuan
- Fei Liu
- Zi-Wei Li
- Zheng-Qiang Wei
- Dong Peng
journal: Canadian Journal of Gastroenterology & Hepatology
year: 2023
pmcid: PMC9988384
doi: 10.1155/2023/2017171
license: CC BY 4.0
---
# The Impact of Serum Parameters Associated with Kidney Function on the Short-Term Outcomes and Prognosis of Colorectal Cancer Patients Undergoing Radical Surgery
## Abstract
### Purpose
The current study was designed to investigate the impact of blood urea nitrogen (BUN), serum uric acid (UA), and cystatin (CysC) on the short-term outcomes and prognosis of colorectal cancer (CRC) patients undergoing radical surgery.
### Methods
CRC patients who underwent radical resection were included from Jan 2011 to Jan 2020 in a single clinical centre. The short-term outcomes, overall survival (OS), and disease-free survival (DFS) were compared in different groups. A Cox regression analysis was conducted to identify independent risk factors for OS and DFS.
### Results
A total of 2047 CRC patients who underwent radical resection were included in the current study. Patients in the abnormal BUN group had a longer hospital stay ($$p \leq 0.002$$) and more overall complications ($$p \leq 0.001$$) than that of the normal BUN group. The abnormal CysC group had longer hospital stay ($p \leq 0.01$), more overall complications (p=$p \leq 0.01$), and more major complications ($$p \leq 0.001$$) than the normal CysC group. Abnormal CysC was associated with worse OS and DFS for CRC patients in tumor stage I ($p \leq 0.01$). In Cox regression analysis, age ($p \leq 0.01$, HR = 1.041, $95\%$ CI = 1.029–1.053), tumor stage ($p \leq 0.01$, HR = 2.134, $95\%$ CI = 1.828–2.491), and overall complications ($$p \leq 0.002$$, HR = 1.499, $95\%$ CI = 1.166–1.928) were independent risk factors for OS. Similarly, age ($p \leq 0.01$, HR = 1.026, $95\%$ CI = 1.016–1.037), tumor stage ($p \leq 0.01$, HR = 2.053, $95\%$ CI = 1.788–2.357), and overall complications ($$p \leq 0.002$$, HR = 1.440, $95\%$ CI = 1.144–1.814) were independent risk factors for DFS.
### Conclusion
In conclusion, abnormal CysC was significantly associated with worse OS and DFS at TNM stage I, and abnormal CysC and BUN were related to more postoperative complications. However, preoperative BUN and UA in the serum might not affect OS and DFS for CRC patients who underwent radical resection.
## 1. Introduction
Colorectal cancer (CRC) is the second most fatal tumor worldwide, and it was estimated that nearly $9.4\%$ of cancer-related deaths would be caused by CRC in 2020 [1–3]. The most effective method for the therapy of CRC is still radical surgery [4–6]. Although great progress was made in the surgical techniques, the prognosis of these patients varied for different reasons, such as tumor stage [7, 8], comorbidities [9–11], and complications [12, 13]. For better clinical decisions and to improve the survival of CRC patients, many biochemical indicators, such as albumin [14, 15] and bilirubin [16, 17], were identified to find patients with high risks of postoperative complications and a poor prognosis.
It was reported that chronic kidney disease (CKD) could increase postoperative complications and worsen the OS for patients who accepted radical surgery [18–20]. CKD is usually identified and classified by the glomerular filtration rate (GFR) [21]. Besides GFR, when the glomerular filtration function began to deteriorate, blood urea nitrogen (BUN) [22], cystatin C (CysC) [23], and serum uric acid (UA) [24] were also elevated. What's more, the changes in CysC and serum UA were more sensitive and prominent than serum creatinine in the early period when glomerular filtration function was impaired [25]. As a result, we deduced that BUN, UA, and CysC might be related to the short-term outcomes and prognosis for CRC patients undergoing radical resection as well.
Both CysC and UA were proved to be interacted with tumor development and invasion. Previous studies reported the CySc was a marker for the prognosis of urinary system carcinoma [26, 27], esophageal cancer [28], and lung cancer patients [29]. Only Kos J et al. reported that CRC patients, after surgery with high cystatin C, had lower survival [30]. Similarly, the level of UA in the serum was correlated with the survival of patients with pancreatic cancer [31], laryngeal cancer [32], and so on, but its specific role in the prognosis for CRC patients remained controversial. Meanwhile, little was known about the predictive value of these factors for short-term outcomes.
As a result, the current study was designed to investigate the impact of BUN, CysC, and UA in serum on the short-term outcomes and prognosis of CRC patients undergoing radical surgery.
## 2.1. Patients
Patients who underwent radical CRC surgery were included from Jan 2011 to Jan 2020 in a single clinical center. The study was approved by the ethics committee of our institution (the First Affiliated Hospital of Chongqing Medical University, 2022-135-2), and all patients signed informed consent forms. This study was conducted in accordance with the World Medical Association Declaration of Helsinki as well.
## 2.2. Inclusion and Exclusion Criteria
Patients who underwent radical CRC surgery were included ($$n = 5473$$). The exclusion criteria were as follows: 1, non-R0 surgery ($$n = 25$$); 2, incomplete clinical data ($$n = 849$$); and 3, incomplete records of BUN, UA, and CysC before surgery ($$n = 2552$$). Finally, a total of 2047 CRC patients were included in this study (Figure 1).
## 2.3. Data Collection
The values of BUN, UA, and CysC were determined by the blood tests conducted a week before surgery. The baseline characteristics collected were as follows: age, sex, body mass index (BMI), smoking, drinking, hypertension, type 2 diabetes mellitus (T2DM), coronary heart disease (CHD), surgical method, tumor location, tumor node metastasis (TNM) stage, and tumor size. The short-term outcomes included operation time, intraoperative blood loss, postoperative hospital stay, overall complications, and major complications. The long-term prognosis was estimated by the OS and DFS. All the data were collected from the electronic medical record system, outpatient visits, and telephone interviews.
## 2.4. Definitions
The TNM stage was identified according to the AJCC 8th Edition [33]. The postoperative complications were classified on the basis of the Clavien-Dindo classification [34], and major complications were regarded as ≥ grade III. OS was defined as the time from surgery to death or loss of follow-up. DFS was calculated from the date of surgery to the date of recurrence or death.
## 2.5. Treatment and Follow-Up
All patients underwent radical surgery according to standard principles, and R0 resection was confirmed by pathology. Patients were regularly followed up every six months in the first three years and every year in the next years.
## 2.6. Optimal Cut-Off and Groups
According to the upper limits of the reference ranges of BUN, UA, and CysC, patients were divided into the abnormal BUN group (BUN>8.2 mmol/L) and the normal BUN group (BUN≤8.2 mmol/L); the abnormal UA group (UA>357 μmol/L) and the normal UA group (UA≤357 μmol/L); as well as the abnormal CysC group (CysC>1.09 mg/L) and the normal CysC group (CysC≤1.09 mg/L).
## 2.7. Statistical Analysis
A normality test was performed on the measurement data. The measurement data conforming to the normal distribution were expressed as mean ± standard deviation (SD), and an independent-samplet-test was used to compare the indicators between groups; the measurement data not conforming to the normal distribution were expressed as the median (minimum value and maximum value), and a Mann−Whitney U test was adopted for comparison between groups. Categorical variables are expressed as absolute values and percentages, and Chi-square test or Fisher's exact test was performed. The Kaplan−Meier method was used to estimate the OS and DFS, and a log-rank test was conducted to compare the OS and DFS between the CysC groups in different tumor stages. Moreover, Cox regression analysis was performed to identify independent risk factors for OS and DFS. Data were analyzed using SPSS (version 22.0) statistical software. A bilateral p value of <0.05 was considered statistically significant.
## 3.1. Patients and Characteristics
A total of 2047 CRC patients who underwent radical resection were included in the current study, and these patients were divided into different groups according to the values of BUN, UA, and CysC.
As a result, there were 1937 patients in the normal BUN group and 110 patients in the abnormal BUN group. The abnormal BUN group had an older age ($p \leq 0.01$), more males ($p \leq 0.01$), higher portion of smoking ($$p \leq 0.001$$), drinking ($$p \leq 0.004$$), hypertension ($p \leq 0.01$), and T2DM ($$p \leq 0.001$$) than the normal BUN group (Table 1).
Similarly, 1756 patients were in the normal UA group, and 291 patients were in the abnormal UA group. The abnormal UA group had an older age ($$p \leq 0.009$$), a higher BMI ($p \leq 0.01$), higher incidence of hypertension ($p \leq 0.01$) and CHD ($$p \leq 0.038$$), and more tumor size< 5 cm ($$p \leq 0.016$$). ( Table 2).
Moreover, 1627 patients and 420 patients were included in the normal CysC group and the abnormal CysC group, respectively. The abnormal CysC group had older age ($p \leq 0.01$), more males ($p \leq 0.01$), a higher portion of smoking ($p \leq 0.01$), and drinking ($$p \leq 0.013$$), a higher incidence of hypertension ($p \leq 0.01$), T2DM ($$p \leq 0.017$$), and CHD ($p \leq 0.01$), more open surgery ($p \leq 0.01$). ( Table 3).
## 3.2. Short-Term Outcomes
The short-term outcomes were compared in different groups. Accordingly, no difference was found between the normal UA group and the abnormal UA group ($p \leq 0.05$). Patients in the abnormal BUN group had a longer hospital stay ($$p \leq 0.002$$) and more overall complications ($$p \leq 0.001$$) than the normal BUN group. The abnormal CysC group had a longer hospital stay ($p \leq 0.01$), more overall complications ($p \leq 0.01$), and more major complications ($$p \leq 0.001$$) than the normal CysC group (Tables 1–3).
## 3.3. Cox Analyses for OS and DFS
Cox regression analyses were conducted to identify the independent risk factors for OS and DFS. As a consequence, age ($p \leq 0.01$, HR = 1.039, $95\%$ CI = 1.028–1.050), sex ($$p \leq 0.009$$, HR = 0.716, $95\%$ CI = 0.558–0.919), tumor stage ($p \leq 0.01$, HR = 2.123, $95\%$ CI = 1.823–2.473), smoking ($$p \leq 0.012$$, HR = 1.356, $95\%$ CI = 1.070–1.717), tumor size ($$p \leq 0.002$$, HR = 1.451, $95\%$ CI = 1.147–1.837), CysC ($$p \leq 0.006$$, HR = 1.441, $95\%$ CI = 1.108–1.875), and overall complications ($p \leq 0.01$, HR = 1.682, $95\%$ CI = 1.311–2.158) were potential risk factors for OS. In multivariate analysis, age ($p \leq 0.01$, HR = 1.041, $95\%$ CI = 1.029–1.053), tumor stage ($p \leq 0.01$, HR = 2.134, $95\%$ CI = 1.828–2.491), and overall complications ($$p \leq 0.002$$, HR = 1.499, $95\%$ CI = 1.166–1.928) were independent risk factors for OS (Table 4).
As for DFS, age ($p \leq 0.01$, HR = 1.026, $95\%$ CI = 1.017–1.036), sex ($$p \leq 0.044$$, HR = 0.797, $95\%$ CI = 0.639–0.994), tumor stage ($p \leq 0.01$, HR = 2.053, $95\%$ CI = 1.791–2.352), smoking ($$p \leq 0.020$$, HR = 1.288, $95\%$ CI = 1.041–1.594), tumor size ($$p \leq 0.007$$, HR = 1.340, $95\%$ CI = 1.084–1.656), CysC ($$p \leq 0.012$$, HR = 1.357, $95\%$ CI = 1.068–1.723), and overall complications ($p \leq 0.01$, HR = 1.542, $95\%$ CI = 1.227–1.937) were potential indicators. Furthermore, age ($p \leq 0.01$, HR = 1.026, $95\%$ CI = 1.016–1.037), tumor stage ($p \leq 0.01$, HR = 2.053, $95\%$ CI = 1.788–2.357), and overall complications ($$p \leq 0.002$$, HR = 1.440, $95\%$ CI = 1.144–1.814) were independent risk factors (Table 5).
However, none of BUN, CysC, or UA were independent risk factors for OS or DFS ($p \leq 0.05$).
## 3.4. Kaplan−Meier Curves in Different TNM Stages
The median follow-up time was 35 (1–114) months. Since CysC was found to be a potential risk factor for OS and DFS, we adopted the Kaplan−Meier method and log-rank test to compare the OS (Figure 2) and DFS (Figure 3) between the abnormal CysC group and the normal CysC group in TNM stages I–IV. Consequently, abnormal CysC were associated with worse OS ($p \leq 0.01$) and DFS ($p \leq 0.01$) for CRC patients in TNM stage I. However, no significant difference was found between the two groups for OS and DFS in stages II–IV ($p \leq 0.05$).
## 4. Discussion
A total of 2047 CRC patients were enrolled in the current study. We investigated the impact of biochemical indicators, including BUN, UA, and CysC, which were associated with kidney function, on the short-term outcomes and prognosis of CRC patients who underwent radical surgery.
It was reported that nearly $15\%$ of CRC patients had CKD [35]. Previous studies found that CRC patients with CKD had more postoperative complications, especially cardiovascular diseases [18–20]. The abnormal renal function also led to an increase in BUN, UA, and CysC in serum. In this study, patients in the abnormal BUN group had longer hospital stay and more overall complications than the normal BUN group, and patients in the abnormal CysC group had a longer hospital stay and more overall complications and major complications than the normal CysC group. However, we found the abnormal level of UA did not affect the short-term outcomes. The CysC was a sensitive indicator which could early identify the injury of kidney filtration function [23]. Thus, the monitoring of preoperative CysC might help to early identify patients with postoperative complication risks.
BUN was one of the main products in protein metabolism, and it was usually used to estimate glomerular filtration function [22]. The BUN in the serum began to increase only if the GFR decreased to less than $50\%$, which reflected the severity of CKD. Sohal DP et al. found elevated BUN before surgery indicated worse OS in pancreatic adenocarcinoma, which was simply explained as that higher BUN might imply subclinical organ dysfunction. However, whether preoperative BUN affected the prognosis of CRC patients was rarely reported, and our study found that BUN was not associated with the OS or DFS. The underlying mechanism needs to be further studied.
UA was an antioxidant as well as a pro-oxidant, which was produced from purine nucleotides, and the process was mediated by xanthine oxidase [36, 37]. It was widely reported that oxidative stress could facilitate the development of tumors; therefore, the prognostic value of UA might be controversial. Dziaman et al. first reported that CRC patients with high levels of UA in their serum had longer survival in a cohort study conducted in Poland [38]. However, in China, Mao et al. obtained the opposite conclusion that lower UA-level patients lived longer than those with higher serum UA [39]. The author attributed the incongruity to racial differences. Moreover, in a retrospective study including 332 patients, it was found that a higher preoperative UA was a risk factor for OS [40]. Nevertheless, different from the conclusions above, we found that preoperative UA had no obvious impact on OS or DFS for CRC patients.
In this study, although higher CysC was found to be associated with worse OS and DFS in CRC patients in tumor stage I, CysC was not an independent risk factor for DFS and OS. Kos demonstrated that patients with higher CysC had worse OS but it was not an independent indicator as well [30]. Besides the capacity to indicate the injury of kidney function, CysC was an inhibitor of cysteine proteinases, and the imbalance between cysteine proteinases and its inhibitors was proved to promote tumor invasion and metastasis [41]. As a result, the level of CysC in the serum might reflect the activity of tumor cells and the intensity of antitumor reactions in the body of cancer patients, which partly helped to explain the correlation between CysC and prognosis. However, it remained unclear why only patients in TNM stage I had worse OS and DFS.
To our knowledge, this was the first study to find that abnormal CysC was associated with more postoperative complications and worse OS and DFS in CRC patients with a relatively large sample size. Meanwhile, we also pointed out that preoperative UA had no obvious impact on OS and DFS for CRC patients, which was inconsistent with previous studies. Nevertheless, there were some limitations in our study as well. For this was a retrospective study conducted in a single clinical center, confounding bias was inevitable. Second, chemotherapeutic information was lacking in TNM III-IV patients, which might impair the reliability of the survival analysis. Therefore, multicenter prospective studies with a large sample size are needed to identify the predictive roles of these indicators.
In conclusion, abnormal CysC was significantly associated with worse OS and DFS at TNM stage I, and abnormal CysC and BUN were related to more postoperative complications. However, preoperative BUN and UA in the serum might not affect OS and DFS for CRC patients who underwent radical resection.
## Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
## Ethical Approval
The study was approved by the Ethics Committee of our institution (the First Affiliated Hospital of Chongqing Medical University, 2022-135-2). This study was conducted in accordance with the World Medical Association Declaration of Helsinki as well.
## Consent
All patients signed informed consent.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## Authors' Contributions
All authors contributed to data collection and analysis, drafting or revising the manuscript, have agreed on the journal to which the manuscript will be submitted, gave final approval of the version to be published, and agree to be accountable for all aspects of the work. Bin Zhang and Xu-Rui Liu contributed equally to this work.
## References
1. Sung H., Ferlay J., Siegel R. L.. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. (2021) **71** 209-249. DOI: 10.3322/caac.21660
2. Peng D., Liu X. Y., Cheng Y. X., Tao W., Cheng Y.. **Improvement of diabetes mellitus after colorectal cancer surgery: a retrospective study of predictive factors for type 2 diabetes mellitus remission and overall survival**. (2021 Jul 6) **11**. DOI: 10.3389/fonc.2021.694997
3. Hossain M. S., Karuniawati H., Jairoun A. A.. **Colorectal cancer: a review of carcinogenesis, global epidemiology, current challenges, risk factors, preventive and treatment strategies**. (2022) **14** p. 1732. DOI: 10.3390/cancers14071732
4. Hashiguchi Y., Muro K., Saito Y.. **Japanese Society for Cancer of the Colon and Rectum (JSCCR) guidelines 2019 for the treatment of colorectal cancer**. (2020) **25** 1-42. DOI: 10.1007/s10147-019-01485-z
5. Cheng Y. X., Tao W., Liu X. Y.. **Hypertension remission after colorectal cancer surgery: a single-center retrospective study**. (2022) **74** 2789-2795. DOI: 10.1080/01635581.2021.2025256
6. Tanis P. J., Buskens C. J., Bemelman W. A.. **Laparoscopy for colorectal cancer**. (2014) **28** 29-39. DOI: 10.1016/j.bpg.2013.11.017
7. Li B. W., Ma X. Y., Lai S., Sun X., Sun M. J., Chang B.. **Development and validation of a prognostic nomogram for colorectal cancer after surgery**. (2021) **9** 5860-5872. DOI: 10.12998/wjcc.v9.i21.5860
8. Liu X. X., Su J., Long Y. Y., He M., Zhu Z. Q.. **Perioperative risk factors for survival outcomes in elective colorectal cancer surgery: a retrospective cohort study**. (2021) **21** p. 169. DOI: 10.1186/s12876-021-01757-x
9. Cheng Y. X., Tao W., Zhang H., Peng D., Wei Z. Q.. **Does liver cirrhosis affect the surgical outcome of primary colorectal cancer surgery? A meta-analysis**. (2021) **19** p. 167. DOI: 10.1186/s12957-021-02267-6
10. Cheng Y., Cheng Y. X., Liu X. Y., Kang B., Tao W., Peng D.. **The effect of type 2 diabetes mellitus on the short-term outcomes and prognosis of stage I-III colorectal cancer: a propensity score matching analysis**. (2022) **14** 205-214. DOI: 10.2147/CMAR.S347242
11. Sakin A., Samanci N. S., Secmeler S.. **The effect of body mass index on location of recurrence and survival in early-stage colorectal cancer**. (2020) **16** S176-S182. DOI: 10.4103/jcrt.JCRT_326_18
12. Aoyama T., Oba K., Honda M.. **Impact of postoperative complications on the colorectal cancer survival and recurrence: analyses of pooled individual patients’ data from three large phase III randomized trials**. (2017) **6** 1573-1580. DOI: 10.1002/cam4.1126
13. Tevis S. E., Kohlnhofer B. M., Stringfield S.. **Postoperative complications in patients with rectal cancer are associated with delays in chemotherapy that lead to worse disease-free and overall survival**. (2013) **56** 1339-1348. DOI: 10.1097/DCR.0b013e3182a857eb
14. Lai C. C., You J. F., Yeh C. Y.. **Low preoperative serum albumin in colon cancer: a risk factor for poor outcome**. (2011) **26** 473-481. DOI: 10.1007/s00384-010-1113-4
15. Chandrasinghe P. C., Ediriweera D. S., Kumarage S. K., Deen K. I.. **Pre-operative hypoalbuminaemia predicts poor overall survival in rectal cancer: a retrospective cohort analysis**. (2013) **13** p. 12. DOI: 10.1186/1472-6890-13-12
16. Jia Z., Zhu Z., Wang Y.. **The prognostic value of serum bilirubin in colorectal cancer patients with surgical resection**. (2021) **36** 14-22. DOI: 10.1177/17246008211036128
17. Zhang Q., Ma X., Xu Q.. **Nomograms incorporated serum direct bilirubin level for predicting prognosis in stages II and III colorectal cancer after radical resection**. (2016) **8**. DOI: 10.18632/oncotarget.11424
18. Hu W. H., Cajas-Monson L. C., Eisenstein S., Parry L., Ramamoorthy S.. **Association of dialysis with adverse postoperative outcomes in colorectal cancer-an analysis of ACS-NSQIP**. (2015) **30** 1557-1562. DOI: 10.1007/s00384-015-2347-y
19. Currie A., Malietzis G., Askari A.. **Impact of chronic kidney disease on postoperative outcome following colorectal cancer surgery**. (2014) **16** 879-885. DOI: 10.1111/codi.12665
20. Nozawa H., Kitayama J., Sunami E., Watanabe T.. **Impact of chronic kidney disease on outcomes of surgical resection for primary colorectal cancer: a retrospective cohort review**. (2012) **55** 948-956. DOI: 10.1097/DCR.0b013e3182600db7
21. Zsom L., Zsom M., Salim S. A., Fülöp T.. **Estimated glomerular filtration rate in chronic kidney disease: a critical review of estimate-based predictions of individual outcomes in kidney disease**. (2022) **14** p. 127. DOI: 10.3390/toxins14020127
22. Jamshidi P., Najafi F., Mostafaei S.. **Investigating associated factors with glomerular filtration rate: structural equation modeling**. (2020) **21** p. 30. DOI: 10.1186/s12882-020-1686-2
23. Pandey V. K., Mani P., Mazumdar M. L.. **Study of serum cystatin C and serum creatinine in different stages of chronic kidney disease patients**. (2022) **70** 11-12
24. Weiner D. E., Tighiouart H., Elsayed E. F., Griffith J. L., Salem D. N., Levey A. S.. **Uric acid and incident kidney disease in the community**. (2008) **19** 1204-1211. DOI: 10.1681/ASN.2007101075
25. Jovanović D., Krstivojević P., Obradović I., Durdević V., Dukanović L.. **Serum cystatin C and**. (2003) **25** 123-133. DOI: 10.1081/jdi-120017475
26. Yang L., Wei Q., Tan P.. **The preoperative serum cystatin-C as an independent prognostic factor for survival in upper tract urothelial carcinoma**. (2019) **21** 163-169. DOI: 10.4103/aja.aja_84_18
27. Guo K., Chen Q., He X.. **Expression and significance of Cystatin-C in clear cell renal cell carcinoma**. (2018) **107** 1237-1245. DOI: 10.1016/j.biopha.2018.08.083
28. Yan Y., Zhou K., Wang L., Wang F., Chen X., Fan Q.. **Clinical significance of serum cathepsin B and cystatin C levels and their ratio in the prognosis of patients with esophageal cancer**. (2017) **10** 1947-1954. DOI: 10.2147/OTT.S123042
29. Zhang X., Hou Y., Niu Z.. **[Clinical significance of detection of cathepsin X and cystatin C in the sera of patients with lung cancer]**. (2013) **16** 411-416. DOI: 10.3779/j.issn.1009-3419.2013.08.04
30. Kos J., Krasovec M., Cimerman N., Nielsen H. J., Christensen I. J., Brünner N.. **Cysteine proteinase inhibitors stefin A, stefin B, and cystatin C in sera from patients with colorectal cancer: relation to prognosis**. (2000) **6** 505-511. PMID: 10690531
31. Stotz M., Szkandera J., Seidel J.. **Evaluation of uric acid as a prognostic blood-based marker in a large cohort of pancreatic cancer patients**. (2014) **9**. DOI: 10.1371/journal.pone.0104730
32. Hsueh C. Y., Shao M., Cao W., Li S., Zhou L.. **Pretreatment serum uric acid as an efficient predictor of prognosis in men with laryngeal squamous cell cancer: a retrospective cohort study**. (2019) **2019** 12. DOI: 10.1155/2019/1821969
33. Weiser M. R.. **AJCC 8th edition: colorectal cancer**. (2018) **25** 1454-1455. DOI: 10.1245/s10434-018-6462-1
34. Clavien P. A., Barkun J., de Oliveira M. L.. **The Clavien-Dindo classification of surgical complications: five-year experience**. (2009) **250** 187-196. DOI: 10.1097/SLA.0b013e3181b13ca2
35. Kozłowski L., Kozłowska K., Małyszko J.. **Hypertension and chronic kidney disease is highly prevalent in elderly patients with colorectal cancer undergoing primary surgery**. (2019) **28** 1425-1428. DOI: 10.17219/acem/104537
36. Lanaspa M. A., Sanchez-Lozada L. G., Choi Y. J.. **Uric acid induces hepatic steatosis by generation of mitochondrial oxidative stress: potential role in fructose-dependent and -independent fatty liver**. (2012) **287**. DOI: 10.1074/jbc.M112.399899
37. Sautin Y. Y., Nakagawa T., Zharikov S., Johnson R. J.. **Adverse effects of the classic antioxidant uric acid in adipocytes: NADPH oxidase-mediated oxidative/nitrosative stress**. (2007) **293** C584-C596. DOI: 10.1152/ajpcell.00600.2006
38. Dziaman T., Banaszkiewicz Z., Roszkowski K.. **8-Oxo-7,8-dihydroguanine and uric acid as efficient predictors of survival in colon cancer patients**. (2014) **134** 376-383. DOI: 10.1002/ijc.28374
39. Mao L., Guo C., Zheng S.. **Elevated urinary 8-oxo-7,8-dihydro-2&#39;-deoxyguanosine and serum uric acid are associated with progression and are prognostic factors of colorectal cancer**. (2018) **11** 5895-5902. DOI: 10.2147/OTT.S175112
40. Üstüner M. A., Dogan L.. **Relationship of preoperative serum uric acid level with survival in colorectal cancer**. (2020) **30** 717-721. DOI: 10.29271/jcpsp.2020.07.717
41. Kos J., Werle B., Lah T., Brunner N.. **Cysteine proteinases and their inhibitors in extracellular fluids: markers for diagnosis and prognosis in cancer**. (2000) **15** 84-89. DOI: 10.1177/172460080001500116
|
---
title: Chitosan Nanoparticles Alleviated the Adverse Effects of Sildenafil on the
Oxidative Stress Markers and Antioxidant Enzyme Activities in Rats
authors:
- Salah A. Sheweita
- Dalia M. Elsayed Alian
- Medhat Haroun
- Mohamed Ismail Nounou
- Ayyub Patel
- Labiba El-Khordagui
journal: Oxidative Medicine and Cellular Longevity
year: 2023
pmcid: PMC9988388
doi: 10.1155/2023/9944985
license: CC BY 4.0
---
# Chitosan Nanoparticles Alleviated the Adverse Effects of Sildenafil on the Oxidative Stress Markers and Antioxidant Enzyme Activities in Rats
## Abstract
Sildenafil (SF) is widely used for erectile dysfunction and other conditions, though with limitations regarding oral absorption and adverse effects. Despite nanotechnological improvements, the effect of nanocarriers on SF hepatotoxicity has not been documented to date. This study aimed at assessing the impact of chitosan nanoparticles either uncoated (CS NPs) or Tween 80-coated (T-CS NPs) on the effects of SF on oxidative stress markers and antioxidant enzyme activities in rats. Test SF-CS NPs prepared by ionic gelation were uniform positively charged nanospheres (diameter 178-215 nm). SF was administered intraperitoneally to male rats (1.5 mg/kg body weight) in free or nanoencapsulated forms as SF-CS NPs and T-SF-CS NPs for 3 weeks. Free SF significantly suppressed the activity of the antioxidant enzymes glutathione S-transferase (GST), glutathione peroxidase (GPx), glutathione reductase (GR), catalase (CAT), and superoxide dismutase (SOD), as well as the levels of glutathione (GSH) and thiobarbituric acid reactive substances (TBARS) as in an indirect measure of free radicals. Interestingly, SF-CS NPs and T-SF-CS-NPs treatments significantly attenuated the inhibitory effects of SF on the activity of these enzymes whereas, GST activity was inhibited. Moreover, the protein expression of GST was downregulated upon treatment of rats with free SF, SF-CS-NPs, and T-SF CS-NPs. In contrast, the activity and protein expression of GPx was induced by SF-CS NPs and T-SF-CS-NPs treatments. The histopathological study showed that SF induced multiple adverse effects on the rat liver architecture which were markedly suppressed particularly by T-SF-CS NPs. In conclusion, chitosan nanoencapsulation of SF counteracted the adverse effects of SF on the activity of antioxidant enzymes and liver architecture. Findings might have significant implications in improving the safety and efficacy of SF treatment of the widely expanding disease conditions.
## 1. Introduction
Sildenafil (SF) is a vasoactive first-generation phosphodiesterase-5 (PDE5) inhibitor approved for the treatment of male erectile dysfunction [1] and pulmonary arterial hypertension [2]. PDE5 inhibitors including sildenafil, tadalafil, and vardenafil act by inhibiting the phosphodiesterase type 5 (PDE5) enzyme present in high concentration in the corpus cavernosum. PDE5 specifically breaks down the cyclic guanosine monophosphate (cGMP), responsible for nitric oxide-induced smooth muscle relaxation and vasodilatation [3]. As the PDE5 enzyme is also distributed in many cells throughout the body, PDE5 inhibitors have the potential for wider use in different clinical conditions such as neurodegenerative disorders and brain injuries [4, 5], heart failure [6], and cancer [7].
SF is usually taken orally as film-coated tablets, though with limitations including low oral bioavailability (~$40\%$) as well as delayed onset (30–45 min) and short (half-life ~3 h) duration of action [8]. SF is extensively metabolized in the liver by cytochrome P450 enzymes, mainly CYP3A4 and to a lesser extent CYP2C9 [9], resulting in the loss of a large part of an oral dose [10]. Conversely, SF and other PDE5 inhibitors may affect the activity of antioxidant enzymes such as glutathione S-transferase (GST), glutathione peroxidase (GPx), glutathione reductase (GR), catalase (CAT), and superoxide dismutase (SOD), the levels of thiobarbituric acid reactive substances (TBARS) and glutathione (GSH) as well as the protein expression of different CYPs isozymes in the livers of experimental animals [11–13]. Increased amounts of reactive oxygen species (ROS), such as superoxide (O2▪−) and hydroxyl radical (OH▪), and reactive nitrogen species (RNS) as nitric oxide (NO▪) and nitrogen dioxide (NO2▪) radicals cause oxidative and nitrosative stress, respectively, when antioxidant systems are suppressed. Such situations play a significant role in the pathophysiology of erectile dysfunction [14] and other disorders [15, 16].
Approaches most adopted to overcome the limitations of oral SF comprise alternative routes of administration to bypass the liver [17, 18] and nanotechnology to improve SF bioavailability, prolong its action and reduce its adverse effects [19, 20]. Despite achievements in this respect, the potential influence of nanocarriers on the hepatic effects of SF has not been documented to date. This initiated our interest in investigating the influence of nanoencapsulation on SF hepatic effects. Chitosan (CS) was selected as the carrier polymer because of its beneficial bioactivities [21] as well as its important role in effective systemic drug delivery and tunable cellular uptake of drugs [22]. Additionally, CS has been shown to protect the liver from hepatocarcinogens and other drug-induced toxicity [23, 24].
This study aimed at assessing the changes induced in the livers of rats upon intraperitoneal administration of SF, SF-CS NPS, and T-SF CS NPs for 21 days both biochemically and histopathologically. Biochemical assessments included the change in the activity of antioxidant enzymes (GST, GR, GPx, SOD, and catalase CAT), the level of GSH and TBARs as an indirect measure of free radicals, and the protein expression of GST and GPx. Biochemical assessments were supported by histopathological examination of changes in the liver of rats.
## 2.1. Materials
Low molecular weight chitosan (50-190 kDa, 75-$85\%$ deacetylated) (SC), sildenafil citrate (SF), 5,5′-dithiobis nitrobenzoic acid (DTNB), nicotinamide adenine dinucleotide phosphate (NADPH), bovine serum albumin (BSA), reduced glutathione (GSH), 1-chloro-2,4-dinitrobenzene (CDNB), epinephrine, acrylamide, bisacrylamide, tetramethylethylenediamine (TEMED), cumene hydroperoxide, and tris-HCl were purchased from Sigma Aldrich, Germany. Sodium tripolyphosphate (TPP) was purchased from Loba Chemie, India. Folin-Cioclateu phenol reagent was purchased from Oxford Lab Chem, India. Tween 80, potassium phosphate, trichloroacetic acid (TCA), thiobarbituric acid (TBA), sodium phosphate, hydrogen peroxide (H2O2), sodium carbonate, magnesium chloride, acetone, sodium hydroxide (NaOH), sodium borate, sodium carbonate (Na2CO3), copper sulfate (CuSO4), Na-K tartrate, sulphosalicylic acid, ammonium persulphate (APS), and sodium dodecyl sulfate (SDS) were purchased from El-Nasr Pharmaceutical Company, Egypt. Primary anti-mouse antibodies for GPx and GST were obtained from Santa Cruz Co., USA.
## 2.2. Preparation of Chitosan-Based Nanoparticles (CS NPs)
Chitosan-based nanoparticles including blank CS NPs, SF-loaded chitosan nanoparticles (SF-CS NPs), and Tween 80-coated SF-loaded CS-NPs (T- SF-CS NPs) were prepared using essentially a TPP ionic gelation method with some modification [25]. In brief, low molecular weight CS was dissolved in an acetic acid solution with magnetic stirring overnight. After pH adjustment to 4.7-4.8 with $20\%$ NaOH, the CS solution was filtered using a 0.45 μm nylon syringe filter. A cold-filtered TPP solution (3 mL) was added to the CS solution (10 mL) in a water bath at 60°C under magnetic stirring for about 10 min and the formed blank SF-CS NPs were separated. To adjust the physical properties of plain CS-NPs, particularly, the particle size and size distribution, a series of preliminary trials based on the preparation of NPs by changing the experimental variables one at a time while keeping other variables constant was undertaken. The variables included the concentration of CS and TPP solutions (0.5 mg/mL vs. 2 mg/mL), the temperature of the CS solution during TPP addition (25°C vs. 60°C) as well and the method of separation of the formed CS NPs (low-speed centrifugation at 3000 rpm for 10 min at ambient temperature vs probe sonication for 10 and 20 min).
SF-CS-NPs were prepared by dissolving SF (2 mg/mL) in the CS solution as reported earlier for berberine-loaded CS NPs [26], and the procedure was completed as described above. For the preparation of T-SF-CS-NPs, freshly prepared SF-CS NPs were resuspended in $1\%$ Tween 80 solution and sonicated for 20 min in a water bath sonicator [27].
## 2.3.1. Physical Properties
The SF-CS-NPs and T-SF-CS-NPs were characterized for particle size and size distribution expressed as polydispersity index (PDI) by dynamic light scattering (DLS) using Zetasizer Nano ZS Series DTS 1060, Malvern Instruments S.A., Worcestershire, UK at a scattering angle of 90° at 25°C using a 4-mW He–Ne laser at 633 nm. The NP dispersions were suitably diluted 1: 80 in deionized water and measurements were performed in triplicate. Zeta potential was determined at 25°C in water using a cell voltage of 150 V and 5 mA current.
## 2.3.2. Transmission Electron Microscopy (TEM)
The morphology of the test NPs was examined by TEM using JEOL, JEM-100 CX Electron Microscope (Tokyo, Japan). Before analysis, NP dispersions were sprayed onto copper grids and stained with $2\%$ w/v uranyl acetate solution. Shots were taken at ×10 k at 80 kV.
## 2.3.3. Entrapment Efficiency (EE%)
The SF entrapment efficiency (EE%) was calculated based on the difference between the amounts of entrapped and unentrapped SF. SF-CS-NPs were separated by centrifugation for 30 min at 15000 rpm at 4°C. Unentrapped SF in the supernatant was determined spectrophotometrically at λmax 293 nm. EE% was calculated as follows: [1]EE%=Total SF mg−Unentrapped SFmgTotal SF mg.
## 2.4. Treatment of Rats with Sildenafil and Sildenafil-Chitosan Nanoparticles
The study protocol was approved by the Research Ethics Committee of the Medical Research Institute, Alexandria University, and complied with the Guide for the Care and Use of Laboratory Animals of the National Research Council (US), Institute for Laboratory Animal Research. Forty-eight male Wistar rats (average weight of 200 ± 20 g) were obtained from the animal house of the Faculty of Agriculture, Alexandria University. Rats were acclimated for 7 days before the experiment and were provided with a balanced commercial diet. The rats were randomly divided into four groups, 12 rats each. Treatments were administered by a daily intraperitoneal (i.p.) injection for 21 days as follows: Group 1 (control): 0.3 mL normal physiological saline Group 2: 1.5 mg/kg SF citrate solution in distilled water (DW) Group 3: SF-CS NPs (equivalent to 1.5 mg/kg SF)
Group 4: T-SF-CS NPs (equivalent to 1.5 mg/kg SF) At the end of the treatment period, rats were anesthetized, sacrificed, and their livers were isolated, washed with saline, and kept at -80°C for further biochemical analyses. Liver biopsies for histological examination were kept in $10\%$ formalin.
## 2.4.1. Assay of Antioxidant Enzymes Activity
The livers of rats were rinsed in cold 0.1 M potassium phosphate buffer (pH 7.4), blotted dry, weighed, and kept on ice. The liver homogenate ($33\%$) was prepared in 3 portions of 0.1 M phosphate buffer (pH 7.4) using a Teflon piston homogenizer on the ice at 4°C. The liver homogenates were then centrifuged for 20 min at 4°C at 11000 rpm to remove intact cell nuclei, mitochondria, and cell debris. The S9 fractions of the livers were stored at -80°C [28].
The GST activity was assayed according to the method of Habig et al. [ 29]. The calculations were performed using a molar extinction coefficient of 9.6 mM/cm. Under the assay conditions, a unit of enzyme activity was defined as the amount of enzyme that catalyzes the synthesis of 1 mM of CDNB conjugate/mg protein/min. Thiobarbituric acid-reactive substances (TBARS) were detected in the supernatant of S9 fractions [30]. The color intensity of the reactants (TBARS) was measured at 532 nm. An extinction coefficient of 156 000 M−1/cm was used in the calculation of the TBARS level. The glutathione levels in the supernatant of liver tissue homogenates were determined using sulfosalicylic acid for protein precipitation and bis-(3-carboxy-4-nitrophenyl)-disulfide for color development [31]. The color intensity at 412 nm was measured using a double-beam spectrophotometer. The activity of glutathione reductase (GR) was determined by monitoring NADPH oxidation at 340 nm in the supernatant of liver tissue homogenates [32]. GR activity was expressed as nmol NADPH oxidized/mg protein/min. The protein concentration was measured using bovine serum albumin as standard [33].
The activity of the SOD enzyme (EC 1.15.1.1) in S9 fractions was measured as reported [34]. The SOD assay was based on the inhibition of epinephrine autoxidation to adrenochrome in an alkaline medium, which is significantly reduced in the presence of SOD. The increase in adrenochrome absorbance was measured spectrophotometrically at 480 nm every 30 s for up to 4 min. The SOD enzyme activity was measured as the quantity of enzyme that prevents epinephrine from being oxidized by $50\%$, with each $50\%$ inhibition equaling one unit (1 U/g tissue). The catalase (CAT) (EC1.11.1.6) activity was measured by the method of Beers and Sizer, 1952. The assay is based on the spectrophotometric measurement of H2O2 decomposition at 240 nm. A known volume (2.5 mL) of H2O2 buffer (0.15 M sodium-potassium phosphate buffer pH 7.0) and 50 μL of S9 enzyme source were used in the experiment. The absorbance was determined spectrophotometrically at 240 nm after 20 and 40 s intervals against blank. The CAT enzyme activity was expressed as unit/mg protein. One unit of CAT is equal to one nmol H2O2/mg protein/min.
## 2.4.2. Western Blotting and Detection of Immobilized Proteins
Aliquots (100 μL) of the S9 fraction from each rat (10 rats per group) were pooled and used to examine the protein expression of GST and GPx. Each group's pooled microsomal proteins (40 μL) were combined with sample application buffer (SAB) and heated for 3 min before loading on a $10\%$ SDS-polyacrylamide gel. Proteins were transferred to nitrocellulose membranes using a semidry transblotter after electrophoresis. They were washed three times with TBS buffer pH 7.3 (8 g NaCl, 0.2 g KCl, and 3 g Tris-base/L) for 10 min after completing the transblotting of proteins on membranes. The membranes were then rinsed in Tris-HCL buffer saline (T-TBS) buffer containing $0.1\%$ Tween 20 for 5 min and then in TBS buffer twice for 10 min after being incubated with $5\%$ fat-free dry milk-TBS buffer for 1 h at room temperature. The membranes were then incubated for 2 h with primary antibodies for anti-GST, and anti-GPx at a dilution of 1: 1000 before being washed twice with Tween 80-TBS (0.2 ml Tween 20/L TBS) for 20 min and TBS for 15 min. After incubation with anti-mouse horseradish peroxidase-conjugated secondary antibody at a dilution of 1: 7000 in TBS, the membranes were washed twice with Tween 80-TBS for 15 min and then twice with TBS for 15 min. The protein expression of various isozymes was identified using an ECL kit and X-ray film. The intensity of the bands was determined using the Quantity One Software Program (version 4.6.9, Bio-Rad Co., California, USA).
## 2.4.3. Histopathological Examination
Small sections of both liver tissues from each rat in each treatment were preserved in a $10\%$ formaldehyde solution, embedded in paraffin wax, and sectioned with a microtome into 3 μm-thick sections which were stained with Hematoxylin and Eosin (H&E) and examined by light microscopy (Olympus BX 50, Japan) to identify histopathological changes [35].
## 2.5. Statistical Analysis
The results were presented as means ± SE. A one-way analysis of variance was used to calculate the differences between groups (ANOVA) using the SPSS Statistics Program version 20. Differences between groups were considered significant at $P \leq 0.05.$
## 3.1. Characteristics of Chitosan-Based Nanoparticles
Chitosan-based NPs prepared using a 2 mg/mL concentration of CS and TPP, a CS: TPP ratio of 3: 1, and probe sonication on ice for 10 min displayed desired physical properties. The physical properties of blank CS NPs, SF-CS NPs, and T-SF-CS NPs are shown in Figure 1. Blank CS NPs had a mean size of 124.2 ± 20.2 nm, a mean polydispersity index (PDI) of 0.221 ± 0.05, and a mean zeta potential of 18.0 ± 2.0 mV. SF-loaded CS NPs (SF-CS NPs) showed a significantly ($P \leq 0.0001$) larger mean size (178 ± 12.1 nm) with no significant changes in PDI and ZP. Tween 80-coated SF-CS NPs (T- SF-CS NPs) exhibited a further significant increase in size (215.3 ± 10.5 nm) relative to SF-Cs NPs ($P \leq 0.0001$) in addition to a small increase in PDI and a reduction in ZP.
TEM imaging of the test CS-based NPs (Figure 2) indicated that blank CS NPs (Figure 2(a)) were uniform nonaggregated nanospheres with a size ranging from 43.2 to 48.2 nm. SF-CS NPs were also spherical but slightly larger (Figure 2(b)). Tween 80 surface coating led to a further increase in the size of SF-CS NPs which appeared surrounded by a clear zone (Figure 2(c)). The mean EE of SF was 27.63 ± $1.25\%$.
## 3.2. Hepatic Biochemical Effects of Chitosan-Based Nanoparticles
As shown in Table 1, SF administration to rats drastically reduced the levels of GSH and TBARs. Compared with the SF-treated group, SF-CS NPs and T-SF-CS NPs treatments considerably boosted GSH levels, though normal levels were not recovered (Table 1). Similarly, both SF-CS NPs attenuated the inhibitory effects of SF on the TBARS level which was not significantly different from the control level following treatment with T-SF-CS NPs. Western immunoblotting data (Figure 3) demonstrated that the protein expression of GST was downregulated which may account for the effects of SF, SF-CS-NPs, and T-SF-CS-NPs treatments on GST activity (Figure 3).
The effect of SF, SF-CS NPs, and T-SF-CS NPs treatments on the activity of the antioxidant enzymes glutathione reductase (GR), catalase (CAT), glutathione peroxidase (GPx), and superoxide dismutase (SOD) is shown in Table 1. In contrast to a significant increase in GPx activity, free SF therapy significantly reduced the activity of GR, CAT, and SOD. Western immunoblotting findings showing an increase in GPx protein expression after SF-CS NPs and T-SF-CS NPs treatments verified the results of GPx activity (Figure 4). It is interesting to note that CAT, GR, GPx, and SOD activities were dramatically increased by SF-CS NPs and T-SF-CS NPs treatments relative to either SF and/or the control groups (Table 1).
## 3.3. Histopathological Examination
Histopathological examination (Figures 5(a)–5(g)) demonstrated changes in the architecture of liver tissues after treatment with SF, SF-CS NPs, and T-SF-CS NPs compared with the control liver. Liver sections from SF-treated rats (Figures 5(a)–5(d)) showed several hepatotoxic effects including necrosis of hepatic cells with loss of nuclei and fibrosis (Figure 5(a)), regeneration of hepatic cells with oval cell hyperplasia (long arrows) and binucleated cells (short arrows) (Figure 5(b)), hemolysis of blood in the portal veins (long arrows) and inflammatory cells with fibrosis in the portal areas (short arrows) (Figure 5(c)) in addition to dilation of central veins that were engorged with blood (Figure 5(d)). Chitosan nanoencapsulation of SF led to different hepatic changes including ballooning degeneration of hepatocytes with white fluffy cytoplasm surrounding the central nucleus and accumulation of inflammatory cells around the central vein and portal area (Figure 5(e)). Such changes were markedly suppressed by T-SF-CS NPs treatment (Figure 5(f)) which protected the normal liver architecture and hepatic cells. The histological characteristics of liver sections of the T-SF-CS NPs-treated group were close to those of the normal control section (Figure 5(g)).
## 4.1. Chitosan-Based Nanoparticles
The CS NPs under study (Blank CS NPs, SF-CS NPs, and T-SF-CS NPs) exhibited accepted mean size (range 124 to 215 nm) and mean polydispersity index (PDI range 0.20–0.25) indicative of nanosize and narrow particle size distribution [36]. The test NPs displayed a positive surface charge as indicated by a zeta potential range of 15.4-20 mV. This is of importance to maintaining the colloidal stability of the NPs. The relatively high ZP of CS NPs is attributed to the protonation of CS amino groups [37]. The mean size of blank CS NPs was increased by SF loading, an observation consistent with literature reports [38, 39]. The further significant increase in mean SF-CS NPs size by Tween 80 coating of SF-CS NPs can be ascribed to adherence of the Tween 80 surfactant molecules to the hydrophilic surface of CS NPs, probably increasing their hydrodynamic radii. Tween 80 coating also slightly decreased the ZP of NPs, as a result of the partial concealing of the surface positive charge of CS NPs CS NPs [40].
TEM imaging provided evidence for the uniformity of shape and size of the three test CS NPs and revealed a clear zone around the T-SF-CS NPs. Despite differences in absolute values, the particle size of the three nanoformulations shown by TEM was in agreement with that determined by DLS. The smaller size of NPs generally observed in TEM images can be explained by the dehydration of NPs during sample preparation [41]. The colloidal properties and morphological traits of the three CS NP formulations were generally appropriate for drug delivery applications [22].
## 4.2. Effect of Chitosan-Based Nanoparticles on the Activity of Antioxidant Enzymes
Glutathione (GSH) and glutathione S-transferase enzyme (GST) play a major role in drug conjugation and detoxification, affecting the efficacy of several chemotherapeutic and alkylating drugs [42]. Accordingly, any organ with low GSH levels and inhibited GST activity is more sensitive to alkylating agents, whereas those with high GSH levels and induced GST activity are more protected and resistant to these agents [43–45]. In the present study, GSH levels and GST activity were significantly decreased after the treatment of rats with SF. Although SF-CS NPs and T-SF CS NPs alleviated the inhibitory effect of SF, normal levels of GSH levels and GST activity were not restored. Data from western immunoblotting confirmed the results of GST activity by demonstrating that all treatments decreased the protein expression of GST in comparison to the control group. This might imply the contribution of CS to the SF-induced suppression of GST protein expression. In support of this assumption, it has been reported that rats fed chitosan also showed lower liver activity of GST [46]. As reported previously, SF therapy reduced GSH levels and inhibited GST activity in the liver tissues of rats [11] and human red blood cells [47]. Possible conversion of reduced glutathione (GSH) into its oxidized form as a result of SF, SF-CS NPs, and T-SF-CS NPs treatments may account for the decline in GSH levels. As oxidized glutathione may be reverted into its reduced form by the enzyme GR, the reduction of GR by SF treatment as reported earlier [11] might account for the findings. While SF-CS-NPs were unable to reverse the SF-induced suppression of GR activity, T-SF-CS NPs restored the activity to its control level.
It has been shown that when GST activity is suppressed and GSH levels are low, the epoxides of well-known chemical carcinogens bind to DNA and other macromolecules more covalently [43, 44, 48] which enhances hepatocarcinogenesis. This entails that SF therapy may increase hepatotoxicity as a result of decreasing GSH levels and GST activity. Interestingly, T-SF CS NPs enhanced both GSH levels and GST activity, suggesting possible liver protection from the hazardous effects of chemical substances produced endogenously or upon exposure to exogenous sources.
Reactive oxygen species (ROS) have been implicated in the induction of oxidative stress [49]. According to data obtained in the present study, the levels of TBARS were significantly lower in the groups treated with free SF, SF CS-NPs, and T-SF CS NPs compared with the control group. Similarly, earlier research demonstrated that SF treatment reduced free radical levels, resulting in subsequent inhibition of lipid peroxidation [50]. In addition, GPx and CAT enzymes are involved in the termination reaction of the ROS pathway. The data of the present study revealed that there was a significant increase in GPx activity and upregulation of its protein expression after treatment of rats with SF CS-NPs and T-SF CS NPs relative to the control group. CAT seems to be the main regulator of hydrogen peroxide metabolism [51]. SF-CS-NPs- and T-SF-CS NPs-treated groups significantly increased CAT and SOD activity in comparison with the control group. As a result, stimulation of GPx and CAT which are acting as radical ions quenchers could explain the mechanism of reduction of free radical levels in SF CS-NPs- and T-SF CS NPs-treated groups.
## 4.3. Histopathological Examination
The results of histopathological examination revealed multiple SF-induced hepatotoxic effects including necrosis of hepatic cells with loss of nuclei and fibrosis; regeneration of hepatic cells with oval cell hyperplasia and binucleated cells; hemolysis of blood in the portal veins and inflammatory cells with fibrosis in the portal areas in addition to dilation of central veins which appeared engorged with blood. These findings raise concerns regarding the safety of SF treatment of different diseases and urge the need for strategies to reduce the adverse effects of SF on the liver. As demonstrated in the present study, nanoencapsulation of SF might influence its hepatotoxic profile. Encapsulation of SF by CS NPs modified the SF-induced histopathological changes which were manifested as an accumulation of inflammatory cells in the central vein and portal area with degeneration of hepatocytes. Likewise, CS NPs with a size of 200 nm were reported to induce severe defects in zebrafish embryos, including a twisted spine, pericardial edema, and an opaque yolk, supporting our findings [52]. In addition, embryos exposed to CS NPs had a higher rate of cell death, higher levels of ROS, and overexpression of heat shock protein 70, verifying the induction of physiological stress in zebrafish by CS NPs [52]. Intriguingly, the Tween 80 coating of SF-CS NPs significantly protected the liver architecture and hepatic cells from the harmful effects caused by SF and uncoated CS NPs as verified by the close-to-normal architecture of liver sections from T-SF-CS NPs-treated rats. In comparison with liver sections of the SF-CS NPs-treated group, sections of T-SF-CS NPs-treated and untreated control groups showed a network of hepatic strands made up of almost normal hepatocytes structures. As such, T-SF-CS NPs were effective in preventing the deleterious hepatic histopathological changes induced by SF and CS NPs.
## 5. Conclusion
A nanotechnological approach was explored to suppress the hepatic adverse effects and improve the safety of sildenafil. The drug was encapsulated in uncoated and Tween 80-coated chitosan nanoparticles characterized for colloidal properties and drug entrapment efficiency. The adverse hepatic effects of sildenafil manifested as suppression of antioxidant enzymes activity and multiple deleterious changes in liver tissues of rats following a 21-day-intraperitoneal treatment, were attenuated particularly by the Tween 80-coated nanoparticles. Findings are of significance regarding the improvement of the safety of sildenafil therapy, a drug currently under active investigation for safer treatment of erectile dysfunction, pulmonary arterial hypertension, and an expanding list of disease conditions.
## Data Availability
All available data are included in the MS.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## References
1. Deger M. D., Madendere S.. **Erectile dysfunction treatment with phosphodiesterase-5 inhibitors: Google trends analysis of last 10 years and COVID-19 pandemic**. (2021) **93** 361-365. DOI: 10.4081/aiua.2021.3.361
2. Ruopp N. F., Cockrill B. A.. **Diagnosis and treatment of pulmonary arterial hypertension**. (2022) **327** 1379-1391. DOI: 10.1001/jama.2022.4402
3. Turko I. V., Ballard S. A., Francis S. H., Corbin J. D.. **Inhibition of cyclic GMP-binding cyclic GMP-specific phosphodiesterase (type 5) by sildenafil and related compounds**. (1999) **56** 124-130. DOI: 10.1124/mol.56.1.124
4. Xiong Y., Wintermark P.. **The role of sildenafil in treating brain injuries in adults and neonates**. (2022) **16, article 879649**. DOI: 10.3389/fncel.2022.879649
5. Ibrahim M. A., Haleem M., AbdelWahab S. A., Abdel-Aziz A. M.. **Sildenafil ameliorates Alzheimer disease via the modulation of vascular endothelial growth factor and vascular cell adhesion molecule-1 in rats**. (2021) **40** 596-607. DOI: 10.1177/0960327120960775
6. Zhu G., Ueda K., Hashimoto M.. **The mitochondrial regulator PGC1**. (2022) **596** 17-28. DOI: 10.1002/1873-3468.14228
7. Cruz-Burgos M., Losada-Garcia A., Cruz-Hernández C. D.. **New approaches in oncology for repositioning drugs: the case of PDE5 inhibitor sildenafil**. (2021) **26, article 627229** p. 11. DOI: 10.3389/fonc.2021.627229
8. Hong J. H., Kwon Y. S., Kim I. Y.. **Pharmacodynamics, pharmacokinetics and clinical efficacy of phosphodiesterase-5 inhibitors**. (2017) **13** 183-192. DOI: 10.1080/17425255.2017.1244265
9. Tang P. F., Zheng X., Hu X. X.. **Functional measurement of CYP2C9 and CYP3A4 allelic polymorphism on sildenafil metabolism**. (2020) **14** 5129-5141. DOI: 10.2147/DDDT.S268796
10. Shim H. J., Kim Y. C., Park K. J.. **Pharmacokinetics of DA‐8159, a new erectogenic, after intravenous and oral administration to rats: hepatic and intestinal first‐pass effects**. (2003) **92** 2185-2195. DOI: 10.1002/jps.10482
11. Sheweita S., Salama B., Hassan M.. **Erectile dysfunction drugs and oxidative stress in the liver of male rats**. (2015) **2** 933-938. DOI: 10.1016/j.toxrep.2015.06.002
12. Sheweita S. A., Meftah A. A., Sheweita M. S., Balbaa M. E.. **Erectile dysfunction drugs altered the activities of antioxidant enzymes, oxidative stress and the protein expressions of some cytochrome P450 isozymes involved in the steroidogenesis of steroid hormones**. (2020) **15**. DOI: 10.1371/journal.pone.0241509
13. Sheweita S. A., Wally M., Hassan M.. **Erectile dysfunction drugs changed the protein expressions and activities of drug-Metabolising enzymes in the liver of male rats**. (2016) **2016** 9. DOI: 10.1155/2016/4970906
14. Chakraborty S., Roychoudhury S.. **Pathological roles of reactive oxygen species in male reproduction**. (2022) **1358** 41-62. DOI: 10.1007/978-3-030-89340-8_3
15. Taysi S., Algburi F. S., Mohammed Z. R., Ali O. A., Taysi M. E.. **Thymoquinone: a review on its pharmacological importance, and its association with oxidative stress, COVID-19, and radiotherapy**. (2022) **22** 1847-1875. DOI: 10.2174/1389557522666220104151225
16. Taysi S., Tascan A. S., Ugur M. G., Demir M.. **Radicals, oxidative/nitrosative stress and preeclampsia**. (2019) **19** 178-193. DOI: 10.2174/1389557518666181015151350
17. Loprete L., Leuratti C., Frangione V., Radicioni M.. **Pharmacokinetics of a novel sildenafil Orodispersible film administered by the supralingual and the sublingual route to healthy men**. (2018) **38** 765-772. DOI: 10.1007/s40261-018-0665-x
18. Atipairin A., Chunhachaichana C., Nakpheng T., Changsan N., Srichana T., Sawatdee S.. **Development of a sildenafil citrate microemulsion-loaded hydrogel as a potential system for drug delivery to the penis and its cellular metabolic mechanism**. (2020) **12** p. 1055. DOI: 10.3390/pharmaceutics12111055
19. Güven E.. **Lipid-based nanoparticles in the treatment of erectile dysfunction**. (2020) **32** 578-586. DOI: 10.1038/s41443-020-0235-7
20. Hosny K. M., Aljaeid B. M.. **Sildenafil citrate as oral solid lipid nanoparticles: a novel formula with higher bioavailability and sustained action for treatment of erectile dysfunction**. (2014) **11** 1015-1022. DOI: 10.1517/17425247.2014.912212
21. Kou S., Peters L., Mucalo M.. **Chitosan: a review of molecular structure, bioactivities and interactions with the human body and micro-organisms**. (2022) **282, article 119132**. DOI: 10.1016/j.carbpol.2022.119132
22. Aibani N., Rai R., Patel P., Cuddihy G., Wasan E. K.. **Chitosan nanoparticles at the biological Interface: implications for drug delivery**. (2021) **13** p. 1686. DOI: 10.3390/pharmaceutics13101686
23. Ashoush Y., El-Sayed S., Abd-Elwahab M.. **Hepatoprotective effects of chitosan and chitosan nanoparticles**. (2022) **7** p. 27. DOI: 10.21608/mjab.2022.263921
24. Dawoud S. F., Al-Akra T. M., Zedan A. M.. **Hepatoprotective effects of chitosan and chitosan nanoparticles against biochemical, genetic, and histological disorders induced by the toxicity of emamectin benzoate**. (2021) **10** 506-517. DOI: 10.52547/rbmb.10.3.506
25. Fan W., Yan W., Xu Z., Ni H.. **Formation mechanism of monodisperse, low molecular weight chitosan nanoparticles by ionic gelation technique**. (2012) **1** 21-27. DOI: 10.1016/j.colsurfb.2011.09.042
26. Soudi S. A., Nounou M. I., Sheweita S. A., Ghareeb D. A., Younis L. K., el-Khordagui L. K.. **Protective effect of surface-modified berberine nanoparticles against LPS-induced neurodegenerative changes: a preclinical study**. (2019) **9** 906-919. DOI: 10.1007/s13346-019-00626-1
27. Widyastama G., Kurniati M.. **Optimization of sonication time and surfactant concentration for chitosan-alginate coated Ketoprofen nanoencapsulation**. (2021) **1805** p. 012003. DOI: 10.1088/1742-6596/1805/1/012003
28. Sheweita S. A.. **Glutathione alleviates the influence of N-nitrosamines on the activity of carcinogen-metabolizing enzymes in the liver of male mice**. (2006) **1** 72-81. DOI: 10.1016/S1658-3612(06)70010-5
29. Habig W. H., Pabst M. J., Fleischner G., Gatmaitan Z., Arias I. M., Jakoby W. B.. **The identity of glutathione S-transferase B with ligandin, a major binding protein of liver**. (1974) **71** 3879-3882. DOI: 10.1073/pnas.71.10.3879
30. Tappel A., Zalkin H.. **Inhibition of lipide peroxidation in mitochondria by vitamin E**. (1959) **80** 333-336. DOI: 10.1016/0003-9861(59)90259-0
31. Mitchell J. R., Thorgeirsson S. S., Potter W. Z., Jollow D. J., Keiser H.. **Acetaminophen-induced hepatic injury: protective role of glutathione in man and rationale for therapy**. (1974) **16** 676-684. DOI: 10.1002/cpt1974164676
32. Suojanen J. N., Gay R. J., Hilf R.. **Influence of estrogen on glutathione levels and glutathione-metabolizing enzymes in uteri and R3230AC mammary tumors of rats**. (1980) **630** 485-496. DOI: 10.1016/0304-4165(80)90003-3
33. Lowry O., Rosebrough N., Farr L., Randall R. J.. **Protein measurement with the Folin phenol reagent**. (1951) **193** 265-275. DOI: 10.1016/S0021-9258(19)52451-6
34. Misra H. P., Fridovich I.. **The role of superoxide anion in the autoxidation of epinephrine and a simple assay for superoxide dismutase**. (1972) **247** p. 3170. DOI: 10.1016/S0021-9258(19)45228-9
35. Drury R. A. B., Wallington E. A., Cameron S. R.. (1967)
36. Schärtl W.. (2007)
37. Martien R., Loretz B., Sandbichler A. M., Schnürch A. B.. **Thiolated chitosan nanoparticles: transfection study in the Caco-2 differentiated cell culture**. (2008) **19**. DOI: 10.1088/0957-4484/19/04/045101
38. Lazaridou M., Christodoulou E., Nerantzaki M.. **Formulation and in-vitro characterization of chitosan-nanoparticles loaded with the iron chelator deferoxamine mesylate (DFO)**. (2020) **12** p. 238. DOI: 10.3390/pharmaceutics12030238
39. Janes K. A., Fresneau M. P., Marazuela A., Fabra A., Alonso M. J.. **Chitosan nanoparticles as delivery systems for doxorubicin**. (2001) **73** 255-267. DOI: 10.1016/s0168-3659(01)00294-2
40. Tao X., Li Y., Hu Q.. **Preparation and drug release study of novel Nanopharmaceuticals with polysorbate 80 surface adsorption**. (2018) **2018** 11. DOI: 10.1155/2018/4718045
41. Nagpal K., Singh S. K., Mishra D. N.. **Optimization of brain targeted chitosan nanoparticles of rivastigmine for improved efficacy and safety**. (2013) **59** 72-83. DOI: 10.1016/j.ijbiomac.2013.04.024
42. Hamilton T. C., Winker M. A., Louie K. G.. **Augmentation of adriamycin, melphalan, and cisplatin cytotoxicity in drug- resistant and -sensitive human ovarian carcinoma cell lines by buthionine sulfoximine mediated glutathione depletion**. (1985) **34** 2583-2586. DOI: 10.1016/0006-2952(85)90551-9
43. Ketterer B.. **Protective role of glutathione and glutathione transferases in mutagenesis and carcinogenesis**. (1988) **202** 343-361. DOI: 10.1016/0027-5107(88)90197-2
44. Malaveille C., Brun G., Hautefeuille A., Bartsch H.. **Effect of glutathione and urifine 5′-diphosphoglucuronic acid on mutagenesis by benzo[**. (1981) **83** 15-24. DOI: 10.1016/0027-5107(81)90067-1
45. Gopalan P., Jensen D. E., Lotlikar P. D.. **Glutathione conjugation of microsome-mediated and synthetic aflatoxin B**. (1992) **64** 225-233. DOI: 10.1016/0304-3835(92)90047-Y
46. Yao H. T., Lii C. K., Chou R. H., Lin J. H., Yang H. T., Chiang M. T.. **Effect of chitosan on hepatic drug-metabolizing enzymes and oxidative stress in rats fed low- and high-fat diets**. (2010) **58** 5187-5193. DOI: 10.1021/jf903857m
47. Tagliabue M., Pinach S., Di Bisceglie C.. **Glutathione levels in patients with erectile dysfunction, with or without diabetes mellitus**. (2005) **28** 156-162. DOI: 10.1111/j.1365-2605.2005.00528.x
48. Gopalan-Kriczky P., Hiruma S., Lotlikar P. D.. **Effect of glutathione levels on aflatoxin B**. (1994) **76** 25-30. DOI: 10.1016/0304-3835(94)90130-9
49. Li N., Xia T., Nel A. E.. **The role of oxidative stress in ambient particulate matter-induced lung diseases and its implications in the toxicity of engineered nanoparticles**. (2008) **44** 1689-1699. DOI: 10.1016/j.freeradbiomed.2008.01.028
50. Sheweita S. A., Tilmisany A. M., Al-Sawaf H.. **Mechanisms of male infertility: role of antioxidants**. (2005) **6** 495-501. DOI: 10.2174/138920005774330594
51. Góth L., Rass P., Páy A.. **Catalase enzyme mutations and their association with diseases**. (2004) **8** 141-149. DOI: 10.1007/BF03260057
52. Hu Y. L., Qi W., Han F., Shao J. Z., Gao J. Q.. **Toxicity evaluation of biodegradable chitosan nanoparticles using a zebrafish embryo model**. (2011) **6** 3351-3359. DOI: 10.2147/IJN.S25853
|
---
title: 'A Pilot Study of Plantar Mechanics Distributions and Fatigue Profiles after
Running on a Treadmill: Using a Support Vector Machine Algorithm'
authors:
- Qian Liu
- Hairong Chen
- Anand Thirupathi
- Meimei Yang
- Julien S. Baker
- Yaodong Gu
journal: Journal of Healthcare Engineering
year: 2023
pmcid: PMC9988392
doi: 10.1155/2023/7461729
license: CC BY 4.0
---
# A Pilot Study of Plantar Mechanics Distributions and Fatigue Profiles after Running on a Treadmill: Using a Support Vector Machine Algorithm
## Abstract
The treadmill is widely used in running fatigue experiments, and the variation of plantar mechanical parameters caused by fatigue and gender, as well as the prediction of fatigue curves by a machine learning algorithm, play an important role in providing different training programs. This experiment aimed to compare changes in peak pressure (PP), peak force (PF), plantar impulse (PI), and gender differences of novice runners after they were fatigued by running. A support vector machine (SVM) was used to predict the fatigue curve according to the changes in PP, PF, and PI before and after fatigue. 15 healthy males and 15 healthy females completed two runs at a speed of 3.3 m/s ± $5\%$ on a footscan pressure plate before and after fatigue. After fatigue, PP, PF, and PI decreased at hallux (T1) and second-fifth toes (T2–5), while heel medial (HM) and heel lateral (HL) increased. In addition, PP and PI also increased at the first metatarsal (M1). PP, PF, and PI at T1 and T2–5 were significantly higher in females than in males, and metatarsal 3–5 (M3–5) were significantly lower in females than in males. The SVM classification algorithm results showed the accuracy was above average level using the T1 PP/HL PF (train accuracy: $65\%$; test accuracy: $75\%$), T1 PF/HL PF (train accuracy: $67.5\%$; test accuracy: $65\%$), and HL PF/T1 PI (train accuracy: $67.5\%$; test accuracy: $70\%$). These values could provide information about running and gender-related injuries, such as metatarsal stress fractures and hallux valgus. Application of the SVM to the identification of plantar mechanical features before and after fatigue. The features of the plantar zones after fatigue can be identified and the learned algorithm of plantar zone combinations with above-average accuracy (T1 PP/HL PF, T1 PF/HL PF, and HL PF/T1 PI) can be used to predict running fatigue and supervise training. It provided an important idea for the detection of fatigue after running.
## 1. Introduction
The most serious threat to health in modern times has been identified as sedentary behavior with insufficient physical activity [1]. Running has long been a popular leisure activity. Athletes have much lower resting heart rate, body weight, body mass index (BMI), and triglyceride levels compared to the general population [2], indicating that regular physical exercise can minimize the risks of cardiovascular disease. At the same time, running carries a considerable risk of injury. In follow-up cases in the population, the incidence of running-related injury was reported to be 2.5 to 33.0 cases per 1000 h [3]. However, the causes of injuries are varied. Most running-related lower limb injuries, for example, are the result of avoidable training errors [4, 5]. In addition, accumulating long and strong training may lead to an increase in shin pain [6].
Muscle tiredness is a complicated physiological state induced not only by changes in muscle capacity but also by the central nervous system's inability to appropriately drive motor neurons [7]. Long-term running has been proven to cause central fatigue, which diminishes the strength of the maximal autonomic plantar flexor muscle. Plantar flexor fatigue can limit the power of these muscles during the propulsion phase of running, and lower limb strength can be lowered by 30 to $40\%$ after running [8, 9]. The biomechanical features of the lower limbs change as a result of exhaustion, which is crucial in preventing sports injuries. Changes in knee angle and moment because of fatigue, for example, can be used to predict anterior cruciate ligament injuries [10].
Several measurement approaches have been utilized in many studies to quantify the association between foot dynamics and lower extremity overuse injuries. Plantar mechanical measurement has been frequently utilized to evaluate overall running performance as a result of this [11, 12]. The second and third metatarsals exhibit a $10\%$ increase in peak pressure immediately after fatigue, and an $11\%$ increase after 30 mins, with a significant $12\%$ drop in load at the first toe [13]. It is worth noting that increased load under the metatarsal bone can produce biomechanical imbalance, which could lead to metatarsalgia [14]. Furthermore, the increasing plantar load will promote stretching stresses on the plantar aponeurosis, which leads to microtraumas and degradation of connective tissues, promoting the development of plantar fasciitis [15, 16]. In conclusion, there is an urgent need to reflect on and evaluate fatigue and fatigue injuries through changes in plantar mechanical parameters. Insole technologies for activity classification couple plantar pressure with accelerometer data, increasing technology cost, and complexity [17, 18]. The advantage of the platform is that it is easy to use because it is stationary and flat and can be well applied to the laboratory environment [19]. Therefore, we used the footscan force platform to detect the mechanical characteristics of the plantar. Treadmills have been widely used in laboratory studies to easily control speed gradients. Previous studies have also shown that treadmill running is different from running on the ground. Whether treadmill running can simulate running on the ground is still a controversial issue [20]. This experiment only examined the change form of plantar mechanical parameters after fatigue running on a treadmill.
Males and females have different bone structures and muscle strength, and studies have shown that females are more likely than males to sustain lower limb injuries while running [21, 22]. Females are more prone than males to have ligamentous laxity of the ankle joint, and females are approximately twice as likely as males to have ankle sprains [23]. Plantar mechanical parameter distributions are affected by several factors, including weight, gender, foot structure, and even how a person stands or walks [24]. Experts in forensic science use variations in foot bones to determine gender [25]. There are, however, no consistent results on the gender differences in plantar pressure characteristics. According to research [26], there are no significant variations in the midfoot contact area and plantar pressure between males and females. The pressure under the toes was higher in female adolescents than in male adolescents, while the pressure was higher in male adolescents only at the hindfoot, and the pressure at the metatarsophalangeal toe increased more significantly in females [27]. The difference in plantar mechanical parameters caused by gender can reflect a lot of practical problems. Therefore, it is necessary to explore the effect of gender differences on plantar mechanical parameters.
In biomechanical research, traditional statistical methodologies have limited the ability to classify groups based on many variables [28]. In recent years, a support vector machine (SVM) has emerged as a new tool for solving biological classification problems [29]. By creating discriminatory parameters to separate groups from one another, the SVM attempts to discover a hyperplane that maximizes the distance between groups [30]. The SVM has the advantage of producing classification results based on limited data sets while minimizing structural and empirical risk [31]. Injuries are common in individual sports and will cause serious physical outcomes. Reduced exercise capacity because of fatigue increases the incidence of musculoskeletal injuries [32]. As a result, forecasting the occurrence of sports injuries is critical to maintaining good health [33]. Previous research [34] used the SVM to predict diabetic foot ulceration based on plantar mechanical parameters. Aguirre et al. [ 35] proposed a computational model for predicting tiredness during exercise from a sitting to a standing posture, which could be useful for rehabilitation. Si et al. [ 36] employed the SVM and fractal analysis for gait recognition and test the identification performance, and the testing outcomes indicate an overall accuracy of $93.57\%$ via radial basis function kernel. Jeong et al. [ 37] used the SVM to classify activity patterns based on plantar pressure characteristics, and the recognition rate reached $95.2\%$. Stetter et al. [ 38] used the SVM and identified the kinematic difference between fatigue and nonfatigue based on principal component analysis, the strides of fatigue and nonfatigue were separated, and the classification accuracy was $99.4\%$. Wang et al. [ 39] used inertial measurement unit (IMU) and SVM to distinguish fatigue and nonfatigue running states, and predict the degree of fatigue. The classification accuracy of tibia and thigh IMUs was $91.10\%$. The characteristics of plantar pressure were evaluated using leave-one-out cross-validation with machine learning algorithms: SVM, decision tree, discriminant analysis, and k-nearest neighbors in the study of Merry et al. [ 17]. The results showed that the SVM and decision tree have higher classification accuracy. In addition, other studies have shown that the SVM has the best performance in distinguishing gait characteristics [40]. Therefore, the SVM was used to predict fatigue in this study. In addition, many researchers have applied SVM to the recognition of gait characteristics before and after fatigue, but few studies have paid attention to the plantar mechanical characteristics before and after fatigue.
As a consequence, this research aimed to explore the differences in peak pressure (PP), peak force (PF), and plantar impulse (PI) before and after long-distance running fatigue in novice runners, as well as gender differences. We also employed the SVM algorithm to predict fatigue based on plantar mechanical parameters. Based on previous studies, we assumed that the change in plantar mechanical parameters before and after fatigue mainly occurred in the toes. It was also assumed that gender differences in plantar mechanical parameters were mainly concentrated in the toes and metatarsal regions. In addition, it was assumed that the SVM can predict fatigue at a high level.
## 2.1. Participants
The experimental subjects for this investigation were 15 healthy males and 15 healthy females [13, 25] who were novice runners (Table 1) with dominant right legs. Participants were recruited from sports clubs at Ningbo University and via social media. There were no health issues, neuromuscular abnormalities, or recognized gait difficulties in any of the participants, and no lower limb injuries in six months before the experiment. High arches and flat feet were not allowed to participate in the recruitment process. All subjects were given and signed written consent granted by the Institutional Review Board before the experiment (RAGH20210922205.6).
## 2.2. Experimental Procedures
Figure 1 depicts the experimental procedure. All of the participants did fatigue-inducing running workouts. The 15-point Borg scale and heart rate monitor (Polar RS100, Polar Electro Oy, Woodbury, NY, USA) were used to record perceived exertion, and heart rate changes per minute during the fatigue intervention. The individuals began the experiment by running at a speed of 1.67 m/s on a treadmill (h/p/cosmos para graphicsR, Germany). During the experiment, the slope was maintained at $1\%$ [41–43]. After which the speed was increased by 0.28 m/s every 2 minutes until the subjects reached a Borg intensity of 13. The subjects then continued at this speed until they reached Borg scale 17 or $90\%$ of maximal heart rate (HRmax calculated at 220-age), at which point they slowly reduced the speed to a speed of their choice [44, 45]. Space constraints, repeatability, and better control of climate, speed, and slope were the reasons why treadmill running was selected by our research team [46].
In this experiment, a footscan pressure plate (Footscan® software 7.0 Gait 2nd Generation, RsScan International) was used to monitor dynamic plantar pressure. The footscan pressure plate was 2 m in length and the acquisition frequency was 126 Hz. Subjects were asked to perform a pressure measurement on the footscan pressure plate before and immediately after fatigue. To avoid injury during the test, the subjects familiarized themselves with the footscan pressure plate before the trial. After familiarity, the subjects were asked to run on the footscan pressure plate at a speed of 3.3 m/s ± $5\%$ [44]. To manage running speed, Brower timing lights (Brower Timing System, Draper, UT, USA) were used. The subjects who completed a full gait cycle on the footscan pressure plate at the specified speed were regarded as successful. 5 groups of valid data were collected from each subject before and after the fatigue intervention. In addition, during the fatigue intervention, we uniformly provided clothes and shoes to the subjects to avoid experimental differences and maintain consistency.
## 2.3. Data Analysis
We analyzed plantar mechanical parameters in the running stance phase. For each trial, ten anatomical zones were automatically identified by the software (Footscan® software 7.0 Gait 2nd Generation, RsScan International) and if necessary, manually corrected by adjusting the pixels per zone (Figure 2): hallux (T1), second-fifth toes (T2–5), metatarsal 1–5 (M1, M2, M3, M4, M5), midfoot (MF), heel medial (HM), and heel lateral (HL). During the adjustment, we performed strict controls to ensure that the adjustment conditions and adjustment levels were rigorous and careful. The average values of PP, PF, and PI for all ten regions were calculated.
## 2.4. Statistical Analysis and SVM Classification Algorithm
The calculated data were exported to a statistical software package SPSS 26.0 (SPSS, Chicago, IL, USA), and the peak pressure, peak force, and plantar impulse of each plantar zone before and after running were statistically processed. The data were initially assessed for normality using a Kolmogorov–Smirnov test. The data were normally distributed. To investigate the effects of fatigue, gender, and their interaction on the plantar mechanical parameters, a two-way analysis of covariance (ANCOVA) was conducted. The significance level was set as $P \leq 0.05.$
When the data sets were not easily separable, the SVM classifier is a supervised machine learning technique that translates the input data space to a higher dimensional space to obtain a more accurate classification [35]. In our study, we used the LIBSVM toolbox based on MATLAB 2016b (Mathworks, MA, USA). The linear kernel was used for the SVM in the study. The cross-validation technique we employed was the hold-out method. $66.7\%$ of the sample size was randomly selected as the training set, and $33.3\%$ of the sample size was used as the test set [41].
The SVM is suitable for small and medium data samples, and nonlinear, high-dimensional classification problems. It maps the feature vector of the instance to some points in space. The purpose of the SVM is to find a line that best distinguishes two types of points, and when new points are added later, this line can also make a good classification. The SVM will find the partitioning hyperplane that distinguishes the two classes and maximizes the separation. For any hyperplane, the data points on both sides have a minimum distance (vertical distance) from it, and the sum of these two minimum distances is the interval.
For this partitioned hyperplane, we can give the following equation:[1]ωTX+$b = 0$,where ω is the weight of each feature and the column vector. b is the displacement value.
The distance from the point xi to the surface is as follows:[2]ωTxi+bω. Then,[3]ωTxi+bω×yi≥d,yi is the predicted value of sample i (−1 or 1, doing sign transformation). d is the distance of the support vector to the hyperplane. We assume that d is (2/‖ω‖).
Making all the points meet:[4]yiωTxi+b≥1.
The hyperplane we need is the one that needs to maximize the minimum interval, i.e.,[5]argmaxω,b1ωminiyiωT∙xi+b. Then, we need to calculate[6]argmaxω,b1ω.
Equivalent to calculate[7]minω,b12ω2,and[8]yiωTxi+b≥1.
Using the Lagrange multiplier method:[9]=12ω2−∑$i = 1$naiyiωTxi+b−1.
The original problem is the minimax problem.[10]minω,bmaxαLω,b,α.
The dual problem of the original problem is a maximin problem:[11]maxω,bminαLω,b,α.
Taking its partial derivative with respect to ω and b and making it equal to 0,[12]ω=∑$i = 1$nαiyixn,∑$i = 1$nαiyi=0.Then,[13]Lω,b,α=∑$i = 1$nαi−12∑$i = 1$nαiαjyiyjxixj.
Combining the abovementioned condition:[14]∑$i = 1$nαiyi=0αi≥0,$i = 1$,2⋯n. Then, we find the maximum value of α.[15]min12∑$i = 1$nαiαjyiyjxixj−∑$i = 1$nαi.
We continue to use Lagrange multipliers:[16]ω∗=∑$i = 1$Nai∗yixi,b∗=yi−∑$i = 1$Nai∗yixixj.
We find the final hyperplane.
## 3.1. The Peak Pressure
According to Figure 3 and Table 2, fatigue mainly affected the PP at T1, T2–5, HM, and HL, and gender factors were mainly reflected at T1, T2–5, and M3–5. Specifically, PP decreased significantly in T1 and T2–5 regions after fatigue ($P \leq 0.05$) and increased significantly in HM and HL ($P \leq 0.05$). PP at T1 and T2–5 was significantly higher in females than in males ($P \leq 0.05$), and PP at M3–5 was significantly higher in males than in females ($P \leq 0.05$).
## 3.2. The Peak Force
According to Figure 3 and Table 2, the fatigue effect was mainly reflected in the T1, T2–5, M1, HM, and HL, while the gender effect was mainly reflected in T1, T2–5, and M3–5. PF was significantly decreased at T1 and T2–5 due to fatigue and significantly increased at M1, HM, and HL ($P \leq 0.05$). PF in females was significantly larger at T1 and T2–5 than that in males, and significantly smaller at M3–5 than that in males ($P \leq 0.05$).
## 3.3. The Impulse
According to Figure 3 and Table 2, the fatigue effect was mainly reflected in the toes, M1, and heel, while the gender effect was mainly reflected in T1, T2–5, and M3–5. PI decreased significantly at T1 and T2–5 after fatigue ($P \leq 0.05$), and increased significantly at M1, HM, and HL ($P \leq 0.05$). In addition, the PI at T1 and T2–5 showed that females were significantly larger than males, and at M3–5, females were significantly smaller than males ($P \leq 0.05$).
## 3.4. SVM Classification Algorithm
We selected combinations of plantar zone parameters with significant differences ($P \leq 0.001.$ Figure 4 shows the best fit separating hyperplane lines of fatigue or not fatigue in different plantar zone parameter combinations. The accuracy of the different plantar zone parameter combinations in predicting fatigue is presented in Table 3. The results showed that the average accuracy was a moderate level (train accuracy: $62.5\%$; test accuracy: $62.5\%$). The accuracies of following combinations were above average and showed a high level: T1 PP/HL PF (train accuracy: $65\%$; test accuracy: $75\%$), T1 PF/HL PF (train accuracy: $67.5\%$; test accuracy: $65\%$), and HL PF/T1 PI (train accuracy: $67.5\%$; test accuracy: $70\%$).
## 4. Discussion
This research aimed to analyze how PP, PF, and PI changed before and after running fatigue in novice runners, as well as gender differences. Based on previous studies, we assumed that the changes in plantar mechanical parameters before and after fatigue mainly occurred in T1 and T2–5. It was also assumed that gender differences in plantar parameters were mainly concentrated in T1 and T2–5 and M1–5. In addition, it was assumed that SVM can predict fatigue at a high level. Our results are largely consistent with our previous assumptions.
The changes in plantar mechanical parameters caused by fatigue were mainly under the T1, T2–5, M1, HM, and HL. The plantar mechanical parameters in the toes region were also reduced in the research of Bisiaux and Moretto [13], Karagounis et al. [ 47], and Willems et al. [ 48]. This may be due to the increased dorsiflexion of the metatarsophalangeal joint after fatigue, which leads to fewer toes contributing to running, and thus less load under the toes [49]. In a study of PP and center of pressure (COP), it was found that the PP under the toes decreased, with a retraction of the COP. According to Stolwijk et al. [ 50], to avoid overuse of the forefoot and the risk of incurring forefoot pain, subjects adjusted their gait pattern. This could explain why plantar mechanical parameters under the toes were decreased. Nagel et al. [ 49] also noted a decline in toes load. However, in the study of Bisiaux and Moretto [13] and Weist et al. [ 51], the phenomenon of decreased load under the toes was not observed. Moreover, in the study of Willems et al. and Wu et al. [ 48, 52], it was also found that PF and PI at M1 increased significantly after fatigue, which was consistent with our findings. Perhaps because of the reduction in mechanical parameters under the toes, the load was transferred to the metatarsal. However, increased submetatarsal load may contribute to the occurrence of metatarsal stress fractures [49]. An investigation also revealed that after running fatigue, the contact area of the HM grew while the HL reduced. Then led to greater pronation in the rearfoot [53]. If fatigued, the quadriceps need to play a greater role to avoid knee instability, resulting in less knee flexion, which leads to increased heel pressure. This explanation has also been confirmed by Stolwijk et al. [ 50].
Several gender-induced differences in plantar mechanical parameters were found. PP, PF, and PI were significantly higher in females than in males at T1 and T2–5 and significantly higher in males than in females at M3–5. This was also reflected in studies of Ferrari et al. [ 54] and Demirbuken et al. [ 27]. The larger plantar mechanical parameters of females' T1 and T2–5 may be related to the fact that females wear high heels, which also raised the risk of chronic paraspinal muscle fatigue, which was linked to postural changes and pain [55]. Although this study did not include cases of hallux valgus (HV), females had a higher load of the hallux than males. Studies reported a meta-analysis that estimated that female HV prevalence ($30\%$) was 2.3 times greater than that in males ($13\%$) [21]. Although many studies cannot reach a unified conclusion, there was no denying that gender differences in plantar mechanical parameters may be one of the reasons for the increase in hallux valgus in females. Males had much more load in the forefoot area than females, which could be due to males' higher body weight, physical structural differences, and females' better flexibility [56, 57]. Further to this, males tend to have a higher vertical center of mass displacement during walking than females. This may also contribute to a higher load in M3–5 [56]. Pressure is equal to force divided by area. In all regions of the foot, males had a considerably higher contact area than females, both statistically and clinically [25]. At the same time, because of the female hormone secretion, the foot ligament relaxation reduces the degree of stiffness and spreads the forces to a larger extent [58, 59]. This may be the reason that the PP, PF, and PI at M3–5 are higher in males than in females to varying degrees. In this study, gender differences in PP, PF, and PI were mainly found in the T1, T2–5, and M3–5. Although the literature's findings are not always consistent, factors including gender and foot anatomy are thought to be linked to metatarsal stress fractures and lower limb injuries [50].
During running, the feet are the only part of the body that makes direct touch with the ground, and they are critical to progress. Running actions may be hampered by muscle exhaustion and physical discomfort. As a result, it is theoretically possible to predict fatigue through plantar mechanical parameters. Previous research has shown that fatigue will affect plantar mechanical parameter distribution and fatigue is correlated with plantar mechanical parameters [48]. Interval maximization is the SVM classification algorithm, which may be characterized as a problem of solving convex quadratic programming and is equivalent to the regularized hinges loss work minimization issue. The SVM classification algorithm is an optimization algorithm for solving convex quadratic programming, as evidenced by our SVM classification algorithm results. The SVM classification algorithm results revealed that the mean accuracy was an above-moderate level. The accuracy was of an above-average level by using the T1 PP/HL PF, T1 PF/HL PF, and HL PF/T1 PI. These indicated that fatigue can be predicted to a certain extent by monitoring plantar mechanical parameters before and after running fatigue. Running fatigue can be predicted using the learned SVM classification algorithm, which can also be used as a useful tool for fatigue supervision. The learned SVM classification algorithm can help coaches to better identify the physical state of athletes from start to the finish of a run by monitoring plantar mechanical parameters. The classification may also be useful in identifying injuries over the running season.
There are some limits of this study. In the experiment, a treadmill was used for the fatigue intervention. We only studied the plantar mechanical parameters under treadmill conditions but did not study the condition of running on the ground. Future studies should include subjects performing at different exercise levels, such as professional athletes. In addition, the sample size should be expanded to improve the accuracy of the SVM classification algorithm.
## 5. Conclusions
We found that the change of plantar mechanical parameters caused by fatigue was mainly concentrated in T1, T2–5, M1, HM, and HL. While the effect of gender was mainly found in the T1, T2–5, and M3–5. These may indicate injuries related to fatigue and gender factors, such as metatarsal stress fractures and HV. Plantar mechanical parameters can be monitored before and after long-distance running to predict fatigue to some extent. The learned algorithm of plantar zone combinations with above-average accuracy (T1 PP/HL PF, T1 PF/HL PF, and HL PF/T1 PI) can predict long-distance running fatigue and provide supervised training strategies.
## Data Availability
The data used to support the findings of this study are available from the corresponding authors upon request.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## Authors' Contributions
Q. L. and H. C. conceived the presented idea, developed the framework, and wrote the manuscript. A. T., J. S. B., M. Y., and Y. G. provided critical feedback and contributed to the final version. All authors were involved in the final direction of the paper and contributed to the final version of the manuscript. All authors have read and agreed to the published version of the manuscript.
## References
1. Cooper D., Kavanagh R., Bolton J., Keaver L.. **A pilot 6-week lifestyle intervention in women aged 50+ in Ireland**. (2022.0) **6** 180-188. DOI: 10.5334/paah.195
2. Schwartz R. S., Kraus S. M., Schwartz J. G.. **Increased coronary artery plaque volume among male marathon runners**. (2014.0) **111** 89-94. PMID: 30323509
3. Hulme A., Nielsen R. O., Timpka T., Verhagen E., Finch C.. **Risk and protective factors for middle-andlong-distancerunning-related injury**. (2017.0) **47** 869-886. DOI: 10.1007/s40279-016-0636-4
4. Paluska S. A.. **An overview of hip injuries in running**. (2005.0) **35** 991-1014. DOI: 10.2165/00007256-200535110-00005
5. Saragiotto B. T., Yamato T. P., Hespanhol Junior L. C., Rainbow M. J., Davis I. S., Lopes A. D.. **What are the main risk factors for running-related injuries?**. (2014.0) **44** 1153-1163. DOI: 10.1007/s40279-014-0194-6
6. Yeung S. S., Yeung E. W., Gillespie L. D.. **Interventions for preventing lower limb soft‐tissue running injuries**. (2011.0) **6** CD001256-142. DOI: 10.1002/14651858.cd001256.pub2
7. Chandra Y., Tewari R., Jain A.. **Experimental studies on acrylic dielectric elastomers as actuator for artificial skeletal muscle application**. (2021.0) **37** 65-82. DOI: 10.1504/ijbet.2021.10040911
8. Saldanha A., Nordlund Ekblom M. M., Thorstensson A.. **Central fatigue affects plantar flexor strength after prolonged running**. (2008.0) **18** 383-388. DOI: 10.1111/j.1600-0838.2007.00721.x
9. Giandolini M., Vernillo G., Samozino P.. **Fatigue associated with prolonged graded running**. (2016.0) **116** 1859-1873. DOI: 10.1007/s00421-016-3437-4
10. Balachandar K., Muralidharan N., Jawahar N., Chockalingam K.. **Development of comfort fit lower limb prosthesis by reverse engineering and rapid prototyping methods and validated with gait analysis**. (2021.0) **35** 362-381. DOI: 10.1504/ijbet.2021.10037469
11. Dowling G. J., Murley G. S., Munteanu S. E.. **Dynamic foot function as a risk factor for lower limb overuse injury: a systematic review**. (2014.0) **7** 53-13. DOI: 10.1186/s13047-014-0053-6
12. Kim H. K., Mirjalili S. A., Fernandez J.. **Gait kinetics, kinematics, spatiotemporal and foot plantar pressure alteration in response to long-distance running: systematic review**. (2018.0) **57** 342-356. DOI: 10.1016/j.humov.2017.09.012
13. Bisiaux M., Moretto P.. **The effects of fatigue on plantar pressure distribution in walking**. (2008.0) **28** 693-698. DOI: 10.1016/j.gaitpost.2008.05.009
14. Kang J.-H., Chen M.-D., Chen S.-C., Hsi W.-L.. **Correlations between subjective treatment responses and plantar pressure parameters of metatarsal pad treatment in metatarsalgia patients: a prospective study**. (2006.0) **7** 95-98. DOI: 10.1186/1471-2474-7-95
15. Ribeiro A. P., Trombini-Souza F., Tessutti V. D., Lima F. R., João S. M., Sacco I. C.. **The effects of plantar fasciitis and pain on plantar pressure distribution of recreational runners**. (2011.0) **26** 194-199. DOI: 10.1016/j.clinbiomech.2010.08.004
16. Chen H., Song Y., Xuan R., Hu Q., Baker J. S., Gu Y.. **Kinematic comparison on lower limb kicking action of fetuses in different gestational weeks: a pilot study**. (2021.0) **9** p. 1057. DOI: 10.3390/healthcare9081057
17. Merry K. J., Macdonald E., MacPherson M.. **Classifying sitting, standing, and walking using plantar force data**. (2021.0) **59** 257-270. DOI: 10.1007/s11517-020-02297-4
18. Sazonov E. S., Fulk G., Hill J., Schutz Y., Browning R.. **Monitoring of posture allocations and activities by a shoe-based wearable sensor**. (2011.0) **58** 983-990. DOI: 10.1109/tbme.2010.2046738
19. Abdul Razak A. H., Zayegh A., Begg R. K., Wahab Y.. **Foot plantar pressure measurement system: a review**. (2012.0) **12** 9884-9912. DOI: 10.3390/s120709884
20. Hong Y., Wang L., Li J. X., Zhou J. H.. **Comparison of plantar loads during treadmill and overground running**. (2012.0) **15** 554-560. DOI: 10.1016/j.jsams.2012.01.004
21. Yamamoto T., Hoshino Y., Kanzaki N.. **Plantar pressure sensors indicate women to have a significantly higher peak pressure on the hallux, toes, forefoot, and medial of the foot compared to men**. (2020.0) **13** 40-47. DOI: 10.1186/s13047-020-00410-2
22. Dempster J., Dutheil F., Ugbolue U. C.. **The prevalence of lower extremity injuries in running and associated risk factors: a systematic review**. (2021.0) **5** 133-145. DOI: 10.5334/paah.109
23. Wolf J. M., Cannada L., Van Heest A. E., O’Connor M. I., Ladd A. L.. **Male and female differences in musculoskeletal disease**. (2015.0) **23** 339-347. DOI: 10.5435/jaaos-d-14-00020
24. Ang C. K., Solihin M. I., Chan W. J., Ong Y. Y.. **Study of plantar pressure distribution**. (2018.0) **237**. DOI: 10.1051/matecconf/201823701016
25. Putti A., Arnold G., Abboud R.. **Foot pressure differences in men and women**. (2010.0) **16** 21-24. DOI: 10.1016/j.fas.2009.03.005
26. Murphy D. F., Beynnon B. D., Michelson J. D., Vacek P. M.. **Efficacy of plantar loading parameters during gait in terms of reliability, variability, effect of gender and relationship between contact area and plantar pressure**. (2005.0) **26** 171-179. DOI: 10.1177/107110070502600210
27. Demirbüken İ., Özgül B., Timurtaş E., Yurdalan S. U., Çekin M. D., Polat M. G.. **Gender and age impact on plantar pressure distribution in early adolescence**. (2019.0) **53** 215-220. DOI: 10.1016/j.aott.2019.01.006
28. Fukuchi R. K., Eskofier B. M., Duarte M., Ferber R.. **Support vector machines for detecting age-related changes in running kinematics**. (2011.0) **44** 540-542. DOI: 10.1016/j.jbiomech.2010.09.031
29. Chan Y.-Y., Fong D. T.-P., Chung M. M.-L.. **Identification of ankle sprain motion from common sporting activities by dorsal foot kinematics data**. (2010.0) **43** 1965-1969. DOI: 10.1016/j.jbiomech.2010.03.014
30. Sain S. R.. (1996.0)
31. Zhang J., Lockhart T. E., Soangra R.. **Classifying lower extremity muscle fatigue during walking using machine learning and inertial sensors**. (2014.0) **42** 600-612. DOI: 10.1007/s10439-013-0917-0
32. Verschueren J., Tassignon B., De Pauw K.. **Does acute fatigue negatively affect intrinsic risk factors of the lower extremity injury risk profile? A systematic and critical review**. (2020.0) **50** 767-784. DOI: 10.1007/s40279-019-01235-1
33. Van Eetvelde H., Mendonça L. D., Ley C., Seil R., Tischer T.. **Machine learning methods in sport injury prediction and prevention: a systematic review**. (2021.0) **8** 27-15. DOI: 10.1186/s40634-021-00346-x
34. Botros F. S., Taher M. F., ElSayed N. M., Fahmy A. S.. **Prediction of diabetic foot ulceration using spatial and temporal dynamic plantar pressure**. 43-47
35. Aguirre A., Pinto M. J., Cifuentes C. A., Perdomo O., Díaz C. A. R., Múnera M.. **Machine learning approach for fatigue estimation in sit-to-stand exercise**. (2021.0) **21** p. 5006. DOI: 10.3390/s21155006
36. Si W., Yang G., Chen X., Jia J.. **Gait identification using fractal analysis and support vector machine**. (2019.0) **23** 9287-9297. DOI: 10.1007/s00500-018-3609-8
37. Jeong G.-M., Truong P. H., Choi S.-I.. **Classification of three types of walking activities regarding stairs using plantar pressure sensors**. (2017.0) **17** 2638-2639. DOI: 10.1109/jsen.2017.2682322
38. Stetter B. J., Möhler F., Krafft F. C., Sell S., Stein T.. **Identification of fatigue-related kinematic changes in elite runners using a support vector machine approach**. (2020.0) **38** p. 264
39. Wang G., Mao X., Zhang Q., Lu A.. **Fatigue detection in running with inertial measurement unit and machine learning**. 85-90
40. Trentzsch K., Schumann P., Śliwiński G.. **Using machine learning algorithms for identifying gait parameters suitable to evaluate subtle changes in gait in people with multiple sclerosis**. (2021.0) **11** p. 1049. DOI: 10.3390/brainsci11081049
41. Aghakeshizadeh F., Letafatkar A., Ataabadi P. A., Hosseinzadeh M.. **The effect of taping on maximum plantar pressure and ground reaction force in people with flat foot after applying a fatigue protocol**. (2021.0) **7** 203-226
42. García-Pinillos F., Cartón-Llorente A., Jaén-Carrillo D.. **Does fatigue alter step characteristics and stiffness during running?**. (2020.0) **76** 259-263. DOI: 10.1016/j.gaitpost.2019.12.018
43. Clansey A. C., Hanlon M., Wallace E. S., Lake M. J.. **Effects of fatigue on running mechanics associated with tibial stress fracture risk**. (2012.0) **44** 1917-1923. DOI: 10.1249/mss.0b013e318259480d
44. Anbarian M., Esmaeili H.. **Effects of running-induced fatigue on plantar pressure distribution in novice runners with different foot types**. (2016.0) **48** 52-56. DOI: 10.1016/j.gaitpost.2016.04.029
45. Koblbauer I. F., van Schooten K. S., Verhagen E. A., van Dieën J. H.. **Kinematic changes during running-induced fatigue and relations with core endurance in novice runners**. (2014.0) **17** 419-424. DOI: 10.1016/j.jsams.2013.05.013
46. García-Pérez J. A., Pérez-Soriano P., Llana S., Martínez-Nova A., Sánchez-Zuriaga D.. **Effect of overground vs. treadmill running on plantar pressure: influence of fatigue**. (2013.0) **38** 929-933. DOI: 10.1016/j.gaitpost.2013.04.026
47. Karagounis P., Prionas G., Armenis E., Tsiganos G., Baltopoulos P.. **The impact of the Spartathlon ultramarathon race on athletes’ plantar pressure patterns**. (2009.0) **2** 173-178. DOI: 10.1177/1938640009342894
48. Willems T. M., De Ridder R., Roosen P.. **The effect of a long-distance run on plantar pressure distribution during running**. (2012.0) **35** 405-409. DOI: 10.1016/j.gaitpost.2011.10.362
49. Nagel A., Fernholz F., Kibele C., Rosenbaum D.. **Long distance running increases plantar pressures beneath the metatarsal heads: a barefoot walking investigation of 200 marathon runners**. (2008.0) **27** 152-155. DOI: 10.1016/j.gaitpost.2006.12.012
50. Stolwijk N. M., Duysens J., Louwerens J. W. K., W Keijsers N. L.. **Plantar pressure changes after long-distance walking**. (2010.0) **42** 2264-2272. DOI: 10.1249/mss.0b013e3181e305f4
51. Weist R., Eils E., Rosenbaum D.. **The influence of muscle fatigue on electromyogram and plantar pressure patterns as an explanation for the incidence of metatarsal stress fractures**. (2004.0) **32** 1893-1898. DOI: 10.1177/0363546504265191
52. Wu W.-L., Chang J.-J., Wu J.-H., Guo L.-Y., Lin H.-T.. **EMG and plantar pressure patterns after prolonged running**. (2007.0) **19** 383-388. DOI: 10.4015/s1016237207000483
53. Escamilla-Martínez E., Martínez-Nova A., Gómez-Martín B., Sánchez-Rodríguez R., Fernández-Seguín L. M.. **The effect of moderate running on foot posture index and plantar pressure distribution in male recreational runners**. (2013.0) **103** 121-125. DOI: 10.7547/1030121
54. Ferrari J., Watkinson D.. **Foot pressure measurement differences between boys and girls with reference to hallux valgus deformity and hypermobility**. (2005.0) **26** 739-747. DOI: 10.1177/107110070502600912
55. Gimunová M., Zvonař M., Mikeska O.. **The effect of aging and gender on plantar pressure distribution during the gait in elderly**. (2018.0) **20** 139-144
56. Chung M.-J., Wang M.-J.. **Gender and walking speed effects on plantar pressure distribution for adults aged 20–60 years**. (2012.0) **55** 194-200. DOI: 10.1080/00140139.2011.583359
57. Zhang B., Lu Q.. **A current review of foot disorder and plantar pressure alternation in the elderly**. (2020.0) **4** 95-106. DOI: 10.5334/paah.57
58. Xu D., Quan W., Zhou H., Sun D., Baker J. S., Gu Y.. **Explaining the differences of gait patterns between high and low-mileage runners with machine learning**. (2022.0) **12** p. 2981. DOI: 10.1038/s41598-022-07054-1
59. Xiang L., Mei Q., Wang A., Shim V., Fernandez J., Gu Y.. **Evaluating function in the hallux valgus foot following a 12-week minimalist footwear intervention: a pilot computational analysis**. (2022.0) **132**. DOI: 10.1016/j.jbiomech.2022.110941
|
---
title: Ultrasonic-assisted enzymatic improvement of polyphenol content, antioxidant
potential, and in vitro inhibitory effect on digestive enzymes of Miang extracts
authors:
- Nalapat Leangnim
- Kridsada Unban
- Patcharapong Thangsunan
- Suriya Tateing
- Chartchai Khanongnuch
- Apinun Kanpiengjai
journal: Ultrasonics Sonochemistry
year: 2023
pmcid: PMC9988395
doi: 10.1016/j.ultsonch.2023.106351
license: CC BY 4.0
---
# Ultrasonic-assisted enzymatic improvement of polyphenol content, antioxidant potential, and in vitro inhibitory effect on digestive enzymes of Miang extracts
## Highlights
•Ultrasonic-assisted enzymatic extraction improved polyphenol and flavonoid contents.•Ultrasonic-assisted enzymatic extraction promoted the release of gallated catechins.•Tannase treated Miang extract exhibited high antioxidant activity.•Tannase treated Miang extract could potentially inhibit digestive enzymes.
## Abstract
The aims of this research were to optimize the ultrasonic-assisted enzymatic extraction of polyphenols under Miang and tannase treatment conditions for the improvement of antioxidant activity of Miang extracts via response surface methodology. Miang extracts treated with and without tannase were investigated for their inhibitory effects on digestive enzymes. The optimal conditions for ultrasonic-assisted enzymatic extraction of the highest total polyphenol (TP) (136.91 mg GAE/g dw) and total flavonoid (TF) (5.38 mg QE/g dw) contents were as follows: 1 U/g cellulase, 1 U/g xylanase, 1 U/g pectinase, temperature (74 °C), and time (45 min). The antioxidant activity of this extract was enhanced by the addition of tannase obtained from *Sporidiobolus ruineniae* A45.2 undergoing ultrasonic treatment and under optimal conditions (360 mU/g dw, 51 °C for 25 min). The ultrasonic-assisted enzymatic extraction selectively promoted the extraction of gallated catechins from Miang. Tannase treatment improved the ABTS and DPPH radical scavenging activities of untreated Miang extracts by 1.3 times. The treated Miang extracts possessed higher IC50 values for porcine pancreatic α-amylase inhibitory activity than those that were untreated. However, it expressed approximately 3 times lower IC50 values for porcine pancreatic lipase (PPL) inhibitory activity indicating a marked improvement in inhibitory activity. The molecular docking results support the contention that epigallocatechin, epicatechin, and catechin obtained via the biotransformation of the Miang extracts played a crucial role in the inhibitory activity of PPL. Overall, the tannase treated Miang extract could serve as a functional food and beneficial ingredient in medicinal products developed for obesity prevention.
## Introduction
Obesity is defined as the abnormal or excessive accumulation of fat that could impair health. This would further support the contention that the fundamental cause of obesity is an energy imbalance between the number of calories consumed and the number of calories expended [1]. It is considered an important risk factor for chronic diseases such as cardiovascular disease, hypertension, hyperlipemia, type 2 diabetes, fatty liver, heart disease, and certain cancers [2], [3]. Accordingly, a trigger for obesity and its associated comorbidities could be intricately linked to an increase in reactive oxygen species (ROS) and a subsequent increase in oxidative stress [4]. Pharmacological and surgical interventions are often employed in strategies administered to prevent obesity; however, they have been associated with a fairly high cost as well as a range of negative effects and potentially hazardous side effects. The development of nutrient digestion and absorption inhibitors to reduce the degree of energy intake through gastrointestinal mechanisms is one of the most promising strategies in the treatment of obesity [5]. Yet, attempts to find a variety of natural products that can inhibit digestive enzymes, along with those that possess antioxidant activity, have received a considerable amount of attention.
Tea (*Camellia sinensis* (L.) Kuntze) is the most widely consumed plant-based beverage in the world. It is commonly known as a rich source of polyphenols. Catechins, the main polyphenols in tea, are considered potentially beneficial biological substances for health and well-being with regard to their antioxidant activity, anti-inflammatory activity, effect on cancer prevention, and regulation of lipid metabolisms [6]. Miang is a traditional fermented tea leaves of C. sinensis var. assamica, which is typically found in northern Thailand. Miang consists of important biological substances and microorganisms that include tannins, catechins and their derivatives, and organic acids, as well as certain potential probiotic microorganisms [7], [8], [9], [10]. Currently, Miang extracts have exhibited antimicrobial, antioxidant, and anti-inflammatory activities [7], [11]; however, other beneficial health promoting effects attributed to the Miang extract would require further investigation. The non-filamentous fungi growth-based process (NFP) used to produce Miang resulted in considerably higher polyphenol content than the Miang produced via the filamentous fungi growth-based process (FFP) [11]; however, higher polyphenol content and epicatechin content were considerably lower than in three other well-known tea beverages namely green, black, and oolong teas [12], [13]. Based on the outcomes of a previous study, the polyphenol contents of the NFP-Miang ranged from 30 to 35 mg/g dry weight (dw) and the major catechins were epicatechin gallate (ECG), epigallocatechin (EGC), gallocatechin (GC), catechin (C), and epicatechin (EC), while epigallocatechin gallate (EGCG) and gallocatechin gallate (GCG) were present in low amounts at approximately 1 mg/g dw [14].
To further utilize and apply the important bioactive compounds present in low contents at catechins and derivatives, it is inevitable that an appropriated extraction method must be established. The effective extraction of bioactive compounds in tea is dependent upon pH, extraction time and temperature, and solubility. Moreover, the extraction technique can directly influence rate, yield, and purity of the compounds of interest. Four potential extraction techniques have been previously proposed, i.e., solvent-based extraction, microwave-assisted water extraction, ultrasonic extraction, and chemical extraction. Solvent-based and chemical extractions require further steps for solvent removal, whereas exposure to high temperatures during the microwave-assisted extraction process can cause degradation of some bioactive compounds [15]. Notably, ultrasonic-assisted enzymatic extraction can overcome these limitations. For the enhancement of antioxidant activity in tea, some tannases have been found to be able to transform the catechins present in the tea into more active forms [16], [17]. However, only tannase with high substrate specificity towards gallated catechins should be considered. Tannase from *Sporidiobolus ruineniae* A45.2 isolated from *Miang is* a thermostable enzyme [18] that exhibits high specificity toward gallated catechins (data not shown), yet it was found to be suitable for the biotransformation of Miang extracts in this study. The aims of this research study were to optimize the ultrasonic-assisted enzymatic extraction of the polyphenols present in Miang and to optimize the necessary conditions for improvement of antioxidant activity of the Miang extract. Furthermore, the Miang extracts with and without treatment of tannase were investigated for their inhibitory effects on digestive enzymes. The results of this research could be used to support the applicability of utilizing Miang extracts in the promotion of functional foods and as a potential component in the development of obesity prevention substances.
## Chemicals and culture media
Food grade cellulase (within a pH range of 3.0 to 6.5 and a temperature range of 35 to 75 °C), xylanase (within a pH range of 4.0 to 9.0 and a temperature range of 25 to 75 °C), and pectinase (within a pH range of 3 to 6.5 and a temperature range of 35 to 75 °C) were purchased from Winovazyme (Beijing, China). Porcine pancreatic α-amylase (with an optimal temperature of 20 °C and an optimal pH of 7.4), porcine pancreatic lipase (with an optimal temperature of 37 °C and an optimal pH of 7), and lysozyme were purchased from Sigma Aldrich (St. Louis, MO, USA). Tannic acid, methyl gallate, gallic acid, rhodanine, 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,2′-azino-bis (3-ethylbenzothiazoline-6-sulfonic acid) (ABTS), Trolox (6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid), potassium sulfate, quercetin, and Folin-Ciocalteu’s phenol reagent were all of analytical grade and of the highest quality available from Sigma-Aldrich. High-performance liquid chromatography (HPLC) grade standard [-]-epigallocatechin gallate (EGCG), [-]-epicatechin gallate (ECG), [-]-epigallocatechin (EGC), [-]-epicatechin (EC), [-]-gallocatechin gallate (GCG), [-]-catechin gallate (CG), [-]-gallocatechin (GC), (+)-catechin (C), caffeine, and gallic acid (GA) were all purchased from Sigma. All chemicals used for antioxidant activity assays and enzyme production were of analytical grade and were obtained from RCI Labscan (Bangkok, Thailand). The medium ingredients used in this study, such as agar, yeast extract, and malt extract, were all purchased from HiMedia (Nashik, India).
## Microorganism, culture conditions, and production of tannase
A single colony of *Sporidiobolus ruineniae* A45.2 was inoculated in yeast extract-malt extract broth (YMB) (3 g/L yeast extract, 3 g/L malt extract, 10 g/L glucose) and incubated at 30 °C on a 150-rpm rotary shaker for 24 h. Accordingly, $10\%$ (v/v) of inoculum was transferred to YMB supplemented with $1\%$ (w/v) filtered sterile tannic acid and incubated at the same conditions as have been described above. After 48 h of cultivation, the culture was harvested by centrifugation. Cell pellets were washed with 20 mM sodium phosphate buffer pH 6.5 supplemented with $0.1\%$ (v/v) Triton X-100 to remove any gallic acid attached to the yeast cell wall and the residual tannic acid. The cell pellets were then suspended with $20\%$ (w/v) sucrose in 30 mM Tris-HCl pH 8.0. To release the tannase associated with the cell envelope [19], the suspension was supplemented with lysozyme solution prepared in 100 mM EDTA pH 7.3 to yield a final concentration of 0.1 mg/mL of lysozyme prior to being incubated on ice for 40 min. The lysozyme-EDTA treated suspension was centrifuged at 17,350 × g for 15 min at 4 °C. The supernatant was then dialyzed against 20 mM sodium phosphate buffer pH 7.0 at 4 °C until equilibrium was reached. The resulting dialyzed enzyme was then used in further experiments.
Tannase activity was determined according to the method described in a previous study [20] with slight modifications. Briefly, 50 μL of the enzyme solution was mixed with 50 μL of the substrate (12.5 mM methyl gallate in 100 mM sodium phosphate buffer pH 6.5). The reaction was carried out at 37 °C for 20 min. Then, 60 μL of $0.667\%$ (w/v) methanolic rhodanine solution was added into the reaction mixture to stop the reaction and to detect the release of gallic acid from tannic acid. After a 5 min period of incubation at room temperature (25 °C), a pinkish purple color was visualized by adding 40 μL of 0.5 M KOH and the mixture was left at room temperature for 5 min. Finally, 800 μL of distilled water was added, the mixture was vigorously mixed, and absorbance was measured at 520 nm. One unit of tannase was defined as the amount of enzyme that released 1 μmol of gallic acid in 1 min under the assay conditions.
## Preparation of Miang
Astringent Miang (7-day fermentation period) made from young tea leaves of C. sinensis var. assamica was purchased from tea plantations located in Mae Taeng District, Chiang Mai Province, Thailand (N19.19441, E98.77654). It was dried using a vacuum dryer at 50–60 °C prior to being ground and sieved through a 30-mesh screen for further use.
## Optimization for the ultrasonic-assisted enzymatic extraction of total polyphenol (TP) and total flavonoid (TF) contents
In this study, only water was employed as an extraction solvent due to the fact that previous studies have shown its importance as an environmentally friendly solvent with high efficiency in the recovery of antioxidant phytochemicals [21]. Plackett and Burman design (PBD) was used to screen for the most effective factors that positively influenced the extraction efficiency as follows: cellulase (1–10 U/g dw tea), xylanase (1–10 U/g dw tea), pectinase (1–10 U/g dw tea), temperature (45–65 °C), and time (10–50 min). All enzymes used in this experiment, including cellulase, xylanase, and pectinase, were standardized by determining the enzyme activity according to the method previously described by Kanpiengjai et al. [ 22] with some modifications. The definition of each enzyme is as follows. One unit of cellulase was defined as the amount of the enzyme that catalyzed the hydrolysis of cellulose ($0.5\%$ w/v, pH 5.5) to release 1 μmol of reducing sugars equivalent to glucose in 1 min under assay conditions (pH 5.5, 50 °C, 10 min). One unit of xylanase was defined as the amount of the enzyme that catalyzed the hydrolysis of xylan from birchwood ($0.5\%$ w/v, pH 5.5) to release 1 μmol of reducing sugars equivalent to xylose in 1 min under the assay conditions (pH 5.5, 50 °C, 10 min). One unit of pectinase was defined as the amount of the enzyme that catalyzed the hydrolysis of citrus pectin ($0.5\%$ w/v, pH 5.5) to release 1 μmol of reducing sugars equivalent to galacturonic acid in 1 min under assay conditions (pH 5.5, 50 °C, 10 min).
Based on the PBD matrix, 12 treatment combinations were generated with three center points for the extraction of Miang. A total of 0.5 g of Miang was mixed with 5 mL of distilled water prior to enzymes being added and a period of incubation being initiated at a specific temperature and time. All treatment combinations were performed in a 40 kHz ultrasonic bath with 150 W ultrasonic power (GT SONIC-D6, GT-Sonic RoHS, Shanghai Tense Electronical Equipment Co., Ltd., China). The ultrasonic bath was connected with a low temperature water circulator (B.E. Marubishi Co. Ltd., Tokyo, Japan) in order to avoid thermal variance due to ultrasonic thermal effect. After the extraction process, the mixture was centrifuged at 17,350 × g. The clear supernatant was immediately collected for determination of TP and TF contents that corresponded to each treatment combination. The experimental responses were fitted with the first-order model. The impact of a range of factors affecting the response variables and model reliability was evaluated by analysis of variance (ANOVA) and regression analysis, respectively. The factors whose p-values were less than 0.05 were considered significant factors and were further optimized by central composite design (CCD). In CCD, five-level coded values of each screened factor were established including factorial points (– 1, + 1), axial points (– α, + α), and the center point [0]. The number of generated treatment combinations are dependent upon the number of screened factors. Extraction of Miang was also performed in an ultrasonic bath under the same conditions that had previously been described. The TP and TF contents were determined to achieve the response variables that fit with the second-order polynomial model. ANOVA and regression analysis were also performed as has been described above. Finally, the regression equation together with the 3D-contour plots were used to predict the optimal values of the experimental factors for the highest TP and TF contents. In addition, validation of the predicted value was performed to ensure optimal model fitting. Under optimal conditions, the effects of ultrasonic or enzyme treatment on the extraction of TP and TF were compared with the control (no ultrasonic and no enzyme treatments). The profiles of gallic acid, caffeine, and catechins, as well as their contents, were determined by high-performance liquid chromatography (HPLC). Fig. 1 summarizes the Miang extraction procedure for untreated Miang extracts. Fig. 1A schematic diagram for extraction of polyphenols and flavonoids from Miang.
## Optimization for ultrasonic-assisted enzymatic treatment of Miang extract
The enhancement of antioxidant activity of the Miang extract was investigated by employing the ultrasonic-assisted tannase treatment. After the optimal conditions were established for the extraction of Miang, the obtained extract was immediately treated with tannase derived from S. ruineniae A45.2. Here, optimal levels of tannase, temperature, and time were investigated using CCD. Five values for tannase (500, 898, 750, 601, and 1000 mU/g tea), temperature (30, 34, 40, 46, and 50 °C), and time (5, 10, 17.5, 25, and 30 min) (Table 1S) were set in order to generate 14 treatment combinations with six center points, thus resulting in a set of 20 conditions being established for the treatment of Miang extract with tannase. After the treatment of Miang extract with tannase, the DPPH and ABTS radical scavenging activities obtained from each treatment combination were determined and expressed as μmol Trolox equivalent (TE)/g dw. Statistical analyses of the response variables were performed as has been previously described. Miang extract obtained from different extraction methods (Section 2.4) was treated with tannase by employing the optimal conditions. The contents of catechins, caffeine, and gallic acid obtained from all treated Miang extracts were determined by HPLC.
## Determination of TP content
The TP content of the extract was determined by employing the Folin-Ciocalteu method according to the procedure described in a previous study [23]. A sample of 0.80 mL was mixed with 0.05 mL of Folin-Ciocalteu reagent. After 1 min, 0.150 mL of $20\%$ (w/v) sodium carbonated was added. The mixture was then allowed to stand at room temperature in the dark for 120 min prior to the absorbance being measured at 750 nm. Gallic acid (GA) was used as the standard. TP content was expressed as mg GA equivalent (GAE)/g dry weight (dw) of Miang.
## Determination of TF content
The TF content of the extract was determined by employing the aluminium chloride method according to the procedure explained in a previous study [24]. A total of 200 μL of Miang extract was mixed with 40 μL of $10\%$ (w/v) AlCl3 solution prepared in methanol, 40 μL of 1 M potassium acetate, and 1.12 mL of distilled water. The mixture was incubated for 30 min at room temperature prior to the absorbance being measured at 415 nm. Quercetin (Q) was used as the standard. The TF content was expressed as mg quercetin equivalent (QE)/g dw.
## Assays of antioxidants
DPPH assay was performed by mixing a sample (0.25 mL) with 2.25 mL of freshly prepared 40 mg/L methanolic DPPH. The reaction was allowed to stand in the dark at room temperature (25 °C). A decrease in absorbance at 517 nm was determined after 30 min of incubation. The concentration of the sample that produced a degree of inhibition between $20\%$ and $80\%$ of the blank absorbance was determined and adapted. Radical scavenging activity was expressed as the concentration of the extract required for inhibition of the initial concentration of DPPH by $50\%$ (IC50) under specified experimental conditions. DPPH radical scavenging activity was expressed as μmol TE/g dw.
ABTS radical scavenging activity was then performed. ABTS of 0.0384 g was dissolved in 10 mL of deionized water. Then, 5 mL of the solution was mixed with 88 μL of 140 mM potassium persulfate and adjusted to 25 mL with deionized water in a volumetric flask for further experimentation. An ABTS solution of 1.75 mL was mixed thoroughly with 0.25 mL of the sample. The reaction was allowed to stand in the dark at room temperature. A decrease in absorbance at 734 nm was determined after 30 min of incubation. The concentration of the sample that produced between $20\%$ and $80\%$ inhibition of the blank absorbance was then determined and adapted. Radical scavenging activity was expressed as the concentration of the extract required for inhibition of the initial concentration of ABTS by $50\%$ (IC50) under specified experimental conditions. ABTS radical scavenging activity was expressed as μmol TE/g dw.
## Determination of gallic acid, caffeine, and catechins by HPLC
The gallic acid, caffeine, and catechin contents were determined using an HPLC system equipped with an Inertsil ODS-3 (5 μm, 25 × 0.46 cm ID) (GL Sciences Inc., Tokyo, Japan) and a UV–Vis detector. The mobile phase consisted of $0.05\%$ phosphoric acid (solvent A) and acetonitrile (solvent B). Initially, the column was equilibrated with a mixture of $90\%$ solvent A and $10\%$ solvent B. Separation was achieved with a linear gradient program as follows: $25\%$ solvent B and $75\%$ solvent A for 15 min, then increased to $60\%$ solvent B and $40\%$ solvent A for 10 min. Flow rate and separation temperature were set up at 1 mL/min and 25 °C, respectively. The catechins were detected by absorbance at 280 nm. Finally, gallic acid, caffeine, and catechin contents were calculated and expressed as mg/g dw.
## Effect of Miang extracts on porcine pancreatic α-amylase activity
The porcine pancreatic α-amylase (PPA) activity was determined by measurement of soluble starch retained after the enzyme – soluble starch reaction. Briefly, 100 μL of $1.5\%$ (w/v) soluble starch prepared in 100 mM sodium phosphate buffer pH 6.5 was mixed with 100 μL of the PPA and 300 μL of the same buffer or the Miang extract and then incubated at 37 °C. After 10 min, the reaction was stopped by adding 0.5 mL of 1 M HCl. An aliquot amount of the mixture (200 μL) was mixed with 800 μL of iodine solution (0.3 g/L I2, 6 g/L KI). A degree of absorbance at 620 nm was then measured in the reaction. One unit (U) of PPA was defined as the amount of the enzyme required to hydrolyze 1 μg soluble starch in 1 min under the standard assay conditions. To determine the inhibition percentage of Miang against the PPA, PPA activity in the reaction was initially set up as 128 U. The PPA activity was then assayed in the presence of various concentrations of the Miang extract as the inhibitor and compared to that without the presence of an inhibitor. The IC50 value was determined from the regression curve and expressed as g/100 mL of the Miang extract.
## Effect of Miang extracts on porcine pancreatic lipase activity
Based on the findings of a previous study [25], lipase activity was determined by measuring the free fatty acid released after the enzyme – olive oil reaction. Briefly, 3 mL of olive oil was mixed with 2.5 mL of deionized water or the Miang extract, 1 mL of 100 mM sodium phosphate buffer pH 6.5, and 0.5 mL of Tween 80. The mixture was then vigorously mixed using a magnetic stirrer for 15 min to obtain an emulsion. The porcine pancreatic lipase (PPL) (100 U) was added to the emulsified mixture and incubated on a 150-rpm rotary shaker at 37 °C for 30 min. At the end of the incubation period, 3 mL of $95\%$ ethanol was added prior to the mixture, which was then titrated with 50 mM NaOH using an automatic potentiometric titrator. The end point for the titration was set at pH 9.0. One unit of lipase activity was defined as the amount of enzyme that catalyzed the hydrolysis of triglycerides to release 1 microequivalent of fatty acids in 1 min under standard assay conditions. To determine the percentage of the inhibitory concentration of Miang against PPL, the PPL activity in the reaction was initially set up as 200 U. The PPL activity was assayed in the presence of various concentrations of the Miang extract as the inhibitor and compared to that without the presence of an inhibitor. The IC50 value was determined from the regression curve and expressed as g/100 mL of Miang extract.
## Molecular docking analysis
For protein structure preparation, the crystal structure of the porcine pancreatic lipase-colipase in a complex with tetraethylene glycol monooctyl ether (TGME) (PDB ID: 1ETH) [26] was retrieved from the RCSB Protein Data Bank. The TGME molecule was separated from the protein structure using Discovery Studio Client v.21.1.0 (Dassault Systèmes Biovia Corp.). The protein was converted from ‘pdb’ to a ‘pdbqt’ format using Python script (Prepare_receptor4.py) the AutoDock Tool (ADT) and the metal charges were then automatically calculated (e.g., zinc ion = +2.0). Resolution of the three-dimensional grid box (x, y, and z) was set at 26 × 34 × 36 for the active pocket [27] and 22 × 30 × 28 for the catechin binding pocket [28] with a grid spacing of 0.375 Å. The center of the grid was set to 56.658, 47.892, and 122.042 Å for the ×, y, and z dimensions of the active site and 63.305, 27.761, and 149.683 Å for the ×, y, and z dimensions of the catechin binding site, respectively.
For ligand structure preparation, catechin and derivative structures were sketched as a ‘mol2′ file using Discovery Studio Client v.21.1.0. The ligand structures were subsequently assigned according to atom type, while energy optimization was performed using the steepest descent algorithm in the MMFF94 force field via the Avogadro v.1.2.0 program. The ligands were then converted from ‘mol2′ into a ‘pdbqt’ format using MGLTools version 1.5.7.
All molecular docking experiments were performed using AutoDock Vina [29] on a Linux operating platform. The docking parameters were set as follows: exhaustiveness = 20 and 50 for the active site and the catechin binding site, respectively, and an energy range of 2 kcal/mol. The best docking conformation for each complex was selected from the output file based on the position and intermolecular interactions in the active pocket and in the catechin binding site of the porcine lipase-colipase protein. Interactions between the proteins and ligands were visualized by Discovery Studio Clients v.21.1.0.
## Statistical design and analysis
In this study, PBD, CCD, ANOVA, and regression analyses were performing using Design Expert software version 7.0 (Stat-Ease Corporation, MN, USA). All experiments were performed in duplicate. The results are presented as values of mean ± standard deviation (SD). Data analysis of the mean values was performed based on a full factorial complete randomized design (CRD). Multiple comparison tests were performed based on all pairwise comparisons using Tukey’s HSD test at a confidence level of $95\%$. The paired t-test was used to compare the results from the experimental and control groups. For comparison tests, all analyses were carried out using the Statistix software version 8.0 (Analytical software, FL, USA). A probability value of $p \leq 0.05$ was considered significant.
## Extraction of polyphenols and flavonoids from Miang
PBD was used to evaluate the most significant variables influencing the extraction of TP and TF from Miang. For the experimental design matrix of PBD, the experimental and predicted values for the extraction of polyphenols and flavonoids from Miang are shown in Table 1. The maximal TP and TF contents were 128 mg GAE/g dw and 4.81 mg QE/g dw, respectively. The experimental TP and TF contents were well-fitted with the least square linear regression model, wherein the significance of the model fit values ($p \leq 0.05$) was aligned with the R2-values and the adjusted R2-values that were higher than 0.90. Temperature and time were the significant factors ($p \leq 0.05$) that enhanced extractability of TP and TF in contrast to cellulase and xylanase, which were found to have a significantly negative effect on the extraction of the compounds. Table 1Experimental design matrix of PBD and response variables for screening of the most significant factors enhancing ultrasonic-assisted enzyme extraction of polyphenols and flavonoids from Miang. RunA: CellulaseB: XylanaseC: PectinaseD: Temperature (°C)E: Time (min)Total polyphenols (mg GAE/g dw)Total flavonoids (mg QE/g dw)ActualPredictedActualPredicted110 (+1)10 (+1)1 [-1]65 (+1)50 (+1)111.09115.444.364.4021 [-1]10 (+1)10 (+1)45 [-1]50 (+1)118.64119.673.143.19310 (+1)1 [-1]10 (+1)65 (+1)10 [-1]118.18120.064.084.1441 [-1]10 (+1)1 [-1]65 (+1)50 (+1)126.55123.674.594.5651 [-1]1 [-1]10 (+1)45 [-1]50 (+1)128.00128.203.533.5761 [-1]1 [-1]1 [-1]65 (+1)10 [-1]127.73126.974.814.74710 (+1)1 [-1]1 [-1]45 [-1]50 (+1)117.73118.654.053.85810 (+1)10 (+1)1 [-1]45 [-1]10 [-1]107.91104.893.103.26910 (+1)10 (+1)10 (+1)45 [-1]10 [-1]106.73106.212.962.82101 [-1]10 (+1)10 (+1)65 (+1)10 [-1]118.73119.764.013.921110 (+1)1 [-1]10 (+1)65 (+1)50 (+1)128.91125.294.274.34121 [-1]1 [-1]1 [-1]45 [-1]10120.27121.653.733.81135.5 [0]5.5 [0]5.5 [0]55 [0]30 [0]122.36120.824.054.05145.5 [0]5.5 [0]5.5 [0]55 [0]30 [0]120.82120.824.064.05155.5 [0]5.5 [0]5.5 [0]55 [0]30 [0]119.27120.824.034.05 On the other hand, pectinase had no effect on the extraction of polyphenols, but it did strongly affect the extraction of flavonoids (Table 2). Based on the results, temperature and time were selected for further optimization. Other factors were fixed at their low levels in terms of their effects and their relevant significant differences (1 U/g cellulase, 1 U/g xylanase, and 1 U/g pectinase).Table 2Regression of coefficients and ANOVA of the first-order model for total polyphenols and total flavonoids in PBD.I) Total polyphenolsSourceCoefficient EstimateSum of SquaresdfMean SquareF-Valuep-ValueProb > FModel110.0943593.39745118.679514.96020.0007*A-Cellulase−0.9141203.06401203.064025.59730.0010*B-Xylanase−0.9478218.29821218.298227.51770.0008*C-Pectinase0.14655.212815.21280.65710.4410D-Temperature0.265984.8492184.849210.69570.0113*E-Time0.130781.9731181.973110.33320.0123*Curvature6.249216.24920.78770.4007Residual63.464287.9330Lack of Fit58.687369.78124.09530.2092Pure Error4.776922.3884Cor Total663.110714 Std. Dev.2.8166R20.9034Mean119.5273Adjusted R20.8430C.V. %2.3564Predicted R20.6298PRESS245.4972Adequate Precision12.1113 II) Total flavonoidsSourceCoefficient EstimateSum of SquaresdfMean SquareF-Valuep-ValueProb > FModel1.76183.84650.769250.9451< 0.0001*A-Cellulase−0.01830.08110.08105.36720.0492*B-Xylanase−0.04250.44010.439829.12720.0006*C-Pectinase−0.04870.57710.576738.19340.0003*D-Temperature0.04672.62112.6211173.6044< 0.0001*E-Time0.00520.12710.12738.43350.0198*Curvature0.06410.06364.20950.0743Residual0.12180.0151Lack of Fit0.12060.020060.34040.0164*Pure Error0.00120.0003Cor Total4.03014 Std. Dev.0.1229R20.9695Mean3.9170Adjusted R20.9505C.V. %3.1369Predicted R20.8804PRESS0.4820Adequate Precision22.8780* Significant difference at $p \leq 0.05.$
For CCD optimization, temperature (A) and time (B) were extended to have broader ranges than those in PBD. The experiments were performed to achieve the experimental TP and TF values (Table 3). The ANOVA results confirmed the PBD in terms of the significant effect of temperature and time on the extraction of TP and TF from Miang (Table 2S). The models for extraction of TP and TF were significantly fitted with the second-order model, thus establishing regression equations based on coded values to predict the extractability of TP and TF from Miang as follows:Totalpolyphenols(mgGAE/gdw)=127.82+6.31A+1.79B-0.32AB-3.02A2-4.75B2Totalflavonoids(mgQE/gdw)=4.51+0.74A+0.14B+0.0715AB-0.30A2-0.25B2Table 3Experimental design matrix of CCD and experimental and predicted values of TP and TF for quantitative determination of optimal temperature and time for ultrasonic-assisted enzymatic extraction from Miang. RunA: Temperature (°C)B: Time (min)Total polyphenols (mg GAE/g dw)Total flavonoids (mg QE/g dw)ExperimentalPredictedExperimentalPredicted145 [-1]20 [-1]110.91111.623.033.15275 (+1)20 [-1]126.45124.884.644.50345 [-1]60 (+1)116.09115.843.013.29475 (+1)60 (+1)130.36127.844.914.91539 (-1.414)40 [0]113.55112.843.122.86681 (+1.414)40 [0]128.18130.704.834.96760 [0]12 (-1.414)115.55115.783.783.82860 [0]68 (+1.414)119.27120.864.394.21960 [0]40 [0]128.45127.824.484.511060 [0]40 [0]126.18127.824.484.511160 [0]40 [0]130.55127.824.524.511260 [0]40 [0]127.27127.824.554.511360 [0]40 [0]126.64127.824.544.51 According to the regression models, response surface plots were employed and are shown in Fig. 2. These models produced acceptable results when $p \leq 0.05$ with R2-values between 0.92 and 0.97. A TP content of 131.12 mg GAE/g dw and a TF content of 4.97 mg QE/g dw were predicted from the regression models and successfully validated at $95.5\%$ (136.91 mg GAE/g dw) and $91.9\%$ (5.38 mg QE/g dw), respectively. This was achieved when Miang was extracted at 74 °C for 45 min by employing the ultrasonic, cellulase, xylanase, and pectinase treatments. Under optimal conditions, the ultrasonic-assisted enzymatic extraction method exhibited the potential to significantly increase TP, TF, and TC contents as well as antioxidant activity (Fig. 3) when compared with other extraction methods including enzyme extraction, ultrasonic extraction, and water extraction. Fig. 2Three-dimensional curves and contour plots demonstrating the effects of time and temperature on extractions of polyphenols (a) and flavonoids (b) derived from Miang. Fig. 3Effects of different extraction methods on extractability of TP (a) and TF (b) derived from Miang.
The highest TC content was 91.28 ± 1.76 mg/g dw which included both epicatechins and non-epimer catechins in the following order: C (25.53 ± 0.25 mg/g dw) > EC (17.86 ± 0.60 mg/g dw) > EGC (11.96 ± 0.08 mg/g dw) = ECG (12.53 ± 0.53 mg/g dw) > EGCG (8.00 ± 0.10 mg/g dw) (Fig. 4). In addition, the ultrasonic-assisted enzymatic extraction encouraged the release of significant amounts of gallated catechins. Fig. 4Profile of gallic acid, caffeine, and catechins derived from Miang extract obtained using different extraction methods.
Different lowercase letters in the columns with the same dept indicate differences in antioxidant activity at $p \leq 0.05.$
## Enhancement of antioxidant activity of Miang extract by yeast tannase
The effect of tannase (A), temperature (B), and time (C) on the antioxidant activity of the Miang extract was evaluated based on the design matrix of CCD (Table 4). Consequently, the experimental ABTS and DPPH radical scavenging activities were achieved. Table 4Experimental design matrix of CCD and response variables for enhancement of antioxidant activity of Miang extract by administering ultrasonic-assisted tannase treatment. RunA: Tannase (mU/g dw)B: Temperature (°C)C: Time (min)Antioxidant activity (μmol TE/g dw)ABTS·DPPH·ExperimentalPredictedExperimentalPredicted1141 [-1]38 [-1]10 [-1]2,189.852,162.151,498.051,532.862409 (+1)38 [-1]10 [-1]2,346.682,328.331,537.361,547.973141 [-1]62 (+1)10 [-1]2,227.712,138.521,573.611,542.594409 (+1)62 (+1)10 [-1]2,406.792,356.871,619.071,590.215141 [-1]38 [-1]25 (+1)2,348.212,355.061,619.921,647.536409 (+1)38 [-1]25 (+1)2,395.602,441.731,699.761,729.527141 [-1]62 (+1)25 (+1)2,361.312,336.591,659.731,647.878409 (+1)62 (+1)25 (+1)2,490.822,475.441,798.431,762.36950 (-1.68)50 [0]17.5 [0]2,160.772,220.131,625.221,613.0010500 (+1.68)50 [0]17.5 [0]2,475.082,476.631,707.991,721.9911275 [0]30 (-1.68)17.5 [0]2,271.372,246.481,606.001,544.2712275 [0]70 (+1.68)17.5 [0]2,169.162,254.961,516.571,580.0713275 [0]50 [0]5 (-1.68)2,179.582,268.921,537.521,545.5114275 [0]50 [0]30 (+1.68)2,559.262,530.831,792.921,786.7015275 [0]50 [0]17.5 [0]2,447.012,475.701,740.811,744.2216275 [0]50 [0]17.5 [0]2,440.002,475.701,764.741,744.2217275 [0]50 [0]17.5 [0]2,488.962,475.701,763.701,744.2218275 [0]50 [0]17.5 [0]2,481.922,475.701,707.131,744.2219275 [0]50 [0]17.5 [0]2,496.082,475.701,720.891,744.2220275 [0]50 [0]17.5 [0]2,510.672,475.701,768.341,744.22 These values significantly fitted with the second-order model at $p \leq 0.05$ by establishing two reliable regression models with R2-values between 0.87 and 0.90 for the models ABTS and DPPH (Table 3S). The highest antioxidant activity was predicted by applying the following coding equations:ABTS·scavengingactivity(μmolTE/gdw)=2475.70+76.25A+2.52B+77.87C+13.04AB-19.88AC+1.29BC-45.01A2-79.54B2-26.81C2DPPH·scavengingactivity(μmolTE/gdw)=1744.22+32.40A+10.64B+71.71C+8.13AB+16.72AC-2.35BC-27.12A2-64.36B2-27.62C2 The convex shape of the 3D surface plots indicates the optimal conditions for elevation of antioxidant activities (Fig. 5). The regression models predicted the maximal ABTS and DPPH radical scavenging activities of 2,544.87 and 1,809.26 μmol TE/g dw, respectively, when the Miang extract was treated with 360 mU/g dw and the conditions involved 51 °C for 25 min. The predicted values were successfully validated with 2,557.83 ± 59.46 μmol TE/g dw and 1,822.65 ± 23.65 μmol TE/g dw, which resulted in $99\%$ validation, while the antioxidant activity was increased by up to 1.3 times for both the ABTS and DPPH radical scavenging activities. Fig. 5Three-dimensional curves and contour plots demonstrating the effects of tannase, temperature, and time on ABTS (a) and DPPH radical scavenging activities of the Miang extract obtained by the ultrasonic-assisted enzymatic extraction method (b).
When considering the extracts obtained from the different extraction methods, it was confirmed that tannase promoted the antioxidant activity of the Miang extract (Fig. 6).Fig. 6ABTS and DPPH radical scavenging activities of untreated and treated Miang extracts obtained from different extraction methods. * Significant differences in antioxidant activity between untreated and treated Miang extracts. Different lowercase letters in the columns with the same dept indicate differences in antioxidant activity at $p \leq 0.05.$
Additionally, the ultrasonic-assisted enzymatic extraction method yielded the extract with the highest degree of antioxidant activity, which was then improved by employing the tannase treatment. The treated extracts revealed different profiles of catechins (Fig. 7A) when compared with those of the untreated Miang extracts (Fig. 4). The gallated catechins were significantly transformed into non-gallated catechins via the reaction of tannase and consequently released significant amounts of gallic acid. In addition, this phenomenon led to a high titer of non-gallated catechins namely C (28.23 ± 0.36 mg/g dw), EC (27.93 ± 0.53 mg/g dw), and EGC (20.72 ± 0.12 mg/g dw). The example HPLC chromatograms of untreated and treated Miang extracts are presented in Fig. 1S. For the untreated Miang extracts, the ABTS and DPPH radical scavenging activities displayed positive significant correlations ($p \leq 0.001$) with EGCG ($r = 0.8898$ and $r = 0.9011$), ECG ($r = 0.9176$ and $r = 0.9501$), and CG ($r = 0.8989$ and $r = 0.9190$). After treatment with tannase, positive significant correlations were found between ABTS radical scavenging activity and gallic acid ($r = 0.9866$), C ($r = 0.9593$), EC ($r = 0.9385$), and EGC ($r = 0.8462$) (Fig. 7B). Similarly, positive significant correlations were also found between the DPPH radical scavenging activity and GC ($r = 0.9195$), EC ($r = 0.9111$), EGC ($r = 0.8756$), and gallic acid ($r = 0.8531$), respectively. Fig. 7Profile of gallic acid, caffeine, and catechins derived from the tannase treated Miang extracts obtained from different extraction methods (a) and their correlation with ABTS and DPPH radical scavenging activities (b). Different lowercase letters in the columns with the same dept indicate differences in antioxidant activity at $p \leq 0.05.$ * Indicates significant differences at $p \leq 0.001.$
## In vitro inhibitory effect of Miang extract on digestive enzymes
Both untreated and treated Miang extracts obtained from different extraction methods were evaluated for their potential in vitro inhibitory effect on PPA and PPL activities. The results indicated that the inhibitory activity of the extract against digestive enzymes was directly influenced by the selected extraction method (Table 5).Table 5Pancreatic α-amylase and lipase inhibitory activities (IC50) of *Miang tea* extracts obtained from different extraction methods. Extraction methodα-Amylase inhibitory activity (g/100 mL)Lipase inhibitory activity (g/100 mL)UntreatedTreatedUntreatedTreatedEnzyme + Ultrasonic2.17 ± 0.01a6.44 ± 0.03d1.66 ± 0.02e0.46 ± 0.01aWater + Ultrasonic2.41 ± 0.01b7.12 ± 0.05e2.44 ± 0.01f0.64 ± 0.01bEnzyme2.45 ± 0.02bc7.49 ± 0.03f2.59 ± 0.04 g0.71 ± 0.01cWater2.50 ± 0.03c9.30 ± 0.04 g2.60 ± 0.03 g1.03 ± 0.02dDifferent lowercase letters indicate significant differences in α-amylase inhibitory activity and lipase inhibitory activity at $p \leq 0.05.$
The untreated Miang extract exhibited a significantly stronger inhibitory effect on PPA activity than that of the treated extract. Accordingly, its IC50 value ranged from 2.17 ± 0.01 to 2.50 ± 0.03 g/100 mL. The most effective fraction for PPA inhibition was expressed by the Miang extract obtained from the ultrasonic-assisted enzymatic extraction, while that of the control displayed the lowest degree of inhibitory activity. On the other hand, the treated Miang extract exhibited a more efficient inhibitory effect on the PPL activity than that of the untreated extract, while its IC50 values ranged from 0.46 ± 0.01 to 1.03 ± 0.02 g/100 mL. Reverse correlation analysis indicated that the gallated catechins that were found in the untreated Miang extract were likely to be associated with the best PPA inhibitors ($p \leq 0.0001$), whereas EGC, EC, and C the most abundant compounds from the treated Miang extract, exhibited the strongest degree of inhibitory activity against PPL ($p \leq 0.0001$) (Fig. 8).Fig. 8Reverse correlation of gallic acid, caffeine, and catechins obtained from untreated and treated Miang extracts and the inhibitory activities of porcine pancreatic α-amylase and porcine pancreatic lipase. * Indicates significant differences at $p \leq 0.001.$
## Molecular docking
Molecular docking analysis was conducted to investigate the mechanisms of interaction between PPL as the receptor and the ligand compound by determining the binding affinity and the binding site of the PPL and the ligand compound. The ligands used in this study were EGCG as the positive control for the PPL inhibitor, as has been suggested in the previous study [30], while the top three major compounds that were found in the untreated and treated Miang extracts included C, EC, EGC, and ECG. The active pocket of PPL is composed of Gly77, Phe78, Ile79, Asp80, Trp86, Tyr115, Ser153, Leu154, Asp177, Pro181, His264, and Leu265 with its catalytic triad includes Ser153, Asp177, and His264 [27]. The epicatechins exhibited the lowest degrees of binding energy in the following order ECG = EC < EGCG < EGC = EC. Alternatively, gallic acid exhibited the highest degree of binding energy at −5.9 kcal/mol (Table 6). The binding energy of these catechins ranged from −9.7 to 9.3 kcal/mol. All tested catechins could effectively interact with the active residues of the PPL via a hydrogen bond and hydrophobic interactions with amino acids at the active pocket of the PPL, along with electrostatic interactions with one of the catalytic triad amino acids. Table 6Different PPL (1ETH)-ligand complexes and their binding energy and interactions.1ETH-ligand complexesBinding energy (kcal/mol)Hydrogen bondingHydrophobic interactionsElectrostatic interactions1ETH-C−9.7GLY77Phe78, Tyr115, Phe216His2641ETH-EC−9.3His152Phe78, Tyr115, Phe216His2641ETH-EGC−9.3Gly77Phe78, Tyr115, Phe216His2641ETH-ECG−9.7Gly77Phe78, Tyr115, Phe216, Leu265His2641ETH-EGCG−9.4Gly77, Asp80Phe78, Ile79, Tyr115, Phe216His2641ETH-GA−5.9Asp80, Arg257Ala261, Leu265–
## Discussion
Despite the fact that Miang has exhibited long and deep-rooted social and cultural integration and relevance with the people of northern Thailand, it is currently less popular among younger generations [31]. Finding a potentially beneficial application for Miang would accordingly conserve the tradition of Miang fermentation. Polyphenols, specifically gallated and non-gallated catechins, are the major bioactive compounds of tea. However, their contents in Miang are relatively lower than those of the most popular varieties of tea including green tea, oolong, and pu-erh [11], [13]. Differences in the polyphenol and catechin contents depend upon the quality of the tea leaves [32] and the tea plantation area from which they were grown [33]. Accordingly, an appropriate extraction technique is a key factor affecting the polyphenol production of the tea. In this study, we attempted to present an interesting extraction strategy, namely ultrasonic-assisted enzymatic extraction, for the extraction of certain bioactive compounds, namely polyphenols and flavonoids. Several studies have revealed that polyphenols, such as tannins, catechins, cyanidin-3-glucoside, and quercetin, can interact with the polysaccharides associated with the plant cell wall, i.e., cellulose, hemicellulose, and pectin via hydrogen bonding, hydrophobic interaction, adsorption, and pi-pi interaction. The network of cellulose and hemicellulose is naturally embedded in the pectin matrix, which is the most complex structure of polysaccharides in the plant cell wall [34], [35]. In order to counteract this process, the effect of carbohydrate-active enzymes on mediating polyphenol-cell wall interactions in Miang would be an alternative strategy to specifically weaken or break down the cell wall structure and significantly contribute to the release of more polyphenol content from Miang, as has been suggested in the previous study [35]. In addition, enzymatic extraction provides several advantages as opposed to conventional extraction methods as follows; mild reaction conditions, processes requiring fewer steps, a substrate specificity that in turn leads to high productivity of bioactive compounds with a high degree of bioavailability and quality, and lower production costs by replacing multiple installations that are needed for the classical extraction processes [36]. The PB results showed that temperature and time were positively significant factors for the extraction of TP and TF contents. Elevated temperatures can increase the solubility but reduce the viscosity and surface tension of the solvent, thus promoting solvent penetration into the matrix and improving the extraction process [37]. Temperature is generally a time dependent factor. Although the addition of the enzyme mixture had a significantly negative effect on the extraction of TP and TF, further results revealed that it is essential to add an enzyme mixture to enhance the degree of extraction efficiency (Fig. 3). It can be determined that a range of each enzyme may be too high to be applied in the extraction. Thus, they were fixed at a level of 1 U/g dw. Optimization for the ultrasonic-assisted enzymatic extraction of Miang was successful with higher titers of TP, TF, and TC contents when compared with the conventional extraction method (no enzyme and ultrasonic treatments). NFP Miang contains TP content of 100 mg GAE/g dw and TC content of 5 mg QE/g dw [14], yet the results of this study indicated 1.5- and 2.5-times higher TP and TC contents, respectively. Green tea made from young tea leaves possessed TP ranging from 108.8 to 323.6 mg/g dw [13], [38], which was higher than that of oolong tea (103.5–297.3 mg/g dw), black tea (130.1–181.7 mg/g dw), and dark tea (78.2–162.9 mg/g dw) [13]. The main classes of flavonoids found in tea are flavanols and flavonols. Although catechins are a flavanol class of flavonoids, it has been observed that the TF content is much lower than the TC content. It can be concluded that the determination of TF by Folin-Ciocalteu does have a limitation when detecting catechins, thus TF may only be referred to as myricetin, quercetin, and kaempferol [39]. Previous studies have reported on the amounts of TF (in terms of total amounts of myricetin, quercetin, and kaempferol) in green tea and black tea in ranges of 4.18–8.95 mg/g dw and 3.0–5.86 mg/g dw, respectively [40], [41], while lower TF contents were found at levels of 2.69 mg/g dw for oolong tea and 1.15 mg/g dw for pu-erh [41]. These reported TF contents are in accordance with the results of this study. The ultrasonic-assisted enzymatic extraction of Miang improved the extractability of TC, particularly gallated catechins both in terms of epicatechins (EGCG and ECG) and non-epimer catechins (GCG and CG). This would affirm that bigger molecules with more hydroxyl groups, like gallated catechins, are retained in greater amounts in the plant cell wall due to hydrogen bonding and hydrophobic interactions [42]. Various catechins in tea possess different antioxidant activities depending upon their type. Based on a molar basis, EGC exhibited the greatest activity, followed by EGCG and GA, and then EC and ECG [23]. Furthermore, the antioxidant activity of EC was comparatively equivalent to that of C [43]. Accordingly, higher antioxidant activity is expected in the resultant concentrations of catechins in the treated Miang extract. Yet, the higher antioxidant activity of the treated Miang extract by tannase could have been due to the presence of high amounts of EGC, EC, C, and GA. In terms of the overall extractability results, it could be stated that carbohydrate-active enzymes can reduce the structural integrity or increase the permeability of the cell wall, thus encouraging acoustic cavitation provided by ultrasonic treatment to disrupt interactions between bioactive compounds and the cell wall [23], [44]. Therefore, the ultrasonic-assisted enzymatic extraction method would be the best extraction method for Miang.
Gallated epicatechins, specifically the EGCG present in tea, display high inhibitory activity against human digestive enzymes, such as amylase, lipase, and trypsin [45], [46], [47], [48], which have been associated with hyperlipidemia and obesity. In this study, the untreated Miang extracts that possessed high amounts of gallated catechins, specifically those that were derived from the optimal conditions for ultrasonic-assisted enzymatic extraction, gave lower IC50 values against the PPA than those of the treated extracts. Surprisingly, the treated Miang extract contents consisted of C, EC, and EGC as the major compounds in that respective order, which indicated significantly lower degrees of IC50 values against the PPL than the untreated Miang extracts. The molecular docking results that were expressed in terms of binding energy were used to explain the results. Lower binding energy would indicate the formation of a more stable ligand-receptor complex [30]. According to the results obtained from reverse correlation, the EGC, EC, and C were indicated as three active compounds that displayed binding energy that was as low as EGCG, thus indicating their high binding affinity against PPL. From the results, it can be noted that EGCG, EGC, ECG, EC, and C possess a similar degree of binding affinity against PPL. The structural-activity relationship of lipase inhibition achieved by different extract methods was found to contain high phenolic compounds, which has been reported to be dose dependent and dependent upon the type of substrate used [49]. Those outcomes are in agreement with the findings of this study, which indicate that the inhibition of the PPL does require significant amounts of EGC, EC, and C. This could explain why the tannase treated Miang extracts exhibited greater potential than the untreated extracts. In this study, when considering olive oil as the substrate for PPL, it could be stated that EGC, EC, and C could act as competitive inhibitors for the substrate. The effect of the individual catechins on PPL determined by using olive oil as a substrate may be required to further confirm these results. Moreover, the most recent study has revealed that catechins as non-competitive inhibitors could enhance the inhibitory effect of cyanidin-3-glucoside (C3G), a competitive inhibitor on PPL via the catechin-C3G mixture [28]. Further investigations of the synergistic PPL inhibitory activity of catechin-EGC and catechin-EC mixtures have garnered significant interest.
## Conclusion
This study has demonstrated a new strategy for ultrasonic-assisted enzymatic treatment for the extraction of bioactive compounds from Miang and the treatment of tannase to increase the antioxidant potential of the Miang extract. After the statistical optimization step, this ultrasonic-assisted enzymatic extraction method exhibited significantly higher antioxidant activity and resulted in greater amounts of catechins, particularly gallated catechins, than when a conventional method was used. It would therefore be useful for the extraction of phenolic compounds from Miang and is recommended to be applied for phenolic compound extraction from other types of tea. The higher antioxidant activity of the Miang extract could be established by the specific yeast tannase treatment, which then contributes to the formation of non-gallate catechins and gallic acid. Moreover, the treated Miang extract exhibited potential inhibitory effects against the PPA and the PPL. Regarding the inhibitory activity against PPL, the molecular docking results indicate that the non-gallated catechins, namely EGC, EC, and C, may be associated with a reduction in PPL activity. Importantly, further biological properties of the treated Miang extract, such as anti-wrinkle and anti-hypertensive activities, are of significant interest for further investigations.
## CRediT authorship contribution statement
Nalapat Leangnim: Methodology, Investigation, Formal analysis, Validation, Writing – review & editing. Kridsada Unban: Supervision, Resources. Patcharapong Thangsunan: Investigation, Writing – original draft. Suriya Tateing: Investigation, Writing – original draft. Chartchai Khanongnuch: Supervision, Resources. Apinun Kanpiengjai: Conceptualization, Funding acquisition, Methodology, Investigation, Resources, Supervision, Data curation, Writing – original draft, Writing – review & editing.
## Declaration of Competing Interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Apinun Kanpiengjai reports financial support was provided by Chiang Mai University.
## Supplementary data
The following are the *Supplementary data* to this article:*Supplementary data* 1
## Data availability
Data will be made available on request.
## References
1. Safaei M., Sundararajan E.A., Driss M., Boulila W., Shapi'i A.. **A systematic literature review on obesity: understanding the causes & consequences of obesity and reviewing various machine learning approaches used to predict obesity**. *Comput. Biol. Med.* (2021.0) **136**. DOI: 10.1016/j.compbiomed.2021.104754
2. Sun Y., Wang Y., Song P., Wang H., Xu N., Wang Y., Zhang Z., Yue P., Gao X.. **Anti-obesity effects of instant fermented teas in vitro and in mice with high-fat-diet-induced obesity**. *Food Funct.* (2019.0) **10** 3502-3513. DOI: 10.1039/C9FO00162J
3. Liu T.-T., Liu X.-T., Chen Q.-X., Shi Y.. **Lipase inhibitors for obesity: a review**. *Biomed. Pharmacother.* (2020.0) **128**. DOI: 10.1016/j.biopha.2020.110314
4. Tun S., Spainhower C.J., Cottrill C.L., Lakhani H.V., Pillai S.S., Dilip A., Chaudhry H., Shapiro J.I., Sodhi K.. **Therapeutic efficacy of antioxidants in ameliorating obesity phenotype and associated comorbidities**. *Front. Pharmacol.* (2020.0) **11** 1234. DOI: 10.3389/fphar.2020.01234
5. Kim D.H., Park Y.H., Lee J.S., Jeong H.I., Lee K.W., Kang T.H.. **Anti-obesity effect of DKB-117 through the inhibition of pancreatic lipase and α-amylase activity**. *Nutrients (Basel)* (2020.0) **12** 3053. DOI: 10.3390/nu12103053
6. Yan Z., Zhong Y., Duan Y., Chen Q., Li F.. **Antioxidant mechanism of tea polyphenols and its impact on health benefits**. *Anim. Nutr.* (2020.0) **6** 115-123. DOI: 10.1016/j.aninu.2020.01.001
7. Unban K., Khatthongngam N., Shetty K., Khanongnuch C.. **Nutritional biotransformation in traditional fermented tea (Miang) from north Thailand and its impact on antioxidant and antimicrobial activities**. *J. Food. Sci. Technol.* (2019.0) **56** 2687-2699. DOI: 10.1007/s13197-019-03758-x
8. Unban K., Kodchasee P., Shetty K., Khanongnuch C.. **Tannin-tolerant and extracellular tannase producing**. *Foods (Basel)* (2020.0) **9** 490. DOI: 10.3390/foods9040490
9. Unban K., Chaichana W., Baipong S., Abdullahi A.D., Kanpiengjai A., Shetty K., Khanongnuch C.. **Probiotic and antioxidant properties of lactic acid bacteria isolated from indigenous fermented tea leaves (Miang) of north Thailand and promising application in synbiotic formulation**. *Fermentation (Basel)* (2021.0) **7** 195. DOI: 10.3390/fermentation7030195
10. Chaikaew S., Baipong S., Sone T., Kanpiengjai A., Chui-Chai N., Asano K., Khanongnuch C.. **Diversity of lactic acid bacteria from Miang, a traditional fermented tea leaf in northern Thailand and their tannin-tolerant ability in tea extract**. *J. Microbiol.* (2017.0) **55** 720-729. DOI: 10.1007/s12275-017-7195-8
11. Abdullahi A.D., Kodchasee P., Unban K., Pattananandecha T., Saenjum C., Kanpiengjai A., Shetty K., Khanongnuch C.. **Comparison of phenolic contents and scavenging activities of Miang extracts derived from filamentous and non-filamentous fungi-based fermentation processes**. *Antioxidants (Basel)* (2021.0) **10** 1144. DOI: 10.3390/antiox10071144
12. Sajilata M.G., Singhal R.S., Kamat M.Y.. **The carotenoid pigment zeaxanthin-a review**. *Compr. Rev. Food Sci. Food Saf.* (2008.0) **7** 29-49. DOI: 10.1111/j.1541-4337.2007.00028.x
13. Zhao C., Li C., Liu S., Yang L.. **The galloyl catechins contributing to main antioxidant capacity of tea made from**. *Sci. World J.* (2014.0) **2014**. DOI: 10.1155/2014/863984
14. Unban K., Khatthongngam N., Pattananandecha T., Saenjum C., Shetty K., Khanongnuch C.. **Microbial community dynamics during the non-filamentous fungi growth-based fermentation process of Miang, a traditional fermented tea of north Thailand and their product characterizations**. *Front. Microbiol.* (2020.0) **11** 1515. DOI: 10.3389/fmicb.2020.01515
15. Pasrija D., Anandharamakrishnan C.. **Techniques for extraction of green tea polyphenols: a review**. *Food Bioprocess Technol.* (2015.0) **8** 935-950. DOI: 10.1007/s11947-015-1479-y
16. Xu X.Y., Meng J.M., Mao Q.Q., Shang A., Li B.Y., Zhao C.N., Tang G.Y., Cao S.Y., Wei X.L., Gan R.Y., Corke H., Li H.B.. **Effects of tannase and ultrasound treatment on the bioactive compounds and antioxidant activity of green tea extract**. *Antioxidants (Basel)* (2019.0) **8**. DOI: 10.3390/antiox8090362
17. Liu T.P.S.L., Brandão Costa R.M.P., de Vasconcelos Freitas D.J., Oliveira Nacimento C., de Souza Motta C.M., Bezerra R.P., Nunes Herculano P., Porto A.L.F.. **Tannase from**. *Int. J. Food Sci. Technol.* (2017.0) **52** 652-661. DOI: 10.1111/ijfs.13318
18. Kanpiengjai A., Khanongnuch C., Lumyong S., Haltrich D., Nguyen T.H., Kittibunchakul S.. **Co-production of gallic acid and a novel cell-associated tannase by a pigment-producing yeast,**. *Microb. Cell Fact.* (2020.0) **19** 95. DOI: 10.1186/s12934-020-01353-w
19. Mateos P.F., Jimenez-Zurdo J.I., Chen J., Squartini A.S., Haack S.K., Martinez-Molina E., Hubbell D.H., Dazzo F.B.. **Cell-associated pectinolytic and cellulolytic enzymes in**. *Appl. Environ. Microbiol.* (1992.0) **58** 1816-1822. DOI: 10.1128/aem.58.6.1816-1822.1992
20. Kanpiengjai A., Chui-Chai N., Chaikaew S., Khanongnuch C.. **Distribution of tannin-'tolerant yeasts isolated from Miang, a traditional fermented tea leaf (**. *Int. J. Food Microbiol.* (2016.0) **238** 121-131. DOI: 10.1016/j.ijfoodmicro.2016.08.044
21. 21A. Nicolescu, M. Babotă, L. Zhang, C.I. Bunea, L. Gavrilaș, D.C. Vodnar, A. Mocan, G. Crișan, G. Rocchetti, Optimized ultrasound-assisted enzymatic extraction of phenolic compounds from Rosa canina L. Pseudo-Fruits (Rosehip) and their biological activity, Antioxidants (Basel), 11 (2022), 1123, 10.3390/antiox11061123.
22. Kanpiengjai A., Nuntikaew P., Wongsanittayarak J., Leangnim N., Khanongnuch C.. **Isolation of efficient xylooligosaccharides-fermenting probiotic lactic acid bacteria from ethnic pickled bamboo shoot products**. *Biology* (2022.0) **11** 638. DOI: 10.3390/biology11050638
23. Hong Y.-H., Jung E.Y., Park Y., Shin K.-S., Kim T.Y., Yu K.-W., Chang U.J., Suh H.J.. **Enzymatic improvement in the polyphenol extractability and antioxidant activity of green tea extracts**. *Biosci. Biotechnol. Biochem.* (2013.0) **77** 22-29. DOI: 10.1271/bbb.120373
24. Aryal S., Baniya M.K., Danekhu K., Kunwar P., Gurung R., Koirala N.. **Total phenolic content, flavonoid content and antioxidant potential of wild vegetables from western Nepal**. *Plants (Basel, Switzerland)* (2019.0) **8** 96. DOI: 10.3390/plants8040096
25. Gulua L., Nikolaishvili L., Jgenti M., Turmanidze T., Dzneladze G.. **Polyphenol content, anti-lipase and antioxidant activity of teas made in Georgia**. *Ann. Agrar. Sci.* (2018.0) **16** 357-361. DOI: 10.1016/j.aasci.2018.06.006
26. Hermoso J., Pignol D., Kerfelec B., Crenon I., Chapus C., Fontecilla-Camps J.C.. **Lipase activation by nonionic detergents. The crystal structure of the porcine lipase-colipase-tetraethylene glycol monooctyl ether complex**. *J. Biol. Chem.* (1996.0) **271** 18007-18016. DOI: 10.1074/jbc.271.30.18007
27. Chen Z.S., Wu Y.D., Hao J.H., Liu Y.J., He K.P., Jiang W.H., Xiong M.J., Lv Y.S., Cao S.L., Zhu J.. **Molecular dynamic simulation of the porcine pancreatic lipase in non-aqueous organic solvents**. *Front. Bioeng. Biotechnol.* (2020.0) **8** 676. DOI: 10.3389/fbioe.2020.00676
28. Wang Y., Chen L., Liu H., Xie J., Yin W., Xu Z., Ma H., Wu W., Zheng M., Liu M., Liu J.. **Characterization of the synergistic inhibitory effect of cyanidin-3-O-glucoside and catechin on pancreatic lipase**. *Food Chem.* (2023.0) **404**. DOI: 10.1016/j.foodchem.2022.134672
29. Trott O., Olson A.J.. **AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading**. *J. Comput. Chem.* (2010.0) **31** 455-461. DOI: 10.1002/jcc.21334
30. Rahim A.T.M.A., Takahashi Y., Yamaki K.. **Mode of pancreatic lipase inhibition activity in vitro by some flavonoids and non-flavonoid polyphenols**. *Food Res. Int.* (2015.0) **75** 289-294. DOI: 10.1016/j.foodres.2015.05.017
31. Khanongnuch C., Unban K., Kanpiengjai A., Saenjum C.. **Recent research advances and ethno-botanical history of Miang, a traditional fermented tea (**. *J. Ethn. Foods* (2017.0) **4** 135-144. DOI: 10.1016/j.jef.2017.08.006
32. Liu X., Le Bourvellec C., Renard C.M.G.C.. **Interactions between cell wall polysaccharides and polyphenols: Effect of molecular internal structure**. *Compr. Rev. Food Sci. Food Saf.* (2020.0) **19** 3574-3617. DOI: 10.1111/1541-4337.12632
33. Chen G.-H., Yang C.-Y., Lee S.-J., Wu C.-C., Tzen J.T.C.. **Catechin content and the degree of its galloylation in oolong tea are inversely correlated with cultivation altitude**. *J. Food Drug Anal.* (2014.0) **22** 303-309. DOI: 10.1016/j.jfda.2013.12.001
34. Phan A.D.T., Flanagan B.M., D'Arcy B.R., Gidley M.J.. **Binding selectivity of dietary polyphenols to different plant cell wall components: Quantification and mechanism**. *Food Chem.* (2017.0) **233** 216-227. DOI: 10.1016/j.foodchem.2017.04.115
35. Siemińska-Kuczer A., Szymańska-Chargot M., Zdunek A.. **Recent advances in interactions between polyphenols and plant cell wall polysaccharides as studied using an adsorption technique**. *Food Chem.* (2022.0) **373**. DOI: 10.1016/j.foodchem.2021.131487
36. Gligor O., Mocan A., Moldovan C., Locatelli M., Crișan G., Ferreira I.C.F.R.. **Enzyme-assisted extractions of polyphenols – A comprehensive review**. *Trends Food Sci. Technol.* (2019.0) **88** 302-315. DOI: 10.1016/j.tifs.2019.03.029
37. Oroian M., Ursachi F., Dranca F.. **Ultrasound-assisted extraction of polyphenols from crude pollen**. *Antioxidants (Basel)* (2020.0) **9** 322. DOI: 10.3390/antiox9040322
38. Wang C., Han J., Pu Y., Wang X.. **Tea (**. *Appl. Sci.* (2022.0) **12** 5874. DOI: 10.3390/app12125874
39. Pękal A., Pyrzynska K.. **Evaluation of aluminium complexation reaction for flavonoid content assay**. *Food Anal. Methods* (2014.0) **7** 1776-1782. DOI: 10.1007/s12161-014-9814-x
40. Wang H., Provan G.J., Helliwell K.. **Tea flavonoids: Their functions, utilisation and analysis**. *Trends Food Sci. Technol.* (2000.0) **11** 152-160. DOI: 10.1016/S0924-2244(00)00061-3
41. Peterson J., Dwyer J., Bhagwat S., Haytowitz D., Holden J., Eldridge A.L., Beecher G., Aladesanmi J.. **Major flavonoids in dry tea**. *J. Food Compos. Anal.* (2005.0) **18** 487-501. DOI: 10.1016/j.jfca.2004.05.006
42. Liu D., Lopez-Sanchez P., Martinez-Sanz M., Gilbert E.P., Gidley M.J.. **Adsorption isotherm studies on the interaction between polyphenols and apple cell walls: Effects of variety, heating and drying**. *Food Chem.* (2019.0) **282** 58-66. DOI: 10.1016/j.foodchem.2018.12.098
43. Grzesik M., Naparło K., Bartosz G., Sadowska-Bartosz I.. **Antioxidant properties of catechins: comparison with other antioxidants**. *Food Chem.* (2018.0) **241** 480-492. DOI: 10.1016/j.foodchem.2017.08.117
44. 44Y. Liu, W. Zhe, R. Zhang, Z. Peng, Y. Wang, H. Gao, Z. Guo, J. Xiao, Ultrasonic-assisted extraction of polyphenolic compounds from Paederia scandens (Lour.) Merr. Using deep eutectic solvent: Optimization, identification, and comparison with traditional methods, Ultrason. Sonochem., 86 (2022), 106005, 10.1016/j.ultsonch.2022.106005.
45. Wang S., Sun Z., Dong S., Liu Y., Liu Y.. **Molecular interactions between (−)-epigallocatechin gallate analogs and pancreatic lipase**. *Plos One* (2014.0) **9** e111143. PMID: 25365042
46. Wu X., He W., Yao L., Zhang H., Liu Z., Wang W., Ye Y., Cao J.. **Characterization of binding interactions of (-)-epigallocatechin-3-gallate from green tea and lipase**. *J. Agric. Food. Chem.* (2013.0) **61** 8829-8835. DOI: 10.1021/jf401779z
47. Cui F., Yang K., Li Y.. **Investigate the binding of catechins to trypsin using docking and molecular dynamics simulation**. *Plos One* (2015.0) **10** e0125848. PMID: 25938485
48. Forester S.C., Gu Y., Lambert J.D.. **Inhibition of starch digestion by the green tea polyphenol, (-)-epigallocatechin-3-gallate**. *Mol. Nutr. Food Res.* (2012.0) **56** 1647-1654. DOI: 10.1002/mnfr.201200206
49. Buchholz T., Melzig M.F.. **Polyphenolic compounds as pancreatic lipase inhibitors**. *Planta Med.* (2015.0) **81** 771-783. DOI: 10.1055/s-0035-1546173
|
---
title: Study of the antibacterial effects of the starch-based zinc oxide nanoparticles
on methicillin resistance Staphylococcus aureus isolates from different clinical
specimens of patients from Basrah, Iraq
authors:
- Reham M. Al-Mosawi
- Hanadi Abdulqadar Jasim
- Athir Haddad
journal: AIMS Microbiology
year: 2023
pmcid: PMC9988410
doi: 10.3934/microbiol.2023006
license: CC BY 4.0
---
# Study of the antibacterial effects of the starch-based zinc oxide nanoparticles on methicillin resistance Staphylococcus aureus isolates from different clinical specimens of patients from Basrah, Iraq
## Abstract
This study aimed to assess the efficacy of starch-based zinc oxide nanoparticles (ZnO-NPs) against methicillin-resistant *Staphylococcus aureus* (MRSA) isolates from clinical specimens in Basrah, Iraq. In this cross-sectional study, 61 MRSA were collected from different clinical specimens of patients in Basrah city, Iraq. MRSA isolates were identified using standard microbiology tests, cefoxitin disc diffusion and oxacillin salt agar. ZnO-NPs were synthesized in three different concentrations (0.1 M, 0.05 M, 0.02 M) by the chemical method using starch as the stabilizer. Starch-based ZnO-NPs were characterized using ultraviolet–visible spectroscopy (UV-Vis), X-ray diffraction (XRD), field emission scanning electron microscopy (FE-SEM), energy dispersive X-ray spectroscopy (EDS), and transmission electron microscopy (TEM). The antibacterial effects of particles were investigated by the disc diffusion method. The minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) of the most effective starch-based ZnO-NPs were determined using a broth microdilution assay. The UV-Vis of all concentrations of starch-based ZnO-NPs exhibited a strong absorption band at 360 nm which was characteristic of the ZnO-NPs. XRD assay confirmed the representative hexagonal wurtzite phase of the starch-based ZnO-NPs, and their purity and high crystallinity. The spherical shape with a diameter of 21.56 ± 3.42 and 22.87 ± 3.91 was revealed for the particles by FE-SEM and TEM, respectively. EDS analysis confirmed the presence of zinc (Zn) (61.4 ± $0.54\%$) and oxygen (O) (36 ± $0.14\%$). The 0.1 M concentration had the highest antibacterial effects (mean ± SD of inhibition zone = 17.62 ± 2.65 mm) followed by the 0.05 M concentration (16.03 ± 2.24 mm) and the 0.02 M concentration (12.7 ± 2.57 mm). The MIC and the MBC of the 0.1 M concentration were in the range of 25–50 µg/mL and 50–100 µg/mL, respectively. Infections caused by MRSA can be treated with biopolymer-based ZnO-NPs as effective antimicrobials.
## Introduction
Staphylococcus aureus is a Gram-positive bacterium and commensal microorganism that colonizes about $30\%$ of the anterior nares of human individuals [1]. This bacterium plays a considerable role in causing both nosocomial and community-acquired infections including skin infections, endocarditis, osteomyelitis, bacteremia, necrotizing pneumonia, toxic shock syndrome, infections associated with foreign bodies, post-operative surgical infections, and food poisoning [2]–[4].
During the past few decades, treating infections caused by S. aureus has become challenging due to the emergence of multidrug-resistant, particularly methicillin-resistant *Staphylococcus aureus* (MRSA) [5]. The antimicrobial resistance including methicillin resistance in the MRSA strains is correlated with the acquisition of a mobile genetic element called staphylococcal chromosomal cassette mec (SCCmec), which harbors both the mecA or mecC genes, which are responsible for the production of proteins with low binding affinity for beta-lactam antibiotics such as PBP2a [5],[6]. Today, the emergence of MRSA strains resistant to linezolid, vancomycin, and daptomycin has been reported [7].
Since antimicrobial resistance has emerged, spread, and endured in MRSA strains, it has become imperative to develop new and effective alternatives to traditional antibiotics to treat the infections caused by these pathogens [5]. In this regard, nanotechnology can be used to develop antimicrobial nanomaterials with more effective properties compared with traditional antibiotics [5]. It is largely due to their nanoscale size and distinct structures that nanomaterials including inorganic nanoparticles have demonstrated a novel and developed biological functions [8]. Recent studies have found that zinc oxide nanoparticles (ZnO-NPs) possess safe and stable properties that make them one of the ideal antibacterial agents [5]–[8]. There has been speculation that the antimicrobial activity of ZnO-NPs comes from a free radical formation on the surface of metal oxide, which destroys bacterial cell walls and inhibits their growth [8]. Since the experiments conducted in this field are rarely seen in Iraq, this study aimed to assess the efficacy of starch-based synthesized ZnO-NPs against MRSA isolates collected from clinical specimens in Basrah city, Iraq.
## Ethics statement
This study was approved by the University of Basrah, Basrah, Iraq according to the Declaration of Helsinki. All methods were performed in accordance with the relevant guidelines and regulations of the University of Basrah, Basrah, Iraq. The clinical samples were collected as routine clinical care for referred and admitted patients and not for this study. Hence, the written informed consent was waived by the University of Basrah, Basrah, Iraq.
## Study design and sample collection
A total of 150 clinical specimens including wound swabs, sputum, throat swabs, nasal swabs, pus, and urine were collected from patients suffering of urinary tract infection (UTI), wound infection, and upper respiratory tract infection. These patients attended the outpatients and inpatients clinics of Alsadr Teaching Hospital and Al-Shefa General Hospital, Basrah, Iraq for a seven-month period from 1 January to 30 July 2022. All samples were collected in sterile conditions with sterile containers and transmitted to the microbiology laboratory of the College of Medicine, University of Basrah for isolation and identification of MRSA isolates.
## Bacterial isolation and identification
Each clinical sample was directly inoculated into plates of mannitol salt agar (MSA, Merck, Darmstadt, Germany) and sheep blood agar (SBA, Merck, Darmstadt, Germany) and incubated at 37 °C for 24–48 h. Then, all colonies from primary cultures were identified depending on the morphological features in culture media as beta hemolytic on blood agar and fermentation of the mannitol sugar on MSA. In addition, a panel of standard microbiology and biochemical tests including Gram staining, catalase, DNase, slide and tube coagulase were performed to confirm the S. aureus isolates [9]–[11]. S. aureus ATCC® 29213™ was used as a quality control strain.
## Identification of MRSA
In vitro detection of MRSA strains were applied by cefoxitin (30 µg) disk diffusion test and oxacillin salt agar that compromised of Mueller-Hinton agar (MHA, Merck, Darmstadt, Germany) containing 6 µg/mL of oxacillin (Sigma, USA) supplemented with $4\%$ NaCl following the Clinical and Laboratory Standards Institute (CLSI) instructions [12]. The plates were inoculated with S. aureus isolates at a concentration of 1.5 × 108 CFU/mL equal to the 0.5 McFarland tube and incubated at 35 °C for 16–18 h for cefoxitin (30 µg) disk diffusion and 24 h for oxacillin salt agar, respectively. In the cefoxitin (30 µg) disk diffusion, the isolates were considered MRSA if the inhibition zone around the disks was recorded ≤ 21 mm [12]. In the oxacillin salt agar, the existence of > 1 colony or light film of growth was considered as MRSA [12]. S. aureus ATCC® 29213™ and S. aureus ATCC® 43300™ were used as negative and positive quality control strains, respectively.
## Antibiotic susceptibility testing of MRSA isolates
Antibiotic susceptibility testing of MRSA isolates were investigated by disk diffusion test on MHA medium according to the CLSI instructions [12]. The following antibiotic disks (Mast, UK) were used: azithromycin (15 µg), norfloxacin (10 µg), erythromycin (15 µg), rifampin (5 µg), chloramphenicol (30 µg), tetracycline (30 µg), and clindamycin (2 µg).
## Preparation of starch-based ZnO-NPs
The previously described method was used to prepare the ZnO-NPs with minor modifications [13]. We examined different parameters to obtain an optimum synthesized ZnO-NPs. ZnO-NPs were prepared by the wet chemical precipitation method using the zinc nitrate 6-hydrate (Zn(NO3)2.6H2O) (Sigma-Aldrich, USA) in three concentrations (0.1 M, 0.05 M, 0.02 M) and sodium hydroxide (NaOH) (Sigma-Aldrich, USA) in concentrations of 0.2 M, 0.1 M, and 0.04 M as a precipitating agent in ratio 2:1. Also, in this method the soluble starch (Sigma-Aldrich, USA) was used as a stabilizing agent in concentrations ($0.5\%$, $0.25\%$, $0.1\%$) for each prepared concentration of precursors mention above. The different concentrations of the primary precursors and stabilizer were prepared with deionized water and stirred vigorously using a magnetic stirrer till complete dissolution. The zinc nitrate solutions were kept under constant stirring at room temperature using a magnetic stirrer for 1 hour. Next, the starch solution was added and the mixture was magnetically stirred to obtain a homogeneous solution. Then, the NaOH solution was slowly added drop by drop at room temperature under vigorous stirring, which resulted in the formation of a white precipitate of the nanoparticles. The solution was allowed to settle overnight. Then, the precipitate was separated by centrifugation (10000 g for 10 min). The produced nanoparticles were washed three times with distilled water to remove the byproducts and the excessive starch particles that were bound to the formed nanoparticles. Finally, the nanoparticles were dried at 60–80 °C for overnight [13].
## Ultraviolet–visible (UV-Vis) spectroscopy analysis
The presence of nanoparticles was proved by UV-*Vis analysis* using the Shimadzu UV-1800 UV/Visible Scanning Spectrophotometer (Shimadzu, Kyoto, Japan) in the Department of Physics, University of Basrah. This device detected the surface plasmon resonance (SPR) peak of the prepared starch-based ZnO-NPs in the scanning range of 200–800 nm. An absorbance test was conducted on 1 cm quartz cells using starch-based ZnO-NPs dispersed in deionized water [13].
## X-ray diffraction (XRD)
Crystalline structure, nature of the phase, lattice parameters, and crystalline grain size of the starch-based ZnO-NPs were evaluated using XRD type Xpert MPD (Empyrean, Malvern Panalytical, Malvern, United Kingdom) in the Department of Physics, University of Basrah. The parameters were as follows: Cu-K 1 radiation (λ = 1.5406 Å) at 40 kV and 40 mA to work in Bragg–Brentano geometry with 2θ = (20–80)°, a speed of 2 sec/step and 0.02° step, and extract analysis 2θ = (0–80)° [14].
## Field emission scanning electron microscopy (FE-SEM) and energy dispersive X-ray spectroscopy (EDS)
The surface morphology and structure (mean particle size) of starch-based ZnO-NPs were evaluated using the FEI Nova NanoSEM 450 (FEI, Hillsboro, OR, USA) equipped with energy dispersive X-ray spectroscopy (EDS) in the Department of Physics University of Basrah. Starch-based ZnO-NPs were mixed with acetone and small drops of each sample were placed on a glass slide and allowed to dry. The samples were coated with thin layers of platinum to be conductive. The device was operated at a vacuum of the order 5–10 Torr. The acceleration voltage of the device was kept in the range of 10–20 kV. In the next step, the compositional analysis of the samples was carried out by EDS attached to the FE-SEM device. EDS analysis was used to determine the elemental compositions of the synthesized ZnO-NPs.
## Transmission electron microscopy (TEM)
The prepared solutions of starch-based ZnO-NPs in distilled water (Milli-Q®, Millipore Corporation, Bedford, MA, USA) were placed on carbon-coated copper grid and allowed to dry under ambient conditions. The particle size and the shape of starch-based ZnO-NPs were observed by a TEM microscope (Tecnai G2 200 kV TEM, FEI Electron Optics) with an accelerating voltage of 200 kV [15].
## In vitro antibacterial assay of starch-based ZnO-NPs
Qualitative and quantitative assays were performed to evaluate the antibacterial effects of starch-based ZnO-NPs. The qualitative antibacterial effect of starch-based ZnO-NPs against clinical MRSA isolates was performed by standard disc diffusion method. The minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) were then determined using a broth microdilution assay as a quantitative assay.
## Disc diffusion method
For the disc diffusion method, the bacterial isolates were grown aerobically in nutrient broth for 24 hrs at 37 °C. Then, 100 µL of the bacterial suspensions (concentration equal to 2 × 108 CFU/mL) was spread on sterile MHA plates. Sterile Whatman filter paper discs (5 mm) (Sigma-Aldrich, USA) saturated with 50 mg/L of prepared starch-based ZnO-NPs in distilled water (Milli-Q®, Millipore Corporation, Bedford, MA, USA) were placed on each inoculated plate. The cultured agar plates were incubated at 37 °C for 24 h. Finally, the zones of inhibition were recorded. Distilled water was used as the negative control [15]. The experiments were performed in triplicate. Results were estimated as mean ± the standard deviations (SD) of three replicates.
## Broth microdilution assay
MIC and MBC were determined for the concentration that showed the highest antibacterial effects with the disc diffusion method. About, 100 µL of starch-based ZnO-NPs was added into a sterile 96 well microtiter plate containing 100 µL of Mueller-Hinton broth (MHB, Merck, Darmstadt, Germany) to reach serially diluted concentrations of 200 to 0.2 µg/mL. Then, 100 µL of the bacterial suspensions (concentration equal to 2 × 106 CFU/mL) was inoculated in each well to reach the concentration of 2 × 105 CFU/mL. The microplate was incubated at 37 °C for 24 h. The MIC was described as the smallest amount of starch-based ZnO-NPs that prevented MRSA growth. Re-culturing (10 µL) of wells with no visible growth was performed on the MHA medium to determine the MBC. Incubation of the MHA plates was conducted aerobically at 37 °C for 24 h. The MBC for the examined strains was based on the starch-based ZnO-NP concentration at which bacterial growth was not detected. MHB inoculated with MRSA suspension and MHB alone were used as positive and negative controls, respectively. These experiments were repeated three times and the best observation was recorded as the final result [16],[17]. An MIC90/MBC90 was defined as a MIC/MBC that inhibits/kills $90\%$ of MRSA isolates, while the MIC50/MBC50 was the MIC/MBC value that inhibits/kills $50\%$ of isolates [18].
## Data analysis
Statistical analysis of the data was performed using GraphPad Prism 9 (GraphPad Software, USA) and repeated measures ANOVA test. Data were presented as mean ± standard deviation (SD). The significant differences were considered based on the P-value < 0.05.
## S. aureus and MRSA isolates
In this study, a total of 61 ($40.7\%$) S. aureus were isolated and identified from 150 clinical samples during the survey period. The isolates that showed round Gram-positive cocci with aggregate in clusters (irregular grapes) phenotype, positive catalase test, positive slide or tube coagulase test, fermentation of mannitol on MSA medium (changing the color of MSA from pink to yellow), and beta hemolysis on SBA were selected as S. aureus isolates. The results of cefoxitin (30 µg) disk diffusion and oxacillin salt agar showed that all 61 isolates were resistant to methicillin and were considered as MRSA. The most prevalence of MRSA isolates was found in ear pus samples ($32.8\%$, $\frac{20}{61}$), followed by sputum ($29.5\%$, $\frac{18}{61}$), urine (18.0, $\frac{11}{61}$), nasopharynx ($9.8\%$, $\frac{6}{61}$), throat ($6.6\%$, $\frac{4}{61}$), and wound samples ($3.3\%$, $\frac{2}{61}$) (Table 1).
**Table 1.**
| Type of clinical specimens | Methicillin-resistant/Number | Staphylococcus aureus isolates/% |
| --- | --- | --- |
| Wound | 2 | 3.3 |
| Sputum | 18 | 29.5 |
| Throat | 4 | 6.6 |
| Nasopharynx | 6 | 9.8 |
| Ear pus | 20 | 32.8 |
| Urine | 11 | 18.0 |
| Total | 61 | 100.0 |
## Antibiotic resistance patterns of MRSA isolates
The antibiotic resistance patterns of the MRSA isolates were shown in the Table 2. Accordingly, the most and the less resistance rates were against rifampin ($60.7\%$) and chloramphenicol ($11.5\%$), respectively.
**Table 2.**
| Antibiotics | Resistance/ N (%) | Susceptible/ N (%) |
| --- | --- | --- |
| Oxacillin (1 µg) | 61 (100) | __ |
| Cefoxitin (30 µg) | 61 (100) | __ |
| Clindamycin (2 µg) | 12 (19.7) | 49 (80.3) |
| Tetracycline (5 µg) | 18 (29.5) | 43 (70.5) |
| Chloramphenicol (30 µg) | 7 (11.5) | 54 (88.5) |
| Rifampin (5 µg) | 37 (60.7) | 24 (39.3) |
| Erythromycin (15 µg) | 21 (34.4) | 40 (65.6) |
| Norfloxacin (10 µg) | 21 (34.4) | 40 (65.6) |
| Azithromycin (15 µg) | 29 (47.5) | 32 (52.5) |
## Characterization of starch-based ZnO-NPs
The pour and dried white powder of 3 different concentrations (0.1 M, 0.05 M, 0.02 M) of the starch-based ZnO-NPs is shown in Figure 1.
**Figure 1.:** *The pour and dried white powder of 3 different concentrations (1: 0.1 M, 2: 0.05 M, 3: 0.02 M) of the synthesized zinc oxide nanoparticles (ZnO-NPs).*
## UV-Vis analysis
The UV-*Vis spectra* of the different three concentrations of starch-based ZnO-NPs (0.1 M, 0.05 M, 0.02 M) that prepared with $0.5\%$, $0.25\%$ and $0.1\%$ of soluble starch were shown in Figure 2. The three different concentrations of starch-based ZnO-NPs exhibited a strong absorption band in the region below 400 nm (at 360 nm) that was the characteristic for the ZnO-NPs (Figure 2).
**Figure 2.:** *Ultraviolet–visible (UV-Vis) spectroscopy analysis of the three different concentrations (C1: 0.1 M, C2: 0.05 M, C3: 0.02 M) of the zinc oxide nanoparticles (ZnO-NPs) that prepared with 0.5%, 0.25% and 0.1% of soluble starch, respectively.*
## XRD analysis
XRD patterns of the three concentrations of starch-based ZnO-NPs were shown in Figure 3. All diffraction peaks were obtained at 2θ values of 31.7°, 34.4°, 36.2°, 47.5°, 56.6°, 62.8°, 66.3°, 67.9° and 72.5° corresponding to [100], [002], [101], [102], [110], [103], [200] and [112] orientation planes, confirming the representative hexagonal wurtzite phase of the ZnO-NPs. The XRD spectra did not exhibit additional peaks associated with impurities, suggesting the high purity of the starch-based ZnO-NPs. Also, as evident from Figure 3, the signal sharpness indicated the high crystallinity of the starch-based ZnO-NPs. There were no differences in XDR patterns of different concentrations of ZnO-NPs as all of them showed the diffraction peaks at 2θ values of 31.7°, 34.4°, 36.2°, 47.5°, 56.6°, 62.8°, 66.3°, 67.9° and 72.5°.
**Figure 3.:** *X-ray diffraction (XRD) of the three different concentrations (C1: 0.1 M, C2: 0.05 M, C3: 0.02 M) of the zinc oxide nanoparticles (ZnO-NPs) that prepared with 0.5%, 0.25% and 0.1% of soluble starch, respectively.*
## FE-SEM and TEM analysis
The morphology of the starch-based ZnO-NPs was investigated by the FE-SEM and TEM as shown in Figure 4 A and B, respectively. The diameter of the ZnO-NPs was in the range of 18.47 to 25.19 (mean ± SD = 21.56 ± 3.09 and 22.87 ± 2.32 nm by FE-SEM and TEM, respectively). Starch-based ZnO-NPs displayed spherical morphology as shown by FE-SEM and TEM images. Also, a smooth surface was generally present on the particles, with uniform sizes and shapes. There were no significant differences (P-value > 0.05) in the mean ± SD of the size of three concentrations of the starch-based ZnO-NPs confirming the similarity of their size. Also, the shape of all synthesized ZnO-NPs showed spherical morphology confirming their shape similarity.
**Figure 4.:** *A: Field emission scanning electron microscopy (FE-SEM) morphology of the zinc oxide nanoparticles (ZnO-NPs). B: Transmission electron microscopy (TEM) morphology of the zinc oxide nanoparticles (ZnO-NPs).*
EDS plots of the FE-SEM of three concentrations of starch-based ZnO-NPs were presented in Figures 5, 6, and 7. EDS analysis confirmed the presence of zinc (Zn) ($54.22\%$) and oxygen (O) ($9.68\%$) in C1 (0.1 M) concentration (Figure 5). Non-intentional dopants including Na ($19.65\%$), Si ($0.81\%$), Br ($7.43\%$), Cu ($6.32\%$), and N ($1.89\%$) elements were also detected (Figure 5). This was probably due to the presence of substrate over which the ZnO-NPs samples were held for analysis. Also the zinc (Zn) ($63.47\%$) and oxygen (O) ($8.98\%$) were found in C2 (0.05 M) concentration (Figure 6). Non-intentional dopants including Na ($24.25\%$), Alu ($0.21\%$), and Br ($3.08\%$) elements were also detected (Figure 6). Meanwhile, the zinc (Zn) ($27.4\%$) and oxygen (O) ($5.39\%$) were found in C3 (0.02 M) concentration (Figure 7). Non-intentional dopants including Cu ($62.86\%$) as the major element, Alu ($3.61\%$), and N ($1.11\%$) elements were also detected (Figure 7). The carbon was not detected in any synthesized nanoparticles because the precipitates were produced from the reaction to obtain ZnO-NPs nanoparticles were separated by centrifugation at 10000 g for 10 min. Then, the produced nanoparticles were washed three times with distilled water to remove the byproducts and starch particles that were bound to the formed nanoparticles because the starch used in this method was a stabilizing agent and when the reaction was complete, we exclude it from the formed nanoparticles.
**Figure 5.:** *Energy dispersive X-ray spectroscopy (EDS) plot of the field emission scanning electron microscopy (FE-SEM) of C1 (0.1 M) starch-based ZnO-NPs.* **Figure 6.:** *Energy dispersive X-ray spectroscopy (EDS) plot of the field emission scanning electron microscopy (FE-SEM) of C2 (0.05 M) starch-based ZnO-NPs.* **Figure 7.:** *Energy dispersive X-ray spectroscopy (EDS) plot of the field emission scanning electron microscopy (FE-SEM) of C3 (0.02 M) starch-based ZnO-NPs.*
## In vitro antibacterial effects of the starch-based ZnO-NPs
The results of the disc diffusion method showed that all concentrations of the starch-based ZnO-NPs had inhibitory effects on MRSA isolates. The 0.1 M concentration had the highest antibacterial effects with the mean ± SD of the inhibition zone of 17.62 ± 2.65 mm followed by the 0.05 M concentration with an inhibition zone of 16.03 ± 2.24 mm and the 0.02 M concentration with an inhibition zone of 12.7 ± 2.57 mm (Table 3). The MIC of the 0.1 M concentration was in the range of 25–50 µg/mL, while the MBC was in the range of 50–100 µg/mL. Also, the MIC90/MBC90 and the MIC50/MBC50 were $\frac{50}{100}$ µg/mL and $\frac{25}{50}$ µg/mL, respectively (Table 3).
**Table 3.**
| Antibacterial effects | Concentrations of ZnO-NPs | Concentrations of ZnO-NPs.1 | Concentrations of ZnO-NPs.2 |
| --- | --- | --- | --- |
| Antibacterial effects | C1 (0.1 Mol) | C2 (0.05 Mol) | C3 (0.02 Mol.) |
| Inhibition zones of synthesized ZnO-NPs (Mean ± SD) (mm) | 2.65 ± 17.62 | 16.03 ± 2.24 | 12.7 ± 2.57 |
| MIC of 0.1 M concentration | 25–50 µg/mL | | |
| MBC of 0.1 M concentration | 50–100 µg/mL | | |
| MIC90/MBC90 of 0.1 M concentration | 50/100 µg/mL | | |
| MIC50/MBC50 of 0.1 M concentration | 25/50 µg/mL | | |
The 0.1 M and 0.05 M concentrations of the starch-based ZnO-NPs showed significantly greater inhibition zones against MRSA isolates compared to 0.02 M concentration (P-value = 0.0001). Likewise, the inhibition zones of the 0.1 M concentration were significantly greater than those of 0.05 M ZnO-NPs (Figure 8).
**Figure 8.:** *Inhibition zones of different concentrations (0.1 M, 0.05 M, 0.02 M) of zinc oxide nanoparticles against methicillin resistance Staphylococcus aureus isolates. The statistically significant differences were according to repeated measures ANOVA test (P-value < 0.05), **** = P < 0.0001.*
## Discussion
Various microbes have been prevented from growing on humans due to the use of zinc salt for decades. Also, there are extensive studies demonstrating the effectiveness of ZnO-NPs against pathogenic bacteria including E. coli and S. aureus [5],[8],[19]. However, the antibacterial effects of starch-based ZnO-NPs on MRSA isolates from *Iraq is* lacking. In this study, 61 MRSA isolates were collected from 150 different clinical samples of Iraqi patients, confirming the prevalence of $40.7\%$. This prevalence rate of MRSA was lower than previous reports from Iran ($78.9\%$) [11] and Iraq ($53.1\%$) [20], and was higher than studies from Italy ($1.1\%$) [21] and Ghana ($17.1\%$) [22]. Differences in prevalence rates may be explained by the differences in bacteria detection methods, examined populations, and studied sample types and sizes in various countries. The MRSA isolates showed relatively high resistance rates against azithromycin and rifampin (more than $40.0\%$), while the other antibiotics including chloramphenicol, tetracycline, clindamycin, norfloxacin, and erythromycin were more effective with resistance rates below $35.0\%$. In comparison to this study, previous research from Iran [11], found a higher resistance rate to azithromycin ($100\%$) and erythromycin ($98.3\%$) among MRSA isolates. However, in a previous study from Fiji [23], MRSA isolates showed a significantly lower resistance rate against clindamycin ($0.0\%$), rifampicin ($0.0\%$), and tetracycline ($12.0\%$) that was in contrast to this study. These differences may be explained by the variations in the patients' demographics and geographical location that influence the resistance rates.
In this study, the qualitative and quantitative antibacterial assays showed promising effects of all synthesized starch-based ZnO-NPs against all MRSA isolates. These observations were in line with the previous studies from Egypt [5], Iraq [8], and Iran [16], in which the strong inhibitory effects of ZnO-NPs were found on multidrug-resistant S. aureus. The highest rate of inhibition was found at 0.1 M concentration with the mean ± SD of the inhibition zone of 17.62 ± 2.65 mm followed by the 0.05 M concentration (16.03 ± 2.24 mm) and the 0.02 M concentration (12.7 ± 2.57 mm). In previous studies, ZnO-NPs showed inhibition zones of 73.95 ± $2.17\%$ at 10 mg/mL against vancomycin-resistant S. aureus (VRSA) and 16–21 mm against various Gram-negative and Gram-positive bacteria [8],[17]. In another study by Kamarajan et al. [ 24] from India, ZnO-NPs at a concentration of 10 µg/mL showed inhibitory effects against *Escherichia coli* (25 mm), *Pseudomonas aeruginosa* (23 mm), S. aureus (22 mm), and *Bacillus subtilis* (21 mm). The discrepancies in the inhibitory zone size in different studies may be due to the bacteria studied, shape, size, concentrations of the synthesized ZnO-NPs, and method used to synthesize ZnO-NPs.
Previous studies have found that the shape, size, concentrations of the synthesized ZnO-NPs, and the method to synthesize ZnO-NPs affect the antibacterial properties of the nanoparticles [24]–[27]. In this study, the highest concentrations of the starch-based ZnO-NPs exhibited significantly greater inhibition zones compared to the lowest concentrations. These results were consistent with the previous studies in which higher concentrations of ZnO-NPs showed stronger antimicrobial effects [8],[25],[26]. However, some studies showed that the inhibition zone of nanoparticles starts to shrink beyond an optimal concentration [28],[29]. One of the possible reasons may be due to the accumulation of nanoparticles in high concentrations and the inability to penetrate into bacterial cells [28],[29].
ZnO-NPs are believed to act in four distinct ways including releasing Zn2+ ions, damaging the cell wall, producing reactive oxygen species (ROS), and by ZnO-NPs internalizing [25]. The antibacterial activity of ZnO-NPs depends on their penetration into bacterial cells. Thus, the antibacterial effects of ZnO-NPs can be evaluated by the broth dilution method as a precise and confirmative assay [5]. In this study, the broth microdilution assay revealed the MIC of the starch-based ZnO-NPs in the range of 25 to 50 µg/mL at the 0.1 M concentration. Also, the MBC was in the range of 50 to 100 µg/mL. These values were lower than a previous study that reported ZnO-NP MICs ranging from 128 to 2048 µg/mL against S. aureus isolates [5]. MIC values in this study were also lower than those reported by Jasim et al. [ 8] against VRSA isolates (625 µg/mL). However, in a previous study by Tănase et al. [ 14] from Romania, the chemical and *Saponaria officinalis* extract-mediated ZnO-NPs showed lower MICs (<20 µg/mL) against standard strains of S. aureus, P. aeruginosa, E. coli, and Candida albicans. The differences among studies may be due to the used methodology, the antibiotic resistance patterns of examined bacteria, and the structural nature of the synthesized ZnO-NPs.
In this study, the structural nature of the starch-based ZnO-NPs were investigated by various methods. The UV-*Vis analysis* showed that three concentrations of starch-based ZnO-NPs (0.1 M, 0.05 M, 0.02 M) exhibited a strong absorption band at 360 nm which was characteristic of the ZnO-NPs. This observation was in good parallel with the previous studies from Egypt [13], Romania [14], and Jordan [30] which showed the absorption peaks of ZnO-NPs below 400 nm. The shape, size, and method of fabrication of ZnO-NPs are all factors influencing the absorption peak. *In* general, ZnO-NPs exhibit a UV-Vis spectroscopic peak between 350 and 390 nm [30]. Moreover, the XRD analysis of the three concentrations of the starch-based ZnO-NPs confirmed the representative hexagonal wurtzite phase, high purity, and high crystallinity of the starch-based ZnO-NPs. These results were consistent with the previous observations from Malaysia [15] and Jordan [30]. Another observation of this study was the spherical morphology of the starch-based ZnO-NPs with a diameter of 21.56 ± 3.42 and 22.87 ± 3.91 nm by FE-SEM and TEM, respectively. There was no any significant difference among three concentrations of the starch-based ZnO-NPs in terms of size and shape. In contrast to this study, Saleemi et al. [ 15], showed rod-shaped morphology of standard ZnO-NPs with the diameter of 49.39 ± 22.54 nm [15]. Alshraiedeh et al. [ 30] reported spherical ZnO-NPs with a size of 100 nm in their study.
From a future perspective, it is recommended to examine the synergistic effects of the synthesized starch-based ZnO-NPs in combination with standard antibiotics or other chemical or plant-based materials against different pathogens and cancer cell lines. Previous studies have shown the significant effects of combining nanoparticles with other materials against microorganisms [31],[32]. Although several studies have investigated the antibacterial effects of ZnO-NPs against MRSA isolates, but in each of them, the different researchers looked forward to finding more effective nanoparticles in terms of their shape, quality, and antibacterial effects. The novelty of this study was the synthetization of relatively smaller nanoparticles in comparison to previous studies. Also, the synthesized nanoparticles showed promising antibacterial effects in low concentrations (0.02 Mol). However, this study had several limitations as follows: lack of investigation of starch-based ZnO-NPs against other Gram-positive and Gram-negative bacteria, lack of in vivo experiment, and lack of time-kill kinetics assay.
## Conclusion
UV-Vis, XRD, FE-SEM, and TEM analysis showed the crystalline organization, spherical shape, and smooth surface of the starch-based ZnO-NPs with a size below 27 nm. Qualitative and quantitative antimicrobial assays showed the promising effects of the starch-based ZnO-NPs against clinical MRSA isolates with MIC ranging from 25–50 µg/mL. Further in vivo experiment is needed to reveal the mechanism of action of the synthesized starch-based ZnO-NPs.
## References
1. Bier K, Schittek B. **Beneficial effects of coagulase-negative staphylococci on**. *Exp Dermatol* (2021) **30** 1442-1452. DOI: 10.1111/exd.14381
2. Turner NA, Sharma-Kuinkel BK, Maskarinec SA. **Methicillin-resistant**. *Nat Rev Microbiol* (2019) **17** 203-218. PMID: 30737488
3. Masters EA, Ricciardi BF, Bentley KL. **Skeletal infections: microbial pathogenesis, immunity and clinical management**. *Nat Rev Microbiol* (2022) **20** 385-400. DOI: 10.1038/s41579-022-00686-0
4. Pal S, Sayana A, Joshi A. *J Family Med Prim Care* (2019) **8** 3600-3606. DOI: 10.4103/jfmpc.jfmpc_521_19
5. Abdelraheem WM, Khairy RMM, Zaki AI. **Effect of ZnO nanoparticles on methicillin, vancomycin, linezolid resistance and biofilm formation in**. *Ann Clin Microbiol Antimicrob* (2021) **20** 54. DOI: 10.1186/s12941-021-00459-2
6. Milheiriço C, Tomasz A, de Lencastre H. **Impact of the stringent stress response on the expression of methicillin resistance in**. *Antibiotics* (2022) **11** 255. DOI: 10.3390/antibiotics11020255
7. Chen CJ, Huang YC, Shie SS. **Evolution of multi-resistance to vancomycin, daptomycin, and linezolid in methicillin-resistant**. *Front Microbiol* (2020) **11** 1414. DOI: 10.3389/fmicb.2020.01414
8. Jasim NA, Al-Gasha'a FA, Al-Marjani MF. **ZnO nanoparticles inhibit growth and biofilm formation of vancomycin-resistant**. *Biocatal Agric Biotechnol* (2020) **29** 101745. DOI: 10.1016/j.bcab.2020.101745
9. Abbasi Montazeri E, Seyed-Mohammadi S, Asarehzadegan Dezfuli A. **Investigation of SCC**. *Biosci Rep* (2020) **40** BSR20200847. DOI: 10.1042/BSR20200847
10. Gadban TH, Al-Amara SS, Jasim HA. **Screening the frequency of panton-valentine leukocidin (**. *Sys Rev Pharma* (2020) **11** 285-290. DOI: 10.31838/srp.2020.11.42
11. Khoshnood S, Shahi F, Jomehzadeh N. **Distribution of genes encoding resistance to macrolides, lincosamides, and streptogramins among methicillin-resistant**. *Acta Microbiol Immunol Hung* (2019) **66** 387-398. DOI: 10.1556/030.66.2019.015
12. 12
Clinical and Laboratory Standards Institute (CLSI)
Performance standards for antimicrobial susceptibility testing
31st Ed.
Malvern
2021. *Performance standards for antimicrobial susceptibility testing* (2021)
13. Hassanein TF, Mohammed AS, Mohamed W. **Optimized synthesis of biopolymer-based zinc oxide nanoparticles and evaluation of their antibacterial activity**. *Egypt J Chem* (2021) **64** 3767-3790. DOI: 10.21608/EJCHEM.2021.75677.3709
14. Tănase MA, Marinescu M, Oancea P. **Antibacterial and photocatalytic properties of ZnO nanoparticles obtained from chemical versus**. *Molecules* (2021) **26** 2072. DOI: 10.3390/molecules26072072
15. Saleemi MA, Alallam B, Yong YK. **Synthesis of zinc oxide nanoparticles with bioflavonoid rutin: characterisation, antioxidant and antimicrobial activities and**. *Antioxidants* (2022) **11** 1853. DOI: 10.3390/antiox11101853
16. Shakerimoghaddam A, Razavi D, Rahvar F. **Evaluate the effect of zinc oxide and silver nanoparticles on biofilm and**. *J Burn Care Res* (2020) **41** 1253-1259. DOI: 10.1093/jbcr/iraa085
17. Ahmad I, Alshahrani MY, Wahab S. **Zinc oxide nanoparticle: an effective antibacterial agent against pathogenic bacterial isolates**. *J King Saud Univ Sci* (2022) **34** 102110. DOI: 10.1016/j.jksus.2022.102110
18. Bashir MH, Hollingsworth A, Thompson JD. **Antimicrobial performance of two preoperative skin preparation solutions containing iodine and isopropyl alcohol**. *Am J Infect Control* (2022) **50** 792-798. DOI: 10.1016/j.ajic.2021.10.031
19. Irfan M, Munir H, Ismail H. *Biomater Res* (2021) **25** 17. DOI: 10.1186/s40824-021-00219-5
20. Awayid HS, Mohammad SQ. **Prevalence and antibiotic resistance pattern of methicillin-resistant**. *Arch Razi Inst* (2022) **77** 1147-1156. DOI: 10.22092/ARI.2022.357391.2031
21. Mascaro V, Squillace L, Nobile CG. **Prevalence of methicillin-resistant**. *Infect Drug Resist* (2019) **12** 2561-2571. DOI: 10.2147/IDR.S211629
22. Kotey FC, Awugah SA, Dayie NT. **High prevalence of methicillin-resistant**. *J Infect Dev Ctries* (2022) **16** 1450-1457. DOI: 10.3855/jidc.14839
23. Loftus MJ, Young-Sharma TE, Wati S. **Epidemiology, antimicrobial resistance and outcomes of**. *Lancet Reg Health West Pac* (2022) **22** 100438. DOI: 10.1016/j.lanwpc.2022.100438
24. Kamarajan D, Anburaj B, Porkalai V. **Green synthesis of ZnO nanoparticles and their photocatalyst degradation and antibacterial activity**. *J Water Environ Nanotechnol* (2022) **7** 180-193. DOI: 10.22090/jwent.2022.02.006
25. Babayevska N, Przysiecka Ł, Iatsunskyi I. **ZnO size and shape effect on antibacterial activity and cytotoxicity profile**. *Sci Rep* (2022) **12** 8148. DOI: 10.1038/s41598-022-12134-3
26. Djearamane S, Loh ZC, Lee JJ. **Remedial aspect of zinc oxide nanoparticles against**. *Front Pharmacol* (2022) **13** 891304. DOI: 10.3389/fphar.2022.891304
27. Alnehia A, Al-Odayni AB, Al-Sharabi A. *J Chem* (2022) **2022** 9647793. DOI: 10.1155/2022/9647793
28. Abass AA, Alaarage WK, Abdulrudha NH. **Evaluating the antibacterial effect of cobalt nanoparticles against multi-drug resistant pathogens**. *J Med Life* (2021) **14** 823-833. DOI: 10.25122/jml-2021-0270
29. Gupta V, Kant V, Sharma AK. **Comparative assessment of antibacterial efficacy for cobalt nanoparticles, bulk cobalt and standard antibiotics: a concentration dependent study**. *Nanosystems Phys Chem Math* (2020) **11** 78-85. DOI: 10.17586/2220-8054-2020-11-1-78-85
30. Alshraiedeh NA, Ammar OF, Masadeh MM. **Comparative study of antibacterial activity of different ZnO nanoparticles, nanoflowers, and nanoflakes**. *Curr Nanosci* (2022) **18** 758-765. DOI: 10.2174/1573413718666220303153123
31. Al-Mosawi RM. **The effectiveness study of altered glass ionomer cement with ZnO and TiO**. *Trends Pharm Nanotechnol* (2020) **2** 52-58. DOI: 10.46610/TPNT.2020.v02i02.004
32. Al-Mosawi RM, Al-Badr RM. **The study effects of dental composite resin as antibacterial agent which contain nanoparticles of zinc oxide on the bacteria associated with oral infection**. *J Dent Med Sci* (2017) **16** 49-55. DOI: 10.9790/0853-1601014955
|
---
title: Mendelian randomization and clinical trial evidence supports TYK2 inhibition
as a therapeutic target for autoimmune diseases
authors:
- Shuai Yuan
- Lijuan Wang
- Han Zhang
- Fengzhe Xu
- Xuan Zhou
- Lili Yu
- Jing Sun
- Jie Chen
- Haochao Ying
- Xiaolin Xu
- Yongfu Yu
- Athina Spiliopoulou
- Xia Shen
- Jim Wilson
- Dipender Gill
- Evropi Theodoratou
- Susanna C. Larsson
- Xue Li
journal: eBioMedicine
year: 2023
pmcid: PMC9988426
doi: 10.1016/j.ebiom.2023.104488
license: CC BY 4.0
---
# Mendelian randomization and clinical trial evidence supports TYK2 inhibition as a therapeutic target for autoimmune diseases
## Body
Research in contextEvidence before this studyDeucravacitinib is a selective inhibitor of tyrosine kinase 2 (TYK2) and has been approved to treat moderate-to-severe plaque psoriasis. TYK2 belongs to the Janus kinase family that exerts effects on a wide range of inflammatory disorders. Thus, TYK2 inhibitors may have the potential in the treatment for autoimmune diseases. However, relatively few clinical trials on autoimmune diseases except psoriasis hinder the assessment of the effectiveness of TYK2 inhibitor treatment on autoimmune diseases. In addition, Janus kinase inhibitors have been associated with increased risk of serious heart-related events and certain cancers, which similarly raises concerns on their safety. No studies have been conducted to systematically explore the possible adverse effects of TYK2 inhibitor. Added value of this studyThis comprehensive study found evidence supporting the efficacy of TYK2 inhibitors for psoriasis and its related disorders. There were Mendelian randomization associations of the TYK2 loss-of-function variant with hypothyroidism, inflammatory bowel disease, primary biliary cirrhosis, and type 1 diabetes. Although only a few clinical trials supported that TYK2 inhibitors appeared to improve disease activity among patients with ulcerative colitis, alopecia areata, atopic dermatitis, or active non-segmental vitiligo, these findings need to be confirmed in larger studies, especially for ulcerative colitis, for which there was conflicting evidence in previous trials. The study identified several potential adverse effects of TYK2 inhibitors, including headache, upper respiratory tract infection, nausea, diarrheal, increased circulating levels of creatinine and liver enzymes, and risk of certain malignant neoplasms, such prostate and breast cancer. Implications of all the available evidenceTYK2 inhibitors may be used to treat psoriasis and possibly other autoimmune diseases, like hypothyroidism, inflammatory bowel disease, primary biliary cirrhosis, and type 1 diabetes. The side effects of TYK2 inhibitors should be assessed, especially on prostate and breast cancer.
## Summary
### Background
To explore the associations of genetically proxied TYK2 inhibition with a wide range of disease outcomes and biomarkers to identify therapeutic repurposing opportunities, adverse effects, and biomarkers of efficacy.
### Methods
The loss-of-function missense variant rs34536443 in TYK2 gene was used as a genetic instrument to proxy the effect of TYK2 inhibition. A phenome-wide Mendelian randomization (MR) study was conducted to explore the associations of genetically-proxied TYK2 inhibition with 1473 disease outcomes in UK Biobank ($$n = 339$$,197). Identified associations were examined for replication in FinnGen ($$n = 260$$,405). We further performed tissue-specific gene expression MR, colocalization analyses, and MR with 247 blood biomarkers. A systematic review of randomized controlled trials (RCTs) on TYK2 inhibitor was performed to complement the genetic evidence.
### Findings
PheWAS-MR found that genetically-proxied TYK2 inhibition was associated with lower risk of a wide range of autoimmune diseases. The associations with hypothyroidism and psoriasis were confirmed in MR analysis of tissue-specific TYK2 gene expression and the associations with systemic lupus erythematosus, psoriasis, and rheumatoid arthritis were observed in colocalization analysis. There were nominal associations of genetically-proxied TYK2 inhibition with increased risk of prostate and breast cancer but not in tissue-specific expression MR or colocalization analyses. Thirty-seven blood biomarkers were associated with the TYK2 loss-of-function mutation. Evidence from RCTs confirmed the effectiveness of TYK2 inhibitors on plaque psoriasis and reported several adverse effects.
### Interpretation
This study supports TYK2 inhibitor as a potential treatment for psoriasis and several other autoimmune diseases. Increased pharmacovigilance is warranted in relation to the potential adverse effects.
### Funding
None.
## Evidence before this study
Deucravacitinib is a selective inhibitor of tyrosine kinase 2 (TYK2) and has been approved to treat moderate-to-severe plaque psoriasis. TYK2 belongs to the Janus kinase family that exerts effects on a wide range of inflammatory disorders. Thus, TYK2 inhibitors may have the potential in the treatment for autoimmune diseases. However, relatively few clinical trials on autoimmune diseases except psoriasis hinder the assessment of the effectiveness of TYK2 inhibitor treatment on autoimmune diseases. In addition, Janus kinase inhibitors have been associated with increased risk of serious heart-related events and certain cancers, which similarly raises concerns on their safety. No studies have been conducted to systematically explore the possible adverse effects of TYK2 inhibitor.
## Added value of this study
This comprehensive study found evidence supporting the efficacy of TYK2 inhibitors for psoriasis and its related disorders. There were Mendelian randomization associations of the TYK2 loss-of-function variant with hypothyroidism, inflammatory bowel disease, primary biliary cirrhosis, and type 1 diabetes. Although only a few clinical trials supported that TYK2 inhibitors appeared to improve disease activity among patients with ulcerative colitis, alopecia areata, atopic dermatitis, or active non-segmental vitiligo, these findings need to be confirmed in larger studies, especially for ulcerative colitis, for which there was conflicting evidence in previous trials. The study identified several potential adverse effects of TYK2 inhibitors, including headache, upper respiratory tract infection, nausea, diarrheal, increased circulating levels of creatinine and liver enzymes, and risk of certain malignant neoplasms, such prostate and breast cancer.
## Implications of all the available evidence
TYK2 inhibitors may be used to treat psoriasis and possibly other autoimmune diseases, like hypothyroidism, inflammatory bowel disease, primary biliary cirrhosis, and type 1 diabetes. The side effects of TYK2 inhibitors should be assessed, especially on prostate and breast cancer.
## Introduction
Deucravacitinib, a selective inhibitor of tyrosine kinase 2 (TYK2), has been approved to treat moderate-to-severe plaque psoriasis.1,2 Given that TYK2 belongs to the Janus kinase (JAK) family that exerts effects on a wide range of inflammatory disorders, TYK2 inhibitors may have the potential in the treatment for other autoimmune diseases, such as inflammatory bowel disease,3 rheumatoid arthritis,4 and type 1 diabetes.5 However, relatively few clinical trials on these outcomes hinder the assessment of the effectiveness of TYK2 inhibitor treatment on autoimmune diseases beyond plaque psoriasis.6,7 In addition, three JAK inhibitors have been recently associated with increased risk of serious heart-related events and certain cancers,8 which similarly raises concerns on their safety. A recent Mendelian randomization (MR) study observed positive associations of a TYK2 loss-of-function mutation that mimic TYK2 inhibition with increased risk of lung cancer, non-Hodgkin lymphoma, and possibly prostate cancer.9 However, no studies have been conducted to systematically explore the possible adverse effects of inhibiting this drug target.
In the absence of long-term randomized controlled trials (RCTs) investigating TYK2 inhibition, MR analysis can be used to assess the effectiveness, repurposing potential, and safety of TYK2 inhibition by utilizing genetic variants in the TYK2 gene that reduce its function as instrumental variables for life-time TYK2 inhibition.10,11 Resembling the RCT study design, the MR approach naturally randomizes participants into groups based on genetically predicted drug target perturbation, and thus diminishes confounding effects from environmental factors since genetic variants are randomly assorted at conception. In addition, this approach can minimize reverse causality as the onset and progression of disease cannot modify the germline genotype. Here, we performed an MR investigation to comprehensively explore disease and biomarker phenotypes associated with a TYK2 loss-of-function genetic variant. To strengthen and complement the MR results, we performed a review of RCTs on TYK2 inhibition to investigate the effectiveness and safety of this drug.
## Study design and ethics permit
The study design overview is presented in Fig. 1. We firstly performed a phenome-wide association study (PheWAS) to comprehensively examine the associations of the loss-of-function mutation in the TYK2 gene with disease outcomes in the UK Biobank study. We then conducted a Mendelian randomization (MR) analysis in the FinnGen study with the aim of replicating the identified PheWAS associations. To further investigate the evidence for causality, tissue-specific gene expression and colocalization analyses were performed to examine the associations between TYK2 gene expression on certain tissue and risk of diseases highlighted in PheWAS-MR. We also explored the MR associations of TYK2 with a wide range of biomarkers, including haematological, biochemical, metabolomic, inflammatory, and immunological traits in data from phenotype-specific genetic consortia and performed mediation analysis of pathophysiological mechanisms pathways from TYK2 inhibition to disease outcomes. Finally, we collected data on published RCTs on TYK2 inhibition to complement the genetic evidence of possible clinical effects. UK Biobank received ethical permits from the Northwest Multi-centre Research Ethics Committee, the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland. All participants provided written informed consent. Fig. 1Study design overview.
## Phenome-wide association study of TYK2 mutation in the UK Biobank
PheWAS analysis of the loss-of-function mutation in TYK2 gene was performed in the UK Biobank study, an ongoing cohort study collecting phenotypic and genetic data from over 500,000 individuals since its initiation in 2006–2010. After removal of participants of other descents to minimize population bias, the current study was based on data from 339,197 (182,072 females and 157,125 males) unrelated White British individuals. Health outcomes were defined by using the PheCODE schema with diagnostic codes (10,750 unique ICD-10 codes and 3113 ICD-9 codes) from national medical records (inpatient hospital episode records, cancer registry, and death registry).12 The PheCODE system provides a scheme to automatically exclude patients that have similar or potentially overlapping disease states from the corresponding control group. We used the International Classification of Diseases (ICD) versions 9 and 10 to identify cases in the medical records, with both incident and prevalent cases included. A map matching ICD-9 and -10 codes to phecodes was used, as previously described (https://phewascatalog.org/phecodes_icd10).13 Detailed information on genotyping and quality control is described in our previous studies.14,15
## Validating PheWAS associations in the FinnGen Biobank
For phenotypes reaching statistical significance after FDR correction in the original PheWAS analysis, we further examined associations with the missense variant rs34536443 of the TYK2 gene in the FinnGen ($$n = 260$$,405) study. The FinnGen study is a growing project combining germline genotype data from Finnish biobanks and health record data on clinically defined outcomes from Finnish health registries in up to 260,405 individuals.16 We performed an MR study in R6 release of the FinnGen study to investigate replication of the identified PheWAS associations (https://finngen.gitbook.io/documentation/).
## Tissue-specific TYK2 expression and related disease outcomes
We carried out tissue-specific expression analysis of TYK2 gene to examine the associations between gene expression levels and related health outcomes identified from the loss-of-function PheWAS analysis, using the PrediXcan software.17 *The analysis* was based on the same sample from UK Biobank as for PheWAS. PrediXcan first uses reference transcriptome datasets to train additive models of gene expression levels, providing the effect sizes of single nucleotide polymorphisms (SNPs) on gene expression (i.e., prediction weights). We used expression weights from 45 tissues in the genotype-tissue expression (GTEx) database18 as reference panels and the prepackaged expression weights can be downloaded directly from the PredictDB data repository. Then, PrediXcan imputed the genetic component of expression by integrating genotype data from large-scale genome-wide association studies (GWASs) and prediction weights from the training sets while accounting for linkage disequilibrium among SNPs. Last, PrediXcan correlates the genetically predicted gene expression with the disease phenotypes using logistic regression methods. We applied a Benjamini-Hochberg correction to account for multiple testing in each tissue and associations with FDR <0.05 were considered as statistically significant.
## Colocalization analysis of TYK2 gene tissue-specific expression with disease outcomes
To further investigate causality of observed MR associations, we performed colocalization analysis of TYK2 gene tissue-specific expression (eQTL) with risk of common autoimmune diseases (including psoriasis,19 rheumatoid arthritis,20 inflammatory bowel disease,21 systemic lupus erythematosus,22 multiple sclerosis,23 and type 1 diabetes24) and related cancers (prostate25 and breast26 cancers) with publicly available genome-wide association data. This colocalization analysis can infer whether TYK2 expression and the risk of above autoimmune disease are affected by the same genetic variant. SNPs in TYK2 gene region ±1000 kb were used as instruments. Data on TYK2 expression in different tissues were obtained from the GTEx database.18 We additionally used data on TYK2 expression in whole blood from the eQTLGen dataset.27 Summary-level data on the associations of used SNPs with the outcomes were obtained from above cited GWASs. We used coloc method to obtain posterior probability for 5 hypotheses (H0–H4) in a Bayesian framework.28 PP.H4 <$80\%$ of the colocalization analysis (H4) indicates absence of strong support for a shared causal variant affecting gene expression and disease risk. We also applied the Sum of Single Effects (SuSiE) colocalization method that allows multiple signals to be distinguished to filter out linkage disequilibrium-contaminated associations.29 The analyses were performed using the default priors (p1 = 1 × 10−4, p2 = 1 × 10−4, and p12 = 1 × 10−5). F statistics were estimated for each eQTL signal across tissues. The analyses were performed using coloc 5.1 package in R 3.5.1.30
## Biomarker-wide association and mediation analyses
We obtained association estimates of the loss-of-function mutation of TYK2 gene with the following biomarkers: (i) 25 serum and urine biomarkers available in the biochemistry panel of the UK Biobank (353,579 individuals); (ii) 36 haematological traits with data derived from the summary statistics of the study by Astle et al. ( 173,480 European individuals)31; (iii) 122 nuclear magnetic resonance-measured serum lipids and metabolites with data derived from the publicly available summary statistics provided by Kettunen et al. ( 24,925 individuals of European ancestry)32; (iv) circulating levels of 41 cytokines and growth factors with data derived from the publicly available summary statistics by Ahola-Olli et al. ( 8293 individuals of Finnish ancestry)33; (v) 3 hemodynamic traits that were available in the UK Biobank (408,228 individuals); (vi) 5 glycaemic traits made publicly available from a series of analyses from the MAGIC Consortium (up to 133,010 individuals)34; and (vii) 16 blood immune cell counts derived from the summary statistics made publicly available by Orrù V et al. ( 3757 individuals).35 *The data* sources for these studies are described in Supplementary Table S1.
To uncover pathophysiological mechanisms pathways from TYK2 inhibitor to autoimmune disease, we performed causal mediation analysis (CMA) for certain identified biomarkers using the mediation R package36 in the UK Biobank study. We obtained an average causal mediation effect (ACME) that is transmitted via mediator to the outcome and an average direct effect that explained by the exposure as well as the proportion of explained variance by the mediator from this analysis.36
## Systematic review of clinical drug trials on TYK2 inhibitors
We conducted a systematic review on clinical trials of TYK2 inhibitors by searching corresponding studies in three databases: MEDLINE, EMBASE, and the clinical trials registration database, published until March 30th, 2022. Full search strategies are shown in Supplementary Table S2. Studies that were not RCT or not based on humans, were excluded. Information on the first author, year of study, National Clinical Trial number, characteristics of included patients, sample size, intervention, phase of trial, status of trial, assessment of efficacy and adverse effects were extracted. The literature search, review process, and data extraction were done in parallel by two authors (S.Y and X.Z.).
## Statistical analysis
The associations of rs34536443 with disease outcomes was estimated by logistic regression, and levels of biomarkers by linear regression. The PheWAS compared the risk of outcomes between individual carrying and not carrying rare TYK2 loss-of-function mutation, and the logistic regression model was adjusted for age, sex, body mass index, assessment centre, and first 10 principal genetic components. MR analysis in FinnGen and tissue-specific gene expression MR analysis was based on logistic regression with an additive [per minor (C) allele] genetic model adjusting for age, sex, 10 genetic principal components, and genotyping batch in FinnGen, and adjusting for age, sex, assessment centre, and first 10 principal genetic components in tissue-specific gene expression MR. Covariates adjusted in biomarker-wide MR analysis are presented in Supplementary Table S1. We applied a Benjamini-Hochberg correction to account for multiple testing in each analysis with FDR <0.05 were considered as statistically significant.
## Role of funding source
The funding sources had no role in the design of this study and did not have any role in the data collection, data analyses, interpretation, writing of report, or decision to submit results.
## PheWAS identified 19 disease outcomes associated with TYK2 inhibition in UK Biobank
The characteristics of 339,197 individual in UK Biobank are displayed in Supplementary Table S3. We defined 1473 phenotypes using the PheCODE schema after removing outcomes with less than 200 cases in UK Biobank (Supplementary Table S4). The MR-PheWAS analysis identified 119 outcomes nominally associated with the loss-of-function mutation of TYK2 (Supplementary Table S5), and sixteen outcomes showed significant associations after multiple-testing correction (Fig. 2a and Table 1). The mappings of ICD codes to these health outcomes are shown in Supplementary Table S6. In detail, the TYK2 loss-of-function mutation was associated with decreased risk of hypothyroidism, psoriasis and its related disorders, psoriasis vulgaris, rheumatoid arthritis and other inflammatory polyarthropathies, psoriatic arthropathy, chronic hepatitis, ulcerative colitis, inflammatory bowel disease and other gastroenteritis and colitis, celiac disease, noninfectious gastroenteritis, type 1 diabetes, disorders of eye, and increased risk of congenital deformities of feet and congenital anomalies of stomach (Table 1).Fig. 2Summary of results from Mendelian randomization (MR) analysis on disease outcomes. a, MR-PheWAS analysis of the associations between TYK2 loss-of-function mutation and health outcomes. b, MR analysis of the health effects of TYK2 inhibition on disease outcomes. c, Tissue-specific gene expression analysis for validating the associations between TYK2 expression and health outcomes. CI, confidence interval; OR, odds ratio; UKB, UK Biobank. Table 1Outcomes associated with the TYK2 loss-of-function mutation in MR-PheWAS analysis in the UK Biobank. PhecodePhenotypeGroupCasesControlsBetaSEORP244.4Hypothyroidismendocrine/metabolic18,503315,717−0.180.030.846.23E-10696.4Psoriasisdermatologic2589301,676−0.460.080.634.11E-08246Other disorders of thyroidendocrine/metabolic21,850315,717−0.140.030.875.26E-08696.41Psoriasis vulgarisdermatologic2751301,676−0.390.080.675.42E-07714.1Rheumatoid arthritismusculoskeletal5906304,719−0.240.050.792.79E-06714Rheumatoid arthritis and other inflammatory polyarthropathiesmusculoskeletal30,060304,719−0.100.020.906.26E-06696.42Psoriatic arthropathydermatologic929301,676−0.660.150.521.59E-05755.1Congenital deformities of feetcongenital anomalies273336,6220.650.161.914.89E-0570.4Chronic hepatitisinfectious diseases341330,659−1.270.330.281.50E-04750.15Congenital anomalies of stomachcongenital anomalies73335,4511.000.272.721.89E-04555.2Ulcerative colitisdigestive3269251,815−0.250.070.782.83E-04555Inflammatory bowel disease and other gastroenteritis and colitisdigestive19,792251,815−0.100.030.912.99E-04557.1Celiac diseasedigestive2185251,815−0.310.080.743.08E-04558Non-infectious gastroenteritisdigestive19,875251,815−0.100.030.913.31E-04250.1Type 1 diabetesendocrine/metabolic2862311,499−0.260.070.773.68E-04379Other disorders of eyesense organs57,586280,543−0.060.020.944.77E-04CI, confidence interval; OR, odds ratio; SE, standard error. The risk of outcomes was calculated by comparing odds between individual carrying and not carrying the rare TYK2 loss-of-function mutation.
## Health effects of TYK2 inhibition on autoimmune diseases were successfully replicated in FinnGen Biobank
The results showed that eleven related disease outcomes were successfully replicated in MR analysis in FinnGen (Fig. 2b and Supplementary Table S7). Per minor (C) allele increase of rs34536443, the odds ratio (OR) was 0.46 ($95\%$ confidence interval [CI] 0.29, 0.71) for primary biliary cirrhosis, 0.51 ($95\%$ CI 0.30, 0.88) for chronic hepatitis, 0.64 ($95\%$ CI 0.52, 0.78) for psoriatic arthropathy, 0.73 ($95\%$ CI 0.66, 0.81) for rheumatoid arthritis, 0.76 ($95\%$ CI 0.65, 0.87) for psoriasis vulgaris, 0.77 ($95\%$ CI 0.69, 0.87) for type 1 diabetes, 0.79 ($95\%$ CI 0.70, 0.88) for psoriasis, 0.83 ($95\%$ CI 0.77, 0.89) for hypothyroidism, 0.83 ($95\%$ CI 0.74, 0.94) for ulcerative colitis, 0.84 ($95\%$ CI 0.76, 0.94) for inflammatory bowel disease, and 0.86 ($95\%$ CI 0.78, 0.95) for systemic connective tissue disorders. No associations were observed between rs34536443 and cystitis, chronic kidney disease, and congenital deformities of feet. No data were available for congenital anomalies of stomach or celiac disease in FinnGen.
## Tissue-specific expression analyses verified the associations between TYK2 expression and disease outcomes across multi-tissues
Tissue-specific gene expression analyses verified that the loss-of-function mutation of rs34536443 was associated with differential expression of TYK2 in multiple tissues, particularly whole blood, visceral adipose, colon, skin, testis (Supplementary Fig. S1). We observed several associations between TYK2 expression and disease outcomes in tissues where disease occurs. Specifically, there were inverse associations of lower TYK2 expression in thyroid with reduced risk of hypothyroidism (OR, 0.86; $95\%$ CI 0.75, 1.00), in skin with psoriasis and its related disorders (OR, 0.49; $95\%$ CI 0.34, 0.69), psoriasis (OR, 0.47; $95\%$ CI 0.33, 0.67), psoriasis vulgaris (OR, 0.53; $95\%$ CI 0.36, 0.78), and psoriatic arthropathy (OR, 0.35; $95\%$ CI 0.18, 0.68) (Fig. 2c). *Differential* gene expression in other tissues also showed associations with diseases in MR-PheWAS where corresponding pathophysiology does not typically manifest (Supplementary Table S8).
## Malignant neoplasm associated with genetically proxied TYK2 inhibition
Even though there were no significant associations between genetically proxied TYK2 inhibition and risk of different cancers after correction for multiple comparison, three malignant neoplasms, including malignant neoplasm of prostate, male genital organs, and breast showed consistent suggestive positive associations with the TYK2 loss-of-function mutation in UK Biobank and FinnGen (Supplementary Table S9). Tissue-specific expression analyses showed reduced expression of TYK2 in breast tissue was associated with increased risk of breast cancer (OR, 1.21; $95\%$ CI 1.02, 1.43), but there were no associations with cancers of the prostate or male genital organs at corresponding tissues. Colocalization analysis observed no associations of TYK2 expression with prostate or breast cancer in any tissues (PP <$50\%$).
## Colocalization analysis of tissue specific TYK2 expression with disease outcomes
In total, 18 of 49 tissues had TYK2 eQTL signals at the genome-wide significant level ($P \leq 5$ × 10−8) and the F statistics of the signals ranged from 16 to 67 across tissues (Supplementary Table S10). Twelve associations of TYK2 gene expression with 6 autoimmune diseases in 8 tissues were identified in colocalization analysis (PP>$80\%$). Specifically, TYK2 gene expression showed colocalized associations with systemic lupus erythematosus in lower leg skin (PP = $100\%$), whole blood (PP = $99\%$), artery tibial (PP = $98\%$), adrenal gland (PP = $98\%$), and stomach (PP = $91\%$), psoriasis in whole blood (PP = $99\%$), ulcerative colitis (PP = $97\%$) and inflammatory bowel disease (PP = $93\%$) in brain hypothalamus, Crohn's Disease in artery tibial (PP = $97\%$), oesophagus muscularis (PP = $92\%$), and oesophagus gastroesophageal junction (PP = $87\%$), and rheumatoid arthritis in whole blood (PP = $88\%$). There were two hits prioritized by SuSiE analysis shared between TYK2 expression and above outcomes in several tissues, and additionally type 1 diabetes in visceral adipose and lung (Supplementary Table S11).
## Effects of genetically proxied TYK2 inhibition on multiple disease-related biomarkers
To gain additional insights into the relationships between TYK2 function and subclinical endophenotypes relevant to human diseases, we explored associations between the TYK2 loss-of-function variant and eight categories of 247 biomarkers derived from different sources, as detailed in Supplementary Table S12. The results, along with the number of individuals examined in each analysis are presented in Supplementary Table S12. Forty-four out of 247 biomarkers were nominally associated with rs34536443 (Supplementary Table S12). The associations for 37 of 44 biomarkers survived after multiple testing correction, mostly belonging to blood immune cell, haematological traits, and serum/urine biochemistry parameters (Fig. 3 and Supplementary Table S12). For each additional minor (C) allele of rs34536443, the levels of rheumatoid factor decreased by −1.21 ($95\%$ CI -1.98, −0.44) and the count of lymphocyte increased by 0.32 ($95\%$ 0.18, 0.47) (Fig. 3).Fig. 3Biomarkers associated with additional minor (C) allele of rs34536443 in TYK2 gene regression. CI, confidence interval. The associations survived after multiple testing were labelled in the volcano plot.
We performed the CMA for Cystatin C, insulin-like growth factor 1, sex hormone binding globulin, and interleukin 18 (Supplementary Table S13). We observed Cystatin C mediated the association of TYK2 mutations with hypothyroidism (P for ACME <0.001), rheumatoid arthritis (P for ACME = 0.02), ulcerative colitis (P for ACME = 0.02), chronic hepatitis (P for ACME <0.001), type 1 diabetes (P for ACME <0.001), Celiac disease (P for ACME <0.001), and diffuse diseases of connective tissue (P for ACME <0.001). Two mediation effects were observed for insulin-like growth factor 1 on the associations for hypothyroidism (P for ACME <0.001) and Celiac disease (P for ACME <0.001). There were no mediations observed for other biomarkers in the association between TYK2 mutations and observed outcomes in the UK Biobank (Supplementary Table S13).
## Review of RCTs on TYK2 inhibitors
A total of 23 published trials were identified in MEDLINE and 110 in EMBASE. After merging papers from two databases and removal of duplicates, 65 studies were included for screening. After title, abstract, and full-text screening, 19 studies were included. Along with 3 additional trials with published results identified in clinicaltrail.gov registration database, we included 21 RCTs on TYK2 inhibitors in this systematic review (Supplementary Fig. S2). The characteristics of 21 included RCTs are presented in Supplementary Table S14. In brief, these RCTs focused on examining the treatment effectiveness of TYK2 inhibitors on plaque psoriasis and a few studied ulcerative colitis, alopecia areata, systemic lupus erythematosus, atopic dermatitis, and active non-segmental vitiligo. These RCTs included both women and men with a wide range of age and the sample size ranged from 30 to 66.
Fifteen studies reported data on effectiveness of TYK2 inhibitors treatment on the target disease (Table 2). For plaque psoriasis, all studies ($$n = 7$$) found improved disease activity measured by the Psoriasis Area and Severity Index in the intervention groups with different doses compared to the control group. Likewise, disease activity improved in the intervention compared to control group among patients with psoriatic arthritis ($$n = 2$$), alopecia areata ($$n = 2$$), atopic dermatitis ($$n = 1$$), or active non-segmental vitiligo ($$n = 1$$) although a few studies were conducted in these diseases. TYK2 inhibitors improved certain clinical measures of ulcerative colitis severity, like improved modified Mayo endoscopic and Mayo rectal bleeding sub-score in the intervention group; however, there was no strong evidence of effect on clinical remission. Possible adverse effects of TYK2 inhibitors identified are presented in Supplementary Table S14. The most common complaints among individuals with TYK2 inhibitors treatment are headache, upper respiratory tract infection, nausea, diarrhoea, and increased circulating levels of creatinine and liver enzymes. Two RCTs reported cancer as the possible adverse effect of TYK2 inhibitor (Supplementary Table S14). Except for the above RCTs, there were some additional trails registered with the aim of exploring the effectiveness of TYK2 inhibitors on inflammatory bowel disease and systemic lupus erythematosus as well as assessing safety (Supplementary Table S15).Table 2Effectiveness assessment of TYK2 inhibitor in randomized controlled trails. StudyNCT numberDrugConditionClinical endpointInterventionNEstimation ParameterEstimated ValueP valueBanfield 2018NCT02310750PF-06700841Plaque psoriasisChange from baseline in PASI score after 4 weeksPBO30 mg QD100 mg QD9147Maximal mean percent changeRef−$67.92\%$−$96.31\%$–––Papp 2018NCT02931838BMS-986165Plaque psoriasis$75\%$ or greater reduction from baseline in PASI score at week 12 (primary)$50\%$ or greater reduction from baseline in PASI score at week $1290\%$ or greater reduction from baseline in PASI score at week $12100\%$ reduction from baseline in PASI score at week 12sPGA score of 0 or 1DLQI score of 0 or 1PBO3 mg QOD3 mg QD3 mg BID6 mg BID12 mg QDPBO3 mg QOD3 mg QD3 mg BID6 mg BID12 mg QDPBO3 mg QOD3 mg QD3 mg BID6 mg BID12 mg QDPBO3 mg QOD3 mg QD3 mg BID6 mg BID12 mg QDPBO3 mg QOD3 mg QD3 mg BID6 mg BID12 mg QDPBO3 mg QOD3 mg QD3 mg BID6 mg BID12 mg QD454444454544454444454544454444454544454444454544454444454544454444454544ProportionPercentage differencePercentage differencePercentage differencePercentage differencePercentage difference$7\%$$9\%$$39\%$$69\%$$67\%$$75\%$Ref12 (−8, 32)37 [18, 56]60 [41, 75]47 [29, 65]58 [41, 74]Ref5 (−16, 25)14 (−7, 33)42 [21, 60]42 [21, 60]41 [20, 58]Ref2 (−18, 23)–9 (−13, 30)18 (−4, 38)25 [4, 44]Ref14 (−7, 33)32 [11, 50]69 [51, 83]58 [38, 74]68 [50, 82]Ref12 (−2, 26)12 (−2, 26)38 [20, 54]56 [38, 71]59 [41, 74]Ref0.49<0.001<0.001<0.001<0.001––––––––––––––––––Ref–––––Ref–––––Forman 2020NCT02969018PF-06700841Plaque psoriasisChange from baseline in PASI score at week 12Proportion of patients achieving $75\%$ reduction from baseline PASI at week 12Proportion of patients achieving $90\%$ reduction from baseline PASI at week 12PBO30 mg QDPBO30 mg QDPBO30 mg QD232923292329LS mean differenceProportionProportionRef−17.3 (−20.0, −14.6)Ref$86.20\%$Ref$51.70\%$Ref<0.0001––––Sandbron 2020NCT02818686TD-1473Ulcerative colitisRates of clinical response and endoscopic response on day 28Rates of modified Mayo endoscopic and Mayo rectal bleeding sub-score improvement from baseline at day 28Change in Robarts Histopathology Index from baseline to day 28PBO20 mg QD80 mg QD270 mg QDPBO20 mg QD80 mg QD270 mg QDPBO20 mg QD80 mg QD270 mg QD910101191010119101011RateRatemean$11\%$ (clinical)$0\%$ (endoscopic)$20\%$ (clinical) $20\%$ (endoscopic)$20\%$ (clinical)$20\%$ (endoscopic)$55\%$ (clinical)$9\%$ (endoscopic)$0\%$ (endoscopy)$44\%$ (rectal bleeding)$20\%$ (endoscopy)$30\%$ (rectal bleeding)$30\%$ (endoscopy)$70\%$ (rectal bleeding)$18\%$ (endoscopy)$73\%$ (rectal bleeding)−2−4.51.8−5.3––––––––––––Armstrong 2021NCT03624127BMS-986165Plaque psoriasisPASI 75 response versus placebo at Week 16sPGA $\frac{0}{1}$ response versus placebo at Week 16PBO6 mg QDApremilast 30 mg BIDPBO6 mg QDApremilast 30 mg BID165322168165322168ProportionProportion$12.70\%$$58.70\%$$35.10\%$$7.20\%$$53.60\%$$32.10\%$Ref 1<0.0001Ref 2Ref 1<0.0001Ref 2Tehliran 2021NCT03210961PF-06826647Plaque psoriasisChange in PASI score at day 28PBO100 mg QD400 mg QD141115LS mean differenceRef−3.49 (−9.48, 2.50)−13.05 (−18.76, −7.35)Ref0.330.00077King 2021NCT02974868PF-06700841Alopecia areataChange from baseline in SALT score at week 24Proportion of patients achieving $30\%$ improvement in SALT score at week 24PBO60 mg QD for 4 ws30 mg QD for 20 wsPBO60 mg QD for 4 ws30 mg QD for 20 ws47474747LS mean differenceProportionRef49.2 (36.6, 61.7)–$64\%$ ($51\%$, $75\%$)Ref<0.001––Mease 2021NCT03963401PF-06700841Psoriatic arthritisACR-20 response at week 16PBO10 mg QD30 mg QD60 mg QD67316059Proportion$29\%$$20\%$$40\%$$44\%$Ref>0.05<0.05<0.05Danese 2022NCT03934216BMS-986165Ulcerative colitisClinical remission evaluated by modified Mayo score at week 12PBO6 mg BID4388Proportion$16.30\%$$14.80\%$Ref0.59Mease 2022NCT03881059BMS-986165Psoriatic arthritisACR-20 response at week 16Change from baseline in HAQ-DI score at week 16PASI-75 response at week 16Change from baseline in SF-36 PCS at week 16PBO6 mg QD12 mg QDPBO6 mg QD12 mg QDPBO6 mg QD12 mg QDPBO6 mg QD12 mg QD667067667067667067667067Adjusted ORMean differenceAdjusted ORMean differenceRef2.4 (1.2, 4.8)3.6 (1.8, 7.4)Ref−0.3 (−0.4, −0.1)−0.3 (−0.5, −0.1)Ref2.9 (1.3, 6.7)5.8 (2.4, 13.8)Ref3.3 (0.9, 5.7)3.5 (1.1, 5.9)Ref0.01340.0004Ref0.0020.0008Ref0.0136<0.0001Ref0.00620.0042Thaci 2022NCT02931838BMS-986165Plaque PsoriasisPercentages of patients who achieved absolute PASI ≤ 1, absolute PASI ≤ 3, absolute PASI ≤ 5Percentages of patients who achieved BSA ≤ $1\%$ and BSA ≤ $3\%$Percentages of patients who achieved ≥ $75\%$ improvement in sPGA × BSAPBO3 mg BID6 mg BID12 mg QDPBO3 mg BID6 mg BID12 mg QDPBO3 mg BID6 mg BID12 mg QD454545444545454445454544ProportionProportionProportion$0\%$, $2.2\%$, $8.9\%$$24.4\%$, $57.8\%$, $73.3\%$$33.3\%$, $53.3\%$, $64.4\%$$34.1\%$, $63.6\%$, $77.3\%$$0\%$, $2.2\%$$26.7\%$, $51.1\%$$37.8\%$, $44.4\%$$38.6\%$, $56.8\%$$13.30\%$$80.00\%$$73.30\%$$81.80\%$––––––––––––Winnette 2022NCT02974868PF-06700841Alopecia AreataChange in AASIS scores at week 24Correlation between SALT scores and AASIS scores at baselineCorrelation between SALT scores and AASIS scores at week 24PBO60 mg QD for 4 ws30 mg QD for 20 wsPBO60 mg QD for 4 ws30 mg QD for 20 wsPBO60 mg QD for 4 ws30 mg QD for 20 ws474747474747LS mean differencePearson correlationPearson correlationRef−1.5 (−2.1, −1.0)Ref0.18 (0.0119, 0.3325)Ref0.51 (0.3602, 0.6327)Ref<0.0001Ref0.0359Ref<0.0001Unpublished1NCT03895372PF-06826647Plaque psoriasisPercentage of participants with a PASI 90 response up to week 16 (investigation period)PBO50 mg QD100 mg QD200 mg QD400 mg QD4222214541Risk differenceRef8.87 (−4.50, 26.26)4.76 (−7.07, 21.48)33.02 (18.01, 47.11)46.46 (30.62, 60.56)Ref0.26210.26210.0004<0.0001Unpublished2NCT03903822PF-06700841Atopic DermatitisPercent change from baseline in Eczema Area and Severity Index total score at week 6PBO QD$0.1\%$ cream QD$0.3\%$ cream QD$1.0\%$ cream QD$3.0\%$ cream QDPBO BID$0.3\%$ cream BID$1.0\%$ cream BID3737363736363637LS mean differenceLS mean differenceRef−13.9 (−32.1, 4.3)−20.2 (−38.3, −2.1)−25.6 (−43.3, −8.0)−23.5 (−41.5, −5.5)Ref−11 (−24.3, 2.4)−27.4 (−40.7, −14.1)Ref0.1040.03340.00860.0158Ref0.08790.0004Unpublished3NCT03715829PF-06700841Active Non-segmental VitiligoPercent change from baseline in Central Read Facial-Vitiligo Area Scoring Index at week 24PBO200 mg + 50 mg QD100 mg + 50 mg QD50 mg QD30 mg QD10 mg QD666567675049LS mean differenceRef−23.2 (−32.53, −13.96)−23.2 (−32.53, −13.93)−20.6 (−30.23, −10.93)−16.7 (−27.77, −5.61)−5.1 (−15.02, 4.91)Ref<0.0001<0.00010.00030.00680.2015PASI, Psoriasis Area and Severity Index; sPGA, Static Physician's Global Assessment; SALT, Severity of Alopecia Tool; ACR-20, American College of Rheumatology-20; HAQ-DI, HAQ-Disability Index; SF-36 PCS, Short Form-36 Health Survey Physical Component Summary; DLQI, Dermatology Life Quality Index; BSA, body surface area; AASIS, Alopecia Areata Symptom Impact Scale; PBO, placebo; QD, once daily; BID, twice daily; QOD, every other day; LS mean difference, least-squares mean difference; adjusted OR, adjusted odds ratio; ∗, $90\%$ confidence interval; Ref, reference.
## Discussion
We comprehensively explored the genetic, phenotypic, and clinical data to investigate the efficacy and safety of TYK2 inhibitors. We found consistent evidence supporting the efficacy of TYK2 inhibitors for psoriasis and its related disorders. MR associations of the TYK2 loss-of-function variant with hypothyroidism, inflammatory bowel disease, primary biliary cirrhosis, and type 1 diabetes supported further investigation of TYK2 inhibitors as a potential treatment for these diseases in future clinical trials. Although only a few clinical trials supported that TYK2 inhibitors appeared to improve disease activity among patients with ulcerative colitis, alopecia areata, atopic dermatitis, or active non-segmental vitiligo, these findings need to be confirmed in larger studies, especially for ulcerative colitis, for which there was conflicting evidence in previous trials. Several potential adverse effects of TYK2 inhibitors, including headache, upper respiratory tract infection, nausea, diarrheal, increased circulating levels of creatinine and liver enzymes, and risk of certain malignant neoplasms, such prostate and breast cancer, should be further explored.
*Human* genetic data can be used to facilitate drug development and have been found to be effective in many scenarios.37 In genome-wide association analyses of common autoimmune diseases, like rheumatoid arthritis,20 psoriasis,19 multiple sclerosis,38 and inflammatory bowel disease,39 the TYK2 gene region has been highlighted, with the allele associated with decreased TYK2 activity showing inverse associations with risk of these diseases. A phenome-wide study on 19 candidate disease targets also indicated that TYK2 loss-of-function mutation might be associated with several autoimmune diseases,11 supporting therapeutic benefit of pharmacological inhibition. Our MR-PheWAS analysis confirmed the inverse associations between genetically proxied TYK2 inhibition and various autoimmune diseases. However, the tissue specific gene expression analysis only validated the inverse effects of genetically proxied TYK2 inhibition on hypothyroidism and psoriasis and its related disorders. In addition, colocalization analysis strengthened the associations for systemic lupus erythematosus, psoriasis, inflammatory bowel disease, and rheumatoid arthritis in appropriate tissues. The findings for psoriasis were supported by RCTs.1,2,40, 41, 42, 43 The finding for hypothyroidism is in line with a recent MR analysis11 and the present analysis went further to support mechanistic relevance specifically in thyroid tissue. For other outcomes associated with genetically proxied TYK2 inhibition, few trials were completed. Thus, the repurposing potential of TYK2 inhibitors for systemic lupus erythematosus and rheumatoid arthritis identified by genetic evidence in our current study needs clinical validation in an RCT setting. Of note, even though MR analysis used a genetic variant to mimic the biological effects of TYK2 inhibitors, several aspects deserve attention when comparing results from the current genetic study and previous trials. First, MR analysis estimated the lifetime exposure to TYK2 inhibitors. Thus, the effect estimates in the current study might be different to that observed in trials that usually last for a short period. In addition, we used loss of function of TYK2 variant to mimic TYK2 inhibitors without a clear definition of dosage in each arm, which prevented the investigation of the dose–response relationship. Compared to clinical trials, participants of the MR study were more heterogenous, and our MR design is unable to study disease progression. But MR study can usually overcome low treatment adherence (especially when the intervention has serious side-effects) and do not study off-target effects.
TYK2 plays an important role in mediating cytokine signalling and regulating group 1 and 2 cytokine pathways.44 Patients carrying TYK2 loss-of-function mutations are usually characterized by immunodeficiency,45 which may increase the risk of health outcomes such as cancer.46 From the family of TYK2 inhibitors, JAK inhibitors have been associated with increased risk of certain cancers.8,47 However, whether TYK2 inhibitors increases cancer risk has not been extensively evaluated given lack of long-term RCTs.2,48 *Our analysis* found inconsistent evidence on the associations of genetically proxied TYK2 inhibition on malignant neoplasms of the prostate or breast. The observed association for prostate cancer is in agreement with a recent MR study where TYK2 inhibition mimicked by a loss-of-function variant in TYK2 (rs34536443) showed associations with lung cancer, non-Hodgkin lymphoma, and advanced prostate cancer.9 Although we observed nominal associations of genetically proxied TYK2 inhibition with prostate and breast cancer risk in both UK Biobank and FinnGen, the tissue-specific gene expression and colocalization analyses did not confirm these associations. From the current evidence, whether TYK2 inhibitor increases the risk of cancer remains undetermined and needs further study, especially in RCTs with a long-term follow-up period.
Other adverse effects reported in previous RCTs include headache, upper respiratory tract infection, nausea, diarrheal, and increased circulating levels of creatinine and liver enzymes.1,40,41,49 However, our MR analysis found a contradictory association of genetically proxied TYK2 inhibition with reduced levels of alkaline phosphatase. One in vivo study found that deletion of TYK2 in myeloid cells reduced lipopolysaccharide-induced interleukin 18 production,50 which is in line with our MR findings on interleukin 18. In addition, the effects of the TYK2 loss-of-function variant on sex hormone binding globulin15 and insulin-like growth factor-I,51,52 which exerts effects on a wide range of diseases, may also hint at other possible pleiotropic effects related to TYK2 inhibitor use.
The present study has several strengths. Firstly, we explored associations of the TYK2 loss-of-function mutation with a wide range of disease outcomes in a large biobank and validated the associations in independent populations. Secondly, we used several analytical approaches to examine the associations, and the consistency of results increase confidence in our findings. Thirdly, we conducted a review of RCTs on TYK2 inhibitors to triangulate the evidence. The consistency between findings of the genetic analysis and RCTs further supports the robustness of our conclusions. Limitations also need to be considered when interpreting our findings. Our analysis may have inadequate power for rare diseases and outcomes with low prevalence. For the analyses of biomarkers, we could not compare the results for biomarkers measured in different units across studies with varying sample sizes. Body mass index was adjusted for in the genome-wide association analysis of cytokines and glycaemic traits, which might introduce collider bias in these MR analyses. Although TYK2 is a protein coding gene, previous studies identified no cis signal in this gene affecting gene expression at the genome-wide significance level,53 which confined colocalization analysis based on protein quantitative levels. The mediation effect should be interpreted with caution given the strong assumptions to be held under the mediation analysis. In addition, our analysis was majorly based on the European population. Whether our findings can be generalized to other populations needs to be examined in future studies. There was no risk of bias assessment of included trials in the review of TYK2 inhibitors due to limited information on several studies. Thus, whether the summarized evidence from published trials is robust needs to be verified.
In summary, using multiple analytic approaches this study found that genetically proxied TYK2 inhibition was associated with lower risk of psoriasis and its related disorders. The association is largely supported by RCT evidence. The observed associations of TYK2 with other autoimmune diseases, including hypothyroidism, systemic lupus erythematosus and rheumatoid arthritis, should help inform future clinical study design. Finally, potential adverse effects of TYK2 inhibitors, including elevated risk of prostate and breast cancer, should be evaluated in studies with long follow-up duration.
## Contributors
X.L. and S.Y. had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. X.L., S.L., S.C.L., and E.T. conceived and designed the study. X.L., S.Y., and L.W. undertook the statistical analyses. S.Y. wrote the first draft of the manuscript. X.L. is the study guarantor. S.Y., L.W., H.Z., F.X., X.Z., L.Y., J.S., J.C., H.Y., X.X., Y.Y., A.S., X.S., J.W., D.G., E.T., S.C.L., and X.L. interpreted data, reviewed the paper, and made critical revision of the manuscript for important intellectual content. All authors read and approved the final version of the manuscript.
## Data sharing statement
Data used in this study can be obtained by a reasonable request to corresponding author. This work has been conducted using the UK Biobank Resource. The UK *Biobank is* an open access resource and bona fide researchers can apply to use the UK Biobank dataset by registering and applying at http://ukbiobank.ac.uk/register-apply/.
## Declaration of interests
DG is employed part-time by Novo Nordisk. The other authors declare no competing interest.
## Supplementary data
Supplementary Figs. S1 and S2 and Tables S1–S15
## References
1. Armstrong A., Gooderham M., Warren R.B.. **POS1042 efficacy and safety of deucravacitinib, an oral, selective tyrosine kinase 2 (TYK2) inhibitor, compared with placebo and apremilast in moderate to severe plaque psoriasis: results from the phase 3 poetyk PSO-1 study**. *Ann Rheum Dis* (2021) **80** 795-796
2. Papp K., Gordon K., Thaçi D.. **Phase 2 trial of selective tyrosine kinase 2 inhibition in psoriasis**. *N Engl J Med* (2018) **379** 1313-1321. PMID: 30205746
3. Sandborn W.J., Feagan B.G., Loftus E.V.. **Efficacy and safety of upadacitinib in a randomized trial of patients with Crohn's disease**. *Gastroenterology* (2020) **158** 2123-2138.e8. PMID: 32044319
4. Taylor P.C., Keystone E.C., van der Heijde D.. **Baricitinib versus placebo or adalimumab in rheumatoid arthritis**. *N Engl J Med* (2017) **376** 652-662. PMID: 28199814
5. Garg S.K., Henry R.R., Banks P.. **Effects of sotagliflozin added to insulin in patients with type 1 diabetes**. *N Engl J Med* (2017) **377** 2337-2348. PMID: 28899222
6. Sandborn W.J., Nguyen D.D., Beattie D.T.. **Development of gut-selective pan-janus kinase inhibitor TD-1473 for ulcerative colitis: a translational medicine programme**. *J Crohns Colitis* (2020) **14** 1202-1213. PMID: 32161949
7. Danese S., Panaccione R., D'Haens G.. **DOP42 Efficacy and safety of deucravacitinib, an oral, selective tyrosine kinase 2 inhibitor, in patients with moderately-to-severely active Ulcerative Colitis: 12-week results from the Phase 2 LATTICE-UC study**. *J Crohn's Colitis* (2022) **16** i091-i092
8. USFaD Administration. (2021)
9. Yarmolinsky J., Amos C.I., Hung R.J.. **Association of germline TYK2 variation with lung cancer and non-Hodgkin lymphoma risk**. *Int J Cancer* (2022) **151** 2155-2160. PMID: 35747941
10. Burgess S., Thompson S.G.. (2015)
11. Diogo D., Tian C., Franklin C.S.. **Phenome-wide association studies across large population cohorts support drug target validation**. *Nat Commun* (2018) **9** 4285. PMID: 30327483
12. Denny J.C., Bastarache L., Ritchie M.D.. **Systematic comparison of phenome-wide association study of electronic medical record data and genome-wide association study data**. *Nat Biotechnol* (2013) **31** 1102-1110. PMID: 24270849
13. Li X., Meng X., Spiliopoulou A.. **MR-PheWAS: exploring the causal effect of SUA level on multiple disease outcomes by using genetic instruments in UK Biobank**. *Ann Rheum Dis* (2018) **77** 1039-1047. PMID: 29437585
14. Li X., Meng X., He Y.. **Genetically determined serum urate levels and cardiovascular and other diseases in UK Biobank cohort: a phenome-wide mendelian randomization study**. *PLoS Med* (2019) **16**
15. Yuan S., Wang L., Sun J.. **Genetically predicted sex hormone levels and health outcomes: phenome-wide Mendelian randomization investigation**. *Int J Epidemiol* (2022) **51** 1931-1942. PMID: 35218343
16. Kurki M.I., Karjalainen J., Palta P.. **FinnGen provides genetic insights from a well-phenotyped isolated population**. *Nature* (2023) **613** 508-518. PMID: 36653562
17. Gusev A., Ko A., Shi H.. **Integrative approaches for large-scale transcriptome-wide association studies**. *Nat Genet* (2016) **48** 245-252. PMID: 26854917
18. Battle A., Brown C.D., Engelhardt B.E., Montgomery S.B.. **Genetic effects on gene expression across human tissues**. *Nature* (2017) **550** 204-213. PMID: 29022597
19. Tsoi L.C., Spain S.L., Knight J.. **Identification of 15 new psoriasis susceptibility loci highlights the role of innate immunity**. *Nat Genet* (2012) **44** 1341-1348. PMID: 23143594
20. Okada Y., Wu D., Trynka G.. **Genetics of rheumatoid arthritis contributes to biology and drug discovery**. *Nature* (2014) **506** 376-381. PMID: 24390342
21. Liu J.Z., van Sommeren S., Huang H.. **Association analyses identify 38 susceptibility loci for inflammatory bowel disease and highlight shared genetic risk across populations**. *Nat Genet* (2015) **47** 979-986. PMID: 26192919
22. Bentham J., Morris D.L., Graham D.S.C.. **Genetic association analyses implicate aberrant regulation of innate and adaptive immunity genes in the pathogenesis of systemic lupus erythematosus**. *Nat Genet* (2015) **47** 1457-1464. PMID: 26502338
23. Consortium I.M.S.G.. **A systems biology approach uncovers cell-specific gene regulatory effects of genetic associations in multiple sclerosis**. *Nat Commun* (2019) **10** 2236. PMID: 31110181
24. Chiou J., Geusz R.J., Okino M.L.. **Interpreting type 1 diabetes risk with genetics and single-cell epigenomics**. *Nature* (2021) **594** 398-402. PMID: 34012112
25. Schumacher F.R., Al Olama A.A., Berndt S.I.. **Association analyses of more than 140,000 men identify 63 new prostate cancer susceptibility loci**. *Nat Genet* (2018) **50** 928-936. PMID: 29892016
26. Zhang H, Ahearn T.U., Lecarpentier J.. **Genome-wide association study identifies 32 novel breast cancer susceptibility loci from overall and subtype-specific analyses**. *Nat Genet* (2020) **52** 572-581. PMID: 32424353
27. Võsa U., Claringbould A., Westra H.J.. **Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression**. *Nat Genet* (2021) **53** 1300-1310. PMID: 34475573
28. Giambartolomei C., Vukcevic D., Schadt E.E.. **Bayesian test for colocalisation between pairs of genetic association studies using summary statistics**. *PLoS Genet* (2014) **10**
29. Wallace C.. **A more accurate method for colocalisation analysis allowing for multiple causal variants**. *PLoS Genet* (2021) **17**
30. Wang G., Sarkar A., Carbonetto P., Stephens M.. **A simple new approach to variable selection in regression, with application to genetic fine mapping**. *J Roy Stat Soc B* (2020) **82** 1273-1300
31. Astle W.J., Elding H., Jiang T.. **The allelic landscape of human blood cell trait variation and links to common complex disease**. *Cell* (2016) **167** 1415-1429.e19. PMID: 27863252
32. Kettunen J., Demirkan A., Wurtz P.. **Genome-wide study for circulating metabolites identifies 62 loci and reveals novel systemic effects of LPA**. *Nat Commun* (2016) **7**
33. Ahola-Olli A.V., Wurtz P., Havulinna A.S.. **Genome-wide association study identifies 27 loci influencing concentrations of circulating cytokines and growth factors**. *Am J Hum Genet* (2017) **100** 40-50. PMID: 27989323
34. 34MAGIC ConsortiumMAGIC (the meta-analyses of glucose and insulin-related traits consortium)2020Available from:https://www.magicinvestigators.org/. (2020)
35. Orrù V., Steri M., Sidore C.. **Complex genetic signatures in immune cells underlie autoimmunity and inform therapy**. *Nat Genet* (2020) **52** 1036-1045. DOI: 10.1038/s41588-020-0684-4
36. Zhang Z., Zheng C., Kim C., Van Poucke S., Lin S., Lan P.. **Causal mediation analysis in the context of clinical research**. *Ann Transl Med* (2016) **4** 425. PMID: 27942516
37. Holmes M.V.. **Human genetics and drug development**. *N Engl J Med* (2019) **380** 1076-1079. PMID: 30865805
38. Beecham A.H., Patsopoulos N.A., Xifara D.K.. **Analysis of immune-related loci identifies 48 new susceptibility variants for multiple sclerosis**. *Nat Genet* (2013) **45** 1353-1360. PMID: 24076602
39. Jostins L., Ripke S., Weersma R.K.. **Host-microbe interactions have shaped the genetic architecture of inflammatory bowel disease**. *Nature* (2012) **491** 119-124. PMID: 23128233
40. Banfield C., Scaramozza M., Zhang W.. **The safety, tolerability, pharmacokinetics, and pharmacodynamics of a TYK2/JAK1 inhibitor (PF-06700841) in healthy subjects and patients with plaque psoriasis**. *J Clin Pharmacol* (2018) **58** 434-447. PMID: 29266308
41. Forman S.B., Pariser D.M., Poulin Y.. **TYK2/JAK1 inhibitor PF-06700841 in patients with plaque psoriasis: phase IIa, randomized, double-blind, placebo-controlled trial**. *J Invest Dermatol* (2020) **140** 2359-2370.e5. PMID: 32311398
42. Tehlirian C., Peeva E., Kieras E.. **Safety, tolerability, efficacy, pharmacokinetics, and pharmacodynamics of the oral TYK2 inhibitor PF-06826647 in participants with plaque psoriasis: a phase 1, randomised, double-blind, placebo-controlled, parallel-group study**. *Lancet Rheumatol* (2021) **3** e204-e213
43. Thaçi D., Strober B., Gordon K.B.. **Deucravacitinib in Moderate to Severe Psoriasis: Clinical and Quality-of-Life Outcomes in a Phase 2 Trial**. *Dermatol Ther (Heidelb)* (2022) **12** 495-510. PMID: 35025062
44. Watford W.T., O'Shea J.J.. **Human tyk2 kinase deficiency: another primary immunodeficiency syndrome**. *Immunity* (2006) **25** 695-697. PMID: 17098200
45. Minegishi Y., Saito M., Morio T.. **Human tyrosine kinase 2 deficiency reveals its requisite roles in multiple cytokine signals involved in innate and acquired immunity**. *Immunity* (2006) **25** 745-755. PMID: 17088085
46. O'Shea J.J., Holland S.M., Staudt L.M.. **JAKs and STATs in immunity, immunodeficiency, and cancer**. *N Engl J Med* (2013) **368** 161-170. PMID: 23301733
47. Yarmolinsky J., Amos C.I., Hung R.J.. **Association of germline TYK2 variation with lung cancer and non-Hodgkin lymphoma risk**. *Int J Cancer* (2022) **151** 2155-2160. PMID: 35747941
48. Mease P.H.P., Silwinska-Stanczyk P., Miakisz M.. **Efficacy and safety of brepocitinib (tyrosine kinase 2/janus kinase 1 inhibitor) for the treatment of active psoriatic arthritis: results from a phase 2b randomized controlled trial**. *Arthritis Rheumatol* (2021)
49. Mease P.J., Deodhar A.A., van der Heijde D.. **Efficacy and safety of selective TYK2 inhibitor, deucravacitinib, in a phase II trial in psoriatic arthritis**. *Ann Rheum Dis* (2022) **81** 815-822. PMID: 35241426
50. Poelzl A., Lassnig C., Tangermann S.. **TYK2 licenses non-canonical inflammasome activation during endotoxemia**. *Cell Death Differ* (2021) **28** 748-763. PMID: 32929218
51. Larsson S.C., Carter P., Vithayathil M., Kar S., Mason A.M., Burgess S.. **Insulin-like growth factor-1 and site-specific cancers: a Mendelian randomization study**. *Cancer Med* (2020) **9** 6836-6842. PMID: 32717139
52. Yuan S., Wan Z.H., Cheng S.L., Michaëlsson K., Larsson S.C.. **Insulin-like growth factor-1, bone mineral density, and fracture: a mendelian randomization study**. *J Clin Endocrinol Metab* (2021) **106** e1552-e1558. PMID: 33462619
53. Pietzner M., Wheeler E., Carrasco-Zanini J.. **Mapping the proteo-genomic convergence of human diseases**. *Science* (2021) **374**
|
---
title: Modeling of dielectric resonator antenna array for retina photoreceptors☆
authors:
- Mahdi NoroozOliaei
- Hamid Riazi Esfahani
- Mohammad Sadegh Abrishamian
journal: Heliyon
year: 2023
pmcid: PMC9988474
doi: 10.1016/j.heliyon.2023.e13794
license: CC BY 4.0
---
# Modeling of dielectric resonator antenna array for retina photoreceptors☆
## Abstract
The retina encompasses several cone and rod photoreceptors at fovea region i.e., 90 million cells of rod photoreceptors and 4.5million cells of cone photoreceptors. The overall photoreceptors determine the vision of every human. An electromagnetic dielectric resonator antenna has been presented for retina photoreceptors in order to model them at fovea and its peripheral retina with the respected angular spectrum. Three coloring primary system of human eye (R, G, B) can be realized based on the model. Three miscellaneous models i.e., simple, graphene coated, and interdigital models have been presented in this paper. The nonlinear property of interdigital structures is one of the best advantages to use for creating the capacitor. The capacitance property helps improving the upper band of visible spectrum. The absorption of light for graphene as an energy harvesting material and its conversion into electrochemical signals is making it one of the best models. The mentioned three electromagnetic models of human photoreceptors have been expressed as a receiver antenna. The proposed electromagnetic models based on dielectric resonator antenna (DRA) are being analyzed for cones and rods photoreceptors of retina in the human eye by Finite Integral Method (FIM) utilized by CST MWS. The results show that the models are so fine for vision spectrum due to its localized near field enhancement property. The results indicate fine parameters of S11 (return loss below -10 dB) with invaluable resonants in a wide range of frequencies from 405 THz to 790 THz (vision spectrum), appropriate S21 (insertion loss 3-dB bandwidth), very good field distribution of electric and magnetic fields for flowing the power and electrochemical signals. Finally, mfERG clinical and experimental results validate the numeric results by the normalized output to input ratio of these models and it points out that these models can stimulate the electrochemical signals in photoreceptor cells for the best suiting of realizing the new retinal implants.
## Introduction
Stimulation of nerves for eyes to recover the lost vision is necessary for patients. The electromagnetic modeling of retina photoreceptors is the best way to realize this vital procedure. Photoreceptors in a normal eye are being located at the outer layers of retina containing photopigments which they are responsible for phototransduction to generate the electrochemical neuronal signals in the presence of light stimuli. These signals can be passed and processed by some neurons at the middle layers of the retina before reaching the retinal ganglion cells (RGC) within the inner layers of retina and the central nervous system (CNS) by optic nerve. Vision restoration of blind people can be achieved by creating devices such as retinal prostheses which they can receive the light and process it. After processing the light, the information can be transmitted in the form of electrical impulses to the remaining inner retinal layers for the visual function. One of the big electromagnetic challenges is analysis and modeling of retina and its optic nerve in order to replace a prosthesis as a bionic eye. Many good endeavors for realizing commercial prosthesis (epi-retinal and sub-retinal implants) are being done so far [1], [2], [3], [4], [5], [6], [7]. In [8], a graphene based antenna model has been introduced for the electromagnetic model of retina photoreceptors. The typical human trichromatic color visual system combines the response of three types of cone cells in retina as shown in Fig. 1 of [8]. In this paper, the proposed electromagnetic models on array antenna of different models of dielectric resonator antenna (DRA) are being simulated for cones and rods in retina photoreceptors of human eye by Finite Integral Method (FIM) utilizing CST MWS.Figure 13 × 3 array of simple cone photoreceptor (a) schematic, (b) S parameters, (c) reflectance, transmittance and absorbance index, and (d) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 1 Two modes of array antenna have been investigated in this paper. Finite array with 9 (3 × 3) elements and infinite array (unit cell boundary conditions). There are three different models: simple, graphene coated and interdigital one. The dielectric constant of photoreceptor (cone or rod shapes) and its media (with a cross section of 5 × 5 μm2) are ϵrPh =2.1 and ϵrM =1.85, respectively. The description of human cone cells dimension has been presented in [8].
The dimensions of human rod cell are being brought into [9] as well.
The distributions of photoreceptors in the retina are defined as the following: 1- Cone photoreceptors at the central region (from 0∘ to 5∘) of retina (fovea) with a high density which it is defined by infinite array (unit cell).
2- Cone photoreceptors at the near central region (from 5∘ to 15∘) of retina with a moderated density which it can be shown by finite 3 × 3 array.
3- Rod photoreceptors (from 20∘-27∘) of retina with a high density which it is defined by infinite array (unit cell).
4- Rod photoreceptors (more than 27∘) of retina with a moderated density which it can be replaced by finite 3 × 3 array.
Every simulated cone photoreceptor array has the equal length of 50 μm and the diameter varies from 4 μm (first point) to 1 μm (end point). Every simulated rod photoreceptor array has the equal length of 100 μm and the diameter is 2 μm. The results of the output to input ratio obtained by numerical method of the mentioned arrangement for photoreceptors array have very good conformance with mfERG ones of a normal subject with a valid interpretation of the signal to noise ratio of the experiments.
The following sections illustrate the response and electromagnetic fields at one frequency sample of each three models.
## Array of simple model (equal cone photoreceptor)
3 × 3 array of simple cone photoreceptor model has been illustrated in Fig. 1(a). The array assumed to be equal in size (both length and width). S parameters (return loss and insertion loss) and reflection, absorption, and transmission indexes of this array have been shown in Figs. 1(b) and 1(c). The main resonant frequency of finite cone photoreceptor array is around 450 THz (Red Region). It is noteworthy that these types of structures (Dielectric Resonant Antenna) have several resonant frequencies at visible spectrum. It has no significant transmission coefficient. The absorption is high up to high frequencies around 780 THz and the reflection increases after frequencies around 650 THz. It will be reached to its maximum at frequencies around 780 THz. Fig. 1(d) shows electric and magnetic fields of 3 × 3 array for simple cone photoreceptor in its maximum possibility of local magnitude at one of frequency samples i.e., 565 THz.
## Array of simple model (equal rod photoreceptor)
3 × 3 array of simple rod photoreceptor model has been illustrated in Fig. 2(a). The array assumed to be equal in size (both length and width). S parameters (return loss and insertion loss) of this array have been shown in Fig. 2(b). There is no significant transmission for rod photoreceptors as shown in Fig. 2(b). It is also fine for these types of photoreceptors as rod cell receives lower luminance of signals in scotopic vision. In other words, it requires less light to function than cone. Fig. 2(c) shows electric and magnetic fields of 3 × 3 array for simple rod photoreceptor at one of frequency samples i.e., 565 THz. Figure 23 × 3 array of simple rod photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 2
## Unit cell of simple model (cone photoreceptor)
The unit cell boundary condition provides criterion which it equals to infinite array. Simple cone photoreceptor infinite array (unit cell) has the equal length and diameters as shown in Fig. 3(a). The return loss (S11) of unit cell is below -15 dB while the insertion loss (S21) has a ripple shape at visible spectrum (Fig. 3(b)). It is also obvious from transmittance and absorbance indexes in Fig. 3(c). Fig. 3(d) shows electric and magnetic fields of unit cell (infinite array) for simple cone photoreceptor at one of frequency samples i.e., 565 THz. Figure 3Unit cell of simple cone photoreceptor (a) schematic, (b) S parameters, (c) reflectance, transmittance and absorbance index, and (d) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 3
## Unit cell of simple model (rod photoreceptor)
Simple rod photoreceptor infinite array (unit cell) has the equal length and the diameter as shown in Fig. 4(a). The return loss (S11) of unit cell shows main resonant at middle frequency of visible spectrum while the insertion loss (S21) has a slight ripple at visible spectrum (Fig. 4(b)). Fig. 4(c) shows electric and magnetic fields of unit cell (infinite array) for simple rod photoreceptor at one of frequency samples i.e., 565 THz. Figure 4Unit cell of simple rod photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 4
## Array of graphene coated model (equal cone photoreceptor)
3 × 3 array of graphene coated cone photoreceptor model has been illustrated in Fig. 5(a). The array assumed to be equal in size (both length and width). A thin graphene layer is being coated around the photoreceptor. S parameters (return loss and insertion loss) have been shown in Fig. 5(b). The return loss (S11) of 3 × 3 array of graphene coated cone photoreceptor is below -10 dB at visible spectrum and the insertion loss (S21) has a slight ripple shape near -10 dB at this spectrum. Fig. 5(c) shows electric and magnetic fields of 3 × 3 array for graphene coated cone photoreceptor at one of frequency samples i.e., 565 THz. Figure 53 × 3 array of graphene coated cone photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 5
## Array of graphene coated model (equal rod photoreceptor)
3 × 3 array of graphene coated rod photoreceptor model has been illustrated in Fig. 6(a). The array assumed to be equal in size (both length and width). A thin graphene layer is being coated around the photoreceptor. S parameters (return loss and insertion loss) have been shown in Fig. 6(b). The return loss (S11) of 3 × 3 array of graphene coated rod photoreceptor is below -10 dB at visible spectrum and the insertion loss (S21) has a ripple shape between near -10 dB and -20 dB at this spectrum. Fig. 6(c) shows electric and magnetic fields of 3 × 3 array for graphene coated rod photoreceptor at one of frequency samples i.e., 565 THz. Figure 63 × 3 array of graphene coated rod photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 6
## Unit cell of graphene coated model (cone photoreceptor)
Graphene coated cone photoreceptor infinite array (unit cell) has the equal length and the diameter as shown in Fig. 7(a). A thin graphene layer is being coated around the photoreceptor. The main resonant frequencies of unit cell are in the green light spectrum as illustrated in Fig. 7(b)). The transmittance, reflectance and absorbance indexes indicate that this is a proper model for considering as array (Fig. 7(c)). Fig. 7(d) illustrates the group delay of transmitting from port 1 and port 2. Fig. 7(e) shows electric and magnetic fields of unit cell (infinite array) for graphene coated cone photoreceptor at one of frequency samples i.e., 565 THz. Figure 7Unit cell of graphene coated cone photoreceptor (a) schematic (b) S parameters, (c) reflectance, transmittance and absorbance index, (d) group delay of transmittance way, and (e) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 7
## Unit cell of graphene coated model (rod photoreceptor)
Graphene coated rod photoreceptor infinite array (unit cell) has the equal length and the diameter as shown in Fig. 8(a). The return loss (S11) of unit cell shows that it is below -50 dB while the insertion loss (S21) has a slight ripple around -10 dB at visible spectrum (Fig. 8(b)). Fig. 8(c) shows electric and magnetic fields of unit cell (infinite array) for graphene coated rod photoreceptor at one of frequency samples i.e., 565 THz. Figure 8Unit cell of graphene coated rod photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 8
## Array of interdigital model (equal cone photoreceptor)
3 × 3 array of interdigital cone photoreceptor model has been illustrated in Fig. 9(a). The array assumed to be equal in size (both length and width). S parameters (return loss and insertion loss) have been shown in Fig. 9(b). The main resonant frequencies are in green and blue lights spectrum. The insertion loss (S21) has a ripple shape near -10 dB upto 650 THz and then it degrades at higher frequencies of visible spectrum. Fig. 9(c) shows electric and magnetic fields of 3 × 3 array for interdigital cone photoreceptor at one of frequency samples i.e., 565 THz. Figure 93 × 3 array of interdigital cone photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 9
## Array of interdigital model (equal rod photoreceptor)
3 × 3 array of interdigital rod photoreceptor model has been illustrated in Fig. 10(a). The array assumed to be equal in size (both length and width). S parameters (return loss and insertion loss) have been shown in Fig. 10(b). The return loss (S11) is below -20 dB and insertion loss (S21) of 3 × 3 array of interdigital rod photoreceptor has a ripple shape around -10 dB to -20 dB at visible spectrum. Fig. 10(c) shows electric and magnetic fields of 3 × 3 array for interdigital rod photoreceptor at one of frequency samples i.e., 565 THz. Figure 103 × 3 array of interdigital rod photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 10
## Unit cell of interdigital model (cone photoreceptor)
Interdigital cone photoreceptor infinite array (unit cell) has the equal length and the diameter as shown in Fig. 11(a). The main resonant frequencies of unit cell are in the green light spectrum as illustrated in Fig. 11(b)). The transmittance, reflectance and absorbance indexes indicate that this is a proper model for considering as array (Fig. 11(c)) like the graphene coated model. Fig. 11(d) illustrates the group delay of transmitting from port 1 and port 2 which it shows a low magnitude around the 0.3 ps for determined phase. It can be conformed by N1 implicit time of mfERG technique. Fig. 11(e) shows electric and magnetic fields of unit cell (infinite array) for interdigital cone photoreceptor at one of frequency samples i.e., 565 THz. Figure 11Unit cell of interdigital cone photoreceptor (a) schematic (b) S parameters, (c) reflectance, transmittance and absorbance index, (d) group delay of transmittance way, and (e) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 11
## Unit cell of interdigital model (rod photoreceptor)
Interdigital rod photoreceptor infinite array (unit cell) has the equal length and the diameter as shown in Fig. 12(a). The return loss (S11) of unit cell shows that it has the main resonant in green light frequencies while the insertion loss (S21) has a slight ripple around -10 dB at visible spectrum (Fig. 12(b)). Fig. 12(c) shows electric and magnetic fields of unit cell (infinite array) for interdigital rod photoreceptor at one of frequency samples i.e., 565 THz. Figure 12Unit cell of interdigital rod photoreceptor (a) schematic, (b) S parameters, and (c) electric and magnetic fields in one sample frequency - green light (565 THz).Figure 12
## mfERG experimental and clinical results
The multifocal ERG responses for a normal subject with a valid interpretation of signal to noise ratio are being shown in Fig. 13 [10], [11], [12], [13], [14], [15]. The field trace, N1 implicit time, and N1 amplitude of left and right eyes are being illustrated in Figs. 13 (a) to (f) correspondingly. The N1 waves are related to the signals of photoreceptors. Figure 13Left eye of a normal subject (a) field trace, (c) N1 implicit time, (e) N1 amplitude, right eye of a normal subject (b) field trace, (d) N1 implicit time, (f) N1 amplitude. Figure 13
## Analysis
Fig. 14(a) illustrates the number and distribution of cones and rods in mm2 with respect to the distance within retina. The cones and rods connection to the retina nerves has been shown in Fig. 14(b). The microscopic demonstration of cones and rods with a line scale of 10 μm has been shown in Fig. 14(c). The form of Fig. 14 shows that the arrangement of photoreceptor array should be like Fig. 15. There is infinite array of cone photoreceptor at the center of retina's field of view (0∘ to 5∘). The distribution of cones decreases from 5∘ to 15∘ meanwhile the distribution of rods increases in this area. The rods distribution will be maximum around 20∘-27∘ and afterwards it begins to decrease upto 30∘ and 70∘. The maximum input power (intensity) of incidence into photoreceptors is 0.5 Wμm2. It is different for any introduced structures in this paper. The ratio between output (electric field at maximum possibility of local magnitude i.e. central locations) and input (intensity) of all structures (models) have been attained and then these values have been normalized to their maximum in each model based on the presented local constellation in Fig. 15. These normalized values (based on numerical method) are being compared with mfERG clinical and experimental ones as illustrated in Fig. 16. The values for simple, graphene coated and interdigital models have been shown in Figs. 16 (a) to (f). The best array model conforms to the mfERG results is the interdigital photoreceptor array model. The reason for this can be concluded by the shape of photoreceptor model as it has been designed with a near structure to the real photoreceptor of human eye at outer segment. The capacitance of outer segment in this model interacts with its inductance property and consequently affects the mentioned ratio (output to input) effectively. The interdigital cone photoreceptor has near 150 interdigital segments at outer segment (OS) and its rod photoreceptor has 1500 ones. It is noteworthy that the interdigital segment thickness and the periodic step of both cone and rod are 10 nm and 25 nm, respectively. Figure 14Retina photoreceptors (a) number and distribution of cones and rods in mm2 with respect to the distance within retina, (b) illustration of cones and rods connection to the nerves of retina, and (c) illustration of cones and rods (line scale is 10 μm) [16], [17].Figure 14Figure 15Modeling of retina photoreceptors (cones and rods) in terms of arrays. Figure 15Figure 16Comparison of numerical method array (at one frequency sample) with mfERG clinical and experimental results in normalized output to input ratio [μV.mW] (a) simple photoreceptor array model-left eye, (b) simple photoreceptor array model-right eye, (c) graphene photoreceptor array model-left eye, (d) graphene photoreceptor array model-right eye, (e) interdigital photoreceptor array model-left eye, and (f) interdigital photoreceptor array model-right eye. Figure 16 The E and H fields of dielectric photoreceptors for every model can be presented here as these fields have been shown with maximum probability location for both conveying and receiving (on Z axis) the signals in the previous figures. Figure 17, Figure 18, Figure 19 are illustrating the electromagnetic fields on dielectric models for simple, graphene coated and interdigital ones, respectively. Figure 17E (left) and H (right) plots for simple model (a), (b) cone - 3 × 3 array, (c), (d) cone - unit cell, (e), (f) rod - 3 × 3 array, and (g), (h) rod - unit cell. Figure 17Figure 18E (left) and H (right) plots for graphene coated model (a), (b) cone - 3 × 3 array, (c), (d) cone - unit cell, (e), (f) rod - 3 × 3 array, and (g), (h) rod - unit cell. Figure 18Figure 19E (left) and H (right) plots for interdigital model (a), (b) cone - 3 × 3 array, (c), (d) cone - unit cell, (e), (f) rod - 3 × 3 array, and (g), (h) rod - unit cell. Figure 19 There are E and H fields for simple cone and rod photoreceptors array in Fig. 17 (a) to (h). The fields for graphene coated photoreceptors array are in Fig. 18 (a) to (h) as well. The E and H fields of interdigital photoreceptors array model are being shown in Fig. 19 (a) to (h).
The impedance bandwidth parameter of DRA severely depends on its permittivity. Decreasing the dielectric constant of the model results in the decrement of the Q-factor and consequently increases the bandwidth of the resonant modes. Moreover, the resonant frequencies of low dielectric constant are higher than those of with the high permittivity one [18].
As one single cone photoreceptor is connected with one nerve through optic nerve, there is a comparison on Z axis for output to input ratio of interdigital cone model (the best suited model) by two numerical methods i.e. FIT (CST MWS) and FEM (COMSOL Multiphysics) at the determined sample frequency, 565 THz (Fig. 20). Fig. 20 (a) illustrates the electric fields of single interdigital cone photoreceptor in CST MWS whereas Fig. 20 (b) shows its form in COMSOL Multiphysics at one determined frequency. The maximum probability location for both conveying and receiving of the signals (on Z axis) are being shown for comparison with these two methods in Fig. 20 (c).Figure 20Single interdigital cone photoreceptor (a) E-field in CST MWS, (b) E-field in COMSOL multiphysics, and (c) output to input ratio (O/I) of numerical methods (near central fovea) on the central axis at specified sample frequency. Figure 20 The aim of this research for showing these results is the inquiry of phototransduction process of photoreceptors mainly the received signals with respect to the input ones. So, dielectric transmission line can be more suitable for using in this note as well.
## Conclusion
In this paper, electromagnetic dielectric resonator antenna (DRA) was presented for retina photoreceptors as a receiver antenna concept based on three models i.e., simple, graphene coated and interdigital structure at outer segment (OS) in order to model them at fovea with the respected angular spectrum. The proposed electromagnetic models based on dielectric resonator antenna (DRA) was analyzed for cones and rods of retina photoreceptors in human eye by utilizing the Finite Integral Method (FIM) associated with CST MWS. The results showed that these models were so fine for vision spectrum with a proper field enhancement in cone photoreceptors due to its sensitivity to light by the localized near field enhancement property. The results indicated fine properties of S11 (return loss below -10 dB) with invaluable resonants in a wide range of frequencies from 405 THz to 790 THz (vision spectrum), adequate S21 (insertion loss 3-dB bandwidth), very good field distribution for flowing the power. Finally, the array of each model was compared with clinical and experimental techniques, mfERG. It was shown that the results obtained from mfERG validate the numeric results. The best array model conformed by the mfERG results was the interdigital photoreceptor array model. The reason for this concluded by the shape of photoreceptor model as it was designed with a near shape to the real photoreceptor of human eye at outer segment. The interaction of capacitance and inductance properties of outer segment in this model was one of these effective impacts on the mentioned ratio.
## Ethical approval
The study followed the tenets of declaration of Helsinki and was approved by the Farabi eye hospital/Tehran University of Medical Science's Institutional Review Board (IRB) and written informed consents were obtained from patients (Ethics committee reference number: ERC/R/332).
## Funding
None.
## Informed consent
Informed consent was taken from all individual participants.
## CRediT authorship contribution statement
Mahdi Norooz Oliaei: Conceived and designed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper. Hamid Riazi Esfahani: Conceived and designed the experiments; Performed the experiments; Contributed reagents, materials, analysis tools or data; Wrote the paper. MohammadSadegh Abrishamian: Contributed reagents, materials, analysis tools or data; Wrote the paper.
## Declaration of Competing Interest
None declared.
## References
1. Humayun M.S., Weiland J.D., Fujii G.Y., Greenberg R., Williamson R., Little J., Mech B., Cimmarusti V., Van Boemel G., Dagnelie G., de Juan E.. **Visual perception in a blind subject with a chronic microelectronic retinal prosthesis**. *Vis. Res.* (November 2003) **43** 2573-2581. DOI: 10.1016/s0042-6989(03)00457-7
2. Caspi A., Dorn J.D., Mc Clure K.H., Humayun M.S., Greenberg R.J., McMahon M.J.. **Feasibility study of a retinal prosthesis: spatial vision with a 16-electrode implant**. *Arch. Ophthalmol.* (2009) **127** 398-401. DOI: 10.1001/archophthalmol.2009.20
3. Humayun M.S., Dorn J.D., da Cruz Lyndon, Dagnelie G., Sahel J.A., Stanga P.E., Cideciyan A.V., Duncan J.L., Eliott D., Filley E., Ho A.C., Santos A., Safran A.B., Arditi A., Del Priore L.V., Greenberg R.J.. **Interim results from the international trial of second sight's visual prosthesis**. *Ophthalmology* (April 2012) **119** 779-788. DOI: 10.1016/j.ophtha.2011.09.028
4. Watterson William J., Montgomery Rick D., Taylor Richard P.. **Modeling the improved visual acuity using photodiode based retinal implants featuring fractal electrodes**. *Front. Neurosci.* (April 2018) **12**. DOI: 10.3389/fnins.2018.00277
5. Watterson W.J., Montgomery R.D., Taylor R.P.. **Fractal electrodes as a generic interface for stimulating neurons**. *Sci. Rep.* (July 2017) **7** 1-9. DOI: 10.1038/s41598-017-06762-3
6. Watterson W.J.. (2017)
7. 7https://blogs.uoregon.edu/richardtaylor/files/2018/01/Fracret-Market-Analysis-xsdk30.ppsx
8. NoroozOliaei M., Riazi Esfahani H., Abrishamian M.S.. **Graphene coated dielectric resonator antenna for modeling the photoreceptors at visible spectrum**. *Heliyon (Cell Press)* (2022) **8**. DOI: 10.1016/j.heliyon.2022.e09611
9. 9R. Milo, R. Philips, Cell Biology by the Numbers, ISBN 978-0-8153-4537-4, Taylor & Francis Group.
10. Hood D.C., Odel J.G., Chen C.S., Winn B.J.. **The multifocal electroretinogram**. *J. Neuro-Ophthalmol.* (2003) **23** 225-235. DOI: 10.1097/00041327-200309000-00008
11. Hood D.C., Bach M., Brigell M., Keating D., Kondo M., Lyons J.S., Marmor M.F., McCulloch D.L., Palmowski-Wolfe A.M.. **International society for clinical electrophysiology of vision, ISCEV standard for clinical multifocal electroretinography (mfERG) (2011 edition)**. *Doc. Ophthalmol.* (2012) **124** 1-13. DOI: 10.1007/s10633-011-9296-8
12. Khojasteh Hassan, Riazi-Esfahani Hamid, Khalili Pour Elias, Faghihi Hooshang, Ghassemi Fariba, Bazvand Fatemeh, Mahmoudzadeh Raziyeh, Salabati Mirataollah, Mirghorbani Masoud, Riazi Esfahani Mohammad. **Multifocal electroretinogram in diabetic macular edema and its correlation with different optical coherence tomography features**. *Int. Ophthalmol.* (2020) **40** 571-581. DOI: 10.1007/s10792-019-01215-4
13. Mazahery Tehrani Neda, Riazi-Esfahani Hamid, Jafarzadehpur Ebrahim, Mirzajani Ali, Talebi Hossein, Amini Abdulrahim, Mazloumi Mehdi, Roohipoor Ramak, Riazi-Esfahani Mohammad. **Multifocal electroretinogram in diabetic macular edema; correlation with visual acuity and optical coherence tomography**. *J. Ophthalmic Vis. Res.* (2015) **10** 165-171. DOI: 10.4103/2008-322X.163773
14. Miguel-Jimenez Juan M., Ortega Sergio, Boquete Luciano, Rodríguez-Ascariz José M., Blanco Román. **Multifocal ERG wavelet packet decomposition applied to glaucoma diagnosis**. *Biomed. Eng. Online* (2011) **10**. DOI: 10.1186/1475-925X-10-37
15. Wright Tom, Cortese Filomeno, Nilsson Josefin, Westall Carol. **Analysis of multifocal electroretinograms from a population with type 1 diabetes using partial least squares reveals spatial and temporal distribution of changes to retinal function**. *Doc. Ophthalmol.* (2012) **125** 31-42. DOI: 10.1007/s10633-012-9330-5
16. 16The Human Eye, Chapter 9https://web.stanford.edu>class>archive>stanford.lecture.04.pdf
17. **Introduction human vision light, color, eyes, etc.**
18. Keyrouz S., Caratelli D.. **Dielectric resonator antennas: basic concepts, design guidelines, and recent developments at millimeter-wave frequencies**. *Int. J. Antennas Propag.* (2016). DOI: 10.1155/2016/6075680
|
---
title: Combining three independent pathological stressors induces a heart failure
with preserved ejection fraction phenotype
authors:
- Yijia Li
- Hajime Kubo
- Daohai Yu
- Yijun Yang
- Jaslyn P. Johnson
- Deborah M. Eaton
- Remus M. Berretta
- Michael Foster
- Timothy A. McKinsey
- Jun Yu
- John W. Elrod
- Xiongwen Chen
- Steven R. Houser
journal: American Journal of Physiology - Heart and Circulatory Physiology
year: 2023
pmcid: PMC9988529
doi: 10.1152/ajpheart.00594.2022
license: CC BY 4.0
---
# Combining three independent pathological stressors induces a heart failure with preserved ejection fraction phenotype
## Abstract
Heart failure (HF) with preserved ejection fraction (HFpEF) is defined as HF with an ejection fraction (EF) ≥ $50\%$ and elevated cardiac diastolic filling pressures. The underlying causes of HFpEF are multifactorial and not well-defined. A transgenic mouse with low levels of cardiomyocyte (CM)-specific inducible Cavβ2a expression (β2a-Tg mice) showed increased cytosolic CM Ca2+, and modest levels of CM hypertrophy, and fibrosis. This study aimed to determine if β2a-Tg mice develop an HFpEF phenotype when challenged with two additional stressors, high-fat diet (HFD) and Nω-nitro-l-arginine methyl ester (l-NAME, LN). Four-month-old wild-type (WT) and β2a-Tg mice were given either normal chow (WT-N, β2a-N) or HFD and/or l-NAME (WT-HFD, WT-LN, WT-HFD-LN, β2a-HFD, β2a-LN, and β2a-HFD-LN). Some animals were treated with the histone deacetylase (HDAC) (hypertrophy regulators) inhibitor suberoylanilide hydroxamic acid (SAHA) (β2a-HFD-LN-SAHA). Echocardiography was performed monthly. After 4 mo of treatment, terminal studies were performed including invasive hemodynamics and organs weight measurements. Cardiac tissue was collected. Four months of HFD plus l-NAME treatment did not induce a profound HFpEF phenotype in FVB WT mice. β2a-HFD-LN (3-Hit) mice developed features of HFpEF, including increased atrial natriuretic peptide (ANP) levels, preserved EF, diastolic dysfunction, robust CM hypertrophy, increased M2-macrophage population, and myocardial fibrosis. SAHA reduced the HFpEF phenotype in the 3-Hit mouse model, by attenuating these effects. The 3-Hit mouse model induced a reliable HFpEF phenotype with CM hypertrophy, cardiac fibrosis, and increased M2-macrophage population. This model could be used for identifying and preclinical testing of novel therapeutic strategies.
NEW & NOTEWORTHY Our study shows that three independent pathological stressors (increased Ca2+ influx, high-fat diet, and l-NAME) together produce a profound HFpEF phenotype. The primary mechanisms include HDAC-dependent-CM hypertrophy, necrosis, increased M2-macrophage population, fibroblast activation, and myocardial fibrosis. A role for HDAC activation in the HFpEF phenotype was shown in studies with SAHA treatment, which prevented the severe HFpEF phenotype. This “3-Hit” mouse model could be helpful in identifying novel therapeutic strategies to treat HFpEF.
## INTRODUCTION
Heart failure (HF) with preserved ejection fraction (HFpEF) is a complex clinical syndrome defined as HF with left ventricular ejection fraction ≥$50\%$ with elevated LV filling pressures at rest or during exercise [1]. It is a major public health problem [2]. HFpEF has a high prevalence of $1.1\%$–$1.5\%$ in the general population and accounts for ∼$50\%$ of all HF cases. HFpEF prevalence is rising by ∼$1\%$ per year, likely because of an aging population and ongoing epidemics of hypertension, obesity, and diabetes mellitus (3–6). HFpEF is also characterized by high morbidity and mortality. After hospitalization for HF, the 5-yr survival rate of HFpEF is $50\%$ [7], and every second patient reenters the hospital within 6 mo after the previous hospitalization [8]. Better understanding and treatment of this disorder are clearly needed.
HFpEF is recognized as a multiorgan, systemic syndrome [4] in which cardiac, pulmonary, renal, skeletal, immune, inflammatory, metabolic, and other components combine to cause symptoms and disease [9]. The cellular and molecular mechanisms that underlie this syndrome are not well understood, but the HFpEF cardiac phenotype includes cardiac hypertrophy, myocardial fibrosis, Ca2+ signaling pathway defects, inflammation, mitochondrial, and metabolic defects [4, 9, 10]. Animal models that recapitulate critical HFpEF features are needed to better understand HFpEF and identify targets for novel therapies.
HFpEF animal models include Dahl salt-sensitive rats [11], spontaneously hypertensive rats [12], mice and rats with aortic constriction [13], aging models [14], aortic-banded cats [15], DOCA and salt-loaded pigs receiving a high-fat diet [16], and Nω-nitro-l-arginine methyl ester (l-NAME) plus high-fat diet (HFD) mice [17]. Although some multifactorial models develop features of human HFpEF, most of the preclinical animal HFpEF models fail to meet the HFpEF clinical criteria like heart failure association pretest assessment, echocardiography and natriuretic peptide, functional testing, Final etiology (HFA-PEFF) diagnostic algorithm [18]. The HFA-PEFF (heart failure association pretest assessment, echocardiography and natriuretic peptide, functional testing, Final etiology) algorithm is a stepwise approach based on expert consensus to establish diagnosis in patients with suspected HFpEF [19]. The HFA-PEFF score is determined by scoring the natriuretic peptide levels and the echocardiographic findings of cardiac function and structure. If the HFA-PEFF score is ≥5 points, HFpEF is diagnosed, and if the HFA-PEFF score is ≤1 point, HFpEF could be excluded. If the HFA-PEFF score is between 2 and 4 points, the diagnosis for HFpEF needs further evaluation. Preclinical animal models with critical HFpEF features could help to define the complex HFpEF pathophysiology and to test putative HFpEF treatments. To date, there is only one successful clinical trial in HFpEF [20], and additional treatments are needed [1].
The aim of the current study was to characterize a novel mouse model that combines three pathological stressors: 1) increased calcium (Ca2+) influx caused by cardiomyocyte (CM)-specific expression of the l-type Ca2+ channel (Cav) β2a-subunit (β2a-Tg mice), 2) high-fat diet (HFD), and 3) l-NAME (LN, nitric oxide synthase inhibitors) (termed the 3-Hit model). Schiattarella et al. [ 17] were the first to present a 2-Hit (HFD and l-NAME) mouse model with the C57BL/6 strain that developed features of human HFpEF. They observed that mice subjected to both stress factors developed a HFpEF phenotype, including lung congestion and reduced exercise tolerance with increased natriuretic peptides. However, this profound HFpEF phenotype was not observed in FVB wild-type (WT) mice subjected to HFD and l-NAME for 4 mo.
Our results show that in FVB WT mice, all three stressors needed to be present to induce phenotypic features of human HFpEF. This study also shows that the expression of histone deacetylases (HDACs, CM hypertrophy regulators) is increased in 3-Hit mice and is associated with CM hypertrophy and cell necrosis, followed by an increased cardiac M2-macrophage population, fibroblast activation, and myocardial fibrosis. The HDAC inhibitor suberoylanilide hydroxamic acid (SAHA) reduced the HFpEF phenotype in 3-Hit mice.
## METHODS
All experiments involving animals conformed to the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health (NIH Publication, 8th ed., Revised 2011) and were approved by the Temple University Institutional Animal Care and Use Committee. In addition, the studies complied with all ethical regulations.
## Experimental Animals
FVB mouse strain was used in our study. Cardiac myocyte specific [α-myosin heavy chain (MHC) promoter] with inducible tetracycline-activator (tTA) expression of the β2a-subunit of L-type Ca2+ channel (Cavβ2a) was used. This β2a-Tg with relatively low expression level was documented (Supplemental Fig. S3; all Supplemental material is available at https://www.doi.org/10.6084/m9.figshare.22068815) [21]. The overexpression of β2a-subunit increases the open probability and membrane trafficking of the pore-forming Cav1.2α1c-subunit, which further increased Ca2+ influx in cardiomyocytes as previously reported [21, 22].
The β2a-Tg mice were established with the inducible (tet-off), bitransgenic system [22] (Supplemental Fig. S3). Mice with the tetracycline transactivator (tTA) driver gene and the Cavβ2a gene (double transgenic, DTG) without the doxycycline-containing chow were used as our β2a-Tg experimental group. Doxycycline is a derivative of tetracycline and hence represses β2a transgene expression. The Cavβ2a transgene was not expressed until adulthood to avoid developmental complications [22]. For each litter, β2a-Tg mice were separated into different treatment cohorts, and their WT mice littermates were separated into corresponding treatments as well. Sex-matched animals (β2a-Tg and WT mice) were given different treatments at the age of 4 mo when the Cavβ2a gene had been fully expressed.
## Special Diet and Water Treatment
Both WT and β2a-Tg mice were housed in an animal room with a 12-h:12-h light/dark cycle from 6:00 am to 6:00 pm, a temperature of 22 ± 3°C, relative humidity of 50 ± $6\%$, and free access to food (Cat. No. 2916, Teklad for normal chow diet groups and D12492, Research Diet for the high-fat diet groups) and water [tap water, or Nω-nitro-l-arginine methyl ester; 0.5 g/L, Cayman Chemical], or l-NAME with suberoylanilide hydroxamic acid (SAHA, vorinostat, 670 mg/L, Biogems; 670 mg SAHA dissolved by 6.7 mL DMSO and 20 mL PGE300) (Detailed study groups provided in Supplemental Table S1).
## Study Design
Four-month-old sex-matched WT and β2a-Tg mice were given either normal chow (WT-N and β2a-N) with normal water or a high-fat diet (HFD) with/without l-NAME in water (WT-HFD, WT-LN, WT-HFD-LN, β2a-HFD, β2a-LN, and β2a-HFD-LN) for 4 mo. SAHA treatment was given to another group of 4-mo-old sex-matched β2a-Tg mice together with HFD and l-NAME (β2a-HFD-LN-SAHA) for 4 mo as well. Echocardiography was performed at baseline (4 mo old, before treatment) and once a month after different treatments. Long-term animal survival and body weight (BW) were recorded during the 4-mo follow-up (Supplemental Fig. S1).
After 4 mo of treatment, mice were terminated using inhaled isoflurane (Butler Shein Animal Health, Dublin Ohio) after measurement of hemodynamics parameters. BW weight was recorded at terminal time points. Organs, including heart, lung, liver, kidney, and spleen, were carefully trimmed and collected, rinsed with PBS, and weighed after blotting off the excess fluid. Organ weight-to-body weight ratios were calculated. Heart tissues were excised and cut into two parts (basal and apical). The basal part containing part of papillary muscles was fixed with $10\%$ formalin, then paraffin embedded for histology following previously described protocols [23, 24]. The apical part was snap-frozen in liquid nitrogen for molecular analysis. Plasma samples were collected for further study. The analysis of histology such as cardiomyocyte cross-sectional area (CSA), Picrosirius red staining, and echocardiography was performed by investigators blinded to groups.
## Echocardiography in Animal Study
Transthoracic echocardiography was performed using a Vevo2100 ultrasound system (VisualSonics; Toronto, ON, Canada). In brief, mice were placed in the supine position on a heated platform with all legs taped to electrocardiographic electrodes for recording. Mice were initially anesthetized with $2\%$ isoflurane and then $1\%$ during the ultrasound procedure to maintain a heart rate between 450 and 500 beats/min. The mouse’s body temperature was maintained within a range of 37.0 ± 0.5°C. Hair was removed from the chest using chemical hair remover before imaging.
Images were obtained in the short-axis B-mode, long-axis B-mode, and M-mode at the level of the midpapillary muscles for analysis of systolic function and dimensions. Parameters include diastolic left ventricular anterior wall thicknesses (LVAWd), end-diastolic left ventricular posterior wall thickness (LVPWd), end-diastolic left ventricular internal diameter (LVIDd), LV ejection fraction (LVEF), LV fractional shortening (LVFS), and left atrial diameter (LA). Diastolic function was determined using B-mode at parasternal long-axis view and apical four-chamber view, pulsed-Doppler (PW), and tissue-Doppler imaging (TDI) as previously described [23, 25]. The left atrial diameter was measured at the parasternal long-axis view. PW was used to obtain the mitral inflow E, and TDI was used to measure the e′ wave. E/e′ was then calculated. Long-axis and short-axis B-mode images were collected for speckle-tracking strain analysis. Parameters were measured offline with VevoLab v3.2.6 (VisualSonics).
## Invasive Hemodynamics (In Vivo Intra-LV Pressure Measurements)
Invasive hemodynamics were performed after 4 mo of treatment. Briefly, intra-LV pressure was measured with a 1.4-Fr Millar pressure catheter (SPR-1000, Millar Instruments, Houston, TX) connected to an AD Instruments Power-Lab $\frac{16}{30}$ (ADInstruments, Colorado Springs, CO) with LabChart Pro 7.0 software as previously reported. After mice were anesthetized with $2\%$ isoflurane to maintain HR in the 450–500 beats/min range, a neck incision along the midline was made, and the right carotid artery was exposed. The pressure catheter was inserted into the right common carotid artery and advanced into the left ventricular (LV) chamber to measure left ventricular pressures and volumes. Blood pressure was recorded when the catheter was in the right common carotid artery. After entering in the LV, the catheter was carefully adjusted to avoid direct contact with the ventricular wall so that smooth intra-LV pressure traces could be clearly recorded. Data were analyzed offline with the blood pressure module in the LabChart7.0 software.
## Histology and Immunofluorescence Staining
Four to six mice were included in each group. After the mice were euthanized, the hearts were excised and cut into two parts (basal and apical). The basal portion containing part of the papillary muscles was fixed in $10\%$ neutral buffered formalin and then embedded in paraffin. Three 5-μm-thick sections of basal pieces (one section per piece) were sliced, deparaffinized (xylene, Fisher Scientific, Fair Lawn, NJ), dehydrated ($100\%$, $95\%$, $70\%$, and $50\%$ ethanol, sequentially; Fisher Scientific, Fair Lawn, NJ) for the following staining: Wheat-germ agglutinin (WGA, Life Technologies W11261, 1:100) [26] and nuclei (4′,6-diamidino-2-phenylinodole, DAPI, 268298 Millipore, 1:1,000) [26] staining was used to determine cardiomyocyte cross-sectional area (CSA). Images were taken using Nikon Eclipse Ti Confocal microscope, and at least 12 fields of view were taken of the left ventricle from three sections of the heart. Cardiomyocyte CSA was analyzed using NIH ImageJ software. Interstitial fibrosis was detected by Picrosirius red staining using a kit (ab150681, Abcam). Von Kossa staining was performed to detect Ca2+ deposits according to the manufacturer’s protocol (ab150687; Abcam). Pictures were taken using a Nikon Eclipse Ti Confocal microscope with DS-Ri2 light camera. At least 15 views from each animal that did not include vessels were analyzed. Fibrosis (red) and nonfibrosis (pink) areas were calculated with the “color threshold” tool from ImageJ software (v.1.49v; NIH).Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining was performed with the DeadEnd Fluorometric TUNEL System kit (Promega, Madison, WI) as previously described [27]. Briefly, after being deparaffinized and dehydrated, the slides were then fixed in $4\%$ paraformaldehyde solution in PBS (Affymetrix, Cleveland, OH). After being washed in PBS, the tissue was digested by incubation with the Proteinase K included in the Promega kit for 8 min as per protocol instructions. Slides were then washed and incubated with the labeling cocktail for 1 h at 37°C. The reaction was then stopped with the included SSC solution, and the slides were washed. The slides were fixed once more with paraformaldehyde before α-sarcomeric actin (Sigma, A2172, 1:500, RRID: AB_476695) [26], WGA, and DAPI immunofluorescence staining. Images were taken using Nikon Eclipse Ti Confocal microscope, and at least 12 fields of view were taken of the left ventricle using a ×20 objective. NIS *Confocal analysis* software was used to analyze the images. When determining the immune response among different groups, primary antibodies anti-CD45 (AF114, R&D Systems, 1:100, RRID: AB_442146) [28], anti-CD68 (MAB10114, R&D Systems, 1:100, RRID: AB_621929), anti-CD206 (AF2535, R&D Systems, 1:100, RRID: AB_2063012), anti-α-smooth muscle actin (ab5694, Abcam, 1:100, RRID: AB_2223021), anti-protein disulfide isomerase/P4HB (NB 300-517, Novus Biologicals, 1:100, RRID: AB_531260) [29], and anti-phospho-Smad2-S$\frac{465}{467}$ + Smad3-S$\frac{423}{425}$ (AP0548, ABclonal, 1:100, RRID: AB_2771541) [30] antibody was used for immunofluorescent staining of heart tissues. The secondary antibodies [26] were rhodamine Red-X (RRX) AffiniPure donkey anti-goat IgG (705-295-147, 1:100, RRID: AB_2340423), rhodamine (TRITC) AffiniPure donkey anti-rabbit IgG (711-095-152, 1:100, RRID: AB_2315776), rhodamine Red-X (RRX) AffiniPure donkey anti-mouse IgM (715-295-020, 1:100, RRID:AB_2340829), and fluorescein (FITC) AffiniPure donkey anti-rabbit (711-095-152, 1:100, RRID:AB_2315776) from Jackson ImmunoResearch. Nuclei were stained with DAPI.
## Quantitative Real-Time PCR
Total RNA was extracted from snap-frozen myocardial tissue using miRNeasy Mini kit (Qiagen) following the manufacturer’s instructions and then digested with DNase I (18068, Invitrogen). cDNA was synthesized with SuperScript III first strand (18068, Invitrogen) as previously described [25, 26]. Real-time PCR was performed using the Quantifast Sybrgreen PCR kit (204057, Qiagen) and the QuantStudio 3 Real-Time PCR System (A28567, Thermo Fisher). Ct values were normalized with respect to β2-microglobulin (β2M). Fold changes were calculated with respect to WT-N mice compared with different treatment groups. Fold changes were calculated with respect to HDAC1 when compared among different HDACs.
The following primer sets were used (forward, reverse): β2M, 5′- ATGTGAGGCGGGTGGAACTG, 5′- CTCGGTGACCCTG GTCTTTCTG; atrial natriuretic peptide (ANP), 5′- GCCCTGAGTGAGCAGACTG, 5′- GGAAGCTGTTGCAGCCTA; HDAC1, 5′- GTCCGGTGTTTGATGGCTTG, 5′- GCAGTGGGTAGTTCACAGCA; HDAC2, 5′- TATCCCGCTCTGTGCCCTAC, 5′- GAGGCTTCATGGGATGACCC; HDAC3, 5′- GACTTCTACCAGCCGACGTG, 5′- GCTTCTGGCCTGCTGTAGTT; HDAC8, 5′- CTGGACATACTTGACCGGGG, 5′- ACCGCTTGCATCAACACACT; HDAC4, 5′- GGGAGCAGCATCATGGTTCAA, 5′- TGAGAACTGGTGGTCCAAGC; HDAC5, 5′- AGAGTGACGTCTCCGAATGTTG, 5′- AGGAGTCCGTGGCAGGATTT; HDAC6, 5′- AGATCTGCGCGAGTGGAAG, 5′- CTCTCTGATGGCATGGAGCC; HDAC7, 5′- TATTCCCTACAGCCTGCCCACT, 5′- ACAGTGGGGCATGAGAGACT; HDAC9, 5′- CCATTGCCACGTGAACAACC, 5′- GACGACAGGATCCACCACAG; TGF-β, 5′- GCCCGAAGCGGACTACTATG,5′-TTTGGGGCTGATCCCGTTG; FN1, 5′- AGAAGACAGGACAGGAAGCTC, 5′- ATGGCGTAATGGGAAACCGT; LOX, 5′- TTCCAAGCTGGTTTCTCGCC, 5′- GTCCGATGTCCCTTGGTTCT; MMP9, 5′- CGCTCATGTACCCGCTGTAT, 5′- TGTCTGCCGGACTCAAAGAC.
## Western Blot Analysis
Lysates from snap-frozen heart tissues were prepared and analyzed using Western blot analysis as previously described [23]. The following primary antibodies were used: GAPDH (EMD Millipore Cat. No. MAB374, 1:1,000, RRID: AB_2107445) [21], anti-CACNB2 (calcium voltage-gated channel auxiliary subunit-β2) (A16037, ABclonal, 1:1,000, RRID:AB_2763475) [31]. The following secondary antibodies were used: 800CW donkey anti-rabbit (Cat. No. 926-32213, 1:5,000, RRID: AB_2715510) [26] and 680RD donkey anti-mouse (Cat. No. 926-68072, 1:5,000, RRID: AB_2814912) [26] purchased from LICOR (Lincoln, NE). Briefly, protein lysates were prepared from heart tissues, followed by denaturation with $12\%$ SDS and derivatization with 1× DNPH (2,4-dinitrophenylhydrazine). Derivatized protein samples (10 μg/well) were used for Western blot analysis and immunodetection. The gel used is Mini-PROTEAN TGX Precast Gels (Cat. No. 456-1086, Bio-RAD, 4–$15\%$). Western blot band intensities were quantified using Li-Cor Image Studio computer software.
## Statistical Analysis
Data are represented as means ± SE. The distributions of all continuous variables were tested for normality assumptions using the normal probability plot along with the Anderson–Darling normality test using GraphPad Prism. For parameters with a single measurement between two groups in the animal study, group comparisons were performed using the two-sample t test or the Mann–Whitney U test, depending on the data distribution. For parameters with a single measurement among multiple groups, the difference was evaluated using one-way ANOVA followed by the Tukey post hoc multiple comparison test. For body weight data and echocardiography parameters with repeated measures over time, linear mixed-effects models were used to estimate mean values at each assessment time point and to test treatment group differences at each time point as well as change versus baseline over time within each treatment group. In each linear mixed-effects model, time and treatment group were included as fixed effects along with its time-by-treatment group interaction term. Pairwise comparisons between various experimental groups under these mixed-effects models were performed using the Tukey post hoc multiple comparison tests. For in vivo data among groups under multiple treatments, the analysis was performed by two-way ANOVA, followed by the Tukey multiple comparisons test. Two-sided testing was used for all statistical comparisons. A P value of <0.05 was considered statistically significant. Data analyses were performed using the GraphPad Prism software (v.8.4.3, GraphPad Inc, La Jolla, CA) and/or SAS (v.9.4, SAS Institute Inc., Cary, NC).
## Effects of High Fat Diet, l-NAME, and HFD + LN in WT Mice
Four-month-old FVB wild-type (WT) mice were fed with either a normal chow diet (WT-N) or treated with a HFD and/or l-NAME in water (WT-HFD, WT-LN, and WT-HFD-LN) for 4 mo (Supplemental Fig. S1). HFD treatment caused an increase in both body weight (BW) and blood pressure, while l-NAME treatment increased the blood pressure of WT mice (Fig. 1B; Supplemental Fig. S2, A and B). There was no significant difference in survival rate among the four groups (Fig. 1A). Four months of HFD and/or l-NAME treatment led to a trend to increase heart weight (HW), but not significantly higher HW to body weight ratio (HW/BW) and HW to tibia length ratio HW/TL (Fig. 1, C and D; Supplemental Fig. S2C) compared with WT-N group. Concentric remodeling was observed in WT-HFD, WT-LN, and WT-HFD-LN groups, but was highest in the WT-HFD-LN group. Concentric remodeling was shown by thicker LV walls (Fig. 1, E and F), as measured by conventional echocardiography (ECHO), and greater cardiomyocyte cross-sectional area (CSA) (Fig. 1, M and O). ECHO analysis did not show significant cardiac systolic or diastolic dysfunction in any group. LVEF (Fig. 1G) was preserved and LV longitudinal and radial strain and left ventricular internal diameter (Fig. 1H, Supplemental Fig. S2, E and F) were not significantly changed in all groups. The similar left atrium (LA) diameter (Fig. 1I) and E/e′ ratio (Fig. 1J) measured by ECHO, as well as similar left ventricular (LV) end-diastolic pressure (EDP) (Fig. 1K) and maximal rate of LV pressure decrease (dP/dtmin) (Fig. 1L) from hemodynamic measurements indicated no significant diastolic dysfunction was present in WT-HFD, WT-LN, and WT-HFD-LN mice. Modest LV fibrosis was observed in the WT-LN and WT-HFD-LN groups (Fig. 1, M and N). WT-HFD-LN mice had the most cardiac remodeling compared versus other groups, but the lung weight (LuW) and gene expression level of atrial natriuretic peptide (ANP), two heart failure indicators, were not significantly increased (Supplemental Fig. S2, D and G). These data suggest that 4 mo of HFD plus l-NAME treatment induced some cardiac hypertrophy did not induce a robust HFpEF phenotype in FVB mice.
**Figure 1.:** *Effects of high-fat diet (HFD), Nω-nitro-l-arginine methyl ester (l-NAME, LN), or HFD + l-NAME on the heart failure with preserved ejection fraction (HFpEF) phenotype in wild-type (WT) mice. A: survival rate from 4-mo follow-up. Body weight (BW; B), heart weight (HW; C), and ratio of HW to tibia length (HW/TL; D) at the time of euthanasia. Conventional and sophisticated echocardiography data showing left ventricular (LV) wall thickness (E and F), LV ejection fraction (LVEF; G), LV longitudinal strain (H), left atrium (LA) diameter (I), and ratio between early mitral inflow velocity (E) and mitral annular early diastolic velocity (e′) (E/e′; J). Hemodynamics data showing LV end-diastolic pressure (LVEDP; K) and maximum rate of pressure decay (dP/dtmin; L). M: representative images of hearts stained with Picrosirius red and wheat germ agglutinin (WGA). N: quantification of the percentage of Picrosirius red-positive area. O: quantification of cardiomyocyte cross-sectional area (CSA). DAPI, 4′,6-diamidino-2-phenylindole; LVAWd, end-diastolic left ventricular anterior wall thicknesses; LVEF, left ventricular ejection fraction; LVPWd, end-diastolic left ventricular posterior wall thickness; N, normal chow diet. Data shown are means ± SE. Tukey post hoc multiple comparison adjusted P values: E–J: P < 0.05, *WT-N vs. WT-HFD-LN, #WT-N vs. WT-HFD, and &WT-N vs. WT-LN; and other panels: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Total number of animals (n) and number of females and males included in each group are reported in the Supplemental Table.*
## Effects of High-Fat Diet and/or l-NAME on Mice with Low Levels of Expression of a Cardiac-Specific LTCC β2a-Subunit (3-Hit Mice)
Low levels of expression of cardiac-specific Cavβ2a-subunit were found in 4-mo-old mice, as previously reported [21] and remained constant through 8 mo of age (2.8-fold higher levels) (Supplemental Fig. S3, A–C). β2a-Tg mice were fed with a normal chow diet (β2a-N) or treated with a high-fat diet and/or l-NAME in water (β2a-HFD, β2a-LN, and β2a-HFD-LN; 3-Hit) for 4 mo (Supplemental Fig. S1). HFD treatment showed a trend to increase the mean BW (Fig. 2B), while l-NAME treatment increased the blood pressure of β2a mice (Supplemental Fig. S3, D and E).
**Figure 2.:** *Effects of cardiac-specific L-type Ca2+ channel (LTCC) β2a-subunit expression together with high-fat diet (HFD) and Nω-nitro-l-arginine methyl ester (l-NAME, LN) treatment (3-Hit) on the heart failure with preserved ejection fraction (HFpEF) phenotype. A: survival rate from 4-mo follow-up. Body weight (BW; B), heart weight (HW; C), and ratio of HW to tibia length (HW/TL; D) at the time of euthanasia. Conventional and sophisticated echocardiography data showing left ventricular (LV) wall thickness (E and F), LV ejection fraction (LVEF; G), LV longitudinal strain (H), left atrium (LA) diameter (I), and ratio between early mitral inflow velocity (E), and mitral annular early diastolic velocity (e′) (E/e′; J). Hemodynamics data showing LV end-diastolic pressure (LVEDP; K) and maximum rate of pressure decay (dP/dtmin; L). M: representative images of hearts stained with Picrosirius red and wheat germ agglutinin (WGA). N: quantification of the percentage of Picrosirius red-positive area. O: quantification of cardiomyocyte cross-sectional area (CSA). β2a, transgenic mouse with low levels of cardiomyocyte (CM)-specific inducible Cavβ2a expression; DAPI, 4′,6-diamidino-2-phenylindole; LVAWd: end-diastolic left ventricular anterior wall thicknesses; LVPWd, end-diastolic left ventricular posterior wall thickness; N, normal chow diet. Data shown are means ± SE. Tukey post hoc multiple comparison adjusted P values: E–J: P < 0.05, *β2a-N vs. β2a-HFD-LN, #β2a-N vs. β2a-HFD, &β2a-N vs. β2a-LN, ^β2a-HFD vs. β2a-HFD-LN, +β2a-LN vs. β2a-HFD-LN; and other panels: *P < 0.05, **P < 0.01, ****P < 0.0001. Total number of animals (n) and number of females and males included in each group are reported in the Supplemental Table.*
The data from four β2a groups (β2a-N, β2a-HFD, β2a-LN, and β2a-HFD-LN; 3-Hit) were compared (see Fig. 2). β2a-HFD-LN (3-Hit) mice had premature mortality compared with β2a-N mice (Fig. 2A), and a more severe cardiac phenotype compared with β2a-N, β2a-HFD, and β2a-LN mice. Cardiac hypertrophy in β2a-HFD-LN mice was shown by a greater HW, HW/TL (Fig. 2, C and D), thicker LV walls (Fig. 2, E and F), and a greater cardiomyocyte cross-sectional area (CSA) (Fig. 2, M and O) versus other groups. ECHO analysis showed all four groups of mice had preserved LVEF (>$50\%$) (Fig. 2G), but β2a-HFD-LN mice had significant decreases in LV longitudinal (<$16\%$) strain (Fig. 2H), suggesting some impairment of systolic function. ECHO and hemodynamic measurements both showed a more significant cardiac diastolic dysfunction in β2a-HFD-LN mice than in any other group. There was a significant increase in left atrium (LA) diameter and E/e′ ratio (Fig. 2, I and J) measured by ECHO, and an increased EDP and decreased dP/dtmin determined by invasive hemodynamics (Fig. 2, K and L) in β2a-HFD-LN mice. In addition, LV fibrosis was most severe in the β2a-HFD-LN group (Fig. 2, M and N). These data document that 3-Hit mice had the most severe pathological phenotype of the mice studied.
## The 3-Hit Mouse Model Produces a Profound HFpEF Phenotype, Which Can Be Reduced by SAHA Treatment
To clarify the similarities and differences between WT and β2a-Tg mice after treatment with HFD and LN, and the effect of SAHA [suberoylanilide hydroxamic acid, vorinostat, a pan-HDAC activity inhibitor [32, 33]] treatment, the data from five groups (WT-N, WT-HFD-LN, β2a-N, β2a-HFD-LN, and β2a-HFD-LN-SAHA) were compared.
The death rate during the 4-mo study was significantly greater in β2a-HFD-LN animals compared with WT-N (Fig. 3A). The BW and blood pressure (systolic and diastolic pressure) were significantly greater in β2a-HFD-LN mice when compared with WT-N mice (Supplemental Fig. S4, A–C). More severe cardiac hypertrophy in β2a-HFD-LN mice was shown by significantly increased HW, HW/TL, HW/BW, and LV wall thickness (Supplemental Fig. S4, D and E; Fig. 3, B–D) versus other groups. In addition, the β2a-Tg mice had thicker LV walls at 4 mo of age compared with WT mice (Fig. 3, C and D), which was consistent with our previous findings [21].
**Figure 3.:** *The 3-Hit mouse model produces a profound heart failure with preserved ejection fraction (HFpEF) phenotype, which can be reduced by suberoylanilide hydroxamic acid (SAHA) treatment. β2a, transgenic mouse with low levels of cardiomyocyte (CM)-specific inducible Cavβ2a expression; HFD, high-fat diet; l-NAME (or LN), Nω-nitro-l-arginine methyl ester; N, normal diet; WT, wild type. Data shown for WT-N, WT-HFD-LN, β2a-N, and β2a-HFD-LN groups in are the same as those shown in Figs. 1 and 2, and Supplemental Fig. S2. Statistical comparisons being made here are unique. The β2a-HFD-LN-SAHA group was added into the statistic comparation. A: survival rate from 4-mo follow-up. B: body weight (BW) at the time of euthanasia. Conventional and sophisticated echocardiography data showing left ventricular (LV) wall thickness (C and D), LV ejection fraction (LVEF; E), LV longitudinal strain (F), and left atrium (LA) diameter (G). H: LA weight-to-BW ratio at the time of euthanasia. I: conventional echocardiography data showing ratio between early mitral inflow velocity (E) and mitral annular early diastolic velocity (e′) (E/e′). Hemodynamics data showing end-diastolic pressure (EDP; J) and LV diastolic time constant (τ; K). L: expression level of atrial natriuretic peptide (ANP) in heart tissues by real-time polymerase chain reaction. Relative expression was calculated with respect to the WT-N group. HW, heart weight; LVAWd, end-diastolic left ventricular anterior wall thicknesses; LVPWd, end-diastolic left ventricular posterior wall thickness; TL, tibia length. Data shown are means ± SE. Tukey post hoc multiple comparison adjusted P values: C–G and I: P < 0.05, *WT-N vs. β2a-HFD-LN, #WT-N vs. β2a-N,@WT-N vs. WT-HFD-LN, &β2a-N vs. β2a-HFD-LN, and $WT-HFD-LN vs. β2a-HFD-LN; and other panels: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Total number of animals (n) and number of females and males included in each group are reported in the Supplemental Table.*
ECHO analysis showed β2a-HFD-LN mice had preserved LVEF (>$50\%$) (Fig. 3E), but significantly decreased LV longitudinal strain (<$16\%$) (Fig. 3F). Histological, ECHO, and hemodynamic measurements showed significant cardiac diastolic dysfunction. The evidence included a significant increase in left atrium (LA) diameter (Fig. 3G), LA weight/BW (Fig. 3H), E/e′ ratio (Fig. 3I), and increased EDP (Fig. 3J), left ventricular diastolic time constant (τ) (Fig. 3K), and decreased dP/dtmin (Supplemental Fig. S4G). When compared with other groups, ANP gene expression was significantly greater in β2a-HFD-LN mice (Fig. 3L). These results show that the combination of three stressors, cardiac-specific β2a-subunit plus HFD plus l-NAME treatment (3-Hit) induces a profound HFpEF phenotype that could cause premature death.
SAHA treatment groups did not show significant decreases in survival rate compared with WT-N group (Fig. 3A). In addition, SAHA treatment caused a significant decrease in HW, HW/BW, HW/TL, and LV wall thickness versus β2a-HFD-LN mice (Fig. 3, B–D and Supplemental Fig. S4, D and E). SAHA treatment did not affect LVEF (Fig. 3E) but prevented the decrease in longitudinal (Fig. 3F) in β2a-HFD-LN mice. ECHO measurements documented significantly decreased LA diameter, LA weight/BW, and absolute E/e′ ratio (Fig. 3, G–I) in β2a-HFD-LN-SAHA mice indicating improved LV diastolic function. Lower EDP and τ in SAHA-treated mice was observed (Fig. 3, J and K). SAHA treatment also caused a significant decrease in ANP expression (Fig. 3L). These data show that SAHA treatment prevented early stage systolic dysfunction and diastolic dysfunction in the 3-Hit mice.
## The 3-Hit Mouse Has More Cardiomyocyte Hypertrophy and Necrosis, and SAHA Treatment Can Alleviate the Phenotype
HDACs are known to be centrally involved in pathological cardiac hypertrophy [34, 35]. Activation of Class I HDACs (HDACs 1, 2, 3, and 8) and Class IIb HDACs (HDAC6) are thought to promote pathological hypertrophy, whereas class II HDACs (HDACs 4, 5, 7, and 9) are thought to suppress cardiac hypertrophy. In WT-N and β2a-HFD-LN mice, the most abundant HDACs (Class I and Class IIb) expressed in the heart were HDACs 1, 2, 3, 8, 4, and 7 (Supplemental Fig. S5). HDAC1, 3, 6, and 8 were significantly higher in β2a-HFD-LN mice compared with WT-N, WT-HFD-LN, and β2a-N groups (Fig. 4A). SAHA treatment group showed a significant decrease in HDAC3 expression, and an increase in HDAC4 and 7 (Fig. 4, A and B).
**Figure 4.:** *The 3-Hit mouse model has a more severe cardiomyocyte hypertrophy and necrosis, which can be reduced by suberoylanilide hydroxamic acid (SAHA) treatment. β2a, transgenic mouse with low levels of cardiomyocyte (CM)-specific inducible Cavβ2a expression; HFD, high-fat diet; N, normal chow diet, l-NAME (LN), Nω-nitro-l-arginine methyl ester; WT, wild type. Representative images of wheat germ agglutinin (WGA)-stained hearts (C) and quantification of cardiomyocyte cross-sectional area (CSA) data (D) in WT-N, WT-HFD-LN, β2a-N, and β2a-HFD-LN groups are the same as those shown in Fig. 1, M and O, and Fig. 2, M and O. The statistical comparisons being made here are different than those made in Figs. 1 and 2. The β2a-HFD-LN-SAHA group was added into the statistic comparison. A and B: expression level of class I histone deacetylases (HDACs) 1, 2, 3 and 8, Class IIb HDAC6 (A) and class IIa HDACs 4 and 7 (B) in heart tissues by real-time polymerase chain reaction. Relative expression was calculated with respect to WT-N mice compared with different treatment groups. C: representative images of WGA-stained hearts and histological assessment of cardiac ventricular pathology by Von Kossa staining. D: quantification of cardiomyocyte CSA. E and F: representative images of terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)-stained hearts (E) and quantification of TUNEL-positive myocyte nuclei from hearts (F): α-sarcomeric actin (α-SA), red; TUNEL, green; and 4′,6-diamidino-2-phenylindole (DAPI), blue. Data shown are means ± SE. Tukey post hoc multiple comparison adjusted P values are reported here. *P < 0.05, **P < 0.01, ****P < 0.0001.*
β2a-HFD-LN, 3-Hit mice had the greatest increase in myocyte CSA compared with WT-N, and WT-HFD-LN, indicating the most severe myocyte hypertrophy (Fig. 4, C and D). The CM size in β2a-HFD-LN-SAHA group was smaller (Fig. 4, C and D) than in β2a-HFD-LN mice. There was no significant difference in the number of myocytes undergoing apoptosis (TUNEL staining) between WT-N and β2a-HFD-LN mice (Fig. 4, E and F). In contrast, Von Kossa staining showed that only β2a-HFD-LN mice had evidence of more Ca2+ deposition, suggestive of CM necrosis [22] (Fig. 4C). Collectively, these results show that SAHA treatment reduced cardiomyocyte hypertrophy and necrosis caused by HFD plus l-NAME treatment in β2a-Tg mice.
## The 3-Hit Mouse Has Increased Cardiac M2-Macrophages and Fibroblast Activation, Which is Reduced by SAHA Treatment
Cardiac hypertrophy and necrosis are associated with interstitial cardiac fibrosis in HFpEF [36, 37]. We counted CD45+, CD68+, and CD206+ cells, to estimate the number of macrophages, and M2-macrophages, respectively. Significantly more CD45+ monocytes were present in β2a-HFD-LN hearts than in other groups (Fig. 5, A and B). The β2a-HFD-LN mice also had a significantly higher ratio of CD68+/CD45+ macrophages, and more CD206+ CD68+ M2-like macrophage than in hearts from other groups (Fig. 5, C and D). SAHA treated β2a-HFD-LN hearts had fewer CD45+ monocytes, CD68+/CD45+ macrophages, and CD206+CD68+ profibrotic M2-macrophage population than in untreated β2a-HFD-LN hearts (Fig. 5, A–D).
**Figure 5.:** *The 3-Hit model has robust cardiac M2-macrophage infiltration and fibroblast activation, which can be reduced by suberoylanilide hydroxamic acid (SAHA) treatment. A: immunofluorescence staining of heart sections to show CD45+ (red) immune cell, CD45+ (red) CD68+ (green) macrophage, CD68+ (green) CD206+ (red) M2-macrophage, and phosphorylated Smad2/3 (pSmad2/3, green) in fibroblast (red). 4′,6-Diamidino-2-phenylindole (DAPI) was stained as blue. Data are expressed as percentage of CD45+ cell/total cells (B), percentage of CD68+ cell/CD45+ cell (C), and CD206+CD68+/total cell (D). E: expression level of transforming growth factor-β (TGFβ) in heart tissues by real-time polymerase chain reaction. Relative expression was calculated with respect to wild-type/normal chow diet (WT-N) mice compared with different treatment groups. F: quantification of mean intensity of pSmad2/3 for each group. β2a, transgenic mouse with low levels of cardiomyocyte (CM)-specific inducible Cavβ2a expression; HFD, high-fat diet; l-NAME (LN), Nω-nitro-l-arginine methyl ester; P4HB, protein disulfide isomerase/prolyl 4-hydroxylase. Data shown are means ± SE. Tukey post hoc multiple comparison adjusted P values are reported here. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.*
Macrophages are a potent transforming growth factor-β (TGFβ) producer [38]. TGFβ mRNA was significantly increased in β2a-HFD-LN hearts (Fig. 5E). The phosphorylation levels of Smad2 and 3 (pSmad$\frac{2}{3}$), two major molecules known to be downstream of TGFβ signaling, were also increased in β2a-HFD-LN hearts (Fig. 5F). The mRNA level of TGFβ and pSmad$\frac{2}{3}$ were significantly lower in β2a-HFD-LN-SAHA mice compared with β2a-HFD-LN mice (Fig. 5, E and F). These results support the idea that in this 3-Hit HFpEF model, profibrotic M2-macrophages activate TGFβ signaling to stimulate cardiac fibroblast activation (differentiation into myofibroblasts) to promote fibrosis, which was reduced by SAHA treatment.
## SAHA Treatment Reduces Myocardial Fibrosis in the 3-Hit Mouse
The expression of α-smooth muscle actin [α-SMA, a biomarker for activated myofibroblast [39, 40]], collagen, and fibronectin 1 (FN1), are upregulated by the transcription factors pSmad$\frac{2}{3}$ [41]. β2a-HFD-LN hearts had a higher α-SMA staining intensity (immunofluorescence) and more interstitial fibrosis (Picrosirius red staining) (Fig. 6, A–C) than found in other groups. In addition, the mRNA levels of FN1 and collagen cross-linking enzyme lysyl oxidase (LOX) were increased in β2a-HFD-LN hearts (Fig. 6, D and E). After SAHA treatment, α-SMA intensity was reduced, as were cardiac interstitial fibrosis, and mRNA level of FN1 and LOX when compared with β2a-HFD-LN mice (Fig. 6, A–E). Expression levels of matrix metalloproteinase-9 (MMP9), a proteinase involved in fibrosis [42], were not different among groups (Supplemental Fig. 4H).
**Figure 6.:** *The 3-Hit model has severe myocardial fibrosis, which can be reduced be suberoylanilide hydroxamic acid (SAHA) treatment. β2a, transgenic mouse with low levels of cardiomyocyte (CM)-specific inducible Cavβ2a expression; HFD, high-fat diet; N, normal chow diet, l-NAME (LN), Nω-nitro-l-arginine methyl ester; WT, wild type. The representative images of Picrosirius red-stained hearts (A) and quantification of the percentage of Picrosirius red-positive area (C) data of WT-N, WT-HFD-LN, β2a-N, and β2a-HFD-LN groups are the same as in Fig. 1, M and N, and Fig. 2, M and N. The statistical comparisons being made are different than those made in Figs. 1 and 2. The β2a-HFD-LN-SAHA group was added into the statistic comparation. A: representative images of hearts from 4 groups with α-smooth muscle actin (α-SMA) (red) immunofluorescence staining and Picrosirius red staining. Quantification of the α-SMA (red) intensity (B) and the percentage of Picrosirius red-positive area (C). Expression level of fibronectin 1 (FN1; D) and collagen cross-linking enzyme lysyl oxidase (LOX; E) in heart tissues by real-time polymerase chain reaction. Relative expression was calculated with respect to WT-N mice compared with different treatment groups. DAPI: 4′,6-diamidino-2-phenylindole. Data shown are means ± SE. Tukey post hoc multiple comparison adjusted P values are reported here. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.*
These results suggest that the combination of three stressors in β2a-HFD-LN hearts produces severe cardiac hypertrophy and cell necrosis with increased profibrotic M2-macrophage populations, TGFβ-dependent cardiac fibroblast activation, and myocardial fibrosis, ultimately leading to a profibrotic HFpEF phenotype. SAHA treatment prevents these HFpEF features.
## DISCUSSION
Heart failure with preserved ejection fraction (HFpEF) is a major public health problem with an increasing prevalence (3–6), high morbidity and mortality [7], and a high rehospitalization rates [8]. HFpEF treatments are not well established [1] and new therapies need to be developed. Animal models that have critical features of human HFpEF pathophysiology could help define the cellular and molecular bases of HFpEF induction and progression and identify targets for new therapies. This study explored the characteristics of a new 3-Hit mouse model [cardiomyocyte-specific increases in Ca2+ influx, plus high-fat diet (HFD) plus l-NAME] to determine if it has crucial phenotypic features of HFpEF.
HFpEF is a complex syndrome that has many causes in humans and the 3-Hit model characterized in this study was developed with the idea that more than one pathological stressor may be needed to induce a profound phenotype reminiscent of human HFpEF. Our 3-Hit model combines cardiac myocyte-specific Ca2+ stress, systemic metabolic stress, and vascular stress. Our results show that each of these three stressors alone induce modest cardiac phenotypes but do not cause a profound HFpEF phenotype. Combining HFD and l-NAME in WT mice did not cause a profound HFpEF phenotype, as also shown in a recent study [17]. Likewise, combining Ca2+ stress with one of the other two stressors also did not result in a profound phenotype. Only the combination of these three stressors was able to induce a profound HFpEF phenotype with premature death.
We identified a mechanistic role for HDAC dependent-CM hypertrophy and necrosis that was associated with increased profibrotic M2-macrophage population, fibroblast activation, and myocardial fibrosis. Finally, we showed that treatment of β2a-HFD-LN mice with SAHA (a pan-HDAC inhibitor) prevented the severe HFpEF phenotype and the associated premature death.
## l-NAME Can Cause Hypertension and Cardiac Remodeling but Alone Does Not Induce a Profound HFpEF Phenotype
Hypertension, and vascular dysfunction can cause cardiac remodeling that includes myocyte hypertrophy and fibrosis [43]. Chronic administration of the nitric oxide (NO) synthase inhibitor l-NAME is known to causes arterial hypertension in mouse model [44]. In the present experiments, treatment of WT mice with l-NAME caused hypertension but only induced modest changes in cardiac structure and function. These changes were not sufficient to cause a profound HFpEF phenotype (Fig. 1).
β2a-Tg mice have a basal hypertrophic phenotype with modest fibrosis (see discussion later). Treatment with l-NAME caused an increase in blood pressure but did not induce a profound exacerbation of the β2a phenotype (Fig. 2). These results show that the effects of l-NAME contribute to a cardiovascular phenotype, but at least with the conditions we employed, a pronounced phenotype was not observed when the effects of l-NAME alone were studied, and the phenotype was not substantially increased when l-NAME treatment was combined with HFD (in WT mice, Fig. 1) or β2a stress (Fig. 2).
## High-Fat Diet Causes Cardiac Remodeling but Alone Does Not Induce a Profound HFpEF Phenotype
HFD induces weight gain and metabolic disturbances in mice [45] and in humans [46]. The obesity epidemic in Western society is now clearly linked to the development of HFpEF in younger and younger patients [47, 48]. In the present study, HFD in WT mice caused modest cardiac hypertrophy and fibrosis but the overall phenotype was mild (Fig. 1). Adding l-NAME to WT mice fed a HFD induced an increase in blood pressure but did not substantially increase the cardiac remodeling response (Fig. 1; Supplemental Fig. S2). Feeding β2a-Tg mice a HFD caused them to gain weight but did not induce a significant exacerbation of the β2a phenotype and did not cause profound HFpEF signs and symptoms (Fig. 2).
A recent study from Dr. Hill’s laboratory [17] that the combination of two stressors (HFD and l-NAME) in C57BL/6 mice induced features of human HFpEF. Their results showed that with HFD + l-NAME treatment, WT mice already had increased BW, blood pressure, HW/TL, lung weight, cardiac ANP expression, cardiomyocyte CSA, and myocardial fibrosis. In addition, this 2-Hit mouse model showed preserved LVEF but decreased LV longitudinal strain, as well as a significantly altered LV diastolic function evaluated by both ECHO and hemodynamic.
In our study, we treated WT FVB mice with HFD and l-NAME for 4 mo. We observed some HFpEF features, including increased BW, blood pressure, preserved LVEF, increased cardiomyocyte CSA, and myocardial fibrosis (Fig. 1; Supplemental Fig. S2). However, we failed to observe a significantly increased HW/TL ratio, lung weight/TL ratio, and cardiac ANP expression (Fig. 1; Supplemental Fig. S2). In addition, the WT-HFD-LN groups did not show significantly decreased LV longitudinal and radial strain and LV diastolic dysfunction measured by both ECHO and hemodynamic (Fig. 1; Supplemental Fig. S2). These results suggest that FVB mice do not develop critical HFpEF features with the HFD + l-NAME.
The potential reasons for our results likely involve the distinct genetic backgrounds of FVB and C57BL/6 mice, which have been reported to have an influence on mouse baseline cardiac function and the response to different stimuli [49]. In addition, FVB and C57BL/6 mouse strains have different behaviors in relation to metabolism [50]. The C57BL/6 mouse also has a more robust effect on the innate immune system than seen in FVB mice after HFD [50]. The FVB mouse is considered diet‐resistant [51] and has low HFD-induced atherosclerosis susceptibility [52].
## Ca2+ Stress Can Induce Cardiac Hypertrophy without a Profound HFpEF Phenotype
It is well established that many cardiac diseases increase the work demands of cardiac myocytes [53, 54]. Increases in Ca2+ influx enhances contraction to meet increased physiological and pathological demands (21, 54–56). The increased cellular Ca2+ causes increased contraction but also elicits a host of responses including cardiac hypertrophy (21, 54–56), metabolic reprogramming [57], and can lead to cell death [22, 58]. There is a robust literature linking increased Ca2+ influx in disease to cardiac remodeling (21, 54–56, 59).
We used a genetically modified mouse with low levels of cardiac myocyte-specific expression of the β2a-subunit of the l-type Ca2+channel [21]. The present experiments show that low levels of β2a expression can induce a hypertrophic phenotype (Figs. 3 and 4). The Ca2+ disturbances induced in this mouse could be similar to those that are present in a variety of mild disease states. We choose this mouse model because it induced low levels of Ca2+ stress without a severe cell death phenotype [22]. Our new results show that adding HFD alone or l-NAME alone to β2a-Tg animals did not induce a severe exacerbation of the β2a phenotype (Fig. 2).
Our results show that any two combinations of Ca2+ stress, HFD, and l-NAME were not sufficient to cause a profound HFpEF phenotype with significant premature death. Why this occurs is not entirely clear since all three stressors cause some degree of cardiac hypertrophy and fibrosis. Our experiments show that when these three independent stressors were applied to FVB mice together, they collectively induced a severe HFpEF phenotype associated with significant levels of premature death (Figs. 3 and 4).
## Three-Hit Mouse Model Meets All Criteria Necessary for HFA-PEFF HFpEF Diagnosis
The clinical complexity of HFpEF and the lack of a single objective marker make diagnosing of HFpEF difficult. Recently, HFA-PEFF [60] clinical algorithms have been developed to improve and standardize the diagnosis of HFpEF. HFA-PEFF diagnostic algorithms, proposed by the Heart Failure Association (HFA) of the European Society of Cardiology, is a stepwise diagnostic algorithm, which can be easily and accurately calculated, and is useful for predicting composite cardiovascular events as well as HF-related events in patients with HFpEF (61–63). It assesses the pretest probability of HFpEF based on clinical features (including age and comorbidities) and cardiac functional and structural echocardiographic data, including morphological aspects of the LA and LV, as well as levels of natriuretic peptides [60]. Several different HFpEF preclinical models had been scored using HFA-PEFF [18]. Our 3-Hit mouse model showed comorbidity burden of heart failure (obesity and increased blood pressure), decreased global longitudinal strain (<$16\%$), diastolic dysfunction (increased absolute E/e′, ≥15), enlarged LA, increased heart weight, thicker LV wall, LV hypertrophy, and increased natriuretic peptides expression in heart tissue (Figs. 2, 3, and 4). These results show that the 3-Hit model meet the criteria necessary for an HFA-PEFF HFpEF diagnosis with a high score. In addition, while our 3-Hit model had preserved LVEF, we found a decreased LV longitudinal and radial strain. Strain is an ECHO parameter widely used in clinical work to detect mild cardiac systolic impairment at an early stage before LVEF decreases. Several clinical studies have already shown that patients with HFpEF are characterized by decreased LV strain [64, 65], and our 3-Hit model also mirrors the mild cardiac systolic impairment phenotype shown in these studies.
## 3-Hit Mouse Model Involves HDAC Activation
Cardiac hypertrophy is one of the best-studied aspects of HFpEF and is a common clinical feature of HFpEF [9]. Patients with HFpEF tend to have normal LV filling volumes, with variable degrees of LV wall thickening (66–68). Gupta et al. [ 68] showed that ∼$75\%$ of patients with HFpEF had significant cardiac hypertrophy. Cardiac hypertrophy is also associated with the diastolic dysfunction and elevated diastolic filling pressure observed in HFpEF [9, 69]. Pathophysiological cardiac hypertrophy in patients with HFpEF likely involves impaired Ca2+ handling [9], myocardial fibrosis [4, 9], oxidative stress [9, 70], cell death [9, 22, 58], metabolic reprogramming [57], and induction of fetal genes [9, 71]. All of these features are induced in the 3-Hit mouse model characterized in the present work.
One of the known molecular contributors to pathological cardiac hypertrophy is the activation of HDACs [72, 73], which remove N-acetyl-lysine from histone and non-histone proteins to induce cardiac hypertrophy, fibrosis, and diastolic dysfunction (35, 74–76). HDACs fall into four distinct classes (I, II, III, and IV). Class I HDACs (HDACs 1, 2, 3, and 8) and Class IIb HDAC (HDAC6) promote pathological hypertrophy, whereas Class IIa HDACs (HDACs 4, 5, 7, and 9) suppress cardiac hypertrophy [35, 77]. Experiments with the 3-Hit mouse model show that they had severe cardiac hypertrophy, diastolic dysfunction, and a profound inflammatory and fibrotic response (Figs. 3, 4, and 5, Supplemental Fig. S4). Significant changes in the expression ratio of different HDACs were observed in heart tissue of WT-N versus β2a-HFD-LN hearts. The main class I and class IIa HDACs expressed in all the hearts were HDACs 1, 2, 3, 8, 4, and 7 (Supplemental Fig. S5). Cardiac hypertrophy-inducing HDACs (HDAC1, 3, 6 and 8) were significantly increased in β2a-HFD-LN hearts (Fig. 4). The role of specific HDAC isoforms will require further study and could lead to novel approaches to abrogate the HFpEF phenotype in the 3-Hit model.
A role for HDAC activation in the HFpEF phenotype of β2a-HFD-LN was further documented in studies with the pan-HDAC inhibitor SAHA, an agent capable of promoting regression of established hypertrophy [32, 33]. Our experiments showed that SAHA prevented the development of a severe HFpEF phenotype in 3-Hit mice. SAHA also reduced the necrosis, inflammation, and fibrosis that were shown to be responsible for the development of the profound HFpEF phenotype seen in the 3-Hit model (Figs. 3, 4, 5, and 6; Supplemental Fig. S6).
## Increased Profibrotic M2-macrophage-TGFβ-Fibroblast Activation-Myocardial Fibrosis Pathway in 3-Hit Mouse Model Mirror the Clinical HFpEF Pathological Features
Obesity, hypertension, and cardiac hypertrophy have long been considered to create a profibrotic environment, stimulating the interstitial cardiac fibrosis that contributes to passive muscle stiffening and reduced chamber compliance in HFpEF [36, 37]. One of the known signaling mechanisms that can contribute to cardiac fibrosis is TGFβ. Macrophages are a potent producer of TGFβ. Macrophages have two main phenotypes: the M1 (classically activated) and M2 (alternatively activated and profibrotic)-macrophages. Westermann et al. [ 78] reported an increased number of TGFβ-expressing leukocytes, features characteristic of M2-macrophages, in HFpEF cardiac biopsies. Glezeva et al. [ 79] reported similar results that increased peripheral inflammation, monocytosis, and monocyte differentiation to anti-inflammatory/profibrotic M2-macrophages were present in HFpEF. The role of M2-macrophage has not been well studied in HFpEF animal models. In one recently published study, uninephrectomy and d-aldosterone infusion were performed to create a HFpEF mouse model. This study reported decreased mRNA expression of M2 markers (Arg1, CD163, and CD206) in HFpEF hearts, which was opposite from clinical patients with HFpEF [80]. Our results show that the number of M2-macrophages was significantly increased in the 3-Hit mouse with a significant HFpEF phenotype (Fig. 5). More work is needed to more clearly define the role of TGFβ signaling and profibrotic inflammatory processes in HFpEF.
TGFβ, secreted by macrophages, binds to the TGF receptor and actives a cascade of intracellular signals through the phosphorylation of Smad2 and -3 [41]. The pSmad$\frac{2}{3}$ translocate into the nucleus and binds to transcription factors (Smad binding element, SBE) on DNA, and then regulates the downstream gene expression, including α-SMA, collagen fiber, FN1, etc. [ 81]. The activation of TGFβ dependent Smad$\frac{2}{3}$ pathway in cardiac fibroblast cells is thought to contribute to the development of fibrosis, where fibroblasts transform into myofibroblast cells [40]. The crosslinking of extracellular matrix proteins regulated by lysyl oxidase (LOX) also potently affects their mechanical properties [82]. In our study, the 3-Hit mouse model developed severe CM hypertrophy with some cell necrosis. These changes were associated with increased cardiac profibrotic M2-macrophage population, TGFβ secretion, phosphorylation of Smad2 and -3, fibroblast activation, expression of FN1 and LOX, and myocardial fibrosis (Figs. 5 and 6). These pathological changes were linked to increased myocardial stiffness and LV filling pressures that underlie diastolic dysfunction in the 3-Hit model.
HDACs inhibitors have been shown to reduce collagen production and decrease markers of cardiac fibrosis in the diseased heart [83, 84]. The present study showed the importance of these HFpEF mechanisms in SAHA treatment experiments. These experiments showed that SAHA treatment of 3-Hit mice can prevent the development of severe cardiac hypertrophy, increased M2-macrophage population, TGFβ secretion, fibroblast activation, and myocardial fibrosis (Figs. 4, 5, and 6). Collectively, these data suggest that the 3-Hit model could be used to further define HFpEF mechanisms and for testing putative new therapies.
## Limitations
HFpEF is a complex clinical syndrome and is increasingly being recognized as a multiorgan, systemic syndrome. HFpEF animal models with one inducing stressor are, therefore, likely to have limitations that make cellular and molecular mechanisms that characterize this syndrome less well understood. Our goal was to develop a preclinical HFpEF animal model with multiple stressors known to be linked to HFpEF that together induce a profound HFpEF phenotype.
Our 3-Hit HFpEF mouse model with cardiomyocyte-specific Ca2+ stress plus HFD and l-NAME treatment mirrored the HFpEF clinical phenotype. A limitation of the current study is the small size of some of the treatment groups and the fact that sex-based effects in the model still need to be determined [85]. Another limitation of the study is the relatively long treatment period, which made the model building more challenging. The mice were 8 mo old when studies were terminated, so aging could be a factor contributing to the results. In addition, our study did not include the exercise exhaustion test, which can detect the exercise intolerance commonly present in patients with HFpEF. An advantage of the current study is that the β2a-N mice had a lower β2a expression and a less pronounced phenotype [22] that allowed us to explore the added stress of HFD and l-NAME.
## Conclusions
In summary, the 3-Hit mouse model produced a profound HFpEF phenotype. The primary mechanisms inducing this phenotype were HDACs dependent-CM hypertrophy, necrosis, increased profibrotic M2-macrophage populations, fibroblast activation, and myocardial fibrosis. A role for HDAC activation in the HFpEF phenotype was shown in studies with SAHA treatment, which prevented the severe HFpEF phenotype. These results suggest that this 3-Hit mouse model can be used for identifying and testing novel therapeutic strategies to treat HFpEF.
## DATA AVAILABILITY
Data will be made available upon reasonable request.
## GRANTS
This study was funded by National Heart, Lung, and Blood Institute Grants HL140071 (to S. R. Houser) and HL147558 (to S. R. Houser and T. A. McKinsey).
## DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the authors.
## AUTHOR CONTRIBUTIONS
Y.L., H.K., J.W.E., X.C., and S.R.H. conceived and designed research; Y.L., J.P.J., D.M.E., R.M.B., and M.F. performed experiments; Y.L. and D.Y. analyzed data; Y.L. interpreted results of experiments; Y.L. prepared figures; Y.L. drafted manuscript; Y.L., D.Y., Y.Y., J.P.J., M.F., T.A.M., J.Y., J.W.E., X.C., and S.R.H., edited and revised manuscript; Y.L., H.K., D.Y., Y.Y., J.P.J., D.M.E., M.F., T.A.M., J.Y., J.W.E., X.C., and S.R.H. approved final version of manuscript.
## References
1. Gevaert AB, Kataria R, Zannad F, Sauer AJ, Damman K, Sharma K, Shah SJ, Van Spall HGC. **Heart failure with preserved ejection fraction: recent concepts in diagnosis, mechanisms and management**. *Heart* (2022) **108** 1342-1350. DOI: 10.1136/heartjnl-2021-319605
2. Roger VL. **Epidemiology of heart failure**. *Circ Res* (2021) **128** 1421-1434. DOI: 10.1161/CIRCRESAHA.121.318172
3. Clark KAA, Velazquez EJ. **Heart failure with preserved ejection fraction: time for a reset**. *JAMA* (2020) **324** 1506-1508. DOI: 10.1001/jama.2020.15566
4. Shah SJ, Borlaug BA, Kitzman DW, McCulloch AD, Blaxall BC, Agarwal R, Chirinos JA, Collins S, Deo RC, Gladwin MT, Granzier H, Hummel SL, Kass DA, Redfield MM, Sam F, Wang TJ, Desvigne-Nickens P, Adhikari BB. **Research priorities for heart failure with preserved ejection fraction**. *Circulation* (2020) **141** 1001-1026. DOI: 10.1161/CIRCULATIONAHA.119.041886
5. Oktay AA, Rich JD, Shah SJ. **The emerging epidemic of heart failure with preserved ejection fraction**. *Curr Heart Fail Rep* (2013) **10** 401-410. DOI: 10.1007/s11897-013-0155-7
6. Virani SS, Alonso A, Benjamin EJ, Bittencourt MS, Callaway CW, Carson AP. **Heart disease and stroke statistics—2020 update: a report from the American Heart Association**. *Circulation* (2020) **141** e139-e596. DOI: 10.1161/CIR.0000000000000757
7. Singh A, Mehta Y. **Heart failure with preserved ejection fraction (HFpEF): implications for the anesthesiologists**. *J Anaesthesiol Clin Pharmacol* (2018) **34** 161-165. DOI: 10.4103/joacp.JOACP_352_16
8. Ovchinnikov AG, Arefieva TI, Potekhina AV, Filatova AY, Ageev FT, Boytsov SA. **The molecular and cellular mechanisms associated with a microvascular inflammation in the pathogenesis of heart failure with preserved ejection fraction**. *Acta Naturae* (2020) **12** 40-51. DOI: 10.32607/actanaturae.10990
9. Mishra S, Kass DA. **Cellular and molecular pathobiology of heart failure with preserved ejection fraction**. *Nat Rev Cardiol* (2021) **18** 400-423. DOI: 10.1038/s41569-020-00480-6
10. Eisner DA, Caldwell JL, Trafford AW, Hutchings DC. **The control of diastolic calcium in the heart: basic mechanisms and functional implications**. *Circ Res* (2020) **126** 395-412. DOI: 10.1161/CIRCRESAHA.119.315891
11. Qu P, Hamada M, Ikeda S, Hiasa G, Shigematsu Y, Hiwada K. **Time-course changes in left ventricular geometry and function during the development of hypertension in Dahl salt-sensitive rats**. *Hypertens Res* (2000) **23** 613-623. DOI: 10.1291/hypres.23.613
12. Chen-Izu Y, Chen L, Bányász T, McCulle SL, Norton B, Scharf SM, Agarwal A, Patwardhan A, Izu LT, Balke CW. **Hypertension-induced remodeling of cardiac excitation-contraction coupling in ventricular myocytes occurs prior to hypertrophy development**. *Am J Physiol Heart Circ Physiol* (2007) **293** H3301-H3310. DOI: 10.1152/ajpheart.00259.2007
13. Valero-Muñoz M, Li S, Wilson RM, Hulsmans M, Aprahamian T, Fuster JJ, Nahrendorf M, Scherer PE, Sam F. **Heart failure with preserved ejection fraction induces beiging in adipose tissue**. *Circ Heart Fail* (2016) **9**. DOI: 10.1161/CIRCHEARTFAILURE.115.002724
14. Karuppagounder V, Arumugam S, Babu SS, Palaniyandi SS, Watanabe K, Cooke JP, Thandavarayan RA. **The senescence accelerated mouse prone 8 (SAMP8): a novel murine model for cardiac aging**. *Ageing Res Rev* (2017) **35** 291-296. DOI: 10.1016/j.arr.2016.10.006
15. Wallner M, Eaton DM, Berretta RM, Borghetti G, Wu J, Baker ST, Feldsott EA, Sharp TE, Mohsin S, Oyama MA, von Lewinski D, Post H, Wolfson MR, Houser SR. **A feline HFpEF model with pulmonary hypertension and compromised pulmonary function**. *Sci Rep* (2017) **7**. DOI: 10.1038/s41598-017-15851-2
16. Sorop O, Heinonen I, van Kranenburg M, van de Wouw J, de Beer VJ, Nguyen ITN, Octavia Y, van Duin RWB, Stam K, van Geuns RJ, Wielopolski PA, Krestin GP, van den Meiracker AH, Verjans R, van Bilsen M, Danser AHJ, Paulus WJ, Cheng C, Linke WA, Joles JA, Verhaar MC, van der Velden J, Merkus D, Duncker DJ. **Multiple common comorbidities produce left ventricular diastolic dysfunction associated with coronary microvascular dysfunction, oxidative stress, and myocardial stiffening**. *Cardiovasc Res* (2018) **114** 954-964. DOI: 10.1093/cvr/cvy038
17. Schiattarella GG, Altamirano F, Tong D, French KM, Villalobos E, Kim SY, Luo X, Jiang N, May HI, Wang ZV, Hill TM, Mammen PPA, Huang J, Lee DI, Hahn VS, Sharma K, Kass DA, Lavandero S, Gillette TG, Hill JA. **Nitrosative stress drives heart failure with preserved ejection fraction**. *Nature* (2019) **568** 351-356. DOI: 10.1038/s41586-019-1100-z
18. Withaar C, Lam CSP, Schiattarella GG, de Boer RA, Meems LMG. **Heart failure with preserved ejection fraction in humans and mice: embracing clinical complexity in mouse models**. *Eur Heart J* (2021) **42** 4420-4430. DOI: 10.1093/eurheartj/ehab389
19. Faxen UL, Venkateshvaran A, Shah SJ, Lam CSP, Svedlund S, Saraste A, Beussink-Nelson L, Lagerstrom Fermer M, Gan L-M, Hage C, Lund LH. **Generalizability of HFA-PEFF and H2FPEF diagnostic algorithms and associations with heart failure indices and proteomic biomarkers: insights from PROMIS-HFpEF**. *J Card Fail* (2021) **27** 756-765. DOI: 10.1016/j.cardfail.2021.02.005
20. Anker SD, Butler J, Filippatos G, Ferreira JP, Bocchi E, Böhm M. **Empagliflozin in heart failure with a preserved ejection fraction**. *N Engl J Med* (2021) **385** 1451-1461. DOI: 10.1056/NEJMoa2107038
21. Chen X, Nakayama H, Zhang X, Ai X, Harris DM, Tang M, Zhang H, Szeto C, Stockbower K, Berretta RM, Eckhart AD, Koch WJ, Molkentin JD, Houser SR. **Calcium influx through Cav1.2 is a proximal signal for pathological cardiomyocyte hypertrophy**. *J Mol Cell Cardiol* (2011) **50** 460-470. DOI: 10.1016/j.yjmcc.2010.11.012
22. Nakayama H, Chen X, Baines CP, Klevitsky R, Zhang X, Zhang H, Jaleel N, Chua BH, Hewett TE, Robbins J, Houser SR, Molkentin JD. **Ca2+- and mitochondrial-dependent cardiomyocyte necrosis as a primary mediator of heart failure**. *J Clin Invest* (2007) **117** 2431-2444. DOI: 10.1172/JCI31060
23. Harper SC, Johnson J, Borghetti G, Zhao H, Wang T, Wallner M, Kubo H, Feldsott EA, Yang Y, Joo Y, Gou X, Sabri AK, Gupta P, Myzithras M, Khalil A, Franti M, Houser SR. **GDF11 decreases pressure overload-induced hypertrophy, but can cause severe cachexia and premature death**. *Circ Res* (2018) **123** 1220-1231. DOI: 10.1161/CIRCRESAHA.118.312955
24. Duran JM, Makarewich CA, Sharp TE, Starosta T, Zhu F, Hoffman NE, Chiba Y, Madesh M, Berretta RM, Kubo H, Houser SR. **Bone-derived stem cells repair the heart after myocardial infarction through transdifferentiation and paracrine signaling mechanisms**. *Circ Res* (2013) **113** 539-552. DOI: 10.1161/CIRCRESAHA.113.301202
25. Wallner M, Duran JM, Mohsin S, Troupes CD, Vanhoutte D, Borghetti G, Vagnozzi RJ, Gross P, Yu D, Trappanese DM, Kubo H, Toib A, Sharp TE, Harper SC, Volkert MA, Starosta T, Feldsott EA, Berretta RM, Wang T, Barbe MF, Molkentin JD, Houser SR. **Acute catecholamine exposure causes reversible myocyte injury without cardiac regeneration**. *Circ Res* (2016) **119** 865-879. DOI: 10.1161/CIRCRESAHA.116.308687
26. Yang Y, Kurian J, Schena G, Johnson J, Kubo H, Travers JG, Kang C, Lucchese AM, Eaton DM, Lv M, Li N, Leynes LG, Yu D, Yang F, McKinsey TA, Kishore R, Khan M, Mohsin S, Houser SR. **Cardiac remodeling during pregnancy with metabolic syndrome**. *Circulation* (2021) **143** 699-712. DOI: 10.1161/CIRCULATIONAHA.120.051264
27. Hobby ARH, Sharp TE, Berretta RM, Borghetti G, Feldsott E, Mohsin S, Houser SR. **Cortical bone-derived stem cell therapy reduces apoptosis after myocardial infarction**. *Am J Physiol Heart Circ Physiol* (2019) **317** H820-H829. DOI: 10.1152/ajpheart.00144.2019
28. Xu K, Chen S, Xie L, Qiu Y, Liu XZ, Bai X, Jin Y, Wang XH, Sun Y. **The protective effects of systemic dexamethasone on sensory epithelial damage and hearing loss in targeted Cx26-null mice**. *Cell Death Dis* (2022) **13**. DOI: 10.1038/s41419-022-04987-3
29. Lo M, Sharir A, Paul MD, Torosyan H, Agnew C, Li A, Neben C, Marangoni P, Xu L, Raleigh DR, Jura N, Klein OD. **CNPY4 inhibits the Hedgehog pathway by modulating membrane sterol lipids**. *Nat Commun* (2022) **13**. DOI: 10.1038/s41467-022-30186-x
30. Ye X, Zhang N, Jin Y, Xu B, Guo C, Wang X, Su Y, Yang Q, Song J, Yu W, Cheng P, Cheng L, Gong Y, Fu X, Sun H. **Dramatically changed immune-related molecules as early diagnostic biomarkers of non-small cell lung cancer**. *FEBS J* (2020) **287** 783-799. DOI: 10.1111/febs.15051
31. Yu H, Zhang P, Chen YR, Wang YJ, Lin XY, Li XY, Chen G. **Temporal changes of spinal transcriptomic profiles in mice with spinal nerve ligation**. *Front Neurosci* (2019) **13**. DOI: 10.3389/fnins.2019.01357
32. Grabarska A, Łuszczki JJ, Nowosadzka E, Gumbarewicz E, Jeleniewicz W, Dmoszyńska-Graniczka M, Kowalczuk K, Kupisz K, Polberg K, Stepulak A. **Histone deacetylase inhibitor SAHA as potential targeted therapy agent for larynx cancer cells**. *J Cancer* (2017) **8** 19-28. DOI: 10.7150/jca.16655
33. Wallner M, Eaton DM, Berretta RM, Liesinger L, Schittmayer M, Gindlhuber J, Wu J, Jeong MY, Lin YH, Borghetti G, Baker ST, Zhao H, Pfleger J, Blass S, Rainer PP, von Lewinski D, Bugger H, Mohsin S, Graier WF, Zirlik A, McKinsey TA, Birner-Gruenberger R, Wolfson MR, Houser SR. **HDAC inhibition improves cardiopulmonary function in a feline model of diastolic dysfunction**. *Sci Transl Med* (2020) **12**. DOI: 10.1126/scitranslmed.aay7205
34. Seto E, Yoshida M. **Erasers of histone acetylation: the histone deacetylase enzymes**. *Cold Spring Harb Perspect Biol* (2014) **6**. DOI: 10.1101/cshperspect.a018713
35. Kee HJ, Bae EH, Park S, Lee KE, Suh SH, Kim SW, Jeong MH. **HDAC inhibition suppresses cardiac hypertrophy and fibrosis in DOCA-salt hypertensive rats via regulation of HDAC6/HDAC8 enzyme activity**. *Kidney Blood Press Res* (2013) **37** 229-239. DOI: 10.1159/000350148
36. Yamamoto K, Masuyama T, Sakata Y, Nishikawa N, Mano T, Yoshida J, Miwa T, Sugawara M, Yamaguchi Y, Ookawara T, Suzuki K, Hori M. **Myocardial stiffness is determined by ventricular fibrosis, but not by compensatory or excessive hypertrophy in hypertensive heart**. *Cardiovasc Res* (2002) **55** 76-82. DOI: 10.1016/s0008-6363(02)00341-3
37. Alex L, Russo I, Holoborodko V, Frangogiannis NG. **Characterization of a mouse model of obesity-related fibrotic cardiomyopathy that recapitulates features of human heart failure with preserved ejection fraction**. *Am J Physiol Heart Circ Physiol* (2018) **315** H934-H949. DOI: 10.1152/ajpheart.00238.2018
38. Nacu N, Luzina IG, Highsmith K, Lockatell V, Pochetuhen K, Cooper ZA, Gillmeister MP, Todd NW, Atamas SP. **Macrophages produce TGF-β-induced (β-ig-h3) following ingestion of apoptotic cells and regulate MMP14 levels and collagen turnover in fibroblasts**. *J Immunol* (2008) **180** 5036-5044. DOI: 10.4049/jimmunol.180.7.5036
39. Shinde AV, Humeres C, Frangogiannis NG. **The role of α-smooth muscle actin in fibroblast-mediated matrix contraction and remodeling**. *Biochim Biophys Acta Mol Basis Dis* (2017) **1863** 298-309. DOI: 10.1016/j.bbadis.2016.11.006
40. Tarbit E, Singh I, Peart JN, Rose'Meyer RB. **Biomarkers for the identification of cardiac fibroblast and myofibroblast cells**. *Heart Fail Rev* (2019) **24** 1-15. DOI: 10.1007/s10741-018-9720-1
41. Zhu X, Kong X, Ma S, Liu R, Li X, Gao S, Ren D, Zheng Y, Tang J. **TGFβ/Smad mediated the polyhexamethyleneguanide areosol-induced irreversible pulmonary fibrosis in subchronic inhalation exposure**. *Inhal Toxicol* (2020) **32** 419-430. DOI: 10.1080/08958378.2020.1836091
42. Yabluchanskiy A, Ma Y, Iyer RP, Hall ME, Lindsey ML. **Matrix metalloproteinase-9: many shades of function in cardiovascular disease**. *Physiology (Bethesda)* (2013) **28** 391-403. DOI: 10.1152/physiol.00029.2013
43. González A, Ravassa S, López B, Moreno MU, Beaumont J, José GS, Querejeta R, Bayés-Genís A, Díez J. **Myocardial remodeling in hypertension**. *Hypertension* (2018) **72** 549-558. DOI: 10.1161/HYPERTENSIONAHA.118.11125
44. Peotta VA, Vasquez EC, Meyrelles SS. **Cardiovascular neural reflexes in L-NAME–induced hypertension in mice**. *Hypertension* (2001) **38** 555-559. DOI: 10.1161/01.hyp.38.3.555
45. de Moura e Dias M, dos Reis SA, da Conceição LL, Sediyama C, Pereira SS, de Oliveira LL, Gouveia Peluzio MC, Martinez JA, Milagro FI. **Diet-induced obesity in animal models: points to consider and influence on metabolic markers**. *Diabetol Metab Syndr* (2021) **13**. DOI: 10.1186/s13098-021-00647-2
46. Kopp W. **How western diet and lifestyle drive the pandemic of obesity and civilization diseases**. *Diabetes Metab Syndr Obes* (2019) **12** 2221-2236. DOI: 10.2147/DMSO.S216791
47. Tromp J, MacDonald MR, Tay WT, Teng T-HK, Hung C-L, Narasimhan C, Shimizu W, Ling LH, Ng TP, Yap J, McMurray JJV, Zile MR, Richards AM, Anand IS, Lam CSP. **Heart failure with preserved ejection fraction in the young**. *Circulation* (2018) **138** 2763-2773. DOI: 10.1161/CIRCULATIONAHA.118.034720
48. Teramoto K, Teng TK, Chandramouli C, Tromp J, Sakata Y, Lam CS. **Epidemiology and clinical features of heart failure with preserved ejection fraction**. *Card Fail Rev* (2022) **8**. DOI: 10.15420/cfr.2022.06
49. Barnabei MS, Palpant NJ, Metzger JM. **Influence of genetic background on ex vivo and in vivo cardiac function in several commonly used inbred mouse strains**. *Physiol Genomics* (2010) **42a** 103-113. DOI: 10.1152/physiolgenomics.00071.2010
50. Gaisler-Silva F, Junho CVC, Fredo-da-Costa I, Christoffolete MA, Carneiro-Ramos MS. **Diet-induced obesity differentially modulates cardiac inflammatory status in the C57 and FVB mouse strains**. *Curr Mol Med* (2022) **22** 365-373. DOI: 10.2174/1566524021666210603163613
51. Kim DH, Gutierrez-Aguilar R, Kim HJ, Woods SC, Seeley RJ. **Increased adipose tissue hypoxia and capacity for angiogenesis and inflammation in young diet-sensitive C57 mice compared with diet-resistant FVB mice**. *Int J Obes (Lond)* (2013) **37** 853-860. DOI: 10.1038/ijo.2012.141
52. Shim HK, Kim SG, Kim TS, Kim SK, Lee SJ. **Total thyroidectomy in the mouse: the feasibility study in the non-thyroidal tumor model expressing human sodium/iodide symporter gene**. *Nucl Med Mol Imaging* (2011) **45** 103-110. DOI: 10.1007/s13139-011-0076-x
53. Saheera S, Krishnamurthy P. **Cardiovascular changes associated with hypertensive heart disease and aging**. *Cell Transplant* (2020) **29**. DOI: 10.1177/0963689720920830
54. Zhang H, Chen X, Gao E, MacDonnell SM, Wang W, Kolpakov M, Nakayama H, Zhang X, Jaleel N, Harris DM, Li Y, Tang M, Berretta R, Leri A, Kajstura J, Sabri A, Koch WJ, Molkentin JD, Houser SR. **Increasing cardiac contractility after myocardial infarction exacerbates cardiac injury and pump dysfunction**. *Circ Res* (2010) **107** 800-809. DOI: 10.1161/CIRCRESAHA.110.219220
55. Eisner DA, Caldwell JL, Kistamás K, Trafford AW. **Calcium and excitation-contraction coupling in the heart**. *Circ Res* (2017) **121** 181-195. DOI: 10.1161/CIRCRESAHA.117.310230
56. Sutanto H, Lyon A, Lumens J, Schotten U, Dobrev D, Heijman J. **Cardiomyocyte calcium handling in health and disease: Insights from in vitro and in silico studies**. *Prog Biophys Mol Biol* (2020) **157** 54-75. DOI: 10.1016/j.pbiomolbio.2020.02.008
57. Chaanine AH. **Metabolic remodeling and implicated calcium and signal transduction pathways in the pathogenesis of heart failure**. *Int J Mol Sci* (2021) **22**. DOI: 10.3390/ijms221910579
58. Sridhar KC, Hersch N, Dreissen G, Merkel R, Hoffmann B. **Calcium mediated functional interplay between myocardial cells upon laser-induced single-cell injury: an in vitro study of cardiac cell death signaling mechanisms**. *Cell Commun Signal* (2020) **18**. DOI: 10.1186/s12964-020-00689-5
59. Durak A, Olgar Y, Genc K, Tuncay E, Akat F, Degirmenci S, Turan B. **STIM1-Orai1 interaction mediated calcium influx activation contributes to cardiac contractility of insulin-resistant rats**. *BMC Cardiovasc Disord* (2022) **22**. DOI: 10.1186/s12872-022-02586-w
60. Pieske B, Tschöpe C, de Boer RA, Fraser AG, Anker SD, Donal E, Edelmann F, Fu M, Guazzi M, Lam CSP, Lancellotti P, Melenovsky V, Morris DA, Nagel E, Pieske-Kraigher E, Ponikowski P, Solomon SD, Vasan RS, Rutten FH, Voors AA, Ruschitzka F, Paulus WJ, Seferovic P, Filippatos G. **How to diagnose heart failure with preserved ejection fraction: the HFA–PEFF diagnostic algorithm: a consensus recommendation from the Heart Failure Association (HFA) of the European Society of Cardiology (ESC)**. *Eur Heart J* (2019) **40** 3297-3317. DOI: 10.1093/eurheartj/ehz641
61. Barandiarán Aizpurua A, Sanders-van Wijk S, Brunner-La Rocca HP, Henkens M, Heymans S, Beussink-Nelson L, Shah SJ, van Empel VPM. **Validation of the HFA-PEFF score for the diagnosis of heart failure with preserved ejection fraction**. *Eur J Heart Fail* (2020) **22** 413-421. DOI: 10.1002/ejhf.1614
62. Kim MN, Park SM. **Heart failure with preserved ejection fraction: insights from recent clinical researches**. *Korean J Intern Med* (2020) **35** 514-534. DOI: 10.3904/kjim.2020.104
63. Egashira K, Sueta D, Komorita T, Yamamoto E, Usuku H, Tokitsu T, Fujisue K, Nishihara T, Oike F, Takae M, Hanatani S, Takashio S, Ito M, Yamanaga K, Araki S, Soejima H, Kaikita K, Matsushita K, Tsujita K. **HFA-PEFF scores: prognostic value in heart failure with preserved left ventricular ejection fraction**. *Korean J Intern Med* (2022) **37** 96-108. DOI: 10.3904/kjim.2021.272
64. DeVore AD, McNulty S, Alenezi F, Ersboll M, Vader JM, Oh JK, Lin G, Redfield MM, Lewis G, Semigran MJ, Anstrom KJ, Hernandez AF, Velazquez EJ. **Impaired left ventricular global longitudinal strain in patients with heart failure with preserved ejection fraction: insights from the RELAX trial**. *Eur J Heart Fail* (2017) **19** 893-900. DOI: 10.1002/ejhf.754
65. Kim HY, Park S-J, Lee S-C, Chang SY, Kim E-K, Chang S-A, Choi J-O, Park SW, Kim S-M, Choe YH, Oh JK. **Comparison of global and regional myocardial strains in patients with heart failure with a preserved ejection fraction vs hypertension vs age-matched control**. *Cardiovasc Ultrasound* (2020) **18**. DOI: 10.1186/s12947-020-00223-0
66. Lam CS, Roger VL, Rodeheffer RJ, Bursi F, Borlaug BA, Ommen SR, Kass DA, Redfield MM. **Cardiac structure and ventricular-vascular function in persons with heart failure and preserved ejection fraction from Olmsted County, Minnesota**. *Circulation* (2007) **115** 1982-1990. DOI: 10.1161/CIRCULATIONAHA.106.659763
67. Katz DH, Beussink L, Sauer AJ, Freed BH, Burke MA, Shah SJ. **Prevalence, clinical characteristics, and outcomes associated with eccentric versus concentric left ventricular hypertrophy in heart failure with preserved ejection fraction**. *Am J Cardiol* (2013) **112** 1158-1164. DOI: 10.1016/j.amjcard.2013.05.061
68. Gupta DK, Shah AM, Castagno D, Takeuchi M, Loehr LR, Fox ER, Butler KR, Mosley TH, Kitzman DW, Solomon SD. **Heart failure with preserved ejection fraction in African Americans: the ARIC (Atherosclerosis Risk In Communities) study**. *JACC Heart Fail* (2013) **1** 156-163. DOI: 10.1016/j.jchf.2013.01.003
69. Slivnick J, Lampert BC. **Hypertension and heart failure**. *Heart Fail Clin* (2019) **15** 531-541. DOI: 10.1016/j.hfc.2019.06.007
70. Budde H, Hassoun R, Mügge A, Kovács Á, Hamdani N. **Current understanding of molecular pathophysiology of heart failure with preserved ejection fraction**. *Front Physiol* (2022) **13**. DOI: 10.3389/fphys.2022.928232
71. Nakamura M, Sadoshima J. **Mechanisms of physiological and pathological cardiac hypertrophy**. *Nat Rev Cardiol* (2018) **15** 387-407. DOI: 10.1038/s41569-018-0007-y
72. Kook H, Lepore JJ, Gitler AD, Lu MM, Wing -M, Yung W, Mackay J, Zhou R, Ferrari V, Gruber P, Epstein JA. **Cardiac hypertrophy and histone deacetylase-dependent transcriptional repression mediated by the atypical homeodomain protein Hop**. *J Clin Invest* (2003) **112** 863-871. DOI: 10.1172/JCI19137
73. Ooi JY, Tuano NK, Rafehi H, Gao XM, Ziemann M, Du XJ, El-Osta A. **HDAC inhibition attenuates cardiac hypertrophy by acetylation and deacetylation of target genes**. *Epigenetics* (2015) **10** 418-430. DOI: 10.1080/15592294.2015.1024406
74. Cao DJ, Wang ZV, Battiprolu PK, Jiang N, Morales CR, Kong Y, Rothermel BA, Gillette TG, Hill JA. **Histone deacetylase (HDAC) inhibitors attenuate cardiac hypertrophy by suppressing autophagy**. *Proc Natl Acad Sci USA* (2011) **108** 4123-4128. DOI: 10.1073/pnas.1015081108
75. Gillette TG. **HDAC inhibition in the heart**. *Circulation* (2021) **143** 1891-1893. DOI: 10.1161/CIRCULATIONAHA.121.054262
76. Travers JG, Wennersten SA, Peña B, Bagchi RA, Smith HE, Hirsch RA, Vanderlinden LA, Lin Y-H, Dobrinskikh E, Demos-Davies KM, Cavasin MA, Mestroni L, Steinkühler C, Lin CY, Houser SR, Woulfe KC, Lam MPY, McKinsey TA. **HDAC inhibition reverses preexisting diastolic dysfunction and blocks covert extracellular matrix remodeling**. *Circulation* (2021) **143** 1874-1890. DOI: 10.1161/CIRCULATIONAHA.120.046462
77. Lemon DD, Horn TR, Cavasin MA, Jeong MY, Haubold KW, Long CS, Irwin DC, McCune SA, Chung E, Leinwand LA, McKinsey TA. **Cardiac HDAC6 catalytic activity is induced in response to chronic hypertension**. *J Mol Cell Cardiol* (2011) **51** 41-50. DOI: 10.1016/j.yjmcc.2011.04.005
78. Westermann D, Lindner D, Kasner M, Zietsch C, Savvatis K, Escher F, von Schlippenbach J, Skurk C, Steendijk P, Riad A, Poller W, Schultheiss HP, Tschöpe C. **Cardiac inflammation contributes to changes in the extracellular matrix in patients with heart failure and normal ejection fraction**. *Circ Heart Fail* (2011) **4** 44-52. DOI: 10.1161/CIRCHEARTFAILURE.109.931451
79. Glezeva N, Voon V, Watson C, Horgan S, McDonald K, Ledwidge M, Baugh J. **Exaggerated inflammation and monocytosis associate with diastolic dysfunction in heart failure with preserved ejection fraction: evidence of M2 macrophage activation in disease pathogenesis**. *J Card Fail* (2015) **21** 167-177. DOI: 10.1016/j.cardfail.2014.11.004
80. Zhang L, Chen J, Yan L, He Q, Xie H, Chen M. **Resveratrol ameliorates cardiac remodeling in a murine model of heart failure with preserved ejection fraction**. *Front Pharmacol* (2022) **12**. DOI: 10.3389/fphar.2021.646240
81. Li P, Wang D, Lucas J, Oparil S, Xing D, Cao X, Novak L, Renfrow MB, Chen YF. **Atrial natriuretic peptide inhibits transforming growth factor beta-induced Smad signaling and myofibroblast transformation in mouse cardiac fibroblasts**. *Circ Res* (2008) **102** 185-192. DOI: 10.1161/CIRCRESAHA.107.157677
82. González-Santamaría J, Villalba M, Busnadiego O, López-Olañeta MM, Sandoval P, Snabel J, López-Cabrera M, Erler JT, Hanemaaijer R, Lara-Pezzi E, Rodríguez-Pascual F. **Matrix cross-linking lysyl oxidases are induced in response to myocardial infarction and promote cardiac dysfunction**. *Cardiovasc Res* (2016) **109** 67-78. DOI: 10.1093/cvr/cvv214
83. Travers JG, Tharp CA, Rubino M, McKinsey TA. **Therapeutic targets for cardiac fibrosis: from old school to next-gen**. *J Clin Invest* (2022) **132**. DOI: 10.1172/JCI148554
84. Wang K, Tang R, Wang S, Wang W, Zhang K, Li J, Li P, Tang YD. **SAHA could inhibit TGF-β1/p38 pathway in MI-induced cardiac fibrosis through DUSP4 overexpression**. *Heart Vessels* (2022) **37** 152-160. DOI: 10.1007/s00380-021-01900-4
85. Sotomi Y, Hikoso S, Nakatani D, Mizuno H, Okada K, Dohi T, Kitamura T, Sunaga A, Kida H, Oeun B, Sato T, Komukai S, Tamaki S, Yano M, Hayashi T, Nakagawa A, Nakagawa Y, Yasumura Y, Yamada T, Sakata Y. **Sex differences in heart failure with preserved ejection fraction**. *J Am Heart Assoc* (2021) **10**. DOI: 10.1161/JAHA.120.018574
|
---
title: Endothelial cell-specific roles for tetrahydrobiopterin in myocardial function,
cardiac hypertrophy, and response to myocardial ischemia-reperfusion injury
authors:
- Surawee Chuaiphichai
- Sandy M. Chu
- Ricardo Carnicer
- Matthew Kelly
- Jenifer K. Bendall
- Jillian N. Simon
- Gillian Douglas
- Mark J. Crabtree
- Barbara Casadei
- Keith M. Channon
journal: American Journal of Physiology - Heart and Circulatory Physiology
year: 2023
pmcid: PMC9988535
doi: 10.1152/ajpheart.00562.2022
license: CC BY 4.0
---
# Endothelial cell-specific roles for tetrahydrobiopterin in myocardial function, cardiac hypertrophy, and response to myocardial ischemia-reperfusion injury
## Abstract
The cofactor tetrahydrobiopterin (BH4) is a critical regulator of nitric oxide synthase (NOS) function and redox signaling, with reduced BH4 implicated in multiple cardiovascular disease states. In the myocardium, augmentation of BH4 levels can impact on cardiomyocyte function, preventing hypertrophy and heart failure. However, the specific role of endothelial cell BH4 biosynthesis in the coronary circulation and its role in cardiac function and the response to ischemia has yet to be elucidated. Endothelial cell-specific Gch1 knockout mice were generated by crossing Gch1fl/fl with Tie2cre mice, generating Gch1fl/flTie2cre mice and littermate controls. GTP cyclohydrolase protein and BH4 levels were reduced in heart tissues from Gch1fl/flTie2cre mice, localized to endothelial cells, with normal cardiomyocyte BH4. Deficiency in coronary endothelial cell BH4 led to NOS uncoupling, decreased NO bioactivity, and increased superoxide and hydrogen peroxide productions in the hearts of Gch1fl/flTie2cre mice. Under physiological conditions, loss of endothelial cell-specific BH4 led to mild cardiac hypertrophy in Gch1fl/flTie2cre hearts. Endothelial cell BH4 loss was also associated with increased neuronal NOS protein, loss of endothelial NOS protein, and increased phospholamban phosphorylation at serine-17 in cardiomyocytes. Loss of cardiac endothelial cell BH4 led to coronary vascular dysfunction, reduced functional recovery, and increased myocardial infarct size following ischemia-reperfusion injury. Taken together, these studies reveal a specific role for endothelial cell Gch1/BH4 biosynthesis in cardiac function and the response to cardiac ischemia-reperfusion injury. Targeting endothelial cell Gch1 and BH4 biosynthesis may provide a novel therapeutic target for the prevention and treatment of cardiac dysfunction and ischemia-reperfusion injury.
NEW & NOTEWORTHY We demonstrate a critical role for endothelial cell Gch1/BH4 biosynthesis in coronary vascular function and cardiac function. Loss of cardiac endothelial cell BH4 leads to coronary vascular dysfunction, reduced functional recovery, and increased myocardial infarct size following ischemia/reperfusion injury. Targeting endothelial cell Gch1 and BH4 biosynthesis may provide a novel therapeutic target for the prevention and treatment of cardiac dysfunction, ischemia injury, and heart failure.
## INTRODUCTION
Cardiovascular diseases including coronary artery disease, myocardial infarction, and heart failure are leading causes of global mortality and disability [1]. A hallmark of cardiovascular diseases is an early reduction in nitric oxide (NO) bioavailability and an increase in reactive oxygen species (ROS) production. Tetrahydrobiopterin (BH4) is a critical regulator of nitric oxide synthases (NOS) function and NOS-derived NO and ROS signaling in cardiovascular physiology [2, 3]. Biosynthesis of BH4 is catalyzed by GTPCH (GTP cyclohydrolase 1, encoded by Gch1). We have previously shown that Gch1 expression is a key determinant of BH4 bioavailability, NOS regulation, and NO generation [4, 5]. When BH4 bioavailability is limited, NOS is unable to generate NO from l-arginine and becomes “uncoupled,” resulting in generation of superoxide anion and other ROS, rather than NO, contributing to disturbed redox signaling (3, 6–8).
Clinically, genetic variants in GCH1 functionally associated with altered GCH1 expression appear to be associated with alterations in markers of cardiac function and cardiovascular risk [9, 10]. *Rare* genetic variants causing loss of BH4 synthesis are also associated with alterations of NOS-mediated vascular function and cardiovascular physiology [9, 10]. These observations from clinical studies have been supported by preclinical models. Reduced BH4 bioavailability and NOS uncoupling are associated with various heart diseases including cardiac hypertrophy [11] and ischemia-reperfusion injury (12–17). Oral supplementation with BH4 or the BH4 precursor sepiapterin has been shown to prevent or reduce cardiac hypertrophy and failure (11, 18–21). For example, pressure overload induced by transverse aortic constriction (TAC) in mice reduced cardiac BH4 levels and promoted eNOS uncoupling, leading to myocyte hypertrophy, cardiac dilation, interstitial fibrosis, and ventricular dysfunction [22]. Treatment with oral BH4 prevented the NOS uncoupling and reduced the TAC-induced hypertrophy. Furthermore, exogenous BH4 was able to recouple eNOS and reverse preexisting cardiac hypertrophy and fibrosis caused by TAC-induced pressure overload [11].
However, these and other studies do not address the important question of which cell types in the heart may mediate the effects of BH4 on cardiac function. NO is generated from coronary endothelial cells by eNOS, whereas in cardiac myocytes, both eNOS and nNOS have been shown to contribute to myocardial function. In the endothelium, NO mediates coronary vascular function and flow, whereas in cardiac myocytes, NO regulates LV relaxation by effects on myofilament calcium sensitivity and calcium handling [23]. Although prior studies have suggested a critical role of endothelial NO and BH4 levels on cardiac function and injury (17, 21, 24–26), no prior study has specifically addressed the requirement for endothelial cell BH4 in the regulation of cardiac function, particularly in myocardial ischemia-reperfusion (I/R) injury where roles for both eNOS and nNOS are implicated [27]. Interpreting the specific roles of NO in the heart using knockouts of either eNOS or nNOS is limited by loss of all NOS-related functions, including ROS generation, subcellular localization, and protein-protein interactions, whereas the requirement for BH4 in the generation of NO by both eNOS and nNOS enables selective targeting of NOS-mediated NO generation, without primary alterations in either NOS protein levels or other NOS functions.
Accordingly, we sought to investigate how selective targeting of endothelial cell BH4 biosynthesis, without alteration of cardiac myocyte BH4, would alter cardiac function, focusing on ischemia-reperfusion injury, where changes in coronary endothelial function play an important pathophysiological role.
## Animals
All animal studies were conducted under project licenses PPL $\frac{30}{3080}$ and P0C27F69A with ethical approval from the Local Ethical Review Committee and in accordance with the United Kingdom Home Office regulations (Guidance on the Operation of Animals, Scientific Procedures Act), 1986, with procedures reviewed by the clinical medicine Animal Care and Ethical Review Body (AWERB). Animals were housed in individually ventilated cages (between 4 and 6 mice per cage of mixed genotypes) in specific pathogen-free conditions. All animals were provided with standard chow (Teklad global $16\%$ protein diet, Harlan Laboratories) and water ad libitum and maintained on a 12-h:12-h light/dark cycle at controlled temperature (20–22°C) and humidity.
## Gch1 Knockout Mice
*We* generated mice with a Gch1 conditional knockout (floxed) allele, as previously described (28–30). Gch1fl/fl animals were bred with Tie2cre transgenic mice to produce Gch1fl/flTie2cre mice where Gch1 is deleted in endothelial cells, generating a mouse model of endothelial cell-specific BH4 deficiency. Since the Tie2cre transgene is active in the female germline, only male animals are used to establish breeding pairs to maintain conditional endothelial cell expression. Mice were genotyped according to the published protocol [29, 31]. Briefly, mice were genotyped by polymerase chain reactions using DNA prepared from ear biopsies. For Gch1fl/fl genotyping, PCR was performed using the following primers: Gch1fl/fl, forward 5′-GTC CTT GGT CTC AGT AAA CTT GCC AGG-3′; Gch1fl/fl, reverse 5′-GCC CAG CCA AGG ATA GAT GCA G-3′. The Gch1 floxed allele showed as 1,030 bp. For Tie2cre genotyping, PCR was performed using the following primers: Tie2cre, forward 5′-GCA TAA CCA GTG AAA CAG CAT TGC TG-3′; Tie2cre, reverse 5′-GGA CAT GTT CAG GGA TCG CCA GGC G-3′. The Tie2cre allele amplified as 280-bp fragment. Adult male Gch1fl/flTie2cre mice and their Gch1fl/fl littermates (hereafter referred to as wild type) on a pure (>10 generations) C57BL6/J background were bred in house and were used for all experiments at 20 to 24 wk.
## Determination of Tissue Tetrahydrobiopterin Levels
BH4 and oxidized biopterins (BH2 and biopterin) were determined by high-performance liquid chromatography (HPLC) followed by electrochemical and fluorescence detection, respectively, following an established protocol [32]. Briefly, frozen heart samples were homogenized in ice-cold resuspension buffer, consisting of (in mmol · L−1) 50 phosphate-buffered saline, 1 dithioerythriol, and 1 EDTA at pH 7.4. After centrifugation at 13,200 rpm for 10 min at 4°C, the supernatant was removed, and ice-cold acid precipitation buffer, consisting of (in mmol · L−1) 1 phosphoric acid, 2 trichloroacetic acid, and 1 dithioerythritol, was added. Samples were vigorously mixed and then centrifuged for 15 min at 13,000 rpm and 4°C. Samples were injected into an isocratic HPLC system and quantified using sequential electrochemical (Coulochem III, ESA, Inc.) and fluorescence (Jasco) detection. HPLC separation was performed using a 250-mm ACE C-18 column (Hichrom) and a mobile phase comprised of 50 mM sodium acetate, 5 mM citric acid, 48 µΜ EDTA, and 160 µΜ dithioerythritol (pH 5.2) (all ultrapure electrochemical HPLC grade) at a flow rate of 1.3 mL/min. Background currents of +500 μA and −50 μA were used for the detection of BH4 on electrochemical cells E1 and E2, respectively. 7,8-BH2 and biopterin were measured using a Jasco FP2020 fluorescence detector set at 510-nm excitation and 595-nm emission. Quantification of BH4, BH2, and B was done by comparison with authentic external standards and normalized to sample protein content.
## Endothelial Cell Isolation
Primary heart endothelial cells were isolated using MACS beads (Miltenyi Biotec), as previously described [29]. Briefly, mice were euthanized by an overdose of inhaled isoflurane. Hearts were harvested and digested in DMEM containing 0.18 U/mL Liberase (Roche) and 0.1 mg/mL DnaseI (Roche) for 1 h at 37°C. The digested tissue was filtered through 100- and 70-μm cell strainers. The cell suspension was then incubated with rat anti-CD31 antibody (BD PharMingen) for 15 min at 4°C and then with anti-rat secondary antibody coated immune magnetic beads for a further 15 min at 4°C. Bead-bound endothelial cells were selected using a magnetic column. Endothelial cells were collected and stored at −80°C for further analysis.
## Cardiomyocyte Isolation
Cardiac myocytes were isolated using an enzymatic dispersion technique [23]. Briefly, the heart was perfused with Ca2+-free isolation solution (37°C, oxygenated) for 3 min and then with 1 mg/mL collagenase type II solution (Worthington Biochemical) for a further 9 min. The myocytes were pelleted by centrifugation at a low speed (600 rpm). The supernatant was then spun down at 10,000 rpm; the pellet was considered as the nonmyocyte fraction.
## Echocardiography
Left ventricular (LV) size and function were investigated in vivo using a high-resolution two-dimensional (2D) echocardiography system (Vevo 2100, VisualSonics, Canada) in isoflurane ($1\%$–$1.5\%$) anesthetized mice. LV wall thickness and chamber dimensions were determined in the parasternal short-axis view (M-mode), from which measures of LVEF and fractional shortening were derived. 2-D images of the heart were obtained from the four-chamber apical view to assess mitral blood inflow and tissue-Doppler velocities.
## Quantification of Superoxide Production by Dihydroethidine (DHE)-HPLC
Superoxide production was quantified by measuring the production of 2-hydroxyethidium from dihydroethidium, using HPLC [29]. Briefly, frozen heart homogenate was preincubated with serum-free DMEM with or without 100 μM NG-nitro-l-arginine methyl ester (l-NAME; Sigma). Samples were then incubated with 25 μM DHE (Invitrogen) for 20 min before being harvested for separation of 2-hydroxyethidium using a gradient HPLC system (Jasco, UK) with an ODS3 reverse phase column (250 mm, 4.5 mm, Hichrom UK) and quantified using a fluorescence detector set at 510 nm (excitation) and 595 nm (emission).
## Langendorff Heart Preparation
Mice were heparinized (300 U) and anesthetized with ketamine (75 mg/kg) plus medetomidine hydrochloride (1 mg/kg), with the adequacy of anesthesia confirmed by the absence of a pedal reflex. Hearts were quickly excised and immersed in KH buffer. The aorta was then cannulated onto the Langendorff perfusion system for retrograde perfusion. The heart was perfused with 37°C KH buffer and gassed with $95\%$ O2-$5\%$ CO2, at 2 mL/min, and cardiac function was assessed using a fluid-filled balloon inserted into the left ventricle, which connected to a pressure transducer and a PowerLab system (ADInstruments). Left ventricular developed pressure (LVDP), calculated by the difference between systolic and diastolic pressure, was recorded continuously via LabChart software v.7.0. After 25-min equilibration, hearts were subjected to 35 min global ischemia followed by 60 min of reperfusion. For triphenyltetrazolium chloride (TTC) staining, hearts were removed from the Langendorff following ex vivo I/R, briefly frozen, and then sliced into six 1-mm-thick transverse sections. To distinguish viable (stained) versus necrotic (pale, unstained) tissue, sections were incubated in $1\%$ TTC for 30 min at 37°C. Sections were then scanned, and the area of infarction (TTC negative) was quantified as a percentage of the area at risk (entire area of the section) using ImageJ.
## Quantification of Gene Expression by Real-Time RT-PCR
RNA was prepared using the RNeasy kit (Qiagen) and was reverse transcribed using Superscript II (Life Technologies) according to standard protocols. RNA equivalent cDNA (5 ng) was used to perform real-time PCR using predesigned tag-man gene expression assays (Life Technologies) using a BioRad CFX1000. Gene expression levels of mouse Gch1, Nos1, Nos2, and Nos3 were normalized to the housekeeping gene GAPDH using the ΔCt method.
## Western Blot Analysis
Immunoblotting in LV homogenates was performed to evaluate protein levels of GTPCH (1:10,000 dilution; a gift from S.Gross, Cornell University; New York), iNOS (1: 1,100 dilution; Abcam), nNOS (1:1,000 dilution; Santa Cruz Biotechnology), eNOS (1:5,000 dilution; BD Bioscience), CD102 (1:1,000; R&D systems), SERCA2A (1:5,000 dilution; Santa Cruz Biotechnology), total phospholamban (1:2,000 dilution; PLB, Badrilla), phosphor-Thr17-PLB (1:2,000 dilution; Badrilla), phosphor-Ser16-PLB (1:2,000 dilution; Badrilla), NCX1 (1:1,000 dilution, Santa Cruz), phospho-extracellular signal-regulated protein kinases (1:500 dilution; ERK$\frac{1}{2}$), total ERK$\frac{1}{2}$ (1:500 dilution;), catalase (1:5,000 dilution; Calbiochem), MnSOD (1:5,000 dilution; Stressgen Bioreagents), EcSOD (1:750 dilution; Stressgen Bioreagents), Cu/ZnSOD (1:500 dilution; Stressgen Bioreagents), and β-tubulin (1:20,000; Abcam), followed by appropriate HRP-conjugated secondary antibody (1:10,000–20,000 dilution; Promega). Protein bands were visualized by enhanced chemiluminescence (Super West Pico Chemiluminescence, Thermo Scientific).
## Blood Pressure Measurement by Tail-Cuff Plethysmography
Systolic blood pressure in conscious wild-type and Gch1fl/flTie2cre mice was determined using the VisitechR computerized tail-cuff plethysmography system (Visitech) following 5 days of training and 3 days baseline periods. Experiments were performed between the hours of 8:00 and 12:00 am. The animal tails were passed through a cylindrical latex tail-cuff and taped down to reduce movement. Twenty readings were taken per mouse of which the first five readings were discarded. The remaining 15 readings were used to calculate the mean systolic blood pressure in each mouse.
## Statistical Analysis
All data are reported as means ± SE. The experimental unit (n) was defined as a single animal, animals of both genotypes were caged together, and animals of both genotypes were derived from more than one cage in all experiments. Statistical analyses were performed using GraphPad Prism v. 9.3.0. ( San Diego, CA). Normality was tested using D’Agostino and *Pearson omnibus* normality test. Groups were compared using the Mann–Whitney U test for nonparametric data or an unpaired Student’s t test for parametric data. When comparing multiple groups, data were analyzed by analysis of variance (ANOVA) with Newman–Keuls posttest for parametric data or Kruskal–Wallis test with Dunn’s posttest for nonparametric data. When more than two independent variables were present, a two-way ANOVA with Tukey’s multiple comparisons test was used. When within-subject repeated measurements were present, a repeated-measures (RM) ANOVA was used. A value of $P \leq 0.05$ was considered statistically significant. Data were collected and analyzed with the operator blind of treatment allocation. Randomization was performed by cage.
## Endothelial Cell-Targeted Gch1 Deletion in the Heart Causes Selective Endothelial Cell BH4 Deficiency
*We* generated matched litters of Gch1fl/flTie2cre and Gch1fl/fl mice (hereafter referred to as wild type) by crossing male Gch1fl/flTie2cre and female Gch1fl/fl mice. Body weights between the groups were similar (36 ± 1.5 g in wild type and 36 ± 1.1 g in Gch1fl/flTie2cre; $$n = 6$$ to 8 animals per group). Genomic polymerase chain reaction demonstrated efficient excision of the floxed Gch1 allele in isolated endothelial cells from Gch1fl/flTie2cre hearts (Fig. 1A). Endothelial cell-specific Gch1 deletion resulted in a significant reduction in Gch1 expression (Fig. 1, B and C), GTPCH protein in whole heart tissue, and barely detectable levels in endothelial cells isolated from the hearts. However, the GTPCH protein in isolated cardiomyocytes was similar between the groups (Fig. 1, D–I). Accordingly, BH4 levels were significantly decreased in hearts and barely detected in isolated endothelial cells from Gch1fl/flTie2cre hearts (Fig. 1, J, K, and M). Despite marked BH4 deficiency, absolute BH2 levels in heart tissue were comparable between wild-type and Gch1fl/flTie2cre mice, such that the BH4/BH2 and biopterin ratio was significantly reduced in Gch1fl/flTie2cre hearts (Fig. 1L). However, the BH4/BH2 and biopterin ratio in isolated endothelial cells was comparable between wild-type and Gch1fl/flTie2cre mice (Fig. 1N). In contrast to the observations in endothelial cells, BH4 levels in isolated cardiomyocytes were similar between wild-type and Gch1fl/flTie2cre mice, indicating that the reduction in overall heart tissue BH4 levels in Gch1fl/flTie2cre mice is due to specific deletion of endothelial cell Gch1. However, BH2 levels were significantly increased in cardiomyocytes isolated from Gch1fl/flTie2cre mice, such that the BH4/BH2 and biopterin ratio was significantly decreased in cardiomyocytes from Gch1fl/flTie2cre mice (Fig. 1, O and P), suggesting that selective endothelial cell BH4 deficiency in the heart leads to secondary effects on BH4 oxidation and/or recycling in cardiomyocytes, independent of changes in de novo BH4 biosynthesis. Importantly, plasma BH4 levels were similar between the groups, indicating that endothelial cell BH4 biosynthesis by GTPCH1 is not a major contributor to circulating BH4 levels (Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.21732581.v1).
**Figure 1.:** *Myocardial endothelial cell targeted Gch1 deletion causes a tissue-specific decrease in Gch1 gene, GTPCH protein, and biopterin content. A: evaluation of Tie2cre-mediated excision of the loxP flanked DNA in heart tissues and primary heart endothelial cells derived from Gch1fl/flTie2cre and Gch1fl/fl
[wild-type mice (WT)]. The predicted 1,030-bp product was detected in WT mice. In the presence of Tie2cre transgene a 1,392-bp knockout allele was detected, with efficient excision in primary endothelial cells from hearts. B and C: quantitative real-time PCR was used to quantify Gch1 gene expression in hearts and primary endothelial cells from hearts (*P < 0.05; n = 4 per group). D–F: representative immunoblot of GTPCH proteins in hearts, isolated primary endothelial cells, and isolated cardiomyocytes from WT and Gch1fl/flTie2cre hearts, respectively, with quantitative data, measured as percent band density in G–I: CD102 and β-tubulin were used as endothelial cell marker and loading control respectively. J: representative chromatograms of BH4 traces in hearts from WT and Gch1fl/flTie2cre mice. K and L: BH4 levels and BH4/BH2 + B ratio were reduced in hearts from Gch1fl/flTie2cre mice compared with wild-type littermates (*P < 0.05; n = 8 and 9 per group). M and N: BH4 levels were barely detectable in primary ECs from Gch1fl/flTie2cre compared with WT mice (*P < 0.05; n = 4–6 per group). O and P: BH4 levels were comparable between primary cardiomyocytes from Gch1fl/flTie2cre mice and wild-type littermates. Absolute BH2 levels in cardiomyocytes were significantly increased in Gch1fl/flTie2cre mice compared with wild-type mice, such that the BH4/BH2 and biopterin ratio was significantly reduced in cardiomyocytes in Gch1fl/flTie2cre mice (*P < 0.05; n = 4 per group). Each data point represents an individual adult male mouse.*
## Endothelial Cell BH4 Deficiency Leads to Cardiac NOS Uncoupling with Increased Superoxide Production and Loss of Cardiac NO Generation
We next determined the effects of altering cardiac endothelial cell BH4 availability on NOS function. We first measured basal superoxide productions in whole heart homogenate by quantification of 2-hydroxyethidium (2-HE) production from dihydroethidine, using high-performance liquid chromatography (HPLC). Basal superoxide production was significantly elevated in hearts from Gch1fl/flTie2cre mice compared with wild-type controls ($P \leq 0.05$, Fig. 2, A and B). In the presence of the nonselective nitric oxide synthase inhibitor l-NAME (100 µM), the levels of NOS-derived superoxide production in wild-type hearts were significantly increased compared with untreated hearts, suggesting a tonic scavenging effect of cardiac NO on superoxide. Furthermore, there was significant inhibition of superoxide production in Gch1fl/flTie2cre hearts by the NOS inhibitor l-NAME, ($P \leq 0.05$, Fig. 2, A–D), suggesting that NOS is a source of superoxide production in Gch1fl/flTie2cre endothelial cells.
**Figure 2.:** *Superoxide production is increased, and nitric oxide bioavailability is reduced in cardiac endothelial cell BH4-deficient mice. Quantification of superoxide production, as measured by 2-hydroxyethidium (2-HE), in whole heart homogenate from Gch1fl/flTie2cre and wild-type (WT) mice using dihydroethidine (DHE) high-performance liquid chromatograph (HPLC). A: superoxide production was markedly increased in hearts from Gch1fl/flTie2cre mice compared with wild-type controls (*P < 0.05, n = 5–6 per group). B: representative trances of 2-HE and ethidium peaks in hearts from WT and Gch1fl/flTie2cre mice detected by DHE HPLC. C and D: nonselective nitric oxide synthase inhibitor NG-nitro-l-arginine methyl ester (l-NAME; 100 µM)-inhibitable fraction and polyethylene glycol superoxide dismutase, PEG-SOD (100 U/ml)-inhibitable fraction were greatly increased in Gch1fl/flTie2cre hearts compared with wild-type controls (*P < 0.05, n = 5–6 per group), respectively. E: nitrite/nitrate production in whole heart homogenate. Nitrite/nitrate production in heart homogenate from Gch1fl/flTie2cre mice was significantly decreased when compared with that from WT controls (*P < 0.05; n = 4–6 animals per group). F: levels of hydrogen peroxide from wild-type and Gch1fl/flTie2cre hearts were determined using an amperometric hydrogen peroxide microsensor electrode. Level of hydrogen peroxide production was significantly increased in endothelial cell BH4-deficient hearts (Gch1fl/flTie2cre) compared with wild-type controls (*P < 0.05; n = 6 per group). G: representative immunoblots for antioxidant proteins: catalase, Es-SOD, Mn-SOD, and Cu/Zn-SOD in Gch1fl/flTie2cre and wild-type hearts, with quantitative data, measured as percent band density in H (n = 6 per group). Each data point represents an individual adult male mouse.*
To determine the effects of endothelial cell BH4 deficiency on NO bioactivity, we measured nitrate and nitrite in heart homogenates using ozone chemiluminescence. Nitrite and nitrate production were significantly reduced in Gch1fl/flTie2cre hearts compared with wild-type controls ($P \leq 0.05$, Fig. 2E). Furthermore, hydrogen peroxide production was significantly elevated in endothelial cell BH4-deficient hearts compared with wild-type controls ($P \leq 0.05$, Fig. 2F). To investigate whether increased productions of superoxide and hydrogen peroxide and reduced NO bioactivity in Gch1fl/flTie2cre heart altered antioxidant defenses, we measured protein levels of antioxidant enzymes by Western blot. There was no change in protein levels of catalase, extracellular superoxide dismutase (ecSOD), manganese superoxide dismutase (MnSOD), or Cu/ZnSOD between Gch1fl/flTie2cre and wild-type hearts (Fig. 2, G and H). Taken together, these data demonstrate that deficiency in coronary endothelial cell BH4 leads to eNOS uncoupling, increased superoxide and hydrogen peroxide productions, and decreased NO bioactivity in myocardium from Gch1fl/flTie2cre mice.
## Specific Loss of Endothelial Cell BH4 Leads to Cardiac Dysfunction and Hypertrophy
We next investigated the effect of endothelial cell Gch1 and BH4 deficiency on cardiac function, using M-mode echocardiography (Fig. 3A). There was no difference in either fractional shortening or ejection fraction or peak systolic velocity in wild-type and Gch1fl/flTie2cre mice (Fig. 3, B–D). However, left ventricular (LV) diastolic volume and LV end dimensions were significantly reduced in Gch1fl/flTie2cre mice compared with wild-type littermates (Fig. 3E). LV diastolic and systolic volume were also significantly reduced in Gch1fl/flTie2cre mice compared with wild-type littermates (Fig. 3, F and G). LV end-diastolic thickness was significantly increased in Gch1fl/flTie2cre mice compared with wild-type littermates (Fig. 3H). Cardiac output was significantly depressed in Gch1fl/flTie2cre mice compared with wild-type controls (Fig. 3I). We next measured blood pressure in Gch1fl/flTie2cre and wild-type littermate controls using tail-cuff plethysmography. We observed that Gch1fl/flTie2cre mice have a mild increased (∼5–7 mmHg) systolic blood pressure compared with wild-type littermate controls (102 ± 3 mmHg WT vs. 109 ± 2 mmHg in Gch1fl/flTie2cre mice; Fig. 3J).
**Figure 3.:** *Loss of cardiac endothelial cell BH4 leads to cardiac dysfunction and hypertrophy. A: example of M-mode echocardiograms from Gch1fl/flTie2cre and wild-type littermates controls. B: ejection fraction. C: LV fractional shortening. D: peak systolic velocity, cardiac output (mL/min). E: systolic and diastolic LV end dimensions (mm). F and G: LV diastolic and systolic volume (μL). H: anterior and posterior LV end-diastolic thickness (mm). I: cardiac output (mL/min) in WT and Gch1fl/flTie2cre mice (*P < 0.05, n = 6 per group). J: systolic blood pressure in Gch1fl/flTie2cre and wild-type littermates were determined using tail-cuff plethysmography. K–M: gene expression of hypertrophic markers and fibrosis marker (N) in hearts from Gch1fl/flTie2cre and WT littermates (*P < 0.05; n = 7–8 per group). Each data point represents an individual adult male mouse.*
In addition, we have undertaken further studies to examine the effect of endothelial cell BH4 deficiency on blood glucose levels and lipid profiles in Gch1fl/flTie2cre mice. First, we observed that blood glucose levels after 6 h of fasting were also comparable between wild-type and Gch1fl/flTie2cre mice (Supplemental Fig. S1). Lipid profiles including total cholesterol, triglycerides, LDL, and HDL were also comparable between the groups (Supplemental Fig. S2). These findings indicate that endothelial cell BH4 deficiency does not affect the metabolic or lipid profile in Gch1fl/flTie2cre mice.
To investigate the downstream effect of endothelial cell BH4 deficiency and NOS uncoupling on myocardial remodeling and hypertrophy, we measured mRNA expression of fetal genes in heart homogenates from Gch1fl/flTie2cre mice and their littermates. The mRNA expression of hypertrophic markers including natriuretic factor type A (Nppa), type B (Nppb), and β-myosin heavy chain (Myh7) were significantly increased in Gch1fl/flTie2cre hearts compared with littermate WT controls (Fig. 3, K–M). In contrast, there was no difference in Col1a1 (a fibrosis marker) expression between the genotypes (Fig. 3N). Taken together, these findings suggest that endothelial cell BH4 deficiency leads to LV dysfunction and hypertrophy.
## Endothelial Cell BH4 Deficiency Leads to Activation of Cardiomyocytes, Increased nNOS, and Decreased eNOS Protein in Gch1fl/flTie2cre Hearts
To investigate the mechanism by which cardiac endothelial cell-specific BH4 deficiency impairs of cardiac function and increases hypertrophy, we isolated primary endothelial cells and primary cardiomyocytes from Gch1fl/flTie2cre and wild-type mice and tested whether NOS uncoupling from endothelial cells could mediate changes in cardiomyocytes. First, we observed that Erk$\frac{1}{2}$ phosphorylation was significantly increased in the whole hearts from Gch1fl/flTie2cre mice compared with wild-type littermate controls (Fig. 4, A and B). Furthermore, we found that Erk$\frac{1}{2}$ phosphorylation was significantly increased in primary cardiomyocytes from Gch1fl/flTie2cre compared with wild-type controls (Fig. 4, A and B). In contrast, we did not observe an increase in Erk$\frac{1}{2}$ phosphorylation in endothelial cells isolated from Gch1fl/flTie2cre hearts compared with littermate controls (Fig. 4, A and B). These findings suggest that loss of BH4 leads to NOS uncoupling in cardiac endothelial cells, which in turn causes changes in cardiomyocyte signaling pathways.
**Figure 4.:** *Increased phosphorylated extracellular signal-regulated kinases 1/2 in hearts and cardiomyocytes from Gch1fl/flTie2cre mice. A: representative immunoblots for phosphorylated and total proteins for extracellular signal-regulated kinases 1/2 (ERK1/2) in hearts, isolated endothelial cells (ECs), and cardiomyocytes from Gch1fl/flTie2cre mice and wild-type littermate controls. B: summary data (*P < 0.05, n = 5–6 per group). Each data point represents an individual adult male mouse.*
We next determined whether cardiac dysfunction and hypertrophy in Gch1fl/flTie2cre hearts are associated with changes in myocardial NOS isoforms. Quantitative real-time PCR and Western blot analysis demonstrated a significant increase in nNOS mRNA and protein in Gch1fl/flTie2cre hearts, accompanied by a reduction in eNOS mRNA and protein (Fig. 5, A–C). There was no significant difference in iNOS mRNA and protein expression between Gch1fl/flTie2cre hearts and wild-type hearts (Fig. 5, A–C). To investigate which cell type is responsible for the upregulation of nNOS and downregulation of eNOS protein, NOS isoforms were determined in isolated cardiomyocytes from Gch1fl/flTie2cre mice, revealing increased nNOS protein and decreased eNOS protein in isolated cardiomyocytes from Gch1fl/flTie2cre hearts (Fig. 5, D and E).
**Figure 5.:** *Deficiency in endothelial cell BH4 causes an increased nNOS expression and reduced eNOS expression in the hearts specifically in cardiomyocytes from Gch1fl/flTie2cre mice. A: quantitative real-time PCR was used to quantify endothelial cell nitric oxide synthase (eNOS), neuronal NOS (nNOS), and inducible NOS (iNOS) gene expression in hearts from Gch1fl/flTie2cre mice and wild-type controls (*P < 0.05; n = 5–9 per group). B: representative immunoblots of eNOS, nNOS, and iNOS isoforms in hearts from Gch1fl/flTie2cre mice and wild-type controls. C: summary data (*P < 0.05; n = 6 per group). D: representative immunoblots of eNOS, nNOS, and iNOS isoforms in cardiomyocytes from Gch1fl/flTie2cre mice and wild-type controls. E: summary data (*P < 0.05; n = 6 per group). Each data point represents an individual adult male mouse.*
We further investigated whether loss of Gch1/BH4 in endothelial cells alters calcium-handling proteins in cardiomyocytes. We found that phospholamban phosphorylation at the calmodulin-dependent kinase II (CaMKII)-specific site (PLB-Thr17) was increased in myocardium from endothelial cell BH4-deficient mice, but not at the protein kinase A-specific site (PLB-Ser16) (Fig. 6, A and B). There was no change in overall protein levels of phospholamban, NCX, or SERCA2A in cardiomyocytes between Gch1fl/flTie2cre and wild-type mice (Fig. 6, C and D). Taken together, these data demonstrate that endothelial cell BH4 deficiency leads to activation of cardiomyocytes, increased nNOS, and decreased eNOS protein in Gch1fl/flTie2cre hearts, associated with alteration in cardiomyocyte calcium handling.
**Figure 6.:** *Calcium-handling proteins in cardiomyocytes from wild-type and Gch1fl/flTie2cre mice. A and B: phosphorylation of phospholamban (PLB) at Thr (17) was significantly increased in cardiomyocytes from Gch1fl/flTie2cre mice (*P < 0.05; n = 6 per group), whereas the phosphorylation of phospholamban at Ser (16) was unchanged (n = 6 per group). C: representative immunoblots for sodium-calcium exchanger (NCX) and sarco(endo)plasmic reticulum Ca2+ ATPase (SERCA) in cardiomyocytes from Gch1fl/flTie2cre mice and wild-type littermate controls. D: summary data (n = 6 per group). Each data point represents an individual adult male mouse.*
## Deficient Endothelial Cell Gch1/BH4 Biosynthesis Leads to Coronary Vascular Dysfunction and Injury Following Cardiac Ischemia-Reperfusion
To investigate the specific role of endothelial cell BH4 in postischemic myocardial function and injury following ischemia-reperfusion, Langendorff perfused wild-type and Gch1fl/flTie2cre hearts were subjected to 35 min global ischemia followed by 60-min reperfusion (Fig. 7A). Myocardial function was measured using a ventricular balloon to determine LV developed pressure, LV end-diastolic pressure, and rate pressure product. LV functional recovery after 35 min of global ischemia was significantly impaired in Gch1fl/flTie2cre hearts compared with wild-type controls. This impairment in recovery was manifest as decreased LV-developed pressure (Fig. 7B) and increased diastolic pressure (Fig. 7C). The maximal rates of contraction (LV dP/dtmax) and relaxation (LV dP/dtmin) were also significantly lower in the Gch1fl/flTie2cre hearts than in the wild-type group ($P \leq 0.05$) (Fig. 7, D and E). The rate pressure product (RPP), an index of workload, was significantly decreased in Gch1fl/flTie2cre hearts compared with wild-type controls (Fig. 7F). The coronary perfusion pressure (CPP), a direct measure of coronary vascular resistance, was greatly increased in Gch1fl/flTie2cre hearts compared with wild-type controls (Fig. 7G). Importantly, the coronary flow was significantly reduced after ischemia in Gch1fl/flTie2cre hearts compared with wild-type hearts (Fig. 7I). In addition, in Gch1fl/flTie2cre hearts, infarct size, assessed by TTC staining, was significantly larger than in wild-type mice (Gch1fl/flTie2cre, 62 ± $6\%$ of the risk region; wild type, 38 ± $7\%$, $P \leq 0.05$; Fig. 7, J and K). Collectively, these data demonstrate a critical role for endothelial cell Gch1/BH4 biosynthesis on coronary vascular function, cardiac function, and myocardial injury following ischemia-reperfusion injury.
**Figure 7.:** *Loss of endothelial cell Gch1/BH4 biosynthesis leads to cardiac dysfunction and injury following cardiac ischemia-reperfusion. Cardiac function and infarct size were measured in isolated wild-type (WT) and Gch1fl/flTie2cre hearts perfused using Langendorff model. A: after baseline stabilization, wild-type and Gch1fl/flTie2cre hearts were subjected to 35 min of global ischemia and 60 min of reperfusion. B: LV developed pressure (LVDP). C: LV diastolic pressure. D: LV dP/dtmax. E: LV dP/dtmin. F: rate pressure product (RPP; mmHg). G: coronary perfusion pressure (CPP; mmHg). H: heart rate (beats/min). I: coronary flow was measured in hearts before ischemia and after periods of ischemia followed by 60 min of reperfusion (n = 7 per group). J and K: infarct size, defined by triphenyltetrazolium chloride staining and measurement as percentage of risk region, was significantly increased in hearts from Gch1fl/flTie2cre mice compared with wild-type littermate controls. Values are the means ± SE (*P < 0.05; n = 4–6 animals per group). Each data point represents an individual adult male mouse.*
## DISCUSSION
In this study, we used a mouse model of endothelial cell-targeted Gch1 deletion to test the specific requirement for endothelial cell BH4 in the regulation of cardiac function under both physiological and pathophysiological conditions. The major findings of this study are as follows: 1) endothelial cell-targeted Gch1 deletion leads to selective loss of endothelial cell BH4 in the myocardium; 2) loss of endothelial cell BH4 leads to NOS uncoupling, increased superoxide and hydrogen peroxide productions, and decreased NO bioactivity, resulting in cardiac dysfunction and mild myocardial hypertrophy; 3) selective loss of endothelial cell BH4 is associated with changes in cardiomyocytes, including increased nNOS protein, loss of eNOS protein, and increased in phospholamban phosphorylation at Ser17; and 4) specific loss of cardiac endothelial cell BH4 leads to coronary vascular dysfunction, cardiac dysfunction, and increased myocardial infarct size following ischemia-reperfusion injury. Collectively, these studies reveal specific effects of endothelial cell Gch1/BH4 biosynthesis on cardiomyocytes and on overall cardiac function under both physiological and pathophysiological conditions.
An important observation underpinning the interpretation of selective and specific roles for endothelial cell BH4 is that endothelial cell-targeted Gch1 deletion abolishes GTPCH protein expression and de novo BH4 biosynthesis in endothelial cells, and this loss of endothelial cell BH4 synthesis is not rescued by normal levels of BH4 in plasma or adjacent cells. This indicates that BH4 in endothelial cells is compartmentalized and is not amenable to uptake or recycling, at least under conditions of normal BH4 levels. High-level supplementation of sepiapterin [20, 21, 33] or BH4 [34, 35] has been shown to prevent NOS uncoupling and improved left ventricular function and I/R injury. However, it is not clear whether these effects of supraphysiological BH4 supplementation are mediated by known BH4 functions, and if so via which cell type and which mechanisms. Previous observations suggest that exogenous BH4 is not sufficient to rescue or augment BH4 levels in endothelial cells [30], whereas cardiomyocytes are amenable to exogenous BH4 supplementation [18, 23, 36].
Our study emphasizes the importance of the cross talk between the coronary endothelium and cardiomyocytes in cardiac function [37]. We found that selective loss of endothelial cell BH4 led to changes in cardiomyocyte redox signaling, gene expression, and function. Increased superoxide and hydrogen peroxide generation in endothelial cell BH4-deficient hearts is consistent with the induction of fetal gene expression (nppa, nppb, and myh7) contributing to cardiac hypertrophy. Indeed, emerging evidence suggests that H2O2 can mediate and cause cardiac dysfunction, hypertrophy, and heart failure (38–40). In eNOS knockout mice, cardiac structure, LV function, and heart rate were similar to that of wild-type mice [41]. These findings suggest that the combination of increased NOS-derived H2O2 and reduced NOS-derived NO production are likely to be important contributors to the development of cardiac dysfunction in Gch1fl/flTie2cre mice and contrasts with eNOS knockout mice where all functions of eNOS (i.e., both NO and ROS generation) are deleted.
Neuronal NOS (nNOS) has also been implicated in the regulation of basal and β-adrenergic inotropy in normal and chronically infarcted hearts. We confirmed that nNOS was upregulated and eNOS downregulated in cardiomyocytes in endothelial cell BH4-deficient mice. Consistent with these findings, increased nNOS and decreased eNOS expression have been observed in the failing rat and human hearts (42–44). It is possible that upregulation of eNOS-derived H2O2 generation from coronary endothelial cells from Gch1fl/flTie2cre mice affects eNOS protein and activity in the cardiomyocytes. Consistent with this idea, several reports have shown that H2O2 decreases eNOS protein expression and activity, at least in part by an inhibition of c-Jun activity and thus leading to a reduction in AP-1 transcription factor binding to the eNOS promoter (45–47). Thus, eNOS expression and activity have been found to be suppressed in the failing hearts, suggesting that the eNOS-mediated regulation of cardiac function and β-adrenergic response may be reduced in myocardial disease.
We have shown that Gch1fl/flTie2cre mice have a slightly increased (∼5–7 mmHg) in systolic blood pressure compared with wild-type littermate controls. Thus, it is possible that increased systolic blood pressure in Gch1fl/flTie2cre mice may contribute to cardiac dysfunction and hypertrophy in this study. However, this degree of blood pressure elevation is very modest and would not usually be considered sufficient to constitute a model of “hypertension.” Indeed, the eNOS knockout mouse has systemic hypertension with a 20–30-mmHg increase in systolic blood pressure, but at baseline, cardiac contractility was reported to be normal in eNOS knockout mice [48, 49]. Therefore, it is unlikely that a mild increase in systolic blood pressure in Gch1fl/flTie2cre mice was responsible for the cardiac dysfunction and hypertrophy in this study.
Increased superoxide and hydrogen peroxide productions in the myocardium have been shown to impair calcium handling and induce pathological cardiac changes such as fibrosis, apoptosis, and hypertrophy (50–52). We found that endothelial cell BH4-deficient mice had increased phospholamban phosphorylation at the calmodulin-dependent kinase II (CaMKII)-specific site (PLB-Thr17) but not at the protein kinase A-specific site (PLB-Ser16). nNOS modulates cardiac relaxation via effects on phospholamban phosphorylation [53]. PLB has inhibitory effect on SERCA activity and reuptake of Ca2+. Therefore, increased phospholamban phosphorylation at Ser17 abrogates the inhibitory effect of PLB on SERCA, thereby increasing SR Ca2+ reuptake and ultimately increased myocyte contractility and relaxation. This finding suggests that increased nNOS gene expression and protein in Gch1fl/flTie2cre hearts and cardiomyocytes may contribute to an increase in contraction and accelerated SR Ca2+ reuptake in cardiomyocytes, possibly by increased basal PLB phosphorylation.
Myocardial ischemia-reperfusion is associated with markedly elevated levels of ROS production [14, 54], and these ROS are central mediators of postischemic injury. In the postischemic heart, changes in coronary endothelial vascular function occur because of a reduction in eNOS-derived NO production which in turn impairs coronary flow [24]. Evidence from isolated rat hearts suggests that ischemia-induced oxidative stress leads to enhanced BH4 oxidation [11, 24], contributing to postischemic eNOS uncoupling with a resultant loss of coronary endothelium-dependent vasodilation [24]. Our observation of greatly increased coronary perfusion pressure (CPP), a direct measure of coronary vascular resistance, and reduced coronary flow in Gch1fl/flTie2cre hearts now demonstrates a specific role of endothelial cell BH4 in determining the response of the coronary microcirculation to ischemia-reperfusion. Importantly, we found that loss of endothelial cell BH4 caused a greater myocardial infarct size compared with wild-type hearts following I/R.
Collectively, these data demonstrate a critical role for endothelial cell Gch1/BH4 biosynthesis in coronary vascular function, cardiac function, and the response to ischemia-reperfusion injury. Thus, targeting endothelial cell Gch1 and BH4 biosynthesis may provide a novel therapeutic target for the prevention and treatment of cardiac dysfunction, ischemia injury, and heart failure.
## DATA AVAILABILITY
Data will be made available upon reasonable request.
## GRANTS
This study was supported by British Heart Foundation (BHF) Programme Grants RG/$\frac{12}{5}$/29576 and RG/$\frac{17}{10}$/32859, BHF Chair Award CH/$\frac{16}{1}$/32013, Wellcome Trust Grant 090532/Z/09/Z, BHF Centre of Research Excellence (Oxford) Grants RE/$\frac{13}{1}$/30181 and RE/$\frac{18}{3}$/34214, and the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre.
## DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the authors.
## AUTHOR CONTRIBUTIONS
S.C. and K.M.C. conceived and designed research; S.C., S.M.C., R.C., M.K., J.K.B., J.N.S., G.D., and M.J.C. performed experiments; S.C., S.M.C., R.C., M.K., J.K.B., J.N.S., G.D., and M.J.C. analyzed data; S.C., B.C., and K.M.C. interpreted results of experiments; S.C. prepared figures; S.C. and K.M.C. drafted manuscript; S.C. and K.M.C. edited and revised manuscript; S.C., S.M.C., R.C., M.K., J.K.B., J.N.S., G.D., M.J.C., B.C., and K.M.C. approved final version of manuscript.
## References
1. Kim H, Yun J, Kwon SM. **Therapeutic strategies for oxidative stress-related cardiovascular diseases: removal of excess reactive oxygen species in adult stem cells**. *Oxid Med Cell Longev* (2016) **2016**. DOI: 10.1155/2016/2483163
2. Bendall JK, Douglas G, McNeill E, Channon KM, Crabtree MJ. **Tetrahydrobiopterin in cardiovascular health and disease**. *Antioxid Redox Signal* (2014) **20** 3040-3077. DOI: 10.1089/ars.2013.5566
3. Xia Y, Tsai AL, Berka V, Zweier JL. **Superoxide generation from endothelial nitric-oxide synthase - A Ca2+/calmodulin-dependent and tetrahydrobiopterin regulatory process**. *J Biol Chem* (1998) **273** 25804-25808. DOI: 10.1074/jbc.273.40.25804
4. Vasquez-Vivar J, Martasek P, Whitsett J, Joseph J, Kalyanaraman B. **The ratio between tetrahydrobiopterin and oxidized tetrahydrobiopterin analogues controls superoxide release from endothelial nitric oxide synthase: an EPR spin trapping study**. *Biochem J* (2002) **362** 733-739. DOI: 10.1042/0264-6021:3620733
5. Crabtree MJ, Tatham AL, Al-Wakeel Y, Warrick N, Hale AB, Cai S, Channon KM, Alp NJ. **Quantitative regulation of intracellular endothelial nitric oxide synthase (eNOS) coupling by both tetrahydrobiopterin-eNOS stoichiometry and biopterin redox status: Insights from cells with tet-regulated GTP cyclohydrolase I expression**. *J Biol Chem* (2009) **284** 1136-1144. DOI: 10.1074/jbc.M805403200
6. Channon KM. **Tetrahydrobiopterin - regulator of endothelial nitric oxide synthase in vascular disease**. *Trends Cardiovasc Med* (2004) **14** 323-327. DOI: 10.1016/j.tcm.2004.10.003
7. Vasquez-Vivar J, Kalyanaraman B, Martasek P, Hogg N, Masters BSS, Karoui H, Tordo P, Pritchard KA. **Superoxide generation by endothelial nitric oxide synthase: The influence of cofactors**. *Proc Natl Acad Sci USA* (1998) **95** 9220-9225. DOI: 10.1073/pnas.95.16.9220
8. Chuaiphichai S, Crabtree MJ, Mcneill E, Hale AB, Trelfa L, Channon KM, Douglas G. **A key role for tetrahydrobiopterin- dependent endothelial NOS regulation in resistance arteries: studies in endothelial cell tetrahydrobiopterin-deficient mice**. *Br J Pharmacol* (2017) **174** 657-671. DOI: 10.1111/bph.13728
9. Zhang L, Rao F, Zhang K, Khandrika S, Das M, Vaingankar SM, Bao X, Rana BK, Smith DW, Wessel J, Salem RM, Rodriguez-Flores JL, Mahata SK, Schork NJ, Ziegler MG, O'Connor DT. **Discovery of common human genetic variants of GTP cyclohydrolase 1 (GCH1) governing nitric oxide, autonomic activity, and cardiovascular risk**. *J Clin Invest* (2007) **117** 2658-2671. DOI: 10.1172/JCI31093
10. Mayahi L, Mason L, Bleasdale-Barr K, Donald A, Trender-Gerhard I, Sweeney MG, Davis MB, Wood N, Mathias CJ, Watson L, Pellerin D, Heales S, Deanfield JE, Bhatia K, Murray-Rust J, Hingorani AD. **Endothelial, sympathetic, and cardiac function in inherited (6R)-L-erythro-5,6,7,8-tetrahydro-L-biopterin deficiency**. *Circ Cardiovasc Genet* (2010) **3** 513-522. DOI: 10.1161/CIRCGENETICS.110.957605
11. Moens AL, Takimoto E, Tocchetti CG, Chakir K, Bedja D, Cormaci G, Ketner EA, Majmudar M, Gabrielson K, Halushka MK, Mitchell JB, Biswal S, Channon KM, Wolin MS, Alp NJ, Paolocci N, Champion HC, Kass DA. **Reversal of cardiac hypertrophy and fibrosis from pressure overload by tetrahydrobiopterin - efficacy of recoupling nitric oxide synthase as a therapeutic strategy**. *Circulation* (2008) **117** 2626-2636. DOI: 10.1161/CIRCULATIONAHA.107.737031
12. Cubberley RR, Alderton WK, Boyhan A, Charles IG, Lowe PN, Old RW. **Cysteine-200 of human inducible nitric oxide synthase is essential for dimerization of haem domains and for binding of haem, nitroarginine and tetrahydrobiopterin**. *Biochem J* (1997) **323** 141-146. DOI: 10.1042/bj3230141
13. Shimizu S, Ishii M, Kawakami Y, Momose K, Yamamoto T. **Protective effects of tetrahydrobiopterin against nitric oxide-induced endothelial cell death**. *Life Sci* (1998) **63** 1585-1592. DOI: 10.1016/s0024-3205(98)00427-5
14. Bolli R, Jeroudi MO, Patel BS, DuBose CM, Lai EK, Roberts R, McCay PB. **Direct evidence that oxygen-derived free radicals contribute to postischemic myocardial dysfunction in the intact dog**. *Proc Natl Acad Sci USA* (1989) **86** 4695-4699. DOI: 10.1073/pnas.86.12.4695
15. Yamashiro S, Noguchi K, Kuniyoshi Y, Koja K, Sakanashi M. **Role of tetrahydrobiopterin on ischemia-reperfusion injury in isolated perfused rat hearts**. *J Cardiovasc Surg (Torino)* (2003) **44** 37-49. PMID: 12627070
16. Settergren M, Bohm F, Malmstrom RE, Channon KM, Pernow J. **L-Arginine and tetrahydrobiopterin protects against ischemia/reperfusion-induced endothelial dysfunction in patients with type 2 diabetes mellitus and coronary artery disease**. *Atherosclerosis* (2009) **204** 73-78. DOI: 10.1016/j.atherosclerosis.2008.08.034
17. Perkins KAA, Pershad S, Chen Q, McGraw S, Adams JS, Zambrano C, Krass S, Emrich J, Bell B, Iyamu M, Prince C, Kay H, Teng JCW, Young LH. **The effects of modulating eNOS activity and coupling in ischemia/reperfusion (I/R)**. *Naunyn Schmiedebergs Arch Pharmacol* (2012) **385** 27-38. DOI: 10.1007/s00210-011-0693-z
18. Takimoto E, Champion HC, Li MX, Ren SX, Rodriguez ER, Tavazzi B, Lazzarino G, Paolocci N, Gabrielson KL, Wang YB, Kass DA. **Oxidant stress from nitric oxide synthase-3 uncoupling stimulates cardiac pathologic remodeling from chronic pressure load**. *J Clin Invest* (2005) **115** 1221-1231. DOI: 10.1172/JCI21968
19. Silberman GA, Fan THM, Liu H, Jiao Z, Xiao HD, Lovelock JD, Boulden BM, Widder J, Fredd S, Bernstein KE, Wolska BM, Dikalov S, Harrison DG, Dudley SC. **Uncoupled cardiac nitric oxide synthase mediates diastolic dysfunction**. *Circulation* (2010) **121** 519-528. DOI: 10.1161/CIRCULATIONAHA.109.883777
20. Tiefenbacher CP, Lee CH, Kapitza J, Dietz V, Niroomand F. **Sepiapterin reduces postischemic injury in the rat heart**. *Pflugers Arch* (2003) **447** 1-7. DOI: 10.1007/s00424-003-1131-y
21. Baumgardt SL, Paterson M, Leucker TM, Fang J, Zhang DX, Bosnjak ZJ, Warltier DC, Kersten JR, Ge ZD. **Chronic co-administration of sepiapterin and L-citrulline ameliorates diabetic cardiomyopathy and myocardial ischemia/reperfusion injury in obese type 2 diabetic mice**. *Circ Heart Fail* (2016) **9**. DOI: 10.1161/CIRCHEARTFAILURE.115.002424
22. Takimoto E, Kass DA. **Role of oxidative stress in cardiac hypertrophy and remodeling**. *Hypertension* (2007) **49** 241-248. DOI: 10.1161/01.HYP.0000254415.31362.a7
23. Carnicer R, Hale AB, Suffredini S, Liu X, Reilly S, Zhang MH, Surdo NC, Bendall JK, Crabtree MJ, Lim GB, Alp NJ, Channon KM, Casadei B. **Cardiomyocyte GTP cyclohydrolase 1 and tetrahydrobiopterin increase NOS1 activity and accelerate myocardial relaxation**. *Circ Res* (2012) **111** 718-727. DOI: 10.1161/CIRCRESAHA.112.274464
24. Dumitrescu C, Biondi R, Xia Y, Cardounel AJ, Druhan LJ, Ambrosio G, Zweier JL. **Myocardial ischemia results in tetrahydrobiopterin (BH4) oxidation with impaired endothelial function ameliorated by BH4**. *Proc Natl Acad Sci USA* (2007) **104** 15081-15086. DOI: 10.1073/pnas.0702986104
25. Giraldez RR, Panda A, Zweier JL. **Endothelial dysfunction does not require loss of endothelial nitric oxide synthase**. *Am J Physiol Heart Circ Physiol* (2000) **278** H2020-H2027. DOI: 10.1152/ajpheart.2000.278.6.H2020
26. Tiefenbacher CP, Chilian WM, Mitchell M, DeFily DV. **Restoration of endothelium-dependent vasodilation after reperfusion injury by tetrahydrobiopterin**. *Circulation* (1996) **94** 1423-1429. DOI: 10.1161/01.cir.94.6.1423
27. Aragón JP, Condit ME, Bhushan S, Predmore BL, Patel SS, Grinsfelder DB, Gundewar S, Jha S, Calvert JW, Barouch LA, Lavu M, Wright HM, Lefer DJ. **Beta3-adrenoreceptor stimulation ameliorates myocardial ischemia-reperfusion injury via endothelial nitric oxide synthase and neuronal nitric oxide synthase activation**. *J Am Coll Cardiol* (2011) **58** 2683-2691. DOI: 10.1016/j.jacc.2011.09.033
28. Chuaiphichai S, Rashbrook VS, Hale AB, Trelfa L, Patel J, McNeill E, Lygate CA, Channon KM, Douglas G. **Endothelial cell tetrahydrobiopterin modulates sensitivity to Ang (Angiotensin) II-induced vascular remodeling**. *Hypertension* (2018) **72** 128-138. DOI: 10.1161/HYPERTENSIONAHA.118.11144
29. Chuaiphichai S, McNeill E, Douglas G, Crabtree MJ, Bendall JK, Hale AB, Alp NJ, Channon KM. **Cell-autonomous role of endothelial GTP cyclohydrolase 1 and tetrahydrobiopterin in blood pressure regulation**. *Hypertension* (2014) **64** 530-540. DOI: 10.1161/HYPERTENSIONAHA.114.03089
30. Chuaiphichai S, Yu GZ, Tan CMJ, Whiteman C, Douglas G, Dickinson Y, Drydale EN, Appari M, Zhang W, Crabtree MJ, McNeill E, Hale AB, Lewandowski AJ, Alp NJ, Vatish M, Leeson P, Channon KM. **Endothelial GTPCH (GTP cyclohydrolase 1) and tetrahydrobiopterin regulate gestational blood pressure, uteroplacental remodeling, and fetal growth**. *Hypertension* (2021) **78** 1871-1884. DOI: 10.1161/HYPERTENSIONAHA.120.1764
31. Chuaiphichai S, Starr A, Nandi M, Channon KM, McNeill E. **Endothelial cell tetrahydrobiopterin deficiency attenuates LPS-induced vascular dysfunction and hypotension**. *Vascul Pharmacol* (2016) **77** 69-79. DOI: 10.1016/j.vph.2015.08.009
32. Crabtree MJ, Tatham AL, Hale AB, Alp NJ, Channon KM. **Critical role for tetrahydrobiopterin recycling by dihydrofolate reductase in regulation of endothelial nitric-oxide synthase coupling: relative importance of the de novo biopterin synthesis versus salvage pathways**. *J Biol Chem* (2009) **284** 28128-28136. DOI: 10.1074/jbc.M109.041483
33. Jo H, Otani H, Jo F, Shimazu T, Okazaki T, Yoshioka K, Fujita M, Kosaki A, Iwasaka T. **Inhibition of nitric oxide synthase uncoupling by sepiapterin improves left ventricular function in streptozotocin-induced diabetic mice**. *Clin Exp Pharmacol Physiol* (2011) **38** 485-493. DOI: 10.1111/j.1440-1681.2011.05535.x
34. Xie L, Talukder MAH, Sun J, Varadharaj S, Zweier JL. **Liposomal tetrahydrobiopterin preserves eNOS coupling in the post-ischemic heart conferring in vivo cardioprotection**. *J Mol Cell Cardiol* (2015) **86** 14-22. DOI: 10.1016/j.yjmcc.2015.06.015
35. Okazaki T, Otani H, Shimazu T, Yoshioka K, Fujita M, Katano T, Ito S, Iwasaka T. **Reversal of inducible nitric oxide synthase uncoupling unmasks tolerance to ischemia/reperfusion injury in the diabetic rat heart**. *J Mol Cell Cardiol* (2011) **50** 534-544. DOI: 10.1016/j.yjmcc.2010.12.010
36. Hashimoto T, Sivakumaran V, Carnicer R, Zhu GS, Hahn VS, Bedja D, Recalde A, Duglan D, Channon KM, Casadei B, Kass DA. **Tetrahydrobiopterin protects against hypertrophic heart disease independent of myocardial nitric oxide synthase coupling**. *J Am Heart Assoc* (2016) **5**. DOI: 10.1161/JAHA.116.003208
37. Zhang M, Shah AM. **ROS signalling between endothelial cells and cardiac cells**. *Cardiovas Res* (2014) **102** 249-257. DOI: 10.1093/cvr/cvu050
38. Raut GK, Manchineela S, Chakrabarti M, Bhukya CK, Naini R, Venkateshwari A, Reddy VD, Mendonza JJ, Suresh Y, Nallari P, Bhadra MP. **Imine stilbene analog ameliorate isoproterenol-induced cardiac hypertrophy and hydrogen peroxide-induced apoptosis**. *Free Radic Biol Med* (2020) **153** 80-88. DOI: 10.1016/j.freeradbiomed.2020.04.014
39. Song R, Zhang J, Zhang LJ, Wang GH, Wo D, Feng J, Li XC, Li J. **H2O2 induces myocardial hypertrophy in H9c2 cells: a potential role of Ube3a**. *Cardiovasc Toxicol* (2015) **15** 23-28. DOI: 10.1007/s12012-014-9264-0
40. Steinhorn B, Sorrentino A, Badole S, Bogdanova Y, Belousov V, Michel T. **Chemogenetic generation of hydrogen peroxide in the heart induces severe cardiac dysfunction**. *Nature communications* (2018) **9**. DOI: 10.1038/s41467-018-06533-2
41. Kuhlencordt PJ, Gyurko R, Han F, Scherrer-Crosbie M, Aretz TH, Hajjar R, Picard MH, Huang PL. **Accelerated atherosclerosis, aortic aneurysm formation, and ischemic heart disease in apolipoprotein E/endothelial nitric oxide synthase double-knockout mice**. *Circulation* (2001) **104** 448-454. DOI: 10.1161/hc2901.091399
42. Damy T, Ratajczak P, Robidel E, Bendall JK, Oliviero P, Boczkowski J, Ebrahimian T, Marotte F, Samuel JL, Heymes C. **Up-regulation of cardiac nitric oxide synthase 1-derived nitric oxide after myocardial infarction in senescent rats**. *FASEB J* (2003) **17** 1-22. DOI: 10.1096/fj.02-1208fje
43. Damy T, Ratajczak P, Shah AM, Camors E, Marty I, Hasenfuss G, Marotte F, Samuel JL, Heymes C. **Increased neuronal nitric oxide synthase-derived NO production in the failing human heart**. *Lancet* (2004) **363** 1365-1367. DOI: 10.1016/S0140-6736(04)16048-0
44. Bendall JK, Damy T, Ratajczak P, Loyer X, Monceau V, Marty I, Milliez P, Robidel E, Marotte F, Samuel JL, Heymes C. **Role of myocardial neuronal nitric oxide synthase-derived nitric oxide in beta-adrenergic hyporesponsiveness after myocardial infarction-induced heart failure in rat**. *Circulation* (2004) **110** 2368-2375. DOI: 10.1161/01.CIR.0000145160.04084.AC
45. Kumar S, Sun X, Wiseman DA, Tian J, Umapathy NS, Verin AD, Black SM. **Hydrogen peroxide decreases endothelial nitric oxide synthase promoter activity through the inhibition of Sp1 activity**. *DNA Cell Biol* (2009) **28** 119-129. DOI: 10.1089/dna.2008.0775
46. Kumar S, Sun XT, Wedgwood S, Black SM. **Hydrogen peroxide decreases endothelial nitric oxide synthase promoter activity through the inhibition of AP-1 activity**. *Am J Physiol Lung Cell Mol Physiol* (2008) **295** L370-L377. DOI: 10.1152/ajplung.90205.2008
47. Wedgwood S, Steinhorn RH, Bunderson M, Wilham J, Lakshminrusimha S, Brennan LA, Black SM. **Increased hydrogen peroxide downregulates soluble guanylate cyclase in the lungs of lambs with persistent pulmonary hypertension of the newborn**. *Am J Physiol Lung Cell Mol Physiol* (2005) **289** L660-L666. DOI: 10.1152/ajplung.00369.2004
48. Huang PL, Huang ZH, Mashimo H, Bloch KD, Moskowitz MA, Bevan JA, Fishman MC. **Hypertension in mice lacking the gene for endothelial nitric-oxide synthase**. *Nature* (1995) **377** 239-242. DOI: 10.1038/377239a0
49. Shesely EG, Maeda N, Kim HS, Desai KM, Krege JH, Laubach VE, Sherman PA, Sessa WC, Smithies O. **Elevated blood pressures in mice lacking endothelial nitric oxide synthase**. *Proc Natl Acad Sci USA* (1996) **93** 13176-13181. DOI: 10.1073/pnas.93.23.13176
50. Zhang M, Prosser BL, Bamboye MA, Gondim ANS, Santos CX, Martin D, Ghigo A, Perino A, Brewer AC, Ward CW, Hirsch E, Lederer WJ, Shah AM. **Contractile function during angiotensin-ii activation increased Nox2 activity modulates cardiac calcium handling via phospholamban phosphorylation**. *J Am Coll Cardiol* (2015) **66** 261-272. DOI: 10.1016/j.jacc.2015.05.020
51. Joseph LC, Avula UMR, Wan EY, Reyes MV, Lakkadi KR, Subramanyam P, Nakanishi K, Homma S, Muchir A, Pajvani UB, Thorp EB, Reiken SR, Marks AR, Colecraft HM, Morrow JP. **Dietary saturated fat promotes arrhythmia by activating NOX2 (NADPH oxidase 2)**. *Circ Arrhythm Electrophysiol* (2019) **12**. DOI: 10.1161/CIRCEP.119.007573
52. Donoso P, Finkelstein JP, Montecinos L, Said M, Sanchez G, Vittone L, Bull R. **Stimulation of NOX2 in isolated hearts reversibly sensitizes RyR2 channels to activation by cytoplasmic calcium**. *J Mol Cell Cardiol* (2014) **68** 38-46. DOI: 10.1016/j.yjmcc.2013.12.028
53. Zhang YH, Zhang MH, Sears CE, Emanuel K, Redwood C, El-Armouche A, Kranias EG, Casadei B. **Reduced phospholamban phosphorylation is associated with impaired relaxation in left ventricular myocytes from neuronal NO synthase-deficient mice**. *Circ Res* (2008) **102** 242-249. DOI: 10.1161/CIRCRESAHA.107.164798
54. Zweier JL, Flaherty JT, Weisfeldt ML. **Direct measurement of free radical generation following reperfusion of ischemic myocardium**. *Proc Natl Acad Sci* (1987) **84** 1404-1407. DOI: 10.1073/pnas.84.5.1404
|
---
title: Association of weight-adjusted-waist index with asthma prevalence and the age
of first asthma onset in United States adults
authors:
- Longshan Yu
- Yan Chen
- Ming Xu
- Rongfu Li
- Juan Zhang
- Shouwei Zhu
- Zongbao He
- Mingwei Chen
- Gaosheng Wang
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9988541
doi: 10.3389/fendo.2023.1116621
license: CC BY 4.0
---
# Association of weight-adjusted-waist index with asthma prevalence and the age of first asthma onset in United States adults
## Abstract
### Objective
The objective of this study was to assess whether the weight-adjusted-waist index(WWI) is associated with the prevalence of asthma and age when first asthma onset appears in US adults.
### Methods
For analysis we selected participants from the National Health and Nutrition Examination Survey(NHANES)database between 2001 and 2018. A dose-response curve was calculated using logistic regression,subgroup analysis,and a dose-response curve.
### Results
The study included 44480 people over the age of 20,including 6061 reported with asthma, and the increase in asthma prevalence was $15\%$ associated with each unit increase in the WWI, after adjusting for all confounders(odds ratio(OR)=1.15,$95\%$ CI:1.11,1.20). The sensitivity analysis was performed by trichotomizing the WWI, and compared to the lowest tertile, the highest tertile WWI group displayed a $29\%$ increase in asthma prevalence(OR=1.29,$95\%$ CI:1.19,1.40). A nonlinear correlation was found between the WWI index and the risk of asthma onset, with a threshold saturation effect indicating an inflection point of 10.53 (log-likelihood ratio test, $P \leq 0.05$), as well as a positive linear correlation with age at first asthma onset.
### Conclusions
A higher WWI index was associated with an increased prevalence of asthma and an older age of first asthma onset.
## Introduction
Asthma is a common chronic respiratory disease. Exacerbations are inevitable for asthma patients,even following medical guidance, resulting in further decline in lung function [1]. In the past few decades, asthma prevalence has steadily increased. According to the 2008-2010 Global Burden of Disease(GBD)study, there are 334 million people worldwide suffer from asthma [2]. Asthma led to 1.6 million hospitalizations and 183,000 emergency department visits in 2017 [3]. In 2009, deaths due to asthma per 10,000 people with asthma were 1.9 in adults and 0.3 in children [4]. A large portion of the direct medical costs of asthma are related to hospitalization for severe or poorly controlled asthma [5]. As a result,asthma has become one of the most common diseases worldwide,resulting in a significant burden on society [6]. By identifying its risk factors,including smoking, alcohol consumption, air pollution, and occupational exposures, asthma can be prevented (6–8).
There is an increase in asthma prevalence due to unhealthy dietary patterns becoming more prevalent. According to research, obese people in the United States are responsible for 250,000 asthma cases a year [9]. Adolescents with obesity and overweight subjects to increased risk of asthma [10]. The metabolic complications associated with obesity are not the same for everyone. Traditionally, body mass index(BMI)is used to assess obesity, but it cannot differentiate between lean body mass and fat mass [11]. The presence of visceral adiposity in conjunction with central obesity can be more relevant to poor metabolic characteristics and is increasingly valued by researchers [12]. Furthermore, numerous studies indicate that visceral adipose tissue is more closely associated with diabetes, hypertension, cardiovascular disease and cardiometabolic risk factors than subcutaneous adipose tissue (13–15). In order to assess body fat amount and distribution, a variety of methods are used, including densitometry(dual-energy X-ray absorptiometry, DXA), magnetic resonance imaging(MRI), computed tomography(CT), and mechanical methods. These methods are characterized with high accuracy in assessing body fat,and the first three provide fat imaging and location [16]. Since these processes are technically complex, time-consuming and high cost, they cannot be routinely used in clinical settings. The weight-adjusted waist index(WWI) was proposed [17] in 2018. Comparing to BMI, the WWI is a better indicator of fat and muscle mass composition, and it primarily reflects central obesity, independent of body weight [18]. Among adults with increased WWI, there was an increased prevalence of hypertension [19], proteinuria [11], cardiovascular mortality [20], and hyperuricemia [21]. Nevertheless, no studies have been conducted to determine, if WWI is associated with asthma prevalence. Our objective of this research is to determine the value of the WWI in estimating asthma prevalence in United States(US)adults.
## Study design and participants
Using baseline clinical data from the National Health and Nutrition Examination Survey(NHANES) from 2001 to 2018, the Centers for Disease Control and Prevention(CDC) monitored US population health every other year using cross-sectional survey methods. A written consent form was submitted by every participant during the NHANES study, which was approved and reviewed by the National Center for Health Statistics Institutional Review Board (NCHS). Surveys were conducted over nine consecutive two-year periods and asthma questionnaires were included in the evaluation. The age at which participants first developed asthma was recorded for those who explicitly answered whether they had asthma and whether it was their first time. Participants in the survey totaled 91,351. The following exclusion criteria were used (Figure 1). Finally a total of 44,480 cases were included in this study, including 6061 self-reported ones.
**Figure 1:** *The participants selecting flow chart.*
## Collection and definition of data
As an exposure variable,the WWI index was designed. The WWI for each participant was calculated as WC in centimeters divided by the square root of weight in kilograms and then rounded to two decimal places. To measure asthma, questionnaires were used, including: The occurrence of asthma and age at first asthma onset were designed as outcome variables.
The multivariate adjusted models summarized potential relationship between the WWI index and asthma. Covariates in our study includes: Besides, the average consumption of the two 24-hour dietary recalls was used for the analysis of all participants.
When continuous variables had a large number of missing values, we converted them into categorical variables. Details of the measurement procedures using the study variables are available at http://www.cdc.gov/nchs/nhanes/s/.
## Statistical methods
A complete statistical analysis of the NHANES data was conducted with three types of sampling weights, stratifications, and clusterings to illustrate the manner of selected participant. Sample weights reflecting selection and response probabilities were used to generate unbiased national estimates. New sampling weights for the combined survey cycle were constructed by dividing the 2-year weights for each cycle by 6 according to the NHANES analysis guidelines. The survey design R package was used to interpret the complex multistage stratified sampling technique of NHANES using the weights provided by the dataset. Continuous variables were represented with weighted survey means and $95\%$ confidence intervals, and categorical variables were represented with weighted survey means and $95\%$ confidence intervals. To exclude cointegration, we used the cointegration test. A sample with VIF greater than 5 was considered to have a cointegration problem. We used multiple logistic regression models to study the relationship between the WWI index, different trichotomies of the WWI index,and asthma prevalence based on the guidelines [22]. In model 1, no covariates were adjusted for. In model 2, gender, age, race, marital status and education level adjustment was applied. In model 3, all the variables listed above was adjusted. Smoothed curve fitting(penalized spline method) and generalized additive model (GAM) regression were performed to further evaluatethe relationship between the WWI index and asthma prevalence. Inflection point values were obtained by a likelihood ratio test when a nonlinear relationship was determined to exist. Multiple regression analyses were performed stratified by sex, age, race, hypertension, diabetes mellitus and presence of blood relatives with asthma. $P \leq 0.05$ was considered statistically significant. All analyses were performed using Empower software. www.empowerstats.com(X&Y Solutions, Inc., Boston, Massachusetts, USA) and R version 4.2.0 (http://www. R-project.org,The R Foundation).
## Participant characteristics
The demographic characteristics of the included participants are shown in Table 1. Compared with the controlgroup, the asthma group had a WWI index of 10.99(10.95,11.03), higher than 10.92(10.90,10.94)in the control group. $58.93\%$ of the participants in the asthma group were females.
**Table 1**
| Characteristic | Non-asthma formers (n=38419) | Asthma formers (n=6061) | P-value |
| --- | --- | --- | --- |
| WWI Index | 10.92 (10.90,10.94) | 10.99 (10.95,11.03) | <0.0001 |
| Gender (%) | | | <0.0001 |
| Male | 49.47 (48.94,50.00) | 41.07 (39.45,42.72) | |
| Female | 50.53 (50.00,51.06) | 58.93 (57.28,60.55) | |
| Race (%) | | | <0.0001 |
| Mexican American | 14.15 (12.63,15.82) | 10.57 (9.20,12.11) | |
| White | 67.80 (65.56,69.95) | 69.91 (67.40,72.31) | |
| Black | 10.88 (9.76,12.12) | 12.74 (11.33,14.28) | |
| Other Race | 7.17 (6.52,7.89) | 6.79 (5.90,7.80) | |
| Education Level (%) | | | 0.0064 |
| Less than high school | 44.32 (42.90,45.75) | 41.53 (39.53,43.56) | |
| High school | 46.31 (45.03,47.59) | 47.90 (45.90,49.91) | |
| More than high school | 9.37 (8.36,10.49) | 10.57 (8.73,12.75) | |
| Marital Status (%) | | | <0.0001 |
| Cohabitation | 65.10 (64.08,66.10) | 58.41 (56.67,60.13) | |
| Solitude | 34.90 (33.90,35.92) | 41.59 (39.87,43.33) | |
| Alcohol (%) | | | 0.433 |
| Yes | 63.03 (61.69,64.35) | 64.13 (62.34,65.88) | |
| No | 20.81 (19.66,22.00) | 20.10 (18.68,21.60) | |
| Unclear | 16.16 (15.37,16.99) | 15.77 (14.45,17.19) | |
| High Blood Pressure (%) | | | <0.0001 |
| Yes | 29.61 (28.79,30.45) | 34.96 (33.23,36.72) | |
| No | 70.39 (69.55,71.21) | 65.04 (63.28,66.77) | |
| Diabetes (%) | | | 0.0002 |
| Yes | 8.47 (8.08,8.89) | 10.20 (9.33,11.14) | |
| No | 91.53 (91.11,91.92) | 89.80 (88.86,90.67) | |
| Smoked (%) | | | <0.0001 |
| Yes | 45.15 (44.18,46.13) | 49.95 (48.05,51.85) | |
| No | 54.85 (53.87,55.82) | 50.05 (48.15,51.95) | |
| Physical Activity (%) | | | 0.7076 |
| Never | 28.89 (28.01,29.80) | 28.24 (26.68,29.86) | |
| Moderate | 31.70 (31.01,32.40) | 31.81 (30.27,33.38) | |
| Vigorous | 39.40 (38.47,40.35) | 39.95 (38.05,41.88) | |
| Blood relatives had asthma (%) | | | <0.0001 |
| Yes | 17.79 (17.21,18.38) | 40.90 (39.33,42.50) | |
| No | 80.39 (79.77,81.01) | 55.63 (53.96,57.28) | |
| Unclear | 1.82 (1.65,2.00) | 3.47 (2.90,4.15) | |
| Coronary Artery Disease | | | 0.0108 |
| Yes | 3.25 (2.96,3.56) | 4.15 (3.48,4.95) | |
| No | 96.75 (96.44,97.04) | 95.85 (95.05,96.52) | |
| Cancers | | | 0.0006 |
| Yes | 9.28 (8.86,9.72) | 11.30 (10.20,12.50) | |
| No | 90.72 (90.28,91.14) | 88.70 (87.50,89.80) | |
| PIR | | | <0.0001 |
| <1.3 | 18.71 (17.78,19.68) | 24.36 (22.78,26.01) | |
| ≥1.3<3.5 | 33.69 (32.72,34.68) | 31.94 (30.01,33.93) | |
| ≥3.5 | 40.82 (39.37,42.28) | 37.31 (35.00,39.68) | |
| Unclear | 6.78 (6.27,7.33) | 6.39 (5.55,7.35) | |
| Total Kcal (%) | | | 0.2072 |
| Lower | 39.97 (39.25,40.71) | 40.44 (38.72,42.17) | |
| Higher | 47.15 (46.28,48.01) | 45.71 (43.84,47.60) | |
| Unclear | 12.88 (12.24,13.54) | 13.85 (12.61,15.19) | |
| Total Sugar (%) | | | 0.1271 |
| Lower | 38.22 (37.51,38.94) | 37.07 (35.50,38.67) | |
| Higher | 39.28 (38.47,40.10) | 38.96 (37.38,40.57) | |
| Unclear | 22.49 (21.84,23.16) | 23.96 (22.57,25.42) | |
| Total Water (%) | | | 0.0165 |
| Lower | 40.07 (39.30,40.85) | 41.81 (39.98,43.65) | |
| Higher | 47.05 (46.19,47.91) | 44.34 (42.39,46.31) | |
| Unclear | 12.88 (12.24,13.54) | 13.85 (12.61,15.19) | |
| Total Sugar (%) | | | 0.0884 |
| Lower | 39.37 (38.64,40.10) | 40.29 (38.58,42.02) | |
| Higher | 47.75 (46.93,48.57) | 45.86 (44.02,47.71) | |
| Unclear | 12.88 (12.24,13.54) | 13.85 (12.61,15.19) | |
| Serum Cholesterol (%) | | | 0.0001 |
| Lower | 46.71 (45.86,47.57) | 50.19 (48.63,51.76) | |
| Higher | 48.98 (48.08,49.88) | 45.31 (43.77,46.85) | |
| Unclear | 4.31 (4.00,4.64) | 4.50 (3.84,5.27) | |
| Serum Glucose (%) | | | 0.7931 |
| Lower | 48.48 (47.62,49.35) | 48.69 (47.04,50.33) | |
| Higher | 47.22 (46.32,48.13) | 46.82 (45.12,48.52) | |
| Unclear | 4.29 (3.98,4.62) | 4.50 (3.84,5.27) | |
| Serum Uric Acid | | | 0.5003 |
| Lower | 46.30 (45.58,47.02) | 47.02 (45.45,48.59) | |
| Higher | 49.39 (48.63,50.15) | 48.45 (46.71,50.19) | |
| Unclear | 4.31 (4.00,4.64) | 4.54 (3.87,5.30) | |
## Higher WWI index was associated with higher asthma prevalence
According to WWI index data, asthma prevalence is positively related to the WWI index. Likewise, the full-adjusted model (model 3) showed a stable relationship between WWI index and asthma(OR=1.15, $95\%$ CI:1.11,1.20), indicating a $15\%$ increase in asthma risk per unit increase. Additionally, we converted the WWI index from a continuous number to a categorical number (triple quantile) to analyze sensitivity. As shown in Table 2, Tertile 3 has a $29\%$higher risk of asthma occurrence(OR=1.29, $95\%$ CI:1.19,1.40) than Tertile 1 in terms of WWI index (Table 2).
**Table 2**
| Characteristic | Model 1 OR (95%CI) | Model 2 OR (95%CI) | Model 3 OR (95%CI) |
| --- | --- | --- | --- |
| WWI Index | 1.09 (1.06,1.12) | 1.26 (1.21,1.30) | 1.15 (1.11,1.20) |
| Categories | Categories | Categories | Categories |
| Tertile 1(7.59-10.38) | 1 | 1 | 1 |
| Tertile 2(10.38-11.71) | 0.93 (0.87,1.00) | 1.13 (1.05,1.21) | 1.05 (0.97,1.13) |
| Tertile 3(11.71-15.70) | 1.17 (1.10,1.25) | 1.52 (1.41,1.64) | 1.29 (1.19,1.40) |
## Subgroup analysis
To assess whether the correlation between the WWI index and asthma is robust,subgroup analyses were conducted. Results including:
**Table 3**
| Characteristic | Model 1 OR (95%CI) | Model 2 OR (95%CI) | Model 3 OR (95%CI) |
| --- | --- | --- | --- |
| Stratified by gender | Stratified by gender | Stratified by gender | Stratified by gender |
| Male | 0.94 (0.89,0.99) | 1.28 (1.20,1.36) | 1.19 (1.11,1.28) |
| Female | 1.13 (1.09,1.18) | 1.27 (1.22,1.33) | 1.17 (1.11,1.23) |
| Stratified by age (years) | Stratified by age (years) | Stratified by age (years) | Stratified by age (years) |
| 20-39 | 1.03 (0.97,1.08) | 1.11 (1.05,1.18) | 1.07 (1.01,1.14) |
| 40-59 | 1.38 (1.29,1.47) | 1.38 (1.29,1.47) | 1.21 (1.12,1.30) |
| 60-85 | 1.32 (1.24,1.42) | 1.33 (1.24,1.42) | 1.25 (1.16,1.35) |
| Stratified by race | Stratified by race | Stratified by race | Stratified by race |
| Mexican American | 1.13 (1.05,1.22) | 1.12 (1.02,1.22) | 1.07 (0.97,1.17) |
| White | 1.09 (1.04,1.15) | 1.31 (1.24,1.39) | 1.19 (1.12,1.27) |
| Black | 1.16 (1.09,1.24) | 1.28 (1.19,1.38) | 1.18 (1.09,1.28) |
| Other Race | 1.20 (1.07,1.34) | 1.36 (1.19,1.55) | 1.27 (1.09,1.47) |
| Stratified by hypertension | Stratified by hypertension | Stratified by hypertension | Stratified by hypertension |
| Yes | 1.15 (1.09,1.22) | 1.29 (1.22,1.38) | 1.22 (1.14,1.31) |
| No | 0.98 (0.94,1.03) | 1.17 (1.12,1.23) | 1.13 (1.07,1.19) |
| Stratified by diabetes | Stratified by diabetes | Stratified by diabetes | Stratified by diabetes |
| Yes | 1.31 (1.19,1.45) | 1.47 (1.32,1.63) | 1.36 (1.22,1.52) |
| No | 1.04 (1.00,1.07) | 1.25 (1.20,1.30) | 1.18 (1.13,1.23) |
| Stratified by blood relative had asthma | Stratified by blood relative had asthma | Stratified by blood relative had asthma | Stratified by blood relative had asthma |
| Yes | 1.17 (1.11,1.23) | 1.29 (1.21,1.37) | 1.21 (1.13,1.29) |
| No | 1.07 (1.03,1.12) | 1.20 (1.14,1.26) | 1.13 (1.07,1.19) |
| Unclear | 0.99 (0.83,1.18) | 1.26 (1.02,1.55) | 1.18 (0.94,1.49) |
## WWI ‘s dose response and threshold effect on asthma prevalence
We further explored the relationship between the WWI index and asthma using a generalized additive model and smoothed curve fitting. In Figure 2; Table 4, we found a nonlinear relationship between WWI index and asthma. Based on a two-segment linear regression model, the inflection point for the WWI is 10.53. ( log-likelihood ratio test, $p \leq 0.001$).
**Figure 2:** *Density dose-response relationship between WWI index with asthma prevalence.*The area* between the upper and lower dashed lines is represented as $95\%$ CI. Each point shows the magnitude of the WWI index and is connected to form a continuous line.Adjusted for all covariates except effect modifier.* TABLE_PLACEHOLDER:Table 4
## Elevated WWI index may delay age of first asthma onset
Fully adjusted model 3 exhibited a 3.89 year delay in asthma onset for every 1-unit increase in the WW index(β=3.89, $95\%$ CI: 3.24, 4.54) (Table 5).
**Table 5**
| Characteristic | Model 1β (95%CI) | Model 2β (95%CI) | Model 3β (95%CI) |
| --- | --- | --- | --- |
| WWI Index | 7.42 (6.87,7.98) | 6.80 (6.22,7.39) | 3.89 (3.24,4.54) |
## WWI ‘s dose response and threshold effect on age of first asthma onset
According to our results (Figure 3), we found a positive linear correlation between the WWI index and first asthma onset.
**Figure 3:** *Density dose-response relationship between WWI index with onset age of prevalence.The area between the upper and lower dashed lines is represented as 95% CI. Each point shows the magnitude of the WWI index and is connected to form a continuous line.Adjusted for all covariates except effect modifier.*
## Discussion
This study demonstrated a positive association between the WWI index and asthma prevalence among US adults. The WWI index increased by 1 unit was associated with a $15\%$ increase in asthma prevalence. By using a generalized additive model and a smoothed curve fit, we were able to visualize the association between WWI and asthma prevalence clearly. Asthma prevalence and the WWI index did not have a linear correlation (Figure 2). The threshold and saturation effects showed that 10.53 was the most crucial inflection point.
Chronic diseases like asthma have placed increasingly huge burdens on health care costs, quality of life, and prevalence around the world [23]. To prevent asthma, it is critical to practice primary prevention, especially targeting populations that are adapted to the WWI index may be the most effective. Accordingly, we conducted a sensitivity subgroup analysis and found out that almost all populations showed a positive association with asthma prevalence. This was excluding Mexican Americans and groups with unknown blood relatives with asthma. The WWI index has a significant positive association with asthma prevalence, indicating that the WWI index is widely used among asthmatic patients. In light of previous studies, we believe that our findings are accurate. First of all, when grouped by age, we found that older people had a higher prevalence of asthma. There was an increased prevalence of asthma with age, with a greater trend among men than among women [24]. Asthma prevalence has also been found to be higher in older adults than in middle-aged adults, especially in men [25]. We also found that high WWI prevalence among men was associated with higher asthma prevalence among women. In contrast, hypertension [26], diabetes mellitus [27, 28] had significantly higher asthma prevalence.
In epidemiological studies on asthma, phenotypes can be categorized according to the age at onset, the duration of the disease,and the clinical features of the disease [29]. Asthma mortality and morbidity in older patients with asthma are higher than those in younger people [30]. Asthma that develops late in life is susceptible to misdiagnosis and improper treatment, which can have profound negative consequences for the health of the patient [31, 32]. It has also been shown in previous studies that abdominal obesity can lead to delayed-onset asthma, rather than early-onset asthma [33]. A significant finding of this study is the correlation between the WWI index and the age of the first asthma attack. As a result of our findings, 3.89 years will be added to the age of asthma onset for each unit increase in the WWI index. WWI and age of first asthma onset were positively correlated even after smoothing curve fitting. It also implies that there is an increase in late-onset asthma. There is no report on this promising finding yet. This result needs to be confirmed by a large multicentre prospective study to further confirm its accuracy.
Asthma and obesity have associated mechanisms that have yet to be fully elucidated. It is possible to suggest several plausible relationships: Furthermore, our study has several advantages. The NHANES 2001-2018 survey was conducted on a representative sample of the general U.S.population following a well-designed study protocol with extensive quality assurance and quality control procedures. Our results are reliable and can be applied to a broader range of individuals since they were adjusted for confounding covariates based on clinical understanding and previous studies. There are also limitations to our study. As a cross-sectional study, we were unable to establish a deterministic relation between the WWI index and asthma. Besides, the diagnosis of asthma was based on a questionnaire. Although previous studies have confirmed the acceptable accuracy of questionnaires [38, 39], recall bias remains. Furthermore, the database did not disclose detailed clinical variables such as medication history or asthma type classification, so further investigation is necessary. This study has many limitations, but its strength lies in its ability to reveal the relationship between the WWI index and asthma onset.
## Conclusion
This study showed an association between the modifiable risk factor WWI index and the prevalence of asthma and age at first asthma onset. A higher WWI index was associated with an increased prevalence of asthma and an earlier age of first asthma onset. Our findings suggest that weight control and a healthy lifestyle can reduce the occurrence of asthma, although the deterministic relation between the two cannot be clearly established, but is still of interest.
## 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.
## Ethics statement
The studies involving human participants were reviewed and approved by approved and reviewed by the National Center for Health Statistics Institutional Review Board(NCHS). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
LY and YC: Conceptualization, methodology, software. MX, RL, JZ and ZH: Data curation, writing original draft. MC and GW: Writing-review & editing. 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. Castillo JR, Peters SP, Busse WW. **Asthma exacerbations: Pathogenesis, prevention, and treatment**. *J Allergy Clin Immunol Pract* (2017) **5**. DOI: 10.1016/j.jaip.2017.05.001
2. Gomez-Llorente MA, Romero R, Chueca N, Martinez-Cañavate A, Gomez-Llorente C. **Obesity and asthma: A missing link**. *Int J Mol Sci* (2017) **18**. DOI: 10.3390/ijms18071490
3. Pate CA, Zahran HS, Qin X, Johnson C, Hummelman E, Malilay J. **Asthma surveillance - united states, 2006-2018**. *MMWR Surveill Summ* (2021) **70** 1-32. DOI: 10.15585/mmwr.ss7005a1
4. Moorman JE, Akinbami LJ, Bailey CM, Zahran HS, King ME, Johnson CA. **National surveillance of asthma: United states, 2001-2010**. *Vital Health Stat* (2012) **3** 1-58
5. Yang G, Han YY, Forno E, Yan Q, Rosser F, Chen W. **Glycated hemoglobin A(1c), lung function, and hospitalizations among adults with asthma**. *J Allergy Clin Immunol Pract* (2020) **8** 3409-15.e1. DOI: 10.1016/j.jaip.2020.06.017
6. Wang H, Bai C, Yi M, Jia Y, Li Y, Jiang D. **Metabolic syndrome and incident asthma in Chinese adults: An open cohort study**. *Diabetes Metab Syndr Obes* (2020) **13**. DOI: 10.2147/DMSO.S274159
7. Beasley R, Semprini A, Mitchell EA. **Risk factors for asthma: is prevention possible**. *Lancet* (2015) **386**. DOI: 10.1016/S0140-6736(15)00156-7
8. Lieberoth S, Backer V, Kyvik KO, Skadhauge LR, Tolstrup JS, Grønbæk M. **Intake of alcohol and risk of adult-onset asthma**. *Respir Med* (2012) **106**. DOI: 10.1016/j.rmed.2011.11.004
9. Beuther DA, Sutherland ER. **Overweight, obesity, and incident asthma: a meta-analysis of prospective epidemiologic studies**. *Am J Respir Crit Care Med* (2007) **175**. DOI: 10.1164/rccm.200611-1717OC
10. Alwarith J, Kahleova H, Crosby L, Brooks A, Brandon L, Levin SM. **The role of nutrition in asthma prevention and treatment**. *Nutr Rev* (2020) **78**. DOI: 10.1093/nutrit/nuaa005
11. Qin Z, Chang K, Yang Q, Yu Q, Liao R, Su B. **The association between weight-adjusted-waist index and increased urinary albumin excretion in adults: A population-based study**. *Front Nutr* (2022) **9**. DOI: 10.3389/fnut.2022.941926
12. Thomas EL, Frost G, Taylor-Robinson SD, Bell JD. **Excess body fat in obese and normal-weight subjects**. *Nutr Res Rev* (2012) **25**. DOI: 10.1017/S0954422412000054
13. Koenen M, Hill MA, Cohen P, Sowers JR. **Obesity, adipose tissue and vascular dysfunction**. *Circ Res* (2021) **128**. DOI: 10.1161/CIRCRESAHA.121.318093
14. Stefan N. **Causes, consequences, and treatment of metabolically unhealthy fat distribution**. *Lancet Diabetes Endocrinol* (2020) **8**. DOI: 10.1016/S2213-8587(20)30110-8
15. Sorimachi H, Obokata M, Takahashi N, Reddy Y, Jain CC, Verbrugge FH. **Pathophysiologic importance of visceral adipose tissue in women with heart failure and preserved ejection fraction**. *Eur Heart J* (2021) **42**. DOI: 10.1093/eurheartj/ehaa823
16. Andreoli A, Garaci F, Cafarelli FP, Guglielmi G. **Body composition in clinical practice**. *Eur J Radiol* (2016) **85**. DOI: 10.1016/j.ejrad.2016.02.005
17. Park Y, Kim NH, Kwon TY, Kim SG. **A novel adiposity index as an integrated predictor of cardiometabolic disease morbidity and mortality**. *Sci Rep* (2018) **8** 16753. DOI: 10.1038/s41598-018-35073-4
18. Kim NH, Park Y, Kim NH, Kim SG. **Weight-adjusted waist index reflects fat and muscle mass in the opposite direction in older adults**. *Age Ageing* (2021) **50**. DOI: 10.1093/ageing/afaa208
19. Li Q, Qie R, Qin P, Zhang D, Guo C, Zhou Q. **Association of weight-adjusted-waist index with incident hypertension: The rural Chinese cohort study**. *Nutr Metab Cardiovasc Dis* (2020) **30**. DOI: 10.1016/j.numecd.2020.05.033
20. Ding C, Shi Y, Li J, Li M, Hu L, Rao J. **Association of weight-adjusted-waist index with all-cause and cardiovascular mortality in China: A prospective cohort study**. *Nutr Metab Cardiovasc Dis* (2022) **32**. DOI: 10.1016/j.numecd.2022.01.033
21. Zhao P, Shi W, Shi Y, Xiong Y, Ding C, Song X. **Positive association between weight-adjusted-waist index and hyperuricemia in patients with hypertension: The China h-type hypertension registry study**. *Front Endocrinol (Lausanne)* (2022) **13**. DOI: 10.3389/fendo.2022.1007557
22. 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**. *Int J Surg* (2014) **12**. DOI: 10.1016/j.ijsu.2014.07.013
23. Toskala E, Kennedy DW. **Asthma risk factors**. *Int Forum Allergy Rhinol* (2015) **5**. DOI: 10.1002/alr.21557
24. Park S, Jung SY, Kwon JW. **Sex differences in the association between asthma incidence and modifiable risk factors in Korean middle-aged and older adults: NHIS-HEALS 10-year cohort**. *BMC Pulm Med* (2019) **19** 248. DOI: 10.1186/s12890-019-1023-3
25. Gershon AS, Guan J, Wang C, To T. **Trends in asthma prevalence and incidence in Ontario, Canada, 1996-2005: A population study**. *Am J Epidemiol* (2010) **172**. DOI: 10.1093/aje/kwq189
26. Jaddi NS, Abdullah S. **Kidney-inspired algorithm with reduced functionality treatment for classification and time series prediction**. *PloS One* (2019) **14**. DOI: 10.1371/journal.pone.0208308
27. Lee KH, Lee HS. **Hypertension and diabetes mellitus as risk factors for asthma in Korean adults: the sixth Korea national health and nutrition examination survey**. *Int Health* (2020) **12**. DOI: 10.1093/inthealth/ihz067
28. Yun HD, Knoebel E, Fenta Y, Gabriel SE, Leibson CL, Loftus EV. **Asthma and proinflammatory conditions: A population-based retrospective matched cohort study**. *Mayo Clin Proc* (2012) **87**. DOI: 10.1016/j.mayocp.2012.05.020
29. Bel EH. **Clinical phenotypes of asthma**. *Curr Opin Pulm Med* (2004) **10** 44-50. DOI: 10.1097/00063198-200401000-00008
30. Gibson PG, McDonald VM, Marks GB. **Asthma in older adults**. *Lancet* (2010) **376**. DOI: 10.1016/S0140-6736(10)61087-2
31. Dunn RM, Busse PJ, Wechsler ME. **Asthma in the elderly and late-onset adult asthma**. *Allergy* (2018) **73**. DOI: 10.1111/all.13258
32. Kitch BT, Levy BD, Fanta CH. **Late onset asthma: epidemiology, diagnosis and treatment**. *Drugs Aging* (2000) **17**. DOI: 10.2165/00002512-200017050-00005
33. Leone N, Courbon D, Berr C, Barberger-Gateau P, Tzourio C, Alpérovitch A. **Abdominal obesity and late-onset asthma: cross-sectional and longitudinal results: the 3C study**. *Obes (Silver Spring)* (2012) **20**. DOI: 10.1038/oby.2011.308
34. Kim S, Yi HA, Won KS, Lee JS, Kim HW. **Association between visceral adipose tissue metabolism and alzheimer's disease pathology**. *Metabolites* (2022) **12**. DOI: 10.3390/metabo12030258
35. Sood A, Ford ES, Camargo CA. **Association between leptin and asthma in adults**. *Thorax* (2006) **61**. DOI: 10.1136/thx.2004.031468
36. Sood A, Cui X, Qualls C, Beckett WS, Gross MD, Steffes MW. **Association between asthma and serum adiponectin concentration in women**. *Thorax* (2008) **63**. DOI: 10.1136/thx.2007.090803
37. Shore SA. **Obesity and asthma: possible mechanisms**. *J Allergy Clin Immunol* (2008) **121**. DOI: 10.1016/j.jaci.2008.03.004
38. Torén K, Brisman J, Järvholm B. **Asthma and asthma-like symptoms in adults assessed by questionnaires. a literature review**. *Chest* (1993) **104**. DOI: 10.1378/chest.104.2.600
39. Leikauf J, Federman AD. **Comparisons of self-reported and chart-identified chronic diseases in inner-city seniors**. *J Am Geriatr Soc* (2009) **57**. DOI: 10.1111/j.1532-5415.2009.02313.x
|
---
title: Is the excellent air quality a protective factor of health problems for Taitung
County in eastern Taiwan? Perspectives from visual analytics
authors:
- Ching-Hsiang Yu
- Shun-Chuan Chang
- En-Chih Liao
journal: Heliyon
year: 2023
pmcid: PMC9988568
doi: 10.1016/j.heliyon.2023.e13866
license: CC BY 4.0
---
# Is the excellent air quality a protective factor of health problems for Taitung County in eastern Taiwan? Perspectives from visual analytics
## Abstract
Taitung, an agricultural country in Eastern Taiwan, was famous for its fresh air with less industrial and petrochemical pollution. Air pollution may induce cardiovascular disease, chronic obstructive pulmonary disease (COPD), asthma, and stroke, poor air quality also resulted in a higher depression rate and less feeling of happiness; therefore, our study aims to use visualization tools to demonstrate the association between air quality index (AQI) and the among negative factors and try to find that whether Taitung got the benefit of good air quality on health issues. We retrieved data from the government of Taiwan and other open sources in the year 2019, then visual maps and generalized association plots with clusters demonstrated the relationship between each factor and each county/city. Taitung had the lowest AQI and asthma attack rate, but AQI had a negative relationship to air pollution-caused death (R = −0.379), happiness index (R = −0.358), and income (R = −0.251). The GAP analysis revealed that smoke and overweight were the nearest to air pollution causing death, also counties and cities were divided into two major clusters initially based on the air pollution-related variables. In conclusion, the World Health Organization (WHO) definition and the weight of each air pollution cause death may not be suitable for Taiwan due to too many confounding factors.
## Introduction
Air quality index (AQI) is an indicator of air pollution, according to the Environmental Protection Administration (EPA) of Taiwan, the concentration of ozone (O3), suspended fine particles of Particulate Matter≦2.5 μm (PM2.5), suspended particles of PM10, carbon monoxide (CO), sulfur dioxide (SO2) and nitrogen dioxide (NO2) in the same day will be converted into many sub-indicators based on their impact on the human body, then the maximum of sub-indicator will become the AQI of that day [1].
Taitung *County is* located in the southeast of Taiwan, its industrial structure is mainly based on the farming industry, which resulted in less industrial air pollution and better air quality, thus these natural advantages of the environment were also featured for the promotion of the tourism industry in recent years [2]. As to the assessment of air quality by the residents, Taitung and Hualien (another rural county in eastern Taiwan) Counties belong to the "excellent" class [3], the data from Environmental Protection Administration also revealed that Taitung County had the lowest AQI in Taiwan, which caused much better visibility than the highly urbanized center (Taipei City, in northern Taiwan) and highly industrialized center (Kaohsiung City, in southwestern Taiwan) [4].
Much evidence supported that air pollution had multiple adverse effects on the human body, World Health Organization (WHO) pointed out that air pollution was estimated to cause about $29\%$ of lung cancer deaths, $43\%$ of chronic obstructive pulmonary disease (COPD) deaths, about $25\%$ of ischemic heart disease deaths and $24\%$ of stroke deaths [5]. Air pollution may exacerbate and even develop asthma, UK's Committee on the Medical Effects of Air Pollutants proposed four main mechanisms: oxidative stress and damage, airway remodeling, inflammatory pathways, and immunological responses, and enhancement of respiratory sensitization to aeroallergens [6,7]. Research in Taiwan also showed the same finding, a study based on the National Health Insurance database revealed that high levels of NOX (NO and NO2) resulted in significant odds ratios of asthma exacerbation up to 1.45 in preschool children [8], another mass screening study among all middle school students demonstrated that students who lived in heavy air pollution areas were 1.8 times more likely to have a history of asthma than who lived in an area without pollution [9].
Happiness can be influenced by air pollution, when the concentration of NOX increased by 1 μg/m3, Chinese residents thought that their possibility of happiness would decrease by $0.034\%$ [10], and a Spanish study found that people would like to pay $1.4\%$ of their income for reducing $1\%$ of air pollution [11]. Moreover, air pollutants and ambient air pollution levels were positively linked to suicide mortality and risk which was supported by studies in Japan, China, and Belgium [[12], [13], [14]].
Therefore, our study aims to evaluate whether good air quality provided health benefits to Taitung people or not and investigate the association between AQI and some bio-psycho-social indicators.
## Data collection
The data of AQI and PM2.5 were downloaded from Taiwan Air Quality Monitoring Network [1], which was preserved by the Environmental Protection Administration, Executive Yuan. We used the average AQI and PM2.5 of each county or city in 2019 for the study.
The data on causes of death were obtained from the Department of Statistics, Ministry of Health and Welfare [15]. The standardized mortality ratio of lung cancer deaths, COPD deaths, ischemic heart disease deaths, and stroke deaths of each county or city was multiplied by the weights of each disease ($29\%$, $43\%$, $25\%$, $24\%$) based on WHO online database "Burden of disease from joint household and ambient Air Pollution” and summed together.
The data on asthma attack incidence rate was retrieved from the Quality Performance of National Health Insurance Disclosure Website. It was a rate of patients involved in the Improvement Plan of Medical Benefit of National Health Insurance (NHI) for Asthma suffering from asthma attacks then being sent to the emergency department [16]. The rates of the four seasons in 2019 were averaged and the means were used for further analysis.
The data on the happiness index of 2019 was gotten from Economy Daily News (Taiwan), this index is based on the structure of the Organization for Economic Co-operation and Development (OECD) Your Better Life Index [17], consisted of two major parts, objective well-being, and subjective well-being, the first one was calculated by governmental statistics, the other was analyzed by public opinion poll [18].
The demographic data including income, smoking rate, and prevalence of overweight was accessed from the Department of Statistics, Ministry of Health and Welfare (Department of Statistics). The percentage of residents who received higher education was downloaded from the Department of Household Registration, Ministry of the Interior [19].
Excel Office 365 (Microsoft, Redmond, WA, USA) with a map template was administered for drawing the visual map of each indicator, the green-yellow-red spectrum was used for filling in color with its corresponding value [20]. Another online map generator (https://pixelmap.amcharts.com/#) was applied for the demonstration of the overview of the 20 counties/cities and their grouping.
## Statistical analysis
SPSS version 24.0 (IBM, Armonk, NY, USA) was used for calculating the Pearson’s correlation coefficient between AQI and each indicator, then we should use the absolute criterion for correlations as the following [21]: 0–0.19: no correlation, 0.2–0.39: low correlation, 0.40–0.59: moderate correlation, 0.60–0.79: moderately high, ≥0.80: high correlation.
The student’s t-test was chosen for p-value under the statistical guidelines [22,23], α level of 0.05 was set for all statistical tests [24], and all indicators were standardized for generalized association plots.
## Generalized association plots (GAP)
Generalized association plots (GAP) was a Java-designed software developed by Chun-Houh Chen from the Institute of Statistical Science Academia Sinica [25], white-black spectrum was used in the raw data matrix, blue-white-red bidirectional color spectrum was chosen for displaying the correlation coefficient in proximity matrices for columns (9 health related-factors) and rows (20 cities and counties), then the average-linkage method was used for hierarchical clustering.
## Subsection
In Fig. 1A, the greener area had the better air quality, so the air quality of Yilan County, Hualien County, and Taitung County (from the northeastern to southeastern coast area) was good, and the region with the poorest air quality was the southwestern area, especially Kaohsiung City. Fig. 1The visual maps of (A) air pollution, (B) air pollution caused death, (C) asthma attack, (D) happiness, and (E) income. AQI ranged from 35 to 75, the SMR of air pollution caused death per million people ranged from 25 to 55, the asthma attack rate ranged from $0\%$ to $1\%$, the happiness index ranged from 25 to 75, and the annual income per person ranged from 250 thousand to 550 thousand of New Taiwan Dollar (30 NTD ≒ 1 USD). The (A), (B), and (C) were negative indicators. The redder was the worse and the greener was the better. The (D) and (E) were positive indicators, so the redder became the better. Fig. 1 The redder area in Fig. 1B implied a higher standardized mortality ratio of air pollution caused death, and Taitung had the highest value, also the other two counties on the eastern coast (Yilan County and Hualien County) were in yellow to red. On the contrary, Kaohsiung City which was in red on the AQI map was filled with light green in the visual map of air pollution caused death, the statistical result also revealed that AQI and air pollution caused death had a low negative correlation (R = −0.379).
Fig. 1C is the visual map of the asthma attack rate, which showed that Taitung County was in dark green, and Pingtung County was in dark red, then AQI and asthma attack had a low positive correlation ($R = 0.263$). The visual map of the happiness index was demonstrated in Fig. 1D, the yellow color in the three east coast counties corresponded to the moderate level of the happiness index, and Taipei City was the most well-being area which was in dark red. AQI and happiness index had a low negative correlation (R = −0.358).
We found that the richest counties/cities were in northern Taiwan which were in orange to red, and the three east coast counties had the lower income, so they were in the green (Fig. 1E). AQI and income had a low negative correlation (R = −0.251).
Although there was no significant correlation between AQI and the above four indicators, interestingly, happiness index and income had a moderate positive correlation ($R = 0.595$).
## The association of air health-related factors
Fig. 2A showed that three clusters were divided, cluster 1 was separated initially comprising happiness index, higher education, and income, cluster 2 was composed of smoking, overweight, and air pollution caused death, cluster 3 included AQI, PM2.5, and asthma attack rate. Based on their composition, cluster 1 was named “positive indicators”, cluster 2 was named “negative indicators”, and cluster 3 was named “air pollution indicators”. Fig. 2The proximity matrix of health-related indicators. When the cut point was set as a correlation coefficient (R) around 0.1, the 9 indicators could be divided into 3 clusters. ( A). The raw Indicator × Indicator matrix; (B). The proximity matrix after partition by cut point and sufficient with the means of each square. Happiness = Happiness index, Education = percentage of people receiving a college education, Income = the annual income per person, Smoking = smoking rate, Overweight = overweight rate, Death = SMR of air pollution caused death, AQI = air quality index, Asthma = asthma attack rate. Fig. 2 To demonstrate the relationship between each cluster clearly, the matrix under-went partition and sufficient with the means of each square then became Fig. 2A. The color square between “positive indicators” and “negative indicators” was dark blue, indicating they had a moderately to highly positive association, and the color square between “positive indicators” and “air pollution indicators”, “negative indicators” and “air pollution indicators” were light blue, implying their associations were weak (Fig. 2B).
After hierarchical clustering, four groups were successfully divided as the following (Table 1):Table 1The characteristics of the four groups of counties/cities. Table 1Group/ClusterPositive indicatorsNegative indicatorsAir pollution-related indicatorsGroup 1LowHighHighGroup 2ModerateHighLowGroup 3Moderate to highLowHighGroup 4HighLowLow to moderate Group 1: Nantou County, Yunlin County, Changhua County, Pingtung County, Chiayi County.
Group 2: Keelung City, Penghu County, Miaoli County, Hualien County, Taitung County, Yilan County.
Group 3: Tainan City, Chiayi City, Kaohsiung City, Taichung City.
Group 4: Hsinchu County, Taoyuan City, Hsinchu City, Taipei City, New Taipei City.
According to the raw data matrix (Indicators × Counties/cities, in the white-black spectrum), Group 1 had a low level of “positive indicators”, a high level of “negative indicators” and “air pollution indicators”; Group 2 had a moderate level of “positive indicators”, high level of “negative indicators”, and low level of “air pollution indicators”; Group 3 had a moderate to the high level of “positive indicators”, low level of “negative indicators”, and high level of “air pollution indicators”; Group 4 had a high level of “positive indicators”, low level of “negative indicators”, and low to moderate level of “air pollution indicators”.
The proximity matrix of row (Indicators × Indicators, at the ride side in the blue-white-red spectrum) also underwent partition and sufficient as Fig. 4, Fig. 5, then the mildly negative correlation between Group 1 & 2, Group 2 & 4, also the highly negative correlation between Group 2 & 3 were demonstrated. Moreover, there was a mildly to the moderately positive correlation between Groups 1 & 3, and Groups 3 & 4.
## Finding the potential confounding factors of air pollution-caused death
Because the correlation between AQI and air pollution caused death was negative, implying that there should be some confounding factors that disturbed the relationship. Four candidates including smoking, overweight, income, and higher education were chosen, the first two were risk factors and the last two seemed to be protective factors. Table 2 showed that both smoking and overweight had moderately to highly associations ($R = 0.70$ and 0.75) to air pollution caused death with significant p-value (0.001 and < 0.001), income and higher education had significantly moderately to highly associations (R = −0.64 and −0.80, p-value = 0.002 and < 0.001) in contrast. Table 2The associations of air pollution-caused death and potential confounding factors. Table 2Potential confounding factorsCorrelation coefficient (R)P-valueSmoking0.700.001*Overweight0.75<0.001*Income−0.640.002*Higher education−0.80<0.001**P-value <0.05.
## Discussion
The difference in air quality among the 20 counties/cities was attributable to two reasons: natural factors and anthropogenic factors. In Fig. 1A, the green predominant area and the red predominant area was almost separated by the Central Mountain Range, which is an important geographic boundary of climate because the direction of the prevailing wind and the path of the weather system had significant intersection angle, caused the different regional climate in the same season [26], also a seasonal difference on the concentration of PM2.5 and PM10 was observed in a 9 years study, the highest mean PM2.5 particle concentration was detected during spring at the Banqiao (in New Taipei City), Hualien, and Taitung air quality monitoring stations [27], which may be exacerbated by the air pollution from *China via* northeast monsoon [28], and there was evidence showing that in spring, sulfate aerosols from remote sources were predominant in another study [29].
The major causes of air pollution in Taiwan were industrial exhaust ($27.5\%$) and automobile emission ($27.5\%$) according to the news of EPA, and the corresponding policies including the cap-and-trade program [30], allowance of replacing the old vehicle with new or electric one has been implemented [31]. Although some economic considerations blocking the enforcement was another obstacle that needed to be overcome [32], the reduction trend from 2005 to 2015, despite the vehicle numbers and energy consumption, industrial output, was similar to those of developed countries [33], implying that the air pollution policy was effective.
The incompatible result of AQI and air pollution caused death should be owing to the many confounding factors which played important parts in “air pollution caused death”, including smoking, being overweight, and higher education. A previous study in Taiwan found that not only fine particulate air pollution level, but also household income, physician density, high school graduate rate, smoking rate, and blue-collar worker percentage had a positive or negative effect on adult life expectancy [34], we needed a novel method to estimate the real air pollution caused death in Taiwan, the weight of each related disease must be adjusted based on further study to match the condition of Taiwan. Furthermore, this result also reminded us that we need to “treat underlying diseases” of “air pollution-caused death” if we want to reduce the mortality rate.
For the higher smoking rate in Taitung, although one of the most effective ways to reduce tobacco consumption was increasing tobacco taxes [[35], [36], [37], [38]], strangely, a 6-year-study in Taiwan revealed that there was no significantly quitting or reducing consumption of tobacco in most smokers because of non-tax-induced price increases, and some of them would even switching to cheaper brands within the same tobacco company [39], so other strategies different from increasing price should be used, such as a mobile application which was developed by Taiwanese researchers, and made use of a combination of I-Change behavioral change model with health recommender system and computer-tailoring for smoking cessation [40]. Moreover, some factors correlated with the success rate of smoking cessation should be early detected, including nicotine dependence level (measured by the FTCD), exhaled air carbon monoxide concentration, and cigarettes smoked per day which was proven by a Taiwanese study [41].
Overweight and even obesity have become an unneglectable health burden of Taitung for more than ten years, Taitung had the top-ranked adult obesity rate ($48.1\%$) in Taiwan in 2011, and for obesity prevention, a series of strategies followed by Ottawa Charters Guideline were instituted from 2011, then decreased obesity rate ($46.9\%$) with increased exercise rate ($67.7\%$) were found at the end of 2012 (" [42], but the prevalence of overweight in Taitung was still the highest in the data obtained in 2019, therefore, the policies for weight control in Taitung could not only follow up Taiwan's Obesity Prevention and Management Strategy which was published by Health Promotion Administration [43], they should be customized based on the characteristics of Taitung, including the higher percentage of the aboriginal population (up to $30\%$) [44], the low-er-income and education accessibility, also the poor public transport [45].
Taitung County had no university until National Taitung University upgraded from National Taitung Teachers’ College in 2003, which was under the policy of “One county, (at least) one university” by President Shui-Bian Chen [46], indicating that the resource of higher education was exile, even though few studies investigated this issues, but some researchers evaluated the percentage of aboriginal students receiving the college education, the highest was in Taipei City ($24.55\%$), and the lowest was in Taitung County ($9.25\%$) which is almost the one-fourth of the means, and the drop-out rate even higher than the graduated rate, especially in Taitung County [47], then may cause that indigenous graduates tend to have lower salary and lower skill requirement jobs [48], and that should be the predisposing factor of health problems.
The rural-urban disparity was also demonstrated in Fig. 3, Fig. 5, Group 1 and 2 contained all counties except Hsinchu County, and Groups 3 and 4 contained all cities except Keelung City. Cities had a higher level of “positive indicators” and a lower level of “negative indicators”, and then counties had a lower level of “positive indicators” and a higher level of “negative indicators” overall. Fig. 3The combined matrices with clustering of health-related indicators and counties/cities. This figure consisted of three parts: the upper one was the proximity matrix of health-related indicators with the clustering tree, which was the same as Fig. 2. The left lower one was the raw data matrix, its columns were health-related indicators, and its rows were 20 counties/cities of Taiwan, all the data had been standardized, and the values lower than −0.5 were painted with the same color of −0.5, also the values upper than 0.5 were painted with the same color of 0.5 to demonstrate more clearly. The right lower one was the proximity matrix of 20 counties/cities of Taiwan with the clustering tree, and four groups were marked with different colors which would be used in Fig. 5 (blue, yellow, red, green).Fig. 3Fig. 4The combined matrices with clustering of health-related indicators and counties/cities, with proximity matrices partition and sufficient. Fig. 4 was the same matrix as Fig. 3 but underwent partition and sufficient with the means of each square to demonstrate the relationships between each group or cluster. Fig. 4Fig. 5The distribution of the four groups of counties/cities on the map of Taiwan. Groups 1 & 3 could correspond to the yellow to red area (moderate to poor air quality) of the AQI visual map, while Groups 2 & 4 correspond to the area with good air quality (green of AQI map).Fig. 5 We found that Group 3 is special, it contained Tainan City, Kaohsiung City, and Taichung City, which were separated into three cities and three counties (Tainan City and County, Kaohsiung City and County, Taichung City and County) before “City-county Consolidation＂ on $\frac{2010}{12}$/25 [49], so they had mixed characteristics of city and county, therefore, Group 3 had a moderately positive correlation to Group 4 (cities in the northern region) and mildly to moderately positive correlation to Group 1 (counties in the western region).
There are some limitations in this study, the major one is that we only evaluated the data in one year, and the time effect should also be considered. Perhaps a multi-year analysis could help us see more trends and changes in the relationship between air pollution and health problems. We chose 2019 to do this study because the medical resources and economy were relatively sufficient and stable compared to the years after the COVID-19 outbreak, perhaps the rural-urban disparity would be enlarged and the effect of air pollution on health would be decreased due to the new factor of a pandemic, which needs further studies to prove this hypothesis.
Moreover, some studies involved all the major air pollutants in their model, such as Lee et al. ’s study presented on J Allergy Clin Immunol. in 2019, using a formula to combine these factors into a new value, then using this value to build their model ([50]. We did not involve all the air pollutants in our study because there’s no previous Taiwanese study that built the multivariable distributed lag nonlinear model for pollutants. Perhaps we can conduct another study by using the data from Taiwan to investigate whether our data can fit this model and whether the same effect also exists in Taiwan or not.
The reason why we involved PM2.5 in our generalized association plots is that a previous study showed PM2.5 is the most frequent primary air pollutant in Taiwan ([51], and our result also demonstrated that the Pearson’s correlation coefficient was more than 0.77 between AQI and PM2.5. Moreover, AQI is easy to understand and has been used in many studies to evaluate the relationship between health problems [52], indicating that AQI is representative enough for air quality.
In conclusion, this study initiated from the relationship between air quality and health problems in Taitung, and due to the incompatible result of AQI and air pollution caused death, we found that many confounding factors affected the relationship and may mask the protective effect of good air quality, then the rural-urban disparity was revealed by using some bio-psycho-social indicators via generalized association plots.
## Author contribution statement
Ching-Hsiang Yu: Performed the experiments; Analyzed and interpreted the data; Wrote the paper.
Shun-Chuan Chang: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.
En-Chih Liao: Conceived and designed the experiments; Wrote the paper.
## Funding statement
Professor En-Chih Liao was supported by $\frac{10.13039}{501100014804}$Mackay Medical College [MMC-RD-110-1E-P004 & MMC-RD-111-1B-P018].
## Data availability statement
Data will be made available on request.
## Declaration of interest’s statement
The authors declare no competing interests.
## Ethics approval and consent to participate
Not applicable.
## Additional information
No additional information is available for this paper.
## Availability of data and materials
The data of AQI and PM2.5 were downloaded from Taiwan Air Quality Monitoring Network. https://airtw.epa.gov.tw/CHT/Query/His_Data.aspx.
The demographic data including income, smoking rate, and prevalence of overweight, causes of death were obtained from the Department of Statistics, Ministry of Health and Welfare. https://dep.mohw.gov.tw/DOS/lp-5069-113-xCat-y108.html.
The data on asthma attack incidence rate was retrieved from the Quality Performance of National Health Insurance Disclosure Website.
https://www.nhi.gov.tw/mqinfo/Map_1.aspx?Type=Asthma&DAID=1740&List=4.
The data on the happiness index of 2019 was gotten from Economy Daily News (Taiwan).
https://money.udn.com/ACT/2019/happy/chart/2019.pdf.
The percentage of residents who received higher education was downloaded from the Department of Household Registration, Ministry of the Interior.
https://www.ris.gov.tw/app/portal/346.
All the datasets in this study are available as references in the section of “Materials and Methods” and are also available from the corresponding author upon reasonable request.
## References
1. 1Environmental Protection AdministrationTaiwan Air Quality Monitoring Network2021https://airtw.epa.gov.tw/ENG/Information/Standard/AirQualityIndicator.aspx. (2021.0)
2. Su Y.-J., Shih M.-L.. **A study on traveler satisfaction with recreational attractions of bicycle trails—a case study of Forest Park, Taitung City**. *J. Stat. Manag. Syst.* (2009.0) **12** 1041-1046
3. Hsueh Y.-C., Cheng J.-H., Hsu C.-P.. **A study on the commonage's perception toward the implementation of improving air quality policies by local government**. *Rep. Geograph. Stud.* (2000.0) 1-31
4. Tsai Y.I., Kuo S.-C., Lee W.-J., Chen C.-L., Chen P.-T.. **Long-term visibility trends in one highly urbanized, one highly industrialized, and two Rural areas of Taiwan**. *Sci. Total Environ.* (2007.0) **382** 324-341. DOI: 10.1016/j.scitotenv.2007.04.048
5. 5World Health OrganizationGlobal Health Observatory Data: Ambient Air Pollution2021World Health Organizationhttps://www.who.int/data/gho/data/themes/topics/indicator-groups/indicator-group-details/GHO/ambient-air-pollution. (2021.0)
6. Gowers A.M., Cullinan P., Ayres J.G., Anderson H.R., Strachan D.P., Holgate S.T., Maynard R.L.. **Does outdoor air pollution induce new cases of asthma? Biological plausibility and evidence; a review**. *Respirology* (2012.0) **17** 887-898. PMID: 22672711
7. Guarnieri M., Balmes J.R.. **Outdoor air pollution and asthma**. *Lancet* (2014.0) **383** 1581-1592. DOI: 10.1016/S0140-6736(14)60617-6
8. Chung H.-Y., Chung H.-Y., Tseng C.-C., Yiin L.-M.. **Associations between outdoor air pollutants and first occurrence of asthma in pre-school children, 2007-2011**. *Chin. J. Publ. Health* (2016.0) **35** 199-208. DOI: 10.6288/TJPH201635104038
9. Wu J.-H., Lin R.S., Hsieh K.-H., Chiu W.-T., Chiu L.-M., Chiou S.-T., Fan S.-H.. **Adolescent asthma in northern Taiwan**. *Chin. J. Public Health* (1998.0) **17** 214-225
10. Yang J.-D., Zhang Y.-R.. **The price of air pollution: an analysis based on happiness data**. *World Econom. Polit.* (2014.0) **12** 162-188
11. Cuñado J., De Gracia F.P.. **Environment and happiness: new evidence for Spain**. *Soc. Indicat. Res.* (2013.0) **112** 549-567
12. Casas L., Cox B., Bauwelinck M., Nemery B., Deboosere P., Nawrot T.S.. **Does air pollution trigger suicide? A case-crossover analysis of suicide deaths over the life span**. *Eur. J. Epidemiol.* (2017.0) **32** 973-981. PMID: 28623424
13. Lin G.-Z., Li L., Song Y.-F., Zhou Y.-X., Shen S.-Q., Ou C.-Q.. **The impact of ambient air pollution on suicide mortality: a case-crossover study in Guangzhou, China**. *Environ. Health* (2016.0) **15** 1-8. PMID: 26739281
14. Ng C.F.S., Stickley A., Konishi S., Watanabe C.. **Ambient air pollution and suicide in Tokyo, 2001–2011**. *J. Affect. Disord.* (2016.0) **201** 194-202. PMID: 27240312
15. 15Department of Statistics. Statistics ＆ Publications. Ministry of Health and Welfare. https://dep.mohw.gov.tw/DOS/np-1714-113.html.
16. 16National Health Insurance Administration. Quality Performance of National Health Insurance Disclosure Website. Ministry of Health and Welfare. https://www.nhi.gov.tw/AmountInfoWeb/index.html.
17. Mizobuchi H.. **Measuring world better life frontier: a composite indicator for OECD better life index**. *Soc. Indicat. Res.* (2014.0) **118** 987-1007
18. 18Economy Daily News2020 Happiness Index of Counties and Cities, Manual of Happiness2020. (2020.0)
19. 19Department of Household Registration. Demographic Statistics Database. Ministry of the Interior. https://www.ris.gov.tw/app/portal/674.
20. Liu W.-X.. (2017.0)
21. Zhu W.. **p< 0.05,< 0.01,< 0.001,< 0.0001,< 0.00001,< 0.000001, or< 0.0000001…**. *J. Sport Heal. Sci.* (2016.0) **5** 77
22. Lee S.W.. **Methods for testing statistical differences between groups in medical research: statistical standard and guideline of Life Cycle Committee**. *Life Cycle* (2022.0) **2**
23. Lee S.W.. **Regression analysis for continuous independent variables in medical research: statistical standard and guideline of Life Cycle Committee**. *Life Cycle* (2022.0) **2**
24. Di Leo G., Sardanelli F.. **Statistical significance: p value, 0.05 threshold, and applications to radiomics—reasons for a conservative approach**. *Euro. Radiol. Experiment.* (2020.0) **4** 1-8
25. Wu H.-M., Tien Y.-J., Chen C.-h.. **GAP: a graphical environment for matrix visualization and cluster analysis**. *Comput. Stat. Data Anal.* (2010.0) **54** 767-778
26. Wu M.-C., Chen Y.-L.. **The classification of the climate in Taiwan**. *Atmospheric Sci.* (1993.0) **21** 55-66
27. Fang G.-C., Chang S.-C.. **Atmospheric particulate (PM10 and PM2.5) mass concentration and seasonal variation study in the Taiwan area during 2000–2008**. *Atmos. Res.* (2010.0) **98** 368-377. DOI: 10.1016/j.atmosres.2010.07.005
28. Shie R.-H., Lee J.-H., Chan C.-C.. **Temporal-spatial distribution of PM_(2.5) concentration and the contribution of sources in Taiwan [Temporal-Spatial distribution of PM_(2.5) concentration and the contribution of sources in Taiwan]**. *Formosan J. Med.* (2016.0) **20** 367-376. DOI: 10.6320/fjm.2016.20(4).4
29. Kishcha P., Wang S.-H., Lin N.-H., da Silva A., Lin T.-H., Lin P.-H., Alpert P.. **Differentiating between local and remote pollution over Taiwan**. *Aerosol Air Qual. Res.* (2018.0) **18** 1788-1798. DOI: 10.4209/aaqr.2017.10.0378
30. Shaw D., Hung M.-F.. **Evolution and evaluation of air pollution control policy in Taiwan**. *Environ. Econ. Pol. Stud.* (2001.0) **4** 141-166
31. Kao J.-Y.. (2019.0)
32. Yang T.-M., Yeh K.-P.. **The air quality of Taiwan and the practice of Total mass based control of air pollution control act in 2013**. *Taiwan Acad. Ecol.* (2014.0) 9-15
33. Lee C.-S., Chang K.-H., Kim H.. **Long-term (2005–2015) trend analysis of PM2.5 precursor gas NO2 and SO2 concentrations in Taiwan**. *Environ. Sci. Pollut. Control Ser.* (2018.0) **25** 22136-22152. DOI: 10.1007/s11356-018-2273-y
34. Chen C.-C., Chen P.-S., Yang C.-Y.. **Relationship between fine particulate air pollution exposure and human adult life expectancy in Taiwan**. *J. Toxicol. Environ. Health, Part A* (2019.0) **82** 826-832. DOI: 10.1080/15287394.2019.1658386
35. Amato M.S., Boyle R.G., Brock B.. **Higher price, fewer packs: evaluating a tobacco tax increase with cigarette sales data**. *Am. J. Publ. Health* (2015.0) **105** e5-e8
36. Bank W.. (1999.0)
37. Chaloupka F.J., Yurekli A., Fong G.T.. **Tobacco taxes as a tobacco control strategy**. *Tobac. Control* (2012.0) **21** 172-180
38. Marquez P.. (2017.0)
39. Gao W., Sanna M., Branston J.R., Chiou H.-Y., Chen Y.-H., Wu A., Wen C.P.. **Exploiting a low tax system: non-tax-induced cigarette price increases in Taiwan 2011–2016**. *Tobac. Control* (2019.0) **28** e126-e132
40. Syed-Abdul S., Malwade S., Hors-Fraile S., Spachos D., Fernandez-Luque L., Su C.-T., Li Y.-C.. **Smoking cessation supported by mobile app in Taiwan [journal article]**. *Tobacco Prevent. Cessat.* (2018.0) **4**. DOI: 10.18332/tpc/91509
41. Huang W.-H., Hsu H.-Y., Chang B.C.-C., Chang F.-C.. **Factors correlated with success rate of outpatient smoking cessation services in Taiwan**. *Int. J. Environ. Res. Publ. Health* (2018.0) **15** 1218
42. **The 12th international congress on obesity**. *Obes. Rev.* (2014.0) **15** e1-e22. DOI: 10.1111/obr.12196
43. 43Health Promotion AdministrationTaiwan's Obesity Prevention and Management Strategy2018Ministry of Health and Welfarehttps://www.hpa.gov.tw/File/Attach/10299/File_11744.pdf. (2018.0)
44. Hung K.-H., Liou K.-C., Hsu K.-N., Hu C.. **Disparities in ischemic stroke subtypes and risk factors between Taiwanese aborigines and Han Chinese in Taitung, Taiwan**. *Int. J. Gerontol.* (2016.0) **10** 17-21
45. Chuang Y.-H.. (2017.0)
46. Chang K.-S.. **Pursuing excellence as a game in the eyes of the devil-on exit mechanism in higher education [pursuing excellence as a game in the eyes of the devil-on exit mechanism in higher education]**. *Formosan Edu. Society* (2012.0) 75-105. DOI: 10.6429/fes.201206.0075
47. Pan Y.-L.. **A review of the current status of Taiwan aboriginal students' attending higher education**. *Taiwan Edu. Rev. Monthly* (2017.0) **6** 9-16
48. Huang C.-K.. **A study on higher education policies for indigenous residents in Taiwan [A study on higher education policies for indigenous residents in Taiwan]**. *J. Edu. Res. Develop.* (2018.0) **14** 33-64. DOI: 10.3966/181665042018091403002
49. Huang T.-Y., Hsieh C.-A.. **Research issues regarding city-county mergers and reorganization: the examples of Kaohsiung, Tainan and Taichung city**. *Soochow J. Political Sci.* (2014.0) **32** 51-130
50. Lee S.W., Yon D.K., James C.C., Lee S., Koh H.Y., Sheen Y.H., Sugihara G.. **Short-term effects of multiple outdoor environmental factors on risk of asthma exacerbations: age-stratified time-series analysis**. *J. Allergy Clin. Immunol.* (2019.0) **144** 1542-1550. PMID: 31536730
51. Lee Y.-Y., Hsieh Y.-K., Chang-Chien G.-P., Wang W.. **Characterization of the air quality index in southwestern Taiwan**. *Aerosol Air Qual. Res.* (2019.0) **19** 749-785
52. Kumari S., Jain M.K.. **A critical review on air quality index**. *Environ. Pollut.* (2018.0) 87-102
|
---
title: Inhibiting O-GlcNAcylation impacts p38 and Erk1/2 signaling and perturbs cardiomyocyte
hypertrophy
authors:
- Kyriakos N. Papanicolaou
- Jessica Jung
- Deepthi Ashok
- Wenxi Zhang
- Amir Modaressanavi
- Eddie Avila
- D. Brian Foster
- Natasha E. Zachara
- Brian O'Rourke
journal: The Journal of Biological Chemistry
year: 2023
pmcid: PMC9988579
doi: 10.1016/j.jbc.2023.102907
license: CC BY 4.0
---
# Inhibiting O-GlcNAcylation impacts p38 and Erk1/2 signaling and perturbs cardiomyocyte hypertrophy
## Body
The enzyme O-GlcNAc transferase (OGT) utilizes UDP-GlcNAc as a substrate and covalently links the GlcNAc moiety to Ser/Thr residues of intracellular proteins [1, 2, 3]. The O-GlcNAcase (OGA) catalyzes the opposing reaction, removing O-GlcNAc from proteins [4, 5]. Collectively known as O-GlcNAcylation, this protein modification is reversible, highly dynamic, and positioned at the intersection of homeostatic and stress pathways [6, 7]. Both Ogt and *Oga* genes, necessary for O-GlcNAc cycling, are essential for mammalian development and survival [8, 9, 10]. In the heart and the cardiac myocyte, O-GlcNAc cycling has been shown to play important roles in ischemia-reperfusion injury [11, 12], hypertrophic dysfunction [13, 14, 15, 16], diabetic arrhythmias and cardiomyopathy [17, 18, 19], and heart failure [20, 21]. The current view suggests that transient elevation of O-GlcNAcylation serves a protective role against stress, whereas prolonged increases exacerbate cardiac pathology [22, 23, 24, 25]. Furthermore, it has been postulated that highly dynamic O-GlcNAcylation overlaps with phosphorylation to provide an additional layer of regulation of protein function [26]. Focused studies in the heart and cardiac myocytes found that O-GlcNAcylation impacts the function of diverse kinases such as the Ca2+/calmodulin-dependent kinase CaMKII [27], the AMP-dependent protein kinase [28], and the growth regulator mTOR [29].
The mitogen-activated protein kinases (MAPKs) are a conserved family of kinases forming an ordered signaling cascade (MAP3K/MAP2K/MAPK), activated by mechanical and neuroendocrine factors acting on G protein–coupled receptors [30, 31]. The main end-effectors of MAPK signaling, Erk$\frac{1}{2}$, p38α-δ, and Jnk1-3, have unique and overlapping effects in stressed cardiac myocytes, leading to proliferation, growth, and cell death. Erk$\frac{1}{2}$ mediate cardiomyocyte proliferation and survival [32, 33, 34] and are also important in physiological hypertrophy and proper structural organization of growing myofibrils [35, 36, 37, 38]. The Gαq agonist phenylephrine (PE) acts through the small G protein Ras to activate the MAP3K Raf, which phosphorylates and activates MAP2Ks MEK$\frac{1}{2}$, upstream activators of Erk$\frac{1}{2}$ [39]. Downstream targets of Erk$\frac{1}{2}$ are the ribosomal S6 kinases Rsk1-4 and Msk$\frac{1}{2}$ [40], which are kinases that relay the signal downstream to the translation machinery but also to transcription factors, such as Gata4, Elk-1, c-Myc, and Creb [41, 42, 43, 44, 45].
While Erk$\frac{1}{2}$ are thought to be associated with adaptive physiological responses, p38 signaling kinases are associated with inflammation, maladaptive remodeling, and reduced cardiac contractility [46, 47, 48]. Gαq agonists mildly promote the phosphorylation of p38 through parallel mechanisms that involve phospholipase C and proteins Gβγ, which in turn activate the Rho family of GTPases (RhoA, Rac, Cdc42), ultimately activating MKK$\frac{3}{6}$ and p38 [49, 50, 51]. The downstream targets of p38 include kinases Msk1, Mk$\frac{2}{3}$, and Mnk$\frac{1}{2}$ that, in turn, phosphorylate their own downstream targets such as the small heat shock protein Hsp27, the translation initiation factor eIF4E, and the transcription factors Elk-1, C/EBPβ, and Creb [49, 52]. Some evidence indicates a direct interplay between O-GlcNAcylation and phosphorylation on transcription factors c-Myc, Creb, C/EBPβ [53, 54, 55], and Hsp27 [56, 57, 58]. However, direct evidence for O-GlcNAcylation on upstream MAPKs, e.g., on Erk$\frac{1}{2}$ and p38, is sparse.
In the present work, we examined whether short-term changes in O-GlcNAcylation could affect basal and stimulated MAPK signaling in cardiac myocytes. Our data indicate that decreasing O-GlcNAcylation with the OGT inhibitor OSMI-1 induces basal p38 phosphorylation at the early phase, whereas it blocks Erk$\frac{1}{2}$ phosphorylation and causes growth impairment at a later phase. We found several MAPK signaling members that could be putative targets for O-GlcNAcylation, including Erk$\frac{1}{2}$ and p38, but also upstream (MEK$\frac{1}{2}$, Tab1) and downstream mediators (Hsp27 and Creb). Our data reveal the widespread and intricate relationship between O-GlcNAcylation and MAPK signaling in cardiac myocytes, paving the way for further investigations into the role of O-GlcNAcylation in MAPK-related pathophysiology, including cardiac hypertrophy and heart failure.
## Abstract
The dynamic cycling of O-linked GlcNAc (O-GlcNAc) on and off Ser/Thr residues of intracellular proteins, termed O-GlcNAcylation, is mediated by the conserved enzymes O-GlcNAc transferase (OGT) and O-GlcNAcase. O-GlcNAc cycling is important in homeostatic and stress responses, and its perturbation sensitizes the heart to ischemic and other injuries. Despite considerable progress, many molecular pathways impacted by O-GlcNAcylation in the heart remain unclear. The mitogen-activated protein kinase (MAPK) pathway is a central signaling cascade that coordinates developmental, physiological, and pathological responses in the heart. The developmental or adaptive arm of MAPK signaling is primarily mediated by Erk kinases, while the pathophysiologic arm is mediated by p38 and Jnk kinases. Here, we examine whether O-GlcNAcylation affects MAPK signaling in cardiac myocytes, focusing on Erk$\frac{1}{2}$ and p38 in basal and hypertrophic conditions induced by phenylephrine. Using metabolic labeling of glycans coupled with alkyne-azide “click” chemistry, we found that Erk$\frac{1}{2}$ and p38 are O-GlcNAcylated. Supporting the regulation of p38 by O-GlcNAcylation, the OGT inhibitor, OSMI-1, triggers the phosphorylation of p38, an event that involves the NOX2–Ask1–MKK$\frac{3}{6}$ signaling axis and also the noncanonical activator Tab1. Additionally, OGT inhibition blocks the phenylephrine-induced phosphorylation of Erk$\frac{1}{2.}$ Consistent with perturbed MAPK signaling, OSMI-1–treated cardiomyocytes have a blunted hypertrophic response to phenylephrine, decreased expression of cTnT (key component of the contractile apparatus), and increased expression of maladaptive natriuretic factors Anp and Bnp. Collectively, these studies highlight new roles for O-GlcNAcylation in maintaining a balanced activity of Erk$\frac{1}{2}$ and p38 MAPKs during hypertrophic growth responses in cardiomyocytes.
## Contribution of MAPKs Erk1/2 and p38 to the hypertrophic response induced by PE
Neonatal rat ventricular myocytes (NRVMs) provide a cellular model to investigate the role of the MAPK signaling cascade in cardiac hypertrophy. Exposure to the alpha-1 adrenergic receptor agonist PE (5 μM, 30 min) increased the phosphorylation of Erk$\frac{1}{2}$, p38, and the downstream target Creb (Fig. 1A). To investigate the PE-induced hypertrophic signaling, we employed SCH772984 and SB202190, chemical inhibitors to kinases Erk$\frac{1}{2}$ and p38α/β, respectively [59, 60, 61]. Hypertrophic growth was assessed at 24 h of PE stimulation by staining for actin using phalloidin (Fig. 1B). As shown in Figure 1, B and C, PE increased NRVM size (2.1-fold increase, $p \leq 0.001$), and this was reduced by the Erk$\frac{1}{2}$ inhibitor SCH772984 ($50\%$ reduction, $p \leq 0.001$). The p38 inhibitor SB202190 did not significantly affect PE-induced growth, while the combination of the two strongly suppressed PE-induced NRVM growth (Fig. 1C). The total numbers of nuclei per random field imaged were not significantly different across groups (Fig. 1D).Figure 1Phenylephrine activates MAPKs Erk$\frac{1}{2}$ and p38 and their inhibition impairs cardiomyocyte growth. A, primary neonatal rat ventricular myocytes (NRVMs) incubated for 24 h in medium without growth factors were stimulated with 5 μM phenylephrine (PE) and harvested 30 min later for Western blot analysis. Phosphorylation of Erk$\frac{1}{2}$, p38, and Creb was assessed with their respective phospho-specific and total antibodies. B, NRVMs were treated with the Erk$\frac{1}{2}$ inhibitor SCH772984 (10 μM), the p38 inhibitor SB202190 (10 μM), or both and 6 h later, they were exposed to 5 μM PE. After 24 h of combined treatment, the cells were fixed and stained with phalloidin-Alexa 594 to detect and quantify fiber actin (F-actin) as index of NRVM cell size (hypertrophy). The scale bar represents 100 μm. C, per cytoplasm integrated signal intensity of phalloidin stain and (D) total number of nuclei identified per field of view. The data in (C and D) were extracted from confocal images imported into Cell Profiler. For the cytoplasmic F-actin signal, the mean value across a given field of view containing ∼850 cells is represented as a single data point. The number of fields quantified in each treatment group are shown in their respective bars. Bars represent means ± standard error. Comparisons across treatment groups were done with one-way ANOVA and Tukey post hoc test. ∗∗∗∗$p \leq 0.0001$; ns; not significantly different. E–J, representative Western blots and band intensity quantifications resulting from the indicated total and phospho-specific blots. Cells were pretreated for 6 h with 10 μM of the indicated inhibitors and followed by 30 min stimulation with 5 μM PE. The bar graphs represent mean band intensities averaged across the indicated number of biological replicates. K and L, cTnT and Anp mRNA expression across the indicated treatment groups. Cells were pretreated for 6 h with 10 μM of the indicated inhibitors followed by 24 h of stimulation with 5 μM PE before RNA isolation and cDNA synthesis for quantitative PCR. Numbers of biological replicates are shown in their respective bars. Bars represent means ± standard error. Comparisons across treatment groups were done with one-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, ns; not significantly different. MAPK, mitogen-activated protein kinase; NRVM, neonatal rat ventricular myocyte; PE, phenylephrine.
SCH772984 was potent in reducing the PE-induced phosphorylation of Erk$\frac{1}{2}$ and, interestingly, this response was also recapitulated by SB202190 (Fig. 1, E and F). In fact, the two inhibitors appeared to have an additive effect in suppressing PE-induced Erk$\frac{1}{2}$ phosphorylation (Fig. 1F). In cells treated with the p38 inhibitor SB202190, we observed a paradoxical increase in p38 phosphorylation. This, however, can be explained by considering that the phosphorylation of p38 is regulated by a feedback circuit, where inhibition of p38 derepresses its upstream kinase TAK1, leading to compensatory phosphorylation of p38, despite inhibition of its activity [62]. Furthermore, the pyridinyl imidazole p38 inhibitors (e.g., SB202190 or SB203580) block p38 at the active site of the kinase without impacting its phosphorylation [63]. These two effects together explain the increased p38 phosphorylation with SB202190 in basal and PE-treated conditions (Fig. 1, E and G).
A downstream target of p38, MK2, phosphorylates the small heat shock protein Hsp27. We confirmed the functional inhibition of p38 by examining the phosphorylation of Hsp27. As expected, PE-induced phosphorylation of Hsp27 was completely suppressed by SB202190 (Fig. 1, H and I). In contrast, SCH772984 and SB202190 alone, or in combination, did not significantly reduce PE-induced phosphorylation of Creb (Fig. 1, H and J).
We then examined the expression levels of cTnT and Anp as markers of hypertrophic signaling. SCH772984 or SB202190 alone prevented the PE-induced cTnT expression and were most potent when used in combination (Fig. 1K). On the other hand, SCH772984, but not SB202190, significantly reduced PE-induced Anp (Fig. 1L), consistent with a regulatory pathway linking Erk$\frac{1}{2}$ activation to GATA-mediated transcription of Anp [64]. Collectively, these experiments demonstrate that Erk$\frac{1}{2}$ and p38 cooperatively contribute to the growth response in NRVMs downstream of PE.
## Metabolic labeling with clickable Ac4GalNAlk identifies O-GlcNAc–modified versions of p38 and Erk1/2
Next, we interrogated whether p38 and Erk$\frac{1}{2}$ were O-GlcNAcylated. To do that, we employed metabolic labeling in cells using the sugar analogs Ac4GalNAz or Ac4GalNAlk. These are biosynthetically converted into UDP-GalNAz or UDP-GalNAlk respectively, serving as substrates of OGT in intracellular protein O-GlcNAcylation [65, 66, 67] (Fig. 2A). In a proof of concept experiment, we treated cells with 200 μM Ac4GalNAz (negative control) or Ac4GalNAlk for 24 h, and protein extracts were used in Alkyne-Azide click reactions with CalFluor 647 azide, a probe that fluoresces only after it reacts with an alkyne [68]. Using in-gel fluorescence, we detected widespread incorporation of the fluorogenic probe in proteins from Ac4GalNAlk-treated cells, but not in control Ac4GalNAz-treated cells (Fig. 2B). Furthermore, there was no background fluorescent reactivity in the absence of the click catalyst (Cu/THPTA, Fig. 2B). Next, we treated NRVMs with Ac4GalNAz or Ac4GalNAlk as above and proteins were extracted from metabolically labeled cells. Alkyne-azide click reactions were performed with biotin azide and biotinylated proteins were pulled-down using streptavidin-conjugated agarose beads. The different fractions from the pull-down were subjected to Western blot with streptavidin IR Dye 800 showing the efficient enrichment of biotinylated proteins with this approach (Fig. 2C).Figure 2Bio-orthogonal metabolic labeling of glycans coupled with enrichment and immunoblotting identifies O-GlcNAcylated members of the MAPK pathway. A, schematic of the experimental setup for the metabolic labeling of cells with the sugar analogs Ac4GalNAz (negative control) or Ac4GalNAlk. The latter serves as the alkyne donor in a copper-catalyzed alkyne-azide ‘click’ reaction. Cells were incubated with 200 μM Ac4GalNAz or Ac4GalNAlk for 24 h prior to protein extraction. Subsequently, protein extracts are subject to ‘click’ with CalFluor Azide and in-gel fluorescence scanning or Biotin Azide for streptavidin pull-down and Western blot detection. B, protein extracts (20 μg) were reacted with CalFluor 647 Azide (final concentrations 20 μM or 50 μM) with or without the copper catalyst. The samples were resolved on a gel and scanned to detect modified proteins by in-gel fluorescence. C, NRVMs metabolically labeled with Ac4GalNAz or Ac4GalNAlk for 24 h were collected and lysed. Five hundred micrograms of extracted protein were reacted with biotin azide and the ‘click’ reaction was cleaned-up by methanol/chloroform precipitation. Precipitated proteins were resuspended in $0.05\%$ SDS, 50 mM Tris HCl, pH 8.0, and $2\%$ of the input was taken before adding the protein suspension to streptavidin-agarose beads for binding (see also Experimental procedures for details). All collected fractions were loaded in duplicates and probed with streptavidin IR Dye 800 in Western blot. D, protein samples obtained after ‘click’ reaction onto biotin azide plus, enriched as described above and analyzed by immunoblotting for the presence of glycosylated and enriched candidates MEK$\frac{1}{2}$, Erk$\frac{1}{2}$, Creb, and Gata4. E, protein samples obtained as described above were analyzed by immunoblotting for the presence of glycosylated and enriched candidates p38 and Hsp27. The samples were loaded in duplicates in this experiment. MAPK, mitogen-activated protein kinase; NRVM, neonatal rat ventricular myocyte; O-GlcNAc, O-linked GlcNAc.
We next analyzed the biotinylated samples to identify evidence for O-GlcNAcylation on MAPKs Erk$\frac{1}{2}$ and p38. Indeed, by Western blotting, we found that the pulled-down fraction contained Erk$\frac{1}{2}$ and also its upstream MAP2Ks, MEK$\frac{1}{2}$ (Fig. 2D). The MAPK-downstream transcription factor Creb was also detected, consistent with previous reports showing that it is O-GlcNAcylated [55]. Another transcription factor regulated by MAPK signaling, Gata4, was also present in the pull-down (Fig. 2D). Moreover, we found that the other MAPK of interest, p38, was pulled-down and so was its downstream target Hsp27 (Fig. 2E). Taken together, these findings indicate that several members in the MAPK signaling pathway are putative targets for O-GlcNAcylation in cardiac myocytes.
## TMG and OSMI-1 respectively increase and decrease protein O-GlcNAcylation in NRVMs without impacting the overall abundance of other glycans
Protein O-GlcNAcylation was manipulated in NRVMs with Thiamet G (TMG) and OSMI-1, inhibitors of OGA and OGT, respectively [69, 70]. Treatment with TMG (200 nM, 6 h) led to more than $50\%$ increase in protein O-GlcNAcylation, while OSMI-1 (25 μM) reduced O-GlcNAcylation by $50\%$ (Fig. 3, A and B). Consistent with earlier reports, OGA protein was significantly upregulated in cells treated with TMG (2.4-fold higher, TMG versus vehicle, Fig. 3, C and D), whereas OGT was significantly upregulated in cells treated with OSMI-1 (2.1-fold higher, OSMI-1 versus vehicle, Fig. 3, C and E). PE treatment (5 μM, 30 min) did not significantly change the levels of O-GlcNAcylation, although longer treatment durations (3 and 6 h) induced modest but significant elevations in protein O-GlcNAcylation ($20.5\%$ and $36.3\%$ increase respectively, Fig. 3, F and G).Figure 3TMG and OSMI-1, chemical inhibitors of OGA and OGT, respectively, increase and decrease protein O-GlcNAcylation with reciprocal changes in the protein abundance of OGA and OGT.A and B, NRVMs incubated for 24 h in medium without growth factors were exposed to 200 nM Thiamet G (TMG) or 25 μM OSMI-1 and were harvested 6 h later for Western blot analysis. The detection of O-GlcNAcylated proteins was carried out using antibody RL2 (mouse IgG), and the cumulative band signal along each individual lane ranging from 50 to 250 kDa was quantified and expressed relative to the total protein load. Summary results are shown in (B). Comparisons across treatment groups were done with one-way ANOVA and Tukey post hoc test, ∗∗∗∗$p \leq 0.0001$, $$n = 3$.$ C–E, NRVMs were pretreated for 6 h with TMG or OSMI-1 and then exposed to 5 μM PE for an additional 30 min. Samples were analyzed for protein abundance of OGA, OGT, and overall protein O-GlcNAcylation. The bar graphs in (D and E) show the mean OGA and OGT abundance across the different treatment groups, and the number of biological replicates are indicated inside the bars. F and G, NRVMs incubated for 24 h in medium without serum were left untreated (Control) or treated with 5 μM PE for 30 min, 3 h, or 6 h after which they were harvested for O-GlcNAc Western blot analysis and quantification as described above. The summary of the quantifications is shown in (G). Comparisons across treatment groups were done with one-way ANOVA and Tukey post hoc test, ∗$p \leq 0.05$, ∗∗ $p \leq 0.01$, ∗∗∗$p \leq 0.001.$ NRVM, neonatal rat ventricular myocyte; O-GlcNAc, O-linked GlcNAc; OGA, O-GlcNAcase; OGT, O-GlcNAc transferase; PE, phenylephrine.
Furthermore, we examined whether TMG or OSMI-1 altered the abundance of GlcNAc-, GalNAc-Galactose–, and GalNAc-Mannose–bearing glycoproteins, whose biosynthesis might be affected by potential alterations in UDP-GlcNAc levels caused by either TMG or OSMI-1. Lectin blots with biotinylated wheat germ agglutinin, peanut agglutinin, and *Dolichos biflorus* agglutinin did not identify changes in glycan abundances between control and TMG- or OSMI-1–treated cells (Fig. S1, A–E). Given OSMI-1’s structural similarity with UDP-GlcNAc, we examined whether OSMI-1 might elicit a confounding impact on N-linked glycosylation. However, lectin blots for Concanavalin A (ConA) revealed that the abundances of N-linked glycoproteins were not impacted significantly by OSMI-1 (Fig. S1, F and G). On the other hand, tunicamycin, which blocks the addition of GlcNAc onto dolichol phosphate, induced a significant reduction in the ConA signal (Fig. S1, F and G).
## OGT inhibition with OSMI-1 induces p38 phosphorylation and downstream signaling
Next, we examined the effects of TMG and OSMI-1 on MAPK phosphorylation with or without PE. We found that OSMI-1 induced a nearly 4-fold increase in p38 phosphorylation (3.9-fold increase versus untreated, $p \leq 0.001$, Fig. 4, A and B). In contrast, TMG alone or in combination with PE did not impact p38 phosphorylation. Moreover, OSMI-1 had no effect on Erk$\frac{1}{2}$ phosphorylation, either at baseline or in PE-treated conditions (Fig. 4, A and C). p38 phosphorylation gradually increased during an 18-h exposure to OSMI-1, mirrored by a progressive increase in Hsp27 phosphorylation (Fig. 4D). Hsp27 phosphorylation was increased in PE-stimulated conditions, although OSMI-1 alone caused higher Hsp27 phosphorylation (Fig. 4, E and F). We next examined if OSMI-1 caused widespread changes in the phosphorylation of MAPK substrates. While we found that phosphorylation of MAPK/CDK substrates was increased by PE treatment, this was not the case in OSMI-1–treated cells (Fig. 4, G and H). These data suggest that the induction of p38 phosphorylation by OSMI-1 is a selective event that impacts the MAPK signaling pathway at a level that is distinct from alpha-1 adrenergic activation and the broad induction of its downstream effectors (e.g., phospholipase C, inositol triphosphate, diacylglycerol, Ca2+). Finally, the p38 inhibitor SB202190 was sufficient to completely abolish the OSMI-1–induced phosphorylation of Hsp27 and it was also effective at decreasing the OSMI-1–induced phosphorylation of the transcription factor Creb (Fig. 4, I–K). Collectively, these experiments indicate that OGT inhibition selectively activates p38 phosphorylation and its downstream signaling. Figure 4OGT inhibition induces phosphorylation of p38 and activates its downstream signaling pathway. A–C, representative western blots of p38 and Erk$\frac{1}{2}$ (phospho-specific and total). Cells were exposed for 6 h to TMG (200 nM), OSMI-1 (25 μM), or a combination of TMG/OSMI-1, and this was followed by treatment for 30 min with or without PE (5 μM). Comparisons across the different samples were performed with one-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, ns; not significant ($$p \leq 0.88$$). The number of biological replicates in each group is indicated in their respective bars. D, Western blot analysis of p38 and Hsp27 phosphorylation during a time-course of 1 to 18 h of exposure to OSMI-1. E and F, Western blot and quantitation of Hsp27 phosphorylation induced by 25 μM OSMI-1 (6 h), 5 μM PE (30 min), or both. Comparisons between groups were done with one-way ANOVA and Tukey post hoc test. ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, four biological replicates per group. G and H, Western blot of proteins with the phosphorylated motifs PXSP or SPXR/K, target sites of members of the MAPK, and cyclin dependent kinases (CDKs). For quantifications, the cumulative band intensities along the lane were obtained from proteins with molecular weights of 50 to 250 kDa and were normalized with the respective cumulative lane intensities from the total protein blots. Comparisons between groups were done with one-way ANOVA and Tukey post hoc test. ∗∗∗$p \leq 0.001$, the comparison between PE versus PE/OSMI-1 yields $$p \leq 0.57$$, $$n = 3$$ biological replicates per group. I–K, Western blot and quantitation of Hsp27 and Creb phosphorylation induced by 25 μM OSMI-1 (6 h) but repressed by the p38 inhibitor SB202190 (10 μM, 6 h). Comparisons between groups were done with one-way ANOVA and Tukey post hoc test. ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, four biological replicates per group. MAPK, mitogen-activated protein kinase; OGT, O-GlcNAc transferase; PE, phenylephrine; TMG, Thiamet G.
Following the finding that reducing protein O-GlcNAcylation with OSMI-1 triggers p38 phosphorylation, we tested if other OGT inhibitors induce p38 phosphorylation. OSMI-4 is structurally related to OSMI-1 and inhibits OGT by competitive docking into the UDP-GlcNAc binding pocket of OGT [71]. In addition, we used 5SGlcNHex, a metabolic inhibitor that is converted to UDP-5SGlcNAc which then functions as a competitive OGT inhibitor [72]. Treating NRVMs with each inhibitor (25 μM, 6 h) induced significant reductions in protein O-GlcNAcylation ($52.2\%$, $25.9\%$, and $42.9\%$ of baseline for OSMI-1, OSMI-4, and 5SGlcNHex, respectively; Fig. 5, A and B). Importantly, all three inhibitors induced significant increases in p38 phosphorylation (3.4-fold, 3.0-fold, and 1.5-fold increases for OSMI-1, OSMI-4, and 5SGlcNHex, respectively; Fig. 5, D and E), confirming that the effect was not an epiphenomenon of the particular chemical structure of OSMI-1.Figure 5Three different OGT inhibitors lower protein O-GlcNAcylation and induce p38 phosphorylation in NRVMs. A–C, NRVMs were treated with the OGT inhibitor OSMI-1 (25 μM), its structurally related derivative OSMI-4 (25 μM, ethylester form -4b), and the precursor metabolic inhibitor 5SGlcNHex (25 μM). After 6 h of treatment, cell extracts were analyzed for changes in protein O-GlcNAcylation and OGT protein abundance. Statistical differences between groups were assessed by one-way ANOVA and Tukey post hoc test. ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, three biological replicates per group. D and E, NRVM extracts obtained as described above were analyzed for p38 phosphorylation. Statistical differences between groups were assessed by one-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, three biological replicates per group. NRVM, neonatal rat ventricular myocyte; O-GlcNAc, O-linked GlcNAc; OGT, O-GlcNAc transferase.
## OGT inhibition with OSMI-1 prevents the O-GlcNAcylation of p38
As shown earlier, using metabolic labeling with Ac4GalNAlk, we found evidence that p38 is directly O-GlcNAcylated (Fig. 2E). Given OSMI-1’s impact on potently increasing p38 phosphorylation, we examined whether OSMI-1 was directly affecting p38 O-GlcNAcylation. To that end, we metabolically labeled HEK293 cells with Ac4GalNAlk, concomitantly with exposure to vehicle, TMG, or OSMI-1 (see also schematic in Fig. S2A). Using streptavidin-mediated pull-down, we found efficient enrichment of biotinylated proteins in Ac4GalNAlk-treated cells but not in Ac4GalNAz negative control cells (Fig. S2B). Follow-up Western blots showed that p38 was present in pull-downs from vehicle- and TMG-treated cells, but not in OSMI-1–treated cells (Fig. S2C). These findings further confirm the previous observations that p38 is O-GlcNAcylated at baseline and demonstrate that this can be prevented by OGT inhibition with OSMI-1.
## Activation of p38 downstream of OSMI-1 is dependent on canonical MAP3K–MAP2K signaling axis
Next, we examined whether upstream kinases might be involved in the phosphorylation of p38. There are currently very few commercially available inhibitors for MKK3 and MKK6, which are the immediate upstream activators of canonical p38 phosphorylation (see also schematic in Fig. 6A). Therefore, we employed siRNA-mediated knockdown of MKK3 and MKK6, to test whether their depletion could impair the observed OSMI-1–induced p38 phosphorylation. Indeed, we found that transfecting NRVMs with siRNAs targeting MKK3 or MKK6 significantly decreased OSMI-1–induced phosphorylation of p38 ($34.7\%$ reduction, OSMI-1/nontargeting control siRNA versus OSMI-1/siMKK3 and $25.9\%$ reduction, OSMI-1/nontargeting control siRNA versus OSMI-1/siMKK6, $p \leq 0.05$, Fig. 6, B and C).Figure 6Phosphorylation of p38 downstream of OGT inhibition is mediated by MAP2Ks MKK$\frac{3}{6}$ and MAP3K Ask1.A, schematic illustrating the canonical pathway of p38 phosphorylation involving MAP3Ks Ask1, Tak1, Mlk3, and MAP2Ks MKK3 and MKK6. B and C, NRVMs were transfected with nontargeting control (NTC), MKK3-, or MKK6-targeting siRNA (alone or combined, final concentration 20 nM) and 48 h later, they were treated with vehicle or OSMI-1 (25 μM) for an additional 6 h. Subsequently, cell extracts were analyzed for the phosphorylation of p38 by Western blot. D–F, NRVMs were transfected with NTC, Map3k5 (Ask1)-, Map3k7 (Tak1)-, or Map3k11 (Mlk3)-targeting siRNAs (final concentration 20 nM) and 48 h later, they were treated with vehicle or OSMI-1 (25 μM) for an additional 6 h. Subsequently, cell extracts were analyzed for the phosphorylation of p38 and Hsp27 by Western blot. Statistical differences were assessed by two-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, ns; not significantly different. NRVM, neonatal rat ventricular myocyte; OGT, O-GlcNAc transferase.
MAP3Ks Ask1, Tak1, and Mlk3 are well-known inducers of MKK$\frac{3}{6}$ in canonical p38 activation and have been described as potential MKK$\frac{3}{6}$ activators in cardiac myocytes (Fig. 6A). The Ask1-specific inhibitor GS-444217 was tested for its potency to abolish OSMI-1–induced p38 phosphorylation (1 μM GS-444217 used alone or together with OSMI-1 for 6 h treatment). However, we observed that this treatment did not significantly reduce OSMI-1–induced p38 phosphorylation or the phosphorylation of downstream Hsp27 (Fig. S3, A–C). Consistently, when a range of GS-444217 concentrations was used (100 nM, 500 nM, 1 μM), we did not observe reductions in OSMI-1–induced p38 phosphorylation, and when a higher concentration of GS-444217 was used (10 μM), it potentiated instead of inhibiting OSMI-1–induced p38 phosphorylation and downstream Hsp27 phosphorylation (Fig. S3, D–F). Next, we used the Tak1-specific inhibitor Takinib, either alone or in combination with GS-444217, and examined whether this treatment could impair OSMI-1–induced p38 phosphorylation. However, OSMI-1–induced p38 phosphorylation was still observed in the presence of Takinib, and it was further potentiated by the presence of GS-444217 (Fig. S3, G and H). Finally, we examined the effect of Mlk3 inhibitor URMC-099. Unexpectedly, this inhibitor was potent in inducing p38 alone (Fig. S3, I and J), making it unsuitable for the study of the OSMI-1–induced p38 phosphorylation mechanism. It is unclear why GS-444217 or URMC-099 induces p38 phosphorylation. One potential explanation could be due to off-target effects on other kinases operating upstream of p38 phosphorylation. Another possibility could be that accumulation of the kinase (e.g., Ask1) in its inactive state in the cell elicits secondary pathways leading to the paradoxical p38 phosphorylation and activation.
Given these drawbacks with the MAP3K inhibitors, we resorted to siRNA-mediated knockdown of the kinases. Importantly, we found that transfecting NRVMs with an Ask1-targeting siRNA induced a significant reduction in OSMI-1–induced p38 phosphorylation (Fig. 6, D and E). A modest reduction in OSMI-1–induced p38 phosphorylation was also observed with Tak1-targeting siRNA, whereas targeting Mlk3 did not reach statistical significance (Fig. 6, D and E). Importantly, the reduction in p38 phosphorylation by Ask1-targeting siRNA led to a significant reduction in downstream Hsp27 phosphorylation (Fig. 6F).
Taken together, these findings with siRNA-mediated MAP3K targeting illustrate that OSMI-1–induced p38 phosphorylation is mediated by MKK$\frac{3}{6}$ MAP2Ks that are in turn activated predominantly by Ask1 with potentially lesser inputs from Tak1 MAP3K.
## OSMI-1–induced p38 phosphorylation requires NADPH oxidase 2 subunits p47phox and gp91phox
Ask1 can be activated due to the accumulation of reactive oxygen species (ROS) in the cytosol. One source of cytosolic ROS in cardiomyocytes is the family of NADPH oxidases (NOX), multicomponent enzymes on the plasma membrane that transfer electrons from NADPH to oxygen leading to superoxide production [73]. To examine whether NOXs are implicated in OSMI-1–induced p38 activation in cardiomyocytes, we used siRNA to target p47phox (encoded by gene neutrophil cytosolic factor 1, Ncf1) a key regulatory subunit that upon activation translocates from the cytosol to the plasma membrane to associate with NOX2 and activate the production of superoxide. In agreement with a role in this pathway, we found that targeting p47phox significantly reduced the phosphorylation of p38 as well as the phosphorylation of downstream target Hsp27 in OSMI-1–treated cells (Fig. 7, A–C). Furthermore, we used siRNA to target the core subunit of NOX2, gp91phox. Similarly, this approach reduced the phosphorylation of p38 and Hsp27 (Fig. 7, D–F), further implicating NOX2 as a causal upstream mediator of OSMI-1–induced p38 phosphorylation through the canonical Ask1–MKK$\frac{3}{6}$ pathway. Figure 7Phosphorylation of p38 downstream of OGT inhibition requires NADPH oxidase 2 subunits p47phox and gp91phox. A–C, NRVMs were transfected with nontargeting control (NTC) or p47phox-targeting siRNA (the protein p47phox is encoded by the gene neutrophil cytosolic factor 1, Ncf1). After 48 h of transfection, cells were treated with vehicle or OSMI-1 (25 μM) for an additional 6 h. Cell extracts were analyzed for the phosphorylation of p38 and its downstream target Hsp27 by Western blot. D-F, NRVMs were transfected with NTC or gp91phox-targeting siRNA (the protein gp91phox, also referred to as Nox2 is encoded by the gene cytochrome B-245 beta chain, Cybb). After 48 h of transfection, cells were treated with vehicle or OSMI-1 (25 μM) for an additional 6 h and the cell extracts were analyzed for the phosphorylation of p38 and Hsp27. Statistical differences between groups were assessed by two-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗∗$p \leq 0.0001$, three biological replicates per group. NRVM, neonatal rat ventricular myocyte; OGT, O-GlcNAc transferase.
## Activation of p38 downstream of OSMI-1 is dependent on the noncanonical signaling axis and requires Tab1
In addition to the canonical activation of p38 by MAP3Ks, p38 can undergo noncanonical activation by autophosphorylation. p38’s autophosphorylation in cardiac myocytes is kept in check by the inhibitory complex Hsp90/Cdc37, which sequesters p38 from interacting with its activating scaffold Tab1 [74, 75]. Additionally, Tab1 is known to undergo O-GlcNAcylation [76]. In agreement, using metabolic labeling with Ac4GalNAlk in cardiac myocytes, we found that both Hsp90 and Tab1 were O-GlcNAcylated (Fig. 8A). These findings indicated that the noncanonical pathway of p38 activation (see also schematic in Fig. 8B) was potentially subject to regulation by OSMI-1. To address that, we first examined whether the Hsp90 inhibitor geldanamycin could induce noncanonical p38 signaling in NRVMs. Indeed, geldanamycin alone increased the phosphorylation of Tab1, and this effect was additive to the induction of Tab1 phosphorylation caused by OSMI-1 (Fig. 8, C and D). Similar to Tab1, geldanamycin increased the phosphorylation of p38 and Hsp27 in a manner that was additive to the effect of OSMI-1 (Fig. 8, E–G). These findings indicate that OSMI-1 is not an inhibitor of Hsp90 per se but rather operates parallel to Hsp90 inhibition to further increase p38 autophosphorylation. Figure 8Phosphorylation of p38 downstream of OGT inhibition is mediated by the noncanonical p38 activation pathway and requires protein Tab1.A, protein samples obtained after ‘click’ reaction and enriched with streptavidin pull-down were analyzed by immunoblotting for the presence of glycosylated targets Hsp90 and Tab1. B, schematic illustrating the noncanonical pathway for p38 activation, including the inhibitory complex Hsp90/Cdc37 and the scaffold protein Tab1. The Hsp90 inhibitor geldanamycin is also shown. C–G, NRVMs were treated with vehicle or OSMI-1 (25 μM) for 5 h, and then vehicle or geldanamycin (2 μM) were added to the cells for the remaining 1 h of incubation. Protein extracts were analyzed for the phosphorylation of Tab1, p38, and Hsp27 across the different treatment groups. Statistical differences between treatment groups were assessed by one-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001.$ H–K, NRVMs were transfected with nontargeting control (NTC) or Tab1-targeting siRNA (final concentration 20 nM) and 48 h later, they were treated with vehicle or OSMI-1 (25 μM) for an additional 6 h. Cell extracts were analyzed for Tab1 protein knockdown and phosphorylation of p38. Statistical differences between treatment groups were assessed by two-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, ns; not significantly different. NRVM, neonatal rat ventricular myocyte; OGT, O-GlcNAc transferase.
To further investigate the role of the noncanonical autophosphorylation pathway in OSMI-1–induced p38 phosphorylation, we knocked down Tab1 in NRVMs with siRNA. This approach led to a significant reduction in total Tab1 protein levels by 50 to $55\%$ in the absence or presence of OSMI-1 (Fig. 8, H and I). Notably, knockdown of Tab1 decreased the OSMI-1–induced phosphorylation of p38 ($23.4\%$ reduction, OSMI-1 versus OSMI-1/Tab1 si, $p \leq 0.05$, Fig. 8, J and K), implicating Tab1 as an additional mediator of p38 phosphorylation in OGT-inhibited NRVMs.
## OGT inhibition with OSMI-1 perturbs PE-induced Erk1/2 and p38 signaling and impairs PE-induced hypertrophic growth of cardiac myocytes
Next, we examined the impact of prolonged OSMI-1 treatment on the PE-induced effects on Erk$\frac{1}{2}$ and p38 phosphorylation. PE-induced phosphorylation of prohypertrophic Erk$\frac{1}{2}$ was significantly blunted by the 24-h OSMI-1 treatment (Fig. 9, A and B). In contrast to reducing the phosphorylation of Erk$\frac{1}{2}$, treatment with OSMI-1 for 24 h further increased the phosphorylation of pathologic p38 (Fig. 9, A and C). Interestingly, the increase in relative p38 phosphorylation was in part due to decreased protein levels of total p38 (Fig. 9A), suggesting perhaps the recruitment of compensatory mechanisms to restrict further pathologic increase in p38 activity. However, the phosphorylation of the p38-downstream factor Hsp27 was also significantly elevated in OSMI-1–treated cells, consistent with increased p38 activity (Fig. 9, A and D). On the other hand, the 24-h OSMI-1 treatment had no significant effect on basal Creb phosphorylation and also did not impact its PE-induced phosphorylation (Fig. 9, E and F).Figure 9OGT inhibition disrupts Erk$\frac{1}{2}$ and p38 phosphorylation and impairs the hypertrophic growth of NRVMs. A–F, Western blot and quantitation of Erk$\frac{1}{2}$, p38, and Hsp27 phosphorylation at 24 h after OSMI-1 exposure. NRVMs were treated with OSMI-1 (25 μM) and 24 h later, they were exposed to PE (5 μM) for an additional 30 min to stimulate signaling. Protein extracts were then analyzed for the phosphorylation of Erk$\frac{1}{2}$, p38, Hsp27, and Creb across the different treatment groups. G, NRVMs were treated with the OGA inhibitor TMG (200 nM), the OGT inhibitor OSMI-1 (25 μM), or both and 6 h later, they were exposed to PE (5 μM). After 24 h of combined treatment, the cells were fixed and stained with phalloidin-Alexa 594 to detect and quantify F-actin by confocal microscopy. The scale bar represents 100 μm. H and I, F-actin signal intensity per cytoplasm and total number of nuclei identified per field of view. The number of fields quantified per treatment group is shown in the respective bar graphs. Bars represent means ± standard error. Comparisons across treatment groups were done with one-way ANOVA and Tukey post hoc test. ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$; ns; not significantly different. J–M, cTnT, RCan1, Anp, and Bnp mRNA expression across the indicated treatment groups. Cells were pretreated with TMG (200 nM), OSMI-1 (25 μM), or both, and 6 h later, they were exposed to PE (5 μM) for 24 h followed by RNA isolation and cDNA synthesis for quantitative PCR. The number of biological replicates in each treatment group are shown in their respective bars. Bars represent means ± standard error. Comparisons across treatment groups were done with one-way ANOVA and Tukey post hoc test. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$, ∗∗∗∗$p \leq 0.0001$, ns; not significantly different. NRVM, neonatal rat ventricular myocyte; OGA, O-GlcNAcase; OGT, O-GlcNAc transferase; PE, phenylephrine; TMG, Thiamet G.
Given the increase in protein abundance of OGT observed after a 6-h OSMI-1 treatment (Fig. 3E), we examined whether the same was true for the 24-h treatment. Indeed, OSMI-1 significantly increased the protein levels of OGT and reciprocally decreased the protein levels of OGA (Fig. S4, A–C).
Finally, we examined whether manipulating O-GlcNAcylation could impact PE-induced NRVM hypertrophy. Cells were pretreated with OSMI-1 or TMG for 6 h followed by concomitant treatment with PE (5 μM, 24 h). PE induced a 2-fold increase in cell size in vehicle- and TMG-treated cells but remarkably, OSMI-1 prevented this hypertrophic growth (Fig. 9, G and H). This is consistent with the perturbed Erk$\frac{1}{2}$ and p38 signaling noted above. Furthermore, assessment of total nuclei numbers showed that OSMI-1, combined with PE, was associated with lower numbers (Fig. 9I), indicating that in addition to its growth-arresting effect, OSMI-1 also impairs myocyte viability in PE-stimulated conditions. At the molecular level, OSMI-1 suppressed the PE-induced increase of prohypertrophic genes cTnT and RCan1 (Fig. 9, J and K) but still allowed the induction of pathologic markers Anp and Bnp (Fig. 9, L and M).
## Discussion
A close relationship between phosphorylation signaling cascades and O-GlcNAcylation has been previously recognized [26, 77]. The MAPK signaling cascade is a key determinant of cardiomyocyte growth and it also confers myocyte adaptation to diverse cardiac stresses. In this work, we investigated the functional intersections between O-GlcNAcylation and MAPK signaling in cardiac myocytes at baseline and during physiological hypertrophic growth. The salient findings of the study are as follows: [1] the MAPKs Erk$\frac{1}{2}$ strongly contribute to the hypertrophic growth of cardiac myocytes downstream of PE stimulation, with lesser input from p38, [2] a subset of both Erk$\frac{1}{2}$ and p38 are O-GlcNAcylated, as well as signaling mediators upstream and downstream of the Erk$\frac{1}{2}$ and p38 kinases, [3] the OGT inhibitor OSMI-1 strongly and specifically reduces O-GlcNAcylation in cardiac myocytes and promotes a robust induction of p38 phosphorylation at early time points, and [4] prolonged exposure to OSMI-1 prevents PE-induced Erk$\frac{1}{2}$ phosphorylation, exacerbates p38 phosphorylation, and blunts the PE-induced growth of cardiomyocytes. Focusing specifically on the activation of p38 by OSMI-1, our findings suggest three potential mechanisms that are not mutually exclusive: (i) activation of a novel signaling axis involving NOX2/Ask1 that activates the canonical MAP2Ks MKK$\frac{3}{6}$ to phosphorylate and activate p38, (ii) activation of the noncanonical p38 autoactivation pathway that requires the scaffold protein Tab1, and (iii) a direct O-GlcNAcylation of p38 that could affect its phosphorylation.
Previously, it was found that cardiac myocytes exposed to high glucose underwent increased O-GlcNAcylation, associated with increased phosphorylation of Erk$\frac{1}{2}$ [78]. A positive correlation between O-GlcNAc levels and Erk$\frac{1}{2}$ phosphorylation was also noted in gastric cancer cells and hippocampal slices [79, 80]. Our findings on Erk$\frac{1}{2}$ phosphorylation during global O-GlcNAc manipulation suggest a more nuanced relationship. Firstly, we found that elevating global O-GlcNAcylation with TMG does not impact basal and PE-induced Erk$\frac{1}{2}$ phosphorylation and the same is true for a 6-h treatment with OSMI-1. However, we found the PE-induced Erk$\frac{1}{2}$ phosphorylation to be significantly blunted in cells exposed to OSMI-1 for 24 h. A search of proteomics compendia [81] yields hits for O-GlcNAcylation on Erk$\frac{1}{2}$ and the upstream kinases MEK$\frac{1}{2.}$ In agreement with this bioinformatic information, we have provided evidence with metabolic labeling, pull-down, and direct Western blotting, that subsets of both Erk$\frac{1}{2}$ and MEK$\frac{1}{2}$ are O-GlcNAcylated. It remains unknown, however, how the loss of O-GlcNAcylation on these targets might lead to reduced Erk$\frac{1}{2}$ phosphorylation. One possibility is that O-GlcNAcylation prevents the phosphorylation on inhibitory T394 of MEK2 leading to more activation of Erk$\frac{1}{2}$ [82]. Regardless of the underlying mechanism, OSMI-1–induced prevention of Erk$\frac{1}{2}$ phosphorylation is likely to play a major role in the blunted hypertrophic response we observe, because Erk$\frac{1}{2}$ is the primary signaling MAPK that drives cardiomyocyte growth during PE stimulation.
Concerning the other MAPK of interest, p38, we found that OSMI-1 was a potent inducer of its phosphorylation and activity, both at early and late time points (6 h and 24 h). Importantly, OSMI-1 induced the phosphorylation of p38 in basally treated cells and it also had an additive effect on p38 phosphorylation when cells were treated with PE. Thus, decreasing O-GlcNAcylation in cardiomyocytes correlates with increased p38 phosphorylation. This is in contrast with previous work in leukocytes and glomerular mesangial cells, where it was found that increasing O-GlcNAcylation (high glucose, treatment with increased glucosamine, PUGNAc, or TMG) potentiates rather than decreasing the phosphorylation of p38 [83, 84]. These studies examined time points from 90 min up to 24 h of O-GlcNAc manipulation, which are similar to our treatments. However, because OSMI-1 was not used in these studies, it is difficult to make direct comparisons as to what might be accounting for the different outcomes in these different cell types. As for the upstream mechanisms driving p38 phosphorylation with increasing O-GlcNAcylation, the first study found increased activity of MKK$\frac{3}{6}$ [84], while the other found evidence for Ask1’s participation in p38 phosphorylation [83]. In our work with cardiomyocytes, although the directionality of the manipulation is opposite (i.e., reducing O-GlcNAcylation increases p38 phosphorylation), we found that the same pathway was involved (i.e., Ask1/MKK$\frac{3}{6}$ leading to p38 phosphorylation).
Regulation of p38 phosphorylation by Ask1 in the context of ROS has been recently demonstrated in cardiac myocytes and hearts [85]. Upstream activators of Ask1 are the ROS-generating NOXs and ablation of p47phox, the regulatory subunit of NOX2, ameliorates Ask1 phosphorylation and cardiac damage due to angiotensin-II infusion [86]. Furthermore, excessively high levels of O-GlcNAc in cardiomyocytes (e.g., due to hyperglycemia) lead to ROS production by NOX2 in cardiomyocytes [87, 88] and vascular smooth muscle cells [89]. Along these lines, we investigated the implications of NOX2 in OSMI-1–induced p38 phosphorylation and found that p47phox and gp91phox, both subunits of NOX2 are required for p38 phosphorylation in OSMI-1–treated cells. There is currently a scarcity of studies identifying O-GlcNAcylation on one or more NOX2 subunits (gp91phox, p67phox, p47phox, p40phox, p22phox, or Rac$\frac{1}{2}$), so it remains to be determined how in our model NOX2 senses reductions in O-GlcNAc levels to activate the downstream pathway leading to p38 phosphorylation.
In addition to the canonical MKK$\frac{3}{6}$ leading to p38 activation during OGT inhibition, we identified a second mechanism of p38 activation that involves Tab1. Tab1 serves as a scaffold protein that recruits p38 to induce its autoactivation [75]. This mode of activation has been described in many cell types including cardiac myocytes where it modulates p38 activity [90]. Here, we found that the knockdown of Tab1 reduced OSMI-1–induced phosphorylation of p38. Notably, Tab1 is a known substrate of OGT [76, 91] and in agreement, we find here that a subset of Tab1 is O-GlcNAcylated in cardiac myocytes. Previously, it has been reported that noncanonical p38 activation is under negative regulation of Hsp90/Cdc37 in cardiac myocytes [74]. Consistently, we found that inhibiting this complex with geldanamycin increases p38’s phosphorylation, which is additive to that induced by OSMI-1, suggesting that the activating effect of OGT inhibition occurs downstream of the release of p38 from its inhibitory complex. Interestingly, we find that like Tab1, Hsp90 also appears to be O-GlcNAcylated in cardiomyocytes. In a potential scenario, OSMI-1 inhibition of OGT could lead to a net loss of O-GlcNAc from Hsp90 and Tab1, which would then facilitate the release from the former and increased interaction with the latter leading to activation of p38. It is noteworthy that the minimum region of Tab1 required for p38 autoactivation, the 46-mer 371 to 416 [92], encompasses a well-known O-GlcNAcylation site at S395 [76]. However, whether O-GlcNAcylation on S395 is critical for p38 activation during stress conditions has yet to be determined and a knock-in mouse with a S395A substitution exhibits a normal baseline phenotype [93].
Results from OGT assays suggest that p38 can be O-GlcNAcylated in vitro [94]. In agreement, using the metabolic labeling approach, coupled with pull-down and Western blotting, we found that p38 was O-GlcNAcylated in a manner sensitive to OSMI-1. There is currently a paucity of known O-GlcNAcylation sites on p38 and consequently, it is difficult to rationalize how the presence or absence of O-GlcNAc on such sites might impact p38 activity. In the simplest scenario, it could be that O-GlcNAcylation on T180 within p38’s activation loop directly hinders its phosphorylation either by MKK$\frac{3}{6}$ or during autoactivation. Furthermore, it was found that a highly conserved threonine within the activation loop, T185, is essential for the Tab1-mediated autoactivation of p38 [95]. Whether O-GlcNAcylation on the aforementioned, or other unknown sites, modulates the activation of p38 warrants further investigation. Finally, it is worth mentioning that enhancing O-GlcNAcylation by glucosamine perfusion blunts the phosphorylation of p38 during cardiac ischemia but increases it during reperfusion, further underscoring a tight but complex relationship between O-GlcNAcylation and p38 phosphorylation in the intact heart [11].
The activity of Erk$\frac{1}{2}$ in cardiomyocytes and hearts is generally associated with physiological responses while that of p38 is associated with pathophysiological outcomes [96]. In the context of OGT inhibition for 24 h, it could be that the simultaneous disruption of Erk$\frac{1}{2}$ and p38 signaling contributes to the impaired hypertrophic response to PE. After treating cells with OSMI-1 for 24 h, we found decreased phosphorylation of Erk$\frac{1}{2}$ and increased phosphorylation of p38. Erk$\frac{1}{2}$ signaling induces prohypertrophic protein synthesis [97], and inhibiting Erk$\frac{1}{2}$ blunts the adaptive cardiac hypertrophy [98]. Consistently, in our experimental model, we find that direct Erk$\frac{1}{2}$ inhibition prevents PE-induced hypertrophic growth, and similarly, reduced Erk$\frac{1}{2}$ phosphorylation in the context of OGT inhibition correlates with a blunted PE-induced myocyte hypertrophy. Along these lines, the PE-induced expression of cTnT, a gene encoding for a key component of myofilaments, is downregulated by Erk$\frac{1}{2}$ inhibition, and this response is also recapitulated during OGT inhibition. Compared to that of Erk$\frac{1}{2}$, the role of p38 in the context of PE-induced hypertrophy is limited [99] and consistently, the p38 inhibitor alone did not significantly block the PE-induced hypertrophic growth. Overactive p38 signaling in the heart induces a fetal gene expression signature, including increased Anp expression [100]. While we found that the induction of Anp downstream of PE stimulation is largely driven by Erk$\frac{1}{2}$, it is still possible that overactive p38 compensates for the decreased Erk$\frac{1}{2}$ signaling and contributes to the uniformly high levels of Anp expression in the context of PE stimulation and OGT inhibition. Consistently, both Erk$\frac{1}{2}$ and p38 are reported to have a convergent role in the regulation of promoters controlling the expression of natriuretic peptides [101].
Cardiomyocyte-specific deletion of OGT leads to early postnatal lethality due to developmental defects [102, 103], highlighting OGT’s role in cardiomyocyte growth and maturation. In agreement, our work with primary neonatal cardiomyocytes found that OGT inhibition blunts hypertrophic growth induced by PE. Consistent with an impaired growth response, OSMI-1 did not reverse the PE-induced expression of pathologic markers Anp and Bnp and it downregulated PE-induced cTnT expression. Together with the knockout results, our findings underscore the important role of OGT in developmental cardiomyocyte hypertrophy. In the adult heart, O-GlcNAcylation is upregulated during the early response to pressure overload, indicating its importance in adaptive hypertrophy [104, 105]. Consistently, hearts with adult-onset OGT deficiency exhibit systolic dysfunction after pressure overload, although this is not accompanied by significant changes in cardiomyocyte size and overall cardiac mass [14, 15]. Collectively, it appears that OGT activity promotes cardiomyocyte growth in postnatally developing hearts and maintains cardiomyocyte function/contractility during early pressure-overload hypertrophy in adult hearts. Interestingly, a balanced cardiomyocyte OGT activity needs to be kept, because mice with cardiac overexpression of OGT throughout embryonic and postnatal development exhibit cardiomyopathy by early adulthood [21].
The underlying mechanisms leading to cardiac dysfunction in hearts with altered OGT expression either at baseline or after challenge with pressure overload are incompletely understood and no immediate connections have been made for the implications of MAPK signaling. However, one study found that the transcription factor Gata4 was strongly downregulated in OGT-deficient hearts challenged with pressure overload [14]. Because Gata4 is a downstream target of Erk$\frac{1}{2}$ signaling, its downregulation in OGT-deficient heart might indicate a dysregulation of Erk$\frac{1}{2}$ signaling. Erk$\frac{1}{2}$ signals through Gata4 to mediate adaptive hypertrophy in cardiac myocytes [106, 107] and consistently, our findings here show that OGT inhibition perturbs Erk$\frac{1}{2}$ phosphorylation concomitantly with blunted hypertrophy. Interestingly, we find that Gata4 is O-GlcNAcylated in agreement with previous findings in the heart [108]. While O-GlcNAcylation appears to promote the transcriptional activity of Gata4 [108], it is unknown whether reduced O-GlcNAcylation can lead to its downregulation. Ultimately, it would be interesting in future studies to address to what extent the gain of Gata4 could reverse any of the phenotypes of OGT inhibition observed here.
In this study, we have used inhibitors extensively, which offers the advantage of acute target manipulation but also has disadvantages such as off-target effects or unexpected activities [109]. Compound SB202190 (an ATP-binding pocket interactor) exhibits high selectivity for p38α and p38β [61], however, it was found to interact with 12 other targets from a panel of 119 protein kinases [110]. Inhibitor SCH772984 (binding to an allosteric pocket) exhibits high selectivity for Erk1 and Erk2 against a panel of 456 kinases but can also interact with 12 other kinases with greater than $90\%$ binding affinity [111]. Inhibitor GS-444217 (interacting with the ATP-binding pocket and thus serving as a competitive inhibitor) exhibits high selectivity for Ask1 with only two other potentially interacting kinases across a panel of 442 kinases [112]. Inhibitor takinib (ATP-binding domain interactor) exhibits high selectivity for Tak1, targeting only five other kinases across a panel of 140 kinases albeit with substantially higher IC50s [113]. On the other hand, compound URMC-099 (a likely ATP-binding pocket interactor) exhibited only moderate selectivity for Mlk3 as it interacted with 111 other kinases across a panel of 442 [114]. Finally, geldanamycin, an Hsp90-binding macrocyclic, was found to affect the expression of 288 kinases [115]. Taken together, most of these compounds are highly selective for their intended kinases, although their potential targeting of other proteins (kinases or not) necessitates the use of caution when interpreting their effects on p38 and Erk$\frac{1}{2}$ signaling pathways examined here.
TMG was tested against human lysosomal β-hexosaminidase and five other glycoside hydrolases and was found to be highly selective for OGA with a Ki value of 21 nM and an IC50 in the nanomolar range [70]. Consistently, in our experiments, we found that while significantly increasing O-GlcNAcylation, TMG did not impact the abundance of cell surface N- or O-linked glycans. OSMI-1 was developed from a quinoline-6-sulfonamide scaffold and was found to target OGT with an IC50 of 2.7 μM and dose dependently decreased O-GlcNAcylation in cells [69]. Due to the lack of glycosyltransferase panels (akin to kinase panels), screening for off-target binding of OSMI-1 to other glycosyltransferases was not possible. Nevertheless, screening with lectin blots did not identify a gross impact of OSMI-1 on the abundance of various N- and O-linked glycans [69]. Similarly, lectin blots with cardiomyocyte extracts performed here did not reveal a significant impact of OSMI-1 on cell surface N- and O-linked glycans. It should be noted that a close structural analog of OSMI-1, PG34, had a sizable negative effect on cell viability after 24 h of exposure despite being a weak OGT inhibitor (IC50 = 68 μM), indicating that OSMI-1 might potentially have off-target effects impacting cell viability [69]. In light of this caveat and while caution should be used in interpreting the results, it should be noted that the majority of the experiments done here involved acute use of OSMI-1 for up to 6 h at an intermediate concentration (25 μM). Furthermore, although not designed to test cell viability per se, our experiments assessing myocyte hypertrophy found that baseline (non-PE–treated) cells exposed to OSMI-1 for 24 h did not have lower nuclear counts.
In summary, we have uncovered a functional relationship between O-GlcNAcylation and MAPK signaling in cardiomyocytes and have identified several potential entry points for regulation. While we provide evidence for O-GlcNAcylation on a number of proteins within the MAPK signaling pathway (e.g., MEK$\frac{1}{2}$, Erk$\frac{1}{2}$, Tab1, p38, Hsp27, and others), it remains an open question whether O-GlcNAcylation on this handful of proteins or on other substrates constitutes the key driving mechanism behind the observed impairment of PE-induced growth of cardiomyocytes. Therefore, further investigation is warranted to identify the most relevant substrate proteins and sites. The work presented here paves the way for such target-focused studies that can ultimately provide new approaches for intervening in the MAPK pathway in cardiac pathophysiology.
## Materials
A detailed list of key reagents including their identifiers, vendors, and catalog numbers is provided in Table S1. Additional information on other reagents can also be found in the description of methods in the text below. The primer sequences used in quantitative real-time PCR are shown in Table S2. Procedures involving the use of animals were approved by the Institutional Animal Care and Use Committees (IACUC) at the Johns Hopkins School of Medicine.
## Primary neonatal myocyte culture and treatments
NRVMs were isolated from P0-P1 rats of either sex using enzymatic dissociation, according to a standard methodology as previously described [116, 117]. Following a preplating step, to remove fibroblasts and other nonmyocytes, the cardiomyocytes (suspensions at 0.5 × 106 cells/ml in Dulbecco’s modified Eagle’s medium supplemented with $10\%$ FBS) were seeded for experiments on culture plates precoated with $0.1\%$ bovine gelatin (Cat. No. G9391, Sigma) and incubated at 37 °C, $5\%$ CO2 for 24 h. Typical seeding densities were 1.0 × 106 cells per well in a 6-well plate for Western blot experiments or 1.25 × 105 cells per chamber in ibidi μ-Slide (Cat. No. 80426) for confocal microscopy. On the next day, the cells were washed twice with prewarmed PBS and the medium was switched to $0\%$ FBS Dulbecco’s modified Eagle’s medium (Cat. No. 10-013-CV, Corning, 4.5 g/l glucose) supplemented with insulin transferrin selenium (Cat. No. 51500056, Gibco). Following 24 h of serum starvation, the cells were stimulated with the indicated agonists and/or inhibitors for the specified durations for each experiment (see also main text and/or figure legends). In experiments for gene knockdown, cells were transfected with a lipofectamine/siRNA mix at the time of seeding. Briefly, dicer-substrate short interfering RNA was complexed with Lipofectamine RNAiMax (Cat. No. 13778075, Thermo Fisher Scientific) at room temperature for 15 min according to manufacturer’s specifications and the transfection mix was added to the cell suspension at a ratio of 40 pmol siRNA/1 × 106 cells.
## Protein extraction and immunoblotting
Cells were washed with ice-cold PBS (3×) and lysed in RIPA buffer (contains 150 mM NaCl, $1.0\%$ IGEPAL CA-630, $0.5\%$ sodium deoxycholate, $0.1\%$ SDS, 50 mM Tris, pH 8.0, Cat. No. R0278, Sigma) supplemented with 2 μM TMG, 1 mM PMSF, and 1× protease and phosphatase inhibitors (cOmplete, mini, EDTA-free, and PhosSTOP respectively, Millipore, Sigma). Lysates were sonicated and protein concentration was quantified with the bicinchoninic acid assay (Cat. No. 23225, Pierce). For Western blot, protein samples were combined with 50 mM DTT and 1× LDS sample buffer (NP0008, NuPAGE, Thermo Fisher Scientific), and 10 μg total protein per lane was resolved by electrophoresis, typically run in NuPAGE Bis-Tris 4 to $12\%$ gradient gels (Cat. No. WG1403, Thermo Fisher Scientific). Proteins were transferred onto nitrocellulose membranes (iBlot2 transfer stacks, Cat. No. IB23001, Thermo Fisher Scientific). The membranes were blocked with $5\%$ bovine serum albumin (BSA) in TBS for 1 h at room temperature and then incubated with primary antibodies, typically at a 1000:1 dilution in $5\%$ BSA in $0.1\%$ Tween20-TBS, overnight at 4 °C (see Table S2 for a list of primary antibodies used). For detection, we used the LI-COR Odyssey system and appropriate host-specific secondary antibodies (e.g., IR Dye 800w goat anti-rabbit, Cat. No. 926-32211, LI-COR) typically at a 5000:1 dilution in $5\%$ BSA in $0.1\%$ Tween20-TBS. For the detection of biotinylated targets, we used IRDye800w Streptavidin (Cat. No. 926-32230, 5000:1 in % BSA in $0.1\%$ Tween20-TBS). Quantification of band intensities was performed with Image Studio Lite (LI-COR). Normalization for protein loading was performed based on intensities obtained by staining the membranes with REVERT 700 Total Protein Stain (Cat. No. 926-11010, LI-COR).
For the detection of glycans with lectin blotting, we used lectin kit 1 (Cat. No. BK-1000, Vector Laboratories) that includes biotinylated conjugates of wheat germ agglutinin, ConA, peanut agglutinin, and D. biflorus agglutinin. For the lectiblots, proteins were run on NuPAGE Bis-Tris 4 to $12\%$ gradient gels and then were transferred using the TransBlot Turbo system (Bio-Rad) onto nitrocellulose membranes (kit Cat. No. 1704271, Bio-Rad). Following blocking with $5\%$ BSA in TBS, membranes were incubated with 5000:1 diluted lectins (final concentration of lectin 0.4 μg/ml) overnight at 4 °C. Following extensive washes with TBS-T, the bound biotinylated lectins were detected with IRDye800w Streptavidin. To control for the reactivity of streptavidin with endogenously biotinylated proteins, some experiments were done as negative controls where no lectin was included in the overnight step. These experiments yielded reactive bands at ∼75 and also at 125 and 250 kDa.
## Metabolic labeling with unnatural sugar analogs, alkyne-azide click reaction, and enrichment of putative O-GlcNAcylated proteins
The procedure was developed based on protocols described previously [65, 66, 67]. Briefly, for the metabolic labeling of glycans with clickable handles, NRVM or HEK293 cells were incubated in growth medium supplemented with 200 μM Ac4GalNAz or 200 μM Ac4GalNAlk for 24 h. Following metabolic labeling, the cells were washed in ice-cold PBS (3×), lysed in a buffer containing $1\%$ SDS, 2 μM TMG, protease and phosphatase inhibitors, 50 mM Tris HCl pH 8.0, and sonicated as described above. In some experiments, after 18 h of treatment with the metabolic chemical reporter, the cells were treated with vehicle, TMG (200 nM), or OSMI-1 (25 μM) for an additional 6 h in the presence of the metabolic reporter. For the labeling of glycosylated proteins with CalFluor 647 azide, the copper-catalyzed alkyne-azide reaction contained 20 μg protein suspension in $0.6\%$ SDS, 2 mM sodium ascorbate, 100 μM THPTA, 1 mM CuSO4.5H2O, and either 20 μM or 50 μM CalFluor 647 azide (Cat. No. 1372, Click Chemistry Tools). As a negative control for the click reaction, a set of reactions was set up as above in the presence of 50 μM CalFluor 647 azide, omitting the catalyst Cu/THPTA. The ‘click’ reaction took place for 1 h at room temperature in the dark and was quenched with 5 mM EDTA. Then, samples were mixed with 1× LDS sample buffer and 50 mM DTT and were resolved on a 4 to $12\%$ Bis-Tris gradient gel which was subsequently scanned on the LI-COR Odyssey system using the 700 channel (685 nm infrared laser).
In other experiments, lysates from Ac4GalNAz- or Ac4GalNAlk-treated cells were used in copper-catalyzed alkyne-azide reactions containing 500 μg protein suspension in $0.6\%$ SDS, 2 mM sodium ascorbate, 100 μM THPTA, 1 mM CuSO4.5H2O, and 40 μM biotin azide plus (Cat. No. 1488, Click Chemistry Tools). Following ‘click’ reaction and quenching, the reaction mix was cleared by methanol/chloroform precipitation as described above and proteins resuspended in $1.0\%$ SDS. A fraction of this suspension was taken as ‘input’. Next, high-capacity streptavidin-agarose slurry (Cat. No. 20357, Pierce) corresponding to 100 μl settled resin was placed into spin columns (Cat. No. 89868, Pierce) and washed twice in water and then twice in 50 mM Tris HCl, pH 8.0. Washes were done by spinning at 1500g for 1 min and discarding the flow-through. The protein suspension was then diluted 5-fold in 50 mM Tris HCl, pH 8.0 ($0.2\%$ SDS final concentration), added to the washed streptavidin resin bed, and allowed to bind for 2 h at room temperature with end-over-end rotation. Afterward, the suspension was spun to obtain the ‘unbound’ fraction, while the agarose beads were washed 5× with IP wash buffer ($1\%$ NP-40, 150 mM NaCl, 50 mM Hepes, pH 7.9). To elute biotinylated proteins from the beads, we used a solution containing $80\%$ acetonitrile, $0.1\%$ formic acid, and $0.2\%$ TFA and heating at 65 °C for 5 min [118]. The released biotinylated proteins were separated from the beads by spinning at 1500g for 1 min and the eluted fractions were dried down by speed-vac centrifugation. The dried protein pellets were resuspended in 1× LDS/50 mM DTT sample buffer and run next to ‘input’ and ‘unbound’ fractions by gel electrophoresis and Western blot. Detection of eluted proteins was performed with IRDye800w streptavidin for total biotinylated proteins or with appropriate antibodies for specific intracellular proteins of interest.
## RNA isolation, reverse transcription, and gene expression analysis by quantitative real-time PCR
Total RNA was extracted from NRVMs seeded at 1.0 × 106 cells/well after appropriate treatments as indicated in the main text (e.g., 24-h PE treatment ± TMG/OSMI-1 treatment). The treated cells were washed with ice-cold PBS (3×) and RNA was extracted with the RNeasy kit (Cat. No. 74704, Qiagen), including a proteinase K digestion step to remove abundant sarcomeric proteins. In-column DNaseI treatment (Cat. No. 79254, Qiagen) was used to digest away genomic DNA. For reverse transcription, we used a QuantiTect reverse transcription kit (Cat. No. 205311, Qiagen) and an input of 850 ng of total RNA. To control for potentially residual genomic DNA carry-over, reactions without reverse-transcriptase were set up in parallel. Real-time quantitative PCR was performed with the Power SYBR Green master mix (Cat. No. 4367659, Thermo Fisher Scientific) in the presence of 500 nM forward and reverse primers (for gene-specific sequences see Table S2). A total of 45 cycles of amplification were done (15 s at 95 °C for melting followed by 30 s at 60 °C for annealing and elongation) using the CFX384 Touch system (Bio-Rad). Each sample was run in technical triplicates. Gene expression was quantified with the ΔCt method using Gapdh and 36B4 as housekeeping genes.
## Cell processing for F-actin staining, imaging, and quantitation of NRVM hypertrophy
Following seeding and treatments in 4-chamber imaging slides, NRVMs were washed with 37 °C prewarmed PBS (3×) and then fixed with prewarmed $4\%$ paraformaldehyde for 15 min. The fixative was then washed off with PBS and cells were permeabilized with $0.1\%$ Triton X-100 for 5 min at room temperature. Permeabilization was followed by PBS washes and finally blocking overnight at 4 °C with $1\%$ BSA in $0.1\%$ tween-20 PBS. To stain F-actin, we exposed cells to 5 units/ml phalloidin-Alexa 594 (Cat. No. A12381, Thermo Fisher Scientific) diluted in $1\%$ BSA, $0.1\%$ Tween-20, PBS for 1 h at room temperature. An additional 15-min staining step was done with 5000-fold diluted Hoechst 33342 (Cat. No. H3570, Thermo Fisher Scientific) to stain nuclei, and then cells were washed with PBS. Imaging was carried out with the FV3000RS confocal microscope (Olympus) using the 10× objective (UPLSAPO10X2, numeric aperture 0.40, working distance 3.1 mm) and 405 nm and 561 nm lasers to excite Hoechst and Alexa 594, respectively. Laser power and gain were kept constant across imaging sessions and four fields of view were captured per chamber.
For data analysis,.oir images were uploaded into a Fiji-run custom-made macro to convert composite images into individual.tiff single-color channels (blue and red). The images were then uploaded into Cell Profiler (latest version 4.2.1., https://cellprofiler.org/). The Cell Profiler pipeline extracts nuclei from the blue channel and uses those as primary objects. Next, using nuclei as primary objects, the pipeline identifies the cells from the red channel (phalloidin) using the ‘Propagation’ method. Next, the cytoplasm is identified by subtracting the nuclear area from the cell area. The main output parameters of the pipeline utilized for quantification of the F-actin signal were the ‘Integrated Cytoplasmic Intensity’ measured as the intensity of the phalloidin signal in the cytoplasmic area. The mean integrated cytoplasmic intensity across all cells within a given field of view was used as a single data point. Additionally, the ‘Nuclear Counts’ measured as the total number of nuclei in a given field of view were used as a secondary parameter of overall cell viability/representation in a given field of view to confirm that sampling was performed from areas with comparable cell density.
## Statistical analysis
Data were handled and plotted using Microsoft Excel and GraphPad Prism. Values in the graphs are reported as means ± standard error. The numbers of biological replicates are shown inscribed in their respective bars. Data were examined for normality of distribution using the Shapiro-Wilk test and then visually inspected (e.g., residual plot and Q-Q plot) for any potential outliers. If necessary, one or more outliers were identified using the ROUT test ($Q = 1$%). To identify statistically significant differences between two groups, we performed unpaired student’s t test. When comparing three or more groups, we used one-way or two-way ANOVA depending on the number of factors (e.g., for comparing the effect across three or more drugs, we used one-way, whereas for comparing groups resulting from a combination of drugs and gene manipulation, we used two-way). To identify statistically significant differences, the data were assessed for normality of distribution. If the majority of groups within an experiment had normally distributed data, we used the Tukey post hoc test. If the distribution of the data in the majority of groups was not normal, we used the Dunnet post hoc test. In all cases, the α level for statistical significance was 0.05. p values less than 0.05, 0.01, 0.001, and 0.0001 are symbolized as ∗, ∗∗, ∗∗∗, and ∗∗∗∗ respectively.
## Data availability
All data presented are included in the article and can be available upon request from Kyriakos N. Papanicolaou, Johns Hopkins University School of Medicine, [email protected].
## Supporting information
The article contains supporting information.
Supporting Figures S1–S4 and Tables S1 and S2
## Conflict of interest
The authors declare no competing interests.
## Author contributions
K. N. P. and B. O. R. conceptualization; K. N. P., J. J., D. A., W. Z., A. M., E. A., D. B. F., N. E. Z., and B. O. R. methodology; K. N. P., D. B. F., N. E. Z., and B. O. R. supervision; K. N. P. writing–original draft; K. N. P., D. B. F., N. E. Z., and B. O. R. writing–review and editing; K. N. P., J. J., D. A., W. Z., A. M., and E. A. investigation; N. E. Z. and B. O. R. resources; K. N. P. and D. A. data curation; K. N. P., J. J., D. A., W. Z., A. M., and E. A. formal analysis; D. B. F., N. E. Z., and B. O. R. funding acquisition.
## Funding and additional information
This work was supported by K12 fellowship HL141952 to K. N. P. and N. E. Z. and $\frac{10.13039}{100000968}$American Heart Association Career Development Award 935823 to K. N. P. D. A. acknowledges support by $\frac{10.13039}{100000002}$NIH training grant T32HL007227. Work in the lab of N. E. Z. is supported by R01HL139640 and U01CA230978 grants. Work in the lab of D. B. F. is supported by an $\frac{10.13039}{100000968}$AHA Transformational Project Award (18TPA34170575) and R01HL134821. Work in the lab of B. O. R. is supported by R01HL134821. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
## References
1. Hart G.W., Housley M.P., Slawson C.. **Cycling of O-linked β-N-acetylglucosamine on nucleocytoplasmic proteins**. *Nature* (2007) **446** 1017-1022. PMID: 17460662
2. Bond M.R., Hanover J.A.. **A little sugar goes a long way: the cell biology of O-GlcNAc**. *J. Cell Biol.* (2015) **208** 869-880. PMID: 25825515
3. Levine Z.G., Walker S.. **The biochemistry of O-GlcNAc transferase: which functions make it essential in mammalian cells?**. *Annu. Rev. Biochem.* (2016) **85** 631-657. PMID: 27294441
4. Toleman C., Paterson A.J., Whisenhunt T.R., Kudlow J.E.. **Characterization of the histone acetyltransferase (HAT) domain of a bifunctional protein with activable O-GlcNAcase and HAT activities**. *J. Biol. Chem.* (2004) **279** 53665-53673. PMID: 15485860
5. Gao Y., Wells L., Comer F.I., Parker G.J., Hart G.W.. **Dynamic O-glycosylation of nuclear and cytosolic proteins: cloning and characterization of a neutral, cytosolic β-N-acetylglucosaminidase from human brain**. *J. Biol. Chem.* (2001) **276** 9838-9845. PMID: 11148210
6. Chatham J.C., Nöt L.G., Fülöp N., Marchase R.B.. **Hexosamine biosynthesis and protein O-glycosylation: the first line of defense against stress, ischemia, and trauma**. *Shock* (2008) **29** 431-440. PMID: 17909453
7. Groves J.A., Lee A., Yildirir G., Zachara N.E.. **Dynamic O-GlcNAcylation and its roles in the cellular stress response and homeostasis**. *Cell Stress Chaperones* (2013) **18** 535-558. PMID: 23620203
8. Shafi R., Iyer S.P.N., Ellies L.G., O'Donnell N., Marek K.W., Chui D.. **The O-GlcNAc transferase gene resides on the X chromosome and is essential for embryonic stem cell viability and mouse ontogeny**. *Proc. Natl. Acad. Sci. U. S. A.* (2000) **97** 5735-5739. PMID: 10801981
9. Yang Y.R., Song M., Lee H., Jeon Y., Choi E.J., Jang H.J.. **O-GlcNAcase is essential for embryonic development and maintenance of genomic stability**. *Aging Cell* (2012) **11** 439-448. PMID: 22314054
10. Keembiyehetty C., Love D.C., Harwood K.R., Gavrilova O., Comly M.E., Hanover J.A.. **Conditional knock-out reveals a requirement for O-linked N-acetylglucosaminase (O-GlcNAcase) in metabolic homeostasis**. *J. Biol. Chem.* (2015) **290** 7097-7113. PMID: 25596529
11. Fülöp N., Zhang Z., Marchase R.B., Chatham J.C.. **Glucosamine cardioprotection in perfused rat heart associated with increased O-Linked N-acetylglucosamine protein modification and altered p38 activation**. *Am. J. Physiol. Heart Circ. Physiol.* (2007) **292** H2227-H2236. PMID: 17208994
12. Jones S.P., Zachara N.E., Ngoh G.A., Hill B.G., Teshima Y., Bhatnagar A.. **Cardioprotection by N-acetylglucosamine linkage to cellular proteins**. *Circulation* (2008) **117** 1172. PMID: 18285568
13. Lunde I.G., Aronsen J.M., Kvaløy H., Qvigstad E., Sjaastad I., Tønnessen T.. **Cardiac O-GlcNAc signaling is increased in hypertrophy and heart failure**. *Physiol. Genomics* (2011) **44** 162-172. PMID: 22128088
14. Dassanayaka S., Brainard R.E., Watson L.J., Long B.W., Brittian K.R., DeMartino A.M.. **Cardiomyocyte Ogt limits ventricular dysfunction in mice following pressure overload without affecting hypertrophy**. *Basic Res. Cardiol.* (2017) **112** 23. PMID: 28299467
15. Zhu W.Z., El-Nachef D., Yang X., Ledee D., Olson A.K.. **O-GlcNAc transferase promotes compensated cardiac function and protein kinase AO-GlcNAcylation during early and established pathological hypertrophy from pressure overload**. *J. Am. Heart Assoc.* (2019) **8**
16. Dassanayaka S., Brittian K.R., Long B.W., Higgins L.A., Bradley J.A., Audam T.N.. **Cardiomyocyte Oga haploinsufficiency increases O-GlcNAcylation but hastens ventricular dysfunction following myocardial infarction**. *PLoS One* (2020) **15**
17. Fricovsky E.S., Suarez J., Ihm S.-H., Scott B.T., Suarez-Ramirez J.A., Banerjee I.. **Excess protein O-GlcNAcylation and the progression of diabetic cardiomyopathy**. *Am. J. Physiol. Heart Circ. Physiol.* (2012) **303** R689-R699
18. Prakoso D., Lim S.Y., Erickson J.R., Wallace R.S., Lees J.G., Tate M.. **Fine-tuning the cardiac O-GlcNAcylation regulatory enzymes governs the functional and structural phenotype of the diabetic heart**. *Cardiovasc. Res.* (2022) **118** 212-225. PMID: 33576380
19. Mesubi O.O., Rokita A.G., Abrol N., Wu Y., Chen B., Wang Q.. **Oxidized CaMKII and O-GlcNAcylation cause increased atrial fibrillation in diabetic mice by distinct mechanisms**. *J. Clin. Invest.* (2021) **131**
20. Watson L.J., Facundo H.T., Ngoh G.A., Ameen M., Brainard R.E., Lemma K.M.. **O-linked β-N-acetylglucosamine transferase is indispensable in the failing heart**. *Proc. Natl. Acad. Sci. U. S. A.* (2010) **107** 17797-17802. PMID: 20876116
21. Umapathi P., Mesubi O.O., Banerjee P.S., Abrol N., Wang Q., Luczak E.D.. **Excessive O-GlcNAcylation causes heart failure and sudden death**. *Circulation* (2021) **143** 1687-1703. PMID: 33593071
22. Dassanayaka S., Jones S.P.. **O-GlcNAc and the cardiovascular system**. *Pharmacol. Ther.* (2014) **142** 62-71. PMID: 24287310
23. Mailleux F., Gélinas R., Beauloye C., Horman S., Bertrand L.. **O-GlcNAcylation, enemy or ally during cardiac hypertrophy development?**. *Biochim. Biophys. Acta* (2016) **1862** 2232-2243. PMID: 27544701
24. Wright J.N., Collins H.E., Wende A.R., Chatham J.C.. **O-GlcNAcylation and cardiovascular disease**. *Biochem. Soc. Trans.* (2017) **45** 545-553. PMID: 28408494
25. Ducheix S., Magré J., Cariou B., Prieur X.. **Chronic O-GlcNAcylation and diabetic cardiomyopathy: the bitterness of glucose**. *Front. Endocrinol.* (2018) **9** 642
26. Hart G.W., Slawson C., Ramirez-Correa G., Lagerlof O.. **Cross talk between O-GlcNAcylation and phosphorylation: roles in signaling, transcription, and chronic disease**. *Annu. Rev. Biochem.* (2011) **80** 825-858. PMID: 21391816
27. Erickson J.R., Pereira L., Wang L., Han G., Ferguson A., Dao K.. **Diabetic hyperglycaemia activates CaMKII and arrhythmias by O-linked glycosylation**. *Nature* (2013) **502** 372. PMID: 24077098
28. Gélinas R., Mailleux F., Dontaine J., Bultot L., Demeulder B., Ginion A.. **AMPK activation counteracts cardiac hypertrophy by reducing O-GlcNAcylation**. *Nat. Commun.* (2018) **9** 374. PMID: 29371602
29. Tran D.H., May H.I., Li Q., Luo X., Huang J., Zhang G.. **Chronic activation of hexosamine biosynthesis in the heart triggers pathological cardiac remodeling**. *Nat. Commun.* (2020) **11** 1-15. PMID: 31911652
30. Heineke J., Molkentin J.D.. **Regulation of cardiac hypertrophy by intracellular signalling pathways**. *Nat. Rev. Mol. Cell Biol.* (2006) **7** 589-600. PMID: 16936699
31. Wang Y.. **Mitogen-activated protein kinases in heart development and diseases**. *Circulation* (2007) **116** 1413-1423. PMID: 17875982
32. Lips D.J., Bueno O.F., Wilkins B.J., Purcell N.H., Kaiser R.A., Lorenz J.N.. **MEK1-ERK2 signaling pathway protects myocardium from ischemic injury in vivo**. *Circulation* (2004) **109** 1938-1941. PMID: 15096454
33. Maillet M., Purcell N.H., Sargent M.A., York A.J., Bueno O.F., Molkentin J.D.. **DUSP6 (MKP3) null mice show enhanced ERK1/2 phosphorylation at baseline and increased myocyte proliferation in the heart affecting disease susceptibility**. *J. Biol. Chem.* (2008) **283** 31246-31255. PMID: 18753132
34. Purcell N.H., Wilkins B.J., York A., Saba-El-Leil M.K., Meloche S., Robbins J.. **Genetic inhibition of cardiac ERK1/2 promotes stress-induced apoptosis and heart failure but has no effect on hypertrophy in vivo**. *Proc. Natl. Acad. Sci. U. S. A.* (2007) **104** 14074-14079. PMID: 17709754
35. Bueno O.F., De Windt L.J., Tymitz K.M., Witt S.A., Kimball T.R., Klevitsky R.. **The MEK1-ERK1/2 signaling pathway promotes compensated cardiac hypertrophy in transgenic mice**. *EMBO J.* (2000) **19** 6341-6350. PMID: 11101507
36. Braz J.C., Bueno O.F., De Windt L.J., Molkentin J.D.. **PKC alpha regulates the hypertrophic growth of cardiomyocytes through extracellular signal-regulated kinase1/2 (ERK1/2)**. *J. Cell Biol.* (2002) **156** 905-919. PMID: 11864993
37. Mutlak M., Kehat I.. **Extracellular signal-regulated kinases 1/2 as regulators of cardiac hypertrophy**. *Front. Pharmacol.* (2015) **6** 149. PMID: 26257652
38. Liu R., van Berlo J.H., York A.J., Vagnozzi R.J., Maillet M., Molkentin J.D.. **DUSP8 regulates cardiac ventricular remodeling by altering ERK1/2 signaling**. *Circ. Res.* (2016) **119** 249-260. PMID: 27225478
39. Molkentin J.D., Dorn G.W.. **Cytoplasmic signaling pathways that regulate cardiac hypertrophy**. *Annu. Rev. Physiol.* (2001) **63** 391-426. PMID: 11181961
40. Cargnello M., Roux P.P.. **Activation and function of the MAPKs and their substrates, the MAPK-activated protein kinases**. *Microbiol. Mol. Biol. Rev.* (2011) **75** 50-83. PMID: 21372320
41. Morimoto T., Hasegawa K., Kaburagi S., Kakita T., Wada H., Yanazume T.. **Phosphorylation of GATA-4 is involved in alpha 1-adrenergic agonist-responsive transcription of the endothelin-1 gene in cardiac myocytes**. *J. Biol. Chem.* (2000) **275** 13721-13726. PMID: 10788492
42. Babu G.J., Lalli M.J., Sussman M.A., Sadoshima J., Periasamy M.. **Phosphorylation of elk-1 by MEK/ERK pathway is necessary for c-fos gene activation during cardiac myocyte hypertrophy**. *J. Mol. Cell. Cardiol.* (2000) **32** 1447-1457. PMID: 10900171
43. Zhu J., Blenis J., Yuan J.. **Activation of PI3K/Akt and MAPK pathways regulates Myc-mediated transcription by phosphorylating and promoting the degradation of Mad1**. *Proc. Natl. Acad. Sci. U. S. A.* (2008) **105** 6584-6589. PMID: 18451027
44. Xing J., Ginty D.D., Greenberg M.E.. **Coupling of the RAS-MAPK pathway to gene activation by RSK2, a growth factor-regulated CREB kinase**. *Science* (1996) **273** 959-963. PMID: 8688081
45. Wiggin G.R., Soloaga A., Foster J.M., Murray-Tait V., Cohen P., Arthur J.S.. **MSK1 and MSK2 are required for the mitogen- and stress-induced phosphorylation of CREB and ATF1 in fibroblasts**. *Mol. Cell. Biol.* (2002) **22** 2871-2881. PMID: 11909979
46. Li M., Georgakopoulos D., Lu G., Hester L., Kass D.A., Hasday J.. **p38 MAP kinase mediates inflammatory cytokine induction in cardiomyocytes and extracellular matrix remodeling in heart**. *Circulation* (2005) **111** 2494-2502. PMID: 15867183
47. Auger-Messier M., Accornero F., Goonasekera S.A., Bueno O.F., Lorenz J.N., van Berlo J.H.. **Unrestrained p38 MAPK activation in Dusp1/4 double-null mice induces cardiomyopathy**. *Circ. Res.* (2013) **112** 48-56. PMID: 22993413
48. Braz J.C., Bueno O.F., Liang Q., Wilkins B.J., Dai Y.-S., Parsons S.. **Targeted inhibition of p38 MAPK promotes hypertrophic cardiomyopathy through upregulation of calcineurin-NFAT signaling**. *J. Clin. Invest.* (2003) **111** 1475-1486. PMID: 12750397
49. Lopez-Ilasaca M.. **Signaling from G-protein-coupled receptors to mitogen-activated protein (MAP)-kinase cascades**. *Biochem. Pharmacol.* (1998) **56** 269-277. PMID: 9744561
50. Yamauchi J., Tsujimoto G., Kaziro Y., Itoh H.. **Parallel regulation of mitogen-activated protein kinase kinase 3 (MKK3) and MKK6 in Gq-signaling cascade**. *J. Biol. Chem.* (2001) **276** 23362-23372. PMID: 11304531
51. Brown J.H., Del Re D.P., Sussman M.A.. **The Rac and Rho hall of fame: a decade of hypertrophic signaling hits**. *Circ. Res.* (2006) **98** 730-742. PMID: 16574914
52. Cuadrado A., Nebreda A.R.. **Mechanisms and functions of p38 MAPK signalling**. *Biochem. J.* (2010) **429** 403-417. PMID: 20626350
53. Kamemura K., Hayes B.K., Comer F.I., Hart G.W.. **Dynamic interplay between O-glycosylation and O-phosphorylation of nucleocytoplasmic proteins: alternative glycosylation/phosphorylation of Thr-58, a known mutational hot spot of c-Myc in lymphomas, is regulated by mitogens**. *J. Biol. Chem.* (2002) **277** 19229-19235. PMID: 11904304
54. Li X., Molina H., Huang H., Zhang Y.-Y., Liu M., Qian S.-W.. **O-linked N-acetylglucosamine modification on CCAAT enhancer-binding protein β. Role during adipocyte differentiation**. *J. Biol. Chem.* (2009) **284** 19248-19254. PMID: 19478079
55. Rexach J.E., Clark P.M., Mason D.E., Neve R.L., Peters E.C., Hsieh-Wilson L.C.. **Dynamic O-GlcNAc modification regulates CREB-mediated gene expression and memory formation**. *Nat. Chem. Biol.* (2012) **8** 253. PMID: 22267118
56. Guo K., Gan L., Zhang S., Cui F.J., Cun W., Li Y.. **Translocation of HSP27 into liver cancer cell nucleus may be associated with phosphorylation and O-GlcNAc glycosylation**. *Oncol. Rep.* (2012) **28** 494-500. PMID: 22664592
57. Netsirisawan P., Chaiyawat P., Chokchaichamnankit D., Lirdprapamongkol K., Srisomsap C., Svasti J.. **Decreasing O-GlcNAcylation affects the malignant transformation of MCF-7 cells via Hsp27 expression and its O-GlcNAc modification**. *Oncol. Rep.* (2018) **40** 2193-2205. PMID: 30106436
58. Balana A.T., Levine P.M., Craven T.W., Mukherjee S., Pedowitz N.J., Moon S.P.. **O-GlcNAc modification of small heat shock proteins enhances their anti-amyloid chaperone activity**. *Nat. Chem.* (2021) **13** 441-450. PMID: 33723378
59. Morris E.J., Jha S., Restaino C.R., Dayananth P., Zhu H., Cooper A.. **Discovery of a novel ERK inhibitor with activity in models of acquired resistance to BRAF and MEK inhibitors**. *Cancer Discov.* (2013) **3** 742-750. PMID: 23614898
60. Young P.R., McLaughlin M.M., Kumar S., Kassis S., Doyle M.L., McNulty D.. **Pyridinyl imidazole inhibitors of p38 mitogen-activated protein kinase bind in the ATP site**. *J. Biol. Chem.* (1997) **272** 12116-12121. PMID: 9115281
61. Davies S.P., Reddy H., Caivano M., Cohen P.. **Specificity and mechanism of action of some commonly used protein kinase inhibitors**. *Biochem. J.* (2000) **351** 95-105. PMID: 10998351
62. Cheung P.C., Campbell D.G., Nebreda A.R., Cohen P.. **Feedback control of the protein kinase TAK1 by SAPK2a/p38α**. *EMBO J.* (2003) **22** 5793-5805. PMID: 14592977
63. Kumar S., Jiang M.S., Adams J.L., Lee J.C.. **Pyridinylimidazole compound SB 203580 inhibits the activity but not the activation of p38 mitogen-activated protein kinase**. *Biochem. Biophys. Res. Commun.* (1999) **263** 825-831. PMID: 10512765
64. Pikkarainen S., Tokola H., Kerkelä R., Ruskoaho H.. **GATA transcription factors in the developing and adult heart**. *Cardiovasc. Res.* (2004) **63** 196-207. PMID: 15249177
65. Batt A.R., Zaro B.W., Navarro M.X., Pratt M.R.. **Metabolic chemical reporters of glycans exhibit cell-type selective metabolism and glycoprotein labeling**. *Chembiochem* (2017) **18** 1177. PMID: 28231413
66. Darabedian N., Pratt M.R.. **Identifying potentially O-GlcNAcylated proteins using metabolic labeling, bioorthogonal enrichment, and western blotting**. *Methods Enzymol.* (2019) **622** 293-307. PMID: 31155058
67. Cioce A., Bineva-Todd G., Agbay A.J., Choi J., Wood T.M., Debets M.F.. **Optimization of metabolic oligosaccharide engineering with Ac4GalNAlk and Ac4GlcNAlk by an engineered pyrophosphorylase**. *ACS Chem. Biol.* (2021) **16** 1961-1967. PMID: 33835779
68. Shieh P., Dien V.T., Beahm B.J., Castellano J.M., Wyss-Coray T., Bertozzi C.R.. **CalFluors: a universal motif for fluorogenic azide probes across the visible spectrum**. *J. Am. Chem. Soc.* (2015) **137** 7145-7151. PMID: 25902190
69. Ortiz-Meoz R.F., Jiang J., Lazarus M.B., Orman M., Janetzko J., Fan C.. **A small molecule that inhibits OGT activity in cells**. *ACS Chem. Biol.* (2015) **10** 1392-1397. PMID: 25751766
70. Yuzwa S.A., Macauley M.S., Heinonen J.E., Shan X., Dennis R.J., He Y.. **A potent mechanism-inspired O-GlcNAcase inhibitor that blocks phosphorylation of tau in vivo**. *Nat. Chem. Biol.* (2008) **4** 483-490. PMID: 18587388
71. Martin S.E., Tan Z.-W., Itkonen H.M., Duveau D.Y., Paulo J.A., Janetzko J.. **Structure-based evolution of low nanomolar O-GlcNAc transferase inhibitors**. *J. Am. Chem. Soc.* (2018) **140** 13542-13545. PMID: 30285435
72. Liu T.W., Zandberg W.F., Gloster T.M., Deng L., Murray K.D., Shan X.. **Metabolic inhibitors of O-GlcNAc transferase that act in vivo implicate decreased O-GlcNAc levels in leptin-mediated nutrient sensing**. *Angew. Chem. Int. Ed. Engl.* (2018) **130** 7770-7774
73. Zhang Y., Murugesan P., Huang K., Cai H.. **NADPH oxidases and oxidase crosstalk in cardiovascular diseases: novel therapeutic targets**. *Nat. Rev. Cardiol.* (2020) **17** 170-194. PMID: 31591535
74. Ota A., Zhang J., Ping P., Han J., Wang Y.. **Specific regulation of noncanonical p38α activation by Hsp90-Cdc37 chaperone complex in cardiomyocyte**. *Circ. Res.* (2010) **106** 1404-1412. PMID: 20299663
75. Ge B., Gram H., Di Padova F., Huang B., New L., Ulevitch R.J.. **MAPKK-independent activation of p38α mediated by TAB1-dependent autophosphorylation of p38α**. *Science* (2002) **295** 1291-1294. PMID: 11847341
76. Pathak S., Borodkin V.S., Albarbarawi O., Campbell D.G., Ibrahim A., Van Aalten D.M.. **O-GlcNAcylation of TAB1 modulates TAK1-mediated cytokine release**. *EMBO J.* (2012) **31** 1394-1404. PMID: 22307082
77. Wang Z., Gucek M., Hart G.W.. **Cross-talk between GlcNAcylation and phosphorylation: site-specific phosphorylation dynamics in response to globally elevated O-GlcNAc**. *Proc. Natl. Acad. Sci. U. S. A.* (2008) **105** 13793-13798. PMID: 18779572
78. Ding F., Yu L., Wang M., Xu S., Xia Q., Fu G.. **O-GlcNAcylation involvement in high glucose-induced cardiac hypertrophy via ERK1/2 and cyclin D2**. *Amino Acids* (2013) **45** 339-349. PMID: 23665912
79. Jiang M., Qiu Z., Zhang S., Fan X., Cai X., Xu B.. **Elevated O-GlcNAcylation promotes gastric cancer cells proliferation by modulating cell cycle related proteins and ERK 1/2 signaling**. *Oncotarget* (2016) **7** 61390-61402. PMID: 27542217
80. Tallent M.K., Varghis N., Skorobogatko Y., Hernandez-Cuebas L., Whelan K., Vocadlo D.J.. **In vivo modulation of O-GlcNAc levels regulates hippocampal synaptic plasticity through interplay with phosphorylation**. *J. Biol. Chem.* (2009) **284** 174-181. PMID: 19004831
81. Wulff-Fuentes E., Berendt R.R., Massman L., Danner L., Malard F., Vora J.. **The human O-GlcNAcome database and meta-analysis**. *Sci. Data* (2021) **8** 1-11. PMID: 33414438
82. Skorobogatko Y.V., Deuso J., Adolf-Bergfoyle J., Nowak M.G., Gong Y., Lippa C.F.. **Human Alzheimer’s disease synaptic O-GlcNAc site mapping and iTRAQ expression proteomics with ion trap mass spectrometry**. *Amino Acids* (2011) **40** 765-779. PMID: 20563614
83. Goldberg H., Whiteside C., Fantus I.G.. **O-linked β-N-acetylglucosamine supports p38 MAPK activation by high glucose in glomerular mesangial cells**. *Am. J. Physiol. Endocrinol. Metab.* (2011) **301** E713-E726. PMID: 21712532
84. Kneass Z.T., Marchase R.B.. **Protein O-GlcNAc modulates motility-associated signaling intermediates in neutrophils**. *J. Biol. Chem.* (2005) **280** 14579-14585. PMID: 15703172
85. Meijles D.N., Cull J.J., Markou T., Cooper S.T., Haines Z.H., Fuller S.J.. **Redox regulation of cardiac ASK1 (apoptosis signal-regulating kinase 1) controls p38-MAPK (mitogen-activated protein kinase) and orchestrates cardiac remodeling to hypertension**. *Hypertension* (2020) **76** 1208-1218. PMID: 32903101
86. Liu F., Fan L.M., Geng L., Li J.-M.. **p47phox-dependent oxidant signalling through ASK1, MKK3/6 and MAPKs in angiotensin II-induced cardiac hypertrophy and apoptosis**. *Antioxidants (Basel)* (2021) **10** 1363. PMID: 34572995
87. Lu S., Liao Z., Lu X., Katschinski D.M., Mercola M., Chen J.. **Hyperglycemia acutely increases cytosolic reactive oxygen species via O-linked GlcNAcylation and CaMKII activation in mouse ventricular myocytes**. *Circ. Res.* (2020) **126** e80-e96. PMID: 32134364
88. Hegyi B., Borst J.M., Bailey L.R., Shen E.Y., Lucena A.J., Navedo M.F.. **Hyperglycemia regulates cardiac K+ channels via O-GlcNAc-CaMKII and NOX2-ROS-PKC pathways**. *Basic Res. Cardiol.* (2020) **115** 1-19
89. Souza-Silva L., Alves-Lopes R., Silva Miguez J., Dela Justina V., Neves K.B., Mestriner F.L.. **Glycosylation with O-linked β-N-acetylglucosamine induces vascular dysfunction via production of superoxide anion/reactive oxygen species**. *Can. J. Physiol. Pharmacol.* (2018) **96** 232-240. PMID: 28793197
90. Lu G., Kang Y.J., Han J., Herschman H.R., Stefani E., Wang Y.. **TAB-1 modulates intracellular localization of p38 MAP kinase and downstream signaling**. *J. Biol. Chem.* (2006) **281** 6087-6095. PMID: 16407200
91. Meek R.W., Blaza J.N., Busmann J.A., Alteen M.G., Vocadlo D.J., Davies G.J.. **Cryo-EM structure provides insights into the dimer arrangement of the O-linked β-N-acetylglucosamine transferase OGT**. *Nat. Commun.* (2021) **12** 6508. PMID: 34764280
92. De Nicola G.F., Martin E.D., Chaikuad A., Bassi R., Clark J., Martino L.. **Mechanism and consequence of the autoactivation of p38α mitogen-activated protein kinase promoted by TAB1**. *Nat. Struct. Mol. Biol.* (2013) **20** 1182-1190. PMID: 24037507
93. Authier F., Muha V., van Aalten D.M.. **A mouse model for functional dissection of TAB1 O-GlcNAcylation**. *Wellcome Open Res.* (2019) **4** 128. PMID: 32676538
94. Dias W.B., Cheung W.D., Hart G.W.. **O-GlcNAcylation of kinases**. *Biochem. Biophys. Res. Commun.* (2012) **422** 224-228. PMID: 22564745
95. Thapa D., Nichols C., Bassi R., Martin E.D., Verma S., Conte M.R.. **TAB1-induced autoactivation of p38α mitogen-activated protein kinase is crucially dependent on threonine 185**. *Mol. Cell. Biol.* (2018) **38** e00409-e00417. PMID: 29229647
96. Rose B.A., Force T., Wang Y.. **Mitogen-activated protein kinase signaling in the heart: angels versus demons in a heart-breaking tale**. *Physiol. Rev.* (2010) **90** 1507-1546. PMID: 20959622
97. Wang L., Proud C.G.. **Ras/Erk signaling is essential for activation of protein synthesis by Gq protein-coupled receptor agonists in adult cardiomyocytes**. *Circ. Res.* (2002) **91** 821-829. PMID: 12411397
98. Harris I.S., Zhang S., Treskov I., Kovacs A., Weinheimer C., Muslin A.J.. **Raf-1 kinase is required for cardiac hypertrophy and cardiomyocyte survival in response to pressure overload**. *Circulation* (2004) **110** 718-723. PMID: 15289381
99. Clerk A., Michael A., Sugden P.H.. **Stimulation of the p38 mitogen-activated protein kinase pathway in neonatal rat ventricular myocytes by the G protein–coupled receptor agonists, endothelin-1 and phenylephrine: a role in cardiac myocyte hypertrophy?**. *J. Cell Biol.* (1998) **142** 523-535. PMID: 9679149
100. Liao P., Georgakopoulos D., Kovacs A., Zheng M., Lerner D., Pu H.. **The in vivo role of p38 MAP kinases in cardiac remodeling and restrictive cardiomyopathy**. *Proc. Natl. Acad. Sci. U. S. A.* (2001) **98** 12283-12288. PMID: 11593045
101. Pikkarainen S., Tokola H., Kerkela R., Majalahti-Palviainen T., Vuolteenaho O., Ruskoaho H.. **Endothelin-1-specific activation of B-type natriuretic peptide gene via p38 mitogen-activated protein kinase and nuclear ETS factors**. *J. Biol. Chem.* (2003) **278** 3969-3975. PMID: 12446726
102. Watson L.J., Long B.W., DeMartino A.M., Brittian K.R., Readnower R.D., Brainard R.E.. **Cardiomyocyte Ogt is essential for postnatal viability**. *Am. J. Physiol. Heart Circ. Physiol.* (2014) **306** H142-H153. PMID: 24186210
103. Mu Y., Yu H., Wu T., Zhang J., Evans S.M., Chen J.. **O-linked β-N-acetylglucosamine transferase plays an essential role in heart development through regulating angiopoietin-1**. *PLoS Genet.* (2020) **16**
104. Facundo H.T., Brainard R.E., Watson L.J., Ngoh G.A., Hamid T., Prabhu S.D.. **O-GlcNAc signaling is essential for NFAT-mediated transcriptional reprogramming during cardiomyocyte hypertrophy**. *Am. J. Physiol. Heart Circ. Physiol.* (2012) **302** H2122-H2130. PMID: 22408028
105. Zhu W.Z., Ledee D., Olson A.K.. **Temporal regulation of protein O-GlcNAc levels during pressure-overload cardiac hypertrophy**. *Physiol. Rep.* (2021) **9**
106. Liang Q., Wiese R.J., Bueno O.F., Dai Y.S., Markham B.E., Molkentin J.D.. **The transcription factor GATA4 is activated by extracellular signal-regulated kinase 1- and 2-mediated phosphorylation of serine 105 in cardiomyocytes**. *Mol. Cell. Biol.* (2001) **21** 7460-7469. PMID: 11585926
107. Van Berlo J.H., Elrod J.W., Aronow B.J., Pu W.T., Molkentin J.D.. **Serine 105 phosphorylation of transcription factor GATA4 is necessary for stress-induced cardiac hypertrophy**. *Proc. Natl. Acad. Sci. U. S. A.* (2011) **108** 12331-12336. PMID: 21746915
108. Cannon M.V., Silljé H.H., Sijbesma J.W., Vreeswijk-Baudoin I., Ciapaite J., van der Sluis B.. **Cardiac LXR α protects against pathological cardiac hypertrophy and dysfunction by enhancing glucose uptake and utilization**. *EMBO Mol. Med.* (2015) **7** 1229-1243. PMID: 26160456
109. Hantschel O.. **Unexpected off-targets and paradoxical pathway activation by kinase inhibitors**. *ACS Chem. Biol.* (2015) **10** 234-245. PMID: 25531586
110. Fabian M.A., Biggs W.H., Treiber D.K., Atteridge C.E., Azimioara M.D., Benedetti M.G.. **A small molecule-kinase interaction map for clinical kinase inhibitors**. *Nat. Biotechnol.* (2005) **23** 329-336. PMID: 15711537
111. Chaikuad A., Tacconi E.M.C., Zimmer J., Liang Y., Gray N.S., Tarsounas M.. **A unique inhibitor binding site in ERK1/2 is associated with slow binding kinetics**. *Nat. Chem. Biol.* (2014) **10** 853-860. PMID: 25195011
112. Liles J.T., Corkey B.K., Notte G.T., Budas G.R., Lansdon E.B., Hinojosa-Kirschenbaum F.. **ASK1 contributes to fibrosis and dysfunction in models of kidney disease**. *J. Clin. Invest.* (2018) **128** 4485-4500. PMID: 30024858
113. Totzke J., Gurbani D., Raphemot R., Hughes P.F., Bodoor K., Carlson D.A.. **Takinib, a selective TAK1 inhibitor, broadens the therapeutic efficacy of TNF-α inhibition for cancer and autoimmune disease**. *Cell Chem. Biol.* (2017) **24** 1029-1039.e7. PMID: 28820959
114. Goodfellow V.S., Loweth C.J., Ravula S.B., Wiemann T., Nguyen T., Xu Y.. **Discovery, synthesis, and characterization of an orally bioavailable, brain penetrant inhibitor of mixed lineage kinase 3**. *J. Med. Chem.* (2013) **56** 8032-8048. PMID: 24044867
115. Wu Z., Gholami A.M., Kuster B.. **Systematic identification of the HSP90 candidate regulated proteome**. *Mol. Cell. Proteomics* (2012) **11**
116. Koitabashi N., Aiba T., Hesketh G.G., Rowell J., Zhang M., Takimoto E.. **Cyclic GMP/PKG-dependent inhibition of TRPC6 channel activity and expression negatively regulates cardiomyocyte NFAT activation: novel mechanism of cardiac stress modulation by PDE5 inhibition**. *J. Mol. Cell. Cardiol.* (2010) **48** 713-724. PMID: 19961855
117. Takimoto E., Champion H.C., Li M., Belardi D., Ren S., Rodriguez E.R.. **Chronic inhibition of cyclic GMP phosphodiesterase 5A prevents and reverses cardiac hypertrophy**. *Nat. Med.* (2005) **11** 214-222. PMID: 15665834
118. Schiapparelli L.M., McClatchy D.B., Liu H.-H., Sharma P., Yates J.R., Cline H.T.. **Direct detection of biotinylated proteins by mass spectrometry**. *J. Proteome Res.* (2014) **13** 3966-3978. PMID: 25117199
|
---
title: 'No insulin degludec dose adjustment required after aerobic exercise for people
with type 1 diabetes: the ADREM study'
authors:
- Linda C. A. Drenthen
- Mandala Ajie
- Evertine J. Abbink
- Laura Rodwell
- Dick H. J. Thijssen
- Cees J. Tack
- Bastiaan E. de Galan
journal: Diabetologia
year: 2023
pmcid: PMC9988601
doi: 10.1007/s00125-023-05893-9
license: CC BY 4.0
---
# No insulin degludec dose adjustment required after aerobic exercise for people with type 1 diabetes: the ADREM study
## Abstract
### Aims/hypothesis
It is generally recommended to reduce basal insulin doses after exercise to reduce the risk of post-exercise nocturnal hypoglycaemia. Based on its long t½, it is unknown whether such adjustments are required or beneficial for insulin degludec.
### Methods
The ADREM study (Adjustment of insulin Degludec to Reduce post-Exercise (nocturnal) hypoglycaeMia in people with diabetes) was a randomised controlled, crossover study in which we compared $40\%$ dose reduction (D40), or postponement and $20\%$ dose reduction (D20-P), with no dose adjustment (CON) in adults with type 1 diabetes at elevated risk of hypoglycaemia, who performed a 45 min aerobic exercise test in the afternoon. All participants wore blinded continuous glucose monitors for 6 days, measuring the incidence of (nocturnal) hypoglycaemia and subsequent glucose profiles.
### Results
We recruited 18 participants (six women, age 38 ± 13 years, HbA1c 56 ± 8 mmol/mol [7.3 ± $0.8\%$], mean ± SD). Time below range (i.e. glucose <3.9 mmol/l) the night after the exercise test was generally low and occurrence did not differ between the treatment regimens. During the subsequent whole day, time below range was lower for D40 compared with CON (median [IQR], 0 [0–23] vs 18 [0–55] min, $$p \leq 0.043$$), without differences in the number of hypoglycaemic events. Time above range (i.e. glucose >10 mmol/l) was greater for D20-P vs CON (mean ± SEM, 584 ± 81 vs 364 ± 66 min, $$p \leq 0.001$$) and D40 (385 ± 72 min, $$p \leq 0.003$$).
### Conclusions/interpretation
Post-exercise adjustment of degludec does not mitigate the risk of subsequent nocturnal hypoglycaemia in people with type 1 diabetes. Although reducing degludec reduced next-day time below range, this did not translate into fewer hypoglycaemic events, while postponing degludec should be avoided because of increased time above range. Altogether, these data do not support degludec dose adjustment after a single exercise bout.
### Trial registration
EudraCT number 2019-004222-22
### Funding
The study was funded by an unrestricted grant from Novo Nordisk, Denmark.
### Supplementary Information
The online version of this article (10.1007/s00125-023-05893-9) contains peer-reviewed but unedited supplementary material.
## Introduction
Regular physical exercise is recommended for people with type 1 diabetes mellitus given its beneficial effects on general well-being, cardiometabolic health and insulin requirements [1]. However, aerobic exercise in people with type 1 diabetes increases the risk of hypoglycaemia due to the inability to adjust for falling insulin requirements [2, 3]. This risk is amplified because muscle glycogen storage needs to be replenished, leading to increased insulin sensitivity and glucose disposal. These effects usually peak 7–11 h after exercise, but can last for up to 24 h [4]. As a consequence, there is an increased risk of nocturnal hypoglycaemia, particularly after engaging in sports in the afternoon or evening [4, 5]. This may be even more pronounced in people with reduced awareness of hypoglycaemia. Nocturnal hypoglycaemia is associated with impaired cognitive function and well-being [6] and is the main barrier for people with type 1 diabetes to engage in sports [7]. Dose reduction of meal-related bolus insulin does not prevent late-onset (nocturnal) hypoglycaemia in people using first-generation long-acting insulins [8]. However, reducing these basal insulins after exercise can prevent exercise-induced (nocturnal) hypoglycaemia [9], as can reducing the basal rate of insulin infusion pumps [10]. To mitigate the risk of nocturnal hypoglycaemia, it is therefore typically recommended to reduce the dose of first-generation long-acting insulin at bedtime or the basal rate of insulin infusion in pump users by $20\%$ after afternoon or evening exercise [3].
Insulin degludec is a second-generation long-acting insulin analogue with a much longer t½ compared with other long-acting insulins, resulting in a more stable glucose-lowering profile and longer duration of action [11, 12]. Use of insulin degludec has been associated with reduced risks of hypoglycaemia, particularly nocturnal events [13]. However, it is suggested that insulin degludec carries the same risk for post-exercise (nocturnal) hypoglycaemia compared with insulin glargine in people with type 1 diabetes [14]. The long t½ of degludec has important implications for dosing adjustments, since a steady state will be reached no earlier than after 2–3 days [15]. One study found that a $25\%$ dose reduction of insulin degludec did not reduce the risk for hypoglycaemia in people with type 1 diabetes during 5 consecutive days of moderate-intensity activity, but this study was very small ($$n = 7$$) [16]. As such, it is unclear what recommendation for insulin dose reduction after exercise is justified for insulin degludec. Therefore, this study compared the effects of two different degludec dose adjustments with no adjustment on the incidence of nocturnal hypoglycaemia and glucose profiles after aerobic exercise in people with type 1 diabetes at elevated risk of hypoglycaemia.
## Study procedures
The ADREM study (Adjustment of insulin Degludec to Reduce post-Exercise (nocturnal) hypoglycaeMia in people with diabetes) was an open-label, randomised controlled, three-way crossover study, conducted in accordance with the principles of the Declaration of Helsinki, the Medical Research Involving Human Subjects Act and applicable International Conference on Harmonization (ICH) Good Clinical Practice guidelines. The study was approved by the local ethics committee and national competent authority. All participants gave their written informed consent before any study-related activity was performed.
## Study participants
Adults aged 18–60 years were eligible for participation when they had been diagnosed with type 1 diabetes for at least 2 years, had been treated with a basal-bolus multi-dose insulin regimen for at least 1 year and were at increased risk of hypoglycaemia. The latter was defined as a history of at least one severe hypoglycaemia event in the past year and/or ≥2 points on the Dutch modified version of the Clarke questionnaire and/or ≥3 points on the Gold score [17–19]. They also had to engage in moderate-intensity exercise for at least 1 h per week and had to have an HbA1c ≤75 mmol/mol ($9\%$). Main exclusion criteria were microvascular complications (except for background retinopathy or a urinary albumin/creatinine of maximum 30 mg/mmol), BMI >30 kg/m2, pregnancy, Modification of Diet in Renal Disease (MDRD) GFR <60 ml/min per 1.73 m2, any contraindication for exercise testing according to the American Heart Association (AHA) guidelines [20] and the use of β-blockers or drugs affecting glucose metabolism other than insulin. Participants were recruited from the outpatient clinic of the Radboud University Medical Center and Rijnstate Hospital and websites of patient associations.
## Screening visit
A schematic overview of the study design is shown in Fig. 1. All participants performed an incremental cardiopulmonary exercise test (CPET) (Lode Excalibur; Lode, Groningen, the Netherlands) on a bicycle ergometer to determine their maximum cardiovascular fitness level (defined as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \dot{V}{\mathrm{O}}_{2\max } $$\end{document}V˙O2max) and maximum heart rate [21]. During CPET, the participant was asked to cycle at a continuous rate of 60 to 80 rotations per minute and the work rate was increased every minute by 15 or 20 W until exhaustion, or an indication to stop according to the American Thoracic Society/American College of Chest Physicians (ATS/ACCP) statement [22]. Directly before and after cycling, venous blood was collected to measure glucose and lactate concentrations using Biosen C-Line (EKF Diagnostics, Cardiff, UK). After screening, all participants were instructed to inject insulin degludec at 23:00 hours during the trial, and participants not on degludec were transferred to it. A 28 day titration run-in was used to reach stable glycaemic control, defined as a self-measured fasting mean glucose concentration below 7 mmol/l. Fig. 1Overview of the study. Ex, exercise day; Scr, screening visit
## Exercise days
Each participant engaged in 3 exercise days and was randomly assigned by the investigator to an order of the three post-exercise degludec treatment regimens, i.e. no adjustment of insulin degludec (CON), a $40\%$ dose reduction of degludec (D40) and 8 h postponement with a $20\%$ dose reduction of degludec (D20-P). For D40, the usual recommended long-acting insulin dose reduction of $20\%$ was doubled, because degludec has a t½ of about twice that of insulin glargine. This means that, in theory, plasma insulin levels will gradually fall and be $20\%$ lower at the time of the next injection, and ~$7\%$ lower the next morning. By postponing the injection by 8 h (i.e. a third of the t½), we expected insulin levels to fall ~$16\%$ overnight. The six potential treatment sequences were evenly distributed among the trial population. The exercise days were each separated by a period of 14 days (±3 days), except that 5 exercise days had to be postponed for up to 14 days because of COVID-19 restrictions. Every exercise day was followed by 6 days of blinded continuous glucose monitoring (CGM) (Dexcom G6; Dexcom, San Diego, CA, USA). Participants were also allowed to simultaneously use their own glucose sensor. Sleep times were recorded using an activity tracker (activPAL3 micro; PAL Technologies, Glasgow, UK) [23]. Participants were requested to refrain from strenuous exercise of all types during the 2 days before and 7 days after the exercise tests. On the exercise days, participants consumed their lunch at home with a $50\%$ dose reduction of their short-acting insulin to prevent hypoglycaemia before and during the exercise test. Between 15:30 and 16:30 hours, participants arrived at the research facility, where CGM was started. Blood was sampled for determination of glucose, lactate, insulin, catecholamines and cortisol at arrival, 5 min before and after the exercise test, and before discharge. Depending on the participant’s glucose concentration and its trend before the exercise test, the participants consumed a carbohydrate-rich snack aiming for a blood glucose concentration of 7–14 mmol/l. At 18:00 hours, participants performed a 45 min exercise test on a bicycle ergometer at $70\%$ of their heart rate reserve using the Karvonen formula, based on the maximum heart rate determined by CPET and the resting heart rate measured during the screening visit [24]. During the exercise test, the participant’s glucose concentration was monitored by measuring their interstitial glucose level using their own glucose sensor and additional carbohydrates were given when necessary. After the exercise test and before discharge, participants consumed a standardised meal (consisting of 45–$50\%$ carbohydrates, 30–$40\%$ protein and 20–$30\%$ fat) with a $25\%$ dose reduction of their short-acting insulin. Participants were instructed not to eat from discharge until getting up the next day, except in case of hypoglycaemia. They were also instructed not to inject any short-acting insulin from discharge until getting up the next day, except in case of profound hyperglycaemia. At 23:00 hours on the exercise day (CON and D40) or 07:00 hours the next day (D20-P), the participants administered insulin degludec. The day after the exercise tests, participants measured their fasting ketones by point-of-care testing before 07:30 hours. They also registered their injected insulin dose for 6 days after the exercise day.
## Study outcomes
The primary outcome was the time below range (i.e. glucose <3.9 mmol/l) in the night (00:00–05:59 hours) following the exercise test. Secondary outcomes included times above range (i.e. glucose >10.0 mmol/l) and in range (i.e. glucose ≥3.9 mmol/l and ≤10.0 mmol/l), mean glucose concentration, number of hypoglycaemic and severe hypoglycaemic (requiring external assistance for recovery) events and total daily dose of short-acting insulin. All outcome variables were calculated during the first and second days (00:00–23:59 hours) after the exercise test as well as for the total 6 days following the exercise test. A hypoglycaemic event was defined as a glucose concentration <3.9 mmol/l for at least 15 consecutive minutes and a new event was calculated if the glucose concentration had been risen above this level for at least 15 min [25]. All CGM outcomes were calculated using R version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria).
## Measurements
Plasma insulin was measured using an in-house radioimmunoassay, using an in-house-generated guinea pig anti-human insulin antibody and 125I-labelled human insulin tracer. 125I-labelled human insulin tracer was generated using 125I (PerkinElmer Nederland) and human insulin (Novo Biolabs cat. no. 471). In this assay, bound–free separation is performed by second antibody/polyethylene glycol precipitation of antibody-bound insulin. The assay is calibrated on World Health Organization international standard $\frac{83}{500.}$ The cross-reactivity in this method is approximately $60\%$ for insulin aspart and $50\%$ for insulin lispro. The cross-reactivity for insulin degludec is not well known. Catecholamines were analysed by an LC-MS/MS method developed and validated in-house after derivatisation with propionic anhydride and subsequent solid-phase extraction [26]. Plasma cortisol was determined using a routine analysis method with an electrochemiluminescent immunoassay on a Cobas E801 random access analyser (Roche Diagnostics, Mannheim, Germany) [27].
## Statistical analyses
The sample size estimation was based on the two treatments for which the smallest difference was expected (D40 and CON). Given this was a crossover trial, it was expected that the other comparisons (with D20-P) would have sufficient power. No ɑ corrections were made to account for the multiple comparisons. Furthermore, it was assumed that the within-person correlation for the response measures on the different treatments was 0.65. We aimed at finding a significant decrease in time below range with at least $20\%$ increase in time in range during the night after the exercise test and $50\%$ increase in next-morning fasting glucose concentration [9]. We calculated that 13 participants would be required to detect a difference at a significance level of 0.05 and a power of $80\%$. To account for the relatively small number of participants involved, a total of 18 participants were enrolled. Data were analysed using IBM SPSS version 25 and Stata version 16. We performed an as-treated analysis. We used random effects models to account for the three measurements for each participant, with period and treatment as independent variables. Given the low incidence of hypoglycaemia the night after the exercise test, the primary outcome was transformed to a binary outcome to represent no or any time below range, and analysed using a logistic random effects model. Differences in continuous variables between the three treatment arms were analysed using a multilevel mixed-effects linear regression model performing restricted maximum-likelihood estimation. Differences in count data between the study arms were analysed using a negative binomial random effects model. Data that were not normally distributed were log transformed or analysed using the related samples Friedman’s two-way analysis. No adjustments were made to account for multiple testing of the secondary endpoints. Every day started at midnight and the night period was defined as 00:00 hours to 05:59 hours. We performed a sensitivity analysis where we repeated all analyses for the CGM data based on the sleep times of the participants instead of the predefined day and night periods. All data are expressed as mean ± SEM or median [IQR], unless otherwise specified. A p value <0.05 was considered statistically significant.
## Results
A total of 19 participants were screened, 18 of whom were included. One participant was withdrawn after screening because of personal reasons unrelated to the study. All 18 included participants completed the study. Their baseline characteristics are shown in Table 1. Nine participants were already on insulin degludec; the other participants were transferred to it with a mean ± SD dose of 87 ± $10\%$ of their pre-study long-acting insulin dose (insulin glargine and detemir). There were no differences in the proportion of glucose readings by the glucose sensor between the treatment regimens during the total 6 day periods (mean ± SD, CON 99 ± $2\%$, D40 95 ± $16\%$, D20-P 98 ± $3\%$). One participant was on real-time CGM and 17 used flash-glucose monitoring, three of whom had the alarm function for low and high glucose concentrations turned on. All participants achieved maximal exhaustion during CPET (electronic supplementary material [ESM] Table 1). No serious adverse events occurred during the study. One participant had mild cellulitis on her foot and one participant was infected by COVID-19 during the study period, but neither were judged to be related to the study, nor to have impact on the study results. Table 1Baseline characteristicsCharacteristicn=18Age, years38 ± 13Male sex12 [67]BMI, kg/m225.0 ± 2.7Duration of diabetes, years12 ± 11HbA1c, mmol/mol56 ± 8HbA1c, %7.3 ± 0.8Total insulin dose, U/day49 ± 26Short-acting insulin Insulin aspart15 [83] Fast-acting insulin aspart2 [11] Insulin lispro1 [6]Score on modified Clarke questionnaire2 [2–2]Gold score2 [2–3]IAH5 [28]Serum creatinine, μmol/l72 ± 16\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \dot{V}{\mathrm{O}}_{2\max } $$\end{document}V˙O2max, ml min−1 kg−140.2 ± 9.6Data are presented as number (%), mean ± SD or median [IQR]IAH, impaired awareness of hypoglycaemia, i.e. ≥3 points on Clarke questionnaire and/or ≥4 points on Gold score
## Exercise tests
The three 45 min exercise tests were performed consistently across all treatment groups regarding heart rate (CON 144 ± 3, D40 145 ± 3, D20-P 144 ± 3 beats per minute). All participants cycled at $70\%$ of their heart rate reserve, except for one participant, who cycled all three tests at $57\%$ heart rate reserve because of mild ST-segment depression during CPET. The increase in blood lactate (Δ lactate CON 0.76 ± 0.15, D40 0.97 ± 0.35, D20-P 0.69 ± 0.14 mmol/l) and decrease in blood glucose levels (Δ glucose CON −3.88 ± 0.95, D40 −4.91 ± 0.50, D20-P −3.69 ± 0.58 mmol/l) during cycling did not differ for the three exercise tests (ESM Fig. 1). Six participants in CON, three in D40 and six in D20-P ingested additional carbohydrates shortly before or during the test to prevent hypoglycaemia. The counter-regulatory hormone responses to exercise did not differ between treatment regimens, except for the cortisol response which was slightly higher during D40 compared with CON (97 ± 37 vs 53 ± 39 nmol/l, $$p \leq 0.03$$) (ESM Fig. 1).
## Time below range and hypoglycaemic events
Time below range in the night after the exercise test was generally low and occurrence did not differ significantly between the treatment regimens (three participants in CON, one in D40, three in D20-P). The day after the exercise test, time below range was greater for CON compared with D40 (18 [0–55] vs 0 [0–23] min, $$p \leq 0.043$$) (ESM Fig. 2), but the number of hypoglycaemic events was similar (ESM Table 2). During the second whole day, D20-P was associated with more time below range (28 [4–46] vs 0 [0–41] min, $$p \leq 0.019$$) and more hypoglycaemic events (20 vs 9, $$p \leq 0.027$$) compared with D40, but neither differed significantly from CON (5 [0–44] min; 13 events). No differences in time below range and hypoglycaemic events were found between the treatment regimens during the total 6 days after the exercise test. No severe hypoglycaemic events occurred during the entire study period.
## Mean glucose concentration
No differences in mean glucose concentration were found between the treatment regimens the night after the exercise test (Fig. 2). The day after the exercise test, D20-P was associated with a higher mean glucose concentration (9.6 ± 0.5 mmol/l) compared with CON (8.5 ± 0.4 mmol/l, $$p \leq 0.015$$) and D40 (8.7 ± 0.5 mmol/l, $$p \leq 0.035$$) (ESM Table 2). The second day after the exercise test, D20-P was associated with a lower mean glucose concentration compared with CON (8.6 ± 0.5 vs 9.8 ± 0.5 mmol/l, $$p \leq 0.014$$), but neither differed significantly from D40 (9.5 ± 0.5 mmol/l). No differences in mean glucose concentration were found between the treatment regimens during the total 6 days after the exercise test. Fig. 2Time course of the mean glucose concentration over the first (a) and second day (b) after the exercise test, according to insulin degludec dosing regimen. Grey, CON; red, D40; blue, D20-P. Values are given as mean ± SEM
## Time above range
No differences in time above range were found between the treatment regimens the night after the exercise test. The day after the exercise test, D20-P led to significantly more time above range (584 ± 81 min) compared with CON (364 ± 66 min, $$p \leq 0.001$$) and D40 (385 ± 72 min, $$p \leq 0.003$$) (Fig. 3). No differences in time above range were found between the treatment regimens during the second day after the exercise test, nor for the total 6 day period. Fig. 3Percentage of time spent above range, in range and below range the first night (a), first day (b), second day (c) and total 6 days (d) after the exercise test, according to insulin degludec dosing regimen. TAR, time above range (yellow); TIR, time in range (green); TBR, time below range (red). Data are given as mean values
## Time in range
The night following the exercise test, D20-P was associated with less time in range compared with D40 (229 ± 30 vs 287 ± 26 min, $$p \leq 0.027$$), but neither differed significantly from CON (256 ± 26 min). The day after the exercise test, D20-P led to significantly less time in range (824 ± 74 min) compared with CON (1041 ± 62 min, $$p \leq 0.001$$) and D40 (1029 ± 70 min, $$p \leq 0.002$$). No differences in time in range were found between the treatment regimens during the second day after the exercise day, nor for the total 6 day period.
## Sensitivity analyses
Repeating the analyses according to the actual sleep times of the participants did not materially change the results, except that the night after the exercise test, the mean glucose concentration was higher for D20-P compared with CON (9.8 ± 0.8 vs 8.5 ± 0.5 mmol/l, $$p \leq 0.044$$), but neither differed significantly from D40 (8.9 ± 0.7 mmol/l) (ESM Table 3). Furthermore, during this night, time in range was lower for D20-P when compared with both D40 (255 ± 44 vs 366 ± 46 min, $$p \leq 0.014$$) and CON (376 ± 35 min, $$p \leq 0.008$$). The day after the exercise test, the mean glucose concentration was higher for D20-P only when compared with CON (9.6 ± 0.5 vs 8.4 ± 0.4 mmol/l, $$p \leq 0.012$$), whereas neither differed significantly from D40 (8.8 ± 0.5 mmol/l). The second day after the exercise test, time below range was higher for D20-P when compared with both D40 (28 [4–65] vs 0 [0–41] min, $$p \leq 0.016$$) and CON (5 [0–44] min, $$p \leq 0.038$$).
## Fasting ketones
Fasting ketones in the morning after the exercise tests were generally low (all ≤0.8 mmol/l), but were significantly higher for D20-P than D40 (0.27 ± 0.04 vs 0.16 ± 0.03 mmol/l, $$p \leq 0.022$$), neither of which differed from CON (0.21 ± 0.05 mmol/l).
## Short-acting insulin and carbohydrate intake
The evening after the exercise test, three people ($17\%$) in CON, one ($6\%$) in D40 and one ($6\%$) in D20-P injected additional short-acting insulin because of profound hyperglycaemia. During this evening, nine people ($50\%$) in CON, five ($28\%$) in D40 and two ($11\%$) in D20-P ingested additional carbohydrates to prevent hypoglycaemia. However, of these participants, only one within each treatment arm ingested carbohydrates after 23:00 hours without experiencing a nocturnal hypoglycaemic event. No differences were found in the total daily dose of short-acting insulin used between the treatment regimens on the first and second days after the exercise test (ESM Fig. 3).
## Discussion
The main finding of this study is that adjustment of insulin degludec dosing after aerobic exercise performed in the afternoon had no effect on the incidence of subsequent nocturnal hypoglycaemia in people with type 1 diabetes. While next-day time below range was slightly reduced in the $40\%$ dose reduction group, this did not translate to fewer hypoglycaemic events. Postponement of degludec to the next morning at a $20\%$ lower dose led to more time above range and less time in range during that day, as well as slightly more time below range on the subsequent second day. Altogether, these results do not support dose adjustments of degludec in people with type 1 diabetes after afternoon aerobic exercise.
Two recent studies have reported a relatively low incidence of nocturnal hypoglycaemia after aerobic exercise in people with type 1 diabetes using insulin degludec [16, 28]. We extend those findings by showing that this is also the case for people at elevated risk for hypoglycaemia. We believe that meticulous adherence to the protocol of short-acting insulin dose reductions at the subsequent meal after exercise was critical for this result. Although this is not sufficient for people using first-generation long-acting insulins [8, 29], reducing the meal-related dose of short-acting insulin after evening exercise may reduce the risk of nocturnal hypoglycaemia in people with type 1 diabetes on insulin degludec [28]. This could be due to the more durable and stable pharmacodynamic profile of insulin degludec when administered at fixed timepoints as compared with first-generation long-acting insulins [13].
Heise et al reported insulin glargine and insulin degludec to have a similar (nocturnal) hypoglycaemic risk profile after performing aerobic exercise of moderate intensity in people with type 1 diabetes, without insulin dose reductions [14]. However, patients at high risk for hypoglycaemia were excluded from participation and participants injected their long-acting insulin in the morning. Since insulin glargine has the strongest glucose-lowering effect during the first 12 h after injection [11], a higher number of nocturnal hypoglycaemic events is plausible for people injecting this type of insulin before bedtime, as is still common practice. In addition, the blood glucose concentration was measured at a few predefined timepoints instead of using CGM, so the occurrence of hypoglycaemia may have been underestimated.
For people with type 1 diabetes, the risk of hypoglycaemia is increased for at least 24 h after aerobic exercise, in particular, when performed in the afternoon [5]. Indeed, half of the participants in CON had at least one episode of time below range the day after the exercise test. Although the number of hypoglycaemic events was not reduced, our data suggest that a $40\%$ dose reduction is needed to reduce next-day time below range, without a concomitant increase in the risk of hyperglycaemia. However, these data seem to contrast with general recommendations to reduce the basal insulin component of insulin regimens by $20\%$ after exercise to achieve this result, but may be specific for insulin degludec [3]. Indeed, a $20\%$ dose reduction in the postponement study arm did not lower the risk of subsequent hypoglycaemia. This is supported by previous research showing that a $25\%$ dose reduction of insulin degludec on 5 consecutive days did not protect against hypoglycaemia during the first 48 h after exercise in people with type 1 diabetes [16].
We chose postponement of insulin degludec as one of the dosing adjustment regimens because of previous data showing that alternating degludec dosing at flexible intervals of 8 to 40 h provided about similar glycaemic control to dosing every 24 h [30]. However, in contrast to that study, we found that our participants spent more time above range and less time in range the day after the exercise test when randomised to the postponement study arm, as compared with the other two study arms. The largest difference in mean glucose concentration was seen in the early morning (Fig. 2a), where the additional dose reduction had no effect yet. One explanation for this apparent discrepancy may be the relatively low insulin dose used in our study, since the duration of action of insulin is dose-dependent, with a longer duration of action with larger doses [31, 32]. Indeed, on average, our participants used insulin degludec at a daily dose of 23 units, approximately 10 units less than in the study by Mathieu et al [30]. Nevertheless, our data argue against postponing insulin degludec to reduce post-exercise hypoglycaemia, particularly when insulin doses are low.
Strengths of our study are the randomised crossover study design, the robust and highly reproducible exercise protocol, the use of CGM and the daily life setting, all of which are relevant in the context of the potential need for dose adjustments.
Our study also has limitations. First, it may be questioned to what extent our results can be generalised to people performing morning exercise or injecting degludec in the morning. However, morning exercise leads to a lower risk of late-onset hypoglycaemia compared with afternoon exercise in people with type 1 diabetes on insulin pump therapy [5]. It is similarly unlikely that morning degludec administration is associated with greater risk for nocturnal hypoglycaemia than evening administration. Therefore, we expect that aerobic exercise in the morning can be safely performed by people with type 1 diabetes on insulin degludec without adjusting the dose and irrespective of injection time. Second, more people in CON ingested carbohydrates in the evening after exercise compared with the other two treatment arms. It could be that participants felt unease in breaking their routine of ingesting carbohydrates after exercise in the control arm, even though we advised against it. Although this may have affected the risk of hypoglycaemia in the first couple of hours, additional intake of carbohydrates after exercise has been found to be insufficient for preventing late-night hypoglycaemia in people using first-generation long-acting insulins [29]. In our study, only three people ingested additional carbohydrates after 23:00 hours, which was evenly distributed across the treatment regimens. Besides, slightly more people in CON injected short-acting insulin that evening because of hyperglycaemia, making it further unlikely that late-evening eating played an important role. Third, three people used the automatic hypoglycaemia alarm function of their glucose sensor; because all three had overt impaired awareness of hypoglycaemia, we deemed it unsafe and unethical for them to turn the alarm function off for the sake of the study. However, alarm function settings were similar for all study periods and all three participants spent time below range during every study period with alarms going off to a similar extent. Finally, using a crossover study design has the potential risk of a carry-over or period effect. To minimise these risks, we used block-randomisation and a wash-out period of 2 weeks between the exercise days. In addition, for our statistical analyses we corrected for a period effect, although we would not expect that to be present.
In conclusion, adjustment of insulin degludec dosing after aerobic exercise in the late afternoon has no effect on subsequent nocturnal hypoglycaemia in people with type 1 diabetes. In fact, postponing the administration of degludec leads to more time above range, which underscores the importance of adhering to insulin degludec dosing around exercise, especially when insulin doses are low. Adjustments in meal-related short-acting insulin both before and after exercise may be advisable for people using degludec, but our data do not provide support for standard insulin degludec dose adjustment after exercise in people with type 1 diabetes. These data add evidence for the ease of use of insulin degludec for most people with type 1 diabetes who want to engage in aerobic exercise.
## Supplementary information
ESM(PDF 450 kb)
## Authors’ relationships and activities
BEdG is associate editor for Diabetologia, but was not involved in the handling of the manuscript during the editorial process. The authors declare that there are no other relationships or activities that might bias, or be perceived to bias, their work.
## Contribution statement
EJA, CJT and BEdG designed the study. LCAD recruited the participants, performed the experiments, collected the data, analysed the data and wrote the first version of the manuscript. LCAD, MA and LR performed the statistical analyses. All authors discussed the results and implications, commented on the manuscript at all stages and approved the final version of the manuscript. The guarantor (BEdG) accepts full responsibility for the work and conduct of the study, has access to the data and controlled the decision to publish.
## References
1. Chimen M, Kennedy A, Nirantharakumar K, Pang TT, Andrews R, Narendran P. **What are the health benefits of physical activity in type 1 diabetes mellitus? A literature review**. *Diabetologia* (2012) **55** 542-551. DOI: 10.1007/s00125-011-2403-2
2. Metcalf KM, Singhvi A, Tsalikian E. **Effects of moderate-to-vigorous intensity physical activity on overnight and next-day hypoglycemia in active adolescents with type 1 diabetes**. *Diabetes Care* (2014) **37** 1272-1278. DOI: 10.2337/dc13-1973
3. Riddell MC, Gallen IW, Smart CE. **Exercise management in type 1 diabetes: a consensus statement**. *Lancet Diabetes Endocrinol* (2017) **5** 377-390. DOI: 10.1016/s2213-8587(17)30014-1
4. McMahon SK, Ferreira LD, Ratnam N. **Glucose requirements to maintain euglycemia after moderate-intensity afternoon exercise in adolescents with type 1 diabetes are increased in a biphasic manner**. *J Clin Endocrinol Metab* (2007) **92** 963-968. DOI: 10.1210/jc.2006-2263
5. Gomez AM, Gomez C, Aschner P. **Effects of performing morning versus afternoon exercise on glycemic control and hypoglycemia frequency in type 1 diabetes patients on sensor-augmented insulin pump therapy**. *J Diabetes Sci Technol* (2015) **9** 619-624. DOI: 10.1177/1932296814566233
6. Graveling AJ, Frier BM. **The risks of nocturnal hypoglycaemia in insulin-treated diabetes**. *Diabetes Res Clin Pract* (2017) **133** 30-39. DOI: 10.1016/j.diabres.2017.08.012
7. Brazeau AS, Rabasa-Lhoret R, Strychar I, Mircescu H. **Barriers to physical activity among patients with type 1 diabetes**. *Diabetes Care* (2008) **31** 2108-2109. DOI: 10.2337/dc08-0720
8. Campbell MD, Walker M, Trenell MI. **Large pre- and postexercise rapid-acting insulin reductions preserve glycemia and prevent early- but not late-onset hypoglycemia in patients with type 1 diabetes**. *Diabetes Care* (2013) **36** 2217-2224. DOI: 10.2337/dc12-2467
9. Campbell MD, Walker M, Bracken RM. **Insulin therapy and dietary adjustments to normalize glycemia and prevent nocturnal hypoglycemia after evening exercise in type 1 diabetes: a randomized controlled trial**. *BMJ Open Diabetes Res Care* (2015) **3** e000085. DOI: 10.1136/bmjdrc-2015-000085
10. Taplin CE, Cobry E, Messer L, McFann K, Chase HP, Fiallo-Scharer R. **Preventing post-exercise nocturnal hypoglycemia in children with type 1 diabetes**. *J Pediatr* (2010) **157** 784-788. DOI: 10.1016/j.jpeds.2010.06.004
11. Heise T, Hövelmann U, Nosek L, Hermanski L, Bøttcher SG, Haahr H. **Comparison of the pharmacokinetic and pharmacodynamic profiles of insulin degludec and insulin glargine**. *Expert Opin Drug Metab Toxicol* (2015) **11** 1193-1201. DOI: 10.1517/17425255.2015.1058779
12. Heise T, Hermanski L, Nosek L, Feldman A, Rasmussen S, Haahr H. **Insulin degludec: four times lower pharmacodynamic variability than insulin glargine under steady-state conditions in type 1 diabetes**. *Diabetes Obes Metab* (2012) **14** 859-864. DOI: 10.1111/j.1463-1326.2012.01627.x
13. Lane W, Bailey TS, Gerety G. **Effect of insulin degludec vs insulin glargine U100 on hypoglycemia in patients with type 1 diabetes: the SWITCH 1 randomized clinical trial**. *JAMA* (2017) **318** 33-44. DOI: 10.1001/jama.2017.7115
14. Heise T, Bain SC, Bracken RM. **Similar risk of exercise-related hypoglycaemia for insulin degludec to that for insulin glargine in patients with type 1 diabetes: a randomized cross-over trial**. *Diabetes Obes Metab* (2016) **18** 196-199. DOI: 10.1111/dom.12588
15. Heise T, Korsatko S, Nosek L. **Steady state is reached within 2-3 days of once-daily administration of degludec, a basal insulin with an ultralong duration of action**. *J Diabetes* (2016) **8** 132-138. DOI: 10.1111/1753-0407.12266
16. Moser O, Eckstein ML, Mueller A. **Reduction in insulin degludec dosing for multiple exercise sessions improves time spent in euglycaemia in people with type 1 diabetes: A randomized crossover trial**. *Diabetes Obes Metab* (2019) **21** 349-356. DOI: 10.1111/dom.13534
17. Clarke WL, Cox DJ, Gonder-Frederick LA, Julian D, Schlundt D, Polonsky W. **Reduced awareness of hypoglycemia in adults with IDDM. A prospective study of hypoglycemic frequency and associated symptoms**. *Diabetes Care* (1995) **18** 517-522. DOI: 10.2337/diacare.18.4.517
18. Gold AE, MacLeod KM, Frier BM. **Frequency of severe hypoglycemia in patients with type I diabetes with impaired awareness of hypoglycemia**. *Diabetes Care* (1994) **17** 697-703. DOI: 10.2337/diacare.17.7.697
19. Geddes J, Wright RJ, Zammitt NN, Deary IJ, Frier BM. **An evaluation of methods of assessing impaired awareness of hypoglycemia in type 1 diabetes**. *Diabetes Care* (2007) **30** 1868-1870. DOI: 10.2337/dc06-2556
20. Gibbons RJ, Balady GJ, Beasley JW. **ACC/AHA Guidelines for Exercise Testing. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on Exercise Testing)**. *J Am Coll Cardiol* (1997) **30** 260-311. DOI: 10.1016/s0735-1097(97)00150-2
21. Janssen L, Frambach S, Allard NAE. **Skeletal muscle toxicity associated with tyrosine kinase inhibitor therapy in patients with chronic myeloid leukemia**. *Leukemia* (2019) **33** 2116-2120. DOI: 10.1038/s41375-019-0443-7
22. Ross RM. **ATS/ACCP statement on cardiopulmonary exercise testing**. *Am J Respir Crit Care Med* (2003) **167** 1451. DOI: 10.1164/ajrccm.167.10.950
23. Ryan CG, Grant PM, Tigbe WW, Granat MH. **The validity and reliability of a novel activity monitor as a measure of walking**. *Br J Sports Med* (2006) **40** 779-784. DOI: 10.1136/bjsm.2006.027276
24. Karvonen MJ, Kentala E, Mustala O. **The effects of training on heart rate; a longitudinal study**. *Ann Med Exp Biol Fenn* (1957) **35** 307-315. PMID: 13470504
25. Battelino T, Danne T, Bergenstal RM. **Clinical targets for continuous glucose monitoring data interpretation: recommendations from the international consensus on time in range**. *Diabetes Care* (2019) **42** 1593-1603. DOI: 10.2337/dci19-0028
26. van Faassen M, Bischoff R, Eijkelenkamp K, de Jong WHA, van der Ley CP, Kema IP. **In matrix derivatization combined with LC-MS/MS results in ultrasensitive quantification of plasma free metanephrines and catecholamines**. *Anal Chem* (2020) **92** 9072-9078. DOI: 10.1021/acs.analchem.0c01263
27. van Meijel LA, Tack CJ, de Galan BE. **Effect of short-term use of dapagliflozin on impaired awareness of hypoglycaemia in people with type 1 diabetes**. *Diabetes Obes Metab* (2021) **23** 2582-2589. DOI: 10.1111/dom.14505
28. McCarthy O, Deere R, Churm R. **Extent and prevalence of post-exercise and nocturnal hypoglycemia following peri-exercise bolus insulin adjustments in individuals with type 1 diabetes**. *Nutr Metab Cardiovasc Dis* (2021) **31** 227-236. DOI: 10.1016/j.numecd.2020.07.043
29. Campbell MD, Walker M, Trenell MI. **A low-glycemic index meal and bedtime snack prevents postprandial hyperglycemia and associated rises in inflammatory markers, providing protection from early but not late nocturnal hypoglycemia following evening exercise in type 1 diabetes**. *Diabetes Care* (2014) **37** 1845-1853. DOI: 10.2337/dc14-0186
30. Mathieu C, Hollander P, Miranda-Palma B. **Efficacy and safety of insulin degludec in a flexible dosing regimen vs insulin glargine in patients with type 1 diabetes (BEGIN: Flex T1): a 26-week randomized, treat-to-target trial with a 26-week extension**. *J Clin Endocrinol Metab* (2013) **98** 1154-1162. DOI: 10.1210/jc.2012-3249
31. Walsh J, Roberts R, Heinemann L. **Confusion regarding duration of insulin action: a potential source for major insulin dose errors by bolus calculators**. *J Diabetes Sci Technol* (2014) **8** 170-178. DOI: 10.1177/1932296813514319
32. Plank J, Bodenlenz M, Sinner F. **A double-blind, randomized, dose-response study investigating the pharmacodynamic and pharmacokinetic properties of the long-acting insulin analog detemir**. *Diabetes Care* (2005) **28** 1107-1112. DOI: 10.2337/diacare.28.5.1107
|
---
title: Bile salt hydrolase of Lactiplantibacillus plantarum plays important roles
in amelioration of DSS-induced colitis
authors:
- Xin Feng
- Zichen Bu
- Hongyu Tang
- Yongjun Xia
- Xin Song
- Lianzhong Ai
- Guangqiang Wang
journal: iScience
year: 2023
pmcid: PMC9988676
doi: 10.1016/j.isci.2023.106196
license: CC BY 4.0
---
# Bile salt hydrolase of Lactiplantibacillus plantarum plays important roles in amelioration of DSS-induced colitis
## Summary
Bile salt hydrolases are thought to be the gatekeepers of bile acid metabolism. To study the role of BSH in colitis, we investigated the ameliorative effects of different BSH-knockout strains of *Lactiplantibacillus plantarum* AR113. The results showed that L. plantarum Δbsh 1 and Δbsh 3 treatments did not improve body weight and alleviate the hyperactivated myeloperoxidase activity to the DSS group. However, the findings for L. plantarum AR113, L. plantarum Δbsh 2 and Δbsh 4 treatments were completely opposite. The double and triple bsh knockout strains further confirmed that BSH 1 and BSH 3 are critical for the ameliorative effects of L. plantarum AR113. In addition, L. plantarum Δbsh 1 and Δbsh 3 did not significantly inhibit the increase in pro-inflammatory cytokines or the decrease in an anti-inflammatory cytokine. These results suggest that BSH 1 and BSH 3 in L. plantarum play important roles in alleviating enteritis symptoms.
## Graphical abstract
## Highlights
•AR113 could alleviate the disease of mice with DSS-induced colitis•L. plantarum Δbsh 1 and Δbsh 3 could not regulate inflammatory cytokine-related genes•Bsh 1 and bsh 3 are critical for ameliorative effects in colitis
## Abstract
Microbiology; Microbiome; Cell biology
## Introduction
Inflammatory bowel diseases (IBDs) are mainly divided into ulcerative colitis (UC) and Crohn’s disease, the typical symptoms of which include abdominal pain, diarrhea, blood in the stool, loss of appetite, fatigue, and weight loss.1 The incidence and prevalence of IBD have increased markedly in recent years, making IBD a global public health problem, which has attracted worldwide attention. Although the exact cause of IBD remains largely unknown, several studies have suggested that it involves a complex interaction between genetic, microbial, and immune factors.1,2,3,4,5 Recent research has revealed that of these factors, the composition of the intestinal microbiota is especially important in IBD pathogenesis.3,4,5 Some recent studies have suggested that the intestinal microbiota affects IBD pathogenesis through metabolites such as short-chain fatty acids (SCFAs) and bile acids (BAs).6,7,8,9,10,11,12 SCFAs metabolized by intestinal bacteria have been shown to ameliorate the typical symptoms in animal models of IBD and have been associated with a reduced risk of IBD.7,8,9,10 In addition, some recent studies have found that changes in BAs and BA metabolism are also linked to IBD.10,11,12,13 Studies have found that although the levels of primary BAs and conjugated BAs (CBAs) are augmented in the fecal samples of healthy individuals, the levels of secondary BAs are decreased in the fecal samples of IBD patients, however, it is not clear why IBD patients manifest these changes and what role these changes play in IBD pathogenesis.11,12 Intestinal microorganisms can promote the conversion of primary BAs to secondary BAs, thus changing the composition of BAs in the body. As bile salt hydrolases (BSHs) can cleave the amide bond in conjugated BAs to increase the concentration of deconjugated BAs, which can subsequently undergo a variety of transformations to generate secondary BAs, they are thought to be the gatekeepers of BA metabolism and host microbiome crosstalk in the gastrointestinal tract.14,15 Labbe et al.13 found a significant reduction in the abundance of the Firmicutes-derived BSH (bsh) gene in IBD patients relative to healthy controls, using large available datasets containing metagenomic information from IBD patients. Studies have also reported a significant negative correlation between the relative abundance of bacterial bsh genes and the retinoic acid receptor-related orphan receptor gamma (RORC) gene, which is related to IBD.11,16 Using chemoproteomic approaches, Parasar et al.17 identified altered BSH activities in a murine model of IBD. In silico analysis showed that the relative abundance of BSH in the gut microbiota was markedly lower in IBD patients than in healthy individuals, and that this reduction was most evident in Firmicutes from the patients.17 Although an association of BSH in the gut microbiota with IBD has been suggested, further research is warranted to provide the evidence.
Current treatment strategies for IBD focus on reducing the inflammatory burden in patients with active disease and maintaining remission in those with inactive disease.18,19 However, they are associated with a wide range of possible severe side effects, and some of these treatments are costly.18,19,20 *As a* result, probiotics are attracting increasing research interest as a safer way of ameliorating disease activity.20 Several studies have found that some probiotics can ameliorate or prevent IBD.20,21,22,23,24 Although the mechanisms of the ameliorative effects of probiotics on IBD are not well defined, many mechanisms have been proposed to explain this action, including antagonism to pathogenic bacteria, modulation of gut microbiota, production of nutrients, and enhancement of anti-inflammatory cytokine levels.20,21,22,23,24 A meta-analysis showed that probiotic VSL#3 is effective in inducing remission in active UC and is equivalent to 5-aminosalicylic acid (5-ASA) in preventing UC relapse.20 Wang et al.21 found that *Lactiplantibacillus plantarum* ZS2058 was an efficient producer of conjugated linoleic acid (CLA) in vitro and that it ameliorated dextran sulfate sodium (DSS)-induced acute colitis by producing CLA locally in mice. Clostridium butyricum producing high levels of SCFAs was revealed to alleviate epithelial damage in rats with DSS-induced colitis.22 Ke et al.23 demonstrated that fucose ameliorated intestinal inflammation by regulating the crosstalk between BAs and the gut microbiota in DSS-treated mice. However, few studies have evaluated the role of BSH in the ameliorative effect of probiotics on IBD. The family *Lactobacillaceae is* rich in BSHs. A total of 551 BSHs from 107 Lactobacillaceae species were identified from 451 genomes of 158 Lactobacillaceae species.24 Through metagenomic analyses, Jones et al.15 demonstrated that BSH activity is a conserved microbial adaptation to the human gut environment with a high level of redundancy in this ecosystem. Therefore, it is necessary to investigate whether probiotics exert ameliorative effects on IBD through BSHs. Our previous work found that L. plantarum AR113 was more effective than other probiotic strains in alleviating epithelial damage, improving colon length, and maintaining the epithelial barrier integrity after DSS treatment.25 We also found that L. plantarum AR113 had the highest BSH activity among the 10 lactic acid bacterial strains tested with the plate assay. Furthermore, using in silico molecular docking, heterologous expression, and knockout experiments, we verified that the bsh 1 and bsh 3 genes were responsible for most of the BSH activity in L. plantarum AR113.26,27 To explore the roles of BSHs in IBD, here we investigated the ameliorative effects of BSH-knockout strains of L. plantarum AR113 on the disease severity of mice with DSS-induced colitis.
## BSH improves colitis symptoms
To explore the roles of BSHs in IBD, the effects of orally administrated BSH-knockout strains on the severity of DSS-induced colitis in mice were evaluated based on body weight loss, DAI, colon length, MPO activity, and histopathology. Mice were fed according to the protocol outlined in Figure 1.Figure 1Schematic diagram of the animal experiment The DSS group showed severe body weight loss from day 4 and demonstrated a final weight loss of $15\%$ relative to the Control group. In the L. plantarum AR113 group, the DSS-treated mice showed evident recovery of body weight, with their weights being almost the same as those in the 5-ASA group and even approaching the Control group (Figure 2A). Compared with L. plantarum AR113, L. plantarum Δbsh 1 and L. plantarum Δbsh 3 led to little recovery of body weights. However, similar to L. plantarum AR113, L. plantarum Δbsh 2 and L. plantarum Δbsh 4 effectively ameliorated the body weight loss of DSS-treated mice (Figures 2A and S1A). Further analysis revealed that the L. plantarum Δbsh 13, L. plantarum Δbsh 132, and L. plantarum Δbsh 134 groups had significant body weight loss with no recovery, similar to that in the DSS group (Figures 2B and S1B). These findings suggest that BSH 1 and BSH 3, not BSH 2 and BSH 4, in L. plantarum AR113 play important roles in the recovery of body weight in DSS-treated mice. Figure 2Physical and chemical indicators of colitis in mice(A) Body weight, (B) body weight at day 12, (C) disease activity index (DAI), (D) DAI at day 12, (E) colon length, (F) myeloperoxidase (MPO) activity. The red bar indicates the group with bsh 2 or bsh 4 knocked out, and the blue bar indicates the group with bsh 1 or bsh 3 knocked out. The statistical significance between the data was assessed using One-way ANOVA by Dunnett’s tests, ∗: $p \leq 0.05$, ∗∗: $p \leq 0.01.$ *All data* are presented as the mean ± standard error of the mean ($$n = 8$$ mice per group).
Compared with the DSS group, the L. plantarum AR113 group showed significantly reduced DAI values (10.21 ± 0.15 vs 4.20 ± 0.40, $p \leq 0.05$) and the L. plantarum Δbsh 2 and L. plantarum Δbsh 4 groups also showed alleviated symptoms and significantly reduced DAI values (Figures 2C, 2D and S1C). However, the DAI values of the L. plantarum Δbsh 1 (6.28 ± 0.26), L. plantarum Δbsh 3 (5.80 ± 0.23), L. plantarumΔbsh 13 (6.00 ± 0.16), L. plantarum Δbsh 13 (5.81 ± 0.22), and L. plantarum Δbsh 134 (6.03 ± 0.17) groups did not differ significantly from one another and were significantly higher than that of the L. plantarum AR113 group (4.20 ± 0.40, $p \leq 0.05$, Figures 2D and S1D).
The colon lengths of mice in different groups were measured after euthanizing the mice (Figure 2E). Compared with the Control group (6.82 ± 0.31), the DSS group (5.90 ± 0.53) showed significantly decreased colon length. The L. plantarum AR113 treatment group (7.1 ± 0.18) effectively prevented the DSS-induced shortening of the colon and recovered the colon length to that observed in the Control and 5-ASA groups. The colon lengths in the L. plantarum Δbsh 1, L. plantarum Δbsh 3, L. plantarum Δbsh 13, L. plantarum Δbsh 132, and L. plantarum Δbsh 134 groups were significantly reduced relative to the L. plantarum AR113 treatment group and were not significantly different from one another.
Figure 2F shows that the MPO activity in the DSS group (1.24 ± 0.04 U/g) was significantly higher than that in the Control group (0.85 ± 0.04 U/g), indicating that the colons of DSS-treated mice had significant neutrophil infiltration and severe inflammation. L. plantarum AR113 (0.82 ± 0.03 U/g) treatment reduced the DSS-induced MPO activity to almost the level in the Control group. Similar to L. plantarum AR113, L. plantarum Δbsh 2 and L. plantarum Δbsh 4 inhibited the MPO activity in DSS-treated mice. In contrast, L. plantarum Δbsh 1 and L. plantarum Δbsh 3 had no inhibitory effect on the MPO activity. The L. plantarum Δbsh 13, L. plantarum Δbsh 132, and L. plantarum Δbsh 134 groups also showed no suppression of MPO activity relative to the DSS group. These findings suggest that L. plantarum AR113 uses BSH 1 or BSH 3 to alleviate the hyperactivated MPO activity in DSS-treated mice.
The effect of each treatment on the pathological changes in mouse colon tissues was observed by hematoxylin and eosin (H&E) staining (Figures 3A–3K). In the Control group (Figure 3A), the intestinal epithelium was structurally intact; the complete crypt structure and many goblet cells were retained. In contrast, the DSS group showed complete destruction of the intestinal epithelial structure, loss of the crypt structure, and infiltration of a large number of inflammatory cells in the colon (Figure 3B). Similar to that in the 5-ASA group (Figure 3C), the L. plantarum AR113 (Figure 3D), L. plantarum Δbsh 2 (Figure 3F), and L. plantarum Δbsh 4 (Figure 3H) groups showed significant alleviation of the DSS-induced colon tissue lesions, with histological injury scores of 4.86 ± 0.34, 5.67 ± 0.87, 5.34 ± 0.38, and 4.23 ± 0.40 (all $p \leq 0.05$), respectively, which were significantly lower than the score of the DSS group (Figure 3L). In contrast, the L. plantarum Δbsh 1, L. plantarum Δbsh 3, L. plantarum Δbsh 13, L. plantarum Δbsh 132, and L. plantarum Δbsh 134 groups showed destruction of the intestinal epithelial structure, loss of the crypt structure, and colonic infiltration of inflammatory cells, similar to those in the DSS group. Figure 3Effects of *Lactiplantibacillus plantarum* AR113 and its seven mutant strains on the colon histopathology of mice with dextran sulfate sodium (DSS)-induced colitis(A–K) Representative histological examination micrographs of colon tissue section slides. Scale bars, 200 mm.(L) Histological injury scores for the mouse colon tissues. The red bar indicates the group with bsh 2 or bsh 4 knocked out, which does not affect the function of AR113, and the blue bar indicates the group with bsh 1 or bsh 3 knocked out. The statistical significance between the data was assessed using One-way ANOVA by Dunnett’s tests, ∗: $p \leq 0.05$, ∗∗: $p \leq 0.01.$
## Effect of BSHs on the inflammatory cytokine expression
Changes in the levels of pro-inflammatory cytokines (TNF-α, IL-1β, and IL-6) and the anti-inflammatory cytokine IL-10 in mouse colon tissues were measured to determine the anti-inflammatory effect of bsh. The sequences of primers are listed in Table 1. As shown in Figure 4, compared with the Control group, the DSS group demonstrated a significant increase in the transcript levels of the colonic pro-inflammatory cytokines TNF-α, IL-1β, and IL-6, but a significant decrease in the transcript level of the anti-inflammatory cytokine IL-10. Similar to the 5-ASA group, the L. plantarum AR113 group showed significantly downregulated expression of the pro-inflammatory cytokines and upregulated expression of the anti-inflammatory cytokine relative to the DSS group. Similar to L. plantarum AR113, L. plantarum Δbsh 2 and L. plantarum Δbsh 4 also downregulated the expression of the pro-inflammatory cytokines and upregulated the expression of the anti-inflammatory cytokine. This indicates that BSH 2 and BSH 4 do not affect the inflammatory cytokine expression. In contrast to the L. plantarum AR113 group, the L. plantarum Δbsh 1 and L. plantarum Δbsh 3 groups showed no significant difference in the expression of most of inflammatory cytokines relative to the DSS group. Furthermore, the cytokine expressions in the L. plantarum Δbsh 13, L. plantarum Δbsh 132, and L. plantarum Δbsh 134 groups were not significantly different from one another or from those in the DSS group. These findings suggest that BSH 1 and BSH 3 in L. plantarum AR113 play crucial roles in regulating the inflammatory cytokine expression in the colon tissues of DSS-treated mice. Table 1Primers sequences use in this studyPrimersForward primers (5′-3′)Reverse primers (5′-3′)RFPinsertTAGGACTAACTCTACCGAAGCAGTCTTGTTCGGATTAATCβ-actinGGCTGTATTCCCCTCCATCGCCAGTTGGTAACAATGCCATGTTNF-αAGGGTCTGGGCCATAGAACTCCACCACGCTCTTCTGTCTACIL-6GAGGATACCACTCCCAACAGACCAAGTGCATCATCGTTGTTCATACAIL-1βCTGAACTCAACTGTGAAATGCTGATGTGCTGCTGCGAGAMUC2ATGCCCACCTCCTCAAAGACGTAGTTTCCGTTGGAACAGTGAAClaudin1CTGTGGATGTCCTGCGTTTCTCATGCACTTCATGCCAATGOccludinTGGCGGATATACAGACCCAACGATCGTGGCAATAAACACCZO-1CTTCTCTTGCTGGCCCTAAACTGGCTTCACTTGAGGTTTCTGFGFR4GTGGTCAGTGGGAAGTCTGGTTGTACCAGTGACGACCACGFGF15GACTGCGAGGAGGACCAAAACAGCCCGTATATCTTGCCGTCYP7A1AACAACCTGCCAGTACTAGATAGCGTGTAGAGTGAAGTCCTCCTTAGCSHPTCTGCAGGTCGTCCGACTATTCAGGCAGTGGCTGTGAGATGCFXRTGGGCTCCGAATCCTCTTAGATGGTCCTCAAATAAGATCCTTGGFigure 4mRNA expression of the cytokines(A) *Tumor necrosis* factor alpha (TNF-α), (B) interleukin (IL)-1β, (C) IL-6, and (D) IL-10 was quantified by RT-qPCR. The data are expressed as the mean ± standard deviation ($$n = 8$$ mice per group). The red bar indicates the group with bsh 2 or bsh 4 knocked out, which does not affect the function of AR113, and the blue bar indicates the group with bsh 1 or bsh 3 knocked out. The statistical significance between the data was assessed using One-way ANOVA by Dunnett’s tests, ∗: $p \leq 0.05$, ∗∗: $p \leq 0.01.$
## Effect of BSH on the expression of tight junction (TJ)-related genes
The relative expression of MUC2 secreted by the mouse colon tissues was analyzed by reverse transcription quantitative polymerase chain reaction (RT-qPCR). The primers information are listed in Table 1.The mRNA expression of MUC2 was significantly downregulated in the DSS group compared with the Control group (Figure 5A). Compared with the DSS group, the 5-ASA, L. plantarum AR113, L. plantarum Δbsh 2, and L. plantarum Δbsh 4 groups showed upregulated mRNA expression of MUC2. However, the mRNA expression of MUC2 in the L. plantarum Δbsh 1, L. plantarum Δbsh 13, L. plantarum Δbsh 132, and L. plantarum Δbsh 134 groups was not significantly different from that in the DSS group. Figure 5Expression of intestinal barrier function-related genes in mice(A) Claudin, (B) MUC2, (C) occludin, (D) ZO-1. The statistical significance between the data was assessed using One-way ANOVA by Dunnett’s tests, ∗: $p \leq 0.05$, ∗∗: $p \leq 0.01.$ *All data* are presented as the mean ± standard error of the mean ($$n = 8$$ mice per group). The red bar indicates the group with bsh 2 or bsh 4 knocked out, and the blue bar indicates the group with bsh 1 or bsh 3 knocked out.
The relative expression levels of TJ-related genes, namely, those encoding ZO-1, occludin, and claudin, in the mouse colon tissues were assessed to investigate the integrity of the intestinal epithelial barrier and epithelial structure. The DSS group showed downregulated mRNA expression of ZO-1, occludin, and claudin relative to the Control group. In contrast, 5-ASA and L. plantarum AR113 treatments alleviated the DSS-induced downregulation of occludin, ZO-1, and claudin expression (Figure 5B–5D). Similarly, the L. plantarum Δbsh 4 treatment upregulated ZO-1 and claudin expression and the L. plantarum Δbsh 2 treatment upregulated occludin and claudin expression relative to the DSS group. However, the TJ protein expressions in the L. plantarum Δbsh 1, L. plantarum Δbsh 13, L. plantarum Δbsh 132, and L. plantarum Δbsh 134 groups were not significantly different from those in the DSS group.
## Effect of BSH on the total bile acid and the BA-specific receptors expression
BA metabolism of IBD patients is damaged because of the impaired microbial enzyme activity, which result in the abnormal change of total bile acid (TBA). The ameliorative effect of BSH on IBD was further demonstrated by measuring the TBA in feces. The DSS group showed significantly higher of TBA compared with the Control group, the L. plantarum AR113 reduce effectively TBA to normal level (Figure 6A). However, the remission effect of L. plantarum Δbsh1 and L. plantarum Δbsh134 was significantly worse than that of the L. plantarum AR113 (Figure 6A).Figure 6Effect of BSH on the total bile acid(A) Determination of TBA and in mice feces.(B) Determination of fluorescence intensity in mice feces. The statistical significance between the data was assessed using One-way ANOVA by Dunnett’s tests, ∗: $p \leq 0.05$, ∗∗: $p \leq 0.01$, ∗∗∗:$p \leq 0.001$, ∗∗∗∗:$p \leq 0.0001.$Results are express as the mean ± standard deviation for each experimental group ($$n = 8$$).
To determine if the change of TBA is because of the strains number, red fluorescent protein (RFP) gene was successfully inserted into the L. plantarum AR113 and knockout strains genomes using CRISPR technology. There is no change of fluorescence intensity of L. plantarum AR113, L. plantarum Δbsh1 and L. plantarum Δbsh134 in mice feces on days 8, 10 and 12, which shows that there was no significant difference in the number of strains (Figure 6B). These findings show that the mutant strains survive in the mice to similar levels as the wild strain.
We further investigated the effect of BSH on the BA-specific receptors expression. Compared with the Control group, the expression of fgfr4, fgf15, shp and fxr are downregulated in DSS group significantly (Figures 7A–7D), but the expression of cyp7a1 is significantly increased (Figure 7E). L. plantarum AR113 treatments, in contrast, effectively alleviated the DSS-induced downregulated of fgfr4, fgf15, shp and upregulated of cyp7a1, even approaching to the normal levels (Figure 7). L. plantarum Δbsh 1 and L. plantarum Δbsh 134 groups have a similar regulatory effect, but their effect was significantly worse than that of L. plantarum AR113 and was closer to that in the DSS group (Figure 7). These results show that L. plantarum can regulate BA-specific receptors expression BA receptor expression through BSH.Figure 7mRNA expression of BA-related genes(A) fgfr4, (B) fgf15, (C) shp, (D) fxr, (E) cyp7a1 were quantified by RT-qPCR. The statistical significance between the data was assessed using One-way ANOVA by Dunnett’s tests, ∗: $p \leq 0.05$, ∗∗: $p \leq 0.01$, ∗∗∗:$p \leq 0.001$, ∗∗∗∗:$p \leq 0.0001.$ *The data* are expressed as the mean ± standard deviation for each experimental group ($$n = 8$$ mice per group).
## Discussion
In our study, L. plantarum AR113 relieved the symptoms of DSS-induced colitis in mice by reducing DAI values, inhibiting the hyperactivated MPO activity, and increasing the colon length, but the bsh 1 or bsh 3 knockout strains did not. To the best of our knowledge, this is the first study to demonstrate probiotics can exert ameliorative effects on IBD through BSHs. BSH is a crucial enzyme that catalyzes an essential gateway reaction in BA metabolism. Our research also found that L. plantarum can affect total bile acid by BSH. This suggests that probiotics may influence the occurrence of inflammation by affecting BA metabolism. Many studies have suggested that BSH activity and subsequent BA modification could significantly impact the pathophysiology of metabolic diseases, such as obesity, diabetes, and atherosclerosis, through perturbations of the BA pool.28,29 Therefore, it is worth further studying the effect of probiotics on these metabolic diseases by changing BA and establish the corresponding relationship between BSH activity and prebiotic function.
In addition, we demonstrated the role of BSH in alleviating inflammation, providing a basis for rational and accurate screening of probiotics. To our best knowledge, few studies have examined the active component of probiotic that alleviate inflammation by constructing knockout strains. SCFAs are metabolized by gut bacteria have been shown to ameliorate inflammatory.10 Although the exact source for the action of SCFA are still not clear, some probiotics that can reduce inflammation do not produce SCFAs. The exact function of SCFAs produced by probiotics has not been studied by knocking out SCFAs synthetic genes. We found that BSH is the active component in alleviating inflammation through single, double, and triple bsh knockout strains. So BSH activity can be included in the criteria for the anti-inflammatory probiotic strain selection.
In our study, L. plantarum AR113 treatment significantly downregulated the expression of TNF-α, IL-6, and IL-1β and upregulated that of IL-10, however, L. plantarum Δbsh 1 and L. plantarum Δbsh 3 treatments did not affect the inflammatory cytokine expression. This indicated that BSH are involved in the regulation of these factors. Some studies have demonstrated that BAs inhibit the secretion of TNF-α, IL-1β, and IL-6 in macrophages, and that this downregulation is mediated by the BA-specific membrane receptor TGR5.30 Some studies also found that BA-dependent farsenoid X receptor (FXR) activation appears to limit mucosal inflammatory responses via reducing pro-inflammatory cytokine (e.g., IL-1β, IL-6).31 *In this* study, L. plantarum AR113 significantly upregulated of FXR expression, whereas L. plantarum Δbsh1 and L. plantarum Δbsh134 did not. This suggests L. plantarum AR113 may modulate inflammatory balance by regulating bile salts. Parasar et al.17 identified altered BSH activities in a murine model of IBD, which led to changes in BA metabolism that could impact host metabolism and immunity. They also reported that Firmicutes-derived BSH was significantly reduced in IBD patients relative to healthy individuals. This suggests from another aspect that oral administration of L. plantarum can regulate BA-dependent immune responses to alleviate DSS-induced colitis through the BSH activity.
## Limitations of the study
This study focuses on the role of BSH in studying the ameliorative effects of BSH in enterohepatic circulation and the ameliorative effects of BSH on the disease severity of mice with DSS-induced colitis, whereas the mechanism of BSH alleviating the colitis needs more detailed research. It is well known that a balanced and intact intestinal environment is closely related to intestinal health and disease treatment; it would thus be interesting to investigate the changes of intestinal flora community, abundance changes and potential mechanisms.
## Key resources table
REAGENT or RESOURCESOURCEIDENTIIERBacterial and virus strainsLactiplantibacillus plantarumShanghai Engineering Research Center of Food Microbiology, University of Shanghai for Science and TechnologyAR113Chemicals, peptides, and recombinant proteinsHieff qPCR SYBR Green Master MixYeasen Biotechnology, China11201ES03HiScript Ⅲ RT SuperMix for qPCRVazyme Biotech, ChinaR323-01Total RNA Extractor (TRIzol)Shenggong bio, ChinaB511311-01005-Aminosalicylic acidShyuanyeS30083Critical commercial assaysTBA determination kit(Tongwei, China)TWP110234MPO Detection KitNanjing Jiancheng Bioengineering Institute, ChinaA044-1-1OligouncleotidesSee Table 1 for primersN/AN/A
## Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author, Professor Guangqiang Wang ([email protected]).
## Materials availability
This work did not generate new unique reagents. Primers sequences used were provided in Table 1, and available upon request to the corresponding author.
## Strains and culture conditions
L. plantarum AR113 and its seven bshmutant strains were obtained from Shanghai Engineering Research Center of Food Microbiology, University of Shanghai for Science and Technology (Shanghai, China).32 All strains were anaerobically cultured in deMan, Rogosa, and Sharpe (MRS) medium at 37°C for 16h before centrifugation (3,000 × g for 3minat 4°C). The cell pellets were washed twice with phosphate buffer solution (pH 7.4) and then resuspended at a density of approximately 5 × 109CFU/mL, which was determined by colony counting on MRS plates, for the following experiments. Each mouse was orally administered with 1 x109 CFU of the strains in 0.2mL (5 × 109CFU/mL) by oral gavage daily from day 5 to day 12 of the experiment.
## Animals and experimental design
Six-week-old male SPF C57BL/6 mice (weight, 18–20 g) were purchased from Shanghai SLRC Laboratory Animal Co. Ltd. (Shanghai, China). The mice were housed at a constant temperature of 23 ± 2°C with a 12-h dark/light cycle and given ad libitum access to standard chow and water at all times. All mice were allowed at least 1 week to adapt to the experimental environment before conducting the following experiments. The experimental procedures were performed in accordance with the institutional and governmental regulations on the ethical use of experimental animals.
The mice were randomly divided into 11 groups of eight (Figure 1). The Control group was given free access to sterile water and fed a normal diet for the whole experimental period (12 days). The DSS group was given free access to $2.5\%$ DSS-containing drinking water for the first 7 days and then fed normal diet and normal water for the next 5 days of the experiment. The DSS-treated mouse groups of L. plantarum AR113 and its seven bshmutant derivatives were administered the corresponding strains (0.2 mL containing 5 × 109CFU/mL) once a day by gavage from day 5 to day 12. The 5-ASA group of DSS-treated mice was administered 5-ASA (0.2mL) once a day by gavage from day 5 to day 12.
## Evaluation of the DAI
The DAI scores were calculated using a scoring system that takes into account weight loss, stool consistency, and hematochezia.20 First, an occult stool blood test was performed using the Occult Blood Kit (Beisuo, Zhuhai, China). Then, the following scoring criteria were used to record the DAI scores: A: weight loss (0, no loss; 1, $1\%$–$5\%$ loss; 2, $6\%$–$10\%$ loss; 3, $11\%$–$15\%$ loss; and 4, >$15\%$ loss); B: stool consistency (0, normal; 2, loose stools; 3 and 4, diarrhea); C: Occult blood or gross bleeding (0 and 1, negative; 2 and 3, hemoccult positive; and 4, gross bleeding).
## Assessment of MPO activity
The colonic MPO activity in each group of mice was determined using an MPO test kit (Nanjing Jiancheng Co., Ltd., Nanjing, China) according to the manufacturer’s instructions. The colon tissue samples were prepared as follows: colon tissue was accurately weighed, and then physiological saline was added as the homogenization medium in a weight-to-volume ratio of 1:19. A $5\%$ tissue homogenate was then prepared in a tissue-dispersing machine for MPO activity determination. MPO activity was measured in U/g of fresh colon tissue, with one unit of MPO defined as the amount required to degrade 1.0 μmol of hydrogen peroxide per minute at 37 °C.
## H&E staining
The colon morphology and histopathological lesions were assessed with H&E staining (Murthy, Cooper, & Shim, 1993). The colon tissues were first cut into 5-mm-thick slices. The slices were then fixed in neutral buffered formalin for 24 h, dehydrated with graded alcohol ($75\%$–$100\%$) solutions, and then embedded in paraffin wax for H&E staining. The severity of colonic histological injury in each mouse was scored using a modified scoring system that took into account the degree of inflammation, mucosal damage, crypt damage, and range of pathological changes.
## RNA extraction and RT-qPCR
Total RNA was extracted from each colon tissue sample using TRIzol reagent (Shenggong bio, Shanghai, China). cDNA was then prepared by reverse transcription using a HiScript Ⅲ RT SuperMix for qPCR (Vazyme Biotech, China) and amplified by RT-qPCR using Hieff qPCR SYBR Green Master Mix (Yeasen Biotech, China) and the appropriate primers (Table 1). The conditions were 40 cycles of 95 °C for 30 s, 95 °C for 5 s, and 60 °C for 30 s. The primers used in this study were synthesized by Beijing Genomics (Shanghai, China). The RNA expression levels of the relevant genes in each group were measured using the 2-ΔΔCt method, where Δ Ct represents the difference in the Ct values between the target gene and the β-actin reference gene. The β-actin mRNA levels in the test groups are expressed as the ratio of its expression in the test group relative to that in the DSS group.
## Fecal TBAs concentration analysis
Faecal samples were prepared for TBA analysis. Mice faeces that were stored at -80 °C were thawed in an ice bath to reduce sample degradation, 100 mg/mL of feces in each group were homogenate with ethanol, which were centrifuged at 4 °C for 10minat 12000 rpm, and the supernatant were collected and added to the reagent according to the instructions of TBA determination kit (TONGWEI, CHINA), the optical density at 405 nm was measured after 3 min for reaction, and content of TBA was calculated by the following equation. TBA=C1xA2−A0A1−A0xVW in which, C1 is the standard quality concentration, A2 represents the absorbance at time $t = 3$ min and A1 represents the absorbance at $t = 1.5$ min, and A0 represents the optical density of blank. V is the volume of the ethanol, W is the weight of sample.
## Faecal fluorescence intensity analysis
As described by Qin,33 red fluorescent protein (RFP) gene was inserted into L. plantarum AR113, L. plantarum Δbsh 1, L. plantarum Δbsh 134 with the CRISPR/Cas 9 gene editing tool. The upstream, downstream homologous arms of the insert plasmids and red fluorescent protein gene (RFP) were amplified by PCR using the L. plantarum AR113 genome and pIB184-RFP, which were ligated to the sgRNA amplified by PCR using plasmid pHSP01 as template to obtain up-RFP-down-sgRNA fragment. The fragment was cloned with the editing plasmid pHSP01 skeleton (digested by ApaI and XbaI) by one-step cloning to obtain the insertion plasmid, and the insertion plasmid was then introduced into L. plantarum AR113, L. plantarum Δbsh 1 and L. plantarum Δbsh 134, cultured at 37 °C for 48 h and screened in erythromycin. The used primer RFPinsert was listed in Table 1.
The mice were randomly divided into 5 groups ($$n = 8$$). The control group was given free access to sterile water and fed a normal diet for the whole experimental period (12 days). The DSS group was given free access to $2.5\%$ DSS-containing drinking water for the first 7 days and then fed normal diet and normal water for the next 5 days of the experiment. The DSS-treated mouse groups of L. plantarum AR113-RFP, L. plantarum Δbsh 1-RFP and L. plantarum Δbsh 134-RFP with 1х109 CFU, once a day by gavage from day 5 to day 12. Faecal samples were collected on day 8, 10 and 12, and weighed as soon as possible to dissolved into $0.6\%$ faecal homogenate in sterile water. The supernatant was collected after centrifugation at 4000g for 10 min, subsequently, 200 μL of the supernatant was added to a 96-well plate, in addition to sterile water as the standard, the absorbance (excitation wavelength, 600 nm; emission wavelength, 635 nm) was measured using a fluorescence spectrophotometer (SpectraMax i3x, Molecular Devices, San Jose, CA, USA).
## Quantification and statistical analysis
Statistical analyses were performed using SPSS software (SPSS Inc, Chicago, U.S.A.) or GraphPad Prism 5 (GraphPad Software, Inc, San Diego, U.S.A.). All data are presented as the mean ± standard deviation ($$n = 8$$ per group). The statistical significance of the results was analyzed by a one-way analysis of variance followed by a Tukey’s post hoc test. A p-value < 0.05 was considered significant.
## Supplemental information
Document S1. Figures S1
## Data and code availability
The data reported in this paper will be shared upon request to the lead corresponding author ([email protected]).This paper dose not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.
## Author contributions
All the authors contributed extensively to the work presented in this paper. A.L.Z. and W.G.Q. designed the project and experiments. F.X. and T.H.Y. performed the experiments and analyzed the data. B.Z.C. wrote the draft. W.G.Q. and F.X. wrote the manuscript. X.Y.J. and S.X. modified the paper. All authors critically reviewed and approved the final version of the manuscript.
## Declaration of interests
The authors declare no competing interests.
## References
1. Hoentjen F., Dieleman L., Gibson G.R., Roberfriod M.B.. *Handbook of Prebiotics* (2008) 341-373. DOI: 10.1201/9780849381829.ch17
2. Hodson R.. **Inflammatory bowel disease**. *Nature* (2016) **540** S97. DOI: 10.1038/540s97a
3. Mravec B.. **Pathophysiology of inflammatory bowel diseases**. *N. Engl. J. Med.* (2021) **384** 1377-1378. DOI: 10.1056/NEJMc2101562
4. Lane E.R., Zisman T.L., Suskind D.L.. **The microbiota in inflammatory bowel disease: current and therapeutic insights**. *J. Inflamm. Res.* (2017) **10** 63-73. DOI: 10.2147/JIR.S116088
5. Morgan X.C., Tickle T.L., Sokol H., Gevers D., Devaney K.L., Ward D.V., Reyes J.A., Shah S.A., LeLeiko N., Snapper S.B.. **Dysfunction of the intestinal microbiome in inflammatory bowel disease and treatment**. *Genome Biol.* (2012) **13** 1-18. DOI: 10.1186/gb-2012-13-9-r79
6. Lavelle A., Sokol H.. **Gut microbiota-derived metabolites as key actors in inflammatory bowel disease**. *Nat. Rev. Gastroenterol. Hepatol.* (2020) **17** 223-237. DOI: 10.1038/s41575-019-0258-z
7. Zhang Z., Zhang H., Chen T., Shi L., Wang D., Tang D.. **Regulatory role of short-chain fatty acids in inflammatory bowel disease**. *Cell Commun. Signal.* (2022) **20** 64. DOI: 10.1186/s12964-022-00869-5
8. Gonçalves P., Araújo J.R., Di Santo J.P.. **A cross-talk between microbiota-derived short-chain fatty acids and the host mucosal immune system regulates intestinal homeostasis and inflammatory bowel disease**. *Inflamm. Bowel Dis.* (2018) **24** 558-572. DOI: 10.1093/ibd/izx029
9. Russo E., Giudici F., Fiorindi C., Ficari F., Scaringi S., Amedei A.. **Immunomodulating activity and therapeutic effects of short chain fatty acids and tryptophan post-biotics in inflammatory bowel disease**. *Front. Immunol.* (2019) **10** 2754. DOI: 10.3389/fimmu.2019.02754
10. Sun M., Wu W., Liu Z., Cong Y.. **Microbiota metabolite short chain fatty acids, GPCR, and inflammatory bowel diseases**. *J. Gastroenterol.* (2017) **52** 1-8. DOI: 10.1007/s00535-016-1242-9
11. Biagioli M., Marchianò S., Carino A., Di Giorgio C., Santucci L., Distrutti E., Fiorucci S.. **Bile acids activated receptors in inflammatory bowel disease**. *Cells* (2021) **10** 1-16. DOI: 10.3390/cells10061281
12. Li N., Zhan S., Tian Z., Liu C., Xie Z., Zhang S., Chen M., Zeng Z., Zhuang X.. **Alterations in bile acid metabolism associated with inflammatory bowel disease**. *Inflamm. Bowel Dis.* (2021) **27** 1525-1540. DOI: 10.1093/ibd/izaa342
13. Labbé A., Ganopolsky J.G., Martoni C.J., Prakash S., Jones M.L.. **Bacterial bile metabolising gene abundance in Crohn’s, Ulcerative Colitis and Type 2 diabetes metagenomes**. *PLoS One* (2014) **9** 1-17. DOI: 10.1371/journal.pone.0115175
14. Foley M.H., O’Flaherty S., Barrangou R., Theriot C.M.. **Bile salt hydrolases: gatekeepers of bile acid metabolism and host-microbiome crosstalk in the gastrointestinal tract**. *PLoS Pathog.* (2019) **15** 1-6. DOI: 10.1371/journal.ppat.1007581
15. Jones B.V., Begley M., Hill C., Gahan C.G.M., Marchesi J.R.. **Functional and comparative metagenomic analysis of bile salt hydrolase activity in the human gut microbiome**. *Proc. Natl. Acad. Sci. USA* (2008) **105** 13580-13585. DOI: 10.1073/pnas.0804437105
16. Hernandez-Rocha C., Borowski K., Turpin W., Smith M., Stempak J., Silverberg M.S.. **A9 bacterial bile salt hydrolase gene abundance is associated with rorc gene expression in intestinal mucosa of inflammatory disease patients**. *J. Can. Assoc. Gastroenterol.* (2020) **3** 11-12. DOI: 10.1093/jcag/gwz047.008
17. Parasar B., Zhou H., Xiao X., Shi Q., Brito I.L., Chang P.V.. **Chemoproteomic profiling of gut microbiota-associated bile salt hydrolase activity**. *ACS Cent. Sci.* (2019) **5** 867-873. DOI: 10.1021/acscentsci.9b00147
18. Greuter T., Rieder F., Kucharzik T., Peyrin-Biroulet L., Schoepfer A.M., Rubin D.T., Vavricka S.R.. **Emerging treatment options for extraintestinal manifestations in IBD**. *Gut* (2021) **70** 796-802. DOI: 10.1136/gutjnl-2020-322129
19. Na S.Y., Moon W.. **Perspectives on current and novel treatments for inflammatory bowel disease**. *Gut Liver* (2019) **13** 604-616. DOI: 10.5009/gnl19019
20. Derwa Y., Gracie D.J., Hamlin P.J., Ford A.C.. **Systematic review with meta-analysis: the efficacy of probiotics in inflammatory bowel disease**. *Aliment. Pharmacol. Ther.* (2017) **46** 389-400. DOI: 10.1111/apt.14203
21. Wang J., Chen H., Yang B., Gu Z., Zhang H., Chen W., Chen Y.Q.. *RSC Adv.* (2016) **6** 14457-14464. DOI: 10.1039/C5RA24491A
22. Araki Y., Andoh A., Takizawa J., Takizawa W., Fujiyama Y.. *Int. J. Mol. Med.* (2004) **13** 577-580. DOI: 10.3892/ijmm.13.4.577
23. Ke J., Li Y., Han C., He R., Lin R., Qian W., Hou X.. **Fucose ameliorate intestinal inflammation through modulating the crosstalk between bile acids and gut microbiota in a chronic colitis murine model**. *Inflamm. Bowel Dis.* (2020) **26** 863-873. DOI: 10.1093/ibd/izaa007
24. Liang L., Yi Y., Lv Y., Qian J., Lei X., Zhang G.. **A comprehensive genome survey provides novel insights into bile salt hydrolase (BSH) in**. *Molecules* (2018) **23** 1-12. DOI: 10.3390/molecules23051157
25. Xia Y., Chen Y., Wang G., Yang Y., Song X., Xiong Z., Zhang H., Lai P., Wang S., Ai L.. *J. Funct.Foods* (2020) **67** 1-13. DOI: 10.1016/j.jff.2020.103854
26. Xiong Z.Q., Wang Q.H., Kong L.H., Song X., Wang G.Q., Xia Y.J., Zhang H., Sun Y., Ai L.Z.. **Short communication: improving the activity of bile salt hydrolases in**. *J. Dairy Sci.* (2017) **100** 975-980. DOI: 10.3168/jds.2016-11720
27. Wang G., Yu H., Feng X., Tang H., Xiong Z., Xia Y., Ai L., Song X.. **Specific bile salt hydrolase genes in**. *LWT* (2021) **145** 1-8. DOI: 10.1016/j.lwt.2021.111208
28. Kiriyama Y., Nochi H.. **Physiological role of bile acids modified by the gut microbiome**. *Microorganisms* (2021) **10** 68. DOI: 10.3390/microorganisms10010068
29. Di Ciaula A., Garruti G., Lunardi Baccetto R., Molina-Molina E., Bonfrate L., Wang D.Q.H., Portincasa P.. **Bile acid physiology**. *Ann. Hepatol.* (2017) **16** S4-S14. DOI: 10.5604/01.3001.0010.5493
30. Ichikawa R., Takayama T., Yoneno K., Kamada N., Kitazume M.T., Higuchi H., Matsuoka K., Watanabe M., Itoh H., Kanai T.. **Bile acids induce monocyte differentiation toward interleukin-12 hypo-producing dendritic cells via a TGR5-dependent pathway**. *Immunology* (2012) **136** 153-162. DOI: 10.1111/j.1365-2567.2012.03554.x
31. Chen M.L., Takeda K., Sundrud M.S.. **Emerging roles of bile acids in mucosal immunity and inflammation**. *Mucosal Immunol.* (2019) **12** 851-861. DOI: 10.1038/s41385-019-0162-4
32. Yue B., Luo X., Yu Z., Mani S., Wang Z., Dou W.. **Inflammatory bowel disease: a potential result from the collusion between gut microbiota and mucosal immune system**. *Microorganisms* (2019) **7** 440. DOI: 10.3390/microorganisms7100440
33. Qin W., Xia Y., Xiong Z., Song X., Ai L., Wang G.. **The intestinal colonization of**. *Food Res. Int.* (2022) **157** 1-12. DOI: 10.1016/j.foodres.2022.111382
|
---
title: Thrombocytopenia and insufficient thrombopoietin production in human small-for-gestational-age
infants
authors:
- Satoru Takeshita
- Hiroki Kakita
- Shimpei Asai
- Takafumi Asai
- Mari Mori
- Hiroko Ueda
- Hiromasa Aoki
- Mineyoshi Aoyama
- Yasumasa Yamada
journal: Pediatric Research
year: 2022
pmcid: PMC9988681
doi: 10.1038/s41390-022-02107-7
license: CC BY 4.0
---
# Thrombocytopenia and insufficient thrombopoietin production in human small-for-gestational-age infants
## Abstract
### Background
Small-for-gestational-age (SGA) infants are at increased risk for transient thrombocytopenia. The aim of this study was to determine whether thrombocytopenia in human SGA infants is due to insufficient thrombopoietin (TPO) production.
### Methods
A prospective study of 202 infants with gestational age less than 37 weeks was conducted; 30 of them were SGA infants, and 172 were non-SGA infants. Thrombocytopenia was seen in 17 of 30 SGA infants and 40 of 172 non-SGA infants.
### Results
Platelet counts were significantly lower in the SGA group than in the non-SGA group at the time of the lowest platelet count within 72 h of birth. The platelet count and immature platelet fraction (IPF) were negatively correlated in non-SGA infants, but not in SGA infants. In addition, the platelet count and TPO were negatively correlated in non-SGA infants. IPF and TPO were significantly lower in SGA than in non-SGA infants with thrombocytopenia.
### Conclusion
IPF increased with thrombocytopenia to promote platelet production in non-SGA infants due to increasing TPO, but not in SGA infants. This study found an association between insufficient TPO production and thrombocytopenia in SGA infants. In addition, this study is important for understanding the etiology of thrombocytopenia in SGA infants.
### Impact
The immature platelet fraction was low, and serum thrombopoietin was not increased in small-for-gestational-age (SGA) infants with thrombocytopenia. Thrombocytopenia in SGA infants is due to insufficient thrombopoietin production. This study is important for understanding the etiology of thrombocytopenia in SGA infants.
## Introduction
Intrauterine growth restriction (IUGR) occurs in approximately $15\%$ of births worldwide.1,2 IUGR is characterized by a restrictive environment that prevents the fetus from meeting its genetic potential for growth, and it occurs often in infants who are small for gestational age (SGA).2 SGA is defined as a birth weight of less than the 10th percentile for gestational age.3 SGA infants are at increased risk of transient thrombocytopenia.3,4 Some reports showed that 31–$53\%$ of SGA infants developed thrombocytopenia, generally defined as a platelet count less than 150 × 103/µL, within the first week after birth.3,4 Christensen et al. reported that, in SGA infants with thrombocytopenia, the lowest platelet counts were typically on day 4, with a mean nadir of 93 × 103/µL, and that the platelet count increased to ≥150 × 103/µL by day 14 in half of infants.3 The cause of thrombocytopenia in SGA infants has been postulated to be a decrease in platelet production.5–8 Sola et al. reported that thrombocytopenic SGA infants had low TPO concentrations and decreased marrow megakaryocytes.6 Murray et al. studied circulating burst-forming unit-megakaryocytes/colony-forming unit-megakaryocytes, total cultured megakaryocyte precursors, and mature megakaryocytes in most of the preterm infants with thrombocytopenia who were growth restricted.7 They suggested that the abnormal hematological characteristics of newborns with intrauterine growth retardation are a consequence of dysregulation of fetal hemopoiesis occurring proximal to committed megakaryocyte and neutrophil progenitors, most likely at the level of the primitive multipotent hemopoietic stem cell.7 Watts et al. reported that platelet counts and megakaryocyte numbers were significantly lower in premature infants than in controls on day 1, and TPO levels at the platelet nadir were significantly lower in neonates than in children.8 They suggested that preterm infants have an impaired TPO response to thrombocytopenia. We have previously reported animal studies that showed that chronic hypoxia in utero causes immaturity of liver function and a decrease in TPO expression in the liver, which in turn suppresses platelet production.9 However, the etiology of thrombocytopenia in human SGA infants remains unclear. The present study attempted to demonstrate that thrombocytopenia in human SGA infants is due to insufficient TPO production.
## Methods
A prospective study of infants admitted to the Aichi Medical University Hospital neonatal intensive care unit was performed. Clinical data from all infants with gestational age less than 37 weeks, born between April 2018 and March 2021, were gathered. Infants with chromosomal abnormalities, neonatal death, hypoxic-ischemic encephalopathy, congenital malformation syndrome, received blood transfusion, maternal-fetal transfusion syndrome, sepsis, suspected fetal infection, or no available data were excluded. SGA infants were defined as those whose weight at birth was less than the 10th percentile, and non-SGA infants were defined as those whose weight at birth was greater than the 10th percentile. Thrombocytopenia was defined as a platelet count less than 150 × 103/µL. A total of 202 infants were enrolled during the study period (Fig. 1); 30 were SGA infants and 172 were non-SGA infants (Fig. 1), and 17 of 30 SGA infants and 40 of 172 non-SGA infants showed thrombocytopenia (Fig. 1). This study was approved by the ethics committee of Aichi Medical University Hospital. Written, informed consent was obtained from a parent. Fig. 1Study profile. SGA small-for-gestational-age.
Laboratory data, including the white blood cell (WBC) count, hemoglobin (Hb), platelet count, and IPF at the time of the lowest platelet count within 72 h after birth, were collected. The platelet counts at 7 and 14 days of age were also collected. Serum TPO was measured using ELISA (Human Thrombopoietin ELISA Kit, R&D Systems Inc., Minneapolis, MN) in infants with thrombocytopenia at the time of the lowest platelet count within 72 h after birth.
All statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R software (The R Foundation for Statistical Computing, Vienna, Austria). For parametric variables, Student’s t-test was used, whereas Fisher’s exact test was used for categorical variables. Coefficients of correlation were tested using Pearson’s two-tailed test. Data are reported as means (interquartile range). Statistical significance was set at $p \leq 0.05.$
## Results
Factors contributing to SGA were hypertensive disorder of pregnancy (HDP) ($\frac{11}{30}$), umbilical cord factors including limbic attachment ($\frac{9}{30}$), twin or triplet gestation ($\frac{7}{30}$), and unknown ($\frac{3}{30}$). Table 1 shows the infants’ characteristics and a comparison between the SGA and non-SGA infants at birth. There were no significant differences in gestational age, head circumference, the ratios of males and cesarean sections, Apgar score, cord blood pH, HCO3-, and base excess between SGA and non-SGA infants. Bodyweight and length were significantly lower in SGA infants than in non-SGA infants, although the head circumferences were not different and greater than the 10th percentile in all SGA infants. In both infant groups, there were no serious complications with thrombocytopenia, such as cerebral and pulmonary hemorrhages. Table 1Characteristics of all infants. Total ($$n = 202$$)SGA ($$n = 30$$)Non-SGA ($$n = 172$$)p valueGestational age (weeks)a32.4(31.0–35.0)32.6(31.2–35.0)32.4(30.6–35.0)nsBodyweight (g)a1582–0.52 SD(1299–2009)(–1.63 to 0.30)1389–2.10 SD(1132–1511)(–2.51 to 1.41)18280.12 SD(1426–2194)(–0.47 to 0.74)<0.05Length (cm)a38.5–0.73 SD(36.0–41.3)(–1.49 to 0.26)37.5–1.84 SD(35.4–39.0)(–2.26 to –1.24)40.50.24 SD(37.9–42.5)(–0.54 to 0.65)<0.05Head circumference (cm)a28.5–0.57 SD(26.5–30.0)(–0.83 to 0.36)28.5–0.77 SD(27.0–29.5)(–1.04 to –0.57)29.40.41 SD(26.6–30.5)(–0.48 to 0.80)nsMaleb106 ($52.4\%$)13 ($43.3\%$)93 ($54.1\%$)nsCesarean sectionb161 ($79.7\%$)26 ($86.7\%$)135 ($78.5\%$)nsApgar score (1‘)a7(5–8)7(5–9)7(5–8)nsApgar score (5’)a8(7–10)9(7–10)8(7–9)nspHa7.324(7.285–7.360)7.316(7.268–7.343)7.341(7.307–7.366)nsHCO3- (mmol/L)a22.8(21.0–25.1)22.4(21.3–24.3)23.1(20.8–25.2)nsBE (mmol/L)a–2.9(–5.0 to –1.1)–3.6(–5.4 to –2.2)–2.2(–4.7 to –0.7)nsns not significant.aValues are shown as median (interquartile range).bValues are shown as numbers (%).
Table 2 shows the WBC count, Hb, and platelet count values at the time of the lowest platelet count within 72 h after birth in each group. There was no significant difference in the WBC count at the time of the lowest platelet count within 72 h after birth between the two groups (SGA: 10,100 (7600–13,100)/µL, non-SGA: 10,200 (7100–11,200)/µL, ns). Hb was significantly higher in SGA infants than in non-SGA infants (SGA: 19.9 (17.7–21.0)/g/dL, non-SGA: 16.5 (13.9–18.3) g/dL, $p \leq 0.05$). The platelet count was significantly lower in SGA infants than in non-SGA infants at the time of the lowest platelet count within 72 h after birth (SGA: 150 (92–216) ×103/µL, non-SGA: 233 (181–294) ×103/µL, $p \leq 0.05$), but not after 7 days of age (Table 2 and Fig. 2). There were no significant differences in IPF and TPO at the time of the lowest platelet count within 72 h after birth (IPF, SGA 3.2 (2.2–4.6)%, non-SGA 3.2 (2.5–4.7)%, ns; TPO, SGA 330 (167–690), non-SGA 470 (216–712) pg/mL, ns) (Fig. 3a, b). However, the platelet count and IPF were negatively correlated in non-SGA infants (r2 = 0.222, $p \leq 0.05$), but not in SGA infants (r2 = 0.067, ns) (Fig. 3c). In addition, the platelet count and TPO were also negatively correlated in non-SGA infants (r2 = 0.104, $p \leq 0.05$), but not in SGA infants (r2 = 0.001, ns) (Fig. 3d). IPF and TPO have been reported to be useful markers for identifying the cause of thrombocytopenia.10Table 2WBC, Hb, and platelet count at the time of the lowest platelet count within 72 h after birth in all infants. Total ($$n = 202$$)SGA ($$n = 30$$)Non-SGA ($$n = 172$$)p valueWBC (/µL)10,100(7450–12,500)10,200(7100–11,200)10,100(7600–13,100)nsHb (g/dL)17.9(15.5–19.5)19.9(17.7–21.0)16.5(13.9–18.3)<0.05Platelet (×103/µL)208(135–281)150(92–216)233(181–294)<0.05Values are shown as median (interquartile range).ns not significant. Fig. 2Time course of platelet counts at the time of the lowest platelet count within 72 h after birth in all infants. The figure shows box plots with whiskers. The horizontal line indicates the median. The box indicates the interquartile range. The whiskers indicate the full range. SGA small-for-gestational-age, ns not significant. Fig. 3Immature platelet fraction and thrombopoietin at the time of the lowest platelet count within 72 h after birth in all infants and the coefficients of correlation with the platelet count. The figure shows box plots with whiskers. The horizontal line indicates the median. The box indicates the interquartile range. The whiskers indicate the full range. SGA small-for-gestational-age, ns not significant. a Immature platelet fraction at the time of the lowest platelet count within 72 h after birth. b Thrombopoietin at the time of the lowest platelet count within 72 h after birth. c Coefficients of correlation between the platelet count and immature platelet fraction. d Coefficients of correlation between the platelet count and thrombopoietin.
Table 3 shows the infants’ characteristics and a comparison between the SGA and non-SGA infants with thrombocytopenia. The etiology of thrombocytopenia in non-SGA infants was unknown, except in three infants with maternal idiopathic thrombocytopenia. There were also no significant differences, except in bodyweight and length, between SGA and non-SGA infants. Table 3Characteristics of infants with thrombocytopenia. Total ($$n = 57$$)SGA ($$n = 17$$)Non-SGA ($$n = 40$$)p valueGestational age (weeks)a32.0(29.4–33.7)32.4(31.1–33.7)30.0(28.5–35.4)nsBodyweight (g)a1429–1.64 SD(1260–1740)(–2.28 to –0.32)1384–2.24 SD(1176–1440)(–2.28 to –0.67)1723–1.08 SD(1391–1898)(–2.15 to –0.15)<0.05Length (cm)a37.5–1.44 SD(36.0–38.8)(–2.20 to –0.73)37.0–2.20 SD(36.0–37.5)(–2.29 to –1.93)38.5–0.73 SD(36.5–40.1)(–1.35 to 0.50)<0.05Head circumference (cm)a27.8–0.72 SD(26.0–29.5)(–0.91 to –0.08)27.8–0.91 SD(26.0–29.1)(–1.24 to –0.85)28.10.07 SD(26.0–30.6)(–0.15 to 0.33)nsMaleb23 ($40.4\%$)8 ($47.1\%$)17 ($42.5\%$)nsCesarean sectionb29 ($50.9\%$)9 ($52.9\%$)20 ($50.0\%$)nsApgar score (1‘)a7(4–8)7(4–7.5)7(4.75–9)nsApgar score (5’)a8(7–10)9(7–10)7.5(6.5–8.5)nspHa7.333(7.247–7.384)7.319(7.182–7.364)7.363(7.278–7.395)nsHCO3- (mmol/L)a23.4(21.5–25.4)21.8(18.6–24.8)24.5(22.4–28.2)nsBE (mmol/L)a–2.8(–6.7 to –1.1)–4.5(–5.1 to –0.9)–1.8(–2.7 to 0.80)nsns not significant.aValues are shown as median (interquartile range).bValues are shown as numbers (%).
Table 4 shows the WBC count, Hb, and platelet count at the time of the lowest platelet count within 72 h after birth in the two groups with thrombocytopenia. There was no significant difference in the WBC count at the time of the lowest platelet count within 72 h after birth between the two groups with thrombocytopenia (SGA: 9600 (6500–11,400)/µL, non-SGA: 11,200 (7700–16,000)/µL, ns) (Table 4). Hb was significantly higher in SGA infants than in non-SGA infants with thrombocytopenia (SGA: 19.3 (16.7–20.5)/g/dL, non-SGA: 16.7 (13.9–20.8) g/dL: $p \leq 0.05$). The platelet count was significantly lower in SGA infants than in non-SGA infants (SGA: 81 (51–119) ×103/µL, non-SGA: 127 (100–137) ×103/µL, $p \leq 0.05$) (Table 4). IPF was significantly lower in SGA infants than in non-SGA infants with thrombocytopenia (SGA: 3.7 (3.0–4.8)%, non-SGA: 5.3 (3.7–8.9)%, $p \leq 0.05$) (Fig. 4a). Similar to IPF, TPO was significantly lower in SGA infants than in non-SGA infants with thrombocytopenia (SGA: 219 (180–322) pg/ml, non-SGA: 554 (328–897) pg/mL, $p \leq 0.05$) (Fig. 4b). In addition, the correlations of the platelet count with IPF and TPO were investigated in the two groups with thrombocytopenia. There were negative correlations in non-SGA infants with thrombocytopenia, but not in SGA infants with thrombocytopenia, between the platelet count and IPF (non-SGA: r2 = 0.426, $p \leq 0.05$, SGA: r2 = 0.002, ns), and between the platelet count and TPO (non-SGA: r2 = 0.240, $p \leq 0.05$, SGA: r2 = 0.024, ns) (Fig. 4c, d). Together, these results showed that IPF increased with thrombocytopenia to promote platelet production in non-SGA infants due to increasing TPO, but not in SGA infants. Table 4WBC, Hb, and platelet count at the time of the lowest platelet count within 72 h after birth in infants with thrombocytopenia. Total ($$n = 57$$)SGA ($$n = 17$$)Non-SGA ($$n = 40$$)p valueWBC (/µL)10,500(6600–13,200)9600(6500–1400)11,200(7700–16,000)nsHb (g/dL)17.7(15.0–20.8)19.3(16.7–20.5)16.7(13.9–20.8)<0.05Platelet (×103/µL)121(83–133)81(51–119)127(100–137)<0.05Values are shown as median (interquartile range).ns not significant. Fig. 4Immature platelet fraction and thrombopoietin at the time of the lowest platelet count within 72 h after birth in infants with thrombocytopenia and the coefficients of correlation with the platelet count. The figure shows box plots with whiskers. The horizontal line indicates the median. The box indicates the interquartile range. The whiskers indicate the full range. SGA small-for-gestational-age, ns not significant. a Immature platelet fraction at the time of the lowest platelet count within 72 h after birth. b Thrombopoietin at the time of the lowest platelet count within 72 h after birth. c Coefficients of correlation between the platelet count and the immature platelet fraction. d Coefficients of correlation between the platelet count and thrombopoietin.
## Discussion
The present study demonstrated that preterm SGA infants had significantly lower platelet counts, and that the platelet count and IPF were negatively correlated in non-SGA infants, but not in SGA infants. In addition, the platelet count and TPO were negatively correlated in non-SGA infants, but not in SGA infants. It was also demonstrated that SGA infants with thrombocytopenia had lower IPF and serum TPO than non-SGA infants. These results suggest that thrombocytopenia in SGA infants is due to insufficient TPO production. These findings are important for understanding the etiology of thrombocytopenia in SGA infants.
SGA is caused by maternal factors such as HDP, malnutrition, psychosocial stress, and smoking, as well as neonatal factors such as multiple births and congenital diseases.11–14 The risk of developing HDP increases with maternal age and thinness in Japan.15 *In this* study, the head circumferences were greater than the 10th percentile in all SGA infants. These results demonstrate that all SGA infants showed asymmetrical growth due to maternal/placental factors.9,12 These SGA infants are at increased risk for complications, such as prematurity, asphyxia, hypothermia, hypoglycemia, hypocalcemia, polycythemia, and thrombocytopenia.1,2 Chronic intrauterine hypoxia caused by maternal factors for SGA infants leads to high fetal erythropoietin and polycythemia.1,2 In the present study, similar to the previous studies, Hb was significantly higher in SGA infants than in non-SGA infants.3,4 The etiology of thrombocytopenia in SGA infants remains unknown. Thus, it is crucial to clarify the etiology and management of thrombocytopenia in SGA infants.
TPO is a major physiological regulator protein that promotes the differentiation and proliferation of megakaryocytes to platelets.16 TPO is produced constantly in the liver and binds to the TPO receptor on the surface of megakaryocytic cells, promoting production of platelets via signaling pathways such as JAK-STAT and RAS-MAPK.17,18 The main causes of thrombocytopenia are increasing destruction/consumption of circulating platelets and decreased platelet production in the bone marrow.10 The cause of thrombocytopenia in SGA infants has been postulated to be a decrease in platelet production.5–9,19 However, there are few studies on infants, and the etiology of thrombocytopenia in human SGA infants remains unclear. Wasiuk et al. evaluated thrombopoiesis in SGA infants and postulated that intrauterine hypoxia is responsible for the increase of erythropoietin and impairment of thrombopoiesis in SGA infants.5 However, they did not compare TPO levels between SGA and non-SGA infants. Amariyo et al. reported that TPO and inflammatory cytokine levels in cord blood samples from SGA infants were significantly higher than in appropriate-for-gestational-age infants.20 They suggested that this increase was caused by a state of inflammation in the IUGR fetus.20 In the present study, the timing of blood sampling was later than in their study, and this discrepancy between study findings might be due to this difference in timing.
We previously demonstrated that a decrease in TPO production due to hepatic dysmaturation resulted in thrombocytopenia in SGA model rats.9 In the present study, SGA infants with thrombocytopenia had lower IPF and serum TPO levels than non-SGA infants with thrombocytopenia. These results suggest that IPF and TPO levels do not increase in response to thrombocytopenia in human SGA infants, similar to SGA model rats. This is the first report to show human infant data in agreement with those of our animal experiments.
The brain is symmetrically smaller in SGA infants of fetal origin, whereas the brain is protected by the brain-sparing effect in SGA infants of non-fetal origin.21,22 In the present study, the length and weight of SGA infants were lower than those of non-SGA infants, but the head circumference was not significantly different from that of non-SGA infants. These results suggest that other organs including the liver, rather than the brain, are dysmature in SGA infants. The decrease in TPO production may also be reflected by this liver dysmaturity, because TPO is mainly produced in the liver, which is susceptible to hypoxia.
There are some limitations to this study. It was a prospective study in a single hospital with a limited number of patients. Biases in patients’ background characteristics and treatment strategies may also have been present. In the present study, most SGA infants were born at a gestational age of more than 30 weeks. The lack of serious complications with thrombocytopenia might have been due to them being fairly mature infants. In addition, whether early administration of a TPO receptor agonist was effective for thrombocytopenia and improved the prognosis in SGA infants was not investigated.
In conclusion, thrombocytopenia in SGA infants could be due to insufficient platelet production caused by a decrease in TPO levels. These results are consistent with previous studies5–8 and are important for understanding the etiology of thrombocytopenia in SGA infants.
## References
1. Kramer MS. **The epidemiology of adverse pregnancy outcome: an overview**. *J. Nutr.* (2003) **133** 1292-1296. DOI: 10.1093/jn/133.5.1592S
2. Sankaran S, Kyle PM. **Aetiology and pathogenesis of IUGR**. *Best. Pract. Res. Clin. Obstet. Gynaecol.* (2009) **23** 765-777. DOI: 10.1016/j.bpobgyn.2009.05.003
3. Christensen RD. **Thrombocytopenia in small for gestational age infants**. *Pediatr* (2015) **136** e361-e370. DOI: 10.1542/peds.2014-4182
4. Fustolo-Gunnink SF. **Early-onset thrombocytopenia in small-for-gestational-age neonates: a retrospective cohort study**. *Plos One* (2016) **11** e0154853. DOI: 10.1371/journal.pone.0154853
5. Wasiluk A. **Thrombopoiesis in small for gestational age newborns**. *Platelets* (2009) **20** 520-524. DOI: 10.3109/09537100903207505
6. Sola MC, Calhoun DA, Hutson AD, Christensen RD. **Plasma thrombopoietin concentrations in thrombocytopenic and non‐thrombocytopenic patients in a neonatal intensive care unit**. *Br. J. Haematol.* (1999) **104** 90-92. DOI: 10.1046/j.1365-2141.1999.01154.x
7. Murray NA, Roberts IA. **Circulating megakaryocytes and their progenitors in early thrombocytopenia in preterm neonates**. *Pediatr. Res.* (1996) **40** 112-119. DOI: 10.1203/00006450-199607000-00020
8. Watts TL, Murray NA, Roberts IA. **Thrombopoietin has a primary role in the regulation of platelet production in preterm babies**. *Pediatr. Res.* (1999) **46** 28-32. DOI: 10.1203/00006450-199907000-00005
9. Takeshita S. **Insufficient thrombopoietin due to hepatic dysmature results in thrombocytopenia in small-for-gestational-age rats**. *Br. J. Haematol.* (2021) **192** e105-e108. DOI: 10.1111/bjh.17294
10. Jeon K. **Immature platelet fraction: a useful marker for identifying the cause of thrombocytopenia and predicting platelet recovery**. *Medicine* (2020) **99** e19096. DOI: 10.1097/MD.0000000000019096
11. Cheng J, Li J, Tang X. **Analysis of perinatal risk factors for small-for-gestational-age and appropriate-for-gestational-age late-term infants**. *Exp. Ther. Med.* (2020) **19** 1719-1724. PMID: 32104225
12. Berger H. **Impact of diabetes, obesity and hypertension on preterm birth: population-based study**. *PloS One* (2020) **15** e0228743. DOI: 10.1371/journal.pone.0228743
13. Hobel C, Culhane J. **Role of psychosocial and nutritional stress on poor pregnancy outcome**. *J. Nutr.* (2003) **133** 1709S-1717S. DOI: 10.1093/jn/133.5.1709S
14. Vrijkotte TG, van der Wal MF, van Eijsden M, Bonsel GJ. **First-trimester working conditions and birthweight: a prospective cohort study**. *Am. J. Public Health* (2009) **99** 1409-1416. DOI: 10.2105/AJPH.2008.138412
15. Takemoto Y, Ota E, Yoneoka D, Mori R, Takeda S. **Japanese secular trends in birthweight and the prevalence of low birthweight infants during the last three decades: a population-based study**. *Sci. Rep.* (2016) **6** 1-6. DOI: 10.1038/srep31396
16. Kato T. **Native thrombopoietin: structure and function**. *Stem Cells* (1998) **16** 11-19. DOI: 10.1002/stem.5530160704
17. Royer Y, Staerk J, Costuleanu M, Courtoy PJ, Constantinescu N. **Janus kinases affect thrombopoietin receptor cell surface localization and stability**. *J. Biol. Chemi* (2005) **280** 27251-27261. DOI: 10.1074/jbc.M501376200
18. Kuter DJ. **The biology of thrombopoietin and thrombopoietin receptor agonists**. *Int. J. Hematol.* (2013) **98** 10-23. DOI: 10.1007/s12185-013-1382-0
19. Cremer M, Weimann A, Hammer H, Bührer C, Dame C. **Immature platelet values indicate impaired megakaryopoietic activity in neonatal early-onset thrombocytopenia**. *Thromb. Haemost.* (2010) **103** 1016-1021. DOI: 10.1160/TH09-03-0148
20. Amarilyo G. **Increased cord serum inflammatory markers in small-for gestational-age neonates**. *J. Perinatol.* (2011) **31** 30-32. DOI: 10.1038/jp.2010.53
21. Roza SJ. **What is spared by fetal brain-sparing? Fetal circulatory redistribution and behavioral problems in the general population**. *Am. J. Epidemiol.* (2008) **168** 1145-1152. DOI: 10.1093/aje/kwn233
22. Simanaviciute D, Gudmundsson S. **Fetal middle cerebral to uterine artery pulsatility index ratios in normal and pre-eclamptic pregnancies**. *Ultrasound Obstet. Gynecol.* (2006) **28** 794-801. DOI: 10.1002/uog.3805
|
---
title: Validation of disease-specific biomarkers for the early detection of bronchopulmonary
dysplasia
authors:
- Alida S. D. Kindt
- Kai M. Förster
- Suzan C. M. Cochius-den Otter
- Andreas W. Flemmer
- Stefanie M. Hauck
- Andrew Flatley
- Juliette Kamphuis
- Stefan Karrasch
- Jürgen Behr
- Axel Franz
- Christoph Härtel
- Jan Krumsiek
- Dick Tibboel
- Anne Hilgendorff
journal: Pediatric Research
year: 2022
pmcid: PMC9988689
doi: 10.1038/s41390-022-02093-w
license: CC BY 4.0
---
# Validation of disease-specific biomarkers for the early detection of bronchopulmonary dysplasia
## Abstract
### Objective
To demonstrate and validate the improvement of current risk stratification for bronchopulmonary dysplasia (BPD) early after birth by plasma protein markers (sialic acid-binding Ig-like lectin 14 (SIGLEC-14), basal cell adhesion molecule (BCAM), angiopoietin-like 3 protein (ANGPTL-3)) in extremely premature infants.
### Methods and results
Proteome screening in first-week-of-life plasma samples of $$n = 52$$ preterm infants <32 weeks gestational age (GA) on two proteomic platforms (SomaLogic®, Olink-Proteomics®) confirmed three biomarkers with significant predictive power: BCAM, SIGLEC-14, and ANGPTL-3. We demonstrate high sensitivity (0.92) and specificity (0.86) under consideration of GA, show the proteins’ critical contribution to the predictive power of known clinical risk factors, e.g., birth weight and GA, and predicted the duration of mechanical ventilation, oxygen supplementation, as well as neonatal intensive care stay. We confirmed significant predictive power for BPD cases when switching to a clinically applicable method (enzyme-linked immunosorbent assay) in an independent sample set ($$n = 25$$, $p \leq 0.001$) and demonstrated disease specificity in different cohorts of neonatal and adult lung disease.
### Conclusion
While successfully addressing typical challenges of clinical biomarker studies, we demonstrated the potential of BCAM, SIGLEC-14, and ANGPTL-3 to inform future clinical decision making in the preterm infant at risk for BPD.
### Trial registration
Deutsches Register Klinische Studien (DRKS) No. 00004600; https://www.drks.de.
### Impact
The urgent need for biomarkers that enable early decision making and personalized monitoring strategies in preterm infants with BPD is challenged by targeted marker analyses, cohort size, and disease heterogeneity. We demonstrate the potential of the plasma proteins BCAM, SIGLEC-14, and ANGPTL-3 to identify infants with BPD early after birth while improving the predictive power of clinical variables, confirming the robustness toward proteome assays and proving disease specificity. Our comprehensive analysis enables a phase-III clinical trial that allows full implementation of the biomarkers into clinical routine to enable early risk stratification in preterms with BPD.
## Introduction
Risk stratification for preterm infants with chronic lung disease (CLD), i.e., bronchopulmonary dysplasia (BPD), early after birth is urgently needed to inform postnatal clinical decision making. As of now, physicians have to rely on the diagnosis at 36 weeks gestational age (GA), solely based on clinical criteria.1 Previous approaches aiming at the identification of such biomarkers have been largely limited by the predominant use of targeted marker analysis, non-sensitive detection techniques, and standard data analysis approaches as opposed to statistical modeling including clinical variables.2,3 In order to overcome these limitations, we identified a combination of three plasma markers (basal cell adhesion molecule (BCAM), sialic acid-binding Ig-like lectin 14 (SIGLEC-14), angiopoietin-like 3 protein (ANGPTL-3)) using unbiased proteome screening (SOMAscan® assay, SomaLogic®, Boulder, CO),4,5 whose expression levels in the first week of life and after 28 days were significantly associated with later BPD development,6 complemented by their verification in paraformaldehyde tissue sections from autopsy lungs of infants with BPD.
In order to now evaluate the potential of these proteins to serve as biomarkers in clinical routine as early as in the first week of life and thereby improve current risk stratification for BPD, we designed a study approach validating the biomarkers’ expression in a larger patient cohort as well as an independent sample set while rigorously assessing different, clinically relevant performance criteria. These included the evaluation of the biomarkers’ added value for disease detection in comparison to the sole use of clinical risk factors, the transfer of the measurement technique to a clinically applicable assay as well as the assessment of the biomarkers’ disease specificity by the use of neonatal and adult cohorts suffering from CLD of different origin. The comprehensive analysis was designed to enable a phase II clinical trial for implementation of the identified biomarkers into routine care for preterm infants.
## Patient characteristics
Sample sets analyzed comprise a training (preterm training cohort) and a validation (preterm validation cohort) cohort of preterm infants, as well as additional samples from a small group of preterm infants recruited at a different study site. Disease specificity was addressed in a cohort of neonatal CLD (CLD-CDH cohort) and a cohort of adult CLD patients (adult CLD cohort).
## Preterm training cohort
We analyzed a total of 55 plasma samples obtained in the first week of life (median day of life 4, range: 0–7) from preterm infants born <32 weeks GA (total number of patients: $$n = 55$$, median GA 27.2 weeks, range: 23.2–30.6; $45.5\%$ males). All infants were born at the Perinatal Center in Munich and prospectively enrolled into the AIRR study (Attention to Infants @ Respiratory Risks) after written informed parental consent. The approval was assigned by the Ethics Committee of the Medical Faculty of Ludwig-Maximilians University in Munich (Ethical vote #195-07). The study was registered at the German Registry for Clinical Trials (No. 00004600; https://www.drks.de). Preterm infants were prospectively included following the in- and exclusion criteria published previously:6 inclusion of preterm infants born <32 weeks GA with the exception of severe congenital malformations (e.g., hypoplastic left-heart syndrome, severe hypoplasia of the lungs or congenital diaphragmatic hernia (CHD)), chromosomal abnormalities (e.g., trisomy 13 or 18), inborn errors of metabolism, and decision for palliative therapy directly after birth). Clinical variables were comprehensively monitored from birth to discharge (Table 1) using the following consented definitions: intrauterine growth restriction: birth weight below the 10th percentile; diagnosis and severity of respiratory distress syndrome (RDS): assessment of anterior-posterior chest radiographs according to Couchard et al.;7 chorioamnionitis: inflammatory alterations of the chorionic plate (histologic examination) or signs of maternal and fetal signs of infection;8 presence of early postnatal systemic infections (early-onset infection (eoi)): one or more clinical and laboratory signs of infection according to Sherman et al.9Table 1Patient characteristics cohort of preterm infants (Munich).Training cohortValidation cohortn5225GA (weeks)27.2 (23.2–30.6)27.0 (23.6–31.0)Birth weight (g)798 (510–1590)820 (480–1620)IUGR6 (11.5)5 [20]Gender (female/male)$\frac{29}{2312}$/13ANCS46 ($88.5\%$)21 ($84\%$)Chorioamnionitis27 ($51.9\%$)10 ($40\%$)Early-onset infection (EOI)14 ($26.9\%$)6 ($24\%$)RDS ≥311 ($21.2\%$)8 ($32\%$)Mechanical ventilation (days)45 (0–109)48 (6–129)Oxygen supplementation (days)32 (0–186)44 (0–129)Postnatal steroids22 ($42.3\%$)12 ($48\%$)ROP15 ($28.9\%$)5 ($20\%$)IVH6 (11.5)1 ($4\%$)NICU stay (days)63 (9–113)65 (9–150)*Bronchopulmonary dysplasia* (BPD) None24 ($46.2\%$)10 ($40\%$) Mild15 ($28.9\%$)5 ($20\%$) Moderate5 ($9.6\%$)2 ($8\%$) Severe8 ($15.4\%$)8 ($32\%$)Pulmonary hypertension (PH) (≥$\frac{2}{3}$ syst. pressure)0 ($0\%$)0 ($0\%$)Samples of the preterm training cohort were analyzed on two analysis platforms (SOMAscan® assay ($$n = 33$$); Proximity Extension Assay ($$n = 19$$)) and validated in an independent sample set (preterm validation cohort) by enzyme-linked immunosorbent assay (ELISA, $$n = 25$$). Data are given as median and range or number and percent of total in group’s respective range. NICU stay not available in $$n = 1$$ infant in the preterm training cohort. GA gestational age, ANCS antenatal corticosteroids, RDS respiratory distress syndrome, ROP retinopathy of prematurity, IVH intraventricular hemorrhage, ICU intensive care unit, BPD bronchopulmonary dysplasia.
BPD was defined according to the NICHD/NHLBI/ORD workshop1 based on the need for oxygen supplementation (>FiO2 0.21) for at least 28 days, followed by a final assessment at 36 weeks postmenstrual age (PMA) or at discharge, whichever came first in preterm infants born <32 weeks GA. Disease grading accordingly assigned infants to having mild BPD (requirement of supplemental oxygen for 28 days, no need for oxygen supplementation at 36 weeks PMA) or moderate BPD (oxygen supplementation <FiO2 0.30 at 36 weeks PMA), and severe BPD (oxygen supplementation >FiO2 0.30 at 36 weeks PMA and/or positive pressure ventilation/continuous positive pressure) with each treatment referring to its continuous application and oxygen supplementation >12 h equaling one day of treatment.1 The infants’ oxygen saturation was assessed by standardized pulse oximetry. No infant was discharged from hospital before 36 weeks’ gestation.
## Preterm validation cohort
To validate the results obtained, we analyzed expression levels for all three proteins in an independent sample set of the AIRR study collected in the first week of life (median day of life 0, range 0–5 days) by the use of commercially available enzyme-linked immunosorbent assay (ELISA). Preterm infants were included following the same in- and exclusion criteria as outlined above (total number of patients: $$n = 25$$, median GA 27.0 weeks, range 23.6–31.0; $52\%$ males). In this cohort, $$n = 10$$ infants did not develop BPD (no BPD ($40\%$)), $$n = 5$$ infants developed mild BPD ($20\%$), and a total of 10 infants were diagnosed with either moderate BPD ($$n = 2$$ ($8\%$)) or severe BPD ($$n = 8$$ ($32\%$)) with no infant being discharged before 36 weeks’ gestation (Table 1).
In an additional step, a small group of preterm infants was recruited after written informed parental consent at a different study site following the same in- and exclusion criteria in order to mimic sampling conditions of a clinical trial, i.e., ongoing recruitment of small sample sets with random distribution of clinical characteristics; total number of patients: $$n = 8$$, median GA: 25.6 weeks, range: 24.1–29.0; median birth weight: 852 g, range: 520–1470 g; eoi $$n = 2$$ ($25\%$), $75\%$ males, $$n = 4$$ no BPD, $$n = 4$$ moderate/severe BPD). The approval was assigned by the Ethics Committee of the University of Schleswig-Holstein (Ethical vote #AZ 15-304).
## Neonatal CLD-CDH cohort
To assess disease specificity of the biomarkers investigated, we repeated their measurement in a neonatal cohort of infants suffering from CDH with and without CLD. CLD was defined according to the need for mechanical ventilation and/or oxygen supplementation beyond day 28 of life, thereby following the BPD definition from the NICHD/NHLBI/ORD workshop for infants >32 weeks PMA.1 The infants were part of the VICI-trial10 and prospectively included after informed parental consent at the ErasmusMC Sophia Children’s Hospital in Rotterdam. The approval was assigned by the Ethics Committee of the University of Rotterdam, the Netherlands (Ethical vote #MEC-2006-260). The cohort included 21 neonates in total with a median GA 38.0 weeks (range 33.6–41.3), $33.3\%$ males. Six infants did not develop CLD (no CLD-CDH) and nine infants developed CLD (CLD-CDH (survivors)). Six infants deceased (CLD-CDH (deceased)) (Table 2).Table 2Patient characteristics Neonatal CLD-CDH cohort (Rotterdam).n21GA (weeks)38.0 (33.6–41.3)Gender (female/male)($\frac{14}{7}$)Death6 ($28.6\%$)Early-onset infection (EOI)No17 ($81\%$)Yes4 ($19\%$)Mechanical ventilationNo8 ($38.1\%$)Yes10 ($47.6\%$)NA3 ($14.3\%$)O2 (day 28)No6 ($28.6\%$)Yes10 ($47.6\%$)NA2 ($9.5\%$)Severity CLDNo CLD6 ($28.6\%$)Mild CLD8 ($38.1\%$)Severe CLD1 ($4.8\%$)PH (first echocardiogram)No6 ($28.6\%$)PH (from $\frac{2}{3}$ of syst. pressure on)14 ($66.7\%$)NA1 ($4.8\%$)ECMONo13 ($61.9\%$)Yes8 ($38.1\%$)iNONo11 ($52.4\%$)Yes10 ($47.6\%$)Data are given as median and range or number and percent of total in group’s respective range. CLD chronic lung disease, ECMO extra corporal membrane oxygenation, EOI early-onset infection, GA gestational age, iNO inhalative nitric oxide, NA not available, PH pulmonary hypertension.
## Adult CLD cohort
Disease specificity of the biomarkers was furthermore assessed in a cohort of adult CLD patients after informed consent (CPC-M bioArchive, Munich, Ethics Committee of the Medical Faculty of Ludwig-Maximilians University in Munich (Ethical vote #19-629)) comprising samples from patients with idiopathic pulmonary fibrosis (IPF, total number of patients $$n = 21$$, median age 56 years (range 30–73), $76.7\%$ males), chronic obstructive pulmonary disease (COPD, total number of patients: $$n = 26$$, median age 50 years (range 14–74), $58\%$ males), and subjects free of lung disease according to clinical history from the KORA cohort11 (total number of patients: $$n = 25$$, median age 60 years (range 53–67), $52\%$ males). KORA (Cooperative Health Research in the Region Augsburg) is a regional research platform for population-based surveys and subsequent follow-up studies with a focus on diabetes, cardiovascular, and lung diseases, including the impact of environmental factors.
## Sampling processing
Serial whole blood samples (200 µl minimum each) obtained during routine laboratory blood drawings were collected using ethylenediaminetetraacetic acid neonatal collection tubes. After pseudonymization samples were processed for proteomic screening by centrifugation (1000 g, 5 min) before supernatants were aliquoted and stored at −80 °C. Time from sample collection to sample processing was standardized according to the study protocol (DRKS No. 00004600).
## Proteomic analysis
Samples from the preterm training cohort were analyzed on two analysis platforms in three subsequent batches (SOMAscanⓇ assays: 1st batch $$n = 16$$, 2nd batch $$n = 20$$; Proximity Extension Assay (PEA): $$n = 19$$; collected at day 4 of life (median), range: 0–7) followed by validation in a sample set of independent patients (preterm validation cohort, $$n = 25$$) that used a clinically applicable analysis technique, i.e., ELISA. Samples from an additional group of preterm infants ($$n = 8$$) were analyzed by PEA. Analysis for disease specificity comprised samples from cohorts with neonatal and adult CLD of different origin, i.e., $$n = 21$$ neonates in the CLD-CDH cohort (PEA (Olink-Proteomics®), $$n = 21$$) and $$n = 72$$ in the adult CLD cohort (SOMAscan® assay (SomaLogic®)).
The SOMAscanⓇ assay (SomaLogic®, Boulder, CO) uses 1129 individual high-affinity molecules (SOMAmer®—slow off-rate modified DNA aptamer—reagents) quantified on a custom Agilent hybridization array.5,12 The PEA (Olink-Proteomics®, Uppsala, Sweden) employs a matched pair of antibodies linked to unique oligonucleotides detected in multiplexed fashion in a high throughput fluidic chip system measuring 630 unique proteins.13 Both techniques are designed for the accurate quantification of human plasma proteins present in concentrations below picogram per milliliter using even low-amount samples. For PEA measurements, SIGLEC$\frac{5}{14}$ was detected in the identical aliquot by ELISA (R&D Systems, MN).
We validated the results obtained by the use of a clinically applicable method (ELISA) in an independent sample set obtained from $$n = 25$$ infants (preterm validation cohort, Table 1). The commercially available ELISAs targeted all three proteins and were performed according to the manufacturer’s instructions. Samples were measured in duplicates and diluted 1:100 for the SIGLEC$\frac{5}{14}$ ELISA (R&D Systems #DY1072), 1:10 for the BCAM ELISA (Thermo Scientific #EHBCAM), and 1:200 for the ANGPTL-3 ELISA (Ray Biotech #ELH-ANGPTL3). Readouts were obtained in a TECAN Spark ELISA reader (Tecan Trading AG, Switzerland).
## Statistical analysis
Three outliers were detected by principal component analysis (prcomp function, R framework, log2-transformed, and pareto scaled data) and removed from further analysis and summary statistics. Preterm training cohort: protein expression obtained from two different sample sets analyzed by SOMAscan® (SomaLogic®) (1st batch $$n = 16$$, 2nd batch $$n = 17$$) and one sample set analyzed by PEA (Olink-Proteomics®) ($$n = 19$$) were batch corrected using the combat function from the sva package (version 3.36) in R (version 4.0; Supplementary Fig. 1). Next, validation was performed in an independent cohort of 25 preterm infants, in which protein concentrations were determined by ELISA. Further, we independently analyzed results obtained by PEA in samples from eight preterm infants recruited at a different study site. Analysis for disease specificity comprised samples from cohorts with neonatal and adult CLD of different origin, i.e., $$n = 21$$ neonates in the CLD-CDH cohort (PEA (Olink-Proteomics®), $$n = 21$$) and $$n = 72$$ in the adult CLD cohort (SOMAscan® assay (SomaLogic®)). Area under the curve (AUCs) were calculated using a leave-one-out cross-validation with a generalized linear model that compared no BPD infants to (a) only moderate and severe BPD cases or (b) all BPD cases as binary outcome and log2 transformed, pareto scaled protein expression data together with the clinical variables GA, birth weight, eoi, and sex as covariates, which are the most important risk factors associated with BPD development.14 These variables were analyzed with and without protein expression levels as covariates in a total of 32 different combinations including a null model. The model with the highest AUC and lowest Akaike’s Information Criterion (AIC) was the one used to perform the prediction of BPD and the continuous variables of duration of mechanical ventilation [days], oxygen supplementation [days], and neonatal intensive care unit (NICU) duration [days]. All prediction models for the preterm training cohort were calculated using a leave-1-out cross-validation with a linear regression model using the covariates from the best model comprising protein expression levels and GA. The models applied for the CLD-CDH cohort and adult CLD cohort, i.e., IPF and COPD, were corrected for GA (neonates) or age (adult patients). Three tests (Bartlett test, Fligner–Killeen test, and Levene test (from the car package)) were used to check for equal variances before ANOVA testing. All protein concentrations were log2 transformed and pareto scaled prior to statistical analysis. Fig. 1Improved performance of BPD prediction by novel plasma protein biomarkers detected in the first week of life. Protein markers (SIGLEC-14, BCAM, ANGPTL-3) significantly improve performance of BPD prediction models solely based on clinical variables. AIC vs. AUC for 31 of the 32 analyzed models (excluding the null model from the graphs) including protein expression levels (pr), gestational age (GA), sex, birth weight (weight) and early-onset infection (eoi) for a no BPD vs. moderate and severe BPD (<32 weeks GA), for b no BPD vs. moderate and severe BPD (<28 weeks GA), and for c no BPD vs. all BPD Grades (<32 weeks GA). AIC Akaike’s Information Criterion, AUC area under the curve, BPD bronchopulmonary dysplasia, BPD grades: 0 = no BPD, 1 = mild BPD, 2 = moderate BPD, 3 = severe BPD.
## Results
In summary, the final dataset after exclusion of outliers comprised 52 samples in the preterm training cohort ($$n = 24$$ no BPD ($46.2\%$), $$n = 15$$ mild BPD ($28.9\%$), $$n = 5$$ moderate BPD ($9.6\%$) and $$n = 8$$ severe BPD ($15.4\%$)) and 25 samples in the preterm validation cohort (Table 1). In addition, 8 samples from a different study site were analyzed to mimic a clinical study setting. To determine disease specificity, 21 samples from a neonatal CLD-CDH cohort and 72 samples from an adult CLD cohort were analyzed.
## Protein markers significantly improve the performance of BPD prediction models based on clinical variables
We compared 32 models that included protein expression levels for BCAM, SIGLEC-14, and ANGPTL-3 as well as critical risk factors for BPD development, i.e., GA, sex, birth weight, and eoi, which are the most important risk factors associated with BPD development,14 to determine the model with the highest sensitivity and specificity (highest AUC) for the separation of moderate and severe BPD cases from no BPD while explaining the data only by the most important variables (lowest AIC). We show that the combination of the three protein markers together with GA (BPD~BCAM+SIGLEC-14+ANGPTL-3+GA) best predicted BPD in the first week of life with an optimized AUC (0.87) and AIC (23.40), thereby being superior to the other models tested, e.g., “GA alone” (AUC = 0.87, AIC = 30.46), and the null model (AIC = 49.97; Fig. 1a). The model furthermore demonstrated high sensitivity (0.92), specificity (0.86), accuracy (0.89) and positive predictive value (0.80) as well as test accuracy (F1-scores: 0.89). These results were confirmed when restricting the analysis to very immature infants, i.e., <28 weeks GA, again demonstrating superiority with increased AUC (0.86) and decreased AIC (10.0) when compared to other models tested (null model AIC: 24.91; Fig. 1b).
When including all BPD grades, the model comprising the protein markers together with GA demonstrated improved performance (AIC: 43.16) when compared to the model with GA alone (51.0) and the null model (73.77; Fig. 1c). In contrast, AUC levels are comparable between the models when comparing no BPD vs. all BPD cases (protein levels and GA: AUC = 0.83, GA alone: AUC = 0.84; Fig. 1c).
## Protein markers enable BPD prediction for all disease grades with significant accuracy at birth
For the model comprising plasma protein levels and GA, receiver operating characteristic curves for no BPD vs. mild, moderate, and severe BPD and no BPD vs. moderate and severe BPD indicate high sensitivity for BPD prediction in the first week of life (Fig. 2a). The model successfully separates infants according to their risk for later BPD while considering different disease grades: no BPD vs. mild, moderate and severe BPD, ANOVA p value = 1.4 × 10−5 (leave-1-out cross-validation); no BPD vs. mild BPD (t-test p value = 3.4 × 10−3); no BPD vs. moderate and severe BPD (t-test p value = 1.4 × 10−6; Fig. 2b (left panel)). The results obtained were confirmed by ELISA measurements in an independent sample set (preterm validation cohort); ANOVA p value = 2.4 × 10−4; t-test p values no BPD vs. mild BPD $$p \leq 0.024$$; no BPD vs. moderate/severe BPD, $$p \leq 1.4$$ × 10−4; Fig. 2b (right panel)). In an additional analysis, we demonstrated the successful separation according to the risk for BPD development at birth by the proteins in combination with GA in a small sample set recruited at a different study site ($$n = 8$$; Fig. 2b (right panel, gray filled squares)).Fig. 2Disease specific biomarkers predict BPD severity with significant accuracy. Protein markers enable BPD prediction for all disease grades with significant accuracy at birth while demonstrating disease specificity. a AUC values calculated for protein expression levels and GA: black bold line (preterm training cohort): no BPD vs. all BPD cases, AUC = 0.83; black line (preterm training cohort): no BPD vs. moderate and severe BPD AUC = 0.87. b Predicted probability of no BPD vs. all BPD cases resulting from leave-1-out cross-validation using the model with protein expression levels and GA for preterm training cohort (ANOVA $$p \leq 1.4$$ × 10−5, gray filled circles); t-test p values for: no BPD vs. mild BPD $$p \leq 3.4$$ × 10−3; no BPD vs. (mod./severe) BPD, $$p \leq 1.4$$ × 10−6; mild BPD vs. (mod./severe) BPD, $$p \leq 0.13$$; and the ELISA measurements (black triangles) for the preterm validation cohort (ANOVA $$p \leq 2.4$$ × 10−4); t-test p values for: no BPD vs. mild BPD $$p \leq 0.024$$; no BPD vs. (mod./severe) BPD, $$p \leq 1.4$$ × 10−4; mild BPD vs. (mod./severe) BPD, $$p \leq 0.56.$$ Samples from another study site ($$n = 8$$, Luebeck, gray filled squares) show homogeneous distribution within the preterm validation cohort. c *Prediction analysis* (leave-1-out model, preterm training cohort) for mechanical ventilation [days] (black squares, black solid line; $r = 0.81$, $$p \leq 2.9$$ × 10−12), O2 supplementation [days] (gray solid line, gray circles; $r = 0.64$, $$p \leq 7.2$$ × 10−7) and duration of NICU stay [days] (gray dashed line, gray triangles; $r = 0.78$, $$p \leq 7.1$$ × 10−11). d Specificity for BPD prediction in the preterm in comparison to neonatal and adult CLD patients: bold black line (adults): AUC = 0.67; black line (CLD-CDH cohort): CLD-CDH deceased vs. no CLD-CDH infants AUC = 0.78; dashed line (CLD-CDH cohort): CLD-CDH survivors vs. no CLD-CDH infants, AUC = 0.69; gray line (preterm training cohort: no BPD vs. moderate and severe BPD (AUC = 0.87) as reference.
The model furthermore correctly predicts the need for mechanical ventilation (duration in days; observed vs. predicted: $r = 0.81$, p value = 2.9 × 10−12, MAE = 8.60, RMSE = 10.61), and oxygen supplementation (duration in days; observed vs. predicted: $r = 0.64$, p value = 7.2 × 10−7, MAE = 23.20, RMSE = 32.12) as well as the duration of NICU stay [(duration in days; observed vs. predicted: $r = 0.78$, p value = 7.1 × 10−11, MAE = 9.57, RMSE = 12.70; Fig. 2c) in the preterm training cohort.
## Protein markers show specificity for BPD prediction in the preterm when compared to neonatal and adult CLD patients
Application of the model including the protein levels and GA in a cohort of neonatal patients with CLD due to CDH (CLD-CDH cohort) showed no discrimination between neonates with CDH and age-matched infants with CLD-CDH when analyzing survivors (AUC 0.69; Fig. 2d). The protein markers only allowed the separation of maximum disease, i.e., fatal outcome from all survivors including no CLD-CDH and CLD-CDH patients (AUC 0.78; Fig. 2d). In adult patients with CLD (adult CLD cohort), protein levels did not separate cases with COPD or IPF from pulmonary healthy controls with sufficient power (Fig. 2d).
The study showed very good power to separate BPD phenotypes in the preterm training cohort (no BPD vs. all BPD grades: $87\%$; no BPD vs. moderate/severe BPD: $97\%$) and good power to separate cases in the CLD-CDH cohort (no CLD-CDH vs. CLD-CDH (survivors): $60\%$; CLD-CDH (fatal outcome) vs. no CLD-CDH: $72\%$).
## Discussion
BPD is a multifactorial disease and remains the most serious lung condition in neonates born premature due to its significant mortality and morbidity. Despite the clinical significance, the diagnostic process solely relies on clinical criteria assessed at 36 weeks PMA. The relatively late diagnosis only inadequately addresses both the need for early risk stratification as well as the diseases’ multifaceted pathology that includes the rarefication of the gas exchange area, interstitial remodeling, and airway pathology, now clustered at a late stage in a non-discriminatory diagnosis.14 The diagnostic gap with regard to timeliness and accuracy is reflected by the limitations of clinical trials aiming at the implementation of new therapeutic strategies15–19 and underscores the need for new markers enabling today’s clinicians to early and sensitively identify CLD in the preterm infant.
We therefore followed a rigorous approach to evaluate protein markers with significant potential to serve as future biomarkers for BPD prediction as early as in the first week of life, previously identified by us using unbiased proteome screening.6 Based on these findings, we first applied a generalized linear model to identify the best combination of proteins and clinical risk factors for BPD prediction and demonstrated the significant impact of the protein expression levels on improving BPD prediction when compared to clinical markers only. Second, we addressed the markers’ ability to successfully predict BPD grades as well as the duration of mechanical ventilation, oxygen supplementation, and intensive care treatment. Third, we successfully validated the biomarker results in an independent cohort of preterm infants using a clinically applicable measurement technique, i.e., ELISA and in a final step defined disease specificity in neonatal and adult patients suffering from CLD of different origins.
With the assessment of critical factors that can affect the biomarkers’ clinical performance including their validation in different cohorts and their resilience toward different measurement techniques, we overcame significant limitations of previous studies.
The assessment of 32 models identified the combination of GA together with the protein expression levels as the best model for BPD prediction while demonstrating the reproducible impact of the proteins on disease stratification when compared to known clinical covariates. The comparison between infants with higher disease grades to infants that did not develop BPD unequivocally identifies infants at risk, whereas the comparison of all BPD cases to patients without BPD is limited by the clinical heterogeneity of cases with mild BPD.14 Nonetheless, the model demonstrates good sensitivity and specificity to identify all BPD cases. Confirmation of the results in infants <28 weeks GA further supports the clinical value of the model by demonstrating superiority in a subcohort of very immature infants, in which GA alone is assumed to predict BPD development with significant power.20,21 The demonstration of the model’s potential to separate infants with different BPD grades from infants without the disease in the first week of life, and the validation of the results in an independent sample set with a clinically applicable protein measurement technique, i.e., ELISA, not only underscores the independence of the results from the measurement technique applied but adds significant insight into the performance of the potential biomarkers that is unmet by previous studies.
Reviewing and discussing the identification of reliable markers to predict BPD by previous studies, the use of clinical disease indicators including intrauterine growth restriction,22 low GA or birth weight, male sex,23 RDS, and invasive mechanical ventilation,24 sepsis, asphyxia, and chorioamnionitis,25 as well as race or ethnicity,26,27 and mode of delivery28,29 show only moderate predictive value.21 The limitations might be enhanced by significant changes in the diagnostic or therapeutic process applied to very low birth weight infants over time.30,31 Addressing the need for additional markers, a variety of studies aimed at identifying protein-based biomarkers, with the majority of studies centered around the detection of inflammation.2,3 Here, even the development of multivariate logistic regression models for the outcome of BPD or death at PMA of 36 weeks using protein expression levels of 25 cytokines as suggested by the Neonatal Research Network of the National Institute of Child Health and Human Development (NICHD NRN)32 did not improve disease prediction significantly, potentially due to the use of a pre-selected set of markers as opposed to their identification by unbiased screening. At the same time, markers derived from metabolomic analysis that showcased a cluster of 53 interesting metabolites associated with BPD development33 are most likely limited by the markers sensitivity toward sample collection and processing as well as the analytical platform used. Furthermore, marker detection in tracheal aspirates requires intratracheal intubation for sample acquisition, which becomes rarer with current postnatal treatment strategies.34 Despite significantly informing other fields of lung disease35,36 genetic BPD markers, although informing pathophysiologic understanding,2,37,38 have not been shown to significantly contribute to risk stratification until now.39–42 In contrast, the use of miRNAs already shows promising potential for BPD diagnosis and treatment.3,43,44 In conclusion, the use of protein-based biomarkers may thus be today’s method of choice when generated by unbiased screening such as in the study by Arjaans et al. that used the SOMAmer technology® for the identification of biomarkers that enable the detection of vascular disease in the preterm infants.45 The approach succeeded showing early postnatal changes in circulating angiogenic peptides in association with disease, only limited by the lack of a validation cohort.
Likely, the reflection of the three most important processes of BPD pathophysiology by the three proteins supports their strength for early risk stratification: As the presence of the immune-activating SIGLEC-1446,47 has been previously associated with invasive infections in human newborns48 and the host response to viral airway infections,49 it holds promising potential to reflect the degree of pulmonary inflammation characterizing the BPD lung.50–53 The laminin receptor BCAM likely mirrors the process of tissue remodeling and its associated cellular cross-talk,51,54 thereby potentially reflecting BPD “activity”.55 In line with the findings of Arjaans et al. ,45 ANGPTL-3 also is associated with angiogenic signaling, playing a role in endothelial development and survival, whereas BCAM holds functions in the basal membrane thereby reflecting significant changes to the surrounding niche of the developing alveoli.
Pulmonary expression of the three markers demonstrated in our previous study by the use of paraformaldehyde tissue sections from autopsy preterm BPD lungs6 supports their potential as indicators of BPD pathophysiology and demonstrates that the circulating proteins largely originate from lung tissue in contrast to other studies.2,3,45 In order to further add to previous studies and to showcase clinical applicability of the biomarker, we simulated the circumstances of a clinical trial by the use of a small dataset characterized by random differences in the patient’s characteristics and successfully demonstrated the fit of the data obtained in the distribution of the expression levels of the larger sample sets. Furthermore, we showed disease specificity of the biomarkers for BPD prediction by their application in a neonatal (CLD-CDH cohort) and adult CLD cohort (adult CLD cohort). Here, the proteins in combination with the confounder GA did not allow the stratification of CDH neonates that developed CLD but only separated infants with fatal outcome, whereas adult patients suffering from emphysema (COPD) or lung fibrosis (IPF) could not be separated from pulmonary healthy individuals at all when using the biomarker combination.
While successfully demonstrating the clinical value of the three biomarkers, limitations of our study include (i) cohort size, which we, however, consider adequate for a study in ELGAN’s (extremely low GA newborns) and (ii) the fact that despite the prospective study design, patients cannot be considered randomly assigned to disease groups. To partially compensate for these limitations, we underscored clinical relevance when demonstrating the proteins’ ability to predict main risk factors associated with BPD, i.e., need for mechanical ventilation, oxygen supplementation and NICU stay in the context of GA, as well as by their validation in an independent sample set, resulting in comparable results despite differences in cohort characteristics (Table 1). The limited cohort size was remedied by combining the expression profiles of the three proteins from several proteomic platforms and correcting the observed batch effect as well as the subsequent validation using a different measurement technique, i.e., ELISA in an independent sample set. Although the use of the SOMAscan® or the PEA assay allows for the extensive study of a large number of peptides, peptides playing a role in the pathogenesis of BPD, however, might be missed. These candidates, when identified and validated by other studies, could be added to the model for further improvement.
In conclusion, we demonstrated the promising potential of the identified proteins to inform clinical decision making while considering critical clinical variables. The study significantly adds to previous biomarker studies in the field as it addresses disease specificity and the biomarkers’ robustness toward changes in cohort characteristics or measurement technique.2,3,56 The currently prepared clinical trial aims to prove the biomarkers’ benefit for guiding clinical care. Future studies will have to address the potential of the biomarkers to inform disease monitoring, supported by our previous findings demonstrating stable protein expression levels at day 28 of life6 or to identify disease subtypes dominated by inflammation, matrix remodeling, or vascular pathology.57,58
## Supplementary information
Supplementary table Supplementary information figure The online version contains supplementary material available at 10.1038/s41390-022-02093-w.
## References
1. Jobe AH, Bancalari E. **Bronchopulmonary dysplasia**. *Am. J. Respir. Crit. Care Med.* (2001.0) **163** 1723-1729. DOI: 10.1164/ajrccm.163.7.2011060
2. Lal CV, Ambalavanan N. **Biomarkers, early diagnosis, and clinical predictors of bronchopulmonary dysplasia**. *Clin. Perinatol.* (2015.0) **42** 739-754. DOI: 10.1016/j.clp.2015.08.004
3. Rivera L, Siddaiah R, Oji-Mmuo C, Silveyra GR, Silveyra P. **Biomarkers for bronchopulmonary dysplasia in the preterm infant**. *Front Pediatr.* (2016.0) **4** 33. DOI: 10.3389/fped.2016.00033
4. Geyer PE. **Proteomics reveals the effects of sustained weight loss on the human plasma proteome**. *Mol. Syst. Biol.* (2016.0) **12** 901. DOI: 10.15252/msb.20167357
5. Rohloff JC. **Nucleic acid ligands with protein-like side chains: modified aptamers and their use as diagnostic and therapeutic agents**. *Mol. Ther. Nucleic Acids* (2014.0) **3** e201. DOI: 10.1038/mtna.2014.49
6. Forster K. **Early Identification of bronchopulmonary dysplasia using novel biomarkers by proteomic screening**. *Am. J. Respir. Crit. Care Med.* (2018.0) **197** 1076-1080. DOI: 10.1164/rccm.201706-1218LE
7. Couchard M, Polge J, Bomsel F. **Hyaline membrane disease: diagnosis, radiologic surveillance, treatment and complications**. *Ann. Radio. (Paris)* (1974.0) **17** 669-683
8. Franz AR, Steinbach G, Kron M, Pohlandt F. **Interleukin-8: a valuable tool to restrict antibiotic therapy in newborn infants**. *Acta Paediatr.* (2001.0) **90** 1025-1032. DOI: 10.1111/j.1651-2227.2001.tb01359.x
9. Sherman MP, Goetzman BW, Ahlfors CE, Wennberg RP. **Tracheal asiration and its clinical correlates in the diagnosis of congenital pneumonia**. *Pediatrics* (1980.0) **65** 258-263. DOI: 10.1542/peds.65.2.258
10. Snoek KG. **Conventional mechanical ventilation versus high-frequency oscillatory ventilation for congenital diaphragmatic hernia: a randomized clinical trial (the Vici-Trial)**. *Ann. Surg.* (2016.0) **263** 867-874. DOI: 10.1097/SLA.0000000000001533
11. Holle R, Happich M, Lowel H, Wichmann HE, Group MKS. **Kora-a research platform for population based health research**. *Gesundheitswesen* (2005.0) **67** S19-S25. DOI: 10.1055/s-2005-858235
12. Gold L. **Aptamer-based multiplexed proteomic technology for biomarker discovery**. *PLoS One* (2010.0) **5** e15004. DOI: 10.1371/journal.pone.0015004
13. Assarsson E. **Homogenous 96-Plex pea immunoassay exhibiting high sensitivity, specificity, and excellent scalability**. *PLoS One* (2014.0) **9** e95192. DOI: 10.1371/journal.pone.0095192
14. Thebaud B. **Bronchopulmonary dysplasia**. *Nat. Rev. Dis. Prim.* (2019.0) **5** 78. DOI: 10.1038/s41572-019-0127-7
15. Askie LM, Henderson-Smart DJ, Irwig L, Simpson JM. **Oxygen-saturation targets and outcomes in extremely preterm infants**. *N. Engl. J. Med* (2003.0) **349** 959-967. DOI: 10.1056/NEJMoa023080
16. Bhandari V. **The potential of non-invasive ventilation to decrease BPD**. *Semin. Perinatol.* (2013.0) **37** 108-114. DOI: 10.1053/j.semperi.2013.01.007
17. Carlo WA. **Minimal ventilation to prevent bronchopulmonary dysplasia in extremely-low-birth-weight infants**. *J. Pediatr.* (2002.0) **141** 370-374. DOI: 10.1067/mpd.2002.127507
18. Davis PG. **Caffeine for apnea of prematurity trial: benefits may vary in subgroups**. *J. Pediatr.* (2010.0) **156** 382-387. DOI: 10.1016/j.jpeds.2009.09.069
19. Engle WA, Kominiarek MA. **Late preterm infants, early term infants, and timing of elective deliveries**. *Clin. Perinatol.* (2008.0) **35** 325. DOI: 10.1016/j.clp.2008.03.003
20. Laughon MM. **Prediction of bronchopulmonary dysplasia by postnatal age in extremely premature infants**. *Am. J. Respir. Crit. Care Med.* (2011.0) **183** 1715-1722. DOI: 10.1164/rccm.201101-0055OC
21. Onland W. **Clinical prediction models for bronchopulmonary dysplasia: a systematic review and external validation study**. *BMC Pediatr.* (2013.0) **13** 207. DOI: 10.1186/1471-2431-13-207
22. Gortner L. **Neonatal outcome in small for gestational age infants: do they really better?**. *J. Perinat. Med.* (1999.0) **27** 484-489. DOI: 10.1515/JPM.1999.065
23. Korhonen P, Tammela O, Koivisto AM, Laippala P, Ikonen S. **Frequency and risk factors in bronchopulmonary dysplasia in a cohort of very low birth weight infants**. *Early Hum. Dev.* (1999.0) **54** 245-258. DOI: 10.1016/S0378-3782(98)00101-7
24. Zhang H, Zhang J, Zhao S. **Airway damage of prematurity: the impact of prolonged intubation, ventilation, and chronic lung disease**. *Semin. Fetal Neonatal Med.* (2016.0) **21** 246-253. DOI: 10.1016/j.siny.2016.04.001
25. Hartling L, Liang Y, Lacaze-Masmonteil T. **Chorioamnionitis as a risk factor for bronchopulmonary dysplasia: a systematic review and meta-analysis**. *Arch. Dis. Child Fetal Neonatal Ed.* (2012.0) **97** F8-F17. DOI: 10.1136/adc.2010.210187
26. Rojas MA. **Changing trends in the epidemiology and pathogenesis of neonatal chronic lung disease**. *J. Pediatr.* (1995.0) **126** 605-610. DOI: 10.1016/S0022-3476(95)70362-4
27. Vesoulis ZA, McPherson CC, Whitehead HV. **Racial disparities in calculated risk for bronchopulmonary dysplasia: a dataset**. *Data Brief.* (2020.0) **30** 105674. DOI: 10.1016/j.dib.2020.105674
28. Chen X. **The utility of comprehensive metabolic panel tests for the prediction of bronchopulmonary dysplasia in extremely premature infants**. *Dis. Markers* (2019.0) **2019** 5681954. DOI: 10.1155/2019/5681954
29. Wang K, Huang X, Lu H, Zhang Z. **A comparison of Kl-6 and Clara cell protein as markers for predicting bronchopulmonary dysplasia in preterm infants**. *Dis. Markers* (2014.0) **2014** 736536. DOI: 10.1155/2014/736536
30. El Faleh I. **Bronchopulmonary dysplasia: a predictive scoring system for very low birth weight infants. a diagnostic accuracy study with prospective data collection**. *Eur. J. Pediatr.* (2021.0) **180** 2453-2461. DOI: 10.1007/s00431-021-04045-8
31. Poindexter BB. **Comparisons and limitations of current definitions of bronchopulmonary dysplasia for the prematurity and respiratory outcomes program**. *Ann. Am. Thorac. Soc.* (2015.0) **12** 1822-1830. DOI: 10.1513/AnnalsATS.201504-218OC
32. Ambalavanan N. **Cytokines associated with bronchopulmonary dysplasia or death in extremely low birth weight infants**. *Pediatrics* (2009.0) **123** 1132-1141. DOI: 10.1542/peds.2008-0526
33. Piersigilli F. **Identification of new biomarkers of bronchopulmonary dysplasia using metabolomics**. *Metabolomics* (2019.0) **15** 20. DOI: 10.1007/s11306-019-1482-9
34. Jain D, Bancalari E. **New developments in respiratory support for preterm infants**. *Am. J. Perinatol.* (2019.0) **36** S13-S17. DOI: 10.1038/s41372-019-0471-1
35. Kropski JA. **Genetic evaluation and testing of patients and families with idiopathic pulmonary fibrosis**. *Am. J. Respir. Crit. Care Med.* (2017.0) **195** 1423-1428. DOI: 10.1164/rccm.201609-1820PP
36. 36.Morrell, N. W. et al. Genetics and genomics of pulmonary arterial hypertension. Eur. Respir. J.53, 1801899 (2019).
37. Bhandari V. **Familial and genetic susceptibility to major neonatal morbidities in preterm twins**. *Pediatrics* (2006.0) **117** 1901-1906. DOI: 10.1542/peds.2005-1414
38. Lavoie PM, Pham C, Jang KL. **Heritability of bronchopulmonary dysplasia, defined according to the consensus statement of the National Institutes of Health**. *Pediatrics* (2008.0) **122** 479-485. DOI: 10.1542/peds.2007-2313
39. Ambalavanan N. **Integrated genomic analyses in bronchopulmonary dysplasia**. *J. Pediatr.* (2015.0) **166** 531-537.e513. DOI: 10.1016/j.jpeds.2014.09.052
40. Floros J. **Il-18r1 and Il-18rap Snps may be associated with bronchopulmonary dysplasia in African-American infants**. *Pediatr. Res.* (2012.0) **71** 107-114. DOI: 10.1038/pr.2011.14
41. Huusko JM. **A study of genes encoding cytokines (Il6, Il10, Tnf), cytokine receptors (Il6r, Il6st), and glucocorticoid receptor (Nr3c1) and susceptibility to bronchopulmonary dysplasia**. *BMC Med. Genet* (2014.0) **15** 120. DOI: 10.1186/s12881-014-0120-7
42. Li J. **Exome sequencing of neonatal blood spots and the identification of genes implicated in bronchopulmonary dysplasia**. *Am. J. Respir. Crit. Care Med.* (2015.0) **192** 589-596. DOI: 10.1164/rccm.201501-0168OC
43. Gronbach J. **The potentials and caveats of mesenchymal stromal cell-based therapies in the preterm infant**. *Stem Cells Int.* (2018.0) **2018** 9652897. DOI: 10.1155/2018/9652897
44. Shrestha A. **A critical role for Mir-142 in alveolar epithelial lineage formation in mouse lung development**. *Cell Mol. Life Sci.* (2019.0) **76** 2817-2832. DOI: 10.1007/s00018-019-03067-8
45. Arjaans S. **Early angiogenic proteins associated with high risk for bronchopulmonary dysplasia and pulmonary hypertension in preterm infants**. *Am. J. Physiol. Lung Cell Mol. Physiol.* (2020.0) **318** L644-L654. DOI: 10.1152/ajplung.00131.2019
46. Angata T, Hayakawa T, Yamanaka M, Varki A, Nakamura M. **Discovery of Siglec-14, a novel sialic acid receptor undergoing concerted evolution with Siglec-5 in primates**. *FASEB J.* (2006.0) **20** 1964-1973. DOI: 10.1096/fj.06-5800com
47. Yamanaka M, Kato Y, Angata T, Narimatsu H. **Deletion polymorphism of Siglec14 and its functional implications**. *Glycobiology* (2009.0) **19** 841-846. DOI: 10.1093/glycob/cwp052
48. Ali SR. **Siglec-5 and Siglec-14 are polymorphic paired receptors that modulate neutrophil and amnion signaling responses to group B Streptococcus**. *J. Exp. Med.* (2014.0) **211** 1231-1242. DOI: 10.1084/jem.20131853
49. Angata T. **Loss of Siglec-14 reduces the risk of chronic obstructive pulmonary disease exacerbation**. *Cell Mol. Life Sci.* (2013.0) **70** 3199-3210. DOI: 10.1007/s00018-013-1311-7
50. Bhandari V. **Hyperoxia-derived lung damage in preterm infants**. *Semin Fetal Neonatal Med.* (2010.0) **15** 223-229. DOI: 10.1016/j.siny.2010.03.009
51. 51.Blackwell, T. S. et al. Nf-Κb signaling in fetal lung macrophages disrupts airway morphogenesis. J. Immunol.187, 2740–2747 (2011).
52. Bose CL, Dammann CE, Laughon MM. **Bronchopulmonary dysplasia and inflammatory biomarkers in the premature neonate**. *Arch. Dis. Child Fetal Neonatal Ed.* (2008.0) **93** F455-F461. DOI: 10.1136/adc.2007.121327
53. Wallace MJ. **Early biomarkers and potential mediators of ventilation-induced lung injury in very preterm lambs**. *Respir. Res.* (2009.0) **10** 19. DOI: 10.1186/1465-9921-10-19
54. Bland RD. **Mechanical ventilation uncouples synthesis and assembly of elastin and increases apoptosis in lungs of newborn mice. prelude to defecite alveolar septation during lung development?**. *Am. J. Physiol. Lung Cell Mol. Physiol.* (2008.0) **294** L3-L14. DOI: 10.1152/ajplung.00362.2007
55. Thibeault DW, Mabry SM, Ekekezie II, Zhang X, Truog WE. **Collagen scaffolding during development and its deformation with chronic lung disease**. *Pediatrics* (2003.0) **111** 766-776. DOI: 10.1542/peds.111.4.766
56. Zhang ZQ, Huang XM, Lu H. **Early biomarkers as predictors for bronchopulmonary dysplasia in preterm infants: a systematic review**. *Eur. J. Pediatr.* (2014.0) **173** 15-23. DOI: 10.1007/s00431-013-2148-7
57. Baker CD, Alvira CM. **Disrupted lung development and bronchopulmonary dysplasia: opportunities for lung repair and regeneration**. *Curr. Opin. Pediatr.* (2014.0) **26** 306-314. DOI: 10.1097/MOP.0000000000000095
58. Camenisch G. **Angptl3 stimulates endothelial cell adhesion and migration via integrin alpha Vbeta 3 and induces blood vessel formation in vivo**. *J. Biol. Chem.* (2002.0) **277** 17281-17290. DOI: 10.1074/jbc.M109768200
|
---
title: The Associations Between Alanine Aminotransferase and Other Biochemical Parameters
in Lean PCOS
authors:
- Cai Liu
- Kai Liu
- Xiao Zhao
- Junhua Zhu
- Yang Liu
- Lina Hao
- Yanyun Gao
- Peng Liu
journal: Reproductive Sciences
year: 2022
pmcid: PMC9988735
doi: 10.1007/s43032-022-01030-w
license: CC BY 4.0
---
# The Associations Between Alanine Aminotransferase and Other Biochemical Parameters in Lean PCOS
## Abstract
To explore the associations of alanine aminotransferase in lean women of polycystic ovary syndrome (PCOS) with other biochemical parameters and the potential risk factors. This is a retrospective cohort study with lean PCOS ($$n = 91$$) and healthy controls ($$n = 45$$); we reviewed the electrical records and databases of the PCOS patients in our infertility clinic between January 2019 and September 2021; independent t-test, linear correlation analysis, and multiple linear regression were used to explore the associations. Higher levels of luteinizing hormone, total testosterone, thyroid stimulating hormone, platelet count, lymphocyte count, homocysteine, alanine aminotransferase (ALT), and uric acid were identified in lean PCOS patients, while follicle-stimulating hormone level was lower in in lean PCOS as expected ($P \leq 0.05$). Of note, the linear correlation showed that BMI, total testosterone, white blood cell count, lymphocyte count, aspartate aminotransferase, and uric acid were positively associated with alanine aminotransferase ($r = 0.232$, 0.318, 0.218, 0.388, 0.602, 0.353 respectively, $P \leq 0.05$). After multiple linear regression was performed, total testosterone and aspartate aminotransferase were independently and positively correlated with alanine aminotransferase in lean PCOS ($B = 0.251$, 0.605 respectively, $P \leq 0.05$). Higher level of ALT was identified in the lean PCOS. BMI, white blood cell count, lymphocyte count, aspartate aminotransferase, uric acid, and total testosterone were positively correlated with ALT in lean PCOS. Total testosterone and aspartate aminotransferase were independently and positively associated with ALT in lean PCOS after multiple linear regression. There might exist a potential risk of afflicting liver impairment for the lean PCOS women in the earlier period. Early examination and intervention might be necessary to prevent or delay the progression of the liver disease as soon as the diagnosis of PCOS.
## Introduction
Polycystic ovary syndrome (PCOS) is a common endocrine disorder and a leading cause of infertility, the prevalence ranges between 5 and $21\%$ of the reproductive women depending on the different definitions and studies [1, 2]. Stein and Leventhal described it in 1935 initially, along with the development of reproductive medicine, evidences have increasingly demonstrated that PCOS is a polygenic, polyfactorial, inflammatory, and autoimmune disease [3]; the physiopathologic mechanism is complicated and still unclear until today. A growing number of studies [4–6] suggest that PCOS is closely correlated with cardiovascular diseases and metabolic syndrome, such as type 2 diabetes mellitus and non-alcoholic fatty liver disease (NAFLD). Since simple steatosis may be a benign course, but steatohepatitis can lead to cirrhosis or hepatocellular carcinoma eventually [7], greatly damaging woman’s health; therefore, it seems extremely crucial to prevent the progression of the liver disease in clinical practice. Alanine aminotransferase (ALT), as a sensitive indicator of liver function, is increasingly significant and widely used in clinical practice; it is often treated as an initial marker of hepatic impairment and liver inflammation, and more sensitive than aspartate aminotransferase (AST) [8]. Studies [8, 9] demonstrated that there was a higher level of ALT among the PCOS patients, mainly owing to the increased prevalence of obesity, hyperandrogenism, insulin resistance, and dyslipidemia of the PCOS. However, the correlation between ALT and obese PCOS was reported more often than the lean one; moreover, the current situation is that the liver function test is not recommended routinely unless the patient is overweight or obese. This may cause the lean patients miss the best opportunity for intervention, such as modify lifestyles and make regular checking to prevent or delay the progression of the disease. This is why we conduct this research to seek for the correlation of ALT with other biochemical parameters in lean PCOS and the potential risk factors, offering some clinical evidences to elucidate the intricate pathogenic mechanism further.
## Study Population
We reviewed the electrical records and databases of the PCOS patients who came to Infertility Clinic of Yulin First Hospital (a tertiary hospital in Shaanxi province, China) between January 2019 and September 2021 for infertility, and the controls were also extracted from our clinic record who came for a pre-pregnancy check-up, and all the tests of the controls are within the normal level and their normal ovulatory functions were confirmed by vaginal sonography. Clinical and laboratory materials were collected from electronic medical records; we also made a call to some of them for gathering the necessary information. This study was approved by the institutional research ethics review board of Yulin First Hospital (2022–001). Given the retrospective nature of the work, no specific consent was required from the patients.
The criteria of PCOS was diagnosed according to *Rotterdam criteria* [10], oligo-/anovulation, clinical or biochemical hyperandrogenism, ultrasound diagnosis of polycystic ovary morphology (PCOM), defined as the presence of at least one ovary > 10 ml, or 12 or more antral follicles 2–9 mm in diameter. PCOS was diagnosed if only two of the three are present, and patients with other causes of hyperandrogenism (congenital adrenal hyperplasia, Cushing’s syndrome, and androgenic-secreting tumors) and ovulation dysfunction, such as functional hypothalamic amenorrhea (FHA), thyroid dysfunction, and hyperprolactinemia (HPRL), were excluded. Exclusion criteria also included patients with diabetes, hypertension, hepatitis, endometriosis, recurrent pregnancy loss, any acute or chronic inflammation of the whole body, the people treated with any medication in 3 months, and with smoking and alcoholism history.
## Variables
Body mass index (BMI) was calculated by weight in kilograms divided by the square of the height in meters (kg/m2), and the one whose BMI ≥ 25 kg/m2 were excluded. The requirement of age was 20–40 years. The hormonal blood test was performed during the of 2–4th day of the menstrual period, and the biochemical blood was asked for an overnight fasting. Finally, we identified 91 patients of lean PCOS women as the study population and 45 healthy women as the controls. We collected general characteristics, the biochemical markers: the luteinizing hormone (LH), follicle-stimulating hormone (FSH), total testosterone (T), prolactin (PRL), thyroid-stimulating hormone (TSH), white blood cell count (WBC), neutrophil count (NEUT), lymphocyte count (LYMPH), monocyte count (MONO), platelet count (PLT), fasting plasma glucose (FPG), homocysteine (Hcy), erythrocyte sedimentation rate (ESR), ALT, AST, uric acid (UA), creatinine, urea, and cystatin levels. Owing to a retrospective study, we failed to get the data of insulin and lipid for they were not routine ones for the lean PCOS in the past.
## Statistical Analysis
PSS 23.0 was applied for all of the analysis, comparison of continuous variables were tested with Independent t-test or Mann–Whitney test. Pearson correlation analyses were used to evaluate correlations between continuous variables. Comparisons between groups were performed using single-factor ANOVA or non-parametric tests, and multiple linear regression was used to analyze independent correlated factors; $P \leq 0.05$ was considered to indicate statistical significance.
## General and Biochemical Indexes of Lean PCOS Women Compared with Healthy Controls
The two groups are comparable in age and BMI. However, higher levels of LH ($t = 6.166$, $$P \leq 0.000$$), total testosterone (Z = − 5.535, $$P \leq 0.000$$), TSH (Z = − 2.130, $$P \leq 0.033$$), platelet count ($t = 2.607$, $$P \leq 0.010$$), lymphocyte count ($t = 2.199$, $$P \leq 0.029$$), homocysteine (Z = − 3.517, $$P \leq 0.000$$), ALT (Z = − 2.436, $$P \leq 0.015$$), and uric acid ($t = 3.715$, $$P \leq 0.000$$) were identified in lean PCOS; FSH level was significantly lower in in lean PCOS as expected (t = − 2.199, $$P \leq 0.029$$). No statistical significance was identified in PRL, ESR, neutrophil count, monocyte count, FPG, AST, creatinine, urea, and cystatin levels between the lean PCOS and controls (Table 1).Table1General and biochemical indexes of lean PCOS compared with controlsLean PCOSControlsP valuean9145Age (y)29.46 ± 3.5030.38 ± 2.976.096BMI (kg/m2)22.22 ± 1.6421.97 ± 1.45.342LH (mIU/ml)11.28 ± 10.045.16 ± 1.99.000FSH (mIU/ml)6.19 ± 1.656.83 ± 1.59.029T (ng/ml)0.3(0.19–0.45)0.15(0.09–0.23).000PRL (ng/ml)14.34 ± 5.5616.11 ± 6.56.094TSH (uIU/ml)2.68(1.94–3.74)2.21(1.7–3.1).029WBC(× 109/L)6.30 ± 1.536.22 ± 1.61.760PLT (× 109/L)252.25 ± 60.06227.60 ± 49.63.010MONO/L (× 109/L)0.39 ± 0.100.40 ±.12.371LYMPH/L (× 109/L)2.16 ±.621.94 ±.53.029NEUT/L (× 109/L)3.65 ± 1.123.75 ± 1.29.603ESR (mm/h)4.40 ± 3.624.93 ± 3.01.390Hcy (umol/L)12.1(9.7–17.7)10.25(8.1–12.3).000FPG (mmol/L)5.23 ±.375.22 ±.31.942ALT (U/L)15(12.0–21.2)12(10–15.7).0015AST (U/L)17.54 ± 4.1317.47 ± 4.26.930AST/ALT1.22 ± 0.551.35 ± 0.34.130UA (umol/L)271.00 ± 61.89228.63 ± 51.28.000Crea(umol/L)51.51 ± 9.0151.99 ± 8.99.767Urea(mmol/L)4.15 ±.894.30 ± 1.12.439Cystatin(mg/L).68 ±.10.67 ±.10.566Values are expressed as mean ± SD for normal distribution or median (IQR) for non-normal distributionaDifferences between two groups were analyzed by independent T-test or Mann–Whitney testBMI, body mass index; LH, luteinizing hormone; FSH, follicle-stimulating hormone; T, total testosterone; PRL, prolactin; TSH, thyroid-stimulating hormone; WBC, white blood cell count; NEUT, neutrophil count; LYMPH, lymphocyte count; MONO, monocyte count; PLT, platelet count; FPG, fasting plasma glucose; Hcy, homocysteine; ESR, erythrocyte sedimentation rate; ALT, alanine aminotransferase; AST, aspartate aminotransferase; UA, uric acid; Crea, creatinine
## Comparison of General and Biochemical Characteristics of Lean PCOS Women Between ALT Tertiles
We divided the lean PCOS patients into three subgroups based on the tertiles of ALT levels for the small sample, lower level of ALT ($$n = 31$$), middle level of ALT ($$n = 31$$), and higher level of ALT ($$n = 29$$). Of note, BMI ($F = 4.8$, $$P \leq 0.011$$), total testosterone (χ2 = 16.659, $$P \leq 0.000$$), lymphocyte count ($F = 5.686$, $$P \leq 0.0005$$), AST level ($F = 7.903$, $$P \leq 0.001$$), and UA ($F = 45.138$, $$P \leq 0.008$$) were significantly different; then, we compared the significant factors of every two groups one by one. The results illustrated a higher BMI ($95\%$CI − 1.94 to − 1.95, $$P \leq 0.003$$) in the H-ALT than L-ALT level, but no statistical significance was identified between the other two groups ($95\%$CI − 1.29 to 0.20, $$P \leq 0.152$$) ($95\%$CI − 1.40 to 0.11, $$P \leq 0.097$$). Compared with the L-ALT group, higher total testosterone were showed in the M-ALT ($95\%$CI − 0.18 to − 0.30, $$P \leq 0.007$$) and H-ALT ($95\%$CI − 0.25 to − 0.94, $$P \leq 0.000$$), but no statistical significance between M-ALT and H-ALT groups ($95\%$CI − 0.01 to 0.14, $$P \leq 0.101$$). AST in the M-ALT ($95\%$CI − 4.38 to − 0.50, $$P \leq 0.014$$) and H-ALT ($95\%$CI − 5.88 to − 1.92, $$P \leq 0.000$$) was remarkably higher than L-ALT level; however, there was no marked difference between the other two groups ($95\%$CI − 3.42 to 0.52, $$P \leq 0.142$$). Higher level of UA was displayed in the H-ALT group than M-ALT ($95\%$CI − 75.69 to − 8.46, $$P \leq 0.015$$) and L-ALT ($95\%$CI − 87.69 to − 17.64, $$P \leq 0.004$$). Lymphocyte count was significantly higher in H-ALT group than L-ALT ($95\%$CI − 0.79 to − 0.17, $$P \leq 0.003$$) and M-ALT group ($95\%$CI − 0.73 to − 0.11, $$P \leq 0.008$$) and no significance between L-ALT and M-ALT group ($95\%$CI − 0.24 to 0.36, $$P \leq 0.708$$). The white blood cell count was growing gradually from the L-ALT group to H-ALT; we also explored it between every two groups. Surprisingly, white blood count level in H-ALT was significantly higher than L-ALT group (($95\%$CI − 1.49 to − 0.01, $$P \leq 0.0047$$), but no statistical significance between the other two groups ($95\%$CI − 0.94 to 0.54, $$P \leq 0.574$$) ($95\%$CI − 1.29 to 0.20, $$P \leq 0.148$$) (Table 2, Fig. 1).Table 2Comparison of general and biochemical characteristics of lean PCOS between ALT tertilesL-ALT (group1)M-ALT (group2)H-ALT (group3)P valuean313129Age (y)30.32 ± 3.5929.77 ± 3.5329.45 ± 3.45.624BMI (kg/m2)21.67 ± 1.6322.21 ± 1.2922.86 ± 1.50.011LH (mIU/ml)8.91 ± 5.5612.24 ± 16.3911.10 ± 6.44.481FSH (mIU/ml)6.52 ± 1.836.16 ± 1.166.02 ± 1.77.482T (ng/ml)0.2(0.16–0.27)0.33(0.23–0.44)0.39(0.22–0.57).000PRL (ng/ml)14.24 ± 5.7014.51 ± 5.4815.66 ± 5.71.617TSH (uIU/ml)2.85 ± 1.062.76 ± 1.483.10 ± 1.39.607WBC(× 109/L)6.02 ± 1.196.23 ± 1.726.78 ± 1.37.122PLT (× 109/L)244.70 ± 63.67242.29 ± 52.57270.93 ± 58.98.119MONO/L (× 109/L)0.35 ±.090.39 ±.110.40 ±.09.134LYMPH/L (× 109/L)1.92 ±.481.97 ±.562.40 ±.73.005NEUT/L (× 109/L)3.59 ± 1.023.70 ± 1.423.84 ±.95.721ESR (mm/h)4.68 ± 4.313.64 ± 2.455.12 ± 3.96.319HCY (umol/L)13.97 ± 6.5415.88 ± 11.0613.50 ± 6.80.565FPG (mmol/L)5.23 ±.275.20 ±.395.21 ±.42.954AST (U/L)15.46 ± 2.9517.91 ± 4.0819.36 ± 4.39.001UA (umol/L)250.46 ± 50.64261.05 ± 59.00303.13 ± 65.57.008Crea (umol/L)51.93 ± 8.3252.58 ± 9.0449.85 ± 9.74.498Urea(mmol/L)4.14 ±.884.23 ±.934.09 ±.89.854Cystatin(mg/L)0.67 ± 0.930.68 ±.100.68 ± 0.10.772Values are expressed as mean ± SD for normal distribution or median (IQR) for non-normal distribution aDifferences between subgroups were analyzed by the single-factor ANOVA or non-parametric testsBMI, body mass index; LH, luteinizing hormone; FSH, follicle-stimulating hormone; T, total testosterone; PRL, prolactin; TSH, thyroid-stimulating hormone; WBC, white blood cell count; NEUT, neutrophil count; LYMPH, lymphocyte count; MONO, monocyte count; PLT, platelet count; FPG, fasting plasma glucose; Hcy, homocysteine; ESR, erythrocyte sedimentation rate; ALT, alanine aminotransferase; AST, aspartate aminotransferase; UA, uric acid; Crea, creatinineFig. 1The specific differences of BMI, lymphocyte count, total testosterone, uric acid, AST, and white blood cell count between every two groups of the ALT tertiles in lean PCOS women. ALT, alanine aminotransferase; WBC, white blood cell count; BMI, body mass index; UA, uric acid; AST, aspartate aminotransferase; tertiles of ALT: 1, lower level of ALT group; 2, middle level of ALT group; 3, higher level of ALT group
## Linear Correlation of ALT Level with the Hormone Indicators in Lean PCOS
As illustrated in Table 3, BMI and total testosterone were significantly and positively associated with ALT level ($r = 0.232$, 0.318 respectively, $P \leq 0.05$). No obvious linear association was identified in the level between ALT and LH, FSH, TSH, and PRL level ($P \leq 0.05$).Table 3The linear correlations of ALT with the hormone indicators in lean PCOSage (y)BMI (kg/m2)LH (mIU/ml)FSH (mIU/ml)TSH (uIU/ml)T (ng/ml)PRL (ng/ml)ALTr − 0.0910.2320.072 − 0.2050.0970.3180.142Pa0.3930.0270.5090.0590.3730.0030.195aP value for test of significance of the associations using the Pearson correlation analysis. BMI, body mass index; LH, luteinizing hormone; FSH, follicle-stimulating hormone; T, total testosterone; PRL, prolactin; TSH, thyroid-stimulating hormone
## Linear Correlation of ALT Level with the Inflammatory Indicators in Lean PCOS
As illustrated in Table 4, white blood cell count and lymphocyte count were significantly and positively associated with ALT ($r = 0.218$,0.388 respectively, $P \leq 0.05$). No obvious linear association was identified in the level between ALT and neutrophil count, monocyte count, platelet count, ESR, and homocysteine level ($P \leq 0.05$).Table 4The linear correlations of ALT with the inflammatory indicators in lean PCOSWBC(× 109/L）NEUT(× 109/L)LIMPH(× 109/L)MONO(× 109/L)PLT(× 109/L)ESR (mm/h)Hcy (umol/L)ALTr0.2180.5520.0000.0930.161 − 0.019 − 0.048Pa0.0380.0630.3880.3830.1280.8650.676aP value for test of significance of the association using the Pearson correlation analysisWBC, white blood cell count; NEUT, neutrophil count; LYMPH, lymphocyte count; MONO, monocyte count; PLT, platelet count; ESR, erythrocyte sedimentation rate; Hcy, homocysteine
## Linear Correlation of ALT Level with the Metabolic Indexes in Lean PCOS
As illustrated in Table 5, AST and uric acid were positively and significantly associated with ALT level ($r = 0.602$, 0.353 respectively, $P \leq 0.05$). No obvious linear association was identified in the level between ALT and FPG, creatinine, urea, and cystatin level ($P \leq 0.05$).Table 5The linear correlations of ALT with the metabolic indexes in lean PCOSFPG (mmol/L)UA (umol/L)Crea (umol/L)Urea(mmol/LAST (U/L)Cystatin(mg/L)ALTr0.0610.353-0.155-0.0360.6020.049Pa0.5740.0030.1440.7660.0000.644aP value for test of significance of the association using the Pearson correlation analysisALT, alanine aminotransferase; FPG, fasting plasma glucose; UA, uric acid; AST, aspartate aminotransferase; Crea, creatinine
## Multiple Linear Regression of Tertiles of ALT with the Metabolic Markers in Lean PCOS
Multiple linear regression was performed to analyze the independent correlations between ALT and other parameters if there was a statistically significant association with ALT in the univariate regression analysis or if it was clinically indicated. Total testosterone ($B = 0.251$, $P \leq 0.01$) and AST ($B = 0.605$, $P \leq 0.01$) were identified to be independently and positively correlated with ALT in lean PCOS. No significantly independent correlation was found in BMI ($B = 0.113$, $$P \leq 0219$$), uric acid ($B = 0.160$, $$P \leq 0.092$$), white blood count ($B = 0.067$, $$P \leq 0.088$$), and lymphocyte count ($B = 0.174$, $$P \leq 0.073$$) with ALT (Table 6).Table 6The multiple linear regression of ALT with the biochemical indexes in lean PCOSParameterBtP valueaBMI (kg/m2)0.1131.2420.219WBC (× 109/L)0.0670.7120.088Lymphocyte (× 109/L)0.1741.8250.073AST (U/L)0.6056.6790.000T (ng/ml)0.2512.7750.007UA (umol/L)0.1601.710.092aP value for test of significance of the association using the multiple linear correlation analysisBMI, body mass index; WBC, white blood cell count; AST, aspartate aminotransferase; T, total testosterone; UA, uric acid
## Discussion
PCOS is a prevalent reproductive endocrine disorder, often accompanied by infertility, metabolic syndrome, and cardiovascular disease. Our report demonstrated consistently with previous researches [11] that LH is higher and FSH is lower in lean PCOS than the controls. Besides, despite we excluded the one with the thyroid dysfunction, TSH level was significantly higher in lean PCOS, revealing a similar result with other reports [12]. PCOS women are prone to have higher TSH levels owing to the disorder of hypothalamic-pituitary-ovarian axis (HPOA) in PCOS as we all know. Additionally, several studies [12, 13] suggest that higher TSH level is also correlated with metabolic syndrome in PCOS; Emel et al. [ 14] revealed that obese children demonstrate an increase in TSH levels as the degree of hepatic steatosis increased. Recently, another study [15] reported a strong link between TSH level and NAFLD proved by biopsy, independent of obesity, suggesting that thyroid hormone directly affect the synthesis and metabolism of cholesterol and fatty acids in an autonomous way by regulating the transcription of target genes involved in liver metabolism. However, our study failed to show statistical significance between ALT and TSH levels in lean PCOS, perhaps owing to the normal range of TSH levels in our study. This issue deserves to be explored further on account of its higher incidence in PCOS. In addition, PRL is negatively associated with AST, ALT, even after adjusting for age and BMI [16], indicating that lower serum PRL may damage liver cells, but the specific mechanism is unclear currently. However, there was no statistical significance showed in our study, and the reason behind this remains unknown, which needs to be studied in the future.
Furthermore, it is well-known that hyperandrogenism is a principal feature of PCOS, excessive production of androgen is the leading cause of the PCOS [17]. Our study demonstrated congruently with the common view that total testosterone is higher than controls in lean PCOS women. What is widespread acknowledged is that hyperandrogenism plays a role in almost all the complications of PCOS, for example, hyperandrogenism is also implicated tightly in elevated level of uric acid [18], which often accompanied by metabolic disease and cardiovascular disease [19]. Total testosterone was the independent risk factor of ALT in lean PCOS in our study, in line with earlier study [20], mainly because androgen can adversely affect mitochondrial function of liver cells, cause the imbalance between apoptosis and autophagy, resulting in liver damage [21]. It also affects the pathway of branched chain amino acid and the degradation of related mitochondrial enzymes, aggravating liver injury [22]. Therefore, PCOS with higher androgen are more predisposed to liver damage, which should be paid more attention in practice, regardless of their weight.
Simultaneously, homocysteine in the lean PCOS was also significantly higher than the healthy controls; this is beneficial in corroborating the higher risk of cardiovascular disease in PCOS for the people with higher homocysteine incline to get microthrombus in the vessels [23], and considered to be an independent risk factor for atherogenic and thrombotic components of various systems [24]. Besides, elevated homocysteine level is also tightly linked with fatty liver and chronic kidney disease [25, 26], while others [27] observed the opposite; they [27, 28] supposed the homocysteine levels are more higher in severe liver disease, but not in the mild one. According to a recent report [29], consensus on this issue has not reached yet, and the mechanism remains unknown. However, no significant correlation between the homocysteine levels with ALT was found in lean PCOS here. Since PCOS women seem to show more higher levels of homocysteine, the specific mechanism and effects on liver disease should be performed further by a well-designed and prospective study to clarify the associations between them.
In this study, the platelet count and lymphocyte count are remarkably higher in lean PCOS than healthy controls, supported by other researches [30–32]; it may be expounded by the mechanism of chronic inflammation in PCOS [3, 33], since the inflammatory state of PCOS may trigger an increased platelet count, but the higher platelet does not correlate with the inflammation markers [32]; therefore, the preexisting procoagulant state in PCOS might be caused by platelet activation and endothelial dysfunction [34]. However, our study failed to demonstrate marked differences in white blood cell count, neutrophil count, monocyte count, and erythrocyte sedimentation rate. Of interest, when we divided the lean PCOS into three subgroups and found that white blood cell count, lymphocyte count is positively associated with ALT levels in the lean PCOS. As mentioned above, being inflammation markers, perhaps white blood cell count and lymphocyte count also play roles in the higher levels of ALT in lean PCOS. Mounts of evidences [3, 35] suggest that PCOS is a state of chronic low-grade inflammation; immune system will activate while sensing the inflammatory factors. As a key metabolic organ, there might exist underlying inflammation in the liver cells in spite of the mild higher or normal range of ALT levels, leading to chronic liver damage.
ALT, a readily available, inexpensive, and routine metabolic marker used in clinical practice [36], has been observed elevated in various metabolic disorders, such as obesity, hyperlipidemia, and diabetes mellitus [37]. Even though our study population are lean ones and the ALT levels did not show clinical significance in practice, a remarkably statistical difference was also demonstrated in lean PCOS women. Being a good predictor of liver damage, ALT reflects more sensitively in variations of the liver [38], which reminds us of the liver injury may exist in the lean PCOS women in an earlier period during which we might ignore before, and they appear to be at higher risks in developing metabolic disease. Perhaps, we should also advise the lean PCOS women to make regular liver function checking and modify their lifestyles, such as making exercise, changing eating habits as early as possible to prevent the progression of liver disease; simultaneously, some researches were published on the reverse of the ALT level; for instance, Javed et al. [ 39] reported an improvement on the marker of liver injury and fibrosis through vitamin D supplementation in overweight and obese PCOS. Certainly, some better indicators and advice are still necessary for us to explore in the future. Additionally, uric acid was also statistically significant between the lean PCOS and controls in our study; however, there were no significant differences in FPG, creatinine, urea, cystatin, and AST levels. It is noting that people who have higher ALT levels tend to have higher uric acid and AST levels in our study. A study published lately demonstrated an independent and significant correlation between hyperuricemia and ALT level, even after adjusting for potential confounders [40], suggesting that insulin resistance, metabolic syndrome, and systemic inflammation might be caused by hyperuricemia, rather than a simple marker [40, 41], leading to steatohepatitis or even aggravating alcoholic or viral hepatitis [42]. Therefore, people with higher ALT level are more likely to develop severe metabolic disease and cardiovascular disease. However, uric acid level is not independently related with ALT level in our study after multiple factor analysis, which might be expounded by a mild metabolic matter in lean PCOS. AST, as another indicator of liver function, is independently correlated with ALT here, which may corroborate the fact that the combination has better sensitivity in clinical practice. Furthermore, our report also failed to show significant differences in FPG between subgroups, which is consistent with Belan M [43] rather than Chen MJ [8]; we speculate this may be related with the lean PCOS in our study population; for the elevated fasting, glucose is more linked with obesity, not PCOS [44]. Above all, ALT level is higher in lean PCOS women than healthy controls, affected by many metabolic parameters, and independently correlated with AST and total testosterone.
Our strength is that our data were truly from the infertility clinic, which represents the generality of lean PCOS patients and the laboratory tests were all from the same lab of our hospital; we excluded the one who had their partial tests out of our clinic, which also contributed to the small sample size. In addition, to our knowledge, this is one of the few studies to focus on the correlations of ALT level with inflammation markers, hormonal indicators, and metabolic indexes together in lean PCOS. What matters most is that the liver damage in lean PCOS are usually be ignored in practice owing to their normal weight. We have to admit that our study is a small and retrospective sample; hence, the incomplete information is unavoidable, such as the relations of ALT in the specific classification of PCOS needs to explore further in the future; the data of insulin resistance and blood lipid were absent for the these were not routine examinations for the lean PCOS in our clinic in the past. Patients with fatty liver were also unable to be specified or excluded due to the lack of ultrasound examination. Therefore, some large, well-designed and prospective studies are extremely necessary to ascertain our findings.
## Conclusions
In our study, higher ALT level was identified in the lean PCOS women. BMI, white blood cell count, lymphocyte count, AST, uric acid, and total testosterone were positively correlated with ALT in lean PCOS. Total testosterone and AST were independently and positively associated with ALT in lean PCOS after multiple linear regression. Our report reminds us of the potential risk of afflicting liver damage for the lean PCOS in the early period. Early examination and intervention might be necessary to prevent or delay the progression of the liver disease as soon as the diagnosis of PCOS, regardless of their weight. Surely, this study is a small sample and restricted to Chinese Han women, and further study is necessary to ascertain our findings.
## References
1. Azziz R, Carmina E, Chen Z, Dunaif A, Laven JS, Legro RS, Lizneva D, Natterson-Horowtiz B, Teede HJ, Yildiz BO. **Polycystic ovary syndrome**. *Nat Rev Dis Primers* (2016) **2** 16057. DOI: 10.1038/nrdp.2016.57
2. Joham AE, Boyle JA, Ranasinha S, Zoungas S, Teede HJ. **Contraception use and pregnancy outcomes in women with polycystic ovary syndrome: data from the Australian Longitudinal Study on Women’s Health**. *Hum Reprod* (2014) **29** 802-808. DOI: 10.1093/humrep/deu020
3. Patel S. **Polycystic ovary syndrome (PCOS), an inflammatory, systemic, lifestyle endocrinopathy**. *J Steroid Biochem Mol Biol* (2018) **182** 27-36. DOI: 10.1016/j.jsbmb.2018.04.008
4. Spremović Rađenović S, Pupovac M, Andjić M, Bila J, Srećković S, Gudović A, Dragaš B, Radunović N. **Prevalence, risk factors, and pathophysiology of nonalcoholic fatty liver disease (NAFLD) in women with polycystic ovary syndrome (PCOS)**. *Biomedicines* (2022) **10** 131. DOI: 10.3390/biomedicines10010131
5. Heida KY, Bots ML, de Groot CJ, van Dunné FM, Hammoud NM, Hoek A. **Cardiovascular risk management aTer reproductive and pregnancy-related disorders: a Dutch multidisciplinary evidence-based guideline**. *Eur J Prev Cardiol* (2016) **17** 1863-1879. DOI: 10.1177/2047487316659573
6. Azziz R. **Polycystic Ovary Syndrome**. *Obstet Gynecol* (2018) **132** 321-336. DOI: 10.1097/AOG.0000000000002698
7. Liu Z, Que S, Xu J, Peng T. **Alanine aminotransferase-old biomarker and new concept: a review**. *Int J Med Sci* (2014) **11** 925-935. DOI: 10.7150/ijms.8951
8. Chen MJ, Chiu HM, Chen CL, Yang WS, Yang YS, Ho HN. **Hyperandrogenemia is independently associated with elevated alanine aminotransferase activity in young women with polycystic ovary syndrome**. *J Clin Endocrinol Metab* (2010) **95** 3332-3341. DOI: 10.1210/jc.2009-2698
9. Chen MJ, Ho HN. **Hepatic manifestations of women with polycystic ovary syndrome**. *Best Pract Res Clin Obstet Gynaecol* (2016) **37** 119-128. DOI: 10.1016/j.bpobgyn.2016.03.003
10. **Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome**. *Fertil Steril* (2004) **81** 19-25. DOI: 10.1016/j.fertnstert.2003.10.004
11. Kinnear HM, Tomaszewski CE, Chang FL, Moravek MB, Xu M, Padmanabhan V, Shikanov A. **The ovarian stroma as a new frontier**. *Reproduction* (2020) **160** R25-R39. DOI: 10.1530/REP-19-0501
12. Fatima M, Amjad S, SharafAli H, Ahmed T, Khan S, Raza M, Inam M. **Correlation of subclinical hypothyroidism with polycystic ovary syndrome (PCOS)**. *Cureus* (2020) **12** e8142. DOI: 10.7759/cureus.8142
13. Bedaiwy MA, Abdel-Rahman MY, Tan J, AbdelHafez FF, Abdelkareem AO, Henry D, Lisonkova S, Hurd WW, Liu JH. **Clinical, hormonal, and metabolic parameters in women with subclinical hypothyroidism and polycystic ovary syndrome: a cross-sectional study**. *J Womens Health (Larchmt)* (2018) **27** 659-664. DOI: 10.1089/jwh.2017.6584
14. Torun E, Özgen IT, Gökçe S, Aydın S, Cesur Y. **Thyroid hormone levels in obese children and adolescents with non-alcoholic fatty liver disease**. *J Clin Res Pediatr Endocrinol* (2014) **6** 34-39. DOI: 10.4274/Jcrpe.1155
15. Nichols PH, Pan Y, May B, Pavlicova M, Rausch JC, Mencin AA, Thaker VV. **Effect of TSH on Non-Alcoholic Fatty Liver Disease (NAFLD) independent of obesity in children of predominantly Hispanic/Latino ancestry by causal mediation analysis**. *PLoS ONE* (2020) **15** e0234985. DOI: 10.1371/journal.pone.0234985
16. Yang H, Di J, Pan J, Yu R, Teng Y, Cai Z, Deng X. **The association between prolactin and metabolic parameters in PCOS women: a retrospective analysis**. *Front Endocrinol (Lausanne)* (2020) **12** 263. DOI: 10.3389/fendo.2020.00263
17. Jonard S, Dewailly D. **The follicular excess in polycystic ovaries, due to intra-ovarian hyperandrogenism, may be the main culprit for the follicular arrest**. *Hum Reprod Update.* (2004) **10** 107-17. DOI: 10.1093/humupd/dmh010
18. Mu L, Pan J, Yang L, Chen Q, Chen Y, Teng Y, Wang P, Tang R, Huang X, Chen X, Yang H. **Association between the prevalence of hyperuricemia and reproductive hormones in polycystic ovary syndrome**. *Reprod Biol Endocrinol* (2018) **16** 104. DOI: 10.1186/s12958-018-0419-x
19. Hu J, Xu W, Yang H, Mu L. **Uric acid participating in female reproductive disorders: a review**. *Reprod Biol Endocrinol* (2021) **19** 65. DOI: 10.1186/s12958-021-00748-7
20. Sarkar MA, Suzuki A, Abdelmalek MF, Yates KP, Wilson LA, Bass NM, Gill R, Cedars M, Terrault N. **Testosterone is associated with nonalcoholic steatohepatitis and fibrosis in premenopausal women with NAFLD**. *Clin Gastroenterol Hepatol* (2021) **19** 1267-1274.e1. DOI: 10.1016/j.cgh.2020.09.045
21. Cui P, Hu W, Ma T. **Long-term androgen excess induces insulin resistance and non-alcoholic fatty liver disease in PCOS like rats [J]**. *J Steroid Biochem Mol Biol* (2021) **208** 105829. DOI: 10.1016/j.jsbmb.2021.105829
22. Anzai Á, Marcondes RR, Gonçalves TH. **Impaired branched-chain amino acid metabolism may underlie the nonalcoholic fatty liver disease-like pathology of neonatal testosterone-treated female rats [J]**. *Sci Rep* (2017) **7** 13167. DOI: 10.1038/s41598-017-13451-8
23. Gözüküçük M, Gürsoy AY, Destegül E, Taşkın S, Şatıroğlu H. **Homocysteine and C-reactive protein levels in women with polycystic ovary syndrome**. *Gynecol Minim Invasive Ther* (2021) **10** 210-214. DOI: 10.4103/GMIT.GMIT_30_20
24. Kondapaneni V, Gutlapalli SD, Poudel S, Zeb M, Toulassi IA, Cancarevic I. **Significance of homocysteine levels in the management of polycystic ovarian syndrome: a literature review**. *Cureus* (2020) **12** e11110. DOI: 10.7759/cureus.11110
25. Karmin O, Siow YL. **Metabolic Imbalance of homocysteine and hydrogen sulfide in kidney disease**. *Curr Med Chem* (2018) **25** 367-377. DOI: 10.2174/0929867324666170509145240
26. Gulsen M, Yesilova Z, Bagci S, Uygun A, Ozcan A, Ercin CN, Erdil A, Sanisoglu SY, Cakir E, Ates Y, Erbil MK, Karaeren N, Dagalp K. **Elevated plasma homocysteine concentrations as a predictor of steatohepatitis in patients with non-alcoholic fatty liver disease**. *J Gastroenterol Hepatol* (2005) **20** 1448-1455. DOI: 10.1111/j.1440-1746.2005.03891.x
27. Brochado MJF, Domenici FA, Martinelli ADLC, Zucoloto S, Cunha SFDCD, Vannucchi H. **Methylenetetrahydrofolate reductase gene polymorphism and serum homocysteine levels in nonalcoholic fatty liver disease**. *Ann Nutr Metab* (2013) **63** 193-199. DOI: 10.1159/000353139
28. Gulsen M, Yesilova Z, Bagci S, Uygun A, Ozcan A, Ercin CN, Erdil A, Sanisoglu SY, Cakir E, Ates Y. **Elevated plasma homocysteine concentrations as a predictor of steatohepatitis in patients with non-alcoholic fatty liver disease**. *J Gastroenterol Hepatol* (2005) **20** 1448-1455. DOI: 10.1111/j.1440-1746.2005.03891.x
29. Werge MP, McCann A, Galsgaard ED, Holst D, Bugge A, Albrechtsen NJW, Gluud LL. **The role of the transsulfuration pathway in non-alcoholic fatty liver disease**. *J Clin Med* (2021) **10** 1081. DOI: 10.3390/jcm10051081
30. Shi Y, Han T, Cui L, Wu G, Zheng R, Xia M, Chen ZJ. **White blood cell differential counts in patients with polycystic ovary syndrome: a pilot study on Chinese women**. *Eur J Obstet Gynecol Reprod Biol* (2013) **170** 162-164. DOI: 10.1016/j.ejogrb.2013.06.002
31. Womack J, Tien PC, Feldman J. **Obesity and immune cell counts in women**. *Metabolism* (2007) **56** 998-1004. DOI: 10.1016/j.metabol.2007.03.008
32. Dasanu CA, Clark BA, Ichim TE, Alexandrescu DT. **Polycystic ovary syndrome: focus on platelets and prothrombotic risk**. *South Med J* (2011) **104** 174-178. DOI: 10.1097/SMJ.0b013e31820c0172
33. Keskin Kurt R, Okyay AG, Hakverdi AU. **The effect of obesity on inflammatory markers in patients with PCOS: a BMI-matched case-control study**. *Arch Gynecol Obstet* (2014) **290** 315-319. DOI: 10.1007/s00404-014-3199-3
34. Rajendran S, Willoughby SR, Chan WP, Liberts EA, Heresztyn T, Saha M, Marber MS, Norman RJ, Horowitz JD. **Polycystic ovary syndrome is associated with severe platelet and endothelial dysfunction in both obese and lean subjects**. *Atherosclerosis* (2009) **204** 509-514. DOI: 10.1016/j.atherosclerosis.2008.09.010
35. Patel S. **Inflammasomes, the cardinal pathology mediators are activated by pathogens, allergens and mutagens: a critical review with focus on NLRP3**. *Biomed Pharmacother* (2017) **92** 819-825. DOI: 10.1016/j.biopha.2017.05.126
36. Kim W, Flamm SL, Di Bisceglie AM, Bodenheimer HC. **Serum activity of alanine aminotransferase (ALT) as an indicator of health and disease**. *Hepatology* (2008) **47** 1363-1370. DOI: 10.1002/hep.22109
37. Clark JM, Brancati FL, Diehl AM. **The prevalence and etiology of elevated aminotransferase levels in the United States**. *Am J Gastroenterol* (2003) **98** 960-967. DOI: 10.1111/j.1572-0241.2003.07486.x10.14740/jocmr3639
38. Minato S, Sakane N, Kotani K, Nirengi S, Hayashi I, Suganuma A, Yamaguchi K, Takakura K, Nagai N. **Prevalence and risk factors of elevated liver enzymes in Japanese women with polycystic ovary syndrome**. *J Clin Med Res.* (2018) **10** 904-910. DOI: 10.14740/jocmr3639
39. Javed Z, Papageorgiou M, Deshmukh H, Kilpatrick ES, Mann V, Corless L, Abouda G, Rigby AS, Atkin SL, Sathyapalan T. **A randomized, controlled trial of vitamin D supplementation on cardiovascular risk factors, hormones, and liver markers in women with polycystic ovary syndrome**. *Nutrients* (2019) **11** 188. DOI: 10.3390/nu11010188
40. Molla NH, Kathak RR, Sumon AH, Barman Z, Mou AD, Hasan A, Mahmud F, Fariha KA, Ali N. **Assessment of the relationship between serum uric acid levels and liver enzymes activity in Bangladeshi adults**. *Sci Rep* (2021) **11** 20114. DOI: 10.1038/s41598-021-99623-z
41. Edwards NL. **The role of hyperuricemia in vascular disorders**. *Curr Opin Rheumatol* (2009) **21** 132-137. DOI: 10.1097/BOR.0b013e3283257b96
42. Afzali A, Weiss NS, Boyko EJ, Ioannou GN. **Association between serum uric acid level and chronic liver disease in the United States**. *Hepatology* (2010) **52** 578-589. DOI: 10.1002/hep.23717
43. Belan M, Pelletier C, Baillargeon JP. **Alanine aminotransferase is a marker of lipotoxicity consequences and hyperandrogenemia in women with polycystic ovary syndrome**. *Metab Syndr Relat Disord* (2017) **15** 145-152. DOI: 10.1089/met.2016.0119
44. Javed A, Lteif AN, Kumar S, Simmons PS, Chang AY. **Fasting glucose changes in adolescents with polycystic ovary syndrome compared with obese controls: a retrospective cohort study**. *J Pediatr Adolesc Gynecol* (2015) **28** 451-456. DOI: 10.1016/j.jpag.2015.01.001
|
---
title: 'Malnutrition risk and frailty in head and neck cancer patients: coexistent
but distinct conditions'
authors:
- Priya Dewansingh
- Linda Bras
- Lies ter Beek
- Wim P. Krijnen
- Jan L. N. Roodenburg
- Cees P. van der Schans
- Gyorgy B. Halmos
- Harriët Jager-Wittenaar
journal: European Archives of Oto-Rhino-Laryngology
year: 2022
pmcid: PMC9988738
doi: 10.1007/s00405-022-07728-6
license: CC BY 4.0
---
# Malnutrition risk and frailty in head and neck cancer patients: coexistent but distinct conditions
## Abstract
### Purpose
Both malnutrition and frailty are associated with adverse treatment outcomes. Malnutrition (risk) and frailty are each commonly present in patients with head and neck cancer (HNC). However, their coexistence and association is unknown. Main goal of this study is to determine the coexistence of, and the association between malnutrition risk and frailty in patients with HNC.
### Methods
In this retrospective analysis on prospectively collected data, newly diagnosed patients with HNC, enrolled in the OncoLifeS databiobank were included. The Patient-Generated Subjective Global Assessment Short Form (PG-SGA SF) was used to assess malnutrition risk. The Groningen Frailty Indicator (GFI) was used to assess frailty status. Multivariate logistic regression analyses were performed, taking into account several patient- and tumor-related factors.
### Results
In total, 197 patients were included. Seventy-six patients ($39\%$) had a medium or high malnutrition risk and 71 patients ($36\%$) were frail. In 38 patients ($19\%$), malnutrition risk coexisted with frailty. Patients with medium and high malnutrition risk were, respectively, 4.0 ($95\%$ CI 1.5–11.2) and 13.4 ($95\%$ CI 4.0–48.7) times more likely to be frail, compared to patients with low malnutrition risk. In turn, frail patients were 6.4 times ($95\%$ CI 2.6–14.9) more likely to have malnutrition risk compared to non-frail patients.
### Conclusions
Malnutrition risk and frailty frequently coexist but not fully overlap in newly diagnosed patients with HNC. Therefore, screening for both conditions is recommended.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00405-022-07728-6.
## Introduction
Malnutrition and frailty are serious health conditions, each commonly present in patients diagnosed with head and neck cancer (HNC) [1–3]. Both conditions are associated with adverse treatment outcomes, such as radiation-induced toxicity, postoperative complications, mortality, and poorer quality of life [4–8]. A combination of patient- and tumor-related factors typically associated with HNC is responsible for the high prevalence of malnutrition and frailty. For example, swallowing problems and pain in the upper aero-digestive tract as consequence of the tumor localization often lead to insufficient oral intake, unintentional weight loss and sarcopenia, especially in patients with mucosal HNC [1, 9–11]. Furthermore, tobacco use and alcohol use are the main risk factors for the development of HNC, but also cause comorbidities, like COPD and liver cirrhosis in these patients [12]. Each of these factors are associated with frailty.
In several patient populations, coexistence and interaction between malnutrition and frailty has been demonstrated, indicating that these conditions share common physical, social, and psychological risk factors [11, 13, 14]. Frailty has been defined as ‘a dynamic state affecting an individual who experiences losses in one or more domains of human functioning (physical, psychological, and social), which is caused by the influence of a range of variables and which increases the risk of adverse outcomes’ [15]. Frail individuals have an increased risk of losing independency in daily activities and an increased risk of mortality. Frailty can be potentially preventable and/or treatable, for example, by nutritional and physical activity interventions [16].
In The Netherlands, screening for malnutrition risk and frailty is routine in the hospital setting, to identify patients who may benefit from nutritional assessment and intervention before and during treatment [17]. To further improve the care pathways around the patients with HNC with multidimensional problems, we aimed to determine the coexistence of, and association between malnutrition risk and frailty.
## Study design and ethics
Since 2014, all newly diagnosed patients with HNC in the multidisciplinary head and neck oncological team of the University Medical Center Groningen (UMCG) were included in OncoLifeS, i.e., a data biobank which has been approved by the Medical Ethical Committee (UMCG METC approval $\frac{2010}{109}$) and complies with the General Data Protection Regulation as stated by the European Union. Patients were enrolled after providing written informed consent. Data were selected from the OncolifeS data biobank, from patients included between August 2015 and January 2017. The study protocol was approved by the Scientific Board of OncoLifeS. The study is conducted in accordance with the current version of the Declaration of Helsinki.
## Study population and data collection
Two groups of patients were included. The first group consisted of patients with mucosal tumors in the upper digestive tract-like oral cavity, oropharyngeal, hypopharyngeal, and laryngeal malignancies. The second group consisted of patients with skin malignancies in the head and neck area. These two groups were distinguished, because the tumor often interferes with food intake in the group with mucosal tumors. During the diagnostic work-up, every patient filled out a set of questionnaires regarding demographic characteristics, comorbidities, smoking and alcohol use, socio-economic factors, cognitive functioning, frailty, and nutritional risk with the assistance of a healthcare professional.
## Variables
Malignancies were staged using the seventh edition of the TNM Classification of Malignant Tumors from the Union for International Cancer Control. Tumor type was categorized as follows: [1] mucosal, i.e., patients with mucosal squamous cell carcinoma of the oral cavity, oropharynx, and hypopharynx, larynx and [2] cutaneous, i.e., patients with a malignancy of the skin in the head and neck area. Comorbidities were assessed using the Adult Comorbidity Evaluation (ACE)-27, which categorizes patients with none, mild, moderate, or severe comorbidities based on 27 predefined items [18]. Cognitive functioning was assessed with the Mini-Mental State Examination (MMSE), in which cognitive impairment was defined by a score ≤ 24 [19].
## Malnutrition risk and frailty
The Dutch version of the Patient-Generated Subjective Global Assessment Short Form (PG-SGA SF) (version 3.7) was used to screen for malnutrition risk [20]. The PG-SGA SF is the patient component of the full PG-SGA, which is the reference method for nutritional assessment in patients with cancer [21]. The PG-SGA SF includes four boxes addressing weight history (Box 1), food intake (Box 2), nutrition impact symptoms (NIS), i.e., symptoms interfering with oral intake (Box 3), and activities and function (Box 4). PG-SGA SF total score ranges from 0 to 36. Patients with a score ≤ 3 were defined as low, ≥ 4 and ≤ 8 as medium, and ≥ 9 as high malnutrition risk [21, 22]. Patients with medium and high malnutrition risk were pooled and classified as ‘malnutrition risk’ for statistical analyses.
Frailty was assessed by the Groningen frailty indicator (GFI). The GFI consists of 15 questions regarding the following domains of life: daily activities, health problems, and psychosocial functioning, generating a score ranging from 0 to 15. Frailty was defined as a GFI score ≥ 4 [23].
## Statistical analysis
Continuous variables are presented as mean ± standard deviation (SD) for normally distributed variables, and as median with interquartile range (IQR) for skewed or ordinal variables. Normality was tested by the Kolmogorov–Smirnov test. The exact binomial Clopper-–Pearson estimation method was used for prevalence numbers and their $95\%$ confidence interval ($95\%$ CI) of frail patients to have malnutrition risk, and patients with malnutrition risk to be frail. Multivariate data imputation was performed for missing data on variables to detect any meaningful differences with the results obtained after casewise deletion [24]. Univariate and multivariate binary logistic regression analyses were used to determine associations between malnutrition risk and frailty. Binary logistic regression analysis was performed separately using malnutrition risk or frailty as dependent outcome variable, respectively. For multivariate logistic regression analyses, the minimum Akaike Information Criterion (min AIC) was used to select and compare models that best predict new outcomes to determine the regression method that would most appropriately model the association between the outcome and the explanatory variables [25, 26]. Two tailed p-values were used with significance set at $p \leq 0.05.$ Associations were presented as odds ratios (ORs) with $95\%$ CIs. Statistical analyses were performed using IBM SPSS version 23.0 (SPSS Inc., Chicago, IL, USA). The Venn diagram, min AIC, and the multivariate logistic regression were produced using R Studio version 1.2.5019.
## Results
In total, 197 patients were included. Table 1 shows baseline characteristics of the patients. The mean age was 70.5 ± 11.5 years. The majority ($68\%$) of patients was male. In total, 54 ($27\%$) patients had medium malnutrition risk, 22 ($11\%$) had high malnutrition risk, and 71 ($36\%$) were frail. Table 1HNC study sample characteristics across malnutrition risk categories by PG-SGA SFNTotal groupN = 197Risk categories by PG-SGA SFLow risk0–3 points121 [61]Medium risk4–8 points54 [27]High risk ≥ 9 points22 [11]Age, mean ± SD19770.5 ± 11.571.3 ± 11.571.7 ± 11.363.1 ± 9.7Gender Male19713482 [61]39 [29]13 [10] Female6339 [62]15 [24]9 [14]BMI, mean ± SD19425.9 ± 4.726.0 ± 4.126.5 ± 5.423.6 ± 5.5Tumor type197 Mucosal12060 [50]40 [33]20 [17] Cutaneous7761 [79]14 [18]2 [3]Tumor localization197 Oral cavity2512 [48]9 [36]4 [16] Oropharynx4112 [29]18 [44]11 [27] Hypopharynxa83 [38]4 [50]1 [13] Supraglottic larynx159 [60]4 [27]2 [13] (Sub)glottic larynx3124 [77]5 [16]2 [7]Skin7761 [79]14 [18]2 [3]Classification for mucosal HNC118 Tis/T1/T2a6440 [63]19 [30]5 [8] T3/T4/Tx5419 [35]21 [39]14 [26]Classification for cutaneous HNC74 Tis/T1/T2a5442 [78]10 [19]2 [4] T3/T4/Tx2016 [80]4 [20]0 [0]Smoking197 Currently smoking6126 [43]21 [34]14 [23] Never smoked/smoked in past13695 [70]33 [24]8 [6]Alcohol, units/day, median (IQR)1741 (0–3)1 (0–2)1 (0–4)2 (0–4)Education196 Lower9253 [58]29 [32]10 [11] Middlea5331 [58]16 [30]6 [11] Higher4532 [71]8 [18]5 [11] Other/unknowna64 [67]1 [17]1 [17]Marital status197 Single/widowed/divorced6640 [61]19 [29]7 [11] Married/living together/not singlea13181 [62]35 [27]15 [11]Comorbidityb181 None/mild9665 [68]21 [22]10 [10] Moderate/severe8542 [49]31 [36]12 [14]Cognitionc, median score (IQR)19728 (25–29)28 (26–29)27 (25–30)29 (25–30)Normal cognition15898 [62]42 [27]18 [11]Impaired cognition3923 [59]12 [31]4 [10]Frailtyd197 Frail7133 [46]25 [35]13 [18] Non-frail12688 [70]29 [23]9 [7]Numbers are shown as n (%) unless reported otherwiseSD standard deviation, IQR interquartile range, PG-SGA SF Patient-Generated Subjective Global Assessment Short Form, BMI body mass indexaPercentages does not sum to 100, due to roundingbAdult Comorbidity Evaluation 27cMini-Mental State ExaminationdGroningen Frailty Indicator
## Coexistence of malnutrition risk and frailty
Figure 1a–c shows proportional Venn diagrams of the coexistence of malnutrition risk and frailty. In total, 109 ($55\%$) patients had malnutrition risk and/or were frail. Coexistence was present in 38 ($19\%$) patients, while 38 ($19\%$) patients had only malnutrition risk, and 33 ($17\%$) patients were only frail. Almost half of the patients ($$n = 88$$, $45\%$) neither had malnutrition risk nor were frail. Fig. 1a–c Proportional Venn diagram of the coexistence of malnutrition risk and frailty in patients with head and neck cancer. N (%), 1Percentages do not sum to 100, due to rounding Coexistence of malnutrition risk and frailty was found in 25 ($21\%$) patients with mucosal HNC, and in 13 ($17\%$) patients with cutaneous HNC. Solely malnutrition risk was present in almost one-third of patients with mucosal HNC ($$n = 35$$, $29\%$), and only in 3 ($4\%$) patients with cutaneous HNC. Solely frailty was more often present in patients with cutaneous HNC ($$n = 19$$, $25\%$) compared to patients with mucosal HNC ($$n = 14$$, $12\%$). Moreover, absence of both malnutrition risk and frailty was more often present in patients with cutaneous HNC ($$n = 42$$, $55\%$) compared to patients with mucosal HNC ($$n = 46$$, $38\%$).
The exact binomial test with exact Clopper–Pearson $95\%$ CI showed a prevalence of having malnutrition risk of $39\%$ ($95\%$ CI 32–$46\%$, $$p \leq 0.002$$) in frail patients. The prevalence of being frail was $36\%$ ($95\%$ CI 29–$43\%$, $p \leq 0.001$) in patients with malnutrition risk.
## Univariate and multivariate analyses of factors associated with malnutrition risk and/or frailty
Results of univariate analyses are presented in Tables 2 and 3. Age was significantly associated with frailty, but not with malnutrition risk. Patients with moderate to severe comorbidities significantly more frequently had malnutrition risk and frailty. Having a partner decreased the odds of being frail. Alcohol use was associated with higher odds of malnutrition risk, but with decreased odds of being frail. Patients with mucosal HNC were—not significantly—less often frail compared to patients with cutaneous HNC. However, patients with mucosal HNC had 4.8 times more often medium or high malnutrition risk compared to patients with cutaneous HNC.Table 2Univariate and multivariate modeling analyses of variables associated with frailty in head and neck cancer patients, $$n = 159$$FrailtyUnivariate modelOR [$95\%$ CI]p-valueMultivariate model AIC OR [$95\%$ CI]p-valueAgea1.06 [1.03–1.09]0.00061.07 [1.03–1.12] < 0.001Smoking, yes1.13 [0.56–2.23]0.7309Sex, female1.73 [0.85–3.52]0.1280Comorbidity Moderate/severe2.26 [1.17–4.43]0.0161Marital status Married, living together, not single0.31 [0.15–0.62]0.00100.14 [0.05–0.35] < 0.001Education level Middle0.44 [0.18–1.03]0.0644 High2.07 [0.32–16.47]0.4416 Unknown0.73 [0.45–1.16]0.1873 Alcohola0.78 [0.64–0.93]0.01080.72 [0.58–0.87]0.002CognitionC ognitive impairment0.15 [0.06–0.38]0.00010.18 [0.05–0.60]0.007Tumor type Mucosal0.64 [0.32–1.26]0.194Medium malnutrition risk2.35 [1.10–5.04]0.02783.99 [1.50–11.20]0.007High malnutrition risk3.79 [1.44–10.34]0.007613.44 [4.04–48.71] < 0.001Bold indicates significant variablesAIC Akaike Information Criterion, OR odds ratio, CI confidence intervalaContinuous variableTable 3Univariate and multivariate modeling analyses of variables associated with malnutrition in head and neck cancer patients, $$n = 172$$MalnutritionUnivariate modelOR [$95\%$ CI]p-valueMultivariate model AIC OR [$95\%$ CI]p-valueAgea0.98 [0.95–1.00]0.0931Smoking, yes3.29 [1.70–6.48]0.00052.08 [0.95–4.61]0.068Sex, female0.82 [0.41–1.62]0.577Comorbidity Moderate/severe2.37 [1.25–4.58]0.0092Marital status Married, living together, not single0.93 [0.50–1.75]0.8452.10 [0.93–4.99]0.084Education level Middle0.59 [0.26–1.30]0.1960 High0.73 [0.10–4.00]0.7304 Unknown0.68 [0.43–1.06]0.0935Alcohola1.15 [1.01–1.33]0.04571.21 [1.02–1.45]0.029Cognition Cognitive impairment1.01 [0.46–2.23]0.981Tumor type Mucosal4.82 [2.34–10.63] < 0.0014.23 [1.79–10.77]0.002Frailty, yes2.75 [1.41–5.44]0.00326.04 [2.63–14.86] < 0.001Bold indicates significant variablesAIC Akaike Information Criterion, OR odds ratio, CI confidence intervalaContinuous variable Table 2 demonstrates that after correction for age, alcohol use, marital status, and cognition, patients with medium and high malnutrition risk were 4.0 and 13.4 times more likely to be frail compared to patients with low malnutrition risk, respectively. Vice versa, Table 3 shows that frail patients were 6.0 times more likely at risk of malnutrition compared to non-frail patients, after correction for smoking status, alcohol use, marital status, and tumor type.
## PG-SGA SF outcomes
The Supplementary file 1 shows scores on the PG-SGA SF for the study population and per frailty status.
## Discussion
This study shows that malnutrition risk and frailty considerably coexist in patients with newly diagnosed HNC. The prevalence of malnutrition risk or frailty alone is comparable to the prevalence of coexistence of malnutrition risk and frailty, i.e., $19\%$ and $17\%$, respectively. Malnutrition risk is strongly positively associated with being frail. Medium and high malnutrition risk is related to 4.0 and 13.4 times more chance of being frail, respectively. In turn, frail patients are 6.0 times more likely to have medium or high malnutrition risk.
This is the first study investigating the coexistence and association between malnutrition risk and frailty in patients with HNC. In populations of older adults, the coexistence varies between 8 and $33\%$ [13, 27–29]. The coexistence of almost $20\%$ in our study is within this range. In line with previous findings in other populations, prevalence of both conditions separately was also considerable in our population [13]. Moreover, our results are in line with previous findings showing that older adults with malnutrition risk have a higher risk of being frail, and vice versa [30, 31]. Unfortunately, comparable studies in HNC populations are not available. Furthermore, comparison of our results with previous research is hampered due to use of different instruments for assessment of malnutrition risk and frailty.
The prevalence of frailty in our HNC study population is comparable with previous findings. Reported prevalence of frailty in HNC patients largely varies, i.e., between 7 and $75\%$ [32–35], possibly depending on the methods used to determine frailty. Lowest percentages were found in retrospective population-based studies on hospitals’ discharge data, while highest percentages were found in studies using prospective multidimensional frailty instruments. Previous results from a comparable cohort of HNC patients showed a frailty prevalence of $40\%$ [36]. Unfortunately, comparison of our findings on prevalence of malnutrition risk with previous studies is hampered, as previous studies in patients with HNC assessed malnutrition by the full PG-SGA rather than malnutrition risk by the PG-SGA SF. In those studies, prevalence of malnutrition varied between 31 and $44\%$ [10, 37–39]. The association between malnutrition risk and frailty in patients with mucosal HNC is less strong compared to the association between malnutrition risk and frailty in our total study population. Despite comparable frailty prevalence in the total study population and the population with mucosal HNC, the latter showed a higher prevalence of malnutrition risk. It is likely that the prominently present swallowing problems in patients with mucosal HNC due to the tumor localization [40] more often result in malnutrition risk, independent of frailty.
The current study shows that alcohol consumption is associated with greater risk of developing malnutrition risk. However, this association is not shown for frailty, in which alcohol consumption even seems protective for being frail. This protective association between alcohol consumption and frailty was also found in a systematic review and meta-analysis [41]. Although the underlying mechanisms for a lower risk of frailty among alcohol consumers compared to non/past drinkers is not clear, it is possible that individuals who consume alcohol also have a stronger social network and stronger social support, which can prevent social isolation and therefore frailty [42].
A limitation of this study is the relatively small sample size, both for the whole study population and for the two subgroups, i.e., patients with mucosal tumors and patients with cutaneous tumors. As result, it was not possible to perform multivariate analyses per patient group. We countered this limitation by including the tumor type in the multivariate analysis.
Based on the current study, several clinical implications and recommendations can be formulated. First, malnutrition risk and frailty both need to be proactively screened for in patients with HNC, since both conditions not only coexist, but also separately occur in these patients. Although patients with mucosal HNC show the highest prevalence of malnutrition risk and frailty, screening for malnutrition risk in patients with complex cutaneous HNC is also relevant, since still one out of five of these patients has medium-to-high malnutrition risk. Screening for both conditions will identify different types of health-related problems per individual and may guide starting different interventions, e.g., nutritional interventions for patients with malnutrition or psychosocial support for frail patients, to optimize the patient’s pretreatment condition. Patients who remain malnourished and/or frail during and after cancer treatment are at risk of body tissue catabolism and wound healing disorders. These adverse advents can lead to a non-optimal treatment, making the already burdensome oncology treatment even harder for patients and can also lead to decreased overall survival [6, 40, 43]. Furthermore, previous studies in the same patient cohort as in the current study showed that medium malnutrition risk and frailty in patients undergoing surgery were both associated with postoperative complications [8]. Pretreatment medium-to-high malnutrition risk and frailty and were also associated with a decline in post-treatment quality of life [7, 44]. These findings also highlight the importance to screen patients with HNC for both conditions. Second, we recommend to screen for malnutrition risk and frailty to create awareness amongst healthcare professionals and patients [45] for potentially treatable factors. Frailty is a dynamic concept and the process of frailty can possibly be reversed [16]. Previous research has shown that nutritional status is prone to further deterioration during HNC treatment [46]. However, more research is needed to gain insight in the development of frailty during the course from diagnosis to rehabilitation and on the effect of specific supportive treatment that might lead to possibly reversing the patients’ frailty status.
## Conclusion
This study demonstrates considerable coexistence and an association between malnutrition risk and frailty in newly diagnosed patients with HNC, but also shows that both conditions considerably occur separately in these patients. Our findings highlight the importance of screening for both conditions in these patients at diagnosis. To potentially reverse malnutrition risk and frailty, targeted interventions are required.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 16 KB)
## References
1. Jager-Wittenaar H, Dijkstra PU, Vissink A, van der Laan BFAM, van Oort RP, Roodenburg JLN. **Critical weight loss in head and neck cancer–prevalence and risk factors at diagnosis: an explorative study**. *Support Care Cancer* (2007.0) **15** 1045-1050. DOI: 10.1007/s00520-006-0212-9
2. Alshadwi A, Nadershah M, Carlson ER, Young LS, Burke PA, Daley BJ. **Nutritional considerations for head and neck cancer patients: a review of the literature**. *J Oral Maxillofac Surg* (2013.0) **71** 1853-1860. DOI: 10.1016/j.joms.2013.04.028
3. Bras L, Driessen DAJJ, de Vries J, Festen S, van der Laan BFAM, van Leeuwen BL. **Patients with head and neck cancer: Are they frailer than patients with other solid malignancies?**. *Eur J Cancer Care (Engl)* (2020.0) **29** e13170. DOI: 10.1111/ecc.13170
4. Kono T, Sakamoto K, Shinden S, Ogawa K. **Pre-therapeutic nutritional assessment for predicting severe adverse events in patients with head and neck cancer treated by radiotherapy**. *Clin Nutr* (2017.0) **36** 1681-1685. DOI: 10.1016/j.clnu.2016.10.021
5. van Deudekom FJ, van der Velden L-A, Zijl WH, Schimberg AS, Langeveld AP, Slingerland M. **Geriatric assessment and 1-year mortality in older patients with cancer in the head and neck region: A cohort study**. *Head Neck* (2019.0) **41** 2477-2483. DOI: 10.1002/hed.25714
6. Jager-Wittenaar H, Dijkstra PU, Vissink A, van der Laan BFAM, van Oort RP, Roodenburg JLN. **Malnutrition and quality of life in patients treated for oral or oropharyngeal cancer**. *Head Neck* (2011.0) **33** 490-496. DOI: 10.1002/hed.21473
7. de Vries J, Bras L, Sidorenkov G, Festen S, Steenbakkers RJHM, Langendijk JA. **Association of deficits identified by geriatric assessment with deterioration of health-related quality of life in patients treated for head and neck cancer**. *JAMA Otolaryngol Head Neck Surg* (2021.0) **147** 1089-1099. DOI: 10.1001/jamaoto.2021.2837
8. Bras L, de Vries J, Festen S, Steenbakkers RJHM, Langendijk JA, Witjes MJH. **Frailty and restrictions in geriatric domains are associated with surgical complications but not with radiation-induced acute toxicity in head and neck cancer patients: a prospective study**. *Oral Oncol* (2021.0) **118** 105329. DOI: 10.1016/j.oraloncology.2021.105329
9. Righini C-A, Timi N, Junet P, Bertolo A, Reyt E, Atallah I. **Assessment of nutritional status at the time of diagnosis in patients treated for head and neck cancer**. *Eur Ann Otorhinolaryngol Head Neck Dis* (2013.0) **130** 8-14. DOI: 10.1016/j.anorl.2012.10.001
10. Kubrak C, Olson K, Jha N, Jensen L, McCargar L, Seikaly H. **Nutrition impact symptoms: key determinants of reduced dietary intake, weight loss, and reduced functional capacity of patients with head and neck cancer before treatment**. *Head Neck* (2010.0) **32** 290-300. DOI: 10.1002/hed.21174
11. Laur Cv, McNicholl T, Valaitis R, Keller HH. **Malnutrition or frailty? Overlap and evidence gaps in the diagnosis and treatment of frailty and malnutrition**. *Appl Physiol Nutr Metab* (2017.0) **42** 449-458. DOI: 10.1139/apnm-2016-0652
12. Paleri V, Wight RG, Silver CE, Haigentz M, Takes RP, Bradley PJ. **Comorbidity in head and neck cancer: a critical appraisal and recommendations for practice**. *Oral Oncol* (2010.0) **46** 712-719. DOI: 10.1016/j.oraloncology.2010.07.008
13. Boulos C, Salameh P, Barberger-Gateau P. **Malnutrition and frailty in community dwelling older adults living in a rural setting**. *Clin Nutr* (2016.0) **35** 138-143. DOI: 10.1016/j.clnu.2015.01.008
14. Ligthart-Melis GC, Luiking YC, Kakourou A, Cederholm T, Maier AB, de van der Schueren MAE. **Frailty, sarcopenia, and malnutrition frequently (Co-)occur in hospitalized older adults: a systematic review and meta-analysis**. *J Am Med Dir Assoc* (2020.0) **21** 1216-1228. DOI: 10.1016/j.jamda.2020.03.006
15. Gobbens RJJ, Luijkx KG, Wijnen-Sponselee MT, Schols JMGA. **Towards an integral conceptual model of frailty**. *J Nutr Health Aging* (2010.0) **14** 175-181. DOI: 10.1007/s12603-010-0045-6
16. Marcucci M, Damanti S, Germini F, Apostolo J, Bobrowicz-Campos E, Gwyther H. **Interventions to prevent, delay or reverse frailty in older people: a journey towards clinical guidelines**. *BMC Med* (2019.0) **17** 193. DOI: 10.1186/s12916-019-1434-2
17. 17.Nederlands Vereniging van Ziekenhuizen NF van UMC. VMS. https://www.vmszorg.nl/wp-content/uploads/2017/11/web_2009.0104_praktijkgids_kwetsbare_ouderen.pdf. Accessed 18 May 2022
18. Piccirillo JF, Tierney RM, Costas I, Grove L, Spitznagel EL. **Prognostic importance of comorbidity in a hospital-based cancer registry**. *JAMA* (2004.0) **291** 2441-2447. DOI: 10.1001/jama.291.20.2441
19. van der Cammen TJ, van Harskamp F, Stronks DL, Passchier J, Schudel WJ. **Value of the Mini-Mental State Examination and informants’ data for the detection of dementia in geriatric outpatients**. *Psychol Rep* (1992.0) **71** 1003-1009. DOI: 10.2466/pr0.1992.71.3.1003
20. Sealy MJ, Haß U, Ottery FD, van der Schans CP, Roodenburg JLN, Jager-Wittenaar H. **Translation and cultural adaptation of the scored patient-generated subjective global assessment: an Interdisciplinary Nutritional Instrument Appropriate for Dutch Cancer Patients**. *Cancer Nurs* (2018.0) **41** 450-462. DOI: 10.1097/NCC.0000000000000505
21. Jager-Wittenaar H, Ottery FD. **Assessing nutritional status in cancer: role of the Patient-Generated Subjective Global Assessment**. *Curr Opin Clin Nutr Metab Care* (2017.0) **20** 322-329. DOI: 10.1097/MCO.0000000000000389
22. ter Beek L, Banning LBD, Visser L, Roodenburg JLN, Krijnen WP, van der Schans CP. **Risk for malnutrition in patients prior to vascular surgery**. *Am J Surg* (2018.0) **216** 534-539. DOI: 10.1016/j.amjsurg.2017.11.038
23. Schuurmans H, Steverink N, Lindenberg S, Frieswijk N, Slaets JPJ. **Old or frail: what tells us more?**. *J Gerontol A Biol Sci Med Sci* (2004.0) **59** M962-M965. DOI: 10.1093/gerona/59.9.m962
24. Hettiarachchi J, Reijnierse EM, Soh CH, Agius B, Fetterplace K, Lim WK. **Malnutrition is associated with poor trajectories of activities of daily living in geriatric rehabilitation inpatients: RESORT**. *Mech Ageing Dev* (2021.0) **197** 111500. DOI: 10.1016/j.mad.2021.111500
25. Akaike H. **A new look at the statistical model identification**. *IEEE Trans Autom Control* (1974.0) **19** 716-723. DOI: 10.1109/TAC.1974.1100705
26. Konishi S, Kitagawa G. *Information criteria and statistical modeling* (2008.0)
27. Gingrich A, Volkert D, Kiesswetter E, Thomanek M, Bach S, Sieber CC. **Prevalence and overlap of sarcopenia, frailty, cachexia and malnutrition in older medical inpatients**. *BMC Geriatr* (2019.0) **19** 120. DOI: 10.1186/s12877-019-1115-1
28. ter Beek L, van der Vaart H, Wempe JB, Krijnen WP, Roodenburg JLN, van der Schans CP. **Coexistence of malnutrition, frailty, physical frailty and disability in patients with COPD starting a pulmonary rehabilitation program**. *Clin Nutr* (2020.0) **39** 2557-2563. DOI: 10.1016/j.clnu.2019.11.016
29. Dorner TE, Luger E, Tschinderle J, Stein Kv, Haider S, Kapan A. **Association between nutritional status (MNA®-SF) and frailty (SHARE-FI) in acute hospitalised elderly patients**. *J Nutr Health Aging* (2014.0) **18** 264-269. DOI: 10.1007/s12603-013-0406-z
30. Chye L, Wei K, Nyunt MSZ, Gao Q, Wee SL, Ng TP. **Strong relationship between malnutrition and cognitive frailty in the Singapore Longitudinal Ageing Studies (SLAS-1 and SLAS-2)**. *J Prev Alzheimers Dis* (2018.0) **5** 142-148. DOI: 10.14283/jpad.2017.46
31. Kim J, Lee Y, Won CW, Lee KE, Chon D. **Nutritional status and frailty in community-dwelling older korean adults: the Korean Frailty and Aging Cohort Study**. *J Nutr Health Aging* (2018.0) **22** 774-778. DOI: 10.1007/s12603-018-1005-9
32. Nieman CL, Pitman KT, Tufaro AP, Eisele DW, Frick KD, Gourin CG. **The effect of frailty on short-term outcomes after head and neck cancer surgery**. *Laryngoscope* (2018.0) **128** 102-110. DOI: 10.1002/lary.26735
33. Pottel L, Lycke M, Boterberg T, Pottel H, Goethals L, Duprez F. **Serial comprehensive geriatric assessment in elderly head and neck cancer patients undergoing curative radiotherapy identifies evolution of multidimensional health problems and is indicative of quality of life**. *Eur J Cancer Care (Engl)* (2014.0) **23** 401-412. DOI: 10.1111/ecc.12179
34. Pitts KD, Arteaga AA, Stevens BP, White WC, Su D, Spankovich C. **Frailty as a predictor of postoperative outcomes among patients with head and neck cancer**. *Otolaryngol Head Neck Surg* (2019.0) **160** 664-671. DOI: 10.1177/0194599818825466
35. Kwon M, Kim S-A, Roh J-L, Lee S-W, Kim S-B, Choi S-H. **An introduction to a head and neck cancer-specific frailty index and its clinical implications in elderly patients: a prospective observational study focusing on respiratory and swallowing functions**. *Oncologist* (2016.0) **21** 1091-1098. DOI: 10.1634/theoncologist.2016-0008
36. Bras L, Peters TTA, Wedman J, Plaat BEC, Witjes MJH, van Leeuwen BL. **Predictive value of the Groningen Frailty Indicator for treatment outcomes in elderly patients after head and neck, or skin cancer surgery in a retrospective cohort**. *Clin Otolaryngol* (2015.0) **40** 474-482. DOI: 10.1111/coa.12409
37. Capuano G, Gentile PC, Bianciardi F, Tosti M, Palladino A, di Palma M. **Prevalence and influence of malnutrition on quality of life and performance status in patients with locally advanced head and neck cancer before treatment**. *Support Care Cancer* (2010.0) **18** 433-437. DOI: 10.1007/s00520-009-0681-8
38. Isenring E, Bauer J, Capra S. **The scored Patient-generated Subjective Global Assessment (PG-SGA) and its association with quality of life in ambulatory patients receiving radiotherapy**. *Eur J Clin Nutr* (2003.0) **57** 305-309. DOI: 10.1038/sj.ejcn.1601552
39. Arribas L, Hurtós L, Milà R, Fort E, Peiró I. **Predict factors associated with malnutrition from patient generated subjective global assessment (PG-SGA) in head and neck cancer patients**. *Nutr Hosp* (2013.0) **28** 155-163. DOI: 10.3305/nh.2013.28.1.6168
40. Jager-Wittenaar H, Dijkstra PU, Vissink A, van Oort RP, van der Laan BFAM, Roodenburg JLN. **Malnutrition in patients treated for oral or oropharyngeal cancer–prevalence and relationship with oral symptoms: an explorative study**. *Support Care Cancer* (2011.0) **19** 1675-1683. DOI: 10.1007/s00520-010-1001-z
41. Kojima G, Liljas A, Iliffe S, Jivraj S, Walters K. **A systematic review and meta-analysis of prospective associations between alcohol consumption and incident frailty**. *Age Ageing* (2018.0) **47** 26-34. DOI: 10.1093/ageing/afx086
42. Sayette MA, Creswell KG, Dimoff JD, Fairbairn CE, Cohn JF, Heckman BW. **Alcohol and group formation: a multimodal investigation of the effects of alcohol on emotion and social bonding**. *Psychol Sci* (2012.0) **23** 869-878. DOI: 10.1177/0956797611435134
43. Noor A, Gibb C, Boase S, Hodge J-C, Krishnan S, Foreman A. **Frailty in geriatric head and neck cancer: a contemporary review**. *Laryngoscope* (2018.0) **128** E416-E424. DOI: 10.1002/lary.27339
44. de Vries J, Bras L, Sidorenkov G, Festen S, Steenbakkers RJHM, Langendijk JA. **Frailty is associated with decline in health-related quality of life of patients treated for head and neck cancer**. *Oral Oncol* (2020.0) **111** 105020. DOI: 10.1016/j.oraloncology.2020.105020
45. Jager-Wittenaar H, de Bats HF, Welink-Lamberts BJ, Gort-van Dijk D, van der Laan BFAM, Ottery FD. **Self-completion of the patient-generated subjective global assessment short form is feasible and is associated with increased awareness on malnutrition risk in patients with head and neck cancer**. *Nutr Clin Pract* (2020.0) **35** 353-362. DOI: 10.1002/ncp.10313
46. Jager-Wittenaar H, Dijkstra PU, Vissink A, Langendijk JA, van der Laan BFAM, Pruim J. **Changes in nutritional status and dietary intake during and after head and neck cancer treatment**. *Head Neck* (2011.0) **33** 863-870. DOI: 10.1002/hed.21546
|
---
title: Does conduction heterogeneity determine the supervulnerable period after atrial
fibrillation?
authors:
- Annejet Heida
- Willemijn F. B. van der Does
- Mathijs S. van Schie
- Lianne N. van Staveren
- Yannick J. H. J. Taverne
- Ad J. J. C. Bogers
- Natasja M. S. de Groot
journal: Medical & Biological Engineering & Computing
year: 2022
pmcid: PMC9988743
doi: 10.1007/s11517-022-02679-w
license: CC BY 4.0
---
# Does conduction heterogeneity determine the supervulnerable period after atrial fibrillation?
## Drs. Annejet Heida
PhD student at the department of Translational Electrophysiology at the Erasmus Medical Center. General practitioner in training.
## Drs. Willemijn F.B. van der Does
PhD student at the department of Translational Electrophysiology at the Erasmus Medical Center. Medical doctor at municipal health service.
## Drs. Mathijs S. Van Schie
PhD student at the department of Translational Electrophysiology at the Erasmus Medical Center. Clinical Technologist.
## Drs. Lianne N. Van Staveren
PhD student at the department of Translational Electrophysiology at the Erasmus Medical Center. Medical doctor.
## Dr. Yannick Y.J.H.J. Taverne
Assistant professor at the department of Cardiothoracic Surgery at the Erasmus Medical Center. Head of department of Translational Cardiothoracic surgery. Thoracic surgeon.
## Prof. Dr. Ad J.J.C. Bogers
Former Head of department of Cardiothoracic Surgery. Thoracic Surgeon.
## Prof. Dr. Natasja M.S. de Groot
Head of department of Translational Electrophysology. Cardiologist-Electrophysiologist.
## Abstract
Atrial fibrillation (AF) resumes within 90 s in $27\%$ of patients after sinus rhythm (SR) restoration. The aim of this study is to compare conduction heterogeneity during the supervulnerable period immediately after electrical cardioversion (ECV) with long-term SR in patients with AF. Epicardial mapping of both atria was performed during SR and premature atrial extrasystoles in patients in the ECV ($$n = 17$$, age: 73 ± 7 years) and control group ($$n = 17$$, age: 71 ± 6 years). Inter-electrode conduction times were used to identify areas of conduction delay (CD) (conduction times 7–11 ms) and conduction block (CB) (conduction times ≥ 12 ms). For all atrial regions, prevalences and length of longest CB and continuous CDCB lines, magnitude of conduction disorders, conduction velocity, biatrial activation time, and voltages did not differ between the ECV and control group during both SR and premature atrial extrasystoles (p ≥ 0.05). Hence, our data suggest that there may be no difference in biatrial conduction characteristics between the supervulnerable period after ECV and long-term SR in AF patients.
### Graphical abstract
The supervulnerable period after AF termination is not determined by conduction heterogeneity during SR and PACs. It is unknown to what extent intra-atrial conduction is impaired during the supervulnerable period immediately after ECV and whether different right and left atrial regions are equally affected. This high-resolution epicardial mapping study (upper left panel) of both atria shows that during SR the prevalences and length of longest CB and cCDCB lines (upper middle panel), magnitude of conduction disorders, CV and TAT (lower left panel), and voltages did not differ between the ECV and control group. Likewise, these parameters were comparable during PACs between the ECV and control group (lower left panel). †Non-normally distributed. cm/s = centimeters per second; mm = millimeter; ms = millisecond; AF = atrial fibrillation; AT = activation time; BB = Bachmann’s bundle; cCDCB = continuous lines of conduction delay and block; CB = conduction block; CD = conduction delay; CT = conduction time; CV = conduction velocity; ECV = electrical cardioversion; LA = left atrium; LAT = local activation times; PAC = premature atrial complexes; PVA = pulmonary vein area; RA = right atrium; SR = sinus rhythm; TAT = total activation time.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s11517-022-02679-w.
## Introduction
The recurrence rate of atrial fibrillation (AF) after electrical cardioversion (ECV) is as high as $57\%$ during the first month after cardioversion, with a peak incidence during the first 5 days [1]. In fact, AF even resumes within 1 or 2 min in up to $27\%$ of patients after restoration of sinus rhythm (SR) [2–5]. This immediate recurrence of AF (IRAF) can be explained by either a high frequency of ectopic beats or the presence of a supervulnerable period immediately after ECV. Duytschaever et al. studied electrophysiological properties in a goat model after spontaneous termination of at least 5 min of AF-induced electrical remodeling and found during SR a transient shortening of the atrial effective refractory period, reduction of intra-atrial conduction velocity (CV), and shortening of the atrial wavelength compared to baseline [2]. During this so-called supervulnerable period, the atria are more susceptible to re-initiation of AF triggered by premature beats [2]. However, heterogeneity in conduction as a result of AF-induced electrical remodeling during this period during SR and premature atrial complexes has never been examined in humans. It is unknown to what extent intra-atrial conduction is impaired during this phase and whether different right and left atrial regions are equally affected. The aim of this case–control study is therefore to compare conduction heterogeneity assessed during the supervulnerable period with long-term SR at a high resolution scale. To our knowledge, this is the first study investigating differences in prevalence and severity of conduction disorders at the epicardial surface of the right atrium, Bachmann’s bundle, and left atrium including the pulmonary vein area immediately after ECV.
## Study population and setting
The study population consisted of participants undergoing elective open-heart surgery in the Erasmus Medical Center. Indications for elective cardiac surgery were either coronary artery disease, aortic valve disease or mitral valve disease, or the combination of these. The case group consisted of AF patients who presented with AF at the onset of the surgical procedure and were electrically cardioverted to SR (structural and electrically remodeled atria). The control group consisted of AF patients who presented with SR (solely structurally remodeled atria as they were in SR for a longer period of time) [6–10]. Thus, only AF-induced electrical remodeling is studied. Participants were matched based on age [11], body mass index [12], and left atrial enlargement [13], known confounders of intra-atrial conduction disorders. In a previous paper of our group [14], we studied the impact of underlying heart disease on conduction heterogeneity during sinus rhythm and did not find any differences between patients with valvular heart disease and ischemic heart disease. Echocardiographic images were used to assess atrial dilatation. This study is approved by the institutional Medical Ethical Committee (resp. MEC 2010–054 [15] and MEC 2014–393 [16]). Prior to the surgical procedure, written informed consent was obtained from all patients. The study complied with the Declaration of Helsinki. Clinical data was extracted from electronic patient files.
## Mapping procedure
High-resolution epicardial mapping was performed during open-heart surgery, prior to extracorporeal circulation [17]. A pacemaker wire temporarily attached to the right atrial free wall functioned as a bipolar reference electrode. A steel wire fixed to the subcutaneous tissue of the thoracic wall was used as an indifferent electrode. Epicardial mapping was performed by shifting an unipolar 128- or a 192-electrode array (electrode diameter respectively 0.65 and 0.45 mm, inter-electrode distances of 2 mm) in a systematic order along predefined sites covering the epicardial surface of both atria (Fig. 1a), including right atrium (from the inferior caval vein up to the right atrial appendage, perpendicular to the caval veins), pulmonary vein area (from the sinus transversus, alongside the borders of the pulmonary veins towards the atrioventricular groove), left atrium (from the lower border of the left pulmonary vein along the left atrioventricular groove towards the left atrial appendage), and Bachmann’s bundle (from the tip of left atrial appendage behind the aorta towards the superior cavo-atrial junction).Fig. 1Epicardial mapping method. a Mapping scheme of RA (RA1-RA4), BB, LA (LA1-LA2), and PVA (PVR and PVL). b An example of a color-coded activation map with isochrones (black lines) drawn at 10 ms. The black arrows indicate the main wavefront directions. An example of calculation of CTs by subtracting the LAT of adjacent electrodes is shown next to the activation map. b An example of the corresponding CB and cCDCB map. ms, milliseconds; BB, Bachmann’s bundle; cCDCB, continuous lines of conduction delay and block lines; CB, conduction block; CD, conduction delay; CT, conduction time; IVC, inferior vena cava; LA, left atrium; LAA, left atrial appendage; LAT, local activation time; PVA, pulmonary vein area; PVL, left pulmonary vein; PVR, right pulmonary vein; RA, right atrium; RAA, right atrial appendage; SVC, superior vena cava At each site, 5 s of SR mapping were recorded, including unipolar epicardial electrograms, a surface electrocardiogram, a bipolar reference electrogram, and a calibration signal (amplitude: 2 mV, duration: 1000 ms). Recordings were sampled with a rate of 1 kHz, amplified (gain: 1000), filtered (bandwidth: 0.5–400 Hz), analog-to-digital-converted (16-bits), and stored on hard disk.
## Mapping data processing
The steepest negative slopes of all atrial potentials were automatically annotated with custom-made software. For each electrode, the local activation time was determined, and color-coded activation maps were reconstructed as illustrated in Fig. 1b [18, 19]. All annotations were visually verified. Mapping sites with less than $50\%$ annotation were excluded from analysis.
## Analysis of intra-atrial conduction disorders
As previously described in a number of mapping studies, inter-electrode conduction times (CTs) were calculated by subtracting the local activation times of each electrode from the adjacent right and lower electrode (Fig. 1b) [18, 19]. Conduction delay (CD) and conduction block (CB) were defined as conduction times of respectively 7–11 ms and ≥ 12 ms, which corresponds to effective conduction velocities of respectively 17 to 29 cm/s and < 17 cm/s [20, 21]. Lines of CB and continuous CDCB (cCDCB) were defined as uninterrupted series of respectively inter-electrode CB or a combination of CD and CB (Fig. 1b). Prevalence of lines of CB and cCDCB lines are expressed as a percentage of the total available number of inter-electrode connections. In all patients, lengths of the longest CB or cCDCB line were assessed at every atrial region. The magnitude of conduction times was defined as the size of inter-electrode time differences in milliseconds and the percentage of patients with conduction times above different magnitudes was calculated. The magnitude of conduction times was analyzed in 10-ms increments. Local CV was computed as an average of velocity estimations between neighboring electrodes (longitudinal, transversal, and diagonal) using discrete velocity vectors as previously described by van Schie et al. [ 22]. From these local CVs, median CV and variation in CV (Δ P5-P95) were calculated for every mapping site. Total activation time and the activation time for each mapping site separately were determined by relating the first and last activation to the reference electrode. Voltage was defined as the peak-to-peak amplitude of the steepest deflection of the unipolar potential. We determined the 5th percentile of the relative frequency histograms of the voltages of all unipolar potentials and compared them between the ECV and control group. Areas of simultaneous activation were excluded from analysis in order to avoid inclusion of far field potentials.
## Intra-atrial conduction disorders during premature beats
To study whether conduction disorders are more pronounced at shorter coupling intervals during the supervulnerable period, conduction heterogeneity during spontaneously occurring premature atrial complexes (PACs) was also investigated. PACs included premature and premature aberrant atrial extrasystoles (Fig. 2). PACs are defined as beats with a shortening in cycle length of ≥ $25\%$ compared to the previous SR beat (Fig. 2a, b). Additionally, the premature aberrant beat has a different direction of propagation compared to the previous SR beat (Fig. 2b) [23]. Prematurity index of PACs was expressed as the ratio between the coupling interval of the PAC and the preceding SR cycle length:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm I}_{\mathrm{prematurity}}=-\frac{{\mathrm{CL}}_{\mathrm{PAC}}}{{\mathrm{CL}}_{\mathrm{SR}}}\bullet100\%$$\end{document}Iprematurity=-CLPACCLSR∙$100\%$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{CL}}_{\mathrm{PAC}}$$\end{document}CLPAC equals the cycle length of the spontaneous PAC and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{CL}}_{\mathrm{SR}}$$\end{document}CLSR the cycle length of the preceding two sinus beats. The difference (Δ) of conduction parameters between the previous SR beat and the PACs were compared between the ECV and the control group. Fig. 2Premature atrial extrasystoles. a Examples of color-coded activation maps during SR (left) and during a premature atrial extrasystole (right) with a shortening in cycle length ≥ $25\%$ compared to the previous SR beat. b Examples of color-coded activation maps during SR (left) and during a prematurely aberrant atrial extrasystole (right) with a shortening in cycle length ≥ $25\%$ and a different direction of propagation compared to the previous SR beat. Isochrones (black lines) drawn at 10-ms increments. The black arrows indicate the main trajectories of activation. ms, milliseconds; mV, millivolts; AES, atrial extrasystoles; LAT, local activation time; SR, sinus rhythm
## Statistical analysis
Statistical analysis was performed with SPSS version 25 (IBM Corporation, Armonk, NY). All data were tested for normality using Shapiro–Wilk test. Normally distributed continuous data were expressed as mean ± standard deviation and skewed data as median (interquartile range). A paired samples t-test or Wilcoxon signed rank test was used to compare continuous parameters for the comparison of SR between the ECV and the control group. For the comparison of continuous data during PACs, an independent samples t-test or Mann–Whitney U test was performed. Categorical data are expressed as absolute numbers (percentages) and analyzed with (McNemar’s symmetry) χ2 or McNemar’s exact test if appropriate. For the comparison of magnitude of conduction times between the ECV and control group, correction for multiple testing was applied. Corrected p values will be reported. A two-sided p value of < 0.05 was considered statistically significant.
## Study population
As presented in Table 1, baseline characteristics between the ECV ($$n = 17$$, 73 ± 7 years; 11 ($64.7\%$) male) and control group ($$n = 17$$, 71 ± 6 years; 12 ($70.6\%$) male) did not differ (all p ≥ 0.05). Participants in the ECV group had either paroxysmal AF ($$n = 6$$, $35.3\%$), persistent AF ($$n = 9$$, $52.9\%$), or longstanding persistent AF ($$n = 2$$, $11.8\%$), while in the control group participants had paroxysmal AF ($$n = 11$$, $64.7\%$) or persistent AF ($$n = 6$$, $35.3\%$) ($$p \leq 1.00$$). Patients in the ECV group had an AF episode duration of 1 month (0.5–4.0) before AF was terminated. Patients in the control group were 54 (13–234) days in SR before surgery. Table 1Characteristics of participantsECV group($$n = 17$$)Control group($$n = 17$$)p valueAge—years (mean ± SD)73 ± 771 ± 60.32Male sex—N (%)11 (64.7)12 (70.6)1.00BMI—kg/m2 (mean ± SD)27 ± 526 ± 50.42History of AF—N (%)17 (100.0)17 (100.0)1.00 Paroxysmal6 (35.3)11 (64.7)p ≥ 0.017* Persistent9 (52.9)6 (35.3)p ≥ 0.017* Longstanding persistent2 (11.8)0 (0.0)p ≥ 0.017*Underlying heart disease—N (%)1.00 IHD2 (11.8)2 (11.8) (i)VHD15 (88.2)15 (88.2) AVD2 (11.8)4 (23.5) AVD and CAD2 (11.8)1 (5.9) MVD9 (52.9)8 (47.1) MVD and CAD2 (11.8)2 (11.8)Echocardiography LVF—N (%)0.13 Normal11 (64.7)16 (94.1) Mild dysfunction3 (17.6)0 (0.0) Moderate dysfunction3 (17.6)0 (0.0) Severe dysfunction0 (0.0)1 (5.9) LAVI—ml/m2 (median (IQR))47 (43–63)†46 (38–66)‡0.72Medication—N (%) Antiarrhythmic drugs0.45 Class I0 (0.0)0 (0.0) Class II11 (64.7)8 (47.1) Class III4 (23.5)4 (23.5) Class IV1 (5.9)1 (5.9) Digoxin5 (29.4)3 (17.6)0.50N, number; SD, standard deviation; AF, atrial fibrillation; AVD, aortic valve disease; BMI, body mass index; CAD, coronary artery disease; ECV, electrical cardioversion; IHD, ischemic heart disease; LA, left atrium; LAVI, left atrial volume index; LVF, left ventricular function; MVD, mitral valve disease; (i)VHD, (ischemic and) valvular heart disease. †$$n = 15$.$ ‡$$n = 12$.$ * Bonferroni correction was applied
## Mapping data characteristics
In the ECV and control group, a total of respectively 164,099 unipolar potentials (9192 potentials/patient (7421–11,250)) and 149,521 unipolar potentials (8418 potentials/patient (7112–10,831)) were available for further analysis ($$p \leq 0.52$$). Due to simultaneous activation, $2.3\%$ of the potentials in the ECV group and $1.1\%$ of the potentials in the control group were excluded from analysis. SR cycle length during epicardial mapping was 788 ms (736–894) in the ECV group and 855 ms (764–962) in the control group ($$p \leq 0.10$$).
## Biatrial conduction
In the entire study population, each patient in the ECV group, as well as in the control group, had areas of CD and CB. Differences in prevalences and length of longest lines of CB and cCDCB in both atria between the ECV and control group are shown in Fig. 3a, c. As illustrated in Fig. 3a, the prevalence of CB and cCDCB in both atria did not differ between the control and ECV group (CB: 3.1 ± $1.7\%$ vs. 3.1 ± $1.9\%$, $$p \leq 0.93$$; cCDCB: 3.7 ± $1.8\%$ vs. 3.9 ± $1.9\%$, $$p \leq 0.78$$). Additionally, the length of the longest lines of both CB and cCDCB was the same in patients immediately after ECV and during long-term SR (CB: 48 mm (31–66) vs. 40 mm (25–53), $$p \leq 0.23$$; cCDCB: 67 ± 26 mm vs. 67 ± 35 mm, $$p \leq 1.00$$) (Fig. 3c).Fig. 3Prevalences and length of longest CB and cCDCB lines. a Prevalence of CB and cCDCB in both atria. b Spatial distribution of prevalences of CB and length of longest CB lines. c Length of longest lines of CB and cCDCB in both atria. d Spatial distribution of prevalences of cCDCB and length of longest cCDCB lines. †Non-normally distributed. mm, millimeter; ms, milliseconds; cCDCB, continuous lines of conduction delay and block; CB, conduction block; ECV, electrical cardioversion Regional differences in prevalences and the length of longest lines of CB and cCDCB between the control and the ECV group are shown in the Fig. 3b, d and Supplemental Table 1. Conduction disorders are observed in both groups at all locations, but mainly at Bachmann’s bundle. Figure 3b, d and Supplemental Table 1 show that both the prevalence of CB and cCDCB as well as the length of the longest CB lines and cCDCB lines at every location did not differ between the ECV and control group (all p ≥ 0.05).
Figure 4 shows the median CV for each patient in both atria and for each location separately. Biatrial median CV was not reduced in the ECV group (90 cm/s (84–99) vs. 89 cm/s (85–95), $$p \leq 0.69$$). Biatrial variation in CV also did not differ between both groups (Δ P5-P95: 127 cm/s (123–132) vs. 125 cm/s (121–136), $$p \leq 0.87$$). Comparing CV for each location separately, again no differences in median CV were found between the ECV and the control group (right atrium: 92 ± 7 cm/s vs. 88 ± 7 cm/s, $$p \leq 0.11$$; Bachmann’s bundle: 80 ± 12 cm/s vs. 84 ± 9 cm/s, $$p \leq 0.22$$; pulmonary vein area: 90 cm/s (77–98) vs. 93 cm/s (85–104), $$p \leq 0.34$$; left atrium: 90 ± 13 cm/s vs. 90 ± 8 cm/s, $$p \leq 0.90$$). As shown in Supplemental Table 1, the variation in CV per location also was comparable between patients in the ECV and control group (all p ≥ 0.05).Fig. 4Biatrial and regional conduction velocity. Left panel: biatrial median conduction velocity displayed for each patient. Right panel: median conduction velocity displayed for each patient per region separately. †Non-normally distributed. cm/s, centimeter per second; BB, Bachmann’s bundle; ECV, electrical cardioversion; LA, left atrium; PVA, pulmonary vein area; RA, right atrium
## Severity of conduction disorders
Figure 5 shows the magnitude of conduction times in both atria for the ECV and control group separately. Each patient in both groups had at least one CT ≥ 32 ms. The magnitude of conduction times was comparable in the ECV and control group (Bonferroni corrected p ≥ 0.006). By comparing the different atrial regions separately between both groups, again there were no differences in the magnitude of conduction times (right atrium: Bonferroni corrected p ≥ 0.008; Bachmann’s bundle: Bonferroni corrected p ≥ 0.005; pulmonary vein area: Bonferroni corrected p ≥ 0.01; left atrium: Bonferroni corrected p ≥ 0.01).Fig. 5Severity of conduction disorders. Magnitude of CTs measured after ECV and during long-term SR for the entire study population in increments of 10 ms. ms, milliseconds; CTs, conduction times; ECV, electrical cardioversion; SR, sinus rhythm
## Relation between conduction heterogeneity and biatrial activation time
Figure 6 illustrates for each patient the total activation times (Fig. 6a) and the activation time per region separately (Fig. 6b). The supervulnerable period was not associated with a prolonged biatrial total activation times (158 ms (137–166) vs. 145 ms (122–160), $$p \leq 0.41$$) or a prolonged activation time for each location separately (all p ≥ 0.05). Activation time was longest at the right atrium (ECV: 82 ms (69–90), control: 84 ms (76–110), $$p \leq 0.57$$).Fig. 6Total activation time. a Biatrial total activation time displayed for each individual patient. b Activation time displayed for each patient per region separately. †Non-normally distributed. ms, milliseconds; BB, Bachmann’s bundle; ECV, electrical cardioversion; LA, left atrium; PVA, pulmonary vein area; RA, right atrium
## Unipolar voltages
Comparison of the 5th percentile of all biatrial voltages between the control and ECV group did not reveal lower unipolar voltages during the supervulnerable period (Table 2; ECV group: 0.8 ± 0.4 mV vs. control group 0.9 ± 0.5 mV; $$p \leq 0.31$$). When comparing the 5th percentile of voltages for each location separately, there were also no differences between both groups (all p ≥ 0.05).Table 2Unipolar voltages in the ECV and control groupECV($$n = 17$$)Control($$n = 17$$)p valueBiatrial—mV (mean ± SD)0.7 ± 0.30.9 ± 0.50.31Right atrium—mV (median (IQR))1.0 (0.7–1.3)0.9 (0.5–1.3)0.83Bachmann’s bundle—mV (mean ± SD)1.0 ± 0.61.1 ± 0.80.49Pulmonary vein area—mV (median (IQR))0.6 (0.5–1.4)0.9 (0.6–1.2)0.94Left atrium—mV (mean ± SD)1.2 ± 0.71.6 ± 0.70.22mV, millivolts; IQR, interquartile range; SD, standard deviation; ECV, electrical cardioversion
## Conduction disorders during premature beats
Seven patients ($41\%$) in the control group had a total of 11 PACs (4 premature atrial extrasystole ($36\%$); 7 premature aberrant atrial extrasystole ($64\%$)) whereas in the ECV group, seven patients ($41\%$) had a total of 22 PACs (6 premature atrial extrasystole ($27\%$); 16 premature aberrant atrial extrasystole ($73\%$)). The prematurity index of the PACs did not differ between the control and ECV group (61.2 ± $10.3\%$ versus 55.6 ± $12.6\%$, $$p \leq 0.22$$).
Table 3 shows for both groups the difference (Δ) in conduction disorders during the PAC compared to the corresponding SR beat. The increase in conduction disorders was not more pronounced during the supervulnerable period, as the Δ prevalence and Δ length of longest CB and cCDCB lines did not differ between both groups (all p ≥ 0.05). Additionally, Δ CV was similar between the control and the ECV group as the CV decreased with respectively 11 ± 13 cm/s and 6 ± 19 cm/s between SR and PACs ($$p \leq 0.48$$). The supervulnerable period was also not associated with a more pronounced decrease of the 5th percentile of the voltage histograms in patients after ECV (− 0.3 mV (− 1.0–0.4) vs. − 0.2 mV (− 2.2–0.6), $$p \leq 0.87$$).Table 3Differences (Δ) in conduction during PACs compared SRECV($$n = 22$$)Control($$n = 11$$)p valueΔ CBPrevalence—% (median (IQR))1.0 (− 0.7–3.6)2.6 (0.0–5.6)0.30Length of longest CB line—mm (median (IQR))6.0 (2.0–10)12.5 (3.0–27.0)0.27Δ cCDCBPrevalence—% (mean ± SD)3.6 ± 6.21.1 ± 5.30.23Length of longest cCDCB line—mm (median (IQR))0.0 (− 2–8)18 (− 12.0–28)0.13Δ CV—cm/s (mean ± SD) − 6 ± 19 − 11 ± 130.48Δ P5 of unipolar voltages—mV (median (IQR)) − 0.3 (− 1.0–0.4) − 0.2 (− 2.2–0.6)0.87cm/s, centimeters per second; SD, standard deviation; IQR, interquartile range; CB, conduction block; cCDCB, continuous lines of conduction delay and block; CV, conduction velocity; PAC, premature atrial complexes
## Key findings
This high-resolution intra-operative mapping study is the first to investigate biatrial heterogeneity in conduction during the so-called supervulnerable period immediately after ECV. Compared to long-term SR, no increased conduction heterogeneity was found immediately after ECV, since the frequency and severity of conduction disorders, as well as CV and TAT, did not differ during SR between the control and ECV group. Additionally, conduction disorders during PACs were not more pronounced immediately after ECV. Hence, our data suggest that the supervulnerable period may not be characterized by impaired intra-atrial conduction.
## Conduction disorders as a predictor for early atrial fibrillation recurrences
Rosenbaum introduced the term “domestication of AF,” meaning that the longer the duration of AF episodes, the more difficult it becomes to achieve SR. After termination of AF, $27\%$ of patients have an IRAF within 1 min after successful ECV [3–5]. At a higher heart rate, e.g., during AF, atrial CV will decrease while the wavelength of the atrial impulse and the atrial effective refractory period shortens [24, 25]. These changes during AF promote re-entry as they reduce the likelihood that a wavefront circling around a line of CB collides with its refractory tail [26]. After AF termination, it is generally assumed that the combination of increased dispersion of the atrial effective refractory period and a reduced CV in combination with the presence of triggers such as PAC may increase the susceptibility to AF recurrence.
Duytschaever et al. examined the supervulnerable phase immediately after AF termination in goats with non-remodeled and electrically remodeled atria (48 h of electrically maintained AF) [2]. Baseline atrial effective refractory period, intra-atrial CV, and atrial wavelength were determined [2]. After the baseline study, AF was induced lasting at least 5 min and all measures were repeated immediately after spontaneous restoration of SR [2]. They found transient shortening of the atrial effective period, reduction of intra-atrial CV during SR, and shortening of the atrial wavelength compared to baseline [2]. These observations implied the existence of a vulnerable substrate for initiation of reentry after AF termination in goats. One possible explanation for slowing of conduction after AF termination is a decrease in sodium and increase in potassium currents due to high atrial rates during AF [27–30]. The resting membrane potential, and as a result the driving force for sodium ion exchange, will decrease resulting in a lower action potential velocity upstroke and thus a lower CV [2]. However, we did not observe a reduction of CV in humans immediately after AF termination. We found that intra-atrial conduction during the supervulnerable period and long-term SR were comparable. Also during PACs, there was no reduction of CV. In other words, our findings suggest that an increased susceptibility to AF re-initiation during the so-called supervulnerable period may be not determined by a reduction in CV.
A possible explanation may be that the normalization of intracellular sodium concentrations is restored within only a few SR beats and is not present for 1 to 2 min as previously suggested. Another explanation may be the duration of AF and its impact on electrical remodeling [31, 32]. In our ECV group, patients had 1-month AF before termination of AF ranging (IQR) between 2 weeks and 4 months. Only 2 patients had longstanding persistent AF (AF duration of 1 year and 1.5 year) before termination, while longer AF episodes are correlated with more electrical remodeling and thus a reduced CV [31, 32].
In humans, atrial conduction during the supervulnerable period has not been previously investigated. However, a few studies reported on the reversal of electrical remodeling over time after termination of AF. Yu et al. performed endocardial mapping of the left atrium and right atrium 30 min ($t = 30$) after restoration of SR in humans and studied conduction times during four consecutive days using two ten-polar electrode catheters positioned at the right atrial appendage and distal coronary sinus [33]. Conduction times were measured from the second to the fifth pairs of electrodes (5-mm inter-pair distance), while the first pair was used for pacing at a basic cycle length of 700 ms [33]. After termination of AF, inter-atrial conduction did not change during these 4 days [33]. We studied conduction times during SR as a measure of inter-electrode differences in local activation time ≥ 12 ms (CB) and as the activation time of the right atrium and left atrium. Moreover, we investigated these conduction properties during the supervulnerable period rather than the reversal of electrical remodeling over time starting at $t = 30$ min. However, our findings that there are no differences in frequency of CB and activation time of the right atrium and left atrium during the supervulnerable period are consistent with these findings. Additionally, Yu et al. [ 33] examined surface electrocardiograms over the same time course after AF termination using the duration of the p wave as a measure of total activation times in patients with persistent AF. They found no change in p wave duration over time [33]. In contrast, in another study, p wave duration was prolonged within 5 to 20 min after AF termination compared to 24 h and 1 month post-conversion [34]. However, in both studies, there were no measurements during the supervulnerable period. The control group in our study was in SR for approximately 1.5 months and still no differences in intra-atrial conduction were found between control and ECV group.
## Premature beats as a trigger for early atrial fibrillation recurrences
Triggers, e.g., PACs, play an important role in AF onset in patients with an IRAF [2, 35]. In our study population, a higher incidence of PACs was present during the supervulnerable period compared to long-term SR. In both groups, seven patients had PACs, but 22 PACs were found in the ECV group, while only 11 PACs were found in the control group. A possible mechanism that enhances PAC-initiated IRAF is the occurrence of intracellular calcium overload due to the previous high-rate AF episode promoting late phase 3 early afterdepolarization-induced PACs [36]. The high AF rates result in an increase in intracellular sodium leading to cellular calcium load mediated by sodium-calcium exchange [36]. After AF termination, strong calcium release in the sarcoplasmic reticulum stimulates extrusion of calcium through sodium-calcium exchanger [36]. As a result, a transient period of hypercontractility occurs [37]. Additionally, the inward current of calcium mediated by the exchanger is most likely responsible for the transient action potential duration prolongation and early afterdepolarizations [36].
Additionally to the occurrence of atrial triggers, IRAF requires a vulnerable substrate for reentry. In our study population, even conduction disorders caused by PACs were not more pronounced during the supervulnerable period. A limitation is that we did not study PACs that did indeed trigger an IRAF, yet the prematurity of PACs is comparable to previously reported PACs inducing AF. In the goat model of AF, all IRAF episodes were triggered by PACs with coupling intervals ranging between 310 and 580 ms; ectopic beats with a coupling interval > 600 ms never resulted in IRAF [2]. In humans, the coupling intervals of PACs initiating IRAF were shorter (418 ms) than PACs which did not initiate IRAF (661 ms) ($p \leq 0.05$) [3]. PACs in our study population had a mean coupling interval of 482 ms (prematurity index: $61.2\%$) and 470 ms (prematurity index: $55.6\%$) in respectively the control and ECV group. Although no IRAF was initiated, these coupling intervals were short enough to initiate AF.
## Potential other mechanisms leading to early atrial fibrillation recurrences
Since intra-atrial conduction is not impaired after AF termination during both SR and PACs, other mechanisms may be responsible for the occurrence of IRAF. As previously mentioned, an increased dispersion of atrial effective refractory period, a reduced CV, and the frequency of triggers may enhance the susceptibility to AF recurrences. In the present study, we did not investigate atrial effective refractory period, as our study was designed to study conduction during SR and PACs. However, several other human studies found a shortening of atrial effective refractory period during the supervulnerable period [34, 38–40]. Additionally, a significant dispersion of atrial refractoriness between different right atrium sites was present [38]. This, in combination with a higher frequency of PACs, may be a possible explanation for the occurrence of IRAF, as it facilitates the likelihood of encountering unidirectional conduction block, which is a prerequisite for development of re-entrant circuits.
In the present study, biatrial CV was not reduced in the ECV group ($$p \leq 0.69$$). However, we have not studied the rate-dependent slowing of CV (CV restitution) which may precede AF initiation [41]. Narayan et al. showed that patients with paroxysmal AF had steep CV restitution interacting with steep action potential restitution, which may cause rapid tachycardias to initiate AF [41, 42]. On the other hand, patients with persistent AF, with more advanced remodeling of the atria, and broad CV restitution developed AF at lower heartrates [41–44]. However, the precise mechanism underlying the relationship between CV restitution and AF initiation is still unclear.
## Low voltage areas during the supervulnerable period
In our present study, we found no relationship between low voltage areas and the supervulnerable period. Little is known about the impact of electrical remodeling on unipolar voltages. As previously mentioned, it is likely that due to electrical remodeling during AF, sodium current is reduced resulting in a decrease of voltages displayed in the unipolar electrogram which may still be present after AF termination [27–30]. However, we did not find low voltage areas during the supervulnerable period.
## Study limitations
Patients with a history of AF may have had variable degrees of atrial remodeling as some patients had persistent or longstanding persistent AF, while other patients had paroxysmal AF. Even if we perform a subanalysis in patients with paroxysmal AF between the ECV and control group, there were no differences found in any of the conduction parameters (see Supplemental Table 2). In patients with persistent and longstanding persistent AF, comparable results were found (see supplemental Table 3). Additionally, PACs triggering AF were not investigated.
## Clinical relevance
This study is the first to investigate conduction disorders due to AF-related electrical remodeling immediately after ECV in high resolution of the entire atrial surface. Since intra-atrial conduction is not impaired after AF termination during both SR and PACs compared to long-term SR, it is suggested that IRAF is not enhanced by conduction disorders. To further investigate conduction impairment during the shortest coupling intervals, programmed electrical stimulation reaching atrial refractoriness should be performed to examine CV restitution in relation to AF initiation. Other mechanisms, such as an increased dispersion of atrial effective refractory period and the frequency of triggers, may also be possible explanations for the occurrence of IRAF. These findings help to better understand the mechanism behind the IRAF and improve treatment strategies aimed at eliminating IRAF.
## Conclusion
This high-resolution intra-operative mapping study is the first to investigate characteristics of biatrial conduction immediately after ECV during the so-called supervulnerable period. Compared to long-term SR, there was no impaired intra-atrial conduction immediately after ECV. These observations suggest that the supervulnerable period is not characterized by increased conduction heterogeneity during SR or PACs. However, to further investigate conduction impairment during the shortest coupling intervals, programmed electrical stimulation reaching atrial refractoriness should be performed to examine CV restitution.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 22 KB)
## References
1. Tieleman RG, Van Gelder IC, Crijns HJ, De Kam PJ, Van Den Berg MP, Haaksma J. **Early recurrences of atrial fibrillation after electrical cardioversion: a result of fibrillation-induced electrical remodeling of the atria?**. *J Am Coll Cardiol* (1998) **31** 167-173. DOI: 10.1016/S0735-1097(97)00455-5
2. Duytschaever M, Danse P, Allessie M. **Supervulnerable phase immediately after termination of atrial fibrillation**. *J Cardiovasc Electrophysiol* (2002) **13** 267-275. DOI: 10.1046/j.1540-8167.2002.00267.x
3. Timmermans C, Rodriguez LM, Smeets JL, Wellens HJ. **Immediate reinitiation of atrial fibrillation following internal atrial defibrillation**. *J Cardiovasc Electrophysiol* (1998) **9** 122-128. DOI: 10.1111/j.1540-8167.1998.tb00893.x
4. Sra J, Biehl M, Blanck Z, Dhala A, Jazayeri MR, Deshpande S. **Spontaneous reinitiation of atrial fibrillation following transvenous atrial defibrillation**. *Pacing Clin Electrophysiol* (1998) **21** 1105-1110. DOI: 10.1111/j.1540-8159.1998.tb00157.x
5. Wellens HJ, Lau CP, Luderitz B, Akhtar M, Waldo AL, Camm AJ. **Atrioverter: an implantable device for the treatment of atrial fibrillation**. *Circulation* (1998) **98** 1651-1656. DOI: 10.1161/01.CIR.98.16.1651
6. Spach MS, Boineau JP. **Microfibrosis produces electrical load variations due to loss of side-to-side cell connections: a major mechanism of structural heart disease arrhythmias**. *Pacing Clin Electrophysiol* (1997) **20** 397-413. DOI: 10.1111/j.1540-8159.1997.tb06199.x
7. Darby AE, Dimarco JP. **Management of atrial fibrillation in patients with structural heart disease**. *Circulation* (2012) **125** 945-957. DOI: 10.1161/CIRCULATIONAHA.111.019935
8. Wijffels MC, Kirchhof CJ, Dorland R, Allessie MA. **Atrial fibrillation begets atrial fibrillation. A study in awake chronically instrumented goats**. *Circulation* (1995) **92** 1954-68. DOI: 10.1161/01.CIR.92.7.1954
9. Everett THt, Li H, Mangrum JM, McRury ID, Mitchell MA, Redick JA. **Electrical, morphological, and ultrastructural remodeling and reverse remodeling in a canine model of chronic atrial fibrillation**. *Circulation* (2000) **102** 1454-60. DOI: 10.1161/01.CIR.102.12.1454
10. Psaty BM, Manolio TA, Kuller LH, Kronmal RA, Cushman M, Fried LP. **Incidence of and risk factors for atrial fibrillation in older adults**. *Circulation* (1997) **96** 2455-2461. DOI: 10.1161/01.CIR.96.7.2455
11. van der Does WFB, Houck CA, Heida A, van Schie MS, van Schaagen FRN, Taverne Y. **Atrial electrophysiological characteristics of aging**. *J Cardiovasc Electrophysiol* (2021) **32** 903-912. DOI: 10.1111/jce.14978
12. Schram-Serban C, Heida A, Roos-Serote MC, Knops P, Kik C, Brundel B. **Heterogeneity in conduction underlies obesity-related atrial fibrillation vulnerability**. *Circ Arrhythm Electrophysiol* (2020) **13** e008161. DOI: 10.1161/CIRCEP.119.008161
13. Schotten U, Verheule S, Kirchhof P, Goette A. **Pathophysiological mechanisms of atrial fibrillation: a translational appraisal**. *Physiol Rev* (2011) **91** 265-325. DOI: 10.1152/physrev.00031.2009
14. Heida A, van der Does WFB, van Staveren LN, Taverne Y, Roos-Serote MC, Bogers A. **Conduction heterogeneity: impact of underlying heart disease and atrial fibrillation**. *JACC Clin Electrophysiol* (2020) **6** 1844-1854. DOI: 10.1016/j.jacep.2020.09.030
15. van der Does L, Yaksh A, Kik C, Knops P, Lanters EAH, Teuwen CP. **QUest for the Arrhythmogenic Substrate of Atrial fibRillation in patients undergoing cardiac surgery (QUASAR study): rationale and design**. *J Cardiovasc Transl Res* (2016) **9** 194-201. DOI: 10.1007/s12265-016-9685-1
16. Lanters EA, van Marion DM, Kik C, Steen H, Bogers AJ, Allessie MA. **HALT & REVERSE: Hsf1 activators lower cardiomyocyt damage; towards a novel approach to REVERSE atrial fibrillation**. *J Transl Med* (2015) **13** 347. DOI: 10.1186/s12967-015-0714-7
17. Yaksh A, Kik C, Knops P, Roos-Hesselink JW, Bogers AJ, Zijlstra F. **Atrial fibrillation: to map or not to map?**. *Neth Heart J* (2014) **22** 259-266. PMID: 24129689
18. Teuwen CP, Yaksh A, Lanters EA, Kik C, van der Does LJ, Knops P. **Relevance of conduction disorders in Bachmann’s bundle during sinus rhythm in humans**. *Circ Arrhythm Electrophysiol* (2016) **9** e003972. DOI: 10.1161/CIRCEP.115.003972
19. Mouws E, van der Does L, Kik C, Lanters EAH, Teuwen CP, Knops P. **Impact of the arrhythmogenic potential of long lines of conduction slowing at the pulmonary vein area**. *Heart Rhythm* (2019) **16** 511-519. DOI: 10.1016/j.hrthm.2018.10.027
20. Allessie M, Ausma J, Schotten U. **Electrical, contractile and structural remodeling during atrial fibrillation**. *Cardiovasc Res* (2002) **54** 230-246. DOI: 10.1016/S0008-6363(02)00258-4
21. de Groot NM, Houben RP, Smeets JL, Boersma E, Schotten U, Schalij MJ. **Electropathological substrate of longstanding persistent atrial fibrillation in patients with structural heart disease: epicardial breakthrough**. *Circulation* (2010) **122** 1674-1682. DOI: 10.1161/CIRCULATIONAHA.109.910901
22. van Schie MS, Heida A, Taverne Y, Bogers A, de Groot NMS. **Identification of local atrial conduction heterogeneities using high-density conduction velocity estimation**. *Europace* (2021) **23** 1815-1825. DOI: 10.1093/europace/euab088
23. Teuwen CP, Kik C, van der Does L, Lanters EAH, Knops P, Mouws E. **Quantification of the arrhythmogenic effects of spontaneous atrial extrasystole using high-resolution epicardial mapping**. *Circ Arrhythm Electrophysiol* (2018) **11** e005745. DOI: 10.1161/CIRCEP.117.005745
24. Smeets JL, Allessie MA, Lammers WJ, Bonke FI, Hollen J. **The wavelength of the cardiac impulse and reentrant arrhythmias in isolated rabbit atrium. The role of heart rate, autonomic transmitters, temperature, and potassium**. *Circ Res* (1986) **58** 96-108. DOI: 10.1161/01.RES.58.1.96
25. Rensma PL, Allessie MA, Lammers WJ, Bonke FI, Schalij MJ. **Length of excitation wave and susceptibility to reentrant atrial arrhythmias in normal conscious dogs**. *Circ Res* (1988) **62** 395-410. DOI: 10.1161/01.RES.62.2.395
26. Mines GR. **On dynamic equilibrium in the heart**. *J Physiol* (1913) **46** 349-383. DOI: 10.1113/jphysiol.1913.sp001596
27. Gaspo R, Bosch RF, Bou-Abboud E, Nattel S. **Tachycardia-induced changes in Na+ current in a chronic dog model of atrial fibrillation**. *Circ Res* (1997) **81** 1045-1052. DOI: 10.1161/01.RES.81.6.1045
28. Bosch RF, Zeng X, Grammer JB, Popovic K, Mewis C, Kuhlkamp V. **Ionic mechanisms of electrical remodeling in human atrial fibrillation**. *Cardiovasc Res* (1999) **44** 121-131. DOI: 10.1016/S0008-6363(99)00178-9
29. Van Wagoner DR, Pond AL, McCarthy PM, Trimmer JS, Nerbonne JM. **Outward K+ current densities and Kv1.5 expression are reduced in chronic human atrial fibrillation**. *Circ Res* (1997) **80** 772-81. DOI: 10.1161/01.RES.80.6.772
30. Nattel S, Maguy A, Le Bouter S, Yeh YH. **Arrhythmogenic ion-channel remodeling in the heart: heart failure, myocardial infarction, and atrial fibrillation**. *Physiol Rev* (2007) **87** 425-456. DOI: 10.1152/physrev.00014.2006
31. Ausma J, Litjens N, Lenders MH, Duimel H, Mast F, Wouters L. **Time course of atrial fibrillation-induced cellular structural remodeling in atria of the goat**. *J Mol Cell Cardiol* (2001) **33** 2083-2094. DOI: 10.1006/jmcc.2001.1472
32. Schotten U, Duytschaever M, Ausma J, Eijsbouts S, Neuberger HR, Allessie M. **Electrical and contractile remodeling during the first days of atrial fibrillation go hand in hand**. *Circulation* (2003) **107** 1433-1439. DOI: 10.1161/01.CIR.0000055314.10801.4F
33. Yu WC, Lee SH, Tai CT, Tsai CF, Hsieh MH, Chen CC. **Reversal of atrial electrical remodeling following cardioversion of long-standing atrial fibrillation in man**. *Cardiovasc Res* (1999) **42** 470-476. DOI: 10.1016/S0008-6363(99)00030-9
34. Manios EG, Kanoupakis EM, Chlouverakis GI, Kaleboubas MD, Mavrakis HE, Vardas PE. **Changes in atrial electrical properties following cardioversion of chronic atrial fibrillation: relation with recurrence**. *Cardiovasc Res* (2000) **47** 244-253. DOI: 10.1016/S0008-6363(00)00100-0
35. Inoue K, Kurotobi T, Kimura R, Toyoshima Y, Itoh N, Masuda M. **Trigger-based mechanism of the persistence of atrial fibrillation and its impact on the efficacy of catheter ablation**. *Circ Arrhythm Electrophysiol* (2012) **5** 295-301. DOI: 10.1161/CIRCEP.111.964080
36. Burashnikov A, Antzelevitch C. **Reinduction of atrial fibrillation immediately after termination of the arrhythmia is mediated by late phase 3 early afterdepolarization-induced triggered activity**. *Circulation* (2003) **107** 2355-2360. DOI: 10.1161/01.CIR.0000065578.00869.7C
37. Denham NC, Pearman CM, Caldwell JL, Madders GWP, Eisner DA, Trafford AW. **Calcium in the pathophysiology of atrial fibrillation and heart failure**. *Front Physiol* (2018) **9** 1380. DOI: 10.3389/fphys.2018.01380
38. Pandozi C, Bianconi L, Villani M, Gentilucci G, Castro A, Altamura G. **Electrophysiological characteristics of the human atria after cardioversion of persistent atrial fibrillation**. *Circulation* (1998) **98** 2860-2865. DOI: 10.1161/01.CIR.98.25.2860
39. Kamalvand K, Tan K, Lloyd G, Gill J, Bucknall C, Sulke N. **Alterations in atrial electrophysiology associated with chronic atrial fibrillation in man**. *Eur Heart J* (1999) **20** 888-895. DOI: 10.1053/euhj.1998.1404
40. Lubinski A, Kempa M, Lewicka-Nowak E, Krolak T, Raczak G, Swiatecka G. **Electrical atrial remodeling assessed by monophasic action potential and atrial refractoriness in patients with structural heart disease**. *Pacing Clin Electrophysiol* (1998) **21** 2440-2444. DOI: 10.1111/j.1540-8159.1998.tb01197.x
41. Lalani GG, Schricker A, Gibson M, Rostamian A, Krummen DE, Narayan SM. **Atrial conduction slows immediately before the onset of human atrial fibrillation: a bi-atrial contact mapping study of transitions to atrial fibrillation**. *J Am Coll Cardiol* (2012) **59** 595-606. DOI: 10.1016/j.jacc.2011.10.879
42. Narayan SM, Kazi D, Krummen DE, Rappel WJ. **Repolarization and activation restitution near human pulmonary veins and atrial fibrillation initiation: a mechanism for the initiation of atrial fibrillation by premature beats**. *J Am Coll Cardiol* (2008) **52** 1222-1230. DOI: 10.1016/j.jacc.2008.07.012
43. Rohr S, Kucera JP, Kleber AG. **Slow conduction in cardiac tissue, I: effects of a reduction of excitability versus a reduction of electrical coupling on microconduction**. *Circ Res* (1998) **83** 781-794. DOI: 10.1161/01.RES.83.8.781
44. Shaw RM, Rudy Y. **Ionic mechanisms of propagation in cardiac tissue Roles of the sodium and L-type calcium currents during reduced excitability and decreased gap junction coupling**. *Circ Res* (1997) **81** 727-41. DOI: 10.1161/01.RES.81.5.727
|
---
title: Microbiome Structure and Mucosal Morphology of Jejunum Appendix and Colon of
Rats in Health and Dysbiosis
authors:
- Chenyi Shao
- Xiaobo Song
- Lili Wang
- Hongying Zhang
- Yinhui Liu
- Chunhao Wang
- Shenmin Chen
- Baowei Ren
- Shu Wen
- Jing Xiao
- Li Tang
journal: Current Microbiology
year: 2023
pmcid: PMC9988748
doi: 10.1007/s00284-023-03224-0
license: CC BY 4.0
---
# Microbiome Structure and Mucosal Morphology of Jejunum Appendix and Colon of Rats in Health and Dysbiosis
## Abstract
Gut microbiota contributes to human health. Plenty of studies demonstrate that antibiotics can disrupt gut ecosystem leading to dysbiosis. Little is known about the microbial variation of appendix and its up/downstream intestine after antibiotic treatment. This study aimed to investigate the microbiome and mucosal morphology of jejunum, appendix, and colon of rats in health and dysbiosis. A rodent model of antibiotic-induced dysbiosis was employed. Microscopy was used to observe mucosal morphological changes. 16S rRNA sequencing was performed for identifying bacterial taxa and microbiome structure. The appendices of dysbiosis were found enlarged and inflated with loose contents. Microscopy revealed the impairment of intestinal epithelial cells. High-throughput sequencing showed the Operational Taxonomic Units changed from 361 ± 33, 634 ± 18, 639 ± 19 in the normal jejunum, appendix, colon to 748 ± 98, 230 ± 11, 253 ± 16 in the disordered segments, respectively. In dysbiosis, Bacteroidetes translocated inversely from the colon and appendix ($0.26\%$, $0.23\%$) to the jejunum ($13.87\%$ ± $0.11\%$); the relative abundance of all intestinal Enterococcaceae increased, while Lactobacillaceae decreased. Several bacterial clusters were found correlated to the normal appendix, whereas nonspecific clusters correlated to the disordered appendix. In conclusion, species richness and evenness reduced in the disordered appendix and colon; similar microbiome patterns were shared between the appendix and colon regardless of dysbiosis; site-specific bacteria were missing in the disordered appendix. Appendix is likely a transit region involving in upper and lower intestinal microflora modulation. The limitation of this study is all the data were derived from rats. We must be cautious about translating the microbiome results from rats to humans.
## Introduction
The human gastrointestinal tract is home to a large number of microorganisms known as gut microbiome. The distal gut microbiota is mainly composed of strict anaerobes, but also some facultative anaerobes. The quantity of bacteria increases along the gastrointestinal tract, while the stomach has a small number of bacteria due to its acid environment [1]. The composition of gut microbiome can be impacted by many factors, such as intestinal pH; environmental temperature; diet, drug therapy, in addition to the kinship [2]. Extensive studies have shown that the gut microorganisms are involved in human metabolism and nutrition. The gut bacteria can produce a variety of vitamins, synthesize all essential and non-essential amino acids [3], and metabolize non-digestible carbohydrates [4]. The gut microbiota also produces antibacterial compounds, and competes for nutrients and attachment loci in the gut wall to prevent the colonization of pathogens [5]. The gastrointestinal tract is a dynamic microecosystem of which equilibrium is essential to our health. An imbalance in the composition and/or activity of the gut microbiome, which may have negative impacts on health, is referred to as dysbiosis. It may lead to irritable bowel syndrome (IBS) and inflammatory bowel diseases (IBD) [6].
The appendix is a part of the digestive organs located at the junction between the small intestine and large intestine. It was regarded as a functionless vestige from evolutionary history. Nowadays, researchers recognized it as a repository for gut commensal microflora and a part of the immune system [7]. Masses of lymphoid tissues in the appendix enable the adaptation of commensal microflora to the intestinal niches. In addition, the appendix forms immune-mediated biofilms where gut probiotics reside in. Compared to other intestinal regions, the appendix assembles much biofilm as a “hotbed” of intestinal flora [8]. Moreover, its pouch-like structure helps to prevent the loss of gut commensals during diarrhoea [7]. Previous culture-based studies showed that diverse microorganisms, such as Bacteroides fragilis, Escherichia coli, Pseudomonas aeruginosa, and Peptostreptococcus species, were frequently isolated from normal and inflammatory appendices [9–11]. In recent years, researchers have started to use gene-sequencing analysis to investigate the microbiome of appendices in health and appendicitis [12–14]. So far, little is known about the ecological variation of appendix and surroundings in the early inflammation, as well as the action of appendix to the gut microbial community during gut dysbiosis.
The present study aimed to compare the microbiome structure and mucosal morphology of the jejunum, appendix and colon of rats in health and dysbiosis. An antibiotic-induced dysbiosis rodent model was established to explore the in situ microbiota in the appendix and its up/downstream intestinal compartments, as well as the pathophysiological changes in the early inflammatory phase.
## Rat Handling and In Situ Sampling
Totally, 12 four-week-old Sprague–Dawley (SD) rats were selected and reared adaptively in the laboratory for 1 week. An experimental study was set after allocating the rats randomly into two groups, the experiment group ($$n = 6$$) and the control group ($$n = 6$$). The rats in the experiment group were administered orally ceftriaxone sodium at 125 mg/ml dissolved in $0.9\%$ saline solution, 2 ml/day for 14 days to induce gastrointestinal dysbiosis [15]. The rats in the control group were administered orally $0.9\%$ saline solution, 2 ml/day for 14 days in parallel. Then, the rats were sacrificed, and intestinal samples were collected as described below. A 20 cm jejunum segment was sectioned 10 cm beneath the ligament of Treitz. The content was extruded into a 2 ml sterile Eppendorf tube. The appendix segment was cut at the tip of the caecum around the ileocecal junction, and the content was extruded into a sterile tube. A 20 cm colon segment was sectioned 2 cm apart from the end of the ileocecal valve, and the content was extruded into a sterile tube. All the intestinal contents were frozen at −80 °C. The jejunum, appendix and colon segment tissues were fixed in a solution of $10\%$ neutral formalin and $5\%$ glutaraldehyde and stored at 4 °C.
The animal study was reviewed and approved by the Biomedical Ethics Committee, Dalian Medical University. All experiments were performed in accordance with the relevant guidelines and regulations of our Ethical Committee.
## Morphology of Intestinal Segments and Structure of Mucosal Barrier
The gross morphology of the jejunum, appendix and colon and their contents were observed and compared between the two groups.
A 1 cm segment of the jejunum, appendix and colon tissues was embedded in paraffin, and then, the tissue blocks were sectioned at 4 μm. The sections were stained with haematoxylin and eosin and observed with an optical microscope (DP73, Olympus, Tokyo, Japan).
Tissue blocks of 1 mm × 3 mm from each segment were placed in precooled $2.5\%$ glutaraldehyde for 2 h, rinsed with 0.1 M phosphate buffer (pH 7.2) for 15–20 min (4 °C) three times, fixed with $1\%$ osmium tetroxide for 2 h (4 °C), rinsed with 0.1 M phosphate buffer for 5 min (4 °C) three times, washed with double-distilled water and dried. The blocks were soaked and embedded in methyloxirane and the embedded liquid and then sectioned into semithin section smears. The sections were stained, dried, dyed (toluidine blue), washed, cleared and sealed and observed under a transmission electron microscope (JEM-2000EX, JEOL, Tokyo, Japan). The experiments were performed in triplicate.
For evaluation of the severity of inflammation, five randomly selected fields in each section (magnification × 100) were inspected and graded by a pathologist blinded to the group allocation. Scores were generated according to the criteria of the modified scale of Bobin-Dubigeon et al. [ 16]. After the 5 fields were graded, the mean score was calculated for each section and is expressed as the histological score.
## DNA Isolation and 16S rRNA Gene Sequencing
An E.Z.N.A. Stool DNA kit (Omega Bio-Tek, Inc., Norcross, GA, United States) was used for whole DNA extraction from the stool. The DNA concentration was measured with Qubit 2.0 (Invitrogen, Carlsbad, CA, United States). Polymerase chain reaction (PCR) was applied to the bulk DNA with the barcoded primers 341F (5' -CCTAYGGGRBGCASCAG- 3') and 806R (5' -GGACTACNNGGGTATCTAAT-3'), which cover the V3–V4 regions of the 16S rRNA gene [17]. PCR reactions were performed on an ABI GeneAmp 9700 PCR system (Applied Biosystems, Foster City, CA, United States).
The 16S amplicons were purified with a GeneJET PCR Purification Kit (Thermo Fisher Scientific, Waltham, MA, United States). A DNA library was constructed by using the Ion Plus Fragment Library Kit 48 rxns (Thermo Fisher Scientific, Waltham, MA, United States). After the library was quantified with a Qubit fluorometer (Qubit 3.0, Invitrogen, Carlsbad, CA, United States) and qualified, it was sequenced by an Ion S5 XL system (Thermo Fisher Scientific, Waltham, MA, United States).
## Bioinformatic and Statistical Analysis
Quality filtering was performed on the raw reads to obtain high-quality clean reads. According to Cutadapt (v1.9.1) [18] (http://cutadapt.readthedocs.io/en/stable/), the reads were compared with the GOLD reference database (http://drive5.com/uchime/uchime_download.html) with the UCHIME algorithm (http://www.drive5.com/usearch/manual/uchime_algo.html) to detect and remove chimaeric sequences to obtain clean reads [19, 20].
Sequence analysis was performed with UPARSE software (Uparse v7.0.1001) (http://drive5.com/uparse/) [21]. Sequences with ≥ $97\%$ similarity were assigned to the same operational taxonomic units (OTUs). Representative sequences for each OTU were screened for further annotation. For each representative sequence, the SSU rRNA [22] database of Silva (http://www.arb-silva.de/) [23] was used based on the Mothur algorithm to annotate taxonomic information (set threshold from 0.8 to 1). For determination of the phylogenetic relationships of different OTUs and the difference in the dominant species in different samples (groups), multiple sequence alignments were conducted using MUSCLE (http://www.drive5.com/muscle/) Software (v3.8.31) [24]. OTUs abundance information was normalized using a standard sequence number corresponding to the sample with the fewest sequences. Subsequent analyses of alpha diversity and beta diversity were all performed based on these output normalized data.
Data are expressed as the mean ± standard error of the mean. Alpha diversity was applied to analyse the complexity of species diversity for a sample through 2 indices, observed species and Chao1 indices. Both of these indices in our samples were calculated with QIIME (Version 1.7.0). The Wilcox test in the agricolae package of R software (Version 2.15.3) was used to analyse the between-group difference in alpha diversity. Beta diversity was applied with Permutational multivariate analysis of variance (Adonis) analysis and the nonmetric multidimensional scaling (NMDS) analysis. NMDS analysis was based on Bray–Curtis dissimilarity and performed by the vegan software package of R software. The correlation between microbiome taxa and rosuvastatin effectiveness was assessed using linear discriminant analysis (LDA) effect size (LEfSe) at various taxonomic ranks [25]. An LDA score greater than 4.0 was defined as significant by default. LEfSe data were analysed using R software, and analysis of variance (ANOVA) was used to identify the relative abundance differences between groups. Tukey’s test was applied to perform post hoc tests, with $P \leq 0.05$ considered a significant difference. PICRUSt2 was performed using the OmicStudio Analysis (https://www.omicstudio.cn/analysis/) to predict the functional profiles of intestinal microbiome. T-test was used for analysing the OTU abundance from the same gut segment between the two groups OmicStudio tools (https://www.omicstudio.cn/tool) was utilized for statistical analyses and visualization of the identified pathways. R software was used for permutational multivariate analysis of variance (Adonis) to analyse the between-group differences in beta diversity. Group comparisons of histological scores were statistically analysed using independent-samples t-tests (SPSS 19.0). Statistical significance was accepted at $P \leq 0.05.$ Twenty-five appendicitis-associated taxa reported previously (Table 1), such as Actinobacteria, Proteobacteria, and Fusobacteria, were analysed from our samples with/without dysbiosis [13, 26–29].Table 1Appendicitis-associated taxa reported in previous studiesPhylumGenusSpeciesFirmicutesBacteroidetesActinobacteriaProteobacteriaFusobacteriaStreptococcusGemellaBacteroidesFaecalibacteriumProteusFusobacteriumRhizobiumPorphyromonasMogibacteriumPrevotellaBilophilaDialisterAnaerofilumBergeyellaPeptostreptococcusFusibacterParvimonasEscherichia coliBacteroides fragilisPorphyromonas endodontalis
## Intestinal Morphology and Mucosal Structure from the Appendix, Jejunum and Colon of the Rats w/o Dysbiosis
The gross morphology of the three intestinal segments was compared between the two groups: experiment group (referred to as ceftriaxone-treated group) and control group (referred to as placebo group). In the experiment group, the jejunum had no obvious changes (Fig. 1A); the appendices became enlarged and inflated with increasing, loose and light-coloured contents (Fig. 1B, D); the colon contents were also loose and light coloured (Fig. 1C). The morphological variations suggested abnormal absorption and/or exudation arose in the appendix and colon. Fig. 1The gross morphologies of jejunum, appendix and colon were compared between control and experiment group. ( A) *Experiment jejunum* had no obvious changes; (B, D) Experiment appendices became enlarged and inflated with loose and light-coloured contents; (C) Experiment colon contents were loose and light coloured (Color figure online) Light microscopy reveals the mucosal morphology of the jejunum (Fig. 2A), appendix (Fig. 2B) and colon (Fig. 2C) in both groups. In the control group, the colonic surface of the mucosa was smooth, and all mucosal epithelial cells were intact. In the experiment group, more than half of the disordered jejunal epithelium was incomplete, and a few epithelial cells fell into the intestinal cavity. The lumen of the ceftriaxone-treated appendix was narrower than that of the control appendix. The local lymph nodules proliferated in the antibiotic-treated appendix, and the adjacent fibrous connective tissue also proliferated. The colonic mucosal layer became thinner; the mucosal epithelium was incomplete, and the cells fell into the intestinal cavity; most of the glands disappeared; the vessels of the lamina propria and submucosa were dilatable and congested; and fibrous connective tissue obviously proliferated. In the rest of the lamina propria and submucosa, the blood vessels were dilated, and hyperaemia and oedema with slight hyperplasia of fibrous connective tissue were observed. The graphs of the histological scores illustrate the differences in the intestinal sections between the groups. Fig. 2Representative histological sections of (A) jejunum, (B) appendix and (C) colon under microscope (above: light microscopy, down: electron microscopy) and histological scores (*significant difference with $P \leq 0.05$). CJ control jejunum, CA control appendix, CC control colon, EJ experiment jejunum, EA experiment appendix, EC experiment colon Transmission electron microscopy reveals the epithelium ultrastructure of the jejunum, appendix and colon (Fig. 2). In the experiment group, the jejunum epithelial cells were oedematous, the mitochondria were swollen, and the microvilli were disorganized; the microvilli in the appendix became short, ruptured and dispersed; the colonic epithelial cells were oedematous, and the microvilli were sparse. All the above impairments indicated that the ceftriaxone treated appendix and the up/downstream intestinal compartments were in the inflammatory responses.
## Diversity and Abundance of Microbiome from the Jejunum, Appendix, and Colon of the Rats w/o Dysbiosis
Based on high-throughput sequencing, 16S rRNA sequence data were processed and analysed with bioinformatic and statistical packages. The numbers of OTUs represent species richness. The mean number of reads was 55576.19 (with two decimal places), and the range was 46836 to 66648.
In the control group, the microbiome OTUs numbers were 361±33, 634 ± 18 and 639 ± 19 in the jejunum, appendix and colon, respectively (Fig. 3A). The species richness, abundance and evenness were high in the appendix and colon compared to the jejunum (Fig. 3A, B). The number of common OTUs was 524 in the three segments (Fig. 3D), and the number of distinct OTUs was 308 in the jejunum, 79 in the appendix and 73 in the colon. The number of OTUs shared by the jejunum and appendix was 65, the number shared by the jejunum and colon was 52, and the number shared by the appendix and colon was 359. More OTUs were shared by the appendix and the colon. Fig. 3Metagenomic analysis of the gut microbiome of control and experiment group. ( A) observed OTUs; (B) Shannon diversity; (C) NMDS is analysed based on Bray–Curtis distance; (D) Venn diagram showing the number of OTUs specific and common to the three sites in the control group; (E) Venn diagram showing the number of OTUs specific and common to the three sites in the experiment group. CJ control jejunum; CA control appendix; CC control colon; EJ experiment jejunum; EA experiment appendix; EC experiment colon In the experiment group, the numbers of OTUs were 748 ± 98, 230 ± 11 and 253 ± 16 in the jejunum, appendix, and colon, respectively (Fig. 3A). The species richness, abundance and distribution evenness were high in the jejunum. The species diversity and evenness fell in the appendix and colon (Fig. 3A, B). The number of common OTUs in the jejunum, appendix and colon was 238 (Fig. 3E), and the number of distinct OTUs was 943, 34 and 33 in the jejunum, appendix and colon, respectively. The number of OTUs shared by the jejunum and appendix was 95, by the jejunum and colon was 119, and by the appendix and colon was 41. Adonis analysis (Table 2) and Non-metric multidimensional scaling (NMDS) analysis (Fig. 3C) showed that the microbiome structure of the jejunum was different from that of the appendix and colon. Similar microbial patterns were identified in the appendix and the colon in both groups. Table 2Adonis analyse of microbiome from jejunum, appendix and colonVs GroupR2P valueCJ–CA0.343110.001CJ–CC0.406680.001CA–CC0.08208NSEJ–EA0.35951NSEJ–EC0.325360.001389EA–EC0.03698NSCJ–EJ0.360040.012CA–EA0.490380.012CC–EC0.494750.001CJ control jejunum, CA control appendix, CC control colon, EJ experiment jejunum, EA experiment appendix, EC experiment colon, NS no significant
## Composition of the Microbiome from the Appendix, Jejunum and Colon of the Rats w/o Dysbiosis
The gut regional microbiome was classified at the phylum, family and genus levels. At each taxon level, the top ten most abundant bacteria were selected for comparative analysis (detailed data are shown in Table 3).Table 3Relative abundance of microbiome from jejunum, appendix, and colon at phylum, family and genus levelTaxonCJ (%)CA (%)CC (%)EJ (%)EA (%)EC (%)Phylum: Firmicutes80.37 ± 0.2873.90 ± 0.1156.87 ± 0.1450.05 ± 0.3979.46 ± 0.2283.40 ± 0.14Bacteroidetes10.85 ± 0.2522.09 ± 0.1340.23 ± 0.14#θ13.87 ± 0.110.23 ± 0.000.26 ± 0.00Tenericutes5.56 ± 0.080.27 ± 0.00β0.22 ± 0.00θ0.98 ± 0.01*'18.65 ± 0.2112.74 ± 0.12Cyanobacteria0.17 ± 0.00α0.19 ± 0.000.45 ± 0.009.64 ± 0.08*'0.02 ± 0.000.02 ± 0.00#'Actinobacteria1.78 ± 0.00α1.24 ± 0.010.64 ± 0.006.70 ± 0.05*'0.07 ± 0.000.07 ± 0.00#'Fusobacteria0.01 ± 0.00α0.00 ± 0.000.01 ± 0.001.06 ± 0.02*'0.00 ± 0.000.00 ± 0.00#'Proteobacteria1.15 ± 0.012.17 ± 0.011.52 ± 0.0115.76 ± 0.130.09 ± 0.000.11 ± 0.00Deinococcus-Thermus0.01 ± 0.00α0.00 ± 0.000.00 ± 0.000.56 ± 0.00*'0.01 ± 0.000.00 ± 0.00#'SR1 (Absconditabacteria)0.00 ± 0.00α0.00 ± 0.000.00 ± 0.000.29 ± 0.00*'0.00 ± 0.000.00 ± 0.00#'Euryarchaeota0.01 ± 0.00α0.00 ± 0.000.00 ± 0.000.16 ± 0.00*'0.00 ± 0.000.00 ± 0.00#'Family: Enterococcaceae0.28 ± 0.00α0.17 ± 0.000.15 ± 0.0035.14 ± 0.4816.84 ± 0.0721.13 ± 0.10Lactobacillaceae68.69 ± 0.28*α22.40 ± 0.1511.24 ± 0.07#2.13 ± 0.010.59 ± 0.001.60 ± 0.02Clostridiales_vadinBB60_group0.17 ± 0.000.37 ± 0.00β0.36 ± 0.00θ0.36 ± 0.00*'46.97 ± 0.2942.04 ± 0.28#'Bacteroidales_S24-7_group10.55 ± 0.2518.16 ± 0.1029.47 ± 0.04#θ2.26 ± 0.010.17 ± 0.000.17 ± 0.00Anaeroplasmatacee0.06 ± 0.000.09 ± 0.00β0.09 ± 0.00θ0.10 ± 0.00*'18.42 ± 0.2111.75 ± 0.12Lachnospiraceae0.51 ± 0.00*29.59 ± 0.09β23.53 ± 0.11#θ2.59 ± 0.010.21 ± 0.000.25 ± 0.00Erysipelotrichaceae6.83 ± 0.091.14 ± 0.011.41 ± 0.020.94 ± 0.0010.05 ± 0.1513.96 ± 0.22Prevotellaceae0.11 ± 0.003.42 ± 0.039.63 ± 0.13#3.36 ± 0.030.02 ± 0.000.03 ± 0.00Ruminococcaceae0.27 ± 0.00*18.49 ± 0.08β18.25 ± 0.06#θ1.78 ± 0.000.15 ± 0.000.19 ± 0.00Mycoplasmataceae5.50 ± 0.08*0.04 ± 0.000.04 ± 0.00#0.84 ± 0.010.23 ± 0.000.99 ± 0.00Genus: Enterococcus0.28 ± 0.000.17 ± 0.00β0.15 ± 0.0035.11 ± 0.4816.83 ± 0.0721.13 ± 0.10Lactobacillus68.69 ± 0.28*α22.40 ± 0.1511.24 ± 0.07#2.10 ± 0.010.59 ± 0.001.60 ± 0.02Anaeroplasma0.06 ± 0.000.09 ± 0.000.09 ± 0.00θ0.14 ± 0.00*'16.41 ± 0.2112.47 ± 0.12Erysipelotrichaceae_UCG-0040.07 ± 0.000.06 ± 0.000.05 ± 0.00θ0.08 ± 0.009.99 ± 0.1513.89 ± 0.22Prevotella_90.06 ± 0.001.29 ± 0.024.79 ± 0.090.44 ± 0.000.01 ± 0.000.02 ± 0.00Mycoplasma5.50 ± 0.08*α0.04 ± 0.000.04 ± 0.00#0.84 ± 0.010.23 ± 0.000.99 ± 0.00Turicibacter5.02 ± 0.080.27 ± 0.000.49 ± 0.000.27 ± 0.000.03 ± 0.000.04 ± 0.00unidentified_Chloroplast0.15 ± 0.000.02 ± 0.00β0.02 ± 0.009.57 ± 0.08*'0.02 ± 0.000.02 ± 0.00#'Bacteroides0.07 ± 0.000.23 ± 0.00β0.49 ± 0.006.56 ± 0.06*'0.01 ± 0.000.04 ± 0.00#'Lachnospiraceae_NK4A136_group0.07 ± 0.00*α7.13 ± 0.03β5.38 ± 0.03#θ0.39 ± 0.000.05 ± 0.000.06 ± 0.00Data are expressed as mean ± standard error of the meanCJ control jejunum, CA control appendix, CC control colon, EJ experiment jejunum, EA experiment appendix, EC experiment colonSignificant difference with $P \leq 0.05$ when comparing the microbiome between *CJ and CA; #CJ and CC; *' EJ and EA; #' EJ and EC; α CJ and EJ; β CA and EA; θ CC and EC At the phylum level, the composition of the microbiome was different among the three segments and between the two groups (Fig. 4A). In the control group, the relative abundance of most phyla decreased down from the jejunum to the colon. Firmicutes and Bacteroidetes were the main constituent microbiomes from all segments. Firmicutes showed a declining relative abundance ($80.37\%$ ± $0.28\%$, $73.90\%$ ± $0.11\%$ and $56.87\%$ ± $0.14\%$) from the jejunum down to the colon. Bacteroidetes showed an increasing relative abundance ($10.85\%$ ± $0.25\%$, $22.09\%$ ± $0.13\%$ and $40.23\%$ ± $0.14\%$) along the intestinal tract. Fusobacteria, Deinococcus-Thermus, and Euryarchaeota were barely detected (detailed data shown in Table 3). In the experiment group, the Firmicutes and Bacteroidetes were opposite to those in the control group. Firmicutes showed an increasing relative abundance ($50.05\%$ ± $0.39\%$, $79.46\%$ ± $0.22\%$ and $83.40\%$ ± $0.14\%$) from the jejunum down to the colon. Bacteroidetes showed a declining relative abundance ($13.87\%$ ± $0.11\%$, $0.23\%$ and $0.26\%$). Tenericutes also varied compared to that in the control group, with the relative abundance being high in the appendix and colon but low in the jejunum. Proteobacteria, Cyanobacteria and Actinobacteria, like Bacteroidetes, were highly abundant in the jejunum but almost undetectable in the appendix and colon. Other phyla were detected in a low relative abundance from the three intestinal segments (detailed data shown in Table 3).Fig. 4Predominant intestinal microbiome and their relative abundance. ( A) top 10 microbiome at phylum level; (B) top microbiome 10 at family level; (C) top 30 microbiome at genus level. CJ control jejunum; CA control appendix; CC control colon; EJ experiment jejunum; EA experiment appendix; EC experiment colon At the family level, the relative abundances of the microbiome were compared between the segments and the groups (Fig. 4B). In the control group, Lactobacillaceae was the main constituent flora from the jejunum to the colon at relative abundances of $68.69\%$ ± $0.28\%$, $22.40\%$ ± $0.15\%$ and $11.24\%$ ± $0.07\%$. Enterococcaceae accounted for a very small proportion of the gut microbiota ($0.28\%$ ± $0.00\%$, $0.17\%$ ± $0.00\%$, $0.15\%$ ± $0.00\%$). Other families showed higher relative abundances in the appendix and colon than in the jejunum (detailed data shown in Table 3). In the experiment group, Enterococcaceae increased in the jejunum, appendix and colon ($35.14\%$ ± $0.48\%$, $16.84\%$ ± $0.07\%$ and $21.13\%$ ± $0.10\%$), whereas Lactobacillaceae decreased in all three segments ($2.13\%$ ± $0.01\%$, $0.59\%$ ± $0.00\%$ and $1.60\%$ ± $0.02\%$). A dramatic increase of Clostridiales_vadinBB60_group was identified in the appendix and colon ($46.97\%$ ± $0.29\%$ and $42.04\%$ ± $0.28\%$). Erysipelotrichaceae and Mycoplasmataceae were abundant in the jejunum ($6.83\%$ ± $0.09\%$, $5.50\%$ ± $0.08\%$) but scant in the appendix ($1.14\%$ ± $0.01\%$, $0.04\%$ ± $0.00\%$) and the colon ($1.41\%$ ± $0.02\%$, $0.04\%$ ± $0.00\%$) despite of dysbiosis. Anaeroplasmataceae and Erysipelotrichaceae increased in the appendix and colon. Prevotellaceae, Lachnospiraceae, Bacteroidales_S24-7_group and Ruminococcaceae were found only in the jejunum at low relative abundances (detailed data shown in Table 3).
At the genus level, statistical analysis was performed to compare the top 30 genera within and between the groups (Fig. 4C). In the control group, the top 30 genera accounted for over $85\%$ of the bacteria in the jejunum, $52\%$ in the appendix and $44\%$ in the colon. Lactobacillus was the dominant genus in all segments ($68.69\%$ ± $0.28\%$, $22.40\%$ ± $0.15\%$ and $11.24\%$ ± $0.07\%$). Enterococcus was rare ($0.28\%$ ± $0.00\%$, $0.17\%$ ± $0.00\%$ and $0.15\%$ ± $0.00\%$). The majority of these genera, except Lachnospiraceae_NK4A136_group and Prevotella_9, showed higher relative abundances in the jejunum than in the appendix and colon (detailed data shown in Table 3). In the experiment group, Enterococcus appeared to be the main genus from the jejunum to the colon ($35.11\%$ ± $0.48\%$, $16.83\%$ ± $0.07\%$ and $21.13\%$ ± $0.10\%$), while Lactobacillus reduced dramatically in the disordered guts ($2.10\%$ ± $0.01\%$, $0.59\%$ ± $0.00\%$ and $1.60\%$ ± $0.02\%$). Bacteroides and unidentified_Chloroplast were detected in the jejunum at relative abundances of $6.56\%$ ± $0.06\%$ and $9.57\%$ ± $0.08\%$ and were almost undetectable in the appendix and colon. Anaeroplasma and Erysipelotrichaceae_UCG-004 were barely detected in the jejunum but were highly detected in the appendix ($16.41\%$ ± $0.21\%$, $9.99\%$ ± $0.15\%$) and the colon ($12.47\%$ ± $0.12\%$, $13.89\%$ ± $0.22\%$) (detailed data shown in Table 3).
## Intergroup/Intragroup Analysis of Microbial Clusters from Different Intestinal Sites
To compare the microbial clusters from the different intestinal sites within and between groups, we used LEfSe software to determine the metagenomic differences of the OTUs derived from 16S rRNA sequences from the jejunum, appendix and colon. In the control group, a total of 28 microbial clusters showed site-specific abundances (LDA value > 4.0), including 11 in the jejunum, eight in the appendix, and nine in the colon (Fig. 5A1). In the experiment group, eight microbial clusters showed site-specific abundances (LDA value > 4.0), including seven in the jejunum and one in the colon (Fig. 5A2). No site-specific clusters were found in the disordered appendix. Comparing the two groups, we found 12 different bacterial clusters in the jejunum (Fig. 5A3), 26 different clusters in the appendix (Fig. 5A4) and 24 different clusters in the colon (Fig. 5A5). Of the jejunum, Enterococcus and Lactobacillus varied greatly between the healthy and disordered jejunum. Of the appendix and the colon, Lachnospiraceae, Bacteroidales_S24_7_group, and Clostridiales_vadinBB60_group varied greatly between the two groups. PICRUSt2 analysis was used to predict metagenomic functions associated with bacterial communities based on 16S rRNA sequencing data. At KEGG levels 3, there are 56 different functional pathways found in the jejunum, and on 100 and 105 functional pathways found in the appendix and colon between the two groups. Figure 5B shows the 20 functional pathways with the highest significant difference in the jejunum, appendix, and colon between the two groups. Fig. 5LEfSe analysis (A) and functional profiles (B) of intestinal microbiome. ( A1) LDA scores of the segmental microbiome of control group; (A2) LDA scores of the segmental microbiome of experiment group; (A3) LDA scores of jejunum microbiome between the two groups; (A4) LDA scores of appendix microbiome between the two groups; (A5) LDA scores of colon microbiome between the two groups (LDA score threshold set at 4); (B1) 20 most different functional pathways between CJ and EJ; (B2) 20 most different functional pathways between CA and EA; (B3) 20 most different functional pathways between CC and EC. CJ control jejunum; CA control appendix; CC control colon; EJ experiment jejunum; EA experiment appendix; EC experiment colon
## Appendicitis-Associated Taxa Identified from the Jejunum, Appendix and Colon of the Rats w/o Dysbiosis
Former researchers reported that 25 bacterial taxa, five phyla, 17 genus and three species were commonly associated with appendicitis (Table 1) [13, 26–29]. We analysed these appendicitis-associated taxa in our samples and found that some microbes had special preferences for the different intestinal sites. At the phylum level, the relative abundance of Actinobacteria was $1.78\%$ ± $0.00\%$, $1.24\%$ ± $0.01\%$ and $0.64\%$ ± $0.00\%$ in the control jejunum, appendix and colon, respectively, but changed to $6.70\%$ ± $0.05\%$, $0.07\%$ ± $0.00\%$ and $0.07\%$ ± $0.00\%$ in the experiment jejunum, appendix and colon, respectively. The relative abundances of Proteobacteria were $1.15\%$ ± $0.01\%$, $2.17\%$ ± $0.01\%$ and $1.52\%$ ± $0.01\%$ from the control jejunum, appendix and colon, respectively, but changed to $15.76\%$ ± $0.13\%$, $0.09\%$ ± $0.00\%$ and $0.11\%$ ± $0.00\%$ in the experiment group. Fusobacteria was barely detected in the control group but increased to $1.06\%$ ± $0.02\%$ in the experiment jejunum. At the genus level, the relative abundances of Fusobacterium, Streptococcus, Porphyromonas and Proteus were increased in the experiment jejunum. Other genus showed no significant change in the experiment group. At the species level, E. coli and B. fragilis increased in the experiment jejunum.
## Discussion
Trillions of microorganisms inhabit the gastrointestinal tract of complex multicellular animals and humans. They play a vital role in dietary metabolism and digestive system health. Due to the development of high-throughput sequencing technology, the impact of gut flora on human health and disease has been explored in recent years. The appendix is an organ of the digestive tract where bulk of commensals inhabit. Some studies have revealed that appendicitis is precipitated by luminal obstruction and subsequent microbial overgrowth [30]. The present study investigated the in situ microbiome and mucosal morphology of rat jejunum, appendix and colon with/without antibiotic-induced dysbiosis. We intended to gain an insight into the ecology of appendix and its up/downstream intestine as well as the protective role of appendix to gut microbial community.
In this experiment, a rodent model of antibiotic-induced dysbiosis was established in order to imitate human dysbiosis and to obtain gut microbiome in situ. Ceftriaxone is a third-generation cephalosporin with a wide spectrum of activity against gram-negative bacilli and most gram-positive bacteria. The present study shows that oral antibiotic administration changes the microbial community and mucosal morphology of the appendix and the up/downstream intestinal compartments. Our metagenomic data revealed that the species richness and evenness were higher in the jejunum than in the appendix and colon of dysbiosis. The reason for this may be that the peristalsis speed of the jejunum is faster than that of the appendix and colon. Ceftriaxone remained longer in the large intestine than the other organs so that it enhanced the inhibitory effect on the gut microbiota. On the other hand, some of the flora that are sensitive to prolonged antibiotic effects might shift retrogradely into the upper intestine. For example, Bacteroides translocated from the colon to the jejunum during dysbiosis. Other bacteria that rely on the end products from these bacteria possibly migrate together to ingest carbon and nitrogen sources [31]. Thus, the community diversity increases in the upper bowel during dysbiosis. Another interesting finding is that the appendix and colon share similar microbiome pattern regardless of dysbiosis. This reflects the similar ecosystem of appendix and colon despite of the challenges from antibiotics or inflammation.
Firmicutes and Bacteroidetes were recovered to be the most dominant phyla among the gut microflora in this study. The relative abundance of Firmicutes was decreasing from the upper to the lower intestine of the healthy rats, but this tendency reversed in the disordered rats. Nevertheless, the relative abundance of Firmicutes remained stable in the appendix. Firmicutes is a phylum of bacteria, most of which have gram-positive cell wall structures. More than 274 genera were assigned to the Firmicutes phylum. *Notable* genera of Firmicutes are Enterococcus and Lactobacillus etc. The main function of *Firmicutes is* to hydrolyse carbohydrates and proteins in the intestine [32]. A sufficient level of Firmicutes in the healthy jejunum complies with the function of the small intestine, where most chemical digestion of carbohydrates, proteins and fats occurs. When the intestinal flora was disrupted, the amount of Firmicutes dropped in the jejunum and increased in the colon. Bacteroidetes distributed inversely to Firmicutes in the guts of both groups. Bacteroidetes are primary colonizers of the colon that involve in the metabolism of steroids, polysaccharides and bile acids, as well as polysaccharide utilization and protein synthesis [33]. A previous work found the same variation in microbiota among obese children that Firmicutes increased and Bacteroidetes decreased in the colon [34]. Schade demonstrated that the cell wall glycopolymers of Firmicutes could influence host microbe interactions through the modulation of bacterial colonization [35].
Lactobacillaceae and Enterococcaceae are two families under the phylum Firmicutes. In the healthy jejunum, Lactobacillaceae abundance was high, while the relative abundance of Enterococcaceae was extremely low, less than $0.3\%$. The jejunum favours the survival and reproduction of Lactobacillaceae because the pH value is around 6.1 in the proximal jejunum and 7.3 in the terminal ileum [36]. When the intestinal flora was disrupted, Enterococcaceae became predominant in the jejunum, and Lactobacillaceae decreased significantly. It is likely that Enterococcaceae and Lactobacillaceae react adversely to each other. Lactobacillaceae is a group of gram-positive, facultative anaerobes that can convert sugars to lactic acid. They are the predominant gut microbiota and the most common probiotics added to foods, and it has antimicrobial potential against pathogens [37].
To compare the in situ microbial composition, we used LEfSe analysis and species abundance clustering to estimate the beta diversity of the bacterial OTUs. In a normal intestinal ecosystem, the distribution of microbiota reflects the tissue tropism of different intestinal parts. Lachnospiraceae was found abundant in the control appendix, whose species showed a negative correlation with the development of obesity and diabetes [38]. Ruminococcaceae, Desulfovibrionaceae and Coriobacteriaceae were also abundant in the control appendix. Ruminococcaceae is the member of commensal bacteria of the caecum and colon [39], which can degrade various polysaccharides and fibres [40]. Desulfovibrionaceae is a group of sulphate-reducing bacteria that can use sulphate as a terminal electron acceptor to form hydrogen sulphide. Hydrogen sulphide (H2S) serves as a gasotransmitter in the maintenance of tissues homeostasis. It also has versatile effects in vasodilation, neuromodulation and anti-inflammation [41, 42]. Coriobacteriaceae plays a key role in the succession of gut microbial consortia in early life in humans [43]. The present study found that these beneficial bacteria were abundant in the appendix, indicating that the appendix may serve as a commensal pool that preserves a mass of probiotic bacteria to maintain the balance of the intestinal ecosystem. When the intestine was disordered, the flora of the jejunum became diverse, and the number of site-specific clusters increased. However, site-specific bacterial clusters were not identified in the disordered appendix. This finding is possibly due to the protective role of appendix biofilms with overwhelming commensal bacteria. The biofilm forms on the layer of mucus that covers the intestinal epithelium. The mucus consists of mucins rich in fucose, galactose, sialic acid, N-acetylgalactosamine, N-acetylglucosamine and mannose, some of which are produced by *Bacillus mesentericus* TO-A. The bacteria in the biofilm produce various producer-derived glycoside hydrolases to establish a metabolic interaction network that favours the growth of organisms that need carbon sources [31].
Former workers proposed a number of bacterial taxa in association with appendicitis in literature [13, 26–29]. We examined the distribution of these appendicitis-associated taxa in the jejunum, appendix and colon, and found that most appendicitis-associated taxa were detected in low level in our samples of both groups, except Firmicutes and Bacteroidetes. Salo et al. recovered Firmicutes in high level in the appendix of the patients with appendicitis, which agrees with our results. Toon Peeters et al. reported that the richness and diversity within the phyla Firmicutes, Actinobacteria, Fusobacteria and Verrucomicrobia were lower in faeces from those with appendicitis than in normal faeces [14]. Zhong D et al. recovered Bacteroides in high level in the normal appendix and the appendix with gangrenous appendicitis and they suggested the abundance of Bacteroides was inversely related to the degree of inflammation. Some scholars demonstrated that bacteria could migrate along the intestinal tract during appendicitis [44, 45]. Combining the findings from our study and from the others, we assume that the Bacteroides translocate retrogradely from the appendix and colon to the jejunum after antibiotic disruption. Bacteroides displacement might be both an outcome and a risk factor for inflammation progression to the appendix. With multicompartment profiling of the appendix-associated bacteria, we find no correlation between the appendicitis-associated taxa and the rat disordered appendix. Researchers have suggested that the excessive growth of appendiceal bacteria is a consequence of appendix obstruction and inflammation [12, 46], for example, Fusobacterium was present in most appendicitis specimens [47], and the number of species increased in parallel with the severity of the disease. In the present study, Fusobacterium, which was not present in the healthy gut, emerged in the dysbiosis jejunum. However, one should keep in mind that the appendix-associated taxa have been selected from the human with the onset of appendicitis.
Our antibiotic-induced dysbiosis model reveals that the ecological imbalance of the jejunum, appendix and colon likely link to gut inflammation and impairment. We found that the mucosal morphology changed apparently in the jejunum and colon but slightly in the appendix in the early stage of appendiceal inflammation. This result might be due to the protective role of the appendix. On the one hand, masses of lymphoid tissues are present in the appendix. On the other hand, appendix biofilms preserve high levels of gut probiotics to maintain community equilibrium. There are several ways of maintaining the integrity of the epithelial barrier and host microbial homeostasis in the gut. Intestinal goblet cells secrete mucin proteins acting as protective coatings that provide structural integrity and regulate macrophage and adaptive T cell responses during inflammatory processes [48]. Gastrointestinal dysfunction is associated with defects in the intestinal mucous lining caused by biotic or abiotic factors. This study characterized the gross morphology, mucosal morphology and cell structure of intestinal segments with light microscopy and electron transmission microscopy. Microvilli, brush-like edges, are the free surface of small intestinal epithelial cells. They are small finger-like protrusions that protrude from fibres in the cell membrane and cytoplasm. The function of microvilli is to expand the surface area of the free cells to enhance nutrient absorption. Microscopy revealed that the jejunal villus epithelial cells were impaired, the brush border was unclear, and mitochondria became swollen from the experiment group. Moreover, the abundance of microbiota such as Betaproteobacteria, Pseudonocardiaceae, Enterococcaceae, Prevotellaceae and Enterococcus durans changed in the experiment jejunum. We observed an inflated appendix with increasing fluids in the dysbiosis rats. Microscopy revealed that the appendix microvilli became sparse and broken. The local lymph nodules of the experiment appendix proliferated obviously, and the adjacent fibrous connective tissue also proliferated. These mucosal defects implicated that the appendix proceeded to early appendiceal inflammation. The role of dysbiotic microbiota in the development of appendiceal inflammation needs to be explored further. In the experiment colon, the epithelial cells were swollen and oedematous, and the microvilli were sparse. The microbiota richness and abundance reduced in the colon of the experiment animals. Microbiota imbalance or even loss of the colonial microflora leads to the exposure of microvilli to ceftriaxone sodium, which leads to the shrinkage of the villi, a reduction in absorption ability, and the accumulation of harmful substances. Microbiome functional profiling reveals the 20 functional pathways with the highest significant difference in the jejunum, appendix, and colon between the two groups. Most of these functional pathways were enriched in the control group regardless of the sites. The experiment appendix and colon shared two identical pathways, tetracycline biosynthesis and valine, leucine, and isoleucine degradation. Oxidative phosphorylation, which was exclusively enriched in the disordered appendix, is a metabolic pathway by which bacteria require oxygen to produce energy for cell survival. Arachidonic acid metabolism pathway, which was enriched in the disordered jejunum, has been indicated in association with inflammatory response [49].
To date, there has been a steady rise in studies of human microecology in association with health and disease. The link between gut intestinal dysbiosis and necrotizing enterocolitis has drawn great attention of researchers. The gut microbiota contributes to ecosystem equilibrium by increasing metabolic capacity, preventing the colonization of pathogenic organisms, providing vitamins and regulating the host immune system in human health [50]. Antibiotics have a comprehensive impact on the microbial community, which not only disrupts microbial survival but also alters the nutritional and pathophysiological status of the intestine.
## Conclusion
The present study investigated the microbiota composition, diversity and relative abundance, as well as mucosal morphology from the different intestinal segments of rats with/without antibiotic interruption. The microbiome species richness and evenness were high in the healthy appendix and colon but low in the disordered appendix and colon; the microbiome structure and relative abundance varied from the site to site inter- and intragroup; the appendix and colon shared similar microbiome patterns regardless of dysbiosis; and site-specific bacteria were missing in the disordered appendices. We find that appendix is likely a transition region involving in upper and lower intestinal microflora modulation. The present study reveals the in situ variation of gut microbiota under dysbiosis, which lays the foundation for exploring the microbial aetiology of diseases. However, we should be cautious about translating the microbiome profiling results from rat to human. Further research is needed on the human gut microbiota in the different stages of dysbiosis, as well as therapeutic effects of potential probiotics.
## References
1. Tropini C, Earle KA, Huang KC, Sonnenburg JL. **The gut microbiome: connecting spatial organization to function**. *Cell Host Microbe* (2017) **21** 433-442. DOI: 10.1016/j.chom.2017.03.010
2. 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
3. Vyas U, Ranganathan N. **Probiotics, prebiotics, and synbiotics: gut and beyond**. *Gastroenterol Res Pract* (2012). DOI: 10.1155/2012/872716
4. Cummings JH, Pomare EW, Branch WJ, Naylor CP, Macfarlane GT. **Short chain fatty acids in human large intestine, portal, hepatic and venous blood**. *Gut* (1987) **28** 1221-1227. DOI: 10.1136/gut.28.10.1221
5. Guarner F, Malagelada JR. **Gut flora in health and disease**. *Lancet* (2003) **361** 512-519. DOI: 10.1016/s0140-6736(03)12489-0
6. Gomaa EZ. **Human gut microbiota/microbiome in health and diseases: a review**. *Antonie Van Leeuwenhoek* (2020) **113** 2019-2040. DOI: 10.1007/s10482-020-01474-7
7. Bollinger R, Barbas AS, Bush EL, Lin SS, Parker W. **Biofilms in the large bowel suggest an apparent function of the human vermiform appendix**. *J Theor Biol* (2007) **249** 826-831. DOI: 10.1016/j.jtbi.2007.08.032
8. Laurin M, Everett ML, Parker W. **The cecal appendix: one more immune component with a function disturbed by post-industrial culture**. *Anat Rec (Hoboken)* (2011) **294** 567-579. DOI: 10.1002/ar.21357
9. Guillet-Caruba C, Cheikhelard A, Guillet M, Bille E, Descamps P, Yin L, Khen-Dunlop N, Zahar JR, Sarnacki S, Revillon Y. **Bacteriologic epidemiology and empirical treatment of pediatric complicated appendicitis**. *Diagn Microbiol Infect Dis* (2011) **69** 376-381. DOI: 10.1016/j.diagmicrobio.2010.11.003
10. Chen CY, Chen YC, Pu HN, Tsai CH, Chen WT, Lin CH. **Bacteriology of acute appendicitis and its implication for the use of prophylactic antibiotics**. *Surg Infect (Larchmt)* (2012) **13** 383-390. DOI: 10.1089/sur.2011.135
11. Roberts JP. **Quantitative bacterial flora of acute appendicitis**. *Arch Dis Child* (1988) **63** 536-540. DOI: 10.1136/adc.63.5.536
12. Guinane CM, Tadrous A, Fouhy F, Ryan CA, Dempsey EM, Murphy B, Andrews E, Cotter PD, Stanton C, Ross RP. **Microbial composition of human appendices from patients following appendectomy**. *MBio* (2013). DOI: 10.1128/mBio.00366-12
13. Zhong D, Brower-Sinning R, Firek B, Morowitz MJ. **Acute appendicitis in children is associated with an abundance of bacteria from the phylum Fusobacteria**. *J Pediatr Surg* (2014) **49** 441-446. DOI: 10.1016/j.jpedsurg.2013.06.026
14. Peeters T, Penders J, Smeekens SP, Galazzo G, Houben B, Netea MG, Savelkoul PH, Gyssens IC. **The fecal and mucosal microbiome in acute appendicitis patients: an observational study**. *Future Microbiol* (2019) **14** 111-127. DOI: 10.2217/fmb-2018-0203
15. Yin J, Prabhakar M, Wang S, Liao SX, Peng X, He Y, Chen YR, Shen HF, Su J, Chen Y. **Different dynamic patterns of beta-lactams, quinolones, glycopeptides and macrolides on mouse gut microbial diversity**. *PLoS One* (2015) **10** e0126712. DOI: 10.1371/journal.pone.0126712
16. Bobin-Dubigeon C, Collin X, Grimaud N, Robert JM, Le Baut G, Petit JY. **Effects of tumour necrosis factor-alpha synthesis inhibitors on rat trinitrobenzene sulphonic acid-induced chronic colitis**. *Eur J Pharmacol* (2001) **431** 103-110. DOI: 10.1016/s0014-2999(01)01410-8
17. Sundberg C, Al-Soud WA, Larsson M, Alm E, Yekta SS, Svensson BH, Sorensen SJ, Karlsson A. **454 pyrosequencing analyses of bacterial and archaeal richness in 21 full-scale biogas digesters**. *FEMS Microbiol Ecol* (2013) **85** 612-626. DOI: 10.1111/1574-6941.12148
18. Martin M. **CUTADAPT removes adapter sequences from high-throughput sequencing reads**. *EMBnetjournal* (2011). DOI: 10.14806/ej.17.1.200
19. Haas BJ, Gevers D, Earl AM, Feldgarden M, Ward DV, Giannoukos G, Ciulla D, Tabbaa D, Highlander SK, Sodergren E. **Chimeric 16S rRNA sequence formation and detection in Sanger and 454-pyrosequenced PCR amplicons**. *Genome Res* (2011) **21** 494-504. DOI: 10.1101/gr.112730.110
20. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. **UCHIME improves sensitivity and speed of chimera detection**. *Bioinformatics* (2011) **27** 2194-2200. DOI: 10.1093/bioinformatics/btr381
21. Edgar RC. **UPARSE: highly accurate OTU sequences from microbial amplicon reads**. *Nat Methods* (2013) **10** 996-998. DOI: 10.1038/nmeth.2604
22. Wang Q, Garrity GM, Tiedje JM, Cole JR. **Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy**. *Appl Environ Microbiol* (2007) **73** 5261-5267. DOI: 10.1128/aem.00062-07
23. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, Peplies J, Glockner FO. **The SILVA ribosomal RNA gene database project: improved data processing and web-based tools**. *Nucleic Acids Res* (2013) **41** D590-596. DOI: 10.1093/nar/gks1219
24. Edgar RC. **MUSCLE: multiple sequence alignment with high accuracy and high throughput**. *Nucleic Acids Res* (2004) **32** 1792-1797. DOI: 10.1093/nar/gkh340
25. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, Huttenhower C. **Metagenomic biomarker discovery and explanation**. *Genome Biol* (2011) **12** R60. DOI: 10.1186/gb-2011-12-6-r60
26. Salo M, Marungruang N, Roth B, Sundberg T, Stenstrom P, Arnbjornsson E, Fak F, Ohlsson B. **Evaluation of the microbiome in children's appendicitis**. *Int J Colorectal Dis* (2017) **32** 19-28. DOI: 10.1007/s00384-016-2639-x
27. Wagner M, Tubre DJ, Asensio JA. **Evolution and current trends in the management of acute appendicitis**. *Surg Clin North Am* (2018) **98** 1005-1023. DOI: 10.1016/j.suc.2018.05.006
28. Schulin S, Schlichting N, Blod C, Opitz S, Suttkus A, Stingu CS, Barry K, Lacher M, Buhligen U, Mayer S. **The intra- and extraluminal appendiceal microbiome in pediatric patients: A comparative study**. *Medicine (Baltimore)* (2017) **96** e9518. DOI: 10.1097/md.0000000000009518
29. Alkadhi S, Kunde D, Cheluvappa R, Randall-Demllo S, Eri R. **The murine appendiceal microbiome is altered in spontaneous colitis and its pathological progression**. *Gut Pathog* (2014) **6** 25. DOI: 10.1186/1757-4749-6-25
30. Lamps LW. **Infectious causes of appendicitis**. *Infect Dis Clin North Am* (2010) **24** 995-1018. DOI: 10.1016/j.idc.2010.07.012
31. Rakoff-Nahoum S, Coyne MJ, Comstock LE. **An ecological network of polysaccharide utilization among human intestinal symbionts**. *Curr Biol* (2014) **24** 40-49. DOI: 10.1016/j.cub.2013.10.077
32. 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
33. Xu J, Bjursell MK, Himrod J, Deng S, Carmichael LK, Chiang HC, Hooper LV, Gordon JI. **A genomic view of the human-Bacteroides thetaiotaomicron symbiosis**. *Science* (2003) **299** 2074-2076. DOI: 10.1126/science.1080029
34. Riva A, Borgo F, Lassandro C, Verduci E, Morace G, Borghi E, Berry D. **Pediatric obesity is associated with an altered gut microbiota and discordant shifts in Firmicutes populations**. *Environ Microbiol* (2017) **19** 95-105. DOI: 10.1111/1462-2920.13463
35. Schade J, Weidenmaier C. **Cell wall glycopolymers of Firmicutes and their role as nonprotein adhesins**. *FEBS Lett* (2016) **590** 3758-3771. DOI: 10.1002/1873-3468.12288
36. Lucas M. **Determination of acid surface pH in vivo in rat proximal jejunum**. *Gut* (1983) **24** 734-739. DOI: 10.1136/gut.24.8.734
37. Al Kassaa I, Hober D, Hamze M, Chihib NE, Drider D. **Antiviral potential of lactic acid bacteria and their bacteriocins**. *Probiotics Antimicrob Proteins* (2014) **6** 177-185. DOI: 10.1007/s12602-014-9162-6
38. Kameyama K, Itoh K. **Intestinal colonization by a Lachnospiraceae bacterium contributes to the development of diabetes in obese mice**. *Microbes Environ* (2014) **29** 427-430. DOI: 10.1264/jsme2.ME14054
39. Donaldson GP, Lee SM, Mazmanian SK. **Gut biogeography of the bacterial microbiota**. *Nat Rev Microbiol* (2016) **14** 20-32. DOI: 10.1038/nrmicro3552
40. Shang Q, Shan X, Cai C, Hao J, Li G, Yu G. **Dietary fucoidan modulates the gut microbiota in mice by increasing the abundance of Lactobacillus and Ruminococcaceae**. *Food Funct* (2016) **7** 3224-3232. DOI: 10.1039/c6fo00309e
41. Zhang-Sun W, Augusto LA, Zhao L, Caroff M. **Desulfovibrio desulfuricans isolates from the gut of a single individual: structural and biological lipid A characterization**. *FEBS Lett* (2015) **589** 165-171. DOI: 10.1016/j.febslet.2014.11.042
42. Magierowski M, Jasnos K, Kwiecien S, Drozdowicz D, Surmiak M, Strzalka M, Ptak-Belowska A, Wallace JL, Brzozowski T. **Endogenous prostaglandins and afferent sensory nerves in gastroprotective effect of hydrogen sulfide against stress-induced gastric lesions**. *PLoS One* (2015) **10** e0118972. DOI: 10.1371/journal.pone.0118972
43. Harmsen HJ, Wildeboer-Veloo AC, Grijpstra J, Knol J, Degener JE, Welling GW. **Development of 16S rRNA-based probes for the Coriobacterium group and the Atopobium cluster and their application for enumeration of Coriobacteriaceae in human feces from volunteers of different age groups**. *Appl Environ Microbiol* (2000) **66** 4523-4527. DOI: 10.1128/aem.66.10.4523-4527.2000
44. Coldewey SM, Hartmann M, Schmidt DS, Engelking U, Ukena SN, Gunzer F. **Impact of the rpoS genotype for acid resistance patterns of pathogenic and probiotic Escherichia coli**. *BMC Microbiol* (2007) **7** 21. DOI: 10.1186/1471-2180-7-21
45. Blod C, Schlichting N, Schulin S, Suttkus A, Peukert N, Stingu CS, Hirsch C, Elger W, Lacher M, Buhligen U. **The oral microbiome-the relevant reservoir for acute pediatric appendicitis?**. *Int J Colorectal Dis* (2018) **33** 209-218. DOI: 10.1007/s00384-017-2948-8
46. Wangensteen OH, Dennis C. **Experimental proof of the obstructive origin of appendicitis in man**. *Ann Surg* (1939) **110** 629-647. DOI: 10.1097/00000658-193910000-00011
47. Swidsinski A, Dorffel Y, Loening-Baucke V, Theissig F, Ruckert JC, Ismail M, Rau WA, Gaschler D, Weizenegger M, Kuhn S. **Acute appendicitis is characterised by local invasion with Fusobacterium nucleatum/necrophorum**. *Gut* (2011) **60** 34-40. DOI: 10.1136/gut.2009.191320
48. Robinson K, Deng Z, Hou Y, Zhang G. **Regulation of the intestinal barrier function by host defense peptides**. *Front Vet Sci* (2015) **2** 57. DOI: 10.3389/fvets.2015.00057
49. Powell WS, Rokach J. **2015 Biosynthesis, biological effects, and receptors of hydroxyeicosatetraenoic acids (HETEs) and oxoeicosatetraenoic acids (oxo-ETEs) derived from arachidonic acid**. *Biochim Biophys Acta* (1851) **4** 340-355. DOI: 10.1016/j.bbalip.2014.10.008
50. Stein RR, Bucci V, Toussaint NC, Buffie CG, Ratsch G, Pamer EG, Sander C, Xavier JB. **Ecological modeling from time-series inference: insight into dynamics and stability of intestinal microbiota**. *PLoS Comput Biol* (2013) **9** e1003388. DOI: 10.1371/journal.pcbi.1003388
|
---
title: Influence of fat-free mass index on the survival of patients with head and
neck cancer
authors:
- Nina Lapornik
- Brigita Avramovič Brumen
- Gaber Plavc
- Primož Strojan
- Nada Rotovnik Kozjek
journal: European Archives of Oto-Rhino-Laryngology
year: 2022
pmcid: PMC9988755
doi: 10.1007/s00405-022-07732-w
license: CC BY 4.0
---
# Influence of fat-free mass index on the survival of patients with head and neck cancer
## Abstract
### Purpose
To determine whether muscle mass, defined by fat-free mass index (FFMI) measured with bioelectrical impedance analysis (BIA), is predictive of survival of head and neck squamous cell carcinoma (HNSCC) patients.
### Methods
HNSCC patients treated between 2014 and 2018 at the Department for Nutrition of the Institute of Oncology Ljubljana were reviewed. The FFMI values from the pretreatment BIA measurements and pretreatment body mass index (BMI) were used to categorize patients into groups with low and normal muscle mass and BMI using the Global Leadership Initiative on malnutrition (GLIM) recommended cutoff values. The impact of FFMI on disease-free survival (DFS) and overall survival (OS) was determined.
### Results
Of the 71 included patients, 31 ($43.7\%$) had normal FFMI, and 40 ($56.3\%$) had low FFMI, whereas 44 ($62\%$) and 27 ($38\%$) of the patients had normal and low BMI, respectively. Between FFMI and BMI values, a significant correlation was found (RP = 0.75, $p \leq 0.001$). Univariate regression analysis showed that FFMI (as a continuous variable) was of prognostic significance for OS ($$p \leq 0.039$$), which was confirmed by multivariate regression analysis ($$p \leq 0.029$$). The model where BMI replaced FFMI negated the prognostic value of BMI (as a continuous variable). Neither FFMI nor BMI was found to be a predictor of DFS on univariate or multivariate analysis.
### Conclusions
In the present group of HNSCC patients, low FFMI adversely influenced OS, emphasizing the importance of using body composition measurement over BMI alone for pretreatment nutritional evaluation of these patients.
## Introduction
Head and neck squamous cell carcinoma (HNSCC) patients are often nutritionally compromised due to lifestyle, location of the tumor growth, and the effects of treatment on food intake [1]. HNSCC patients have the second highest prevalence of malnutrition, with pretreatment severe weight loss ranging between $19\%$ and $57\%$ [2–4].
Malnutrition leads to altered body composition with depletion of fat mass and lean body mass, resulting in reduced physical and mental functioning and poorer clinical outcome [5]. Approximately $70\%$ of weight loss in cancer patients is thought to be due to loss of lean body mass [6–8]. Reduction of skeletal muscle mass is a good indicator of lean body mass loss and one of the established diagnostic criteria for assessing nutritional status [5, 9]. It leads to an increased risk of rehospitalizations, falls, fractures, loss of independence and death in hospitalized patients [10, 11].
Several studies in patients with HNSCC found an association between computed tomography (CT) determined decreased muscle mass and worse survival [12–15]. Although CT scans could be an important tool for assessing muscle mass [16–18], they are rarely used in clinical routine for this purpose [19]. According to the Global Leadership Initiative on Malnutrition (GLIM) criteria from 2019, reduced muscle mass is one out of three possible phenotypic criteria for diagnosing malnutrition in cancer patients and can be determined by fat-free mass index (FFMI) measurement using bioelectrical impedance analysis (BIA) [20].
Currently, BIA is a widely available, simple, non-invasive, and inexpensive method, routinely used in clinical settings [21]. Measuring the impedance of body tissues to the flow of electric current at a fixed frequency or range of frequencies determines the electrically conductive properties of the body and predicts body composition [22]. The principle of BIA is that lean tissue, consisting of water and electrolytes, is a good electrical conductor; on the contrary, fat is a poor electrical conductor as it does not contain water. Fat-free mass (FFM) assessed by BIA using special regression equations calibrated against the direct measurement of FFM can be and is used for FFMI calculation [23].
Under standard conditions, BIA measurements showed good correlation with the assessment of muscle mass by dual-energy X-ray absorptiometry (DEXA) [24, 25], magnetic resonance imaging (MRI) [26], and CT [27]. In HNSCC patients, body composition as determined with BIA was found to correlate strongly with CT-based estimates, although HNSCC patients represent a challenging population given wide fluctuations in their hydration status [19]. However, information on the impact of BIA-derived FFMI on the disease-free survival (DFS) and overall survival (OS) of HNSCC patients is limited in the literature [28]. The aim of the present study was to determine whether the FFMI determined by BIA can be used as a prognosticator for DFS and OS in this challenging group of patients.
## Patient eligibility
The study retrospectively included patients with HNSCC treated with curative intent between 2014 and 2018 who had pretreatment BIA (Bodystat® Quadscan 4000 (Douglas, GB)) evaluation of their nutritional status at the Department for Clinical Nutrition of Institute of Oncology Ljubljana, Slovenia. All tumors were histologically confirmed and without systemic metastases located in the oral cavity, oropharynx, hypopharynx, or larynx. Patients were treated with definitive or postoperative (chemo)radiotherapy (RT) and had to complete their treatment as planned. Linac-based intensity-modulated radiotherapy and concurrent weekly cisplatin (40 mg/m2 IV, in patients at high risk for in-field recurrence) were employed as indicated by the Multidisciplinary Head and Neck Tumor Board. Exclusion criteria were prior treatment in the head and neck area and any synchronous cancer except basal cell carcinoma of the skin.
## Study design
Demographic data and tumor-, treatment- and survival-related information were extracted from the clinical records of the patients. The tumors were staged using the criteria of the International Union Against Cancer (UICC) TNM staging system, 7th edition [29]. The p16 and/or human papillomavirus (HPV) status in patients with oropharyngeal cancer was determined by immunohistochemistry and/or in situ hybridization studies. Patients who had stopped smoking more than 2 years prior to diagnosis were considered ex-smokers. Comorbidities of patients at the time of HNSCC diagnosis were assessed using Charlson comorbidity Index, where the index cancer was not considered comorbidity [30].
FFMI was determined during the first consultation with a clinical dietitian using BIA, which was performed with the BodyStat BIA device (Douglas, GB) according to the standards of the National Health Institute, as previously described [31, 32]. To differentiate between normal and reduced FFMI values, cutoff points determined by the GLIM criteria for malnutrition were employed: for men, < 17 kg/m2 and for women < 15 kg/m2 [20]. Body mass index (BMI) values were also calculated and categorized according to the GLIM criteria (low BMI, < 70 years: < 20 kg/m2; > 70 years: < 22 kg/m2) [20].
## Statistical methods
The study was conducted according to the guidelines of the Declaration of Helsinki, and the study protocol was approved by the Committee for Medical Ethics and the Protocol Review Board of the Institute of Oncology Ljubljana (ERIDEK-$\frac{0044}{2021}$, 21.5.2021).
Statistical analysis was performed using R Studio, version 1.4.1106 (R-3.6.3). Categorical variables are presented as frequencies, and for continuous variables, arithmetic mean, standard deviation and range were calculated. The association of FFMI with categorical and continuous variables was tested with the chi-squared test (or Fisher’s exact test if the number of subjects in any of the cells was < 5) and the t test, respectively. The parametric correlation test (Pearson) was used to measure a linear dependence between the FFMI and BMI values in individual patients. The aims of the survival analysis were DFS (locoregional failure, distant metastasis, or death from any cause considered as an event) and OS (death from any cause considered as an event), which were defined as the time between the date of histological verification of the tumor and event or close-out date. The probability of DFS and OS was assessed using the Kaplan‒Meier method, and the log-rank test was used for curve comparison. The influence of FFMI and other variables on the OS of patients was tested with the univariate Cox regression model, where the FFMI and BMI were analyzed as continuous and categorical variables. Because several different covariates can potentially affect patient prognosis, a multivariate Cox regression model was used to examine the effect of different variables. In this model, the effect of FFMI and BMI on DFS and OS was examined separately, considering other variables that showed an impact ($p \leq 0.1$) on patient survival in univariate analysis. Thus, several models with different sets of variables were used in the multivariate analysis of DFS and OS. The performance of different models was compared by the corrected Akaike information criterion (cAIC), a tool for assessing the quality of various statistical models relative to each other and for the selection of the best model [33]. A p value of ≤ 0.05 was considered statistically significant.
## Study population
Out of 569 patients, 71 patients fulfilled the inclusion criteria. The overall mean FFMI value of all patients was 16.4 kg/m2 (SD ± 2.6, range 10.7–4.1), in women 14.0 kg/m2 (SD ± 2.2, range 10.7–17.0) and in men 17.0 kg/m2 (SD ± 2.4, range 11.8–24.1). Thirty-one ($43.7\%$) patients had normal FFMI, and 40 ($56.3\%$) patients had low FFMI. Considering BMI, 44 ($62\%$) patients were classified into the group with a normal BMI and 27 ($38\%$) into the group with a reduced BMI. A significant correlation was found between FFMI and BMI values measured in individual patients (RP = 0.75, $95\%$ confidence interval [CI] 0.63–0.84, $p \leq 0.001$). The demographic, clinical and nutritional characteristics of the study group are shown in Table 1. Smokers and ex-smokers were more likely to have low FFMI ($$p \leq 0.003$$), which was also associated with low BMI ($p \leq 0.001$).Table 1Demographic and clinical characteristics of patients grouped by fat free mass index (low: men, < 17 kg/m2; women, < 15 kg/m2)CharacteristicAll ($$n = 71$$)Normal FFMI ($$n = 31$$; $43.7\%$)Low FFMI ($$n = 40$$; $56.3\%$)p valueAge Mean age (± SD)61 (12.1)60.89 (13.60)61.10 (11.00)0.945Sex Men56 ($78.9\%$)25 ($80.6\%$)31 ($77.5\%$)0.748 Women15 ($21.1\%$)6 ($19.4\%$)9 ($22.5\%$)*Performance status* (WHO) 0–155 ($77.5\%$)23 ($74.2\%$)32 ($80.0\%$)0.561 216 ($22.5\%$)8 ($25.8\%$)8 ($20.0\%$)BMI Mean BMI (± SD)22.25 (4.12)25.68 (3.39)19.59 (2.24)< 0.001 Low BMI27 ($38\%$)0 ($0\%$)27 ($67.5\%$)< 0.001 Normal BMI44 ($62\%$)31 ($100\%$)13 ($32.5\%$)*Smoking status* ($$n = 63$$) Non-smokers and ex-smokers22 ($31\%$)15 ($55.6\%$)7 ($19.4\%$)0.003 Smokers41 ($57.7\%$)12 ($44.4\%$)29 ($80.6\%$) Unknown8 ($11.3\%$)Comorbidities (CCI) 048 ($67.6\%$)18 ($58.1\%$)30 ($75.0\%$)0.130 1–323 ($32.4\%$)13 ($41.9\%$)10 ($25.0\%$) 112 ($16.9\%$) 26 ($8.5\%$) 35 ($7.0\%$)Primary tumor location Oropharynx32 ($45.1\%$)12 ($38.7\%$)20 ($50.0\%$)0.374 HPV + 8 ($25\%$) HPV−24 ($75\%$) Hypopharynx and larynx27 ($38.1\%$)13 ($41.9\%$)14 ($35.0\%$) Oral cavity12 ($16.9\%$)6 ($19.4\%$)6 ($15.0\%$)Overall stage I–III13 ($18.3\%$)8 ($25.8\%$)5 ($12.5\%$)0.151 IV58 ($81.7\%$)23 ($74.2\%$)35 ($87.5\%$)Surgical resection Yes32 ($45.1\%$)18 ($58.1\%$)14 ($35.0\%$)0.053 R027 ($84.4\%$) R12 ($6.2\%$) R23 ($9.4\%$) No39 ($54.9\%$)13 ($41.9\%$)26 ($65.0\%$)Addition of ChT RT32 ($45.1\%$)17 ($54.8\%$)15 ($37.5\%$)0.069 ChRT36 ($50.7\%$)14 ($45.2\%$)22 ($55.0\%$) Induction ChT (→ RT/ChRT)3 ($4.2\%$)0 ($0\%$)3 ($7.5\%$)FFMI fat free mass index, BMI body mass index, SD standard deviation, WHO World Health Organization, HPV Human papillomavirus, ChRT Chemoradiotherapy, RT radiotherapy, ChT chemotherapy, BIA bioelectrical impedance analysisp ≤ 0.05 statistically significant are in bold On the close-out date, 53 ($74.6\%$) of the patients were dead, either due to disease progression (25, $35\%$) or other causes (28, $39.4\%$). The mean time to malignant disease progression/recurrence or death was 1.5 years (range 0–5.6). Surviving patients were followed-up between 2.6 and 7.4 years (mean 4.4).
The DFS rates at 3 years of patients with low and normal FFMI was $27.0\%$ ($95\%$ CI 0.16–0.45) and $44.9\%$ ($95\%$ CI 0.30–0.67), respectively ($$p \leq 0.06$$, Fig. 1) and the OS rates $29.4\%$ ($95\%$ CI 0.18–0.48) and $47.7\%$ ($95\%$ CI 0.33–0.69), respectively ($$p \leq 0.06$$, Fig. 2). DFS was $36.2\%$ ($95\%$ CI 0.24–0.54) in patients with normal BMI and $32.4\%$ ($95\%$ CI 0.19–0.57) in those with low BMI ($$p \leq 0.80$$). The corresponding OS rates were $40.5\%$ ($95\%$ CI 0.28–0.58) and $32.1\%$ ($95\%$ CI 0.18–0.56), respectively ($$p \leq 0.60$$).Fig. 1Disease-free survival of patients with low and normal fat-free mass index (FFMI) as determined by bioelectrical impedance analysis (BIA) ($$p \leq 0.06$$)Fig. 2Overall survival of patients with low and normal fat-free mass index (FFMI) as determined by bioelectrical impedance analysis (BIA) ($$p \leq 0.06$$) In the univariate Cox regression model for DFS, only treatment type (surgical vs. non-surgical, $$p \leq 0.038$$) had a statistically significant effect (Table 2). In three different multivariate analysis models, only PS consistently showed statistical significance on DFS: neither FFMI (as a continuous or binary variable), nor BMI (continuous) was retained in the final model (Table 2).Table 2Univariate and multivariate Cox regression analyses of disease-free survival ($$n = 71$$)VariableUnivariate analysisMultivariate analysis 1Multivariate analysis 2Multivariate analysis 3HR ($95\%$ CI)p valueHR ($95\%$ CI)p valueHR ($95\%$ CI)p valueHR ($95\%$ CI)p valueFFMI (normal vs. low)1.72 (0.98–3.01)0.0581.96 (0.75–5.11)0.171FFMI (continuous)0.90 (0.81–1.00)0.0600.96 (0.78–1.18)0.691BMI (normal vs. low)1.09 (0.62–1.89)0.771BMI (continuous)0.96 (0.89–1.02)0.1850.92 (0.81–1.04)0.174Sex (men vs. women)1.19 (0.63–2.27)0.593Age (continuous)1.01 (0.99–1.03)0.592Comorbidities (0 vs. 1–3)1.46 (0.82–2.59)0.200Tumor location (OP/OC vs. HP/LX)1.02 (0.59–1.78)0.939Overall stage (I–III vs. IV)1.40 (0.66–2.98)0.378Treatment type (surgical vs. non-surgical)1.79 (1.03–3.11)0.0380.98 (0.34–2.84)0.9600.79 (0.27–2.32)0.6741.12 (0.41–3.07)0.833Performance status, WHO (0–1 vs. 2)1.71 (0.93–3.16)0.0864.75 (1.51–14.97)0.0084.50 (1.45–13.99)0.0095.12 (1.63–16.08)0.005Smoking status (non-/ex-smokers vs. smokers)1.04 (0.57–1.90)0.89p16/HPV status (negative vs. positive), $$n = 32$$*0.35 (0.18–1.05)0.060.44 (0.13–1.50)0.1900.39 (0.11–1.38)0.1360.45 (0.13–1.53)0.201FFMI fat free mass index, BMI body mass index, OP oropharynx, OC oral cavity, HP hypopharynx, LX larynx, WHO World Health Organization, n number of patients, HR hazard ratio, CI confidence intervalp ≤ 0.05 statistically significant are in bold*Only patients with oropharyngeal primary tumors In the univariate Cox regression model, World Health Organization (WHO) performance status (PS) (0–1 vs. 2, $$p \leq 0.016$$) and FFMI (as a continuous variable, $$p \leq 0.039$$) had a statistically significant effect on OS (Table 2). In the first multivariable analysis model, which included FFMI as a continuous variable, both variables remained statistically significant. In the second model with BMI (as a continuous variable), only PS and treatment type showed statistical significance (Table 3). For both multivariate analysis models, the cAIC was obtained to reveal the model with the lowest cAIC value (Table 4). The model that included FFMI was shown to be more accurate and informative in terms of OS prediction than the model with BMI.Table 3Univariate and multivariate Cox regression analyses of overall survival ($$n = 71$$)VariableUnivariate analysisMultivariate analysis 1Multivariate analysis 2Hazard ratio ($95\%$ CI)p valueHazard ratio ($95\%$ CI)p valueHazard ratio ($95\%$ CI)p valueFFMI (normal vs. low)1.71 (0.98–2.99)0.06FFMI (continuous)0.89 (0.81–0.99)0.0390.88 (0.79–0.99)0.029BMI (normal vs. low)1.17 (0.67–2.05)0.582BMI (continuous)0.95 (0.89–1.02)0.150.93 (0.86–1.00)0.061Sex (men vs. women)1.15 (0.60–2.18)0.677Age (continuous)1.01 (0.98–1.03)0.525Tumor location (OP/OC vs. HP/LX)1.00 (0.57–1.74)0.991Overall stages (I–III vs. IV)1.63 (0.77–3.47)0.203Treatment type (surgical vs. non-surgical)1.65 (0.96–2.86)0.0721.64 (0.92–2.91)0.0911.92 (1.08–3.39)0.025Performance status, WHO (0–1 vs. 2)2.14 (1.15–3.98)0.0162.85 (1.48–5.47)0.0022.90 (1.49–5.64)0.002Smoking status (non-/ex-smokers vs. smokers)0.92 (0.50–1.68)0.788P16/HPV status (negative vs. positive), $$n = 32$$*0.43 (0.15–1.27)0.127FFMI fat free mass index, BMI body mass index, OP oropharynx, OC oral cavity, HP hypopharynx, LX larynx, WHO World Health Organization, n number of patients, CI confidence intervalp ≤ 0.05 statistically significant are in bold*Only patients with oropharyngeal primary tumorsTable 4Prognostic performance of multivariate analysis modelsModelKAICcOverall survival MVA model with FFMI as a continuous variable3376.42 MVA model with BMI as a continuous variable3377.46Disease-free survival MVA model with FFMI as a continuous variable4134.08 MVA model with FFMI as a categorical variable4134.08 MVA model with BMI as a continuous variable4135.89MVA multivariable analysis, FFMI fat free mass index, BMI body mass index, K number of parameters in the model, cAIC corrected Akaike information criterion
## Discussion
The present study confirms that body composition as measured by BIA, but not BMI, is an independent prognostic factor for predicting OS in HNSCC patients, in addition to their PS. This speaks in favor of BIA as a more practical bedside procedure that, e.g., CT, is also noninvasive, reproducible and inexpensive [21, 34]. Although BIA may result in incorrect assessment of muscle mass with FFMI in poorly hydrated patients [34], a good correlation was generally reported between BIA and CT measurements of skeletal muscle mass [19].
The GLIM consensus for malnutrition recognized 3 phenotypic criteria for the diagnosis of malnutrition, i.e., weight loss (in %), decrease in muscle mass and BMI [20]. Before the start of treatment, $38\%$ of our patients had a low BMI. After categorizing patients according to GLIM criteria, $56.3\%$ of the patients had reduced pretreatment FFMI. This almost $20\%$ difference in the share of malnourished patients further supports the importance of the use of several criteria to determine malnutrition [20]. In studies that used either CT scans or BIA for the determination of muscle mass, the prevalence of patients with low muscle mass differs significantly (20.5–$54.5\%$) [15, 35], reflecting the characteristics of the studied population. In our case, only HNSCC patients who were directed to our department before oncological treatment due to clinically identifiable and already existing or threatened malnutrition were included in the study. This could be the reason for the higher proportion of patients with low FFMI than in some other studies [26, 35, 36].
We found no correlation between FFMI and age, which is contrary to the general premise that muscle mass decreases with age [15, 37]. Furthermore, the FFMI of our patients also did not correlate with sex, PS, primary tumor location, overall disease stage, or type of treatment (definitive or postoperative (chemo)radiotherapy). This also contradicts the findings of some other authors, i.e., the relationship between low muscle mass and female sex [15, 38, 39] or higher overall disease stage [14, 38]. However, we observed an association between reduction of muscle mass and smoking, as did Bril et al. [ 38] but not also Wendrich et al. [ 15] and Huiskamp et al. [ 40], possibly reflecting the problem of the reliability of data obtained from patients. In addition, reduced muscle mass correlated with a lower mean BMI in our patients, which was also previously described [15, 38].
The 3-year DFS and OS in our group of HNSCC patients were only $34.9\%$ and $37.5\%$, respectively, probably because of a selection bias by including mainly already nutritionally compromised patients ($94.4\%$) in the advanced stage of the disease (IVA-B, $81.7\%$) with at least one comorbidity ($32.4\%$). This probably masked the differences in survival between individual categories of patients (e.g., regarding the origin of the tumor, the stage of the disease, the type of treatment). An additional, albeit related, reason was the high prevalence of low muscle mass ($56.3\%$) in our patients, which turned out to be an independent adverse prognostic factor for OS (but not also for DFS) in multivariate analysis. Several other studies demonstrated a negative prognostic impact of low muscle mass on the survival of HNSCC patients [14, 19, 39]. Moreover, despite the statistically significant correlation between FFMI and BMI values measured in individual patients, only FFMI proved to be of significance for predicting OS in multivariate analysis. Although the association between reduced muscle mass and weight reduction (and thus lower BMI) is to be expected, it should be noted that BMI alone is not a good predictor of lower muscle mass and altered body composition [41]. Moreover, one should be aware that weight loss is not necessarily present in sarcopenia and, on the other hand, muscle mass may also be reduced in individuals with sarcopenic obesity [42]. In contrast to the OS analysis, neither FFMI nor BMI was found to be an independent predictor of DFS.
Limitations of this study relate primarily to its retrospective nature and its inherent shortcomings (selection bias, sometimes deficient data that are questionably reliable). Furthermore, inclusion criteria that further curtailed the available set of patients and resulted in a relatively small sample size added to the selection bias and resulted in a problem with the relativity of small series statistics. For a reliable determination of the relationship between FFMI and OS, all patients with HNSCC treated over a selected period should have been included. However, due to the lack of research in HNSCC using BIA-determined FFMI, the presented results could be a valuable source of scientific data to the existing knowledge.
To conclude, in the present study, we retrospectively demonstrated that BIA determined low FFMI as a measure of body muscle mass, but BMI also did not appear to be a negative prognostic factor for OS in HNSCC patients. This emphasizes the importance of using body composition measurements, such as FFMI, over BMI alone in these patients for prognostic evaluation. Although our findings are consistent with the general opinion of the experts that low muscle mass is prognostic for negative oncological outcomes, further studies with prospective recruitment of all HNSCC patients are needed for confirmation.
## References
1. Almada-Correia I, Neves PM, Mäkitie A, Ravasco P. **Body composition evaluation in head and neck cancer patients: a review**. *Front Oncol* (2019.0) **9** 1112. DOI: 10.3389/fonc.2019.01112
2. Ferrão B, Neves PM, Santos T, Capelas ML, Mäkitie A, Ravasco P. **Body composition changes in patients with head and neck cancer under active treatment: a scoping review**. *Support Care Cancer* (2020.0) **28** 4613-4625. DOI: 10.1007/s00520-020-05487-w
3. Jager-Wittenaar H, Dijkstra PU, Vissink A, van der Laan BFAM, van Oort RP, Roodenburg JLN. **Critical weight loss in head and neck cancer-prevalence and risk factors at diagnosis: an explorative study**. *Support Care Cancer* (2007.0) **15** 1045-1050. DOI: 10.1007/s00520-006-0212-9
4. 4.van Bokhorst-de van der Schueren MA, van Leeuwen PA, Sauerwein HP, Kuik DJ, Snow GB, Quak JJ (1997) Assessment of malnutrition parameters in head and neck cancer and their relation to postoperative complications. Head Neck 19:419–425. 10.1002/(sici)1097-0347(199708)19:5<419::aid-hed9>3.0.co;2-2
5. Cederholm T, Barazzoni R, Austin P, Ballmer P, Biolo G, Bischoff SC. **ESPEN guidelines on definitions and terminology of clinical nutrition**. *Clin Nutr* (2017.0) **36** 49-64. DOI: 10.1016/j.clnu.2016.09.004
6. Jackson W, Alexander N, Schipper M, Fig L, Feng F, Jolly S. **Characterization of changes in total body composition for patients with head and neck cancer undergoing chemoradiotherapy using dual-energy x-ray absorptiometry**. *Head Neck* (2014.0) **36** 1356-1362. DOI: 10.1002/hed.23461
7. Solís-Martínez O, Plasa-Carvalho V, Phillips-Sixtos G, Trujillo-Cabrera Y, Hernández-Cuellar A, Queipo-García GE, Meaney-Mendiolea E, Ceballos-Reyes GM, Fuchs-Tarlovsky V. **Effect of eicosapentaenoic acid on body composition and inflammation markers in patients with head and neck squamous cell cancer from a public hospital in Mexico**. *Nutr Cancer* (2018.0) **70** 663-670. DOI: 10.1080/01635581.2018.1460678
8. Lønbro S, Dalgas U, Primdahl H, Johansen J, Nielsen JL, Aagaard P, Hermann AP, Overgaard J, Overgaard K. **Progressive resistance training rebuilds lean body mass in head and neck cancer patients after radiotherapy: results from the randomized DAHANCA 25B trial**. *Radiother Oncol* (2013.0) **108** 314-319. DOI: 10.1016/j.radonc.2013.07.002
9. Landi F, Camprubi-Robles M, Bear DE, Cederholm T, Malafarina V, Welch AA, Cruz-Jentoft AJ. **Muscle loss: the new malnutrition challenge in clinical practice**. *Clin Nutr* (2019.0) **38** 2113-2120. DOI: 10.1016/j.clnu.2018.11.021
10. Prado CM, Purcell SA, Alish C, Pereira SL, Deutz NE, Heyland DK, Goodpaster BH, Tappenden KA, Heymsfield SB. **Implications of low muscle mass across the continuum of care: a narrative review**. *Ann Med* (2018.0) **50** 675-693. DOI: 10.1080/07853890.2018.1511918
11. Gariballa S, Alessa A. **Impact of poor muscle strength on clinical and service outcomes of older people during both acute illness and after recovery**. *BMC Geriatr* (2017.0) **17** 123. DOI: 10.1186/s12877-017-0512-6
12. Nishikawa D, Hanai N, Suzuki H, Koide Y, Beppu S, Hasegawa Y. **The impact of skeletal muscle depletion on head and neck squamous cell carcinoma**. *ORL* (2018.0) **80** 1-9. DOI: 10.1159/000485515
13. Thureau S, Lebret L, Lequesne J, Cabourg M, Dandoy S, Gouley C, Lefebvre L, Mallet R, Mihailescu SD, Moldovan C, Rigal O, Veresezan O, Modzewelski R, Clatot F. **Prospective evaluation of sarcopenia in head and neck cancer patients treated with radiotherapy or radiochemotherapy**. *Cancers* (2021.0) **13** 753. DOI: 10.3390/cancers13040753
14. Rijn-Dekker MI, Bosch L, Hoek JGM, Bijl HP, Aken ESM, Hoorn A, Oosting SF, Halmos GB, Witjes MJH, van der Laan HP, Langendijk JA, Steenbakkers RJHM. **Impact of sarcopenia on survival and late toxicity in head and neck cancer patients treated with radiotherapy**. *Radiother Oncol* (2020.0) **147** 103-110. DOI: 10.1016/j.radonc.2020.03.014
15. Wendrich AW, Swartz JE, Bril SI, Wegner I, de Graeff A, Smid EJ, de Bree R, Pothen AJ. **Low skeletal muscle mass is a predictive factor for chemotherapy dose-limiting toxicity in patients with locally advanced head and neck cancer**. *Oral Oncol* (2017.0) **71** 26-33. DOI: 10.1016/j.oraloncology.2017.05.012
16. Shen W, Punyanitya M, Wang Z, Gallagher D, St-Onge M-P, Albu J, Heymsfield SB. **Heshka S (2004) Total body skeletal muscle and adipose tissue volumes: estimation from a single abdominal cross-sectional image**. *J Appl Physiol* (1985.0) **97** 2333-2338. DOI: 10.1152/japplphysiol.00744.2004
17. Mourtzakis M, Prado CMM, Lieffers JR, Reiman T, McCargar LJ, Baracos VE. **A practical and precise approach to quantification of body composition in cancer patients using computed tomography images acquired during routine care**. *Appl Physiol Nutr Metab* (2008.0) **33** 997-1006. DOI: 10.1139/H08-075
18. Cesari M, Vellas B. **Sarcopenia: a novel clinical condition or still a matter for research?**. *J Am Med Dir Assoc* (2012.0) **13** 766-767. DOI: 10.1016/j.jamda.2012.07.020
19. Grossberg AJ, Rock CD, Edwards J, Mohamed ASR, Ruzensky D, Currie A, Rosemond P, Phan J, Gunn GB, Frank SJ, Morrison WH, Garden AS, Fuller CD, Rosenthal DI. **Bioelectrical impedance analysis as a quantitative measure of sarcopenia in head and neck cancer patients treated with radiotherapy**. *Radiother Oncol* (2021.0) **159** 21-27. DOI: 10.1016/j.radonc.2021.03.005
20. Cederholm T, Gl J, Correia MITD, Gonzalez MC, Fukushima R, Higashiguchi T. **GLIM criteria for the diagnosis of malnutrition—a consensus report from the global clinical nutrition community**. *J Cachexia Sarcopenia Muscle* (2019.0) **10** 207-217. DOI: 10.1002/jcsm.12383
21. Deutz NEP, Ashurst I, Ballesteros MD, Bear DE, Cruz-Jentoft AJ, Genton L, Landi F, Laviano A, Norman K, Prado CM. **The underappreciated role of low muscle mass in the management of malnutrition**. *J Am Med Dir Assoc* (2019.0) **20** 22-27. DOI: 10.1016/j.jamda.2018.11.021
22. Kuriyan R. **Body composition techniques**. *Indian J Med Res* (2018.0) **148** 648-658. DOI: 10.4103/ijmr.IJMR_1777_18
23. Gonzalez MC, Pastore C, Orlandi S, Heymsfield S. **Obesity paradox in cancer: new insights provided by body composition**. *Am J Clin Nutr* (2014.0) **99** 999-1005. DOI: 10.3945/ajcn.113.071399
24. Leahy S, O’Neill C, Sohun R, Jakeman P. **A comparison of dual energy X-ray absorptiometry and bioelectrical impedance analysis to measure total and segmental body composition in healthy young adults**. *Eur J Appl Physiol* (2012.0) **112** 589-595. DOI: 10.1007/s00421-011-2010-4
25. Kim M, Shinkai S, Murayama H, Mori S. **Comparison of segmental multifrequency bioelectrical impedance analysis with dual-energy X-ray absorptiometry for the assessment of body composition in a community-dwelling older population**. *Geriatr Gerontol Int* (2015.0) **15** 1013-1022. DOI: 10.1111/ggi.12384
26. Janssen I, Heymsfield SB, Baumgartner RN. **Ross R (2000) Estimation of skeletal muscle mass by bioelectrical impedance analysis**. *J Appl Physiol* (1985.0) **89** 465-471. DOI: 10.1152/jappl.2000.89.2.465
27. Aleixo GFP, Shachar SS, Nyrop KA, Muss HB, Battaglini CL, Williams GR. **Bioelectrical impedance analysis for the assessment of sarcopenia in patients with cancer: a systematic review**. *Oncologist* (2020.0) **25** 170-182. DOI: 10.1634/theoncologist.2019-0600
28. Willemsen ACH, Hoeben A, Lalisang RI, Van Helvoort A, Wesseling FWR, Hoebers F, Baijens LWJ, Schols AMWJ. **Disease-induced and treatment-induced alterations in body composition in locally advanced head and neck squamous cell carcinoma**. *J Cachexia Sarcopenia Muscle* (2020.0) **11** 145-159. DOI: 10.1002/jcsm.12487
29. Sobin LH, Gospodarowicz M, Wittekind C. *International Union Against Cancer (UICC) (2010) TNM Classification of Malignant Tumours* (2003.0)
30. Charlson ME, Pompei P, Ales KL, MacKenzie R. **A new method of classifying prognostic comorbidity in longitudinal studies: development and validation**. *J Chron Dis* (1987.0) **40** 373-383. DOI: 10.1016/0021-9681(87)90171-8
31. Gosak M, Gradišar K, RotovnikKozjek RN, Strojan P. **Psychological distress and nutritional status in head and neck cancer patients: a pilot study**. *Eur Arch Otorhinolaryngol* (2020.0) **277** 1211-1217. DOI: 10.1007/s00405-020-05798-y
32. Stegel P, Kozjek NR, Brumen BA, Strojan P. **Bioelectrical impedance phase angle as indicator and predictor of cachexia in head and neck cancer patients treated with (chemo)radiotherapy**. *Eur J Clin Nutr* (2016.0) **70** 602-606. DOI: 10.1038/ejcn.2016.13
33. Cavanaugh JE, Neath AA. **The Akaike information criterion: Background, derivation, properties, application, interpretation, and refinements**. *WIREs Comput Stat* (2019.0) **11** e1460. DOI: 10.1002/wics.1460
34. Cruz-Jentoft AJ, Bahat G, Bauer J, Boirie Y, Bruyère O, Cederholm T. **Writing Group for the European Working Group on Sarcopenia in Older People 2 (EWGSOP2), and the Extended Group for EWGSOP2. Sarcopenia: revised European consensus on definition and diagnosis**. *Age Ageing* (2019.0) **48** 16-31. DOI: 10.1093/ageing/afy169
35. Cereda E, Pedrazzoli P, Lobascio F, Masi S, Crotti S, Klersy C, Turri A, Stobäus N, Tank M, Franz K, Cutti S, Giaquinto E, Filippi AR, Norman K, Caccialanza R. **The prognostic impact of BIA-derived fat-free mass index in patients with cancer**. *Clin Nutr* (2021.0) **40** 3901-3907. DOI: 10.1016/j.clnu.2021.04.024
36. Lundberg M, Nikander P, Tuomainen K, Orell-Kotikangas H, Mäkitie A. **Bioelectrical impedance analysis of head and neck cancer patients at presentation**. *Acta Oto-Laryngol* (2017.0) **137** 417-420. DOI: 10.1080/00016489.2016.1266510
37. Janssen I, Heymsfield SB, Wang ZM, Ross R. **Skeletal muscle mass and distribution in 468 men and women aged 18–88 yr**. *J Appl Physiol (1985)* (2000.0) **89** 81-88. DOI: 10.1152/jappl.2000.89.1.81
38. Bril SI, Pezier TF, Tijink BM, Janssen LM, Braunius WW, de Bree R. **Preoperative low skeletal muscle mass as a risk factor for pharyngocutaneous fistula and decreased overall survival in patients undergoing total laryngectomy**. *Head Neck* (2019.0) **41** 1745-1755. DOI: 10.1002/hed.25638
39. Chargi N, Wegner I, Markazi N, Smid E, de Jong P, Devriese L, de Bree R. **Patterns, predictors, and prognostic value of skeletal muscle mass loss in patients with locally advanced head and neck cancer undergoing cisplatin-based chemoradiotherapy**. *J Clin Med* (2021.0) **10** 1762. DOI: 10.3390/jcm10081762
40. Huiskamp LFJ, Chargi N, Devriese LA, de Jong PA, de Bree R. **The predictive and prognostic value of low skeletal muscle mass for dose-limiting toxicity and survival in head and neck cancer patients receiving concomitant cetuximab and radiotherapy**. *Eur Arch Otorhinolaryngol* (2020.0) **277** 2847-2858. DOI: 10.1007/s00405-020-05972-2
41. Gonzalez MC, Correia MITD, Heymsfield SB. **A requiem for BMI in the clinical setting**. *Curr Opin Clin Nutr Metab Care* (2017.0) **20** 314-321. DOI: 10.1097/MCO.0000000000000395
42. Donini LM, Busetto L, Bischoff SC, Cederholm T, Ballesteros-Pomar MD, Batsis J. **Definition and diagnostic criteria for sarcopenic obesity: ESPEN and EASO consensus statement**. *Obes Facts* (2022.0) **15** 321-335. DOI: 10.1159/000521241
|
---
title: 'Symptoms and Comorbidities Differ Based on Race and Weight Status in Persons
with HIV in the Northern United States: a Cross-Sectional Study'
authors:
- Kierra R. Butler
- Faye R. Harrell
- Bridgett Rahim-Williams
- Jeffrey M. Robinson
- Xuemin Zhang
- Adwoa Gyamfi
- Judith A. Erlen
- Wendy A. Henderson
journal: Journal of Racial and Ethnic Health Disparities
year: 2022
pmcid: PMC9988761
doi: 10.1007/s40615-022-01271-0
license: CC BY 4.0
---
# Symptoms and Comorbidities Differ Based on Race and Weight Status in Persons with HIV in the Northern United States: a Cross-Sectional Study
## Abstract
### Background
Persons with HIV (PWHIV) on highly active antiretroviral treatments (HAART) may require specialized care based on health and demographic indicators. This study investigated the association of comorbidities, race, weight status, and gastrointestinal (GI) and cardiovascular (CV) symptoms among PWHIV.
### Methods
The Symptom Checklist, Co-Morbidity Questionnaire, and Sociodemographic Questionnaire were used to assess weight status and GI and CV symptoms among 283 PWHIV. Data were analyzed using latent class analysis on John’s Macintosh Project 13 Platform.
### Results
Participants were majority Black ($50\%$), $69\%$ male, and $35\%$ AIDS diagnosed. Ages were 25 to 66. Clusters included least symptomatic status, weight gain, and weight loss by Black and non-Black participants. The non-Black weight gain cluster reported a higher incidence of AIDS ($70.6\%$ vs $38.2\%$), nausea ($70.6\%$ vs $17.6\%$), diarrhea ($70.6\%$ vs $26.5\%$), and shortness of breath ($58.8\%$ vs $20.6\%$) compared to the Black weight gain cluster. The Black weight loss cluster reported a higher incidence of CV symptoms such as chest palpitations ($42.2\%$ vs $2.7\%$), chest pain ($44.4\%$ vs $8.1\%$), and shortness of breath ($73.3\%$ vs $35.1\%$). Moreover, the Black weight loss cluster reported a higher incidence of all GI symptoms with the most prominent being diarrhea ($71.1\%$ vs $48.6\%$) compared to the non-Black weight loss cluster.
### Conclusions
The existing racial disparities in health-related quality of life for PWHIV may be improved through precision health and nutrition modifications. Continued research is needed investigating differential health outcomes among PWHIV on HAART.
### Clinical Trial Registration Number
NCT00222716. Registered 22 September 2005. Retrospectively registered, https://clinicaltrials.gov/ct2/show/NCT00222716?term=NCT00222716&draw=2&rank=1
## Background
Human immunodeficiency virus (HIV), a viral infection previously known as the wasting disease, has become a chronic yet manageable disease due to the advent of highly active antiretroviral treatments (HAART) [1]. HIV is characterized by chronic systemic inflammation including inflammation of the gastrointestinal (GI) barrier [2]. The virus depletes specific cluster of differentiation 4 [CD4+] thymus cells [T-cells] in the gut-associated lymphoid tissue causing the intestinal barrier to become inflamed, leaky, and more permeable [3]. Inflammation in the gut may contribute to persons with HIV (PWHIV) having GI symptoms, such as nausea, vomiting, and diarrhea. Many of these symptoms cause weight changes, affecting adherence to HAART [4]. As the gut becomes more permeable, microbes can translocate into the systemic circulation and cause systemic inflammation. Chronic systemic inflammation may contribute to the progression of the disease, systemic dysregulation, and cardiovascular (CV) instability [5]. Microbial translocation is associated with hypertension and GI symptoms in PWHIV [6]. Hypertension and the inflammatory process driven by HIV alter the endothelial function of these patients [5].
Although HAART has prolonged life, studies suggest that it may also be associated with weight gain [7–9], CV disease, and GI symptoms [4, 7]. Weight gain after initiating HAART occurs frequently and is associated with lower mortality. Thus, it is looked upon as a favorable outcome [9]. However, excessive weight gain may lead to an increased risk of chronic conditions such as hypertension, diabetes mellitus, and CV disease [7]. In fact, CV disease is the leading cause of death among PWHIV [8].
Current literature has shown that weight changes in PWHIV on HAART have significant implications for health outcomes. CV disease and PLWHIV on HAART must have ongoing investigations to identify factors that will guide care over the lifespan. Gastrointestinal symptoms such as nausea, vomiting, diarrhea, and loss of appetite affect health outcomes for PWHIV on HAART. Increases in body mass index (BMI) are common in HIV-positive minorities and women. Symptoms were found to vary with patient race, age, and disease progression [9, 10]. Racial differences in conjunction with symptom presentation influence treatment options. Socioeconomic status and neighborhood disadvantage contribute to chronic stress for many Black/African American PWHIV and must be considered in their care when relevant [11, 12]. The clinician selects the proper treatment based on weight status (e.g., weight gain, weight loss), race, and symptoms (especially loss of appetite, acquired immunodeficiency syndrome [AIDS] classification, diarrhea, vomiting, and overeating). The selection of a specified treatment option is of paramount importance because the grid of care options varies; care options even oppose one another for certain groups. The care for Black/African American PWHIV on HAART who lose weight and have a loss of appetite is quite different than the care for Black/African American PWHIV on HAART who gain weight and overeat. Thus, precision care involving the proper choice of investigation and treatment improves patients’ outcomes and accuracy of care provided by practitioners [13].
Limited clinical research exists on the topics of changes in weight status, race classification, gastrointestinal health, and CV health among PWHIV on HAART. Clinicians benefit from knowledge regarding this population that will guide personalized care that targets each specific demographic group based on the needs associated with their weight status, their demographics, and presenting symptoms. The health disparities for Black/African American patients in terms of CV disease and HIV are well documented. There is a higher burden of CV risk factors and CV disease in patients who are Black/Africa American [14]. Additionally, it is 13 times more likely that a Black/African American over the age of 50 will receive an HIV diagnosis than a White/Caucasian over 50. To address these disparities, health professionals must provide ongoing and personalized care for this population [15]. As such, monitoring of PWHIV on HAART may require stratification and personalization based on demographics and symptoms. The differences found may be attributable to patients’ race, sex, weight status, comorbidities, and presenting symptoms. Moreover, as PWHIV on HAART age, comorbidities require closer examination, especially in terms of racial differences [16].
Current literature associated with weight status and multiple morbidities in PWHIV either examines the change in weight status/body mass index (BMI) related to race after HAART initiation [17], changes in BMI across a life span [18], existence of multiple morbidities and obesity without examining the effects of race[19], or multiple morbidities and aging [20], with some attention to weight/obesity [8]. Although research has been conducted on this topic, there is limited clinical research investigating these variables in a clinical population. Most research has included epidemiological large, cross-sectional studies and literature reviews. The current study was a secondary data analysis of a clinical population. The aim of the study was to investigate the association of race (Black and non-Black [Asian, mixed-race, and Whites]), weight status (weight gain and weight loss), gastrointestinal and cardiovascular symptoms (nausea, vomiting, shortness of breath, chest pain), and comorbidities (hypertension, coronary artery disease, heart failure) among PWHIV/AIDS. The findings of this study will contribute to the growing body of research addressing adverse effects experienced by PWHIV on HAART and provide important information related to personalized treatment, especially related to race.
## Methods
A secondary analysis was performed on data from the parent study, Improving Adherence to Antiretroviral Therapy (R01 NR04749, PI, J. A. Erlen, University of Pittsburgh). The aim of the study was to improve adherence to antiretroviral therapy in PWHIV through a nurse-delivered, telephone-based intervention. The parent study recruited 356 PWHIV on HAART from the Northern United States, specifically, western Pennsylvania and eastern Ohio community hospitals, university-based clinics, comprehensive HIV care centers, and through self-referral [21]. Inclusion criteria for the parent study required a positive HIV diagnosis by a healthcare provider being treated with antiretroviral medication, access to a telephone, and consent to participate in the study.
In this cross-sectional, secondary analysis, participants were included who submitted responses to all CV and GI symptoms and comorbidity questions on the Symptom Checklist, Co-Morbidity Questionnaire, and Sociodemographic Questionnaire (Center for Research in Chronic Disorders, University of Pittsburgh School of Nursing, 1999). Applying these inclusionary criteria to the parent study sample resulted in 283 participants for the current study.
To analyze data from the 283 participants, latent class analysis (LCA) was implemented with John’s Macintosh Project (JMP) 13 to perform an analysis of self-reported data on the three aforementioned questionnaires (Fig. 1). LCA is an unsupervised, multivariate grouping method that fits a model and determines the most likely “latent class” of each participant, in a pre-selected number of discreet classes (Statistical Analysis System [SAS] Institute Inc., 2016). In this analysis, three groups were selected a priori. Fig. 1Latent class analysis results
## Results
Of the participants, approximately $50\%$ self-identified as Black, $69\%$ as male, and $35\%$ as having AIDS. Participants’ ages ranged from 25 to 66 years (mean age = 43.70 years) as shown in Table 1. Participants were grouped into clusters by race, Black and non-Black. Within the Black and non-Black groups, a pattern developed among the three clusters. Each racial group had a cluster of PWHIV who reported the lowest incidence of symptoms (weight loss, weight gain, vomiting, etc.) and comorbidities and a cluster characterized with a high incidence of weight gain and weight loss. Thus, each cluster was labeled based on the most prevalent symptom. Within each racial group (Black and non-Black), the clusters were the least symptomatic cluster, weight gain cluster, and weight loss cluster. After the LCA was conducted, descriptive statistics were calculated using International Business Machines Corporation Statistical Package for the Social Sciences (IBM SPSS) version 25 to ascertain the overall and average number of self-reported symptoms and comorbidities for each cluster within the Black and non-Black groups. A chart of the percentage of all GI and CV symptoms and comorbidities reported can be found in Table 2.Table 1Demographic characteristics of sampleDemographic measuresOverall group($$n = 283$$)Black($$n = 144$$)Non-Black($$n = 139$$)Gender % (N)Male68.90 [195]32.51 [92]36.40 [103]Female31.10 [88]18.37 [52]12.72 [36]AIDS % (N)Male34.98 [99]11.31 [32]13.78 [39]Female5.30 [15]4.59 [13]AgeMean = 43.69CD4 + countMean = 436.03CD4 count $$n = 182$$Table 2Cluster demographic and variable percentagesBlack least symptomatic clustern = 65 (%)Non-Black least symptomatic clustern = 85 (%)Black weight loss clustern = 43 (%)Non-Black weight loss clustern = 37 (%)Black weight gain clustern = 34 (%)Non-Black weight gain clustern = 17 (%)Sex70.8 male76.5 male66.7 male70.3 male47.1 male70.6 maleAge ± Std Dv44.49 ± 8.845.54 ± 8.8141.07 ± 7.344.24 ± 7.7641.32 ± 5.5041.94 ± 6.82Mean CD4 ± Std Dv496.34 ± 313.09420.92 ± 231.72345.16 ± 302.99425.73 ± 308.46494.31 ± 311.10500.73 ± 269.75AIDS2329.442.240.538.270.6Mean BMI26. 15 ± 6.1025.4 ± 4.6625.40 ± 6.7523.86 ± 5.3029.82 ± 5.2427.94 ± 5.82High blood pressure3217.62021.632.429.4Weight loss32.32.442.283.8029.4Weight gain1.530.628.910.810064.7Nausea13.810.662.256.817.670.6Vomiting02.426.7272.935.3Diarrhea23.131.871.148.626.570.6Abdominal pain4.67.135.6278.841.2Constipation07.128.921.620.611.8Loss of appetite208.255.683.85.929.4Overeating4.610.624.410.855.958.8Shortness of breath2017.673.335.120.6100Chest palpitations1.5042.22.7082.4Chest pain01.244.48.12.958.8Heart attack7.75.90014.711.8Hospitalized or treated for heart failure4.65.9008.85.9Coronary artery disease4.63.52.208.80Irregular heart rate7.711.805.411.85.9Heart valve disorder1.54.74.405.90
## Least Symptomatic Clusters
Participants in both Black and non-Black least symptomatic clusters reported a lower incidence of GI and CV symptoms and comorbidities (heart attack, irregular heart rate) compared to the weight gain and weight loss clusters. However, participants in the Black least symptomatic cluster reported a higher incidence of high blood pressure ($32.0\%$ vs $17.6\%$) and weight loss ($32.3\%$ vs $2.4\%$) than their non-Black cluster counterparts. Weight gain ($30.6\%$ vs $1.5\%$), diarrhea ($31.8\%$ vs $23.1\%$), and AIDS ($29.4\%$ vs $23.1\%$) were reported more in the non-Black least symptomatic cluster compared to the Black least symptomatic cluster.
## Weight Loss Clusters
Participants who self-identified as Black in the weight loss cluster reported a higher incidence of all GI symptoms than non-Blacks with the most prominent being diarrhea ($71.1\%$ vs $48.6\%$) and nausea ($62.2\%$ vs $56.8\%$). CV symptoms including chest palpitations ($42.2\%$ vs $2.7\%$), chest pain ($44.4\%$ vs $8.1\%$), and shortness of breath ($73.3\%$ vs $35.1\%$) were more common in the Black cluster compared to the non-Black cluster. Interestingly, there were few to no reports of CV comorbidities (i.e., heart attack, heart failure, coronary artery disease) in the non-Black group. The most prominent symptoms of the non-Black weight loss cluster compared to the Black weight loss cluster were loss of appetite ($83.8\%$ vs $55.6\%$) and weight loss ($83.8\%$ vs $42.4\%$). Although participants in the non-Black cluster reported some CV symptoms (shortness of breath, chest palpitations, and chest pain), the incidence of CV comorbidities (irregular heart rate $2.5\%$) was very low.
## Weight Gain Clusters
Results revealed a high incidence of weight gain and overeating among participants in the weight gain clusters. Compared to the Black weight gain cluster, the non-Black weight gain cluster reported the highest incidence of AIDS ($70.6\%$ vs $38.2\%$), nausea ($70.6\%$ vs $17.6\%$), diarrhea ($70.6\%$ vs $26.5\%$), and shortness of breath ($58.8\%$ vs $20.6\%$). The Black weight gain cluster reported low incidence of GI symptoms (i.e., vomiting, abdominal pain), but a higher incidence of CV comorbidities than any other cluster ($14.7\%$).
## Discussion
The aim of this study was to investigate the association among race (Black, non-Black), weight status, (weight gain and weight loss), and symptoms/comorbidities in PWHIV. With advances in medications used to treat and manage HIV, PWHIV are living longer [22]. However, longevity of life predisposes individuals to developing chronic disease conditions common in aging [23]. Additionally, long-term use of HAART by individuals living with HIV can affect weight [17]. This combination of multiple morbidities and weight status affects health-related quality of daily life for PWHIV.
## Multiple Morbidities and Microbial Translocation/Disease Progression
Antiretroviral treatments aid in replenishing the CD4+ T-cell count and decreasing the viral load, which reduces inflammation in the gut-associated lymphoid tissue and promotes immune reconstitution. Results from the administered surveys suggested that the Black and non-Black participants in the least symptomatic clusters were able to manage their HIV status. Of the participants in the Black and non-Black least symptomatic cluster, $23\%$ vs $29.4\%$, respectively, reported an AIDS diagnosis. The lower incidence of AIDS had an association with fewer reports of GI and CV symptoms and comorbidity compared to the weight gain and weight loss clusters. This finding may indicate a reduced occurrence of the translocation of microbes and/or restored or improved CD4+ T-cell count due to an early initiation of HAART [24]. It may also be possible to surmise that the absence of an AIDS diagnosis is indicative of the efficacy and tolerability of HAART in participants in the least symptomatic cluster, which in turn suppresses the progression of HIV.
## Race/Ethnicity and Weight Status
In this study, PWHIV differed by race, weight status, and types of chronic disease conditions. Race included Black and non-Black participants. Weight status of the weight loss cluster and weight gain cluster differed by race in terms of GI and CV symptoms and comorbidities. While both Black and non-Black participants had increased weight loss and loss of appetite for the weight loss cluster, Black participants experienced the most incidence of diarrhea and non-Black participants had the least amount of CV symptoms. The weight gain cluster for Black participants had decreased GI symptoms with only one participant reporting vomiting. The non-Black weight gain group had a higher percentage of AIDS diagnosis. Several studies [17, 25] have likewise found these differences by race/ethnicity, comorbidities, and weight status among PWHIV.
The findings of this study related to the racial differences of symptoms within the same weight status cluster help to elucidate the need for precision healthcare. Precision care should also incorporate stress of neighborhood disadvantage and employment/socioeconomic status that often affects Black PWHIV. The practitioner examining a Black PWHIV that is having weight gain must consider CV symptoms and comorbidities, neighborhood disadvantage, and employment/socioeconomic status (SES) [11]. The non-Black PWHIV needs consideration for GI and CV symptoms and comorbidities along with an AIDS diagnosis. Simply treating PWHIV based on history may be ineffective as it may omit the required investigation and treatment required by each racial group [13].
Additionally, among the participants diagnosed with AIDS in this study, non-Black men had the highest prevalence of the disease. However, according to the Centers for Disease Control and Prevention, Black men have the highest prevalence of AIDS diagnoses [26]. This atypical burden of a higher number of reported AIDS diagnoses in non-Black males may be attributable to more non-Black males with AIDS enrolled in the study compared to their Black male counterparts ($52\%$ vs $47\%$). The higher prevalence of non-Black men with AIDS enrolled in this study could also be indicative of factors such as later diagnosis and treatment [27], or a lower tolerability of HAART.
## HAART and Weight
Initiating HAART has been associated with increasing BMI, and long-term use can lead to obesity [28]. The Black weight gain cluster had a higher BMI than the non-Black weight gain cluster ($29.8\%$ vs $27.9\%$). The higher BMI and self-reported weight gain could be attributed to a “return to health,” depression, or other factors such as unemployment [25] or metabolic syndrome [29]. According to several studies [17, 25], race attributed to the differences between weight gain and BMI. Participants who identified as Black were found more likely to have higher BMI and weight gain than their counterparts. The observed differences may be due to higher CD4+ T-cell count, a longer duration of HAART [7], and/or a higher pretreatment CD4+ T-cell count [17]. The chronic stress of neighborhood disadvantage and SES are associated with increased BMI and may contribute to the increased BMI for Black participants [11, 12]. The non-Black weight gain group reported a higher incidence of GI symptoms (loss of appetite, diarrhea, vomiting) compared to their counterparts in the Black weight gain group. Such symptoms typically are associated with weight loss [30]. A lower BMI has been shown to be associated with a higher mortality risk for patients on HAART [28].
Gastrointestinal symptoms such as nausea, vomiting, weight loss, and diarrhea are common in PWHIV [4, 31]. Although such symptoms were reported in both the weight loss and weight gain clusters, these symptoms were reported more frequently in the weight loss clusters. The GI symptoms could be attributable to the disease itself or a side effect of HAART medication [32]. A possible explanation for the higher incidence of weight loss and loss of appetite in the non-Black weight loss cluster could be a combination of multiple factors such as early vs late HAART initiation, antiretroviral efficacy and tolerability, disease progression, and race.
## Strengths
This study adds to the growing body of literature investigating health status among PWHIV. Our findings contribute new knowledge to the limited research examining the role of weight and its effects on GI symptoms and cardiovascular risks among PWHIV. Insights into the relationships among this tetrad (race, weight, GI symptoms, and cardiovascular risks) suggest a call for primary prevention to address and promote healthy weight status to improve health-related quality of life for PWHIV. The study reported herein purposefully implemented the latent class analysis (LCA), a type of structural equation modeling, to find groups or subsets of cases within the multivariant categorical data. Many analyses use race as a covariant and adjust for race and ethnicity and outcomes, thereby not purposefully including race and ethnicity as the main effect in the model. Our goal was to allow the data in an unsupervised fashion to be fit via the LCA method. Our findings are novel, specifically because of the analytical method.
## Limitations
The research design included a secondary data analysis that utilized a sample size of approximately 283 individuals with complete information on three questionnaires. As such, the sample size prevents (or limits) generalizability of the results as the secondary data analysis was not purposefully prospectively powered. Limited data points prevented in-depth statistical analysis. Additional care should be used in the interpretation of the findings as model fit analysis does not imply causation.
## Conclusions
Weight changes affect both GI and cardiovascular symptoms for PWHIV. As such, nutritional interventions may be beneficial for managing weight and reducing adverse effects for PWHIV [33]. Moreover, although the use of HAART medication is beneficial for treating and managing HIV infection, the long-term use of HAART medications may be problematic for healthy weight maintenance. Therefore, monitoring of weight status is important for PWHIV for reducing chronic disease conditions such as cardiovascular disease, hypertension, and diabetes; comorbid conditions that adversely affect health-related quality of life. Moreover, there is an unmet need for healthcare providers to recognize the symptomatic and comorbid factors that affect health outcomes in persons of diverse racial/ethnic backgrounds. Precision health initiatives that “take into account individual differences in people’s genes, environments and lifestyles,” and with the goal of “revolutionizing how we improve and treat disease,”[34] hold promise for healthcare treatments for PWHIV; such advances may indeed continue to prolong years of healthy living across the life span.
## References
1. Lakey W, Yang LY, Yancy W, Chow SC, Hicks C. **Short communication: From wasting to obesity: initial antiretroviral therapy and weight gain in HIV-infected persons**. *AIDS Res Hum Retroviruses.* (2013.0) **29** 435-40. DOI: 10.1089/AID.2012.0234
2. Mudd JC, Brenchley JM. **Gut mucosal barrier dysfunction, microbial dysbiosis, and their role in HIV-1 disease progression**. *J Infect Dis* (2016.0) **214** S58-66. DOI: 10.1093/infdis/jiw258
3. 3.Yoder AC, Guo K, Dillon SM, Phang T, Lee EJ, Harper S, et al. The transcriptome of HIV-1 infected intestinal CD4 + T cells exposed to enteric bacteria. 2017; 10.1371/journal.ppat.1006226.
4. Hall VP. **Common gastrointestinal complications associated with human immunodeficiency virus/AIDS: an overview**. *Crit Care Nurs Clin North Am [Internet]* (2018.0) **30** 101-107. DOI: 10.1016/j.cnc.2017.10.009
5. Ballocca F, D’Ascenzo F, Gili S, Grosso Marra W, Gaita F. **Cardiovascular disease in patients with HIV**. *Trends Cardiovasc Med [Internet]* (2017.0) **27** 558-563. DOI: 10.1016/j.tcm.2017.06.005
6. Manner IW, Baekken M, Kvale D, Oektedalen O, Pedersen M, Nielsen SD. **Markers of microbial translocation predict hypertension in HIV-infected individuals**. *HIV Med* (2013.0) **14** 354-361. DOI: 10.1111/hiv.12015
7. Achhra AC, Mocroft A, Reiss P, Sabin C, Ryom L, De WS. **Short-term weight gain after antiretroviral therapy initiation and subsequent risk of cardiovascular disease and diabetes : the D : A : D study ***. *HIV Med* (2016.0) **17** 20-25. DOI: 10.1111/hiv.12294
8. Lake JE. **The fat of the matter: obesity and visceral adiposity in treated HIV infection**. *Curr HIV/AIDS Rep* (2017.0) **14** 211-219. DOI: 10.1007/s11904-017-0368-6
9. Yuh B, Tate J, Butt AA, Crothers K, Freiberg M, Leaf D. **Weight change after antiretroviral therapy and mortality**. *Clin Infect Dis* (2015.0) **60** 1852-1859. DOI: 10.1093/cid/civ192
10. So Armah K, Freiberg MS. **HIV and cardiovascular disease: update on clinical events, special populations, and novel biomarkers**. *Curr HIV/AIDS Rep.* (2018.0) **15** 233-244. DOI: 10.1007/s11904-018-0400-5
11. Chirinos DA, Garcini LM, Seiler A, Murdock KW, Peek K, Stowe RP, Fagundes CDA. **Psychological and biological pathways linking perceived neighborhood characteristics and body mass index**. *Ann Behav Med* (2019.0) **53** 827-838. DOI: 10.1093/abm/kay092
12. Weihrauch-Blüher S, Richter M, Staege MS. **Body weight regulation, socioeconomic status and epigenetic alterations**. *Metabolism* (2018.0) **85** 109-115. DOI: 10.1016/j.metabol.2018.03.006
13. Hekler E, Tiro JA, Hunter CM, Nebeker C. **Precision health: the role of the social and behavioral sciences in advancing the vision**. *Ann Behav Med* (2020.0) **54** 805-826. DOI: 10.1093/abm/kaaa018
14. Carnethon MR, Pu J, Howard G, Albert MA, Anderson CAM, Bertoni AG. **Cardiovascular health in African Americans: a scientific statement from the American Heart Association**. *Circulation* (2017.0) **139** 393-423
15. Abara WE, Smith L, Zhang S, Fairchild AJ, Heiman HJ, Rust G. **The influence of race and comorbidity on the timely initiation of antiretroviral therapy among older persons living with HIV/AIDS**. *Am J Public Health* (2014.0) **104** e135-e141. DOI: 10.2105/AJPH.2014.302227
16. Kirk JB, Goetz MB. **Human immunodeficiency virus in an aging population, a complication of success**. *J Am Geriatr Soc* (2009.0) **57** 2129-2138. DOI: 10.1111/j.1532-5415.2009.02494.x
17. Koethe JR, Jenkins CA, Lau B, Shepherd BE, Justice AC, Tate JP. **Rising obesity prevalence and weight gain among adults starting antiretroviral therapy in the United States and Canada**. *AIDS Res Hum Retroviruses* (2016.0) **32** 50-58. DOI: 10.1089/aid.2015.0147
18. Brennan AT, Berry KM, Rosen S, Stokes A, George J, Raal F. **Growth curve modelling to determine distinct body mass index trajectory groups in HIV-positive adults on ART in South Africa**. *AIDS* (2021.0) **33** 2049-2059. DOI: 10.1097/QAD.0000000000002302.Growth
19. Kim DJ, Westfall AO, Chamot E, Willig AL, Mugavero MJ, Ritchie C. **Multimorbidity patterns in HIV-infected patients**. *JAIDS J Acquir Immune Defic Syndr* (2012.0) **61** 600-605. DOI: 10.1097/qai.0b013e31827303d5
20. 20.Guaraldi G, Malagoli A, Calcagno A, Mussi C, Celesia BM, Carli F, et al. The increasing burden and complexity of multi-morbidity and polypharmacy in geriatric HIV patients: a cross sectional. BMC Geriatr [Internet]. 2018;18:1–10G. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5910563/pdf/12877_2018_Article_789.pdf.
21. Henderson WA. *Testing a model of health-related quality of life in persons living with HIV and liver disease* (2007.0)
22. Ballocca F, Gili S, D’Ascenzo F, Marra WG, Cannillo M, Calcagno A. **HIV infection and primary prevention of cardiovascular disease: lights and shadows in the HAART era**. *Prog Cardiovasc Dis [Internet]* (2016.0) **58** 565-576. DOI: 10.1016/j.pcad.2016.02.008
23. D’Ascenzo F, Cerrato E, Calcagno A, Grossomarra W, Ballocca F, Omedè P. **High prevalence at computed coronary tomography of non-calcified plaques in asymptomatic HIV patients treated with HAART: a meta-analysis**. *Atherosclerosis* (2015.0) **240** 197-204. DOI: 10.1016/j.atherosclerosis.2015.03.019
24. Allers K, Puyskens A, Epple HJ, Schürmann D, Hofmann J, Moos V. **The effect of timing of antiretroviral therapy on CD4+ T-cell reconstitution in the intestine of HIV-infected patients**. *Mucosal Immunol* (2016.0) **9** 265-274. DOI: 10.1038/mi.2015.58
25. Olaleye AO, Owhonda G, Daramola O, Adejo I, Olayiwola H, Inyang JI. **Factors associated with weight gain among adult patients initiating antiretroviral therapy in Port Harcourt, Nigeria: a retrospective cohort study**. *Infect Dis (Auckl)* (2017.0) **49** 635-638. DOI: 10.1080/23744235.2017.1306102
26. 26.Centers for Disease Control and Prevention. Estimated HIV incidence and prevalence in the United States, 2014–2018. HIV Surveill Suppl Rep [Internet]. 2021;26(1), 1–81. http://www.cdc.gov/hiv/library/reports/hiv-surveillance.html. http://wwwn.cdc.gov/dcs/ContactUs/Form
27. Shiu ATY, Choi KC, Lee DTF, Yu DSF, Man NW. **Application of a health-related quality of life conceptual model in community-dwelling older Chinese people with diabetes to understand the relationships among clinical and psychological outcomes**. *J Diabetes Investig* (2014.0) **5** 677-686. DOI: 10.1111/jdi.12198
28. 28.de Pee, S. & Semba RD. Role of nutrition in HIV infection: review of evidence for more effective programming in resource-limited settings. Food Nutr Bull [Internet]. 2010;31(4):S313–344. https://www.ncbi.nlm.nih.gov/pubmed/21214036.
29. Dimala CA, Atashili J, Mbuagbaw JC, Wilfred A, Monekosso GL. **Prevalence of hypertension in HIV/AIDS patients on highly active antiretroviral therapy (HAART) compared with HAART-naïve patients at the Limbe Regional Hospital**. *Cameroon PLoS One* (2016.0) **11** 1-11. DOI: 10.1371/journal.pone.0148100
30. Evans D, McNamara L, Maskew M, Selibas K, Van Amsterdam D, Baines N. **Impact of nutritional supplementation on immune response, body mass index and bioelectrical impedance in HIV-positive patients starting antiretroviral therapy**. *Nutr J* (2013.0) **12** 1-14. DOI: 10.1186/1475-2891-12-111
31. 31.WHO. HIV and AIDS [Internet]. 2019. WHO, “HIV and AIDS.” 2019, [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/hiv-aids.
32. Gupta R, Ordonez RM, Koenig S. **Global impact of antiretroviral therapy-associated diarrhea**. *AIDS Patient Care STDS* (2012.0) **26** 711-713. DOI: 10.1089/apc.2012.0299
33. Mankal PK, Kotler DP. **From wasting to obesity, changes in nutritional concerns in HIV/AIDS**. *Endocrinol Metab Clin North Am [Internet]* (2014.0) **43** 647-663. DOI: 10.1016/j.ecl.2014.05.004
34. 34.The White House. Precision medicine initiatives. https://obamawhitehouse.archives.gov/precision-medicine
|
---
title: Surgery-related characteristics, efficacy, safety and surgical team satisfaction
of three-dimensional heads-up system versus traditional microscopic equipment for
various vitreoretinal diseases
authors:
- Xin-yu Zhao
- Qing Zhao
- Ning-ning Li
- Li-hui Meng
- Wen-fei Zhang
- Er-qian Wang
- You-xin Chen
journal: Graefe's Archive for Clinical and Experimental Ophthalmology
year: 2022
pmcid: PMC9988774
doi: 10.1007/s00417-022-05850-z
license: CC BY 4.0
---
# Surgery-related characteristics, efficacy, safety and surgical team satisfaction of three-dimensional heads-up system versus traditional microscopic equipment for various vitreoretinal diseases
## Abstract
### Purpose
To compare the three-dimensional (3D) heads-up surgery with the traditional microscopic (TM) surgery for various vitreoretinal diseases.
### Methods
A medical record review of patients that underwent 3D heads-up or TM vitreoretinal surgeries was performed from May 2020 to October 2021 in this retrospective case–control study. Main outcome measures included surgery-related characteristics, efficacy, safety, and satisfaction feedback from the surgical team.
### Results
A total of 220 ($47.6\%$) and 242 ($52.4\%$) eyes were included in the 3D and TM groups, respectively. The 3D heads-up system significantly benefits delicate surgical steps, like the epiretinal membrane (ERM) peeling for ERM and internal limiting membrane peeling for idiopathic macular holes ($P \leq 0.05$). The 3D heads-up system could facilitate a significantly better visual outcome for pathologic myopic foveoschisis ($$P \leq 0.049$$), while no difference by TM surgery ($$P \leq 0.45$$). For the satisfaction feedback, the 3D heads-up system was rated significantly higher in most subscales and the overall score ($P \leq 0.05$). The surgeons’ ratings on operating accuracy and the first assistants’ rating on operating accuracy and operation cooperation were significantly higher in the TM group than in the 3D group ($P \leq 0.05$). Besides that, the 3D heads-up surgery was comparable with TM surgery in the surgery-related characteristics, choice of tamponades, postoperative VA, primary anatomic success, and perioperative complications ($P \leq 0.05$).
### Conclusion
The efficacy and safety of the 3D heads-up surgery were generally comparable to the TM surgery. The 3D heads-up system could significantly benefit delicate surgical steps and achieve better surgical team satisfaction.
## Introduction
Three-dimensional (3D) heads-up surgery was first developed and applied in ophthalmic surgery in 2009, mostly in anterior segment surgery [1, 2]. In 2016, Eckardt et al. [ 3] firstly reported the application of the 3D heads-up system in more complicated vitrectomy surgery, increasing the popularity among ophthalmologists for treating vitreoretinal diseases. Nowadays, 3D heads-up systems have been widely used in multiple vitreoretinal surgeries, including macular membrane peeling for epiretinal membrane (ERM), repair of macular hole (MH) and rhegmatogenous retinal detachment (RRD), and vitrectomy for non-clearing vitreous hemorrhage (VH) and tractional retinal detachments (TRD). The current clinical or research- used 3D heads-up systems in ophthalmic surgeries included the Alcon NGENUITY® 3D Visualization System (Alcon Laboratories, Fort Worth, TX), the TrueVision Visualization System (Santa Barbara, CA), and the NCVideo3D system (NewComm, Beijing, China).
The 3D heads-up surgery system was reported to have multiple advantages over the traditional microscopic (TM) system, including high magnification performance, superior stereoscopic sensation, wide visual field, expanded depth of field, and reduced retinal phototoxicity, display image manipulation, improved ergonomics, and enhanced surgical team communication and education [3–9]. Previously reported disadvantages included the costly equipment, the learning curve required to use it efficiently, and the time latency between surgical interventions and their visualization [10, 11], while recent studies found that the time latency of the current 3D heads-up system may not jeopardize the surgical performance and outcome [12, 13].
However, several issues still needed to be settled. [ 1] *Most previous* studies evaluating the 3D heads-up system in vitreoretinal surgeries focused only on one single vitreoretinal disease [11, 14]. Studies with a larger sample size, multiple vitreoretinal diseases, and more comprehensive analysis were needed to better describe the merit and demerit of the 3D-heads-up system; [2] only a few studies have compared the outcomes of surgeries, for example, visual acuity (VA), primary anatomic success, and postoperative complications between surgeries using the 3D heads-up system and TM equipment, and their sample size was also limited. Thus, their conclusion might be unsolid [15, 16]; [3] whether 3D heads-up surgery was associated with longer surgical duration or longer learning curve remained controversial. Some studies reported 3D heads-up surgery with a longer learning curve and longer surgical duration, while other studies found no significant difference [16–18].
Our study aimed to investigate the surgery-related characteristics, efficacy, safety, and surgical team satisfaction feedback between the 3D heads-up surgery and TM surgery for common vitreoretinal diseases. The duration of specific surgical steps, visual outcomes, primary anatomic success rate, perioperative complications, and subjective assessment from the surgery team were compared in detail to obtain a more comprehensive description of the 3D heads-up surgery system and provide references for ophthalmologists.
## Study design
A medical record review of patients who underwent vitreoretinal surgeries using the 3D heads-up system (3D group) or TM equipment (TM group) was performed from May 2020 to October 2021. All patients were examined and treated by the same surgeon (YXC) at the Ophthalmology Department of Peking Union Medical College Hospital (PUMCH) in Beijing, China. This retrospective study was approved by the Institutional Review Board/Ethics Committee of PUMCH (No. S-K1944) and was conducted following the tenets of the Declaration of Helsinki. Written informed consent was provided to each patient before the surgery. All the healthcare staff presented in Fig. 1a had given informed consent for publication. Fig. 1a The surgical team during the surgery using the 3D heads-up visualization system. Every member in the surgical team wore passive polarized 3D glasses and viewed the surgical field on the Advanced NGENUITY® 3D 4 K OLED Display. The NGENUITY® 3D Visualization System and the CONSTELLATION® Vision System establish an Integrated Surgical Platform, which could monitor the real-time IOP, flow rate, and other surgical parameters. b The Alcon NGENUITY® 3D Visualization System. Abbreviations: ILM = internal limiting membrane; IOP = intraocular pressure; 2D = two-dimensional; 3D = three-dimensional
## Inclusion and exclusion criteria
The following inclusion criteria were used: [1] patients underwent vitreoretinal surgeries for ERM, vitreomacular traction syndrome (VMT), VH, TRD, MH, RRD, pathologic myopic (PM) foveoschisis, silicone oil removal (SOR), and vitreous opacities using the 3D heads-up system or the TM equipment; [2] patients with detailed medical records and underwent comprehensive ophthalmologic examination including the Snellen best-corrected visual acuity (BCVA), intraocular pressure (IOP), axial lengths (AL), slit-lamp biomicroscopy, optical coherence tomography (OCT), and fundus photograph (FP); [3] a minimum follow-up period of 3 months after the surgery. The exclusion criteria were the following: [1] any other concomitant ocular diseases that could confound the results of the included vitreoretinal diseases; [2] patients with insufficient medical data or lost to follow-up. When both eyes of one patient were eligible, both eyes were included in the study.
## Surgical procedure
All surgeries were performed with the Alcon Constellation surgery system (Alcon Laboratories, Fort Worth, TX) by the same surgeon (YXC) with experience in vitreoretinal surgeries for more than 30 years. The TM group used the traditional microscopic system (OPMI-Lumera 700 with ReSight; Carl Zeiss Meditec AG; Jena, Germany), and the 3D group used Alcon NGENUITY® 3D Visualization System (Alcon Laboratories, Fort Worth, TX). This 3D visualization system was mainly composed of the 3D High Dynamic Range NGENUITY® Camera, advanced NGENUITY® 3D 4 K OLED Display, and NGENUITY® DAVS Console (see Fig. 1b). All patients underwent standard 23-gauge or 25-gauge three-port pars plana vitrectomy (PPV) under local retrobulbar anesthesia or general anesthesia. All pre-, peri-, and pos-toperative regimens were the same in these two groups. After the eyes were disinfected with $5\%$ povidone-iodine and the conjunctiva was displaced by approximately 1–2 mm, trocar cannulas were inserted at a 20–30° angle into the conventional inferotemporal, superotemporal, and superonasal quadrants 3.5–4 mm posterior to the limbus. Surgical procedures vary according to the surgical indicators. Triamcinolone acetonide (TA), indocyanine green (ICG), liquid perfluorocarbon (C3F8), endodiathermy, retinotomy, and endolaser coagulation were applied as surgical adjuncts if necessary. The inverted internal limiting membrane (ILM) flap or the ILM insertion was applied in eyes with idiopathic MHs or PM-related MHs. The ILM around the fovea was peeled in the eyes with PM foveoschisis. Fluid-air exchange and tamponades of air, balanced salt solution (BSS), 10–$14\%$ C3F8, or silicone oil were performed based on the operating surgeon’s discretion when indicated.
## Data collection
Information extracted from the medical records of patients included age, gender, operative eye, AL, ocular and surgical history, diagnosis and surgical indicators, surgical procedures, choice of tamponades, pre- and pos-toperative Snellen BCVA, pre- and pos-toperative IOP, perioperative complications, general surgical duration, and duration of specific steps (e.g., ILM peeling). Twenty surgeries were randomly selected in a 1:1 ratio from the 3D group and the TM group by an independent analyzer. The satisfaction questionnaires evaluating surgery-related characteristics (e.g., resolution, magnification, depth of the field) and general satisfaction feedback to the surgical system were requested to be finished by the surgeon, first assistant, instrument nurses, and visitors immediately after the surgery. The postoperative follow-up was scheduled at approximately 1 week, 1 month, and 3 months, with the measurement of Snellen BCVA, IOP, FP, etc.
## Outcome measures
The main outcome measures included Snellen BCVA, primary anatomic success, general surgical duration, duration of specific steps, perioperative complications, and satisfaction feedback from the surgical team. The Snellen BCVA was converted to the logarithm of the minimum angle of resolution (logMAR) equivalents for statistical analysis [19]. No light perception (NLP) was set at 2.90 logMAR, light perception (LP) at 2.60 logMAR, hand movements (HM) at 2.30 logMAR, and fingers counting (FC) at 1.85 logMAR [20]. The definition of anatomic success varied according to the surgical indicators and was evaluated by two retinal specialists (XYZ and QZ). The primary anatomic success was defined as complete removal of ERM for eyes with ERM, relieving of VMT for VMT, clearance of VH and vitreous opacities for VH, reattachment of the retina for TRD and RRD, closure of MH for MH, recovery of PM foveoschisis for PM foveoschisis, removal of silicone oil and the attachment of retina for SOR, and disappearance of vitreous opacities for vitreous opacities. The duration of ILM peeling was defined as from the starting of ICG injection to the finishing of ILM peeling. *The* general surgical duration was defined as starting the trocar insertion to finishing the wound sealing. Ocular hypertension was defined as IOP ≥ 21 mmHg during the postoperative follow-up.
## Statistical analysis
Data were analyzed by univariable analysis by comparing each aforementioned parameter between the 3D group and the TM group. Analyses were performed independently for each subgroup of vitreoretinal disease. Continuous variables were summarized as mean ± standard deviation (SD) and categorical data were presented as frequency (percentages). The independent t-test and two-tailed, paired t-test were used to evaluate comparative statistical analyses. The chi-squared test or Fisher’s exact test was used to examine categorical variables. All statistical analyses were performed with Stata SE 12.0 software (StataCorp, College Station, TX, USA). The two-tailed p-value < 0.05 was considered statistically significant.
## Results
A total of 426 patients and 462 eyes were finally enrolled, of which 220 ($47.6\%$) and 242 ($52.4\%$) eyes were included in the 3D and TM groups, respectively. Among the included patients, 218 ($51.2\%$) were female and 208 ($48.8\%$) were male, with a mean age of 55.00 ± 14.36 years. Age, AL, and pseudophakic eye showed no statistical differences ($P \leq 0.05$) between the 3D and TM groups (see Table 1).Table 1Baseline demographics of patients in the 3D group and the TM groupSurgical indicatorsPatients (eyes)Group Size (eyes)Age (mean ± SD, years)AL (mean ± SD, mm)Pseudophakic eye (n, %)TM3DTM3DPTM3DPTM3DPIdiopathic ERM66[70]393164.35 ± 14.1861.08 ± 14.130.3423.17 ± 1.2123.66 ± 1.260.107(17.9)9(29.0)0.27VMT12[12]5761.25 ± 13.6764.67 ± 12.790.6724.67 ± 1.2723.47 ± 1.050.100[0]1(14.3)0.86VH With TRD42[47]252248.23 ± 12.4646.45 ± 12.070.5122.88 ± 1.4323.58 ± 1.720.138(32.0)9(40.9)0.53 Without TRD80[85]513457.19 ± 11.8755.58 ± 12.380.5523.74 ± 0.9923.08 ± 2.260.079(17.6)11(32.4)0.12MH Idiopathic MH32[34]142057.75 ± 14.2862.47 ± 12.560.3123.79 ± 1.4823.63 ± 1.390.752(14.3)4(20.0)0.98 PM-related MH24[29]121760.57 ± 10.9157.13 ± 10.360.4028.53 ± 2.1828.89 ± 2.210.672(16.7)2(11.8)0.87RRD Primary RRD43[45]252053.83 ± 12.2548.67 ± 13.280.1825.56 ± 3.4425.58 ± 4.570.993(12.0)1(5.0)0.77 PVR-related RRD29[33]191445.43 ± 15.3950.90 ± 16.050.3324.69 ± 2.7523.88 ± 2.990.434(21.1)6(42.9)0.18PM foveoschisis13[13]4949.75 ± 16.5252.75 ± 13.750.7427.52 ± 2.1428.08 ± 1.880.640[0]0[0]NASOR For RRD34[34]142055.27 ± 14.1951.32 ± 14.710.4425.76 ± 3.8225.81 ± 4.230.974(28.5)7(35.0)0.98 For TRD47[52]302246.89 ± 16.6450.09 ± 14.830.4824.69 ± 2.7523.88 ± 2.990.327(23.3)5(22.7)0.96Vitreous opacities4[8]4449.23 ± 15.9145.00 ± 20.790.7632.44 ± 2.5134.17 ± 3.990.490[0]0 0)NAAll426[462]24222054.79 ± 14.1954.32 ± 14.540.7324.52 ± 2.2824.74 ± 2.630.3446(19.0)55(25.0)0.12AL axial length, ERM epiretinal membrane, MH macular hole, PM pathologic myopic, PVR proliferative vitreoretinopathy, RRD rhegmatogenous retinal detachment, SD standard deviation, SOR silicone oil removal, TM traditional microscopic, TRD tractional retinal detachments, 3D three-dimensional, VH vitreous hemorrhage, VMT vitreomacular traction syndrome The duration of ERM or ILM peeling for eyes with ERM and idiopathic MH was significantly shorter in the 3D group than in the TM group (ERM: 6.12 ± 2.45 versus 9.55 ± 5.34, $P \leq 0.01$; idiopathic MH: 6.03 ± 2.12 versus 9.01 ± 4.06, $$P \leq 0.01$$). Compared with the TM group, the 3D group was associated with significantly shorter general surgical duration for eyes with ERM (20.14 ± 6.72 versus 24.81 ± 8.62, $$P \leq 0.02$$) and idiopathic MH (22.13 ± 6.58 versus 26.75 ± 5.94, $$P \leq 0.04$$). No significant difference existed in the duration of complete vitrectomy and proliferating membrane peeling, the choice of tamponades, and the incidence of iatrogenic retinal breaks between the 3D group and TM group ($P \leq 0.05$) (see Table 2).Table 2Surgical duration and the choice of tamponades in the 3D group and the TM groupTM3DPDuration of ILM peeling (mean ± SD, min) ERM9.55 ± 5.346.12 ± 2.45 < 0.01* VMT8.69 ± 4.295.99 ± 3.790.28 Idiopathic MH9.01 ± 4.066.03 ± 2.120.01* PM-related MH15.74 ± 7.9312.97 ± 5.830.29 PM foveoschisis16.74 ± 8.4112.89 ± 6.740.40Duration of complete vitrectomy (mean ± SD, min) VH with TRD21.14 ± 14.8419.31 ± 8.190.61 VH without TRD14.33 ± 3.3215.57 ± 5.880.22 Primary RRD16.49 ± 6.8517.71 ± 7.420.57 PVR-related RRD23.82 ± 15.6922.37 ± 14.280.79Duration of proliferating membrane peeling (mean ± SD, min) VH with TRD22.98 ± 15.0925.71 ± 16.270.55 PVR-related RRD24.15 ± 14.3322.62 ± 13.110.76General surgical duration (mean ± SD, min) ERM24.81 ± 8.6220.14 ± 6.720.02* VH With TRD54.98 ± 24.9249.04 ± 26.830.44 Without TRD37.13 ± 22.0535.62 ± 19.280.75 MH Idiopathic MH26.75 ± 5.9422.13 ± 6.580.04* PM-related MH29.72 ± 15.8927.72 ± 9.920.68 RRD Primary RRD37.41 ± 8.7736.51 ± 9.290.74 PVR-related RRD54.74 ± 21.0748.55 ± 25.010.45 PM foveoschisis34.21 ± 12.2132.11 ± 9.180.74 SOR For RRD16.67 ± 8.1416.09 ± 5.220.80 For TRD25.45 ± 14.3926.88 ± 15.980.74 Vitreous opacities12.88 ± 3.3113.64 ± 4.270.79Choice of tamponades (n, %)3931 ERM Air6(15.4)5(16.1) BSS30(76.9)25(80.6) C3F83(7.7)1(3.2) Silicone oil0[0]0[0]0.17 VMT57 Air1(20.0)2(28.6) BSS4(80.0)5(71.4) C3F80[0]0[0] Silicone oil0[0]0[0]0.74 VH with TRD2522 Air0[0]0[0] BSS0[0]0[0] C3F80[0]0[0] Silicone oil25[100]22[100]NA VH without TRD5134 Air10(19.6)6(17.6) BSS35(68.6)24(70.6) C3F86(11.8)4(11.8) Silicone oil0[0]0[0]0.97 Idiopahtic MH1420 Air10(71.4)16(80.0) C3F84[29]4(20.0) Silicone oil0[0]0[0]0.56 PM-related MH1217 Air0[0]0[0] C3F89(75.0)15(88.2) Silicone oil3(25.0)2(11.8)0.35 PM foveoschisis49 C3F82(50.0)5(55.6) Silicone oil2(50.0)4(44.4)0.85 Primary RRD2520 C3F811(44.0)9(45.0) Silicone oil14(56.0)11(55.0)0.95 PVR-related RRD1914 C3F86(31.5)2(14.3) Silicone oil13(68.4)12(85.7)Incidence of iatrogenic retinal breaks (n, %)4(1.6)6(2.7)0.11*$P \leq 0.05$BSS balanced salt solution, C3F8 perfluoropropane, ERM epiretinal membrane, ILM internal limiting membrane, MH macular hole, NA not available, PM pathologic myopic, PVR proliferative vitreoretinopathy, RRD rhegmatogenous retinal detachment, SD standard deviation, SOR silicone oil removal, TM traditional microscopic, TRD tractional retinal detachments, 3D three-dimensional, VH vitreous hemorrhage, VMT vitreomacular traction syndrome For eyes with idiopathic ERM, VH, idiopathic MH, RRD, and SOR, postoperative VA improved significantly compared with the preoperative VA both in the 3D group and the TM group ($P \leq 0.05$). For eyes with PM foveoschisis, significant postoperative VA improvement was noticed in the 3D group (0.57 ± 0.38 versus 1.00 ± 0.47, $$P \leq 0.049$$) but not in the TM group (0.63 ± 0.55 versus 0.97 ± 0.63, $$P \leq 0.45$$). No significant difference in preoperative VA, postoperative VA, primary anatomic success rate, and perioperative complications between the 3D group and the TM group ($P \leq 0.05$) (see Table 3). Table 3Visual acuity, anatomic success, and perioperative complications in the 3D group and the TM groupPreoperative BCVAPostoperative BCVAP (Pre-versus Post-)Primary anatomic success rate (n,%)ComplicationsOcular hypertension (n,%)VH (n,%)RD (n,%)MH (n,%)TM3DPTM3DPTM3DTM3DPTM3DPTM3DPTM3DPTM3DPIdiopathic ERM ($\frac{39}{31}$)0.78 ± 0.610.81 ± 0.570.840.45 ± 0.400.48 ± 0.590.800.01*0.03*39[100]31[100]NA0[0]2(6.5)0.380[0]0[0]NA0[0]0[0]NA0[0]0[0]NAVMT($\frac{5}{7}$)0.95 ± 0.510.91 ± 0.680.910.56 ± 0.380.59 ± 0.480.910.210.335[100]7[100]NA1(20.0)2(28.6)0.740[0]0[0]NA0[0]0[0]NA0[0]0[0]NAVH With TRD ($\frac{25}{22}$)1.66 ± 0.641.77 ± 0.620.551.02 ± 0.551.07 ± 0.830.81 < 0.01* < 0.01*20(80.0)18(81.8)0.835[20]4(18.2)0.832(8.0)2(9.1)0.703(12.0)2(9.1)0.880[0]0[0]NA Without TRD ($\frac{51}{34}$)1.61 ± 0.751.65 ± 0.760.810.39 ± 0.620.31 ± 0.790.60 < 0.01* < 0.01*48(94.1)33(97.1)0.925(9.8)5(14.7)0.733(5.9)1(2.9)0.920[0]0[0]NA0[0]0[0]NAMH Idiopathic MH ($\frac{14}{20}$)1.02 ± 0.300.95 ± 0.350.550.74 ± 0.320.69 ± 0.280.630.02*0.01*12(85.7)20[100]0.322(14.3)[0]0.320[0]0[0]NA0[0]0[0]NA2(14.3)0[0]0.32 PM-related MH ($\frac{12}{17}$)1.17 ± 0.691.12 ± 0.530.830.97 ± 0.920.95 ± 0.680.950.550.429(75.0)15(88.2)0.673(25.0)4(23.5)0.730[0]0[0]NA0[0]0[0]NA3(25.0)2(11.8)0.67RRD Primary RRD ($\frac{25}{20}$)1.49 ± 0.771.44 ± 0.680.820.98 ± 0.740.91 ± 0.540.730.02*0.01*22(88.0)18(90.0)0.797(28.0)3(15.0)0.500[0]0[0]NA3(12.0)2(10.0)0.790[0]0[0]NA PVR-related RRD ($\frac{19}{14}$)1.62 ± 0.731.65 ± 0.580.901.10 ± 0.591.09 ± 0.610.940.02*0.02*14(73.6)11(78.6)0.936(31.6)2(14.3)0.460[0]0[0]NA5(26.3)3(21.4)0.930[0]0[0]NAPM foveoschisis($\frac{4}{9}$)0.97 ± 0.631.00 ± 0.470.930.63 ± 0.550.57 ± 0.380.820.450.049*3(75.0)8(88.9)0.851(25.0)2(22.2)0.550[0]0[0]NA0[0]0[0]NA0[0]0[0]NASOR For RRD ($\frac{14}{20}$)1.09 ± 0.561.13 ± 0.640.850.69 ± 0.450.66 ± 0.490.860.047*0.01*12(85.7)19(95.0)0.751(7.1)1(5.0)0.630[0]0[0]NA2(14.3)1(5.0)0.750[0]0[0]NA For TRD ($\frac{30}{22}$)1.21 ± 0.621.18 ± 0.570.860.80 ± 0.510.79 ± 0.620.950.01*0.045*24(80.0)18(81.8)0.872(6.7)1(4.5)0.780[0]0[0]NA6(20.0)4(18.2)0.850[0]0[0]NAVitreous opacities ($\frac{4}{4}$)0.14 ± 0.230.11 ± 0.190.850.10 ± 0.200.09 ± 0.290.960.800.914[100]4[100]NA0[0]0[0]NA0[0]0[0]NA0[0]0[0]NA0[0]0[0]NAAll ($\frac{242}{220}$)1.29 ± 0.661.31 ± 0.620.740.70 ± 0.550.69 ± 0.600.85 < 0.01* < 0.01*212(87.6)202(91.8)0.1433(13.6)26(11.8)0.565(2.1)3(1.4)0.8319(7.9)12(5.5)0.305(2.1)2(0.9)0.53*$P \leq 0.05$BCVA best-corrected visual acuity, ERM epiretinal membrane, MH macular hole, NA not available, PM pathologic myopic, PVR proliferative vitreoretinopathy, RD retinal detachment, RRD rhegmatogenous retinal detachment, SOR silicone oil removal, TM traditional microscopic, TRD tractional retinal detachments, 3D three-dimensional, VH vitreous hemorrhage, VMT vitreomacular traction syndrome *In* general, satisfaction feedback to the surgical system, the 3D heads-up system was rated significantly higher in most of the subscales ($P \leq 0.05$) and the overall score (212.48 ± 18.52 versus 160.17 ± 25.32, $P \leq 0.01$). The surgeons’ rating on instrument adjustment was comparable in these two groups (8.89 ± 1.30 versus 9.15 ± 0.62, $$P \leq 0.58$$). The TM group was rated significantly higher in the operating accuracy of the surgeon (9.56 ± 0.53 versus 8.22 ± 1.30, $$P \leq 0.01$$) and the operating accuracy (9.28 ± 0.80 versus 5.12 ± 2.21, $P \leq 0.01$) and operation cooperation (9.34 ± 0.71 versus 6.06 ± 2.43, $P \leq 0.01$) of the first assistant (see Table 4).Table 4General satisfaction feedback to the surgical system in the 3D group and the TM groupTM3DPSurgeonResolution of the lesion7.44 ± 1.519.56 ± 0.73 < 0.01*Magnification6.33 ± 1.229.44 ± 0.73 < 0.01*Depth of Field7.78 ± 0.979.00 ± 0.71 < 0.01*Operating accuracy9.56 ± 0.538.22 ± 1.300.01*Comfort level6.00 ± 1.129.44 ± 0.73 < 0.01*Instrument adjustment9.15 ± 0.628.89 ± 1.300.58Operation cooperation5.81 ± 1.789.51 ± 0.72 < 0.01*General satisfaction8.12 ± 1.599.37 ± 0.880.04*First assistantResolution of the lesion6.89 ± 1.769.12 ± 0.82 < 0.01*Magnification5.17 ± 2.029.15 ± 0.81 < 0.01*Depth of Field7.41 ± 1.128.91 ± 0.95 < 0.01*Operating accuracy9.28 ± 0.805.12 ± 2.21 < 0.01*Operation cooperation9.34 ± 0.716.06 ± 2.43 < 0.01*Comfort level5.87 ± 1.769.30 ± 0.68 < 0.01*General satisfaction7.25 ± 1.699.21 ± 0.81 < 0.01*Instrument nursesUnderstanding of surgical process5.05 ± 2.159.00 ± 0.75 < 0.01*Instrument preparation5.75 ± 1.849.10 ± 0.78 < 0.01*Active operation cooperation6.02 ± 1.689.33 ± 0.66 < 0.01*Comfort level5.00 ± 2.159.25 ± 0.72 < 0.01*General satisfaction6.50 ± 1.759.13 ± 0.68 < 0.01*VisitorUnderstanding of surgical process4.82 ± 2.659.25 ± 0.75 < 0.01*Resolution of the lesion4.13 ± 2.719.31 ± 0.66 < 0.01*Magnification3.75 ± 2.899.45 ± 0.62 < 0.01*Comfort level3.50 ± 2.208.52 ± 1.29 < 0.01*General satisfaction4.25 ± 2.559.40 ± 0.58 < 0.01*OverallOverall score160.17 ± 25.32212.48 ± 18.52 < 0.01**$P \leq 0.05$TM traditional microscopic, 3D three-dimensional
## Discussion
Our study evaluated the surgery-related characteristics, surgical outcomes, perioperative complications, and surgical team satisfaction between the 3D heads-up surgery system and the TM surgery for multiple vitreoretinal diseases. The results suggested that the 3D heads-up surgery system could significantly benefit in some delicate surgical steps, like ERM peeling for ERM and ILM peeling for idiopathic MH. The 3D heads-up system could facilitate a significantly better visual outcome for PM foveoschisis, while no difference existed with TM surgery. For the general satisfaction feedback, the 3D heads-up system was rated significantly higher in most of the subscales and the overall score, while the surgeons’ rating on operating accuracy, as well as the first assistants’ rating on operating accuracy and operation cooperation, were significantly higher in the TM group than the 3D group. Apart from the aforementioned findings, the 3D heads-up surgery system was as effective and safe as TM surgery in regards to the surgery-related characteristics, choice of tamponades, postoperative VA, primary anatomic success, and perioperative complications.
In our study, the duration of ILM peeling was significantly shorter in the 3D group than in the TM group for eyes with ERM or idiopathic MH. The possible reason could be that the 3D heads-up surgery has the advantage of high image magnification at a wider visual field compared with the TM system [21], the OPMI-Lumera 700 with ReSight, which enables surgeons to view the fine structures of the retina and then perform the membrane peeling more precisely. For surgical steps with less request of precision, such as complete vitrectomy, the duration was comparable between the 3D heads-up surgery and the TM surgery. This suggested that the advantage of the 3D heads-up system was more obvious when performing surgical steps of high precision [10, 21]. However, the inferiority of the TM system in handling precise surgical steps might also be associated with the specific TM operating system. Further research with other TM viewing operating systems was expected to compare the ability to handle precise surgical steps of the 3D heads-up system and the TM system.
Previous studies reported difficulties in controlling the depth of surgical operation using the 3D heads-up system. This might induce intraoperative complications like iatrogenic retinal breaks and require changing intraocular tamponades. Piccirillo et al. [ 11] reported the occurrence of 3 iatrogenic macular soft contusions in 10 procedures using the 3D heads-up system for vitreoretinal surgery, with no major retinal hemorrhages occurring. Our study found no difference in the incidence of iatrogenic retinal breaks or the choice of vitreous tamponades between the TM group and 3D group, and no differences existed in the occurrence of perioperative complications, including ocular hypertension, VH, RD, and MH. This indicated that the safety of the 3D heads-up surgery is comparable to the TM surgery [14].
The 3D heads-up surgery was comparable to the TM surgery in postoperative VA and primary anatomic success rate, which was in accordance with the previous studies [3, 10, 11, 16, 18, 21–23]. Although the 3D heads-up system could significantly shorten the duration of ILM peeling for eyes with ERM or idiopathic MH, no significant difference existed in the postoperative VA. However, the postoperative VA for eyes with PM foveoschisis significantly improved in the 3D group but not in the TM group. This indicated that the higher resolution of the 3D heads-up system enables more precise operations and better releasing of the retina, therefore promoting the recovery of the PM foveoschisis and the rehabilitation of visual function. The possible bias might exist due to the limited sample size of PM foveoschisis, and this finding should be confirmed in further studies.
In TM surgery, only the surgeon can have a high-grade stereo view of the surgical field, while the remaining observers (e.g. first assistant, instrument nurses, and visitors) could not appreciate the depth of field and the 3D view necessary for fine operations. However, by viewing a larger high-resolution screen of the 3D heads-up system, all members of the surgical team present can have access to the same live surgical image just as the surgeon. This provides significant improvements in operation cooperation and achieves additional pedagogical advantages [13, 24]. In this light, the surgeon could teach more easily and allow students or trainee surgeons to operate by reducing their installation time. In previous studies, surgeons and residents have rated the 3D system with improved ergonomics over TM [9]. The TM equipment was associated with more complaints of musculoskeletal pain because of the prolonged static unnatural neck-bent positions [25]. In contrast, by wearing polarized 3D glasses and viewing the surgical field on the 3D monitor without looking through microscope eyepieces in the neck-bent position, surgeons could turn their heads up through the 3D heads-up surgery system. Results obtained from the questionnaire designed ourselves showed that the 3D heads-up system was favored over the TM system in all subscales by instrument nurses and visitors. However, the first assistant rated lower scores in the subscale of “Operation cooperation” when using the 3D heads-up system, which was in accordance with the previously published study [7]. Rizzo et al. [ 7] evaluated the perceptions of the surgical team to the 3D surgical viewing system and recorded the first assistants’ dissatisfaction with the question “second surgeon’s comfort during surgery.” The first assistant has to rotate his head uncomfortably to look at the screen. The first assistant has to bear the inconsistency between the direction of the screen and the direction of the surgical steps such as trimming and pressuring, further increasing the difficulties of the surgery. However, this disadvantage could be overcome if the first assistant performed these surgical steps using the assistant microscope.
We summarized five keys to the satisfactory surgical experience of using the 3D heads-up system. Firstly, set the aperture at $30\%$ to get the best brightness and depth of field. Secondly, set the white balance to eliminate chromatic aberration and restore the original color. Thirdly, set the display screen 1–1.2 m away from the surgeon’s eye level to achieve the best resolution and 3D effect. And the center of the display screen should be set perpendicular to the surgeon’s visual axis to reduce the double image of peripheral images. Fourthly, the focus should be adjusted occasionally. For the anterior segment, zoom in to the maximum, fine the focus on the iris to get a clear image, and then zoom out to the suitable image size. For the posterior segment, ensure that the non-contact lens is in the shortest position and is placed well. In the targeted surgical area, zoom in to the maximum, focus to the clearest, and then zoom out to the appropriate image size. Finally, the surgical field should fill the display screen, which could bring a better depth of field, resolution, and contrast, making the image more immersive.
Several limitations of our study should be noted. Firstly, OCT and FP were not assigned to every patient in each follow-up, and some information could not be extracted due to the retrospective nature of this study. Secondly, the last follow-up period was only 3 months which might have underestimated the rate of perioperative complications and overestimated the primary anatomic success rate. Thirdly, the detailed satisfaction feedback for surgically treating specific types of vitreoretinal disease was not investigated in our study because of the limited amount of feedback. Further studies with a prospective design and a longer follow-up period are needed to confirm our findings.
In summary, the efficacy and safety of the 3D heads-up surgery were generally comparable to the TM surgery. The 3D system could significantly benefit delicate surgical steps. Moreover, the 3D heads-up surgery performed better in the surgical team satisfaction. The 3D heads-up system, with all these advantages, could be recommended for patients with vitreoretinal diseases, especially those with ERM or idiopathic MH.
## References
1. Bressler SB, Almukhtar T, Bhorade A. **Repeated intravitreous ranibizumab injections for diabetic macular edema and the risk of sustained elevation of intraocular pressure or the need for ocular hypotensive treatment**. *JAMA Ophthalmol* (2015.0) **133** 589-597. DOI: 10.1001/jamaophthalmol.2015.186
2. Galvis V, Berrospi RD, Arias JD, Tello A. **Bernal JC (2017) Heads up Descemet membrane endothelial keratoplasty performed using a 3D visualization system**. *J Surg Case Rep* (2017.0) **11** rjx231. DOI: 10.1093/jscr/rjx231
3. Eckardt C, Paulo EB. **Heads-up surgery for vitreoretinal procedures: an experimental and clinical study**. *Retina* (2016.0) **36** 137-147. DOI: 10.1097/iae.0000000000000689
4. Adam MK, Thornton S, Regillo CD, Park C, Ho AC, Hsu J. **Minimal endoillumination levels and display luminous emittance during three-dimensional heads-up vitreoretinal surgery**. *Retina* (2017.0) **37** 1746-1749. DOI: 10.1097/iae.0000000000001420
5. Kunikata H, Abe T, Nakazawa T. **Heads-up macular surgery with a 27-gauge microincision vitrectomy system and minimal illumination**. *Case Rep Ophthalmol* (2016.0) **7** 265-269. DOI: 10.1159/000452993
6. Chhaya N, Helmy O, Piri N, Palacio A, Schaal S. **Comparison of 2d and 3d video displays for teaching vitreoretinal surgery**. *Retina* (2018.0) **38** 1556-1561. DOI: 10.1097/iae.0000000000001743
7. Rizzo S, Abbruzzese G, Savastano A. **3D surgical viewing system in ophthalmology: perceptions of the surgical team**. *Retina* (2018.0) **38** 857-861. DOI: 10.1097/iae.0000000000002018
8. Nariai Y, Horiguchi M, Mizuguchi T, Sakurai R, Tanikawa A. **Comparison of microscopic illumination between a three-dimensional heads-up system and eyepiece in cataract surgery**. *Eur J Ophthalmol* (2021.0) **31** 1817-1821. DOI: 10.1177/1120672120929962
9. Zhang Z, Wang L, Wei Y, Fang D, Fan S, Zhang S. **The preliminary experiences with three-dimensional heads-up display viewing system for vitreoretinal surgery under various status**. *Curr Eye Res* (2019.0) **44** 102-109. DOI: 10.1080/02713683.2018.1526305
10. Talcott KE, Adam MK, Sioufi K. **Comparison of a three-dimensional heads-up display surgical platform with a standard operating microscope for macular surgery**. *Ophthalmol Retina* (2019.0) **3** 244-251. DOI: 10.1016/j.oret.2018.10.016
11. Piccirillo V, Sbordone S, Sorgente F. **To estimate the safety and efficacy of a 3-D visualization helmet for vitreoretinal surgery**. *Acta Ophthalmol* (2021.0) **99** e346-e351. DOI: 10.1111/aos.14591
12. 12.Ta Kim D, Chow D (2021) The effect of latency on surgical performance and usability in a three-dimensional heads-up display visualization system for vitreoretinal surgery. Graefes Arch Clin Exp Ophthalmol 260(2):471–476. 10.1007/s00417-021-05388-6
13. Berquet F, Henry A, Barbe C. **Comparing heads-up versus binocular microscope visualization systems in anterior and posterior segment surgeries: a retrospective study**. *Ophthalmologica* (2020.0) **243** 347-354. DOI: 10.1159/000507088
14. Asani B, Siedlecki J, Schworm B. **3D Heads-up display vs. standard operating microscope vitrectomy for rhegmatogenous retinal detachment**. *Front Med (Lausanne)* (2020.0) **7** 615515. DOI: 10.3389/fmed.2020.615515
15. Kantor P, Matonti F, Varenne F. **Use of the heads-up NGENUITY 3D visualization system for vitreoretinal surgery: a retrospective evaluation of outcomes in a french tertiary center**. *Sci Rep* (2021.0) **11** 10031. DOI: 10.1038/s41598-021-88993-z
16. Zhang T, Tang W, Xu G. **comparative analysis of three-dimensional heads-up vitrectomy and traditional microscopic vitrectomy for vitreoretinal diseases**. *Curr Eye Res* (2019.0) **44** 1080-1086. DOI: 10.1080/02713683.2019.1612443
17. Palácios RM, Maia A, Farah ME, Maia M. **Learning curve of three-dimensional heads-up vitreoretinal surgery for treating macular holes: a prospective study**. *Int Ophthalmol* (2019.0) **39** 2353-2359. DOI: 10.1007/s10792-019-01075-y
18. Romano MR, Cennamo G, Comune C. **Evaluation of 3D heads-up vitrectomy: outcomes of psychometric skills testing and surgeon satisfaction**. *Eye (Lond)* (2018.0) **32** 1093-1098. DOI: 10.1038/s41433-018-0027-1
19. Tiew S, Lim C, Sivagnanasithiyar T. **Using an excel spreadsheet to convert Snellen visual acuity to LogMAR visual acuity**. *Eye (Lond)* (2020.0) **34** 2148-2149. DOI: 10.1038/s41433-020-0783-6
20. Schulze-Bonsel K, Feltgen N, Burau H, Hansen L, Bach M. **Visual acuities “hand motion” and “counting fingers” can be quantified with the freiburg visual acuity test**. *Invest Ophthalmol Vis Sci* (2006.0) **47** 1236-1240. DOI: 10.1167/iovs.05-0981
21. Kannan NB, Jena S, Sen S, Kohli P, Ramasamy K. **A comparison of using digitally assisted vitreoretinal surgery during repair of rhegmatogenous retinal detachments to the conventional analog microscope: A prospective interventional study**. *Int Ophthalmol* (2021.0) **41** 1689-1695. DOI: 10.1007/s10792-021-01725-0
22. Palácios RM, de Carvalho ACM, Maia M, Caiado RR, Camilo DAG, Farah ME. **An experimental and clinical study on the initial experiences of Brazilian vitreoretinal surgeons with heads-up surgery**. *Graefes Arch Clin Exp Ophthalmol* (2019.0) **257** 473-483. DOI: 10.1007/s00417-019-04246-w
23. Palácios RM, Kayat KV, Morel C. **Clinical study on the initial experiences of french vitreoretinal surgeons with heads-up surgery**. *Curr Eye Res* (2020.0) **45** 1265-1272. DOI: 10.1080/02713683.2020.1737136
24. Mendez BM, Chiodo MV, Vandevender D, Patel PA. **Heads-up 3D microscopy: an ergonomic and educational approach to microsurgery**. *Plast Reconstr Surg Glob Open* (2016.0) **4** e717. DOI: 10.1097/gox.0000000000000727
25. Diaconita V, Uhlman K, Mao A, Mather R. **Survey of occupational musculoskeletal pain and injury in Canadian ophthalmology**. *Can J Ophthalmol* (2019.0) **54** 314-322. DOI: 10.1016/j.jcjo.2018.06.021
|
---
title: Impact of fasting on thyrotropin and thyroid status during Ramadan in 292 previously
well controlled hypothyroid patients. IFTAR study
authors:
- Tamer Mohamed Elsherbiny
journal: Endocrine
year: 2022
pmcid: PMC9988775
doi: 10.1007/s12020-022-03242-1
license: CC BY 4.0
---
# Impact of fasting on thyrotropin and thyroid status during Ramadan in 292 previously well controlled hypothyroid patients. IFTAR study
## Abstract
### Purpose
Fasting during Ramadan affects thyrotropin both in healthy subjects and hypothyroid patients on adequate levothyroxine replacement. Few studies have addressed this effect in hypothyroid patients with pre-Ramadan euthyroidism. This study aims to report the impact of fasting in a relatively large cohort.
### Methods
This was a prospective study including hypothyroid patients who fasted Ramadan during the years 2018, 2019, and 2020 in Alexandria, Egypt. All patients were euthyroid. Patients chosen one of three levothyroxine regimens during Ramadan, regimen 1: 60 min before Iftar, regimen 2: 3–4 h after Iftar, 60 min before Suhor, regimen 3: before the start of next fast, 3–4 h after an early Suhor. Thyroid status was assessed in pre-Ramadan visit and reassessed in post-Ramadan visit within 6 weeks from the end of Ramadan.
### Results
The study included 292 hypothyroid patients. Most patients were adherent, 249 patients ($85.3\%$), one sixth of patients were non-adherent, 43 patients ($14.7\%$). Post-Ramadan TSH was 2.13 ± 1.88 mIU/L versus 1.60 ± 0.96 mIU/L pre-Ramadan [$$P \leq 0.001$$]. Most patients were still euthyroid post-Ramadan, 233 patients ($79.8\%$), while 59 patients ($20.2\%$) were dysthyroid. Post-Ramadan TSH significantly correlated to pre-Ramadan TSH [$P \leq 0.001$]. Post-Ramadan TSH was significantly higher in non-adherent patients, 3.57 ± 3.11 mIU/L compared to adherent patients, 1.88 ± 1.44 mIU/L [$P \leq 0.001$].
### Conclusion
Fasting Ramadan in well controlled hypothyroid patients resulted in a significant increase in post-Ramadan TSH, yet $80\%$ the patients remain euthyroid after Ramadan. Post-Ramadan TSH and euthyroidism are related to adherence and pre-Ramadan TSH.
## Introduction
Levothyroxine (L-T4) - being the mainstay treatment for hypothyroidism - has been one of the top ten prescribed medications in the past decade [1]. However, $11\%$ to $42\%$ of hypothyroid patients on L-T4 are under-replaced [2]. Low or non-adherence was reportedly high among L-T4 users, with rates up to $52\%$ in USA and $55\%$ in Lebanon [3, 4]. This lack of adherence to L-T4 is partly related to the fact that it has to be taken daily on an empty stomach, one hour before food and beverages, and avoid interfering drugs for 4 or more hours [5].
Fasting Ramadan entails abstaining from food and drinks from dawn – more than an hour before sunrise – till sundown. In Egypt, years 2018 and 2019, duration of fasting ranged from 15 h:02 min to 15 h:29 min and 14 h:44 min to 15 h:23 min respectively, leaving only 9 h or less for 2 main meals and L-T4 intake for hypothyroid patients [6].
Fasting Ramadan in healthy subjects significantly and gradually increased thyrotropin (TSH) within normal limits – compared to pre-Ramadan values – reaching a maximum at the end of the month and returning back to normal around two months after Ramadan [7]. Fasting Ramadan also flattens TSH diurnal circadian rhythm with decreased midnight and increased afternoon values at the later one third of the month [8].
The aim of the present study is to report the impact of fasting Ramadan on TSH and thyroid status in previously well controlled hypothyroid patients in Alexandria, Egypt.
## Materials and methods
This was a prospective study including Muslim hypothyroid patients willing to fast Ramadan during the years 2018, 2019, and 2020 attending endocrinology outpatient clinic, Alexandria faculty of medicine, Alexandria university, Egypt. All included patients were euthyroid, and stable on the same L-T4 dose for at least 3 months before the start of Ramadan. Exclusion criteria were thyroid cancer patients requiring suppressive therapy, central hypothyroidism, pregnancy, chronic heart failure, liver cirrhosis, renal failure, acute medical or surgical illness at the time of evaluation, to avoid acute and chronic non thyroidal illness syndromes. All patients were explained the nature and aim of the study and signed an informed written consent. The protocol of the study was approved by the ethical committee of Alexandria faculty of medicine, [IRB number 12098].
Patients were free to follow one of three L-T4 regimens during Ramadan, explained in detail previously [6]. In short, regimen 1: to take L-T4 60 min before Iftar and beverages, regimen 2: to take L-T4 3–4 h after Iftar, 60 min before Suhor meal, regimen 3: to take L-T4 before the start of next fast 3–4 h after an early Suhor around midnight. If patients mixed between regimens 1 and 2, this was labeled regimen 4.
Adherence was assessed by interviewing participants during post-Ramadan visit. Non-adherence was defined as stopping food and beverages for less than 3 h before L-T4 tablet(s) or stopping food and beverages for less than 45 min after L-T4 tablet(s). Patients who skipped L-T4 treatment for 2 or more days without making up for their missed doses were excluded from the study.
Thyroid status was assessed for recruited patients in pre-Ramadan visit using TSH. The institution uses electrochemiluminescence immunoassay [ECLIA] on Cobas e 411 (Roche Diagnostics GmbH, Mannheim, Germany). Patients were considered euthyroid when TSH was 0.3–4 mIU/L for patients less than 70 years of age, and 1–5 mIU/L for patients more than 70 years of age according to European thyroid association recommendations [9].
Thyroid status was reassessed in post-Ramadan visit using TSH, provided that this visit comes within 6 weeks from the end of Ramadan. Patients were excluded from the study if post-Ramadan visit was delayed beyond 6 weeks after Ramadan or if they did not report TSH during post-Ramadan visit.
## Statistical methods
Data were fed to the computer and analyzed using IBM SPSS software package version 20.0. ( Armonk, NY: IBM Corp). Qualitative data were described using number and percent. Quantitative data were described using range mean, standard deviation, median (Range). Chi-square test for categorical variables, to compare between different groups. Monte Carlo correction for chi-square when more than $20\%$ of the cells have expected count less than 5. Student’s t test for normally distributed quantitative variables, to compare between two studied groups F-test (ANOVA) for normally distributed quantitative variables, to compare between more than two groups. Mann–Whitney test for abnormally distributed quantitative variables, to compare between two studied groups Kruskal–Wallis test for abnormally distributed quantitative variables, to compare between more than two studied groups Wilcoxon signed ranks test for abnormally distributed quantitative variables, to compare between two periods. Spearman coefficient to correlate between two distributed abnormally quantitative variables. Significance of the obtained results was judged at the $5\%$ level.
## Baseline characteristics
The study included 292 hypothyroid patients, the majority were females [280 patients ($95.9\%$)], and only 12 male patients were included in the study ($4.1\%$). Of the total patients, 68 patients were included in the year 2018 ($23.3\%$), 122 patients in the year 2019 ($41.7\%$), and 102 patients in the year 2020 ($35\%$) [Fig. 1]. Age ranged from 19 to 73 years; and the mean age was 43.5 years. Causes of hypothyroidism included: *Hashimoto thyroiditis* in 214 patients ($73.3\%$), thyroidectomy in 46 patients ($15.8\%$), unclassified in 21 patients ($7.2\%$), radioiodine ablation in 7 patients ($2.4\%$), and post-partum thyroiditis in 4 patients ($1.4\%$) [Table 1].Fig. 1Flow chart of patient’s exclusions in the studyTable 1Demographic, clinical data, and thyroid status of studied patients ($$n = 292$$)ParameterNo. (%) Sex Female280 ($95.9\%$) Male12 ($4.1\%$)Age (years) <203 ($1\%$) 20–<3039 ($13.4\%$) 30–<4090 ($30.8\%$) ≥40160 ($54.8\%$) Mean ± SD.43.50 ± 13.25 Median (Min.–Max.)41 (19–73)Diagnosis HT214 ($73.3\%$) Tx46 ($15.8\%$) RAI7 ($2.4\%$) PPT4 ($1.4\%$) Un Classified21 ($7.2\%$)Adherence Non-adherent43 ($14.7\%$) Adherent249 ($85.3\%$)L-T4 Regimen 1101 ($34.6\%$) 2127 ($43.5\%$) 314 ($4.8\%$) 450 ($17.1\%$)*Thyroid status* Pre Euthyroid292 ($100\%$) Under-replaced0 ($0\%$) Over-replaced0 ($0\%$) Post Euthyroid233 ($79.8\%$) Dysthyroid59 ($20.2\%$) Under-replaced33 ($11.3\%$) Over-replaced26 ($8.9\%$)HT Hashimoto thyroiditis, Tx thyroidectomy, RAI radioiodine ablation, PPT post-partum thyroiditis L-T4 regimen preferences were 101 patients ($34.6\%$), 127 patients ($43.5\%$), 14 patients ($4.8\%$), and 50 patients ($17.1\%$) for regimens 1, 2, 3, and 4 respectively. Most patients were adherent, 249 patients ($85.3\%$), and only one sixth of patients were non-adherent, 43 patients ($14.7\%$) [Table 1].
Fasting Ramadan significantly increased post-Ramadan TSH relative to pre-Ramadan TSH. Post-Ramadan TSH was 2.13 ± 1.88 mIU/L versus 1.60 ± 0.96 mIU/L pre-Ramadan [$$P \leq 0.001$$]. Post-Ramadan TSH was significantly higher in non-adherent patients, 3.57 ± 3.11 mIU/L compared to adherent patients, 1.88 ± 1.44 mIU/L [$P \leq 0.001$] [Fig. 2].Fig. 2Mean pre-Ramadan TSH, post-Ramadan TSH in the total patients, adherent, and non-adherent patients. *: Statistically significant at p ≤ 0.05, p value for comparing between pre- and post-Ramadan TSH using Wilcoxon signed ranks test. **: Statistically significant at p ≤ 0.05, p value for comparing between adherent and non-adherent patients using Mann–Whitney test Most patients were still euthyroid post-Ramadan, 233 patients ($79.8\%$), while 59 patients ($20.2\%$) were dysthyroid in the post Ramadan visit [Table 1]. Adherence rates were significantly higher in patients with post-Ramadan euthyroidism ($89.3\%$) versus only ($69.5\%$) in patients with post-Ramadan dysthyroidism [$P \leq 0.001$].
To show the effect of age on post-Ramadan TSH, patients were classified into 5 age groups as shown in Table 2. Post-Ramadan TSH increased significantly only in the 2 younger age groups, less than 30 years, and 30 to <40 years [Table 2]. Rates of adherence increased with increasing age, with the lowest adherence in patients < 30 years of age $\frac{33}{42}$ ($78.6\%$), intermediate adherence in patients from 30 to <60 years $\frac{175}{205}$ ($85.4\%$), and highest adherence in patients 60 years or older $\frac{41}{45}$ ($91.1\%$).Table 2Comparison between pre-Ramadan and post-Ramadan TSH according to age, etiology of hypothyroidism, and L-T4 regimen preferenceTSHNPre-Ramadan TSHPost-Ramadan TSHZpTotal Mean ± SD.2921.60 ± 0.962.13 ± 1.883.199*0.001* Median (Min.–Max.)1.4 (0.3–4.80)1.68 (0.01–15.40)Age (years)<30 Mean ± SD.421.55 ± 0.912.56 ± 2.142.313*0.021*30–<40 Mean ± SD.901.61 ± 0.992.41 ± 2.342.386*0.017*>40–<50 Mean ± SD.511.52 ± 0.882.03 ± 1.411.7110.08750–<60 Mean ± SD.641.46 ± 0.811.70 ± 1.360.8530.394≥60 Mean ± SD.451.93 ± 1.181.91 ± 1.540.5870.577DiagnosisHT Mean ± SD.2141.60 ± 0.952.10 ± 1.732.673*0.008*Tx Mean ± SD.461.48 ± 0.971.95 ± 2.540.4150.678RAI Mean ± SD.71.23 ± 1.272.32 ± 2.111.3520.176PPT Mean ± SD.41.64 ± 0.933.77 ± 1.631.4610.144Un Classified Mean ± SD.212.06 ± 0.882.46 ± 1.521.3210.187Regimen number1 Mean ± SD.1011.57 ± 0.952.08 ± 1.701.8340.0672 Mean ± SD.1271.57 ± 0.942.24 ± 2.142.563*0.010*3 Mean ± SD.142.16 ± 1.241.80 ± 1.340.5970.5514 Mean ± SD.501.60 ± 0.932.06 ± 1.631.3470.178HT Hashimoto thyroiditis, Tx thyroidectomy, RAI radioiodine ablation, PPT post-partum thyroiditis, Z Wilcoxon signed ranks test, p p value for comparing between pre and post*Statistically significant at p ≤ 0.05 Also, patients were classified according to the etiology of hypothyroidism, post-Ramadan TSH was significantly raised only in those who have Hashimoto thyroiditis. When patients were classified according to their preferred L-T4 regimen, post-Ramadan TSH was significantly raised only in patients who followed regimen 2. [ Table 2] Post-Ramadan TSH significantly correlated to pre-Ramadan TSH (Spearman coefficient 0.268 [$P \leq 0.001$]), and age (Spearman coefficient –0.117 [$$P \leq 0.046$$]).
## Discussion
The present study showed that in previously well controlled hypothyroid patients, fasting Ramadan was associated with a significant increase in TSH from 1.60 ± 0.96 mIU/L to 2.13 ± 1.88 mIU/L [$$P \leq 0.001$$]. Almost $80\%$ of patients were still euthyroid post-Ramadan. $85.3\%$ of the patients were adherent to L-T4 regimen of their choice. Non-adherence was associated with significantly higher TSH, whereas adherent patients constituted almost $90\%$ of patients who maintained euthyroidism in post-Ramadan. Pre-Ramadan TSH and age were significantly associated with post-Ramadan TSH.
Four published studies reported the impact of fasting Ramadan in previously well controlled hypothyroid patients with numbers of included patients ranging from 47 patients by Karoli et al. in 2013 to 112 patients by El-Kaissi et al. in 2020. The present study included 292 well controlled hypothyroid patients [10–13].
In healthy subjects fasting Ramadan, TSH increases significantly within normal reference range, and returns to pre-Ramadan values two months after Ramadan [7]. In adequately replaced hypothyroid patients, post-Ramadan TSH increased significantly compared to pre-Ramadan TSH in 2 studies by Dabbous et al. and El-Kaissi et al. whereas more than $60\%$ of patients showed an increase in TSH without statistical significance in the other 2 studies by Karoli et al. and Dellal et al. [ 10–13]. El-Kaissi et al. reported that $32\%$ of their patients were dysthyroid after Ramadan, whereas only $20.2\%$ of the patients in the present study were dysthyroid [13].
Post-Ramadan TSH was negatively correlated with age (Spearman coefficient –0.117 [$$P \leq 0.046$$]), meaning that older age was associated with lower post-Ramadan TSH. Also, post-Ramadan TSH increased significantly only in patients less than 40 years of age, whereas older patients did not show a significant increase. An opposite finding was reported by El-Kaissi et al. who found that older patients had a higher post-Ramadan TSH [13]. Current findings may be explained by a better adherence to L-T4 with increasing age, recently reported in a middle eastern country [14]. When patients were classified by age, the highest adherence rate was found in patients 60 years or older ($91.1\%$), whereas the lowest adherence rate was found in patients younger than 30 years ($78.6\%$).
Only hypothyroid patients due to *Hashimoto thyroiditis* showed a significant increase in post-Ramadan TSH. Hashimoto thyroiditis accounted for $73.3\%$ of the recruited patients. Number of patients attributed to other causes may have been too small to show statistical significance. Also, only patients following L-T4 regimen 2 showed a significant increase in post-Ramadan TSH.
Levothyroxine intake regimens offered to our patients are based on American thyroid association recommendations as explained before [15]. L-T4 regimens offered to patients in previous studies included: with iftar, 30 min before iftar or 30 min before suhor, or 2 h after iftar or food [10–13]. These regimens are not consistent with ATA recommendations to optimize absorption; although, it was considered convenient by most patients in one study by Dabbous et al. however, the authors of this study considered these time intervals insufficient for optimal absorption due to the nature of foods commonly consumed during Ramadan [11].
$85\%$ of the patients in the present study were adherent to L-T4 regimen of their choice. Reported rates of adherence range from $25\%$ reported by Karoli et al., 35–$42\%$ by Dabbous et al. and up to $75\%$ reported by Sheikh et al. [ 10, 11, 16]. $82\%$ adherence was previously reported by our institution following the same L-T4 regimens used in the present study [6]. Having multiple L-T4 regimen and freedom of choice of the regimen that suits every patient’s priority may have contributed to this high rate of adherence.
Non-adherence resulted in significantly higher TSH compared to adherence whereas $89\%$ of patients who maintained euthyroidism post-Ramadan were adherent. Karoli et al. found that meal-levothyroxine interval correlated negatively to post-Ramadan TSH [10].
With an estimated number of about 48 million Muslims with hypothyroidism requiring L-T4 replacement and practicing Ramadan fasting yearly, the currently available literature addressing the impact of fasting Ramadan on thyroid status is quite limited and unsatisfactory compared to the literature for example covering the topic of Ramadan fasting and diabetes [6, 17]. Future research is needed to report experience with fasting Ramadan in different locations and different times of the year, and to answer research questions like the need to raise L-T4 dose temporarily during Ramadan and the utility of weekly or biweekly L-T4 during Ramadan fasting.
The present study has several strengths. The only study to be conducted in 3 consecutive Ramadan months, in 3 consecutive years allowing the inclusion of a large number of patients. The largest number of recruited well controlled hypothyroid patients fasting during Ramadan in a single study. In fact, the number of included patients – 292 patients – is almost equivalent to the numbers of included patients in the 4 similarly designed studies combined, 317 patients [10–13]. The only study to offer L-T4 regimens in accordance with the latest American thyroid association recommendations for optimal L-T4 absorption, as opposed to regimens in previous studies with time intervals that may cause a suboptimal absorption, which may explain the highest rate of post-Ramadan euthyroidism ($79.8\%$) versus previously reported ($68\%$) [13]. The only study to be conducted in a north African country – Egypt – as opposed to other studies conducted in countries from west Asia in Turkey to south Asia in India, thus, our patients were of different ethnicities, social and dietary habits.
A major limitation for this study and others dealing with fasting *Ramadan is* the fact that duration of fasting and atmospheric temperatures differ from one country to another according to latitude and for the same country from year to year as Ramadan moves from one season to another every 8 years, meaning that results can’t be generalized for all countries or even for the same country in different seasons of the year.
In conclusion, – based on the present data from 292 hypothyroid patients – fasting Ramadan in well controlled hypothyroid patients resulted in a significant increase in post-Ramadan TSH, yet $80\%$ the patients remain euthyroid after Ramadan. Post-Ramadan TSH and euthyroidism are related to adherence, L-T4 regimen preference, age, and pre-Ramadan TSH.
## References
1. Jonklaas J, DeSale S. **Levothyroxine prescriptions trends may indicate a downtrend in prescribing**. *Ther. Adv. Endocrinol. Metab.* (2020) **11** 2042018820920551. DOI: 10.1177/2042018820920551
2. Eligar V, Taylor PN, Okosieme OE, Leese GP, Dayan CM. **Thyroxine replacement: a clinical endocrinologist’s viewpoint**. *Ann. Clin. Biochem* (2016) **53** 421-433. DOI: 10.1177/0004563216642255
3. Hepp Z, Wyne K, Manthena SR, Wang S, Gossain V. **Adherence to thyroid hormone replacement therapy: a retrospective, claims database analysis**. *Curr. Med Res Opin.* (2018) **34** 1673-1678. DOI: 10.1080/03007995.2018.1486293
4. El Helou S, Hallit S, Awada S, Al-Hajje A, Rachidi S, Bawab W, Salameh P, Zein S. **Adherence to levothyroxine among patients with hypothyroidism in Lebanon**. *East Mediterr. Health J.* (2019) **25** 149-159. DOI: 10.26719/emhj.18.022
5. Rajput R, Pathak V. **The Effect of Daily versus Weekly Levothyroxine Replacement on Thyroid Function Test in Hypothyroid Patients at a Tertiary Care Centre in Haryana**. *Eur. Thyroid J.* (2017) **6** 250-254. DOI: 10.1159/000477348
6. Elsherbiny TM. **Preference, Adherence, and Maintenance of Euthyroidism Using 3 Different Regimens of Levothyroxine Intake during the Fasting Month of Ramadan**. *Dubai Diabetes Endocrinol. J.* (2021) **27** 6-13. DOI: 10.1159/000513927
7. Sajid KM, Akhtar M, Malik GQ. **Ramadan fasting and thyroid hormone profile**. *J. Pak. Med Assoc.* (1991) **41** 213-216. PMID: 1744968
8. Bogdan A, Bouchareb B, Touitou Y. **Ramadan fasting alters endocrine and neuroendocrine circadian patterns. Meal-time as a synchronizer in humans?**. *Life Sci.* (2001) **68** 1607-1615. DOI: 10.1016/S0024-3205(01)00966-3
9. Pearce SH, Brabant G, Duntas LH. **2013 ETA Guideline: Management of Subclinical Hypothyroidism**. *Eur. Thyroid J.* (2013) **2** 215-228. DOI: 10.1159/000356507
10. Karoli R, Fatima J, Chandra A, Mishra PP. **Levothyroxine replacement and Ramadan fasting**. *Indian J. Endocr. Metab.* (2013) **17** 318-319. DOI: 10.4103/2230-8210.109700
11. Dabbous Z, Alowainati B, Darwish S, Ali H, Farook S, Al Malaheem M, Abdalrubb A, Gul W, Haliqa WA. **A Prospective Study Comparing Two-Time Points of Thyroid Hormone Replacement during the Holy Month of Ramadan**. *Int J. Endocrinol.* (2019) **22** 9843961
12. Dellal FD, Ogmen B, Ozdemir D, Alkan A, Cuhaci Seyrek FN, Polat SB, Ersoy R, Cakir B. **Effect of Ramadan Fasting on Thyroid Hormone Levels in Patients on Levothyroxine Treatment**. *J. Coll. Physicians Surg. Pak.* (2020) **30** 1009-1014. DOI: 10.29271/jcpsp.2020.10.1009
13. El-Kaissi S, Dajani R, Lee-St John TJ, Ann Santarina S, Makia F, AlTakruri M, Ahmed Y. **Impact of Lifestyle Changes During Ramadan on Thyroid Function Tests in Hypothyroid Patients Taking Levothyroxine**. *Endocr. Pr.* (2020) **26** 748-753. DOI: 10.4158/EP-2019-0505
14. El Helou S, Hallit S, Awada S, Al-Hajje A, Rachidi S, Bawab W, Salameh P, Zein S. **Adherence to levothyroxine among patients with hypothyroidism in Lebanon**. *East Mediterr. Health J.* (2019) **25** 149-159. DOI: 10.26719/emhj.18.022
15. Jonklaas J, Bianco AC, Bauer AJ. **Guidelines for the treatment of hypothyroidism: prepared by the american thyroid association task force on thyroid hormone replacement**. *Thyroid* (2014) **24** 1670-1751. DOI: 10.1089/thy.2014.0028
16. Sheikh A, Mawani M, Mahar SA. **Impact of Ramadan Fasting on Thyroid Status and Quality of Life in Patients with Primary Hypothyroidism: a Prospective Cohort Study from karachi, pakistan**. *Endocr. Pr.* (2018) **24** 882-888. DOI: 10.4158/EP-2018-0038
17. Hassanein M, Afandi B, Yakoob Ahmedani M. **Diabetes and Ramadan: Practical guidelines 2021**. *Diabetes Res Clin. Pract.* (2022) **185** 109185. DOI: 10.1016/j.diabres.2021.109185
|
---
title: Preoperative Serum Cortisol Level Is Predictive of Weight Loss After Laparoscopic
Sleeve Gastrectomy in Men with Severe Obesity but Not Women
authors:
- Hironori Bando
- Hiroshi Miura
- Seiichi Kitahama
- Shinsuke Nakajima
- Tetsuya Takahashi
- Toshihiko Mihara
- Teppei Momono
- Maki Kimura-Koyanagi
- Kazuhiko Sakaguchi
- Tomoichiro Mukai
- Wataru Ogawa
- Yoshikazu Tamori
journal: Obesity Surgery
year: 2023
pmcid: PMC9988780
doi: 10.1007/s11695-022-06415-z
license: CC BY 4.0
---
# Preoperative Serum Cortisol Level Is Predictive of Weight Loss After Laparoscopic Sleeve Gastrectomy in Men with Severe Obesity but Not Women
## Abstract
### Background
Bariatric surgery is an effective treatment for severe obesity and its associated medical problems. Preoperative factors that predict postoperative weight loss remain to be fully characterized, however.
### Methods
Anthropometric and laboratory data were collected retrospectively for severely obese patients who underwent laparoscopic sleeve gastrectomy (LSG) between April 2016 and July 2019 at our hospital. Preoperative factors that predicted weight loss at 1 year after LSG were investigated.
### Results
A total of 122 subjects (45 men and 77 women) underwent LSG. The mean ± SD age and body mass index at surgery were 44.4 ± 10.4 years and 40.7 ± 6.7 kg/m2. The percent total weight loss (%TWL) was 27.0 ± 8.6 among all subjects, 26.4 ± 8.0 among men, and 27.4 ± 8.9 among women, with no significant difference between the sexes. The %TWL showed a significant inverse correlation with serum cortisol level in men and with age and the visceral/subcutaneous fat area ratio in women. Multivariable regression analysis revealed the presence of type 2 diabetes and the serum cortisol concentration to be negatively associated with %TWL among all subjects and men, respectively. Receiver operating characteristic curve analysis identified an optimal cutoff of 10 µg/dL for prediction of a %TWL of ≥ 25 in men by serum cortisol level.
### Conclusions
Serum cortisol concentration was identified as a predictor for postoperative weight loss in men. Our results may thus help inform the decision to perform LSG or more effective surgical procedures in men with severe obesity.
## Introduction
Obesity triggers a wide variety of health problems and thus leads to a decline in the quality and expectancy of life [1]. Whereas treatment of obesity remains a clinical challenge, the mortality of obese individuals decreases with weight loss regardless of the treatment approach, either surgical or nonsurgical [2]. In the case of severe obesity, for which the effect of antiobesity medications is limited, surgery is an effective and reliable treatment option [3].
Bariatric surgery is effective not only for weight loss but also for associated medical problems of obesity such as type 2 diabetes mellitus (T2DM) [4], and it has been rapidly and widely adopted as a treatment for severe obesity—in particular, since the introduction of lower-risk laparoscopic procedures [5]. Approximately 700,000 bariatric surgeries are now performed annually worldwide [6]. Laparoscopic sleeve gastrectomy (LSG) is one of the most commonly performed types of bariatric surgery, given its simplicity of procedure and good outcomes [7].
Several factors have been found to be associated with weight loss after LSG, including body mass index (BMI), age, sex, and comorbid T2DM [8–14]. The pattern of fat accumulation differs between the sexes [15], however, suggesting that the pathophysiology as well as the efficacy of treatment for obesity might also differ. As far as we are aware, no previous study has analyzed sex-specific predictive factors for postoperative weight loss in individuals who undergo bariatric surgery.
The number of bariatric surgeries performed in Japan has been increasing rapidly but is still small compared with that in Western countries [16]. Vulnerability to obesity and its associated medical problems appears to differ among ethnicities [17], and information relating to factors that influence the outcome of bariatric surgery in Japanese individuals is limited. Our aim is therefore identifying sex-specific predictive factors to develop individualized treatment strategies for efficient weight loss for severe obesity in Japanese. We have now performed a retrospective study to examine preoperative anthropometric and laboratory data for 122 Japanese individuals with severe obesity who underwent LSG in an attempt to identify sex-specific predictors of postoperative weight loss.
## Patients
This retrospective observational study was approved by the institutional review board of our hospital (approval no. 20211111A) and conformed to the provisions of the Declaration of Helsinki (as revised in 2013). The study subjects were all patients who underwent LSG at our hospital by a bariatric fellowship-trained surgeon in the setting of a comprehensive multidisciplinary program between April 2016 and July 2019. All participants had a BMI of ≥ 35 kg/m2 with obesity-related medical problems and failed to achieve a substantial weight reduction despite more than 6 months of medical therapy. They provided written informed consent for the collection and use of their data for research purposes only. Patients were excluded if they had Cushing’s syndrome or other hormonal abnormalities, or if they had received steroid treatment by multiple endocrinologists. We defined the presence of diabetes mellitus as the taking of hypoglycemic agents or as a fasting plasma glucose concentration of ≥ 126 mg/dL and a hemoglobin A1c (HbA1c) level of ≥ $6.5\%$, according to the diagnostic criteria of the Japan Diabetes Society [18]. T2DM was diagnosed by confirming the absence of antibodies to glutamic acid decarboxylase as well as by excluding diabetes due to other specific mechanisms or diseases in subjects with diabetes mellitus.
## Study Design and Measurements
Anthropometric and laboratory data were obtained immediately before and 1 year after the surgery. The serum cortisol concentration was measured in the morning before breakfast and at rest during hospitalization for surgery. The amount of skeletal muscle and fat mass were measured by the bioelectrical impedance method with an In-Body S20 body composition analyzer (Biospace, Seoul, Korea). The subcutaneous fat area (SFA) and visceral fat area (VFA) were calculated from abdominal computed tomography images obtained at the navel level. The visceral/subcutaneous fat area ratio (VSR) was obtained as 100 × VFA/SFA. The percent total weight loss (%TWL) was calculated as 100 × (operative weight − follow-up weight) / operative weight. The percent excess weight loss (%EWL) was calculated as 100 × (operative weight − follow-up weight) / (operative weight − ideal body weight). The skeletal muscle index (SMI) was calculated by dividing skeletal muscle weight (kg) by body weight (kg). The primary end point of the study was the identification of significant preoperative predictors of %TWL.
## Surgical Procedures
The abdomen was entered under direct vision with a forward-viewing trocar. A 5-mm, three 12-mm, and one 15-mm ports were placed. A Nathanson liver retractor was used to elevate the liver. We mobilized the periesophageal fat pad to visually position the stapler to leave approximately 1 cm of gastric tissue lateral to the angle of His. The pylorus was identified, and an area approximately 4 cm from the pylorus was chosen to begin ligating and transecting the greater curvature vessels with a vessel-sealer device. The greater curvature of the stomach was mobilized to the angle of His, with particular attention paid to mobilizing the entire fundus to the mid-portion of the left crura of the diaphragm. A 34-French bougie was passed by an anesthesiologist and positioned in the distal antrum. Resection of the antrum was started tangentially from the right lateral port using a linear stapler, positioning the tip of the stapler to give a distance of 1 cm from the bougie at the area of the incisura angularis. Resection of the body and fundus of the stomach was achieved via the 12-mm left mid-clavicular port site to the angle of His. It was our practice to oversew the staple line with 2–0 nonabsorbable suture. The 12- and 15-mm port sites were closed with absorbable sutures.
## Statistical Analysis
All statistical analysis was performed with the use of JMP Statistical Database Software version 12.2.0 (SAS Institute, Cary, NC, USA). Analysis of variance (ANOVA), correlation analysis, multivariable regression analysis, and receiver operating characteristic (ROC) curve analysis were performed as appropriate. Multivariable regression analysis was performed to identify potential independent predictors of postsurgery weight loss. Age, BMI, SMI, VSR, T2DM, and serum cortisol level were included as covariates. These factors were selected as explanatory covariates because aging is associated with changes in body composition and physical function [19], BMI is associated with obesity-related outcomes [20], the amount of skeletal muscle is associated with energy expenditure [21], T2DM is associated with difficulty in losing body weight [22], insufficient weight loss in bariatric surgery was associated with a history of hypertension [23], and cortisol promotes body weight gain and obesity [24]. VSR is known to be different between the sexes [15] and has a strong association with cardiometabolic risks [25], but its effect on weight loss is unclear. All reported P values are two-tailed, and those of < 0.05 were considered statistically significant.
## Patient Characteristics and Effectiveness of LSG
A total of 122 severely obese individuals (45 men and 77 women) underwent LSG. The characteristics of these study patients are shown in Table 1. The mean ± SD age and BMI at the time of surgery were 44.4 ± 10.4 years and 40.7 ± 6.7 kg/m2, and the prevalence of T2DM was $47.5\%$. BMI had decreased from 40.7 to 28.0 kg/m2, and the effectiveness of surgery as evaluated by %TWL was 27.0 ± 8.6 at 1 year after surgery. Whereas skeletal muscle weight decreased after LSG, SMI increased, suggesting that surgery reduced fat composition predominantly. Blood pressure was decreased, and parameters of glucose metabolism were improved after surgery. With regard to lipid metabolism, triglyceride and high-density lipoprotein cholesterol levels were decreased and increased after LSG, respectively. Serum adiponectin and leptin concentrations were increased and decreased, respectively. Analysis according to sex revealed that BMI was decreased in both men and women, and %TWL did not differ between the sexes (26.4 ± 8.0 in men and 27.4 ± 8.9 in women; $$p \leq 0.5609$$, Student’s t-test). Low-density lipoprotein cholesterol was decreased after surgery only in men. Preoperative serum cortisol concentrations were < 16 μg/dL in all subjects and did not differ between the sexes ($$p \leq 0.3690$$, Student’s t-test).Table 1Clinical characteristics of the patients before and after laparoscopic sleeve gastrectomyTotal ($$n = 122$$)Male ($$n = 45$$)Female ($$n = 77$$)PreoperativePostoperativep-value (pre vs post)PreoperativePostoperativep-value (pre vs post)PreoperativePostoperativep-value (pre vs post)Age (years)44.4 ± 10.443.6 ± 9.944.7 ± 10.7BMI (kg/m2)40.7 ± 6.728.0 ± 5.4 < 0.0001*42.3 ± 8.1†29.7 ± 6.0§ < 0.0001*39.7 ± 5.6†27.0 ± 4.7§ < 0.0001*Body weight (kg)104 ± 20.275.7 ± 17.4 < 0.0001*121 ± 19.6†89.5 ± 17.2§ < 0.0001*93.4 ± 12.4†68.1 ± 12.2§ < 0.0001*Skeletal muscle weight (kg)31.5 ± 728.0 ± 6.3 < 0.0001*39.3 ± 4.5†34.9 ± 4.7§ < 0.0001*27.2 ± 3.3†24.3 ± 3.1§ < 0.0001*SMI (%)30.3 ± 3.537.5 ± 5.3 < 0.0001*32.6 ± 3.8†39.7 ± 6.0§ < 0.0001*29.0 ± 2.6†36.3 ± 4.5§ < 0.0001*SBP (mmHg)134 ± 17.2121 ± 16.6 < 0.0001*137 ± 18.6124 ± 18.50.0122*131.5 ± 16119 ± 15.10.0003*DBP (mmHg)83.1 ± 12.675.4 ± 11.20.0004*83.2 ± 14.276.7 ± 12.00.0080*83.1 ± 11.774.6 ± 10.70.0239*Prevalence of hypertension (%)72.479.168.5Total fat area (cm2)688 ± 185341 ± 161 < 0.0001*752 ± 196.6†369 ± 201 < 0.0001*651.2 ± 168.0†325 ± 132 < 0.0001*VSR (%)50.1 ± 25.930.6 ± 20.7 < 0.0001*59.6 ± 30.0†34.0 ± 27.2 < 0.0001*44.6 ± 21.6†28.7 ± 15.7 < 0.0001*Serum total protein (g/dL)6.96 ± 0.46.96 ± 0.520.06926.96 ± 0.417.02 ± 0.540.86597.0 ± 0.406.92 ± 0.500.0298*Prevalence of T2DM (%)47.557.841.6HbA1c (%)6.13 ± 0.875.59 ± 0.55 < 0.0001*6.22 ± 1.055.48 ± 0.59 < 0.0001*6.1 ± 0.745.65 ± 0.52 < 0.0001*FPG (mg/dL)106 ± 18.795.5 ± 16.60.0002*110 ± 22.3100 ± 14.2§0.0639103.5 ± 16.093.0 ± 17.4§0.0012*Serum TG (mg/dL)143 ± 72.990.1 ± 46.2 < 0.0001*157 ± 72.298.5 ± 50.6 < 0.0001*138.3 ± 72.985.4 ± 43.2 < 0.0001*Serum CPR (ng/mL)3.09 ± 1.471.84 ± 0.73 < 0.0001*3.48 ± 1.80†2.15 ± 0.86§ < 0.0001*2.9 ± 1.19†1.68 ± 0.59§ < 0.0001*Serum HDL-Chol (mg/dL)47.0 ± 10.472.1 ± 19.3 < 0.0001*44.0 ± 10.5†65.7 ± 21.7§ < 0.0001*48.8 ± 10.0†75.7 ± 17.1§ < 0.0001*Serum LDL-Chol (mg/dL)117 ± 43114 ± 33.80.3007112 ± 32.399.3 ± 26.5§0.0173*120.1 ± 34.9123 ± 34.7§0.7478LDL/HDL-Chol ratio2.63 ± 1.021.72 ± 0.70 < 0.0001*2.74 ± 1.121.73 ± 0.8 < 0.0001*2.6 ± 0.961.71 ± 0.65 < 0.0001*Serum adiponectin concentration (µg/dL)6.74 ± 3.715.2 ± 10.4 < 0.0001*5.95 ± 3.2113.3 ± 8.3 < 0.0001*7.2 ± 3.916.2 ± 11.2 < 0.0001*Serum leptin concentration (ng/dL)51.0 ± 32.317.2 ± 12.3 < 0.0001*42.8 ± 26.8†12.3 ± 9.9§ < 0.0001*55.7 ± 34.4†19.8 ± 12.7§ < 0.0001*Serum cortisol concentration (μg/dL)7.9 ± 3.18.2 ± 3.37.7 ± $3.0\%$TWL27.0 ± 8.626.4 ± 8.027.4 ± 8.9Data are means ± SD. BMI, body mass index; SMI, skeletal muscle index; SBP, systolic blood pressure; DBP, diastolic blood pressure; VSR, visceral/subcutaneous fat area ratio; T2DM, type 2 diabetes mellitus; HbA1c, glycosylated hemoglobin; FPG, fasting plasma glucose; TG, triacylglycerol; CPR, C-peptide reactivity; HDL-chol, high-density lipoprotein cholesterol; LDL-chol, low-density lipoprotein cholesterol; %TWL, percent total weight loss. * $p \leq 0.05$ (preoperative vs postoperative). †$p \leq 0.05$ (preoperative male vs preoperative female). § $p \leq 0.05$ (postoperative male vs postoperative female)
## Factors that Correlate with %TWL
The linear correlation between preoperative anthropometric or laboratory parameters and %TWL at 1 year after surgery is shown in Table 2. Among all patients, age (r = − 0.2438, $$p \leq 0.0122$$) and VSR (r = − 0.1938, $$p \leq 0.0487$$) showed a significant negative correlation with %TWL. Serum cortisol level was significantly and negatively correlated with %TWL in men (r = − 0.4629, $$p \leq 0.0067$$), whereas age (r = − 0.3035, $$p \leq 0.0119$$) and VSR (r = − 0.2814, $$p \leq 0.0201$$) were weakly (but significantly) and negatively correlated with %TWL in women. Table 2Correlation between preoperative clinical parameters and %TWL at 1 year after laparoscopic sleeve gastrectomyTotalMaleFemalerp-valuerp-valuerp-valueAge − 0.24380.0122* − 0.11040.5154 − 0.30350.0119*BMI − 0.02420.8067 − 0.00010.9536 − 0.01500.9035SMI − 0.05450.58670.10120.5628 − 0.16040.1947SBP − 0.07230.4772-0.11950.5010 − 0.03360.7904DBP0.06580.5190 − 0.04200.81370.12200.3329VSR − 0.19380.0487* − 0.03310.8480 − 0.28140.0201*FPG − 0.07230.4638 − 0.01670.9207 − 0.10100.4123HbA1c − 0.13910.1589 − 0.14230.4008 − 0.13580.2374Adiponectin0.18390.06700.09870.57260.20370.1036Leptin − 0.02110.8356 − 0.15220.39030.00130.9917CPR0.00360.9708 − 0.04810.77730.06860.5784TG − 0.13690.1648 − 0.15420.3623 − 0.11640.3445HDL-chol0.12150.21700.28360.08900.02620.8318LDL-chol − 0.03020.75970.07150.6741 − 0.08310.5003Cortisol − 0.19370.0706 − 0.46290.0067* − 0.04910.7217BMI, body mass index; SMI, skeletal muscle index; SBP, systolic blood pressure; DBP, diastolic blood pressure; VSR, visceral/subcutaneous fat area ratio; FPG, fasting plasma glucose; HbA1c, glycosylated hemoglobin; CPR, C-peptide reactivity; TG, triacylglycerol; HDL-chol, high-density lipoprotein cholesterol; LDL-chol, low-density lipoprotein cholesterol; %TWL, percent total weight loss. * p-value < 0.05
## Preoperative Factors Predicting Efficient Weight Loss at 1 Year After LSG
To identify independent predictors of weight loss outcome, we performed multivariable regression analysis. We selected seven preoperative factors—age, BMI, SMI, VSR, T2DM, hypertension, and serum cortisol level—as explanatory variables (Table 3). This analysis revealed that T2DM was negatively associated with %TWL among all patients ($$p \leq 0.0267$$), whereas the serum cortisol level was negatively associated with %TWL in men ($$p \leq 0.0215$$), and there were no independent predictors of %TWL for women. Table 3Multivariable regression analysis of %TWL and various preoperative clinical parametersFactorsTotalMaleFemalet-valuep-valueVIFt-valuep-valueVIFt-valuep-valueVIFAge − 1.430.15601.3045 − 0.290.77681.4800 − 1.180.24291.3589BMI − 0.460.64381.3620 − 0.250.80153.4824 − 0.880.38491.5464SMI − 0.440.66331.55680.120.90273.5720 − 0.940.35371.8590VSR − 0.400.69071.4078 − 0.630.53821.5316 − 0.610.54211.3589T2DM − 2.260.0267*1.3621 − 1.920.28531.2148 − 0.980.33441.8160Hypertension − 0.070.94581.0993 − 0.380.71141.26410.020.98771.2662Cortisol − 1.000.31871.1969 − 2.490.0215*1.3802 − $0.150.88401.3589\%$TWL, percent total weight loss; BMI, body mass index; SMI, skeletal muscle index; VSR, visceral/subcutaneous fat area ratio; T2DM, type 2 diabetes mellitus; and VIF, variance inflation factor. * p-value < 0.05 To confirm the potential of serum cortisol level as a predictor of postoperative body weight in men, we also performed multivariable regression analysis using another weight loss marker, %EWL as well. We found that only serum cortisol level was negatively associated with %EWL in men ($$p \leq 0.0308$$), whereas BMI was negatively associated with %EWL in total subjects (p = < 0.0001) and in women ($$p \leq 0.0009$$).
Recent studies have suggested that a good response to bariatric surgery should be defined as a %TWL of ≥ 25 [26, 27]. We, therefore, divided the study patients into two groups with a %TWL of ≥ 25 or < 25. Male patients with a %TWL of ≥ 25 had a lower serum cortisol concentration before surgery than did those with a %TWL of < 25, whereas there was no such difference between the two groups among all patients or women (Fig. 1A). We performed ROC curve analysis to determine the cutoff value of the serum cortisol level for prediction of a good efficacy for LSG (%TWL of ≥ 25) in men. We found that the optimal cutoff was 10.0 µg/dL, which provided a sensitivity of $94.2\%$ and a specificity of $50.0\%$ (Fig. 1B).Fig. 1Serum cortisol level before surgery predicts the efficiency of laparoscopic sleeve gastrectomy (LSG) at 1 year after surgery in men. A Serum cortisol concentration according to a percent total weight loss (%TWL) cutoff of 25 in all study subjects as well as in male and female subjects separately. The box-and-whisker plots represent the minimum and maximum values (whiskers), the second and third quartiles (box), and the median (midline). * $P \leq 0.05$ (Student’s t-test). B Receiver operating characteristic (ROC) curve analysis of serum cortisol level for prediction of a good response (%TWL of ≥ 25) to LSG in men. The area under the curve (AUC) was 0.717, and the optimal cutoff of 10 µg/dL is indicated
## Relation Between Preoperative Serum Cortisol Level and the Metabolic Improvement
Preoperative serum cortisol level was negatively correlated with the decrease of HbA1c (r = − 0.36, $$p \leq 0.037$$) and LDL-C (r = − 0.50, $$p \leq 0.0036$$) as well as the increase of HDL-C ($r = 0.48$, $$p \leq 0.0055$$) at 1 year after surgery only in men. In addition, men with a serum cortisol concentration ≥ 10.0 µg/dL showed smaller improvement in HbA1c level (− $0.17\%$) after LSG compared to those with cortisol levels < 10.0 µg/dL (− $0.88\%$) ($$p \leq 0.038$$). These data indicated that preoperative cortisol levels were also related to weight loss–induced metabolic improvement after LSG in men.
## Discussion
We have here identified the serum cortisol concentration as a predictive factor for efficient weight loss in Japanese men who underwent LSG. As far as we are aware, our study is the first to analyze sex-specific predictive factors for the outcome of bariatric surgery and to identify serum cortisol level as such a factor.
Cortisol is an obesogenic hormone that stimulates food intake [24]. The reason why the serum level of cortisol was correlated with weight loss only in men is unclear. Estrogen promotes the production of cortisol-binding globulin [28] and thereby alters the ratio of free to total cortisol levels. Given that estrogen levels decline with age after menopause, the total concentration of cortisol measured in the present study might not well reflect the free cortisol level in postmenopausal female subjects. About $35\%$ of the female study subjects were > 50 years of age. A previous study of a Japanese obese cohort consisting of 37 men and 46 women found that the concentration of cortisol in saliva, which reflects well the free cortisol level in serum, was negatively correlated with the extent of weight reduction after nonsurgical treatment [29].
Cortisol concentrations can be measured from hair, saliva, serum, and urine. A recent meta-analysis showed that hair cortisol concentration was associated with adiposity-related outcomes [30]. Measuring hair cortisol is a noninvasive method and stably reflects the long-term effect of cortisol compared with serum cortisol levels which show circadian variations. Hair cortisol concentration may therefore become a more potent predictor for weight loss in the future.
The circulating level of testosterone has been found to be high in men with a low BMI [31], and a reduced level of this hormone is associated with the development and progression of obesity in men [32, 33]. Glucocorticoids inhibit testosterone production by directly influencing Leydig cell function [34, 35]. The serum cortisol concentration may thus affect the outcome of bariatric surgery in men through its effect on testosterone production, given that reduced testosterone enhanced the activity of lipoprotein lipase, resulting in a rise in triacylglycerol uptake to adipose tissues and subsequent obesity [36]. It will thus be of interest to determine whether the circulating testosterone level serves as a predictive factor for the outcome of bariatric surgery in men.
Consistent with previous studies [8, 9, 13, 14], we found that T2DM negatively influenced the extent of weight loss after surgery in the present study. Although the underlying mechanism of this association remains unclear, insufficient adherence to diet therapy might be responsible, as suggested by a previous study [22]. Moreover, certain antidiabetes medications, such as insulin, have been shown to increase weight gain in patients whose diabetes did not improve even after LSG [37]. Whereas previous studies have found that preoperative BMI and age were correlated with weight loss after bariatric surgery [8–13], we did not observe such relations in our study. This apparent discrepancy might be attributable to differences in the number, ethnicity, or male/female ratio of study subjects, in surgical procedure (sleeve gastrectomy, gastric bypass, or gastric banding), in postoperative medical support, or in explanatory variables selected for multivariable regression analysis.
These are several limitations of our study. The study was retrospective in nature, was restricted to a single specialized center, and had a relatively small sample size. In addition, we analyzed only the total cortisol level, not the free cortisol concentration, the latter of which directly reflects the action of the hormone. Furthermore, we did not have data concerning the dexamethasone suppression test, thus could not thoroughly exclude the possibility of autonomous cortisol secretion. We did not evaluate certain clinical parameters that might affect postoperative weight loss, such as mental state, alcohol consumption, diet adherence, and physical activity. In addition, the AUC of the ROC curve analysis was not very robust (0.72). Longer follow-up might increase the robustness of this study. Finally, we cannot exclude possible effects of concurrent medical treatments that reduce body weight, such as diabetes therapy with glucagon-like peptide-1 (GLP-1) receptor agonists and sodium-glucose cotransporter 2 (SGLT2) inhibitors.
## Conclusion
Our data show that the serum cortisol level is an independent predictor of weight loss after LSG in severely obese Japanese men. Such a predictor may be helpful for the choice of surgical procedure, such as LSG or other surgical methods which can be expected to achieve greater weight loss than LSG [38, 39]. Further studies are warranted to clarify the mechanism by which serum cortisol limits the effectiveness of LSG as well as validate the usefulness of cortisol level as a marker for effective weight loss in clinical settings in male patients with severe obesity.
## References
1. Haslam DW, James WP. **Obesity**. *Lancet* (2005) **366** 1197-1209. DOI: 10.1016/S0140-6736(05)67483-1
2. Gloy VL, Briel M, Bhatt DL, Kashyap SR, Schauer PR, Mingrone G. **Bariatric surgery versus non-surgical treatment for obesity: a systematic review and meta-analysis of randomised controlled trials**. *BMJ* (2013) **347** f5934. DOI: 10.1136/bmj.f5934
3. Bray GA, Frühbeck G, Ryan DH, Wilding JP. **Management of obesity**. *Lancet* (2016) **387** 1947-1956. DOI: 10.1016/S0140-6736(16)00271-3
4. Schauer PR, Bhatt DL, Kirwan JP, Wolski K, Aminian A, Brethauer SA. **Bariatric surgery versus intensive medical therapy for diabetes-5-year outcomes**. *N Engl J Med* (2017) **376** 641-651. DOI: 10.1056/NEJMoa1600869
5. Angrisani L, Santonicola A, Iovino P, Formisano G, Buchwald H, Scopinaro N. **Bariatric surgery worldwide 2013**. *Obes Surg* (2015) **25** 1822-1832. DOI: 10.1007/s11695-015-1657-z
6. Angrisani L, Santonicola A, Iovino P, Ramos A, Shikora S, Kow L. **Bariatric surgery survey 2018: similarities and disparities among the 5 IFSO chapters**. *Obes Surg* (2021) **31** 1937-1948. DOI: 10.1007/s11695-020-05207-7
7. Rosenthal RJ, Diaz AA, Arvidsson D, Baker RS, Basso N, Bellanger D. **International sleeve gastrectomy expert panel consensus statement: best practice guidelines based on experience of >12,000 cases**. *Surg Obes Relat Dis* (2012) **8** 8-19. DOI: 10.1016/j.soard.2011.10.019
8. Dixon JB, Dixon ME, O’Brien PE. **Pre-operative predictors of weight loss at 1-year after Lap-Band surgery**. *Obes Surg* (2001) **11** 200-207. DOI: 10.1381/096089201321577884
9. Ma Y, Pagoto SL, Olendzki BC, Hafner AR, Perugini RA, Mason R. **Predictors of weight status following laparoscopic gastric bypass**. *Obes Surg* (2006) **16** 1227-1231. DOI: 10.1381/096089206778392284
10. Saboor Aftab SA, Halder L, Piya MK, Reddy N, Fraser I, Menon V. **Predictors of weight loss at 1 year after laparoscopic adjustable gastric banding and the role of presurgical quality of life**. *Obes Surg* (2014) **24** 885-890. DOI: 10.1007/s11695-014-1184-3
11. Al-Khyatt W, Ryall R, Leeder P, Ahmed J, Awad S. **Predictors of inadequate weight loss after laparoscopic gastric bypass for morbid obesity**. *Obes Surg* (2017) **27** 1446-1452. DOI: 10.1007/s11695-016-2500-x
12. Nickel F, de la Garza JR, Werthmann FS, Benner L, Tapking C, Karadza E. **Predictors of risk and success of obesity surgery**. *Obes Facts* (2019) **12** 427-439. DOI: 10.1159/000496939
13. Ohira M, Watanabe Y, Yamaguchi T, Saiki A, Oshiro T, Tatsuno I. **Low serum insulin-like growth factor-1 level is a predictor of low total weight loss percentage after sleeve gastrectomy**. *Surg Obes Relat Dis* (2020) **16** 1978-1987. DOI: 10.1016/j.soard.2020.07.033
14. Sisik A, Basak F. **Presurgical predictive factors of excess weight loss after laparoscopic sleeve gastrectomy**. *Obes Surg* (2020) **30** 2905-2912. DOI: 10.1007/s11695-020-04624-y
15. Hattori K, Numata N, Ikoma M, Matsuzaka A, Danielson RR. **Sex differences in the distribution of subcutaneous and internal fat**. *Hum Biol* (1991) **63** 53-63. PMID: 2004744
16. Ohta M, Kasama K, Sasaki A, Naitoh T, Seki Y, Inamine S. **Current status of laparoscopic bariatric/metabolic surgery in Japan: the sixth nationwide survey by the Japan Consortium of Obesity and Metabolic Surgery**. *Asian J Endosc Surg* (2021) **14** 170-177. DOI: 10.1111/ases.12836
17. Ma RC, Chan JC. **Type 2 diabetes in East Asians: similarities and differences with populations in Europe and the United States**. *Ann N Y Acad Sci* (2013) **1281** 64-91. DOI: 10.1111/nyas.12098
18. Seino Y, Nanjo K, Tajima N, Kadowaki T, Kashiwagi A, Araki E. **Report of the committee on the classification and diagnostic criteria of diabetes mellitus**. *J Diabetes Investig* (2010) **1** 212-228. DOI: 10.1111/j.2040-1124.2010.00074.x
19. Cruz-Jentoft AJ, Baeyens JP, Bauer JM, Boirie Y, Cederholm T, Landi F. **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
20. Sun YQ, Burgess S, Staley JR, Wood AM, Bell S, Kaptoge SK. **Body mass index and all cause mortality in HUNT and UK Biobank studies: linear and non-linear mendelian randomisation analyses**. *BMJ* (2019) **364** 1042. DOI: 10.1136/bmj.l1042
21. Zurlo F, Larson K, Bogardus C, Ravussin E. **Skeletal muscle metabolism is a major determinant of resting energy expenditure**. *J Clin Invest* (1990) **86** 1423-1427. DOI: 10.1172/JCI114857
22. Wing RR, Marcus MD, Epstein LH, Salata R. **Type II diabetic subjects lose less weight than their overweight nondiabetic spouses**. *Diabetes Care* (1987) **10** 563-566. DOI: 10.2337/diacare.10.5.563
23. Cadena-Obando D, Ramírez-Rentería C, Ferreira-Hermosillo A, Albarrán-Sanchez A, Sosa-Eroza E, Molina-Ayala M. **Are there really any predictive factors for a successful weight loss after bariatric surgery?**. *BMC Endocr Disord* (2020) **20** 20. DOI: 10.1186/s12902-020-0499-4
24. Tataranni PA, Larson DE, Snitker S, Young JB, Flatt JP, Ravussin E. **Effects of glucocorticoids on energy metabolism and food intake in humans**. *Am J Physiol* (1996) **271** E317-E325. PMID: 8770026
25. Kaess BM, Pedley A, Massaro JM, Murabito J, Hoffmann U, Fox CS. **The ratio of visceral to subcutaneous fat, a metric of body fat distribution, is a unique correlate of cardiometabolic risk**. *Diabetologia* (2012) **55** 2622-2630. DOI: 10.1007/s00125-012-2639-5
26. Vennapusa A, Panchangam RB, Kesara C, Chivukula T. **Factors predicting weight loss after “sleeve gastrectomy with loop duodenojejunal bypass” surgery for obesity**. *J Obes Metab Syndr* (2020) **29** 208-214. DOI: 10.7570/jomes20044
27. Tu Y, Pan Y, Han J, Pan J, Zhang P, Jia W. **A total weight loss of 25% shows better predictivity in evaluating the efficiency of bariatric surgery**. *Int J Obes (Lond)* (2021) **45** 396-403. DOI: 10.1038/s41366-020-00690-5
28. Cobey F, Leone L, Taliaferro I. **Effect of diethylstilbestrol on plasma 17-hydroxycorticosteroid levels in humans**. *Proc Soc Exp Biol Med* (1956) **92** 742-744. DOI: 10.3181/00379727-92-22599
29. Himeno A, Satoh-Asahara N, Usui T, Wada H, Tochiya M, Kono S. **Salivary cortisol levels are associated with outcomes of weight reduction therapy in obese Japanese patients**. *Metab* (2012) **61** 255-261. DOI: 10.1016/j.metabol.2011.06.023
30. Ma L, Liu X, Yan N, Gan Y, Wu Y, Li Y. **Associations between different cortisol measures and adiposity in children: a systematic review and meta-analysis**. *Front Nutr* (2022) **9** 879256. DOI: 10.3389/fnut.2022.879256
31. Vermeulen A, Kaufman JM, Deslypere JP, Thomas G. **Attenuated luteinizing hormone (LH) pulse amplitude but normal LH pulse frequency, and its relation to plasma androgens in hypogonadism of obese men**. *J Clin Endocrinol Metab* (1993) **76** 1140-1146. PMID: 8496304
32. Kelly DM, Jones TH. **Testosterone and obesity**. *Obes Rev* (2015) **16** 581-606. DOI: 10.1111/obr.12282
33. Mah PM, Wittert GA. **Obesity and testicular function**. *Mol Cell Endocrinol* (2010) **316** 180-186. DOI: 10.1016/j.mce.2009.06.007
34. Bambino TH, Hsueh AJ. **Direct inhibitory effect of glucocorticoids upon testicular luteinizing hormone receptor and steroidogenesis in vivo and in vitro**. *Endocrinol* (1981) **108** 2142-2148. DOI: 10.1210/endo-108-6-2142
35. Cumming DC, Quigley ME, Yen SS. **Acute suppression of circulating testosterone levels by cortisol in men**. *J Clin Endocrinol Metab* (1983) **57** 671-673. DOI: 10.1210/jcem-57-3-671
36. Kelly DM, Jones TH. **Testosterone: a metabolic hormone in health and disease**. *J Endocrinol* (2013) **217** R25-45. DOI: 10.1530/JOE-12-0455
37. Hodish I. **Insulin therapy, weight gain and prognosis**. *Diabetes Obes Metab* (2018) **20** 2085-2092. DOI: 10.1111/dom.13367
38. Biertho L, Lebel S, Marceau S, Hould FS, Lescelleur O, Marceau P. **Laparoscopic sleeve gastrectomy: with or without duodenal switch? A consecutive series of 800 cases**. *Dig Surg* (2014) **31** 48-54. DOI: 10.1159/000354313
39. Shimon O, Keidar A, Orgad R, Yemini R, Carmeli I. **Long-term effectiveness of laparoscopic conversion of sleeve gastrectomy to a biliopancreatic diversion with a duodenal switch or a Roux-en-Y gastric bypass due to weight loss failure**. *Obes Surg* (2018) **28** 1724-1730. DOI: 10.1007/s11695-017-3086-7
|
---
title: 'When and why patients drop out from benign thyroid nodules follow-up: a single
centre experience'
authors:
- Ilenia Pirola
- Mario Rotondi
- Elena Di Lodovico
- Letizia Chiara Pezzaioli
- Barbara Agosti
- Maurizio Castellano
- Alberto Ferlin
- Carlo Cappelli
journal: Endocrine
year: 2022
pmcid: PMC9988786
doi: 10.1007/s12020-022-03256-9
license: CC BY 4.0
---
# When and why patients drop out from benign thyroid nodules follow-up: a single centre experience
## Abstract
### Purpose
Drop-out in clinical long-term follow-up is a general problem that is potentially harmful to patients. No data about patients that drop out from thyroid ultrasound follow-up is available literature. The aim of the present retrospective study was to evaluate the characteristics of patients that dropped out from ultrasound thyroid nodule follow-up.
### Patients and methods
We reviewed medical records of all consecutive patients who underwent a fine needle aspiration from January 2007 to March 2009 in our department. All the patients with benign nodule(s) were recommended annual ultrasounds; patients who had dropped out from follow-up were included and a telephone interview was obtained to evaluate the reasons for dropping out.
### Results
$\frac{289}{966}$ ($30\%$) of patients with benign nodules dropped out during follow-up; $94\%$ of them within the first 5 years. Phone interviews were obtained from $\frac{201}{289}$ ($70\%$) of the patients. In the $57\%$ of cases, the main declared reason for dropping out was nodular dimension stability during the first 2-3 years; $8.7\%$ of them had forgotten about the appointment; $6.4\%$ of subjects claimed to check only serum TSH, and $3.2\%$ stated that they would undergo an ultrasound only if the nodule(s) were symptomatic. Finally, $10.7\%$ patients continued follow-up in other centres.
### Conclusion
we showed that a third of patients miss their thyroid ultrasound follow-ups, and that the major cause is the low perceived threat coming from the disease. As a certain amount of drop-out is inevitable, attempting to reinforce our patients’ awareness regarding their own health state is mandatory.
### Trial registration
Trial registration: no. 4084.
## Introduction
Thyroid nodules are a common clinical problem worldwide [1]. In addition, the number of incidental thyroid lesions detected on imaging examinations performed for different reasons is growing, probably as a consequence of the increasing use of imaging exams [2].
Fine needle aspiration cytology (FNA) is the gold standard exam to differentiate benign from malignant lesions; in fact, it is safe, cost-effective and can be performed in an outpatient setting [3]. Fortunately, more than $90\%$ of detected nodules are clinically benign lesions [4]. In this case, clinicians recommend serial ultrasound exams in accordance with the most authoritative guidelines [5, 6]. The purpose of monitoring is to identify nodular growth, with the assumption that this variable is an indication of malignancy and suggestive of the false diagnosis of a benign nodule at first evaluation.
We have recently observed in a large group of patients that a small number of benign lesions showed evidence of nodule growth during a follow-up period of 10 years [7]. However, in the same observational study $\frac{289}{1248}$ patients ($23\%$) dropped out during follow-up [7].
Drop-out in clinical long-term follow-up is a general problem that is potentially harmful to patients [8]. To the best of our knowledge, no data about patients that drop out from thyroid ultrasound follow-up have been reported in the available literature.
The aim of the present observational study was to evaluate the characteristics of patients that dropped out from ultrasound thyroid nodule follow-up.
## Subjects and methods
We reviewed the medical and imaging records of patients admitted to the FNA from January 2007 to March 2009. The indication to FNA was in accordance with the most authoritative guidelines edited in 2006 [9, 10].
All the patients with benign nodule(s) underwent annual ultrasound evaluations in our department. We used a LOGIQ 9 (GE, Healthcare, Milwaukee, WI, USA) or AplioTM500 (Toshiba Medical Systems Corp, Otawara, Japan) ultrasonographic scanner fitted with a 10–14-MHz linear transducer for morphological study.
We selected only patients who had dropped out from follow-up. Drop-out was considered for all patients who did not show up at annual ultrasound follow-up after the benign cytology report. A telephone interview was conducted in order to evaluate the reasons for dropping out.
The study was conducted according to the principles of the Helsinki Declaration and the guidelines of the Institutional Ethical Committee. All patients gave written consent for the storage and use of their data.
The study was approved by the Comitato Etico di Brescia (no. 4084).
## Statistical analysis
All data were collected in an electronic case report database. Statistical analyses were performed using SPSS 20.0 software (SPSS, Inc., Evanston, IL, USA).
Comparisons between groups and differences between proportions were calculated using χ2 for categorical variables and ANOVA test for quantitative variables, as appropriate. The Kaplan-Meier curve was fitted to determine the dropout time. Two-tailed $p \leq 0.05$ was considered statistically significant.
## Results
From January 2007 to March 2009, 1248 patients underwent thyroid fine needle aspiration cytology in our department. Among them, 966 ($77.4\%$) patients were given a cytological diagnosis of benign nodule. During follow-up, $\frac{289}{966}$ ($30\%$) patients (202 women and 87 men), with a mean age of 52.5 (21–72) years dropped out.
The baseline features of the study population, comparing patients who did drop-out towards those who did not, are reported in Table 1. Patients were comparable for gender, TSH values and nodule characteristics. In detail, hypoechoic pattern, blurred margins, microcalcifications and intranodular vascular pattern were fewer (but not significant) sings of malignancy of the drop-out group. Differently, they were older (52.5 ± 14.3 vs 45.6 ± 13.9 yrs, $p \leq 0.0001$, respectively), and showed a higher number of multiple nodules than those who were retained (52.2 vs $44.8\%$, $$p \leq 0.019$$, respectively).Table 1Baseline characteristics of the study population, comparing patients who did drop-out towards those who did notPatients who did not drop-outPatients who dropped-outp ValueGender (F/M)$\frac{474}{203202}$/870.514Mean (SD) age45.6 (13.9)52.5 (14.3)0.0001Mean (SD) TSH, mIU/L2.4 (0.8)2.5 (0.7)0.814Nodule Characteristics No of multiple nodules (%)303 ($44.8\%$)151 ($52.2\%$)0.019 Mean nodule volume in ml2.4 (1.2)2.4 (1.0)0.568 *Mean maximum* diameter in mm (SD)27.5 (9.4)27.1 (8.8)0.486Ultrasound findings Hypoechoic pattern (%)484 ($71.5\%$)197 ($68.2\%$)0.299 Blurred margins (%)356 ($52.5\%$)106 ($36.7\%$)0.238 Microcalcifications (%)428 ($63.2\%$)164 ($56.7\%$)0.269 Intranodular vascular pattern (%)287 ($42.3\%$)121 ($41.8\%$)0.880 Significantly, $\frac{271}{289}$ patients ($94\%$) dropped out within the first 5 years of follow-up, as evidenced by the Kaplan–Meier curve, shown in Fig. 1. In detail, $2.8\%$ of patients withdrew on the first appointment; 10, 14.9, and $11.1\%$ of subjects dropped out at the second, third and fourth follow-up appointments, respectively. In the fifth year of follow-up, $55\%$ of the patients did not show up for the scheduled check-up. Fig. 1The Kaplan Meier dropout estimation among patients submitted to ultrasound follow-up Phone interviews were obtained from $\frac{201}{289}$ ($70\%$) of the patients who were not retained, and the results are shown in Fig. 2.Fig. 2Reasons for missing scheduled clinic appointments In detail, $\frac{28}{201}$ ($14\%$) of patients did not provide a reason for dropping out, $8.7\%$ had forgotten about the appointment, 13 ($6.4\%$) subjects claimed to check only serum TSH, $3.2\%$ stated that they would undergo an ultrasound evaluation only if the nodule(s) were bothering them, whereas 21 ($10.7\%$) patients did proceed with follow-ups, but in a hospital closer to their home. In these last patients, no cases of malignancy were reported by phone interview.
Finally, $57\%$ of the subjects declared that they dropped out from the follow-up as the nodule was stable during the first years, without any further growth. Among these patients, $89\%$ dropped out at fifth years of follow-up.
## Discussion
This retrospective study showed that $30\%$ of patients with benign thyroid nodules do not attend ultrasound follow-up. Moreover, $94\%$ of them fail to attend within the first 5 years of monitoring.
## The American Thyroid Association (ATA) guidelines suggest ultrasound follow-up stratified accordingly to different ultrasound features, in patients with benign nodules
In particular, an ultrasound evaluation at 6–12 months for follow-up in high-risk ultrasound patterns, and no need in very low-risk group [5]. However, it is important to underline that these suggestions are based on an average-low quality of evidence, as reported by the ATA guidelines themselves [5]. Moreover, the American Association of Clinical Endocrinologists, American College of Endocrinology, and Associazione Medici Endocrinologi suggested to perform ultrasound examination in benign nodules in ~12 months after FNA. If nodules are unchanged repeat the exam after 24 months; this assumption was again based on low quality of evidence, as reported by the guidelines themselves [6].
In addition, a survey obtained from a large sample of patients referred to the Thyroid Nodule Clinic at the Brigham and Women’s Hospital in Boston, found that most benign thyroid nodule can be safely recommended for follow-up at 2–4 years with no risk of mortality or likelihood of harm [11]. Recently, our group showed that in a longer period of follow-up only $11\%$ of patients with benign lesions showed evidence of nodule growth [7]. However, it is important to emphasise that the growth occurred in a linear fashion, starting from the first year of observation, and more important nodule growth started within the first 5 years of follow-up for $79\%$ of the patients [7]. This is a key point, because in the present study we showed that almost all patients ($94\%$) who dropped out of follow-up, were lost within the first 5 years of monitoring. Furthermore, the dimensional stability of the lesions within the first three years of follow-up was the main cause of dropout in the belief that the nodule (s) would no longer grow.
It is important to highlight that the purpose of monitoring is to identify nodular growth, considering this variable as an indicator of malignancy and suggesting a false diagnosis of a benign nodule at the first evaluation [12]. Indeed, how relevant is nodule growth as a marker of the development of malignancy is still matter of controversy [13].
Several factors related to patient, provider, disease, treatment, clinical administration, or environment, may affect clinic attendance. From orthodontic and orthopaedic surveys, it has been reported that clinical compliance decreases proportionally to the interval between the completion of treatment and the follow-up examination. The most common reasons for this include change of address, and death [8]. In diabetic patients, drop-out rates are quite high, ranging from 4 to $50\%$ in different countries. The major reasons for this include low perceived concern for the disease and conflicts with work [14, 15]. To the best of our knowledge, no studies have specifically evaluated this topic in patients affected by thyroid diseases. In the present study, we showed that a large number ($30\%$) of patients are lost during follow-up and we found that the majority of patients dropped out after a few checkups, in the first 5 years of monitoring, due to a ‘general’ low perceived concern for the disease (Fig. 2).
In agreement with Grover et al. [ 16], in our study drop-out was also associated with higher age, even if our population set was younger than those reported by colleagues. This result was also evidenced by a larger set of patients [17].
In few observation studies, a longer waiting time from referral to scheduled appointment was significantly associated with missed appointments [18].
Moreover, Murdock et al. reported forgetting an appointment, particularly in case of long waiting time, as the most common reason ($36\%$) for nonattendance, in a survey-based study of gastroenterology patients [19]. Conversely, in our study less than $9\%$ of subjects forgot their appointments. We have no explanation for this difference, as our institute has not adopted any reminder (i.e. letter, phone call or SMS) days before appointment, as was done in the above-mentioned study [19]. On the other hand, in our study, low perceived concern for the disease clearly emerged as the major cause of follow-up drop out. We believe that standardised recordings of sonographic features, as EU-TIRADS [20], will improve the selection of patients with potentially malignant nodules, who could be introduce in a recall system.
In conclusion, we showed that a large set of patients miss their ultrasound thyroid follow-ups and a low perceived threat from the disease appears to be the major cause of this. As a certain amount of ‘reasons’ for patient drop-out are inevitable, attempting to reinforce our patients’ awareness regarding their own state of health is necessary.
## References
1. Sharbidre KG, Lockhart ME, Tessler FN. **Incidental thyroid nodules on imaging: relevance and management**. *Radio. Clin. North Am.* (2021.0) **59** 525-533. DOI: 10.1016/j.rcl.2021.03.004
2. Tan GH, Gharib H. **Thyroid incidentalomas: management approaches to nonpalpable nodules discovered incidentally on thyroid imaging**. *Ann. Intern. Med.* (1997.0) **126** 226-231. DOI: 10.7326/0003-4819-126-3-199702010-00009
3. Cappelli C, Pirola I, Agosti B, Tironi A, Gandossi E, Incardona P, Marini F, Guerini A, Castellano M. **Complications after fine-needle aspiration cytology: a retrospective study of 7449 consecutive thyroid nodules**. *Br. J. Oral. Maxillofac. Surg.* (2017.0) **55** 266-269. DOI: 10.1016/j.bjoms.2016.11.321
4. Hegedüs L. **Clinical practice. The thyroid nodule**. *N. Engl. J. Med.* (2004.0) **351** 1764-1771. DOI: 10.1056/NEJMcp031436
5. Haugen BR, Alexander EK, Bible KC. **2015 American Thyroid Association Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer: The American Thyroid Association Guidelines Task Force on Thyroid Nodules and Differentiated Thyroid Cancer**. *Thyroid* (2016.0) **26** 1-133. DOI: 10.1089/thy.2015.0020
6. Gharib H, Papini E, Garber JR. **American Association of Clinical Endocrinologists, American College of Endocrinology, and Associazione Medici Endocrinologi Medical Guidelines for Clinical Practice for the Diagnosis and Management of Thyroid Nodules–2016 Update**. *Endocr. Pract.* (2016.0) **22** 622-639. DOI: 10.4158/EP161208.GL
7. Cappelli C, Pirola I, Gandossi E, Rotondi M, Casella C, Lombardi D, Agosti B, Ferlin A, Castellano M. **Ultrasound of benign thyroid nodules: a 120 months follow-up study**. *Clin. Endocrinol.* (2021.0) **94** 866-871. DOI: 10.1111/cen.14408
8. Kahl B, Fischbach H, Schwarze CW. **How to deal with the drop-out in clinical follow-up studies: results of a long-term follow-up study of orthodontically treated patients**. *Am. J. Orthod. Dentofac. Orthop.* (1995.0) **108** 415-420. DOI: 10.1016/s0889-5406(95)70040-4
9. Cooper DS, Doherty GM, Haugen BR. **Management guidelines for patients with thyroid nodules and differentiated thyroid cancer**. *Thyroid* (2006.0) **16** 109-42. DOI: 10.1089/thy.2006.16.109
10. Gharib H, Papini E, Valcavi R. **American Association of Clinical Endocrinologists and Associazione Medici Endocrinologi medical guidelines for clinical practice for the diagnosis and management of thyroid nodules**. *Endocr. Pract.* (2006.0) **12** 63-102. DOI: 10.4158/EP.12.1.63
11. Nou E, Kwong N, Alexander LK, Cibas ES, Marqusee E, Alexander EK. **Determination of the optimal time interval for repeat evaluation after a benign thyroid nodule aspiration**. *J. Clin. Endocrinol. Metab.* (2014.0) **99** 510-516. DOI: 10.1210/jc.2013-3160
12. Durante C, Costante G, Lucisano G. **The natural history of benign thyroid nodules**. *JAMA* (2015.0) **313** 926-935. DOI: 10.1001/jama.2015.0956
13. Singh Ospina N, Maraka S, Espinosa DeYcaza A. **Diagnostic accuracy of thyroid nodule growth to predict malignancy in thyroid nodules with benign cytology: systematic review and meta-analysis**. *Clin. Endocrinol.* (2016.0) **85** 122-131. DOI: 10.1111/cen.12975
14. 14.Kahl B, Long-term results of a clinical follow-up evaluation of orthodontically and orthopaedically treated patients stability, relapse or new anomaly. In: University of Cologne (1993).
15. Graber AL, Davidson P, Brown AW, McRae JR, Woolridge K. **Dropout and relapse during diabetes care**. *Diabetes Care* (1992.0) **15** 1477-1483. DOI: 10.2337/diacare.15.11.1477
16. Grover S, Dua D, Chakrabarti S, Avasthi A. **Dropout rates and factors associated with dropout from treatment among elderly patients attending the outpatient services of a tertiary care hospital**. *Indian J. Psychiatry* (2018.0) **60** 49-55. DOI: 10.4103/psychiatry.IndianJPsychiatry_410_17
17. Torres O, Rothberg MB, Garb J, Ogunneye O, Onyema J, Higgins T. **Risk factor model to predict a missed clinic appointment in an urban, academic, and underserved setting**. *Popul Health Manag.* (2015.0) **18** 131-136. DOI: 10.1089/pop.2014.0047
18. Shrestha MP, Hu C, Taleban S. **Appointment wait time, primary care provider status, and patient demographics are associated with nonattendance at outpatient gastroenterology clinic**. *J. Clin. Gastroenterol.* (2017.0) **51** 433-438. DOI: 10.1097/MCG.0000000000000706
19. Murdock A, Rodgers C, Lindsay H, Tham TCK. **Why do patients not keep their appointments? Prospective study in a gastroenterology outpatient clinic**. *J. R. Soc. Med.* (2002.0) **95** 284-286. DOI: 10.1258/jrsm.95.6.284
20. Russ G, Bonnema SJ, Erdogan MF, Durante C, Ngu R, Leenhardt L. **European Thyroid Association Guidelines for ultrasound malignancy risk stratification of thyroid nodules in adults: the EU-TIRADS**. *Eur. Thyroid J.* (2017.0) **6** 225-237. DOI: 10.1159/000478927
|
---
title: 'Descending necrotizing mediastinitis: etiopathogenesis, diagnosis, treatment
and long-term consequences—a retrospective follow-up study'
authors:
- Thea Charlott Reuter
- Valentina Korell
- Jens Pfeiffer
- Gerd Jürgen Ridder
- Manuel Christoph Ketterer
- Christoph Becker
journal: European Archives of Oto-Rhino-Laryngology
year: 2022
pmcid: PMC9988808
doi: 10.1007/s00405-022-07769-x
license: CC BY 4.0
---
# Descending necrotizing mediastinitis: etiopathogenesis, diagnosis, treatment and long-term consequences—a retrospective follow-up study
## Abstract
### Purpose
The primary aim of this retrospective study was to analyze the progression of descending necrotizing mediastinitis (DNM), evaluate the impact of comorbidities on complications and mortality and to observe long-term consequences of DNM on dysphagia and measurements quality of life. DNM is a serious infectious disease that requires multimodal treatment. Current literature varies in conclusions of risk factors, management and outcome of DNM. In addition, little is known about persisting effects on quality of life.
### Methods
Retrospective data analysis of 88 patients with DNM representing the largest single-center study. Recording data of patients and diseases as well as clinical progression from 1997 to 2018. Two questionnaires were sent to the participants to measure quality of life and to detect dysphagia.
### Results
88 patients were included. The most frequently found pathogen were Streptococcus spp. ( $52\%$). $75\%$ of the patients underwent multiple surgeries, mean count of surgical procedures was 4.3 times. $84\%$ received intensive care treatment. Median length of stay on the intensive care unit was 7 days. $51\%$ had pre-existing comorbidities associated with reduced tissue oxygenation (e.g., diabetes). The most common complication was pleural effusion ($45\%$). During the observation period, the mortality rate was $9\%$. 12 questionnaires could be evaluated. $67\%$ of the participants were affected by dysphagia at the time of the survey.
### Conclusions
Descending necrotizing mediastinitis (DNM) is a severe disease requiring an immediate initiation of multimodal treatment. Although quality of life usually isn´t impaired permanently, dysphagia may often persist in patients after DNM.
## Introduction
Descending necrotizing mediastinitis (DNM) is a rare but rapidly progressive and often life-threatening form of mediastinitis. The infection originates from a head and neck source, mostly an odontogenic or oropharyngeal focus, extends via the deep fascial planes and descends into the mediastinum [1–3]. Formerly quite high mortality rates up to $40\%$ decreased in recent years to an reported overall mortality rate of $17.5\%$, credited to the widespread use of antibiotics and improvement in diagnostic, surgical and interdisciplinary management [2, 4–6]. Chest pain, high fever and crackling on palpation are described symptoms of DNM, though these unspecific disorders do not necessarily be present. All the more expeditious recognition of DNM is crucially important to promptly initiate an appropriate surgical and wide-spread antibiotic treatment to reduce morbidity and mortality [4, 7–10]. In terms of potential risk factors to suffer from a DNM several pre-existing comorbidities, especially systemic conditions with a reduced tissue oxygenation, are described, such as diabetes mellitus, severe chronic nicotine and alcohol abuse, present tumor illness and chronic pulmonary and cardiovascular diseases [4, 11, 12]. In 1983 Estrera et al. established the following diagnostic criteria: [1] Clinical manifestations of severe infection; [2] Demonstration of characteristic radiographic findings; [3] Documentation of necrotizing mediastinal infection in operation and [4] Establishment of oropharyngeal/cervical infection with descending necrotizing mediastinitis relationship [1]. Computed tomography (CT) is an essential diagnostic medium which is used for clinical diagnosis and to identify the extent of mediastinal involvement and serves as a point of comparison for postoperative control [2, 5, 13, 14]. Microbiological culture usually reveals aerobic/anaerobic coinfections corresponding to its pharyngeal or odontogenic origins [4, 11]. Severe complications with high mortality of DNM are multiple organ dysfunction syndrome, sepsis or septic shock [2, 9, 15]. The prognostic impact of certain clinical predictors such as laboratory findings and patient specific data regarding outcome and intensity of the course of DNM is differently discussed [4, 11, 12]. In 2010 Ridder et al. published a study with 45 patients suffering from DNM treated at Department of Otorhinolaryngology Head and Neck Surgery, University of Freiburg between 1997 and 2008 [4]. Our retrospective study enlarges this collective of patients until December 31, 2018. With 88 patients the present study currently represents to our knowledge the biggest cohort treated at a single center. Our aim was not only to describe patient data and the individual course of DNM but also to characterize risk factors and outcome predictors. Moreover, we wanted to present data about long-term conditions, quality of life and dysphagia.
## Methods
This is an observational descriptive retrospective cohort study of 88 cases of DNM treated at the Department of Otorhinolaryngology and Head and Neck Surgery at the University Hospital of Freiburg Germany over a period of 21 years (January 1997–December 2018). Individual patients and diseases information as well as data of individual clinical progression of DNM were taken from in house software: symptoms, medical condition and clinical findings at the moment of presentation, laboratory, microbiological and radiographical studies were recorded. Beyond that we observed pre-existing comorbidities, source of infection, antibiotic and surgical treatment, interval between arising of symptoms and presentation, respectively, initiation of operation, duration of hospitalization and intensive care unit (ICU) stay and the number of operations performed. Subsequently we sent two questionnaires to all patients supposed to be alive: The Eating Assessment Tool 10 (EAT-10) was used to measure persistent swallowing difficulties and the Short Form [36] Health Survey (SF-36) was used to assess the actual state of health at the moment of interrogation. Diagnosis of DNM was established by clinical, radiographical and intraoperative findings. The diagnostic criteria of DNM defined by Estrera et al. were fulfilled [1]. The classification of Endo et al. was used to divide cases of DNM based on the anatomical extent in CT findings [16]. We excluded patients with mediastinitis from non-descending cause and also those with an iatrogenic etiology, e.g., postoperatively.
Unless stated otherwise, data are presented as mean (SD). Statistical analysis was done using R-based software (Jamovi, jamoviproject [2018], version 0.9.2.3.). Shapiro–Wilk test showed that data were not normally distributed, the Mann–Whitney U test was applied to continuous variables. To determine group differences for category variables Fisher’s exact test was applied. P ≤ 0.05 was considered statistically significant. To analyze the influencing factors on the dependent variable death we performed logistic regression analyzes. The study was approved by the ethics committee of the Albert-Ludwig-University of Freiburg (Votum Nr., $\frac{369}{18}$, 07.02.2019).
## Demographic data
The mean age of the 88 patients included in the study was 54.7 years (median 55.5 years; youngest 3 years and oldest 84 years). 51 were male ($57.9\%$) and 37 female ($42.1\%$).
## Symptoms and source
At the time of presentation frequent symptoms were sore throat ($$n = 48$$, $54.6\%$), swelling and redness ($$n = 37$$, $42.1\%$), odynophagia ($$n = 30$$, $34.1\%$), dyspnea ($$n = 26$$, $29.6\%$) and dysphagia ($$n = 21$$, $23.9\%$). Furthermore, symptoms were fever ($$n = 17$$, $19.3\%$), neck pain and headache ($$n = 13$$, $14.8\%$), worsening of general condition and gnathospasm (respectively, $$n = 7$$, $7.9\%$), hoarseness ($$n = 6$$, $6.8\%$), foreign body sensation and upper back and/or shoulder pain (respectively, $$n = 5$$, $5.7\%$), sternal pain ($$n = 4$$, $4.6\%$) and stridor or stupor ($$n = 2$$, $2.3\%$). In most of the cases the etiology could not be reported with absolute certainty ($$n = 33$$, $37.5\%$). In 28 patients ($31.8\%$), DNM originated from a previous infection, in 4, $4.6\%$, from post-radiogenic situation and one patient suffered from injection abscess. The time interval between onset of symptoms to admission was 3.53 days on average (median 3 days).
## Comorbidities
More than half of the patients ($$n = 45$$, $51.1\%$), suffered from pre-existing diseases with reduced tissue oxygenation: cardiac insufficiency ($$n = 28$$, $31.8\%$), diabetes mellitus ($$n = 16$$, $18.2\%$), respiratory insufficiency ($$n = 12$$, $13.6\%$), adiposities per magna ($$n = 6$$, $6.8\%$), post cervical radiation ($$n = 4$$, $4.6\%$) and peripheral arterial obstructive disease ($$n = 3$$, $3.4\%$). We calculated the individual Charlson Comorbidity Index of each patient, mean 2.9 points and median 2 points. $76.1\%$ had ≥ 1 points. Chronic nicotine, alcohol and other substance abuse was detected in 31, 13 and 7 patients, respectively.
## Diagnostics
First step of the diagnostic pathway in any case was a thorough otolaryngologic examination. Amongst others laboratory findings on admission consisted of white blood cell (WBC) count and C-reactive protein (CRP). The WBC count ranged from 1.4 to 34.5 cells/mm3 (mean 14.9 cells/mm3, median 13.4 cells/mm3). CRP on admission ranged from 0 to 495 mg/L (mean 94.8 mg/L, median 39.6 mg/L). Microbiological examinations were obtained in 70 patients. In $50\%$ of the microbiological examinations a mixed anaerobic and aerobic spectrum was detected. In 33 patients microbiological examination showed an anaerobic infection. In two patients, microbiological examination showed an exclusively aerobic infection. A detailed distribution of microorganisms is shown in Fig. 1. Depending on the patient`s condition B-mode ultrasound was performed. Another essential part of the diagnostic pathway were radiographic examinations. All of the 88 patients underwent imaging diagnostic. 86 patients received CT scans, at most 15 times (mean 3.14, median 2). 69 patients received CT scans in the follow-up. 16 patients ($18.2\%$) underwent magnetic resonance imaging (MRI). In five cases MRI was performed multiple times. With the results of the radiographic imaging the conclusion of the localization of the DNM was achieved. Most affected site of the mediastinum was the upper part above the carina in 71 patients ($80.7\%$) corresponding to Endo Type I. In 12 ($13.6\%$), respectively, 10 ($11.4\%$) patients the lower anterior (Endo Type IIA) or posterior (Endo Type IIB) mediastinum was involved as well. Fig. 1Distribution of microorganisms causing descending necrotizing mediastinitis
## Therapy, trends, complications and mortality
Before presenting in our department 38 patients ($43.1\%$) had already been treated with an antibiotic. At the latest at time of admission an empiric antibiotic therapy was started in all patients. Most frequently prescribed antibiotics were cephalosporine of the 2nd or 3rd generation (64 patients, $72.7\%$) often combined with metronidazole (61 patients, $69.3\%$) to cover the anaerobically spectrum of the infectious organisms. As soon as the results of the polymicrobial samples or the appropriate antibiograms were given a modification of the antibiotic treatment was performed if necessary. Indeed this was required in 50 patients of 88 ($56.8\%$) in the course of the hospitalization. Complementing the antibiotic therapy a surgery was performed in 83 patients. 69 patients ($83.1\%$) underwent a panendoscopy, 53 patients ($63.7\%$) outer transcervical drainage, 48 patients ($57.8\%$) collar mediastinotomy. Inner drainage of the abscess was executed in 31 patients ($37.6\%$), tracheotomy in 29 patients ($34.9\%$) and foreign bodies were extracted in five patients ($6\%$). 15 patients were operated only a single time. 66 patients underwent multiple surgical procedures (mean 4.26, median 3, maximum 26). To ensure wound drainage, cervical incisions were sutured incompletely and with the help of large lumen tubes inserted in the wounds the situs was daily irrigated with antiseptic solutions. If required this took place in repeated sessions under general anesthesia. The mean duration of hospitalization was 25.5 days (median 18 days, maximum 139 days). 74 ($84.1\%$) patients were admitted to an intensive care unit (ICU). The mean duration of ICU stay was 13.3 days (median 7 days, maximum 110 days). During the treatment of DNM, a total of 64 patients ($72.7\%$) developed complications (Fig. 2). 8 of our 88 patients with DNM died, thus the mortality rate was $9.1\%$. Death occurred meanly after 11.4 days past admission (median 2.5).Fig. 2Complications during the treatment of descending necrotizing mediastinitis
## Group differences and influencing factors on mortality
Analysis of relevant group differences between survivors ($$n = 80$$) and deceased ($$n = 8$$) was conducted. This revealed that those patients who died were distinctly older (mean age 67.1 years in patients who died, mean age 53.5 in patients who survived DNM; $$p \leq 0.04$$) and had higher CRP levels on admission (mean CRP 239.7 mg/L in patients who died, mean CRP 80.3 mg/L in patients who survived DNM; $$p \leq 0.05$$). No significant differences between survivors and deceased were found in the WBC count on admission, extent of involved mediastinum and interval between onset of symptoms and admission. In addition, pre-existing comorbidities, diabetes mellitus, nicotine abuse and antibiotic therapy before admission were not relevantly different between the two groups. Of all the observed factors CRP level on admission ($$p \leq 0.01$$) and greater age of the patients ($$p \leq 0.05$$) showed a significant impact on the mortality of our patients.
## Quality of life and dysphagia in the long-term
Of all the patients being sent the questionnaires only 12 patients answered. The mean interval between DNM and interrogation was 106.7 months (median 107.5 months). The mean age of the patients at the time of interrogation was 53.2 years (median 60.5 years; minimum 12 years and maximum 84 years). Concerning the EAT-10 questionnaire helping to detect swallowing restrictions and dysphagia 8 of the 12 patients reached a score ≥ 3 points, indicating abnormal high values. The highest score was 22 points (mean 8.2 points, median 5.5 points). One patient declared not to have any problems at all (0 points). Highest score was assessed to the statement “I cough when I eat” (mean 1.5 points, median 1.5 points), whereas odynophagia was weighted least (mean 0.3 points, median 0 points). The SF-36 questionnaire represents the health status during the last 4 weeks at the time of filling in. Comparative values are represented by average data of a general population. Our patients´ answers were partially incomplete which made it impossible to receive complete records of all of the 12 patients. Health-related quality of life is sectioned into physical and mental health status. SF-36 revealed, that the physical health summary is above average in 5 patients and below average in 4 patients. The mental health score lays about average in 7 patients and below average in 3 patients.
## Discussion
First described in 1938 by Pearson et al. DNM remains a serious, aggressive disease often associated with fatal outcome [3, 5, 17, 18]. Even though it is a rare disease DNM should not be underestimated. It can still lead to sepsis and death, while once high mortality rates from $49\%$ in 1938 decreased during the last years to $17.5\%$ described in a review by Prado-Calleros in 2016 as diagnostics and multimodal therapy concepts improved [4, 5, 19–21]. Deep neck infections, DNI, develop from odontogenic or oropharyngeal infections and rapidly spread via deep fascial planes downward into mediastinum resulting in DNM [2, 11, 22, 23]. Based on its pattern of spreading Endo et al. developed a classification of CT findings: (I) focal form: localized to the upper mediastinal space above the carina. ( II) Diffuse form reaches out below the tracheal bifurcation and is subdivided into (IIA): lower anterior mediastinum and (IIB): lower posterior mediastinum [16].
Common literature agrees that a diagnosis at early stage of DNM as well as a prompt initiation of appropriate medical and radical surgical treatment are imperative [2, 4, 9, 11, 17, 24–29].
However, an immediate recognition of DNM can be challenging due to the following reasons: DNM is a rare disease not only ENT experts and general surgeons but also family doctors, pediatricians and general practitioners should keep in mind. Though only 2–$3\%$ of deep space neck infections develop to more serious infections, such as mediastinitis, one should always be aware of such a severe course as it can progress very fast [7, 30, 31]. Furthermore, it can affect all age groups as the results of our investigation, age range 3–84 years, show in concordance with the literature. In doubt the clinical picture is sufficient to suspect DNM and the patient should promptly be admitted to a hospital with ENT specialist for thorough examination of the pharynx and larynx including fiberoptic transnasal endoscopy and efficient initiation of imaging diagnostics [4, 32]. Moreover, symptoms may not be distinct. The infection is often clinically silent especially at the beginning and may be veiled by analgesics delaying the diagnosis. Therefore, a clear association between symptom, severity and extent of the disease is difficult [4, 12, 13, 32, 33].
In our study most common clinical symptoms, pain, swelling and odynophagia, at the time of presentation are related to common oropharyngeal infections and DNIs. However, disorders correlating explicitly with mediastinitis such as chest pain and mediastinal emphysema are less frequent in our patients [15, 27, 33, 34].
The clear relation between infections of an odontogenic or pharyngeal source and the development of DNM is undoubted [9, 17, 35]. The anatomic continuity of the posterior pharyngeal, parapharyngeal and submandibular spaces with the mediastinum explains this smooth transition. Once an infection enters one of these spaces a spreading downward is promoted because of gravity, respiration, intrathoracic negative pressure and absence of barriers in fascial planes [4, 35, 36]. The so-called danger space lies posterior to the alar fascia, runs from skull base to diaphragm and allows even a contralateral spread of the infection [9].
Also, in our patient group previous infection of odontogenic or pharyngeal origin is often mentioned. Four patients even reported a post-radiogenic situation of the neck. This underlines the importance of a diligent execution of medical history as these patients are more likely to suffer from wound healing disorders [37].
Several predisposing risk factors are well-recognized referring to patients with DNM, such as diabetes mellitus, poor dental or oral hygiene, immunosuppression, renal and liver failure, high blood pressure and recent steroids. Moreover, chronic nicotine, alcohol and IV drug abuse happens to appear frequently in patients with DNM [9, 11, 12, 15, 38–40]. The description that patients with DNM suffer significantly more often from comorbidities compared to patients with DNI underlines the high impact of these pre-existing risk factors to develop DNM [10, 12]. Our analyses turned the attention especially to comorbidities with a reduced tissue oxygenation as these expedite the development of DNM [4]. More than half of our patients suffered from pre-existing diseases associated with impaired tissue oxygenation, e.g., cardiac or respiratory insufficiency, diabetes mellitus, adiposities per magna and post cervical radiation. This coincides with the observations of Kocher et al. [ 2]. Nonetheless also young, healthy patients with no medical history can suffer from DNM [41–43].
The Charlson Comorbidity Index, CCI, is a well-approved statistical test to predict the mortality of patients based on underlying comorbidities and has widely been used since its first description in 1987. Patients are divided in four groups (0 points, 1–2 points, 3–4 points and 5 points) correspondent to an increasing mortality risk. We used the age adjusted variant [44, 45]. Only $23.9\%$ of our patients had an index of 0 points. Park et al. observed a CCI ≥ 1 in $54.1\%$ in a group of 135 patients with DNI [46]. The results of CCI in our study underlines even more the role of comorbidities in patients with DNM and indicates, that patients with pre-existing reduced health conditions may have a higher risk to develop DNM.
The polymicrobiological nature of DNM as often described in diverse studies makes sense considering its origin as an oropharyngeal infection that once it has penetrated the mucosa spreads downward to the mediastinum [2, 33]. Likewise, Palma et al. half of our patients had mixed anaerobic and aerobic infections [11]. Nonetheless no result regarding extensive microbiological examinations was obtained in 18 of our patients. This lack of microbiological results was observed as well by other authors. It might be due to the fact that quite a high number of patients, e.g., in our study $43.1\%$, has already been treated with antibiotics before microbiological examinations was obtained [4, 12, 32].
In more than three-quarter of our patients the upper part of mediastinum above the carina was affected corresponding to Endo Type I. CT scans are an early part of the diagnostic pathway and crucial in recognizing DNM before it spreads even deeper. Radiation exposure should be accepted in case of deep neck infection or just its clinical suspicion to exclude DNM yet at the beginning. Freitas et al. even suggests an algorithm with serial CT scans every 24–48 h to assess disease progression [32]. In line with that also most of our patients received CT scans in the follow-up to monitor the course of DNM and if applicable to act immediately. The expressiveness of CT scans after repeated surgical interventions with changed tissue textures can be hindered. Moreover, laboratory findings as WBC or CRP levels can be indistinct, e.g., if they do not decrease conspicuously. In case of doubt, we, therefore, favor a liberal decision to revision surgery.
The surgical drainage of the affected parts of mediastinum is without any doubt mandatory. However, there is still some controversy about the optimal approach. A proper approach should thoughtful be chosen according to patients condition, the extent of the disease and also the surgeons experience [2]. Responsible for the mortality of DNM is not only a delayed diagnosis but as well an inappropriate drainage of the mediastinum. Therefore, the latter should also carefully be focused on [11]. As most of our patients suffer from DNM Endo Type I, where infection is located in the upper part of the mediastinum, we are of the opinion that in these cases a sole transcervical drainage is sufficient for surgical debridement and necrotic tissue removal. A highly aggressive surgical treatment irrespective of the extent of DNM as formerly advocated by many authors can also implicate disadvantages and may go along with a higher complication rate [6, 13, 28, 33, 36]. In case of an advanced stage of DNM, Endo Type II, we also certainly support a thoracic approach via thoracotomy in combination with transcervical debridement for drainage of the upper and lower parts of mediastinum. Most accomplished is a posterolateral thoracotomy on the more affected side [5, 18]. As healing of DNM may often be protracted and revision surgery is commonly needed we usually perform incomplete closure and insertion of large tubes in the wounds combined with daily irrigation with antiseptic solutions.
The infectious process of DNM can frequently lead to pharyngolaryngeal edema and consequently cause dyspnea. We recommend in these cases of foreseeable difficult intubation an awake or at least video assisted intubation if necessary in preparedness of coniotomy. If upper airway is compromised or patients seem to need a long-term treatment because of severity of DNM we approve tracheotomy. Nevertheless, we refuse to overhasty decisions for routine tracheotomy in patients with DNM as spreading of cervical infection may occur [2, 5].
Amongst our patients mortality was significantly influenced by higher age and higher CRP-levels. Although not relevant to mortality in the present survey we do think other factors as comorbidities, extent of involved mediastinal site and duration of discomfort until treatment still have an impact on progress and outcome of DNM. Therefore, the overall impression of the patient must not be disregarded.
We are confronted with a modest number of questionnaires sent back by our patients who were also treated at ICU. Besides we face a heterogeneous distribution of time passed, since patients suffered from DNM. In addition, Gerth et al. report a loss of patients in follow-up controls in their review regarding changes in health-related quality of life after ICU [47]. Amongst others a reason to not sending back the questionnaire might be the long period of time of more than 20 years included. Patients suffering from DNM years ago not only became essentially older as well as their health condition presumably impaired and, therefore, precluded answering and sending back the questionnaires. Basically the SF-36 we used is well-approved and the most employed questionnaire to evaluate quality of life after critical illness. Despite the quantity our findings implicate a tendency that patients post DNM rather suffer from impaired physical than mental health in long-term. This is concomitant with the results of Gerth et al. [ 47]. Kramer et al. indeed reported that physical-functionary health-related quality-of-life subscales remain depressed even years after the acute illness in patients after ICU [48].
Our findings implicate that persisting dysphagia after DNM actually occurs hence two-thirds of our patients had above average high values in EAT-10 questionnaire. This corresponds to the observation that dysphagia is well-recognized as a late complication of DNM by several authors. We likewise subscribe to the advice of comprehensive long-term physiotherapy and logopedic support in patients with DNM not only to prevent aspirations but also to improve their quality of life [8, 17, 40].
## Conclusions
Despite great advances in diagnostics and treatment of DNM it remains a severe disease that requires a prompt, strenuous and multidisciplinary approach to prevent fatal outcomes. Particularly high CRP levels and old age have a remarkable influence on mortality. Nonetheless comorbidities, e.g., associated with impaired tissue oxygenation should not be disregarded. Even though acuteness of DNM is conquered long-term influence as dysphagia and in some cases declined quality of life may persist.
## References
1. Estrera AS, Landay MJ, Grisham JM, Sinn DP, Platt MR. **Descending necrotizing mediastinitis**. *Surg Gyn Obstet* (1983) **157** 545-552
2. Kocher GJ, Hoksch B, Caversaccio M. **Diffuse descending necrotizing mediastinitis: surgical therapy and outcome in a single-centre series**. *Eur J Cardiothorac Surg* (2012) **42** e66-72. DOI: 10.1093/ejcts/ezs385
3. Pearse HE. **Mediastinitis following cervical suppuration**. *Ann Surg* (1938) **108** 588-611. DOI: 10.1097/00000658-193810000-00009
4. Ridder GJ, Maier W, Kinzer S. **Descending necrotizing mediastinitis: contemporary trends in etiology, diagnosis, management, and outcome**. *Ann Surg* (2010) **251** 528-534. DOI: 10.1097/SLA.0b013e3181c1b0d1
5. Prado-Calleros HM, Jiménez-Fuentes E, Jiménez-Escobar I. **Descending necrotizing mediastinitis: systematic review on its treatment in the last 6 years, 75 years after its description**. *Head Neck* (2016) **38** E2275-E2283. DOI: 10.1002/hed.24183
6. Kiernan PD, Hernandez A, Byrne WD. **Descending cervical mediastinitis**. *Ann Thorac Surg* (1998) **65** 1483-1488. DOI: 10.1016/s0003-4975(98)00142-8
7. Pulst-Korenberg A, Morris SC. **Descending necrotizing mediastinitis resulting from pharyngitis with perforation of the aryepiglottic fold**. *Case Rep Emerg Med* (2020) **2020** 4963493. DOI: 10.1155/2020/4963493
8. Fujiwara K, Koyama S, Fukuhara T, Takeuchi H. **Successful surgical treatment for dysphagia secondary to descending necrotizing mediastinitis**. *Yonago Acta Med* (2019) **62** 253-257. DOI: 10.33160/yam.2019.09.002
9. Taylor M, Patel H, Khwaja S, Rammohan K. **Descending cervical mediastinitis: the multidisciplinary surgical approach**. *Eur Arch Otorhinolaryngol* (2019) **276** 2075-2079. DOI: 10.1007/s00405-019-05471-z
10. Chen K-C, Chen J-S, Kuo S-W. **Descending necrotizing mediastinitis: a 10-year surgical experience in a single institution**. *J Thorac Cardiovasc Surg* (2008) **136** 191-198. DOI: 10.1016/j.jtcvs.2008.01.009
11. Palma DM, Giuliano S, Cracchiolo AN. **Clinical features and outcome of patients with descending necrotizing mediastinitis: prospective analysis of 34 cases**. *Infection* (2016) **44** 77-84. DOI: 10.1007/s15010-015-0838-y
12. Kimura A, Miyamoto S, Yamashita T. **Clinical predictors of descending necrotizing mediastinitis after deep neck infections**. *Laryngoscope* (2020) **130** E567-E572. DOI: 10.1002/lary.28406
13. Misthos P, Katsaragakis S, Kakaris S. **Descending necrotizing anterior mediastinitis: analysis of survival and surgical treatment modalities**. *J Oral Maxillofac Surg* (2007) **65** 635-639. DOI: 10.1016/j.joms.2006.06.287
14. Zhang JL, Chen WX, Li JJ. **Clinical analysis of 27 cases with descending necrotizing mediastinitis**. *Zhonghua Er Bi Yan Hou Tou Jing Wai Ke Za Zhi* (2019) **54** 919-923. DOI: 10.3760/cma.j.issn.1673-0860.2019.12.007
15. Weaver E, Nguyen X, Brooks MA. **Descending necrotising mediastinitis: two case reports and review of the literature**. *Eur Resp Rev* (2010) **19** 141-149. DOI: 10.1183/09059180.00001110
16. Endo S, Murayama F, Hasegawa T. **Guideline of surgical management based on diffusion of descending necrotizing mediastinitis**. *Jpn J Thorac Cardiovasc Surg* (1999) **47** 14-19. DOI: 10.1007/BF03217934
17. Sandner A, Börgermann J, Kösling S. **Descending necrotizing mediastinitis: early detection and radical surgery are crucial**. *J Oral Maxillofac Surg* (2007) **65** 794-800. DOI: 10.1016/j.joms.2005.11.075
18. Sumi Y. **Descending necrotizing mediastinitis: 5 years of published data in Japan**. *Acute Med Surg* (2015) **2** 1-12. DOI: 10.1002/ams2.56
19. Dajer-Fadel WL, Ibarra-Pérez C, Sánchez-Velázquez LD. **Descending necrotizing mediastinitis below the tracheal carina**. *Asian Cardiovasc Thorac Ann* (2014) **22** 176-182. DOI: 10.1177/0218492313485589
20. Deu-Martín M, Saez-Barba M, Sanz IL. **Mortality risk factors in descending necrotising mediastinitis**. *Arch Bronconeumol* (2010) **46** 182-187. DOI: 10.1016/j.arbres.2010.01.008
21. Vieira F, Allen SM, Stocks RMS, Thompson JW. **Deep neck infection**. *Otolaryngol Clin N Am* (2008) **41** 459-483. DOI: 10.1016/j.otc.2008.01.002
22. Papalia E. **Descending necrotizing mediastinitis: surgical management**. *Eur J Cardiothoracic Surg* (2001) **20** 739-742. DOI: 10.1016/s1010-7940(01)00790-4
23. Kinzer S, Pfeiffer J, Becker S, Ridder GJ. **Severe deep neck space infections and mediastinitis of odontogenic origin: clinical relevance and implications for diagnosis and treatment**. *Acta Otolaryngol* (2009) **129** 62-70. DOI: 10.1080/00016480802008181
24. Sarna T, Sengupta T, Miloro M, Kolokythas A. **Cervical necrotizing fasciitis with descending mediastinitis: literature review and case report**. *J Oral Maxillofac Surg* (2012) **70** 1342-1350. DOI: 10.1016/j.joms.2011.05.007
25. Biasotto M, Chiandussi S, Costantinides F, Di Lenarda R. **Descending necrotizing mediastinitis of odontogenic origin**. *Recent Pat Antiinfect Drug Discov* (2009) **4** 143-150. DOI: 10.2174/157489109788490299
26. Ma C, Zhou L, Zhao J-Z. **Multidisciplinary treatment of deep neck infection associated with descending necrotizing mediastinitis: a single-centre experience**. *J Int Med Res* (2019) **47** 6027-6040. DOI: 10.1177/0300060519879308
27. Sancho LMM, Minamoto H, Fernandez A. **Descending necrotizing mediastinitis: a retrospective surgical experience**. *Eur J Cardiothorac Surg* (1999) **16** 200-205. DOI: 10.1016/s1010-7940(99)00168-2
28. Marty-Ané C-H, Berthet J-P, Alric P. **Management of descending necrotizing mediastinitis: an aggressive treatment for an aggressive disease**. *Ann Thorac Surg* (1999) **68** 212-217. DOI: 10.1016/s0003-4975(99)00453-1
29. Bakir S, Tanriverdi MH, Gün R. **Deep neck space infections: a retrospective review of 173 cases**. *Am J Otolaryngol* (2012) **33** 56-63. DOI: 10.1016/j.amjoto.2011.01.003
30. Marioni G, Staffieri A, Parisi S. **Rational diagnostic and therapeutic management of deep neck infections: analysis of 233 consecutive cases**. *Ann Otol Rhinol Laryngol* (2010) **119** 181-187. DOI: 10.1177/000348941011900306
31. Huang T-T, Liu T-C, Chen P-R. **Deep neck infection: analysis of 185 cases**. *Head Neck* (2004) **26** 854-860. DOI: 10.1002/hed.20014
32. de Freitas RP, Fahy CP, Brooker DS. **Descending necrotising mediastinitis: a safe treatment algorithm**. *Eur Arch Otorhinolaryngol* (2007) **264** 181-187. DOI: 10.1007/s00405-006-0174-z
33. Roccia F, Pecorari GC, Oliaro A. **Ten years of descending necrotizing mediastinitis: management of 23 cases**. *J Oral Maxillofac Surg* (2007) **65** 1716-1724. DOI: 10.1016/j.joms.2006.10.060
34. Kluge J. **Die akute und chronische Mediastinitis**. *Chirurg* (2016) **87** 469-477. DOI: 10.1007/s00104-016-0172-7
35. Qu L, Xu H, Liang X. **A retrospective cohort study of risk factors for descending necrotizing mediastinitis caused by multispace infection in the maxillofacial region**. *J Oral Maxillofac Surg* (2020) **78** 386-393. DOI: 10.1016/j.joms.2019.11.017
36. Wheatley MJ, Stirling MC, Kirsh MM. **Descending necrotizing mediastinitis: Transcervical drainage is not enough**. *Ann Thorac Surg* (1990) **49** 780-784. DOI: 10.1016/0003-4975(90)90022-x
37. Haubner F, Ohmann E, Pohl F. **Wound healing after radiation therapy: review of the literature**. *Radiat Oncol* (2012) **24** 162. DOI: 10.1186/1748-717X-7-162
38. Bottin R, Marioni G, Rinaldi R. **Deep neck infection: a present-day complication. A retrospective review of 83 cases (1998–2001)**. *Eur Arch Otorhinolaryngol* (2003) **260** 576-579. DOI: 10.1007/s00405-003-0634-7
39. Byers J, Lowe T, Goodall CA. **Acute cervico-facial infection in Scotland 2010: patterns of presentation, patient demographics and recording of systemic involvement**. *Br J Oral Maxillofac Surg* (2012) **50** 626-630. DOI: 10.1016/j.bjoms.2011.11.013
40. Glen P, Morrison J. **Diffuse descending necrotising mediastinitis and pleural empyema secondary to acute odontogenic infection resulting in severe dysphagia**. *BMJ Case Rep* (2016) **2016** bcr2015212145. DOI: 10.1136/bcr-2015-212145
41. Makeieff M, Gresillon N, Berthet JP. **Management of descending necrotizing mediastinitis**. *Laryngoscope* (2004) **114** 772-775. DOI: 10.1097/00005537-200404000-00035
42. Ochi N, Wakabayashi T, Urakami A. **Descending necrotizing mediastinitis in a healthy young adult**. *Ther Clin Risk Manag* (2018) **14** 2013-2017. DOI: 10.2147/TCRM.S176520
43. Ye R-H, Yang J-C, Hong H-H. **Descending necrotizing mediastinitis caused by Streptococcus constellatus in an immunocompetent patient: case report and review of the literature**. *BMC Pulm Med* (2020) **20** 43. DOI: 10.1186/s12890-020-1068-3
44. 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
45. Quan H, Li B, Couris CM. **Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries**. *Am J Epidemiol* (2011) **173** 676-682. DOI: 10.1093/aje/kwq433
46. Park MJ, Kim JW, Kim Y. **Initial nutritional status and clinical outcomes in patients with deep neck infection**. *Clin Exp Otorhinolaryngol.* (2018) **11** 293-300. DOI: 10.21053/ceo.2018.00108
47. Gerth AMJ, Hatch RA, Young JD, Watkinson PJ. **Changes in health-related quality of life after discharge from an intensive care unit: a systematic review**. *Anaesthesia* (2019) **74** 100-108. DOI: 10.1111/anae.14444
48. Kramer CL. **Intensive care unit-acquired weakness**. *Neurol Clin* (2017) **35** 723-736. DOI: 10.1016/j.ncl.2017.06.008
|
---
title: The effect of a-Lipoic acid (ALA) on oxidative stress, inflammation, and apoptosis
in high glucose–induced human corneal epithelial cells
authors:
- Zhen Li
- Yu Han
- Yan Ji
- Kexin Sun
- Yanyi Chen
- Ke Hu
journal: Graefe's Archive for Clinical and Experimental Ophthalmology
year: 2022
pmcid: PMC9988813
doi: 10.1007/s00417-022-05784-6
license: CC BY 4.0
---
# The effect of a-Lipoic acid (ALA) on oxidative stress, inflammation, and apoptosis in high glucose–induced human corneal epithelial cells
## Abstract
### Purpose
Oxidative stress and inflammation had been proved to play important role in the progression of diabetic keratopathy (DK). The excessive accumulation of AGEs and their bond to AGE receptor (RAGE) in corneas that cause the formation of oxygen radicals and the release of inflammatory cytokines, induce cell apoptosis. Our current study was aimed to evaluate the effect of ALA on AGEs accumulation as well as to study the molecular mechanism of ALA against AGE-RAGE axis mediated oxidative stress, apoptosis, and inflammation in HG-induced HCECs, so as to provide cytological basis for the treatment of DK.
### Methods
HCECs were cultured in a variety concentration of glucose medium (5.5, 10, 25, 30, 40, and 50 mM) for 48 h. The cell proliferation was evaluated by CCK-8 assay. Apoptosis was investigated with the Annexin V- fluorescein isothiocyanate (V-FITC)/PI kit, while, the apoptotic cells were determined by flow cytometer and TUNEL cells apoptosis Kit. According to the results of cell proliferation and cell apoptosis, 25 mM glucose medium was used in the following HG experiment. The effect of ALA on HG-induced HCECs was evaluated. The HCECs were treated with 5.5 mM glucose (normal glucose group, NG group), 5.5 mM glucose + 22.5 mM mannitol (osmotic pressure control group, OP group), 25 mM glucose (high glucose group, HG group) and 25 mM glucose + ALA (HG + ALA group) for 24 and 48 h. The accumulation of intracellular AGEs was detected by ELISA kit. The RAGE, catalase (CAT), superoxide dismutase 2 (SOD2), cleaved cysteine-aspartic acid protease-3 (Cleaved caspase-3), Toll-like receptors 4 (TLR4), Nod-like receptor protein 3 (NLRP3) inflammasome, interleukin 1 beta (IL-1 ß), and interleukin 18 (IL-18) were quantified by RT-PCR, Western blotting, and Immunofluorescence, respectively. Reactive oxygen species (ROS) production was evaluated by fluorescence microscope and fluorescence microplate reader.
### Results
When the glucose medium was higher than 25 mM, cell proliferation was significantly inhibited and apoptosis ratio was increased ($P \leq 0.001$). In HG environment, ALA treatment alleviated the inhibition of HCECs in a dose-dependent manner, 25 μM ALA was the minimum effective dose. ALA could significantly reduce the intracellular accumulation of AGEs ($P \leq 0.001$), activate protein and genes expression of CAT and SOD2 ($P \leq 0.001$), and therefore inhibited ROS-induced oxidative stress and cells apoptosis. Besides, ALA could effectively down-regulate the protein and gene level of RAGE, TLR4, NLRP3, IL-1B, IL-18 ($P \leq 0.05$), and therefore alleviated AGEs-RAGE-TLR4-NLRP3 pathway–induced inflammation in HG-induced HCECs.
### Conclusion
Our study indicated that ALA could be a desired treatment for DK due to its potential capacity of reducing accumulation of advanced glycation end products (AGEs) and down-regulating AGE-RAGE axis–mediated oxidative stress, cell apoptosis, and inflammation in high glucose (HG)–induced human corneal epithelial cells (HCECs), which may provide cytological basis for therapeutic targets that are ultimately of clinical benefit.
## Introduction
Diabetes mellitus (DM) is a chronic metabolic disease. According to the statistics, there are about 537 million adults (20–79 years) are living with diabetes (1 in 10). This number is predicted to rise to 643 million by 2030 and 783 million by 2045 1. DM had been widely proved to be the most important risk factor for developing all kinds of chronic ocular diseases, such as dry eye disease (DED), delayed epithelial wound healing, corneal edema, recurrent erosions, superficial punctate keratitis, corneal ulcers, and so on 2. These problems seem to be affecting up to $70\%$ of diabetic patients 3. Although the pathogenic mechanisms of ocular surface damage in DM are very complicated, most of the studies suggested that oxidative stress and inflammation played an important role in the progression of DK 4–6.
AGEs are a heterogeneous group of irreversible adducts from glucose-protein condensation reactions, as well as lipids and nucleic acids exposed to reducing sugars. The factors like hyperglycemia or aging have been reported to involve in the generation of AGEs. The bonding of AGEs to their receptors (i.e., RAGE) attributes oxidative stress, which may induce cellular dysfunction and pathophysiological effects. AGEs-mediated ROS generation emerges in the activation of discrete sets of transcription factors and associated genes contributing to various pathological consequences including cardiovascular diseases, cancer, chronic inflammation, neurological disorders and DK 7–11. AGE-RAGE axis also appears to have tight relationship in increasing oxidative stress and inflammation during diabetes 12,13.
α-Lipoic acid (ALA; 1,2-dithiolane-3-pentanoic acid) also known as thioctic acid, is traditionally recognized as an essential cofactor in mitochondrial respiratory enzymes that catalyze the oxidative decarboxylation reaction 14. Due to its powerful antioxidant value, which has been widely used to prevent and treat some metabolic diseases, such as diabetic peripheral neuropathy, reducing plasma cholesterol, protecting liver and heart damage, inhibiting the occurrence of cancer, inhibiting inflammation caused by allergy, arthritis, asthma and anti-aging 15,16. However, the effect and molecular mechanisms of ALA on DK were unknown. Therefore, our aim was to evaluate the effect of ALA on AGEs accumulation and study the molecular mechanism of ALA against AGE-RAGE axis mediated oxidative stress, apoptosis and inflammation in HG-induced HCECs, so as to provide cytological basis for the treatment of DK.
## Cell culture and treatment
HCECs (Purchased from BNCC, Beijing, China) were cultured in MEM medium (Basal Media, Shanghai, China) supplemented with $10\%$ fetal bovine serum (FBS, Gibco, USA) and Penicillin–Streptomycin Solution (10 kU/ml Penicillin, 10 mg/ml Streptomycin, Procell, Wuhan, China) in a humidified incubator at 37 °C containing $5\%$ CO2. According to the groups, D-Glucose (Procell, Wuhan, China) was dilated to different concentration. The whole research procedure was approved by the Ethics Committee of Chongqing Medical University and was performed following the Declaration of Helsinki.
## Cell proliferation assay
The capacity of cell proliferation was evaluated by CCK-8 assay (Shanghai Yi Sheng Biotechnology Co., Ltd., Shanghai, China). This commercially available kit was used following the manufacturer’s instructions. In brief, HCECs (1 × 104 cells per well) were seeded in 96 well plates. The culture medium was replaced after the cells adhered to the wall. The culture medium was replaced with 5.5, 10, 25, 30, 40, and 50 mM glucose for 48 h and then 10 μl CCK-8 solution was added into the microplate. Two hours after incubating, the microplate reader was used to detect the absorbance at 450 nm. According to the results of cell proliferation, 25 mM glucose was used in following HG stimulation. The treating concentration of ALA was determined by the capacity of proliferation. In brief, HCECs (1 × 104 cells per well) were seeded in 96 well plates, the culture medium was replaced after the cells adhered to the wall. The medium was replaced with various concentration of ALA (25, 50, 125, 250, 500 μM, MCE, China) for 24 h at 37 °C in an atmosphere containing $5\%$ CO2. According to the cell viability, 25 and 50 μM ALA were used in the following experiment. In order to avoid the influence of osmotic pressure changes caused by high glucose on experimental results, we set 5.5 mM glucose + 22.5 mM mannitol as mannitol hypertonic control. After the HCECs adhered to the wall, the medium was, respectively, replaced with 5.5 mM glucose, 5.5 mM glucose + 22.5 mM mannitol, 25 mM glucose, 25 mM glucose + 25 μM ALA and 25 mM glucose + 50 μM ALA for 24 h at 37 °C in an atmosphere containing $5\%$ CO2. 10 μl CCK-8 solution was added into the microplate, 2 h after incubating, the microplate reader was used to detect the absorbance at 450 nm.
## Measurement of intracellular ROS
HCECs were cultured in a 96-well plate with the density of 2 × 104 cells per well. After the cells adhered to the wall, according to the groups, the culture medium was, respectively, replaced with 5.5 mM glucose, 5.5 mM glucose + 22.5 mM mannitol, 25 mM glucose and 25 mM glucose + 25 μM ALA for 24 and 48 h. The production of intracellular ROS was measured with an ROS assay kit (Beyotime Biotechnology, Shanghai, China) using 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) (10 Um) as a fluorescence probe in dark. After incubation with 200 μl per wall diluted DCFH-DA for 30 min at 37 °C, cells were washed three times with PBS. The fluorescence released was detected using a fluorescence microplate reader (Molecular Devices, Sunnyvale, CA, USA) at an excitation and emission wave length of 480 nm and 520 nm. A fluorescence microscope (Leica Microsystems, Wetzlar, Germany) was used to obtain the images. The percentage increase in fluorescence per well was calculated by the formula 【(Ft30-Ft0)/Ft0 × 100】, where Ft30 = fluorescence at time 30 min and Ft0 = fluorescence at time 0 min 13.
## Cell apoptosis assay
HCECs were seeded in a six-well plate according to the groups. When the cells adhered to the wall, the culture medium was, respectively, replaced according to the groups. After incubating for 24 and 48 h, cells were harvested and suspended in 1 × Annexin V Bingding Buffer. Then, cells were stained with Annexin V-FITC and propidium iodide (PI) using a cell apoptosis detection kit (Procell Life Scinece & Technology Co., Ltd, China). After incubating for 20 min, cell apoptosis was analyzed by a flow cytometry (BD Biosciences, NJ, USA). The percentage of apoptotic cells was analyzed using FlowJo software (Becton–Dickinson-San Jose, CA, USA). TUNEL cells apoptosis staining was detected by commercial TUNEL cell apoptosis detection Kit (Beyotime, Shanghai, China).
## Protein expression
The protein expression of RAGE, CAT, SOD2, TLR4, NLRP3, Cleaved caspase-3,IL-1β and IL-18 was evaluated by Western blotting (WB). GAPDH was considered as control. In brief, the medium was removed from the six-well plates and cells were washed with iced PBS for three times. The RIPA lysis buffer (Beyotime, Shanghai, China) was used to extract total proteins. Then, a BCA Kit (Beyotime, Shanghai, China) was executed for detecting the concentrations of proteins. A total of 20 μg protein samples in cell lysates were separated in SDS-PAGE gels (FuturePAGETM$4\%$-$20\%$, ACE Biotechnology, Nanjing, China) followed by transferring to polyvinylidene difluoride membranes (Millipore, Bedford, MA, USA). The unspecific bands were blocked with QuickBlock™ Blocking Buffer (Beyotime, Shanghai, China) for 15 min. Then, the membranes were washed with TBST for 15 min. Subsequently, these membranes were, respectively, probed with primary antibodies and stored in 4℃ overnight (RAGE Rabbit pAb, Catalase Rabbit pAb, SOD2 Rabbit mAb, TLR4 Rabbit pAb, NLRP3 Rabbit pAb, Cleaved caspase-3, IL1 beta Rabbit pAb, IL18 Rabbit pAb, GAPDH Rabbit mAb, AB clonal, China). After having been washed with TBST for three times, these membranes were incubated with HRP-conjugated Goat Anti-Rabbit IgG (AB clonal, Wuhan, China). The bands were visualized using an Odyssey Infrared Imaging Scanner (LI-COR Biosciences). Intensity of bands was examined using Image-J software (National Institutes of Health, Bethesda, MA, USA). The protein expression was normalized to GAPDH levels. The cells were lysed with RIPA and the supernatant was collected for detecting the concentration of AGEs by commercially available Enzyme-linked immunosorbent assay (Glucose-derived AGEs ELISA Kit) (Shanghai Xitang Biotechnology Co., Ltd., Shanghai, China).
## Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) analysis
Total RNA was extracted from cells using Trizol reagent (Invitrogen). Then, first-strand cDNA was synthesized from RNA using a Sensiscript RT kit (Takara Biotechnology Co., Ltd., Japan) following the manufacturer’s recommendations. Then, qPCR was conducted with SYBR Green Supermix (Bio-Rad Laboratories, Inc.) on the ABI 7500 system (Applied Biosystems; Thermo Fisher Scientific, Inc.). *Relative* gene expression was normalized to GAPDH. The relative primers were described in Table 1.Table 1Nucleotide sequence of specific primers used for polymerase chain reaction amplification (human)Gene primerNucleotide sequenceGen IDCAT847Forward primerTGGAGCTGGTAACCCAGTAGGReverse primerCCTTTGCCTTGGAGTATTTGGTASOD26648Forward primerGGAAGCCATCAAACGTGACTTReverse primerCCCGTTCCTTATTGAAACCAAGCRAGE177Forward primerACTACCGAGTCCGTGTCTACCReverse primerGGAACACCAGCCGTGAGTTTLR47099Forward primerTCCATAAAAGCCGAAAGGTGReverse primerGATACCAGCACGACTGCTCANLRP3114,548Forward primerCGTGAGTCCCATTAAGATGGAGTReverse primerCCCGACAGTGGATATAGAACAGAIL-183606Forward primerTCTTCATTGACCAAGGAAATCGGReverse primerTCCGGGGTGCATTATCTCTACIL-1B3553Forward primerATGATGGCTTATTACAGTGGCAAReverse primerGTCGGAGATTCGTAGCTGGAGAPDH2597Forward primerGGAGCGAHATCCCTCCAAAATReverse primerGGCTGTTGTCATACTTCTCATGG
## Immunofluorescence
HCECs were seeded in 24-well plates according to the groups. When the cells adhered to the wall, the culture medium was, respectively, replaced according to the groups. After incubating for 48 h, the HCECs were fixed by $4\%$ PFA for 10 min, rinsed with PBS to remove PFA, then incubated for 12 min in $1\%$ Triton X-100 (Aladdin, Shanghai, China). After being washed, cells were incubated in $3\%$ BSA for blocking. The permeabilized cells were incubated with primary antibody overnight at 4 °C. Then, after being served with PBS again, cells were incubated with goat anti-rabbit IgG (H + L), fluorescein Isothiocyanate conjugate (TransGen, Beijing, China) for 1 h at room temperature and DAPI for 7 min. Cover the cell slides on the micro slides with Mounting Medium, antifading, and around the cell slides with neutral balsam to prevent drying. Specimens were examined with a Leica SP8 Laser Scanning confocal microscope (Leica, Wetzlar, Germany).
## Statistical analysis
Representative data from three independent experiments are presented as means ± standard deviation (SD). A significant difference between two groups using Mann–Whitney U test/Student’s test and more than two groups by One-way ANOVA followed by Bonferroni’s multiple comparison test using Graph pad prism 8 (Graphpad Holdings, LLC, San Diego, CA, USA). Statistical significance was defined as $p \leq 0.05.$
## Effect of HG stimulation on cell proliferation and apoptosis
From the result of Fig. 1A, we found that when the concentration of glucose medium reached to 25 mM, the cell proliferation rate (%) of HCECs had been decreased to 87.374 ± 1.157. However, compared with 5.5 mM glucose medium, HCECs treated with 10 mM glucose showed a higher cell proliferation rate (%), which was increased to 130.05 ± 26.77. The results of the above indicated that slightly high glucose environment was favorable for cell proliferation, but when the glucose medium concentration was higher than 25 mM, cell proliferation would be remarkably inhibited ($$P \leq 0.00046$$). From the results of flow cytometry (Fig. 1B), we got insight that the number of apoptosis of HCECs was increased after the stimulation with 25 mM glucose. From 25 to 50 mM glucose, the total apoptosis rate of HCECs was gradually increased from 16.1 to $21.53\%$. Concurrently, the expression of apoptosis-related proteins cleaved caspase-3 were evaluated by Western blotting. As presented in Fig. 1C and D, compared to 5.5 mM glucose medium group, from 25 to 50 mM glucose medium, the cleaved caspase-3 protein expression of HCECs was significantly up-regulated ($P \leq 0.0001$). The data suggested that glucose medium higher than 25 mM would inhibit HCECs proliferation and increase cell apoptosis. Therefore, 25 mM glucose was used in the following HG experiments. Fig. 1Effect of high glucose on cell proliferation and apoptosis. ( A) The cell proliferation of HCECs was evaluated by CCK-8 assay. Data are the mean ± SD of three independent experiments for all groups. The cell proliferation ratio (%) of HCECs in 5.5 mM glucose medium was 130.05 ± 26.77, in 25 mM glucose medium was 87.374 ± 1.157. ( B) *Cells apoptosis* was evaluated by Flow cytometry. From 25 mM glucose to 50 mM glucose, the total apoptosis rate of HCECs was gradually increased from 16.1 to $21.53\%$. Q3 = early apoptosis, Q2 = late apoptosis, total apoptosis = Q2 + Q3. ( C and D) The expression of apoptosis-related proteins cleaved caspase-3 was evaluated by Western blotting. GAPDH was considered as control. Data shown were from three independent experiments. Compared to 5.5 mM glucose medium group, from 25 to 50 mM glucose medium, the cleaved caspase-3 protein expression of HCECs was significantly up-regulate. “##” $P \leq 0.001$
## Effect of ALA on cell proliferation in HG-induced HCECs
After treatment with different doses of ALA (25, 50, 125, 250, and 500 μM) in normal glucose medium for 24 h, the effect of ALA on cell proliferation was detected by CCK-8 assay. From the result of Fig. 2A, we found that compared to normal glucose medium without ALA, the administration of 50 μM ALA presented a notable effect on the HCECs proliferation ($$P \leq 0.002$$). However, the administration of 125 μM ALA presented a remarkable inhibition on cell proliferation ($$P \leq 0.013$$). In the following experiment, both 25 and 50 μM ALA were used to investigate the cell proliferation in HG environment. As exhibited in Fig. 2B, compared to HG group, the administration of 25 μM ALA presented significant effect on promoting cell proliferation ($P \leq 0.001$).Fig. 2Effect of ALA on cell proliferation. ( A) Compared to normal glucose medium without ALA, administration of 50 uM ALA presented a notably effect on the cell proliferation. However, the administration of 125 uM ALA presented a remarkable inhibition on cell proliferation. Data shown were from three independent experiments. “#” $P \leq 0.05.$ ( B) Compared to HG group, the administration of 25 uM ALA presented significant effect on promoting cell proliferation. Data shown were from three independent experiments. “##” $P \leq 0.001$
## Effect of ALA on accumulation of AGEs and expression of RAGE in HG-induced HCECs
As presented in Fig. 3A, compared to NG group, the production of AGEs in HG group was significantly increased after the stimulation of 25 mM glucose for 24 ($P \leq 0.001$) and 48 h ($P \leq 0.001$), but which could be significantly inhibited by the treatment of 25 uM ALA ($P \leq 0.001$). The concentration of AGEs in HG group and HG + ALA group was positively correlated with the treatment time. We were surprised to observe that the administration of 25 μM ALA significantly alleviated the accumulation of AGEs and down-regulated protein expression of RAGE ($$p \leq 0.012$$) in HG induced HCECs (Fig. 3B and C). The mRNA level of RAGE was further confirmed by RT-PCR (Fig. 3D). In HG-induced HCECs, 25 μM ALA treatment significantly down-regulated the mRNA expression of RAGE (1.86-fold, $$p \leq 0.00085$$). The same results also were observed in immunofluorescence images (Fig. 3E). Collectively, our results suggested that ALA treatment could alleviate the accumulation of AGEs, down-regulate the expression of RAGE in HG-induced HCECs. Fig. 3Effect of ALA on accumulation of AGEs and expression of RAGE in HECEs. ( A) The intracellular concentration of AGEs was detected by commercially available ELISA Kit. Data were from three independent experiments for all groups. Compared to NG group, the production of AGEs in HG group was significantly increased after the stimulation of 25 mM glucose for 24 and 48 h ($P \leq 0.001$), but which could be significantly inhibited by the treatment of 25 uM ALA. “##” $P \leq 0.001.$ ( B and C) The relative expression of RAGE protein was evaluated by Western blotting. GAPDH was considered as control. Data were from three independent experiments for all groups. Compared to NG group, the protein expression of RAGE was significantly increased in HG group, however, the administration of ALA significantly down-regulated protein expression of RAGE in high glucose medium. “#” $P \leq 0.05.$ ( D) The relative expression of RAGE mRNA was evaluated by RT-qPCR. GAPDH was considered as control. Data were from three independent experiments for all groups. Compared to NG group, AGER mRNA was significantly up-regulated in HG group. However, compared to HG group, the administration of ALA significantly down-regulated the mRNA level of AGER in HG + ALA group. “##” $P \leq 0.001.$ ( E) Captured immunofluorescent staining images of RAGE. DAPI and their merged images were also demonstrated (scale bar = 400 µm)
## Effect of ALA on oxidative stress in HG-induced HCECs
To confirm the protective effect of ALA on oxidative stress induced by HG stimulation, the production of intracellular ROS, the oxidative stress–related proteins and genes were detected. From the results of Fig. 4A and B, we found that the increased fluorescence in HG group ($$P \leq 0.017$$) was obviously higher than that in NG group, but the increased fluorescence could be obviously alleviated after the treatment of ALA ($$P \leq 0.008$$). Meanwhile, we found that compared to NG group, the relative expression of oxidative stress–related CAT($$P \leq 0.014$$) and SOD2 ($$P \leq 0.011$$) protein was obviously decreased in HG group. We also observed that the CAT ($$P \leq 0.002$$) and SOD2 ($$P \leq 0.008$$) mRNA level was significantly down-regulated. But the protein and gene expression of CAT ($P \leq 0.001$) and SOD2 ($P \leq 0.001$) could be significantly activated after the treatment of ALA (Fig. 4C to E). These results demonstrated that ALA could inhibit the formation of ROS, activate the protein and gene expression of CAT and SOD2, and therefore alleviated oxidative stress induced by HG stimulation. Fig. 4Effect of ALA on oxidative stress in high glucose induced HCECs. ( A and B) Intracellular ROS was detected with an ROS assay kit ((DCFH-DA). ROS positive cells were stained with green fluorescence (scale bar = 100 µm). Data were from three independent experiments for all groups. Compared to NG group, the increased fluorescence was significantly increased in HG group. However, the administration of ALA significantly inhibited the increased fluorescence. “#” $P \leq 0.05.$ ( C and D) The relative protein expression of CAT and SOD2 was evaluated by Western blotting. GAPDH was considered as control. Data were from three independent experiments for all groups. Compare to NG group, the relative protein expression of CAT and SOD2 was significantly inhibited in HG group. However, the administration of ALA could significantly activate the expression of CAT and SOD2 in high glucose induced HCECs. “#” $P \leq 0.05$, “##” $P \leq 0.001.$ ( E) The relative expression of AGER mRNA was evaluated by RT-qPCR. GAPDH was considered as control. Data were from three independent experiments for all groups. Compared to NG group, CAT and SOD2 mRNA was significantly down-regulated in HG group. However, compared to HG group, the administration of ALA significantly up-regulated the mRNA level of CAT and SOD2 in HG + ALA group. “#” $P \leq 0.05$, “##” $P \leq 0.001$
## Effect of ALA on cell apoptosis in HG-induced HCECs
As presented in Fig. 5A, we found that compared to NG group, the early and total apoptotic cells in HG group were increased after the stimulation of HG. In HG group, the early apoptotic cells ratio accounted to $8.65\%$ and the total apoptotic cells ratio accounted to $21.45\%$. But, after the treatment of ALA, the cell apoptosis ratio was decreased, especially the early cell apoptosis was decreased from 8.65 to $4.79\%$. The result of cleaved caspase-3 protein expression also proved that ALA treatment could have significantly inhibited HG-induced HCECs apoptosis ($P \leq 0.001$). ( Fig. 5B and C). TUNEL apoptotic cells staining also confirmed the similar results to flow cytometry (Fig. 5D).Fig. 5Effect of ALA on cell apoptosis in high glucose induced HCECs. ( A) In HG group, the early apoptotic cells accounted to $8.65\%$ and the total apoptotic cells accounted to $21.45\%$. However, after the administration of ALA, the cell apoptosis ratio was decreased, especially the early cell apoptosis was decreased from 8.65 to $4.79\%$. Q3 = early apoptosis, Q2 = late apoptosis, total apoptosis = Q2 + Q3. ( B and C) The protein expression of apoptosis-related cleaved caspase-3 was evaluated by Western blotting. GAPDH was considered as control. Data were from three independent experiments for all groups. Compared to NG group, the relative protein expression of cleave caspse-3 was significantly increased in HG group. However, compared to HG group, the relative expression of cleave caspse-3 protein could be remarkably down-regulated by the treatment of ALA. “##” $P \leq 0.001.$ ( D) TUNEL staining was used to detect cell apoptosis. DAPI (blue), TUNEL positive staining (red), scale bar = 100 µm
## Effect of ALA on inflammation in HG-induced HCECs
In order to evaluate the effect of ALA on inflammation in HG stimulation, we detected the protein and gene level of TLR4, NLRP3, IL-1β and IL-18 in HG induce HCECs. As presented in Fig. 6A and B, we found that compared to NG group, the relative protein expression of TLR4 ($P \leq 0.001$), NLRP3 ($$P \leq 0.002$$), IL-1ß ($$P \leq 0.0003$$) and IL-18 ($$P \leq 0.005$$) was significantly increased in the HG group. However, the administration of ALA presented significant effect on down-regulating the expression of TLR4 ($P \leq 0.001$), NLRP3 ($$P \leq 0.005$$), IL-1ß ($$P \leq 0.002$$) and IL-18 ($$P \leq 0.008$$). Meanwhile, the results of RT-PCR (Fig. 6C) also proved that ALA could effectively down-regulate HG-induced mRNA level of TLR4 ($$P \leq 0.0181$$), NLRP3 ($$P \leq 0.01$$), IL-1ß ($$P \leq 0.001$$) and IL-18 ($$P \leq 0.025$$). The same results also could be observed in immunofluorescence images (Fig. 7). Hence, we speculated that ALA could play an anti-inflammatory role through down-regulation of TLR4-NLRP3 signaling pathway. Fig. 6Effect of ALA on inflammation in high glucose induced HCECs. ( A and B) The protein expression of TLR4, NLRP3, IL-1β and IL-18 was evaluated by Western blotting. GAPDH was considered as control. Data were from three independent experiments for all groups. Compared to NG group, the relative protein expression of TLR4, NLRP3, IL-1β and IL-18 was significantly increased in HG group. However, compared to HG group, the relative protein expression of TLR4, NLRP3, IL-1β and IL-18 could be remarkably down-regulated by the treatment of ALA. “#” $P \leq 0.05$, “##” $P \leq 0.001.$ ( C) The relative expression of TLR4, NLRP3, IL-1β and IL-18 was evaluated by RT-qPCR. GAPDH was considered as control. Data were from three independent experiments for all groups. Compared to NG group, TLR4, NLRP3, IL-1β and IL-18 mRNA were significantly up-regulated in HG group. However, compared to HG group, the administration of ALA significantly down-regulated the mRNA level of TLR4, NLRP3, IL-1β and IL-18 in HG + ALA group. “#” $P \leq 0.05$, “##” $P \leq 0.001$Fig. 7Immunofluorescent staining images. ( A) Captured immunofluorescent staining images of TLR4. DAPI and their merged images were also demonstrated (scale bar = 400 µm). ( B) Captured immunofluorescent staining images of NLRP3. DAPI and their merged images were also demonstrated (scale bar = 400 µm). ( C) Captured immunofluorescent staining images of IL-1β. DAPI and their merged images were also demonstrated (scale bar = 400 µm). ( D) Captured immunofluorescent staining images of IL-18. DAPI and their merged images were also demonstrated (scale bar = 400 µm)
## Discussion
Diabetes is a public health problem that concern all countries in the world. DK, the most frequent clinical condition affecting the human cornea, is a potential sight-threatening condition caused mostly by epithelial disturbances that are of clinical and research attention due to its severity. Diabetic keratopathy exhibits several clinical manifestations, including persistent corneal epithelial erosion, superficial punctate keratopathy, delayed epithelial regeneration, and decreased corneal sensitivity that may lead to compromised visual acuity or permanent vision loss 17.
A large number of studies have proved that sustained hyperglycemia causes non-enzymatic glycosylation of various proteins in vivo and AGEs are important in the pathogenesis of chronic complications of diabetes 18. Besides the diabetic complications, AGEs also play important role in many widespread age-related pathology such as Alzheimer’s disease, decreased skin elasticity 19–21, male erectile dysfunction 22,23 and atherosclerosis promotion, progression, and prevention 24. AGEs can accumulate in and out of cells and is therefore associated with a variety of eye diseases, such as diabetic retinopathy 25, cataract 26, glaucoma 27, age-related macular degeneration 28, dry eye 29, diabetic keratopathy 30, and so on. The pathogenic mechanism involved may be mainly through non-receptor–mediated and receptor-mediated. In the process of AGEs formation, proteins are modified to promote cross-linking between proteins, resulting in abnormal structure and function of corresponding organ or cells. The glycosylation of collagen and elastin cloud weakens the compliance of tissues and organs, while the glycosylation of laminin and fibronectin could affect cell adhesion, migration and differentiation. Glycosylation of enzymes, receptors and DNA molecules in cells could affect normal physiological functions and metabolic activities, resulting in cell degeneration and death. AGEs also can bind to RAGE and therefore induce local cell signaling towards apoptotic and proliferative pathways, as well as increases oxidative stress and inflammation. In our study, we evaluated the intracellular concentration of AGEs and the expression level of RAGE, we found that the concentration of AGEs and the relative expression level of RAGE were significantly higher in HG group, the concentration of AGEs in cells was positively correlated with the duration of HG stimulation. We still found that the HCECs viability would be significantly inhibited, while the ratio of apoptosis would be greatly increased in HG stimulation. Besides the above results, we also observed that HG stimulation would increase the accumulation of ROS and induce up-regulating of TLR4, NLRPE, IL-1B, IL-18 in HCECs. We speculated that HG stimulation could increase the accumulation of AGEs and accelerate AGEs binding to RAGE, which would activate ROS production, and therefore induce apoptosis, up-regulate TLR4 signaling pathway and activate inflammatory cytokine release in HCECs. These speculations were consistent with some former reports 31,32.
It is likely that reducing the accumulation of AGEs and inhibiting AGEs binding to RAGE may be a potential treatment strategy for diabetic keratopathy. ALA is a kind of multifunctional antioxidant with chemical structure of 6, 8-dilipoic acid, which can be worked as a coenzyme to participate in the acyl transfer in the metabolism of substances in our body and eliminate free radical–induced aging and disease. The potential antioxidant effect of ALA can be ascribed to the direct reactive oxygen species (ROS) scavenging capacity, metal ion chelating ability and ability to restore the cellular antioxidants such as reduced glutathione (GSH), coenzyme Q, vitamins C and E levels. In laboratory experiments, the effect of ALA on protein glycation and AGEs formation had been investigated both in vitro and in vivo. Dietary supplementation of ALA in rats fed chronically with glucose significantly decreased mitochondrial superoxide in the heart and AGEs formation in the aorta 33. Supplementation of ALA in fructose-fed-rats significantly attenuated AGEs-mediated skin-collagen crosslinking and other physicochemical abnormalities 34. ALA could protect against fructose-mediated myoglobin glycation in vitro by inhibiting the early and intermediate glycation reactions involved in the formation of AGEs and therefore ALA supplementation is beneficial in the prevention of AGEs-mediated diabetic and cardiovascular complications 35. At the cellular level, ALA is reduced to dihydrolipoic acid (DHLA), which has a number of cellular actions including free radical scavenging and modulating oxidative stress and inflammatory pathways 36. Moreover, ALA could markedly suppress AGEs-induced activation of NF-kB in cultured vascular endothelial cells and in retinal endothelial cells 37. Exogenous administration of ALA diminished AGEs-induced endothelial expression of vascular cell adhesion molecule-1 (VCAM-1) and monocyte binding to endothelium 38. Furthermore, ALA prevented the up-regulation of AGEs-induced inducible nitric oxide synthase (iNOS) expression and nitric oxide (NO) production in murine microglial cells 39. In the field of eye diseases, ALA had been successfully employed in a variety of in vivo models, such as diabetic retinal vascular lesions 40, cataract 41, glaucoma42, and diabetic corneal diseases43, all of which include complex and intimate association between increased oxidative stress and increased inflammation 44. However, none of the previous studies had focused on the effects of ALA on AGEs accumulation and the molecular mechanisms on AGE-RAGE signaling pathway in HG-induced HCECs.
In our study, we observed that 25 μM ALA showed a significant effect on inhibiting AGEs accumulation and down-regulating RAGE expression in HG-induced HCECs. The possible mechanism of action of ALA in inhibiting the formation of AGEs might include (a) blocking the amino groups of protein, thus preventing its glycation with free sugar, (b) blocking the carbonyl groups of reducing sugars, (c) blocking the Amadori products and dicarbonyl intermediates which may reduce glycation, as well as AGEs formation, preventing autoxidation of fructose and glycoxidation. Mechanistic studies on the effects of ALA on the redox status of insulin-responsive cells revealed that ALA stimulated glucose uptake by affecting components of the insulin-signaling pathway and prevented excesses glucose converting to AGEs 45. Besides, we were surprised to find that the ROS production was significantly inhibited after the treatment of ALA, meanwhile, the oxidative stress–related CAT and SOD2 expression was significantly activated. In the study of AM et al. 46, they also proved the similar results. In HG group, the early apoptotic cells accounted to $8.65\%$ and the total apoptotic cells accounted to $21.45\%$. However, after the treatment of ALA, the cell apoptosis ratio was obviously decreased, especially the early cell apoptosis was decreased from 8.65 to $4.79\%$. We speculated that the anti-apoptotic effect of ALA was mainly through inhibiting AGEs accumulation and RAGE binding, thus inhibiting AGE-RAGE-ROS pathway–induced cell apoptosis.
In our study, we also studied the anti-inflammatory effects of ALA on HG-induced HCECs. We were delighted to find that ALA had predominant effect on down-regulating the expression of TLR4, NLRP3, IL-1β and IL-18. Some previous studies had reported that AGEs could bind to RAGE or up-regulate the expression of RAGE, and therefore activate TLR4, NLRP3, FOXC2, pPKCβ 1, JNK, p38 MAPK and NF-κB pathways 12,32,47–51. All of the results were related to extracellular and intracellular oxidative stress and inflammation. Hu et al. 52 had demonstrated that TLR4-NLRP3 pathway plays a critical role in the inflammation and apoptosis of retinal ganglion cells induced by high glucose. Garibotto et al. 53 also proved that TLR4, nucleotide-binding oligomerization domain-containing protein 2 (NOD2), and NLRP3 inflammasome are involved in the production and persistence of inflammation in diabetic nephropathy. In the study of DK, Wan et al. 54 studied the effect of the NLRP3 inflammasome on diabetic corneal wound healing and never regeneration. They proved that NLRP3 inflammasome–mediated inflammation and pyroptosis contributed to DK pathogenesis. They also revealed that the accumulated AGEs promoted hyperactivation of the NLRP3 inflammasome through ROS production, genetically and pharmacologically blocking the AGEs/ROS/NLRP3 inflammasome axis significantly expedited diabetic corneal epithelial wound closure and nerve regeneration. Therefore, we speculated that the anti-inflammatory effects of ALA on HG-induced HCECs were probably through blocking the AGEs-RAGE-TLR4-NLRP3 axis. However, the specific signaling pathway of ALA against mediated oxidative stress, inflammation and apoptosis is still worthy of further study.
## Conclusion
It is likely that reducing the accumulation of AGEs and inhibiting AGEs binding to RAGE may be a potential treatment strategy for diabetic keratopathy. Our study indicated that ALA could be a desired treatment for diabetic keratopathy due to its potential capacity of alleviating AGEs accumulation, inhibiting AGEs-RAGE-ROS–mediated oxidative stress and cells apoptosis, as well as down-regulating AGEs-RAGE-TLR4-NLRP3 axis–induced inflammation. Understanding the mechanisms of ALA on HG-induced HCECs may provide cytological basis for therapeutic targets that are ultimately of clinical benefit.
## References
1. 1.The IDF Diabetes Atlas 10th edition. https://diabetesatlas.org/
2. Co Shih K, Lam KS-L, Tong L. **A systematic review on the impact of diabetes mellitus on the ocular surface**. *Nutrition & Diabetes.* (2017.0) **7** e251. DOI: 10.1038/nutd.2017.4
3. Ljubimov AV. **Diabetic complications in the cornea**. *Vision Res* (2017.0) **139** 138-152. DOI: 10.1016/j.visres.2017.03.002
4. Gunay M, Celik G, Yildiz E. **Ocular surface characteristics in diabetic children**. *Curr Eye Res* (2016.0) **41** 1526-1531. DOI: 10.3109/02713683.2015.1136421
5. Yagci A, Gurdal C. **The role and treatment of inflammation in dry eye disease**. *Int Ophthalmol* (2014.0) **34** 1291-2130. DOI: 10.1007/s10792-014-9969-x
6. Yoo TK, Ein Oh. **Diabetes mellitus is associated with dry eye syndrome: a meta-analysis**. *Int Ophthalmol* (2019.0) **39** 2611-2620. DOI: 10.1007/s10792-019-01110-y
7. Kuhla A, Ludwig SC, Kuhla B. **Advanced glycation end products are mitogenic signals and trigger cell cycle reentry of neurons in Alzheimer’s disease brain**. *Neurobiol Aging* (2015.0) **36** 753-761. DOI: 10.1016/j.neurobiolaging.2014.09.025
8. 8.Pathak C, Vaidya FU, Waghela BN et al (2021) Advanced glycation end products mediated oxidative stress and regulated cell death signaling in cancer. In: Chakraborti S., Ray B.K., Roychowdhury S. (eds) Handbook of oxidative stress in cancer: mechanistic aspects. Springer, Singapore. 10.1007/978-981-15-4501-6_44-1
9. Amara F, Hafez S, Orabi A. **Review of diabetic polyneuropathy: pathogenesis, diagnosis and management according to the consensus of Egyptian experts**. *Curr Diabetes Rev* (2019.0) **15** 340-345. DOI: 10.2174/1573399815666190226150402
10. Kaji Y, Usui T, Oshika T. **Advanced glycation end products in diabetic corneas**. *Invest Ophthalmol Vis Sci* (2000.0) **41** 362-368. PMID: 10670463
11. Yucel I, Yucel G, Akar Y. **Transmission electron microscopy and autofluorescence findings in the cornea of diabetic rats treated with aminoguanidine**. *Can J Ophthalmol* (2006.0) **41** 60-66. DOI: 10.1016/S0008-4182(06)80068-2
12. Kim J, Kim CS, Sohn E. **Involvement of advanced glycation end products, oxidative stress and nuclear factor-kappaB in the development of diabetic keratopathy**. *Graefes Arch Clin Exp Ophthalmol* (2011.0) **249** 529-536. DOI: 10.1007/s00417-010-1573-9
13. Kim J, Kim CS, Kim H. **Protection against advanced glycation end products and oxidative stress during the development of diabetic keratopathy by KIOM-79**. *J Pharm Pharmacol* (2011.0) **63** 524-530. DOI: 10.1111/j.2042-7158.2010.01206.x
14. Packer L, Witt EH, Tritschler HJ. **Alpha-Lipoic acid as a biological antioxidant**. *Free Radic Biol Med* (1995.0) **19** 227-250. DOI: 10.1016/0891-5849(95)00017-R
15. Evans JL. **Alpha-lipoic acid: a multifunctional antioxidant that improves insulin sensitivity in patients with type 2 diabetes**. *Diabetes Technol* (2000.0) **2** 401-413. DOI: 10.1089/15209150050194279
16. Biewenga GP, Haenen GR, Bast A. **The pharma-cology of the antioxidant lipoic acid**. *Gen Pharmacol* (1997.0) **29** 315-331. DOI: 10.1016/S0306-3623(96)00474-0
17. 17.Priyadarsini S, Whelchel A, Nicholas S et al (2020) Diabetic keratopathy: insights and challenges. Surv Ophthalmol. 10.1016/j.survophthal.2020.02.005
18. Bejarano E, Taylor A. **Too sweet: problems of protein glycation in the eye**. *Exp Eye Res.* (2019.0) **178** 255-262. DOI: 10.1016/j.exer.2018.08.017
19. Munch G, Schinzel R, Loske C. **Alzheimer’s disease—synergistic effects of glucose deficit, oxidative stress and advanced glycation endproducts**. *J Neural Transm* (1998.0) **105** 439-461. DOI: 10.1007/s007020050069
20. Monnier VM, Cerami A. **Nonenzymatic browning in vivo: possible process for aging of long-lived proteins**. *Science* (1981.0) **211** 491-493. DOI: 10.1126/science.6779377
21. Sell DR, Monnier VM. **End-stage renal disease and diabetes catalyze the formation of a pentose-derived crosslink from aging human collagen**. *J Clin Invest* (1990.0) **85** 380-340. DOI: 10.1172/JCI114449
22. Jiaan DB, Seftel AD, Fogarty J. **Age-related increase in an advanced glycation end product in penile tissue**. *World J Urol* (1995.0) **13** 369-375. DOI: 10.1007/BF00191219
23. Seftel AD, Vaziri ND, Ni Z. **Advanced glycation end products in human penis: elevation in diabetic tissue, site of deposition, and possible effect through iNOS or eNOS**. *Urology* (1997.0) **50** 1016-1026. DOI: 10.1016/S0090-4295(97)00512-8
24. Stitt AW, He C, Friedman S. **Elevated AGE-modified ApoB in the sera of euglycemic, normolipidemic patients with atherosclerosis**. *Mol Med* (1997.0) **3** 617-627. DOI: 10.1007/BF03401819
25. Ma K, Xu Y, Wang C. **A cross talk between class A scavenger receptor and receptor for advanced glycation end-products contributes to diabetic retinopathy**. *Am J Physiol Endocrinol Metab* (2014.0) **307** E1153-E1165. DOI: 10.1152/ajpendo.00378.2014
26. Swathi Chitra P, Chaki D, Naveen K. **Status of oxidative stress markers, advanced glycation index, and polyol pathway in age-related cataract subjects with and without diabetes**. *Exp Eye Res* (2020.0) **200** 108230. DOI: 10.1016/j.exer.2020.108230
27. Han MY, Dong HL, Yu JL. **Research progress of advanced glycation end products in glaucoma**. *International Eye Science* (2020.0) **20** 990-994. DOI: 10.3980/j.issn.1672-5123.2020.6.14
28. AnandBabu K, Sen P, Angayarkanni N. **Oxidized LDL, homocysteine, homocysteine thiolactone and advanced glycation end products act as pro-oxidant metabolites inducing cytokine release, macrophage infiltration and proangiogenic effect in ARPE-19 cells**. *PLoS ONE* (2019.0) **14** e0216899. DOI: 10.1371/journal.pone.0216899
29. Alves M, Calegari VC, Cunha DA. **Increased expression of advanced glycation end-products and their receptor, and activation of nuclear factor kappa-B in lacrimal glands of diabetic rats**. *Diabetologia* (2005.0) **48** 2675-2681. DOI: 10.1007/s00125-005-0010-9
30. Jing D, Qingjun Z, Lixin X. **The research progress of relationship between advanced glycation end products and diabetic keratopathy**. *Zhonghua Yan Ke Za Zhi* (2018.0) **54** 475-480. DOI: 10.3760/cma.j.issn.0412-4081.2018.06.018
31. Shi L, Chen H, Xiaoming Yu. **Advanced glycation end products delay corneal epithelial wound healing through reactive oxygen species generation**. *Mol Cell Biochem* (2013.0) **383** 253-259. DOI: 10.1007/s11010-013-1773-9
32. Ramya R, Coral K, Bharathidevi SR. **RAGE silencing deters CML-AGE induced inflammation and TLR4 expression in endothelial cells**. *Exp Eye Res* (2021.0) **206** 108519. DOI: 10.1016/j.exer.2021.108519
33. Midaoui AE, Elimadi A, Wu L. **Lipoic acid prevents hypertension, hyperglycemia, and the increase in heart mitochondrial superoxide production**. *Am J Hypertens* (2003.0) **16** 173-179. DOI: 10.1016/S0895-7061(02)03253-3
34. Thirunavukkarasu V, Nandhini AT, Anuradha CV. **Fructose diet-induced skin collagen abnormalities are prevented by lipoic acid**. *Exp Diabesity Res* (2004.0) **5** 237-244. DOI: 10.1080/154386090506148
35. Ghelani H, Razmovski-Naumovski V, Pragada RR. **(R)-α-Lipoic acid inhibits fructose-induced myoglobin fructation and the formation of advanced glycation end products (AGEs) in vitro**. *BMC Complement Altern Med* (2018.0) **18** 13. DOI: 10.1186/s12906-017-2076-6
36. Shay KP, Moreau RF, Smith EJ. **Alpha-lipoic acid as a dietary supplement: molecular mechanisms and therapeutic potential**. *Biochim Biophys Acta* (2009.0) **1790** 1149-1160. DOI: 10.1016/j.bbagen.2009.07.026
37. Kowluru RA. **Effect of advanced glycation end products on accelerated apoptosis of retinal capillary cells under in vitro conditions**. *Life Sci* (2005.0) **76** 1051-1060. DOI: 10.1016/j.lfs.2004.10.017
38. Kunt T, Forst T, Wilhelm A. **Alphalipoic acid reduces expression of vascular cell adhesion molecule-1 and endothelial adhesion of human monocytes after stimulation with advanced glycation end products**. *Clin Sci (Lond)* (1999.0) **96** 75-82. DOI: 10.1042/CS19980224
39. Wong A, Dukic-Stefanovic S, Gasic-Milenkovic J. **Anti-inflammatory antioxidants attenuate the expression of inducible nitric oxide synthase mediated by advanced glycation endproducts in murine microglia**. *Eur J Neurosci* (2001.0) **14** 1961-1967. DOI: 10.1046/j.0953-816x.2001.01820.x
40. Chen C-L, Cheng W-S, Chen J-L. **Potential of nonoral a-Lipoic acid aqueous formulations to reduce ocular microvascular complications in a streptozotocin-induced diabetic rat model**. *J Ocul Pharmacol Ther* (2013.0). DOI: 10.1089/jop.2012.0147
41. Kojima M, Sun L, Hata I. **Efficacy of α-Lipoic acid against diabetic cataract in rat**. *Jpn J Ophthalmol* (2007.0) **51** 10-13. DOI: 10.1007/s10384-006-0384-3
42. Inman DM, Lambert WS, Calkins DJ. **A-Lipoic acid antioxidant treatment limits glaucoma-related retinal ganglion cell death and dysfunction**. *PLoS ONE* (2013.0) **8** e65389. DOI: 10.1371/journal.pone.0065389
43. Alvarez-Rivera F, Fernández-Villanueva D, Concheiro A. **α-Lipoic Acid in Soluplus(®) polymeric nanomicelles for ocular treatment of diabetes-associated corneal diseases**. *J Pharm Sci* (2016.0) **105** 2855-2863. DOI: 10.1016/j.xphs.2016.03.006
44. Rochette L. **Direct and indirect antioxidant properties of -lipoic acid and therapeutic potential**. *Mol Nutr Food Res* (2013.0) **57** 114-125. DOI: 10.1002/mnfr.201200608
45. 45.Konrad D, Somwar R, Sweeney G et al (2001) The antihyperglycemic drug alpha-lipoic acid stimulates glucose uptake via both GLUT4 translocation and GLUT4 activation: potential role of p38 mitogen-activated protein kinase in GLUT4 activation. Diabetes 0:1464–1471
46. Fayez AM, Zakaria S, Moustafa D. **Alpha lipoic acid exerts antioxidant effect via Nrf2/HO-1 pathway activation and suppresses hepatic stellate cells activation induced by methotrexate in rats**. *Biomed Pharmacother* (2018.0) **105** 428-433. DOI: 10.1016/j.biopha.2018.05.145
47. Xing Y, Pan S, Zhu L. **Advanced glycation end products induce atherosclerosis via RAGE/TLR4 signaling mediated-M1 macrophage polarization-dependent vascular smooth muscle cell phenotypic conversion**. *Oxid Med Cell Longev* (2022.0) **2022** 9763377. DOI: 10.1155/2022/9763377.PMID:35069982;PMCID:PMC8776434
48. Behl Tapan, Sharma Eshita, Sehgal Aayush. **Expatiating the molecular approaches of HMGB1 in diabetes mellitus: highlighting signalling pathways via RAGE and TLRs**. *Molecular Biology Reports* (2021.0) **48** 1869-1881. DOI: 10.1007/s11033-020-06130-x
49. Zhao Liding, Li Ya, Tian Xu. **Dendritic cell-mediated chronic low-grade inflammation is regulated by the RAGE-TLR4-PKCβ signaling pathway in diabetic atherosclerosis**. *Molecular Medicine* (2022.0) **28** 4. DOI: 10.1186/s10020-022-00431-6
50. Chen L, Hu W, Tan S. **Genome-wide identification and analysis of MAPK and MAPKK gene families in Brachypodium distachyon**. *PLoS ONE* (2012.0) **7** e46744. DOI: 10.1371/journal.pone.0046744
51. Shi L, Yu X, Yang H. **Advanced glycation end products induce human corneal epithelial cells apoptosis through generation of reactive oxygen species and activation of JNK and p38 MAPK pathways**. *PLoS ONE* (2013.0) **8** e66781. DOI: 10.1371/journal.pone.0066781
52. Lili Hu, Yang H, Ai M. **Inhibition of TLR4 alleviates the inflammation and apoptosis of retinal ganglion cells in high glucose**. *Graefes Arch Clin Exp Ophthalmol* (2017.0) **255** 2199-2210. DOI: 10.1007/s00417-017-3772-0
53. Garibotto G, Carta A, Picciotto D, Viazzi F, Verzola D. **Toll-like receptor-4 signaling mediates inflammation and tissue injury in diabetic nephropathy**. *J Nephrol* (2017.0) **30** 719-727. DOI: 10.1007/s40620-017-0432-8
54. Wan L, Bai X, Zhou Q. **The advanced glycation end-products (AGEs)/ROS/NLRP3 inflammasome axis contributes to delayed diabetic corneal wound healing and nerve regeneration**. *Int J Biol Sci* (2022.0) **18** 809-825. DOI: 10.7150/ijbs.63219
|
---
title: Socioeconomic and Medical Vulnerabilities Among Syrian Refugees with Non-communicable
Diseases Attending Médecins Sans Frontières Services in Irbid, Jordan
authors:
- Antonio Isidro Carrion-Martin
- Ahmad Alrawashdeh
- Georgios Karapanagos
- Refqi Mahmoud
- Nashaat Ta’anii
- Mais Hawari
- Stefanie Dittmann
- Luna Hammad
- Geertje Huisman
- Mark Sherlock
- Amulya Reddy
journal: Journal of Immigrant and Minority Health
year: 2022
pmcid: PMC9988815
doi: 10.1007/s10903-022-01408-7
license: CC BY 4.0
---
# Socioeconomic and Medical Vulnerabilities Among Syrian Refugees with Non-communicable Diseases Attending Médecins Sans Frontières Services in Irbid, Jordan
## Abstract
Non-communicable diseases (NCDs) are high-prevalence health problems among Syrian refugees. In 2014, Médecins Sans Frontières (MSF) identified unmet NCD care needs and began providing free-of-charge services for Syrian refugees in Irbid, Jordan. This study aimed to describe current socioeconomic and medical vulnerabilities among MSF Irbid Syrian refugee patients and their households and raise awareness of their ongoing health needs that must be addressed. A cross-sectional survey among Syrian refugees attending MSF NCD services in Irbid Governorate, Jordan was conducted by telephone interviews in January 2021 to query sociodemographic characteristics, economic situation, self-reported NCD prevalence, and Ministry of Health (MoH) policy awareness. Descriptive analysis of indicators included proportions or means presented with $95\%$ confidence intervals. The survey included 350 patient-participants in 350 households and 2157 household members. Mean age was 28.3 years. Only $13.5\%$ of household members had paid or self-employed work; $44\%$ of households had no working members. Mean monthly income was 258.3 JOD ($95\%$CI: 243.5–273.1) per household. Mean expenditures were 320.0 JOD ($95\%$CI: 305.1–334.9). Debt was reported by $93\%$ of households. NCD prevalence among adults was $42\%$ ($95\%$CI: 40–45). Hypertension was most prevalent ($31.1\%$, $95\%$CI: 28.7–33.7), followed by diabetes ($21.8\%$, $95\%$CI: 19.7–24.1) and cardiovascular diseases ($14.4\%$, $95\%$CI: 12.6–16.4). Only $23\%$ of interviewees were aware of subsidized MoH rates for NCD care. Twenty-nine percent stated they will not seek MoH care, mainly due to the unaffordable price. Our findings highlight increased vulnerability among MSF Irbid Syrian refugee NCD patients and their households, including: an older population; a high percentage of unemployment and reliance on cash assistance; higher proportion of households in debt and a high number of households having to resort to extreme coping mechanisms when facing a health emergency; and a higher proportion of people with multiple comorbid NCDs and physical disability. Their awareness of subsidised MoH care was low. MoH care is expected to be unaffordable for many. These people are at increased risk of morbidity and mortality. It is vital that health actors providing care for Syrian refugees take action to reduce their risk, including implementing financial support mechanisms and free healthcare.
## Background
Since its start in 2011, the Syrian civil war has forced over 5.6 million refugees into neighbouring countries [1], with $95\%$ not living in camps but in urban, peri-urban and rural areas. Jordan has received $11.9\%$ of these refugees. Among Syrian refugees in Jordan, 2018 prevalence of chronic diseases was reported as $29\%$ [2] and a 2019 systematic review noted non-communicable diseases to be the most common health problems [3]. A 2019 Jordan National Stepwise Survey showed comparable hypertension prevalence of $22\%$ and diabetes $20\%$ among Jordanians and Syrians [4].
The Jordan Ministry of Health (MoH) provided Syrian refugees with free healthcare until 2014, when it implemented subsidized uninsured Jordanian rates. In late 2014, Médecins Sans Frontières (MSF) began providing free-of-charge non-communicable diseases (NCD) care for Syrian refugees living in Irbid Governorate, which registers 136,498 Syrians living in non-camp circumstances [1]. MSF’s primary care level clinic received Syrian refugees and vulnerable Jordanians with diabetes, hypertension, cardiovascular diseases, and chronic respiratory diseases (asthma, chronic obstructive pulmonary disease (COPD)) in a model of care that also provided ancillary laboratory, health education, physiotherapy and mental health support services.
In 2019, MSF began planning handover of its Jordan programme, aiming to transfer patients by the end of 2021 to MoH and other partners. A population-based survey in Irbid Governorate was planned for 2020 since a 2017 MSF survey had shown $22\%$ NCD prevalence, with $44\%$ multi-morbidity among NCD patients, $23\%$ of whom did not seek care due to unaffordability [5].
Due to the COVID pandemic, however, this changed to a telephone survey in 2021. The survey aimed to describe current socioeconomic and medical vulnerabilities among MSF Irbid Syrian refugee NCD patients and their households and raise awareness of their ongoing healthcare needs that must be addressed.
## Methods
This was a cross-sectional survey among Syrian refugees attending MSF NCD services in Irbid Governorate, Jordan, using telephone interviews and electronic data collection tools. Telephones are in wide use and incoming calls do not incur charges in Jordan. Telephone interviews were conducted between 14 January and 21 January 2021. There were 3581 patients in the cohort, 328 of whom receive home-based care given multiple morbidities and poor mobility.
## Inclusion Criteria and Definitions
A person was included in the study if they satisfied all the following criteria: (i) NCD patient in the MSF cohort; (ii) Syrian refugee arrived in Jordan in 2012 or after and (iii) resident in Irbid for more than six months prior to the interview.
A household is a person or group of people who live in the same housing unit and share living arrangements (e.g., meals, resources). A permanent household member is a person that resides regularly in the household and is mainly dependent on it. NCDs queried included diabetes, hypertension, cardiovascular diseases, chronic respiratory diseases, thyroid disease and/or cancer. The medical severity score (MSS) classifies patients by clinical criteria for the queried NCDs. The MSS ranges between zero and three, with zero indicating stable/controlled medical status and scores between one and three indicating unstable/uncontrolled medical status (MSS = 3 reflects home visit service eligibility). Physical disability indicates need for assistance during routine daily activities.
The distribution of MSF Irbid cohort NCD patients by medical severity score at the time of data collection is shown in Table 1.Table 1Distribution of NCD patients by Medical Severity ScoreMedical severity scoren%0174653.9212959.11291928.3832788.59Total3238100.00
## Sampling and Sample Size
We used simple random sampling of MSF Irbid Syrian NCD patients, who were then divided into two groups: (i) stable (MSS = zero) and (ii) unstable (MSS = one to three). Sampling was done by Irbid clinic staff.
Sample size was calculated to estimate the socioeconomic indicators (at the household level) and knowledge of recent MoH policy changes (among MSF Syrian NCD patients). We used a conservative approach, considering $50\%$ anticipated prevalence and $5\%$ precision, and arrived at a sample size of 342 patients-participants and their households. We inflated the random selection for a high non-response rate of $40\%$, which resulted in 484 patients-participants and their households.
## Data Collection and Analysis
The household interviews followed a 4-part questionnaire: (i) sociodemographic characteristics (all household members)—age, sex, educational level, employment, UNHCR and Ministry of Interior (MOI) registration; (ii) economic situation (household level, in the last month)—income, non-salary income, non-monetary assistance, household expenditure, debt, resort to financial adaptative mechanisms when encountering an urgent health problem; (iii) self-reported NCD prevalence (all household members)—hypertension, diabetes, cardiovascular disease, chronic respiratory disease, thyroid disease and cancer; and (iv) MoH policy awareness (for MSF NCD patients)—knowledge of policy changes and willingness to seek MoH care after MSF closes. The questionnaire included items not discussed in this paper.
The survey questionnaire was designed in English, translated to Arabic and back translated to English for verification. Interviewers were medical students who received a 2-day training on the study, questionnaire and daily call sheet. Interviewers were trained on potential bias, respectful behaviour, communication skills, and how to start a telephone interview, explaining the purpose of their call [6]. The questionnaire was piloted (20 interviews) and some small changes were subsequently made. Data were entered using a mobile data collection system with the open-source toolbox KOBO (https://www.kobotoolbox.org/).
All data collected were anonymized (no names, exact location or telephone numbers were collected); electronic files were stored password-protected by MSF. Data cleaning and analysis were conducted using STATA 16 (StataCorp, College Station, TX, USA). Descriptive analysis of indicators included proportions or means, which are presented with $95\%$ confidence intervals ($95\%$CI). Differences in proportions were measured using Pearson χ2 test; differences in means were measured using Student t test for parametric variables and Kruskal-Wallis quality-of-populations rank test for non-parametric variables (p-values are presented).
## Results
Of 484 randomly selected NCD patients, 390 were reached by the clinical team ($$n = 94$$, $19\%$ of participants were not reached on the phone). Fifteen ($3\%$) subsequently did not consent. Among the 375 who accepted, 9 were not reached by the survey team and 13 did not meet inclusion criteria. Data were lost for 3 patients (data tool errors). The survey therefore included 350 patient-participants in 350 households and 2157 household members, with characteristics as shown in Table 2.Table 2Characteristics of Syrian participants and household members. MSF patients $$n = 350$$Entire sample $$n = 2157$$n (%)$95\%$ CIn (%)$95\%$ CIAge ≤174 (1.1)0.4–3.0842 (39.0)35.1–42.3 18–3916 (4.6)2.8–7.3673 (31.2)29.9–33.0 40–59178 (50.9)45.6–56.1388 (18.0)16.1–19.8 ≥60152 (43.4)38.3–48.7254 (12.0)10.1–13.7Gender Male152 (43.4)38.3–48.71029 (47.7)45.2–49.5 Female198 (56.6)51.3–61.71128 ($52.3\%$)(50.2–54.4).Education No formal education76 (21.7)17.7–26.48.37.2–9.6 Primary209 (59.7)54.5–64.71352 (62.7)61.3–64.1 Secondary/post-secondary65 (18.6)14.8–23.0121 (9.2)7.8–10.9Work status Not working316 (90.3)86.7–93.01877 (87.0)86.7–88.6 Working (paid or self-employed)34 (9.7)7.0–13.3280 (13.0)11.5–14.7Legal status UNHCR337 (96.3)93.7–97.82071 (96.0)95.1–96.8 MOI card343 [98]95.9–99.02026 (93.9)92.8–94.9 UNCHR & MOI card333 (95.1)92.3–97.01993 (92.4)91.2–93.4 Prevalence of one or more NCDs350 [100]–601 (27.9)26.0–29.8NCD prevalence Diabetes210 (60.0)54.8–65.0184 (25.9)22.8–29.2 HTN278 (79.4)74.9–83.4238 (33.5)30.1–37.0 CVD127 (36.3)31.4–41.599 (13.9)11.6–16.7 Respiratory41 (11.7)8.7–15.5104 (4.8)4.0.–5.8 Thyroid31 (8.9)6.3–12.363 (2.9)2.3–3.7 Cancer5 (1.4)0.6–3.416 (0.7)0.5–1.2
## Demographic and Economic Characteristics
Mean household member age was 28 years (range: 1–90, standard deviation (SD): 21.3). Females constituted $52\%$ of individuals ($\frac{1128}{2157}$, $95\%$CI: 50–54). The proportion of adults (≥18 years) was $61\%$ ($\frac{1315}{2157}$, $95\%$CI: 59–63) and those aged >65 years $7.5\%$ ($\frac{162}{2157}$, $95\%$CI: 6.5–8.7). Mean household size was 6.2 members (range: 1–20, SD: 3.0). A total of 25 ($7.2\%$) households arrived before 2012 and only 24 ($6.9\%$) households arrived between 2014 and 2018. Households have been in Jordan for a mean 8.4 years ($95\%$ CI, 8.3–8.5).
Only 280 of 2157 ($13\%$, $95\%$CI: 12–14) individuals were in paid work ($$n = 273$$) or self-employed ($$n = 7$$) (Table 3). Forty-four percent of households ($\frac{153}{350}$, $95\%$CI: 39–49) had no working members; $39\%$ ($\frac{137}{350}$, $95\%$CI: 34–44) had only one working member. Table 3Work status of adults (18–65 years)*Work status* (adults 18–65 years)n%$95\%$ CIWorking for wages or salary26523.020.6–25.5Self-employed70.60.3–1.3Available and actively looking for work15613.511.7–15.6In school/training867.56.1–9.1Doing home duties39634.431.7–37.1Not working due to chronic health condition22719.717.5–22.1Retired70.60.3–1.3Don’t know90.80.4–1.5 Most individuals were registered with UNHCR ($$n = 2071$$, $96\%$, $95\%$CI: 95–97) and had an MOI card ($$n = 2026$$, $94\%$, $95\%$CI: 93–95). Full legal status (UNHCR and MOI) was reported for $92\%$ ($$n = 1993$$, $95\%$CI: 91–93). Main reasons for not registering with UNHCR were ‘not needed’, ‘not applicable/other nationality’ and ‘UNHCR is closed’; ‘lack of time’ was the main reason for lack of MOI registration.
Mean monthly income was 258 JOD (range: 0–2004, SD: 169) per household and 48 JOD (range-0–501, SD: 39) for each member, while mean expenditures were 320 JOD (range: 305–335, SD: 8) and 62 JOD (range: 0–300, SD: 2) respectively. Debt was reported by $93\%$ of households ($\frac{324}{350}$, $95\%$CI: 89–95). Main sources of household income were UNHCR cash assistance ($\frac{184}{350}$, $53\%$, $95\%$CI: 47–58) and work ($\frac{149}{350}$, $43\%$, $95\%$CI: 37–48).
Seventy percent ($\frac{243}{350}$, $95\%$CI: 64–74) of households reported that they resorted to financial adaptive approaches for urgent health problems as detailed in Table 4.Table 4Adaptive responses to urgent health problems in families of Syrian refugees attending Médecins Sans Frontières NCD services in Irbid Governorate, Jordan, 2021.Responsesn%$95\%$CIBorrow money from friend/relative2096054–65Reduce standard household spending2075954–64Turn down work or missing workdays1073126–36Sell household assets631814–23Borrow money from usurer123.42.0–6.0 The proportion of adults ($61\%$) is higher than the UNHCR estimate of $51\%$ for Syrian refugees and asylum seekers registered in Jordan, the $46\%$ found by the 2017 MSF survey of Syrian refugees residing in Irbid and the $56\%$ estimate for the Jordanian population (19 years and older) [5, 7, 8]. Mean age in this study was higher than that found in the 2017 MSF survey (28.5 vs. 21.3 respectively). These differences in age, as well as other differences presented in this discussion, are likely to be explained by the fact that our study included only households with Syrian refugees residing in Irbid Governorate and attending MSF NCD services. Since NCD prevalence increases with age this would explain why this population is older. As a consequence of the increased age and NCD prevalence, increased healthcare needs and expenditures are expected.
The average household size of 6.2 is higher than the 5.9 average for Syrian refugees living in Jordan found by the Vulnerability Assessment Framework of Syrian Refugees in Jordan [9] and the 4.7 average household size in the Jordanian population [8], but lower than the 6.9 in the 2017 MSF survey [5].
Only $24\%$ of adults were in paid work or self-employed. At household level, a high proportion ($44\%$) had no working members. This is comparable to the findings of the Syrian refugees’ living conditions survey where just over half of households relied on work income [10]. For the $7.5\%$ were in school/training, this likely reflects additional expense that should be queried further with respect to later employment/rewards.
The proportion of individuals registered with UNCHR in our study was similar to the 2017 MSF survey and the $\frac{2017}{18}$ living conditions survey ($96\%$, $95\%$ and $97\%$ respectively), while the proportion of MOI registration was higher ($94\%$ vs. $81\%$ and $86\%$ respectively) [5, 10]. This may indicate that registration status of MSF NCD patients has improved in recent years.
The mean household income was similar to the 2017 MSF survey and the $\frac{2017}{18}$ living conditions survey (258 JOD, 239 JOD and 260 JOD respectively) [5, 10]; however, it was lower than the 376 JOD reported by the World Food Programme (WFP) 2018 assessment of Syrian refugees living in Jordan [11], and much lower than the 2017 national average reported by the Jordan Department of Statistics (889 JOD) [8]. Per capita income, given the average household size of 6.2 in our survey, is 42 JOD, which is lower than that of vulnerable Jordanians supported by the National Aid Fund (57 JOD) reported by WFP, and $38\%$ lower than the existing 68 JOD poverty line [11].
Mean household expenditure was lower than that reported by the Jordan Department of Statistics for Jordanian national households (320 JOD vs 627 JOD) [8], and lower than in the earlier surveys (320 JOD vs. 359 JOD and 429 JOD respectively) [5, 10]. Although this indicates that mean debt was lower in our study, we found a higher proportion of households in debt compared with previous surveys, as $93\%$ of households reported being in debt compared with the $79\%$ in the 2017 MSF survey and $67\%$ in the $\frac{2017}{18}$ living conditions survey [5, 10].
The main source of household income in our study population was cash assistance ($53\%$), a much higher proportion than the $31\%$ in the 2017 MSF survey [12] and higher than the $39\%$ reported for Syrian refugees in the 2018 WFP assessment [11]. In our survey, the second source of household income was work, reported by $43\%$ of households; this was similar to the $45\%$ reported by the 2017 MSF survey and lower than the $53\%$ reported for Syrian refugees by the 2018 WFP assessment but higher than the $26\%$ reported for vulnerable Jordanians [5, 11].
Almost $70\%$ of households reported having to respond to an urgent health problem in the previous six months. Although we have not found other reports containing adequate indicators to compare this, the fact that a high proportion of households reported having to borrow money ($60\%$), reduce standard household living spending ($59\%$) or even sell household assets ($18\%$) reveals their extreme financial vulnerability.
The proportion of people with physical disability ($12\%$) was threefold higher than $3.9\%$ found in the 2017 MSF survey [12]. It was even higher when compared to the $1.6\%$ of people with a handicap/functional difficulty found by the $\frac{2017}{18}$ living conditions survey, albeit using a different definition [10].
## Non-communicable Diseases and Physical Disability
Twenty-eight percent of household members ($\frac{601}{2157}$, $95\%$CI: 26–30) reported having one or more NCDs, including diabetes, hypertension, cardiovascular diseases, chronic respiratory disease, thyroid disease, and/or cancer. Adult (≥18 years) NCD prevalence was $42\%$ ($\frac{556}{1.315}$, $95\%$CI: 40–45). Women had higher prevalence than men ($46\%$, $95\%$CI: 42–50 and $38\%$, $95\%$CI: 34–42, $$p \leq 0.003$$).
Among adults, hypertension was most prevalent ($31\%$, $95\%$CI: 29–34), followed by diabetes ($22\%$, $95\%$CI: 20–24) and cardiovascular disease ($14\%$, $95\%$CI: 13–16). Among minors, chronic respiratory disease was most prevalent NCD ($3.1\%$, $95\%$CI: 2.1–4.5) (Table 5). The prevalence of most NCDs increased with age; this was most strongly observed for diabetes, hypertension and cardiovascular conditions. The prevalence of most NCDs was higher in women than in men (Table 5).Table 5NCD prevalence among Syrian refugees attending Médecins Sans Frontières NCD services in Irbid Governorate, Jordan, 2021Entire population ($$n = 2157$$)Minors (<18 years) ($$n = 842$$)Adults (≥18 years) ($$n = 1315$$)Women ($$n = 711$$)Men ($$n = 604$$)n (%)$95\%$CIn (%)$95\%$CIn (%)$95\%$CIn (%)$95\%$CIn (%)$95\%$CIDiabetes296 [14]12–159 (1.1)0.6–2.0287 [22]20–24184 [26]23–29103 [17]14–20Hypertension410 [19]17–211 (0.1)0.0–0.8409 (31.1)29–34238 (33.5)30–37171 [28]25–32Cardiovascular diseases198 (9.2)8.0–109 (1.1)0.6–2.0189 [14]13–1699 (13.9)12–1790 [15]12–18Respiratory diseases104 (4.8)4.0–5.826 (3.1)2.1–4.578 (5.9)4.8–7.348 (6.8)5.1–8.830 (5.0)3.5–7.0Thyroid diseases63 (2.9)2.3–3.71 (0.1)0.0–0.862 (4.7)3.7–6.052 (7.3)5.6–9.510 (1.7)0.9–3.1Cancer16 (0.7)0.5–1.22 (0.2)0.1–0.914 (1.1)0.6–1.87 (1.0)0.5–2.17 (1.2)0.6–2.4 Individuals with more than one NCD constituted $15\%$ ($95\%$CI: 14–17). Among adults, $59\%$ ($95\%$CI: 55–63) reported more than one NCD. Given the design, all households had at least one NCD patient; over half had more than one ($53\%$, $95\%$CI: 48–59).
Twelve percent ($\frac{251}{2157}$, $95\%$CI: 10–3) of individuals had a physical disability requiring medical support.
## Awareness of Subsidized MoH Rates
Only $23\%$ of interviewees ($\frac{82}{350}$, $95\%$CI: 19–28) were aware of the subsidized MoH rate for NCD patients and $7.1\%$ ($\frac{25}{350}$, $95\%$CI: 4.9–10) knew that it is guaranteed until 2023.
Twenty-nine percent ($\frac{100}{250}$, $95\%$CI: 24–34) stated they will not seek MoH care after MSF closes, mainly due to the unaffordable price even with the subsidised rate ($61\%$, $\frac{61}{100}$, $95\%$CI: 51–70).
## Discussion
Our findings highlight increased socioeconomic and medical vulnerability among MSF Irbid Syrian refugee NCD patients and their households when compared to earlier studies.
## Non-communicable Diseases
Comparison of self-reported NCD prevalence results with other (population-based) studies is limited by our redesign to include only MSF Irbid clinic patients and their households. However, the finding of $53\%$ of households with more than one NCD patient was similar to a 2019 scoping review on the burden of non-communicable diseases among Syrian refugees [13].
The percentage of adults reporting at least one NCD under study ($42\%$) was almost double the $22\%$ in the 2017 MSF survey [5] and even higher than the $16\%$ chronic health failure among Syrian refugees in Jordan reported in the $\frac{2017}{18}$ living conditions survey [10]. Sixty-five percent of our patients had more than one NCD, which highlights the medical complexity and vulnerability of these people.
We found the most prevalent NCDs were hypertension and diabetes, aligning with findings of the 2017 MSF survey [5], the 2019 scoping review [13] and a recent study by Ratnayake, et al. [ 14]. We found higher prevalence of self-reported hypertension and diabetes ($31\%$ and $22\%$ respectively) compared to the 2017 MSF survey (which found $14\%$ and $9\%$ respectively) [5] and the Ratnayake, et al. study (which found $17\%$ and $10\%$) [14]. This is expected given our sampling strategy. The self-reported results are likely to be an underestimation of true biologically-based disease burden. The Ratnayake et al. study measured biologically-based prevalence to be approximately double the self-reported prevalence. Except for cardiovascular disease, NCD prevalence was higher among females, aligning with the 2017 MSF survey.
## Awareness of MoH Subsidized Healthcare
Awareness of the MoH subsidised rate for NCD care was low, with only $23\%$ of interviewees being aware and only $7\%$ being aware that it is guaranteed until 2023. This may indicate lack of access to appropriate and relevant information concerning health care rights or lack of information seeking in the MSF NCD cohort. Almost $30\%$ of interviewees reported they would not seek MoH care after MSF closes; almost two thirds of these people stated the MoH subsidized price would still be unaffordable, highlighting their strained financial situation.
## Limitations
Our study included only households with Syrian refugees residing in Irbid Governorate and attending MSF NCD services, limiting generalizability to the full Irbid Syrian refugee population. Nonetheless, comparison of our results with those from previous surveys and available Syrian refugee and vulnerable Jordanian indicators aid understanding of this cohort’s characteristics and vulnerabilities.
All information collected was self-reported; we did not ask for proof. To decrease as much as possible the risk of lack of accuracy on the information provided by one person, the interviewers offered the interviewee the option of delegating or asking another household member to be present to assist specific responses if they were not sure about the answers. The interviewers were Jordanians who were trained on communication skills to make respondents feel comfortable. However, Syrian respondents may have perceived a difference in origin and dialect, which could have had an impact and potential bias in their responses.
Some interviewees may not have reported true household income and expenditure figures due to inaccurate calculations or preference for under- or overestimation. Debt accuracy and exchange mechanisms were not assessed in the data collection. Interviewees could have underreported individuals not registered with UNHCR or the MOI fearing legal repercussions. The research team was aware of these potential limitations and interviewers were trained to clearly state there were no incentives for survey participation and that all information remains anonymous.
## Conclusions
This study highlights increased vulnerability of MSF NCD patients in relation to the general Syrian refugee population residing in Jordan: this is an older population with high unemployment and reliance on cash assistance, high household debt and use of extreme coping mechanisms when facing a health emergency, and high multiple NCD comorbidity and physical disability. The continuous care that NCDs require creates an ongoing cost burden in households where expenditures already exceed income, risking further debt in households already living below the poverty line in Jordan.
Awareness of the MoH subsidised rate was low. This should be further explored as it may indicate lack of access to appropriate and relevant information concerning healthcare rights, lack of information seeking and/or other issues.
Almost one third of interviewees reported that if the free-of-charge MSF option is not available, they would not seek MoH care, most commonly stating it is unaffordable. These people are at increased risk of morbidity and mortality. It is vital that health actors providing care for Syrian refugees take action to reduce their risk, including making financial support mechanisms such as cash for health widely available and offering free healthcare.
## References
1. 1.UNHCR. UNCHR Operational Portal. Syrian regional Refugee response: Total Persons of Concern. [Internet]. 2021 https://data2.unhcr.org/en/situations/syria Accessed 29 April 2021
2. Dator W, Abunab H, Dao ayen N. **Health challenges and access to health care among Syrian refugees in Jordan: a review**. *East Mediter Health J* (2018.0) **24** 680-6. DOI: 10.26719/2018.24.7.680
3. El Arnaout N, Rutherford S, Zreik T, Nabulsi D, Yassin N, Saleh S. **Assessment of the health needs of Syrian refugees in Lebanon and Syria’s neighboring countries**. *Confl Health* (2019.0) **13** 31. DOI: 10.1186/s13031-019-0211-3
4. 4.Jordan Ministry of Health. Jordan National Stepwise Survey (STEPs) for Noncommunicable Diseases Risk Factors 2019. [Internet]. 2019 http://www.emro.who.int/jor/jordan-news/results-of-jordan-national-stepwise-survey-steps-of-noncommunicable-diseases-and-their-risk-factors-2019.html#:~:text=30%20September%202020%2C%20Amman%20%E2%80%93%20Noncommunicable,diabetes%20and%20chronic%20respiratory%20diseases. Accessed 1 Aug 2022
5. Rehr M, Shoaib M, Ellithy S, Okour S, Ariti C, Ait-Bouziad I. **Prevalence of non-communicable diseases and access to care among non-camp Syrian refugees in northern Jordan**. *Conflict Health* (2018.0). DOI: 10.1186/s13031-018-0168-7
6. 6.World Health Organization, Council for International Organizations of Medical SciencesInternational ethical guidelines for health-related research involving humans2017GenevaCIOMS. *International ethical guidelines for health-related research involving humans* (2017.0)
7. 7.UNHCR. Registered Persons of Concern Refugees and Asylum Seekers in Jordan- Syria [Internet]. 2021 Mar. https://data2.unhcr.org/en/documents/details/85499
8. 8.Department of Statistics, Jordan. Department of Statistics, Jordan. [Internet] http://dosweb.dos.gov.jo/population/population-2/. Accessed 25 March 2021
9. 9.UNHCR, ACF, ILO. Vulnerability Assessment Framework of regsitered Syrian refugees living in Jordan, 2019 [Internet]. Jordan; 2019. https://data2.unhcr.org/en/documents/details/68856. Accessed 25 March 2021
10. 10.Åge A. Tiltnes, Huafeng Zhang, Jon Pedersen. The living conditions of Syrian refugees in Jordan Results from the 2017–2018 survey of Syrian refugees inside and outside camps [Internet]. 2019. https://reliefweb.int/sites/reliefweb.int/files/resources/67914.pdf
11. 11.World Food Programme, REACH. Jordan – Comprehensive Food Security and Vulnerability Assessment, 2018 [Internet]. World Food Programme; 2018 https://reliefweb.int/report/jordan/jordan-comprehensive-food-security-and-vulnerability-assessment-2018-april-2019. Accessed 29 March 2021
12. 12.Médecins Sans Frontières. Health Service Access Survey among Non-camp Syrian Refugees in Irbid Governorate, Jordan (full report). [Internet]. 2017. https://fieldresearch.msf.org/handle/10144/619235. Accessed 25 March 2021
13. Naja F, Shatila H, El Koussa M, Meho L, Ghandour L, Saleh S. **Burden of non-communicable diseases among Syrian refugees: a scoping review**. *BMC Public Health* (2019.0). DOI: 10.1186/s12889-019-6977-9
14. Ratnayake R, Rawashdeh F, AbuAlRub R, Al-Ali N, Fawad M, Bani Hani M. **Access to care and prevalence of hypertension and diabetes among Syrian Refugees in Northern Jordan**. *JAMA Network Open* (2020.0) **3** e2021678. DOI: 10.1001/jamanetworkopen.2020.21678
15. 15.Council for International Organizations of Medical Sciences. International Guidelines for Ethical Review of Epidemiological Studies [Internet]. Council for International Organizations of Medical Sciences; 1991 https://cioms.ch/publications/product/1991-international-guidelines-for-ethical-review-of-epidemiological-studies/ Accessed 29 April 2021
|
---
title: Association between triglyceride to high-density lipoprotein cholesterol ratio
and type 2 diabetes risk in Japanese
authors:
- Huijuan Wang
- Changming Wang
- Xiuping Xuan
- Zhouni Xie
- Yuanyuan Qiu
- Huiping Qin
- Zhong Xiaoning
journal: Scientific Reports
year: 2023
pmcid: PMC9988840
doi: 10.1038/s41598-022-25585-5
license: CC BY 4.0
---
# Association between triglyceride to high-density lipoprotein cholesterol ratio and type 2 diabetes risk in Japanese
## Abstract
Abnormal lipid metabolism is known to increases the risk for metabolic diseases, such as type 2 diabetes mellitus(T2DM). The relationship between baseline ratio of triglyceride to HDL cholesterol (TG/HDL-C) and T2DM in Japanese adults was investigated in this study. Our secondary analysis included 8419 male and 7034 female Japanese subjects who were free of diabetes at baseline. The correlation between baseline TG/HDL-C and T2DM was analyzed by a proportional risk regression model, the nonlinear correlation between baseline TG/HDL-C and T2DM was analyzed by a generalized additive model (GAM), and the threshold effect analysis was performed by a segmented regression model. We conducted subgroup analyses in different populations. During the median 5.39 years follow-up, 373 participants, 286 males and 87 females, developed diabetes mellitus. After full adjustment for confounders, the baseline TG/HDL-C ratio positively correlated with the risk of diabetes (hazard ratio 1.19, $95\%$ confidence interval 1.09–1.3), and smoothed curve fitting and two-stage linear regression analysis revealed a J-shaped relationship between baseline TG/HDL-C and T2DM. The inflection point for baseline TG/HDL-C was 0.35. baseline TG/HDL-C > 0.35 was positively associated with the development of T2DM (hazard ratio 1.2, $95\%$ confidence interval 1.10–1.31). Subgroup analysis showed no significant differences in the effect between TG/HDL-C and T2DM in different populations. A J-shaped relationship was observed between baseline TG/HDL-C and T2DM risk in the Japanese population. When TG/HDL-C was higher than 0.35, there was a positive relationship between baseline TG/HDL-C and the incidence of diabetes mellitus.
## Introduction
Diabetes is a chronic disease that seriously affects human health and has a widespread global impact, it has become a growing public health problem worldwide and it has been recognized as a global public health challenge1. According to an estimate provided by the International Diabetes Federation (IDF), by 2021, 537 million adults are living with diabetes, and this number is expected to rise to 643 million by 2030 and to 783 million by 20452. In Japan, the prevalence of diabetes has been increasing dramatically since 1997, especially in males3. Many Asian countries, including China, India, Singapore, and Japan, have a considerable prevalence of diabetes, and the prevalence of diabetes has increased extremely rapidly in these regions in recent years3. The prevalence of diabetes, primarily type 2 diabetes, poses enormous social and economic problems that may hinder national and global development. Many factors such as unhealthy diet, obesity, and a sedentary lifestyle are thought to contribute to T2DM4. Given the global economic and social burden of diabetes, understanding the risk factors for diabetes that can be intervened to enhance diabetes prevention can help reduce the economic burden on countries and individuals. Active and effective prevention of diabetes mellitus can lead to early detection of diabetic patients, facilitate timely and effective treatment, reduce, and delay the occurrence and development of diabetic complications, improve the quality of life of patients, reduce the disability rate and prolong life expectancy.
The pathogenesis of diabetes mellitus is very complex, and as research continues, it is found that diabetes mellitus is the result of multiple factors and mechanisms acting together. Abnormal lipid metabolism is both an important factor in the development of diabetes mellitus and an important cause of its complications5,6. Diabetes is characterized by insulin resistance (IR), insufficient insulin secretion and increased hepatic glucose output7,8. IR is a condition in which the efficiency of insulin in promoting glucose uptake and utilization by cells is reduced for various reasons, i.e., the biological effects produced by insulin do not work properly9. The gold standard for assessing β-cell function and insulin sensitivity is the hyperinsulinemic euglycemic clamp technique10. However, this technique is more challenging to apply in clinical practice due to its inconvenience and high cost, so other simple biomarkers that reflect IR or β-cell dysfunction may be beneficial for screening for diabetic diabetes. Disorders of lipid metabolism, including elevated serum triglyceride levels and decreased serum high-density lipoprotein cholesterol, among others. Lipid metabolism disorders plays a crucial role in the pathogenesis of diabetes11. Previous studies have shown that the ratio of triglycerides to HDL cholesterol (TG/HDL-C) is strongly associated with insulin resistance, and it is expected to be a simple and easy predictor of IR12–15. However, despite the potential of the TG/HDL-C ratio as a simple and easy metric to predict diabetes risk, there are very limited reports of prospective cohort studies investigating the relationship between baseline TG/HDL-C ratio and T2DM risk. Although studies on the relationship between TG/HDL-C and T2DM have only been reported in Chinese Singaporean, Korean, Chinese, and Iranian populations13,16–19, there have been few studies on data from the Japanese population. To the best of our knowledge, one study data from Ibaraki-Prefecture, Japan, showed that the TG/HDL-C ratio was positively associated with incident diabetes20; However, their study did not clarify the linear/nonlinear effect of TG/HDL-C on diabetes risk and the appropriate cutoff value.
In our present investigation, we re-analyze the data from the previously published study by Okamura et al.21. TG/HDL-C was utilized as an independent variable in the secondary analysis, and the outcome variables and other covariates were the same as in the original study.
## Data source
Information of participants in this study was acquired from the NAGALA database. The data package for this study was collected, organized by Okamura et al. and submitted to the Dryad database for free use by the researchers. We conducted a secondary analysis of this dataset22. Variables included in this dataset: age, gender, body mass index (BMI), weight, waist circumference (WC), high-density lipoprotein cholesterol (HDL-C), γ-glutamyl transpeptidase (GGT), triglycerides (TG), total cholesterol (TC), hemoglobin A1c (HbA1c), diastolic blood pressure (DBP), systolic blood pressure (SBP), alanine aminotransferase (ALT), fasting plasma glucose (FPG), ethanol consumption, aspartate aminotransferase (AST), smoking status, alcohol consumption, follow-up time, fatty liver, diabetes mellitus, and exercise habits.
## Study participants
NAGALA is a longitudinal cohort study in the Gifu Area of Japan analyzing21. The study project collected data from participants in a health checkup program at Murakami Memorial Hospital in Japan21. In this medical checkup program, $60\%$ of the participants received one or two medical checkups per year21. Previous researchers recruited 20,944 participants from individuals who participated in medical checkup programs from 2004 to 2015. Participants with the following conditions at baseline will be excluded: Missing relevant data (including exercise, alcohol consumption, height, HDL-cholesterol, and abdominal ultrasound), viral /alcoholic hepatitis, alcohol abuse, impaired fasting blood glucose, diabetes, or Any medication used at baseline examination. The final 15,453 subjects were eligible for our study. Because previous studies were submitted to the ethics committee of Murakami Memorial Hospital for approval21, the current study is exempt from ethical review.
## Variables measurement and definitions
As previously mentioned21, the researchers used questionnaires, physical examinations, and blood tests to obtain baseline data from the participants. Subjects were categorized depending on average weekly ethanol and type of alcohol intake. Alcohol intake of less than 40 g per week is defined as no or minimal alcohol consumption23. Weekly alcohol intake of 40 g to 140 g is defined as light alcohol consumption23. Alcohol intake of 140 g to 280 g per week is defined as moderate alcohol consumption23. Weekly alcohol intake greater than 280 g is defined as heavy alcohol consumption23. Participants were divided into non-smokers, ex-smokers, and current smokers based on their smoking status at baseline. Non-smokers were defifined as participants who never smoked cigarettes, ex-smokers as participants who had smoked in the past but who quit smoking until the baseline visit, and current-smokers as participants who smoked at the baseline visit21.
Regular participation in sports > 1x/week is defined as regular exercise. 24. Body mass index is calculated as the number of kilograms of body weight divided by the square of the number of meters of height25. The ratio of TG/HDL-C was measured by dividing the fasting triglyceride level (mmol/L) into the fasting High-density lipoprotein cholesterol level (mmol/L). Gastroenterologists diagnose fatty liver by reviewing abdominal ultrasound based on four known criteria (liver brightness, liver, and kidney echo contrast, vascular blurring, and depth attenuation)26.
## Definition of T2DM
Participant self-report,FPG ≥ 7 mmol/L, or HbA1c ≥ $6.5\%$,were used to identify incident type 2 diabetes21,27.
## Statistical analysis
Frequencies or percentages was used to express categorical variables. The mean ± standard deviation was used to represent normally distributed continuous variables, and the median (P25, P75) is used to represent skewed continuous variables. One-way ANOVA, Kruskal–Wallis H-test, and chi-square test were used to compare the differences between groups. Univariate regression models were used to examine the effect of each variable on T2DM. The covariates found to be significantly different in the univariate analysis were screened for confounding factors. Multivariate Cox proportional risk models were used to examine the risk prediction of exposure variables on outcome variables, and the risk ratios (HRs) with $95\%$ confidence intervals (CIs) were estimated to assess the risk of the outcome variables. The confounding factors screened included: sex, age, alcohol consumption, exercise habits, smoking status, fatty liver, BMI, fasting plasma glucose, total cholesterol, and HbA1c. We show three models: unadjusted analysis model (model 1); partially adjusted analysis (model 2): adjusted for sex and age only; and fully adjusted analysis (model 3), adjusted for all screened confounders. Researchers used a generalized additive model (GAM, restricted cubic spline function) to examine if there was a nonlinear relationship between baseline TG/HDL-C and the risk of T2DM. Threshold effects were evaluated by smoothed curve fitting and segmented regression models. A stratified logistic regression model was used to perform subgroup analyses based ages, genders, exercise habits, and smoking status. The likelihood ratio test was used to test the interactions among subgroups. We used R version 3.4.3 and Empower (R) version 2.0 to perform statistical analysis of the study data. $P \leq 0.05$(bilateral) is the criterion for significance.
## Ethical approval
Approval of the research protocol: *The data* comes from the public database. In the previously published article21. Takuro Okamura et al. has clearly stated that: the study was approved by the ethics committee of Murakami Memorial Hospital.
## Participants' baseline characteristics
In our study, 15,453 individuals free of diabetes at baseline were included. The mean age of the subjects was 43.71 ± 8.90 years, and $45.52\%$ of the subjects were females.
The median follow-up time was 5.39 years. During this period, 373 individuals developed diabetes, 87 females and 286 males. The prevalence of diabetes was $2.4\%$. Table 1 shows the baseline characteristics of subjects grouped according to quartiles of TG/HDL-C ratio. In the group with higher TG/HDL-C, participants were older and had higher BMI, weight, waist circumference and blood pressure (SBP and DBP). In the presence of elevated TG/HDL-C quartiles, ALT, GGT, FPG and TC gradually increased. There were considerably more smokers and heavy drinkers in the group Q4 than in the other three groups (Q1-Q3). The proportion of those with fatty liver and diabetes increased with increasing TG/HDL-C.Table 1Baseline characteristics of participants by categories of the baseline TG/HDL-C in the NAGALA study, 2004–2015.VariableTG/HDL-C quartilesP-valueQ1 (0.21 ± 0.06)Q2 (0.39 ± 0.06)Q3 (0.68 ± 0.11)Q4 (1.68 ± 0.98)Participants (n)3699396739163871Age, year41.18 ± 8.3943.34 ± 8.8544.97 ± 9.0645.23 ± 8.68 < 0.001Ethanol consumption, g/week1.00 (0.00–22.00)1.00 (0.00–60.00)2.80 (0.00–84.00)12.00 (1.00–90.00) < 0.001BMI, kg/m220.26 ± 2.2821.33 ± 2.6422.54 ± 2.9324.27 ± 3.07 < 0.001WC, cm70.51 ± 6.8673.85 ± 7.9277.94 ± 8.1983.36 ± 7.96 < 0.001ALT, IU/L14.00 (11.00–17.00)15.00 (12.00–20.00)18.00 (14.00–23.00)23.00 (17.00–32.00) < 0.001AST, IU/L16.00 (13.00–20.00)17.00 (14.00–20.00)17.00 (14.00–21.00)19.00 (16.00–24.00) < 0.001GGT, IU/L12.00 (10.00–15.50)13.00 (11.00–18.00)16.00 (12.00–23.00)22.00 (16.00–33.00) < 0.001HDL-C, mmol/L1.85 ± 0.381.58 ± 0.291.35 ± 0.251.09 ± 0.21 < 0.001TC, mmol/L4.86 ± 0.805.01 ± 0.815.17 ± 0.865.45 ± 0.87 < 0.001TG, mmol/L0.38 ± 0.120.62 ± 0.130.91 ± 0.201.73 ± 0.78 < 0.001HbA1c, %5.15 ± 0.295.14 ± 0.315.18 ± 0.335.21 ± 0.34 < 0.001FPG, mmol/L4.97 ± 0.395.10 ± 0.405.22 ± 0.395.35 ± 0.37 < 0.001SBP, mmHg108.32 ± 12.99111.94 ± 14.07116.35 ± 14.70121.13 ± 14.88 < 0.001DBP, mmHg66.95 ± 9.2169.67 ± 9.8272.96 ± 10.1776.56 ± 10.25 < 0.001Gender < 0.001Female2838 ($76.72\%$)2241 ($56.49\%$)1380 ($35.24\%$)575 ($14.85\%$)Male861 ($23.28\%$)1726 ($43.51\%$)2536 ($64.76\%$)3296 ($85.15\%$)Fatty liver < 0.001No3624 ($97.97\%$)3686 ($92.92\%$)3208 ($81.92\%$)2198 ($56.78\%$)Yes75 ($2.03\%$)281 ($7.08\%$)708 ($18.08\%$)1673 ($43.22\%$)WC ≥ 90 in men, ≥ 80 in women < 0.001No3489 ($94.32\%$)3617 ($91.18\%$)3360 ($85.80\%$)2975 ($76.85\%$)Yes210 ($5.68\%$)350 ($8.82\%$)556 ($14.20\%$)896 ($23.15\%$)Habit of exercise < 0.001Yes3002 ($81.16\%$)3258 ($82.13\%$)3207 ($81.89\%$)3280 ($84.73\%$)No697 ($18.84\%$)709 ($17.87\%$)709 ($18.11\%$)591 ($15.27\%$)Alcohol consumption < 0.001Non3135 ($84.75\%$)3085 ($77.77\%$)2901 ($74.08\%$)2681 ($69.26\%$)Light300 ($8.11\%$)444 ($11.19\%$)496 ($12.67\%$)514 ($13.28\%$)Moderate211 ($5.70\%$)323 ($8.14\%$)363 ($9.27\%$)460 ($11.88\%$)Heavy53 ($1.43\%$)115 ($2.90\%$)156 ($3.98\%$)216 ($5.58\%$)*Smoking status* < 0.001Never2918 ($78.89\%$)2611 ($65.82\%$)2012 ($51.38\%$)1486 ($38.39\%$)Past452 ($12.22\%$)680 ($17.14\%$)858 ($21.91\%$)959 ($24.77\%$)Current329 ($8.89\%$)676 ($17.04\%$)1046 ($26.71\%$)1426 ($36.84\%$)Incident diabetes < 0.001No3677 ($99.41\%$)3921 ($98.84\%$)3833 ($97.88\%$)3649 ($94.27\%$)Yes22 ($0.59\%$)46 ($1.16\%$)83 ($2.12\%$)222 ($5.73\%$)Data were mean ± SD or median (P25–P75)/N (%) for skewed variables or numbers (proportions) for categorical variables. BMI body mass index, WC waist circumference, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT γ-glutamyl transferase, HDL-C high-density lipoprotein cholesterol, TC total cholesterol, TG triglyceride, HbA1c hemoglobin A1c, FPG fasting plasma glucose, SBP systolic blood pressure, DBP diastolic blood pressure.
## Univariate analysis
We performed the univariate analysis of the relationship of each variable with T2DM, and the results are shown in Table 2. Without adjusting for other variables, all of the covariables, except light and moderate ethanol consumption, were related to the occurrence of T2DM; of these, exercise habits and HDL-C were negatively related to the onset of diabetes, and the other variables were positively associated with diabetes. The study also observed that males had a greater to acquire diabetes than females, that subjects with fatty liver were at significantly higher risk of acquiring T2DM than subjects without fatty liver, and that subjects with thicker waist circumference (WC ≥ 90 in males and WC ≥ 80 in females) had a stronger chance of developing T2DM, and those current and former smokers were more likely to develop diabetes than never smokers. Heavy drinkers have a higher risk of developing T2DM than those who do no or minimal alcohol consumption. Table 2Univariate analysis for incident diabetes. CovariateStatisticsOR ($95\%$CI)P-valueGenderFemale7034 ($45.52\%$)ReferenceMale8419 ($54.48\%$)2.81 (2.20, 3.58) < 0.0001Age, year43.71 ± 8.901.04 (1.03, 1.05) < 0.0001Ethanol consumption, g/week47.71 ± 82.311.00 (1.00, 1.00)0.0001Fatty liverNo12,716 ($82.29\%$)ReferenceYes2737 ($17.71\%$)7.43 (6.02, 9.18) < 0.0001BMI, kg/m222.12 ± 3.131.26 (1.23, 1.29) < 0.0001WC, cm76.47 ± 9.111.10 (1.09, 1.11) < 0.0001WC ≥ 90 in men, ≥ 80 in womenNo13,441 ($86.98\%$)ReferenceYes2012 ($13.02\%$)4.09 (3.29, 5.07) < 0.0001Baseline BMI ≥ 25No12,932 ($83.69\%$)ReferenceYes2521 ($16.31\%$)4.64 (3.77, 5.71) < 0.0001ALT, IU/L19.99 ± 14.351.03 (1.02, 1.03) < 0.0001AST, IU/L18.40 ± 8.641.03 (1.02, 1.04) < 0.0001Body weight, kg60.63 ± 11.621.06 (1.05, 1.07) < 0.0001Habit of exerciseNo12,747 ($82.49\%$)ReferenceYes2706 ($17.51\%$)0.74 (0.55, 1.00)0.0491GGT, IU/L20.31 ± 18.141.01 (1.01, 1.02) < 0.0001HDL-C, mmol/L1.46 ± 0.400.10 (0.07, 0.14) < 0.0001TC, mmol/L5.13 ± 0.861.47 (1.32, 1.64) < 0.0001TG, mmol/L0.91 ± 0.662.00 (1.83, 2.19) < 0.0001HbA1c, %5.17 ± 0.3236.72 (26.06, 51.73) < 0.0001Alcohol consumptionNon11,802 ($76.37\%$)ReferenceLight1754 ($11.35\%$)1.01 (0.72, 1.42)0.9441Moderate1357 ($8.78\%$)1.22 (0.86, 1.72)0.2722Heavy540 ($3.49\%$)2.55 (1.73, 3.76) < 0.0001Smoking statusNever9027 ($58.42\%$)ReferencePast2949 ($19.08\%$)1.64 (1.24, 2.17)0.0005Current3477 ($22.50\%$)2.78 (2.21, 3.50) < 0.0001FPG, mmol/L5.16 ± 0.4124.58 (17.91, 33.74) < 0.0001SBP, mmHg114.49 ± 14.971.03 (1.02, 1.04) < 0.0001DBP, mmHg71.58 ± 10.501.05 (1.04, 1.06) < 0.0001TG (mmol/L)/HDL-C(mmol/L)0.74 ± 0.751.75 (1.62, 1.89) < 0.0001CI confidence interval, OR odds ratio, BMI body mass index, WC waist circumference, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT γ-glutamyl transferase, HDL-C high-density lipoprotein cholesterol, TC total cholesterol, TG triglyceride, HbA1c hemoglobin A1c, FPG fasting plasma glucose, SBP systolic blood pressure, DBP diastolic blood pressure.
## Independent effect of baseline TG/HDL-C on the risk of T2DM
A Cox proportional risk regression model was used to analyze the association with baseline TG/HDL-C and T2DM risk. Table 3 shows the results of the analyses in the no-adjusted, partially adjusted, and fully adjusted models, respectively. When TG/HDL-C was used as a continuous variable, it was strongly related to diabetes risk in the unadjusted model, with each 1-unit increase in TG/HDL-C associated with a $46\%$ increase in diabetes risk. After partially adjustment and full adjustment for confounding, a positive association between TG/HDL-C and T2DM risk remained. We then categorized participants into quartiles based on baseline TG/HDL-C. The results showed that the relationship between TG/HDL-C and T2DM first decreased and then increased in model 3. HRs and $95\%$CIs for Q2–Q4 were 0.90 (0.54–1.51), 0.86 (0.52–1.42), and 1.12 (0.68–1.84), respectively, when compared to Q1.Table 3Relationship between TG/HDL-C and incident diabetes. OutcomeModel 1Model 2Model 3HR ($95\%$ CI)P-valueHR ($95\%$ CI)P-valueHR ($95\%$ CI)P-valueTG/HDL-C1.46 (1.40, 1.52) < 0.0011.41 (1.35, 1.48) < 0.0011.19 (1.09, 1.30) < 0.001TG/HDL-C(quartile)Q1ReferenceReferenceReferenceQ21.61 (0.97, 2.68)0.06461.41 (0.84, 2.34)0.19260.90 (0.54, 1.51)0.6869Q32.90 (1.81, 4.64) < 0.00012.22 (1.37, 3.61)0.00130.86 (0.52, 1.42)0.5549Q47.59 (4.90, 11.77) < 0.00015.51 (3.45, 8.80) < 0.00011.12 (0.68, 1.84)0.6593Model 1: not adjusted other covariants. Model 2: adjusted for gender and age. Model 3: adjusted for gender, age, exercise habits, body mass index, hemoglobin A1c, fatty liver, total cholesterol, smoking situation, alcohol consumption, fasting plasma glucose. HR hazard ratio, CI confidence interval, TG triglyceride, HDL-C high-density lipoprotein cholesterol.
## Threshold effect analysis
As Cox regression analysis showed inconsistent results on the prevalence for diabetes when TG/HDL-C was considered as a categorical and continuous variable. The nonlinear relationship between TG/HDL-C and T2DM was analyzed using the generalized additive model (GAM). A J-shaped association of TG/HDL-C with T2DM was found through smoothed curve fitting (Fig. 1).Figure 1GAM and smoothed curve fitting were used to investigate the relationship between TG/HDL-C ratio and the incidence of T2DM. The red solid line indicates the estimated risk of developing T2DM. The green dashed line indicates the $95\%$ confidence interval of the fit. After adjusting for gender, age, exercise habits, body mass index, fatty liver, total cholesterol, hemoglobinA1c, smoking status, alcohol consumption, fasting plasma glucose, a J-shaped relationship was detected between TG/HDL-C ratio and the incidence of T2DM, with the risk of developing T2DM decreasing with increasing TG/HDL-C on the left side of the inflection point and the opposite relationship observed on the right side of the inflection point. Association between TG/HDL-C and T2DM in the Japanese population: J-shaped association between TG/HDL-C and T2DM. The solid red line indicates the smoothed curve fit between the variables. The green dashed line indicates the $95\%$ confidence interval of the fit. Adjusted for gender, age, exercise habits, body mass index, fatty liver, total cholesterol, hemoglobinA1c, smoking status, alcohol consumption, fasting plasma glucose.
Threshold effects analysis was performed by smoothed curve fitting and segmented regression models to determine the inflection point of the association of TG/HDL-C with T2DM. The inflection point of TG/HDLC was 0.35. TG/HDL-C < 0.35, TG/HDL-C was negatively associated with T2DM, whereas TG/HDL-C > 0.35, TG/HDL-C was positively associated with T2DM (Table 4).Table 4The results two-piecewise linear regression mode. OutcomesHR ($95\%$ CI)P-valueOne-line linear regression model1.19 (1.09, 1.30) < 0.001Inflection point of TG/HDL-C< 0.350.27 (0.01, 6.55),0.4234> 0.351.20 (1.10, 1.31), < 0.0001Log-likelihood ratio test0.0376Adjusted for gender, age, exercise habits, body mass index, hemoglobin A1c, fatty liver, total cholesterol, smoking situation, alcohol consumption, fasting plasma glucose. HR hazard ratio, CI confidence interval, TG triglyceride, HDL-C high-density lipoprotein cholesterol.
## The results of subgroup analyses
The correlation between T2DM and TG/HDL-C in different subgroups was shown in Table 5. We grouped the variables of age, gender, smoking status, exercise habits, and alcohol consumption and performed subgroup and interaction analyses, respectively. The results showed that the correlation between TG/HDL-C and T2DM was stable across subgroups, and further interaction analysis did not reveal any significant differences between subgroups. Table 5Subgroup analyses of the association between TG/HDL-C and incident type 2 diabetes. CharacteristicNo. of participantsHR ($95\%$CI)P -valueP for interactionAge (year)0.862618–3530911.24 (0.93, 1.64)0.140836–3924931.29 (1.09, 1.53)0.002840–4432811.19 (1.01, 1.40)0.039845–5133731.15 (0.96, 1.37)0.131452–7932151.15 (0.96, 1.37)0.1314Gender0.6478Female70341.26 (0.99, 1.61)0.0629Male84191.19 (1.08, 1.30)0.0003Alcohol consumption0.6951Non11,8021.17 (1.05, 1.30)0.0033Light17541.28 (1.04, 1.57)0.0204Moderate13571.35 (0.78, 1.54)0.0384Heavy5401.10 (1.10, 1.30)0.5894Fatty liver0.8435No12,7161.19 (1.03, 1.38)0.0171Yes27371.21 (1.08, 1.34)0.0005Smoking status0.6490Never90271.07 (0.90, 1.28)0.4385Past29491.32 (1.11, 1.56)0.0015Current34771.22 (1.07, 1.40)0.0033Habit of exercise0.5374No12,7471.21 (1.10, 1.33) < 0.0001Yes27061.11 (0.86, 1.44)0.4247Adjusted for age, gender, fatty liver, BMI, habit of exercise, total cholesterol, HbA1c, alcohol consumption, smoking status, fasting plasma glucose except the subgroup variable.
## Discussion
This report explored the relationship between TG/HDLC and T2DM. The correlation between TG/HLD-C and T2DM has been investigated in several studies on populations in different regions of China, and their studies12,13 showed a non-linear connection between TG/HDL-C and diabetes events. This is consistent with our finding. Unexpectedly, in this study, T2DM showed a J-shaped relationship with baseline TG/HDL-C after adjusting for confounders gender, age, exercise habits, BMI, fatty liver, total cholesterol, FPG, HbA1c, alcohol consumption, and smoking status (Table 4 and Fig. 1). In addition, we calculated a threshold value of 0.35 for TG/HDL-C by threshold effect analysis. It's worth noticing that the link between TG/HDL-C and T2DM had the opposite effect on different sides of threshold value. Participants had the lowest risk of T2DM when TG/HDL-C was approximately 0.35, and when TG/HDL-C was lower than 0.35, TG/HDL-C was negatively linked with T2DM. However, the risk was not statistically significant (HR 0.27, $95\%$ CI 0.01–6.55, P-value 0.4234). When TG/HDL-C > 0.35, there was a positive association between TG/HDL-C and the risk of diabetes, suggesting that the risk of diabetes is increased with either a high or low TG/HDL-C. The mechanism by which a high TG/HDL-C ratio increases the incidence of diabetes is unclear. Previous studies have suggested dyslipidemia as a causal factor of insulin resistance28. An increase in TG and decreased HDL-C levels through genetic variants in lipid-related genes could cause insulin resistance29. It results in compensatory hyperinsulinemia, leading to aggravation of hypertriglyceridemia. However, the J-shaped association between TG/HDL-C and T2DM and the mechanisms behind the threshold value are unclear. The issue has significant physiological and clinical implications based on the impact of dyslipidemia on diabetes.
According to studies, disorders of lipid metabolism have a major role in the development of T2DM, and the impact of dyslipidemia on developing type 2 diabetes cannot be ignored5,30–32. Elevated triglycerides decrease insulin sensitivity and increase the risk of developing diabetes33–35, while high-density lipoprotein cholesterol plays a protective role36. Elevated plasma triglycerides and decreased HDL-C are danger markers and predictors of diabetic events and insulin resistance in the population37. It was shown that elevated triglycerides, TG/HDL-C and decreased HDL-C can contribute to the onset and progression of diabetes38. The ratio of TG/HDL-C is a relatively easy, convenient, and low-cost indicator obtained during routine clinical care or physical examination and is considered by some authors to be a highly sensitive and specific predictive indicator of diabetes31,38. TG/HDL-C is better than TG or HDL-C alone on predicting diabetes risk38. Therefore, it is more than recommended to use TG/HDL-C for predicting impaired beta-cell and insulin resistance39–43. Studies have reported that the predictive function of TG/HDL-C for diabetes is race-specific, and it has been suggested that TG/HDL-C could be a useful predictive indicator for diabetes in Chinese Singaporean, Hispanic, and African American, as well as Chinese populations16,17,41,44. We reviewed the relevant literature and found several studies that associated TG/HDL-C with T2DM. Liu's research team and Kim's research team noted a close connection between the TG/HDL-C and T2DM13,19. A cohort study of 114,787 Chinese participants showed a positive relationship between TG/HDL-C and diabetes risk, using subjects in the lowest quartile of TG/HDL-C as a reference, subjects in the highest quartile of TG/HDL-C were more susceptible to acquiring T2DM13. Similar results were found by Uruska et al. for the study of the TG/HDL-C ratio to assess IR in patients with type 1 diabetes45. Our results also show that TG/HDL-C is positively associated with the risk of T2DM by a proportional hazards model, and their relationship remains positive after adjusting for different confounders, and the results suggest an independent relationship between them.
Glucose and lipid metabolism are influenced by various factors, and it remains controversial whether TG/HDL-C was correlation with diabetes differs between genders. Liu et al. showed that the results of subgroup analysis indicated that the correlation of TG/HDL-C ratio on the incidence of T2DM was not significantly different between genders, with a P-value of 0.53 for their interaction12. Similarly, a cohort study by Chen et al. showed the same results for gender-specific subgroups, with $$P \leq 0.058$$ for their interaction13. To investigate the differences between Japanese men and Japanese women in the association of TG/HDL-C withT2DM, we performed a subgroup analysis in this study (Table 5). Our study showed a non-significant difference in TG/HDL-C and T2DM between genders, female (HR = 1.26, $$P \leq 0.06$$) vs. male (HR = 1.19, $P \leq 0.01$), with a P-value of 0.65 for the interaction.
Our findings show there was no significant difference in the risk of diabetes with increased TG/HDL-C by gender in the Japanese population. And the finding indicated with increased TG/HDL-C that the risk of diabetes was consistent between genders in the Japanese population. However, other studies have obtained different results: for example, some studies in Iranian, Chinese, Chinese Singaporeans, and Japanese populations suggest that the correlation between TG/HDL-C and T2DM as significantly higher in females than in males17,19,20,46. Another study concluded that high TG/HDL-C was an influential factor in incident diabetes in men participants and that TG/HDL-C was available to infer the risk of T2DM in male, but their study did not include female participants47. Qin et al. investigated the effect of TG/HDL-C on diabetes in Chinese adults and whether there were differences between genders and discovered that the correlation between TG/HDL-C and diabetes was independent44. This association was significant only in Chinese adult males44. Similarly, Zhang et al. reported that the trajectory of TG/HDL-C was only observed to be correlated with the progression of diabetes in men, but not in women16. Whether gender affects the relationship between lipid metabolism and diabetes can be further investigated. We have analysed these studies that are inconsistent with our results and speculate that the reasons for the different results may be due to the following factors: Firstly, it is thought that the higher risk of diabetes in females than in males is due to the dysregulation of glucose and lipid metabolism caused by the decline in oestrogen levels in women after menopause48, which may put women at greater risk of developing T2DM. The mean age of the women in our study population was 43.25 years and our findings are limited by the fact that the original data did not register whether the women were menopausal or not. Second, Song S et al. thought that differences in dietary patterns were responsible for gender differences in the effect of TG/HDL-C on the risk of diabetes in Korean adults49. Third, the association between TG/HDL-C and insulin resistance is race-specific, and gender differences may vary between races.
Our study has several strengths. First, this investigation is the first to show a J-shaped correlated between baseline TG/HDL-C and T2DM risk. Second, this is a large cohort study, involving a relatively large number of people, and is highly representative of the Japanese population. Third, to improve the stability of the results, we analyzed TG/HDL-C as categorical and continuous variables, respectively. Fourth, to explore the impact of baseline TG/HDL-C on T2DM in different populations, a subgroup analysis was conducted in this study.
Despite its strengths, the study has some limitations. First, some covariates were not available in the study due to the limitations of the original study data; therefore, residual confounding may be present in this study. Second, this investigation did not distinguish between types of diabetes mellitus. But, type 1 diabetes is not extremely common in the Japanese population50. Therefore, we inferred that almost all new-onset diabetes in this study was T2DM. Third, in this investigation, oral glucose tolerance tests were not used to screen for T2DM, so the findings may have underestimated the risk of developing T2DM. Fourth, given that T2DM prevalence is associated with race and region, and the study population in this study was Japanese, this result is not necessarily generalizable to populations outside of Japan.
## Conclusion
Our findings indicate an independent relationship between baseline TG/HDL-C and T2DM in the Japanese population. This investigation is the first to show a J-shaped relationship between baseline TG/HDL-C and T2DM risk. People with TG/HDL-C higher than 0.35 are at greater risk of developing diabetes, and we should pay more attention to this group for diabetes prevention.
## References
1. Zhou B. **Worldwide trends in diabetes since 1980: a pooled analysis of 751 population-based studies with 4.4 million participants**. *Lancet* (2016) **387** 1513-1530. DOI: 10.1016/S0140-6736(16)00618-8
2. Sun H. **IDF diabetes atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045**. *Diabetes Res. Clin. Pract.* (2022) **183** 109119. DOI: 10.1016/j.diabres.2021.109119
3. Nanditha A. **Diabetes in Asia and the Pacific: Implications for the global epidemic**. *Diabetes Care* (2016) **39** 472-485. DOI: 10.2337/dc15-1536
4. Hills AP. **Epidemiology and determinants of type 2 diabetes in south Asia**. *Lancet Diabetes Endocrinol.* (2018) **6** 966-978. DOI: 10.1016/S2213-8587(18)30204-3
5. Wu L, Parhofer KG. **Diabetic dyslipidemia**. *Metabolism* (2014) **63** 1469-1479. DOI: 10.1016/j.metabol.2014.08.010
6. von Eckardstein A, Sibler RA. **Possible contributions of lipoproteins and cholesterol to the pathogenesis of diabetes mellitus type 2**. *Curr. Opin. Lipidol.* (2011) **22** 26-32. DOI: 10.1097/MOL.0b013e3283412279
7. Zheng Y, Ley SH, Hu FB. **Global aetiology and epidemiology of type 2 diabetes mellitus and its complications**. *Nat. Rev. Endocrinol.* (2018) **14** 88-98. DOI: 10.1038/nrendo.2017.151
8. Kahn SE. **The relative contributions of insulin resistance and beta-cell dysfunction to the pathophysiology of Type 2 diabetes**. *Diabetologia* (2003) **46** 3-19. DOI: 10.1007/s00125-002-1009-0
9. Matthaei S. **Pathophysiology and pharmacological treatment of insulin resistance**. *Endocr. Rev.* (2000) **21** 585-618. PMID: 11133066
10. DeFronzo RA, Tobin JD, Andres R. **Glucose clamp technique: A method for quantifying insulin secretion and resistance**. *Am. J. Physiol.* (1979) **237** E214-E223. PMID: 382871
11. Ozder A. **Lipid profile abnormalities seen in T2DM patients in primary healthcare in Turkey: A cross-sectional study**. *Lipids Health Dis.* (2014) **13** 183. DOI: 10.1186/1476-511X-13-183
12. Liu H. **Association of the ratio of triglycerides to high-density lipoprotein cholesterol levels with the risk of type 2 diabetes: A retrospective cohort study in Beijing**. *J. Diabetes Res.* (2021) **2021** 5524728. DOI: 10.1155/2021/5524728
13. Chen Z. **Association of Triglyceride to high-density lipoprotein cholesterol ratio and incident of diabetes mellitus: A secondary retrospective analysis based on a Chinese cohort study**. *Lipids Health Dis.* (2020) **19** 33. DOI: 10.1186/s12944-020-01213-x
14. Giannini C. **The triglyceride-to-HDL cholesterol ratio: Association with insulin resistance in obese youths of different ethnic backgrounds**. *Diabetes Care* (2011) **34** 1869-1874. DOI: 10.2337/dc10-2234
15. Liu H. **Triglyceride to high-density lipoprotein cholesterol (TG/HDL-C) ratio, a simple but effective indicator in predicting type 2 diabetes mellitus in older adults**. *Front. Endocrinol.* (2022) **13** 828581. DOI: 10.3389/fendo.2022.828581
16. Zhang Y. **Association of TG/HDLC ratio trajectory and risk of type 2 diabetes: A retrospective cohort study in China**. *J. Diabetes* (2020) **1** 1-10
17. Wang YL. **Association between the ratio of triglyceride to high-density lipoprotein cholesterol and incident type 2 diabetes in Singapore Chinese men and women**. *J. Diabetes* (2017) **9** 689-698. DOI: 10.1111/1753-0407.12477
18. Kim J. **Positive association between the ratio of triglycerides to high-density lipoprotein cholesterol and diabetes incidence in Korean adults**. *Cardiovasc. Diabetol.* (2021) **20** 183. DOI: 10.1186/s12933-021-01377-5
19. Hadaegh F. **Lipid ratios and appropriate cut off values for prediction of diabetes: A cohort of Iranian men and women**. *Lipids Health Dis.* (2010) **9** 85. DOI: 10.1186/1476-511X-9-85
20. Fujihara K. **Utility of the triglyceride level for predicting incident diabetes mellitus according to the fasting status and body mass index category: The Ibaraki Prefectural Health Study**. *J. Atheroscler. Thromb.* (2014) **21** 1152-1169. DOI: 10.5551/jat.22913
21. Okamura T. **Ectopic fat obesity presents the greatest risk for incident type 2 diabetes: A population-based longitudinal study**. *Int. J. Obes.* (2019) **43** 139-148. DOI: 10.1038/s41366-018-0076-3
22. Okamura TEA. *Dataset* (2019). DOI: 10.5061/dryad.8q0p192
23. Hashimoto Y. **Modest alcohol consumption reduces the incidence of fatty liver in men: A population-based large-scale cohort study**. *J. Gastroenterol. Hepatol.* (2015) **30** 546-552. DOI: 10.1111/jgh.12786
24. Okamura T. **Creatinine-to-bodyweight ratio is a predictor of incident non-alcoholic fatty liver disease: A population-based longitudinal study**. *Hepatol. Res.* (2020) **50** 57-66. DOI: 10.1111/hepr.13429
25. Du T. **Clinical usefulness of lipid ratios, visceral adiposity indicators, and the triglycerides and glucose index as risk markers of insulin resistance**. *Cardiovasc. Diabetol.* (2014) **13** 146. DOI: 10.1186/s12933-014-0146-3
26. Hamaguchi M. **The severity of ultrasonographic findings in nonalcoholic fatty liver disease reflects the metabolic syndrome and visceral fat accumulation**. *Am. J. Gastroenterol.* (2007) **102** 2708-2715. DOI: 10.1111/j.1572-0241.2007.01526.x
27. **Standards of medical care in diabetes: 2011**. *Diabetes Care* (2011) **34** S11-61. PMID: 21193625
28. Reaven G. **Insulin resistance and coronary heart disease in nondiabetic individuals**. *Arterioscler. Thromb. Vasc. Biol.* (2012) **32** 1754-1759. DOI: 10.1161/ATVBAHA.111.241885
29. Li N. **Are hypertriglyceridemia and low HDL causal factors in the development of insulin resistance?**. *Atherosclerosis* (2014) **233** 130-138. DOI: 10.1016/j.atherosclerosis.2013.12.013
30. Gupta M. **An update on pharmacotherapies in diabetic dyslipidemia**. *Prog. Cardiovasc. Dis.* (2019) **62** 334-341. DOI: 10.1016/j.pcad.2019.07.006
31. Schulze MB. **Use of multiple metabolic and genetic markers to improve the prediction of type 2 diabetes: The EPIC-Potsdam Study**. *Diabetes Care* (2009) **32** 2116-2119. DOI: 10.2337/dc09-0197
32. Sokooti S. **Triglyceride-rich lipoprotein and LDL particle subfractions and their association with incident type 2 diabetes: The PREVEND study**. *Cardiovasc. Diabetol.* (2021) **20** 156. DOI: 10.1186/s12933-021-01348-w
33. Zhao J. **Triglyceride is an independent predictor of type 2 diabetes among middle-aged and older adults: A prospective study with 8-year follow-ups in two cohorts**. *J. Transl. Med.* (2019) **17** 403. DOI: 10.1186/s12967-019-02156-3
34. Bhowmik B. **Serum lipid profile and its association with diabetes and prediabetes in a rural Bangladeshi population**. *Int. J. Environ. Res. Public Health* (2018) **15** 9. DOI: 10.3390/ijerph15091944
35. Tirosh A. **Changes in triglyceride levels over time and risk of type 2 diabetes in young men**. *Diabetes Care* (2008) **31** 2032-2037. DOI: 10.2337/dc08-0825
36. He Y, Kothari V, Bornfeldt KE. **High-density lipoprotein function in cardiovascular disease and diabetes mellitus**. *Arterioscler. Thromb. Vasc. Biol.* (2018) **38** e10-e16. DOI: 10.1161/ATVBAHA.117.310222
37. Bonora E. **Prevalence of insulin resistance in metabolic disorders: The Bruneck study**. *Diabetes* (1998) **47** 1643-1649. DOI: 10.2337/diabetes.47.10.1643
38. He S. **Higher ratio of triglyceride to high-density lipoprotein cholesterol may predispose to diabetes mellitus: 15-year prospective study in a general population**. *Metabolism* (2012) **61** 30-36. DOI: 10.1016/j.metabol.2011.05.007
39. Kannel WB. **Usefulness of the triglyceride-high-density lipoprotein versus the cholesterol-high-density lipoprotein ratio for predicting insulin resistance and cardiometabolic risk (from the Framingham Offspring Cohort)**. *Am. J. Cardiol.* (2008) **101** 497-501. DOI: 10.1016/j.amjcard.2007.09.109
40. Maturu A. **The triglyceride to high-density lipoprotein cholesterol (TG/HDL-C) ratio as a predictor of β-cell function in African American women**. *Metabolism* (2015) **64** 561-565. DOI: 10.1016/j.metabol.2015.01.004
41. Young KA. **The triglyceride to high-density lipoprotein cholesterol (TG/HDL-C) ratio as a predictor of insulin resistance, β-cell function, and diabetes in Hispanics and African Americans**. *J. Diabetes Compl.* (2019) **33** 118-122. DOI: 10.1016/j.jdiacomp.2018.10.018
42. Zhou M. **The triglyceride to high-density lipoprotein cholesterol (TG/HDL-C) ratio as a predictor of insulin resistance but not of β cell function in a Chinese population with different glucose tolerance status**. *Lipids Health Dis* (2016) **15** 104. DOI: 10.1186/s12944-016-0270-z
43. Tushuizen ME. **Pancreatic fat content and beta-cell function in men with and without type 2 diabetes**. *Diabetes Care* (2007) **30** 2916-2921. DOI: 10.2337/dc07-0326
44. Qin H. **Triglyceride to high-density lipoprotein cholesterol ratio is associated with incident diabetes in men: A retrospective study of Chinese individuals**. *J. Diabetes Investig.* (2020) **11** 192-198. DOI: 10.1111/jdi.13087
45. Uruska A. **TG/HDL-C ratio and visceral adiposity index may be useful in assessment of insulin resistance in adults with type 1 diabetes in clinical practice**. *J. Clin. Lipidol.* (2018) **12** 734-740. DOI: 10.1016/j.jacl.2018.01.005
46. Cheng C. **Dose-response association between the triglycerides: High-density lipoprotein cholesterol ratio and type 2 diabetes mellitus risk: The rural Chinese cohort study and meta-analysis**. *J. Diabetes* (2019) **11** 183-192. DOI: 10.1111/1753-0407.12836
47. Vega GL. **Triglyceride-to-high-density-lipoprotein-cholesterol ratio is an index of heart disease mortality and of incidence of type 2 diabetes mellitus in men**. *J. Investig. Med.* (2014) **62** 345-349. DOI: 10.2310/JIM.0000000000000044
48. Faulds MH. **The diversity of sex steroid action: regulation of metabolism by estrogen signaling**. *J. Endocrinol.* (2012) **212** 3-12. DOI: 10.1530/JOE-11-0044
49. Song S, Lee JE. **Dietary patterns related to triglyceride and high-density lipoprotein cholesterol and the incidence of type 2 diabetes in KOREAN men and women**. *Nutrients* (2018) **11** 1-8. DOI: 10.3390/nu11010008
50. Tajima N, Morimoto A. **Epidemiology of childhood diabetes mellitus in Japan**. *Pediatr. Endocrinol. Rev.* (2012) **10** 44-50. PMID: 23330245
|
---
title: Comparison of moderate-intensity continuous training and high-intensity interval
training effects on the Ido1-KYN-Ahr axis in the heart tissue of rats with occlusion
of the left anterior descending artery
authors:
- Pouria Nori
- Rouhollah Haghshenas
- Younes Aftabi
- Hakimeh Akbari
journal: Scientific Reports
year: 2023
pmcid: PMC9988842
doi: 10.1038/s41598-023-30847-x
license: CC BY 4.0
---
# Comparison of moderate-intensity continuous training and high-intensity interval training effects on the Ido1-KYN-Ahr axis in the heart tissue of rats with occlusion of the left anterior descending artery
## Abstract
Myocardial infarction (MI) affects many molecular pathways in heart cells, including the Ido1-KYN-Ahr axis. This pathway has recently been introduced as a valuable therapeutic target in infarction. We examined the effects of moderate-intensity continuous training (MICT) and high-intensity interval training (HIIT) on the axis in the heart tissue of male Wistar rats with occluded left anterior descending (OLAD). Thirty rats (age 10–12 weeks, mean weight 275 ± 25 g) were divided into five groups with 6 animals: Control (Ct) group, MICT group, rats with OLAD as MI group, rats with OLAD treated with MICT (MIMCT group) and rats with OLAD treated with HIIT (MIHIIT group). Rats performed the training protocols for 8 weeks, 5 days a week. HIIT included 7 sets of 4 min running with an intensity of 85–$90\%$ VO2max and 3 min of recovery activation between sets. MICT included continuous running at the same distance as HIIT with an intensity of 50–$60\%$ VO2max for 50 min. The expressions of Ahr, Cyp1a1, and Ido1 were assayed by real-time PCR. Malondialdehyde (MDA) and Kynurenine levels, and AHR, CYP1A1, and IDO1 proteins were detected using ELISA. Data were analyzed using the ANOVA and MANOVA tests. Compared to the CT group, MI caused an increase in all studied factors, but only statistically significant ($P \leq 0.05$) for MDA and IDO1. With a greater effect of HIIT, both protocols significantly lowered the proteins expressions in the MIHIIT and MIMCT groups, compared with the MI group ($P \leq 0.001$). In healthy rats, only AHR protein significantly decreased in the MICT group compared to the Ct group ($P \leq 0.05$). HIIT and MICT protocols significantly reduced the gene and protein expression of Cyp1a1 ($P \leq 0.05$) and Ido1 ($P \leq 0.01$), and HIIT had a greater effect. In conclusion, both protocols were effective at reducing the levels of Ido1-Kyn-*Ahr axis* components and oxidative stress in the infarcted heart tissue and HIIT had a higher significant effect.
## Introduction
Myocardial infarction (MI) as a type of ischemic cardiovascular disease promotes adverse remodeling of the left ventricle by affecting cardiomyocytes and vascular cells, which altogether develop the first cause of morbidity and mortality worldwide1. These cellular changes involve a wide range of molecular pathways including the Kynurenine (KYN) metabolism and Ido1-KYN-Ahr axis2. KYN is a metabolite produced from the amino acid tryptophan (TRP) by the activity of tryptophan 2,3-dioxygenase and indoleamine 2,3-dioxygenase 1 (IDO1) and 2 enzymes, which regulates various immune and physiological responses via binding to and activation of the aryl hydrocarbon receptor (AHR)3,4. AHR is a ligand-activated transcription factor with myriad functions in health and diseases, which works upon binding to many intrinsic and extrinsic chemicals in physiological and immune responses and Cyp1a1 upregulation is one of the main hallmarks of its activation5,6. Ahr knockout in a mouse model has been shown that it has highly crucial roles in maintaining the function, health, and physiological homeostasis of cardiac cells and tissues7. In MI, the Ido1-KYN-*Ahr axis* has paracrine effects on cardiomyocyte apoptosis and contractility and cardiac remodeling and function2. Furthermore, the levels of catabolites of this pathway in body fluids have been suggested as markers positively associated with MI and associated mortality2. KYN generation through IDO is markedly induced after MI and KYN metabolites may increase inflammation, oxidative stress, and apoptosis of smooth muscle cells and endothelial cells including cardiomyocytes8–10. It is shown that IDO1 activity has an inverse association with ischemic heart disease and therefore it has been introduced as a potential therapeutic target for this disease11.
Recently, it has been shown that exercise affects the Ido1-KYN-*Ahr axis* in different cells and tissues. Physical exercise has been shown to impact the KYN pathway (KP) in response to both acute and chronic exercise training stimuli and currently, it is accepted that exercise-induced KP may contribute to the prevention and treatment of chronic diseases12,13. Surprisingly, due to the effects of KP metabolites on skeletal muscle, adipose tissue, the immune system, and brain physiology, some researchers proposed that some of these metabolites could be suggested as exercise-induced myokines14. Also, other members of the Ido1-KYN-*Ahr axis* have been shown to interact with exercise metabolism. Pal et al., found that both acute and chronic endurance training may regulate NK cell function via the AHR/IDO axis15.
Increasing evidence confirmed exercise has improving effects on the cardiac function of MI patients16,17. All types of exercise can effectively inhibit skeletal muscle atrophy via reducing oxidative stress and protein degradation, increasing the antioxidant capacity, and regulating the growth factors expression18. However, there are different opinions about the impacts of exercise types on MI improvement. Cai et al. believe that in the early stages of MI, moderate-intensity exercise is the best choice to improve the outcomes for MI patients18. And Dun et al., reported that compared to moderate-intensity continuous training (MICT), supervised high-intensity interval training (HIIT) results in greater improvements in MI patients with metabolic syndrome16. Also, Moholdt et al. found that in MI patients aerobic interval training increases peak oxygen uptake more than usual exercise training19. Furthermore, different types of exercise affect KP in different ways. Joisten et al., reported that HIIT consistently led to greater effects than MICT on KP in persons with multiple sclerosis20. Currently, experimental studies aiming for a deeper understanding of cellular and molecular mechanisms underlying exercise interaction with the Ido1-KYN-*Ahr axis* in heart tissues of MI patients are still lacking. To fill some gaps in the current knowledge, here we conducted an in vivo investigation to compare the effects of MICT and HIIT on the axis in the heart tissue of rats with occlusion of the left anterior descending artery.
## Experimental animals
Thirty male Wistar rats (age 10–12 weeks, mean weight 275 ± 25 g) were purchased from the Pasteur institute of Iran. Rats were kept in polycarbonate cages (Three rats per cage) on a 12 h light/dark cycle and a humidity of 65 ± $5\%$ and a temperature of 25 °C and were provided food (rat chow) and water ad libitum. All experimental protocols were approved by the Committee on the Care of Laboratory Animal Resources, Semnan university of medical science (IR.SEMUMS.REC.1399.158) and were carried out following the Declaration of Helsinki, the ARRIVE guidelines and the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health. Eighteen rats underwent surgery for occlusion of the left anterior descending (LAD). After evaluating and confirming MI using echocardiography, based on ejection fraction and fractional shortening the rats were divided into five groups including six members in each: Control group (Ct), a group treated with MICT, rats with occluded LAD (OLAD) as a model of MI, rats with OLAD and treated with MICT (MIMCT), and a group with OLAD and treated with HIIT (MIHIIT). Care of the animals was performed according to the European Convention for the Protection of Vertebrate Animals (ECPVA). The number of members in the groups was calculated following Charan and Biswas’s descriptions21.
## Occlusions of left anterior descending
After anesthesia of the animal using a combination of ketamine and xylazine, the surgical site on the animal’s chest was disinfected with $70\%$ alcohol. After keeping the animal fixed on the operating desk, using an otoscope number 3 and a green angiocatheter, the animal was intubated and connected to a ventilator (inter med Bear) (inhaul to exhalation ratio of 1 to 2 and 80–90 breaths per minute with a volume of 8 ml). In the space between the third and fourth ribs, the chest was cut to a length of 10 mm. With this incision, the LAD vessel was identified as a bright red pulsating spike that flows in the middle of the heart wall from under the left atrium to the apex of the heart. The LAD vessel was closed with the help of 0.6 mm polypropylene suture 1–2 mm below the level of the tip of the left atrium and was completely closed by tying two knots at this point. Left ventricular anterior wall infarction was confirmed by sudden myocardial coloration (discoloration). An increase in ST was also observed after ligation. Then, the chest, muscle layers, and skin were sewn in three layers using 0.5 proline suture and the animal's skin was sutured with 0.3 proline suture. When the rats regained consciousness, they were removed from the ventilator. After 48 h, the rats were anesthetized again and with echo vivid7 probe 10 s (MHz), an echo was performed to determine MI. In addition, cefazolin and tramadol as antibiotics and analgesics were injected twice a day, 1 day before surgery and 3 days after surgery. Rats in the MI group underwent all surgeries without occlusion of the left coronary artery. Also, Ct group rats did not receive any intervention and were kept only in the laboratory.
## Exercise training
After evaluating and confirming MI and a week of rest MIHIIT, MICT, and MIMCT groups trained for three sessions per week for 2 weeks (each session 10–15 min at speed of 10 to 15 m per minute with a treadmill) to get acquainted with the training protocols. After 2 weeks of initial familiarization with the treadmill, the intensity of the training program was obtained in terms of VO2max and the relationship between VO2max and the speed and incline of the treadmill22. The aerobic capacity of the rats was evaluated after the initial warm-up. The test started with an initial speed of 6 m/min. The speed of the treadmill was increased every 2 min by 1.8–2 m/min until the rats reached the exhaustive stage. After obtaining the test speed and time, their average was calculated, and based on it, the protocol of training was designed. Then, the rats performed the two designed protocols of training based on previous reports22–24 for 8 weeks, 5 days a week with a slope of zero degrees (Table 1): (A) HIIT, which consisted of seven sets of four minutes running with the intensity of 85–$90\%$ VO2max and three minutes of recovery activation between sets with the intensity of 50–$60\%$ VO2max. ( B) MICT, which consisted of continuous running in the same distance as HIIT with the intensity of 50–$60\%$ VO2max for 50 min. In both groups, warming and cooling down periods were performed for 5 min before and after the training with an intensity of $40\%$ VO2max. The intensity. Table 1Protocol of exercise training. Warm-upHIITMICTCooldownTotal time7 sets of 4 min3 minIntensity VO2max$40\%$85–$90\%$$40\%$60–$65\%$$40\%$60 min1-2th week5*20–221115–1753-4th week5*22–241317–1955-6th week5*24–261519–2157-8th week5*26–281721–235*Meter/Minute: The intensity is expressed in meter/minute and the duration in minutes (min).HIIT high intensity of interval training, MICT moderate intensity of continues training.
## Tissue sample preparation
Tissue resection was performed at the end of the eighth week and 72 h after the last training session. This was after performing anesthesia with CO2 gas and blood sampling from the heart. Then, heart tissue was extracted and after washing in physiological serum placed in microtubes and transferred to − 70 °C.
## Determination of malondialdehyde and KYN
For quantitative assay of KYN and malondialdehyde (MDA) in the heart tissue lysate, Kynurenine ELISA Kit, ZellBio GmbH (Cat. No: ZB-11203C-R9648; Germany) and MDA Assay kit (CAT No. ZB-MDA-96A; Germany) were used following the manufacturer’s instruction.
## Gene expression analysis
Total RNA was isolated from heart samples using TRIzol Reagent (Invitrogen, USA), treated with Dnase I, and quantified by NanoDrop (Thermo Fisher Scientific). RNA quality was determined by examining the $\frac{260}{280}$ ratio > 1.8. A total of 1 µg RNA was then reverse transcribed to cDNA using the RevertAid First Strand cDNA Synthesis kit (Thermo Scientific) according to the manufacturer’s instructions. Expression of Ahr, Cyp1a1, and Ido1, was measured using specific primers produced by SinaColon, Iran (Table 2). Amplification was performed in ABI Prism 7500 sequence detection system; Life Technologies real-time RT-PCR device using the condition: initial denaturation (95 °C for 15 min for all genes); start of the cycle with denaturation (95 °C for 30 s for Gapdh, and 95 °C for 5 min for Ahr, Cyp1a1, and Ido1); annealing (55 °C for 30 s for Gapdh, 48 °C for 105 s for Ido1, 56 °C for 90 s for Cyp1a1and 60 °C for 90 s for Ahr); and extension (60 °C for 30 min for all genes). At the end of amplification cycles, reactions were given a final extension step at 60–95 °C. All data were analyzed with the ΔΔCt method and the expression of glyceraldehyde-3-phosphate dehydrogenase (Gapdh) was used as the internal standard. Table 2Primers. Gene accession noSequencesAmplicon size (bp)Annealing Tm (°C)RefAhrFW: 5′-TCACTGCGCAGAATCCCACATCC-3′1866048NM_013149RV: 5’-TCGCGTCCTTCTTCATCCGTTAGC-3’Cyp1a1FW: 5′-GTCCCGGATGTGGCCCTTCTCAAA-3′109 bp56NM_012540RV: 5′-TAACTCTTCCCTGGATGCCTTCAA-3′Ido1Fw: 5′-GACTTCGTGGATCCAGAC-3′277 bp4849NM_02397.1RV: 5′-TCTAAGGAGGAGAGGAAG-3′GapdhFW: 5′-GCCAAGGTCATCCATGACAAC-3′600 bp55NM_017008.4RV: 5′-GTCCACCACCCTGTTGCTGTA-3′
## Protein expression analysis by ELISA methods
ELISA kits ZellBio GmbH (Germany) were used for determining AHR (Cat. No: ZB-16349C-R), CYP1A1 (Cat. No: ZB-11059C-R9648), and IDO1 (Cat. No: ZB-10730C-R9648) levels in the heart tissue lysate following the manufacturer’s protocol.
## Statistical analysis
For data analysis, we used multivariate analysis of variance (MANOVA), which was previously reported as an efficient test for assessing multiple dependent variables simultaneously25. Data are presented as means ± standard deviation in the text and tables. The Pearson and partial correlation coefficients were used to measure the relationship between MDA, KYN, and protein levels. A one-way MANOVA was first performed to determine the effect of the factor ‘group’ on the latent variable. Since the result of the multivariate analysis was significant, univariate analyses were done to discover the effect of a significant factor on each indicator variable (AHR, CYP1A1, IDO1, KYN, and MDA) of the latent variable, which was followed by the Tukey’s post hoc test. Assumptions of analysis of variance were verified with the Shapiro–Wilk test, Levene’s test, Doornik-Hansen test, Box's M-test, and evaluation of the homogeneity of covariance matrices. The results were considered significant with P ≤ 0.05. All statistical computations were done using Stata version 16 (College Station, TX: Stata Corp LLC; 2019) and GraphPad Prism version 9.0.0 for Windows, GraphPad Software, San Diego, California USA, www.graphpad.com.
## Ethics approval
All experimental protocols were approved by the Regional Research Ethics Committee of Semnan University of Medical Sciences and Health Services (IR.SEMUMS.REC.1399.291) and were carried out following the Declaration of Helsinki, the ARRIVE guidelines and the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health.
## Informed consent
Informed consent was obtained from animal’s owner involved in the study.
## Effect of MI and exercise on components of Ido1-KYN-Ahr axis
The results in Fig. 1 depict that MI increased the level of KYN, but this increase was not statistically significant compared to the Ct group. Both training protocols significantly reduced the level of KYN in the heart tissues of rats with MI ($P \leq 0.05$) and the effect of HIIT was greater than MICT (Table 3B). Also, MICT significantly reduced KYN in healthy rats ($$P \leq 0.003$$), which in general shows that exercise reduces KYN in heart tissue (Fig. 1A). MI significantly increased the level of MDA in the heart tissue of rats with MI ($P \leq 0.011$) and both HIIT ($P \leq 0.001$) and MICT ($$P \leq 0.001$$) significantly reduced its level. Figure 1KYN (A), and MDA (B) levels in heart tissue of rats. MDA and KYN levels of the heart tissue were affected by the MI condition. Training had different effects on the treated groups. Abbreviations: Ct control group, MI the group with occlusion of the left coronary artery, MICT the group treated with moderate intensity of continuous training, MIMCT rats with left coronary artery occlusion treated with moderate intensity of continuous training, MIHIIT rats with left coronary artery occlusion treated with high-intensity interval training. Table 3Relationships between variables and parameter estimates.(A) The relationships between variables in liver tissue by Pearson and Partial correlationVariablesCorrelationAHRIDO1P-valueCYP1A1P-valueKYNP-valueMDAP-valueCoefCoefCoefCoefAHRr10.62* < 0.0010.70 < 0.0010.530.0030.65 < 0.001rp0.140.4980.350.075− 0.200.3110.360.066IDO1r10.81 < 0.0010.79 < 0.0010.67 < 0.001rp0.470.0130.490.010− 0.080.678CYP1A1r10.71 < 0.0010.69 <.001rp0.090.6680.110.587KYNr10.78 < 0.001rp0.550.003MDAr1rp(B) Parameter estimates of variables in the five groupsGroupControlAHRIDO1CYP1A1KYNMDAMIHIITβ− 0.07− 10.33− 1.77− 5.92− 7.83P-value0.0430.007 < 0.0010.0010.021MIMCTβ− 0.07− 2.75− 0.20− 3.50− 2.83P-value0.0270.4400.4390.0430.381MICTβ− 0.15− 10.50− 1.77− 6.83− 8.17P-value < 0.0010.006 < 0.001 < 0.0010.017MIβ0.0411.670.671.8311.67P-value0.2350.0030.0150.2740.001r Pearson correlation, rp Partial correlation, Coef Coefficient, β Parameter estimates.
The results in Fig. 2, indicated that the MI increased the mRNA levels of Ahr, Ido1, and Cyp1a1 compared to the control group, but this increase was not statistically significant. Also, the changes in proteins level were not significant except for IDO1, and only levels of this protein significantly ($P \leq 0.027$) increased in the MI group in comparison to Ct rats (Fig. 2F). Both training protocols significantly reduced the levels of AHR, IDO, and CYP1A1 proteins in MIHIIT and MIMCT groups, compared with the MI group ($P \leq 0.001$), in which HIIT training had a greater effect than MICT. In healthy male rats, only protein expression of AHR significantly ($P \leq 0.05$) decreased in the MICT group compared to the control group (Fig. 2D). Both HIIT and MICT training protocols significantly reduced the Cyp1a1 and Ido1 expressions and CYP1A1 and IDO1 proteins levels in heart tissue, and a greater effect of HIIT was observed (Fig. 2B,C,E,F). Figure 2Ahr (A), Ido1 (B), and Cyp1a1 (C) gene expression, and AHR (D), IDO1 (E), and CYP1A1 (F) protein levels in the heart tissue of rats. Gene expressions and protein levels were increased in the MI condition and training showed decreasing effects in them. Abbreviations: Ct control group, MI the group with occlusion of the left coronary artery, MICT the group treated with moderate intensity of continuous training, MIMCT rats with left coronary artery occlusion treated with moderate intensity of continuous training, MIHIIT rats with left coronary artery occlusion treated with high-intensity interval training.
## Effects of interventional factors on variables
Box’s M-test showed the observed covariance matrices of the dependent variables are equal across groups ($$P \leq 0.064$$). The significant effect of the group on the latent variables (AHR, CYP1A1, KYN, IDO1, and MDA simultaneously) was observed (Wilks’ λ = 0.027, Partial Eta Square = 0.60, $P \leq 0.001$). Considering the significant effect on the latent variable, individual univariate analysis was done and showed that there is a significant effect of group on variables as AHR (F5,42 = 10.32, $P \leq 0.001$), CYP1A1 (F5,42 = 37.71, $P \leq 0.001$), KYN (F5,42 = 10.36, $P \leq 0.001$), IDO1 (F5,42 = 13.52, $P \leq 0.001$) and MDA (F5,42 = 12.70, $P \leq 0.001$).
According to Table 3A, based on the r coefficient there was a highly positive significant correlation between the studied factors. Also, in the bivariate analysis, there was a moderate positive correlation between factors (based on the rp coefficient), however, it was significant just for some of them (Table 3A). The results of the Sidak post hoc test and pairwise comparison of groups are presented in Figs. 1 and 2.
Table 3B presented the parameter estimates of variables that the control group was considered as a reference and other groups were compared to it. This test showed that MI resulted in the elevation of all variables and HIIT and MICT significantly reduced these factors levels and only in the CYP1A1 variable, MICT couldn’t exert a significant effect. It seems that between the two types of exercise, HIIT had better performance (Table 3B).
## Discussion
The evidence has confirmed that exercise training can improve cardiac function in heart diseases such as MI and improve the quality of life in patients. Although for the early exercise rehabilitation following MI, resistance training and MICT have been proposed as beneficial choices18 supervised HIIT was reported to result in greater improvements in MI patients16. A randomized controlled study found that aerobic interval training increases peak oxygen uptake more than usual exercise training in myocardial infarction patients19. It has been shown that aerobic exercise with moderate intensity could improve physical capacity and other cardiovascular variables26. However, HIIT has shown a relatively low rate of major adverse cardiovascular events for patients with coronary artery disease or heart failure when applied within cardiac rehabilitation settings17. Exercise intensity is an important factor for reversing left ventricular remodeling and improving aerobic capacity, endothelial function, and quality of life in patients with postinfarction heart failure27.
The results of the present study showed that both training protocols significantly reduced the level of KYN in the heart tissue of rats with MI, however, the effect of HIIT was greater than MICT (Fig. 1A). KYN is a new and valuable biomarker of chronic heart failure, with the ability to predict mortality and reflect exercise capacity28. Impairment of heart rhythm and observations of myocardial cell failure induced by KYN have been reported earlier and elevated plasma levels of KP metabolites and their ratios are associated with increased mortality, independent of coronary artery disease, in patients with heart failure10,29. Considering that KYN metabolites may increase inflammation, oxidative stress, and apoptosis of smooth muscle cells and endothelial cells it has been suggested that alterations of tryptophan metabolism might have an impact on the bioenergetic activities of heart mitochondria and might be involved in the development of clinical symptoms such as cardiomyopathy8,30,31. KYN generation through IDO is markedly induced after MI and genetic deletion or pharmacological inhibition of IDO limits cardiac injury and cardiac dysfunction after MI9, and here we showed that exercise reduces KYN levels in heart tissue significantly.
Exercise-induced KP modulates energy homeostasis and may contribute to the prevention and treatment of chronic diseases. Physical exercise has been shown to impact the KP in response to both acute and chronic exercise training stimuli12. Physical exercise can modulate KP metabolism in skeletal muscle and thus change the concentrations of select compounds in peripheral tissues and the central nervous system14. The effects of KP metabolites on skeletal muscle, adipose tissue, the immune system, and the brain suggest that some of these compounds could qualify as exercise-induced myokines14. Endurance exercise training can change KP metabolites by changing the levels of KP enzymes in skeletal muscle. This leads to a metabolite pattern that favors energy expenditure and an anti-inflammatory immune cell profile and reduces toxic metabolites32.
MI significantly increases MDA levels in heart tissue and here we showed that MICT and HIIT significantly reduced the levels (Fig. 1B). MDA is a known oxidative stress marker for coronary artery disease severity and plaque sensitivity33 and its reduction is considered to be an indicator of the healing of myocardial ischemia injuries34.
Also, the AHR pathway has been reported to exert cardioprotective effects against cardiotoxicity and produce heart-specific transcriptional responses35,36. KYN induces cardiomyocyte apoptosis through reactive oxygen species production in an AHR–dependent mechanism9. The expression level of circulating AHR may affect the susceptibility and progression of coronary arterial disease37 and it participates in myocardial ischemia–reperfusion injury by regulating mitochondrial apoptosis38. The results of the present study showed that OLAD resulted in an increase of *Ahr* gene expression in the heart tissue of rats with MI and MICT decreased it but none of these changes were statistically significant. However, at the protein level, both training protocols significantly reduced the level of AHR in the heart tissue of rats with OLAD. Additionally, even in healthy controls, MICT reduced the level of AHR protein in cardiac tissue. *In* general, both training protocols were able to reduce the AHR protein level in heart tissue. Previous studies have reported that in the myocardial ischemia model with OLAD, AHR is abundantly expressed in necrotic myocardium and it is shown that acute myocardial ischemia can activate AHR and induce inflammation39. Also, some animal studies reported a decrease in AHR after improvement of cardiac condition40. It seems that some AHR ligands, such as BaiCalin1, could reduce myocardial necrosis and inflammation by inhibiting the cardiac expression of AHR38,41. The heart and its vasculature system express all AHR-regulated genes and cardiac AHR-regulated CYPs are involved in the pathogenesis of cardiovascular diseases42. *Increased* gene expression of Cyp1a1 with MI was not statistically significant. However, in comparison to MICT HIIT reduces Cyp1a1 expression significantly. Regarding protein levels, both exercise protocols significantly reduce CYP1A1, although the effect of HIIT was greater (Fig. 2C,F). Parameter Estimation in Table 3 shows that among protein expressions HIIT had the greatest effect and only in CYP1A1, MICT training was most effective. It has been clearly shown that cardiac AHR-regulated CYPs are involved in the pathogenesis of cardiovascular diseases42.
In the present study, MI increased the level of *Ido1* gene expression but it was not statistically significant. Both types of HIIT and MICT significantly reduced the *Ido1* gene expression in heart tissue, however, HIIT had a greater effect and the same changes were observed for IDO1 protein levels (Fig. 2B and E). IDO1 and the IDO1-associated pathway constitute critical mediating agents associated with immunoinflammatory responses such as atherosclerosis in the heart tissue43. IDO1 promotes cardiomyocyte hypertrophy partially via PI3K-AKT-mTOR-S6K1 signaling44 and its suppression could potentially reduce the inflammatory response in cardiomyocyte injury45. IDO1 was inversely associated with ischemic heart disease with a directionally consistent estimate for stroke and might be a potential therapeutic target for this disease11. Based on the results in Table 3A, the protein expression of the studied variables is significantly correlated with each other. Also, it is shown that these variables are significantly related and affected by each other. A decrease in AHR led to a decrease in IDO1 and IDO1 had a significant relationship with KYN and CYP1A1 (Table 3A).
It is reported previously that HIIT was more effective than moderate-intensity training for improving oxygen pulse (O2P) slope in coronary heart disease patients, while ventilation and carbon dioxide production (VE/VCO2) slope and oxygen uptake efficiency slope were similarly improved by aerobic training regimens versus controls46. A systematic review and meta-analysis reported that HIIT is superior to MICT in improving cardiorespiratory fitness in participants of cardiac rehabilitation47. Our results are somehow consistent with the finding of these studies, however, there were some limitations. Measuring the slopes related to oxygen physiology, analyzing the activity of the enzymes involved in the oxidative/antioxidative system in heart tissue, assessing other components of the Ido1-Kyn-*Ahr axis* and immunohistochemical studies could add highly valuable data to future studies on this topic.
In conclusion, based on this study, it is possible to conclude that myocardial infarction alters the Ido1-Kyn-*Ahr axis* in heart tissue cells and imbalances it, as well as triggering oxidative stress. Both high-intensity interval training and moderate-intensity continuous training were effective at reducing the levels of the axis components and HIIT had a more significant effect. The intensity of exercise appears to be a prominent factor in ameliorating this molecular axis in cells of the infarcted heart.
## References
1. Contessotto P, Pandit A. **Therapies to prevent post-infarction remodelling: From repair to regeneration**. *Biomaterials* (2021.0) **275** 120906. DOI: 10.1016/j.biomaterials.2021.120906
2. Melhem NJ, Taleb S. **Tryptophan: From diet to cardiovascular diseases**. *Int. J. Mol. Sci.* (2021.0) **22** 9904. DOI: 10.3390/ijms22189904
3. Wirthgen E, Hoeflich A, Rebl A, Günther J. **Kynurenic acid: The Janus-faced role of an immunomodulatory tryptophan metabolite and its link to pathological conditions**. *Front. Immunol.* (2018.0) **8** 1957. DOI: 10.3389/fimmu.2017.01957
4. Seok SH. **Trace derivatives of kynurenine potently activate the aryl hydrocarbon receptor (AHR)**. *J. Biol. Chem.* (2018.0) **293** 1994-2005. DOI: 10.1074/jbc.RA117.000631
5. Rothhammer V, Quintana FJ. **The aryl hydrocarbon receptor: An environmental sensor integrating immune responses in health and disease**. *Nat. Rev. Immunol.* (2019.0) **19** 184-197. DOI: 10.1038/s41577-019-0125-8
6. Kou Z, Dai W. **Aryl hydrocarbon receptor: Its roles in physiology**. *Biochem. Pharmacol.* (2021.0) **185** 114428. DOI: 10.1016/j.bcp.2021.114428
7. Vasquez A. **A role for the aryl hydrocarbon receptor in cardiac physiology and function as demonstrated by AhR knockout mice**. *Cardiovasc. Toxicol.* (2003.0) **3** 153-163. DOI: 10.1385/CT:3:2:153
8. Metghalchi S. **Indoleamine 2 3-dioxygenase knockout limits angiotensin II-induced aneurysm in low density lipoprotein receptor-deficient mice fed with high fat diet**. *PLoS One* (2018.0) **13** e0193737. DOI: 10.1371/journal.pone.0193737
9. Melhem NJ. **Endothelial cell Indoleamine 2, 3-dioxygenase 1 alters cardiac function after myocardial infarction through kynurenine**. *Circulation* (2021.0). DOI: 10.1161/CIRCULATIONAHA.120.050301
10. Lund A. **Plasma kynurenines and prognosis in patients with heart failure**. *PLoS ONE* (2020.0) **15** 1-15
11. Li M, Kwok MK, Fong SSM, Schooling CM. **Indoleamine 2,3-dioxygenase and ischemic heart disease: A Mendelian randomization study**. *Sci. Rep.* (2019.0). DOI: 10.1038/s41598-019-44819-7
12. Joisten N, Walzik D, Metcalfe AJ, Bloch W, Zimmer P. **Physical exercise as kynurenine pathway modulator in chronic diseases: Implications for immune and energy homeostasis**. *Int. J. Tryptophan Res.* (2020.0) **13** 117864692093868. DOI: 10.1177/1178646920938688
13. Agudelo LZ. **Erratum: Skeletal muscle PGC-1α1 modulates kynurenine metabolism and mediates resilience to stress-induced depression (Cell (2014) 159 (33–45))**. *Cell* (2015.0) **160** 351. DOI: 10.1016/j.cell.2014.12.025
14. Martin KS, Azzolini M, Ruas JL. **The kynurenine connection: How exercise shifts muscle tryptophan metabolism and affects energy homeostasis, the immune system, and the brain**. *Am. J. Physiol. Cell Physiol.* (2020.0) **318** C818-C830. DOI: 10.1152/ajpcell.00580.2019
15. Pal A. **Different endurance exercises modulate NK cell cytotoxic and inhibiting receptors**. *Eur. J. Appl. Physiol.* (2021.0) **121** 3379-3387. DOI: 10.1007/s00421-021-04735-z
16. Dun Y. **High-intensity interval training improves metabolic syndrome and body composition in outpatient cardiac rehabilitation patients with myocardial infarction**. *Cardiovasc. Diabetol.* (2019.0) **18** 1-11. DOI: 10.1186/s12933-019-0907-0
17. Wewege MA, Ahn D, Yu J, Liou K, Keech A. **High-intensity interval training for patients with cardiovascular disease-is it safe? A systematic review**. *J. Am. Heart Assoc.* (2018.0) **7** e009305. DOI: 10.1161/JAHA.118.009305
18. Cai M. **Effects of different types of exercise on skeletal muscle atrophy, antioxidant capacity and growth factors expression following myocardial infarction**. *Life Sci.* (2018.0) **213** 40-49. DOI: 10.1016/j.lfs.2018.10.015
19. Moholdt T. **Aerobic interval training increases peak oxygen uptake more than usual care exercise training in myocardial infarction patients: A randomized controlled study**. *Clin. Rehabil.* (2012.0) **26** 33-44. DOI: 10.1177/0269215511405229
20. 20.Joisten, N. et al. Exercise Diminishes Plasma Neurofilament Light Chain and Reroutes the Kynurenine Pathway in Multiple Sclerosis. Neurol. Neuroimmunol. neuroinflammation8, (2021).
21. Charan J, Biswas T. **How to calculate sample size for different study designs in medical research?**. *Indian J. Psychol. Med.* (2013.0) **35** 121-126. DOI: 10.4103/0253-7176.116232
22. Høydal MA, Wisløff U, Kemi OJ, Ellingsen Ø. **Running speed and maximal oxygen uptake in rats and mice: Practical implications for exercise training**. *Eur. J. Prev. Cardiol.* (2007.0) **14** 753-760. DOI: 10.1097/HJR.0b013e3281eacef1
23. 23.Wang, B. et al. Effect of high-intensity interval training on cardiac structure and function in rats with acute myocardial infarct. Biomed. Pharmacother.131, (2020).
24. Khalafi M. **The impact of moderate-intensity continuous or high-intensity interval training on adipogenesis and browning of subcutaneous adipose tissue in obese male rats**. *Nutrients* (2020.0) **12** 925. DOI: 10.3390/nu12040925
25. Gilani N, Haghshenas R, Esmaeili M. **Application of multivariate longitudinal models in SIRT6, FBS, and BMI analysis of the elderly**. *Aging Male* (2019.0) **22** 260-265. DOI: 10.1080/13685538.2018.1477933
26. Izeli NL. **Aerobic training after myocardial infarction: Remodeling evaluated by cardiac magnetic resonance**. *Arq. Bras. Cardiol.* (2016.0) **106** 311-318. PMID: 26959403
27. Wisløff U. **Superior cardiovascular effect of aerobic interval training versus moderate continuous training in heart failure patients: A randomized study**. *Circulation* (2007.0) **115** 3086-3094. DOI: 10.1161/CIRCULATIONAHA.106.675041
28. Dschietzig TB. **Plasma kynurenine predicts severity and complications of heart failure and associates with established biochemical and clinical markers of disease**. *Kidney Blood Press. Res.* (2019.0) **44** 765-776. DOI: 10.1159/000501483
29. Rudzite V. **Bradyarrythmias and myocardial cell failure induced By kynurenine**. *Prog. Tryptophan Serotonin Res.* (2019.0) **1986** 131-136. DOI: 10.1515/9783110854657-038
30. Wang Q. **Activation of NAD(P)H oxidase by Tryptophan-derived 3-Hydroxykynurenine accelerates endothelial apoptosis and dysfunction in Vivo**. *Circ. Res.* (2014.0) **114** 480-492. DOI: 10.1161/CIRCRESAHA.114.302113
31. Baran H. **Effects of various kynurenine metabolites on respiratory parameters of rat brain, liver and heart mitochondria**. *Int. J. Tryptophan Res.* (2016.0) **9** 17-29. PMID: 27226722
32. Valente-Silva P. **Effects of tryptophan supplementation and exercise on the fate of kynurenine metabolites in mice and humans**. *Metabolites* (2021.0) **11** 508. DOI: 10.3390/metabo11080508
33. Matsuo Y. **Circulating malondialdehyde-modified low-density lipoprotein levels are associated with the presence of thin-cap fibroatheromas determined by optical coherence tomography in coronary artery disease**. *Eur. Heart J. Cardiovasc. Imaging* (2013.0) **14** 43-50. DOI: 10.1093/ehjci/jes094
34. 34.Liu, C. et al. Protective Effects and Mechanisms of Recombinant Human Glutathione Peroxidase 4 on Isoproterenol-Induced Myocardial Ischemia Injury. Oxid. Med. Cell. Longev.2021, (2021).
35. Volkova M, Palmeri M, Russell KS, Russell RR. **Activation of the aryl hydrocarbon receptor by doxorubicin mediates cytoprotective effects in the heart**. *Cardiovasc. Res.* (2011.0) **90** 305-314. DOI: 10.1093/cvr/cvr007
36. Carney SA. **Aryl hydrocarbon receptor activation produces heart-specific transcriptional and toxic responses in developing zebrafish**. *Mol. Pharmacol.* (2006.0) **70** 549-561. DOI: 10.1124/mol.106.025304
37. Huang S. **AhR expression and polymorphisms are associated with risk of coronary arterial disease in Chinese population**. *Sci. Rep.* (2015.0) **5** 8022. DOI: 10.1038/srep08022
38. Wang B, Xu A. **Aryl hydrocarbon receptor pathway participates in myocardial ischemia reperfusion injury by regulating mitochondrial apoptosis**. *Med. Hypotheses* (2019.0) **123** 2-5. DOI: 10.1016/j.mehy.2018.12.004
39. Yi T. **Aryl hydrocarbon receptor: A new player of pathogenesis and therapy in cardiovascular diseases**. *Biomed. Res. Int.* (2018.0) **2018** 1-11
40. Vilahur G. **Reperfusion-triggered stress protein response in the myocardium is blocked by post-conditioning. Systems biology pathway analysis highlights the key role of the canonical aryl-hydrocarbon receptor pathway**. *Eur. Heart J.* (2013.0) **34** 2082-2093. DOI: 10.1093/eurheartj/ehs211
41. Xue Y. **Baicalin inhibits inflammation and attenuates myocardial ischaemic injury by aryl hydrocarbon receptor**. *J. Pharm. Pharmacol.* (2015.0) **67** 1756-1764. DOI: 10.1111/jphp.12484
42. Korashy H, El-Kadi A. **The role of aryl hydrocarbon receptor in the pathogenesis of cardiovascular diseases**. *Drug Metab. Rev.* (2006.0) **38** 411-450. DOI: 10.1080/03602530600632063
43. Liu G, Chen S, Zhong J, Teng K, Yin Y. **Crosstalk between tryptophan metabolism and cardiovascular disease, mechanisms, and therapeutic implications**. *Oxid. Med. Cell. Longev.* (2017.0) **2017** 1-5
44. Liu Y. **Indoleamine 2,3-dioxygenase 1 (IDO1) promotes cardiac hypertrophy via a PI3K-AKT-mTOR-dependent mechanism**. *Cardiovasc. Toxicol.* (2021.0) **21** 655-668. DOI: 10.1007/s12012-021-09657-y
45. Guo G, Sun L, Yang L, Xu H. **IDO1 depletion induces an anti-inflammatory response in macrophages in mice with chronic viral myocarditis**. *Cell Cycle* (2019.0) **18** 2598-2613. DOI: 10.1080/15384101.2019.1652471
46. Cardozo GG, Oliveira RB, Farinatti PTV. **Effects of high intensity interval versus moderate continuous training on markers of ventilatory and cardiac efficiency in coronary heart disease patients**. *Sci. World J.* (2015.0) **2015** 1-8. DOI: 10.1155/2015/192479
47. Hannan A. **High-intensity interval training versus moderate-intensity continuous training within cardiac rehabilitation: A systematic review and meta-analysis**. *Open Access J. Sport. Med.* (2018.0) **9** 1-17. DOI: 10.2147/OAJSM.S150596
48. Brauze D, Widerak M, Cwykiel J, Szyfter K, Baer-Dubowska W. **The effect of aryl hydrocarbon receptor ligands on the expression of AhR, AhRR, ARNT, Hif1α, CYP1A1 and NQO1 genes in rat liver**. *Toxicol. Lett.* (2006.0) **167** 212-220. DOI: 10.1016/j.toxlet.2006.09.010
49. Kanth VR, Lavanya K, Srinivas J, Naga Raju T. **Elevated expression of indoleamine 2,3-dioxygenase (IDO) and accumulation of kynurenic acid in the pathogenesis of STZ-induced diabetic cataract in wistar rats**. *Curr. Eye Res.* (2009.0) **34** 274-281. DOI: 10.1080/02713680902725954
|
---
title: 'Transcriptomic analysis of the cerebral hippocampal tissue in spontaneously
hypertensive rats exposed to acute hypobaric hypoxia: associations with inflammation
and energy metabolism'
authors:
- Wei Chang
- Jinxiu Cui
- Yajuan Li
- Kehai Zang
- Xutao Zhang
- Zhuoru Zhang
- Yihong Jiang
- Qianqian Ma
- Shuai Qu
- Fengzhou Liu
- Junhui Xue
journal: Scientific Reports
year: 2023
pmcid: PMC9988845
doi: 10.1038/s41598-023-30682-0
license: CC BY 4.0
---
# Transcriptomic analysis of the cerebral hippocampal tissue in spontaneously hypertensive rats exposed to acute hypobaric hypoxia: associations with inflammation and energy metabolism
## Abstract
We evaluated the effect of acute hypobaric hypoxia (AHH) on the hippocampal region of the brain in early-stage spontaneously hypertensive male rats. The rats were classified into a control (ground level; ~ 400 m altitude) group and an AHH experimental group placed in an animal hypobaric chamber at a simulated altitude of 5500 m for 24 h. RNA-*Seq analysis* of the brains and hippocampi showed that differentially expressed genes (DEGs) were primarily associated with ossification, fibrillar collagen trimer, and platelet-derived growth factor binding. The DEGs were classified into functional categories including general function prediction, translation, ribosomal structure and biogenesis, replication, recombination, and repair. Pathway enrichment analysis revealed that the DEGs were primarily associated with relaxin signaling, PI3K-Akt signaling, and amoebiasis pathways. Protein–protein interaction network analysis indicated that 48 DEGs were involved in both inflammation and energy metabolism. Further, we performed validation experiments to show that nine DEGs were closely associated with inflammation and energy metabolism, of which two (Vegfa and Angpt2) and seven (Acta2, Nfkbia, Col1a1, Edn1, Itga1, Ngfr, and Sgk1) genes showed up and downregulated expression, respectively. Collectively, these results indicated that inflammation and energy metabolism-associated gene expression in the hippocampus was altered in early-stage hypertension upon AHH exposure.
## Introduction
Studies have shown the prevalence of cerebrovascular lesions and cognitive impairment in early-stage hypertensive patients and animal models1–3. Abnormalities in vascular structure and function, including endothelial dysfunction, increased oxidative stress and vascular remodeling, and decreased compliance are considered symptoms of early-stage hypertension. These phenotypes play an important role in the development of hypertension4. Furthermore, varying degrees of brain damage are common under conditions involving acute hypobaric hypoxic (AHH)5. Upon first arriving at high altitudes, significant changes are observed in the cerebral hemodynamics of individuals that live at low altitudes. As the degree of hypobaric hypoxia increases, arterial vasodilatation gradually increases, including that in the middle cerebral artery. In turn, pathological changes such as increased intracapillary pressure, loss of autoregulatory function, and abnormal neural and humoral regulation are observed6. Inflammation is associated with the susceptibility to and development of high-altitude cerebral edema. Hypoxia enhances lipopolysaccharide-induced inflammation and mediates the onset and development of cerebral edema in mice at high altitudes. This phenotype can be attributed to the disruption of blood–brain barrier integrity and activation of the microglia7. Hypobaric hypoxia results in abnormal alterations in the energy metabolism of the body, including changes in various branched-chain amino acids, succinate, lactate, and pyruvate8–10. Compared with healthy individuals, those with hypertension exhibit a significantly higher probability of developing acute altitude sickness when entering high-altitude regions11. AHH conditions, in addition to hypertension, may lead to an increase in the severity of cerebral damage. However, the underlying pathophysiological mechanisms remain unclear.
Transcriptome analysis is widely used in the study of biological processes (physiological mechanisms, pathways, or genes) associated with cardiovascular and cerebrovascular diseases. RNA-sequencing (RNA-seq) and subsequent bioinformatics analysis have been used to investigate the potential molecular targets associated with energy metabolism under conditions involving hypertension during diabetes. The results have led to the discovery of novel therapeutic targets12. A Japanese study performed transcriptomic characterization of samples from individuals under conditions involving AHH. The transcriptional profile of individuals underwent rapid changes under conditions involving acute hypoxic, and these changes may affect individual adaptation to the hypoxic environment13.
To investigate the effects of AHH on the hippocampus of early-stage hypertensive brains, we used six-week-old male spontaneously hypertensive rats (SHRs) in our study. SHR rats are the most commonly used animal model to study the physiological mechanism underlying hypertension. The blood pressure of SHR rats began to increase from the fourth to the sixth week14. Since sex plays a crucial role in the development of hypertension, and the incidence of early hypertension is significantly higher in men than in women, we considered only male animals in the selection of study subjects15. A large number of neural cells exist in the hippocampal region, and their function is closely associated with cognition. Moreover, the hippocampal region has a rich blood supply. The normality of vascular function has a remarkable influence on the role of the hippocampal region. Therefore, we exposed six-week-old SHRs to a hypobaric hypoxic environment at 5500 m or a control ground-level environment (~ 400 m above sea level) for 24 h. Subsequently, we anesthetized the rats and isolated the hippocampal region of the brain for high-throughput RNA-seq. Based on the sequencing results, inflammation-associated genes and energy metabolism-associated genes were screened, and a protein–protein interaction (PPI) network was constructed. Furthermore, the data were validated using quantitative polymerase chain reaction (qPCR). This study aimed to investigate the effects of AHH on the hippocampus in the brain during early-stage hypertension. Furthermore, we aimed to validate the role of inflammation and energy metabolism via transcriptome sequencing and PPI network construction.
## Animals and blood pressure measurements
All experiments were performed using six-week-old male SHRs (systolic blood pressure: 151.5 ± 3.5 mmHg; diastolic blood pressure: 109.6 ± 6.0 mmHg) purchased from the Animal Experiment Center of Air Force Medical University, China (Supplementary Table 1). The systolic blood pressure of the caudal artery was measured using a BP-2010A automatic non-invasive blood pressure meter (Softron Biotechnology Co., Ltd, Beijing, China) at 8:00 am in the rats during the resting state. Five measurements were obtained for each rat, and the mean of the five readings was considered as the systolic blood pressure of the rat. All animals (license lot number SCXK 2017–0021) were considered eligible for the experiment. During the experiment, animals were housed at 21 ± 1 °C with ad libitum access to food and water. All animal experiments and operations were performed under the guidance of the Animal Research Committee and approved by the Air Force Medical University Animal Ethics Committee of the Institute (No. IACUC-20220392). All animal feeding and experimental procedures followed the guidelines of the institutional ethics review board of the Air Force Medical University and strictly performed according to the ARRIVE guidelines16.
## RNA-seq
Six male SHRs were equally and randomly classified into the ground-level (~ 400 m above sea level) control or the AHH experimental groups. During the experiment, the rats in the control group were fed the same food and water as those in the AHH group. The rats in the AHH group were placed in an animal hypobaric pressure chamber (Department of Aerospace Medicine of Air Force Medical University and Hongyuan Oxygen Industry Co., Ltd., Xi'an, China), elevated to 5500 m at 10–15 m/s, and returned to ground level at a rate of 15–20 m/s after 24 h. The rats in both groups were euthanized with an overdose of $1.5\%$ sodium pentobarbital administered via intraperitoneal injection, and the brains were harvested.
The hippocampus of the brain was isolated, and total RNA was extracted from the tissue using TRIzol (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. Subsequently, total RNA was quantified and characterized using a Nanodrop spectrophotometer and an Agilent 2100 Bioanalyzer (Thermo Fisher Scientific, Waltham, MA, USA), respectively (28S/18S > 1.0, RIN > 7.0). Hieff NGS® DNA Selection Beads (Yeasen, Shanghai, China) were used for mRNA purification. The purified mRNA was fragmented into small fragments using a fragmentation buffer at an appropriate temperature. First-strand cDNA was generated using random hexamer-initiated reverse transcription, followed by second-strand cDNA synthesis, according to the kit instructions (Takara Bio Inc., Beijing, China). End repair was performed by incubation with an A-tail mixture and an RNA index adapter (Yeasen, Shanghai, China). The cDNA fragment obtained was amplified using qPCR, purified with AMPure XP beads, and eluted with an elution buffer solution. The PCR products were checked for quality using an Agilent Technologies 2100 Bioanalyzer. The double-stranded PCR products obtained were denatured by heating and cyclized with a splint oligo sequence to obtain the final library. Single-stranded circular DNA was used as the final library. DNA nanoballs (DNBs) were amplified using phi29; one molecule had more than 300 copies. The DNBs were loaded into patterned nanoarrays to generate 150 base-pair end reads on the DNBSEQ-T7 platform (Tsingke Biotechnology Co. Ltd., Beijing, China).
## Quality control and identification
The raw reads obtained from the sequencing data were filtered to obtain clean reads and compared to the reference genome of *Rattus norvegicus* mRatBN7.2($\frac{7}{7}$/2022). Six samples were sequenced using the Illumina platform, yielding 44.09 GB of data. The percentage of bases with quantitative values higher than 30 (Q30) was ≥ $94.24\%$. The clean reads of each sample were aligned relative to the designated reference genome separately, and the matching efficiency ranged from 96.66 to $97.27\%$. The sequencing data were of good quality and met the requirements for subsequent analysis (Table 1). After preprocessing and filtering, gene expression levels were normalized to transcript per kilobase million (TPM) values. Table 1Statistical analysis of transcriptome sequencing of the hippocampal tissue of spontaneous hypertensive rats (SHRs) exposed to acute hypobaric hypoxia (AHH) for 24 h in the AHH group compared to those in the control group. Sample NameClean ReadsClean BasesClean Reads Q30 (%)Control-24 h-121,095,0346,307,222,11694.66Control-24 h-222,356,5676,683,650,60694.47Control-24 h-331,972,1149,556,379,68095.79AHH-24 h-119,693,5415,885,567,86095.12AHH-24 h-223,146,5566,903,908,26496.50AHH-24 h-329,285,7978,754,402,34294.24Sample name: sample name of the sample information sheet. Clean reads: total number of paired-end reads in the clean data. Clean bases: total number of bases in the clean data number. Clean reads Q30: the percentage of bases with clean data quality value higher than or equal to 30.
## GO and KEGG enrichment analysis of differentially expressed genes (DEGs)
Gene expression was calculated for each sample as TPM values, and the DEseq2 R package (version: 4.1.3) was used for analyzing DEGs with a screening threshold of ∣log2(fold change)∣ ≥ 1.2, $p \leq 0.0517.$ The functional enrichment analysis of DEG was performed using the clusterProfiler package based on the Gene Ontology (GO) (http://www.geneontology.org) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases (http://www.genome.jp/kegg/)17. The conditional threshold for GO and KEGG analysis was $p \leq 0.05.$
## Screening of inflammation and energy metabolism-associated genes and PPI network construction
Based on previous studies18, genes associated with inflammation were retrieved by searching the GeneCards database (https://www.genecards.org/). A similar search strategy was employed for energy metabolism-associated genes. Based on this, we obtained DEGs associated with both inflammation and energy metabolism. These DEGs were entered into the STRING database (https://string-db.org/) and analyzed, and a PPI network was constructed. The network file was downloaded and imported into Cytoscape (version 3.9.1) and visualized based on the degree of the topological properties of this network19.
## Reverse transcription qPCR validation
Total RNA was extracted using TRIzol (DBI Bioscience, Shanghai, China). cDNA was obtained using a reverse transcription kit (DBI Bioscience, Shanghai, China) and amplified using SYBR Green PCR Master Mix (DBI Bioscience, Shanghai, China) according to the manufacturer’s instructions. β-actin was used to normalize gene expression, and the relative gene expression values were calculated using the 2–∆∆Ct method20. Primer sequences were retrieved from PrimerBank 40 (Supplementary Table 2), and primer specificity was further validated using the National Center for Biotechnology Information Primer-BLAST tool.
## Statistical analysis of GO, KEGG enrichment, and qPCR results
Based on the GO (http://www.geneontology.org) and KEGG databases (http://www.genome.jp/kegg/), the functions of DEGs can be discerned, and the associated pathways can be enriched using the clusterProfiler package of R. The cutoff of GO enrichment analysis was set to $p \leq 0.05.$ Significant KEGG pathways were screened based on $p \leq 0.05.$ *The data* of reverse transcription qPCR were statistically analyzed using GraphPad Prism 8.0 software. The data are expressed as the mean ± standard deviation. Comparisons between groups were performed using the independent sample t-test, and differences were considered statistically significant at $p \leq 0.05$, $$n = 6$.$
## Analysis of DEGs
Boxplots for the distribution of DEGs across all six data sets are shown in Fig. 1A. The density plots show that the two groups (control-24 h and AHH-24 h) each had three replicates with a similar distribution of the six curves (Fig. 1B). In addition, the cluster heatmap shows differences in the gene expression profiles between the AHH and control groups (Fig. 1C). To determine the differential expression of these genes in the hypobaric hypoxic environment, we analyzed changes between the two groups using volcano plots (Fig. 1D). DEGs were filtered based on the criterion ∣log2(fold change)∣ ≥ 1.2, $p \leq 0.05$, and 112 DEGs were subsequently obtained by comparing the hippocampal tissue between the AHH and control groups. Of these 112 genes, 25 and 87 genes showed up- and downregulated expression, respectively, in the AHH group. Figure 1(A) Box plot of log2 (TPM + 1) expression values for each sample. The horizontal coordinates in the plot represent the different samples; the vertical coordinates indicate the expression level of the sample log2 (TPM + 1) values. The plot indicates the expression level of each sample in terms of the overall dispersion of expression. Box line plots for each region correspond to six statistical parameters (upper, outlier, upper quartile, median, lower quartile, and lower limit from top to bottom). ( B) Density plot of the distribution of log2 (TPM + 1) values for each sample. Different colored curves in the graph represent different samples; the horizontal coordinates of the points on the curves indicate the log values of TPM + 1 for the corresponding samples, and the vertical coordinates of the points indicate the probability density. ( C) *Heatmap analysis* of 112 differentially expressed genes (DEGs) in the two groups: control-24 h and acute hypobaric hypoxia (AHH)-24 h groups. *All* genes shown are significantly differentially expressed between the two groups (adjusted $p \leq 0.05$). Colored bars from blue to red represent the increasing levels of gene expression from low to high. ( D) Volcano plot analysis of 112 DEGs in the control-24 h and AHH-24 h groups. Each dot represents a gene that was detected in both groups. Red and blue dots indicate genes with significantly up and downregulated expression, respectively. Gray dots show genes that were not significantly differentially expressed between the control-24 h and AHH-24 h groups. The horizontal axis indicates the log ratio (changes in gene expression ploidy across samples), and the vertical axis indicates the probability of differential expression of each gene. The DEGs were identified using a false discovery rate of 0.05.
## GO-, clusters of orthologous gene (COG)-, and KEGG-based functional classification of DEGs
A total of 112 DEGs were annotated using GO analysis and classified into three categories, i.e., biological processes, cellular components, and molecular functions (Fig. 2A and Supplementary Table 3). The upregulated DEGs were primarily enriched in the biological process and cellular component categories. In the biological process module, DEGs were primarily associated with the regulation of cell–cell adhesion, positive regulation of cell adhesion, positive regulation of cell–cell adhesion, regulation of alpha–beta T cell activation, and regulation of leukocyte cell–cell adhesion. In the cellular component module, DEGs with upregulated expression were primarily associated with the cytoplasmic side of the plasma membrane. In addition, DEGs with upregulated expression were not involved in molecular functions (Fig. 2B and Supplementary Table 4). However, DEGs with downregulated expression were highly enriched in several GO functional categories, including collagen fibril organization, axon guidance, and neuron projection guidance in the biological process module, fibrillar collagen trimer, banded collagen fibril; a complex of collagen trimers in the cellular components module; and platelet-derived growth factor binding, extracellular matrix structural constituent, and growth factor binding in the molecular functions module. Interestingly, many subcategories of DEGs showed downregulated expression (Fig. 2C and Supplementary Table 5).Figure 2(A) Gene ontology (GO) annotation analysis of 112 differentially expressed genes (DEGs) for the control-24 h and acute hypobaric hypoxia (AHH)-24 h groups. The top eight categories with the smallest p values for each classification were screened for GO annotations. BP, biological process; CC, cellular component; MF, molecular function. The bar color in the graph represents the p value distributed from blue to red; the closer to red, the smaller the p-value, and vice versa. ( B) DEGs upregulated in the AHH-24 h group compared with those in the control-24 h group subjected to GO annotation analysis. ( C) DEGs downregulated in the AHH-24 h group compared with those in the control-24 h group analyzed by GO annotation. ( D) Clusters of orthologous gene (COG) classification analysis of 112 DEGs. The vertical axis indicates the frequency of DEGs in specific functional clusters, and the horizontal axis indicates the functional class. ( E) Kyoto Encyclopedia of Genes and Genomes (KEGG) transcript classification analysis of 112 DEGs in the control-24 h and AHH-24 h groups. The left and right-colored bars represent the two groups of DEGs and the KEGG pathways, respectively. Lines connecting the KEGG pathway show the enrichment of the two sets of DEGs. The horizontal axis of the bubble plot indicates the gene ratio (the ratio of the number of DEGs enriched in the corresponding pathway to the number of all DEGs entered for enrichment analysis). The vertical axis indicates the KEGG pathway term, and the bubble size represents the number of differentially annotated genes in a term, with larger bubbles indicating more genes. The color represents the enrichment significance p value, with a higher intensity of the red color representing a smaller value (indicating stronger significance). ( F) KEGG analysis of DEGs upregulated in the AHH-24 h group compared with those in the control-24 h group. KEGG transcripts for taxonomic analysis. ( G) KEGG transcript classification analysis of the DEGs downregulated in the AHH-24 h group compared with those in the control-24 h group.
In the COG database, 112 DEGs were classified into 18 functional categories. The main category was general function prediction, followed by translation, ribosomal structure and biogenesis, replication, recombination, and repair, carbohydrate transport and metabolism, posttranslational modification, protein turnover, and chaperones (Fig. 2D and Supplementary Table 6).
KEGG pathway enrichment analysis was performed for both groups of DEGs, and the threshold for determining gene enrichment was $p \leq 0.05$ (Fig. 2E and Supplementary Table 7). In the KEGG enrichment analysis, the most enriched category was the PI3K-Akt signaling pathway, followed by neuroactive ligand-receptor interaction, relaxin signaling pathway, and human papillomavirus infection. In parallel, analysis of enriched pathways for DEGs with up- and downregulated expression was performed separately. For DEGs with upregulated expression, the most significantly enriched pathways were the Ras and PI3K-Akt signaling pathways (Fig. 2F and Supplementary Table 8). For DEGs with downregulated expression, neuroactive ligand-receptor interaction, PI3K-Akt signaling pathway, human papillomavirus infection, and relaxin signaling pathway were significantly enriched (Fig. 2G and Supplementary Table 9).
## Screening for inflammation and energy metabolism-associated genes
To elucidate the role of DEGs related to inflammation and energy metabolism, the term “energy metabolism” was searched on the GeneCards website to obtain a list of associated genes. Inflammation-associated genes were retrieved in a similar manner. The overlaps between the resultant 9730 energy metabolism-associated genes and 11,109 inflammation-associated genes were analyzed along with the 112 DEGs, yielding 48 overlapping genes (Fig. 3A). We constructed a PPI network using the STRING database to further investigate the biological roles of the DEGs associated with energy metabolism and inflammation (Fig. 3B). Among them, genes encoding vascular endothelial growth factor A (Vegfa) and actin alpha 2 (Acta2) showed a central pivotal position. Genes encoding nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor alpha (Nfkbia), collagen type I alpha 1 (Col1a1), endothelin 1 (Edn1), and serum/glucocorticoid regulated kinase 1 (Sgk1) also showed numerous associations with other genes. The ten most significant pathways were screened using KEGG enrichment analysis ($p \leq 0.05$; Fisher’s exact test, followed by the Bonferroni test). The results showed significant enrichment of the relaxin signaling and PI3K-Akt signaling pathways, among others (Fig. 3C,D, and Supplementary Table 10). The DEGs evaluated via PPI analysis indicate a vital role of signaling pathways related to energy metabolism and inflammation in the response to AHH.Figure 3(A) Venn diagram of common differentially expressed genes (DEGs) between groups. The three circles represent energy metabolism-associated genes, inflammation-associated genes, and DEGs. The intersecting regions of the circles indicate the intersecting genes of different groups. ( B) Proteins interacting with inflammation- and energy metabolism-associated genes. The protein–protein interaction network of DEGs was visualized using Cytoscape (version 3.9.1) according to its topological properties. The higher the color intensity and the lower the distance to the center, the more the number of genes it interacted with in the network. The hub sub-networks were screened by the MCODE plug-in. ( C) Relaxin signaling pathway (www.kegg.jp/entry/map04926). Red and green boxes represent DEGs that were significantly up or downregulated in this pathway, respectively. ( D) PI3K-Akt signaling pathway (www.kegg.jp/entry/map04151). Red and green boxes represent DEGs that were significantly up or downregulated in this pathway, respectively.
## Validation of RNA-seq data using qPCR
To validate the reliability of the transcriptome sequencing data obtained by *Illumina analysis* in our study, we selected nine core DEGs screened by PPI analysis for reverse transcription-qPCR using the same RNA samples. The DEGs included Vegfa, Acta2, Nfkbia, Col1a1, Edn1, Angpt2, integrin subunit alpha 1 (Itga1), nerve growth factor receptor (Ngfr), and Sgk1 (Fig. 4 and Supplementary Table 11). The results confirmed that the RNA-seq data were reliable. Figure 4Reverse transcription-quantitative polymerase chain reaction analysis results of nine core differentially expressed genes (DEGs) screened via the protein–protein interaction network. The bars represent geometric means ± standard deviations. mRNA levels of these genes differed between the ground-level control and acute hypobaric hypoxia experimental groups ($p \leq 0.05$, $$n = 6$$).
## Discussion
In this study, 112 DEGs were identified using high-throughput transcriptome sequencing analysis of samples derived from the hippocampal region of rats in the AHH experimental and control (ground-level) groups. We observed 25 upregulated and 87 down-regulated genes in the AHH group. GO, COG, and KEGG functional classification analyses were performed on the DEGs. GO functional analysis showed that the identified DEGs were primarily associated with ossification, fibrillar collagen trimer, and platelet-derived growth factor binding. Additionally, the DEGs were primarily enriched in biological process modules. In the COG functional annotation, the DEGs were classified into 18 functional categories, including general function prediction, translation, ribosomal structure and biogenesis, and replication, recombination, and repair. KEGG enrichment analysis showed that the identified DEGs were associated with the relaxin, PI3K-Akt, and amoebiasis pathways. Among the DEGs, Vegfa, Acta2, Nfkbia, Col1a1, Edn1, Angpt2, Itga1, Ngfr, and Sgk1 play vital roles in these pathways. Inflammation plays an important role in mediating hypobaric hypoxic brain injury7. Furthermore, alterations in brain energy metabolism have been demonstrated in many hypoxic environments21–24. However, the effects of AHH on the brain have not been reported in early-stage hypertensive rats, and the involvement of inflammation and energy metabolism in this process is not well understood. Therefore, we screened for inflammation- and energy metabolism-associated DEGs by constructing PPI networks. The results revealed that both groups of genes play crucial roles in early-stage hypertension during exposure to AHH.
VEGFA was originally identified as an endothelial growth factor as well as a regulator of vascular permeability. It is produced by most cells in vivo and is significantly upregulated in response to hypoxia25. VEGFA induces a series of cascade responses, such as proliferation and survival, cell migration, vascular permeability, invasion of surrounding tissues, and endothelial cell inflammation, via the activation of VEGFR226. VEGFA-induced expression of VEGFR2 is associated with multiple signaling pathways, including the phospholipase Cγ-extracellular regulated kinase and the PI3K-Akt pathways, both of which are closely associated with inflammation and energy metabolism27,28. Upregulation of VEGFA expression in the hippocampal region of SHRs under conditions involving exposure to AHH can alter the proliferation efficiency of hippocampal cells. Moreover, VEGFA, as an important vascular growth factor, affects the blood supply in the hippocampal region by altering vascular function.
Acta2 encodes α-smooth muscle actin, which is primarily expressed in the vascular smooth muscle. Alterations in this gene are associated with several vascular diseases29. Changes in Acta2 expression under conditions involving AHH may represent alterations in vasodilatory function in the hippocampus30.
NFKBIA is a member of a family of cellular proteins that inhibit NF-κB transcription factors. NFKBIA inhibits NF-κB by masking the nuclear localization signal of NF-κB and maintaining it in an inactive state in the cytoplasm31. In addition, NFKBIA can prevent NF-κB from functioning by blocking its binding to DNA32. NF-κB is a key regulator of pro-inflammatory gene expression, inducing the transcription of pro-inflammatory cytokines, chemokines, adhesion molecules, matrix metalloproteinases, cyclooxygenase 2, and inducible nitric oxide synthase33. Inhibition of NF-κB activity controls the development of inflammatory diseases34. Thus, overexpression of Nfkbia inhibits NF-κB activity and suppresses inflammatory responses in various diseases. In contrast, decreased expression of Nfkbia promotes the development of inflammatory responses35. Reduced expression of *Nfkbia is* observed in the hippocampus of SHRs under conditions involving AHH, which may result in the overactivation of inflammation-related pathways and aggravate AHH-induced cerebral damage in SHRs.
COL1A1 influences the development and prognosis of various tumors by involvement in tumor cell metastasis, proliferation, and apoptosis36–38. COL1A1 activates multiple signaling pathways (including epithelial-mesenchymal transition, tumor growth factor-beta, and PI3k/Akt pathways), enhances energy metabolism, promotes cell metastasis, and inhibits apoptosis39–41. The invasive and migratory abilities of hepatocellular carcinoma cells are significantly inhibited after the knockout of Col1a136. In addition, the expression of the cell proliferation factor cyclin D1 and the apoptosis marker BCL-2 is decreased, whereas that of the apoptosis regulator BAX is increased after the knockdown of Col1a142. In this study, AHH exposure resulted in the downregulation of Colla1 expression in the hippocampal region of SHRs, which may affect normal energy metabolism and cell proliferation in the hippocampus, thereby aggravating cerebral damage.
EDN1, a strong vasoconstrictor, is closely associated with pathophysiological changes in blood vessels, and its expression is affected under hypoxic conditions43. Prolonged hypoxic exposure leads to elevated EDN1 expression44. However, upon exposure to short-term hypobaric hypoxic conditions, the expression of EDN1 is reduced in brain neurons, astrocytes, and endothelial cells45. This response may be a protective mechanism of brain cells against early-stage hypoxia. However, the reason why EDN1 expression is altered with prolonged hypoxia remains unknown.
ANGPT2 is an important molecule involved in the process of angiogenesis and acts as a marker of inflammation46. ANGPT2 levels are low under normal physiological conditions but increase during inflammation46. ANGPT2 acts on endothelial cells, increasing endothelial permeability. It also acts on pericytes, mediating their detachment from the basement membrane and further inducing vascular leakage47.
AHH can exacerbate cerebral damage in SHRs by promoting inflammation. Itga1 encodes the integrin α1 chain, which binds to the α1 chain (ITGB1) to form a heterodimer that acts as a dual laminin/collagen receptor in neuronal and hematopoietic cells. Integrin α1 plays an important role in both fracture healing and cartilage remodeling48,49. The role played by Itga1 in the hippocampal region of SHRs under AHH conditions remains to be elucidated. NGFR is a transmembrane glycoprotein. As a nerve growth factor receptor, it is involved in the mitogen-activated protein kinase, Ras, PI3K-Akt, and the apoptosis signaling pathways in several species50–53. NGFR expression is closely associated with cell growth, proliferation, and apoptosis. Sgk1 is transcriptionally regulated by serum and glucocorticoids54. SGK1 activates several ion channels, transporter proteins, transcription factors, and enzymes50, and its expression is strongly upregulated in a variety of cardiovascular diseases, which are closely associated with vascular calcification55,56. Downregulation of Sgk1 alleviates inflammation via inhibition of the NF-κB signaling pathway57. However, Sgk1 knockdown reduces the potency of protective mechanisms associated with hypoxia/reoxygenation injury in cardiomyocytes58. Therefore, the effects attributed to the downregulation of Sgk1 expression in AHH should be further investigated in SHRs.
Our study provides the first demonstration of the significant influence of AHH exposure on gene expression changes in the hippocampal region during early spontaneous hypertension in rats. Furthermore, our results showed that energy metabolism and inflammation play important roles in early-stage hypertension under conditions involving AHH.
This study has few limitations. First, the experimental conditions employed were limited. We used six-week-old SHRs exposed to AHH to examine the changes in gene expression in the hippocampal region of rats in the early stages of hypertension. The next step would be to select SHRs of different ages to investigate the gene expression changes in the hippocampal region during intermediate and late-stage hypertension upon AHH exposure. Alternatively, Wistar Kyoto rats can be used as experimental animals to exclude the effect of AHH exposure in normal rats. Furthermore, we have performed RNA-seq and qPCR validation of selected genes to determine the alterations in the expression of related genes in the hippocampal region of the early-stage SHR model under AHH exposure. Follow-up studies are necessary to compare the effects of AHH exposure on hypertensive brain injury in different periods by increasing the sample size and further experimental grouping. Additionally, the experimental analyses should be enriched for a more in-depth study of the alterations in the associated genes.
## Supplementary Information
Supplementary Information 1.Supplementary Information 2.Supplementary Information 3.Supplementary Information 4.Supplementary Information 5.Supplementary Information 6.Supplementary Information 7.Supplementary Information 8.Supplementary Information 9.Supplementary Information 10.Supplementary Information 11. The online version contains supplementary material available at 10.1038/s41598-023-30682-0.
## References
1. Chen X, Wen W, Anstey KJ, Sachdev PS. **Prevalence, incidence, and risk factors of lacunar infarcts in a community sample**. *Neurology* (2009) **73** 266-272. DOI: 10.1212/WNL.0b013e3181aa52ea
2. Suvila K. **Early-but not late-onset hypertension is related to midlife cognitive function**. *Hypertension* (2021) **77** 972-979. DOI: 10.1161/HYPERTENSIONAHA.120.16556
3. Li Y. **MRI study of cerebral blood flow, vascular reactivity, and vascular coupling in systemic hypertension**. *Brain Res.* (2021) **1753** 147224. DOI: 10.1016/j.brainres.2020.147224
4. Feihl F, Liaudet L, Levy BI, Waeber B. **Hypertension and microvascular remodelling**. *Cardiovasc. Res.* (2008) **78** 274-285. DOI: 10.1093/cvr/cvn022
5. Hackett PH, Roach RC. **High-altitude illness**. *N. Engl. J. Med.* (2001) **345** 107-114. DOI: 10.1056/NEJM200107123450206
6. Ainslie PN, Subudhi AW. **Cerebral blood flow at high altitude**. *High Alt. Med. Biol.* (2014) **15** 133-140. DOI: 10.1089/ham.2013.1138
7. Zhou Y. **Hypoxia augments LPS-induced inflammation and triggers high altitude cerebral edema in mice**. *Brain Behav. Immun.* (2017) **64** 266-275. DOI: 10.1016/j.bbi.2017.04.013
8. Xie H, Xu G, Aa J, Gu S, Gao Y. **Modulation of perturbed cardiac metabolism in rats Under high-altitude hypoxia by combination treatment With L-carnitine and trimetazidine**. *Front. Physiol.* (2021) **12** 671161. DOI: 10.3389/fphys.2021.671161
9. Xu G. **DL-3-n-butylphthalide improved physical and learning and memory performance of rodents exposed to acute and chronic hypobaric hypoxia**. *Mil. Med. Res.* (2021) **8** 23. PMID: 33766114
10. Li J. **Acute high-altitude hypoxic brain injury: Identification of ten differential proteins**. *Neural Regen. Res.* (2013) **8** 2932-2941. PMID: 25206614
11. Rimoldi SF. **High-altitude exposure in patients with cardiovascular disease: Risk assessment and practical recommendations**. *Prog. Cardiovasc. Dis.* (2010) **52** 512-524. DOI: 10.1016/j.pcad.2010.03.005
12. Pauza AG. **GLP1R attenuates sympathetic response to high glucose via carotid body inhibition**. *Circ. Res.* (2022) **130** 694-707. DOI: 10.1161/CIRCRESAHA.121.319874
13. Yasukochi Y, Shin S, Wakabayashi H, Maeda T. **Transcriptomic changes in Young Japanese Males After exposure to acute hypobaric hypoxia**. *Front. Genet.* (2020) **11** 559074. DOI: 10.3389/fgene.2020.559074
14. Warshaw DM, Mulvany MJ, Halpern W. **Mechanical and morphological properties of arterial resistance vessels in young and old spontaneously hypertensive rats**. *Circ Res.* (1979) **45** 250-259. DOI: 10.1161/01.RES.45.2.250
15. Colafella KMM, Denton KM. **Sex-specific differences in hypertension and associated cardiovascular disease**. *Nat. Rev. Nephrol.* (2018) **14** 185-201. DOI: 10.1038/nrneph.2017.189
16. Lilley E. **ARRIVE 2.0 and the British Journal of Pharmacology: Updated Guidance for 2020**. *Brit. J. Pharmacol.* (2020) **177** 3611-3616. DOI: 10.1111/bph.15178
17. 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
18. Liu Y. **Prognostic implications of autophagy-associated gene signatures in non-small cell lung cancer**. *Aging (Albany NY)* (2019) **11** 11440-11462. DOI: 10.18632/aging.102544
19. Shannon P. **Cytoscape: A software environment for integrated models of biomolecular interaction networks**. *Genome Res.* (2003) **13** 2498-2504. DOI: 10.1101/gr.1239303
20. 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. DOI: 10.1006/meth.2001.1262
21. Sun S, Guo Z, Fu H, Zhu J, Ge X. **Integrated metabolomic and transcriptomic analysis of brain energy metabolism in the male Oriental river prawn (Macrobrachium nipponense) in response to hypoxia and reoxygenation**. *Environ. Pollut.* (2018) **243** 1154-1165. DOI: 10.1016/j.envpol.2018.09.072
22. Wang X. **Rhodiola crenulata attenuates apoptosis and mitochondrial energy metabolism disorder in rats with hypobaric hypoxia-induced brain injury by regulating the HIF-1alpha/microRNA 210/ISCU1/2(COX10) signaling pathway**. *J. Ethnopharmacol.* (2019) **241** 111801. DOI: 10.1016/j.jep.2019.03.028
23. Farhat E, Devereaux MEM, Pamenter ME, Weber JM. **Naked mole-rats suppress energy metabolism and modulate membrane cholesterol in chronic hypoxia**. *Am. J. Physiol. Regul. Integr. Comp. Physiol.* (2020) **319** R148-R155. DOI: 10.1152/ajpregu.00057.2020
24. Kolesnikova EE, Soldatov AA, Golovina IV, Sysoeva IV, Sysoev AA. **Effect of acute hypoxia on the brain energy metabolism of the scorpion fish Scorpaena porcus Linnaeus, 1758: The pattern of oxidoreductase activity and adenylate system**. *Fish Physiol. Biochem.* (2022) **48** 1105-1115. DOI: 10.1007/s10695-022-01103-2
25. Ferrara N. **Vascular endothelial growth factor: Basic science and clinical progress**. *Endocr. Rev.* (2004) **25** 581-611. DOI: 10.1210/er.2003-0027
26. Claesson-Welsh L, Welsh M. **VEGFA and tumour angiogenesis**. *J. Intern. Med.* (2013) **273** 114-127. DOI: 10.1111/joim.12019
27. Jiang BH, Liu LZ. **PI3K/PTEN signaling in angiogenesis and tumorigenesis**. *Adv. Cancer Res.* (2009) **102** 19-65. DOI: 10.1016/S0065-230X(09)02002-8
28. Zhang S, Lachance BB, Mattson MP, Jia X. **Glucose metabolic crosstalk and regulation in brain function and diseases**. *Prog. Neurobiol.* (2021) **204** 102089. DOI: 10.1016/j.pneurobio.2021.102089
29. Yuan SM. **Alpha-Smooth muscle actin and ACTA2 gene expressions in vasculopathies**. *Braz. J. Cardiovasc. Surg.* (2015) **30** 644-649. PMID: 26934405
30. Hallbäck M, Weiss L. **Mechanisms of spontaneous hypertension in rats**. *Med. Clin. North Am.* (1977) **61** 593-609. DOI: 10.1016/S0025-7125(16)31319-0
31. Crépieux P. **I kappaB alpha physically interacts with a cytoskeleton-associated protein through its signal response domain**. *Mol. Cell. Biol.* (1997) **17** 7375-7385. DOI: 10.1128/MCB.17.12.7375
32. Prigent M, Barlat I, Langen H, Dargemont C. **IkappaBalpha and IkappaBalpha/NF-kappa B complexes are retained in the cytoplasm through interaction with a novel partner, rasgap SH3-binding protein 2(J)**. *J. Biol. Chem.* (2000) **275** 36441-36449. DOI: 10.1074/jbc.M004751200
33. Baeuerle PA, Baichwal VR. **NF-kappaB as a frequent target for immunosuppressive and anti-inflammatory molecules**. *Adv. Immunol.* (1997) **65** 111-137. DOI: 10.1016/S0065-2776(08)60742-7
34. Tak PP, Firestein GS. **NF-kappaB: A key role in inflammatory diseases**. *J. Clin. Invest.* (2001) **107** 7-11. DOI: 10.1172/JCI11830
35. Bondeson J, Foxwell B, Brennan F, Feldmann M. **Defining therapeutic targets by using adenovirus: blocking NF-kappaB inhibits both inflammatory and destructive mechanisms in rheumatoid synovium but spares anti-inflammatory mediators**. *Proc. Natl Acad. Sci. U. S. A.* (1999) **96** 5668-5673. DOI: 10.1073/pnas.96.10.5668
36. Ma HP. **Collagen 1A1 (COL1A1) is a reliable biomarker and putative therapeutic target for hepatocellular carcinogenesis and metastasis**. *Cancers (Basel)* (2019) **11** 786. DOI: 10.3390/cancers11060786
37. Zheng Z, Chen Y, Wang Y, Li Y, Cheng Q. **MicroRNA-513b-5p targets COL1A1 and COL1A2 associated with the formation and rupture of intracranial aneurysm**. *Sci Rep. Sci. Rep.* (2021) **11** 14897. DOI: 10.1038/s41598-021-94116-5
38. Zhang C. **COL1A1 is a potential prognostic biomarker and correlated with immune infiltration in mesothelioma**. *BioMed Res. Int.* (2021) **2021** 5320941. PMID: 33490271
39. Wang Q. **CircCSPP1 functions as a ceRNA to promote colorectal carcinoma cell EMT and liver metastasis by upregulating COL1A1**. *Front. Oncol.* (2020) **10** 850. DOI: 10.3389/fonc.2020.00850
40. Guo Y. **miR-133b suppresses invasion and migration of gastric cancer cells via the COL1A1/TGF-beta axis**. *Onco. Targets Ther.* (2020) **13** 7985-7995. DOI: 10.2147/OTT.S249667
41. Li M. **Microenvironment remodeled by tumor and stromal cells elevates fibroblast-derived COL1A1 and facilitates ovarian cancer metastasis**. *Exp. Cell Res.* (2020) **394** 112153. DOI: 10.1016/j.yexcr.2020.112153
42. Zhao W, Jiang X, Yang S. **lncRNA TUG1 promotes cell proliferation, migration, and invasion in hepatocellular carcinoma via regulating miR-29c-3p/ COL1A1 axis**. *Cancer Manag. Res.* (2020) **12** 6837-6847. DOI: 10.2147/CMAR.S256624
43. Yamashita K, Discher DJ, Hu J, Bishopric NH, Webster KA. **Molecular regulation of the endothelin-1 gene by hypoxia. Contributions of hypoxia-inducible factor-1, activator protein-1, GATA-2, AND p300/CBP**. *J. Biol. Chem.* (2001) **276** 12645-12653. DOI: 10.1074/jbc.M011344200
44. Jiang S. **Testosterone attenuates hypoxia-induced hypertension by affecting NRF1-mediated transcriptional regulation of ET -1 and ACE**. *Hypertens. Res.* (2021) **44** 1395-1405. DOI: 10.1038/s41440-021-00703-4
45. Kanazawa F. **Expression of endothelin-1 in the brain and lung of rats exposed to permanent hypobaric hypoxia**. *Brain Res.* (2005) **1036** 145-154. DOI: 10.1016/j.brainres.2004.12.019
46. Wu Q, Xu WD, Huang AF. **Role of angiopoietin-2 in inflammatory autoimmune diseases: A comprehensive review**. *Int. Immunopharmacol.* (2020) **80** 106223. DOI: 10.1016/j.intimp.2020.106223
47. Akwii RG, Sajib MS, Zahra FT, Mikelis CM. **Role of angiopoietin-2 in vascular physiology and pathophysiology**. *Cells* (2019) **8** 471-1785. DOI: 10.3390/cells8050471
48. Ekholm E. **Diminished callus size and cartilage synthesis in alpha 1 beta 1 integrin-deficient mice during bone fracture healing**. *Am. J. Pathol.* (2002) **160** 1779-1785. DOI: 10.1016/S0002-9440(10)61124-8
49. Li WF. **Genetics of osteoporosis: Accelerating pace in gene identification and validation**. *Hum. Genet.* (2010) **127** 249-285. DOI: 10.1007/s00439-009-0773-z
50. Liang L. **CD271(+) cells are diagnostic and prognostic and exhibit elevated MAPK activity in SHH medulloblastoma**. *Cancer Res.* (2018) **78** 4745-4759. DOI: 10.1158/0008-5472.CAN-18-0027
51. Elshaer SL. **Modulation of the p75 neurotrophin receptor using LM11A-31 prevents diabetes-induced retinal vascular permeability in mice via inhibition of inflammation and the RhoA kinase pathway**. *Diabetologia* (2019) **62** 1488-1500. DOI: 10.1007/s00125-019-4885-2
52. Wu R, Li K, Yuan M, Luo KQ. **Nerve growth factor receptor increases the tumor growth and metastatic potential of triple-negative breast cancer cells**. *Oncogene* (2021) **40** 2165-2181. DOI: 10.1038/s41388-021-01691-y
53. Li J. **Resveratrol induces autophagy and apoptosis in non-small-cell lung cancer cells by activating the NGFR-AMPK-mTOR pathway**. *Nutrients* (2022) **14** 2413. DOI: 10.3390/nu14122413
54. Firestone GL, Giampaolo JR, O'Keeffe BA. **Stimulus-dependent regulation of serum and glucocorticoid inducible protein kinase (SGK) transcription, subcellular localization and enzymatic activity**. *Cell. Physiol. Biochem.* (2003) **13** 1-12. DOI: 10.1159/000070244
55. Lang F. **(Patho)physiological significance of the serum- and glucocorticoid-inducible kinase isoforms**. *Physiol. Rev.* (2006) **86** 1151-1178. DOI: 10.1152/physrev.00050.2005
56. Lang F, Voelkl J. **Therapeutic potential of serum and glucocorticoid inducible kinase inhibition**. *Expert Opin. Investig. Drugs* (2013) **22** 701-714. DOI: 10.1517/13543784.2013.778971
57. Voelkl J. **SGK1 induces vascular smooth muscle cell calcification through NF-kappaB signaling**. *J. Clin. Invest.* (2018) **128** 3024-3040. DOI: 10.1172/JCI96477
58. Cong B. **SGK1 is involved in cardioprotection of urocortin-1 against hypoxia/reoxygenation in cardiomyocytes**. *Can. J. Cardiol.* (2014) **30** 687-695. DOI: 10.1016/j.cjca.2014.03.024
|
---
title: 'The role of SMAD signaling in hypertrophic obstructive cardiomyopathy: an
immunohistopathological study in pediatric and adult patients'
authors:
- Zhengjie Zhang
- Fengzhi Zhang
- Mingkui Zhang
- Hui Xue
- Lixin Fan
- Yan Weng
journal: Scientific Reports
year: 2023
pmcid: PMC9988847
doi: 10.1038/s41598-023-30776-9
license: CC BY 4.0
---
# The role of SMAD signaling in hypertrophic obstructive cardiomyopathy: an immunohistopathological study in pediatric and adult patients
## Abstract
Hypertrophic obstructive cardiomyopathy (HOCM) can bring a high risk of sudden cardiac death in young people. It is particularly urgent to understand the development and mechanism of HOCM to prevent unsafe incidents. Here, the comparison between pediatric and adult patients with HOCM has been performed to uncover the signaling mechanism regulating pathological process through histopathological analysis and immunohistochemical analysis. We found SMAD proteins played an important role during myocardial fibrosis for HOCM patients. In patients with HOCM, Masson and HE staining showed that myocardial cells were diffusely hypertrophied with obvious disorganized myocardial fiber alignment, and myocardial tissue was more damaged and collagen fibers increased significantly, which come early in childhood. Increased expressions of SMAD2 and SMAD3 contributed to myocardial fibrosis in patients with HOCM, which happened early in childhood and continued through adulthood. In addition, decreased expression of SMAD7 was closely related to collagen deposition, which negatively expedited fibrotic responses in patients with HOCM. Our study indicated that the abnormal regulation of SMAD signaling pathway can lead to severe myocardial fibrosis in childhood and its fibrogenic effects persist into adulthood, which is a crucial factor in causing sudden cardiac death and heart failure in HOCM patients.
## Introduction
Sudden cardiac death and progressive heart failure are considered to be the most significant complications of hypertrophic obstructive cardiomyopathy (HOCM), while myocardial fibrosis is the main cause of lethal arrhythmias and heart failure in patients with HOCM1. Some studies confirmed the high burden of myocardial fibrosis in both young patients who suffered sudden cardiac death2 or older patients with HOCM suffering from advanced heart failure3. Our previous histopathological study demonstrated that pediatric patients with HOCM may present with severe myocardial fibrosis and reduced microvascular density4.
The transforming growth factor-β (TGF-β) superfamily proteins play a critical role in regulation of cardiac fibrotic responses, which is mediated via intracellular effectors (the Smads), or via activation of Smad-independent cascades5,6. The members of the Smad family are considered the best characterized intracellular effectors of TGF-β, which can be divided into three functional classes: (a). Receptor-activated Smads (R-Smads): including Smad1, Smad2, Smad3, Smad5 and Smad8, responsible for the TGF-β superfamily signaling pathway; (b). the co-mediator Smads (Co-Smads), the Smad4 forms a signaling complex with R-Smads; (c). the inhibitory Smads (I-Smads): Smad6, Smad7, involved in the negative regulation of the R-Smad-mediated cascades7. Extensive experimental evidences suggest fibroblast-specific deletion of Smad$\frac{2}{3}$ from cardiac fibroblasts prevented the gene program for fibrosis and extracellular matrix (ECM) remodeling8. Activation of Smad$\frac{2}{3}$ cascade is critical in ECM gene expression and regulation of fibrous tissue deposition9. Smad3 is a key activating signal for cardiac fibroblast to function at inducing myofibroblast conversion, stimulating the transcription of extracellular matrix proteins, and promoting a matrix-preserving phenotype10. In vitro studies also suggest that Smad2 plays a significant role in activation of a fibrogenic transcriptional program7. Smad2 knockdown was proved to inhibit incorporation of α-SMA into myofibroblast stress fibers9. In contrast, Smad3 appears more important than Smad2 in mediating fibroblast activation and fibrosis in vivo8, Smad7 can negatively feedback block the activation of Smad2/Smad3, so the Smad7 has a protective role in MF and ventricular remodeling11. Aberrant Smad family expression can occur in a variety of cardiac pathophysiological states. Fibroblasts, vascular cells and cardiomyocytes, the major cellular effector cells of cardiac fibrosis, are highly responsive to TGF-β and play a critical role in regulating the fibrotic process by activating Smad-dependent signaling pathways.
In this study, we analyzed the expression levels of SMAD proteins in septal myocardial tissue samples from pediatric and adult patients with HOCM after Morrow procedure using immunohistopathological methods, focusing on the effect of SMAD signaling on myocardial fibrosis to further elucidate the evolution of myocardial fibrosis in patients HOCM.
## Study population
Pediatric and adult patients with hypertrophic obstructive cardiomyopathy who underwent transaortic septal myocardial resection (Morrow procedure) were selected from March 2004 to December 2018 at the heart center of the First Hospital of Tsinghua University. The control group was obtained from ventricular septal myocardial tissue of pediatric and adult autopsy individuals who died of non-cardiac disease. Diagnostic criteria for hypertrophic obstructive cardiomyopathy: left ventricular wall thickness, or septal thickness ≥ 15 mm, left ventricular outflow tract pressure difference at rest > 30 mmHg, or > 50 mmHg at provocation, and can exclude other causes of ventricular hypertrophy. In pediatric patients, the increase in LV wall thickness exceeded the mean LV wall thickness plus 2 standard deviations (or Z value > 2) for children of the same age, sex, and body surface area, and other conditions causing increased cardiac load were excluded. Exclusion criteria: coronary artery disease (> $50\%$ stenosis on coronary angiography), uncontrollable hypertension (defined as blood pressure above $\frac{140}{90}$ mmHg), valvular heart disease, infection, or renal insufficiency. Twelve adult and five pediatric patients with hypertrophic obstructive cardiomyopathy were included in this study. Control myocardium from the LV septal wall was collected at autopsy of five adults and four pediatric individuals, who died of non-cardiac causes. Clinical information was obtained from the hospital electronic medical record system (Table 1).Table 1Baseline characteristics of adult and pediatric patients with HOCM.VariableAdult with HOCM ($$n = 12$$)Pediatric with HOCM ($$n = 5$$)Age (years)47.4 ± 7.99.30 ± 4.00Male, n (%)11 ($92\%$)2($40\%$)BMI (kg/m2)26.22 ± 4.1618.60 ± 4.64Symptoms Chest pain5 ($41.67\%$)1 ($20\%$) Syncope2 ($16.67\%$) Dyspnea8 ($66.67\%$)1 ($20\%$)Echocardiography LVOT PG at rest(mmHg)96.8 ± 44.189.4 ± 29.4 Septal wall thickness (mm)25.7 ± 9.416.2 ± 1.2 LV end-diastolic diameter (mm)47.67 ± 7.4622.60 ± 4.28 LVEF (%)63.48 ± 5.2972.80 ± 7.12 Family history of HCM n (%)2 ($16.67\%$)Medications, n (%) β-Blockers8 ($66.67\%$)1 ($20\%$) Calcium channel blockers3 ($25\%$)
## Echocardiography
Standard transthoracic M-mode, 2-dimensional, pulsed or continuous wave doppler images were acquired using a Philips IE33 color Doppler system (Philips Healthcare, Andover, MA, USA). Parameter acquisition was performed according to the American Society of Echocardiography guidelines12. The detection of peak velocity and differential pressure across the left ventricular outflow tract was calculated using the simplified Bernoulli equation. Left ventricular outflow tract obstruction was defined as hypertrophic cardiomyopathy diagnosed as left ventricular thickness or septal thickness ≥ 15 mm or left ventricular outflow tract differential pressure > 30 mmHg at rest or > 50 mmHg at provocation13.
## Histological analysis
The septal myocardium samples were fixed in $10\%$ formalin, embedded in paraffin. The samples were cut into 5 μm sections and stained with hematoxylin and eosin (H&E) to analyze the morphology. The samples were also stained with Masson trichrome to analyze the extent of myocardial fibrosis14.
## Immunohistochemical staining
Septal myocardial tissue specimens were fixed in $10\%$ buffered formalin, embedded in paraffin, and cut into 5um thick tissue sections for immunohistochemical staining. After deparaffinization and rehydration with graded concentrations of ethanol to deionized water, the sections were heated in EDTA antigen retrieval buffer (pH9.0, ZSGB-BIO) in a pressure cooker for 2.5 min for antigen retrieval. Endogenous peroxidase activity was bloked with $3\%$ (vol/vol) H2O2. Then, the sections were incubated with $10\%$ goat serum for 1 h at room temperature, followed by different primary antibodies at 4 °C overnight: anti-Smad2 (diluted 1:20, Abcam), or anti-Smad3 (diluted 1:500, Abcam), or anti-MADH7/SMAD7 (diluted 1:100, Abcam), or anti-Collagen I (diluted 1:1000, Abcam). The sections were subjected to horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (diluted 1:1000, Abcam) for 1 h at room temperature and then visualized with 3,3-diaminobenzidine (DAB) solution (ZSGB-BIO). Finally, the sections were counterstained using hematoxylin.
Images were captured using Motic EasyScan 60 and analyzed with Image -pro plus software (Image Pro Plus 6.0; Media Cybernetics Inc. Media Cybernetics, Inc; 4340 East–West Hwy, Suite 400 Bethesda, MD; USA). Ten areas per section were randomly selected, and the images were magnified × 40 magnification, and the positive expression areas were brown in color. Image-Pro Plus 6.0 image analysis software was used to calculate the integrated optical density (IOD) value, IOD indicates the protein expression of each index, the higher the IOD value, the higher the protein expression.
## Statistical methods
Statistical data were analyzed using SPSS 25.0 (SPSS Inc., Chicago, IL, USA). Data were expressed as mean ± standard deviation. Continuous variables were compared between the two groups using an independent t-test. p values less than 0.05 were used as a test for significant differences.
## Ethics declarations
The study was supported by the Ethics Committee of the First Hospital of Tsinghua University, and informed consent was obtained from patients or legal guardians. All experiments were performed in accordance with relevant guidelines and regulations.
## Clinical baseline characteristics of patients with HOCM
Demographic and clinical baseline characteristics of the study population are presented in Table 1. The mean age of the 12 adult HOCM patients and 5 pediatric HOCM patients was 41.5 ± 15.4 years and 9.30 ± 4.00 years, respectively. The left ventricular outflow tract PG max (mmHg) was 96.8 ± 44.1 mmHg in the adult HOCM group and 89.4 ± 29.4 mmHg in the pediatric HOCM group.
## Changes in ventricular septal myocardial tissue and cell morphology in patients with HOCM
HOCM is a genetic disorder characterized by thickened septum between the ventricles, which is usually accompanied by dynamic left ventricular outflow tract obstruction. First, we depicted the changes of cell gross morphology, muscle fiber arrangement, and organization of the architecture in the hearts with HOCM using histological analysis. HE staining showed that in the adult control and pediatric control groups, myocardial cells were structurally intact and well-arranged; however, in the adult HOCM and pediatric HOCM groups, myocardial cells were diffusely hypertrophied with obvious disorganized myocardial fiber alignment (Fig. 1a). Masson staining showed that myocardial fibrotic areas were blue and myocardial cells were red (Fig. 1b): myocardial tissue was neatly arranged and evenly stained with fewer collagen fibers in adult control and pediatric control groups; while myocardial tissue was more damaged and collagen fibers increased significantly in adult HOCM and pediatric HOCM patients. Figure 1Morphological examinations of myocardial tissue from patients with HOCM. ( a) Representative images of haematoxylin and eosin (H&E) staining of ventricular septal myocardial tissue from adult HOCM and pediatric HOCM patients, compared to their controls. Scale bars, 400 μm and 200 μm (enlarged views of boxed areas). ( b) Representative images of Masson’s Trichrome staining of heart sections from adult HOCM and pediatric HOCM patients, compared to their controls. Scale bars, 400 μm and 200 μm (enlarged views of boxed areas).
Therefore, in patients with HOCM, myocardial cells were diffusely hypertrophied with obvious disorganized myocardial fiber alignment, and myocardial tissue was more damaged and collagen fibers increased significantly, which come early in childhood.
## Increased SMAD2 and SMAD3 expressions in ventricular septal myocardial tissue from patients with HOCM
Given that increased fibrosis appeared in HOCM hearts, and TGF-β-Smad$\frac{2}{3}$ signaling is considered as principal mediators of the fibrotic response, Smad2 and Smad3 even serving as their major downstream effectors15, we hypothesized that expressions of SMAD2 and SMAD3 might be upregulated in patients with HOCM. Using immunohistochemical staining, we found expressions of SMAD2 and SMAD3 in myocardial tissue of adult HOCM patients were significantly higher than those of their controls (1014.00 ± 290.10 vs. 230.50 ± 60.91 $$p \leq 0.0001$$ Fig. 2a,b; 846.80 ± 274.30 vs. 162.40 ± 102.50 $$p \leq 0.0003$$ Fig. 3a,b). Similarly, expressions of SMAD2 and SMAD3 in myocardial tissue of pediatric with HOCM were significantly higher than those of their controls (1197.00 ± 383.80 vs. 229.00 ± 155.10 $$p \leq 0.0008$$ Fig. 2c,d; 867.30 ± 217.50 vs. 301.30 ± 158.31 $$p \leq 0.0015$$ Fig. 3c,d). In addition, there was no significant difference in expressions of SMAD2 and SMAD3 in myocardial tissues between adult HOCM patients and pediatric HOCM patients (Fig. 6a,b).Figure 2Immunohistochemical analysis of SMAD2 expression in ventricular septal myocardial tissue from patients with HOCM. ( a) Representative images of immunohistochemical staining of SMAD2 in ventricular septal myocardial tissue from the control and adult HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). b. Quantification of SMAD2 expression level in ventricular septal myocardial tissue from the control and adult HOCM patients. Data are mean ± SEM, ***$p \leq 0.001.$ c. Representative images of immunohistochemical staining of SMAD2 in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). d. Quantification of SMAD2 expression level in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Data are mean ± SEM, ***$p \leq 0.001.$Figure 3Immunohistochemical analysis of SMAD3 expression in ventricular septal myocardial tissue from patients with HOCM. ( a) Representative images of immunohistochemical staining of SMAD3 in ventricular septal myocardial tissue from the control and adult HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). ( b) Quantification of SMAD3 expression level in ventricular septal myocardial tissue from the control and adult HOCM patients. Data are mean ± SEM, ***$p \leq 0.001.$ c. Representative images of immunohistochemical staining of SMAD3 in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). ( d) Quantification of SMAD3 expression level in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Data are mean ± SEM, **$p \leq 0.01.$
Thus, increased expressions of SMAD2 and SMAD3 contributed to myocardial fibrosis in patients with HOCM, which happened early in childhood and continued through adulthood.
## Decreased SMAD7 expression and collagen I deposition in ventricular septal myocardial tissue from patients with HOCM
Smad7 is an important negative regulator of TGF-β/Smad signaling that may inhibit cardiac fibrosis, whose overexpression in cardiac fibroblasts restrains collagen synthesis16. In adult patients with HOCM, SMAD7 expression was remarkably reduced in ventricular septal myocardial tissue when compared to the control group (164.70 ± 76.11 vs. 316.70 ± 119.20 $$p \leq 0.0110$$ Fig. 4a,b), likewise, SMAD7 expression in myocardial tissue of pediatric patients with HOCM was obviously lower than that of their controls (195.10 ± 43.37 vs. 534.80 ± 195.00 $$p \leq 0.0052$$ Fig. 4c,d). In addition, there was no significant difference in the expression of Smad7 in myocardial tissues between adult HOCM patients and pediatric HOCM patients (Fig. 6c).Figure 4Immunohistochemical analysis of SMAD7 expression in ventricular septal myocardial tissue from patients with HOCM. ( a) Representative images of immunohistochemical staining of SMAD7 in ventricular septal myocardial tissue from the control and adult HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). ( b) Quantification of SMAD7 expression level in ventricular septal myocardial tissue from the control and adult HOCM patients. Data are mean ± SEM, *$p \leq 0.05.$ ( c) Representative images of immunohistochemical staining of SMAD7 in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). ( d) Quantification of SMAD7 expression level in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Data are mean ± SEM, **$p \leq 0.01.$
Collagen is a major component of the myocardium17. Fibrosis is typically characterized by an accumulation of fibrillar collagens, especially of collagen type I18. The expression of type I collagen in myocardial tissues of adult HOCM patients was significantly increased when compared with the control group (139.80 ± 63.76 vs. 7.26 ± 5.26 $$p \leq 0.0039$$ Fig. 5a,b), and its expression in myocardial tissue of pediatric with HOCM dramatically higher than that of their controls (210.80 ± 74.92 vs. 20.41 ± 5.42 $$p \leq 0.0053$$ Fig. 5c,d). There was no significant difference in the expression of type I collagen in myocardial tissues between adult HOCM patients and pediatric HOCM patients (Fig. 6d).Figure 5Immunohistochemical analysis of type I collagen expression in ventricular septal myocardial tissue from patients with HOCM. ( a) Representative images of immunohistochemical staining of type I collagen in ventricular septal myocardial tissue from the control and adult HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). ( b) Quantification of type I collagen expression level in ventricular septal myocardial tissue from the control and adult HOCM patients. Data are mean ± SEM, **$p \leq 0.01.$ ( c) Representative images of immunohistochemical staining of type I collagen in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Scale bars, 200 μm and 50 μm (enlarged views of boxed areas). ( d) Quantification of type I collagen expression level in ventricular septal myocardial tissue from the control and pediatric HOCM patients. Data are mean ± SEM, **$p \leq 0.01.$Figure 6Comparison of expression levels of SMAD2, 3, 7, and type I collagen in adult and pediatric HOCM patients. ( a) Comparison of the expression level of SMAD2 in adult and pediatric HOCM patients. Data are mean ± SEM, ns, non-significant, ($p \leq 0.05$). ( b) Comparison of the expression level of SMAD3 in adult and pediatric HOCM patients. Data are mean ± SEM, ns, non-significant, ($p \leq 0.05$). ( c) Comparison of the expression level of SMAD7 in adult and pediatric HOCM patients. Data are mean ± SEM, ns, non-significant, ($p \leq 0.05$). ( d) Comparison of the expression level of type I collagen in adult and pediatric HOCM patients. Data are mean ± SEM, ns, non-significant, ($p \leq 0.05$).
Overall, besides of positive regulation to myocardial fibrosis by increased expression of SMAD2 and SMAD3, decreased expression of SMAD7 was closely related to collagen deposition, which negatively expedited fibrotic responses in patients with HOCM. As for HOCM, SMAD signaling may start as early as childhood.
## Discussion
This study demonstrated that significantly increased expression levels of SMAD$\frac{2}{3}$ and sharply reduced expression levels of SMAD7 occurred in ventricular septal myocardial tissue samples from pediatric and adult patients with HOCM who underwent Morrow surgery, while there were no significant differences between pediatric and adult patients. It indicates that abnormal regulation of SMAD signaling in HOCM patients may contribute to the progressive exacerbation of myocardial fibrosis and cause severe cardiac failure.
Smad proteins are the main downstream effector molecules in the TGF-β signaling pathway. Under certain pathophysiological conditions, the expression of TGF-β and its superfamily members is enhanced in myocardial tissue, which triggers a heterotetrameric complex formation through binding to their receptors, TGF-β receptor I (TβRI) or TGF-β receptor II (TβRII). Activated TβRI recruits and phosphorylates downstream receptor-regulated Smad proteins (Smad2 and Smad3), and activated Smad2 and Smad3 bind to Smad4 to form Smad complexes and enter the nucleus, initiating the expression of fibrosis-related transcription factors and proteins7. Clinical investigations suggested TGF-β1 expression were upregulated in myocardial tissue of patients with congenital hypertrophic cardiomyopathy19. Circulating TGF-β levels were elevated and associated with adverse cardiac events in patients with hypertrophic cardiomyopathy20. Smad activation in fibrotic cardiac condition are considered as the results of induction of TGF-β superfamily members7. *Previous* genetic-based studies demonstrated that canonical Smad2 and Smad3 signaling cascade were more dedicated to control fibrosis-mediating genes in activated fibroblast-specific Tgfbr$\frac{1}{2}$ and Smad$\frac{2}{3}$-deleted mice8. Smad$\frac{2}{3}$ deletion in mouse fibroblasts comprises TGF-β-induced fibrosis. In myocardial infarction studies, activation of Smad2 and Smad3 led to myocardial fibrosis and cardiac remodeling21. Smad3 stimulates transcription of structural and matrix extracellular matrix proteins, including type I and III collagen, fibronectin, and periostin, and induces synthesis of matrix cross-linking enzymes8,22. In contrast to the critical effects in mediating myofibroblast conversion and cardiac fibrosis, Smad3 function dominantly over Smad2. Smad2 plays an important role in regulation of the reparative response following cardiac remodeling of the infarcted heart8. Loss of Smad2 did not affect scar organization, but was associated with delayed dilative remodeling21. Although the Smad signaling cascade is well defined in regulation of fibrotic responses in murine hearts, the knowledge of its function in patients with HOCM is still limited. In this histopathological comparison of 12 adult and 5 pediatric patients with HOCM, we demonstrated that Smad$\frac{2}{3}$ protein expression was significantly increased in the ventricular septal tissue in HOCM patients with extensive myocardial fibrosis, which indicated that the Smad$\frac{2}{3}$ signaling cascade is a persistent stimulus for the development of hypertrophic cardiomyopathy. Thus, Smad$\frac{2}{3}$ induced myocardial fibrosis is an important cause of sudden cardiogenic death in pediatric patients and late heart failure in adults.
Smad7 serves as a negative regulator of TGF-β1/Smads signaling pathway, which binds to activated TβRI to inhibit the phosphorylation and activation of R-Smads and antagonize TGF-β1 signaling23. Smad7 activation acts as a TGF-β-induced endogenous inhibitory signal that prevent myocardial fibrosis. Decreased Smad7 expression results in cardiac fibrosis in the infarcted rat heart, whereas Smad7 overexpression in cardiac fibroblasts inhibits collagen synthesis16,24. In our study, we found that SMAD7 expression significantly reduced in myocardial tissue of pediatric and adults with HOCM while SMAD$\frac{2}{3}$ expression increased, which is consistent with a previous report that Smad7 loss accentuated Smad$\frac{2}{3}$ activation25. Thus, with respect to patients with HOCM, the steady state controlled by the antifibrotic effects of SMAD7 and the fibrotic responses of SMAD$\frac{2}{3}$ might be disturbed.
The characteristic pathophysiological changes in hypertrophic cardiomyopathy are myocardial cell abnormalities, disorganized arrangement and interstitial fibrosis. The main components of the extracellular matrix (ECM) are collagen and glycosaminoglycans (GAGs)26. In hypertrophic cardiomyopathy, the most typical ECM components are collagens I and III, and the collagen content is heavily aggregated, affecting cardiac diastolic function17,27. In this study, we found that myocardial tissue collagen I expression was significantly elevated in pediatric and adults with HOCM, while SMAD2 and SMAD3 increased and inhibitory SMAD7 decreased in myocardial tissue, which was closely related to myocardial fibrosis described above. Therefore, activated SMAD signaling may be the direct cause of the fibrotic outcome of HOCM. And in our previous study, we found that patients with HOCM had reduced microvascular density and impaired myocardial perfusion4. The main role of MMP is to degrade type I and III collagen, and its elevated activity can cause increased fibrillar collagen degradation, extracellular matrix remodeling, and ventricular dilation28. Bi et al. found in patients with HOCM that myocardial MMP-2/TIMP-1 ratio was elevated by reduced microvasculature and that high plasma PICP/ICTP and MMP-2/TIMP-1 ratio were independent predictors of adverse outcomes in patients with HOCM29. So, our results have extended the understanding of the mechanism of HOCM, which helps to explore the new therapeutic treatments for HOCM patients as early as possible.
To date, more and more important molecules and crosstalk involved in the development of HOCM were revealed by bioinformatics technology, such as lncRNAs (XIST, MALAT1, and H19), TFs (SPI1 and SP1), and miRNAs (has-miR-29b-39 and has-miR-29a-3p)30. Transcription factors (TFs) can control the molecular network that regulates cardiomyocyte hypertrophy. Among them, SP1 is able to affect cardiomyocyte hypertrophy by SP1/GATA4 signaling pathways30, MRTF-A-Sp1-PDE5 axis31, and ROCK1-Sp1-PKCγ axis32. The study earlier reported that SP1 plays a crucial role in the pathogenesis and the progression of human glomerulonephritis probably via cooperation with pSmad$\frac{2}{3}$ and p30033. More importantly, many of the functions of TGF-β are mediated through Sp1 in cooperation with Smad signaling34. p300 is a transcriptional co-activator that may be involved in the activation of Smad complexes35. So, in our further research, Sp1 may be an important candidate that help uncover the more complicated pathogenic mechanism of HOCM.
## Limitations
This study also has some limitations. As a retrospective study, the research methods restricted to the sample size and sample types (only apply to the relevant examinations). First, the number of cases in this manuscript is slightly less, particularly the samples from pediatric patients with HOCM are very rare, but they are crucial to understand the development and mechanism of HOCM. Second, formalin/PFA- fixed paraffin-embedded tissues are long-lasting and easy to archive, but these are only fit for immunohistochemical investigations, which limits its applications. For these reasons, we will try to use some advanced technologies to the small but rare sample, such as using single cell RNA sequencing (scRNA-seq) to display comprehensive gene expression profiling of different cell types from the HOCM myocardial tissue, and to reveal the gene regulatory network or the crosstalk among different cell types by combining with bioinformatics technology. Also, we will collect RNA samples and protein samples from fresh tissues for data validations. Further studies with multi-level confirmations using various sample types, correlations between gene or transcript levels, and clinical prognosis will be carried out in the next part of our plans.
## Conclusions
In myocardial tissue from patients with HOCM, SMAD$\frac{2}{3}$ and type I collagen expression levels are elevated, whereas inhibitory SMAD7 protein expression levels are reduced, and their expression levels do not differ significantly between pediatric and adult patients with HOCM. These indicate that myocardial fibrosis due to SMAD signaling pathway activation occurs in childhood and that its fibrogenic effects persist and are a crucial factor in causing sudden cardiac death and heart failure in HOCM patients.
## References
1. Raman B. **Progression of myocardial fibrosis in hypertrophic cardiomyopathy: Mechanisms and clinical implications**. *Eur. Heart J. Cardiovasc. Imaging* (2019.0) **20** 157-167. DOI: 10.1093/ehjci/jey135
2. Shirani J, Pick R, Roberts WC, Maron BJ. **Morphology and significance of the left ventricular collagen network in young patients with hypertrophic cardiomyopathy and sudden cardiac death**. *J. Am. Coll. Cardiol.* (2000.0) **35** 36-44. DOI: 10.1016/S0735-1097(99)00492-1
3. Galati G. **Histological and histometric characterization of myocardial fibrosis in end-stage hypertrophic cardiomyopathy**. *Circ. Hear. Fail.* (2016.0) **9** 1-10
4. Zhang M-K, Zhang Z, Xue H, Fan L, Wen Y. **Microvascular rarefaction and myocardial fibrosis in hypertrophic obstructive cardiomyopathy: A histopathological comparison of pediatric and adult patients**. *Heart Surg. Forum* (2022.0) **25** E042-E047. DOI: 10.1532/hsf.4277
5. Segura AM, Frazier OH, Buja LM. **Fibrosis and heart failure**. *Heart Fail. Rev.* (2014.0) **19** 173-185. DOI: 10.1007/s10741-012-9365-4
6. Hanna A, Frangogiannis NG. **The role of the TGF-β superfamily in myocardial infarction**. *Front. Cardiovasc. Med.* (2019.0) **6** 1-15. DOI: 10.3389/fcvm.2019.00140
7. Hanna A, Humeres C, Frangogiannis NG. **The role of Smad signaling cascades in cardiac fibrosis**. *Cell. Signal.* (2021.0) **77** 1-29. DOI: 10.1016/j.cellsig.2020.109826
8. Khalil H. **Fibroblast-specific TGF-β-Smad2/3 signaling underlies cardiac fibrosis**. *J. Clin. Investig.* (2017.0) **127** 3770-3783. DOI: 10.1172/JCI94753
9. Dobaczewski M. **Smad3 signaling critically regulates fibroblast phenotype and function in healing myocardial infarction**. *Circ. Res.* (2010.0) **107** 418-428. DOI: 10.1161/CIRCRESAHA.109.216101
10. Frangogiannis NG. **Transforming growth factor–ß in tissue fibrosis**. *J. Exp. Med.* (2020.0) **217** 1-16. DOI: 10.1084/jem.20190103
11. Hu HH. **New insights into TGF-β/Smad signaling in tissue fibrosis**. *Chem. Biol. Interact.* (2018.0) **292** 76-83. DOI: 10.1016/j.cbi.2018.07.008
12. Nagueh SF. **American society of echocardiography clinical recommendations for multimodality cardiovascular imaging of patients with hypertrophic cardiomyopathy: Endorsed by the American society of nuclear cardiology, society for cardiovascular magnetic resonance, and**. *J. Am. Soc. Echocardiogr.* (2011.0) **24** 473-498. DOI: 10.1016/j.echo.2011.03.006
13. Schulz-Menger J. **Left ventricular outflow tract planimetry by cardiovascular magnetic resonance differentiates obstructive from non-obstructive hypertrophic cardiomyopathy**. *J. Cardiovasc. Magn. Reson.* (2006.0) **8** 741-746. DOI: 10.1080/10976640600737383
14. van de Vlekkert D, Machado E, d’Azzo A. **Analysis of generalized fibrosis in mouse tissue sections with Masson’s trichrome staining**. *Bio-Protoc.* (2020.0) **10** 1-16
15. Huang S. **The role of Smad2 and Smad3 in regulating homeostatic functions of fibroblasts in vitro and in adult mice**. *Biochim. Biophys. Acta Mol. Cell Res.* (2020.0) **1867** 118703. DOI: 10.1016/j.bbamcr.2020.118703
16. Wang B. **Regulation of collagen synthesis by inhibitory Smad7 in cardiac myofibroblasts**. *Am. J. Physiol. Hear. Circ. Physiol.* (2007.0) **293** 1282-1290. DOI: 10.1152/ajpheart.00910.2006
17. Bloksgaard M, Lindsey M, Martinez-Lemus LA. **Extracellular matrix in cardiovascular pathophysiology**. *Am. J. Physiol. Hear. Circ. Physiol.* (2018.0) **315** H1687-H1690. DOI: 10.1152/ajpheart.00631.2018
18. Hosper NA. **Epithelial-to-mesenchymal transition in fibrosis: Collagen type I expression is highly upregulated after EMT, but does not contribute to collagen deposition**. *Exp. Cell Res.* (2013.0) **319** 3000-3009. DOI: 10.1016/j.yexcr.2013.07.014
19. Li RK. **Overexpression of transforming growth factor-β1 and insulin-like growth factor-I in patients with idiopathic hypertrophic cardiomyopathy**. *Circulation* (1997.0) **96** 874-881. DOI: 10.1161/01.CIR.96.3.874
20. Ayça B. **Increased transforming growth factor-β levels associated with cardiac adverse events in hypertrophic cardiomyopathy**. *Clin. Cardiol.* (2015.0) **38** 371-377. DOI: 10.1002/clc.22404
21. Huang S. **Distinct roles of myofibroblast-specific Smad2 and Smad3 signaling in repair and remodeling of the infarcted heart**. *J. Mol. Cell. Cardiol.* (2019.0) **132** 84-97. DOI: 10.1016/j.yjmcc.2019.05.006
22. Biernacka A. **Smad3 signaling promotes fibrosis, while preserving cardiac and aortic geometry in obese diabetic mice**. *Circ. Hear. Fail.* (2015.0) **8** 788-798. DOI: 10.1161/CIRCHEARTFAILURE.114.001963
23. Nakao A. **Identification of Smad7, a TGFβ-inducible antagonist of TGF-β signalling**. *Nature* (1997.0) **389** 631-635. DOI: 10.1038/39369
24. Wang B. **Decreased Smad 7 expression contributes to cardiac fibrosis in the infarcted rat heart**. *Am. J. Physiol. Hear. Circ. Physiol.* (2002.0) **282** 1685-1696. DOI: 10.1152/ajpheart.00266.2001
25. 25.Humeres, C. et al. Smad7 effects on TGF-β and ErbB2 restrain myofibroblast activation and protect from postinfarction heart failure. J. Clin. Investig.132, (2022).
26. Fan D, Kassiri Z. **Modulation of cardiac fibrosis in and beyond cells**. *Front. Mol. Biosci.* (2021.0) **8** 1-7. DOI: 10.3389/fmolb.2021.750626
27. Lombardi R. **Myocardial collagen turnover in hypertrophic cardiomyopathy**. *Circulation* (2003.0) **108** 1455-1460. DOI: 10.1161/01.CIR.0000090687.97972.10
28. Spinale FG, Villarreal F. **Targeting matrix metalloproteinases in heart disease: Lessons from endogenous inhibitors**. *Biochem. Pharmacol.* (2014.0) **90** 7-15. DOI: 10.1016/j.bcp.2014.04.011
29. Bi X. **Matrix metalloproteinases increase because of hypoperfusion in obstructive hypertrophic cardiomyopathy**. *Ann. Thorac. Surg.* (2021.0) **111** 915-922. DOI: 10.1016/j.athoracsur.2020.05.156
30. Qin X. **Multi-factor regulatory network and different clusters in hypertrophic obstructive cardiomyopathy**. *BMC Med. Genom.* (2021.0) **14** 1-11. DOI: 10.1186/s12920-021-01036-4
31. Wu T. **An MRTF-A–Sp1–PDE5 axis mediates Angiotensin-II-induced cardiomyocyte hypertrophy**. *Front. Cell Dev. Biol.* (2020.0) **8** 1-14. DOI: 10.3389/fcell.2020.00839
32. Bai L. **Protocatechuic acid attenuates isoproterenol-induced cardiac hypertrophy via downregulation of ROCK1–Sp1–PKCγ axis**. *Sci. Rep.* (2021.0) **11** 1-16. DOI: 10.1038/s41598-021-96761-2
33. Kassimatis TI. **Transcription factor Sp1 expression is upregulated in human glomerulonephritis: Correlation with pSmad2/3 and p300 expression and renal injury**. *Ren. Fail.* (2010.0) **32** 243-253. DOI: 10.3109/08860220903411164
34. Pardali K. **Role of Smad proteins and transcription factor Sp1 in p21Waf1/Cip1 regulation by transforming growth factor-β**. *J. Biol. Chem.* (2000.0) **275** 29244-29256. DOI: 10.1074/jbc.M909467199
35. Massague J. **New EMBO members review: Transcriptional control by the TGF-beta/Smad signaling system**. *EMBO J.* (2000.0) **19** 1745-1754. DOI: 10.1093/emboj/19.8.1745
|
---
title: Longitudinal follow-up study of the association with gout and Alzheimer’s disease
and Parkinson’s disease in Korea
authors:
- Eun Jae Lee
- So Young Kim
- Hyo Geun Choi
- Yoo Hwan Kim
- Mi Jung Kwon
- Joo-Hee Kim
- Heui Seung Lee
- Jae Keun Oh
- In Bok Chang
- Joon Ho Song
- Ji Hee Kim
journal: Scientific Reports
year: 2023
pmcid: PMC9988850
doi: 10.1038/s41598-023-30379-4
license: CC BY 4.0
---
# Longitudinal follow-up study of the association with gout and Alzheimer’s disease and Parkinson’s disease in Korea
## Abstract
To date, no clear conclusion on the relationships of gout with the occurrence of typical neurodegenerative diseases, Alzheimer’s disease (AD) and Parkinson’s disease (PD), has been reached. This study aimed to determine whether the patients with gout are at a lower or higher probability of developing AD or PD than those without gout. Longitudinal follow-up data of a representative sample of Korean adults were assessed. 18,079 individuals diagnosed with gout between 2003 and 2015 were enrolled in the gout group. The comparison group comprised 72,316 demographics-matched individuals not diagnosed with gout. Longitudinal associations of gout with AD or PD were estimated using Cox proportional hazard regression adjusting for potential confounders. The adjusted hazard ratios (HRs) of AD and PD in the gout group were 1.01 and 1.16 times higher than controls, but these differences were not statistically significant ($95\%$ confidence interval [CI] = 0.92–1.12 and $95\%$ CI = 0.97–1.38, respectively). Although there was no significant association in the entire sample, AD and PD probabilities in patients with gout were significantly higher in participants < 60 years, and PD probabilities in patients with gout were significantly higher in overweight participants. Our findings identify significant correlations of gout with AD and PD in participants < 60 years and gout with PD in those with overweight, indicating that gout may play a role in the development of neurodegenerative diseases in younger or overweight populations. Further investigations should be performed to corroborate these findings.
## Introduction
Gout is a common systemic inflammatory disease characterized by severe joint pain. It results from the deposition of monosodium urate crystals in the synovial fluid of joints and in other tissues caused by a disorder of purine metabolism that leads to hyperuricemia. Gout presents either as recurrent acutely painful arthritis in a few joints or as chronic inflammatory polyarthritis affecting small and large joints of the extremities, which significantly impairs patients’ health-related quality of life. Hyperuricemia is the result of an increased production of uric acid, hypoexcretion of uric acid by the kidneys, or both1.
Alzheimer’s disease (AD), the most frequent form of dementia, is an irreversible, progressive neurodegenerative disorder of the central nervous system that is characterized by impairments of memory and learning. Its pathogenesis has been attributed to extracellular aggregations of beta-amyloid plaques and intracellular neurofibrillary tangles consisting of hyperphosphorylated tau protein2. Although numerous risk factors for AD have been suggested, it remains unknown whether any of these factors lead to amyloid deposition and tauopathy in humans. Parkinson’s disease (PD) is the second most common neurodegenerative disorder; it produces not only cardinal motor symptoms but also a wide range of nonmotor symptoms. Although significant progress has been made in elucidating the pathogenesis of PD, the cause of the progressive degeneration of dopaminergic neurons in the substantia nigra and the pathologic characteristics of PD remain unclear.
Because uric acid is a potent natural antioxidant and oxidative stress is one of the major pathogenic mechanisms of neurodegenerative diseases, close associations between gout due to hyperuricemia and neurodegenerative diseases, such as AD and PD, seem highly plausible. However, regarding whether gout, with its associated high serum levels of uric acid, protects against neurodegenerative disorders remains debatable. Some studies have reported that serum uric acid is not related to a lower risk of AD3; in contrast, a study of the United Kingdom Health Improvement Network reported that the multivariate-adjusted hazard ratio (HR) for AD among gout patients was 0.71 ($95\%$ confidence interval [CI] = 0.62–0.80), implicating a meaningful protective effect4. Similar to the findings of AD, studies on the relationship between gout and PD have also reached mixed conclusions5,6. One UK study using a large population-based database concluded that individuals with a previous history of gout had a lower risk of developing PD (odds ratio [OR] = 0.69, $95\%$ CI = 0.48–0.99)5. In contrast, a report derived from US Medicare claims data showed that gout was associated with a $14\%$ increased risk of incident PD6. Overall, epidemiological studies have reported conflicting results. Moreover, recent data have called into question the neuroprotective role of uric acid; therefore, the association between gout and neurodegenerative diseases remains equivocal7,8.
We thus aimed to determine the associations of gout with incident AD and PD in a Korean population through a longitudinal follow-up study, adjusting for various potential confounders (including metabolic syndrome-related risk factors).
## Results
Although there was a minor difference in the distribution of obesity between the gout group and the comparison group (standardized difference = 0.27), the absolute value of the standardized difference for most variables was less than 0.2, indicating that the intergroup differences in most baseline characteristics were well balanced after matching (Table 1).Table 1General characteristics of participants. CharacteristicsTotal participantsGout (n, %)Control (n, %)Standardized differenceAge (years old)0.00 40–44579 (3.2)2316 (3.2) 45–492049 (11.3)8196 (11.3) 50–543460 (19.1)13,840 (19.1) 55–593357 (18.6)13,428 (18.6) 60–642826 (15.6)11,304 (15.6) 65–692476 (13.7)9904 (13.7) 70–741838 (10.2)7352 (10.2) 75–791062 (5.9)4248 (5.9) 80–84357 (2.0)1428 (2.0) 85+75 (0.4)300 (0.4)Sex0.00 Male14,490 (80.1)57,960 (80.1) Female3589 (19.9)14,356 (19.9)Income0.00 1 (lowest)2514 (13.9)10,056 (13.9) 22235 (12.4)8940 (12.4) 32753 (15.2)11,012 (15.2) 43805 (21.0)15,220 (21.0) 5 (highest)6772 (37.5)27,088 (37.5)Residential area0.00 Urban7677 (42.5)30,708 (42.5) Rural10,402 (57.5)41,608 (57.5)Obesity†0.27 Underweight240 (1.3)1810 (2.5) Normal4548 (25.2)25,158 (34.8) Overweight4978 (27.5)20,141 (27.9) Obese I7573 (41.9)23,459 (32.4) Obese II740 (4.1)1748 (2.4)Smoking status0.08 Nonsmoker10,315 (57.1)40,314 (55.7) Past smoker3567 (19.7)12,869 (17.8) Current smoker4197 (23.2)19,133 (26.5)Alcohol consumption0.09 < 1 time a week9408 (52.0)41,025 (56.7) ≥ 1 time a week8671 (48.0)31,291 (43.3)Systolic blood pressure0.13 < 120 mmHg4226 (23.4)19,927 (27.6) 120–139 mmHg8798 (48.7)35,949 (49.7) ≥ 140 mmHg5055 (28.0)16,440 (22.7)Diastolic blood pressure0.13 < 80 mmHg6769 (37.4)30,597 (42.3) 80–89 mmHg6849 (37.9)27,248 (37.7) ≥ 90 mmHg4461 (24.7)14,471 (20.0)Fasting blood glucose0.05 < 100 mg/dL10,647 (58.9)43,630 (60.3) 100–125 mg/dL5740 (31.7)21,259 (29.4) ≥ 126 mg/dL1692 (9.4)7427 (10.3)Total cholesterol0.09 < 200 mg/dL9388 (51.9)40,304 (55.7) 200–239 mg/dL5979 (33.1)23,181 (32.1) ≥ 240 mg/dL2712 (15.0)8831 (12.2)Charlson comorbidity index0.12 011,345 (62.8)49,835 (68.9) 12796 (15.5)9979 (13.8) ≥ 23938 (21.8)12,502 (17.3)Alzheimer’s disease519 (2.9)1989 (2.8)0.01Parkinson’s disease166 (0.9)563 (0.8)0.02†Obesity (BMI, body mass index, kg/m2) was categorized as < 18.5 (underweight), ≥ 18.5 to < 23 (normal), ≥ 23 to < 25 (overweight), ≥ 25 to < 30 (obese I), and ≥ 30 (obese II).
The incidence rate of AD was 4.91 per 1000 person-years in the gout group and 4.71 per 1000 person-years in the comparison group. In patients with gout, the crude HR (matched on age, sex, income, and residential area) of AD was 1.05 ($95\%$ CI = 0.95–1.15, $$P \leq 0.360$$). After adjustment for obesity, smoking status, alcohol consumption, blood pressure, fasting blood glucose, total cholesterol, CCI score, and PD, the HR of AD during the follow-up period was 1.01 for patients in the gout group ($95\%$ CI = 0.92–1.12, $$P \leq 0.825$$, Table 2). The Kaplan–Meier survival curves and log-rank test of the gout and comparison groups indicated that the probability of AD was not significantly higher in the gout group than in the comparison group during the follow-up period ($$P \leq 0.367$$ in log-rank test, Fig. 1).Table 2Crude and adjusted hazard ratios of gout for Alzheimer’s disease with subgroup analyses stratified based on covariates. Incidence rate per 1000 person-yearsIncidence rate difference per 1000 person-years ($95\%$ confidence interval)Hazard ratio for Alzheimer’s disease ($95\%$ confidence interval)Crude†P-valueAdjusted†‡P-valueTotal participants ($$n = 90$$,395) Gout4.910.21 (− 0.26 to 0.67)1.05 (0.95–1.15)0.3601.01 (0.92–1.12)0.825 Control4.71Age < 60 ($$n = 47$$,225) Gout0.900.36 (0.14 to 0.57)1.65 (1.22–2.25)0.001*1.46 (1.07–2.00)0.018* Control0.5511Age ≥ 60 ($$n = 43$$,170) Gout11.12− 0.06 (− 1.19 to 1.08)1.00 (0.90–1.11)0.9990.97 (0.87–1.08)0.559 Control11.1711Men ($$n = 72$$,450) Gout4.170.01 (− 0.48 to 0.49)0.99 (0.88–1.11)0.8920.97 (0.86–1.09)0.642 Control4.1611Women ($$n = 17$$,945) Gout8.171.13 (− 0.20 to 2.46)1.19 (1.00–1.42)0.0521.10 (0.92–1.31)0.322 Control7.0411Underweight ($$n = 2050$$) Gout7.49− 2.59 (− 8.53 to 3.35)0.73 (0.37–1.45)0.3710.61 (0.30–1.24)0.174 Control10.0811Normal weight ($$n = 29$$,706) Gout5.830.43 (− 0.55 to 1.42)1.08 (0.91–1.29)0.3750.99 (0.83–1.19)0.946 Control5.4011Overweight ($$n = 25$$,119) Gout4.480.43 (− 0.39 to 1.26)1.11 (0.91–1.34)0.3070.95 (0.78–1.15)0.601 Control4.0511Obese ($$n = 33$$,520) Gout4.630.43 (− 0.24 to 1.09)1.10 (0.95–1.28)0.2041.10 (0.94–1.28)0.236 Control4.2011SBP < 140 mmHg and DBP < 90 mmHg ($$n = 64$$,040) Gout4.410.36 (− 0.18 to 0.90)1.09 (0.96–1.24)0.1871.03 (0.90–1.17)0.701 Control4.0511SBP ≥ 140 mmHg or DBP ≥ 90 mmHg ($$n = 26$$,355) Gout5.75− 0.42 (− 1.30 to 0.45)0.93 (0.81–1.08)0.3511.00 (0.86–1.16)0.979 Control6.1711Fasting blood glucose < 100 mg/dL ($$n = 54$$,277) Gout4.380.20 (− 0.36 to 0.75)1.05 (0.92–1.19)0.4831.01 (0.89–1.16)0.840 Control4.1911Fasting blood glucose ≥ 100 mg/dL ($$n = 36$$,118) Gout5.760.17 (− 0.65 to 0.99)1.03 (0.89–1.19)0.6541.01 (0.87–1.17)0.904 Control5.5911Total cholesterol < 200 mg/dL ($$n = 49$$,692) Gout5.010.11 (− 0.56 to 0.77)1.02 (0.90–1.17)0.7290.94 (0.83–1.08)0.412 Control4.9111Total cholesterol ≥ 200 mg/dL ($$n = 40$$,703) Gout4.820.35 (− 0.29 to 1.00)1.08 (0.94–1.24)0.2861.10 (0.95–1.27)0.186 Control4.4611SBP systolic blood pressure, DBP diastolic blood pressure. * Stratified or unstratified cox proportional hazard regression model, Significance at $P \leq 0.05.$ †Matched model based one age, sex, income, and residential area. ‡Adjusted for obesity, smoking status, alcohol consumption, SBP, DBP, fasting blood glucose, total cholesterol, Charlson comorbidity index scores, and Parkinson’s disease. Figure 1Kaplan‒Meier probability of the incidence of Alzheimer’s disease in gout and the control participants within 13 years of the index date.
The incidence rate of PD (per 1000 person-years) was 1.56 in the gout group and 1.32 in the comparison group. The longitudinal association between gout and PD had a crude HR of 1.18 ($95\%$ CI = 0.99–1.41, $$P \leq 0.057$$) and an adjusted HR of 1.16 ($95\%$ CI = 0.97–1.38, $$P \leq 0.100$$, Table 3); these values were not statistically significant. In the analysis of the cumulative incidence of PD according to the presence of gout, no significant between-group differences were found ($$P \leq 0.059$$ in log-rank test, Fig. 2).Table 3Crude and adjusted hazard ratios of gout for Parkinson’s disease with subgroup analyses stratified based on covariates. Independent variables and subgroupIncidence rate per 1000 person-yearsIncidence rate difference per 1000 person-years ($95\%$ confidence interval)Hazard ratio for Parkinson’s disease ($95\%$ confidence interval)Crude†P-valueAdjusted†‡P-valueTotal participants ($$n = 90$$,395) Gout1.560.24 (− 0.01 to 0.49)1.18 (0.99–1.41)0.0571.16 (0.97–1.38)0.100 Control1.3211Age < 60 ($$n = 47$$,225) Gout0.560.22 (0.05 to 0.39)1.63 (1.11–2.40)0.013*1.55 (1.04–2.31)0.032* Control0.3511Age ≥ 60 ($$n = 43$$,170) Gout3.080.27 (− 0.30 to 0.84)1.10 (0.91–1.33)0.3401.07 (0.88–1.31)0.492 Control2.8111Men ($$n = 72$$,450) Gout1.470.18 (− 0.10 to 0.45)1.13 (0.93–1.38)0.2091.14 (0.93–1.39)0.207 Control1.3011Women ($$n = 17$$,945) Gout1.950.52 (− 0.09 to 1.12)1.37 (0.95–1.97)0.0871.28 (0.89–1.86)0.188 Control1.4411Underweight ($$n = 2050$$) Gout4.141.73 (− 1.32 to 4.78)1.70 (0.65–4.47)0.2821.37 (0.49–3.88)0.551 Control2.4111Normal weight ($$n = 29$$,706) Gout1.590.18 (− 0.32 to 0.68)1.12 (0.80–1.57)0.4911.05 (0.75–1.48)0.758 Control1.4111Overweight ($$n = 25$$,119) Gout1.780.64 (0.18 to 1.10)1.56 (1.13–2.14)0.007*1.41 (1.02–1.95)0.037* Control1.1411Obese ($$n = 33$$,520) Gout1.350.05 (− 0.32 to 0.42)1.04 (0.79–1.37)0.7860.99 (0.75–1.31)0.966 Control1.3111SBP < 140 mmHg and DBP < 90 mmHg ($$n = 64$$,040) Gout1.420.23 (− 0.06 to 0.53)1.20 (0.95–1.50)0.1211.15 (0.91–1.45)0.237 Control1.1911SBP ≥ 140 mmHg or DBP ≥ 90 mmHg ($$n = 26$$,355) Gout1.790.17 (− 0.29 to 0.63)1.10 (0.85–1.44)0.4641.16 (0.88–1.51)0.293 Control1.6211Fasting blood glucose < 100 mg/dL ($$n = 54$$,277) Gout1.500.23 (− 0.08 to 0.54)1.18 (0.94–1.48)0.1431.12 (0.89–1.41)0.331 Control1.2711Fasting blood glucose ≥ 100 mg/dL ($$n = 36$$,118) Gout5.764.11 (3.62 to 4.59)1.17 (0.89–1.54)0.2481.20 (0.91–1.57)0.201Control1.6511Total cholesterol < 200 mg/dL ($$n = 49$$,692) Gout1.720.28 (− 0.08 to 0.65)1.20 (0.95–1.51)0.1261.14 (0.90–1.44)0.274 Control1.4311Total cholesterol ≥ 200 mg/dL ($$n = 40$$,703) Gout1.410.22 (− 0.12 to 0.55)1.18 (0.91–1.53)0.2071.15 (0.88–1.50)0.302 Control1.1911SBP systolic blood pressure, DBP diastolic blood pressure. * Stratified or unstratified cox proportional hazard regression model, Significance at $P \leq 0.05.$ †Matched model based one age, sex, income, and residential area. ‡Adjusted for obesity, smoking status, alcohol consumption, SBP, DBP, fasting blood glucose, total cholesterol, Charlson comorbidity index scores, and Alzheimer’s disease. Figure 2Kaplan‒Meier probability of the incidence of Parkinson’s disease in gout and the control participants within 13 years of the index date.
To determine if covariates contributed to the nonsignificant relationships of gout with AD and PD, we further conducted subgroup analyses stratified by multiple covariates. Gout remained a nonsignificant factor for AD in all subgroups, except for the age subgroup of individuals < 60 years old (adjusted HR = 1.46, $95\%$ CI = 1.07–2.00, $$P \leq 0.018$$, Table 2). Similarly, gout was not associated with the probability of PD in most covariates-stratified subgroups. A positive correlation of gout with PD remained only in patients < 60 years old (adjusted HR = 1.55, $95\%$ CI = 1.04–2.31, $$P \leq 0.032$$) and in those who were overweight (adjusted HR = 1.41, $95\%$ CI = 1.02–1.95, $$P \leq 0.037$$, Table 3).
## Discussion
In this large population-based study with a sample representative of the Korean population, we did not find any significant associations of gout with the two neurodegenerative diseases (AD and PD) after adjustment for not only age, sex, monthly income, and residential area but also body mass index (BMI), lifestyle factors, metabolic syndrome-related risk factors and comorbidities. The probabilities of AD and PD were only significantly higher in the gout group compared to the comparison group among individuals under the age of 60 years.
Recent epidemiological studies have reached inconsistent conclusions about the associations of gout with AD and PD. One study utilized claims data from a nationwide representative sample from Taiwan and revealed that patients with gout had a lower risk of both nonvascular (HR = 0.77, $95\%$ CI = 0.72–0.83) and vascular dementia (HR = 0.76, $95\%$ CI = 0.65–0.88)9. Similarly, according to a large population-based study, a $22\%$ lower risk of AD was observed among individuals with higher serum uric acid levels (HR = 0.78, $95\%$ CI = 0.66–0.91)10. A large longitudinal study, the Dutch Rotterdam study, also concluded that the risk of AD was lower among those with higher serum uric acid concentrations (HR = 0.89, $95\%$ CI = 0.80–0.99)11. These previous studies reported an inverse association between gout and dementia, supporting the purported potential neuroprotective role of uric acid, a powerful antioxidant.
Conversely, other observational studies have shown that hyperuricemia is linked to a higher risk of dementia and cognitive decline12,13. A large French population-based study in people 65 years or older reported that hyperuricemia was associated with a higher risk of AD (HR = 1.55, $95\%$ CI = 0.92–2.61) as well as age-related brain changes identified by MRI, demonstrating a clinical-pathological correlation3. In addition, one report using US Medicare claims data showed that gout was independently associated with a $15\%$ higher risk of incident dementia in older adults (HR = 1.15, $95\%$ CI = 1.12–1.18)14. These epidemiological data demonstrating a higher risk of dementia in patients with gout mainly focused on the role of uric acid in oxidative stress and inflammation.
As in our results, several other reports have also found no association between gout and AD. A recent study using 2-sample Mendelian randomization (MR) analysis, which is not susceptible to bias from unmeasured confounders or reverse causation, determined that epidemiological evidence did not support a causal relationship between gout and AD15. Other MR analyses also found no support of a causal role of genetically elevated serum levels of uric acid on the risk of AD (OR = 1.02, $95\%$ CI = 0.93–1.12)16.
Likewise, epidemiological data regarding the role of high uric acid levels in the development of PD have provided inconsistent results. A number of studies have revealed strong evidence of an inverse relationship of serum levels of uric acid (or gout) with PD, which suggests that individuals with gout may have a lower risk of PD17,18. In other words, these observations highlight the potential involvement of oxidative stress in PD and the protective role of uric acid, an antioxidant, against the development of PD. On the other hand, a UK study found that in the gout group, the increase in the risk of subsequent PD was modest (risk ratio [RR] = 1.11, $95\%$ CI 1.05–1.17)19. They suggested that the inflammatory state associated with the development of symptomatic gout or recurrent inflammatory arthritis may have counteracted any potential protective effect of uric acid as an antioxidant. Similar to our results, recent studies failed to find a robust association between gout and PD20. Two large MR analyses also reported no relationship between uric acid and the risk of PD, which suggests there is no clear causal association between serum levels of uric acid and the risk of PD21,22.
Because the present study had an observational design, we are unable to determine the plausible reason or mechanism underlying the lack of association between gout and neurodegenerative diseases. Based on previous research, possible biological explanations may involve the neutralization or incapacitation of two powerful functions of uric acid. Uric acid has antioxidative properties; specifically, uric acid scavenges reactive oxygen species23 and exerts a neuroprotective function by inhibiting oxyradical accumulation and preserving mitochondrial function, suppressing the cytotoxic action of lactoperoxidase, repairing DNA damaged by free radicals, and protecting against dopamine-induced apoptosis24. Gout flares occur due to the sudden release of monosodium urate crystal deposits in the joints, setting off an inflammatory cascade that manifests as an acute gouty arthritis attack25. Uric acid thus appears to be able to activate the immune response and, in that context, mediate the inflammatory process via the inflammasome26. Although this acute phase is self-limiting, one study found that monosodium urate crystals remain in the synovial fluid, causing persistent low-grade inflammation during the intercritical period27. As a result, acute and chronic inflammatory conditions of gout can yield both inflammatory responses and antioxidant effects, making it impossible to predict the associations of gout with AD or PD due to the interaction or counteraction between two competing actions.
Intriguingly, several studies (including a meta-analysis) reporting an inverse association between gout and incident PD described a stronger association in men than in women28. A recent study demonstrated that higher serum levels of uric acid were correlated with higher dopamine transporter uptake as assessed by a positron emission tomography (PET) scan, indicating a neuroprotective effect of gout against PD; this neuroprotective effect was more evident in women than in men29. Sex-specific analyses in a randomized placebo-controlled trial showed a correlation between an increase in serum uric acid levels and a slower rate of PD progression in women but not in men30. This sex difference in the association between gout and subsequent PD may suggest a greater neuroprotective effect of uric acid in women than in men. However, in sex-stratified analyses in the present study, this difference was not identified. Rather, in age-stratified analyses of both AD and PD probabilities, a positive correlation with gout was shown individuals under 60 years of age. Likewise, in the weight-stratified analysis of PD probability, a positive association with gout was found among participants with overweight. If gout is diagnosed at a relatively young age or when it is overweight, it may be assumed that the role of uric acid in the development of neurodegenerative diseases is more likely to be an inflammatory response than an antioxidant effect; but these postulations remain to be demonstrated, and further well-designed randomized controlled studies are needed to support this assumption.
This study should be interpreted with caution due to some limitations. First, this study was conducted using data extracted from a medical registry database, and diagnoses were defined based on diagnostic codes, which may increase concerns about inaccuracy of diagnoses of AD, PD, and gout due to codes. We further defined the presence of gout, AD, and PD as individuals with relevant claim codes based on at least two clinic visits, which may compensate for these shortcomings. Second, information on laboratory data was not available in this health insurance database; therefore, actual serum levels of uric acid were not included in the analyses. Third, we did not investigate drug use that could affect AD or PD, such as various gout management drugs, including uric acid-lowering drugs, nonsteroidal anti-inflammatory drugs (NSAIDs), steroids, interleukin (IL)-1 blockers, and colchicine. Further studies on the effects of gout medications are needed. Fourth, to identify the effect modification of age on the association between gout and AD or PD, interactions between age and gout should be tested using the Cox proportional hazard model with an added interaction term and covariates. Similarly, the same analysis is needed to test the effects of pooled normal and overweight on the risk of PD, but we could not consider them in this study. Finally, other unmeasured confounding factors, such as ethnicity, genetics or a familial history of neurodegenerative diseases, could not be completely excluded. Thus, it is possible that residual confounding effects and bias in the associations of gout with AD and PD may have been present.
In conclusion, our nationwide longitudinal follow-up study supports the possibility that gout does not significantly influence the occurrences of AD and PD. Additional evidence from future studies will complement these findings, further clarify these associations, and elucidate the pathophysiology of AD and PD.
## Ethics
The Ethics Committee of Hallym University [2019-10-023] approved this study. The need for written informed consent was waived by the Institutional Review Board. All analyses in this study were performed in accordance with the guidelines and regulations of the Ethics Committee of Hallym University. The study utilizes data from the Korean National Health Insurance Service-Health Screening Cohort. A random sample of about $10\%$ of the total population (approximately 515,000 out of 5,150,000) that underwent a health check-up between 2002 and 2003 was selected by the Korean National Health Insurance Service (NHIS). The age and sex specific distributions of the cohort population is available online. All Korean over the age of 40 and their families are required to undergo bi-annual health checks at no cost. The study benefits from the fact that all Korean citizens have a lifelong, 13-digit resident registration number, allowing for accurate population statistics to be calculated. This number is used in all hospitals and clinics in Korea and ensures that medical records do not overlap even if a patient moves. The Korean Health Insurance Review and Assessment (HIRA) system manages all medical treatments in Korea and the causes and dates of death, as recorded on death certificates by medical doctors, are officially reported. This NHIS data includes health insurance claim codes, diagnostic codes based on the International Classification of Disease-10 (ICD-10), death records, socioeconomic data, and health check-up information (including BMI, drinking and smoking habits, blood pressure, urinalysis results, hemoglobin levels, fasting glucose, lipid parameters, creatinine, and liver enzyme levels) for each participant from 2002 to 2013. Details regarding the Korean National Health Insurance Service-Health Screening *Cohort data* have been presented elsewhere31.
## Gout cases
According to the methods of a previous study32, gout cases were defined as participants who had visited the clinic or hospital more than twice with a diagnosis of gout according to the 10th revision of the International Classification of Diseases and Related Health Problems (ICD-10) code M10 (gout).
## Ascertainment of AD
Participants were considered to have AD if they were diagnosed with AD (ICD-10 code: G30) or Dementia in Alzheimer's disease (ICD-10 code: F00). To ensure the accuracy of diagnosis, we included only participants who had been treated ≥ 2 times for AD, as in our earlier studies33.
## Ascertainment of PD
Participants were considered to have PD if they were diagnosed with PD (ICD-10 code: G20). To ensure the accuracy of diagnosis, we recruited only participants who had visited hospitals or clinics ≥ 2 times for the treatment of PD, as in our previous study34.
## Selection of case and comparison groups
Gout participants were selected from 514,866 participants with 615,488,428 medical claim codes between 2002 and 2015 ($$n = 20$$,739); the remaining participants were included in the comparison group ($$n = 494$$,127). Individuals diagnosed with gout in 2002 ($$n = 2451$$) were excluded from the analysis such that the gout group consisted solely of participants first diagnosed with gout after 2002 (a washout period). Participants without blood pressure records in the gout group ($$n = 1$$) and those diagnosed with gout once previously in the comparison group ($$n = 10$$,255) were excluded. Gout participants were 1:4 matched with comparison participants in terms of age, sex, monthly income, and residential area (urban or rural). To limit selection bias in the matching process, participants without gout were randomly arranged and then included from the top of the list to the bottom. We assumed that comparison participants were evaluated simultaneously with each corresponding gout participant; in other words, that both participants had the same index date. Participants who died before the index date and those who had a history of AD or PD before the index date were excluded from each group. As a result of the matching process, 208 and 411,556 participants were excluded from the gout group and the comparison group, respectively, and thus, 18,079 gout participants and 72,316 comparison participants were included in the analysis (Fig. 3).Figure 3A schematic illustration of the group of participants selected for analysis of this study from the National Health Insurance database. Of the 514,866 participants, 18,079 gout participants were identified, and 72,316 control participants without gout were matched at a 1:4 ratio on age, sex, income, and residential area.
## Covariates
Participants were divided into 10 age groups at 5-year intervals and 5 income groups from class 1 (lowest income) to class 5 (highest income). The residential area was categorized as urban or rural following our previous study35. Smoking status, alcohol consumption, and obesity (using BMI, kg/m2) were categorized in the same manner as in our prior study36. The records of systolic blood pressure (SBP, mmHg), diastolic blood pressure (DBP, mmHg), fasting blood glucose (mg/dL), and total cholesterol (mg/dL) were also included in the analysis.
The Charlson Comorbidity Index (CCI), which assesses the burden of disease using 17 comorbidities, was calculated as a continuous variable from 0 (no comorbidities) to 29 (multiple comorbidities; excluding dementia) in the analysis.
## Statistical analyses
The standardized difference in means was used to evaluate whether the distribution of baseline characteristics between the gout and comparison groups was balanced after matching. Stratified Cox proportional hazard models were used to examine the associations of gout with AD or PD. Crude (unadjusted) models included participants matched by age, sex, income, and residential region. The adjusted models controlled for obesity, smoking status, alcohol consumption, SBP, DBP, fasting blood glucose, total cholesterol, CCI, AD diagnosis and PD diagnosis and provided HRs and $95\%$ CIs. We also used the Kaplan–Meier estimates to map the survival curves of the gout and comparison groups, and differences in their survival rates were evaluated with a log-rank test.
In addition, we further carried out subgroup analyses to examine the effects of various covariates on probabilities of AD and PD in patients with gout. Stratified Cox proportional hazard models were performed for age (< 60 years old and ≥ 60 years old) and sex. Unstratified Cox proportional hazard models were used for obesity (underweight, normal weight, overweight and obese), blood pressure (normal and hypertension), fasting blood glucose (< 100 mg/dL and ≥ 100 mg/dL), and total cholesterol (< 200 mg/dL and ≥ 200 mg/dL).
Two-tailed analyses were conducted, and the significance threshold was set at P values less than 0.05. SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) was used for statistical analyses.
## Ethic declarations
The Ethics Committee of Hallym University [2019-10-023] approved this study.
## References
1. Corrado A, D'Onofrio F, Santoro N, Melillo N, Cantatore FP. **Pathogenesis, clinical findings and management of acute and chronic gout**. *Minerva Med.* (2006) **97** 495-509. PMID: 17213786
2. Tiwari S, Atluri V, Kaushik A, Yndart A, Nair M. **Alzheimer’s disease: Pathogenesis, diagnostics, and therapeutics**. *Int. J. Nanomedicine* (2019) **14** 5541-5554. DOI: 10.2147/IJN.S200490
3. Latourte A. **Uric acid and incident dementia over 12 years of follow-up: A population-based cohort study**. *Ann. Rheum. Dis.* (2018) **77** 328-335. DOI: 10.1136/annrheumdis-2016-210767
4. Lu N. **Gout and the risk of Alzheimer's disease: A population-based, BMI-matched cohort study**. *Ann. Rheum. Dis.* (2016) **75** 547-551. DOI: 10.1136/annrheumdis-2014-206917
5. Alonso A, Rodríguez LAG, Logroscino G, Hernán MA. **Gout and risk of Parkinson disease: A prospective study**. *Neurology* (2007) **69** 1696-1700. DOI: 10.1212/01.wnl.0000279518.10072.df
6. Singh JA, Cleveland JD. **Gout and the risk of Parkinson’s disease in older adults: A study of US Medicare data**. *BMC Neurol.* (2019) **19** 4. DOI: 10.1186/s12883-018-1234-x
7. Hershfield MS, Roberts LJ, Ganson NJ, Kelly SJ. **Treating gout with pegloticase, a PEGylated urate oxidase, provides insight into the importance of uric acid as an antioxidant in vivo**. *Proc. Natl. Acad. Sci. U.S.A.* (2010) **107** 14351-14356. DOI: 10.1073/pnas.1001072107
8. Latourte A, Dumurgier J, Paquet C, Richette P. **Hyperuricemia, gout, and the brain—An update**. *Curr. Rheumatol. Rep.* (2021) **23** 82. DOI: 10.1007/s11926-021-01050-6
9. Hong J-Y. **Gout and the risk of dementia: A nationwide population-based cohort study**. *Arthritis Res. Ther.* (2015) **17** 139. DOI: 10.1186/s13075-015-0642-1
10. Scheepers LE. **Urate and risk of Alzheimer's disease and vascular dementia: A population-based study**. *Alzheimers Dement.* (2019) **15** 754-763. DOI: 10.1016/j.jalz.2019.01.014
11. Euser S, Hofman A, Westendorp R, Breteler MM. **Serum uric acid and cognitive function and dementia**. *Brain* (2009) **132** 377-382. DOI: 10.1093/brain/awn316
12. Cicero AFG. **Serum uric acid and impaired cognitive function in a cohort of healthy young elderly: Data from the Brisighella Study**. *Intern. Emerg. Med.* (2015) **10** 25-31. DOI: 10.1007/s11739-014-1098-z
13. Beydoun MA. **Serum uric acid and its association with longitudinal cognitive change among urban adults**. *J. Alzheimers Dis.* (2016) **52** 1415-1430. DOI: 10.3233/JAD-160028
14. Singh JA, Cleveland JD. **Gout and dementia in the elderly: A cohort study of Medicare claims**. *BMC Geriatr.* (2018) **18** 281. DOI: 10.1186/s12877-018-0975-0
15. Lee YH. **Gout and the risk of Alzheimer’s disease: A Mendelian randomization study**. *Int. J. Rheum. Dis.* (2019) **22** 1046-1051. DOI: 10.1111/1756-185X.13548
16. Yuan H, Yang W. **Genetically determined serum uric acid and Alzheimer’s disease risk**. *J. Alzheimers Dis.* (2018) **65** 1259-1265. DOI: 10.3233/JAD-180538
17. Annanmaki T, Muuronen A, Murros K. **Low plasma uric acid level in Parkinson's disease**. *Mov. Disord.* (2007) **22** 1133-1137. DOI: 10.1002/mds.21502
18. Winquist A, Steenland K, Shankar A. **Higher serum uric acid associated with decreased Parkinson's disease prevalence in a large community-based survey**. *Mov. Disord.* (2010) **25** 932-936. DOI: 10.1002/mds.23070
19. Pakpoor J, Seminog OO, Ramagopalan SV, Goldacre MJ. **Clinical associations between gout and multiple sclerosis, Parkinson’s disease and motor neuron disease: Record-linkage studies**. *BMC Neurol.* (2015) **15** 16. DOI: 10.1186/s12883-015-0273-9
20. Hu L-Y. **Risk of Parkinson’s disease following gout: A population-based retrospective cohort study in Taiwan**. *BMC Neurol.* (2020) **20** 338. DOI: 10.1186/s12883-020-01916-9
21. Kia DA. **Mendelian randomization study shows no causal relationship between circulating urate levels and Parkinson's disease**. *Ann. Neurol.* (2018) **84** 191-199. DOI: 10.1002/ana.25294
22. Kobylecki CJ, Nordestgaard BG, Afzal S. **Plasma urate and risk of Parkinson's disease: A Mendelian randomization study**. *Ann. Neurol.* (2018) **84** 178-190. DOI: 10.1002/ana.25292
23. Rao A, Balachandran B. **Role of oxidative stress and antioxidants in neurodegenerative diseases**. *Nutr. Neurosci.* (2002) **5** 291-309. DOI: 10.1080/1028415021000033767
24. Jones DC. **Cyanide enhancement of dopamine-induced apoptosis in mesencephalic cells involves mitochondrial dysfunction and oxidative stress**. *Neurotoxicology* (2003) **24** 333-342. DOI: 10.1016/S0161-813X(03)00042-1
25. Schumacher HR. **The pathogenesis of gout**. *Clevel. Clin. J. Med.* (2008) **75** S2-S4. DOI: 10.3949/ccjm.75.suppl_5.s2
26. Krishnan E. **Inflammation, oxidative stress and lipids: The risk triad for atherosclerosis in gout**. *Rheumatology* (2010) **49** 1229-1238. DOI: 10.1093/rheumatology/keq037
27. Pascual E. **Persistence of monosodium urate crystals and low-grade inflammation in the synovial fluid of patients with untreated gout**. *Arthritis Rheumatol.* (1991) **34** 141-145. DOI: 10.1002/art.1780340203
28. Shen C, Guo Y, Luo W, Lin C, Ding M. **Serum urate and the risk of Parkinson's disease: Results from a meta-analysis**. *Can. J. Neurol. Sci.* (2013) **40** 73-79. DOI: 10.1017/s0317167100012981
29. Oh YS. **Gender difference in the effect of uric acid on striatal dopamine in early Parkinson's disease**. *Eur. J. Neurol.* (2020) **27** 258-264. DOI: 10.1111/ene.14070
30. Schwarzschild MA. **Sex differences by design and outcome in the Safety of Urate Elevation in PD (SURE-PD) trial**. *Neurology* (2019) **93** e1328-e1338. DOI: 10.1212/WNL.0000000000008194
31. Kim SY, Min C, Oh DJ, Choi HG. **Tobacco smoking and alcohol consumption are related to benign parotid tumor: A nested case–control study using a national health screening cohort**. *Clin. Exp. Otorhinolaryngol.* (2019) **12** 412-419. DOI: 10.21053/ceo.2018.01774
32. Kim J-W. **Prevalence and incidence of gout in Korea: Data from the national health claims database 2007–2015**. *Rheumatol. Int.* (2017) **37** 1499-1506. DOI: 10.1007/s00296-017-3768-4
33. Kim SY, Min C, Oh DJ, Choi HG. **Risk of neurodegenerative dementia in asthma patients: A nested case–control study using a national sample cohort**. *BMJ Open* (2019) **9** e030227. DOI: 10.1136/bmjopen-2019-030227
34. Choi HG, Lim J-S, Lee YK, Sim S, Kim M. **Mortality and cause of death in South Korean patients with Parkinson’s disease: A longitudinal follow-up study using a national sample cohort**. *BMJ Open* (2019) **9** e029776. DOI: 10.1136/bmjopen-2019-029776
35. Kim SY, Min C, Oh DJ, Choi HG. **Bidirectional association between GERD and asthma: Two longitudinal follow-up studies using a national sample cohort**. *J. Allergy Clin. Immunol. Pract.* (2020) **8** 1005-1013. DOI: 10.1016/j.jaip.2019.10.043
36. Kim SY, Oh DJ, Park B, Choi HG. **Bell’s palsy and obesity, alcohol consumption and smoking: A nested case–control study using a national health screening cohort**. *Sci. Rep.* (2020) **10** 4248. DOI: 10.1038/s41598-020-61240-7
|
---
title: The prevalence and risk factors for anxiety and depression symptoms among migrants
in Morocco
authors:
- Firdaous Essayagh
- Meriem Essayagh
- Sanah Essayagh
- Ikram Marc
- Germain Bukassa
- Ihsane El otmani
- Mady Fanta Kouyate
- Touria Essayagh
journal: Scientific Reports
year: 2023
pmcid: PMC9988851
doi: 10.1038/s41598-023-30715-8
license: CC BY 4.0
---
# The prevalence and risk factors for anxiety and depression symptoms among migrants in Morocco
## Abstract
Humanitarian migration can result in mental health issues among migrants. The objective of our study is to determine the prevalence of anxiety and depression symptoms among migrants and their risk factors. A total of 445 humanitarian migrants in the Orientale region were interviewed. A structured questionnaire was used in face-to-face interviews to collect socio-demographic, migratory, behavioral, clinical, and paraclinical data. The Hospital Anxiety and Depression Scale was used to assess anxiety and depression symptoms. Risk factors for anxiety and depression symptoms were identified using multivariable logistic regression. The prevalence of anxiety symptoms was $39.1\%$, and the prevalence of depression symptoms was $40.0\%$. Diabetes, refugee status, overcrowding in the home, stress, age between 18 and 20 years, and low monthly income were associated with anxiety symptom. The associated risk factors for depression symptoms were a lack of social support and a low monthly income. Humanitarian migrants have a high prevalence of anxiety and depression symptoms. Public policies should address socio-ecological determinants by providing migrants with social support and adequate living conditions.
## Introduction
In recent decades, migration has increased to 281 million international migrants worldwide, accounting for $3.6\%$ of the world's population1. Human migration has an impact on both developed and developing countries2. With 86,000 migrants in 20143, Morocco, long known as a transit country to Europe, became a host country for migrants from West and Central Africa, as well as Syria4. Migration places people in situations that can have an impact on their physical, social, and mental health. The migration process can cause stress, feelings of loss, and social marginalization5, making migrants more vulnerable to health problems, including anxiety and depression. Depression is widely regarded as the leading cause of suicides and suicide attempts. Indeed, depression accounts for $40\%$ to $80\%$ of all suicide attempts globally6.
In 2017, anxiety disorders affected 260 million people worldwide, and depression affected 300 million, with economic consequences totaling at least $1000 billion (US) in lost productivity per year7. In 2009, a meta-analysis of populations exposed to conflict and refugees revealed a $30.8\%$ prevalence of depression8. A systematic review of 8,176 resettled Syrian refugees in ten countries published in 2019 found a prevalence of anxiety of $26\%$ and depression of $40\%$9.
To the best of our knowledge, no epidemiological study has focused on mental health among migrants in Morocco. Given the dynamics of migration and its impact on migrants' health, it is critical to investigate the prevalence of self-reported health issues among vulnerable populations. This would provide evidence on which policymakers in host countries and humanitarian organizations could rely to strengthen mental health services for undocumented migrants, asylum seekers, and refugees and meet their health needs in accordance with the Sustainable Development Goals and the global action plan "Promoting the health of refugees and migrants" (2019–2023)9.
Addressing mental health among undocumented migrants, asylum seekers, and refugees would also increase the chances of successful social integration, which would benefit the host country's socioeconomic capital in the long run. As a result, the objective of our study was to determine the prevalence of self-reported anxiety and depression symptoms among migrants in Morocco, as well as the risk factors associated with them.
## Setting of the study, study design, and population
The Orientale Region covers an area of approximately 90,130 km2 and has a population of 2,314,346 people, or $6.8\%$ of Morocco's total population10. The Mediterranean Sea borders it to the north, and Algeria borders it to the south. It is notable for its maritime coast, which stretches for 200 km. The region serves as a gateway to Africa, a focal point for the Maghreb, and a Mediterranean interface open to Europe. It is one of three major Moroccan regions that house asylum seekers and refugees.
Between November and December 2021, we conducted a cross-sectional survey among the migrant population that has relationships with associations working to improve the health of migrants in the Orientale region as stated in their statutes. The Prefecture of Oujda provided an exhaustive list of these associations. The sampling was divided into two stages. A random sample drawing was made to select the primary unit, consisting of 17 associations from the 30 associations present in Oujda. The secondary unit consisted of migrants aged 18 and older who were present in the Orientale region on the days of the survey and agreed to participate in the study. To determine the secondary unit in each selected primary unit, migrants were assigned a queuing number upon their arrival in the association, and these numbers were drawn at random to select the participants until the required sample size was achieved. A migrant was defined as any person of foreign origin in Morocco, regardless of their date of entry, duration of stay, or even settlement. The migrants were divided into three groups: (i) undocumented migrants are those who do not have a valid Moroccan residence permit; this situation could have resulted from entering the country without valid travel documents; being the child of undocumented parents; overstaying an entry visa; or losing a valid residence permit11; (ii) asylum seekers are people who are seeking safety from persecution or harm in a country other than their own and are waiting for a response to their application for refugee status; (iii) and refugees, who are defined as anyone who is recognized by the host country as being unable or unwilling to return to their country of origin due to a well-founded fear of persecution because of their race, religion, nationality, membership in a social group, or political beliefs12.
## Sample size determination
No previous study has measured the prevalence of self-reported anxiety and depression symptoms among migrants in Morocco. Hence, we used the hypothesis that the estimated prevalence of self-reported anxiety and depression symptoms is $50\%$. Considering the $95\%$ confidence level, the $5\%$ margin of error, and the $50\%$ prevalence of self-reported anxiety and depression symptoms, the estimated minimum sample size was 384 migrants.
## Data collection
We collected data on the participant's socio-demographic background, behavioral habits, comorbidities, and paraclinical parameters using a standardized structured questionnaire during a face-to-face individual interview. The questionnaire was given in Arabic, French, or English, which are the main languages spoken by migrants in Morocco. The interviewers translated the questionnaire into French and English. The French-translated questionnaire was given to two community workers for retranslation into Arabic. These community agents were bilingual; French was their mother tongue, and Arabic their second language. This technique ensured that the French version and the version written in Arabic were identical. The same was true for the English language questionnaire, but this time with two community workers whose first language is English. To ensure data standardization, we trained and involved the same interviewers. The data were collected in a closed room to ensure confidentiality and anonymity. Participants were not compensated in order to reduce selection bias.
## Anxiety symptom assessment
The WHO defines anxiety as a feeling of undetermined imminent danger accompanied by uneasiness, agitation, helplessness, and even annihilation7. In our study, participants self-rated their anxiety using the Hospital-Anxiety-and-Depression-Scale-A (HADS-A). On this scale, the participant reported how he or she felt over the previous two weeks13, this sub-scale consists of seven items: (i) I feel tense or irritable, (ii) I have a fear that something terrible will happen to me, (iii) I worry, (iv) I can sit quietly doing nothing and feel relaxed, (v) I have feelings of fear and my stomach is knotted, (vi) I am on the move and I can't keep it in place, and (vii) I have sudden feelings of panic. Each of these items was assigned a score ranging from 0 to 3, with "0" for no symptoms and "3" for the most severe symptoms. The sub-score for overall anxiety ranged from 0 to 21. Scores of 11 or higher were considered to signal elevated levels of anxiety symptomatology13.
## Depression symptom assessment
Depression is defined as sadness, loss of interest or pleasure, feelings of guilt or low self-esteem, sleep or appetite disturbances, feeling tired and lack of concentration7. In our study, participants self-rated their depression using the Hospital-Anxiety-and-Depression-Scale-D subscale (HADS-D). The participant reported his or her feelings over the previous two weeks on this scale13.
This sub-scale consists of seven items: (i) I still enjoy the same things I used to, (ii) I laugh easily and see the bright side of things, (iii) I'm in a good mood, (iv) I feel like I'm idling, (v) I'm no longer concerned with how I look, (vi) I look forward to doing certain things, and (vii) I enjoy a good book as well as a good radio or television show. Each of these items was assigned a score ranging from 0 to 3, with "0" for no symptoms and "3" for the most severe symptoms. The sub-score for overall depression ranged from 0 to 21. Scores of 11 or higher were considered to signal elevated levels of depression symptomatology.
Our study's variables included sociodemographic, migratory, behavioral, and clinical characteristics.
Social support was reported when a participant stated receiving assistance from close friends or associations.
The consumption of alcohol and tobacco was reported when the participant declared having consumed them during the last two months preceding the survey.
Overweight and obesity have been categorized according to the WHO body mass index classification14,15.
The stress was reported when the participant declared being stressed.
The countries of origin of migrants have been divided into two groups: Sub-Saharan Africa and the Eastern Mediterranean Region.
## Data management and statistical analysis
Epi Info version 7.2.0.1 was used to enter and analyze data. The precautionary measures to protect the confidentiality of the collected information and the anonymity of the participants were strictly followed. All tests were two-sided, with statistical significance set at less than 0.05.
Continuous variables were expressed as mean and standard deviation, while categorical variables were expressed as numbers and percentages. In bivariable analysis, the proportions of categorical variables were compared using the Pearson chi-2 test or, where applicable, Fisher's exact test. Where applicable, continuous variables were compared using the analysis of variance test or the Mann–Whitney test. We included in the multiple logistic regression any variables with a p-value up to 0.05 in the bivariable analysis.
We used a multiple logistic regression procedure to determine the full model. We revealed the association between each risk factor and the presence of symptoms of anxiety or depression using the adjusted odds ratio (AOR) and its $95\%$ confidence interval (CIs).
## Ethics approval and consent to participate
The study adhered to the Helsinki Declaration. Potential participants were informed of the study's main goal and procedure. All subjects who took part in the study signed an informed written statement of consent. The study protocol was reviewed and approved by the ethical review board of the faculty of Medicine and Pharmacy in Rabat, Morocco (#$\frac{33}{21}$).
## Socio-economic, demographic, and migration-specific characteristics
Table 1 summarizes the socio-economic, demographic, and migration-specific characteristics. During the study period, 445 participants were recruited, with 174 ($39.1\%$) reporting anxiety symptoms and 178 ($40.0\%$) reporting depression symptoms. The participants’ average age was 27.9 ± 10.9 years, ranging from 18 to 73 years, with 306 ($68.8\%$) males. There were 288 ($51.2\%$) undocumented migrants, 177 asylum seekers ($39.8\%$) and 40 refugees ($9.0\%$). A total of 109 ($24.5\%$) participants were homeless; 40 ($11.9\%$) lived in households with more than ten people; 300 ($67.4\%$) had a low monthly income; and 221 ($49.7\%$) had no social support. Table 1Socio-demographic and migrant characteristics, Morocco, 2021.All participants n (%)Anxiety symptoms n (%)Absence of anxiety symptoms n (%)Total participants445 [100]174 (39.1)271 (60.9)Mean age in years ± sd27.9 ± 10.928.1 ± 12.327.8 ± 09.9Age group in years 18–20119 (26.7) 58 (33.3) 61 (22.5) 21–25120 (27.0) 39 (22.4) 81 (29.9) 26–30 85 (19.1) 33 (19.0) 52 (19.2) 31 and older121 (27.2) 44 (25.3) 77 (28.4)Sex Male306 (68.8)126 (72.4)180 (66.4) Female139 (31.2) 48 (27.6) 91 (33.6)*Marital status* Partnered†140 (31.5) 57 (32.8) 83 (30.6) Single‡305 (68.5)117 (67.2)188 (69.4)Education Illiterate112 (25.2) 47 (27.0) 65 (24.0) Elementary144 (32.4) 57 (32.8) 87 (32.1) Middle school 56 (12.6) 22 (12.6) 34 (12.5) High school 92 (20.6) 37 (21.3) 55 (20.3) College 41 (09.2) 11 (06.3) 30 (11.1)Native country Eastern Mediterranean Region100 (22.5) 46 (26.4) 54 (20.0) Sub-Saharan Africa345 (77.5)128 (73.6)217 (80.0)Length of stay in Morocco (in years) ≥ 5 94 (21.1) 39 (22.4) 55 (20.3) < 5351 (78.9)135 (77.6)216 (79.7)*Legal status* Refugee 40 (09.0) 30 (17.2) 10 (03.7) Asylum seeker177 (39.8) 73 (42.0)104 (38.4) Undocumented migrant228 (51.2) 71 (40.8)157 (57.9)Number of countries crossed ($$n = 430$$) ≥ 3134 (31.2) 56 (33.3) 78 (29.8) < 3296 (68.8)112 (66.7)184 (70.2)Housing type Homeless109 (24.5) 47 (27.0) 62 (22.9) House*336 (75.5)127 (73.0)209 (77.1)Number of persons per house ($$n = 336$$) ≥ 10 40 (11.9) 25 (19.8) 15 (07.1) [5–9]174 (51.8) 63 (50.0)111 (52.9) ≤ 4122 (36.3) 38 (30.2) 84 (40.0)Occupation No430 (96.6)170 (97.7)260 (96.0) Yes 15 (03.4) 4 (02.3) 11 (04.0)Monthly income ($) ≤ 150300 (67.4)140 (80.5)160 (59.0) > 150145 (32.6) 34 (19.5)111 (41.0)Health insurance No444 (99.8)173 (99.4)271 (100.0) Yes 1 (00.2) 1 (00.6) 0 (00.0)Social support No221 (49.7)79 (45.4)142 (52.4) Yes224 (50.3)95 (54.6)129 (47.6)All participants n (%)Depression symptoms n (%)Absence of depression symptoms n (%)Total participants445 [100]178 (40.0)267 (60.0)Mean age in years ± sd27.9 ± 10.926.7 ± 11.828.7 ± 10.3Age group in years 18–20119 (26.7) 61 (34.3) 58 (21.7) 21–25120 (27.0) 52 (29.2) 68 (25.5) 26–30 85 (19.1) 28 (15.7) 57 (21.3) 31 and older121 (27.2) 37 (20.8) 84 (31.5)Sex Male306 (68.8)132 (74.2)174 (65.2) Female139 (31.2) 46 (25.8) 93 (34.8)*Marital status* Partnered†140 (31.5) 56 (31.5) 84 (31.5) Single‡305 (68.5)122 (68.5)183 (68.5)Education Illiterate112 (25.2) 36 (20.2) 76 (28.5) Elementary144 (32.4) 74 (41.6) 70 (26.2) Middle school 56 (12.6) 28 (15.7) 28 (10.5) High school 92 (20.6) 27 (15.2) 65 (24.3) College 41 (09.2) 13 (07.3) 28 (10.5)Native country Eastern Mediterranean region100 (22.5) 43 (24.2) 57 (21.3) Sub-Saharan Africa345 (77.5)135 (75.8)210 (78.7)Length of stay in Morocco (in years) ≥ 5 94 (21.1) 30 (16.8) 64 (24.0) < 5351 (78.9)148 (83.2)203 (76.0)*Legal status* Refugee 40 (09.0) 8 (04.5) 32 (12.0) Asylum seeker177 (39.8) 43 (24.2)134 (50.2) Undocumented migrant228 (51.2)127 (71.3)101 (37.8)Number of countries crossed ($$n = 430$$) ≥ 3134 (31.2) 56 (32.4) 78 (30.4) < 3296 (68.8)117 (67.6)179 (69.6)Housing type Homeless109 (24.5) 66 (37.1) 43 (16.1) House*336 (75.5)112 (62.9)224 (83.9)Number of persons per house ($$n = 336$$) ≥ 10 40 (11.9) 19 (17.1) 21 (09.3) [5–9]174 (51.8) 60 (54.1)114 (50.7) ≤ 4122 (36.3) 32 (28.8) 90 (40.0)Occupation No430 (96.6)176 (98.9)254 (95.1) Yes 15 (03.4) 2 (01.1) 13 (04.9)Monthly income ($) ≤ 150300 (67.4)148 (83.1)152 (56.9) > 150145 (32.6) 30 (16.9)115 (43.1)Health insurance No444 (99.8)177 (99.4)267 (100.0) Yes 1 (00.2) 1 (00.6) 0 (00.0)Social support No221 (49.7)131 (73.6) 90 (33.7) Yes224 (50.3) 47 (26.4)177 (66.3)sd standard deviation. * House means living in house or apartment or reception center. †A partnered means to be married or to be in concubine. ‡Single means to be single or divorced or widower.
## Behavioral characteristics
The data in Table 2 show that 96 ($21.6\%$) of the participants used tobacco, 72 ($16.2\%$) drank alcohol, 44 ($9.9\%$) were physically inactive, 320 ($71.9\%$) were stressed, and 18 ($4.0\%$) were diabetic. Table 2Behavioral characteristics and clinical features of migrants, Morocco, 2021.All participants n (%)Anxiety symptoms ($$n = 174$$)Absence of anxiety symptoms ($$n = 271$$)Tobacco consumption Yes 96 (21.6) 43 (24.7) 53 (19.6) No349 (78.4)131 (75.3)218 (80.4)Alcohol consumption Yes 72 (16.2) 29 (16.7) 43 (15.9) No373 (83.8)145 (83.3)228 (84.1)Physical activity Unsatisfactory 44 (09.9) 28 (16.1) 16 (05.9) Satisfactory401 (90.1)146 (83.9)255 (94.1)Stress Yes320 (71.9)145 (83.3)175 (64.6) No125 (28.1) 29 (16.7) 96 (35.4)Comorbidity Yes 15 (03.4) 10 (05.8) 5 (01.8) No430 (96.6)164 (94.2)266 (98.2)Overweight/obesity Yes177 (39.8) 68 (39.1)109 (40.2) No268 (60.2)106 (60.9)162 (59.8)Diabetes Yes 18 (04.0) 13 (07.5) 5 (01.9) No427 (96.0)161 (92.5)266 (98.1)Hypertension Yes124 (27.9) 41 (23.6) 83 (30.6) No321 (72.1)133 (76.4)188 (69.4)All participants n (%)Depression symptoms ($$n = 178$$)Absence of depression symptoms ($$n = 267$$)Tobacco consumption Yes 96 (21.6) 20 (11.2) 76 (28.5) No349 (78.4)158 (88.8)191 (71.5)Alcohol consumption Yes 72 (16.2) 10 (05.6) 62 (23.2) No373 (83.8)168 (94.4)205 (76.8)Physical activity Unsatisfactory 44 (09.9) 16 (09.0) 28 (10.5) Satisfactory401 (90.1)162 (91.0)239 (89.5)Stress Yes320 (71.9) 94 (52.8)226 (84.6) No125 (28.1) 84 (47.2) 41 (15.4)Comorbidity Yes 15 (03.4) 04 (02.2) 11 (04.1) No430 (96.6)174 (97.8)256 (95.9)Overweight/obesity Yes177 (39.8) 63 (35.4)114 (42.7) No268 (60.2)115 (64.6)153 (57.3)Diabetes Yes 18 (04.0) 5 (02.8) 13 (04.9) No427 (96.0)173 (97.2)254 (95.1)Hypertension Yes124 (27.9) 51 (28.6) 73 (27.3) No321 (72.1)127 (71.4)194 (72.7)
## Anxiety symptoms
In Table 1, a total of 174 ($39.5\%$) of the 445 participants screened showed symptoms of anxiety. Their average age was 28.1 ± 12.3 years, 126 ($72.4\%$) were male and 48 ($27.6\%$) were female. According to the socio-demographic data, 73 ($42.0\%$) were asylum seekers, 71 ($40.8\%$) were undocumented migrants, and 30 ($17.2\%$) were refugees. 170 participants ($97.7\%$) were unemployed, and 140 ($80.4\%$) had a low monthly income.
The behavioral characteristics reported in Table 2 show that: 28 ($16.1\%$) of participants reported inadequate physical activity; 145 ($83.3\%$) reported stress; and 10 ($5.7\%$) reported comorbidities. Diabetes was reported in 13 ($7.5\%$) of the participants.
The cut-off p-value after the bivariable analysis was set at a p-value ≤ 0.05. The legal status of undocumented migrants was considered the reference group during the bivariable analyses. This choice is based on the assumption that undocumented migrants are a population that still hopes for a better future in the host country after their legal status has been regularized. On the other hand, refugees are a population who witnesses the persistence of difficulties in accessing basic services even after regularization of their administrative situation, this realization leads to a higher level of anxiety or depression than that of undocumented migrants and asylum seekers. As mentioned in Table 3, according to the bivariable analysis, we identified eight factors associated with self-reported anxiety symptoms: (i) age between 18 and 20 years compared to 31 years or older (OR of 1.66; $95\%$ CI [0.99–2.78] (ii) refugee compared to undocumented migrant (OR of 6.63; $95\%$ CI [3.07–14.3]); (iii) living with more than ten people per household compared to four people or fewer each household (OR of 3.68; $95\%$ CI [1.74–7.76]); (iv) low monthly income (OR of 2.85; $95\%$ CI [1.82–4.46] (v) unsatisfactory physical activity (OR of 3.05; $95\%$ CI [1.60–5.83]); (vi) stress (OR of 2.74; $95\%$ CI [1.71–4.38]); (vii) presence of comorbidity (OR of 3.24; $95\%$ CI [1.08–9.65]); and (viii) having diabetes (OR of 4.29; $95\%$ CI [1.50–12.2]).Table 3Multivariable analysis (odds ratio, p-value) of risk factors associated with anxiety symptoms among migrants, Morocco, 2021.Bivariable analysisMultivariable analysis complete modelCOR ($95\%$ CI)p-valueAOR ($95\%$ CI)p-valueAge group in years 18–20/ ≥ 311.66 [0.99–2.78]0.053.56 [1.55–8.16]0.002 21–25/ ≥ 310.84 [0.49–1.43]0.521.14 [0.54–2.39]0.79 26–30/ ≥ 311.11 [0.62–1.96]0.711.76 [0.84–3.66]0.12Legal status Refugee/undocumented migrant6.63 [3.07–14.3] < 0.0016.26 [2.34–16.7]0.0003 Asylum seeker/undocumented migrant1.55 [1.03–2.33]0.031.38 [0.75–2.55]0.29Number of persons per house ≥ 10/ ≤ 43.68 [1.74–7.76]0.00064.08 [1.70–9.76]0.001 5–9/ ≤ 41.25 [0.76–2.05]0.361.28 [0.72–2.27]0.38Monthly income ≤ 150 ($)2.85 [1.82–4.46] < 0.0013.64 [2.08–6.37] < 0.001Unsatisfactory physical activity3.05 [1.60–5.83]0.00052.02 [0.90–4.53]0.08Stress2.74 [1.71–4.38] < 0.0014.30 [1.88–9.83]0.0005Having comorbiditie3.24 [1.08–9.65]0.030.75 [0.18–3.07]0.69Having diabetes4.29 [1.50–12.2]0.00318.6 [2.80–124.4]0.002COR Crude odds ratio, AOR Adjusted odds ratio, CI Confidence interval.
## Depression symptoms
In Table 1, a total of 178 ($40.0\%$) of the 445 participants screened showed symptoms of depression. Their average age was 26.7 ± 11.8 years, 132 ($74.2\%$) were males and 46 ($25.8\%$) were females. According to socioeconomic data, 127 ($71.3\%$) were undocumented migrants, 43 ($24.2\%$) were asylum seekers, and 8($4.5\%$) were refugees. 139 ($78.1\%$) participants were from Sub-Saharan Africa, and, 148 ($83.1\%$) had a low monthly income. According to Table 2, 131 ($73.6\%$) participants reported no social support, while 94 ($52.8\%$) reported stress.
The cut-off p-value after the bivariable analysis was set at p-value ≤ 0.05. As mentioned in Table 4, according to the bivariable analysis, we identified 16 factors associated with the self-reported depression symptoms: (i) age between 18 and 20 years compared to 31 years or older (OR of 2.38; $95\%$ CI [1.40–4.04]); (ii) age between 21 and 25 years compared to 31 years or older (OR of 1.73; $95\%$ CI [1.02–2.94]); (iii) male sex (OR of 1.53; $95\%$ CI [1.00–2.23]); (iv) elementary level of education compared to college (OR of 2.27; $95\%$ CI [1.09–4.74]); (v) middle school level of education compared to college (OR of 2.15; $95\%$ CI [0.92–4.99]); (vi) refugee compared to undocumented migrant (OR of 0.19; $95\%$ CI [0.08–0.45]); (vii) asylum seeker (OR of 0.25; $95\%$ CI [0.16–0.39]); (viii) homeless (OR of 3.06; $95\%$ CI [1.96–4.79]); (ix) number of people per household more than or equal to ten compared to four people or fewer each household (OR of 2.54; $95\%$ CI [1.21–5.33]); (x) number of people per household between five and nine compared to four people or fewer each household (OR of 1.48; $95\%$ CI [0.88–2.46]); (xi) unemployment (OR of 4.50; $95\%$ CI [1.00–20.1]); (xii) low monthly income (OR of 3.73; $95\%$ CI [2.35–5.91]); (xiii) lack of social support (OR of 5.48; $95\%$ CI [3.60–8.33]); (xiv) tobacco consumption (OR of 0.31; $95\%$ CI [0.18–0.54]); (xv) alcohol consumption (OR of 0.19; $95\%$ CI [0.09–0.39]); and (xvi) stress (OR of 0.20; $95\%$ CI [0.13–0.31]).Table 4Multivariable analysis (odds ratio, p-value) of risk factors associated with depression symptoms among migrants, Morocco, 2021.Bivariable analysisMultivariable analysis complete modelCOR ($95\%$ CI)p-valueAOR ($95\%$ CI)p-valueAge group in years 18–20/ ≥ 312.38 [1.40–4.04]0.0011.01 [0.42–2.41]0.96 21–25/ ≥ 311.73 [1.02–2.94]0.041.64 [0.79–3.40]0.17 26–30/ ≥ 311.11 [0.61–2.02]0.711.57 [0.73–3.36]0.23Sex male/female1.53 [1.00–2.33]0.040.85 [0.46–1.56]0.60Education Illiterate/college1.02 [0.47–2.19]0.951.89 [0.60–5.94]0.27 Elementary/college2.27 [1.09–4.74]0.021.41 [0.46–4.33]0.54 Middle school/college2.15 [0.92–4.99]0.071.29 [0.35–4.73]0.69 High school/college0.89 [0.40–1.98]0.782.13 [0.65–6.91]0.20Legal status Refugee/undocumented migrant0.19 [0.08–0.45]0.00010.39[0.13–1.21]0.10 Asylum seeker/undocumented migrant0.25 [0.16–0.39] < 0.0010.63 [0.32–1.21]0.16Homeless3.06 [1.96–4.79] < 0.0010.0 [0.00- > 1.0E12]0.96Number of persons per house ≥ 10/ ≤ 42.54 [1.21–5.33]0.011.57 [0.63–3.93]0.32 5–9/ ≤ 41.48 [0.88–2.46]0.131.47 [0.79–2.73]0.21Unemployed4.50 [1.00–20.1]0.042.08 [0.39–11.0]0.38Monthly income ≤ 150 ($)3.73 [2.35–5.91] < 0.0012.72 [1.50–4.91]0.0009Lack of social support5.48 [3.60–8.33] < 0.0013.48 [1.82–6.64]0.0001Tobacco consumption0.31 [0.18–0.54] < 0.0010.71 [0.32–1.56]0.40Alcohol consumption0.19 [0.09–0.39] < 0.0010.33 [0.13–0.82]0.01Stress0.20 [0.13–0.31] < 0.0010.66 [0.30–1.41]0.28COR Crude odds ratio, AOR Adjusted odds ratio, CI Confidence interval.
## Multivariable analysis
After adjusting for the other variables, we identified the following factors as risk factors associated with self-reported anxiety symptoms: (i) having diabetes (AOR of 18.68; $95\%$ CI [2.80–124.42]); (ii) refugee compared to undocumented migrants (AOR of 6.26; $95\%$ CI [2.34–16.8]); (iii) living with more than ten people per household compared to four people or fewer each household (AOR of 4.08; $95\%$ CI [1.70–9.76]); (iv) stress (AOR of 4.30; $95\%$ CI [1.88–9.83]); (v) age between 18 and 20 years compared to 31 years or older (AOR of 3.56; $95\%$ CI [1.55–8.16]); and (vi) low monthly income (AOR of 3.64; $95\%$ CI [2.08–6.37])“Table 3”.
For the depression symptoms, after adjustment for the other variables, two risk factors were associated: (i) lack of social support (AOR of 3.48; $95\%$ CI [1.82–6.64]); and (ii) low monthly income (AOR of 2.72; $95\%$ CI [1.50–4.91]). Alcohol consumption was a protective factor for depression symptoms (AOR of 0.33; $95\%$ CI [0.13–0.82]) “Table 4”.
## Discussion
To the best of our knowledge, this is Morocco's first study on the self-reported prevalence of anxiety and depression symptoms among migrants. In the literature, community assessment tools for mental diseases exist, but the emphasis tends to be on trauma and resilience factors or as a component part of "quality of life" rating scales. The Hospital Anxiety and Depression Score, which was not originally intended for community use, has been widely used in community settings around the world, including in our study16,17. The prevalence of self-reported anxiety symptoms among undocumented migrants, refugees, and asylum seekers in our survey was $39.1\%$, which is similar to Richter's study in 2018, where the prevalence of anxiety was $39.2\%$ among 296 migrants18. However, it remains higher than the $4.7\%$ reported in Nepal out of 574 migrants19 and the $7.7\%$ reported in Lebanon out of 194 migrants12.
For self-reported symptoms of depression, the prevalence was $40.0\%$ in our survey, which is consistent with the $39.5\%$ reported in Turkey on 238 migrants20. However, it is still lower than the $53.4\%$ reported in the Kizilhan study on 296 migrants in 201821.
In our study, the young age of migrants was associated with self-reported anxiety symptoms. This could be explained by discrimination, feelings of social isolation, a lack of financial resources, and uncertainty about the future. The discussion carried out with migrants had shown that the young subjects who left their country, their comfort zone, for a new country, came with enormous expectations and a lot of energy and motivation, but once in the host country, they came up against a reality marked by the difficulties of rapid access to adequate employment22. This situation pushed migrants to want to seek answers to the reasons for their unemployment, and it is then that migrant began to devalue themselves and loose self-esteem, which would lead to anxiety23.
Refugee status was associated with self-reported anxiety symptoms in our study. Indeed, we know that migrants, whether refugees, asylum seekers, or undocumented migrants, are a vulnerable population whose mental health can suffer. They may have been victims of traumatic events such as war, the loss of a family member, or physical and/or sexual violence before their arrival in the host country4,22, all of which can lead to anxiety and depressive disorders. Added to this is the negative perception of migrants that the population of the host country has, the cumbersomeness of the administrative procedures, and the length of the regularization procedures24 Once granted a refugee status, migrants find themselves caught between an ideal discourse on asylum and the reality in the host country. They come up against the persistence of discrimination, difficulties in accessing health services, access to work, adequate housing, and professional deskilling. All of these are factors that lead to the onset of anxiety symptoms in refugees.
In our study, low monthly income was associated with self-reported symptoms of anxiety and depression among migrants. Indeed, the low monthly income limits access to basic necessities such as housing, food, and medical care.
According to the literature, the presence of social support is associated with a low level of mental disorders. It facilitates migrants' integration into the host country and overcomes isolation and anti-migrant sentiment25. A lack of social support was associated with self-reported depression symptoms in our study.
Stress, considered a risk factor for several diseases, particularly anxiety and depression, could be linked to migrants' awareness that they have lost everything, that they no longer have any control over certain aspects of their lives, and that they have no social status in the host country26. It could also be the result of structural barriers caused by migrants' ignorance of the services available to them in the host countries. In addition, there is the fear of stigmatization, cultural and linguistic differences, post-migration living conditions, and the migrant's legal status in the host country. Dissatisfaction with the healthcare system can also be a source of stress and a major challenge for migrants. A healthcare system that does not provide services tailored to cultural differences can be a source of stress and an impediment to the well-being of migrants. In our study, stress was associated with self-reported anxiety symptoms.
We know that physical activity, when practiced in sufficient quantity, can improve an individual's health. It has the potential to prevent the onset of mental disorders such as anxiety and depression. According to the literature, people who engage in regular physical activity have a high level of life satisfaction27. Wieland's 2015 study of 127 migrants found that a lack of physical activity was associated with the development of mental disorders28. The disruption of daily life, economic pressure, ignorance of the environment, loss of social network, and lack of motivation in the face of extreme stress and uncertainty about the future could be the main barriers to the practice of physical activity among migrants. In our study, unsatisfactory activity physique was not associated with self-reported anxiety or depression.
Overcrowding and insufficient housing have long been associated with mental health issues, including among humanitarian migrants as a result of financial constraints, discrimination, and housing policies. In our study, living in overcrowded environments was associated with self-reported anxiety symptoms. A similar result was reported from a study of 681 migrants in Sweden29. This may be explained by the lack of privacy, autonomy, and isolation from the local community.
In terms of conditions, diabetes was associated with self-reported anxiety symptoms in our study. Indeed, these chronic illnesses necessitate medical attention and medication. Failure to meet these needs due to migrants' limited resources can exacerbate their health problems22, including mental health.
In our study, alcohol consumption was a protective factor against self-reported depression among migrants. The literature review showed that light to moderate alcohol consumption can provide short-term relief from self-reported depressive symptoms due to its euphoric and stimulating effects30. It gives migrants a feeling of relief, distancing themselves from their problems, and allows them to feel a tranquilizing or soothing effect. Excessive alcohol consumption allows them to anesthetize the emotions that assail them and that they are unable to manage and express. They drink to forget their ill-being, to endure their suffering, or even to get to sleep. However, in the long term, chronic consumption can become a triggering factor or aggravator of the depressive state30.
## Limitations of the study
Despite the guarantee of anonymity and the use of community actors during data collection, our study had some limitations, including prevarication bias in relation to monthly income and behavior characteristics during data collection.
Although the sampling suggests the presence of a selection bias in the study, which includes only migrants attending associations, this is not entirely correct. Indeed, in the Orientale region, the associations are well-known for the assistance and support they provide in terms of access to food, health care, and other basic life necessities for vulnerable people. As a result, they are frequented by refugees, asylum seekers, undocumented migrants, and even newly arrived migrants.
## Conclusion
Undocumented migrants, refugees, and asylum seekers are vulnerable populations to mental illnesses. The findings of our study revealed a high prevalence of self-reported anxiety and depression symptoms. Self-reported anxiety symptoms were associated with factors such as diabetes, young age, refugee status, crowding, low monthly income, and stress. A lack of social support and a low income were associated with self-reported depression symptoms. Following these findings, it is critical to emphasize the importance of preventing mental disorders among migrants by addressing social, economic, and environmental determinants. Future research measuring mental health status at arrival then at six months/12-month intervals would be beneficial to track changes.
## References
1. 1.McAulife, M. and A. Triandafyllidou (eds.), (2021). World Migration Report 2022. International Organization for Migration
(IOM), Geneva
2. 2.The United Nation High Commissioner for Refugees. HCR-Aperçu statistique. https://www.unhcr.org/fr/apercu-statistique.html.
3. 3.Haut Commissariat au Plan Maroc. Enquête national sur la migration internationale, 2018–2019, https://www.hcp.ma/downloads/Enquete-Nationale-sur-la-Migration_t21608.html. (2020).
4. Essayagh T. **Prevalence and determinants of intercourse without condoms among migrants and refugees in Morocco, 2021: A cross-sectional study**. *Sci. Rep.* (2022.0) **12** 22491. DOI: 10.1038/s41598-022-26953-x
5. Gkiouleka A. **Depressive symptoms among migrants and non-migrants in Europe: Documenting and explaining inequalities in times of socio-economic instability**. *Eur. J. Public Health* (2018.0) **28** 54-60. DOI: 10.1093/eurpub/cky202
6. 6.Corruble, E. Dépression et risque suicidaire. In: Philippe Courtet éd., Suicides et tentatives de suicide (pp. 120-123). Cachan: Lavoisier. 10.3917/lav.court.2010.01.0120
7. 7.World Health Organization. Monitoring mental health systems and services in the WHO European Region: Mental Health Atlas, 2017. (2019).
8. Steel Z. **Association of torture and other potentially traumatic events with mental health outcomes among populations exposed to mass conflict and displacement**. *JAMA* (2009.0) **302** 537. DOI: 10.1001/jama.2009.1132
9. 9.World Health Organization. Promoting the Health of Refugees and Migrants: Draft Global Action Plan. Seventy Second World Health Assembly. 2019.https://apps.who.int/gb/ebwha/pdf_files/WHA72/A72_25-en.pdf.
10. 10.Haut Commissariat au Plan. Données du Recensement Général de la Population et de l’Habtat de 2014 : Niveau National, Rabat. https://www.hcp.ma/downloads/RGPH-2014_t17441.html. (2018).
11. 11.International Organization for Migration. World migration report 2020. (2020).
12. Llosa AE. **Mental disorders, disability and treatment gap in a protracted refugee setting**. *Br. J. Psychiatry* (2014.0) **204** 208-213. DOI: 10.1192/bjp.bp.112.120535
13. Zigmond AS, Snaith RP. **The hospital anxiety and depression scale**. *Acta Psychiatr. Scand.* (1983.0) **67** 361-370. DOI: 10.1111/j.1600-0447.1983.tb09716.x
14. Essayagh T, Essayagh M, Essayagh S. **Drug non-adherence in hypertensive patients in Morocco, and its associated risk factors**. *Eur. J. Cardiovasc. Nurs.* (2021.0) **20** 324. DOI: 10.1093/eurjcn/zvaa002
15. Essayagh T. **Prevalence of uncontrolled blood pressure in Meknes, Morocco, and its associated risk factors in 2017**. *PLoS ONE* (2019.0) **14** e0220710. DOI: 10.1371/journal.pone.0220710
16. Harvey MR. **A multidimensional approach to the assessment of trauma impact, recovery and resiliency: Initial psychometric findings**. *J. Aggress. Maltreat Trauma* (2003.0) **6** 87-106. DOI: 10.1300/J146v06n02_05
17. Peddle N. **Assessing trauma impact, recovery, and resiliency in refugees of war**. *J. Aggress. Maltreat Trauma* (2007.0) **14** 185-204. DOI: 10.1300/J146v14n01_10
18. Richter K. **Prevalence of psychiatric diagnoses in asylum seekers with follow-up**. *BMC Psychiatry* (2018.0) **18** 206. DOI: 10.1186/s12888-018-1783-y
19. van Ommeren M. **Mental illness among bhutanese shamans in Nepal**. *J. Nerv. Ment. Dis.* (2004.0) **192** 313-317. DOI: 10.1097/01.nmd.0000122381.09491.7f
20. Tekin A. **Prevalence and gender differences in symptomatology of posttraumatic stress disorder and depression among Iraqi Yazidis displaced into Turkey**. *Eur. J. Psychotraumatol.* (2016.0) **7** 28556. DOI: 10.3402/ejpt.v7.28556
21. Kizilhan JI. **PTSD of rape after IS (“Islamic State”) captivity**. *Arch. Womens Ment. Health* (2018.0) **21** 517-524. DOI: 10.1007/s00737-018-0824-3
22. Essayagh F. **Disease burden among migrants in Morocco in 2021: A cross-sectional study**. *PLoS ONE* (2023.0) **18** e0281129. DOI: 10.1371/journal.pone.0281129
23. Chemali Z, Borba CPC, Johnson K, Khair S, Fricchione GL. **Needs assessment with elder Syrian refugees in Lebanon: Implications for services and interventions**. *Glob. Public Health* (2018.0) **13** 1216-1228. DOI: 10.1080/17441692.2017.1373838
24. Bogic M. **Factors associated with mental disorders in long-settled war refugees: refugees from the former Yugoslavia in Germany, Italy and the UK**. *Br. J. Psychiatry* (2012.0) **200** 216-223. DOI: 10.1192/bjp.bp.110.084764
25. Pannetier J, Lert F, JauffretRoustide M, du Loû AD. **Mental health of sub-saharan african migrants: The gendered role of migration paths and transnational ties**. *SSM Popul. Health* (2017.0) **3** 549-557. DOI: 10.1016/j.ssmph.2017.06.003
26. Abbott A. **The mental-health crisis among migrants**. *Nature* (2016.0). DOI: 10.1038/538158a
27. Netz Y, Wu M-J, Becker BJ, Tenenbaum G. **Physical activity and psychological well-being in advanced age: A meta-analysis of intervention studies**. *Psychol. Aging* (2005.0) **20** 272-284. DOI: 10.1037/0882-7974.20.2.272
28. Wieland ML. **Perspectives on physical activity among immigrants and refugees to a small urban community in Minnesota**. *J. Immigr. Minor Health* (2015.0) **17** 263-275. DOI: 10.1007/s10903-013-9917-2
29. Mangrio E, Zdravkovic S. **Crowded living and its association with mental ill-health among recently-arrived migrants in Sweden: A quantitative study**. *BMC Res. Notes* (2018.0) **11** 609. DOI: 10.1186/s13104-018-3718-6
30. Holzhauer CG, Gamble SA. **Depressive symptoms mediate the relationship between changes in emotion regulation during treatment and abstinence among women with alcohol use disorders**. *Psychol. Addict. Behav.* (2017.0) **31** 284-294. DOI: 10.1037/adb0000274
|
---
title: Falcaria vulgaris leaves extract as an eco-friendly corrosion inhibitor for
mild steel in hydrochloric acid media
authors:
- Mohammadreza Alimohammadi
- Mohammad Ghaderi
- Ahmad Ramazani S.A.
- Mohammad Mahdavian
journal: Scientific Reports
year: 2023
pmcid: PMC9988855
doi: 10.1038/s41598-023-30571-6
license: CC BY 4.0
---
# Falcaria vulgaris leaves extract as an eco-friendly corrosion inhibitor for mild steel in hydrochloric acid media
## Abstract
Undoubtedly, metal corrosion is one of the most challenging problems faced by industries. Introducing corrosion inhibitors is a reasonable approach to protecting the metal surface. Due to environmental concerns and the toxicity of industrial organic corrosion inhibitors, researchers are continually exploring acceptable replacements. The current study focused on the application of Falcaria Vulgaris (FV) leaves extract to mitigate mild steel (MS) corrosion in a 1 M HCl environment. The polarization findings demonstrated that the corrosion current density decreased from 264.0 µA/cm2 (for the sample submerged in the blank solution) to 20.4 µA/cm2 when the optimal concentration of 800 ppm of FV leaves extract was added to the acid solution. Electrochemical impedance spectroscopy (EIS) analysis revealed an inhibition efficiency of $91.3\%$ at this concentration after 6 h of immersion. It was determined by analyzing several adsorption isotherms that this corrosion inhibitor obeys the Frumkin isotherm. AFM, FE-SEM, and GIXRD surface analyses also supported the findings that adding FV leaves extract can reduce metal damage by adsorption on the metal surface.
## Introduction
Nowadays, metal corrosion is one of the most severe challenges confronting industries. Mild steel (MS), one of the most ubiquitous building materials, is highly susceptible to corrosive ions despite its remarkable qualities, including excellent mechanical capabilities and affordability1. Various methods are used to remove contaminants and rust from the MS surface, the most common of which is acid washing (especially using HCl)2. Consequently, it is crucial to employ strategies that minimize the rate of metal dissolution. Anti-corrosion coatings, corrosion inhibitors, anodic and cathodic protection, and other approaches have been proposed for this goal3. Among all these methods, corrosion inhibitors stand out as a promising approach. In order to select the proper corrosion inhibitor, three crucial factors must be taken into account: [1] effective adsorption, and competence in safeguarding the metal surface, [2] environmentally friendliness, and [3] affordability.
Generally, corrosion inhibitors containing heteroatoms (such as sulfur, oxygen, nitrogen, and phosphorus) can form a bond with the iron's vacant d orbital via their non-bonding electron pair, preventing metal corrosion by producing a protective layer4,5. Furthermore, compounds comprising aromatic rings and polar groups (such as C=O, –NH2, –OH, etc.) can be readily adsorbed on the metal surface via electrostatic attraction6,7. Even though some industrial organic corrosion inhibitors have the property mentioned above and exhibit potent inhibition against harsh ions, they suffer from the absence of two other qualities (they might be toxic and expensive)8–11. Hence, it is vital to discover an alternative with all the desirable features.
Recently, green organic corrosion inhibitors, including plant extracts12, expired drugs13, and ionic liquids14 with effective compounds, have been introduced as a substitution for toxic convectional corrosion inhibitors. Plant extracts, comprising leaf15, fruit 16, and seed 17 extracts, are generally biocompatible, biodegradable, and cost-effective. Also, donor electron components such as aromatic groups, heteroatoms, and compounds with π electrons in plant extract can further confirm their potential to be used as potent corrosion inhibitors18–21.
Many efforts have shown the role of different plant extracts in the protection of MS surfaces in acidic media. Mostafatabar et al. evaluated the inhibitory effect of carrot pomace extract. They clarified that the extract molecules could be adsorbed physicochemically and generate a protective layer on the MS panel, leading to $95\%$ efficiency at 400 ppm extract concentration according to polarization assessment22. In another report, garcinia cambogia fruit rind extract derived from aqueous and alcoholic media was introduced as a green corrosion inhibitor demonstrating mix mode (cathodic and anodic) protection via Langmuir and Temkin adsorption isotherm, respectively23. Dehghani et al. investigated the inhibition action of the rosemary extract at different concentrations and temperatures. Their results showed that increasing the rosemary extract concentration to 800 ppm enhanced the corrosion inhibition efficiency to $92.0\%$24. Moreover, exploration of the inhibition performance of other plant extracts, including Mish Gush25, Xanthium strumarium26, *Eriobotrya japonica* Lindl27, Cardaria draba28, Urtica dioica29, Arbutus unedo L30, *Euphorbia heterophylla* L31 and Thymus vulgaris32, Onion–garlic33 obviously endorsed the potential of plant extracts as corrosion inhibitors.
Falcaria vulgaris (FV) is a species of the Apiaceae family found in West Asia, Europe, and the United States. This plant has been used for medical applications such as healing skin and gastrointestinal diseases in many regions of Iran. Furthermore, the antibacterial and antioxidant properties of FV have been approved due to the existence of carvacrol and spathulenol in its structure. In addition to the mentioned compounds, FV leaves extract contains genistin, rutin, quercetin-3-O-glucoside, and quercetin (Fig. 1)34,35.Figure 1The chemical structure of Carvacrol (a), Spathulenol (b), Genistin (c), Rutin (d), Quercetin-3-O-glucoside (e), and Quercetin (f).
In current study, we aimed to investigate the inhibitory performance of FV extract towards corrosion mitigation of MS submerged in 1 M HCl medium. Electrochemical measurements such as EIS and polarization were used to explore this compound's anti-corrosion properties. In addition, the morphology and topology of the surface were examined by exploiting energy-dispersive X-ray analysis (EDX), field-emission scanning electron microscopy (FE-SEM), and atomic force microscopy (AFM). Also, Fourier-transform infrared spectroscopy (FTIR) and grazing incidence X-ray diffraction (GIXRD) were used to assess the FV-based layer formed on the MS surface.
## Extraction process
FV leaves were collected from the Markazi province, Iran (with permission of the landowner) and dried and powdered after washing with distilled water. For extract preparation, 15 g of FV leaves powder was poured into 500 ml of deionized water and agitated with a heater stirrer for 12 h at 70 °C. After that, filtration was performed using filtering paper to separate dark brown liquid and solid, and in the second step, the obtained solution was centrifuged at 4000 rpm for 5 min. Finally, the prepared extract was dried at 45 °C overnight. All methods were performed in accordance with the relevant guidelines and regulations. The main chemical structure of FV leaves extracts is illustrated in Fig. 1, as already mentioned.
## Sample preparation
MS plates (ST12, Foolad-E-Mobarakeh Co., Iran) were used as a working electrode. Hydrochloric acid ($37\%$, Doctor-Mojalali Co., Iran) was diluted with distilled water to prepare 1 M HCl. Different grades of silicone carbide paper (400–1000) were used to remove the surface scales. After that, the surface was cleaned using industrial-grade of acetone. Finally, different concentrations of FV extract in 1 M HCl (0, 400, 600, 800, and 1000 ppm) were prepared for electrochemical measurement (the solubility of FV leaves extract was higher than 5 g L−1).
## Electrochemical measurements
CorrTest (CS350, China) potentiostat instrument was exploited to check the corrosion inhibition effect of FV leaves extract. The electrochemical setup comprised three-electrode, including MS (contact surface = 1 cm2), Calomel and Pt rod. EIS analysis was conducted by applying 10 mV AC voltage at open circuit potential (OCP) in the frequency range of 0.01 to 10,000 Hz. Also, the polarization test was performed within the voltage range of −250 to + 250 mV versus OCP with a 0.5 mV/s sweep rate. Three experiments were performed for each concentration to guarantee the reproducibility of data.
## Surface study
To study surface characteristics, MS was soaked in an acidic solution with and without 800 ppm of FV extract for 6 h. After that surface of MS was rinsed with distilled water two times and then dried at room temperature. Morphological and elemental analysis of the MS surface were assessed using FE-SEM (TE-Scan—MIRA3) and EDX (Oxford—X-MAX-80). Also, the surface topology was examined using AFM (Bruker, Icon, United States). The adsorption of FV extract molecules on the MS surface was analyzed using FTIR (Thermo, Avatar, United States), Ultraviolet–visible spectroscopy (UV–Vis, Thermo, Biomate5, United States), and GIXRD (X'Pert PRO MPD PANalytical Company).
## OCP
The OCP assessment of the MS substrate was conducted during 1500 s immersion in the HCl medium, in the presence and absence of the FV leaves extract. The data depicted in Fig. 2 reveals that the OCP reached a state of equilibrium prior to the completion of the 1500 s, as no discernible deviation from the OCP values was observed after 1000 s. The OCP trend for both conditions was congruent, commencing at a lower potential and progressively augmenting over time, a phenomenon that can be attributed to the formation of an oxide layer on the MS surface as a result of the corrosive species attack until a steady state is attained36. Notably, the initial OCP value in the FV-containing solution was higher than that of the blank solution, which may be indicative of the adsorption of inhibitors onto the MS surface36. Upon examination of the OCP values in the presence of FV, it was determined that the maximum deviation of OCP was under 85 mV, indicating that the FV extract displays properties of a mixed-type corrosion inhibitor for MS in a 1 M HCl solution37. However, a detailed analysis of the data presented in Fig. 2 highlights that FV primarily exerts its inhibition effects through inhibiting anodic reactions as OCP shows a positive shift upon addition of FV extract38.Figure 2OCP vs time diagrams for MS submerged in an acidic solution without and containing different concentration of FV leaves extract.
## EIS
EIS analysis was used to scrutinize the corrosion-inhibiting competence of FV leaves extract. Figure 3 illustrates the Nyquist and Bode plots of MS immersed in acidic electrolytes containing various concentration of FV extract solution (400–1000 ppm) and blank solution. A depressed semicircle, representing the roughness of the tested sample's surface, can be seen in all Nyquist diagrams39. It is of note that the Nyquist graph's shape remained unchanged when various inhibitor concentrations were added, indicating that the mechanism of the corrosion reaction remained unaltered. Furthermore, only one time constant is visible on the Bode-phase angle diagram, revealing that the charge transfer mechanism predominates at the metal/electrolyte interface40.Figure 3Nyquist (left) and Bode (right) plots of MS soaked in 1 M HCl solution without (a,b) and containing 400 (c,d), 600 (e,f), 800 (g,h), and 1000 (i,j) ppm of FV extract.
The EIS data were fitted by exploiting a one-time constant equivalent electrical circuit (chi-square < 0.005) utilizing ZView software for a more precise examination, and the related parameters are reported in Table 1. In this table, the terms Rs, CPEdl, Y0, and n connote the electrolyte resistance, constant phase element, admittance, and power of the CPE, respectively. Furthermore, Rp stands for polarization resistance, calculated by adding the charge transfer and film resistance (Rp = Rct + Rf) 19. As demonstrated in Fig. 3, the smallest diameter of the Nyquist diagram (which indicates Rp) corresponds to the sample immersed in the inhibitor-free solution and displays a declining trend over time. The corrosion reactions instigate as soon as the working electrode is soaked in 1 M HCl medium. As a result, the attack of corrosive ions induces MS dissolution and the degradation of the porous oxide layer, culminating in poor MS resistance in the absence of corrosion inhibitors. In contrast, the diameter of the Nyquist diagram was considerably boosted by introducing diverse concentrations of FV leaves extract in the acidic solution. To gain a better insight into the corrosion-inhibiting properties of FV leaves extract, the inhibition efficiency (η) was computed using Eq. [ 1]41:Table 1Electrochemical parameters attained from fitted EIS data. Concentration (ppm)Immersion time (h)Rsa (Ω cm2)Rctb (Ω cm2)CPECdle (µF cm−2)log |Z|0.01 Hz (Ω cm2)η (%)τ (s)θEISY0c (µsn Ω−1 cm−2)ndBlank21.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.02100.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.3272.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.230.94 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0140.62.00–0.004–41.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0292.1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.2184.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.450.94 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0147.91.96–0.004–61.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0386.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.999.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.670.94 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0157.01.94–0.004–241.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0266.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.23237.2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 2.450.95 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.01150.91.81–0.01–40021.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.08429.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 5.3652.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 3.180.89 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0115.62.6276.60.00661.541.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.04477.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 4.6754.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.880.89 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0116.92.6780.70.00864.661.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.15482.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 8.3255.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 2.120.89 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0116.92.6682.10.00870.3241.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.04257.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 6.8983.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 3.470.86 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0118.72.474.30.00487.660021.2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.15682.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 11.1231.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 5.100.88 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.017.62.8485.30.00581.241.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.13712.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 8.7833.1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 3.660.89 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.019.42.8587.00.00680.361.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.12694.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 9.5637.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 3.780.89 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0110.62.8387.60.00781.4241.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.08364.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 5.8378.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 2.220.85 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0117.22.5381.80.00688.680021.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.31778.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 13.8745.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 8.160.88 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0112.32.8987.10.00969.741.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.13954.1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 12.1245.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 4.120.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0114.13.0090.30.01370.661.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.16988.1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 12.4947.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 3.130.89 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0116.23.0091.30.01671.5241.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.17633.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 8.4463.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 2.110.88 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0119.02.8089.50.01287.4100022.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.30506.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 16.1989.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 5.550.91 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0119.02.7180.20.00953.242.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.27654.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 1.277.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 3.980.91 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0118.82.8285.90.01260.662.2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.21674.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 12.9164.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 3.450.91 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0118.42.8387.80.01267.8242.2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.12535.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 9.9082.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 2.200.92 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document}± 0.0121.42.7487.60.0185.8aStandard deviation changed between 1.5 and $2.1\%$bStandard deviation changed between 2.6 and $3.2\%$cStandard deviation changed between 2.2 and $4.7\%$dStandard deviation changed between 1.9 and $3.5\%$eStandard deviation changed between 3.1 and $5.4\%$.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\eta \%=\left(1-\frac{{R}_{p.0}}{{R}_{p. i}}\right)100\%$$\end{document}η%=1-Rp.0Rp.i$100\%$ *In this* equation, Rp,0 and Rp, i denote the polarization resistance of MS in the inhibitor-free and inhibitor-containing solution, respectively. Compared to the blank solution, the sample submerged in the inhibitor-containing solution has a greater Rp, as shown in Table 1. It is worth mentioning that as the concentration of the corrosion inhibitor increased, the Rp elevated, reaching a maximum of Rp = 988.1 Ω cm2 and η % = $91.3\%$ at 800 ppm after 6 h of immersion. This remarkable improvement is due to corrosion inhibitor adsorption on the MS surface, which establishes a protective layer and blocks the access of the corrosive electrolyte to the metal surface. Notably, the η has dropped slightly after 24 h of immersion, suggesting the extract's long-term capacity to mitigate corrosion of the MS surface.
The Bode diagram can provide more information regarding the corrosion-inhibition activity of FV leaves extract. According to the literature, rising |Z| at the lowest frequency (0.01 Hz) and decreasing the phase angle towards -90° (pure capacitor phase angle) at the highest frequency (10,000 Hz) suggests an increase in corrosion resistance36. Table 1 and the Bode graphs show that these two parameters (|Z|0.01 Hz and phase angle at 10 kHz) for the coupons dipped in the inhibitor-based solution are greater than the blank one, confirming the development of the FV-based protective film on the working electrode's surface. Further beneficial parameters in corrosion studies are capacitance (Cdl) and relaxation time (τ), which are expressed by Eqs. [ 2] and [3], respectively36,42, and their values are given in Table 1.2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{dl}={Y}_{0}^\frac{1}{n}\times {(\frac{{R}_{s}\times {R}_{p}}{{R}_{s}+{R}_{p}})}^{\frac{1-n}{n}}= \frac{{\varepsilon }^{0} \varepsilon A}{d}$$\end{document}Cdl=Y01n×(Rs×RpRs+Rp)1-nn=ε0εAd3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau ={R}_{p}\times {C}_{dl}$$\end{document}τ=Rp×Cdl In Eq. [ 2], ε0 and ε symbolize the dielectric constants of air and the double layer, while A and d signify the surface area of MS and the electric double layer thickness, respectively. A close inspection of Table 1 implies that when the inhibitor is added to the acidic solution, Cdl significantly lowered in comparison to the inhibitor-free solution. These findings might be related to the replacement of inhibitor components with water molecules, leading to a drop in ε and an increase in d, resulting in a decrease in Cdl43. According to the inferences mentioned above, the decline in Cdl can be linked to a reduction in the contact area between corrosive water molecules and the MS surface. Furthermore, the increase in τ of the sample soaked in the inhibitor-based solution might be linked to the inhibitor molecule adsorption on the MS surface, which ultimately resulted in a delay in reaching equilibrium following charge distribution44.
Surface coverage (θEIS), which is given by Eq. [ 4]45,46, is another parameter that may be extracted from EIS data:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\theta }_{EIS} \%=\left(1-\frac{{C}_{dl. i}}{{C}_{dl. 0}}\right)100\%$$\end{document}θEIS%=1-Cdl.iCdl$.0100\%$ *In this* equation, Cdl,0 and Cdl,i represent the capacitance of MS in solutions without and with FV leaves extract, respectively. The orientation of corrosion inhibitor molecules on the MS surface (horizontal or vertical) can be assessed by comparing θEIS with η. When a corrosion inhibitor is adsorbed horizontally on a metal surface, it blocks the access of corrosive ions to the metal surface and leads to an increase in Rp and η. Consequently, it is demonstrated that η > θEIS when the horizontal adsorption of inhibitor molecules occurs on the MS surface43,47,48. Conversely, when corrosion inhibitor molecules are vertically oriented on the MS surface, the electric double layer thickness rises and, as a result, Cdl decreases. Meanwhile, *Rp is* less affected because corrosion inhibitor molecules replace just a limited number of water molecules. Thus, η < θEIS reflects the vertical adsorption of corrosion inhibitors43,47,48.
A careful evaluation of Table 1 discloses that the FV leaves extract molecules (which contain diverse compounds such as carvacrol, spathulenol, genistin, etc.) are horizontally adsorbed on the MS surface during 6 h of immersion (η > θEIS). It is worth noting that some compounds in FV leaves extract may be desorbed from the metal surface after 24 h of immersion. Thus, as shown in Table 1, it is not surprising that the orientation of the remaining molecules on the MS surface has altered and become vertical (η < θEIS) after 24 h of immersion at concentrations of 400 and 600 ppm49. Higher concentrations (800 and 1000 ppm) also show a similar trend, and after 6 h, η declines and θEIS rises. Although η is still bigger than θEIS, the fact that these two values are so close to one another after 24 h compared to other times suggests that the corrosion inhibitors have the propensity to switch from horizontal to vertical orientation. This observation can be interpreted as evidence for either the durability of the protective film generated on the MS surface or the existence of more active FV molecules in the solution, which can quickly and efficiently replace the desorbed molecules.
As shown in Table 1, the maximum inhibition efficiency was observed at a concentration of 800 ppm (optimum concentration), whereas raising the concentration to 1000 ppm decreased the polarization resistance and inhibition efficiency. Based on this table, the Cdl value has increased in the solution with 800 ppm FV extract compared to 1000 ppm one, revealing a reduction in double-layer thickness and, as a result, a reduction in the thickness of adsorbed corrosion inhibitor on the surface. The reduction of the adsorbed layer thickness might be a symptom of the inhomogeneous adsorption of corrosion inhibitors on the MS surface. The following approach can be considered to analyze this behavior; Due to the concentration gradient, corrosion inhibitors tend to be adsorbed onto the metal surfaces when introduced to acid solutions. The adsorption process is optimized when the corrosion inhibitor molecules have the least interaction with each other and the most interaction with the metal surface. It is evident that elevating the concentration over the optimal concentration considerably increases the intermolecular attraction leading to the creation of clusters50. Therefore, a passageway might be created for the penetration of corrosive ions (Fig. 4). In fact, the inhibitor's intermolecular interaction at high concentrations to form clusters is more thermodynamically favored, which competes with adsorption forces between the inhibitor and the metal surface preventing them from developing a compact monolayer. Figure 4The corrosion inhibitor adsorption process: at optimum concentration (a) and above optimum concentration (b).
## Polarization
Further investigation on the effect of FV leaves extract on cathodic and anodic reactions was investigated using polarization analysis after 24 h of submersion of MS in the electrolyte without and with various concentrations of the extract. Polarization parameters, comprising corrosion current density (icorr), corrosion potential (Ecorr) and slope of anodic and cathodic branches (βa and βc), are listed in Table 2. According to Fig. 5, it can be realized that in the presence of FV leaves extracts, the current density has decreased in both anodic and cathodic branches compared to the blank. The increase in current density in the anodic branch at -0.3 V/SCE indicates the desorption of corrosion inhibitor molecules from the MS surface at high potential51. This potential is known as the desorption potential. The FV extract seems to have a more significant impact on the anodic reaction mechanism, as observed through changes in the anodic slopes. Meanwhile, the cathodic branch diagrams remain parallel in form, indicating that the cathodic reaction mechanism remains unchanged. The corrosion inhibition efficiency (ξ) can be calculated using Eq. [ 5], where i0 and ii represent the MS corrosion current density in the blank and inhibitor-containing solution, respectively. Table 2Parameters acquired from polarization analysis.βa (mV/dec)−βc (mV/dec)i0 (µA/cm2)E0 (V)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\xi$$\end{document}ξ (%)θPDPBlank80.5109.0264.0−0.446––400 ppm109.3116.2185.1−0.47829.80.298600 ppm93.5100.457.7−0.47978.10.781800 ppm84.5103.820.4−0.47192.20.9221000 ppm99.795.133.5−0.48387.30.873Figure 5Polarization curves of MS submerged in blank and inhibitor-containing solutions.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\xi \%=\left(1-\frac{{i}_{i}}{{i}_{0}}\right) 100\%$$\end{document}ξ%=1-iii$0100\%$ According to Table 2, the corrosion current densities' values have decreased with the addition of corrosion inhibitor, reaching 20.4 µA/cm2 at the optimum concentration (800 ppm), which is significantly lower than the blank (264.0 µA/cm2). Moreover, the inhibition efficiency at this concentration (800 ppm) is 92.2 %, highlighting this inhibitor's excellent ability to protect the MS against harsh species. Besides, since the difference in Ecorr compared to blank is less than 85 mV, it can be concluded that FV extract acts as a mixed corrosion inhibitor52.
## Adsorption isotherms
In order to explore the interaction of FV molecules and MS surface, different adsorption isotherms comprising Frumkin, Temkin, Freundlich and Langmuir were investigated. The adsorption isotherm equations are given below (Eqs. 6–9), and the corresponding diagram is shown in Fig. 6:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{Frumkin}:\mathrm{ ln}\frac{C(1-{\theta }_{Pol})}{{\theta }_{Pol}}= -2\alpha {\theta }_{Pol}-\mathrm{ln}{K}_{ads}$$\end{document}Frumkin:lnC(1-θPol)θPol=-2αθPol-lnKads7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{Temkin}: \mathit{exp}\left(-2\alpha {\theta }_{Pol}\right)={K}_{ads}C$$\end{document}Temkin:exp-2αθPol=KadsC8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{Freundlich}: {\theta }_{Pol}={K}_{ads}{C}^{n}$$\end{document}Freundlich:θPol=KadsCn9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{Langmuir}: \frac{C}{{\theta }_{Pol}}=\frac{1}{{K}_{ads}}+C$$\end{document}Langmuir:CθPol=1Kads+C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\theta }_{Pol}(=\frac{\xi \%}{100})$$\end{document}θPol(=ξ%100), α, and Kads in the above equations symbolize surface coverage based on polarization data, lateral interaction between FV leaves extract and MS surface, and adsorption/desorption equilibrium constant, respectively. According to Fig. 6, the best fit among several adsorption isotherms is the Frumkin isotherm with R2 = 0.9932. The values of Kads and α derived from the intercept and slope of the Frumkin isotherm curve are shown in Table 3.Figure 6Frumkin (a), Temkin (b), Freundlich (c), and Langmuir (d) adsorption isotherms diagrams. Table 3Valuable parameters extracted from Frumkin isotherm. ParametersValueKads (L/g)0.345α1.86 The protective effects of natural extracts are believed to come from the presence of multiple molecules at different contents. One key factor is the synergistic interaction between these components, which increases their inhibiting effect5,7,11. The high concentration of compounds with aromatic rings containing heteroatoms (nitrogen, oxygen, and sulfur) enhances their ability to adsorb onto the metal surface and thus provides protection against corrosive ions such as H+ or Cl−. The possible adsorption mechanism of FV leaves extract molecules on the MS surface is shown in Fig. 7.Figure 7Schematic of the proposed mechanism of adsorption of FV leaves extract on MS surface.
## FE-SEM/EDX
The surface morphology of MS was investigated after submerging an MS coupon in the acidic solution in the absence and presence of 800 ppm FV extract for 6 h. As shown in FE-SEM images, the MS surface, which was exposed to the corrosive solution without inhibitor, is severely damaged (Fig. 8a1,a2); However, adding the FV extract to the acidic solution prevented the formation of corrosion products and a relatively uniform and smooth surface can be seen (Fig. 8b1,b2). Lower damage to the metal surface in the presence of FV extract could be due to the efficient adsorption of corrosion inhibitors, which diminish the contact between the corrosive ions and the MS surface and lessens the dissolution rate53,54.Figure 8FE-SEM images of MS surface soaked in acidic solution without (a1, a2) and with 800 ppm FV extract (b1, b2).
EDX analysis was also performed to demonstrate the elemental composition of the MS surface, which can be seen in Table 4. According to this table, Fe, C, and O elements were detected on the surface of MS. When the FV extract was added to the acidic solution, the percentage of the carbon was increased on the metal surface, confirming the FV-based protective film's creation. Also, the decrease in oxygen percentage on the surface can be due to the reduction of corrosion products29,55.Table 4Elemental composition of MS surface immersed in acid media and the acidic solution containing 800 ppm FV extract for 6 h.Weight %SampleOCFe5.46.688.0Blank2.57.090.5800 ppm
## AFM
AFM analysis was conducted to demonstrate the effect of FV extract on the roughness and microstructure of the MS surface, which was submerged in 1 M HCl solution (Fig. 9). Based on the AFM results, the average roughness (Ra) of the MS coupons in the blank sample was 520.4 nm, while in the presence of FV extract, this value decreased to 111.2 nm suggesting a decline in metal corrosion. Besides the Ra, other parameters such as average height (Hm), peak to valley (Rp-v), and root mean square deviation (Rq) are reported in Table 5. The reduction of these parameters establishes that the FV extract molecules adsorbed on the steel surface limited corrosion attacks on the MS surface56,57.Figure 9AFM micrograph of MS surface in 1 M HCl without (a) and with 800 ppm FV extract (b) after 6 h of immersion. Table 5AFM parameters of MS surface dipped in 1 M HCl solution in the presence and absence of 800 ppm FV extract. SamplesParametersRa (nm)Rq (nm)Rp-v (µm)Hm (µm)Blank520.4684.44.6881.778800 ppm111.2143.31.1150.512
## FTIR
The FTIR spectrum was recorded to characterize the MS surface after immersing in an acidic solution containing FV extract. According to the FTIR spectrum of FV extract, which can be seen in Fig. 10, the peak at 3434 cm−1 is related to O–H stretching58. Moreover, the appearance of peaks at 2856 cm−1 and 2930 cm−1 are linked to the aliphatic CH2 stretching43,58. The peaks related to C=C stretching of the aromatic ring and C=O stretching are observed at 1605 cm−1 and 1715 cm−1, respectively58,59. The peaks between 1295 and 1490 cm−1 are correlated to C-H bending59. Besides, the peaks at 1056 cm−1 and 1260 cm−1 are linked to C–O and C–O–C stretching60. The peaks between 400 and 1000 cm−1 can be due to the aliphatic and aromatic C-H bending30. The presence of the FV extract peaks in the FTIR spectrum of the MS surface reveals that the adsorption of FV extract molecules on the metal surface occurs during immersion. Figure 10FTIR spectrum of the FV powder and MS immersed in the acidic solution containing 800 ppm FV extract after 6 h of immersion.
Additionally, the appearance of a new peak at 476 cm−1 is connected to the Fe–O bonds, revealing the interaction of the oxygen-containing groups of FV extract molecules and metal surfaces61. Furthermore, shifting the peaks of C=O stretching from 1715 to 1728 cm−1 and C=C stretching from 1605 to 1624 cm−1 can be due to the development of a complex between FV extract and iron ions 21. The results showed that functional groups, aromatic rings, and oxygen-containing groups in the FV extract structure could cause the interaction between FV extract compounds and the MS surface.
## UV–Vis
The possibility of complex development among the FV extract compounds and iron ions was studied using the UV–Vis test. Figure 11 shows the UV–*Vis spectrum* before and after soaking the MS in the acidic solution containing FV extract. The UV–*Vis spectrum* of the solution before immersion, included three peaks centered at 206 nm, 265 nm, and 328 nm. Intensive absorption at 206 nm is related to π–π* transitions of C=C bonds in aromatic rings, and the absorption at 265 nm and 328 nm are associated with n-π* transitions of C=O and O–H bonds, respectively61,62. After immersion of the MS in the acidic solution containing the FV extract, the intensity of the absorption peaks related to π–π* and n–π* transitions dramatically decreased, and the UV–*Vis spectrum* shifted to a lower wavelength value (blue shift). These observations may correspond to the adsorption of FV extract molecules on the MS surface and the formation of an organic–inorganic complex via the interaction between pair electrons of FV extract molecule and vacant orbital of Fe2+/Fe3+, which causes the construction of a shielding film against corrosive species63.Figure 11UV–*Vis spectra* of 1 M HCl solution containing 800 ppm FV extract before and after MS immersion.
## GIXRD
GIXRD was recorded to investigate the crystalline composition of the MS surface after soaking in an acidic solution in the presence and absence of FV extract. As shown in Fig. 12 the peak at 2θ = 27.3° is related to γ-FeOOH, and peaks at 2θ = 35.4°, 36.3°, and 36.9° are attributed to Fe2O3/FeCl3 which are due to the presence of corrosion product on the blank specimen surface. Also, peaks at 2θ = 45.2°, and 65.6° are related to iron metal64,65. Comparing the GIXRD patterns shows that the corrosion product peaks disappeared after adding FV extract to the acidic solution. Furthermore, the intensification of the peak at 45.2° and the appearance of a peak at 65.6° (peaks related to Fe metal) can confirm the assumption of the interaction between FV extract compounds and metal surface and verify the production of less corrosion products. Figure 12GIXRD patterns of the MS surface immersed in 1 M HCl (a) and 1 M HCl solution containing 800 ppm FV extract (b) after 6 h of immersion.
## Comparative study
Table 6 summarizes the corrosion inhibition capabilities of several plant extracts in terms of solvent utilized, optimal inhibitor concentration, inhibition efficiency at optimal concentration, and adsorption isotherm. According to this table, the FV leaves extract with an inhibition efficacy above $90\%$ can be designated a robust MS corrosion inhibitor. Moreover, the extraction procedure is considered environmentally friendly because water was employed as the solvent. Table 6Comparison between different plant extract used as corrosion inhibitor for MS in HCl medium. Plant extractExtraction solventOptimum concentration (ppm)Inhibition efficiency at the optimum concentration (%)Adsorption isothermRefSoybeanWater30062Langmuir66Rheum ribes rootWater200085.9Langmuir67Origanum compactumEthanol40090Langmuir68Golpar leavesWater100091Langmuir69Garcinia livingstonei leavesEthanol400096.84Langmuir70CabbageWater10095.66Langmuir2Pomelo peelEthanol800084.07Langmuir71Arbutus unedo L. leavesEthanol/water50091.72Langmuir30Aerva lanata flowersWater60095.07Langmuir72Ceratonia siliqua L seedsChloroform10093.84Langmuir51Thevetia peruviana flowerAcetone/water20091.69Langmuir73Dolichandra unguis-cati leavesEthanol76093.33Langmuir74Falcaria vulgaris leavesWater80091.3FrumkinPresent study
## Conclusion
The current study sought to explore the application of FV extract as a sustainable and effective corrosion inhibitor for MS in hydrochloric acid. The results from EIS and polarization studies indicated that at an optimized concentration of FV (800 ppm), the inhibitor demonstrated a remarkable polarization resistance of 988.1 Ω cm2 and a $91.3\%$ inhibition efficiency after 6 h of immersion. Additionally, there was a substantial reduction in corrosion current density by $92.2\%$ compared to the blank sample. Surface coverage and inhibition efficiency data analyses revealed that the inhibitor was adsorbed horizontally on the metal surface, following the Frumkin adsorption isotherm. Furthermore, SEM and AFM analysis showed a smoother metal surface with fewer corrosion products, while EDX and FTIR verified the formation of an FV-based layer on the metal surface. In conclusion, these results demonstrate the potential of FV extract as an eco-friendly and efficient corrosion inhibitor for MS in hydrochloric acid media.
## References
1. Ghaderi M, Saadatabadi AR, Mahdavian M, Haddadi SA. **pH-sensitive polydopamine–La (III) complex decorated on carbon nanofiber toward on-demand release functioning of epoxy anti-corrosion coating**. *Langmuir* (2022) **38** 11707-11723. DOI: 10.1021/acs.langmuir.2c01801
2. Sun X, Qiang Y, Hou B, Zhu H, Tian H. **Cabbage extract as an eco-friendly corrosion inhibitor for X70 steel in hydrochloric acid medium**. *J. Mol. Liq.* (2022) **362** 119733. DOI: 10.1016/j.molliq.2022.119733
3. Honarvar Nazari M. **Nanocomposite organic coatings for corrosion protection of metals: A review of recent advances**. *Prog. Org. Coatings* (2022) **162** 106573. DOI: 10.1016/j.porgcoat.2021.106573
4. Ouakki M, Galai M, Cherkaoui M. **Imidazole derivatives as efficient and potential class of corrosion inhibitors for metals and alloys in aqueous electrolytes: A review**. *J. Mol. Liq.* (2022) **345** 117815. DOI: 10.1016/j.molliq.2021.117815
5. Khadom AA, Abd AN, Ahmed NA. *S. Afr. J. Chem. Eng.* (2018) **25** 13-21
6. Berrissoul A. **Exploitation of a new green inhibitor against mild steel corrosion in HCl: Experimental, DFT and MD simulation approach**. *J. Mol. Liq.* (2022) **349** 118102. DOI: 10.1016/j.molliq.2021.118102
7. Bouknana D, Hammouti B, Messali M, Aouniti A, Sbaa M. **Phenolic and non-phenolic fractions of the olive oil mill wastewaters as corrosion inhibitor for steel in HCl medium**. *Port. Electrochim. Acta* (2014) **32** 1-19. DOI: 10.4152/pea.201401001
8. Alibakhshi E. **Progress in organic coatings epoxy nanocomposite coating based on calcium zinc phosphate with dual active / barrier corrosion mitigation properties**. *Prog. Org. Coatings* (2022) **163** 106677. DOI: 10.1016/j.porgcoat.2021.106677
9. Hamed S, Yop K, Verma C, Quraishi MA, Ebenso EE. **Challenges and advantages of using plant extract as inhibitors in modern corrosion inhibition systems : Recent advancements**. *J. Mol. Liq.* (2021) **321** 114666. DOI: 10.1016/j.molliq.2020.114666
10. Arash S. **Progress in organic coatings synthesis of methyltriethoxysilane-modified calcium zinc phosphate nanopigments toward epoxy nanocomposite coatings : Exploring rheological, mechanical, and anti-corrosion properties**. *Prog. Org. Coatings* (2022) **171** 107055. DOI: 10.1016/j.porgcoat.2022.107055
11. Bammou L. **Thermodynamic properties of**. *Green Chem. Lett. Rev.* (2010) **3** 173-178. DOI: 10.1080/17518251003660121
12. Ahanotu CC, Madu KC, Chikwe IS, Chikwe OB. **The inhibition behaviour of extracts from**. *J. Mater. Environ. Sci.* (2022) **13** 1025-1036
13. Singh AK, Quraishi MA. **Effect of Cefazolin on the corrosion of mild steel in HCl solution**. *Corros. Sci.* (2010) **52** 152-160. DOI: 10.1016/j.corsci.2009.08.050
14. Li Y, Zhang S, Ding Q, Qin B, Hu L. **Versatile 4, 6-dimethyl-2-mercaptopyrimidine based ionic liquids as high-performance corrosion inhibitors and lubricants**. *J. Mol. Liq.* (2019) **284** 577-585. DOI: 10.1016/j.molliq.2019.04.042
15. Li X, Deng S, Fu H. **Inhibition of the corrosion of steel in HCl, H2SO4 solutions by bamboo leaf extract**. *Corros. Sci.* (2012) **62** 163-175. DOI: 10.1016/j.corsci.2012.05.008
16. Liao LL, Mo S, Luo HQ, Li NB. **Corrosion protection for mild steel by extract from the waste of lychee fruit in HCl solution: Experimental and theoretical studies**. *J. Colloid Interface Sci.* (2018) **520** 41-49. DOI: 10.1016/j.jcis.2018.02.071
17. Hassannejad H, Nouri A. **Sunflower seed hull extract as a novel green corrosion inhibitor for mild steel in HCl solution**. *J. Mol. Liq.* (2018) **254** 377-382. DOI: 10.1016/j.molliq.2018.01.142
18. Chauhan LR, Gunasekaran G. **Corrosion inhibition of mild steel by plant extract in dilute HCl medium**. *Corros. Sci.* (2007) **49** 1143-1161. DOI: 10.1016/j.corsci.2006.08.012
19. Shahmoradi AR. **Theoretical and surface/electrochemical investigations of walnut fruit green husk extract as effective inhibitor for mild-steel corrosion in 1M HCl electrolyte**. *J. Mol. Liq.* (2021) **338** 116550. DOI: 10.1016/j.molliq.2021.116550
20. Sin HLY. *Meas. J. Int. Meas. Confed.* (2017) **109** 334-345. DOI: 10.1016/j.measurement.2017.05.045
21. Shahini MH, Keramatinia M, Ramezanzadeh M, Ramezanzadeh B, Bahlakeh G. **Combined atomic-scale/DFT-theoretical simulations & electrochemical assessments of the chamomile flower extract as a green corrosion inhibitor for mild steel in HCl solution**. *J. Mol. Liq.* (2021) **342** 117570. DOI: 10.1016/j.molliq.2021.117570
22. Mostafatabar AH, Dehghani A, Ghahremani P, Bahlakeh G, Ramezanzadeh B. **Molecular-dynamic/DFT-electronic theoretical studies coupled with electrochemical investigations of the carrot pomace extract molecules inhibiting potency toward mild steel corrosion in 1 M HCl solution**. *J. Mol. Liq.* (2022) **346** 118344. DOI: 10.1016/j.molliq.2021.118344
23. Rani ATJ, Thomas A, Arshad M, Joseph A. **The influence of aqueous and alcoholic extracts of**. *Theor. Electroanal. Stud.* (2022) **346** 117873
24. Dehghani A, Ramezanzadeh B. **Rosemary extract inhibitive behavior against mild steel corrosion in tempered 1 M HCl media**. *Ind. Crops Prod.* (2023) **193** 116183. DOI: 10.1016/j.indcrop.2022.116183
25. Shahini MH, Ramezanzadeh M, Bahlakeh G, Ramezanzadeh B. **Superior inhibition action of the Mish Gush (MG) leaves extract toward mild steel corrosion in HCl solution: Theoretical and electrochemical studies**. *J. Mol. Liq.* (2021) **332** 115876. DOI: 10.1016/j.molliq.2021.115876
26. Khadom AA, Abd AN, Ahmed NA, Kadhim MM, Fadhil AA. **Combined influence of iodide ions and**. *Curr. Res. Green Sustain. Chem.* (2022) **5** 100278. DOI: 10.1016/j.crgsc.2022.100278
27. Nikpour S, Ramezanzadeh M, Bahlakeh G, Ramezanzadeh B, Mahdavian M. *Constr. Build. Mater.* (2019) **220** 161-176. DOI: 10.1016/j.conbuildmat.2019.06.005
28. Khadom AA, Abd AN, Ahmed NA. **Results in chemistry synergistic effect of iodide ions on the corrosion inhibition of mild steel in 1 M HCl by**. *Introduction* (2022) **4** 1-6
29. Ramezanzadeh M, Bahlakeh G, Sanaei Z, Ramezanzadeh B. **Studying the**. *J. Mol. Liq.* (2018) **272** 120-136. DOI: 10.1016/j.molliq.2018.09.059
30. Abdelaziz S. **Green corrosion inhibition of mild steel in HCl medium using leaves extract of**. *Colloids Surf. A Physicochem. Eng. Asp.* (2021) **619** 126496. DOI: 10.1016/j.colsurfa.2021.126496
31. Akinbulumo OA, Odejobi OJ, Odekanle EL. **Thermodynamics and adsorption study of the corrosion inhibition of mild steel by**. *Results Mater.* (2020) **5** 1-6
32. Lashgari SM, Bahlakeh G, Ramezanzadeh B. **Detailed theoretical DFT computation/molecular simulation and electrochemical explorations of**. *J. Mol. Liq.* (2021) **335** 115897. DOI: 10.1016/j.molliq.2021.115897
33. Rashid KH, Khadom AA, Abbas SH. **Optimization, kinetics, and electrochemical investigations for green corrosion inhibition of low-carbon steel in 1 M HCl by a blend of onion-garlic leaves wastes**. *Bioresour. Technol. Rep.* (2022) **19** 101194. DOI: 10.1016/j.biteb.2022.101194
34. Jaberian H, Piri K, Nazari J. **Phytochemical composition and in vitro antimicrobial and antioxidant activities of some medicinal plants**. *Food Chem.* (2013) **136** 237-244. DOI: 10.1016/j.foodchem.2012.07.084
35. Abdulmanea K, Prokudina EA, Lanková P, Vaní L. **Immunochemical and HPLC identification of iso flavonoids in the Apiaceae family**. *Biochem. Syst. Evol.* (2012) **45** 237-243. DOI: 10.1016/j.bse.2012.08.002
36. Rahimi A. **Novel sucrose derivative as a thermally stable inhibitor for mild steel corrosion in 15 % HCl medium : An experimental and computational study**. *Chem. Eng. J.* (2022) **446** 136938. DOI: 10.1016/j.cej.2022.136938
37. Farhadian A. **A theoretical and experimental study of castor oil-based inhibitor for corrosion inhibition of mild steel in acidic medium at elevated temperatures**. *Corros. Sci.* (2020) **175** 108871. DOI: 10.1016/j.corsci.2020.108871
38. El-Azabawy OE. **Studying the temperature influence on carbon steel in sour petroleum media using facilely-designed Schiff base polymers as corrosion inhibitors**. *J. Mol. Struct.* (2022) **1275** 134518. DOI: 10.1016/j.molstruc.2022.134518
39. Singh A, Ansari KR, Alanazi AK, Quraishi MA, Banerjee P. **Biological macromolecule as an eco-friendly high temperature corrosion inhibitor for P110 steel under sweet environment in NACE brine ID196: Experimental and computational approaches**. *J. Mol. Liq.* (2022) **345** 117866. DOI: 10.1016/j.molliq.2021.117866
40. Shahini MH, Keramatinia M, Ramezanzadeh M, Ramezanzadeh B, Bahlakeh G. **Combined atomic-scale/DFT-theoretical simulations & electrochemical assessments of the chamomile flower extract as a green corrosion inhibitor for mild steel in HCl solution**. *J. Mol. Liq.* (2021) **342** 117570. DOI: 10.1016/j.molliq.2021.117570
41. Damej M. **An environmentally friendly formulation based on**. *Colloids Surf. A Physicochem. Eng. Asp.* (2022) **643** 128745. DOI: 10.1016/j.colsurfa.2022.128745
42. Daoudi W. **Essential oil of**. *J. Mol. Liq.* (2022) **363** 119839. DOI: 10.1016/j.molliq.2022.119839
43. Ghaderi M. **Corrosion inhibition of a novel antihistamine-based compound for mild steel in hydrochloric acid solution: Experimental and computational studies**. *Sci. Rep.* (2022) **12** 13450. DOI: 10.1038/s41598-022-17589-y
44. El Basiony NM, Badr EE, Baker SA, El-Tabei AS. **Experimental and theoretical (DFT&MC) studies for the adsorption of the synthesized Gemini cationic surfactant based on hydrazide moiety as X-65 steel acid corrosion inhibitor**. *Appl. Surf. Sci.* (2021) **539** 148246. DOI: 10.1016/j.apsusc.2020.148246
45. Alibakhshi E, Ramezanzadeh M, Bahlakeh G, Ramezanzadeh B. *J. Mol. Liq.* (2018) **255** 185-198. DOI: 10.1016/j.molliq.2018.01.144
46. Cao C. **On electrochemical techniques for interface inhibitor research**. *Corros. Sci.* (1996) **38** 2073-2082. DOI: 10.1016/S0010-938X(96)00034-0
47. Mahdavian M. **Corrosion of mild steel in hydrochloric acid solution in the presence of two cationic Gemini surfactants with and without hydroxyl substituted spacers**. *Corros. Sci.* (2018) **137** 62-75. DOI: 10.1016/j.corsci.2018.03.034
48. Motamedi M, Tehrani-bagha AR, Mahdavian M. **Effect of aging time on corrosion inhibition of cationic surfactant on mild steel in sulfamic acid cleaning solution**. *Corros. Sci.* (2013) **70** 46-54. DOI: 10.1016/j.corsci.2013.01.007
49. Hammouti B, Aouniti A, Taleb M, Brighli M, Kertit S. **L-Methionine methyl ester hydrochloride as a corrosion inhibitor of iron in acid chloride solution**. *Corrosion* (1995) **51** 06. DOI: 10.5006/1.3293606
50. Teymouri F, Allahkaram SR, Shekarchi M, Azamian I, Johari M. **A comprehensive study on the inhibition behaviour of four carboxylate-based corrosion inhibitors focusing on efficiency drop after the optimum concentration for carbon steel in the simulated concrete pore solution**. *Constr. Build. Mater.* (2021) **296** 123702. DOI: 10.1016/j.conbuildmat.2021.123702
51. Abbout S. *J. Mol. Struct.* (2021) **1240** 130611. DOI: 10.1016/j.molstruc.2021.130611
52. Caldona EB. **Corrosion inhibition of mild steel in acidic medium by simple azole-based aromatic compounds**. *J. Electroanal. Chem.* (2021) **880** 114858. DOI: 10.1016/j.jelechem.2020.114858
53. Muthukrishnan P, Jeyaprabha B, Prakash P. **Mild steel corrosion inhibition by aqueous extract of**. *Int. J. Ind. Chem.* (2014) **5** 1-11. DOI: 10.1007/s40090-014-0005-9
54. Wang D, Li Y, Chen B, Zhang L. **Novel surfactants as green corrosion inhibitors for mild steel in 15% HCl: Experimental and theoretical studies**. *Chem. Eng. J.* (2020) **402** 126219. DOI: 10.1016/j.cej.2020.126219
55. Tabatabaei Majd M, Akbarzadeh S, Ramezanzadeh M, Bahlakeh G, Ramezanzadeh B. **A detailed investigation of the chloride-induced corrosion of mild steel in the presence of combined green organic molecules of Primrose flower and zinc cations**. *J. Mol. Liq.* (2020) **297** 111862. DOI: 10.1016/j.molliq.2019.111862
56. Bahlakeh G, Dehghani A, Ramezanzadeh B, Ramezanzadeh M. **Highly effective mild steel corrosion inhibition in 1 M HCl solution by novel green aqueous mustard seed extract: Experimental, electronic-scale DFT and atomic-scale MC/MD explorations**. *J. Mol. Liq.* (2019) **293** 111559. DOI: 10.1016/j.molliq.2019.111559
57. Ostovari A, Hoseinieh SM, Peikari M, Shadizadeh SR, Hashemi SJ. **Corrosion inhibition of mild steel in 1 M HCl solution by henna extract: A comparative study of the inhibition by henna and its constituents (lawsone, gallic acid, α-d-glucose and tannic acid)**. *Corros. Sci.* (2009) **51** 1935-1949. DOI: 10.1016/j.corsci.2009.05.024
58. Kohsari I. **In vitro antibacterial property assessment of silver nanoparticles synthesized by**. *J. Sol-Gel Sci. Technol.* (2019) **90** 380-389. DOI: 10.1007/s10971-019-04961-0
59. Li X-H, Deng S-D, Fu H. **Inhibition by**. *J. Appl. Electrochem.* (2010) **40** 1641-1649. DOI: 10.1007/s10800-010-0151-5
60. Zangeneh MM, Zangeneh A, Pirabbasi E, Moradi R, Almasi M. *Appl. Organomet. Chem.* (2019) **33** e5246. DOI: 10.1002/aoc.5246
61. Majd MT, Ramezanzadeh M, Bahlakeh G, Ramezanzadeh B. **Probing molecular adsorption/interactions and anti-corrosion performance of poppy extract in acidic environments**. *J. Mol. Liq.* (2020) **304** 112750. DOI: 10.1016/j.molliq.2020.112750
62. Li H, Qiang Y, Zhao W, Zhang S. **A green**. *Colloids Surf. A Physicochem. Eng. Asp.* (2021) **616** 126077. DOI: 10.1016/j.colsurfa.2020.126077
63. Haldhar R. **Investigation of plant waste as a renewable biomass source to develop efficient, economical and eco-friendly corrosion inhibitor**. *J. Mol. Liq.* (2021) **335** 116184. DOI: 10.1016/j.molliq.2021.116184
64. Haddadi SA, Ramazani SAA, Mahdavian M, Arjmand M. **Epoxy nanocomposite coatings with enhanced dual active/barrier behavior containing graphene-based carbon hollow spheres as corrosion inhibitor nanoreservoirs**. *Corros. Sci.* (2021) **185** 109528. DOI: 10.1016/j.corsci.2021.109428
65. Keshmiri N, Najmi P, Ramezanzadeh M, Ramezanzadeh B. **Designing an eco-friendly lanthanide-based metal organic framework (MOF) assembled graphene-oxide with superior active anti-corrosion performance in epoxy composite**. *J. Clean. Prod.* (2021) **319** 128732. DOI: 10.1016/j.jclepro.2021.128732
66. Wan S. **Soybean extract firstly used as a green corrosion inhibitor with high efficacy and yield for carbon steel in acidic medium**. *Ind. Crops Prod.* (2022) **187** 115354. DOI: 10.1016/j.indcrop.2022.115354
67. Kaya F, Solmaz R, Geçibesler İH. **Adsorption and corrosion inhibition capability of**. *J. Mol. Liq.* (2023) **372** 121219. DOI: 10.1016/j.molliq.2023.121219
68. Berrissoul A. **Assessment of corrosion inhibition performance of**. *Ind. Crops Prod.* (2022) **187** 115310. DOI: 10.1016/j.indcrop.2022.115310
69. Ghahremani P, Tehrani MEHN, Ramezanzadeh M, Ramezanzadeh B. **Golpar leaves extract application for construction of an effective anti-corrosion film for superior mild-steel acidic-induced corrosion mitigation at different temperatures**. *Colloids Surf. A Physicochem. Eng. Asp.* (2021) **629** 127488. DOI: 10.1016/j.colsurfa.2021.127488
70. Rathod MR, Rajappa SK, Kittur AA. *Mater. Today Proc.* (2022) **54** 786-796. DOI: 10.1016/j.matpr.2021.11.084
71. Yee YP, Saud SN, Hamzah E. **Pomelo peel extract as corrosion inhibitor for steel in simulated seawater and acidic mediums**. *J. Mater. Eng. Perform.* (2020) **29** 2202-2215. DOI: 10.1007/s11665-020-04774-1
72. Hynes NRJ. *Chem. Pap.* (2021) **75** 1165-1174. DOI: 10.1007/s11696-020-01361-5
73. Haque J, Verma C, Srivastava V, Nik WBW. **Corrosion inhibition of mild steel in 1M HCl using environmentally benign**. *Sustain. Chem. Pharm.* (2021) **19** 100354. DOI: 10.1016/j.scp.2020.100354
74. Rathod MR, Rajappa SK, Praveen BM, Bharath DK. **Investigation of**. *Curr. Res. Green Sustain. Chem.* (2021) **4** 100113. DOI: 10.1016/j.crgsc.2021.100113
|
---
title: USP7- and PRMT5-dependent G3BP2 stabilization drives de novo lipogenesis and
tumorigenesis of HNSC
authors:
- Nan Wang
- Tianzi Li
- Wanyu Liu
- Jinhua Lin
- Ke Zhang
- Zhenhao Li
- Yanfei Huang
- Yufei Shi
- Meilan Xu
- Xuekui Liu
journal: Cell Death & Disease
year: 2023
pmcid: PMC9988876
doi: 10.1038/s41419-023-05706-2
license: CC BY 4.0
---
# USP7- and PRMT5-dependent G3BP2 stabilization drives de novo lipogenesis and tumorigenesis of HNSC
## Abstract
GTPase-activating protein-binding protein 2 (G3BP2) is a key stress granule-associated RNA-binding protein responsible for the formation of stress granules (SGs). Hyperactivation of G3BP2 is associated with various pathological conditions, especially cancers. Emerging evidence indicates that post-translational modifications (PTMs) play critical roles in gene transcription, integrate metabolism and immune surveillance. However, how PTMs directly regulate G3BP2 activity is lacking. Here, our analyses identify a novel mechanism that PRMT5-mediated G3BP2-R468me2 enhances the binding to deubiquitinase USP7, which ensures the deubiquitination and stabilization of G3BP2. Mechanistically, USP7- and PRMT5-dependent G3BP2 stabilization consequently guarantee robust ACLY activation, which thereby stimulating de novo lipogenesis and tumorigenesis. More importantly, USP7-induced G3BP2 deubiquitination is attenuated by PRMT5 depletion or inhibition. PRMT5-activity dependent methylation of G3BP2 is required for its deubiquitination and stabilization by USP7. Consistently, G3BP2, PRMT5 and G3BP2 R468me2 protein levels were found positively correlated in clinical patients and associated with poor prognosis. Altogether, these data suggest that PRMT5-USP7-G3BP2 regulatory axis serves as a lipid metabolism reprogramming mechanism in tumorigenesis, and unveil a promising therapeutic target in the metabolic treatment of head and neck squamous carcinoma.
## Introduction
In addition to genetic and epigenetics alterations, aberrant lipid metabolism in cancer is becoming increasingly recognized as a hallmarker [1, 2]. Accumulating evidence revealed that hyperactive fatty-acid synthesis, uptake and storage pathways lead to tumorigenesis. Hence, the inhibition of lipid metabolic pathways might be a potential therapeutic target in cancers. Nevertheless, the association between lipid metabolism and the pathological development of cancer is not well illustrated.
GTPase-activating protein-binding proteins 1 and 2 (G3BP1 and G3BP2) are originally recognized as core components that contributes to stress granules (SGs) assembly [3]. Several lines of study underline the importance of G3BP2 in regulating the formation of SGs, RNA metabolism and human diseases, including cardiac hypertrophy and diabetic nephropathy [4, 5]. Inhibition of G3BP2 prevents tumor growth and augments the accumulation of nuclear p53 in mice [6, 7]. G3BP2 is also required for the translocation of p53 and RanBP2-mediated p53 SUMOylation in prostate cancer cells [8]. In addition, G3BP2-MG53 complex participates in the formation of GSs, which regulates lung tumor progression [9]. To date, no known role for G3BP2 in lipogenesis and its molecular detail functions in specific processes by post-translational modification (PTM) have been previously identified. Much more need to be studied regarding the role of G3BP2 in tumorigenesis.
Protein arginine methyltransferase 5 (PRMT5), a dominant type II protein arginine methyltransferase, plays an important role by interacting with cytoplasmic and nuclear molecules, such as transcription factors, elongation factors and splicing factors [10]. Importantly, PRMT5 exerts oncogenic driver and its dysregulation has been linked to human diseases including malignancies [11–13]. Recent studies revealed that epigenetic modifications arise as important mechanisms involved in reprogramming of metabolic patterns. For instance, PRMT5-methylation of KLF4 and then retards its ubiquitination by pVHL and stabilizes KLF4, thereby enhancing oncogenic signaling in breast cancer [14, 15]. PRMT5-mediated SREBP1 methylation inhibits its phosphorylation by GSK3β, suggesting the PTMs role of PRMT5 in HCC [16]. Despite some studies have highlighted the importance of PRMT5 in cancers, the PTMs role of this enzyme in de novo lipogenesis needs further exploration. Therefore, understanding the regulatory relationship between epigenetic modifications and dysfunctional energy metabolism will aid to elucidate the mechanisms of cancer development.
In the current work, we show that PRMT5 is responsible for G3BP2 accumulation by inducing its binding and deubiquitination by Ubiquitin-Specific Peptidase 7 (USP7). USP7-mediated G3BP2 deubiquitination and stabilization are attenuated by PRMT5 depletion or chemical inhibition of methyl-transferase activity of PRMT5. PRMT5 and USP7 serve as a G3BP2-sensitive ‘switch’ regulating G3BP2 stabilization and lipid metabolism reprogramming, which leads to lipogenesis and aggressive malignancy of head and neck squamous carcinoma (HNSC). PRMT5-USP7-G3BP2 regulatory complex is an essential driver for tumorigenesis.
## Cell lines and transfections
Human embryonic kidney cells (HEK293) were purchased from the type culture collection of the Chinese Academy of Sciences (Shanghai, China). HNSC cells (Tu686, Tu212, CAL-27, SCC25, HSC3) were purchased from Guangzhou Juyan Biological Technology (Guangzhou, China). Cells were maintained in RPMI-1640 or in Dulbecco’s modified Eagle’s medium (DMEM, Gibco). All culture media were supplemented with $10\%$ FBS and $1\%$ Penicillin-Streptomycin in a 37 °C humidified incubator with $5\%$ CO2. All of the cells were authenticated and no mycoplasma contamination was detected in the cell lines used in this study. Wild-type G3BP2, PRMT5, truncated and R468K-mutant G3BP2 plasmids were synthesized by GeneCopoeia (Rockville, MD, USA). Constructs, siRNAs, and shRNAs were transfected into HEK293 and HNSC cell lines in accordance with the instructions of Lipofectamine 3000 reagents, respectively (Life Technologies/Invitrogen). Supplementary Table 1 contains detailed information about the shRNA and siRNA sequences. The reagent used in this study: cycloheximide (CHX) 50 μg/mL, MG132 40 μM.
## Immunoprecipitation and Immunoblotting analysis
The indicated cells were lysed with RIPA buffer (10 Mm Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, and $0.5\%$ Nonidet P-40) supplemented with protease inhibitor cocktail. After incubation on ice for 30 min, 30 μl protein A/G magnetic beads was used to preclear cell extracts, followed by incubation with either IgG (1:100, 3900 S, Cell Signaling Technology), USP7 primary antibody (1:100, GTX125894, GeneTex) or PRMT5 (1:50, 79998, Cell Signaling Technology) antibody overnight at 4 °C. The immune complexes were followed by incubation with 40 μl protein A/G beads for 2 h. Immunoprecipitates were washed and subjected to Western blot, silver staining and mass spectrometry analysis. Silver staining was performed using the Fast Silver Stain Kit (Beyotime) and mass spectrometry was carried out by Novogene, in Beijing, China. Immunoblotting assay was performed as previously described [17]. Antibodies specific to human proteins were anti-PRMT5 (ab109451, Abcam), anti-G3BP2 (16276-1-AP; Proteintech and ab86135; Abcam), anti-Flag (9A3; Cell Signaling Technology), anti-α-Tubulin (T6074, MilliporeSigma), anti-USP7 (4833 S, Cell Signaling Technology), anti-ACLY (4332, Cell Signaling Technology), anti-FASN (3180, Cell Signaling Technology), anti-PPARγ (33436 M, Bioss Antibodies), anti-SCD1 (3787 R, Bioss Antibodies), anti-nSREBP1 (ab159577, Abcam) anti-ELOVL6 (ab69857, Abcam), anti-ACSL3 (ab151959, Abcam) and anti-methy-G3BP2 was a custom-made polyclonal antibody.
## Luciferase reporter analysis
The ACLY and FASN promoters were prepared by PCR and inserted into pGL3-Basic vectors. HEK293 and PRMT5 (KO) cells were grown in DMEM and transiently co-transfected with plasmids containing 3’-UTR of wild-type (WT) or R468K (MUT) from G3BP2 and PRMT5, along with ACLY or FASN luciferase reporter plasmids using Lipofectamine 3000. Luciferase activities were measured 60 h after transfection using the dual luciferase reporter assay system kit (Promega, Madison, USA) and normalized to Renilla luciferase activity, each experiment was repeated in triplicate.
## CRISPR-Cas9 knockout cell line
*To* generate PRMT5-Tu686 knockout cell lines, the sgRNA sequences were cloned into LentiCRISPRv2 plasmid, then the recombinant viral plasmids and viral packaging plasmids (psPAX2 and pMD2G) were co-transfected into HEK293 cells. After transfection for 48 h, the viral supernatants were harvested and filtered through a 0.45 μm filter. Targeted cells were then infected with the viral supernatant and selected with puromycin (1 mg/ml) for 2 weeks. sgRNA sequences targeting PRMT5 and USP7 were designed using the CRISPR designer (http://crispr.mit.edu).
## Cell proliferation and colony-formation assay
For cell proliferation assay, indicated cells were plated in 96-well plates at 1 × 103 cells per well, and then detected using CCK-8 (HY-K0301, MCE) according to the manufacture’s instruction. The absorbance was measured using a Spark® multimode microplate reader (Tecan, Männedorf, Switzerland). For colony formation assay, 500 cells were seeded into 6-well plates and cultured in medium for 10 days to 2 weeks. After macroclones (>50 cells) formed, colonies were fixed with methanol for 30 min and subsequently stained with $0.1\%$ (m/v) crystal violet for 1 h at room temperature. The colonies were captured using microscope.
## GST pull-down assay
GST, GST-PRMT5, GST-PRMT5 truncated recombinant proteins were expressed in *Escherichia coli* BL21-DE3 by induction with 1 mM IPTG (Amersco, SF, USA). GST-tagged proteins were purified using Glutathione Sepharose 4B beads (GE Healthcare). G3BP2-His and G3BP2-His truncated proteins were purified with Ni-NTA agarose beads (Qiagen). Different purified GST- and His-tagged recombinant proteins were incubated with pull-down buffer (50 mM Tris-Cl, pH 8.0, 200 mM NaCl, 10 mM MgCl2, $1\%$ NP-40, 1 mM EDTA,1 mM DTT) for 2 h at 4 °C before immunoblotting assay.
## Co-immunoprecipitation assay
Flag tagged PRMT5 full-length and its truncations were inserted into pcDNA3.1-Flag vector. The human G3BP2 cDNA were inserted into pcDNA3.1-HA vector. Indicated plasmids were transiently transfected into HEK293 cells, and after 48 h of incubation, cells were lysed in immunoprecipitation buffer as described. After washing three times, proteins in elutes were detected with anti-Flag or anti-HA antibody.
## Metabolomics analysis
Tu212 cells with ectopic G3BP2 were collected and flash-frozen in liquid nitrogen for extracting metabolites. Non-targeted liquid chromatography mass spectroscopy (LC-MS/MS) analysis and data preprocessing and annotation were performed at Shanghai Biotree Biotech Co. Ltd. Briefly, about 1000 μl extract sample (acetonitrile: methanol: water = 2: 2: 1) were homogenized and sonicated for 5 min in the ice-water bath. LC-MS/MS analyses were performed using UHPLC system (Vanquish, Thermo Fisher Scientific) with a UPLC BEH Amide column (2.1 × 100 mm, 1.7 μm) coupled to Q Exactive HFX mass spectrometer (Orbitrap MS, Thermo). The QE HFX mass spectrometer was used to acquire MS/MS spectra on information-dependent acquisition (IDA) mode in the control of acquisition software (Xcalibur,Thermo). The raw data were converted to mzXML format using ProteoWizard and processed by R package XCMS (version 3.2). R package CAMERA was used for peak annotation after XCMS data processing. The metabolites annotation were performed by accurate mass search and MS/MS spectral match using an in-house MS2 database, and adjusted P (FDR) values < 0.05 were considered statistically significant.
## In vitro methylation assay
5 µg of recombinant Flag-PRMT5 proteins purified from HEK293 cells or GST-G3BP2-WT or R468K proteins purified from bacteria were incubated with immunoprecipitated Flag-PRMT5 in the presence of adenosyl-L-methionine, S-[methyl-3H]. The reactions were performed in the methylation buffer (50 mM Tris pH 8.0, 20 mM KCl, $0.01\%$ Triton X-100, 10 mM MgCl2, 5 mM β-mercaptoethanol, and 120 mM sucrose) at room temperature for 2 h and stopped by adding 5× SDS loading buffer and was resolved by SDS-PAGE.
## Transwell migration and invasion assay
For migration assay, 5 × 104 indicated cells were plated with 300 μl serum-free medium into the uncoated or Matrigel-coated upper chamber (24-well insert, 8 μm pore size, Corning, USA) for migration or invasion assay. The bottom chamber were filled with 500 μl medium containing $20\%$ FBS. After culturing for 36 h, the migratory or invasive cells were fixed in $4\%$ paraformaldehyde and stained with $0.5\%$ crystal violet solution. Five randomly selected fields from each well were imaged and counted under a microscope (magnification, ×200).
## Wound healing assay
Indicated cells were seeded in 6-well plates at a density of 2 × 106 cell/well and incubated for 20 h. Then, a sterile 10-μl pipette tip was used to make the scratch line. The indicated cells were incubated with serum-free medium for 48 h. The images of each wound were observed by microscope and the wound healing rate was calculated using Image J software (http://rsb.info.nih.gov/ij/).
## Xenograft mouse model and Immunohistochemistry
4–5-week-old nude mice were purchased from Vital River (Beijing, China). The mice were randomly separated into four groups and subcutaneously injected with 1 × 107 Tu686 cells, with five mice per group. Tumor volumes were measured by the formula: length × width2 × 0.52. When the volume of xenograft tumors reached 800 mm3, mice were treated with indicated stable cell lines and shRNA for up to 8 weeks. Tumor volumes were measured every 3 days and at the end of the eighth weekend, and xenograft tumors were extracted and paraffin-embedded for further analysis. The animal study was conducted in accordance with protocols approved by the Animal Ethics Committee of Jiaying University.
Immunohistochemistry assay was performed as previously described [17]. Experiments using human HNSC tissues were approved by the Ethics Committee of the Sun Yat-sen University Cancer Center as described before [17]. Informed consent was obtained from the patients. The sections were incubated with primary antibodies G3BP2 (16276-1-AP; Proteintech, 1:80), PRMT5 (79998, Cell Signaling Technology, 1:50) and Ki-67 (9027, Cell Signaling Technology, 1:800). The IHC results were reviewed by two independent pathologists. The staining of G3BP2, methy-G3BP2, PRMT5 and USP7 were evaluated by IHC scores [18].
## Quantitative real-time PCR assay
RT-PCR analysis was carried out as previously described [17]. The first-strand cDNA were synthesized using a HiScript Q RT SuperMix kit (Vazyme Biotech, Nanjing, China). Quantitative real-time PCR was performed using SYBR Green Master Mix (Vazyme Biotech) on LightCycler 480 (Roche, USA). The expression levels of lipid metabolism-related genes were normalized to GAPDH. Each sample was run in triplicate. Primers used for RT-PCR in this study were listed in Supplementary Table 2.
## Statistical analysis
All statistical analysis were performed using GraphPad Prism 7.0 (GraphPad Software, USA) and SPSS 19.0 (SPSS, USA). Student’s two-tailed t-test was performed to analyze the significance of the differences between two groups. All data were presented as mean ± SD. Survival curves of HNSC patients were plotted by the Kaplan–Meier method. Multivariate Cox regress analyses was performed to determine independent risk factors. All data were collected from at least three independent experiments. P values of less than 0.05 were considered to be statistically significant.
## PRMT5 associates with and methylates G3BP2
Relatively few post-translational modifications of G3BP2 in human cancer cells have been reported. Our previous work identified PRMT5 is highly expressed in laryngeal cancer cells [17], and involved in regulating protein arginine methylation. To explore whether G3BP2 is a novel substrate of PRMT5, immunoprecipitation assay was performed to detect potential binding partners of PRMT5. The subsequent mass spectrometry (MS) analysis identified G3BP2, a RasGAP-binding protein, as a strong PRMT5-interacting partner (Fig. 1A). The physical association between PRMT5 and G3BP2 was validated by co-immunoprecipitation in HEK293 cells (Fig. 1B, C). In addition, we performed western blot analysis to compare the expression levels of G3BP2 and PRMT5 in HNSC cell lines (Supplementary Fig. S1A, B) and confirmed this interaction between PRMT5 and G3BP2 in Tu686 and Tu212 cells (Fig. 1D).Fig. 1PRMT5 interacts with and methylates G3BP2.A PRMT5 plasmid was transfected into HEK293 and Tu686 cells. After incubated for 48 h, cell lysates were measured by Immunoprecipitation (IP) assays using anti-PRMT5 beads, and the silver staining showed the location of PRMT5 and its associated protein G3BP2. Lane 1,3 for HEK293 cell. Lane 2,4 for Tu686 cell. B, C Co-immunoprecipitation (Co-IP) of PRMT5 and G3BP2 was performed in HEK293 cells. IgG was used as a negative control. D Association of PRMT5 with G3BP2 were verified by Co-IP assay with anti-Flag in Tu212 cells transfected with Flag-PRMT5 or Flag-G3BP2, respectively. E Schematic diagram showed the structure of PRMT5 (left) and the deletion constructs were co-transfected with Flag-G3BP2 into HEK293 cells. Cell lysates were precipitated with GST beads. G3BP2 was blotted with an anti-Flag antibody. Immunoblotting and Coomassie Brilliant blue staining were shown. F The methylation of G3BP2 was detected by western blot analysis using a custom-made methy-G3BP2 antibody in HEK293 and Tu212 cells transfected with or without PRMT5 plasmid. G Methylation site of G3BP2 was examined by liquid chromatography- mass spectrometry (LC-MS). R represents potential methylation site. H Sequences of the evolutionarily conserved residue R468 (red) in G3BP2. I Validation of R468 as the PRMT5-catalyzed G3BP2 methylation site in Tu212 and Tu686 cells. J In vitro methylation of G3BP2 in the presence of 3H-SAM. Recombinant GST-G3BP2-WT and G3BP2-R468K proteins were purified from bacteria and Flag-PRMT5 proteins were immunopurified from HEK293 cells. K Tu686 cells were treated with the indicated amounts of GSK3326595 for 24 h, protein levels of G3BP2 and methy-G3BP2 were assessed by western blot.
Next, we performed a series of truncated fragments to examine which domain of PRMT5 is responsible for the binding with G3BP2. *We* generated three GST-tagged truncated PRMT5 for pull-down assays, which is composed from aa 1 to 292, 293 to 420, and 421 to 637 region. The results showed that both the FL-PRMT5 and the construct with enzymatic activity region (421-637aa) could specifically interact with G3BP2 (Fig. 1E). These data strongly demonstrate that PRMT5 interacts with G3BP2, and the enzymatic activity region of PRMT5 is required for G3BP2-PRMT5 interaction.
Since GR and/or GRG repeats represent potential preferred methylation sites of PRMT5 [19], we investigate whether G3BP2 could also in fact be methylated by PRMT5. Surprisingly, PRMT5 could methylate G3BP2 as detected by a custom-made methy-G3BP2 antibody that specifically recognizes symmetric dimethyl R468 (Fig. 1F). The subsequent mass spectrometry analysis of immunopurified G3BP2 protein from HEK293 and Tu686 all suggest that arginine-468 (R468) residue was a symmetric dimethylation site (Fig. 1G and data not shown). Of note, G3BP2-R468 is evolutionarily conserved from *Rattus norvegicus* to Homo sapiens (Fig. 1H). Next, we verified the PRMT5 methylation site on G3BP2 by co-transfected with Flag-PRMT5 and G3BP2 into HEK293, in which the levels of G3BP2 methylation were detected. As shown in Fig. 1I, a construct containing the Arg residues at position 468 is important for PRMT5 methylates G3BP2, as R468K mutation markedly attenuated PRMT5-mediated methylation of G3BP2 compared to WT, indicating that the C-terminal GRG repeats of G3BP2 are responsible for its interaction with PRMT5. In in vitro methyltransferase assays, R468K mutants also showed lower G3BP2 methylation compared to WT (Fig. 1J). Moreover, in the presence of GSK3326595 [20], a specific inhibitor of PRMT5 methyltransferase, methylation of G3BP2 is dramatically reduced (Fig. 1K and Supplementary Fig. S1C). Taken together, these results indicated that G3BP2 interacts with and methylates by PRMT5 in an enzyme activity-dependent manner.
## PRMT5-dependent methylation of G3BP2 promotes its deubiquitination by USP7
To investigate the mechanism of which PRMT5-mediated G3BP2 methylation affects G3BP2 expression and function, we inhibited PRMT5 expression in Tu686 and Tu212 cells. We found that knockdown PRMT5 resulted in a decrease in G3BP2 protein but not affect G3BP2 mRNA levels (Fig. 2A, B). In addition, the deubiquitinase (DUB) that protects ubiquitinated G3BP2 from degradation remains elusive. To assess the role of G3BP2 in proteasome degradation, we used the proteasome inhibitor MG-132 to determine the effect of PRMT5 depletion on G3BP2 (Fig. 2C). Likewise, we knocked down PRMT5 using specific short interfering RNAs (siRNA) and measured the half-life of G3BP2 protein. The half-life of G3BP2 protein was shorter in PRMT5-depleted cells than in the control in Tu686 cells (Fig. 2D). Furthermore, we detected the effect of PRMT5 on ubiquitination of G3BP2 by overexpressing indicated fragments into Tu212 cells. Compared with control vector and truncated group, transfection of PRMT5-WT and the construct with enzymatic region fragments significantly inhibited G3BP2 ubiquitination (Fig. 2E). However, knockdown of PRMT5 led to an enhancement of G3BP2 ubiquitination in Tu686 cells (Fig. 2F). Thus, the above results show that PRMT5 can stabilize G3BP2 by an ubiquitin-mediated pathway. Fig. 2PRMT5 and UPS7 stabilizes G3BP2 by deubiquitination pathway. A, B Western blot and qPCR analysis of G3BP2 and PRMT5 expression in Tu686 and Tu212 cells transfected with siNC or siPRMT5. ns, no significant difference, **$P \leq 0.01.$ C Tu686 cells were transfected with PRMT5 siRNAs and then incubated with or without MG132 (40 μM) for 6 h. Cell lysates were analyzed by immunoblotting. D Tu686 cells were transfected with negative control or PRMT5 siRNA and then applied with Cycloheximide (CHX, 50 μg/ml) for 2, 4 or 6 h. Immunoblotting analysis was used to measure the expression of G3BP2. * $P \leq 0.05.$ E Cell lysates were immunoprecipitated with Flag-tag antibody and then immunoblotted by HA-tag antibody. F Cell lysates were immunoprecipitated with G3BP2 antibody before immunoblotting with HA-tag antibody. G HEK293 cells were transfected different deubiquitinating enzymes (DUBs) and then lysed for immunoblotting to detect the expression of G3BP2. H HEK293 cells were co-transfected with the indicated vectors for 48 h, followed by treated with MG132 for 6 h. Cell lysates were immunoprecipitated with anti-Flag antibody and immunoblotting with anti-Myc antibody. I, J Western blot and RT-PCR analysis of USP7 expression in Tu212 and Tu686 cells transfected with siNC or siUSP7. ns, no significant difference. K Tu686 cells were transfected with negative control or USP7 siRNA and then applied with 50 μg/ml CHX for the indicated times and cell lysates were assessed by immunoblotting. * $P \leq 0.05.$ L Western blot analysis of Tu686 and Tu212 cells with or without P5091, followed by treatment with DMSO and MG132 for 6 h, respectively. M Tu212 cells were transfected with G3BP2 and HA-Ub plasmids for 48 h, the purified G3BP2-Ubn was added 40 ng or 80 ng recombinant GST-USP7 proteins before immunoblotting analysis. N USP7 knockdown in Tu686 cells increased G3BP2 ubiquitination. Tu686 cells were co-transfected with HA-Ub and USP7 siRNA or control siRNAs, followed by treated with MG132 for 6 h. Cell lysates were immunoprecipitated using an anti-G3BP2 antibody and then subjected to immunoblotting.
Based on the result that Arg-468 methylation decreased G3BP2 ubiquitination, we hypothesized that PRMT5 functions with an unknown DUB at post-transcriptional level to regulate G3BP2 expression. Combined with the LC-MS/MS analysis, we transfected a list of DUBs cDNA plasmids into HEK293 cells, and surprisingly found that USP7, USP8, USP39 upregulated G3BP2 expression (Fig. 2G). Of note, only USP7 of these DUBs decreased G3BP2 ubiquitination (Supplementary Fig. S2A, B). Similar observation was obtained under denaturing conditions (Fig. 2H).
Considering USP7 stabilizes its substrates, including p53, Sirt1, MDM2 [21, 22], by promoting their deubiquitination, we speculated that USP7 may affect the ubiquitination of G3BP2. To test this, we detected the effect of USP7 on G3BP2 stability. In Tu686 and Tu212 cells, USP7 depletion decreased the G3BP2 protein level, while the mRNA expression was nearly unchanged (Fig. 2I, J). The reduction of G3BP2 protein by USP7 knockdown could be due to the decreased G3BP2 stability (Fig. 2K), suggesting that USP7 stabilized G3BP2 by inhibiting its degradation through proteasome. We then evaluated whether P5091, an inhibitor of USP7, decrease G3BP2 expression. Consistently, MG132, a proteasome inhibitor, blocked P5091-induced G3BP2 decrease in both cell lines (Fig. 2L). In addition, we found that G3BP2 ubiquitination was reduced when incubating with recombinant USP7 in vitro deubiquitination analysis (Fig. 2M). Moreover, knockdown of USP7 increased G3BP2 ubiquitination in Tu686 cells (Fig. 2N). Taken together, these results suggest that USP7 regulates the stability of G3BP2 by inhibiting proteasomal degradation.
## G3BP2 deubiquitination depends on its methylation by PRMT5
In order to confirm whether G3BP2 deubiquitination by USP7 depends on PRMT5-mediated G3BP2 methylation. *We* generated three truncated fragments of G3BP2, and the arginine-glycine rich motif, from 287 to 482 aa (contain PRMT5 methylation site) was required for G3BP2 interacts with USP7 (Fig. 3A), suggesting a potential link between G3BP2 methylation and deubiquitination. Accordingly, PRMT5 depletion in Tu686 cells retarded the interaction between USP7 and G3BP2 (Fig. 3B, C). G3BP2-R468K mutant significantly decreased the USP7-G3BP2 interaction in the presence of PRMT5 (Fig. 3D). In addition, the level of G3BP2 deubiquitination was attenuated after PRMT5 knockdown (Fig. 3E). Similarly, G3BP2-R468K mutation reduced the G3BP2 deubiquitination by USP7 in the presence of PRMT5 (Fig. 3F). These results demonstrate that PRMT5-dependent G3BP2 methylation is critical for G3BP2 interaction with and deubiquitination by USP7. Overall, G3BP2 acts as a central epigenetic regulatory role in the lipogenesis and PRMT5-driven laryngeal tumorigenesis. Fig. 3PRMT5 augment G3BP2 binding with and deubiquitination by methylation. A A series of G3BP2 constructs were co-transfected with HA-USP7 plasmid into HEK293 cells. Cell lysates were immunoprecipitated with anti-Flag antibody and then analyzed by immunoblotting with HA-USP7 antibody. B, C Tu686 cells were transfected with either negative control siRNA or PRMT5 siRNA, then treated with 40 μΜ MG132 for 6 h. Cell lysates were immunoprecipitated with anti-USP7 or G3BP2 antibody, followed by immunoblotting analysis. D Flag-G3BP2 wild type or R468K mutant was co-transfected with PRMT5 and HA-USP7 into HEK293 cells, and then treated with MG132 for 6 h. Cell lysates were immunoprecipitated with anti-HA antibody then analyzed by immunoblotting using anti-Flag antibody. E Cells were transfected with Myc-Ub and HA-USP7 and negative control siRNA or PRMT5 siRNA, followed by treated with MG132 for 6 h. Cell lysates were immunoprecipitated with anti-G3BP2 antibody then analyzed by immunoblotting. F Flag-G3BP2 wild type or R468K mutant, Myc-Ub and PRMT5 were co-transfected into HEK293 cells, and then treated with MG132 for 6 h. Cell lysates were immunoprecipitated using anti-Flag antibody and then analyzed by immunoblotting using anti-Myc antibody.
## PRMT5-G3BP2 complex activates lipid metabolism reprogramming
To explore the biological functions of methylation G3BP2R468 on tumor cells, KEGG pathway combined untargeted metabolomics analysis were performed. As expected, Glycerophospholipid metabolism pathway belonged to lipid metabolism process was significantly enriched (Fig. 4A), suggesting that G3BP2 plays a crucial role in the lipogenesis of HNSC cells. Furthermore, RNA-seq indicated that G3BP2 conferred lipid metabolism, such as fatty acid metabolism, fatty acid elongation, biosynthesis of unsaturated fatty acids and central carbon metabolism in cancer (Fig. 4B), indicating that G3BP2 globally functions in the metabolic reprogramming of carcinomas. Then, we applied qPCR and immunoblotting analysis to confirm the effect of G3BP2 and G3BP2R468 on the expression of lipid metabolism-related genes. As shown in Fig. 4C, G3BP2 upregulated the mRNA levels of ACLY, FASN, ACSL3, SCD1, and PPARγ in Tu212 cells. Immunoblotting analysis demonstrated that G3BP2 increased the protein levels of FASN, ACLY, PPARγ, SCD1, suggesting G3BP2-WT displayed a more stronger effect than G3BP2R468 in the de novo fatty-acid biosynthesis (Fig. 4D). Next, we explored whether these lipogenic genes transcription activity were regulated by PRMT5 via the modification on R468. Consistent with previous research in lung adenocarcinoma [23], we found knockdown PRMT5 attenuated the activity of SREBP1 in Tu686 cells. Conversely, ACLY and FASN luciferase activity was dramatically increased upon ectopic expression of PRMT5 wild type but not the enzymatic inactive mutant (Fig. 4E and Supplementary Fig. S3A, B). In addition, either G3BP2-WT or R468K-mediated ACLY and FASN transcription activity was compromised by PRMT5 depletion (Fig. 4F, G). Similarly, Real-time PCR analysis showed that both of the target lipogenic genes were elevated by G3BP2-WT but not G3BP2-R468K when PRMT5 depletion (Supplementary Fig. S3C, D). These data suggested that PRMT5-G3BP2 interaction as a major co-activator complex to be recruited to the promoter of lipogenic genes for lipid metabolic reprogramming. Fig. 4Methylation of G3BP2 by PRMT5 confers lipid metabolism reprogramming. A The KEGG pathway analysis of the lipid metabolism-associated pathway affected by G3BP2 in Tu212 cells. B The KEGG pathway enrichment analysis of G3BP2-conferred metabolism pathways in Tu212 cells. C The effect of G3BP2-WT or G3BP2-R468K on the expression of lipid metabolism-associated genes were measured by RT-PCR analysis in Tu212 cells. D The effect of G3BP2-WT and G3BP2-R468K on the lipid metabolism-associated proteins were measured by immunoblotting analysis. E ACLY and FASN luciferase activity were determined by co-transfected with PRMT5-WT/PRMT5-MUT and pSV-Renilla to Tu686 cells. Luciferase activities were measured 60 h later. ACSL3 gene was used as an negative control. F, G The G3BP2-WT or G3BP2-R468K along with ACLY and FASN luciferase reporter plasmids were co-transfected into HEK293 or Tu686-PRMT5(KO) cells for 60 h. The luciferase activities were analyzed by dual-luciferase reporter assay, and normalized to the activity of Renilla. The data are presented as the mean ± SD; ns no significant difference; *$P \leq 0.05$; ** $P \leq 0.01$,***$P \leq 0.001.$
We next tested the possible regulation of G3BP2 methylation to lipid metabolism. As expected, G3BP2-WT elevated the level of triglycerides and fatty acids, but not cholesterols in Tu686 cells, supporting that methylation of G3BP2 induces lipid metabolism reprogramming in tumor cells (Fig. 5A). PRMT5 knockout efficiency were also identified (Supplementary Fig. S4A, B). Similarly, ectopic expression of G3BP2-WT and PRMT5 elevated the triglycerides and fatty acids, but not intracellular cholesterols when compared with the G3BP2R468K group in PRMT5(KO)-Tu686 cells (Fig. 5B). To further confirm the function of G3BP2 in lipid accumulation and lipid droplets (LDs) formation, we performed Oil red O staining analysis in PRMT5 knockout Tu686 and CAL-27 cells. We found that lipid droplets formation were dramatically decreased and less abundant when G3BP2 inhibition in Tu686 and CAL-27 cells, while there were more lipid droplets formation in PRMT5(KO)-Tu686 and CAL-27 cells with G3BP2 overexpression (Fig. 5C, D and Supplementary Fig. S4C, D). Consistent with that, co-transfection of G3BP2-WT along with PRMT5 enhanced the levels of lipid droplets, suggesting that methylation of G3BP2 R468 was required for PRMT5-G3BP2 complex mediated accumulation of lipid droplets formation (Fig. 5E, F). The above results demonstrating the synergistic activation of lipogenesis by PRMT5 and G3BP2.Fig. 5G3BP2 methylation contributes to PRMT5-induced lipids synthesis. A, B The levels of triglycerides, fatty acids and cholesterols were measured after transfected with indicated plasmids, such as pcDNA3.1-G3BP2-WT, pcDNA3.1-G3BP2-R468K, pcDNA3.1-PRMT5, pcDNA3.1-PRMT5 + G3BP2-WT or pcDNA3.1-PRMT5 + G3BP2-R468K, in PRMT5-knockout Tu686 cells. C, D The effect of G3BP2 inhibition or overexpression on lipogenesis was determined by Oil Red O staining in Tu686 and PRMT5-knockout Tu686 cells. Quantification of lipid droplets were performed by Image J. Scale bar: 75 μm, 25 μm. E, F Representative images and quantification of G3BP2 or G3BP2R468K with or without PRMT5 on lipogenesis by Oil Red O staining in PRMT5-knockout cells. Scale bar: 75 μm. PRMT5i for PRMT5 inhibitor. The data are presented as the mean ± SD; ns no significant difference, *$P \leq 0.05$; **$P \leq 0.01$,***$P \leq 0.001.$
## Stabilization of G3BP2 by PRMT5 and USP7 promotes tumorigenesis of carcinoma cells
Previous studies have shown that higher levels of G3BP2 status correlated with malignant progression and poor prognosis [7], we sought to confirm if it also occurs in a methylation-dependent way. Hence, the function of PRMT5-G3BP2 interaction on tumorigenesis was further explored. In the USP7-stable Tu686 cell lines, CCK-8 and colony formation assays showed that the ectopic expression of G3BP2-WT increased PRMT5-mediated cell proliferation compared with the G3BP2-MUT (Fig. 6A, B). Additionally, the enhanced proliferation induced by overexpression of G3BP2-WT was blocked when the absence of PRMT5, suggesting that the methylation of G3BP2 is involved in the event of PRMT5 promotes carcinoma cell growth (Fig. 6C, D). Moreover, transwell and wound healing assays were used to confirm this observation. G3BP2-WT and PRMT5 co-transfected cells dramatically promoted the migration and invasion, whereas the G3BP2-R468K + PRMT5 and G3BP2-WT + siPRMT5 group cells presented a little bit faster migration capacity than G3BP2-WT cells (Fig. 6E, F). Meanwhile, we measured a series of metastasis-related genes, which were reported to be associated with epithelial-mesenchymal transition (EMT). The results revealed that G3BP2-R468K and PRMT5 apparently abolished the expression levels of EMT-related genes (Fig. 6G). Thus, we conclude that methylation of G3BP2 R468 promotes the proliferation and migration of cancer cells in vitro. Fig. 6PRMT5-USP7-G3BP2 axis is required for tumorigenesis of HNSC.A, C Cell viability was analyzed by Cell counting kit-8 (CCK-8). B, D Colony formation assays were performed to evaluate the proliferation ability of the cells with indicated treatments. Cells were seeded into six-well plates at a density of 1000 cells/well, and cultured for 12 d, and then stained with crystal violet. The colonies were captured and counted. E Transwell migration and invasion assays were used to estimate the migration and invasion ability of G3BP2-WT and G3BP2-R468K with or without PRMT5 expression. Representative pictures are shown on the left, and quantifications on the right. Scale bar: 100 μm. F Representative pictures shown the wound healing assays of G3BP2-WT and G3BP2-R468K with or without PRMT5 expression. Scale bar: 100 μm. G *Immunoblotting analysis* validated the protein expression levels of EMT and lipid metabolic-related markers. H Representative images of tumor-bearing nude mice. Tu686 cells were treated as indicated, such as EV, shG3BP2, shUSP7, and G3BP1 + shUSP7, and then injected into nude mice ($$n = 5$$). I Tumor diameters were measured every 3 days, and tumor volumes were calculated. J Average tumor weight of the xenografts in each group were weighed. K Representative images of tumor-bearing nude mice. G3BP2-WT and G3BP2-R468K overexpression Tu686 cells with shPRMT5 were respectively injected into nude mice ($$n = 5$$). L, M Tumor diameters and tumor volumes were calculated. N The levels of proliferation marker of Ki-67 were examined in different groups of tissues. Scar bar: 200 μm; 50 μm. O The H score of Ki-67 in different groups. P Representative images of Oil red O staining of tumor tissues from nude mice. scale bar: 100 μm; 25 μm. Q The quantification of the number of lipid droplets per cell among 200 cells in different groups. The data are presented as the mean ± SD *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$
A xenograft model was established to further explore whether PRMT5 and USP7 regulate the tumorigenesis through G3BP2. To this end, we performed G3BP2-WT and the methylation-deficient mutant G3BP2-R468K stable cell lines in Tu686 cells. Notably, depletion of G3BP2 and USP7 independently retarded tumor growth, but the efficacy of USP7 silencing was rescued by G3BP2 expression (Fig. 6H–J). Next, we pinpoint the role of PRMT5-dependent methylation of G3BP2-R468 in vivo. Compared to G3BP2-R468K mutation, G3BP2-WT dramatically rescued the inhibitory effect of PRMT5 silencing on tumor growth (Fig. 6K–M). IHC analysis was performed to observe the proliferation of tumors by using an antibody against Ki-67. As expected, tumors derived from G3BP2-R468K group were less proliferative than those from G3BP2-WT (Fig. 6N, O and Supplementary Fig. S5). Oil red O staining assay was carried out to evaluate the lipid accumulation in tumors, whereas G3BP2-WT rescued PRMT5 silencing-mediated intracellular lipid accumulation (Fig. 6P, Q). Altogether, PRMT5-mediated G3BP2-R468 methylation is an important step for its activation and oncogenic function in HNSC cell models.
## G3BP2 expression is correlated with PRMT5 and USP7 in HNSC specimens
The results of G3BP2 methylation by PRMT5 promotes HNSC cells de novo lipogenesis and growth capacity prompted us to explore whether the methylation of G3BP2R468, PRMT5, and USP7 are positively correlated in HNSC tissues. 50 pairs of HNSC samples with adjacent normal tissues were analyzed by IHC staining using an antibody against R468-G3BP2. As shown in Fig. 7A, the majority of HNSC biopsies displayed PRMT5, USP7 and G3BP2-positive staining in both cytoplasmic and nuclei of HNSC tissue cells, whereas weak diffused in adjacent normal HNSC tissue cells. H score analysis showed a significant higher expression levels of PRMT5, USP7 and methylated G3BP2 in HNSC tissues compared with adjacent normal tissues (Fig. 7B, C). Strikingly, we observed the levels of methylated G3BP2 was correlated with tumor grade of HNSC specimens, and survival analysis suggested that elevated levels of methy-G3BP2 associated with poor overall survival (Fig. 7D, E). Similarly, we validated that the expression of G3BP2 was higher in HNSC than normal tissues in the TCGA-HNSC cohort, and high G3BP2 expression associated with tumor grade (Fig. 7F, G). Likewise, the USP7 and PRMT5 expression was also increased in the TCGA-HNSC tissues than normal tissues (Supplementary Fig. S6A, B). Moreover, the relationship between methy-G3BP2 expression and clinicopathological factors was summarized in Supplementary Table 3. In addition, Pearson’s correlation analysis was performed to validate the association of PRMT5, USP7, and G3BP2 in TCGA-HNSC samples (Fig. 7H). Immunoblotting assays verified the upregulation of methy-G3BP2, PRMT5, USP7 and lipid metabolism-related proteins in HNSC samples. The methy-G3BP2 levels was positive correlated with the levels of ACLY, but not FASN (Fig. 7I). Oil red O staining assay was performed to estimate the intracellular lipid droplet formation, the accumulation lipid droplets were significantly much more abundant in the tumors (Fig. 7J). All these findings suggested that PRMT5-dependent methylation of G3BP2 is partly responsible for the tumorigenicity of HNSC. Dysregulated G3BP2 expression in HNSC is driven by high-level PRMT5 and USP7. Based on these results, we propose an epigenetic de novo lipogenesis and tumorigenesis model (Fig. 7K). The ectopic expression of PRMT5 interacts with G3BP2, G3BP2 in turn deubiquitination and stabilization by USP7, and then recruitment of methy-G3BP2-PRMT5-USP7 complex to lipogenic gene promoters, thereby activating lipogenesis and tumorigenesis. Fig. 7Methylated G3BP2 is up-regulated in HNSC tissues and predicts poor prognosis. A Representative IHC images of methy-G3BP2, USP7, and PRMT5 proteins in HNSC tumor samples. Scale bar: 100 μm. Histogram shows the H score of methy-G3BP2 with PRMT5 (B) and USP7 (C), note that the scores of some samples overlapped. D IHC analysis estimate the expression of methy-G3BP2 in different tumor grade of HNSC. E Kaplan–Meier analyses of HNSC specimens survival stratified by the expression of methy-G3BP2. F mRNA expression of G3BP2 in HNSC tissues and normal tissues from the TCGA database. G Correlation between G3BP2 expression and tumor grade of HNSC patients in the TCGA cohort. H Pearson correlation coefficient between G3BP2, USP7, and PRMT5 in TCGA-HNSC cohort. I *Immunoblotting analysis* of indicated proteins expression level in HNSC and corresponding adjacent non-tumor tissues. J Representative images and quantification of the number of lipid droplets per cell among 200 cells of Oil red O staining in tumor and non-tumor tissues. ** $P \leq 0.01$, ***$P \leq 0.001.$ K Proposed model to describe the role of G3BP2-R468 methylation on promoting USP7- and PRMT5-dependent lipogenesis and tumorigenesis.
## Discussion
Hyperactivation of the metabolic reprogramming, including de no lipogenesis pathway is a hallmark of human cancers, which can be caused by genetic alterations or post-translational modifications [24, 25]. Notably, increased lipid synthesis in HNSC has been observed; however, the underlying molecular mechanisms remain not fully investigated. In this study, we performed experimental evidence showing that USP7-mediated G3BP2 deubiquitination and stabilization is attenuated by PRMT5 depletion or chemical inhibition of methyl-transferase activity of PRMT5. PRMT5 and USP7 serve as a G3BP2-sensitive ‘switch’ in regulating its stabilization and lipid metabolism reprogramming, which leads to lipogenesis and aggressive malignancy of HNSC. The PRMT5-USP7-G3BP2 signal is critical for tumorigenesis, which provides a potential rationale for further pharmacological study as well.
G3BP2 was preliminary identified as an androgen-responsive gene and consists of a nuclear transport factor 2-like domain, an RNA recognition motif, and an arginine and glycine rich region [26]. Of note, G3BP2 is relatively poorly documented when compared with G3BP1, another G3BPs family members. In the current study, we confirmed that G3BP2 binds to PRMT5 via its arginine and glycine rich box motif. The C-terminal GRG repeats of G3BP2 are required for their interactions. As RNA binding proteins have been engaged in posttranscriptional modification by modulating RNA splicing, mRNA localization and stability [27, 28]. Therefore, RNA binding proteins are essential for maintaining homeostasis of gene expression. As a RNA binding protein, abnormal expression of G3BP2 exerts critical roles in gene expression homeostasis and cause tumor progression [4, 8, 29]. In the present study, we demonstrate that G3BP2 was methylated at R468 to allow direct interaction with USP7 and via an enhance in ACLY and FASN transcription levels, thereby methylation-stabilized G3BP2 increased de novo lipogenesis and accelerated the growth of HNSC cells in vivo and in vitro. Moreover, we identified G3BP2 as an independent prognostic marker in HNSC, consistent with prior reports in prostate cancer, breast and lung cancer [8, 9, 30]. Importantly, we suggest G3BP2 is involved in PRMT5-dependent activation of the lipogenic promoter in HNSC cells, and the repression of G3BP2 decreases triglyceride levels. We ascertained G3BP2-dependent growth induction was eliminated by USP7 inhibition. G3BP2 methylation by PRMT5 promotes G3BP2 binding with and then deubiquitination by USP7, which provides a novel epigenetic link between substrates methylation and deubiquitination.
As a well-known oncogenic enzyme, PRMT5 is ubiquitously expressed in the cytoplasm and nucleus of carcinoma cells. The function of PRMT5 in malignancy is due to the context with substrate and tumor types. PRMT5 dysregulation has been involved in cellular hyperproliferation, metastasis, apoptosis, EMT and differentiation [31]. We previously reported that PRMT5 is highly expressed in laryngeal carcinoma, of which the most malignant subtype of HNSC [17]. PRMT5 catalyzes a series of substrates, including histone and non-histone proteins [32–34]. Its methylation modification of arginine residues has been linked to several cellular processes, such as cell growth, differentiation, and lipid metabolism [35, 36]. PRMT5 arginine methylation of p53 gene was required in altered nuclear localization and activation in promoting lymphomagenesis [37, 38], therefore, we speculated if PRMT5 was involved in the lipogenesis through epigenetic modification. Through mRNA-seq and KEGG analysis, we suggested that overexpression PRMT5 activates head and neck squamous carcinoma lipid metabolism pathway. Here, we found ectopic expression of PRMT5 could interact and methylate G3BP2 dramatically increased HNSC progression. The data from the TCGA database in addition to our cohort strongly show that PRMT5 and G3BP2 are frequently up-regulated in HNSC tissues. Notably, our data show that depletion the enzyme activity of PRMT5 by GSK3326595, which is being tested in clinical trials, dramatically abrogate the methylation of G3BP2. These findings indicate that G3BP2 methylated by PRMT5 is in an enzyme activity-dependent manner. Consistent with our results, both PRMT5 and G3BP2 have emerged as crucial therapeutic targets for several diseases [39–41]. However, it is expected that whether different types of PTMs on the same protein will exhibit different effects, which ensure cancer cells effectively coordinate metabolic regulation in order to maximize their survivability.
Ubiquitination, a reversible process, is driven by a series of enzymes in most eukaryotic cells. Protein deubiquitination is well certified to be reserved by deubiquitinating enzymes. Regulation of G3BP2 ubiquitination and deubiquitination in cancer cells is of great interest but remains unclear. Our present study identified that G3BP2 methylation affects its recognition by deubiquitinating enzymes (DUBs). USP7, a member of ubiquitin-specific processing proteases, was regarded as a biomarker for predicting metastasis and recurrence of several malignant tumors [42–44]. USP7 regulates p53 and its E3 ligase MDM2 by preventing their degradation in order to control protein network [45]. USP7 could also deubiquitinate and stabilize EZH2 in prostate cancer cells [46]. As far as we know, little is known about the function of USP7 in HNSC tumorigenesis and elucidated new substrates of USP7 is of great interest. Our present study identified that methylation G3BP2 by PRMT5 increased its binding with and deubiquitination by USP7. Moreover, USP7 was highly expressed in HNSC, and was recruited to the ACLY gene promoter, in which USP7 stabilizes G3BP2 and epigenetically enhancing lipogenic genes expression and lipogenesis. In contrast, other deubiquitinases such as USP10, which is reported to mediate G3BP2/G3BP1 deubiquitination has no correlation with G3BP2R468me2 under our system [7]. Recent studies reported that USP7 acts as a co-activator in tumor initiation by stabilizing Axin and hnRNPA1 [47, 48]. Consistent with these results, our study indicates that abolish USP7 expression retards tumor growth of HNSC cells. Our present study identified that G3BP2 methylation by PRMT5 promotes G3BP2 binding with and then deubiquitination by USP7, which suggests a link between substrates methylation and deubiquitination. Despite considerable analysis arranged in this study on the USP7 and G3BP2 regulatory circuit, it is not yet fully clear how USP7-G3BP2 promoted lipid metabolic reprogram cells to help the cancer state.
Our results extend the knowledge regarding of G3BP2 methylation and stabilization for tumor progression. High-level PRMT5 and USP7 are responsible for the accumulation of, and enhancing G3BP2 deubiquitination, which leads to lipogenesis and aggressive malignancy of HNSC. A PRMT5-USP7-G3BP2 regulatory complex is a potential therapeutic strategy for HNSC and other lipid metabolic diseases as well.
## Supplementary information
Supplementary tables Supplementary figure legends Supplementary Figure S1 Supplementary Figure S2 Supplementary Figure S3 Supplementary Figure S4 Supplementary Figure S5 Supplementary Figure S6 checklist Original Data File The online version contains supplementary material available at 10.1038/s41419-023-05706-2.
## References
1. Bian X, Liu R, Meng Y, Xing D, Xu D, Lu Z. **Lipid metabolism and cancer**. *J Exp Med* (2021) **218** e20201606. DOI: 10.1084/jem.20201606
2. Abramson HN. **The lipogenesis pathway as a cancer target**. *J Med Chem* (2011) **54** 5615-38. DOI: 10.1021/jm2005805
3. Kedersha N, Panas MD, Achorn CA, Lyons S, Tisdale S, Hickman T. **G3BP-Caprin1-USP10 complexes mediate stress granule condensation and associate with 40S subunits**. *J Cell Biol* (2016) **212** 845-60. DOI: 10.1083/jcb.201508028
4. Hong HQ, Lu J, Fang XL, Zhang YH, Cai Y, Yuan J. **G3BP2 is involved in isoproterenol-induced cardiac hypertrophy through activating the NF-kappaB signaling pathway**. *Acta Pharm Sin* (2018) **39** 184-94. DOI: 10.1038/aps.2017.58
5. Zhao B, Li H, Liu J, Han P, Zhang C, Bai H. **MicroRNA-23b targets Ras GTPase-activating protein SH3 domain-binding protein 2 to alleviate fibrosis and albuminuria in diabetic nephropathy**. *J Am Soc Nephrol* (2016) **27** 2597-608. DOI: 10.1681/ASN.2015030300
6. Takayama KI, Suzuki T, Tanaka T, Fujimura T, Takahashi S, Urano T. **TRIM25 enhances cell growth and cell survival by modulating p53 signals via interaction with G3BP2 in prostate cancer**. *Oncogene* (2018) **37** 2165-80. DOI: 10.1038/s41388-017-0095-x
7. Takayama KI, Suzuki T, Fujimura T, Takahashi S, Inoue S. **Association of USP10 with G3BP2 inhibits p53 signaling and contributes to poor outcome in prostate cancer**. *Mol Cancer Res* (2018) **16** 846-56. DOI: 10.1158/1541-7786.MCR-17-0471
8. Ashikari D, Takayama K, Tanaka T, Suzuki Y, Obinata D, Fujimura T. **Androgen induces G3BP2 and SUMO-mediated p53 nuclear export in prostate cancer**. *Oncogene* (2017) **36** 6272-81. DOI: 10.1038/onc.2017.225
9. Li H, Lin PH, Gupta P, Li X, Zhao SL, Zhou X. **MG53 suppresses tumor progression and stress granule formation by modulating G3BP2 activity in non-small cell lung cancer**. *Mol Cancer* (2021) **20** 118. DOI: 10.1186/s12943-021-01418-3
10. Rengasamy M, Zhang F, Vashisht A, Song WM, Aguilo F, Sun Y. **The PRMT5/WDR77 complex regulates alternative splicing through ZNF326 in breast cancer**. *Nucleic Acids Res* (2017) **45** 11106-20. DOI: 10.1093/nar/gkx727
11. Deng X, Shao G, Zhang HT, Li C, Zhang D, Cheng L. **Protein arginine methyltransferase 5 functions as an epigenetic activator of the androgen receptor to promote prostate cancer cell growth**. *Oncogene* (2017) **36** 1223-31. DOI: 10.1038/onc.2016.287
12. Liu M, Yao B, Gui T, Guo C, Wu X, Li J. **PRMT5-dependent transcriptional repression of c-Myc target genes promotes gastric cancer progression**. *Theranostics* (2020) **10** 4437-52. DOI: 10.7150/thno.42047
13. Powers MA, Fay MM, Factor RE, Welm AL, Ullman KS. **Protein arginine methyltransferase 5 accelerates tumor growth by arginine methylation of the tumor suppressor programmed cell death 4**. *Cancer Res* (2011) **71** 5579-87. DOI: 10.1158/0008-5472.CAN-11-0458
14. Gamper AM, Qiao X, Kim J, Zhang L, DeSimone MC, Rathmell WK. **Regulation of KLF4 turnover reveals an unexpected tissue-specific role of pVHL in tumorigenesis**. *Mol Cell* (2012) **45** 233-43. DOI: 10.1016/j.molcel.2011.11.031
15. Hu D, Gur M, Zhou Z, Gamper A, Hung MC, Fujita N. **Interplay between arginine methylation and ubiquitylation regulates KLF4-mediated genome stability and carcinogenesis**. *Nat Commun* (2015) **6** 8419. DOI: 10.1038/ncomms9419
16. Liu L, Zhao X, Zhao L, Li J, Yang H, Zhu Z. **Arginine methylation of SREBP1a via PRMT5 promotes de novo lipogenesis and tumor growth**. *Cancer Res* (2016) **76** 1260-72. DOI: 10.1158/0008-5472.CAN-15-1766
17. Wang N, Yan H, Wu D, Zhao Z, Chen X, Long Q. **PRMT5/Wnt4 axis promotes lymph-node metastasis and proliferation of laryngeal carcinoma**. *Cell Death Dis* (2020) **11** 864. DOI: 10.1038/s41419-020-03064-x
18. Feng J, Li L, Zhang N, Liu J, Zhang L, Gao H. **Androgen and AR contribute to breast cancer development and metastasis: an insight of mechanisms**. *Oncogene* (2017) **36** 2775-90. DOI: 10.1038/onc.2016.432
19. Dacwag CS, Bedford MT, Sif S, Imbalzano AN. **Distinct protein arginine methyltransferases promote ATP-dependent chromatin remodeling function at different stages of skeletal muscle differentiation**. *Mol Cell Biol* (2009) **29** 1909-21. DOI: 10.1128/MCB.00742-08
20. Smith E, Zhou W, Shindiapina P, Sif S, Li C, Baiocchi RA. **Recent advances in targeting protein arginine methyltransferase enzymes in cancer therapy**. *Expert Opin Ther Targets* (2018) **22** 527-45. DOI: 10.1080/14728222.2018.1474203
21. Huang YT, Cheng AC, Tang HC, Huang GC, Cai L, Lin TH. **USP7 facilitates SMAD3 autoregulation to repress cancer progression in p53-deficient lung cancer**. *Cell Death Dis* (2021) **12** 880. DOI: 10.1038/s41419-021-04176-8
22. Song N, Cao C, Tian S, Long M, Liu L. **USP7 Deubiquitinates and Stabilizes SIRT1**. *Anat Rec (Hoboken)* (2020) **303** 1337-45. DOI: 10.1002/ar.24252
23. Liu L, Yan H, Ruan M, Yang H, Wang L, Lei B. **An AKT/PRMT5/SREBP1 axis in lung adenocarcinoma regulates de novo lipogenesis and tumor growth**. *Cancer Sci* (2021) **112** 3083-98. DOI: 10.1111/cas.14988
24. Faubert B, Solmonson A, DeBerardinis RJ. **Metabolic reprogramming and cancer progression**. *Science* (2020) **368** eaaw5473. DOI: 10.1126/science.aaw5473
25. Furuta E, Okuda H, Kobayashi A, Watabe K. **Metabolic genes in cancer: their roles in tumor progression and clinical implications**. *Biochim Biophys Acta* (2010) **1805** 141-52. PMID: 20122995
26. Prigent M, Barlat I, Langen H, Dargemont C. **IkappaBalpha and IkappaBalpha /NF-kappa B complexes are retained in the cytoplasm through interaction with a novel partner, RasGAP SH3-binding protein 2**. *J Biol Chem* (2000) **275** 36441-9. DOI: 10.1074/jbc.M004751200
27. Ye J, Blelloch R. **Regulation of pluripotency by RNA binding proteins**. *Cell Stem Cell* (2014) **15** 271-80. DOI: 10.1016/j.stem.2014.08.010
28. Chen Q, Hu G. **Post-transcriptional regulation of the pluripotent state**. *Curr Opin Genet Dev* (2017) **46** 15-23. DOI: 10.1016/j.gde.2017.06.010
29. Zheng Y, Wu J, Deng R, Lin C, Huang Y, Yang X. **G3BP2 regulated by the lncRNA LINC01554 facilitates esophageal squamous cell carcinoma metastasis through stabilizing HDGF transcript**. *Oncogene* (2022) **41** 515-26. DOI: 10.1038/s41388-021-02073-0
30. Gupta N, Badeaux M, Liu Y, Naxerova K, Sgroi D, Munn LL. **Stress granule-associated protein G3BP2 regulates breast tumor initiation**. *Proc Natl Acad Sci USA* (2017) **114** 1033-8. DOI: 10.1073/pnas.1525387114
31. Stopa N, Krebs JE, Shechter D. **The PRMT5 arginine methyltransferase: many roles in development, cancer and beyond**. *Cell Mol Life Sci* (2015) **72** 2041-59. DOI: 10.1007/s00018-015-1847-9
32. Cho EC, Zheng S, Munro S, Liu G, Carr SM, Moehlenbrink J. **Arginine methylation controls growth regulation by E2F-1**. *EMBO J* (2012) **31** 1785-97. DOI: 10.1038/emboj.2012.17
33. Huang J, Vogel G, Yu Z, Almazan G, Richard S. **Type II arginine methyltransferase PRMT5 regulates gene expression of inhibitors of differentiation/DNA binding Id2 and Id4 during glial cell differentiation**. *J Biol Chem* (2011) **286** 44424-32. DOI: 10.1074/jbc.M111.277046
34. Hsu JM, Chen CT, Chou CK, Kuo HP, Li LY, Lin CY. **Crosstalk between Arg 1175 methylation and Tyr 1173 phosphorylation negatively modulates EGFR-mediated ERK activation**. *Nat Cell Biol* (2011) **13** 174-81. DOI: 10.1038/ncb2158
35. Jia Z, Yue F, Chen X, Narayanan N, Qiu J, Syed SA. **Protein arginine methyltransferase PRMT5 regulates fatty acid metabolism and lipid droplet biogenesis in white adipose tissues**. *Adv Sci (Weinh)* (2020) **7** 2002602. DOI: 10.1002/advs.202002602
36. Kanamaluru D, Xiao Z, Fang S, Choi SE, Kim DH, Veenstra TD. **Arginine methylation by PRMT5 at a naturally occurring mutation site is critical for liver metabolic regulation by small heterodimer partner**. *Mol Cell Biol* (2011) **31** 1540-50. DOI: 10.1128/MCB.01212-10
37. Durant ST, Cho EC, La Thangue NB. **p53 methylation-the Arg-ument is clear**. *Cell Cycle* (2009) **8** 801-2. DOI: 10.4161/cc.8.6.7850
38. Li Y, Chitnis N, Nakagawa H, Kita Y, Natsugoe S, Yang Y. **PRMT5 is required for lymphomagenesis triggered by multiple oncogenic drivers**. *Cancer Disco* (2015) **5** 288-303. DOI: 10.1158/2159-8290.CD-14-0625
39. Vinet M, Suresh S, Maire V, Monchecourt C, Nemati F, Lesage L. **Protein arginine methyltransferase 5: A novel therapeutic target for triple-negative breast cancers**. *Cancer Med* (2019) **8** 2414-28. DOI: 10.1002/cam4.2114
40. Liu H, Bai Y, Zhang X, Gao T, Liu Y, Li E. **SARS-CoV-2 N protein antagonizes stress granule assembly and IFN production by interacting with G3BPs to facilitate viral replication**. *J Virol* (2022) **96** e0041222. DOI: 10.1128/jvi.00412-22
41. Zhang H, Zhang S, He H, Zhao W, Chen J, Shao RG. **GAP161 targets and downregulates G3BP to suppress cell growth and potentiate cisplaitin-mediated cytotoxicity to colon carcinoma HCT116 cells**. *Cancer Sci* (2012) **103** 1848-56. DOI: 10.1111/j.1349-7006.2012.02361.x
42. Dai X, Lu L, Deng S, Meng J, Wan C, Huang J. **USP7 targeting modulates anti-tumor immune response by reprogramming Tumor-associated Macrophages in Lung Cancer**. *Theranostics* (2020) **10** 9332-47. DOI: 10.7150/thno.47137
43. Zhang T, Periz G, Lu YN, Wang J. **USP7 regulates ALS-associated proteotoxicity and quality control through the NEDD4L-SMAD pathway**. *Proc Natl Acad Sci USA* (2020) **117** 28114-25. DOI: 10.1073/pnas.2014349117
44. Qi SM, Cheng G, Cheng XD, Xu Z, Xu B, Zhang WD. **Targeting USP7-mediated deubiquitination of MDM2/MDMX-p53 pathway for cancer therapy: Are we there yet?**. *Front Cell Dev Biol* (2020) **8** 233. DOI: 10.3389/fcell.2020.00233
45. Tavana O, Gu W. **Modulation of the p53/MDM2 interplay by HAUSP inhibitors**. *J Mol Cell Biol* (2017) **9** 45-52. DOI: 10.1093/jmcb/mjw049
46. Lee JE, Park CM, Kim JH. **USP7 deubiquitinates and stabilizes EZH2 in prostate cancer cells**. *Genet Mol Biol* (2020) **43** e20190338. DOI: 10.1590/1678-4685-gmb-2019-0338
47. Ji L, Lu B, Zamponi R, Charlat O, Aversa R, Yang Z. **USP7 inhibits Wnt/beta-catenin signaling through promoting stabilization of Axin**. *Nat Commun* (2019) **10** 4184. DOI: 10.1038/s41467-019-12143-3
48. Zhang H, Deng T, Liu R, Ning T, Yang H, Liu D. **CAF secreted miR-522 suppresses ferroptosis and promotes acquired chemo-resistance in gastric cancer**. *Mol Cancer* (2020) **19** 43. DOI: 10.1186/s12943-020-01168-8
|
---
title: Optical coherence tomography in healthy human subjects in the setting of prolonged
dark adaptation
authors:
- Erin H. Su
- Niranjana Kesavamoorthy
- Hossein Ameri
journal: Scientific Reports
year: 2023
pmcid: PMC9988879
doi: 10.1038/s41598-023-30747-0
license: CC BY 4.0
---
# Optical coherence tomography in healthy human subjects in the setting of prolonged dark adaptation
## Abstract
Human studies have established that short periods of dark adaptation can induce outer retinal thinning and various band intensity changes that can be detected with Optical Coherence Tomography (OCT). Similar findings were observed in mice, including a positive correlation between the degree of outer retinal changes and dark adaptation duration. We decided to assess potential retinal structural changes following prolonged dark adaptation in humans. 40 healthy subjects without any ocular diseases participated in this study. For each subject, one eye was covered for dark adaptation for four hours, and the other eye was left uncovered as a control. Before and after the dark adaptation period, both eyes were assessed with OCT. Using the Heidelberg Spectralis system, basic statistical functions, and qualitative and quantitative analysis, we were able to compare retinal layer thicknesses and band intensities between covered (dark adapted) versus uncovered (control) eyes. Prolonged dark adaptation did not induce any significant thickness, volume, or intensity changes in the outer retina or in the inner or overall retina. These observations thus alter our current understanding of the mechanisms underlying dark adaptation’s neuroprotective effects in preventing blindness and require further study.
## Introduction
Dark adaptation describes the recovery of visual sensitivity in the dark after light exposure. Multiple processes occur simultaneously to allow for retinal homeostasis, including modulation of blood flow1, interphotoreceptor matrix protein and photoreceptor signaling protein redistribution2,3, and metabolic energy flow reversal4. Because visual perception in the dark is rod-mediated and affected by rod photoreceptor health and homeostasis, delayed dark adaptation is often an early symptom of various retinal diseases, including retinitis pigmentosa5, diabetic retinopathy6, and age-related macular degeneration7. Deeper understanding of dark adaptation, as well as early clinical detection, are therefore essential to prevent further irreversible vision loss in these subjects8.
Previous animal studies have established that dark adaptation induces outer retinal structural changes related to maintaining retinal homeostasis8–13. Concordant studies that have focused on human subjects have shown similar results, mostly in diseased retinas (i.e. diabetic retinopathy, macular diseases, etc.); considerably fewer studies, however, have been performed on healthy human or mouse retinas, and those that have used bleaching or flash light stimuli9,14–20.
An ideal imaging modality for studying such structural changes is Optical Coherence Tomography (OCT), a noninvasive, high-resolution, in vivo imaging modality used for monitoring and diagnosing retinal diseases. OCT takes images of optical tissue sections using an infrared light source and has been used for time-lapse retinal imaging to study light-induced morphologic changes.
Many studies have shown the utility of OCT for studying retinal structural changes, especially in the context of dark adaptation, though the dark adaptation periods in these and other clinical studies were done over short periods of time only21. To the best of our knowledge, the longest studied period of dark adaptation in human subjects has been 30 min22. Given previous mouse study findings of more marked structural changes with increased periods of dark adaptation9, our study sought to assess retinal structural changes following prolonged dark adaptation in humans and explore its potential clinical significance.
## Results
The participants’ mean age was 27.65 years (Table 1; 21 subjects male, 19 subjects female). Using high-resolution OCT imaging, we examined healthy human retinal structure under dark adaptation and control conditions. Table 1Subject demographics. Includes age, sex, best corrected visual acuity (BCVA), refraction (sphere) for each eye, and which eye was dark adapted/patched (aexcluded from analysis).OCT participant #AgeSexBCVARefraction (SPH)DA/Patched eye125F$\frac{20}{20}$ OU− 2.75 OD, − 3.5 OSOS225F$\frac{20}{20}$ OU0 OUOD325F$\frac{20}{20}$ OU− 5 OUOD425M$\frac{20}{20}$ OU0 OUOS526M$\frac{20}{20}$ OU− 2.5 OUOD629M$\frac{20}{20}$ OU− 6.5 OUOS726F$\frac{20}{20}$ OU− 5.5 OD, − 4.5 OSOS830M$\frac{20}{20}$ OU− 2 OD, 0 OSOD933M$\frac{20}{20}$ OU− 4.5 OUOD1025M$\frac{20}{20}$ OU− 2.5 OUOS1127M$\frac{20}{20}$ OU0 OUOS1226M$\frac{20}{20}$ OU− 4.25 OD, − 4.5 OSOS1325F$\frac{20}{20}$ OU− 2.5 OUOD1429M$\frac{20}{20}$ OU0 OUOD1525F$\frac{20}{20}$ OU− 8.25 OUOS1619F$\frac{20}{20}$ OU− 3.25 OD, − 2.75 OSOD1723F$\frac{20}{20}$ OU− 1.25 OD, − 2.5 OSOS1826M$\frac{20}{20}$ OU− 6 OD, − 5.5 OSOD1924M$\frac{20}{20}$ OU0 OUOS2031M$\frac{20}{20}$ OU− 5 OD, − 5.5 OSOD2119M$\frac{20}{20}$ OU− 1.5 OD, − 1.75 OSOD2228F$\frac{20}{20}$ OU− 3.75 OD, − 3.25 OSOS2326M$\frac{20}{20}$ OU0 OUOD2422F$\frac{20}{20}$ OU0 OUOS2525F$\frac{20}{20}$ OU− 2 OD, − 3 OSOD2631M$\frac{20}{20}$ OU− 4.5 OUOS2724F$\frac{20}{20}$ OU− 2.25 OUOD2828M$\frac{20}{20}$ OU0 OUOS2925F$\frac{20}{20}$ OU0 OUOD3025F$\frac{20}{20}$ OU0 OUOS3122F$\frac{20}{20}$ OU0 OUOD3225M$\frac{20}{20}$ OU0 OUOS3328F$\frac{20}{20}$ OU− 2 OD, − 3 OSOD3430M$\frac{20}{20}$ OU− 1 OD, − 2 OSOS3531M$\frac{20}{20}$ OU0 OUOS3629F$\frac{20}{20}$ OU− 5.75 OUOD3756M$\frac{20}{20}$ OU− 7 OUOD3860F$\frac{20}{20}$ OU− 2.75 OD, 0 OSOS39a25F$\frac{20}{20}$ OU− 2.25 OD, − 2.5 OSOD40a23F$\frac{20}{20}$ OU− 4 OD, − 2.25 OSOS Average thicknesses and volumes of the three retinal layers were calculated across the foveal, and four inner and four outer perifoveal regions (Table 2, Supplemental Table 2). The average thickness values for the outer, inner, and overall retina for dark adaptation and controlled conditions are represented in Fig. 1. Neither thickness nor volume parameters showed any significant changes between any of the three analyzed layers when comparing control versus dark adapted eyes (Figs. 1, 2, Tables 2, 3, Supplemental Figs. 1–4, Supplemental Tables 1, 2); the control and dark adaptation conditions yielded similar results across the overall, inner, and outer retinal layers. Thickness and volume differences across all nine regions (foveal, inner and outer perifoveal) mostly converged upon zero, with both layer thickening and thinning occurring (Figs. 1B, 2B, Supplemental Figs. 1–4B). Paired t-testing using the absolute values of the thickness and volume differences also yielded results that were not statistically significantly different; calculations resulted in p-values of 0.89, 0.35, and 0.20 for overall, inner, and outer retinal layer thickness changes, and p-values of 0.39, 0.14, and 0.33 for overall, inner, and outer retinal layer volume changes respectively (Table 3, Supplemental Table 2).Table 2Average thickness measurements and p-values for all 9 regions of interest (foveal and inner and outer perifoveal regions) between control and dark adaptation (DA) conditions. Region of interestAvg thickness (µm)Overall retinaInner retinaOuter retinaFoveaDA272.71 ± 23.46180.74 ± 24.1492.13 ± 3.65Control274.64 ± 19.04182.02 ± 19.6592.53 ± 4.03p-valuep = 0.41p = 0. 57p = 0.90Temporal inner maculaDA336.63 ± 14.51253.14 ± 12.8283.45 ± 3.33Control337.56 ± 16.97254.09 ± 15.4483.54 ± 3.30p-valuep = 0. 70p = 0.60p = 0.87Superior inner maculaDA344.82 ± 15.64262.01 ± 14.1382.97 ± 3.09Control345.60 ± 14.27262.59 ± 12.9683.04 ± 3.06p-valuep = 0.37p = 0.37p = 0. 99Nasal inner maculaDA338.42 ± 18.06255.00 ± 16.2283.53 ± 3.43Control335.90 ± 31.88254.68 ± 13.6183.87 ± 3.29p-valuep = 0.93p = 0.91p = 0.63Inferior inner maculaDA339.03 ± 14.69257.18 ± 13.6282.01 ± 3.65Control339.36 ± 14.89258.06 ± 13.3181.85 ± 3.32p-valuep = 0.99p = 0.160.44Temporal outer maculaDA302.86 ± 18.68222.05 ± 18.5681.57 ± 6.01Control301.37 ± 22.67221.07 ± 22.0080.27 ± 3.55p-valuep = 0.97p = 0.93p = 0.11Superior outer maculaDA303.95 ± 12.16222.87 ± 10.6481.21 ± 3.53Control302.79 ± 12.19222.32 ± 11.1680.50 ± 2.85p-valuep = 0.26p = 0.95p = 0.14Nasal outer maculaDA302.47 ± 23.93222.53 ± 23.1980.21 ± 3.48Control302.94 ± 19.86222.97 ± 19.0680.13 ± 3.09p-valuep = 0.97p = 0.98p = 0.43Inferior outer maculaDA291.26 ± 15.07212.63 ± 13.4678.84 ± 3.34Control290.65 ± 13.28211.89 ± 12.4778.72 ± 3.40p-valuep = 0.97p = 0.93p = 0.11Table 3Absolute values of average change in thickness, standard deviations, and p-values for overall, inner, and outer retinal layers between control and dark adapted (DA) eyes respectively. Avg Δ (µm), abs valueOverall retina thickness, controlOverall retina thickness, DAInner retina thickness, controlInner retina thickness, DAOuter retina, thickness controlOuter retina thickness, DAAverage2.572.612.402.161.561.79St dev1.581.331.351.0670.940.75P-value0.890.350.20Figure 1Outer retinal thickness measurements before and after dark adaptation (DA) at baseline and four hours later in the dark adapted and control eyes. ( A) Bar graph comparing dark adapted (orange) and control (blue) average thickness measurements for the 9 regions of interest (foveal, and inner and outer perifoveal regions) with a labeled thickness map example template included (superior-sup., nasal-n., temporal-t., inferior-inf.). ( B) Average differences for the outer retinal layer for control and dark adaptation conditions respectively. Data points were taken from the Heidelberg-generated thickness maps from all 9 regions (foveal and inner and outer perifoveal regions) for each subject, and the values before the dark adaptation period were subtracted from those after the dark adaptation period to calculate the difference in values. Negative values show layer thinning and positive values show layer thickening in the setting of dark adaptation. There was no statistically significant difference within the respective nine regions. Figure 2Overall retinal thickness measurements before and after dark adaptation (DA) at baseline and four hours later in the dark adapted and control eyes. ( A) Bar graph comparing dark adapted (orange) and control (blue) average thickness measurements for the 9 regions of interest (foveal, and inner and outer perifoveal regions), with a labelled thickness map example template included. ( B) Average differences for the overall retinal layer for control and dark adaptation conditions respectively. Data points were taken from the Heidelberg-generated thickness maps from all 9 regions (foveal and inner and outer perifoveal regions) for each subject, and the values before the dark adaptation period were subtracted from those after the dark adaptation period to calculate the difference in values. Negative values show layer thinning and positive values show layer thickening in the setting of dark adaptation. There was no statistically significant difference within the respective nine regions.
Band intensities for the ellipsoid zone (EZ) band were also assessed and no significant changes were noticed ($$p \leq 0.46$$). The interdigitation zone (IZ) band length was measured as well and served as a proxy for determining the strength of the hyporeflective band between the inner segment (IS) and retinal pigment epithelium (RPE) bands. However, no IZ band changes were noted ($$p \leq 0.32$$).
## Discussion
Prolonged dark adaptation over a period of hours did not induce any significant thickness, volume, or intensity changes in the overall, inner, or outer retina in our healthy subjects; however, this was not the case in previous studies. Previous human OCT studies have shown that short periods of dark adaptation (the longest being 30 min) can lead to outer retinal thinning, thinning or absence of the hyporeflective band seen in light between the outer segment (OS) and RPE bands (which can also be interpreted as IZ band blurring), and inner segment ellipsoid (ISe) and EZ band intensity dimming in the dark; the same changes have been shown in mice for short dark adaptation periods (the longest being 2 h) and overnight dark adaptation8,9,11,12,22.
The lack of these changes could be due to several factors including homeostatic processes. Given the nature of homeostasis, it is possible that the photoreceptor dark current that causes the increased adenosine triphosphate (ATP) consumption in dark adaptation is not maintained over prolonged periods of time, as the eye has already acclimated to the dark. If the eye has already fully adjusted to its new light settings, there is no continued need to adjust, and retinal ATP consumption can return to its baseline levels. Furthermore, dark adaptation is known to slow with age23. Given our study participants were almost exclusively young, healthy adults, dark adaptation is expected to be faster. It is therefore possible that the previously reported dark adaptation structural changes are transient and occurred long before our four-hour dark adaptation period was completed in our young and healthy subjects, and then returned to baseline. Kim et al. performed dark adaptation over a five-minute period, imaging mice throughout the entirety of the dark adaptation period. They noted that band intensity changes generally came in phases, depending on the duration of dark adaptation; ISe intensity significantly reduced after only 5 min of dark adaptation, but inner plexiform layer (IPL), outer plexiform layer (OPL) and RPE band dimming only occurred at a later phase of dark adaptation8. Collectively, their data suggested that retinal changes occur linearly, and that the retina can respond incredibly quickly during dark adaptation. Dark adaptation recovery time in humans, however, has not yet been established in OCT studies. The lack of structural changes in our results necessitates further study to establish the longest dark adaptation period with structural changes.
However, mouse studies like Li et al. 2016 and Li et al., 2018 threaten to refute the possibility of quick dark adaptation recovery time in the setting of homeostatic processes9,10; overnight dark adaptation in these studies yielded structural changes as well, with a longer dark adaptation period that was also on the scale of several hours. While prolonged dark adaptation may well cause such structural changes, these changes may be due to other confounding variables, namely, diurnal rhythm. During disc shedding, photoreceptors are renewed through transient loss (then subsequent replacement) of the outer segment, thus leading to times of temporary outer segment shortening; as disc shedding is linked to the circadian cycle, the most active period of shedding is the early morning, and the least is in the late evening24–30. Therefore, any associated retinal thinning or shortening seen after an overnight period may be attributable to other confounding variables rather than dark adaptation.
Another potential variable that may account for the discrepancy between our data and those of previous studies is the imaging software used during OCT collection. Since the Spectralis system uses the fundal image as a reference during eye tracking, our study’s OCT image follow-ups were therefore performed at the same exact retinal location to allow for exact qualitative and quantitative comparison10. The Spectralis system uses the fundus image as a reference during eye tracking, so "follow-up mode" is a Spectralis feature that allows one to image the same exact retinal location upon later imaging sessions to allow for exact qualitative and quantitative comparison. Given that layer thickness varies across different retinal cross sections, eye tracking is a necessary tool for precise comparison between conditions. Since this feature is not widely available across different imaging systems, lack of eye tracking may account for some of these differences in results. As discussed in Krebs et al., variations in retinal thickness measurements can also be attributed to the different imaging systems’ methodologies, and therefore image quality, layer detection, and segmentation31. Other considerable differences are caused by localization control, scan line density, algorithms and segmentation line positioning.
In summary, we were unable to detect any significant retinal structural changes on OCT following four hours of dark adaptation in healthy humans. Further exploration of dark adaptation and retinal homeostasis, including analysis of choroidal thickness measurements and modulation of blood flow, may lend further insight into the mechanisms behind dark adaptation and prove clinically useful. Further studies across different age groups and with various intervals of dark adaptation, ranging on a scale of minutes to a few hours, and different times throughout the day and night, also should be performed to detect potential structural photoresponses, and if these photoresponses can indeed serve as future clinical tools in diagnosing and treating retinal pathologies.
## Dark adaptation and OCT imaging protocol
OCT images were taken from both eyes at baseline and four hours later. Between the two OCT imaging, one eye was dark adapted and the other eye was kept as control (Fig. 3A,B). The baseline OCT image of both eyes were taken in the dark between 12 p.m. and 1 p.m. to avoid peak disc shedding time as a possible confounder, similar to Lu et al.22. Subjects who wore contact lenses were instructed to remove their contact lenses for the duration of the experiment. Subjects who wore glasses were instructed to remove their glasses during OCT imaging and could elect to wear their glasses during the dark adaptation period over the eye patch if they so chose. Figure 3Example of Spectral Domain Optical Coherence Tomography (SD OCT) retinal images obtained (A) B-scan of one subject’s retina through the fovea in the dark adapted/patched eye (left) and unpatched control eye (right), before and after the dark adaptation (DA) period. ( B) Side-by-side comparison of the OCT images for the same local retinal region (outlined by the blue rectangles in [A]) for patched and unpatched eyes, before and after the DA period. ILM internal limiting membrane, ELM external limiting membrane, EZ ellipsoid zone, IZ interdigitation zone, RPE retinal pigment epithelium (C) Example of the segmentation performed by the Heidelberg Spectralis software that was used to generate thickness maps for the outer retinal layer. BM basement membrane.
After this initial baseline OCT image was taken, we randomly selected one of each subject’s two eyes and covered that eye with an eye patch that was designed specifically for this study and blocked all potential light from getting through. The eye patch involved taping three layers to the area around the subject’s selected eye: first, a disposable cotton eye pad; second, a flexible sheet of black plastic impenetrable to light; third, a final layer of gauze. After the addition of each eyepatch layer, the subject was asked to shut the other unpatched eye and confirm that no light was visible through or around the eye covering.
The subjects were instructed to keep the eye patch on for four hours in between the two imaging sessions. During this time, the subjects were instructed to stay in a relatively well-lit setting and were instructed to not sleep or spend the patching period in a dark place. These instructions were given to ensure the unpatched eye would be exposed to light for the full duration of the experiment as opposed to the patched eye, which would remain in the dark throughout the experiment. Of the 40 participants, 20 were observed throughout the entirety of the four-hour dark adaptation period and were confirmed to have followed instructions. The remaining 20 gave verbal affirmation they followed instructions. After the dark adaptation period was over, the subjects returned for a second OCT imaging session where both eyes imaged again. The room was kept dark during both imaging sessions to ensure higher quality OCT images. The room was also kept dark during imaging to ensure the patched eye remained dark adapted during post-patching imaging and was not exposed to light until after the patch was removed. The dark adapted eye was imaged immediately after removal of the patch. No follow-up was needed with the subjects after their OCT images were collected.
## OCT settings
The Spectralis SD OCT instrument (Heidelberg Engineering, Heidelberg, Germany) was used to obtain 20° × 20° OCT cube scans (49 sections, 1024 A-scans in each B-scan, Automated Real Time 20 frames) centered on the fovea of each eye. Images of each eye, both before and after patching, were taken. During the study period, all scans were performed using the follow-up mode after setting the scans before the dark adaptation period as reference. Images of two subjects were eliminated from the study due to lack of using follow-up mode (Table 1).
## Analysis
Segmentation of retinal layer boundaries on each OCT retinal B-scan and thickness maps of the desired retinal cell layers were performed and generated by Heidelberg software (Fig. 3C). Internal limiting membrane (ILM), external limiting membrane (ELM) and basement membrane (BM refers to the basement membrane of the choriocapillaries/Bruch’s membrane) were used for segmentation. The Heidelberg Spectralis segmentation software reliably detects these structures and provides accurate segmentation. Given that previous studies have shown significant thickness changes in the outer retinal layer, for the thickness maps, the retinal cell layers chosen for analysis were the outer retinal layer (between BM and ELM), inner retinal layer (between ELM and ILM), and overall retina (between BM and ILM).
For each analyzed retinal layer, the thicknesses across the 49 scans were averaged across nine regions of interest, including the foveal and the inner and outer perifoveal (superior, nasal, inferior, and temporal) regions (Fig. 1A, Supplementary Figs. 1–4A).
Averaged thicknesses across the foveal and inner and outer perifoveal regions were compared before and after dark adaptation for each subject. The absolute values of the before-and-after differences were compared between dark adapted/patched versus unpatched control eyes by paired t testing, two-tailed. Statistical significance was defined as $p \leq 0.05.$
For band intensity evaluation, all B-scans for the dark adapted/patched and unpatched control eyes, before and after the dark adaptation period, were masked and quantitatively and assessed for their EZ and IZ band intensities.
For assessing EZ band intensity changes, we used EZ:ELM ratio to eliminate potential intensity changes introduced to each scan because of factors such as media clarity, head tilt etc. Invitrogen GelQuant Express software was used to evaluate EZ band intensity changes over the central 2600 µm region on the central line scan. This region was split into 11 boxes, and due to irregular contour of bands in the fovea, the central-most box was excluded from analysis. The EZ and ELM band intensities were measured and the EZ:ELM ratios were calculated and averaged across the remaining 10 boxes. The EZ:ELM ratios, after vs before the dark adaptation period, were compared between light and dark adaptation conditions using paired t-testing, two-tailed.
In addition to EZ band intensity, we evaluated OCT images for any clinically significant changes. In a pilot observation we noticed some changes in the length of visibility of IZ band before and after dark adaptation. Thereafter, the IZ band lengths, before and after the dark adaptation period, were compared using paired t-testing, two-tailed.
## Ethics declaration
This methodology has been approved by the USC Biomedical Institutional Review Board ethics committee (HS-21-00842) in accordance with current guidelines and regulations. Informed consent was obtained from all participants.
## Supplementary Information
Supplementary Figure 1.Supplementary Figure 2.Supplementary Figure 3.Supplementary Figure 4.Supplementary Table 5.Supplementary Table 6. The online version contains supplementary material available at 10.1038/s41598-023-30747-0.
## References
1. Kwan CC, Lee HE, Schwartz G, Fawzi AA. **Acute hyperglycemia reverses neurovascular coupling during dark to light adaptation in healthy subjects on optical coherence tomography angiography**. *Investig. Opthalmol. Vis. Sci.* (2020) **61** 38. DOI: 10.1167/iovs.61.4.38
2. Calvert PD, Strissel KJ, Schiesser WE, Pugh EN, Arshavsky VY. **Light-driven translocation of signaling proteins in vertebrate photoreceptors**. *Trends Cell Biol.* (2006) **16** 560-568. DOI: 10.1016/j.tcb.2006.09.001
3. Ishikawa M, Sawada Y, Yoshitomi T. **Structure and function of the interphotoreceptor matrix surrounding retinal photoreceptor cells**. *Exp. Eye Res.* (2015) **133** 3-18. DOI: 10.1016/j.exer.2015.02.017
4. Linton JD. **Flow of energy in the outer retina in darkness and in light**. *Proc. Natl. Acad. Sci.* (2010) **107** 8599-8604. DOI: 10.1073/pnas.1002471107
5. Herse P. **Retinitis pigmentosa: Visual function and multidisciplinary management**. *Clin. Exp. Optom.* (2005) **88** 335-350. DOI: 10.1111/j.1444-0938.2005.tb06717.x
6. Hsiao C-C, Hsu H-M, Yang C-M, Yang C-H. **Correlation of retinal vascular perfusion density with dark adaptation in diabetic retinopathy**. *Graefes Arch. Clin. Exp. Ophthalmol.* (2019) **257** 1401-1410. DOI: 10.1007/s00417-019-04321-2
7. Jackson GR. **Diagnostic sensitivity and specificity of dark adaptometry for detection of age-related macular degeneration**. *Investig. Opthalmol. Vis. Sci.* (2014) **55** 1427. DOI: 10.1167/iovs.13-13745
8. Kim T-H, Ding J, Yao X. **Intrinsic signal optoretinography of dark adaptation kinetics**. *Sci. Rep.* (2022) **12** 2475. DOI: 10.1038/s41598-022-06562-4
9. Li Y, Fariss RN, Qian JW, Cohen ED, Qian H. **Light-induced thickening of photoreceptor outer segment layer detected by ultra-high resolution OCT imaging**. *Investig. Opthalmol. Vis. Sci.* (2016) **57** OCT105. DOI: 10.1167/iovs.15-18539
10. Li Y. **Light-dependent OCT structure changes in photoreceptor degenerative rd 10 mouse retina**. *Investig. Opthalmol. Vis. Sci.* (2018) **59** 1084. DOI: 10.1167/iovs.17-23011
11. Berkowitz BA. **Mitochondrial respiration in outer retina contributes to light-evoked increase in hydration in vivo**. *Investig. Opthalmol. Vis. Sci.* (2018) **59** 5957. DOI: 10.1167/iovs.18-25682
12. Gao S. **Functional regulation of an outer retina hyporeflective band on optical coherence tomography images**. *Sci. Rep.* (2021) **11** 10260. DOI: 10.1038/s41598-021-89599-1
13. Asteriti S, Gargini C, Cangiano L. **Mouse rods signal through gap junctions with cones**. *Elife* (2014) **3** e01386. DOI: 10.7554/eLife.01386
14. Mathis T. **Light-induced modifications of the outer retinal hyperreflective layers on spectral-domain optical coherence tomography in humans: An experimental study**. *Acta Ophthalmol. (Copenh.)* (2021) **99** 765-772. DOI: 10.1111/aos.14723
15. Bavinger JC. **The effects of diabetic retinopathy and pan-retinal photocoagulation on photoreceptor cell function as assessed by dark adaptometry**. *Investig. Opthalmol. Vis. Sci.* (2016) **57** 208. DOI: 10.1167/iovs.15-17281
16. Pandiyan VP. **The optoretinogram reveals the primary steps of phototransduction in the living human eye**. *Sci. Adv.* (2020) **6** eabc1124. DOI: 10.1126/sciadv.abc1124
17. Godara P. **Assessing retinal structure in complete congenital stationary night blindness and Oguchi disease**. *Am. J. Ophthalmol.* (2012) **154** 987-1001.e1. DOI: 10.1016/j.ajo.2012.06.003
18. Staurenghi G, Cereda M, Giani A, Luiselli C. **Autofluorescence and Oct characteristics in area of resolved long-lasting neurosensory retinal serous detachment**. *Invest. Ophthalmol. Vis. Sci.* (2008) **49** 910-910. PMID: 18326711
19. Borrelli E. **Impact of bleaching on photoreceptors in different intermediate AMD phenotypes**. *Transl. Vis. Sci. Technol.* (2019) **8** 5. DOI: 10.1167/tvst.8.6.5
20. Abràmoff MD. **Human photoreceptor outer segments shorten during light adaptation**. *Investig. Opthalmol. Vis. Sci.* (2013) **54** 3721. DOI: 10.1167/iovs.13-11812
21. Jonnal RS. **Toward a clinical optoretinogram: A review of noninvasive, optical tests of retinal neural function**. *Ann. Transl. Med.* (2021) **9** 1270. DOI: 10.21037/atm-20-6440
22. Lu CD. **Photoreceptor layer thickness changes during dark adaptation observed with ultrahigh-resolution optical coherence tomography**. *Investig. Opthalmol. Vis. Sci.* (2017) **58** 4632. DOI: 10.1167/iovs.17-22171
23. Jackson GR, Owsley C, McGwin G. **Aging and dark adaptation**. *Vis. Res.* (1999) **39** 3975-3982. DOI: 10.1016/S0042-6989(99)00092-9
24. Kocaoglu OP. **Photoreceptor disc shedding in the living human eye**. *Biomed. Opt. Express* (2016) **7** 4554. DOI: 10.1364/BOE.7.004554
25. LaVail MM. **Rod outer segment disk shedding in rat retina: Relationship to cyclic lighting**. *Science* (1976) **194** 1071-1074. DOI: 10.1126/science.982063
26. Goldman AI, Teirstein PS, O’Brien PJ. **The role of ambient lighting in circadian disc shedding in the rod outer segment of the rat retina**. *Investig. Ophthalmol. Vis. Sci.* (1980) **19** 1257-1267. PMID: 7429762
27. Goldman AI. **The sensitivity of rat rod outer segment disc shedding to light**. *Investig. Ophthalmol. Vis. Sci.* (1982) **22** 695-700. PMID: 7076414
28. LaVail MM. **Outer segment disc shedding and phagocytosis in the outer retina**. *Trans. Ophthalmol. Soc. U. K.* (1983) **103** 397-404. PMID: 6380008
29. Bobu C, Hicks D. **Regulation of retinal photoreceptor phagocytosis in a diurnal mammal by circadian clocks and ambient lighting**. *Investig. Opthalmol. Vis. Sci.* (2009) **50** 3495. DOI: 10.1167/iovs.08-3145
30. Baker BN, Moriya M, Williams TP. **Alteration of disk-shedding patterns by light-onset of higher than normal intensity**. *Exp. Eye Res.* (1986) **42** 535-546. DOI: 10.1016/0014-4835(86)90044-8
31. Krebs I. **Quality and reproducibility of retinal thickness measurements in two spectral-domain optical coherence tomography machines**. *Investig. Opthalmol. Vis. Sci.* (2011) **52** 6925. DOI: 10.1167/iovs.10-6612
|
---
title: Antigenotoxic properties of the halophyte Polygonum maritimum L. highlight
its potential to mitigate oxidative stress-related damage
authors:
- Daniela Oliveira
- Maria Inês Dias
- Lillian Barros
- Luísa Custódio
- Rui Oliveira
journal: Scientific Reports
year: 2023
pmcid: PMC9988880
doi: 10.1038/s41598-022-20402-5
license: CC BY 4.0
---
# Antigenotoxic properties of the halophyte Polygonum maritimum L. highlight its potential to mitigate oxidative stress-related damage
## Abstract
Long-term exposure to dietary xenobiotics can induce oxidative stress in the gastrointestinal tract, possibly causing DNA damage and contributing to the initiation of carcinogenesis. Halophytes are exposed to constant abiotic stresses, which are believed to promote the accumulation of antioxidant metabolites like polyphenols. The aim of this study was to evaluate the antioxidant and antigenotoxic properties of the ethanol extract of the aerial part of the halophyte *Polygonum maritimum* L. (PME), which can represent a dietary source of bioactive compounds with potential to attenuate oxidative stress-related damage. The PME exhibited a high antioxidant potential, revealed by the in vitro capacity to scavenge the free radical DPPH (IC50 = 2.29 ± 0.10 μg/mL) and the improved viability of the yeast *Saccharomyces cerevisiae* under oxidative stress ($p \leq 0.001$, 10 min). An antigenotoxic effect of PME against H2O2-induced oxidative stress was found in S. cerevisiae ($p \leq 0.05$) with the dominant deletion assay. In vitro colorimetric assays and LC-DAD-ESI/MSn analysis showed that PME is a polyphenol-rich extract composed of catechin, (epi)catechin dimer and trimers, quercetin and myricetin glycosides. Hence, P. maritimum is a source of antioxidant and antigenotoxic metabolites for application in industries that develop products to provide health benefits.
## Introduction
Industrial advances have led to an increasing consumption of foods contaminated with xenobiotics, such as those formed due to food processing and cooking at high temperatures or that arise from direct contact between foods and their packaging, which can perturb homeostasis and contribute to the development of human diseases1. Recurrent ingestion of xenobiotics may induce inflammation in the gastrointestinal tract, leading to the production of reactive oxygen species [ROS, e.g., hydrogen peroxide (H2O2) and hydroxyl radical (•OH)] and reactive nitrogen species (e.g., nitric oxide)2. In addition, inflammatory cytokines are produced, and these can activate enzymes capable of generating oxidants3. The redox imbalance caused by this adverse situation may lead to oxidative stress and consequent oxidative damage of cellular constituents, such as proteins, lipids and DNA, potentially promoting the development of oxidative stress-related illnesses. These include gastrointestinal diseases and cancers2, neurodegenerative ailments3, diabetes4 and others. Some xenobiotics can be carcinogenic due to their capability to induce DNA damage directly or after metabolic activation by generating carcinogenic metabolites able to interact with DNA and ROS that can cause oxidative damage5. If not repaired, DNA damage can induce mutations and genomic instability and may eventually lead to cancer risks.
Polyphenol rich-diets have been linked to health benefits that include lower mortality risk and preventive effects against diseases such as cancers of the gastrointestinal tract6. These benefits are often attributed to the antioxidant and anti-inflammatory properties of polyphenols7. Additionally, these phytochemicals have been associated with protection against DNA damage8. Halophytes are unique plant species able to live in saline environments due to the development of adaptive responses that allow them to counteract the extreme adversity inherent to their habitat, such as the accumulation of non-enzymatic antioxidants to neutralize ROS and prevent oxidative damage9. These antioxidants include vitamins, carotenoids and polyphenols, which are responsible for beneficial nutritional and medicinal properties of such plants. In fact, local communities in the coastal areas, such as those in the Mediterranean basin, have been using halophytes as food or in traditional medicine to relieve several human ailments, including cancer and diabetes10. Due to the harsh conditions that halophytes have to endure to survive, it is not surprising that these plants contain a higher level of phenolic compounds than glycophytes which correlates to a higher antioxidant activity that, in some cases, is even more potent than that of synthetic antioxidants11.
Polygonum maritimum L., or sea knotgrass, is a halophyte that can be found on the sandy coasts of America, Europe, South Africa and in the Mediterranean region12. Previous studies with extracts of P. maritimum from the Mediterranean region revealed a high content in phenolic compounds, particularly in flavonoids, which were suggested to contribute to a considerable antioxidant activity13–15. Other interesting bioactivities attributed to P. maritimum extracts include anti-inflammatory and anti-diabetic16, neuroprotective14 and anti-microbial13, suggesting its potential application in several industrial fields. Although extracts of P. maritimum have been previously reported to be rich in phenolic compounds and to display antioxidant properties, to our knowledge there are no reports of its antigenotoxicity. Thus, the aim of this study was to evaluate the antioxidant and antigenotoxic properties of an ethanol extract of P. maritimum from the south of Portugal, which could reinforce its potential application in the food and nutraceutical areas.
## Plant material and extraction
The aerial part of *Polygonum maritimum* L. was collected in August 2018 at the Faro beach (south of Portugal; coordinates: 37° 0′ 30.0852′′ N, 7° 59′ 45.2616′′ W). The plant species was identified by Dr. Luísa Custódio (Centre of Marine Sciences, University of Algarve, Portugal) and a voucher specimen was kept in the XtremeBio laboratory (voucher code MBH22). All methods used in the experimental research on plants, including the collection of plant material, complied with relevant institutional, national, namely by the Portuguese Institute for Nature Conservation and Forests (ICNF), and international guidelines and legislation. The plant material was washed, oven dried at 40 °C for five days and reduced to a fine powder. The extraction occurred in absolute ethanol in a flask protected from the light at 25 °C and 100 revolutions per minute (rpm) for 24 h. The P. maritimum L. aerial part ethanol extract (PME) was filtered (Whatman grade 4 paper and with syringe filter 0.2 µm), the major part of the solvent was evaporated in a Rotavapor under low pressure at 40 °C and 50 rpm, and the remaining solvent in the concentrated extract was evaporated with gaseous nitrogen. The dried extract was stored at − 20 °C until use.
## Total phenolic content
The total phenolic content (TPC) of PME was estimated by the Folin-Ciocalteu method17 adapted for microplate assay. For this assay, 10 μL of sample (25–175 μg/mL; dissolved in absolute ethanol) were mixed with 50 μL of a diluted solution of Folin-Ciocalteu reagent (1:10 v/v in deionized water) and 40 μL of Na2CO3 (75 mg/mL in deionized water), incubated at room temperature in the dark for 1 h, and the absorbance was measured at 760 nm. The assay included blanks and controls, where the Folin-Ciocalteu reagent and Na2CO3 were replaced by deionized water and the samples by absolute ethanol, respectively. The absorbance values calculated (absorbance = absorbance sample − absorbance blank − absorbance control) were correlated with a gallic acid standard curve to express the TPC of PME in mg of gallic acid equivalents (GAE)/g of dry weight (DW).
## Ortho-diphenol content
The ortho-diphenol content (ODC) of PME was determined according to Domínguez-Perles et al.18, with some alterations for use in microplate. In brief, 160 μL of sample (150–300 μg/mL; dissolved in $50\%$ ethanol) were added to the microplate, followed by the addition of 40 μL of sodium molybdate (50 mg/mL in $50\%$ ethanol). The microplate was incubated at room temperature in the dark for 15 min and the absorbance was read at 370 nm. Blanks and controls were included in the experiment, where sodium molybdate and samples were substituted by ethanol $50\%$, respectively. The ODC of PME was expressed in mg of GAE/g of DW by correlating the absorbance values of the samples (absorbance = absorbance sample − absorbance blank − absorbance control) with a gallic acid standard curve.
## Total flavonoid content
The total flavonoid content (TFC) of PME was estimated using the methodology described by Kumazawa et al.19, with minor modifications to adapt for microplate assay. Briefly, 50 μL of AlCl3 (20 mg/mL in absolute ethanol) were added to 50 μL of sample (1000–2000 μg/mL; dissolved in absolute ethanol), the mixture was incubated at room temperature in the dark for 1 h and the absorbance was measured at 420 nm. The assay included blanks and controls, where AlCl3 and the samples were replaced by absolute ethanol, respectively. The absorbance values (absorbance = absorbance sample − absorbance blank − absorbance control) were correlated with a quercetin standard curve to express the TFC of PME in mg of quercetin equivalents (QE)/g of DW.
## Standards and reagents
Acetonitrile ($99.9\%$) was of High-Performance Liquid Chromatography (HPLC) grade from Fisher Scientific (Lisbon, Portugal). Formic acid was purchased from Panreac Química S.L.U. (Barcelona, Spain), and phenolic standards were from Extrasynthèse (Genay, France). Water was treated in a Milli-Q water purification system (TGI Pure Water Systems, USA).
## Phenolic compounds analysis
The phenolic profile of PME was determined in the dried extract re-dissolved in a methanol:water (80:20, v/v) mixture by liquid chromatography with diode array detection and electrospray ionization tandem mass spectrometry (LC-DAD-ESI/MSn) (Dionex Ultimate 3000 UPLC, Thermo Scientific, San Jose, CA, USA), as previously described by Bessada et al.20. For the double online detection, 280 and 370 nm were used as preferred wavelengths for diode array detection (DAD) and in a mass spectrometer (MS) connected to HPLC system via the DAD cell outlet. The MS detection was performed in negative mode, using a Linear Ion Trap LTQ XL mass spectrometer (ThermoFinnigan, San Jose, CA, USA) equipped with an electrospray ionization (ESI) source. The identification of the phenolic compounds was performed using standard compounds, when available, by comparing their retention times, UV–Vis and mass spectra; and also, comparing the obtained information with available data reported in the literature giving a tentative identification. For quantitative analysis, a 7-level calibration curve for each available phenolic standard was constructed based on the UV signal [catechin ($y = 84.950$x − 23.200, R2 = 0.999, LOD (Limit of detection) = 0.17 μg/mL; LOQ (Limit of quantification) = 0.68 μg/mL, peaks 1, 2, 3, 4, and 5), myricetin ($y = 23287$x − 581,708, R2 = 0.9988, LOD = 61.21 µg/mL and LOQ = 185.49 µg/mL, peaks 6 and 7) and quercetin-3-O-glucoside ($y = 34843$x − 160,173, R2 = 0.9998; LOD = 0.21 μg/mL; LOQ = 0.71 μg/mL, peak 8)]. For the identified phenolic compounds for which a commercial standard was not available, the quantification was performed through the calibration curve of the most similar available standard. The results were expressed as mg/g of extract DW.
## Eukaryotic model and growth conditions
The budding yeast *Saccharomyces cerevisiae* was selected as eukaryotic cell model for this work. The strains BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) and CEN.PK dDEL (HIS3 Δ::dDEL; created by Silva et al.21) were used for measurement of antioxidant activity in viability assays (colony-forming units, CFU’s) and antigenotoxicity using the dominant deletion (dDEL) assay, respectively. The strain CEN.PK dDEL was created by replacing the HIS3 gene from the plasmid pPS1 with the dDEL cassette, which is limited by two partial alleles of the hphMX6Δ marker that comprise between them sequences with homology and a marker for geneticin (G418) resistance, in the laboratory strain CEN.PK 102-3A (Mata ura3-52 HIS3 leu2-3112 TRP1 MAL2-8c SUC2)21. Upon double-strand break in the region comprised between the partial alleles of the hphMX6Δ marker, the homologous recombination repair pathway is activated, leading to the reversion of the marker. Due to this process, the strain loses the marker for geneticin resistance and becomes resistant to hygromycin B (HygB). The strains BY4741 and CEN.PK dDEL were cultured every week on solid rich medium [YPDA; composed of $1\%$ (w/v) yeast extract (Acros Organics), $2\%$ (w/v) peptone (BD Bacto), $2\%$ (w/v) glucose, $2\%$ (w/v) agar (Liofilchem)] and YPDA supplemented with 300 µg/mL geneticin, respectively, and stored at 4 °C. For each experiment, one colony from the cultures at 4 °C was suspended in liquid rich medium (YPD; same composition as YPDA, except agar) and YPD supplemented with 400 µg/mL geneticin for BY4741 and CEN.PK dDEL strains, respectively, and incubated at 30 °C and 200 rpm. Cell proliferation was monitored by measuring the optical density at 600 nm (OD600). In the dDEL assay, the recombinant cells were selected on YPDA medium supplemented with 100 µg/mL HygB.
## DPPH radical scavenging assay
The capacity of PME to scavenge the free radical 2,2-diphenyl-1-picrylhydrazyl (DPPH) was assessed with an adapted procedure from Mitra and Uddin22. In brief, 50 μL of sample (1–10 μg/mL; dissolved in absolute ethanol) were mixed with 100 μL of DPPH solution (0.04 mg/mL), incubated at room temperature in the dark for 20 min and the absorbance was measured at 517 nm. The assay included blanks and controls, where DPPH and samples were replaced by absolute ethanol, respectively. Gallic acid was used as standard. The absorbance values obtained were used to calculate the percentage of DPPH inhibition as follows: (control absorbance − (sample absorbance − blank absorbance))/control absorbance. The IC50 values were calculated and represent the concentration required to inhibit $50\%$ of the initial amount of DPPH radical.
## Cell viability under H2O2-induced oxidative stress
A liquid culture of BY4741 grown overnight was diluted to OD600 = 0.2 and incubated for at least two generations further (~ 4 h; OD600 = 0.8). The culture was then divided into aliquots and 100 μL were taken at time 0 min. Next, the aliquots were treated as follows: negative control, $1.3\%$ absolute ethanol; positive control, 5 mM H2O2 and $1.3\%$ absolute ethanol; control of extract, 250 μg/mL PME; co-incubation, 250 μg/mL PME and 5 mM H2O2; followed by incubation at 30 °C and 200 rpm. Aliquots of 100 μL were taken at 10, 20 and 30 min after the addition of treatments, serially diluted in sterile deionized water to 10−4 and spread on YPDA, followed by incubation at 30 °C for 48 h. The viability (%) was assessed by counting CFU’s and dividing the number of colonies in each time point by the number of colonies at 0 min (before the addition of treatment). Time 0 min was assumed as $100\%$ viability.
## Antigenotoxicity
The antigenotoxicity of PME was evaluated using the dDEL assay according to Silva et al.21. Briefly, a liquid culture of CEN.PK dDEL grown overnight was diluted to OD600 = 0.2 and incubated for 4 h until OD600 = 0.8. The culture was then divided into aliquots, 200 μL were taken at time 0 min, followed by the addition of treatments as described in the previous section, except for the use of 4 mM H2O2 instead of 5 mM. Then, 200 μL aliquots were taken from each situation after 10 min and diluted to the final volume of 1 mL. The samples were centrifuged at maximum speed for 30 s, 950 μL of supernatant were discarded and the pellet was suspended in 950 μL of YPD, followed by incubation at 30 °C, 200 rpm, for 2 h to guarantee sufficient expression of the hphMX6 marker. Afterwards, each sample was serially diluted in sterile deionized water to 10−4. For each situation, 100 μL from the undiluted samples were placed on YPDA supplemented with 100 μg/mL HygB and 100 μL from the 10−4 dilution on YPDA medium, followed by incubation at 30 °C for 48 h. The dDEL recombination frequency was calculated by dividing the number of recombinant colonies on YPDA + HygB medium by the number of viable colonies on YPDA medium, and then normalized to the positive control (4 mM H2O2).
Excessive levels of ROS can cause genotoxic damage, which may contribute to the first stages of carcinogenesis. Polyphenol-rich extracts with strong antioxidant properties may protect the DNA from oxidative damage and, in this way, provide an antigenotoxic effect. Thus, the potential antigenotoxicity of PME was assessed after exposure to H2O2-induced oxidative stress in S. cerevisiae (CEN.PK dDEL), using the dDEL assay. This assay is based on the principle that the dDEL recombination frequency increases if an agent induces DNA double-strand breaks within the dDEL cassette, which activates the homologous recombination DNA repair pathway and leads to the reversion of the hphMX6 marker, turning the strain resistant to HygB21. Therefore, the dDEL recombination frequency reflects the number of DNA double-strand breaks that occurred in each treatment. The time-point 10 min was chosen in accordance with the result from the viability assay (Fig. 2) that showed the most significant protective effect against oxidative stress ($p \leq 0.001$). The dDEL recombination frequency observed in the negative control was considered the basal level and it was similar to that of the extract (Fig. 3), indicating that 250 µg/mL PME did not induce DNA damage. On the other hand, the treatment with 4 mM H2O2 increased the dDEL recombination frequency, implying that H2O2 caused DNA double-strand breaks (Fig. 3). This genotoxic effect was mitigated by the treatment with 250 µg/mL PME, since the dDEL recombination frequency in the co-treatment was lower than in the positive control ($p \leq 0.05$, Fig. 3). These results indicate that PME prevented H2O2-induced DNA double-strand breaks in a eukaryotic cell model. Figure 3Dominant deletion (dDEL) recombination frequency after exposure to H2O2-induced oxidative stress and P. maritimum aerial part ethanol extract (PME) in S. cerevisiae (CEN.PK dDEL). Cultures were treated for 10 min with: 4 mM H2O2 in the positive control; 4 mM H2O2 and 250 µg/mL PME in the co-treatment; solvent of the extract in the negative and positive controls; 250 µg/mL PME in the control of extract. The dDEL recombination frequency (%) was calculated by dividing the number of recombinant colonies on medium supplemented with hygromycin B by the number of viable colonies on medium without hygromycin B. Data are presented as the mean of three independent experiments ± SEM normalized to the positive control. Statistical analysis was assessed between positive control and co-treatment with Student’s t-test, *$p \leq 0.05.$
Oxidative stress imposed by H2O2 led to a genotoxic effect in S. cerevisiae, causing DNA double-strand breaks that triggered the activation of the homologous recombination DNA repair pathway measured with the dDEL assay. The co-incubation with PME decreased the dDEL recombination frequency imposed by H2O2, suggesting a lower DNA damage and consequently the homologous recombination pathway was activated at a lower extent. This DNA protective effect provided by PME likely results from the neutralization of H2O2, preventing the Fenton reaction, or of •OH before its potential interaction with the DNA. In line with this, extracts from other Polygonum species, Polygonum aviculare L. ethanol extract and *Polygonum cuspidatum* Sieb. et Zucc. ethanol and ethyl acetate root extracts, protected ΦX174 RF1 supercoiled DNA from hydroxyl radical-induced DNA strand scission (generated by UV photolysis of H2O2) in a dose-dependent manner, suggesting an in vitro DNA protective effect under oxidative stress44,45. Other plant extracts rich in polyphenols46,47 or isolated polyphenols, like quercetin48 and myricetin41, have also been reported to exhibit antigenotoxic effects against H2O2, in part, attributed to their antioxidant properties. Interestingly, a procyanidin extract revealed a stronger antigenotoxic effect against H2O2 than catechin and epicatechin, corroborating the stronger antioxidant activity suggested for oligomers48. Hence, PME has the capacity to prevent oxidative DNA damage.
The antioxidant and antigenotoxic properties of PME presented in our study should be further complemented with experiments performed in human cell lines and in animal models, using established methodologies to measure the redox state and DNA damage like the cell-permeable probe 2',7'-dichlorodihydrofluorescein diacetate and the single-cell gel electrophoresis assay, respectively, as these are expected to produce results closer to an in vivo situation where the extract may be consumed by humans to prevent genotoxicity.
## Statistical analysis
All data are presented as the mean of three independent experiments with associated standard error of the mean (mean ± SEM). The statistical analysis was performed with the GraphPad Prism (version 8.2.1), using the Student’s t-test to analyze the differences between samples. Differences were considered statistically significant when $p \leq 0.05.$
## Chemical composition
Phenolic compounds are often associated with a strong antioxidant capacity and with other bioactivities of interest, such as anti-inflammatory and antigenotoxic. Thus, the chemical composition of PME was assessed in terms of total phenol, ortho-diphenol and flavonoid contents by colorimetric methods. The results indicated that PME has a high content in total phenols and ortho-diphenols, and that part of them may be flavonoids (Table 1). The PME showed TPC similar to an ethanol extract of P. maritimum from the south of Portugal (241 ± 4 mg GAE/g DW)15, but lower than a methanol extract from the Algerian coast (352.49 ± 18.03 mg GAE/g DW)13 and higher than methanol extracts of stems (75.9 mg GAE/g DW) and leaves (70.40 mg GAE/g DW) of the same species collected in southern Tunisia23, suggesting that the geographical location may have influenced the TPC of P. maritimum extracts. Phenolic compounds with ortho-diphenol structure are likely to provide a strong antioxidant activity. Hydroxyl groups can neutralize free radicals and their adjacent position in the aromatic ring allows the stabilization of the resultant phenoxyl radical by forming an intramolecular hydrogen bond24. Flavonoids are expected to play a major part in the antioxidant capacity of PME, as they are considered main secondary metabolites in the Polygonum genus25. Accordingly, flavonoids were indicated as the predominant phenolic constituent of a P. maritimum aerial part acetone extract from the south of Portugal that revealed antioxidant properties15. Flavonoids were detected in PME; however, its TFC value was lower than that of P. maritimum methanol extracts from southern Tunisia23, perhaps due to differences in the geographical location or harvesting conditions. Table 1Chemical composition of P. maritimum aerial part ethanol extract (PME) determined by in vitro colorimetric assays: total phenolic content (TPC) and ortho-diphenol content (ODC) expressed as mg of gallic acid equivalents (GAE)/g of dry weight (DW) and total flavonoid content (TFC) expressed as mg of quercetin equivalents (QE)/g of DW. Data are presented as mean of three independent experiments performed in triplicate with associated standard error of the mean (mean ± SEM).SampleTPC (mg GAE/g DW)ODC (mg GAE/g DW)TFC (mg QE/g DW)PME249.03 ± 1.28315.71 ± 4.6117.18 ± 0.56 The chemical composition of PME was also analysed by LC-DAD-ESI/MSn, which confirmed the presence of phenolic compounds and flavonoids detected in the in vitro colorimetric assays. Eight phenolic compounds were tentatively identified and quantified in the extract (Table 2), revealing the dominance of flavan-3-ols over flavonol glycosides. A representative chromatogram of the phenolic profile of PME, recorded at 280 and 370 nm, is presented in Fig. 1. Flavan-3-ols and flavonol glycosides have been previously described in extracts of P. maritimum12,14,15,23 and were suggested to be correlated with their biological properties, namely antioxidant, anti-inflammatory, anti-melanogenic and neuroprotective activities. Additionally, β-type (epi)catechin oligomers were found in the rhizome extract of *Polygonum paleaceum* Wall. ex Hook. f.28. The concentration of flavan-3-ols was 29.5 ± 1.2 mg/g extract DW, which represented approximately ~ $68\%$ of the total phenolic compounds found (43.3 ± 0.3 mg/g extract DW) and double of the obtained for the flavonols (13.7 ± 0.2 mg/g extract DW, ~ $32\%$ of the total phenolic). Flavan-3-ols detected included catechin (monomer) and four β-type (epi)catechin oligomers (dimer and trimers). Flavonols included two myricetin glycosides and one quercetin glycoside. Flavan-3-ols and flavonols have attracted considerable attention due to their recognized bioactivities that include antioxidant, anti-inflammatory, cardioprotective, anticancer, anti-viral and anti-microbial properties7,29–31. Overall, these results indicate that PME is composed of phenolic compounds that are likely to provide many valuable properties, including a remarkable antioxidant activity. Table 2Chemical composition of the P. maritimum aerial part ethanol extract determined by liquid chromatography with diode array detection and electrospray ionization tandem mass spectrometry (LC-DAD-ESI/MSn). The retention times (Rt), wavelengths of maximum absorption (λmax) in the UV–Vis region, main [M–H]− and respective fragment ions (MS2), tentative identification, chemical class, and quantification (mg/g of extract DW) of phenolic compounds are presented. NoRt (min)λmax (nm)[M-H]- (m/z)MS2 (m/z)Tentative identificationChemical classQuantification (mg/g DW)14.90278577577[72], 575[39], 425[16], 407[10], 289[6], 287[12]β-type (epi)catechin dimer26Favan-3-ol3.6 ± 0.225.08280865739[83], 695[100], 577[74], 575[47], 425[12], 289[10], 287[8]β-type (epi)catechin trimer26Favan-3-ol4.4 ± 0.435.95280865739[74],695[100],577[36], 575[62],425[15],289[12],287[14]β-type (epi)catechin trimer26Favan-3-ol4.6 ± $\frac{0.146.32281}{311289245}$[35], 203[15], 137[29](+)-Catechin15,26Favan-3-ol7.5 ± 0.257.42280865739[74], 695[100], 577[36], 575[62], 425[15], 289[12], 287[14]β-type (epi)catechin trimer26Favan-3-ol9.4 ± 0.1616.73349463317[100]Myricetin-O-deoxyhexoside27Flavonol8.0 ± 0.1718.65348463317[100]Myricetin-O-deoxyhexoside27Flavonol5.1 ± 0.1821.43342447301[100]Quercetin-O-deoxyhexoside[27]Flavonol0.64 ± 0.01Total phenolic compounds43.3 ± 0.3Total flavan-3-ols29.5 ± 1.2Total flavonols13.7 ± 0.2Figure 1Representative phenolic profile of the P. maritimum aerial part ethanol extract recorded at 280 nm (a) and 370 nm (b). Peak identification is presented in Table 2.
## Antioxidant activity
The in vitro antioxidant activity of PME was evaluated by measuring its capacity to scavenge the free radical DPPH. The extract showed the capacity to scavenge $50\%$ of the free radical at low concentration (IC50 = 2.29 ± 0.10 µg/mL, data not shown), revealing a strong antioxidant effect. The DPPH scavenging capacity of PME was lower than that of the standard used, gallic acid (IC50 = 0.86 ± 0.03 µg/mL, $p \leq 0.001$, data not shown), but higher than the activities reported for phenolic extracts recognized for their antioxidant properties such as *Ginkgo biloba* L. (IC50 = 33.91 ± 1.16 µg/mL) or *Camellia sinensis* L. (IC50 = 14.50 ± 1.69 µg/mL)32, which are often used as standards in in vitro antioxidant assays. In addition, PME revealed higher DPPH scavenging capacity than P. maritimum methanol extracts prepared with aerial parts from Algeria (IC50 = 7.71 ± 1.88 µg/mL)13 or with leaves from the south of Portugal (IC50 = 26.0 ± 0.7 µg/mL)16, possibly due to differences in the weather conditions at the time of harvesting. The remarkable antioxidant activity observed in our study seems to be a feature of halophytes, although typically not as strong as the one displayed by PME. Studies have reported relatively low IC50 values in the DPPH scavenging assay for extracts of halophytes, such as the species *Tamarix gallica* L. (aerial parts methanol extract, IC50 = 14.05 ± 0.66 µg/mL)33, *Limonium delicatulum* (methanol extracts of leaves and roots, IC50 = 10.58 ± 0.18 µg/mL and IC50 = 5.79 ± 0.05 µg/mL, respectively)34, and *Arthrocnemum indicum* (Willd.) Moq. ( shoots ethanol extract, IC50 = 7.17 ± 1.26 μg/mL)35, associated to a high content in phenolic compounds. Although fruits and vegetables are the common source of polyphenols in the human diet and whose consumption has been associated with many health benefits, in part due to their antioxidant properties, it should be noted that most of these foods do not exhibit an antioxidant activity comparable to that of halophytes like P. maritimum. For instance, from a study with methanol extracts of 8 wild edible leafy vegetables, only the species *Cassia tora* (IC50 = 9.898 µg/mL) revealed a high antioxidant activity, but still considerably lower than PME, whereas the rest of the 7 vegetables showed lower antioxidant activity as indicated by the higher IC50 values that ranged from 33.82 to 45.68 µg/mL36. Moreover, methanol extracts of commonly consumed vegetables like spinach (IC50 = 200.4 ± 2.1 µg/mL), carrot (IC50 = 97.6 ± 2.1 µg/mL), and radish (IC50 = 155.7 ± 0.8 µg/mL)37 revealed a considerably higher IC50 than PME, evidencing their lower antioxidant capacity. Therefore, halophytes like P. maritimum could represent a source of polyphenols with more potent antioxidant properties than those that are provided by conventional foods.
Since PME revealed a strong in vitro antioxidant activity, it would be important to understand if the extract could provide an antioxidant effect in a biological context. Due to the similarities shared with cells of higher eukaryotes regarding oxidative stress response mechanisms, the yeast S. cerevisiae (BY4741) was selected as eukaryotic cell model to test if PME can protect cells from H2O2-induced oxidative stress. The results showed that the treatment with the solvent used in the extraction (ethanol, negative control) or with the PME, at the concentration of 250 µg/mL, up to 30 min caused no effect on cell viability (Fig. 2), suggesting absence of toxicity. Treatment with 5 mM H2O2 caused loss of cell viability over time in the positive control, whereas in the co-treatment with 250 µg/mL PME this effect was attenuated at 10, 20 and 30 min ($p \leq 0.001$, $p \leq 0.05$ and $p \leq 0.05$, respectively; Fig. 2). This protective effect likely results from the neutralization of H2O2, which is in line with the strong in vitro antiradical activity attributed to PME and, therefore, suggests a potential antiradical action of the extract against the radical •OH formed by H2O2 and intracellular transition metal ions in the Fenton reaction. The protective effect of PME against H2O2 is in accordance with results reported by Rodrigues et al.14, where the co-treatment of a P. maritimum methanol extract with H2O2 prevented oxidative stress-induced cytotoxicity in the human neuroblastoma cell line SH-SY5Y. Moreover, a P. maritimum methanol extract was reported to have a higher in vitro H2O2 scavenging activity than the antioxidants α-tocopherol and butylated hydroxyanisole13. The capacity to scavenge H2O2 is considered a valuable feature since it can prevent the Fenton reaction that generates the highly reactive OH, which is able to damage the DNA, and overall can help control intracellular ROS levels to avert oxidative damage. Figure 2Effect of P. maritimum aerial part ethanol extract (PME) on S. cerevisiae (BY4741) viability under oxidative stress. Exponentially growing cultures were treated with 5 mM H2O2 (positive control; diamond) or 5 mM H2O2 and 250 µg/mL PME (co-treatment; square). The negative (circle) and positive controls contained the solvent of the extract and the control of extract contained 250 µg/mL PME (triangle). Colony-forming units were counted to calculate the viability (%) in each time point, assuming $100\%$ at 0 min. Data are presented as the mean of three independent experiments ± SEM. Statistical analysis was performed between positive control and co-treatment with Student’s t-test at 10, 20 and 30 min ($p \leq 0.001$, $p \leq 0.05$ and $p \leq 0.05$, respectively).
The antioxidant effects observed in our study reflect the polyphenol-rich composition of PME. Compounds related to those found in PME like catechin, epicatechin, myricetin and quercetin have been described to display higher DPPH scavenging activities than the antioxidant Trolox38 and to provide a high protective effect against H2O2 in cell-based assays39–41. Although quercetin and myricetin are present in PME in their glycosylated forms, which are usually associated with a lower antioxidant effect than their respective aglycones42, they are still expected to contribute significantly to the protective effect described in our study. In line with this, Rha et al.30 reported that a flavonol glycoside fraction from green tea induced a relevant protective effect against H2O2 in an intracellular oxidation assay in rat pheochromocytoma PC-12 cells. ( Epi)catechin oligomers found in PME may also play a major role in the antioxidant effect displayed by the extract since the degree of oligomerization of (epi)catechin seems to influence this bioactivity. Accordingly, Wang et al.28 reported that (epi)catechin trimers revealed higher DPPH scavenging activity than monomers and dimers present in P. paleaceum rhizome extract and than the antioxidants ascorbic acid and gallic acid. In addition, Roig et al.43 reported that a procyanidin extract from grape seed, composed of monomers, dimers, trimers and higher polymers, was the most effective in decreasing the lipid peroxidation and the levels of oxidized glutathione caused by H2O2 in comparison with the respective monomers in FaO rat hepatoma cell line. Altogether, our results indicate that PME exhibits antioxidant effect in cell-free and cell-based assays, suggesting a potential to prevent oxidative stress-related damage.
## Conclusion
In this work, PME showed high in vitro antioxidant capacity evidenced by the low IC50 value determined in the DPPH radical scavenging assay, possibly linked to its high content in phenolic compounds. The antioxidant properties of PME were confirmed by the protection of cell viability against H2O2-induced oxidative stress and its antigenotoxic potential was revealed by the decrease in the dDEL recombination frequency imposed by H2O2 in the model organism S. cerevisiae. To our knowledge, this is the first time that a P. maritimum extract is reported for its capacity to protect the DNA from H2O2-induced oxidative damage, which is likely to result from the ROS scavenging activity provided by polyphenols. Our study suggests that P. maritimum represents a source of antioxidant and antigenotoxic metabolites with potential application in food and nutraceutical industries to mitigate oxidative and genotoxic effects such as those induced in the gastrointestinal tract by dietary xenobiotics. Thus, P. maritimum or its metabolites may play a role in the prevention of malignancy as a source of natural bio-antioxidants or as a functional food ingredient.
## References
1. Elmassry MM, Zayed A, Farag MA. **Gut homeostasis and microbiota under attack: Impact of the different types of food contaminants on gut health**. *Crit. Rev. Food Sci. Nutr.* (2022) **62** 738-763. DOI: 10.1080/10408398.2020.1828263
2. Bhattacharyya A, Chattopadhyay R, Mitra S, Crowe SE. **Oxidative stress: An essential factor in the pathogenesis of gastrointestinal mucosal diseases**. *Physiol. Rev.* (2014) **94** 329-354. DOI: 10.1152/physrev.00040.2012
3. Thanan R. **Oxidative stress and its significant roles in neurodegenerative diseases and cancer**. *Int. J. Mol. Sci.* (2015) **16** 193-217. DOI: 10.3390/ijms16010193
4. Asmat U, Abad K, Ismail K. **Diabetes mellitus and oxidative stress—A concise review**. *Saudi Pharm. J.* (2016) **24** 547-553. DOI: 10.1016/j.jsps.2015.03.013
5. Penning TM, Penning TM. **Metabolic activation of chemical carcinogens**. *Chemical Carcinogenesis* (2011) 135-158
6. Costa C. **Current evidence on the effect of dietary polyphenols intake on chronic diseases**. *Food Chem. Toxicol.* (2017) **110** 286-299. DOI: 10.1016/j.fct.2017.10.023
7. Zhang H, Tsao R. **Dietary polyphenols, oxidative stress and antioxidant and anti-inflammatory effects**. *Curr. Opin. Food Sci.* (2016) **8** 33-42. DOI: 10.1016/j.cofs.2016.02.002
8. Azqueta A, Collins A. **Polyphenols and DNA damage: A mixed blessing**. *Nutrients* (2016) **8** 785. DOI: 10.3390/nu8120785
9. Ozgur R, Uzilday B, Sekmen AH, Turkan I. **Reactive oxygen species regulation and antioxidant defence in halophytes**. *Funct. Plant Biol.* (2013) **40** 832-847. DOI: 10.1071/FP12389
10. Petropoulos SA, Karkanis A, Martins N, Ferreira ICFR. **Edible halophytes of the Mediterranean basin: Potential candidates for novel food products**. *Trends Food Sci. Technol.* (2018) **74** 69-84. DOI: 10.1016/j.tifs.2018.02.006
11. Qasim M. **Antioxidant properties, phenolic composition, bioactive compounds and nutritive value of medicinal halophytes commonly used as herbal teas**. *S. Afr. J. Bot.* (2017) **110** 240-250. DOI: 10.1016/j.sajb.2016.10.005
12. Kazantzoglou G, Magiatis P, Kalpoutzakis E, Skaltsounis A-L. **Polygonophenone, the first MEM-substituted natural product, from**. *J. Nat. Prod.* (2009) **72** 187-189. DOI: 10.1021/np800762x
13. El-Haci IA, Bekkara FA, Mazari W, Hassani F, Didi MA. **Screening of biological activities of**. *Asian Pac. J. Trop. Biomed.* (2013) **3** 611-615. DOI: 10.1016/S2221-1691(13)60124-0
14. Rodrigues MJ. **In vitro and in silico approaches to appraise**. *Ind. Crops Prod.* (2018) **111** 391-399. DOI: 10.1016/j.indcrop.2017.10.046
15. Rodrigues MJ. **Sea knotgrass (**. *Ind. Crops Prod.* (2019) **128** 391-398. DOI: 10.1016/j.indcrop.2018.11.038
16. Rodrigues MJ. **Unlocking the in vitro anti-inflammatory and antidiabetic potential of**. *Pharm. Biol.* (2017) **55** 1348-1357. DOI: 10.1080/13880209.2017.1301493
17. Singleton VL, Orthofer R, Lamuela-Raventós RM. **Analysis of total phenols and other oxidation substrates and antioxidants by means of folin-ciocalteu reagent**. *Methods Enzymol.* (1999) **299** 152-178. DOI: 10.1016/S0076-6879(99)99017-1
18. Domínguez-Perles R, Teixeira AI, Rosa E, Barros AI. **Assessment of (poly)phenols in grape (**. *Food Chem.* (2014) **164** 339-346. DOI: 10.1016/j.foodchem.2014.05.020
19. Kumazawa S, Hamasaka T, Nakayama T. **Antioxidant activity of propolis of various geographic origins**. *Food Chem.* (2004) **84** 329-339. DOI: 10.1016/S0308-8146(03)00216-4
20. Bessada SMF, Barreira JCM, Barros L, Ferreira ICFR, Oliveira MBPP. **Phenolic profile and antioxidant activity of**. *Ind. Crops Prod.* (2016) **89** 45-51. DOI: 10.1016/j.indcrop.2016.04.065
21. Silva PC, Domingues L, Collins T, Oliveira R, Johansson B. **Quantitative assessment of DNA damage in the industrial ethanol production strain**. *FEMS Yeast Res.* (2018) **18** foy101. DOI: 10.1093/femsyr/foy101
22. Mitra K, Uddin N. **Total phenolics, flavonoids, proanthrocyanidins, ascorbic acid contents and in-vitro antioxidant activities of newly developed isolated soya protein**. *Dis. J. Agri. Food Sci.* (2014) **2** 160-168
23. Mahmoudi M. **Analysis of**. *Anal. Lett.* (2021) **54** 2940-2955. DOI: 10.1080/00032719.2021.1906267
24. Visioli F, Galli C. **Olive oil phenols and their potential effects on human health**. *J. Agric. Food Chem.* (1998) **46** 4292-4296. DOI: 10.1021/jf980049c
25. Shen B-B. **Analysis of the phytochemistry and bioactivity of the genus**. *Digit. Chin. Med.* (2018) **1** 19-36. DOI: 10.1016/S2589-3777(19)30005-9
26. Barros L. **The powerful in vitro bioactivity of**. *Ind. Crops Prod.* (2015) **76** 318-322. DOI: 10.1016/j.indcrop.2015.05.086
27. Llorent-Martínez EJ. **Phytochemical characterization, in vitro and in silico approaches for three**. *New J. Chem.* (2018) **42** 5204-5214. DOI: 10.1039/C8NJ00347E
28. Wang K-J, Zhang Y-J, Yang C-R. **Antioxidant phenolic compounds from rhizomes of**. *J. Ethnopharmacol.* (2005) **96** 483-487. DOI: 10.1016/j.jep.2004.09.036
29. Aron PM, Kennedy JA. **Flavan-3-ols: Nature, occurrence and biological activity**. *Mol. Nutr. Food Res.* (2008) **52** 79-104. DOI: 10.1002/mnfr.200700137
30. Rha C-S. **Antioxidative, anti-inflammatory, and anticancer effects of purified flavonol glycosides and aglycones in green tea**. *Antioxidants* (2019) **8** 278. DOI: 10.3390/antiox8080278
31. Barreca D. **Food flavonols: Nutraceuticals with complex health benefits and functionalities**. *Trends Food Sci. Technol.* (2021) **117** 194-204. DOI: 10.1016/j.tifs.2021.03.030
32. Stanković MS, Petrović M, Godjevac D, Stevanović ZD. **Screening inland halophytes from the central Balkan for their antioxidant activity in relation to total phenolic compounds and flavonoids: Are there any prospective medicinal plants?**. *J. Arid Environ.* (2015) **120** 26-32. DOI: 10.1016/j.jaridenv.2015.04.008
33. Lefahal M. **The cosmetic potential of the medicinal halophyte**. *Comb. Chem. High Throughput Screen.* (2021) **24** 1671-1678. DOI: 10.2174/1386207323666201204141541
34. Bakhouche I. **Phenolic contents and in vitro antioxidant, anti-tyrosinase, and anti-inflammatory effects of leaves and roots extracts of the halophyte**. *S. Afr. J. Bot.* (2021) **139** 42-49. DOI: 10.1016/j.sajb.2021.01.030
35. Hajlaoui H. **HPLC-MS profiling, antioxidant, antimicrobial, antidiabetic, and cytotoxicity activities of**. *Plants* (2022) **11** 232. DOI: 10.3390/plants11020232
36. Aryal S. **Total phenolic content, flavonoid content and antioxidant potential of wild vegetables from western Nepal**. *Plants* (2019) **8** 96. DOI: 10.3390/plants8040096
37. Ghoora MD, Haldipur AC, Srividya N. **Comparative evaluation of phytochemical content, antioxidant capacities and overall antioxidant potential of select culinary microgreens**. *J. Agric. Food Res.* (2020) **2** 100046
38. Zhang D. **Analysis of the antioxidant capacities of flavonoids under different spectrophotometric assays using cyclic voltammetry and density functional theory**. *J. Agric. Food Chem.* (2011) **59** 10277-10285. DOI: 10.1021/jf201773q
39. Chen Z. **Effects of quercetin on proliferation and H**. *Molecules* (2018) **23** 2012. DOI: 10.3390/molecules23082012
40. Ruijters EJB, Weseler AR, Kicken C, Haenen GRMM, Bast A. **The flavanol (-)-epicatechin and its metabolites protect against oxidative stress in primary endothelial cells via a direct antioxidant effect**. *Eur. J. Pharmacol.* (2013) **715** 147-153. DOI: 10.1016/j.ejphar.2013.05.029
41. Wang ZH. **Myricetin suppresses oxidative stress-induced cell damage via both direct and indirect antioxidant action**. *Environ. Toxicol. Pharmacol.* (2010) **29** 12-18. DOI: 10.1016/j.etap.2009.08.007
42. Xiao J. **Dietary flavonoid aglycones and their glycosides: Which show better biological significance?**. *Crit. Rev. Food Sci. Nutr.* (2017) **57** 1874-1905. PMID: 26176651
43. Roig R, Cascón E, Arola L, Bladé C, Salvadó MJ. **Procyanidins protect Fao cells against hydrogen peroxide-induced oxidative stress**. *Biochim. Biophys. Acta Gen. Subj.* (2002) **1572** 25-30. DOI: 10.1016/S0304-4165(02)00273-8
44. Hsu C-Y. **Antioxidant activity of extract from**. *Biol. Res.* (2006) **39** 281-288. DOI: 10.4067/S0716-97602006000200010
45. Lin Y-W, Yang F-J, Chen C-L, Lee W-T, Chen R-S. **Free radical scavenging activity and antiproliferative potential of**. *J. Nat. Med.* (2010) **64** 146-152. DOI: 10.1007/s11418-009-0387-8
46. Oliveira D, Latimer C, Parpot P, Gill CIR, Oliveira R. **Antioxidant and antigenotoxic activities of**. *Eur. J. Nutr.* (2020) **59** 465-476. DOI: 10.1007/s00394-019-01915-8
47. Sassi A. **Assessment in vitro of the genotoxicity, antigenotoxicity and antioxidant of**. *Regul. Toxicol. Pharmacol.* (2016) **77** 117-124. DOI: 10.1016/j.yrtph.2016.02.009
48. Llópiz N. **Antigenotoxic effect of grape seed procyanidin extract in Fao cells submitted to oxidative stress**. *J. Agric. Food Chem.* (2004) **52** 1083-1087. DOI: 10.1021/jf0350313
|
---
title: Deficiency of Perry syndrome-associated p150Glued in midbrain dopaminergic
neurons leads to progressive neurodegeneration and endoplasmic reticulum abnormalities
authors:
- Jia Yu
- Xuan Yang
- Jiayin Zheng
- Carmelo Sgobio
- Lixin Sun
- Huaibin Cai
journal: NPJ Parkinson's Disease
year: 2023
pmcid: PMC9988887
doi: 10.1038/s41531-023-00478-0
license: CC BY 4.0
---
# Deficiency of Perry syndrome-associated p150Glued in midbrain dopaminergic neurons leads to progressive neurodegeneration and endoplasmic reticulum abnormalities
## Abstract
Multiple missense mutations in p150Glued are linked to Perry syndrome (PS), a rare neurodegenerative disease pathologically characterized by loss of nigral dopaminergic (DAergic) neurons. Here we generated p150Glued conditional knockout (cKO) mice by deleting p150Glued in midbrain DAergic neurons. The young cKO mice displayed impaired motor coordination, dystrophic DAergic dendrites, swollen axon terminals, reduced striatal dopamine transporter (DAT), and dysregulated dopamine transmission. The aged cKO mice showed loss of DAergic neurons and axons, somatic accumulation of α-synuclein, and astrogliosis. Further mechanistic studies revealed that p150Glued deficiency in DAergic neurons led to the reorganization of endoplasmic reticulum (ER) in dystrophic dendrites, upregulation of ER tubule-shaping protein reticulon 3, accumulation of DAT in reorganized ERs, dysfunction of COPII-mediated ER export, activation of unfolded protein response, and exacerbation of ER stress-induced cell death. Our findings demonstrate the importance of p150Glued in controlling the structure and function of ER, which is critical for the survival and function of midbrain DAergic neurons in PS.
## Introduction
Motor protein dynein and its activator, dynactin, move along the microtubule towards the minus end, playing crucial roles in mitosis and intracellular transport1–4. Genetic mutations in genes encoding components of dynein and dynactin are associated with various neurological diseases5–7. The G59S missense mutation in dynactin subunit p150Glued, encoded by the DCTN1 gene, has been linked to an autosomal dominant motor neuron disease (MND)8. Since then, another 13 missense mutations (F52L, K56R, G67D, K68E, G71A/R/E/V, T72P, Q74P, Y78C/H, and Q93H) in p150Glued have been identified as the genetic cause of Perry syndrome (PS), an autosomal dominant neurodegenerative disease characterized by parkinsonism with mental depression, weight loss, and central hypoventilation9–13. Severe loss of dopaminergic (DAergic) neurons in the substantia nigra (SN) and dysfunction of dopamine (DA) transmission in the striatum were reported in PS patients10–16. However, the pathogenic mechanism underlying the degeneration of midbrain DAergic neurons in PS remains elusive.
P150Glued contains tandem microtubule-binding domains (MTBDs) at its N-terminus: the cytoskeleton-associated protein and glycine-rich (CAP-Gly) domain and the basic domain17–19. The CAP-Gly domain has the binding affinity for microtubules and microtubule end-binding proteins17–19. The DCTN1 gene, composed of 32 exons, not only synthesizes the MTBDs-containing p150Glued but also encodes p135 and other short splicing isoforms, collectively called p135+, which lack the MTBDs18–22. The MTBDs in p150Glued possess neuron-specific functions, such as facilitating the initiation of retrograde transport from the distal axons and enhancing the stability of microtubules along axons23–25. The MND- and PS-related mutations in p150Glued occur within or close to the CAP-Gly domain, disrupt the association of p150Glued with microtubules and microtubule end-binding proteins, and thereby inhibit the initiation of retrograde transport and destabilize microtubules in axons8–13,23–26. These findings indicate a critical role of MTBDs in the pathogenesis of neurodegeneration induced by the mutations of p150Glued.
Knock-in (KI) and transgenic mice expressing MND-related G59S mutant p150Glued, as well as the Dctn1LoxP/LoxP;Thy1-Cre mice which lack the MTBDs-containing p150Glued but express p135+ in the forebrain and spinal neurons, have been generated22,27–29. Studies on these mouse models reveal multiple subcellular abnormalities in spinal motor neurons which might contribute to the neurodegeneration induced by the mutation or lack of MTBDs in p150Glued, including impaired distal axonal integrity, and retrograde axonal transport, aberrant endoplasmic reticulum (ER)-Golgi secretory pathway and autophagosome-lysosome degradative pathway, and augmentations of postsynaptic glutamate receptors and susceptibility to excitotoxicity22,27–29. While KI and transgenic mice expressing PS-related mutant p150Glued have generated and recapitulated some of PS’s clinical and pathological features30–32, the vital intracellular pathways and the underlying molecular mechanisms responsible for the selective degeneration of midbrain DAergic neurons induced by the mutation or lack of MTBDs in p150Glued remain to be determined.
In this study, we investigated the importance of p150Glued and its MTBDs in the survival and function of midbrain DAergic neurons using genetically engineered mouse models. Through crossbreeding Dctn1LoxP/ mice22 and Th-Cre transgenic mice33, we generated Dctn1 LoxP/LoxP;Th-Cre conditional knockout (cKO) mice, which lacked the MTBDs-containing p150Glued but expressed p135+ in midbrain DAergic neurons. We then performed a series of in vivo and in vitro experiments to identify behavioral, neuropathological, electrochemical, subcellular, and molecular abnormalities in the cKO mice. The cKO mice displayed aggravated impairment of motor coordination during aging and progressive degeneration of midbrain DAergic neurons. Furthermore, p150Glued deficiency leads to multiple changes in ER of midbrain DAergic neurons, including the reorganization of ER structure, dysfunction of ER export, activation of unfolded protein response, and increased susceptibility to ER stress-induced cell death, suggesting the abnormal ER structure and function contribute to the DAergic neurodegeneration in PS.
## Generation of cKO mice with deletion of p150Glued in midbrain DAergic neurons
In our previous work, we generated Dctn1LoxP/ mice, which had LoxP sites flanking exon 2–4 of the *Dctn1* gene (Fig. 1a)22. Cre recombinase-mediated deletion of exon 2–4 from the floxed Dctn1 abolishes the expression of the MTBDs-containing p150Glued but keeps the expression of the MTBDs-lacking p135+ in the Cre-expressing cells of Dctn1LoxP/LoxP;Cre mice22. In the current study, to delete p150Glued in midbrain DAergic neurons, we crossbred Dctn1LoxP/ mice with Th-Cre mice33. We obtained Dctn1+/+ [referred to as wild-type (WT)], Dctn1+/+;Th-Cre (referred to as Cre), Dctn1Loxp/LoxP [referred to as control (Ctrl)], and Dctn1LoxP/LoxP;Th-Cre (referred to as cKO) mice. The composition of offspring with different genotypes followed the Mendelian ratio, indicating a normal embryonic development of cKO mice. To examine the selective depletion of p150Glued in cKO mice, we used the antibody specific for the N-terminus of p150Glued (recognizing only p150Glued but not p135+) and the antibody specific for the C-terminus of p150Glued (recognizing both p150Glued and p135+). Western blotting revealed a substantial decrease (approximately $70\%$) of p150Glued but a marked increase (about $64\%$) of p135+ in the midbrain tissues dissected from 1-month-old cKO mice compared to the age-matched WT, Cre, and Ctrl mice (Fig. 1b, c). On the other hand, the levels of all the protein isoforms of the *Dctn1* gene (p150Glued & p135+) and other dynactin subunits (including DCTN4, p50, and ARP1) were not significantly changed in the cKO mice (Fig. 1b). The residual p150Glued protein detected in the cKO samples is likely derived from the non-DAergic neurons and glia in the tissue preparations (Fig. 1b, c). Indeed, immunohistochemistry of midbrain sections from 1-month-old cKO mice demonstrated a complete loss of p150Glued staining in approximately $100\%$ of tyrosine hydrogenase (TH)-positive DAergic neurons, while the p135+ staining remained (Fig. 1d, e). Additionally, the levels of p150Glued, p135+, and other dynactin subunits were comparable in the olfactory bulb, cerebral cortex, hippocampus, striatum, cerebellum, and brainstem of Ctrl and cKO mice (Fig. 1f, g). Therefore, we deleted p150Glued but kept p135+ expression in midbrain DAergic neurons of cKO mice. Apart from the midbrain DAergic neurons, we examined the Cre-mediated deletion of p150Glued in other types of TH-expressing cells in the central nervous system and peripheral tissues of 1-month-old cKO mice. Interestingly, deletion of p150Glued was observed in approximately $46\%$, $2\%$, $99\%$, and $98\%$ of TH+ cells in the locus coeruleus, olfactory bulb, superior cervical ganglion, and adrenal medulla, respectively, indicating the differential efficiencies of Cre-mediated recombination in various populations of TH-expressing cells in cKO mice (Supplementary Fig. 1).Fig. 1Genetic deletion of p150Glued in midbrain DAergic neurons of cKO mice.a The schematic diagram depicts the full-length p150Glued and the floxed Dctn1 allele of mice. The N-terminal CAP-Gly domain and the adjacent basic domain (encoded by the exon 1–3 and 3–8 of the *Dctn1* gene, respectively) form the tandem MTBDs of p150Glued. The floxed Dctn1 allele has two LoxP sites inserted at the intron 1 and 4 of the *Dctn1* gene locus. b, c Western blots show the expression of p150Glued, p135+, DCTN4, p50, ARP1, and TH in the midbrain of 1-month-old WT (Dctn1+/+), Cre (Dctn1+/+;Th-Cre), Ctrl (Dctn1LoxP/LoxP), and cKO (Dctn1LoxP/LoxP;Th-Cre) mice. The protein level was quantified as mean ± SEM ($$n = 3$$ animals per genotype). One-way ANOVA, ***$$p \leq 0.0002$$, ****$p \leq 0.0001.$ d, e Immunofluorescent images show the staining of p150Glued (green), p150Glued & p135+ (red), and TH (blue) in the midbrain of 1-month-old Ctrl and cKO mice. Scale bar: 250 μm. The percentages of p150Glued-positive and p150Glued-negative TH+ neurons were quantified as mean ± SEM ($$n = 6$$ animals per genotype and 6 sections per animal). Unpaired t test, ****$p \leq 0.0001.$ f, g Western blots show the expression of p150Glued, p135+, DCTN4, p50, ARP1, and TH in the olfactory bulb (OB), cortex (CX), hippocampus (HP), striatum (ST), midbrain (MB), cerebellum (CB), and brain-stem (BS) of 1-month-old Ctrl and cKO mice. The protein level was quantified as mean ± SEM ($$n = 3$$ animals per genotype). Unpaired t test, **$$p \leq 0.0019.$$
In a parallel study, we crossbred Dctn1Loxp/ mice with the tamoxifen-inducible Cre/Esr1 transgenic mice34 to generate Dctn1+/+ (WT), Dctn1+/+;Cre/Esr1 (Cre), Dctn1Loxp/LoxP (Ctrl), and Dctn1LoxP/LoxP;Cre/Esr1 [inducible knockout (iKO)] mice. We isolated the midbrain tissues from neonatal iKO pups and littermate controls for primary neuronal cultures. At 1 day in vitro (DIV), we added 1 μM 4-OHT to the medium of primary cultures to induce Cre recombinase activity within cells. At 14 DIV, western blotting and immunocytochemistry confirmed the inducible deletion (~$97\%$) of p150Glued and the compensatory increase (around $92\%$) of p135+ in the iKO midbrain cell culture, including the DAergic neurons (Supplementary Fig. 2). The Dctn1Loxp/LoxP;Cre/Esr1 iKO mice allow for in-depth cell biology and biochemical studies on the functional significance of p150Glued and its MTBDs.
## P150Glued cKO mice exhibit aggravated deterioration of motor coordination during aging
P150Glued cKO mice developed normally with no gross physical or behavioral abnormalities. Although weight loss is a typical clinical manifestation of PS patients9–13, the cKO mice weighed similarly to the Ctrl mice at 1, 3, 6, 12, and 18 months of age (Fig. 2a). In the open-field test, the cKO mice displayed similar locomotor activities (ambulatory, rearing, and fine movement) and time spent in the center of the arena with the Ctrl mice at 1, 3, 6, 12, and 18 months of age (Fig. 2b–e). In the rotarod test, the cKO mice performed as well as the Ctrl mice at 1 and 3 months, but exhibited a more profound deterioration of rotarod performance than the Ctrl mice starting at 6 months of age (Fig. 2f). Considering the crucial role of dopamine in motor control, the accelerated impairment of motor coordination observed in the cKO mice could result from progressive dysfunction/degeneration of midbrain DAergic neurons during aging. Fig. 2Aggravated deterioration of motor coordination in cKO mice during aging. Cohorts of male Ctrl and cKO mice ($$n = 16$$ animals per genotype) were repeatedly assessed for body weight and behavioral performance at 1, 3, 6, 12, and 18 months of age. a The body weight of Ctrl and cKO mice. b–e The ambulatory movement (b), rearing movement (c), fine movement (d), and center time (e) of Ctrl and cKO mice in the open-field test. f The latency to fall of Ctrl and cKO mice in the rotarod test. Data were presented as mean ± SEM. Two-way ANOVA with Sidak’s multiple comparisons test was used for statistical analysis. In the rotarod test (f), *$$p \leq 0.0104$$ (6 M), **$$p \leq 0.0027$$ (12 M), ***$$p \leq 0.0012$$ (18 M).
## P150Glued cKO mice display progressive degeneration of DAergic neurons in the midbrain
Since the loss of midbrain DAergic neurons is the cardinal neuropathological feature of PS10–13, we used unbiased stereology to count the TH-positive neurons in the substantia nigra pars compacta (SNc) and the ventral tegmental area (VTA) of Ctrl and cKO mice. Compared with the age-matched controls, the cKO mice had similar numbers of midbrain DAergic neurons at 6 and 12 months of age, but significantly fewer DAergic neurons at 24 months of age, with approximately 23 and $26\%$ neuronal loss in the SNc and the VTA, respectively (Fig. 3a, b). Therefore, genetic deletion of p150Glued in midbrain DAergic neurons contribute to late-onset loss of DAergic neurons. Fig. 3Progressive degeneration of DAergic neurons in the midbrain of cKO mice.a, b Immunohistochemical images show TH staining in the midbrain of 24-month-old Ctrl and cKO mice. Scale bar: 400 μm. Unbiased stereological estimation of the number of TH-positive DAergic neurons in the SNc and VTA of 6-, 12-, and 24-month-old Ctrl and cKO mice ($$n = 4$$ animals per genotype per time point). Data were presented as mean ± SEM. Two-way ANOVA, ****$p \leq 0.0001.$ c, d Representative images show TH staining in the midbrain of 6-month-old Ctrl and cKO mice. Arrows point to the dystrophic DAergic dendrites in the SNr of cKO mice. Dystrophic DAergic dendrites were defined as TH-positive neuritic varicosity ≥25 μm2. Scale bar: 200 μm (low magnification), 100 μm (high magnification). The bar graph quantifies the density of dystrophic DAergic dendrites in the SNr of 6-, 12-, and 24-month-old Ctrl and cKO mice (at each time point, $$n = 4$$ animals per genotype and 5 sections per animal). Data were presented as mean ± SEM. Two-way ANOVA, ****$p \leq 0.0001.$ e, f Immunofluorescent images show the staining of GFAP (green) and TH (red) in the midbrain of 24-month-old Ctrl and cKO mice. Scale bar: 20 μm. The area fraction of GFAP-positive astrocytes was quantified as mean ± SEM ($$n = 4$$ animals per genotype and 5 sections per animal). Unpaired t test, *$$p \leq 0.0178$$, **$$p \leq 0.0016.$$
The dendrites of DAergic neurons located in the ventral SNc often protrude perpendicularly into the underneath substantia nigra par reticulata (SNr) region and form synaptic connections with the incoming axon fibers35. Interestingly, compared with the age-matched controls, the cKO mice displayed substantial and progressive dystrophy of DAergic dendrites (≥25 μm2) in the SNr starting at 6 months of age (Fig. 3c, d). Thus, genetic deletion of p150Glued in midbrain DAergic neurons leads to early-onset dystrophy of DAergic dendrites.
In addition to the cell loss and dendritic dystrophy of DAergic neurons, increased astrogliosis was detected in the SNc and the SNr of 24-month-old cKO mice (Fig. 3e, f). As the neuronal loss, the sphere-like neurite swelling, and the gliosis were also observed in the SN of PS patients9,36,37, the cKO mice with deletion of p150Glued in the midbrain DAergic neurons recapitulate some key pathological features of PS.
## Abnormal accumulation of α-synuclein but not TDP-43 in the midbrain DAergic neurons of aged p150Glued cKO mice
As abnormal TAR DNA-binding protein 43 (TDP-43)-positive cytoplasmic inclusions were often identified in the SN and other brain regions of PS patients9,36,37, we examined the subcellular location of TDP-43 in the midbrain DAergic neurons of 24-month-old Ctrl and cKO mice. However, we did not detect any apparent accumulation of TDP-43 in the cytosol of either Ctrl or cKO DAergic neurons (Fig. 4a, b). By contrast, although very sparse Lewy bodies composed of α-synuclein aggregates were identified in the brain of PS patients10, we observed a substantial increase (approximately $140\%$) of cytoplasmic and nuclear accumulation of α-synuclein in the midbrain DAergic neurons of 24-month-old cKO mice compared with the Ctrl mice (Fig. 4c, d). While α-synuclein is typically enriched in the axon terminals, the abnormal accumulation of α-synuclein in cytosol and nucleus is implicated in neurodegeneration38,39. It has been reported that α-synuclein in synucleinopathy lesions is extensively phosphorylated at Ser12940. In line with this finding, immunostaining revealed a significant increase (~$220\%$) of p-α-synuclein (Ser129) immunoreactivity in the soma of midbrain DAergic neurons of 24-month-old cKO mice compared with the Ctrl mice (Fig. 4e, f). Emerging evidence suggests that α-synuclein pathology can spread intercellularly and between interconnected brain areas41. However, we detected no apparent accumulation of α-synuclein in the soma of TH-negative neurons in the SNc, SNr, dorsal striatum, hippocampal CA1 area, and prefrontal cortex of 24-month-old cKO mice (Supplementary Fig. 3). This indicates that α-synuclein pathology in the midbrain DAergic neurons of aged cKO mice rarely spreads to other neurons in the neighboring and interconnected brain areas. Fig. 4Abnormal accumulation of α-synuclein but not TDP-43 in the midbrain DAergic neurons of cKO mice.a, b Immunofluorescent images show the staining of TDP-43 (green) and TH (red) in the midbrain of 24-month-old Ctrl and cKO mice. Scale bar: 10 μm. The cytoplasmic/nuclear ratio of TDP-43 in the soma of DAergic neurons was quantified as mean ± SEM ($$n = 4$$ animals per genotype and ≥30 neurons per animal). Unpaired t test, ns ($$p \leq 0.8547$$). c, d Immunofluorescent images show the staining of α-synuclein (green) and TH (red) in the midbrain of 24-month-old Ctrl and cKO mice. Note the increase of cytoplasmic and nuclear α-synuclein in the midbrain DAergic neurons of cKO mice. Scale bar: 10 μm. The staining intensity of α-synuclein in the soma of DAergic neurons was quantified as mean ± SEM ($$n = 4$$ animals per genotype and ≥30 neurons per animal). Unpaired t test, **$$p \leq 0.0016.$$ e, f Immunofluorescent images show the staining of p-α-synuclein (Ser129) (green) and TH (red) in the midbrain of 24-month-old Ctrl and cKO mice. Note the increase of somatic p-α-synuclein (Ser129) in the midbrain DAergic neurons of cKO mice. Scale bar: 10 μm. The staining intensity of p-α-synuclein in the soma of DAergic neurons was quantified as mean ± SEM ($$n = 4$$ animals per genotype and ≥30 neurons per animal). Unpaired t test, ****$p \leq 0.0001.$
## P150Glued cKO mice show progressive degeneration of DAergic axon terminals in the striatum
As axon dystrophy was reported in the brains of PS patients9,36,37, we examined the density and morphology of DAergic axon terminals in the dorsal striatum of Ctrl and cKO mice. Compared with the age-matched controls, the cKO mice had similar densities of striatal DAergic axon terminals at 6 and 12 months of age, but significantly sparser striatal DAergic axon terminals at 24 months of age, with approximately $32\%$ of axonal loss (Fig. 5a, b). This finding agrees with the late-onset loss of midbrain DAergic neurons in the cKO mice (Fig. 3a, b). TH staining also revealed gradually increasing numbers of abnormally swollen DAergic axon terminals (≥3 μm2) in the dorsal striatum of 6-, 12-, and 24-month-old cKO mice compared with the age-matched controls (Fig. 5a, c). In addition, the swollen axon terminals in cKO mice were packed with VMAT2- and synaptophysin-positive vesicles (Fig. 5d, e). These data show that genetic deletion of p150Glued in midbrain DAergic neurons induces early-onset swelling and late-onset loss of DAergic axon terminals in the striatum. Fig. 5Progressive degeneration of DAergic axon terminals in the striatum of cKO mice.a–c Representative images show TH staining in the dorsal striatum of 24-month-old Ctrl and cKO mice. The arrow points to the swollen DAergic axon terminal (defined as TH-positive neuritic varicosity ≥3 μm2) of cKO mice. Scale bar: 10 μm. The area fraction of DAergic axon terminals (b) and density of swollen DAergic axonal terminals (c) in the dorsal striatum of 6-, 12-, and 24-month-old Ctrl and cKO mice were quantified as mean ± SEM (at each time point, n ≥ 4 animals per genotype and ≥5 sections per animal). Two-way ANOVA, **$$p \leq 0.0015$$, ****$p \leq 0.0001.$ d, e Immunofluorescent images show the co-staining of VMAT2 (d, green) or synaptophysin (e, green) with TH (red) in the dorsal striatum of 6-month-old Ctrl and cKO mice. Note the accumulation of VMAT2-positive (d) and synaptophysin-positive (e) vesicles within the swollen DAergic axon terminal. Scale bar: 2 μm. f, g Western blots show the expression of DAT and TH in the striatum of 6-, 12-, and 24-month-old Ctrl and cKO mice. The protein level was quantified as mean ± SEM (at each time point, $$n = 4$$ animals per genotype). Two-way ANOVA, ****$p \leq 0.0001.$ h, i Immunofluorescent images show the staining of DAT (green) and TH (red) in the dorsal striatum of 6-month-old Ctrl and cKO mice. Scale bar: 2 μm. The area fraction of DAT-positive puncta was quantified as mean ± SEM ($$n = 4$$ animals per genotype and 5 sections per animal). Unpaired t test, **$$p \leq 0.0060.$$ j HPLC measures the content of DA, DOPAC, and 5-HT in the dorsal striatum of 12-month-old Ctrl ($$n = 3$$) and cKO ($$n = 4$$) mice. Data were presented as mean ± SEM. Unpaired t test, ns (p ≥ 0.05). k–m FSCV quantifies the kinetics of evoked DA release in the dorsal striatum of 12-month-old Ctrl and cKO mice ($$n = 3$$ animals per genotype and 3 sections per animal). Data were presented as mean ± SEM. For the peak evoked DA release following single-pulse electrical stimulation of different stimulus intensities (k), Two-way ANOVA genotype factor: F[1, 16] = 25.11, ***$$p \leq 0.0001.$$ For the peak evoked DA release in response to burst electrical stimulation (50 Hz, 5 pulses) of different stimulus intensities (l), Two-way ANOVA genotype factor: F[1, 16] = 23.09, ***$$p \leq 0.0002.$$ For the time constant of slope decay (τ) following single-pulse electrical stimulation of different stimulus intensities (m), Two-way ANOVA genotype factor: F[1, 16] = 6.035, *$$p \leq 0.0258.$$
As neuroimaging has shown marked loss of dopamine transporter (DAT) binding in the striatum of PS patients14–16, we examined the protein level of DAT, which is crucial for dopamine uptake, in the striatal homogenates of 6-, 12-, and 24-month-old Ctrl and cKO mice. Western blotting revealed a substantial decrease of DAT in the striatum of cKO mice compared with age-matched controls, with ~$26\%$, $27\%$, and $39\%$ of DAT loss at 6, 12, and 24 months of age, respectively (Fig. 5f, g). Immunostaining further confirmed a marked reduction (around $32\%$) of DAT-positive puncta in the dorsal striatum of 6-month-old cKO mice compared with Ctrl mice (Fig. 5h, i). These results demonstrate that genetic deletion of p150Glued in midbrain DAergic neurons contribute to the early-onset reduction of DAT in the striatum.
In parallel, we examined the protein level of TH, the rate-limiting enzyme in DA synthesis, in the striatal homogenates of 6-, 12-, and 24-month-old Ctrl and cKO mice. Compared with the age-matched controls, the cKO mice showed no marked change of TH at 6 and 12 months of age but a substantial reduction (about $28\%$) of TH at 24 months of age (Fig. 5f, g), consistent with the late-onset loss of striatal DAergic axon terminals in cKO mice (Fig. 5a, b). In correlation with no apparent loss of DAergic axon terminals or TH expression in the cKO mice at 12 months of age (Fig. 5b, f, g), HPLC revealed similar content of DA, dopamine metabolite 3,4-dihydroxyphenylacetic acid (DOPAC), or 5-hydroxytryptamine (5-HT) in the dorsal striatum of 12-month-old cKO mice compared with controls (Fig. 5j).
Considering the early-onset morphological and biochemical changes of DAergic axon terminals in the cKO mice (Fig. 5c–i), we further evaluated the kinetics of DA release and uptake in the striatal slices prepared from 12-month-old Ctrl and cKO mice using fast-scan cyclic voltammetry (FSCV). In response to either single- or burst-pulse stimulation, the peak evoked DA release was substantially higher from the DAergic axon terminals of the cKO mice than the controls (Fig. 5k, l). Additionally, the time constant of slope decay (τ), which measures the kinetics of dopamine uptake, was also significantly increased in the DAergic axon terminals of the cKO mice compared with the controls (Fig. 5m). Taken together, these data show that genetic deletion of p150Glued in midbrain DAergic neurons leads to early-onset swelling of DAergic axon terminals accumulated with DA-containing vesicles, reduction of DAT, and the resultant dysregulation of DA transmission in the dorsal striatum, which may contribute to the early-onset impairment of motor coordination in the cKO mice.
## P150Glued deficiency induces reorganization of ER structure and increases ER tubule-shaping protein reticulon 3 (RTN3)
To investigate the composition of dystrophic DAergic dendrites in the SNr of cKO mice, we co-stained the midbrain sections from 6-month-old Ctrl and cKO mice with antibodies against TH and various intracellular organelle markers. The occupancy of ER (visualized by BiP, calnexin, or PDI staining), Golgi (visualized by RCAS1 staining), endosomes (visualized by EEA1 staining), autophagosomes (visualized by SQST1 staining), and lysosomes (visualized by cathepsin D staining) in the dystrophic DAergic dendrites was ~$80\%$, $0\%$, $7\%$, $11\%$, and $12\%$, respectively (Fig. 6a and Supplementary Fig. 4a–g). These findings indicate that the reorganized ER is the major organelle within the dystrophic DAergic dendrite of cKO mice. Three-dimensional (3D) reconstruction of the ER and DAergic dendrites in the SNr of 6-month-old Ctrl and cKO mice further demonstrated that the reorganized ER occupied a large portion of the intracellular space of the dystrophic DAergic dendrite in cKO mice (Supplementary Fig. 4h and Supplementary Movie 1 and 2). Moreover, the number of dystrophic DAergic dendrites with ER reorganization in the SNr of cKO mice increased in an age-dependent manner, ~8.42, 12.91, and 18.25 per mm2 at 6, 12, and 24 months of age, respectively (Fig. 6a, b).Fig. 6P150Glued deficiency induces reorganization of ER structure and increases ER tubule-shaping protein RTN3.a, b Immunofluorescent images show the staining of BiP (the ER marker, green) and TH (red) in the SNr of 6-month-old Ctrl and cKO mice. Arrows point to the abnormal reorganization of ER within the dystrophic DAergic dendrite of cKO mice. Scale bar: 10 μm. The Bar graph quantifies the density of dystrophic DAergic dendrites with ER reorganization in the SNr of 6-, 12-, and 24-month-old Ctrl and cKO mice (at each time point, $$n = 6$$ animals per genotype and 4 sections per animal). Data were presented as mean ± SEM. Two-way ANOVA, **$$p \leq 0.0020$$, ****$p \leq 0.0001.$ c, d Immunofluorescent images show the co-staining of ER tubule-shaping protein RTN3 (c, green) or ER sheet-shaping protein CLIMP63 (d, green) with TH (red) in the SNr of 6-month-old Ctrl and cKO mice. Note the positive immunoreactivity of RTN3 (c) and the negative immunoreactivity of CLIMP63 (d) in the dystrophic DAergic dendrite of cKO mice. Scale bar: 10 μm. e, f Western blots, show the expression of ER-resident proteins (calnexin, PDI, BiP, VAPB, and ERp72) and ER-shaping proteins (CLIMP63, ATL1, ATL2, ATL3, RTN1, RTN3, and RTN4) in the Ctrl and iKO fibroblasts at 14 DIV. The protein level was quantified as mean ± SEM ($$n = 4$$ independent cultures). Unpaired t test, ***$$p \leq 0.0001.$$ g–i Immunofluorescent images show the staining of BiP (green), p150Glued (red), α-tubulin (magenta), and DAPI (blue) in the Ctrl and iKO fibroblasts at 14 DIV. Note the obvious expansion of perinuclear ER and the abnormal reorganization of peripheral ER in the iKO fibroblast. Scale bar: low magnification, 20 μm; high magnification, 5 μm. The percentage of cells with ER expansion (h) and the area fraction of ER in the peripheral regions were quantified as mean ± SEM ($$n = 3$$ independent cultures and ≥60 cells per genotype). Unpaired t test, **$$p \leq 0.0028$$, ****$p \leq 0.0001.$
The ER network consisting of interconnected tubules and sheets extends throughout the neuron and adopts heterogeneous architecture in soma, axon, and dendrites42,43. Several classes of proteins are enriched in ER tubules or sheets and participate in the shaping of ER membranes44–46. We detected positive immunoreactivity of ER tubule-shaping protein RTN3 in the dystrophic DAergic dendrites of 6-month-old cKO mice (Fig. 6c), but negative immunoreactivity of ER sheet-shaping protein CLIMP63 (Fig. 6d). These observations suggest the participation of RTN3 in the ER reorganization within the dystrophic DAergic dendrites of cKO mice. Western blotting of Ctrl and iKO fibroblasts revealed that p150Glued deficiency selectively increased (approximately $43\%$) the protein level of RTN3 without changing the expression of ER-resident proteins (including calnexin, PDI, BiP, VAPB, and ERp72) and other ER-shaping proteins (including CLIMP63, ATL1, ATL2, ATL3, RTN1, RTN4) (Fig. 6e, f). Using immunocytochemistry, we further examined the subcellular distribution pattern and fine structure of ER in the Ctrl and iKO fibroblasts. BiP, α-tubulin, and p150Glued co-staining revealed well-organized ER architecture in the Ctrl fibroblasts, with cohesive ER sheets concentrating in the perinuclear regions and three-way junction of ER tubules distributing in the peripheral areas (Fig. 6g–i). By contrast, over $89\%$ iKO fibroblasts exhibited expansion of perinuclear ER (Fig. 6g–i). Compared with the Ctrl fibroblasts, the iKO fibroblasts exhibited a significantly increased (~$82\%$) area fraction of ER in the peripheral regions, with evident clustering of ER tubules and the existence of ER sheets (Fig. 6g–i). Together, these findings demonstrate that p150Glued deficiency induces ER reorganization and selectively increases ER tubule-shaping protein RTN3 level. Interestingly, RTN3-immunoreactive dystrophic neurites have been found in Alzheimer’s disease brains, and DCTN6 deficiency enhanced RTN3 protein level and the formation of RTN3-immunoreactive dystrophic neurites in the hippocampus of aging mice47,48.
## P150Glued deficiency induces defects in the early secretory pathway and impairs the COPII-mediated ER export
The early secretory pathway, consisting of ER, ER-Golgi intermediate compartment (ERGIC), and Golgi apparatus, ensures the proper supply of secretory and membrane materials for neurons, playing an essential role in neuronal survival and function49,50. Immunohistochemistry revealed a severe accumulation of DAT inside the reorganized ER within the dystrophic DAergic dendrite of 6-month-old cKO mice, indicating the potential dysfunction of ER export (Fig. 7a, b). The findings that DAT protein was trapped in the reorganized ER of p150Glued-deficient DAergic neurons may explain the reduction of DAT protein in the axon terminals of cKO mice (Fig. 5f–i). Furthermore, we found a marked reduction (approximately $51\%$) of the area fraction of ERGIC (visualized by ERGIC53 staining) in the soma of midbrain DAergic neurons from 6-month-old cKO mice compared with the controls (Fig. 7c, d). In addition, the area fraction of cis-Golgi (visualized by GM130 staining) was also significantly decreased (~$39\%$) in the soma of p150Glued-deficient DAergic neurons, while Golgi fragmentation was apparent in individual p150Glued-deficient DAergic neurons of the cKO mice at 6 months of age (Fig. 7e, f). These data demonstrate that p150Glued deficiency induces defects in the early secretory pathway of midbrain DAergic neurons. Fig. 7P150Glued deficiency induces defects in the early secretory pathway and impairs the COPII-mediated ER export.a, b Immunofluorescent images show the staining of DAT (green), BiP (red), TH (magenta), and DAPI (blue) in the SNr of 6-month-old Ctrl and cKO mice. The arrow points to the accumulation of DAT inside the reorganized ER within the dystrophic DAergic dendrite of cKO mice. Scale bar: 10 μm. The density of reorganized ERs with DAT accumulation was quantified as mean ± SEM ($$n = 4$$ animals per genotype and 4 sections per animal). Unpaired t test, **$$p \leq 0.0016.$$ c–f Immunofluorescent images show the co-staining of ERGIC53 (c, the ERGIC marker, green) or GM130 (e, the cis-Golgi marker, green) with TH (red) in the midbrain of 6-month-old Ctrl and cKO mice. Scale bar: 10 μm. The area fractions of ERGIC (d) and cis-Golgi (f) in the soma of DAergic neurons were quantified as mean ± SEM ($$n = 4$$ animals per genotype and ≥30 neurons per animal). Unpaired t test, **$$p \leq 0.0039$$ (d), **$$p \leq 0.0079$$ (f). g Co-IP reveals interaction between p150Glued and COPII component Sec31 in the mouse brains’ total lysate and ER microsomes. h, i Western blots show the expression of COPII components (Sec13 and Sec31) in the midbrains’ total lysate and ER microsomes from 4-month-old Ctrl and iKO mice. The Ctrl and iKO mice were intraperitoneally injected with tamoxifen at 3 months of age to induce Cre recombinase activity and the resultant deletion of p150Glued. Calnexin and VAPB were used as markers of ER microsomes. The protein level in the ER microsomes was quantified as mean ± SEM ($$n = 4$$ independent experiments). Unpaired t test, *$$p \leq 0.0187$$, **$$p \leq 0.0090.$$ j, k Western blots show the expression of COPII components (Sec13, Sec23, and Sec31) in the total lysate and ER microsomes of Ctrl and iKO fibroblasts at 14 DIV. The protein level in the ER microsomes was quantified as mean ± SEM ($$n = 6$$ independent cultures). Unpaired t test, **$$p \leq 0.0032$$ (Sec13), **$$p \leq 0.0097$$ (Sec23), ***$$p \leq 0.0005.$$ l, m Western blots show the expression of mature, immature, and unglycosylated nicastrin in the Ctrl and iKO fibroblasts at 14 DIV. The protein level and mature/immature ratio of nicastrin were quantified as mean ± SEM ($$n = 8$$ independent cultures). Unpaired t test, ***$$p \leq 0.0002$$, ****$p \leq 0.0001.$
Previous studies demonstrate that p150Glued stabilizes the coat protein complex II (COPII) at ER exit site (ERES) through interaction with COPII component Sec2351,52. Consistent with the early findings, co-immunoprecipitation (co-IP) revealed endogenous protein-protein interaction between p150Glued and COPII component Sec31 in mouse brain’s total lysate and ER microsomes (Fig. 7g). Moreover, we found that COPII components Sec13 and Sec31 substantially decreased in the midbrain’s ER microsomes of 4-month-old iKO mice compared with the Ctrl mice (Fig. 7h, i). The reduction of Sec13, Sec23, and Sec31 was also observed in the ER microsomes isolated from the cultured iKO fibroblasts compared to the Ctrl cells (Fig. 7j, k). The concentration of COPII at ERES is critical for ER export of secretory and membrane proteins, such as type-I integral membrane protein nicastrin, which is partially glycosylated in ER as the immature form and then exported to *Golgi apparatus* for fully glycosylation and maturation53. The ratio of mature versus immature nicastrin can be used to estimate the efficiency of ER export in cells53,54. Accordingly, compared with Ctrl cells, the iKO fibroblasts displayed a substantial decrease in the level of mature nicastrin and the ratio of mature/immature nicastrin (Fig. 7l, m), implying dysfunction of ER export in the p150Glued-deficient cells. Therefore, our studies further support that p150Glued deficiency destabilizes the COPII complex at ERES and compromises the COPII-mediated ER export.
## P150Glued deficiency activates unfolded protein response and exacerbates ER stress-induced DAergic neuron death
Since the disruption of ER structural or functional homeostasis triggers ER stress and unfolded protein response (UPR)55,56, we investigated the effects of p150Glued deficiency on a series of proteins implicated in the UPR pathway. Using immunohistochemistry, we detected a marked increase (approximately $96\%$) of phosphorylated eIF2α (Ser51) [p-eIF2α (Ser51)] in the midbrain DAergic neurons of 18-month-old cKO mice compared with Ctrl mice (Fig. 8a, b). Additionally, a significant upregulation (approximately $108\%$) of phosphorylated IRE1α (Ser724) [p-IRE1α (Ser724)] was observed in the midbrain DAergic neurons of 18-month-old cKO mice compared with Ctrl mice (Fig. 8c, d). By western blotting, we found a substantial increase of phosphorylated PERK (Thr982) [p-PERK (Thr982)], p-eIF2α (Ser51), ATF4, and CHOP in the iKO fibroblasts compared to the Ctrl cells (Fig. 8e, f), indicating that p150Glued deficiency leads to the activation of the PERK-eIF2α-ATF4-CHOP branch of UPR pathway. We also found a significant upregulation of IRE1α, p-IRE1α (Ser724), spliced XBP1, and phosphorylated SPK/JNK (Thr183/Tyr185) [p-SPK/JNK (Thr183/Tyr185)], as well as a marked decrease of unspliced XBP1, in the iKO fibroblasts (Fig. 8e, f), demonstrating the activation of the IRE1-XBP1-SAPK/JNK branch of UPR pathway due to p150Glued deficiency. Additionally, the increased expression of ATF6 revealed the activation of the ATF6 branch of UPR pathway in the iKO fibroblasts (Fig. 8e, f). Moreover, the elevated level of cleaved caspase-3 indicates the activation of the cell death pathway in the iKO fibroblasts (Fig. 8e, f).Fig. 8P150Glued deficiency activates unfolded protein response and exacerbates ER stress-induced cell death of DAergic neurons.a–d Immunofluorescent images show the co-staining of p-eIF2α (Ser51) (a, green) or p-IRE1α (Ser724) (c, green) with TH (red) in the midbrain of 18-month-old Ctrl and cKO mice. Scale bar: 10 μm. The staining intensity of p-eIF2α (b) or p-IRE1α (d) in the soma of DAergic neurons was quantified as mean ± SEM ($$n = 5$$ animals per genotype and ≥20 neurons per animal). Unpaired t test, **$$p \leq 0.0016$$ (b), **$$p \leq 0.0011$$ (d). e, f Western blots show the expression of UPR (PERK-eIF2α-ATF4-CHOP, IRE1-XBP1-SAPK/JNK, and ATF6) and cell death (cleaved caspase-3) regulatory proteins in the Ctrl and iKO fibroblasts at 14 DIV. The protein level was quantified as mean ± SEM ($$n = 4$$ independent cultures). Unpaired t test, **$$p \leq 0.0017$$ [p-PERK (Thr982)], *$$p \leq 0.0159$$ [p-eIF2α (Ser51)], **$$p \leq 0.0028$$ (ATF4), *$$p \leq 0.0405$$ (CHOP), **$$p \leq 0.0096$$ (IRE1α), ***$$p \leq 0.0002$$ [p-IRE1α (Ser724)], *$$p \leq 0.0237$$ (unspliced XBP1), *$$p \leq 0.0451$$ (spliced XBP1), *$$p \leq 0.0190$$ [p-SAPK/JNK (Thr183/Tyr185)], ***$$p \leq 0.0003$$ (ATF6), *$$p \leq 0.0317$$ (cleaved caspase-3). g Bar graphs show the survival rates of Ctrl and iKO midbrain DAergic neurons (14 DIV) treated with 0 or 10 nM thapsigargin for 48 and 96 h (at each time point, $$n = 12$$ coverslips per genotype per condition). 10 nM thapsigargin was used to induce ER stress. Data were presented as mean ± SEM. One-way ANOVA, ****$p \leq 0.0001.$ h The bar graph shows the survival rates of Ctrl and iKO midbrain DAergic neurons (14 DIV) cultured in the presence of 1× or 0.1× N2/B27 supplement for 48 h ($$n = 14$$ coverslips per genotype per condition). The content of N2/B27 supplement in the culture medium was lowed from 1× to 0.1× to induce growth factor deprivation. Data were presented as mean ± SEM. One-way ANOVA, ***$$p \leq 0.0008$$, ****$p \leq 0.0001.$
Despite the beneficial role of UPR in sensing and mitigating ER stress, the sustained UPR activation under irremediable ER stress can result in cell death, contributing to the pathogenesis of a wide range of human diseases, including neurodegeneration55,56. To determine whether p150Glued deficiency affects the susceptibility of midbrain DAergic neurons to ER stress-induced cell death, we treated Ctrl and iKO midbrain neuron cultures (14 DIV) with 0 or 10 nM thapsigargin, an inhibitor of ER Ca2+-ATPase. After treating with 10 nM thapsigargin for 48 h, the survival rates of Ctrl and iKO midbrain DAergic neurons were ~74 and $48\%$, respectively (Fig. 8g). After treating with 10 nM thapsigargin for 96 h, the survival rates of Ctrl and iKO midbrain DAergic neurons further fell to 44 and $23\%$, respectively (Fig. 8g). At both time points, thapsigargin-induced cell death was exacerbated significantly in the iKO midbrain DAergic neurons, as compared with the Ctrl midbrain DAergic neurons (Fig. 8g). Similarly, the iKO midbrain DAergic neurons were more susceptible to cell death induced by growth factor deprivation (Fig. 8h), another type of cellular stress previously shown to induce apoptosis57. To investigate whether inhibition of UPR signaling has protective activity against ER stress-induced cell death of midbrain DAergic neurons, we applied GSK2606414 (an inhibitor of PERK) and KIRA8 (an inhibitor of IRE1α). Co-application of 100 nM GSK2606414 or 100 nM KIRA8 with 10 nM thapsigargin for 48 h significantly mitigated thapsigargin-induced cell death of both the Ctrl and iKO midbrain DAergic neurons (Supplementary Fig. 5). Collectively, our data demonstrate that p150Glued deficiency activates unfolded protein response, making the DAergic neurons more susceptible to ER stress-induced cell death.
## Discussion
In the present study, we generated Dctn1LoxP/LoxP;Th-Cre cKO mice by abolishing p150Glued expression but keeping p135+ expression in midbrain DAergice neurons. We found the accelerated deterioration of motor coordination and progressive degeneration of nigrostriatal DAergic pathway in the cKO mice. Compared with the age-matched controls, the cKO mice exhibited late-onset loss of DAergic neurons and axon terminals, as well as somatic accumulation of α-synuclein. Prior to the neuronal and axonal loss, the cKO mice exhibited early-onset dystrophy of DAergic dendrites, swelling of DAergic axon terminals with the accumulation of VMAT2-and synaptophysin-positive vesicles, reduction of striatal DAT, and the resultant dysregulation of striatal DA transmission. At the subcellular and molecular level, we revealed that p150Glued deficiency induced ER reorganization within the dystrophic DAergic dendrites and upregulated ER tubule-shaping protein RTN3. Meanwhile, p150Glued deficiency led to multiple defects in the early secretory pathway of DAergic neurons, including DAT accumulation inside the reorganized ER and diminished area fractions of ERGIC and cis-Golgi. Moreover, we demonstrated that p150Glued interacted with COPII component Sec31, and p150Glued deficiency impaired the ER export mediated by COPII. Finally, we provided evidence that p150Glued deficiency activated UPR in DAergic neurons and made DAergic neurons more susceptible to ER stress-induced cell death, suggesting that ER structural and functional abnormalities may contribute to the degeneration of midbrain DAergic neurons in PS.
The identification of p150Glued mutations as the genetic cause of PS promotes the studies on the roles of p150Glued in the survival and function of midbrain DAergic neurons9–13. Recently, G71A p150Glued KI mice have been generated31. Homozygous G71A p150Glued KI mice show the same early embryonic lethality as the homozygous p150Glued KO mice27,31, indicating that the PS-linked missense mutations compromise the function of p150Glued. Heterozygous G71A p150Glued KI mice are viable and fertile, displaying behavioral defects in tail-suspension and beam-walking tests and a decrease in the TH immunoreactivity of DAergic neurons31. Additionally, using a tetracycline-controlled transcriptional regulation system, we generated transgenic mice with selective overexpression of G71R p150Glued in the midbrain DAergic neurons32. The G71R p150Glued transgenic mice exhibited early-onset dysregulation of striatal DA transmission and the resultant motor abnormalities, as well as late-onset loss of DAergic neurons and axon terminals32. In line with these earlier findings, our current study further highlighted the essential role of p150Glued and its MTBDs in the survival and function of midbrain DAergic neurons. We observed early-onset impairment of motor coordination, dystrophy of DAergic dendrites, swelling of DAergic axon terminals, and dysregulated DA transmission in the cKO mice. We also found late-onset loss of DAergic neurons and axons in the cKO mice. These observations demonstrate that the loss-of-function of MTBDs in p150Glued is sufficient to lead to dysfunction and degeneration of midbrain DAergic neurons.
The MTBDs in p150Glued are required to initiate retrograde axonal transport and to maintain axon terminals’ integrity23–25. Accordingly, axonal pathologies of spinal motor neurons, such as axonal swellings, axon terminal degeneration, and accumulation of synaptic vesicles in neuromuscular junctions, have been observed in the KI and transgenic mice expressing MND-related G59S mutant p150Glued, as well as the Dctn1LoxP/LoxP;Thy1-Cre mice which lack the MTBDs-containing p150Glued but express p135+ in the forebrain and spinal neurons22,27–29. Similarly, substantial loss and abnormal swellings of DAergic axon terminals have been found in transgenic mice with selective overexpression of PS-related G71R mutant p150Glued in the midbrain DAergic neurons32. In line with these findings, our current study revealed early-onset swelling of DAergic axon terminals with the accumulation of DA-containing vesicles, reduction of DAT, and the resultant dysregulation of DA transmission in the dorsal striatum of cKO mice. In addition, we detected late-onset loss of DAergic axon terminals in the dorsal striatum of cKO mice. Interestingly, the cKO mice displayed early-onset dystrophy of DAergic dendrites in the SNr, indicating an essential role of p150Glued in maintaining the structural integrity of both axons and dendrites of DAergic neurons. The substantial pathological changes of DAergic axons and dendrites might compromise the function of DAergic circuitry and contribute to the eventual death of DAergic neurons. Thus, future studies will be needed to unmask the mechanisms underlying the early-onset degeneration of DAergic axons and dendrites in PS.
Since the MND- and PS-linked missense mutations in p150Glued all reside within or close to the CAP-Gly domain and compromise the function of MTBDs8–13,23–26, the MTBDs in p150Glued must participate in some distinct cellular processes essential for the survival and function of spinal motor neurons and midbrain DAergic neurons. In a recent study, we genetically deleted p150Glued but kept p135+ expression in the forebrain and spinal neurons and found no widespread neuronal loss but selective degeneration of spinal motor neurons22. Further histological and biochemical assays revealed that the lack of MTBDs in p150Glued promoted acetylation of microtubules, impairment of the autophagosome-lysosome pathway, cell surface targeting of ionotropic glutamate receptors, and vulnerability to glutamate-induced excitotoxicity in spinal motor neurons22. Here when we genetically deleted p150Glued but kept p135+ in midbrain DAergic neurons, we found that p150Glued deficiency particularly affected the structure and function of ER. The lack of MTBDs in p150Glued led to a substantial reorganization of ER in the dystrophic DAergic dendrites, a preferential increase of ER tubule-shaping protein RTN3, accumulation of DAT within the reorganized ER, decreased ER targeting of COPII, dysfunction of ER export, activation of UPR, and increased susceptibility to ER stress-induced cell death in midbrain DAergic neurons. These findings highlight the important role of p150Glued and its MTBDs in maintaining the structural and functional homeostasis of ER, which is critical for the survival and function of midbrain DAergic neurons. Since the carboxyl terminus of DAT protein also binds to COPII component SEC24 and genetic knockdown of SEC24 impairs the cell surface targeting of DAT58,59, the impaired ER export likely also contributes to the reduced DAT level in the axon terminals of p150Glued-deficient DAergic neurons. Further investigations will be required to determine how the PS-related mutations in p150Glued detrimentally affect the structure and function of ER, exerting pathogenic impacts on the survival and function of the midbrain DAergic neurons.
While we focused the present study on the role of p150Glued in midbrain DAergic neurons and PS-related parkinsonism, it remains to be determined the underlying pathological mechanisms of PS-related psychiatric symptoms such as depression and apathy, which appeared as an early clinical feature prior to parkinsonism10–13. Except for neuronal loss in the midbrain, the postmortem brains of PS patients also displayed variable degrees of neurodegeneration in other nuclei in basal ganglia, as well as in the hypothalamus, locus coeruleus (LC), periaqueductal gray matter, ventrolateral medulla, dorsal raphe nucleus, and pontine reticular formation10–13. In our cKO mice, deletion of p150Glued was observed in ~$100\%$, $46\%$, $2\%$, $99\%$, and $98\%$ of TH+ cells in the midbrain, locus coeruleus, olfactory bulb, superior cervical ganglion, and adrenal medulla, respectively, suggesting differential efficiencies of Cre-mediated recombination in various populations of TH-expressing cells. On the one hand, this makes the cKO mice a valuable tool for studying the role of p150Glued in the survival and function of DAergic neurons in the midbrain, sympathetic neurons in the superior cervical ganglion, and chromaffin cells in the adrenal medulla. On the other hand, this makes the cKO mice less suitable for studying the role of p150Glued in the survival and function of TH-expressing neurons in the locus coeruleus and olfactory bulb. Future studies are needed to crossbreed the Dctn1LoxP/ mice with different Cre lines to continue exploring the role of p150Glued in LC adrenergic neurons and other cell types. For example, the loss of serotonergic neurons in the medullary raphe and ventrolateral medulla may be responsible for the respiratory failure seen in PS patients60. It would be interesting to test this hypothesis with a selective deletion of p150Glued in those serotonergic neurons using a suitable Cre line.
While TDP-43–positive cytoplasmic inclusions were often identified in the SN and other brain regions of PS patients9,36,37, we did not detect any apparent accumulation of TDP-43 in the cytosol of p150Glued-deficient DAergic neurons. This observation suggests the TDP-43 pathology might not contribute to the loss of midbrain DAergic neurons in the cKO mice during aging. Nonetheless, future studies would be needed to determine whether the TDP-43 pathology contributes to the degeneration of specific types of neurons with p150Glued deficiency or mutations. In contrast, although the α-synuclein-containing Lewy bodies were rarely spotted in the postmortem brain of PS patients10, we observed somatic accumulation of α-synuclein and p-α-synuclein (Ser129) in the midbrain DAergic neurons of aged cKO mice. Besides, we found no apparent accumulation of α-synuclein in the soma of TH-negative neurons in the SNc, SNr, dorsal striatum, hippocampal CA1 area, and prefrontal cortex of aged cKO mice, indicating that the intercellular spread of α-synuclein pathology from midbrain DAergic neurons to other neurons in the neighboring and interconnected brain areas is a relatively rare event in our mouse model. Our findings are in line with previous studies that the abnormal accumulation of α-synuclein in cytosol and nucleus contributes to neurodegeneration38,39. However, it remains to determine how the deficiency of p150Glued leads to α-synuclein pathology and how the accumulation of α-synuclein and the ER defects interplay within the p150Glued-deficient DAergic neurons.
In conclusion, to understand how the loss-of-function of p150Glued protein contributes to PS-related parkinsonism and DAergic neuron loss, we genetically deleted the MTBDs-containing p150Glued but kept the MTBD-lacking p135+ in the midbrain DAergic neurons of Dctn1LoxP/LoxP;Th-Cre cKO mice. The cKO mice developed progressive impairment of motor coordination and dysfunction/degeneration of DAergic neurons, including early-onset dendritic dystrophy, axonal swelling, DAT reduction, and dysregulated DA transmission, as well as late-onset neuronal death, axonal loss, and α-synuclein accumulation. We further revealed the impacts of p150Glued deficiency on the ER in midbrain DAergic neurons, such as the reorganization of ER in dystrophic DAergic dendrites, the upregulation of ER tubule-shaping protein RTN3, the accumulation of DAT in reorganized ERs, the dysfunction of COPII-mediated ER export, the activation of UPR pathway, and the exacerbation of ER stress-induced cell death. Our studies raise the significance of defective ER structure and function in the degeneration of p150Glued-deficient DAergic neurons. The cKO mice may serve as a valuable animal model to further investigate the pathogenic mechanism and therapeutic targets of midbrain DAergic neuron degeneration in PS.
## Animals
Dctn1LoxP/ mice (JAX, #032428) with two LoxP sites inserted in intron 1 and 4 of the *Dctn1* gene locus were generated in our previous work22. Dctn1LoxP/ mice were further crossbred with Cre recombinase (Cre) mouse lines to obtain Dctn1LoxP/LoxP;Cre mice, in which Cre recombinase-mediated deletion of Dctn1 exons 2 to 4 abolished the expression of p150Glued in the Cre-expressing cells22.
In this project, Dctn1LoxP/ mice were mated with Th-Cre mice (MMRRC, #029177-UCD), which have functional Cre expression in catecholamine DAergic and noradrenergic neurons33. This breeding strategy produced Dctn1LoxP/LoxP;Th-Cre mice [referred to as conditional knockout (cKO) mice], which had selective deletion of p150Glued in midbrain DAergic neurons. The cKO mice and littermate controls were used for in vivo study.
In this project, Dctn1LoxP/ mice were also mated with Cre/Esr1 mice (JAX, #004682) which have a tamoxifen-inducible form of Cre, capable of deleting floxed sequences in widespread cells or tissues34. This breeding strategy produced Dctn1LoxP/LoxP;Cre/Esr1 mice [referred to as inducible knockout (iKO) mice], which had inducible deletion of p150Glued expression after exposure to tamoxifen or 4-hydroxytamoxifen (4-OHT). The neonatal iKO pups and littermate controls were used for primary cell culture and in vitro study. Cohorts of 3-month-old Ctrl and iKO mice were intraperitoneally injected with tamoxifen (Sigma-Aldrich, solubilized in $100\%$ sunflower seed oil) at the dosage of 100 mg/kg body weight for five consecutive days. One month after tamoxifen injection, midbrains of Ctrl and cKO mice were collected and used for ER microsomes preparation and biochemical assays.
All the mice were housed in a 12-h light/dark cycle and fed a regular diet ad libitum. All mouse work followed the guidelines approved by the Institutional Animal Care and Use Committees of the National Institute on Aging (No. 13-040) and Beijing Geriatric Hospital (No. BGH-2020-001).
## Genotyping
Genomic DNA was prepared from tail biopsy using DirectPCR Lysis Reagent (Viagen Biotech) and subjected to PCR amplification using specific sets of PCR primers for wild-type or floxed *Dctn1* gene (5′-CAGCTGCAAAGACCAGCAAA-3′ and 5′-CACACCACCTTCTTAGGCTTCA-3′) and Cre transgene (5′-CATTTGGGCCAGCTAAACAT-3′ and 5′-TGCATGATCTCCGGTATTGA-3′)22.
## Behavior tests
Body weight and motor function were repeated assessed on cohorts of male cKO mice ($$n = 16$$) and their littermate controls ($$n = 16$$) at 1, 3, 6, 12, and 18 months of age. Test performers were blinded to the genotypes of the mice.
Open-field test-Mice were placed in the open-field apparatus with infrared photobeam sensors. Locomotor activities (including ambulatory, rearing, and fine movement) and time spent in the center area (~$40\%$ of the total surface of the arena) of mice were measured by the Flex-Field Activity System (San Diego Instruments)22. Flex-Field software was used to trace and quantify mouse movement in the unit as the number of beam breaks per 30 min.
Rotarod test-Mice were placed onto a rotating rod with auto-acceleration from 0 rpm to 40 rpm for 1 min (San Diego Instruments)22. The length of time the mouse stayed on the rotating rod was recorded. Three measurements were taken for each animal during each test.
## Immunohistochemistry and light microscopy
Mice were sacrificed and transcardially perfused with $4\%$ paraformaldehyde (PFA) in cold phosphate-buffered saline (PBS). Mouse brains, superior cervical ganglions, and adrenal glands were collected, post-fixed in $4\%$ PFA/PBS solution overnight, submerged in $30\%$ sucrose in PBS for at least 72 h, and sectioned at 40 μm thickness using CM1950 cryostat (Leica)22. Frozen sections were stained with antibodies specific to p150Glued (amino acid 3–202 at the N-terminus of p150Glued, BD Biosciences, #610474, 1:200, recognizing p150Glued but not p135+), p150Glued & p135+ (amino acid 1266–1278 at the C-terminus of p150Glued, Abcam, #ab11806, 1:500, recognizing both p150Glued and p135+), tyrosine hydroxylase (TH, Pel-Freez, #P40101-150, 1:2500; ImmunoStar, #22941, 1:500; Synaptic Systems, #213104, 1:500), dopamine transporter (DAT, Millipore, #MAB369, 1:500), vesicular monoamine transporter 2 (VMAT2, Synaptic Systems, #138302, 1:1000), glial fibrillary acidic protein (GFAP, Abcam, #ab7260, 1:1000), TAR DNA-binding protein 43 (TDP-43, Proteintech, #10782-2-AP, 1:500), α-synuclein (Santa Cruz, #sc-7011-R, 1:500; Santa Cruz, #sc-69977, 1:500), phosphorylated α-synuclein (Ser129) [p-α-synuclein (Ser129), Abcam, #ab51253, 1:500], neuronal nuclei (NeuN, Millipore, #ABN91, 1:500), synaptophysin (Millipore, #AB9272, 1:500), binding immunoglobulin protein (BiP, also referred to as GRP78, Abcam, #ab21685, 1:500), reticulon 3 (RTN3, Proteintech, #12055-2-AP, 1:500), 63 kDa cytoskeleton-linking membrane protein (CLIMP63, Proteintech, #16686-1-AP, 1:500), calnexin (Abcam, #ab22595, 1:500), protein disulfide isomerase (PDI, Proteintech, #11245-1-AP, 1:500), receptor binding cancer antigen expressed on SiSo cells (RCAS1, Cell Signaling Technology, #12290, 1:500), early endosome antigen 1 (EEA1, Cell Signaling Technology, #3288, 1:500), sequestosome 1 (SQSTM1, MBL, #PM066, 1:500), cathepsin D (R&D Systems, #AF1029, 1:500), ER-Golgi intermediate compartment 53 kDa protein (ERGIC53, Sigma-Aldrich, #E1031, 1:500), 130 kDa cis-Golgi matrix protein (GM130, BD Biosciences, #610822, 1:500), phosphorylated eukaryotic translation initiation factor 2α (Ser51) [p-eIF2α (Ser51), Abcam, #ab32157, 1:500], and phosphorylated inositol-requiring enzyme 1α (Ser724) [p-IRE1α (Ser724), Abcam, #ab48187, 1:500] as suggested by manufacturers. Alexa Fluor 488-, 546-, or 647-conjugated secondary antibody (Invitrogen, 1:500) was used to visualize the staining. Fluorescent images were captured using LSM 880 laser-scanning confocal microscope with Zen software (Zeiss) in conventional or Airyscan mode. As a high-resolution imaging modality, the Airyscan technology is reported to improve resolution 2-fold and signal-to-noise ratio 8-fold relative to the conventional confocal microscopy61. The paired images in all the figures were collected at the same gain and offset settings. Post-collection processing was applied uniformly to all paired images. The images were presented as a single optic layer after acquisition in z-series stack scans at 1.0 μm intervals from individual fields or displayed as maximum-intensity projection or three-dimensional (3D) reconstruction to represent confocal stacks.
## Stereology
Unbiased stereology was performed to estimate the number of midbrain DAergic neurons62,63. According to the mouse brain in stereotaxic coordinates, a series of 40-μm-thick coronal sections across the midbrain (every fourth section from Bregma −2.54 to −4.24 mm, ten sections per case) were stained with an antibody specific to TH (Pel-Freez, #P40101-150, 1:2500) and subsequently visualized with Vectastain Elite ABC Kit and DAB Kit (Vector Laboratories). Bright-field images were captured by Axio microscope Imager A1 (Zeiss). The number of TH-positive neurons was assessed using the optical fractionator function of Stereo Investigator 10 (MicroBrightField). Four or more mice were used per genotype at each time point. Counters were blinded to the genotypes of the samples. The sampling scheme was designed to have a coefficient of error <$10\%$ in order to obtain reliable results.
## Image analysis
For the quantitative assessment of various marker protein distributions, images were taken using identical settings and exported to ImageJ (NIH) for imaging analysis. Images were converted to an 8-bit color scale (fluorescence intensity from 0 to 255) by ImageJ (NIH). The areas of interest were first selected with Polygon or Freehand selection tools and then subjected to measurement by mean optical intensities or area fractions. The mean intensity for the background area was subtracted from the selected area to determine the net mean intensity. Data analyzers were blinded to the genotypes of the samples.
For quantitative analysis of the dystrophy of DAergic dendrites32, five tile images per animal (5 sections per animal and 1 tile image per section) from the SNr were taken with a ×40 lens. Dystrophic DAergic dendrites were defined as TH-positive neuritic varicosity ≥25 μm2. The number of dystrophic DAergic dendrites and the area of the SNr were quantified with ImageJ (NIH).
For quantitative analysis of the loss of DAergic axon terminals32, forty images per animal (10 sections per animal and 4 images per section) from the dorsal striatum were taken with a ×63 lens. The area fraction of DAergic axon terminals in each image was quantified with ImageJ (NIH).
For quantitative analysis of the swelling of DAergic axon terminals, five tile images per animal (5 sections per animal and 1 tile image per section) from the dorsal striatum were taken with a ×63 lens. Swollen DAergic axon terminals were defined as TH-positive neuritic varicosity ≥3 μm2. The number of swollen DAergic axon terminals and the dorsal striatum area were quantified with ImageJ (NIH).
## High-performance liquid chromatography (HPLC)
HPLC was performed to measure the content of DA and its metabolites in the striatum32,63,64. Mouse dorsal striatum was dissected, weighted, and homogenized in 500 μl buffer (0.1 N perchloric acid containing 100 μM EDTA) per 100 mg of tissue. After sonication and centrifugation, the supernatant was collected, frozen, and stored at −80 °C until assayed for DA, 3,4-dihydroxyphenylacetic acid (DOPAC), and 5-hydroxytryptamine (5-HT) by liquid chromatography with electrochemical detection. Briefly, the mobile-phase solution containing octanesulfonic acid as an ion-pairing agent was pumped isocratically through a reversed-phase liquid chromatographic column. DA, DOPAC, and 5-HT were quantified by the current produced after exposure of the eluate to a flow-through electrode set to oxidizing and then reducing potentials in series, with recordings from the last electrode reflecting reversibly oxidized species.
## Fast-scan cyclic voltammetry (FSCV)
To investigate the kinetics of DA release evoked by electrical stimulation, FSCV was performed in 400-μm-thick slices of the dorsal striatum32,63. Striatal slices were bathed in 32 °C oxygenated artificial cerebrospinal fluid [aCSF: 126 mM NaCl, 2.5 mM KCl, 1.2 mM NaH2PO4, 2.4 mM CaCl2, 1.2 mM MgCl2, 25 mM NaHCO3, 11 mM glucose, 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 0.4 mM L-ascorbic acid]. Cylindrical carbon-fiber microelectrodes (50–100 μm of exposed fiber) were prepared with T650 fibers (6 μm diameter, Goodfellow) and inserted into a glass pipette. The carbon-fiber electrode was held at −0.4 V, and the potential was increased to 1.2 V and back at 400 V/s every 100 ms using a triangle waveform. DA release was evoked by rectangular, electrical pulse stimulation (100–600 μA, 0.6 ms per phase, biphasic) applied every 5 min. Data collection and analysis were performed using the Demon Voltammetry and Analysis software suite65. Ten cyclic voltammograms of charging currents were recorded as background before stimulation, and the average of these responses was subtracted from data collected during and after stimulation. Maximum amplitudes of extracellular DA transients were obtained from input/output function (I/O) curves. I/O curves were constructed by plotting stimulus current against the concentration of DA response amplitude over a range of stimulus intensities. The time constant of the slope decay (τ) was used for uptake kinetic analysis of evoked DA release. Following experiments, electrodes were calibrated using solutions of 1 and 10 μM DA in aCSF.
## Primary midbrain neuronal culture
Mouse primary midbrain neuronal cultures were prepared from newborn Dctn1LoxP/LoxP;Cre/Esr1 pups and littermate controls on P022,62. Briefly, midbrain tissues containing SNc and VTA were dissected and subjected to papain digestion (5 U/ml, Worthington Biochemicals) for 40 min at 37 °C. The digested tissue was carefully triturated into single cells using increasingly smaller pipette tips. The cells were then centrifuged at 250 × g for 5 min and resuspended in warm medium [Basal Medium Eagle (BME, Sigma-Aldrich), 1× N2/B27 supplement (the optimized serum-free supplement used to support the growth and viability of neurons, 100× stock, Invitrogen), 1× GlutaMax (100× stock, Invitrogen), $0.45\%$ D-glucose (Sigma-Aldrich), 10 U/ml penicillin (Invitrogen), and 10 μg/ml streptomycin (Invitrogen)] supplemented with $5\%$ heat-inactivated fetal bovine serum (FBS, Invitrogen). Dissociated midbrain neurons (~3 × 105 cells per coverslip) were plated onto 12-mm round coverslips precoated with poly-D-lysine and laminin (BD Bioscience) in a 24-well plate and maintained at 37 °C in the $95\%$ O2- and $5\%$ CO2-humidified incubator. Twenty-four hours after seeding, the cultures were switched to the serum-free medium supplemented with 5 μM cytosine β-D-arabinofuranoside (Sigma-Aldrich) to suppress the proliferation of glia and 1 μM 4-OHT to induce CRE recombinase activity. From 5 days in vitro (DIV), culture medium was changed twice weekly.
## Assessment of DAergic neuron survival after chemical treatment or growth factor deprivation
For chemical treatment, the primary midbrain neurons at 14 DIV were exposed to vehicle (dimethyl sulfoxide, DMSO), 10 nM thapsigargin (an ER stress inducer, Sigma-Aldrich, #T9033), 100 nM GSK2606414 (a PERK inhibitor, MedChemExpress, #HY-18072)66, or 100 nM KIRA8 (an IRE1α inhibitor, MedChemExpress, #HY-114368)67 for 48 or 96 h. For growth factor deprivation, the primary midbrain neurons at 14 DIV were deprived of growth factor by lowering the content of N2/B27 supplement in the culture medium from 1× to 0.1× for 48 h. After chemical treatment or growth factor deprivation, neurons were fixed with $4\%$ PFA/PBS and immunostained with TH antibody and secondary antibody. The number of TH-positive neurons on each coverslip was counted under the confocal microscope with a ×40 objective. Counters were blinded to the genotypes and treatments of the samples. The survival rate of DAergic neurons was calculated by dividing the number of TH-positive neurons on each coverslip by the number of TH-positive neurons on the control coverslip (neurons from Dctn1LoxP/LoxP P0 pups treated with vehicle or cultured in normal medium)62.
## Primary fibroblast culture
Mouse fibroblast cultures were prepared from the dorsal skin of newborn Dctn1LoxP/LoxP;Cre/Esr1 pups and littermate controls on postnatal day 0 (P0)54. Briefly, skin tissues were collected, rinsed in sterile PBS, and minced into small pieces. The minced tissues were triturated with 2 ml medium [DMEM (Invitrogen) supplemented with $10\%$ FBS, 10 U/ml penicillin, and 10 μg/ml streptomycin], sparsely plated into a 10-cm culture dish, and maintained at 37 °C in the $95\%$ O2- and $5\%$ CO2-humidified incubator. After overnight (tissue fragments usually attach firmly to the dish), 8 ml fresh medium was added. After 3–4 days (fibroblasts grown out of tissue fragments usually start to undergo rapid proliferation), the medium was changed. At 7 DIV, fibroblasts were trypsinized with TrypLE (Invitrogen), passaged into new dishes, and treated with 1 μM 4-hydroxytamoxifen (4-OHT) to induce CRE recombinase activity. From 11 DIV on, the medium was changed every 3–4 days.
## Immunocytochemistry and light microscopy
Cultured cells on coverslips were fixed with $4\%$ PFA in PBS for 15 min, permeabilized by $0.1\%$ Triton X-100 for 5 min, and blocked in $10\%$ normal donkey serum (Invitrogen) in PBS for 1 h at room temperature54,62. Cells were then labeled with primary antibodies against p150Glued (BD Biosciences, #610474, 1:200, recognizing p150Glued but not p135+), tyrosine hydroxylase (TH, Pel-Freez, #P40101-150, 1:2000), microtubule-associated protein 2 (MAP2, Abcam, #92434, 1:1000), binding immunoglobulin protein (BiP, also referred to as GRP78, Abcam, #ab21685, 1:500), and α-tubulin (Abcam, #ab89984, 1:1000) overnight at 4 °C in a humidified chamber. After three washes with PBS, secondary antibodies conjugated to Alexa Fluor (Invitrogen, 1:1000) were applied and incubated for 1 h at room temperature in the dark. After extensive washes, coverslips were mounted on glass slides with prolonged diamond antifade reagent containing DAPI (Invitrogen), and fluorescence signals were detected using LSM 880 laser-scanning confocal microscope (Zeiss). The paired images in all the figures were collected at the same acquisition settings, uniformly processed, presented as either a single optic layer or maximum-intensity projection of confocal stacks, and analyzed with ImageJ (NIH).
## Preparation of subcellular fractions
According to the manufacturer’s instructions, subcellular fractions were prepared using an endoplasmic reticulum isolation kit (Sigma-Aldrich). All procedures were performed at 4 °C. Mouse midbrains were isolated, cut into small pieces, and homogenized in four volumes of ice-cold Isotonic Extraction Buffer (10 mM HEPES, pH 7.8, 250 mM sucrose, 25 mM KCl, 1 mM EGTA, and 1× Protease and Phosphatase Inhibitor Cocktail) with Dounce homogenizer (12 strokes). Cultured cells were harvested, washed with ten volumes of PBS, and spun down at 600 × g for 5 min. The cell pellet was suspended and incubated in three volumes of ice-cold Hypotonic Extraction Buffer (10 mM HEPES, pH 7.8, 25 mM KCl, 1 mM EGTA, and 1× Protease and Phosphatase Inhibitor Cocktails) for 20 min to allow the cells to swell. Swollen cells were centrifuged at 600 × g for 5 min. The new cell pellet was homogenized in two volumes of ice-cold Isotonic Extraction Buffer with Dounce homogenizer (10 strokes). The homogenate (referred to as total lysate) from brain tissues or cultured cells was spun at 1000 × g for 10 min. The supernatant (S1) was collected, and the pellet (P1) was saved as crude nuclei fraction. S1 was centrifuged at 12,000 × g for 15 min. The supernatant (S2) was collected, and the pellet (P2) was saved as crude mitochondria fraction. S2 was further centrifuged at 100,000 × g for 60 min. The supernatant (S3) was collected as cytosol fraction, and the pellet (P3) was saved as ER microsomes fraction. Crude nuclei fraction, crude mitochondria fraction, and ER microsomes fraction were lysed in $1\%$ SDS buffer. SDS was added to the total lysate and cytosol fraction to $1\%$ final concentration. Equal amounts of protein from total lysate and each fraction were resolved in SDS-PAGE and applied to western blot analysis.
## Co-immunoprecipitation (co-IP)
Mouse brains or ER microsomes fraction of mouse brains were homogenized in IP buffer (50 mM Tris, pH 7.5, 150 mM NaCl, $10\%$ glycerol, 50 mM NaF,10 mM glycerolphosphate, 2 mM EGTA, 2 mM EDTA, $1\%$ NP-40, and 1× Protease and Phosphatase Inhibitor Cocktails) with Dounce homogenizer (10 strokes)54. Lysates were centrifuged at 15,000 × g for 15 min at 4 °C, and the supernatants were collected. The protein concentration of the lysates was measured and adjusted to 1 mg/ml. After pre-clearing with Protein G agarose (Thermo Fisher Scientific), the lysates were incubated with antibody-bound Protein G agarose for 1 h at 4 °C. After five washes of the agarose beads with IP buffer at 4 °C, the immune complexes were eluted with SDS sample buffer (Thermo Fisher Scientific) and examined by western blotting. The mouse-derived specific antibody against p150Glued (BD Biosciences, #610474, 1:1000, recognizing p150Glued but not p135+) and normal mouse IgG (Santa Cruz, #sc-2025) were used for co-IP.
## Western blotting
Cultured cells or mouse tissues were homogenized by sonication in ice-cold lysis buffer [50 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, pH 7.5, $1\%$ SDS, and 1× Protease and Phosphatase Inhibitor Cocktails (Thermo Fisher Scientific)]22,32. Lysates were centrifuged at 15,000 × g for 15 min at 4 °C. The supernatants were collected and quantified for protein content using the bicinchoninic acid (BCA) assay kit (Thermo Fisher Scientific). Equal amounts of total protein were separated by NuPage 4–$12\%$ Bis-Tris gel using MES or MOPS running buffer (Thermo Fisher Scientific). The separated proteins were then transferred to nitrocellulose membranes using the Trans-Blot Turbo Transfer system (Bio-Rad) and incubated with specific primary antibodies. The antibodies used for western blot analysis included p150Glued (BD Biosciences, #610474, 1:1000, recognizing p150Glued but not p135+), p150Glued & p135+ (Abcam, #ab11806, 1:1000, recognizing both p150Glued and p135+), dynactin subunit 4 (DCTN4, Abcam, #ab170107, 1:1000), dynactin subunit p50 (BD Biosciences, #611002, 1:1000), dynactin subunit actin-related protein 1 (ARP1, Sigma-Aldrich, #A5601, 1:1000), tyrosine hydroxylase (TH, Sigma-Aldrich, #T1299, 1:1000), dopamine transporter (DAT, Millipore, #MAB369, 1:1000), VAMP (vesicle-associated membrane protein)-associated protein B (VAPB, Proteintech, #14477-1-AP, 1:1000), calnexin (Abcam, #ab22595, 1:1000), protein disulfide isomerase (PDI, Proteintech, #11245-1-AP, 1:1000), binding immunoglobulin protein (BiP, also referred to as GRP78, Abcam, #ab21685, 1:1000), endoplasmic reticulum-resident protein 72 (ERp72, Cell Signaling Technology, #5033, 1:1000), 63 kDa cytoskeleton-linking membrane protein (CLIMP63, Proteintech, #16686-1-AP, 1:1000), atlastin 1 (ATL1, Cell Signaling Technology, #12728, 1:1000), atlastin 2 (ATL2, Proteintech, #16688-1-AP, 1:1000), atlastin 3 (ATL3, Proteintech, #16921-1-AP, 1:1000), reticulon 1 (RTN1, Proteintech, #105048-1-AP, 1:1000), reticulon 3 (RTN3, Proteintech, #12055-2-AP, 1:1000), reticulon 4 (RTN4, Proteintech, #10950-1-AP, 1:1000), Sec13 [component of the coat protein complex II (COPII), Santa Cruz, #SC-514308, 1:1000], Sec23 (component of COPII, Thermo Fisher Scientific, #PA1-069A, 1:1000), Sec31 (component of COPII, BD Biosciences, #612351, 1:1000), nicastrin (Cell Signaling Technology, #3632, 1:1000), protein kinase-like endoplasmic reticulum kinase (PERK, Proteintech, #20582-1-AP, 1:1000), p-PERK (Thr982) (Thermo Fisher Scientific, #PA5-40294, 1:1000), eukaryotic translation initiation factor 2α (eIF2α, Cell Signaling Technology, #5324, 1:1000), phosphorylated eIF2α (Ser51) [p-eIF2α (Ser51), Abcam, #ab32157, 1:1000], activating transcription factor 4 (ATF4, Proteintech, #10835-1-AP, 1:1000), C/EBP-homologous protein (CHOP, Cell Signaling Technology, #2895, 1:1000), inositol-requiring enzyme 1α (IRE1α, Cell Signaling Technology, #3294, 1:1000), phosphorylated IRE1α (Ser724) [p-IRE1α (Ser724), Abcam, #ab48187, 1:1000], unspliced X-box binding protein 1 (unspliced XBP1, Proteintech, #25997-1-AP, 1:1000), spliced XBP1 (Proteintech, #24858-1-AP, 1:1000), stress-activated protein kinase/Jun-amino-terminal kinase (SAPK/JNK, Cell Signaling Technology, #9252, 1:1000), phosphorylated SAPK/JNK (Thr183/Tyr185) [p-SAPK/JNK (Thr183/Tyr185), Cell Signaling Technology, #4668, 1:1000], activating transcription factor (ATF6, Proteintech, #24169-1-AP, 1:1000), cleaved caspase-3 (Cell Signaling Technology, #9664, 1:1000), glyceraldehyde-3-phosphate dehydrogenase (GAPDH, Sigma-Adrich, #G9545, 1:1000), α-tubulin (Abcam, #ab7291, 1:5000), and β-actin (Sigma-Aldrich, #A1978, 1:5000). Protein signals were visualized by IRDye secondary antibodies and Odyssey system (LI-COR Biosciences) and quantified with ImageJ (NIH). Data analyzers were blinded to the genotypes of the samples.
## Statistical analysis
Statistical analysis was performed using GraphPad Prism 9 (GraphPad Software). Data were presented as mean ± SEM. Statistical significance was determined by comparing means of different groups using unpaired t test, one-way ANOVA with Tukey’s multiple comparisons test, two-way ANOVA with post-hoc Bonferroni test, or two-way ANOVA with Sidak’s multiple comparisons test. Not significant (ns), p ≥ 0.05; *$p \leq 0.05$; **$p \leq 0.01$; ***$p \leq 0.001$; ****$p \leq 0.0001.$
## Supplementary information
Supplementary Information Supplementary Movie 1 Supplementary Movie 2 The online version contains supplementary material available at 10.1038/s41531-023-00478-0.
## References
1. Reck-Peterson SL, Redwine WB, Vale RD, Carter AP. **The cytoplasmic dynein transport machinery and its many cargoes**. *Nat. Rev. Mol. Cell Biol.* (2018.0) **19** 382-398. DOI: 10.1038/s41580-018-0004-3
2. Canty JT, Yildiz A. **Activation and regulation of cytoplasmic dynein**. *Trends Biochem. Sci.* (2020.0) **45** 440-453. DOI: 10.1016/j.tibs.2020.02.002
3. Radler MR, Suber A, Spiliotis ET. **Spatial control of membrane traffic in neuronal dendrites**. *Mol. Cell Neurosci.* (2020.0) **105** 103492. DOI: 10.1016/j.mcn.2020.103492
4. Cason SE, Holzbaur ELF. **Selective motor activation in organelle transport along axons**. *Nat. Rev. Mol. Cell Biol.* (2022.0) **23** 699-714. DOI: 10.1038/s41580-022-00491-w
5. Lipka J, Kuijpers M, Jaworski J, Hoogenraad CC. **Mutations in cytoplasmic dynein and its regulators cause malformations of cortical development and neurodegenerative diseases**. *Biochem. Soc. Trans.* (2013.0) **41** 1605-1612. DOI: 10.1042/BST20130188
6. Cianfrocco MA, DeSantis ME, Leschziner AE, Reck-Peterson SL. **Mechanism and regulation of cytoplasmic dynein**. *Annu. Rev. Cell. Dev. Biol.* (2015.0) **31** 83-108. DOI: 10.1146/annurev-cellbio-100814-125438
7. Jaarsma D, Hoogenraad CC. **Cytoplasmic dynein and its regulatory proteins in Golgi pathology in nervous system disorders**. *Front. Neurosci.* (2015.0) **9** 397. DOI: 10.3389/fnins.2015.00397
8. Puls I. **Mutant dynactin in motor neuron disease**. *Nat. Genet.* (2003.0) **33** 455-456. DOI: 10.1038/ng1123
9. Farrer MJ. **DCTN1 mutations in Perry syndrome**. *Nat. Genet.* (2009.0) **41** 163-165. DOI: 10.1038/ng.293
10. Konno T. **DCTN1-related neurodegeneration: Perry syndrome and beyond**. *Parkinsonism Relat. Disord.* (2017.0) **41** 14-24. DOI: 10.1016/j.parkreldis.2017.06.004
11. Mishima T. **Establishing diagnostic criteria for Perry syndrome**. *J. Neurol. Neurosurg. Psychiatry* (2018.0) **89** 482-487. DOI: 10.1136/jnnp-2017-316864
12. Tsuboi Y, Mishima T, Fujioka S. **Perry disease: concept of a new disease and clinical diagnostic criteria**. *J. Mov. Disord.* (2021.0) **14** 1-9. DOI: 10.14802/jmd.20060
13. Dulski J. **Clinical, pathological and genetic characteristics of Perry disease-new cases and literature review**. *Eur. J. Neurol.* (2021.0) **28** 4010-4021. DOI: 10.1111/ene.15048
14. Chung EJ. **Expansion of the clinicopathological and mutational spectrum of Perry syndrome**. *Parkinsonism Relat. Disord.* (2014.0) **20** 388-393. DOI: 10.1016/j.parkreldis.2014.01.010
15. Felicio AC. **In vivo dopaminergic and serotonergic dysfunction in DCTN1 gene mutation carriers**. *Mov. Disord.* (2014.0) **29** 1197-1201. DOI: 10.1002/mds.25893
16. Mishima T. **Cytoplasmic aggregates of dynactin in iPSC-derived tyrosine hydroxylase-positive neurons from a patient with Perry syndrome**. *Parkinsonism Relat. Disord.* (2016.0) **30** 67-72. DOI: 10.1016/j.parkreldis.2016.06.007
17. Schroer TA. **Dynactin**. *Annu. Rev. Cell Dev. Biol.* (2004.0) **20** 759-779. DOI: 10.1146/annurev.cellbio.20.012103.094623
18. Dixit R, Levy JR, Tokito M, Ligon LA, Holzbaur EL. **Regulation of dynactin through the differential expression of p150Glued isoforms**. *J. Biol. Chem.* (2008.0) **283** 33611-33619. DOI: 10.1074/jbc.M804840200
19. Zhapparova ON. **Dynactin subunit p150Glued isoforms notable for differential interaction with microtubules**. *Traffic* (2009.0) **10** 1635-1646. DOI: 10.1111/j.1600-0854.2009.00976.x
20. Tokito MK, Howland DS, Lee VM, Holzbaur EL. **Functionally distinct isoforms of dynactin are expressed in human neurons**. *Mol. Biol. Cell.* (1996.0) **7** 1167-1180. DOI: 10.1091/mbc.7.8.1167
21. Hammesfahr B, Kollmar M. **Evolution of the eukaryotic dynactin complex, the activator of cytoplasmic dynein**. *BMC Evol. Biol.* (2012.0) **12** 95. DOI: 10.1186/1471-2148-12-95
22. Yu J. **Genetic ablation of dynactin p150(Glued) in postnatal neurons causes preferential degeneration of spinal motor neurons in aged mice**. *Mol. Neurodegener.* (2018.0) **13** 10. DOI: 10.1186/s13024-018-0242-z
23. Lloyd TE. **The p150(Glued) CAP-Gly domain regulates initiation of retrograde transport at synaptic termini**. *Neuron* (2012.0) **74** 344-360. DOI: 10.1016/j.neuron.2012.02.026
24. Moughamian AJ, Holzbaur EL. **Dynactin is required for transport initiation from the distal axon**. *Neuron* (2012.0) **74** 331-343. DOI: 10.1016/j.neuron.2012.02.025
25. Lazarus JE, Moughamian AJ, Tokito MK, Holzbaur EL. **Dynactin subunit p150(Glued) is a neuron-specific anti-catastrophe factor**. *PLoS Biol.* (2013.0) **11** e1001611. DOI: 10.1371/journal.pbio.1001611
26. Levy JR. **A motor neuron disease-associated mutation in p150Glued perturbs dynactin function and induces protein aggregation**. *J. Cell Biol.* (2006.0) **172** 733-745. DOI: 10.1083/jcb.200511068
27. Lai C. **The G59S mutation in p150(glued) causes dysfunction of dynactin in mice**. *J. Neurosci.* (2007.0) **27** 13982-13990. DOI: 10.1523/JNEUROSCI.4226-07.2007
28. Laird FM. **Motor neuron disease occurring in a mutant dynactin mouse model is characterized by defects in vesicular trafficking**. *J. Neurosci.* (2008.0) **28** 1997-2005. DOI: 10.1523/JNEUROSCI.4231-07.2008
29. Chevalier-Larsen ES, Wallace KE, Pennise CR, Holzbaur EL. **Lysosomal proliferation and distal degeneration in motor neurons expressing the G59S mutation in the p150Glued subunit of dynactin**. *Hum. Mol. Genet.* (2008.0) **17** 1946-1955. DOI: 10.1093/hmg/ddn092
30. Mishima T. **Behavioral defects in a DCTN1(G71A) transgenic mouse model of Perry syndrome**. *Neurosci. Lett.* (2018.0) **666** 98-103. DOI: 10.1016/j.neulet.2017.12.038
31. Deshimaru M. **Behavioral profile in a Dctn1(G71A) knock-in mouse model of Perry disease**. *Neurosci. Lett* (2021.0) **764** 136234. DOI: 10.1016/j.neulet.2021.136234
32. 32.Yu, J. et al. Selective expression of neurodegenerative diseases-related mutant p150Glued in midbrain dopaminergic neurons causes progressive degeneration of nigrostriatal pathway. Ageing Neurodegener. Dis.2, 10.20517/and.2022.07 (2022).
33. Gong S. **Targeting Cre recombinase to specific neuron populations with bacterial artificial chromosome constructs**. *J. Neurosci.* (2007.0) **27** 9817-9823. DOI: 10.1523/JNEUROSCI.2707-07.2007
34. Hayashi S, McMahon AP. **Efficient recombination in diverse tissues by a tamoxifen-inducible form of Cre: a tool for temporally regulated gene activation/inactivation in the mouse**. *Dev. Biol.* (2002.0) **244** 305-318. DOI: 10.1006/dbio.2002.0597
35. Crittenden JR. **Striosome-dendron bouquets highlight a unique striatonigral circuit targeting dopamine-containing neurons**. *Proc. Natl. Acad. Sci. USA* (2016.0) **113** 11318-11323. DOI: 10.1073/pnas.1613337113
36. Wider C. **Pallidonigral TDP-43 pathology in Perry syndrome**. *Parkinsonism Relat. Disord.* (2009.0) **15** 281-286. DOI: 10.1016/j.parkreldis.2008.07.005
37. Wider C. **Elucidating the genetics and pathology of Perry syndrome**. *J. Neurol. Sci.* (2010.0) **289** 149-154. DOI: 10.1016/j.jns.2009.08.044
38. Cooper AA. **Alpha-synuclein blocks ER-Golgi traffic and Rab1 rescues neuron loss in Parkinson’s models**. *Science* (2006.0) **313** 324-328. DOI: 10.1126/science.1129462
39. Kontopoulos E, Parvin JD, Feany MB. **Alpha-synuclein acts in the nucleus to inhibit histone acetylation and promote neurotoxicity**. *Hum. Mol. Genet.* (2006.0) **15** 3012-3023. DOI: 10.1093/hmg/ddl243
40. Fujiwara H. **Alpha-Synuclein is phosphorylated in synucleinopathy lesions**. *Nat. Cell Biol.* (2002.0) **4** 160-164. DOI: 10.1038/ncb748
41. Uemura N, Uemura MT, Luk KC, Lee VM, Trojanowski JQ. **Cell-to-cell transmission of Tau and alpha-Synuclein**. *Trends Mol. Med.* (2020.0) **26** 936-952. DOI: 10.1016/j.molmed.2020.03.012
42. Wu Y. **Contacts between the endoplasmic reticulum and other membranes in neurons**. *Proc. Natl. Acad. Sci. USA* (2017.0) **114** E4859-E4867. DOI: 10.1073/pnas.1701078114
43. 43.Sree, S., Parkkinen, I., Their, A., Airavaara, M. & Jokitalo, E. Morphological heterogeneity of the endoplasmic reticulum within neurons and its implications in neurodegeneration. Cells10, 10.3390/cells10050970 (2021).
44. Westrate LM, Lee JE, Prinz WA, Voeltz GK. **Form follows function: the importance of endoplasmic reticulum shape**. *Annu. Rev. Biochem.* (2015.0) **84** 791-811. DOI: 10.1146/annurev-biochem-072711-163501
45. Zhang H, Hu J. **Shaping the endoplasmic reticulum into a social network**. *Trends Cell Biol.* (2016.0) **26** 934-943. DOI: 10.1016/j.tcb.2016.06.002
46. 46.Wang, N. & Rapoport, T. A. Reconstituting the reticular ER network - mechanistic implications and open questions. J. Cell Sci.132, 10.1242/jcs.227611 (2019).
47. Sharoar MG. **Dysfunctional tubular endoplasmic reticulum constitutes a pathological feature of Alzheimer’s disease**. *Mol. Psychiatry* (2016.0) **21** 1263-1271. DOI: 10.1038/mp.2015.181
48. Sharoar MG, Zhou J, Benoit M, He W, Yan R. **Dynactin 6 deficiency enhances aging-associated dystrophic neurite formation in mouse brains**. *Neurobiol. Aging* (2021.0) **107** 21-29. DOI: 10.1016/j.neurobiolaging.2021.07.006
49. Wang B. **The COPII cargo adapter SEC24C is essential for neuronal homeostasis**. *J. Clin. Invest.* (2018.0) **128** 3319-3332. DOI: 10.1172/JCI98194
50. Tang BL. **Defects in early secretory pathway transport machinery components and neurodevelopmental disorders**. *Rev. Neurosci.* (2021.0) **32** 851-869. DOI: 10.1515/revneuro-2021-0020
51. Watson P, Forster R, Palmer KJ, Pepperkok R, Stephens DJ. **Coupling of ER exit to microtubules through direct interaction of COPII with dynactin**. *Nat. Cell Biol.* (2005.0) **7** 48-55. DOI: 10.1038/ncb1206
52. Verissimo F, Halavatyi A, Pepperkok R, Weiss M. **A microtubule-independent role of p150glued in secretory cargo concentration at endoplasmic reticulum exit sites**. *J. Cell Sci.* (2015.0) **128** 4160-4170. PMID: 26459637
53. Dries DR, Yu G. **Assembly, maturation, and trafficking of the gamma-secretase complex in Alzheimer’s disease**. *Curr. Alzheimer Res.* (2008.0) **5** 132-146. DOI: 10.2174/156720508783954695
54. Cho HJ. **Leucine-rich repeat kinase 2 regulates Sec16A at ER exit sites to allow ER-Golgi export**. *EMBO J.* (2014.0) **33** 2314-2331. DOI: 10.15252/embj.201487807
55. Hetz C, Papa FR. **The unfolded protein response and cell fate control**. *Mol. Cell* (2018.0) **69** 169-181. DOI: 10.1016/j.molcel.2017.06.017
56. Metcalf MG, Higuchi-Sanabria R, Garcia G, Tsui CK, Dillin A. **Beyond the cell factory: Homeostatic regulation of and by the UPR(ER)**. *Sci. Adv.* (2020.0) **6** eabb9614. DOI: 10.1126/sciadv.abb9614
57. Collins MK, Perkins GR, Rodriguez-Tarduchy G, Nieto MA, Lopez-Rivas A. **Growth factors as survival factors: regulation of apoptosis**. *Bioessays* (1994.0) **16** 133-138. DOI: 10.1002/bies.950160210
58. Sucic S. **The serotonin transporter is an exclusive client of the coat protein complex II (COPII) component SEC24C**. *J. Biol. Chem.* (2011.0) **286** 16482-16490. DOI: 10.1074/jbc.M111.230037
59. Bu M, Farrer MJ, Khoshbouei H. **Dynamic control of the dopamine transporter in neurotransmission and homeostasis**. *NPJ Parkinsons Dis.* (2021.0) **7** 22. DOI: 10.1038/s41531-021-00161-2
60. Tsuboi Y. **Neurodegeneration involving putative respiratory neurons in Perry syndrome**. *Acta Neuropathol.* (2008.0) **115** 263-268. DOI: 10.1007/s00401-007-0246-1
61. Wu X, Hammer JA. **ZEISS airyscan: optimizing usage for fast, gentle, super-resolution imaging**. *Methods Mol. Biol.* (2021.0) **2304** 111-130. DOI: 10.1007/978-1-0716-1402-0_5
62. Liu G. **Aldehyde dehydrogenase 1 defines and protects a nigrostriatal dopaminergic neuron subpopulation**. *J. Clin. Invest.* (2014.0) **124** 3032-3046. DOI: 10.1172/JCI72176
63. Liu G. **Selective expression of Parkinson’s disease-related Leucine-rich repeat kinase 2 G2019S missense mutation in midbrain dopaminergic neurons impairs dopamine release and dopaminergic gene expression**. *Hum. Mol. Genet.* (2015.0) **24** 5299-5312. DOI: 10.1093/hmg/ddv249
64. Holmes C, Eisenhofer G, Goldstein DS. **Improved assay for plasma dihydroxyphenylacetic acid and other catechols using high-performance liquid chromatography with electrochemical detection**. *J. Chromatogr. B Biomed. Appl.* (1994.0) **653** 131-138. DOI: 10.1016/0378-4347(93)E0430-X
65. Yorgason JT, Espana RA, Jones SR. **Demon voltammetry and analysis software: analysis of cocaine-induced alterations in dopamine signaling using multiple kinetic measures**. *J. Neurosci. Methods* (2011.0) **202** 158-164. DOI: 10.1016/j.jneumeth.2011.03.001
66. Axten JM. **Discovery of 7-methyl-5-(1-[3-(trifluoromethyl)phenyl]acetyl-2,3-dihydro-1H-indol-5-yl)-7H-pyrrolo[2,3-d]pyrimidin-4-amine (GSK2606414), a potent and selective first-in-class inhibitor of protein kinase R (PKR)-like endoplasmic reticulum kinase (PERK)**. *J. Med. Chem.* (2012.0) **55** 7193-7207. DOI: 10.1021/jm300713s
67. Morita S. **Targeting ABL-IRE1alpha signaling spares ER-stressed pancreatic beta cells to reverse autoimmune diabetes**. *Cell Metab.* (2017.0) **25** 883-897 e888. DOI: 10.1016/j.cmet.2017.03.018
|
---
title: Cannabidiol improves muscular lipid profile by affecting the expression of
fatty acid transporters and inhibiting de novo lipogenesis
authors:
- Patrycja Bielawiec
- Sylwia Dziemitko
- Karolina Konstantynowicz-Nowicka
- Adrian Chabowski
- Janusz Dzięcioł
- Ewa Harasim-Symbor
journal: Scientific Reports
year: 2023
pmcid: PMC9988888
doi: 10.1038/s41598-023-30872-w
license: CC BY 4.0
---
# Cannabidiol improves muscular lipid profile by affecting the expression of fatty acid transporters and inhibiting de novo lipogenesis
## Abstract
Obesity is one of the principal public health concerns leading to disturbances in glucose and lipid metabolism, which is a risk factor for several chronic diseases, including insulin resistance, type 2 diabetes mellitus, and cardiovascular diseases. In recent years, it turned out that cannabidiol (CBD) is a potential therapeutic agent in the treatment of obesity and its complications. Therefore, in the present study, we used CBD therapy (intraperitoneal injections in a dose of 10 mg/kg of body mass for 14 days) in a rat model of obesity induced by a high-fat diet (HFD). Gas–liquid chromatography and Western blotting were applied in order to determine the intramuscular lipid content and total expression of selected proteins in the white and red gastrocnemius muscle, respectively. Based on fatty acid composition, we calculated de novo lipogenesis ratio (16:$\frac{0}{18}$:2n-6), desaturation ratio (18:1n-$\frac{9}{18}$:0), and elongation ratios (18:$\frac{0}{16}$:0, 20:$\frac{0}{18}$:0, 22:$\frac{0}{20}$:0 and 24:$\frac{0}{22}$:0), in the selected lipid fractions. Two-week CBD administration significantly reduced the intramuscular fatty acids (FAs) accumulation and inhibited de novo lipogenesis in different lipid pools (in the free fatty acid, diacylglycerol, and triacylglycerol fractions) in both muscle types, which coincided with a decrease in the expression of membrane fatty acid transporters (fatty acid translocase, membrane-associated fatty acid binding protein, and fatty acid transport proteins 1 and 4). Moreover, CBD application profoundly improved the elongation and desaturation ratios, which was in line with downregulated expression of enzymes from the family of elongases and desaturases regardless of the metabolism presented by the muscle type. To our knowledge, this study is the first that outlines the novel effects of CBD action on skeletal muscle with different types of metabolism (oxidative vs. glycolytic).
## Introduction
Obesity is one of the most common medical concerns worldwide and its incidence is anticipated to rise alarmingly. The ongoing increase in obesity rates reflects lifestyle changes, particularly changes in nutrition. Although the reasons for its development are well known, including genetic factors, overnutrition, and a sedentary lifestyle, an effective prevention and treatment method is still being sought1. In the course of obesity, increased availability of fatty acids (FAs) in the diet leads to excessive storage of lipids in adipocytes and, subsequently, in other metabolically important tissues such as the liver, cardiac and skeletal muscle2. Several lines of evidence have shown that intramuscular lipid accumulation, especially in the fractions of triacylglycerol (TAG), diacylglycerol (DAG), and ceramide (CER), is due to increased transmembrane transport of long-chain fatty acids (LCFAs)3. Moreover, numerous studies indicate the detrimental effect of some lipid molecules (DAG, CER), occurring in increased amounts intracellularly, on the downstream insulin signal transduction pathway4,5. In turn, this results in dysregulation of glucose and lipid metabolism, and thus disturbed cellular and, subsequently, whole-body energy homeostasis6. A shift in the energy balance is the major cause of metabolic comorbidities, including insulin resistance (IR), type 2 diabetes mellitus (T2DM), and nonalcoholic fatty liver disease (NAFLD), which together account for a large number of obesity-related deaths7,8.
The LCFAs transport across the plasma membrane of cells is of fundamental importance as these compounds perform a variety of functions, among others, they are the building blocks of membranes, fuel for energy supply, and play a role in different signaling pathways9,10. They can be derived directly from the diet or they may be synthesized de novo through lipogenesis11. A large number of studies performed in the last two decades has shown that FAs are transported across the plasma membrane of skeletal muscle not only by passive diffusion but mainly with the use of highly specialized protein transporters including fatty acid translocase (CD36), membrane-associated fatty acid binding protein (FABPpm) and fatty acid transport proteins 1 and 4 (FATP-1,4) (Fig. 9)12,13. Subsequently, inside the myocyte, exogenous FAs derived from the diet along with palmitic acid (16:0), which is the primary end product of de novo lipogenesis, undergo further elongation processes producing a variety of long-chain and very long-chain fatty acids, including saturated fatty acids (SFAs), monounsaturated fatty acids (MUFAs) and polyunsaturated fatty acids (PUFAs)14. The enzymes involved in the elongation process are referred to as the elongation of very long-chain fatty acids proteins (ELOVL)15. There are 7 different elongase subtypes (ELOVL1-7), which show different substrate selectivity and mediate a wide range of specific elongation reactions16. To date, it has been established that ELOVL1,3,6 and 7 are selective for SFA and MUFA, while ELOVL2,4 and 5 preferentially utilize PUFA17. Moreover, functional diversity of lipids is possible not only due to the variability of their chain length but also the degree of their unsaturation, which is conditioned by the acyl-coenzyme A (CoA) desaturases. This family of enzymes introduces a double bond in a specific position on the acyl chain of LCFA, and so far two subgroups of desaturases have been identified, namely stearoyl-coenzyme A (CoA) desaturases (SCDs) and fatty acid desaturases (FADS)14,17. SCD1 is a central lipogenic enzyme that plays a pivotal role in the regulation of MUFA biosynthesis17. It catalyzes the formation of palmitoleate (16:1n-7) and oleate (18:1n-9), from SFA palmitate (16:0) and stearate (18:0), respectively18. However, when it comes to the second subgroup, the ∆-5 (FADS1) and ∆-6 desaturase (FADS2) are the key enzymes in the metabolism of n-3 and n-6 PUFA, enabling the formation of long-chain metabolites from α-linoleic acid (ALA) and linoleic acid (LA)19. Disturbances in the elongation and desaturation process may disrupt the FAs balance and, thereby, proper cellular functioning and metabolic homeostasis, which may be implicated in various diseases related to obesity occurrence20.
For several years, cannabidiol (CBD) has been in the spotlight as a potential therapeutic agent in the treatment of obesity. Among phytocannabinoids isolated from the *Cannabis sativa* plant (more than 120 compounds), CBD exhibits a great safety profile and lack of psychoactive properties21,22. Many studies have demonstrated CBD’s therapeutic potential due to its anti-inflammatory, antioxidant, anxiolytic, anticonvulsant, and neuroprotective properties23,24. In recent years, CBD has been shown to exert its effect on the endogenous system of signaling lipids called the endocannabinoid system (ECS), the definition of which has evolved into the extended ECS or the endocannabinoidome (eCBome)25. The eCBome encompassing endocannabinoids (e.g., N-arachidonoylethanolamine-anandamide (AEA) and 2-arachidonoylglycerol (2-AG)) and numerous long-chain fatty acid-derived congeners, their metabolic enzymes, and the receptors of these lipid compounds, including cannabinoid receptors (such as CB1 and CB2), orphan G protein-coupled receptors (such as GPR55 and GPR18), thermosensitive transient receptor potential (TRP) channels (such as TRPV1), and peroxisome proliferator-activated receptors (PPARs such as PPARα and PPARγ)26,27. So far, numerous studies have shown that CBD exhibits a weak affinity for CB1 and CB2 receptors, whereas it modulates to a much greater extent other molecular targets28,29. Furthermore, it has also been shown that CBD alters the eCBome tone by increasing the concentration of AEA, due to fatty acid amide hydrolase (FAAH) inhibition30. Research in the last decade has significantly increased our knowledge of the complexity of CBD’s action, however, its potential mechanisms for obesity treatment are not yet fully revealed. Therefore, the main purpose of this study was to investigate the influence of CBD on the expression of selected fatty acid transporters (FAT/CD36, FABPpm, and FATP-1,4), the content of various lipid fractions (free fatty acid (FFA), DAG, TAG, and phospholipid (PL)), de novo lipogenesis as well as stearoyl CoA-desaturase (SCD1) activity, and elongation ratios of selected FAs in the mentioned above lipid fractions in the white and red skeletal muscle (musculus gastrocnemius) of rats with obesity induced by a high-fat diet.
## Effect of chronic CBD administration on the total expression of proteins involved in fatty acid uptake in rats subjected to standard and high-fat diets
Our experiment revealed that induction of obesity by high-fat diet feeding significantly raised the levels of both FABPpm and FATP-1 in the white skeletal muscle (+ $46.73\%$ and + $32.40\%$, respectively, $p \leq 0.05$; Fig. 1A) compared to the control group. However, in the red skeletal muscle, levels of all examined transporters were considerably elevated after HFD feeding (CD36: + $23.38\%$, FABPpm: + $28.77\%$, FATP-1: + $49.54\%$, FATP-4: + $21.01\%$, vs. the control group, $p \leq 0.05$; Fig. 1B). Moreover, the intramuscular expression of fatty acid transporters, i.e., CD36, FABPpm as well as FATP-1 and FATP-4, were significantly changed in the group fed HFD after the introduction of CBD (FABPpm: + $44.31\%$, FATP-1: + $19.21\%$, vs. the control group, FATP 1: $9.96\%$, FATP-4: − $29.82\%$, CD36: − $33.11\%$, vs. the HFD group, $p \leq 0.05$; Fig. 1A) in the white skeletal muscle. Concomitantly, CBD injections to animals fed HFD caused a substantial decrease in the expression of each transporter in the red gastrocnemius muscle (CD36: − $40.35\%$, FABPpm: − $18.37\%$, FATP-1: − $32.01\%$, FATP-4: − $16.69\%$, $p \leq 0.05$; Fig. 1B) compared to the HFD group. Figure 1The total expression of proteins involved in fatty acid uptake: fatty acid translocase (CD36), fatty acid binding protein (FABPpm) and fatty acid transport protein 1 and 4 (FATP-1 and -4), in the white and red gastrocnemius muscle. The total expressions of the abovementioned proteins are presented as percentage differences compared to the control group which was set as $100\%$. The data are expressed as mean values ± SD, $$n = 6$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.
## Effect of chronic CBD administration on total free fatty acid, diacylglycerol, triacylglycerol and phospholipid content in rats subjected to standard and high-fat diets
The study presented a significant increase in the pool of FFA in the group fed HFD after two-week CBD treatment (+ $30.46\%$, vs. the control group; + $25.79\%$, vs. the HFD group, $p \leq 0.05$; Fig. 2A) in the white gastrocnemius muscle. DAG fraction was markedly reduced after the administration of CBD (− $13.36\%$, vs. the control group, $p \leq 0.05$; Fig. 2A), whereas in the TAG pool not only the decrease in the same group occurred (− $29.51\%$, vs. the control group, $p \leq 0.05$), but also pronounced changes in the HFD (+ $206.58\%$, vs. the control group, $p \leq 0.05$) and HFD + CBD groups were observed (+ $108.15\%$, vs. the control group; − $32.11\%$, vs. the HFD group, $p \leq 0.05$; Fig. 2A) in the white skeletal muscle. PL fraction was substantially diminished in the groups treated with CBD as well as fed HFD (− $5.19\%$ and − $5.70\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 2A) with simultaneous elevation in the HFD + CBD group (+ $5.09\%$, vs. the control group; + $11.44\%$, vs. the HFD group, $p \leq 0.05$; Fig. 2A) in the white gastrocnemius muscle. In contrast, in the red skeletal muscle the above fraction was increased in CBD, HFD as well as HFD + CBD groups (+ $14.90\%$, + $16.39\%$, + $38.86\%$, respectively, $p \leq 0.05$; Fig. 1B) in comparison with control animals. Moreover, in the same tissue, significant elevation also occurred in the rats fed HFD and treated with CBD (+ $19.31\%$, $p \leq 0.05$; Fig. 2B) compared to rats fed HFD. Additionally, the content of FFA was considerably heightened in all examined groups (CBD: + $20.81\%$, HFD: + $101.02\%$, HFD + CBD: + $81.88\%$, respectively, $p \leq 0.05$; Fig. 1B) compared to the control group, and substantially diminished in the HFD + CBD group (− $9.52\%$, $p \leq 0.05$; Fig. 1B) compared to the HFD group. The levels of DAG and TAG fractions in the red gastrocnemius muscle were markedly elevated in animals fed HFD as well as fed HFD and administered with CBD (DAG: + $61.42\%$, + $51.41\%$, TAG: + $196.71\%$, $130.68\%$, respectively, $p \leq 0.05$; Fig. 2B) in comparison to control animals. In contrast, the content of both fractions was lessened in rats receiving HFD and CBD (DAG: − $6.20\%$ and TAG: − $22.25\%$, $p \leq 0.05$; Fig. 2B) compared to the rats receiving only HFD (see Supplementary File S1 for the total fatty acid composition of the individual lipid fractions in the white and red gastrocnemius muscle).Figure 2Total free fatty acid (FFA), diacylglycerol (DAG), triacylglycerol (TAG) and phospholipid (PL) content in the white and red gastrocnemius muscle. Values are expressed in nmol per mg of tissue. The data are expressed as mean values ± SD, $$n = 10$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.
## Effect of chronic CBD administration on intramuscular de novo lipogenesis ratio in the free fatty acid, diacylglycerol, triacylglycerol and phospholipid fractions in rats subjected to standard and high-fat diets
In the white skeletal muscle, the de novo lipogenesis ratio in the FFA and DAG fractions was elevated in the HFD (+ $21.54\%$, + $20.02\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 3A) but diminished in the CBD group (− $33.48\%$, − $11.80\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 3A) as well as in HFD group administered with CBD (− $33.92\%$, − $13.32\%$ vs. the control group, respectively; − $45.63\%$, − $27.78\%$, vs. the HFD group, respectively, $p \leq 0.05$; Fig. 3A). Interestingly, in the red skeletal muscle significant changes in FFA and DAG pools were observed only in rats receiving a diet rich in fatty acids and CBD injections (FFA: − $10.16\%$, DAG: − $34.57\%$, vs. the control group, respectively; DAG: − $27.14\%$, vs. the HFD group, $p \leq 0.05$; Fig. 3B). The 16:$\frac{0}{18}$:2n-6 ratio in TAG fraction was markedly diminished in each examined group (CBD: -$33.75\%$, HFD: − $47.20\%$, HFD + CBD: − $39.92\%$, vs. the control group, $p \leq 0.05$; Fig. 3A) In the white gastrocnemius muscle, while in the red skeletal muscle it was substantially changed in HFD (+ $56.22\%$, vs. the control group, $p \leq 0.05$) and HFD + CBD groups (+ $31.57\%$, vs. the control group; − $15.78\%$, vs. the HFD group, $p \leq 0.05$; Fig. 3B). Introduction of CBD to animals fed standard and rich in fatty acid diet led to significant changes of 16:$\frac{0}{18}$:2n-6 ratio in PL pool in the red muscle (CBD: − $7.81\%$ vs. the control group, HFD + CBD: − $8.67\%$, vs. HFD group, $p \leq 0.05$; Fig. 3B) as well as in the white muscle (CBD: − $6.64\%$, HFD + CBD: + $40.94\%$ vs. the control group, − $10.39\%$, vs. the HFD group, $p \leq 0.05$; Fig. 3A). Concomitantly, in both types of muscle tissue in the above-mentioned fraction, we observed a substantial rise after induction of obesity by high-fat diet feeding (white muscle: + $57.28\%$, Fig. 3A; red muscle: + $10.22\%$ $p \leq 0.05$; Fig. 3B) compared to the control group. Figure 3Intramuscular de novo lipogenesis in the free fatty acid (FFA), diacylglycerol (DAG), triacylglycerol (TAG) and phospholipid (PL) fractions in the white and red gastrocnemius muscle. The data are expressed as mean values ± SD, $$n = 10$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.
## Effect of chronic CBD administration on intramuscular stearoyl-coenzyme A desaturase 1 level in the free fatty acid, diacylglycerol, triacylglycerol and phospholipid fractions in rats subjected to standard and high-fat diets
As shown in Fig. 4A, we noticed pronounced changes in the 18:1n-$\frac{9}{18}$:0 ratio in FFA (CBD: + $25.66\%$, HFD: + $63.57\%$, HFD + CBD: $69.67\%$, vs. the control group, $p \leq 0.05$), DAG (CBD: − $32.97\%$, HFD: + $40.37\%$, HFD + CBD: − $18.91\%$, vs. the control group, − $42.23\%$, vs. the HFD group, $p \leq 0.05$) and TAG fractions (CBD: − $13.61\%$, HFD: + $42.67\%$, HFD + CBD: + $19.71\%$, vs. the control group, − $16.10\%$, vs. the HFD group, $p \leq 0.05$;), whereas in the PL pool changes occurred only in the group treated with CBD (− $22.16\%$, vs. the control group, $p \leq 0.05$) in white gastrocnemius muscle. In red gastrocnemius muscle, administration of CBD to animals fed standard chow led to the diminishment of the 18:1n-$\frac{9}{18}$:0 ratio only in DAG and PL fractions (− $18.15\%$ and − $12.18\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 4B), however, CBD treatment in animals on high-fat chow led to significant changes in each examined lipid fraction (FFA: + $107.56\%$, vs. the control group, − $28.25\%$, vs. the HFD; DAG: + $110.65\%$, vs. the control group, − $33.99\%$, vs. the HFD; TAG: − $21.08\%$ vs. the HFD; PL: + $22.94\%$, vs. the control group, $p \leq 0.05$; Fig. 4B). In the same type of tissue, we also observed substantial elevation in SCD1 in rodents receiving HFD (FFA: + $189.28\%$; DAG: + $219.12\%$; TAG: + $32.89\%$; PL: + $23.18\%$, $p \leq 0.05$; Fig. 4B) compared to the rodents on the standard diet. Figure 4Intramuscular stearoyl-coenzyme A desaturase 1 (SCD1) activity in the free fatty acid (FFA), diacylglycerol (DAG), triacylglycerol (TAG) and phospholipid (PL) fractions in the white and red gastrocnemius muscle. The data are expressed as mean values ± SD, $$n = 10$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.
## Effect of chronic CBD administration on intramuscular elongation ratios in the free fatty acid, diacylglycerol, triacylglycerol and phospholipid fractions in rats subjected to standard and high-fat diets
Our experiment demonstrated that the elongation 18:$\frac{0}{16}$:0 ratio was significantly altered after the introduction of CBD to rodents on standard and rich in fatty acids diet in FFA (CBD: − $9.14\%$, HFD + CBD: + $7.89\%$, vs. the control group and + $6.30\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5), DAG (CBD: + $8.57\%$, CBD + HFD: + $40.55\%$, vs. the control group and + $33.43\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5), TAG (CBD: + $8.08\%$, HFD + CBD: + $54.10\%$, $p \leq 0.05$, vs. the control group and + $3.0\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5) and PL pools (CBD: + $5.28\%$, HFD + CBD: + $15.66\%$, vs. the control group and + $3.66\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5) in the white skeletal muscle. Additionally, induction of obesity by a diet rich in fats led to a substantial rise in the 18:$\frac{0}{16}$:0 elongation ratio in selected lipid fractions: DAG, TAG as well as PL (+ $5.34\%$, + $49.64\%$ and + $11.58\%$, respectively, $p \leq 0.05$; Fig. 5) compared to the control group in the white skeletal muscle. In the same type of tissue, the elongation 20:$\frac{0}{18}$:0 ratio was significantly changed in examined fractions of FFA (CBD: − $13.14\%$, HFD: + $16.45\%$, HFD + CBD: − $20.93\%$, vs. the control group and − $32.10\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5), DAG (HFD: + $25.08\%$, HFD + CBD: − $29.32\%$, vs. the control group and − $43.49\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5), TAG (CBD: + $38.77\%$, HFD: − $54.92\%$, HFD + CBD: − $52.26\%$, vs. the control group, $p \leq 0.05$; Fig. 5) and PL (HFD + CBD: − $27.22\%$, vs. the control group and − $26.49\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5). We also observed that the elongation 22:$\frac{0}{20}$:0 ratio was considerably heightened in the high-fat diet fed rats (FFA: + $15.93\%$, DAG: + $20.93\%$ and PL: + $54.94\%$, $p \leq 0.05$; Fig. 5) compared to rats fed standard chow in white gastrocnemius muscle. Administration of CBD in the HFD group led to the elevation of the above-mentioned ratio in DAG and PL pools (+ $6.98\%$ and + $19.08\%$, respectively, $p \leq 0.05$; Fig. 5) in comparison to the control, yet it caused a substantial diminishment in the FFA, DAG and PL fractions (− $20.40\%$, − $11.54\%$, − $23.15\%$, respectively, $p \leq 0.05$; Fig. 5) in comparison to the HFD group in white gastrocnemius muscle. Finally, the elongation 24:$\frac{0}{22}$:0 ratio in the white skeletal muscle was markedly altered in 3 out of 4 examined lipid fractions: FFA (CBD: − $27.31\%$, HFD: + $25.38\%$, vs. the control group, HFD + CBD: − $27.13\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5), DAG (HFD: + $26.41\%$, HFD + CBD: − $19.71\%$, vs. the control group and − $36.48\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5) and TAG (HFD: − $20.87\%$, HFD + CBD: − $65.16\%$, vs. the control group and − $55.97\%$, vs. the HFD group, $p \leq 0.05$; Fig. 5). Regarding the red gastrocnemius muscle, the same ratio was substantially increased by prolonged administration of fatty acids in selected lipid fractions: FFA, DAG as well as PL (+ $33.74\%$, + $21.40\%$ and + $42.12\%$, respectively, $p \leq 0.05$; Fig. 6) in comparison with the control group. Concomitantly, the introduction of CBD in the HFD group led to the substantial diminishment of this ratio in FFA, DAG and TAG fractions (− $15.95\%$, -$24.93\%$ and − $24.74\%$, respectively, $p \leq 0.05$; Fig. 6) in comparison with the HFD group. We noticed that elongation 18:$\frac{0}{16}$:0 ratio in the red muscle after administration of CBD to animals on standard and rich in fats chow was heightened in FFA (CBD: + $9.64\%$, HFD + CBD: + $20.34\%$, vs. the control group and + $21.08\%$, vs. the HFD group, $p \leq 0.05$; Fig. 6), DAG (CBD: + $13.58\%$, HFD + CBD: + $44.36\%$, vs. the control group and + $21.62\%$, vs. the HFD group, $p \leq 0.05$; Fig. 6), TAG (CBD: + $11.72\%$, HFD + CBD: + $107.40\%$, vs. the control group and + $3.02\%$, $p \leq 0.05$, vs. the HFD group; Fig. 6) and PL fractions (CBD: + $7.34\%$, HFD + CBD: + $23.32\%$, vs. the control group, $p \leq 0.05$; Fig. 6), whereas lessened only in PL fractions (HFD + CBD: − $12.50\%$, vs. the HFD group, $p \leq 0.05$; Fig. 6). The described elongation ratio was markedly increased after HFD feeding in DAG, TAG as well as PL (+ 18.70, + $101.32\%$ and + $40.93\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 6) pools in the red skeletal muscle. As shown in Fig. 6, in the same type of tissue, the experimental induction of the obesity by HFD altered the elongation 20:$\frac{0}{18}$:0 (FFA: + $17.30\%$, DAG: + $21.08\%$, TAG: − $44.21\%$, vs. the control group, $p \leq 0.05$) as well as 22:$\frac{0}{20}$:0 ratio (FFA: + $16.73\%$, TAG: − $35.79\%$, PL: + $12.46\%$, vs. the control group, $p \leq 0.05$) in examined lipid pools. Moreover, CBD injections to animals fed HFD caused a substantial decrease of both elongation 20:$\frac{0}{18}$:0 and 22:$\frac{0}{20}$:0 ratios in all examined lipid fractions (20:$\frac{0}{18}$:0: FFA: − $11.95\%$, TAG: − $37.19\%$, PL: − $29.12\%$, vs. the control group, PL: − $17.05\%$, DAG: − $26.03\%$, FFA: − 24.93, vs. the HFD group; 22:$\frac{0}{20}$:0: DAG: − $14.75\%$, TAG: − $41.98\%$, vs. the control group, FFA: − $19.70\%$, DAG: − $10.26\%$, PL: − $18.50\%$, vs. the HFD group, $p \leq 0.05$; Fig. 6) in red gastrocnemius muscle. Additionally, the two-week CBD treatment of animals receiving standard chow considerably influenced only the 22:$\frac{0}{20}$:0 ratio in PL fraction (− $16.08\%$, vs. the control group, $p \leq 0.05$; Fig. 6) in the red skeletal muscle. Figure 5Intramuscular 18:$\frac{0}{16}$:0, 20:$\frac{0}{18}$:0, 22:$\frac{0}{20}$:0 and 24:$\frac{0}{22}$:0 elongation ratios in the free fatty acid (FFA), diacylglycerol (DAG), triacylglycerol (TAG) and phospholipid (PL) fractions in the white gastrocnemius muscle. The data are expressed as mean values ± SD, $$n = 10$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.Figure 6Intramuscular 18:$\frac{0}{16}$:0, 20:$\frac{0}{18}$:0, 22:$\frac{0}{20}$:0 and 24:$\frac{0}{22}$:0 elongation ratios in the free fatty acid (FFA), diacylglycerol (DAG), triacylglycerol (TAG) and phospholipid (PL) fractions in the red gastrocnemius muscle. The data are expressed as mean values ± SD, $$n = 10$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.
## Effect of chronic CBD administration on the total expression of proteins involved in fatty acid synthesis and metabolism in rats subjected to standard and high-fat diets
In our experiment, we revealed that a high-fat chow administration caused a significant increase in the total expression of sterol regulatory element-binding protein 1 (SREBP1) in the white gastrocnemius muscle (+ $53.74\%$, $p \leq 0.05$, vs. the control group; Fig. 7), which was further reduced by the CBD implementation (− $10.82\%$, $p \leq 0.05$, vs. the HFD group; Fig. 7). On the contrary, in the red skeletal muscle, induction of obesity by high-fat feeding resulted in a slight downregulation of SREBP1 expression in comparison with the control rats. However, two-week CBD treatment substantially decreased the SREBP1 expression in standard chow-fed control rats (− $15.66\%$, $p \leq 0.05$, vs. the control group) and in the lipid oversupply conditions in relation to the untreated control and HFD group alone (− $27.26\%$ and − $21.23\%$, $p \leq 0.05$, respectively; Fig. 8). As shown in Fig. 7, in the white muscle homogenates, we observed decreased fatty acid synthase (FAS) expression after the introduction of CBD to rats on diet rich in fatty acids (− $24.13\%$, $p \leq 0.05$, vs. the HFD group). Moreover, induction of obesity by high-fat diet feeding markedly altered the expression of each examined isoform of ELOVL, namely ELOVL1, ELOVL3, and ELOVL6 (+ $65.04\%$, + $36.19\%$, − $37.34\%$ and + $65.81\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 7) in the white skeletal tissue. Additionally, CBD injections to animals on high-fat chow resulted in an increase in ELOVL1 expression in the white skeletal muscle (+ $138.15\%$, $p \leq 0.05$; Fig. 7) compared to the group receiving standard chow, whereas the content of ELOVL3 and ELOVL6 was substantially diminished in the described group (− $23.80\%$ and − $40.04\%$, respectively, $p \leq 0.05$; Fig. 7) in comparison with the HFD group. Concomitantly, in red skeletal muscle induction of obesity by a diet rich in fatty acids significantly enhanced only ELOVL1 and ELOVL6 (+ $33.09\%$, + $35.54\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 8). Additionally, administration of CBD to the HFD group only diminished the expression of ELOVL6 isoform (− $32.98\%$, vs. the HFD group, $p \leq 0.05$; Fig. 8) in the red skeletal muscle. Concomitantly, we observed elevated intramuscular expression of FADS1 only in animals receiving HFD and treated with CBD (+ $24.81\%$, vs. the control group; + $31.92\%$, vs. the HFD, $p \leq 0.05$; Fig. 7) in the white gastrocnemius muscle, whereas FADS2 expression in the same type of tissue was increased in CBD group (+ $28.47\%$, vs. the control group, $p \leq 0.05$; Fig. 7) and HFD group (+ $40.47\%$, vs. the control group, $p \leq 0.05$; Fig. 7), yet diminished in HFD + CBD group (− $16.45\%$, vs. the HFD group, $p \leq 0.05$; Fig. 7). In contrast, in the red skeletal muscle, we did not notice any pronounced changes in total FADS2 expression, however, induction of obesity by high-fat chow feeding significantly heightened intramuscular expression of FADS1 as well as FAS (+ $54.06\%$ and + $34.12\%$, vs. the control group, respectively, $p \leq 0.05$; Fig. 8). Importantly, CBD treatment during the course of high-fat feeding considerably reduced FADS1 and FAS expression (− $35.38\%$ and − $26.74\%$, vs. the HFD group, respectively, $p \leq 0.05$; Fig. 8) in the red gastrocnemius muscle. Figure 7The total expression of proteins involved in the fatty acid synthesis and metabolism: sterol regulatory element-binding protein 1 (SREBP1), fatty acid synthase (FAS), fatty acid desaturase 1 and 2 (FADS1 and FADS2) as well as fatty acid elongase 1, 3, and 6 (ELOVL1, ELOVL3, and ELOVL6) in the white gastrocnemius muscle. The total expressions of the abovementioned proteins are presented as percentage differences compared to the control group which was set as $100\%$. The data are expressed as mean values ± SD, $$n = 6$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.Figure 8The total expression of proteins involved in the fatty acid synthesis and metabolism: sterol regulatory element-binding protein 1 (SREBP1), fatty acid synthase (FAS), fatty acid desaturase 1 and 2 (FADS1 and FADS2) as well as fatty acid elongase 1, 3, and 6 (ELOVL1, ELOVL3, and ELOVL6) in the red gastrocnemius muscle. The total expressions of the abovementioned proteins are presented as percentage differences compared to the control group which was set as $100\%$. The data are expressed as mean values ± SD, $$n = 6$$ in each group; ap < 0.05 indicates a significant difference: the control group vs. the examined group; bp < 0.05 indicates a significant difference: HFD vs. HFD + CBD.
## Discussion
In this study, we demonstrated for the first time that CBD, a non-psychotropic plant-origin cannabinoid, reduces the expression of LCFA transport proteins and intramuscular lipid accumulation with simultaneous inhibition of de novo lipogenesis in skeletal muscle of rats in the course of obesity induced by a high-fat diet. Moreover, we have shown that CBD under these conditions is a modulator of the elongation and desaturation processes, which positively influenced FAs metabolism.
In our study, as we expected, the induction of obesity by feeding rats a HFD significantly increased the expression of all examined LCFAs transporters (CD36, FABPpm, FATP-1, and FATP-4) in the red gastrocnemius muscle, while in the white skeletal muscle only expression of FABPpm and FATP-1 was enhanced. These differences are probably related to the metabolism presented by these types of muscles, where white skeletal muscle exhibits glycolytic metabolism and red skeletal muscle exhibits oxidative metabolism31. In addition, we found that CBD appears to function as a regulator of the expression of LCFAs transport proteins, as its two-week administration markedly attenuated the total expression of specific LCFAs transporters, i.e., CD36, FABPpm, FATP-1, and FATP-4 in both types of skeletal muscles. In line with that, we observed similar effects of CBD action on the expression of fatty acid protein carriers in the myocardial tissue, which we described in our previous work32. It is worth emphasizing, that CBD had the greatest effect on the muscular expression of CD36, which is believed to play a pivotal role in maintaining cellular lipid homeostasis12,33. Numerous studies have shown that complete deletion or downregulation of CD36 expression significantly reduces the uptake of LCFAs, thereby causing changes in fatty acid metabolism, especially concerning fatty acid oxidation, which is important in the working heart and skeletal muscles34. Other researchers have also shown that the overexpression of CD36 caused by lipid oversupply, increases the influx of LCFAs into the cell, exceeding the oxidative capacity of mitochondria, which in turn results in the storage of excessive amounts of FAs in the TAG fraction and its further conversion to more metabolically active lipid metabolites (i.e., DAG, CER), which are known to inhibit the insulin signaling pathway, promoting the development of IR35. Furthermore, recent studies have shown the unexpected properties of CD36 to promote de novo lipogenesis in hepatocytes by regulating SREBP1 processing36. Zeng et al. found that CD36 directly interacted with an insulin-induced gene (INSIG) to attenuate its inhibitory effects on SREBP1 proteolysis. Increased processing of SREBP1 upregulated the expression of de novo lipogenesis enzymes such as acetyl-CoA carboxylase α (ACCα) and FAS, which resulted in elevated hepatic lipid accumulation36. Considering the above, CBD downregulates protein-mediated LCFAs transport in myocytes of obese rats, which is desired therapeutic effect in future clinical studies.
It is well known that in the course of obesity there is the deposition of lipids in non-adipose tissues such as skeletal muscle9. In our research, we confirmed it by noting a significant increase in intramuscular accumulation of FAs in various lipid fractions including FFA, DAG, TAG, and PL in both muscle types in fatty acids oversupply conditions, which is in line with a significant upregulation of LCFAs transport proteins expression in either white and red skeletal muscle. However, we have observed a much greater FAs accumulation in the red gastrocnemius muscle in all examined lipid fractions. It is consistent with the metabolism shown by this type of muscle tissue, since the red muscle fibers use FAs as the primary source of energy in the oxidation process, while white skeletal muscles obtain energy through the glycolytic pathway31. Moreover, it is also associated with a more pronounced expression of membrane LCFAs transport proteins in the red skeletal muscle31. Particular attention should be paid to the increased accumulation of lipids not only in the TAG fraction but also in the fraction of DAG, which was observed in our study in the red skeletal muscle. DAG is one of the prime candidates among the lipid derivatives responsible for the development of IR. Many studies have shown that it disrupts the insulin signaling pathway by activating both classic and atypical protein kinase C (PKC), which dephosphorylates insulin receptor substrate 1 (IRS1) and thus blocks further downstream insulin signaling, thereby reducing insulin-stimulated glucose uptake37,38. In contrast, TAG is considered to be a relatively safe fraction that does not exhibit intracellular lipotoxic activity, because both the activation and inhibition of TAG intramuscular hydrolysis do not significantly affect the mitochondrial function and cellular insulin sensitivity39. Accordingly, the herein study demonstrated that in animals receiving a HFD after a two-week administration of CBD, we noticed a remarkable decrease in TAG content in the white gastrocnemius muscle, whereas in the red skeletal muscle CBD injections significantly reduced lipid accumulation in FFA, DAG, and TAG fractions. This is in line with the results obtained by Silvestri et al.22 where they showed that CBD dose-dependently inhibits TAG accumulation in hepatocytes and 3T3-L1 adipocytes treated with oleic acid (OA). Interestingly, only in the case of the PL fraction, we observed that chronic CBD treatment caused a pronounced increase in the content of accumulated lipids in the rats fed a HFD in both white and red skeletal muscle. We can hypothesize that CBD enhances the incorporation of FAs into the PL fraction and significantly influences the composition of the above-mentioned fraction in favor of n-3 PUFA, as we demonstrated in our previous work40. Concomitantly, in our experiment, we reported intensified muscular de novo lipogenesis in rats subjected to high-fat feeding in both white and red gastrocnemius muscle as indicated by increased lipogenic indexes. De novo lipogenesis is a complex metabolic pathway in which excess carbohydrates are converted into FAs, which further are esterified and stored as TAG fraction41.This also underlines the importance of de novo lipogenesis process in the skeletal muscle, especially in the state of IR related to obesity42.
Furthermore, recent research has shown the relationship between ECS and FAs metabolism. Osei-Hyiaman et al.43 demonstrated that hepatic CB1 activation results in enhanced de novo fatty acid synthesis via the induction of the lipogenic transcription factor SREBP-1c and its target enzymes including FAS. Interestingly, in our previous research6, we revealed that CBD downregulated muscle CB1 receptor expression, however, we can assume that CBD in skeletal muscle affects FAS expression through a different molecular mechanism than in the liver, due to the fact that both tissues differ significantly in terms of metabolic processes occurring in them. In addition, CB1 has also been shown to be associated with SCD1 activity. A study conducted by Liu et al. revealed that various overlapping mechanisms contribute to the development of IR, more specifically stimulation of the ECS by CB1 receptor activation in the course of high-fat diet-induced obesity promotes the activity of SCD1, an enzyme that catalyzes the biosynthesis of MUFA44,45. The same study also revealed that endogenous MUFAs (palmitoleic and oleic acid) act as fatty acid amide hydrolase (FAAH) inhibitors leading to a reduction in AEA degradation and an elevation in its concentration, which stimulates the ECS. Furthermore, elevated levels of MUFA activate CB1 receptors that promote de novo lipogenesis by inducing the expression of lipogenic genes, including SCD1, forming a positive feedback loop45. This is in accordance with our studies, which also showed a pronounced upregulation of SCD1 activity in rats fed a HFD in all examined lipid fractions in the red and white gastrocnemius muscle, only with the exception of the PL fraction in the glycolytic type of muscle. The two-week administration of CBD significantly reduced the level of SCD1 in the DAG and TAG fractions in the white muscle, again affecting the red muscle to a greater extent, where it decreased SCD1 activity in the FFA, DAG, and TAG lipid pools. Based on the above data, we may suspect that the effects of CBD are related to the inhibition of the endocannabinoids action at the CB1 receptor level since CBD has been well-described as a negative allosteric modulator of this cannabinoid receptor46. Additionally, in our experimental model, we examined the muscle expression of fatty acid desaturase 1 (FADS1; also known as delta-5 desaturase (∆5D)) and fatty acid desaturase 2 (FADS2; referred as delta-6 desaturase (∆6D)), enzymes involved in the double bond introduction at delta ∆5 and ∆6 positions of PUFA chain, respectively19. Our analysis, after treatment of HFD-fed rats with CBD, indicated substantially decreased expression of FADS1 in the red gastrocnemius muscle and FADS2 in the white skeletal muscle. We believe that this may be considered as a cell protection mechanism since Yashiro et al. reported that FADS1 inhibition alleviates IR and reduce body weight in a HFD-induced obese (DIO) mouse model47. Moreover, FADS1 catalyzes the production of arachidonic acid (20:4, AA), which is a precursor of pro-inflammatory eicosanoids, therefore, its accumulation promotes the synthesis of AA-derived mediators that contribute to the development of chronic inflammation related to obesity48. Similarly, a growing number of studies demonstrated a positive correlation between increased FADS2 expression and IR as well as TD2M occurrence, whereas FADS2 inhibition resulted in resistance to HFD-induced obesity in a mouse model49,50. Intriguingly, in the white gastrocnemius muscle we also noticed a substantial rise in the FADS1 expression in the HFD + CBD group. This is in line with the results from our previous publication, where we discovered that the most pronounced changes considering the AA and eicosapentaenoic acid (EPA) content (the synthesis of which is controlled by FADS1), were observed in the same experimental group in different lipid fractions in the white skeletal muscle40. However, a stronger protective effect of CBD treatment was observed in the red gastrocnemius muscle, where it caused a significant shift in the n-6/n-3 PUFA balance towards the anti-inflammatory n-3 PUFA in examined lipid fractions, which we can associate with the metabolism exhibited by this muscle type. This taken altogether, can lead to the presumption that CBD in conditions of elevated FAs consumption inhibits de novo fatty acids synthesis and reduces their intracellular accumulation in the skeletal muscle in obesity.
In addition, we examined for the first time the effect of CBD action on the expression of the elongase family members, where we revealed substantial alternations in the elongation process in both white and red skeletal muscle. Overall, LCFAs supplied with the diet and formed during de novo lipogenesis (mainly 16:0; palmitic acid) can be elongated by adding two carbon units to the carboxyl end of LCFAs, using malonyl-CoA and fatty acyl-CoA as substrates. This process is catalyzed by the ELOVLs (Fig. 9). Our findings indicate that a HFD increases elongation of 16∶0 to stearic acid (18∶0) in the DAG, TAG, and PL lipid fractions in the skeletal muscle of rats regardless of the “presented metabolism”. Intriguingly, CBD treatment in high-fat feeding markedly elevated the 18:$\frac{0}{16}$:0 FA elongation ratio in the following fractions: FFA, DAG, and TAG in both muscle types. We also observed similar effects of CBD administration in rats subjected to the standard diet, where it enhanced a ratio of 18:$\frac{0}{16}$:0 elongation in the DAG, TAG, and PL fractions in the white gastrocnemius muscle, and in all examined lipid fractions in the red gastrocnemius muscle. Possibly, increased elongation of 16:0 by treatment with CBD can help prevent a buildup of 16∶0 in favor of 18∶0 in the skeletal muscle of obese rats, whichcan be considered as a cell protection mechanism. Accordingly, the increased elongation of palmitic acid to stearic acid reflects a significant rise in the expression of the ELOVL6 enzyme in rats fed a HFD in both muscle types, which was subsequently diminished by CBD. This is worth emphasizing since Matsuzaka et al.51 showed that mice with *Elovl6* gene deletion were resistant to the development of diet-induced IR despite the presence of obesity, suggesting that inhibition of elongase 6 may represent a novel approach to treat IR. In the case of further elongation of the FAs chain, our results show that the supply of a HFD to rats significantly increased the elongation ratio of stearic acid (18:0) to arachidic acid (20:0), arachidic acid to behenic acid (22:0) and behenic acid to lignoceric acid (24:0) in the FFA, DAG and PL lipid pools in either muscle with glycolytic and oxidative metabolism, whereas injections with CBD caused a significant reduction of the above-mentioned indexes. In addition, our research showed that the TAG fraction was affected to a lesser extent, and surprisingly chronic exposure to a HFD resulted in a decrease in the 20:$\frac{0}{18}$:0 and 24:$\frac{0}{22}$:0 indexes in the white gastrocnemius muscle, while in the red skeletal muscle, 20:$\frac{0}{18}$:0 and 22:$\frac{0}{20}$:0 ratios were reduced. However, the implementation of CBD treatment significantly diminished the elongation of 22:0 to 24:0 in both muscle types as well. Moreover, in our study, we investigated the muscular expressions of ELOVL1 and 3, the enzymes that elongate mostly SFA and MUFA providing precursors for the synthesis of sphingolipids52. The total expression of ELOVL1 and ELOVL3 was significantly enhanced in both white and red gastrocnemius muscles of the rats subjected to a HFD, which is consistent with results obtained by Kozawa et al.53, and coincides with the increase in elongation ratios determined in our study (18:$\frac{0}{16}$:0, 20:$\frac{0}{18}$:0, 22:$\frac{0}{20}$:0 and 24:$\frac{0}{22}$:0). Although, in the oxidative fibers, we did observe only a trend towards downregulation in the muscular expression of elongase 1 and 3 in the HFD + CBD group, while in the case of anaerobic fibers, two-week CBD administration significantly reduced total ELOVL3 expression and surprisingly increased the expression of ELOVL1 in the same group of animals. Collectively, these findings implicate that lipid oversupply to skeletal muscle, results in increased elongation of SFA chains, which, take part in further metabolic transformations, and intensify the accumulation of compounds such as CER and DAG, since most SFAs in mammals are found to be components of sphingolipids family52. These active biomolecules, as indicated by a substantial body of evidence, impair the insulin-signaling cascade by inhibiting insulin-stimulated phosphorylation of protein kinase B (Akt) and glycogen synthase kinase 3 (GSK3β), and subsequent GLUT4 translocation from the intracellular compartments to the plasma membrane, which partially was also showed in our previous work6,54. Collectively, the interactions between the above processes are quite complex, and examined enzymes are involved in different synthetic pathways that can alter the levels of various LCFAs pools in response to distinct factors, which is the basis for further research examining the exact mechanisms of CBD action. Figure 9Effects of a high-fat diet (HFD) and two-week cannabidiol (CBD) administration on the fatty acid transport proteins expression, de novo lipogenesis, elongation and desaturation processes in rat myocytes. ↑, increase; ↓, decrease; red arrow inside the circle indicates the effects of seven weeks of high-fat diet feeding; green arrow inside the circle indicates the effects of two weeks of CBD treatment in HFD-fed rats; long-chain fatty acids (LCFA); fatty acid transport protein 1 and 4 (FATP-1 and -4); cytoplasmic and membrane-associated fatty acid binding protein (FABPc, FABPpm); fatty acid translocase (CD36); palmitic acid (C16:0); palmitoleic acid (C16:1n-7); stearic acid (C18:0); oleic acid (C18:1n-9); arachidic acid (C20:0); behenic acid (C22:0); lignoceric acid (C24:0); ATP citrate lyase (ACL); acetyl-CoA carboxylase (ACC); fatty acid synthase (FAS); fatty acid desaturase 2 (FADS2); stearoyl-coenzyme A (SCD1); elongation of very long-chain fatty acids proteins 1,3,6 and 7 (ELOVL1,3,6 and 7); glycerol-3-phosphate acyltransferase (GPAT); lysophosphatidic acid (LPA); 1-acylglycerol-3-phosphate-O-acyltransferase (AGPAT); phosphatidic acid (PA); phosphatidic acid phosphatase (PAP); diacylglycerol (DAG); diglyceride acyltransferase (DGAT); triacylglycerol (TAG). Figure was created in BioRender by the authors.
## Conclusions
Taken altogether, the study presented herein has demonstrated the novel effects of CBD action as a potential drug to combat obesity and its complications. Under normal conditions, there appear to be closely coordinated regulations between LCFAs intracellular transport, de novo lipogenesis, elongation, and desaturation. Whereas the excess delivery of LCFA significantly affects these processes, leading to dysregulation of lipid metabolism, which contributes to the development of IR. However, the results obtained here indicate that CBD is involved in the regulation of intramuscular FAs uptake and thus reduces the accumulation of lipids in the skeletal muscles and has a positive effect on their further metabolic fate. Hence, CBD properties appear to be an attractive therapeutic strategy for obesity treatment.
## Animals and experimental protocol
All experimental procedures were conducted on male Wistar rats, initially weighing 70–100 g. The animals were delivered from the Center for Experimental Medicine of the Medical University of Bialystok, Poland. Rats were kept in approved animal holding facilities (22 ± 2 °C with a reverse light–dark cycle of $\frac{12}{12}$ h) and had free access to water and standard rodent chow (Labofeed B, Animal Feed Manufacturer “Morawski”, Kcynia, Poland). After seven days of acclimatization, rodents were randomly divided into four experimental groups, each consisting of 10 animals: [1] control group—receiving a basal rodent diet (12.4 kcal% fat, 57.1 kcal% carbohydrates, and 30.5 kcal% protein), [2] CBD group—receiving a basal rodent diet and treated with CBD, [3] HFD group—receiving a rodent diet rich in fatty acids (60 kcal% fat, 20 kcal% carbohydrates, and 20 kcal% protein (cat. no.: D12492, Research Diets Inc., New Brunswick, NJ, USA55)) and [4] HFD + CBD group—receiving a rodent diet rich in fatty acids and treated with CBD (Table 1). According to the described groups, animals were fed an appropriate diet for seven weeks. Simultaneously, intraperitoneal (i.p.) injections of CBD or its vehicle were given to the animals, starting from the fifth week. Daily injections with synthetic CBD (10 mg/kg, purity ≥ $99\%$; THC Pharm GmbH, Frankfurt, Germany) or its solvent for control and HFD groups (3:1:16, ethanol, Tween-80, and $0.9\%$ NaCl) lasted for two weeks56,57. Then, at the end of the 7th week of the experiment and 24 h after the last dose of CBD or its vehicle, rats from all experimental groups were anesthetized i.p. with pentobarbital (80 mg/kg). Additionally, the body weight of each animal was monitored during the whole experiment and we reported substantially increased body mass in rats subjected to the high fat feeding, however two-week CBD treatment did not significantly affect the body weight in rats fed both the standard chow and HFD (the results were published in our previous publication32). Muscle samples (musculus gastrocnemius) were excised, then immediately frozen with aluminum tongs precooled in liquid nitrogen and stored at − 80 °C for subsequent analysis. To obtain plasma samples, a puncture through the inferior vena cava was performed, blood was collected to heparinized test tubes and centrifuged. The experimental protocol was approved by the local Animal Ethics Committee in Olsztyn under license number $\frac{71}{2018}$ and all methods were carried out in accordance with relevant guidelines and regulations. Methods are reported in accordance with ARRIVE guidelines. Table 1The nutritional composition of the high-fat diet (HFD).Class descriptionIngredientHFD (g)ProteinCasein, lactic, 30 meshCystine, L200.003.00CarbohydrateLodex 10Sucrose, fine granulated125.0072.80FiberSolka floc, FCC20050.00FatSoybean oil, USPLard25.00245.00MineralS10026B50.00VitaminCholine bitartrateV10001C2.001.00
## Skeletal muscle lipid analysis
Gas–liquid chromatography (GLC) was performed to analyze intramuscular lipid fractions, namely DAG, TAG, PL, and FFA fractions. The pulverization of frozen muscle samples in mortar precooled in liquid nitrogen was followed by overnight extraction in chloroform–methanol solution (2:1, vol/vol) according to the method of Folch58, with the addition of butylated hydroxytoluene as an antioxidant and heptadecanoic acid as an internal standard. Afterward, the samples were centrifuged and the lower layer was collected for subsequent analysis. The above-mentioned lipid fractions were separated by thin-layer chromatography (TLC) on silica gel plates (Silica Plate 60, 0.25 mm; Merck, Darmstadt, Germany), using a heptane/isopropyl ether/acetic acid (60:40:3, vol/vol/vol) as a resolving solution. Visualization of dried silica plates under ultraviolet light enabled the identification of target lipid fractions. Thereafter, gel bands corresponding to selected lipid fractions were scrapped and eluted. DAG, TAG, PL, and FFA fractions were eluted in appropriate solutions and the organic phase was transmethylated in a $14\%$ boron trifluoride-methanol (BF3) solution. Samples with the addition of hexane were examined by a Hewlett Packard 5890 Series II Gas Chromatograph (Agilent Technologies, CA, USA) containing a capillary column (50 m × 0.25 mm inner diameter) and a flame ionization detector—HP-INNOWax. Individual fatty acids in each fraction were identified. Based on a sum of the particular fatty acid species content in each target fraction, the concentration of total DAG, TAG, PL, and FFA was calculated and expressed in nanomoles per gram of tissue. The de novo lipogenesis ratio was calculated as palmitic/linoleic acid (16:$\frac{0}{18}$:2n-6) ratio; SCD1 was measured as oleic/stearic acid (18:1n-$\frac{9}{18}$:0) ratio; elongation was estimated as stearic/palmitic acid (18:$\frac{0}{16}$:0) ratio, arachidic/stearic acid (20:$\frac{0}{18}$:0) ratio, behenic/arachidic acid (22:$\frac{0}{20}$:0) ratio as well as lignoceric/behenic acid (24:$\frac{0}{22}$:0) ratio.
## Western blotting
To examine selected protein expression, a Western Blotting procedure was performed. Obtained muscle samples were homogenized in radioimmunoprecipitation assay (RIPA) buffer containing a cocktail of protease and phosphatase inhibitors (Roche Diagnostics GmbH, Manheim, Germany). Then, the bicinchoninic acid method (BCA) with bovine serum albumin (BSA) as a standard was performed in order to measure total protein concentration. Thereafter, the homogenates were reconstituted in Laemmli buffer (Bio-Rad, Hercules, CA, USA) and applied on CriterionTM TGX Stain-Free precast gels (Bio-Rad, Hercules, CA, USA). After the electrophoresis, the separated proteins were transferred onto polyvinylidene fluoride (PVDF) (semi-dry transfer) or nitrocellulose membranes (wet transfer). The next step was blocking membranes in the Tris-buffered saline with Tween-20 (TBST) and $5\%$ non-fat dry milk or $5\%$ BSA, subsequently, the membranes were incubated overnight at 4 °C with selected primary antibodies: CD36 (sc-7309, Santa Cruz Biotechnology, Inc., Dallas, TX, USA), FABPpm (ab180162, Abcam, Cambridge, United Kingdom), FATP-1 (sc-25541, Santa Cruz Biotechnology, Inc., Dallas, TX, USA), FATP-4 (sc-5834, Santa Cruz Biotechnology, Inc., Dallas, TX, USA), SREBP1 (sc-367, Santa Cruz Biotechnology, Inc., Dallas, TX, USA), FAS (3180S, Cell Signalling Technology, Danvers, MA, USA), FADS1 (ab126706, Abcam, Cambridge, United Kingdom), FADS2 (ab232898, Abcam, Cambridge, United Kingdom), ELOVL1 (ab230634, Abcam, Cambridge, United Kingdom), ELOVL3 (sc-54878, Santa Cruz Biotechnology, Inc., Dallas, TX, USA) and ELOVL6 (sc-385127, Santa Cruz Biotechnology, Inc., Dallas, TX, USA). The membranes were then incubated with a secondary antibody conjugated with horseradish peroxidase (HRP) (Cell Signaling Technology, Danvers, MA, USA). The addition of the appropriate substrate for HRP (Clarity Western ECL Substrate; Bio-Rad, Hercules, CA, USA) was followed by the visualization of protein bands using a ChemiDoc visualization system (Image Laboratory Software Version 6.0.1; Bio-Rad, Warsaw, Poland). Stain-free gels and the total protein normalization method (Bio-Rad, Hercules, CA, USA) were applied to quantify the expression of the examined proteins (see Supplementary File S2). The total expressions of the abovementioned proteins were presented as percentage differences compared to the control group which was set as $100\%$, and are based on six independent determinations.
## Statistical analysis
All data obtained from the experiment are expressed as mean values ± standard deviation (SD). Statistical analysis was performed with the use of GraphPad Prism version 7.0 for Windows (GraphPad Software, La Jolla, CA, USA). The normality of the result distribution was checked using the Shapiro–Wilk test and the homogeneity of the variance with the use of Bartlett’s test. Then, a two-way test ANOVA and an appropriate post hoc test were carried out to indicate statistical differences. Values $p \leq 0.05$ were considered significant for all results.
## Supplementary Information
Supplementary Information 1.Supplementary Information 2. The online version contains supplementary material available at 10.1038/s41598-023-30872-w.
## References
1. Barazzoni R, Gortan Cappellari G, Ragni M, Nisoli E. **Insulin resistance in obesity: An overview of fundamental alterations**. *Eat. Weight Disord. EWD* (2018) **23** 149-157. DOI: 10.1007/s40519-018-0481-6
2. Samuel VT, Petersen KF, Shulman GI. **Lipid-induced insulin resistance: unravelling the mechanism**. *Lancet* (2010) **375** 2267-2277. DOI: 10.1016/S0140-6736(10)60408-4
3. Lopaschuk GD. **Fatty Acid Oxidation and Its Relation with Insulin Resistance and Associated Disorders**. *Ann. Nutr. Metab.* (2016) **68** 15-20. DOI: 10.1159/000448357
4. Consitt LA, Bell JA, Houmard JA. **Intramuscular lipid metabolism, insulin action, and obesity**. *IUBMB Life* (2009) **61** 47-55. DOI: 10.1002/iub.142
5. Yu C. **Mechanism by which fatty acids inhibit insulin activation of insulin receptor substrate-1 (IRS-1)-associated phosphatidylinositol 3-kinase activity in muscle**. *J. Biol. Chem.* (2002) **277** 50230-50236. DOI: 10.1074/jbc.M200958200
6. Bielawiec P, Harasim-symbor E, Konstantynowicz-nowicka K. **Chronic cannabidiol administration attenuates skeletal muscle de novo ceramide synthesis pathway and related metabolic effects in a rat model of high-fat diet-induced obesity**. *Biomolecules* (2020) **10** 1241. DOI: 10.3390/biom10091241
7. Koutaki D, Michos A, Bacopoulou F, Charmandari E. **The emerging role of Sfrp5 and Wnt5a in the pathogenesis of obesity: Implications for a healthy diet and lifestyle**. *Nutrients* (2021) **13** 2459. DOI: 10.3390/nu13072459
8. Kahn SE, Hull RL, Utzschneider KM. **Mechanisms linking obesity to insulin resistance and type 2 diabetes**. *Nature* (2006) **444** 840-846. DOI: 10.1038/nature05482
9. Morales PE, Bucarey JL, Espinosa A. **Muscle lipid metabolism: Role of lipid droplets and perilipins**. *J. Diabetes Res.* (2017) **2017** 1789395. DOI: 10.1155/2017/1789395
10. Glatz JFC, Luiken JJFP. **Time for a détente in the war on the mechanism of cellular fatty acid uptake**. *J. Lipid Res.* (2020) **61** 1300-1303. DOI: 10.1194/jlr.6192020LTE
11. Jeppesen J, Kiens B. **Regulation and limitations to fatty acid oxidation during exercise**. *J. Physiol.* (2012) **590** 1059-1068. DOI: 10.1113/jphysiol.2011.225011
12. Glatz JFC, Luiken J. **From fat to FAT (CD36/SR-B2): Understanding the regulation of cellular fatty acid uptake**. *Biochimie* (2017) **136** 21-26. DOI: 10.1016/j.biochi.2016.12.007
13. Schwenk RW, Holloway GP, Luiken JJFP, Bonen A, Glatz JFC. **Fatty acid transport across the cell membrane: Regulation by fatty acid transporters**. *Prostaglandins Leukot. Essent. Fatty Acids* (2010) **82** 149-154. DOI: 10.1016/j.plefa.2010.02.029
14. Bond LM, Miyazaki M, O’Neill LM, Ding F, Ntambi JM, Ridgway ND, McLeod RS. **Fatty acid desaturation and elongation in mammals**. *Biochemistry of Lipids, Lipoproteins and Membranes* (2016) 185-208
15. Deák F, Anderson RE, Fessler JL, Sherry DM. **Novel cellular functions of very long chain-fatty acids: Insight from ELOVL4 mutations**. *Front. Cell. Neurosci.* (2019) **13** 428. DOI: 10.3389/fncel.2019.00428
16. Jump DB. **Mammalian fatty acid elongases**. *Methods Mol. Biol.* (2009) **579** 375-389. DOI: 10.1007/978-1-60761-322-0_19
17. Guillou H, Zadravec D, Martin PGP, Jacobsson A. **The key roles of elongases and desaturases in mammalian fatty acid metabolism: Insights from transgenic mice**. *Prog. Lipid Res.* (2010) **49** 186-199. DOI: 10.1016/j.plipres.2009.12.002
18. Ralston JC, Badoud F, Cattrysse B, McNicholas PD, Mutch DM. **Inhibition of stearoyl-CoA desaturase-1 in differentiating 3T3-L1 preadipocytes upregulates elongase 6 and downregulates genes affecting triacylglycerol synthesis**. *Int. J. Obes.* (2014) **38** 1449-1456. DOI: 10.1038/ijo.2014.35
19. Tosi F, Sartori F, Guarini P, Olivieri O, Martinelli N. **Delta-5 and delta-6 desaturases: Crucial enzymes in polyunsaturated fatty acid-related pathways with pleiotropic influences in health and disease**. *Adv. Exp. Med. Biol.* (2014) **824** 61-81. DOI: 10.1007/978-3-319-07320-0_7
20. Flowers MT, Ntambi JM. **Role of stearoyl-coenzyme A desaturase in regulating lipid metabolism**. *Curr. Opin. Lipidol.* (2008) **19** 248-256. DOI: 10.1097/MOL.0b013e3282f9b54d
21. Turner SE, Williams CM, Iversen L, Whalley BJ, Kinghorn AD, Falk H, Gibbons S, Kobayashi J. **Molecular pharmacology of phytocannabinoids**. *Phytocannabinoids: Unraveling the Complex Chemistry and Pharmacology of Cannabis sativa* (2017) 61-101
22. Silvestri C. **Two non-psychoactive cannabinoids reduce intracellular lipid levels and inhibit hepatosteatosis**. *J. Hepatol.* (2015) **62** 1382-1390. DOI: 10.1016/j.jhep.2015.01.001
23. Ligresti A, De Petrocellis L, Di Marzo V. **From phytocannabinoids to cannabinoid receptors and endocannabinoids: Pleiotropic physiological and pathological roles through complex pharmacology**. *Physiol. Rev.* (2016) **96** 1593-1659. DOI: 10.1152/physrev.00002.2016
24. Iannotti FA. **Effects of non-euphoric plant cannabinoids on muscle quality and performance of dystrophic mdx mice**. *Br. J. Pharmacol.* (2019) **176** 1568-1584. DOI: 10.1111/bph.14460
25. Veilleux A, Di Marzo V, Silvestri C. **The expanded endocannabinoid system/endocannabinoidome as a potential target for treating diabetes mellitus**. *Curr. Diabetes Rep.* (2019) **19** 1-12. DOI: 10.1007/s11892-019-1248-9
26. Forte N, Fernández-Rilo AC, Palomba L, Di Marzo V, Cristino L. **Obesity affects the microbiota-gut-brain axis and the regulation thereof by endocannabinoids and related mediators**. *Int. J. Mol. Sci.* (2020) **21** 1554. DOI: 10.3390/ijms21051554
27. Di Marzo V, Silvestri C. **Lifestyle and metabolic syndrome: Contribution of the endocannabinoidome**. *Nutrients* (2019) **11** 1-24. DOI: 10.3390/nu11081956
28. McPartland JM, Glass M, Pertwee RG. **Meta-analysis of cannabinoid ligand binding affinity and receptor distribution: Interspecies differences**. *Br. J. Pharmacol.* (2007) **152** 583-593. DOI: 10.1038/sj.bjp.0707399
29. Morales P, Hurst DP, Reggio PH. **Molecular targets of the phytocannabinoids: A complex picture**. *Prog. Chem. Org. Nat. Prod.* (2017) **103** 103-131. PMID: 28120232
30. Leweke FM. **Cannabidiol enhances anandamide signaling and alleviates psychotic symptoms of schizophrenia**. *Transl. Psychiatry* (2012) **2** e94. DOI: 10.1038/tp.2012.15
31. Zierath JR, Hawley JA. **Skeletal muscle fiber type: Influence on contractile and metabolic properties**. *PLoS Biol.* (2004) **2** e348. DOI: 10.1371/journal.pbio.0020348
32. Charytoniuk T. **Cannabidiol Downregulates Myocardial de Novo Ceramide Synthesis Pathway in a Rat Model of High-Fat Diet-Induced Obesity**. *Int. J. Mol. Sci.* (2022) **23** 2232. DOI: 10.3390/ijms23042232
33. Glatz JFC, Nabben M, Luiken JJFP. **CD36 (SR-B2) as master regulator of cellular fatty acid homeostasis**. *Curr. Opin. Lipidol.* (2022) **33** 103-111. DOI: 10.1097/MOL.0000000000000819
34. Kuang M, Febbraio M, Wagg C, Lopaschuk GD, Dyck JRB. **Fatty acid translocase/CD36 deficiency does not energetically or functionally compromise hearts before or after ischemia**. *Circulation* (2004) **109** 1550-1557. DOI: 10.1161/01.CIR.0000121730.41801.12
35. Zhu B. **Lipid oversupply induces CD36 sarcolemmal translocation via dual modulation of PKCζ and TBC1D1: An early event prior to insulin resistance**. *Theranostics* (2020) **10** 1332-1354. DOI: 10.7150/thno.40021
36. Zeng H. **CD36 promotes de novo lipogenesis in hepatocytes through INSIG2-dependent SREBP1 processing**. *Mol. Metab.* (2022) **57** 101428. DOI: 10.1016/j.molmet.2021.101428
37. Itani SI, Ruderman NB, Schmieder F, Boden G. **Lipid-induced insulin resistance in human muscle is associated with changes in diacylglycerol, protein kinase C, and IkappaB-alpha**. *Diabetes* (2002) **51** 2005-2011. DOI: 10.2337/diabetes.51.7.2005
38. Qu X, Seale JP, Donnelly R. **Tissue and isoform-selective activation of protein kinase C in insulin-resistant obese Zucker rats-effects of feeding**. *J. Endocrinol.* (1999) **162** 207-214. DOI: 10.1677/joe.0.1620207
39. Sitnick MT. **Skeletal muscle triacylglycerol hydrolysis does not influence metabolic complications of obesity**. *Diabetes* (2013) **62** 3350-3361. DOI: 10.2337/db13-0500
40. Bielawiec P, Harasim-Symbor E, Sztolsztener K, Konstantynowicz-Nowicka K, Chabowski A. **Attenuation of Oxidative stress and inflammatory response by chronic cannabidiol administration is associated with improved n-6/n-3 PUFA ratio in the white and red skeletal muscle in a rat model of high-fat diet-induced obesity**. *Nutrients* (2021) **13** 1603. DOI: 10.3390/nu13051603
41. Solinas G, Borén J, De Dulloo AG. **novo lipogenesis in metabolic homeostasis: More friend than foe?**. *Mol. Metab.* (2015) **4** 367-377. DOI: 10.1016/j.molmet.2015.03.004
42. Funai K. **Muscle lipogenesis balances insulin sensitivity and strength through calcium signaling**. *J. Clin. Investig.* (2013) **123** 1229-1240. DOI: 10.1172/JCI65726
43. Osei-Hyiaman D. **Endocannabinoid activation at hepatic CB 1 receptors stimulates fatty acid synthesis and contributes to diet-induced obesity**. *J. Clin. Investig.* (2005) **115** 1298-1305. DOI: 10.1172/JCI200523057
44. Chu K, Miyazaki M, Man WC, Ntambi JM. **Stearoyl-coenzyme A desaturase 1 deficiency protects against hypertriglyceridemia and increases plasma high-density lipoprotein cholesterol induced by liver X receptor activation**. *Mol. Cell Biol.* (2006) **26** 6786-6798. DOI: 10.1128/MCB.00077-06
45. Liu J. **Monounsaturated fatty acids generated via stearoyl CoA desaturase-1 are endogenous inhibitors of fatty acid amide hydrolase**. *Proc. Natl. Acad. Sci. U. S. A.* (2013) **110** 18832-18837. DOI: 10.1073/pnas.1309469110
46. Laprairie RB, Bagher AM, Kelly MEM, Denovan-Wright EM. **Cannabidiol is a negative allosteric modulator of the cannabinoid CB1 receptor**. *Br. J. Pharmacol.* (2015) **172** 4790-4805. DOI: 10.1111/bph.13250
47. Yashiro H. **A novel selective inhibitor of delta-5 desaturase lowers insulin resistance and reduces body weight in diet-induced obese C57BL/6J mice**. *PLoS ONE* (2016) **11** e0166198. DOI: 10.1371/journal.pone.0166198
48. Inoue K, Kishida K, Hirata A, Funahashi T, Shimomura I. **Low serum eicosapentaenoic acid/arachidonic acid ratio in male subjects with visceral obesity**. *Nutr. Metab.* (2013) **10** 25. DOI: 10.1186/1743-7075-10-25
49. Stoffel W. **Obesity resistance and deregulation of lipogenesis in Δ6-fatty acid desaturase (FADS2) deficiency**. *EMBO Rep.* (2014) **15** 110-120. DOI: 10.1002/embr.201338041
50. Warensjö E, Ohrvall M, Vessby B. **Fatty acid composition and estimated desaturase activities are associated with obesity and lifestyle variables in men and women**. *Nutr. Metab. Cardiovasc. Dis.* (2006) **16** 128-136. DOI: 10.1016/j.numecd.2005.06.001
51. Matsuzaka T. **Crucial role of a long-chain fatty acid elongase, Elovl6, in obesity-induced insulin resistance**. *Nat. Med.* (2007) **13** 1193-1202. DOI: 10.1038/nm1662
52. Kihara A. **Very long-chain fatty acids: Elongation, physiology and related disorders**. *J. Biochem.* (2012) **152** 387-395. DOI: 10.1093/jb/mvs105
53. Kozawa S. **Induction of peroxisomal lipid metabolism in mice fed a high-fat diet**. *Mol. Med. Rep.* (2011) **4** 1157-1162. PMID: 21850377
54. Chavez JA, Summers SA. **Characterizing the effects of saturated fatty acids on insulin signaling and ceramide and diacylglycerol accumulation in 3T3-L1 adipocytes and C2C12 myotubes**. *Arch. Biochem. Biophys.* (2003) **419** 101-109. DOI: 10.1016/j.abb.2003.08.020
55. Zalewska A, Maciejczyk M, Szulimowska J, Imierska M, Błachnio-Zabielska A. **High-fat diet affects ceramide content, disturbs mitochondrial redox balance, and induces apoptosis in the submandibular glands of mice**. *Biomolecules* (2019) **9** 877. DOI: 10.3390/biom9120877
56. Rajesh M. **Cannabidiol attenuates cardiac dysfunction, oxidative stress, fibrosis, and inflammatory and cell death signaling pathways in diabetic cardiomyopathy**. *J. Am. Coll. Cardiol.* (2010) **56** 2115-2125. DOI: 10.1016/j.jacc.2010.07.033
57. Ignatowska-Jankowska B, Jankowski MM, Swiergiel AH. **Cannabidiol decreases body weight gain in rats: Involvement of CB2 receptors**. *Neurosci. Lett.* (2011) **490** 82-84. DOI: 10.1016/j.neulet.2010.12.031
58. Folch J, Lees M, Sloane Stanley G. **A simple method for the isolation and purification of total lipides from animal tissues**. *J. Biol. Chem.* (1987) **55** 999-1033
|
---
title: 'Spousal Similarities in Cardiovascular Risk Factors in Northern China: A Community-Based
Cross-Sectional Study'
authors:
- Binbin Lin
- Li Pan
- Huijing He
- Yaoda Hu
- Ji Tu
- Ling Zhang
- Ze Cui
- Xiaolan Ren
- Xianghua Wang
- Jing Nai
- Guangliang Shan
journal: International Journal of Public Health
year: 2023
pmcid: PMC9988901
doi: 10.3389/ijph.2023.1605620
license: CC BY 4.0
---
# Spousal Similarities in Cardiovascular Risk Factors in Northern China: A Community-Based Cross-Sectional Study
## Abstract
Objectives: The aim of this study was to explore spousal similarities in cardiovascular risk factors in northern China.
Methods: We conducted a cross-sectional study of married couples from Beijing, Hebei, Gansu, and Qinghai provinces between 2015 and 2019. A total of 2,020 couples were included in the final analyses. The spousal similarities for metabolic indicators and cardiovascular risk factors (including lifestyle factors and cardiometabolic diseases) were evaluated using Spearman’s correlation and logistic regression analyses, respectively.
Results: All metabolic indicators showed positive spousal correlations ($p \leq 0.001$), with the strongest for fasting blood glucose ($r = 0.30$) and the lowest for high-density lipoprotein cholesterol ($r = 0.08$). Significant husband-wife associations were observed for several cardiovascular risk factors except for hypertension in multivariable models, with the strongest association for physical inactivity (odds ratios with $95\%$ confidence intervals of 3.59 [2.85, 4.52] and 3.54 [2.82, 4.46] for husbands and wives, respectively). In addition, the interaction of age with spousal overweight/obesity status was statistically significant, and the association was stronger in people ≥50 years.
Conclusion: There were spousal similarities in cardiovascular risk factors. The finding may have public health implications that targeted screening and interventions for spouses of people with cardiovascular risk factors.
## Introduction
The occurrence of cardiovascular diseases continues to increase in China due to the dual pressures of population ageing and the steady rise in the prevalence of metabolic risk factors. In 2019, the number of people with cardiovascular diseases was nearly 330 million, with two out of five deaths attributable to cardiovascular diseases [1]. *Both* genetic and environmental factors may contribute to the development of cardiovascular diseases [2]. The well-recognized environmental risk factors for cardiovascular diseases are smoking, alcohol consumption, high body mass index (BMI), high blood pressure, high cholesterol, and high fasting blood glucose [3]. In addition, the prevalence of hyperuricemia in the Chinese population has significantly increased in recent decades [4]. Several studies have suggested that hyperuricemia significantly increases the risk of cardiovascular diseases through the potential mechanism of inducing inflammation, oxidative stress, and subsequent endothelial dysfunction [5, 6]. Most cardiovascular risk factors are modifiable or manageable, and effective interventions for individuals with these risk factors may reduce the burden of cardiovascular diseases [7].
Couples are not genetically related but are likely to face the same health problems due to sharing environmental factors, adopting similar behaviors, or experiencing assortative mating (non-random partner selection based on the similarity of observable characteristics) [8]. There have been several studies on the spousal aggregation of obesity [9], hypertension [10, 11], diabetes [12, 13], depression [14], and cancer [15]. Likewise, the behaviors between spouses might influence each other, which suggests that interventions targeted at couples may be more effective than those targeted at individuals [9]. A few studies have demonstrated concordance of cardiovascular risk factors between Chinese couples [2, 16, 17]. However, two of these studies were with participants from cities in southern China [2, 17]. Cultural differences between northern and southern China may affect the concordance of couples’ lifestyles. Liao et al. focused only on middle-aged and elderly couples and merely studied the concordance of chronic disease, not lifestyles [16]. Moreover, to the best of our knowledge, concordance of hyperuricemia between couples has not been reported.
The aim of our study is to investigate spousal associations for modifiable lifestyles (smoking, alcohol drinking, leisure-time physical activity, and overweight/obesity) and cardiometabolic diseases (hypertension, diabetes mellitus, dyslipidemia, and hyperuricemia), which are major risk factors for cardiovascular disease, in married couples from four provinces in northern China. We also assessed whether the concordance differed by age.
## Study Population
The current study was based on the China National Health Survey (CNHS) and the cohort study of the general population in the Beijing-Tianjin-Hebei (an abbreviation of the Chinese name is Jing-Jin-Ji) area (J-J-J Cohort Study). The CNHS is a nationwide cross-sectional study conducted from 2012 to 2017, and the J-J-J Cohort *Study is* a population-based prospective cohort study initiated in 2017. Details of the CNHS and the J-J-J Cohort Study designs and methods have been described elsewhere [18, 19]. In brief, the CNHS and the J-J-J Cohort Study investigated community populations aged 18 years and above, collecting data on demographic and socioeconomic information, lifestyle factors, anthropometric measures, laboratory tests, and clinical profiles to explore risk factors for major chronic diseases. The CNHS and the J-J-J Cohort Study were approved by the ethics committee of the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences. All participants provided written informed consent.
In this study, we used data from Qinghai, Gansu, and Hebei provinces, three CNHS survey sites surveyed from 2015 to 2017, and Beijing, one of the J-J-J Cohort Study survey sites conducted in 2019, as additional information on family relationships was collected at these four survey sites. We identified 2,073 couples who were married at the time of the interview and further excluded couples if either member had missing data for sociodemographic information (age, sex, education, and income), lifestyle factors (smoking status, alcohol consumption, and exercise), medical history (hypertension and diabetes), BMI, blood pressure, and blood biochemistry. A total of 2,020 couples were included in the current analyses. Supplementary Figure S1 shows a flow chart for selecting participants.
## Data Collection
The questionnaire and equipment were uniform and the criteria was consistent for the measurements that were used by the CNHS and the J-J-J Cohort Study. All investigators underwent a training program to guarantee their capability to conduct precise data collection. Information on sociodemographic characteristics, lifestyle factors, and personal and family medical history was obtained from a standard questionnaire through face-to-face interviews. To ensure data accuracy, questionnaires were checked and verified by inspectors after the completion of each questionnaire. All participants were asked, “Has your spouse also come to participate in the survey?” Furthermore, the ID and name of the participant’s spouse were collected if they replied “yes” to the first question. Couples could be combined using the family ID variable.
A physical examination, including anthropometry measurements (height and body weight) and blood pressure measurements, was conducted by well-trained staff. Body weight and height were measured using standard procedures with the participants wearing light clothes and no shoes. BMI was calculated as body weight (kg) divided by height (m) squared. Blood pressure was measured 3 times using an automated electronic device (OMRON, HEM-907). The mean value of the 3 readings was used for analysis. Fasting blood samples were collected from participants for measurements of fasting blood glucose (FBG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), and serum uric acid (UA) via a standard protocol.
## Modifiable Lifestyle Factors
Lifestyle behaviors, including smoking, alcohol consumption, and physical activity, were defined according to the validated self-report questionnaires. Subjects were classified as current smokers or non-smokers. Drinking status was categorized into current drinkers or non-drinkers. Physical activity level was divided into two categories according to the frequency of leisure time exercise in the last year. Physical inactivity was defined as exercising less than once per week. Overweight/obesity was defined as BMI ≥24 kg/m2 based on criteria for Chinese individuals [20].
## Cardiometabolic Diseases
Participants who reported having been diagnosed with hypertension by a doctor or who had an average measured systolic blood pressure (SBP) ≥ 140 mmHg or diastolic blood pressure (DBP) ≥ 90 mmHg were defined as having hypertension. Diabetes was defined as having been diagnosed with diabetes by a doctor or having a measured FBG concentration ≥7.0 mmol/L. According to Chinese guidelines on the prevention and treatment of dyslipidemia in adults, dyslipidemia was defined as having at least one of the following: high TC (≥6.2 mmol/L), low HDL-C (<1.0 mmol/L), high LDL-C (≥4.1 mmol/L), and high TG (≥2.3 mmol/L) [21]. Hyperuricemia was defined as serum UA > 360 μmol/L in women and serum UA > 420 μmol/L in men.
## Exposure and Outcome Variables and Covariates
The exposure variables in our study were spousal cardiovascular risk factors, including current smoking, current drinking, physical inactivity, overweight/obesity, hypertension, diabetes, dyslipidemia, and hyperuricemia. The corresponding cardiovascular risk factors for individuals were defined as the outcome. Sociodemographic covariates included age (20–29, 30–39, 40–49, 50–59, 60–69, ≥70 years), education level (middle school or below, high school or above), annual income (<24,000 RMB, ≥24,000 RMB), and geographic region (Qinghai, Gansu, Hebei, and Beijing). The abovementioned lifestyle factors and a family history of hypertension or diabetes were also included as covariates in the models for cardiometabolic diseases. A family history of hypertension or diabetes was defined as having at least one first-degree relative with hypertension or diabetes.
## Statistical Analysis
Characteristics of the study population were presented as the mean ± SD or median (IQR) for continuous variables and the number (frequency) for categorical variables according to sex. Student’s t tests or Wilcoxon rank sum tests were used to compare the differences between groups for continuous variables. The χ 2 test was used to compare the differences between groups for categorical variables. Concordance for cardiovascular risk factors was defined as a case in which both spouses had the same response for a category of variables.
The univariate Spearman correlation coefficient (r) and the Phi coefficient were used to assess the correlations of continuous and categorical variables within couples, respectively. Spousal correlations for metabolic indicators (BMI, SBP, DBP, FBG, TC, HDL-C, LDL-C, TG, and UA) might emerge because of the associations of these indicators with age. Therefore, we fitted a regression of metabolic indicators against age and derived the residual for these indicators adjusted for individuals’ age. We then used the residuals to calculate the spousal correlations in the metabolic indicators; namely, the correlation coefficients were corrected for the age of both partners (Model 1). Since previous studies reported that BMI is a surrogate for assortative mating [22], we further adjusted the BMI of both spouses in Model 2 using the same method.
The χ 2 test was performed to explore the crude association of cardiovascular risk factors between couples. The odds ratios (ORs) and their corresponding $95\%$ confidence intervals (CIs) were estimated by age-adjusted and multivariable-adjusted logistic regression models to indicate the spousal association for each given cardiovascular risk factor. The multivariable model of lifestyle factors adjusted the age, education, annual income, and geographic regions of the individuals. The multivariable model of cardiometabolic disease was further adjusted for individuals’ smoking, drinking, leisure-time physical activity, and overweight/obesity status. Family history of hypertension and diabetes was additionally adjusted in the models of hypertension and diabetes, respectively. Spousal similarity was defined as r > 0 and its corresponding p value <0.05 for continuous variables and ORs >1 and their corresponding $95\%$ CIs excluding 1 for categorical variables.
To explore potential changes in spousal similarities with age, roughly representing marriage duration, subgroup analyses were performed according to the individuals’ age (20–50 years, ≥50 years). Multiplicative interaction was calculated by cross-product interaction terms in multivariable logistic regression models. A sensitivity analysis was performed, excluding couples with an age difference ≥5 years, to evaluate the robustness of the results.
All analyses were performed separately for husbands and wives. The significance level was set as a 2-sided p value < 0.05. SAS 9.4 (SAS Institute Inc., Cary, NC, United States.) was used to conduct all analyses.
## Results
The characteristics of the women and men are presented in Table 1. A total of 2,020 married couples, with a mean (SD) age of 55.4 (12.1) years among men and 53.7 (11.9) years among women, were included in the study. The study population was geographically diverse ($32.48\%$ from Beijing, $28.32\%$ from Hebei, $31.68\%$ from Gansu, and $7.52\%$ from Qinghai). Husbands were more likely than their wives to have higher levels of education, annual income, BMI, SBP, DBP, FBG, and UA.
**TABLE 1**
| Variables | Overall (N = 4,040) | Male (n = 2,020) | Female (n = 2,020) | p value |
| --- | --- | --- | --- | --- |
| Demographic characteristics | Demographic characteristics | Demographic characteristics | Demographic characteristics | Demographic characteristics |
| Age (years), mean (SD) | 54.5 ± 12.0 | 55.4 ± 12.1 | 53.7 ± 11.9 | <0.0001 |
| Geographic regions, n (%) | | | | — |
| Beijing | 1,312 (32.48) | 656 (32.48) | 656 (32.48) | |
| Hebei | 1,144 (28.32) | 572 (28.32) | 572 (28.32) | |
| Gansu | 1,280 (31.68) | 640 (31.68) | 640 (31.68) | |
| Qinghai | 304 (7.52) | 152 (7.52) | 152 (7.52) | |
| Annual income (RMB), median (IQR) | 24,000 (12,000–42,000) | 30,000 (13,333–48,000) | 24,000 (12,000–36,000) | <0.0001 |
| Educated to high school or above, n (%) | 1,400 (34.7) | 762 (37.7) | 638 (31.6) | <0.0001 |
| Metabolic indicators | Metabolic indicators | Metabolic indicators | Metabolic indicators | Metabolic indicators |
| BMI (kg/m2), mean (SD) | 25.0 ± 3.5 | 25.3 ± 3.4 | 24.8 ± 3.5 | <0.0001 |
| SBP (mmHg), mean (SD) | 128.0 ± 18.2 | 130.2 ± 16.8 | 125.8 ± 19.2 | <0.0001 |
| DBP (mmHg), mean (SD) | 76.3 ± 10.9 | 78.5 ± 10.7 | 74.1 ± 10.6 | <0.0001 |
| FBG (mmol/L), mean (SD) | 5.9 ± 1.7 | 6.1 ± 1.9 | 5.8 ± 1.5 | <0.0001 |
| TC (mmol/L), mean (SD) | 4.8 ± 1.0 | 4.8 ± 1.0 | 4.9 ± 1.1 | <0.0001 |
| HDL-C (mmol/L), mean (SD) | 1.3 ± 0.3 | 1.2 ± 0.3 | 1.3 ± 0.3 | <0.0001 |
| LDL-C (mmol/L), mean (SD) | 2.8 ± 0.8 | 2.8 ± 0.8 | 2.8 ± 0.8 | 0.0010 |
| TG (mmol/L), median (IQR) | 1.5 (1.1–2.2) | 1.5 (1.1–2.3) | 1.5 (1.1–2.1) | <0.0001 |
| UA (μmol/L), mean (SD) | 322.4 ± 88.2 | 362.7 ± 87.7 | 282.1 ± 67.9 | <0.0001 |
The within-couple concordance (the proportion of couples in which both or neither of the spouses had a certain risk factor) for the cardiovascular risk factors ranged from a low of $40.70\%$ for current drinkers to a high of $74.1\%$ for leisure-time physical activity (Figure 1). The proportion of couples who were both overweight/obese and hypertensive accounted for approximately $39\%$ and $24\%$ of couples, respectively. The correlation of leisure-time physical activity within couples was the strongest among cardiovascular risk factors (phi = 0.3972, see Supplementary Table S1).
**FIGURE 1:** *Percentage of couples in concordance categories of cardiovascular risk factors. H+ W+ indicates couples in which both the husband and wife have the characteristic; H- W-, couples in which neither have it; H- W+ and H+ W- indicate discordant pairs. Qinghai, Gansu, Hebei, and Beijing, China, 2015–2019.*
Table 2 shows that there were significant correlations between spouses in BMI ($r = 0.17$; $p \leq 0.001$), SBP ($r = 0.26$; $p \leq 0.001$), DBP ($r = 0.11$; $p \leq 0.001$), FBG ($r = 0.30$; $p \leq 0.001$), TC ($r = 0.16$; $p \leq 0.001$), HDL-C ($r = 0.08$, $p \leq 0.001$), LDL-C ($r = 0.16$, $p \leq 0.001$), TG ($r = 0.13$, $p \leq 0.001$), and UA ($r = 0.14$; $p \leq 0.001$). After adjusting for the age and BMI of both spouses, the correlations remained significant for all traits, but the spousal correlations for SBP and FBG decreased substantially (SBP: from 0.26 to 0.12; FBG: from 0.30 to 0.17), whereas the spousal correlations for TG and UA increased (TG: from 0.13 to 0.19; UA: from 0.14 to 0.18). As shown in Supplementary Table S2, the correlations of BMI, SBP, DBP, and LDL-C between couples were stronger in those aged 50 years and older, but the correlation of HDL-C and UA was stronger in young couples (20–50 years).
**TABLE 2**
| Metabolic indicators | Unadjusted | Unadjusted.1 | Model 1 | Model 1.1 | Model 2 | Model 2.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Metabolic indicators | r | p Value | r | p Value | r | p Value |
| BMI | 0.17 | <0.001 | 0.17 | <0.001 | — | — |
| SBP | 0.26 | <0.001 | 0.16 | <0.001 | 0.12 | <0.001 |
| DBP | 0.11 | <0.001 | 0.12 | <0.001 | 0.11 | <0.001 |
| FBG | 0.30 | <0.001 | 0.21 | <0.001 | 0.17 | <0.001 |
| TC | 0.16 | <0.001 | 0.16 | <0.001 | 0.15 | <0.001 |
| HDL-C | 0.08 | <0.001 | 0.08 | <0.001 | 0.09 | <0.001 |
| LDL-C | 0.16 | <0.001 | 0.15 | <0.001 | 0.15 | <0.001 |
| TG | 0.13 | <0.001 | 0.16 | <0.001 | 0.19 | <0.001 |
| UA | 0.14 | <0.001 | 0.18 | <0.001 | 0.18 | <0.001 |
Significant spousal associations were observed in four lifestyle factors after adjusting for individuals’ age, education, annual income, and geographic regions (Table 3). Men whose wives were current smokers or current drinkers had significantly higher odds of being current smokers (OR: 1.90; $95\%$ CI: 1.07, 3.38) or current drinkers (OR:1.59; $95\%$ CI: 1.20, 2.10), and vice versa. Physical inactivity showed the strongest spousal association among cardiovascular risk factors, with ORs ($95\%$ CIs) of 3.59 (2.85, 4.52) and 3.54 (2.82, 4.46) for husbands and wives, respectively. Participants with spouses who were overweight/obese were at an increased risk of being overweight/obese (husbands: OR = 1.37 [$95\%$ CI: 1.13, 1.66]; wives: OR = 1.40 [$95\%$ CI: 1.15, 1.70]).
**TABLE 3**
| Spouses’ status | Husbands | Husbands.1 | Wives | Wives.1 |
| --- | --- | --- | --- | --- |
| Spouses’ status | Model 1 | Model 2 | Model 1 | Model 2 |
| Modifiable lifestyles a | Modifiable lifestyles a | Modifiable lifestyles a | Modifiable lifestyles a | Modifiable lifestyles a |
| Current smoker | Current smoker | Current smoker | Current smoker | Current smoker |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 1.75 (0.99, 3.08) | 1.90 (1.07, 3.38) | 1.73 (0.98, 3.06) | 1.81 (1.02, 3.23) |
| Current drinker | Current drinker | Current drinker | Current drinker | Current drinker |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 1.62 (1.23, 2.14) | 1.59 (1.20, 2.10) | 1.61 (1.22, 2.12) | 1.56 (1.18, 2.07) |
| Physical inactivity | Physical inactivity | Physical inactivity | Physical inactivity | Physical inactivity |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 4.96 (4.00, 6.15) | 3.59 (2.85, 4.52) | 4.93 (3.97, 6.12) | 3.54 (2.82, 4.46) |
| Overweight/obesity | Overweight/obesity | Overweight/obesity | Overweight/obesity | Overweight/obesity |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 1.53 (1.27, 1.85) | 1.37 (1.13, 1.66) | 1.50 (1.24, 1.82) | 1.40 (1.15, 1.70) |
| Cardiometabolic disease b | Cardiometabolic disease b | Cardiometabolic disease b | Cardiometabolic disease b | Cardiometabolic disease b |
| Hypertension | Hypertension | Hypertension | Hypertension | Hypertension |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 1.42 (1.16, 1.74) | 1.21 (0.97, 1.51) | 1.34 (1.09, 1.64) | 1.09 (0.87, 1.37) |
| Diabetes | Diabetes | Diabetes | Diabetes | Diabetes |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 1.82 (1.39, 2.38) | 1.73 (1.29, 2.31) | 1.82 (1.38, 2.39) | 1.55 (1.16, 2.07) |
| Dyslipidemia | Dyslipidemia | Dyslipidemia | Dyslipidemia | Dyslipidemia |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 1.32 (1.09, 1.60) | 1.29 (1.06, 1.57) | 1.34 (1.10, 1.62) | 1.28 (1.05, 1.57) |
| Hyperuricemia | Hyperuricemia | Hyperuricemia | Hyperuricemia | Hyperuricemia |
| No (ref) | 1 | 1 | 1 | 1 |
| Yes | 1.47 (1.07, 2.02) | 1.41 (1.01, 1.96) | 1.49 (1.09, 2.04) | 1.42 (1.02, 1.97) |
There were statistically significant associations between the wives’ health conditions and the corresponding health conditions in their husbands, with multivariable-adjusted ORs ($95\%$ CIs) of 1.73 (1.29, 2.31) for diabetes, 1.29 (1.06, 1.57) for dyslipidemia, and 1.41 (1.01, 1.96) for hyperuricemia. Husbands’ health condition statuses had similar effects on their wives. However, the wife-husband association was not significant for hypertension in the multivariable model. In addition, the spousal associations varied in different types of dyslipidemia in the multivariable model, with significant wife-husband associations observed in low HDL-C and high TG (Supplementary Table S3). The women whose spouses had high LDL-C were at an increased risk of having high LDL-C than those whose spouses did not have high LDL-C, while the spousal association was not significant in men.
Figure 2 shows that the husband-wife associations for cardiovascular risk factors did not differ appreciably according to age group, except for overweight/obesity. We observed a significant interaction between age and overweight/obesity status (for men, P for interaction = 0.0112; for women, P for interaction = 0.016). The association was stronger in individuals aged ≥50 years old than in those aged <50 years old (OR: 1.61 [$95\%$ CI: 1.27, 2.04] compared with 0.90 [0.63, 1.29] for men; 1.70 [1.34, 2.18] compared with 0.95 [0.68, 1.32] for women), which suggests that having a spouse with overweight/obesity in middle-aged and elderly individuals would have a higher risk of becoming overweight/obese. The spousal associations for physical inactivity were significant across age groups. To test the robustness of our findings, we excluded couples with an age difference of ≥5 years and found that the results were not substantially changed (Supplementary Table S4).
**FIGURE 2:** *The forest plot of spousal associations for cardiovascular risk factors by age. The plot shows the adjusted odds ratios (ORs) and 95% confidence intervals (CIs) of the association between each given cardiovascular risk factor for individuals with the same condition as spouses. The reference to all risk factors is “no.” Wives’ influence on husbands (A). Husbands’ influence on wives (B). Models for current smoker, current drinker, physical inactivity, and overweight/obesity were adjusted for individuals’ age, education, annual income, and geographic regions. Models for hypertension, diabetes, dyslipidemia, and hyperuricemia were further adjusted for individuals’ smoking status, drinking status, physical activity, and overweight/obesity. Family history of hypertension and diabetes was additionally adjusted in the models of hypertension and diabetes, respectively. Qinghai, Gansu, Hebei, and Beijing, China, 2015–2019.*
## Discussion
In this community-based, cross-sectional study, we found statistically significant spousal correlations in metabolic indicators among Chinese couples. Significant wife-husband associations were observed for several cardiovascular risk factors, with stronger associations for behavioral risk factors. No statistically significant association was found for hypertension after adjusting for their personal risk factors. There were no sex differences in our findings.
Consistent with previous studies [2, 17], husbands had a higher level of education but also had higher levels of BMI, SBP, DBP, FBG, and UA than their wives, which may be attributed to the traditional gender norm that women tend to marry men with a higher level of education [23] and sex differences in the distribution of these metabolic indicators [24]. Additionally, the mean age of husbands was slightly older than that of wives, and BMI, SBP, DBP, and FBG significantly increased with age [25]. We found statistically significant but generally weak and modest spousal correlations for metabolic indicators. The magnitude of these correlations was in line with previous studies [26, 27]. After adjusting for both partners’ age and BMI, the correlation coefficients for SBP and FBG decreased significantly, indicating that spousal correlations for the two traits were partly due to the within-couple correlations for age and BMI (age: $r = 0.97$; BMI: $r = 0.17$ in our study). Nakaya et al. found spousal similarities in cardiometabolic risk factors among 5,391 random male-female pairs with similar ages, but the significant correlations disappeared after adjustment for age [28]. However, the correlation coefficients between couples for metabolic indicators remained significant after adjusting for both partners’ age and BMI in our study, suggesting that other important environmental factors may contribute to spousal similarities in these indicators. Wilson et al. found that health behavior concordance (e.g., diet and sleep) and marital quality affect cardiometabolic similarity between couples [29]. We admit that some small effect sizes (e.g., the spousal correlation of 0.08 for HDL-C) may not be very meaningful clinically, but we suggest that shared environmental factors may play an important role in spousal similarities for metabolic profiles.
Our results demonstrated a positive association for lifestyles among couples, consistent with previous studies [17, 30]. We found that the risk of being a current smoker when having a currently smoking spouse was more than 1.8 times that of having a current non-smoking spouse. However, Dutch and American couples showed substantially stronger associations for current smokers, with ORs ($95\%$ CIs) of 6.9 (6.3, 7.5) and 5.4 (4.7, 6.3), respectively [26, 31]. Notably, $48.7\%$ and $2.6\%$ of Chinese husbands and wives were current smokers, respectively, compared with $15.5\%$ and $11.4\%$ in Dutch couples [26] and $12.6\%$ and $6.5\%$ in American couples [31]. Similarly, Dutch couples with smaller sex differences in the prevalence of current drinking had a stronger association for current drinking (OR: 5.14; 95 CI%: 4.70, 5.61) than our study (1.59; 1.20, 2.10). Therefore, the degree of spousal concordance for current smoking and drinking may be strongly influenced by the disparities in current smoking and drinking prevalence across sex. Further, the phenomenon may be attributed to cultural differences in the acceptance and freedom of smoking and alcohol consumption among women and men [32].
Increased intake of energy-dense foods and decreased physical activity appear to account for the dramatically increased prevalence of obesity throughout the past few decades [33]. In the current study, the highest proportion of cardiovascular risk factors shared by both spouses was overweight/obesity, which accounted for approximately $40\%$. The odds of being overweight/obese were nearly 1.4 times higher among individuals whose spouses were overweight/obese than among those whose spouses were not overweight/obese, which was in line with earlier evidence [14]. The spousal concordance for overweight/obesity may be attributed to couples’ similar diet and physical activity habits [9]. Middle-aged and elderly couples with a stronger spousal association for overweight/obesity may become the key group for weight management.
A meta-analysis suggested that people are likely to have hypertension (OR: 1.41; $95\%$ CI: 1.08, 1.45) if their spouse has hypertension [34]. However, most of the studies included in the meta-analysis adjusted only for demographic characteristics variables such as age and education. We found that the age-adjusted husband-wife association for hypertension was statistically significant, but the association was not statistically significant after further adjusting for individuals’ education level, income, geographic locations, and lifestyle factors, which may imply that the spousal association for hypertension can be mainly explained by spousal concordance for these covariates. Spouses of those affected by diabetes are at a more than 1.5-fold higher risk of diabetes after adjusting for individuals’ risk factors, which was in line with previous studies [10, 27].
Uric acid is the end product of purine metabolism in the human body. Although the causal relationship between uric acid and cardiovascular disease remains unclear, many epidemiological studies have suggested that hyperuricemia is strongly associated with cardiovascular disease [5, 35]. To our knowledge, this is the first study to report spousal similarity in hyperuricemia. The odds of having hyperuricemia when the spouse had hyperuricemia were significantly increased in our study. We also found that the correlation coefficient of UA within young couples was greater than that of older couples. Notably, a recent study reported that the rising prevalence of hyperuricemia in the younger population [36], which implies that we should improve the management of uric acid levels in young couples.
The following two theories have been widely publicized in an attempt to explain the spousal concordance of cardiovascular risk factors: [1] the observed associations may be due to the tendency to select spouses based on a preference for similar characteristics (assortative mating) or [2] the impact of the shared environments and convergence in lifestyles during long-term cohabitation [2, 34]. In our opinion, having a spouse with a cardiometabolic disease increases the risk of developing the same disease due to an indirect association. First, couples had similar sociodemographic characteristics, such as similar age, which is one of the most significant risk factors for cardiometabolic diseases. Second, couples are likely to adopt similar lifestyles, which are well-recognized environmental risk factors for cardiometabolic diseases. Helga et al. reported that spousal concordance in lifestyles was due to assortative mating and convergence over time [37]. For example, individuals more often choose a spouse with similar exercise habits. Over the course of long-term cohabitation, spousal interactions and shared environmental resources (e.g., access to community sports facilities) may also lead to similar spousal physical activity participation [26]. Finally, the husband-wife associations remained significant after adjusting for sociodemographic characteristics and lifestyles, which suggests that important unobserved factors shared by couples impact their likelihood of developing the same diseases, such as dietary patterns. The two theories jointly explain the spousal resemblance in cardiovascular risk factors, and it is difficult to determine the most important factors.
Our findings have important public health implications. Spousal similarities in cardiovascular risk factors were observed among couples from northern China, and there was a higher burden of cardiovascular disease [38]. Smoking, alcohol consumption, physical inactivity, and overweight/obesity are well-known modifiable risk factors for cardiovascular diseases [3]. A previous study suggested that individuals are likely to make similar changes when their spouses improve their behaviors (i.e., quit smoking, quit drinking, start to exercise) [30]. It has been shown that programs targeted at couples can be more effective in improving health behavior than individual interventions [39], which may provide new ideas for cardiovascular disease prevention strategies in China. For example, couple-based lifestyle interventions can be implemented in the community to encourage couples to participate in regular physical activity and weight control together. Additionally, regular screening may be recommended for the spouses of patients with cardiometabolic diseases and interventions as early as possible to reduce the risk of these diseases in their spouses. Exploring the pattern and effects of couple-based intervention programs may be increasingly important in the future.
## Limitations
There were several limitations in this study. The marriage duration of couples was unavailable, so we could not determine the long-term and short-term effects of marriage on cardiovascular risk factors. However, a previous study reported that age and relationship length are closely correlated ($r = 0.78$) [29]. Therefore, a stratified analysis of age was performed to roughly assess the effect of marital length on the associations of cardiovascular risk factors between couples in our study. Second, since the study population was from four northern provinces of China, the results could not represent the situation in other regions of China. Nevertheless, it could suggest that we should pay attention to the health status of individuals whose spouses have cardiovascular risk factors. Finally, a cross-sectional design cannot assess the associations of cardiovascular risk factors within couples over time.
## Conclusion
The aggregation of risk factors for cardiovascular disease in couples from northern China was depicted in this study. The results suggested that if wives had unhealthy lifestyles and cardiometabolic diseases, then their husbands were susceptible to the same behavior or disease and vice versa. Overall, these observations may imply that intervention and regular screening of spouses of patients with cardiovascular risk factors are warranted.
## Ethics Statement
The studies involving human participants were reviewed and approved by the ethics committee of the Institute of Basic Medical Sciences Chinese Academy of Medical Sciences. The patients/participants provided their written informed consent to participate in this study.
## Author Contributions
GS, LP, and HH: funding acquisition, validation, project administration, and supervision. All authors contributed to the material preparation and data collection. BL: methodology, formal analysis, and writing. GS and YH critically revised the manuscript. All authors read and approved the final 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.2023.1605620/full#supplementary-material
## References
1. **Report on Cardiovascular Health and Diseases in China 2021: An Updated Summary**. *Biomed Environ Sci* (2022) **35** 573-603. DOI: 10.3967/bes2022.079
2. Retnakaran R, Wen SW, Tan H, Zhou S, Ye C, Shen M. **Spousal Concordance of Cardiovascular Risk Factors in Newly Married Couples in China**. *JAMA Netw Open* (2021) **4** e2140578. DOI: 10.1001/jamanetworkopen.2021.40578
3. **Cardiovascular Disease, Chronic Kidney Disease, and Diabetes Mortality burden of Cardiometabolic Risk Factors from 1980 to 2010: a Comparative Risk Assessment**. *Lancet Diabetes Endocrinol* (2014) **2** 634-47. DOI: 10.1016/S2213-8587(14)70102-0
4. Wu J, Qiu L, Cheng XQ, Xu T, Wu W, Zeng XJ. **Hyperuricemia and Clustering of Cardiovascular Risk Factors in the Chinese Adult Population**. *Sci Rep* (2017) **7** 5456. DOI: 10.1038/s41598-017-05751-w
5. Chen JH, Chuang SY, Chen HJ, Yeh WT, Pan WH. **Serum Uric Acid Level as an Independent Risk Factor for All-Cause, Cardiovascular, and Ischemic Stroke Mortality: a Chinese Cohort Study**. *Arthritis Rheum* (2009) **61** 225-32. DOI: 10.1002/art.24164
6. Saito Y, Tanaka A, Node K, Kobayashi Y. **Uric Acid and Cardiovascular Disease: A Clinical Review**. *J Cardiol* (2021) **78** 51-7. DOI: 10.1016/j.jjcc.2020.12.013
7. Sattar N, Gill JMR, Alazawi W. **Improving Prevention Strategies for Cardiometabolic Disease**. *Nat Med* (2020) **26** 320-5. DOI: 10.1038/s41591-020-0786-7
8. Knuiman MW, Divitini ML, Bartholomew HC, Welborn TA. **Spouse Correlations in Cardiovascular Risk Factors and the Effect of Marriage Duration**. *Am J Epidemiol* (1996) **143** 48-53. DOI: 10.1093/oxfordjournals.aje.a008656
9. Cobb LK, McAdams-DeMarco MA, Gudzune KA, Anderson CA, Demerath E, Woodward M. **Changes in Body Mass Index and Obesity Risk in Married Couples over 25 Years: The ARIC Cohort Study**. *Am J Epidemiol* (2016) **183** 435-43. DOI: 10.1093/aje/kwv112
10. Watanabe T, Sugiyama T, Takahashi H, Noguchi H, Tamiya N. **Concordance of Hypertension, Diabetes and Dyslipidaemia in Married Couples: Cross-Sectional Study Using Nationwide Survey Data in Japan**. *BMJ Open* (2020) **10** e036281. DOI: 10.1136/bmjopen-2019-036281
11. Hippisley-Cox J, Pringle M. **Are Spouses of Patients with Hypertension at Increased Risk of Having Hypertension? A Population-Based Case-Control Study**. *Br J Gen Pract* (1998) **48** 1580-3. PMID: 9830183
12. Sun J, Lu J, Wang W, Mu Y, Zhao J, Liu C. **Prevalence of Diabetes and Cardiometabolic Disorders in Spouses of Diabetic Individuals**. *Am J Epidemiol* (2016) **184** 400-9. DOI: 10.1093/aje/kwv330
13. Leong A, Rahme E, Dasgupta K. **Spousal Diabetes as a Diabetes Risk Factor: a Systematic Review and Meta-Analysis**. *BMC Med* (2014) **12** 12. DOI: 10.1186/1741-7015-12-12
14. Jun SY, Kang M, Kang SY, Lee JA, Kim YS. **Spousal Concordance Regarding Lifestyle Factors and Chronic Diseases Among Couples Visiting Primary Care Providers in Korea**. *Korean J Fam Med* (2020) **41** 183-8. DOI: 10.4082/kjfm.18.0104
15. Izumi S, Imai K, Nakachi K. **Excess Concordance of Cancer Incidence and Lifestyles in Married Couples (Japan): Survival Analysis of Paired Rate Data**. *Cancer Causes Control* (2004) **15** 551-8. DOI: 10.1023/B:CACO.0000036162.86921.09
16. Liao J, Zhang J, Xie J, Gu J. **Gender Specificity of Spousal Concordance in the Development of Chronic Disease Among Middle-Aged and Older Chinese Couples: A Prospective Dyadic Analysis**. *Int J Environ Res Public Health* (2021) **18** 2886. DOI: 10.3390/ijerph18062886
17. Jurj AL, Wen W, Li HL, Zheng W, Yang G, Xiang YB. **Spousal Correlations for Lifestyle Factors and Selected Diseases in Chinese Couples**. *Ann Epidemiol* (2006) **16** 285-91. DOI: 10.1016/j.annepidem.2005.07.060
18. He H, Pan L, Pa L, Cui Z, Ren X, Wang D. **Data Resource Profile: The China National Health Survey (CNHS)**. *Int J Epidemiol* (2018) **47** 1734-5f. DOI: 10.1093/ije/dyy151
19. He H, Pan L, Hu Y, Tu J, Zhang L, Zhang M. **The Diverse Life-Course Cohort (DLCC): Protocol of a Large-Scale Prospective Study in China**. *Eur J Epidemiol* (2022) **37** 871-80. DOI: 10.1007/s10654-022-00894-1
20. Zhou B. **Predictive Values of Body Mass index and Waist Circumference to Risk Factors of Related Diseases in Chinese Adult Population**. *Zhonghua Liu Xing Bing Xue Za Zhi* (2002) **23** 5-10. PMID: 12015100
21. **2016 Chinese Guidelines for the Management of Dyslipidemia in Adults**. *J Geriatr Cardiol* (2018) **15** 1-29. DOI: 10.11909/j.issn.1671-5411.2018.01.011
22. Di Castelnuovo A, Quacquaruccio G, Arnout J, Cappuccio FP, de Lorgeril M, Dirckx C. **Cardiovascular Risk Factors and Global Risk of Fatal Cardiovascular Disease Are Positively Correlated between Partners of 802 Married Couples from Different European Countries. Report from the IMMIDIET Project**. *Thromb Haemost* (2007) **98** 648-55. DOI: 10.1160/th07-01-0024
23. Li W, Li S, Feldman MW.. **Marriage Aspiration, Perceived Marriage Squeeze, and Anomie Among Unmarried Rural Male Migrant Workers in China**. *Am J Mens Health* (2019) **13** 1557988319856170. DOI: 10.1177/1557988319856170
24. Wu J, Cheng X, Qiu L, Xu T, Zhu G, Han J. **Prevalence and Clustering of Major Cardiovascular Risk Factors in China: A Recent Cross-Sectional Survey**. *Medicine (Baltimore)* (2016) **95** e2712. DOI: 10.1097/MD.0000000000002712
25. Chen GY, Li L, Dai F, Li XJ, Xu XX, Fan JG. **Prevalence of and Risk Factors for Type 2 Diabetes Mellitus in Hyperlipidemia in China**. *Med Sci Monit* (2015) **21** 2476-84. DOI: 10.12659/MSM.894246
26. Nakaya N, Xie T, Scheerder B, Tsuchiya N, Narita A, Nakamura T. **Spousal Similarities in Cardiometabolic Risk Factors: A Cross-Sectional Comparison between Dutch and Japanese Data from Two Large Biobank Studies**. *Atherosclerosis* (2021) **334** 85-92. DOI: 10.1016/j.atherosclerosis.2021.08.037
27. Di Castelnuovo A, Quacquaruccio G, Donati MB, de Gaetano G, Iacoviello L. **Spousal Concordance for Major Coronary Risk Factors: a Systematic Review and Meta-Analysis**. *Am J Epidemiol* (2009) **169** 1-8. DOI: 10.1093/aje/kwn234
28. Nakaya N, Nakaya K, Tsuchiya N, Sone T, Kogure M, Hatanaka R. **Similarities in Cardiometabolic Risk Factors Among Random Male-Female Pairs: a Large Observational Study in Japan**. *BMC Public Health* (2022) **22** 1978. DOI: 10.1186/s12889-022-14348-6
29. Wilson SJ, Peng J, Andridge R, Jaremka LM, Fagundes CP, Malarkey WB. **For Better and Worse? the Roles of Closeness, Marital Behavior, and Age in Spouses' Cardiometabolic Similarity**. *Psychoneuroendocrinology* (2020) **120** 104777. DOI: 10.1016/j.psyneuen.2020.104777
30. Falba TA, Sindelar JL. **Spousal Concordance in Health Behavior Change**. *Health Serv Res* (2008) **43** 96-116. DOI: 10.1111/j.1475-6773.2007.00754.x
31. Shiffman D, Louie JZ, Devlin JJ, Rowland CM, Mora S. **Concordance of Cardiovascular Risk Factors and Behaviors in a Multiethnic US Nationwide Cohort of Married Couples and Domestic Partners**. *JAMA Netw Open* (2020) **3** e2022119. DOI: 10.1001/jamanetworkopen.2020.22119
32. Ma GX, Shive SE, Ma XS, Toubbeh JI, Tan Y, Lan YJ. **Social Influences on Cigarette Smoking Among Mainland Chinese and Chinese Americans: A Comparative Study**. *Am J Health Stud* (2013) **28** 12-20. PMID: 24511220
33. Swinburn BA, Sacks G, Hall KD, McPherson K, Finegood DT, Moodie ML. **The Global Obesity Pandemic: Shaped by Global Drivers and Local Environments**. *Lancet* (2011) **378** 804-14. DOI: 10.1016/S0140-6736(11)60813-1
34. Wang Z, Ji W, Song Y, Li J, Shen Y, Zheng H. **Spousal Concordance for Hypertension: A Meta-Analysis of Observational Studies**. *J Clin Hypertens (Greenwich)* (2017) **19** 1088-95. DOI: 10.1111/jch.13084
35. Kivity S, Kopel E, Maor E, Abu-Bachar F, Segev S, Sidi Y. **Association of Serum Uric Acid and Cardiovascular Disease in Healthy Adults**. *Am J Cardiol* (2013) **111** 1146-51. DOI: 10.1016/j.amjcard.2012.12.034
36. He H, Pan L, Ren X, Wang D, Du J, Cui Z. **The Effect of Body Weight and Alcohol Consumption on Hyperuricemia and Their Population Attributable Fractions: A National Health Survey in China**. *Obes Facts* (2022) **15** 216-27. DOI: 10.1159/000521163
37. Ask H, Rognmo K, Torvik FA, Roysamb E, Tambs K. **Non-random Mating and Convergence over Time for Alcohol Consumption, Smoking, and Exercise: the Nord-Trondelag Health Study**. *Behav Genet* (2012) **42** 354-65. DOI: 10.1007/s10519-011-9509-7
38. Zhao D, Liu J, Wang M, Zhang X, Zhou M. **Epidemiology of Cardiovascular Disease in China: Current Features and Implications**. *Nat Rev Cardiol* (2019) **16** 203-12. DOI: 10.1038/s41569-018-0119-4
39. Arden-Close E, McGrath N. **Health Behaviour Change Interventions for Couples: A Systematic Review**. *Br J Health Psychol* (2017) **22** 215-37. DOI: 10.1111/bjhp.12227
|
---
title: Neuronal nitric oxide synthase is required for erythropoietin stimulated erythropoiesis
in mice
authors:
- Jeeyoung Lee
- Soumyadeep Dey
- Praveen K. Rajvanshi
- Randall K. Merling
- Ruifeng Teng
- Heather M. Rogers
- Constance T. Noguchi
journal: Frontiers in Cell and Developmental Biology
year: 2023
pmcid: PMC9988911
doi: 10.3389/fcell.2023.1144110
license: CC BY 4.0
---
# Neuronal nitric oxide synthase is required for erythropoietin stimulated erythropoiesis in mice
## Abstract
Introduction: Erythropoietin (EPO), produced in the kidney in a hypoxia responsive manner, is required for red blood cell production. In non-erythroid tissue, EPO increases endothelial cell production of nitric oxide (NO) and endothelial nitric oxide synthase (eNOS) that regulates vascular tone to improve oxygen delivery. This contributes to EPO cardioprotective activity in mouse models. Nitric oxide treatment in mice shifts hematopoiesis toward the erythroid lineage, increases red blood cell production and total hemoglobin. In erythroid cells, nitric oxide can also be generated by hydroxyurea metabolism that may contribute to hydroxyurea induction of fetal hemoglobin. We find that during erythroid differentiation, EPO induces neuronal nitric oxide synthase (nNOS) and that neuronal nitric oxide synthase is required for normal erythropoietic response.
Methods: Wild type (WT) mice and mice with targeted deletion of nNOS (nNOS−/−) and eNOS (eNOS−/−) were assessed for EPO stimulated erythropoietic response. Bone marrow erythropoietic activity was assessed in culture by EPO dependent erythroid colony assay and in vivo by bone marrow transplantation into recipient WT mice. Contribution of nNOS to EPO stimulated cell proliferation was assessed in EPO dependent erythroid cells and in primary human erythroid progenitor cell cultures.
Results: EPO treatment increased hematocrit similarly in WT and eNOS−/− mice and showed a lower increase in hematocrit nNOS−/− mice. Erythroid colony assays from bone marrow cells were comparable in number from wild type, eNOS−/− and nNOS−/− mice at low EPO concentration. Colony number increased at high EPO concentration is seen only in cultures from bone marrow cells of wild type and eNOS−/− mice but not from nNOS−/− mice. Colony size with high EPO treatment also exhibited a marked increase in erythroid cultures from wild type and eNOS−/− mice but not from nNOS−/− mice. Bone marrow transplant from nNOS−/− mice into immunodeficient mice showed engraftment at comparable levels to WT bone marrow transplant. With EPO treatment, the increase in hematocrit was blunted in recipient mice that received with nNOS−/− donor marrow compared with recipient mice that received WT donor marrow. In erythroid cell cultures, addition of nNOS inhibitor resulted in decreased EPO dependent proliferation mediated in part by decreased EPO receptor expression, and decreased proliferation of hemin induced differentiating erythroid cells.
Discussion: EPO treatment in mice and in corresponding cultures of bone marrow erythropoiesis suggest an intrinsic defect in erythropoietic response of nNOS−/− mice to high EPO stimulation. Transplantation of bone marrow from donor WT or nNOS−/− mice into recipient WT mice showed that EPO treatment post-transplant recapitulated the response of donor mice. Culture studies suggest nNOS regulation of EPO dependent erythroid cell proliferation, expression of EPO receptor and cell cycle associated genes, and AKT activation. These data provide evidence that nitric oxide modulates EPO dose dependent erythropoietic response.
## 1 Introduction
Erythropoietin (EPO), a hormone produced in the kidney in a hypoxia inducible manner, is the principle regulator of red blood cell production for transport of oxygen from the lungs to the tissues (Bunn, 2013). EPO acts by binding to its cell surface EPO receptor (EPOR) on erythroid progenitor cells to promote survival, proliferation, and differentiation to produce mature erythrocytes (Bhoopalan et al., 2020). Loss of EPO or EPOR causes death in utero due to severe anemia (Wu et al., 1995; Lin et al., 1996). EPOR is also expressed in non-erythroid tissues contributing to EPO activity beyond red blood cell production (Suresh et al., 2019). Bioavailability of nitric oxide (NO) in the vasculature contributes to regulation of local blood flow affecting oxygen delivery. EPOR expression in endothelial cells mediates EPO stimulated activation of endothelial nitric oxide synthase (eNOS) and endothelial production of NO (Beleslin-Cokic et al., 2004; Mihov et al., 2009a; Premont et al., 2020). EPO stimulated increase in endothelial eNOS activation and NO production were required for EPO cardioprotection in an acute mouse model of ischemia-reperfusion injury (Mihov et al., 2009a; Teng et al., 2011). EPO also promotes angiogenesis and an increase in intracellular calcium mediated in part by transient receptor potential vanilloid type 1 (TRPV1) and required phospholipase C-gamma1 activity (Xue et al., 2011; Yu et al., 2017). Absence of EPO signaling during embryogenesis results in angiogenic defects (Kertesz et al., 2004; Watanabe et al., 2005; Nakano et al., 2007). In a mouse myocardial ischemia/reperfusion model, EPO can promote nNOS expression contributing to angiogenesis and protection against ventricular arrhythmia (Burger et al., 2009; Wen et al., 2014). The presence of eNOS in erythrocytes and the potential for EPO to activate eNOS and NO production in red blood cells further advances EPO regulation of NO bioavailability and vascular tone (Kleinbongard et al., 2006; Mihov et al., 2009b; Simmonds et al., 2014).
NO and NOS may directly affect EPO stimulated erythropoiesis. NO treatment in wild type mice shifted hematopoiesis toward the erythroid lineage and reduced leukocyte counts, mediated in part by activation of soluble guanylate cyclase (Ikuta et al., 2016). A role for NO to modify the erythroid program was suggested by hydroxyurea induction of fetal hemoglobin mediated by NO-dependent activation of soluble guanylyl cyclase (Cokic et al., 2003). In patients with end stage renal disease, levels of erythrocyte asymmetric dimethylarginine, a naturally occurring inhibitor of NOS, was associated with low hemoglobin levels and erythropoietin resistance (Yokoro et al., 2017). Similarly, a mouse model for advanced chronic kidney disease exhibited decreased hemoglobin, hematocrit and splenic EPOR gene expression as well as increased erythrocyte asymmetric dimethylarginine, suggesting that erythrocyte accumulation of asymmetric dimethylarginine and suppression of EPOR contribute to impaired erythropoietin response (Yokoro et al., 2017).
In the current study, NOS requirement for EPO stimulated erythropoiesis was determined by EPO treatment in mice with deletion of eNOS (eNOS−/−) or nNOS (nNOS−/−). nNOS−/− mice showed a blunted erythropoietic response compared with wild type (WT) mice. Cultures of hematopoietic progenitor cells from eNOS−/− and nNOS−/− mice treated with EPO confirmed the requirement of nNOS for erythroid colony formation, especially at high EPO concentration. Bone marrow transplantation provided additional evidence for nNOS activity in hematopoietic progenitor cells for normal EPO dependent erythropoietic response. Treatment of erythroid cell cultures indicated that nNOS inhibition decreased EPOR expression and EPO dependent erythroid cell proliferation mediated in part by altered cell cycle gene expression and decreased AKT activation.
## 2.1 Animals
Animal procedures were approved by the National Institute of Diabetes and Digestive and Kidney Diseases Animal Care and Use Committee and carried out in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. Mice were maintained under a specific pathogen–free and thermostable environment (23°C) and photoperiod conditions ($\frac{12}{12}$ h light/dark cycle) with free access to food (NIH-31, $14\%$ kcal/fat, 3.0 kcal/g, Teklad Diets) and water. All mice were on a C57BL/6 background. eNOS−/− mice and nNOS−/− mice were purchased from Jackson Laboratory (stock no. 002684 and 002986, respectively). nNOS−/− mice were bred by mating heterozygous females to homozygous males or by mating homozygous females to heterozygous males. Mouse genotype was determined at weaning by PCR analysis of extracted DNA. Pep Boy with CD45.2 mice (Jackson Laboratory, stock no. 002014) were used as recipient mice for bone marrow transplantation.
## 2.2 EPO treatment and hematocrit measurements
Mice (10–12 weeks old) were treated with EPO (recombinant human EPO, Epogen, Amgen, Thousand Oaks, CA, United States) at 3,000 units/kg three times/week for 7–10 days as indicated. EPO was administrated by intraperitoneal injection. For hematocrit, blood was collected from the tail vein in heparin coated capillary tubes. The tubes were centrifuged using a micro-hematocrit centrifuge (Unico, NJ, United States) and hematocrits were measured using a VIN micro-hematocrit capillary tube reader (Veterinary Information Network Bookstore, CA, United States).
## 2.3 Colony formation assay
125,000 de novo isolated cells from mouse bone marrow were plated in MethoCult M3334 (Stem Cell Technologies, Vancouver, Canada) containing SCF, IL-3, IL-6 (PeproTech, Rocky Hill, NJ, United States), and EPO (Amgen, Thousand Oaks, CA, United States) at indicated doses, and cultured at 37°C and $5\%$ CO2. Colonies were counted and scored at 14 days post-plating using phase contrast microscopy.
## 2.4 Bone marrow transplantation
Recipient female Pep Boy (CD45.2) mice at 8–10 weeks received 20 mg/kg of pharmaceutical grade busulfan by intraperitoneal injection (Hayakawa et al., 2009) 1 day prior to transplantation. Bone marrow cells were collected from female WT and nNOS−/− mice at 8–10 weeks and were infused by direct injection intravenously into the tail vein of Pep Boy mice. After 11–12 weeks following transplantation, blood was drawn from the recipient Pep Boy mice by nicking of the tail vein to analyze engraftment rate. Mice were then treated with EPO (Amgen, Thousand Oaks, CA, United States) at 3,000 units/kg three times/week for 1 week and hematocrit levels were determined.
## 2.5 FACS analysis
After erythrocyte lysis, bone marrow cells were incubated with 0.5 µg of anti-mouse CD16/CD32 antibody for blocking Fc receptors. For analyzing hematopoietic cells in the bone marrow, cells were incubated with CD45.1 (#553775, BD Bioscience) and CD 45.2 (#558707, BD Bioscience) for 30 min, followed by washing in staining buffer, and analysis by FACS Calibur (BD Bioscience).
## 2.6 HCD57 and K562 cell culture
EPO-dependent murine HCD57 erythroleukemia cells were grown in Iscove’s modified Dulbecco’s medium (IMDM), $25\%$ fetal bovine serum (FBS), 10 μg/mL gentamicin (Invitrogen, Thermo Fisher Scientific, Grand Island, NY, United States), at 37°C in a $5\%$ CO2 environment with either 0.2 U/mL or 2 U/mL EPO (Amgen, Thousand Oaks, CA, United States) (Sawyer and Jacobs-Helber, 2000). The human K562 chronic myeloid leukemia cell line was cultured in RPMI 1640 medium, $10\%$ fetal bovine serum (FBS), and $1\%$ Penicillin/Streptomycin (Invitrogen) at 37°C with $5\%$ CO2. To investigate the effect of nNOS inhibitor, HCD57 and K562 cells were cultured with nNOS inhibitor, 7-Nitroindazole (7-NI; #00240 Biotium, Fremont, CA, United States) at different concentrations (10, 100, 200 μM). HCD57 cells were harvested at day 4 for analysis of gene and protein expression. For erythroid differentiation of K562 cells, cells were stimulated with 30 μM of hemin (Millipore Sigma, St. Louis, MO, United States).
## 2.8 Human peripheral blood erythroid progenitor cell cultures
Buffy coat packs were prepared from units of whole blood collected from volunteer donors at the National Institutes of Health, Department of Transfusion Medicine. They were distributed for research use in an anonymized manner, through an exemption from full IRB review granted by the National Institutes of Health Office of Human Subjects Research Protections. Erythroid progenitors were harvested and grown in a two-phase liquid culture system (Fibach, 1998). Mononuclear cells were isolated by centrifugation on Ficoll-Hypaque (1.077 Density, Mediatech, Inc. Corning, Manassas, VA, United States). Cells were then cultured in α-minimal essential medium (αMEM) supplemented with $10\%$ fetal bovine serum (FBS), 1.5 mM L-Glutamine (100 mM), $1\%$ Penicillin/Streptomycin (Invitrogen, Thermo Fisher Scientific, Grand Island, NY, United States), $10\%$ conditioned medium from bladder carcinoma 5,637 cultures, and 1 mg/mL cyclosporin A (Millipore Sigma, St. Louis, MO, United States). Cultures were incubated at 37°C, $5\%$ CO2. After 5–7 days, non-adherent cells were washed twice with Dulbecco’s phosphate-buffered saline without Ca2+ and Mg2+ and transferred to α-minimal essential medium supplemented with $30\%$ fetal bovine serum, $1\%$ Penicillin/Streptomycin (Invitrogen), $1\%$ deionized bovine serum albumin, 10−6 M dexamethasone, 10−5 M β-mercapthoethanol, 0.3 mg/mL human holo-transferrin (Millipore Sigma), 10 ng/mL human recombinant stem cell factor (PeproTech, Rocky Hill, NJ, United States), and 2 U/mL EPO (Amgen, Thousand Oaks, CA, United States). Cultures were incubated at 37°C, $5\%$ CO2 for up to 12 days.
## 2.9 Quantitative real-time RT-PCR
Quantitative real-time RT-PCR analyses were carried out using gene-specific primers (Supplementary Table S1) and fluorescently labeled SYBR Green dye (Roche, Indianapolis, IN, United States) in a 7900 Sequence Detector (Applied Biosystems, Thermo Fisher Scientific, Foster City, CA, United States). For relative mRNA quantification, Ct values were normalized with RPL13a as an internal control using the delta-delta CT method.
## 2.10 Western blotting
Cellular proteins were extracted by AllPrep RNA/Protein Kit (Qiagen, Germantown, MD, United States). Isolated proteins were resolved by $4\%$–$20\%$ Tris-glycine SDS/PAGE, transferred to nitrocellulose membranes, blotted using an XCell SureLock Mini-Cell system (Invitrogen, Thermo Fisher Scientific, Grand Island, NY, United States) and visualized using protein specific antibodies (Supplementary Table S2). Quantitative analysis was performed by measuring the integrated density using NIH ImageJ and normalized to GAPDH.
## 2.11 Apoptosis assays
To measure the number of apoptotic cells, annexin V and propidium iodide (PI) staining and flow cytometry were used. After 7-NI treatment for 4 days, HCD57 cells were resuspended in cell staining buffer (#420201, Biolegend) and Allophycocyanin (APC)–labeled and PI (#640932, Biolegend) were added to cells and incubated for 15 min. The number of annexin V+/PI− and annexin V+/PI+ were quantified using FACS Calibur (BD Bioscience).
## 2.12 Statistical analysis
The data are expressed as mean ± s.e.m. Comparisons between two groups were made using Student’s two-tailed non-paired t-test. p values of <0.05 were regarded as statistically significant.
## 3.1 EPO treatment in nNOS−/− mice results in a blunted erythropoietic response while increased hematocrit in EPO treated eNOS−/− mice is comparable to EPO treated WT mice
EPO is the primary regulator of erythropoiesis and EPO treatment (3,000 units/kg, three times/week for 10 days) in WT mice beginning at 10 weeks of age stimulates red blood cell production and increases hematocrit from $51.5\%$ ± $1\%$ to $67.5\%$ ± $1.1\%$ (Figure 1A). To examine if nNOS and eNOS contribute to EPO stimulated erythroid response, we also treated 10-week-old eNOS−/− and nNOS−/− mice with EPO (3,000 units/kg, three times/week for 10 days). EPO treatment in eNOS−/− mice resulted in increased hematocrit from $52\%$ ± $1.2\%$ to $66\%$ ± $1.9\%$ (Figure 1A), comparable to EPO stimulated erythropoiesis in WT mice. In contrast, nNOS−/− mice showed a blunted erythropoietic response to EPO treatment with only a modest increase in hematocrit from $45\%$ ± $1.6\%$ to $49\%$ ± $1\%$ (Figure 1A). To determine if EPO stimulated erythroid response is differentially regulated in hematopoietic progenitor cells from nNOS−/− mice compared with WT and eNOS−/− mice, in vitro cultures of isolated bone marrow cells from WT, eNOS−/− and nNOS−/− mice were assessed for erythroid colony formation. At EPO concentration of 2–3 U/mL, erythroid colony formation from bone marrow cells isolated from WT, eNOS−/−, nNOS−/− mice were similar in number (133 ± 6.8, 130.25 ± 5.9 and 126 ± 6.4, respectively) and in colony size. In contrast, bone marrow cultures treated with higher dose of EPO (20 U/mL) resulted in increased erythroid colony numbers from WT and eNOS−/− mice (211 ± 7.5 and 217 ± 5.8, respectively) (Figure 1B) and corresponding increased colony size (Figure 1C), while nNOS−/− mice exhibited little or only modest increases in colony number (145 ± 2.5) and size (Figures 1B, C). Without EPO treatment, male nNOS−/− mice ($$n = 17$$) exhibit a trend toward reduced hematocrit compared with male WT mice ($$n = 24$$) that was significant between female nNOS−/− mice ($$n = 36$$) and female WT mice ($$n = 28$$) (Figure 1D).
**FIGURE 1:** *EPO stimulated erythropoiesis in nNOS−/− mice. nNOS−/− mice showed a blunted erythropoietic response to EPO while erythropoietic response in eNOS−/− mice was comparable to WT. (A) Hematocrit levels (%) in WT, eNOS−/−, and nNOS−/− mice were determined after 10 days of EPO treatment (3000 U/kg; 3 times per week). (B) Methylcellulose hematopoietic colony-forming assays were carried out using bone marrow hematopoietic cells from WT, eNOS−/−, and nNOS−/− mice and methylcellulose-based media supplemented with hematopoietic growth factors, EPO at 3 U/mL or 20 U/mL, and scored for total hematopoietic colonies after 14 days. (C) Photographs of CFU-granulocyte, erythroid, macrophage, megakaryocyte (CFU-GEMM) colonies derived from bone marrow hematopoietic cells cultured with EPO at 2 U/mL or 20 U/mL from WT, eNOS−/− and nNOS−/− mice (40X). (D) Hematocrit levels (%) in male and female WT and nNOS−/− mice (∗p < 0.05, and ∗∗p < 0.01).*
## 3.2 Mice transplanted with nNOS−/− donor marrow show reduced response and smaller increase in hematocrit with EPO treatment compared to mice transplanted with WT donor marrow
Bone marrow transplantation was used to demonstrate that the decreased EPO stimulated erythropoietic response in nNOS−/− mice was directly related to a blunted EPO stimulated erythroid response of nNOS−/− hematopoietic cells. Bone marrow cells isolated from WT (CD45.1) and nNOS−/− (CD45.1) mice were transplanted into Pep Boy mice (CD45.2) after Busulfan conditioning (Figure 2A). At 11–12 weeks posttransplant, transplanted donor marrow cell (CD45.1) from nNOS−/− mice (Figure 2B) and WT mice (Supplementary Figure S1) were counted by fluorescence activated cell sorting (FACS) analysis in peripheral blood of recipient Pep Boy mice (CD45.2). This was compared with untransplanted mice. Similar level of engraftment was observed for transplanted bone marrow cells from nNOS−/− ($81.5\%$ ± $0.7\%$) and WT mice ($82.8\%$ ± $0.7\%$). Without EPO treatment, hematocrit levels were comparable for all mouse groups, untransplanted WT and nNOS−/− mice, transplanted Pep Boy mice with WT donor and transplanted Pep Boy with nNOS−/− mice (Figure 2C). Mice were treated with EPO (3,000 U/kg; 3 times per week) for 1 week, 11–12 weeks after transplant. Transplanted mice receiving WT donor marrow showed the greatest increase in hematocrit ($65\%$ ± $1.0\%$) in contrast to the reduced increase in hematocrit with EPO treatment in transplanted mice receiving nNOS−/− donor marrow ($58.6\%$ ± $1.2\%$) (Figure 2D). This difference was comparable to hematocrits after EPO treatment in untransplanted WT mice and untransplanted nNOS−/− mice, respectively (Figure 2D). The higher hematocrit levels in EPO treated transplanted Pep Boy mice compared with untransplanted mice may reflect a more robust erythropoietic response in Pep Boy mice compared to WT mice. Engraftment of transplanted WT donor bone marrow was consistent between $80\%$ and $83\%$, while engraftment of nNOS−/− donor marrow exhibited a larger range between $70\%$ and $90\%$ and a negative correlation between hematocrit (%) and engraftment (%) (Figure 2E). Mice transplanted with nNOS−/− donor marrow that had the lowest level of % engraftment was associated with the highest hematocrit.
**FIGURE 2:** *EPO stimulated erythropoiesis in mice transplanted with nNOS−/− bone marrow. Mice transplanted with nNOS−/− bone marrow exhibited a blunted erythropoietic response to EPO treatment with a smaller increase in hematocrit compared to mice that received WT donor marrow. (A) Bone marrow from WT (CD45.1) and nNOS−/− (CD45.1) donor mice was transplanted into Pep Boy (CD45.2) recipient mice. (B) Representative FACS plots for CD45.1 and CD45.2 cells from peripheral blood of Pep Boy mice with transplant from nNOS−/− mice and without transplant. Engraftment efficiency was measured after 10–11 weeks following transplant. (C,D) Hematocrit levels (%) in WT and nNOS−/− mice and in Pep Boy recipient mice transplanted with WT and nNOS−/− donor marrow were assessed before (C) and after (D) 1 week of EPO treatment (3000 U/kg; 3 times per week). Only mice with more than 70% transplant rate were analyzed. (E) Hematocrit levels (%) were plotted with engraftment (%) for recipient Pep Boy mice with WT (black squares) and nNOS−/− (red circles) donor marrow.*
## 3.3 nNOS inhibitor decreased cell proliferation of EPO dependent mouse erythroid HCD57 cells and of hemin induced erythroid differentiating human K562 cells
Cultures of EPO dependent (HCD57) and EPO independent (K562) cells were used to determine if nNOS contributes to erythroid cell growth. Mouse erythroleukemia HCD57 cells are EPO-responsive and require EPO for survival. Human myelogenous leukemia K562 cells are EPO independent and can be chemically induced to undergo erythroid differentiation (Jacobs-Helber et al., 2000). Cultures were treated with increasing amounts of nNOS inhibitor 7-NI (10, 100, 200 μM). HCD57 cells cultured with EPO at 0.2 and 2.0 U/mL exhibited EPO concentration dependent proliferation that was inhibited by increasing concentrations of 7-NI especially at 100 and 200 μM (Figures 3A, B). The treatment with 7-NI at high dose compared with PBS after 4 days of culture in HCD57 cells increased cell death (annexin V + PI+: 7-NI 200 μM at $7.81\%$ versus PBS at $4.83\%$) and apoptosis (annexin V + PI: 7-NI 200 μM at $17.6\%$ versus PBS at $16.5\%$) in flow cytometry (Supplementary Figure S1). However, these changes appear to be small compared with the reduction in cell number by more than $50\%$. Expression of EPOR mRNA and protein showed a decreasing trend in HCD57 cell cultures treated with increasing concentrations of 7-NI (Figure 3C), consistent with the decreased growth response to 7-NI exposure. EPO independent proliferation of K562 cells was not affected by treatment with 7-NI at concentrations from 10 to 200 μM (Figure 3D), suggesting that nNOS may be required for EPO stimulated erythroid cell proliferation, but not for EPO independent erythroid cell growth. Hemin induction of erythroid differentiation of K562 cells decreased cell proliferation. Exposure of hemin induced differentiating K562 cells to 7-NI further decreased cell proliferation (Figure 3E) indicating that nNOS may contribute to maintaining cell growth and erythroid lineage expansion during erythroid differentiation.
**FIGURE 3:** *Inhibition of nNOS in proliferating erythroid cells. Treatment with nNOS inhibitor decreases proliferation of EPO-dependent HCD57 cells and proliferation of erythroid differentiating EPO-independent K562 cells. (A,B) Growth curves of EPO-dependent HCD57 cells cultured with 0.2 U/mL (A) and 2 U/mL (B) EPO were determined with addition of increasing concentrations of nNOS inhibitor 7-NI (n = 3/group; indicated are $, *, #p < 0.05, and $$, **, ##p < 0.01; $ means control vs. 10 μM 7-NI, * means control vs. 100 μM 7-NI, # means 200 μM 7-NI). (C)
EPOR gene expression (left panel) and EPOR protein expression determined by western blotting (n = 3–4/group) in HCD57 cells were assessed with EPO treatment at 2 U/mL and increasing concentration of nNOS inhibitor. HCD57 cells were harvested at day 4. (D,E) Growth curves of EPO-independent K562 cells were determined with addition of nNOS inhibitor cultured without (D) and with (E) hemin induced erythroid differentiation (n = 4/group; indicated are &, *, #p < 0.05, and &&, **, ##p < 0.01; & means control vs. hemin, * means hemin vs. 100 μM 7-NI + hemin, # means hemin vs. 200 μM 7-NI + hemin).*
## 3.4 Reduced EPO dependent erythroid cell proliferation by nNOS inhibitor is associated with cell cycle gene regulation
EPO stimulated proliferating HCD57 cells were treated with increasing concentrations of 7-NI (10, 100 and 200 μM) and changes in expression of specific cell-cycle associated genes were determined. EPO has been reported to regulate cell-cycle-associated genes such as Cyclin D2, Cyclin G2, Gspt1, Nupr1, Egr1, Nab2, Myc, p27, and Bcl6, with increasing expression of Cyclin D2, Nupr1, Egr1, Nab2, MYC, and decreasing expression of Cyclin G2 (an inhibitory Cyclin) (Fang et al., 2007). Concomitant with decreased proliferation, EPO stimulated HCD57 cells treated with nNOS inhibitor 7-NI decreased Cyclin D1, Cyclin D2, Egr1 and Nab2 expression and increased Cyclin G2 expression with Myc and Nupr1 expression unchanged (Figure 4A). Egr1 and Nab2 have been indicated in the regulation of progression from G1 to S-phase, and Cyclin G2 expression was increased upon EPO withdrawal in primary erythroblasts and was indicated to restrict proliferation potential and inhibit cell cycle progression at S-phase in UT-7/EPO cells, an EPO dependent cell line (Fang et al., 2007). EPO inhibition of Cyclin G2 may enhance DNA replication and sustain erythroblast proliferation. These data suggest that the reduced cell proliferation by nNOS inhibition may be associated with inhibition of the G1 to S-phase transition by reduction of Egr1 and Nab2 or S-phase by increased Cyclin G2. Western blotting confirmed that nNOS inhibitor treatment in HCD57 cells decreased Cyclin D1, Cyclin D2, and Egr1 protein levels, and decreased pAKT activation (Figure 4B). EPO-induced differentiation of erythroid cells has been shown to be dependent on the PI3K/Akt signaling pathway (Myklebust et al., 2002; Tothova et al., 2021). In addition, EPO has been reported to induce progression of the cell cycle through upregulation of Cyclin D3, Cyclin E and Cyclin A, and the regulation of Cyclin expression is dependent on activation of PI3- and Akt-kinase pathways (Sivertsen et al., 2006). pAKT regulates the expression of Cyclin D2, Cyclin G2, and p27 (Adlung et al., 2017). These results provide evidence that the reduced EPO dependent cell proliferation by treatment with nNOS inhibitor might be mediated by regulating cell cycle genes via the AKT-kinase pathway.
**FIGURE 4:** *nNOS and cell cycle-associated gene expression in EPO-dependent proliferating erythroid cells. Treatment with nNOS inhibitor altered cell cycle-associated gene expression in HCD57 cells. (A) Gene expression was assessed by quantitative RT-PCR analysis for cell cycle-associated genes, Cyclin D1, Cyclin D2, Egr1, Nab2, Myc, Nupr1 and Cyclin G2 in HCD57 cells cultured with EPO (2 U/mL) and increasing concentrations of nNOS inhibitor 7-NI at 0, 10, 100 and 200 μM, and harvested at day 4 (n = 3/group). (B) Protein expression of Cyclin D1, Cyclin D2, EGR1, pAKT (Ser473), AKT and GAPDH as control was assessed by Western blotting (n = 3–4/group). (Indicated are ∗ p < 0.05, and ∗∗ p < 0.01).*
## 3.5 Inhibition of nNOS decreases proliferation of differentiating human primary erythroid progenitor cells and affects cell cycle related gene expression
Primary human peripheral blood erythroid progenitor cells exhibited EPO dependent proliferation as shown for cultures treated with EPO at 0.2 U/mL (Figure 5A) and at 2 U/mL (Figure 5B), as expected (Fibach, 1998; Rogers et al., 2008). Proliferation of primary erythroid progenitor cells was sensitive to nNOS inhibition and decreased with increasing dose of 7-NI (10, 100, 200 μM; Figures 5A, B), providing further evidence that nNOS inhibition decreases EPO dependent erythroid cell proliferation as also observed with 7-NI treated proliferating HCD57 cells (Figures 3A, B). EPOR increased during EPO stimulated erythroid differentiation as reported previously (Rogers et al., 2008), but was decreased with 7-NI incubation in a dose dependent manner (Figure 5C), consistent with the decreased EPO stimulated erythroid cell proliferation (Figures 5A, B). 7-NI treatment dose-dependently modified expression of cell cycle associated genes, decreasing Cyclin D1, Egr1, and Myc expression with a tendency toward decreased Nab2 expression (Figure 5C). The decreases in proliferation and the modification of cell cycle associated gene expression by nNOS inhibition are consistent with observations in EPO stimulated HCD57 cells treated with 7-NI (Figures 3A, B, 4A). As observed in HCD57 cells, Western blotting demonstrated that nNOS inhibition in EPO stimulated differentiating human erythroid progenitor cells decreased Cyclin D2 with a trend toward decreased AKT activation (Figure 5D) in addition to decreasing cell proliferation (Figures 5A, B).
**FIGURE 5:** *nNOS modulates proliferation of differentiating human erythroid progenitor cells. Treatment with nNOS inhibitor during EPO stimulated differentiation of primary human peripheral blood erythroid progenitor cells decreased proliferation and affected cell cycle related genes. (A,B) Growth curves of primary human erythroid progenitor cells stimulated with 0.2 U/mL (A) and 2 U/mL (B) EPO were treated with increasing doses of nNOS inhibitor 7-NI (0, 10, 100, 200 μM) for 12 days (n = 3/group; indicated are $, *, #p < 0.05, and $$, **, ##p < 0.01; $ means control vs. 10 μM 7-NI, * means control vs. 100 μM 7-NI, # means 200 μM 7-NI). (C) At day 10 of culture with EPO treatment at 2 U/mL, expression of cell cycle-associated genes, Cyclin D1, Egr1, Myc and Nab2, and EPOR was determined by quantitative RT-PCR (n = 3/group). (D) Protein expression of Cyclin D2 with GAPDH as control and pAKT (Ser473)/AKT was determined by Western blotting (n = 3–4/group). (Indicated are ∗ p < 0.05, and ∗∗ p < 0.01).*
## 4 Discussion
In non-erythroid tissue, EPO regulation of nitric oxide synthase, especially eNOS and iNOS, contributes to protection of vascular endothelium and inflammation, and nNOS is involved in EPO mediated neural protection. EPO neuroprotective activity including neural cell survival and prevention of apoptosis has been shown to be mediated via NO production and neural cell expression of EPOR (Chen et al., 2010). NO regulated neural cell transcription of EPOR and NO treatment or hypoxia induced NO increased EPOR reporter gene activity. Primary mouse neural cell cultures treated with NO or subjected to hypoxia induced NO showed increased number of EPOR expressing neurons that was inhibited by treatment with nNOS inhibitor. EPO stimulation of vascular endothelium in culture and animal models promotes eNOS expression and NO production, and is associated with prevention or improvement of endothelial dysfunction (Beleslin-Cokic et al., 2004; Mihov et al., 2009a; Teng et al., 2011; Serizawa et al., 2015). In rodent models of inflammation including sepsis and seizure, EPO protective activity was associated with increased eNOS, suppression of proinflammatory cytokines, and decreased iNOS expression (Contaldo et al., 2011; Kandasamy et al., 2016; Peng et al., 2020). EPO also reduced iNOS expression in inflammation during diet induced obesity (Alnaeeli et al., 2014; Lee et al., 2021). iNOS plays a critical role in development of inflammatory response in conditions such as acute lung injury, septic shock and burn injury (Nakazawa et al., 2017; Wang et al., 2020; Golden et al., 2021). Chronic inflammation and inflammatory cytokine production contributed to decreased EPO stimulated erythropoiesis resulting in anemia of chronic disease (Weiss et al., 2019; Paulson et al., 2020a). The link between stress erythropoiesis and inflammation suggests a potential role for iNOS in stress erythropoiesis (Paulson et al., 2020b). Association between iNOS and erythropoiesis is also suggested by erythroid progenitor cells from β-thal/HbE patients that exhibited increased sensitivity in vitro to cytokine-induced apoptosis mediated by iNOS activity (Kheansaard et al., 2011). Increased pro-inflammatory cytokines resulted in a shift in hematopoietic stem cells toward myeloid lineage commitment and reduced differentiation into erythroid and lymphoid lineages (Pietras et al., 2016). In addition to inflammatory cytokines, anemia of chronic disease has been associated with other factors such as IL-33 mediated by binding to its receptor ST2 on erythroid progenitors and high mobility group box-1 protein HMGB1 binding to its receptor on erythroid precursors, that is proposed to interfere with EPO binding to EPOR (Swann et al., 2020; Dulmovits et al., 2022).
Using EPO stimulation of erythropoiesis in WT, eNOS−/− and nNOS−/− mouse models and in corresponding bone marrow cell cultures, we provide evidence that nNOS, but not eNOS, is required for normal erythropoietic response to EPO treatment (Figure 1). EPO treatment in WT and eNOS−/− mice stimulated red blood cell production and increased hematocrit while hematocrit from nNOS−/− mice showed a markedly reduced response with EPO treatment. Furthermore, cultures of bone marrow cells isolated from WT, eNOS−/− and nNOS−/− mice suggest that nNOS is required for robust erythropoietic response of hematopoietic cells to high EPO. Erythroid colony formation assay of isolated bone marrow showed high EPO increased the number of erythroid colonies and colony size in cultures from WT and eNOS−/− mice. Minimal difference was evident in erythroid colony number and colony size from nNOS−/− mice exposed to low and high EPO. These observations suggest that nNOS is essential for sensitivity of erythropoietic response to EPO level in mice in vivo and in cultures of bone marrow hematopoietic cells. Transplantation of bone marrow from donor WT and nNOS−/− mice into immunodeficient mice demonstrated in vivo that nNOS contributed importantly to erythropoietic response of bone marrow hematopoietic cells to high EPO treatment (Figure 2). In erythroid cell cultures, nNOS supports proliferation in response to EPO stimulation and during erythroid differentiation (Figure 4). Treatment with NO inhibitor 7-NI showed a dose dependent decrease in cell proliferation of EPO dependent erythroid HCD57 cells and of EPO stimulated differentiating primary human peripheral blood erythroid progenitor cells. In EPO independent K562 cells, 7-NI treatment did not affect proliferation in undifferentiated erythroid cells. Decreased EPOR expression is concomitant with the decrease in EPO stimulated proliferation with nNOS inhibition in HCD57 cells and primary human erythroid progenitor cells. This is consistent with nNOS regulation of EPOR transcription in neuronal cells (Chen et al., 2010). In contrast, K562 cells that proliferate independent of EPO express very low level of EPOR (Fraser et al., 1988) (Shinjo et al., 1997). Inhibition of nNOS in EPO stimulated HCD57 cells and primary human erythroid progenitor cells decreased proliferation concomitant with decreased AKT activation and decreased expression of proteins associated with cell cycle progression such as Cyclin D1, Cyclin D2, and EGR (Figures 4, 5). During EPO stimulation of erythroid progenitor cells, AKT activation is required for erythroid differentiation and increases phosphorylation of GATA-1 and enhances GATA-1 activity to upregulate red blood cell gene expression including EPOR (Chin et al., 1995, Zhao et al., 2006). Inhibition of nNOS reduced proliferation of hemin induced erythroid differentiating K562 cells suggesting a role for nNOS in promoting erythroid differentiation independent of EPO stimulated proliferation. In addition to its role in erythropoietic response to EPO stimulation, nNOS modulates granulopoiesis and neutrophil differentiation via NO generation (Sadaf et al., 2021). Treatment with nNOS inhibitor 7-NI in mice abrogated granulopoiesis and decreased the numbers of bone marrow progenitor and mature neutrophils. In cultures of human hematopoietic CD34+ cells and K562 cells, treatment with NO donor enhanced neutrophil differentiation, and treatment with NO inhibitor in CD34+ cells or silencing nNOS in K562 cells reduced neutrophil differentiation. These data together with the blunted erythropoietic EPO response of nNOS−/− mice suggest that nNOS contributes to both erythroid and myeloid differentiation of hematopoietic stem/progenitor cells that includes EPO stimulated erythropoiesis, as well as granulopoiesis and neutrophil differentiation.
A critical role for nNOS in proliferation and tissue development in non-erythroid tissues is exemplified by bone formation in nNOS−/− mice that show reduced chondrocyte proliferation and bone growth (Yan et al., 2012). nNOS−/− mice exhibit a reduction in growth plate replicating cells with decreased Cyclin D1, slower cell cycle progression and premature cell cycle exit, thinner cortical bone and fewer trabeculae. In the nervous system, rat Schwan cells treated with nNOS inhibitor showed arrested cell cycle progression and decreased proliferating cell nuclear antigen levels (Shen et al., 2008). The proliferative activity associated with nNOS is not observed with eNOS. A major function of eNOS is regulation of vascular tone via NO production and of vascular endothelial growth factor induced angiogenesis (Fukumura et al., 2001; Smith et al., 2021). In a rodent model of hind limb ischemia, intramuscular gene transfer of an eNOS expression vector increased eNOS, NO, vascular endothelial growth factor and angiogenesis (Namba et al., 2003). In contrast, overexpression of eNOS directly in endothelial cell cultures inhibited endothelial cell proliferation and knockdown of eNOS increased proliferation, reduced cell cycle inhibitor p21 and increased proliferation marker Ki67, but reduced angiogenesis (Kader et al., 2000; Bu et al., 2022). Post-translational mechanisms and proper localization to intracellular compartments contribute further to nNOS and eNOS regulation. For example, NOSIP can negatively modulate NOS activity and affect endothelial cell eNOS translocation from the plasma membrane and other subcellular compartments to impair NO production and promote cell cycle regulated inactivation of eNOS and impaired NO production (Dedio et al., 2001; Schleicher et al., 2005). In neuronal cells, NOSIP influences nNOS subcellular distribution and may regulate synaptic availability and activity of NOS to protect against excessive NO production in neurons (Dreyer et al., 2004). In hematopoietic cells, NOSIP impacts on nNOS induction of neutrophil differentiation (Sadaf et al., 2021).
The range of processes linked to NO that include cell proliferation, modulation of cell cycle, angiogenesis and inflammation, appear to be differentially regulated by NO concentration that may reflect, in part, varying sensitivity to NO of specific proteins such as AKT, ERK, HIF1α, p53 and caspase (Thomas et al., 2008). Generally lower NO concentrations have been linked to survival and proliferation while higher NO concentration has been linked to cell cycle arrest and apoptosis. Furthermore, low NO concentrations can promote mitochondria respiration that is inhibited at high concentrations (Bailey et al., 2019; Dynnik et al., 2020). Here, we provide evidence that nNOS is a critical contributor to regulation of EPO dependent erythroid cell proliferation and EPOR expression, and expression of cell cycle associated genes. More detailed studies are required to determine the role of nNOS translocation during EPO stimulated erythropoiesis and of nNOS activation of the erythroid program including EPOR expression during differentiation.
## 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 animal study was reviewed and approved by the National Institute of Diabetes and Digestive and Kidney Diseases Animal Care and Use Committee.
## Author contributions
JL designed the experiments, conducted the studies, analyzed the data, and wrote the manuscript. SD, PR, and HR designed the experiments, conducted the studies, analyzed the data, and reviewed and edited the manuscript. RM and RT contributed to experimental design, data generation, and reviewed and edited the manuscript. CN contributed to the experimental design and discussion of the data and wrote 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.2023.1144110/full#supplementary-material
## References
1. Adlung L., Kar S., Wagner M. C., She B., Chakraborty S., Bao J.. **Protein abundance of AKT and ERK pathway components governs cell type-specific regulation of proliferation**. *Mol. Syst. Biol.* (2017) **13** 904. DOI: 10.15252/msb.20167258
2. Alnaeeli M., Raaka B. M., Gavrilova O., Teng R., Chanturiya T., Noguchi C. T.. **Erythropoietin signaling: A novel regulator of white adipose tissue inflammation during diet-induced obesity**. *Diabetes* (2014) **63** 2415-2431. DOI: 10.2337/db13-0883
3. Bailey J. D., Diotallevi M., Nicol T., Mcneill E., Shaw A., Chuaiphichai S.. **Nitric oxide modulates metabolic remodeling in inflammatory macrophages through TCA cycle regulation and itaconate accumulation**. *Cell Rep.* (2019) **28** 218-230 e7. DOI: 10.1016/j.celrep.2019.06.018
4. Beleslin-Cokic B. B., Cokic V. P., Yu X., Weksler B. B., Schechter A. N., Noguchi C. T.. **Erythropoietin and hypoxia stimulate erythropoietin receptor and nitric oxide production by endothelial cells**. *Blood* (2004) **104** 2073-2080. DOI: 10.1182/blood-2004-02-0744
5. Bhoopalan S. V., Huang L. J., Weiss M. J.. **Erythropoietin regulation of red blood cell production: From bench to bedside and back**. *F1000Res* (2020) **9** F1000. DOI: 10.12688/f1000research.26648.1
6. Bu S., Nguyen H. C., Nikfarjam S., Michels D. C. R., Rasheed B., Maheshkumar S.. **Endothelial cell-specific loss of eNOS differentially affects endothelial function**. *PLoS One* (2022) **17** e0274487. DOI: 10.1371/journal.pone.0274487
7. Bunn H. F.. **Erythropoietin**. *Cold Spring Harb. Perspect. Med.* (2013) **3** a011619. DOI: 10.1101/cshperspect.a011619
8. Burger D. E., Xiang F. L., Hammoud L., Jones D. L., Feng Q.. **Erythropoietin protects the heart from ventricular arrhythmia during ischemia and reperfusion via neuronal nitric-oxide synthase**. *J. Pharmacol. Exp. Ther.* (2009) **329** 900-907. DOI: 10.1124/jpet.109.150896
9. Chen Z. Y., Wang L., Asavaritkrai P., Noguchi C. T.. **Up-regulation of erythropoietin receptor by nitric oxide mediates hypoxia preconditioning**. *J. Neurosci. Res.* (2010) **88** 3180-3188. DOI: 10.1002/jnr.22473
10. Chin K., Oda N., Shen K., Noguchi C. T.. **Regulation of transcription of the human erythropoietin receptor gene by proteins binding to GATA-1 and Sp1 motifs**. *Nucleic Acids Res.* (1995) **23** 3041-3049. DOI: 10.1093/nar/23.15.3041
11. Cokic V. P., Smith R. D., Beleslin-Cokic B. B., Njoroge J. M., Miller J. L., Gladwin M. T.. **Hydroxyurea induces fetal hemoglobin by the nitric oxide-dependent activation of soluble guanylyl cyclase**. *J. Clin. Invest.* (2003) **111** 231-239. DOI: 10.1172/JCI16672
12. Contaldo C., Lindenblatt N., Elsherbiny A., Hogger D. C., Borozadi M. K., Vetter S. T.. **Erythropoietin requires endothelial nitric oxide synthase to counteract TNF-[alpha]-induced microcirculatory dysfunction in murine striated muscle**. *Shock* (2011) **35** 315-321. DOI: 10.1097/SHK.0b013e3181fd0700
13. Dedio J., Konig P., Wohlfart P., Schroeder C., Kummer W., Muller-Esterl W.. **NOSIP, a novel modulator of endothelial nitric oxide synthase activity**. *FASEB J.* (2001) **15** 79-89. DOI: 10.1096/fj.00-0078com
14. Dreyer J., Schleicher M., Tappe A., Schilling K., Kuner T., Kusumawidijaja G.. **Nitric oxide synthase (NOS)-interacting protein interacts with neuronal NOS and regulates its distribution and activity**. *J. Neurosci.* (2004) **24** 10454-10465. DOI: 10.1523/JNEUROSCI.2265-04.2004
15. Dulmovits B. M., Tang Y., Papoin J., He M., Li J., Yang H.. **HMGB1-mediated restriction of EPO signaling contributes to anemia of inflammation**. *Blood* (2022) **139** 3181-3193. DOI: 10.1182/blood.2021012048
16. Dynnik V. V., Grishina E. V., Fedotcheva N. I.. **The mitochondrial NO-synthase/guanylate cyclase/protein kinase G signaling system underpins the dual effects of nitric oxide on mitochondrial respiration and opening of the permeability transition pore**. *FEBS J.* (2020) **287** 1525-1536. DOI: 10.1111/febs.15090
17. Fang J., Menon M., Kapelle W., Bogacheva O., Bogachev O., Houde E.. **EPO modulation of cell-cycle regulatory genes, and cell division, in primary bone marrow erythroblasts**. *Blood* (2007) **110** 2361-2370. DOI: 10.1182/blood-2006-12-063503
18. Fibach E.. **Techniques for studying stimulation of fetal hemoglobin production in human erythroid cultures**. *Hemoglobin* (1998) **22** 445-458. DOI: 10.3109/03630269809071542
19. Fraser J. K., Lin F. K., Berridge M. V.. **Expression and modulation of specific, high affinity binding sites for erythropoietin on the human erythroleukemic cell line K562**. *Blood* (1988) **71** 104-109. DOI: 10.1182/blood.v71.1.104.bloodjournal711104
20. Fukumura D., Gohongi T., Kadambi A., Izumi Y., Ang J., Yun C. O.. **Predominant role of endothelial nitric oxide synthase in vascular endothelial growth factor-induced angiogenesis and vascular permeability**. *Proc. Natl. Acad. Sci. U. S. A.* (2001) **98** 2604-2609. DOI: 10.1073/pnas.041359198
21. Golden T. N., Venosa A., Gow A. J.. **Cell origin and iNOS function are critical to macrophage activation following acute lung injury**. *Front. Pharmacol.* (2021) **12** 761496. DOI: 10.3389/fphar.2021.761496
22. Hayakawa J., Hsieh M. M., Uchida N., Phang O., Tisdale J. F.. **Busulfan produces efficient human cell engraftment in NOD/LtSz-Scid IL2Rgamma(null) mice**. *Stem Cells* (2009) **27** 175-182. DOI: 10.1634/stemcells.2008-0583
23. Ikuta T., Sellak H., Odo N., Adekile A. D., Gaensler K. M.. **Nitric oxide-cGMP signaling stimulates erythropoiesis through multiple lineage-specific transcription factors: Clinical implications and a novel target for erythropoiesis**. *PLoS One* (2016) **11** e0144561. DOI: 10.1371/journal.pone.0144561
24. Jacobs-Helber S. M., Ryan J. J., Sawyer S. T.. **JNK and p38 are activated by erythropoietin (EPO) but are not induced in apoptosis following EPO withdrawal in EPO-dependent HCD57 cells**. *Blood* (2000) **96** 933-940. DOI: 10.1182/blood.v96.3.933.015k52_933_940
25. Kader K. N., Akella R., Ziats N. P., Lakey L. A., Harasaki H., Ranieri J. P.. **eNOS-overexpressing endothelial cells inhibit platelet aggregation and smooth muscle cell proliferation**. *Tissue Eng.* (2000) **6** 241-251. DOI: 10.1089/10763270050044425
26. Kandasamy K., Choudhury S., Singh V., Addison M. P., Darzi S. A., Kasa J. K.. **Erythropoietin reverses sepsis-induced vasoplegia to norepinephrine through preservation of α1d-adrenoceptor mRNA expression and inhibition of GRK2-mediated desensitization in mouse aorta**. *J. Cardiovasc Pharmacol. Ther.* (2016) **21** 100-113. DOI: 10.1177/1074248415587968
27. Kertesz N., Wu J., Chen T. H., Sucov H. M., Wu H.. **The role of erythropoietin in regulating angiogenesis**. *Dev. Biol.* (2004) **276** 101-110. DOI: 10.1016/j.ydbio.2004.08.025
28. Kheansaard W., Panichob P., Fucharoen S., Tanyong D. I.. **Cytokine-induced apoptosis of beta-thalassemia/hemoglobin E erythroid progenitor cells via nitric oxide-mediated process**. *Acta Haematol.* (2011) **126** 224-230. DOI: 10.1159/000329903
29. Kleinbongard P., Schulz R., Rassaf T., Lauer T., Dejam A., Jax T.. **Red blood cells express a functional endothelial nitric oxide synthase**. *Blood* (2006) **107** 2943-2951. DOI: 10.1182/blood-2005-10-3992
30. Lee J., Walter M. F., Korach K. S., Noguchi C. T.. **Erythropoietin reduces fat mass in female mice lacking estrogen receptor alpha**. *Mol. Metab.* (2021) **45** 101142. DOI: 10.1016/j.molmet.2020.101142
31. Lin C. S., Lim S. K., D'Agati V., Costantini F.. **Differential effects of an erythropoietin receptor gene disruption on primitive and definitive erythropoiesis**. *Genes Dev.* (1996) **10** 154-164. DOI: 10.1101/gad.10.2.154
32. Mihov D., Bogdanov N., Grenacher B., Gassmann M., Zund G., Bogdanova A.. **Erythropoietin protects from reperfusion-induced myocardial injury by enhancing coronary endothelial nitric oxide production**. *Eur. J. Cardiothorac. Surg.* (2009a) **35** 839. DOI: 10.1016/j.ejcts.2008.12.049
33. Mihov D., Vogel J., Gassmann M., Bogdanova A.. **Erythropoietin activates nitric oxide synthase in murine erythrocytes**. *Am. J. Physiol. Cell Physiol.* (2009b) **297** C378-C388. DOI: 10.1152/ajpcell.00543.2008
34. Myklebust J. H., Blomhoff H. K., Rusten L. S., Stokke T., Smeland E. B.. **Activation of phosphatidylinositol 3-kinase is important for erythropoietin-induced erythropoiesis from CD34(+) hematopoietic progenitor cells**. *Exp. Hematol.* (2002) **30** 990-1000. DOI: 10.1016/s0301-472x(02)00868-8
35. Nakano M., Satoh K., Fukumoto Y., Ito Y., Kagaya Y., Ishii N.. **Important role of erythropoietin receptor to promote VEGF expression and angiogenesis in peripheral ischemia in mice**. *Circ. Res.* (2007) **100** 662-669. DOI: 10.1161/01.RES.0000260179.43672.fe
36. Nakazawa H., Chang K., Shinozaki S., Yasukawa T., Ishimaru K., Yasuhara S.. **iNOS as a driver of inflammation and apoptosis in mouse skeletal muscle after burn injury: Possible involvement of sirt1 S-Nitrosylation-Mediated acetylation of p65 NF-κB and p53**. *PLoS One* (2017) **12** e0170391. DOI: 10.1371/journal.pone.0170391
37. Namba T., Koike H., Murakami K., Aoki M., Makino H., Hashiya N.. **Angiogenesis induced by endothelial nitric oxide synthase gene through vascular endothelial growth factor expression in a rat hindlimb ischemia model**. *Circulation* (2003) **108** 2250-2257. DOI: 10.1161/01.CIR.0000093190.53478.78
38. Paulson R. F., Hariharan S., Little J. A.. **Stress erythropoiesis: Definitions and models for its study**. *Exp. Hematol.* (2020a) **89** 43-54 e2. DOI: 10.1016/j.exphem.2020.07.011
39. Paulson R. F., Ruan B., Hao S., Chen Y.. **Stress erythropoiesis is a key inflammatory response**. *Cells* (2020b) **9** 634. DOI: 10.3390/cells9030634
40. Peng B., Kong G., Yang C., Ming Y.. **Erythropoietin and its derivatives: From tissue protection to immune regulation**. *Cell Death Dis.* (2020) **11** 79. DOI: 10.1038/s41419-020-2276-8
41. Pietras E. M., Mirantes-Barbeito C., Fong S., Loeffler D., Kovtonyuk L. V., Zhang S.. **Chronic interleukin-1 exposure drives haematopoietic stem cells towards precocious myeloid differentiation at the expense of self-renewal**. *Nat. Cell Biol.* (2016) **18** 607-618. DOI: 10.1038/ncb3346
42. Premont R. T., Reynolds J. D., Zhang R., Stamler J. S.. **Role of nitric oxide carried by hemoglobin in cardiovascular physiology: Developments on a three-gas respiratory cycle**. *Circ. Res.* (2020) **126** 129-158. DOI: 10.1161/CIRCRESAHA.119.315626
43. Rogers H. M., Yu X., Wen J., Smith R., Fibach E., Noguchi C. T.. **Hypoxia alters progression of the erythroid program**. *Exp. Hematol.* (2008) **36** 17-27. DOI: 10.1016/j.exphem.2007.08.014
44. Sadaf S., Nagarkoti S., Awasthi D., Singh A. K., Srivastava R. N., Kumar S.. **nNOS induction and NOSIP interaction impact granulopoiesis and neutrophil differentiation by modulating nitric oxide generation**. *Biochim. Biophys. Acta Mol. Cell Res.* (2021) **1868** 119018. DOI: 10.1016/j.bbamcr.2021.119018
45. Sawyer S. T., Jacobs-Helber S. M.. **Unraveling distinct intracellular signals that promote survival and proliferation: Study of erythropoietin, stem cell factor, and constitutive signaling in leukemic cells**. *J. Hematother Stem Cell Res.* (2000) **9** 21-29. DOI: 10.1089/152581600319586
46. Schleicher M., Brundin F., Gross S., Muller-Esterl W., Oess S.. **Cell cycle-regulated inactivation of endothelial NO synthase through NOSIP-dependent targeting to the cytoskeleton**. *Mol. Cell Biol.* (2005) **25** 8251-8258. DOI: 10.1128/MCB.25.18.8251-8258.2005
47. Serizawa K., Yogo K., Tashiro Y., Aizawa K., Kawasaki R., Hirata M.. **Epoetin beta pegol prevents endothelial dysfunction as evaluated by flow-mediated dilation in chronic kidney disease rats**. *Eur. J. Pharmacol.* (2015) **767** 10-16. DOI: 10.1016/j.ejphar.2015.09.034
48. Shen A., Gao S., Ben Z., Wang H., Jia J., Tao T.. **Identification and potential role of PSD-95 in Schwann cells**. *Neurol. Sci.* (2008) **29** 321-330. DOI: 10.1007/s10072-008-0989-z
49. Shinjo K., Takeshita A., Higuchi M., Ohnishi K., Ohno R.. **Erythropoietin receptor expression on human bone marrow erythroid precursor cells by a newly-devised quantitative flow-cytometric assay**. *Br. J. Haematol.* (1997) **96** 551-558. DOI: 10.1046/j.1365-2141.1997.d01-2071.x
50. Simmonds M. J., Detterich J. A., Connes P.. **Nitric oxide, vasodilation and the red blood cell**. *Biorheology* (2014) **51** 121-134. DOI: 10.3233/BIR-140653
51. Sivertsen E. A., Hystad M. E., Gutzkow K. B., Dosen G., Smeland E. B., Blomhoff H. K.. **PI3K/Akt-dependent Epo-induced signalling and target genes in human early erythroid progenitor cells**. *Br. J. Haematol.* (2006) **135** 117-128. DOI: 10.1111/j.1365-2141.2006.06252.x
52. Smith T. L., Oubaha M., Cagnone G., Boscher C., Kim J. S., El Bakkouri Y.. **eNOS controls angiogenic sprouting and retinal neovascularization through the regulation of endothelial cell polarity**. *Cell Mol. Life Sci.* (2021) **79** 37. DOI: 10.1007/s00018-021-04042-y
53. Suresh S., Rajvanshi P. K., Noguchi C. T.. **The many facets of erythropoietin physiologic and metabolic response**. *Front. Physiol.* (2019) **10** 1534. DOI: 10.3389/fphys.2019.01534
54. Swann J. W., Koneva L. A., Regan-Komito D., Sansom S. N., Powrie F., Griseri T.. **IL-33 promotes anemia during chronic inflammation by inhibiting differentiation of erythroid progenitors**. *J. Exp. Med.* (2020) **217** e20200164. DOI: 10.1084/jem.20200164
55. Teng R., Calvert J. W., Sibmooh N., Piknova B., Suzuki N., Sun J.. **Acute erythropoietin cardioprotection is mediated by endothelial response**. *Basic Res. Cardiol.* (2011) **106** 343-354. DOI: 10.1007/s00395-011-0158-z
56. Thomas D. D., Ridnour L. A., Isenberg J. S., Flores-Santana W., Switzer C. H., Donzelli S.. **The chemical biology of nitric oxide: Implications in cellular signaling**. *Free Radic. Biol. Med.* (2008) **45** 18-31. DOI: 10.1016/j.freeradbiomed.2008.03.020
57. Tothova Z., Semelakova M., Solarova Z., Tomc J., Debeljak N., Solar P.. **The role of PI3K/AKT and MAPK signaling pathways in erythropoietin signalization**. *Int. J. Mol. Sci.* (2021) **22** 7682. DOI: 10.3390/ijms22147682
58. Wang Y., Wang K., Fu J.. **HDAC6 mediates macrophage iNOS expression and excessive nitric oxide production in the blood during endotoxemia**. *Front. Immunol.* (2020) **11** 1893. DOI: 10.3389/fimmu.2020.01893
59. Watanabe D., Suzuma K., Matsui S., Kurimoto M., Kiryu J., Kita M.. **Erythropoietin as a retinal angiogenic factor in proliferative diabetic retinopathy**. *N. Engl. J. Med.* (2005) **353** 782-792. DOI: 10.1056/NEJMoa041773
60. Weiss G., Ganz T., Goodnough L. T.. **Anemia of inflammation**. *Blood* (2019) **133** 40-50. DOI: 10.1182/blood-2018-06-856500
61. Wen C. T., He T., Xing Y. Q.. **Erythropoietin promotes retinal angiogenesis in a mouse model**. *Mol. Med. Rep.* (2014) **10** 2979-2984. DOI: 10.3892/mmr.2014.2593
62. Wu H., Liu X., Jaenisch R., Lodish H. F.. **Generation of committed erythroid BFU-E and CFU-E progenitors does not require erythropoietin or the erythropoietin receptor**. *Cell* (1995) **83** 59-67. DOI: 10.1016/0092-8674(95)90234-1
63. Xue Y., Lim S., Yang Y., Wang Z., Jensen L. D., Hedlund E. M.. **PDGF-BB modulates hematopoiesis and tumor angiogenesis by inducing erythropoietin production in stromal cells**. *Nat. Med.* (2011) **18** 100-110. DOI: 10.1038/nm.2575
64. Yan Q., Feng Q., Beier F.. **Reduced chondrocyte proliferation, earlier cell cycle exit and increased apoptosis in neuronal nitric oxide synthase-deficient mice**. *Osteoarthr. Cartil.* (2012) **20** 144-151. DOI: 10.1016/j.joca.2011.11.014
65. Yokoro M., Nakayama Y., Yamagishi S. I., Ando R., Sugiyama M., Ito S.. **Asymmetric dimethylarginine contributes to the impaired response to erythropoietin in CKD-anemia**. *J. Am. Soc. Nephrol.* (2017) **28** 2670-2680. DOI: 10.1681/ASN.2016111184
66. Yu Y. B., Su K. H., Kou Y. R., Guo B. C., Lee K. I., Wei J.. **Role of transient receptor potential vanilloid 1 in regulating erythropoietin-induced activation of endothelial nitric oxide synthase**. *Acta Physiol. (Oxf)* (2017) **219** 465-477. DOI: 10.1111/apha.12723
67. Zhao W., Kitidis C., Fleming M. D., Lodish H. F., Ghaffari S.. **Erythropoietin stimulates phosphorylation and activation of GATA-1 via the PI3-kinase/AKT signaling pathway**. *Blood* (2006) **107** 907-915. DOI: 10.1182/blood-2005-06-2516
|
---
title: 'Spontaneous NETosis in diabetes: A role of hyperglycemia mediated ROS and
autophagy'
authors:
- Anam Farhan
- Ghulam Hassan
- Sheikha Hina Liaqat Ali
- Zainab Yousaf
- Kandeel Shafique
- Amir Faisal
- Bilal bin Younis
- Shaper Mirza
journal: Frontiers in Medicine
year: 2023
pmcid: PMC9988915
doi: 10.3389/fmed.2023.1076690
license: CC BY 4.0
---
# Spontaneous NETosis in diabetes: A role of hyperglycemia mediated ROS and autophagy
## Abstract
Type 2-diabetes, particularly poorly controlled diabetes, is a risk factor for several infections such as lower respiratory tract and skin infections. Hyperglycemia, a characteristic downstream effect of poorly controlled diabetes, has been shown to impair the function of immune cells, in particular neutrophils. Several studies have demonstrated that hyperglycemia-mediated priming of NADPH oxidase results in subsequent elevated levels of reactive oxygen species (ROS). In healthy neutrophils, ROS plays an important role in pathogen killing by phagocytosis and by induction of Neutrophil Extracellular Traps (NETs). Given the key role of ROS in autophagy, phagocytosis and NETosis, the relationship between these pathways and the role of diabetes in the modulation of these pathways has not been explored previously. Therefore, our study aimed to understand the relationship between autophagy, phagocytosis and NETosis in diabetes. We hypothesized that hyperglycemia-associated oxidative stress alters the balance between phagocytosis and NETosis by modulating autophagy. Using whole blood samples from individuals with and without type 2-diabetes (in the presence and absence of hyperglycemia), we demonstrated that (i) hyperglycemia results in elevated levels of ROS in neutrophils from those with diabetes, (ii) elevated levels of ROS increase LCIII (a marker for autophagy) and downstream NETosis. ( iii) Diabetes was also found to be associated with low levels of phagocytosis and phagocytic killing of S. pneumoniae. ( iv) Blocking either NADPH oxidase or cellular pathways upstream of autophagy led to a significant reduction in NETosis. This study is the first to demonstrate the role of ROS in altering NETosis and phagocytosis by modulating autophagy in type 2-diabetes.
GRAPHICAL ABSTRACT
## 1. Introduction
Type 2-diabetes mellitus (T2D), a metabolic syndrome characterized by chronic hyperglycemia, remains one of the largest emerging threats to global health in the 21st century. During the last decade, evidence has emerged linking T2D to an increased risk of several infections (1–3), of which the most important are respiratory tract infections [4, 5] including tuberculosis (TB), pneumonia and influenza [6, 7]. In addition to respiratory tract infections, diabetes is also associated with a 1.5-fold increased risk of surgical site infections [8], a 2-fold increased risk of urinary tract infections (UTIs) [9], and a 2-to 3-fold increased risk of bacteremia [10]. This increased susceptibility is attributed to impairments in sentinel cells of the immune system, such as neutrophils [11].
Neutrophils (PMNs), one of the principal effector cells of the immune system, are short-lived phagocytes [12]. They are known for their role in the surveillance and killing of invading microbes. By adopting several strategies, these leukocytes provide defense against a broad range of pathogens, thus limiting their spread [13]. Recently, new horizons of neutrophil activity have been revealed, suggesting their long-term involvement in infection and inflammation [14, 15]. Brinkmann and co-workers identified a novel strategy of trapping and killing microbes by neutrophils and coined the term neutrophil extracellular trap (NET) [16]. While phagocytosis requires engulfing and killing extracellular bacteria, in NETosis, neutrophils undergo multiple morphological changes to release NETs that are characterized by a web of DNA decorated with cytoplasmic antimicrobial compounds (17–19).
While the molecular mechanisms driving NETosis remain largely unknown, the role of NADPH oxidase-dependent ROS in the induction of NETosis is well established (20–22). The NADPH oxidase is a multi-subunit enzyme complex, with subunits localized in both cytoplasm and cell membrane. Upon activation, cytosolic subunits migrate to the membrane resulting in the formation of an active enzyme complex [23]. Reactive oxygen species thus generated are considered an important second messenger required for NETs generation and several other processes, including phagocytosis and autophagy [24, 25]. Autophagy is a well-conserved cellular mechanism that, under certain conditions, allows for the degradation of cytoplasmic material and organelles to recycle their constituents. It also plays a significant role in infection and inflammation, eliminating phagocytosed microbes and limiting inflammasome activation [26, 27]. Autophagy-related signaling has been implicated in NETosis [28, 29].
NETosis is a physiological process that, although exacerbated in diabetes, is instrumental to protecting against invasive infections [22]. However, exacerbation of NETs in T2D is far from being completely understood. It has recently been described that high glucose in vitro and hyperglycemia in vivo increase the release of NETs, characterized by elevated levels of circulating peptidyl-arginine-deiminase, an enzyme important in chromatin decondensation and DNA release, in T2D (30–34). However, these studies did not address the biological processes associated with the observed increase in NETosis, neither did these studies elaborate on associations between hyperglycemia, NETosis, phagocytosis and autophagy. It is plausible that, NETosis and autophagy operates as partners for carrying out antibacterial activity in a coordinated/cooperative manner.
While NETs formation is intended to have protective effect against pathogen, it can be detrimental to the surrounding tissues (35–37). NETs have been shown to cause both epithelial and endothelial cytotoxicity to respiratory cells during pneumonia [38, 39]. It is therefore likely that high bacterial burden and toxicity from protective mechanisms like NETosis may lead to poor outcomes of pneumonia in those with diabetes. It therefore makes it imperative to understand these mechanisms in diabetes and identify relevant molecules and pathways that can be harnessed to reduce the burden of infection in those with diabetes.
The current study is therefore designed to expand our knowledge regarding the molecular cross-talk among pathways that regulate neutrophil antibacterial activity. This understanding will likely facilitate manipulation of the neutrophil machinery in chronic diseases such as diabetes for successful design of future inflammation modulatory therapeutics.
## 2. Materials and methods
The human subject’s research has been approved by the Lahore University of Management Sciences (LUMS) and by Shalamar Hospital Institutional Review Boards (IRB) to ensure risks to humans are minimized. Written informed consent were obtained from all the participants, and appropriately documented according to all rules and regulations for compliance with the University IRB.
## 2.1. Sample size
Primary goal of this study is to determine the role of hyperglycemia mediated oxidative stress on activation of autophagy and subsequent NETosis in individuals with and without type 2-diabetes. Using convenient sampling, blood samples were collected from a total of 60 participants, which included 30 individuals with diagnosed T2D and 30 age and sex matched healthy control.
## 2.1.1. Inclusion criteria
Individuals 18 years and older diagnosed with type 2-diabetes were included in the study. Diabetes among study participants was either self-reported or doctor diagnosed. For diagnosis, diabetes was defined as follows: A fasting blood glucose of >100 mg/dl, glycated hemoglobin (HbA1c) of >$6.5\%$ and on medication for glucose (ADA guidelines 2010). Poorly controlled diabetes was further defined as fasting blood glucose of >126 mg/dl and HbA1c of >$8\%$ and on medication for glucose. Healthy control were defined as; age and sex matched individuals with no diagnosis of diabetes HbA1c <$5.6\%$, fasting blood glucose of <100 mg/dl and not on medication for diabetes.
## 2.1.2. Exclusion criteria
Participants were excluded from the study if they confirmed to any of the following (i) Individuals on antibiotics for any current infections (ii) diagnosed with tuberculosis and on ATT and (iii) HIV positive individuals.
## 2.2. Recruitment and informed consent
Participants who confirmed to our inclusion and exclusion criteria were approached and were requested to participate in the study. Participants with type 2-diabetes were recruited from Sakina Institute of Diabetes and Endocrinology Research (SiDER) at Shalamar Hospital, a 250 bed private tertiary care hospital in the south of the city. Whereas, age and sex matched healthy controls were recruited from either SiDER or from outpatient clinic at the department of medicine, Shalamar Hospital. Participants presenting at SiDER diabetes clinic or outpatient clinic in the department of medicine were invited to participate. Study was explained to each participant in local language that they were most familiar with, which include either Urdu (national language) or Punjabi (language of the natives of province of Punjab). Individuals who agreed to participate in the study were given a consent form in the language that they were most comfortable with (English, Urdu or Punjabi). Following signing of consent forms, a health questionnaire was administered by the project coordinator. The questionnaire was designed to obtain information on demographic, anthropometric and health related variables. After filling out of the questionnaire, participants were requested to donate at least 10 ml of blood. A computer generated, unique study code was assigned to every participant. Data collected from participants was saved in an excel sheet which was only accessible to the PI and project coordinator. The laptop computer was password protected and password was only provided to study coordinator and the Principal Investigator of the study. Confidentiality of all subjects was completely ensured.
## 2.3. Sample processing
A total of 10 ml of blood was drawn using 18-gauge needle and immediately transferred into a 15 ml conical tube containing 10 μls of heparin sulphate (heparin concentration is approximately 15 USP (US Pharmacopeia) units of heparin per milliliter of blood) to prevent coagulation. Blood was transported to the laboratory at the Department of Life Sciences, Lahore University of Management Sciences (LUMS), within 2 h of bleeding. Blood was processed as follows:
## 2.3.1. Purification of peripheral blood mononuclear cells and granulocytes cells from whole blood
A total of 10 ml of whole blood collected from individuals with T2D or healthy volunteers, was layered on equal volume of a sucrose gradient, (Polymorphprep - Axis Shield UK Co) for separation of peripheral blood mononuclear cells (PBMC) and neutrophils (PMNs). The gradient in polymorphprep allows for the separation of PBMC and PMNs into two separate layers. PMNs obtained using this method are $99\%$ pure. On average 4-6 × 106 cells/ml RPMI were isolated from 10 ml of blood. Purified PMNs were re-suspended in 1 × PBS (pH 7.0) for the measurement of ROS and the remaining PMNs were re-suspended in assay medium (RPMI supplemented with autologous plasma) for use in NETs induction. PMNs were kept at room temperature and used the same day.
## 2.3.2. Cells counting and cell number adjustment
Washed PMNs were re-suspended in RPMI (Sigma 1,640) and cell viability was determined using trypan blue solution (Gibco™, $0.4\%$), cells were suspended in RPMI+ 10 mM HEPES and $10\%$ autologous human serum and were adjusted to the final concentration of 1×105 PMNs/ml RPMI.
## 2.3.3. Glucose treatment of neutrophils (PMNs)
For both ROS generation and induction of NETs, 1 × 105 PMNs/ml RPMI were treated with 5 and 15 mM glucose for varying time intervals which are explained below alongside each experiment.
## 2.4. Advanced glycation end products preparation
Advanced glycation end products were generated by incubating bovine serum albumin (20 mg/ml) with 5 M D-Glucose and 0.2 M phosphate buffer (pH.7.4) for 60 days at 60°C under sterilized conditions. After incubation, amount of glycated BSA was 20 mg/ml. Sample was lyophilized, followed by dialysis against un-bound salt using Slide-A-Lyzer MINI dialysis device (Thermo-Fisher Scientific). The amount of glycated BSA was 18 mg/ml after removing the impurities. The glycated albumin was used in assays at a final concentration of 200 μg/ml.
## 2.5. Measurement of reactive oxygen species
A quantitative assay was performed to measure the generation of ROS in different conditions. ROS generated in response to glucose treatment at different time points was measured using Luminol/HRP chemiluminescence assay for ROS detection. Briefly, PMNs were seeded in 96-well black plate (Corning 3,991 polystyrene flat bottom) in triplicates and 50 μM Luminol (Sigma- Aldrich A4685) and 1.2 U/ml horseradish peroxidase (Sigma- Aldrich P8250) were added to each well. As a positive control, PMNs were stimulated with 600 nM PMA (Sigma P8139). Plates were gently tapped to ensure mixing of all reagents. Luminescence was immediately measured by taking multiple readouts for up to 5 min in a luminometer (Perkin Elmer).
## 2.6. Induction of NETosis in the presence and absence of glucose
To measure the effect of glucose on induction of NETs, purified PMNs were treated with different concentrations of glucose that mimics normoglycemic (5 mM = HbA1C $4.8\%$) and hyperglycemic conditions (15 mM = HBA1C $11\%$). Briefly, washed PMNs were re-suspended in buffer containing RPMI (Sigma 1640), 10 mM HEPES (H0887 Sigma-Aldrich) and $10\%$ autologous heat inactivated serum and seeded into pre-made chambered slides coated with poly-l-lysine ($0.1\%$ (w/v) in H2O P8920 Sigma-Aldrich) at a concentration of 1 × 105. PMNs were allowed to adhere to the wells for 30 min at 37°C in the presence of $5\%$ CO2. Adherent PMNs were either incubated with (5 mM and 15 mM) glucose (158,968 Sigma-Aldrich) or with 600 nM PMA as a positive control for 4 h at 37°C in the presence of $5\%$ CO2. To determine the impact of in vivo hyperglycemia on NETosis, PMNs isolated from those with T2D were incubated in the absence of glucose. Expression of NETs, was determined by staining extracellular DNA with propidium iodide (Invitrogen™ P3566). Cells were washed and slides were observed under confocal microscope (Nikon C1). Percentage of NETosis was determined by counting a total of sixteen fields per slide and averaging the counted fields.
## 2.7. Nets in hyperglycemia in the presence and absence of ROS inhibitor
The effect of high glucose mediated ROS on NETs generation was confirmed by using catalase (150 units) (Sigma-Aldrich), a specific inhibitor of hydrogen peroxide (H202). After isolation, 1 × 105 cells/ml RPMI were seeded into 24 well plate on poly-l-lysine coated coverslips and stimulated with different glucose concentrations (5 mM, 15 mM) in the presence and absence of catalase (150 units). Stimulation of each concentration lasted for 4 h. Finally, the structure and location of NETs were observed under fluorescence microscopy. DNA quantitation was done using the standard protocol mentioned below.
## 2.8. Quantitation of NETs
Extracellular DNA was collected by scraping from the coverslip and solubilized using 20 units of DNAseI (Bio Basic DD0649), for 5 min at 37°C. Reaction was stopped by adding 0.05 M EDTA, extracellular DNA was quantitated using nano drop (Thermo scientific).
## 2.9. Measurement of RAGE, LCIII B, and neutrophil elastase using immunoblotting
Immunoblotting was used to measure expression of receptor for AGEs (RAGE), and intracellular markers of autophagy (LCIII B) and NETosis (neutrophil elastase). To determine the impact of hyperglycemia on NETosis and autophagy, PMNs isolated from those with poorly controlled diabetes were left stimulated. PMNs isolated from age and sex matched healthy volunteers were incubated in the presence or absence of 5 and 15 mM glucose for 30 and 120 min. On completion of incubation, samples were processed for western blotting according to the standard protocol. Briefly, glucose treated and untreated neutrophils were lysed in 2X laemmli sample buffer, sonicated (to shear DNA) at $50\%$ amplitude for 15 s on ice, boiled at 95°C for 10 min, and loaded on to $12\%$ (wt/vol) polyacrylamide gels for separation of proteins. Once the 6 kDa band of protein ladder (See Blue Plus-Invitrogen) reached the dye front at the end of the gel, the gels were removed from the cast and prepared for transfer to nitrocellulose membranes. Nitrocellulose membranes containing neutrophil proteins were blocked for 1 h in PBS containing $5\%$ non-fat milk and $0.1\%$ (wt/vol) Tween, to prevent non-specific binding of antibodies. After blocking, blots were washed and incubated with anti–neutrophil elastase (Santa Cruz) polyclonal antibody in 1:1000 dilution or anti–RAGE (Sigma-Aldrich) in 1:1000 or anti-LCIIIB (Abcam) or anti-GAPDH (santa-cruz) in 1:1000 dilution in PBST (Phosphate Buffered Saline pH7.4 + $0.1\%$ Tween) containing $5\%$ non-fat milk. Bound antibody was detected with enhanced chemiluminescence using horseradish peroxidase–conjugated anti–mouse-Ig secondary antibodies (southern biotech) in a dilution of 1:1000.
## 2.10. PMN activation and fractionation for measurement of activation of NADPH oxidase
PMNs isolated from patients with diabetes were not stimulated and processed directly. However, PMNs purified from non-diabetes individuals were stimulated with glucose (5 mM and 15 mM) with continuous shaking. After different time intervals, cells were pelleted and re-suspended in cell lysis buffer (10 mM PIPES [piperazine-N,N = -bis (2-ethanesulfonic acid)], pH 7.3, 100 mM KCl, 3 mM NaCl, 1.25 mM EGTA, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride [PMSF], 20 g/ml leupeptin, 20 g/ml pepstatin). Samples were sonicated at $50\%$ amplitude for 15 s on ice, and unbroken cells and nuclei were pelleted by centrifugation at 800 g for 10 min at 4°C. The supernatant was centrifuged at 50,000 ×g for 12 min at 4°C to pellet membranes. The pellet was re-suspended in solubilization buffer (20 mM Tris, pH 7.5, $1\%$ SDS, 1 mM PMSF, 20 g mL leupeptin, and 20 g mL pepstatin) in a volume equal to the initial total volume. Cytosol and membranes (1 × 106 cell equivalents) were incubated for 60 min on ice with solubilization buffer and then centrifuged at 15,000 g for 5 min to remove insoluble material. Equal volumes of each fraction, were mixed with 1X SDS sample buffer, and separated on $12\%$ SDS-PAGE followed by transfer to polyvinylidene difluoride (PVDF) (Millipore) membrane. After transfer, membranes were blocked with $5\%$ non-fat milk and $0.1\%$ (wt/vol) tween, and probed with rabbit anti-p40phox (Santa Cruz), or mouse anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Santa Cruz) at room temperature with continuous shaking. Blots were washed after 1 h with PBST and incubated with horseradish peroxidase labeled secondary goat anti- rabbit (Thermo Scientific) or anti-mouse IgG (southern biotech). Blots were developed with an ECL detection kit (GE Healthcare). Images were acquired on a Gel Doc System using Image Lab software (Bio-Rad).
## 2.11. NET detection by immunofluorescence
Poly-l-lysine coated coverslips were placed in a 24-well plate. PMNs were seeded at a concentration of 1 × 105 cells/ml of RPMI on coverslips, stimulated, and exposed to NET-inducing stimuli. After formation, NETs were fixed with $4\%$ PFA and incubated for 10 min at room temperature. Samples were washed twice with 300 μl sterile PBS for 10 min each time at room temperature. After washing, samples were permeabilized with $0.05\%$ Triton X-100, and blocked using 300 μl blocking solution ($2\%$ BSA) for 30 min at room temperature. Primary antibody (anti-human neutrophil elastase) was added into samples and incubated overnight at 4°C in dark. Coverslips were washed thrice with sterile PBS for 5 min at room temperature. Fluorochrome-labelled secondary antibody diluted in blocking solution ($2\%$ BSA) was added to coverslips for 1 h at room temperature in dark. Coverslips were washed three times in PBS for 5 min at room temperature. Antibody solutions was aspirated and after washing, a 10,000-fold diluted DAPI was added to coverslips for 2 min at room temperature. Coverslips were washed twice in PBS for 5 min each time at room temperature to remove unbound secondary antibodies and DNA stain. Coverslips were dried on a paper towel, and carefully placed on clean, degreased microscope slide with 10 μl gold anti-fade mounting medium. The edges of the coverslip were sealed using sealing liquid and after drying observed under confocal microscope (Nikon C1).
## 2.12. LDH cytotoxicity assay
LDH Cytotoxicity assay was performed using Pierce LDH cytotoxicity assay kit (Thermo scientific 88,954) according to manufacturer’s instructions. Briefly, the optimal number of cells (1 × 105 cells/ml) were seeded in 100 μl of RPMI medium in triplicate wells in a 96-well tissue culture plate. A complete medium control without cells was used to determine LDH background activity present in sera used for media supplementation. A serum-free media control was included to determine the amount of LDH activity in sera. Additional cells were plated in triplicate wells for spontaneous LDH activity controls (water) and maximum LDH activity controls (10X Lysis Buffer). The plate was incubated in an incubator at 37°C, $5\%$ CO2 for 30 min. To the set of triplicate wells serving as the maximum LDH activity controls, 10 μl of lysis buffer (10X) was added, and mixed by gentle tapping. Different concentrations of glucose (5 mM and 15 mM), and inhibitors AZD6244 (10 μM), GDC0941 (2 μM), GDC0068 (2 μM), were added to cells. The plate was again incubated in an incubator at 37°C, $5\%$ CO2 for 45 min. Each sample medium (50 μl) (e.g., complete medium, serum-free medium, Spontaneous LDH Activity Controls, glucose-treated, inhibitors-treated and Maximum LDH Activity Controls) was added to a 96-well flat-bottom plate in triplicate wells. Reaction mixture (50 μl) was transferred to each sample well and mixed using a multichannel pipette. The plate was incubated at room temperature for 30 min protected from light. Stop solution (50 μl) was added to each sample well and mixed by gentle tapping. The absorbance was measured at 490 nm and 680 nm. To determine LDH activity, the 680 nm absorbance value (background) was subtracted from the 490 nm absorbance before calculation of % cytotoxicity [(LDH at 490 nm) − (LDH at 680 nm)]. The % cytotoxicity was calculated as follows: % Cytotoxicity = Glucose/inhibitors-treated LDH activity − Spontaneous LDH activity/Maximum LDH activity − Spontaneous LDH activity × 100.
## 2.13. Phagocytosis
In order to measure phagocytosis, PMNs were isolated using polymorphprep as described previously. Cell count was adjusted to 1 × 105 cells/ml of RPMI. Streptococcus pneumoniae (D39 strain) were grown to an OD600nm of 0.4. To measure phagocytosis in normoglycemic or hyperglycemic conditions, PMNs were incubated with S. pneumonia (multiplicity of infection 1:50) in the presence or absence of glucose (5 mM and 15 mM) for varying lengths of time (30 min, 60 min and 120 min). To allow for phagocytosis to occur, mixture containing pneumococci and PMNs were incubated at 37°C in a shaking incubator and samples were removed at pre-determined time points. After incubation, samples were centrifuged at 3000 rpm for 10 min at room temperature. Supernatant was discarded and pellets were washed twice with 1 X PBS. Ampicillin (10 μg/ml) was added to eliminate all the extracellular bacteria. PMNs were washed again to remove ampicillin with 1 X PBS, and lysed with $0.05\%$ triton X-100 to release all the intracellular bacteria. The lysate was plated immediately on blood agar plates and left at 37°C incubator overnight in the candle jar. Next day the colonies were counted that gave an estimate of the phagocytosed bacteria.
## 2.14. Analysis of phagocytosis through FACS
The capsular type 2 strain of S. pneumoniae, D39, was grown to an OD600nm of 0.4, centrifuged at 3,000 rpm for 10 min and labelled for 1 h at 4°C in PBS containing 250 μg/ml FITC. After labelling, bacteria were washed thoroughly to remove excessive dye and stored at −20°C. Isolated PMNs (2 × 106 cells/ml) were incubated in different glucose conditions (5 mM and 15 mM) for 1 h at 37°C shaking incubator. Phagocytosis was performed in RPMI supplemented with $10\%$ FBS, 10 mM HEPES and different concentrations of glucose. In a 24 well plate 100 μl of FITC labeled bacteria (4 × 108 CFU/ml) were mixed with 1 ml of PMNs (2 × 106 cells/mL) to reach a 20:1 MOI. The reaction was incubated for other 30 min at 37°C on a plate thermoshaker (170 rpm). Samples were then fixed for 1 h with 1 ml cold formaldehyde at the final concentration of $2\%$. Samples were acquired immediately after 1 h of fixation at 4°C and run on FACS Verse flow cytometer. PMNs without bacteria were gated to eliminate cell debris and measure their auto florescence on FITC-H to differentiate FITC+ PMNs in other samples. A total of 10,000 events were collected for each sample gated on neutrophils.
## 2.15. Statistical analysis
All experiments in this study were repeated in triplicates in at least three independent experiments. In case of patient samples, experiments were repeated on three different days with different diabetic samples. Statistical analysis was performed using GraphPad Prism 7 software and SPSS 22.0 (IBM, United States). Data are presented in the form of graphs and expressed as mean, ± standard error of the mean (SEM). Unpaired t-test, was performed to make comparison between two groups at one time point assuming unequal variances. One-way analysis of variance (ANOVA) was used to calculate differences between more than two groups. Two way ANOVA used to compare more than two groups with more than one time point. The criterion for statistical significance was taken as $p \leq 0.05$ (2 sided). Probability values of $p \leq 0.05$ and $p \leq 0.001$ were statistically significant.
## 3.1. Demographic analysis of diabetes participants
The demographic and clinical characteristics of study participants are shown in Table 1. Table 2 categorizes those with diabetes on the basis of their most recent HbA1c values. A total of 60 cases were investigated during the study period, of which 30 individuals had diagnosed T2D whereas 30 were healthy controls. Most of the healthy controls were individuals visiting the outpatient clinic for complaints other than diabetes or were attendants of those with diabetes. The median age was 47 years for patients with T2D and 42.7 years for healthy controls. The male to female ratio was $43\%$ males and $56\%$ females among those with T2D, while in the healthy controls, $56\%$ were males and $43\%$ females. Among those with T2D, 23 out of 30 participants ($76\%$) had a family history of diabetes, whereas nine participants ($30\%$) in the non-diabetes group reported a family history of diabetes. Among those with T2D, 18 patients ($60\%$) also had a history of hypertension; however, four ($13\%$) non-diabetes participants reported a history of hypertension. A significant majority of those with diabetes were poorly controlled as suggested by their HbA1c values, where the lowest HbA1C value was 6.4 mM and the highest 14.1 mM with a mean HbA1C 10.15 ± 1.90. Similarly, individuals with T2D demonstrated high levels of random blood sugar (BSR), which ranged between 70 and 551 mg/dl, with a mean of 250 mg/dl. In comparison, BSR among the healthy controls was less than 110 mg/dl. None of these participants were diagnosed with tuberculosis or human immunodeficiency virus (HIV).
## 3.2. PMNs isolated from individuals with T2D undergo spontaneous NETosis
Hyperglycemia has been shown to aggravate NETosis, indicating the significance of this event in the underlying aetiology of diabetes [40]. We therefore wanted to measure NETosis in normoglycemic (5 mM = $4.8\%$ HbA1c) and hyperglycemic (15 mM = $11\%$ HbA1c) conditions. Incubation of neutrophils from healthy donors with 15 mM glucose for 4 h resulted in significant NETosis (Figure 1A). To confirm if extracellular structures represent NETs, immunofluorescence was done using antibodies to neutrophil elastase. Labeling for neutrophil elastase, when overlayed with DNA labeling (performed by DAPI), further confirmed that the extracellular matrix associated with neutrophils was indeed NETs (Figure 1B). Quantitation of DNA using nanodrop indicated that hyperglycemic conditions were associated with increased levels of extracellular DNA (58.64 ng/μl) as compared to normoglycemic conditions (10.52 ng/μl) $p \leq 0.001$ (Figure 1C). To further confirm if released DNA is the result of NETosis, we performed western blotting using antibodies to neutrophil elastase as a marker for NETs formation. An elevated level of protein was observed in neutrophils incubated with hyperglycemic conditions (15 mM glucose) as compared to normal glycemic conditions (5 mM glucose) or PMA. Densitometric analysis was performed to determine the ratio of neutrophil elastase/GAPDH from three independent experiments. Bar graph representing significantly elevated ratio of neutrophil elastase/GAPDH in the presence of 15 mM as compared to healthy PMNs (Figure 1D). To differentiate between cell lysis and NETosis, we performed lactate dehydrogenase assay. The presence of measurable quantities of extracellular lactate dehydrogenase (LDH), indicates the loss of viability of the cell and subsequent cell lysis. Low levels of LDH were identified associated with NETs suggesting the absence of cell lysis. LDH cytotoxicity assay has shown that neutrophils are not undergoing lysis upon incubation with 15 mM glucose which supports our hypothesis that the neutrophils are undergoing NETosis (Figure 1E).
**Figure 1:** *Spontaneous NETosis in PMNs isolated from individuals with T2D. (A) Fluorescent microscopy using propidium iodide DNA staining. The induction of NETosis was conducted on PMNs from healthy controls and treated with different concentrations of glucose (5 mM, 15 mM). PMA used as a positive control. The morphology of NETs observed after 4 h. Magnification: 20× (B)
in vitro NETosis by immunofloresence for neutrophil elastase (red) and DAPI (Blue). (C) Quantification of DNA released during NETosis from neutrophils using nanodrop. Data were analyzed using one way ANOVA with multiple comparisons test for group wise comparison. p < 0.001 (D) Western blot showing levels of neutrophil elastase. PMNs isolated from healthy subjects were treated with different glucose concentration (5 mM and 15 mM) for 4 h and processed for western blotting. Lower panel showing GAPDH used as a loading control. Bar graph showing densitometry measurement of the ratio of neutrophil elastase/GAPDH (three representative blots from independent experiments). Columns are mean values; error bars are SEM (E) LDH cytotoxicity assay done to determine cellular cytotoxicity upon incubating healthy PMNs with 15 mM glucose. Graph shows hyperglycemic glucose concentration (15 mM glucose) does not cause cytotoxicity. (F) Diabetic PMNs exhibited increased spontaneous NETosis. Fluorescent microscopy using propidium iodide staining, indicative of increased spontaneous NETosis observed in PMNs isolated from representative patients with diabetes compared to healthy donors. Magnification: 20× (G)
in vitro NETosis by immunofluorescence for neutrophil elastase (red) and DAPI (blue). Immunofluorescence Image showing the presence of neutrophil elastase on DNA strands. (H) Graph showing DNA Quantitation data using nanodrop. Data represented is indicative of three independent experiments. Data analyzed using one way ANOVA with multiple comparisons test. Data are represented as mean ± SE. p < 0.001. (I) Western blot showing basal level expression of elastase protein in PMNs isolated from diabetic subjects with different HbA1c values. PMNs from healthy subject with HbA1c less than 5 is taken as a control. Bar graph representing densitometry analysis of the ratio of neutrophil elastase/GAPDH. (J) Relationship of random blood glucose and percent glycated hemoglobin A1C (HbA1c) with DNA released by diabetic PMNs. Both random blood glucose (r = 0.609: p < 0.001) and HbA1c (r = 0.603: p = 0.01) showed a significant positive relationship with released DNA. Correlations were calculated using the Pearson’s correlation test.*
In order to validate these findings ex vivo, PMNs were isolated from individuals with well-controlled (HbA1C 6–$8\%$) or poorly controlled (HbA1C 9–$14\%$) diabetes, where diabetes control was determined by HbA1c values. PMNs were incubated in the absence of a stimulus for 4 h, and NETosis was visualized as described in the material and methods section. PMNs isolated from individuals with poorly controlled T2D demonstrated NETosis in the absence of any stimuli. PMNs isolated from those with no diabetes showed very little NETosis. In order to obtain NETosis, comparable to those with poorly controlled diabetes, PMNs from those without diabetes had to be incubated with PMA (positive control; Figure 1F). Immunofluorescence showed neutrophil elastase associated with DNA strands indicating the extracellular DNA is the result of NETosis and not cell lysis in neutrophils from T2D subjects (Figure 1G). Quantitation of DNA using nanodrop indicated that hyperglycemic conditions were associated with increased levels of extracellular DNA as compared to normoglycemic conditions $p \leq 0.001$ (Figure 1H). NETosis was also confirmed by the presence of neutrophil elastase through western blotting *Densitometry analysis* showing significantly higher ratio of neutrophil elastase/GAPDH isolated from diabetic PMNs as compared to healthy PMNs (Figure 1I). Higher levels of neutrophil elastase were found associated with neutrophils from poorly controlled diabetes individuals, which correlated well with the DNA concentration and NETosis. The expression of GAPDH was used as a loading control. Next, we wanted to determine if there is a correlation between released DNA and blood glucose and HbA1c values. Results of Pearson Correlation demonstrated that there was a moderately positive ($r = 0.6$) association between blood glucose, HbA1c and release of DNA from diabetic neutrophils (Figure 1J).
## 3.3. Priming of neutrophils and subsequent NETosis in the presence of advanced glycation end products
Hyperglycemia has been shown to impact cellular functions in multiple ways [41]. One of the most common mechanisms is via glycation of various structural and functional proteins, thereby resulting in the generation of advanced glycation endproducts or AGE [42]. These glycated proteins bind to their cognate receptors called Receptors for Advance Glycation End Products or RAGE. The RAGE receptor is a scavenger receptor and is ubiquitously expressed on endothelial cells, including immune cells, such as macrophages and neutrophils [43]. Therefore, in addition to hyperglycemia, we wanted to determine the impact of downstream products of hyperglycemia, such as AGE, on the formation of NETs. As demonstrated previously, the AGE binds to RAGE and that presence of AGE increases the expression of RAGE on the surface of PMNs. We first wanted to determine if there is an increase in the expression of RAGE on the surface of PMNs isolated from those with diabetes. Additionally, we also wanted to determine if transient acute glycemia, in response to incubation of healthy PMNs in hyperglycemic conditions, would lead to the expression of RAGE. Our results indicated that incubation of PMNs with 15 mM glucose for 120 min led to increased surface expression of RAGE, whereas low levels of hyperglycemia or incubation of PMNs with high levels of glucose for a short period of time did not impact RAGE levels on the surface of PMNs. Densitometry analysis showing relative protein expression of RAGE/GAPDH (Figure 2A). Next, to determine if PMNs isolated from those with different levels of glucose control, would also show differences in RAGE levels, we performed similar measurements on PMNs isolated from those with diabetes showing varying levels of HbA1c. Interestingly we observed differences in levels of RAGE associated with PMNs, where highest levels of RAGE expression was observed in individuals with HbA1c ranging from 7 to 10. Control non-diabetes individuals had very little or non-detectable levels of surface RAGE confirmed through densitometry analysis of RAGE/GAPDH (Figure 2B).
**Figure 2:** *Priming of Neutrophils and subsequent NETosis in the presence of advance glycation end products. (A) Representative western blot analysis of RAGE expression in PMNs. Cells were treated with various concentrations of glucose (5 mM and 15 mM) for varying time points (30 min and 120 min). After given time points samples were prepared according to the standard protocol. A total of 10 μg protein loaded onto each well. Hyperglycemic glucose concentration (15 mM) leads to elevated expression of RAGE after 120 min of incubation. GAPDH used as a loading control. Bar graphs showing densitometric analysis of ratio of RAGE/GAPDH using two independent blots from two different experiments. Columns are mean values; error bars are SEM. Asterisks show level of significant difference from basal *p < 0.01 (B) Western blot showing RAGE protein expression in PMNs isolated from diabetic subjects with different HbA1C values. Healthy subjects with HbA1c value <5 are used as control. RAGE; receptor for advanced glycation end products. Bar graph representing densitometric measurements of RAGE/GAPDH in diabetic PMNs. Experiments were performed thrice, with data from a representative experiment shown (C) Representation of in-vitro NET release. Indicative of increased NETosis observed in PMNs isolated from representative healthy donors and incubated with AGE for 4 h. NETosis was decreased in the presence of Anti-RAGE antibody (D) Graph showing DNA Quantitation data using nanodrop. Data are represented as mean ± SE. ***p < 0.001. (E) Western blot showing levels of neutrophil elastase protein expression after incubating cells with AGE. Lower panel showing GAPDH as a loading control. Bar graph showing densitometry analysis of neutrophil elastase/GAPDH (two independent experiments). Columns are mean values; error bars are SEM.*
In order to determine if binding of AGE to RAGE leads to NETosis, we performed assays to detect NETS generation. We observed extensive NETosis of neutrophils in the presence of AGE as compared to absence of AGE (Figure 2C) which was further confirmed by DNA quantitation using nanodrop (Figure 2D) and release of neutrophil elastase in the presence of AGE (Figure 2E). Western blot showing that the release of neutrophil elastase by AGE was blocked by incubation with the anti-RAGE antibody confirming the specificity of AGE action. Densitometry analysis demonstrating the relative protein expression of RAGE/GAPDH in response to AGE in the presence and absence of anti-RAGE antibody.
## 3.4. Increased levels of ROS are associated with T2D
Reactive oxygen species serves as a second messenger for the generation of NETs [44]. We, therefore, tested the hypothesis that hyperglycemia might prime PMNs to produce elevated levels of ROS, which leads to spontaneous NETosis in the absence of second stimuli. The levels of ROS in PMNs isolated from healthy volunteers in normoglycemia and hyperglycemic conditions were measured. For that, PMNs isolated from the healthy subject were incubated with 5 mM (physiological concentration) and 15 mM (hyperglycemic concentration) glucose for 30 min. Phorbol myristate acetate (PMA), known to induce ROS via activation of NADPH, was used as a positive control. Levels of ROS were measured using Luminol/HRP assay according to the standard protocol. Resting PMNs isolated from healthy donors demonstrated low levels of ROS, which did not change significantly in the presence of normoglycemic conditions. However, up to 3 fold increase in ROS was observed in PMNs incubated with 15 mM of glucose, indicating that hyperglycemia increases levels of ROS, which was significantly higher compared to normoglycemic condition ($p \leq 0.001$) (Figure 3A).
**Figure 3:** *Increased levels of ROS were associated with T2D. (A) ROS generation was quantified in response to glucose in healthy peripheral blood neutrophils (PMNs) using luminol HRP assay. PMNs were incubated in 5 mM glucose (physiological concentration) and 15 mM glucose (high glucose concentration) and RLUs measured for upto 5 min in luminometer. Bar graph showing data from three independent experiments. Data analyzed using one way ANOVA with multiple comparisons. p < 0.001 (B) Total reactive oxygen species produced by PMNs of non-diabetic and diabetic subjects. PMNs isolated from non-diabetic individuals and diabetic patients subjected for ROS secretion at resting condition. The maximal production of ROS at any time point over a period of 5 min was recorded and compared using one way ANOVA. PMNs isolated from diabetic patients showed a higher and consistent secretion of ROS in resting condition. Data expressed as mean ± SEM and the statistical significance was determined at p ≤ 0.01. ROS: Reactive oxygen species; RLUs: Relative luminisence units. (C) Levels of ROS generated after incubating PMNs with AGE in the presence and absence of anti-RAGE antibody. Significantly elevated levels of ROS were recorded in the presence of ROS as compared to healthy control and AGE in the presence of anti-RAGE antibody. Data analyzed using one way ANOVA with multiple comparison test for comparing different groups. Statistical significance was determined at p < 0.001.*
To validate these findings further, we isolated PMNs from those with T2D with HbA1c values, ranging from 6 to $14\%$, and measured levels of ROS generated intrinsically. Samples collected from those with poorly controlled diabetes demonstrated significantly elevated levels of ROS as compared to healthy PMNs ($$p \leq 0.01$$; Figure 3B).
While hyperglycemia results in AGing of proteins and lipids, binding of AGE to RAGE than continues the cycle of ROS generation and subsequent oxidative stress [42]. We thus determined if the binding of AGE to RAGE is also impacting the levels of ROS in healthy PMNs. Incubation of healthy PMNs with AGE for 30 min demonstrated a significant increase in levels of ROS ($p \leq 0.001$; Figure 3C). To demonstrate that the observed increase in the levels of ROS were resulting from the interaction between AGE and RAGE, samples were incubated for another 30 min before measuring levels of ROS. A significant decrease ($60\%$) in the levels of ROS was observed in the presence of antibodies to RAGE, indicating that binding of AGE to RAGE is associated with an increase in levels of ROS ($p \leq 0.001$) (Figure 3C).
## 3.5. Elevated levels of ROS are associated with the activation of NADPH oxidase in hyperglycemic conditions
NADPH oxidase is a multicomponent enzyme system [45]. In resting cells, this enzyme system is not activated (not assembled), and the components are divided between the membrane and the cytosol. The neutrophil NADPH oxidase is comprised of plasma membrane-bound subunits (gp91phox and p22phox) comprising flavoCytochrome b 558 and cytosolic subunits (p47phox, p67phox, p40phox, and Rac2). The activation of PMNs by stimuli, such as fMLP or PMA, causes the phosphorylation and translocation of the cytosolic components to the plasma membrane, where they interact with flavoCytochrome b 558. NADPH oxidase activation can result from either an increase in the expression of one or several subunits or by translocation of cytosolic subunits to the plasma membrane [46]. We, therefore, wanted to test the hypothesis that ROS production as observed in the previous section (Figure 3A), is the result of the priming or partial activation of NADPH oxidase in response to hyperglycemic conditions. To confirm the activation of NADPH oxidase complex, we measured the expression and translocation of p40 (cytosolic subunit), which when active, translocates from the cytosol to the membrane. Our results indicated that in resting PMNs, obtained from healthy individuals, p40 was predominantly cytoplasmically located; however, in hyperglycemic conditions (15 mM), partial translocation of p40 subunit from cytosolic to membrane fraction was observed in 120 min post glucose exposure. CD11b, a membrane receptor, was used as a control for membrane fraction, and GAPDH was the control for the cytosolic fraction. To determine if hyperglycemic conditions in those with poorly controlled diabetes are also associated with translocation of p40 subunit and subsequent activation of NADPH in PMNs, we measured translocation of p40 subunit in PMNs isolated from individuals with T2D. Fractionation of PMNs isolated from those with T2D, demonstrated an increased membrane localization of p40 subunit compared to the PMNs from healthy individuals, indicating that the NADPH oxidase complex was primed in those with T2D (Figure 4A).
**Figure 4:** *Elevated levels of ROS are associated with activation of NADPH oxidase in hyperglycemic conditions. (A) Western blot showing translocation of NADPH oxidase p40phox cytoplasmic subunit to neutrophil membranes. Primary human neutrophils (PMNs) isolated from healthy subjects were incubated with 15 mM glucose for 120 min. PMNs isolated from diabetic individuals with different HbA1C (7 and 9) were not stimulated with glucose. PMNs were separated into membrane and cytosol fractions, and the amounts of p40phox in neutrophil cytosol and membrane fractions were analyzed by western blotting. The membrane-associated Cd11b protein served as a loading control for membrane fractions and GAPDH as a control for cytosolic fraction. The results presented are representative of three independent experiments. (B) Western blot showing activation of NADPH oxidase complex as indicated by translocation of p40phox subunit from cytosol to membrane fraction after incubating healthy PMNs with AGE (200 μg/ml) in the presence and absence of anti-RAGE antibody. Membrane associated Cd11b protein is used as a control for membrane fraction. GAPDH used as a loading control for cytosolic fraction (C) Florescence microscope images showing NETS formation in the presence and absence of NADPH oxidase inhibitor, Diphenylene ionodonium (DPI) (10 μM) confirming the role of NADPH oxidase generated ROS in the process of NETs generation. (D) DNA quantitation showing significantly decreased levels of ROS in the presence of DPI. One way ANOVA used to compare groups. ***p < 0.001. (E) Western blot showing the decreased expression of neutrophil elastase protein in the presence of DPI as compared to 15 mM glucose. Lower panel showing GAPDH used as a loading control. Bar graph representing densitometric measurements of ratio of neutrophil elastase/GAPDH using two independent blots. Columns are mean values; error bars are SEM.*
We next determined if the binding of AGE-RAGE would activate NADPH-oxidase, a multiprotein enzyme complex responsible for the activation and release of ROS from PMNs. PMNs isolated from healthy donors were incubated with 200 μg/ml AGE for 120 min, and membrane localization of p40 subunit was measured. Our results demonstrated that in PMNs incubated with glycated BSA, p40 subunit translocated to the cytoplasm after 120 min incubation (Figure 4B), whereas in the presence of antibody to RAGE, we found p40 to be both membrane and cytoplasmically located.
Next, we wanted to determine if ROS generated in response to priming of PMNs by activation of NADPH results in induction of NETs generation. For that, we blocked the activation of the NADPH oxidase complex by using Diphenyleneiodonium (DPI), a potent inhibitor of flavoenzymes. Our results indicated that pretreatment of PMNs with DPI (10 μM) for 1 h blocked NETs formation, despite activation with 15 mM glucose (Figure 4C); this was further confirmed by DNA quantitation ($p \leq 0.001$; Figure 4D). Western blot also confirmed that after blocking NADPH oxidase complex with DPI, the level of neutrophil elastase protein complex is decreased as compared to 15 mM glucose. Densitometry measurements showing relative protein expression of neutrophil elastase/GAPDH in the presence and absence of DPI (Figure 4E).
## 3.6. Elevated levels of ROS and subsequent NETosis are associated with poorly controlled T2D
To explore the relationship between ROS production and NETs extrusion, PMNs were isolated from healthy subjects and NETs were generated using hydrogen peroxide (H202) (100 μM), the well-established ROS. PMNs were seeded into wells in the presence or absence of hydrogen peroxide. When DNA is extruded extracellularly, it is bound by propidium iodide and fluoresces when excited. After 4 h, the samples were fixed and mounted and observed in a fluorescence microscope to morphologically confirm the presence of NETs. As little as 100 μM H202 was sufficient to induce PMNs to release extracellular structures that fluoresced in the presence of propidium iodide (Figure 5A). When observed through immunofloresence, neutrophil elastase appeared to be associated with DNA strands, confirming the release of NETs (Figure 5B) and quantitated using nanodrop (Figure 5C). Neutrophil extracellular traps (NETs), extruded from neutrophils upon activation, has been shown to be dependent on reactive oxygen species (ROS) generation. However, NET release can also occurs through a rapid ROS-independent mechanism [47]. Thus, to examine the requirement of ROS in the generation of NETs and to confirm that these assays were compatible with detecting ROS-dependent release of NETs, western blotting was performed to check the levels of neutrophil elastase protein expression in the presence and absence of hydrogen peroxide (H202). An elevated level of neutrophil elastase expression was observed in the presence of ROS as depicted by densitometry analysis (Figure 5D). We further aimed to block ROS, particularly hydrogen peroxide using catalase, to confirm if the inhibition of pathways involved in ROS generation leads to reduction of NETs formation. For that, we treated PMNs isolated from a healthy volunteer with different concentrations of glucose (5 mM and 15 mM) in the presence and absence of catalase (150 units), a potent inhibitor of hydrogen peroxide (H202). NETs generation assay was performed to qualitatively visualize NETs using a florescence microscope. Our results demonstrated that the addition of antioxidant catalase significantly reduced the release of NETs even in the presence of hyperglycemic glucose concentration (Figure 5E). We further quantified DNA released during NETs generation using nanodrop. We observed a significant decrease ($p \leq 0.001$) in the levels of released DNA after the addition of antioxidant (12.75 ng/μl) as compared to 15 mM glucose (61.5 ng/μl; Figure 5F), suggesting a role of ROS in the induction of spontaneous NETosis in those with T2D. Western blot analysis demonstrated lower expression of neutrophil elastase in the presence of an antioxidant, further demonstrated by densitometry analysis showing relative protein expression of neutrophil elastase/GAPDH (Figure 5G).
**Figure 5:** *An association was observed between elevated levels of ROS and NETosis in poorly controlled T2D. (A) Representative images of NETs released from PMNs stimulated with H202 (100 μM), and unstimulated control. Results are representative of three independent experiments. Magnification 20×. (B) Immunofluorescence image showing association of neutrophil elastase (red) with DNA stained with DAPI (blue). (C) DNA quantitation values using nanodrop. Bar graph representing data from three independent experiments. Data analyzed using unpaired T-test. Data expressed as mean ± SEM and the statistical significance was determined at p ≤ 0.001. (D) Western blot showing the levels of neutrophil elastase upon stimulating healthy PMNs with hydrogen peroxide (100 uM) for 4 h. Bar graph showing densitometric analysis of neutrophil elastase/GAPDH from two independent experiments. Column represents mean values while error bar represents SEM (E) Antoixidant block NETs production in vitro. The induction of NETosis was conducted on PMNs from healthy controls and treated with PMA, 5 mM and hyperglycemic concentrations of glucose (15 mM) in the presence and absence of antioxidant (catalase, 150 units) for 4 h. (F) DNA quantitation using nanodrop. Data analyzed using one way ANOVA with multiple comparisons. Data expressed as mean ± SEM and the statistical significance was determined at p value ≤0.001. (G) Western blot showing the levels of neutrophil elastase protein expression after incubating neutrophils with 15 mM glucose in the presence and absence of catalase. PMA used as a positive control. Lower panel showing GAPDH as a loading control. Bar graph showing densitometric analysis of ratio of neutrophil elastase/GAPDH from three independent experiments. Column represents mean: error bar represents SEM. *p < 0.01.*
## 3.7. Elevated levels of ROS impair the phagocytic ability of PMNs in T2D
While NETosis, helps to contain the infection, phagocytosis remains an important killing mechanism employed by PMNs [48]. Reactive oxygen species have been shown to be important for both NETosis and phagocytic killing [49]. However, it remains to be determined if elevated levels of ROS oscillate the bactericidal activity towards NETosis or phagocytosis. To determine how hyperglycemia impacts the phagocytic ability of PMNs to engulf and kill S. pneumoniae, healthy PMNs were incubated with hyperglycemic conditions followed by incubation with S. pneumoniae. Aliquots were removed at varying time intervals to measure phagocytosis, as explained in the material and methods section. Our results (Figure 6A) showed that compared to hyperglycemic conditions where the phagocytic ability of PMNs decreased overtime, optimum phagocytic activity of peripheral PMNs was observed in the absence of glucose and in the presence of normoglycemic conditions. To determine whether the phagocytosed bacteria were viable, the number of viable bacteria were determined by colony forming unit (CFU) analysis. Compared to the no glucose controls, the relative CFU measured at 15 mM glucose were significantly lower when incubated for 2 h (Figure 6A) as compared to no glucose or normoglycemic conditions (5 mM). In order to validate these findings, we further quantified levels of phagocytosis through FACS. Our results demonstrated that the ability of healthy PMNs to phagocytose fluorescently labeled D39 strain of S.pneumoniae after stimulating with high glucose (15 mM) was markedly decreased as compared to no glucose and 5 mM glucose controls. Thus, confirming that the observed decrease in the number of colonies in Figure 6A represents impaired phagocytosis of bacteria in the presence of high glucose (Figure 6B). Diabetic PMNs were also assessed for their ability to phagocytose S. pneumoniae ex vivo. PMNs isolated from those with diabetes showed decreased phagocytosis as suggested by less intracellular bacteria compared to healthy PMNs (Figure 6C). At each time point, the ratio of extracellular to intracellular S. pneumoniae was very high in PMNs isolated from those with diabetes as compared to age and sex matched controls ($p \leq 0.001$), suggesting a markedly reduced phagocytic capacity of PMNs isolated from those with diabetes.
**Figure 6:** *Hyperglycemia mediated ROS impairs phagocytic ability of PMNs in T2D. (A) Analysis of neutrophil phagocytosis in the presence of different glucose concentrations. Percentage of PMNs phagocytically active as determined by the percentage of PMNs which have phagocytosed S. pneumonia (D39 strain). Log CFUs of ingested S. pneumoniae as an indication of neutrophil phagocytic activity at different time points and at different glucose concentration. Bar graphs representing data from three independent experiments. Data analyzed using two way ANOVA with multiple comparison test. Data expressed as mean ± SEM and the statistical significance was determined at p value ≤0.001 (B) Analysis of phagocytosis through FACS. Uptake of FITC-labelled D39 strain of S. pneumonaie by healthy PMNs. PMNs isolated from healthy controls were incubated with 5 mM and 15 mM glucose for 1 h and incubated with FITC-labelled D39 for 30 min at 37°C. Phagocytic activities were analyzed using flow cytometry. Data are presented as histograms. (C) Phagocytosis of S. pneumonia (D39 strain) by human PMNs (diabetics vs. healthy controls) for 2 h. PMNs were isolated from diabetics and healthy individuals and incubated with S. pneumoniae for 2 h. At each time point diabetic PMNs showed significantly reduced phagocytosis activity as compared to healthy controls. Statistical significance was determined by comparing samples between healthy and diabetics at different time points (two-way ANOVA, Bonferroni) with significance p < 0.001. (D) Phagocytosis of S. pneumoniae by human PMNs in the presence and absence of hydrogen peroxide (H202) (100 μM) for 2 h. In the presence of hydrogen peroxide (H202), phagocytosis was decreased significantly. Data expressed as mean ± SEM and the statistical significance was determined at p value ≤0.01. (E) Phagocytosis of S. pneumoniae by human PMNs in the presence and absence of antioxidant catalase (150 units) and 15 mM glucose for 2 h. Graph represent results of three independent experiments. Data analyzed using two way ANOVA with multiple comparison test.*
While ROS is responsible for the killing of bacteria during the process of phagocytosis, excessive production of ROS, as observed in diabetes, has been shown to impair bacterial phagocytosis and intracellular killing [50]. To determine if ROS generated in hyperglycemic conditions is associated with impairment in phagocytosis, we used hydrogen peroxide H202 (100 μM). Incubation of PMNs with capsule type 2 strain of *Streptococcus pneumoniae* D39, in the presence of H202, showed a decrease in the phagocytosis of S. pneumoniae. However, in the absence of H202, the phagocytosis increased and peaked in 2 h, further confirming our hypothesis that hyperglycemia-mediated ROS impairs phagocytosis and intracellular killing of bacteria (Figure 6D). Furthermore, we measured the phagocytosis of S. pneumoniae in PMNs treated with catalase (150 units). Our results indicated that contrary to what we observed in NETosis, phagocytosis of S. pneumoniae, was restored in the presence of catalase. We started to observe an increase in the number of intracellular colonies at all-time points in catalase treated PMNs, suggesting an increase in the number of phagocytosed bacteria (Figure 6E).
## 3.8. ROS generated through hyperglycemia and AGE/RAGE signaling induces autophagy that leads to NETosis
Reactive oxygen species have been shown to modulate several cellular processes. One such process is autophagy, which is a mechanism of self-sustenance [24, 25]. In the previous section, we observed inhibition of phagocytosis by blocking ROS. Since ROS plays an important role in NETosis, phagocytosis and autophagy, we next wanted to determine how blocking ROS would impact autophagy. As autophagy is an important cellular pathway that facilitates both NETosis and phagocytosis, it was therefore imperative to address the following scenarios (i) how hyperglycemia modulates autophagy (ii) does modulation of autophagy impact the bactericidal activity of neutrophils (iii) which bactericidal process (phagocytosis and NETosis) is most impacted. To investigate the effects of ROS on autophagy, and its downstream impact on phagocytosis and NETosis, PMNs from healthy individuals were subjected to glucose induction using 5 mM and 15 mM of glucose. PMNs were exposed to each concentration of glucose for 30 min and 120 min. At the end of each incubation, we measured LCIIIB using western blotting. An increase in levels of LCIIIB was observed, indicating induction of autophagy at 15 mM at both 30 min and 120 min time points as shown by densitometry measurements (Figure 7A). In comparison to induced PMNs, we observed basal levels of LCIIIB in PMNs incubated in the absence of glucose or in the presence of 5 mM (normoglycemic conditions) of glucose. Next, we compared levels of LCIIIB in individuals with diabetes and age and sex matched controls. As shown in Figure 7B, we observed significantly elevated levels of LCIIIB in individuals with diabetes which correlated with their blood glucose control as indicated by HbA1c values. Basal levels of LCIIIB were observed in age and sex matched controls and in individuals with well controlled diabetes. Our previous results demonstrated that hyperglycemia can either directly (Figure 3A) or indirectly (via AGE-RAGE association) (Figure 3C) activate ROS. Therefore we next wanted to investigate if hyperglycemia mediated up-regulation of RAGE and binding of AGE to RAGE would also induce autophagy that results in subsequent NETs formation (as shown in Figure 2C). Incubation of PMNs with AGE in the presence and absence of antibody to RAGE for 30 and 120 showed an increased expression of LCIIIB after addition of AGE, however in the presence of anti-RAGE LCIIIB expression was markedly decreased. Bar graph showing densitometry measurements of LCIIIB/GAPDH expression (Figure 7C). To determine if this binding of AGE-RAGE is resulting in ROS generation, particularly hydrogen peroxide (H202), which further contributes to up-regulation of autophagy, we used catalase enzyme (150 units) to specifically block hydrogen peroxide (H202), and measured the expression of LCIIIB. Our results showed that after using catalase enzyme, the expression of LCIIIB is decreased confirming the role of hydrogen peroxide (H202) in the induction of autophagy (Figure 7D). Next, we wanted to correlate the association of AGE with RAGE interaction, subsequent ROS production and activation of autophagy on induction of NETosis. For that, we scavenged hydrogen peroxide (H202), using catalase (150 units) and checked for the expression of NETs using NETs generation assay and western blotting. Our results indicated that AGE resulted in up-regulation of autophagy which led to the formation of NETs. However, the addition of catalase (Figures 7E,F) lowered autophagy and NETosis, suggesting the role of AGE-mediated ROS, particularly hydrogen peroxide, in the induction of autophagy that led to NETosis. ( Figure 7G).
**Figure 7:** *ROS generated through hyperglycemia and AGE/RAGE signaling induces Autophagy that leads to NETosis: (A) Western blot analysis of LCIIIB I and LCIIIB II proteins levels in human PMNs isolated from healthy individuals. Cells were incubated with different glucose concentrations (5 mM and 15 mM) for varying time points (30 min and 120 min). Where 15 mM glucose resulted in enhanced expression of LCIIIB protein after the given time points. Bar graphs representing densitometry measurements of ratio of LCIIIB/GAPDH (two independent experiments). Column represents mean, error bar represents SEM (B) Western blot showing basal level protein expression of LCIIIB I and LCIIIB II proteins in PMNs isolated from diabetic individuals with different HbA1C values mentioned. PMNs from healthy individual with HBA1c values <5 were used as healthy control. Lower panel showing GAPDH as a loading control. Western blots are representative of three independent experiments (n = 3). Densitometry analysis represented using bar graph. (C) Representative western blot analysis of LCIIIB I and LCIIIB II levels in human PMNs pre-treated as indicated with AGE (200 μg/ ml) for 120 min and anti-RAGE antibody as compared to untreated cells. Incubation with AGE lead to enhanced expression of LCIIIB protein. GAPDH used as a loading control. Densitometry analysis of the autophagy signal (LCIIIB) in healthy PMNs treated with AGE in the presence and absence of anti-RAGE antibody. Columns are mean values of LCIIIB/GAPDH; error bars are SEM. Asterisks indicates significant reduction in autophagy in the presence of anti-RAGE antibody. (D) Representative western blot image of LCIIIB I and LCIIIB II protein using human PMNs treated with AGE in the presence and absence of antioxidant catalase (150 units). Cells left untreated were used as a control. Lower panel showing GAPDH used as a loading control. Bar graph showing densitometry analysis of ratio of LCIIIB/GAPDH from two independent experiments (E) Florescence microscopic images representing in-vitro NET release. Increased NETosis observed in PMNs isolated from healthy subjects incubated with AGE for 4 h. NETosis was decreased in the presence of antioxidant catalase (150 units). (F) DNA Quantitation data using nanodrop. Bar graphs represent mean DNA release from three independent experiments. Error bars represent SEM. Data analyzed using t-test. Asterisks represent significant difference of DNA release. ***p < 0.001. (G) Western blot showing levels of neutrophil elastase protein expression after incubating cells with AGE in the presence and absence of catalase. Lower panel showing GAPDH as a loading control. Bar graph representing ratio of neutrophil elastase/GAPDH from representative experiment.*
## 3.9. Inhibition of autophagy prevents hyperglycemia-induced NETosis and decreases phagocytosis
Previous studies have shown that both autophagy and ROS are required for the formation of NETs [51]. However, a three-way relationship between ROS, autophagy and NETosis remains to be determined. Additionally, it is also not known how conditions such as diabetes that impact both ROS and autophagy would modulate NETosis. We, therefore, investigated the hypothesis that hyperglycemia-mediated ROS activates autophagy which primes neutrophils for NETosis. To validate this hypothesis, we measured NETosis in PMNs in the presence and absence of inhibitors of autophagy. In individuals without diabetes, PMNs were stimulated with different concentrations of glucose (5 mM and 15 mM). The presence of PI-3kinase inhibitor (GDC0941; 2 μM) inhibited autophagy and blocked NETs formation, as indicated by qualitative analysis of NETs induction (Figure 8A) and quantitated by DNA quantitation analysis (Figure 8B) and measurement of levels of neutrophils elastase by western blotting and densitometry analysis (Figure 8C). Blocking autophagy also impacts phagocytosis as we observed a significant decrease in the log CFUs after incubating PMNs with PI-3Kinase inhibitor GDC0941, both in the presence and absence of 15 mM glucose. ( Figure 8D). The PI3K–AKT–mTOR axis has widely been reported to connect autophagy and NETs formation, where inhibition of mTOR leads to enhances NET generation through activation of autophagy [52]. Thus, to further validate the role of autophagy in induction of NETosis, we set out to block MAPK/Erk $\frac{1}{2}$ and PI3K-I/AKT signaling pathways upstream of mTOR; a negative regulator of autophagy. We aimed to suppress mTOR expression using two inhibitors AZD6244 (inhibits upstream kinase of ERK), and GDC0068 a highly selective pan-Akt inhibitor, targeting Akt$\frac{1}{2}$/3. PMNs isolated from healthy individuals were treated with GDC0068 (2 μM), and AZD6244 (10 μM) in the presence and absence of 5 mM and 15 Mm glucose. Our results have demonstrated that the pan-Akt inhibitor GDC0068 and ERK inhibitor AZD6244 achieves significant upregulation in autophagy that results in increased NETosis in the presence of normoglycemic glucose concentration (i.e., 5 mM glucose). Thus confirming that activation of autophagy is required for induction of NETosis (Figure 8E). Lesser degree of NETosis observed with hyperglycemic glucose concentration (i.e., 15 mM glucose, may indicate that hyperglycemia triggers NETosis independently of mTOR). This however, needs further investigation. LDH cytotoxicity assay was performed to confirm that the concentrations of inhibitors AZD6244 (10 μM), GDC0941 (2 μM), GDC0068 (2 μM), used during the assays are not toxic to healthy PMNs. The percent cytotoxicity displayed by the three inhibitors is significantly lower as compared to the positive control. ( Figure 8F).
**Figure 8:** *Inhibition of autophagy prevents hyperglycemia induced NETosis and decreases phagocytosis. (A) Fluorescent microscopy to visualize NETs in untreated (basal) PMNs as negative control, treated with 5 mM and 15 mM glucose in the presence and absence of GDC0941 (2 μM). (B) Bar graph showing DNA quantitation data. DNA (ng/μl) released after treating healthy PMNs with 5 mM and 15 mM glucose in the presence and absence of GDC0941, an inhibitor of autophagy. Data analyzed using one way ANOVA with multiple comparison test. Columns show mean; error bars are SEM. Statistical significance determined at p < 0.01. (C) Western blot analysis showing expression of neutrophil elastase protein in human PMNs left unstimulated, treated with normal glucose concentration (5 mM) and hyperglycemic glucose concentration (15 mM) in the presence and absence of an inhibitor of autophagy GDC0941. Lower panel showing GAPDH as a loading control. Bar graph showing densitometry analysis of neutrophil elastase/GAPDH of representative experiment. (D) Graph showing phagocytosis of D39 strain of S. pneumoniae by human PMNs in the presence and absence of autophagy inhibitor GDC0941 (2 μM). Data representative of three independent experiments analyzed using two way ANOVA. (E) Fluorescent microscopic images representing NETosis using propidium iodide DNA staining. PMNs isolated from healthy controls were treated with ERK inhibitor AZD6244 (10 μM), and pan-Akt inhibitor GDC0068 (2 μM) in the presence and absence of glucose (5 mM and 15 mM) to induce NETosis. PMA used as a positive control. The morphology of NETs observed after 4 h. Magnification: 20×. (F) LDH Cytotoxicity assay to determine cellular cytotoxicity upon incubating healthy PMNs with different inhibitors of autophagy AZD6244 (10 μM), GDC0068 (2 μM) and GDC0941 (2 μM). Graph shows the used concentrations of inhibitors are not cytotoxic for healthy PMNs.*
## 4. Discussion
Hyperglycemia, an important downstream complication of T2D, has been implicated in the dysfunctioning of immune cells such as neutrophils (PMNs). Recent reports suggest that hyperglycemia leads to metabolic reprograming, characterized by excessive glycolysis and pentose phosphate pathway (PPP), and subsequent elevated levels of metabolites which feeds into neutrophil functions such as activation of NADPH-oxidase and production of ROS [53]. This metabolic reprogramming has also been implicated in trained immunity, where both neutrophils and macrophages demonstrate a legacy of hyperglycemia, by epigenetic changes, leading to priming of macrophages and neutrophils into a more pro-inflammatory phenotype [54]. Moreover, elevated levels of glucose and its associated metabolic reprograming in diabetes demonstrates marked increase in glycation of proteins and fatty acids through polyol and hexosamine pathways. End products thus made are referred to AGE [42, 55]. Association of AGE to their cognate receptors on surface of PMNs leads to activation of protein kinase C [56, 57]. These metabolic and biochemical disturbances, both inside and outside of PMNs, translates into increase in superoxide production, and activation of pro-inflammatory conditions in PMNs [53, 54, 58]. While several reports present data on impact of hyperglycemic condition on impairments in PMNs functions, in particular induction of NETosis, none of these reports provide a mechanism for induction of NETosis in diabetes. We developed this study to address this gap in knowledge and performed a series of experiments to explain impairment in processes of NETosis and imbalance between induction of NETs and phagocytosis. Our results demonstrated spontaneous NETosis in conditions of hyperglycemia and in the presence of AGE. Moreover, PMNs isolated from those with poorly controlled T2D showed similar phenotype, where spontaneous NETosis was observed, in the absence of PMA. These results correlated with previous findings presented from several laboratories [30, 31] implicating hyperglycemia in spontaneous NETosis.
Binding of AGE with RAGE increase RAGE protein expression in “feed-forward” loop, which results in activation of cytosolic NADPH-oxidase and production of ROS [59, 60]. Incubation of glycated-albumin demonstrated elevated surface expression of RAGE in healthy PMNs, whereas, in PMNs isolated from those with diabetes showed surface expression of RAGE which correlated with glycemic control. Interaction of AGE-RAGE showed increase levels of ROS and subsequent NETosis. High concentration of glucose has been reported to promote ROS mediated NFKB activation that further increase in the expression of RAGE receptor [61]. These results further confirmed that hyperglycemic conditions and its associated metabolic impairments can sensitize/primes PMNs to NETs expression, while reducing phagocytosis. The low phagocytic capacity of PMNs in the presence of hyperglycemia and low bacterial-killing by diabetic PMNs, could be explained by the fact that spontaneous NETosis was reducing the number of available PMNs required for phagocytosis. Towards that we demonstrated a low level of pneumococcal phagocytosis as compared to NETosis in PMNs cultured in the conditions of hyperglycemia, or in PMNs isolated from those with diabetes. However, blocking NETosis by blocking ROS, restored phagocytosis, suggesting a ROS mediated mechanism that favors NETosis. Another reason for the observed decrease in phagocytosis could be the release of macrovesicles from dying PMNs which are priming neighboring PMNs to commit to NETosis. Alternatively, it was also likely that hyperglcyemia was inducing a downstream pathway that facilitated NETosis in comparison to phagocytosis. The observation that hyperglycemia increases NETosis while reducing phagocytosis was interesting and warranted further investigation.
Previous reports (62–64) have demonstrated glucose mediated autophagy in neurons and renal cells. Autophagy is a key quality control mechanism known to maintain cellular integrity [65, 66]. In condition of hyperglycemia, mitochondrial depolarization, endoplasmic reticulum stress and miss folding of proteins leads to induction of autophagy, targeted to removal of miss folded proteins [67]. These observations suggests that autophagy induction may be a default mechanism to prevent high glucose-induced cellular damage. Elaborating on the role of hyperglycemia in activation of autophagy and subsequent NETosis, we demonstrated that high glucose induces autophagy in human neutrophil in classical PI-3 K dependent fashion. The fact that we observed LCIIIB associated with PMNs isolated from those with poorly controlled diabetes, suggested that metabolic reprograming in response to hyperglycemia leads to induction of autophagy in diabetes. Autophagy leads to chromatin decondensation which is essential for NETosis [20]. Therefore to confirm if autophagy in hyperglycemia will also lead to NETosis, we blocked formation of LCIIIB, using GDC0941, a PI-3kinase inhibitor, which inhibited autophagy and downstream NETosis, even in condition of hyperglycemia. Furthermore, we also observed a strong association of autophagy with NADPH-oxidase, suggesting that reactive oxygen species generated during hyperglycemia, conditions neutrophils for NETosis by activating autophagy. These observations are also supported by previous studies demonstrating that AGE binding to RAGE enhances the expression of Beclin and LCIIIB in cardiomyocytes, increased the number of autophagic vacuoles, and most importantly reduced cell viability in dose dependent manner. Inhibition of RAGE by pretreatment of PMNs with soluble RAGE, decreased Beclin-1 and LCIII expression (68–70). We made similar observations where inhibition of either NADPH oxidase or blocking of pathway upstream of autophagy attenuated NETosis.
In conclusion, our results provide insight into impairment in mechanisms of bacterial killing by neutrophils in the presence of hyperglycemia. Our study is the first to demonstrate that hyperglycemia directly and via a secondary mechanism involving AGE-RAGE association induces oxidative stress, through activation of NADPH-oxidase leading to activation of autophagy and spontaneous NETosis, while reducing phagocytosis. NETs thus generated were short-lived and disintegrated when incubated with pneumococci.
## 5. Conclusion
To our knowledge data presented in this paper provides a holistic understanding of how hyperglycemia induce spontaneous NETosis. We have demonstrated that hyperglycemia directly and via AGEs-RAGE interaction induces oxidative stress, at least in part through activation of NADPH oxidase, resulted in the activation of autophagy and ultimately lead to excessive NETosis and decreased phagocytosis in neutrophils. In conclusion, our findings provides a mechanism for the observed relationship between hyperglycemia and poor bactericidal activity. We were able to demonstrate that the link between hyperglycemia and poor bactericidal activity is elevated levels of ROS, which modulates autophagy switching the bactericidal activity from phagocytic killing to release of NETs.
## 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/s.
## Ethics statement
The studies involving human participants were reviewed and approved by Lahore University of Management Sciences (LUMS), and Shalamar Hospital Institutional Review Boards (IRB). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
AnF and SM designed the research and wrote the manuscript. AnF, GH, SA, ZY, and KS performed the experiments. BY helped with the provision of blood samples. AmF critically reviewed the research design and the manuscript. All authors contributed to the article and approved the submitted version.
## Funding
This work is supported by the Higher Education Commission of Pakistan, grant 5931/Punjab/NRPU/HEC and grant 9319/Punjab/NRPU/HEC.
## 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/fmed.2023.1076690/full#supplementary-material
## References
1. Lin X, Xu Y, Pan X, Xu J, Ding Y, Sun X. **Global, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025**. *Sci Rep* (2020) **10** 1-11. DOI: 10.1038/s41598-020-71908-9
2. Toniolo A, Cassani G, Puggioni A, Rossi A, Colombo A, Onodera T. **The diabetes pandemic and associated infections: suggestions for clinical microbiology**. *Rev Med Microbiol* (2019) **30** 1-17. DOI: 10.1097/MRM.0000000000000155
3. Berbudi A, Rahmadika N, Tjahjadi AI, Ruslami R. **Type 2 diabetes and its impact on the immune system**. *Curr Diabetes Rev* (2020) **16** 442-9. DOI: 10.2174/1573399815666191024085838
4. Carey IM, Critchley JA, DeWilde S, Harris T, Hosking FJ, Cook DG. **Risk of infection in type 1 and type 2 diabetes compared with the general population: a matched cohort study**. *Diabetes Care* (2018) **41** 513-21. DOI: 10.2337/dc17-2131
5. de Santi F, Zoppini G, Locatelli F, Finocchio E, Cappa V, Dauriz M. **Type 2 diabetes is associated with an increased prevalence of respiratory symptoms as compared to the general population**. *BMC Pulm Med* (2017) **17** 101-8. DOI: 10.1186/s12890-017-0443-1
6. Yoo JE, Kim D, Han K, Rhee SY, Shin DW, Lee H. **Diabetes status and association with risk of tuberculosis among Korean adults**. *JAMA Netw Open* (2021) **4** e2126099-9. DOI: 10.1001/jamanetworkopen.2021.26099
7. van Crevel R, Critchley JA. **The interaction of diabetes and tuberculosis: translating research to policy and practice**. *Trop Med Infect Dis* (2021) **6** 8. DOI: 10.3390/tropicalmed6010008
8. Martin ET, Kaye KS, Knott C, Nguyen H, Santarossa M, Evans R. **Diabetes and risk of surgical site infection: a systematic review and meta-analysis**. *Infect Control Hosp Epidemiol* (2016) **37** 88-99. DOI: 10.1017/ice.2015.249
9. Mardhia M, Mahyarudin M, Irsan A. **Antibiotic sensitivity pattern among diabetic outpatients with urinary tract infection in Pontianak**. *Microbiology* (2020) **14** 89-94. DOI: 10.5454/mi.14.3.1
10. Thomsen RW, Hundborg HH, Lervang HH, Johnsen SP, Schonheyder HC, Sorensen HT. **Diabetes mellitus as a risk and prognostic factor for community-acquired bacteremia due to enterobacteria: a 10-year, population-based study among adults**. *Clin Infect Dis* (2005) **40** 628-31. DOI: 10.1086/427699
11. Insuela D, Coutinho D, Martins M, Ferrero M, Carvalho V. **Neutrophil function impairment is a host susceptibility factor to bacterial infection in diabetes**. *Cells Immune Syst* (2019) **2019** 1-22. DOI: 10.5772/intechopen.86600
12. Rosales C. **Neutrophil: a cell with many roles in inflammation or several cell types?**. *Front Physiol* (2018) **9** 113. DOI: 10.3389/fphys.2018.00113
13. Fine N, Tasevski N, McCulloch CA, Tenenbaum HC, Glogauer M. **The neutrophil: constant defender and first responder**. *Front Immunol* (2020) **11** 571085. DOI: 10.3389/fimmu.2020.571085
14. Kruger P, Saffarzadeh M, Weber ANR, Rieber N, Radsak M, von Bernuth H. **Neutrophils: between host defence, immune modulation, and tissue injury**. *PLoS Pathog* (2015) **11** e1004651. DOI: 10.1371/journal.ppat.1004651
15. De Filippo K, Rankin SM. **The secretive life of neutrophils revealed by intravital microscopy**. *Front Cell Dev Biol* (2020) **8** 603230. DOI: 10.3389/fcell.2020.603230
16. Brinkmann V, Reichard U, Goosmann C, Fauler B, Uhlemann Y, Weiss DS. **Neutrophil extracellular traps kill bacteria**. *Science* (2004) **303** 1532-5. DOI: 10.1126/science.1092385
17. Chen F, Yu M, Zhong Y, Wang L, Huang H. **Characteristics and role of neutrophil extracellular traps in asthma**. *Inflammation* (2021) **45** 6-13. DOI: 10.1007/s10753-021-01526-8
18. Morales-Primo AU, Becker I, Zamora-Chimal J. **Neutrophil extracellular trap-associated molecules: a review on their immunophysiological and inflammatory roles**. *Int Rev Immunol* (2022) **41** 253-74. DOI: 10.1080/08830185.2021.1921174
19. Rosazza T, Warner J, Sollberger G. **NET formation–mechanisms and how they relate to other cell death pathways**. *FEBS J* (2021) **288** 3334-50. DOI: 10.1111/febs.15589
20. Azzouz D, Khan MA, Palaniyar N. **ROS induces NETosis by oxidizing DNA and initiating DNA repair**. *Cell Death Discov* (2021) **7** 1-10. DOI: 10.1038/s41420-021-00491-3
21. Leung HH, Perdomo J, Ahmadi Z, Yan F, McKenzie SE, Chong BH. **Inhibition of NADPH oxidase blocks NETosis and reduces thrombosis in heparin-induced thrombocytopenia**. *Blood Adv* (2021) **5** 5439-51. DOI: 10.1182/bloodadvances.2020003093
22. Vorobjeva N, Chernyak B. **NETosis: molecular mechanisms, role in physiology and pathology**. *Biochem Mosc* (2020) **85** 1178-90. DOI: 10.1134/S0006297920100065
23. Moghadam ZM, Henneke P, Kolter J. **From flies to men: ROS and the NADPH oxidase in phagocytes**. *Front Cell Dev Biol* (2021) **9** 628991. DOI: 10.3389/fcell.2021.628991
24. Kenno S, Perito S, Mosci P, Vecchiarelli A, Monari C. **Autophagy and reactive oxygen species are involved in neutrophil extracellular traps release induced by**. *Front Microbiol* (2016) **7** 879. DOI: 10.3389/fmicb.2016.00879
25. Skendros P, Mitroulis I, Ritis K. **Autophagy in neutrophils: from granulopoiesis to neutrophil extracellular traps**. *Front Cell Dev Biol* (2018) **6** 109. DOI: 10.3389/fcell.2018.00109
26. Pang Y, Wu L, Tang C, Wang H, Wei Y. **Autophagy-inflammation interplay during infection: balancing pathogen clearance and host inflammation**. *Front Pharmacol* (2022) **13** 832750. DOI: 10.3389/fphar.2022.832750
27. Pradel B, Robert-Hebmann V, Espert L. **Regulation of innate immune responses by autophagy: a goldmine for viruses**. *Front Immunol* (2020) **11** 578038. DOI: 10.3389/fimmu.2020.578038
28. Guo Y, Gao F, Wang X, Pan Z, Wang Q, Xu S. **Spontaneous formation of neutrophil extracellular traps is associated with autophagy**. *Sci Rep* (2021) **11** 1-10. DOI: 10.1038/s41598-021-03520-4
29. Liang X, Liu L, Wang Y, Guo H, Fan H, Zhang C. **Autophagy-driven NETosis is a double-edged sword–review**. *Biomed Pharmacother* (2020) **126** 110065. DOI: 10.1016/j.biopha.2020.110065
30. Menegazzo L, Ciciliot S, Poncina N, Mazzucato M, Persano M, Bonora B. **NETosis is induced by high glucose and associated with type 2 diabetes**. *Acta Diabetol* (2015) **52** 497-503. DOI: 10.1007/s00592-014-0676-x
31. Wang L, Zhou X, Yin Y, Mai Y, Wang D, Zhang X. **Hyperglycemia induces neutrophil extracellular traps formation through an NADPH oxidase-dependent pathway in diabetic retinopathy**. *Front Immunol* (2019) **9** 3076. DOI: 10.3389/fimmu.2018.03076
32. Bryk AH, Prior SM, Plens K, Konieczynska M, Hohendorff J, Malecki MT. **Predictors of neutrophil extracellular traps markers in type 2 diabetes mellitus: associations with a prothrombotic state and hypofibrinolysis**. *Cardiovasc Diabetol* (2019) **18** 1-12. DOI: 10.1186/s12933-019-0850-0
33. Wong SL, Wagner DD. **Peptidylarginine deiminase 4: a nuclear button triggering neutrophil extracellular traps in inflammatory diseases and aging**. *FASEB J* (2018) **32** 6258-370. DOI: 10.1096/fj.201800691R
34. Wong SL, Demers M, Martinod K, Gallant M, Wang Y, Goldfine AB. **Diabetes primes neutrophils to undergo NETosis, which impairs wound healing**. *Nat Med* (2015) **21** 815-9. DOI: 10.1038/nm.3887
35. Mutua V, Gershwin LJ. **A review of neutrophil extracellular traps (NETs) in disease: potential anti-NETs therapeutics**. *Clin Rev Allergy Immunol* (2021) **61** 194-211. DOI: 10.1007/s12016-020-08804-7
36. Zhu S, Yu Y, Ren Y, Xu L, Wang H, Ling X. **The emerging roles of neutrophil extracellular traps in wound healing**. *Cell Death Dis* (2021) **12** 1-9. DOI: 10.1038/s41419-021-04294-3
37. Niedźwiedzka-Rystwej P, Repka W, Tokarz-Deptuła B, Deptuła W. **“In sickness and in health” – how neutrophil extracellular trap (NET) works in infections, selected diseases and pregnancy**. *J Inflamm* (2019) **16** 15. DOI: 10.1186/s12950-019-0222-2
38. Rossi GA, Fanous H, Colin AA. **Viral strategies predisposing to respiratory bacterial superinfections**. *Pediatr Pulmonol* (2020) **55** 1061-73. DOI: 10.1002/ppul.24699
39. Twaddell SH, Baines KJ, Grainge C, Gibson PG. **The emerging role of neutrophil extracellular traps in respiratory disease**. *Chest* (2019) **156** 774-82. DOI: 10.1016/j.chest.2019.06.012
40. Shafqat A, Abdul Rab S, Ammar O, al Salameh S, Alkhudairi A, Kashir J. **Emerging role of neutrophil extracellular traps in the complications of diabetes mellitus**. *Front Med* (2022) **9** 9. DOI: 10.3389/fmed.2022.995993
41. Stefano GB, Challenger S, Kream RM. **Hyperglycemia-associated alterations in cellular signaling and dysregulated mitochondrial bioenergetics in human metabolic disorders**. *Eur J Nutr* (2016) **55** 2339-45. DOI: 10.1007/s00394-016-1212-2
42. Khalid M, Petroianu G, Adem A. **Advanced glycation end products and diabetes mellitus: mechanisms and perspectives**. *Biomol Ther* (2022) **12** 542. DOI: 10.3390/biom12040542
43. Bayarsaikhan G, Bayarsaikhan D, Lee J, Lee B. **Targeting scavenger receptors in inflammatory disorders and oxidative stress**. *Antioxidants* (2022) **11** 936. DOI: 10.3390/antiox11050936
44. Shah MNAD. **The role of free radicals and reactive oxygen species in biological systems - a comprehensive review**. *Int J Drug Res Dent Sci* (2022) **4** 28-41. DOI: 10.36437/ijdrd.2022.4.3.E
45. Belambri SA, Rolas L, Raad H, Hurtado-Nedelec M, Dang PMC, el-Benna J. **NADPH oxidase activation in neutrophils: role of the phosphorylation of its subunits**. *Eur J Clin Investig* (2018) **48** e12951. DOI: 10.1111/eci.12951
46. Vermot A, Petit-Härtlein I, Smith SME, Fieschi F. **NADPH oxidases (NOX): an overview from discovery, molecular mechanisms to physiology and pathology**. *Antioxidants* (2021) **10** 890. DOI: 10.3390/antiox10060890
47. Rochael NC, Guimarães-Costa AB, Nascimento MTC, DeSouza-Vieira TS, Oliveira MP, Garcia e Souza LF. **Classical ROS-dependent and early/rapid ROS-independent release of neutrophil extracellular traps triggered by Leishmania parasites**. *Sci Rep* (2015) **5** 1-11. DOI: 10.1038/srep18302
48. Uribe-Querol E, Rosales C. **Phagocytosis: our current understanding of a universal biological process**. *Front Immunol* (2020) **11** 1066. DOI: 10.3389/fimmu.2020.01066
49. Winterbourn CC, Kettle AJ, Hampton MB. **Reactive oxygen species and neutrophil function**. *Annu Rev Biochem* (2016) **85** 765-92. DOI: 10.1146/annurev-biochem-060815-014442
50. Toller-Kawahisa JE, O'Neill LA. **How neutrophil metabolism affects bacterial killing**. *Open Biol* (2022) **12** 220248. DOI: 10.1098/rsob.220248
51. Huang Z, Zhang H, Fu X, Han L, Zhang H, Zhang L. **Autophagy-driven neutrophil extracellular traps: the dawn of sepsis**. *Pathol Res Pract* (2022) **234** 153896. DOI: 10.1016/j.prp.2022.153896
52. Jiang G-M, Tan Y, Wang H, Peng L, Chen HT, Meng XJ. **The relationship between autophagy and the immune system and its applications for tumor immunotherapy**. *Mol Cancer* (2019) **18** 1-22. DOI: 10.1186/s12943-019-0944-z
53. Joshi MB, Ahamed R, Hegde M, Nair AS, Ramachandra L, Satyamoorthy K. **Glucose induces metabolic reprogramming in neutrophils during type 2 diabetes to form constitutive extracellular traps and decreased responsiveness to lipopolysaccharides**. *Biochim Biophys Acta Mol Basis Dis* (2020) **1866** 165940. DOI: 10.1016/j.bbadis.2020.165940
54. Choudhury RP, Edgar L, Rydén M, Fisher EA. **Diabetes and metabolic drivers of trained immunity: new therapeutic targets beyond glucose**. *Arterioscler Thromb Vasc Biol* (2021) **41** 1284-90. DOI: 10.1161/ATVBAHA.120.314211
55. Kopytek M, Ząbczyk M, Mazur P, Undas A, Natorska J. **Accumulation of advanced glycation end products (AGEs) is associated with the severity of aortic stenosis in patients with concomitant type 2 diabetes**. *Cardiovasc Diabetol* (2020) **19** 92. DOI: 10.1186/s12933-020-01068-7
56. Metzemaekers M, Gouwy M, Proost P. **Neutrophil chemoattractant receptors in health and disease: double-edged swords**. *Cell Mol Immunol* (2020) **17** 433-50. DOI: 10.1038/s41423-020-0412-0
57. Jangde N, Ray R, Rai V. **RAGE and its ligands: from pathogenesis to therapeutics**. *Crit Rev Biochem Mol Biol* (2020) **55** 555-75. DOI: 10.1080/10409238.2020.1819194
58. Injarabian L, Devin A, Ransac S, Marteyn BS. **Neutrophil metabolic shift during their lifecycle: impact on their survival and activation**. *Int J Mol Sci* (2019) **21** 287. DOI: 10.3390/ijms21010287
59. Bongarzone S, Savickas V, Luzi F, Gee AD. **Targeting the receptor for advanced glycation Endproducts (RAGE): a medicinal chemistry perspective**. *J Med Chem* (2017) **60** 7213-32. DOI: 10.1021/acs.jmedchem.7b00058
60. Otazu GK, Dayyani M, Badie B. **Role of RAGE and its ligands on inflammatory responses to brain tumors**. *Front Cell Neurosci* (2021) **15** 770472-2. DOI: 10.3389/fncel.2021.770472
61. Hu Z, Fang W, Liu Y, Liang H, Chen W, Wang H. **Acute glucose fluctuation promotes RAGE expression via reactive oxygen species-mediated NF-κB activation in rat podocytes**. *Mol Med Rep* (2021) **23** 1-9. DOI: 10.3892/mmr.2021.11969
62. Pontrelli P, Oranger A, Barozzino M, Divella C, Conserva F, Fiore MG. **Deregulation of autophagy under hyperglycemic conditions is dependent on increased lysine 63 ubiquitination: a candidate mechanism in the progression of diabetic nephropathy**. *J Mol Med* (2018) **96** 645-59. DOI: 10.1007/s00109-018-1656-3
63. Cui Y, Fang J, Guo H, Cui H, Deng J, Yu S. **Notch3-mediated mTOR signaling pathway is involved in high glucose-induced autophagy in bovine kidney epithelial cells**. *Molecules* (2022) **27** 3121. DOI: 10.3390/molecules27103121
64. Lai L, Wang Y, Peng S, Guo W, Wei G, Li L. **Bupivacaine induces ROS-dependent autophagic damage in DRG neurons via TUG1/mTOR in a high-glucose environment**. *Neurotox Res* (2022) **40** 111-26. DOI: 10.1007/s12640-021-00461-8
65. Aman Y, Schmauck-Medina T, Hansen M, Morimoto RI, Simon AK, Bjedov I. **Autophagy in healthy aging and disease**. *Nat Aging* (2021) **1** 634-50. DOI: 10.1038/s43587-021-00098-4
66. Nwose EU, Bwititi PT. **Autophagy in diabetes pathophysiology: oxidative damage screening as potential for therapeutic management by clinical laboratory methods**. *Front Cell Dev Biol* (2021) **9** 651776. DOI: 10.3389/fcell.2021.651776
67. Burillo J, Marqués P, Jiménez B, González-Blanco C, Benito M, Guillén C. **Insulin resistance and diabetes mellitus in Alzheimer’s disease**. *Cells* (2021) **10** 1236. DOI: 10.3390/cells10051236
68. Hou X, Hu Z, Xu H, Xu J, Zhang S, Zhong Y. **Advanced glycation endproducts trigger autophagy in cadiomyocyte via RAGE/PI3K/AKT/mTOR pathway**. *Cardiovasc Diabetol* (2014) **13** 78-8. DOI: 10.1186/1475-2840-13-78
69. Sruthi C, Raghu K. **Advanced glycation end products and their adverse effects: the role of autophagy**. *J Biochem Mol Toxicol* (2021) **35** e22710. DOI: 10.1002/jbt.22710
70. Scavello F, Zeni F, Milano G, Macrì F, Castiglione S, Zuccolo E. **Soluble receptor for advanced glycation end-products regulates age-associated cardiac fibrosis**. *Int J Biol Sci* (2021) **17** 2399-416. DOI: 10.7150/ijbs.56379
|
---
title: 'Genotypic variation in Na, K and their ratio in 45 commercial cultivars of
Indian tropical onion: A pressing need to reduce hypertension among the population'
authors:
- Hira Singh
- Mauro Lombardo
- Abhishek Goyal
- Amrender Kumar
- Anil Khar
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9988931
doi: 10.3389/fnut.2023.1098320
license: CC BY 4.0
---
# Genotypic variation in Na, K and their ratio in 45 commercial cultivars of Indian tropical onion: A pressing need to reduce hypertension among the population
## Abstract
The intake of diets with higher sodium (Na) and lower potassium (K) has been considered a leading factor for the development of hypertension (HTN). Majority of junk, processed and packaged food have higher Na contents. To counter the effects of diet on HTN, the identification of high K/Na ratio plant-based food is needed. Among fruits and vegetables, onion could be the ideal option since it contains high K content. Keeping this in mind, 45 commercially well adapted short day Indian onion cultivars were evaluated for K and Na content and their ratio to isolate suitable cultivars to prevent HTN in the Indian population. The data suggested wide variation among the genotypes for K, Na, and K/Na ratio ranging from 490.2 ± 17.0 to 9160.0 ± 96.7 mg/kg on dry matter basis, 52.7 ± 3.0 to 458.2 ± 61.7 mg/kg on dry matter basis and 3.1 ± 0.7 to 109.5 ± 17.3, respectively. The K content was recorded as significantly highest in the yellow-coloured bulb variety “Arka Pitamber” (9160.1 ± 96.7) followed by Pusa Sona (7933.2 ± 292.8). On the other hand, minimal K was assessed in the white-coloured bulb variety “Agrifound White” (490.3 ± 17.0) followed by Udaipur Local (732.9 ± 93.4). Twelve cultivars exhibited > 7000 mg K content, while nine cultivars recorded < 1500 mg. On the contrary, Na was recorded as significantly highest in the dark-red-coloured bulbs and the lowest in white bulbs. Furthermore, it was determined that there was a more than 35-fold difference observed between the highest (109.5) and lowest (3.1) K/Na ratio in the bulbs of tested cultivars. Cluster analysis revealed three major groups comprising of 23, 13 and 9 genotypes. This information could form the base for public health, food and onion researchers to design suitable cultivars to prevent HTN as a population-wide approach. The next century is going to be food-based for the amelioration of human diseases in a sustainable way without any after-effects on the human body.
## 1. Introduction
Globally, non-communicable diseases (NCDs) are becoming the foremost cause of death. Approximately $80\%$ of deaths due to NCDs occur in countries with low to middle incomes [1]. Principally, NCDs consisting of cardiovascular diseases (CVDs), diabetes, various types of cancer and chronic lung dysfunction are responsible for the majority of deaths. According to one of the estimates in 2010, about 1.39 billion individuals ($31.1\%$) in the adult population were suffering from hypertension (HTN); this number has been constantly increasing since then. Therefore, HTN has now become a major health issue worldwide [2]. Among the various factors causing NCDs and metabolic disorders, unhealthy diets are a prominent cause of HTN. According to WHO observations, people from low to middle income countries consume table salt much more than is recommended [1]. According to a previous study, global mean sodium intake was quite high (3.95 gm per day) in 2010 compared to the recommended intake (< 2.3 gm per day) in major published guidelines.
Onion (*Allium cepa* L., 2n = 2× = 16), a bulbous vegetable and condiment crop, belonging to the Amaryllidaceae family, is one of the most important crop which has been domesticated and cultivated worldwide for more than 5000 years due to their peculiar properties as food, their therapeutic value and ethnopharmacological properties. This crop is grown in all climates worldwide (3–6) and is the third most important horticultural crop after potato and tomato [7]. Its bulbs are an enriched source of various health promoting phytochemicals and nutrients and this crop has an utmost valorisation globally due to its multifarious uses in every community and society across the globe. The Queen of French cuisine, Julia Child, stated: “*It is* hard to imagine a civilisation without onions.” Every community, region and country have various traditional and folk remedies but there is a great need to document them in a systematic way to form the foundation of more scientific and modern research on that particular aspect.
The World Health Organization (WHO) also endorses the use of fresh onion extracts for treating colds, coughs, bronchitis, asthma, and appetite loss, as well as relieving hoarseness and preventing atherosclerosis [8]. Because of its naturally possession of higher amounts of flavonoids and widely popular across the world, the onion crop became an interesting and fascinating vegetable [9, 10]. Onions are recommended to lighten blood and lymph stagnation and to improve sexual debility or weakness. Regularly taken on an empty stomach, a mixture of white onion and honey was considered as an exceptional aphrodisiac tonic [11]. Being the leading country in onion production, Indian farmers harvested 26.7 million tons from 1.4-million-hectares [12].
The 21st century is going to work on the principle of “Food as Medicine” and onion will surely play a large role in this. Since antiquity, the bulbous onion has played an important role in human health as it is being used in every kitchen in India. Most of the breeding experiments focused only on enhancing yield and yield-attributing components. However, little focus has been given to improving various quality characteristics. In Indian onions, not much scientific data on nutritional properties are available [13]. Onion bulbs are enriched with potassium, vitamin C, folic acid and dietary fibre, also possessing good amounts of iron and calcium; however, they are lower in sodium and fat (8, 14–16). In USA onion cultivars, Metrani et al. [ 17] quantified 13,550.1 mg/kg potassium in red onion bulbs.
The comprehensive information of the genotypic difference in the potassium and sodium concentration in onion bulbs could be an epitomized contribution for people who are suffering from HTN and prone to CVD. The ratio of potassium and sodium across the genotypes varies with bulb colour and geographical location, which may support the development of future cultivars which are nutrition- and disease-specific.
HTN and CVD may be the result of metabolic syndrome; this has received the global attention of nutrition and health researchers [18]. Potassium and sodium are the most important elements, being essential for normal and proper cellular functioning in the body. As it enhances the risks of high blood pressure, HTN [19, 20], CVD [21, 22], and obesity [23, 24] higher sodium and lower potassium dietary intake has become a serious global health challenge. Global research reports revealed that adverse ratio of both electrolytes is strongly linked to blood pressure [20, 21, 25]. It is well documented that the dietary Na:K ratio is an independent risk factor for metabolic syndrome. Furthermore, it was suggested to modify ratios, including lower Na intakes and higher K intakes, to prevent metabolic disorders [18].
Keeping this in mind, the present study was conducted with the aim (a) to evaluate the potential cultivars representing diverse bulb colours and geographical locations with higher available potassium levels for utilisation in future onion breeding programs, (b) to identify the genotypes of Indian onions possessing the lowest sodium content in their bulbs, and (c) to select cultivars exhibiting higher potassium and sodium ratios for the regulation of blood pressure in hypertensive people.
## 2.1. Location and climate
This experiment was carried out in the Division of Vegetable Science, ICAR-Indian Agricultural Research Institute, New Delhi, which is situated at 28.63oN latitude and 77.15oE longitudes and a mean height of 228 m above mean sea level. This geographical location falls in the Trans-Gangetic agro-climatic zone of India.
## 2.2. Plant material
A total of 45 different commercially grown varieties (Table 1) comprising different bulb colours from white to dark red were collected from different states of the country (representing more than 10 onion-producing states) and evaluated. Seeds of all varieties were maintained and produced during 2019–2020 at the Vegetable Research Farm, Division of Vegetable Science, IARI, New Delhi. After proper cleaning, harvested seeds were stored under ambient conditions. In October 2020, fungicide-treated seeds were sown for nursery production. After 6–7 weeks, the seedlings of all genotypes were transplanted in January 2021. All of the agronomical packages and practices recommended by the IARI for raising successful bulb crops were followed. This experiment was laid out in a Randomised Block Design, with three replications. Each replication included about 200 plants per plot of each variety.
**TABLE 1**
| S. No | Variety | Code | Bulb colour | Institute/ University | Releasing state | Country region |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | Akola Safed | AKLS | White | IARI | Maharashtra | W |
| 2 | Early Grano | EG | Yellow | IARI | New Delhi | N |
| 3 | Bhima Shubra | BSBR | White | DOGR | Maharashtra | W |
| 4 | JWO-1 | JWO1 | White | JAU | Gujarat | W |
| 5 | Bhima Shweta | BSWT | White | DOGR | Maharashtra | W |
| 6 | Pusa Riddhi | PRDI | Red | IARI | New Delhi | N |
| 7 | Pusa White Round | PWR | White | IARI | New Delhi | N |
| 8 | NHRDF Fursungi | NFRS | Red | NHRDF | Maharashtra | W |
| 9 | PKV White | PKVW | White | NHRDF | Maharashtra | W |
| 10 | Bhima Shakti | BSKT | Red | DOGR | Maharashtra | W |
| 11 | Pusa White Flat | PWF | White | IARI | New Delhi | N |
| 12 | VL Pyaz | VLPZ | Red | VPKAS | Uttarakhand | N |
| 13 | RO-252 | R252 | Red | RAU | Rajasthan | N |
| 14 | Udaipur Local | ULCL | Red | RAU | Rajasthan | N |
| 15 | GJWO-3 | GJW3 | White | JAU | Gujarat | W |
| 16 | GJWO-11 | GJ11 | White | JAU | Gujarat | W |
| 17 | JNDWO-085 | JNW8 | White | JAU | Gujarat | W |
| 18 | Arka Pitamber | APTB | Yellow | IIHR | Karnataka | S |
| 19 | Bhima Kiran | BKRN | Red | DOGR | Maharashtra | W |
| 20 | Phursungi Local | PHLC | Pink | NHRDF | Maharashtra | W |
| 21 | Pusa Shobha | PSOB | Brown | IARI | New Delhi | N |
| 22 | Agrifound White | AFW | White | NHRDF | Maharashtra | W |
| 23 | Pusa Sona | PSON | Yellow | IARI | New Delhi | N |
| 24 | Talaja Red | TZRD | Red | JAU | Gujarat | W |
| 25 | JRO-11 | JR11 | Red | JAU | Gujarat | W |
| 26 | Bhima Raj | BRAJ | Red | DOGR | Maharashtra | W |
| 27 | HOS-4 | HOS4 | Red | CCSHAU | Haryana | N |
| 28 | Bhima Light Red | BLRD | Red | DOGR | Maharashtra | W |
| 29 | Pusa Madhavi | PMDV | Red | IARI | New Delhi | N |
| 30 | Arka Bheem | ARBM | Red | IIHR | Karnataka | S |
| 31 | NHRDF Red-4 | NRD4 | Red | NHRDF | Maharashtra | W |
| 32 | L-819 | L819 | Red | NHRDF | Haryana | N |
| 33 | Punjab Naroya | PBNR | Red | PAU | Punjab | N |
| 34 | Bhima Super | BSPR | Red | DOGR | Maharashtra | W |
| 35 | Hisar-2 | HSR2 | Red | CCSHAU | Haryana | N |
| 36 | B-780 | B780 | Red | MPKV | Maharashtra | W |
| 37 | Bhima Safed | BMSF | White | DOGR | Maharashtra | W |
| 38 | Pusa Red | PRED | Red | IARI | New Delhi | N |
| 39 | PRO-6 | PRO6 | Red | PAU | Punjab | N |
| 40 | Kalyanpur Round Red | KRR | Red | CSAUAT | Uttar Pradesh | N |
| 41 | Sukhsagar | SSR | Red | LOCAL | West Bengal | W |
| 42 | Bhima Dark Red | BDR | Red | DOGR | Maharashtra | W |
| 43 | RO-59 | RO59 | Red | RAU | Rajasthan | N |
| 44 | NHRDF-Red L-28 | NL28 | Red | NHRDF | Haryana | N |
| 45 | XP Red | XPR | Red | Local | New Delhi | N |
## 2.3. Estimation of potassium and sodium content
Replication-wise, fully dried samples of the edible portion of the bulb were homogenised using a pestle and mortar. Half a gram of powdered sample (three replications) was taken for digestion in 20 ml of an acid solution of nitric acid (HNO3) and 4-perchloric acid in the ratio of 9:4 and placed in a 500 ml conical flask. The corresponding mixture was kept overnight and was placed on a hot plate the next morning for digestion until white fumes had appeared for about 2 h. After digestion, the clear solution was diluted with double-distilled autoclaved water up to 100 ml. After dilution, the mixture was filtered with Whatman Filter Paper Number-1. An Atomic Absorption Spectrophotometer (Model AA-6880, Shimadzu, Japan) was used to measure absorbance and calculate sodium and potassium contents (Table 2). Air acetylene gas was used for this study. Each sample was measured twice ($$n = 6$$ for each variety, 3 replications and two replicates) to avoid any handling mistakes.
**TABLE 2**
| Element | Symbol | Burner height (mm) | Wavelength for OD value (nm) | R-value |
| --- | --- | --- | --- | --- |
| Potassium | K | 7 | 766.4 | 0.95 |
| Sodium | Na | 7 | 588.0 | 0.94 |
## 2.4. Statistics
Analysis of variance (ANOVA), box plot analysis, DMRT and cluster analysis was calculated by the use of SAS software version 9.3 (SAS Institute, Cary, NC, USA). For cluster analysis, hierarchical clustering technique was used and calculating the distance between the two clusters, a complete linkage algorithm was used which works on the principle of distant neighbours or dissimilarities.
## 3.1. Potassium (K) content (mg/kg of DWB)
A wide genotypic variation in K concentrations was recorded in the onion bulbs (Table 3). The average potassium content in onion bulbs was recorded to be 4679.3 mg/kg on DWB (dry weight basis), whereas it ranged from 490.3 to 9160.1. It was determined that there was a more than 18-fold difference between the highest and lowest potassium contents in the bulbs of tested onion cultivars. It was also observed that bulb colour impacted K concentration.
**TABLE 3**
| S. No | Variety name | Potassium (mg/kg) | Sodium (mg/kg) |
| --- | --- | --- | --- |
| 1 | Arka Pitamber | 9160.1 ± 96.7a | 120.9 ± 13.8mnopq |
| 2 | Pusa Sona | 7933.2 ± 292.8b | 73.8 ± 13.9opq |
| 3 | GJWO-11 | 7875.9 ± 572.8bc | 341.0 ± 31.6de |
| 4 | PRO-6 | 7690.0 ± 188.2bcd | 244.2 ± 13.9fghi |
| 5 | VL Pyaz | 7660.5 ± 68.7bcd | 347.8 ± 39.8de |
| 6 | Bhima Safed | 7514.5 ± 181.8bcde | 292.8 ± 28.8ef |
| 7 | Pusa Red | 7467.1 ± 252.7bcde | 379.7 ± 11.9bcd |
| 8 | Punjab Naroya | 7428.6 ± 332.9bcde | 205.6 ± 38.3hijk |
| 9 | Bhima Dark Red | 7258.5 ± 155.2bcde | 181.9 ± 17.9ijklm |
| 10 | Pusa Riddhi | 7245.7 ± 236.5bcde | 103.4 ± 11.9nopq |
| 11 | Kalyanpur Round Red | 7214.3 ± 130.2bcde | 279.0 ± 46.7efg |
| 12 | Bhima Super | 7186.6 ± 106.4bcde | 303.8 ± 39.5ef |
| 13 | Pusa White Round | 6942.4 ± 220.8bcde | 202.9 ± 29.4hijkl |
| 14 | Pusa Madhavi | 6796.3 ± 433.5bcde | 298.0 ± 39.3ef |
| 15 | Sukhsagar | 6621.9 ± 71.3bcde | 280.6 ± 41.3efg |
| 16 | XP Red | 6539.3 ± 183.4bcde | 446.8 ± 17.6ab |
| 17 | Bhima Kiran | 6485.3 ± 145.3cde | 147.0 ± 12.9jklmno |
| 18 | Akola Safed | 6340.8 ± 135.1de | 68.8 ± 4.2pq |
| 19 | NHRDF-Red L-28 | 6333.3 ± 411.3de | 458.2 ± 61.7a |
| 20 | HOS-4 | 6322.5 ± 284.8de | 202.2 ± 12.4hijkl |
| 21 | RO-252 | 6321.8 ± 125.6de | 421.9 ± 16.9abc |
| 22 | Pusa White Flat | 6142.1 ± 168.3e | 395.2 ± 43.2abcd |
| 23 | Bhima Light Red | 4831.7 ± 69.7f | 129.2 ± 15.0lmnop |
| 24 | Pusa Shobha | 4681.3 ± 109.9fg | 236.5 ± 21.5fghi |
| 25 | Arka Bheem | 4017.2 ± 148.6fgh | 291.6 ± 21.3ef |
| 26 | B-780 | 3647.7 ± 88.0fghi | 352.6 ± 28.9cde |
| 27 | JNDWO-085 | 3429.2 ± 251.3ghi | 382.4 ± 11.2bcd |
| 28 | NHRDF Fursungi | 3368.3 ± 159.9hi | 183.3 ± 11.2ijklm |
| 29 | JRO-11 | 3279.0 ± 10.3hi | 70.2 ± 8.7pq |
| 30 | Hisar-2 | 3207.1 ± 119.9hi | 92.7 ± 11.9nopq |
| 31 | L-819 | 3035.5 ± 161.3hi | 119.3 ± 12.7mnopq |
| 32 | Bhima Shakti | 2529.2 ± 142.8ij | 306.3 ± 32.9ef |
| 33 | Phursungi Local | 1740.7 ± 203.2jk | 133.2 ± 6.7klmnop |
| 34 | Early Grano | 1699.9 ± 178.7jk | 209.4 ± 11.7ghij |
| 35 | NHRDF Red-4 | 1662.2 ± 129.2jk | 263.3 ± 15.7fgh |
| 36 | PKV White | 1551.9 ± 143.2jk | 202.4 ± 12.0hijkl |
| 37 | RO-59 | 1496.3 ± 27.2jk | 291.8 ± 10.4ef |
| 38 | JWO-1 | 1415.8 ± 123.4jk | 121.9 ± 8.5mnopq |
| 39 | Bhima Shubhra | 1230.6 ± 13.5jk | 94.2 ± 6.3nopq |
| 40 | Talaja Red | 1165.1 ± 121.1k | 292.8 ± 40.5ef |
| 41 | Bhima Raj | 928.3 ± 70.8k | 83.7 ± 10.9nopq |
| 42 | Bhima Shweta | 892.7 ± 76.7k | 155.2 ± 13.8jklmn |
| 43 | GJWO-3 | 752.7 ± 112.9k | 112.1 ± 15.2mnopq |
| 44 | Udaipur Local | 732.9 ± 93.4k | 239.0 ± 23.8fghi |
| 45 | Agrifound White | 490.3 ± 17.0k | 52.7 ± 3.0q |
The K content was recorded to be significantly highest in the yellow-coloured bulb variety from Southern India “Arka Pitamber” (9160.1 ± 96.7 mg/kg of DWB) followed by the Pusa Sona (7933.2 ± 292.8 mg/kg of DWB), GJWO-11 (7875.9 ± 572.8 mg/kg of DWB), and PRO-6 (7690.0 ± 188.2 mg/kg of DWB) varieties. However, minimum K content was assessed in the white-coloured bulb variety “Agrifound White” (490.3 ± 17.0 mg/kg of DWB) followed by Udaipur Local (732.9 ± 93.4 mg/kg of DWB), GJWO-3 (752.7 ± 112.9 mg/kg of DWB), Bhima Shweta (892.7 ± 76.7 mg/kg of DWB), and Bhima Raj (928.3 ± 70.8 mg/kg of DWB).
It was concluded that yellow-coloured bulb varieties (Arka Pitamber and Pusa Sona) exhibited significantly higher K contents than red and white varieties. Twenty-three varieties showed higher potassium contents than the overall mean (4679.3), while 22 exhibited lower values.
Twelve cultivars, including Arka Pitamber, Pusa Sona, GJWO-11, PRO-6, VL Pyaz, Bhima Safed, Pusa Red, Punjab Naroya, Bhima Dark Red, Pusa Riddhi, Kalyanpur Round Red, and Bhima Super, exhibited K content of more than 7000 mg/kg on DWB, whereas nine cultivars, including RO-59, JWO-1, Bhima Shubhra, Talaja Red, Bhima Raj, Bhima Shweta, GJWO-3, Udaipur Local and Agrifound White, had a DWB content of less than 1500 mg/kg in their bulbs when evaluated under the trans-gangetic plain zone of New Delhi conditions.
## 3.2. Sodium (Na) content (mg/kg of DWB)
Like K, Na also exhibited broad variation in its concentrations among the tested genotypes (Table 3). The overall average sodium content in onion bulbs was 229.0 mg/kg of DWB, ranging from 52.7 to 458.2. It was determined that there was a more than eightfold difference observed between the highest and lowest Na content in the bulbs of tested cultivars.
In the reverse trend, like K, the Na content was found to be significantly higher in the dark-red-coloured bulb variety “NHRDF-Red L-28” (458.2 ± 61.7 mg/kg of DWB) followed by XP Red (446.8 ± 17.6), RO-252 (421.9 ± 16.9), and Pusa White Flat (395.2 ± 43.2). However, the lowest content was recorded in the white-coloured bulb variety Agrifound White (52.7 ± 3.0 mg/kg of DWB) followed by Akola Safed (68.8 ± 4.15), JRO-11 (70.2 ± 8.7), Pusa Sona (73.8 ± 13.9), and Bhima Raj (83.7 ± 10.9). Among white varieties, the highest Na content was recorded in Pusa White Flat (395.2 ± 43.2), whereas the lowest was found in Agrifound White (52.7 ± 3.0). Twenty-two varieties possessed higher Na than the overall mean (229.0), while 23 recoded values less than this. On the whole, it was further observed that red-coloured varieties elicited higher sodium contents compared to yellow, brown, and white bulb-coloured varieties on a dry weight basis.
Eleven cultivars, including NHRDF-Red L-28, XP Red, RO-252, Pusa White Flat, JNDWO-085, Pusa Red, B-780, VL Pyaz, GJWO-11, Bhima Shakti and Bhima Super, exhibited a DWB Na content of more than 300 mg/kg, whereas ten cultivars including L-819, GJWO-3, Pusa Riddhi, Bhima Shubhra, Hisar-2, Bhima Raj, Pusa Sona, JRO-11, Akola Safed, and Agrifound White elicited a level of less than 120 mg/kg of DWB.
## 3.3. Potassium and sodium (K/Na) ratio
The overall average K/Na ratio in onion bulbs was shown to be 25.5, ranging from 3.1 to 109.6. A greater than 35-fold difference was observed between the highest and lowest ratio in the bulbs of tested onion cultivars (Figure 1).
**FIGURE 1:** *Potassium to sodium ratio in the 45 Indian short day onion cultivars.*
This ratio was significantly higher in the yellow-coloured Northern Indian variety “Pusa Sona” (109.6 ± 17.3) followed by Akola Safed (92.5 ± 7.5), Arka Pitamber (76.5 ± 9.3), Pusa Riddhi (70.6 ± 6.3), and JRO-11 (47.3 ± 6.3). On the other hand, the minimum ratio was estimated in the red-coloured bulb variety from Rajasthan “Udaipur Local” (3.1 ± 0.7) followed by Talaja Red (4.1 ± 0.9), RO-59 (5.1 ± 0.3), Bhima Shweta (5.8 ± 0.9), and NHRDF Red-4 (6.3 ± 0.5). Among the white-coloured bulbs, the maximum ratio was record in Akola Safed (92.6 ± 7.5), while the minimum ratio was reported in GJWO-3 (6.8 ± 1.1). Sixteen varieties showed a higher ratio than the overall mean, while 29 were lower than this value. On the whole, it was further observed that yellow-coloured varieties elicited higher potassium and sodium ratios compared to red, brown, and white bulb-coloured varieties.
Thirteen cultivars, including Pusa Sona, Akola Safed, Arka Pitamber, Pusa Riddhi, JRO-11, Bhima Kiran, Bhima Dark Red, Bhima Light Red, Punjab Naroya, Hisar-2, Pusa White Round, PRO-6 and HOS-4, exhibited a K/Na ratio of more than 30, whereas 11 cultivars, including Agrifound White, JNDWO-085, Bhima Shakti, Early Grano, PKV White, GJWO-3, NHRDF Red-4, Bhima Shweta, RO-59, Talaja Red and Udaipur Local, showed a ratio of less than 10 when evaluated under the trans-gangetic plain zone of New Delhi conditions. Varieties (RO-59 and Udaipur Local) selected from Rajasthan, showed a significant yet very low ratio. While two varieties released by IARI, New Delhi viz., Pusa White Flat (15.7 ± 2.1) and Pusa White Round (34.6 ± 3.8), showed highly significant differences, almost twofold differences were obtained, these varied depending on the shape of the bulbs. This confirmed that the shape of bulbs is also associated with the K/Na ratio.
## 3.4. Impact of bulb colour and region
The box plot analysis, based on region, showed that the mean Na and K were highest in the varieties from the North Indian region (NI) followed by South India (SI) and Western India (WI). The highest mean Na/K ratio was observed in onions from SI followed by NI and WI grown onions (Figure 2). In terms of colour, the highest Na was found in brown onions, followed by red-, white-, pink-, and yellow-coloured onions. The highest K was observed in yellow-coloured onions, followed by brown, red, white, and pink onions. The highest mean Na/K ratio was observed in yellow onion, followed by red, brown, pink, and white onion bulbs (Figure 3).
**FIGURE 2:** *Box plot analysis of K, Na, and K/Na ratio on the basis of region.* **FIGURE 3:** *Box plot analysis of K, Na, and K/Na ratio on the basis of bulb colour.*
## 3.5. Cluster analysis
The Hierarchical Clustering of all the genotypes was done based on distant neighbours or dissimilarities. The dendrogram is presented in Figure 4. The genotypes grouped in clusters and sub-clusters are presented in Table 4. The dendrogram exhibited three major clusters including 23, 13, and 9 genotypes in cluster C1, C2, and C3, respectively. The cluster I was divided into two groups; group C1A and C1B. The group C1A contained one genotype, i.e., Arka Pitamber which is a yellow-coloured bulb variety. The cluster C1B again subdivided into two categories. The cluster C2 grouped 13 genotypes and divided into two categories included 5 and 8 genotypes. The third cluster C3 contained nine genotypes and mainly consisted of red coloured bulb varieties except JNDWO-085. Except few, most of the northern Indian cultivars grouped into cluster C1.
**FIGURE 4:** *Hierarchical Clustering of 45 genotypes for K, Na, and P/Na ratio based on the distant neighbour and dissimilarities.* TABLE_PLACEHOLDER:TABLE 4
## 4. Discussion
Across the globe, people are facing major health issues in the form of metabolic disorders and chronic NCDs. Coronary artery and cerebrovascular (heart stroke) are the most prevalent among cardiovascular diseases [26] and the frequency of deaths due to CVDs increased significantly [27]. HTN was found to be the major risk factor (28–30) for CVDS. The intake of higher amounts of sodium and lower amounts of potassium is one of the main reasons for HTN. There is strong evidence that increasing dietary potassium intake reduces both systolic and diastolic blood pressure. Intake of potassium-enriched diets not only reduces blood pressure, but also the risk factors for various CVD, even in elderly and obese subjects with HTN (31–37). A renowned Canadian physician stated that the “prevalence of arterial hypertension on this continent is in large part due to potash poor diet and an excessive use of salt” [38]. Current diets contain lower potassium (70–80 mmol/day) and higher sodium (150–200 mmol/day), whereas ancestral diets contained much higher potassium (230–300 mmol/day) and negligible (1–10 mmol/day) sodium [39, 40]. With the advancements in diet and lifestyle, a major sustainable change was observed after using artificial salt in cooking led to reduction in dietary intake of potassium [31, 41, 42].
In a recent study, Bibbins-Domingo et al. [ 43] estimated that a reduction in dietary sodium intake of only 1,200 mg per day would reduce the number of stroke cases in the USA from 32,000 to 66,000. Now, it has been scientifically established that a reduction in dietary sodium intake can reduce the risk factors for various CVDs [44]. Furthermore, according to the WHO, K is important for blood pressure regulation in hypertensive persons and recommends at least 3510 mg of potassium per day to maintain blood pressure and reduce the risk of CVD. A meta-analysis of 33 randomised controlled trials concluded that potassium supplementation led to a significant reduction in mean systolic and diastolic blood pressure of 3.1 and 2 mmHg, respectively [45].
As an enriched source of dietary K, along with other beneficial bioactive compounds, onion bulbs could be a potential source for hypertensive people to reduce their elevated blood pressure. In low- and middle-income countries, rapid demographic growth and the lower availability of resources have become essential to create cheaper plant-based functional foods that could replace high-cost allopathic medicines, avoid their side effects and ensure nutritional security. Being a versatile crop, onion bulbs are the best option for Indian people since this crop is used in almost all Indian kitchens. Since antiquity, various physicians have prescribed this crop to prevent various ailments and diseases, well documented in the historical literature. India is also a global leader in onion production. Wide genotypic variation in potassium and sodium concentrations and their ratios was recorded in the Indian short day onion bulbs. Our results were supported by the findings of Metrani et al. [ 17] in the USA. They estimated 12720.7 and 13550.1 mg/kg of DWB K concentrations in red onion long day varieties. However, they found significant differences in the concentration of Na between the two genotypes: 314.1 and 1001.3 mg/kg of DWB. This shows that concentrations of Na and K are highly dependent on the genotype and growing environmental conditions. In onion, growing environmental conditions, agronomic management and genetic makeup of the cultivars determined the variation minerals composition in bulbs.
On the basis of bulb colour, yellow varieties recorded the highest K contents. Varieties like Arka Pitamber and Pusa Sona could be recommended to the Indian population after clinical assessment and bioavailability studies. Twenty-three varieties showed higher potassium contents than the overall mean (4679.3), which clearly showed that the Indian onion population has a higher K content. The major reason of this variability is likely to be due to the genotypic makeup of the cultivar [46]. However, this aspect was not even considered for the exploration of a potential source of dietary K. Our results suggest that onion bulbs may offer a sufficient amount of potassium to meet the recommended dietary allowance (RDA) for humans. However, there is no clear-cut RDA for potassium [47].
The human body requires traces of sodium for some metabolic functions, but the consumption of too much sodium results in elevated blood pressure [1], eventually leading to CVD or heart failure. In the current scenario, the identification of plant-based food with low sodium contents is important for decreasing or minimising the risk of CVD and heart attacks. In the present study, an overall average of 229.0 mg/kg of DWB, ranging from 52.7 to 458.2 was observed. Much higher differences were determined between the highest and lowest values. On the contrary, Na content was significantly highest in the dark-red-coloured variety, while K was highest in yellow-coloured onions. Here, our interest is to identify the genotype with low sodium and high potassium contents for further breeding programs. Eleven cultivars exhibited Na content of more than 300 mg/kg of DWB, whereas ten cultivars recorded values of less than 120. Being versatile, with the peculiar flavour of Indian onions, no commercial hybrid is there at national level from the public sector. Cultivators mostly grow open pollinated varieties, so this study might be useful for breeders aiming to develop hybrids with higher contents of potassium using identified genotypes as parents [8].
Various scientific studies have proven that unhealthy diets and HTN play a chief role in the development of various heart diseases. The intake of higher dietary salt and the lower intake of vegetables and fruits are directly associated with a higher risk of CVDs [41, 48] because of elevated blood pressure. The higher intake of potassium and lower intake of sodium is quietly helpful for regulating blood pressure and decreasing the risk of CVDs, especially in hypertensive adults [44, 49]. Therefore, the WHO recommended reducing the intake of sodium to less than 2000 mg per day [50] and significantly enhancing the intake of potassium in the diet to a minimum of 3510 mg per day to reduce blood pressure [51]. Sodium is usually considered responsible for enhancing blood pressure while potassium antagonistically acts to keep blood pressure within the desired range. Instead of looking at these two elements distinctly, the ratio of the two in the diet has greater significance than the amount of either one alone. Interestingly, the recorded data pertaining to this ratio on the dry weight basis exhibited significant differences and it was further determined that there was a more than 35-fold difference observed between the highest and lowest values.
The K/Na ratio was highest in the yellow-coloured bulb varieties than in red and white onions (Table 5). Sixteen varieties showed higher ratios than the overall mean, while 29 were lower than this. Despite that, the excessive intake of any mineral may prevent other mineral elements from being properly absorbed and utilised in the body. Therefore, the K/Na ratio in onion bulbs is more important to avoid any imbalance. Results of the current research work provide beneficial preliminary information for nutritionists and dieticians involved in developing diet plans and potassium-restricted meals for hypertensive individuals.
**TABLE 5**
| S. No | Variety name | K/Na ratio ± SD |
| --- | --- | --- |
| 1 | Pusa Sona | 109.6 ± 17.3a |
| 2 | Akola Safed | 92.5 ± 7.5b |
| 3 | Arka Pitamber | 76.5 ± 9.3c |
| 4 | Pusa Riddhi | 70.6 ± 6.3c |
| 5 | JRO-11 | 47.3 ± 6.3d |
| 6 | Bhima Kiran | 44.3 ± 3.1de |
| 7 | Bhima Dark Red | 40.2 ± 4.6def |
| 8 | Bhima Light Red | 37.7 ± 4.7efg |
| 9 | Punjab Naroya | 37.1 ± 8.0efg |
| 10 | Hisar-2 | 35.1 ± 5.6fg |
| 11 | Pusa White Round | 34.6 ± 3.8fg |
| 12 | PRO-6 | 31.6 ± 2.4gh |
| 13 | HOS-4 | 31.3 ± 1.4gh |
| 14 | Kalyanpur Round Red | 26.3 ± 4.0hi |
| 15 | Bhima Safed | 25.9 ± 2.9hi |
| 16 | L-819 | 25.5 ± 1.4hi |
| 17 | Sukhsagar | 23.9 ± 3.5hij |
| 18 | Bhima Super | 23.9 ± 2.9hij |
| 19 | GJWO-11 | 23.2 ± 2.5ijk |
| 20 | Pusa Madhavi | 23.0 ± 2.8ijk |
| 21 | VL Pyaz | 22.2 ± 2.5ijkl |
| 22 | Pusa Red | 19.7 ± 1.1ijklm |
| 23 | NHRDF Fursungi | 18.4 ± 2.0ijklmn |
| 24 | Pusa Shobha | 17.2 ± 1.3jklmno |
| 25 | Pusa White Flat | 15.7 ± 2.1klmnop |
| 26 | RO-252 | 15.0 ± 0.9lmnopq |
| 27 | XP Red | 14.6 ± 0.3lmnopqr |
| 28 | NHRDF-Red L-28 | 14.1 ± 2.6mnopqrs |
| 29 | Arka Bheem | 13.8 ± 0.6mnopqrs |
| 30 | Bhima Shubhra | 13.6 ± 0.9mnopq |
| 31 | Phursungi Local | 13.1 ± 2.1mnopqrst |
| 32 | JWO-1 | 11.7 ± 1.8mnopqrstu |
| 33 | Bhima Raj | 11.2 ± 0.9nopqrstuv |
| 34 | B-780 | 10.4 ± 1.0nopqrstuv |
| 35 | Agrifound White | 9.3 ± 0.7opqrstuv |
| 36 | JNDWO-085 | 9.0 ± 0.4pqrstuv |
| 37 | Bhima Shakti | 8.3 ± 0.6pqrstuv |
| 38 | Early Grano | 8.1 ± 0.5pqrstuv |
| 39 | PKV White | 7.7 ± 0.5pqrstuv |
| 40 | GJWO-3 | 6.8 ± 1.1qrstuv |
| 41 | NHRDF Red-4 | 6.3 ± 0.5rstuv |
| 42 | Bhima Shweta | 5.8 ± 0.9stuv |
| 43 | RO-59 | 5.1 ± 0.3tuv |
| 44 | Talaja Red | 4.1 ± 0.9uv |
| 45 | Udaipur Local | 3.1 ± 0.7v |
One of the possible contraindications to onion consumption is the FODMAPs- content. FODMAPs are a category of carbs and fibres that many people cannot tolerate. They may cause unpleasant digestive symptoms, such as bloating, gas, cramping, and diarrhoea. Individuals suffering from irritable bowel syndrome are often intolerant to FODMAPs and may need to avoid onions. [ 52].
## Conclusion
From the findings of the current study, it could be determined that the Indian onion may be considered a potential source of potassium, as well as having low sodium contents. Yellow bulb onions recorded higher potassium levels than red- and white-coloured onions, which may be useful for hypertensive individuals to prevent various CVDs. Further, it should be explored using comprehensive investigations as a potential source of dietary potassium. That information could be used to develop diet plans and further breeding programs to develop specific cultivars for populations with improved potency. Although the comprehensive clinical and physiological implications remain to be established, our findings and information generated on Indian onion further cater to emphasise the need for future studies focused on the development of functional foods as a public health approach. Still, imperative questions with respect to bioavailability and physiological and molecular pathways still need much more attention from plant physiologists. Additionally, molecular level studies for the identification of genotypes with significantly higher potassium-to-sodium ratios will be beneficial for breeders and geneticists aiming to develop new climate smart resilient cultivars with desired quality and nutrition factors. Ultimately, all such novel approaches may help to alleviate HTN effects in the global population, which is an alarming worldwide challenge faced by public health and plant scientists. In the 21st century, dietary interventions to reduce the occurrence of HTN should have immense potential to considerably decrease CVD morbidity and mortality. Furthermore, well-designed clinical trials are required to test the probable effects of various Indian varieties with high potassium-to-sodium ratios on blood pressures and various cardiovascular and cerebrovascular events.
## Data availability statement
The original contributions presented in this study are included in the article/supplementary material, further inquiries can be directed to the corresponding authors.
## Author contributions
HS and AnK: conceptualization, formal analysis, and methodology. HS: investigation and writing – original draft. HS, AnK, and AmK: software. AnK: supervision. HS: writing – original draft. AG, AmK, AnK, HS, and ML: interpretation of data. AnK, ML, and AG: writing – review and editing. All authors read and approved 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. 1.WHO.
Global Status Report on Noncommunicable Diseases.
Geneva: World Health Organization (2010).. (2010)
2. 2.WHO.
World Health Organization: A Global Brief on Hypertension.
Geneva: World Health Organization (2013).. (2013)
3. Bednarz F. **A generalized picture of the onion (**. (1993) **358** 321-4. DOI: 10.17660/ActaHortic.1994.358.53
4. Rabinowitch HD, Currah L.. (2002) 515. DOI: 10.1079/9780851995106.0000
5. Brewster JL.. (2008) 448. DOI: 10.1079/9781845933999.0000
6. Khokhar KM. **Environmental and genotypic effects on bulb development in onion–a review.**. (2017) **92** 448-54. DOI: 10.1080/14620316.2017.1314199
7. 7.FAOSTAT.
Onion Production, Area and Productivity.
Rome: FAOSTA (2022).. (2022)
8. Singh H, Verma P, Khar A. **Screening of short day onion cultivars of India for vitamin-C content.**. (2022) **79** 160-7. DOI: 10.5958/0974-0112.2022.00022.6
9. Griffiths G, Trueman L, Crowther T, Thomas B, Smith B. **Onions–A global benefit to health.**. (2002) **16** 603-15. DOI: 10.1002/ptr.1222
10. Jaggi RC. **Sulphur as production and protection agent in onion (**. (2005) **75** 805-8
11. Singh H, Khar A. **Potential of onion (**. (2022) **92** 11-7. DOI: 10.56093/ijas.v92i11.123235
12. 12.FAO.
World Food and Agriculture – Statistical Yearbook 2021. (2022). Available online at: https://www.fao.org/3/cb4477en/cb4477en.pdf (accessed December 6, 2022).. (2022)
13. Islam S, Khar A, Singh S, Tomar BS. **Variability, heritability and trait association studies for bulb and antioxidant traits in onion (**. (2019) **89** 450-7. DOI: 10.56093/ijas.v89i3.87588
14. Abhayawick L, Laguerre J, Tauzin V, Duquenoy A. **Physical properties of three onion varieties as affected by the moisture content.**. (2002) **55** 253-62. DOI: 10.1016/S0260-8774(02)00099-7
15. Nemeth K, Piskula MK. **Food content, processing, absorption and metabolism of onion flavonoids.**. (2007) **47** 397-409. DOI: 10.1080/10408390600846291
16. Nile SH, Park SW. **Total phenolics, antioxidant and xanthine oxidase inhibitory activity of three coloured onions (**. (2013) **7** 224-8. DOI: 10.1080/21553769.2014.901926
17. Metrani R, Singh J, Acharya P, Jayaprakasha K, Patil SB. **Comparative metabolomics profiling of polyphenols, nutrients and antioxidant activities of two red onion (**. (2020) **9**. DOI: 10.3390/plants9091077
18. Li X, Guo B, Jin D, Wang Y, Jiang Y, Zhu B. **Association of dietary sodium: potassium ratio with the metabolic syndrome in Chinese adults.**. (2018) **120** 612-8. DOI: 10.1017/S0007114518001496
19. Zhang Z, Cogswell ME, Gillespie C, Fang J, Loustalot F, Dai S. **Association between usual sodium and potassium intake and blood pressure and hypertension among U.S. adults: NHANES 2005– 2010.**. (2013) **8**. DOI: 10.1371/journal.pone.0075289
20. Park J, Kwock C, Yang Y. **The effect of the sodium to potassium ratio on hypertension prevalence: a propensity scores matching approach.**. (2016) **8**. DOI: 10.3390/nu8080482
21. Cook NR, Obarzanek E, Cutler JA, Buring JE, Rexrode KM, Kumanyika SK. **Joint effects ofsodium and potassium intake on subsequent cardiovascular disease: the Trials of Hypertension Prevention follow-up study.**. (2009) **169** 32-40. DOI: 10.1001/archinternmed.2008.523
22. Yang Q, Liu T, Kuklina EV, Flanders WD, Hong Y, Gillespie C. **Sodium and potassium intake and mortality among US adults: prospective data from the third national health and nutrition examination survey.**. (2011) **171** 1183-91. DOI: 10.1001/archinternmed.2011.257
23. Ge Z, Zhang J, Chen X, Yan L, Guo X, Lu Z. **Are 24 h urinary sodium excretion and sodium:potassium independently associated with obesity in Chinese adults?**. (2016) **19** 1074-80. DOI: 10.1017/S136898001500230X
24. Jain N, Minhajuddin AT, Neeland IJ, Elsayed EF, Vega GL, Hedayati SS. **Association of urinary sodium-to-potassium ratio with obesity in a multiethnic cohort.**. (2014) **99** 992-8. DOI: 10.3945/ajcn.113.077362
25. Okayama A, Okuda N, Miura K, Okamura T, Hayakawa T, Akasaka H. **Dietary sodiumto-potassium ratio as a risk factor for stroke, cardiovascular disease and all-cause mortality in Japan: the NIPPON DATA80 cohort study.**. (2016) **6**. DOI: 10.1136/bmjopen-2016-011632
26. Mendis S, Puska P, Norrving B.. (2011)
27. Krishnamurthi R, Moran A, Feigin V, Barker-Collo S, Norrving B, Mensah G. **Stroke prevalence, mortality and disability-adjusted life years in adults aged 20-64 years in 1990-2013: data from the global burden of disease 2013 study.**. (2015) **45** 190-202. DOI: 10.1159/000441098
28. Roger V, Go A, Lloyd-Jones D, Benjamin E, Berry J, Borden W. **Heart disease and stroke statisticsd2 012 update.**. (2012) **125** e2-220. PMID: 22179539
29. 29.WHO.
The World Health Report 2002: Reducing Risks, Promoting Healthy Life.
Geneva: World Health Organization (2002).. (2002)
30. Appel LJ, Frohlich E, Hall J, Pearson T, Sacco R, Seals D. **The importance of population-wide sodium reduction as a means to prevent cardiovascular disease and stroke.**. (2011) **123** 1138-43. DOI: 10.1161/CIR.0b013e31820d0793
31. Penton D, Czogalla J, Loffing J. **Dietary potassium and the renal control of salt balance and blood pressure.**. (2015) **467** 513-30. DOI: 10.1007/s00424-014-1673-1
32. Tannen R. **Effects of potassium on blood pressure control.**. (1983) **98** 773-80. DOI: 10.7326/0003-4819-98-5-773
33. Barri Y, Wingo C. **The effects of potassium depletion and supplementation on blood pressure: a clinical review.**. (1997) **314** 37-40. DOI: 10.1097/00000441-199707000-00008
34. Aburto N, Hanson S, Gutierrez H, Hooper L, Elliott P, Cappuccio F. **Effect of increased potassium intake on cardiovascular risk factors and disease: systematic review and meta-analyses.**. (2013) **346**. DOI: 10.1136/bmj.f1378
35. Mente A, O’Donnell M, Rangarajan S, McQueen M, Poirier P, Wielgosz A. **Association of urinary sodium and potassium excretion with blood pressure.**. (2014) **371** 601-11. DOI: 10.1056/NEJMoa1311989
36. O’Donnell M, Mente A, Rangarajan S, McQueen M, Wang X, Liu L. **Urinary sodium and potassium excretion, mortality, and cardiovascular events.**. (2014) **371** 612-23. PMID: 25119607
37. He F, MacGregor G. **Beneficial effects of potassium on human health.**. (2008) **133** 725-35. DOI: 10.1111/j.1399-3054.2007.01033.x
38. Addison W. **The use of sodium chloride, potassium chloride, sodium bromide, and potassium bromide in cases of arterial hypertension which are amenable to potassium chloride.**. (1988) **46** 295-6. DOI: 10.1111/j.1753-4887.1988.tb05460.x
39. Meneton P, Loffing J, Warnock D. **Sodium and potassium handling by the aldosterone-sensitive distal nephron: the pivotal role of the distal and connecting tubule.**. (2004) **287** 593-601. DOI: 10.1152/ajprenal.00454.2003
40. Castro H, Raij L. **Potassium in hypertension and cardiovascular disease.**. (2013) **33** 277-89. DOI: 10.1016/j.semnephrol.2013.04.008
41. He FJ, MacGregor GA. **Reducing population salt intake worldwide: from evidence to implementation.**. (2010) **52** 363-82. DOI: 10.1016/j.pcad.2009.12.006
42. Aaron K, Sanders P. **Role of dietary salt and potassium intake in cardiovascular health and disease: a review of the evidence.**. (2013) **88** 987-95. DOI: 10.1016/j.mayocp.2013.06.005
43. Bibbins-Domingo K, Chertow G, Coxson P, Moran A, Lightwood J, Pletcher M. **Projected effect of dietary salt reductions on future cardiovascular disease.**. (2010) **362** 590-9. DOI: 10.1056/NEJMoa0907355
44. Jayedi A, Ghomashi F, Zargar MS, Shab-Bidar S. **Dietary sodium, sodium-to-potassium ratio, and risk of stroke: a systematic review and nonlinear dose-response meta-analysis.**. (2019) **38** 1092-100. DOI: 10.1016/j.clnu.2018.05.017
45. Whelton PK, He J, Cutler JA, Brancati FL, Appel LJ, Follmann D. **Effects of oral potassium on blood pressure: meta-analysis of randomized controlled clinical trials.**. (1997) **277** 1624-32. DOI: 10.1001/jama.277.20.1624
46. Chope GA, Terry LA. **Use of canonical variate analysis to differentiate onion cultivars by mineral content as measured by ICP-AES.**. (2009) **115** 1108-13. DOI: 10.1016/j.foodchem.2008.12.090
47. Kumssa D, Joy E, Broadley M. **Global trends (1961–2017) in human dietary potassium supplies.**. (2021) **13**. DOI: 10.3390/nu13041369
48. Aune D, Giovannucci E, Boffetta P, Fadnes LT, Keum N, Norat T. **Fruit and vegetable intake and the risk of cardiovascular disease, total cancer and all-cause mortality—A systematic review and dose-response meta-analysis of prospective studies.**. (2017) **46** 1029-56. DOI: 10.1093/ije/dyw319
49. O’Donnell M, Mente A, Rangarajan S, McQueen MJ, O’Leary N, Yin L. **Joint association of urinary sodium and potassium excretion with cardiovascular events and mortality: prospective cohort study.**. (2019) **364**. DOI: 10.1136/bmj.l772
50. 50.WHO.
World Health Organization Guideline: Sodium Intake for Adults and Children.
Geneva: World Health Organization (2012).. (2012)
51. 51.WHO.
World Health Organization Guideline: Potassium Intake for Adults and Children.
Geneva: World Health Organization (2012).. (2012)
52. Böhn L, Störsrud S, Liljebo T, Collin L, Lindfors P, Törnblom H. **Diet low in FODMAPs reduces symptoms of irritable bowel syndrome as well as traditional dietary advice: a randomized controlled trial.**. (2015) **149** 1399-1407.e2. DOI: 10.1053/j.gastro.2015.07.054
|
---
title: Inhibition of matrix metalloproteinases by HIV-1 integrase strand transfer
inhibitors
authors:
- Emma G. Foster
- Nicholas Y. Palermo
- Yutong Liu
- Benson Edagwa
- Howard E. Gendelman
- Aditya N. Bade
journal: Frontiers in Toxicology
year: 2023
pmcid: PMC9988942
doi: 10.3389/ftox.2023.1113032
license: CC BY 4.0
---
# Inhibition of matrix metalloproteinases by HIV-1 integrase strand transfer inhibitors
## Abstract
More than fifteen million women with the human immunodeficiency virus type-1 (HIV-1) infection are of childbearing age world-wide. Due to improved and affordable access to antiretroviral therapy (ART), the number of in utero antiretroviral drug (ARV)-exposed children has exceeded a million and continues to grow. While most recommended ART taken during pregnancy suppresses mother to child viral transmission, the knowledge of drug safety linked to fetal neurodevelopment remains an area of active investigation. For example, few studies have suggested that ARV use can be associated with neural tube defects (NTDs) and most notably with the integrase strand transfer inhibitor (INSTI) dolutegravir (DTG). After risk benefit assessments, the World Health Organization (WHO) made recommendations for DTG usage as a first and second-line preferred treatment for infected populations including pregnant women and those of childbearing age. Nonetheless, long-term safety concerns remain for fetal health. This has led to a number of recent studies underscoring the need for biomarkers to elucidate potential mechanisms underlying long-term neurodevelopmental adverse events. With this goal in mind, we now report the inhibition of matrix metalloproteinases (MMPs) activities by INSTIs as an ARV class effect. Balanced MMPs activities play a crucial role in fetal neurodevelopment. Inhibition of MMPs activities by INSTIs during neurodevelopment could be a potential mechanism for adverse events. Thus, comprehensive molecular docking testing of the INSTIs, DTG, bictegravir (BIC), and cabotegravir (CAB), against twenty-three human MMPs showed broad-spectrum inhibition. With a metal chelating chemical property, each of the INSTI were shown to bind Zn++ at the MMP’s catalytic domain leading to MMP inhibition but to variable binding energies. These results were validated in myeloid cell culture experiments demonstrating MMP-2 and 9 inhibitions by DTG, BIC and CAB and even at higher degree than doxycycline (DOX). Altogether, these data provide a potential mechanism for how INSTIs could affect fetal neurodevelopment.
## Introduction
Pregnant women and women of child bearing age infected with the human immunodeficiency virus type-1 (HIV-1) infection have benefited by antiretroviral therapy (ART) in the reduction of maternal fetal viral transmission (The U.S. Department of Health and Human Services, 2015; World Health Organization (WHO), 2019a). Currently, more than 15.5 million women of child-bearing age are HIV-1 infected, worldwide (The Joint United Nations Programme on HIV/AIDS (UNAIDS), 2021a). In 2020, eighty five percent of HIV-1-infected pregnant women were on ART (The Joint United Nations Programme on HIV/AIDS (UNAIDS), 2021a). Due to such broad usage of ART during pregnancy, the rate of vertical transmission of HIV-1 has reduced to less than $1\%$ (The Centers for Disease Control and Prevention (CDC), 2018; Peters et al., 2017; Schnoll et al., 2019; Rasi et al., 2022; The Joint United Nations Programme on HIV/AIDS (UNAIDS), 2021b). This includes resource-limited countries (RLCs), which currently hold up to two-thirds of the world’s total HIV-1 infected population (The Joint United Nations Programme on HIV/AIDS (UNAIDS), 2021b). However, along with the significant benefits in reducing infection-associated morbidities and mortalities, there remains risks of ART-linked adverse events (Hill et al., 2018). As over a million ARV-exposed HIV-1 uninfected children are born each year (Ramokolo et al., 2019; Crowell et al., 2020), an appreciation of adverse pregnancy events, in particular, related to ARVs is certainly warranted.
Herein, we particularly focused on HIV-1 integrase strand transfer inhibitors (INSTIs), a relatively new class of ARVs. Raltegravir (RAL), elvitegravir (EVG), dolutegravir (DTG), bictegravir (BIC), and cabotegravir (CAB) are the US Food and Drug Administration (FDA) approved INSTIs for the treatment of HIV-1 infected patients (Smith et al., 2021). In recent years, widespread usage of INSTIs have emerged related to their efficacy and high barrier to viral drug resistance (Smith et al., 2021). Indeed, these antiretrovirals are currently part of preferred first- and second-line ART regimens (World Health Organization (WHO), 2016; Department of Health and Human Services (DHHS), 2022). Moreover, increasing pretreatment resistance to non-nucleoside reverse transcriptase inhibitors (NNRTIs) in RLCs, especially in women, increases usage of INSTI-based regimens (World Health Organization (WHO), 2019a; World Health Organization (WHO), 2019b). During pregnancy, DTG and RAL are preferred drugs in combination therapy with a preferred dual-nucleoside reverse transcriptase inhibitor (NRTI) backbone. EVG, BIC or CAB are not recommended during pregnancy due to limited safety data (The U.S. Department of Health and Human Services, 2015). Recently DTG was found to be potentially associated with birth defects (NTDs) and postnatal neurodevelopmental abnormalities (Hill et al., 2018; Cabrera et al., 2019; Zash et al., 2019; Crowell et al., 2020; Mohan et al., 2020; Bade et al., 2021). Given the widescale usage of DTG as a part of first-line regimens worldwide (Hill et al., 2018; Dorward et al., 2018; World Health Organization (WHO), 2018; The Lancet, 2020) and emerging potent INSTIs such as BIC and CAB, uncovering any INSTIs-associated adverse effects and thus, the underlying mechanisms is of importance.
Pre-clinical and clinical research have served to evaluate interaction between folate levels or transport pathways and DTG or other INSTIs for any developmental toxicity (Cabrera et al., 2019; Chandiwana et al., 2020; Gilmore et al., 2022). However, results have failed to conclusively establish cause-and-effect relationships (Cabrera et al., 2019; Chandiwana et al., 2020; Gilmore et al., 2022). No other biomarker linked to INSTI drug-induced adverse events has been explored. We demonstrated that DTG is a broad-spectrum inhibitor of matrix metalloproteinases (MMPs) (Bade et al., 2021). MMPs are known to play a role in many neurodevelopmental processes, including, but not limited to axonal growth and guidance, synaptic development and plasticity (Ethell and Ethell, 2007; Agrawal et al., 2008; Fujioka et al., 2012; Reinhard et al., 2015; De Stefano and Herrero, 2017). Therefore, dysregulation of their activities could affect fetal neurodevelopment (Reinhard et al., 2015; Bade et al., 2021). Docking assessments against five MMPs showed that DTG binds to Zn++ at the catalytic domain of an MMP to inhibit the enzyme’s activity. Moreover, such MMPs inhibition can affect mice fetal neurodevelopment following DTG administration to pregnant dams at the time of conception. Clinical reports of adverse events associated with INSTIs have demonstrated class effects. Therefore, it is prudent to determine whether other ARVs from the INSTI class are inhibitors of MMPs and consider this as a potential mechanism of INSTIs-related adverse neurodevelopmental outcomes. Moreover, such biomarker discovery against MMP enzymes will help to understand potential genetic susceptibility. Herein, we show, for the first time, comprehensive computational molecular docking assessments of DTG, BIC or CAB against each one of the twenty-three human MMP enzymes. Further, inhibition potency of each INSTI was validated using a cell culture model. To this end, we show that inhibition of MMPs activities is an INSTI class effect and warranting assessments to determine the effect of drug-induced effects on the gestational environment and fetal neurodevelopment.
## Molecular docking
Homology models of all 23 known human MMPs (MMP-1, 2, 3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20, 21, 23, 24, 25, 26, 27, and 28) were generated. This was done on a template of MMP-2 (PDB ID: 1HOV) using the Homology Modeling module of the YASARA Structure program package (Krieger and Vriend, 2014). The Schrodinger software suite release 2020–4 (New York, NY) was used for all molecular dynamic simulations and molecular docking calculations. All molecules were parametrized using the OPLS3e force field (Harder et al., 2016). Each homology model was placed in an orthorhombic box of TIP4P water with periodic boundaries; at least 10 Å from any solute molecule. The simulation cells were neutralized with the addition of Na + or Cl− ions. Production molecular dynamics were run for 500 ns with default settings. The representative structure of the largest cluster from each simulation was chosen for docking calculations. Induced-fit binding as implemented in Schrodinger was used with default settings, except that the high-accuracy XP mode was chosen for Glide docking. All ranked poses were required to have at least one bond with the active site zinc ion; other poses were not considered.
## Gelatin zymography
Gelatin zymography was performed to assess MMP-9 and -2 activity following treatment of THP-1 cells with DTG, CAB, BIC, or DOX. This assay was used as preliminary confirmation of the inhibition of MMPs by individual INSTIs. Due to the nature of this assay, only the gelatinases, MMP-2 and -9, could be assessed. Cells were plated at a density of 1 × 106 in 12 well plates and treated with phorbol-12-myristate-13-acetate (PMA) for 24 h. This was done to promote cell differentiation to stimulate MMP secretion. Following PMA treatment, cells were treated with DTG, CAB, BIC, or DOX at concentrations of 25, 50, 75, or 100 µM or control vehicle for 24 h. In our previous study, no DTG-induced cytotoxicity was recorded in PMA-stimulated THP-1 cells up to 100 µM (Bade et al., 2021). Thus, for comparative assessments among different INSTIs (DTG, BIC, and CAB) and DOX (positive control) drug concentrations of up to 100 µM were utilized. Each of the experimental tests were performed in triplicate. Following treatment, media was collected and centrifuged at 15,000 x g for 10 min at 4°C. Supernatant was collected and stored at −80°C for further analysis. For gelatin zymography, 3 µg of protein from cell medium was loaded in a $10\%$ SDS-polyacrylamide gel containing $0.1\%$ gelatin. Gels were ran at 55 V until the loading dye passed through its bottom. The gel was then removed and washed with water for 15 min, then incubated with renaturation buffer [$2.5\%$ (v/v) Triton X-100 in Milli-Q water] for 90 min at room temperature. The used renaturation buffer was replaced with fresh buffer every 30 min. Renaturation buffer was then replaced with developing buffer (50 mM Tris–HCl, pH 7.5, 5 mM CaCl2, 0.2 M NaCl, and $0.02\%$ Brij-35) and the gel was incubated at 37°C in a shaker (Innova 42, New Brunswick Scientifc, Edison, NJ) for 48 h. After 48 h, the gel was washed with water for 15 min and then stained using $0.2\%$ Coomassie Brilliant Blue R-250 (BIO-RAD, Hercules, CA) for 1 h. After staining, the gel was washed with water for 15 min before washing with destaining solution ($30\%$ methanol, $10\%$ acetic acid, $60\%$ water) for 45 min. The gel was then washed with water for 20 min to remove any destaining solution. Finally, the stained gel was imaged using the iBright 750 Imaging System (Invitrogen, Carlsbad, CA). ImageJ software was used to quantitate band density recorded as a measure of relative MMP activity.
## Statistical analysis
Statistical analyses were conducted using GraphPad Prism 7.0 software (La Jolla, CA). Data from in vitro studies were expressed as mean ± standard error of the mean (SEM) with a minimum of 3 biological replicates. A one-way ANOVA followed by Tukey’s or Dunnett’s test was used to compare three or more groups. Statistical significance was denoted as *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$, ****$p \leq 0.0001.$
## INSTIs chelate Zn++ at the catalytic domain of MMPs
Molecular dockings were completed using Schrodinger’s software to identify the mechanism through which each INSTI interacts with the catalytic domain structures of the human MMPs. Here, we used DTG, BIC and CAB for assessments. MMPs are Zn++ dependent endopeptidases. INSTIs possess a prominent metal-binding pharmacophore (MBP) also, referred to as a metal-binding group or MBG in their chemical structure. Based on these chemical abilities for binding to the metal ions we hypothesized that DTG, BIC or CAB can inhibit MMPs activities by binding to Zn++ at the catalytic domain of the protein structure. Herein, induced fit docking used a combination of the Glide and Prime programs in the Schrodinger suite. All docking scores used the highest accuracy Glide XP mode. Previously, we reported interaction of DTG with five MMPs, MMP-2, 8, 9, 14, and 19 and interaction of CAB or BIC with MMP-2 and -14 as proof-of-concept evaluations (Bade et al., 2021). Herein, as 23 MMPs are known to be found in humans, molecular docking interaction was tested against each of these enzymes to find the highest binding interaction for DTG, BIC or CAB and determine whether any individual MMP could have genetic susceptibility against these INSTIs.
DTG formed a metal coordination complex with Zn++. This was recorded in the catalytic domain of each of the MMPs tested. Metal coordination of DTG with Zn++ occurred at Zn 166, 166, 479, 269, 469, 709, 478, 490, 472, 473, 584, 671, 609, 605, 510, 485, 571, 485, 647, 564, 263, 515, and 522 receptors of MMP-1, 2, 3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20, 21, 23, 24, 25, 26, 27, and 28, respectively. DTG also formed other interactions with Zn++, which included cation pi interactions. These interactions occurred at Zn 490, 605, and 522 receptors of MMP-11, 17, and 28 respectively. Other interactions included pi stacking and hydrogen bonding. Pi stacking occurred with histidine amino acids of all tested MMPs except MMP-7, 9, 19, and 27. Pi stacking also occurred with tyrosine amino acids, but only with MMP-3, 11, and 16. Hydrogen bond interactions occurred at glutamate amino acid residues of MMP-1, 3, 10, 12, 15, 16, 19, 20, 21, 23, 24, 25, and 26; alanine amino acid residues of MMP-2, 7, 8, 11, 14, 17, and 19; leucine amino acid residues of MMP-7, 8, 9, 10, 12, 15, 16, 20, 23, 24, 25, and 26; glycine amino acid residues of MMP-9 and 27; tyrosine amino acid residues of MMP-9 and 27; asparagine 170 amino acid residue of MMP-8; proline 421 amino acid residue of MMP-9; phenylalanine 249 of MMP-21; and glutamine 247 of MMP-21. The distances of all the receptor-ligand bonds are shown in the respective DTG-MMP interaction table (Table 1). Docking simulation of DTG into individual MMP showed binding energy of −6.032, −6.450, −6.253, −7.243, −8.330, −9.430, −6.686, −6.210, −6.713, −6.461, −9.040, −5.885, −7.325, −6.810, −7.130, −6.097, −6.179, −6.213, −6.917, −6.865, −7.198, −7.427 or −6.012 kcal/mol for MMP-1 to −28, respectively (Table 4). Overall, observed high binding energies from the docking simulation validated docking interactions in Table 1. Moreover, these docking assessments confirmed that DTG is a broad-spectrum inhibitor, and it inhibits all MMPs activities by chelating Zn++ at the catalytic domain.
**TABLE 1**
| MMP-1 interactions | MMP-1 interactions.1 | MMP-1 interactions.2 | MMP-1 interactions.3 |
| --- | --- | --- | --- |
| Ligand | Receptor | Type | Distance (Å) |
| Ar1 | His 218 | Pi stacking | 3.60 |
| NH | Glu 219 | Hydrogen bond | 1.98 |
| O2 | Zn 166 | Metal coordination | 2.01 |
Notably, CAB also formed a metal coordination complex with Zn++ in the catalytic domain of all tested MMPs. Metal coordination of CAB with Zn++ occurred at Zn 471, 166, 479, 269, 469, 709, 478, 490, 472, 473, 584, 671, 609, 605, 510, 485, 571, 392, 647, 564, 263, 515, and 522 receptors of MMP-1, 2, 3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20, 21, 23, 24, 25, 26, 27, and 28, respectively. However, CAB also formed salt bridges with Zn++ of all tested MMPs. In addition, salt bridge interactions occurred with glutamate amino acids of MMP-1, 8, 9, 10, 12, 14, 17, 19, 21, 23, 24, and 26. Salt bridges with Zn++ or with other amino acids were not observed with any of DTG-MMP interactions. Other interactions between CAB and MMPs were cation pi, pi stacking and hydrogen bonding. Cation pi interactions occurred at histidine 263 and phenylalanine 205 of MMP-15 and 16 respectively. Pi stacking interactions occurred with histidine amino acids of MMP-1, 2, 11, 12, 15, 17, 19, 21, 23, 25, and 28; phenylalanine amino acids of MMP-1 and 16; tyrosine amino acids of MMP-2, 14, and 19. Hydrogen bonding of CAB with tested MMPs was found, except MMP-17. Like DTG, CAB was found to produce hydrogen bonding with leucine, alanine, glutamate, phenylalanine, glycine, proline, asparagine, and tyrosine amino acid residues. However, other hydrogen bonding occurred at serine 239 amino acid residue of MMP-1, valine 233 amino acid residue of MMP-19, and arginine 240 amino acid residue of MMP-23. The distances of all the receptor-ligand bonds are shown in the respective CAB-MMP interaction table (Table 2). Docking simulation of CAB into individual MMP showed binding energy of −14.251, −8.588, −15.222, −12.305, −14.337, −14.222, −14.592, −10.352, −12.19, −11.798, −12.983 −9.632, −12.249, −11.666, −12.718, −12.413, −14.389, −13.109, −15.339, −11.62, −12.381, −12.614, or −10.11 kcal/mol for MMP-1 to −28, respectively (Table 4). Altogether, observed high binding energies from the docking simulation and docking interactions (Table 1; Table 4) evaluations confirmed that CAB is a broad-spectrum inhibitor, and it inhibits all MMPs activities by binding to Zn++ at the catalytic domain.
**TABLE 2**
| MMP-1 Interactions | MMP-1 Interactions.1 | MMP-1 Interactions.2 | MMP-1 Interactions.3 |
| --- | --- | --- | --- |
| Ligand | Receptor | Type | Distance (Å) |
| Ar1 | His 218 | Pi stacking | 4.9 |
| Ar1 | Phe 242 | Pi stacking | 5.28 |
| O2 | Ser 239 | Hydrogen bond | 2.28 |
| O2 | Zn 471 | Salt bridge | 2.14 |
| O3 | Zn 471 | Metal coordination | 2.08 |
| O3 | Zn 471 | Salt bridge | 2.08 |
| N2 | Glu 219 | Salt bridge | 4.91 |
Further, docking simulation confirmed that BIC formed a metal coordination complex with Zn++ in the catalytic domain of all tested MMPs, validating that all INSTIs possess chemical abilities to inhibits MMPs activities by chelating Zn++ at the catalytic domain. Metal coordination of BIC with Zn++ occurred at Zn 471, 166, 479, 269, 469, 709, 478, 490, 472, 473, 584, 671, 609, 605, 510, 485, 571, 392, 647, 564, 263, 515, and 522 receptors of MMP-1, 2, 3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20, 21, 23, 24, 25, 26, 27, and 28, respectively. Like CAB, BIC also formed salt bridges with Zn++ of all tested MMPs. Along with Zn++, BIC was found to form salt bridge interactions with glutamate amino acids of all tested MMPs except MMP-2, 13, 15, 24 and 26. Cation pi interactions occurred only at phenylalanine 241 of MMP-24. Pi stacking interactions occurred with histidine amino acids of MMP-1, 9, 14, 17, 24, 26, 27, and 28; tyrosine amino acids of MMP-3, 10, 16, and 19, phenylalanine amino acids of MMP-12 and 24; tryptophan amino acid of MMP-26. Hydrogen bonding of BIC was observed with all MMPs except MMP-2, 19, 28. Like DTG and CAB, BIC was found to form hydrogen bonds with asparagine, histidine, leucine, alanine, proline, glutamate, valine, phenylalanine, glycine, and arginine amino acid residues. The distances of all the receptor-ligand bonds are shown in the respective BIC-MMP interaction table (Table 3). Docking simulation of BIC into individual MMP showed binding energy of −14.55, −10.972, −15.082, −11.385, −15.771, −13.415, −16.021, −12.143, −13.482, −12.73, −14.539, −8.877, −14.625, −12.409, −11.695, −11.923, −12.698, −12.25, −10.181, −11.716, −11.971, −12.54, and −10.823 kcal/mol for MMP-1 to −28, respectively (Table 4). Observed high binding energies and docking interactions (Table 1; Table 4) confirmed that BIC is a broad-spectrum MMPs inhibitor.
For comparative evaluations, docking simulations were also performed using the known broad-spectrum MMPs inhibitor DOX. DOX is the only US Food and Drug Administration (FDA)-approved broad-spectrum MMPs inhibitor. These were performed against five MMPs. These included MMP-2, 8, 9, 14, and 19. These MMPs were selected as INSTIs have higher binding energies with these enzymes compared to others and each enzyme represented different class of the MMP family. DOX was found to form metal coordination with Zn++ for all tested MMPs. These metal co-ordinations occurred at Zn 166, 469, 709, 584, and 510 receptors of MMP-2, 8, 9, 14, and 19 respectively. Pi stacking interactions occurred with histidine amino acids of MMP-2 and 14. Hydrogen bonding of DOX occurred with all five MMPs tested. DOX was found to form hydrogen bonds with glutamate, alanine, aspartic acid, tyrosine, asparagine, serine, and proline amino acid residues. The distances of receptor-ligand bonds are shown in the DOX-MMP interaction table (Supplementary Table S1). Further, docking simulations of DOX into individual MMPs showed binding energies of −6.595, −7.024, −7.658, −7.114, and −6.488 kcal/mol for MMP-2, 8, 9, 14, and 19 respectively. In comparison to DOX, all three INSTIs (DTG, CAB and BIC) showed higher binding energies with each tested MMP. Interestingly, both, CAB and BIC, showed significantly higher energies compared to DOX and DTG, suggesting CAB or BIC may have comparatively stronger inhibition effect on MMPs (Supplementary Table S2).
The lower binding energy of DOX compared to INSTIs can be explained by its fit within the catalytic binding site. An overlay of the docked DTG, CAB, BIC and DOX on catalytic domain of MMP-9 and -14 showed that DOX (yellow color) has more solvent exposed area than DTG (magenta color), BIC (light blue color) and CAB (red color) (Figure 1). Further, solvent accessible surface area (SASA) calculations confirmed the higher solvent exposure of DOX compared to any of the INSTI (Supplementary Table S3). The higher SASA values of docking complex indicate that the DOX is interacting at lesser extent with MMP’s structural binding site and has a higher affinity to form bonds with the solvent compared to INSTIs. These data confirmed that INSTIs fit the MMP binding pocket with greater efficiency than DOX.
**FIGURE 1:** *Superior affinity of DTG, CAB, BIC compared to DOX at MMPs catalytic binding site. (A and B) 3D representative images of overlapping molecular docking complexes of DTG, CAB, BIC, and DOX on MMP-9 or -14 catalytic domain containing Zn++ (green ball) are shown in ribbon (gray color) format. The color scheme utilized for drugs is as follows: DTG - Magenta; CAB - Red; BIC - Light blue; and DOX - Yellow.*
## INSTIs-induced inhibition of MMPs activities
To affirm that inhibition of MMPs activity is an INSTI class effect, gelatin zymography, a commonly used assay to study MMPs activity and their inhibitors, was performed. For gelatin zymography, cell culture of THP-1 cells was utilized. Cells were treated with phorbol-12-myristate-13-acetate (PMA) to induce differentiation of THP-1 cells into macrophage like cells and to promote MMPs secretion. Herein, PMA-stimulated THP-1 cells were treated with escalating concentrations (25, 50, 75 or 100 µM) of DTG, CAB or BIC for 24 h in serum-free culture medium. Further, to validate the outcome, DOX was utilized as an MMP inhibitor control and the same treatment conditions were employed. To determine the proteolytic activity of MMP-2 and -9 (gelatinases), equal amount of protein (3 µg) from cell culture medium was loaded on SDS-PAGE containing gelatin. Gel area digested by both MMPs was visualized using Coomassie blue stain (Figures 2A, C, E). A decrease in activity of MMP-2 and -9 was observed following treatment with all three INSTIs compared to vehicle-treated controls on gelatin zymogram (Figures 2A, C, E). Relative activity of the pro forms of MMP-2 and MMP-9 was significantly decreased in a concentration-dependent manner at each tested concentration after treatment with DTG, CAB or BIC compared to controls (Figures 2B, D, F). Relative activity of the active form of MMP-2 was significantly reduced at all concentrations of DTG. However, for CAB and BIC, relative activity of the active form of MMP-2 was significantly reduced at 75 or 100 µM. Interestingly, there was a significant increase in relative activity of the active form of MMP-2 at 25 µM BIC compared to controls. Further, DOX treatment, showed a significant decrease in relative activity of pro-form of MMP-9 in a concentration dependent manner (Figures 2G, H). However, variable inhibition of MMP-2 was observed after DOX treatment (Figures 2G, H). Relative activity of the pro form of MMP-2 was significantly increased at 25 µM DOX, but significantly decreased in a concentration-dependent manner at 75 and 100 µM concentrations. Relative activity of the active form of MMP-2 was significantly increased at all treatment concentrations of DOX. When comparing DOX and INSTIs at the same treatment concentration, DTG, CAB, and BIC showed higher MMP inhibition compared to DOX (Supplementary Figure S1A–H). Overall, gelatin zymography results confirmed that inhibition of MMPs activities is an INSTI class effect. These results demonstrated that INSTIs inhibit MMP at a higher degree than the known broad-spectrum MMP inhibitor, DOX.
**FIGURE 2:** *Inhibition of MMPs by INSTIs. (A, C, E, and G) Gelatin zymogram. Activity of MMP-2 and MMP-9 was evaluated in serum-free medium of THP-1 cells following treatment with DTG, CAB, BIC or DOX (25, 50, 75 or 100 µM). Vehicle treated cells were used as controls. (B, D, F, and H) Relative activity of MMP-9 or -2 was measured following treatment with DTG, CAB, BIC or DOX. A one-way ANOVA followed by Dunnett’s test was used to compare activity of individual MMP between each treatment concentration of individual drug and respective control (*p < 0.05, **p < 0.01, ***p < 0.001, ***p < 0.0001). Data are expressed as the mean ± SEM, N = 3 biological replicates. Experiments were repeated independently three times with equivalent results.*
## Discussion
The risk of pre- or post-natal neurodevelopmental deficits due to gestational exposure to ARVs remains possible (Hill et al., 2018; Cassidy et al., 2019; Zash et al., 2019; Crowell et al., 2020; Williams et al., 2020). Works outlined in this report provide unique insights into the underlying mechanisms linked to such adverse events. Recently, clinical and pre-clinical studies reported a potential association between DTG usage at the time of conception and NTDs (Hill et al., 2018; Raesima et al., 2019; Zash et al., 2019; Kreitchmann et al., 2021) and postnatal neurological abnormalities(Crowell et al., 2020). Due to mass usage of DTG-based regimens worldwide, reports highlighted the need to find an underlying mechanism of potential DTG-related adverse neurodevelopmental outcomes. With the introduction of new potent ARVs from the INSTI class to treatment regimens, it is essential to establish if such mechanism can be linked to other ARVs from the INSTI class. Herein, we show that INSTIs including DTG, CAB, and BIC possess chemical abilities to interact with Zn++ at the catalytic domain of all twenty-three MMPs observed in humans and thus, can be classified as broad-spectrum MMPs inhibitors. Such secondary mechanism of MMPs inhibition introduces potential for adverse effects, especially during critical periods of fetal brain development.
All ARVs from the INSTI class possess metal-binding pharmacophore, MBP in their chemical structure. This chemical property enables INSTIs to interact with active metal ion (Mg++) sites in the HIV-1 integrase enzyme to block its action of insertion of the viral genome into the host cellular DNA (Smith et al., 2021). With such inherent metal chelating chemical property, INSTIs have potential to interact with other metalloenzymes that are critical for normal cellular functions such as cell proliferation, differentiation, cell signaling, protein cleavage, etc. MMPs are well recognized Zn++ dependent metalloenzymes (Ethell and Ethell, 2007; Page-McCaw et al., 2007; Agrawal et al., 2008; van Hinsbergh and Koolwijk, 2008; Loffek et al., 2011; Fujioka et al., 2012; Reinhard et al., 2015; Rempe et al., 2016; Small and Crawford, 2016; De Stefano and Herrero, 2017; Shinotsuka et al., 2018; Kanda et al., 2019). The active site of these enzymes is highly conserved, and comprised of three histidine residues that are bound to the catalytic zinc (Laronha and Caldeira, 2020). Dysregulation of activities of MMPs through chelation of Zn++ can cause detrimental effects on structural and functional development of the CNS. The chemical property of INSTIs to chelate divalent cations enables them to engage with Zn++ in the catalytic domain of all twenty-three human MMPs. Our comprehensive molecular docking assessments confirmed that inhibition of MMPs activity is an INSTI class effect. Notably, interaction of individual DTG, CAB or BIC with each MMP was variable with different binding energy. Thus, studies evaluating drug-induced inhibitions of individual MMPs under biological conditions is needed in the future to identify the susceptibility of individual MMP enzymes under normal and genetic polymorphism conditions.
The role of MMPs in normal neural development is of critical importance. MMPs expression is at high levels during early CNS development and decreases into adulthood (Vaillant et al., 1999; Ayoub et al., 2005; Ulrich et al., 2005; Larsen et al., 2006; Bednarek et al., 2009; Aujla and Huntley, 2014; Reinhard et al., 2015). Due to their proteolytic activities, MMPs are ubiquitously expressed during neural development and their expression has been majorly studied in hippocampus, cortex and cerebellum (Fujioka et al., 2012; Reinhard et al., 2015; Small and Crawford, 2016; Beroun et al., 2019). The principal function of MMPs is to degrade extracellular matrix components (Lukes et al., 1999). However, it is well recognized that MMPs functions are essential for the regulation of several neurodevelopmental processes including neurogenesis, neurite outgrowth, migration of newly born neurons, myelination, axonal guidance, synaptic plasticity and angiogenesis (Fujioka et al., 2012; Reinhard et al., 2015; Small and Crawford, 2016). Dysregulation of MMPs activities during critical periods of fetal brain development during gestation could significantly affect these processes, resulting in adverse neurodevelopmental outcomes (Fujioka et al., 2012; Reinhard et al., 2015; Small and Crawford, 2016). Notably, previously we observed that DTG inhibits MMPs activities in rodent embryo brain during gestation leading to neuroinflammation and neuronal injury in the CNS of mice pups during postnatal assessments (Bade et al., 2021). This study identified DTG-induced inhibition of MMPs activities as a neurotoxicity biomarker. However, the previous study was proof of concept and mainly focused on DTG, but comprehensive docking assessments for consideration of each of MMP was missing. The current study confirmed that all INSTIs possess abilities to inhibit MMPs activities. Therefore, drug-induced differences in MMP activities or MMP expression levels could serve as a biomarker for INSTI-associated neurodevelopmental impairments. In addition to pregnancy outcomes, INSTIs also have been recognized to be associated with neuropsychiatric adverse events (NPAEs) in adults (Yombi, 2018; Amusan et al., 2020; Senneker and Tseng, 2021) and clinically significant weight-gain, especially in females (NAMSAL ANRS 12313 Study Group et al., 2019; Venter et al., 2019; Bourgi et al., 2020; Caniglia et al., 2020; Sax et al., 2020). Impaired MMPs activity, expression and related cellular pathways have been identified as biomarkers in both disorders (Vandenbroucke and Libert, 2014; Jaoude and Koh, 2016; Shinotsuka et al., 2018; Beroun et al., 2019; Ruiz-Ojeda et al., 2019; Gorwood et al., 2020; Li et al., 2020). Further, due to critical functions of MMPs, their dysregulation has also served as a biomarker for several types of tumors and atherosclerosis (Goncalves et al., 2015; Huang, 2018). Thus, this work provides a potential mechanistic biomarker for neurodevelopmental assessments following in utero exposure to ART regimens with an INSTI component.
Identifying the roles of the MMPs and impact of their independent or broad-spectrum inhibition in physiological or pathological conditions is complex. This has reflected in cessation of clinical trials of more than fifty broad-spectrum MMP inhibitors due to adverse events following prolonged treatment (Vandenbroucke and Libert, 2014). Thus, an understanding of the inhibitory effect of INSTIs against each MMP enzyme is essential to define the mechanism linked to neurodevelopment. The INSTIs utilized for testing in this study DTG, CAB and BIC showed strong binding energy with the catalytic domains of all twenty-three MMPs. Moreover, comparison assessments against DOX, a clinically used broad-spectrum inhibitor of MMPs, confirmed the high potency of DTG, CAB or BIC against MMPs. For example, binding energies (kcal/mol) with the catalytic domain of MMP-9 were −9.430 (DTG), −14.222 (CAB), and −13.415 (BIC) against −7.658 (DOX). These molecular docking assessment differences against MMP-9 were further apparent on the confirmatory gelatin zymography biological tests. Interestingly, CAB and BIC exhibited higher binding energies for all tested MMPs compared to DTG, suggesting that these newer INSTIs may be more potent MMPs inhibitors. Such observations will need biological validations along with determination of half maximal inhibitory concentration (IC50) values of each INSTI against individual enzymes in the future.
Of the twenty-three human MMPs, few MMPs are well characterized for their role during neurodevelopment. MMP-2 and MMP-9 have been studied extensively and are shown to be essential for the neuronal development, migration, axonal guidance and synaptic plasticity (Ethell and Ethell, 2007; Page-McCaw et al., 2007; Agrawal et al., 2008; Fujioka et al., 2012; Reinhard et al., 2015; Small and Crawford, 2016; De Stefano and Herrero, 2017). Widespread expression of MMP-3 has been identified in neurons in the brain and spinal cord of rodents during critical timepoints of axonal outgrowth (Van Hove et al., 2012a). Moreover, MMP-24 is also expressed in neurons in the brain and spinal cord during development, signifying its role in neuronal development, and MMP-2 and -14 are involved in angiogenesis and in establishment and/or maintenance of the blood-brain barrier (BBB) (Girolamo et al., 2004; Lehti et al., 2005; Lehti et al., 2009; Ikonomidou, 2014; Rempe et al., 2016; Kanda et al., 2019). Knock out or knockdown models of the mentioned MMPs have proved that deficiency in these MMPs can affect neurodevelopmental processes (Oh et al., 2004; Luo, 2005; Van Hove et al., 2012b; Kanda et al., 2019). Interestingly, MMPs are expressed abundantly in neural stem cells (Frolichsthal-Schoeller et al., 1999). Inhibition of MMPs activity by synthetic inhibitors was shown to reduce proliferation and differentiation of neural stem cells (Szymczak et al., 2010; Wojcik-Stanaszek et al., 2011). Thus, identifying the impact of INSTI-induced inhibition of MMPs activities on neurodevelopment and unravelling genetic susceptibility increasing the severity of adverse effects will be critical.
Clinical and pre-clinical studies showed high levels of transplacental transfer of INSTI drugs (Schalkwijk et al., 2016; Mulligan et al., 2018; Mandelbrot et al., 2019; Waitt et al., 2019; Bollen et al., 2021; Bukkems et al., 2021). Studies of DTG have found high placental transfer of DTG from mother to fetus with median cord blood to maternal blood drug level ratios from 1.21 up to 1.29 (Schalkwijk et al., 2016; Mulligan et al., 2018; Mandelbrot et al., 2019; Waitt et al., 2019; Bollen et al., 2021). Further, DTG was also found to accumulate in the fetus with noted prolonged elimination of drug from infants after birth (Mulligan et al., 2018; Waitt et al., 2019). Although few studies have addressed placental transfer of CAB and BIC, evidence does suggest that these INSTIs also cross the placental barrier (Pencole et al., 2020; Bukkems et al., 2021; Le et al., 2022). Our previous work investigated pharmacokinetic (PK) and biodistribution (BD) of DTG during pregnancy in mice and confirmed that DTG levels are detectable in brain tissues of embryos following daily oral administration at supratherapeutic dosage (Bade et al., 2021). Our work validated clinical reports of high placental transfer of DTG and was the first to show drug levels in the fetal developing brain during gestation. Such transplacental transfer of DTG indicated that direct exposure of the embryo brain to DTG during critical periods of development could have an adverse impact on neurodevelopment. Therefore, understanding the PK and BD profiles of new INSTIs during pregnancy and their effects on neurodevelopmental processes is needed for better mechanistic assessments.
It is acknowledged that despite the occurrence of birth defects has been a concern, both the United States Department of Health and Human Services (DHHS) and World Health Organization recommend DTG as a preferred first-line ARV during pregnancy (The U.S. Department of Health and Human Services, 2015; World Health Organization (WHO), 2019a). This decision was based on risk benefit ratios offered by DTG as an ARV compared to rate of associated risk. These included fewer mother-to-child HIV-1 transmission and maternal deaths, and cost-effective (Dugdale et al., 2019; Phillips et al., 2020). Moreover, DTG’s high genetic barrier to drug resistance would address the critical problem of rising pretreatment resistance (PDR) to non-nucleoside reverse transcriptase inhibitors (NNRTIs) in RLCs, especially in women (World Health Organization (WHO), 2019a; World Health Organization (WHO), 2019b). Moreover, most updated data from Tsepamo study (Botswana) reported declined rate of birth defects and was comparative between DTG and other ARVs at the time of conception in late breaking abstract at the 24th International AIDS Conference, 2022. Nonetheless, assessment of birth defects in *Botswana is* an ongoing study and recommended guidelines were based on higher benefits offered by DTG. Yet, risk of long-term neurodevelopmental deficits persists. Particularly, there is a research gap of known adverse events reflecting DTG-associated long-term impact on postnatal neurodevelopment. Therefore, with large number of fetuses being exposed to DTG worldwide, continuous research efforts are critical to uncover any adverse effects of DTG exposures on pre- or post-natal neurodevelopment and elucidate underlying mechanism.
Although the current study provides evidence of an INSTI class effect on the inhibition of MMPs, it was limited to laboratory cell-based assessments. Future studies are necessary in order to affirm mechanistic links between altered MMP activities and adverse developmental outcomes following in utero INSTI exposures. Dose dependent effects of each INSTI on MMPs activities at different stages of neurodevelopment during gestation and early postnatal period need to be studied in animal models. This work would need to include detailed BD drug profiles within the fetal CNS and related MMPs activities. Moreover, with a metal chelating motif, INSTIs possess potential to inhibit other metalloenzymes required for fetal brain development such as Zn++ dependent a disintegrin and metalloproteinase (ADAM) family members (Jorissen et al., 2010; Vandenbroucke and Libert, 2014). Whether inhibition of these metalloenzymes, even at minimal extent, in addition to MMPs could augment the developmental adverse events needs consideration. Thus, comprehensive computational modeling against other metalloenzymes along with biological validations are critical in the future. Moreover, development of ultra-long acting nanoformulations of DTG and assessment of these as a safe drug delivery system for neuroprotective outcomes will be the focus our own future work. We hypothesize that neuroprotective effect would be the outcome of lower drug biodistribution in the embryo brain preventing MMPs inhibition. Such lower drug biodistribution in fetal brain while maintaining therapeutic drug levels in maternal blood is expected due to long-acting pharmacological properties of formulations and lower total drug administration compared to daily oral drug administration. For example, the 8-week cumulative dose of daily oral CAB (VOCABRIA) is 1,680 mg. Whereas, a 600 mg bi-monthly single intramuscular injection of LA-CAB (CABENUVA or APRETUDE) results in a 3-fold reduction in drug exposure with equivalent duration of action (US Food and Drug Administration (FDA), 2021; US Food and Drug Adminis tration (FDA), 2021). Importantly, scientific exchange between basic science mechanistic findings and the clinical assessment of INSTI-exposed children will be required in the future to provide cross-validation of scientific findings and rigorous assessments of neurodevelopment. Overall, it is timely to elucidate any potential ARV-induced secondary effects during pregnancy, in order to provide effective care for women and their fetuses. This study confirms that INSTIs are broad-spectrum MMPs inhibitors. As balanced regulation of MMP activities are crucial for neurodevelopment, the enzyme’s inhibition could underlie INSTI-related adverse neurodevelopmental outcomes.
## 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
EF: performed experiments, collected, analyzed and interpreted data sets, co-wrote, edited and reviewed the manuscript; NP: performed molecular docking experiments, collected, analyzed data sets, YL: provided overall project and technical guidance and edited the manuscript; BE: provided overall project and technical guidance and edited the manuscript; HG: provided overall project and technical guidance, edited and reviewed the manuscript; AB: conceived project, devised central hypothesis and the project’s scientific approach, analyzed and interpreted data, co-wrote, edited and reviewed the manuscript and provided funding acquisition. All authors critically evaluated and approved the final manuscript prior to submission.
## Conflict of interest
HG and BE are Co-founders of Exavir Therapeutics, Inc., a biotechnology company focused on the development of long-acting antiretroviral medicines and HIV-1 cure strategies.
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/ftox.2023.1113032/full#supplementary-material
## References
1. Agrawal S. M., Lau L., Yong V. W.. **MMPs in the central nervous system: Where the good guys go bad**. *Semin. Cell Dev. Biol.* (2008) **19** 42-51. DOI: 10.1016/j.semcdb.2007.06.003
2. Amusan P., Power C., Gill M. J., Gomez D., Johnson E., Rubin L. H.. **Lifetime antiretroviral exposure and neurocognitive impairment in HIV**. *J. Neurovirol* (2020) **26** 743-753. DOI: 10.1007/s13365-020-00870-z
3. Aujla P. K., Huntley G. W.. **Early postnatal expression and localization of matrix metalloproteinases-2 and -9 during establishment of rat hippocampal synaptic circuitry**. *J. Comp. Neurol.* (2014) **522** 1249-1263. DOI: 10.1002/cne.23468
4. Ayoub A. E., Cai T. Q., Kaplan R. A., Luo J.. **Developmental expression of matrix metalloproteinases 2 and 9 and their potential role in the histogenesis of the cerebellar cortex**. *J. Comp. Neurol.* (2005) **481** 403-415. DOI: 10.1002/cne.20375
5. Bade A. N. M. J., Liu Y., Edagwa B. J., Gendelman H. E.. **Dolutegravir inhibition of matrix metalloproteinases affects mouse neurodevelopment**. *Mol. Neurobiol.* (2021) **58** 5703-5721. DOI: 10.1007/s12035-021-02508-5
6. Bednarek N., Clement Y., Lelievre V., Olivier P., Loron G., Garnotel R.. **Ontogeny of MMPs and TIMPs in the murine neocortex**. *Pediatr. Res.* (2009) **65** 296-300. DOI: 10.1203/PDR.0b013e3181973aee
7. Beroun A., Mitra S., Michaluk P., Pijet B., Stefaniuk M., Kaczmarek L.. **MMPs in learning and memory and neuropsychiatric disorders**. *Cell Mol. Life Sci.* (2019) **76** 3207-3228. DOI: 10.1007/s00018-019-03180-8
8. Bollen P., Freriksen J., Konopnicki D., Weizsacker K., Hidalgo Tenorio C., Molto J.. **The effect of pregnancy on the pharmacokinetics of total and unbound dolutegravir and its main metabolite in women living with human immunodeficiency virus**. *Clin. Infect. Dis.* (2021) **72** 121-127. DOI: 10.1093/cid/ciaa006
9. Bourgi K., Rebeiro P. F., Turner M., Castilho J. L., Hulgan T., Raffanti S. P.. **Greater weight gain in treatment-naive persons starting dolutegravir-based antiretroviral therapy**. *Clin. Infect. Dis.* (2020) **70** 1267-1274. DOI: 10.1093/cid/ciz407
10. Bukkems V. E., Hidalgo-Tenorio C., Garcia C., van Hulzen A. G. W., Richel O., Burger D. M.. **First pharmacokinetic data of bictegravir in pregnant women living with HIV**. *AIDS* (2021) **35** 2405-2406. DOI: 10.1097/QAD.0000000000003032
11. Cabrera R. M., Souder J. P., Steele J. W., Yeo L., Tukeman G., Gorelick D. A.. **The antagonism of folate receptor by dolutegravir: Developmental toxicity reduction by supplemental folic acid**. *AIDS* (2019) **33** 1967-1976. DOI: 10.1097/QAD.0000000000002289
12. Caniglia E. C., Shapiro R., Diseko M., Wylie B. J., Zera C., Davey S.. **Weight gain during pregnancy among women initiating dolutegravir in Botswana**. *EClinicalMedicine* (2020) **29-30** 100615. DOI: 10.1016/j.eclinm.2020.100615
13. Cassidy A. R., Williams P. L., Leidner J., Mayondi G., Ajibola G., Makhema J.. *Pediatr. Infect. Dis. J.* (2019) **38** 828-834. DOI: 10.1097/INF.0000000000002332
14. Chandiwana N. C., Chersich M., Venter W. F., Akpomiemie G., Hill A., Simmons B.. **Unexpected interactions between dolutegravir and folate: Randomised trial evidence from south Africa**. *AIDS* (2020) **35** 205-211. DOI: 10.1097/QAD.0000000000002741
15. Crowell C. S., Williams P. L., Yildirim C., Van Dyke R. B., Smith R., Chadwick E. G.. **Safety of**. *AIDS* (2020) **34** 1377-1387. DOI: 10.1097/QAD.0000000000002550
16. De Stefano M. E., Herrero M. T.. **The multifaceted role of metalloproteinases in physiological and pathological conditions in embryonic and adult brains**. *Prog. Neurobiol.* (2017) **155** 36-56. DOI: 10.1016/j.pneurobio.2016.08.002
17. Department of Health and Human Services (DHHS) (2022). Panel on antiretroviral guidelines for adults and adolescents. Guidelines for the use of antiretroviral agents in adults and adolescents living with HIV. Available at: https://clinicalinfo.hiv.gov/sites/default/files/guidelines/documents/AdultandAdolescentGL.pdf
.. *Panel on antiretroviral guidelines for adults and adolescents. Guidelines for the use of antiretroviral agents in adults and adolescents living with HIV* (2022)
18. Dorward J., Lessells R., Drain P. K., Naidoo K., de Oliveira T., Pillay Y.. **Dolutegravir for first-line antiretroviral therapy in low-income and middle-income countries: Uncertainties and opportunities for implementation and research**. *Lancet HIV* (2018) **5** e400-e404. DOI: 10.1016/S2352-3018(18)30093-6
19. Dugdale C. M., Ciaranello A. L., Bekker L. G., Stern M. E., Myer L., Wood R.. **Risks and benefits of dolutegravir- and efavirenz-based strategies for South African women with HIV of child-bearing potential: A modeling study**. *Ann. Intern Med.* (2019) **170** 614-625. DOI: 10.7326/M18-3358
20. Ethell I. M., Ethell D. W.. **Matrix metalloproteinases in brain development and remodeling: Synaptic functions and targets**. *J. Neurosci. Res.* (2007) **85** 2813-2823. DOI: 10.1002/jnr.21273
21. Frolichsthal-Schoeller P., Vescovi A. L., Krekoski C. A., Murphy G., Edwards D. R., Forsyth P.. **Expression and modulation of matrix metalloproteinase-2 and tissue inhibitors of metalloproteinases in human embryonic CNS stem cells**. *Neuroreport* (1999) **10** 345-351. DOI: 10.1097/00001756-199902050-00025
22. Fujioka H., Dairyo Y., Yasunaga K., Emoto K.. **Neural functions of matrix metalloproteinases: Plasticity, neurogenesis, and disease**. *Biochem. Res. Int.* (2012) **2012** 789083. DOI: 10.1155/2012/789083
23. Gilmore J. C., Hoque M. T., Dai W., Mohan H., Dunk C., Serghides L.. **Interaction between dolutegravir and folate transporters and receptor in human and rodent placenta**. *EBioMedicine* (2022) **75** 103771. DOI: 10.1016/j.ebiom.2021.103771
24. Girolamo F., Virgintino D., Errede M., Capobianco C., Bernardini N., Bertossi M.. **Involvement of metalloprotease-2 in the development of human brain microvessels**. *Histochem Cell Biol.* (2004) **122** 261-270. DOI: 10.1007/s00418-004-0705-x
25. Goncalves I., Bengtsson E., Colhoun H. M., Shore A. C., Palombo C., Natali A.. **Elevated plasma levels of MMP-12 are associated with atherosclerotic burden and symptomatic cardiovascular disease in subjects with type 2 diabetes**. *Arterioscler. Thromb. Vasc. Biol.* (2015) **35** 1723-1731. DOI: 10.1161/ATVBAHA.115.305631
26. Gorwood J., Bourgeois C., Pourcher V., Pourcher G., Charlotte F., Mantecon M.. **The integrase inhibitors dolutegravir and raltegravir exert pro-adipogenic and profibrotic effects and induce insulin resistance in human/simian adipose tissue and human adipocytes**. *Clin. Infect. Dis.* (2020) **71** e549-e560. DOI: 10.1093/cid/ciaa259
27. Harder E., Damm W., Maple J., Wu C., Reboul M., Xiang J. Y.. **OPLS3: A force field providing broad coverage of drug-like Small molecules and proteins**. *J. Chem. Theory Comput.* (2016) **12** 281-296. DOI: 10.1021/acs.jctc.5b00864
28. Hill A., Clayden P., Thorne C., Christie R., Zash R.. **Safety and pharmacokinetics of dolutegravir in HIV-positive pregnant women: A systematic review**. *J. Virus Erad.* (2018) **4** 66-71. DOI: 10.1016/s2055-6640(20)30247-8
29. Huang H.. **Matrix metalloproteinase-9 (MMP-9) as a cancer biomarker and MMP-9 biosensors: Recent advances**. *Sensors (Basel)* (2018) **18** 3249. DOI: 10.3390/s18103249
30. Ikonomidou C.. **Matrix metalloproteinases and epileptogenesis**. *Mol. Cell Pediatr.* (2014) **1** 6. DOI: 10.1186/s40348-014-0006-y
31. Jaoude J., Koh Y.. **Matrix metalloproteinases in exercise and obesity**. *Vasc. Health Risk Manag.* (2016) **12** 287-295. DOI: 10.2147/VHRM.S103877
32. Jorissen E., Prox J., Bernreuther C., Weber S., Schwanbeck R., Serneels L.. **The disintegrin/metalloproteinase ADAM10 is essential for the establishment of the brain cortex**. *J. Neurosci.* (2010) **30** 4833-4844. DOI: 10.1523/JNEUROSCI.5221-09.2010
33. Kanda H., Shimamura R., Koizumi-Kitajima M., Okano H.. **Degradation of extracellular matrix by matrix metalloproteinase 2 is essential for the establishment of the blood-brain barrier in Drosophila**. *iScience* (2019) **16** 218-229. DOI: 10.1016/j.isci.2019.05.027
34. Kreitchmann R., Oliveira F. R., Sprinz E.. **Two cases of neural tube defects with dolutegravir use at conception in south Brazil**. *Braz J. Infect. Dis.* (2021) **25** 101572. DOI: 10.1016/j.bjid.2021.101572
35. Krieger E., Vriend G.. **YASARA View - molecular graphics for all devices - from smartphones to workstations**. *Bioinformatics* (2014) **30** 2981-2982. DOI: 10.1093/bioinformatics/btu426
36. Laronha H., Caldeira J.. **Structure and function of human matrix metalloproteinases**. *Cells* (2020) **9** 1076. DOI: 10.3390/cells9051076
37. Larsen P. H., DaSilva A. G., Conant K., Yong V. W.. **Myelin formation during development of the CNS is delayed in matrix metalloproteinase-9 and -12 null mice**. *J. Neurosci.* (2006) **26** 2207-2214. DOI: 10.1523/JNEUROSCI.1880-05.2006
38. Le M. P., Ferre V. M., Mazy F., Bourgeois-Moine A., Damond F., Matheron S.. **Bictegravir pharmacokinetics in a late-presenting HIV-1-infected pregnant woman: A case report**. *J. Antimicrob. Chemother.* (2022) **77** 851-853. DOI: 10.1093/jac/dkab424
39. Lehti K., Allen E., Birkedal-Hansen H., Holmbeck K., Miyake Y., Chun T. H.. **An MT1-MMP-PDGF receptor-beta axis regulates mural cell investment of the microvasculature**. *Genes Dev.* (2005) **19** 979-991. DOI: 10.1101/gad.1294605
40. Lehti K., Rose N. F., Valavaara S., Weiss S. J., Keski-Oja J.. **MT1-MMP promotes vascular smooth muscle dedifferentiation through LRP1 processing**. *J. Cell Sci.* (2009) **122** 126-135. DOI: 10.1242/jcs.035279
41. Li X., Zhao Y., Chen C., Yang L., Lee H. H., Wang Z.. **Critical role of matrix metalloproteinase 14 in adipose tissue remodeling during obesity**. *Mol. Cell Biol.* (2020) **40** e00564-19. DOI: 10.1128/MCB.00564-19
42. Loffek S., Schilling O., Franzke C. W.. **Series "matrix metalloproteinases in lung health and disease": Biological role of matrix metalloproteinases: A critical balance**. *Eur. Respir. J.* (2011) **38** 191-208. DOI: 10.1183/09031936.00146510
43. Lukes A., Mun-Bryce S., Lukes M., Rosenberg G. A.. **Extracellular matrix degradation by metalloproteinases and central nervous system diseases**. *Mol. Neurobiol.* (1999) **19** 267-284. DOI: 10.1007/BF02821717
44. Luo J.. **The role of matrix metalloproteinases in the morphogenesis of the cerebellar cortex**. *Cerebellum* (2005) **4** 239-245. DOI: 10.1080/14734220500247646
45. Mandelbrot L., Ceccaldi P. F., Duro D., Le M., Pencole L., Peytavin G.. **Placental transfer and tissue accumulation of dolutegravir in the**. *PLoS One* (2019) **14** e0220323. DOI: 10.1371/journal.pone.0220323
46. Mohan H., Lenis M. G., Laurette E. Y., Tejada O., Sanghvi T., Leung K. Y.. **Dolutegravir in pregnant mice is associated with increased rates of fetal defects at therapeutic but not at supratherapeutic levels**. *EBioMedicine* (2020) **63** 103167. DOI: 10.1016/j.ebiom.2020.103167
47. Mulligan N., Best B. M., Wang J., Capparelli E. V., Stek A., Barr E.. **Dolutegravir pharmacokinetics in pregnant and postpartum women living with HIV**. *AIDS* (2018) **32** 729-737. DOI: 10.1097/QAD.0000000000001755
48. Kouanfack C., Mpoudi-Etame M., Omgba Bassega P., Eymard-Duvernay S., Leroy S., Boyer S.. **Dolutegravir-based or low-dose efavirenz-based regimen for the treatment of HIV-1**. *N. Engl. J. Med.* (2019) **381** 816-826. DOI: 10.1056/nejmoa1904340
49. Oh J., Takahashi R., Adachi E., Kondo S., Kuratomi S., Noma A.. **Mutations in two matrix metalloproteinase genes, MMP-2 and MT1-MMP, are synthetic lethal in mice**. *Oncogene* (2004) **23** 5041-5048. DOI: 10.1038/sj.onc.1207688
50. Page-McCaw A., Ewald A. J., Werb Z.. **Matrix metalloproteinases and the regulation of tissue remodelling**. *Nat. Rev. Mol. Cell Biol.* (2007) **8** 221-233. DOI: 10.1038/nrm2125
51. Pencole L., Le M. P., Bouchet-Crivat F., Duro D., Peytavin G., Mandelbrot L.. **Placental transfer of the integrase strand inhibitors cabotegravir and bictegravir in the**. *AIDS* (2020) **34** 2145-2149. DOI: 10.1097/QAD.0000000000002637
52. Peters H., Francis K., Sconza R., Horn A., Peckham C. S., Tookey P. A.. **UK mother-to-child HIV transmission rates continue to decline: 2012-2014**. *Clin. Infect. Dis.* (2017) **64** 527-528. DOI: 10.1093/cid/ciw791
53. Phillips A. N., Bansi-Matharu L., Venter F., Havlir D., Pozniak A., Kuritzkes D. R.. **Updated assessment of risks and benefits of dolutegravir versus efavirenz in new antiretroviral treatment initiators in sub-saharan africa: Modelling to inform treatment guidelines**. *Lancet HIV* (2020) **7** e193-e200. DOI: 10.1016/S2352-3018(19)30400-X
54. Raesima M. M., Ogbuabo C. M., Thomas V., Forhan S. E., Gokatweng G., Dintwa E.. **Dolutegravir use at conception - additional surveillance data from Botswana**. *N. Engl. J. Med.* (2019) **381** 885-887. DOI: 10.1056/NEJMc1908155
55. Ramokolo V., Goga A. E., Slogrove A. L., Powis K. M.. **Unmasking the vulnerabilities of uninfected children exposed to HIV**. *BMJ* (2019) **366** l4479. DOI: 10.1136/bmj.l4479
56. Rasi V., Peters H., Sconza R., Francis K., Bukasa L., Thorne C.. **Trends in antiretroviral use in pregnancy in the UK and Ireland**. *HIV Med.* (2022) **23** 397-405. DOI: 10.1111/hiv.13243
57. Reinhard S. M., Razak K., Ethell I. M.. **A delicate balance: Role of MMP-9 in brain development and pathophysiology of neurodevelopmental disorders**. *Front. Cell Neurosci.* (2015) **9** 280. DOI: 10.3389/fncel.2015.00280
58. Rempe R. G., Hartz A. M. S., Bauer B.. **Matrix metalloproteinases in the brain and blood-brain barrier: Versatile breakers and makers**. *J. Cereb. Blood Flow. Metab.* (2016) **36** 1481-1507. DOI: 10.1177/0271678X16655551
59. Ruiz-Ojeda F. J., Mendez-Gutierrez A., Aguilera C. M., Plaza-Diaz J.. **Extracellular matrix remodeling of adipose tissue in obesity and metabolic diseases**. *Int. J. Mol. Sci.* (2019) **20** 4888. DOI: 10.3390/ijms20194888
60. Sax P. E., Erlandson K. M., Lake J. E., McComsey G. A., Orkin C., Esser S.. **Weight gain following initiation of antiretroviral therapy: Risk factors in randomized comparative clinical trials**. *Clin. Infect. Dis.* (2020) **71** 1379-1389. DOI: 10.1093/cid/ciz999
61. Schalkwijk S., Greupink R., Colbers A. P., Wouterse A. C., Verweij V. G., van Drongelen J.. **Placental transfer of the HIV integrase inhibitor dolutegravir in an**. *J. Antimicrob. Chemother.* (2016) **71** 480-483. DOI: 10.1093/jac/dkv358
62. Schnoll J. G., Temsamrit B., Zhang D., Song H., Ming G. L., Christian K. M.. **Evaluating neurodevelopmental consequences of perinatal exposure to antiretroviral drugs: Current challenges and new approaches**. *J. Neuroimmune Pharmacol.* (2019) **16** 113-129. DOI: 10.1007/s11481-019-09880-z
63. Senneker T., Tseng A.. **An update on neuropsychiatric adverse effects with second-generation integrase inhibitors and nonnucleoside reverse transcriptase inhibitors**. *Curr. Opin. HIV AIDS* (2021) **16** 309-320. DOI: 10.1097/COH.0000000000000705
64. Shinotsuka N., Yamaguchi Y., Nakazato K., Matsumoto Y., Mochizuki A., Miura M.. **Caspases and matrix metalloproteases facilitate collective behavior of non-neural ectoderm after hindbrain neuropore closure**. *BMC Dev. Biol.* (2018) **18** 17. DOI: 10.1186/s12861-018-0175-3
65. Small C. D., Crawford B. D.. **Matrix metalloproteinases in neural development: A phylogenetically diverse perspective**. *Neural Regen. Res.* (2016) **11** 357-362. DOI: 10.4103/1673-5374.179030
66. Smith S. J., Zhao X. Z., Passos D. O., Lyumkis D., Burke T. R., Hughes S. H.. **Integrase strand transfer inhibitors are effective anti-HIV drugs**. *Viruses* (2021) **13** 205. DOI: 10.3390/v13020205
67. Szymczak P., Wojcik-Stanaszek L., Sypecka J., Sokolowska A., Zalewska T.. **Effect of matrix metalloproteinases inhibition on the proliferation and differentiation of HUCB-NSCs cultured in the presence of adhesive substrates**. *Acta Neurobiol. Exp. (Wars)* (2010) **70** 325-336. PMID: 21196941
68. The Centers for Disease Control and Prevention (CDC) (2018). HIV and pregnant women, infants, and children. Available at: https://www.cdc.gov/hiv/group/gender/pregnantwomen/index.html
.. *HIV and pregnant women, infants, and children* (2018)
69. The Joint United Nations Programme on HIV/AIDS (UNAIDS) (2021). Global HIV & AIDS statistics fact sheet— 2021.. *Global HIV & AIDS statistics fact sheet— 2021* (2021)
70. The Joint United Nations Programme on HIV/AIDS (UNAIDS) (2021). Start free, stay free, AIDS free final report on 2020 targets. Available at: https://www.unaids.org/en/resources/documents/2021/start-free-stay-free-aids-free-final-report-on-2020-targets
.. *Start free, stay free, AIDS free final report on 2020 targets* (2021)
71. The Lancet H.. **End resistance to dolutegravir roll-out**. *Lancet HIV* (2020) **7** e593. DOI: 10.1016/S2352-3018(20)30231-9
72. The U.S. Department of Health and Human Services (2015). Recommendations for the use of antiretroviral drugs in pregnant women with HIV infection and interventions to reduce perinatal HIV transmission in the United States. Available at: https://clinicalinfo.hiv.gov/sites/default/files/guidelines/documents/Perinatal_GL.pdf
.. *Recommendations for the use of antiretroviral drugs in pregnant women with HIV infection and interventions to reduce perinatal HIV transmission in the United States* (2015)
73. Ulrich R., Gerhauser I., Seeliger F., Baumgartner W., Alldinger S.. **Matrix metalloproteinases and their inhibitors in the developing mouse brain and spinal cord: A reverse transcription quantitative polymerase chain reaction study**. *Dev. Neurosci.* (2005) **27** 408-418. DOI: 10.1159/000088455
74. US Food and Drug Administration (FDA) (2021). FDA approves first injectable treatment for HIV pre-exposure prevention. Available at: https://www.fda.gov/news-events/press-announcements/fda-approves-first-injectable-treatment-hiv-pre-exposure-prevention
.. *FDA approves first injectable treatment for HIV pre-exposure prevention* (2021)
75. US Food and Drug Administration (FDA) (2021). FDA approves cabenuva and vocabria for the treatment of HIV-1 infection. Available at: https://www.fda.gov/drugs/human-immunodeficiency-virus-hiv/fda-approves-cabenuva-and-vocabria-treatment-hiv-1-infection
.. *FDA approves cabenuva and vocabria for the treatment of HIV-1 infection* (2021)
76. Van Hove I., Verslegers M., Buyens T., Delorme N., Lemmens K., Stroobants S.. **An aberrant cerebellar development in mice lacking matrix metalloproteinase-3**. *Mol. Neurobiol.* (2012) **45** 17-29. DOI: 10.1007/s12035-011-8215-z
77. Vaillant C., Didier-Bazes M., Hutter A., Belin M. F., Thomasset N.. **Spatiotemporal expression patterns of metalloproteinases and their inhibitors in the postnatal developing rat cerebellum**. *J. Neurosci.* (1999) **19** 4994-5004. DOI: 10.1523/JNEUROSCI.19-12-04994.1999
78. van Hinsbergh V. W., Koolwijk P.. **Endothelial sprouting and angiogenesis: Matrix metalloproteinases in the lead**. *Cardiovasc Res.* (2008) **78** 203-212. DOI: 10.1093/cvr/cvm102
79. Van Hove I., Lemmens K., Van de Velde S., Verslegers M., Moons L.. **Matrix metalloproteinase-3 in the central nervous system: A look on the bright side**. *J. Neurochem.* (2012) **123** 203-216. DOI: 10.1111/j.1471-4159.2012.07900.x
80. Vandenbroucke R. E., Libert C.. **Is there new hope for therapeutic matrix metalloproteinase inhibition?**. *Nat. Rev. Drug Discov.* (2014) **13** 904-927. DOI: 10.1038/nrd4390
81. Venter W. D. F., Moorhouse M., Sokhela S., Fairlie L., Mashabane N., Masenya M.. **Dolutegravir plus two different prodrugs of tenofovir to treat HIV**. *N. Engl. J. Med.* (2019) **381** 803-815. DOI: 10.1056/NEJMoa1902824
82. Waitt C., Orrell C., Walimbwa S., Singh Y., Kintu K., Simmons B.. **Safety and pharmacokinetics of dolutegravir in pregnant mothers with HIV infection and their neonates: A randomised trial (DolPHIN-1 study)**. *PLoS Med.* (2019) **16** e1002895. DOI: 10.1371/journal.pmed.1002895
83. Williams P. L., Yildirim C., Chadwick E. G., Van Dyke R. B., Smith R., Correia K. F.. **Association of maternal antiretroviral use with microcephaly in children who are HIV-exposed but uninfected (SMARTT): A prospective cohort study**. *Lancet HIV* (2020) **7** e49-e58. DOI: 10.1016/S2352-3018(19)30340-6
84. Wojcik-Stanaszek L., Sypecka J., Szymczak P., Ziemka-Nalecz M., Khrestchatisky M., Rivera S.. **The potential role of metalloproteinases in neurogenesis in the gerbil hippocampus following global forebrain ischemia**. *PLoS One* (2011) **6** e22465. DOI: 10.1371/journal.pone.0022465
85. World Health Organization (WHO) (2016). Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection: Recommendations for a public health approach. Second edition. Available at: https://www.who.int/hiv/pub/arv/chapter4.pdf?ua=1
.. *Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection: Recommendations for a public health approach* (2016)
86. World Health Organization (WHO) (2018). Dolutegravir (DTG) and the fixed dose combination (FDC) of tenofovir/lamivudine/dolutegravir (TLD): Briefing note. Available at: http://www.who.int/hiv/pub/arv/DTG-TLD-arv_briefing_2018.pdf
.. *Dolutegravir (DTG) and the fixed dose combination (FDC) of tenofovir/lamivudine/dolutegravir (TLD): Briefing note* (2018)
87. World Health Organization (WHO) (2019). HIV drug resistance report 2019.. *HIV drug resistance report 2019* (2019)
88. World Health Organization (WHO) (2019). Update of recommendations on first- and second-line antiretroviral regimens. Available at: https://apps.who.int/iris/bitstream/handle/10665/325892/WHO-CDS-HIV-19.15-eng.pdf
.. *Update of recommendations on first- and second-line antiretroviral regimens* (2019)
89. Yombi J. C.. **Dolutegravir neuropsychiatric adverse events: Specific drug effect or class effect**. *AIDS Rev.* (2018) **20** 14-26. PMID: 29628511
90. Zash R., Holmes L., Diseko M., Jacobson D. L., Brummel S., Mayondi G.. **Neural-tube defects and antiretroviral treatment regimens in Botswana**. *N. Engl. J. Med.* (2019) **381** 827-840. DOI: 10.1056/NEJMoa1905230
|
---
title: 'Type 2 diabetes and glycemic traits are not causal factors of delirium: A
two-sample mendelian randomization analysis'
authors:
- Jing Li
- Mingyi Yang
- Pan Luo
- Gang Wang
- Buhuai Dong
- Peng Xu
journal: Frontiers in Genetics
year: 2023
pmcid: PMC9988945
doi: 10.3389/fgene.2023.1087878
license: CC BY 4.0
---
# Type 2 diabetes and glycemic traits are not causal factors of delirium: A two-sample mendelian randomization analysis
## Abstract
This study aims to explore the genetic causal association between type 2 diabetes (T2D) and glycemic traits (fasting glucose [FG], fasting insulin [FI], and glycated hemoglobin [HbA1c]) on delirium using Mendelian randomization (MR). Genome-wide association studies (GWAS) summary data for T2D and glycemic traits were obtained from the IEU OpenGWAS database. GWAS summary data for delirium were obtained from the FinnGen Consortium. All the participants were of European ancestry. In addition, we used T2D, FG, FI, and HbA1c as exposures and delirium as outcomes. A random-effects variance-weighted model (IVW), MR Egger, weighted median, simple mode, and weighted mode were used to perform MR analysis. In addition, MR-IVW and MR-Egger analyses were used to detect heterogeneity in the MR results. Horizontal pleiotropy was detected using MR-Egger regression and MR pleiotropy residual sum and outliers (MR-PRESSO). MR-PRESSO was also used to assess outlier single nucleotide polymorphisms (SNPs). The “leave one out” analysis was used to investigate whether the MR analysis results were influenced by a single SNP and evaluate the robustness of the results. In this study, we conducted a two-sample MR analysis, and there was no evidence of a genetic causal association between T2D and glycemic traits (T2D, FG, FI, and HbA1c) on delirium (all $p \leq 0.05$). The MR-IVW and MR-Egger tests showed no heterogeneity in our MR results (all p values >0.05). In addition, The MR-Egger and MR-PRESSO tests showed no horizontal pleiotropy in our MR results (all $p \leq 0.05$). The MR-PRESSO results also showed that there were no outliers during the MR analysis. In addition, the “leave one out” test did not find that the SNPs included in the analysis could affect the stability of the MR results. Therefore, our study did not support the causal effects of T2D and glycemic traits (FG, FI, and HbA1c) on delirium risk.
## 1 Introduction
Delirium is a distressing acute encephalopathy characterized by the acute onset of deficits in attention, awareness, and cognition that fluctuate in severity over time (Wilson et al., 2020). Delirium is the most common surgical complication among older adults, with an incidence of 15–$25\%$ after major elective surgery and $50\%$ after high-risk procedures such as hip-fracture repair and cardiac surgery (Gibb et al., 2020). Delirium is associated with prolonged length of hospital stay and costs, higher morbidity and mortality, cognitive decline, dementia, and poorer overall outcomes (Gleason et al., 2015; Marcantonio, 2017; Mattison, 2020). With the impact of the recent and ongoing coronavirus disease pandemic, these costs are likely to have increased many times over (Wilson et al., 2020).
Some leading mechanisms postulated to contribute to delirium include neurotransmitters, inflammation, physiological stressors, metabolic derangements, electrolyte disorders, and genetic factors (Inouye et al., 2014; Oldham and Holloway, 2020). The lack of effective treatments calls for the identification of modifiable risk factors and strategies for prevention. In cohort studies, type 2 diabetes (T2D) shows associations with a higher risk for delirium independently of other risk factors (Milisen et al., 2020; Haynes et al., 2021; Liu et al., 2022), but still conflicting findings exist (Bramley et al., 2021). The differences in the study population, methodology, and surgery type may explain these different results. In addition, abnormal glycemic traits, including fasting glucose (FG), fasting insulin (FI), and hemoglobin A1c (HbA1c) levels, have been reported to be associated with delirium (Lin et al., 2021; Song et al., 2022). However, these studies were subject to various methodological limitations, such as a high risk of selection bias, unclear outcome definitions, or retrospective data collection.
No conclusion has yet been reached regarding the relationship between T2D, glycemic traits, and delirium. Mendelian randomization (MR) may help clarify these associations. The MR approach uses genetic information as an instrumental variable to address some of the limitations of observational studies and to estimate causality so that the results are generally independent of environmental confounders and less subject to reverse causation (Holmes et al., 2017; O’Donnell and Sabatine, 2018). In this study, we used large-scale genome-wide association studies (GWAS) data and performed a two-sample MR analysis to investigate the causal effect of T2D and related glycemic traits (FG, FI, and HbA1c) on delirium.
## 2.1 Study design and data sources
GWAS summary data for T2D and glycemic traits were obtained from the IEU OpenGWAS database. The data of T2D included 12,931 patients and 57,196 controls, with a total of 14,277,791 single nucleotide polymorphisms (SNPs). FG data included 200,622 samples and 31,008,728 SNPs. FI data included 151,013 samples and 29,664,438 SNPs. The data for HbA1c levels included 146,806 samples and 30,649,064 SNPs. All participants were of European ancestry, and informed consent was obtained. Each participating cohort underwent study-level quality control (QC), imputation, and association analyses following a shared analysis plan.
Cohorts were genotyped using commercially available genome-wide arrays or the Illumina CardioMetabochip (Metabochip) array. Before imputation, each cohort underwent stringent sample and variant QC to ensure that only high-quality variants were retained in the genotype scaffold for imputation. Sample QC checks included removing samples with a low call rate of less than $95\%$, extreme heterozygosity, sex mismatch with X-chromosome variants, duplicates, first- or second-degree relatives (unless by design), or ancestry outliers. More details on the data can be found in the published study (Chen et al., 2021). GWAS summary data for delirium were obtained from the FinnGen Consortium. We used publicly available data from 1,269 patients with delirium and 209,487 controls of Finnish ancestry. A total of 16,380,452 SNPs were identified. All cases were defined using the code M13 in the International Classification of Diseases, 10th Revision (ICD-10). Detailed information on participants, genotyping, imputation, and QC can be found on the FinnGen website (http://finngen.gitbook.io/documentation/).
## 2.2 Instrumental variable selection
MR analysis of exposure and outcome was performed using strictly censored instrumental variables (IVs). We obtained SNPs that were strongly associated ($p \leq 5$ × 10−8, F > 10) with four exposures (T2D, FG, FI, and HbA1c levels). Because strong linkage disequilibrium (LD) among the selected SNPs may lead to biased results, the clumping process (r 2 < 0.001, clumping distance = 10,000 kb) was carried out to eliminate the LD between the included IVs (Chen et al., 2022). Furthermore, palindromic SNPs with intermediate allele frequencies were excluded to guarantee that the impact of SNPs on exposure corresponded to the same allele as the effect on outcome (Cao et al., 2022). In addition, we applied the PhenoScanner database (http://www.phenoscanner.medschl.cam.ac.uk/phenoscanner) to assess whether the selected SNPs were associated with other traits at the genome-wide significance levels (Shu et al., 2022). When SNPs were not available from GWAS results, proxy SNPs were identified using the online platform LDlink (https://ldlink.nci.nih.gov/).
## 2.3 Statistical analysis
The “TwoSampleMR” package of the R software (version 4.1.2) was used to perform two-sample MR analysis of exposure and outcome. We used a random-effects variance-weighted model (IVW), MR-Egger, weighted median, simple mode, and weighted mode to perform MR analysis (Cao et al., 2022). With random-effects IVW as the main method and weighted median, simple mode, and weighted mode as supplementary methods. We used the I2 index and Cochran’s Q statistic for MR-IVW analyses and Rucker’s Q statistic for MR-Egger analyses to detect heterogeneity of the effects of SNPs related to T2D, FG, FI, and HbA1c on delirium, and $p \leq 0.05$, indicating no heterogeneity (Hemani et al., 2018). We used the MR-Egger method to test for horizontal pleiotropy, and $p \leq 0.05$, indicating no horizontal pleiotropy (Shu et al., 2022). Since MR-Egger may show lower accuracy in some cases, the MR pleiotropy residual sum and outlier (MR-PRESSO) method was also used to assess outlier SNPs and potential horizontal pleiotropy. We also performed a ‘leave one out’ analysis to investigate whether the causal relationship between exposure and outcome was influenced by a single SNP (Lee, 2019).
## 3.1 Instrumental variable selection
After a series of quality controls, we obtained 14 SNPs as IVs for MR analysis of T2D and delirium (Supplementary Table S1). We obtained 59 SNPs as IVs for MR analysis of FG and delirium, among which there were 3 palindrome SNPs (Supplementary Table S2). We obtained 20 SNPs as IVs for the MR analysis of FI and delirium (Supplementary Table S3). We obtained 21 SNPs as IVs for MR analysis of HbA1c levels and delirium (Supplementary Table S4). Among the IVs obtained, none of the SNPs were proxied.
## 3.2 Results of mendelian randomization analysis
The random-effects IVW results suggest that T2D ($$p \leq 0.322$$, odds ratio [OR] ($95\%$ confidence interval [CI]) = 1.080 [0.927–1.258]), FG ($$p \leq 0.400$$, OR [$95\%$ CI] = 0.803 [0.482–1.338]), FI ($$p \leq 0.413$$, OR [$95\%$ CI] = 0.547 [0.129–2.321]), and HbA1c ($$p \leq 0.427$$, OR [$95\%$ CI] = 1.428 [0.592–3.445]) have no genetic causal relationship with delirium. In addition, MR-Egger, weighted median, simple mode, and weighted mode analyses also showed that T2D, FG, FI, and HbA1c had no genetic causal relationship with delirium (Figures 1, 2).
**FIGURE 1:** *MR analysis results of the four exposures (T2D, FG, FI, and HbA1c) and outcome (delirium).* **FIGURE 2:** *Scatter plot of the MR results between exposures and outcome. (A) T2D and delirium. (B) FG and delirium. (C) FI and delirium. (D) HbA1c and delirium.*
The IVW test showed no heterogeneity in the MR analysis results for T2D ($$p \leq 0.717$$), FG ($$p \leq 0.634$$), FI ($$p \leq 0.144$$), and HbA1c ($$p \leq 0.889$$) with delirium. Likewise, the MR-Egger test showed no heterogeneity in the MR analysis results for T2D ($$p \leq 0.750$$), FG ($$p \leq 0.702$$), FI ($$p \leq 0.112$$), and HbA1c ($$p \leq 0.932$$) with delirium. The MR-Egger test showed no horizontal pleiotropy in the MR analysis results for T2D ($$p \leq 0.281$$), FG ($$p \leq 0.099$$), FI ($$p \leq 0.865$$), and HbA1c ($$p \leq 0.205$$) with delirium (Table 1). The results of MR-PRESSO showed no horizontal pleiotropy in the MR analysis of T2D ($$p \leq 0.552$$), FG ($$p \leq 0.665$$), FI ($$p \leq 0.149$$), and HbA1c ($$p \leq 0.902$$) with delirium. The MR-PRESSO results showed no outliers during the MR analysis (Table 1). In addition, the “leave-one-out” analysis showed that the results of our MR analysis were not affected by a single SNP (Figure 3).
## 4 Discussion
By leveraging large-scale GWAS data in MR analysis, we investigated the causal associations between T2D, glycemic traits, and delirium. There was no evidence of positive or negative genetic causality between T2DM, FG, FI, and HbA1c with delirium. The F values of IVs indicate that the variables satisfy the strong correlation assumption of MR analysis, and the instrument bias is weak; therefore, estimates of causal effects are not materially affected. In addition, we used MR-IVW and MR-Egger analyses to detect heterogeneity, and MR-Egger regression and MR-PRESSO were used to detect horizontal pleiotropy. MR-PRESSO was also used to assess outlier SNPs. The results showed no heterogeneity, horizontal pleiotropy, or outliers. Furthermore, the “leave one out” analysis showed that our MR analysis results were not affected by a single SNP, which indicated the reliability of the MR results.
Although the exact mechanism of delirium is not known, it is generally thought to involve abnormal nerve transmission and neuroinflammation (Inouye et al., 2014; Oldham and Holloway, 2020). It has been suggested that cumulative neuroinflammation and neurodegeneration resulting from oxidative stress caused by chronic hyperglycemia in patients with T2D might result in progressive structural abnormalities in the brain that can lead to delirium (Rom et al., 2019). The current observational studies concerning the association between T2D and delirium show conflicting results (Bramley et al., 2021). In addition to differences in population, methods, and type of surgery, these prior studies may have been subject to other confounding factors, did not define diabetes and did not report information on glycemic control. Considering these residual confounders, the strength of the association may be heterogeneous, and the causal association between T2D and delirium has not yet been determined. Hence, conducting a more in-depth study on the correlation between T2D and delirium at the genetic level is necessary.
No previous studies have investigated the association between T2D and delirium using large-scale GWAS data. Recent studies (Garfield et al., 2021; Ware et al., 2021) used MR approaches to estimate the effects of T2D on cognitive outcomes. Of note, a recent analysis using a cumulative genetic risk score for T2D as a valid instrument showed no non-causal association between a history of T2D and cognitive impairment or non-dementia in European ancestry. In addition, using IVW, the genetic liability for T2D was not associated with reaction time or visual memory. In our study, we performed MR analysis with random-effects IVW as the main method and weighted median, simple mode, and weighted mode as supplementary methods, and the results suggest that T2D has no genetic causal relationship with delirium. However, other factors may also play mediating roles. Important factors associated with T2D, such as low vitamin D levels, prior cognitive status, and cardiovascular complications, contribute to the development of delirium (Mattison, 2020; Wilson et al., 2020).
We also considered three glycemic traits (FG, FI, and HbA1c) closely related to T2D and conducted multivariable analyses to avoid bias of confounders caused by these traits. Several observational studies (van Keulen et al., 2018; Windmann et al., 2019; Song et al., 2022) have consistently found that perioperative acute hyperglycemia, independent of diabetes, is associated with delirium. Van Keulen et al. ( van Keulen et al., 2018) was the first to explore the association between diabetes, glucose dysregulation, and their interplay in relation to delirium. They reported that glucose dysregulation was associated with the transition to intensive care unit (ICU) delirium in non-diabetic patients, and diabetes was not associated with an increased risk of ICU delirium. Similarly, Windmann et al. ( Windmann et al., 2019), found that intraoperative hyperglycemia was associated with postoperative delirium independent of age, sex, diabetes, American Society of Anesthesiologists status, duration, and type of surgery; in particular, hyperglycemic non-diabetic patients might be at high risk for postoperative delirium. Furthermore, high preoperative HbA1c levels and poor glycemic control have been reported to increase the risk of postoperative delirium following cardiovascular surgery (Kotfis et al., 2019; Lin et al., 2021).
In clinical practice, hyperglycemia is common in hospitalized patients, and stress hyperglycemia is thought to be caused by inflammation and neurohormonal disturbances that occur during an acute illness. In fact, hyperglycemia not only results from poor control of chronic diabetes, but it can also be due to acute stress. Higher glycemic levels may reflect a more severe inflammatory state and neuroendocrine response, both of which are associated with the development of delirium (van Niekerk et al., 2019; Zhao et al., 2021).
Several studies (Arcambal et al., 2019; Rom et al., 2019; Watt et al., 2020) have reported that abnormally elevated glucose concentrations can promote the release of proinflammatory cytokines, disrupt the blood-brain barrier, and induce neuroinflammation, which may eventually lead to neural network disturbances that can induce delirium. However, it is unclear whether hyperglycemia is to blame or whether vascular risk factors (e.g., hypertension, dyslipidemia, and inflammation) mediate the link between diabetes and poorer brain function. We did not find evidence of a causal relationship between FG, FI, or HbA1c levels and the risk of delirium. Stress hyperglycemia usually results from inflammation and neuroendocrine disruption during acute illness; thus, it is likely that abnormal glycemic traits alone do not explain the increased risk of delirium in patients with T2D and could be a marker of vulnerability with diminished reserve capacity.
This study has several strengths. First, to the best of our knowledge, our study is the first to investigate the causal association between T2D and the related glycemic traits with delirium by leveraging large-scale GWAS. The two-sample MR method can overcome the limitations of some observational studies, such as reverse causality, confounding factors, and various biases. Second, to evaluate the robustness of the MR results, tests for heterogeneity and pleiotropy were conducted as additional means of sensitivity analysis.
However, some limitations of this study cannot be ignored. First, all the participants included in the GWAS were of European ancestry. Consequently, it remains to be determined whether our findings can be generalized to other populations and regions. Second, although we used the IVW and MR-Egger methods to detect and adjust for pleiotropy of genetic variants, there may still be confounding factors between exposure and outcome, such as level of education, personality, and nutrition that may have caued bias in our results. Third, only summary-level GWAS data were available, and the associated effects of sex, age, and specific exposure types on outcomes require further investigation.
## 5 Conclusion
In summary, the association between T2D and delirium is complex and dependent on multiple factors. However, given that the instrumental variable analysis findings are less likely to be biased than that of the observational estimates, our two-sample MR analysis did not suggest significant causal effects of T2D risk, FG, FI, and HbA1c on delirium.
## 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
PX and JL conceptualized and designed the study. MY and PL provided the “TwoSampleMR” package codes in R language and analyzed the data in the study. JL drafted the manuscript. GW and BD gave constructive suggestions when writing the manuscript. All authors have read 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/fgene.2023.1087878/full#supplementary-material
## Abbreviations
FG, fasting glucose; FI, fasting insulin; GWAS, genome-wide association studies; HbA1c, glycated hemoglobin; ICU, intensive care unit; IVs, instrumental variables; IVW, random-effects variance-weighted model; LD, linkage disequilibrium; MR-PRESSO, MR pleiotropy residual sum and outliers; MR, Mendelian randomization; QC, quality control; SNPs, single nucleotide polymorphisms; T2D, type 2 diabetes.
## References
1. Arcambal A., Taile J., Rondeau P., Viranaicken W., Meilhac O., Gonthier M. P.. **Hyperglycemia modulates redox, inflammatory and vasoactive markers through specific signaling pathways in cerebral endothelial cells: Insights on insulin protective action**. *Free Radic. Biol. Med.* (2019) **130** 59-70. DOI: 10.1016/j.freeradbiomed.2018.10.430
2. Bramley P., McArthur K., Blayney A., McCullagh I.. **Risk factors for postoperative delirium: An umbrella review of systematic reviews**. *Int. J. Surg.* (2021) **93** 106063. DOI: 10.1016/j.ijsu.2021.106063
3. Cao Z., Wu Y., Li Q., Li Y., Wu J.. **A causal relationship between childhood obesity and risk of osteoarthritis: Results from a two-sample mendelian randomization analysis**. *Ann. Med.* (2022) **54** 1636-1645. DOI: 10.1080/07853890.2022.2085883
4. Chen J., Spracklen C. N., Marenne G., Varshney A., Corbin L. J., Luan J.. **The trans-ancestral genomic architecture of glycemic traits**. *Nat. Genet.* (2021) **53** 840-860. DOI: 10.1038/s41588-021-00852-9
5. Chen Y., Shen J., Wu Y., Ni M., Deng Y., Sun X.. **Tea consumption and risk of lower respiratory tract infections: A two-sample mendelian randomization study**. *Eur. J. Nutr.* (2022) **30** 1-9. DOI: 10.1007/s00394-022-02994-w
6. Garfield V., Farmaki A., Fatemifar G., Eastwood S. V., Mathur R., Rentsch C. T.. **Relationship between glycemia and cognitive function, structural brain outcomes, and dementia: A mendelian randomization study in the UK biobank**. *Diabetes* (2021) **70** 2313-2321. DOI: 10.2337/db20-0895
7. Gibb K., Seeley A., Quinn T., Siddiqi N., Shenkin S., Rockwood K.. **The consistent burden in published estimates of delirium occurrence in medical inpatients over four decades: A systematic review and meta-analysis study**. *Age ageing* (2020) **49** 352-360. DOI: 10.1093/ageing/afaa040
8. Gleason L. J., Schmitt E. M., Kosar C. M., Tabloski P., Saczynski J. S., Robinson T.. **Effect of delirium and other major complications on outcomes after elective surgery in older adults**. *JAMA Surg.* (2015) **150** 1134-1140. DOI: 10.1001/jamasurg.2015.2606
9. Haynes M. S., Alder K. D., Toombs C., Amakiri I. C., Rubin L. E., Grauer J. N.. **Predictors and sequelae of postoperative delirium in a geriatric patient population with hip fracture**. *J. Am. Acad. Orthop. Surg. Glob. Res. Rev.* (2021) **5** e20.00221. DOI: 10.5435/JAAOSGlobal-D-20-00221
10. Hemani G., Zheng J., Elsworth B., Wade K. H., Haberland V., Baird D.. **The MR-Base platform supports systematic causal inference across the human phenome**. *Elife* (2018) **7** e34408. DOI: 10.7554/eLife.34408
11. Holmes M. V., Ala-Korpela M., Smith G. D.. **Mendelian randomization in cardiometabolic disease: Challenges in evaluating causality**. *Nat. Rev. Cardiol.* (2017) **14** 577-590. DOI: 10.1038/nrcardio.2017.78
12. Inouye S. K., Westendorp R. G., Saczynski J. S.. **Delirium in elderly people**. *Lancet* (2014) **383** 911-922. DOI: 10.1016/S0140-6736(13)60688-1
13. Kotfis K., Szylińska A., Listewnik M., Brykczyński M., Ely E. W., Rotter I.. **Diabetes and elevated preoperative HbA1c level as risk factors for postoperative delirium after cardiac surgery: An observational cohort study**. *Neuropsychiatr. Dis. Treat.* (2019) **15** 511-521. DOI: 10.2147/NDT.S196973
14. Lee Y. H.. **Causal association between smoking behavior and the decreased risk of osteoarthritis: A mendelian randomization**. *Z Rheumatol.* (2019) **78** 461-466. DOI: 10.1007/s00393-018-0505-7
15. Lin Y. J., Lin L. Y., Peng Y. C., Zhang H. R., Chen L. W., Huang X. Z.. **Association between glucose variability and postoperative delirium in acute aortic dissection patients: An observational study**. *J. Cardiothorac. Surg.* (2021) **16** 82. DOI: 10.1186/s13019-021-01456-4
16. Liu K., Song Y., Yuan Y., Li Z., Wang X., Zhang W.. **Type 2 diabetes mellitus with tight glucose control and poor pre-injury stair climbing capacity may predict postoperative delirium: A secondary analysis**. *Brain Sci.* (2022) **12** 951. DOI: 10.3390/brainsci12070951
17. Marcantonio E. R.. **Delirium in hospitalized older adults**. *N. Engl. J. Med.* (2017) **377** 1456-1466. DOI: 10.1056/NEJMcp1605501
18. Mattison M. L. P.. **Delirium**. *Ann. Intern Med.* (2020) **173** ITC49-ITC64. DOI: 10.7326/AITC202010060
19. Milisen K., Van Grootven B., Hermans W., Mouton K., Al Tmimi L., Rex S.. **Is preoperative anxiety associated with postoperative delirium in older persons undergoing cardiac surgery? Secondary data analysis of a randomized controlled trial**. *BMC Geriatr.* (2020) **20** 478. DOI: 10.1186/s12877-020-01872-6
20. O’Donnell C. J., Sabatine M. S.. **Opportunities and challenges in Mendelian randomization studies to guide trial design**. *JAMA Cardiol.* (2018) **3** 967. DOI: 10.1001/jamacardio.2018.2863
21. Oldham M. A., Holloway R. G.. **Delirium disorder: Integrating delirium and acute encephalopathy**. *Neurology* (2020) **95** 173-178. DOI: 10.1212/WNL.0000000000009949
22. Rom S., Zuluaga-Ramirez V., Gajghate S., Seliga A., Winfield M., Heldt N. A.. **Hyperglycemia-driven neuroinflammation compromises BBB leading to memory loss in both diabetes mellitus (DM) Type 1 and Type 2 mouse models**. *Mol. Neurobiol.* (2019) **56** 1883-1896. DOI: 10.1007/s12035-018-1195-5
23. Shu M. J., Li J. R., Zhu Y. C., Shen H.. **Migraine and ischemic stroke: A mendelian randomization study**. *Neurol. Ther.* (2022) **11** 237-246. DOI: 10.1007/s40120-021-00310-y
24. Song Q., Dai M., Zhao Y., Lin T., Huang L., Yue J.. **Association between stress hyperglycemia ratio and delirium in older hospitalized patients: A cohort study**. *BMC Geriatr.* (2022) **22** 277. DOI: 10.1186/s12877-022-02935-6
25. van Keulen K., Knol W., Belitser S. V., van der Linden P. D., Heerdink E. R., Egberts T. C. G.. **Diabetes and glucose dysregulation and transition to delirium in ICU patients**. *Crit. Care Med.* (2018) **46** 1444-1449. DOI: 10.1097/CCM.0000000000003285
26. van Niekerk G., Davis T., Patterton H. G., Engelbrecht A. M.. **How does inflammation-induced Hyperglycemia cause mitochondrial dysfunction in immune cells?**. *Bioessays* (2019) **41** e1800260. DOI: 10.1002/bies.201800260
27. Ware E. B., Morataya C., Fu M., Bakulski K. M.. **Type 2 diabetes and cognitive status in the health and retirement study: A mendelian randomization approach**. *Front. Genet.* (2021) **12** 634767. DOI: 10.3389/fgene.2021.634767
28. Watt C., Sanchez-Rangel E., Hwang J. J.. **Glycemic variability and CNS inflammation: Reviewing the connection**. *Nutrients* (2020) **12** 3906. DOI: 10.3390/nu12123906
29. Wilson J. E., Mart M. F., Cunningham C., Shehabi Y., Girard T. D., MacLullich A. M. J.. **Delirium**. *Nat. Rev. Dis. Prim.* (2020) **6** 90. DOI: 10.1038/s41572-020-00223-4
30. Windmann V., Spies C., Knaak C., Wollersheimm T., Piper S. K., Vorderwulbecke G.. **Intraoperative hyperglycemia increases the incidence of postoperative delirium**. *Minerva Anestesiol.* (2019) **85** 1201-1210. DOI: 10.23736/S0375-9393.19.13748-0
31. Zhao M., Wang S., Zuo A., Zhang J., Wen W., Jiang W.. **HIF-1α/JMJD1A signaling regulates inflammation and oxidative stress following hyperglycemia and hypoxia-induced vascular cell injury**. *Cell. Mol. Biol. Lett.* (2021) **26** 40. DOI: 10.1186/s11658-021-00283-8
|
---
title: Requirement for ER-mitochondria Ca2+ transfer, ROS production and mPTP formation
in L-asparaginase-induced apoptosis of acute lymphoblastic leukemia cells
authors:
- Jung Kwon Lee
- Jesusa L. Rosales
- Ki-Young Lee
journal: Frontiers in Cell and Developmental Biology
year: 2023
pmcid: PMC9988955
doi: 10.3389/fcell.2023.1124164
license: CC BY 4.0
---
# Requirement for ER-mitochondria Ca2+ transfer, ROS production and mPTP formation in L-asparaginase-induced apoptosis of acute lymphoblastic leukemia cells
## Abstract
Acute lymphoblastic leukemia (aLL) is a malignant cancer in the blood and bone marrow characterized by rapid expansion of lymphoblasts. It is a common pediatric cancer and the principal basis of cancer death in children. Previously, we reported that L-asparaginase, a key component of acute lymphoblastic leukemia chemotherapy, causes IP3R-mediated ER Ca2+ release, which contributes to a fatal rise in [Ca2+]cyt, eliciting aLL cell apoptosis via upregulation of the Ca2+-regulated caspase pathway (Blood, 133, 2222–2232). However, the cellular events leading to the rise in [Ca2+]cyt following L-asparaginase-induced ER Ca2+ release remain obscure. Here, we show that in acute lymphoblastic leukemia cells, L-asparaginase causes mitochondrial permeability transition pore (mPTP) formation that is dependent on IP3R-mediated ER Ca2+ release. This is substantiated by the lack of L-asparaginase-induced ER Ca2+ release and loss of mitochondrial permeability transition pore formation in cells depleted of HAP1, a key component of the functional IP3R/HAP1/Htt ER Ca2+ channel. L-asparaginase induces ER Ca2+ transfer into mitochondria, which evokes an increase in reactive oxygen species (ROS) level. L-asparaginase-induced rise in mitochondrial Ca2+ and reactive oxygen species production cause mitochondrial permeability transition pore formation that then leads to an increase in [Ca2+]cyt. Such rise in [Ca2+]cyt is inhibited by Ruthenium red (RuR), an inhibitor of the mitochondrial calcium uniporter (MCU) that is required for mitochondrial Ca2+ uptake, and cyclosporine A (CsA), an mitochondrial permeability transition pore inhibitor. Blocking ER-mitochondria Ca2+ transfer, mitochondrial ROS production, and/or mitochondrial permeability transition pore formation inhibit L-asparaginase-induced apoptosis. Taken together, these findings fill in the gaps in our understanding of the Ca2+-mediated mechanisms behind L-asparaginase-induced apoptosis in acute lymphoblastic leukemia cells.
## Introduction
Acute lymphoblastic leukemia (aLL) is a devastating cancer of immature lymphocytes. It largely afflicts children, representing more than a quarter of all childhood cancers, and causing most of the fatalities from cancer in children (Hunger and Mullighan, 2015). L-asparaginase is a key component of aLL chemotherapy. Regimens that consist of L-asparaginase give rise to greater induction of remission compared to L-asparaginase-free regimens (Egler et al., 2016). L-asparaginase is thought to trigger asparagine insufficiency, causing protein synthesis inhibition and subsequent aLL cell death. However, treatment of L-asparaginase comes with the risk of resistance.
Using genome-wide RNA interference screening, we discovered huntingtin-associated protein 1 (HAP1) (Lee et al., 2019) as a novel biomarker for L-asparaginase resistance in aLL cells. Loss of HAP1 expression in aLL patient primary leukemic cells corresponds to L-asparaginase resistance, indicating that L-asparaginase induces aLL cell apoptosis (Kang et al., 2017; Lee et al., 2019) through a novel non-canonical pathway that involves HAP1. HAP1 binds to huntingtin (Htt) and the intracellular inositol 1,4,5- trisphosphate (IP3) receptor (IP3R) Ca2+ channel to form a functional HAP1-Htt-IP3R complex that regulates IP3-stimulated ER Ca2+ release. HAP1 loss inhibits HAP1-Htt-IP3R formation and thus L-asparaginase stimulation of ER Ca2+ release (Lee et al., 2019). Loss of HAP1 also reduces entry of external Ca2+, inhibiting an overwhelming increase in [Ca2+]i, and downregulating the Ca2+-activated calpain 1, Bid, and caspase-$\frac{3}{12}$ apoptotic pathway, which result in L-asparaginase resistance (Lee et al., 2019). These findings indicate that L-asparaginase causes aLL cell apoptosis through perturbation of intracellular Ca2+ homeostasis and subsequent upregulation of the Ca2+-activated calpain 1, Bid, and caspase-$\frac{3}{12}$ apoptotic pathway. The ability of the Ca2+ chelator, BAPTA-AM, to almost completely reverse aLL cell apoptosis establishes an association between an increase in [Ca2+]cyt and L-asparaginase-stimulated apoptosis (Lee et al., 2019). However, the cellular events that lead to a lethal increase in [Ca2+]i following L-asparaginase-stimulated ER Ca2+ release are still unknown.
Mitochondria are cell organelles that regulate Ca2+ homeostasis and apoptosis. The outer mitochondrial membrane (OMM) is easily permeable to Ca2+ while the inner mitochondrial membrane (IMM) consists of the mitochondrial calcium uniporter (MCU) complex that mediates mitochondrial Ca2+ influx (Collins et al., 2001). As MCU has low affinity to Ca2+, increased Ca2+ concentration is required for MCU activity (Paupe and Prudent, 2018). Uptake of mitochondrial Ca2+ via MCU channels is made possible by the proximity between the mitochondria and the ER (Rizzuto et al., 2009; Grimm, 2012), the major intracellular Ca2+ store. A typical mechanism for ER-mitochondria communication is via the mitochondria-associated ER membrane (MAM) (Vance, 2014), the ER-mitochondria interface. MAMs are associated with several proteins such as IP3R Ca2+ channels (Patergnani et al., 2011) and voltage-dependent anion channels (VDACs) (Ma et al., 2017). These channels regulate ER Ca2+ transport to the mitochondria (Patergnani et al., 2011). Once ER Ca2+ is released through IP3R channels, mitochondria Ca2+ uptake occurs (Patergnani et al., 2011) via the OMM VDACs, and the IMM MCU channel (Rizzuto et al., 2009; Shoshan-Barmatz et al., 2017). However, overload of mitochondrial Ca2+ is related to not only increased or sustained formation of the mitochondrial permeability transition pore (mPTP) (Moore, 1971; Duchen, 2000; Finkel et al., 2015) but also the generation of mitochondrial ROS (Ermak and Davies, 2002; Gorlach et al., 2015), which also contributes to mPTP formation (Zorov et al., 2000; NavaneethaKrishnan et al., 2020) that allows ROS release into the cytoplasm (Zorov et al., 2014). The mPTP channel regulates the IMM permeabilization. Although transient mPTP opening serves as a mitochondrial Ca2+ efflux channel under normal conditions (Altschuld et al., 1992; Ichas et al., 1997), sustained mPTP formation triggers swelling of mitochondria and secretion of cytochrome C and other intermembrane space (IMS) proteins, causing caspase-regulated apoptosis (Lemasters et al., 2002; Kinnally et al., 2011).
In the current study, we utilized SEM patient aLL cells expressing or depleted of HAP1 by retroviral transfection, and demonstrate that L-asparaginase-induced aLL cell apoptosis triggered by a lethal rise in [Ca2+]cyt is caused by mPTP formation that results from ER-mitochondria Ca2+ transfer and subsequent ROS production. Thus, our findings define the Ca2+-mediated mechanisms through which L-asparaginase perturbs intracellular Ca2+ homeostasis to cause apoptosis in aLL cells.
## Materials
RPMI 1640 media, fetal bovine serum, penicillin-streptomycin, Mag-Fluo-4 AM, Rhod-2 AM, Fluo-4 AM, Annexin V-FITC staining kit, Image-IT live mitochondria permeability transition pore assay kit, MitoSOX Red, MitoTracker green and DCFDA were from Thermo Fisher Scientific (Burlington, ON, Canada). L-asparaginase (ab73439) was from Abcam (Toronto, ON, Canada). 2,5-di-tert-butylhydroquinone (TBHQ) was from Sigma (Oakville, ON, Canada). Xestospongin-C (XeC), ruthenium red (RuR), and cyclosporine A (CsA) were from Bio-Techne (Oakville, ON, Canada). HAP1 (D-12) and actin (I19) antibodies, and Mito-Tempo were from Santa Cruz Biotech. ( Dallas, TX, United States of America).
## Cell culture
SEM cells were originally derived from a relapsed 5-year-old female patient diagnosed with pre-B aLL (Greil et al., 1994). These cells, which were prepared by high-density culture of blast cells, exhibited continuous growth and survival in vitro (Lee et al., 2019). SEM cells (*) infected with retrovirus carrying an empty pRS vector (*+pRS) or pRS-shHAP1 (*+pRS-shHAP1) were generated as we described previously (Lee et al., 2019). These cells were cultured in RPMI 1640, containing $10\%$ FBS and 100 μg/ml penicillin-streptomycin, at 37°C in $5\%$ CO2.
## mPTP formation
Formation of mPTP was assessed using the Image-IT live mitochondria permeability transition pore assay kit following the manufacturer’s instructions. * +pRS or *+pRS-shHAP1 cells (0.1 × 106) loaded with 1 µM calcein-AM then pre-treated with 3 μM RuR or 1 μM CsA for 10 min were stimulated with 100 mIU L-asparaginase for 30 min then treated with 1 mM CoCl2 for 15 min. Treatments were performed at 37°C. Cells were rinsed in HBSS, resuspended in ice-cold 1x PBS, and analyzed by flow cytometry using a fluorescein isothiocyanate filter (530 nm).
## Western blot analysis
Cell lysates were resolved by $12.5\%$ SDS-PAGE, transferred to a nitrocellulose membrane, and immunoblotted using the indicated antibodies. Western blot images were captured using a ChemiDoc Imager (Bio-Rad) set at optimal exposure. Chemiluminescence intensity ratios of protein bands of interest vs. actin were determined after densitometry of blots using the National Institutes of Health ImageJ 1.61 software.
## Ca2+ measurement
To measure ER Ca2+ release, *+pRS and *+pRS-shHAP1 cells (0.5 × 106) loaded with 2.5 μM Mag-Fluo-4 AM [in Ca2+-free Krebs-Ringer-Henseleit (KRH) buffer containing 25 mM HEPES, pH 7.4, 125 mM NaCl, 5 mM KCl, 6 mM glucose, and 1.2 mM MgCl2 + 5 μM EGTA] for 30 min then stimulated with 100 mIU L-asparaginase were analyzed using a Shimadzu RF 5301 PC spectrofluorometer (Tokyo, Japan) at λex = 495nm and λem = 530nm.
To measure [Ca2+]mt, *+pRS and *+pRS-shHAP1 (0.5 × 106) loaded with 2 μM of Rhod-2 AM [in Ca2+-free KRH buffer containing 5 μM EGTA] for 1 h then pre-treated with 2 μM XeC or 3 μM RuR were stimulated with 100 mIU L-asparaginase and analyzed using a Shimadzu RF 5301 PC spectrofluorometer at λex = 550nm and λem = 588nm. Peak amplitudes were quantified as ratios of fluorescence (F/F0) after addition of L-asparaginase. F0 represents basal fluorescence or fluorescence before stimulation with L-asparaginase.
To measure [Ca2+]cyt,*+pRS and *+pRS-shHAP1 cells (0.5 × 106) grown on poly-L-ornithine-coated glass coverslips were loaded with 5 μM Fluo-4 AM [in Ca2+-free KRH buffer] for 1 h then pre-treated with 2 μM XeC, 3 μM RuR or 1 μM CsA and stimulated with 100 mIU L-asparaginase. Ca2+ transients were analyzed by single-cell Ca2+ imaging using an Olympus X71 inverted microscope (Tokyo, Japan) at λex = 485 nm and λem = 530 nm. Fluorescence intensities were measured in individual cells ($$n = 10$$) every 2 s. Data were analyzed using ImageJ 1.4.1 (NIH, United States of America). The integrated Ca2+ signals (area under the curve0 were calculated at 60 s–240 s following treatment.
## Measurement of Reactive Oxygen Species (ROS)
*+pRS and *+pRS-shHAP1 cells seeded on poly-ornithine coated coverslips and pre-treated with RuR (3 μM), CsA (1 μM) or Mito-Tempo (5 μM) then stimulated with 100 mIU L-asparaginase for 12 h were stained with MitoSOX red (5 μM) and MitoTracker green (200 nM) or DCFDA (5 μM) for 30 min at 37°C. MitoTracker green was used to label mitochondria in live cells. Cell images were acquired using an Olympus 1 × 71 inverted microscope (Tokyo, Japan) at 160 to ×360 magnification. Fluorescence intensity of captured images (from a field with at least 200 cells) were measured using the ImageJ software. Values from cells stimulated with L-asparaginase alone were normalized to 1.
## Apoptosis
*+pRS and *+pRS-shHAP1 cells (1×104) seeded on 96-well plates coated with 0.2 mg/ml poly-L-ornithine and pre-treated with RuR (3 μM), CsA (1 μM) or Mito-Tempo (5 μM) for 30 min then stimulated with 100 mIU L-asparaginase for 12 h were stained with Hoechst 34580 and FITC-Annexin V. FITC-positive apoptotic cells were counted 12 h post-treatment at ×10 magnification using a I×71 Olympus inverted microscope attached to a 37°C incubator with $5\%$ CO2. The percentage of FITC-positive apoptotic cells was determined from a field of ∼100 Hoechst 34580-stained cells using the Olympus CellSens software (Olympus, Japan).
## Statistical analysis
Student’s t-test (unpaired, two-tailed) was performed at $p \leq 0.05$ for experiments involving two treatment groups. For experiments involving more than two treatment groups, one-way Analysis of Variance (ANOVA) with Tukey Honestly Significantly Different (HSD) post hoc tests were performed.
## Results
L-asparaginase-induced ER Ca2+ release that is mediated by IP3R causes mPTP formation. To investigate if L-asparaginase-induced IP3R-mediated ER Ca2+ release (Lee et al., 2019) causes mPTP formation, SEM cells (*) infected with retrovirus carrying an empty pRS vector (*+pRS) were loaded with calcein-AM and stimulated with L-asparaginase. Cells were then treated with CoCl2 and analyzed by flow cytometry. Cell-permeable calcein-AM dye disperses and gets confined into subcellular organelles such as mitochondria (Petronilli et al., 1998). CoCl2 removes calcein staining in all subcellular compartments except the mitochondria, which are surrounded by a CoCl2-resistant inner mitochondrial membrane (IMM), when mPTP is closed (Petronilli et al., 1998). Thus, CoCl2 treatment permits detection of status of mPTP formation (Petronilli et al., 1998). As shown in Figure 1A, L-asparaginase caused an obvious shift in calcein-stained population of *+pRS cells, indicating clear removal of calcein staining and, therefore, mPTP formation in these cells. To establish a link between L-asparaginase-induced IP3R-mediated ER Ca2+ release and mPTP formation, SEM cells (*) stably depleted of HAP1 by infection with retrovirus carrying pRS-shHAP1 (*+pRS-shHAP1) (Lee et al., 2019) were used. Lack of HAP1 (Figure 1B), a key component of the functional ER Ca2+channel, IP3R/HAP1/Htt ternary complex (Lee et al., 2019), inhibited L-asparaginase-induced ER Ca2+ release (Figure 1C). Treatment with TBHQ, an ER Ca2+ pump inhibitor, caused ER Ca2+ release in both *+pRS and *+pRS-shHAP1 cells, indicating viability of these cells during analysis. In *+pRS-shHAP1 cells where ER Ca2+ release was blocked due to HAP1 loss, L-asparaginase caused a modest shift in calcein-stained population (Figure 1A), indicating high and greater retention of calcein staining in these cells compared to *+pRS cells, and, therefore, closed mPTP. These findings indicate that L-asparaginase-induced IP3R-mediated ER Ca2+ release, which was observed in *+pRS cells, causes mPTP formation.
**FIGURE 1:** *L-asparaginase-induced IP3R-mediated ER Ca2+ release causes mPTP formation. (A) SEM cells (*) infected with retrovirus carrying an empty pRS vector (*+pRS) or pRS-shHAP1 (*+pRS-shHAP1) and loaded with calcein-AM were stimulated with L-asparaginase. Cells were then treated with CoCl2 and subjected to flow cytometry analysis. Data on the left are from one of three independent experiments showing similar results. The chart on the right shows quantitative analysis of the relative calcein fluorescence in *+pRS and *+pRS-shHAP1 cells treated (or untreated) with L-asparaginase. Readings from untreated cells were normalized to 1.0. Values are means ± SEM from the three independent experiments (n = 3). **p < 0.05. (B) Lysates of *+pRS and *+pRS-shHAP1 cells were resolved by SDS-PAGE and immunoblotted for HAP1. Blots (upper panel) shown represent one of three blots with similar results. The actin blot serves as loading control. The bottom panel shows ratios of HAP1 vs. actin levels based on densitometric analysis of blots from the three independent experiments (n = 3) using the NIH ImageJ 1.61 software. Actin values were normalized to 1.0. **p < 0.05. (C) *+pRS and *+pRS-shHAP1 cells (0.5 × 106 cells) loaded with Mag-Fluo-4 AM, an ER Ca2+ probe (Takahashi et al., 1999), then treated with L-asparaginase were analyzed for ER Ca2+ release by spectrofluorometry. Tracings on the upper panel are from one of three independent experiments showing similar results. The chart (bottom panel) shows ER Ca2+ release in *+pRS and *+pRS-shHAP1 cells stimulated with L-asparaginase. Values are means ± SEM from the three independent experiments (n = 3). **p < 0.05.*
L-asparaginase-induced mPTP formation, which is dependent on IP3R-mediated ER Ca2+ release, results from Ca2+ entry into mitochondria. Since mPTP formation is associated with Ca2+ overload in mitochondria (Duchen, 2000; Contreras et al., 2010; Finkel et al., 2015; NavaneethaKrishnan et al., 2020) that could be mediated by the MCU located in the IMM (Moore, 1971), we tested the involvement of MCU in L-asparaginase-induced mPTP formation. To do so, *+pRS and *+pRS-shHAP1 cells pre-treated with Ruthenium Red (RuR), a potent MCU inhibitor (Moore, 1971), and stimulated with L-asparaginase were examined for mPTP formation as described above. As shown in Figures 2A, B, RuR increased calcein staining in *+pRS cells stimulated with L-asparaginase (left panel), indicating inhibition of L-asparaginase-induced mPTP formation. As expected, RuR had no effect on calcein fluorescence intensity in *+pRS-shHAP1 cells stimulated with L-asparaginase (right panel). Cyclosporine A (CsA), an mPTP inhibitor (Crompton et al., 1988), was used as positive control. These findings indicate that in aLL cells, L-asparaginase-induced mPTP formation, which depends on IP3R-mediated ER Ca2+ release, involves the MCU channel that is linked to mitochondrial Ca2+ uptake (Giorgi et al., 2018).
**FIGURE 2:** *The MCU channel is involved in L-asparaginase-induced mPTP formation. *+pRS and *+pRS-shHAP1 cells loaded with calcein-AM then pre-treated with RuR or CsA were stimulated with L-asparaginase then treated with CoCl2. Quantitative analysis of the relative calcein fluorescence in *+pRS and *+pRS-shHAP1 cells was performed by flow cytometry. (A) Data are from one of three independent experiments showing similar results. (B) Readings from untreated cells were normalized to 1.0. Values are means ± SEM from the three independent experiments (n = 3). **p < 0.05.*
We then examined if L-asparaginase induces a rise in mitochondrial Ca2+ level ([Ca2+]mt). To do so, *+pRS and *+pRS-shHAP1 cells loaded with the cell permeable mitochondrial Ca2+ dye, Rhod-2 AM33, were treated with L-asparaginase, and analyzed for mitochondrial Ca2+ increase by spectrofluorometry. As shown in Figure 3, L-asparaginase, which causes ER Ca2+ release in *+pRS cells (Figure 1C), induced mitochondrial Ca2+ increase in these cells, but not in HAP1-depleted *+pRS-shHAP1 cells [where ER Ca2+ release is inhibited (Figure 1C)]. To further establish a link between L-asparaginase-induced IP3R-mediated ER Ca2+ release and increased [Ca2+]mt, *+pRS cells were pre-treated with Xestospongin C (XeC), a potent inhibitor of IP3R (Gafni et al., 1997), prior to L-asparaginase treatment, and spectrofluorometric Ca2+ analysis was performed. XeC dramatically reduced mitochondrial Ca2+ increase in *+pRS cells, indicating that the rise in [Ca2+]mt is associated with L-asparaginase-induced ER Ca2+ release and subsequent transfer to the mitochondria. Pre-treatment with RuR also inhibited L-asparaginase-induced ER-mitochondria Ca2+ transfer, further indicating the involvement of MCU in the process.
**FIGURE 3:** *L-asparaginase causes Ca2+ transfer from the ER to the mitochondria. *+pRS and *+pRSshHAP1 cells loaded with Rhod-2 AM and pre-treated with XeC or RuR were stimulated with L-asparaginase and analyzed for mitochondrial Ca2+ uptake by spectrofluorometry. Data on the left is from one of three independent experiments showing similar results. Peak amplitudes were quantified as ratios of fluorescence (F/F0) after addition of L-asparaginase. F0 represents basal fluorescence or fluorescence before stimulation with L-asparaginase. The chart on the right shows mitochondrial Ca2+ uptake in *+pRS and *+pRS-shHAP1 cells treated as described above. Values are means ± SEM from the three independent experiments (n = 3). **p < 0.05.*
L-asparaginase-induced rise in [Ca2+]mt evokes an increase in reactive oxygen species (ROS) in aLL cells. Since a rise in [Ca2+]mt has been associated with the generation of mitochondrial ROS (Ermak and Davies, 2002; Gorlach et al., 2015), which also contributes to mPTP formation (Zorov et al., 2000; NavaneethaKrishnan et al., 2020) that allows ROS release into the cytoplasm (Zorov et al., 2014), we examined if L-asparaginase-induced rise in [Ca2+]mt upregulates mitochondrial and cytosolic ROS levels in aLL cells. To determine mitochondrial superoxide anion levels, *+pRS and *+pRS-shHAP1 cells pre-treated with RuR, then stimulated with L-asparaginase were stained with MitoSOX and examined by microscopy. As shown in Figure 4, L-asparaginase induced a rise in mitochondrial ROS level in *+pRS cells, which was inhibited by pre-treatment with RuR and more so by HAP1 loss in *+pRS-shHAP1 cells. We then examined cytosolic ROS levels in *+pRS and *+pRS-shHAP1 cells pre-treated with RuR or CsA or Mito-Tempo, a mitochondrial ROS scavenger (NavaneethaKrishnan et al., 2018), then stimulated with L-asparaginase. Cells were stained with 2′,7′-dichlorofluorescin diacetate (DCFDA) and examined by microscopy. Figure 5 shows that L-asparaginase induced a rise in cytosolic ROS level in *+pRS cells, which was inhibited by pre-treatment with RuR, CsA or Mito-Tempo and more so by HAP1 loss in *+pRS-shHAP1 cells. These findings indicate that L-asparaginase-induced rise in [Ca2+]mt is accompanied by mitochondrial and cytosolic ROS increases in aLL cells.
**FIGURE 4:** *L-asparaginase induces an increase in mitochondrial ROS production. *+pRS and *+pRS-shHAP1 cells pre-treated with RuR then stimulated with L-asparaginase were stained with MitoSOX red and MitoTracker green, and examined by microscopy. Cell images were acquired using an Olympus 1 × 71 inverted microscope at ×360 magnification. Bar size = 20 µm. The bar graph shows mean fluorescence intensity of captured images measured using the ImageJ software with values from cells treated with L-asparaginase alone normalized to 1. Values are means ± SEM from three independent experiments (n = 3). **p < 0.05.* **FIGURE 5:** *L-asparaginase causes an increase in cytoplasmic ROS level. *+pRS and *+pRS-shHAP1 cells pre-treated with XeC, RuR, CsA or Mito-Tempo then treated with L-asparaginase were stained with DCFDA and examined by microscopy. Cell images were acquired using an Olympus 1 × 71 inverted microscope at ×160 magnification. The bar graph shows mean fluorescence intensity of captured images measured using the ImageJ software with values from cells stimulated with L-asparaginase normalized to 1. Values are means ± SEM from three independent experiments (n = 3). **p < 0.05.*
L-asparaginase-induced ER-mitochondria Ca2+ transfer and subsequent mPTP formation cause a rise in [Ca2+]cyt. We then examined whether a rise in [Ca2+]cyt observed in L-asparaginase-treated aLL cells (Lee et al., 2019) is due to ER-mitochondria Ca2+ transfer and subsequent mPTP formation. To do so, *+pRS cells loaded with Fluo-4 AM were pre-treated with XeC, RuR or CsA then treated with L-asparaginase and analyzed for Ca2+ transients by single-cell Ca2+ imaging. HAP1-depleted *+pRS-shHAP1 cells were again examined to test whether L-asparaginase-induced IP3R-mediated ER Ca2+ release is linked to a rise in [Ca2+]cyt. As shown in Figure 6, L-asparaginase caused an increase in [Ca2+]cyt in *+pRS cells, which was inhibited by pre-treatment with RuR or CsA and more so by XeC, and by HAP1 loss in *+pRS-shHAP1 cells. These findings indicate that IP3R-stimulated ER Ca2+ release and transfer to the mitochondria, and subsequent mPTP formation account for L-asparaginase-induced rise in [Ca2+]cyt.
**FIGURE 6:** *L-asparaginase causes a rise in [Ca2+]cyt. *+pRS and *+pRS-shHAP1 cells loaded with Fluo-4 AM and pre-treated with XeC, RuR or CsA then stimulated with L-asparaginase were analyzed by single-cell Ca2+ imaging. The left panel shows the average Ca2+ tracing from 10 cells measured every 2 s. Data are from one of three independent experiments showing similar results. The chart on the right shows integrated Ca2+ signals (area under the curve from 60 s to 240 s following treatment) in *+pRS and *+pRSshHAP1 cells treated as described above. Values are means ± SEM from the three independent experiments (n = 3). **p < 0.05.*
L-asparaginase-induced apoptosis is inhibited by blocking ER-mitochondria Ca2+ transfer, mPTP formation and/or mitochondrial ROS production. As indicated above, L-asparaginase causes aLL cell apoptosis by triggering IP3R-mediated ER Ca2+ release that results in a lethal rise in [Ca2+]i and upregulation of the Ca2+-activated calpain-1-Bid-caspase-$\frac{3}{12}$ pathway (Lee et al., 2019). Treatment with the Ca2+ chelator, BAPTA-AM, in aLL cells reversed L-asparaginase-induced apoptotic cell death, indicating a link between [Ca2+]cyt increase and apoptosis in aLL cells (Lee et al., 2019). Thus, we sought to determine whether L-asparaginase-induced ER-mitochondria Ca2+ transfer and subsequent ROS production, which cause mPTP opening that leads to increased [Ca2+]cyt, are linked to L-asparaginase-induced apoptosis in aLL cells. To do so, *+pRS cells pre-treated with RuR, CsA or Mito-Tempo then stimulated with L-asparaginase were stained with Hoechst 34580 and FITC-Annexin V. FITC-positive apoptotic cells were counted 12 h post-treatment and the percentage of apoptotic cells was determined. As shown in Figure 7, L-asparaginase-induced aLL cell apoptosis was inhibited by RuR, CsA or Mito-Tempo in *+pRS cells (left panel) but not in *+pRS-shHAP1 cells (right panel), which showed no ER Ca2+ release upon stimulation with L-asparaginase (Figure 1C). Altogether, our findings indicate that L-asparaginase-induced aLL cell apoptosis caused by an IP3R-mediated ER Ca2+ release and subsequent lethal rise in [Ca2+]cyt (Lee et al., 2019) results from ER-mitochondria Ca2+ transfer and ROS production, which lead to mPTP formation.
**FIGURE 7:** *Blocking ER-mitochondria Ca2+ transfer by RuR, mPTP formation by CsA and/or mitochondrial ROS production by Mito-Tempo inhibit L-asparaginase-induced apoptosis. *+pRS and *+pRS-shHAP1 cells pre-treated with RuR, CsA or Mito-Tempo then stimulated with L-asparaginase were stained with Hoechst 34580 and FITC-Annexin V. FITC-positive apoptotic cells were counted 12 h post-treatment at ×10 magnification using a I×71 Olympus inverted microscope attached to a 37°C incubator with 5% CO2. The percentage of apoptotic cells was determined from a field of ∼100 Hoechst 34580-stained cells using the Olympus CellSens software (Olympus, Japan). Values are means ± SEM from three independent experiments (n = 3). **p < 0.05. N.S. not significant.*
## Discussion
Mitochondria regulate a number of cellular processes, including Ca2+ homeostasis and apoptosis. They communicate dynamically with the ER and store part of the released ER Ca2+. In this study, we demonstrate that L-asparaginase-induced IP3R-mediated ER Ca2+ release in aLL cells causes mPTP formation, which is inhibited in cells lacking HAP1 or upon inhibition of IP3R. These findings establish a link between L-asparaginase-induced IP3R-mediated ER Ca2+ release and mPTP formation. RuR inhibition of L-asparaginase-induced mPTP formation indicates the involvement of the MCU channel that is important for mitochondrial Ca2+ uptake (Giorgi et al., 2018). The fact that L-asparaginase also evokes [Ca2+]mt increase in aLL cells, but not in those depleted of HAP1 or upon inhibition of IP3R suggests a link between L-asparaginase-induced Ca2+ transfer from the ER to the mitochondria and the rise in [Ca2+]mt.
Our finding that L-asparaginase-induced rise in [Ca2+]mt is accompanied by mitochondrial and cytosolic ROS increase in aLL cells is consistent with the notion that a rise in [Ca2+]mt contributes to mitochondrial ROS production (Ermak and Davies, 2002; Gorlach et al., 2015), which stimulates mPTP formation (Zorov et al., 2000; NavaneethaKrishnan et al., 2020) that facilitates the release of ROS into the cytoplasm (Zorov et al., 2014). Although we observed significant RuR inhibition of L-asparaginase-induced rise in both [Ca2+]mt and mitochondrial ROS level in *+pRS cells, the degree of inhibition of mitochondrial ROS level is less than that of [Ca2+]mt. This difference may arise from the different time of measurement: Ca2+ response occurs within seconds and thus was measured immediately; on the other hand, ROS response is slower and was measured 12 h following L-asparaginase treatment. In addition, since ROS is produced in mitochondria and leaks into the cytoplasm through mPTP, there will be dynamic changes in mitochondrial levels of ROS which eventually accumulates in the cytoplasm. This explains the similar degree of RuR inhibition of [Ca2+]mt and cytoplasmic ROS levels. Thus, differences in the extent of RuR inhibition in [Ca2+]mt and mitochondrial ROS can be attributed to differences in method and time of measurement. Non-etheless, it is clear that MCU-mediated inhibition of [Ca2+]mt increase also causes inhibition of ROS increase in both mitochondria and cytoplasm.
As for our view that the rise in [Ca2+]cyt results from L-asparaginase-induced IP3R-mediated ER Ca2+ release and transfer to the mitochondria and subsequent mPTP formation, this is substantiated by the observed inhibition of the process in cells lacking HAP1 or when ER Ca2+ release, MCU, or mPTP is inhibited. Our finding that cells with inhibited MCU show greater [Ca2+]cyt compared to cells with inhibited IP3R or depleted of HAP1 suggests that MCU inhibition causes released ER Ca2+ to bypass the mitochondria and go directly into the cytosol.
Overall, our findings align with previous studies showing that loss of cyclin-dependent kinase 5 (Cdk5) in breast cancer cells or knocking out Cdk5 in primary mouse embryonic fibroblasts (MEFs) is associated with mPTP formation, ROS increase [Ca2+]mt and [Ca2+]cyt increase, and caspase-mediated apoptosis (NavaneethaKrishnan et al., 2018; NavaneethaKrishnan et al., 2020). Our notion that L-asparaginase-induced apoptosis in aLL cells involves ER-mitochondria Ca2+ transfer, ROS production and mPTP formation is validated by inhibition of apoptosis upon inhibition of MCU channel, ROS production and/or mPTP formation.
Since stimulation of µ-opioid receptors (µ-ORs) was shown to trigger Gβγ-mediated rise in [Ca2+]cyt through phospholipase C (PLC) (Yoon et al., 1999; Charles et al., 2003; Celik et al., 2016; Machelska and Celik, 2018; Lee et al., 2021), we propose a model (Figure 8) whereby L-asparaginase causes aLL cell apoptosis via our previously identified Ca2+-mediated calpain-1-Bid-caspase-$\frac{3}{12}$ apoptotic pathway (in purple) (Lee et al., 2019). In this model, we show that L-asparaginase stimulates GPCR (e.g., PAR2 (Peng et al., 2016)) in aLL cells, causing Gβγ-stimulation of PLC, which causes a rise in [Ca2+]cyt through IP3R-mediated release of ER Ca2+ and mitochondrial Ca2+ uptake that leads to ROS production, which together induce mPTP formation.
**FIGURE 8:** *Proposed mechanism by which L-asparaginase induces mPTP-mediated aLL cell apoptosis via IP3R-dependent ER Ca2+ release. Previously, we have shown that L-asparaginase causes aLL cell apoptosis via the Ca2+-mediated calpain-1-Bid-caspase-3/12 apoptotic pathway (in purple) (Lee et al., 2019). In this study, we demonstrate that L-asparaginase-induced ER Ca2+ release triggers a rise in [Ca2+]cyt by causing mitochondrial Ca2+ uptake and subsequent ROS production that leads to mPTP formation and subsequent aLL cell apoptosis. Since activation of opioid receptors has been shown to cause Gβγ-mediated rise in [Ca2+]i
via PLC (Yoon et al., 1999; Charles et al., 2003; Celik et al., 2016; Machelska and Celik, 2018; Lee et al., 2021), we propose that L-asparaginase activation of GPCR (e.g., PAR2 (Peng et al., 2016)) in aLL cells causes Gβγ-mediated stimulation of PLC, which triggers a rise in [Ca2+]cyt through IP3R-mediated ER Ca2+ release, mitochondrial Ca2+ uptake and subsequent ROS production, causing mPTP opening that leads to aLL cell apoptosis.*
In conclusion, our findings indicate that L-asparaginase-induced aLL cell apoptosis requires ER-mitochondria Ca2+ transfer, ROS production and mPTP formation. Thus, results from our studies not only fill in the gaps in our understanding of the Ca2+-mediated mechanisms by which L-asparaginase induces aLL cell apoptosis, but also offer a fresh perspective on targeting ER Ca2+ release, ER-mitochondria Ca2+ transport, ROS and/or mPTP in leukemic cells for aLL therapy.
## 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
JL performed all the experiments, analyzed the data, and drafted the manuscript. K-YL conceived the study and contributed to the analysis and interpretation of data. JR and K-YL provided constructive comments, critically revised the manuscript for important intellectual content, and wrote the final 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.
## References
1. Altschuld R. A., Hohl C. M., Castillo L. C., Garleb A. A., Starling R. C., Brierley G. P.. **Cyclosporin inhibits mitochondrial calcium efflux in isolated adult rat ventricular cardiomyocytes**. *Am. J. Physiology-Heart Circulatory Physiology* (1992) **262** H1699-H1704. DOI: 10.1152/ajpheart.1992.262.6.H1699
2. Celik M. O., Labuz D., Henning K., Busch-Dienstfertig M., Gaveriaux-Ruff C., Kieffer B. L.. **Leukocyte opioid receptors mediate analgesia via Ca(2+)-regulated release of opioid peptides**. *Brain Behav. Immun.* (2016) **57** 227-242. DOI: 10.1016/j.bbi.2016.04.018
3. Charles A. C., Mostovskaya N., Asas K., Evans C. J., Dankovich M. L., Hales T. G.. **Coexpression of delta-opioid receptors with micro receptors in GH3 cells changes the functional response to micro agonists from inhibitory to excitatory**. *Mol. Pharmacol.* (2003) **63** 89-95. DOI: 10.1124/mol.63.1.89
4. Collins T. J., Lipp P., Berridge M. J., Bootman M. D.. **Mitochondrial Ca2+ uptake depends on the spatial and temporal profile of cytosolic Ca2+ signals**. *J. Biol. Chem.* (2001) **276** 26411-26420. DOI: 10.1074/jbc.M101101200
5. Contreras L., Drago I., Zampese E., Pozzan T.. **Mitochondria: The calcium connection**. *Biochimica Biophysica Acta (BBA)-Bioenergetics* (2010) **1797** 607-618. DOI: 10.1016/j.bbabio.2010.05.005
6. Crompton M., Ellinger H., Costi A.. **Inhibition by cyclosporin A of a Ca2+-dependent pore in heart mitochondria activated by inorganic phosphate and oxidative stress**. *Biochem. J.* (1988) **255** 357-360. PMID: 3196322
7. Duchen M. R.. **Mitochondria and calcium: From cell signalling to cell death**. *J. physiology* (2000) **529** 57-68. DOI: 10.1111/j.1469-7793.2000.00057.x
8. Egler R. A., Ahuja S. P., Matloub Y.. **L-asparaginase in the treatment of patients with acute lymphoblastic leukemia**. *J. Pharmacol. Pharmacother.* (2016) **7** 62-71. DOI: 10.4103/0976-500X.184769
9. Ermak G., Davies K. J.. **Calcium and oxidative stress: From cell signaling to cell death**. *Mol. Immunol.* (2002) **38** 713-721. DOI: 10.1016/s0161-5890(01)00108-0
10. Finkel T., Menazza S., Holmström K. M., Parks R. J., Liu J., Sun J.. **The ins and outs of mitochondrial calcium**. *Circulation Res.* (2015) **116** 1810-1819. DOI: 10.1161/CIRCRESAHA.116.305484
11. Gafni J., Munsch J. A., Lam T. H., Catlin M. C., Costa L. G., Molinski T. F.. **Xestospongins: Potent membrane permeable blockers of the inositol 1, 4, 5-trisphosphate receptor**. *Neuron* (1997) **19** 723-733. DOI: 10.1016/s0896-6273(00)80384-0
12. Giorgi C., Marchi S., Pinton P.. **The machineries, regulation and cellular functions of mitochondrial calcium**. *Nat. Rev. Mol. Cell Biol.* (2018) **19** 713-730. DOI: 10.1038/s41580-018-0052-8
13. Gorlach A., Bertram K., Hudecova S., Krizanova O.. **Calcium and ROS: A mutual interplay**. *Redox Biol.* (2015) **6** 260-271. DOI: 10.1016/j.redox.2015.08.010
14. Greil J., Gramatzki M., Burger R., Marschalek R., Peltner M., Trautmann U.. **The acute lymphoblastic leukaemia cell line SEM with t (4; 11) chromosomal rearrangement is biphenotypic and responsive to interleukin-7**. *Br. J. Haematol.* (1994) **86** 275-283. DOI: 10.1111/j.1365-2141.1994.tb04726.x
15. Grimm S.. **The ER–mitochondria interface: The social network of cell death**. *Biochimica Biophysica Acta (BBA)-Molecular Cell Res.* (2012) **1823** 327-334. DOI: 10.1016/j.bbamcr.2011.11.018
16. Hunger S. P., Mullighan C. G.. **Acute lymphoblastic leukemia in children**. *N. Engl. J. Med.* (2015) **373** 1541-1552. DOI: 10.1056/NEJMra1400972
17. Ichas F., Jouaville L. S., Mazat J-P.. **Mitochondria are excitable organelles capable of generating and conveying electrical and calcium signals**. *Cell* (1997) **89** 1145-1153. DOI: 10.1016/s0092-8674(00)80301-3
18. Kang S., Rosales J., Meier-Stephenson V., Kim S., Lee K., Narendran A.. **Genome-wide loss-of-function genetic screening identifies opioid receptor μ 1 as a key regulator of L-asparaginase resistance in pediatric acute lymphoblastic leukemia**. *Oncogene* (2017) **36** 5910-5913. DOI: 10.1038/onc.2017.211
19. Kinnally K. W., Peixoto P. M., Ryu S-Y., Dejean L. M.. **Is mPTP the gatekeeper for necrosis, apoptosis, or both?**. *Biochimica Biophysica Acta (BBA)-Molecular Cell Res.* (2011) **1813** 616-622. DOI: 10.1016/j.bbamcr.2010.09.013
20. Lee J., Rosales J. L., Byun H. G., Lee K. Y. D.. **D,L-Methadone causes leukemic cell apoptosis via an OPRM1-triggered increase in IP3R-mediated ER Ca**. *Sci. Rep.* (2021) **11** 1009. DOI: 10.1038/s41598-020-80520-w
21. Lee J. K., Kang S., Wang X., Rosales J. L., Gao X., Byun H. G.. **HAP1 loss confers l-asparaginase resistance in ALL by downregulating the calpain-1-Bid-caspase-3/12 pathway**. *Blood* (2019) **133** 2222-2232. DOI: 10.1182/blood-2018-12-890236
22. Lemasters J. J., Qian T., He L., Kim J. S., Elmore S. P., Cascio W. E.. **Role of mitochondrial inner membrane permeabilization in necrotic cell death, apoptosis, and autophagy**. *Antioxidants Redox Signal.* (2002) **4** 769-781. DOI: 10.1089/152308602760598918
23. Ma J. H., Shen S., Wang J. J., He Z., Poon A., Li J.. **Comparative proteomic analysis of the mitochondria-associated ER membrane (MAM) in a long-term type 2 diabetic rodent model**. *Sci. Rep.* (2017) **7** 2062-2117. DOI: 10.1038/s41598-017-02213-1
24. Machelska H., Celik M. O.. **Advances in achieving opioid analgesia without side effects**. *Front. Pharmacol.* (2018) **9** 1388. DOI: 10.3389/fphar.2018.01388
25. Moore C. L.. **Specific inhibition of mitochondrial Ca++ transport by ruthenium red**. *Biochem. Biophys. Res. Commun.* (1971) **42** 298-305. DOI: 10.1016/0006-291x(71)90102-1
26. NavaneethaKrishnan S., Rosales J. L., Lee K. Y.. **Loss of Cdk5 in breast cancer cells promotes ROS-mediated cell death through dysregulation of the mitochondrial permeability transition pore**. *Oncogene* (2018) **37** 1788-1804. DOI: 10.1038/s41388-017-0103-1
27. NavaneethaKrishnan S., Rosales J. L., Lee K. Y.. **mPTP opening caused by Cdk5 loss is due to increased mitochondrial Ca(2+) uptake**. *Oncogene* (2020) **39** 2797-2806. DOI: 10.1038/s41388-020-1188-5
28. Patergnani S., Suski J. M., Agnoletto C., Bononi A., Bonora M., De Marchi E.. **Calcium signaling around mitochondria associated membranes (MAMs)**. *Cell Commun. Signal.* (2011) **9** 19-10. DOI: 10.1186/1478-811X-9-19
29. Paupe V., Prudent J.. **New insights into the role of mitochondrial calcium homeostasis in cell migration**. *Biochem. biophysical Res. Commun.* (2018) **500** 75-86. DOI: 10.1016/j.bbrc.2017.05.039
30. Peng S., Gerasimenko J. V., Tsugorka T., Gryshchenko O., Samarasinghe S., Petersen O. H.. **Calcium and adenosine triphosphate control of cellular pathology: Asparaginase-induced pancreatitis elicited via protease-activated receptor 2**. *Philos. Trans. R. Soc. Lond B Biol. Sci.* (2016) **371** 20150423. DOI: 10.1098/rstb.2015.0423
31. Petronilli V., Miotto G., Canton M., Colonna R., Bernardi P., Lisa F. D.. **Imaging the mitochondrial permeability transition pore in intact cells**. *Biofactors* (1998) **8** 263-272. DOI: 10.1002/biof.5520080314
32. Rizzuto R., Marchi S., Bonora M., Aguiari P., Bononi A., De Stefani D.. **Ca2+ transfer from the ER to mitochondria: When, how and why**. *Biochimica Biophysica Acta (BBA)-Bioenergetics* (2009) **1787** 1342-1351. DOI: 10.1016/j.bbabio.2009.03.015
33. Shoshan-Barmatz V., De S., Meir A.. **The mitochondrial voltage-dependent anion channel 1, Ca2+ transport, apoptosis, and their regulation**. *Front. Oncol.* (2017) **7** 60. DOI: 10.3389/fonc.2017.00060
34. Takahashi A., Camacho P., Lechleiter J. D., Herman B.. **Measurement of intracellular calcium**. *Physiol. Rev.* (1999) **79** 1089-1125. DOI: 10.1152/physrev.1999.79.4.1089
35. Vance J. E.. **MAM (mitochondria-associated membranes) in mammalian cells: Lipids and beyond**. *Biochimica Biophysica Acta (BBA)-Molecular Cell Biol. Lipids* (2014) **1841** 595-609. DOI: 10.1016/j.bbalip.2013.11.014
36. Yoon S. H., Lo T. M., Loh H. H., Thayer S. A.. **Delta-opioid-induced liberation of Gbetagamma mobilizes Ca2+ stores in NG108-15 cells**. *Mol. Pharmacol.* (1999) **56** 902-908. DOI: 10.1124/mol.56.5.902
37. Zorov D. B., Filburn C. R., Klotz L. O., Zweier J. L., Sollott S. J.. **Reactive oxygen species (ROS)-induced ROS release: A new phenomenon accompanying induction of the mitochondrial permeability transition in cardiac myocytes**. *J. Exp. Med.* (2000) **192** 1001-1014. DOI: 10.1084/jem.192.7.1001
38. Zorov D. B., Juhaszova M., Sollott S. J.. **Mitochondrial reactive oxygen species (ROS) and ROS-induced ROS release**. *Physiol. Rev.* (2014) **94** 909-950. DOI: 10.1152/physrev.00026.2013
|
---
title: 'Sleep traits, fat accumulation, and glycemic traits in relation to gastroesophageal
reflux disease: A Mendelian randomization study'
authors:
- Xiaoyan Zhao
- Rui Ding
- Chengguo Su
- Rensong Yue
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9988956
doi: 10.3389/fnut.2023.1106769
license: CC BY 4.0
---
# Sleep traits, fat accumulation, and glycemic traits in relation to gastroesophageal reflux disease: A Mendelian randomization study
## Abstract
### Background
Sleep traits, fat accumulation, and glycemic traits are associated with gastroesophageal reflux disease (GERD) in observational studies. However, whether their associations are causal remains unknown. We performed a Mendelian randomization (MR) study to determine these causal relationships.
### Methods
*Independent* genetic variants associated with insomnia, sleep duration, short sleep duration, body fat percentage, visceral adipose tissue (VAT) mass, type 2 diabetes, fasting glucose, and fasting insulin at the genome-wide significance level were selected as instrumental variables. Summary-level data for GERD were derived from a genome-wide association meta-analysis including 78,707 cases and 288,734 controls of European descent. Inverse variance weighted (IVW) was used for the main analysis, with weighted median and MR-Egger as complements to IVW. Sensitivity analyses were performed using Cochran’s Q test, MR-Egger intercept test, and leave-one-out analysis to estimate the stability of the results.
### Results
The MR study showed the causal relationships of genetically predicted insomnia (odds ratio [OR] = 1.306, $95\%$ confidence interval [CI] 1.261 to 1.352; $$p \leq 2.24$$ × 10−51), short sleep duration (OR = 1.304, $95\%$ CI: 1.147 to 1.483, $$p \leq 4.83$$ × 10−5), body fat percentage (OR = 1.793, $95\%$ CI 1.496 to 2.149; $$p \leq 2.68$$ × 10−10), and visceral adipose tissue (OR = 2.090, $95\%$ CI 1.963 to 2.225; $$p \leq 4.42$$ × 10−117) with the risk of GERD. There was little evidence for causal associations between genetically predicted glycemic traits and GERD. In multivariable analyses, genetically predicted VAT accumulation, insomnia, and decreased sleep duration were associated with an increased risk of GERD.
### Conclusion
This study suggests the possible roles of insomnia, short sleep, body fat percentage, and visceral adiposity in the development of GERD.
## Introduction
Gastroesophageal reflux disease (GERD) is defined as a condition that develops when the reflux of stomach contents causes troublesome symptoms and/or complications [1]. GERD is a highly prevalent disease, affecting approximately $13\%$ of the worldwide population and $20\%$ of the adult population in high-income countries [2, 3]. GERD is linked to the increased risk of esophagitis, esophageal strictures, Barrett esophagus, and esophageal adenocarcinoma [2].
Sleep disturbance is commonly associated with GERD [4]. Of 11,685 survey respondents with GERD, $88.9\%$ experienced nighttime GERD symptoms, and $68.3\%$ had sleep difficulties [5]. Also, obesity increased the risk of GERD [6] and a causal association between body mass index (BMI) and GERD has been found in a Mendelian randomization study [7]. However, BMI is a crude indicator of obesity because it does not account for body composition [8]. Obesity is clinically defined based on the measurement of body fat percentage, subcutaneous adipose tissue, and visceral adipose tissue (VAT) [9]. And VAT is considered unique pathogenic fat depots [10]. In addition, diabetic patients suffer various complications, among which esophageal dysfunction are common [11]. A meta-analysis suggested that patients with diabetes are at greater risk of GERD than those without diabetes [12]. There is no evidence to demonstrate a causal relationship between fasting insulin, fasting glucose and GERD. Existing large-scale meta-analyses and observational studies have revealed several possible risk factors for GERD, including sleep disturbances [13, 14], excess adiposity (15–17), and diabetes mellitus [12]. However, unobserved confounding factors, reverse causality, and other biases may affect these results in observational studies. Determining the causal relationship of sleep traits, fat accumulation, and glycemic traits with GERD is very important for the prevention and management of GERD.
Mendelian randomization (MR) is a credible and powerful method to investigate the causal relationship by using genetic variants associated with the specific exposures as instrumental variables (IVs) [18]. The MR design can minimize the biases including residual confounding and reverse causality, because genetic variants are randomly allocated at conception [19]. In this study, we performed a two-sample MR study to examine whether insomnia, sleep duration, short sleep duration, body fat percentage, VAT accumulation, type 2 diabetes mellitus (T2DM), fasting glucose, and fasting insulin were causally associated with the risk of GERD.
## Study design
The study design overview was shown in Figure 1. This study was based on summary-level data on measures of insomnia, sleep duration, short sleep duration, body fat percentage, visceral adipose tissue mass, type 2 diabetes, fasting glucose, fasting insulin, and GERD from published genome-wide association studies (GWASs). To obtain unbiased causal evaluations, the MR study needs to satisfy three assumptions: [1] the genetic variants are robustly associated with the exposure; [2] genetic variants are not associated with potential confounders; and [3] genetic variants affect the risk of the outcome only through the exposure. All analyses in our study were based on publicly available GWAS data, and no additional ethics approval was required.
**Figure 1:** *Study design overview. SNPs, single-nucleotide polymorphisms; MR, Mendelian randomization; MR-PRESSO, MR-pleiotropy residual sum and outlier.*
## Data sources for sleep traits, fat accumulation, and glycemic traits
Summary-level data for insomnia were obtained from a large meta-analysis of GWAS, including 1,331,010 European-ancestry individuals from UK Biobank ($$n = 386$$,533) and 23andMe ($$n = 944$$,477) [20]. Insomnia complaints were measured using questionnaire data; an independent sample (the Netherlands Sleep Register), which provides similar question data and clinical interviews evaluating insomnia, was used to validate the specific questions, making them good proxies for insomnia [20]. Genetic association data for sleep duration were obtained from a recently published GWAS of 446,118 European-ancestry participants [21]. Participants were asked: “How many hours do you sleep in every 24 h?” This genome-wide association study identified 78 loci for self-reported sleep duration. Furthermore, the 78 loci were associated with accelerometer-derived sleep duration ($$n = 85$$,499) [21]. Sleep duration was a continuous variable, and sleep duration <7 h was defined as short sleep, while sleep duration ≥9 h was considered long sleep [21]. The number of single-nucleotide polymorphisms (SNPs) for long sleep duration was insufficient, which was under the risk of weak instrument bias. Therefore, long sleep duration was not included in our study.
IVs for body fat percentage (BF%) were available from a genome-wide association meta-analysis of 100,716 individuals [22]. Of these, summary statistics from 65,831 individuals of European ancestry were included in our study [22]. BF% was measured either with dual-energy X-ray absorptiometry (DEXA) or bioimpedance analysis (BIA), as described in detail before [23]. The summary data for VAT accumulation were derived from a large-scale GWAS including 325,153 participants [24]. This GWAS study constructed two sub-cohorts to calculate VAT mass: a VAT-training dataset with VAT mass estimated by DEXA, to which the prediction models were calibrated; and a VAT-application dataset, in which VAT mass was determined by the calibrated prediction models [24].
We extracted summarized data for T2DM based on 659,361 participants, fasting glucose based on 200,622 participants, and fasting insulin based on 151,013 participants from the MRC IEU Open GWAS database (ID: ebi-a-GCST006867 for T2DM, ebi-a-GCST90002232 for fasting glucose, ebi-a-GCST90002238 for fasting insulin) (25–28).
## Data source for gastroesophageal reflux disease
IVs for GERD were obtained from a recent GWAS meta-analysis of the QSKIN study and UK Biobank study including 78,707 cases and 288,734 controls of European descent [29]. The QSkin cohort [30] is a population-based cohort study to investigate risk factors for skin cancers and other complex diseases in Queensland, Australia. The UK Biobank [31] is a large-scale population-based cohort consisting of over 500,000 participants aged 40–69 years recruited from the United Kingdom. GERD cases were defined based on a combination of self-reported GERD symptoms, international classification of diseases diagnosis, and the use of GERD-related medication [29]. Individuals that did not have any history or occurring conditions of disorders in the upper digestive system were defined as controls. In the multi-trait genetic association analysis, 88 loci associated with GERD were identified [29]. The GWASs details in the MR study were summarized in Table 1.
**Table 1**
| Phenotype | Author, published year | Sample size | No. of cases (Binary trait) | PMID |
| --- | --- | --- | --- | --- |
| Insomnia | Jansen PR et al., 2019 (20) | 1331010 | 288557.0 | 30804565 |
| Sleep duration | Dashti HS et al., 2019 (21) | 446118 | | 30846698 |
| Short sleep duration | Dashti HS et al., 2019 (21) | 411934 | 106192.0 | 30846698 |
| Body fat percentage | Lu YC et al., 2016 (22) | 65831 | | 26833246 |
| Visceral adipose tissue | Karlsson T et al., 2019 (24) | 325153 | | 31501611 |
| Type 2 diabetes | Xue A et al., 2018 (26) | 659316 | 62892.0 | 30054458 |
| Fasting glucose | Chen J et al., 2021 (25) | 200622 | | 34059833 |
| Fasting insulin | Chen J et al., 2021 (25) | 151013 | | 34059833 |
| Gastroesophageal reflux disease | Ong JS et al., 2022 (29) | 367441 | 78707.0 | 34187846 |
## Selection of genetic instruments
To filter eligible instrumental variables, we performed rigorous filtering steps before MR analysis. First, we selected genetic instrumental variables that were significantly associated with the exposures at a genome-wide significance level of $p \leq 5$ × 10−8. Second, the PLINK clumping algorithm (r2 = 0.001 and window size = 10,000 kb) was performed to evaluate the linkage disequilibrium. Third, we selected the SNPs with F statistic >10. IVs with F statistic less than 10 were considered weak genetic instruments [32]. Fourth, ambiguous and palindromic SNPs derived from harmonizing processes were excluded. Finally, SNPs with potential pleiotropy were removed using the MR-pleiotropy residual sum and outlier (MR-PRESSO) analysis. MR-PRESSO analysis could correct horizontal pleiotropy via outlier removal [33].
## Mendelian randomization estimates
Several MR approaches, including inverse variance weighted (IVW), weighted median (WM), and MR-Egger were performed to evaluate the causal association of exposure with outcome. IVW was used as the main statistical method (the random-effects model for the exposure constructed by ≥ 3 SNPs) [7], with WM and MR-Egger methods as supplements to IVW. IVW analysis provides the most precise results when all selected SNPs are valid IVs [34]. The WM method can provide a consistent estimate even when up to $50\%$ of the information comes from invalid IVs [35]. MR-*Egger analysis* can be used for detecting violations of the instrumental variable assumptions, but causal estimates from the MR-Egger may be biased and have inflated type 1 error rates [36]. For quality control, the intercept term derived from the MR-Egger regression was used to evaluate horizontal pleiotropy. Cochran’s Q-test was performed to assess heterogeneity. We also performed leave-one-out analyses by excluding each SNP to evaluate whether causal estimates are driven by a single SNP.
## Multivariable Mendelian randomization
Obesity is a potential confounder affecting the risk of GERD [29, 37]. Therefore, we performed a multivariable IVW MR analysis to confirm the direct effects of sleep traits and fat accumulation after adjusting for body mass index (BMI). Genetic variables on BMI were obtained from the Genetic Investigation of Anthropometric Traits (GIANT) consortium [38]. Detailed information on data source for the multivariable MR study was displayed in Supplementary Table S1.
Considering obesity plays an important role in the pathogenesis of GERD, we then conducted an IVW-based multivariable MR to estimate the effects of insomnia, sleep duration, body fat percentage, and VAT on GERD accounting for the confounding effect from BMI. In multivariable IVW analysis, after adjusting for BMI, the effects of insomnia (OR = 3.225, $95\%$ CI 2.347 to 4.431; $$p \leq 5.07$$ × 10−13), sleep duration (OR = 0.642, $95\%$ CI 0.508 to 0.813; $$p \leq 0.0002$$), and VAT (OR = 1.223, $95\%$ CI 1.025 to 1.459; $$p \leq 0.026$$) on GERD did not alter substantially, respectively. However, the association between body fat percentage and risk of GERD was not significant after adjusting for BMI (Table 3).
**Table 3**
| Exposure | Adjustment | MR method | P value | OR (95% CI) |
| --- | --- | --- | --- | --- |
| Insomnia | BMI | IVW | 5.07e-13 | 3.225 (2.347, 4.431) |
| Sleep duration | BMI | IVW | 0.0002 | 0.642 (0.508, 0.813) |
| Body fat percentage | BMI | IVW | 0.127 | 1.254 (0.938, 1.676) |
| Visceral adipose tissue | BMI | IVW | 0.026 | 1.223 (1.025, 1.459) |
## Statistical analysis
All analyses were performed using the TwoSampleMR package (version 0.5.6) in the R software (version 4.2.1). The MR estimates were shown as odds ratios (OR) with corresponding $95\%$ confidence intervals (CI). To account for multiple testing, we used the Bonferroni correction to adjust the thresholds for significance. Associations with p value < 0.006 ($\frac{0.05}{8}$ exposures) were regarded as significant associations, and associations with p value ≥ 0.006 and < 0.05 were considered suggestive associations.
## Genetic instruments
In total, there were 105 SNPs as instrumental variables for insomnia, 36 SNPs for sleep duration, 14 SNPs for short sleep duration, 7 SNPs for body fat percentage, 136 SNPs for VAT, 100 SNPs for T2DM, 44 SNPs for fasting glucose, and 16 SNPs for fasting insulin. The SNP instruments for the causal relationship between these traits and GERD were detailed in Supplementary Tables S2–S9.
## Causal associations between sleep traits and gastroesophageal reflux disease
Genetically predicted insomnia was significantly associated with an increased risk of GERD in both the IVW analysis (OR = 1.306, $95\%$ CI 1.261 to 1.352; $$p \leq 2.24$$ × 10−51) and the WM analysis (OR = 1.294, $95\%$ CI 1.244 to 1.347; $$p \leq 5.10$$ × 10−37). And a suggestive association was presented in the MR-*Egger analysis* (OR = 1.274, $95\%$ CI 1.074 to 1.151; $$p \leq 0.007$$; Figures 2, 3A). Moreover, IVW and WM analyses demonstrated that short sleep duration was significantly associated with the risk of GERD (OR = 1.304, $95\%$ CI: 1.147 to 1.483, $$p \leq 4.83$$ × 10−5; OR = 1.192, $95\%$ CI: 1.061 to 1.338, $$p \leq 0.003$$), whereas MR-Egger presented a consistent direction but nonsignificant result (Figures 2, 3B). In addition, the results from IVW suggested that there was a suggestive association between sleep duration and GERD (OR = 0.996, $95\%$ CI 0.994 to 0.999; $$p \leq 0.014$$), whereas the causal evaluations from MR-Egger showed an inconsistent direction (Figure 2).
**Figure 2:** *Association of genetically proxied sleep traits, fat accumulation, and glycemic traits with gastroesophageal reflux disease in Mendelian randomization analyses. MR, Mendelian randomization; IVW, inverse-variance weighted; WM, weighted median.* **Figure 3:** *Scatter plot of Mendelian randomization analysis for the associations of insomnia (A), short sleep duration (B), body fat percentage (C), and visceral adipose tissue (D) with the risk of gastroesophageal reflux disease.*
## Causal associations between body fat percentage, visceral adipose tissue, and gastroesophageal reflux disease
A significant positive correlation between genetically proxied body fat percentage and GERD risk was detected using IVW analysis (OR = 1.793, $95\%$ CI 1.496 to 2.149; $$p \leq 2.68$$ × 10−10) and weighted median (OR = 1.715, $95\%$ CI 1.404 to 2.095; $$p \leq 1.25$$ × 10−7), and MR-*Egger analysis* showed a similar causal estimate, although the association was not statistically significant (Figures 2, 3C). In addition, a significant positive causality between VAT accumulation and GERD risk was presented in the IVW analysis (OR = 2.090, $95\%$ CI 1.963 to 2.225; $$p \leq 4.42$$ × 10−117), weighted median (OR = 2.017, $95\%$ CI 1.870 to 2.176; $$p \leq 8.36$$ × 10−74), and MR-egger (OR = 1.739, $95\%$ CI 1.366 to 2.214; $$p \leq 1.49$$ × 10−5; Figures 2, 3D).
## Causal associations between T2DM, fasting glucose, fasting insulin, and gastroesophageal reflux disease
In our primary analysis, genetically predicted T2DM was associated with an increased risk of GERD (OR = 1.039, $95\%$ CI 1.013 to 1.065; $$p \leq 0.003$$). A suggestive association was also shown in the weighted median analysis (OR = 1.031, $95\%$ CI 1.002 to 1.061; $$p \leq 0.038$$). However, the causal inference from MR-Egger showed an inconsistent direction (OR = 0.947, $95\%$ CI 0.881 to 1.017; $$p \leq 0.138$$). Additionally, we did not observe any association of genetically predicted fasting glucose or fasting insulin with GERD risk in three MR analyses (Figure 2).
## Sensitivity analyses
Cochran’s Q test and MR-Egger intercept test were performed to evaluate the robustness of these causal estimates (Table 2). Although heterogeneity was observed for some results in the Cochran’s Q test, it did not invalidate the MR results from the random-effect IVW method, which might balance the pooled heterogeneity [39]. Horizontal pleiotropy was detected in the MR-Egger intercept test of T2DM (P for intercept < 0.05). However, the p-values of other evaluations were all > 0.05, indicating no horizontal pleiotropy bias was introduced into the MR estimates in the context of heterogeneity [39]. Leave-one-out analyses demonstrated that the estimates in our study were not biased by any single SNP.
**Table 2**
| Outcome | Exposure | Cochran Q test | Cochran Q test.1 | MR-Egger | MR-Egger.1 |
| --- | --- | --- | --- | --- | --- |
| Outcome | Exposure | Q value | P | Intercept | P |
| Gastroesophageal reflux disease | Insomnia | 220.834 | 2.06E-10 | 0.001 | 0.773 |
| Gastroesophageal reflux disease | Short sleep duration | 31.344 | 0.003 | 0.002 | 0.851 |
| Gastroesophageal reflux disease | Sleep duration | 80.442 | 3.04E-05 | 0.001 | 0.822 |
| Gastroesophageal reflux disease | Body fat percentage | 13.313 | 0.038 | 0.009 | 0.646 |
| Gastroesophageal reflux disease | Visceral adipose tissue | 258.186 | 9.33E-10 | 0.004 | 0.125 |
| Gastroesophageal reflux disease | Type 2 diabetes | 225.847 | 6.59E-12 | 0.007 | 0.009 |
| Gastroesophageal reflux disease | Fasting glucose | 81.392 | 0.0004 | −0.002 | 0.259 |
| Gastroesophageal reflux disease | Fasting insulin | 27.696 | 0.024 | −0.008 | 0.324 |
## Discussion
In the current study, MR was performed to investigate the causal relationships of insomnia, sleep duration, short sleep duration, body fat percentage, VAT accumulation, T2DM, fasting glucose, and fasting insulin with GERD. We found some evidence that genetically predicted insomnia, short sleep duration, body fat percentage, and VAT accumulation were associated with an increased risk of GERD. Obesity is a known GERD risk factor. Our study showed that VAT accumulation, insomnia, and decreased sleep duration were associated with an increased risk of GERD after adjusting for BMI.
Sleep disturbances were associated with GERD. A total of $61.7\%$ of 33,391 French patients with GERD had regular GERD-related sleep disturbances [40]. A cohort study including 3,813 GERD cases and 15,252 controls indicated that GERD might increase the risk of sleep disorders [41]. A large population-based study suggested that the association between sleep problems and GERD might be bidirectional [42]. Among the many lifestyle factors, poor quality of sleep is a strong risk factor for GERD [13]. Previous studies observed that persistent insufficient and/or short sleep increased the risk of GERD [13, 14]. A prospective study including 2,316 adults showed that insomnia was associated with the risk of GERD [43]. Consistent with previous studies, our MR study confirmed the causal association of insomnia and short sleep with the increased risk of GERD. The study suggests that treatments to improve sleep quality may decrease GERD symptoms in patients with GERD. Poor sleep quality due to sleep fragmentation and sleep deprivation from awakenings is related to GERD symptoms [44]. Several potential mechanisms may contribute to the causal relationship between sleep and GERD. A study in healthy adults has shown a hyperalgesic effect associated with sleep deprivation and an analgesic effect associated with slow-wave sleep recovery [45] and sleep deprivation may cause esophageal hyperalgesia evidenced by the acid perfusion testing, which provided an underlying mechanism for the GERD symptoms in patients with poor sleep quality [46]. In addition, we considered the interpretation of the results in terms of psychological factors. Insomnia is a prevalent mental disorder [20]. Previous studies have shown a bidirectional association between sleep disorders and depression [47, 48]. A study indicated that depressive symptoms and poor sleep quality are associated with the presence of GERD [49]. Psychopathology plays a role in GERD pathogenesis [50]. A study suggested that psychological symptomatology, mood and anxiety disorders are positively associated with GERD symptoms [50]. Another study observed the associations between anxiety, poor sleep quality and GERD [51]. In GERD patients, there was a strong relationship between psychological stress (anxiety and depression) and sleep disturbances [52]. Therefore, sleep may affect GERD through factors such as depression.
Overweight and obesity were correlated with an increased risk of GERD [6, 15]. A meta-analysis including 18,346 patients with GERD demonstrated a positive relationship between increasing BMI and GERD risk [6]. Another meta-analysis showed that obesity was associated with a significant increase in the risk for GERD and its complications (erosive esophagitis, and esophageal adenocarcinoma) [15]. And the risk of these diseases seems to progressively increase with weight gain [15]. Weight loss was dose-dependently correlated with both a reduction of GERD symptoms and an increase of treatment success with antireflux medication [16]. BMI is calculated with weight and height, which does not consider body composition such as body fat mass and VAT mass. Body fat distribution, especially the abdominal adiposity (such as visceral adiposity), is an important factor in the association of obesity with GERD, and it is more strongly associated with GERD than BMI [53]. Our study showed that genetically predicted body fat percentage and VAT mass were positively associated with the risk of GERD. In multivariable IVW analysis, the association between body fat percentage and GERD risk was not significant after adjusting for BMI, suggesting that this association could be affected by BMI. In addition, a positive association between VAT and GERD after adjusting for BMI was observed, which confirmed the robustness of the result. The accumulation of VAT is more harmful than the accumulation of adipose tissue at other locations [54, 55]. Several potential mechanisms may explain the association, including increased intra-abdominal pressures, delayed gastric emptying, low esophageal sphincter abnormalities, and increased frequency of transient sphincter relaxation [15, 37]. And visceral adipose tissue, which is metabolically active, secretes adipokines and inflammatory cytokines that may predispose to GERD and its complications [37].
Gastrointestinal symptoms are common in patients with diabetes. A meta-analysis involving 9,067 cases and 81,968 controls suggested that individuals with diabetes were at greater risk of GERD than those without diabetes (overall OR = 1.61) [12]. Our main analysis showed the causal nature of the positive association between T2DM and GERD, but three MR analyses presented an inconsistent result. In addition, our MR estimates could not provide any association of genetically predicted fasting glucose or fasting insulin with GERD. Inconsistent with the previous studies, the correlation between glycemic traits and GERD could not be determined in our study. But it’s worth noting that diabetic upstream factors (e.g., obesity), duration of diabetes, and diabetic autonomic neuropathy were associated with gastrointestinal symptoms [12, 56, 57]. It has been reported that patients with diabetic complications are more likely to report reflux symptoms, and the quality of diabetic control and use of oral hypoglycemic agents may influence the incidence of GERD [58]. Studies have reported on the association between diabetes and esophageal dysfunction; however, no consensus has been reached. The mechanism for this association needs to be further explored.
Our study has several important strengths. First, MR is an analytic approach using genetic variants as IVs to explore the causal association between exposure and outcome. MR design diminished unobserved confounding and reverse causality that are common in observational studies. Second, several MR methods and sensitivity analyses were performed to ensure the stability of the results. Finally, our study strengthened the requirement for consistent beta direction in all MR analyses [39]. IVW has higher statistical power than other MR analyses, especially MR-Egger [59]. Therefore, it is not surprising that the causal estimates derived from the MR-Egger may have nonsignificant p-values and wider confidence compared to IVW estimates in our study [39].
Our study also has several limitations. First, most participants in the datasets were of European ancestry, which may limit the generalizability of the findings to other ethnic groups. Second, it is difficult to completely exclude pleiotropy in MR analyses. Horizontal pleiotropy affects the stability of MR results, but vertical pleiotropy where exposure acts on outcome through other factors with the same causal pathway is acceptable. Importantly, the MR-Egger regression method detected no evidence of horizontal pleiotropy for important causal evaluations in our study. Third, self-reported sleep data may introduce potential bias into the study. Despite limitations of imprecision in self-report, the GWAS study observed largely consistent effects of the 78 signals for self-reported sleep duration with accelerometer-estimated sleep duration [21]. Finally, GERD cases are defined using various sources (self-reported GERD, medication use, and clinical diagnosis), which may introduce outcome misclassification. However, the GERD genome-wide association study found very high-genetic correlations (rg > 0.9) between the different GERD phenotypes [29], indicating a good validity of GERD data. Thus, the bias caused by GERD data is not a major issue.
## Conclusion
Our study suggested that genetically predicted insomnia, short sleep duration, and visceral adipose tissue accumulation were correlated with an increased risk of GERD. Thus, improving sleep quality and reducing visceral adiposity may be potential intervention targets for preventing GERD.
## 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
XZ, CS, RD, and RY designed the manuscript. XZ and CS are responsible for the statistical analyses and manuscript writing. All authors contributed to the article and approved the submitted version.
## Funding
This project was supported by Natural Science Foundation of Sichuan Province (2022NSFSC0853).
## 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.2023.1106769/full#supplementary-material
## References
1. Patti MG. **An evidence-based approach to the treatment of gastroesophageal reflux disease**. *JAMA Surg* (2016) **151** 73-8. DOI: 10.1001/jamasurg.2015.4233
2. Maret-Ouda J, Markar SR, Lagergren J. **Gastroesophageal reflux disease: a review**. *JAMA* (2020) **324** 2536-47. DOI: 10.1001/jama.2020.21360
3. Richter JE, Rubenstein JH. **Presentation and epidemiology of gastroesophageal reflux disease**. *Gastroenterology* (2018) **154** 267-76. DOI: 10.1053/j.gastro.2017.07.045
4. Oh JH. **Gastroesophageal reflux disease: recent advances and its association with sleep**. *Ann N Y Acad Sci* (2016) **1380** 195-203. DOI: 10.1111/nyas.13143
5. Mody R, Bolge SC, Kannan H, Fass R. **Effects of gastroesophageal reflux disease on sleep and outcomes**. *Clin Gastroenterol Hepatol* (2009) **7** 953-9. DOI: 10.1016/j.cgh.2009.04.005
6. Corley DA, Kubo A. **Body mass index and gastroesophageal reflux disease: a systematic review and meta-analysis**. *Am J Gastroenterol* (2006) **101** 2619-28. DOI: 10.1111/j.1572-0241.2006.00849.x
7. Yuan S, Larsson SC. **Adiposity, diabetes, lifestyle factors and risk of gastroesophageal reflux disease: a Mendelian randomization study**. *Eur J Epidemiol* (2022) **37** 747-54. DOI: 10.1007/s10654-022-00842-z
8. Rothman KJ. **BMI-related errors in the measurement of obesity**. *Int J Obes* (2008) **32** S56-9. DOI: 10.1038/ijo.2008.87
9. Yan B, Yang J, Zhao B, Wu Y, Bai L, Ma X. **Causal effect of visceral adipose tissue accumulation on the human longevity: a mendelian randomization study**. *Front Endocrinol* (2021) **12** 12722187. DOI: 10.3389/fendo.2021.722187
10. Mahabadi AA, Massaro JM, Rosito GA, Levy D, Murabito JM, Wolf PA. **Association of pericardial fat, intrathoracic fat, and visceral abdominal fat with cardiovascular disease burden: the Framingham heart study**. *Eur Heart J* (2009) **30** 850-6. DOI: 10.1093/eurheartj/ehn573
11. Kinekawa F, Kubo F, Matsuda K, Fujita Y, Tomita T, Uchida Y. **Relationship between esophageal dysfunction and neuropathy in diabetic patients**. *Am J Gastroenterol* (2001) **96** 2026-32. DOI: 10.1111/j.1572-0241.2001.03862.x
12. Sun XM, Tan JC, Zhu Y, Lin L. **Association between diabetes mellitus and gastroesophageal reflux disease: a meta-analysis**. *World J Gastroenterol* (2015) **21** 3085-92. DOI: 10.3748/wjg.v21.i10.3085
13. Yamamichi N, Mochizuki S, Asada-Hirayama I, Mikami-Matsuda R, Shimamoto T, Konno-Shimizu M. **Lifestyle factors affecting gastroesophageal reflux disease symptoms: a cross-sectional study of healthy 19864 adults using FSSG scores**. *BMC Med* (2012) **10** 1045. DOI: 10.1186/1741-7015-10-45
14. Emilsson ÖI, Al YH, Theorell-Haglöw J, Ljunggren M, Lindberg E. **Insufficient sleep and new onset of nocturnal gastroesophageal reflux among women: a longitudinal cohort study**. *J Clin Sleep Med* (2022) **18** 1731-7. DOI: 10.5664/jcsm.9928
15. Hampel H, Abraham NS, El-Serag HB. **Meta-analysis: obesity and the risk for gastroesophageal reflux disease and its complications**. *Ann Intern Med* (2005) **143** 199-211. DOI: 10.7326/0003-4819-143-3-200508020-00006
16. Ness-Jensen E, Lindam A, Lagergren J, Hveem K. **Weight loss and reduction in gastroesophageal reflux. A prospective population-based cohort study: the HUNT study**. *Am J Gastroenterol* (2013) **108** 376-82. DOI: 10.1038/ajg.2012.466
17. Ness-Jensen E, Hveem K, El-Serag H, Lagergren J. **Lifestyle intervention in gastroesophageal reflux disease**. *Clin Gastroenterol Hepatol* (2016) **14** 175-182.e3. DOI: 10.1016/j.cgh.2015.04.176
18. Richmond RC, Davey SG. **Mendelian randomization: Concepts and scope**. *Cold Spring Harb Perspect Med* (2022) **12** a040501. DOI: 10.1101/cshperspect.a040501
19. Davey SG, Hemani G. **Mendelian randomization: genetic anchors for causal inference in epidemiological studies**. *Hum Mol Genet* (2014) **23** R89-98. DOI: 10.1093/hmg/ddu328
20. Jansen PR, Watanabe K, Stringer S, Skene N, Bryois J, Hammerschlag AR. **Genome-wide analysis of insomnia in 1,331,010 individuals identifies new risk loci and functional pathways**. *Nat Genet* (2019) **51** 394-403. DOI: 10.1038/s41588-018-0333-3
21. Dashti HS, Jones SE, Wood AR, Lane JM, van Hees VT, Wang H. **Genome-wide association study identifies genetic loci for self-reported habitual sleep duration supported by accelerometer-derived estimates**. *Nat Commun* (2019) **10** 1100. DOI: 10.1038/s41467-019-08917-4
22. Lu Y, Day FR, Gustafsson S, Buchkovich ML, Na J, Bataille V. **New loci for body fat percentage reveal link between adiposity and cardiometabolic disease risk**. *Nat Commun* (2016) **7** 710495. DOI: 10.1038/ncomms10495
23. Kilpeläinen TO, Zillikens MC, Stančákova A, Finucane FM, Ried JS, Langenberg C. **Genetic variation near IRS1 associates with reduced adiposity and an impaired metabolic profile**. *Nat Genet* (2011) **43** 753-60. DOI: 10.1038/ng.866
24. Karlsson T, Rask-Andersen M, Pan G, Höglund J, Wadelius C, Ek WE. **Contribution of genetics to visceral adiposity and its relation to cardiovascular and metabolic disease**. *Nat Med* (2019) **25** 1390-5. DOI: 10.1038/s41591-019-0563-7
25. Chen J, Spracklen CN, Marenne G, Varshney A, Corbin LJ, Luan J. **The trans-ancestral genomic architecture of glycemic traits**. *Nat Genet* (2021) **53** 840-60. DOI: 10.1038/s41588-021-00852-9
26. Xue A, Wu Y, Zhu Z, Zhang F, Kemper KE, Zheng Z. **Genome-wide association analyses identify 143 risk variants and putative regulatory mechanisms for type 2 diabetes**. *Nat Commun* (2018) **9** 2941. DOI: 10.1038/s41467-018-04951-w
27. Elsworth B, Lyon M, Alexander T, Liu Y, Matthews P, Hallett J. **The MRC IEU open GWAS data infrastructure**. *bio Rxiv* (2020) 2020-8. DOI: 10.1101/2020.08.10.244293
28. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D. **The MR-base platform supports systematic causal inference across the human phenome**. *Elife* (2018) **7** e34408. DOI: 10.7554/eLife.34408
29. Ong JS, An J, Han X, Law MH, Nandakumar P, Schumacher J. **Multitrait genetic association analysis identifies 50 new risk loci for gastro-oesophageal reflux, seven new loci for Barrett's oesophagus and provides insights into clinical heterogeneity in reflux diagnosis**. *Gut* (2022) **71** 1053-61. DOI: 10.1136/gutjnl-2020-323906
30. Olsen CM, Green AC, Neale RE, Webb PM, Cicero RA, Jackman LM. **Cohort profile: the QSkin Sun and health study**. *Int J Epidemiol* (2012) **41** 929-929i. DOI: 10.1093/ije/dys107
31. Fry A, Littlejohns TJ, Sudlow C, Doherty N, Adamska L, Sprosen T. **Comparison of sociodemographic and health-related characteristics of UK biobank participants with those of the general population**. *Am J Epidemiol* (2017) **186** 1026-34. DOI: 10.1093/aje/kwx246
32. Burgess S, Thompson SG. **Avoiding bias from weak instruments in Mendelian randomization studies**. *Int J Epidemiol* (2011) **40** 755-64. DOI: 10.1093/ije/dyr036
33. Ong JS, MacGregor S. **Implementing MR-PRESSO and GCTA-GSMR for pleiotropy assessment in Mendelian randomization studies from a practitioner's perspective**. *Genet Epidemiol* (2019) **43** 609-16. DOI: 10.1002/gepi.22207
34. Yin KJ, Huang JX, Wang P, Yang XK, Tao SS, Li HM. **No genetic causal association between periodontitis and arthritis: a bidirectional two-sample mendelian randomization analysis**. *Front Immunol* (2022) **13** 13808832. DOI: 10.3389/fimmu.2022.808832
35. Bowden J, Davey SG, Haycock PC, Burgess S. **Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator**. *Genet Epidemiol* (2016) **40** 304-14. DOI: 10.1002/gepi.21965
36. Burgess S, Thompson SG. **Interpreting findings from Mendelian randomization using the MR-egger method**. *Eur J Epidemiol* (2017) **32** 377-89. DOI: 10.1007/s10654-017-0255-x
37. Chang P, Friedenberg F. **Obesity and GERD**. *Gastroenterol Clin N Am* (2014) **43** 161-73. DOI: 10.1016/j.gtc.2013.11.009
38. Locke AE, Kahali B, Berndt SI, Justice AE, Pers TH, Day FR. **Genetic studies of body mass index yield new insights for obesity biology**. *Nature* (2015) **518** 197-206. DOI: 10.1038/nature14177
39. Chen X, Hong X, Gao W, Luo S, Cai J, Liu G. **Causal relationship between physical activity, leisure sedentary behaviors and COVID-19 risk: a Mendelian randomization study**. *J Transl Med* (2022) **20** 216. DOI: 10.1186/s12967-022-03407-6
40. Cadiot G, Delaage PH, Fabry C, Soufflet C, Barthélemy P. **Sleep disturbances associated with gastro-oesophageal reflux disease: prevalence and impact of treatment in French primary care patients**. *Dig Liver Dis* (2011) **43** 784-7. DOI: 10.1016/j.dld.2011.06.004
41. You ZH, Perng CL, Hu LY, Lu T, Chen PM, Yang AC. **Risk of psychiatric disorders following gastroesophageal reflux disease: a nationwide population-based cohort study**. *Eur J Intern Med* (2015) **26** 534-9. DOI: 10.1016/j.ejim.2015.05.005
42. Jansson C, Nordenstedt H, Wallander MA, Johansson S, Johnsen R, Hveem K. **A population-based study showing an association between gastroesophageal reflux disease and sleep problems**. *Clin Gastroenterol Hepatol* (2009) **7** 960-5. DOI: 10.1016/j.cgh.2009.03.007
43. Zhang J, Lam SP, Li SX, Yu MWM, Li AM, Ma RCW. **Long-term outcomes and predictors of chronic insomnia: a prospective study in Hong Kong Chinese adults**. *Sleep Med* (2012) **13** 455-62. DOI: 10.1016/j.sleep.2011.11.015
44. Lim KG, Morgenthaler TI, Katzka DA. **Sleep and nocturnal gastroesophageal reflux: An update**. *Chest* (2018) **154** 963-71. DOI: 10.1016/j.chest.2018.05.030
45. Onen SH, Alloui A, Gross A, Eschallier A, Dubray C. **The effects of total sleep deprivation, selective sleep interruption and sleep recovery on pain tolerance thresholds in healthy subjects**. *J Sleep Res* (2001) **10** 35-42. DOI: 10.1046/j.1365-2869.2001.00240.x
46. Schey R, Dickman R, Parthasarathy S, Quan SF, Wendel C, Merchant J. **Sleep deprivation is hyperalgesic in patients with gastroesophageal reflux disease**. *Gastroenterology* (2007) **133** 1787-95. DOI: 10.1053/j.gastro.2007.09.039
47. Fang H, Tu S, Sheng J, Shao A. **Depression in sleep disturbance: a review on a bidirectional relationship, mechanisms and treatment**. *J Cell Mol Med* (2019) **23** 2324-32. DOI: 10.1111/jcmm.14170
48. Pandi-Perumal SR, Monti JM, Burman D, Karthikeyan R, BaHammam AS, Spence DW. **Clarifying the role of sleep in depression: a narrative review**. *Psychiatry Res* (2020) **291** 113239. DOI: 10.1016/j.psychres.2020.113239
49. Ju G, Yoon IY, Lee SD, Kim N. **Relationships between sleep disturbances and gastroesophageal reflux disease in Asian sleep clinic referrals**. *J Psychosom Res* (2013) **75** 551-5. DOI: 10.1016/j.jpsychores.2013.10.004
50. Sanna L, Stuart AL, Berk M, Pasco JA, Girardi P, Williams LJ. **Gastro oesophageal reflux disease (GORD)-related symptoms and its association with mood and anxiety disorders and psychological symptomology: a population-based study in women**. *BMC Psychiatry* (2013) **13** 13194. DOI: 10.1186/1471-244X-13-194
51. On ZX, Grant J, Shi Z, Taylor AW, Wittert GA, Tully PJ. **The association between gastroesophageal reflux disease with sleep quality, depression, and anxiety in a cohort study of Australian men**. *J Gastroenterol Hepatol* (2017) **32** 1170-7. DOI: 10.1111/jgh.13650
52. Okuyama M, Takaishi O, Nakahara K, Iwakura N, Hasegawa T, Oyama M. **Associations among gastroesophageal reflux disease, psychological stress, and sleep disturbances in Japanese adults**. *Scand J Gastroenterol* (2017) **52** 44-9. DOI: 10.1080/00365521.2016.1224383
53. Lagergren J. **Influence of obesity on the risk of esophageal disorders**. *Nat Rev Gastroenterol Hepatol* (2011) **8** 340-7. DOI: 10.1038/nrgastro.2011.73
54. Fox CS, Massaro JM, Hoffmann U, Pou KM, Maurovich-Horvat P, Liu CY. **Abdominal visceral and subcutaneous adipose tissue compartments: association with metabolic risk factors in the Framingham heart study**. *Circulation* (2007) **116** 39-48. DOI: 10.1161/CIRCULATIONAHA.106.675355
55. Deng P, Yu Q, Tang H, Lu Y, He Y. **Age at menarche mediating visceral adipose Tissue's influence on pre-eclampsia: a Mendelian randomization study**. *J Clin Endocrinol Metab* (2022) **108** 405-13. DOI: 10.1210/clinem/dgac566
56. Kinekawa F, Kubo F, Matsuda K, Kobayashi M, Furuta Y, Fujita Y. **Esophageal function worsens with long duration of diabetes**. *J Gastroenterol* (2008) **43** 338-44. DOI: 10.1007/s00535-008-2169-6
57. Lluch I, Ascaso JF, Mora F, Minguez M, Peña A, Hernandez A. **Gastroesophageal reflux in diabetes mellitus**. *Am J Gastroenterol* (1999) **94** 919-24. DOI: 10.1111/j.1572-0241.1999.987_j.x
58. Nishida T, Tsuji S, Tsujii M, Arimitsu S, Sato T, Haruna Y. **Gastroesophageal reflux disease related to diabetes: analysis of 241 cases with type 2 diabetes mellitus**. *J Gastroenterol Hepatol* (2004) **19** 258-65. DOI: 10.1111/j.1440-1746.2003.03288.x
59. Lin Z, Deng Y, Pan W. **Combining the strengths of inverse-variance weighting and egger regression in Mendelian randomization using a mixture of regressions model**. *PLoS Genet* (2021) **17** e1009922. DOI: 10.1371/journal.pgen.1009922
|
---
title: Downregulation of KLF10 contributes to the regeneration of survived renal tubular
cells in cisplatin-induced acute kidney injury via ZBTB7A-KLF10-PTEN axis
authors:
- Yang Zhang
- Siyu Bao
- Daxi Wang
- Wei Lu
- Sujuan Xu
- Weiran Zhou
- Xiaoyan Wang
- Xialian Xu
- Xiaoqiang Ding
- Shuan Zhao
journal: Cell Death Discovery
year: 2023
pmcid: PMC9988960
doi: 10.1038/s41420-023-01381-6
license: CC BY 4.0
---
# Downregulation of KLF10 contributes to the regeneration of survived renal tubular cells in cisplatin-induced acute kidney injury via ZBTB7A-KLF10-PTEN axis
## Abstract
Acute kidney injury (AKI) is a common clinical dysfunction with complicated pathophysiology and limited therapeutic methods. Renal tubular injury and the following regeneration process play a vital role in the course of AKI, but the underlining molecular mechanism remains unclear. In this study, network-based analysis of online transcriptional data of human kidney found that KLF10 was closely related to renal function, tubular injury and regeneration in various renal diseases. Three classical mouse models confirmed the downregulation of KLF10 in AKI and its correlation with tubular regeneration and AKI outcome. The 3D renal tubular model in vitro and fluorescent visualization system of cellular proliferation were constructed to show that KLF10 declined in survived cells but increased during tubular formation or conquering proliferative impediment. Furthermore, overexpression of KLF10 significantly inhibited, whereas knockdown of KLF10 extremely promoted the capacity of proliferation, injury repairing and lumen-formation of renal tubular cells. In mechanism, PTEN/AKT pathway were validated as the downstream of KLF10 and participated in its regulation of tubular regeneration. By adopting proteomic mass spectrum and dual-luciferase reporter assay, ZBTB7A were found to be the upstream transcription factor of KLF10. Our findings suggest that downregulation of KLF10 positively contributed to tubular regeneration in cisplatin induced acute kidney injury via ZBTB7A-KLF10-PTEN axis, which gives insight into the novel therapeutic and diagnostical target of AKI.
## Introduction
Acute kidney injury, a syndrome manifested as a rapid decline of renal function, is a common organ dysfunction caused by chemotherapy, surgery, sepsis and etc. According to a meta-analysis of world incidence of AKI, 1 in 5 adults and 1 in 3 children worldwide experience AKI during a hospital episode of care [1]. Nowadays, there are limited diagnostic or therapeutic options for early-stage AKI, thereby molecular mechanisms underlying the progression and recovery of AKI need excavating urgently. Cisplatin received FDA approval for the treatment of kinds of cancer in 1978, which still prevails in the current clinical treatment [2], such as cancers of breast, cervical, esophageal, bladder, small cell lung, and testicular [3–7]. However, AKI occurs frequently even with low-dose cisplatin administration [8], especially the acute tubular necrosis (ATN) [9]. Reversely, Pathophysiology of cisplatin-induced AKI remains shrouded in mystery.
Renal tubules are the most susceptible part of the kidney, especially in cisplatin-induced AKI [10]. AKI causes the injured tubular cells to shed and the remaining tubule begins to repair [11]. However, the mechanism of tubular regeneration still remains obscure. Several studies have found that resident survived tubular cells instead of circulatory stem cells play a significant role in repairing during AKI [12–14], either localized at the urinary pole of Bowman’s capsule or segment-specific and localized separately [11, 15–18]. CD133, CD24 and Vcam1 positive cells tended to be resource proliferative cells responsible for regeneration [19–21]. As activators, molecules such as Pax2, Sox9 and Foxm1 [14, 21, 22] and pathway like EGFR, Wnt-β catenin and Hippo signaling also got involved in the proliferation progress [23–25]. However, the inhibitors in tubular regeneration in AKI are lack of study, for which the subtle cellular proliferation progress cannot be fully understood. Thus, investigating the mechanisms of resident cell-cycling reentry and cease during AKI progression and repairing is of great significance.
Krüppel-like factor 10 (KLF10) is a well-known tumor suppressor of KLFs family because of its inhibitory effect on cell proliferation, and has become a research target for lung cancer, pancreatic cancer and liver cancer [26], which is first discovered as TGFβ inducible early gene 1 in osteoblasts [27]. Moreover, several studies have shown that KLF10 is also involved in the pathophysiology of various acute tissue injury diseases by inhibiting the proliferation of injured cells. In cerebral ischemia-reperfusion injury, down-regulation of KLF10 can inhibit N-myc/PTEN signaling pathway and promote the proliferation and repair of brain nerve cells [28]. In lower extremity ischemia-reperfusion injury, down-regulation of KLF10 can inhibit TGFβ/Smad pathway, alleviate cell cycle arrest caused by injury, and promote proliferation and repair of lower extremity vascular endothelial cells [29]. In the field of nephrology, KLF10 was reported to aggravate diabetic podocyte dysfunction while its role in renal tubular cell remains unclear [30]. Apart from that, it is lack of exploration in kidney, especially in AKI. Therefore, the effect of KLF10 on the proliferation of renal tubular epithelial cells in AKI is worth studying.
In our current work, we used transcriptome combined with clinical data to uncover the relationship between KLF10 and renal function or tubular proliferation. Through not only adoption of classical AKI mouse models in vivo but also construction of 3D renal tubular model in vitro, the role of KLF10 in injury repairing, tubular proliferation and lumen-formation during the course of AKI was further validated. Mechanistically, the up- and down-stream of KLF10 in regulating tubular proliferation were also investigated. Collectively, these findings reveal that KLF10 hinders the regeneration of renal tubules in cisplatin-induced AKI via ZBTB7A-KLF10-PTEN axis, which provides a potential approach for the diagnosis and treatment of AKI.
## The expression of KLF10 in kidney was highly correlated with renal function, renal tubular proliferation and various renal diseases in human datasets
In order to excavate the clinical meaning and biological function of KLF10, we analyzed human renal transcriptomic data of GSE1563 [31]. After outlier analysis (Fig. 1A), well-functioning transplants with no clinical evidence of rejection (clinical status: 1, $$n = 10$$), transplants with acute renal dysfunction without rejection (clinical status: 2, $$n = 5$$), and kidneys undergoing acute rejection (clinical status: 3, $$n = 6$$, 1 outlier) were included in the following analysis. To evaluate the tubular proliferation level of each sample, gene set in corresponding GO terms were taken into the gene set variation analysis labeled with Epithelial Proliferation Score (ProScore, Table 1). ( Considering that the kidney from healthy living donor did not conquer transplant attack or related drug usage which was far more different from transplant kidney and affected the quality of analysis, we used well-functioning transplants as the control instead.) Interestingly, the ProScore was positively correlated to Log2SCr in the scope of the dataset, which confirmed that the renal tubules prone to proliferate as kidney injury progressed (Fig. 1B). The baseline characteristics of the patient were depicted in the Table 2. *The* genes with NA value or duplicate values except the maximum one was removed and 21 samples and 9155 genes were obtained finally. The soft thresholding power was set as 7.0 with correlation over 0.8 after careful consideration (Fig. 1C, D). Then the gene expression profiles could be transformed into the adjacency matrix, TOM, and dissTOM and 10 co-expression modules were constructed subsequently (Fig. 1E). Moreover, dissTOMplot showed that the modules were relatively independent of each other suggesting efficient clustering (Fig. 1F). By adopting Cytoscape, we transferred the dissTOM into protein-to-protein interaction network based on the correlation strength among each genes. The KLF10 with its closely related genes was finally obtained (Fig. 1G), which showed that both KLF10 and its neighboring genes belonged mainly to the turquoise module. Fig. 1Construction of weighted gene co-expression network and clinical and functional analysis of KLF10.A Heatmap with hierarchical clustering dendrogram of kidney samples in GSE1563. B Negative correlation between Epithelial Proliferation Score (ProScore) and Log2 (Serum Creatinine). C Analysis of appropriate soft-thresholding powers for WGCNA. D Average network connectivity under weighting coefficients of WGCNA. E Clustering dendrograms of genes in 21 samples, from which 10 coexpression modules were constructed with different colors. F The Heatmap plot depicts the TOM among all genes in the analysis, which shows the interactions among co-expression modules. The stronger intensity of orange indicates greater overlaps. G Interaction network analysis based on WGCNA and network of KLF10 with its closely interacted genes. Each dot is color-coded by its corresponding module. H Histogram of functional enrichment analysis of genes closely interacted with KLF10. I Bubble plot of the relationship between the coexpression module and clinical data. The size shows the corresponding -log10 (P value) and the color represents the correlation coefficient. J Bubble plot of the relationship between the KLF family and clinical data with histogram displaying the number of closely interacted genes with each KLF. The size shows the corresponding -log10 (P value) and the color represents the correlation coefficient. K Negative correlation between KLF10 and ProScore/Log2 (Serum Creatinine). L Scatter plot of relative expression of KLF10 in well function group versus acute dysfunction group. M ROC curve with AUC value of KLF10. N Scatter plot of tubulointerstitial relative expression of KLF10 in control groups versus renal disease groups. *** $p \leq 0.001$ and ****$p \leq 0.0001$ vs control group at the same experimental conditions. Table 1Selection of GO terms of Epithelial Proliferation Score (ProScore).Screening conditionAND/NOTPositive regulation of epithelial cell proliferation (PRECP)Negative regulation of epithelial cell proliferation (NRECP)GO TermANDGO:0050679GO:0050680OrganismANDHomo sapiensHomo sapiensTypeANDproteinproteinEvidenceANDexperimental evidenceexperimental evidenceGO classNOTpositive regulation of endothelial cell proliferationnegative regulation of endothelial cell proliferationpositive regulation of vascular endothelial cell proliferationnegative regulation of blood vessel endothelial cell proliferation involved in sprouting angiogenesispositive regulation of blood vessel endothelial cell proliferation involved in sprouting angiogenesisnegative regulation of vascular endothelial cell proliferationNumber of Genes for Gene Set Variation Analysis (GSVA)5533Epithelial Proliferation Score (ProScore)= PRECP − NRECPTable 2Baseline characteristics of the patient taken into WGCNA.Well functioning graftAcutely dysfunctional graft without rejectionGraft with acute rejectionP-valuen1056-Age (mean (SD))44.70 (12.77)42.20 (15.77)32.83 (13.33)0.259Male (%)7 (70.0)4 (80.0)4 (66.7)0.88Cadaveric Donor (%)5 (50.0)1 (20.0)5 (83.3)0.109SCr (mean (SD))1.32 (0.27)3.76 (1.94)4.42 (4.03)0.0001Days post transplant (mean (SD))769.90 (105.85)143.20 (184.32)430.50 (519.88)0.031ProScore (mean (SD))−0.08 (0.21)0.02 (0.22)0.10 (0.24)0.363 Functional enrichment analysis of the KLF10 cluster above indicated that KLF10 might participate in the ‘mitotic cell cycle’ and ‘regulation of growth’ (Fig. 1H). Then, the correlation coefficients for the modules/KLF family and the clinical data were calculated to identify how each module or gene was related to each clinical trait. It was obvious that both KLF10 and its belonging turquoise module were negatively related to clinical status, serum creatinine and regeneration of renal tubules (Fig. 1I–K). KLF10 also showed the most closely related genes among the KLF family which illustrated that KLF10 might played the greatest role in AKI versus the other KLF members (Fig. 1J). Correspondence to the results above, KLF10 decreased significantly in acute dysfunction group and ROC curve with AUC value over 0.9 showed high specificity and sensitivity of KLF10 to AKI (Fig. 1L, M). Furthermore, renal tubulointerstitial RNA-seq data of other kidney diseases from Nephroseq database also displayed significant downregulation of KLF10 (Fig. 1N), which validated the vital role of KLF10 in injury renal tubule. Above all, the level of KLF10 in kidney was clinically correlated to renal function and renal tubular regeneration.
## KLF10 dramatically decreased in AKI mouse models and in 3D renal tubular injury model, correlated tightly with the proliferative states of renal epithelial cells
Relevance between downregulation of KLF10 and AKI was further investigated in three classical AKI mouse models. Mice were treated with 30 min IR or 20 mg/kg cisplatin intraperitoneally administered or cecum ligation and puncture (CLP) and then sacrificed at corresponding time point respectively (Fig. 2A). All three models showed significant decrease in renal expression of KLF10 versus control group. Moreover, proliferation marker PCNA of IR group and cisplatin group and CCNB1 of CLP group increased, suggesting the occurrence of cellular proliferation in AKI (Fig. 2B and Supplementary Figs. 2A, 3A).Fig. 2KLF10 was down-regulated in multiple AKI mouse models and 3D renal tubule injury models. A Flowchart detailing bilateral IR or CLP surgery and cisplatin treatment regime in C57BL/6 J mice. B Representative immunoblot analysis of KLF10 and PCNA in kidney tissues from cisplatin-induced AKI mouse model. ACTIN served as the standard. $$n = 6$$ per group. C, D The morphology and injury rate of 3D renal tubular model treated with cisplatin of different concentration. E Representative immunoblot analysis of KLF10 in cisplatin treated 3D renal tubular model. ACTIN served as the standard. F Fluorescence images of 3D tubular cisplatin-induced AKI model (75 μM) at different time points reflecting the injury degree (LM - light microscope and GFP) and proliferative activity (Fucci-overall, Merge and Fucci). *** $p \leq 0.001$ vs control group at the same experimental conditions.
To validate the in vivo results, 3D renal tubular injury in vitro model was constructed. We observed time- and dose-dependent changes in tubular morphology, which were highly similar to tubular changes of AKI in vivo, including cell shedding, apoptosis and necrosis (Fig. 2C, D and Supplementary Figs. 1A, B). Downregulation of KLF10 was also found in this model (Fig. 2E). The degrees of tubular injury and proliferation at different time points were then observed (Fig. 2F). The proliferation was significantly activated as the damage progressed (Fig. 2F - LM/GFP) and peaked at 9 h (Fig. 2F - Merge/Fucci/Overall), which was consistent with the changes of KLF10 in vitro/in vivo and PCNA in vivo. However, the proliferation was suppressed at 12 h because the structure of tubular organoid was lost mostly and tubular cell death occurred the most. Thus, KLF10 might get involved in the tubular injury and tubular proliferation of cisplatin-induced AKI.
Given that KLF10 got involved in the tubular injury process of AKI, we further explored the change of KLF10 as the AKI recovered. A dose of 20 mg/kg cisplatin was intraperitoneally administered to C57BL/6 J mice with kidney harvested 36, 72 and 120 h later (Fig. 3A). H&E staining also showed obvious morphological changes on day 3 including loss of tubular brush, vacuolar degeneration and acute tubular necrosis, which got milder on day 5 (Fig. 3B). SCr, KIM1, and NGAL ascended the most significantly at 72 h and descending at 120 h, suggesting that the renal function got worst on day 3 and moderately recovered on day 5 (Fig. 3D). Interestingly, levels of cell proliferation markers PCNA, FOXM1 and Ki67 as well as KLF10 developed the same changing trend (Fig. 3C–E). Moreover, KLF10 was observed in plasm and nuclei of renal tubules and both levels were reduced and then regained as AKI progressed and then recovered (Fig. 3G–I). Regression analysis was then undertaken based on KLF10 protein level and SCr/blood urea nitrogen (BUN)/KIM1, for further validation of the relevance between KLF10 and renal function and injury (Fig. 3F and Supplementary Fig. 4). The decrease and recovery of KLF10 could be observed in IR- and CLP-induced mouse model as well (Supplementary Figs. 2B, 3B).Fig. 3Downregulation of KLF10 was correlated with cell proliferation and the outcome of cisplatin-induced AKI in vivo. A Flowchart detailing cisplatin treatment regime in C57BL/6 J mice. $$n = 6$$ per group. B Representative H&E staining image of kidney sections from the model. C Representative immunofluorescence staining image of Ki67 in the model and counts of positive cells per 40x field (objective lens). D Renal dysfunction was determined in cisplatin (20 mg/kg, ip) treated mouse model. Serum creatinine (SCr) was measured in sera. Degree of renal injury was determined in the model through relative mRNA levels of KIM1, NGAL and IGFBP7 in mouse kidney tissues. Degree of renal cellular proliferation was determined in the model through relative mRNA levels of PCNA, FOXM1 and Ki67 in mice kidney tissues. E Representative immunoblot analysis of KLF10, KIM1 and PCNA in the model. ACTIN served as the standard. F *Regression analysis* was undertaken to determine correlation between relative protein level of KLF10 in mice kidney tissues and SCr. Relative protein level of KLF10 was examined and normalized by Fiji. G–I Representative immunofluorescence staining image and rate of positive nuclear of KLF10 in the model. * $p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$, ns no significantly difference vs control group at the same experimental conditions.
To specified KLF10 changes in renal tubular cells further, 3D renal tubular injury model in vitro was treated with 75 μM cisplatin for 6 h or 9 h and then cultured in normal condition without cisplatin for another 15 h to recover (Fig. 4A). The KLF10 decreased after 9 h treatment and rose up after cisplatin withdrawal for 15 h (Fig. 4B). Fucci sensor cellular system [32] was constructed to visualize cell cycle progression, which showed red fluorescence (mCherry) in G2/M cells (Fig. 4C). To dissect the change of KLF10 in the survived cells from that in the apoptotic or necrotic one, Fucci renal tubule cell (without green fluorescence) and puromycin-resistant Fucci renal tubule cell (with green fluorescence) were mixed at the ratio of 1:1 and then cultured in matrix for 6 days to form 3D renal tubule. Puromycin was added for 12 h to induce apoptosis of the sensitive cells, the rest continued to be cultured for another 12 h or 24 h (Fig. 4D). The density of red fluorescence raised dramatically at 12 h and 24 h and then declined at 36 h (Fig. 4F Fucci fluorescence) as the tubule fully repaired (Fig. 4F GFP fluorescence), which was confirmed further by the change of PCNA (Fig. 4E). Consistent with results in vivo and in vitro above, the expression pattern of KLF10 was opposite to that of PCNA (Fig. 4E). Therefore, the change of KLF10 expression level reflects the injury, healthy or recovery state of renal tubules throughout the course of AKI both in vivo and in vitro. Fig. 4Downregulation of KLF10 in survived tubular cells was reversed as the integrity of injury 3D renal tubules recovered in vitro. A Diagram detailing cisplatin treatment regime in 3D renal tubular model. B Representative immunoblot analysis of KLF10 and PCNA in model of Fig. 4A. ACTIN served as the standard. C Diagram detailing the change in flurescent color of Fucci-cell during different stages of cell cycle. D Diagram detailing puromycin treatment regime in 3D renal tubular model. E Representative immunoblot analysis of KLF10 and PCNA in model of Fig. 4D. ACTIN served as the standard. F Fluorescence images of 3D survived tubular repair model according to the treatment in Fig. 4D, reflecting the forming integrity of survived tubular cells (LM and GFP) and proliferative activity (Fucci-overall, Merge and Fucci).
## KLF10 was an inhibitor of tubular cell proliferation in AKI
Validation of the relationship between KLF10 and cell proliferation was then carried out. Firstly, analysis of datasets published on GEPIA indicated that renal tumor tissue (KICH, KIRP) had much lower level of KLF10 than the normal tissue (Fig. 5A) [33]. Secondly, immunofluorescence staining also demonstrated that there was less KLF10 in tubular cells in G2/M phase versus those in non-G2/M phase, especially in the nuclei (Fig. 5B). Moreover, when cell proliferation was inhibited by physically contraction, an elevated expression of KLF10 was observed compared to low density group (Fig. 5C, D). In addition, we mimicked the regeneration pattern of normal renal tubule in vitro, upregulation of KLF10 was observed at the late stage of tubular formation concomitant with lower rate of proliferation through levels of CCNB1, CCND1 and Fucci fluorescence (Fig. 5E–G).Fig. 5KLF10 increased when experiencing physical inhibition to proliferation or during the formation of 3D renal tubule model in vitro. A Datasets published on GEPIA were examined for KLF10 transcript expression between tumor tissue (T) and normal tissue (N). KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma. B Representative immunofluorescence staining image of KLF10 and Fucci in 2D cultured renal tubular cells. C Representative immunoblot analysis of KLF10, CCNB1 and CCND1 in low-/middle-/high-density group. ACTIN served as the standard. D Representative fluorescence image of Fucci in low-/middle-/high-density group. Rate of proliferation was calculated via percentage of Fucci positive cells. E Representative immunoblot analysis of KLF10, CCNB1 and CCND1 in early/middle/late stages of tubular formation in 3D renal tubular model. F Relative mRNA levels of KLF10 in different stages of tubular formation in 3D renal tubular model were examined. G Representative fluorescence image of Fucci during different stages of tubular formation in 3D renal tubular model (early stage: day 1–2, middle stage: day 3–4, late stage: day 5–6). ns p ≥ 0.05, **$p \leq 0.01$, ***$p \leq 0.001$ vs control group at the same experimental conditions.
To gain more insight into the role of KLF10 in cell proliferation, tubular formation and repairing, overexpression of KLF10 was carried out in vitro (Fig. 6A, B). Expression of PCNA and CCND1 decreased in KLF10-overexpressed cells (Fig. 6C). Proliferation rate declined significantly as well in both 2D and 3D cultured in vitro model (Fig. 6F, G), which was validated further through CCK-8 assay (Fig. 6D). Overexpression of KLF10 also inhibited the ability to repair based on wound healing assay (Fig. 6E). We could also observe that KLF10-overexpressed 3D model showed larger rate of small tubules and less rate of large tubules (Fig. 6H). Then, we overexpressed KLF10 in 3D tubular organoid under cisplatin stimuli (Fig. 6I). As we expected, overexpression of KLF10 not only exacerbated tubular injury such as more shedding and death of tubular cells and loss of tubular structure (Fig. 6I - LM/GFP), but also inhibited the proliferation and activation of proliferation after cisplatin stimuli strongly (Fig. 6I - Merge/Fucci/Overall), validating the negative role of highly expressed KLF10 on cell proliferation under condition of cisplatin stimuli. Consistently, knockdown of KLF10 exerted converse effects on proliferation (Fig. 6J–L). Considering together, KLF10 got involved not only in the cell proliferation but also in the formation and repair of renal tubules in AKI.Fig. 6Overexpression or suppression of KLF10 affected the ability of proliferation, repair and lumen-formation of renal tubular cells. A, B Efficient overexpression of KLF10 was confirmed by immunoblot analysis and relative mRNA levels. C Representative immunoblot analysis of PCNA and CCND1 in KLF10 overexpressed group and control group. ACTIN served as the standard. D The cell viability of KLF10 overexpressed group and control group were measured by CCK-8 assay. E The ability to repair was measured by wound healing assay. F, G Representative fluorescence image of KLF10 overexpressed group and control group (2D/3D cultured). H Distribution of relative volume of tubule model was calculated and analyzed by Fiji. I Fluorescence images of 3D tubular cisplatin-induced AKI model (75 μM) for 9 h with or without overexpression of KLF10 reflecting the injury degree (LM and GFP) and proliferative activity (Fucci-overall, Merge and Fucci). J Relative mRNA levels of KLF10 in siRNA transfected group and control group were examined. K Representative immunoblot analysis of KLF10 and PCNA in siRNA transfected group and control group. ACTIN served as the standard. L Representative fluorescence image of Fucci in siRNA transfected group and control group. * $p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ vs control group at the same experimental conditions.
## KLF10 impeded tubular regeneration via PTEN/AKT signaling
To elucidate the mechanism by which KLF10 inhibits cell proliferation, we studied the PTEN/AKT pathway activation by KLF10 in AKI mouse models and renal tubular epithelial cells, because mounting evidence suggests that PTEN/AKT pathway has great negative effects on cellular proliferation and cerebral ischemia-reperfusion injury [28, 34, 35]. Decreasing PTEN with increasing phosphorylation of AKT was observed in IR-, cisplatin- and CLP-induced AKI mouse models (Fig. 7A and Supplementary Figs. 2C, 3C), which could be also examined in cisplatin treated 3D model in vitro (Fig. 7B). Consistent with the trend of KLF10 expression, PTEN elevated significantly not only in high-density cultured tubular cells but also in the late stage of tubular formation (Fig. 7C, D). Phosphorylation of AKT reduced obviously as well when tubular proliferation stagnated due to high-density. Interestingly, in the late stage of tubular formation, pAKT bumped up on the contrary which might be due to other potential mechanism in 3D cultured condition (Fig. 7D). Thus, PTEN/AKT also participated in the progression of AKI and tubular regeneration. Fig. 7KLF10 inhibited tubular regeneration via PTEN/AKT pathway. A Representative immunoblot analysis of PTEN, pAKT and AKT in cisplatin-induced AKI group and control group. ACTIN served as the standard. $$n = 6$$ per group. B Representative immunoblot analysis of PTEN, pAKT and AKT in 3D AKI group and control group. ACTIN served as the standard. C Representative immunoblot analysis of PTEN, pAKT and AKT in low-/middle-/high-density group. ACTIN served as the standard. D Representative immunoblot analysis of PTEN, pAKT and AKT in early/middle/late stages of tubular formation in 3D renal tubular model. ACTIN served as the standard. E Representative immunoblot analysis of PTEN, pAKT and AKT in siRNA transfected group and control group. ACTIN served as the standard. F Representative immunoblot analysis of PTEN, pAKT and AKT in KLF10 overexpressed group and control group. ACTIN served as the standard. G Representative immunoblot analysis of PTEN, pAKT and AKT in bpV treatment group with different concentration. ACTIN served as the standard. H The cell viability in bpV (4 μM) treated/KLF10 overexpression/bpV (4 μM) treated + KLF10 overexpression group and control group were measured by CCK-8 assay. I Representative fluorescence image of Fucci in bpV (4 μM) treated/ KLF10 overexpression/ bpV (4 μM) treated + KLF10 overexpression group and control group (2D/3D cultured). J Representative image of 3D renal tubular model in cisplatin (75 μM) treated/cisplatin (75 μM) treated + KLF10 overexpression/cisplatin (75 μM) treated + KLF10 overexpression + bpV (4 μM) treated group and control group. ns p ≥ 0.05, *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ vs control group at the same experimental conditions.
Next, in order to clarify the relationship between PTEN/AKT and KLF10, overexpression and suppression of KLF10 were carried out in vitro. Knockdown of KLF10 led to reduced PTEN and increased pAKT (Fig. 7E). On the other hand, PTEN increased evidently with decreased pAKT in KLF10 overexpressed group (Fig. 7F). Therefore, PTEN/AKT pathway was the downstream of KLF10.
Furthermore, we adopted bpV, an efficient PTEN-specific inhibitor [36], to regulate PTEN/AKT pathway in vitro. Treatment with bpV of 4 μM showed the weakest PTEN and the strongest pAKT in renal tubular cells (Fig. 7G), which was adopted in the following treatment. Through CCK8 assay, we found that inhibition of PTEN could reverse the negative effect of KLF10 overexpression on tubular proliferation (Fig. 7H), which was confirmed by the Fucci fluorescence in both 2D and 3D cultured model (Fig. 7I). Moreover, tubular injury deteriorated by overexpression of KLF10 while inhibition of PTEN alleviated the worsen condition (Fig. 7J). Overall, KLF10 impeded tubular regeneration to worsen renal injury via PTEN/AKT pathway.
## The inhibitive effect of KLF10 on tubular regeneration in AKI was regulated by transcription factor ZBTB7A
In view of the important role of KLF10 in the regulation of renal epithelial proliferation, we finally investigated the upstream of KLF10. Previous study and the results above showed that the expression of KLF10 in renal tubular cells could be induced by TGFβ1 treatment [27] and high-density culture condition. The nucleoprotein from control group, high-density cultured group and TGFβ1 treated group was extracted respectively and then co-incubated with the biotin-tagged promoter segment of KLF10. Through overnight incubation and purification, the nucleoprotein combining to the promoter of KLF10 was obtained (Fig. 8A). Silver staining SDS-PAGE gel of nucleoprotein precipitated with promoter of KLF10 uncovered a significant enhanced band around 60 kD in both high-density cultured group and TGFβ1 treated group versus the control (Fig. 8B). After identification by protein mass spectrometry assay, the protein in both high-density cultured group and TGFβ1 treated group and exclusive of protein from control group was obtained (N1 = 125). All of the pulldown protein in classical transcription factor databases (TFDB [37] and JASPAR, https://jaspar.genereg.net, [38]) exclusive of protein from control group was then obtained (N2 = 2). The intersection between N1 and N2 was ZBTB7A (Fig. 8C), which could be detected in both high-density cultured group and TGFβ1 treated group instead of control group (Fig. 8D). Thus, ZBTB7A might be the transcription factor combining to the promoter of KLF10 especially when the proliferation of renal tubular cell slowed down. Fig. 8ZBTB7A promoted KLF10 expression by combining to the promotor regions of KLF10.A Flowchart detailing extraction and analyses of transcription factors combining KLF10 promoter in C57BL/6 J mice. B Silver staining SDS-PAGE gel of nucleoprotein precipitated with promoter of KLF10. C Venn plot of precipitated proteins in high-density, TGFβ1 treated group and control group. UpSetR plot of common transcription factors database and proteins only in high-density and TGFβ1 treated group not in control group (marked in red in Venn plot). D Protein mass spectrometer of ZBTB7A in high-density/TGFβ1 treated group and not detected in control group. E Scatter plot of relative expression of KLF10 in normal function group versus dysfunction group in Flechner database. F Negative correlation between ZBTB7A and SCr in Saint-Mezard database. G, H Representative immunoblot analysis and relative mRNA levels of ZBTB7A in cisplatin-induced AKI group and control group. ACTIN served as the standard. $$n = 6$$ per group. I Positive correlation between KLF10 and ZBTB7A in TCGA and GTEx normal kidney tissue database. J–L Representative immunoblot analysis and relative mRNA levels of ZBTB7A and KLF10 in ZBTB7A overexpressed group and control group. M Luciferase activity quantified in 293 T cells of interaction between ZBTB7A and KLF10 promoter. * $p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ vs control group at the same experimental conditions.
Then, we explored public transcriptomic data [31, 39] and found that ZBTB7A decreased significantly in acute dysfunctional kidney versus well-functioning kidney and negatively correlated to SCr (Fig. 8E, F). Consistent with the expression pattern of KLF10, the declining of ZBTB7A was further validated in cisplatin-, CLP- and IR-induced AKI mouse models (Fig. 8G/H and Supplementary Figs. 2D/E, 3D/E). Therefore, ZBTB7A also took part in the progression of AKI as KLF10 did.
In order to validate the transcription factor role of ZBTB7A to KLF10, transcriptomic data of kidney para-carcinoma tissue in TCGA and normal kidney tissue in GTEx from GEPIA [33] showed significant positive correlation between KLF10 and ZBTB7A (Fig. 8I). Moreover, overexpression of ZBTB7A in 293 T induced the expression of KLF10 (Fig. 8J–L). Furthermore, luciferase activity of the 293 T cells was significantly activated when co-transfecting pGL3-KLF10-promoter and pCMV-ZBTB7A, confirming that ZBTB7A promoted the expression of KLF10 by combing to its promoter (Fig. 8M). Taken together, proliferative inhibitive effect of KLF10 in AKI was positively regulated by transcription factor ZBTB7A.
## Discussion
Cell proliferation is the key process for tissue regeneration after injury. Cytokines and mechanisms that promote renal epithelial cell proliferation were well studied [14, 19–25], but factors that inhibit or cease the proliferate process were rarely defined. In this study, we provided novel evidence showing that renal epithelial cell down-regulated ZBTB7A in response to injury, and then decreased KLF10 expression which further inhibited PTEN/AKT signaling for initiating cell proliferation. We first unveiled the downregulation of KLF10 in human renal transcriptomic database of acute kidney dysfunction, 3 classical AKI mouse models and 3D AKI tubular model, which suggested evidently that KLF10 was closely related to renal function and tubular proliferation. Consistently, as injury alleviated in AKI model in vivo and in vitro, KLF10 rose again. KLF10 also elevated obviously in the group with contact inhibition and the late stage of tubular formation. Overexpression or knockdown of KLF10 directly influenced the capacity of renal tubular cells to proliferate, form lumen and repair, which further clarified the causal connection between KLF10 and tubular regeneration. Moreover, ZBTB7A-KLF10-PTEN axis was found to be the specific mechanism of regenerative regulation of KLF10.
The transplant kidney data were first adopted for analysis, since the resource of human kidneys with common AKI has been always limited due to such kidneys being seldomly biopsied. Instead, kidney transplants become the valuable resources, for kidney transplants often experience AKI and are extensively followed up with detailed clinical data and kidney punctures, even in the well-functioning stages [39, 40]. Considering that we aimed to excavate the renal injury and renal function related changes alone rather than changes owing to transplant process or drug usage, the kidneys from healthy living donor not conquered transplant/ drug attack were excluded. Well-functioning transplants were used as the control instead. Therefore, the downregulation of KLF10 and its clinical meaning for renal function were demonstrated convincingly from realistic human data to experimental multiple mouse models in vivo, and finally to cellular model in vitro (Figs. 1–3). Moreover, transcriptomic data from tubulointerstitium of other kidney diseases showing decreasing KLF10 revealed the significant role of KLF10 in renal tubules.
Renal repair process is always the front burner topic in the area of kidney injury, especially in AKI. Tubular injury and regeneration, the most common issue in AKI, thereby becomes the vital core in the pathophysiological process of AKI [10, 11]. However, only AKI model in vivo cannot distinguish the exact changes in renal tubules from other tissue parts and common cellular culture is far from the condition in vivo for loss of epithelial polarity. Thus, we constructed 3D tubular injury model in vitro to mimic the conditions in AKI progress in vivo, which was much more persuasive. As the manifestation in vivo, cell shedding, apoptosis and necrosis was also observed in 3D tubular model (Fig. 2). Moreover, we separated the survived cell from injury dead cells to specify the changes of KLF10 (Fig. 4). Furthermore, Fucci cell cycle visualization cellular system was also applied into our model, which could monitor the proliferation activity continuously in a real-time living way [32]. And when tubular proliferation slowed down in high-density culture condition and the late stage of tubular formation, KLF10 elevated significantly (Fig. 5).
For the experiments above only verified the simultaneity between KLF10 and tubular regeneration in AKI, overexpression and knockdown of KLF10 were carried out to validate their causal relationship. The ability to proliferate, repair and form lumen were all inhibited remarkably (Fig. 6), whereby KLF10 might also get involved in the renal development and self-renewal process.
To specify the exact mechanism, PTEN/AKT signaling was focused on for its roles of tumor suppressor and downstream of KLF10 in previous tumor study [28, 34, 35] and the same change of expression as KLF10 in AKI. At the same time, elevation of PTEN was examined in physical inhibition model and the late stage of tubular formation. Moreover, inhibiting PTEN could reverse the negative effect of KLF10 on tubular regeneration in cisplatin-induced AKI, which confirmed the downstream role of PTEN/AKT signaling (Fig. 7). Interestingly, phosphorylation of AKT increased at the late stage of tubular formation, which might be due to other effects of pAKT in lumen formation and needs further study. In the meanwhile, the upstream transcription factor ZBTB7A was found by protein mass spectrometer assay of nucleoprotein precipitated with the promoter of KLF10. Previous study has reported that ZBTB7A suppressed the proliferation of several epithelial tumor [41, 42]. Thus, other than the same expression pattern as KLF10 in vivo, the transcription factor role of ZBTB7A was validated by dual-luciferase reporter assay, which integrated the up- and down-stream regulative mechanism of KLF10.
There were also several limitations in our study. The human samples in this study were from public database and limited to the transcriptomic level, lacking further validation of protein levels due to unavailability to samples of healthy and AKI human kidney. Meanwhile, mouse models in vivo were not further validated by employing KLF10 knockout mice due to time and financial constraints. The main object of this study was cisplatin-induced AKI and in the in vivo models, we also extended our idea in IRI- and CLP-induced AKI, suggesting the potential universal significance of KLF10 in AKI with various etiology. However, only cisplatin was used to induce 3D tubular injury in our in vitro study owing to the limitation of equipment for hypoxia culture and huge differences of the LPS treated 3D tubular model in vitro from sepsis model in vivo, which would be explored and constructed in the future when available.
## Conclusions
Collectively, our study proposed a novel mechanism wherein downregulation of KLF10 contributed to the proliferation of survived renal tubular cells in cisplatin-induced acute kidney injury via ZBTB7A-KLF10-PTEN axis, which might be a promising diagnostic and therapeutic target of AKI.
## Data collection, weighted gene co-expression network analysis, functional enrichment analysis, and code availability
GSE1563 dataset was obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) [31], WGCNA R package (version: 1.70-3) was performed to construct a scale-free network. The soft-thresholding power was set as 7 with a correlation coefficient of 0.8. Cluster analysis divided highly correlated genes into gene modules after proper infiltration. After topological overlap matrix (TOM) was obtained, the data was transferred to the network into the edge and node list files Cytoscape could read. Subsequently, the network diagrams were plotted by Cytoscape 3.9.0. Functional enrichment analysis was carried out by Metascape (https://metascape.org) [43]. Renal tubulointerstitial RNA-seq data of other kidney diseases were obtained from Nephroseq database (www.nephroseq.org, version November 2022, University of Michigan, Ann Arbor, MI). The tutorial and source code of WGCNA could be found on the website (https://horvath.genetics.ucla.edu/html/CoexpressionNetwork/Rpackages/WGCNA/Tutorials/).
## Animal models
Male C57BL/6 J mice (7–9 weeks of age, weighing 20–25 g) were obtained from Shanghai Jihui Laboratory Animal Care Co. LTD, Shanghai, China. All the protocols were approved by the Animal Care and Use Committee of Zhongshan Hospital and were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All the experiments were replicated at least twice and mice were randomized to each group.
To establish IR-induced AKI model, bilateral renal pedicles were clamped for 30 min for mice in the IR group with mice body temperature retained at 35–36 °C during all surgical procedures. The same operation except clamping of renal pedicles for mice in the sham group. Kidneys were collected for analysis at different time points after reperfusion.
To establish the cisplatin-induced AKI model, mice in the cisplatin group were intraperitoneally injected with a single dose of cisplatin (Sigma–Aldrich) at 20 mg/kg and mice of the control group were received saline alone. Kidneys were collected for analysis at different time points after treatment.
To establish the sepsis-induced AKI model, the cecum was exteriorized and ligated distal of the ileocecal valve. Then, the cecum was perforated three times using a 20-gauge needle and squeezed to extrude fecal contents that were spread around the cecum using a cotton swab, with mice body temperature retained at 35–36 °C during all surgical procedures. In sham animals, the cecum was exteriorized without ligation and puncture. Kidneys were collected for analysis at different time points after surgery.
## Cell culture and treatment
Madine-Darby Canine Kidney (MDCK) cells were purchased from the American Type Culture Collection. Cells were cultured in Minimum Essential Medium Eagle (Sigma-Aldrich) supplemented with $5\%$ FBS (Gibco) under the condition of 37 °C, $5\%$ CO2 and saturation humidity. The plasmids harboring KLF10/NC siRNA manufactured by Shanghai Genechem Co., Ltd and lentivirus vectors harboring CDS of KLF10 and GFP/puromycin-resistance and GFP/control manufactured by HanBio Co, Ltd, Shanghai, China were transfected into cells according to the manufacturer’s guidelines.
MDCK cells were also used for the 3D culture. The Matrigel matrix (Corning, USA) and the cells were mixed well and spread on a Matrigel matrix-coated culture plate. After culturing for 5–6 days, 3D renal tubules were formed.
MDCK-Fucci cells were kindly provided by professor Cai Liang in school of life sciences at Fudan University, which would show red fluorescence during G2 to M phase.
Cisplatin was dissolved to 1 mg/mL in saline and diluted with MEM to different final concentrations before usage. Cells were collected at different time points after cisplatin treatment. Recombinant TGFβ1 (10804-HNAC, SinoBiological) was dissolved to 2 μg/mL in in sterile 4 mM HCl containing 1 mg/mL bovine serum albumin and diluted with MEM to different final concentrations before usage. Cells were collected 24 h after TGFβ1 treatment. PTEN specific inhibitor bpV (HOpic) (S8651, Selleck) was dissolved to 1 mM in ddH2O and diluted with MEM to different final concentrations before usage. Cells were collected 72 h after bpV treatment.
## Western blot analyses
Proteins from cultured cells or mice kidneys were extracted with TRI Reagent (Sigma–Aldrich) according to the manufacturer’s guidelines. Samples were separated by SDS-PAGE and transferred onto PVDF membranes (IPVH00010, Millipore). After blocking the membranes with $5\%$ milk, we used primary antibodies to incubate the membranes overnight at 4 °C against the following proteins: KLF10 (1:1000, ab184182, Abcam), CCNB1 (1:1000, GTX100911, GeneTex), CCND1 (1:1000, GTX108624, GeneTex), PCNA (1:1000, 101239-T46, SinoBiological), KIM1 (1:1000, AF1817, R&D), Anti-FLAG (1:1000, 8146, CST), ACTIN (1:2000, GTX11003, GeneTex), ZBTB7A (1:1000, ab175918, Abcam), PTEN (1:1000, AB170941, Abcam), pAKT (1:1000, 4060 S, CST), AKT (1:1000, 4691 S, CST). The membranes were washed by TBST before incubated with secondary antibodies (1:2000, Jackson ImmunoResearch Inc). The bands of the target proteins were visualized by the LAS-3000 detection system and were quantitively analyzed by Fiji based on ACTIN and control groups respectively.
## Real-time RT-PCR
Total RNA from MDCK and kidney tissues were extracted using TRI Reagent (Sigma–Aldrich). The reverse transcription and real-time RT-PCR were carried out using PrimeScript RT Master Mix and SYBR Premix ExTaqTM (TaKaRa) on QuantStudio 5 (Thermo Fisher Scientific). Gene expression was measured relative to ACTIN or GAPDH and 2−ΔΔCt method was used to calculate the fold change differences of the experimental groups compared to the control group. The following primers (Sangon Biotech) 5’ to 3’ were used:*Canis lupus* familiarisKLF10FCTCCCGGGTACACCTGATTTTRGCAATGTGAGGCTTGGCAGTATCACTINFTGCGGCATCCATGAAACTACRACAGCACTGTGTTGGCATAGMus musculusKLF10FATGCTCAACTTCGGCGCTTRCGCTTCCACCGCTTCAAAGHAVCR1FAGGCGCTGTGGATTCTTATGRAAGCAGAAGATGGGCATTGCLCN2FTGGCCCTGAGTGTCATGTGRCTCTTGTAGCTCATAGATGGTGCIGFBP7FTAACCTGCGAATCCATGAGCRAGAGAAGTGTGTCAGGCAAGAGPCNAFTTTGAGGCACGCCTGATCCRGGAGACGTGAGACGAGTCCATFOXM1FGGACATCTACACTTGGATTGAGGRTGTCATGGAGAGAAAGGTTGTGMKi67FATCATTGACCGCTCCTTTAGGTRGCTCGCCTTGATGGTTCCTGAPDHFAGGTCGGTGTGAACGGATTTGRGGGGTCGTTGATGGCAACAACTINFAGCCATGTACGTAGCCATCCRGCTGTGGTGGTGAAGCTGTAHomo SapiensZBTB7AFGCTTGGGCCGGTTGAATGTARGGCTGTGAAGTTACCGTCGGKLF10FCTTCCGGGAACACCTGATTTTRGCAATGTGAGGTTTGGCAGTATCACTINFTCACCCACACTGTGCCCATCTACGARCAGCGGAACCGCTCATTGCCAATGG
## Immunofluorescence and hematoxylin-eosin (H&E) staining
After treatment, renal tissues were fixed with $4\%$ paraformaldehyde, embedded in paraffin wax and sliced into 3-μm-thick sections for hematoxylin–eosin (H&E) staining. The slides were incubated in the mixed primary antibodies overnight at 4 °C: KLF10 (1:100, 11881-1-AP, Proteintech) and Ki67 (1:50, GB111141, Servicebio). The slices were then incubated with Alexa Fluor® 488-conjugated goat anti-rabbit IgG (1:200, ab150077, Abcam). Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI).
After treatment, MDCK cells which crawled on the slide were fixed in $4\%$ paraformaldehyde for 15 min. Cells were permeabilized by $0.5\%$ Triton X-100 in PBS for 10 min. After blocking with $5\%$ BSA in PBS, sections or cells were incubated with primary antibody against KLF10 (1:100, 11881-1-AP, Proteintech) followed by Alexa Fluorophore 488-conjugated secondary antibody (1:200, abs20020, Absin Bioscience). Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). The results were visualized by Olympus microscope (Tokyo, Japan) and the light microscopy (Leica DM 6000B; Leica Microsystems, Wetzeler, Germany).
## Serum creatine ELISA
Serum creatinine levels of mice were determined using a QuantiChromTM Creatinine Assay Kit (BioAssay Systems, Hayward, CA, USA) following the manufacturer’s guidelines.
## Cell proliferation assay
2000 MDCK cells overexpressed KLF10 or control were seeded into a 96-well plate per well. CCK-8 reagent (Dojindo) was added after treatment and incubated for 90 min at 37 °C. Cell viability was assessed by absorbance at 450 nm using the spectrophotometer (Thermo Fisher Scientific) 0, 24, 48 and 72 h after adherence.
## Wound healing assay
4 × 105 MDCK cells overexpressed KLF10 or control were seeded into 6-well plate per well and cultured until $100\%$ confluence. Then confluent cultures were scratched using a pipette tip. After scratching, the dishes were gently washed twice with medium to remove the detached cells. Scratched cultures were photographed under a microscope at 0, 12 and 24 h. Capacity to repair of cells was established by measuring the width of the scratched area at each time point in the scratched area.
## DNA pulldown, nucleoprotein extraction and mass spectrometry
Biotin labeled KLF10 promoter DNA was prepared by PCR amplifying using forward primer: 5’-biotin/TTCTCGAGTCACAAGTCAAGACCGCTCCCT-3’ and reverse primer: 5’-biotin/ TTAAGCTTTTGAGCTCGGTGTAGCTGAAGTTTAAA-3’. The resulting 1100 bp DNA fragment were further gel extracted and purified by DNA gel extraction kit (AxyPrepTM, Ap-GX-25). MDCK nuclear extracts were prepared using NE_PER Nuclear and Cytoplasmic Extraction kit (Thermo Fisher, 78833), and quantified by BCA. Briefly, following nuclear extraction, 1 mg of lysate was incubated with 3 pmol of biotinylated DNA along with 120 μL of streptavidin magnetic beads (Thermo Fisher, 88816). The final volume was adjusted to 500 μL using NER buffer. The mixture was incubated in rocker at 4 centigrade degrees for 24 h. The samples were placed on a magnetic stand and washed with ice cold PBS three times followed by one wash with NER buffer. The beads were resuspended in 16 μL PBS and 4 μL 5x SDS sample buffer, boiled at 70 °C for 10 min and the proteins were separated by SDS-PAGE. The gel was visulized by silver staining and identified by mass spectrometry (MS) (H. Wayen Biotechnology, Shanghai). The silver-stained gel was treated with decolorization, reductive alkylation, enzymolysis and desalination in order before MS examination. The samples were then isolation by EASY-nLC 1000 (Thermo Scientific, USA), examined by Orbitrap Fusion Lumos (Thermo Scientific, USA) and analyzed by Proteome Discoverer 2.4 software (Sequent HT), (Thermo Scientific, USA) for MS.
## Dual-luciferase activity assay
KLF10 promoter were first cloned into the dual-luciferase reporter plasmid vector and then was co-transfected with pCMV-ZBTB7A or negative control (Shanghai Genechem Co., Ltd) into 293 T cells. Dual-luciferase activity was measured using the Dual-Glo Luciferase Assay System (E2920, Promega).
## Statistical analysis
Data were expressed as means ± standard error of the mean. Two-tailed unpaired t-tests, Two-Way ANOVA, correlation and linear regression were performed using Microsoft Excel for Mac and GraphPad Prism 9.0 software for Mac. ns p ≥ 0.05, *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$, ****$p \leq 0.0001$ vs control group at the same experimental conditions.
## Supplementary information
Supplementary Figures Supplementary Mass Spectrometry Data Original Data File The online version contains supplementary material available at 10.1038/s41420-023-01381-6.
## References
1. Susantitaphong P, Cruz DN, Cerda J, Abulfaraj M, Alqahtani F, Koulouridis I. **World incidence of AKI: a meta-analysis**. *Clin J Am Soc Nephrol* (2013) **8** 1482-93. DOI: 10.2215/CJN.00710113
2. Rottenberg S, Disler C, Perego P. **The rediscovery of platinum-based cancer therapy**. *Nat Rev Cancer* (2021) **21** 37-50. DOI: 10.1038/s41568-020-00308-y
3. Wang S, Xie J, Li J, Liu F, Wu X, Wang Z. **Cisplatin suppresses the growth and proliferation of breast and cervical cancer cell lines by inhibiting integrin β5-mediated glycolysis**. *Am J Cancer Res* (2016) **6** 1108-17. PMID: 27294003
4. Li Z, Zhang P, Ma Q, Wang D, Zhou T. **Cisplatin-based chemoradiotherapy with 5-fluorouracil or pemetrexed in patients with locally advanced, unresectable esophageal squamous cell carcinoma: a retrospective analysis**. *Mol Clin Oncol* (2017) **6** 743-7. DOI: 10.3892/mco.2017.1222
5. Hussain SA, Palmer DH, Lloyd B, Collins SI, Barton D, Ansari J. **A study of split-dose cisplatin-based neo-adjuvant chemotherapy in muscle-invasive bladder cancer**. *Oncol Lett* (2012) **3** 855-9. PMID: 22741006
6. Chan BA, Coward JI. **Chemotherapy advances in small-cell lung cancer**. *J Thorac Dis* (2013) **5** S565-78. PMID: 24163749
7. Einhorn LH, Donohue J. **Cis-diamminedichloroplatinum, vinblastine, and bleomycin combination chemotherapy in disseminated testicular cancer**. *Ann Intern Med* (1977) **87** 293-8. DOI: 10.7326/0003-4819-87-3-293
8. Bennis Y, Savry A, Rocca M, Gauthier-Villano L, Pisano P, Pourroy B. **Cisplatin dose adjustment in patients with renal impairment, which recommendations should we follow?**. *Int J Clin Pharm* (2014) **36** 420-9. DOI: 10.1007/s11096-013-9912-7
9. Miller RP, Tadagavadi RK, Ramesh G, Reeves WB. **Mechanisms of Cisplatin nephrotoxicity**. *Toxins (Basel)* (2010) **2** 2490-518. DOI: 10.3390/toxins2112490
10. Yang L, Brooks CR, Xiao S, Sabbisetti V, Yeung MY, Hsiao LL. **KIM-1-mediated phagocytosis reduces acute injury to the kidney**. *J Clin Investig* (2015) **125** 1620-36. DOI: 10.1172/JCI75417
11. Rinkevich Y, Montoro DT, Contreras-Trujillo H, Harari-Steinberg O, Newman AM, Tsai JM. **In vivo clonal analysis reveals lineage-restricted progenitor characteristics in mammalian kidney development, maintenance, and regeneration**. *Cell Rep* (2014) **7** 1270-83. DOI: 10.1016/j.celrep.2014.04.018
12. Humphreys BD, Valerius MT, Kobayashi A, Mugford JW, Soeung S, Duffield JS. **Intrinsic epithelial cells repair the kidney after injury**. *Cell Stem Cell* (2008) **2** 284-91. DOI: 10.1016/j.stem.2008.01.014
13. Bonventre JV. **Dedifferentiation and proliferation of surviving epithelial cells in acute renal failure**. *J Am Soc Nephrology: JASN* (2003) **14** S55-61. DOI: 10.1097/01.ASN.0000067652.51441.21
14. Chang-Panesso M, Kadyrov FF, Lalli M, Wu H, Ikeda S, Kefaloyianni E. **FOXM1 drives proximal tubule proliferation during repair from acute ischemic kidney injury**. *J Clin Investig* (2019) **129** 5501-17. DOI: 10.1172/JCI125519
15. Sagrinati C, Netti GS, Mazzinghi B, Lazzeri E, Liotta F, Frosali F. **Isolation and characterization of multipotent progenitor cells from the Bowman’s capsule of adult human kidneys**. *J Am Soc Nephrology: JASN* (2006) **17** 2443-56. DOI: 10.1681/ASN.2006010089
16. Lasagni L, Romagnani P. **Glomerular epithelial stem cells: the good, the bad, and the ugly**. *J Am Soc Nephrology: JASN* (2010) **21** 1612-9. DOI: 10.1681/ASN.2010010048
17. Ronconi E, Sagrinati C, Angelotti ML, Lazzeri E, Mazzinghi B, Ballerini L. **Regeneration of glomerular podocytes by human renal progenitors**. *J Am Soc Nephrology: JASN* (2009) **20** 322-32. DOI: 10.1681/ASN.2008070709
18. Gao C, Zhang L, Chen E, Zhang W. **Aqp2(+) progenitor cells maintain and repair distal renal segments**. *J Am Soc Nephrology: JASN* (2022) **33** 1357-76. DOI: 10.1681/ASN.2021081105
19. Lindgren D, Boström AK, Nilsson K, Hansson J, Sjölund J, Möller C. **Isolation and characterization of progenitor-like cells from human renal proximal tubules**. *Am J Pathol* (2011) **178** 828-37. DOI: 10.1016/j.ajpath.2010.10.026
20. Angelotti ML, Ronconi E, Ballerini L, Peired A, Mazzinghi B, Sagrinati C. **Characterization of renal progenitors committed toward tubular lineage and their regenerative potential in renal tubular injury**. *Stem Cells* (2012) **30** 1714-25. DOI: 10.1002/stem.1130
21. Lazzeri E, Angelotti ML, Peired A, Conte C, Marschner JA, Maggi L. **Endocycle-related tubular cell hypertrophy and progenitor proliferation recover renal function after acute kidney injury**. *Nat Commun* (2018) **9** 1344. DOI: 10.1038/s41467-018-03753-4
22. Kumar S, Liu J, Pang P, Krautzberger AM, Reginensi A, Akiyama H. **Sox9 activation highlights a cellular pathway of renal repair in the acutely injured mammalian kidney**. *Cell Rep* (2015) **12** 1325-38. DOI: 10.1016/j.celrep.2015.07.034
23. Chen J, Chen JK, Harris RC. **Deletion of the epidermal growth factor receptor in renal proximal tubule epithelial cells delays recovery from acute kidney injury**. *Kidney Int* (2012) **82** 45-52. DOI: 10.1038/ki.2012.43
24. Chen J, You H, Li Y, Xu Y, He Q, Harris RC. **EGF receptor-dependent YAP activation is important for renal recovery from AKI**. *J Am Soc Nephrology: JASN* (2018) **29** 2372-85. DOI: 10.1681/ASN.2017121272
25. Schunk SJ, Floege J, Fliser D, Speer T. **WNT-β-catenin signalling - a versatile player in kidney injury and repair**. *Nat Rev Nephrol* (2021) **17** 172-84. DOI: 10.1038/s41581-020-00343-w
26. Memon A, Lee WK. **KLF10 as a tumor suppressor gene and its TGF-β signaling**. *Cancers (Basel)* (2018) **10** 161. DOI: 10.3390/cancers10060161
27. Subramaniam M, Harris SA, Oursler MJ, Rasmussen K, Riggs BL, Spelsberg TC. **Identification of a novel TGF-beta-regulated gene encoding a putative zinc finger protein in human osteoblasts**. *Nucleic Acids Res* (1995) **23** 4907-12. DOI: 10.1093/nar/23.23.4907
28. Xiao Y, Zheng S, Duan N, Li X, Wen J. **MicroRNA-26b-5p alleviates cerebral ischemia-reperfusion injury in rats via inhibiting the N-myc/PTEN axis by downregulating KLF10 expression**. *Hum Exp Toxicol* (2021) **40** 1250-62. DOI: 10.1177/0960327121991899
29. Xu YL, Zhang MH, Guo W, Xue Y, Du X, Zhang T. **MicroRNA-19 restores vascular endothelial cell function in lower limb ischemia-reperfusion injury through the KLF10-dependent TGF-β1/Smad signaling pathway in rats**. *J Cell Biochem* (2018) **119** 9303-15. DOI: 10.1002/jcb.27207
30. Lin CL, Hsu YC, Huang YT, Shih YH, Wang CJ, Chiang WC. **A KDM6A-KLF10 reinforcing feedback mechanism aggravates diabetic podocyte dysfunction**. *EMBO Mol Med* (2019) **11** e9828. DOI: 10.15252/emmm.201809828
31. Flechner SM, Kurian SM, Head SR, Sharp SM, Whisenant TC, Zhang J. **Kidney transplant rejection and tissue injury by gene profiling of biopsies and peripheral blood lymphocytes**. *Am J Transpl* (2004) **4** 1475-89. DOI: 10.1111/j.1600-6143.2004.00526.x
32. Sakaue-Sawano A, Kurokawa H, Morimura T, Hanyu A, Hama H, Osawa H. **Visualizing spatiotemporal dynamics of multicellular cell-cycle progression**. *Cell* (2008) **132** 487-98. DOI: 10.1016/j.cell.2007.12.033
33. 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** W98-102. DOI: 10.1093/nar/gkx247
34. Perry JM, He XC, Sugimura R, Grindley JC, Haug JS, Ding S. **Cooperation between both Wnt/{beta}-catenin and PTEN/PI3K/Akt signaling promotes primitive hematopoietic stem cell self-renewal and expansion**. *Genes Dev* (2011) **25** 1928-42. DOI: 10.1101/gad.17421911
35. Shalova IN, Lim JY, Chittezhath M, Zinkernagel AS, Beasley F, Hernández-Jiménez E. **Human monocytes undergo functional re-programming during sepsis mediated by hypoxia-inducible factor-1α**. *Immunity* (2015) **42** 484-98. DOI: 10.1016/j.immuni.2015.02.001
36. Jiang B, Wu X, Meng F, Si L, Cao S, Dong Y. **Progerin modulates the IGF-1R/Akt signaling involved in aging**. *Sci Adv* (2022) **8** eabo0322. DOI: 10.1126/sciadv.abo0322
37. Hu H, Miao YR, Jia LH, Yu QY, Zhang Q, Guo AY. **AnimalTFDB 3.0: a comprehensive resource for annotation and prediction of animal transcription factors**. *Nucleic Acids Res* (2019) **47** D33-d8. DOI: 10.1093/nar/gky822
38. Kent WJ, Sugnet CW, Furey TS, Roskin KM, Pringle TH, Zahler AM. **The human genome browser at UCSC**. *Genome Res* (2002) **12** 996-1006. DOI: 10.1101/gr.229102
39. Saint-Mezard P, Berthier CC, Zhang H, Hertig A, Kaiser S, Schumacher M. **Analysis of independent microarray datasets of renal biopsies identifies a robust transcript signature of acute allograft rejection**. *Transpl Int* (2009) **22** 293-302. DOI: 10.1111/j.1432-2277.2008.00790.x
40. Israni AK, Salkowski N, Gustafson S, Snyder JJ, Friedewald JJ, Formica RN. **New national allocation policy for deceased donor kidneys in the United States and possible effect on patient outcomes**. *J Am Soc Nephrology: JASN* (2014) **25** 1842-8. DOI: 10.1681/ASN.2013070784
41. Wang G, Lunardi A, Zhang J, Chen Z, Ala U, Webster KA. **Zbtb7a suppresses prostate cancer through repression of a Sox9-dependent pathway for cellular senescence bypass and tumor invasion**. *Nat Genet* (2013) **45** 739-46. DOI: 10.1038/ng.2654
42. Liu F, Tang F, Lan J, Jiao W, Si Y, Lu W. **Stable knockdown of ZBTB7A promotes cell proliferation and progression in nasopharyngeal carcinoma**. *Tumori* (2018) **104** 37-42. DOI: 10.5301/tj.5000706
43. Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O. **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
|
---
title: Reprogramming of palmitic acid induced by dephosphorylation of ACOX1 promotes
β-catenin palmitoylation to drive colorectal cancer progression
authors:
- Qiang Zhang
- Xiaoya Yang
- Jinjie Wu
- Shubiao Ye
- Junli Gong
- Wai Ming Cheng
- Zhanhao Luo
- Jing Yu
- Yugeng Liu
- Wanyi Zeng
- Chen Liu
- Zhizhong Xiong
- Yuan Chen
- Zhen He
- Ping Lan
journal: Cell Discovery
year: 2023
pmcid: PMC9988979
doi: 10.1038/s41421-022-00515-x
license: CC BY 4.0
---
# Reprogramming of palmitic acid induced by dephosphorylation of ACOX1 promotes β-catenin palmitoylation to drive colorectal cancer progression
## Abstract
Metabolic reprogramming is a hallmark of cancer. However, it is not well known how metabolism affects cancer progression. We identified that metabolic enzyme acyl-CoA oxidase 1 (ACOX1) suppresses colorectal cancer (CRC) progression by regulating palmitic acid (PA) reprogramming. ACOX1 is highly downregulated in CRC, which predicts poor clinical outcome in CRC patients. Functionally, ACOX1 depletion promotes CRC cell proliferation in vitro and colorectal tumorigenesis in mouse models, whereas ACOX1 overexpression inhibits patient-derived xenograft growth. Mechanistically, DUSP14 dephosphorylates ACOX1 at serine 26, promoting its polyubiquitination and proteasomal degradation, thereby leading to an increase of the ACOX1 substrate PA. Accumulated PA promotes β-catenin cysteine 466 palmitoylation, which inhibits CK1- and GSK3-directed phosphorylation of β-catenin and subsequent β-Trcp-mediated proteasomal degradation. In return, stabilized β-catenin directly represses ACOX1 transcription and indirectly activates DUSP14 transcription by upregulating c-Myc, a typical target of β-catenin. Finally, we confirmed that the DUSP14-ACOX1-PA-β-catenin axis is dysregulated in clinical CRC samples. Together, these results identify ACOX1 as a tumor suppressor, the downregulation of which increases PA-mediated β-catenin palmitoylation and stabilization and hyperactivates β-catenin signaling thus promoting CRC progression. Particularly, targeting β-catenin palmitoylation by 2-bromopalmitate (2-BP) can efficiently inhibit β-catenin-dependent tumor growth in vivo, and pharmacological inhibition of DUSP14-ACOX1-β-catenin axis by Nu-7441 reduced the viability of CRC cells. Our results reveal an unexpected role of PA reprogramming induced by dephosphorylation of ACOX1 in activating β-catenin signaling and promoting cancer progression, and propose the inhibition of the dephosphorylation of ACOX1 by DUSP14 or β-catenin palmitoylation as a viable option for CRC treatment.
## Introduction
Metabolic reprogramming is critical for malignant transformation and tumor initiation and progression1. Alterations of intracellular and extracellular metabolites caused by metabolic reprogramming have profound effects on gene expression, protein modification, cellular differentiation, and the tumor microenvironment2–5. Metabolic enzyme acyl-CoA oxidase 1 (ACOX1), a rate-limiting enzyme in peroxisomal fatty acid β-oxidation, catalyzes acyl-CoA conversion to enoyl-CoA6. ACOX1 preferentially oxidizes long or very long straight-chain fatty acids6–9, while the related enzymes ACOX2 and ACOX3 catabolize branched-chain fatty acids and intermediates involved in bile acid synthesis10. Knockout of ACOX1 promotes hepatocellular carcinoma in mice11,12, and overexpression of ACOX1 inhibits oral cancer progression13. In addition, ACOX1 acts as a target gene of mir-15B-5p to inhibit tumor cell metastasis14. These studies indicate the inhibitory role of ACOX1 in cancer11–15. However, the role of metabolic reprogramming caused by dysregulation of the metabolic enzyme ACOX1’s post-translational modification in colorectal cancer (CRC) remains elusive.
Palmitic acid (PA), an ACOX1 substrate7 and a dominant fatty acid in a high-fat diet16, has been shown to produce energy and regulate intracellular signaling molecules involved in the development of cancer17. Previous studies have identified that PA promotes metastasis in melanoma, breast cancer, and gastric cancer in a CD36-dependent manner18,19, and also promotes the growth of prostate cancer by activating STAT3 signaling20. Recent research has revealed that dietary metabolite PA alters transcriptional and chromatin programs by modulating H3K4me3 in oral carcinomas and melanoma21. Furthermore, PA can modify cysteine residues in a process termed palmitoylation22–24. Increasing evidence suggests that palmitoylation of proteins (such as PDL1, GULT1, STAT3, and IFNGR1) affects protein functions and tumor progression24–27. Therefore, whether ACOX1-mediated PA reprogramming affects tumor progression by regulating protein palmitoylation remains unknown.
β-catenin signaling is essential for maintaining cell homeostasis and embryonic development and is related to tumor cell proliferation, apoptosis, invasion, stemness, and chemotherapy resistance28,29. Studies have shown that β-catenin signaling is abnormally activated in more than $90\%$ of patients with CRC30. Post-translational modifications (such as phosphorylation, ubiquitination, acetylation, and glycosylation) of β-catenin have been demonstrated to regulate β-catenin signaling31–34. In addition, emerging evidence indicates that PA complements the β-catenin signaling activity19. However, whether β-catenin could be palmitoylated by PA remains unclear.
Here, we demonstrate that ACOX1 is significantly underexpressed in CRC through a systematic bioinformatics screen and propose that reprogramming of PA induced by dysregulation of ACOX1 post-translational modification promotes CRC progression by activating β-catenin signaling via PA-mediated β-catenin palmitoylation and stabilization.
## ACOX1 is downregulated and associated with progression in CRC
To identify metabolism-related genes playing crucial roles in colorectal tumorigenesis, the transcriptional levels of 2752 metabolism-related genes35 were analyzed in at least 1000 CRCs from various datasets, including The Cancer Genome Atlas (TCGA) CRC RNA-SeqV2, TCGA CRC RNA-Seq, and Gene Expression Omnibus (GEO) (Supplementary Table S1). Additionally, protein levels of these metabolism-related genes were also analyzed in at least 100 CRCs from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) dataset and our quantitative mass spectrometry (MS) of clinical samples (Supplementary Table S1). Eleven metabolism-related genes that were significantly altered in CRCs, were selected by overlapping analysis (Fig. 1a; Supplementary Table S1). Specifically, ACOX1, the only metabolic rate-limiting enzyme, was identified for subsequent analysis. Fig. 1ACOX1 is downregulated and associated with progression in CRC.a Venn diagram exhibiting 11 differentially expressed genes (DEGs) in CRCs based on transcription levels and protein levels in the TCGA, GSE25070, CPTAC datasets, and our protein quantitative MS. b Analysis of ACOX1 expression in adjacent normal tissues, primary tumor tissues and metastatic tumors from gene chip data. c Analysis of ACOX1 expression in adjacent normal tissues, polyp tissues, and tumor tissues from GSE68468. d Unpaired and paired analysis of ACOX1 expression in adjacent normal tissues versus primary tumor samples from the Sixth Affiliated Hospital of Sun Yat-sen University. e Analysis of ACOX1 protein expression in our protein quantitative MS. f–h Expression of ACOX1 protein in control colon tissues (Ctrl) and tumor samples from APCMin/+/DSS-inducted CRC mice (APCMin/+/DSS) (f) or AOM/DSS-inducted CRC mice (AOM/DSS) (g) or AOM-inducted CRC mice (AOM) (h) analyzed by immunoblotting (left) and quantified by densitometry (right). i, j Kaplan–Meier overall survival curves of human CRC patients with low versus high ACOX1 mRNA or protein expression, based on CRC TMA (i), and TCGA RNA-SeqV2 (j). k ACOX1 expression is an independent prognostic factor for poor survival. Forest plot showing univariate (left) and multivariate (right) Cox regression analysis of different clinical parameters for CRC patients in TMA. HR, hazard ratio; CI, confidence interval. l, m Analysis of ACOX1 protein in patients with different T stages (l) and lymph node metastases (m) in CRC TMA. n Pie charts showing the relationship between clinicopathologic factors and ACOX1 protein expression in CRC TMA. Data were analyzed using unpaired Student’s t-test (b–d, f–h, l, m), paired Student’s t-test (e), log-rank test (i, j) or χ2 test (n). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; n, number of patient samples.
Analysis of the BioGPS Gene Expression Atlas indicated that the transcriptional level of ACOX1 (not ACOX2 or ACOX3) was notably downregulated in CRCs (Supplementary Fig. S1a). Similarly, analysis of TCGA and GEO datasets showed a significant downregulation of ACOX1 mRNA in CRCs (Supplementary Fig. S1b). Additional datasets, such as public gene chip data36, TCGA and GEO databases revealed that ACOX1 mRNA was negatively correlated with advanced disease (Fig. 1b, c; Supplementary Fig. S1c, d). Consistently, decreased ACOX1 mRNA was also observed in early-stage CRC (TNM, Stage I, and II) (Fig. 1d; Supplementary Table S2). Importantly, the classification of CRC intrinsic-consensus molecular subtypes (iCMSs)37 based on TCGA transcriptomics showed that ACOX1 expression was significantly dysregulated in iCMS2 tumor samples, where β-catenin signaling is hyperactivated, relative to iCMS3 tumor samples (Supplementary Fig. S1e–g). In addition to the transcriptomic level, a fuller analysis showed that ACOX1 protein was also markedly downregulated in CRCs (Fig. 1e; Supplementary Fig. S1h). Immunohistochemistry (IHC) analysis of our clinical samples also revealed decreased ACOX1 protein in CRCs (Supplementary Fig. S1i), further validating the result in the Human Protein Atlas (HPA) database (Supplementary Fig. S1j). Furthermore, we also found a decrease in ACOX1 protein in azoxymethane/dextran sulfate sodium (AOM/DSS)38, DSS (APCMin/+/DSS)39 and AOM40-induced mouse CRC models (Fig. 1f–h; Supplementary Fig. S1k–p). Given the low mutation frequency of ACOX1 alleles in CRC patients (Supplementary Fig. S1q), we suggested that ACOX1 downregulation is the main cause of ACOX1 inactivation in CRC. These results confirmed that ACOX1 is poorly expressed at the transcriptional and protein levels in CRC.
Next, we evaluated our CRC tissue microarray (TMA) containing 192 CRC tissues by IHC (Supplementary Table S3), and observed that CRC patients with low levels of ACOX1 exhibited poor survival (Fig. 1i). This observation was validated in TCGA, GEO, and Vasaikar’s CPTAC41 datasets (Fig. 1j; Supplementary Fig. S2a–c). Univariate and multivariate Cox regression analysis was carried out to assess the importance of ACOX1 expression for CRC prognosis together with other risk factors including age, TNM stage, or tumor differentiation. The results showed that ACOX1 expression was an independent prognostic factor for CRC (Fig. 1k; Supplementary Fig. S2d). Moreover, ACOX1 protein expression was also significantly associated with the clinical stage, T stage, and lymph node metastases (N) of CRC (Fig. 1l–n). Collectively, these results demonstrated that ACOX1 expression is negatively correlated with the progression of CRC.
## ACOX1 depletion promotes colorectal tumorigenesis
To define whether ACOX1 is a tumor suppressor in CRC, we ectopically expressed or silenced ACOX1 using Flag-tagged ACOX1 or ACOX1-specific short hairpin RNAs (shRNAs) in CRC cell lines (HCT15, RKO, HCT8, HCT116, and SW620), respectively (Supplementary Fig. S3a, b). We observed that depletion of ACOX1 promoted CRC cell proliferation and colony formation (Fig. 2a, b), while overexpression of ACOX1 inhibited CRC cell proliferation and migration (Supplementary Fig. S3b–d).Fig. 2Depletion of ACOX1 promotes CRC cell proliferation in vitro and colorectal tumorigenesis in mice.a Enhanced CRC cell viability by ACOX1 depletion. Cell viability of ACOX1-depleted CRC cells (HCT15, RKO, SW620, HCT8, and HCT15) was analyzed for CCK-8. b Colony formation of RKO and HCT15 cells stably expressing the indicated vectors (left), and bar graphs showing the relative colony numbers (right). c–f Representative colonoscopy (c), macroscopic morphologies (d), tumor numbers (e), and tumor sizes (f) of APCMin/+ mice in control and ACOX1-depleted groups. g–j Representative colonoscopy (g), macroscopic morphologies (h), tumor numbers (i) and tumor sizes (j) of C57BL/6 mice in control and ACOX1-depleted groups. k Schematic diagram showing the experimental design for PDX model. l Suppression of PDX growth by ACOX1 overexpression. BALB/c nude mice were subcutaneously transplanted with PDXs into flanks and injected with lentivirus expressing Ctrl or Flag-ACOX1 every 2 days for 18 days. At day 21, PDXs were collected, and relative tumor growth was calculated (ratio of volume: day 21/day 0). m Decreased tumor cell proliferation by ACOX1 overexpression in PDXs. H&E and IHC for ACOX1 and Ki67. Data were analyzed using unpaired Student’s t-test (a, b, e, f, i, j, l). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; n, number of mouse samples.
To further confirm that ACOX1 inhibits colorectal tumorigenesis in vivo, we built two CRC mouse models: AOM/DSS and APCMin/+/DSS (Supplementary Fig. S3e, f). As expected, mice with ACOX1 depletion presented significantly more tumors, larger tumors, and markedly more histologic dysplasia (Fig. 2c–j; Supplementary Fig. S3g–j). To better elucidate the inhibitory effect of ACOX1 in CRC, we collected clinical CRC samples and constructed patient-derived xenograft (PDX) models (Fig. 2k). Consistent with the result above, ACOX1 overexpression by lentivirus inhibited tumor growth in PDX models (Fig. 2l, m; Supplementary Fig. S3k, l). These findings suggest that ACOX1 inhibits colorectal tumorigenesis in vitro and in vivo.
## DUSP14 promotes ACOX1 degradation in a ubiquitination-dependent manner
To uncover the functional effectors regulating ACOX1, we expressed Flag-tagged ACOX1 in HEK293T cells, immunoprecipitated the epitope-tagged protein, and analyzed the precipitate by MS. Combined with Huttlin’s MS data (thousands of cell lines (HCTT16 and 293 T) with each expressing a tagged version of a protein were lysed and immunoprecipitated, followed by MS to identify their biophysically interacting proteins)42, we identified DUSP14 as a candidate ACOX1 interactor (Fig. 3a; Supplementary Fig. S4a), and the endogenous interaction was further validated by Co-immunoprecipitation (Co-IP) assays in HCT15 and RKO cells (Fig. 3b; Supplementary Fig. S4b). Subsequent Co-IP assays revealed that the N-terminal domain of ACOX1 was responsible for its binding to DUSP14, while the C-terminal domain of DUSP14 was required for its interaction with ACOX1 (Fig. 3c, d). A time-course analysis following a cycloheximide block showed that depletion of DUSP14 significantly extended the half-life of endogenous ACOX1 in HCT15 and RKO cells (Fig. 3e; Supplementary Fig. S4c).Fig. 3DUSP14 promotes ACOX1 degradation in a ubiquitination-dependent manner.a Venn diagram exhibiting DUSP14 as an ACOX1 interactor. b Endogenous interaction of ACOX1 and DUSP14 in HCT15 cells. HCT15 cells were treated with MG132 (20 μM) for 6 h before harvest and cell lysates were analyzed for Co-IP. c, d ACOX1–DUSP14 interaction via N-terminal domain and CCD domain. Generation of ACOX1-mutant constructs (c) and DUSP14-mutant constructs (d). HEK293T cell lysates transfected with indicated plasmids analyzed for Co-IP. e Time-course analysis of ACOX1 protein levels in DUSP14-depleted HCT15 cells (left). ACOX1 proteins quantified by densitometry, with β-actin as a normalizer (right). f Increased ACOX1 polyubiquitination by DUSP14 WT but not DUSP14 Dead. Myc-Ub was co-transfected with Flag-ACOX1 and HA-DUSP14 (WT or Dead) into HEK293T cells, and the cell lysates were subjected to immunoprecipitation. g DUSP14 mediates K48-linked ubiquitination of ACOX1. Flag-ACOX1 was co-transfected with HA-DUSP14 and Myc-Ub (WT, K6O, K11O, K27O, K29O, K33O, or K63O) into HEK293T cells, and the cell lysates were subjected to immunoprecipitation. KXO represents substitutions of arginine for all lysine resides except the lysine at X position. h DUSP14 mediates ubiquitination of ACOX1 at K643. Myc-Ub was co-transfected with HA-DUSP14 and Flag-ACOX1 (WT, K29R, K241R, K$\frac{255}{260}$ R, K446R, or K63R) into HEK293T cells, and the cell lysates were subjected to immunoprecipitation. i Alignment of lysine 643 and adjacent amino acids of ACOX1 among multiple species. Data were analyzed using unpaired Student’s t-test (e). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01.$
Ubiquitin-mediated degradation is critical for protein stability43,44. Therefore, to explore whether DUSP14 promotes ACOX1 degradation via the ubiquitin-proteasome system, we transfected HEK293T cells with Myc-Ub, ACOX1, and DUSP14 wild type (WT) or DUSP14 mutant (DUSP14 Dead) with mutation of cysteine 111 to serine45, which damaged the phosphatase activity. Ubiquitination assays showed that DUSP14 WT (not DUSP14 Dead) overexpression markedly increased the polyubiquitination of ACOX1 (Fig. 3f), and DUSP14 promoted K48-linked ubiquitination of ACOX1, but not other position-linked ubiquitination of ACOX1 (Fig. 3g). To further explore the ubiquitination site(s) of ACOX1 mediated by DUSP14, we screened 6 candidate sites (K29, K241, K255, K260, K446, and K643) in the Phospho-Site Plus database and identified that DUSP14-mediated K48-linked ubiquitination of ACOX1 at K643, an evolutionally conserved residue among multiple species (Fig. 3h, i).
To determine the functional role of DUSP14 in CRC, we re-analyzed the public databases mentioned earlier and found that DUSP14 was highly expressed in CRCs (Supplementary Fig. S5a, b) and was positively correlated with advanced disease (Supplementary Fig. S5c–e). Further analysis showed that upregulation of DUSP14 mRNA may be the result of DUSP14 copy number amplification (Supplementary Fig. S5f, g). Meanwhile, DUSP14 mRNA upregulation strongly correlated with poor overall survival in CRC patients (Supplementary Fig. S5h–j). Collectively, all of the findings suggest that DUSP14 is highly expressed in CRC, thus promoting ACOX1 degradation via the ubiquitin-proteasome system.
## Dephosphorylation of ACOX1 at S26 by DUSP14 is critical for CRC growth
Considering that DUSP14 is a multitarget phosphatase45 and DUSP14 regulates ACOX1 stability, we postulated that DUSP14 promotes ubiquitination and degradation of ACOX1 via dephosphorylation. To prove this, immunoprecipitation assays were performed, which revealed that DUSP14 specifically decreased ACOX1 serine phosphorylation rather than threonine phosphorylation and tyrosine phosphorylation (Supplementary Fig. S6a), implicating that DUSP14 dephosphorylates ACOX1 at serine residue(s). Next, we identified three serine phosphorylation sites (serine 26, serine 126, and serine 127) by MS analysis (Supplementary Fig. S6b). Further analysis showed that DUSP14 failed to promote the degradation and serine dephosphorylation of ACOX1 S26A mutant (Fig. 4a; Supplementary Fig. S6c). ACOX1 S26 is evolutionally conserved among vertebrates (Fig. 4b). Moreover, the phosphorylation-mimic mutant ACOX1 S26D exhibited an extended half-life and decreased ubiquitination levels, whereas the S26A mutant exhibited an opposite effect (Supplementary Fig. S6d, e). Additional structural analysis and glutaraldehyde cross-linking experiments revealed that DUSP14-mediated ACOX1 dephosphorylation did not affect the formation of ACOX1 homodimerization (Supplementary Fig. S6f, g). These studies indicated that dephosphorylation of ACOX1 at S26 by DUSP14 is a critical determinant of the ACOX1 protein stability. Fig. 4Dephosphorylation of ACOX1 by DUSP14 promotes CRC growth.a Dephosphorylation of ACOX1 S26 by DUSP14. Flag-ACOX1 (WT or S26A) was co-transfected with HA-DUSP14 (WT or Dead) into HEK293T cells, and the cell lysates were subjected to immunoprecipitation. b Alignment of serine 26 and adjacent amino acids of ACOX1 among multiple species. c Schematic diagram showing the experimental design for mouse CRC model. d–g Representative macroscopic morphologies (d), tumor numbers (e), tumor sizes (f) and H&E, Ki67 staining (g) of mice in control and ACOX1 (WT, S26A or S26D) groups. Data were analyzed using unpaired Student’s t-test (e, f). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01$; n, number of mouse samples.
Next, the lentiviruses encoding ACOX1 WT, ACOX1 S26A, or ACOX1 S26D were delivered intraperitoneally to DSS-treated APCMin/+ mice (Fig. 4c). Mice treated with ACOX1 WT or ACOX1 S26D exhibited fewer tumors, smaller tumors, and less histologic dysplasia (Fig. 4d–g; Supplementary Fig. S6h). In contrast, mice treated with ACOX1 S26A exhibited similar tumor burdens and pathological features as the controls (Fig. 4d–g; Supplementary Fig. S6h). These results show that DUSP14-mediated ACOX1 dephosphorylation is critical for CRC growth.
## ACOX1 depletion stabilizes β-catenin and enhances its transcriptional activity via PA
To explore the CRC-related cellular signaling regulated by ACOX1, gene set enrichment analysis (GSEA) was performed in the TCGA database. GSEA revealed that ACOX1 negatively correlated with Wnt signaling, but not other cancer-related signaling pathways (Supplementary Fig. S7a). Meanwhile, β-catenin target genes were upregulated in tumor tissues and metastasis samples with lower levels of ACOX1 in public databases (Fig. 5a, b; Supplementary Fig. S7b, c). Interestingly, ACOX1 depletion did not affect CTNNB1 (encoding β-catenin protein) mRNA levels but increased β-catenin abundance in HCT15 and RKO cells (Fig. 5c; Supplementary Fig. S7d, e). Depletion of ACOX1 markedly increased β-catenin target gene expression in HCT15 and RKO cells (Fig. 5d; Supplementary Fig. S7f). To further validate that ACOX1 inhibits CRC cell growth by impairing β-catenin-mediated target gene transcription, CCK-8 assays were performed in HCT15 and RKO cells. As expected, overexpression of ACOX1 inhibited CRC cell viability, which was rescued by β-catenin overexpression (Fig. 5e; Supplementary Fig. S7g). Conversely, ACOX1 depletion increased CRC cell viability, which was inhibited by iCRT14, a β-catenin transcriptional activity inhibitor that disrupts the binding of β-catenin to TCF (Fig. 5e; Supplementary Fig. S7g).Fig. 5Depletion of ACOX1 stabilizes β-catenin and enhances the transcriptional activity via PA.a Negative correlation between ACOX1 and β-catenin target genes. GSEA of β-catenin target gene set in the expression profiles of TCGA RNA-SeqV2 according to the expression of ACOX1. b High expression of β-catenin target genes in metastatic tissues. GSEA of β-catenin target gene set in the expression profiles of normal tissues versus metastatic tissues from GSE68468. c Expression of β-catenin and ACOX1 analyzed by immunoblotting. ACOX1-depleted or ACOX1-overexpressed HCT15 cell lysates were subjected to immunoblotting. d Increased β-catenin target gene expression by ACOX1 depletion. HCT15 cells stably expressing control shRNA or ACOX1 shRNA (shACOX1) were analyzed by RT-qPCR. e ACOX1-induced inhibition of HCT15 cell viability rescued by β-catenin overexpression. HCT15 cells stably expressing Flag-ACOX1 or Flag-β-catenin were cultured for 5 days and counted by CCK-8 (left). Suppression of shACOX1-induced cell hyper-viability by β-catenin inhibition. HCT15 cells (shCtrl or shACOX1 stably expressed) were treated with iCRT14 (100 μM) for 5 days and counted by CCK-8 (right). f Promoted β-catenin polyubiquitination by ACOX1 WT and S26D mutant but not S26A mutant. Myc-Ub was co-transfected with Flag-β-catenin and HA-ACOX1 (WT, S26A, or S26D) into HEK293T cells, and the cell lysates were subjected to immunoprecipitation. g Promoted human colonic organoid growth by PA treatment. Organoids were derived from human intestinal normal tissues, treated with PA (300 μM), and assessed by diameter size. Scale bars, 100 μm. h ACOX1-mediated β-catenin inhibition rescued by PA treatment. HCT15 cells transfected with Flag-ACOX1 plasmid were treated with MG132 (20 μM) for 6 h or PA (20 μM, or 100 μM) for 24 h as indicated and cell lysates were subjected to immunoblotting. i *Decreased endogenous* β-catenin polyubiquitination by PA treatment. HEK293T cells were treated with PA (50 μM, 100 μM, or 200 μM) for 24 h, and the cell lysates were subjected to immunoprecipitation. j Expression of β-catenin phosphorylation. HEK293T cells were treated with PA (100 μM) for 24 h and the cell lysates were subjected to immunoblotting. k Decreased interactions between β-catenin and CK1/GSK3β/β-Trcp by PA treatment. HEK293T cells transfected with Flag-CK1 and Flag-GSK3β were treated with MG132 (20 μM) for 6 h or PA (100 μM) for 24 h and cell lysates were subjected to Co-IP. Data were analyzed using unpaired Student’s t-test (d, e, g). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$
β-catenin is prominently degraded in a ubiquitin-proteasome manner46. Thus, we speculated whether ACOX1 reduces the stability of β-catenin via promoting its polyubiquitination. As expected, both ACOX1 WT and ACOX S26D increased β-catenin polyubiquitination levels (Fig. 5f), whereas ACOX1 S26A failed to do so (Fig. 5f), supporting that DUSP14-mediated ACOX1 dephosphorylation is critical for colorectal tumorigenesis.
ACOX1 is a rate-limiting enzyme in peroxisomal fatty acid β-oxidation6. PA, a substrate of ACOX17, has been reported to promote tumor cell growth and migration18–20. Hence, we speculated that ACOX1 regulates β-catenin stability via PA. Overexpression of ACOX1 significantly decreased PA levels in both HCT15 and RKO cells (Supplementary Fig. S7h). Furthermore, PA markedly promoted human colonic organoid growth (Fig. 5g). Overexpression of ACOX1 decreased β-catenin abundance, which was rescued by PA treatment (Fig. 5h). In addition, proteomic analysis showed that PA promotes expression of β-catenin and its target genes in HCT15 cells (Supplementary Fig. S7i, j). These results suggest that ACOX1 depletion stabilizes β-catenin via PA.
To determine whether PA regulated ubiquitination of β-catenin, we performed ubiquitination assays in HEK293T cells, and the results suggested that PA decreased endogenous β-catenin polyubiquitination in a dose-dependent manner (Fig. 5i). Previous studies have demonstrated that phosphorylation of β-catenin at Ser45 by CK1 could trigger sequential phosphorylation of Thr41, Ser37, and Ser33 by GSK3 (preferentially by GSK3β)47,48, leading to the recognization of phosphorylated β-catenin by E3 ubiquitin ligase β-TrCP and subsequent degradation by the ubiquitin-proteasome system49. In addition, AKT phosphorylates β-catenin at Ser552 to promote β-catenin accumulation in both the cytosol and the nucleus and thus enhances its transcriptional activity50. To explore the mechanisms of β-catenin stabilization regulated by PA, immunoblotting analysis was performed, which revealed that PA significantly decreased β-catenin Ser33, Ser37, Thr41, and Ser45 phosphorylation. However, it failed to affect the Ser552 phosphorylation of β-catenin (Fig. 5j; Supplementary Fig. S7k), suggesting that PA inhibits CK1- and GSK3-mediated β-catenin phosphorylation. Co-IP assays revealed that PA suppressed the interactions between β-catenin and CK1, GSK3, and β-TrCP (Fig. 5k). Together, these results suggest that ACOX1 depletion stabilizes β-catenin and enhances its transcriptional activity via PA.
## PA-mediated β-catenin palmitoylation inhibits the ubiquitination of β-catenin
Previous studies demonstrated that PA is the substrate of protein palmitoylation22–24, and that protein palmitoylation can alter the protein–protein interaction24. Therefore, we hypothesized that β-catenin is palmitoylated, which subsequently inhibits its interactions with CK1/GSK3/β-TrCP. As expected, palmitoylation of endogenous β-catenin was confirmed in HEK293T cells (Supplementary Fig. S7l). Interestingly, endogenous palmitoylated β-catenin accounted for $31.7\%$ and $28.4\%$ of total β-catenin in HCT15 and RKO cells, respectively (Supplementary Fig. S7m). Importantly, palmitoylation of β-catenin was increased by PA but decreased by 2-bromopalmitate (2-BP), a palmitoylation inhibitor (Fig. 6a). Inhibition of palmitoylation by 2-BP also decreased β-catenin abundance (Fig. 6b), and promoted β-catenin polyubiquitination (Fig. 6c). These data reveal that β-catenin stabilization can be regulated through a novel palmitoylation modification. Fig. 6PA-mediated β-catenin palmitoylation inhibits the ubiquitination of β-catenin.a Increased and decreased β-catenin palmitoylation by PA and 2-BP, respectively. HEK293T cells transfected with Flag-β-catenin were treated with 2-BP (100 μM) for 6 h or PA (100 μM) for 24 h and cell lysates were subjected to immunoprecipitation. Palmitoylated β-catenin was detected with HRP-conjugated streptavidin antibody. b Decreased β-catenin abundance by 2-BP treatment. HCT15 and RKO cells were treated with 2-BP (100 μM) for 6 h and cell lysates were subjected to immunoblotting. c Increased β-catenin polyubiquitination by 2-BP treatment. HEK293T cells transfected with Flag-β-catenin and/or Myc-Ub were treated with 2-BP (100 μM) or DMSO for 6 h and cell lysates were subjected to immunoprecipitation. d Interactions of β-catenin and CK1/GSK3/β-TrCP were enhanced by 2-BP treatment. HEK293T cells were treated with 2-BP (100 μM) or DMSO for 6 h and cell lysates were subjected to immunoprecipitation. e Alignment of cysteine 466 and adjacent amino acids of β-catenin among multiple species. f Unalterated β-catenin C466A protein abundance by 2-BP treatment. HCT15 and RKO cells transfected with Flag-β-catenin C466A were treated with 2-BP (100 μM) for 6 h, and cell lysates were subjected to immunoblotting. g Abolished β-catenin palmitoylation by β-catenin C466A mutation. HEK293T cells were transfected with Flag-β-catenin or Flag-β-catenin C466A and cell lysates were subjected to immunoprecipitation. h Increased interactions of β-catenin and CK1/GSK3/β-TrCP by β-catenin C466A mutant. HEK293T cells were transfected with Flag-β-catenin or Flag-β-catenin C466A, and cell lysates were analyzed by Co-IP. i Unalterated β-catenin C466A poly-ubiquitination by 2-BP treatment. HEK293T cells transfected with Flag-β-catenin or Flag-β-catenin C466A were treated with 2-BP (100 μM) or DMSO for 6 h and cell lysates were subjected to immunoprecipitation. j Decreased β-catenin half-life by C466A mutation. Time-course analysis of β-catenin levels in Flag-β-catenin WT-, or C466A-overexpressed HEK293T cells (upper). β-catenin quantified by densitometry, with β-actin as a normalizer (lower). k Nude mice carrying HCT15 tumors were intraperitoneally injected with 2-BP, and tumor volume was evaluated. l The tumor weight in the subcutaneous xenograft model. Data were analyzed using unpaired Student’s t-test (j, k, l). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01.$
To clarify whether β-catenin palmitoylation affected the interactions between β-catenin and CK1/GSK3/β-TrCP, we performed Co-IP assays and found that 2-BP increased the β-catenin–CK1/GSK3/β-TrCP interactions (Fig. 6d). Next, we used the predictor Swiss-Palm51 and identified cysteine 466 as a candidate conservative palmitoylation site in β-catenin across species (Fig. 6e). Mutation of cysteine 466 to alanine substantially abolished β-catenin palmitoylation (Fig. 6f; Supplementary Fig. S7n), suggesting that this cysteine residue is the major palmitoylation site of β-catenin. In addition, this mutation rendered resistance to 2-BP-mediated β-catenin downregulation, promoted β-catenin–CK1/GSK3/β-TrCP interactions, increased β-catenin polyubiquitination, and shortened the half-life of β-catenin (Fig. 6g–j).
To investigate the effect of cysteine 466 palmitoylation of β-catenin on CRC cell growth, we constructed a series of stable cell lines for CCK-8 and colony formation assays. The results revealed that depletion of β-catenin inhibited CRC cell colony formation and proliferation, which were substantially rescued by the re-expression of β-catenin WT, but not its C466A mutant (Supplementary Fig. S7o–q). To further pharmacologically inhibit β-catenin palmitoylation in vivo, 2-BP was tested in the subcutaneous xenograft model. Intraperitoneal injection of 2-BP (40 mg/kg; one injection per day) in nude mice carrying HCT15 tumors inhibited tumor growth (Fig. 6k, l; Supplementary Fig. S7r). Consistently, 2-BP injection significantly decreased β-catenin palmitoylation and protein abundance in tumor tissues (Supplementary Fig. S7s). Together, these results demonstrate that PA-mediated β-catenin palmitoylation is essential for inhibiting the β-catenin–CK1/GSK3/β-TrCP interactions, thereby enhancing β-catenin stability, and that targeting β-catenin palmitoylation by 2-BP can efficiently suppress tumor growth.
## β-catenin directly suppresses ACOX1 transcription and indirectly activates DUSP14 transcription via c-Myc
To explore the cause of ACOX1 mRNA downregulation, we analyzed ACOX1 copy number; however, there was no difference between normal tissue and tumor samples (Supplementary Fig. S8a), suggesting that ACOX1 copy number alteration is not the cause of ACOX1 mRNA downregulation. PPARA has been considered to be the main transcription factor of ACOX152,53; however, normal tissues and tumor samples from the GEO datasets showed no difference in their expressions of PPARA mRNA (Supplementary Fig. S8b). Given the correlation between ACOX1 and β-catenin targets in CRC (Fig. 5a; Supplementary Fig. S7b, c), we assessed whether β-catenin regulates ACOX1 in CRC. Interestingly, iCRT14-treated CRC cells exhibited upregulation of ACOX1 transcripts and protein, but downregulation of DUSP14 transcripts and protein (Fig. 7a, b). However, iCRT14 did not appear to affect their transcripts or protein expression in normal human intestinal epithelial cells HIEC-6 (Fig. 7a, b), indicating that background-level β-catenin does not affect ACOX1 and DUSP14 transcription. Fig. 7β-catenin directly represses ACOX1 transcription and indirectly activates DUSP14 transcription via c-Myc.a, b Upregulated ACOX1 expression and downregulated DUSP14 and c-Myc expression by β-catenin inhibition in CRC cells, but not in HIEC-6 cells. Indicated cells were treated with iCRT14 (100 μM) for 24 h and analyzed by RT-qPCR (a) and immunoblotting (b). c, d Schematic presentation of TCF/LEFs-binding element on the ACOX1 locus (c) and c-Myc-binding sites on the DUSP14 locus (d). TBE, TCF/LEFs-binding element; RE, c-Myc-responsive element. Consensus sequence mutations are shown as TBE Mut and RE Mut. e Downregulated ACOX1 expression and upregulated DUSP14 expression by β-catenin overexpression. f Upregulated DUSP14 expression by c-Myc overexpression. HIEC-6 cells transfected with Flag-β-catenin (e) or Flag-c-Myc (f) were subjected to western blot analysis. g, h β-catenin occupancy on the ACOX1 promoter (g) and c-Myc occupancy on the DUSP14 promoter (h). HCT15 and RKO cells were analyzed by ChIP assays. i Assessed Luciferase reporter activities in the presence of exogenous β-catenin (left) and c-Myc (right) in HCT15 and RKO cells. j, k Downregulation of ACOX1 and upregulation of DUSP14 in APCMin/+ intestinal tumors. ACOX1 and DUSP14 expression was analyzed in normal small intestinal tissues and intestinal adenoma samples from APCMin/+ mouse (21 weeks of age) by RT-qPCR (j) and western blot analysis (k). Data were analyzed using unpaired Student’s t-test (a, i, j). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; n, number of mouse samples.
Motifmap website analysis identified the potential TCF/LEFs-binding elements (TBE; CTTTGA/TA/T) in the ACOX1 promoter regions (–5 to +2 kb) (Fig. 7c) and the putative c-Myc E-box response elements (RE; CCACGTG) in the DUSP14 promoter regions (–5 to +2 kb) (Fig. 7d). Ectopic expression of β-catenin downregulated ACOX1 (Fig. 7e) and c-Myc ectopic expression upregulated DUSP14 (Fig. 7f). Chromatin immunoprecipitation (ChIP) assays revealed that β-catenin occupied the promoter of ACOX1 while c-Myc bound the promoter of DUSP14 in HCT15 and RKO cells (Fig. 7g, h). Luciferase reporter assays confirmed that β-catenin significantly suppressed TBE, and that c-Myc activated RE, as compared to TBE or RE mutant (TBE or RE Mut) in HCT15 and RKO cells (Fig. 7i).
Next, we examined whether β-catenin regulates ACOX1 and DUSP14 in vivo. Both ACOX1 mRNA and protein levels were significantly decreased, whereas the mRNA and protein levels of DUSP14 were highly increased in intestinal adenomas of APCMin/+ mice (Fig. 7j, k). These results support the notion that β-catenin directly or indirectly regulates the ACOX1 and DUSP14 transcription in CRC, thus constituting a reciprocal regulation among β-catenin, ACOX1, and DUSP14.
## The DUSP14-ACOX1-PA-β-catenin axis is dysregulated in human CRC
To illustrate the correlation among DUSP14, ACOX1, and β-catenin in CRC, we used IHC of matched patient samples from the HPA dataset (Supplementary Fig. S9a) and validated the negative correlation between ACOX1 and DUSP14 and the negative correlation between ACOX1 and β-catenin (Supplementary Fig. S9b, c). Interestingly, we found that the DUSP14-ACOX1-β-catenin axis is dysregulated in early-stage CRC (Supplementary Fig. S9d). To better validate this observation, 24 early-stage CRC samples (T) with adjacent normal colon tissues (N) were collected (Supplementary Fig. S9e). We observed that ACOX1 protein was significantly decreased in these CRC samples, whereas DUSP14 and β-catenin were markedly increased (Fig. 8a–c). Moreover, the negative correlations between DUSP14 and ACOX1, and ACOX1 and β-catenin, and the positive correlation between β-catenin and DUSP14 were also confirmed in our samples (Fig. 8d, f, h). These results were further validated in our CRC TMA (Fig. 8e, g, i; Supplementary Fig. S9f). More importantly, PA levels were higher in tumor samples than those in paired normal samples (Fig. 8j).Fig. 8Dysregulation of the DUSP14-ACOX1-PA-β-catenin axis in human CRC.a–c Relative protein levels of DUSP14 (a), ACOX1 (b) and β-catenin (c). The proteins were quantified by densitometry, with β-actin as a normalizer, as shown in Supplementary Fig. S9d. d, e Pearson correlation analysis of DUSP14 and ACOX1 proteins from human CRCs (d) and CRC TMA (e). f, g Pearson correlation analysis of β-catenin and ACOX1 proteins from human CRCs (f) and CRC TMA (g). h, i Pearson correlation analysis of DUSP14 and β-catenin proteins from human CRCs (h) and CRC TMA (i). j Analysis of PA levels in adjacent normal tissues versus matched primary tumor tissues. k Predicted compounds and target genes for inhibiting DUSP14-ACOX1-β-catenin axis from DeSigN. l Drug sensitivity analysis for inhibiting DUSP14-ACOX1-β-catenin axis. Correlations of drug target genes and DUSP14, ACOX1, and β-catenin were shown by the heatmap, based on TCGA RNA-SeqV2. The red rectangular frame indicates high drug sensitivity. m Nu-7441 effectively inhibits the DUSP14-ACOX1-β-catenin axis in HCT15 and RKO cells. HCT15 and RKO cells were treated with DMSO, Cytarabine (10 nM), AZ628 (10 nM), BMS-536924 (5 nM) or Nu-7441 (20 nM) for 24 h, and cells were subjected to RT-qPCR. Data were analyzed using unpaired Student’s t-test (a–c, m) or paired Student’s t-test (j). Data are presented as means ± SD; *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; n, number of patient samples.
To explore whether there are known small molecules or drugs that inhibit the DUSP14-ACOX1-β-catenin axis, the DeSigN54 database (Fig. 8k) and TCGA dataset were used, and they showed that Nu-7441 and Cytarabine inhibitors may inhibit the DUSP14-ACOX1-β-catenin axis (Fig. 8l). Only Nu-7441 could effectively inhibit this signal axis in HCT15 and RKO cells (Fig. 8m). Moreover, Nu-7441 significantly inhibited the proliferation of HCT15 and RKO cells (Supplementary Fig. S9g). Taken together, these data support the idea that the DUSP14-ACOX1-PA-β-catenin axis plays a crucial role in CRC progression and that Nu-7441 may be a potential strategy for the treatment of CRC by inhibiting the DUSP14-ACOX1-PA-β-catenin axis.
## Discussion
In this study, we demonstrated a crucial role of dephosphorylation in regulating the stability of ACOX1 protein, which reveals the crosstalk of dephosphorylation and ubiquitination of ACOX1. Accumulation of PA induced by enhanced ACOX1 dephosphorylation promotes palmitoylation of β-catenin, providing an additional layer of regulation to enhance β-catenin signaling in cancer. These findings establish a link between cancer metabolism and the β-catenin signaling and reveal modulation of these post-translational modifications as a promising therapeutic strategy against cancer.
Metabolic reprogramming affects tumorigenesis and tumor progression by maintaining deregulated proliferation and preserving a dedifferentiated state1. Previous studies have shown that dysregulation of metabolic enzyme ME1 phosphorylation and acetylation promotes lipid metabolism and colorectal tumorigenesis55, and that hyperactivation of metabolic enzyme PKM2 methylation promotes aerobic glycolysis and tumorigenesis56. Metabolic enzyme ACOX1, a rate-limiting enzyme in peroxisomal fatty acid β-oxidation6–8, is expressed in multiple tissues. Accumulating evidence has revealed the aberrantly low expression of ACOX1 in many cancers such as lymphoma57, oral squamous cell carcinoma13, and bladder cancer58. Herein, through systematic bioinformatics screening and a series of molecular and cellular experiments, we revealed that reprogramming of PA induced by dephosphorylation of ACOX1 is critical for CRC progression. These findings extend our understanding of metabolic reprogramming induced by post-translational modifications of important metabolic enzymes in cancer.
Emerging evidence suggests that metabolites can regulate the epigenetic modification of proteins59,60. For example, acetyl-CoA derived from hepatic fatty acid oxidation promotes Raptor acetylation59, and S-adenosylmethionine provides methyl donor for protein methylation60. PA can modify protein palmitoylation22–24, which is known to regulate protein functions24–27. In this study, accumulated PA caused by the dephosphorylation of ACOX1 by DUSP14 modifies β-catenin C466 palmitoylation, and palmitoylation of β-catenin suppresses the phosphorylation of β-catenin by GSK3 and CK1, thereby preventing β-Trcp-mediated β-catenin trafficking to the proteasome, increasing the protein level and transcriptional activity of β-catenin (Supplementary Fig. S10). In summary, we discovered a novel β-catenin modification, palmitoylation, and the mechanism by which palmitoylation regulates β-catenin stability, which complements our knowledge of canonical β-catenin signaling.
In addition to being a potent DNA-PK inhibitor, Nu-7441 has also been shown to inhibit PI3K, mTOR, and non-homologous end joining pathway61–63, suggesting robust anti-cancer ability. Here, our data reveal that Nu-7441 significantly inhibits CRC cell growth by targeting the DUSP14-ACOX1-β-catenin axis. This finding shows that Nu-7441 may be a potential drug for the treatment of CRC.
Although popular models suggest that β-catenin phosphorylation/ubiquitination should be inhibited by APC mutations, it has been documented that different APC mutation types have different degradation efficiencies for β-catenin, which contributes to different levels of tumor progression64–66. Therefore, we believe that the DUSP14-ACOX1-β-catenin axis is still suitable for some CRC cell lines with APC mutations (such as HCT15 cell line). In addition, TCF/β-catenin not only stimulates gene transcription but can also repress it67,68. This is possibly mediated by directly recruiting repressive factors, such as Reptin or Fhit that associate with the TCF/β-catenin complex and thus repress β-catenin-mediated transcription69,70, which may explain the mechanism behind the repressive effect of the TCF/β-catenin complex on ACOX1 expression in this study. Furthermore, the E3 ubiquitin ligase and protein kinase of ACOX1, as well as the palmitoyl transferase and de-palmitoyl transferase of β-catenin have not yet been discovered. Another limitation of this study is that specific antibody recognizing phosphorylated S26 of ACOX1 has not yet been developed. We will conduct follow-up research to address the above concerns in the future. In addition, given that fatty acid β-oxidation in peroxisome and mitochondria share some common substrates (mainly some long-chain fatty acids)8,71, we speculate that mitochondria-mediated PA β-oxidation may also affect CRC progression.
The results of our study have revealed that ACOX1 is a tumor suppressor and critical for the supervision of β-catenin signaling by regulating PA-mediated β-catenin palmitoylation and stabilization. We have also proposed that inhibition of the dephosphorylation of ACOX1 by DUSP14 or β-catenin palmitoylation may be a viable option for CRC treatment.
## Cell culture and transfection
Human HCT15, RKO, HCT8, SW620, HCT116, HIEC-6, and HEK293T cells were obtained from the American Type Culture Collection (ATCC). Cells were cultured in DMEM medium (Gibco, NY, USA) supplemented with $10\%$ fetal bovine serum (Gibco, NY, USA) and $1\%$ penicillin-streptomycin (Gibco, CA, USA) at 37 °C in a $5\%$ CO2 incubator. For transfection, after growing to $70\%$ confluence, cells were transfected using Lipofectamine 3000 (Invitrogen, Carlsbad, CA) or HighGene (ABclonal, Wuhan, China), according to the manufacturer’s instructions.
## Reagents and plasmids
Proteasome inhibitor MG132 (HY-13259), Nutlin-3 (HY-50696), Cytarabine (HY-13605), BMS-536924 (HY-10262), AZ628 (HY-11004), and palmitic acid (HY-N0830) were purchased from MedChemExpress. Cycloheximide (R750107), Hydroxylamine (HAM, 467804), 2-BP [238422], and DAPI (D9542) were purchased from Sigma-Aldrich. iCRT14 (sc-362746) was purchased from Santa Cruz. A dual-luciferase reporter assay kit (DL-101-01) was purchased from Vazyme. N-Ethylmaleimide (NEM, A600450-0005) and BMCC-biotin ((1-Biotinamido)-4-[4′-(maleimidomethyl)cyclohexanecarboxamido]hexane, C100222-0050) were purchased from Sangon Biotech. Human palmitic acid ELISA kits (MM-51627H2) were purchased from MeiMian (Jiangsu, China). Anti-Flag agarose beads [23101] and Nu-7441 [503468-95-9] were purchased from Selleck (Houston, USA). RNase A (CW2105) was purchased from CWBIO. All antibodies used in this study are indicated in Supplementary Table S4. The human DUSP14, ACOX1, CTNNB1, and c-Myc coding sequences were amplified from HEK293T cDNA and cloned into pCMV-HA and pHAGE-CMV-MCS-PGK vectors. The human GSK3β and CK1 coding sequences were amplified from HEK293T cDNA and cloned into the pHAGE-CMV-MCS-PGK vector. The human ACOX1-TBE and DUSP14-RE were amplified from HCT15 gDNA and cloned into the pGL3-basic luciferase vector. The mouse ACOX1 coding sequence was amplified from mouse colon cDNA and cloned into pCDH-CMV-MCS-EF1-GFP+Puro vector. Mutations in the DUSP14, ACOX1, β-catenin, and Ubiquitin cDNAs were generated by overlap extension PCR. Deletion mutants from DUSP14 and ACOX1 were cloned into the pHAGE-CMV-MCS-PGK vector. Human DUSP14, CTNNB1, ACOX1, and mouse Acox1 shRNAs were designed and synthesized by RuiBiotech (Guangzhou, China), subsequently annealed, and inserted into the pLKO.1-puro vector. All primers for construction are presented in Supplementary Table S5.
## Animal studies
All animal studies were approved by the Animal Care Committee of Sun Yat-sen University. All mice were maintained in micro isolator cages in the Experimental Animal Center of Sun Yat-sen University.
AOM/DSS-induced mouse CRC model was performed following previously described methods38. Briefly, eight-week-old C57BL/6 mice were injected intraperitoneally with 10 mg/kg AOM (Sigma-Aldrich). After 7 days, the mice were given drinking water containing $2.5\%$ DSS (MP Biomedicals, Santa Ana, CA, USA) for a week, followed by regular drinking water for 2 weeks. Then, the mice were fed with $2.5\%$ DSS water for two rounds for 1 week and sacrificed on the 120th day.
APCMin/+/DSS-induced mouse CRC model was performed following the previously described methods39. Briefly, the eight-week-old APCMin/+ mice were fed with $1.5\%$ DSS water for 1 week and sacrificed on the 75th day.
Lentivirus production was performed following previously described methods72. Briefly, lentivirus was produced using polyethyleneimine-mediated transfection of a second-generation packaging system in HEK293T cells. Supernatant containing lentivirus was harvested at 72 h after transfection and filtered using a 0.45-μm filter. The lentivirus containing supernatant was then mixed with concentrate solution ($5\%$ PEG8000 and 0.5 M NaCl) overnight and concentrated by centrifugation at 4 °C. Virus titers were determined by ELISA kit.
For administration of lentivirus to APCMin/+/DSS-induced or AOM/DSS-induced CRC mice, the eight-week-old mice were randomly assigned to indicated groups. The control group of mice was treated with lentivirus-expressing Ctrl or shCtrl and the indicated group(s) was/were treated with lentivirus-expressing indicated protein or shRNA. Concentrated lentivirus was delivered intraperitoneally to indicated mice twice per week for 3 weeks55. On day 75 or 120, the indicated mice were sacrificed and their tumor burdens were evaluated. Tumors larger than 1 mm were counted and measured. Colon tissues were collected for RNA extraction, protein assays, and pathological studies.
For PDX transplantation, patients-derived tumor tissues were subcutaneously transplanted into BALB/c nude mice. When the tumors reached a certain size, the subcutaneous tumor was dissected and decomposed into a tumor of ~1 mm, and then inoculated into the subcutaneous of BALB/c nude mice again. One week after transplantation, mice were randomly assigned to 2 groups and injected with lentivirus expressing Ctrl or Flag-ACOX1 every 2 days for 9 total times. At day 21, PDXs were collected for tumor volume measurement and IHC analysis. Tumor volumes were calculated by the equation V (mm3) = a × b$\frac{2}{2}$, where a is the length and b is the width.
For subcutaneous xenograft model, the experiment was performed following previously described methods43. Briefly, four-week-old female BALB/c nude mice were purchased from GemPharmatech (Guangzhou, China), and 5 × 106 HCT15 cells were suspended in 100 μL PBS and injected subcutaneously in the flanks of animals. Twelve days after transplantation, 2-BP (40 mg/kg) was delivered intraperitoneally to indicated nude mice once per day for 12 days. Tumor growth was monitored every two days for a total period of 24 days. Tumor volumes were calculated by the equation V (mm3) = a × b$\frac{2}{2}$, where a is the length and b is the width.
## Human CRC specimens
Forty-five early CRC samples and 51 normal adjacent tissues were used to analyze ACOX1 transcript levels, and 24 pairs of these samples were used to analyze ACOX1, DUSP14, and β-catenin protein levels. Fifteen fresh CRC samples and matched normal adjacent tissues were used to analyze PA levels. 192 CRC samples were made into TMA to analyze indicated protein levels and overall survival. All samples were obtained from the Sixth Affiliated Hospital of Sun Yat-sen University. The diagnosis of CRCs was verified by histological review. Our study was approved by the Ethics Committee of the Sixth Affiliated Hospital of Sun Yat-sen University (2020ZSLYEC-232). All patients signed written informed consent forms before treatment.
## Stable cell lines
Stable cell line construction was performed as described previously43. Briefly, indicated lentiviral vectors were packaged in HEK293T cells. HCT15 or RKO cells were infected with lentiviruses in the presence of polybrene and were selected with 1 µg/mL puromycin for two weeks to obtain stable clones. The indicated protein expression in stable clones was validated by western blotting.
## Real-time quantitative PCR (RT-qPCR)
RT-qPCR assays were performed as described previously43. Briefly, total RNA was isolated from cells or tissues and subsequent reverse transcription was performed. qPCR was then performed with SYBR Green Supermix (Bio-Rad, Hercules, CA) using standard procedures. The β-catenin targets in this study were obtained from previous study73. All primer sequences used are listed in Supplementary Table S5. GAPDH was used as an internal control.
## Co-IP and immunoblot analysis
Co-IP and immunoblot analysis were performed as described previously43. Briefly, cells transfected with the indicated plasmids were lysed in 1 mL Lysis buffer. For immunoprecipitation, the anti-Flag agarose beads were washed with 1 mL lysis buffer three times, and then 0.9 mL of cell lysate was added into the indicated group and incubated overnight at 4 °C. The next day, the agarose beads were centrifuged and the supernatant was discarded. Subsequently, the agarose beads were washed three times and mixed in a 2× SDS sample buffer. Lysate samples were boiled for 10 min and were analyzed by immunoblotting with the indicated antibodies.
## Palmitoylation assays
For detecting protein palmitoylation, the acyl-biotin exchange (ABE) method was used74,75. Briefly, cells transfected with the indicated plasmids were lysed in 1 mL Lysis buffer containing 50 mM NEM, followed by centrifugation (20 min, 12,000 rpm, 4 °C) and immuno-precipitation overnight with anti-Flag agarose beads. After washing three times, precipitates were divided evenly into two sections, with $\frac{1}{2}$ used for the –HAM control, and the remaining $\frac{1}{2}$ was used for the +HAM for 1 h at room temperature. The precipitates were gently washed once with Wash Buffer (1 M Tris-HCl, pH 6.5), and incubated with BMCC-biotin Buffer (50 mM Tris-HCl, pH 6.5, 150 mM NaCl, 5 mM EDTA, $1\%$ Triton X-100, and 5 μM BMCC-biotin) for 1 h at 4 °C. Then the precipitates were gently washed two times again with Wash Buffer. After washing samples were analyzed by SDS-PAGE and blotting, palmitoylated β-catenin was detected with HRP-conjugated streptavidin (Sangon Biotech; 1:200 in $0.5\%$ BSA).
## Streptavidin pulldown-based quantification of palmitoylated β-catenin
Streptavidin pulldown-based quantification of palmitoylated protein was performed as previously described76. Briefly, cells were lysed in Lysis buffer, and 80 μL supernatant was saved as input. The remaining supernatant was used for ABE experiments. Then palmitoylated proteins were enriched using streptavidin agarose (Cytiva) with rotation overnight at 4 °C. Samples were centrifugated at 3000 rpm for 5 min. 80 μL supernatant was saved as output. Protein-bound streptavidin agaroses were washed three times with Wash Buffer and bound proteins were eluted with SDS loading buffer for 10 min at 95 °C. Samples were subjected to SDS-PAGE. The fraction of palmitoylated β-catenin was determined by western blotting and calculated based on the β-catenin protein level in input and output samples. β-actin was blotted as a loading control.
## ChIP assay
Cells were cross-linked in situ with $1\%$ formaldehyde for 10 min, quenched with 0.125 M glycine for 5 min at room temperature, and lysed in SDS Lysis buffer. Total lysates were sonicated to smash chromatin DNA to a size range of 200–1000 bp. The supernatant was diluted 10 times in ChIP Dilution buffer and precleared with 50 μL agarose beads for 2 h at 4 °C. Then the supernatant was collected by centrifugation, and the indicated antibodies (2 μg) were added to the supernatant. Then, the mixture was rotated overnight at 4 °C. The next day, 50 μL agarose beads were added, and rotation was continued for 2 h at 4 °C. Subsequent de-crosslinked DNA was subjected to PCR analysis using specific primers listed in Supplementary Table S5.
## Ubiquitination assay
Ubiquitination assays were performed as described previously43. Briefly, HEK293T cells were transfected with the indicated plasmids and treated with 20 μM MG132 for 6 h before collection. The cells were then lysed in RIPA lysis buffer and denatured by heating at 95 °C for 5 min. Immunoprecipitation analysis was performed as described above. The samples were boiled for 10 min in SDS sample buffer and analyzed by immunoblotting with the indicated antibodies.
## Protein half-life assay
Protein half-life assays were performed as described previously43. Briefly, the cells were transfected with the indicated plasmids, and 36 h later, the cells were treated with cycloheximide (CHX, 100 µg/mL) for the indicated time periods before collection. The cells were lysed and proteins were detected by immunoblotting with the indicated antibodies.
## Luciferase reporter assays
0.3 μg pGL3 vector expressing ACOX1-TBE, DUSP14-RE, or indicated mutant and 50 ng Renilla luciferase reporter were transfected in triplicates into HCT15 or RKO cells. After 36 h, luciferase activities were determined by the Dual-Luciferase Reporter Assay System. The Renilla activity was used as an internal control.
## MS analysis
For protein qualitative analysis, HEK293T cells were transfected with the Flag-ACOX1 plasmid, lysed in Lysis buffer, and immune-precipitated with anti-Flag agarose beads. After SDS-PAGE and Coomassie Blue staining of the Flag-ACOX1-associated complexes, the bands were cut, subjected to in-gel trypsin digestion, and dried. The protein composition and protein site modification were analyzed by MS according to the protocols described previously77.
For protein quantitative analysis of CRC samples, five fresh CRC samples and matched normal adjacent tissues were collected from the Sixth Affiliated Hospital of Sun Yat-sen University, and lysed in Urea Lysis buffer. Extracted proteins were subjected to LC-MS/MS (Thermo Fisher Scientific, Rockford, IL, USA) analysis according to the standard protocols78. Proteins were identified by Firmiana, a one-stop proteomic data processing platform79. Briefly, Mascot (Matrix Science, version 2.3.01) and Nation Center for Biotechnology Information (NCBI) Ref-Seq human proteome database (updated on 04-07-2013) were used in the identification and quantification processes. Peptides (FDR < 0.01) were selected, and the proteins that contain high-quality and unique peptides were considered qualified. Length of minimal peptide was seven amino acids. Label-free intensity-based absolute quantification (iBAQ) was used to quantify proteins80. Fraction of total (FOT) was defined as the iBAQ value per protein divided by the sum of all protein iBAQ values. The FOT value was multiplied by 105 for easy representation.
For cell proteome analysis, HCT15 cells were treated with PA (100 μM) or DMSO for 72 h, and lysed in Urea Lysis buffer. Extracted proteins were further digested, purified and measured. One microgram of protein per sample was subjected to LC-MS/MS analysis according to the standard protocols78–80.
## Organoid assays
Organoid assays were performed as described previously81. Briefly, human intestinal normal epithelial cells were maintained with organoid culture advanced DMEM/F12 containing growth factors (100 ng/mL Noggin (Peprotech), 500 ng/mL R-spondin (Peprotech), 50 ng/mL epidermal growth factor (Peprotech) and 10 μM Y-27632 (Abmole)) and treated with PA (300 μM). After spheroid organoid formation for 5 days, organoids were photographed and measured in diameter.
## IHC assays
Colon samples were fixed and embedded in paraffin according to standard protocols. H&E staining was performed in paraffin-embedded sections using hematoxylin and eosin (Servicebio). The analysis of IHC was performed using indicated antibodies against ACOX1 (Abcam) and Ki67 (Servicebio). The IHC staining results were scored considering both the intensity of staining and the proportion of tumor cells with positive reaction. The intensity of staining was scored as follows: 0, negative; 1, weak; 2, medium; and 3, strong. The frequency of positive cells was scored as follows: 0, < $5\%$; 1, $1\%$–$25\%$; 2, $25\%$–$50\%$; 3, $50\%$–$75\%$; 4, > $75\%$. Total score ranging from 0 to 12 was determined by multiplying the score of staining intensity and the score of positive area.
## Statistical analysis
For data analysis, GraphPad Prism 8.3.0, Microsoft Excel, and IBM SPSS Statistics 26 were used. Statistical significance ($P \leq 0.05$) was performed using the unpaired or paired Student’s t-test or χ2 test. Data are presented as the means ± SD.
## Supplementary information
Supplementary Information The online version contains supplementary material available at 10.1038/s41421-022-00515-x.
## References
1. Yoshida GJ. **Metabolic reprogramming: the emerging concept and associated therapeutic strategies**. *J. Exp. Clin. Cancer Res.* (2015.0) **34** 111. DOI: 10.1186/s13046-015-0221-y
2. Pavlova NN, Thompson CB. **The emerging hallmarks of cancer metabolism**. *Cell Metab.* (2016.0) **23** 27-47. DOI: 10.1016/j.cmet.2015.12.006
3. Lu M. **ACOT12-fependent alteration of acetyl-CoA drives hepatocellular carcinoma metastasis by epigenetic induction of epithelial-mesenchymal transition**. *Cell Metab.* (2019.0) **29** 886-900.e5. DOI: 10.1016/j.cmet.2018.12.019
4. Ringel AE. **Obesity shapes metabolism in the tumor microenvironment to suppress anti-tumor immunity**. *Cell* (2020.0) **183** 1848-1866.e26. DOI: 10.1016/j.cell.2020.11.009
5. Ryall JG, Cliff T, Dalton S, Sartorelli V. **Metabolic reprogramming of stem cell epigenetics**. *Cell Stem Cell* (2015.0) **17** 651-662. DOI: 10.1016/j.stem.2015.11.012
6. Hashimoto T. **Peroxisomal beta-oxidation: enzymology and molecular biology**. *Ann. N. Y. Acad. Sci.* (1996.0) **804** 86-98. DOI: 10.1111/j.1749-6632.1996.tb18610.x
7. Van Veldhoven PP, Vanhove G, Assselberghs S, Eyssen HJ, Mannaerts GP. **Substrate specificities of rat liver peroxisomal acyl-CoA oxidases: palmitoyl-CoA oxidase (inducible acyl-CoA oxidase), pristanoyl-CoA oxidase (non-inducible acyl-CoA oxidase), and trihydroxycoprostanoyl-CoA oxidase**. *J. Biol. Chem.* (1992.0) **267** 20065-20074. DOI: 10.1016/S0021-9258(19)88666-0
8. Reddy JK, Mannaerts GP. **Peroxisomal lipid metabolism**. *Annu. Rev. Nutr.* (1994.0) **14** 343-370. DOI: 10.1146/annurev.nu.14.070194.002015
9. Singh I, Moser AE, Goldfischer S, Moser HW. **Lignoceric acid is oxidized in the peroxisome: implications for the Zellweger cerebro-hepato-renal syndrome and adrenoleukodystrophy**. *Proc. Natl. Acad. Sci. USA* (1984.0) **81** 4203-4207. DOI: 10.1073/pnas.81.13.4203
10. Van Veldhoven PP. **Biochemistry and genetics of inherited disorders of peroxisomal fatty acid metabolism**. *J. Lipid Res.* (2010.0) **51** 2863-2895. DOI: 10.1194/jlr.R005959
11. Fan CY. **Hepatocellular and hepatic peroxisomal alterations in mice with a disrupted peroxisomal fatty acyl-coenzyme A oxidase gene**. *J. Biol. Chem.* (1996.0) **271** 24698-24710. DOI: 10.1074/jbc.271.40.24698
12. Huang J. **Progressive endoplasmic reticulum stress contributes to hepatocarcinogenesis in fatty acyl-CoA oxidase 1-deficient mice**. *Am. J. Pathol.* (2011.0) **179** 703-713. DOI: 10.1016/j.ajpath.2011.04.030
13. Lai YH. **MiR-31-5p-ACOX1 axis enhances tumorigenic fitness in oral squamous cell carcinoma via the promigratory prostaglandin E2**. *Theranostics* (2018.0) **8** 486-504. DOI: 10.7150/thno.22059
14. Sun LN. **SIRT1 suppresses colorectal cancer metastasis by transcriptional repression of miR-15b-5p**. *Cancer Lett.* (2017.0) **409** 104-115. DOI: 10.1016/j.canlet.2017.09.001
15. Wen L, Han Z. **Identification and validation of xenobiotic metabolism-associated prognostic signature based on five genes to evaluate immune microenvironment in colon cancer**. *J. Gastrointest. Oncol.* (2021.0) **12** 2788-2802. DOI: 10.21037/jgo-21-655
16. Fatima S. **High-fat diet feeding and palmitic acid increase CRC growth in β2AR-dependent manner**. *Cell Death Dis.* (2019.0) **10** 711. DOI: 10.1038/s41419-019-1958-6
17. Fatima S. **Palmitic acid is an intracellular signaling molecule involved in disease development**. *Cell Mol. Life Sci.* (2019.0) **76** 2547-2557. DOI: 10.1007/s00018-019-03092-7
18. Pascual G. **Targeting metastasis-initiating cells through the fatty acid receptor CD36**. *Nature* (2017.0) **541** 41-45. DOI: 10.1038/nature20791
19. Pan J. **CD36 mediates palmitate acid-induced metastasis of gastric cancer via AKT/GSK-3β/β-catenin pathway**. *J. Exp. Clin. Cancer Res.* (2019.0) **38** 52. DOI: 10.1186/s13046-019-1049-7
20. Kwan HY. **Signal transducer and activator of transcription-3 drives the high-fat diet-associated prostate cancer growth**. *Cell Death Dis.* (2019.0) **10** 637. DOI: 10.1038/s41419-019-1842-4
21. Pascual G. **Dietary palmitic acid promotes a prometastatic memory via Schwann cells**. *Nature* (2021.0) **599** 485-490. DOI: 10.1038/s41586-021-04075-0
22. Nile AH, Hannoush RN. **Fatty acylation of Wnt proteins**. *Nat. Chem. Biol.* (2016.0) **12** 60-69. DOI: 10.1038/nchembio.2005
23. Janda CY, Garcia KC. **Wnt acylation and its functional implication in Wnt signalling regulation**. *Biochem. Soc. Trans.* (2015.0) **43** 211-216. DOI: 10.1042/BST20140249
24. Du W. **Loss of optineurin drives cancer immune evasion via palmitoylation-dependent IFNGR1 lysosomal sorting and degradation**. *Cancer Discov.* (2021.0) **11** 1826-1843. DOI: 10.1158/2159-8290.CD-20-1571
25. Zhang M. **A STAT3 palmitoylation cycle promotes T(H)17 differentiation and colitis**. *Nature* (2020.0) **586** 434-439. DOI: 10.1038/s41586-020-2799-2
26. Yao H. **Inhibiting PD-L1 palmitoylation enhances T-cell immune responses against tumours**. *Nat. Biomed. Eng.* (2019.0) **3** 306-317. DOI: 10.1038/s41551-019-0375-6
27. Zhang Z. **DHHC9-mediated GLUT1 S-palmitoylation promotes glioblastoma glycolysis and tumorigenesis**. *Nat. Commun.* (2021.0) **12** 5872. DOI: 10.1038/s41467-021-26180-4
28. Lee MA. **Wnt3a expression is associated with MMP-9 expression in primary tumor and metastatic site in recurrent or stage IV colorectal cancer**. *BMC Cancer* (2014.0) **14** 125. DOI: 10.1186/1471-2407-14-125
29. Yuan S. **Role of Wnt/β-catenin signaling in the chemoresistance modulation of colorectal cancer**. *Biomed. Res. Int.* (2020.0) **2020** 9390878. DOI: 10.1155/2020/9390878
30. 30.Comprehensive molecular characterization of human colon and rectal cancer. Nature487, 330–337 (2012).
31. Bustos VH. **The first armadillo repeat is involved in the recognition and regulation of beta-catenin phosphorylation by protein kinase CK1**. *Proc. Natl. Acad. Sci. USA* (2006.0) **103** 19725-19730. DOI: 10.1073/pnas.0609424104
32. Wu G. **Structure of a beta-TrCP1-Skp1-beta-catenin complex: destruction motif binding and lysine specificity of the SCF(beta-TrCP1) ubiquitin ligase**. *Mol. Cell* (2003.0) **11** 1445-1456. DOI: 10.1016/S1097-2765(03)00234-X
33. Chocarro-Calvo A, García-Martínez JM, Ardila-González S, De la Vieja A, García-Jiménez C. **Glucose-induced β-catenin acetylation enhances Wnt signaling in cancer**. *Mol. Cell* (2013.0) **49** 474-486. DOI: 10.1016/j.molcel.2012.11.022
34. Ha JR. **β-catenin is O-GlcNAc glycosylated at Serine 23: implications for β-catenin’s subcellular localization and transactivator function**. *Exp. Cell Res.* (2014.0) **321** 153-166. DOI: 10.1016/j.yexcr.2013.11.021
35. Possemato R. **Functional genomics reveal that the serine synthesis pathway is essential in breast cancer**. *Nature* (2011.0) **476** 346-350. DOI: 10.1038/nature10350
36. Bartha Á, Győrffy B. **TNMplot.com: A web tool for the comparison of gene expression in normal, tumor and metastatic tissues**. *Int. J. Mol. Sci.* (2021.0) **22** 2622. DOI: 10.3390/ijms22052622
37. Joanito I. **Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer**. *Nat. Genet.* (2022.0) **54** 963-975. DOI: 10.1038/s41588-022-01100-4
38. Parang B, Barrett CW, Williams CS. **AOM/DSS model of colitis-associated cancer**. *Methods Mol. Biol.* (2016.0) **1422** 297-307. DOI: 10.1007/978-1-4939-3603-8_26
39. He Z. **Campylobacter jejuni promotes colorectal tumorigenesis through the action of cytolethal distending toxin**. *Gut* (2019.0) **68** 289-300. DOI: 10.1136/gutjnl-2018-317200
40. Neufert C. **Inducible mouse models of colon cancer for the analysis of sporadic and inflammation-driven tumor progression and lymph node metastasis**. *Nat. Protoc.* (2021.0) **16** 61-85. DOI: 10.1038/s41596-020-00412-1
41. Vasaikar S. **Proteogenomic analysis of human colon cancer reveals new therapeutic opportunities**. *Cell* (2019.0) **177** 1035-1049.e19. DOI: 10.1016/j.cell.2019.03.030
42. Huttlin EL. **Dual proteome-scale networks reveal cell-specific remodeling of the human interactome**. *Cell* (2021.0) **184** 3022-3040.e28. DOI: 10.1016/j.cell.2021.04.011
43. Zhang Q. **The MAP3K13-TRIM25-FBXW7α axis affects c-Myc protein stability and tumor development**. *Cell Death Differ.* (2020.0) **27** 420-433. DOI: 10.1038/s41418-019-0363-0
44. Li Y. **FBXL6 degrades phosphorylated p53 to promote tumor growth**. *Cell Death Differ.* (2021.0) **28** 2112-2125. DOI: 10.1038/s41418-021-00739-6
45. Wang S. **Hepatocyte DUSP14 maintains metabolic homeostasis and suppresses inflammation in the liver**. *Hepatology* (2018.0) **67** 1320-1338. DOI: 10.1002/hep.29616
46. Rubinfeld B. **Binding of GSK3beta to the APC-beta-catenin complex and regulation of complex assembly**. *Science* (1996.0) **272** 1023-1026. DOI: 10.1126/science.272.5264.1023
47. Liu C. **Control of beta-catenin phosphorylation/degradation by a dual-kinase mechanism**. *Cell* (2002.0) **108** 837-847. DOI: 10.1016/S0092-8674(02)00685-2
48. Xing Y, Clements WK, Kimelman D, Xu W. **Crystal structure of a beta-catenin/axin complex suggests a mechanism for the beta-catenin destruction complex**. *Genes Dev.* (2003.0) **17** 2753-2764. DOI: 10.1101/gad.1142603
49. Hart M. **The F-box protein beta-TrCP associates with phosphorylated beta-catenin and regulates its activity in the cell**. *Curr. Biol.* (1999.0) **9** 207-210. DOI: 10.1016/S0960-9822(99)80091-8
50. Fang D. **Phosphorylation of beta-catenin by AKT promotes beta-catenin transcriptional activity**. *J. Biol. Chem.* (2007.0) **282** 11221-11229. DOI: 10.1074/jbc.M611871200
51. Blanc M. **SwissPalm: Protein palmitoylation database**. *F1000Res.* (2015.0) **4** 261. DOI: 10.12688/f1000research.6464.1
52. Mandard S, Müller M, Kersten S. **Peroxisome proliferator-activated receptor alpha target genes**. *Cell. Mol. Life Sci.* (2004.0) **61** 393-416. DOI: 10.1007/s00018-003-3216-3
53. Rakhshandehroo M, Knoch B, Müller M, Kersten S. **Peroxisome proliferator-activated receptor alpha target genes**. *PPAR Res.* (2010.0) **2010** 612089. DOI: 10.1155/2010/612089
54. Lee BK. **DeSigN: connecting gene expression with therapeutics for drug repurposing and development**. *BMC Genomics* (2017.0) **18** 934. DOI: 10.1186/s12864-016-3260-7
55. Zhu Y. **Dynamic regulation of ME1 phosphorylation and acetylation affects lipid metabolism and colorectal tumorigenesis**. *Mol. Cell* (2020.0) **77** 138-149.e5. DOI: 10.1016/j.molcel.2019.10.015
56. Liu F. **PKM2 methylation by CARM1 activates aerobic glycolysis to promote tumorigenesis**. *Nat. Cell Biol.* (2017.0) **19** 1358-1370. DOI: 10.1038/ncb3630
57. Zheng FM. **ACOX1 destabilizes p73 to suppress intrinsic apoptosis pathway and regulates sensitivity to doxorubicin in lymphoma cells**. *BMB Rep.* (2019.0) **52** 566-571. DOI: 10.5483/BMBRep.2019.52.9.094
58. Xie JY. **The prognostic significance of DAPK1 in bladder cancer**. *PLoS One* (2017.0) **12** e0175290. DOI: 10.1371/journal.pone.0175290
59. He A. **Acetyl-CoA derived from hepatic peroxisomal β-oxidation inhibits autophagy and promotes steatosis via mTORC1 activation**. *Mol. Cell* (2020.0) **79** 30-42.e4. DOI: 10.1016/j.molcel.2020.05.007
60. Bauerle MR, Schwalm EL, Booker SJ. **Mechanistic diversity of radical S-adenosylmethionine (SAM)-dependent methylation**. *J. Biol. Chem.* (2015.0) **290** 3995-4002. DOI: 10.1074/jbc.R114.607044
61. Li Y. **Protein phosphatase 2A and DNA-dependent protein kinase are involved in mediating rapamycin-induced Akt phosphorylation**. *J. Biol. Chem.* (2013.0) **288** 13215-13224. DOI: 10.1074/jbc.M113.463679
62. Bergs JW. **Inhibition of homologous recombination by hyperthermia shunts early double strand break repair to non-homologous end-joining**. *DNA Repair* (2013.0) **12** 38-45. DOI: 10.1016/j.dnarep.2012.10.008
63. Rajput M, Singh R, Singh N, Singh RP. **EGFR-mediated Rad51 expression potentiates intrinsic resistance in prostate cancer via EMT and DNA repair pathways**. *Life Sci.* (2021.0) **286** 120031. DOI: 10.1016/j.lfs.2021.120031
64. Yang J. **Adenomatous polyposis coli (APC) differentially regulates beta-catenin phosphorylation and ubiquitination in colon cancer cells**. *J. Biol. Chem.* (2006.0) **281** 17751-17757. DOI: 10.1074/jbc.M600831200
65. Polakis P. **Wnt signaling and cancer**. *Genes Dev.* (2000.0) **14** 1837-1851. DOI: 10.1101/gad.14.15.1837
66. Matsumoto T. **Serrated adenoma in familial adenomatous polyposis: relation to germline APC gene mutation**. *Gut* (2002.0) **50** 402-404. DOI: 10.1136/gut.50.3.402
67. Delmas V. **Beta-catenin induces immortalization of melanocytes by suppressing p16INK4a expression and cooperates with N-Ras in melanoma development**. *Genes Dev.* (2007.0) **21** 2923-2935. DOI: 10.1101/gad.450107
68. Valenta T, Hausmann G, Basler K. **The many faces and functions of beta-catenin**. *EMBO J.* (2012.0) **31** 2714-2736. DOI: 10.1038/emboj.2012.150
69. Bauer A. **Pontin52 and reptin52 function as antagonistic regulators of beta-catenin signalling activity**. *EMBO J.* (2000.0) **19** 6121-6130. DOI: 10.1093/emboj/19.22.6121
70. Weiske J, Albring KF, Huber O. **The tumor suppressor Fhit acts as a repressor of beta-catenin transcriptional activity**. *Proc. Natl. Acad. Sci. USA* (2007.0) **104** 20344-20349. DOI: 10.1073/pnas.0703664105
71. Schrader M, Costello J, Godinho LF, Islinger M. **Peroxisome-mitochondria interplay and disease**. *J. Inherit. Metab. Dis.* (2015.0) **38** 681-702. DOI: 10.1007/s10545-015-9819-7
72. Bonci D. **The miR-15a-miR-16-1 cluster controls prostate cancer by targeting multiple oncogenic activities**. *Nat. Med.* (2008.0) **14** 1271-1277. DOI: 10.1038/nm.1880
73. Fang L. **ERK2-Dependent phosphorylation of CSN6 is critical in colorectal cancer development**. *Cancer Cell* (2015.0) **28** 183-197. DOI: 10.1016/j.ccell.2015.07.004
74. Drisdel RC, Alexander JK, Sayeed A, Green WN. **Assays of protein palmitoylation**. *Methods* (2006.0) **40** 127-134. DOI: 10.1016/j.ymeth.2006.04.015
75. Kokkola T. **Somatostatin receptor 5 is palmitoylated by the interacting ZDHHC5 palmitoyltransferase**. *FEBS Lett.* (2011.0) **585** 2665-2670. DOI: 10.1016/j.febslet.2011.07.028
76. Wang L. **CARM1 methylates chromatin remodeling factor BAF155 to enhance tumor progression and metastasis**. *Cancer Cell* (2014.0) **25** 21-36. DOI: 10.1016/j.ccr.2013.12.007
77. Shevchenko A, Wilm M, Vorm O, Mann M. **Mass spectrometric sequencing of proteins silver-stained polyacrylamide gels**. *Anal. Chem.* (1996.0) **68** 850-858. DOI: 10.1021/ac950914h
78. Ge S. **A proteomic landscape of diffuse-type gastric cancer**. *Nat. Commun.* (2018.0) **9** 1012. DOI: 10.1038/s41467-018-03121-2
79. Feng J. **Firmiana: towards a one-stop proteomic cloud platform for data processing and analysis**. *Nat. Biotechnol.* (2017.0) **35** 409-412. DOI: 10.1038/nbt.3825
80. Schwanhäusser B. **Global quantification of mammalian gene expression control**. *Nature* (2011.0) **473** 337-342. DOI: 10.1038/nature10098
81. Jung YS. **TMEM9 promotes intestinal tumorigenesis through vacuolar-ATPase-activated Wnt/β-catenin signalling**. *Nat. Cell Biol.* (2018.0) **20** 1421-1433. DOI: 10.1038/s41556-018-0219-8
|
---
title: Determining Regional Differences in Barriers to Accessing Health Care Among
Farmworkers Using the National Agricultural Workers Survey
authors:
- Sheila Soto
- Aaron Meck Yoder
- Benjamin Aceves
- Tomas Nuño
- Refugio Sepulveda
- Cecilia Ballesteros Rosales
journal: Journal of Immigrant and Minority Health
year: 2022
pmcid: PMC9988993
doi: 10.1007/s10903-022-01406-9
license: CC BY 4.0
---
# Determining Regional Differences in Barriers to Accessing Health Care Among Farmworkers Using the National Agricultural Workers Survey
## Abstract
Farmworkers are an essential workforce in the U.S. We assessed the regions in the National Agricultural Workers Survey on the difficulty of accessing health care among farmworkers in the U.S. The study included 9577 farmworkers. Farmworkers in all regions were more likely to report having difficulty accessing health care because it was too expensive. The overall odds ratio for difficulty accessing health care was lower in the MW after adjusting. Farmworkers employed in the SE had greater difficulty accessing health care because of language barriers. Farmworkers employed in CA had difficulty accessing health care in the U.S. because it was too expensive or far away. Results follow previous studies on barriers to access health care among the farmworker population. Understanding regional disparities in the presence of barriers to accessing health care among farmworkers is an essential step to improving equitable health care access in the U.S.
## Introduction
The farmworker population residing in the U.S. is an essential workforce critical to our economy and food system productivity [1, 2]. Presently, there are nearly four million farmworkers in the U.S., with estimates up to 3 million workers classified as migrant or seasonal [1, 3]. Even with exposures to occupational hazards and injury rates as high as $12.5\%$, farmworkers lack occupational health standards and fair labor laws [4, 5]. Farmworkers face low socioeconomic factors, including high rates of poverty, low levels of educational attainment, and inadequate housing conditions [6, 7]. Despite being an essential labor force, farmworkers encounter barriers to accessing health care in the U.S.
Multiple studies have reported barriers to accessing health care among farmworkers. Many farmworkers live and work in rural areas, which present unique challenges in reaching care, such as difficulty attaining transportation to and from appointments [8–10]. Farmworkers, specifically those working in certain areas or crops are often immigrants, or have family members who are immigrants, and have reported fear of immigration enforcement as a barrier to accessing health care [8–10]. Similarly, many farmworkers with limited English proficiency have reported language barriers resulting in both access and quality of care issues [8–10]. The increasing cost of health care services is a national problem, but with a large portion of farmworkers living below the poverty line—health care costs have become an astronomical barrier in accessing services [8–10].
Barriers to health care access have lasting adverse effects on farmworker health outcomes. For instance, farmworkers lacking critical primary care, which limits primary and secondary prevention efforts, have an increased risk of suffering from an undiagnosed chronic disease [8, 11]. This in turn hinders access to early treatment and management to prevent further complications [8, 11, 12]. In addition, populations that have difficulty accessing primary care tend to rely on urgent or emergency care, which is both economically impractical and dangerous for patients [10].
Improving health outcomes for farmworkers in the U.S. requires a broad approach, including policy changes to labor laws and expanding access to health care for all farmworkers. Assessing and understanding disparities in farmworkers’ access to health care services can improve the allocation and development of resources to ensure farmworkers do not face barriers when accessing health care. A patchwork of local, state, and federal laws currently attempts to ensure the agricultural workforce’s health and safety; thus, discrepancies in barriers to accessing health care may exist depending on geographic location [13]. Also, the type of crop work, migratory patterns, season, citizenship, or visa status of farmworkers may be vastly different per geographical region of employment [14, 15] The objective of this study is to determine regional differences in barriers farmworkers face when accessing health care in the U.S. using the National Agriculture Workers Survey (NAWS).
## Study Design
The use of de-identified secondary publicly available data exempts this research from approval by the Institutional Review Board or Ethics Review Committee. We verified that all data was de-identified upon extraction of the datasets from the NAWS 2013–2016 before beginning the statistical analysis [16]. The NAWS is a national survey conducted by the U.S. Department of Labor (DOL) designed to provide a representative sample of hired farmworkers. The DOL, Employment and Training Administration (ETA) makes formal request for approval to the Office of Management and Budget (OMB) for the methodology of the NAWS. The NAWS uses an intricate, multistage survey design to account for seasonal and regional farmworker employment changes. There are seven layers of sampling for the primary study including crop cycle, region, farm labor area, county, zip code, employer, and crop workers. NAWS samples 3 times a year in 12 geographic regions and randomly selects employers within randomly sampled Farm Labor Areas (FLA) resulting in 36-time-by-space strata. The selection of cycles, regions, and farm labor-areas is dependent on the amount of farm labor in a region during the collection cycle [17]. The selection of counties, ZIP codes, and employers is determined by the farm expenditures in FLA and information from employers from the Quarterly Census of Employment and Wages and commercial lists of employers or other related data [16]. Between 1500 and 3600 farmworkers are randomly selected for interviews at their worksite during breaks or before or after work on an annual basis [17]. The NAWS excludes H2A Visa holders and farmworkers who carry out farm-related tasks not related to crop work such as those who exclusively work with livestock or dairy [16]. Most workers have a place of birth in Mexico ($68.5\%$), U.S. ($26.0\%$), Central America ($5.0\%$), or other ($1.0\%$) [18]. Length of time in the U.S. in years is not captured by the NAWS, instead variables of migrant type recorded including settled (did not migrate; $82.5\%$), shuttle migrant ($10.0\%$), follow-the-crop migrant ($5.0\%$), or foreign-born newcomer ($3.0\%$) [18].
## Exposure
This study assessed the effect of employment in the six U.S. regions on the difficulty of accessing health care among farmworkers in the U.S. We analyzed the regions using the NAWS predetermined subgroups of Northeast (NE), Southeast (SE), Midwest (MW), Southwest (SW), Northwest (NW), and California (CA) all on the 17 USDA-designated regions (Fig. 1). Due to the size and number of farmworkers in *California a* separate subgroup was created during the primary study for the state to not heavily impact if included in the surrounding regions. We assessed by regions to have an illustrative sample of the various types of agricultural crops and work seasons throughout the U.S. We calculated descriptive statistics for farmworkers in each region and CA, as the reference region, in all models.
Fig. 1Geographical regions where farmworkers were selected to participate in the NAWS
## Measures
All variables in this study used during secondary data analysis were self-reported during interviews conducted during the NAWS primary study. The outcomes are the barriers that make accessing health care services difficult for farmworkers. We focused on three barriers (cost, language, and transportation). Each outcome was a binary response (Yes or No) and was assessed in the NAWS through the question: “When you want to get health care in the U.S. what are the main difficulties you face?” with the three options: “no transportation, too far away”, “do not speak my language”, and “too expensive” given separately.
We calculated descriptive statistics for age, gender, hourly wage, health insurance status, the origin of birth, and education level, gender (male or female), health insurance status (has health insurance, does not, or does not know), and origin of birth (U.S. or other) are categorical variables and include unweighted counts and percentages. The descriptive statistics, age, hourly wage, and education level are continuous variables, and the NAWS public dataset help determine the mean and standard error, the survey weights in the NAWS public dataset.
## Statistical Methods
First, we use a separate logistic regression model for each outcome, which adjusted for the NAWS survey weights to determine the odds of each outcome by region with California as the reference group. An adjusted odds ratio (OR) is an odds ratio that controls for other predictor variables. Then, three more logistic regression models adjusted for NAWS survey weights and hourly wage as a continuous response: age (< 30, 30–44, 45–59, 60 ≤ years old); gender (male, female); and health insurance status (insured, uninsured, or does not know), to determine the adjusted OR for each region. Age was categorized to meet the assumptions of logistic regression.
We performed a sensitivity analysis using imputed values for missing outcome and covariate data, using hot deck imputation for complex surveys [19]. After imputation, a logistic regression model for each outcome was fit to the data, adjusted for survey weights and previously mentioned covariate using SAS software version 9.4.16.
## Descriptive Statistics
The study included 9577 farmworkers with 1229 in the NE region, 1280 in the SE region, 1076 in the MW region, 727 in the SW region, 1470 in the NW region, and 3791 in CA. Descriptive statistics for age, gender, hourly wage, health insurance status, origin of birth, and education level stratified by region of employment are in Table 1. Overall, the mean age was similar for farmworkers in each region (lowest 36.3 in the NW and highest 41.9 in the SW). Approximately three-quarters of farmworkers in each region were male. The mean hourly wage was lowest in the SW region at $9.17 and highest among farmworkers in the MW at $11.25. About half ($46.2\%$) of the farmworkers in the MW reported having health insurance, while only $25.8\%$, $33.4\%$, and $33.8\%$ of farmworkers in the SE, NW, and SW regions had health insurance. Nearly half ($44.9\%$) of farmworkers in the MW were born in the U.S., while only $8.4\%$, $18.7\%$, and $26\%$ of farmworkers in CA, NW, and SW were born in the U.S., respectively. The mean education level was 10.5 years among farmworkers in the MW, while farmworkers in other regions averaged between 7.6 and 9.0 years of education.
Table 1Descriptive statistics, farmworkers interviewed in the NAWS 2013–2016 ($$n = 9577$$)NE ($$n = 1229$$)SE ($$n = 1280$$)MW ($$n = 1076$$)SW ($$n = 727$$)NW ($$n = 1470$$)CA ($$n = 3791$$)Age, mean (SndEr)37.3 (0.8)37.1 (0.9)40.2 (1.1)41.9 (1.0)36.3 (0.7)38.6 (0.5)Male gender, n (%)954 ($77.6\%$)887 ($69.3\%$)799 ($74.3\%$)603 ($82.9\%$)1,137 ($77.3\%$)3,016 ($79.6\%$)Hourly wage, mean (SndEr)$10.10 (0.2)$9.61 (0.1)$11.25 (0.2)$9.17 (0.1)$11.00 (0.3)$10.49 (0.1)Has health insurance, n (%)411 ($33.4\%$)330 ($25.8\%$)497 ($46.2\%$)246 ($33.8\%$)518 ($35.2\%$)1,716 ($45.3\%$)Born in the U.S., n (%)398 ($32.4\%$)401 ($31.3\%$)483 ($44.9\%$)189 ($26.0\%$)275 ($18.7\%$)317 ($8.4\%$)Education level, mean (SndEr)9.0 (0.3)8.5 (0.2)10.5 (0.3)8.0 (0.3)8.1 (0.2)7.6 (0.1)Mean and standard error calculated using NAWS weights to adjust for sample designNE Northeast, SE Southeast, MW Midwest, SW Southwest, NW Northwest, CA California
## Primary Outcomes
Farmworkers in all regions were more likely to report having difficulty accessing health care in the U.S. because it was too expensive (between 20.1 and $36.1\%$) compared to language or transportation problems (0.9–3.9 and 0.6–$1.9\%$, respectively). The total count of farmworkers in all regions who reported having difficulty accessing health care because of expense was 2568, compared to 163 who reported having difficulty because of language barriers and 85 who reported having difficulty because of transportation or services were too far away.
The odds of farmworkers having difficulty accessing health care because it was too expensive for each region compared to CA can be found in Table 2. After adjusting for survey weights, age, gender, income, and health insurance status, the odds of reporting difficulty accessing care because of cost were lower in regions other than CA, and the results were significant in all regions except for in the SE region where the OR was 0.92 with $95\%$ CI (0.79, 1.08).
Table 2Association between regions and having difficulty accessing health care because it is too expensive, NAWS 2013–2016 ($$n = 9577$$)Count (%)aOR unadjusted model ($95\%$ CI)OR adjusted modelb ($95\%$ CI)OR sensitivity analysisc ($95\%$ CI)NE291 ($23.8\%$)0.88 (0.75, 1.02)0.72 (0.61, 0.84)0.72 (0.62, 0.85)SE459 ($36.1\%$)1.21 (1.04, 1.40)0.92 (0.79, 1.08)0.92 (0.79, 1.07)MW215 ($20.1\%$)0.66 (0.56, 0.77)0.69 (0.58, 0.82)0.68 (0.57, 0.80)SW183 ($25.3\%$)0.87 (0.72, 1.05)0.70 (0.58, 0.86)0.72 (0.59, 0.87)NW374 ($25.5\%$)0.95 (0.83, 1.09)0.73 (0.63, 0.85)0.75 (0.64, 0.87)CA1,046 ($27.8\%$)ReferenceReferenceReferenceNE Northeast, SE Southeast, MW Midwest, SW Southwest, NW Northwest, CA Californiaa53 observations had missing outcome databAdjusted for age, gender, income, and health insurance statuscOdds ratio for adjusted model with imputed data for missing values among covariates Results for farmworkers having difficulty accessing health care because providers do not speak farmworkers’ language are in Table 3. The adjusted odds that farmworkers in the NE and SE region reported having difficulty accessing health care because of language barriers was 1.66 ($95\%$ CI 0.98, 2.79) and 3.38 ($95\%$ CI 2.19, 5.22), respectively. Adjusted odds were lower in the MW but higher in all other regions than CA.
Table 3Association between regions and having difficulty accessing health care because providers do not speak farmworker’s language, NAWS 2013–2016 ($$n = 9577$$)Count (%)aOR unadjusted model ($95\%$ CI)OR adjusted modelb ($95\%$ CI)OR sensitivity analysisc ($95\%$ CI)NE23 ($1.9\%$)1.95 (1.17, 3.25)1.66 (0.98, 2.79)1.63 (0.97, 2.73)SE49 ($3.9\%$)4.45 (2.91, 6.79)3.38 (2.19, 5.22)3.31 (2.16, 5.10)MW10 ($0.9\%$)0.68 (0.32, 1.43)0.91 (0.43, 1.93)0.85 (0.40, 1.80)SW12 ($1.7\%$)1.52 (0.78, 2.98)1.31 (0.66, 2.61)1.29 (0.64, 2.55)NW20 ($1.4\%$)1.25 (0.71, 2.18)1.16 (0.65, 2.04)1.12 (0.64, 1.97)CA49 ($1.3\%$)ReferenceReferenceReferenceNE Northeast, SE Southeast, MW Midwest, SW Southwest, NW Northwest, CA Californiaa55 observations had missing outcome databAdjusted for age, sex, income, and health insurance statuscOdds ratio for adjusted model with imputed data for missing values among covariates Table 4 shows the estimates for farmworkers’ difficulty accessing health care because it is too far away or lacks transportation. The adjusted odds were higher in CA than in all regions except for the SW, where it was 1.02 ($95\%$ CI 0.50, 2.08), compared to CA. The adjusted OR was significantly lower in the NE, MW, and NW.
Table 4Association between regions and having difficulty accessing health care because it is too far away or farmworkers do not have transportation, NAWS 2013–2016 ($$n = 9577$$)Count (%)aOR unadjusted model ($95\%$ CI)OR adjusted modelb ($95\%$ CI)OR sensitivity analysisc ($95\%$ CI)NE7 ($0.6\%$)0.43 (0.19, 1.00)0.31 (0.13, 0.73)0.32 (0.13, 0.74)SE18 ($1.4\%$)0.82 (0.43, 1.58)0.55 (0.27, 1.13)0.66 (0.34, 1.29)MW9 ($0.8\%$)0.24 (0.08, 0.70)0.28 (0.09, 0.86)0.29 (0.10, 0.88)SW14 ($1.9\%$)1.15 (0.57, 2.30)1.02 (0.50, 2.08)1.03 (0.50, 2.10)NW13 ($0.9\%$)0.44 (0.21, 0.96)0.36 (0.17, 0.78)0.36 (0.17, 0.79)CA24 ($0.6\%$)ReferenceReferenceReferenceNE Northeast, SE Southeast, MW Midwest, SW Southwest, NW Northwest, CA Californiaa61 observations had missing outcome databAdjusted for age, sex, income, and health insurance statuscOdds ratio for adjusted model with imputed data for missing values among covariates
## Sensitivity Analysis
There was a total of 53, 55, and 61 total observations that had missing data for the outcomes too expensive, language barriers, or transportation barriers, sequentially. There were three observations of missing data for health insurance status, four that were missing age data, and 206 that were missing hourly wage data. Sensitivity analysis resulted in adjusted OR estimates similar (within 0.11) to estimates found in the primary analysis.
## Discussion
We calculated the odds of specific barriers that cause farmworkers to have difficulty accessing health care in the U.S. Our study found regional variations in both the existence of barriers like cost, transportation, and language and demographic differences, including education, immigration, income, and health insurance status. Understanding regional disparities in the presence of barriers to accessing health care among farmworkers is an essential step to improving equitable health care access in the U.S. Variations between and within national regions may directly result from the nature of work within each region.
The overall OR for difficulty accessing health care was lower in the MW than all other regions for each potential barrier after adjusting for age, gender, income, and health insurance status. In the MW, farmworkers showed higher income, education, health insurance rates and were more likely to be born in the U.S. than farmworkers in all other regions. There were multiple patterns identified within the barriers to accessing health care. These patterns included lower odds of farmworkers employed in the MW reporting any of the barriers in this study compared to other regions. The disproportionate amount of MW farmworkers born in the U.S. compared to other regions could explain the lower odds of experiencing barriers. In line with previous research, English proficiency, health insurance status, education, and cultural differences influence health care access [20]. Being U.S.-born increases access to health insurance, higher wages, and education which in turn lessen barriers [21].
Farmworkers employed in the SE had greater difficulty accessing health care because of language barriers than farmworkers in other regions. Hoerster and colleagues found that individual-level factors, including gender, immigration status, English proficiency, transportation, and use of services outside the U.S., impacted farmworkers’ ability to access care [9]. In the U.S., speaking the English language increases the ability to identify and access health care services [22]. Unsurprisingly, when health care services are obtained, language barriers can result in negative perceptions of health care experiences and quality of care [22]. Language barriers not only affects health care disparities but also makes receiving preventative health information challenging [22].
Farmworkers employed in CA had difficulty accessing health care in the U.S. because it was too expensive or far away more than farmworkers in other regions. Not counting on health insurance causes concern to afford health care. In California, seven in ten adults are “somewhat” or “very” worried about their ability to afford care [23]. Many Californians rely on public programs such as Medicare and Medi-Cal (California’s Medicaid program) to support vulnerable populations. While the expansion of the Affordable Cares Act (ACA) helped increase enrollment many individuals still do not count on coverage [23]. On the positive end, California recently changed their Medi-Cal program to expand access to undocumented individuals, meaning it is likely many low-income farmworker families qualify for assistance through this modification [24]. It is possible that this change can help make health care services more affordable for vulnerable populations including farmworkers.
## Strengths
One strength of the study was using the NAWS dataset, implemented for over 30 years, providing a representative sample of farmworkers in the U.S. This study included nearly 10,000 participants, with the smallest region representing over 700 total participants. Only 274 ($2.8\%$) participants were missing outcome or covariate data, and our sensitivity analysis suggested missing data was not a large factor in our results. Farmworkers are also an understudied population, and more studies are needed to improve the understanding of barriers farmworkers encounter when accessing health care [9].
## Limitations
This study focused on regional differences among farmworkers; however, challenges to accessing health care within each region may vary. Availability of state and county-level data could improve the understanding of geographical variations of barriers among farmworkers. While NAWS does use these variables during collection, employer and farm type data are not publicly available for analysis. Another limitation was that the collection relied on the self-reported outcomes and variable data from on-site interviews. The presence of other farmworkers or supervisors may influence survey responses. Additionally, the NAWS only collects data on farmworkers engaged in crop-related labor in the past year and does not represent farmworkers focused on other farm labor like mechanics or farmworkers who have been out of the labor force for over a year. Farmworkers also reported that they were not aware of barriers to accessing health care in the U.S. because they had never needed to access health care in the U.S. leading to concerns of underutilization of services that available to underserved communities. Lastly, the is limited research findings on differences of health care access by region. While much of the farmworker literature have common findings, there are not many findings that are specific to a region with the exception of California that is home to many farmworker-based research projects. Other states could benefit from their methodology to have precise needs for their farmworker communities which can help address health inequalities.
## Conclusion
A comprehensive strategy that incorporates individual and policy-level changes to expand access to health care for farmworkers should factor and tailor programs and interventions appropriate for farmworkers based on individual, community, state-level, and regional characteristics. For instance, other regions should look towards the MW and study the current policies in place that positively impact the low barriers to health care such as wages and employer-based health coverage. The SE could observe other region’s health care systems and include creative methods to help alleviate the language barrier between patients and providers, such as increasing the use of bilingual and bi-cultural community health workers. While California had higher reports of inability to afford health care cost, they are actively taking steps to attempt to reduce this barrier by modifying policies. Farmworkers who do not have health insurance and struggle to afford health care could be referred to Federally Qualified Health Centers (FQHCs) or Migrant Clinics. Yet, often vulnerable populations rely on emergency care as their main form of health care because they worry about the cost of routine care. Increased education on sliding fee scales can help bring awareness to a population who may not know of these local medical homes that are low cost and, in the long-run, more affordable than the emergency room. Furthermore, expanding access to affordable health care plans, mobile clinics, and other policies and programs that make the health system easier to navigate would help address barriers to accessing health care among farmworkers [9, 10]. Understanding regional disparities in the presence of barriers to accessing health care among farmworkers is an essential step to improving equitable access to care and highlight possible leads to issues that need to be addressed in their respective area.
## References
1. 1.American Public Health Association (2017). Improving working conditions for U.S. farmworkers and food production workers. https://www.apha.org/policies-and-advocacy/public-health-policy-statements/policy-database/2018/01/18/improving-working-conditions.
2. McCoy HV, Williams ML, Atkinson JS, Rubens M. **Structural characteristics of migrant farmworkers reporting a relationship with a primary care physician**. *J Immigr Minor Health* (2016.0) **18** 710-4. DOI: 10.1007/s10903-015-0265-2
3. 3.National Center for Farmworker Health (2018). Agricultural worker demographics. http://www.ncfh.org/uploads/3/8/6/8/38685499/fs_demographics_2018.pdf.
4. Hansen E, Donohoe M. **Health issues of migrant and seasonal farmworkers**. *J Health Care Poor Underserved* (2003.0) **14** 153-64. DOI: 10.1353/hpu.2010.0790
5. Ramos AK. **A human rights-based approach to farmworker health: an overarching framework to address the social determinants of health**. *J Agromedicine* (2018.0) **23** 25-31. DOI: 10.1080/1059924X.2017.1384419
6. TePoel M, Rohlman D, Shaw M. **The impact of work demand and gender on occupational and psychosocial stress in hispanic farmworkers**. *J Agric Saf Health* (2017.0) **23** 109-23. DOI: 10.13031/jash.11753
7. Marsh B, Milofsky C, Kissam E, Arcury TA. **Understanding the role of social factors in farmworker housing and health**. *New Solut* (2015.0) **25** 313-33. DOI: 10.1177/1048291115601020
8. Baker D, Chappelle D. **Health status and needs of Latino dairy farmworkers in Vermont**. *J Agromed* (2012.0) **17** 277-87. DOI: 10.1080/1059924X.2012.686384
9. Hoerster KD, Mayer JA, Gabbard S. **Impact of individual-, environmental-, and policy-level factors on health care utilization among US farmworkers**. *Am J Public Health* (2011.0) **101** 685-92. DOI: 10.2105/AJPH.2009.190892
10. Tulimiero M, Garcia M, Rodriguez M, Cheney AM. **Overcoming barriers to health care access in rural Latino communities: an innovative model in the Eastern Coachella valley**. *J Rural Health* (2021.0) **37** 635-44. DOI: 10.1111/jrh.12483
11. Moyce S, Hernandez K, Schenker M. **Diagnosed and undiagnosed diabetes among agricultural workers in California**. *J Health Care Poor Underserved* (2019.0) **30** 1289-301. DOI: 10.1353/hpu.2019.0102
12. Becot F, Inwood S, Bendixsen C, Henning-Smith C. **Health care and health insurance access for farm families in the United States during COVID-19: essential workers without essential resources?**. *J Agromed* (2020.0) **25** 374-7. DOI: 10.1080/1059924X.2020.1814924
13. Perreira KM, Pedroza JM. **Policies of exclusion: implications for the health of immigrants and their children**. *Annu Rev Public Health* (2019.0) **40** 147-66. DOI: 10.1146/annurev-publhealth-040218-044115
14. Quandt SA, Brooke C, Fagan K, Howe A, Thornburg TK, McCurdy SA. **Farmworker housing in the United States and its impact on health**. *New Solut* (2015.0) **25** 263-86. DOI: 10.1177/1048291115601053
15. Villarejo D. **The health of U.S. hired farm workers**. *Annu Rev Public Health* (2003.0) **24** 175-93. DOI: 10.1146/annurev.publhealth.24.100901.140901
16. 16.Department of Labor. Employment and training administration. Methodology. dol.gov. n.d. https://www.dol.gov/agencies/eta/national-agricultural-workers-survey/methodology.
17. 17.Department of labor, employment and training administration. Overview. dol.gov. n.d. https://www.dol.gov/agencies/eta/national-agricultural-workers-survey/overview.
18. 18.Interstate migrant education council symposium. Presentation: changing trends in crop agriculture and migrant crop workers (2015). Dol.gov. https://www.dol.gov/sites/dolgov/files/ETA/naws/pdfs/IMEC_pres_Oct2015.pdf.
19. 19.SAS/STAT® 141 User’s Guide2015North CarolinaSAS Institute Inc. *1 User’s Guide* (2015.0)
20. Arcury TA, Quandt SA. **Delivery of health services to migrant and seasonal farmworkers**. *Annu Rev Public Health* (2007.0) **28** 345-63. DOI: 10.1146/annurev.publhealth.27.021405.102106
21. Kao D, Park J, Min S, Myers D. **Occupational status and health insurance among immigrants: effects by generation, length of residence in U.S., and race**. *J Immigr Minor Health* (2010.0) **12** 290-301. DOI: 10.1007/s10903-008-9189-4
22. 22.Garcia CM, Duckett LJ. NO TE ENTIENDO Y TÚ NO ME, ENTIENDES: LANGUAGE BARRIERS AMONG IMMIGRANT LATINO ADOLESCENTS SEEKING HEALTH CARE. J Cult Divers. 2009;16(3):120–6.
23. 23.Public Policy Institute of California. California’s Future: Health Care. 2020. https://www.ppic.org/wp-content/uploads/californias-future-health-care-january-2020.pdf.
24. 24.Department of Health Care Services. Older Adult Expansion. Ca.gov. 2022. https://www.dhcs.ca.gov/services/medi-cal/eligibility/Pages/OlderAdultExpansion.aspx.
|
---
title: Adherence to glucose monitoring with intermittently scanned continuous glucose
monitoring in patients with type 1 diabetes
authors:
- Carolina Sousa
- João Sérgio Neves
- Cláudia Camila Dias
- Rute Sampaio
journal: Endocrine
year: 2022
pmcid: PMC9988994
doi: 10.1007/s12020-022-03288-1
license: CC BY 4.0
---
# Adherence to glucose monitoring with intermittently scanned continuous glucose monitoring in patients with type 1 diabetes
## Abstract
### Purpose
This study aims to predict the Intermittently scanned continuous glucose monitoring (isCGM) adherence behavior of patients with Type 1 Diabetes.
### Methods
Patients with Type 1 *Diabetes mellitus* using FreeStyle Libre™ System (FL), a isCGM device, that attended the “Insulin Infusion Pump clinic” at Centro Hospitalar de São João were enrolled and evaluated for sociodemographic and clinical characterization, beliefs and concerns about Diabetes Mellitus, as well as isCGM’s perceptions. Intermittently scanned continuous glucose monitoring data were collected to characterize monitoring patterns and to measure isCGM’s adherence—FL average of scans/day.
### Results
Seventy-two patients with a mean of 30.36 years (sd=11.35) participate in this study. A median of 7 scans/day was performed. The adherence predictors found was Age (β = 0.191, $$p \leq 0.006$$), Time in target (β = 0.530, $$p \leq 0.002$$), isCGM Necessity (β = 2.631, $$p \leq 0.048$$), Body Mass Index (β = −0.549, $$p \leq 0.017$$) and Sex (β = −3.996; $$p \leq 0.011$$).
### Conclusions
This study emphasizes the relevance of glucose monitoring adherence in disease control and shows that males of younger ages, presenting with higher body mass index levels, lower time in target, and reporting lower isCGM necessity are less adherent to isCGM. Therefore, these patients should be closely followed and object of personalized strategies to promote treatment adherence.
## Introduction
Diabetes is a worldwide health concern, affecting more than 536 million patients in 2021, and is estimated to reach 783 million in 2045 [1]. In Portugal, this condition is particularly prevalent, as in the latest data available in 2018, it affects $13.6\%$ of the Portuguese population between 20 and 79 years, representing $8\%$ of Portugal´s health expenses [2].
Treatment adherence is essential to maintain adequate metabolic control and avoid diabetes complications. Therapeutic adherence is defined as “the extent to which a person’s behavior (in terms of taking medications, following diets, or executing lifestyle changes) coincides with medical or health advice” [3]. In chronic conditions such as diabetes, treatment usually follows a complex plan which combines the patient’s education to recognize hyper and hypoglycemia symptoms [4] and self-care behaviors—such as glucose monitoring, diet, medication, regular physical activity, foot care, and regular medical visits [5]. These therapeutic interventions require the commitment of the patient and an active role in managing his or her disease [6].
Measuring treatment adherence in diabetes can be challenging, as it must assess the adherence level in each treatment regimen independently component [3]. This brings challenges like the scarcity of appropriate methods to evaluate adherence to insulin treatment [7] or variations in recommended blood glucose monitoring frequencies for patients with type 1 diabetes [3]. Hence, some studies showed low adherence to different therapy components [3, 8, 9].
Frequent glucose monitoring is recommended for patients treated with insulin – all patients with type 1 diabetes mellitus (T1D) and some patients with type 2 diabetes mellitus (T2D) - so they can adjust insulin dose, depending on factors such as diet and exercise [10]. For many years, patients had to use test strips and finger-stick blood samples to monitor glucose levels [10]. However, this method has barriers that can be reduced with continuous glucose monitoring (CGM) and intermittently scanned continuous glucose monitoring (isCGM), particularly benefiting patients unable or unwilling to self-monitor blood glucose (SMBG) due to pain or discomfort [11]. FreeStyle Libre™ System (FL) is an isCGM that also overcomes some CGM barriers. It doesn´t constantly update the glucose measurements, like CGM, but the current glucose value is quickly available when needed [12]; it doesn’t require calibration; has an extended sensor lifetime of 14 days; and is relatively affordable [13]. These advantages may explain patient satisfaction and the market expansion of this device [13]. It is important to emphasize that isCGM also shows higher monitoring rates when compared to SMBG [13, 14], which can be associated with improved glycaemic control [13].
Therefore, given its increasing usage, it´s urgent to understand the adherence patterns of isCGM´s users and establish possible associations with sociodemographic and clinical variables, isCGM monitoring variables, believes and concerns about Diabetes Mellitus, and isCGM perceptions. More specifically, using a conceptual framework of Illness and Treatment Representations [15] (which focuses on how a patient’s beliefs and expectations about an illness determine a person’s appraisal and coping with perceiving it as manageable or threatening) and understanding the *Necessities versus* Concerns [16, 17] about self-monitoring, this study aims at identify patients that won´t adhere to isCGM and why, so clinicians can implement strategies to enhance its use and reduce diabetes morbidity.
## Study design
This is an observational and retrospective study performed between 20th October 2021 and 1st June 2022 at the Centro Hospitalar de São João, E.P.E. - Hospital de São João.
## Participants
The participants were adult patients with Type 1 *Diabetes mellitus* using FreeStyle Libre (FL), who presential attended the “Insulin Infusion Pump clinic” and signed the informed consent to participate. The exclusion criteria were: inability to communicate in Portuguese; presenting psychiatric and cognitive disorders precluding the interview; using other devices for reading FL; less than a month since FL’s first prescription; and without records on LibreView System.
Of the 158 patients followed in the “Insulin Infusion Pump clinic”, 91 met the inclusion criteria. Of these, 1 refused and 18 patients were excluded: 13 for using other devices to read the FL, 4 with missing information in LibreView; 1 with less than one month since the FL prescription.
## Data collection
Data was collected by asking direct clinical questions to patients and from medical records, by application of questionnaires and Libre View consultation, respectively. All participants were informed about the study objectives and data collection procedures before being invited to sign an informed consent form which included authorization to use the gathered information.
## Instruments
A questionnaire was given to all the participants. It addressed sociodemographic and clinical questions, questions based on the Brief Illness Perception Questionnaire (Brief IPQ) [18], and an Intermittently scanned Continuous Glucose Monitoring questionnaire based on some of the factors that contribute to adherence [3]. Were also collected isCGM Monitoring Data by LibreView platform.
## Sociodemographic questionnaire
Information about sex, age, marital status, education level, and the professional situation was gathered using this questionnaire.
## Clinical questionnaire
Data about diabetes course- age at diagnosis and T1D duration, comorbidities, hospital or emergency department admissions; body mass index (BMI); method of insulin administration (pump/pen); HbA1c at the consult, was obtained using a structured clinical questionnaire. To complete the missing data, medical records were used.
## Brief illness perception questionnaire (Brief IPQ) [18]
Brief IPQ is a validated nine-item scale designed to rapidly assess illness’s cognitive and emotional representations [18], using a 10-point Likert scale, with a total of eight items. Five of the items assess cognitive illness representations: consequences (item 1), timeline [2], personal control [3], treatment control [4], and identity [5]; two of the items assess emotional representations: concern [6] and emotions [8]; one item assesses illness comprehensibility [7]. High scores (total result) reveal a more threatening perception of the illness.
In the present study, we did not use the timeline (item 2), as diabetes is a chronic disease without a fully understood etiology related to genetic and environmental factors [3]. Treatment control (Item 4) also wasn´t implemented as insulin is crucial and the only available treatment.
## Intermittently scanned continuous glucose monitoring questionnaire
Based on the dimensions defined by WHO to understand the phenomenon of adherence [3], information about some of the most notable therapy-related factors was collected by a 5-point Likert scale questionnaire, namely: Information about isCGM; isCGM efficacy and sufficient in disease control; isCGM necessity; Patient satisfaction with isCGM monitoring; self-efficacy in isCGM monitoring; concerns with isCGM use; familiar support in isCGM monitoring. It was also asked about the individual perception of the average frequency of isCGM scans per day and if they consider that number enough.
## Intermittently scanned continuous glucose monitoring data
In this study, it was used the FreeStyle Libre device for glucose monitoring, and to access monitoring data, the LibreView platform. This platform keeps the information about glucose monitoring and patterns in a cloud so professionals and patients can access their account and create reports, facilitating diabetes follow-up.
On the day of the medical appointment, data from the previous 28 days were accessed: the average scans/day with FreeStyle Libre; percentage of time in glucose target, below and above; average glucose; the number of low glucose events; and glucose variability. As the FL sensor has a 14 days lifetime, to prevent possible bias with sensor application adherence, patients with a gap between sensors exchange higher than three days, we collected data only from the 14 days after the sensor application instead of 28 days.
The average of scans/day with Freestyle Libre- monitoring frequency- was used as an accurate way to measure adherence to isCGM.
## Ethics
This study had the permission of the Centro Hospitalar de São João Ethics Committee, with approval number $\frac{254}{21}$, and all patients enrolled gave their written consent after they were given the information about the study.
## Statistical analyses
Categorical variables were described as absolute frequencies (n) and relative frequencies (%). Medians and percentiles were used for continuous variables.
When testing a hypothesis about continuous variables, nonparametric tests Mann–were used as appropriate, taking into account normality assumptions and the number of groups compared; when testing a hypothesis about categorical variables, a chi-square test and Fisher’s exact test were used, as appropriate. To understand the adherence patterns of isCGM´s users and identify possible associations with sociodemographic and clinical variables, univariate and multivariate linear regression modeling was used. Coefficient regression (beta), $95\%$ confidence intervals ($95\%$ CI) and R2 as a measure of goodness of fit were presented. Models were built according to the backward stepwise approach.
The significance level used was 0.05.
Statistical analysis was performed using the software Statistical Package for the Social Sciences v. 27.0.
## Sociodemographic characterization
The socio-demographic data is presented in Table 1. The final sample has 72 patients, 29 males and 43 females. The age of the participants ranged from 18 to 66, with a mean of 30.36 years (sd=11.35). The majority of patients completed high school or college ($93.1\%$), had a full-time job ($65.3\%$), and were single ($57.7\%$).Table 1Sociodemographic and clinical dataVariableSex, n(%) Male29 (40.3) Female43 (59.7)Age Range (years)18–66 Mean (SD)30.4 (11.4) Median (P25-P75)25.50 (21–37) Years n(%) 18–2939 (54.2) 30–3918 [25] 40–499 (12.5) ≥506 (8.3)Education, n(%)a Low5 (6.9) High67 (93.1)Profession, n(%) Student21 (29.2) Unemployed3 (4.2) Full Time Job47 (65.3) Retired1 (1.4)Marital Status, n(%)b Single41 (57.0) Married30 (41.6)Insulin administration, n(%) Pen17 (23.6) Pump55 (76.4)BMI Range (kg/m2)18.3–32.0 Mean (SD)24.6 (3.4) Median (P25-P75)23.71 (22.0–27.0) Categories n(%)c Underweight/ Normal42 (58.3) Overweight22 (30.6) Obese8 (11.1)Age at diagnosis Mean (SD)13.7 (9.4) Median (P25-P75)11 (8–19)T1D duration (years) Mean (SD)16.6 (8.3) Median (P25-P75)15.5 (9.2–22.0)HbA1c Mean (SD)7.51 (0.96) Median (P25-P75)7.3 (6.9–8.1)Past year hospital or ED admissions, n (%) Yes7 (9.7) No65 (90.3)Diabetes comorbidities, n (%) Yes15 (20.8) No57 (79.2)BMI body mass index, T1D type 1 diabetes, ED emergency department, SD standard deviation, P25-P75 25th percentile and 75th percentile (representing the interquartile range) aLow: up to 9 gradebOne non responsecUnderweight/ Normal <25 kg/m2; Overweight 25–29,9 kg/m2; Obese >29,9 kg/m2
## Clinical characterization
The clinical characterization is presented in Table 1. All participants were followed at the “Insulin Infusion Pump clinic”, however, only 55 ($76.4\%$) used an insulin pump, and the remaining 17 patients ($23.6\%$) used insulin pens. The majority had a low or normal BMI ($58.3\%$), $30.6\%$ of the patients had overweight, and $11.1\%$ had obesity. The mean age at diagnosis was 13.7 years and T1D duration was 16.6 years. HbA1c had an average of $7.5\%$. When asked if they had diabetes-related comorbidities, $20.8\%$ responded yes, and $9.7\%$ of the patients reported hospital or emergency department (ED) admissions in the previous year.
## Intermittently scanned continuous glucose monitoring data (isCGM)
The median (P25-P75) percentage of time in glucose target, above and below target, was respectively $51\%$ (39.25–66), $43\%$ (28–56) and $4\%$ (2–7). The patient’s average glucose had a median of 172 mg/dL (149–199), with a variation of glucose of $40.65\%$ (35.7–44.4) and a median number of low glucose events of 15 (6–26).
## Adherence patterns of isCGM
IsCGM adherence was measured by the FL monitoring frequency of the patient, and it showed a median (P25-P75) of 7 scans/day (5–12).
There were not significant differences between monitoring frequency and sociodemographic characteristics of the sample. Neither the presence of comorbidities, the ED or hospital admissions, nor the way of insulin administration had significant monitoring frequency differences between groups. However, patients with obesity had a lower monitoring frequency than underweight/normal BMI ($$p \leq 0.012$$), with a median of 3.5 versus eight scans/day.
The perceived adherence to isCGM monitoring, given by the average perceived number of scans per day, had a median (P25-P75) of 8 scans/day (5.25–10), where $11.3\%$ of the patients had the same number of perceived and effective scans, but $40.8\%$ underestimate and $47.9\%$ overestimate this number. Patients that overestimated the adequate number of scans/day had higher BMI levels ($$p \leq 0.005$$), higher HbA1C ($$p \leq 0.028$$), and lower scores on the emotional item ($$p \leq 0.011$$) than the ones that underestimated this number. The majority of the patients, $68.1\%$, believed the number of scans/day performed was enough for diabetes control. This same group had higher isCGM adherence when compared to patients who believed their number of scans was not enough ($$p \leq 0.016$$).
When analyzing the association between monitoring frequency and different study variables, it was found a positive and statistically significant association with age (β = 0.21, $$p \leq 0.011$$) and percentage of time in glucose target (β = 0.28, $p \leq 0.001$); and also a negative and significative relation with HbA1c (β = −3.43, $p \leq 0.001$), BMI (β = −0.62, $$p \leq 0.025$$), average glucose (β = −0.09, $p \leq 0.001$), percentage of time with glucose above target (β = −0.22, $p \leq 0.001$) and glucose variability (β = −0.34, $$p \leq 0.013$$).
## Cognitive and emotional representations of diabetes and adherence to isCGM
The assessment of cognitive diabetes representation resulted in a median (P25-P75) score of 5 (3–6) points for the disease consequences representation, 3 (2–4) in personal control and 5.50 (4–7) in identity; illness comprehensibility had a median score of 2 (1–2); the emotional representation of diabetes resulted in a median score of 8 (6–9) in disease concern and 6 (2–8) in emotional item.
It was not found a significant difference between cognitive or emotional illness representation and sociodemographic variables. However, patients with underweight/normal BMI showed better diabetes comprehensibility, with a median score of 9 points, when compared to patients with overweight or obesity, that had both a median of 8 points, respectively $$p \leq 0.013$$ and $$p \leq 0.011.$$ The presence of comorbidities or how insulin is administered didn’t show differences in illness representation. However, patients with ED or hospital admissions in the past year had higher scores on cognitive illness representation, revealing them to be more threatened by diabetes ($$p \leq 0.007$$).
When adjusting for Brief IPQ questions, a statistically significant and positive association was found between monitoring frequency and the perception of illness consequences (β = 1.07, $$p \leq 0.017$$).
## Perceived isCGM necessity and concerns and adherence to isCGM
Patients gave a median (P25-P75) score of 5 points for isCGM necessity (4–5); 4 for isCGM monitoring information (4–5), efficacy and sufficiency (4–4), self-efficacy (4–5), satisfaction (4–5), and familial support (3–5); and 2 for concern (1–3).
When analyzing differences between perceived isCGM necessity and concerns and sociodemographic and clinical characteristics, it wasn´t found relevant associations.
When adjusting for isCGM monitoring questions, a statistically significant association was found between monitoring frequency- adherence- and the isCGM necessity (β = 4.825, $$p \leq 0.002$$) and satisfaction (β = −2.57, $$p \leq 0.037$$).
## Predictors of adherence to isCGM
A multivariate regression was performed, adjusting for variables: Sex, BMI, Age, HbA1c, Time in target, Time above target, Average glucose, Glucose variability, Illness consequences and identity, isCGM necessity, and isCGM satisfaction. The results of this analysis are presented in Table 2.Table 2Predictors of adherence to isCGMMeasurementsβCI $95\%$ [Inferior, Upper Limit]pSex FemaleRef Male−3.996[−7.029, −0.962]0.011BMI, kg/m2−0.549[−0.995, −0.103]0.017Age, years0.191[0.057, 0.325]0.006Time in target, %0.530[0.196, 0.864]0.002Time above target, %0.290[−0.002, 0.583]0.052isCGM necessity2.631[0.026, 5.235]0.048isCGM satisfaction−2.020[−4.042, 0.002]0.050Constant−24.965[−60.118, 10.189]0.161Dependent variable: average scans/day with FreeStyle Libre; Independent variables: Sex, BMI, Age, Age at diagnosis, Time in target, Time above target, Illness consequences, Identity, FL necessity, FL satisfaction; Statistic methods: BACKWARDS; R2 = 0.457Β beta, BMI body mass index, isCGM intermittently scanned continuous glucose monitoring Men had worst isCGM adherence compared to women, with almost four fewer scans/day (β = −3.996; $$p \leq 0.011$$). Patients with higher BMI (β = −0.549, $$p \leq 0.017$$) showed to be less adherent to isCGM monitoring. On the contrary, older patients (β = 0.191, $$p \leq 0.006$$) and patients with a higher percentage of time in glucose target (β = 0.530, $$p \leq 0.002$$) demonstrated a higher isCGM monitoring adherence. Perceived isCGM necessity also positively related to monitoring frequency, showing that patients who reported higher isCGM necessity had better adherence rates (β = 2.631, $$p \leq 0.048$$).
## Discussion
To our knowledge, this is the first study aiming to create a model able to predict the adherence behavior to self-monitoring with intermittently scanned continuous glucose monitoring and therefore leading physicians to implement adherence strategies in patients with lower adherence patterns.
The patients enrolled in this study showed a lower glucose monitoring rate median (7 daily scans), when compared to more extensive studies like the Real-world Flash Glucose Monitoring study [13], with a median of 14 daily scans, or Flash Glucose Monitoring in Israel [19], with a median of 12 daily scans. Nevertheless, this study reaffirms the significative relation between glucose markers and monitoring rate, where patients with better HbA1C, time in target, time above target, glucose variability and average glucose had a higher glucose monitoring rate [13, 19–21].
BMI showed an important relation with isCGM adherence. Underweight/ normal BMI patients may be more concerned with following a healthier diet and practicing exercise- leading to better diabetes self-management [22] and, as acknowledged in our results, higher diabetes comprehensibility. This may explain why this population is more adherent in glucose management when compared with patients with obesity.
Previous works showed lower therapy adherence in patients with a more extended history of diabetes [3] and, consequently, the worst glucose monitoring adherence. However, in our study, there wasn’t a significant relation between monitoring frequency and TD1 duration. Nonetheless, older patients had better monitoring rates, an association also shown for the SMBG monitoring frequency [9].
Illness perceptions are shaped by past experiences and illness-related episodes [23], so it’s reasonable that patients with past year ED or hospital admissions revealed higher expectations of diabetes effects, higher perception of diabetes symptoms (data not shown), and also to be more threatened by this disease. Although this group of patients didn’t show better adherence patterns to glucose monitoring, it seems that a higher perception of illness consequences leads to better adherence.
IsCGM necessity was revealed to be an essential factor in monitoring adherence, as an extra point on the question about device necessity was associated with almost five more scans/day. Unexpectedly, and despite some studies showing positive correlations between more frequent monitoring and device satisfaction [24, 25], patients reporting less satisfaction in isCGM monitoring demonstrated better adherence. We hypothesize that fewer satisfaction levels could be related to lower precision of glucose value compared to SMBG, leading to an increase in scan rate and therefore explaining this relation.
When producing a model to predict isCGM adherence, the “patient pattern” of lower isCGM monitoring adherence is characterized as younger males with higher BMI levels, lower time in target, and less perception of isCGM necessity. The perception of isCGM necessity probably is where an intervention could be more fruitful, so it is suggested that clinicians must emphasize and create methods to enhance the need for isCGM monitoring in diabetes control.
The strengths of the present study are its real-life setting, focusing only on patients with type 1 diabetes, the knowledge of our sample sociodemographic characteristics, and the unrestricted exclusion criteria applied. All participants had the same information and support about FL management since Centro Hospitalar de São João provides a nurse appointment to explain and applicate the first sensor to every FL user. Clinical data was as complete and accurate as possible, considering that medical records were retrospectively retrieved when needed, and despite FL giving an estimated HbA1c value, we used laboratory values for Hba1C results to obtain the most precise measurement.
Nevertheless, this study has caveats that should be acknowledged. First, the small size of the sample. Although Centro Hospitalar de São João is the largest hospital center in the north region of Portugal, only 158 patients are followed in the “Insulin Infusion Pump clinic”. Also, this study relies on self-report questionnaires that could jeopardize data accuracy. We tried to mitigate this factor with the questionnaires being delivered on hand by an investigator available to clarify any possible doubts of the patients.
To conclude, this study reflects the relevance of glucose monitoring adherence in disease control. Our results could help predict which patients would need more guidance from health professionals to achieve better isCGM adherence. Males of younger ages, presenting with higher BMI levels, lower time in target, and reporting lower isCGM necessity, should be closely followed and require an application of personalized adherence strategies. However, these findings must be interpreted carefully as further studies are needed to confirm the adherence patterns identified.
## References
1. 1.International Diabetes Federation: IDF Diabetes Atlas. https://www.diabetesatlas.org (2021). Accessed 7 october 2022
2. 2.Sociedade Portuguesa de Diabetologia: Diabetes: Factos e Números – O Ano de 2016, 2017 e 2018. In. Lisboa, (2019)
3. 3.Sabaté, E.: In: Adherence to long-term therapies: evidence for action World Health Organization, Geneva, (2003)
4. 4.Kaufman, F.R. Patient Self-Management Education. In: Medical Management of Type 1 Diabetes. pp. 37–48. American Diabetes Association, Alexandria, Virginia (2012)
5. 5.Standards of medical care for patients with diabetes mellitus. Diabetes Care 25(1), 213–229 (2002). 10.2337/diacare.25.1.213
6. Sampaio R, Pereira MG, Winck JC. **A new characterization of adherence patterns to auto-adjusting positive airway pressure in severe obstructive sleep apnea syndrome: clinical and psychological determinants**. *Sleep. Breath.* (2013.0) **17** 1145-1158. DOI: 10.1007/s11325-013-0814-7
7. Stolpe S, Kroes MA, Webb N, Wisniewski T. **A systematic review of insulin adherence measures in patients with diabetes**. *J. Managed Care Specialty Pharm.* (2016.0) **22** 1224-1246. DOI: 10.18553/jmcp.2016.22.11.1224
8. Aladhab R, Alabbood M. **Adherence of patients with diabetes to a lifestyle advice and management plan in Basra, Southern Iraq**. *Int. J. Diabetes Metab.* (2019.0) **25** 100-105. DOI: 10.1159/000500915
9. Moström P, Ahlén E, Imberg H, Hansson P-O, Lind M. **Adherence of self-monitoring of blood glucose in persons with type 1 diabetes in Sweden**. *BMJ Open Diabetes Res. Care* (2017.0) **5** e000342. DOI: 10.1136/bmjdrc-2016-000342
10. 10.Klimek, M., Knap, J., Reda, M., Masternak, M.: History of glucose monitoring: past, present, future. J. Educ. Health Sport (2019). 10.5281/zenodo.3397600
11. 11.Palylyk-Colwell, E., Ford, C. Flash glucose monitoring system for diabetes. In: CADTH Issues in Emerging Health Technologies. pp. 1-13. Canadian Agency for Drugs and Technologies in Health, Ottawa (ON) (2016)
12. Heinemann L, Freckmann G. **CGM Versus FGM; or, continuous glucose monitoring is not flash glucose monitoring**. *J. Diabetes Sci. Technol.* (2015.0) **9** 947-950. DOI: 10.1177/1932296815603528
13. Dunn TC, Xu Y, Hayter G, Ajjan RA. **Real-world flash glucose monitoring patterns and associations between self-monitoring frequency and glycaemic measures: A European analysis of over 60 million glucose tests**. *Diabetes Res. Clin. Pract.* (2018.0) **137** 37-46. DOI: 10.1016/j.diabres.2017.12.015
14. Castellana M, Parisi C, Di Molfetta S, Di Gioia L, Natalicchio A, Perrini S, Cignarelli A, Laviola L, Giorgino F. **Efficacy and safety of flash glucose monitoring in patients with type 1 and type 2 diabetes: a systematic review and meta-analysis**. *BMJ Open Diabetes Res. Care* (2020.0) **8** e001092. DOI: 10.1136/bmjdrc-2019-001092
15. 15.Leventhal, H., Zimmerman, R. and Gutmann, M. Compliance: a self-regulation perspective. In: Gentry, D. (ed.) Handbook of behaviour medicine. pp. 369–434. Pergamon Press, New York (1984)
16. 16.Horne, R.: Treatment perceptions and self-regulation. In: Linda Cameron, H.L. (ed.) The self-regulation of health and illness behaviour. pp. 138–153. Routledge, London (2003)
17. Horne R, Chapman SCE, Parham R, Freemantle N, Forbes A, Cooper V. **Understanding patients’ adherence-related beliefs about medicines prescribed for long-term conditions: a meta-analytic review of the necessity-concerns framework**. *PLoS ONE* (2013.0) **8** e80633. DOI: 10.1371/journal.pone.0080633
18. 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
19. Eldor R, Roitman E, Merzon E, Toledano Y, Alves C, Tsur A. **Flash glucose monitoring in israel: understanding real-world associations between self-monitoring frequency and metrics of glycemic control**. *Endocr. Pract.* (2022.0) **28** 472-478. DOI: 10.1016/j.eprac.2022.02.004
20. 20.Al-Harbi, M.Y., Albunyan, A., Alnahari, A., Kao, K., Brandner, L., El Jammal, M., Dunn, T.C. Frequency of flash glucose monitoring and glucose metrics: real-world observational data from Saudi Arabia. Diabetol. Metab. Syndrome 14(1) (2022). 10.1186/s13098-022-00831-y
21. Lameijer A, Lommerde N, Dunn TC, Fokkert MJ, Edens MA, Kao K, Xu Y, Gans ROB, Bilo HJG, Van Dijk PR. **Flash Glucose Monitoring in the Netherlands: Increased monitoring frequency is associated with improvement of glycemic parameters**. *Diabetes Res. Clin. Pract.* (2021.0) **177** 108897. DOI: 10.1016/j.diabres.2021.108897
22. 22.Cervantes‐Torres, L., Romero‐Blanco, C. Longitudinal study of the flash glucose monitoring system in type 1 diabetics. J. Clin. Nurs. (2022). 10.1111/jocn.16523
23. 23.Leventhal, H., Brissette, I., Leventhal, E.A. The common-sense model of self-regulation of health and illness. In: The self-regulation of health and illness behaviour. pp. 42–65. Routledge, New York, NY, US (2003)
24. Tansey M, Laffel L, Cheng J, Beck R, Coffey J, Huang E, Kollman C, Lawrence J, Lee J, Ruedy K, Tamborlane W, Wysocki T, Xing D. **Satisfaction with continuous glucose monitoring in adults and youths with Type 1 diabetes**. *Diabet. Med.* (2011.0) **28** 1118-1122. DOI: 10.1111/j.1464-5491.2011.03368.x
25. Al Hayek AA, Robert AA, Al Dawish MA. **Differences of freestyle libre flash glucose monitoring system and finger pricks on clinical characteristics and glucose monitoring satisfactions in type 1 diabetes using insulin pump**. *Clin. Med. Insights: Endocrinol. Diabetes* (2019.0) **12** 117955141986110. DOI: 10.1177/1179551419861102
|
---
title: The pattern of epistaxis recurrence in patients taking prophylactic acetylsalicylic
acid (ASA) from a 10 year cohort
authors:
- Petar Stanković
- Stephan Hoch
- Stefan Rudhart
- Stefan Stojković
- Thomas Wilhelm
journal: European Archives of Oto-Rhino-Laryngology
year: 2022
pmcid: PMC9988998
doi: 10.1007/s00405-022-07666-3
license: CC BY 4.0
---
# The pattern of epistaxis recurrence in patients taking prophylactic acetylsalicylic acid (ASA) from a 10 year cohort
## Abstract
### Objectives
Epistaxis is the most common otolaryngological emergency and one-third of epistaxis patients regularly take low-dose acetylsalicylic acid (ASA) for the prevention of cardiovascular disease (CVD). The shift in contemporary guidelines identifies little benefit of ASA intake in patients who have not previously had an infarction. Existing evidence confirms ASA intake as a factor for severe epistaxis, while the evidence concerning its impact on recurrence is ambiguous. There are no available studies which justify the administration of these drugs nor are there any studies correlating the effects of these drugs to the SCORE2 CVD risk stratifying scale.
### Study design
A retrospective analysis of all admitted epistaxis patients in a tertiary academic hospital for the 10 year period 2011 to 2021.
### Methods
Patient data were analysed using the hospital information software. A recurrence was defined as an epistaxis episode requiring hospital readmittance for at least one night. Patients taking anticoagulants were excluded ($$n = 421$$).
### Results
444 patients were included: 246 were taking ASA and 198 were not (NoASA). ASA patients had more frequent recurrence in general ($$p \leq 0.03$$), more recurrences per patient ($$p \leq 0.002$$), and more changes in bleeding localisation ($$p \leq 0.04$$). Recurrence in the ASA group was associated with lower haemoglobin values (HR 0.62, $p \leq 0.0001$), while surgery (HR 6.83, $p \leq 0.0001$) was associated with recurrence in the NoASA group. ASA patients had a statistically significant (r 0.33, $$p \leq 0.032$$) correlation between the total number of epistaxis recurrences and SCORE2. The indication for drug intake was highly questionable in as much as $40\%$ of ASA patients. Follow-up time was 5.27 years.
### Conclusions
Epistaxis patients taking prophylactic ASA are significantly more burdened by recurrence, because they have more frequent recurrences, a greater number of recurrences per patient, and more changes in bleeding localisations when compared to control patients. The drug indication is questionable in up to $40\%$ of ASA patients, exposing them unnecessarily to recurrence.
### Level of evidence
4.
## Introduction
Epistaxis is one of the most common emergencies in medicine. It mostly affects the elderly and accounts for 1 in 200 general emergency department (ED) visits as well as up to one-third of otolaryngological ED visits [1, 2]. One-third of epistaxis patients are on antiplatelet therapy, most frequently acetylsalicylic acid (ASA) [3–6]. Low-dose ASA (75–150 mg) is not only traditionally widely recommended for patients who have previously suffered vascular events (secondary prophylaxis), but also for patients who have a moderately raised risk of cardiovascular disease (CVD) without previous infarction (primary prophylaxis) [7, 8] as well as the general population above the age of 55 [9, 10].
However, the latest [2019] comprehensive American Heart Association guidelines issued on the prevention of CVD question whether the use of ASA as a primary prophylaxis in patients younger than 40 and older than 70 is beneficial [11] as in this particular population group, the risk of severe haemorrhage surpasses the benefit of prophylaxis. The Canadian and European (ESC) cardiology guidelines from 2020 and 2021, respectively, concur and conclude that only a few people benefit from primary prevention [12, 13]. According to the 2021 ESC guidelines, low dose aspirin is only recommended (class II b recommendation) as primary prevention for people with either Diabetes Mellitus or a very high CVD risk [13].
ASA has been found to be associated with severe epistaxis in previous studies [3, 5, 14]. It remains unclear, however, whether ASA leads to more recurrence in epistaxis patients as the results of current studies are ambigous [5, 15]. It is also unknown what fraction of epistaxis patients taking ASA has a reasonable indication for CVD prevention, particularly in the light of current guideline changes.
The aim of our study was to comprehensively analyse epistaxis patients taking ASA, including the indication for drug intake, and compare them to control patients not taking any anticoagulant or antiplatelet drugs as well as to investigate the factors leading to recurrence. We sought to thoroughly document each recurrent in-hospital stay to investigate location changes, the intervals between recurrences and the cumulative days spent in hospital.
## Materials and methods
We performed a retrospective study analysing patient records in our hospital management software for all admitted adult epistaxis patients in the 10 year period of January 2011 to September 2021. All patients were treated at the Department of Otolaryngology, Head and Neck Surgery, Sana Kliniken Leipziger Land in Borna, Germany, a tertiary academic hospital.
We collected data on each patient’s age, gender, date and season of admission, length of in-hospital stay, intake and dosage of anticoagulants and antiplatelet drugs, the indication for antiplatelet drug(s), systolic and diastolic blood pressure (BP) and whether there was a need to lower said BP, the localisation of epistaxis (left or right side as well as anterior, posterior of diffuse), various laboratory data (haemoglobin, PTT, platelet count, INR, creatinine, grade of chronic renal insufficiency, non-HDL cholesterol), therapy, smoking status, the need for transfusion, death, and sepsis. The seasons were defined as follows: winter (December–February), spring (March–May), summer (June–August) and autumn (September–November). For each readmission epistaxis episode, the same data were collected and the interval between hospital stays for epistaxis was calculated. We noted the total number of recurrences and the cumulative length of hospital stays due to epistaxis. A recurrence of epistaxis was defined as an epistaxis episode which required hospital admittance for at least one night, and where the patient had previously been admitted to hospital for treatment of epistaxis. The Systematic Coronary Risk Estimation 2 (SCORE2) was calculated according to European guidelines for each patient [13]. The indication was marked as justified according to the most recent guidelines [11–13], secondary prophylaxis or primary prophylaxis in patients aged between 40 and 69 with SCORE2 ≥ 10. Follow-up was done until March 2022.
The main inclusion criteria was the regular intake of prophylactic ASA (100 mg/day). The control group was made up of patients taking neither anticoagulant nor antiplatelet drugs (NoASA). Accordingly, all epistaxis patients taking Vitamin-K-Antagonists (VKA), direct oral anticoagulants (DOAC) or single antiplatelet drugs other than ASA were excluded from the study. Other exclusion criteria were: septal perforation, active malignant disease, known hematopoietic diseases, and acute liver and/or kidney failure.
The decision to admit an epistaxis patient for in-hospital treatment was made according to the standard operating procedure (SOP) of our clinic (Fig. 1): upon arrival in the emergency room, conservative measures were applied first, for example: local cooling of the neck, nasal compression, raising the upper body, and measuring blood pressure (BP). In patients with severely high BP, sublingual Nitrendipin was applied. Subsequently, an anterior rhinoscopy was performed to identify the bleeding localisation and in the case of an anterior epistaxis, cautery was performed. If these actions proved sufficient to stop the epistaxis and the patient was in a stable condition, he or she was discharged and the treatment was considered outpatient. If not, nasal packing was performed and the patients were admitted to a ward. This included anterior bleeders, for example, in cases where nasal packing in addition to cautery was needed for sufficient control; each patient with nasal packing must be treated in-hospital due to medico legal issues and danger of packing aspiration. In cases where these measures brought epistaxis under control, the packing was removed after 24 h and the patients were observed for a further 12–24 h before leaving the clinic. In the remaining cases re-packing, surgery or maxillary artery embolization were performed based on case-by-case decision. Fig. 1Flow-chart of the treatment of epistaxis patients (standard operating procedure—SOP) *Statistical analysis* was performed using MedCalc® Statistical Software version 20 (Ostend, Belgium). Fisher’s exact test, Pearson’s chi-squared test and Mann–Whitney U test was used as applicable. Univariate Cox regressions were used to perform crude and adjusted association between various predictors and recurrence. Bonferroni correction was made for multiple comparisons. The Kaplan–Meier curve was used to describe incidence of recurrence. Correlation was calculated according to Pearson’s r. The threshold for statistical significance was set at $p \leq 0.05.$
## Results
A total of 1039 epistaxis patients were screened for the study. After excluding 421 of those patients due to oral anticoagulant intake and another 174 due to other exclusion criteria, a total of 444 patients were included in the study, all of whom underwent stationary treatment. 246 Patients were taking prophylactic ASA (ASA-Group) and 198 patients were not taking any antiplatelet drugs (NoASA). The decision to admit the patient to the ward was always made according to the SOP of our institution. ASA patients were significantly older, with lower haemoglobin values, worse kidney function, and higher SCORE2, whereas NoASA patients displayed posterior epistaxis more often and needed BP reduction more frequently (Table 1).Table 1Comparison between epistaxis patients taking prophylactic acetylsalicylic acid (ASA) and patients not taking any antiplatelet drug (NoASA)N [444]ASANoASAp246198Female99 (40.2)90 (45.5)0.289Age74.74 ± 1.4864.14 ± 2.44< 0.0001Days in hospital2.55 ± 0.132.74 ± 0.210.344Recurrence42 (17.1)20 (10.1)0.034SCORE2 [13]23.36 ± 1.715.52 ± 2.3< 0.00001Treatment Packing28 (11.4)42 (21.2)0.005 Packing and cautery201 (81.7)134 (67.7)0.001 Surgery/embolization17 (6.9)22 (11.1)0.131Localisation Anterior213 (86.6)147 (74.2)0.001 Posterior26 (10.6)36 (18.2)0.027 Diffuse7 (2.8)15 (7.6)0.028 Systolic BP (mmHg)147.2 ± 2.97149.9 ± 3.770.217 Diastolic BP (mmHg)83.61 ± 1.5286.65 ± 20.020 BP lowering21 (8.5)36 (18.2)0.004 Haemoglobin (g/dl)12.83 ± 1.9613.22 ± 0.260.004 Creatinine (µmol/l)93.67 ± 4.7588.53 ± 8.720.0001 Platelets (× 109/l)237.84 ± 10.22242.25 ± 10.780.201 PTT (s)29.4 ± 0.9729.79 ± 1.140.529 INR1.02 ± 0.021 ± 0.010.187Chronic kidney disease Stage 138 (15.5)64 (32.3) Stage 2113 (45.9)72 (36.4) Stage 365 (26.4)26 (13.1) Stage 416 (6.5)8 (4.1) No data14 (5.7)28 (14.1)< 0.00001 Transfusion4 (1.6)2 [1]0.696 Death00 Sepsis00Significant p values are in boldSCORE2 Systematic Coronary Risk Estimation 2, BP blood pressure At least one recurrence was noted in 42 patients ($17.1\%$) in the ASA group, significantly more ($$p \leq 0.034$$) than in the NoASA group, where recurrence occurred in only 20 patients ($10.1\%$) (Table 1, Fig. 2). According to univariate analysis, lower haemoglobin values were associated with recurrence in the ASA group. Surgery during the first admission was associated with recurrence in the NoASA group (Table 2).Fig. 2Kaplan–Meier curve of epistaxis recurrence ($$p \leq 0.03$$), ASA acetylsalicylic acid. Within 30 days, $59.6\%$ of all recurrences in the ASA and $44.6\%$ in the NoASA-Group occurred; within 180 days, $78.4\%$ in the ASA and $80.2\%$ in the NoASA-GroupTable 2Univariate analysis for recurrence in epistaxis patients taking prophylactic acetylsalicylic acid (ASA) and patients not taking any antiplatelet drug (NoASA), according to Bonferroni correction, significance was set at $p \leq 0.0025$ASA $\frac{42}{246}$NoASA $\frac{20}{198}$HR$95\%$ CIpHR$95\%$ CIpAge0.9930.969–1.0180.5971.0090.983–1.0350.509Female1.1220.609–2.0670.7130.9630.399–2.3240.933BP lowering1.5280.601–3.890.3730.490.114–2.1120.339Haemoglobin0.6180.491–0.777< 0.00010.7120.497–1.020.660Creatinine1.0050.998–1.0120.1720.9860.963–1.0090.216Platelets10.996–1.0040.8331.0040.998–1.0090.188PTT0.9440.769–1.160.5841.0750.913–1.2670.386INR0.9750.786–1.210.8190.8830.651–1.2110.394Localisation Anterior0.9310.392–2.210.8710.790.304–2.0560.629 Posterior1.4590.615–3.4630.3922.6011.038–6.520.042 Surgery/Emb1.5820.565–4.4330.3836.8292.786–16.74< 0.0001 Transfusion7.3762.267–24.0040.0100.966 Smoking0.3480.084–1.4380.1450.6090.141–2.6260.506 SCORE2 [13]0.9840.96–1.0090.2051.0080.973–1.0440.659Season Winter1.2460.648–2.3970.5100.6190.207–1.8500.390 Spring1.2450.662–2.3410.4963.6591.516–8.8320.004 Summer0.7820.347–1.760.5520.1980.027–1.4770.114 Fall0.6960.309–1.5670.3820.7520.251–2.250.610 DAPT1.020.472–2.2040.971Significant p values are in boldHR hazard ratio, CI confidence interval, SCORE2 Systematic Coronary Risk Estimation 2, BP blood pressure, DAPT dual antiplatelet therapy There were a total of 77 recurrences in the 42 patients in the ASA group and 24 recurrences in the 20 patients in the NoASA group, resulting in 119 and 44 epistaxis in-hospital stays, respectively. The results showed that the ASA group displayed significantly more recurrence episodes per patient ($$p \leq 0.002$$), more patients exceeding 2 recurrences ($$p \leq 0.010$$), more changes of localisation on recurrence ($$p \leq 0.038$$) and more recurrences within a month after initial discharge ($$p \leq 0.008$$) (Table 3). Surgery or embolization was performed more frequently to treat recurrences in the NoASA group ($$p \leq 0.0009$$). The mean follow-up time was 5.27 ± 0.24 years (Table 3).Table 3Comparison between epistaxis patients with recurrence taking prophylactic acetylsalicylic acid (ASA) and patients not taking any antiplatelet drug (NoASA)RecurrencesASANoASAN (Patients)4220N (*Total epistaxis* episodes)11944pNo. recurrence per patient1.83 ± 0.471.2 ± 0.30.002Cumulative days in hospital7.83 ± 1.557.2 ± 1.590.599Time to next recurrence (days)155.13 ± 66.73191.79 ± 128.790.952Next recurrence < 30 days44 (57.1)10 (41.7)0.008 > 2 Recurrences18 (42.9)2 [10]0.010Localisation change15 (35.7)2 [10]0.038SCORE2 [13]21.14 ± 2.9116.65 ± 5.680.044Treatment Packing14 (11.8)7 (15.9)0.599 Winter82 Spring42 Summer11 Fall12 Packing and cautery100 [84]28 (63.6)0.003 Winter333 Spring2712 Summer193 Fall2110 Surgery/embolization5 (4.2)9 (20.5)0.0009 Winter25 Spring12 Summer10 Fall12 Follow-up (years)5.08 ± 0.315.49 ± 0.380.168Significant p values are in boldSCORE2 Systematic Coronary Risk Estimation 2 A statistically significant correlation (r 0.332, $95\%$ CI [0.031–0.578], $$p \leq 0.032$$) between the total number of epistaxis recurrences and SCORE2 was found in patients with recurrences in the ASA group (Fig. 3).Fig. 3Epistaxis patients with recurrence taking prophylactic acetylsalicylic acid (ASA) ($$n = 42$$), showing a statistically significant ($$p \leq 0.03$$) correlation between the total number of epistaxis recurrences and SCORE2 (Systematic Coronary Risk Estimation 2), $95\%$ confidence interval shaded blue. Spots represent individual patients with overlapping according to the heat map—bottom row from left: 4th spot 2 patients, 6th spot 5 patients, 8th spot 2 patients, 13th spot 4 patients, second row from bottom: last spot on the right 2 patients
There was no appropriate indication for drug intake according to the newest guidelines in 99 out of 246 ($40.2\%$) ASA patients. $18.2\%$ of these patients had 1.89 ± 0.41 episodes of recurrent epistaxis, not significantly different to patients with appropriate indication, where $16.3\%$ patients had 1.79 ± 0.31 episodes of recurrence. The patients without justifiable indication were significantly older, had a higher SCORE2, and needed surgery more often (Table 4).Table 4Comparison of epistaxis patients taking acetylsalicylic acid (ASA) according to the indication justificationAppropriateInappropriatepN [246]147 (59.8)99 (40.2)Primary prophylaxis5899Secondary prophylaxis89/Female53 (36.1)46 (46.5)0.113Age72.92 ± 0.977.43 ± 1.290.0002 < 501 (0.7)4 [4]0.161 50–6950 [34]16 (16.2)0.002 ≥ 7096 (65.3)79 (79.8)0.015Days in hospital2.41 ± 0.072.75 ± 0.130.085Posterior12 (8.2)14 (14.1)0.136Recurrence24 (16.3)18 (18.2)0.732No. of recurrence per patient1.79 ± 0.311.89 ± 0.410.943Surgery / Embolization6 (4.1)11 (11.1)0.041Transfusion1 (0.7)3 [3]0.306SCORE2 [13]20.96 ± 1.0527.1 ± 1.430.0003Significant p values are in boldSCORE2 Systematic Coronary Risk Estimation 2
## Discussion
The findings of our study are that epistaxis patients taking low-dose ASA for CVD prevention are extremely burdened by recurrence. This is supported by evidence of a higher recurrence rate in the ASA group when compared to that of the control group ($17.1\%$ vs $10.1\%$, $$p \leq 0.034$$), the fact that more recurrences per patient occurred in the ASA group (1.83 ± 0.47 vs 1.2 ± 0.3, $$p \leq 0.002$$), and more frequent changes in bleeding localisation in the ASA group ($35.7\%$ vs $10\%$, $$p \leq 0.038$$). There is discrepancy in previous evidence of recurrence rates in large cohorts, some studies concurring with our findings [5, 15] and others differing [15–17]. The comparability of the groups may seem to be diminished by the older age and decreased kidney function of the ASA patients and the raised diastolic BP and increased lowering of BP in the NoASA patients. However, literature reports ambiguous results concerning the impact of old age [16–18] and BP [15, 17] on epistaxis recurrence and as a result these factors should be considered with reservation.
The recurrences were more likely to happen within 1 month of discharge in the ASA group compared to the control group ($$p \leq 0.008$$). These facts highlight pressure on the health system as each stationary treatment incurs high costs: 7000–22,000 USD in the USA [19] and 11,000 USD on average in Switzerland (with currency conversion) [20].
Regarding the dynamics of the initial recurrence episode, we found that in a follow-up period of over 5 years, $80\%$ of first recurrence episodes in both groups occurred within 6 months of discharge. Our results are in concordance with current literature [15]. However, we are the first to report the dynamics of ASA and NoASA patients separately. These results show a steeper trend in the ASA group with $60\%$ of all first recurrences occurring within the first month compared to $44.6\%$ in the control group ($$p \leq 0.033$$). In addition, univariate analysis revealed low haemoglobin values as a significant factor leading to recurrence in the ASA group.
This indicates the need for intensive nasal mucosa care with ointment as well as rigorous anaemia control and post discharge treatment of these patients. Anaemia was also found to be a risk factor for recurrence, independent of ASA intake, in the study of Cohen et al. [ 15]. Existing evidence points out a strong relationship between lower baseline haemoglobin values and major bleeding in CVD patients [21]. Furthermore, it has been shown that low-dose ASA for primary prevention has a negative impact on the haemoglobin values of the elderly [22]. It may, therefore, be assumed that a vicious circle exists in which ASA intake in the elderly leads to lower haemoglobin values, which in turn is compounded by more frequent epistaxis recurrences.
In the light of these facts, particularly concerning the elderly with ASA intake for primary prevention, the importance of the correct indication cannot be emphasized strongly enough. We identified 157 patients ($63.8\%$) with primary prevention as indication and 89 patients ($36.2\%$) with secondary prevention in our cohort of 246 epistaxis patients in the ASA group. After thoroughly reviewing the patient documentation to justify primary prevention in light of the recent guideline changes, we found that 99 of the 157 patients taking ASA for primary prevention did not have a valid indication. As previously mentioned, the newest guidelines question primary prevention in patients with a low CVD risk profile who are younger than 40 and older than 70 years [11–13]. Accordingly, $40\%$ ($\frac{99}{246}$) of patients in the ASA group in our cohort that had no justification for low dose ASA intake. This thorough analysis was unfortunately performed post-hoc and not in “real-time”, because we initially relied on the indication identified by the family doctor. Therefore, ASA was not stopped at the time of the first bleeding episode despite the fact that it was not indicated according to the most recent guidelines.
In surveys of general population, low-dose ASA is predominantly taken for primary prevention and accounts for just over $50\%$ of all ASA intake patients, whereby $20\%$ do not have a legitimate indication [23]. Only ¾ of primary prevention patients are in the appropriate age group of 40 to 70. This means that the potential fraction of ASA patients in the general population without a significant indication exceeds $50\%$.
Bearing the aforementioned treatment costs in mind, it would be interesting to calculate hospital/health system expenses for treating epistaxis in patients without a justified indication for ASA intake. These costs should be weighed against actual cardiovascular events in patients taking aspirin for primary prevention.
The patients with inappropriate indication in our cohort were significantly older, highlighting widespread ASA therapy of the elderly for inappropriate primary prevention and mirroring the recommendations of studies from the 2000s [9, 10]. The need for surgical intervention in the case of a recurrent epistaxis episode was significantly higher in this group of patients. It seems, according to our results, that a plethora of patients are exposed to unnecessary ASA intake which in turn leads to more epistaxis recurrence and more surgery for recurrence. Consequently, the indication for low-dose ASA intake needs to be verified immediately during the first inpatient treatment for epistaxis, because, as previously mentioned, up to $40\%$ of patients do not have a reasonable indication for drug intake and most recurrences occur within 30 days.
To determine the CVD risk of patients, we estimated the systematic coronary risk using the SCORE2 scale, according to the European guidelines [13] for CVD prevention. This scale was chosen, because it is the newest, most relevant (class I recommendation [13]) and validated prediction score available. The scale range of 1–49 is determined by age, sex, smoking status, systolic BP, and non-HDL cholesterol and it stratifies patients into low to moderate, high, and very high CVD risk. This allowed the assessment of drug indication for primary prevention. On univariate analysis, SCORE2 was not a significant factor leading to recurrence in general. However, a significant relationship between the number of recurrence episodes of individual patients and SCORE2 in the ASA group was observed. Potential patients with ASA intake and SCORE2 > 20 need to be screened and drug indication thoroughly checked as this group has a high probability of 2 + epistaxis recurrences, according to our findings.
Concerning the NoASA group, we found that previous surgery was a significant risk factor for recurrent epistaxis. Bleeding points that are not easily accessible, as found in posterior epistaxis, have previously been identified as a risk factor for recurrence [16]. It can be postulated that in this group of patients, early recurrent bleeding episodes were likely the result of insufficient bleeding control in the initial treatment. This cannot be applied to the ASA group as a significant change of localisation was found in these patients. Therefore, global factors due to ASA intake must be heldaccountable.
Existing evidence states that epistaxis occurs more often in winter [18]. Our results showed that first episodes in spring led to more recurrence in the NoASA group. We defined winter according to the meteorological definition as December to February, whereas the aforementioned study used the approximate astronomical definition of January to March and registered most cases in January and March. This could explain the discrepancy. Nevertheless, it seems that seasonal variation, predominantly the cold months, plays a significant role in NoASA patients while having less impact on ASA patients.
## Limitations
Certain limitations of our study need to be disclosed. The retrospective methodology of the study is definitely a drawback as unintentional errors in data acquisition cannot be ruled out. We were able to obtain very few cholesterol values as this is not routinely done in our institution. We, therefore, approximated the SCORE2 in patients with missing values in a uniform way, using the median value. It is possible that this led to data distortion concerning this scale. We only analysed recurrence episodes requiring in-hospital treatment to determine the burden on the tertiary health system thereby excluding outpatient episodes that are very relevant to primary care.
Further noteworthy limitations are: no control group of patients taking ASA without epistaxis could be put together, data concerning bleeding in other locations as well as data about drugs interacting with ASA was not provided, and the data from excluded patients taking anticoagulants could have been used as a comparative group.
Finally, adherence to ASA therapy was not analysed in our study. It is possible that some of the ASA patients did not take ASA regularly in the days prior to admission, which may in turn have interfered with our results.
## Conclusions
The findings of our study show that epistaxis patients taking prophylactic ASA are significantly more burdened by recurrence, because they have recurrence more often, in a greater number per patient, and with more bleeding location changes compared to controls. Recurrence in ASA patients is more likely to occur in patients with lower haemoglobin values. The number of epistaxis recurrence episodes rises with a rising SCORE2. In almost half of these patients the drug indication is questionable, exposing them unnecessarily to recurrence.
## References
1. Pallin DJ, Chng YM, McKay MP, Emond JA, Pelletier AJ, Camargo CA. **Epidemiology of epistaxis in US emergency departments, 1992 to 2001**. *Ann Emerg Med* (2005.0) **46** 77-81. DOI: 10.1016/j.annemergmed.2004.12.014
2. Walker TWM, Macfarlane TV, McGarry GW. **The epidemiology and chronobiology of epistaxis an investigation of Scottish hospital admissions 1995–2004**. *Clin Otolaryngol* (2007.0) **32** 361-365. DOI: 10.1111/j.1749-4486.2007.01530.x
3. Yaniv D, Zavdy O, Sapir E, Levi L, Soudry E. **The impact of traditional anticoagulants, novel anticoagulants, and antiplatelets on epistaxis**. *Laryngoscope* (2021.0) **131** 1946-1951. DOI: 10.1002/lary.29417
4. Stankovic P, Georgiew R, Frommelt C. **Shorter hospital stays in epistaxis patients with atrial fibrillation when taking rivaroxaban or apixaban versus phenprocoumon**. *J Thromb Thrombolysis* (2019.0) **47** 384-391. DOI: 10.1007/s11239-019-01824-x
5. Soyka MB, Rufibach K, Huber A, Holzmann D. **Is severe epistaxis associated with acetylsalicylic acid intake?**. *Laryngoscope* (2010.0) **120** 200-207. DOI: 10.1002/lary.20695
6. Kallenbach M, Dittberner A, Boeger D. **Hospitalization for epistaxis: a population-based healthcare research study in Thuringia, Germany**. *Eur Arch Otorhinolaryngol* (2020.0) **277** 1659-1666. DOI: 10.1007/s00405-020-05875-2
7. Pearson TA, Blair SN, Daniels SR. **AHA guidelines for primary prevention of cardiovascular disease and stroke: 2002 update: consensus panel guide to comprehensive risk reduction for adult patients without coronary or other atherosclerotic vascular diseases. American Heart Association Science Advisory and Coordinating Committee**. *Circulation* (2002.0) **106** 388-391. DOI: 10.1161/01.CIR.0000020190.45892.75
8. 8.JBS 2. Joint British Societies’ guidelines on prevention of cardiovascular disease in clinical practice. Heart; 91:v1–52
9. Bulugahapitiya U, Siyambalapitiya S, Sithole J, Fernando DJ, Idris I. **Age threshold for vascular prophylaxis by aspirin in patients without diabetes**. *Heart* (2008.0) **94** 1429-1432. DOI: 10.1136/hrt.2008.150698
10. Wald NJ, Law MR. **A strategy to reduce cardiovascular disease by more than 80%**. *BMJ* (2003.0) **326** 1419-1425. DOI: 10.1136/bmj.326.7404.1419
11. Arnett DK, Blumenthal RS, Albert MA. **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* (2019.0) **140** e596-e646. PMID: 30879355
12. Wein T, Lindsay MP, Gladstone DJ. **Canadian Stroke Best Practice Recommendations, seventh edition: acetylsalicylic acid for prevention of vascular events**. *CMAJ* (2020.0) **192** E302-E311. DOI: 10.1503/cmaj.191599
13. Visseren FLJ, Mach F, Smulders YM. **2021 ESC Guidelines on cardiovascular disease prevention in clinical practice**. *Eur Heart J* (2021.0) **42** 3227-3337. DOI: 10.1093/eurheartj/ehab484
14. Glikson E, Chavkin U, Madgar O. **Epistaxis in the setting of antithrombotic therapy: a comparison between factor Xa inhibitors, warfarin, and antiplatelet agents**. *Laryngoscope* (2019.0) **129** 119-123. DOI: 10.1002/lary.27400
15. Cohen O, Shoffel-Havakuk H, Warman M. **Early and late recurrent epistaxis admissions: patterns of incidence and risk factors**. *Otolaryngol Head Neck Surg* (2017.0) **157** 424-431. DOI: 10.1177/0194599817705619
16. Ando Y, Iimura J, Arai S. **Risk factors for recurrent epistaxis: importance of initial treatment**. *Auris Nasus Larynx* (2014.0) **41** 41-45. DOI: 10.1016/j.anl.2013.05.004
17. Bui R, Doan N, Chaaban MR. **Epidemiologic and outcome analysis of epistaxis in a tertiary care center emergency department**. *Am J Rhinol Allergy* (2020.0) **34** 100-107. DOI: 10.1177/1945892419876740
18. Purkey MR, Seeskin Z, Chandra R. **Seasonal variation and predictors of epistaxis**. *Laryngoscope* (2014.0) **124** 2028-2033. DOI: 10.1002/lary.24679
19. Villwock JA, Jones K. **Recent trends in epistaxis management in the United States: 2008–2010**. *JAMA Otolaryngol Head Neck Surg* (2013.0) **139** 1279-1284. DOI: 10.1001/jamaoto.2013.5220
20. Nikolaou G, Holzmann D, Soyka MB. **Discomfort and costs in epistaxis treatment**. *Eur Arch Otorhinolaryngol* (2013.0) **270** 2239-2244. DOI: 10.1007/s00405-012-2339-2
21. Bassand JP, Afzal R, Eikelboom J. **Relationship between baseline haemoglobin and major bleeding complications in acute coronary syndromes**. *Eur Heart J* (2010.0) **31** 50-58. DOI: 10.1093/eurheartj/ehp401
22. Gaskell H, Derry S, Moore RA. **Is there an association between low dose aspirin and anemia (without overt bleeding)?: narrative review**. *BMC Geriatr* (2010.0). DOI: 10.1186/1471-2318-10-71
23. Owen JL, Rush JL, Armbrecht ES, Javaheri A, Behshad R. **A survey of aspirin knowledge among the general public**. *J Gen Intern Med* (2021.0) **37** 1799-1801. DOI: 10.1007/s11606-021-06983-3
|
---
title: 'Association between the systemic immune-inflammation index and kidney stone:
A cross-sectional study of NHANES 2007-2018'
authors:
- Xingpeng Di
- Shaozhuang Liu
- Liyuan Xiang
- Xi Jin
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9989007
doi: 10.3389/fimmu.2023.1116224
license: CC BY 4.0
---
# Association between the systemic immune-inflammation index and kidney stone: A cross-sectional study of NHANES 2007-2018
## Abstract
### Background
The incidence rate of kidney stones increased over the past decades globally, which brought medical expenditure and social burden. The systemic immune-inflammatory index (SII) was initially identified as a prognosis of multiple diseases. We performed an updated analysis on the impact of SII on kidney stones.
### Methods
This compensatory cross-sectional study enrolled participants from the National Health and Nutrition Examination Survey 2007-2018. Univariate and multivariate logistic regression analyses were performed to investigate the association between SII and kidney stones.
### Results
Of the 22220 participants, the mean (SD) age was 49.45 ± 17.36 years old, with a $9.87\%$ incidence rate of kidney stones. A fully adjusted model showed that SII higher than 330 x 109/L was parallel associated with kidney stones (Odds ratio [OR] = 1.282, $95\%$ Confidence interval [CI] = 1.023 to 1.608, $$P \leq 0.034$$) in adults aged 20-50. However, no difference was found in the elderly subgroup. Multiple imputation analyses confirmed the robustness of our results.
### Conclusions
Our findings suggested SII was positively associated with a high risk of kidney stones in US adults aged less than 50. The outcome compensated for previous studies that still needed more large-scale prospective cohorts for validation.
## Introduction
In the past few decades, the prevalence of kidney stones increased worldwide. For instance, the incidence rate of kidney stones increased from $3.2\%$ in the 1980s to $9.6\%$ currently in the United States (US) [1, 2]. Specifically, the incidence of symptomatic kidney stones is higher in the male population [3]. For age stratification, the peak incidence rate in males is 40 to 60 years old in males, and 50 years old in females [4, 5]. For the increasing trend of kidney stones, the risk factors include obesity, diabetes mellitus (DM), high intake of salt, animal protein, and added sugar (6–10). The health care of kidney stones has aroused great attention for the high medical expenditure and social burden [11].
Recently, kidney stone was reported to be associated with various inflammatory responses. For instance, C-reactive protein (CRP) concentration and the erythrocyte sedimentation rate (ESR) were identified as biomarkers for inflammatory diseases [12]. In the systematic inflammatory system of the human body, immune cells are critical in multiple diseases. Several studies suggested that the neutrophil-to-lymphocyte ratio (NLR) was a crucial marker for kidney stone [13, 14]. Moreover, studies demonstrated that platelet also engages in inflammatory response [15]. The platelet-to-lymphocyte ratio was identified as an ideal predictor for systematic inflammatory response syndrome in patients undergoing percutaneous nephrolithotomy (PNL). Furthermore, researchers found that integrated peripheral lymphocyte, neutrophil, and platelet counts might be better to predict the inflammatory state, which also served as indicators for many diseases. The systemic immune-inflammatory index (SII) was initially identified as a prognosis of cancer, intracerebral hemorrhage, coronary stenosis, and others (16–18). However, the impact of SII on kidney stones is not fully elucidated, and little is known about its prognostic ability for kidney stones. For the relatively clarified association between NLR and PLR with kidney stone, we solely focused on SII in the current study. We hypothesized SII was a predictor for the risk of kidney stones. We performed the current study to investigate the association between SII and kidney stones.
## Study population
The current study was performed using a cross-sectional design of the NHANES dataset [19]. Data from the NHANES database is collected every two years. All the protocols were approved by the National Center for Health Statistics institutional review board, and informed consent was required from participants. We enrolled 59842 participants from the year 2007 to 2018. The exclusion criteria included: (a) missing kidney stone data; (b) incomplete and extreme SII data (neutrophil, lymphocyte, and platelet count); (c) missing data of covariates. Finally, 2220 participants were included for complete case analysis (Figure 1).
**Figure 1:** *Overview of participants screening. NHANES, National Health and Nutrition Examination Survey; SII, systemic immune-inflammation index.*
## Assessment of kidney stone
Kidney stone history was defined by “Have you ever had kidney stones?” [ 2]. The participants who reported an answer to the question indicated a history of diagnosed kidney stones.
## Assessment of systemic immune-inflammation index
The hypothesis of SII was first used by Hu et al. [ 20], to evaluate the prognostic value of multiple diseases. The SII was composed of peripheral neutrophil (N), lymphocyte (L), and platelet (P), which was defined as P x N/L (109/L). We set the cutoff value at 330 x 109/L for all subsequent analyses based on previous studies of the NHANES [21].
## Covariates definition
Participants self-reported age, gender, race/ethnicity, education level, family income-to-poverty ratio, smoking history, and alcohol drinking history [1, 22, 23]. Race/ethnicity was divided into non-Hispanic Black, non-Hispanic White, Hispanic/Mexican, and other races. The family income-to-poverty ratio was classified into low (<1.3), median (1.3-3.5), and high (>3.5). Alcohol drinking history was defined as None (< 1 time per week), slight (1-3 times per week), and severe (to 4 times per week). The body mass index (BMI) was classified as lower than 20 kg/m2, 20-25 kg/m2, 25-30 kg/m2, and over 30 kg/m2. Smoking history, DM, and coronary heart disease were recorded by “Yes/No”. DM was diagnosed based on previous studies [24].
## Statistical analyses
The comparison in the SII subgroups was performed using survey-weighted logistic regression for continuous variables (mean ± standard deviation [SD]) and survey-weighted Chi-square test for categorical variables (counting number, [n]). Multivariate logistic regression analysis was conducted to investigate the relationship between SII and kidney stones. The crude model was adjusted for no covariates. Model 1 was a minimally-adjusted model adjusted for age, gender, race/ethnicity, family income-to-poverty ratio, and education level. Model 3 was adjusted for BMI, smoking history, alcohol drinking history, DM, and coronary heart disease. Age-stratified subgroup analysis was performed to clarify the impact of SII. Further stratified logistic regression analysis was conducted to identify variables that modify the association in participants aged 20-50 [25]. Sensitivity analyses using the multiple imputation (MI) methods were performed. MI was an approach to compensate for the missing data based on five replications and a chained equation method in the R MI procedure to account for missing data on education level, family income-to-poverty ratio, BMI, smoking history, alcohol drinking history, DM, and coronary heart disease [26].
All analyses were performed utilizing R software version 4.1 (http://www.R-project.org; The R Foundation) and EmpowerStats (http://www.empowerstats.com, X&Y Solutions, Inc.). $P \leq 0.05$ (two-sided) was set for a significant difference.
## Results
There were 59842 participants from 2007-2018 included in this study. 25163 participants were excluded for missing kidney stone data, and 3054 participants were excluded for missing SII data. After removing 9405 participants missing covariates and extreme data for SII, 2220 participants were finally enrolled. Of the 22220 participants, there were 11755 males and 10465 females (Table 1). The prevalence rate of kidney stones was $9.87\%$. Compare with SII lower than 330 x 109/L, participants with SII higher than 330 x 109/L were more non-Hispanic White, smokers, DM, and kidney stones ($P \leq 0.05$).
**Table 1**
| Characteristics | Overall | SII (109/L) | SII (109/L).1 | P value |
| --- | --- | --- | --- | --- |
| Characteristics | Overall | < 330 | ≥ 330 | P value |
| Number | 22220 | 5722 | 16498 | |
| Age | 49.45 ± 17.36 | 49.25 ± 17.24 | 49.52 ± 17.40 | <0.001 |
| Family income-to-poverty ratio | 2.59 ± 1.64 | 2.58 ± 1.63 | 2.59 ± 1.64 | 0.363 |
| BMI (kg/m2) | 29.34 ± 7.04 | 28.47 ± 6.35 | 29.64 ± 7.25 | <0.001 |
| Gender | | | | <0.001 |
| Male | 11755 (52.90%) | 2414 (42.19%) | 8051 (48.80%) | |
| Female | 10465 (47.10%) | 3308 (57.81%) | 8447 (51.20%) | |
| Race/Ethnicity | | | | <0.001 |
| Non-Hispanic Black | 4510 (20.30%) | 1833 (32.03%) | 2677 (16.23%) | |
| Non-Hispanic White | 10171 (45.77%) | 2000 (34.95%) | 8171 (49.53%) | |
| Hispanic/Mexican | 5276 (23.74%) | 1234 (21.57%) | 4042 (24.50%) | |
| Other Races | 2263 (10.18%) | 655 (11.45%) | 1608 (9.75%) | |
| Education level | | | | 0.408 |
| ≤ High school | 4752 (21.39%) | 1239 (21.65%) | 3513 (21.29%) | |
| > High school | 17468 (78.61%) | 4483 (78.35%) | 12985 (78.71%) | |
| Smoking history | | | | 0.010 |
| Non-smoker | 11093 (49.92%) | 3004 (52.50%) | 8089 (49.03%) | |
| Smoker | 11127 (50.08%) | 2718 (47.50%) | 8409 (50.97%) | |
| Alcohol drinking history (drinks/week) | | | | 0.156 |
| < 1 | 14262 (64.19%) | 3633 (63.49%) | 10629 (64.43%) | |
| 1-3 | 5645 (25.41%) | 1500 (26.21%) | 4145 (25.12%) | |
| ≥ 4 | 2313 (10.41%) | 589 (10.29%) | 1724 (10.45%) | |
| Diabetes mellitus | | | | <0.001 |
| No | 18067 (81.31%) | 4709 (82.30%) | 13358 (80.97%) | |
| Yes | 4153 (18.69%) | 1013 (17.70%) | 3140 (19.03%) | |
| Coronary heart disease | | | | 0.090 |
| No | 21278 (95.76%) | 5472 (95.63%) | 15806 (95.81%) | |
| Yes | 942 (4.24%) | 250 (4.37%) | 692 (4.19%) | |
| Kidney stone | | | | 0.010 |
| No | 20026 (90.13%) | 5245 (91.66%) | 14781 (89.59%) | |
| Yes | 2194 (9.87%) | 477 (8.34%) | 1717 (10.41%) | |
Subsequently, logistic regression analysis found no significant difference between SII subgroups and kidney stones after adjusting for covariates in the whole population (Table 2). After the age was stratified by 50 years old, the univariate and multivariate analyses demonstrated high SII over 330 x 109/L was associated with a higher risk of kidney stones in the crude model (Odds ratio [OR] = 1.485, $95\%$ Confidence interval [CI] = 1.195 to 1.845, $P \leq 0.001$) and model 1 (OR = 1.387, $95\%$ CI = 1.108 to 1.736, $$P \leq 0.005$$). After adjusting for all confounding factors, high SII was still positively associated with a kidney stone in the population aged 20-50 years old. ( OR = 1.282, $95\%$ CI = 1.023 to 1.608, $$P \leq 0.034$$) The difference was not found in participants aged 50 years old and above. Stratified logistic regression analysis in the 20-50 years old group suggested no potential modifiers in the relationship between SII and kidney stone in 20-50 population (Supplementary Table S1).
**Table 2**
| Age stratification | SII (109/L) | SII (109/L).1 | P value |
| --- | --- | --- | --- |
| Age stratification | < 330 (OR, 95% CI) | ≥ 330 (OR, 95% CI) | P value |
| Overall | Overall | Overall | Overall |
| Crude model a | 1.0 (Reference) | 1.201 (1.043,1.383) | 0.013 |
| Model 1 b | 1.0 (Reference) | 1.150 (0.995,1.328) | 0.061 |
| Model 2 c | 1.0 (Reference) | 1.075 (0.928,1.246) | 0.337 |
| 20-50 | 20-50 | 20-50 | 20-50 |
| Crude model | 1.0 (Reference) | 1.485 (1.195,1.845) | <0.001 |
| Model 1 | 1.0 (Reference) | 1.387 (1.108,1.736) | 0.005 |
| Model 2 | 1.0 (Reference) | 1.282 (1.023,1.608) | 0.034 |
| ≥ 50 | ≥ 50 | ≥ 50 | ≥ 50 |
| Crude model | 1.0 (Reference) | 1.036 (0.866,1.240) | 0.701 |
| Model 1 | 1.0 (Reference) | 0.987 (0.820,1.188) | 0.889 |
| Model 2 | 1.0 (Reference) | 0.938 (0.776,1.133) | 0.506 |
For further validation of the outcomes, multiple imputations were performed. The baseline characters distribution was depicted in Supplementary Table S2. Sensitivity analysis demonstrated a similar result in the fully-adjusted model (OR = 1.251, $95\%$ CI = 1.015 to 1.541, $$P \leq 0.036$$) (Table 3).
**Table 3**
| Age | SII < 330 x 109/L (OR, 95% CI) | SII ≥ 330 x 109/L (OR, 95% CI), P | SII ≥ 330 x 109/L (OR, 95% CI), P.1 |
| --- | --- | --- | --- |
| Age | SII < 330 x 109/L (OR, 95% CI) | Complete case | Multiple imputation |
| 20-50 | 20-50 | 20-50 | 20-50 |
| Model 2 a | 1.0 (Reference) | 1.282 (1.023,1.608), 0.034 | 1.251 (1.015,1.541),0.036 |
| ≥ 50 | ≥ 50 | ≥ 50 | ≥ 50 |
| Model 2 a | 1.0 (Reference) | 0.938 (0.776,1.133), 0.506 | 0.924 (0.785,1.088), 0.345 |
## Discussion
The immune system has long been identified as a prognostic factor of multiple diseases. In this cross-sectional study, we identified higher SII level was independently associated with kidney stones in people lower than aged 50 years old. However, we did not find such an association in the elderly population. The relationship between SII and kidney stones was confirmed by multiple imputation sensitivity analyses. Importantly, the measurement of SII was based on a standard laboratory approach of peripheral neutrophils, lymphocytes, and platelet in clinical practice. Herein, SII can be identified as a biomarker for kidney stones in younger adults.
Several studies suggested the relationship between inflammatory biomarkers and diseases consistent with previous studies. A study investigated the prognostic value of SII on kidney stone former and passage from NHANES 2007-2014. The results demonstrated SII>444.37 indicated a higher incidence of kidney stones [27]. However, our findings from NHANES 2007-2018 with updated data demonstrated that SII was the only indicator for kidney stones in the 20-50 years old group. Although there was a significant difference in all age groups, the difference disappeared after being adjusted for confounding factors.
Immune inflammation response was identified as engaging in multiple disease processes. A retrospective study compared the diagnostic values of leukocytes, neutrophils, neutrophil-to-lymphocyte (NLR), and platelet-to-lymphocyte (PLR) in distinguishing appendicitis and ureteral stones [28]. The study identified NLR as a predictor in distinguishing appendicitis and ureteral stones. In addition, NLR and PLR were inversely associated with the spontaneous ureteral stone passage. However, a recent study demonstrated that inflammatory biomarkers such as NLR and PLR were not related to spontaneous ureteral stone passage [29]. SII was recognized as a valuable and convenient inflammatory biomarker to predict the risk of low bone mineral density or osteoporosis among postmenopausal women aged ≥ 50 years old [25]. Moreover, SII has also been reported concerning all-cause mortality in arteriosclerotic cardiovascular disease [30].
Recently, researchers suggested that multiple inflammatory processes engaged in kidney stone formation. Idiopathic calcium oxalate stones often attach to Randall’s plaque that was associated with the activation of M1 macrophages [31]. While M2 macrophage-related genes are associated with the inhibition of stone formation [32]. The renal crystal deposition is also related to reactive oxygen species (ROS) production, and inflammasome activation [33]. The exosomes released by macrophage promoted the IL-8 production, facilitated neutrophil migration, and enhanced the crystal inflammatory response [34]. Meanwhile, the exosomes reduced T-cell activation as well. For lymphocytes, a lower peripheral lymphocyte count indicated a higher SII level. Lymphocytes are critical components of leukocytes that mediate both innate and adaptive immune responses. A blood lymphocyte analysis in 36 kidney stones patients found three patients had lymphocyte depletion [35]. In addition, platelets are increasingly identified as crucial modulators of inflammation response. Activated platelets trigger an intrinsic coagulation cascade that contributes to multiple diseases. Platelets can also accelerate inflammatory state [36]. The platelet interacts with monocytes, neutrophils, and lymphocytes, and regulates innate and adaptive responses.
Interestingly, our findings only demonstrated the association between SII and kidney stones in people aged lower than 50. Indeed, obesity, older age, metabolic syndrome, and DM are all risk factors for diseases in the older population [37]. Since there are more risk factors in older people than in younger ones, the impact of SII on kidney stones might be compensated by other confounding factors. In addition, for men aged lower than 60, a higher intake of calcium had a multivariate relative risk of 0.69 [38]. However, no such trend was found in older people. Unfortunately, the relationship between the inflammatory markers and kidney stones was not able to be identified in the younger age group since the data for participants under 20 years old was unincluded. In addition, SII was reported to be an indicator of cardiovascular disease [39]. Therefore, after the model was adjusted for DM and cardiovascular diseases in elderly people, the effect of SII was eliminated.
In the current study, we performed analyses to explore the association between SII and kidney stones based on the cross-sectional prospective NHANES dataset. Furthermore, adjustment for covariates allowed us to identify confounding factors that affected kidney stones. Our findings were validated by multiple imputation sensitivity analyses. Compare with the previous study, we updated the dataset to 2018 with a larger sample size and subgroup analysis to illustrate the risk of higher SII on kidney stone formation.
There were also some limitations. First, the cross-sectional design of the NHANES indicated that causal links cannot be established. Second, the interview forum for data collection may lead to potential bias. Third, some asymptomatic kidney stones without physical examination were missed in the database. Fourth, the immune cell count was obtained from one blood test only. Serial testing may be more reliable than one blood test on account of the blood cell life span. Finally, there were still some unobserved confounding factors that might be missed.
## Conclusion
This cross-sectional study suggested SII was positively associated with kidney stones in US adults aged less than 50. The outcome compensated for previous studies that still needed more large-scale prospective cohorts for validation.
## Data availability statement
Publicly available datasets were analyzed in this study. This data can be found here: NHANES database (https://www.cdc.gov/nchs/nhanes/index.htm).
## Author contributions
Conceptualization and Methodology: XJ and X-PD. Data curation and Project administration: X-PD and L-YX. Investigation and formal analysis:L-YX and S-ZL. Manuscript Writing- Original draft:X-PD. Manuscript editing and manuscript review:XJ and S-ZL. All authors contributed to the article and approved the submitted version.
## Conflict of 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.
## 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.2023.1116224/full#supplementary-material
## References
1. Mao W, Hu Q, Chen S, Chen Y, Luo M, Zhang Z. **Polyfluoroalkyl chemicals and the risk of kidney stones in US adults: A population-based study**. *Ecotoxicol Environ Saf.* (2021) **208** 111497. DOI: 10.1016/j.ecoenv.2020.111497
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. Kittanamongkolchai W, Vaughan LE, Enders FT, Dhondup T, Mehta RA, Krambeck AE. **The changing incidence and presentation of urinary stones over 3 decades**. *Mayo Clin Proc* (2018) **93**. DOI: 10.1016/j.mayocp.2017.11.018
4. Hiatt RA, Dales LG, Friedman GD, Hunkeler EM. **Frequency of urolithiasis in a prepaid medical care program**. *Am J Epidemiol.* (1982) **115**. DOI: 10.1093/oxfordjournals.aje.a113297
5. Lieske JC, Peña de la Vega LS, Slezak JM, Bergstralh EJ, Leibson CL, Ho KL. **Renal stone epidemiology in Rochester, Minnesota: an update**. *Kidney Int* (2006) **69**. DOI: 10.1038/sj.ki.5000150
6. Chooi YC, Ding C, Magkos F. **The epidemiology of obesity**. *Metabolism* (2019) **92**. DOI: 10.1016/j.metabol.2018.09.005
7. Geiss LS, Wang J, Cheng YJ, Thompson TJ, Barker L, Li Y. **Prevalence and incidence trends for diagnosed diabetes among adults aged 20 to 79 years, united states, 1980-2012**. *JAMA* (2014) **312**. DOI: 10.1001/jama.2014.11494
8. Meyer KA, Harnack LJ, Luepker RV, Zhou X, Jacobs DR, Steffen LM. **Twenty-two-year population trends in sodium and potassium consumption: the Minnesota heart survey**. *J Am Heart Assoc* (2013) **2**. DOI: 10.1161/JAHA.113.000478
9. Daniel CR, Cross AJ, Koebnick C, Sinha R. **Trends in meat consumption in the USA**. *Public Health Nutr* (2011) **14**. DOI: 10.1017/S1368980010002077
10. Bleich SN, Wang YC, Wang Y, Gortmaker SL. **Increasing consumption of sugar-sweetened beverages among US adults: 1988-1994 to 1999-2004**. *Am J Clin Nutr* (2009) **89**. DOI: 10.3945/ajcn.2008.26883
11. Trinchieri A. **Epidemiological trends in urolithiasis: impact on our health care systems**. *Urol Res* (2006) **34**. DOI: 10.1007/s00240-005-0029-x
12. Assasi N, Blackhouse G, Campbell K, Hopkins RB, Levine M, Richter T. *Comparative value of erythrocyte sedimentation rate (ESR) and c-reactive protein (CRP) testing in combination versus individually for the diagnosis of undifferentiated patients with suspected inflammatory disease or serious infection: A systematic review and economic analysis* (2015)
13. Lee KS, Ha JS, Koo KC. **Significance of neutrophil-to-Lymphocyte ratio as a novel indicator of spontaneous ureter stone passage**. *Yonsei Med J* (2017) **58**. DOI: 10.3349/ymj.2017.58.5.988
14. Abou Heidar N, Labban M, Bustros G, Nasr R. **Inflammatory serum markers predicting spontaneous ureteral stone passage**. *Clin Exp Nephrol.* (2020) **24**. DOI: 10.1007/s10157-019-01807-5
15. Cetinkaya M, Buldu I, Kurt O, Inan R. **Platelet-to-Lymphocyte ratio: A new factor for predicting systemic inflammatory response syndrome after percutaneous nephrolithotomy**. *Urol J* (2017) **14**
16. Liu J, Li S, Zhang S, Liu Y, Ma L, Zhu J. **Systemic immune-inflammation index, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio can predict clinical outcomes in patients with metastatic non-small-cell lung cancer treated with nivolumab**. *J Clin Lab Anal* (2019) **33**. DOI: 10.1002/jcla.22964
17. Trifan G, Testai FD. **Systemic immune-inflammation (SII) index predicts poor outcome after spontaneous supratentorial intracerebral hemorrhage**. *J Stroke Cerebrovasc Dis* (2020) **29** 105057. DOI: 10.1016/j.jstrokecerebrovasdis.2020.105057
18. Liu Y, Ye T, Chen L, Jin T, Sheng Y, Wu G. **Systemic immune-inflammation index predicts the severity of coronary stenosis in patients with coronary heart disease**. *Coron Artery Dis* (2021) **32**. DOI: 10.1097/MCA.0000000000001037
19. Curtin LR, Mohadjer LK, Dohrmann SM, Kruszon-Moran D, Mirel LB, Carroll MD. **National health and nutrition examination survey: sample design, 2007-2010**. *Vital Health Stat 2* (2013) **160**
20. Hu B, Yang X-R, Xu Y, Sun Y-F, Sun C, Guo W. **Systemic immune-inflammation index predicts prognosis of patients after curative resection for hepatocellular carcinoma**. *Clin Cancer Res* (2014) **20**. DOI: 10.1158/1078-0432.CCR-14-0442
21. Li H, Wu X, Bai Y, Wei W, Li G, Fu M. **Physical activity attenuates the associations of systemic immune-inflammation index with total and cause-specific mortality among middle-aged and older populations**. *Sci Rep* (2021) **11** 12532. DOI: 10.1038/s41598-021-91324-x
22. Lee JA, Johns TS, Melamed ML, Tellechea L, Laudano M, Stern JM. **Associations between socioeconomic status and urge urinary incontinence: An analysis of NHANES 2005 to 2016**. *J Urol.* (2020) **203**. DOI: 10.1097/JU.0000000000000542
23. Xie R, Xiao M, Li L, Ma N, Liu M, Huang X. **Association between SII and hepatic steatosis and liver fibrosis: A population-based study**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.925690
24. Draznin B, Aroda VR, Bakris G, Benson G, Brown FM, Freeman R. **9. pharmacologic approaches to glycemic treatment: Standards of medical care in diabetes-2022**. *Diabetes Care* (2022) **45**. DOI: 10.2337/dc22-S009
25. 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
26. Su YS, Gelman A, Hill J, Yajima M. **Multiple imputation with diagnostics (mi) in r: Opening windows into the black box**. *J OF Stat Software* (2011) **45** 1-31. DOI: 10.18637/jss.v045.i02
27. Mao W, Wu J, Zhang Z, Xu Z, Xu B, Chen M. **Neutrophil-lymphocyte ratio acts as a novel diagnostic biomarker for kidney stone prevalence and number of stones passed**. *Transl Androl Urol.* (2021) **10** 77-86. DOI: 10.21037/tau-20-890
28. Chen L, Zhang Y. **Comparison of the diagnostic values of leukocytes, neutrophils, neutrophil-to-Lymphocyte ratio and platelet-to-Lymphocyte ratio in distinguishing between acute appendicitis and right ureterolithiasis**. *Clin Lab* (2020) **66**. DOI: 10.7754/Clin.Lab.2019.190632
29. Senel C, Aykanat IC, Asfuroglu A, Keten T, Balci M, Aslan Y. **What is the role of inflammatory markers in predicting spontaneous ureteral stone passage**. *Aktuelle Urol* (2022) **53**. DOI: 10.1055/a-1703-3099
30. He L, Xie X, Xue J, Xie H, Zhang Y. **Association of the systemic immune-inflammation index with all-cause mortality in patients with arteriosclerotic cardiovascular disease**. *Front Cardiovasc Med* (2022) **9**. DOI: 10.3389/fcvm.2022.952953
31. Kumar P, Yang Z, Lever JM, Chávez MD, Fatima H, Crossman DK. **Hydroxyproline stimulates inflammation and reprograms macrophage signaling in a rat kidney stone model**. *Biochim Biophys Acta Mol Basis Dis* (2022) **1868** 166442. DOI: 10.1016/j.bbadis.2022.166442
32. Taguchi K, Okada A, Hamamoto S, Unno R, Moritoki Y, Ando R. **M1/M2-macrophage phenotypes regulate renal calcium oxalate crystal development**. *Sci Rep* (2016) **6** 35167. DOI: 10.1038/srep35167
33. 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
34. Singhto N, Thongboonkerd V. **Exosomes derived from calcium oxalate-exposed macrophages enhance IL-8 production from renal cells, neutrophil migration and crystal invasion through extracellular matrix**. *J Proteomics.* (2018) **185** 64-76. DOI: 10.1016/j.jprot.2018.06.015
35. Revúsová V, Pekníková D, Polakovicová D, Breza J. **Blood lymphocyte magnesium in kidney stone formers**. *Int Urol Nephrol.* (1989) **21**. DOI: 10.1007/BF02559632
36. Mezger M, Nording H, Sauter R, Graf T, Heim C, von Bubnoff N. **Platelets and immune responses during thromboinflammation**. *Front Immunol* (2019) **10**. DOI: 10.3389/fimmu.2019.01731
37. Hoffman A, Braun MM, Khayat M. **Kidney disease: Kidney stones**. *FP Essent* (2021) **509**
38. Taylor EN, Stampfer MJ, Curhan GC. **Dietary factors and the risk of incident kidney stones in men: new insights after 14 years of follow-up**. *J Am Soc Nephrol.* (2004) **15**. DOI: 10.1097/01.ASN.0000146012.44570.20
39. Xu M, Chen R, Liu L, Liu X, Hou J, Liao J. **Systemic immune-inflammation index and incident cardiovascular diseases among middle-aged and elderly Chinese adults: The dongfeng-tongji cohort study**. *Atherosclerosis* (2021) **323**. DOI: 10.1016/j.atherosclerosis.2021.02.012
|
---
title: 'The effect of intermittent fasting on the clinical and hematological parameters
of patients with sickle cell disease: A preliminary study'
authors:
- Khalid Ahmed
- Yasamin Abdu
- Sief Khasawneh
- Ahmed Shukri
- Ehab Adam
- Salma Mustafa
- Mohammad Affas
- Mohamed Izham Mohamed Ibrahim
- Abdullah Al Zayed
- Mohamed A. Yassin
journal: Frontiers in Medicine
year: 2023
pmcid: PMC9989014
doi: 10.3389/fmed.2023.1097466
license: CC BY 4.0
---
# The effect of intermittent fasting on the clinical and hematological parameters of patients with sickle cell disease: A preliminary study
## Abstract
### Introduction
Sickle cell disease is a genetic disorder that frequently presents with vaso-occlusive crisis (VOC). Most patients with sickle cell disease in Qatar are Muslims; hence, they practice intermittent fasting during the holy month of Ramadan. However, there is a paucity of literature describing the effect of intermittent fasting on the occurrence of severe VOC. As a result, there is a lack of guidelines or standardized protocols that can help physicians advise patients with sickle cell disease who wish to practice intermittent fasting. Therefore, this study's aim was to investigate the effect of intermittent fasting on the clinical and hematological parameters of individuals with sickle cell disease.
### Methods
We conducted a retrospective study for 52 Muslim patients with sickle cell disease in Qatar aged ≥18 years who were confirmed to be fasting during the holy month of Ramadan during any of the years 2019–2021. The difference in the occurrence of severe VOC, hemolytic crisis, and other clinical, hematological, and metabolic parameters were studied one month before, during, and one month after the intermittent fasting of Ramadan using the patient's medical records. Mean (sd), median (IQR), and frequency (%) described the data. One-way with repeated measures ANOVA with a Greenhouse-Geisser correction and Friedman tests (*) were used at alpha level 0.05.
### Results
The study participants' (mean±sd) age was (31.1±9.2) years, $51.9\%$ were males, and $48.1\%$ were females. Roughly seventy percent of the participants were of Arab ethnicity, while the rest were either African or Asian. Most of the patients were homozygotes (SS) ($90.4\%$). The median number of severe VOC ($$P \leq 0.7$$) and hemolytic crisis ($$P \leq 0.5$$) was not found to be significantly different before, during, or after Ramadan. Significant differences, however, were found in platelet count ($$P \leq 0.003$$), reticulocyte count ($P \leq 0.001$), and creatinine level ($$P \leq 0.038$$) with intermittent fasting.
### Discussion
In this preliminary study, intermittent fasting does not seem to influence the rate of occurrence of severe vaso-occlusive crisis or hemolytic crisis in patients with sickle cell disease; however, it was found to be associated with differences in platelet count, reticulocytes count, and creatinine level. The statistical and clinical significance of these findings needs to be confirmed in studies with a larger sample size.
## Introduction
Sickle cell disease (SCD) is a genetic disorder that results from a mutation in the beta-globin gene resulting in the replacement of the amino acid glutamic acid with valine at position number 6, Patients who are homozygous for the mutation have sickle cell anemia and represent the most prevalent and severe form of the disease. The disease can also result from compound heterozygote mutations such as S/C or S/β thalassemia with variable clinical expression [1]. The most common complication of SCD is the vaso-occlusive crisis (VOC) which presents with severe body pain. SCD is also a systemic disease that can potentially involve all organs [2]. Acute complications of SCD include infection, acute chest syndrome, priapism, stroke, splenic sequestration, hepatobiliary complications, and acute kidney injury [3]. Chronic complications include avascular bone necrosis, pulmonary hypertension, heart failure, renal insufficiency requiring dialysis, retinopathy, and leg ulcers [3].
Intermittent fasting has recently gained much popularity as an effective method to reduce weight. It is now practiced by many people, including Muslims, during the holy month of Ramadan. There are different types of intermittent fasting; Ramadan fasting represents the $\frac{16}{8}$ method (fasting for 16 h a day) [4]. Calorie deprivation for some time can induce dynamic cellular changes, which can manifest in clinical changes. Since Ramadan fasting represents one of the five pillars of the Islamic faith, there is a solid public structure to support fasting during Ramadan. We observe an increase in the number of SCD patients visiting healthcare facilities during the holy month of Ramadan; however, there is not much literature available about the effect of intermittent fasting on developing SCD complications such as severe VOC; as a result, no guidelines or standardized protocols exist which can help physicians advise individuals with sickle cell disease who wish to practice intermittent fasting.
The prevalence of sickle cell disease in the Arabian gulf countries is reported in the literature from Qatar ($3.9\%$), Bahrain ($2.1\%$), Oman ($3.8\%$), Yemen ($0.95\%$), United Arab Emirates ($1.9\%$), and Saudi Arabia (0.01–$0.1\%$) (5–9).
This study investigated the effect of intermittent fasting during Ramadan on developing severe VOC and hemolytic crisis, as well as differences in the hematological and metabolic parameters of SCD patients before, during, and after intermittent fasting. Subsequently, the results of this study will enable our physicians to have some data upon which they can advise individuals with SCD about intermittent fasting. It will also lay a foundation for further studies addressing this topic. In the long term, the recommendations from this study and future similar studies can help reduce the healthcare cost that results from frequent admissions to the hospital and the utilization of laboratory services, medications, and other healthcare services to address sickle cell disease complications.
## Methods
After obtaining approval from the institutional research board (IRB), we retrospectively reviewed the medical records of 145 adult Muslim patients diagnosed with sickle cell disease, living in Qatar, and following up with the hematology department in a tertiary care center. Fifty-two patients were confirmed to be fasting during part or all of Ramadan for any of the years 2019–2021; the research team confirmed the fasting status through a telephone script. The participants fulfilled the inclusion criteria of being 18 years or older, Muslim, and residing in Qatar. We excluded patients younger than 18 years, non-Muslim, sickle cell trait, pregnant ladies, patients with established chronic kidney disease, or patients who were confirmed to have an infection, conditions with fluid loss that can potentially lead to dehydration (e.g., severe diarrhea, vomiting, polyuria, etc.) or any other apparent precipitating factor for VOC during fasting.
The following outcome variables before (within 1 month of Ramadan), during, and post-Ramadan (till 1 month after) were obtained by screening the electronic medical records of the study participants: Excel and SPSS program v 29 were used for data management and analysis. Mean (sd), median (IQR), and frequency (%) described the data. Inferential statistics were applied to determine the differences and associations between variables. One-way with repeated measures ANOVA with a Greenhouse-Geisser correction and Friedman tests (*) were used to measure the difference over time. A priori significance level was set at 0.05.
## Results
Table 1 below describes the patients' demographic characteristics and their fasting profiles. Patients' average age (mean) ± (SD) was (31.1) ± (9.2). Males were slightly more than females ($51.9\%$). Most study participants were Arab ($69.2\%$) and non-Qatari ($63.6\%$). The homozygote form (SS) represented the most common type of SCD ($90.4\%$). The average number of days fasted by the participants (mean) ± (SD) was 25.2 ± [6] indicating that the participants fasted most of the holy month of Ramadan. The (mean) ± (SD) for the severe vaso-occlusive crisis among the study participants in the year immediately preceding the intermittent fasting of Ramadan was 2.1 ± (1.2). The (mean) ± (SD) of Hemoglobin F among the study participants was ($14.3\%$) ± ($6.3\%$). Table 1 also describes whether the study participants received a top-up (simple PRBCs) transfusion 1 month prior to, during, or up to 1 month after Ramadan, most patients did not receive a transfusion in any of the three time periods.
**Table 1**
| Item | Unnamed: 1 | Mean (SD) | Median (IQR) | n (%) |
| --- | --- | --- | --- | --- |
| Age | | 31.1 (9.2) | 31.0 (23.3–35.8) | |
| Gender | Male | | | 27 (51.9) |
| | Female | | | 25 (48.1) |
| Nationality | Qatari | | | 19 (36.5) |
| | Non-qatari | | | 33 (63.6) |
| Region | Arab | | | 36 (69.2) |
| | African | | | 10 (19.2) |
| | Asian | | | 6 (11.5) |
| Type of SCD | SS | | | 47 (90.4) |
| | SC | | | 2 (3.8) |
| | SD | | | 3 (5.8) |
| Number of fasting days | | 25.2 (6.0) | 28.5 (22.0–30.0) | |
| Compliance with hydroxyurea during Ramadan* | Yes | | | 20 (38.5) |
| | No | | | 16 (30.8) |
| | Not applicable | | | 16 (30.8) |
| Number of VOC up to 1 year before Ramadan | | 2.1 (1.2) | 2 (1–3) | |
| Hemoglobin F level before Ramadan | | 14.3 (6.3) | 13.2 (9.8–17.5) | |
| Simple PRBCs transfusion 1 month prior to Ramadan | Yes | | | 2 (3.8) |
| | No | | | 50 (96.2) |
| Simple PRBCs transfusion during Ramadan | Yes | | | 2 (3.8) |
| | No | | | 50 (96.2) |
| Simple PRBCs transfusion 1 month after Ramadan | Yes | | | 1 (1.9) |
| | No | | | 51 (98.1) |
Table 2 illustrates the clinical and laboratory parameters among SCD patients 1 month before, during, and 1 month after Ramadan. One-way with repeated measures ANOVA and Friedman test indicated no significant difference in all the clinical parameters tested over time. One-way with repeated measures ANOVA with a Greenhouse-Geisser correction indicated significant differences for changes in platelet count ($$P \leq 0.003$$), reticulocytes count (<0.001), and creatinine level ($$P \leq 0.038$$) over time. Pairwise comparisons using the Bonferroni test indicated significant differences for the following: for platelet count, the significant difference was seen during intermittent fasting compared to post-intermittent fasting, with platelet count being lower during the fasting period; for reticulocyte count, the significant difference was seen when comparing the pre intermittent fasting values to the values during intermittent fasting as well as when comparing the values during intermittent fasting to the values in the post intermittent fasting period with reticulocytes count being lower during the fasting period, regarding creatinine the significant difference was found between the values in the fasting period compared to the post fasting period with creatinine being higher during Ramadan. With the consideration of covariates (age, gender, region, type of SCD, days of fasting, and compliance with hydroxyurea), there is a significant influence of region on the platelet count (P ≤ 0.001), none influenced the reticulocyte count (P ≥ 0.05) or creatinine level (P ≥ 0.05).
**Table 2**
| Item | Normal range | Unnamed: 2 | 1 month before | During | 1 month after | p-value |
| --- | --- | --- | --- | --- | --- | --- |
| Number of VOC | | Mean (SD) | 0.21 (0.41) | 0.27 (0.45) | 0.23 (0.43) | 0.715 |
| Number of hemolytic crises | | Mean (SD) | 0.08 (0.27) | 0.13 (0.35) | 0.13 (0.35) | 0.535 |
| Length of hospital stay (days) | | Mean (SD) | 1.12 (2.4) | 1.37 (3.52) | 0.88 (1.88) | 0.623 |
| The total dose of morphine (mg) | | Mean (SD) | 24.8 (59.0) | 15.8 (55.0) | 7.12 (16.96) | 0.182 |
| Use of antibiotics | | Yes (n, %) | 6 (11.5) | 7 (13.5) | 10 (19.2) | 0.486* |
| | | No (n, %) | 46 (88.5) | 45 (86.5) | 42 (80.8) | |
| The need for exchange transfusion | | Yes (n, %) | 3 (5.8) | 2 (3.8) | 1 (1.9) | 0.607* |
| | | No (n, %) | 49 (94.2) | 50 (96.2) | 51 (98.1) | |
| Admission to ICU | | Yes (n, %) | 3 (5.8) | 2 (3.8) | 1 (1.9) | 0.607* |
| | | No (n, %) | 49 (94.2) | 50 (96.2) | 51 (98.1) | |
| Number of ICU days | | Mean (SD) | 0.13 (0.63) | 0.12 (0.51) | 0.00 (0.00) | 0.596 |
| WBC count × 103/μL | 4–10 | Mean (SD) | 8.62 (3.66) | 7.75 (3.03) | 9.07 (4.76) | 0.212 |
| Hb (g/dL) | 13–17 | Mean (SD) | 9.86 (1.50) | 9.74 (1.91) | 9.97 (1.49) | 0.761 |
| Platelet count × 103/μL | 150–410 | Mean (SD) | 271.39 (138.92) | 234.35 (104.40) | 317.08 (158.90) | 0.003 |
| Reticulocytes count × 103/ μL | 50–100 | Mean (SD) | 198.84 (76.23) | 150.34 (55.58) | 197.36 (77.17) | < 0.001 |
| Urea (mmol/L) | 2.5–7.8 | Mean (SD) | 2.85 (1.31) | 3.08 (0.98) | 6.04 (10.95) | 0.098 |
| Creatinine (μmol/L) | 62–106 | Mean (SD) | 65.69 (2.42) | 67.90 (19.35) | 60.38 (22.05) | 0.038 |
| Bilirubin (μmol/L) | 0–21 | Mean (SD) | 33.89 (26.10) | 25.27 (13.88) | 31.87 (24.35) | 0.063 |
| LDH (U/L) | 105–333 | Mean (SD) | 365.89 (150.12) | 355.67 (153.83) | 444.00 (136.03) | NA** |
## Discussion
This preliminary study intended to measure the effect of intermittent fasting on the clinical, metabolic, and hematological parameters of patients with sickle cell disease. Fasting is the voluntary abstinence from or reduction of some or all food, drink, or both (absolute) for a period typically between 12 h and 3 weeks, i.e., in a short-term, long-term/prolonged, or intermittent pattern [10]. Intermittent fasting has many types, including the Eat-Stop-Eat diet, 24 h fast, 5:2 diet; fasting for 2 days per week, once or twice per week, alternate day fasting, the warrior diet; fasting during the day and eating a big meal at night, spontaneous meal skipping and the $\frac{16}{8}$ method; fasting for 16 h each day [4]. Ramadan fasting is primarily consistent with the $\frac{16}{8}$ method; it is practiced by millions of Muslims all over the world for a whole lunar month (29–30 days) every year, and it involves abstaining from all types of food or drinks from dawn to sunset time [11]. A unique difference between Ramadan fasting and the $\frac{16}{8}$ method of intermittent fasting is that most Muslims take a meal 1–2 h before dawn, in this sense it allows for good hydration before starting the fasting. Calorie restriction during intermittent fasting can affect metabolic regulation, e.g., by altering circadian biology, the gut microbiome and modifiable lifestyle behavior and this, in turn, has been shown to have major public health benefits [4, 12]. A systematic review by Stephanie et al. of 27 trials addressing weight loss in obese and overweight individuals (18 randomized controlled trials and nine trials that compared weight post intermittent fasting to baseline weight without a control group) found that studies with intermittent fasting reported a 0.8 to $13\%$ weight loss compared to the baseline weight without serious adverse events; moreover, better glycemic control was seen in the studies that recruited patients with type 2 diabetes [13]. In Germany, a one-year follow-up of 1,422 individuals on an intermittent fasting diet reported lower systolic and diastolic blood pressure in those who fasted for a longer time; the proposed mechanism for the reduction in blood pressure is increased parasympathetic activity, increased renal excretion of norepinephrine and improved sensitivity to insulin and natriuretic peptide [14]. A systematic review and meta-analysis of six studies involving 417 patients with non-alcoholic fatty liver disease reported a significant difference in body weight, body mass index, and aspartate transaminase between the fasting and the control group and no significant differences in the triglyceride level and the total cholesterol [15].
The examples mentioned above illustrate the role intermittent fasting may play in the prevention and management of chronic diseases; nevertheless, no studies have investigated the role of intermittent fasting in patients with sickle cell disease. With sickle cell disease being one of the most prevalent genetic diseases worldwide and particularly in the middle east region, where most of the population are Muslims, it becomes essential to study the effect of intermittent fasting on the course of this chronic disease.
Our findings indicate that there is no statistically significant difference in the rate of occurrence of severe vaso-occlusive crisis or hemolytic crisis in individuals with sickle cell disease who practiced intermittent fasting. Our research, however, investigated only one type of intermittent fasting, which is the $\frac{16}{8}$ method; further research needs to be directed toward sickle cell disease individuals who practice other types of intermittent fasting and increasing the sample size by recruiting more patients from neighboring middle east countries.
Our research indicates a significant difference in platelet count during intermittent fasting compared to post-intermittent fasting, with platelet count being lower during the fasting period, however, the mean did not fall in the thrombocytopenia range (< 150 × 103/μL) and none of the study participants developed any bleeding during the fasting of Ramadan. Whether this reduction in platelets count with fasting could have a clinical significance on patients with sickle cell disease (e.g., would it affect the rate of occurrence of thrombotic events? Can intermittent fasting be used to lower platelet count in SCD patients?) or remains a statistical finding is not yet clear and more research is needed to confirm this finding and answer this question. Likewise, the reticulocyte count was significantly different during intermittent fasting compared to both before and post-intermittent fasting, indicating that the reticulocyte count acutely drops during fasting and quickly recovers when fasting is over. These two findings suggest that intermittent fasting can affect hematopoiesis; however, the exact mechanism is yet to be studied, and reduced calorie intake may be postulated as a cause for the slightly depressed platelet and reticulocyte count during intermittent fasting. Regarding kidney function, our research indicates that there is a significant difference in creatinine during intermittent fasting as compared to post-intermittent fasting; the repeated prolonged periods of reduced water intake may account for this, given that patients with sickle cell disease have varying degrees of sickle nephropathy which may render their kidneys sensitive to any periods of reduced perfusion. Interestingly an experimental study conducted in Senegal investigated the effect of Ramadan fasting on the hematocrit and blood viscosity by measuring these values in an experimental group of 10 patients with sickle cell trait and a control group of 10 patients without sickle cell trait during the fasting of Ramadan and 6 weeks after Ramadan. The measurements were done twice (at 8 AM and 6 PM) to account for intraday variations. While hematocrit did not differ between the two groups, it was greater in the evening compared to the morning regardless of the fasting status, the hematocrit mean was reported as (43 ± 1) in the morning for both groups compared to (45 ± 1) in the evening for both groups with $P \leq 0.001.$ On the other hand, while no significant difference occurred in the blood viscosity for the control group throughout the day whether fasting or not, patients with sickle cell trait had a significant increase in blood viscosity in the evening during the fasting which might indicate a higher risk of impaired blood flow in the microcirculation [16].
## Conclusion
In this preliminary study, intermittent fasting during the holy month of Ramadan does not seem to influence the rate of occurrence of severe vaso-occlusive crisis or hemolytic crisis in patients with sickle cell disease; however, it can be associated with differences in platelet count, reticulocyte count, and creatinine level. Until stronger evidence is available from larger studies the decision to allow fasting for sickle cell disease patients should be carefully considered on a one-by-one basis considering the clinical characteristics of each patient. close follow-up from the treating physician is required with advice on when to break fasting. More studies with a bigger sample size are needed to confirm these results. Additionally, more studies are needed to look at types of intermittent fasting other than the $\frac{16}{8}$ method.
## 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 Institutional Review Board-Hamad Medical Corporation. The patients/participants provided their written informed consent to participate in this study.
## 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. Ware RE, de Montalembert M, Tshilolo L, Abboud MR. **Sickle cell disease**. *Lancet.* (2017) **390** 311-23. DOI: 10.1016/S0140-6736(17)30193-9
2. Ballas SK, Lieff S, Benjamin LJ, Dampier CD, Heeney MM, Hoppe C. **Definitions of the phenotypic manifestations of sickle cell disease**. *Am J Hematol.* (2010) **85** 6-13. DOI: 10.1002/ajh.21550
3. Habibi A, Arlet JB, Stankovic K, Gellen-Dautremer J, Ribeil JA, Bartolucci P, Lionnet F. **Recommandations françaises de prise en charge de la drépanocytose de l'adulte: actualisation 2015**. *La Revue de Médecine Interne.* (2015) **36** 5S3-84. DOI: 10.1016/S0248-8663(15)60002-9
4. Patterson RE, Sears DD. **Metabolic effects of intermittent fasting**. *Annu Rev Nutr.* (2017) **37** 64634. DOI: 10.1146/annurev-nutr-071816-064634
5. Elmoneim AA A, Al Hawsawi ZM, Mahmoud BZ, Bukhari AA, Almulla AA, Sonbol AM. **Causes of hospitalization in sickle cell diseased children in western region of Saudi Arabia. A single center study**. *Saudi Med J.* (2019) **40** 401-4. DOI: 10.15537/smj.2019.4.24049
6. Al-Riyami A, Ebrahim GJ. **Genetic blood disorders survey in the Sultanate of Oman**. *J Trop Pediatr.* (2003) **49** i1-20. PMID: 12934793
7. 7.Sickle Cell Anemia. Centre for Arab Genomic Studies. Bur Dubai, United Arab Emirates (2018). Available online at: http://www.cags.org.ae/ctga/details.aspx?id=286&keyword=Sickle+Cell+Disease&se=Latest#epidemiology (accessed June, 2018).. *Centre for Arab Genomic Studies* (2018)
8. Mohammed AM, Al-Hilli F, Nadkarni KV, Bhagwat GP, Bapat JP. **Hemoglobinopathies and glucose-6-phosphate dehydrogenase deficiency in hospital births in Bahrain**. *Ann Saudi Med.* (1992) **12** 536-9. PMID: 17587043
9. White JM, Byrne M, Richards R, Buchanan T, Katsoulis EL, Weerasingh KA. **Red cell genetic abnormalities in Peninsular Arabs: sickle haemoglobin, G6PD deficiency, and alpha and beta thalassaemia**. *J Med Genet.* (1986) **23** 245-51. PMID: 3723553
10. Patterson RE, Laughlin GA, Sears DD, LaCroix AZ, Marinac C, Gallo LC. **Intermittent fasting and human metabolic health**. *J Cad Nutr Diet.* (2015) **115** 1203-12. DOI: 10.1016/j.jand.2015.02.018
11. Ali MM. *The Holy Quran* (2011) 183-185
12. Brown JE, Mosley M, Aldred S. **Intermittent fasting: a dietary intervention for prevention of diabetes and cardiovascular disease?**. *Br J Diabetes Vasc.* (2013) **13** 68-72. DOI: 10.1177/1474651413486496
13. Welton S, Minty R, O'Driscoll T, Willms H, Poirier D, Madden S. **Intermittent fasting and weight loss: systematic review**. *Can Family Phys.* (2020) **66** 117-25. PMID: 36126910
14. Toledo FW, Grundler F, Bergouignan A, Drinda S, Michalsen A. **Safety, health improvement and well-being during a 4 to 21-day fasting period in an observational study including 1,422 subjects**. *PLoS ONE.* (2019) **14** e0209353. DOI: 10.1371/journal.pone.0209353
15. Yin C, Li Z, Xiang Y, Peng H, Yang P, Yuan S. **Effect of intermittent fasting on non-alcoholic fatty liver disease: systematic review and meta-analysis**. *Front Nutr.* (2021) **8** 709683. DOI: 10.3389/fnut.2021.709683
16. Diaw M, Connes P, Samb A, Sow AK, Sall ND, Sar FB. **Intraday blood rheological changes induced by Ramadan fasting in sickle cell trait carriers**. *Chronobiol Int.* (2013) **30** 1116-22. DOI: 10.3109/07420528.2013.804083
|
---
title: How information processing and risk/benefit perception affect COVID-19 vaccination
intention of users in online health communities
authors:
- Hao Liu
- Liyue Gong
- Cao Wang
- Yunyun Gao
- Yi Guo
- Minhan Yi
- Hao Jiang
- Xusheng Wu
- Dehua Hu
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9989022
doi: 10.3389/fpubh.2023.1043485
license: CC BY 4.0
---
# How information processing and risk/benefit perception affect COVID-19 vaccination intention of users in online health communities
## Abstract
### Objective
To investigate the relationship among information processing, risk/benefit perception and the COVID-19 vaccination intention of OHCs users with the heuristic-systematic model (HSM).
### Methods
This study conducted a cross-sectional questionnaire via an online survey among Chinese adults. A structural equation model (SEM) was used to examine the research hypotheses.
### Results
Systematic information processing positively influenced benefit perception, and heuristic information processing positively influenced risk perception. Benefit perception had a significant positive effect on users' vaccination intention. Risk perception had a negative impact on vaccination intention. Findings revealed that differences in information processing methods affect users' perceptions of risk and benefit, which decide their vaccination intention.
### Conclusion
Online health communities can provide more systematic cues and users should process information systematically to increase their perceived benefits, consequently increase their willingness to receive COVID-19 vaccine.
## 1. Introduction
The COVID-19 pandemic has caused a major health crisis in humans [1]. As of 13 Sep 2022, the World Health Organization has reported more than 600 million cumulative confirmed cases of COVID-19 and more than 6 million cumulative deaths [2]. Ensuring that all people are vaccinated will help control the spread of COVID-19 [3] and thus protect the public from COVID-19 [4].
In China, the government legislated an emergency authorization of COVID-19 vaccine for people at high risk in June 2020 [5], and subsequently approved COVID-19 vaccines for public use in December 2020 [5, 6]. As of March 2020, China has passed through its peak of the pandemic. However, *China is* still experiencing a small increase in cases due to the impact of COVID-19 mutations and the importation of cases from abroad. Vaccination and testing are the main vaccination policies in China. As of July 22, 2022, the first full-round vaccination rate reached $89.7\%$ and the booster vaccination rate was $71.7\%$ [7]. Although overall the vaccination rate is relatively high. However, the booster vaccination rate is much lower compared to the first full vaccination rate. In order to strengthen the protective efficacy of the vaccine against COVID-19 mutations, ensuring the booster vaccination rate is an effective measure. In this context, continued attention to the factors influencing the intention to vaccinate against COVID-19 can inform the maintenance of the intention to vaccinate against COVID-19, the improvement of booster vaccination rates, and the development of vaccination policies.
Previous studies have explored the effect of perceived risk/benefit on intention to vaccinate for COVID-19 (8–10). For example, a study by Liora Shmueli showed that perceived benefit was the most important predictor of acceptance of the COVID-19 vaccine [11]. Another study showed a strong correlation between risk perception and vaccine acceptance [12]. And vaccine-related information affects users' perceptions of the risk and benefit of vaccines. Users' processing of vaccine-related information shapes their perception of vaccines [13, 14]. Some studies have found that having the correct knowledge is directly related to a higher perception of risk in the older population [15]. Knowledge about vaccines was associated with how individuals perceived the relevant risks and benefits of those aspects of the vaccine [16]. Information related to the efficacy and safety of vaccines critically influences the acceptance of COVID-19 vaccines [4]. Knowledge formation comes from information processing. However, there are no studies that have explored the effects of perceived risks/benefits on the willingness to vaccinate for COVID-19 from an information processing perspective. Processing of vaccine-related information is a key factor in the formation of people's perceived attitudes. As such, further research is necessary to determine how information processing affects risk/benefit perception associated with COVID-19 vaccination intention.
As access to information through the Internet has the advantage of being quick and convenient, online health communities (OHCs) have become one of the most important channels through which people obtain information during a pandemic (17–19). People used OHCs to learn about COVID-19 and seek information about available vaccines [20, 21]. OHCs are online interactive platforms with health-related features, such as online consultation, health information exchange and experience sharing, which provide users with information and emotion support [19]. Users can also benefit from OHCs by adopting healthier behaviors [22].
Therefore, in the context of the rapid development of Internet medicine and the normalization of COVID-19 prevention and control. Our study explores the impact of users' vaccine-related information processing in OHCs and their risk/benefit perception of vaccines, consequent on COVID-19 vaccination intention. Our study findings may help relevant health authorities to take more effective measures to increase vaccination rates, maintain COVID-19 vaccination intentions among Chinese residents, and provide a reference for the development of Internet healthcare.
## 2.1.1. The information behavior model
Wilson's information behavior theory suggests that users engage in information seeking through formal or informal means in order to satisfy their information needs, and then process and use the information [23]. This process is influenced by activating mechanisms (e.g., stress/coping theory, risk/reward theory) and intervening variables (e.g., psychological, demographic, role-related or interpersonal). This model has been widely used in studies related to user information behavior [24, 25]. Our study examines the factors associated with information behavior and the effect of perceived risk/benefit on willingness to vaccinate.
## 2.1.2. The heuristic-systematic model
The heuristic-systematic model (HSM) of information processing includes two types of information processing: heuristic information processing and systematic information processing [26], where systematic information processing involves a more comprehensive analysis and understanding of information. On the other hand, heuristic information processing requires only simple decision rules such as intuition and experience to form judgments [27]. The HSM has been widely used to explain people's attitude or behavior responses to information. The model considers information processing as a precursor to attitude formation or change, and therefore proposes two basic information processing patterns that people may adopt after acquiring information and assessing and judging risks or things.
## 2.2.1. The antecedents of information processing
Information needs are also known as information insufficiency, where people lack sufficient information to make informed decisions [28]. Information needs arise when the information that people want to know is more than the knowledge they have. During the COVID-19 pandemic, people need sufficient information to make decisions about whether to receive the COVID-19 vaccine [29]. Previous research found that users' demand for information on the prevention of COVID-19 accounted for $36.11\%$ in OHCs [21], which shows their great concerns and information needs about COVID-19 prevention. Some studies have suggested that information needs predict information seeking [8, 30]. We assume that information needs about the COVID-19 vaccine positively influence information seeking (H1a). People satisfy their information needs by seeking information [31]. In addition, if people do not have enough information to cope with emergencies, the more intense their information needs are, and the more actively they will use systematic processing [8, 32]. Conversely, the heuristic processing will become more active [13, 33]. Therefore, it is assumed that information needs positively influence systematic information processing (H1b) and negatively influence heuristic information processing (H1c): Hypothesis 1a (H1a). Information needs positively influence information seeking.
Hypothesis 1b (H1b). Information needs negatively influence heuristic information processing.
Hypothesis 1c (H1c). Information needs positively influence systematic information processing.
Information seeking is a dynamic process of acquiring information and knowledge [34]. People seek information through various approaches to obtain reliable information (35–37). OHCs provide a platform for people to seek and obtain information. Kahlor thinks information seeking is the precondition for information processing [38]. The research of Guo found that information seeking positively affects systematic information processing [26]. Information seeking intention is positively correlated with systematic processing and heuristic processing [39]. We proposed that when people actively seek information about COVID-19 vaccination, both heuristic information processing (H2a) and systematic information processing will improve (H2b): Hypothesis 2a (H2a). Information seeking positively affects heuristic information processing.
Hypothesis 2b (H2b). Information seeking positively affects systematic information processing.
## 2.2.2. Heuristic-systematic information processing
The heuristic-systematic information processing model states that people use one or two types of information processing to help them evaluate information to make decisions [13]. Most people will only make decisions based on superficial information cues [40]. Systematic information processing requires more comprehensive cognition and analysis by individuals [13, 41], and the process of systematic information processing consumes more time and effort on the part of the individual. Therefore, when people carry out systematic information processing, more reliable and effective information can be obtained [26]. Trumbo demonstrated that heuristic information processing negatively affects risk perception, while systematic information processing positively affects risk perception in his study about cancer [13]. Smerecnik et al. [ 42] used an adapted HSM scale to test the relationship between information processing and risk perception about hypertension. In a study about the risk associated with the companies of a petrochemical complex, systematic processing has a direct, positive, and significant influence on risk perception [43]. For benefit perception, both heuristic and systematic information processing are linked to higher benefits of using of nanotechnology [44]. However, few researchers have focused on the relationships between information processing and risk/benefit perception in the context of online health communities. Therefore, we established the following hypotheses: Hypothesis 3a (H3a). Heuristic information processing has a negative effect on risk perception.
Hypothesis 3b (H3b). Heuristic information processing has a positively impact on benefit perception.
Hypothesis 4a (H4a). Systematic information processing positively affects risk perception.
Hypothesis 4b (H4b). Systematic information processing positively affects benefit perception.
## 2.2.3. Risk/benefit perception
With the experiment and implementation of the COVID-19 vaccine, the side effects and adverse effects of the vaccination began to appear [45], which increased people's risk perception of the COVID-19 vaccine. Kelly defined risk perception as potential adverse events or side effects from taking the drug [46]. For example, myocarditis/pericarditis was a rare complication of COVID-19 mRNA vaccinations, especially in young and adolescent males [47]. The lack of effectiveness of COVID-19 vaccines may also threaten people's life, thus people perceive the risk of vaccination, which will lead to vaccine hesitancy and anti-vaccination movements [48]. Some scholars have found that risk perception negatively affects behavioral intentions [8, 49]. This means that when people are aware of the potential risks of vaccination, they may refuse to receive it. Therefore, we proposed hypothesis H5.
Contrary to risk perception, benefit perception is considered as the perception of the benefits of vaccination, such as disease prevention and self-protection [10, 50]. Wong et al. [ 9] and Yu et al. [ 10] proved the positive impact of benefit perception on vaccination intention with the health belief model. However, it has not been studied in the context of online health communities. We believed that the perceived benefit of COVID-19 vaccination information will positively affect users' willingness to vaccinate (H6): Hypothesis 5 (H5). Risk perception will weaken OHCs users' intention to receive the COVID-19 vaccine.
Hypothesis 6 (H6). Benefit perception will increase OHCs users' intention to receive the COVID-19 vaccine.
## 2.2.4. Model building
In accordance with our research hypotheses above, a new model was constructed by integrating HSM with risk/benefit perception to examine the mechanisms influencing users' willingness to vaccinate against COVID-19 in OHCs. As shown in Figure 1, in which information needs and information seeking are antecedents of information processing, information processing is assumed to predict risk/benefit perception, and risk/benefit perception directly influences vaccination intention.
**Figure 1:** *The conceptual research model. Hypothesis: H1–H6. “+” means positive effect, “–” means negative effect. The arrows point to the affected variables.*
## 3.1. Study design
This study aims to investigate the relationship between users' information needs, information seeking, heuristic-systematic information processing, risk perception, benefit perception, and vaccination intention against COVID-19 in OHCs. We conducted an online survey via Questionnaire Star (https://www.wjx.cn/ accessed on 30 June 2021). The questionnaire includes two parts: the first part is sociodemographic characteristics, namely, gender, age, education level, occupation, income, and health status; the second part is the scale measurement part, using a five-point Likert scale from “strongly disagree [1]” to “strongly agree [5],” which was adapted from previous studies. We revised it into Chinese scale and pre-tested. According to the advice of the pre-test participants and the experts group, we revised some sentences and words, for example, we changed “never” into “rarely” in the heuristic information processing items. The final scale settings are shown in Table 1.
**Table 1**
| Constructs | Items | Contents | Source |
| --- | --- | --- | --- |
| Information needs (IN) | IN1 | I need more knowledge about COVID-19 vaccination. | Ter Huurne and Gutteling (35) |
| | IN2 | I need a lot of information to decide whether to get the COVID-19 vaccine. | |
| | IN3 | I want to get more information about the COVID-19 vaccine from the OHCs. | |
| Information seeking (IS) | IS1 | I am familiar with OHCs related to the COVID-19 vaccine at home and abroad. | Ter Huurne and Gutteling (35), Che and Hu (36) |
| | IS2 | I will search for information on COVID-19 vaccinations through the OHCs. | |
| | IS3 | I will use various methods to search for more information on the COVID-19 vaccine. | |
| | IS4 | I will follow the latest information on COVID-19 vaccinations in the OHCs every day. | |
| Heuristic information processing (HIP) | HIP1 | I rarely find useful information about COVID-19 vaccinations. | Smerecnik et al. (42) |
| | HIP2 | I rarely comment on the quality of the information about COVID-19 vaccinations. | |
| | HIP3 | I rarely consider other relevant information. | |
| Systematic information processing (SIP) | SIP1 | I will think and follow up based on the information I get. | Smerecnik et al. (42) |
| | SIP2 | I will think about the importance of the information about the COVID-19 vaccine. | |
| | SIP3 | I will link COVID-19 vaccine information in the OHCs to the current COVID-19 outbreak. | |
| Risk perception (RP) | RP1 | I think the COVID-19 vaccine will bring adverse reactions. | Costa-Font and Gil (51) |
| | RP2 | I think the COVID-19 vaccine will bring unknown sequelae. | |
| | RP3 | I think getting the COVID-19 vaccine is life-threatening. | |
| Benefit perception (BP) | BP1 | I think getting vaccinated against COVID-19 will make me feel safer. | Costa-Font and Gil (51) |
| | BP2 | I think getting the COVID-19 vaccine will reduce the spread of COVID-19. | |
| | BP3 | I feel COVID-19 vaccination provides me with a new option for COVID-19 protection. | |
| Vaccination intention (VI) | VI1 | I would like to get the COVID-19 vaccine immediately. | Cheng et al. (52) |
| | VI2 | I want to get the COVID-19 vaccine in recent time. | |
| | VI3 | I plan to get the COVID-19 vaccine shortly. | |
## 3.2. Procedures and participants
Before the formal survey, this study obtained ethical approval from the Institutional Review Board of the College of Life Sciences at Central South University (Reference No. 2021-1-23). We explained the purpose and significance of our research, and provided privacy protection to all participants. All participants agreed to join this study. The investigation was administered from 1 May 2021 to 15 June 2021.
A total of 525 participants completed the questionnaire. Then we excluded the respondents who had never used OHCs before by using the option “Never used an OHC.” By removing duplicates and anomalies to ensure the validity of the data, we ended up with 410 valid questionnaires. According to the minimum sample size requirement, it must be at least 10–15 times the number of scale items [53], so our effective sample size was reasonable. Valid participants were older than 18. Among them, $64.1\%$ are women, $89.7\%$ of the participants have a bachelor's degree or above, and $78.5\%$ of users are under 30 years old. Sociodemographic characteristics are presented in Table 2.
**Table 2**
| Variables | Categories | Frequency (N = 410) | Percentage (%) |
| --- | --- | --- | --- |
| Gender | Male | 147 | 35.9 |
| | Female | 263 | 64.1 |
| Age | 18–29 | 322 | 78.5 |
| | 30–49 | 79 | 19.3 |
| | ≥50 | 9 | 2.2 |
| Education Level | Senior high school or below | 9 | 2.2 |
| | Junior college | 33 | 8.0 |
| | University | 263 | 64.2 |
| | Master's degree or PHD | 105 | 25.6 |
| Occupation | Student | 243 | 59.3 |
| | Private organization staff | 89 | 21.7 |
| | Medical staff | 19 | 4.6 |
| | Others | 59 | 14.4 |
| Income (CNY) | < 5,000 | 272 | 66.3 |
| | 5,000–10,000 | 90 | 22.0 |
| | >10,000 | 48 | 11.7 |
| Health status | Poor | 22 | 5.3 |
| | General | 152 | 37.1 |
| | Good | 236 | 57.6 |
## 3.3. Statistical analysis
We used the structural equation model (SEM) to test our theoretical model. SEM is a multivariate statistical technique for testing hypotheses about the influences of sets of variables on other variables [54]. It can measure the interrelation of latent variables that are not directly observable and is widely used in social sciences. Latent variables are measured by their corresponding observation variables, namely, scale items. In this research, latent variables include information needs (IN), information seeking (IS), heuristic information processing (HIP), systematic information processing (SIP), risk perception (RP), benefit perception (BP) and vaccination intention (VI). Because there are many latent variables and their relationship is complex, SEM is selected for verification. SEM incorporates two analytical procedures [55]. Firstly, we conducted a confirmatory factor analysis (CFA), which evaluates the measurement component of a theoretical model. After, we carried out a path analysis, which evaluates the relationship between latent variables.
## 4.1. Measurement model testing
We used SPSS 26.0 and AMOS 23.0 of IBM company (Chicagao, America) to analyze the reliability and validity of the measurement model. The Cronbach Alpha coefficient [56] of the scale was 0.897, and the Cronbach Alpha coefficient of each latent variable was >0.7, indicating that the internal stability and consistent reliability of the questionnaire were good.
To further examine the convergent validity of the questionnaire, a confirmatory factor analysis was conducted by AMOS23.0 to obtain values of the average variance extracted (AVE) and standardized loadings of items. As shown in Table 3, all the standardized loadings of items were >0.6 [57]. Values of composite reliability (CR) were between 0.796 and 0.870, which were higher than 0.70, indicating that the constructs have good convergent validity [54]. The AVEs of all structures were greater than the benchmark value of 0.5, indicating that the overall model is valid [58].
**Table 3**
| Variables | Items | Factor loadings | AVE | CR | Alpha |
| --- | --- | --- | --- | --- | --- |
| IN | IN1 | 0.798 | 0.628 | 0.835 | 0.835 |
| | IN2 | 0.779 | | | |
| | IN3 | 0.801 | | | |
| IS | IS1 | 0.751 | 0.756 | 0.84 | 0.84 |
| | IS2 | 0.785 | | | |
| | IS3 | 0.753 | | | |
| | IS4 | 0.722 | | | |
| HIP | HIP1 | 0.801 | 0.583 | 0.807 | 0.808 |
| | HIP2 | 0.714 | | | |
| | HIP3 | 0.773 | | | |
| SIP | SIP1 | 0.739 | 0.615 | 0.827 | 0.826 |
| | SIP2 | 0.803 | | | |
| | SIP3 | 0.809 | | | |
| RP | RP1 | 0.816 | 0.69 | 0.87 | 0.869 |
| | RP2 | 0.859 | | | |
| | RP3 | 0.817 | | | |
| BP | BP1 | 0.828 | 0.647 | 0.846 | 0.844 |
| | BP2 | 0.753 | | | |
| | BP3 | 0.829 | | | |
| VI | VI1 | 0.839 | 0.569 | 0.796 | 0.801 |
| | VI2 | 0.762 | | | |
| | VI3 | 0.649 | | | |
Then we tested the model's fit indicators by AMOS23.0 [55], which are showed as Table 4: the ratio of Chi-square to the degree of freedom (χ2/df) was 2.879, which was smaller than the desired threshold of 3.0. The values of the comparative fit index (CFI), incremental fit index (IFI), and Tucker-Lewis index (TLI) were 0.927, 0.928, and 0.911, respectively. Moreover, the root mean square error of approximation (RMSEA) value was 0.068, which was lower than 0.08. These figures reveal a good fit between the measurement model and the dataset.
**Table 4**
| Fit indicators | χ2/df | CFI | IFI | TLI | RMSEA |
| --- | --- | --- | --- | --- | --- |
| Recommended value | <3.0 | >0.9 | >0.9 | >0.9 | <0.08 |
| Measured value | 2.879 | 0.927 | 0.928 | 0.911 | 0.068 |
## 4.2. Structural equation model analysis
The results of CFA ensure the reliability of our following analysis. By using AMOS 23.0 to set up the structural model, a path analysis was performed to test the relationships among the constructs in the model framework. The standard path coefficients (β) and p-value can be seen in Table 5. All the hypothesized relationships were supported, except H3a, H3b, and H4a. Information needs had a positive effect on information seeking (H1a: β = 0.66, $p \leq 0.001$) and systematic information processing (H1c: β = 0.42, $p \leq 0.001$). Hence, H1a and H1c were supported. The relationship between information needs and heuristic information processing was the opposite (H1b: β = −0.31, $p \leq 0.001$), so H1b was supported. Information seeking had a positive effect on heuristic information processing (H2a: β = 0.72, $p \leq 0.001$) and systematic information processing (H2b: β = 0.41, $p \leq 0.001$). Thus, H2a and H2c were supported.
**Table 5**
| Hypotheses | Path | Standard path coefficients | Standard errors | p -value | Results |
| --- | --- | --- | --- | --- | --- |
| H1a | IN → IS | 0.66 | 0.07 | <0.001 | Supported |
| H1b | IN → HIP | −0.31 | 0.11 | <0.001 | Supported |
| H1c | IN → SIP | 0.42 | 0.06 | <0.001 | Supported |
| H2a | IS → HIP | 0.72 | 0.1 | <0.001 | Supported |
| H2b | IS → SIP | 0.41 | 0.06 | <0.001 | Supported |
| H3a | HIP → RP | 0.61 | 0.06 | <0.001 | Not Supported |
| H3b | HIP → BP | −0.02 | 0.04 | 0.644 | Not Supported |
| H4a | SIP → RP | −0.06 | 0.08 | 0.217 | Not Supported |
| H4b | SIP → BP | 0.73 | 0.07 | <0.001 | Supported |
| H5 | RP → VI | −0.08 | 0.06 | <0.05 | Supported |
| H6 | BP → VI | 0.84 | 0.04 | <0.001 | Supported |
The results show that heuristic information processing had a positive relationship with risk perception (H3a: β = 0.61, $p \leq 0.001$), since this was contrary to our hypothesis, and it had an insignificant negtive effect on the benefit perception (H3b: β = −0.02, $$p \leq 0.644$$ > 0.05). Thus, H3a and H3b were not supported. Systematic information processing had a positive effect on benefit perception (H4b: β = 0.73, $p \leq 0.001$), so H4b was supported. While systematic information processing to risk perception was insignificant (H4a: β = −0.06, $$p \leq 0.217$$ > 0.05), so H4a was not supported. Finally, risk perception had a negative impact on the vaccination willingness of OHCs users (H5: β = −0.08, $$p \leq 0.046$$ < 0.05), and benefit perception had a significant positive impact on the vaccination willingness of OHCs users (H6: β = 0.84, $p \leq 0.001$), indicating that H5 and H6 were supported. The model of final research results was as seen in Figure 2.
**Figure 2:** *The model of final research results, including path coefficients (β), and explained variances (R2). For example, R2 = 0.44, means that the predictors of information seeking explain 44% of its variance.*
## 5.1. Main findings
Our research found a strong positive correlation between vaccine information needs and information seeking among users of OHCs. The COVID-19 pandemic has severely affected the everyday life of people around the world. Even though COVID-19 sometimes mutates, vaccines are still an effective means of prevention [59], and there is often uncertainty about people's attitudes toward emerging technologies [30, 60]. When it comes to the COVID-19 vaccine, this manifests itself as concerns about the safety and efficacy of the vaccine [61]. To reduce uncertainty, people require accurate and effective information, which leads to further product information seeking and information processing behaviors. Savolainen [62] thinks that information needs are the fundamental factor that motivates people to identify and access information sources and the driver that stimulates them to continuously search for information. Our study adds to the empirical evidence, in which vaccine information seeking behavior was largely explained by information needs (R2 = 0.44). A study by Zhou [63] found that pandemic risk stimulates information needs and thus positively influences information seeking behavior. This point is consistent with the findings of our study.
The relationship between vaccine information needs and information processing styles was also explored in this study. The need for vaccine-related information positively influenced systematic information processing, and negatively influenced heuristic information processing. According to Griffin et al. [ 32], people are more likely to process information systematically when they have a greater desire for information. Hubner and Hovick [64] proved that information insufficiency is positively associated with systematic processing. Our study implies that Griffin's model also applies to users' vaccine-related information processing within the online health community. On the other hand, our work reflects that users prioritize risk information when they look for vaccine-related information. Users need comprehensive information to determine vaccination risks and to make vaccination decisions. In this process, information seeking positively correlated with both heuristic and systematic processing. This result suggests that information seeking behavior further facilitates information processing behavior, which is the same as Zhu et al. [ 39]'s research findings.
The focus of this study was to investigate the relationship between information processing, risk/benefit perception, and vaccination intention. Heuristic information processing negatively influences vaccination intention of OHCs users by positively affecting their risk perception, and systematic information processing positively influences vaccination intention of OHCs users by positively affecting their benefit perception. Fast, intuition-based information processing is more likely to elicit users' perception of risk. This feature may be influenced by a large amount of information on vaccine side effects available on the Internet. When people employ heuristic information processing, they verify information less from multiple sources. In addition, because heuristic information processing makes it easier to make quick decisions, it is more difficult for them to spend time searching for more comprehensive information. On online communication platforms, people tend to spread the side effects of vaccines more often than the positive effects of vaccines [65]. Concerns about the side effects are a barrier to achieving high vaccination rates [66, 67], and the perceived risk of vaccines reduces the willingness of people to receive COVID-19 vaccines.
A comprehensive and systematic approach to information processing helps people to perceive the benefits of vaccines and thus promotes their intention to receive vaccinations, and Jing et al. found that parents were more likely to accept childhood vaccinations when they systematically described and processed information [68]. Jing suggested that systematic information processing could lead parents to adopt an “objective” or “balanced” approach, which is compatible with their perceived benefits of childhood vaccination, thus promoting vaccination [68]. Our findings further establish the relationship between systematic information processing, perceived benefits, and vaccination intentions. Another study concluded that active participation in information behavior helps reduce the public's uncertainty and mitigate risk about COVID-19 pandemic [69]. In our study, users more actively involved in information dissemination were more likely to have comprehensive exposure to vaccine information, adopt systematic information processing, and enhance perceived benefits, thus increasing their vaccination intention.
Our study also found that systematic information processing failed to positively influence perceived risk. In contrast, a previous study showed that systematic processing positively influenced perceived risk and thus protective behavioral intentions [49], which is inconsistent with our findings. The possible reason for this is that the information processing in their study was for the vaccine scandal, whereas the participants in our study were exposed to comprehensive vaccine information. Participants who used more systematic processing may have been more likely to use authoritative, official information, whereas authoritative information in China showed more benefits of vaccination. The failure of heuristic processing to predict benefit perception may also be related to the rapidity of the information processing subject. The way decisions are made are based on surface information cues, and the complexity of internet information.
## 5.2. Implications
In the context of the global COVID-19 pandemic, we combined information behavior theory and the heuristic-systematic information processing model to study the COVID-19 vaccination intention of OHC users. Compared with previous studies, we innovatively explored the relationship between cognitive processing and risk/benefit perceptions from two pathways (heuristic or systematic processing). It also provides new theoretical guidance for the application of HSM. Also our study found that the online health community, an information platform, plays a significant role in disseminating information and improving users' vaccination confidence during the pandemic. It provides a basis in the literature for using the network platform to promote public health.
There are several practical applications of our research to increase people's intention to vaccinate against COVID-19. Firstly, the government administration should strengthen the regulation of information about the COVID-19 pandemic and propagandize knowledge of COVID-19 vaccine among the public. Secondly, managers of online health communities should provide users with systematic and positive clues, such as more informative and effective information (e.g., a list of vaccination places, time or price). They are also recommended to expand the retrieval systems of OHCs so that searching is as easy as possible for users. In addition, both government administration and OHCs managers should enhance the dissemination of benefit-related information. Appropriate risk information is needed but must be truthful, as it can stimulate informative behavior and thus promote vaccination intentions. Finally, users should be more proactive in adopting a systematic approach to information processing and making comprehensive judgments about information, for example, by comparing information from different sources (experts, other users or third-party organizations).
## 6. Conclusion
The study confirmed the adaptability of HSM in the background of online health communities and the COVID-19 pandemic, where information needs and information seeking remain the antecedents of information processing. Differences in how OHCs users process information cause differences in perception, with heuristic information processing leading to risk perception and systematic information processing leading to benefit perception. In contrast, risk perception and benefit perception directly influence OHCs users' willingness to vaccinate against COVID-19. Although the negative effect of risk perception on vaccination intention is small but present, the positive effect of benefit perception on vaccination intention is much more significant.
## 7. Limitations and future work
There are some limitations of this study. First, the study was only on Chinese online health community users. Therefore, our findings lack generalizability. Considering the differences between China and foreign countries in terms of COVID-19 pandemic prevention policies and internet management, the impact of online health communities in different countries can be further studied in the future. Second, our quantification of user information behavior through scales may be subjective. More objective measurement tools (i.e., eye-tracking) or methods (i.e., in-depth interview method) could be used further in future studies. Third, cross-sectional studies are limited to confirming the relationship between variables at a particular time period. As vaccination intentions of OHCs users may vary with the development of the COVID-19 pandemic and policies, longitudinal studies can be adopted in future studies to discover further changes in users' vaccination intentions against COVID-19. Finally, since our target population are people who frequently use online health communities and can agree to participate in our survey, some people will refuse to participate in the survey due to disease privacy concerns. Although our sample size meets the minimum requirements for SEM, it is still relatively small. Future studies can expand the sample size; additionally, studies might investigate populations not using online health communities to obtain more comprehensive conclusions.
## 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 Institutional Review Board of College of Life Sciences, Central South University. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
HL and DH: conceptualization and methodology. HL, LG, CW, and YuG: software, validation, investigation, and formal analysis. YiG and MY: resources and data curation. HL and LG: writing—original draft preparation. HJ, XW, and DH: writing—review and editing. DH: visualization, supervision, project administration, and funding acquisition. 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.
## References
1. Abdulazeem HM, von Groote TC, Jayarajah U, Weerasekara I, Borges do Nascimento IJ, Cacic N. **Novel coronavirus infection (COVID-19) in humans: a scoping review and meta-analysis**. *J Clin Med.* (2020) **9** 941. DOI: 10.3390/jcm9040941
2. 2.World Health Organization. WHO Coronavirus (COVID-19) Dashboard. Geneva: WHO (2022). Available online at: https://covid19.who.int (accessed September 13, 2022).. *WHO Coronavirus (COVID-19) Dashboard* (2022)
3. Rahman IU, Ali N, Ijaz F, Afzal A, Abd-Allah EF. **COVID-19: important considerations for developing and using a vaccine**. *Hum Vacc Immunotherap.* (2021) **17** 414-5. DOI: 10.1080/21645515.2020.1781507
4. Laine C, Cotton D, Moyer DV. **COVID-19 vaccine: promoting vaccine acceptance**. *Ann Intern Med.* (2021) **174** 252-3. DOI: 10.7326/M20-8008
5. Joint COVID-19. **Prevention and control mechanism of the State Council, China**. *Conditional Listing of New Coronavirus Vaccine* (2020)
6. HCA Plan. **The central people's government of the People's Republic of China**. *China's New Coronavirus Vaccine Approved for Marketing! 6 Important News!* (2020)
7. 7.National Health Commission. National Health Commission of the People's Republic of China. Beijing: National Health Commission (2022). Available online at: http://www.nhc.gov.cn/xcs/s3574/202207/f920442a8a4845e7833ca8d7e7cd876e.shtml (accessed October 21, 2022).. *National Health Commission of the People's Republic of China* (2022)
8. Guo QZ, Yao NZ, Zhu WW. **How consumers' perception and information processing affect their acceptance of genetically modified foods in China: a risk communication perspective**. *Food Res Int.* (2020) **137** 109518. DOI: 10.1016/j.foodres.2020.109518
9. Wong LP, Alias H, Wong PF, Lee HY, AbuBakar S. **The use of the health belief model to assess predictors of intent to receive the COVID-19 vaccine and willingness to pay**. *Hum Vaccin Immunother.* (2020) **16** 2204-14. DOI: 10.1080/21645515.2020.1790279
10. Yu Y, Lau JTF, She R, Chen X, Li L, Li L. **Prevalence and associated factors of intention of COVID-19 vaccination among healthcare workers in China: application of the health belief model**. *Hum Vaccin Immunother.* (2021) **17** 2894-902. DOI: 10.1080/21645515.2021.1909327
11. Shmueli L. **Predicting intention to receive COVID-19 vaccine among the general population using the health belief model and the theory of planned behavior model**. *BMC Public Health.* (2021) **21** 804. DOI: 10.1186/s12889-021-10816-7
12. Nusair MB, Arabyat R, Khasawneh R, Al-Azzam S, Nusir AT, Alhayek MY. **Assessment of the relationship between COVID-19 risk perception and vaccine acceptance: a cross-sectional study in Jordan**. *Hum Vaccin Immunother.* (2022) **18** 2017734. DOI: 10.1080/21645515.2021.2017734
13. Trumbo CW. **Information processing and risk perception: an adaptation of the heuristic-systematic model**. *J Commun.* (2002) **52** 367-82. DOI: 10.1111/j.1460-2466.2002.tb02550.x
14. Kahneman D, Tversky A. **Prospect theory: an analysis of decision under risk**. *Econometrica.* (1979) **47** 263-91
15. Gallè F, Sabella EA, Roma P, Ferracuti S, Da Molin G, Diella G. **Knowledge and lifestyle behaviors related to COVID-19 pandemic in people over 65 years old from Southern Italy**. *Int J Environ Res Public Health.* (2021) **18** 10872. DOI: 10.3390/ijerph182010872
16. Robertson DA, Mohr KS, Barjaková M, Lunn PD. **lack of perceived benefits and a gap in knowledge distinguish the vaccine hesitant from vaccine accepting during the COVID-19 pandemic**. *Psychol Med.* (2021) **31** 1-4. DOI: 10.1017/S0033291721003743
17. Liu S, Yang L, Zhang C, Xiang YT, Liu Z, Hu S. **Online mental health services in China during the COVID-19 outbreak**. *Lancet Psychiatry.* (2020) **7** 17-8. DOI: 10.1016/S2215-0366(20)30077-8
18. Bento AI, Nguyen T, Wing C, Lozano-Rojas F, Ahn YY, Simon K. **Evidence from internet search data shows information-seeking responses to news of local COVID-19 cases**. *Proc Natl Acad Sci U S A.* (2020) **117** 11220-2. DOI: 10.1073/pnas.2005335117
19. Lu Y. **Automatic topic identification of health-related messages in online health community using text classification**. *Springerplus.* (2013) **2** 309. DOI: 10.1186/2193-1801-2-309
20. Jung G, Jang SH. **Understanding the role of ethnic online communities during the COVID-19 pandemic: a case study of Korean immigrant women's information-seeking behaviors**. *Asian J Soc Sci.* (2022) **50** 292-300. DOI: 10.1016/j.ajss.2022.04.001
21. Wang J, Wang L, Xu J, Peng Y. **Information needs mining of COVID-19 in Chinese online health communities**. *Big Data Res.* (2021) **24** 100193. DOI: 10.1016/j.bdr.2021.100193
22. Li Y, Yan X. **How could peers in online health community help improve health behavior**. *Int J Environ Res Public Health.* (2020) **17** 2995. DOI: 10.3390/ijerph17092995
23. Wilson TD. **Information behavior: an interdisciplinary perspective**. *Inf Process Manag.* (1997) **33** 551-72. DOI: 10.1016/S0306-4573(97)00028-9
24. Xiaoli Hu Y, Shaofeng Z. **Information processing in the “not-in-my-backyard” strategy: an empirical study of anti-nuclear behavioral responses**. *Hum Ecol Risk Assess Int J.* (2020) **26** 2266-87. DOI: 10.1080/10807039.2019.1672138
25. Diwanji V, Reed A, Ferchaud A, Seibert J, Weinbrecht V. **Don't just watch, join in: exploring information behavior and copresence on Twitch**. *Comput Hum Behav.* (2020) **105** 106221. DOI: 10.1016/j.chb.2019.106221
26. Chaiken S. **Heuristic vs. systematic information processing and the use of source vs message cues in persuasion**. *J Person Soc Psychol.* (1980) **39** 752-66. DOI: 10.1037/0022-3514.39.5.752
27. Kim J, Paek HJ. **Information processing of genetically modified food messages under different motives: an adaptation of the multiple-motive heuristic-systematic model**. *Risk Anal.* (2009) **29** 1793-806. DOI: 10.1111/j.1539-6924.2009.01324.x
28. Moore N. **A model of social information need**. *J Inf Sci.* (2002) **28** 297-303. DOI: 10.1177/016555150202800404
29. Lin Y, Hu Z, Zhao Q, Alias H, Danaee M, Wong LP. **Understanding COVID-19 vaccine demand and hesitancy: a nationwide online survey in China**. *PLoS Negl Trop Dis.* (2020) **14** e0008961. DOI: 10.1371/journal.pntd.0008961
30. Zeng J, Wei JC, Zhao DT, Zhu WW, Gu JB. **Information-seeking intentions of residents regarding the risks of nuclear power plant: an empirical study in China**. *Nat Hazards.* (2017) **87** 739-55. DOI: 10.1007/s11069-017-2790-x
31. Wilson TD. **Human information behavior**. *Inform Sci Int J Emerg Transdiscip.* (2000) **3** 49-55. DOI: 10.1002/sdr.4260070210
32. Griffin RJ, Dunwoody S, Neuwirth K. **Proposed model of the relationship of risk information seeking and processing to the development of preventive behaviors**. *Environ Res.* (1999) **80** 230-45. DOI: 10.1006/enrs.1998.3940
33. Johnson BB. **Testing and expanding a model of cognitive processing of risk information**. *Risk Anal.* (2005) **25** 631-50. DOI: 10.1111/j.1539-6924.2005.00609.x
34. Kuhlthau CC. **Inside the search process: information seeking from the user's perspective**. *J Am Soc Inform Sci.* (1991) **42** 361-71. DOI: 10.1002/(SICI)1097-4571(199106)42:5<361::AID-ASI6>3.0.CO;2-#
35. Ter Huurne E, Gutteling J. **Information needs and risk perception as predictors of risk information seeking**. *J Risk Res.* (2008) **11** 847-62. DOI: 10.1080/13669870701875750
36. Che D, Hu DH. **Developing the information-seeking behavior scale for undergraduates**. *J Data Inform Sci.* (2013) **6** 78-96
37. Lu C, Xu W, Shen H, Zhu J, Wang KZ. **MIMO channel information feedback using deep recurrent network**. *IEEE Commun Lett.* (2019) **23** 188-91. DOI: 10.1109/LCOMM.2018.2882829
38. Kahlor LA, Yang ZJ, Liang MC. **Risky politics: applying the planned risk information seeking model to the 2016 US presidential election**. *Mass Commun Soc.* (2018) **21** 697-719. DOI: 10.1080/15205436.2018.1498900
39. Zhu WW, Wu TT, Liao CH. **Impact of information processing on individuals' intentions toward reducing PM2.5: evidence from Hefei City, China**. *J Environ Plan Manag.* (2022) **22** 1-18. DOI: 10.1080/09640568.2022.2036601
40. Smith SW, Hitt R, Russell J, Nazione S, Silk K, Atkin CK. **Risk belief and attitude formation from translated scientific messages about PFOA, an environmental risk associated with breast cancer**. *Health Commun.* (2017) **32** 279-87. DOI: 10.1080/10410236.2016.1138350
41. Chaiken S, Liberman A, Eagly AH, Bargh JS, and Uleman JA. **Heuristic and systematic information processing within and beyond the persuasion context**. *Unintended Thought* (1989) 212-52
42. Smerecnik CM, Mesters I, Candel MJ, De Vries H, De Vries NK. **Risk perception and information processing: the development and validation of a questionnaire to assess self-reported information processing**. *Risk Anal.* (2012) **32** 54-66. DOI: 10.1111/j.1539-6924.2011.01651.x
43. Tortosa-Edo V, Lopez-Navarro MA, Llorens-Monzonis J, Rodriguez-Artola RM. **The antecedent role of personal environmental values in the relationships among trust in companies, information processing and risk perception**. *J Risk Res.* (2014) **17** 1019-35. DOI: 10.1080/13669877.2013.841726
44. Kim J, Yeo SK, Brossard D, Scheufele DA, Xenos MA. **Disentangling the influence of value predispositions and risk/benefit perceptions on support for nanotechnology among the American public**. *Risk Anal.* (2014) **34** 965-80. DOI: 10.1111/risa.12141
45. Yigit M, Ozkaya-Parlakay A, Senel E. **Evaluation of COVID-19 vaccine refusal in parents**. *Pediatr Infect Dis J.* (2021) **40** 134-6. DOI: 10.1097/INF.0000000000003042
46. Kelly BJ, Rupert DJ, Aikin KJ, Sullivan HW, Johnson M, Bann CM. **Development and validation of prescription drug risk, efficacy, and benefit perception measures in the context of direct-to-consumer prescription drug advertising**. *Res Soc Adm Pharm.* (2021) **17** 942-55. DOI: 10.1016/j.sapharm.2020.07.028
47. Lazarus JV, Ratzan SC, Palayew A, Gostin LO, Larson HJ, Rabin K. **A global survey of potential acceptance of a COVID-19 vaccine**. *Nat Med.* (2021) **27** 225-8. DOI: 10.1038/s41591-020-1124-9
48. Luxi N, Giovanazzi A, Capuano A, Crisafulli S, Cutroneo PM, Fantini MP. **COVID-19 vaccination in pregnancy, paediatrics, immunocompromised patients, and persons with history of allergy or prior SARS-CoV-2 infection: overview of current recommendations and pre- and post-marketing evidence for vaccine efficacy and safety**. *Drug Saf.* (2021) **44** 1247-69. DOI: 10.1007/s40264-021-01131-6
49. Yan J, Ouyang Z, Vinnikova A, Chen MX. **Avoidance of the threats of defective vaccines: how a vaccine scandal influences parents' protective behavioral response**. *Health Commun.* (2021) **36** 962-71. DOI: 10.1080/10410236.2020.1724638
50. Yan E, Lai DWL, Lee VWP. **Predictors of intention to vaccinate against COVID-19 in the general public in Hong Kong: findings from a population-based, cross-sectional survey**. *Vaccines.* (2021) **9** 696. DOI: 10.3390/vaccines9070696
51. Costa-Font M, Gil JM. **Structural equation modelling of consumer acceptance of genetically modified (GM) food in the Mediterranean Europe: a cross country study**. *Food Qual Prefer.* (2009) **20** 399-409. DOI: 10.1016/j.foodqual.2009.02.011
52. Cheng P, Ouyang Z, Liu Y. **The effect of information overload on the intention of consumers to adopt electric vehicles**. *Transportation.* (2020) **47** 2067-86. DOI: 10.1007/s11116-019-10001-1
53. Anthoine E, Moret L, Regnault A, Sébille V, Hardouin JB. **Sample size used to validate a scale: a review of publications on newly-developed patient reported outcomes measures**. *Health Qual Life Outcomes.* (2014) **12** 176. DOI: 10.1186/s12955-014-0176-2
54. Gefen DS DW, Boudreau MC. **Structural equation modeling and regression: guidelines for research practice**. *Commun Assoc Inform Syst.* (1978) **4** 1-70. DOI: 10.17705/1CAIS.00407
55. McVeigh J, MacLachlan M, Vallieres F, Hyland P, Stilz R, Cox H. **Identifying predictors of stress and job satisfaction in a sample of merchant seafarers using structural equation modeling**. *Front Psychol.* (2019) **10** 70. DOI: 10.3389/fpsyg.2019.00070
56. Zhang YY, Liu CY, Luo SM, Xie YT, Liu F, Li X. **Factors influencing patients' intentions to use diabetes management apps based on an extended unified theory of acceptance and use of technology model: web-based survey**. *J Med Internet Res.* (2019) **21** e15023. DOI: 10.2196/15023
57. Amini A, Alimohammadlou M. **Toward equation structural modeling: an integration of interpretive structural modeling and structural equation modeling**. *J Manag Anal.* (2020) **8** 693-714. DOI: 10.1080/23270012.2021.1881927
58. Hoque R, Sorwar G. **Understanding factors influencing the adoption of mHealth by the elderly: an extension of the UTAUT model**. *Int J Med Inform.* (2017) **101** 75-84. DOI: 10.1016/j.ijmedinf.2017.02.002
59. Fernandes Q, Inchakalody VP, Merhi M, Mestiri S, Taib N, El-Ella DMA. **Emerging COVID-19 variants and their impact on SARS-CoV-2 diagnosis, therapeutics and vaccines**. *Ann Med.* (2022) **54** 524-40. DOI: 10.1080/07853890.2022.2031274
60. Bearth A, Siegrist M. **Are risk or benefit perceptions more important for public acceptance of innovative food technologies: a meta-analysis**. *Trends Food Sci Technol.* (2016) **49** 14-23. DOI: 10.1016/j.tifs.2016.01.003
61. Kaplan RM, Milstein A. **Influence of a COVID-19 vaccine's effectiveness and safety profile on vaccination acceptance**. *Proc Natl Acad Sci U S A.* (2021) **118** e2021726118. DOI: 10.1073/pnas.2021726118
62. Savolainen R. **Information need as trigger and driver of information seeking: a conceptual analysis**. *Aslib J Inf Manag.* (2017) **69** 2-21. DOI: 10.1108/AJIM-08-2016-0139
63. Zhou SH. **Impact of perceived risk on epidemic information seeking during the outbreak of COVID-19 in China**. *J Risk Res.* (2021) **24** 477-91. DOI: 10.1080/13669877.2021.1907609
64. Hubner AY, Hovick SR. **Understanding risk information seeking and processing during an infectious disease outbreak: the case of Zika virus**. *Risk Anal.* (2020) **40** 1212-25. DOI: 10.1111/risa.13456
65. Okuhara T, Ishikawa H, Okada M, Kato M, Kiuchi T. **Contents of Japanese pro- and anti-HPV vaccination websites: a text mining analysis**. *Patient Educ Couns.* (2018) **101** 406-13. DOI: 10.1016/j.pec.2017.09.014
66. Shoots-Reinhard B, Lawrence ER, Schulkin J, Peters E. **Excluding numeric side-effect information produces lower vaccine intentions**. *Vaccine.* (2022) **40** 4262-9. DOI: 10.1016/j.vaccine.2022.06.001
67. Holzmann-Littig C, Frank T, Schmaderer C, Braunisch MC, Renders L, Kranke P. **COVID-19 vaccines: fear of side effects among German health care workers**. *Vaccines.* (2022) **10** 689. DOI: 10.3390/vaccines10050689
68. Jing Y, Wei JC, Zhe OY, Vinnikova A, Zhao DT, Zhang HB. **The influence of parents' information processing on childhood vaccine acceptance after a vaccine crisis in China**. *Health Risk Soc.* (2019) **21** 284-303. DOI: 10.1080/13698575.2019.1619672
69. Jin XL, Lane D. **To know or not to know? Exploring COVID-19 information seeking with the risk information seeking and processing model**. *J Inform Sci.* (2022) **48** 25325. DOI: 10.1177/01655515221125325
|
---
title: 'Delay in seeking health care from community residents during a time with low
prevalence of COVID-19: A cross-sectional national survey in China'
authors:
- Ziyu Wang
- Yurong Tang
- Yu Cui
- Hanwen Guan
- Xiaoqian Cui
- Yuan Liu
- Yanni Liu
- Zheng Kang
- Qunhong Wu
- Yanhua Hao
- Chaojie Liu
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9989024
doi: 10.3389/fpubh.2023.1100715
license: CC BY 4.0
---
# Delay in seeking health care from community residents during a time with low prevalence of COVID-19: A cross-sectional national survey in China
## Abstract
### Background
The pandemic of COVID-19 has significant implications on health resources allocation and health care delivery. Patients with non-COVID illness may have to change their care seeking behaviors to mitigate the risk of infections. The research aimed to investigate potential delay of community residents in seeking health care at a time with an overall low prevalence of COVID-19 in China.
### Methods
An online survey was conducted in March 2021 on a random sample drawn from the registered survey participants of the survey platform Wenjuanxing. The respondents who reported a need for health care over the past month ($$n = 1$$,317) were asked to report their health care experiences and concerns. Logistic regression models were established to identify predictors of the delay in seeking health care. The selection of independent variables was guided by the Andersen's service utilization model. All data analyses were performed using SPSS 23.0. A two-sided p value of <0.05 was considered as statistically significant.
### Key results
About $31.4\%$ of respondents reported delay in seeking health care, with fear of infection ($53.5\%$) as a top reason. Middle (31–59 years) age (AOR = 1.535; $95\%$ CI, 1.132 to 2.246), lower levels of perceived controllability of COVID-19 (AOR = 1.591; $95\%$ CI 1.187 to 2.131), living with chronic conditions (AOR = 2.008; $95\%$ CI 1.544 to 2.611), pregnancy or co-habiting with a pregnant woman (AOR = 2.115; $95\%$ CI 1.154 to 3.874), access to Internet-based medical care (AOR = 2.529; $95\%$ CI 1.960 to 3.265), and higher risk level of the region (AOR = 1.736; $95\%$ CI 1.307 to 2.334) were significant predictors of the delay in seeking health care after adjustment for variations of other variables. Medical consultations ($38.7\%$), emergency treatment ($18.2\%$), and obtainment of medicines ($16.5\%$) were the top three types of delayed care, while eye, nose, and throat diseases ($23.2\%$) and cardiovascular and cerebrovascular diseases ($20.8\%$) were the top two conditions relating to the delayed care. Self-treatment at home was the most likely coping strategy ($34.9\%$), followed by Internet-based medical care ($29.2\%$) and family/friend help ($24.0\%$).
### Conclusions
Delay in seeking health care remained at a relatively high level when the number of new COVID-19 cases was low, which may present a serious health risk to the patients, in particular those living with chronic conditions who need continuous medical care. Fear of infection is the top reason for the delay. The delay is also associated with access to Internet-based medical care, living in a high risk region, and perceived low controllability of COVID-19.
## Introduction
The World Health Organization (WHO) declared COVID-19 as a global pandemic in March 2020 [1]. As of 1 May 2022, over 500 million confirmed cases of COVID-19 and over six million deaths had been reported worldwide [2]. Along with the direct health threats of COVID-19, there have been disruptions to health services [3]. History shows that the Ebola outbreak in 2014–2015 created serious interruptions on the availability, uptake, and demand of health care services in Sierra Leone [4].
COVID-19 has put health care services under serious stress all over the world. China had adopted a “dynamic zero-COVID” policy prior to December 2022, which required a quick response from local governments to cut off the chain of community transmission through imposing restrictions and mobilizing available health resources once a new COVID-19 case was identified [5, 6]. Community residents could experience additional barriers in seeking health care [7]. This has raised serious concerns about the delay or avoidance of health care [8, 9].
Delay or avoidance of medical assessment [10], treatment of bacteremia [11], thrombolysis for stroke [12], and treatment of botulinum toxin [13] has been reported during the outbreak of COVID-19 in various countries. Meanwhile, many people missed the opportunity of early detection of new conditions and failed to manage their existing chronic conditions properly (14–16). In Japan, $5.6\%$ of patients living with chronic conditions reported worse health [17]. The state of Victoria in Australia witnessed significant decline in patient visits to hospital emergency departments and the diagnosis of five common cancers dropped by approximately one third, prompting urgent calls for the public to seek timely medical attention [18]. Similarly, reduced screening, referrals and presentations for lung and colorectal cancers in the UK also led to a projection of 4.8 and $16.5\%$ increased deaths from the two cancers, respectively [19, 20]. A study suggests that efforts to reduce COVID-19-related care avoidance are warranted even in regions with low COVID-19 prevalence [21].
Delay/avoidance of health care can be caused by patient choice and/or as a result of fear of infection and process delays (including disruptions of supply chain) [22, 23]. In May 2020, the WHO conducted a global assessment of health services, which showed that service provision had been damagingly impacted by COVID-19 (24–26). In some countries, elective surgeries were suspended to mobilize resources to fight COVID-19 [27]. However, most existing studies have attributed delay/avoidance of care during COVID-19 to fear of infection. Meanwhile, fear of losing job, being separated from friends, and falling into financial difficulties have also been acknowledged as the underlined reasons of avoidance of seeking testing for COVID-19 [28].
There is a paucity in the literature documenting the effect of COVID-19 on health care seeking behaviors of consumers. The current study aimed to investigate potential delay of community residents in seeking health care at a time with an overall low prevalence of COVID-19 in China.
## Study setting and participants
A cross-sectional online survey was conducted in mainland China. The study protocol was approved by the Research Ethics Committee of Harbin Medical University (IRB code HMUIRB20200004). The survey was anonymous. Participant information sheet was provided and implied informed consent was required from each participant prior to proceeding to the survey.
Study participants were recruited through the online survey platform Wenjuanxing (www.wjx.cn). It has reach to the largest pool of survey participants in mainland China, covering all regions: more than one million questionnaire responses are recorded by Wenjuanxing every day. Eligible participants in this study were the adults over 18 years of age. They were identified randomly through an automation process embedded in the Wenjuanxing sampling services. The identity of the invited participants remained anonymous and unknown to the research team.
## Data collection and retention
Data were collected in March 2021. Only one submission per IP address was allowed. The survey closed when returned questionnaires reached 4,383. To ensure quality of the data, the returned questionnaires containing logical errors (contradictory answers) and those that were completed within 10 min were excluded. Our pilot test showed that at least 10 min would be needed to read through the questionnaire. This resulted in a final sample size of 4,325 ($98.75\%$ of the returned questionnaires). Of those, 1,317 ($30.45\%$) reported health problems and intention to seek medical care.
## Study measurements
The questionnaire development was informed by the existing literature [29] and was adapted to the specific context of COVID-19 in line with the relevant guidelines issued by the WHO [30], the National Health Commission and the national CDC in China [31].
Delay in seeking health care was the main interest of the study. Participants were asked to report their self-assessment of health and intention to seek health care over the past month, when the seven-day rolling average of daily new confirmed COVID-19 cases ranged from 7.57 to 56.71 [32]. For those who intended to seek health care, their experiences in obtaining the needed care were further investigated through a series of questions, which included whether they ‘delayed care due to concerns related to COVID-19' (1 = “yes”; 0 = “no/unsure”), for what condition (cardiovascular and/or cerebrovascular diseases, digestive diseases, bone diseases; respiratory disease, eye nose throat diseases, diabetes mellitus, tumor, accident and injury, and others), through what service (medical consultation, emergency treatment (care for immediate life-threatening conditions), obtainment of medicines, follow-up examination, hospital admission, surgical procedure, and others), and from which provider (local provincial/municipal public hospitals, primary healthcare network, cross-provincial/municipal public hospitals, private clinics or private hospitals). In addition, they were asked to identify one or more reasons for the delay, if applied, from the following list: fear of infection; discouragement from relatives and friends; difficulties with online appointments; long waiting time in facilities; complex service procedure; facility closure; denied access to facilities; transfer to infection/fever clinics; movement restrictions; community lockdown; complex referral procedure; and others.
The study participants who reported delay in health care were also asked to identify one or more consequences they anticipated (including disrupted medication, slow recovery, complications, missed optimal timing of treatment, deterioration of illness conditions, dissatisfaction with care provision, increased costs, increased mental burden on family, and others) and how they coped with the delay (including use of Internet-based medical services, self-treatment at home, family/friend help, telemedicine, government assistance, and others).
The selection of independent variables was guided by the Andersen's service utilization model [29], which categorizes predictors of health service utilization into predisposing, enabling, and need factors. In this study, gender (male vs. female), age (≤ 30, 31–59, ≥60 years), and marital status (married vs. others) were deemed as predisposing factor, while residency (urban vs. rural), risk level of regions (high vs low), educational attainment (with or without a university degree), personal income (≥average [5,000] vs. < 5,000 Yuan per month [33]), health insurance coverage (yes vs. no), pregnancy or co-habitant with a pregnant woman (yes vs. no), and using Internet medical services (yes vs. no) measured enabling factor. In China, more than 50 accumulative active cases of COVID-19 in a municipality over a period of 14 days would be classified as high risk [7].
Need factor was measured by COVID-19 risk perception and chronic conditions (yes vs. no). The measurement of risk perception followed the definition of Bauer from Harvard [34], considering people's cognition, feeling, and comprehension of the risk characteristics, not the actual risk [35]. Adams and Smith [36] pointed out that individual risk perception is closely associated the severity and controllability of the risk. Risk perception is a pivotal determinant of care seeking behaviors [37]. In this study, a three-dimensional scale was adopted to measure COVID-19 risk perception, covering perceived susceptibility to COVID-19 infection, perceived severity of the consequences of COVID-19, and perceived controllability of COVID-19 outbreaks. Each dimension contains three items rated on a six-point Likert scale ranging from 1 (strongly disagree) to 6 (strongly agree). A summed average score was calculated for each dimension, with 1–3 indicating a low level and 4–6 indicating a high level of risk perception. The risk perception scale has been validated in a previous study [38]. Good internal consistency (Cronbach's α = 0.824) and construct validity (GFI = 0.982, AGFI = 0.961, IFI = 0.972, CFI = 0.972, RMSEA = 0.062 in confirmatory factor analysis) of the scale were also evident in this study. Chronic conditions were defined as a general term for the diagnosed diseases with an insidious onset and prolonged course [39], which include cardiovascular diseases, cerebrovascular diseases, diabetes, and others [40].
## Data analysis
The percentage distributions of the study participants with different characteristics were described and compared between those living in high and low risk regions using Chi-square tests. A multivariate logistic regression model was then established to determine the significant predictors of delay in seeking health care after adjustment for variations in other variables. The reasons of delay and perceived consequences were described and ranked in order using percentage distributions.
All data analyses were performed using SPSS 23.0. A two-sided p value of < 0.05 was considered as statistically significant.
## Participant characteristics
Of the 1,317 study participants who reported a need to seek health care, 832 ($63.2\%$) lived in high risk regions over the study period. Over half were female ($54.1\%$), married ($54.5\%$), obtained a university qualification ($52.2\%$), and had no chronic conditions ($55.4\%$) at the time of the survey. The vast majority were younger than 60 years ($95.2\%$), resided in an urban area ($67.1\%$), earned < 5,000 Yuan per month ($61.7\%$), had health insurance coverage ($93.5\%$), and were not pregnant or living with a pregnant woman ($95.9\%$). Although $81.9\%$ of respondents perceived high severity of COVID-19, $70.1\%$ perceived high levels of controllability, and $87.9\%$ perceived low levels of susceptibility. Compared with the respondents from a region with low mobility restrictions, those experiencing high mobility restrictions were older ($p \leq 0.001$), and were more likely to be married ($p \leq 0.001$), obtained no university qualifications ($$p \leq 0.01$$), earned a low level of income ($$p \leq 0.012$$), had no health insurance coverage ($p \leq 0.001$), lived with chronic conditions ($$p \leq 0.01$$), and perceived higher levels of susceptibility ($p \leq 0.001$) and lower levels of controllability ($$p \leq 0.012$$) (Table 1).
**Table 1**
| Characteristics | Total | Total.1 | High-risk regions (n = 832) | High-risk regions (n = 832).1 | Low-risk regions (n = 485) | Low-risk regions (n = 485).1 | p |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | n | % | n | % | n | % | |
| Gender | Gender | Gender | Gender | Gender | Gender | Gender | 0.347 |
| Male | 605 | 45.9 | 374 | 45.0 | 231 | 47.6 | |
| Female | 712 | 54.1 | 458 | 55.1 | 254 | 52.4 | |
| Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | < 0.001 |
| ≤ 30 | 593 | 45 | 301 | 36.2 | 292 | 60.2 | |
| 31–59 | 661 | 50.2 | 471 | 56.6 | 190 | 39.2 | |
| ≥60 | 63 | 4.8 | 60 | 7.2 | 3 | 0.6 | |
| Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | < 0.001 |
| Married | 718 | 54.5 | 491 | 59.0 | 227 | 46.8 | |
| Others | 599 | 45.5 | 341 | 41.0 | 258 | 53.2 | |
| Residency | Residency | Residency | Residency | Residency | Residency | Residency | 0.304 |
| Urban | 884 | 67.1 | 550 | 66.1 | 334 | 68.9 | |
| Rural | 433 | 32.9 | 282 | 33.9 | 151 | 31.1 | |
| Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | 0.010 |
| University | 688 | 52.2 | 412 | 49.5 | 276 | 56.9 | |
| Without university | 629 | 47.8 | 420 | 50.5 | 209 | 43.1 | |
| Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | 0.012 |
| High (≥5,000) | 504 | 38.3 | 297 | 35.7 | 207 | 42.7 | |
| Low (< 5,000) | 813 | 61.7 | 535 | 64.3 | 278 | 57.3 | |
| Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | < 0.001 |
| Yes | 1231 | 93.5 | 760 | 91.4 | 471 | 97.1 | |
| No | 86 | 6.5 | 72 | 8.7 | 14 | 2.9 | |
| Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | 0.263 |
| Yes | 54 | 4.1 | 38 | 4.6 | 16 | 3.3 | |
| No | 1263 | 95.9 | 794 | 95.4 | 469 | 96.7 | |
| Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | 0.072 |
| Yes | 531 | 40.3 | 320 | 38.5 | 211 | 43.5 | |
| No | 786 | 59.7 | 512 | 61.5 | 274 | 56.5 | |
| Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | 0.010 |
| Yes | 588 | 44.6 | 394 | 47.4 | 194 | 40.0 | |
| No | 729 | 55.4 | 438 | 52.6 | 291 | 60.0 | |
| Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | 0.958 |
| High | 1079 | 81.9 | 682 | 82.0 | 397 | 81.9 | |
| Low | 238 | 18.1 | 150 | 18.0 | 88 | 18.1 | |
| Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | 0.012 |
| High | 923 | 70.1 | 563 | 67.7 | 360 | 74.2 | |
| Low | 394 | 29.9 | 269 | 32.3 | 125 | 25.8 | |
| Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | < 0.001 |
| High | 160 | 12.1 | 125 | 15.0 | 35 | 7.2 | |
| Low | 1157 | 87.9 | 707 | 85.0 | 450 | 92.78 | |
## Delay in seeking healthcare services
Overall, $31.4\%$ of the study participants who had a care need experienced delay in seeking health care: $35.8\%$ in those residing in a high risk region compared with $23.7\%$ in those from low risk regions ($p \leq 0.001$). However, living in high risk regions (AOR = 1.736 [$95\%$ CI 1.307–2.334]) was not the only predictor of delay in seeking health care. An age between 31 and 59 years (AOR = 1.535 [$95\%$ CI 1.132–2.246]), lower levels of perceived controllability (AOR = 1.591 [$95\%$ CI 1.187–2.131]), living with chronic conditions (AOR = 2.008 [$95\%$ CI 1.544–2.611]), pregnancy or co-habitant with a pregnant woman (AOR = 2.115 [$95\%$ CI 1.154–3.874]), and access to Internet-based medical services (AOR = 2.529 [$95\%$ CI 1.960–3.265]) were also associated with delay in seeking health care according to the results of the multivariate modeling. The associations of delay in seeking health care with middle age, marriage, urban residency, higher personal income, and absence of health insurance coverage became statistically insignificant after adjustment for variations of other variables. The multivariate model explained $19.2\%$ of variance (R2) in delay of seeking health care (Table 2).
**Table 2**
| Unnamed: 0 | Unnamed: 1 | Delay in seeking health care (n = 413) | Delay in seeking health care (n = 413).1 | Delay in seeking health care (n = 413).2 | Delay in seeking health care (n = 413).3 | Delay in seeking health care (n = 413).4 | Delay in seeking health care (n = 413).5 | Delay in seeking health care (n = 413).6 | Delay in seeking health care (n = 413).7 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Predictor | Respondents reporting a need to seek health care | N | % | Unadjusted OR | p | Adjusted OR | 95% Confidence interval | 95% Confidence interval | p |
| Predisposing factor | Predisposing factor | Predisposing factor | Predisposing factor | Predisposing factor | Predisposing factor | Predisposing factor | Predisposing factor | Predisposing factor | Predisposing factor |
| Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender |
| Female (Reference) | 712 | 237 | 33.3 | | | | | | |
| Male | 605 | 176 | 29.1 | 0.822 | 0.102 | 0.874 | 0.677 | 1.129 | 0.303 |
| Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) |
| ≤ 30 (Reference) | 593 | 137 | 23.1 | | | | | | |
| 31–59 | 661 | 252 | 38.1 | 2.051 | 0.000 | 1.535 | 1.132 | 2.246 | 0.007 |
| ≥60 | 63 | 24 | 38.1 | 2.048 | 0.010 | 1.268 | 0.961 | 1.673 | 0.093 |
| Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status |
| Others (Reference) | 599 | 149 | 24.9 | | | | | | |
| Married | 718 | 264 | 36.8 | 1.756 | < 0.001 | 1.232 | 0.907 | 1.672 | 0.182 |
| Enabling factor | Enabling factor | Enabling factor | Enabling factor | Enabling factor | Enabling factor | Enabling factor | Enabling factor | Enabling factor | Enabling factor |
| Residency | Residency | Residency | Residency | Residency | Residency | Residency | Residency | Residency | Residency |
| Rural (Reference) | 433 | 120 | 27.7 | | | | | | |
| Urban | 884 | 293 | 33.1 | 1.293 | 0.046 | 1.194 | 0.897 | 1.589 | 0.225 |
| Risk regions | Risk regions | Risk regions | Risk regions | Risk regions | Risk regions | Risk regions | Risk regions | Risk regions | Risk regions |
| Low (Reference) | 485 | 115 | 23.7 | | | | | | |
| High | 832 | 298 | 35.8 | 1.795 | < 0.001 | 1.736 | 1.307 | 2.334 | < 0.001 |
| Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment | Educational attainment |
| Without university degree (Reference) | 629 | 189 | 30.1 | | | | | | |
| With university degree | 688 | 224 | 32.6 | 1.124 | 0.327 | 1.308 | 0.986 | 1.737 | 0.063 |
| Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) | Monthly personal income (Yuan) |
| Low (< 5,000) (Reference) | 813 | 237 | 29.2 | | | | | | |
| High (≥5,000) | 504 | 176 | 34.9 | 1.304 | 0.028 | 1.055 | 0.794 | 1.401 | 0.713 |
| Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage |
| Yes (Reference) | 1231 | 376 | 30.5 | | | | | | |
| No | 86 | 37 | 43.0 | 1.718 | 0.017 | 0.622 | 0.378 | 1.025 | 0.062 |
| Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman | Living with a pregnant woman |
| No (Reference) | 1263 | 383 | 30.3 | | | | | | |
| Yes | 54 | 30 | 55.6 | 2.872 | < 0.001 | 2.115 | 1.154 | 3.874 | 0.015 |
| Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service | Using Internet medical service |
| No (Reference) | 786 | 179 | 22.8 | | | | | | |
| Yes | 531 | 234 | 44.1 | 2.672 | < 0.001 | 2.529 | 1.960 | 3.265 | < 0.001 |
| Need factor | Need factor | Need factor | Need factor | Need factor | Need factor | Need factor | Need factor | Need factor | Need factor |
| Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease | Chronic disease |
| No (Reference) | 729 | 171 | 23.5 | | | | | | |
| Yes | 588 | 242 | 41.2 | 2.282 | < 0.001 | 2.008 | 1.544 | 2.611 | < 0.001 |
| Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 | Perceived severity of COVID-19 |
| Low (Reference) | 238 | 75 | 31.5 | | | | | | |
| High | 1079 | 338 | 31.3 | 1.009 | 0.955 | 0.565 | 0.279 | 1.145 | 1.113 |
| Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 | Perceived controllability of COVID-19 |
| High (Reference) | 923 | 310 | 33.6 | | | | | | |
| Low | 394 | 103 | 26.1 | 1.429 | 0.008 | 1.591 | 1.187 | 2.131 | 0.002 |
| Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 | Perceived susceptibility of COVID-19 |
| Low (Reference) | 1157 | 354 | 30.6 | | | | | | |
| High | 160 | 59 | 36.9 | 1.325 | 0.109 | 1.029 | 0.750 | 1.412 | 0.113 |
Table 3 shows that the types of care being delayed appeared to be consistent between those from the high and low risk regions, despite some differences in the percentage distributions. Medical consultations ($38.7\%$), emergency treatment ($18.2\%$), and obtainment of medicines ($16.5\%$) were the top three types of delayed care. Eye, nose, and throat diseases ($23.2\%$) and cardiovascular and cerebrovascular diseases ($20.8\%$) were the top two conditions relating to the delayed care, followed by digestive diseases ($10.1\%$) in those living in a region with high risk and bone diseases ($12.2\%$) and respiratory diseases ($12.2\%$) in those living in a region with low risk. Most of the delayed services were planned to be obtained from local provincial/municipal hospitals ($68.0\%$) and primary health care networks ($17.2\%$). Self-treatment at home was the most likely coping strategy ($34.9\%$), followed by Internet-based medical care ($29.2\%$) and family/friend help ($24.0\%$) (Table 3).
**Table 3**
| Delayed care | Total | Total.1 | High-risk regions (n = 298) | High-risk regions (n = 298).1 | Low-risk regions (n = 115) | Low-risk regions (n = 115).1 |
| --- | --- | --- | --- | --- | --- | --- |
| | n | % | n | % | n | % |
| Purpose of intended visit | Purpose of intended visit | Purpose of intended visit | Purpose of intended visit | Purpose of intended visit | Purpose of intended visit | Purpose of intended visit |
| Medical consultation | 160 | 38.7 | 117 | 39.3 | 33 | 28.7 |
| Emergency treatment | 75 | 18.2 | 52 | 17.5 | 23 | 20.0 |
| Obtainment of medicines | 68 | 16.5 | 43 | 14.4 | 25 | 21.7 |
| Follow-up examination | 50 | 12.1 | 40 | 13.4 | 10 | 8.7 |
| Hospital admission | 29 | 7.0 | 23 | 7.7 | 6 | 5.2 |
| Surgical procedure | 21 | 5.1 | 13 | 4.4 | 8 | 7.0 |
| Others | 10 | 2.4 | 10 | 3.4 | 10 | 8.6 |
| Illness condition to be treated | Illness condition to be treated | Illness condition to be treated | Illness condition to be treated | Illness condition to be treated | Illness condition to be treated | Illness condition to be treated |
| Eye, nose, throat diseases | 96 | 23.2 | 68 | 22.8 | 28 | 24.4 |
| Cardiovascular and/or cerebrovascular diseases | 86 | 20.8 | 57 | 19.1 | 29 | 25.2 |
| Digestive diseases | 35 | 8.5 | 30 | 10.1 | 5 | 4.4 |
| Bone diseases | 34 | 8.2 | 20 | 6.7 | 14 | 12.2 |
| Diabetes mellitus | 33 | 8.0 | 26 | 8.7 | 7 | 6.1 |
| Respiratory diseases | 30 | 7.3 | 16 | 5.4 | 14 | 12.2 |
| Tumor | 29 | 7.0 | 26 | 8.7 | 3 | 2.6 |
| Accident and injury | 19 | 4.6 | 12 | 4.0 | 7 | 6.1 |
| Others | 51 | 12.3 | 43 | 14.4 | 8 | 7.0 |
| Institution in which care to be sought | Institution in which care to be sought | Institution in which care to be sought | Institution in which care to be sought | Institution in which care to be sought | Institution in which care to be sought | Institution in which care to be sought |
| Local provincial/municipal public hospitals | 281 | 68.0 | 198 | 66.4 | 83 | 72.2 |
| Primary healthcare network | 71 | 17.2 | 52 | 17.5 | 19 | 16.5 |
| Cross-provincial/municipal public hospitals | 38 | 9.2 | 31 | 10.4 | 7 | 6.1 |
| Private clinics or private hospitals | 23 | 5.6 | 17 | 5.7 | 6 | 5.2 |
| Coping strategy | Coping strategy | Coping strategy | Coping strategy | Coping strategy | Coping strategy | Coping strategy |
| Self-treatment at home | 144 | 34.9 | 105 | 35.2 | 40 | 34.8 |
| Internet-based medical services | 121 | 29.2 | 84 | 28.2 | 37 | 32.2 |
| Family/friend help | 99 | 24.0 | 74 | 24.8 | 26 | 22.6 |
| Seeking government support | 28 | 6.8 | 17 | 5.7 | 10 | 8.7 |
| Others | 21 | 5.1 | 18 | 6.0 | 2 | 1.7 |
Figure 1 shows that the top three reasons of delay in seeking health care were fear of infection (52.01−$57.39\%$), complex service procedure ($40.94\%$~$43.48\%$), and long waiting time in facilities (21.74−$22.15\%$) according to the reports of the respondents from both regions with high and low COVID-19 risk. Of the other reasons, those from the regions with high risk were more likely to report facility closure (20.13 vs. $14.78\%$), but less likely to report transference to infection/fever clinics (18.26 vs. $10.07\%$) as a reason of the delay compared with their counterparts from the regions with low risk (Figure 1).
**Figure 1:** *Reasons of delay in seeking hospital care.*
Figure 2 shows that the perceived consequences of delayed care followed the same pattern between those living in the regions with high and low risk. Slow recovery (43.29−$50.43\%$) was the top concern, which was followed by disruptions in medication (39.13−$42.62\%$), psychological distress (37.92−$38.26\%$), missed optimal timing of treatment (27.85−$33.04\%$), and deterioration of illness conditions ($14.78\%$−$16.44\%$) (Figure 2).
**Figure 2:** *Perceived consequences of delay in seeking hospital care.*
## Discussion
The COVID-19 pandemic has disrupted healthcare services around the world, which may have serious implications on population health outcomes. Our study shows that $31.4\%$ of patients in mainland China experienced delay in seeking health care over a period with low prevalence of COVID-19. This level of delay is relatively lower compared to those experienced by other countries, whether in the settings with high prevalence of COVID-19 such as the US ($65.7\%$) [41], or in the settings with low prevalence of COVID-19 such as New Zealand ($55\%$) [42]. There is emerging evidence indicating serious consequences of delayed health care seeking. A study in a tertiary care center in India showed that $44.7\%$ of pregnancy complications over the period of COVID-19 outbreak were resulted from delay in health care seeking [43].
In our study, eye, nose and throat diseases and cardiovascular/cerebrovascular diseases were reported by respondents as the top two conditions for which needed care was delayed. The delay was system wide, with the majority ($68\%$) occurred with planned care in local provincial/municipal hospitals. A systematic review concludes that COVID-19 has a significant impact on health care seeking behaviors of patients with cardiovascular diseases, causing longer delays between the onset of the symptoms and hospital treatment [44]. However, there are limited studies reporting delay in care for eye, nose and throat diseases.
In our study, fear of infection was identified as the major reason for delay in seeking health care, followed by complex services procedure. These findings are consistent with the results of previous studies [45]. COVID-19 outbreaks have caused worldwide shortage of health workforce and medicine supplies, transportation difficulties, and even closure of certain medical care services [46]. Thaddeus and Maine [47] described three common delays—in seeking care, in reaching the facility, and in receiving adequate treatment. The high contagious nature of the SARS-CoV-2 virus and its relatively high death toll has fueled serious fear of nosocomial infection. Indeed, among the early cases of COVID-19, nearly $41\%$ were suspected to be infected during their hospital visits [48]. The risk of nosocomial exposure can trigger psychological panic and avoidance of medical care, especially in the vulnerable populations [49].
Control of the pandemic of COVID-19 has drained tremendous resources that may otherwise be used for other health care services, and led to increased complexity in services procedure. Many hospitals in China started to only accept patients who had an online appointment, despite the challenge of making online appointments by some patients, especially the elderly [50]. Normal body temperature was required to get access to non-COVID related treatment, despite a lack of robust evidence to support such practice [51]. Additional precaution measures were put into place for hospital services and surgical treatment, including infection risk assessment, nucleic acid testing, and even chest computerized tomography (CT) [52, 53]. The increased complexity has added barriers for patients to seek timely care. Similar findings were also reported in other countries. In the United States, for example, patients experienced longer waiting time and the existing racial and socioeconomic inequities in health care were exacerbated by the COVID-19 outbreak [54, 55]. Mobility restriction measures, such as travel restriction, suspension of public transport, isolation of infected cases, and quarantine of close contacts are also associated with delay in seeking health care according to the findings of our study and the exiting literature [56, 57]. In some countries, governments even imposed nationwide lockdown to contain COVID-19 [58].
The predictors of delay in seeking health care identified in our study cover all of the three categories of factors proposed by Anderson [59]: predisposing factor (age), enabling factor (mobility restriction, living with a pregnant woman, using Internet-based medical services), and needs factor (chronic conditions, risk perception). Indeed, delay in health care seeking is a result of balancing act that is shaped by the felt urgency of care need, perceived risk of infection, and self-coping ability under a constraint environment [42]. We found that the residents aged between 31 and 59 years are most likely to experience delay in seeking health care after adjustment for variations in other variables. This is consistent with the findings of the studies conducted elsewhere. The potential reasons include economic and functional limitations, while the biological and pathological factors may also contribute to the delay of treatment among 31–59 years old adults [60]. We also found that residents living in high risk regions were more likely to report delay in seeking health care, which may be associated with higher levels of mobility restrictions. Meanwhile, it is also interesting to note that pregnancy or living with a pregnant woman and use of Internet-based medical services are associated with delay in seeking health care. This may have reflected the common value and coping strategies adopted by the Chinese people: priority in family protection of the pregnant women and unborn babies and using the Internet-based medical care to minimize risk of infection [61]. COVID-19 has triggered a surge of Internet-based medical care [62]. However, it is considered as part of self-management in China, which is not considered a complete patient care [63]. Delay in seeking health care is more likely to be seen in those with chronic conditions according to the findings of this study and others [64]. Empirical evidence shows that patients with chronic conditions are particularly vulnerable to COVID-19. They have a much higher COVID mortality rate than the general population [65], and tend to take extra precaution to avoid health facilities for fear of infection [66]. Both COVID-19 and chronic conditions have been proven to be associated with anxiety and depression [67, 68]. It is common in the public to see health care facilities as the most dangerous place due to the high risk of nosocomial infection and high death toll of COVID-19 [69, 70]. Our study revealed consistently high levels of perceived severity of COVID-19 in the respondents across the regions with high and low risk. However, low levels of perceived susceptibility and high levels of perceived controllability are also evident in this study, which may offer some explanation about the relatively low level of delay in seeking health care in China.
## Limitations
This study adopted a cross-sectional design, which does not allow us to compare the levels of delay in seeking health care before and after the outbreak of COVID-19. No causal conclusions can be established either. The population was not stratified for sampling although a simple random sampling strategy was adopted through the Wenjuanxing platform. We intended to obtain a maximal sample size without calculating the statistical power. The final sample size far exceeds the requirements of a statistical power of 0.8, with an α of 0.05. The final study sample was biased toward those residing in the regions with high risk and those younger than 60 years. The measurement of delay is also subject to recall bias. Attempts to generalize the results to the entire population in China need to be cautious.
## Conclusions
Delay in seeking health care remained at a relatively high level in mainland China (albeit lower than in some other countries) when the prevalence of COVID-19 cases was low: more than $30\%$ patients delayed or avoided needed care. This may present a serious health risk to the patients, in particular those living with chronic conditions who need continuous medical care. Fear of infection and complex service procedures are the major underlying reasons of delay/avoidance of health care, in particular in relation to eye, nose, and throat diseases and cardiovascular and cerebrovascular diseases. Access to Internet-based self-care, restrictions on population movements in high risk regions, and perceived low controllability of COVID-19 are also associated with delay in seeking health care during COVID-19 in China. Although self-treatment at home with support from the Internet-based advices may mitigate some consequences, further studies are needed to unveil the full consequences of delay/avoidance in seeking health care.
## 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 Harbin Medical University. IRB code is HMUIRB20200004. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
YH and CL took overall responsibility for the study design, coordination of the survey, setting up the study framework, and writing. ZW, YT, and YC drafted the manuscript, conducted the survey, and data analyses. XC, HG, YuL, and YaL participated in the literature review and data analyses. ZK and QW participated in the design of the research and revision of 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. Frustaci AM, Pioltelli ML, Ravano E, Di Ruscio F, Campisi DA, Puoti M. **Intrahospital COVID-19 infection outbreak management: Keep calm and carry on**. *Hematol Oncol.* (2021) **39** 431-3. DOI: 10.1002/hon.2873
2. 2.World Health Organization. Coronavirus disease (COVID-19) Weekly Epidemiological Update and Weekly Operational Update. (2022). Available online at: https://www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports. (2022)
3. Xiao H. **The impact of the COVID-19 pandemic on health services utilization in China: Time-series analyses for 2016-2020**. *Soc Sci Electron Publish.* (2021) **9** 100122. DOI: 10.1016/j.lanwpc.2021.100122
4. Jones SA, Gopalakrishnan S, Ameh CA, White S, van den Broek NR. **Women and babies are dying but not of Ebola': the effect of the Ebola virus epidemic on the availability, uptake and outcomes of maternal and newborn health services in Sierra Leone**. *BMJ Global Health.* (2016) **1** e000065. DOI: 10.1136/bmjgh-2016-000065
5. 5.National Health Commission. (2022). Available online at: http://www.nhc.gov.cn/xcs/kpzs/202002/7c70747c793a4e35ad135f68db70d7d8.shtml. (2022)
6. Hong Z, Li N, Li D, Li J, Li B, Xiong W. **Telemedicine During the COVID-19 Pandemic: Experiences From Western China**. *J Med Internet Res.* (2020) **22** e19577. DOI: 10.2196/19577
7. Burki T. **China's successful control of COVID-19**. *Lancet Infect Dis* (2020) **20** 1240-41. DOI: 10.1016/S1473-3099(20)30706-4
8. Kim D, Bonham CA, Konyn P, Cholankeril G, Ahmed A. **Mortality trends in chronic liver disease and cirrhosis in the United States, before and during COVID-19 pandemic**. *Clin Gastroenterol Hepatol.* (2021) **19** 2664-6. DOI: 10.1016/j.cgh.2021.07.009
9. Masroor S. **Collateral damage of COVID-19 pandemic: delayed medical care**. *J Card Surg.* (2020) **35** 1345-7. DOI: 10.1111/jocs.14638
10. Regala P, White PB, Bitterman AD, Katsigiorgis G, Dicpinigaitis PA. **Delay in treatment of a bimalleolar ankle fracture during coronavirus disease-19 COVID-19 pandemic leading to amputation**. *J Orthop Case Rep.* (2021) **11** 28-32. DOI: 10.13107/jocr.2021.v11.i04.2138
11. Miyagami T, Uehara Y, Harada T, Watari T, Shimizu T, Nakamura A. **Delayed treatment of bacteremia during the COVID-19 pandemic**. *Diagnosis.* (2021) **8** 327-32. DOI: 10.1515/dx-2020-0114
12. Drumm B, Bentley P, Brown Z, Anna LD, Dolkar T, Halse O. **Impact of the Covid-19 Pandemic on Stroke Thrombolysis Rate and Delay to Thrombolysis Treatment in a Regional Stroke Centre in London, UK**. *Stroke* (2021) **52** 99. DOI: 10.1161/str.52.suppl_1.P99
13. Samadzadeh S, Brauns R, Rosenthal D, Hefter H. **The impact of SARS-CoV-2 pandemic lockdown on a botulinum toxin outpatient clinic in Germany**. *Toxins.* (2021) **13** 101. DOI: 10.3390/toxins13020101
14. Mark C, Marynak K, Clarke K, Salah Z, Shakya L, Howard ME. **Delay or avoidance of medical care because of COVID-19–related concerns—United States, June 2020**. *MMWR* (2020) **69** 1250-7. DOI: 10.15585/mmwr.mm6936a4
15. Leone JE, Rovito MJ, Gray KA, Mallo R, Boston MA, Orlando FL. **Practical strategies for improving men's health: maximizing the patient-provider encounter**. *Int J Men's Soc Community Health.* (2021) **4** 1-16. DOI: 10.22374/ijmsch.v4i1.17
16. Graboyes EM, Kompelli AR, Neskey DM, Brennan E, Nguyen S, Sterba KR. **Day. Association of treatment delays with survival for patients with head and neck cancer: a systematic review**. *JAMA Otolaryngol Head Neck Surg.* (2018) **145** 166-77. DOI: 10.1001/jamaoto.2018.2716
17. Takakubo T, Odagiri Y, Machida M, Takamiya T, Fukushima N, Kikuchi H. **Changes in the medical treatment status of Japanese outpatients during the coronavirus disease 2019 pandemic**. *J Gen Fam Med.* (2021) **22** 246-61. DOI: 10.1002/jgf2.432
18. Andrews D. *Hospitals Ready Waiting to Support all Victorians* (2020)
19. Dinmohamed AG, Visser O, Verhoeven RHA, Louwman MWJ, van Nederveen FH, Willems SM. **Fewer cancer diagnoses during the COVID-19 epidemic in the Netherlands**. *Lancet Oncol.* (2020) **21** 750-1. DOI: 10.1016/S1470-2045(20)30265-5
20. Maringe C, Spicer J, Morris M, Purushotham A, Nolte E, Sullivan R. **The impact of the COVID-19 pandemic on cancer deaths due to delays in diagnosis in England, UK: a national, population-based, modelling study**. *Lancet Oncol.* (2020) **21** 1023-34. DOI: 10.1016/S1470-2045(20)30388-0
21. Czeisler MÉ, Kennedy JL, Wiley JF, Facer-Childs ER, Robbins R, Barger LK. **Delay or avoidance of routine, urgent and emergency medical care due to concerns about COVID-19 in a region with low COVID-19 prevalence: Victoria, Australia**. *Respirology.* (2021) **26** 707-12. DOI: 10.1111/resp.14094
22. An K. **Pre-hospital delay in treatment after acute myocardial infarction**. *J Korean Acad Nurs.* (2001) **31** 1141. DOI: 10.4040/jkan.2001.31.7.1141
23. Ansar A, Lewis V, McDonald CF, Liu C, Rahman A. **Defining timeliness in care for patients with lung cancer: protocol for a scoping review**. *BMJ Open.* (2020) **10** e039660. DOI: 10.1136/bmjopen-2020-039660
24. Grover S, Mehra A, Sahoo S, Avasthi A, Tripathi A, D'Souza A. **State of mental health services in various training centers in India during the lockdown and COVID-19 pandemic**. *Indian J Psychiatry.* (2020) **62** 363-9. DOI: 10.4103/psychiatry.IndianJPsychiatry_567_20
25. Neely Barnes S, Hunter A, Meiman J, Malone C, Hirschi M, Delavega E. **Leaning into the crisis: managing COVID-19 in social services and behavioral health agencies**. *Hum Serv Org.* (2021) **45** 293-306. DOI: 10.1080/23303131.2021.1915905
26. Kendzerska T, Zhu DT, Gershon AS, Edwards JD, Peixoto C, Robillard R. **The effects of the health system response to the COVID-19 pandemic on chronic disease management: a narrative review**. *Risk Manag Healthc Policy.* (2021) **14** 575-84. DOI: 10.2147/RMHP.S293471
27. Zarrintan S. **Surgical operations during the COVID-19 outbreak: Should elective surgeries be suspended?**. *Int J Surg.* (2020) **78** 5-6. DOI: 10.1016/j.ijsu.2020.04.005
28. Nabe-Nielsen K, Nilsson CJ, Juul-Madsen M, Bredal C, Hansen LOP, Hansen ÅM. **COVID-19 risk management at the workplace, fear of infection and fear of transmission of infection among frontline employees**. *Occup Environ Med.* (2020) **78** 248-54. DOI: 10.1136/oemed-2020-106831
29. Park JD. **A study on the factors for the elderly living alone at home to determine their participation in a health promotion activity program: with the application of anderson model**. *Soc Welf Policy* (2011) **38** 1-23. DOI: 10.15855/swp.2011.38.4.1
30. 30.WHO. Coronavirus disease (COVID-19) advice for the public. (2022). Available online at: https://www.who.int/emergencies/diseases/novel-coronavirus-2019/advice-for-public. (2022)
31. 31.National Health Commission. (2022). Available online at: http://www.nhc.gov.cn/cms-search/xxgk/searchList.htm?type=search. (2022)
32. 32.Our world in data. (2022). Available online at: https://ourworldindata.org/coronavirus/country/china. (2022)
33. Chai N, Stevens R, Fang XZ, Mao C, Wang D. **The impact of compensation upon urban village residents satisfaction with the land expropriation process Empirical evidence from Hangzhou, China**. *Int J Law Built Environ.* (2019) **11** 186-216. DOI: 10.1108/JPPEL-03-2019-0011
34. Bauer RA. **Consumer behavior as risk taking[A]**. *Dynamic marketing for a changing world[C]* (1960) 389-98
35. **Perception of Risk**. *Science* (1987) **236** 280-5. DOI: 10.1126/science.3563507
36. Adams AM, Smith AF. **Risk perception and communication: recent developments and implications for anaesthesia**. *Anaesthesia.* (2001) **56** 745-55. DOI: 10.1046/j.1365-2044.2001.02135.x
37. Park T, Ju I, Ohs JE, Hinsley A. **Optimistic bias and preventive behavioral engagement in the context of COVID-19**. *Res Soc Adm Pharm.* (2020) **17** 1859-66. DOI: 10.1016/j.sapharm.2020.06.004
38. Yajun D. **Establishment and evaluation on reliability and validity of public risk perception scale for public health emergencies**. *Chin J Public Health.* (2020) **36** 227-31. DOI: 10.11847/zgggws1119744
39. Loza E, Abásolo L, Jover JA, Carmona L. **Burden of disease across chronic diseases: A health survey that measured prevalence, function, and quality of life**. *J Rheumatol.* (2008) **35** 159-65. DOI: 10.1080/10582450802479693
40. 40.WHO International Classification of Diseases (ICD) (who.int). (2022). Available online at: https://www.who.int. (2022)
41. Chertcoff A, Bauer J, Silva BA, Aldecoa M, Eizaguirre MB, Rodriguez R. **Changes on the health care of people with multiple sclerosis from Latin America during the COVID-19 pandemic**. *Mult Scler Relat Disord.* (2021) **54** 103120-103120. DOI: 10.1016/j.msard.2021.103120
42. Imlach F, McKinlay E, Kennedy J, Pledger M, Middleton L, Cumming J. **Seeking healthcare during lockdown: challenges, opportunities and lessons for the future**. *Int J Health Policy Manag* (2021) **21** 1-14. DOI: 10.34172/ijhpm.2021.26
43. Goyal M, Singh P, Singh K, Shekhar S, Agrawal N, Misra S. **The effect of the COVID-19 pandemic on maternal health due to delay in seeking health care: experience from a tertiary center**. *Int J Gynaecol Obstet.* (2020) **152** 231-5. DOI: 10.1002/ijgo.13457
44. Kiss P, Carcel C, Hockham C, Peters SAE. **The impact of the COVID-19 pandemic on the care and management of patients with acute cardiovascular disease: a systematic review**. *Eur Heart J Qual Care Clin Outcomes.* (2021) **7** 18-27. DOI: 10.1093/ehjqcco/qcaa084
45. Lai AYK, Sit SMM, Wu SYD, Wang MP, Wong BYM, Ho SY. **Corrigendum: associations of delay in doctor consultation with COVID-19 related fear, attention to information, and fact-checking**. *Front Public Health.* (2022) **10** 847603. DOI: 10.3389/fpubh.2022.847603
46. Mohseni M, Ahmadi S, Azami-Aghdash S, Mousavi Isfahani H, Moosavi A, Fardid M. **Challenges of routine diabetes care during COVID-19 era: a systematic search and narrative review**. *Prim Care Diabetes.* (2021) **15** 918-22. DOI: 10.1016/j.pcd.2021.07.017
47. Thaddeus S, Maine D. **Too far to walk: maternal mortality in context**. *Soc Sci Med.* (1994) **38** 1091-110. DOI: 10.1016/0277-9536(94)90226-7
48. Wang D, Hu B, Hu C, Zhu F, Liu X, Zhang J. **Clinical characteristics of 138 hospitalized patients with 2019 novel coronavirus infected pneumonia in Wuhan, China**. *JAMA.* (2020) **323** 1061-9. DOI: 10.1001/jama.2020.1585
49. Lu M, Sue YM, Hsu HL, Zhang JF, Liu YJ, Yen YC. **Tuberculosis treatment delay and nosocomial exposure remain important risks for patients undergoing regular hemodialysis**. *J Microbiol Immunol Infect* (2021) **8** 926-34. DOI: 10.1016/j.jmii.2021.08.011
50. Ye J. **Health information system's responses to COVID-19 pandemic in China: a national cross-sectional study**. *Appl Clin Inform.* (2021) **12** 399-406. DOI: 10.1055/s-0041-1728770
51. Pană BC, Lopes H, Furtunescu F, Franco D, Rapcea A, Stanca M. **Real world evidence: The low validity of temperature screening for COVID19 triage**. *Front. Public Health* (2020) **9** 672698. DOI: 10.3389/fpubh.2021.672698
52. Cheung SSL, Wong CYK, Chan JCK, Chan CKM, Lam NM, Yuen HKL. **Ophthalmology in the time of COVID-19: experience from Hong Kong Eye Hospital**. *Int J Ophthalmol.* (2020) **13** 851-9. DOI: 10.18240/ijo.2020.06.01
53. Hong L, Ye E, Sun G, Wang X, Zhang S, Wu Y. **Clinical and radiographic characteristics, management and short-term outcomes of patients with COVID-19 in Wenzhou, China**. *BMC Infect Dis.* (2020) **20** 841. DOI: 10.1186/s12879-020-05528-z
54. Laster Pirtle WN. **Racial capitalism: a fundamental cause of novel coronavirus (COVID-19) pandemic inequities in the United States**. *Health Educ Behav.* (2020) **47** 504-8. DOI: 10.1177/1090198120922942
55. Davillas A, Jones AM. **Unmet health care need and income-Related horizontal equity in use of health care during the COVID-19 pandemic**. *Health Econ.* (2021) **30** 1711-6. DOI: 10.1002/hec.4282
56. Ac A, Rps B. **Decline in PM**. *Environ Res.* (2020) **187** 109634. DOI: 10.1016/j.envres.2020.109634
57. Chen Z, Hao X, Zhang X, Chen F. **Have traffic restrictions improved air quality? A shock from COVID-19**. *J Clean Prod.* (2021) **279** 123622. DOI: 10.1016/j.jclepro.2020.123622
58. Shrivastava S, Rai S, Sivakami M. **Challenges for pregnant women seeking institutional care during the COVID-19 lockdown in India: a content analysis of online news reports**. *Indian J Med Ethics* (2021) **VI** 21-4. DOI: 10.20529/IJME.2021.004
59. Andersen RM. **Revisiting the behavioral model and access to medical care: Does it matter?**. *J Health Soc* (1995) **36** 1-10. DOI: 10.2307/2137284
60. Zheng W, Kämpfen F, Huang Z. **Health-seeking and diagnosis delay and its associated factors: a case study on COVID-19 infections in Shaanxi Province, China**. *Sci Rep.* (2021) **11** 17331. DOI: 10.1038/s41598-021-96888-2
61. Bermejo-Sánchez FR, Peña-Ayudante WR, Espinoza-Portilla E. **Perinatal depression in times of COVID-19: the role of social media on the internet**. *Acta Med Peru* (2020) **37** 88-93. DOI: 10.35663/amp.2020.371.913
62. Huang W, Cao B, Yang G, Luo N, Chao N. **Turn to the internet first? Using online medical behavioral data to forecast COVID-19 epidemic trend**. *Inf Process Manag.* (2020) **58** 102486. DOI: 10.1016/j.ipm.2020.102486
63. Peng YY, Li XL, Zhao SZ, He XL, Shi ZY. **Survey of online outpatient clinic usage experiences and analysis of factors influencing retreatment**. *J Nurs* (2021) **68** 43-53. DOI: 10.6224/JN.202102_68(1).07
64. Murewanhema G, Makurumidze R. **Essential health services delivery in Zimbabwe during the COVID-19 pandemic: perspectives and recommendations**. *Pan Afr Med J.* (2020) **35** 143-143. DOI: 10.11604/pamj.supp.2020.35.2.25367
65. Abraham DA, Vijayakumar TM, Rajanandh MG. **Challenges of non-COVID-19 patients with chronic illness during the pandemic**. *J Res Pharm Pract.* (2020) **9** 155. DOI: 10.4103/jrpp.JRPP_20_64
66. Jindal S, Jindal A, Moitra S. **Problems of management of non-corona respiratory diseases in the era of COVID-19**. *Int J Noncommun Dis.* (2020) **5** 63-9. DOI: 10.4103/jncd.jncd_30_20
67. Evans S, Alkan E, Bhangoo JK, Tenenbaum H, Ng-Knight T. **Effects of the COVID-19 lockdown on mental health, wellbeing, sleep, and alcohol use in a UK student sample**. *Psychiatry Res.* (2021) **298** 113819. DOI: 10.1016/j.psychres.2021.113819
68. Ganson KT, Weiser SD, Tsai AC, Nagata JM. **Associations between anxiety and depression symptoms and medical care avoidance during COVID-19**. *J Gen Intern Med.* (2020) **35** 3406-8. DOI: 10.1007/s11606-020-06156-8
69. Abbas M, Nunes TR, Martischang R, Zingg W, Iten A, Pittet D. **Nosocomial transmission and outbreaks of coronavirus disease 2019: the need to protect both patients and healthcare workers**. *Antimicrob Resist Infect Control.* (2021) **10** 7. DOI: 10.1186/s13756-020-00875-7
70. Su C, Zhang Z, Zhao X, Peng HL, Hong Y, Huang LL. **Changes in Prevalence of Nosocomial Infection Pre- and Post-COVID-19 pandemic from a Tertiary Hospital in China**. *BMC Infect Dis* (2021) **21** 693. DOI: 10.1186/s12879-021-06396-x
|
---
title: Sclerostin, vascular risk factors, and brain atrophy in excessive drinkers
authors:
- Candelaria Martín-González
- Ana María Godoy-Reyes
- Pedro Abreu-González
- Camino María Fernández-Rodríguez
- Esther Martín-Ponce
- María José Sánchez-Pérez
- Julio César Alvisa-Negrín
- Melchor Rodríguez-Gaspar
- Emilio González-Reimers
journal: Frontiers in Human Neuroscience
year: 2023
pmcid: PMC9989031
doi: 10.3389/fnhum.2023.1084756
license: CC BY 4.0
---
# Sclerostin, vascular risk factors, and brain atrophy in excessive drinkers
## Abstract
### Objective
Heavy alcohol consumption causes several organic complications, including vessel wall calcification. Vascular damage may be involved in the development of brain atrophy and cognitive impairment. Recently, sclerostin (whose levels may be altered in alcoholics) has emerged as a major vascular risk factor. The objective of the present study is to analyze the prevalence of vascular calcifications in alcoholics, and the relationships of these lesions with brain atrophy, as well as the role of sclerostin on these alterations.
### Patients and methods
A total of 299 heavy drinkers and 32 controls were included. Patients underwent cranial computed tomography, and several indices related to brain atrophy were calculated. In addition, patients and controls underwent plain radiography and were evaluated for the presence or absence of vascular calcium deposits, cardiovascular risk factors, liver function, alcohol intake, serum sclerostin, and routine laboratory variables.
### Results
A total of 145 ($48.47\%$) patients showed vascular calcium deposits, a proportion significantly higher than that observed in controls (χ2 = 16.31; $p \leq 0.001$). Vascular calcium deposits were associated with age ($t = 6.57$; $p \leq 0.001$), hypertension ($t = 5.49$; $p \leq 0.001$), daily ethanol ingestion ($Z = 2.18$; $$p \leq 0.029$$), duration of alcohol consumption ($Z = 3.03$; $$p \leq 0.002$$), obesity (χ2 = 4.65; $$p \leq 0.031$$), total cholesterol ($Z = 2.04$; $$p \leq 0.041$$), triglycerides ($Z = 2.05$; $$p \leq 0.04$$), and sclerostin levels ($Z = 2.64$; $$p \leq 0.008$$). Calcium deposits were significantly related to Bifrontal index ($Z = 2.20$; $$p \leq 0.028$$) and Evans index ($Z = 2.25$; $$p \leq 0.025$$). Serum sclerostin levels were related to subcortical brain atrophy, assessed by cella media index ($Z = 2.43$; $$p \leq 0.015$$) and Huckmann index (ρ = 0.204; $$p \leq 0.024$$). Logistic regression analyses disclosed that sclerostin was the only variable independently related to brain atrophy assessed by altered cella media index. Sclerostin was also related to the presence of vascular calcifications, although this relationship was displaced by age if this variable was also included.
### Conclusion
Prevalence of vascular calcification in alcoholics is very high. Vascular calcium deposits are related to brain atrophy. Serum sclerostin is strongly related to brain shrinkage and also shows a significant relationship with vascular calcifications, only displaced by advanced age.
## Introduction
Ethanol is a toxic compound for human beings. Although, classically, the boundary of the amount of ethanol consumption associated with organic complications was situated at about 50 g/day for men (21 drinks a week) and 30 g/day for women (Reid et al., 1999), more recent epidemiological studies show that drinking more than 100 g/week may shorten lifespan (Wood et al., 2018), and consumption of more than 30 g/day among men or 5–15 g/day among women may be associated with increased cancer risk (Cao et al., 2015). Therefore, the definition of the safe alcohol consumption limits is an issue subjected to debate.
Heavy drinkers develop many important, life-threatening complications. Although liver, cancer, or pancreatic disease constitute outstanding alcohol-related disorders, alcoholic cardiomyopathy, osteosarcopenia/osteosarcopenic adiposity, or brain affectation are very commonly associated with ethanol consumption and importantly contribute to morbidity and mortality of these patients.
Brain damage may severely impair the quality of life of alcoholics. There is general agreement regarding the deleterious influence of heavy ethanol consumption on brain structure and function, but some controversy exists in relation to the effect of light-to moderate ethanol consumption on brain alterations (Ridley et al., 2013; Rehm et al., 2019; Peng et al., 2020). This controversy may be due to the presence of many confounding factors associated with alcoholism, that add to the widespread direct or indirect changes caused by ethanol on metabolic pathways potentially involved in adequate brain function. In addition to the direct inhibitory effects of ethanol on neurogenesis, proinflammatory cytokines and oxidative stress may cause neuroinflammation and neurodegeneration (Crews and Nixon, 2009; Qin and Crews, 2012). Repeated microtrauma associated with the bizarre style of life of many alcoholics, prone to aggression and violence may also contribute (Harris et al., 2019). Altered nutritional status (Romero-Acevedo et al., 2019), or several micronutrient deficiencies also probably play a role (González-Reimers et al., 2014), especially thiamine deficiency (Topiwala and Ebmeier, 2018).
One of the hallmarks of the alcoholic brain shrinkage and cognitive impairment is the potential recovery after alcohol cessation (Pfefferbaum et al., 1995), something that supports the idea of a transient effect of ethanol on some metabolic pathways, that recover their function after alcohol withdrawal. However, in many patients brain functional and/or morphological recovery may be incomplete despite alcohol cessation (Ridley et al., 2013; de La Monte and Kril, 2014), an observation possibly related to the existence of an already established organic damage. In this sense, some authors have reported that heavy alcohol consumption increases τ phosphorylation and β amyloid accumulation [features of Alzheimer disease, (Peng et al., 2020)] but a direct correlation between alcohol consumption and Alzheimer disease has not been described (Ehrlich et al., 2012).
In heavy alcoholics, in addition to the presence of lesions like those observed in Alzheimer disease, brain lesions derived from vascular alterations may be also present. Vessel wall calcifications constitute a hallmark of vascular damage and are associated to increased vascular risk (Rennenberg et al., 2009). Vessel wall calcifications are frequent in heavy drinkers (Shi et al., 2020), as pointed out in previous studies. For instance, Pletcher et al. [ 2005] report a clear-cut, independent association among coronary artery calcification and ethanol consumption. Oros et al. [ 2012] have clearly shown that ethanol promotes vascular smooth muscle cells calcification and transition of these cells to osteoblastic-like cells, providing a strong support to the finding of radiologically detectable vascular calcification in excessive drinkers, independently of the concomitant presence of hypertension. In a study performed in Korea, there was a parallel increase in the incidence of coronary artery calcification and ethanol ingestion, as well as a relationship between ethanol ingestion and hypertension (Yun et al., 2017). Indeed, increased prevalence of hypertension has been reported in alcoholics (Saunders, 1987; van Leer et al., 1994; Fuchs et al., 2001; Fuchs and Fuchs, 2021). Hypertension in alcoholics may develop either by the effects of ethanol by itself (Marmot et al., 1994) or through other factors associated with alcoholism, such as tobacco consumption, gender, or age. It is well known that hypertension is associated with vascular damage, smooth muscle cell remodeling, and vessel wall calcification (Shi et al., 2020). Therefore, in alcoholic patients, vessel wall calcification may be related both to the effect of ethanol by itself and/or to associated hypertension.
The chronic “smoldering” inflammatory status associated with heavy ethanol drinking surely plays a major role on the development of vascular calcifications, together with the simultaneous presence of other risk factors, such as diabetes, or dyslipidemia. Moreover, the multisystemic effects of ethanol may alter the expression and/or functional activity of diverse compounds involved in vascular damage. Recently, the role of sclerostin as a new vascular risk factor has gained attention (Catalano et al., 2020). Sclerostin is a member of the so called osteokines, i.e., bone derived cytokines able to exert a variety of functions in the intermediate metabolism (Kirk et al., 2020). In the last decade several authors have analyzed the role of this molecule on vascular calcifications, observing in in vitro studies that sclerostin was involved in medial vascular smooth muscle cells calcification (Zhu et al., 2011), in the formation of the atherosclerotic plaque, as shown by Leto et al. [ 2019] in 46 patients undergoing carotid endarterectomy, and also in the calcification of the aortic valve (Koos et al., 2013). In other clinical studies serum sclerostin levels were related to vascular calcification in 51 patients with end-stage renal disease (Li et al., 2019). Pelletier et al. [ 2015] also found a marked association between high sclerostin levels and aortic calcification in 53 patients with chronic kidney disease. Some studies have analyzed the behavior of sclerostin among patients with alcoholism or liver disease -although with conflicting reported results (González-Reimers et al., 2013; Wakolbinger et al., 2020; Jadzic et al., 2022; Martín González et al., 2022) but the relationship of sclerostin with vascular changes and/or brain alterations among alcoholics has received little attention.
Based on these facts, in the present study we want to analyze the prevalence of vascular calcifications and the relationship of these vascular lesions with computed tomography (CT)-assessed brain alterations in alcoholics, and to explore the relationship of the osteokine sclerostin with vascular calcification and brain damage in a subset of these patients.
## Patients and methods
In this observational study we included 299 patients consecutively admitted via emergency room to the Internal Medicine Unit of our Hospital due to organic problems related to alcohol consumption. The sample included 271 men and 28 women, who underwent cranial CT at admission (in most cases by withdrawal syndrome), and calculation of several classic indices related to brain shrinkage (Huckmann, Evans, bifrontal, ventricular, bicaudate, cella media; Figure 1). The only selection criteria required were chronic consumption of a daily amount of at least 80 g pure ethanol (men) or 40 g (women) for at least the last 5 years before admission; and the clinical indication of a brain CT study (mainly coma, seizures, withdrawal syndrome, traumatism). Patients with meningitis/meningoencephalitis, intracranial bleeding, brain abscess or tumor, were excluded. The last 122 patients were included in a prospective study devoted to analyzing the behavior of sclerostin in relation with brain alterations and vascular calcifications.
**FIGURE 1:** *Brain computed tomography, indicating the different indices used in this study and how they were calculated: Bifrontal index = maximum width of frontal horns/skull width at the same level = A/D. Evans index = maximum width of frontal horns/skull width at the level of the third ventricle = A/F. Bicaudate index = minimum width of lateral ventricles/skull width at the same level = B/E. Ventricular index = minimum width of lateral ventricles/maximum width of frontal horns = B/A. Cella media index = maximum width of the skull/width of lateral ventricles = G/H.*
All the patients were heavy drinkers. The absolute amount of alcohol consumed was estimated by direct inquiry, both to the patients and close relatives, recording type of beverage(s) and daily amount ingested to calculate amount (in g) of ethanol consumed as: degree of beverage (in%) × beverage volume × 0.8 (alcohol density). Included patients drank a median daily ethanol amount of 197 g [interquartile range (IQR) = 100–250 g] during 31 (IQR = 24–40) years. Patients who consumed any other drug besides tobacco (smoked by 202 patients) were not included in this study. All the patients underwent a complete laboratory evaluation. Body mass index (BMI) was calculated as weight (kg)/height (m2).
A plain thoracic X-ray film was performed to all these patients. In 295 cases the presence or not of calcium deposits in the aortic arch was assessed (in the remaining 4 cases poor X-ray quality precluded accurate evaluation).
Plain X-ray film was also performed to 32 sanitary workers, drinkers of less than 10 g ethanol/day, with similar sex distribution (29 men, 3 women; χ2 = 0; $$p \leq 1$$) and age (53.25 ± 11.11 years; $t = 1.93$; $$p \leq 0.06$$) than the alcoholic patients (57.35 ± 11.40) years. The individuals belonging to the control group were randomly selected among hospital workers, previous informed consent. The only criteria for selection were to be either teetotalers or occasional drinkers of less than 10 g/day; having an approximate age/sex distribution to that of the patients; and having not suffered any brain illness, intervention, or traumatic event.
In addition to complete clinical evaluation, patients also underwent abdominal ultrasound (US) examination. The presence of splenomegaly and/or portal dilatation and a heterogeneous liver structure and irregular shape, together with altered levels either of albumin, bilirubin, or prothrombin activity, served us to classify the patients as cirrhotics, a condition fulfilled by 126 out of the 299 patients, whereas the remaining 173 were classified as non-cirrhotics. We also recorded the presence or not of liver steatosis (that was observed in 128 patients), based on US examination. To achieve a global assessment of liver function, we applied the Child-Pugh score to the whole sample, despite being aware that this score was initially designed as a prognostic tool only for cirrhotics. Child score is based on the alteration of the following variables: serum albumin, bilirubin, prothrombin activity, and presence/severity of ascites and/or encephalopathy (Child and Turcotte, 1964; Pugh et al., 1973). The presence of hypertension (previous or current diagnosis) or diabetes were also recorded.
## Laboratory assessment
All the patients underwent complete routine laboratory analysis. Blood samples were taken at 8.00 am in fasting conditions, in order to determine serum levels of variables related to ethanol consumption such as gamma glutamyl transferase (GGT) and mean corpuscular volume (MCV); liver function variables such as bilirubin, albumin, and prothrombin activity; serum creatinine; and variables related to metabolic syndrome, such as total, LDL and HDL cholesterol, triglycerides, uric acid, and glycated hemoglobin. Serum sclerostin was determined to 122 patients and 31 controls by ELISA method, using a commercial kit purchased from Thermo Scientific Laboratories (Thermo Fisher Scientific Co., Waltham, MA, USA). The calibration curve of ELISA was set 0–10,000 pg/ml. The assay was evaluated with a 4PL algorithm. The correlation analysis between absorbance units (AU) and standards was 0.9945. The λ max of the analysis was established at 450 nm, using a microplate spectrophotometer reader (Spectra MAX-190, Molecular Devices, Sunnyvale, CA, USA). The lower limit of detection (zero + 2 SD) of this assay was 12 pg/ml. Intra and inter-assay coefficients of variation (CV) were $4.32\%$ and $5.18\%$, respectively. The final serum concentration of sclerostin was expressed in pmol/L (conversion factor: 1 pg/ml = 0.044 pmol/L, molecular weight = 22.5 kDa).
## Statistical analysis
The Kolmogorov–Smirnov test was used to test for normal or Gaussian distribution, a condition not fulfilled by several variables. Therefore, non-parametric tests, such as Mann–Whitney’s U test and Kruskal–Wallis test and Spearman’s correlation analysis were used to analyze differences or correlations among non-parametric variables. When the variables subjected to analysis showed a normal distribution, Student’s t test, variance analysis and Pearson’s correlation analysis were used. Stepwise logistic regression analyses (dichotomizing the selected variables according to medians) were used to discern if a given result obtained in the univariate analyses was independent of confounding factors. Multiple linear regression analyses were also used to disclose the confounding effect of age (or other continuous variables) on significant results observed in the univariate analyses. In addition, the ability of sclerostin as a diagnostic marker of vascular damage and/or brain atrophy was also explored using ROC curves analysis. Considering that hypertension is a well-known factor involved in vascular damage, the sensitivity and specificity of sclerostin levels over the median in the diagnosis of vascular calcification or brain shrinkage (assessed by ROC curves) were tested both in the whole group and in the non-hypertensive group. All these analyses were performed with the SPSS program (Chicago, IL, USA).
The study protocol was approved by the local ethical committee of our Hospital (number $\frac{2017}{50}$) and conforms to the ethical guidelines of the 1975 Declaration of Helsinki. All the patients gave their written informed consent.
## Vascular calcifications
One hundred and forty-five patients ($48.47\%$) showed calcium deposits in the thorax plain X-ray films, a proportion by far higher than that observed among controls (χ2 = 16.31; $p \leq 0.001$, Table 1). Age was significantly higher among patients with vascular calcifications ($t = 6.57$; $p \leq 0.001$).
**TABLE 1**
| Unnamed: 0 | Vascular calcium deposits | Vascular calcium deposits.1 | Vascular calcium deposits.2 | Hypertension | Hypertension.1 | Hypertension.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | Yes (n = 143) | No (n = 152) | T (Z); p | Yes (n = 117) | No (n = 177) | T (Z); p |
| Age (years) | 61.62 ± 10.89 | 53.50 ± 10.34 | T = 6.57; p < 0.001 | 61.80 ± 11.50 | 54.69 ± 10.43 | T = 5,49; p < 0.001 |
| Daily ethanol ingestión (g) | 191.32 ± 133.40156 (100–225) | 200.58 ± 84.36200 (132–274) | Z = 2.18; p = 0.029 | 197.63 ± 93.44200 (120–250) | 193.46 ± 120.94180 (100–250) | Z = 0.96; NS |
| Duration of alcohol consumption (years) | 33 ± 1230 (26–40) | 29 ± 1130 (20–35) | Z = 3.03; p = 0.002 | 34 ± 1230 (25–40) | 29 ± 1130 (20–35) | Z = 3.21; p = 0.001 |
| Tobacco consumption | 103/141 | 96/152 | X2 = 2.84; NS | 77/115 | 123/177 | X2 = 0.11; NS |
| Packets/year index | 48.86 ± 33.2840 (30–60) | 41.70 ± 25.5536 (25–50) | Z = 1.75; p = 0.08 | 40.40 ± 27.7735 (25–43) | 48.15 ± 30.6240 (30–60) | Z = 2.14; p = 0.033 |
| Obesity (BMI > 30 kg/m2) | 26/113 | 15/128 | X2 = 4.65; p = 0.031 | 27/93 | 14/147 | X2 = 13.96; p < 0.001 |
| Liver cirrhosis | 57/143 | 67/152 | X2 = 0.38; NS | 47/117 | 75/177 | X2 = 0.07; NS |
| Prothrombin activity (%) | 78.80 ± 21.1185 (64–100) | 77.43 ± 21.4385 (57–100) | Z = 0.48; NS | 78.41 ± 20.1985 (65–98) | 77.65 ± 21.9984 (57–100) | Z = 0.04; NS |
| Serum bilirubin (mg/dl) | 2.34 ± 3.321.0 (1.0–2.3) | 2.96 ± 4.551.4 (1.0–3.4) | Z = 0.84; NS | 2.53 ± 3.601.1 (1.0 –2.30) | 2.75 ± 4.271.2 (1.0–3.3) | Z = 0.45; NS |
| Serum albumin (g/dL) | 3.48 ± 0.733.5 (3.1–3.9) | 3.53 ± 0.753.7 (2.9–4.1) | Z = 1.04; NS | 3.61 ± 0.773.7 (3.1–4.2) | 3.43 ± 0.723.5 (3.0–4.0) | Z = 1.66; NS |
| MCV (fL) | 99.73 ± 9.31100.0 (95.1–104.6) | 99.36 ± 8.28100.2 (94.6–104.6) | Z = 0.01; NS | 98.64 ± 8.4099,1 (93.8–103.6) | 100.35 ± 9.01100.6 (95.8–105.7) | Z = 2.00; p = 0.046 |
| Platelet count (x103/mm3) | 198505.59 ± 111806.46180000 (117000–267000) | 178532.89 ± 134912.77142000 (87000–234000) | Z = 2.36; p = 0.018 | 195341.88 ± 114401.23176000 (108000–243000) | 184634.46 ± 130445.28150000 (94500–243000) | Z = 1.33; NS |
| Serum GGT (U/L) | 237.36 ± 332.31113.0 (50.0–301.0) | 309.66 ± 473.17157.5 (77.3–362.5) | Z = 2.56; p = 0.011 | 289.59 ± 398.06128.0 (62.5–361.5) | 256.36 ± 408.38151.0 (66.5–319.0) | Z = 0.26; NS |
| Diabetes mellitus | 31/143 | 42/150 | X2 = 1.24; NS | 47/117 | 26/177 | X2 = 23.16; p < 0.001 |
| Liver steatosis | 57/133 | 71/131 | X2 = 2.96; NS | 51/109 | 77/154 | X2 = 0.15; NS |
| Serum sclerostin (pmol/L) | 40.41 ± 36.5134.01 (17.47–54.08) | 26.70 ± 22.7817.02 (12.09–35.71) | Z = 2.64; p = 0.008 | 46.81 ± 42.7738.32 (16.58 63.50) | 29.66 ± 23.8422.19 (14.20–39.53) | Z = 2.65; p = 0.008 |
| Total cholesterol (mg/dL) | 149.70 ± 49.49 | 164.62 ± 60.17 | T = 2.32; p = 0.021 | 160.09 ± 56.09 | 155.15 ± 55.16 | T = 0.74; NS |
| Triglycerides (mg/dL) | 104.59 ± 47.6093.0 (71.5–129.0) | 134.53 ± 147.11107 (76–148) | Z = 2.05; p = 0.040 | 118.16 ± 63.09102 (77–140) | 121.35 ± 134.73100 (67–143) | Z = 0.98; NS |
| Serum creatinine (mg/dL) | 0.92 ± 0.480.8 (0.62–1.08) | 0.88 ± 0.550.8 (0.6–1.0) | Z = 1.23; NS | 1.04 ± 0.600.9 (0.7–1.1) | 0.81 ± 0.430,7 (0.6–0.9) | Z = 4.40; p < 0.001 |
| Uric acid (mg/dL) | 5.18 ± 3.404.5 (3.5–6.4) | 5.13 ± 2.035.0 (3.7–6.6) | Z = 0.43; NS | 5.73 ± 2.395.3 (3.9–7.5) | 4.72 ± 1.954.4 (3.4–5.6) | Z = 3.14; p = 0.002 |
| HbA1C (%) | 5.99 ± 3.405.4 (5.0–5.9) | 5.85 ± 1.515.5 (5.0–6.2) | Z = 1.15; NS | 5.95 ± 1.425.5 (5.2–6.2) | 5.96 ± 3.865.3 (4.9–5.7) | Z = 2.38; p = 0.017 |
Daily ethanol ingestion was greater among patients with vascular calcifications ($Z = 2.18$; $$p \leq 0.029$$). Duration of alcohol consumption was related to vascular calcifications ($Z = 3.03$; $$p \leq 0.002$$) but this relationship was displaced by age in the multivariate analysis.
No association was observed between vascular calcium deposits and liver cirrhosis (χ2 = 0.38; $$p \leq 0.54$$; NS). We also failed to find any relationship between prothrombin activity, serum bilirubin, or serum albumin (as variables related to liver failure) and calcium deposits, but calcium deposits were more frequently observed among Child A patients (χ2 = 7.00; $$p \leq 0.03$$), especially when only cirrhotics were considered (χ2 = 9.00; $$p \leq 0.011$$). Platelet count ($Z = 2.34$; $$p \leq 0.018$$; possibly related to portal hypertension) and serum GGT ($Z = 2.56$; $$p \leq 0.011$$), possibly related to ethanol consumption, were higher in patients with vascular calcium deposits, but not MCV ($Z = 1.21$; $$p \leq 0.23$$; NS).
Vascular calcium deposits were associated with obesity (BMI > 30; χ2 = 4.65; $$p \leq 0.031$$); $23\%$ of patients with calcium deposits were obese vs. $11.71\%$ of patients without calcium deposits. BMI was significantly higher among patients with calcium deposits than those without calcium deposits ($t = 2.07$; $$p \leq 0.039$$), a difference that kept a marginally statistical significance when patients with ascities were excluded ($Z = 1.96$; $$p \leq 0.05$$). Serum creatinine levels were non-significantly higher among patients with calcium deposits (0.92 ± 0.48 mg/dl) than among patients without calcium deposits (0.88 ± 0.55 mg/dl; $t = 0.72$; $$p \leq 0.47$$), and no association was observed among calcium deposits and chronic kidney failure (CKF, defined as serum creatinine in stable conditions >1.40 mg/dl; χ2 = 1.45; $$p \leq 0.23$$).
One hundred and seventeen patients ($39.13\%$) were affected by hypertension, a proportion like to that observed in the control population (χ2 = 0.03; $$p \leq 0.86$$; NS, Table 1). A significant relationship was found between calcium in the X-ray plain film and hypertension (χ2 = 4.22; $$p \leq 0.04$$), although only 66 patients with vascular calcium deposits ($46.48\%$) were diagnosed with hypertension.
Age was significantly higher among patients with vascular calcifications ($t = 6.57$; $p \leq 0.001$) and also among patients with hypertension ($t = 5.49$; $p \leq 0.001$). Hypertension was strongly associated with older age ($t = 5.49$; $p \leq 0.0001$), but it was not related to cirrhosis (χ2 = 0.07; $$p \leq 0.80$$; NS) or liver steatosis (χ2 = 0.57; $$p \leq 0.45$$; NS). Duration of alcohol consumption was related to hypertension ($Z = 3.21$; $$p \leq 0.001$$), but this relationship was displaced by age when this variable was also introduced in a multivariate analysis. A strong association was observed between hypertension and diabetes ($t = 5.49$; $p \leq 0.001$), between calcium deposits and diabetes (χ2 = 23.16; $p \leq 0.001$) and hypertension and obesity (χ2 = 13.96; $p \leq 0.001$). A total of 24 patients showed creatinine values (after stabilization) over 1.40 mg/dl. A strong association was observed among hypertension and CKF (χ2 = 9.11; $$p \leq 0.003$$).
Two hundred and two patients ($68\%$) were also smokers. Tobacco consumption was not associated with hypertension (χ2 = 0.11; $$p \leq 0.74$$; NS) and showed a non-significant trend with the presence of vascular calcifications (χ2 = 2.85; $$p \leq 0.09$$). The index packets/year showed a trend to higher values among patients with vascular calcifications ($Z = 1.75$; $$p \leq 0.08$$) and was significantly higher among patients with hypertension ($Z = 2.14$; $$p \leq 0.033$$).
## Serum sclerostin and other factors related to the development of vascular calcifications
Serum sclerostin levels were slightly, non-significantly higher, among patients (Figure 2). Sclerostin levels were related to age (ρ = 0.30, $p \leq 0.001$), but no differences were observed among men and women ($Z = 0.25$; $$p \leq 0.80$$; NS). Patients with diabetes ($Z = 2.10$; $$p \leq 0.035$$) or hypertension ($Z = 2.65$; $$p \leq 0.008$$) showed higher sclerostin levels than patients without diabetes or hypertension (Figure 3). Cirrhotics showed non-significantly higher levels of sclerostin that non-cirrhotics ($Z = 1.51$; $$p \leq 0.13$$; NS).
**FIGURE 2:** *Serum sclerostin levels were slightly, non-significantly higher among patients (Z = 0.08; NS). °Represent the outliers and *represent the extreme outliers.* **FIGURE 3:** *Serum sclerostin levels were higher in hypertensive patients (Z = 2.65; p = 0.008). °Represent the outliers and *represent the extreme outliers.*
Significant relationships were recorded comparing vascular calcifications with total cholesterol ($Z = 2.04$; $$p \leq 0.041$$) and triglycerides ($Z = 2.05$; $$p \leq 0.04$$). Sclerostin levels were significantly higher ($Z = 2.64$; $$p \leq 0.008$$, Figure 4) among patients with vascular calcifications, but not serum creatinine ($Z = 1.23$; $$p \leq 0.22$$; NS) or uric acid ($Z = 0.43$; $$p \leq 0.66$$; NS).
**FIGURE 4:** *Serum sclerostin levels were higher in patients with vascular calcium deposits (Z = 2.64; p = 0.008). °Represent the outliers and *represent the extreme outliers.*
In summary, vascular calcifications are very frequent among alcohol misusers, and are related to age, diabetes, hypertension, sclerostin, cholesterol, triglycerides and daily ethanol intake, and marginally, to obesity assessed by BMI in patients without ascites. A logistic regression analysis including (as independent variables) sclerostin, total cholesterol, triglycerides (classified as dichotomic variables according to medians), CKF, cirrhosis, diabetes, and hypertension showed that sclerostin was the only variable independently related to the presence of vascular calcifications ($$p \leq 0.022$$; odds ratio for calcifications if sclerostin is over the median = 2.65 ($95\%$ CI = 1.14–6.13). However, this relationship was displaced by age if this variable was also included.
The ability of sclerostin to diagnose vascular calcifications can be also observed with the ROC curve with an AUC of 0.676 ± 0.052 ($95\%$ CI 0.574–0.777; $$p \leq 0.002$$, Figure 5). As previously commented, only 66 of the patients with vascular calcifications were also affected by hypertension. Considering only the non-hypertensive patients, the relationship of sclerostin with vascular calcifications was even more marked, as shown in Figure 6, with an AUC of 0.702 ($95\%$ CI = 0.580–0.825), a standard error of 0.063 and a p-value of 0.006. Therefore, among alcoholics, sclerostin constitutes a risk factor for the development of vascular calcification, that, in our study, seems to be more important than the classic risk factors hypertension, kidney failure, cholesterol, or triglycerides.
**FIGURE 5:** *Relationship of sclerostin with vascular calcifications. ROC curve with an AUC of 0.676 ± 0.052 (95% CI 0.574–0.777); p = 0.002.* **FIGURE 6:** *Relationship of sclerostin with vascular calcifications in non-hypertensive patients. ROC curve with an AUC = 0.702 ± 0.063 (95% CI = 0.580–0.825), p = 0.006.*
## Brain atrophy
An expert radiologist evaluated the vast majority [296] CT studies and classified the patients as affected by cortical atrophy [218] or cerebellar atrophy [210], or not. As expected, CT indices were all significantly different among patients with cortical or cerebellar atrophy (Table 2) apart from Evans index, that was similar among patients with cerebellar atrophy and patients without cerebellar atrophy. A close association was observed among the presence of cerebellar atrophy and cortical atrophy (χ2 = 155; $p \leq 0.001$), although 12 patients with cerebellar atrophy ($5.71\%$) did not show cortical atrophy, and 20 patients with cortical atrophy ($9.17\%$) did not show cerebellar atrophy.
**TABLE 2**
| Unnamed: 0 | Cortical atrophy | Cortical atrophy.1 | Cortical atrophy.2 | Cerebellar atrophy | Cerebellar atrophy.1 | Cerebellar atrophy.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | Yes (n = 218) | No (n = 78) | T; p | Yes (n = 210) | No (n = 86) | T; p |
| Bicaudate index | 0.18 ± 0.04 | 0.15 ± 0.03 | T = 6.50; p < 0.001 | 0.18 ± 0.04 | 0.16 ± 0.04 | T = 4.63; p < 0.001 |
| Bifrontal index | 0.35 ± 0.050.36 (0.33–0.38) | 0.33 ± 0.050.33 (0.30–0.35) | Z = 4.89; p < 0.001 | 0.35 ± 0.050.35 (0.33–0.38) | 0.33 ± 0.060.33 (0.30–0.36) | Z = 3.53; p < 0.001 |
| Evans index | 0.30 ± 0.040.30 (0.28–0.33) | 0.29 ± 0.060.29 (0.26–0.30) | Z = 4.24; p < 0.001 | 0.30 ± 0.040.30 (0.28–0.33) | 0.29 ± 0.060.29 (0.26–0.31) | Z = 3.17; p = 0.002 |
| Ventricular index | 0.54 ± 0.100.53 (0.47–0.60) | 0.48 ± 0.110.46 (0.41–0.54) | Z = 4.45; p < 0.001 | 0.53 ± 0.100.52 (0.47–0.60) | 0.50 ± 0.110.48 (0.42–0.57) | Z = 3.31; p = 0.001 |
| Cella media index | 4.02 ± 0.933.94 (3.49–4.42) | 4.58 ± 1.094.52 (3.89–5.24) | Z = 4.42; p < 0.001 | 4.00 ± 0.913.93 (3.50–4.39) | 4.60 ± 1.124.54 (3.90–5.24) | Z = 4.99; p < 0.001 |
| Huckmann index | 0.54 ± 0.08 | 0.48 ± 0.07 | T = 5.65; p < 0.001 | 0.54 ± 0.07 | 0.49 ± 0.09 | T = 3.97; p < 0.001 |
Age was significantly related to all the indices (Huckmann ρ = 0.34; Evans ρ = 0.29; Bifrontal ρ = 0.34, bicaudate ρ = 0.28; $p \leq 0.001$ in all the cases), ventricular ρ = 0.15, and cella media (ρ = 0.16; $p \leq 0.01$ in both cases). Duration of addiction was also significantly related to all the indices (Huckmann ρ = 0.29; Evans ρ = 0.23; Bifrontal ρ = 0.27, bicaudate ρ = 0.22; $p \leq 0.001$ in all the cases), and ventricular (ρ = 0.12) and cella media (ρ = 0.15; $p \leq 0.033$ in both cases), but these relationships were displaced by age when multivariate analyses were performed. No relationships were observed among CT indices and daily ethanol consumption.
No relationships were observed between obesity and brain CT indices, cirrhosis, and brain CT indices, or steatosis and brain CT indices. We also failed to find any relationship among CT indices and liver function assessed by Child-Pugh’s score. Patients with hypertension showed a greater atrophy estimated by Huckmann index ($T = 2.02$; $$p \leq 0.044$$, Figure 7). No differences were observed when the indices were compared among patients with or without CKF, besides cella media ($Z = 1.97$; $$p \leq 0.049$$), and no relationships were observed between CT indices and creatinine, besides a direct one between serum creatinine and ventricular index (ρ = 0.13; $$p \leq 0.03$$). All the indices besides Evan’s index were altered in diabetics (ventricular index $Z = 3.16$; $$p \leq 0.002$$; bicaudate index $Z = 3.47$; $$p \leq 0.001$$; bifrontal index $Z = 2.07$; $$p \leq 0.038$$; cella media $Z = 1.98$; $$p \leq 0.048$$; Huckmann index $Z = 2.99$; $$p \leq 0.003$$). All these differences were displaced by age (logistic regression analysis including diabetes, vascular calcifications, CKF, hypertension, and age, cholesterol and triglycerides, as dichotomic variables). However, diabetes still showed an independent relationship with Huckmann index, but in the second place, after age [OR = 2.28 (1.29–4.06; $$p \leq 0.005$$)].
**FIGURE 7:** *Patients with hypertension showed a greater atrophy estimated by Huckmann index (T = 2.02; p = 0.044). °Represent the outliers and *represent the extreme outliers.*
In Table 3 we show the differences in CT indices in relation with vascular calcium deposits. Calcium deposits were significantly related to Bifrontal index ($Z = 2.20$; $$p \leq 0.028$$, Figure 8) and Evans index ($Z = 2.25$; $$p \leq 0.025$$, Figure 9), although these relationships were displaced by age in both cases when indices and age were classified as dichotomic variable and logistic regression analyses were performed comparing each of the CT indices as dependent variables with age and calcium deposits.
Sclerostin levels showed a significant correlation with Huckmann index (ρ = 0.204; $$p \leq 0.024$$). In addition, cella media index was significantly different when patients with sclerostin values over the median were compared with patients with sclerostin values below the median ($Z = 2.43$; $$p \leq 0.015$$). The relationship between sclerostin and cella media index was also evident when a ROC curve was depicted in order to analyze sensitivity and specificity of sclerostin to detect patients with cella media values over the median, with an AUC of 0.666 ± 0.050 ($95\%$ CI = 0.568–0.764; $$p \leq 0.002$$, Figure 10).
**FIGURE 10:** *Relationship between sclerostin and cella media values over the median. ROC curve with an AUC = 0.666 ± 0.050 (95% CI = 0.568–0.764); p = 0.002.*
A stepwise logistic regression analysis comparing cella media values over the median (as the dependent variable) with sclerostin, cholesterol, triglycerides, daily ethanol consumption (as dichotomic variables according to median values), vascular calcifications, hypertension, and CKF, showed that sclerostin (over the median) was the only variable selected (odds ratio = 2.5, $95\%$ confidence interval = 1.16–5.39; $$p \leq 0.019$$), and the same happened when the variable age (dichotomized) was also introduced. Considering only patients without vascular calcification, the ROC curve comparing sclerostin with the cella media index over or below the median was not significant at all (AUC = 0.518 ± 0.105; $$p \leq 0.86$$), in contrast with what was observed when only patients with vascular calcifications were included (AUC = 0.708 ± 0.057; $95\%$ CI = 0.596–0.819; $$p \leq 0.001$$, Figure 11). Therefore, the relationship of sclerostin with brain atrophy is especially marked in patients with vascular calcifications.
**FIGURE 11:** *Relationship between sclerostin and cella media values over the median in patients with vascular calcifications. ROC curve with an AUC = 0.708 ± 0.057; (95% CI = 0.596–0.819); p = 0.001.*
In hypertensive patients, the ROC curve comparing sclerostin and cella media index was not statistically significant (AUC = 0.588 ± 0.095; $95\%$ CI = 0.401–0.775; $$p \leq 0.31$$), but considering only non-hypertensive patients, ROC curve analysis yielded an AUC even greater [0.743 ± 0.060 ($95\%$ CI = 0.626–0.861; $$p \leq 0.001$$)] than that observed in the whole group, as shown in the Figure 12. Therefore, when hypertension is present, the relationship of sclerostin with brain atrophy is less marked.
**FIGURE 12:** *Relationship between sclerostin and cella media values over the median in non-hypertensive patients. ROC curve with an AUC = 0.743 ± 0.060 (95% CI = 0.626–0.861); p = 0.001.*
Lastly, the ability of sclerostin in the diagnosis of brain atrophy (cella media index) is even more marked in non-hypertensive patients with vessel wall calcium deposits. AUC reaches 0.802 ± 0.064; ($95\%$ CI = 0.675–0.928; $p \leq 0.001$; Figure 13).
**FIGURE 13:** *Relationship between sclerostin and cella media values over the median in non-hypertensive patients with vessel wall calcium deposits. AUC = 0.802 ± 0.064; (95% CI = 0.675–0.928); p < 0.001.*
Also, a stepwise logistic regression analysis within no-hypertensive patients, comparing cella media values over the median (as the dependent variable) with sclerostin, cholesterol, triglycerides, daily ethanol consumption (as dichotomic variables according to median values), vascular calcifications, and CKF, showed that sclerostin (over the median) was the only variable selected (odds ratio = 2.93; $95\%$ CI = 1.08–7.94; $$p \leq 0.035$$), and the same happened when the variable age (dichotomized) and/or diabetes were also introduced.
Therefore, it seems that sclerostin also constitutes a risk factor for subcortical brain atrophy (assessed by cella media index), both in the global alcoholic population and in non-hypertensive patients, independent of age and other common risk factors such as vascular calcifications, cholesterol, triglycerides, or diabetes.
## Discussion
In this study we estimated the prevalence of vascular calcifications in the aortic arch as a marker of vascular injury, and the relationship of this feature with brain atrophy in alcoholics. We also pursued to analyze the relationship of an emerging new vascular risk factor, namely sclerostin with these alterations.
We found that the prevalence of vascular calcifications is very high among alcoholic patients, and vascular calcifications are significantly related to brain shrinkage in these patients. Although in an observational study like this we cannot establish a causal link between both features (brain shrinkage and calcium vessel wall deposits), our results suggest that vascular damage may be a contributory non-functional, but organic, factor involved in the brain alterations observed in these patients.
The relationship of vascular calcifications with chronic ethanol intake has been pointed out in previous studies. For instance, the relationship between coronary artery calcification and ethanol consumption has been recorded by several authors (Pletcher et al., 2005; Yun et al., 2017), who also report an increased prevalence of hypertension in excessive drinkers, in accordance with former observations (Saunders, 1987; van Leer et al., 1994; Fuchs et al., 2001). However, in this study, prevalence of hypertension among alcoholics is similar to that of the controls, and also similar to that reported for the general population of our geographical environment (Cabrera De León et al., 2008; Zubeldia Lauzurica et al., 2016). The markedly higher prevalence of vascular calcification among alcoholics, despite a similar prevalence of hypertension in patients and controls, strongly suggests that factors other than hypertension should play a role in the calcification of vessel walls.
As previously commented. ethanol may exert direct effects on smooth muscle cells of the vessel walls, promoting vascular smooth muscle cells calcification and transition of these cells to osteoblastic-like cells (Oros et al., 2012). Therefore, radiologically detectable vascular calcifications may be observed in excessive drinkers, independent of the concomitant presence of hypertension. Oros’ research also lends support to the findings of this study, in which calcium deposits were recorded in 145 ($48\%$) excessive drinkers, but only 66 out of 145 patients with vascular calcifications were also affected by hypertension, suggesting that other factors also play a pathogenetic role.
In this study we tested the relationship of sclerostin both with vascular damage and brain atrophy in the whole sample, considering together hypertensive and non-hypertensive patients, and also in the subgroups of hypertensive and non-hypertensive patients, these last lacking a classic major risk factor for vascular damage. As commented in the introduction of this study, sclerostin may favor calcification of vessel walls (Zhu et al., 2011), atherosclerotic plaques (Leto et al., 2019), and aortic valves (Koos et al., 2013). We found that sclerostin levels were higher among patients with vascular calcifications. Therefore, our results are in accordance with these observations. The fact that sclerostin is the sole independent factor related to vascular calcification in a logistic regression analysis, being displaced only by age, is a striking result. Of similar importance is the finding observed in non-hypertensive patients, in whom the relation of sclerostin with vascular calcification is even stronger. This may suggest that, at least in alcoholics, sclerostin may be a factor related to vascular damage, independent on hypertension and/or kidney failure (a suggestion also derived from the reported results of the logistic regression analysis).
Importantly, sclerostin levels were not only related to vascular calcification in our study, but also to a possible consequence of vascular calcification, namely brain atrophy. In addition to the already commented many factors involved in the development of brain atrophy in alcoholics, the role of altered vascular supply in alcoholics has been also pointed out, not only by the commented relationship between hypertension and alcoholism, but also as an additional direct functional effect of ethanol. Since several decades, ethanol is known to cause cerebral arterial spasm, bleeding, and alterations of blood flow in certain areas of the brain (Altura and Altura, 1984), such as temporal, frontal and occipital cortices, corpus callosum, and basal ganglia. These effects, together with hypertension, may explain the association of ethanol consumption and stroke (Schuckit, 2009), but also, possibly, the association of ethanol with brain atrophy and cognitive impairment. In this sense, we found a clear-cut relationship between calcium deposits in the thoracic vessel walls and brain atrophy, suggesting a role of vascular lesions on ethanol-mediated brain alterations. As expected, age was the main factor responsible for vascular calcifications and brain atrophy, but it is important to remark the independent relationship of sclerostin with vessel wall calcification and brain alterations (even displacing age), results that are in accordance with the current consideration of sclerostin as a major vascular risk factor. Strikingly, the best relationship with brain atrophy (assessed by cella media index) was observed among non-hypertensive patients with vascular calcifications. Taken together the results of the ROC analyses and the logistic regression analyses, it could be hypothesized that, in absence of hypertension as a major risk factor, the role of sclerostin on brain atrophy gains prominence probably by inducing vessel wall calcification, perhaps sharing or potentiating the effects of ethanol. This potentiating effect is a speculative possibility supported by the observation that the control group of non-drinkers show a very low proportion of vascular calcifications despite a similar prevalence of hypertension and similar sclerostin levels than the patients. Future research devoted to disentangling these possible connections is needed.
Brain damage in alcoholics affects both grey matter and white matter (de La Monte and Kril, 2014), with neuronal shrinkage and loss of dendritic spines, as well as axonal damage, atrophy, and ventricular dilatation. One of the most striking features of alcohol-mediated brain shrinkage is that it in many patients it is almost fully reversible with alcohol withdrawal. Therefore, transient, reversible, metabolic alterations, timely related to heavy alcohol consumption, probably play a major role in brain atrophy in these patients. Perhaps sclerostin is one of these factors, its deleterious effects on vascular structure being triggered by heavy alcohol consumption. In any case, the findings of this study do not support any role of ethanol (at least, in heavy consumers) as a “protective” vascular factor.
One limitation of this study resides in the fact that it was devoted to analyzing radiographically assessed alterations of brain and vascular lesions in alcoholics, but it did not include a functional evaluation. However, previous reports have shown a definite relationship between brain atrophy and/or ventricular dilatation and cognitive impairment in these patients, so brain atrophy can be considered as the structural alteration underlying the functional derangement (de La Monte and Kril, 2014). Many confounding factors, not assessed in this study, such as altered levels of vitamins, micronutrients, and aspects related to style of life and education may be surely involved in the brain alterations of alcoholic patients (as well as in non-alcoholics). Possibly, their correction may contribute to the reversible nature of brain damage and cognitive impairment of alcoholics after drinking cessation, although this study strongly suggest that potentially irreversible vascular lesions may play a contributory role in brain atrophy; and also, perhaps, that ethanol may interact with molecules involved in vessel wall metabolism, such as sclerostin, potentiating its deleterious effects.
## Conclusion
We conclude that prevalence of vascular calcification in alcoholics is very high, despite the relatively young age of the included individuals. Among nearly 300 individuals, vascular calcium deposits were strongly related to brain atrophy. Interestingly in addition to age, sclerostin was strongly related to vascular calcifications, and it was also independently related to brain atrophy, underscoring the role of osteokines on vascular disorders, at least in excessive drinkers.
## Data availability statement
The original contributions presented in this 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 Institutional Ethics Committee of Hospital Universitario de Canarias, and all subjects provided informed written consent (Approval number: 2017_50). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
CM-G and EG-R contributed to the conceptualization, writing—review and editing, formal analysis, and writing—original draft. AG-R, CF-R, EM-P, MS-P, JA-N, and MR-G contributed to the investigation. PA-G contributed to the methodology. 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. Altura B. M., Altura B. T.. **Alcohol, the cerebral circulation and strokes.**. (1984) **1** 325-331. DOI: 10.1016/0741-8329(84)90056-9
2. Cabrera De León A., Pérez M. D. C. R., González D. A., Coello S. D., Jaime A. A., Díaz B. B.. **[Presentation of the “CDC de Canarias” cohort: objectives, design and preliminary results].**. (2008) **82** 519-534. DOI: 10.1590/S1135-57272008000500007
3. Cao Y., Willett W. C., Rimm E. B., Stampfer M. J., Giovannucci E. L.. **Light to moderate intake of alcohol, drinking patterns, and risk of cancer: results from two prospective US cohort studies.**. (2015) **351**. DOI: 10.1136/BMJ.H4238
4. Catalano A., Bellone F., Morabito N., Corica F.. **Sclerostin and vascular pathophysiology.**. (2020) **21** 1-14. DOI: 10.3390/IJMS21134779
5. Child C. G., Turcotte J. G.. **Surgery and portal hypertension.**. (1964) **1** 1-85. PMID: 4950264
6. Crews F. T., Nixon K.. **Mechanisms of neurodegeneration and regeneration in alcoholism.**. (2009) **44** 115-127. DOI: 10.1093/ALCALC/AGN079
7. de La Monte S. M., Kril J. J.. **Human alcohol-related neuropathology.**. (2014) **127** 71-90. DOI: 10.1007/S00401-013-1233-3
8. Ehrlich D., Pirchl M., Humpel C.. **Effects of long-term moderate ethanol and cholesterol on cognition, cholinergic neurons, inflammation, and vascular impairment in rats.**. (2012) **205** 154-166. DOI: 10.1016/J.NEUROSCIENCE.2011.12.054
9. Fuchs F. D., Chambless L. E., Whelton P. K., Nieto F. J., Heiss G.. **Alcohol consumption and the incidence of hypertension: the atherosclerosis risk in communities study.**. (2001) **37** 1242-1250. DOI: 10.1161/01.HYP.37.5.1242
10. Fuchs F. D., Fuchs S. C.. **The effect of alcohol on blood pressure and hypertension.**. (2021) **23**. DOI: 10.1007/S11906-021-01160-7
11. González-Reimers E., Fernández-Rodríguez C. M., Martín-González M. C., Hernández-Betancor I., Abreu-González P., de la Vega-Prieto M. J.. **Antioxidant vitamins and brain dysfunction in alcoholics.**. (2014) **49** 45-50. DOI: 10.1093/alcalc/agt150
12. González-Reimers E., Martín-González C., de la vega-Prieto M. J., Pelazas-González R., Fernández-Rodríguez C., López-Prieto J.. **Serum sclerostin in alcoholics: a pilot study.**. (2013) **48** 278-282. DOI: 10.1093/alcalc/ags136
13. Harris T. C., de Rooij R., Kuhl E.. **The shrinking brain: cerebral atrophy following traumatic brain injury.**. (2019) **47** 1941-1959. DOI: 10.1007/S10439-018-02148-2
14. Jadzic J., Milovanovic P. D., Cvetkovic D., Zivkovic V., Nikolic S., Tomanovic N.. **The altered osteocytic expression of connexin 43 and sclerostin in human cadaveric donors with alcoholic liver cirrhosis: potential treatment targets.**. (2022) **240** 1162-1173. DOI: 10.1111/JOA.13621
15. Kirk B., Feehan J., Lombardi G., Duque G.. **Muscle, bone, and fat crosstalk: the biological role of myokines, osteokines, and adipokines.**. (2020) **18** 388-400. DOI: 10.1007/S11914-020-00599-Y
16. Koos R., Brandenburg V., Mahnken A. H., Schneider R., Dohmen G., Autschbach R.. **Sclerostin as a potential novel biomarker for aortic valve calcification: an**. (2013) **22** 317-325. PMID: 24151757
17. Leto G., D’Onofrio L., Lucantoni F., Zampetti S., Campagna G., Foffi C.. **Sclerostin is expressed in the atherosclerotic plaques of patients who undergoing carotid endarterectomy.**. (2019) **35**. DOI: 10.1002/DMRR.3069
18. Li M., Zhou H., Yang M., Xing C.. **Relationship between serum sclerostin, vascular sclerostin expression and vascular calcification assessed by different methods in ESRD patients eligible for renal transplantation: a cross-sectional study.**. (2019) **51** 311-323. DOI: 10.1007/S11255-018-2033-4
19. Marmot M. G., Elliott P., Shipley M. J., Dyer A. R., Ueshima H. U., Beevers D. G.. **Alcohol and blood pressure: the INTERSALT study.**. (1994) **308**. DOI: 10.1136/BMJ.308.6939.1263
20. Martín González C., Fernández Rodríguez C. M., Abreu González P., García Rodríguez A., Alvisa Negrín J. C., Cabañas Perales E.. **Sclerostin in excessive drinkers: Relationships with liver function and body composition**. (2022) **14**. DOI: 10.3390/nu14132574
21. Oros M., Zavaczki E., Vadasz C., Jeney V., Tosaki A., Lekli I.. **Ethanol increases phosphate-mediated mineralization and osteoblastic transformation of vascular smooth muscle cells.**. (2012) **16** 2219-2226. DOI: 10.1111/J.1582-4934.2012.01533.X
22. Pelletier S., Confavreux C. B., Haesebaert J., Guebre-Egziabher F., Bacchetta J., Carlier M. C.. **Serum sclerostin: the missing link in the bone-vessel cross-talk in hemodialysis patients?**. (2015) **26** 2165-2174. DOI: 10.1007/s00198-015-3127-9
23. Peng B., Yang Q., Joshi R. B., Liu Y., Akbar M., Song B. J.. **Role of alcohol drinking in Alzheimer’s disease, Parkinson’s disease, and amyotrophic lateral sclerosis.**. (2020) **21**. DOI: 10.3390/IJMS21072316
24. Pfefferbaum A., Sullivan E. V., Mathalon D. H., Shear P. K., Rosenbloom M. J., Lim K. O.. **Longitudinal changes in magnetic resonance imaging brain volumes in abstinent and relapsed alcoholics.**. (1995) **19** 1177-1191. DOI: 10.1111/J.1530-0277.1995.TB01598.X
25. Pletcher M. J., Varosy P., Kiefe C., Lewis C. E., Sidney S., Hulley S. B.. **Alcohol consumption, binge drinking, and early coronary calcification: findings from the Coronary Artery Risk Development in Young Adults (CARDIA) study.**. (2005) **161** 423-433. DOI: 10.1093/AJE/KWI062
26. Pugh R. N. H., Murray-Lyon I. M., Dawson J. L., Pietroni M. C., Williams R.. **Transection of the oesophagus for bleeding oesophageal varices.**. (1973) **60** 646-649. DOI: 10.1002/BJS.1800600817
27. Qin L., Crews F. T.. **Chronic ethanol increases systemic TLR3 agonist-induced neuroinflammation and neurodegeneration.**. (2012) **9**. DOI: 10.1186/1742-2094-9-130
28. Rehm J., Hasan O. S. M., Black S. E., Shield K. D., Schwarzinger M.. **Alcohol use and dementia: a systematic scoping review.**. (2019) **11**. DOI: 10.1186/S13195-018-0453-0
29. Reid M. C., Fiellin D. A., O’Connor P. G.. **Hazardous and harmful alcohol consumption in primary care.**. (1999) **159** 1681-1689. DOI: 10.1001/ARCHINTE.159.15.1681
30. Rennenberg R. J. M. W., Kessels A. G. H., Schurgers L. J., van Engelshoven J. M. A., de Leeuw P. W., Kroon A. A.. **Vascular calcifications as a marker of increased cardiovascular risk: a meta-analysis.**. (2009) **5** 185-197. DOI: 10.2147/VHRM.S4822
31. Ridley N. J., Draper B., Withall A.. **Alcohol-related dementia: an update of the evidence.**. (2013) **5**. DOI: 10.1186/ALZRT157
32. Romero-Acevedo L., González-Reimers E., Martín-González M. C., González-Díaz A., Quintero-Platt G., Reyes-Suárez P.. **Handgrip strength and lean mass are independently related to brain atrophy among alcoholics.**. (2019) **38** 1439-1446. DOI: 10.1016/j.clnu.2018.06.965
33. Saunders J. B.. **Alcohol: an important cause of hypertension.**. (1987) **294** 1045-1046. DOI: 10.1136/BMJ.294.6579.1045
34. Schuckit M. A.. **Alcohol-use disorders.**. (2009) **373** 492-501. DOI: 10.1016/S0140-6736(09)60009-X
35. Shi J., Yang Y., Cheng A., Xu G., He F.. **Metabolism of vascular smooth muscle cells in vascular diseases.**. (2020) **319** H613-H631. DOI: 10.1152/AJPHEART.00220.2020
36. Topiwala A., Ebmeier K. P.. **Effects of drinking on late-life brain and cognition.**. (2018) **21** 12-15. DOI: 10.1136/eb-2017-102820
37. van Leer E. M., Seidell J. C., Kromhout D.. **Differences in the association between alcohol consumption and blood pressure by age, gender, and smoking.**. (1994) **5** 576-582. DOI: 10.1097/00001648-199411000-00004
38. Wakolbinger R., Muschitz C., Wallwitz J., Bodlaj G., Feichtinger X., Schanda J. E.. **Serum levels of sclerostin reflect altered bone microarchitecture in patients with hepatic cirrhosis.**. (2020) **132** 19-26. DOI: 10.1007/S00508-019-01595-8
39. Wood A. M., Kaptoge S., Butterworth A., Nietert P. J., Warnakula S., Bolton T.. **Risk thresholds for alcohol consumption: combined analysis of individual-participant data for 599 912 current drinkers in 83 prospective studies.**. (2018) **391** 1513-1523. DOI: 10.1016/S0140-6736(18)30134-X
40. Yun K. E., Chang Y., Yun S. C., Smith G. D., Ryu S., Cho S.. **Alcohol and coronary artery calcification: an investigation using alcohol flushing as an instrumental variable.**. (2017) **46** 950-962. DOI: 10.1093/IJE/DYW237
41. Zhu D., Mackenzie N. C. W., Millán J. L., Farquharson C., MacRae V. E.. **The appearance and modulation of osteocyte marker expression during calcification of vascular smooth muscle cells.**. (2011) **6**. DOI: 10.1371/JOURNAL.PONE.0019595
42. Zubeldia Lauzurica L., Quiles Izquierdo J., Mañes Vinuesa J., Redón Más J.. **[Prevalence of hypertension and associated factors in population aged 16 to 90 years old in valencia region, Spain].**. (2016) **90**. PMID: 27032998
|
---
title: Tranexamic acid is associated with improved hemostasis in elderly patients
undergoing coronary-artery surgeries in a retrospective cohort study
authors:
- Enshi Wang
- Yang Wang
- Yuan Li
- Shengshou Hu
- Su Yuan
journal: Frontiers in Surgery
year: 2023
pmcid: PMC9989169
doi: 10.3389/fsurg.2023.1117974
license: CC BY 4.0
---
# Tranexamic acid is associated with improved hemostasis in elderly patients undergoing coronary-artery surgeries in a retrospective cohort study
## Abstract
### Background
More elderly patients undergo coronary artery bypass surgery (CABG) than younger patients. Whether tranexamic acid (TA) is still effective and safe in elderly patients undergoing CABG surgeries is still unclear.
### Methods
In this study, a cohort of 7,224 patients ≥70 years undergoing CABG surgery were included. Patients were categorized into the no TA group, TA group, high-dose group, and low-dose group according whether TA was administered and the dose administered. The primary endpoint was blood loss and blood transfusion after CABG. The secondary endpoints were thromboembolic events and in-hospital death.
### Results
The blood loss at 24 and 48 h and the total blood loss after surgery in patients in the TA group were 90, 90, and 190 ml less than those in the no-TA group, respectively ($p \leq 0.0001$). The total blood transfusion was reduced 0.38-fold with TA administration compared to that without TA (OR = 0.62, $95\%$ CI 0.56–0.68, $p \leq 0.0001$). Blood component transfusion was also reduced. High-dose TA administration reduced the blood loss by 20 ml 24 h after surgery ($$p \leq 0.032$$) but had no relationship with the blood transfusion. TA increased the risk of perioperative myocardial infarction (PMI) by 1.62-fold [$$p \leq 0.003$$, OR = 1.62, $95\%$ CI (1.18–2.22)] but reduced the hospital stay time in patients who were administered TA compared to that of patients who did not receive TA ($$p \leq 0.026$$).
### Conclusion
We revealed that elderly patients undergoing CABG surgeries had better hemostasis after TA administration but increased the risk of PMI. High-dose TA was effective and safe compared with low-dose TA administration in elderly patients undergoing CABG surgery.
## Introduction
With economic and medical development, the aging of society will inevitably lead to an increasing proportion of patients over 70 years who undergo cardiac surgery [1]. As the population ages, the incidence of age-related complications (including diabetes, peripheral vascular diseases, kidney diseases and cardiovascular diseases) has increased. The number of potential candidates for geriatric surgery is increasing, as approximately a quarter of people over the age of 75 develop cardiovascular disease, and more than half of all heart surgeries are performed in this age group [2]. However, the long-term and short-term mortality and morbidity in elderly patients were only acceptable if accurate selection, a multifactorial risk evaluation, was adopted and nonelective operations were not performed [3]. Compared to younger patients, elderly patients had a higher re-exploration rate in cardiac surgery [4, 5], which was associated with a higher blood transfusion rate. Based on the natural properties in older patients, bleeding and blood transfusion were increased during cardiac surgeries and thereby were combined with more complications, more deaths and greater costs [6, 7].
To avoid bleeding and blood comorbidities in the aged population, antifibrinolytic drugs, such as tranexamic acid (TA), have been used during cardiac surgery to maintain hemostasis [8]. As a double-edged sword, TA also has a risk of thromboembolic complications after cardiac surgery [9]. Nevertheless, controversy regarding TA thrombosis during cardiac surgery remains (9–13). In elderly patients, the physiological reserve decreases, and the reduction in coronary flow reserve and the progression of atherosclerosis lead to a higher coronary artery disease incidence [2]. In addition, during coronary artery bypass grafting (CABG), the activation of the hemostatic system, which showed more obvious fibrinolysis and platelet activation, increased in elderly patients compared with young patients [14]. Furthermore, TA more easily accumulated in older patients due to the higher renal dysfunction complications in the perioperative period, and $90\%$ of TA was unchanged and eliminated by the patients’ kidneys [2, 15, 16]. All the above factors pose a higher thromboembolic risk in elderly patients undergoing CABG with TA administration. Therefore, the blood management and safety issues of TA in the older population undergoing CABG were taken into consideration in this study.
## Patient information
From January 2009 to December 2019, 7,526 patients ≥70 years undergoing CABG surgery were considered for inclusion in this study. Among them, 23 patients were excluded due to enrollment in RCT clinical trials, and 279 patients were excluded due to missing values. The remaining 7,224 patients ($95.98\%$) were ultimately included in this study (Figure 1). According to the administration of TA, the elderly patients were divided into two groups. The TA group consisted of 4,963 patients, and the no-TA group included 2,261 patients. After propensity score matching (PSM), the TA group included 1,910 patients, and the non-TA group included 1,910 patients (Figure 1, Table 1). In the TA group, a total of 4,963 patients were divided into the high-dose group and the low-dose group with a cutoff value of 50 mg/kg according to previous studies (17–23). Patients who received a TA dose greater than 50 mg/kg were categorized as the high-dose TA group, and patients who were administered a TA dose less than 50 mg/kg were defined as the low-dose TA group. In this study, the high-dose TA group included 2,887 patients, and the low-dose TA group included 2,076 patients. After PSM, there were 1,396 patients in the high-dose and low-dose groups (Figure 1, Table 2). The study was carried out according to the guidelines of the Declaration of Helsinki 1964 and its subsequent amendments. The Medical Ethics Committee of Fuwai Hospital approved the protocol. The requirement for informed consent was waived by the Medical Ethics Committee due to the retrospective nature of the study.
**Figure 1:** *Flowchart of elderly patients undergoing CABG surgery with or without TA administration. This study included 7,526 patients older than 70 years treated by CABG surgery. Patients enrolled in RCT studies or with missing values were removed from this study. Finally, a total of 7,224 elderly patients were selected for this study. Among them, the TA group had 4,963 patients, and the no-TA group had 2,261 patients. In the TA group, 2,887 patients were administered a TA dose ≥50 mg/kg, and the remaining 2,076 patients were administered a TA dose <50 mg/kg. The outcomes between the TA and no-TA groups or the high-dose and low-dose groups were compared by PSM and subsequent statistical analysis. CABG, coronary artery bypass grafts; RCT, random control trial; TA, tranexamic acid; PSM, propensity score matching.* TABLE_PLACEHOLDER:Table 1 TABLE_PLACEHOLDER:Table 2
## Operation details
Colleagues collected all patients’ data from the electronic hospital records in the information center of our hospital. All patients in this study underwent CABG surgeries. Among them, isolated CABG accounted for $85.8\%$ of patients in the TA group and $92.2\%$ of patients in the no-TA group (Supplementary Table S1). The remaining CABG procedures were carried out with aneurysm surgeries, valve surgery, or aortic or arch surgery. Elective surgery accounted for $97.0\%$ of surgeries in the TA group and $95.4\%$ of surgeries in the no-TA group (Supplementary Table S1). On-pump surgeries were performed on $58.5\%$ of patients in the TA group and $36.2\%$ of patients in the no-TA group (Supplementary Table S1). Isolated CABGs accounted for $81.8\%$ of patients in the high-dose TA group and $97.2\%$ of patients in the low-dose TA group. In the high-dose TA group, $65.5\%$ of surgeries were on-pump surgeries, while in the low-dose TA group, $48.8\%$ of surgeries were on-pump surgeries (Supplementary Table S2).
We defined high-risk surgery as emergent CABG surgery, CABG with a history of previous cardiac surgery, CABG plus aortic or arch operation, or CABG plus valve surgery. Open chamber surgeries included CABG plus aneurysm resection, aortic operation or valve surgery. Perioperative myocardial infarction (PMI) is diagnosed by an isolated elevation of creatine kinase myocardial isoenzyme (CK-MB) to ≥10 × 99th percentile upper reference limit (URL) or cardiac troponin (cTn) (I or T) to ≥70 × URL during the first 48 h following CABG surgery with or without ECG or imaging changes of MI [24, 25].
## Hemostasis during surgery
TA was administered after anesthesia induction. The administration of TA was under the consideration of anesthesiologists. TA was given at a dose of 54.83 mg/kg (median) with an interquartile range (IQR) from 42.73 to 72.46 mg/kg in elderly patients. The median TA dose was 70.42 mg/kg (IQR 60.24–80.64 mg/kg) in patients in the high-dose group, while the median was 40.76 mg/kg (IQR 35.30–44.78 mg/kg) in patients in the low-dose group. Cell salvage was used during the operation for blood conservation (Fresenius Kabi C.A.T.S.®plus, Fresenius Kabi AG, Bad Homburg, Germany). For on-pump surgeries, the activated clotting time (ACT) was maintained higher than 410 s with a heparin dose of 400 IU/kg, and for off-pump surgeries, the ACT was higher than 300 s with a heparin dose of 200 IU/kg. Additional doses of heparin were given according to the dynamic changes in act during the operation. The ratio of protamine to heparin (1 mg protamine: 100 units heparin) was 1:1; this ratio was used for neutralization. More protamine was added in consideration of hemostasis, ACT value and the recommendation of surgeons.
## Outcome definition
The primary outcome was defined as chest tube drainage and blood transfusion after surgery. The chest tube drainage volume at 24 and 48 h and the total chest tube drainage after CABG were considered as the blood loss after surgery. Blood transfusion after CABG included red blood cell (RBC) infusion, fresh frozen plasma (FFP) infusion and platelet (PLT) infusion. Secondary outcomes were safety issues, including in-hospital deaths and thromboembolic events [perioperative myocardial infarction (PMI), stroke, acute renal injury (AKI), and pulmonary embolism]. The details about the definitions are provided in the supplementary materials.
## Statistical analysis
The normal distribution was expressed as the means ± standard deviations (SDs), and nonnormal distributions were expressed as the medians and interquartile ranges (IQRs). Categorical variables are presented as numbers and percentages.
Before PSM, the baseline information between the TA group and no-TA group or the high-dose TA and low-dose TA groups was compared by Student’s t test for normally distributed continuous variables, the Mann‒Whitney U test for nonnormally distributed variables, and the χ2 test or Fisher’s exact test for categorical variables. A paired t test was used for normally distributed data, the Wilcoxon rank test was used for nonnormally distributed data, and McNemar’s test was used for categorical data after PSM.
Conditional logistic regression was performed in the binary outcome evaluation with an odds ratio (OR) and $95\%$ confidence intervals (CIs) after PSM.
Propensity score matching was performed according to 1:1 matching between the TA and no-TA groups or the high-dose and low-dose groups. The caliper width was 0.01, and the nearest-neighbor matching method without replacement was selected. There were 30 variables chosen based on the clinical and statistical significance in the PSM of the TA and no-TA groups (Supplementary Table S3), while 29 variables were matched in the high-dose and low-dose groups (except ticagrelor within 5 days) (Supplementary Table S4). A total of 1,910 patients in the TA group and 1,910 patients in the no-TA group were matched well. Balances were well maintained in the demographical and perioperative data with standardized differences <0.1 (Figure 2A). Additionally, 1,396 patients in the high-dose group and 1,396 in the low-dose group were matched and balanced well with standardized differences <0.1 (Figure 2B).
**Figure 2:** *Elderly patients undergoing CABG surgery with or without TA or with high- or low-dose TA administration were matched by propensity score. (A) PSM was used and well balanced in elderly patients with or without TA administration. The standardized difference was less than 0.1 in 30 variables; (B) elderly patients with high-dose and low-dose TA administration were matched well by propensity score. The standardized difference was less than 0.1 in all 29 covariates. Insulin-dependent diabetes was defined as the treatment of patients’ diabetes dependent on insulin therapy. Left ventricular dysfunction was defined as a patient’s ejection fraction ≤40%. Surgeons who performed more than 100 CABG surgeries per year assigned a value of “1”, and surgeons who performed less than 100 CABG surgeries per year were assigned a value of “0”. The risk factors for bleeding were age older than 70 years, female sex, low-molecular-weight heparin or an antiplatelet drug less than 5 days before surgery, renal impairment (estimated glomerular filtration rate, <60 ml per minute), and insulin-dependent diabetes.*
Binary logistic regression was used in the sensitivity analysis in the entire elderly cohort. Covariates were 30 variables and TA administration. Binary logistic regression was also applied to patients in the TA cohort with 29 variables and TA dosage as covariates. The dependent variables were binary outcomes. The “enter” method was used, and adjusted ORs and $95\%$ CIs of outcome variables were generated.
$p \leq 0.05$ was considered statistically significant. All statistical analyses were carried out with IBM SPSS Statistics for Windows, version 22.0 (IBM Corp., Armonk, NY, USA).
## The primary outcome in the TA and no-TA groups
After PSM, TA application was not related to the re-exploration rate due to massive hemorrhage and pericardial tamponade ($$p \leq 0.663$$) (Table 3). However, the blood loss after surgery was significantly reduced in the TA groups compared to that in the no-TA group (Table 3). The blood loss in 24 h after surgery in the TA group was 90 ml less than that in the no-TA group ($p \leq 0.0001$). The blood loss in 48 h after surgery in patients with TA was also 90 ml less than that in those without TA ($p \leq 0.0001$). After surgery, the total blood loss for patients with TA was 190 ml less than that in those without TA ($p \leq 0.0001$).
**Table 3**
| Outcomes | TA group (n = 1,910) | No TA group (n = 1,910) | OR (95% CI) by PSM | p value | OR (95% CI) by logistic regression | p value.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Primary outcome | Primary outcome | Primary outcome | Primary outcome | Primary outcome | Primary outcome | Primary outcome |
| Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation |
| Reoperation due to major hemorrhage or cardiac tamponade, n (%) | 40 (2.1) | 44 (2.3) | 0.91 (0.59–1.40) | 0.663 | 0.93 (0.64–1.35) | 0.705 |
| Blood loss in 24 h after surgery (ml), mean ± SD | 450 (320–600) | 530 (470–750) | | <0.0001 | | |
| Blood loss in 48 h after surgery (ml), mean ± SD | 720 (530–940) | 810 (710–1,072) | | <0.0001 | | |
| Total Blood loss after surgery (ml), mean ± SD | 940 (680–1,340) | 1,130 (900–1,430) | | <0.0001 | | |
| Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) |
| Blood transfusion | 658 (34.5) | 1,063 (55.7) | 0.62 (0.56–0.68) | <0.0001 | 0.38 (0.34–0.43) | <0.0001 |
| RBC | 568 (29.7) | 879 (46.0) | 0.65 (0.58–0.72) | <0.0001 | 0.44 (0.39–0.50) | <0.0001 |
| FFP | 296 (15.5) | 614 (32.1) | 0.48 (0.42–0.55) | <0.0001 | 0.37 (0.32–0.42) | <0.0001 |
| PLT | 63 (3.3) | 78 (4.1) | 0.81 (0.58–1.13) | 0.207 | 0.75 (0.54–1.04) | 0.081 |
| Secondary outcome, n (%) | 235 (12.3) | 212 (11.1) | 1.11 (0.92–1.34) | 0.277 | 1.08 (0.89–1.27) | 0.475 |
| Hospital death | 10 (0.5) | 14 (0.7) | 0.71 (0.32–1.61) | 0.416 | 0.59 (0.27–1.30) | 0.191 |
| Myocardial infarction | 102 (5.3) | 63 (3.3) | 1.62 (1.18–2.22) | 0.003 | 1.45 (1.08–1.95) | 0.014 |
| Stroke | 25 (1.3) | 22 (1.2) | 1.14 (0.64–2.02) | 0.662 | 0.97 (0.59–1.59) | 0.905 |
| Acute renal injury | 130 (6.8) | 130 (6.8) | 1.00 (0.78–1.28) | 1.000 | 1.02 (0.81–1.28) | 0.864 |
| Pulmonary embolism | 4 (0.2) | 3 (0.2) | 1.33 (0.30–5.96) | 0.706 | 1.06 (0.30–3.70) | 0.930 |
| Postoperative course | Postoperative course | Postoperative course | Postoperative course | Postoperative course | Postoperative course | Postoperative course |
| Intensive care (h), median (IQR) | 48 (24–96) | 48 (24–96) | | 0.004 | | |
| Hospital stay (day), mean ± SD | 17.50 ± 8.50 | 18.13 ± 9.25 | | 0.026 | | |
| Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) |
| Death from any cause within 30 days | 25 (1.3) | 24 (1.3) | 1.04 (0.60–1.82) | 0.886 | 0.82 (0.48–1.41) | 0.474 |
| Seizure | 1 (0.1) | 4 (0.2) | 0.25 (0.03–2.24) | 0.215 | 0.44 (0.10–1.87) | 0.256 |
In accordance with the decline in blood loss after TA administration, blood transfusion after surgery was also reduced in the TA group compared to that in the no-TA group (Table 3). The total blood transfusion was reduced 0.38-fold with TA administration compared to that in those without TA (OR = 0.62, $95\%$ CI 0.56–0.68, $p \leq 0.0001$). Blood component transfusion was also reduced. TA administration reduced the RBC transfusion rate 0.35-fold (OR = 0.65, $95\%$ CI 0.58–0.72, $p \leq 0.0001$); the FFP transfusion rate was reduced 0.52-fold (OR = 0.48, $95\%$ CI 0.42–0.55, $p \leq 0.0001$) by TA administration, but there was no reduction in the PLT transfusion rate after using TA (OR = 0.81, $95\%$ CI 0.58–1.13, $$p \leq 0.207$$).
The sensitivity analysis yielded results similar to those of the PSM analysis (Table 3).
## The primary outcome in the high-dose and low-dose TA groups
After PSM, the reoperation rate in the high-dose TA group was not different from that in the low-dose group [$$p \leq 0.397$$, OR = 1.27, $95\%$ CI (0.73–2.23)] (Table 4). However, the 24-h blood loss after the operation in the high-dose TA group was 20 ml less than that in the low-dose TA group statistically ($$p \leq 0.032$$) (Table 4). The blood loss within 48 h in the high-dose group was also lower than that in the low-dose group statistically (median 690 vs. 730 ml, $$p \leq 0.014$$). There were no differences in the total blood loss between the two groups ($$p \leq 0.174$$). The blood transfusion rate or components were not associated with the TA dosage ($p \leq 0.05$).
**Table 4**
| Outcome | TA group (n = 1,396) | No-TA group (n = 1,396) | OR (95% CI) by PSM | p value | OR (95% CI) by logistic regression | p value.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Primary outcome | Primary outcome | Primary outcome | Primary outcome | Primary outcome | Primary outcome | Primary outcome |
| Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation | Blood loss after operation |
| Reoperation due to major hemorrhage or cardiac tamponade, n (%) | 28 (2.0) | 22 (1.6) | 1.27 (0.73–2.23) | 0.397 | 1.20 (0.75–1.91) | 0.448 |
| Blood loss in 24 h after surgery (ml), mean ± SD | 430 (310–590) | 450 (320–620) | | 0.032 | | |
| Blood loss in 48 h after surgery (ml), mean ± SD | 690 (520–900) | 730 (530–957) | | 0.014 | | |
| Total Blood loss after surgery (ml), mean ± SD | 930 (670–958) | 960 (686–1,364) | | 0.174 | | |
| Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) | Blood transfusion after operation, n (%) |
| Blood transfusion | 479 (34.3) | 477 (34.2) | 1.00 (0.89–1.14) | 0.948 | 1.04 (0.90–1.20) | 0.605 |
| RBC | 430 (30.8) | 414 (29.7) | 1.04 (0.91–1.19) | 0.582 | 1.10 (0.95–1.28) | 0.206 |
| FFP | 202 (14.5) | 226 (16.2) | 0.89 (0.74–1.08) | 0.246 | 0.92 (0.77–1.11) | 0.384 |
| PLT | 39 (2.8) | 42 (3.0) | 0.93 (0.60–1.44) | 0.739 | 1.06 (0.71–1.60) | 0.770 |
| Secondary outcome, n (%) | 170 (12.2) | 169 (12.1) | 1.01 (0.81–1.25) | 0.957 | 1.02 (0.84–1.25) | 0.832 |
| Hospital death | 6 (0.4) | 9 (0.6) | 0.67 (0.24–1.87) | 0.442 | 0.50 (0.17–1.48) | 0.207 |
| Myocardial infarction | 74 (5.3) | 71 (5.1) | 1.04 (0.75–1.44) | 0.803 | 1.10 (0.83–1.46) | 0.520 |
| Stroke | 20 (1.4) | 15 (1.1) | 1.33 (0.68–2.60) | 0.400 | 1.26 (0.69–2.29) | 0.450 |
| Acute renal injury | 86 (6.2) | 99 (7.1) | 0.87 (0.65–1.16) | 0.340 | 0.93 (0.71–1.22) | 0.607 |
| Pulmonary embolism | 6 (0.4) | 3 (0.2) | 2.00 (0.50–8.00) | 0.327 | 1.24 (0.29–5.35) | 0.773 |
| Postoperative course | Postoperative course | Postoperative course | Postoperative course | Postoperative course | Postoperative course | Postoperative course |
| Intensive care (h), median (IQR) | 48 (24–96) | 48 (24–96) | | 0.201 | | |
| Hospital stay (day), mean ± SD | 17.38 ± 8.27 | 16.76 ± 7.67 | | 0.041 | | |
| Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) | Adverse events after surgery, n (%) |
| Death from any cause within 30 days | 15 (1.1) | 17 (1.2) | 0.88 (0.44–1.77) | 0.724 | 0.67 (0.36–1.26) | 0.215 |
| Seizure | 2 (0.1) | 0 (0.0) | - | 0.500 | 0.92 (0.11–7.92) | 0.938 |
## Safety issues in the TA and no-TA groups
The secondary outcome was not associated with TA administration [$$p \leq 0.277$$, OR = 1.11, $95\%$ CI (0.92–1.34)] (Table 3). No differences were noted between the two groups for stroke, AKI, pulmonary embolism or in-hospital death. However, TA increased the risk of PMI 1.62-fold [$$p \leq 0.003$$, OR = 1.62, $95\%$ CI (1.18–2.22)] (Table 3).
The administration of TA also reduced the hospital stay length compared to that of patients without TA ($$p \leq 0.026$$). TA was not associated with the risk of death within 30 days or seizure ($p \leq 0.05$).
The results from the sensitivity analysis were similar to the PSM results (Table 3).
## Safety issues in the TA dosage group
The TA dosage did not influence the secondary outcomes or its constitutive components ($p \leq 0.05$) (Table 4). No differences were observed in death within 30 days or seizure risk between the high- and low-dose groups ($p \leq 0.05$). However, high-dose TA administration reduced the length of hospital stay compared to the low-dose subsets (16.76 ± 7.67 vs. 17.38 ± 8.27 days, $$p \leq 0.041$$).
Similar results were obtained by sensitivity analysis (Table 4).
## Discussion
This study evaluated the TA and dosage effects on hemostasis and safety issues in elderly patients undergoing CABG surgeries in a retrospective cohort study. TA decreased chest tube drainage and blood infusion in elderly patients. However, the PMI risk in patients who received TA was increased significantly. High-dose TA administration also reduced blood loss but was not associated with the risk of blood transfusion. The dosage effects had no associations with the risk of safety issues.
Elderly patients undergoing cardiac surgeries had a higher risk of re-exploration, thereby increasing the risk of blood transfusion [4, 5]. In a meta-analysis of 557,923 patients undergoing cardiac surgery [26], the patients who underwent re-exploration were significantly older, and re-exploration significantly increased the risk of postoperative mortality and morbidity. Murphy et al. [ 7] revealed that RBC infusion in cardiac operations was closely related to infection and the incidence of postoperative ischemia, length of hospital stay, and increased early and late mortality and hospitalization expenses. Ibrahim et al. [ 27] also found in a 14,281-patient cohort who received cardiac surgery that RBC infusion became an independent risk factor for readmission and mortality. Elderly patients undergoing CABG manifested higher hemostatic activation than young patients for increased fibrinolysis and platelet activation [14]. This situation undoubtedly leads to a higher blood transfusion risk. Guri et al. reported that the number of RBC transfusions was reduced by TA in 64 patients aged ≥70 years undergoing CABG plus aortic valve replacement [8]. However, the small number of patients limits the clinical significance. In this study, a larger CABG cohort including 7,224 patients ≥70 years old revealed that TA administration significantly reduced the blood loss and transfusion rate compared that observed when TA was not administered. Therefore, TA fosters sound blood management effects in elderly patients and thereby could benefit the prognosis of those undergoing CABG surgeries.
The double-blade effects of TA were investigated since it was introduced into clinical application. There was no doubt that TA could maintain hemostasis during CABG surgeries in the studies conducted previously (9–11) and in this study. The evidence on the safety issues of TA during cardiac surgery did not lead to a consensus (9–13). Myers et al. [ 11] reported no associations between thromboembolic events and TA administration. However, our previous study found that the risk of PMI was significantly increased after TA administration [9]. Furthermore, Zhou et al. found that intraoperative TA was associated with postoperative stroke in patients undergoing cardiac surgery. For elderly patients undergoing CABG surgery, less research on the safety issues of TA has been conducted. Elderly patients undergoing cardiac surgery had higher morbidity and mortality, and $78\%$ of deaths and major complications occurred in patients ≥75 years [2]. Wilson et al. found that patients undergoing CABG aged over 75 years had a longer hospital stay times and higher perioperative mortality than those under 75 years [28]. Based on the pathophysiology of elderly patients undergoing CABG surgery and the TA properties of thrombosis, the emphasis was on the safety issues after TA administration in this subset.
Our study found that the TA group showed an increased risk of PMI during hospitalization. This result was different from the study by Myles et al. [ 11]. The ATACAS trial [11] revealed that TA did not increase the risk of thrombotic complications within 30 days after CABG. We tried to explain it in three aspects. First, the CABG composition was different in the ATACAS trial from that in our retrospective data. The CABG in the ATACAS trial were mainly on-pump ($97\%$ in the TXA group vs. $96.8\%$ in the placebo group). However, the on-pump rate in CABG surgeries was not so high in the real world. Our study found that after PSM, the on-pump CABG in TA group and no-TA group accounted for $38.7\%$ and $40.2\%$ respectively (Supplementary Table S1). Second, different definitions of PMI were used. The PMI in the ATACAS trial was defined a bit complicatedly. It included the dynamic change of creatine kinase isoenzyme (CK-MB) or Troponin I (cTnI), the ECG change, or autopsy results, and Troponin I > 10 ng/ml, or Troponin T > 4.0, or CKMB > three times upper reference limit (URL) at any time >12 h post-CABG. However, the derivation of ECG or imaging data was difficult to find in our large retrospective cohort spanning 11 years. So the definition in this clinical study was according to the Society for Cardiovascular Angiography and Interventions (SCAI), which is mainly based on cTnI ≥ 70 URL or CK-MB ≥ 10 URL in the 48-h post-operative period [24, 25]. Third, different populations were focused on. In the study by Myles et al., the mean age in the TA group was 66.8 ± 9.8 years, while 67.0 ± 9.6 years in the placebo group. Our study focused on older patients ≥70 years old. So the mean age in the TA group was 73.54 ± 3.00 years, and 73.49 ± 3.00 in the no-TA group after PSM. Our study population was significantly older than those in the ATACAS trial. According to the three main differences from the ATACAS trial, we believed the inconsistency of PMI results between ATACAS trial and our study was reasonable.
The blood management guidelines have already led to widespread administration of TA in adult cardiac surgery [29], but no consensus has been reached on the TA administration scheme. Some clinical trials recommended a high dose of TA for effective blood management [30, 31]. Sigaut et al. [ 30]. found that, compared with a low dose (10 mg/kg bolus + 1 mg/kg/h infusion), a high dose of TA (30 mg/kg bolus + 16 mg/kg/h infusion) decreased blood transfusion needs, blood loss, and the reoperation rate. However, other researchers discovered that low-dose TA administration is sufficient to reduce postoperative blood loss and RBC infusion in cardiac surgery (32–34). A model-based meta-analysis comprised 49,817 patients undergoing cardiopulmonary bypass operations and discovered that low-dose TA administration could reduce bleeding outcomes without increasing seizure risk [32]. Waldow et al. found that high-dose TA administration over 100 mg/kg was an independent risk predictor of early seizure, and the postoperative blood loss and blood infusion needs were similar between the low-dose and high-dose TA groups [33]. The TA dosage effects in elderly patients have been less evaluated. In this study, we discovered that high-dose TA administration could significantly decrease blood loss, but the blood transfusion between the two dose groups was not different. For safety issues, TA dose was not associated with thromboembolic events, in-hospital death, seizure, ICU stay time, or death within 30 days; conversely, high-dose TA administration could reduce the hospital stay time. In this respect, high-dose TA administration ≥50 mg/kg was both effective and safe for elderly patients.
However, there were several limitations in this study. First, clinical studies and evidence on TA in cardiac surgery have increased over time. Even at present, the safety issue of TA application in cardiac surgery is still under debate [9, 13]. This study ranged in duration from 2009 to 2019. During this time, many anesthesiologists participated in CABG surgeries, making it hard to explain why TA was used or not during CABG. Therefore, we would like to conclude that TA was used under the consideration of anesthesiologists’ judgment and preference. Second, the 11-year period inevitably makes it difficult to obtain all the variables we needed for this analysis. Third, although PSM was chosen to simulate an RCT, potential cofounding covariates were possibly present. Therefore, a large RCT is needed for further study. Fourth, the study population was from China. The Asians appeared to have the lowest BMI, whereas the African American and Hispanic adults have the highest [35]. The mean BMI in this study was around 25 kg/m2 either in the TA group or the no-TA group. As we know, lower BMI leads to more bleeding after cardiac surgery even in elderly patients (36–38). Thereby, the discrepancy in BMI among different ethnic groups may lead to different blood management effects after TA usage. Hence we could not simply extend our results to the world without consideration of racial differences.
We revealed that elderly patients undergoing CABG surgery had better hemostasis after TA administration but had increased PMI risk. High-dose TA administration was effective and safe for elderly patients who underwent CABG surgery.
## 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 Medical Ethics Committee of the Fuwai Hospital. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
EW analyzed and interpreted the patient data. EW and YW performed the statistical analysis, EW and YL draft the manuscript. SY and SH review and editing 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/fsurg.2023.1117974/full#supplementary-material.
## References
1. Wiedemann D, Bernhard D, Laufer G, Kocher A. **The elderly patient and cardiac surgery: a mini-review**. *Gerontology* (2010) **56** 241-9. DOI: 10.1159/000248761
2. Yaffee DW, Williams MR. **Cardiovascular surgery in the elderly**. *Semin Thorac Cardiovasc Surg* (2016) **28** 741-7. DOI: 10.1053/j.semtcvs.2016.08.007
3. Speziale G, Nasso G, Barattoni MC, Esposito G, Popoff G, Argano V. **Short-term and long-term results of cardiac surgery in elderly and very elderly patients**. *J Thorac Cardiovasc Surg* (2011) **141** 725-31. DOI: 10.1016/j.jtcvs.2010.05.010
4. Moulton MJ, Creswell LL, Mackey ME, Cox JL, Rosenbloom M. **Reexploration for bleeding is a risk factor for adverse outcomes after cardiac operations**. *J Thorac Cardiovasc Surg* (1996) **111** 1037-46. DOI: 10.1016/s0022-5223(96)70380-x
5. Dacey LJ, Munoz JJ, Baribeau YR, Johnson ER, Lahey SJ, Leavitt BJ. **Reexploration for hemorrhage following coronary artery bypass grafting: incidence and risk factors. Northern new England cardiovascular disease study group**. *Arch Surg* (1998) **133** 442-7. DOI: 10.1001/archsurg.133.4.442
6. Kuduvalli M, Oo AY, Newall N, Grayson AD, Jackson M, Desmond MJ. **Effect of peri-operative red blood cell transfusion on 30-day and 1-year mortality following coronary artery bypass surgery**. *Eur J Cardiothorac Surg* (2005) **27** 592-8. DOI: 10.1016/j.ejcts.2005.01.030
7. Murphy GJ, Reeves BC, Rogers CA, Rizvi SI, Culliford L, Angelini GD. **Increased mortality, postoperative morbidity, and cost after red blood cell transfusion in patients having cardiac surgery**. *Circulation* (2007) **116** 2544-52. DOI: 10.1161/CIRCULATIONAHA.107.698977
8. Greiff G, Stenseth R, Wahba A, Videm V, Lydersen S, Irgens W. **Tranexamic acid reduces blood transfusions in elderly patients undergoing combined aortic valve and coronary artery bypass graft surgery: a randomized controlled trial**. *J Cardiothorac Vasc Anesth* (2012) **26** 232-8. DOI: 10.1053/j.jvca.2011.07.010
9. Wang E, Yuan X, Wang Y, Chen W, Zhou X, Hu S. **Blood conservation outcomes and safety of tranexamic acid in coronary artery bypass graft surgery**. *Int J Cardiol* (2022) **348** 50-6. DOI: 10.1016/j.ijcard.2021.12.017
10. Wang E, Yuan X, Wang Y, Chen W, Zhou X, Hu S. **Tranexamic acid administered during off-pump coronary artery bypass graft surgeries achieves good safety effects and hemostasis**. *Front Cardiovasc Med* (2022) **9** 1-9. DOI: 10.3389/fcvm.2022.775760
11. Myles PS, Smith JA, Forbes A, Silbert B, Jayarajah M, Painter T. **Tranexamic acid in patients undergoing coronary-artery surgery**. *N Engl J Med* (2017) **376** 136-48. DOI: 10.1056/NEJMoa1606424
12. Henry DA, Carless PA, Moxey AJ, O’Connell D, Stokes BJ, Fergusson DA. **Anti-fibrinolytic use for minimising perioperative allogeneic blood transfusion**. *Cochrane Database Syst Rev* (2011) **2011** CD001886. DOI: 10.1002/14651858.CD001886.pub3
13. Zhou ZF, Zhang FJ, Huo YF, Yu YX, Yu LN, Sun K. **Intraoperative tranexamic acid is associated with postoperative stroke in patients undergoing cardiac surgery**. *PLoS One* (2017) **12** e0177011. DOI: 10.1371/journal.pone.0177011
14. Pleym H, Wahba A, Videm V, Asberg A, Lydersen S, Bjella L. **Increased fibrinolysis and platelet activation in elderly patients undergoing coronary bypass surgery**. *Anesth Analg* (2006) **102** 660-7. DOI: 10.1213/01.ane.0000196526.28277.45
15. Ghanta RK, Shekar PS, McGurk S, Rosborough DM, Aranki SF. **Nonelective cardiac surgery in the elderly: is it justified?**. *J Thorac Cardiovasc Surg* (2010) **140** 103-9, 9 e1. DOI: 10.1016/j.jtcvs.2009.10.001
16. Sharma V, Fan J, Jerath A, Pang KS, Bojko B, Pawliszyn J. **Pharmacokinetics of tranexamic acid in patients undergoing cardiac surgery with use of cardiopulmonary bypass**. *Anaesthesia* (2012) **67** 1242-50. DOI: 10.1111/j.1365-2044.2012.07266.x
17. Hodgson S, Larvin JT, Dearman C. **What dose of tranexamic acid is most effective and safe for adult patients undergoing cardiac surgery?**. *Interact Cardiovasc Thorac Surg* (2015) **21** 384-8. DOI: 10.1093/icvts/ivv134
18. Shi J, Zhou C, Liu S, Sun H, Wang Y, Yan F. **Outcome impact of different tranexamic acid regimens in cardiac surgery with cardiopulmonary bypass (optimal): rationale, design, and study protocol of a multicenter randomized controlled trial**. *Am Heart J* (2020) **222** 147-56. DOI: 10.1016/j.ahj.2019.09.010
19. Koster A, Borgermann J, Zittermann A, Lueth JU, Gillis-Januszewski T, Schirmer U. **Moderate dosage of tranexamic acid during cardiac surgery with cardiopulmonary bypass and convulsive seizures: incidence and clinical outcome**. *Br J Anaesth* (2013) **110** 34-40. DOI: 10.1093/bja/aes310
20. Murkin JM, Falter F, Granton J, Young B, Burt C, Chu M. **High-dose tranexamic acid is associated with nonischemic clinical seizures in cardiac surgical patients**. *Anesth Analg* (2010) **110** 350-3. DOI: 10.1213/ANE.0b013e3181c92b23
21. Faraoni D, Cacheux C, Van Aelbrouck C, Ickx BE, Barvais L, Levy JH. **Effect of two doses of tranexamic acid on fibrinolysis evaluated by thromboelastography during cardiac surgery: a randomised, controlled study**. *Eur J Anaesthesiol* (2014) **31** 491-8. DOI: 10.1097/EJA.0000000000000051
22. Guo J, Gao X, Ma Y, Lv H, Hu W, Zhang S. **Different dose regimes and administration methods of tranexamic acid in cardiac surgery: a meta-analysis of randomized trials**. *BMC Anesthesiol* (2019) **19** 129. DOI: 10.1186/s12871-019-0772-0
23. Sander M, Spies CD, Martiny V, Rosenthal C, Wernecke KD, von Heymann C. **Mortality associated with administration of high-dose tranexamic acid and aprotinin in primary open-heart procedures: a retrospective analysis**. *Crit Care* (2010) **14** R148. DOI: 10.1186/cc9216
24. Moussa ID, Klein LW, Shah B, Mehran R, Mack MJ, Brilakis ES. **Consideration of a new definition of clinically relevant myocardial infarction after coronary revascularization: an expert consensus document from the society for cardiovascular angiography and interventions (SCAI)**. *J Am Coll Cardiol* (2013) **62** 1563-70. DOI: 10.1016/j.jacc.2013.08.720
25. Thielmann M, Sharma V, Al-Attar N, Bulluck H, Bisleri G, Bunge JJH. **Esc joint working groups on cardiovascular surgery and the cellular biology of the heart position paper: perioperative myocardial injury and infarction in patients undergoing coronary artery bypass graft surgery**. *Eur Heart J* (2017) **38** 2392-407. DOI: 10.1093/eurheartj/ehx383
26. Biancari F, Mikkola R, Heikkinen J, Lahtinen J, Airaksinen KE, Juvonen T. **Estimating the risk of complications related to Re-exploration for bleeding after adult cardiac surgery: a systematic review and meta-analysis**. *Eur J Cardiothorac Surg* (2012) **41** 50-5. DOI: 10.1016/j.ejcts.2011.04.023
27. Sultan I, Bianco V, Brown JA, Kilic A, Habertheuer A, Aranda-Michel E. **Long-term impact of perioperative red blood cell transfusion on patients undergoing cardiac surgery**. *Ann Thorac Surg* (2021) **112** 546-54. DOI: 10.1016/j.athoracsur.2020.10.023
28. Wilson MF, Baig MK, Ashraf H. **Quality of life in octagenarians after coronary artery bypass grafting**. *Am J Cardiol* (2005) **95** 761-4. DOI: 10.1016/j.amjcard.2004.11.031
29. Boer C, Meesters MI, Milojevic M, Benedetto U. **2017 Eacts/eacta guidelines on patient blood management for adult cardiac surgery**. *J Cardiothorac Vasc Anesth* (2018) **32** 88-120. DOI: 10.1053/j.jvca.2017.06.026
30. Sigaut S, Tremey B, Ouattara A, Couturier R, Taberlet C, Grassin-Delyle S. **Comparison of two doses of tranexamic acid in adults undergoing cardiac surgery with cardiopulmonary bypass**. *Anesthesiology* (2014) **120** 590-600. DOI: 10.1097/ALN.0b013e3182a443e8
31. Karski JM, Dowd NP, Joiner R, Carroll J, Peniston C, Bailey K. **The effect of three different doses of tranexamic acid on blood loss after cardiac surgery with mild systemic hypothermia (32 degrees C)**. *J Cardiothorac Vasc Anesth* (1998) **12** 642-6. DOI: 10.1016/s1053-0770(98)90235-x
32. Zufferey PJ, Lanoiselee J, Graouch B, Vieille B, Delavenne X, Ollier E. **Exposure-response relationship of tranexamic acid in cardiac surgery**. *Anesthesiology* (2021) **134** 165-78. DOI: 10.1097/ALN.0000000000003633
33. Kalavrouziotis D, Voisine P, Mohammadi S, Dionne S, Dagenais F. **High-dose tranexamic acid is an independent predictor of early seizure after cardiopulmonary bypass**. *Ann Thorac Surg* (2012) **93** 148-54. DOI: 10.1016/j.athoracsur.2011.07.085
34. Waldow T, Szlapka M, Haferkorn M, Burger L, Plotze K, Matschke K. **Prospective clinical trial on dosage optimizing of tranexamic acid in non-emergency cardiac surgery procedures**. *Clin Hemorheol Microcirc* (2013) **55** 457-68. DOI: 10.3233/CH-131782
35. **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
36. Birkmeyer NJ, Charlesworth DC, Hernandez F, Leavitt BJ, Marrin CA, Morton JR. **Obesity and risk of adverse outcomes associated with coronary artery bypass surgery. Northern new England cardiovascular disease study group**. *Circulation* (1998) **97** 1689-94. DOI: 10.1161/01.cir.97.17.1689
37. Maurer MS, Luchsinger JA, Wellner R, Kukuy E, Edwards NM. **The effect of body mass index on complications from cardiac surgery in the oldest old**. *J Am Geriatr Soc* (2002) **50** 988-94. DOI: 10.1046/j.1532-5415.2002.50251.x
38. Lopes CT, Brunori EF, Cavalcante AM, Moorhead SA, Swanson E, Lopes Jde L. **Factors associated with excessive bleeding after cardiac surgery: a prospective cohort study**. *Heart Lung* (2016) **45** 64-9 e2. DOI: 10.1016/j.hrtlng.2015.09.003
|
---
title: 'Intake of omega-3 polyunsaturated fatty acids and fish associated with prevalence
of low lean mass and muscle mass among older women: Analysis of Korea National Health
and Nutrition Examination Survey, 2008-2011'
authors:
- Yeji Kim
- Yongsoon Park
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9989170
doi: 10.3389/fnut.2023.1119719
license: CC BY 4.0
---
# Intake of omega-3 polyunsaturated fatty acids and fish associated with prevalence of low lean mass and muscle mass among older women: Analysis of Korea National Health and Nutrition Examination Survey, 2008-2011
## Abstract
The effects of dietary n-3 PUFA and fish on the risk of sarcopenia and muscle mass remain unclear. The present study investigated the hypothesis that intake of n-3 PUFA and fish is negatively associated with the prevalence of low lean mass (LLM) and positively correlated with muscle mass in older adults. Data from the Korea National Health and Nutrition Examination Survey, 2008-2011, 1,620 men and 2,192 women aged over 65 years were analyzed. LLM was defined as appendicular skeletal muscle mass divided by body mass index < 0.789 kg for men and <0.512 kg for women. Women and men with LLM consumed less eicosapentaenoic acid (EPA) docosahexaenoic acid (DHA) and fish. In women, but not men, the prevalence of LLM was associated with the intake of EPA and DHA (odds ratio, 0.65; $95\%$ confidence interval, 0.48-0.90; $$p \leq 0.002$$) and fish (odds ratio, 0.59; $95\%$ confidence interval, 0.42-0.82; $p \leq 0.001$). Muscle mass was also positively associated with the intake of EPA, DHA ($$p \leq 0.026$$), and fish ($$p \leq 0.005$$) in women, but not men. α-Linolenic acid intake was not associated with the prevalence of LLM and was not correlated with muscle mass. The findings suggest that consumption of EPA, DHA, and fish are negatively associated with the prevalence of LLM, and positively correlated with muscle mass in Korean older women, but not in older men.
## 1. Introduction
Sarcopenia, an age-associated loss of muscle mass and, strength, or performance is associated with increased adverse outcomes including falls, functional decline, frailty, and mortality, and has become a serious health issue among older adults [1]. There are various complicated risk factors for sarcopenia, including aging, body composition, physical activity, comorbidities, and dietary intake [1]. Malnutrition is well-known risk factor for sarcopenia, but the effect of individual nutrient such as protein, vitamin D, and n-3 polyunsaturated fatty acids (PUFA) on sarcopenia is unclear [2, 3]. N-3 PUFA, eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA), which are abundant in fish, and α-linolenic acid (ALA), which is abundant in plants, have anti-inflammatory effects [4]. It is becoming increasingly clear that inflammation processes play an important role in the pathogenesis of age-related sarcopenia [4]. Low lean mass is one of the first factors to diagnose sarcopenia [1].
The risk of sarcopenia is negatively associated with the intake of total n-3 PUFA in kidney transplant patients [5] and serum levels of total n-3 PUFA in Korean older adults [6]. Similarly, the ratio of daily intake of total n-3 PUFA to energy intake has been significantly associated with the risk of sarcopenic obesity in Korean older women, but not men, suggesting that n-3 PUFA have beneficial effects only in women [7]. In kidney transplant patients, the intake of total n-3 PUFA have also been positively associated with muscle mass and negatively associated with the risk of low muscle mass [5].
Consistent with muscle mass, intakes of total n-3 PUFA, EPA, and DHA were positively associated with muscle function in American older adults [8], Japanese men [9], Finnish women [10], and Korean women [11]. Sarcopenic older adults consumed less total n-3 PUFA and had lower erythrocyte EPA levels in the Maastricht Sarcopenia Study [12]. Plasma levels of total n-3 PUFA, EPA, and DHA, as indicators of dietary intake of n-3 PUFA, were positively associated with muscle function among older adults in Europe (12–14), America [15], and Korea [6, 16]. In addition, a meta-analysis of clinical trials found that supplementation with EPA and DHA increased muscle mass and muscle performance measured by timed up-and-go and gait speed in older adults [17]. n-3 PUFA have been suggested to have anabolic and anti-catabolic properties in skeletal muscles by regulating the mammalian target of rapamycin (mTOR) signaling pathway and inflammatory factors [18].
Dietary intake of fatty fish was positively associated with grip strength in older adults in the Hertfordshire cohort study [19] and UK Biobank study [20]. Similarly, adherence to the Mediterranean diet, which is known to contain abundant n-3 PUFA, was associated with better physical performance in postmenopausal women [21] and a lower risk of sarcopenia in Iranian older adults [22].
ALA intake was positively associated with muscle function measured by gait speed, one-leg stance, squat, Short Physical Performance Battery (SPPB), and grip strength, but not with muscle mass in older women in Finland [10]. Supplementation with ALA increased muscle mass but not muscle function in older men but not in women [23].
To our knowledge, no study has shown an association between dietary intake of EPA and DHA, ALA, and fish and the prevalence of sarcopenia and muscle mass in older adults. Therefore, the purpose of the present study was to investigate the hypothesis that consumption of n-3 PUFA and fish is negatively associated with the prevalence of LLM and positively correlated with muscle mass in older men and women.
## 2.1. Participants
This study was based on data obtained from the Korea National Health and Nutrition Examination Survey (KNHANES) from 2008 to 2011. KNHANES was performed using a rolling sampling design involving a complex, stratified, multistage, probability-cluster survey of a representative sample of the non-institutionalized civilian population in South Korea [24]. The survey was performed by the Korean Ministry of Health and Welfare and consisted of three components: a health interview survey, health examination survey, and nutrition survey. All participants signed an informed consent form [24]. The study protocol was approved by the Institutional Review Board of Hanyang University (HYUIRB-202208-003).
Of the 37,753 participants, 33,941 were excluded for the following reasons: age 65 years or younger ($$n = 31$$,383); missing data on body mass index (BMI, kg/m2), appendicular skeletal muscle mass (ASM), and energy intake ($$n = 2$$,480); and extreme energy intake of less than 500 kcal/day or more than 4,000 kcal/day ($$n = 78$$). Finally, 1,620 men and 2,192 women were included in the present study (Figure 1).
**FIGURE 1:** *Flowchart of the inclusion and exclusion of participants.*
## 2.2. Definition of low lean mass
Muscle mass was measured by dual-energy X-ray absorptiometry using a DISCOVERY-W fan-beam densitometer (Hologic, Marlborough, MA, USA). ASM (kg) was calculated as the sum of lean soft tissue in the bilateral upper and lower limbs. LLM was defined as < 0.789 kg of ASM/BMI in men and < 0.512 kg of ASM/BMI in women [25].
## 2.3. Study variables
Trained interviewers and medical staff assessed a wide range of covariates according to a standardized protocol [24]. Anthropometry of waist circumference (WC) was measured at the midpoint between the inferior margin of the last rib and the iliac crest in the horizontal plane while the subject exhaled. Height and weight were measured to the nearest 0.1 cm and 0.1 kg, respectively, with participants wearing light clothing and being barefooted.
A questionnaire related to sociodemographic characteristics that included age, sex, smoking status, drinking status, regular exercise, living arrangement, and comorbidities was administered in the health interview. Smoking was defined as never (a person who has never smoked or has smoked less than five packs of cigarettes during their lifetime), former (a person who smoked more than five packs of cigarettes but who did not currently smoke), or current (a person who smokes more than five packs of cigarettes during their lifetime). Drinking was defined as the alcohol once or more times in a month. Regular exercise was defined as moderate exercise for 30 min, ≥ 5 times a week, vigorous exercise for 20 min, ≥ 3 times a week. Living arrangements were classified into two groups according to whether or not they live alone. Comorbidities were defined as participants with at least one medical history of hypertension, dyslipidemia, stroke, myocardial infarction, angina, osteoarthritis, rheumatoid arthritis, kidney failure, diabetes mellitus, or cancer.
Dietary intake data were assessed using a one-day 24-h dietary recall method during the household interview. Trained dietitians interviewed the participants to recall and describe all the foods and beverages they had consumed in the previous day. Fish was classified according to the Composition Table of Marine Products in Korea 2018 of the National Institute of Fisheries Science [26], and amount of n-3 PUFA in individual fish, as g/day was calculated based on the Food Composition Table developed by the Korea Rural Development Administration in 2018 [27].
## 2.4. Statistical analyses
Descriptive analysis was conducted using clustering and stratifying variables, using a survey procedure that applied individual weights to the analysis [28]. Continuous variables were analyzed using the independent sample t-test and are expressed as the mean ± standard error of the mean. Categorical variables were analyzed using the chi-square test and are expressed as frequencies and percentages.
Multiple regression models were used to determine unsuitable potential covariates and examine the association between the prevalence of LLM and dietary intake of n-3 PUFA and fish after adjusting for potential covariates. In multivariate models, covariates with $p \leq 0.20$ were selected as confounding factors and included in the fully adjusted model [29]. The participants were subdivided into three groups according to tertiles of dietary n-3 PUFA and fish intake, separately [30]. Analysis of covariance (ANCOVA) with Bonferroni correction was performed to assess the mean differences in ASM/BMI among the intake tertile groups after adjustment for confounding variables. The relationship between LLM and dietary intake of n-3 PUFA and fish was analyzed using multivariable logistic regression analysis. This analysis was used to obtain odds ratios (ORs) and $95\%$ confidence intervals (CIs) adjusted for confounding variables. p-values for linear trends were estimated using the median values within each tertile of dietary intake, considering the unequal distances between tertiles. $p \leq 0.05$ was considered statistically significant. Statistical analysis was performed using SPSS 27.0 software (SPSS Inc., Chicago, IL, USA).
## 3.1. Baseline characteristics of participants
Compared to those without LLM, men and women with LLM were older, had higher BMI, greater WC, increased prevalence of obesity and abdominal obesity, more comorbidities, consumed less alcohol, and had reduced energy intake (Table 1). There were no statistically significant differences in the prevalence of LLM, smoking status, living alone, and ALA intake between the LLM and non-LLM groups. Women with LLM exercised less regularly and consumed less EPA, DHA, and fish than women without LLM. The total population with LLM was older, had higher BMI, greater WC, increased prevalence of obesity and abdominal obesity, more comorbidities, reduced exercise activities, and consumed less alcohol, energy, EPA, DHA, and fish than those without LLM (Supplementary Table 1).
**TABLE 1**
| Variables | Men | Men.1 | p-Value* | Women | Women.1 | p-Value*.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | Non-LLM (n = 1,167) | LLM (n = 453) | | Non-LLM (n = 1,629) | LLM (n = 563) | |
| Age (year) | 71.36 ± 0.13 | 73.04 ± 0.22 | <0.001 | 71.81 ± 0.12 | 73.01 ± 0.19 | <0.001 |
| BMI (kg/m2) | 22.62 ± 0.08 | 24.29 ± 0.14 | <0.001 | 23.46 ± 0.08 | 25.96 ± 0.14 | <0.001 |
| Obesity, n (%) | 222 (19.0) | 171 (37.7) | <0.001 | 495 (30.4) | 335 (59.5) | <0.001 |
| WC (cm) | 83.46 ± 0.26 | 87.60 ± 0.42 | <0.001 | 81.83 ± 0.23 | 87.10 ± 0.40 | <0.001 |
| Abdominal obesity, n (%) | 285 (24.5) | 185 (40.9) | <0.001 | 602 (37.1) | 329 (58.9) | <0.001 |
| Smoking status, n (%) | | | 0.069 | | | 0.054 |
| Never | 200 (17.3) | 74 (16.7) | | 1,446 (90.1) | 501 (90.6) | |
| Former | 646 (56.0) | 275 (61.9) | | 69 (4.3) | 31 (5.6) | |
| Current | 308 (26.7) | 95 (21.4) | | 90 (5.6) | 21 (3.8) | |
| Drinking status, n (%) | 684 (59.3) | 224 (50.5) | 0.005 | 279 (17.4) | 72 (13.0) | 0.037 |
| Regular exercise, n (%) | 260 (22.3) | 88 (19.4) | 0.305 | 322 (19.8) | 68 (12.1) | <0.001 |
| Living alone, n (%) | 80 (6.9) | 30 (6.7) | 0.957 | 418 (25.7) | 142 (25.3) | 0.214 |
| Comorbidities | 720 (62.1) | 350 (78.5) | < 0.001 | 1,289 (80.1) | 493 (88.4) | <0.001 |
| Dietary intake | Dietary intake | Dietary intake | Dietary intake | Dietary intake | Dietary intake | Dietary intake |
| Energy intake (kcal/day) | 1,897.07 ± 17.56 | 1,710.18 ± 26.21 | < 0.001 | 1,459.16 ± 12.41 | 1,365.54 ± 19.94 | <0.001 |
| EPA+DHA (g/day) | 0.84 ± 0.04 | 0.76 ± 0.06 | 0.830 | 0.50 ± 0.02 | 0.40 ± 0.03 | 0.002 |
| ALA (g/day) | 1.47 ± 0.06 | 1.24 ± 0.08 | 0.393 | 1.14 ± 0.07 | 1.01 ± 0.06 | 0.088 |
| Fish (g/day) | 43.56 ± 2.21 | 36.83 ± 3.03 | 0.545 | 26.44 ± 1.37 | 18.65 ± 1.74 | <0.001 |
## 3.2. Associations between prevalence of LLM and intakes of n-3 PUFA and fish
Logistic regression analysis revealed that the prevalence of LLM was negatively associated with the intake of EPA and DHA, and fish, but not ALA in women, after adjusting for potential confounders (Table 2). However, there was no significant association between the prevalence of LLM and the intake of n-3 PUFA and fish in men. After full adjustment, the intake of EPA and DHA and fish were also negatively associated in the total study population (Supplementary Table 2). We divided women into 4 groups; non-sarcopenic non-obesity, sarcopenic non-obesity, non-sarcopenic obesity, and sarcopenic obesity. The prevalence of LLM was significantly associated with intake of EPA and DHA, and fish in women with non-sarcopenic non-obesity, sarcopenic non-obesity, and sarcopenic obesity (Supplementary Table 3).
**TABLE 2**
| Variables | Tertiles of n-3 PUFA and fish intake | Tertiles of n-3 PUFA and fish intake.1 | Tertiles of n-3 PUFA and fish intake.2 | p-Trend |
| --- | --- | --- | --- | --- |
| | T1 | T2 | T3 | |
| Men | Men | Men | Men | Men |
| EPA+DHA (g/day), range | <0.13 | 0.13 ≤ to < 0.67 | ≥0.67 | 0.522 |
| No. with/without LLM | 155/385 | 168/372 | 130/410 | |
| OR (95% CI) | 1 | 1.214 (0.881 – 1.673) | 0.962 (0.673– 1.376) | |
| ALA (g/day), range | < 0.58 | 0.58 ≤ to < 1.30 | ≥ 1.30 | 0.404 |
| No. with/without LLM | 175/365 | 156/384 | 122/418 | |
| OR (95% CI) | 1 | 0.897 (0.657 – 1.225) | 0.842 (0.582 – 1.217) | |
| Fish (g/day), range | < 0.24 | 0.24 ≤ to < 34.20 | ≥ 34.20 | 0.567 |
| No. with/without LLM | 152/388 | 169/371 | 132/408 | |
| OR (95% CI) | 1 | 1.375 (0.983 – 1.922) | 1.038 (0.748 – 1.442) | |
| Women | Women | Women | Women | Women |
| EPA+DHA (g/day), range | < 0.06 | 0.06 ≤ to < 0.40 | ≥ 0.40 | 0.002 |
| No. with/without LLM | 207/523 | 195/536 | 161/570 | |
| OR (95% CI) | 1 | 1.088 (0.818 – 1.447) | 0.654 (0.478–0.896) | |
| ALA (g/day), range | < 0.43 | 0.43 ≤ to < 0.95 | ≥ 0.95 | 0.602 |
| No. with/without LLM | 189/541 | 194/537 | 180/551 | |
| OR (95% CI) | 1 | 1.272 (0.940 – 1.722) | 1.151 (0.829–1.597) | |
| Fish (g/day), range | < 0.00 | 0.00 ≤ to < 15.33 | ≥ 15.33 | <0.001 |
| No. with/without LLM | 239/626 | 165/431 | 159/572 | |
| OR (95% CI) | 1 | 1.214 (0.926 – 1.592) | 0.590 (0.423–0.823) | |
## 3.3. Associations between muscle mass and intakes of dietary n-3 PUFA and fish
After adjusting for potential confounders, ANCOVA revealed a significant positive association between muscle mass and intake of EPA, DHA, and fish, but not ALA in women, as a continuous and non-continuous variable (Table 3). However, there were no associations between muscle mass and intakes of n-3 PUFA and fish in men (Table 3) and the total population (Supplementary Table 4).
**TABLE 3**
| Variables (g/day) | Variables (g/day).1 | Tertiles of n-3 PUFA and fish intake | Tertiles of n-3 PUFA and fish intake.1 | Tertiles of n-3 PUFA and fish intake.2 | p-Trend* | Continuous | Continuous.1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | T1 | T2 | T3 | | r | p-Value |
| Men | EPA+DHA, range | < 0.13 | 0.13 ≤ to < 0.67 | ≥ 0.67 | | | |
| | ASM/BMI | 0.85 ± 0.004 | 0.84 ± 0.004 | 0.85 ± 0.004 | 0.743 | 0.012 | 0.689 |
| | ALA, range | < 0.58 | 0.58 ≤ to < 1.30 | ≥ 1.30 | | | |
| | ASM/BMI | 0.84 ± 0.004 | 0.84 ± 0.004 | 0.86 ± 0.004 | 0.754 | 0.017 | 0.603 |
| | Fish, range | < 0.24 | 0.24 ≤ to < 34.20 | ≥ 34.20 | | | |
| | ASM/BMI | 0.85 ± 0.004 | 0.83 ± 0.004 | 0.85 ± 0.004 | 0.52 | 0.027 | 0.316 |
| Women | EPA+DHA, range | < 0.06 | 0.06 ≤ to < 0.40 | ≥ 0.40 | | | |
| | ASM/BMI | 0.56 ± 0.003 | 0.56 ± 0.003 | 0.57 ± 0.003 | 0.026 | 0.054 | 0.013 |
| | ALA, range | < 0.43 | 0.43 ≤ to < 0.95 | ≥ 0.95 | | | |
| | ASM/BMI | 0.56 ± 0.003 | 0.56 ± 0.003 | 0.57 ± 0.003 | 0.445 | <0.001 | 0.964 |
| | Fish, range | < 0.00 | 0.00 ≤ to < 15.33 | ≥ 15.33 | | | |
| | ASM/BMI | 0.55 ± 0.003 | 0.56 ± 0.003 | 0.57 ± 0.003 | 0.005 | 0.083 | <0.001 |
## 4. Discussion
The present study demonstrates that consumption of EPA, DHA, and fish were negatively associated with the prevalence of LLM and positively correlated with muscle mass in Korean older women, but not in men, after adjusting for potential confounders in KNHANES data from 2008 to 2011. Consistent with the present study, a higher total n-3 PUFA intake was associated with a lower risk of sarcopenia among kidney transplant patients in Brazil [5]. The risk of sarcopenia was also negatively associated with serum levels of total n-3 PUFA and decreased by $71\%$ with each standard deviation increment of serum level of total n-3 PUFA in Korean older adults [6]. Yang et al. [ 7] also showed that the ratio of daily total n-3 PUFA intake to energy intake was negatively associated with the risk of sarcopenic obesity in older Korean women, but not in men.
Dos Reis et al. [ 5] observed that the intake of total n-3 PUFA was positively associated with muscle mass, estimated by ASM divided by the height squared, in kidney transplant patients. The Maastricht Sarcopenia Study of a community-dwelling population found that older adults with sarcopenia had a significantly lower intake of total n-3 PUFA and lower erythrocyte EPA levels than those without sarcopenia [12]. A meta-analysis of clinical trials reported that supplementation with EPA and DHA elicited an approximately 0.33 kg increase in muscle mass in older adults, especially when more than 2 g/day of EPA and DHA was administered [17]. n-3 PUFA increases the rate of muscle protein synthesis by stimulating the mTOR signaling pathway, and prevents muscle degradation by reducing inflammatory signaling pathway signaling [18].
Epidemiologic studies have consistently reported that the intake of total n-3 PUFA, EPA, and DHA was associated with muscle function measured by leg strength, time to rise from a chair, timed up-and-go, gait speed, handgrip strength, and SPPB in American older adults [8], Japanese men [9], Finnish women [10], and Korean women [11]. Similarly, high plasma concentrations of total n-3 PUFA, EPA, and DHA were associated with gait speed, handgrip strength, and SPPB among older adults in a Three-City study [13] as well as in Korea [6], United States [15], and Italy [14]. In particular, our previous study showed that erythrocyte EPA and DHA levels are associated with gait speed in Korean older adults [16]. A meta-analysis of clinical trials found that supplementation with EPA and DHA improved gait speed in older adults [17]. However, Rossato et al. [ 31] reported that intake of EPA and DHA was not associated with muscle strength or voluntary peak isokinetic knee extensor strength in American older adults, and their average intake of EPA and DHA was 0.1 g/day. On the other hand, the average intake of EPA and DHA was 0.6 g/day among Korean older adults in the present study, which was six times higher than that in Americans. The Food and Agriculture Organization of the United Nations reported that the intake of aquatic food, a rich source of n-3 PUFA, was more than 50 kg per capita per year among Koreans, the highest in the world, and only 20-30 kg per capita per year among Americans [32].
To our knowledge, no study has investigated the association between fish intake and muscle mass. However, previous studies have reported that fish intake is positively associated with muscle function. Muscle function measured by grip strength was positively associated with consumption of fatty fish and oily fish, but not white fish and shells, in older adults from the Hertfordshire cohort [19] and from the UK Biobank study [20]. In addition, a Mediterranean diet, known to contain many fish, was negatively associated with the risk of sarcopenia and positively associated with gait speed in older Iranian adults [22]. The Finnish Osteoporosis Risk Factor and Prevention Fracture Prevention Study also found that the Mediterranean diet score was significantly associated with gait speed in older women [21]. Rondanelli et al. [ 33] suggested that fish contain anti-sarcopenic compounds, such as n-3 PUFA, proteins, vitamin D, magnesium, and carnitine, which could reduce inflammation and improve muscle response to exercise and diet.
The Osteoporosis Risk Factor and Prevention Fracture Prevention Study reported that ALA intake was positively associated with muscle function, but not muscle mass, among Finnish older women [10]. The Maastricht Sarcopenia *Study analysis* found that ALA intake was significantly associated with muscle function, but erythrocyte levels of ALA, a marker for dietary intake, were not associated with muscle function in Dutch older adults [12]. In addition, supplementation of 14 g/day of ALA with a resistance training program increased knee flexor muscle thickness and had an effect on muscle functions, such as chest and leg press, in Canadian older men [23]. However, the increased muscle thickness might not be due to the intake of ALA but to exercise, since all participants in the trial were on resistance training programs [23]. Thus, the documented effect of ALA intake on muscle function is inconsistent and might not be associated with muscle mass, which supports the results of the present study.
A noteworthy point of our findings was that intake of EPA, DHA, and fish was associated with the prevalence of LLM and muscle mass in older women, but not in older men. Consistent with the present study, a higher ratio of daily total n-3 PUFA intake to energy intake was negatively associated with the risk of sarcopenic obesity [7], and intake of EPA and DHA was positively associated with handgrip strength [11] in Korean older women, but not in men, suggesting that intake of n-3 PUFA might be beneficial for sarcopenia among Korean women. In the present study, intakes of EPA and DHA, and ALA were analyzed separately instead of n-3 PUFA, and appendicular skeletal muscle mass was evaluated instead of handgrip strength. Asian people, especially Asian women, tend to have lower muscle mass and higher body fat mass with central adiposity than Western populations [34, 35]. In the present study, the average BMI and WC of sarcopenic women were 26 kg/m2 and 87 cm, respectively, indicating that sarcopenic women were mostly abdominal obese, but sarcopenic men were not. Thus, in the present study, muscle mass was calculated based on ASM divided by BMI, but not by height squared (m2), to consider sarcopenic obesity. Previous epidemiological studies have shown that patients with sarcopenic obesity have higher levels of cytokines, such as C-reactive protein (CRP), interleukin-6 (IL-6), and monocyte chemotactic protein-1 than those with sarcopenia (36–38). Inflammation is a well-known component of the pathophysiology of muscle wasting [39]. A meta-analysis of clinical trials showed that n-3 PUFA supplementation reduced inflammatory biomarkers such as CRP and IL-6 in older adults [40]. Al-Safi et al. [ 41] further reported that supplementation with EPA and DHA decreased the levels of IL-1 and tumor necrosis factor-alpha (TNF-α), and the reduction in cytokines was greater among obese women than among normal-weight women, suggesting that n-3 PUFA could be more beneficial for those with inflammation. Additionally, women in general have a greater capacity to convert ALA to DHA than men due to estrogen; thus, the plasma level of DHA is higher in women than in men [42]. Canon et al. [ 43] reported that estrogen decreases the levels of CRP and IL-6 and inhibits chronic inflammation. Furthermore, Smith et al. [ 44] observed that supplementation of EPA and DHA enhanced muscle anabolism when plasma leucine concentrations were clamped at 165-175 μmol/L in healthy adults. However, McGlory et al. [ 45] found that supplementation with EPA and DHA had no effect on muscle protein synthesis during peak plasma leucine concentrations of 250-300 μmol/L achieved by the ingestion of 30 g of whey protein in young men. It is possible that ingestion of 30 g of whey protein could maximize the rate of muscle protein synthesis to the extent that fish oil supplementation would not have exerted a further anabolic influence [46, 47]. Thus, previous studies have suggested that a beneficial effect of n-3 PUFA on muscle synthesis might be observed when protein is insufficient (44–47). In the present study, the average protein intake was 62 g/day in men aged 65 years or older, similar to the Korean dietary reference intake (KDRIs) of 60 g/day, but the average protein intake was 45 g/day in women aged 65 years or older, which is lower than the KDRIs of 50 g/day [48, 49]. Additionally, protein intake was lower among older men with LLM (57 g/day vs. 63 g/day; $$p \leq 0.002$$) and women with LLM (43 g/day vs. 45 g/day; $$p \leq 0.015$$) than those without LLM, although protein intake as g/kcal/day was not different between men with and without LLM (33 mg/kcal vs. 33 mg/kcal; $$p \leq 0.573$$) and between women with and without LLM (31 mg/kcal vs. 31 mg/kcal; $$p \leq 0.460$$). As Korean women consume insufficient amounts of protein, the beneficial effect of n-3 PUFA on LLM might be observed only in Korean women.
The major strength of the present study was that the data were gathered from a nationally representative survey throughout Korea; thus, the findings can be generalized to older adults in Korea. However, some limitations should be considered when interpreting the results of this study. First, the cross-sectional study design was unable to establish a causal relationship between the prevalence of LLM and the intake of n-3 PUFA and fish. Second, muscle strength or performance were not measured in the KNHANES 2008-2011, and sarcopenia could not be diagnosed. However, loss of muscle mass with aging is clinically important because it leads to diminished strength and exercise capacity [50]. Third, the dietary intake of one day was assessed using the 24-h recall method, which could have recall bias and did not reflect the usual dietary intake.
## 5. Conclusion
The present study demonstrates that consumption of high levels of EPA, DHA, and fish could have beneficial effects on the prevention of LLM by improving muscle mass in older women. Further studies are needed to verify the preventive effects of EPA, DHA, and fish consumption on sarcopenia in large population-based longitudinal studies of diverse ethnic origins.
## 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 below: https://knhanes.kdca.go.kr/knhanes/sub03/sub03_01.do.
## Ethics statement
The studies involving human participants were reviewed and approved by Institutional Review Board of Hanyang University (HYUIRB-202208-003). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
YK performed statistical analyses and wrote the manuscript. YP designed the study, revised the manuscript, and was responsible for this work. Both 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/fnut.2023.1119719/full#supplementary-material
## References
1. Cruz-Jentoft A, Sayer A. **Sarcopenia.**. (2019) **393** 2636-46. DOI: 10.1016/S0140-6736(19)31138-9
2. Ganapathy A, Nieves J. **Nutrition and sarcopenia—what do we know?**. (2020) **12**. DOI: 10.3390/nu12061755
3. Robinson S, Reginster J, Rizzoli R, Shaw S, Kanis J, Bautmans I. **Does nutrition play a role in the prevention and management of sarcopenia?**. (2018) **37** 1121-32. DOI: 10.1016/j.clnu.2017.08.016
4. Dupont J, Dedeyne L, Dalle S, Koppo K, Gielen E. **The role of omega-3 in the prevention and treatment of sarcopenia.**. (2019) **31** 825-36. DOI: 10.1007/s40520-019-01146-1
5. Dos Reis A, Limirio L, Santos H, de Oliveira E. **Intake of polyunsaturated fatty acids and ω-3 are protective factors for sarcopenia in kidney transplant patients.**. (2021) **81**. DOI: 10.1016/j.nut.2020.110929
6. Jang I, Jung H, Park J, Kim J, Lee S, Lee E. **Lower serum n-3 fatty acid level in older adults with sarcopenia.**. (2020) **12**. DOI: 10.3390/nu12102959
7. Yang W, Lee J, Kim Y, Lee J, Kang H. **Increased omega-3 fatty acid intake is inversely associated with sarcopenic obesity in women but not in men, based on the 2014–2018 Korean National Health and Nutrition Examination Survey.**. (2020) **9**. DOI: 10.3390/jcm9123856
8. Rousseau J, Kleppinger A, Kenny A. **Self-reported dietary intake of omega-3 fatty acids and association with bone and lower extremity function.**. (2009) **57** 1781-8. DOI: 10.1111/j.1532-5415.2008.01870.x
9. Takayama M, Arai Y, Sasaki S, Hashimoto M, Shimizu K, Abe Y. **Association of marine-origin N-3 polyunsaturated fatty acids consumption and functional mobility in the community-dwelling oldest old.**. (2013) **17** 82-9. DOI: 10.1007/s12603-012-0389-1
10. Isanejad M, Tajik B, McArdle A, Tuppurainen M, Sirola J, Kröger H. **Dietary omega-3 polyunsaturated fatty acid and alpha-linolenic acid are associated with physical capacity measure but not muscle mass in older women 65–72 years.**. (2022) **61** 1813-21. DOI: 10.1007/s00394-021-02773-z
11. Bae Y, Cui X, Shin S. **Increased omega-3 fatty acid intake is associated with low grip strength in elderly Korean females.**. (2022) **14**. DOI: 10.3390/nu14122374
12. Ter Borg S, de Groot L, Mijnarends D, de Vries J, Verlaan S, Meijboom S. **Differences in nutrient intake and biochemical nutrient status between sarcopenic and nonsarcopenic older adults – results from the Maastricht Sarcopenia Study.**. (2016) **17** 393-401. DOI: 10.1016/j.jamda.2015.12.015
13. Frison E, Boirie Y, Peuchant E, Tabue Teguo M, Barberger Gateau P, Féart C. **Plasma fatty acid biomarkers are associated with gait speed in community-dwelling older adults: the three-city-Bordeaux study.**. (2017) **36** 416-22. DOI: 10.1016/j.clnu.2015.12.008
14. Abbatecola A, Cherubini A, Guralnik J, Lacueva C, Ruggiero C, Maggio M. **Plasma polyunsaturated fatty acids and age-related physical performance decline.**. (2009) **12** 25-32. DOI: 10.1089/rej.2008.0799
15. Hutchins-Wiese H, Kleppinger A, Annis K, Liva E, Lammi-Keefe C, Durham H. **The impact of supplemental n-3 long chain polyunsaturated fatty acids and dietary antioxidants on physical performance in postmenopausal women.**. (2013) **17** 76-80. DOI: 10.1007/s12603-012-0415-3
16. Kim D, Won C, Park Y. **Association between erythrocyte levels of n-3 polyunsaturated fatty acids and risk of frailty in community-dwelling older adults: the Korean Frailty and aging Cohort study.**. (2021) **76** 499-504. DOI: 10.1093/gerona/glaa042
17. Huang Y, Chiu W, Hsu Y, Lo Y, Wang Y. **Effects of omega-3 fatty acids on muscle mass, muscle strength and muscle performance among the elderly: a meta-analysis.**. (2020) **12**. DOI: 10.3390/nu12123739
18. McGlory C, Calder P, Nunes E. **The influence of omega-3 fatty acids on skeletal muscle protein turnover in health, disuse, and disease.**. (2019) **6**. DOI: 10.3389/fnut.2019.00144
19. Robinson S, Jameson K, Batelaan S, Martin H, Syddall H, Dennison E. **Diet and its relationship with grip strength in community-dwelling older men and women: the Hertfordshire cohort study.**. (2008) **56** 84-90. DOI: 10.1111/j.1532-5415.2007.01478.x
20. Gedmantaite A, Celis-Morales C, Ho F, Pell J, Ratkevicius A, Gray S. **Associations between diet and handgrip strength: a cross-sectional study from UK Biobank.**. (2020) **189**. DOI: 10.1016/j.mad.2020.111269
21. Isanejad M, Sirola J, Mursu J, Rikkonen T, Kröger H, Tuppurainen M. **Association of the Baltic Sea and Mediterranean diets with indices of sarcopenia in elderly women,**. (2018) **57** 1435-48. DOI: 10.1007/s00394-017-1422-2
22. Hashemi R, Motlagh A, Heshmat R, Esmaillzadeh A, Payab M, Yousefinia M. **Diet and its relationship to sarcopenia in community dwelling Iranian elderly: a cross sectional study.**. (2015) **31** 97-104. DOI: 10.1016/j.nut.2014.05.003
23. Cornish S, Chilibeck P. **Alpha-linolenic acid supplementation and resistance training in older adults.**. (2009) **34** 49-59. DOI: 10.1139/h08-136
24. Kweon S, Kim Y, Jang M, Kim Y, Kim K, Choi S. **Data resource profile: the Korea national health and nutrition examination survey (KNHANES).**. (2014) **43** 69-77. DOI: 10.1093/ije/dyt228
25. Studenski S, Peters K, Alley D, Cawthon P, McLean R, Harris T. **The FNIH sarcopenia project: rationale, study description, conference recommendations, and final estimates.**. (2014) **69** 547-58. DOI: 10.1093/gerona/glu010
26. 26.National Institute of Fisheries Science.
Composition Table of Marine Products in Korea 2018.
8th ed. Busan: National Institute of Fisheries Science (2018). p. 205–91.. (2018) p. 205-91
27. Park S, Kim S, Lee S, Choe J, Choi Y. **Development of 9th revision Korean food composition table and its major changes.**. (2018) **23** 352-65. DOI: 10.5720/kjcn.2018.23.4.352
28. Saylor J, Friedmann E, Lee H. **Navigating complex sample analysis using national survey data.**. (2012) **61** 231-7. DOI: 10.1097/NNR.0b013e3182533403
29. Greenland S, Pearce N. **Statistical foundations for model-based adjustments.**. (2015) **36** 89-108. DOI: 10.1146/annurev-publhealth-031914-122559
30. Willett W, Howe G, Kushi L. **Adjustment for total energy intake in epidemiologic studies.**. (1997) **65** 1220S-8S. DOI: 10.1093/ajcn/65.4.1220S
31. Rossato L, de Branco F, Azeredo C, Rinaldi A, de Oliveira E. **Association between omega-3 fatty acids intake and muscle strength in older adults: a study from National Health and Nutrition Examination Survey (NHANES) 1999-2002.**. (2020) **39** 3434-41. DOI: 10.1016/j.clnu.2020.03.001
32. 32.FAO.
In Brief to The State of World Fisheries and Aquaculture 2022. Towards blue transformation.
Rome: FAO (2022). 10.4060/cc0461en. (2022). DOI: 10.4060/cc0461en
33. Rondanelli M, Rigon C, Perna S, Gasparri C, Iannello G, Akber R. **Novel insights on intake of fish and prevention of sarcopenia: all reasons for an adequate consumption.**. (2020) **12**. DOI: 10.3390/nu12020307
34. Wu Y, Hwang A, Liu L, Peng L, Chen L. **Sex differences of sarcopenia in Asian populations: the implications in diagnosis and management.**. (2016) **7** 37-43. DOI: 10.1016/j.jcgg.2016.04.001
35. Lim U, Ernst T, Buchthal S, Latch M, Albright C, Wilkens L. **Asian women have greater abdominal and visceral adiposity than Caucasian women with similar body mass index.**. (2011) **1**. DOI: 10.1038/nutd.2011.2
36. 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
37. Batsis J, Mackenzie T, Jones J, Lopez-Jimenez F, Bartels S. **Sarcopenia, sarcopenic obesity and inflammation: results from the 1999–2004 national health and nutrition examination survey.**. (2016) **35** 1472-83. DOI: 10.1016/j.clnu.2016.03.028
38. Lim J, Leung B, Ding Y, Tay L, Ismail N, Yeo A. **Monocyte chemoattractant protein-1: a proinflammatory cytokine elevated in sarcopenic obesity.**. (2015) **10** 605-9. DOI: 10.2147/cia.S78901
39. Dalle S, Rossmeislova L, Koppo K. **The role of inflammation in age-related sarcopenia.**. (2017) **8**. DOI: 10.3389/fphys.2017.01045
40. Custodero C, Mankowski R, Lee S, Chen Z, Wu S, Manini T. **Evidence-based nutritional and pharmacological interventions targeting chronic low-grade inflammation in middle-age and older adults: a systematic review and meta-analysis.**. (2018) **46** 42-59. DOI: 10.1016/j.arr.2018.05.004
41. Al-Safi Z, Liu H, Carlson N, Chosich J, Harris M, Bradford A. **Omega-3 fatty acid supplementation lowers serum FSH in normal weight but not obese women.**. (2016) **101** 324-33. DOI: 10.1210/jc.2015-2913
42. Giltay E, Gooren L, Toorians A, Katan M, Zock P. **Docosahexaenoic acid concentrations are higher in women than in men because of estrogenic effects.**. (2004) **80** 1167-74. DOI: 10.1093/ajcn/80.5.1167
43. Canon M, Crimmins E. **Sex differences in the association between muscle quality, inflammatory markers, and cognitive decline.**. (2011) **15** 695-8. DOI: 10.1007/s12603-011-0340-x
44. Smith G, Atherton P, Reeds D, Mohammed B, Rankin D, Rennie M. **Omega-3 polyunsaturated fatty acids augment the muscle protein anabolic response to hyperinsulinaemia–hyperaminoacidaemia in healthy young and middle-aged men and women.**. (2011) **121** 267-78. DOI: 10.1042/CS20100597
45. McGlory C, Wardle S, Macnaughton L, Witard O, Scott F, Dick J. **Fish oil supplementation suppresses resistance exercise and feeding-induced increases in anabolic signaling without affecting myofibrillar protein synthesis in young men.**. (2016) **4**. DOI: 10.14814/phy2.12715
46. Moore D, Robinson M, Fry J, Tang J, Glover E, Wilkinson S. **Ingested protein dose response of muscle and albumin protein synthesis after resistance exercise in young men.**. (2008) **89** 161-8. DOI: 10.3945/ajcn.2008.26401
47. Witard O, Jackman S, Breen L, Smith K, Selby A, Tipton K. **Myofibrillar muscle protein synthesis rates subsequent to a meal in response to increasing doses of whey protein at rest and after resistance exercise.**. (2013) **99** 86-95. DOI: 10.3945/ajcn.112.055517
48. 48.Ministry of Health and Welfare.
Dietary Reference Intakes for Koreans : Energy and Macronutrients.
Sejong: Ministry of Health and Welfare (2020). p. 6.. (2020) p. 6
49. 49.The Division of Chronic Disease Surveillance, Korea Centers for Disease Control and Prevention, The Fourth and Fifth Korea National Health and Nutrition Examination Survey (KNHANES IV & V). (2008–2011). Available online at: https://knhanes.kdca.go.kr/knhanes/sub03/sub03_01.do
(accessed September 7, 2021).. (2008–2011)
50. Rosenberg I. **Sarcopenia: origins and clinical relevance.**. (1997) **127** 990S-1S. DOI: 10.1093/jn/127.5.990S
|
---
title: Vitamin D supplementation is effective for olanzapine-induced dyslipidemia
authors:
- Zijian Zhou
- Takuya Nagashima
- Chihiro Toda
- Mone Kobayashi
- Takahide Suzuki
- Kazuki Nagayasu
- Hisashi Shirakawa
- Satoshi Asai
- Shuji Kaneko
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9989177
doi: 10.3389/fphar.2023.1135516
license: CC BY 4.0
---
# Vitamin D supplementation is effective for olanzapine-induced dyslipidemia
## Abstract
Olanzapine is an atypical antipsychotic drug that is clinically applied in patients with schizophrenia. It increases the risk of dyslipidemia, a disturbance of lipid metabolic homeostasis, usually characterized by increased low-density lipoprotein (LDL) cholesterol and triglycerides, and accompanied by decreased high-density lipoprotein (HDL) in the serum. In this study, analyzing the FDA Adverse Event Reporting System, JMDC insurance claims, and electronic medical records from Nihon University School of Medicine revealed that a co-treated drug, vitamin D, can reduce the incidence of olanzapine-induced dyslipidemia. In the following experimental validations of this hypothesis, short-term oral olanzapine administration in mice caused a simultaneous increase and decrease in the levels of LDL and HDL cholesterol, respectively, while the triglyceride level remained unaffected. Cholecalciferol supplementation attenuated these deteriorations in blood lipid profiles. RNA-seq analysis was conducted on three cell types that are closely related to maintaining cholesterol metabolic balance (hepatocytes, adipocytes, and C2C12) to verify the direct effects of olanzapine and the functional metabolites of cholecalciferol (calcifediol and calcitriol). Consequently, the expression of cholesterol-biosynthesis-related genes was reduced in calcifediol- and calcitriol-treated C2C12 cells, which was likely to be mediated by activating the vitamin D receptor that subsequently inhibited the cholesterol biosynthesis process via insulin-induced gene 2 regulation. This clinical big-data-based drug repurposing approach is effective in finding a novel treatment with high clinical predictability and a well-defined molecular mechanism.
## Introduction
Schizophrenia is a group of chronic psychiatric disorders characterized by hallucinations, delusions, reduced motivation, and blunt affect (Saha et al., 2005). Atypical antipsychotics are currently used as first-line drugs in patients with schizophrenia because they potently improve positive symptoms without causing extrapyramidal symptoms (Campbell et al., 1999). Among these, olanzapine is one of the most widely used drugs (Kochi et al., 2017). However, olanzapine causes dyslipidemia as a significant adverse effect, characterized by an increase in total blood cholesterol and triglyceride levels (Lieberman et al., 2005). A recent meta-analysis further revealed a relationship between olanzapine treatment and an increase in low-density lipoprotein (LDL) cholesterol in the blood (Li et al., 2020). Although olanzapine has excellent cardiac safety (Kennedy et al., 2001), the obesity induced by olanzapine treatment has been proven to be a cardiovascular risk factor (Wang et al., 2014; Correll et al., 2015), thereby making its treatment a significant risk factor that contributes to excess premature mortality in patients with schizophrenia (Brown et al., 2000). At present, preventing this adverse effect is impossible due to the incomplete understanding of the underlying mechanisms of olanzapine-induced dyslipidemia (Yan et al., 2013). Thus, there exists a clinical demand for novel target-screening approaches.
Drug repurposing from market-accessible drugs is a classic approach for refining the potential therapeutic applications. Repurposing studies have shown that topiramate, a newer anticonvulsant, was able to attenuate the obesity-inducing effect of olanzapine (Narula et al., 2010), and diabetes-treating drugs, such as metformin (Chen et al., 2008) and liraglutide (Larsen et al., 2017) have relieved the excessive body weight gain due to olanzapine treatment. However, a recently published meta-analysis review on the clinical trials focusing on blood lipid levels concluded that the available lipid-lowering agents are not effective in treating patients who are prescribed with atypical antipsychotics, including olanzapine (Kanagasundaram et al., 2021). Therefore, current practices need to be optimized, considering that existing drug repurposing approaches are inefficient in refining interventions with high clinical efficiency for treating olanzapine-induced dyslipidemia.
In this study, an updated approach, “reverse translational drug repurposing” (Kaneko and Nagashima, 2020), was adopted, making it viable to conduct a retrospective analysis with clinical big data and screen for hypothetical drug-drug pairs that potently ameliorate olanzapine-induced dyslipidemia in patients. We have previously utilized the self-reports of adverse events extracted from the FDA Adverse Events Reporting System (FAERS) to study the underlying mechanisms of hyperglycemia induced by an atypical antipsychotic, quetiapine, and identified the agent with treatment potential (Nagashima et al., 2016). Furthermore, to improve the drawbacks of FAERS (lacking time stamps and the population size of patients who were prescribed the specified drugs), another source of real-world clinical big data, insurance claims from JMDC Inc., was utilized in our recently published studies (Nagaoka et al., 2021; Siswanto et al., 2021). In the present study, in addition to the FAERS and JMDC databases, electronic medical records stored in the clinical data warehouse of Nihon University School of Medicine (NUSM’s CDW) containing detailed demographic, diagnostic, and laboratory data from patients at three hospitals affiliated with NUSM (Nagashima et al., 2022) were also used. Electronic medical records can provide continuous clinical information, leading to unattainably precise outcomes of the onset pattern of specific drug-induced adverse effects, treatment significance, and validated medication safety. To summarize, the corroborating results from the analysis of the three data sources mentioned above enable a more sensitive retrospective analysis for detecting causal relationships between the resultant adverse events and specific drug treatment, and the lowered incidence of adverse events and potential rescuing drugs causing this change. This hypothesis has been validated in animal experiments.
## Analysis of the FAERS database
Adversary event reports from 2004 to 2019 were downloaded from the FDA website (https://www.fda.gov/drugs/drug-approvals-anddatabases/fda-adverse-event-reporting-system-faers). Duplicate reports were deleted as previously reported (Banda et al., 2016), and 11,438,031 preserved reports were analyzed in the present study. Arbitrary drug names, including trade names and abbreviations, were manually mapped to unified generic names using the Medical Subject Headings (MeSH) descriptor ID. Reports of dyslipidemia were defined according to the preferred terms listed in Supplementary Table S1 of MedDRA (version 23.0). The FAERS data analysis was performed as previously described (Nagashima et al., 2016). Adversary event risk was evaluated by calculating the reporting odds ratio (ROR) along with the $95\%$ confidence interval (CI) and Z-score (see Supplementary Tables S2, S3 for details). Z-scores were applied instead of p-values to save space graphically.
## Analyzing the JMDC claims database
Insurance claims data collected by JMDC Inc. from January 2005 to March 2018 were purchased, which contained the medical and prescription claims of 5,550,241 individuals and their dependents on a monthly basis. Health checkup data, including blood test results, body mass index (BMI), and waist circumferences, were provided by 2,278,697 individuals. The patients were mainly aged ≤ 65 years, and no patients aged ≥ 75 years were included. The drug names were coded using the Anatomical Therapeutic Chemical Classification System.
On retrospectively analyzing the profile of the first event onsets, the users of olanzapine were identified as those who were prescribed olanzapine more than 2 months after being included in the JMDC claims database. Patients without specific health checkup data were also excluded. Thus, 1,853 patients with olanzapine were included in the present study. Among them, vitamin D users were defined as patients whose first prescription of vitamin D was ahead of olanzapine treatment. The pre-prescription period for olanzapine was defined as the period within 12 months, while the post-prescription period was defined as the period within 12 months after initiating olanzapine treatment.
The blood test results included in the present study were triglyceride, LDL cholesterol, and high-density lipoprotein (HDL) cholesterol levels apart from BMI and waist circumstances. Test results were collected for each individual on the nearest date before initiating olanzapine treatment in the pre-prescription period and on the closest date in the post-prescription period after initiating olanzapine treatment. To reduce bias in population backgrounds, the propensity score matching method (greedy 1:1 matching) was used by balancing covariates between settings. The propensity score for vitamin D was obtained by fitting a logistic regression model that included all covariates of interest (Table 2). The propensity score of each patient with or without vitamin D treatment was subsequently matched using the nearest neighbor method. Using the matched outcomes, an unpaired two-tailed t-test with Welch’s correction for continuous variables and Fisher’s exact test for categorical data were conducted to compare the differences in baseline characteristics between patients with or without vitamin D exposure.
## Analysis of the NUSM electronic medical records
Electronic medical records were obtained from the Nihon University School of Medicine’s Clinical Data Warehouse (NUSM’s CDW). This database includes detailed diagnostic, demographic, and laboratory data for inpatients and outpatients. We received written informed consent and agreement for the secondary use after anonymization at three hospitals affiliated with the NUSM (Nagashima et al., 2022). Similar to the JMDC claims data analysis, the pre-prescription period of olanzapine was defined as the period within 12 months of the first prescription. Contrastingly, the post-prescription period was defined as the period within 12 months after initiating olanzapine treatment. Among them, preceding users of vitamin D were defined as patients whose first prescription of vitamin D was ahead of the first prescription of olanzapine for at least 1 day. Matching was not performed in this study because there were no significant differences in the baseline characteristics (Table 3). The blood laboratory values were extracted from the database, and the baseline value (month 0) was selected from the nearest test prior to the first prescription of olanzapine. The values at each checkpoint (months 3, 6, 9, and 12) were selected from the furthest test results from the first prescription of olanzapine. The missing values were imputed using the last observation carried forward method. An unpaired two-tailed t-test with Welch’s correction for continuous variables and Fisher’s exact test for categorical data were conducted to compare the differences in baseline characteristics between patients with or without vitamin D exposure.
## Animals
All animal experiments were approved by the Kyoto University Animal Research Committee in accordance with the ethical guidelines of the committee. All experiments were designed to minimize the use of animals and the number of required experiments. Male and female C57BL/6J mice were purchased from Japan SLC (Shizuoka, Japan) and housed at a constant ambient temperature (24 ± 1°C) and humidity ($55\%$ ± $10\%$) on a $\frac{12}{12}$ h light/dark cycle. Mice were fed an ad libitum diet consisting of water and chow.
## Reagents and treatments (in vivo)
Olanzapine was purchased from the Tokyo Chemical Industry (Tokyo, Japan). The medium-fat diet (containing 1.37 IU of cholecalciferol/g) and cholecalciferol-supplemented medium-fat diet (containing 200 IU of cholecalciferol/g) were purchased from Oriental Yeast (Tokyo, Japan). Olanzapine was dissolved in water with $0.5\%$ of carboxymethyl cellulose before use. Cholecalciferol is usually administered orally as a chow supplement, and the intake dose may vary slightly among mice. Mice were first randomized into two groups and fed a medium-fat diet or a cholecalciferol-supplemented medium-fat diet for 1 week. Mice from each group were further randomized to be treated with olanzapine (10 mg/kg, orally administered) or vehicle ($0.5\%$ of carboxymethyl cellulose solution) for another 5 days. One day after the last dose of olanzapine, the mice were anesthetized and dissected for blood collection by cardiac puncture. Mouse serum was isolated from blood samples by centrifugation. Serum total cholesterol levels were measured using a LabAssay Cholesterol kit (Wako, Osaka, Japan), while serum LDL cholesterol, HDL cholesterol, and triglyceride levels were measured by Nagahama Lifescience (Oriental Yeast Co., Ltd., Shiga, Japan).
## Cell culture
Primary mouse hepatocytes were prepared as previously described (Charni-Natan and Goldstein, 2020; Jung et al., 2020) with modifications. Livers from 6 to 8 weeks old male mice were perfused with Hanks’ balanced salt solution (HBSS) without calcium, magnesium, and phenol red, and subsequently supplemented with 0.5 mM of ethylenediaminetetraacetic acid (EDTA) and 25 mM of hydroxyethyl piperazineethanesulfonic acid (HEPES). The livers were then perfused with a digestive enzyme mix [collagenase, type I (Worthington, NJ, United States) 0.15 mg/mL; collagenase, type II (Worthington, NJ, United States) 0.15 mg/mL; Dispase, type II (Gibco, MA, United States) 0.15 mg/mL] solution, dissolved in standard HBSS with phenol red, and supplemented with 25 mM of HEPES. Hepatocytes were released into high-glucose Dulbecco’s modified Eagle’s medium (DMEM; D5796; Sigma-Aldrich, MO, United States) from digested livers, filtered through 70 μm cell strainers (Corning, NY, United States), and further purified using a $45\%$ Percoll® cushion by density gradient centrifugation. The purified cells were resuspended in William’s E medium (A1217601; Gibco) supplemented with $5\%$ of fetal bovine serum (FBS; Sigma-Aldrich), 1 μM of dexamethasone (Nacalai Tesque, Kyoto, Japan), $1\%$ of Penicillin–Streptomycin Mixed Solution (P/S; Nacalai Tesque), 5 μg/mL of human recombinant insulin (Sigma-Aldrich), 2 mM of GlutaMAX supplement (Sigma-Aldrich), and 15 mM of HEPES. Cells were plated onto dishes coated with 0.1 mg/mL of collagen type I (Nippi, Tokyo, Japan) and incubated at 37°C in a humidified chamber containing $5\%$ CO2 for 3 h. After letting cells attach to the surface, the medium was refreshed with serum-free William’s E supplemented with 1 μM of dexamethasone, $0.5\%$ of P/S, $1\%$ of ITS + premix (Corning), 2 mM of GlutaMAX, and 15 mM of HEPES. Cells were used for treatment within 48 h.
Primary mouse adipocytes were prepared as previously described (Lequeux et al., 2009; Galmozzi et al., 2021) with modifications. Subcutaneous adipose tissues were collected from neonatal mouse pups and digested with 0.1 mg/mL collagenase type II and 0.1 mg/mL dispase type II dissolved in standard HBSS. Digested tissues were homogenized by pipetting, filtered through 70 μm cell strainers, and resuspended in high-glucose DMEM supplemented with $20\%$ of FBS, $1\%$ of P/S, and 10 mM of HEPES. The collected preadipocytes were plated onto dishes coated with $0.1\%$ of gelatin (Nacalai Tesque) in distilled water. Cells were flushed twice with HBSS and refreshed with the same medium. After the cells reached $90\%$ confluence, they were transferred into high-glucose DMEM supplemented with $10\%$ of FBS, $1\%$ of P/S, 10 mM of HEPES, 170 nM of human recombinant insulin, 1 μM of dexamethasone, 0.5 mM of 3-isobutyl-1-methylxanthine, 1 nM of triiodothyronine, and 10 nM of hydrocortisone for differentiation into adipocytes. After 48 h, the cells were transferred to high-glucose DMEM supplemented with $10\%$ of FBS, $1\%$ of P/S, 10 mM of HEPES, 170 nM of human recombinant insulin, and 1 nM of triiodothyronine. After 5 days of maintenance, the cells were ready for treatment.
Mouse C2C12 cells were prepared as previously described (Nagashima et al., 2016) with some modifications. The cells were obtained from Prof. H. Takeshima (Kyoto University Graduate School of Pharmaceutical Sciences, Kyoto, Japan). Cells were thawed in a water bath at 37°C and plated in high-glucose DMEM supplemented with $10\%$ of FBS, $1\%$ of P/S, and 10 mM of HEPES. After reaching $70\%$ confluence, the cells were digested with $0.25\%$ of trypsin solution (Nacalai Tesque) and passaged in the same medium. When the cells reached $90\%$ confluence, the medium was changed to high-glucose DMEM supplemented with $2\%$ of horse serum (Sigma-Aldrich), $1\%$ of P/S, and 10 mM of HEPES for differentiation. The medium was refreshed 3 days after the changes, and the cells were ready on day five.
## Reagents and treatments (in vitro)
Calcifediol was purchased from Selleckchem (TX, United States) and calcitriol was purchased from Cayman (MI, United States. ZK159222 was purchased from Cayman. All reagents were reconstructed and preserved in DMSO as a vehicle at −80°C, and thus DMSO was added as a blank reference in control groups. Primary mouse hepatocytes were treated with maintenance medium. Primary mouse adipocytes and C2C12 cells were treated in Advanced DMEM/F12 medium (12634010, Gibco) supplemented with $1\%$ of P/S, 2 mM of GlutaMAX supplement, and 10 mM of HEPES.
## RNA-seq and quantitative reverse transcriptase polymerase chain reaction (qRT-PCR)
RNA expression levels in cultured cells were evaluated using RNA-seq and qRT-PCR. Total RNA was isolated from cultured cells following the standard protocols provided by Isogen reagents (Nippon Gene, Tokyo, Japan).
For RNA-seq analysis, poly(A)+ RNA was selected from total RNA and sequenced using DNBseq (BGI, Shenzhen, China). Total reads were filtered by SOAPnuke (Cock et al., 2010) and clean reads were mapped to the mouse reference genome GRCm38.p6 using HISAT, version 2.0.4 (Kim et al., 2015) and Bowtie, version 2.2.5 (Langmead and Salzberg, 2012) in parallel. Gene expression was calculated using RSEM, version 1.2.8 (Li and Dewey, 2011). *The* gene set related to cholesterol biosynthesis was obtained from MGI (http://www.informatics.jax.org/vocab/gene_ontology/GO:0006695) (Hayamizu et al., 2005). All the genes from the three types of cultured cells (hepatocytes, adipocytes, and C2C12 cells) were mapped to this gene set to generate a list of cell-specific cholesterol biosynthetic process-related genes. Among the mapped genes, those with low expression levels (transcript per million (TPM) < 1) in vehicle-treated cells were removed from the analysis. Differentially expressed genes were detected between vehicle and olanzapine, olanzapine and olanzapine-calcifediol co-treated cells, and olanzapine and olanzapine-calcitriol co-treated cells based on PossionDis (Audic and Claverie, 1997). Genes with fold-change (FC) ≥ 2 and false discovery rate (FDR) ≤ 0.05 were recognized as differentially expressed. The analyzed results were visualized using volcano plots drawn using Prism 9.4.1 (GraphPad Software, CA, United States), with −log10 (FDR) in the y-direction and log2(FC) in the x-direction. The original sequence datasets were deposited in the NCBI sequence read archive with the accession number GSE221683.
For qRT-PCR, the extracted RNA was translated to cDNA using ReverTra Ace (Toyobo, Osaka, Japan) and subjected to StepOne real-time PCR (Life Technologies, Carlsbad, CA, United States) with Thunderbird SYBR qPCR Mix (Toyobo). The amplification process was set as follows: 10 min at 95°C, followed by 40 cycles of looping from 15 s at 95°C to 60 s at 60°C. Oligonucleotide primers were purchased from Sigma-Aldrich, and the sequences were as follows (gene name:5′-forward-3′, 5′-reverse-3′): Mus ribosomal protein, large, P0 (Rplp0): GCT TCG TGT TCA CCA AGG A, GTC CTA GAC CAG TGT TCT GAG G; Mus 3-hydroxy-3-methylglutaryl-CoA reductase (Hmgcr): GCT CGT CTA CAG AAA CTC CAC G, GCT TCA GCA GTG CTT TCT CCG T; Mus insulin-induced gene 2 (Insig2): GCC CAT CCA GAA CCT CTG AC, AGA CGG GGC AAA AGG ACT TC. The expression levels of each mRNA were normalized to those of Rplp0 mRNA.
## Western blotting
Western blotting was performed as previously described (Ohashi et al., 2018) with some modifications. Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer (Nacalai Tesque), diluted to 1 μg/mL with $1\%$ sodium dodecyl sulfate (SDS) aqueous solution, and then buffered with NuPAGE® lithium dodecyl sulfate (LDS) sample buffer (4×; Life Technologies, CA, United States). The samples were loaded onto a $10\%$ SDS-polyacrylamide gel and blotted onto a ClearTrans® PVDF Membrane (Wako). After blocking with Blocking One (Nacalai Tesque), the membranes were split into two pieces to simultaneously blot the target protein, vitamin D receptor (VDR), and the endo-reference protein, glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The membranes were incubated overnight at 4°C with anti-VDR antibody (1:500 dilution, 12,550, Cell Signaling Technology, MA, United States) and anti-GAPDH (1:50,000 dilution, 32,233, Santa Cruz, CA, United States) in Tris-buffered saline supplemented with $0.1\%$ of Tween 20 (TBS-T) and $10\%$ of Blocking One. After washing with TBS-T, the membranes were incubated with peroxidase-conjugated donkey anti-rabbit IgG (1:5000 dilution, NA934V, GE Healthcare, IL, United States) and Peroxidase AffiniPure Goat Anti-Mouse IgG (1:5000 dilution, 115-035-003; Jackson ImmunoResearch, PA, United States) for 2 h at RT. Specific bands were detected using Immobilon Western Chemiluminescent HRP Substrate (Millipore) and visualized using EZCapture MG (ATTO, Tokyo, Japan). The expression level of VDR was normalized to that of GAPDH.
## Statistics
Statistical analysis of the FAERS database was performed using SQL Software (IBM Research Laboratory, NY, United States) and that of the JMDC database was performed using R version 3.5.1 Software (The R Foundation for Statistical Computing, Vienna, Austria). In the JMDC database, paired t-tests were used to compare the mean values within the pre-prescription and post-prescription periods. Differences in continuous variables between the two groups were compared using an unpaired two-tailed t-test with Welch’s correction. In the NUSM database, mixed-effects models were used to analyze repeated measures data.
Statistical analysis of the animal experiments was performed using GraphPad Prism 9.3.1, and the data are expressed as mean ± standard error of the mean (SEM). The blood profile data were analyzed using two-way analysis of variance (ANOVA) with post hoc Tukey’s multiple comparison test. Revalidation using qRT-PCR of the C2C12 cell line on RNA-seq resultant data was analyzed by two-way ANOVA. The dose-dependent changes in specific genes responding to calcifediol, calcitriol, and ZK159222 in the C2C12 cell line were analyzed by one-way ANOVA. The western blotting results of VDR cell-specific enrichment were analyzed using Student’s t-test. All reported p-values of < 0.05 were considered statistically significant.
## Analysis of FAERS
First, the association between the treatment with a particular drug and the incidence of dyslipidemia in the FAERS database was investigated using disproportionality analysis by calculating the ROR and Z-score of each drug (Figure 1A). The known reporting bias and lack of incidence denominators accompanied by self-reports (Alatawi and Hansen, 2017) only allowed these values to be limited demonstrations of the real-world incidence rate. Nevertheless, the high ROR and Z-score of olanzapine indicated that it was one of the drugs that exhibited the strongest association between its use and the onset of dyslipidemia (the complete set of data is provided in Supplementary Table S2). Olanzapine was selected for the present study because of its higher ROR than that of other atypical antipsychotics (Table 1).
**FIGURE 1:** *Increased incidence of dyslipidemia with the prescription of drugs and confounding effects of concomitant drugs on the olanzapine-induced dyslipidemia in FDA Adverse Event Reporting System (FAERS) data. Volcano plots for visualizing the reporting odds ratio (ROR, on a log scale) and its statistical significance (absolute Z-score) are shown. Each circle indicates an individual drug, and the size of the circle reflects the number of patients taking the drug. (A) Strong and significant increases in the ROR for dyslipidemia were seen in patients using olanzapine. (B) Within the population taking olanzapine, confounding effects of concomitantly used drugs on the incidence of olanzapine-induced dyslipidemia were calculated thoroughly and plotted. Overall values are presented in (Supplementary Tables S2, S3).* TABLE_PLACEHOLDER:TABLE 1 The confounding effects of all drug combinations used in the olanzapine-treated patients were also calculated using the ROR and Z-score (Figure 1B). Vitamin D was found to be impressively effective in lowering the incidence of olanzapine-induced dyslipidemia. Although vitamin D was not among the drugs with the lowest ROR, which suggested higher effects on dyslipidemia, vitamin D had the highest Z-score due to its sufficient number of cases (the complete set of data is provided in Supplementary Table S3). However, vitamin D itself was associated with a slightly increased risk of dyslipidemia according to the FAERS data (ROR = 2.55, Z-score = 39.2; also see Supplementary Table S2), which may be due to the lack of time information to establish an accurate causal relationship between the reports and the effects of vitamin D. Thus, further analyses to reinforce the present findings were performed using the JMDC data.
## Analysis of JMDC insurance claims
To further investigate whether the resultant dyslipidemia is associated with the clinical consequences of olanzapine treatment, the JMDC insurance claims data were introduced for analysis. Blood test data from the JMDC insurance claims of olanzapine-exposed 1,853 patients with or without vitamin D treatment were extracted ($$n = 22$$ for with vitamin D, $$n = 1$$,831 for without vitamin D). Since vitamin D is predominately prescribed to female elderly, a balancing approach, propensity score matching, was applied to eliminate possible systemic bias. After adjusting for populations, no significant differences in baseline characteristics between patients with or without vitamin D treatment (Table 2) were detected, and the batch size was balanced ($$n = 20$$). The effect of vitamin D on propensity score-matched data was investigated. In both groups of patients exposed only to olanzapine and to olanzapine and vitamin D simultaneously, the triglyceride level was not influenced (Figure 2A); however, the level of LDL cholesterol significantly increased by 11 mg/dL in the olanzapine-only group, while this value in patients with vitamin D co-treatment did not significantly increase (Figure 2B). Furthermore, the HDL cholesterol level in the blood significantly decreased by 5 mg/dL in the olanzapine-only group, while vitamin D supplementation significantly reversed the declining trend and even slightly increased HDL cholesterol levels by 4 mg/dL (Figure 2C). These results indicate that vitamin D counteracted the influence of olanzapine on blood lipid profiles.
In addition to blood lipid profiles, the effects of vitamin D on body mass, fat distribution, and blood glucose levels in olanzapine-treated patients were investigated by analyzing related data from JMDC insurance claims. Within the same groups providing blood test data, BMI, an indicator used to roughly estimate whether one is overweight (as the higher BMI refers to a higher risk of being overweight), significantly increased by olanzapine treatment, while the co-treatment of vitamin D diminished this trend (Figure 2D). The mean value of the waist circumference, an approximate indicator of body fat mass, also significantly increased in olanzapine-only treated patients but not in olanzapine-vitamin D co-treated individuals (Figure 2E); however, the blood glucose indicator, HbA1c, was neither influenced in the olanzapine-only treated patients nor in the olanzapine-vitamin D co-treated individuals (Figure 2F). To summarize, these results indicate that using olanzapine aggravates the lipid profile, condition of being overweight, and excessive fat accumulation, while the combined use of vitamin D has beneficial effects on all three health aspects to some extent or at least provides no unfavorable effects. The blood and body test data in the JMDC insurance claims were annually collected from medical examinations provided by the Japan Health Insurance Association, which limited the precision of the data of the onset and development of olanzapine-induced dyslipidemia. Thus, to improve this flaw, electronic medical records were obtained from NUSM’s CDW, which is recorded in units of days, allowing tracing of blood profile changes on a smaller but more precise scale than that of the JMDC insurance claims data.
## Analysis of electronic medical records from NUSM’s CDW
The blood test data of patients exposed to olanzapine with or without vitamin D treatment, including TG, LDL cholesterol, and HDL cholesterol values, were obtained from the electronic medical records from NUSM’s CDW. No significant bias was detected between patients with and without vitamin D supplementation (Table 3). By continuously tracking the blood lipid profile changes within 1 year after the first exposure to olanzapine, a significant increase in the blood triglyceride level was detected as early as 6 months after initiating the treatment and kept increasing by 54 mg/dL at the end of the year, while co-treatment with vitamin D significantly diminished this trend and remained stable (Figure 3A). A significant increase in blood LDL cholesterol in patients without vitamin D supplementation was detected not earlier than 12 months from the first exposure to olanzapine and increased by 15 mg/dL, but this trend was not observed in patients with vitamin D supplementation (Figure 3B). Contrastingly, HDL cholesterol in the blood of patients without vitamin D supplementation significantly reduced by 5 mg/dL after 6 months of follow-up, while vitamin D supplementation especially prevented this risk from developing (Figure 3C). Based on NUSM’s CDW electronic medical recordings, the beneficial effects of vitamin D treatments on olanzapine-induced dyslipidemia can also be detected within a shorter time scale in contrast to the insurance claims from the JMDC.
## Effects of vitamin D3 on olanzapine-induced dyslipidemia mouse model
Previous studies have reported non-alcoholic fatty liver disease and LDL cholesterol or HDL cholesterol imbalance in the blood phenotypes of olanzapine-treated wild-type mouse models (Chen et al., 2018; Jiang et al., 2019; Liu et al., 2019; Zhu et al., 2022) under chronic treatment (at least 4 weeks). In the present study, in order to study the primary stage of dyslipidemia development, the treatment was limited to 5 days (Figure 4A). To mimic the calorie intake of human patients, mice were provided with $30\%$ of fructose in drinking water during the acclimatization period of 1 week and this was continued in the following runs of treatment (Faeh et al., 2005). The results showed that neither a scheduled administration of olanzapine (10 mg/kg, orally administered by gavage, daily for 5 days) nor vitamin D supplementation (administrated via a cholecalciferol (vitamin D3)-supplemented diet containing 200 IU of cholecalciferol/g) caused changes in blood triglyceride (Figure 4B) and total cholesterol levels (Figure 4C). However, a significant increase in the blood LDL cholesterol level and the attenuating effects of vitamin D supplementation were detected (Figure 4D). In addition, the reduction in HDL cholesterol level caused by olanzapine treatment was attenuated by cholecalciferol supplementation (Figure 4E). These findings suggest that short-term administration of olanzapine is sufficient to induce dyslipidemia by challenging the balance between blood LDL and HDL cholesterol levels without influencing blood triglyceride and total cholesterol levels. Simultaneously, cholecalciferol as dietary supplementation diminished these effects.
**FIGURE 4:** *Olanzapine treatment caused dyslipidemia in a rodent model while vitamin D3 supplements attenuated the worsening of the condition. (A) C57BL/6J mice were fed for 1 week with the medium-fat diet (containing 1.37 IU of cholecalciferol/g of chow) or cholecalciferol-supplemented medium-fat diet (containing 200 IU of cholecalciferol/g of chow). Each group of mice were divided into two groups and orally given 10 mg/kg/day of olanzapine or vehicle for 5 days; resulting 4 groups of mice (each n = 10) received normal chow and vehicle as control (C), normal chow and olanzapine (O), vitamin D3-supplemented chow and vehicle (D), and vitamin D3-supplemented chow and olanzapine (DO). Blood samples were collected from these mice 1 day after the last dose of olanzapine without fasting. (B) the triglycerides, (C) total cholesterol levels, (D) LDL cholesterol levels, and (E) HDL cholesterol levels were measured. Data are shown as means ± SEM. The multiple comparisons of olanzapine and cholecalciferol influences were compared by two-way analysis of variance (ANOVA) with post hoc Tukey’s multiple comparison test. *p < 0.05, **p < 0.01.*
## Molecular mechanisms of vitamin D on olanzapine-induced dyslipidemia
To investigate the molecular mechanisms underlying cholesterol dyshomeostasis in blood accompanied by sub-chronic olanzapine and the rescuing effects of cholecalciferol supplementation, changes in expression of genes in cells related to cholesterol metabolism were analyzed using RNA-seq. Purified primary cultures of mouse hepatocytes, adipocytes, and fully differentiated C2C12 cells (myotubes) were used instead of liver, adipose, and skeletal muscle tissues, since these tissues are generally the mixtures of various cells with different origins and functions, which may blur the direct effects of olanzapine and vitamin D metabolites. All three types of cells were treated under the same condition, a 24 h pre-exposure to the circulating form of vitamin D3, 25-hydroxycholecarciferol, (calcifediol at 10 μM) or the active form of vitamin D3, 1,25-dihydroxycholecarciferol, (calcitriol at 0.1 μM), followed by another 12 h co-treatment with olanzapine (1 μM). From the MGI database, only cholesterol biosynthesis-related genes were selected for analysis (Hayamizu et al., 2005) because only cholesterol homeostasis was disrupted in the blood according to in vivo modeling. Comparing these gene expressions under different treatments in various types of cells with PossionDis (Audic and Claverie, 1997), the expression of a set of genes was significantly changed in olanzapine-calcifediol-treated C2C12 cells and slightly influenced in olanzapine-calcitriol-treated C2C12 cells. However, olanzapine displayed passive effects in all three cell types (Figure 5A). The regulatory relationship within this set of genes is illustrated in Figure 5B. In this scheme, Insig2, a suppressive regulator of cholesterol biosynthesis (Yabe et al., 2002), was upregulated by vitamin D metabolites, while genes participating in the mevalonate pathway (DeBose-Boyd, 2008), such as 3-hydroxy-3-methylglutaryl-CoA reductase (Hmgcr), 3-hydroxy-3-methylglutaryl-coenzyme A synthase 1 (Hmgcs1), farnesyl diphosphate synthetase (Fdps), farnesyl diphosphate farnesyl transferase 1 (Fdft1), lanosterol synthase (Lss), cytochrome P450, family 51 (Cyp51), NAD(P) dependent steroid dehydrogenase-like (Nsdhl), and 24-dehydrocholesterol reductase (Dhcr24) were significantly downregulated.
**FIGURE 5:** *Volcano plots of gene expression in the cultured mice cells treated with vehicle, olanzapine, or olanzapine co-treated with calcifediol or calcitriol. The changing ranges were expressed as log2-transformed fold-change (FC), and the significance was expressed as −log10-transformed false discovery rate (FDR). (A) Demonstrations of the cholesterol biosynthesis-related gene changes under olanzapine treatment and calcifediol or calcitriol-olanzapine co-treatment. The genes significantly downregulated in the olanzapine-calcifediol co-treated C2C12 cells are tagged with blue dots, and the genes upregulated are tagged with red dots in all the data sets. (B) A schematic diagram illustrating the functions of downregulated genes (blue) and upregulated gene (red) in the process of cholesterol biosynthesis. Genes plotted with FC absolute value > 2 and FDR value > 0.05 were considered to be significantly changed by treatment.*
To validate these findings from the RNA-seq analysis, an enlarged batch of evaluations was performed on C2C12 cells, which displayed the highest sensitivity to vitamin D metabolites, using qRT-PCR. Hmgcr was selected as a representative gene for cholesterol biosynthesis because it is the most representative gene involved in the process of cholesterol biosynthesis (Horton et al., 2003). The qRT-PCR results revealed that both the suppression of Hmgcr expression (Figure 6A) and the induction of the expression of Insig2 (Figure 6B) by calcifediol and calcitriol were significant. A previous report claimed that calcifediol, but not calcitriol, independently reduced cholesterol biosynthesis without VDR actions (Asano et al., 2017). However, further validations by changing the concentration of either calcifediol or calcitriol in the treatment medium showed that the expression of Hmgcr (Figure 6C) and Insig2 (Figure 6D) changed in a dose-dependent manner under calcifediol treatment. In parallel, the changes in the expression of Hmgcr (Figure 6E) and Insig2 (Figure 6F) also displayed a dose-dependent effect in calcitriol treatment, with approximately 5.8 times higher sensitivity (IC50 of calcifediol on Hmgcr: 0.68 µM; IC50 of calcitriol on Hmgcr: 0.12 µM). Since calcitriol is the bioactive form of vitamin D, which acts only through VDR, the role of VDR in cholesterol synthesis suppression in this model should not be underestimated and requires further investigation.
**FIGURE 6:** *Validation by quantitative RT-PCR results showing the expression of cholesterol biosynthesis-associated genes in the cultured C2C12 cell line. (A) 3-hydroxy-3-methylglutaryl-CoA reductase (Hmgcr) and (B) Insulin-induced gene 2 (Insig2) were validated in the same treating condition as that in the RNA-seq experiments (n = 12). Results were normalized to ribosomal protein, large, P0 (Rplp0). (C)
Hmgcr and (D)
Insig2 were validated in calcifediol with gradient-changed concentrations increasing from 0 to 10 μM (n = 8–9). (E)
Hmgcr and (F)
Insig2 were validated in calcitriol with gradient-changed concentrations increasing from 0.01 μM to 1 μM. Data are shown as means ± SEM (n = 9). **p < 0.01, ****p < 0.0001. The comparisons of olanzapine and vitamin D influences were compared by two-way analysis of variance (ANOVA). The validation of the dose-dependent effects of calcifediol and calcitriol were compared by one-way ANOVA.*
To investigate whether the actions of calcifediol and calcitriol were mediated by VDR activity, western blotting of VDR levels in specific cell types was performed using RNA-seq. VDR was highly expressed in C2C12 cells, which was much higher than that in adipocytes and hepatocytes (Figure 7A). The varied tissue-specific enrichment of VDR indicates that suppression of Hmgcr expression is mediated by VDR activity. A partial agonist of VDR (Teske et al., 2016), ZK159222, was used to validate the action of VDR. ZK159222, like calcifediol and calcitriol, suppressed Hmgcr expression (Figure 7B) and induced Insig2 expression (Figure 7C) in a dose-dependent manner, further confirming the essential role of VDR activity.
**FIGURE 7:** *Validation by quantitative RT-PCR results showing calcitriol regulating the cholesterol biosynthesis-associated genes mediated by VDR in the C2C12 cell line. (A) The results of vitamin D receptor (VDR) protein levels in C2C12 cells, adipocytes, and hepatocytes, normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (n = 3). (B)
Hmgcr and (C)
Insig2 were validated in ZK159222 with gradient-changed concentrations increasing from 0 to 3 μM (n = 6). Data are shown as means ± SEM. *p < 0.05, **p < 0.01. The validation of the dose-dependent effects of ZK150222 was compared by one-way analysis of variance (ANOVA).*
## Discussion
In this study, we showed for the first time that vitamin D supplementation is effective in improving the lipid profiles in patients treated with olanzapine. This finding was revealed from three independent sources of clinical data. This hypothesis was validated by using the blood profiles of mice models fed with high-calorie diets, and the mechanism of action was investigated at the molecular level using cells that were metabolically active with cholesterol.
Drug repurposing using FAERS has been proven to be a reliable tool for refining unexpected drug-drug interactions with beneficial effects in treating adverse effects (Nagaoka et al., 2021; Siswanto et al., 2021). In the present study, olanzapine was one of the most dyslipidemia-inducing drugs causing a large ROR in more than tens of thousands of cases, according to the FAERS database, providing a solid foundation to further explore rescue drugs. Additionally, olanzapine had a greater tendency to induce severe dyslipidemia than other atypical antipsychotics that share similar molecular structures, such as clozapine and quetiapine, in the FAERS database (Table 1). Notably, patients are more heavily dosed with clozapine (Cmax > 1.07 μM) (Jann et al., 1993) and quetiapine (Cmax ≈ 0.21 μM) (DeVane and Nemeroff, 2001) than olanzapine (Cmax < 0.064 μM) (Callaghan et al., 1999) in clinical practice. These two facts prioritize screening drug targets for treating olanzapine-induced dyslipidemia from other atypical antipsychotics. Contrastingly, vitamin D treatment is recognized as a dyslipidemia risk factor other than olanzapine, which is likely caused by the confounding in identifying causes and results. Since patients with dyslipidemia are treated with this drug, the onset of the disease as per the FAERS data, is also likely to be positively related to the usage of this drug. This confusion is due to the lack of information regarding the prescription period to help define the time point of drug exposure and disease onset. However, cross-validations using health databases with records on prescription time can help solve this issue.
The hypothesis developed from the FAERS data analysis was verified in a retrospective cohort analysis based on the JMDC insurance claims data. The onset rate of dyslipidemia and the drug exposure period can be analyzed in units of years using the raw blood test results of patients under 75 years of age that the JMDC claims can annually provide as a supplement to the FAERS data. In the present study, propensity score matching revealed olanzapine-induced dyslipidemia and weight gain in the matched groups. However, the patient population characteristics are strongly sex-biased (primarily women), because vitamin D is predominantly prescribed to women in cases where they are more susceptible to osteoporosis (Bohon and Goolsby, 2013). Nevertheless, olanzapine has been reported to increase LDL cholesterol and triglyceride levels in adolescents (Kryzhanovskaya et al., 2009) and young adults (Narula et al., 2010), regardless of gender variance, indicating that the analysis results from JMDC claims still remain highly consistent with clinical trials. The present study found that vitamin D supplementation during the exposure period to olanzapine suppressed the increase in the LDL cholesterol level, decrease in the HDL cholesterol level, increase in BMI, and increase in waist circumference. This is the first study to demonstrate the clinical significance of vitamin D in improving blood lipid profiles, since only one previous trial has shown that vitamin D helps in maintaining waist circumference in olanzapine-treated patients (Kaviyani et al., 2017). Despite already being efficient in discovering the olanzapine-vitamin D interaction at the laboratory test level, the JMDC is still unable to precisely locate the onset time pattern of olanzapine-induced dyslipidemia due to the long data collection intervals. Further improvement was achieved by utilizing the electronic medical records from NUSM’s CDW, since the electronic records were updated daily and allowed for a flexible time-course setting. The blood test results were collected every 3 months after initiating the olanzapine treatment. The triglyceride and LDL cholesterol levels were reported to increase monotonically as the exposure time increased, and the appearance of abnormality in triglyceride levels was prior to that in the LDL and HDL cholesterol levels. Differentiating from JMDC claims, patients treated with olanzapine only exhibited a strong trend in developing triglyceride abnormality, while triglyceride levels from patients in JMDC claims were passively influenced. This inconsistency may be due to the variance in population background characteristics, since patients included in the electronic records from NUSM’s CDW patients were more balanced in terms of sex (female percentage = $61.3\%$) than those from the JMDC claims (female rate = $90.0\%$), and higher estrogen levels in females are generally believed to be the reason for better triglyceride control than males (Carr, 2003). Vitamin D treatment counteracted these deteriorative changes by maintaining the blood lipid profile at stable levels throughout the observation window, consistent with the outcomes of the JMDC claims. By combining these three outcomes based on different data sources, a causal relationship between lipid profiles’ deterioration and olanzapine treatment was detected in the present study, and the suppressive effects of vitamin D supplementation on these influences were also confirmed.
Using drug-induced animal models, significant efforts have been devoted to identifying the mechanism of olanzapine-induced dyslipidemia. In several rodent models, strong hepatoxicity, represented by non-alcoholic fatty liver diseases, has generally been reported after chronic treatments ranging from 4 to 12 weeks (Chen et al., 2018; Jiang et al., 2019; Liu et al., 2019; Zhu et al., 2022). However, these models cannot accurately represent clinical observations from patients, since olanzapine rarely causes clinically apparent hepatotoxicity (Larrey and Ripault, 2013). Furthermore, it generally takes years for non-alcoholic fatty liver disease to develop under constant environmental stress (Friedman et al., 2018). Contrastingly, according to the electronic medical records from NUSM’s CDW in the present study, the onset of dyslipidemia was as early as 6 months after initiating olanzapine treatment, which suggests that the development of olanzapine-induced dyslipidemia is not necessarily related to the onset of hepatic diseases. This assumption is supported by the in vivo observation in mouse models in the present study that olanzapine exposure as short as 1 week is already enough to cause the LDL and HDL cholesterol levels to increase and drop, respectively. A recent clinical study revealed an alternative explanation to olanzapine-induced dyslipidemia that an increased appetite is essential for developing olanzapine-induced dyslipidemia (Huang et al., 2020), and moreover, enhanced food reward circuitry has been detected in olanzapine-treated patients (Mathews et al., 2012). This phenotype was previously observed in rodent models and attributed to the antagonistic effects on histamine H1 receptors in the central nervous system by olanzapine (Kim et al., 2007; He et al., 2014). Nevertheless, a slight increase in the expression of genes related to cholesterol biosynthesis was detected in the hepatocytes in the present study. However, the changing ranges were too limited to be considered investigation-worthy. In conclusion, hepatotoxicity is not causally related to olanzapine-induced dyslipidemia and the direct influence of olanzapine on lipid metabolic processes in related tissues lacks physiological significance.
However, the role of vitamin D in lipid metabolism remains controversial. Several studies have shown that vitamin D is necessary for lipid synthesis and accumulation in adipose tissue. In VDR-knockout mouse models, the absence of VDR showed substantial protective effects against high-fat diet-induced adipose tissue accumulation (Narvaez et al., 2009; Wong et al., 2009), whereas in another knock-in model, mice with overexpressed human VDR in adipose tissue developed obesity even under a standard diet (Wong et al., 2011). Further investigation into this mechanism revealed that vitamin D potentiates subcutaneous preadipocyte differentiation via the VDR pathway. Simultaneously, the expression level of VDR significantly decreased as differentiation progressed until the full maturity of adipocytes (Nimitphong et al., 2012). Contrastingly, vitamin D alleviated atherosclerosis progression in patients by preventing foam cell formation (Oh et al., 2009) and suppressing oxidative stress in blood vessel endothelial cells (Kassi et al., 2013). However, the role of vitamin D in lipid homeostasis is not necessarily associated with maintaining the blood lipid profile. Clinical observations have reported improved blood triglyceride and LDL cholesterol levels in populations supplied with extra vitamin D, with varied background characteristics, according to two meta-analyses (Mirhosseini et al., 2018; Dibaba, 2019). The big data mining-based results in the present study are consistent with these findings. However, according to a previous study on the effects of calcium supplements on the blood lipid profile, the primary medical function of vitamin D, which improves calcium absorption, is not likely to be related to its role in treating dyslipidemia (Reid et al., 2010). To date, studies revealing the mechanisms of these clinical results are lacking. Nevertheless, a recent study established an inhibitory connection between vitamin D and lipid synthesis by proving that calcifediol, but not calcitriol, inhibits cholesterol and fatty acid biosynthesis by preventing sterol regulatory element binding transcription factors from maturing in a VDR-independent pattern (Asano et al., 2017). However, this result was not reproduced in primary hepatocytes and adipocytes, which may be due to the vast metabolic variance between highly differentiated primary cells and immortalized cells in that report (CHO cell line). Another study suggested that VDR-mediated Insig2 expression by calcitriol in HepG2 cells mainly contributed to its suppressive effects on cholesterol synthesis (Li et al., 2016). This result is doubtful because VDR is absent in normal hepatocytes (Gascon-Barre et al., 2003). This understanding is consistent with the lack of VDR expression in primary cultured hepatocytes in the western blot results of the present study. Nevertheless, the VDR activation-induced Insig2 expression increase is reproduced in this study. Insig2 is an essential regulator of cholesterol biosynthesis that not only prevents sterol regulatory element binding transcription factors from entering the cell nucleus to suppress their transcription function (Yabe et al., 2002) but also facilitates HMGCR degradation (Jo et al., 2011). A functional vitamin D response element has been localized in the promoter of the mouse Insig2 (Lee et al., 2005), and further comparisons between rodent and human Insig2 promoters have confirmed that this vitamin D response element is also included in the human genome (Fernandez-Alvarez et al., 2010). Although in contrast to the liver, the role of the muscle tissues is minor in the essentiality of cholesterol production and balance, cholesterol production in peripheral tissues should not be overlooked. A report has shown myopathy development in skeletal muscle-specific Hmgcr knockout mice (Osaki et al., 2015). Additionally, humans with skeletal muscle Hmgcr deficiency easily develop myopathy (Lorenzo-Villalba et al., 2021). Furthermore, as VDR is widely expressed in tissues throughout the body, vitamin D may improve the blood cholesterol condition by limiting tissue self-biosynthesis of cholesterol and simultaneously inducing cholesterol absorption in peripheral tissues. A primary limitation of this research is the failure to clarify the mechanism of olanzapine-induced dyslipidemia through investigations on lipid metabolism. Future explorations will focus on the neurological role of olanzapine in increasing appetites. Additionally, the investigation on vitamin D target tissues and organs is also limited, since the enterocytes in the intestine are also essential for lipid absorption and highly enriched with VDR. The possible role of enterocytes is also scheduled for future studies.
Although the exploration of the potential of VDR-mediated cholesterol biosynthesis suppression in new druggable target development still has a long way to cover, from the clinical findings in this research, it is safe to suggest vitamin D supplements in patients under olanzapine treatment, especially in those diagnosed as vitamin D deficient. Moreover, this treatment is also recommended to be simultaneously applied with a calorie-limited diet for patients under olanzapine treatment for a better outcome.
## 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 below: NCBI Gene Expression Omnibus (GEO) [https://www.ncbi.nlm.nih.gov/geo/], GSE221683.
## Ethics statement
The animal study protocol was reviewed and approved by the Kyoto University Animal Research Committee.
## Author contributions
TN and SK designed the study. TN, CT, and TS performed clinical data analysis. ZZ and MK conducted the experiments and analyzed the data. ZZ and SK wrote the manuscript. KN, HS, SA, and SK provided materials and technical advice.
## 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 potential conflicts 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.2023.1135516/full#supplementary-material
## References
1. Alatawi Y. M., Hansen R. A.. **Empirical estimation of under-reporting in the U.S. Food and drug administration adverse event reporting system (FAERS)**. *Expert Opin. Drug Saf.* (2017) **16** 761-767. DOI: 10.1080/14740338.2017.1323867
2. Asano L., Watanabe M., Ryoden Y., Usuda K., Yamaguchi T., Khambu B.. **Vitamin D metabolite, 25-hydroxyvitamin D, regulates lipid metabolism by inducing degradation of SREBP/SCAP**. *Cell Chem. Biol.* (2017) **24** 207-217. DOI: 10.1016/j.chembiol.2016.12.017
3. Audic S., Claverie J. M.. **The significance of digital gene expression profiles**. *Genome Res.* (1997) **7** 986-995. DOI: 10.1101/gr.7.10.986
4. Banda J. M., Evans L., Vanguri R. S., Tatonetti N. P., Ryan P. B., Shah N. H.. **A curated and standardized adverse drug event resource to accelerate drug safety research**. *Sci. Data* (2016) **3** 160026. DOI: 10.1038/sdata.2016.26
5. Bohon T. M., Goolsby M. A.. **The role of vitamin D supplements in women's health**. *Clin. Med. Insights Womens Health* (2013) **6** 67-70. DOI: 10.4137/CMWH.S11067
6. Brown S., Inskip H., Barraclough B.. **Causes of the excess mortality of schizophrenia**. *Br. J. Psychiatry* (2000) **177** 212-217. DOI: 10.1192/bjp.177.3.212
7. Callaghan J. T., Bergstrom R. F., Ptak L. R., Beasley C. M.. **Olanzapine. pharmacokinetic and pharmacodynamic profile**. *Clin. Pharmacokinet.* (1999) **37** 177-193. DOI: 10.2165/00003088-199937030-00001
8. Campbell M., Young P. I., Bateman D. N., Smith J. M., Thomas S. H.. **The use of atypical antipsychotics in the management of schizophrenia**. *Br. J. Clin. Pharmacol.* (1999) **47** 13-22. DOI: 10.1046/j.1365-2125.1999.00849.x
9. Carr M. C.. **The emergence of the metabolic syndrome with menopause**. *J. Clin. Endocrinol. Metab.* (2003) **88** 2404-2411. DOI: 10.1210/jc.2003-030242
10. Charni-Natan M., Goldstein I.. **Protocol for primary mouse hepatocyte isolation**. *Star. Protoc.* (2020) **1** 100086. DOI: 10.1016/j.xpro.2020.100086
11. Chen C. H., Chiu C. C., Huang M. C., Wu T. H., Liu H. C., Lu M. L.. **Metformin for metabolic dysregulation in schizophrenic patients treated with olanzapine**. *Prog. Neuropsychopharmacol. Biol. Psychiatry* (2008) **32** 925-931. DOI: 10.1016/j.pnpbp.2007.11.013
12. Chen C. H., Shyue S. K., Hsu C. P., Lee T. S.. **Atypical antipsychotic drug olanzapine deregulates hepatic lipid metabolism and aortic inflammation and aggravates atherosclerosis**. *Cell Physiol. Biochem.* (2018) **50** 1216-1229. DOI: 10.1159/000494573
13. Cock P. J., Fields C. J., Goto N., Heuer M. L., Rice P. M.. **The Sanger FASTQ file format for sequences with quality scores, and the Solexa/Illumina FASTQ variants**. *Nucleic Acids Res.* (2010) **38** 1767-1771. DOI: 10.1093/nar/gkp1137
14. Correll C. U., Joffe B. I., Rosen L. M., Sullivan T. B., Joffe R. T.. **Cardiovascular and cerebrovascular risk factors and events associated with second-generation antipsychotic compared to antidepressant use in a non-elderly adult sample: Results from a claims-based inception cohort study**. *World Psychiatry* (2015) **14** 56-63. DOI: 10.1002/wps.20187
15. DeBose-Boyd R. A.. **Feedback regulation of cholesterol synthesis: Sterol-accelerated ubiquitination and degradation of HMG CoA reductase**. *Cell Res.* (2008) **18** 609-621. DOI: 10.1038/cr.2008.61
16. DeVane C. L., Nemeroff C. B.. **Clinical pharmacokinetics of quetiapine: An atypical antipsychotic**. *Clin. Pharmacokinet.* (2001) **40** 509-522. DOI: 10.2165/00003088-200140070-00003
17. Dibaba D. T.. **Effect of vitamin D supplementation on serum lipid profiles: A systematic review and meta-analysis**. *Nutr. Rev.* (2019) **77** 890-902. DOI: 10.1093/nutrit/nuz037
18. Faeh D., Minehira K., Schwarz J. M., Periasamy R., Park S., Tappy L.. **Effect of fructose overfeeding and fish oil administration on hepatic de novo lipogenesis and insulin sensitivity in healthy men**. *Diabetes* (2005) **54** 1907-1913. DOI: 10.2337/diabetes.54.7.1907
19. Fernandez-Alvarez A., Soledad Alvarez M., Cucarella C., Casado M.. **Characterization of the human insulin-induced gene 2 (INSIG2) promoter: The role of Ets-binding motifs**. *J. Biol. Chem.* (2010) **285** 11765-11774. DOI: 10.1074/jbc.M109.067447
20. Friedman S. L., Neuschwander-Tetri B. A., Rinella M., Sanyal A. J.. **Mechanisms of NAFLD development and therapeutic strategies**. *Nat. Med.* (2018) **24** 908-922. DOI: 10.1038/s41591-018-0104-9
21. Galmozzi A., Kok B. P., Saez E.. **Isolation and differentiation of primary white and Brown preadipocytes from newborn Mice**. *J. Vis. Exp.* (2021) **167** e62005. DOI: 10.3791/62005
22. Gascon-Barre M., Demers C., Mirshahi A., Neron S., Zalzal S., Nanci A.. **The normal liver harbors the vitamin D nuclear receptor in nonparenchymal and biliary epithelial cells**. *Hepatology* (2003) **37** 1034-1042. DOI: 10.1053/jhep.2003.50176
23. Hayamizu T. F., Mangan M., Corradi J. P., Kadin J. A., Ringwald M.. **The adult mouse anatomical dictionary: A tool for annotating and integrating data**. *Genome Biol.* (2005) **6** R29. DOI: 10.1186/gb-2005-6-3-r29
24. He M., Zhang Q., Deng C., Wang H., Lian J., Huang X. F.. **Hypothalamic histamine H1 receptor-AMPK signaling time-dependently mediates olanzapine-induced hyperphagia and weight gain in female rats**. *Psychoneuroendocrinology* (2014) **42** 153-164. DOI: 10.1016/j.psyneuen.2014.01.018
25. Horton J. D., Shah N. A., Warrington J. A., Anderson N. N., Park S. W., Brown M. S.. **Combined analysis of oligonucleotide microarray data from transgenic and knockout mice identifies direct SREBP target genes**. *Proc. Natl. Acad. Sci. U. S. A.* (2003) **100** 12027-12032. DOI: 10.1073/pnas.1534923100
26. Huang J., Hei G.-R., Yang Y., Liu C.-C., Xiao J.-M., Long Y.-J.. **Increased appetite plays a key role in olanzapine-induced weight gain in first-episode schizophrenia patients**. *Front. Pharmacol.* (2020) **11** 739. DOI: 10.3389/fphar.2020.00739
27. Jann M. W., Grimsley S. R., Gray E. C., Chang W. H.. **Pharmacokinetics and pharmacodynamics of clozapine**. *Clin. Pharmacokinet.* (1993) **24** 161-176. DOI: 10.2165/00003088-199324020-00005
28. Jiang T., Zhang Y., Bai M., Li P., Wang W., Chen M.. **Up-regulation of hepatic fatty acid transporters and inhibition/down-regulation of hepatic OCTN2 contribute to olanzapine-induced liver steatosis**. *Toxicol. Lett.* (2019) **316** 183-193. DOI: 10.1016/j.toxlet.2019.08.013
29. Jo Y., Lee P. C., Sguigna P. V., DeBose-Boyd R. A.. **Sterol-induced degradation of HMG CoA reductase depends on interplay of two Insigs and two ubiquitin ligases, gp78 and Trc8**. *Proc. Natl. Acad. Sci. U. S. A.* (2011) **108** 20503-20508. DOI: 10.1073/pnas.1112831108
30. Jung Y., Zhao M., Svensson K. J.. **Isolation, culture, and functional analysis of hepatocytes from mice with fatty liver disease**. *Star. Protoc.* (2020) **1** 100222. DOI: 10.1016/j.xpro.2020.100222
31. Kanagasundaram P., Lee J., Prasad F., Costa-Dookhan K. A., Hamel L., Gordon M.. **Pharmacological interventions to treat antipsychotic-induced dyslipidemia in schizophrenia patients: A systematic review and meta analysis**. *Front. Psychiatry* (2021) **12** 642403. DOI: 10.3389/fpsyt.2021.642403
32. Kaneko S., Nagashima T.. **Drug repositioning and target finding based on clinical evidence**. *Biol. Pharm. Bull.* (2020) **43** 362-365. DOI: 10.1248/bpb.b19-00929
33. Kassi E., Adamopoulos C., Basdra E. K., Papavassiliou A. G.. **Role of vitamin D in atherosclerosis**. *Circulation* (2013) **128** 2517-2531. DOI: 10.1161/CIRCULATIONAHA.113.002654
34. Kaviyani S., Bahadoram M., Houshmand G., Bahadoram S.. **Therapeutic impact of cholecalciferol in patients with psychiatric disorders receiving olanzapine**. *J. Parathyr. Dis.* (2017) **6** 19-22. DOI: 10.15171/jpd.2018.07
35. Kennedy J. S., Bymaster F. P., Schuh L., Calligaro D. O., Nomikos G., Felder C. C.. **A current review of olanzapine's safety in the geriatric patient: From pre-clinical pharmacology to clinical data**. *Int. J. Geriatr. Psychiatry* (2001) **16** S33-S61. DOI: 10.1002/1099-1166(200112)16:1+<::aid-gps571>3.0.co;2-5
36. Kim S. F., Huang A. S., Snowman A. M., Teuscher C., Snyder S. H.. **From the cover: Antipsychotic drug-induced weight gain mediated by histamine H**. *Proc. Natl. Acad. Sci. U. S. A.* (2007) **104** 3456-3459. DOI: 10.1073/pnas.0611417104
37. Kim D., Langmead B., Salzberg S. L.. **HISAT: A fast spliced aligner with low memory requirements**. *Nat. Methods* (2015) **12** 357-360. DOI: 10.1038/nmeth.3317
38. Kochi K., Sato I., Nishiyama C., Tanaka-Mizuno S., Doi Y., Arai M.. **Trends in antipsychotic prescriptions for Japanese outpatients during 2006-2012: A descriptive epidemiological study**. *Pharmacoepidemiol. Drug Saf.* (2017) **26** 642-656. DOI: 10.1002/pds.4187
39. Kryzhanovskaya L., Schulz S. C., McDougle C., Frazier J., Dittmann R., Robertson-Plouch C.. **Olanzapine versus placebo in adolescents with schizophrenia: A 6-week, randomized, double-blind, placebo-controlled trial**. *J. Am. Acad. Child. Adolesc. Psychiatry* (2009) **48** 60-70. DOI: 10.1097/CHI.0b013e3181900404
40. Langmead B., Salzberg S. L.. **Fast gapped-read alignment with Bowtie 2**. *Nat. Methods* (2012) **9** 357-359. DOI: 10.1038/nmeth.1923
41. Larrey D., Ripault M.-P.. **Hepatotoxicity of psychotropic drugs and drugs of abuse**. *Drug-induced liver disease* (2013) 443-462
42. Larsen J. R., Vedtofte L., Jakobsen M. S. L., Jespersen H. R., Jakobsen M. I., Svensson C. K.. **Effect of liraglutide treatment on prediabetes and overweight or obesity in clozapine- or olanzapine-treated patients with schizophrenia spectrum disorder: A randomized clinical trial**. *JAMA Psychiatry* (2017) **74** 719-728. DOI: 10.1001/jamapsychiatry.2017.1220
43. Lee S., Lee D. K., Choi E., Lee J. W.. **Identification of a functional vitamin D response element in the murine Insig-2 promoter and its potential role in the differentiation of 3T3-L1 preadipocytes**. *Mol. Endocrinol.* (2005) **19** 399-408. DOI: 10.1210/me.2004-0324
44. Lequeux C., Auxenfans C., Mojallal A., Sergent M., Damour O.. **Optimization of a culture medium for the differentiation of preadipocytes into adipocytes in a monolayer**. *Biomed. Mater Eng.* (2009) **19** 283-291. DOI: 10.3233/BME-2009-0593
45. Li B., Dewey C. N.. **RSEM: Accurate transcript quantification from RNA-seq data with or without a reference genome**. *BMC Bioinforma.* (2011) **12** 323. DOI: 10.1186/1471-2105-12-323
46. Li S., He Y., Lin S., Hao L., Ye Y., Lv L.. **Increase of circulating cholesterol in vitamin D deficiency is linked to reduced vitamin D receptor activity via the Insig-2/SREBP-2 pathway**. *Mol. Nutr. Food Res.* (2016) **60** 798-809. DOI: 10.1002/mnfr.201500425
47. Li R., Zhang Y., Zhu W., Ding C., Dai W., Su X.. **Effects of olanzapine treatment on lipid profiles in patients with schizophrenia: A systematic review and meta-analysis**. *Sci. Rep.* (2020) **10** 17028. DOI: 10.1038/s41598-020-73983-4
48. Lieberman J. A., Stroup T. S., McEvoy J. P., Swartz M. S., Rosenheck R. A., Perkins D. O.. **Effectiveness of antipsychotic drugs in patients with chronic schizophrenia**. *N. Engl. J. Med.* (2005) **353** 1209-1223. DOI: 10.1056/NEJMoa051688
49. Liu X. M., Zhao X. M., Deng C., Zeng Y. P., Hu C. H.. **Simvastatin improves olanzapine-induced dyslipidemia in rats through inhibiting hepatic mTOR signaling pathway**. *Acta Pharmacol. Sin.* (2019) **40** 1049-1057. DOI: 10.1038/s41401-019-0212-1
50. Lorenzo-Villalba N., Andres E., Meyer A.. **Chronic onset form of anti-HMG-CoA reductase myopathy**. *Eur. J. Case Rep. Intern Med.* (2021) **8** 002672. DOI: 10.12890/2021_002672
51. Mathews J., Newcomer J. W., Mathews J. R., Fales C. L., Pierce K. J., Akers B. K.. **Neural correlates of weight gain with olanzapine**. *Arch. Gen. Psychiatry* (2012) **69** 1226-1237. DOI: 10.1001/archgenpsychiatry.2012.934
52. Mirhosseini N., Rainsbury J., Kimball S. M.. **Vitamin D supplementation, serum 25(OH)D concentrations and cardiovascular disease risk factors: A systematic review and meta-analysis**. *Front. Cardiovasc Med.* (2018) **5** 87. DOI: 10.3389/fcvm.2018.00087
53. Nagaoka K., Nagashima T., Asaoka N., Yamamoto H., Toda C., Kayanuma G.. **Striatal TRPV1 activation by acetaminophen ameliorates dopamine D2 receptor antagonist-induced orofacial dyskinesia**. *JCI Insight* (2021) **6** e145632. DOI: 10.1172/jci.insight.145632
54. Nagashima T., Shirakawa H., Nakagawa T., Kaneko S.. **Prevention of antipsychotic-induced hyperglycaemia by vitamin D: A data mining prediction followed by experimental exploration of the molecular mechanism**. *Sci. Rep.* (2016) **6** 26375. DOI: 10.1038/srep26375
55. Nagashima T., Hayakawa T., Akimoto H., Minagawa K., Takahashi Y., Asai S.. **Identifying antidepressants less likely to cause hyponatremia: Triangulation of retrospective cohort, disproportionality, and pharmacodynamic studies**. *Clin. Pharmacol. Ther.* (2022) **111** 1258-1267. DOI: 10.1002/cpt.2573
56. Narula P. K., Rehan H. S., Unni K. E., Gupta N.. **Topiramate for prevention of olanzapine associated weight gain and metabolic dysfunction in schizophrenia: A double-blind, placebo-controlled trial**. *Schizophr. Res.* (2010) **118** 218-223. DOI: 10.1016/j.schres.2010.02.001
57. Narvaez C. J., Matthews D., Broun E., Chan M., Welsh J.. **Lean phenotype and resistance to diet-induced obesity in vitamin D receptor knockout mice correlates with induction of uncoupling protein-1 in white adipose tissue**. *Endocrinology* (2009) **150** 651-661. DOI: 10.1210/en.2008-1118
58. Nimitphong H., Holick M. F., Fried S. K., Lee M. J.. **25-hydroxyvitamin D₃ and 1,25-dihydroxyvitamin D₃ promote the differentiation of human subcutaneous preadipocytes**. *PLoS One* (2012) **7** e52171. DOI: 10.1371/journal.pone.0052171
59. Oh J., Weng S., Felton S. K., Bhandare S., Riek A., Butler B.. **1,25(OH)2 vitamin d inhibits foam cell formation and suppresses macrophage cholesterol uptake in patients with type 2 diabetes mellitus**. *Circulation* (2009) **120** 687-698. DOI: 10.1161/CIRCULATIONAHA.109.856070
60. Ohashi K., Deyashiki A., Miyake T., Nagayasu K., Shibasaki K., Shirakawa H.. **TRPV4 is functionally expressed in oligodendrocyte precursor cells and increases their proliferation**. *Pflugers Arch.* (2018) **470** 705-716. DOI: 10.1007/s00424-018-2130-3
61. Osaki Y., Nakagawa Y., Miyahara S., Iwasaki H., Ishii A., Matsuzaka T.. **Skeletal muscle-specific HMG-CoA reductase knockout mice exhibit rhabdomyolysis: A model for statin-induced myopathy**. *Biochem. Biophys. Res. Commun.* (2015) **466** 536-540. DOI: 10.1016/j.bbrc.2015.09.065
62. Reid I. R., Ames R., Mason B., Bolland M. J., Bacon C. J., Reid H. E.. **Effects of calcium supplementation on lipids, blood pressure, and body composition in healthy older men: A randomized controlled trial**. *Am. J. Clin. Nutr.* (2010) **91** 131-139. DOI: 10.3945/ajcn.2009.28097
63. Saha S., Chant D., Welham J., McGrath J.. **A systematic review of the prevalence of schizophrenia**. *PLoS Med.* (2005) **2** e141. DOI: 10.1371/journal.pmed.0020141
64. Siswanto S., Yamamoto H., Furuta H., Kobayashi M., Nagashima T., Kayanuma G.. **Drug repurposing prediction and validation from clinical big data for the effective treatment of interstitial lung disease**. *Front. Pharmacol.* (2021) **12** 635293. DOI: 10.3389/fphar.2021.635293
65. Teske K. A., Yu O., Arnold L. A.. **Inhibitors for the Vitamin D receptor-coregulator interaction**. *Vitam. Horm.* (2016) **100** 45-82. DOI: 10.1016/bs.vh.2015.10.002
66. Wang J., Liu Y. S., Zhu W. X., Zhang F. Q., Zhou Z. H.. **Olanzapine-induced weight gain plays a key role in the potential cardiovascular risk: Evidence from heart rate variability analysis**. *Sci. Rep.* (2014) **4** 7394. DOI: 10.1038/srep07394
67. Wong K. E., Szeto F. L., Zhang W., Ye H., Kong J., Zhang Z.. **Involvement of the vitamin D receptor in energy metabolism: Regulation of uncoupling proteins**. *Am. J. Physiol. Endocrinol. Metab.* (2009) **296** E820-E828. DOI: 10.1152/ajpendo.90763.2008
68. Wong K. E., Kong J., Zhang W., Szeto F. L., Ye H., Deb D. K.. **Targeted expression of human vitamin D receptor in adipocytes decreases energy expenditure and induces obesity in mice**. *J. Biol. Chem.* (2011) **286** 33804-33810. DOI: 10.1074/jbc.M111.257568
69. Yabe D., Brown M. S., Goldstein J. L.. **Insig-2, a second endoplasmic reticulum protein that binds SCAP and blocks export of sterol regulatory element-binding proteins**. *Proc. Natl. Acad. Sci. U. S. A.* (2002) **99** 12753-12758. DOI: 10.1073/pnas.162488899
70. Yan H., Chen J. D., Zheng X. Y.. **Potential mechanisms of atypical antipsychotic-induced hypertriglyceridemia**. *Psychopharmacol. Berl.* (2013) **229** 1-7. DOI: 10.1007/s00213-013-3193-7
71. Zhu W., Ding C., Huang P., Ran J., Lian P., Tang Y.. **Metformin Ameliorates Hepatic Steatosis induced by olanzapine through inhibiting LXRα/PCSK9 pathway**. *Sci. Rep.* (2022) **12** 5639. DOI: 10.1038/s41598-022-09610-1
|
---
title: Life expectancy tables for dogs and cats derived from clinical data
authors:
- Mathieu Montoya
- Jo Ann Morrison
- Florent Arrignon
- Nate Spofford
- Hélène Charles
- Marie-Anne Hours
- Vincent Biourge
journal: Frontiers in Veterinary Science
year: 2023
pmcid: PMC9989186
doi: 10.3389/fvets.2023.1082102
license: CC BY 4.0
---
# Life expectancy tables for dogs and cats derived from clinical data
## Abstract
There are few recent and methodologically robust life expectancy (LE) tables for dogs or cats. This study aimed to generate LE tables for these species with clinical records from >1,000 Banfield Pet hospitals in the USA. Using Sullivan's method, LE tables were generated across survey years 2013–2019, by survey year, and for subpopulations defined by sex, adult body size group (purebred dogs only: toy, small, medium, large and giant), and median body condition score (BCS) over life. The deceased population for each survey year comprised animals with a recorded date of death in that year; survivors had no death date in that year and were confirmed living by a veterinary visit in a subsequent year. The dataset totaled 13,292,929 unique dogs and 2,390,078 unique cats. LE at birth (LEbirth) was 12.69 years ($95\%$ CI: 12.68–12.70) for all dogs, 12.71 years (12.67–12.76) for mixed-breed dogs, 11.18 years (11.16–11.20) for cats, and 11.12 (11.09–11.14) for mixed-breed cats. LEbirth increased with decreasing dog size group and increasing survey year 2013 to 2018 for all dog size groups and cats. Female dogs and cats had significantly higher LEbirth than males: 12.76 years (12.75–12.77) vs. 12.63 years (12.62–12.64), and 11.68 years (11.65–11.71) vs. 10.72 years (10.68–10.75), respectively. Obese dogs (BCS $\frac{5}{5}$) had a significantly lower LEbirth [11.71 years (11.66–11.77)] than overweight dogs (BCS $\frac{4}{5}$) [13.14 years (13.12–13.16)] and dogs with ideal BCS $\frac{3}{5}$ [13.18 years (13.16–13.19)]. The LEbirth of cats with BCS $\frac{4}{5}$ [13.67 years (13.62–13.71)] was significantly higher than cats with BCS $\frac{5}{5}$ [12.56 years (12.45–12.66)] or BCS $\frac{3}{5}$ [12.18 years (12.14–12.21)]. These LE tables provide valuable information for veterinarians and pet owners and a foundation for research hypotheses, as well as being a stepping-stone to disease-associated LE tables.
## 1. Introduction
Human health metrics and life expectancy computations have been used for centuries. The first quantification of life expectancy was reported in England in the 19th century [1]. Life expectancy tables have diverse uses for humans, which include monitoring the overall health status of populations over time, exploring health inequalities in socio-economic groups, and charting the progress of less developed nations [2, 3]. They are valuable for understanding the acute and long-term, direct and indirect effects of health conditions and medical interventions on survival. Life expectancy data can be used to inform planning for healthcare provisions, economic policies, life insurance premiums and pension annuities.
Accurately calculating the life expectancy of dogs and cats would be of greatest interest in helping to understand the benefits of preventative care and the success of increasingly advanced medical interventions. However, there are relatively few studies on lifespan or all-cause mortality in either of these species, especially in cats. Four studies in dogs have compiled age of death statistics from surveys of owners belonging to the UK Kennel club [4, 5], the Danish Kennel club [6] and owners on a UK insurance database or attending the Crufts dog show [7]. Analyses of the primary-care veterinary records of both dogs [8] and cats [9] in England have provided age of death data as the percentage of pets dying within pre-specified age brackets. Kaplan-Meier survival curves for dogs attending the Banfield Pet Hospital network of primary care hospitals in the USA have been published [10], and a study of life-insured Swedish cats generated survival curves for cats using a methodology of age standardized cox regression [11]. These studies have provided valuable data, but they do not provide the same information as life expectancy tables, which are a standard way of gauging the health status of large human populations.
Survival data are analyzed in two ways: the life-table method divides time into intervals and calculates survival in each interval; the Kaplan-Meier method calculates survival each time an event occurs (death). Both methods produce a graph (survival curve) that shows the cumulative probability (hazard ratio) of the event in the total period of observation. Kaplan-Meier survival analysis may be more appropriate for studies with longitudinal follow-up focused on the time until events occur. In contrast, the life-table methodology is a statistical technique that provides a summary of the mortality and life expectancy experiences of a large population of different ages. This study used life table methodology to analyze survival, in part to provide information equivalent to that for large human populations, and in part because of the high dropout rates of animals (right-censored individuals relative to animals with a confirmed date of death) due to discontinuity in veterinary visits in the clinical dataset.
Life expectancy tables for pets have been reported previously from a total of four studies in Japan and one in the UK (12–16). The oldest of these studies generated a life expectancy table using the cemetery records of 4,915 dogs in the Kyoto region of Japan that died between 1981 and 1982 [12]. Similarly there was a census of 12,039 dogs from cemeteries in Tokyo between 2012 and 2015 [14]. The only life expectancy tables found for cats were also from Japanese cemetery data in Tokyo (3,936 cats identified from between 1981 and 1982) [16]. Using such data sources comprising only dead animals for a life-table methodology is unsatisfactory because of a “hidden” bias of right censorship, since they cannot account for animals in a cohort still alive at the time of sampling [17]. This limitation did not apply to the fourth Japanese study, which generated life expectancy tables from the insurance records of 299,555 dogs insured between 2010 and 2011 and followed for 1 year [13]. The first life tables for dogs in the UK were based on the VetCompass™ database of primary-care veterinary clinics, in which there were 30,563 dogs with at least one clinical record in 2016 and a confirmed death between January 2016 and July 2020 [15]. Despite differences between studies in data sources and the computed life expectancy, they all found substantial differences in life expectancies between different dog sizes or breeds when this was investigated.
Other studies in dog populations comprising multiple breeds and both sexes have summarized overall ages of death without computing life expectancy tables (4–8). Age of death was 10 years (median) in a Danish survey of 2,928 purebred and mixed-breed dogs in 1997 [6], 10.33 years (median) in a survey of 5,663 Kennel Club-registered dogs in the UK in 2014 [4], 11.08 years (mean) in 3,000 British purebred and mixed-breed dogs [7], 11.25 years (median) in 15,881 purebred dogs in the UK over the 10 year prior to 2003 [5], and 12 years (median) in a UK VetCompass™ dataset of 5,095 dogs between 2009 and 2011 [8]. Age-at-death data are scarce for cat populations comprising multiple breeds and/or mixed-breed cats. A study that used a random sample of 4,009 purebred and mixed-breed cats with confirmed deaths in the UK VetCompass™ database between 2009 and 2012 reported a median age of death of 14.0 years [interquartile range (IQR): 9.1–17.0] [9].
The more recent studies demonstrate the opportunities available in the current era of veterinary “big data”, which is giving researchers access to reliable and extensive high quality data over long periods of time [18]. There is a need now to leverage the power of other large datasets to characterize life expectancy in countries with different disease risks, attitudes toward veterinary care, and accessibility of such care. Furthermore, life expectancy tables are needed for pets with and without specific known or hypothesized risk factors for disease, to help quantify the potential benefits of improving modifiable risk factors, and to support conversations with owners on proactive lifestyle choices for the health of their pets. Obesity in pets is a condition of particular interest. It is the most common nutritional disorder in companion animals [19], and obese dogs have a reduced lifespan compared with dogs of normal body condition [20], while severe obesity in cats is associated with reduced survival and reduced lifespan [21]. Obesity in pets is associated with a wide range of functional impairments and health conditions (22–30) that could potentially have a direct or indirect impact on life expectancy. Examples include associations with diabetes mellitus [25, 26], metabolic derangements similar to some of those seen in metabolic syndrome in humans [31], and changes in the structure and function of the heart [29] and kidneys [24]. The cause and effect of associations between obesity, health and reduced life expectancy or lifespan is a multifaceted and complex field of active research in both humans and pets. Understanding the impact of obesity as either a risk factor for the development of specific diseases or as a consequence of them is important for preventative medicine, disease signalment and disease management.
The current study aimed to generate life expectancy tables for dogs and cats in the USA, computed for each survey year between 2013 and 2019, and across survey years according to breed size, sex, and body condition score (BCS). Clinical data were obtained from the electronic medical records (EMRs) of healthy and sick dogs and cats attending Banfield hospitals throughout the USA. Computations followed Sullivan's method for life expectancy tables, which can also account for health status assessed as a cross-sectional variable [32]. This methodology is one of the most popular methods for life expectancy tables and is used by the World Health Organization.
## 2.1. Study populations
The clinical dataset was derived from the EMRs of 1,152 different Banfield hospitals in the USA between January 1st, 2013, and July 31st, 2022. The number of hospitals varied between years due to the closure of some hospitals and the opening of new ones, however most hospitals contributed data for each of the 9 years of the study (n2013 = 854; n2014 = 893; n2015 = 933; n2016 = 984; n2017 = 1,015; n2018 = 1,044; n2019 = 1,074; n2020 = 1,084; n2021 = 1,069; n2022 = 1,054). All hospitals were linked by a proprietary practice management system (PetWare®) and data from all visits were uploaded to a central database, resulting in a large amount of aggregated, structured data.
The dataset was cleaned by removing animals with no recorded birth date, anomalous birth or death dates (date of birth posterior to visit, or date of death anterior to visit), a visit age >30 years, or incomplete data for sex, age or BCS. The resulting dataset is referred to as the cleaned dataset.
A deceased and survivor population for each survey year was identified from the clinical dataset as outlined in Figure 1; the sum of these two populations comprised the total study population for that year. Dogs and cats with no death date in the EMRs were only considered to be survivors for a specific year if there was a new hospital visit in any of the following study years. Dogs and cats with no death date and that only visited the clinic in the course of 1 year, with no return visits in a subsequent year of the study, were excluded because they could not be confirmed as either a survivor in any year or a deceased animal.
**Figure 1:** *Methodology used to build the study population in each calendar survey year from the clinical dataset.*
The return rate of animal visits across all study years was calculated to support the validity of the range of years selected for the analysis. On average, dogs returned in 111 days and cats in 146 days; median return rate was 70 days (IQR: 21–169) for dogs and 90 days (21–190) for cats. The Tukey's outlier detection method applied to the return rate in days yielded an upper fence (UF = Q3 + 1.5*IQR) of 391 days for dogs and 443 days for cats. At least two full years of data after a survey year were needed for life table calculation. It was decided to analyze life expectancy in the dataset from January 1st, 2013 to December 31st, 2019. The last 2.5 years of data (January 1st, 2020 to July 31st, 2022) were used to define if a pet without a death date could be confirmed as a survivor in any of the previous years.
Subpopulations of each species were defined by sex (male or female), and median BCS was recorded for individuals between January 1st, 2013 and July 31st, 2022. The sterilization status was available in the clinical dataset, but it was not included in the study because of the way these data were collected. Pets that were sterilized at a certain age were identified as neutered from birth to death. This introduced a bias due to mortality in the first few months of life when pets were not yet spayed or neutered, which impacted the early mortality rate for entire animals. Body condition score was assessed on a 5-point scale between 2013 and 2018, and on a 9-point scale between 2018 and 2022. Data from the 9-point scale were converted to the 5-point scale (Table 1). Separate analyses were conducted for purebred and mixed-breed animals. Mixed-breed dogs were recorded as such in the database, whereas cats were identified as mixed breed if they were recorded as Domestic Short Hair, Domestic Long Hair, or Domestic Medium Hair.
**Table 1**
| BCS 9-point scale | BCS 9-point scale.1 | BCS 5-point scale | BCS 5-point scale.1 |
| --- | --- | --- | --- |
| Score | Description | Score | Description |
| 1 | Emaciated | 1 | Very thin |
| 2 | Very thin | 2 | Underweight |
| 3 | Thin | 2 | Underweight |
| 4 | Ideal weight | 3 | Ideal weight |
| 5 | Ideal weight | 3 | Ideal weight |
| 6 | Overweight | 4 | Overweight |
| 7 | Heavy | 4 | Overweight |
| 8 | Obese | 5 | Obese |
| 9 | Severely obese | 5 | Obese |
In addition, the population of purebred dogs was subdivided into five size categories based on adult bodyweight: toy, small, medium, large and giant. Firstly, dogs weighing more than 70 and <0.5 kg were removed from the uncleaned dataset of purebred dogs. Bodyweight outliers for each breed were then excluded using the Tukey's outlier detection method (lower fence = Q1 − 1.5*IQR; upper fence = Q3 + 1.5*IQR). The average adult bodyweight of each breed in the resulting dataset was calculated from dogs aged 2–5 years, and was used to assign a size group (Table 2) to each breed in the cleaned dataset.
**Table 2**
| Size group | Adult bodyweight* | Number of breeds in size group |
| --- | --- | --- |
| Toy | < 5.5 kg | 19 |
| Small | ≥5.5 to < 11 kg | 94 |
| Medium | ≥11 to < 26 kg | 215 |
| Large | ≥26 to < 45 kg | 154 |
| Giant | ≥45 kg | 26 |
| Mixed breed | All adult bodyweights | 1 (mixed breed) |
## 2.2. Construction of life tables
Life expectancy tables for dogs and cats were generated using the probabilistic approach of Sullivan's method [32], based on the deceased and survivor populations of dogs and cats described in Section 2.1. This methodology is applied in studies on human health assessments (33–38). Uncertainties were assessed with Monte Carlo methods [33, 34, 37, 39, 40].
The variables used in life table calculations were the number of dead pets d(x) in a 1-year age interval (x, x+1), and the study population size (dead plus living pets) at the mid-year point P(x). Central death rate was computed as m(x)= d(x)P(x). Pets who died in the interval (x, x+1) had lived x complete years plus a fraction a(x) of the last year. It was assumed that deaths were uniformly distributed in each year interval, occurring on average midway in the age range (x, x+1); therefore a(x) was uniformly taken to be 0.5 [41].
The conditional probability of death in a 1-year age interval (x, x+1) was calculated as q(x)= m(x)(1+a(x)*m(x)). The number of pets surviving to age x, denoted as l(x), was estimated from a hypothetical population of 100,000 pets at birth i.e., l[0] = 100, 000. As these pets were assumed to be subject throughout their lives to the conditional probability of death in the 1-year age interval (x, x+1), the calculation was l(x) = (1−q(x−1))*l(x−1). The number of pet years lived at age x (any age other than the first year of life and the final age interval), denoted as L(x), was the sum of pets surviving until age (x+1) and the fraction a(x) of pets surviving in age interval (x, x+1), so that L(x) = l(x+1)+a(x)*(l(x)−l(x+1)). Given that a(x) = 0.5, the equation was simplified to L(x)=(l(x)+l(x+1))2. For the first year of life, the assumption was made that $80\%$ of the deaths happened in the first months of life, leading to L[0] = 0.2*l[0]+0.8*l[1]. For the last age interval ω, as no information was available after this age, it was calculated as L(ω)=l(ω)m(ω). The total number of years lived beyond age x was T(x)= ∑i=xωL(x). The final equation for life expectancy at year x was e(x)=T(x)l(x). Life expectancy at birth (LEbirth) corresponded to life expectancy for animals in the 0–1 age interval. Uncertainties were calculated with Monte-Carlo methods, with $$n = 1$.106$ iterations, drawing d(x) from a parametrized binomial distribution where each sample was equal to the number of successes over n trials, the probability of success being d(x)P(x) and the number of successes P(x). Life expectancy e(x) was calculated n times, using the Monte-Carlo-derived d(x) in the previous equations. Confidence intervals (CIs) at the $95\%$ level were calculated for each year using the mean e(x) and standard deviation [denoted as SLE(x)] for the n iterations: CI$95\%$(x)=e(x)± 1.96*SLE(x).
Life tables were generated for each species across all survey years, for each survey year, for the global population and the subpopulations defined in Section 2.1. For each subpopulation, the variables P(x) and d(x) specific to that subpopulation were used to recalculate the intermediate variables, m(x), q(x), l(x), L(x) and T(x). Only data for purebred dogs were used for life expectancy estimations by dog size group.
## 2.3. Software applications
Only structured data were extracted from the PetWare® database; there was no searching of free text entries. Software applications used for data processing and building life tables were Databricks (runtime 11.0, with Apache Spark 3.3.0) on Microsoft Azure, with Python 3.9.5 (libraries Numpy 1.20.3, Pandas 1.3.4, Lifelines 0.27.2 and Plotly 5.6.0).
## 3.1. Study population
There were 64.76 million PetWare® visit records for 9.29 million unique dogs, and 9.85 million visits for 2.51 million unique cats. Of these animals, $1.74\%$ dogs and $1.82\%$ cats were removed to produce the cleaned dataset as described in the methods. The total population sizes (deceased and survivor combined) in the cleaned dataset are presented by survey year for each species in Figure 2A and Supplementary Table 1. Pets had a mean 2.5 visits during a survey year, and over the 9 years studied had a mean 6.3 visits and a median of three visits (IQR: 1–8). The percentage of dogs returning within 2 years after the survey year increased from $74.8\%$ in 2013 to $77.6\%$ in 2017, and subsequently decreased to $75.9\%$ in 2019. For cats, the percentage of returns was $61.9\%$ in 2013, it increased each year to a maximum of $65.4\%$ in 2017, then it dropped to $62.5\%$ in 2019 (Supplementary Table 2).
**Figure 2:** *Total population sizes (deceased and survivor) for life expectancy calculations. Dogs were categorized as purebred dogs in size groups toy, small, medium, large and giant, and as mixed-breed dogs. Cats were categorized as purebred and mixed-breed cats. Population sizes are presented by survey year (A), sex (B), age interval (C), and median body condition score (D). Body condition scoring was on a 5-point scale from 1 = very thin to 5 = obese. BCS, body condition score.*
In summary, the mean total dog population size (deceased and survivor populations) was 1,899,048 individual dogs per year (26,099 giant, 592,413 large, 251,831 medium, 545,988 small, 409,204 toy, and 73,513 mixed breed). The mean deceased dog population was 85,651 per year, and a mean of 621,284 ($25\%$) dogs were excluded each year on account of having no date of death and no additional visits in subsequent years (Figure 1). The mean age at death for dogs was 10.66 years, the median age at death was 11.57 years (IQR: 8.4–13.9), and the median survival time was 14.13 years.
The mean total population size for cats was 314,823 individuals per year (54,478 purebred and 287,345 mixed breed). The mean deceased cat population was 28,308 cats per year, and 172,523 ($34\%$) cats per year were excluded according to the study methodology (Figure 1). The mean age at death was 11.15 years, the median age at death was 12.30 years (IQR: 7.0–15.7), and the median survival time was 15.7 years.
The overall proportion of male dogs was $52.0\%$; for purebred dogs it was 51.9, 53.7, 51.1, 51.2, and $55.9\%$ for toy, small, medium, large and giant sizes, respectively, and for mixed-breed dogs it was $48.3\%$ (Figure 2B; Supplementary Table 3). The proportion of males in the total cat population was $49.8\%$ ($45.9\%$ for purebred and $50.5\%$ for mixed-breed cats). In total, $91.3\%$ of female and $86.6\%$ of male dogs were sterilized. The percentage of sterilized cats was $96.1\%$ for females and $96.9\%$ for males.
Data were sparse for dogs and cats older than 17 years, and uncertainty in life expectancy calculations became unacceptably large. Therefore, dogs and cats living more than 17 years were grouped into an age interval of 17+ years. Figure 2C and Supplementary Table 4 show the distribution of population headcount by pet group and age interval. The cat population was smaller in each age interval compared with the dog population except for the age interval 17+ years, in which the cat population was more than two times larger than the dog population.
The median BCS was 3 for $72.7\%$ of dogs (toy $77.0\%$, small $72.4\%$, medium $67.6\%$, large $71.9\%$, giant $79.4\%$, and mixed breed $72.3\%$) (Figure 2D; Supplementary Table 5). The prevalence of obesity (median BCS 5) was $1.5\%$ (toy $1.2\%$, small $1.5\%$, medium $2.1\%$, large $1.6\%$, giant $1.0\%$, and mixed breed $1.4\%$). In the cat population, the median BCS was 3 for $58.5\%$ of animals (purebred $62.4\%$, $57.7\%$ mixed breed) (Figure 2D; Supplementary Table 5) and the prevalence of obesity (median BCS 5) was $3.6\%$ ($2.9\%$ purebred, $3.7\%$ mixed breed). The percentage of dogs with median BCS 3 increased each year from $68.7\%$ in 2013 to $74.5\%$ in 2019, while for median BCS 5 it decreased from $2.1\%$ in 2013 to $1.3\%$ in 2017, subsequently remaining constant until 2019. The percentage of cats with median BCS 3 increased from 53.2 to $60.3\%$ between 2013 and 2019, whereas for median BCS 5 it decreased from $4.3\%$ in 2013 to $3.5\%$ in 2015, and then fluctuated between 3.5 and $3.3\%$ until 2019 (Supplementary Table 6).
In summary, the study dataset for dogs covered all bodyweight size categories; the large and small dog sizes had the greatest representation. Less than $5\%$ of dogs were of mixed breed, whereas in the cat dataset there were ~5 times more mixed-breed cats than purebred. For both cats and dogs the sex ratio was well-balanced between males and females, and the vast majority of both sexes in both species were sterilized. Median body condition was ideal for nearly three quarters of dogs and obese for $1.5\%$. In contrast, for cats, median body condition was ideal for slightly more than two thirds and obese for $3.6\%$.
## 3.2. Life expectancy tables
The global life expectancy table with intermediate variables is presented in Table 3 for dogs and Table 4 for cats. The life expectancy of dogs decreased by ~0.9 years each age interval between 1–2 and 7–8 years. Subsequently the decrease in life expectancy diminished from interval to interval, and from 12 to 13 years onwards, the decrease was <0.5 years per age interval. For cats, life expectancy decreased by approximately 0.7 years each age interval between 1–2 and 8–9 years. After this, the reduction in life expectancy decreased, and was 0.5 years or less after the 10–11 years age interval.
## 3.3. Life expectancies by survey year
The LEbirth of dogs across all survey years and dog sizes was 12.69 years ($95\%$ CI: 12.68–12.70) (Table 2). Life expectancy at birth varied by dog size, being 9.51 years (9.45–9.58) for giant, 11.51 years (11.49–11.52) for large, 12.7 years (12.68–12.72) for medium, 13.53 years (13.52–13.55) for small, and 13.36 years (13.33–13.38) for toy (Supplementary Table 7). Mixed-breed dogs had an LEbirth of 12.71 years (12.67–12.76) (Supplementary Table 7). The LEbirth of cats across all survey years was 11.18 years (11.16–11.20) (Table 2). Life expectancy at birth varied by cat group, being 11.54 years (11.48–11.6) for purebred and 11.12 years (11.09–11.14) for mixed-breed cats (Supplementary Table 7).
The LEbirth increased progressively for survey years 2013–2018 for all dog size groups and for cats (Figure 3; Supplementary Table 7). A small decrease in life expectancy was observed in 2019 for toy and giant dog size groups and also for cats; the decreases were not significant for either species. Changes in cat LEbirth over successive survey years were greater than those for all size groups of dogs in all time periods, with the exception of the change between 2013 and 2014 for giant dogs. Across the entire study period, LEbirth increased by 0.72 years ($5.62\%$), 0.66 years ($5.0\%$), 0.56 years ($4.51\%$), 0.48 years ($4.27\%$), 0.6 years ($6.59\%$), and 0.83 years ($6.81\%$) for toy, small, medium, large, giant and mixed-breed dogs, respectively, and by 1.01 years ($9.31\%$) for purebred and 1.41 years ($13.69\%$) for mixed-breed cats.
**Figure 3:** *Life expectancies at birth of purebred and mixed-breed dogs and cats by survey year. Dogs were categorized as purebred dogs in size groups toy, small, medium, large and giant, and as mixed-breed dogs. Cats were categorized as purebred and mixed-breed cats. Error bars represent 95% confidence intervals.*
Life expectancy for each age interval by survey year is presented in Figure 4 and Supplementary Table 8. For dogs, life expectancy was significantly higher in 2019 than in 2013 from the age interval 0–1 year to 15–16 years for toy and medium size groups, from 0–1 year to 16–17 years for small dogs, from 0–1 year to 13–14 years for large dogs, from 0–1 year to 8–9 years for giant dogs, and from 0–1 year to 11–12 years for mixed-breed dogs. Life expectancy was significantly higher in 2019 than in 2013 for mixed-breed cats at all age intervals, and for purebred cats from 0–1 year until 16–17 years.
**Figure 4:** *Life expectancies of purebred and mixed-breed dogs and cats by age interval and survey year. Life expectancies are shown for purebred dog size groups toy (A), small (B), medium (C), large (D) and giant (E), for mixed-breed dogs (F), for purebred cats (G), and for mixed-breed cats (H). Error bars represent 95% confidence intervals.*
## 3.4. Life expectancies by sex
Female dogs had a slightly but significantly higher LEbirth than male dogs: 12.76 years (12.75–12.77) compared with 12.63 years (12.62–12.64) for males (Figure 5A; Supplementary Table 9). Life expectancy was significantly higher for female dogs over age intervals 0–1 year until 6–7 years compared with male dogs. From 11–12 years to 15–16 years male dogs had a significantly higher life expectancy than females. Sex differences in life expectancy varied according to dog size (Figure 6; Supplementary Table 10). Male toy dogs had a slightly but significantly higher LEbirth than female toy dogs [13.39 years (13.36–13.42) vs. 13.32 years (13.28–13.35)]. There were no significant sex differences in the life expectancy of small dogs in any age interval; LEbirth was 13.52 years for males (13.5–13.55) and 13.55 years (13.52–13.57) for females. In all other dog-size groups, females' LEbirth was significantly higher than that of males: medium 12.8 years (12.77–12.83) vs. 12.6 years (12.57–12.63), large 11.74 years (11.72–11.76) vs. 11.28 years (11.26–11.3), giant 9.76 years (9.66–9.85) vs. 9.33 years (9.24–9.41). Female mixed-breed dogs also had a significantly higher LEbirth than their male counterparts-−12.81 years (12.27–12.87) compared with 12.61 years (12.54–12.67).
**Figure 5:** *Life expectancies of dogs (A) and cats (B) by age interval and sex. Animal populations were all dogs and all cats regardless of size and breed. Error bars represent 95% confidence intervals.* **Figure 6:** *Life expectancies of purebred and mixed-breed dogs and cats by age interval and sex. Life expectancies are shown for purebred dog size groups toy (A), small (B), medium (C), large (D) and giant (E), for mixed-breed dogs (F), for purebred cats (G), and for mixed-breed cats (H). Error bars represent 95% confidence intervals.*
Female cats had a significantly higher LEbirth than male cats, by ~1 year [11.68 years (11.65–11.71) vs. 10.72 years (10.68–10.75)] (Figure 5B; Supplementary Table 9). Differences in life expectancies between cat sexes progressively decreased until the age of 16–17 years, at which point they were similar, with overlapping confidence intervals. The LEbirth of purebred and mixed-breed cats were ~1 year higher than those of their male counterparts: purebred 11.98 years (11.91–12.06) vs. 11.05 years (10.97–11.14), mixed breed 11.62 (11.58–11.65) vs. 10.66 (10.63–10.7) (Figure 6; Supplementary Table 10).
## 3.5. Life expectancies by body condition score
Dogs with median BCS 3 and dogs with median BCS 4 had similar LEbirth of 13.18 years (13.16–13.19) and 13.14 years (13.12–13.16), respectively (Figure 7A; Supplementary Table 11). Life expectancy was significantly higher for dogs with median BCS 3 vs. median BCS 4 over the age interval 1–2 year until 17+ years. Dogs with median BCS 5 had a significantly lower LEbirth [11.71 years (11.66–11.77)] than dogs with median BCS 3 or median BCS 4. This continued to be the case for life expectancies in subsequent age intervals until 15–16 years, when the difference between median BCS 5 and median BCS 4 was no longer significant, while the difference between median BCS 5 and median BCS 3 remained significant until the age interval 17+ years. Dogs with median BCS 1 or median BCS 2 had significantly lower life expectancies compared with dogs of all other median BCS at all age intervals; LEbirth was 1.54 years (1.49–1.60) and 3.91 years (3.86–3.97), respectively. The difference in life expectancy between dogs with median BCS 1 and those with median BCS 2 diminished progressively after the age of 2 years, but remained significant until the age of ≥17 years.
**Figure 7:** *Life expectancies of dogs (A) and cats (B) by age interval and median life-time body condition score. Animal populations were all dogs and all cats regardless of size, breed and sex. Error bars represent 95% confidence intervals. Body condition scoring was on a 5-point scale from 1 = very thin to 5 = obese. BCS, body condition score.*
Cats with median BCS 3 [12.18 years (12.14–12.21)] had a significantly lower LEbirth than cats with median BCS 4 [13.67 years (13.62–13.71)] or median BCS 5 [12.56 years (12.45–12.66)] (Figure 7B; Supplementary Table 12). Life expectancy continued to be lower in cats with median BCS 3 compared with median BCS 4 until the age interval of 9–10, when it was 6.16 years (6.13–6.19) vs. 6.15 years (6.11–6.19). After the age interval 10–11 and until 17+ years, median BCS 3 cats had a greater life expectancy than median BCS 4 cats. Life expectancies were significantly higher for cats of median BCS 5 compared with median BCS 3 only for age interval 0–1, after which they were significantly lower. The average life expectancy over all age intervals was ~1.39 years for cats with median BCS 1, and 2.49 years for cats with median BCS 2.
The LEbirth of dogs by year of survey for median BCS 3 was 13.73 years (13.68–13.78) in 2013. This was significantly higher than for subsequent years, during which LEbirth increased slightly from 13.04 years (13.0–13.08) in 2014 to 13.15 years (13.12–13.18) in 2019, with few significant differences between years in this range (Figure 8A; Supplementary Table 13). The LEbirth of dogs with median BCS 5 also decreased between 2013 [11.47 years (11.3–11.63)] and 2014 [11.08 years (10.94–11.22)], but then increased by almost 1.5 to 12.57 years (12.42–12.72) in 2019. The mean LEbirth over all survey years for dogs with median BCS 1 was 1.51 years with no significant differences between years. For dogs with median BCS 3 or 4, the LEbirth was always significantly higher than that of dogs with BCS 5 regardless of the survey year.
**Figure 8:** *Life expectancies at birth of dogs (A) and cats (B) by year of survey and median life-time body condition score. Animal populations were all dogs and all cats regardless of size, breed and sex. Error bars represent 95% confidence intervals. Body condition scoring was on a 5-point scale from 1 = very thin to 5 = obese. BCS, body condition score.*
The trends in LEbirth for cats were similar to those of dogs. Cats had a mean LEbirth over all survey years of 1.04 years with no significant differences between years in the survey (Figure 8B; Supplementary Table 13). The LEbirth for cats with median BCS 3 was 12.88 years (12.76–13.01) in 2013, dropping to 11.87 years (11.77–1.97) in 2014 before starting to increase slowly to 12.47 years (12.38–12.56) in 2018, falling back however to 12.19 years (12.1–12.27) in 2019. For cats with median BCS 5, there was an initial a drop in LEbirth from 12.12 years (11.85–12.39) in 2013 to 11.58 years (11.31–11.85) in 2014, followed by a 1.83 years increase in life expectancy between 2014 and 2019 [13.41 years (13.1–13.72)]. The LEbirth of cats with median BCS 4 was always significantly higher than that of cats with BCS 3 or 5 regardless of the survey year.
## 4. Discussion
This study generated annual life expectancy tables over a period of 7 years for dogs and cats aged from <1 to >17 years in the USA. Life expectancy tables were computed by survey year and by age in 1-year increments for each sex and each of five median BCS values.
## 4.1. Life expectancies overall and by size group or breed heritage
The overall LEbirth for dogs reported here [12.69 years (12.68–12.7)] was higher than that computed from the UK VetCompass™ dataset [11.23 years (11.19–11.27)] [15] and lower than that for dogs in two Japanese studies (13.7 years) [13, 14]. The LEbirth in another Japanese study of dogs that died between 1981 and 1982 was only 8.3 years [12]. This is more consistent with ours when consideration is given to the increase in life expectancy of dogs over time that we observed. The LEbirth by survey year increased by 1.3 months ($0.93\%$) in each successive sampling year (Figure 2). A backwards projection estimates LEbirth to be 9.1 years for dogs in 1981. For comparative purposes with other studies, we also calculated the mean and median age of death of dogs; these were 10.55 and 11.47 years, respectively, which was within the range previously reported for dog populations comprising multiple breeds and both sexes (4–8).
Differences in overall life expectancy between studies might reflect in part different proportions of size categories, breeds and mixed-breed dogs. Japanese life tables by dog size showed the same order of decreasing LEbirth by size as observed in this study: small > toy > medium > large > giant [13]. In an indirect comparison between the Japanese life expectancies and ours for size groups with the most comparable definitions, the LEbirth for small dogs was 14.2 years (14.0–14.4) compared with 13.53 years (13.52–13.55), respectively, and for toy dogs was 13.34 years (13.32–13.36) in japanese study and 13.36 years (13.33–13.38) in our study. The inverse relationship between dog size and longevity is well-described [42], however, even within a size category, the breed composition may affect the life expectancy computation. For example, in the UK, the two highest life expectancies by breed were for the Jack Russell Terrier (small size) and the Yorkshire terrier (toy size), but the Border Collie (medium size) ranked third and the Chihuahua (toy) ranked below this in 15th place [15].
Multiple other differences between studies might also explain differences in life expectancies: population size and source, case ascertainment, husbandry and veterinary practices of the countries concerned and the dates the data relate to. The life expectancy of insured dogs might be enhanced compared with uninsured dogs by both a greater willingness of owners to seek veterinary care and by representation of breeds predisposed to health conditions that incur higher insurance premiums. Frequency of veterinary visits could indicate the presence of a chronic health condition, but might also reflect participation in a wellness program of routine preventative veterinary care, including vaccination. The study country is likely to influence these and other factors. Dogs in the USA will have genetically different breeding lines from the UK and Japan. Broad cultural differences in human populations can affect the lifestyle, environment, and nutrition of their pets, as well as owners' access to and attitude toward veterinary interventions. In terms of methodological variations in life expectancy studies, the reliability of information from veterinary medical records such as used in this study may be different to that from owners, as used in the most recent Japanese cemetery survey [14]. However, veterinary records will of course exclude all pets never visiting a veterinary clinic, which is likely to include many puppies dying between birth and weaning [e.g., $9\%$ of purebred puppies in Norway [43]].
In the only life expectancy tables we found for cats the LEbirth was only 4.2 years, which is approximately 7 years less than in the current study [16]. However, the data were from 1981 and 1982. In our study, life expectancy increased on average by 2.6 months ($2.01\%$) each successive sampling year from 2013 to 2019 (Figure 2), and a backwards projection estimates LEbirth to be 5.42 years for cats in 1981.
Age-at-death data are also scarce for cat populations comprising multiple breeds and or mixed-breed cats. The median age-at-death of cats between 2009 and 2012 in a UK study was 14.0 years (IQR: 9.1–17.0) [9]. This is substantially higher than for our population, which we calculated for comparative purposes to be 12.18 years (IQR: 6.7–15.6). It is difficult to attribute this difference to a specific factor or combination of factors, although the country context is likely to have played a role. Many of the inter-study differences in populations, methodology, pet husbandry and veterinary care that were discussed for dogs might also have contributed to differences in cat data. Another factor to consider is that in the USA, although the owned cat population exceeds the owned dog population, the mean number of veterinary visits per household per year was estimated to be 2.4 for dogs and 1.3 for cats in 2016, and the percentage of yearly routine or preventive care visits was $78.8\%$ for dogs compared with $47.2\%$ for cats [44]. Our study population contained proportionately fewer cats than dogs between the ages of 3 and 11 years, and proportionally more from the age of 12 years, when chronic conditions and conditions of old age will be most apparent. The threshold of owner-perceived severity of illness that will prompt a veterinary visit is higher for cats than dogs. Taking these factors together may indicate that our study underestimated life expectancy of cats.
The concept of hybrid vigor has led to a perception that mixed-breed pets live longer than purebred pets, and indeed, this has been demonstrated for size-matched dogs [45]. Our study was not designed to test this hypothesis, and life expectancies for mixed-breed dogs of any size were similar to those of medium-sized purebred dogs. In contrast, purebred cats had consistently longer life expectancies than mixed-breed cats. This might be explained if a greater percentage of purebred cats were kept indoors, where they would be at a lower risk of fights and accidents [46].
Overall, for both dogs and cats, LEbirth progressively increased except for a small decrease in 2019, which was significant for cats but not for dogs. This might be explained by the impact of the COVID-19 pandemic in 2020 and 2021. During this time in the USA, veterinary practices were subject to emergency regulations, and practice visits were curtailed in many areas to “essential” services only, while telemedicine was actively encouraged [3, 47]. Pet owners were therefore likely to have made veterinary visits only for seriously ill or injured animals, which would tend to have decreased the survivor population but not affected the deceased population. Visits in 2020 and 2021, from which the 2019 survival population was estimated, probably impacted the return rate of visits within 2 years (Supplementary Table 2). Indeed, the return rate was increasing between 2013 ($74.8\%$ for dogs, $61.9\%$ for cats) and 2017 ($77.6\%$ for dogs, $65.4\%$ for cats), but was meaningfully lower in 2019 ($75.9\%$ for dogs and $62.5\%$ for cats) compared to 2018 ($77\%$ for dogs and $65.1\%$ for cats). Such a reduction would have artificially increased the number of excluded animals. Extending the follow-up to 2022 may have allowed a more accurate estimation of survivors in 2019.
## 4.2. Life expectancies by sex
Although female dogs had a significantly higher LEbirth than male dogs, the difference was small (~1.5 months) and did not persist in older age groups. A 4-month difference was shown between the sexes in a UK dog population [LEbirth 11.41 years (11.35–11.47) for females vs. 11.07 years (11.01–11.13) for males], and this also declined with increasing age.
The finding that LEbirth for female cats was 1 year higher compared with that for male cats reflected a higher death rate for males between the ages of 1 and 9 years. It was hypothesized that these sex differences were due to the higher probability of infection with feline leukemia virus and feline immunodeficiency virus [48, 49], and also a higher probability of lower urinary tract obstruction, which occurs almost exclusively in males [50]. Accidents and fights are also more common for males between 1 and 2 years of age than for females. Neoplasia and other diseases associated with chronic retroviral infection, and complications of urethral obstruction can be fatal [49, 50]. A preliminary investigation of the causes of death for male vs. female cats in the current study's database found the main contributor between birth and 10 years was urinary tract obstruction, which accounted for ~$15\%$ of male deaths in the 3–4 years age interval (data not presented).
## 4.3. Life expectancies by body condition score
Our finding that obese dogs (median BCS 5) had 1.5 years lower LEbirth than dogs with median BCS closer to the ideal (median BCS 3) is consistent with another primarily USA analysis that found overweight and obesity to be associated with a higher instantaneous risk of death and a reduced life span [20]. In that study, the hazard ratio for death in overweight compared with normal weight dogs varied by breed from 1.35 ($99.79\%$ CI: 1.05–1.73) to 2.86 ($99.79\%$ CI: 2.14–3.83) [20]. We observed that the impact of overweight or obese body condition on life expectancy was lower in the oldest dogs. It is hypothesized that this could be the influence of the “obesity paradox”, whereby a higher body condition score or increase in weight may have some protective survival effect in dogs (and other animals and humans) with chronic disease, such as chronic kidney disease [51] and heart failure [52].
Obesity in dogs is associated with a wide diversity of disease types, ranging for example from metabolic such as diabetes mellitus [26], to orthopedic [19], and cancer [53], some of which might impact lifespan. In terms of the relationship between body condition and disease, obesity is likely to be a causative risk factor for many of these conditions e.g., diabetes, but it can be a consequence of some other diseases such as hyperadrenocorticism and hypothyroidism (54–56). Obesity could also be an unrelated comorbidity that exacerbates the severity and/or prognosis of another condition, or contributes to a decision on euthanasia. It is not possible to attribute adiposity in our study as a cause of reduced LEbirth, but since the database included both healthy and diseased animals it is feasible that it played an indirect role through disease association.
A further consideration is that we used the median BCS over a pet's medical history in the database to reflect its weight status over life. This approach could lead to misinterpretation in cases where an age-related disease affected BCS. For example, if an animal with a BCS of 5 in the first third of its life contracted a disease in the second third of its life that led to a reduction in BCS to BCS 3, and with disease progression this declined to BCS 1 in the last third of its life, the animal would nevertheless be analyzed in the median BCS 3 category.
Less than ideal median BCS (1 and 2) was associated with much lower life expectancy than an ideal or obese median BCS. In adult dogs, underweight is a general signalment for serious illness or a poor prognostic factor for diagnosed disease. For example, low BCS in dogs with chronic enteropathy is a predictor for lack of response to treatment and death [57]. Underweight in young animals is perceived as a failure to thrive, and low birth weight puppies have an increased risk of neonatal death [58]. In our population, life expectancy increased significantly between the age interval 0–1 year and 1–2 years for dogs with median BCS 1, and, to a lesser extent, dogs with median BCS 2.
Similar to dogs, overweight and obesity in cats are associated with a wide range of disorders, including endocrine and lipid metabolism disorders, neoplasia, lower urinary tract disease, gastrointestinal disease and dermatoses [30, 59]. However, in contrast to dogs, overweight cats (median BCS 4) had a higher LEbirth than cats of ideal body condition (median BCS 3) as well as obese cats (Figure 4); the difference persisted although reduced with increasing age. This is not so surprising in the light of an Australian study that found that cats with a maximum BCS of 6–8 on a 9-point scale (i.e., overweight to obese) had higher survival over time compared with cats with maximum BCS 3–5 or BCS 9 (obese) [21]. The authors hypothesized that having a BCS of 6–8 might have the least negative effects on lifespan, or that it might be associated with owners who provide better care for their pets. The relationship between BCS and survival is complex. The same study found that the age at which maximum BCS was reached influenced the relative effects of different BCS on lifespan. A factor that might be relevant to this is the association of age with susceptibility to different types of diseases, including acute or chronic, easily treatable or life-threatening. For example, risk of serious viral diseases such as feline infectious peritonitis is higher in younger cats [60], while diabetes mellitus, cardiovascular disease and feline lymphoma are associated with old age (61–63). Finally, the same limitations mentioned above of using median BCS in dogs apply to cats.
As for dogs, underweight cats had a lower life expectancy at all ages than cats with ideal, overweight or obese median BCS. However, life expectancy in underweight cats did not decline; cats of median BCS 1 consistently had almost 1 year of life expectancy and cats of median BCS 2 consistently had ~2 years of life expectancy. A range of interacting factors might influence the life expectancy of cats with low median BCS in different ways. For example, weight loss is one sign of chronic kidney disease in cats [64]; it is often present before CKD has been diagnosed, and lower bodyweight is associated with reduced survival time from diagnosis [65]. It is plausible that owners delay seeking veterinary care for weight loss in cats until CKD is advanced and has a poor prognosis. In Banfield medical records, $52\%$ of dogs and cats diagnosed with CKD die within 18 months and $10\%$ are euthanized within 18 months (unpublished data).
For both dogs and cats, the LEbirth of animals with median BCS 3 was quite stable over survey years, while for animals with median BCS 5 a relatively large increase of LEbirth was observed between 2014 and 2019: 1.51 years for dogs and 1.93 years for cats. Although these findings may be related to advances in veterinary medicine that would allow obese dogs to live longer, they also raise new questions about potential changes in veterinarians' perception of, and attitudes toward obesity. It is possible that a dog considered to be BCS 5 in 2014 would be considered to be BCS 3 according to perceptions in 2019. Increasing awareness of obesity and its health implications in pets could also have driven greater proactivity in interventions to treat obesity-associated diseases.
Our clinical dataset relied upon owners visiting a veterinarian, and the animal having a recorded BCS. There may be different thresholds of bodyweight loss or bodyweight gain that would prompt owners to seek veterinary advice. Delaying a visit until a dog or cat has reached an extreme BCS will worsen the prognosis for the underlying etiology. The decision of an owner to opt for the euthanasia of their dog or cat might also vary by BCS and thereby impact life expectancy. Pet lifestyle can impact survival for medical and non-medical reasons. An indoor lifestyle is associated with feline obesity at 1 year of age [66], and outdoor access poses a greater risk of road traffic accidents [46].
## 4.4. Strengths and limitations of the study
This study benefited from a large dataset of 8.9 million dogs and 2.4 million cats, which is more than 10-fold greater than the dataset used to generate life expectancy tables for Japanese dogs [13] and 100-fold greater than the dataset for life expectancy study of dogs in the UK. The database was large enough for a sub-analysis by BCS, and CIs remained modest with the exception of LEbirth for cats with BCS 5, and life expectancy for dogs and cats with BCS 5 over the age interval 14–15 years, likely due to the small number of obese animals in these age intervals.
Limitations that should be considered include the retrospective nature of the study design, a degree of uncertainty of pet age for dogs and cats acquired as rescued animals or adopted from animal shelters, and the inherent limitations of body condition scoring. In clinical practice body condition scoring is routinely used to estimate adiposity and weight status, scoring systems have been validated against measurement of body fat by dual-energy X-ray absorptionmetry (67–72) and good reproducibility of scoring between different veterinarians has been reported [73]. Nevertheless, body condition scoring does have the disadvantage of being a subjective and semi-quantitative measure, and there is a risk of inconsistencies arising between numerous veterinarians scoring body condition over a number of years. A general shift in veterinarians' perception of obesity over time could also impact the median BCS as previously noted. In this study, medical records were only available for a 9-year period rather than over each pets' lifetime. We used median BCS between 2013 and 2022 to address the complication of changes in BCS over time and to focus on pets that had BCS data covering the majority of their life. A limitation of this was that the number of visits and BCS evaluations was different between individuals in any survey year and across all survey years. The method could not account for the duration of overweight or underweight status, or the age at which the median BCS occurred. Other drawbacks to using median BCS have been discussed above. Another point to note is that although it is used in practice in young animals, the BCS system has not been validated in puppies and kittens. The accuracy of body condition scoring in young and aged animals may be reduced because it does not account for “natural” differences and changes in body composition with age.
The intentional omission of some aspects of life expectancy in our analyses might be regarded as limitations. We decided not to analyze any effect of neuter status, because ~$90\%$ of owned dogs and cats in the USA are neutered, and this is reflected in the clinical Banfield population. Differences in life expectancies between breeds within body size were not captured. The intent was to maintain accuracy in life expectancy tables while recognizing the large size differences between breeds. As population numbers were low for many of the more than 300 breeds in our dataset, we favored statistical accuracy over breed differences. Parallel analyses of health status, recorded causality of death, or reason for euthanasia were beyond the scope of this study and this limited the interpretation of differences between subpopulations.
Most pets that visit Banfield Pet Hospitals have veterinary care as part of an Optimal Wellness Plan, which is a suite of pre-paid services focused on preventive health care, routine diagnostic testing, comprehensive physical examination, and unlimited clinic visits. Wellness Plans could result in a pet being seen more frequently and having an earlier diagnosis and treatment of chronic conditions; they might also drive a proactive reduction in risk factors for disease and improvement in prognosis for disease at diagnosis. Conversely, the use of medical records results in missed visits by animals never presented to veterinary clinics.
A major statistical consideration with databases involving millions of individuals is that tests are often highly significant. A widely used method for assessing the standard error of life expectancy is to use the variance of the conditional probability of death. To address the issue of having such large populations, instead of implementing this method we used Monte Carlo simulations to sample dead populations and recalculated life expectancies 1.106 times. This mitigated the issue, but statistical significances were still high whatever significance level. A complementary strategy would have been to sample the mid-year population in the Monte Carlo simulation.
The analyses presented do not extend to estimations of disability free life expectancy, healthy life expectancy or healthy life years lost. A method has been developed for this [74] and applied to life expectancy data from dogs in the UK [15] and Japan [13] for different breeds and cross-breed dogs [75]. The Banfield database provides additional opportunities to explore healthy life expectancy and specific disease-associated life expectancy in future work.
## 4.5. Summary and closing remarks
To our knowledge this is the first study to generate life expectancy tables for dogs and cats in the USA and the first study to present these by median BCS. Some of the main findings were expected based on direct and indirect evidence in the literature, namely decreased life expectancy for the largest sizes of dogs compared to the smallest, increased life expectancy of female versus male dogs and cats, significantly lower life expectancy at all ages for underweight dogs and cats compared to those with an ideal, overweight or obese median BCS, and a progressive increase in LEbirth for both species between 2013 and 2018. Unexpected findings included the relatively constant life expectancy of underweight cats at all ages, and the higher life expectancy of overweight cats compared with cats that were ideal median BCS. Further study is needed to elucidate apparently complex, interrelated and dynamic relationships between median BCS and life expectancy.
The life expectancy tables presented provide valuable information for practicing veterinarians and a foundation for research hypotheses. A better knowledge of life expectancy is important at the population level as a broad indicator of the general health status of the pet population, reflecting owner attitudes, the accessibility of veterinary care, and advances in veterinary care including preventive medicine. At the level of the individual owner, these life expectancy tables may help inform the choice of species and breed size. Life expectancy tables are especially relevant for adopters of adult animals who need to understand the likely duration of their commitment. Quantification of the reduction in life expectancy associated with overweight and/or obesity is a vital step toward making proactive bodyweight management a healthcare priority for veterinarians and owners. Furthermore, this work is a stepping-stone to the generation of disease-associated life expectancy tables.
## Data availability statement
The data analyzed in this study is subject to the following licenses/restrictions: *Raw data* cannot be shared for legal and privacy restrictions. Requests to access these datasets should be directed to [email protected].
## Author contributions
MM contributed to the study design, data processing and interpretation of the results, performed the statistical analyses, wrote the first draft of the manuscript, and revised the intellectual content of subsequent drafts. FA contributed substantially to the design of the study and the statistical analysis work and critically reviewed the intellectual content of the manuscript. JM, NS, VB, HC, and M-AH contributed to data processing and interpretation of the results and critically reviewed the intellectual content of the manuscript. All authors have approved the final content of the manuscript for publication.
## Conflict of interest
MM, HC, M-AH, and VB were employed by Royal Canin. JM and NS were employed by Banfield Pet Hospital. FA was employed by MAD Environnement. Royal Canin SAS and Banfield Pet Hospital are subsidiaries of Mars Inc.
## 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/fvets.2023.1082102/full#supplementary-material
## References
1. Gompertz B. **XXIV. On the nature of the function expressive of the law of human mortality, and on a new mode of determining the value of life contingencies**. *Philos Transact R Soc London.* (1825.0) **115** 513-83. DOI: 10.1098/rstl.1825.0026
2. 2.Office for National Statistics. Life Expectancy Releases and Their Different Uses. (2018). Available from: https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthandlifeexpectancies/articles/lifeexpectancyreleasesandtheirdifferentuses/2018-12-17 (acessed July 11, 2022).. *Life Expectancy Releases and Their Different Uses.* (2018.0)
3. 3.American Veterinary Medical Association. Covid 19. Available from: https://www.avma.org/resources-tools/one-health/covid-19 (acessed July 10, 2022).. *Covid 19*
4. Lewis TW, Wiles BM, Llewellyn-Zaidi AM, Evans KM, O'Neill DG. **Longevity and mortality in Kennel Club registered dog breeds in the UK in 2014**. *Canine Genet Epidemiol.* (2018.0) **5** 10. DOI: 10.1186/s40575-018-0066-8
5. Adams VJ, Evans KM, Sampson J, Wood JL. **Methods and mortality results of a health survey of purebred dogs in the UK**. *J Small Anim Pract.* (2010.0) **51** 512-24. DOI: 10.1111/j.1748-5827.2010.00974.x
6. Proschowsky HF, Rugbjerg H, Ersbøll AK. **Mortality of purebred and mixed-breed dogs in Denmark**. *Prev Vet Med.* (2003.0) **58** 63-74. DOI: 10.1016/S0167-5877(03)00010-2
7. Michell AR. **Longevity of British breeds of dog and its relationships with sex, size, cardiovascular variables and disease**. *Vet Rec.* (1999.0) **145** 625-9. DOI: 10.1136/vr.145.22.625
8. O'Neill DG, Church DB, McGreevy PD, Thomson PC, Brodbelt DC. **Longevity and mortality of owned dogs in England**. *Vet J.* (2013.0) **198** 638-43. DOI: 10.1016/j.tvjl.2013.09.020
9. O'Neill DG, Church DB, McGreevy PD, Thomson PC, Brodbelt DC. **Longevity and mortality of cats attending primary care veterinary practices in England**. *J Feline Med Surg.* (2015.0) **17** 125-33. DOI: 10.1177/1098612X14536176
10. Urfer SR, Wang M, Yang M, Lund EM, Lefebvre SL. **Risk factors associated with lifespan in pet dogs evaluated in primary care veterinary hospitals**. *J Am Anim Hosp Assoc.* (2019.0) **55** 130-7. DOI: 10.5326/JAAHA-MS-6763
11. Egenvall A, Nødtvedt A, Häggström J, Ström Holst B, Möller L, Bonnett BN. **Mortality of life-insured Swedish cats during 1999–2006: age, breed, sex, and diagnosis**. *J Vet Intern Med.* (2009.0) **23** 1175-83. DOI: 10.1111/j.1939-1676.2009.0396.x
12. Hayashidani H, Omi Y, Ogawa M, Fukutomi K. **Epidemiological studies on the expectation of life for dogs computed from animal cemetery records**. *Nihon Juigaku Zasshi.* (1988.0) **50** 1003-8. DOI: 10.1292/jvms1939.50.1003
13. Inoue M, Hasegawa A, Hosoi Y, Sugiura K. **A current life table and causes of death for insured dogs in Japan**. *Prev Vet Med.* (2015.0) **120** 210-8. DOI: 10.1016/j.prevetmed.2015.03.018
14. Inoue M, Kwan NCL, Sugiura K. **Estimating the life expectancy of companion dogs in Japan using pet cemetery data**. *J Vet Med Sci.* (2018.0) **80** 1153-8. DOI: 10.1292/jvms.17-0384
15. Teng KT, Brodbelt DC, Pegram C, Church DB, O'Neill DG. **Life tables of annual life expectancy and mortality for companion dogs in the United Kingdom**. *Sci Rep.* (2022.0) **12** 6415. DOI: 10.1038/s41598-022-10341-6
16. Hayashidani H, Omi Y, Ogawa M, Fukutomi K. **Epidemiological studies on the expectation of life for cats computed from animal cemetery records**. *Nihon Juigaku Zasshi.* (1989.0) **51** 905-8. DOI: 10.1292/jvms1939.51.905
17. Urfer SR. **Right censored data (‘cohort bias') in veterinary life span studies**. *Vet Rec.* (2008.0) **163** 457-8. DOI: 10.1136/vr.163.15.457
18. Paynter AN, Dunbar MD, Creevy KE, Ruple A. **Veterinary big data: when data goes to the dogs**. *Animals.* (2021.0) **11** 1872. DOI: 10.3390/ani11071872
19. German AJ. **The growing problem of obesity in dogs and cats**. *J Nutr.* (2006.0) **136** 1940S-6S. DOI: 10.1093/jn/136.7.1940S
20. Salt C, Morris PJ, Wilson D, Lund EM, German AJ. **Association between life span and body condition in neutered client-owned dogs**. *J Vet Intern Med.* (2019.0) **33** 89-99. DOI: 10.1111/jvim.15367
21. Teng KT, McGreevy PD, Toribio JL, Raubenheimer D, Kendall K, Dhand NK. **Strong associations of nine-point body condition scoring with survival and lifespan in cats**. *J Feline Med Surg.* (2018.0) **20** 1110-8. DOI: 10.1177/1098612X17752198
22. Bach JF, Rozanski EA, Bedenice D, Chan DL, Freeman LM, Lofgren JL. **Association of expiratory airway dysfunction with marked obesity in healthy adult dogs**. *Am J Vet Res.* (2007.0) **68** 670-5. DOI: 10.2460/ajvr.68.6.670
23. Chiang CF, Villaverde C, Chang WC, Fascetti AJ, Larsen JA. **Prevalence, risk factors, and disease associations of overweight and obesity in cats that visited the veterinary medical teaching hospital at the University of California, Davis from January 2006 to December 2015**. *Top Comp Anim Med.* (2022.0) **47** 100620. DOI: 10.1016/j.tcam.2021.100620
24. Henegar JR, Bigler SA, Henegar LK, Tyagi SC, Hall JE. **Functional and structural changes in the kidney in the early stages of obesity**. *J Am Soc Nephrol.* (2001.0) **12** 1211-7. DOI: 10.1681/ASN.V1261211
25. Lund EM, Armstrong PJ, Kirk CA, Klausmer JS. **Prevalence and risk factors for obesity in adult cats from private US veterinary practices**. *Int J Appl Res Vet Med.* (2005.0) **3** 88-96
26. Lund EM, Armstrong PJ, Kirk CA, Klausner JS. **Prevalence and risk factors for obesity in adult dogs from private US veterinary practices**. *Int J Appl Res Vet Med.* (2006.0) **4** 177-86
27. Manens J, Bolognin M, Bernaerts F, Diez M, Kirschvink N, Clercx C. **Effects of obesity on lung function and airway reactivity in healthy dogs**. *Vet J.* (2012.0) **193** 217-21. DOI: 10.1016/j.tvjl.2011.10.013
28. Parker VJ, Orcutt E, Love L, Cline MG, Murphy M. **Pathophysiology of obesity: comorbidities and anesthetic considerations**. *Obesity in the Dog and Cat* (2019.0) 40-61
29. Tropf M, Nelson OL, Lee PM, Weng HY. **Cardiac and metabolic variables in obese dogs**. *J Vet Intern Med.* (2017.0) **31** 1000-7. DOI: 10.1111/jvim.14775
30. Teng KT, McGreevy PD, Toribio J, Raubenheimer D, Kendall K, Dhand NK. **Associations of body condition score with health conditions related to overweight and obesity in cats**. *J Small Anim Pract.* (2018.0) **59** 603-15. DOI: 10.1111/jsap.12905
31. Tvarijonaviciute A, Ceron JJ, Holden SL, Cuthbertson DJ, Biourge V, Morris PJ. **Obesity-related metabolic dysfunction in dogs: a comparison with human metabolic syndrome**. *BMC Vet Res.* (2012.0) **8** 147. DOI: 10.1186/1746-6148-8-147
32. Sullivan DF. **single index of mortality and morbidity**. *HSMHA Health Rep.* (1971.0) **86** 347-54. DOI: 10.2307/4594169
33. Imai K, Soneji S. **On the estimation of disability-free life expectancy: sullivan's method and its extension**. *J Am Stat Assoc.* (2007.0) **102** 1199-211. DOI: 10.1198/016214507000000040
34. Jagger C, Zeng YI, Crimmins EM, Carrière Y, Robine J-M. **Can we live longer, healthier lives?**. *Longer Life and Healthy Aging* (2006.0) 7-22
35. Lopez AD, Ahmad OB, Guillot M, Inoue M, Ferguson BD, Salomon JA, Murray CJL, Evans DB. **Life tables for 191 countries for 2000: data, methods, results**. *Health Systems Performance Assessment: Debates, Methods and Empiricism* (2001.0) 335
36. Silcocks PBS, Jenner DA, Reza R. **Life expectancy as a summary of mortality in a population: Statistical considerations and suitability for use by health authorities**. *J Epidemiol Community Health.* (2001.0) **55** 38-43. DOI: 10.1136/jech.55.1.38
37. Toson B, Baker A. *Life Expectancy at Birth: Methodological Options for Small Populations* (2003.0) 27
38. Wolfson MC. **Health-adjusted life expectancy**. *Health Rep.* (1996.0) **8** 41-5. PMID: 8844180
39. Salomon JA, Murray CJ. **Compositional models for mortality by age, sex and cause**. *Global Programme on Evidence for Health Policy GPE Discussion Paper Series: No 11* (2001.0)
40. Shkolnikov VM, Andreev EM, McKee M, Leon DA. **Components and possible determinants of the decrease in Russian mortality in 2004-2010**. *Demogr Res.* (2010.0) **28** 917-50. DOI: 10.4054/DemRes.2013.28.32
41. Antolin P. *Longevity risk and Private Pensions. OECD Working Papers on Insurance and Private Pensions* (2007.0) 28
42. Kraus C, Pavard S, Promislow DE. **The size-life span trade-off decomposed: why large dogs die young**. *Am Nat.* (2013.0) **181** 492-505. DOI: 10.1086/669665
43. Tønnessen R, Borge KS, Nødtvedt A, Indrebø A. **Canine perinatal mortality: a cohort study of 224 breeds**. *Theriogenology.* (2012.0) **77** 1788-801. DOI: 10.1016/j.theriogenology.2011.12.023
44. 44.AVMA. Pet Ownership & Demographics Sourcebook: 2017-2018 edition. Washington, DC: American Veterinary Medical Association (2017).. *Pet Ownership & Demographics Sourcebook: 2017-2018 edition* (2017.0)
45. Yordy J, Kraus C, Hayward JJ, White ME, Shannon LM, Creevy KE. **Body size, inbreeding, and lifespan in domestic dogs**. *Conserv Genet* (2020.0) **21** 137-48. DOI: 10.1007/s10592-019-01240-x
46. Tan SML, Stellato AC, Niel L. **Uncontrolled outdoor access for cats: an assessment of risks and benefits**. *Animals* (2020.0) **10** 258. DOI: 10.3390/ani10020258
47. 47.US Food & Drug Administration. Coronavirus (COVID-19) Update: FDA Helps Facilitate Veterinary Telemedicine During Pandemic. (2020). Available from: https://www.fda.gov/news-events/press-announcements/coronavirus-covid-19-update-fda-helps-facilitate-veterinary-telemedicine-during-pandemic (accessed January 27, 2022).. *Coronavirus (COVID-19) Update: FDA Helps Facilitate Veterinary Telemedicine During Pandemic.* (2020.0)
48. Chhetri BK, Berke O, Pearl DL, Bienzle D. **Comparison of risk factors for seropositivity to feline immunodeficiency virus and feline leukemia virus among cats: a case-case study**. *BMC Vet Res* (2015.0) **11** 30. DOI: 10.1186/s12917-015-0339-3
49. Little S, Levy J, Hartmann K, Hofmann-Lehmann R, Hosie M, Olah G. **2020 AAFP feline retrovirus testing and management guidelines**. *J Feline Med Surg* (2020.0) **22** 5-30. DOI: 10.1177/1098612X19895940
50. Segev G, Livne H, Ranen E, Lavy E. **Urethral obstruction in cats: predisposing factors, clinical, clinicopathological characteristics and prognosis**. *J Feline Med Surg* (2011.0) **13** 101-8. DOI: 10.1016/j.jfms.2010.10.006
51. Parker VJ, Freeman LM. **Association between body condition and survival in dogs with acquired chronic kidney disease**. *J Vet Intern Med* (2011.0) **25** 1306-11. DOI: 10.1111/j.1939-1676.2011.00805.x
52. Slupe JL, Freeman LM, Rush JE. **Association of body weight and body condition with survival in dogs with heart failure**. *J Vet Intern Med* (2008.0) **22** 561-5. DOI: 10.1111/j.1939-1676.2008.0071.x
53. Marchi PH, Vendramini TH, Perini MP, Zafalon RV, Amaral AR, Ochamotto VA. **Obesity, inflammation, and cancer in dogs: review and perspectives**. *Front Vet Sci* (2022.0) **9** 1004122. DOI: 10.3389/fvets.2022.1004122
54. Cho KD, Paek J, Kang JH, Chang D, Na KJ, Yang MP. **Serum adipokine concentrations in dogs with naturally occurring pituitary-dependent hyperadrenocorticism**. *J Vet Intern Med* (2014.0) **28** 429-36. DOI: 10.1111/jvim.12270
55. German A. **Obesity in companion animals**. *Int Pract* (2010.0) **32** 42-50. DOI: 10.1136/inp.b5665
56. Scott-Moncrieff JC. **Clinical signs and concurrent diseases of hypothyroidism in dogs and cats**. *Vet Clin North Am Small Anim Pract.* (2007.0) **37** 709-22. DOI: 10.1016/j.cvsm.2007.03.003
57. Benvenuti E, Pierini A, Bottero E, Pietra M, Gori E, Salvadori S. **Immunosuppressant-responsive enteropathy and non-responsive enteropathy in dogs: Prognostic factors, short- and long-term follow up**. *Animals* (2021.0) **11** 2637. DOI: 10.3390/ani11092637
58. Mugnier A, Mila H, Guiraud F, Brévaux J, Lecarpentier M, Martinez C. **Birth weight as a risk factor for neonatal mortality: breed-specific approach to identify at-risk puppies**. *Prev Vet Med.* (2019.0) **171** 104746. DOI: 10.1016/j.prevetmed.2019.104746
59. German AJ, Ryan VH, German AC, Wood IS, Trayhurn P. **Obesity, its associated disorders and the role of inflammatory adipokines in companion animals**. *Vet J.* (2010.0) **185** 4-9. DOI: 10.1016/j.tvjl.2010.04.004
60. Worthing KA, Wigney DI, Dhand NK, Fawcett A, McDonagh P, Malik R. **Risk factors for feline infectious peritonitis in Australian cats**. *J Feline Med Surg.* (2012.0) **14** 405-12. DOI: 10.1177/1098612X12441875
61. Prahl A, Guptill L, Glickman NW, Tetrick M, Glickman LT. **Time trends and risk factors for diabetes mellitus in cats presented to veterinary teaching hospitals**. *J Feline Med Surg.* (2007.0) **9** 351-8. DOI: 10.1016/j.jfms.2007.02.004
62. Isomura R, Yamazaki M, Inoue M, Kwan NC, Matsuda M, Sugiura K. **The age, breed and sex pattern of diagnosis for veterinary care in insured cats in Japan**. *J Small Anim Pract.* (2017.0) **58** 89-95. DOI: 10.1111/jsap.12617
63. Economu L, Stell A, O'Neill DG, Schofield I, Stevens K, Brodbelt D. **Incidence and risk factors for feline lymphoma in UK primary-care practice**. *J Small Anim Pract.* (2021.0) **62** 97-106. DOI: 10.1111/jsap.13266
64. Sparkes AH, Caney S, Chalhoub S, Elliott J, Finch N, Gajanayake I. **ISFM consensus guidelines on the diagnosis and management of feline chronic kidney disease**. *J Feline Med Surg.* (2016.0) **18** 219-39. DOI: 10.1177/1098612X16631234
65. Freeman LM, Lachaud MP, Matthews S, Rhodes L, Zollers B. **Evaluation of weight loss over time in cats with chronic kidney disease**. *J Vet Intern Med.* (2016.0) **30** 1661-6. DOI: 10.1111/jvim.14561
66. Rowe E, Browne W, Casey R, Gruffydd-Jones T, Murray J. **Risk factors identified for owner-reported feline obesity at around one year of age: dry diet and indoor lifestyle**. *Prev Vet Med.* (2015.0) **121** 273-81. DOI: 10.1016/j.prevetmed.2015.07.011
67. Burkholder WJ. **Use of body condition scores in clinical assessment of the provision of optimal nutrition**. *J Am Vet Med Assoc.* (2000.0) **217** 650-4. DOI: 10.2460/javma.2000.217.650
68. Jeusette I, Greco D, Aquino F, Detilleux J, Peterson M, Romano V. **Effect of breed on body composition and comparison between various methods to estimate body composition in dogs**. *Res Vet Sci.* (2010.0) **88** 227-32. DOI: 10.1016/j.rvsc.2009.07.009
69. Laflamme D. **Development and validation of a body condition score system for dogs**. *Canine Pract.* (1997.0) **22** 10-5
70. Laflamme D. **Development and validation of a body condition score system for cats: a clinical tool**. *Feline Pract.* (1997.0) **25** 13-8
71. Mawby DI, Bartges JW, d'Avignon A, Laflamme DP, Moyers TD, Cottrell T. **Comparison of various methods for estimating body fat in dogs**. *J Am Anim Hosp Assoc.* (2004.0) **40** 109-14. DOI: 10.5326/0400109
72. Santarossa A, Parr JM, Verbrugghe A. **The importance of assessing body composition of dogs and cats and methods available for use in clinical practice**. *J Am Vet Med Assoc.* (2017.0) **251** 521-9. DOI: 10.2460/javma.251.5.521
73. Shoveller AK, DiGennaro J, Lanman C, Spangler D. **Trained vs untrained evaluator assessment of body condition score as a predictor of percent body fat in adult cats**. *J Feline Med Surg.* (2014.0) **16** 957-65. DOI: 10.1177/1098612X14527472
74. Skiadas CH, Skiadas C, Skiadas CH, Skiadas C. **Direct healthy life expectancy estimates from life tables with a sullivan extension. Bridging the gap between hale and eurostat estimates**. *Demography of Population Health, Aging and Health Expenditures* (2020.0) 25-42
75. Skiadas CH. **Expanding the life tables for companion dogs in UK and Japan to include the healthy life expectancy. Version 1, last edited September 6, 2022**. *SocArXiv.* (2022.0). DOI: 10.31235/osf.io/ftky9
|
---
title: Extracellular vesicles secreted by human gingival mesenchymal stem cells promote
bone regeneration in rat femoral bone defects
authors:
- Situo Wang
- Ziwei Liu
- Shuo Yang
- Na Huo
- Bo Qiao
- Tong Zhang
- Juan Xu
- Quan Shi
journal: Frontiers in Bioengineering and Biotechnology
year: 2023
pmcid: PMC9989199
doi: 10.3389/fbioe.2023.1098172
license: CC BY 4.0
---
# Extracellular vesicles secreted by human gingival mesenchymal stem cells promote bone regeneration in rat femoral bone defects
## Abstract
Extracellular vesicles (EVs), important components of paracrine secretion, are involved in various pathological and physiological processes of the body. In this study, we researched the benefits of EVs secreted by human gingival mesenchymal stem cells (hGMSC-derived EVs) in promoting bone regeneration, thereby providing new ideas for EVs-based bone regeneration therapy. Here, we successfully demonstrated that hGMSC-derived EVs could enhance the osteogenic ability of rat bone marrow mesenchymal stem cells and the angiogenic capability of human umbilical vein endothelial cells. Then, femoral defect rat models were created and treated with phosphate-buffered saline, nanohydroxyapatite/collagen (nHAC), a grouping of nHAC/hGMSCs, and a grouping of nHAC/EVs. The results of our study indicated that the combination of hGMSC-derived EVs and nHAC materials could significantly promote new bone formation and neovascularization with a similar effect to that of the nHAC/hGMSCs group. Our outcomes provide new messages on the role of hGMSC-derived EVs in tissue engineering, which exhibit great potential in bone regeneration treatment.
## 1 Introduction
The repair and regeneration of bone defects caused by tumors, trauma, and infection have always been a hot issue in the field of orthopedics and stomatology (Nguyen, et al., 2018). Autologous bone grafts are considered the “gold standard” for bone repair (Kumar, et al., 2016). However, they have some disadvantages, such as the need for a secondary operation, defects in the donor site, and unpredictable autogenous bone absorption (Valtanen, et al., 2021).
In recent years, tissue engineering strategies based on mesenchymal stem cells (MSCs) have been widely used in the field of bone regeneration. Human gingival mesenchymal stem cells (hGMSCs) are adult stem cells isolated from the gingival lamina propria with multidirectional differentiation potential and high proliferation characteristics; in addition, they have abundant sources and can be easily harvested minimal invasively (Fawzy El-Sayed and Dörfer, 2016). Compared with bone marrow mesenchymal stem cells (BMSCs), hGMSCs have the advantages of faster proliferation and more stable morphology in vitro (Tomar, et al., 2010). In addition, as the majority of hGMSCs are derived from cranial neural crest cells, hGMSCs have good tissue regeneration and immunomodulation functions (Xu, et al., 2013). hGMSCs have been extensively investigated for bone regeneration and have shown good application effects (Al-Qadhi, et al., 2020; Hasani-Sadrabadi, et al., 2020). However, there are disadvantages to MSC transplantation, such as a low survival rate of transplanted cells, tumorigenic effects and immunological rejection (Eggenhofer, et al., 2014). Therefore, avoiding the risk of using MSCs or finding substitutes for MSCs to achieve cell-free therapy is one of the problems to be solved at present.
A recent basic study indicated that the tissue repair function of MSCs is mainly exerted by paracrine secretion of bioactive molecules (Liang, et al., 2014; Najar, et al., 2021; Williams et al., 2022). As an important paracrine factor, extracellular vesicles (EVs) are lipid bilayer nanovesicles secreted by living cells and their classification and nomenclature were formulated by the International Society of Extracellular Vesicles (ISEV) (Boere, et al., 2018). EVs carry a variety of bioactive molecules, such as microRNAs (miRNAs), mRNAs, lipids, and proteins, and are widely distributed in body fluids, such as breast milk, saliva, urine and bile (Trajkovic, et al., 2008; Rani, et al., 2015). After being secreted, EVs can be absorbed by receptor cells through ligand/receptor recognition, membrane fusion or phagocytosis and can regulate cell-to-cell communication by transmitting bioactive molecules (Tarasov, et al., 2021). It has been reported that MSC-derived EVs have shown remarkable therapeutic effects in many disease models, such as cardiovascular diseases, nervous system diseases and immune system diseases (Moghadasi, et al., 2021). Accumulating studies have shown that MSC-derived EVs can effectively promote the repair and regeneration of bone defects, and this effect is closely related to the regulation of osteogenesis and angiogenesis-related cells by MSC-derived EVs (Qin, et al., 2016). However, the therapeutic effect of hGMSC-derived EVs on bone defect repair and regeneration is unclear.
Therefore, in this study, we examined the effect of hGMSC-derived EVs on osteogenesis and angiogenesis by treating rat bone marrow mesenchymal stem cells (rBMSCs) and human umbilical vein endothelial cells (HUVECs) with hGMSC-derived EVs. In addition, we evaluated the bone repair capacity of nanohydroxyapatite/collagen (nHAC) scaffolds loaded with hGMSC-derived EVs on rat femoral defects. Our study showed that the combination of hGMSC-derived EVs/nHAC could promote the repair and regeneration of bone defects by accelerating new bone formation and angiogenesis, potentially providing application value for the treatment of bone defects.
## 2.1 Cell isolation and culture
The methods for extraction and culture of primary hGMSCs were as previously described (Shi, et al., 2017). Gingival tissue was obtained from healthy young patients undergoing tooth crown lengthening operation, impacted third molar extraction and secondary implant surgery. Briefly, the gingival tissue was digested in 2 mg/ml dispase (Roche) at 4°C for 12 h after several rinses with phosphate-buffered saline (PBS). Then, the lamina propria was separated from the gingival tissue, minced and digested with 2 mg/ml collagenase IV (Roche) at 37 °C for 1 h. Afterward, the cell and tissue pellets were cultured in Dulbecco’s modified Eagle’s medium, nutrient mixture F-12 (DMEM/F12, Gibco) supplemented with $10\%$ fetal bovine serum (FBS, Gibco), 100 U/ml penicillin, 100 μg/ml streptomycin and 0.25 μg/ml amphotericin B at 37°C with $5\%$ CO2. hGMSCs at passages three to six were used in this experiment. The methods for extraction and culture of primary rBMSCs were as previously described (Liu, et al., 2020). Briefly, bone marrow was flushed from the femoral bones of SD suckling rats using α-minimum essential medium (α-MEM, Gibco) supplemented with $10\%$ fetal bovine serum (FBS, Gibco), 100 U/ml penicillin, 100 μg/ml streptomycin and 0.25 μg/ml amphotericin B. The cell and tissue pellets were then cultured in α-MEM complete medium. HUVECs were purchased from PROCELL (Wuhan, China) and cultured in HUVEC special medium (Procell). All experimental procedures obtained approval from Clinical Ethics Committee of the Chinese PLA General Hospital.
## 2.2 Isolation and identification of hGMSC-Derived EVs
hGMSCs were cultured in exosome-free FBS medium to collect conditioned medium (Théry, et al., 2006). First, the cells were cultured in osteogenic induction medium (OM):α-MEM supplemented with $10\%$ FBS, 0.1 μmol/L dexamethasone (Gibco), 10 mmol/L β-glycerol sodium phosphate (Gibco) and 50 μg/mL ascorbic acid (Gibco) for 3 days, then OM was replaced with α-MEM medium supplemented with $10\%$ exosome-free FBS for culture with an additional 2 days to collect the conditioned medium. Afterward, the conditioned medium was centrifuged at 500 × g for 10 min and 1,000 × g for 30 min and then filtered through a 0.22 μm sterilized filter. The filtered medium was added to an ultrafiltration centrifuge tube (15 ml Amicon Ultra 30kD, Millipore) and centrifuged at 5000 × g for 20 min to concentrate the medium. Subsequently, the concentrated supernatant was ultracentrifuged at 100,000 × g for 60 min, and then the supernatant was replaced with PBS for the same operation for 60 min to obtain the EVs. The EVs were stored at −80°C.
The morphology of EVs was observed by transmission electron microscopy (TEM, HITACHI). Briefly, 5 μl of EVs was loaded onto a copper grid for 5 min, and the excess liquid was removed by filter paper. After staining with $2\%$ uranyl acetate dihydrate for 1 min, the sample was detected by TEM. The particle size distribution was examined by using nanoparticle tracking analysis (NTA). In addition, Western blotting was performed according to standard protocol as previously reported (Swanson, et al., 2020) to detect the EVs marker CD9 (ab236630, Abcam), tumor susceptibility gene (Tsg) 101 (ab133586, Abcam) and heat shock protein (Hsp) 70 (ab5439, Abcam). All antibodies were diluted at a concentration ratio of 1:1,000. The protein concentrations of the EVs were measured by using the BCA Protein Assay Kit (Servicebio).
## 2.3 BMSC osteogenic differentiation assay
Four groups were established as follows: 1) OM (control), 2) OM complemented with 25 μg/ml hGMSC-derived EVs (25 μg/ml EVs), 3) OM complemented with 50 μg/ml hGMSC-derived EVs (50 μg/ml EVs), and 4) OM complemented with 100 μg/ml hGMSC-derived EVs (100 μg/ml EVs).
Alkaline phosphatase (ALP) staining (Beyotime) and an ALP assay kit (Beyotime) were used to assess ALP activity after 14 days of osteoinduction. Alizarin red staining (Solarbio) was conducted to assess mineralization following 14 days of osteoinduction. To quantify the matrix calcifications, the calcium was deposited with $10\%$ cetylpyridinium chloride (Sigma) for 60 min and measured by the absorbance at 562 nm.
To further examine the expression of osteogenesis-related genes and proteins, real-time qPCR and Western blotting were conducted. The operation steps of real-time qPCR are briefly described as follows. Total RNA was extracted from cells by TRIzol (Servicebio) after 14 days of osteoinduction and then synthesized into cDNA by using StarScript III RT Mix (Genstar). Afterward, quantitative polymerase chain reaction was performed using StarScript III SYBR Mix (Genstar). The primers for ALP, osteocalcin (OCN) and runt-related transcription factor 2 (RUNX2) are presented in Table 1. In addition, total protein was isolated from cells using cell lysis buffer (Beyotime) after 14 days of osteoinduction. Western blotting was performed to detect the expression of osteogenesis-related proteins ALP (No. 60294-1-Ig, Proteintech), OCN (GTX64348, GeneTex) and RUNX2 (No. 20700-1-AP, Proteintech).
**TABLE 1**
| Gene | Forward (5′-3′) | Reverse (3′-5′) |
| --- | --- | --- |
| ALP | TGGTACTCGGACAATGAGATGC | GCTCTTCCAAATGCTGATGAGGT |
| OCN | AGGGCAGTAAGGTGGTGAATAGA | GAAGCCAATGTGGTCCGCTA |
| RUNX2 | CAGTATGAGAGTAGGTGTCCCGC | AAGAGGGGTAAGACTGGTCATAGG |
| GAPDH | GGCACAGTCAAGGCTGAGAATG | ATGGTGGTGAAGACGCCAGTA |
## 2.4 HUVEC angiogenic differentiation assay
HUVECs were cultured in HUVEC special medium with or without different concentrations of EVs (25 μg/ml, 50 μg/ml, and 100 μg/ml). Tube formation assays were performed to assess the impact of hGMSC-derived EVs on angiogenesis. HUVECs pretreated with or without EVs were seeded into 24-well plates covered with Matrigel (BD Biosciences). Images of tube formation were obtained by microscopy after 6 h of culture. ImageJ software was used to quantitatively analyze the number and total length of tubes.
## 2.5.1 Preparation and characterization of nHAC-containing cells and EVs
nHAC materials with diameters of 3.5 mm, comprising collagen I and nanohydroxyapatite, were purchased from Allgens Medical Co., Ltd. (Beijing, China). Small pieces 4 mm in length were cut from the nHAC material with a scalpel as scaffolds. Each scaffold was injected with 50 μl of 4 × 106 cells/ml cell solution and then transferred to 24-well plates.
nHAC materials with hGMSCs were cultured in DMEM/F12 complete medium for 3 days. Scaffolds with or without hGMSCs were observed under a scanning electron microscope (JEOL).
Then, 100 μl of EVs at a concentration of 1 μg/μl was injected into each scaffold. To investigate the loading of EVs in the scaffold, EVs were labeled green with DIO (green) dye (Abmole) according to the manufacturer’s protocol. The control group was injected with the same volume of PBS. Fluorescence expression was examined under a laser scanning confocal microscope (LSCM, ZEISS).
## 2.5.2 Critical-sized femoral defect model
The animal experiments in this study were approved by Animal Care and Use Committee of Chinese PLA General Hospital. A total of 60 male Sprague-Dawley rats (12 weeks old, SPF) were purchased from Sibeifu Biotechnology Co., Ltd. (Beijing, China). The rats were randomly divided into four groups as follows: 1) defects with PBS treatment (control, $$n = 15$$); 2) defects treated with nHAC scaffolds (nHAC, $$n = 15$$); 3) defects treated with nHAC scaffolds loaded with hGMSCs (nHAC/hGMSCs, $$n = 15$$); and 4) defects treated with nHAC scaffolds loaded with EVs (nHAC/EVs, $$n = 15$$). The femoral defect model was established as previously described (Wang, et al., 2020). Briefly, the rats were anesthetized by intraperitoneal injection of $2\%$ sodium pentobarbital solution (45 mg/kg). Then, Critical-sized defects of 4 × 4 × 4 mm3 were created at the lateral femoral condyle. After 4, 8, and 12 weeks, five animals were sacrificed in each group. Then, the femoral condyle defect sites were obtained and fixed in $4\%$ paraformaldehyde for 48 h.
## 2.5.3 Gross observation and imaging examination
The specimens were examined under a stereomicroscope (Nikon). X-ray images were then obtained by a Faxitron cabinet X-ray system to observe defect healing. The femoral condyles with defects were scanned with a micro-CT scanner (Skyscan). Three-dimensional (3D) reconstruction was performed and analyzed using 3D visualization software (Skyscan). The BMD, bone volume/tissue volume (BV/TV%), trabecular thickness (Tb. Th), and trabecular separation/spacing (Tb. Sp) were calculated.
## 2.5.4 Histological and immunohistochemical (IHC) analysis
After micro-CT analysis, the specimens were decalcified using $10\%$ EDTA (pH 7.4) for 30 days, dehydrated and embedded in paraffin. Ultimately, the specimens were cut into 4-μm-thick sections. HE, Masson and Goldner staining were conducted to assess bone healing in the defect sites. To further assess new bone formation and neovascularization in femoral condyle defect sites, immunohistochemical staining for osteogenesis-related protein OCN and angiogenesis-related protein CD34 was performed. The primary antibodies anti-OCN (Servicebio) and anti-CD34 (Servicebio) were diluted 1:500 and used according to the manufacturer’s instructions.
## 2.6 Statistical analysis
All data are presented as the mean ± standard deviation for three experiments per group. Student’s t-test was used for two-group comparisons, and one-way ANOVA was used for comparisons among three or four groups. $p \leq 0.05$ was considered statistically significant.
## 3.1 Characterization of hGMSC-derived EVs
The TEM analysis showed that hGMSC-derived EVs had a cup-shaped morphology with a bilayer membrane structure (Figure 1A). The NTA analysis revealed that the peak of the diameter distribution of these nanoparticles was approximately 120 nm, and it was 127.1 ± 37.6 nm in the quantitative analysis (Figure 1B). The Western blotting results demonstrated that hGMSC-derived EVs expressed CD9, TSG101 and HSP70 (Figure 1C).
**FIGURE 1:** *Characterization of hGMSC-derived EVs (A) hGMSC-derived EVs morphology observed by TEM (B) Particle size distribution of hGMSC-derived EVs detected by NTA (C) Western blotting results of the EVs surface markers CD9, Tsg101, and Hsp 70.*
## 3.2 hGMSC-derived EVs promote the osteogenic differentiation of BMSC
To investigate the effect of hGMSC-derived EVs on the osteogenic differentiation of rBMSCs, rBMSCs were cultured in OM with or without different concentrations of EVs (25 μg/ml, 50 μg/ml, and 100 μg/ml). Following 14 days of induction, ALP staining and ALP activity in rBMSCs were significantly increased in the EVs groups compared with the control group, among which the 50 μg/ml EVs group had the best effect (Figures 2A, B). Additionally, alizarin red staining revealed that the mineralization capacity of rBMSCs was enhanced by EVs with the best effect of 50 μg/ml (Figures 2A, C). Likewise, osteogenic mRNA and protein expression (ALP, OCN and RUNX2) was upregulated by EVs, with the highest level in the 50 μg/ml EVs group (Figures 2D, E).
**FIGURE 2:** *Effects of EVs on the osteogenic differentiation of rBMSCs (A) The observation of ALP and alizarin red staining (B) Quantification of ALP activity (C) Quantification of alizarin red staining (D) The expression of osteogenic genes (ALP, OCN and RUNX2) (E) The expression of osteogenic proteins (ALP, OCN and RUNX2). a, p < 0.05 compared with the control group; b, p < 0.05 compared with the 25 μg/ml EVs group; c, p < 0.05 compared with the 50 μg/ml EVs group.*
## 3.3 hGMSC-derived EVs promote the angiogenic capacity of HUVEC
To evaluate the effect of hGMSC-derived EVs on the angiogenic ability of HUVECs, a tube formation assay was conducted. As shown in Figure 3A, the HUVECs in the EVs groups exhibited stronger angiogenic ability than those in the control group, and this ability was enhanced with increasing EVs concentration. Similarly, the quantitative analysis also showed that the number and total length of tubes were significantly higher in the EVs groups (Figures 3B, C).
**FIGURE 3:** *Effects of EVs on the angiogenic capacity of HUVECs (A) Image of the tube formation assay (B) Quantification analysis of tube numbers (C) Quantification analysis of tube length. a, p < 0.05 compared with the control group; b, p < 0.05 compared with the 25 μg/ml EVs group; c, p < 0.05 compared with the 50 μg/ml EVs group.*
## 3.4 hGMSC and EVs detection from the nHAC scaffold
SEM showed that the nHAC material had a porous structure with a uniform pore size of approximately 50–150 μm (Figure 4A). After 3 days of culture, hGMSCs on the nHAC scaffold grew well and adhered to the surface of the material. Moreover, secreted filamentous extracellular matrix around cells could be observed under high magnification (Figure 4B). After DIO-labeled EVs were added to the nHAC material, a large amount of green fluorescence was observed on the material by LSCM, and no fluorescence was detected in the control group (Figure 4C).
**FIGURE 4:** *Detection of hGMSCs and EVs on the nHAC scaffold (A) SEM image of the nHAC scaffold (a: 200×; b: 600×) (B) SEM image of the nHAC scaffold with hGMSCs (a: 1100×; b: 2200×) (C) LSCM images of the nHAC scaffold with DiO-labeled EVs.*
## 3.5 Cross observation and imaging analysis of bone regeneration
A total of 60 Sprague-Dawley rats with femoral defects were divided into four groups (control, nHAC, nHAC/hGMSCs and nHAC/EVs; $$n = 15$$/group) and euthanized by dislocation of cervical vertebrae under deep anesthesia in three different times (at 4, 8, and 12 weeks). As shown in Figure 5A, defects gradually decreased with healing time, as observed under a stereomicroscope. As expected, the defects of each group implanted with nHAC healed better than those of the control group. Among them, the defect healing of the nHAC/hGMSCs group and nHAC/EVs group was better than that of the nHAC group. Moreover, the results of X-ray and micro-CT imaging also showed that the bone healing effect of the nHAC/hGMSCs group and nHAC/EVs group was better than that of the other two groups, and the defects in the nHAC group healed better (Figures 5B, C). According to 3D reconstruction analysis, the BMD, BV/TV%, Tb. Th, and Tb. Sp results further revealed that more new bone formation was discovered in the HAC/hGMSCs group and nHAC/EVs group than in the nHAC group and control group. In addition, no significant difference was found in the amount of new bone formation between the HAC/hGMSCs group and the nHAC/EVs group (Figure 5C).
**FIGURE 5:** *Cross observation and imaging analysis of bone regeneration (A) Cross observation images of bone defect sites (B) X-ray images of bone defect sites (C) Micro-CT images and analysis of bone formation using BMD, BV/TV%, Tb.Th, and Tb. Sp. a, p < 0.05 compared with the control group; b, p < 0.05 compared with the nHAC group.*
## 3.6 Histological results of bone regeneration
The HE staining results at 4, 8, and 12 weeks are shown in Figure 6. The regenerative bone mass in the femoral defect area increased over time in each group, although it was higher in groups implanted with nHAC, among which the nHAC/hGMSCs group and nHAC/EVs group were better. The Masson and Goldner staining results indicated that the nHAC/hGMSCs group and nHAC/EVs group at each time point had more collagen deposition and new bone formation than the other two groups (Figures 7, 8).
**FIGURE 6:** *Representative images of HE staining of the bone defect area. The 200× image is an enlargement in the dashed box of the 40× image.* **FIGURE 7:** *Masson staining results of bone regeneration (A) Representative images of Masson staining of the bone defect area. The 200× image is an enlargement in the dashed box of the 40× image (B) Quantitative analysis of new bone. a, p < 0.05 compared with the control group; b, p < 0.05 compared with the nHAC group.* **FIGURE 8:** *Goldner staining results of bone regeneration (A) Representative images of Goldner staining of the bone defect area. The 200× image is an enlargement in the dashed box of the 40× image (B) Quantitative analysis of new bone. a, p < 0.05 compared with the control group; b, p < 0.05 compared with the nHAC group.*
## 3.7 Immunohistochemical staining results of bone regeneration
As an important indicator for evaluating osteogenic ability, OCN-positive areas can be stained dark brown by immunohistochemical staining. IHC staining indicated that the expression levels of OCN in groups implanted with nHAC at each time point were higher than those in the control group. Moreover, the nHAC/hGMSCs group and nHAC/EVs group had the highest OCN expression with no significant difference (Figures 9A, B). Additionally, more angiogenesis-related protein CD34 was observed in the nHAC/hGMSCs group and nHAC/EVs group than in the other two groups, suggesting that more new blood vessels were formed in the nHAC/hGMSCs group and nHAC/EVs group (Figures 9C, D).
**FIGURE 9:** *Immunohistochemical staining results of bone regeneration (A) Representative images of OCN staining (B) Quantitative analysis of OCN expression (C) Representative images of CD34 staining (D) Quantitative analysis of CD34 expression. a, p < 0.05 compared with the control group; b, p < 0.05 compared with the nHAC group.*
## 4 Discussion
*In* general, bone tissue shows good tissue repair function after trauma, but the repair of large-scale bone defects is still a difficult problem in clinical therapy. Bone regeneration is a complex process involving many aspects, such as angiogenesis, osteogenesis, and anti-inflammation (Dimitriou, et al., 2011). Although MSC transplantation therapy has shown good application effects for the repair of bone defects, recent studies have shown that its specific mechanism is mainly accomplished by paracrine effects of MSCs (Rani, et al., 2015; Reis, et al., 2018). As critical paracrine factors secreted by cells, EVs can mediate intercellular communication by delivering proteins, mRNAs, miRNAs and other substances to recipient cells, thereby regulating the biological functions of target cells (Hu, et al., 2021; Tarasov, et al., 2021).
hGMSCs are a kind of MSC with multiple differentiation potential and strong self-renewal ability isolated from the human gingival lamina propria (Fawzy El-Sayed and Dörfer, 2016). The special tissue living environment of the gingiva also makes hGMSCs different from other MSCs. Compared with other MSCs, hGMSCs are easy to obtain, rich in sources, and have good biological properties, showing good application prospects in cell therapy and regenerative medicine (Fawzy El-Sayed and Dörfer, 2016). Several in vivo studies have shown that hGMSCs can promote bone regeneration (Xu, et al., 2014; Al-Qadhi, et al., 2020; Hasani-Sadrabadi, et al., 2020). Xu et al. ( Xu, et al., 2014) found that hGMSCs could promote the repair of mandibular defects in mice by intravenous injection of hGMSCs applied to the defect. In addition, Al-Qadhi et al. ( Al-Qadhi, et al., 2020) indicated that hGMSCs had osteogenic ability similar to that of BMSCs in a tibial defect animal model. However, few studies have reported on bone tissue engineering with hGMSC-derived EVs. A study performed by Jiang et al. [ 2020] demonstrated that hGMSC-derived EVs could promote the migration and osteogenic differentiation of preosteoblasts. However, the effect of hGMSC-derived EVs on bone defect repair in vivo has not been reported. Therefore, the study was conducted in vitro and in vivo to deeply explore the bone repair effect of hGMSC-derived EVs.
With the deepening of research, it was found that the osteogenic effect of MSC-derived EVs without osteogenic induction is not obvious (Zhang, et al., 2019). Moreover, Liu et al. ( Liu, et al., 2020) indicated that the osteogenic differentiation capacity of MSCs could be enhanced by osteogenic induction and that the enhancement effect was related to the time of osteogenic induction of MSCs. This study further showed that the osteogenic effect of EVs after 3 days and 14 days of induction was better (Liu, et al., 2020). However, the stemness and paracrine capacity of MSCs were reduced as the induction time increased (Yeo, et al., 2013). Therefore, in this study, we chose EVs derived from hGMSCs after 3 days of osteogenic induction.
Bone regeneration involves the participation of a variety of cells, and bone-related cells and blood vessel-related cells play an important role (Dimitriou, et al., 2011; Sun, et al., 2022). BMSCs are adult stem cells present in the bone marrow stroma that are activated and mobilized upon injury and serve as the main repair cell type in bone regeneration (Deschaseaux, et al., 2009).
Endothelial cells (ECs) are the first cells to enter the bone marrow after bone tissue injury and coordinate tissue development, maintenance, and regeneration by secreting beneficial vascular secretory signals (Kenswil, et al., 2021). When bone defects occur, BMSCs and ECs can synergistically regulate the bone microenvironment of the defect site and promote bone regeneration by accelerating angiogenesis (Cheng, et al., 2022). Therefore, in this study, we selected rat bone marrow mesenchymal stem cells (rBMSCs) and human umbilical vein endothelial cells (HUVECs) to evaluate the effects of hGMSC-derived EVs on the osteogenic ability of osteoblasts and the angiogenic capability of endothelial cells in vitro.
In this study, we found that hGMSC-derived EVs could promote osteogenic differentiation and upregulate the expression of ALP, OCN and RUNX2 osteogenic genes and proteins in rBMSCs. To evaluate the effect of different concentrations of EVs on the osteogenic ability of rBMSCs, we treated rBMSCs with 25 μg/ml, 50 μg/ml and 100 μg/ml EVs. The results showed that 50 μg/ml EVs had the strongest osteogenic ability. In addition, we also demonstrated that hGMSC-derived EVs could enhance the angiogenic ability of HUVECs in vitro such that the higher the EVs concentration was, the better the enhancement effect. MSC-derived EVs could promote the osteogenesis of BMSCs and the angiogenesis of HUVECs, which may contribute to the repair of bone defects. Wu et al. ( Wu, et al., 2019) found that EVs derived from stem cells from human exfoliated deciduous teeth could enhance the repair of alveolar bone defects through the regulation of osteogenesis of BMSCs and angiogenesis of HUVECs. Moreover, Zhang et al. ( Zhang, et al., 2020) also found that BMSC-derived EVs could accelerate fracture healing of non-union through the promotion of osteogenesis and angiogenesis. In this study, hGMSC-derived EVs enhanced the osteogenic ability of rBMSCs and the angiogenic capability of HUVECs. Therefore, we speculated that hGMSC-derived EVs could be an effective approach for bone regeneration in vivo.
In animal models of femoral defects, nanohydroxyapatite/collagen was selected as scaffolding material to carry EVs to the site of defects to verify the repair effect of hGMSC-derived EVs on bone defects. In addition, we set up an nHAC/hGMSC group to better evaluate the role of hGMSC-derived EVs. Our data showed that nHAC materials were biocompatible and could be used as application vectors for cells and EVs. At 4, 8 and 12 weeks postsurgery, the bone repair effect of the nHAC/hGMSCs group and the nHAC/EVs group was better than that of the nHAC group and the control group, while the bone repair effect of the nHAC/hGMSCs group and the nHAC/EVs group was not significantly different. Moreover, HE staining, Masson staining, Goldner staining, and immunohistochemical staining for OCN and CD34 showed that more new bone and new blood vessels were produced after treatment with hGMSC-derived EVs, with effects comparable with those of transplanted hGMSCs. In bone tissue engineering, the combination of bone repair materials and bioactive molecules that induce osteogenesis could enhance the function of biomaterials and promote the aggregation and differentiation of osteoblasts, thereby accelerating the repair and regeneration of bone defects (Zuo, et al., 2022). At present, nHAC materials have been widely used in the repair of bone defects, and it is a feasible application strategy to combine them with nanoactive molecular EVs that have good bone induction ability and proangiogenic ability. In this study, hGMSC-derived EVs were able to stimulate osteogenesis of rBMSCs and angiogenesis of HUVECs in vitro. In addition, OCN and CD34 were highly expressed in the bone defect areas of the EVs treatment group. Therefore, hGMSC-derived EVs may promote osteogenesis and angiogenesis in bone defect areas by influencing the biological function of endogenous cells, such as BMSCs and ECs. Our findings suggested that the combination of hGMSC-derived EVs with nHAC scaffolds was a reliable method for bone defect repair.
However, this study has some limitations. First, this study did not explore in depth the possible mechanisms of hGMSC-derived EVs in osteogenic differentiation and bone repair at the molecular level. Second, bone regeneration is achieved through bone formation and bone resorption with the participation of osteoblasts and osteoclasts. This study explored the effect of hGMSC-derived EVs on the osteogenic differentiation of osteoblasts, but there is a lack of studies on the biological characteristics of osteoclasts. Finally, in this study, the femoral defect of ordinary rats was selected as the model, and whether it is suitable for femoral defects in osteoporotic rats is not known.
## 5 Conclusion
In this study, we successfully extracted EVs derived from hGMSCs and combined the EVs with a nanohydroxyapatite/collagen scaffold for bone defect repair. Our results demonstrated that the combination of hGMSC-derived EVs and nHAC could significantly promote bone regeneration by advancing osteogenesis and angiogenesis. Therefore, this strategy could serve as a clinical therapy for bone regeneration.
## 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 Clinical Ethics Committee of the Chinese PLA General Hospital. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Animal Care and Use Committee of Chinese PLA General Hospital.
## Author contributions
QS, JX, and TZ conceived of the study and designed the experiments jointly. SW, ZL, SY, NH, and BQ carried out the experiments and analyzed the experimental results. SW, ZL, and SY summarized the literature and wrote the manuscript. All authors approved 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. Al-Qadhi G., Soliman M., Abou-Shady I., Rashed L.. **Gingival mesenchymal stem cells as an alternative source to bone marrow mesenchymal stem cells in regeneration of bone defects:**. *Tissue Cell* (2020) **63** 101325. DOI: 10.1016/j.tice.2019.101325
2. Boere J., Malda J., van de Lest C., van Weeren P. R., Wauben M.. **Extracellular vesicles in joint disease and therapy**. (2018) **9** 2575. DOI: 10.3389/fimmu.2018.02575
3. Cheng P., Cao T., Zhao X., Lu W., Miao S., Ning F.. **Nidogen1-enriched extracellular vesicles accelerate angiogenesis and bone regeneration by targeting Myosin-10 to regulate endothelial cell adhesion**. *Bioact. Mater* (2022) **12** 185-197. DOI: 10.1016/j.bioactmat.2021.10.021
4. Deschaseaux F., Sensébé L., Heymann D.. **Mechanisms of bone repair and regeneration**. *Trends Mol. Med.* (2009) **15** 417-429. DOI: 10.1016/j.molmed.2009.07.002
5. Dimitriou R., Jones E., McGonagle D., Giannoudis P. V.. **Bone regeneration: Current concepts and future directions**. *BMC Med.* (2011) **9** 66. DOI: 10.1186/1741-7015-9-66
6. Eggenhofer E., Luk F., Dahlke M. H., Hoogduijn M. J.. **The life and fate of mesenchymal stem cells**. *Front. Immunol.* (2014) **5** 148. DOI: 10.3389/fimmu.2014.00148
7. Fawzy El-Sayed K. M., Dörfer C. E.. **Gingival mesenchymal stem/progenitor cells: A unique tissue engineering gem**. *Stem Cells Int.* (2016) **2016** 1-16. DOI: 10.1155/2016/7154327
8. Hasani-Sadrabadi M. M., Sarrion P., Pouraghaei S., Chau Y., Ansari S., Li S.. **An engineered cell-laden adhesive hydrogel promotes craniofacial bone tissue regeneration in rats**. *Sci. Transl. Med.* (2020) **12** eaay6853. DOI: 10.1126/scitranslmed.aay6853
9. Hu Y., Wang Y., Chen T., Hao Z., Cai L., Li J.. **Exosome: Function and application in inflammatory bone diseases**. *Oxid. Med. Cell Longev.* (2021) **2021** 1-17. DOI: 10.1155/2021/632491210.1155/2021/6324912
10. Jiang S., Xu L.. **Exosomes from gingival mesenchymal stem cells enhance migration and osteogenic differentiation of pre-osteoblasts**. *Pharmazie* (2020) **75** 576-580. DOI: 10.1691/ph.2020.0652
11. Kenswil K., Pisterzi P., Sánchez-Duffhues G., van Dijk C., Lolli A., Knuth C.. **Endothelium-derived stromal cells contribute to hematopoietic bone marrow niche formation**. *Cell Stem Cell* (2021) **28** 653-670.e11. DOI: 10.1016/j.stem.2021.01.006
12. Kumar B. P., Venkatesh V., Kumar K. A., Yadav B. Y., Mohan S. R.. **Mandibular reconstruction: Overview**. *J. Maxillofac. Oral Surg.* (2016) **15** 425-441. DOI: 10.1007/s12663-015-0766-5
13. Liang X., Ding Y., Zhang Y., Tse H. F., Lian Q.. **Paracrine mechanisms of mesenchymal stem cell-based therapy: Current status and perspectives**. *Cell Transpl.* (2014) **23** 1045-1059. DOI: 10.3727/096368913X667709
14. Liu T., Hu W., Zou X., Xu J., He S., Chang L.. **Human periodontal ligament stem cell-derived exosomes promote bone regeneration by altering MicroRNA profiles**. *Stem Cells Int.* (2020) **2020** 1-13. DOI: 10.1155/2020/8852307
15. Moghadasi S., Elveny M., Rahman H. S., Suksatan W., Jalil A. T., Abdelbasset W. K.. **A paradigm shift in cell-free approach: The emerging role of MSCs-derived exosomes in regenerative medicine**. *J. Transl. Med.* (2021) **19** 302. DOI: 10.1186/s12967-021-02980-6
16. Najar M., Melki R., Khalife F., Lagneaux L., Bouhtit F., Moussa Agha D.. **Therapeutic mesenchymal stem/stromal cells: Value, challenges and optimization**. *Front. Cell Dev. Biol.* (2021) **9** 716853. DOI: 10.3389/fcell.2021.716853
17. Nguyen M. K., Jeon O., Dang P. N., Huynh C. T., Varghai D., Riazi H.. **RNA interfering molecule delivery from**. *Acta Biomater.* (2018) **75** 105-114. DOI: 10.1016/j.actbio.2018.06.007
18. Qin Y., Sun R., Wu C., Wang L., Zhang C.. **Exosome: A novel approach to stimulate bone regeneration through regulation of osteogenesis and angiogenesis**. *Int. J. Mol. Sci.* (2016) **17** 712. DOI: 10.3390/ijms17050712
19. Rani S., Ryan A. E., Griffin M. D., Ritter T.. **Mesenchymal stem cell-derived extracellular vesicles: Toward cell-free therapeutic applications**. *Mol. Ther.* (2015) **23** 812-823. DOI: 10.1038/mt.2015.44
20. Reis M., Mavin E., Nicholson L., Green K., Dickinson A. M., Wang X. N.. **Mesenchymal stromal cell-derived extracellular vesicles attenuate dendritic cell maturation and function**. *Front. Immunol.* (2018) **9** 2538. DOI: 10.3389/fimmu.2018.02538
21. Shi Q., Qian Z., Liu D., Sun J., Wang X., Liu H.. **GMSC-derived exosomes combined with a chitosan/silk hydrogel sponge accelerates wound healing in a diabetic rat skin defect model**. *Front. Physiol.* (2017) **8** 904. DOI: 10.3389/fphys.2017.00904
22. Sun Y., Wan B., Wang R., Zhang B., Luo P., Wang D.. **Mechanical stimulation on mesenchymal stem cells and surrounding microenvironments in bone regeneration: Regulations and applications**. *Front. Cell Dev. Biol.* (2022) **10** 808303. DOI: 10.3389/fcell.2022.808303
23. Swanson W. B., Zhang Z., Xiu K., Gong T., Eberle M., Wang Z.. **Scaffolds with controlled release of pro-mineralization exosomes to promote craniofacial bone healing without cell transplantation**. *Acta Biomater.* (2020) **118** 215-232. DOI: 10.1016/j.actbio.2020.09.052
24. Tarasov V. V., Svistunov A. A., Chubarev V. N., Dostdar S. A., Sokolov A. V., Brzecka A.. **Extracellular vesicles in cancer nanomedicine**. *Semin. Cancer Biol.* (2021) **69** 212-225. DOI: 10.1016/j.semcancer.2019.08.017
25. Théry C., Amigorena S., Raposo G., Clayton A.. **Isolation and characterization of exosomes from cell culture supernatants and biological fluids**. *Curr. Protoc. Cell Biol.* (2006) **30** Unit 3.22. DOI: 10.1002/0471143030.cb0322s30
26. Tomar G. B., Srivastava R. K., Gupta N., Barhanpurkar A. P., Pote S. T., Jhaveri H. M.. **Human gingiva-derived mesenchymal stem cells are superior to bone marrow-derived mesenchymal stem cells for cell therapy in regenerative medicine**. *Biochem. Biophys. Res. Commun.* (2010) **393** 377-383. DOI: 10.1016/j.bbrc.2010.01.126
27. Trajkovic K., Hsu C., Chiantia S., Rajendran L., Wenzel D., Wieland F.. **Ceramide triggers budding of exosome vesicles into multivesicular endosomes**. *Science* (2008) **319** 1244-1247. DOI: 10.1126/science.1153124
28. Valtanen R. S., Yang Y. P., Gurtner G. C., Maloney W. J., Lowenberg D. W.. **Synthetic and Bone tissue engineering graft substitutes: What is the future**. *Injury* (2021) **52** S72-S77. DOI: 10.1016/j.injury.2020.07.040
29. Wang L., Wang J., Zhou X., Sun J., Zhu B., Duan C.. **A new self-healing hydrogel containing hucMSC-derived exosomes promotes bone regeneration**. *Front. Bioeng. Biotechnol.* (2020) **8** 564731. DOI: 10.3389/fbioe.2020.564731
30. Williams T., Salmanian G., Burns M., Maldonado V., Smith E., Porter R. M.. **Versatility of mesenchymal stem cell-derived extracellular vesicles in tissue repair and regenerative applications**. *Biochimie* (2022) **207** 33-48. DOI: 10.1016/j.biochi.2022.11.011
31. Wu J., Chen L., Wang R., Song Z., Shen Z., Zhao Y.. **Exosomes secreted by stem cells from human exfoliated deciduous teeth promote alveolar bone defect repair through the regulation of angiogenesis and osteogenesis**. *ACS Biomater. Sci. Eng.* (2019) **5** 3561-3571. DOI: 10.1021/acsbiomaterials.9b00607
32. Xu Q. C., Wang Z. G., Ji Q. X., Yu X. B., Xu X. Y., Yuan C. Q.. **Systemically transplanted human gingiva-derived mesenchymal stem cells contributing to bone tissue regeneration**. *Int. J. Clin. Exp. Pathol.* (2014) **7** 4922-4929. PMID: 25197363
33. Xu X., Chen C., Akiyama K., Chai Y., Le A., Wang Z.. **Gingivae contain neural-crest- and mesoderm-derived mesenchymal stem cells**. *J. Dent. Res.* (2013) **92** 825-832. DOI: 10.1177/0022034513497961
34. Yeo R. W., Lai R. C., Zhang B., Tan S. S., Yin Y., Teh B. J.. **Mesenchymal stem cell: An efficient mass producer of exosomes for drug delivery**. *Adv. Drug Deliv. Rev.* (2013) **65** 336-341. DOI: 10.1016/j.addr.2012.07.001
35. Zhang L., Jiao G., Ren S., Zhang X., Li C., Wu W.. **Exosomes from bone marrow mesenchymal stem cells enhance fracture healing through the promotion of osteogenesis and angiogenesis in a rat model of nonunion**. *Stem Cell Res. Ther.* (2020) **11** 38. DOI: 10.1186/s13287-020-1562-9
36. Zhang Y., Hao Z., Wang P., Xia Y., Wu J., Xia D.. **Exosomes from human umbilical cord mesenchymal stem cells enhance fracture healing through HIF-1α-mediated promotion of angiogenesis in a rat model of stabilized fracture**. *Cell Prolif.* (2019) **52** e12570. DOI: 10.1111/cpr.12570
37. Zuo Y., Li Q., Xiong Q., Li J., Tang C., Zhang Y.. **Naringin release from a nano-hydroxyapatite/collagen scaffold promotes osteogenesis and bone tissue reconstruction**. *Polym. (Basel)* (2022) **14** 3260. DOI: 10.3390/polym14163260
|
---
title: Relationship between preoperative malnutrition, frailty, sarcopenia, body composition,
and anthropometry in elderly patients undergoing major pancreatic and biliary surgery
authors:
- Lijuan Wang
- Pengxue Li
- Yifu Hu
- Bo Cheng
- Lili Ding
- Lei Li
- Jinghai Song
- Junmin Wei
- Jingyong Xu
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9989266
doi: 10.3389/fnut.2023.1135854
license: CC BY 4.0
---
# Relationship between preoperative malnutrition, frailty, sarcopenia, body composition, and anthropometry in elderly patients undergoing major pancreatic and biliary surgery
## Abstract
### Objective
To analyze the correlation between preoperative nutritional status, frailty, sarcopenia, body composition, and anthropometry in geriatric inpatients undergoing major pancreatic and biliary surgery.
### Methods
This is a cross-sectional study of the database from December 2020 to September 2022 in the department of hepatopancreatobiliary surgery, Beijing Hospital. Basal data, anthropometry, and body composition were recorded. NRS 2002, GLIM, FFP 2001, and AWGS 2019 criteria were performed. The incidence, overlap, and correlation of malnutrition, frailty, sarcopenia, and other nutrition-related variables were investigated. Group comparisons were implemented by stratification of age and malignancy. The present study adhered to the STROBE guidelines for cross-sectional study.
### Results
A total of 140 consecutive cases were included. The prevalence of nutritional risk, malnutrition, frailty, and sarcopenia was 70.0, 67.1, 20.7, and $36.4\%$, respectively. The overlaps of malnutrition with sarcopenia, malnutrition with frailty, and sarcopenia with frailty were 36.4, 19.3, and $15.0\%$. There is a positive correlation between every two of the four diagnostic tools, and all six p-values were below 0.002. Albumin, prealbumin, CC, GS, 6MTW, ASMI, and FFMI showed a significantly negative correlation with the diagnoses of the four tools. Participants with frailty or sarcopenia were significantly more likely to suffer from malnutrition than their control groups with a 5.037 and 3.267 times higher risk, respectively (for frailty, $95\%$ CI: 1.715–14.794, $$p \leq 0.003$$ and for sarcopenia, $95\%$ CI: 2.151–4.963, $p \leq 0.001$). Summarizing from stratification analysis, most body composition and function variables were worsen in the ≥70 years group than in the younger group, and malignant patients tended to experience more intake reduction and weight loss than the benign group, which affected the nutrition diagnosis.
### Conclusion
Elderly inpatients undergoing major pancreatic and biliary surgery possessed high prevalence and overlap rates of malnutrition, frailty, and sarcopenia. Body composition and function deteriorated obviously with aging.
## 1. Introduction
The geriatric syndrome refers to a range of multifactorial health conditions representing the accumulation of multiple system impairments in older adults. Malnutrition, frailty, and sarcopenia are three common geriatric syndromes, which can substantially lead to poor outcomes, such as disability, dysfunction, falls, and perioperative complications, and thereby increase the length of hospital stay (LOS) and the cost of hospitalization, and result in long-term care or even mortality (1–5).
Malnutrition or undernutrition refers to deficiencies in nutritional intake resulting in altered body composition, and approximately $\frac{1}{3}$ of Chinese geriatric inpatients experience malnutrition [6]. In the department of hepatopancreatobiliary surgery, the prevalence of nutritional risk and malnutrition are as high as 69.7 and $56.6\%$ in our former study [7]. Frailty is characterized by a cumulative decline in the physiological capacity of multiple organ systems and increased vulnerability to endogenous and exogenous stressors, with an estimated prevalence ranging from 18.8 to $41.9\%$ in geriatric surgical patients and from 10.4 to $37.0\%$ in general surgical patients [8, 9]. Sarcopenia is an age-related syndrome characterized by progressive and generalized loss of skeletal muscle mass and strength, which accounts for $17.4\%$ of Chinese community-dwelling and hospitalized elderly [10]. In pancreatic surgery, the prevalence is $38.8\%$ determined by the total psoas area index in CT scan [7].
Frailty, sarcopenia, and malnutrition have independent diagnostic criteria, but share many components, such as weight loss, muscle mass, or strength loss, and often coexist or overlap in elderly inpatients [11]. In a recent systematic review, it was concluded that about half of the hospitalized older patients suffer from 2 or perhaps 3 of these debilitating conditions, and standardized screening for these conditions is highly controversial to guide nutritional and physical interventions [12]. Due to the significant influence of these three clinical problems on outcomes, respectively, it is important to understand the current situation and provide basal data for further cohort study. So our study aims to investigate the prevalence and overlap of these conditions in the elderly who are going to receive major pancreatic and biliary surgery.
## 2.1. Participants
This study is a cross-sectional study analyzing the daily database of the Department of hepatopancreatobiliary surgery, Beijing Hospital. From December 2020 to September 2022, 205 consecutive patients undergoing major pancreatic and biliary surgery were screened, and then, 140 elderly patients were recruited in this study.
The inclusion criteria of this study are as follows: [1] age ≥60 years old, which is the age cut-off of older adults defined by the Nation Health Commission of China [13]; [2] major pancreatic and biliary surgery, containing pancreatectomy (Whipple procedure, distal pancreatectomy, and local pancreatectomy), bile-enteral bypass due to malignant obstructive, and bile duct exploration; [3] voluntary enrollment and signed informed consent. Exclusion criteria contain [1] emergency operation; [2] cancer patients who underwent adjuvant therapy before operation; [3] severe disability or dementia, inability to cooperate with frailty and sarcopenia assessment or effective communication; [4] refusal of informed consent. The Ethics Committee of Beijing Hospital approved the study protocol and written informed consents were obtained from all participants. ( Approval letter No. 2020BJYYEC-218-01). The present study adhered to the STROBE guidelines for cross-sectional study. Figure 1 shows the flowchart of this study.
**FIGURE 1:** *Flowchart of the study.*
## 2.2. Basal characteristics, anthropometry, and body composition
The basal data include sex, age, height, weight, body mass index (BMI), co-morbidities, and serum examination (complete blood count, liver function, renal function, albumin, glucose, et al.). According to the standard of the guidelines for prevention and control of overweight and obesity in Chinese adults, a BMI < 18.5 kg/m2 was defined as underweight, 18.5 kg/m2 ≤ BMI < 24 kg/m2 was normal weight, 24 kg/m2 ≤ BMI < 28 kg/m2 was considered overweight, and BMI ≥ 28 kg/m2 was considered obesity [14].
A diet survey was conducted after admission. We recorded the change of diet before and after the diagnosis of the original disease, and calculated the contents composition, containing protein, carbohydrate, fat, and total energy.
Anthropometry was done 1 to 2 days after admission, including calf circumference (CC) and grip strength (GS), both of which, we used the average value of the left and right sides. To assess the functional status, 15-foot and 6-meter timed walk speed (6MTW) was conducted to get the walking speed. Bioelectrical impedance analysis (BIA) was applied with the InBody 720 bioimpedance body composition analyzer (Biospace Co., Ltd., Korea). Appendicular skeletal muscle mass index (ASMI) was calculated, which was the sum of the lean muscle mass of the upper and lower extremities adjusted with height. Also, the fat-free mass index (FFMI) was recorded. Visceral fat area (VFA), waist-hip ratio (WHR), and body fat percentage (BTP) were included to reflect fat metabolism.
## 2.3. Nutritional risk screening
We used Nutritional Risk Screening 2002 (NRS 2002) for nutritional screening for each patient within 24 h after admission, which was recommended by the European Society of Parenteral Enteral Nutrition (ESPEN) [15]. NRS2002 contains three aspects: nutritional impairment: weight loss, intake reduction, and lower BMI (score 0–3), the severity of disease (score 0–3), and age [(score 0–1) (< 70 years: 0 scores and ≥ 70 years: 1 score)]. Scores for the final screening take into account all these three sections range from 0 to 7 and classify patients into one of two nutritional risk stages (or groups): at low nutritional risk group (NRS 2002 score < 3), and (moderate/high) risk of malnutrition group (NRS 2002 score ≥ 3). In pancreatic surgery, an NRS2002 score of more and equal to 5 was considered at high nutritional risk with remarkable clinical meaning [16].
## 2.4. Malnutrition diagnosis and grading
The Global Leadership Initiative on Malnutrition (GLIM) criteria were implemented for malnutrition diagnosis and grading among patients with nutritional risk determined by NRS2002 [17]. The framework of GLIM criteria includes three phenotypic criteria and two etiologic criteria, and the detailed items and cut-off values could be determined and modified in different centers and populations [18]. In this study, we used GLIM criteria in a traditional way with the original criteria. Phenotypic criteria include [1] unintentional weight loss (WT): WT > $5\%$ within the past 6 months, or WT > $10\%$ beyond 6 months; [2] low BMI: BMI < 18.5 kg/m2 if age < 70 years, BMI < 20 kg/m2 if age ≥ 70 years; [3] reduced muscle mass: in our study, we used AMMI and FFMI assessed by BIA. AMMI < 7 kg/m2 or FFMI < 17 kg/m2 in men were considered patients with reduced muscle mass, and AMMI < 5.7 kg/m2 or FFMI < 15 kg/m2 in women were considered positive. Etiologic criteria include: [1] Reduced food intake: ≤$50\%$ of needs from 1 to 2 weeks, or any reduction for >2 weeks; [2] Disease burden or inflammation: in this study, most of the patients were suffering from malignancies and the co-morbidities were also taken into account. If at least one criterion was fulfilled in each section, malnutrition can be diagnosed.
The grading of malnutrition also followed the GLIM criteria. Unintentional weight loss (WT) > $10\%$ within the past 6 months or WT > $20\%$ beyond 6 months or low BMI (BMI < 17.0 kg/m2 if age < 70 years or BMI < 17.8 kg/m2 if age ≥ 70 years) or severe muscle deficit were defined as severe malnutrition. 5–$10\%$ Unintentional weight loss (WT) within the past 6 months or 10–$20\%$ WT beyond 6 months or low BMI (17.0 ≤ BMI < 20.0 kg/m2 if age < 70 years or 17.8 ≤ BMI < 22.0 kg/m2 if age ≥ 70 years) or Mild-to-Moderate muscle deficit were the grading criteria for moderate malnutrition.
## 2.5. Diagnosis of sarcopenia
In this study, we used the criteria for sarcopenia diagnosis recommended by the Asian Working Group for Sarcopenia (AWGS) [19]. For patients in acute to chronic health care or clinical research settings, a two-step protocol was used: finding cases and diagnosis. In the first step, we tended to use objective criterion, so calf circumference (CC) (<34 cm in male, <33 cm in female) was facilitated to find cases at risk of sarcopenia, based on which, in the second step, sarcopenia can be diagnosed as follows: [1] Muscle strength: men with grip strength (GS) < 28 kg, women with GS < 18 kg; [2] Physical performance: 6-meter walk < 1 m/s; [3] AMMI: men with AMMI < 7 kg/m2, women with AMMI < 5.7 kg/m2. The result containing low ASMI and low muscle strength or low physical performance was sarcopenia, and the result containing all three criteria was severe sarcopenia.
## 2.6. Diagnosis of frailty
The Fried Frailty Phenotype (FFP) is a recommended assessment tool for frailty in geriatric patients by Chinese expert group consensus [20]. FFP criteria include five physical items: [1] Shrinking: Unintentional weight loss: ≥$5\%$ of body weight in the prior year; [2] Poor endurance and energy: self-reported exhaustion; [3] Weakness: poorer GS; [4] Slowness: lower walk speed; [5] Low physical activity. Patients who fulfilled none of these five criteria were classified as the non-frailty group, who fulfilled 1 or 2 criteria were classified as the pre-frailty group, and who fulfilled ≥3 criteria were considered as the frailty group. The thresholds of GS and gait speed were referred to the AWGS criteria. Table 1 shows the comparison of all the above diagnostic tools we used in this study.
**TABLE 1**
| Unnamed: 0 | NRS2002 | GLIM | AWGS 2019 | FFP 2001 |
| --- | --- | --- | --- | --- |
| Weight loss | > 5% within past 3 months(1 score) >5% within past 2 months(2 scores) >5% within past 1 months(3 scores) | WT > 5% within past 6 months WT > 10% beyond 6 months | Unintentional weight loss | Unintentional weight loss: of 10 pounds in prior year, or of 5% of body weight in prior year at follow-up |
| BMI | BMI < 18.5 kg/m2 (3 scores) | BMI < 18.5 kg/m2 if age < 70 years BMI < 20 kg/m2 if age ≥ 70 years | – | – |
| Muscle mass reduction | – | Reduced by validated body composition measuring techniques: FFMI by DXA or BIA, CT or MRI Anthropometric measures: calf circumferences Functional assessment: hand-grip strength | ASMI by DXA or BIA CC, SARC-F, or SARC-CalF Handgrip strength | Grip strength |
| Intake reduction | 50–75% of normal requirement in preceding week (1 score) 25–50% of normal requirement in preceding week (2 scores) 0–25% of normal requirement in preceding week (3 scores) | ≤ 50% of needs from 1 to 2 weeks any reduction for >2 weeks | – | – |
| Disease and inflammation burden | Patient with chronic disease, admitted to hospital due to complications. Protein requirement can be covered by oral diet or supplements (1 score) Patient confined to bed due to illness. Protein requirement can be covered by artificial feeding (2 scores) Patient in intensive care. Protein requirement is increased and cannot be covered even by artificial feeding (3 scores) | Acute disease/injury-related: Severe inflammation and mild-to-moderate inflammation Chronic disease-related: Chronic or recurrent mild-to-moderate inflammation Transient inflammation of a mild degree is excluded | Malnutrition Chronic conditions | —— |
| Physical performance | – | – | 6-meter walk time 5-time chair stand test SPPB | 15-feet walk time Low physical activity level |
| Other | – | – | Depressive mood Cognitive impairment Repeated falls | Self-reported exhaustion |
## 2.7. Statistical analysis
The sample size was calculated by PASS software 11.0 (NCSS LLC., Kaysville, UT, USA). The confidence level was set at 0.8. According to our former study, the prevalence of malnutrition was $56.6\%$ and we set the proportion at $60\%$ [7]. The tolerance error was set at $10\%$, so the two-sided confidential interval width was 0.12. The final sample size was 125. All statistical analysis was performed by IBM SPSS Statistics for Windows, version 27.0 (IBMCorp., Armonk, NY, USA). Measurement data that correspond to normal distribution were presented as mean with standard deviation (SD) and analyzed by Student’s t-test. Measurement data that did not correspond to normal distribution were presented as median with interquartile range (IQR) and analyzed by Mann–Whitney U test. Categorical data were presented as counts and percentages, and compared by chi-square (χ2) test. Correlations were analyzed by Spearman’s correlation coefficient analysis according to the classification of variables. Multivariate analysis was performed by binary logistic regression to identify potential associated factors of malnutrition. A p-value < 0.05 were declared as statistically significant. All figures including flowchart, overlap bubble chart, and correlation heatmap were designed and drawn by Microsoft Office (Version 2016), and the regression analysis figure was drawn by GraphPad Prism version 7.0.0 for Windows (GraphPad Software, San Diego, CA, USA).
## 3.1. Basal characteristics and nutrition status
A total of 140 participants were included with a mean age of 70.0 ± 7.3 years. $58.6\%$ ($\frac{82}{140}$) were male. $75\%$ ($\frac{105}{140}$) of cases were malignancies, of which, 71 cases were pancreatic duct adenocarcinoma. The details of the history and blood test at admission are shown in Table 2.
**TABLE 2**
| Variables | Basal data, n = 140 |
| --- | --- |
| Sociodemographics | Sociodemographics |
| Age, mean (SD), years | 70.0 (7.3) |
| 60–69, n (%) | 76 (54.3) |
| ≥ 70, n (%) | 64 (45.7) |
| Male sex, n (%) | 82 (58.6) |
| Admission diagnosis | Admission diagnosis |
| Malignancies, n (%) | 105 (75.0) |
| Pancreatic cancer, n (%) | 71 (50.7) |
| Bile duct cancer, n (%) | 18 (12.9) |
| Duodenal cancer, n (%) | 3 (2.1) |
| Ampulla cancer, n (%) | 9 (6.4) |
| Other malignancies, n (%) | 4 (2.9) |
| Benign diseases, n (%) | 35 (25.0) |
| History | History |
| Diabetes, n (%) | 47 (33.6) |
| Chronic obstructive pulmonary disease, n (%) | 4 (2.9) |
| Cardia-cerebral disease, n (%) | 88 (62.9) |
| Smoking, n (%) | 53 (37.9) |
| Drinking, n (%) | 32 (22.9) |
| Blood test at admission | Blood test at admission |
| White blood cell, mean (SD) × 109/L | 6.0 (1.7) |
| Hemoglobin. mean (SD) g/L | 122.6 (17.3) |
| Platelet, mean (SD) × 109/L | 210.3 (61.7) |
| Fasting glucose, mean (SD) g/L | 6.6 (2.9) |
| Total protein, mean (SD) g/L | 64.3 (5.3) |
| Albumin, mean (SD) g/L | 37.3 (4.3) |
| Pre-albumin, mean (SD) g/L | 18.0 (7.7) |
| Alanine aminotransferase, median (IQR) U/L | 21.0 (89.5)2 |
| Creatine, mean (SD) μmoI/L | 64.7 (16.5) |
| Triglyceride, mean (SD) mmol/L | 1.5 (1.0) |
| Total cholesterol, mean (SD) mmol/L | 4.5 (1.3) |
Table 3 shows the data for nutrition assessment. The mean BMI was 23.5 ± 3.6 kg/m2. 83 cases ($59.3\%$) experienced weight loss to varying degrees, in which, 66 cases exceeded $5\%$. According to NRS 2002, $70.0\%$ ($$n = 98$$) of cases were at risk of nutrition. 94 cases ($67.1\%$) were malnutrition and 49 cases ($35.0\%$) were severe malnutrition according to GLIM criteria. Based on FFP criteria, $53.6\%$ ($$n = 75$$) participants were pre-frailty, and $20.7\%$ ($$n = 29$$) were frailty. According to the AWGS 2019 consensus, at the step of finding cases, $52.9\%$ ($$n = 74$$) cases were at risk of sarcopenia determined by reduced calf circumference, among which, in the second step, $36.4\%$ ($$n = 31$$) participants were diagnosed as sarcopenia, $24.2\%$ ($$n = 34$$) fulfilled the criteria of severe sarcopenia. We also reported every diagnostic criterion in each tool in Table 3 to reflect the composition of every diagnosis.
**TABLE 3**
| Variables | Nutrition data, n = 140 |
| --- | --- |
| Nutrition assessment | Nutrition assessment |
| BMI, mean (SD) kg/m2 | 23.5 (3.6) |
| BMI < 18.5 kg/m2, n (%) | 8 (5.7) |
| 18.5 ≤ BMI < 24 kg/m2, n (%) | 74 (52.9) |
| 24 ≤ BMI < 28 kg/m2, n (%) | 46 (32.9) |
| BMI ≥ 28 kg/m2, n (%) | 12 (8.6) |
| Weight at admission, mean (SD) kg | 63.4 (11.0) |
| Weight loss, n (%) | 83 (59.3) |
| Weight loss ≥ 5%, n (%) | 66 (47.1) |
| Weight loss amount at admission, median (IQR) kg | 3.0 (6.4) |
| Weight loss percentage at admission, median (IQR)% | 4.3 (9.1) |
| NRS 2002–nutritional risk (score ≥ 3), n (%) | 98 (70.0) |
| Low risk (score 3–4), n (%) | 44 (31.4) |
| High risk (score 5–7), n (%) | 54 (38.6) |
| Nutrition impairment | Nutrition impairment |
| Weight loss score 0/1/2/3, n (%) | 58 (41.4)/12 (8.6)/10 (7.1)/60 (42.9) |
| Intake reduction score 0/1/2/3, n (%) | 71 (50.7)/23 (16.4)/36 (25.7)/10 (7.1) |
| BMI score 0/3, n (%) | 132 (94.3)/8 (5.7) |
| Disease burden score 0/1/2/3, n (%) | 0 (0.0)/0 (0.0)/140 (100.0)/0 (0.0) |
| Age score 0/1, n (%) | 76 (54.3)/64 (45.7) |
| GLIM – malnutrition, n (%) | 94 (67.1) |
| Moderate-mid Malnutrition, n (%) | 45 (32.1) |
| Severe malnutrition, n (%) | 49 (35.0) |
| Phenotype criteria | Phenotype criteria |
| Weight loss meeting diagnostic criteria, n (%) | 66 (47.1) |
| BMI meeting diagnostic criteria, n (%) | 11 (7.9) |
| FFMI meeting diagnostic criteria, n (%) | 62 (44.3) |
| Etiologic criteria | Etiologic criteria |
| Intake reduction meeting diagnostic criteria, n (%) | 67 (47.9) |
| Disease burden meeting diagnostic criteria, n (%) | 140 (100.0) |
| FFP 2001 | FFP 2001 |
| Pre-frailty, n (%) | 75 (53.6) |
| Frailty, n (%) | 29 (20.7) |
| Unintentional weight loss, n (%) | 73 (52.1) |
| Self-reported exhaustion, n (%) | 30 (21.4) |
| Low grip strengthen, n (%) | 68 (48.6) |
| Low walking speed, n (%) | 108 (77.1) [0.2pt] |
| Low physical activity, n (%) | 14 (10.0) |
| AWGS 2019 | AWGS 2019 |
| At risk of sarcopenia, n (%) | 74 (52.9) |
| Low calf circumference, n (%) | 74 (52.9) |
| Sarcopenia, n (%) | 31 (36.4) |
| Severe sarcopenia, n (%) | 31 (22.1) |
| Low grip strengthen, n (%) | 68 (48.6) |
| Low walking speed, n (%) | 108 (77.1) |
| Low ASMI, n (%) | 53 (37.9) |
| Diet survey | Diet survey |
| Energy reduction after diagnosis, median (IQR) kcal/d | 76 (640.5) |
| Energy reduction percentage after diagnosis, median (IQR)% | 5.3 (34.8) |
| Protein reduction after diagnosis, median (IQR) g/day | 1.0 (25.2) |
| Protein reduction percentage after diagnosis, median (IQR)% | 1.6 (46.8) |
| Fat reduction after diagnosis, median (IQR) g/day | 0.0 (24.2) |
| Fat reduction after percentage after diagnosis, median (IQR)% | 0.0 (49.3) |
| Anthropometry | Anthropometry |
| Calf circumference, mean (SD) cm | 33.2 (3.5) |
| < 34 cm in male, n (%), N = 82 | 42 (51.2) |
| < 33 cm in female, n (%), N = 58 | 32 (55.2) |
| Grip strength, mean (SD) kg | 24.9 (8.5) |
| < 28 kg in male, n (%), N = 82 | 50 (61.0) |
| < 18 kg in female, n (%), N = 58 | 18 (31.0) |
| 6-meter timed walking speed, mean (SD) m/s | 0.85 (0.21) |
| < 1 m/s, n (%) | 108 (77.1) |
| Body composition | Body composition |
| Harris-Benedict equation, mean (SD) kcal/d | 1279.8 (173.5) |
| ASMI, mean (SD) kg/m2 | 6.7 (0.9) |
| < 7.0 kg/m2 in male, n (%), N = 82 | 39 (47.6) |
| < 5.7 kg/m2 in female, n (%), N = 58 | 14 (24.1) |
| FFMI, mean (SD) kg/m2 | 16.5 (1.6) |
| < 17.0 kg/m2 in male, n (%), N = 82 | 41 (50.0) |
| < 15.0 kg/m2 in female, n (%), N = 58 | 21 (36.2) |
| Body fat percentage, mean (SD)% | 27.8 (8.3) |
| Waist hip ratio, mean (SD) | 0.92 (0.07) |
| Visceral fat area, mean (SD) cm2 | 83.6 (27.6) |
Figure 2 displays the overlap of these three conditions, besides which, 21 ($15.0\%$) cases fulfilled all three criteria, and 26 ($18.6\%$) cases were considered normal by all three criteria. Furthermore, we did a stratification analysis between the patients who fulfilled three criteria and healthy patients. The results showed that there were more patients with malignant diseases ($54.3\%$ vs. $16.7\%$, $$p \leq 0$$,024) and older age ($70.8\%$ vs. $17.4\%$, $p \leq 0.001$) in the fulfill-three-criteria group.
**FIGURE 2:** *Overlaps between malnutrition, frailty, and sarcopenia. (A) Overlap between malnutrition and sarcopenia. (B) Overlap between malnutrition and frailty. (C) Overlap between sarcopenia and frailty.*
A diet survey showed 71 cases had a decline in the intake of total energy, protein, and fat before and after the diagnosis of the disease. In Table 3, the amounts and percentages of reduction of energy, protein, and fat were shown in detail. Results of anthropometry and body composition analysis are also recorded in Table 3 and more men than women suffered from a decline in muscle mass and muscle-related function variables such as CC, GS, ASMI, and FFMI.
## 3.2.1. Stratified by age
The patients were stratified by age and divided into the <70 years group and ≥70 years group. In Table 4, results show that the prevalence of nutritional risk, severe malnutrition, frailty, and sarcopenia were all significantly higher in the older group. Though there was no difference in the change in daily diet and weight, an obvious decline was found in both body composition and function. The changes in body composition appeared not only on the protein-related blood tests like hemoglobin, total protein, albumin, and pre-albumin, but also on the reduction of muscle mass (CC, ASMI, and FFMI), which logically affected the muscle function (e.g., GS and 6MTW).
**TABLE 4**
| Variables, n = 140 | < 70 years | ≥ 70 years | P | Malignant | Benign | P.1 |
| --- | --- | --- | --- | --- | --- | --- |
| N | 76 | 64 | | 105 | 35 | |
| Nutrition related blood test | Nutrition related blood test | Nutrition related blood test | Nutrition related blood test | Nutrition related blood test | Nutrition related blood test | Nutrition related blood test |
| Hemoglobin, mean (SD) g/L | 126.1 (16.1) | 118.3 (18.1) | 0.001 | 122.2 (18.0) | 124.0 (15.6) | 0.629 |
| Fasting glucose, mean (SD) g/L | 6.9 (3.5) | 6.2 (1.9) | 0.140 | 6.8 (3.2) | 5.7 (1.2) | 0.059 |
| Total protein, mean (SD) g/L | 65.44 (4.9) | 62.8 (5.5) | 0.004 | 64.1 (5.5) | 64.7 (4.7) | 0.542 |
| Albumin, mean (SD) g/L | 38.6 (3.9) | 35.5 (4.3) | <0.001 | 37.0 (4.2) | 38.0 (4.6) | 0.208 |
| Pre-albumin, mean (SD) g/L | 21.1 (8.0) | 14.3 (5.4) | <0.001 | 16.6 (6.0) | 21.3 (10.2) | 0.008 |
| Triglyceride, mean (SD) mmol/L | 1.3 (0.7) | 1.7 (1.3) | 0.050 | 1.6 (1.0) | 1.3 (1.0) | 0.157 |
| Total cholesterol, mean (SD) mmol/L | 4.4 (1.1) | 4.5 (1.6) | 0.826 | 4.6 (1.4) | 4.1 (0.8) | 0.108 |
| Nutrition assessment | Nutrition assessment | Nutrition assessment | Nutrition assessment | Nutrition assessment | Nutrition assessment | Nutrition assessment |
| BMI, mean (SD) kg/m2 | 24.1 (3.7) | 22.7 (3.4) | 0.016 | 23.2 (3.6) | 24.3 (3.6) | 0.109 |
| Weight loss, n (%) | 42 (55.3) | 41 (64.1) | 0.291 | 70 (66.7) | 13 (37.1) | 0.002 |
| Weight loss ≥ 5%, n (%) | 35 (46.1) | 31 (48.4) | 0.778 | 56 (53.3) | 10 (28.6) | 0.011 |
| Weight loss amount at admission, median (IQR) kg | 3.0 (7.0) | 3.0 (6.0) | 0.709 | 3.8 (7.0) | 0.0 (4.0) | 0.017 |
| Weight loss percentage at admission, median (IQR)% | 4.0 (9.2) | 4.7 (9.2) | 0.521 | 5.3 (9.5) | 0.0 (5.8) | 0.012 |
| NRS 2002–nutritional risk, n (%) | 42 (55.3) | 56 (87.5) | <0.001 | 85 (81.0) | 13 (37.1) | <0.001 |
| High risk, n (%) | 20 (26.3) | 34 (53.1) | 0.001 | 47 (44.8) | 7 (20.0) | 0.009 |
| GLIM–malnutrition, n (%) | 47 (61.8) | 47 (73.4) | 0.146 | 75 (71.4) | 19 (54.3) | 0.061 |
| Severe malnutrition, n (%) | 18 (23.7) | 31 (48.4) | 0.002 | 41 (39.0) | 8 (22.9) | 0.082 |
| FFP 2001 | | | <0.001 | | | 0.029 |
| Pre-frailty, n (%) | 44 (57.9) | 31 (48.4) | | 57 (54.3) | 18 (51.4) | |
| Frailty, n (%) | 6 (7.9) | 23 (35.9) | | 26 (24.8) | 3 (8.6) | |
| Self-reported exhaustion, n (%) | 19 (29.7) | 11 (14.5) | 0.029 | 27 (25.7) | 3 (8.6) | 0.032 |
| Low physical activity, n (%) | 12 (18.8) | 2 (2.6) | 0.002 | 11 (10.5) | 3 (8.6) | 0.745 |
| AWGS 2019 | AWGS 2019 | AWGS 2019 | AWGS 2019 | AWGS 2019 | AWGS 2019 | AWGS 2019 |
| At risk of sarcopenia, n (%) | 31 (40.8) | 43 (67.2) | 0.002 | 55 (52.4) | 19 (54.3) | 0.845 |
| Sarcopenia, n (%) | 18 (23.7) | 33 (51.6) | 0.001 | 40 (38.1) | 11 (31.4) | 0.478 |
| Severe sarcopenia, n (%) | 8 (10.5) | 26 (40.6) | <0.001 | 30 (28.6) | 4 (11.4) | 0.041 |
| Diet survey | Diet survey | Diet survey | Diet survey | Diet survey | Diet survey | Diet survey |
| Energy reduction after diagnosis, median (IQR) kcal/day | 0.0 (523.0) | 263.0 (817.0) | 0.116 | 299.0 (817.0) | 0.0 (0.0) | <0.001 |
| Energy reduction percentage after diagnosis, median (IQR)% | 0.0 (27.9) | 14.5 (49.3) | 0.074 | 18.7 (47.6) | 0.0 (0.0) | <0.001 |
| Protein reduction after diagnosis, median (IQR) g/d | 0.0 (22.0) | 10.0 (32.0) | 0.105 | 13.0 (34.0) | 0.0 (0.0) | <0.001 |
| Protein reduction percentage after diagnosis, median (IQR)% | 0.0 (32.8) | 17.2 (54.1) | 0.085 | 19.7 (51.9) | 0.0 (0.0) | <0.001 |
| Fat reduction after diagnosis, median (IQR) g/d | 0.0 (12.0) | 4.0 (33.0) | 0.087 | 4.0 (28.0) | 0.0 (0.0) | 0.001 |
| Fat reduction after percentage after diagnosis, median (IQR)% | 0.0 (23.1) | 8.9 (66.7) | 0.070 | 8.9 (55.1) | 0.0 (0.0) | 0.001 |
| Anthropometry | Anthropometry | Anthropometry | Anthropometry | Anthropometry | Anthropometry | Anthropometry |
| Calf circumference, mean (SD) cm | 34.3 (3.6) | 32.0 (2.9) | <0.001 | 33.1 (3.6) | 33.6 (3.3) | 0.513 |
| Grip strength, mean (SD) kg | 28.0 (8.8) | 21.3 (6.5) | <0.001 | 24.6 (8.5) | 25.9 (8.5) | 0.426 |
| 6-meter timed walk speed, <1 m/s, n (%) | 0.92 (0.17) | 0.74 (0.21) | <0.001 | 79 (75.2) | 29 (82.9) | 0.353 |
| Body composition | Body composition | Body composition | Body composition | Body composition | Body composition | Body composition |
| Harris-Benedict equation, mean (SD) kcal/day | 1346.5 (167.4) | 1199.0 (145.0) | <0.001 | 1273.1 (168.8) | 1300.7 (188.3) | 0.429 |
| ASMI, mean (SD) kg/m2 | 6.9 (0.9) | 6.4 (0.8) | 0.001 | 6.6 (0.9) | 6.8 (0.9) | 0.238 |
| FFMI, mean (SD) kg/m2 | 16.9 (1.7) | 16.1 (1.3) | 0.003 | 16.4 (1.7) | 16.7 (1.5) | 0.172 |
| Body fat percentage, mean (SD)% | 28.0 (8.4) | 27.5 (8.2) | 0.742 | 27.6 (8.6) | 28.2 (7.6) | 0.710 |
| Waist hip ratio, mean (SD) | 0.92 (0.06) | 0.91 (0.07) | 0.572 | 0.92 (0.06) | 0.91 (0.07) | 0.690 |
| Visceral fat area, mean (SD) cm2 | 84.4 (28.7) | 82.8 (26.5) | 0.739 | 83.0 (27.8) | 85.6 (27.5) | 0.636 |
## 3.2.2. Stratified by malignant and benign disease
When the patients were divided into malignant and benign groups, the results were completely different from the results in the groups stratified by age as above (Table 4). The main differences between malignant and benign groups were the changes in daily diet and weight, and no difference was found in body composition and function. Only prealbumin showed a significant decline in malignant disease in the benign disease group (16.6 ± 6.0 vs. 21.3 ± 10.2, $$p \leq 008$$), which was a sensitive variable to indicate recent nutrition changes. Both nutritional risk and severe nutritional risk were significantly higher in the malignant group, but no difference was found in the prevalence of GLIM-defined malnutrition. The malignant group possessed higher rates of frailty and sarcopenia.
## 3.3. Correlation between variables
Figure 3 is a heatmap showing the correlation between variables. From the perspective of overall color composition, the blue area shows a negative correlation between variables of serum test, body composition, and anthropometry with the four diagnostic tools. The red area could be divided into two parts: the part on the upper left of the blue area shows a positive correlation between every two of the four tools, and all six p-values were below 0.05; the part on the lower right of the blue area shows a positive correlation between variables of serum test, body composition, and anthropometry. In serum tests, hemoglobin, albumin, and prealbumin show a significant correlation with body composition and anthropometry. Body composition (BMI, ASMI, and FFMI) are well correlated with anthropometry (CC, GS, and 6MTW) with statistical significance.
**FIGURE 3:** *Correlation heatmap. The correlation coefficient numbers (r) are presented in the triangle, red for positive association, and blue for negative association. Darker colors indicate stronger associations (larger coefficient numbers). The significance levels for coefficients are presented below the r. CC, calf circumference; GS, grip strength; 6MWS, 6-meter walking speed; BMI, body mass index; ASMI, appendicular skeleton muscle index; FFMI, fat free mass index; HGB, hemoglobin; TP, total protein; ALB, albumin; PreALB, prealbumin.*
The first column shows the correlation between age and other variables. The prevalence of nutritional risk, malnutrition, frailty, and sarcopenia were all positively correlated with age with significance, and all body composition and anthropometry variables were negatively correlated with age. Meanwhile, in the second column, nutritional risk, malnutrition, and frailty were proved to positively correlate with malignant diseases with statistical significance. However, a significant negative correlation was only found in prealbumin in all body composition and anthropometry variables (r = −0.248, $$p \leq 0.018$$).
## 3.4. Multivariate logistic regression analysis
After adjustment for age, malignant diseases, frailty, and sarcopenia with malnutrition as the dependent variables, multivariate logistic regression analysis showed that participants with frailty or sarcopenia were significantly more likely to suffer from malnutrition than their control groups with a 5.037 and 3.267 times higher risk, respectively (for frailty, $95\%$ CI: 1.715–14.794, $$p \leq 0.003$$ and for sarcopenia, $95\%$ CI: 2.151–4.963, $p \leq 0.001$) (Figure 4).
**FIGURE 4:** *Multivariate regression analysis for malnutrition.*
## 4. Discussion
With increasing global aging problems, aging-related debilitating disorders, so-called geriatric syndrome, are becoming the hotspots of geriatric research. All frailty, sarcopenia, and malnutrition are components of geriatric syndrome and are closely interrelated and interdependent. Surgical patients suffer from a double attack of disease and aging. In our study, the prevalence of nutritional risk, and malnutrition are 70.0 and $67.1\%$, respectively, which are higher than in elderly patients with other gastrointestinal diseases [21]. The prevalence of frailty is $20.7\%$, which is familiar to former articles and at a relatively higher proportion [9]. The prevalence of sarcopenia is $36.4\%$, which is nearly the same as our data collected in pancreatic surgery diagnosed by a CT scan [7].
Diagnosis of malnutrition, frailty, and sarcopenia depend on the diagnostic tools, different tools might lead to different prevalence [22]. In a recent systemic review, 18 tools of frailty diagnosis were reported, in which, FFP was the most commonly used one. Meanwhile, EWGSOP (European Working Group on Sarcopenia in Older People) criteria was the most commonly used in all and AWGS was the most commonly implemented in Asia. And for malnutrition, about thirteen tools were mentioned, besides which, BMI only and BMI with albumin were considered to be diagnostic criteria in three articles [12]. However, it is difficult to avoid bias when calculating overlap data between different tools and it is still a controversy in this field. So in our study, we chose FFP, WGS, NRS2002, and GLIM to avoid selection bias.
Table 1 displays the comparison of FFP, WGS, NRS2002, and GLIM, which reflect the commonality and individuality of the tools. Weight loss was the only criterion shared by the four tools, which is not only for nutrition assessment but also a sensitive precursor for tumor diagnosis, especially for pancreatic cancer [23]. Besides weight loss, NRS2002 and GLIM contain age, BMI, intake reduction, and assessment of disease (inflammation burden), which are relatively more comprehensive to assess the nutrition status. But no muscle assessment was contained in NRS2002, and GLIM contains the evaluation of muscle, but with a large range of measuring techniques. AWGS2019 and FFP2001 criteria are based on muscle assessment. The overlaps between frailty or sarcopenia and malnutrition were 19.3 and $36.4\%$, which are higher than was reported before [12]. AWFS2019 criteria focus on both muscle mass and muscle function to diagnose sarcopenia, meanwhile, FFP2001 criteria only focus on muscle function and function-related symptoms like cognitive and behavioral impairment. So the overlap of sarcopenia and frailty ($15.0\%$) was not as large as expected. Moreover, in this study, 21 ($15.0\%$) cases fulfilled all three criteria, and 26 ($18.6\%$) cases were considered normal or no risk by all three criteria. Therefore, due to different clinical values and low overlap rates, these diagnostic criteria would still coexist, and more comprehensive tools may be created and validated in the future.
According to the guidelines of ESPEN, malnutrition, sarcopenia, and frailty were treated as parallel definitions [24]. The links between malnutrition and sarcopenia or frailty have already been explored in several cross-sectional studies, especially in older patients with chronic disease (25–27). In our study, in surgical patients, the correlations between these conditions were proved to be statistically significant, which were shown in Figure 3. However, it is difficult to judge the causal relationship between any two of these three statuses. Theoretically speaking, in this population, original surgical diseases lead to intake reduction and weight loss, which affected nutrients digestion and absorption, and then gave rise to the change in body composition, especially the change of muscle mass, sequentially muscular dysfunction. Nutrition risk or malnutrition seems to be the initiating factor [28, 29]. Our results indicated that sarcopenia and frailty seemed to be risk factors for malnutrition, however, longitudinal studies are needed.
From the perspective of body composition in the criteria, BMI may not be sensitive enough to be used in the surgical population, only $5.7\%$ of cases were lower than 18.5 kg/m2, and nearly $40\%$ of patients suffered from overweight and obesity, in which, nearly $20\%$ were sarcopenia [7]. Even in pancreatic surgery, higher BMI was treated as a risk factor for a fatty pancreas and postoperative pancreatic fistula rather than a nutrition parameter [30]. FFMI and ASMI, which reflect the real change in muscle mass, had become the focus of diagnostic criteria. In this study, FFMI accounted for $44.3\%$ of phenotype criteria in GLIM, second only to weight loss. And it was proved to be well consistent in GLIM-defined malnutrition in this study and other reports [31]. ASMI was the sum of the lean muscle mass of the upper and lower extremities adjusted with height, which was reported to be used in GLIM and well related to sarcopenia and frailty [32, 33]. So with the improvement of availability and simplification of the examination method, ASMI and FFMI will become more popular in clinical practice.
In this study, we did stratification analyses by age and malignancy. Like reports from other centers, it was no doubt that nutritional status became worse with aging and malignant diagnosis [34]. However, from Table 4, by comparing the data from the two stratifications, an interesting phenomenon was notable. In the age stratification, the significant differences were mainly in the changes in body composition and its related parameters, including basic metabolic rate (Harris-Benedict equation), muscle mass (CC, ASMI, and FFMI), muscle function (GS and 6MWS), BMI, and serum test (hemoglobin, total protein, albumin, and prealbumin), all of which reflected the long-term changes of the body due to aging rather than disease. Meanwhile, in the stratification of malignant diseases, the significant differences between malignant and benign groups were only weight loss and intake reduction, which were short-term changes due to the pathophysiologic characteristics of cancer, but no change in body composition existed. In the serum test, only pre-albumin was significantly lower in the malignant group, which has been proposed to be a useful nutritional biomarker due to its shorter half-life than albumin and correlated with different nutritional markers and higher mortality risk [35]. So when referring to preoperative therapy, for patients with advanced age, we must pay attention to both nutrition support and function exercise, to improve long-term nutrition and function problems caused by aging, and increase preoperative reservation, which was defined as “prehabilitation” and needed a relatively longer period [36]. And for cancer patients with nutritional risk or malnutrition, we should commit to increasing intake and improving nutrition status by different support routes even for a short period [16, 37]. Prealbumin might be a biomarker to monitor the effectiveness of preoperative nutrition support but needs further study.
As we know, few researchers have been reported to study the effect of two of the three conditions in older adults, but the three conditions are rarely studied together [38]. This study is the first one to study the overlap of these three conditions in pancreatic and biliary surgery. Since this is a cross-sectional study, we tried our best to follow the STROBE statement, but there must be some limitations that are difficult to avoid. First, we used a relatively lower confidence level (0.8) and prevalence of malnutrition to determine the sample size, which may underestimate the sample size, especially when we did the stratification analysis; second, a cross-sectional study could not verify the causal relationship. Although we used multivariate regression analysis, the aim was to explore the possible relevance and provide the necessary direction for future cohort studies. Third, the sample population is elderly, so whether the tangent point value can represent other populations should be a deeper study field and need more work.
## 5. Conclusion
Elderly inpatients undergoing major pancreatic and biliary surgery had a high prevalence and overlap rates of malnutrition, frailty, and sarcopenia. Body composition and function deteriorated obviously with aging. Patients with malignant diseases often suffer from short-term nutrition changes like intake reduction. Simple and effective biomarker needs to be explored and validated. Rational preoperative prehabilitation containing nutrition support and exercise should be considered in this population to reduce postoperative complications and mortality.
## Data availability statement
The data supporting this study’s findings are available from the corresponding author upon reasonable request. Requests to access the datasets should be directed to JX, [email protected].
## Ethics statement
The studies involving human participants were reviewed and approved by the Ethics Committee of Beijing Hospital. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
JX and JW: conception, design, and administrative support. JX, LL, JS, and JW: provision of study materials or patients. JX, YH, PL, LW, LD, and BC: collection, assembly of data, data analysis, and interpretation. LW, YH, and JX: manuscript writing. LW, PL, YH, BC, LD, LL, JS, JW, and JX: final approval of 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. Bieniek J, Wilczyński K, Szewieczek J. **Fried frailty phenotype assessment components as applied to geriatric inpatients.**. (2016) **11** 453-9. DOI: 10.2147/CIA.S101369
2. Walston J, Buta B, Xue Q. **Frailty screening and interventions: considerations for clinical practice.**. (2018) **34** 25-38. DOI: 10.1016/j.cger.2017.09.004
3. Morley J, Vellas B, van Kan G, Anker S, Bauer J, Bernabei R. **Frailty consensus: a call to action.**. (2013) **14** 392-7. PMID: 23764209
4. Chen R, Zhao W, Zhang X, Liang H, Song N, Liu Z. **Relationship between frailty and long-term care needs in Chinese community-dwelling older adults: a cross-sectional study.**. (2022) **12**. DOI: 10.1136/bmjopen-2021-051801
5. Ruiz A, Buitrago G, Rodríguez N, Gómez G, Sulo S, Gómez C. **Clinical and economic outcomes associated with malnutrition in hospitalized patients.**. (2019) **38** 1310-6. PMID: 29891224
6. Liang Y, Zhang Y, Li Y, Chen Y, Xu J, Liu M. **Identification of frailty and its risk factors in elderly hospitalized patients from different wards: a cross-sectional study in China.**. (2019) **14** 2249-59. DOI: 10.2147/CIA.S225149
7. Xu J, Li C, Zhang H, Liu Y, Wei J. **Total psoas area index is valuable to assess sarcopenia, sarcopenic overweight/obesity and predict outcomes in patients undergoing open pancreatoduodenectomy.**. (2020) **13** 761-70. DOI: 10.2147/RMHP.S257677
8. Darvall J, Gregorevic K, Story D, Hubbard R, Lim W. **Frailty indexes in perioperative and critical care: a systematic review.**. (2018) **79** 88-96. DOI: 10.1016/j.archger.2018.08.006
9. Hewitt J, Long S, Carter B, Bach S, McCarthy K, Clegg A. **The prevalence of frailty and its association with clinical outcomes in general surgery: a systematic review and meta-analysis.**. (2018) **47** 793-800. DOI: 10.1093/ageing/afy110
10. Ren X, Zhang X, He Q, Du L, Chen K, Chen S. **Prevalence of sarcopenia in Chinese community-dwelling elderly: a systematic review.**. (2022) **22**. DOI: 10.1186/s12889-022-13909-z
11. Jeejeebhoy K. **Malnutrition, fatigue, frailty, vulnerability, sarcopenia and cachexia: overlap of clinical features.**. (2012) **15** 213-9. DOI: 10.1097/MCO.0b013e328352694f
12. Ligthart-Melis G, Luiking Y, Kakourou A, Cederholm T, Maier A, de van der Schueren M. **Frailty, sarcopenia, and malnutrition frequently (Co-)occur in hospitalized older adults: a systematic review and meta-analysis.**. (2020) **21** 1216-28. DOI: 10.1016/j.jamda.2020.03.006
13. 13.Nation Health Commission of the People’s Republic of China.
Standard for Healthy Chinese Older Adults (WS/T 802—2022).
Beijing: Nation Health Commission of the People’s Republic of China (2022).. (2022)
14. Chen C, Lu F. **The guidelines for prevention and control of overweight and obesity in Chinese adults.**. (2004) **17** 1-36
15. Kondrup J, Allison S, Elia M, Vellas B, Plauth M. **ESPEN guidelines for nutrition screening 2002.**. (2003) **22** 415-21. PMID: 12880610
16. Xu J, Tian X, Song J, Chen J, Yang Y, Wei J. **Preoperative nutrition support may reduce the prevalence of postoperative pancreatic fistula after open pancreaticoduodenectomy in patients with high nutritional risk determined by NRS2002.**. (2021) **2021**. DOI: 10.1155/2021/6691966
17. Cederholm T, Jensen G, Correia M, Gonzalez M, Fukushima R, Higashiguchi T. **GLIM criteria for the diagnosis of malnutrition – a consensus report from the global clinical nutrition community.**. (2019) **10** 207-17. PMID: 30920778
18. Correia M, Tappenden K, Malone A, Prado C, Evans D, Sauer A. **Utilization and validation of the Global Leadership Initiative on Malnutrition (GLIM): a scoping review.**. (2022) **41** 687-97. DOI: 10.1016/j.clnu.2022.01.018
19. Chen L, Woo J, Assantachai P, Auyeung T, Chou M, Iijima K. **Asian working group for sarcopenia: 2019 consensus update on sarcopenia diagnosis and treatment.**. (2020) **21** 300-7.e2. DOI: 10.1016/j.jamda.2019.12.012
20. Fried L, Tangen C, Walston J, Newman A, Hirsch C, Gottdiener J. **Frailty in older adults: evidence for a phenotype.**. (2001) **56** M146-56. PMID: 11253156
21. Liu C, Lu Z, Li Z, Xu J, Cui H, Zhu M. **Influence of malnutrition according to the GLIM criteria on the clinical outcomes of hospitalized patients with cancer.**. (2021) **8**. DOI: 10.3389/fnut.2021.774636
22. Xu J, Zhu M, Zhang H, Li L, Tang P, Chen W. **A cross-sectional study of GLIM-defined malnutrition based on new validated calf circumference cut-off values and different screening tools in hospitalised patients over 70 years old.**. (2020) **24** 832-8. DOI: 10.1007/s12603-020-1386-4
23. Nicholson B, Thompson M, Hobbs F, Nguyen M, McLellan J, Green B. **Measured weight loss as a precursor to cancer diagnosis: retrospective cohort analysis of 43 302 primary care patients.**. (2022) **13** 2492-503. DOI: 10.1002/jcsm.13051
24. Cederholm T, Barazzoni R, Austin P, Ballmer P, Biolo G, Bischoff S. **ESPEN guidelines on definitions and terminology of clinical nutrition.**. (2017) **36** 49-64. PMID: 27642056
25. Vettoretti S, Caldiroli L, Armelloni S, Ferrari C, Cesari M, Messa P. **Sarcopenia is associated with malnutrition but not with systemic inflammation in older persons with advanced CKD.**. (2019) **11**. DOI: 10.3390/nu11061378
26. Marco E, Sánchez-Rodríguez D, Dávalos-Yerovi V, Duran X, Pascual E, Muniesa J. **Malnutrition according to ESPEN consensus predicts hospitalizations and long-term mortality in rehabilitation patients with stable chronic obstructive pulmonary disease.**. (2019) **38** 2180-6. DOI: 10.1016/j.clnu.2018.09.014
27. Laur C, McNicholl T, Valaitis R, Keller H. **Malnutrition or frailty? Overlap and evidence gaps in the diagnosis and treatment of frailty and malnutrition.**. (2017) **42** 449-58. PMID: 28322060
28. Scott D, Jones G. **Impact of nutrition on muscle mass, strength, and performance in older adults.**. (2014) **25** 791-2. PMID: 24057482
29. Beaudart C, Sanchez-Rodriguez D, Locquet M, Reginster J, Lengelé L, Bruyère O. **Malnutrition as a strong predictor of the onset of sarcopenia.**. (2019) **11**. DOI: 10.3390/nu11122883
30. Zhou L, Xiao W, Li C, Gao Y, Gong W, Lu G. **Impact of fatty pancreas on postoperative pancreatic fistulae: a meta-analysis.**. (2021) **11**. DOI: 10.3389/fonc.2021.622282
31. Gao X, Liu H, Zhang L, Tian H, Zhou D, Li G. **The application value of preoperative fat-free mass index within global leadership Initiative on Malnutrition-defined malnutrition criteria for postoperative outcomes in patients with esophagogastric cancer.**. (2022) **102**. DOI: 10.1016/j.nut.2022.111748
32. Liu H, Gao X, Zhang L, Zhang Y, Wang X. **Application of the GLIM criteria in patients with intestinal insufficiency and intestinal failure at nutritional risk on admission.**. (2022) **76** 1003-9. DOI: 10.1038/s41430-022-01084-8
33. Gillis C, Fenton T, Gramlich L, Sajobi T, Culos-Reed S, Bousquet-Dion G. **Older frail prehabilitated patients who cannot attain a 400 m 6-min walking distance before colorectal surgery suffer more postoperative complications.**. (2021) **47** 874-81. DOI: 10.1016/j.ejso.2020.09.041
34. Wang J, Zhuang Q, Tan S, Xu J, Zhang Y, Yan M. **Loss of body weight and skeletal muscle negatively affect postoperative outcomes after major abdominal surgery in geriatric patients with cancer.**. (2022) **106**. DOI: 10.1016/j.nut.2022.111907
35. Bretscher C, Buergin M, Gurzeler G, Kägi-Braun N, Gressies C, Tribolet P. **The association between prealbumin, all-cause mortality and response to nutritional treatment in patients at nutritional risk. Secondary analysis of a randomized-controlled trial.**. (2023). DOI: 10.1002/jpen.2470
36. Baimas-George M, Watson M, Elhage S, Parala-Metz A, Vrochides D, Davis B. **Prehabilitation in frail surgical patients: a systematic review.**. (2020) **44** 3668-78. DOI: 10.1007/s00268-020-05658-0
37. Weimann A, Wobith M. **ESPEN Guidelines on Clinical nutrition in surgery – special issues to be revisited.**. (2022). DOI: 10.1016/j.ejso.2022.10.002
38. AlMohaisen N, Gittins M, Todd C, Burden S. **What is the overlap between malnutrition, frailty and sarcopenia in the older population? Study protocol for cross-sectional study using UK Biobank.**. (2022) **17**. DOI: 10.1371/journal.pone.0278371
|
---
title: A progression analysis of motor features in Parkinson's disease based on the
mapper algorithm
authors:
- Ling-Yan Ma
- Tao Feng
- Chengzhang He
- Mujing Li
- Kang Ren
- Junwu Tu
journal: Frontiers in Aging Neuroscience
year: 2023
pmcid: PMC9989279
doi: 10.3389/fnagi.2023.1047017
license: CC BY 4.0
---
# A progression analysis of motor features in Parkinson's disease based on the mapper algorithm
## Abstract
### Background
Parkinson's disease (PD) is a neurodegenerative disease with a broad spectrum of motor and non-motor symptoms. The great heterogeneity of clinical symptoms, biomarkers, and neuroimaging and lack of reliable progression markers present a significant challenge in predicting disease progression and prognoses.
### Methods
We propose a new approach to disease progression analysis based on the mapper algorithm, a tool from topological data analysis. In this paper, we apply this method to the data from the Parkinson's Progression Markers Initiative (PPMI). We then construct a Markov chain on the mapper output graphs.
### Results
The resulting progression model yields a quantitative comparison of patients' disease progression under different usage of medications. We also obtain an algorithm to predict patients' UPDRS III scores.
### Conclusions
By using mapper algorithm and routinely gathered clinical assessments, we developed a new dynamic models to predict the following year's motor progression in the early stage of PD. The use of this model can predict motor evaluations at the individual level, assisting clinicians to adjust intervention strategy for each patient and identifying at-risk patients for future disease-modifying therapy clinical trials.
## 1. Introduction
Parkinson's disease is a neurodegenerative disease with a broad spectrum of motor symptoms including bradykinesia, rigidity, resting tremor, and postural and gait impairments (Selikhova et al., 2009). In the clinical course of PD, both linear (Gottipati et al., 2017; Holden et al., 2018) and non-linear progression (Vu et al., 2012; Reinoso et al., 2015) have been reported in the advancement of motor and non-motor symptoms. The substantial heterogeneity in the presentation of clinical phenotypes, genetics, pathology, and disease progression (Foltynie et al., 2002; Selikhova et al., 2009; Ma et al., 2015) and lack of reliable progression markers of neurodegeneration present a major challenge for prediction of progression and accurate prognoses, hampering advances in PD trials, and the clinical routine determining therapeutic efficacy. In an era of increasing focus on individualized management and disease-modifying therapies, there is a need to develop useful tools to predict each patient's motor progression with high accuracy.
The current literature on PD progression consists largely of associative analyses and a few prognostic models. The prognostic models include logistic regression and Bayesian classification models to predict cognitive impairment (Schrag et al., 2017; Hogue et al., 2018; Gramotnev et al., 2019), machine-learning, random survival forests to predict time to initiation of symptomatic treatment (Simuni et al., 2016) and disease progression (Latourelle et al., 2017; Severson et al., 2021). Besides, partial least squares path modeling (PLS-PM), combined with MRI biomarkers, were used to predict progression subtypes and cognitive impairment in prodromal PD (Pyatigorskaya et al., 2021; Rahayel et al., 2021). Based on the Parkinson's Progression Markers Initiative (PPMI) database, we previously built five regression models to predict PD motor progression represented by the coming year's Unified Parkinson's Disease Rating Scale (MDS-UPDRS) Part III score, finding adjusted R2 values of three different categories of regression model, linear, Bayesian, and ensemble, all reached 0.75 (Ma et al., 2020).
In this study, we propose a new approach to disease progression analysis based on topological data analysis (TDA), or the mapper algorithm to be precise. The mapper algorithm was introduced in Singh et al. [ 2007] by Singh-Memoli-Carlsson as a way of capturing topological/geometric informations of a point cloud dataset possibly in a high dimensional Euclidean space. Roughly speaking, it may be viewed as an algorithm to compute a given dataset's geometric “shape" by certain combinatorial object which, in the simplest form, may be a graph or a polyhedron. In the case of analyzing patients' data, the method has been successfully implemented in a variety of circumstances (see for example Nicolau et al., 2011; Li et al., 2015; Rossi-deVries et al., 2018; Dagliati et al., 2020).
It is always difficult to predict PD because of great heterogeneity, including subtypes, markers, and various scales. Only by combining clinical presentation and mathematical methods, selecting appropriate parameters and applying appropriate methods can the accuracy of prediction model be improved. Based on PPMI data and our previous predicting models, we aim to improve our multiple dynamic prediction model via mapper algorithm in this study. Similarly, general information and classical clinical scales, which are routinely and easily performed in clinical activities, were used to predict motor progression, displayed in the form of the MDS–UPDRS Part III score. These inexpensive and easily readily available clinical data can facilitate widespread implementation of this cost-efficient predictive model in real world applications.
## 2.1. Feature selection and data pre-processing
The data were obtained from the PPMI database. The PPMI is an international, multicenter, prospective study designed to discover, and validate biomarkers of disease progression in newly diagnosed PD participants (National Clinical Trials identifier NCT01141023). Each PPMI recruitment site received approval from an institutional review board or ethics committee on human experimentation before study initiation. Written informed consent for research was obtained from all individuals participating in the study. The PPMI database was accessed on December 16, 2022, to obtain data from 943, 379, 324, 256, 268 visits for five consecutive years. For up-to-date information on the study, please visit www.ppmi-info.org.
Since the mapper algorithm is a way of computing the “shape" of a given data set in ℝN, if the dimension N is too large while the data set is relatively small, the shape would only be a collection of sparse points. Thus, our first step uses a topological method to reduce the number of features introduced in Kraft [2016]. The idea behind this feature selection method is that we could eliminate a feature if it does not cause a big change in the underlying topology (calculated using persistent homology) of the data sets. We refer to the article (Kraft, 2016) for more details.
In our case, for the feature selection, we first consider the following listed 29 features mostly used in Ma et al. [ 2020]. We have added a feature “symptom” which is given by the sum symptom1, symptom2, symptom3, and symptom4, with All these variables are binary such that it is 0 if No symptom or unknown; 1 if Symptom present at diagnosis.
*In* general, the features we consider are inexpensive and easily readily available clinical data. Each of the coordinates is normalized to [0, 1]. In the coordinate given by the UPDRS III score, we also performed a clamping at 0.7.1 These features are listed as follows: updrs3, age, NP1APAT, scopa, YEAR, NP1FATG, moca, symptom, NP1ANXS, gds, PD_MED_USE, symptom2, NP1HALL, ageonset, NP1COG, NP1DPRS, rem, ess, symptom1, DOMSIDE, “PATNO,” gen, symptom4, fampd_new, NP1DDS, duration, td_ pig, quip, symptom3.
In the above we have ordered the features according to their Pearson's correlation coefficients with the UPDRS III score. An important point to note is that, excluding the UPDRS III score itself, the maximal of these Pearson's correlation coefficients is 0.27, which shows that their correlation with the UPDRS III score is in general highly non-linear. This is an ideal context to use our topological data analysis (TDA) method as it is a tool developped to handle non-linear correlations.
Then, we use the persistent homology to reduce the number of features (Kraft, 2016). In our case, Figure 1 illustrates the persistent homology, when passing from the first eight features to seven features, has a big difference. This tells us we should stop eliminating features. The remaining 8 selected features are listed as in Table 1.
**Figure 1:** *The barcodes and rips diagrams both illustrate the persistent homology (Lum et al., 2013) of a given dataset. The two pictures compare the persistent homology in the case of seven features with the case of eight features. Observe that the right hand side has considerably more persistence (in both black and red markings) compared with the left hand side.* TABLE_PLACEHOLDER:Table 1 From the PPMI data, we select these features for each patient's data to form a point cloud SPPMI ⊂ ℝ8, of size |SPPMI| = 2, 389, consisting of 481 distinguished patients.
## 2.2. The mapper algorithm
The mapper algorithm introduced by Singh-Memoli-Carlsson (Singh et al., 2007) is a method to analyze high dimensional data based on ideas from topology—a branch of mathematics to study complex shapes of geometric objects.
Roughly speaking, the mapper algorithm consists of several steps as illustrated in Figure 2.
**Figure 2:** *Illustration of the mapper algorithm in the case of a point cloud S in ℝ2, and with the filter function given by the horizontal projection. The outcome of this algorithm is the bottom graph. (A) Original point cloud, (B) Coloring by filter value, (C) Binning by filter value, and (D) Clustering and network construction.*
## 2.3. Construction of Markov chains
We shall apply the mapper algorithm to the point cloud SPPMI from the previous subsection. Recall that SPPMI ⊂ ℝ8 is the sample space of patients' data extracted from the raw PPMI data. Using the mapper algorithm, assume that we have obtained m clusters C1, …, Cm so that SPPMI=C1∪⋯∪Cm. Note that these clusters can possibly intersect with each other.
Let P ⊂ SPPMI × SPPMI be a subset. We proceed to use P to obtain a Markov chain on the set of clusters C1, …, Cm. For a pair of data (x, y) ∈ P, if x ∈ Ci and y ∈ Cj, we consider it as an arrow from the cluster Ci to Cj. This yields a multi-graph (possibly with multiple edges between vertices) whose vertices are the clusters C1, …, Cm. Then we use informations of this multi-graph to obtain a Markov matrix. More precisely, for each pair of indices (i, j) with 1 ≤ i, j ≤ m, we define
## 2.3.1. Computing expected growth
For each 1 ≤ j ≤ m, denote by Ej:=1|Cj|·∑y∈Cjupdrs3(y) the expected value of the UPDRS III score of the cluster Cj.
The expected growth of a patient's UPDRS III of a fixed PD medication type i is computed as follows.
## 2.4. Prediction models
As a second application, we use the Markov chains obtained in the previous paragraph to build a prediction model for a patient's UPDRS III score in the next year. This is done in several steps: The first step (a) needs more explanation, and is realized as follows. Fix a positive integer μ > 0, and a positive real number σ > 0. We find the first μ nearest point a1,… aμ∈SPPMI to the given point x. Then use the equation to determine a constant c. For each 1 ≤ k ≤ μ, the point ak may belong to several clusters. Denote its multiplicity by *At this* point, it is tempted to set the initial probability vector by formula
However, observe that already in the definition of MP (see Equation 1), it is possible that a cluster *Ci is* not the source of any arrows, i.e., *In this* case, it is not possible to use such type of clusters to make predictions for the next year's data. Thus, we set the initial probability at such a cluster by zero, and rescale the resulting vector by a constant to obtain the desired initial probability vector. Explicitly, we set the initial probability vector v = (v1, v2, …, vm) by *In this* paper, we shall fix the parameters to be μ = 14 and σ = 0.0378.
## 3.1. Mapper outputs
We apply the Kepler mapper program 1.4.1 (van Veen et al., 2019a,b) to the point cloud set SPPMI with a 2-dimensional filter function given by two coordinate projections in the direction of “age" and “updrs3." The output graph is shown in Figure 3.
**Figure 3:** *The pictures are out-puts of the mapper algorithm with parameters given by n1 = 20, n2 = 40, p = 0.05 and n1 = 20, n2 = 40, p = 0.3, respectively.*
As expected by the formation of the mapper algorithm, larger percentage of overlaps naturally leads to more non-empty intersections between clusters, and hence the graph on the right appears to have more edges than the left one.
In the two dimensional mapper algorithm, there are three parameters to choose: There exists no general method to determine appropriate parameters in the mapper algorithm. In the next section, we shall use the mapper output to construct a prediction model for the UPDRS III scores of patients. We then use the precision value of the resulting prediction model to evaluate and thus optimize the parameters.
## 3.2. Markov chains
From the PPMI data, there are eight different types of patients according to their usage of PD medications, as shown in Table 2.
**Table 2**
| Type index | Medication |
| --- | --- |
| 0 | Unmedicated |
| 1 | Levodopa |
| 2 | Dopamine agonist |
| 3 | Other |
| 4 | Levodopa + other |
| 5 | Levodopa + dopamine agonist |
| 6 | Dopamine agonist + other |
| 7 | Levodopa + dopamine agonist + other |
Denote by Pi⊂SPPMI×SPPMI,0≤i≤7 the subset consisting of pairings (x, y) such that the data x and y are two consecutive years' data from the same patient (i.e., a progression by 1 year), and that the patient's usage of PD medication is of type i in the above table. For $i = 0$ and $i = 1$ we have depicted the corresponding two Markov chains in Figure 4 (with mapper parameters set to be n1 = 20, n2 = 40, $$p \leq 0.05$$).
**Figure 4:** *The two figures illustrate two Markov chains associated with medication type 0 and 1, respectively. Its nodes are derived from the outputs of the mapper algorithm.*
## 3.3. PD medication type analysis
As a first application of the Markov chains MPi obtained from the previous paragraph. We use it to compute the expected growth of a patient's UPDRS III score according to the patient's PD medication type. The computed results are shown in Table 3.
**Table 3**
| PD medication type index | Expected growth of UPDRS III score |
| --- | --- |
| 0.0 | 2.17 |
| 1.0 | 2.24 |
| 2.0 | 2.51 |
| 3.0 | 3.37 |
| 4.0 | 2.19 |
| 5.0 | 0.3 |
| 6.0 | 0.88 |
| 7.0 | 1.85 |
The expected growth of PD patients with a particular type of medication certainly may depend on the particular choice of medication to begin with. Thus, it makes sense to perform an un-biased comparison with what happens if the medication type i≠0 group of patients were not given any medication. To do this, consider the following probability distribution (p1, …, pm) on the set of clusters defined by We can calculate the expected growth viewed as un-medicated patients under same distribution using the Markov chain MjkP0: The difference between Δi′ and the actual expected growth Δi would measure the benefit of the i-th type medication to reduce the growth of patients' UPDRS III scores. Calculations demonstrate solid medication effects in the cases of type 4, 5, and 6, as shown in Table 4. Observe that patients in medication type 5 and 6 have relatively small expected growth of UPDRS score in Table 3. The un-biased analysis gives at least a partial explanation for this: for these two groups of patients medication effects are rather significant.
**Table 4**
| Medication type i | Δi | Δi′ | Δi′-Δi |
| --- | --- | --- | --- |
| 4.0 | 2.19 | 2.5 | 0.31 |
| 5.0 | 0.3 | 1.3 | 1.0 |
| 6.0 | 0.88 | 2.29 | 1.41 |
## 3.4. Statistics of the prediction models
To test the validity of our prediction model described above, for each PD medication type index 0 ≤ i ≤ 7, we perform a statistical study of its accuracy as follows.
Table 5 shows the statistics of our prediction models in each PD medication type. The R2 score, MAE, MSE and Max Error are well-known statistical measures. We explain the last column “hit percentage." In the evaluation of UPDRS III score (a total of 132 points), medical experiences usually permits a variation of ±5 points. In our data set SPPMI, the difference between maximal score and the minimal score is 80. Since we have normalized this score to [0, 1], a variation of ±5 absolute points would corresponds to ±0.0625 after normalization. The “hit-in percentage” is the percentage of the prediction score p(x0) “hit-in” the interval [y0−0.0625, y0+0.0625] since we regard such a prediction as being a successful one.
**Table 5**
| Medication | R2 score | MAE | MSE | Max error | Hit in percentage % |
| --- | --- | --- | --- | --- | --- |
| 0 | 0.67 | –0.0121 | 0.00597 | 0.216 | 62.6 |
| 1 | 0.726 | –0.00807 | 0.00607 | 0.222 | 82.7 |
| 2 | 0.966 | –0.00542 | 0.000396 | 0.0127 | 97.0 |
| 3 | 0.872 | –0.000186 | 0.0015 | 0.14 | 88.4 |
| 4 | 0.642 | –0.0142 | 0.00691 | 0.247 | 77.6 |
| 5 | 0.749 | –0.012 | 0.00515 | 0.217 | 87.7 |
| 6 | 0.499 | –0.0023 | 0.00698 | 0.274 | 78.6 |
| 7 | 0.953 | 0.000792 | 0.00073 | 0.0663 | 94.3 |
## 3.5. Comparison with classical regression methods
The statistics shown above should be compared with an earlier prediction model (Ma et al., 2020). In loc. cit. the authors used classical methods such as Linear Regression, Bayesian Regression, and so on. For example, in the case of P1, the comparison of statistics of our TDA method with classical methods is shown in Table 6.
**Table 6**
| Unnamed: 0 | R2 score | MAE | MSE | Max error | Hit in percentage % |
| --- | --- | --- | --- | --- | --- |
| TDA | 0.726 | –0.00807 | 0.00607 | 0.222 | 82.7 |
| Linear regression | 0.607 | 0.0632 | 0.00704 | 0.384 | 55.7 |
| Ridge regression | 0.642 | 0.0693 | 0.00794 | 0.315 | 46.8 |
| Bayesian regression | 0.689 | 0.0635 | 0.0069 | 0.283 | 55.5 |
| Random forest | 0.733 | 0.0372 | 0.00593 | 0.562 | 78.6 |
| Gradient boosting | 0.783 | 0.0364 | 0.00481 | 0.311 | 79.2 |
This shows that the mapper algorithm combined with Markov chain construction is more efficient than the more classical regression methods in the study of progression analysis of Parkinson's disease.
## 4. Discussion
In this study, we develop a new predictive model for motor progression in patients with early PD by mapper algorithm, which we report $62.5\%$ accuracy in the group of un-medicated patients (Medication type 0); while in other medication types, the accuracy increased, fluctuating between 77.6 and $97\%$ (Medication type 1–7). Also, we compared different methods in the analysis of PD progression and found that mapper algorithm combined with Markov chain construction is more efficient than the more classical regression methods. This prediction model is an upgrade of our previous prediction model, which improves the accuracy and has better stability. Our findings indicate that the models can practically predict the MDS-UPDRS Part III score of the coming year based on the clinically available characteristics obtained in the current year.
There are a growing number of clinical prediction models of the progression of PD, which vary from the choices of predictive values according to different objectives. Latourelle et al. developed and validated a comprehensive multivariable prognostic model based on the PPMI database (Latourelle et al., 2017). In this model, they obtained a R2 of $41\%$ in PPMI database and $9\%$ in LABS-PD database that used for external validation. This reduction of R2 could be offset by increasing the sample size. As in Lu et al. they developed a progression model based on the videos of MDS-UPDRS tests to estimate the motor severity of PD, in which they obtained a classification accuracy of $72\%$ and F1-score of 0.51 (Lu et al., 2021).
Eight variables were enrolled in this model, including age, MDS-UPDRS III, NP1 apathy score, NP1 fatigue score, NP1 anxiety score, MOCA, SCOPA-AUT, and initial symptoms. These variables contain quantification of motor (MDS-UPDRS III) and non-motor symptoms (apathy, cognitive dysfunction, fatigue and anxiety), all of which contribute to the progression of PD.
Previous studies have identified that cognitive impairment at baseline is correlated with faster disease progression and greater motor impairment (Velseboer et al., 2013; Fereshtehnejad et al., 2015; Reinoso et al., 2015). Apart from UPDRS values, signs of cognitive decline, orthostatic hypotension and rapid eye movement sleep behavior disorder at baseline, could also suggest a much faster decline in motor symptoms. An increase in L-dopa non-responsive symptoms, which suggest a diffuse destruction of extra-nigrostriatal pathways in parallel with the nigrostriatal pathway (Velseboer et al., 2013) may in part explain the situation.
Overall, PD is a neurodegeneration disease and all the patients suffer from progressive aggravation. The expected growth of motor score varies greatly due to different medication types. The rate of progress of patients with no medication is 2.17 per year, which is representative of PD's natural course. Anti-PD drugs can improve patients' motor symptoms, while the expected growth of UPDRS III score in patients taking medicine is lower than type 1. We also found the expected growth of UPDRS III score in groups 5 (levodopa + dopamine agonist) and 6 (dopamine agonist + other) is lower than other types, indicating that dopamine agonists might improve motor dysfunction better or exist potential disease-modifying effect. However, given the complexity of drugs regulation and interactions with patients, further interpretation should be given cautiously. In addition, according to the type of medication used by the patients, the accuracy of prediction model in the patients taking the anti-PD medication was improved compared to patients with no medication, ranging from 77.6 to $97\%$. The reason is that in the type 0 case, patients' UPDRS III score could experience a “jumping" phenomenon, thus making our continuous topological method not as effective as in the case of other medication types. In fact, identifying features of this jumping phenomenon is itself an interesting question which we plan to further investigate in a future work.
There are also some limitations in this study. First, the variability and subjectiveness of measures of the motor and non-motor scores within the PPMI dataset may exist. Second, due to limited PD patients, only uniform predictions across subtypes were made without consideration of PD subtypes. Third, we just predict the MDS-UPDRS Part III total score in the predict model, and no subdivision prediction was made for a single item or symptom category score (such as limb rigidity, central axis slowing, tremor, gait, etc.). Finally, our analysis was based on the early stage of PD. As a result, this model cannot be apply to patients with advanced PD for motor prediction.
In this study, by using mapper algorithm, we apply relatively fewer parameters to achieve better results than the previous models, provide accuracy in the range of 62.5 − $97.0\%$ in predicting motor progression depending on different medication types. The use of this model can predict motor evaluations at the individual level, assisting clinicians to adjust intervention strategy for each patient and identifying at-risk patients for future disease-modifying therapy clinical trials.
## Data availability statement
Publicly available datasets were analyzed in this study. 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 at: www.ppmi-info.org.
## Author contributions
L-YM: research directions throughout the process and provides medical advise for feature selection together with TF. TF: medical advise for feature selection, suggest to produce practical applications of the algorithm, and such as comparing different medication types using the algorithm. CH: algorithm implementations and coding. ML: data pre-processing. KR: current collaboration and discussions throughout different stages of the program. JT: algorithm implementations involved in the program. All authors contributed to the article and approved the submitted version.
## Conflict of interest
KR was employed by GYENNO Science 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. The handling editor JM declared a shared parent affiliation with the authors L-YM and TF at the time of review.
## 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. Dagliati A., Geifman N., Peek N., Holmes J. H., Sacchi L., Bellazzi R.. **Using topological data analysis and pseudo time series to infer temporal phenotypes from electronic health records**. *Artif. Intell. Med* (2020) **108** 101930. DOI: 10.1016/j.artmed.2020.101930
2. Fereshtehnejad S. M., Romenets S. R., Anang J. B., Latreille V., Gagnon J. F., Postuma R. B.. **New clinical subtypes of Parkinson disease and their longitudinal progression: a prospective cohort comparison with other phenotypes**. *JAMA Neurol* (2015) **72** 863-873. DOI: 10.1001/jamaneurol.2015.0703
3. Foltynie T., Brayne C., Barker R. A.. **The heterogeneity of idiopathic Parkinson's disease**. *J Neurol* (2002) **249** 138-145. DOI: 10.1007/PL00007856
4. Gottipati G., Karlsson M. O., Plan E. L.. **Modeling a composite score in Parkinson's disease using item response theory**. *AAPS J* (2017) **19** 837-845. DOI: 10.1208/s12248-017-0058-8
5. Gramotnev G., Gramotnev D. K., Gramotnev A.. **Parkinson's disease prognostic scores for progression of cognitive decline**. *Sci. Rep* (2019) **9** 17485. DOI: 10.1038/s41598-019-54029-w
6. Hogue O., Fernandez H. H., Floden D. P.. **Predicting early cognitive decline in newly-diagnosed Parkinson's patients: a practical model**. *Parkinsonism Relat. Disord* (2018) **56** 70-75. DOI: 10.1016/j.parkreldis.2018.06.031
7. Holden S. K., Finseth T., Sillau S. H., Berman B. D.. **Progression of MDS-UPDRS scores over five years in**. *Mov. Disord. Clin. Pract* (2018) **5** 47-53. DOI: 10.1002/mdc3.12553
8. Kraft R.. *Illustrations of data analysis using the mapper algorithm and persistent homology. TRITA-MAT-E* (2016)
9. Latourelle J. C., Beste M. T., Hadzi T. C., Miller R. E., Oppenheim J. N., Valko M. P.. **Large-scale identification of clinical and genetic predictors of motor progression in patients with newly diagnosed Parkinson's disease: a longitudinal cohort study and validation**. *Lancet Neurol* (2017) **16** 908-916. DOI: 10.1016/S1474-4422(17)30328-9
10. Li L., Cheng W. Y., Glicksberg B. S., Gottesman O., Tamler R., Chen R.. **Identification of type 2 diabetes subgroups through topological analysis of patient similarity**. *Sci. Transl. Med* (2015) **7** 311ra174. DOI: 10.1126/scitranslmed.aaa9364
11. Lu M., Zhao Q., Poston K. L., Sullivan E. V., Pfefferbaum A., Shahid M.. **Quantifying Parkinson's disease motor severity under uncertainty using MDS-UPDRS videos**. *Med. Image Anal* (2021) **73** 102179. DOI: 10.1016/j.media.2021.102179
12. Lum P. Y., Singh G., Lehman A., Ishkanov T., Vejdemo-Johansson M., Alagappan M.. **Extracting insights from the shape of complex data using topology**. *Sci. Rep* (2013) **3** 1236. DOI: 10.1038/srep01236
13. Ma L. Y., Chan P., Gu Z. Q., Li F. F., Feng T.. **Heterogeneity among patients with Parkinson's disease: cluster analysis and genetic association**. *J Neurol Sci* (2015) **351** 41-45. DOI: 10.1016/j.jns.2015.02.029
14. Ma L. Y., Tian Y., Pan C. R., Chen Z. L., Ling Y., Ren K.. **Motor progression in early-stage parkinson's disease: a clinical prediction model and the role of cerebrospinal fluid biomarkers**. *Front. Aging Neurosci* (2020) **12** 627199. DOI: 10.3389/fnagi.2020.627199
15. Nicolau M., Levine A. J., Carlsson G.. **Topology based data analysis identifies a subgroup of breast cancers with a unique mutational profile and excellent survival**. *Proc. Natl. Acad. Sci. U.S.A* (2011) **108** 7265-7270. DOI: 10.1073/pnas.1102826108
16. Pyatigorskaya N., Yahia-Cherif L., Valabregue R., Gaurav R., Gargouri F., Ewenczyk C.. **Parkinson disease propagation using MRI biomarkers and partial least squares path modeling**. *Neurology* (2021) **96** e460-e471. DOI: 10.1212/WNL.0000000000011155
17. Rahayel S., Postuma R. B., Montplaisir J., Miši,ć B, Tremblay C., Vo A.. **A prodromal brain-clinical pattern of cognition in synucleinopathies**. *Ann. Neurol* (2021) **89** 341-357. DOI: 10.1002/ana.25962
18. Reinoso G., Allen J. C., Au W. L., Seah S. H., Tay K. Y., Tan L. C.. **Clinical evolution of Parkinson's disease and prognostic factors affecting motor progression: 9-year follow-up study**. *Eur. J. Neurol* (2015) **22** 457-463. DOI: 10.1111/ene.12476
19. Rossi-deVries J., Pedoia V., Samaan M. A., Ferguson A. R., Souza R. B., Majumdar S.. **Using multidimensional topological data analysis to identify traits of hip OA**. *J. Magn. Reson. Imaging* (2018) **48** 1046-1058. DOI: 10.1002/jmri.26029
20. Schrag A., Siddiqui U. F., Anastasiou Z., Weintraub D., Schott J. M.. **Clinical variables and biomarkers in prediction of cognitive impairment in patients with newly diagnosed Parkinson's disease: a cohort study**. *Lancet Neurol* (2017) **16** 66-75. DOI: 10.1016/S1474-4422(16)30328-3
21. Selikhova M., Williams D. R., Kempster P. A., Holton J. L., Revesz T., Lees A. J.. **A clinico-pathological study of subtypes in Parkinson's disease**. *Brain* (2009) 2947-2957. DOI: 10.1093/brain/awp234
22. Severson K. A., Chahine L. M., Smolensky L. A., Dhuliawala M., Frasier M., Ng K.. **Discovery of Parkinson's disease states and disease progression modelling: a longitudinal data study using machine learning**. *Lancet Digit. Health* (2021) **3** e555-e564. DOI: 10.1016/S2589-7500(21)00101-1
23. Simuni T., Long J. D., Caspell-Garcia C., Coffey C. S., Lasch S., Tanner C. M.. **Predictors of time to initiation of symptomatic therapy in early Parkinson's disease**. *Ann. Clin. Transl. Neurol* (2016) **3** 482-494. DOI: 10.1002/acn3.317
24. Singh G., Memoli F., Carlsson G. E.. **“Topological methods for the analysis of high dimensional data sets and 3d object recognition,”**. *Eurographics Symposium on Point-Based Graphics* (2007) 91-100
25. van Veen H. J., Saul N., Eargle D., Mangham S. W.. **Kepler mapper: a flexible python implementation of the mapper algorithm**. *J. Open Sourc. Softw* (2019a) **4** 1315. DOI: 10.21105/joss.01315
26. van Veen H. J., Saul N., Eargle D., Mangham S. W.. *Kepler Mapper: A Flexible Python Implementation of the Mapper Algorithm (Version 1.4.1)* (2019b). DOI: 10.5281/zenodo.4077395
27. Velseboer D. C., Broeders M., Post B., van Geloven N., Speelman J. D., Schmand B.. **Prognostic factors of motor impairment, disability, and quality of life in newly diagnosed PD**. *Neurology* (2013) **80** 627-633. DOI: 10.1212/WNL.0b013e318281cc99
28. Vu T. C., Nutt J. G., Holford N. H.. **Progression of motor and nonmotor features of Parkinson's disease and their response to treatment**. *Br. J. Clin. Pharmacol* (2012) **74** 267-283. DOI: 10.1111/j.1365-2125.2012.04192.x
|
---
title: Identification of biomarkers and prediction of upstream miRNAs in diabetic
nephropathy
authors:
- Dapeng Yin
- Zhixin Guo
- Xinyu Zhang
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9989306
doi: 10.3389/fendo.2023.1144331
license: CC BY 4.0
---
# Identification of biomarkers and prediction of upstream miRNAs in diabetic nephropathy
## Abstract
### Objective
To explore biomarkers of diabetic nephropathy (DN) and predict upstream miRNAs.
### Methods
The data sets GSE142025 and GSE96804 were obtained from Gene Expression Omnibus database. Subsequently, common differentially expressed genes (DEGs) of renal tissue in DN and control group were identified and protein-protein interaction network (PPI) was constructed. *Hub* genes were screened from in DEGs and made an investigation on functional enrichment and pathway research. Finally, the target gene was selected for further study. The receiver operating characteristic (ROC) curve was used to evaluate the diagnostic efficiency of target gene and predicted its upstream miRNAs.
### Results
130 common DEGs were obtained through analysis, and 10 *Hub* genes were further identified. The function of *Hub* genes was mainly related to extracellular matrix (ECM), collagen fibrous tissue, transforming growth factor (TGF) -β, advanced glycosylation end product (AGE) -receptor (RAGE) and so on. Research showed that the expression level of *Hub* genes in DN group was significantly higher than that in control group. ( all $P \leq 0.05$). The target gene matrix metalloproteinase 2 (MMP2) was selected for further study, and it was found to be related to the fibrosis process and the genes regulating fibrosis. Meanwhile, ROC curve analysis showed that MMP2 had a good predictive value for DN. miRNA prediction suggested that miR-106b-5p and miR-93-5p could regulate the expression of MMP2.
### Conclusion
MMP2 can be used as a biomarker for DN to participate in the pathogenesis of fibrosis, and miR-106b-5p and miR-93-5p may regulate the expression of MMP2 as upstream signals.
## Introduction
Diabetic nephropathy (DN) is a common microvascular complication of diabetes and a major cause of end-stage renal disease (ESRD) throughout the world [1]. DN is characterized by glomerular basement membrane thickening, mesangial expansion and accumulation of matrix, and eventually progression to fibrosis of glomeruli and renal tubules. Its pathogenesis is very complex, involving advanced glycosylation end products (AGE), inflammation, oxidative stress, apoptosis, autophagy and various signaling mechanisms leading to extracellular matrix (ECM) deposition and renal interstitial fibrosis [2]. Finally, it leads to the appearance of albuminuria and the progressive decrease of estimated glomerular filtration rate (eGFR). Therefore, new identification and therapeutic tools to improve DN are urgently needed.
miRNA is an endogenous non-coding RNA (about 22 nucleotides long) that usually targets the 3 ‘untranslated region (3’ -UTRs) of mRNA, resulting in post-transcriptional gene silencing [3]. A large amount of evidence has shown that miRNA is involved in the occurrence of diabetes and its complications, such as DN, neuropathy, retinopathy, cardiomyopathy and wound healing. Zhong et al. found that miR-21 is highly expressed in the renal cortex of diabetes mice with microalbuminuria and positively regulated the expression of ECM by regulating transforming growth factor (TGF)- β, Nuclear factor kB (NF kB) and Smad homolog 7 (SMAD7) play a pathological role in renal fibrosis and inflammation [4]. Therefore, regulation of miRNA is promising in the treatment of DN.
Bioinformatics, which research objects mainly focus on gene and protein, has played a vital role in the research of life science with its rapidly development in recent years. Although many mechanisms are involved in the pathogenesis of DN, the potential bioinformatics driving its pathogenesis is rarely fully elucidated. In this study, the gene expression data of DN kidney tissue from the Gene Expression Omnibus (GEO) was analyzed to predict the key genes causing renal fibrosis and explore their upstream miRNAs. This provides a new idea for the early diagnosis and treatment of DN. The specific working flow chart is shown in Figure 1.
**Figure 1:** *Work flow diagram includes data preparation, processing and analysis.*
## Microarray data
The DN gene expression data sets were retrieved in the NCBI-GEO database (https://www.ncbi.nlm.nih.gov/geo/) using the keywords “diabetic nephropathy” and “diabetic kidney disease”. Finally, two renal tissue gene expression data sets GSE142025 and GSE96804 with relatively large sample size were finally obtained. GSE142025 is based on GPL20301 platform, including 27 cases of DN and 9 cases of normal control group [5]. GSE96804 is based on GPL17586 platform, including 41 cases of DN and 20 cases of renal tissue control not affected by tumor [6]. After downloading the gene expression matrix, the probe identification number (ID) was converted into gene symbol using R software (version 4.2.2). For multiple probes corresponding to a gene, the average expression value was taken as the gene expression value.
## Differential expression analysis
Limma software package can perform differential expression analysis on RNA sequencing data and high-throughput data, which contains powerful functions of reading, standardizing and exploring data [7]. The R software and the limma package in Bioconductor were used to identify differentially expressed genes (DEGs) between DN and control group. The ratio of gene expression between the two groups (fold change, FC) was calculated, and the logarithm base 2 (log2FC) and adjusted P-value were selected. In line with the adjusted P - value < 0.05, | log2FC | > 1 genes are considered to be DEGs. Negative values represent down- regulated gene, while positive values represent up-regulated gene. Use ggplot 2 package (R software) to make volcanic maps. Then, through the online website (http://bioinformatics.psb.ugent.be/webtools/Venn/) draw Venn diagram to identify common DEGs for subsequent analysis.
## PPI network construction and hub gene identification
STRING(https://cn.string-db.org/) (version 11.5) is an online tool for interacting gene retrieval, which provides new insights into the molecular mechanism of disease occurrence by integrating protein interactions [8]. Protein-protein interaction (PPI) networks for common DEGs were predicted by STRING. The confidence score (>0.40) was used as the screening criteria, and the results were visualized by the CytoScape software (version 3.8.0). CytoScape is an open source bioinformatics software platform for visualizing molecular interaction networks. CytoHubba is a tool for defining network topology to find *Hub* genes. Subsequently, MCC algorithm of CytoHubba plugin was used to select the top 10 genes with the highest node connection closeness as the *Hub* genes, and boxplot was drawn to show the expression level.
## Biological function and pathway analysis
DAVID database (https://david.ncifcrf.gov/) (version DAVID2021), an online gene function classification tool, through the powerful aggregation algorithm, complete the annotation, visualization and integration for gene and protein function [9]. Gene Ontology (GO) analysis is a bioinformatics method that annotates genes and their protein products. It is widely used in analyzing the functional similarity between genes and identifying biological functions and pathways related to diseases by high-throughput biological data analysis [10]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) is a public and comprehensive database for understanding the advanced functions of cells and organisms from genomics and other high-throughput data sets, integrating genomes, biological pathways, diseases, drugs and chemicals [11, 12]. In order to understand the function of *Hub* genes, we used the DAVID database for GO functional analysis and KEGG pathway enrichment. Typically, GO analyses are annotated with biological processes (BP), cellular components (CC), and molecular functions (MF). $P \leq 0.05$ was considered statistically significant. If there are more than 10 results, this study will only show the top 10 results in P-value ranking.
## Genomic variation analysis and correlation analysis
The first gene ranked by MCC algorithm among *Hub* genes was selected as the target gene analysis. Each sample in the dataset GSE142025 was scored for fibrosis enrichment using the GSVA package in Bioconductor and R software, and the results were visualized using heat maps. Subsequently, the correlation between target genes and common fibrosis genes was analyzed by Pearson correlation. *The* genes related to fibrosis function were collected from the GSEA (http://www.gsea-msigdb.org/gsea/index.jsp) (version 4.3.2).
## ROC curve analysis
In order to effectively distinguish patients with DN from the control group, we analyzed the ROC curve of the target gene using the pROC package (R software).
## miRNA prediction
Encyclopedia of RNA Interactors (ENCORI) database (https://starbase.sysu.edu.cn/) is an open source network tool for studying the interactions among ncRNAs and RNA-RNA [13]. Based on the ENCORI database, we predicted the upstream miRNAs of the *Hub* genes, and the screening results were jointly calculated by miRanda and Targetscan plugins. The results were presented in the network diagram by the CytoScape software.
## Statistical analysis
Graphpad Prism (version 9.3.0) and R software were used for statistical analysis. The *Hub* genes expression analysis uses t test, and the pROC package (R software) was used for ROC curve analysis to calculate the area under the curve (AUC).
## Identification of DEGs
Take adjusted P-value<0.05, | log2FC |>1 as the screening criteria. 1166 DEGs were obtained from the GSE142025 dataset, including 660 up-regulated and 506 down-regulated genes. 609 DEGs were obtained from GSE96804 dataset, including 283 up-regulated and 326 down-regulated genes. The volcano maps were drawn separately to show the DEGs of two data sets (Figures 2A, B), and 130 common DEGs were identified by Venn diagram (Figure 2C). Table 1 lists common DEGs, in which NPIPB5, FAM151A and NKG7 have opposite expressions in the two data sets are not listed.
**Figure 2:** *(A) Volcano maps of differentially expressed genes in data sets GSE1422025, (B) GSE96804. (C) Common differentially expressed genes in the two data sets. (D) PPI network of differentially expressed genes. (E) The PPI network of the top 10 *Hub* genes.* TABLE_PLACEHOLDER:Table 1
## PPI network and hub gene screening
PPI network analysis was performed on 130 common DEGs using STRING and Cytoscape for visualization (Figure 2D). The CytoHubba plugin was used for PPI network correlation analysis, and 10 *Hub* genes were identified according to MCC algorithm (Figure 2E). In order of rank, they were MMP2, FN1, COL1A2, COL3A1, COL6A3, LUM, VCAN, FBN1, THBS2 and FOS. Finally, MMP2, which was ranked first, was selected as the target gene for further analysis.
## GO function and KEGG pathway enrichment analysis
GO functional analysis of the *Hub* genes showed that biological processes were mainly enriched in skeletal system development, heart development, cellular response to amino acid stimulus, collagen fibril organization, TGF-β receptor signaling pathway, cell adhesion, extracellular matrix organization, cellular response to reactive oxygen species and other functions (Figure 3A). In terms of cell components, genes were mainly enriched in ECM, extracellular region, and organelles of endocellular structures (Figure 3B). At the molecular level, genes were enriched in ECM structural components, ECM structures related to tensile and compressive strength, protease binding, integrin binding and heparin binding (Figure 3C). KEGG pathway analysis showed that the *Hub* genes was enriched in ECM-receptor interaction, AGE-RAGE signaling pathway in diabetic complications, relaxin signaling pathway focal adhesion, PI3K-Akt signaling pathway, and so on (Figure 3D). ( all $P \leq 0.05$). These findings indicate that *Hub* gene is involved in the occurrence of DN and plays an important role in regulating ECM formation and fibrosis.
**Figure 3:** *Gene Ontology (GO) function and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of Hub genes. (A) Biological Processes (BP), (B) cellular components (CC), (C) molecular functions (MF), (D) KEGG.*
## MMP2 participates in the development of fibrosis
Renal fibrosis is one of the main causes of DN, which is characterized by the activation and proliferation of fibroblasts and the deposition of ECM [14]. Matrix metalloproteinase-2(MMP2) play an important role in renal fibrosis and DN [15]. Therefore, we investigated the effect of MMP2 activation on fibrotic pathways and cytokine characteristics. GSVA was used to determine the enrichment fraction of the fibrosis process. The results showed that in the GSE142025 database, MMP2 expression was positively correlated with most fibrosis functions, including stress fiber assembly, elastic fiber assembly, outer dense fiber, contraction fiber and other functions (Figure 4A). The correlation was tested by Person correlation analysis. In both data sets, the correlation coefficients between MMP2 and common fibrosis genes were positive numbers (Table 2). It was shown that MMP2 was positively correlated with common fibrosis genes such as TGFB1, FN1 and CXCL6 (Figures 4B, C). These results support the hypothesis that MMP2 is involved in the regulation of fibrosis during the development of DN.
**Figure 4:** *(A) Heat map of correlation between MMP2 expression and fibrosis function enrichment score. The samples were arranged in ascending order of MMP2 expression. (B) Correlation between MMP2 and fibrosis gene in data sets GSE142025, (C) GSE96804. Correlation coefficients are shown in the lower left corner. In the upper right, the correlation coefficients are also shown to be pie chart proportions.* TABLE_PLACEHOLDER:Table 2
## The expression of Hub gene and its diagnostic value in DN
In the GSE142025 dataset, compared with the control group, the expressions of MMP2, FN1, COL1A2, COL3A1, COL6A3, LUM, VCAN, FBN1, THBS2 in DN renal tissue were up-regulated, while FOS expression was down-regulated ($P \leq 0.05$). This result was verified in the GSE96804 dataset (Figures 5A, B). The target gene MMP2 was selected for ROC curve analysis, and the results show that AUC in the data sets GSE142025 and GSE96804 were $95.1\%$ and $90.4\%$ respectively (Figures 5C, D). This suggests that MMP2 is significantly enriched in DN and can be used as a potential biomarker to predict its occurrence.
**Figure 5:** *(A) Expression levels of hub genes in data sets GSE142025, (B) GSE96804, all P-value<0.05. (C) Receiver Operating Characteristic (ROC) curves of MMP2 in GSE142025, (D) GSE96804. The area under the curve (AUC) were 95.1% and 90.0%, respectively.*
## Prediction of miRNA regulating MMP2
In this study, the upstream miRNAs of MMP2 were predicted through ENCORI database, 13 miRNAs were predicted in TargetScan plugin, and 21 miRNAs were predicted in miRanda plugin. The intersection of the two algorithms was selected. Finally, the results showed that the upstream miRNAs of MMP2 included hsa-miR-17-5p, hsa-miR-20a-5p, hsa-miR-93-5p, hsa-miR-106a-5p, hsa-miR-106b-5p, and hsa-miR-106B-5P. hsa-miR-20b-5p and hsa-miR-519d-3p (Figure 6).
**Figure 6:** *Regulatory network of MMP2 and upstream miRNAs.*
## Discussion
DN is the most common microvascular complication of diabetes, increasing in incidence year on year and becomes the main cause of end-stage renal disease [16]. At present, there are few studies on the genome of DN. Exploring its changes in the genome, analyzing the involved biological functions and predicting the possible molecular mechanisms will be helpful for the diagnosis and treatment of DN.
In this study, the common DEGs of DN and the control group were obtained based on the gene chip matrix, and 10 *Hub* genes were finally selected for GO functional analysis and KEGG pathway enrichment. It is found that the functions of *Hub* genes are mainly concentrated in “collagen fiber tissue, ECM components, TGF-β receptor signaling pathway” and other aspects. This is consistent with the previous study. Some scholars found that genes related to angiogenesis, ECM secretion and immune cell infiltration were involved in the occurrence of diseases through single cell sequencing of DN glomerulus. At the same time, this study revealed that although eGFR was relatively normal at the early stage of the disease, there were mild to moderate glomerulosclerosis and interstitial fibrosis [17]. In DN patients and animal models, urinary TGF-β level was higher than control group, and plasma TGF-β1 level was closely correlated with the severity of renal dysfunction [18]. Overproduction of TGF-β1 in the glomerular apparatus led to proteinuria, polyuria, decreased eGFR, and increased ECM deposition in the glomeruli, indicating that TGF- β participated in the occurrence of DN [19]. These studies are basically consistent with the functions enriched by *Hub* genes, suggesting that these genes have diagnostic and therapeutic value for DN.
The enrichment of KEGG pathway showed significant enrichment in “ECM-receptor interaction, AGE-RAGE signaling pathway in diabetic complications, PI3K-Akt signaling pathway, adhesion plaque” and other aspects. It was found that AGE can induce collagen production, and inhibition of collagen production can significantly reduce the severity of DN. At the same time, AGE can activate TGF- β/Smad pathway and enhanced TGF- β Transcriptional activity. In addition, the crosstalk mechanism of AGE (RAGE)-ERK/p38MAKPs-Smad can activate Smad$\frac{2}{3}$ to play an important role in DN [19]. This indicates that AGE-RAGE and TGF-β pathway related genes play an important role in the process of DN fibrosis. On the other hand, AGE-RAGE signaling pathway is related to the production of reactive oxygen species (ROS) and mitochondrial dysfunction. The inflammatory reaction mediated by NF-kB and PI3K-Akt signaling pathways triggered by ROS aggravates the renal injury in diabetes [20]. Therefore, it is speculated that genes involved in the PI3K-Akt pathway and AGE-RAGE pathway have a hand in DN in the induction of oxidative stress and inflammation response.
For this reason, improving renal fibrosis plays a positive role in the treatment of DN. In this context, we aimed to further explore the related genes involved in fibrosis, finally selected the target gene MMP2 for further research, explored the miRNAs acting on this gene, and find a new therapeutic target to improve DN.
MMP2 plays a key role in the development of chronic kidney disease by remodeling the extracellular matrix, distorting the structure of the glomerular basement membrane, contributing to the development of tubulointerstitial fibrosis and leading to progressive kidney injury [15]. The study found that the expression of MMP2 in renal tissue of DN patients increased [21], and the serum level of MMP-2 was significantly higher than that of normal healthy subjects [22]. Serum creatinine, eGFR and proteinuria were remarkable correlated with serum MMP-2 levels in patients with DN [22]. This was consistent with our research findings, the expression level of MMP2 in DN nephridial tissue was significantly higher than in the control group. At the same time, GSVA analysis indicated that MMP2 was bound up with the most fibrosis processes. Correlation analysis showed that the expression of MMP2 was positively correlated with common fibrosis genes such as TGFB1, FN1, CXCL6, etc. Based on the above results, we predicted that MMP2 could be used as a relevant target for the diagnosis of DN and drug intervention. Subsequently, ROC curve analysis verified that MMP2 could effectively distinguish patients with DN from the control group. The involvement of MMP2 in the occurrence of chronic kidney disease also contains other mechanisms, such as inducing the production of mitochondrial ROS, mitochondrial autophagy, activating systemic inflammatory response, and infiltration of renal monocytes, which cause structural abnormalities in the kidney [23]. This requires us to further study its mechanism in DN.
miRNA is found to regulate the signaling pathways related to inflammation, autophagy, apoptosis and fibrosis, and participate in the occurrence of DN. It is reported that many miRNAs, like miR-21, miR-217, miR-216a, miR-200, miR-195, miR-451, miR-141, miR-93, miR-29, etc regulate DN signal transduction by combining with 3’UTR of target genes [24]. Similarly, multiple drugs play a protective role in DN by regulating miRNA expression. Resveratrol upregulates autophagy and inhibits apoptosis by inhibiting the expression of miR-383-5p [25]. It could also inhibit apoptosis by up regulating the expression of miR-18a-5p, thus having beneficial effects on the kidney [26]. Hyperoside could cause epigenetic changes by down-regulation of mi-R21 and increase the level of MMP-9 protein [27]. Through the above mechanism, hyperoside could regulate ECM and have an impact on the renal function of DN mice [28]. For this reason, this study predicted the upstream miRNAs regulating MMP2, and identified that hsa-miR-17-5p, hsa-miR-20a-5p, hsa-miR-93-5p, hsa-miR-106a-5p, hsa-miR-106b-5p, hsa-miR-20b-5p and hsa-miR-519d-3p could regulate the expression of MMP2. Previous study has shown that miR-106b-5p is involved in the development of chronic thromboembolic pulmonary hypertension by negatively regulating the expression of MMP2 [29]. miR-93-5p inhibits the proliferation, invasion and migration of tumor cells by targeting MMP2 in glioma [30]. This provides a theoretical basis for miR-106b-5p and miR-93-5p to regulate MMP2 and thus delay the development of DN, but the specific regulatory mechanism needs further investigation.
There are some limitations for this research. First of all, there are few data sets available for us to choose. The number of samples of DN patients is limited and clinical data of samples cannot be obtained. Secondly, when GSVA analysis is conducted, due to differences in sequencing genes obtained from different samples, all data sets cannot be used for analysis. Thirdly, the biomarkers identified in this study have not been experimentally verified, and the relevant mechanisms will be further investigated in future studies.
## Conclusions
We determined that MMP2 is a gene closely related to DN fibrosis, which can effectively distinguish patients with DN from the control group. It is predicted that miR-106b-5p and miR-93-5p are upstream signals that regulate the expression of MMP2 and thereby produce a marked effect in DN.
## Data availability statement
Publicly available datasets were analyzed in this study. This data can be found here: https://www.ncbi.nlm.nih.gov/geo/, (Gene Expression Omnibus), (GSE142025,GSE96804).
## Author contributions
Research design, data processing and interpretation, statistical analysis and manuscript drafting were completed by DY. Research concept, data results check and manuscript knowledge revision were completed by ZG. Figures and diagrams drawing were completed by XZ. 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. Qi C, Mao X, Zhang Z, Wu H. **Classification and differential diagnosis of diabetic nephropathy**. *J Diabetes Res* (2017) **2017** 8637138. DOI: 10.1155/2017/8637138
2. Li A, Yi B, Han H, Yang S, Hu Z, Zheng L. **(vitamin d receptor) regulates defective autophagy in renal tubular epithelial cell in streptozotocin-induced diabetic mice**. *Autophagy* (2022) **18**. DOI: 10.1080/15548627.2021.1962681
3. Ferragut Cardoso AP, Banerjee M, Nail AN, Lykoudi A, States JC. **miRNA dysregulation is an emerging modulator of genomic instability**. *Semin Cancer Biol* (2021) **76**. DOI: 10.1016/j.semcancer.2021.05.004
4. Zhong X, Chung AC, Chen HY, Dong Y, Meng XM, Li R. **miR-21 is a key therapeutic target for renal injury in a mouse model of type 2 diabetes**. *Diabetologia* (2013) **56**. DOI: 10.1007/s00125-012-2804-x
5. Fan Y, Yi Z, D'Agati VD, Sun Z, Zhong F, Zhang W. **Comparison of kidney transcriptomic profiles of early and advanced diabetic nephropathy reveals potential new mechanisms for disease progression**. *Diabetes* (2019) **68**. DOI: 10.2337/db19-0204
6. Pan Y, Jiang S, Hou Q, Qiu D, Shi J, Wang L. **Dissection of glomerular transcriptional profile in patients with diabetic nephropathy: SRGAP2a protects podocyte structure and function**. *Diabetes* (2018) **67**. DOI: 10.2337/db17-0755
7. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W. **Limma powers differential expression analyses for RNA-sequencing and microarray studies**. *Nucleic Acids Res* (2015) **43**. DOI: 10.1093/nar/gkv007
8. Szklarczyk D, Gable AL, Lyon D, Junge A, Wyder S, Huerta-Cepas J. **STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets**. *Nucleic Acids Res* (2019) **47**. DOI: 10.1093/nar/gky1131
9. Huang DW, Sherman BT, Tan Q, Collins JR, Alvord WG, Roayaei J. **The DAVID gene functional classification tool: a novel biological module-centric algorithm to functionally analyze large gene lists**. *Genome Biol* (2007) **8** R183. DOI: 10.1186/gb-2007-8-9-r183
10. Chen L, Zhang YH, Wang S, Zhang Y, Huang T, Cai YD. **Prediction and analysis of essential genes using the enrichments of gene ontology and KEGG pathways**. *PloS One* (2017) **12**. DOI: 10.1371/journal.pone.0184129
11. Kanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. **KEGG as a reference resource for gene and protein annotation**. *Nucleic Acids Res* (2016) **44**. DOI: 10.1093/nar/gkv1070
12. Xie R, Li B, Jia L, Li Y. **Identification of core genes and pathways in melanoma metastasis**. *Int J Mol Sci* (2022) **23** :794. DOI: 10.3390/ijms23020794
13. Lan Z, Yao X, Sun K, Li A, Liu S, Wang X. **The interaction between lncRNA SNHG6 and hnRNPA1 contributes to the growth of colorectal cancer by enhancing aerobic glycolysis through the regulation of alternative splicing of PKM**. *Front Oncol* (2020) **10** 363. DOI: 10.3389/fonc.2020.00363
14. Li H, Rong P, Ma X, Nie W, Chen Y, Zhang J. **Mouse umbilical cord mesenchymal stem cell paracrine alleviates renal fibrosis in diabetic nephropathy by reducing myofibroblast transdifferentiation and cell proliferation and upregulating MMPs in mesangial cells**. *J Diabetes Res* (2020) **2020** 3847171. DOI: 10.1155/2020/3847171
15. Cruz JO, Silva AO, Ribeiro JM, Luizon MR, Ceron CS. **Epigenetic regulation of the n-terminal truncated isoform of matrix metalloproteinase-2 (NTT-MMP-2) and its presence in renal and cardiac diseases**. *Front Genet* (2021) **12** 637148. DOI: 10.3389/fgene.2021.637148
16. Samsu N. **Diabetic nephropathy: Challenges in pathogenesis, diagnosis, and treatment**. *BioMed Res Int* (2021) **2021** 1497449. DOI: 10.1155/2021/1497449
17. Wilson PC, Wu H, Kirita Y, Uchimura K, Ledru N, Rennke HG. **The single-cell transcriptomic landscape of early human diabetic nephropathy**. *Proc Natl Acad Sci USA* (2019) **116**. DOI: 10.1073/pnas.1908706116
18. Hathaway CK, Gasim AM, Grant R, Chang AS, Kim HS, Madden VJ. **Low TGFβ1 expression prevents and high expression exacerbates diabetic nephropathy in mice**. *Proc Natl Acad Sci USA* (2015) **112**. DOI: 10.1073/pnas.1504777112
19. Wang L, Wang HL, Liu TT, Lan HY. **TGF-beta as a master regulator of diabetic nephropathy**. *Int J Mol Sci* (2021) **22** 7881. DOI: 10.3390/ijms22157881
20. Lee HW, Gu MJ, Lee JY, Lee S, Kim Y, Ha SK. **Methylglyoxal-lysine dimer, an advanced glycation end product, induces inflammation**. *Mol Nutr Food Res* (2021) **65**. DOI: 10.1002/mnfr.202000799
21. Kim SS, Shin N, Bae SS, Lee MY, Rhee H, Kim IY. **Enhanced expression of two discrete isoforms of matrix metalloproteinase-2 in experimental and human diabetic nephropathy**. *PloS One* (2017) **12**. DOI: 10.1371/journal.pone.0171625
22. Soliman AR, Sadek KM, Thabet KK, Ahmed DH, Mohamed OM. **The role of matrix metalloproteinases 2 in atherosclerosis of patients with chronic kidney disease in type 2 diabetes**. *Saudi J Kidney Dis Transplant an Off Publ Saudi Center Organ Transplant Saudi Arabia* (2019) **30**. DOI: 10.4103/1319-2442.256846
23. Ceron CS, Baligand C, Joshi S, Wanga S, Cowley PM, Walker JP. **An intracellular matrix metalloproteinase-2 isoform induces tubular regulated necrosis: implications for acute kidney injury**. *Am J Physiol Renal Physiol* (2017) **312**. DOI: 10.1152/ajprenal.00461.2016
24. Malakoti F, Mohammadi E, Akbari Oryani M, Shanebandi D, Yousefi B, Salehi A. **Polyphenols target miRNAs as a therapeutic strategy for diabetic complications**. *Crit Rev Food Sci Nutr* (2022) **7** 1-17. DOI: 10.1080/10408398.2022.2119364
25. Huang SS, Ding DF, Chen S, Dong CL, Ye XL, Yuan YG. **Resveratrol protects podocytes against apoptosis**. *Sci Rep* (2017) **7** 45692. DOI: 10.1038/srep45692
26. Xu XH, Ding DF, Yong HJ, Dong CL, You N, Ye XL. **Resveratrol transcriptionally regulates miRNA-18a-5p expression ameliorating diabetic nephropathy**. *Eur Rev Med Pharmacol Sci* (2017) **21**
27. Zhang J, Fu H, Xu Y, Niu Y, An X. **Hyperoside reduces albuminuria in diabetic nephropathy at the early stage through ameliorating renal damage and podocyte injury**. *J Natural medicines* (2016) **70**. DOI: 10.1007/s11418-016-1007-z
28. An X, Zhang L, Yuan Y, Wang B, Yao Q, Li L. **Hyperoside pre-treatment prevents glomerular basement membrane damage in diabetic nephropathy by inhibiting podocyte heparanase expression**. *Sci Rep* (2017) **7** 6413. DOI: 10.1038/s41598-017-06844-2
29. Miao R, Dong X, Gong J, Wang Y, Guo X, Li Y. **Hsa-miR-106b-5p participates in the development of chronic thromboembolic pulmonary hypertension**. *Pulm Circ* (2020) **10** 2045894020928300. DOI: 10.1177/2045894020928300
30. Wu H, Liu L, Zhu JM. **MiR-93-5p inhibited proliferation and metastasis of glioma cells by targeting MMP2**. *Eur Rev Med Pharmacol Sci* (2019) **23**. DOI: 10.26355/eurrev_201911_19446
|
---
title: 'Stroke and the risk of gastrointestinal disorders: A Mendelian randomization
study'
authors:
- Jingru Song
- Wenjing Chen
- Wei Ye
journal: Frontiers in Neurology
year: 2023
pmcid: PMC9989308
doi: 10.3389/fneur.2023.1131250
license: CC BY 4.0
---
# Stroke and the risk of gastrointestinal disorders: A Mendelian randomization study
## Abstract
### Background
The issue of whether a stroke is causally related to gastrointestinal disorders was still not satisfactorily understood. Therefore, we investigated if there is a connection between stroke and the most prevalent gastrointestinal disorders, including peptic ulcer disease (PUD), gastroesophageal reflux disease (GERD), irritable bowel syndrome (IBS), and inflammatory bowel disease (IBD).
### Methods
We applied two-sample Mendelian randomization to investigate relationships with gastrointestinal disorders. We obtained genome-wide association study (GWAS) summary data of any stroke, ischemic stroke, and its subtypes from the MEGASTROKE consortium. From the International Stroke Genetics Consortium (ISGC) meta-analysis, we acquired GWAS summary information on intracerebral hemorrhage (ICH), including all ICH, deep ICH, and lobar ICH. Several sensitivity studies were performed to identify heterogeneity and pleiotropy, while inverse-variance weighted (IVW) was utilized as the most dominant estimate.
### Results
No evidence for an effect of genetic predisposition to ischemic stroke and its subtypes on gastrointestinal disorders were found in IVW. The complications of deep ICH are a higher risk for PUD and GERD. Meanwhile, lobar ICH has a higher risk of complications for PUD.
### Conclusion
This study provides proof of the presence of a brain–gut axis. Among the complications of ICH, PUD and GERD were more common and associated with the site of hemorrhage.
## 1. Introduction
Stroke is one of the leading causes of death and disability worldwide [1, 2]. Based on neuropathology, there are two main categories of stroke: ischemic stroke (IS) and hemorrhagic stroke. Of the two major types of stroke, IS is the more frequent type [3]. There are various subtypes of IS, such as large artery stroke, cardioembolic stroke, and small vessel stroke [4]. Hemorrhagic stroke includes subarachnoid hemorrhage (SAH) and intracerebral hemorrhage (ICH). After a stroke, most patients will have varying degrees of motor impairment, cognitive impairment, speech dysphagia, depression, and other sequelae [5]. In addition, up to $50\%$ of patients usually experience gastrointestinal sequelae [6]. The most common gastrointestinal disorders include PUD, GERD, IBS, and IBD. Among these four diseases, the prevalence of GERD is the highest, up to 18.1–$27.8\%$ in North America, followed by IBS and PUD, and the prevalence of IBD is lower. Patients with IBD commonly have abdominal pain, diarrhea, and bloody stools, while IBS has abdominal pain and altered bowel habits. GERD is usually characterized by regurgitation symptoms and heartburn, while PUD symptoms are not specific and abdominal pain is common (7–10). They sometimes have similar symptoms, such as abdominal pain, and the development of these disorders is all related to the brain–gut axis (11–13). Some observational studies have given attention to the relationship between stroke and peptic ulcer disease (PUD) [14] and also stroke and gastroesophageal reflux disease (GERD) [15]. The study found that the GERD risk of patients with stroke is about 1.51 times that of patients without stroke [15]. However, so far, it is not clear whether there is a causal relationship between the two diseases.
A growing number of observational studies have demonstrated complex interactions between stroke and gastrointestinal disorders (16–18). Furthermore, studies have shown that stroke promotes the destruction of the intestinal barrier and the imbalance of gut microbiota [19, 20]. These proved that there is bidirectional communication between the brain and the gut, usually referred to as the brain–gut or gut–brain axis [21]. After a stroke, the bidirectional communications between the brain and the gut may relate to the dysfunction of the autonomic nervous system, resulting in gastrointestinal disorders [22, 23]. However, the exact mechanism accounting for the brain–gut axis is still widely considered as unsatisfactorily understood.
In systematic reviews and meta-analyses, their causal relationship is unclear or confusing. Mendelian randomization (MR) is a research method using a genetic variation to evaluate the causal relationship between exposures and outcomes based on Mendel's second law. MR overcomes the limitations of observational research by exposing potential causal links and has proved valuable in exploring the causality by using single-nucleotide polymorphisms (SNPs). SNPs are required to be associated with exposures and should not be independently associated with outcomes, except through exposures. Furthermore, SNPs must not be associated with confounders [24, 25]. Moreover, we can further explore the outcomes of insufficient data in RCT through large samples in the genome-wide association study (GWAS). To our knowledge, there are relatively few studies on the causal relationship between stroke and gastrointestinal disorders, and gastrointestinal disorders have received less attention than other stroke complications, yet gastrointestinal disorders after stroke may lead to poor prognosis or even death [26]. PUD, GERD, irritable bowel syndrome (IBS), and inflammatory bowel disease (IBD) are common diseases of the digestive system [27, 28]. Therefore, we are committed to studying the causal effects of stroke and its subtypes and common gastrointestinal disorders by applying two-sample Mendelian randomization.
## 2. Material and methods
The conceptual MR framework is presented in Figure 1.
**Figure 1:** *Conceptual MR framework.*
## 2.1. Study design and ethical approval
According to the Strengthening the Reporting of Observational Studies in Epidemiology-Mendelian Randomization (STROBE-MR) recommendations [29], the MR design was based on three hypotheses: [1] in this investigation, genetic variation was highly linked with the exposure of interest (stroke and its subtypes); [2] genetic variation was not associated with possible confounders; and [3] genetic variation solely had an impact on the outcome through the exposure of interest (gastrointestinal disorders in this study).
## 2.2. Data sources for stroke and gastrointestinal disorders
To investigate the potential causative relationship between stroke and gastrointestinal disorders such as PUD, GERD, IBS, and IBD, we used a two-sample MR method. The largest meta-analysis of genome-wide association studies (GWASs) produced by the MEGASTROKE consortium provided pooled statistics for any stroke, any ischemic stroke, and its subtypes (cardioembolic stroke, small vessel stroke, and large artery stroke) confirmed by clinical and imaging criteria [30]. The International Stroke Genetics Consortium (ISGC), a group with European roots, provided the exposure dataset for hemorrhagic stroke (Table 1) [31]. Regarding the outcome dataset, we selected the results according to Wu et al. [ 28]. PUD, GERD, IBS, and IBD are common gastrointestinal diseases.
**Table 1**
| Phenotype | Data source | Sample size | %European |
| --- | --- | --- | --- |
| Exposures | Exposures | Exposures | Exposures |
| Any stroke | MEGASTROKE (30) | 40,585 cases/406, 111 controls | 100% |
| Any ischemic stroke | MEGASTROKE (30) | 34,217 cases/406,111 control | 100% |
| Cardioembolic stroke | MEGASTROKE (30) | 7,193 cases/406,111 control | 100% |
| Small vessel stroke | MEGASTROKE (30) | 5,386 cases/406,111 control | 100% |
| Large artery stroke | MEGASTROKE (30) | 4,373 cases/406,111 control | 100% |
| All ICH | ISGC (31) | 1,545 cases/1,481 controls | 100% |
| Deep ICH | ISGC (31) | 664 cases/1,481 controls | 100% |
| Lobar ICH | ISGC (31) | 881 cases/1,481 controls | 100% |
| Outcomes | Outcomes | Outcomes | Outcomes |
| PUD | Wu et al. (28) | 16,666 cases/406, 111 controls | 100% |
| GERD | Wu et al. (28) | 54,854 cases/401,473 controls | 100% |
| IBS | Wu et al. (28) | 29,524 cases/426,803 controls | 100% |
| IBD | Wu et al. (28) | 7,045 cases/449,282 controls | 100% |
## 2.3. Selection of genetic instruments
First, in line with the findings of Kwok et al. [ 32], we relaxed the correlation threshold with $P \leq 5$ × 10−6 and linkage disequilibrium (LD) (r2 < 0.001) to obtain the top independent SNPs. This was done in light of the small number of SNPs ($P \leq 5$ × 10−8) that reached genome-wide significance. This strategy has been applied extensively in earlier MR investigations [33, 34]. Second, the results of MR analysis are believed to be unaffected by weak instrumental bias if there is an F-statistic larger than 10. We used the following: Third, we extracted the secondary phenotypes of each SNP from a PhenoScanner V2 [35] and the GWAS library to exclude any putative polymorphism effects. The radial MR and MR pleiotropy residual sum and outlier (MR-PRESSO) tests were used to eliminate outliers before each MR analysis.
## 2.4. Statistical analysis
Three methods, including MR-Egger, weighted median, and random effect inverse-variance weighting (IVW), were utilized in the MR analysis to evaluate robust effects. The primary analysis method was the IVW method with various models, depending on the heterogeneity. At least half of the data for the Mendelian randomization study must originate from reliable instruments to use the weighted median estimator [36, 37]. The effectiveness of potential pleiotropic tools must be independent of their direct relationships with the outcome for MR-Egger regression to be valid. Radial MR-Egger was used to estimate the horizontal pleiotropy and to identify outlier variants [38]. Heterogeneity was also assessed using Cochran's Q-test. With the Cochran Q test (statistics were deemed to be significant if $P \leq 0.05$) and the intercept from MR-Egger regression (statistics were deemed to be significant if $P \leq 0.05$), we evaluated heterogeneity between Mendelian randomization estimates. We also evaluated potential directional polymorphisms using funnel plots. We used fixed-effects IVW and limited our instrument selection for sensitivity analyses to a lower LD correlation threshold. In conclusion, we conducted a thorough investigation of causation using all these techniques. Given the 32 MR estimates, the Bonferroni-corrected P-value for the study of gastrointestinal disorders was set at $\frac{0.05}{32}$ (1.563 × 10−3), and $P \leq 0.05$ was regarded as nominally significant. The statistical study was performed using R (version 4.2.0) and the “TwoSampleMR” and “RadialMR” packages.
## 3. Results
The SNPs of stroke subtypes on gastrointestinal disorders are listed in Supplementary Tables 1–8. Looking over the Phenoscanner, three SNPs (rs10850001, rs10774624, and rs3184504) were associated with smoking and were removed when analyzing PUD-associated SNPs. A total of 10 SNPs (rs12932445, rs1537375, rs2107595, rs2466455, rs4444878, rs4932370, rs6536024, rs6838973, rs72700114, and rs2634074) were related to the anticoagulant use, which was analyzed for PUD-related SNPs removed during the analysis. A total of 10 SNPs (rs10774624, rs1549758, rs1975161, rs2107595, rs2284665, rs34416434, rs42039, rs616154, rs78893982, and rs8103309) were associated with obesity and were removed in the analysis of GERD-related SNPs.
We performed a comprehensive MR study of stroke and its subtypes on gastrointestinal diseases (Supplementary Table 9). Among them, using IVW as the primary analysis, it could be seen that genetics predicted that any ischemic stroke had a normal significance with GERD ($P \leq 0.05$). All ICHs had normal significance with PUD and IBD ($P \leq 0.05$). Meanwhile, deep ICH had signed with the PUD and GERD ($P \leq 1.563$ × 10−3). Lobar ICH had signed with the PUD and IBS ($P \leq 1.563$ × 10−3). A bubble plot was used to show the statistical significance of the analysis (Figure 2). After that, the MR analyses with significant P-values were demonstrated in a forest plot (Figure 3). For ischemic stroke, there was no significant causal relationship with gastrointestinal disorders. For hemorrhagic stroke, the result of IVW showed that deep ICH [odds ratio (OR): 1.020; $95\%$ confidence interval (CI): 1.010–1.030; $$P \leq 4.740$$ × 10−5] was associated with an increased risk of PUD and greater disease severity with the weight median method (OR: 1.020; $95\%$ CI: 1.010–1.030; $$P \leq 4.740$$ × 10−5). The results of the MR-Egger method showed consistent directions but were not statistically significant (OR: 0.997; $95\%$ CI: 0.905–1.099; $$P \leq 0.954$$). In addition, similar causal estimates of lobar ICH on PUD were obtained, and IVW (OR: 1.026; $95\%$ CI: 1.016–1.037; $$P \leq 9.018$$ × 10−7) and weight median (OR: 1.042; $95\%$ CI: 1.027–1.057; $$P \leq 4.077$$ × 10−8) were included, while the same result was observed using the MR-Egger method but without any statistical difference (OR: 1.008, $95\%$ CI: 0.968–1.049, $$P \leq 0.700$$). Deep ICH was associated with an increased risk of GERD with the IVW (OR: 1.028; CI: 1.022–1.034; $$P \leq 1.663$$ × 10−21) and weight median (OR: 1.032; CI: 1.024–1.030; $$P \leq 9.994$$ × 10−17); however, there was no statistical difference in the MR-Egger method (OR: 1.036; CI: 0.971–1.106; $$P \leq 0.290$$), where all $p \leq 0.05$ for the MR-Egger intercept test, except for the MR analysis of lobar ICH on the IBS of lingual without weight median, indicated no horizontal pleiotropy. For significance and nominal significance estimates, Cochran's Q-test, the MR-Egger intercept test, the leave-one-out analysis, and the funnel plot were used to assess horizontal multiplicity (Supplementary Figures 1–4). Finally, we determined that deep ICH and labor ICH were causally related to PUD, and deep ICH was causally related to GERD.
**Figure 2:** *Bubble plot of MR study of stroke on gastrointestinal disorders derived from IVW.* **Figure 3:** *Forest plot for the causal effect of stroke on the risk of gastrointestinal disorders. OR, odds ratio; CI, confidence interval.*
## 4. Discussion
Previous studies have not found a clear causal relationship between stroke and gastrointestinal diseases. In our study, the relationship between stroke and its subtypes of gastrointestinal disorders was determined by the MR analysis. It is reported that obesity, smoking, anticoagulant, and other risk factors are often related to gastrointestinal diseases (39–41). The GWAS of GERD and PUD found genetic overlapping with the identified aforementioned hazardous factors [42, 43]. We cannot rule out that SNP affects the outcome through other related variables. Therefore, we should try our best to reduce the bias caused by pleiotropy. To reduce pleiotropy, we look over the PhenoScanner and eliminate those pleiotropic genetic variants. Thus, we successfully removed SNPs that were highly correlated with possible confounders such as obesity, smoking, and anticoagulation therapy. In addition, we also conducted some sensitivity analyses, such as a leave-one-out analysis and a funnel plot, and other methods such as Cochran's Q-test and the MR-Egger intercept test to assess horizontal multiplicity.
Stroke is often associated with PUD. A retrospective review including 808 cases found that the incidence of gastrointestinal bleeding caused by PUD in patients with ICH was $26.7\%$ [18]. Moreover, the incidence of gastrointestinal bleeding was significantly higher in patients who often use stress ulcer prophylaxis (SUP) for stress ulcer prevention compared with patients not receiving SUP [18]. Another observational study examined 177 patients with acute stroke by gastroscopy, of which 92 ($52\%$) had gastric changes, 10 of which were acute ulcers [44]. For patients with severe ICH, an observational study found that $28.0\%$ of 715 patients with severe ICH developed stress-related gastrointestinal bleeding (SGIB) or stress ulcers during hospitalization [45]. Regrettably, none of these observational studies had a large sample size. Our MR study suggested that stroke has a causal impact on PUD but only on deep ICH and lobar ICH and not ischemic stroke.
The pathogenesis of ICH complicated by peptic ulcers is still unclear and may be related to the damage to the thalamus and the subthalamus. To summarize various studies, the possible mechanisms are as follows: first, patients with acute ICH often experience intracranial hypertension and cerebral edema, which directly or indirectly causes damage to the brain stem, the hypothalamus, and other parts and finally affects their normal physiologic functions, leading to a dysfunction of the autonomic nervous system and gastric hyperchlorhydria. It lessens the blood flow of gastrointestinal mucosa and damages the gastric mucosal barrier, resulting in stress gastrointestinal ulcer peptic ulcers as well as peptic ulcer bleeding [46]. According to previous studies, the development of stress ulcers in patients with ICH can be better predicted by the hematoma volume of ICH [47, 48]. Mechanistically, larger hematomas in the case of cerebral hemorrhage are more likely to lead to increased intracranial pressure [45]. As mentioned earlier, elevated intracranial pressure may cause strong sympathetic excitation and gastrointestinal vasoconstriction, causing a decrease in gastrointestinal blood flow, which subsequently leads to mucosal ischemia and increased gastric acid secretion. Second, post-stroke sepsis plays a very important role in the development of stress ulcers induced by severe ICH [45]. Inflammatory cytokines are released in large amounts in the development of sepsis, thus exacerbating the ischemia of the gastrointestinal mucosa caused by intracerebral hemorrhage and driving the development of stress ulcers [49, 50]. In an observational study, the incidence of gastrointestinal bleeding in patients with ischemic stroke was $7.8\%$, $74\%$ of which were caused by peptic ulcers [51]. Combined with our findings, it is clear that hemorrhagic strokes are more likely to develop peptic ulcers than ischemic strokes. The development of peptic ulcers in ischemic stroke may be associated with vagal hyperactivity, stress, and neuroendocrine dysregulation [51, 52]. However, the trigger for gastrointestinal bleeding in most patients with ischemic stroke is not stress, and stress ulcers due to acute ischemic brain injury may be very rare [52, 53]. One possible explanation for the aforementioned results is that compared to ischemic strokes, hemorrhagic strokes are a more devastating subtype of stroke [54]. Hemorrhagic stroke may have a stronger effect on the brain–gut axis than ischemic stroke. The study finding that hemorrhagic stroke disrupts the gut microbiota more than ischemic stroke may prove this [55]. The incidence of stress ulcer bleeding in patients with brain injury is closely related to the severity of the injury [56]. In our MR study, the risk of lobar ICH is more associated with an increased risk of PUD compared to the risk of deep ICH. The size of the hematoma, sepsis, and prognosis have been reported to be the strongest predictors of gastrointestinal bleeding in patients with ICH in previous research [48]. A Japanese observational study found a higher rate of poor prognosis in patients with lobar ICH than in those with non-lobar ICH [57]. Even lobar ICH is associated with more severe cognitive impairment [58]. It suggests clinical vigilance for PUD for hemorrhagic stroke.
Several studies have proposed an association between stroke and GERD. A population-based Taiwanese cohort study including 18,412 patients with stroke and 18,412 without stroke found that the risk of GERD in patients with stroke was 1.51 times higher than that in patients without stroke ($95\%$ CI, 1.40–1.67) [15]. Moreover, they separated the stroke cohort into two subgroups: hemorrhagic stroke and ischemic stroke. Compared with the subjects without stroke, the HRs for GERD in the intracerebral hemorrhage and ischemic stroke cohorts were 1.45 and 1.52 ($95\%$ CI, 1.22–1.71 and $95\%$ CI, 1.39–1.67) [15]. Our MR study found that the higher risk of GERD is complicated by the risk of deep ICH, and there is a positive causal relationship between them. We reviewed the relevant literature to explain the mechanisms by which stroke leads to the development of GERD. For ischemic stroke, GERD may be induced by drugs used to treat IS, such as aspirin. One of the independent risk factors associated with the clinical symptoms of GERD is NSAIDs. The study also found an increased incidence of GERD in patients with stroke treated with antiplatelet therapy [15, 59]. In addition, ischemic stroke may disrupt the neural regulation of oropharyngeal, esophageal, and gastrointestinal motility, resulting in an extensive impairment of oropharyngeal and gastrointestinal motility and a reduced tone of the lower esophageal sphincter [52]. ICH has a similar effect on the vagus nerve, resulting in the malfunction of esophageal peristalsis, gastrointestinal motility, and the lower esophageal sphincter [48, 60]. Parasympathetic dysfunction in patients with stroke may lead to impaired esophageal motility, the abnormal transmission of food, and the abnormal relaxation of the lower esophageal sphincter [15, 61]. Hypertension is one of the most important risk factors for stroke, and treatment to lower blood pressure to prevent stroke, including the use of calcium channel blockers, often leads to lower esophageal sphincter (LES) pressure and eventually GERD [62, 63]. A community study found that calcium channel blockers were independently associated with GERD symptoms as a risk factor [63]. To explain why deep ICH is more prone to GERD than other subtypes of stroke, we looked through many studies. Deep ICH is often thought to be closely associated with hypertension, while lobar ICH is often thought to be caused by cerebral amyloid angiopathy (CAA) [64]. Among the drugs used to treat hypertension, calcium channel blockers have the effect of lowering the pressure of the LES and impeding gastric emptying, thus inducing GERD [65]. Therefore, compared to lobar ICH, deep ICH is more prone to GERD.
According to our MR results, intracerebral hemorrhage is more likely to cause gastrointestinal disease than ischemic stroke, and we are thinking about the reasons for this result. It is well-known that ischemic stroke and intracerebral hemorrhage do not occur by similar mechanisms, and their degree of criticality is different. ICH is the most severe subtype of stroke. Furthermore, the most devastating type of pathology among the subtypes of stroke is ICH [54]. *In* general, ICH produces more severe strokes than cerebral infarct [66, 67]. ICH typically manifests as elevated intracranial pressure, hematoma compression, and serious cerebral edema, which can cause many negative effects, such as neuroinflammation, mitochondrial dysfunction, and apoptosis, resulting in a sudden disruption of the blood–brain barrier [68]. Contrary to ICH, the structural stability of brain cells and the blood–brain barrier is retained for a longer length of time following the beginning of symptoms in ischemic stroke [69]. One possible explanation for our findings is that compared to ischemic stroke, ICH is more damaging to the brain–gut axis, causing a more severe dysbiosis in the gut microbiota, abnormal gastrointestinal motor function, and impaired gastrointestinal motility, which leads to gastrointestinal disorders. Compared to patients with ischemic stroke, patients with ICH have more severe gut microbiota destruction [70]. ICH causes rapid damage to astrocytes and the blood–brain barrier in patients [69]. *Contemporary* genetics considers stroke not as a disease but as a syndrome. Stroke is an acute manifestation of a range of chronic cerebrovascular diseases [71]. Another possible explanation for our findings is that some subtypes of ischemic stroke present additional phenotypic dilemmas, such as the cardiogenic stroke subtype, whereas the phenotype of ICH is more uniform [71]. Moreover, genetic factors are important in the pathogenesis of ICH [72]. It is estimated that up to $44\%$ of cases of ICH are heritable, and possessing an ICH first-degree relative increases the risk of developing the condition by a factor of six [73].
To explain the causal relationship between hemorrhagic stroke and several gastrointestinal diseases, we have found several possible mechanisms. Intracerebral smoke can affect the function of the autonomic nervous system. Through the enteric nervous system, the extrinsic and autonomic nervous systems can regulate the motor, sensory, and secretory functions of the gastrointestinal tract. ICH affects gastrointestinal function in this way, mainly with motor dysfunction [74]. For example, strokes are often complicated by dysphagia, which may be due to cranial nerve involvement in the region of the vertebrobasilar artery [75]. This is one of the possible causes of stroke complicating gastrointestinal motility disorder-related disease. Moreover, the change in gut microbes caused by intracerebral hemorrhage may be one of the causes of some gastrointestinal diseases [68]. A prospective case–control study found that compared with the control group, the intestinal microbiota composition of both patients with ischemic stroke and patients with intracerebral hemorrhage changed [55]. More specifically, compared with the control group, the abundance of invasive aerobic bacterial genera (Enterococcus species and Escherichia/Shigella species) in all patients with stroke increased, while obligate anaerobic genera decreased [55]. The authors observed that the extent of gut microbiota destruction was positively associated with the severity of stroke. An intracerebral hemorrhage causes more severe disruption of the gut microbiota than an ischemic stroke [55]. The autonomic nervous system abnormally releases norepinephrine to the intestine, which may change the intestinal microbiota [23]. Another study found that the immune system of model mice is disturbed after intracerebral hemorrhage. Furthermore, the gut barrier function of model mice was impaired, and intestinal permeability increased [70]. In addition, experimental studies also found that inflammatory cytokines were upregulated in the intestine, malondialdehyde (MDA) levels were elevated, the superoxide (SOD) dismutase activity was reduced, severe intestinal mucosal damage and plasma endotoxin levels were elevated 2 h after intracerebral hemorrhage in model mice, and intestinal propulsion was reduced 12 h later, and these symptoms persisted for 7 days after the appearance of the above symptoms [76]. These suggest that intracerebral hemorrhage significantly increases inflammatory cytokine levels and myeloperoxidase activity, which, in turn, promotes an inflammatory response in the intestine, leading to gastrointestinal disorders associated with intestinal motility and barrier dysfunction. In contrast, elevated malondialdehyde levels and reduced superoxide dismutase also suggest that intracerebral hemorrhage induced excessive oxygen radical production in the intestine during ischemia-reperfusion. The pathological imbalance of the intestinal oxidative–antioxidant system may also be involved in the pathogenesis of gastrointestinal disorders after intracerebral hemorrhage [76]. In a word, intracerebral hemorrhage may lead to impaired communication between the brain and intestinal axis, which may directly result in gastrointestinal motility dysfunction or intestinal flora disorders. Although there are many studies on how the brain–gut axis interacts, the exact mechanism has not been clarified.
Our MR study has some strengths. First, compared with one-sample MR, our research has a larger sample size and higher statistical efficiency. Second, our research overcame the shortcomings of traditional causal inference. Since the alleles followed the principle of random assignment, we obtained results independent of the confounding factors and reversed causal associations found in traditional epidemiological studies. Furthermore, there is Cochran's Q-test, the MR-Egger intercept test, and sensitivity analysis to test the pleiotropy of instrumental variables, which enhances the reliability of the results. At the same time, our analysis has some limitations. First of all, the estimates mentioned in our MR study cannot be directly compared with those of other observational studies. Second, we have selected only four common gastrointestinal disorders, and it is unknown whether a stroke has a causal effect on other gastrointestinal disorders. Third, the dataset on which our study is primarily based includes only individuals of European ancestry and thus may not be applicable to other humans, which would make our findings not generalizable. Finally, because of the limitation of the number of SNPs, the p-value limits were adjusted in our article.
Our MR study provides evidence for a causal relationship between deep ICH on PUD and GERD and a causal relationship between lobar ICH on PUD, and our results add to the gap in observational studies in this regard and warrant further research for the prevention of gastrointestinal disorders after deep ICH and lobar ICH.
## 5. Conclusion
Our research supports a possible causal link between stroke and its subtypes and gastrointestinal disorders. Early gastrointestinal disease risk assessment and prevention in hemorrhagic stroke is crucial and could aid in the introduction of tailored treatment as soon as possible.
## 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
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
JS contributed to the methodology and wrote the manuscript. WC contributed to conceptualization and investigation. WY contributed to the funding, writing, reviewing, and editing. 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/fneur.2023.1131250/full#supplementary-material
## References
1. Foreman KJ, Marquez N, Dolgert A, Fukutaki K, Fullman N, McGaughey M. **Forecasting life expectancy, years of life lost, and all-cause and cause-specific mortality for 250 causes of death: reference and alternative scenarios for 2016-40 for 195 countries and territories**. *Lancet.* (2018) **392** 2052-90. DOI: 10.1016/S0140-6736(18)31694-5
2. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990-2019: a systematic analysis for the global burden of disease study 2019**. *Lancet.* (2020) **396** 1204-22. DOI: 10.1016/S0140-6736(20)30925-9
3. Petro M, Jaffer H, Yang J, Kabu S, Morris VB, Labhasetwar V. **Tissue plasminogen activator followed by antioxidant-loaded nanoparticle delivery promotes activation/mobilization of progenitor cells in infarcted rat brain**. *Biomaterials.* (2016) **81** 169-80. DOI: 10.1016/j.biomaterials.2015.12.009
4. Regenhardt RW, Das AS, Lo EH, Caplan LR. **Advances in understanding the pathophysiology of lacunar stroke: a review**. *JAMA Neurol.* (2018) **75** 1273-81. DOI: 10.1001/jamaneurol.2018.1073
5. Rost NS, Brodtmann A, Pase MP, van Veluw SJ, Biffi A, Duering M. **Post-Stroke cognitive impairment and dementia**. *Circ Res.* (2022) **130** 1252-71. DOI: 10.1161/CIRCRESAHA.122.319951
6. Pluta R, Januszewski S, Czuczwar SJ. **The role of gut microbiota in an ischemic stroke**. *Int J Mol Sci.* (2021) **22** 915. DOI: 10.3390/ijms22020915
7. 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** 2769-78. DOI: 10.1016/S0140-6736(17)32448-0
8. Sandhu DS, Fass R. **Current trends in the management of gastroesophageal reflux disease**. *Gut Liver.* (2018) **12** 7-16. DOI: 10.5009/gnl16615
9. Lanas A, Chan FKL. **Peptic ulcer disease**. *Lancet.* (2017) **390** 613-24. DOI: 10.1016/S0140-6736(16)32404-7
10. Defrees DN, Bailey J. **Irritable bowel syndrome: epidemiology, pathophysiology, diagnosis, and treatment**. *Prim Care.* (2017) **44** 655-71. DOI: 10.1016/j.pop.2017.07.009
11. Tache Y. **The peptidergic brain-gut axis: influence on gastric ulcer formation**. *Chronobiol Int.* (1987) **4** 11-7. DOI: 10.1080/07420528709078504
12. Ancona A, Petito C, Iavarone I, Petito V, Galasso L, Leonetti A. **The gut-brain axis in irritable bowel syndrome and inflammatory bowel disease**. *Dig Liver Dis.* (2021) **53** 298-305. DOI: 10.1016/j.dld.2020.11.026
13. Yadlapati R, Gyawali CP, Pandolfino JE. **Aga clinical practice update on the personalized approach to the evaluation and management of gerd: expert review**. *Clin Gastroenterol Hepatol.* (2022) **20** 984-94.e1. DOI: 10.1016/j.cgh.2022.01.025
14. Xu Z, Wang H, Lin Y, Zhai Q, Sun W, Wang Z. **The impacts of peptic ulcer on functional outcomes of ischemic stroke**. *J Stroke Cerebrovasc Dis.* (2019) **28** 311-6. DOI: 10.1016/j.jstrokecerebrovasdis.2018.09.056
15. Chang CS, Chen HJ, Liao CH. **Patients with cerebral stroke have an increased risk of gastroesophageal reflux disease: a population-based cohort study**. *J Stroke Cerebrovasc Dis.* (2018) **27** 1267-74. DOI: 10.1016/j.jstrokecerebrovasdis.2017.12.001
16. Satou Y, Oguro H, Murakami Y, Onoda K, Mitaki S, Hamada C. **Gastroesophageal reflux during enteral feeding in stroke patients: a 24-hour esophageal ph-monitoring study**. *J Stroke Cerebrovasc Dis.* (2013) **22** 185-9. DOI: 10.1016/j.jstrokecerebrovasdis.2011.07.008
17. Kristensen SL, Lindhardsen J, Ahlehoff O, Erichsen R, Lamberts M, Khalid U. **Increased risk of atrial fibrillation and stroke during active stages of inflammatory bowel disease: a nationwide study**. *Europace.* (2014) **16** 477-84. DOI: 10.1093/europace/eut312
18. Yang TC, Li JG, Shi HM, Yu DM, Shan K, Li LX. **Gastrointestinal bleeding after intracerebral hemorrhage: a retrospective review of 808 cases**. *Am J Med Sci.* (2013) **346** 279-82. DOI: 10.1097/MAJ.0b013e318271a621
19. Wen SW, Wong CHY. **An unexplored brain-gut microbiota axis in stroke**. *Gut Microbes.* (2017) **8** 601-6. DOI: 10.1080/19490976.2017.1344809
20. Stanley D, Moore RJ, Wong CHY. **An insight into intestinal mucosal microbiota disruption after stroke**. *Sci Rep.* (2018) **8** 568. DOI: 10.1038/s41598-017-18904-8
21. Arya AK, Hu B. **Brain-Gut axis after stroke**. *Brain Circ.* (2018) **4** 165-73. DOI: 10.4103/bc.bc_32_18
22. Yang Z, Wei F, Zhang B, Luo Y, Xing X, Wang M. **Cellular immune signal exchange from ischemic stroke to intestinal lesions through brain-gut axis**. *Front Immunol.* (2022) **13** 688619. DOI: 10.3389/fimmu.2022.688619
23. Houlden A, Goldrick M, Brough D, Vizi ES, Lénárt N, Martinecz B. **Brain injury induces specific changes in the caecal microbiota of mice via altered autonomic activity and mucoprotein production**. *Brain Behav Immun.* (2016) **57** 10-20. DOI: 10.1016/j.bbi.2016.04.003
24. Holmes MV, Ala-Korpela M, Smith GD. **Mendelian randomization in cardiometabolic disease: challenges in evaluating causality**. *Nat Rev Cardiol.* (2017) **14** 577-90. DOI: 10.1038/nrcardio.2017.78
25. Cui G, Li S, Ye H, Yang Y, Huang Q, Chu Y. **Are neurodegenerative diseases associated with an increased risk of inflammatory bowel disease? A two-sample mendelian randomization study**. *Front Immunol.* (2022) **13** 956005. DOI: 10.3389/fimmu.2022.956005
26. Wang WJ, Lu JJ, Wang YJ, Wang CX, Wang YL, Hoff K. **Clinical characteristics, management, and functional outcomes in chinese patients within the first year after intracerebral hemorrhage: analysis from china national stroke registry**. *CNS Neurosci Ther.* (2012) **18** 773-80. DOI: 10.1111/j.1755-5949.2012.00367.x
27. Freuer D, Linseisen J, Meisinger C. **Asthma and the risk of gastrointestinal disorders: a mendelian randomization study**. *BMC Med.* (2022) **20** 82. DOI: 10.1186/s12916-022-02283-7
28. Wu Y, Murray GK, Byrne EM, Sidorenko J, Visscher PM, Wray NR. **Gwas of peptic ulcer disease implicates helicobacter pylori infection, other gastrointestinal disorders and depression**. *Nat Commun.* (2021) **12** 1146. DOI: 10.1038/s41467-021-21280-7
29. Skrivankova VW, Richmond RC, Woolf BAR, Yarmolinsky J, Davies NM, Swanson SA. **Strengthening the reporting of observational studies in epidemiology using mendelian randomization: the strobe-Mr statement**. *Jama.* (2021) **326** 1614-21. DOI: 10.1001/jama.2021.18236
30. Malik R, Chauhan G, Traylor M, Sargurupremraj M, Okada Y, Mishra A. **Publisher correction: multiancestry genome-wide association study of 520,000 subjects identifies 32 loci associated with stroke and stroke subtypes**. *Nat Genet.* (2019) **51** 1192-3. DOI: 10.1038/s41588-019-0449-0
31. Woo D, Falcone GJ, Devan WJ, Brown WM, Biffi A, Howard TD. **Meta-Analysis of genome-wide association studies identifies 1q22 as a susceptibility locus for intracerebral hemorrhage**. *Am J Hum Genet.* (2014) **94** 511-21. DOI: 10.1016/j.ajhg.2014.02.012
32. Kwok MK, Kawachi I, Rehkopf D, Schooling CM. **The role of cortisol in ischemic heart disease, ischemic stroke, type 2 diabetes, and cardiovascular disease risk factors: a bi-directional mendelian randomization study**. *BMC Med.* (2020) **18** 363. DOI: 10.1186/s12916-020-01831-3
33. Du W, Wang T, Zhang W, Xiao Y, Wang X. **Genetically supported causality between benign prostate hyperplasia and urinary bladder neoplasms: a mendelian randomization study**. *Front Genet.* (2022) **13** 1016696. DOI: 10.3389/fgene.2022.1016696
34. Müller R, Freitag-Wolf S, Weiner J, Chopra A, Top T, Dommisch H. **Case-only design identifies interactions of genetic risk variants at siglec5 and plg with the lncrna Ctd-2353f22.1 implying the importance of periodontal wound healing for disease aetiology**. *J Clin Periodontol.* (2023) **50** 90-101. DOI: 10.1111/jcpe.13712
35. Kamat MA, Blackshaw JA, Young R, Surendran P, Burgess S, Danesh J. **Phenoscanner V2: an expanded tool for searching human genotype-phenotype associations**. *Bioinformatics.* (2019) **35** 4851-3. DOI: 10.1093/bioinformatics/btz469
36. Bowden J, Davey Smith G, Burgess S. **Mendelian randomization with invalid instruments: effect estimation and bias detection through egger regression**. *Int J Epidemiol.* (2015) **44** 512-25. DOI: 10.1093/ije/dyv080
37. Burgess S, Bowden J, Fall T, Ingelsson E, Thompson SG. **Sensitivity analyses for robust causal inference from mendelian randomization analyses with multiple genetic variants**. *Epidemiology.* (2017) **28** 30-42. DOI: 10.1097/EDE.0000000000000559
38. Bowden J, Spiller W, Del Greco MF, Sheehan N, Thompson J, Minelli C. **Improving the visualization, interpretation and analysis of two-sample summary data mendelian randomization via the radial plot and radial regression**. *Int J Epidemiol.* (2018) **47** 1264-78. DOI: 10.1093/ije/dyy101
39. Bilski J, Mazur-Bialy A, Wojcik D, Surmiak M, Magierowski M, Sliwowski Z. **Role of obesity, mesenteric adipose tissue, and adipokines in inflammatory bowel diseases**. *Biomolecules.* (2019) **9** 780. DOI: 10.3390/biom9120780
40. Korman MG, Hansky J, Eaves ER, Schmidt GT. **Influence of cigarette smoking on healing and relapse in duodenal ulcer disease**. *Gastroenterology.* (1983) **85** 871-4. DOI: 10.1016/0016-5085(83)90438-9
41. Kawamura N, Ito Y, Sasaki M, Iida A, Mizuno M, Ogasawara N. **Low-Dose aspirin-associated upper gastric and duodenal ulcers in japanese patients with no previous history of peptic ulcers**. *BMC Res Notes.* (2013) **6** 455. DOI: 10.1186/1756-0500-6-455
42. Ong JS, An J, Han X, Law MH, Nandakumar P, Schumacher J. **Multitrait genetic association analysis identifies 50 new risk loci for gastro-oesophageal reflux, seven new loci for barrett's oesophagus and provides insights into clinical heterogeneity in reflux diagnosis**. *Gut.* (2022) **71** 1053-61. DOI: 10.1136/gutjnl-2020-323906
43. Li Z, Chen H, Chen T. **Genetic liability to obesity and peptic ulcer disease: a mendelian randomization study**. *BMC Med Genomics.* (2022) **15** 209. DOI: 10.1186/s12920-022-01366-x
44. Kitamura T, Ito K. **Acute gastric changes in patients with acute stroke. Part 1: with reference to gastroendoscopic findings**. *Stroke.* (1976) **7** 460-3. DOI: 10.1161/01.STR.7.5.460
45. Liu S, Wang Y, Gao B, Peng J. **A nomogram for individualized prediction of stress-related gastrointestinal bleeding in critically ill patients with primary intracerebral hemorrhage**. *Neuropsychiatr Dis Treat.* (2022) **18** 221-9. DOI: 10.2147/NDT.S342861
46. Alhazzani W, Jaeschke R. **Stress ulcer prophylaxis in critical care: a 2016 perspective Dr. Waleed Alhazzani in an interview with Dr. Roman Jaeschke: part 2**. *Pol Arch Med Wewn.* (2016) **126** 796-7. DOI: 10.20452/pamw.3606
47. Liu BL, Li B, Zhang X, Fei Z, Hu SJ, Lin W. **A randomized controlled study comparing omeprazole and cimetidine for the prophylaxis of stress-related upper gastrointestinal bleeding in patients with intracerebral hemorrhage**. *J Neurosurg.* (2013) **118** 115-20. DOI: 10.3171/2012.9.JNS12170
48. Misra UK, Kalita J, Pandey S, Mandal SK. **Predictors of gastrointestinal bleeding in acute intracerebral haemorrhage**. *J Neurol Sci.* (2003) **208** 25-9. DOI: 10.1016/S0022-510X(02)00415-X
49. Pastores SM, Katz DP, Kvetan V. **Splanchnic ischemia and gut mucosal injury in sepsis and the multiple organ dysfunction syndrome**. *Am J Gastroenterol.* (1996) **91** 1697-710. PMID: 8792684
50. Overhaus M, Tögel S, Pezzone MA, Bauer AJ. **Mechanisms of polymicrobial sepsis-induced ileus**. *Am J Physiol Gastrointest Liver Physiol.* (2004) **287** G685-94. DOI: 10.1152/ajpgi.00359.2003
51. Hsu HL, Lin YH, Huang YC, Weng HH, Lee M, Huang WY. **Gastrointestinal hemorrhage after acute ischemic stroke and its risk factors in Asians**. *Eur Neurol.* (2009) **62** 212-8. DOI: 10.1159/000229018
52. Schaller BJ, Graf R, Jacobs AH. **Pathophysiological changes of the gastrointestinal tract in ischemic stroke**. *Am J Gastroenterol.* (2006) **101** 1655-65. DOI: 10.1111/j.1572-0241.2006.00540.x
53. Wijdicks EF, Fulgham JR, Batts KP. **Gastrointestinal bleeding in stroke**. *Stroke.* (1994) **25** 2146-8. DOI: 10.1161/01.STR.25.11.2146
54. Tsai CF, Jeng JS, Anderson N, Sudlow CLM. **Comparisons of risk factors for intracerebral hemorrhage versus ischemic stroke in chinese patients**. *Neuroepidemiology.* (2017) **48** 72-8. DOI: 10.1159/000475667
55. Haak BW, Westendorp WF, van Engelen TSR, Brands X, Brouwer MC, Vermeij JD. **Disruptions of anaerobic gut bacteria are associated with stroke and post-stroke infection: a prospective case-control study**. *Transl Stroke Res.* (2021) **12** 581-92. DOI: 10.1007/s12975-020-00863-4
56. Kamada T, Fusamoto H, Kawano S, Noguchi M, Hiramatsu K. **Gastrointestinal bleeding following head injury: a clinical study of 433 cases**. *J Trauma.* (1977) **17** 44-7. DOI: 10.1097/00005373-197701000-00006
57. Matsukawa H, Shinoda M, Fujii M, Takahashi O, Yamamoto D, Murakata A. **Factors associated with lobar vs. non-lobar intracerebral hemorrhage**. *Acta Neurol Scand.* (2012) **126** 116-21. DOI: 10.1111/j.1600-0404.2011.01615.x
58. Tveiten A, Ljøstad U, Mygland Å, Naess H. **Functioning of long-term survivors of first-ever intracerebral hemorrhage**. *Acta Neurol Scand.* (2014) **129** 269-75. DOI: 10.1111/ane.12185
59. Ercelep OB, Caglar E, Dobrucali A. **The prevalence of gastroesophageal reflux disease among hospital employees**. *Dis Esophagus.* (2014) **27** 403-8. DOI: 10.1111/j.1442-2050.2012.01402.x
60. Cunningham KM, Horowitz M, Riddell PS, Maddern GJ, Myers JC, Holloway RH. **Relations among autonomic nerve dysfunction, oesophageal motility, and gastric emptying in gastro-oesophageal reflux disease**. *Gut.* (1991) **32** 1436-40. DOI: 10.1136/gut.32.12.1436
61. Orlando RC. **Pathophysiology of gastroesophageal reflux disease**. *J Clin Gastroenterol.* (2008) **42** 584-8. DOI: 10.1097/MCG.0b013e31815d0628
62. Schrader J, Lüders S, Kulschewski A, Hammersen F, Plate K, Berger J. **Morbidity and mortality after stroke, eprosartan compared with nitrendipine for secondary prevention: principal results of a prospective randomized controlled study (moses)**. *Stroke.* (2005) **36** 1218-26. DOI: 10.1161/01.STR.0000166048.35740.a9
63. Mohammed I, Nightingale P, Trudgill NJ. **Risk factors for gastro-oesophageal reflux disease symptoms: a community study**. *Aliment Pharmacol Ther.* (2005) **21** 821-7. DOI: 10.1111/j.1365-2036.2005.02426.x
64. Labovitz DL, Sacco RL. **Intracerebral hemorrhage: update**. *Curr Opin Neurol.* (2001) **14** 103-8. DOI: 10.1097/00019052-200102000-00016
65. Osadchuk AM, Davydkin IL, Gricenko TA, Osadchuk MA. *Ter Arkh.* (2019) **91** 135-40. DOI: 10.26442/00403660.2019.08.000228
66. Jørgensen HS, Nakayama H, Raaschou HO, Olsen TS. **Intracerebral hemorrhage versus infarction: stroke severity, risk factors, and prognosis**. *Ann Neurol.* (1995) **38** 45-50. DOI: 10.1002/ana.410380110
67. Ratha Krishnan R, Yeo EQY, Lim CJ, Chua KSG. **The impact of stroke subtype on recovery and functional outcome after inpatient rehabilitation: a retrospective analysis of factors**. *Life.* (2022) **12** 1295. DOI: 10.3390/life12091295
68. Zou X, Wang L, Xiao L, Wang S, Zhang L. **Gut microbes in cerebrovascular diseases: gut flora imbalance, potential impact mechanisms and promising treatment strategies**. *Front Immunol.* (2022) **13** 975921. DOI: 10.3389/fimmu.2022.975921
69. Zhang J, Zhang CH, Lin XL, Zhang Q, Wang J, Shi SL. **Serum glial fibrillary acidic protein as a biomarker for differentiating intracerebral hemorrhage and ischemic stroke in patients with symptoms of acute stroke: a systematic review and meta-analysis**. *Neurol Sci.* (2013) **34** 1887-92. DOI: 10.1007/s10072-013-1541-3
70. Zhang H, Huang Y, Li X, Han X, Hu J, Wang B. **Dynamic process of secondary pulmonary infection in mice with intracerebral hemorrhage**. *Front Immunol.* (2021) **12** 767155. DOI: 10.3389/fimmu.2021.767155
71. Rost NS. **Clinical neurogenetics: stroke**. *Neurol Clin.* (2013) **31** 915-28. DOI: 10.1016/j.ncl.2013.05.001
72. Ekkert A, Šliachtenko A, Utkus A, JatuŽis D. **Intracerebral hemorrhage genetics**. *Genes.* (2022) **13** 1250. DOI: 10.3390/genes13071250
73. Hostettler IC, Seiffge DJ, Werring DJ. **Intracerebral hemorrhage: an update on diagnosis and treatment**. *Expert Rev Neurother.* (2019) **19** 679-94. DOI: 10.1080/14737175.2019.1623671
74. Camilleri M. **Gastrointestinal motility disorders in neurologic disease**. *J Clin Invest.* (2021) **131** e143771. DOI: 10.1172/JCI143771
75. Martino R, Foley N, Bhogal S, Diamant N, Speechley M, Teasell R. **Dysphagia after stroke: incidence, diagnosis, and pulmonary complications**. *Stroke.* (2005) **36** 2756-63. DOI: 10.1161/01.STR.0000190056.76543.eb
76. Cheng Y, Zan J, Song Y, Yang G, Shang H, Zhao W. **Evaluation of intestinal injury, inflammatory response and oxidative stress following intracerebral hemorrhage in mice**. *Int J Mol Med.* (2018) **42** 2120-8. DOI: 10.3892/ijmm.2018.3755
|
---
title: Effects of ambient particulate exposure on blood lipid levels in hypertension
inpatients
authors:
- Yanfang Gao
- Chenwei Li
- Lei Huang
- Kun Huang
- Miao Guo
- Xingye Zhou
- Xiaokang Zhang
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9989317
doi: 10.3389/fpubh.2023.1106852
license: CC BY 4.0
---
# Effects of ambient particulate exposure on blood lipid levels in hypertension inpatients
## Abstract
### Background
With modernization development, multiple studies of atmospheric particulate matter exposure conducted in China have confirmed adverse cardiovascular health effects. However, there are few studies on the effect of particulate matter on blood lipid levels in patients with cardiovascular disease, especially in southern China. The purpose of this study was to investigate the association between short- and long-term exposure to ambient particulate matter and the levels of blood lipid markers in hypertension inpatients in Ganzhou, China.
### Methods
Data on admission lipid index testing for hypertension inpatients which were divided into those with and without arteriosclerosis disease were extracted from the hospital's big data center from January 1, 2016 to December 31, 2020, and air pollution and meteorology data were acquired from the China urban air quality real time release platform from January 1, 2015 to December 31, 2020 and climatic data center from January 1, 2016 to December 31, 2020, with data integrated according to patient admission dates. A semi-parametric generalized additive model (GAM) was established to calculate the association between ambient particulate matter and blood lipid markers in hypertension inpatients with different exposure time in 1 year.
### Results
Long-term exposure to particulate matter was associated with increased Lp(a) in three kinds of people, and with increased TC and decreased HDL-C in total hypertension and hypertension with arteriosclerosis. But particulate matter was associated with increased HDL-C for hypertension inpatients without arteriosclerosis, at the time of exposure in the present study. It is speculated that hypertension inpatients without arteriosclerosis has better statement than hypertension inpatients with arteriosclerosis on human lipid metabolism.
### Conclusion
Long-term exposure to ambient particulate matter is associated with adverse lipid profile changes in hypertension inpatients, especially those with arteriosclerosis. Ambient particulate matter may increase the risk of arteriosclerotic events in hypertensive patients.
## 1. Introduction
Air pollution has become an environmental risk factor that seriously affects health, bringing a huge economic burden of disease and health loss to society [1, 2]. According to a Comment by the World Health Organization (WHO), the number of premature deaths caused by air pollution has exceeded 7 million each year [3]. Among air pollutants, ambient particulate matter (fine particulate matter with aerodynamic diameter <2.5 μm [PM2.5] and respirable particulate matter with aerodynamic diameter <10 μm [PM10]) is widely recognized as an important toxic component of air pollution mixtures [4]. Most studies have found that ambient particulate matter exposure can increase the incidence of cardiovascular disease events [5, 6], among which hypertension as a risk factor for most cardiovascular diseases has been confirmed by most studies to be related to the health hazard effects of ambient particulate matter exposure [7, 8].
Blood lipids are the general term for neutral fats (triglycerides) and lipids (phospholipids, glycolipids, sterols, steroids) in plasma [9]. Adverse alterations in lipid levels in humans can cause deposition of lipids in the vessel wall, which in turn triggers vascular stiffening they are also recognized as a risk factor for cardio - and cerebrovascular disease (10–12), and play an important role in chronic diseases such as hypertension, myocardial infarction, and ischemic stroke (13–16). It has been found that long-term exposure to ambient particulate matter can change the lipid level in the population [17, 18], but there is no consistency about the lipid effects of air pollution described by different studies [19, 20].
Ganzhou *City is* located in the southern of China. It has a typical humid subtropical monsoon climate. The precipitation is concentrated in spring and summer, the climate is mild, the heat and rainfall are abundant, the duration of cold and hot air flow is short, and the frost-free period is long. In recent years, Ganzhou's rare earth mining industry, tourism, furniture construction and other industries have developed rapidly, driving the steady growth of the city's economy. With the influx of population, the continuous increase in traffic flow, and the urbanization, the urban vitality of Ganzhou has been greatly improved, but it has also increased pollution emissions, posing challenges to air governance. As the economy develops, there are no studies on associations between air pollution exposure and blood lipids in Ganzhou. Therefore, this study collected ambient particulate matter exposure concentrations in Ganzhou City and blood lipid detection data of hypertensive inpatients in a tertiary hospital, and explored the impacts of ambient particulate matter pollution on blood lipid levels in hypertensive populations.
## 2.1. Blood lipid markers in hypertensive hospitalized patients
The hospitalization records of cardiovascular medicine from January 1, 2016 to December 31, 2020 were extracted from the big data center of a tertiary hospital (two campuses) in Ganzhou City, including the patient's admission date, main diagnosis, age, gender, and biochemical tests. According to the 10th revised International Classification of Diseases (ICD-10), whose code is Hypertension (ICD-10: I10-I13), we screened the information of patients with “hypertension” as their discharge diagnosis and matched them with data of blood lipid markers collected during hospitalization. Included lipid markers included triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and lipoprotein [Lp(a)]. Since LP(a) is skewed distribution data, we logarithmically transform LP(a). The hypertensive inpatients were divided into “hypertension with arteriosclerosis population” and “hypertension without arteriosclerosis population” according to whether other discharge diagnoses included arteriosclerosis. The sum of the two populations was called “total hypertension population.”
## 2.2. Air pollution and meteorology data of Ganzhou City
The air pollutant concentration data of Ganzhou City from January 1, 2015 to December 31, 2020 were obtained on the China Urban Air Quality Real-time Release Platform. Calculate and sort out the 24-h average concentrations of five air pollutants such as PM2.5, PM10, carbon monoxide (CO), Nitrogen dioxide (NO2), and sulfur dioxide (SO2) and the daily maximum 8-h average ozone (O3) concentration. Air pollution exposure concentrations were matched according to the patients' hospitalization dates to generate a total of 15 air pollution concentration data series including L0, L0-6, L0-13, L0-29, L0-59, L0-89, L0-119, L0-149, L0-179, L0-209, L0-239, L0-269, L0-299, L0-329, and L0-359. Where L0 is equivalent to the same day exposure concentration of the respective air pollutant and L0-6 are equivalent to the moving average exposure concentrations at a lag of 6 days between hospital admissions for patients with hypertension. Meteorological data comes from the National Climate Data Center, which collects indicators such as the daily average temperature (°C) and relative humidity (%) in Ganzhou from January 1, 2016 to December 31, 2020. The locations of air pollutant data monitoring sites and hospital admissions (two campuses) are shown in Figure 1.
**Figure 1:** *The locations of air pollutant data monitoring sites (green symbol) and hospital admissions (two campuses, red symbol).*
## 2.3. Statistical analysis
SPSS was used to describe the mean ± standard deviation (Mean ± SD) of blood lipid indexes in three groups of “hypertension with arteriosclerosis,” “hypertension without arteriosclerosis” and “total hypertension.” Age, gender and hospitalization date were used as covariates for covariance analysis to compare the differences in the mean concentrations of blood lipid indexes among the three groups. If differences were statistically significant, Bonferroni's method was used for multiple comparisons [21]. The central and discrete trends of air pollution and meteorological data were described, and the differences in air pollution concentrations between adjacent years were compared using the Kruskal-Wallis test [22]. Spearman correlation analysis was used to evaluate the associations between air pollutants and meteorological data [23].
R4.0.4 was used to establish a semi-parametric generalized additive model (GAM) [24, 25] to adjust the temperature, atmospheric relative humidity, age and gender of patients and balance the time trend to calculate the effect of changes in the unit concentration of air pollution on blood lipids in the study population. On this basis, choose the lag time period in which the ambient particulate matter has the strongest correlation with the blood lipid concentration of the population in the lag of 1 year to establish a dual-pollutant model, and observe the impacts of ambient particulate matter exposure on the blood lipid concentration of the three groups of people after adjusting for the concentration of gaseous pollutants. The partial regression coefficient and $95\%$ confidence interval of the effect of ambient particulate matter on the concentration of blood lipid markers were obtained, and converted into the change value of blood lipid concentration by exp[(β × 10)-1] × 100. That is, the corresponding change in blood lipid concentration for every 10 μg/m3 increase in PM2.5 or PM10 concentration, and its $95\%$ confidence interval.
## 3.1. Characteristics of lipid markers in study subjects
Among the hypertensive hospitalized patients, 2,269 cases of TC and TG were matched respectively, among which 685 cases had arteriosclerosis, 2,268 cases of HDL-C had 685 cases of arteriosclerosis, 2,266 cases of LDL-C had 685 cases of arteriosclerosis, 2,241 cases of Lp(a) of which 674 had arteriosclerosis. As shown in Table 1, the mean values of the five lipid markers in the total hypertensive population were: TG 1.74 mmol/l, TC 4.45 mmol/l, HDL-C 1.1 mmol/l, LDL-C 2.69 mmol/l, and LP (a) 1.81 μmol/L. There were no significant differences in lipid concentrations among the three groups by covariance analysis.
**Table 1**
| Unnamed: 0 | Total hypertension | Total hypertension.1 | Hypertension with arteriosclerosis | Hypertension with arteriosclerosis.1 | Hypertension without arteriosclerosis | Hypertension without arteriosclerosis.1 | P |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | N | Mean (SD) | N | Mean (SD) | N | Mean (SD) | |
| TG (mmol/L) | 2269 | 1.74 (1.53) | 685 | 1.66 (1.31) | 1584 | 1.78 (1.62) | 1.0 |
| TC (mmol/L) | 2269 | 4.45 (7.23) | 685 | 4.33 (1.10) | 1584 | 4.51 (8.63) | 0.846 |
| HDL-C (mmol/L) | 2268 | 1.1 (0.29) | 685 | 1.08 (0.29) | 1583 | 1.05 (0.29) | 0.754 |
| LDL-C (mmol/L) | 2266 | 2.69 (0.89) | 685 | 2.71 (0.91) | 1581 | 2.68 (0.88) | 0.147 |
| Lp(a) (μmol/L) | 2241 | 1.81 (0.62) | 674 | 1.83 (0.63) | 1567 | 1.81 (0.62) | 0.772 |
## 3.2. Characteristics of air pollution and meteorological data
Table 2 shows that the average concentrations of CO, NO2, O3, PM10, PM2.5, and SO2 are 1.27 mg/m3, 23.171 μg/m3, 70.80 μg/m3, 59.89 μg/m3, 38.17 μg/m3 and 20.94 μg/m3, respectively. The daily average temperature and relative humidity were 19.72°C and $74.37\%$, respectively. Figure 2 shows the change in the annual average concentration of air pollutants from 2015 to 2020 compared with the previous year. From 2018 to 2020, the annual average concentrations of PM2.5, PM10, NO2 and CO showed a clear downward trend year by year. The annual mean concentration of SO2 decreased significantly in 2018 and 2019. The annual average concentration of O3 dropped significantly in 2016 and again in 2017 was significantly higher than the previous year.
As shown in Table 3, positive correlations were observed among all air pollutants except O3, for which ambient particulate matter had stronger correlations with NO2 and SO2. There was a negative correlation between both air temperature and humidity and with ambient particulate matter, but positive correlations between air temperature and O3 or SO2.as well as humidity and CO were observed.
**Table 3**
| Unnamed: 0 | CO | NO2 | O3 | PM10 | PM2.5 | SO2 | Temperature |
| --- | --- | --- | --- | --- | --- | --- | --- |
| NO2 | 0.401** | | | | | | |
| O3 | −0.182** | −0.095** | | | | | |
| PM10 | 0.386** | 0.655** | 0.327** | | | | |
| PM2.5 | 0.445** | 0.618** | 0.257** | 0.958** | | | |
| SO2 | 0.292** | 0.510** | 0.220** | 0.705** | 0.687** | | |
| Temperature | −0.320** | −0.394** | 0.370** | −0.126** | −0.211** | 0.066** | |
| Humidity | 0.210* | −0.073** | −0.642** | −0.359** | −0.228** | −0.293** | −0.316** |
## 3.3. Blood lipids effects on hypertension inpatients with arteriosclerosis
As shown in Figure 3, it was observed in the hypertensive population with arteriosclerosis that PM2.5 and PM10 maintained a significant positive correlation with TC starting from lag0-59. At lag0-209, every 10 μg/m3 increase in PM2.5 and PM10 caused $41.8\%$ ($95\%$CI: 19.39, 68.42) and $25.25\%$ ($95\%$CI: 12.67, 39.23) increases in TC concentration, respectively. From lag 0-179, HDL-C maintained a long negative correlation with PM2.5 and PM10. At lag 0-239, ambient particulate matter had the most significant effect on HDL-C, and every 10μg/m3 increase in PM2.5 and PM10 caused $5.61\%$ ($95\%$CI: 1.76, 9.31) and 3.42 ($95\%$CI: 1.04, 5.74) decrease in HDL-C, respectively. The effects of PM2.5 and PM10 on LDL-C and TG in hospitalized patients with hypertension and arteriosclerosis were not observed with a lag of 359 days in the moving average of ambient particulate matter exposure. Significant positive correlations between PM2.5 and PM10 and Lp(a) were observed only at lag0-359. Lp(a) increased by $38.52\%$ ($95\%$CI: 5.09, 82.59) and $26.47\%$ ($95\%$CI: 6.03, 50.84) for each 10 μg/m3 increase in PM2.5 and PM10, respectively.
**Figure 3:** *Effect of PM2.5 and PM10 on blood lipid in hypertension inpatients with arteriosclerosis.*
## 3.4. Blood lipids effects on hypertension inpatients without arteriosclerosis
TC and TG in hypertension inpatients without arteriosclerosis were not associated with ambient particulate matter exposure during the study period. As show in Figure 4, HDL-C and LDL-C were both significantly positively correlated with ambient particulate matter at lag0-6, and HDL-C showed a positive correlation with ambient particulate matter again at lag0-89 to lag0-209. During the study period, PM2.5 was positively correlated with Lp(a) concentration in hypertensive population without arteriosclerosis only at lag 0-359. Lp(a) increased by $29.19\%$ ($95\%$CI: 8.75, 53.46) for every 10 μg/m3 increase in PM2.5. PM10 showed a positive correlation with Lp(a) concentrations in this population at various time periods, including lag0-6, lag0-59 to lag0-89, and lag0-299 days later. These results revealed that hypertension inpatients without arteriosclerosis has better statement than hypertension inpatients with arteriosclerosis on human lipid metabolism.
**Figure 4:** *Effect of PM2.5 and PM10 on blood lipid in hypertension inpatients without arteriosclerosis.*
## 3.5. Blood lipids effects on total hypertension inpatients
As show in Figure 5, PM2.5 and PM10 were significantly positively correlated with TC levels in total hypertensive patients at lag0-119 to lag0-239 and lag0-89 to lag0-239, respectively. For every 10 μg/m3 increase in PM2.5 and PM10 at lag 0-209, TC in total hypertensive patients increased by $16.77\%$ ($95\%$CI: 5.52, 9.22) and $10.54\%$ ($95\%$CI: 3.60, 17.94), respectively. PM2.5 and PM10 exposures were positively correlated with Lp(a) concentrations in hypertensive patients from lag0-29 to lag0-119, the correlation disappeared at lag0-149, and then reappeared from lag0-359 and lag0-229, respectively. Lp(a) concentrations in patients with a 10μg/m3 increase in PM2.5 and PM10 at lag0-359 increased by $31.41\%$ ($95\%$CI: 10.16, 56.76) and $22.25\%$ ($95\%$CI: 9.16, 36.91), respectively. PM2.5 and PM10 were negatively correlated with HDL-C levels in this population at lag0-239 to lag0-229 and lag0-209 to lag0-229, respectively. For each 10 μg/m3 increase in PM2.5 and PM10 at lag 0-239, HDL-C concentrations in patients decreased by $3.63\%$ ($95\%$CI: 1.71, 5.50) and $2.49\%$ ($95\%$CI: 1.26, 3.69), respectively. No correlation was observed between ambient particulate matter exposure and LDL-C and TG in the total hypertensive population.
**Figure 5:** *Effect of PM2.5 and PM10 on blood lipid in total hypertensive inpatients.*
## 3.6. Blood lipids effects after adjusting for gaseous pollutants
The positive correlation between ambient particulate matter and TC in the population with total hypertension and hypertension with arteriosclerosis remained unchanged after adjusting for gaseous pollutants, but it disappeared significantly after adjusting for NO2 concentration. The positive correlation between particulate matter and Lp(a) of the three groups of people remained unchanged after adjusting for the effects of gaseous pollutants, but the significant effects of particulate matter disappeared after adjusting for SO2, O3 and NO2. In the total hypertensive population, ambient particulate matter was negatively correlated with HDL-C, and the correlation changed positively after CO adjustment. The correlation between PM10 and HDL-C changed from positive to negative after adjusting for NO2 and SO2 in the hypertensive population without arteriosclerosis. The positive correlation between ambient particulate matter and LDL-C levels in hypertensive patients without arteriosclerosis was reversed after adjusting for CO and NO2, but there was not significant. The detailed results are shown in Figure 6. These results revealed that long-term exposure to ambient particulate matter is associated with adverse lipid profile changes in hypertension inpatients, especially those with arteriosclerosis
**Figure 6:** *Effect of ambient particulate matter on blood lipid after adjusting for gaseous pollutants. (A) Effect of ambient particulate matter on blood lipid after adjusting for gaseous pollutants in total hypertensive inpatients, (B) Effect of ambient particulate matter on blood lipid after adjusting for gaseous pollutants in hypertension inpatients with arteriosclerosis, (C) Effect of ambient particulate matter on blood lipid after adjusting for gaseous pollutants in hypertension inpatients without arteriosclerosis. +CO, +NO2, +O3, +SO2 two-pollutant model after adjusting for the concentration of CO, NO2, O3, and SO2 respectively.*
## 4. Discussion
In recent years, Ganzhou has stepped up its air pollution control actions. Since 2017, the concentration of ambient particulate matter has decreased significantly year by year. Ganzhou City belongs to southern China and has not established a unified heating system. However, in the cold winter, residents will generally have individual combustion or energy-consuming heating behaviors, resulting in an increase in the emission of ambient particulate matter and a negative correlation between the concentration of particulate matter and meteorological factors. In addition, the low temperatures in winter may cause poor diffusion of pollutants in the air, leading to widespread peaks of pollutants in the cold season [26]. Similar to most southern cities, the main sources of ambient particulate matter emissions in Ganzhou are construction dust and vehicle exhaust from urbanization. From 2015 to 2020, the average daily concentration of PM10 in Ganzhou was 59.89 μg/m3, which was higher than the first-level standard of the national ambient air quality standard (GB3095-2012), but lower than the second-level standard (the first-level standard was 40 μg/m3, and the second-level standard was 70 μg/m3). The average daily concentration of PM2.5 was 38.17 μg/m3, which was higher than the NAQS secondary standard (15 μg/m3 for the primary standard and 35 μg/m3 for the secondary standard).
Total cholesterol (TC) refers to the sum of cholesterol contained by all lipoproteins in the blood, including free cholesterol and cholesterol esters. Most of the cholesterol is mainly synthesized by the body itself and from food sources as a minor supplement. There are these lipid transfer proteins which are HDL-C, LDL-C, and very low-density lipoprotein cholesterol. High density lipoprotein (HDL) can uptake low-density lipoprotein, cholesterol, triglycerides, sedimented from the intimal lining of the vessel wall for excretion to the liver, which helps to resist the stiffening of blood vessels caused by hyperlipidemia [27] and maintain cardiovascular health (28–30). Low density lipoprotein (LDL) is a lipoprotein particle that carries cholesterol into tissue cells and carries cholesterol accumulation across the arterial wall to cause arteriosclerosis [31]. Lipoprotein [Lp (a)] is synthesized by the liver and is a specialized cholesterol rich macromolecular lipoprotein that promotes atherosclerosis [32].
This study found differences in the effects of ambient particulate matter on TC, HDL-C, LDL-C, and Lp (a) content among hypertension with or without arteriosclerosis. Combined with the analysis of biochemical associations among lipid markers to identify differences in the associations of each lipid marker with ambient particulate matter among different populations, we raise several conjectures.
Ambient particulate matter exposure exhibited health hazard effects of increasing TC levels and decreasing HDL-C levels for both the hypertensive with arteriosclerosis population and the total hypertensive population. However, in the hypertensive without arteriosclerosis population, there are no association with TC levels was present but increasing HDL-C levels emerged with ambient particulate matter exposure. There was a similar trend for Lp (a) levels to increase with higher ambient particulate matter concentrations in the three groups. A positive association with short-term particulate matter exposure was observed only in people with hypertension without arteriosclerosis. It is speculated that ambient particulate matter exposure mainly cause increased cholesterol and lipoprotein synthesis and decreased high-density lipoprotein synthesis in hypertensive patients. But in patients without arteriosclerosis, the regulatory mechanisms of lipid homeostasis are normal, so that lipid transfer proteins such as HDL-C and LDL-C were temporarily increases to maintain stable total cholesterol levels. When cholesterol synthesis continues increasing with long-term exposure, beneficial HDL-C is more producted to metabolize excess cholesterol. Meanwhile LDL-C is inhibited temporarily to protect blood vessels. The phenomenon of compensatory increases in HDL-C and LDL-C with ambient particulate matter exposure has thus emerged in hypertensive population without vascular stiffening.
Long-term exposure to air pollution was found to be associated with increased TC and LP (a) levels in the elderly, obese adolescents and cardiovascular disease (CVD) patients in other studies on associations between air pollution and lipid concentrations (33–35), and additional reports showed that lower HDL-C levels were associated with particulate matter exposure [19, 34, 36]. Two studies of the association between air pollution and blood lipids in adults in China both showed that particulate matter had a deleterious effect on blood lipid markers, and this effect was more pronounced in people who were overweight or obese [37]. These studies are similar to the results of ambient particulate matter and lipid-related changes that we observed in hypertensive patients with arteriosclerosis and total hypertension. Particulate matter exposure may have a stronger effect on raising TC and lowering HDL-C in individuals already at high risk, a study of older U.S. men suggests [38]. This corresponds to our findings and conjectures.
The cardiovascular harm of ambient particulate matter exposure has been widely confirmed, and it has been proposed that oxidative stress [39, 40], inflammatory response [41, 42], and DNA methylation [43, 44] are the main damage mechanisms. Dyslipidemia is the most sensitive metabolic risk factor to air pollution exposure [45]. Ambient particulate matter can cause systemic inflammation and oxidative stress, inducing adverse lipid metabolism and oxidation [46]. The inflammatory response further mediates a lipid compensatory response against the invasion of hazardous substances from ambient particulate matter and aids tissue repair. When the repair of injury is saturated by lipid compensatory responses, cascading repetitive stimulation of the inflammatory response enhances atherosclerotic lesion formation [47]. Additionally ambient particulate matter can also lead to specific gene methylation related to lipid metabolism by reducing the activity of DNA methyltransferases [48]. The above studies exhibited different mechanistic pathways by which ambient particulate matter altered blood lipids, which could directly provoke adverse lipid metabolism and also provoke lipid compensatory responses in the pre hazard period, corresponding to the differences in relevant lipid changes among different populations in this study and the speculation raised previously.
Two-pollutant model calculation found that the correlations between PM10 and PM2.5 exposures and adverse changes in blood lipid markers in the total hypertensive population and the hypertensive patients with arteriosclerosis remained basically stable after adjusting for the four types of gaseous pollutants respectively. Moreover, the blood lipid changes related to PM10 exposure were more significant than those of PM2.5 in the dual-pollutant model. Therefore, we believe that compared with gaseous pollutants, ambient particulate matter exposure is more strongly associated with increased TC, Lp(a) and decreased HDL-C, and the harmful effect of PM10 is greater than that of PM2.5. A longitudinal study in Shijiazhuang, China also suggests that ambient particulate matter may have a greater impact on lipid health than gaseous pollutants [49]. In a hypertensive population with arteriosclerosis, the positive associations between ambient particulate matter exposure and HDL-C and LDL-C were partially reversed after adjustment for gaseous pollutants. Based on the results of the dual-pollutant model, we believe that ambient particulate matter exposure is not the main pollutant affecting the elevated HDL-C and LDL-C concentrations in this population. Combined with the previous conjecture, we suggest that the phenomenon of higher HDL-C associated with particulate matter exposure in people with hypertension without vascular stiffness, on the one hand, may be due to compensatory increases in HDL-C caused by the body's own protective mechanisms against lipid homeostasis. On the other hand, gaseous pollutants, which were positively correlated with each other and with particulate matter, exhibited stronger associations with HDL-C elevation. Based on the above results, we speculate that ambient particulate matter exposure can cause adverse changes in blood lipid levels in hypertensive patients and increase the risk of arteriosclerosis events.
This study has several strengths. First, we applied the detailed blood lipid detection data of inpatients collected by the hospital's big data center, and the real-time monitoring data of air pollution concentration collected by the environmental protection department to ensure the accuracy of the data source. We selected hypertensive patients as research subjects and grouped them according to whether they were accompanied by arteriosclerosis, so as to study the correlation between ambient particulate matter exposure and the risk of blood lipids in hypertensive patients with different disease states. Second, the data analysis applied a semiparametric generalized additive model. After adjusting for the effects of gender, age, weather, and time, the correlation between ambient particulate matter pollution and blood lipids in different lag time periods in a year was more accurately explored, and the relationship between each blood lipid index and air pollution exposure was calculated. Third, this study is the first to explore the relationship between air pollution and human blood lipid concentrations in southern Jiangxi, China, and its findings can provide reference for other regions with similar development levels and geographic latitudes.
This study also has some limitations. First, in addition to environmental factors, blood lipid levels are affected by a variety of factors, including genetics, behavior, and medication [50, 51]. This study was not able to collect this information, so the results ignore the role of some valuable confounding factors and may be biased. Secondly, the air pollution concentration is the average of the data of the five monitoring points in Ganzhou City, and it is impossible to accurately understand the actual exposure concentration of air pollution of each research object. This exposed misclassification is likely to reduce the significance of the association [52]. In addition, our study only analyzed the lipid effects associated with particulate matter exposure during a one-year period, and did not investigate the effect of a longer lag. Significant associations of LP (a) with ambient PM generally appeared at lag 359 days. Looking at the trends of LP (a) changes associated with exposure at different lag times, more significant associations with LP (a) increases may occur at longer exposure times.
## 5. Conclusions
Ambient particulate matter exposure was associated with higher TC, LP (a) and lower HDL-C in hypertensive patients, and PM10 exposure was more strongly associated with changes in lipid markers than PM2.5. There are association of long-term exposure to ambient particulate matter with the risk of arteriosclerosis in hypertensive patients, and such impacts have stronger harmful effects in patients with arteriosclerosis.
## 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 Gannan Medical University. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
YG: conceptualization, writing–reviewing and editing, and funding acquisition. CL and LH: methodology, software, data curation, formal analysis, writing–original draft, review, and editing. XZho: software, formal analysis, investigation, and data curation. KH and MG: sensitivity analysis. XZha: conceptualization, data acquisition, 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. Nemery B, Hoet PH, Nemmar A. **The meuse valley fog of 1930: an air pollution disaster**. *Lancet.* (2001) **357** 704-8. DOI: 10.1016/S0140-6736(00)04135-0
2. Brauer M, Amann M, Burnett RT, Cohen A, Dentener F, Ezzati M. **Exposure assessment for estimation of the global burden of disease attributable to outdoor air pollution**. *Environ Sci Technol.* (2012) **46** 652-60. DOI: 10.1021/es2025752
3. Neira MP. **Air pollution and human health: a comment from the World Health Organization**. *Annals of Global Health.* (2019) **85** 141. DOI: 10.5334/aogh.2712
4. Wang X, Leng M, Liu Y, Qian ZM, Zhang J, Li Z. **Different sized particles associated with all-cause and cause-specific emergency ambulance calls: a multicity time-series analysis in China**. *Sci Total Environ.* (2021) **783** 147060. DOI: 10.1016/j.scitotenv.2021.147060
5. Zeng Q, Li P, Ni Y, Li GX, Wang DZ, Pan XC. **Research on the relationship between atmospheric inhalable particulate matter and cardiovascular diseases burden in Tianjin]**. (2018) **46** 50-5. DOI: 10.3760/cma.j.issn.0253-3758.2018.01.009
6. Tian Y, Liu H, Wu Y, Si Y, Song J, Cao Y. **Association between ambient fine particulate pollution and hospital admissions for cause specific cardiovascular disease: time series study in 184 major Chinese cities**. *BMJ.* (2019) **367** l6572. DOI: 10.1136/bmj.l6572
7. Soldevila BN, Vinyoles BE, Agudo UJ, Camps VL. **Air pollution, cardiovascular risk and hypertension**. *Hipertens Riesgo Vasc.* (2018) **35** 177-84. DOI: 10.1016/j.hipert.2018.03.001
8. Mannucci PM, Ancona C. **Noise and air pollution as triggers of hypertension**. *Eur Heart J.* (2021) **42** 2085-7. DOI: 10.1093/eurheartj/ehab104
9. Mao S, Chen G, Liu F, Li N, Wang C, Liu Y. **Long-term effects of ambient air pollutants to blood lipids and dyslipidemias in a Chinese rural population**. *Environ Pollut.* (2020) **256** 113403. DOI: 10.1016/j.envpol.2019.113403
10. Islam T, Gauderman WJ, Berhane K, McConnell R, Avol E, Peters JM. **Relationship between air pollution, lung function and asthma in adolescents**. *Thorax.* (2007) **62** 957-63. DOI: 10.1136/thx.2007.078964
11. Hwang BF, Chen YH, Lin YT, Wu XT, Leo LY. **Relationship between exposure to fine particulates and ozone and reduced lung function in children**. *Environ Res.* (2015) **137** 382-90. DOI: 10.1016/j.envres.2015.01.009
12. Spyratos D, Sioutas C, Tsiotsios A, Haidich AB, Chloros D, Triantafyllou G. **Effects of particulate air pollution on nasal and lung function development among Greek children: a 19-year cohort study**. *Int J Environ Health Res.* (2015) **25** 480-9. DOI: 10.1080/09603123.2014.979775
13. Franssen R, Monajemi H, Stroes ES, Kastelein JJ. **Obesity and dyslipidemia**. *Med Clin North Am.* (2011) **95** 893-902. DOI: 10.1016/j.mcna.2011.06.003
14. Lee MH, Kim HC, Ahn SV, Hur NW, Choi DP, Park CG. **Prevalence of dyslipidemia among Korean adults: Korea National Health and Nutrition Survey 1998-2005**. *Diabetes Metabol J.* (2012) **36** 43-55. DOI: 10.4093/dmj.2012.36.1.43
15. Kulkarni H, Mamtani M, Blangero J, Curran JE. **Lipidomics in the study of hypertension in metabolic syndrome**. *Curr Hypertens Rep.* (2017) **19** 7. DOI: 10.1007/s11906-017-0705-6
16. Da SC, Di Domenico M, Hilario NSP, Ramos RC. **Subchronic air pollution exposure increases highly palatable food intake, modulates caloric efficiency and induces lipoperoxidation**. *Inhal Toxicol.* (2018) **30** 370-80. DOI: 10.1080/08958378.2018.1530317
17. Heydari H, Abroudi M, Adli A, Pirooznia N, Najafi ML, Pajohanfar NS. **Maternal exposure to ambient air pollution during pregnancy and lipid profile in umbilical cord blood samples; a cross-sectional study**. *Environ Pollut.* (2020) **261** 114195. DOI: 10.1016/j.envpol.2020.114195
18. Kim KN, Ha B, Seog W, Hwang IU. **Long-term exposure to air pollution and the blood lipid levels of healthy young men**. *Environ Int.* (2022) **161** 107119. DOI: 10.1016/j.envint.2022.107119
19. Poursafa P, Mansourian M, Motlagh ME, Ardalan G, Kelishadi R. **Is air quality index associated with cardiometabolic risk factors in adolescents? The CASPIAN-III study**. *Environ Res.* (2014) **134** 105-9. DOI: 10.1016/j.envres.2014.07.010
20. Sorensen M, Hjortebjerg D, Eriksen KT, Ketzel M, Tjonneland A, Overvad K. **Exposure to long-term air pollution and road traffic noise in relation to cholesterol: a cross-sectional study**. *Environ Int.* (2015) **85** 238-43. DOI: 10.1016/j.envint.2015.09.021
21. Hong X, Zhang B, Liang L, Zhang Y, Ji Y, Wang G. **Postpartum plasma metabolomic profile among women with preeclampsia and preterm delivery: implications for long-term health**. *BMC Med.* (2020) **18** 277. DOI: 10.1186/s12916-020-01741-4
22. Mura MC, De Felice M, Morlino R, Fuselli S. **Short-term monitoring of benzene air concentration in an urban area: a preliminary study of application of Kruskal-Wallis non-parametric test to assess pollutant impact on global environment and indoor**. *Ann Ist Super Sanita.* (2010) **46** 444-50. DOI: 10.4415/ANN.10.04.13
23. Tian H, Xu B, Wang X, Wang J, Zhong C. **Study on the correlation between ambient environment-meteorological factors and the number of visits of acute otitis media, Lanzhou, China**. *J Otol.* (2020) **15** 86-94. DOI: 10.1016/j.joto.2020.01.002
24. Forbes LJ, Patel MD, Rudnicka AR, Cook DG, Bush T, Stedman JR. **Chronic exposure to outdoor air pollution and markers of systemic inflammation**. *Epidemiology.* (2009) **20** 245-53. DOI: 10.1097/EDE.0b013e318190ea3f
25. Chen SY, Chan CC, Su TC. **Particulate and gaseous pollutants on inflammation, thrombosis, and autonomic imbalance in subjects at risk for cardiovascular disease**. *Environ Pollut.* (2017) **223** 403-8. DOI: 10.1016/j.envpol.2017.01.037
26. Masson O, Steinhauser G, Wershofen H, Mietelski JW, Fischer HW, Pourcelot L. **Potential source apportionment and meteorological conditions involved in airborne (131)I Detections in January/February 2017 in Europe**. *Environ Sci Technol.* (2018) **52** 8488-500. DOI: 10.1021/acs.est.8b01810
27. van Capelleveen JC, Bochem AE, Motazacker MM, Hovingh GK, Kastelein JJ. **Genetics of HDL-C: A causal link to atherosclerosis?**. *Curr Atheroscler Rep.* (2013) **15** 326. DOI: 10.1007/s11883-013-0326-8
28. Collins P, Webb CM, de Villiers TJ, Stevenson JC, Panay N, Baber RJ. **Cardiovascular risk assessment in women - an update**. *Climacteric.* (2016) **19** 329-36. DOI: 10.1080/13697137.2016.1198574
29. Jabir NR, Siddiqui AN, Firoz CK, Ashraf GM, Zaidi SK, Khan MS. **Current updates on therapeutic advances in the management of cardiovascular diseases**. *Curr Pharm Des.* (2016) **22** 566-71. DOI: 10.2174/1381612822666151125000746
30. Townsend N, Wilson L, Bhatnagar P, Wickramasinghe K, Rayner M, Nichols M. **Cardiovascular disease in Europe: epidemiological update 2016**. *Eur Heart J.* (2016) **37** 3232-45. DOI: 10.1093/eurheartj/ehw334
31. Karagiannis AD, Mehta A, Dhindsa DS, Virani SS, Orringer CE, Blumenthal RS. **How low is safe? The frontier of very low (<30 mg/dL) LDL cholesterol**. *Eur Heart J.* (2021) **42** 2154-69. DOI: 10.1093/eurheartj/ehaa1080
32. Jang AY, Han SH, Sohn IS, Oh PC, Koh KK. **Lipoprotein(a) and Cardiovascular Diseases- Revisited**. *Circ J.* (2020) **84** 867-74. DOI: 10.1253/circj.CJ-20-0051
33. Chuang KJ, Yan YH, Chiu SY, Cheng TJ. **Long-term air pollution exposure and risk factors for cardiovascular diseases among the elderly in Taiwan**. *Occup Environ Med.* (2011) **68** 64-8. DOI: 10.1136/oem.2009.052704
34. Ghosh R, Gauderman WJ, Minor H, Youn HA, Lurmann F, Cromar KR. **Air pollution, weight loss and metabolic benefits of bariatric surgery: a potential model for study of metabolic effects of environmental exposures**. *Pediatr Obes.* (2018) **13** 312-20. DOI: 10.1111/ijpo.12210
35. McGuinn LA, Schneider A, McGarrah RW, Ward-Caviness C, Neas LM, Di Q. **Association of long-term PM2.5 exposure with traditional and novel lipid measures related to cardiovascular disease risk**. *Environ Int.* (2019) **122** 193-200. DOI: 10.1016/j.envint.2018.11.001
36. Chuang KJ, Yan YH, Cheng TJ. **Effect of air pollution on blood pressure, blood lipids, and blood sugar: a population-based approach**. *J Occup Environ Med.* (2010) **52** 258-62. DOI: 10.1097/JOM.0b013e3181ceff7a
37. Yang BY, Bloom MS, Markevych I, Qian ZM, Vaughn MG, Cummings-Vaughn LA. **Exposure to ambient air pollution and blood lipids in adults: The 33 Communities Chinese Health Study**. *Environ Int.* (2018) **119** 485-92. DOI: 10.1016/j.envint.2018.07.016
38. Bind MA, Peters A, Koutrakis P, Coull B, Vokonas P, Schwartz J. **Quantile regression analysis of the distributional effects of air pollution on blood pressure, heart rate variability, blood lipids, and biomarkers of inflammation in elderly American men: the normative aging study**. *Environ Health Perspect.* (2016) **124** 1189-98. DOI: 10.1289/ehp.1510044
39. Zhao L, Zhang L, Chen M, Dong C, Li R, Cai Z. **Effects of Ambient Atmospheric PM2.5, 1-Nitropyrene and 9-Nitroanthracene on DNA Damage and Oxidative Stress in Hearts of Rats**. *Cardiovasc Toxicol.* (2019) **19** 178-190. DOI: 10.1007/s12012-018-9488-5
40. Long YM, Yang XZ, Yang QQ, Clermont AC, Yin YG, Liu GL. **PM**. *J Hazard Mater.* (2020) **386** 121659. DOI: 10.1016/j.jhazmat.2019.121659
41. Xu X, Wang H, Liu S, Xing C, Liu Y, Zhou W. **TP53-dependent autophagy links the ATR-CHEK1 axis activation to proinflammatory VEGFA production in human bronchial epithelial cells exposed to fine particulate matter (PM2.5)**. *Autophagy.* (2016) **12** 1832-48. DOI: 10.1080/15548627.2016.1204496
42. Hu H, Asweto CO, Wu J, Shi Y, Feng L, Yang X. **Gene expression profiles and bioinformatics analysis of human umbilical vein endothelial cells exposed to PM**. *Chemosphere.* (2017) **183** 589-98. DOI: 10.1016/j.chemosphere.2017.05.153
43. Jiang Y, Li J, Ren F, Ji C, Aniagu S, Chen T. **PM**. *Environ Pollut.* (2019) **255** 113331. DOI: 10.1016/j.envpol.2019.113331
44. Zhao L, Zhang M, Bai L, Zhao Y, Cai Z, Yung K. **Real-world PM2.5 exposure induces pathological injury and DNA damage associated with miRNAs and DNA methylation alteration in rat lungs**. *Environ Sci Pollut Res Int* (2022) **29** 28788-803. DOI: 10.1007/s11356-021-17779-7
45. Yang BY, Guo Y, Markevych I, Qian ZM, Bloom MS, Heinrich J. **Association of long-term exposure to ambient air pollutants with risk factors for cardiovascular disease in China**. *JAMA Netw Open.* (2019) **2** e190318. DOI: 10.1001/jamanetworkopen.2019.0318
46. Shanley RP, Hayes RB, Cromar KR, Ito K, Gordon T, Ahn J. **Particulate air pollution and clinical cardiovascular disease risk factors**. *Epidemiology.* (2016) **27** 291-8. DOI: 10.1097/EDE.0000000000000426
47. Esteve E, Ricart W, Fernandez-Real JM. **Dyslipidemia and inflammation: an evolutionary conserved mechanism**. *Clin Nutr.* (2005) **24** 16-31. DOI: 10.1016/j.clnu.2004.08.004
48. Chen R, Meng X, Zhao A, Wang C, Yang C, Li H. **DNA hypomethylation and its mediation in the effects of fine particulate air pollution on cardiovascular biomarkers: a randomized crossover trial**. *Environ Int.* (2016) **94** 614-9. DOI: 10.1016/j.envint.2016.06.026
49. Zhang K, Wang H, He W, Chen G, Lu P, Xu R. **The association between ambient air pollution and blood lipids: a longitudinal study in Shijiazhuang, China**. *Sci Total Environ.* (2021) **752** 141648. DOI: 10.1016/j.scitotenv.2020.141648
50. Schoeler M, Caesar R. **Dietary lipids, gut microbiota and lipid metabolism**. *Rev Endocr Metab Disord.* (2019) **20** 461-72. DOI: 10.1007/s11154-019-09512-0
51. Felzer-Kim IT, Visker JR, Ferguson DP, Hauck JL. **Infant blood lipids: a systematic review of predictive value and influential factors**. *Expert Rev Cardiovasc Ther.* (2020) **18** 381-94. DOI: 10.1080/14779072.2020.1782743
52. Hutcheon JA, Chiolero A, Hanley JA. **Random measurement error and regression dilution bias**. *BMJ.* (2010) **340** c2289. DOI: 10.1136/bmj.c2289
|
---
title: Calcium isotopes as a biomarker for vascular calcification in chronic kidney
disease
authors:
- Anthony Dosseto
- Kelly Lambert
- Hicham I Cheikh Hassan
- Andrew Fuller
- Addison Borst
- Florian Dux
- Maureen Lonergan
- Theo Tacail
journal: 'Metallomics: Integrated Biometal Science'
year: 2023
pmcid: PMC9989339
doi: 10.1093/mtomcs/mfad009
license: CC BY 4.0
---
# Calcium isotopes as a biomarker for vascular calcification in chronic kidney disease
## Abstract
Calcium balance is abnormal in adults with chronic kidney disease (CKD) and is associated with the development of vascular calcification. It is currently not routine to screen for vascular calcification in CKD patients. In this cross-sectional study, we investigate whether the ratio of naturally occurring calcium (Ca) isotopes, 44Ca and 42Ca, in serum could be used as a noninvasive marker of vascular calcification in CKD.
We recruited 78 participants from a tertiary hospital renal center: 28 controls, 9 subjects with mild–moderate CKD, 22 undertaking dialysis and 19 who received a kidney transplant. For each participant, systolic blood pressure, ankle brachial index, pulse wave velocity, and estimated glomerular filtration rate were measured, along with serum markers. Calcium concentrations and isotope ratios were measured in urine and serum. While we found no significant association between urine Ca isotope composition (noted δ$\frac{44}{42}$Ca) between the different groups, δ$\frac{44}{42}$Ca values in serum were significantly different between healthy controls, subjects with mild–moderate CKD and those undertaking dialysis ($P \leq 0.01$). Receiver operative characteristic curve analysis shows that the diagnostic utility of serum δ$\frac{44}{42}$Ca for detecting medial artery calcification is very good (AUC = 0.818, sensitivity $81.8\%$ and specificity $77.3\%$, $P \leq 0.01$), and performs better than existing biomarkers. Although our results will need to be verified in prospective studies across different institutions, serum δ$\frac{44}{42}$Ca has the potential to be used as an early screening test for vascular calcification.
## Graphical Abstract
Graphical AbstractThe calcium isotope composition (noted δ$\frac{44}{42}$Ca) of blood serum of adults with chronic kidney disease can be used as a screening tool to detect medial vascular calcification, where a δ$\frac{44}{42}$Ca value greater than −0.53‰ indicates a high risk of medial vascular calcification.
## Introduction
Chronic kidney disease (CKD) is a global health problem affecting almost $10\%$ of the global population.1 The largest contributor to morbidity and mortality in patients with CKD is cardiovascular disease, which can affect almost half of patients with severe CKD, in addition to being the leading cause of death in those with kidney failure (KF).2 One of the strongest predictors of cardiovascular risk is vascular calcification.3 To date no intervention has been shown to be effective in reversing vascular calcification in CKD and only a few have been shown to reduce the progression.4 One reason for a lack of therapeutic intervention is the difficulty in treating vascular calcification directly. Current therapies have instead focused on restoring the disrupted mineral balance often seen in CKD such as phosphate control, vitamin D deficiency, vitamin K deficiency, and control of hyperparathyroidism. Another reason why vascular calcification has been a difficult therapeutic target to aim for is the lack of appropriate tools to diagnose vascular calcification in the early stage, when an intervention or lifestyle modification may achieve reversible results. Current tools to diagnose vascular calcification rely exclusively on imaging such as echocardiogram, nuclear medicine scans, or computed tomographic (CT) coronary angiography. These tests are expensive and require exposing the patient to radiation. They are also inappropriate tests to implement as a tool to screen the CKD population due to expense, time needed for each individual test and experienced personnel required to perform and interpret the test. Another limitation is the inability to detect an abnormality until after permanent and irreversible changes have set in (such as the case of CT coronary angiography detecting the presence of permanent vascular calcification).5 The ideal screening test for vascular calcification in the CKD population would be easily obtained (such as a blood or urine sample), run in bulk and with results that can be interpreted easily by health professionals without having to be reported by specialists.
Calcium (Ca) has several naturally occurring isotopes, including 40Ca, 42Ca, 43Ca, and 44Ca. The distribution of these isotopes in biological fluids and tissues changes in response to biological processes. This distribution is expressed as the ratio of two isotopes, generally 44Ca/42Ca, thereafter termed Ca isotope ratio and noted δ$\frac{44}{42}$Ca (see “Methods”). Calcium isotopes, as measured in biological fluids such as blood or urine, reflect the global mineral balance of the organism as the result of two combined processes. First, bone is depleted in heavy isotopes (low δ$\frac{44}{42}$Ca), because of a preferential uptake of light Ca isotopes from blood upon bone mineralization (by −0.30‰ or less, e.g. Toepfer et al.6). Hence, release or uptake of bone Ca tends to, respectively, induce a decrease or increase of serum δ$\frac{44}{42}$Ca values. Second, Ca renal excretion results in the preferential loss of heavy Ca isotopes in urine, higher by +1.2‰ when compared to blood. As a feedback effect, the organism tends to retain light Ca isotopes, δ$\frac{44}{42}$Ca of blood being on average decreased by ca. −0.6‰.7,8 The increase or decrease of Ca renal excretion (occurring in negative or positive mineral balance, respectively) contributes to a decrease or increase of urine and blood δ$\frac{44}{42}$Ca, respectively.
Overall, on the one hand, negative mineral balance (e.g. induced or pathological bone loss) produces a decrease in serum and urine isotope compositions (e.g. Morgan et al., Heuser et al., and Eisenhauer et al.9–11). Morgan et al.9 for instance proposed that the Ca isotope ratio of urine could be used as a tracer of bone loss, and their bed rest study showed that urine Ca isotope ratios decreased after 1 wk, signaling bone loss long before it could be detected by densitometry. On the other hand, calcification in the body, such as in net bone accretion, is illustrated by an increase in the Ca isotope ratio of urine or blood.11–13 Recently, Shroff et al.14 used Ca isotopes in blood, urine and feces of children with CKD to successfully identify changes in bone Ca balance. While Ca isotopes have been tested and conceptualized in the context of non-ectopic calcification, they are yet to be used for ectopic calcification.
Given the pathological changes occurring in CKD, where bone de-mineralization and disrupted mineral balance lead to an increased risk of vascular calcification, calcium isotopes may be a potential therapeutic test to identify patients at risk of vascular calcification. We therefore set out to examine whether the Ca isotope ratio of urine or blood could be used as an effective biomarker of CKD vascular calcification by conducting an observational cohort study in a population of healthy volunteers, patients with CKD and patients with KF on dialysis. The aims are to (i) assess changes of Ca isotope ratio in urine or blood across CKD stages including dialysis and transplant patients, (ii) examine any association between the Ca isotope ratio of urine or blood, and traditional markers of vascular changes (e.g. FGF-2315) and pulse wave velocity (PWV), (iii) evaluate the diagnostic accuracy of urine and serum Ca isotopes for identifying vascular calcification.
## Methods
A total of 78 adult participants were recruited in 2018 from a single renal center ($$n = 28$$ controls, $$n = 9$$ with mild–moderate CKD, $$n = 22$$ with KF undertaking dialysis, $$n = 19$$ with KF who had received a kidney transplant). Sample size was determined by assessing the previous literature and the study population used to ensure statistical power. A previous study conducted by Channon et al. ,16 recruited 12 healthy individuals (8 males) and found statistical significance between the parameters investigated despite the small sample size. We decided to recruit at minimum the same sample size for each group, with a maximum number of participants at 28, due to funding and time constraints. Recruitment took place at the Wollongong Hospital Renal Unit. We recruited patients with any stage of CKD, patients on dialysis and patients with a functional kidney transplant. Participants were recruited by a renal nurse upon presentation for routine urine analysis at the Wollongong Hospital Renal Unit. Patients who agreed to participate in the study were given study information sheets as well as provided informed consent. Kidney transplant recipients and home hemodialysis patients who do not visit a renal center routinely were sent an invitation to participate. A follow up phone call by an independent research nurse at the renal unit was used to ascertain eligibility and interest in the study. Further amendments to ethics applications (2019/ETH03747) allowed mail out invitations to focus on stage 3 and transplant participants. A convenience sample of healthy controls was recruited. These consisted of staff members or partners/family of staff or participants with CKD. Ethics approval was granted by the Human Research Ethics Committee (Health and Medical) of the University of Wollongong (HREC number: HREC/18/WGONG/188). Inclusion criteria for the study were: (i) individuals suffering any stage of CKD over the age of 18, and (ii) healthy controls that were free from a diagnosis of CKD and were over the age of 18. Exclusion criteria were: (i) adults with a diagnosis of advanced dementia or severe cognitive impairment that would impact informed consent or compliance with study instructions; (ii) individuals with metastatic cancer in the bones to reduce confounding results, as well as reduce added burden to the individual; (iii) individuals with multiple myeloma or other conditions known to impact bone turnover not associated with CKD; and (iv) individuals who were aneuric (<5 mL urine per day).
Urine and blood were collected following overnight fasting. For each participant, age, presence of comorbidities and gender were recorded. Blood pressure, ankle brachial index (ABI), brachial-ankle pulse wave velocity (ba-PWV), and estimated glomerular filtration rate (eGFR) were measured, along with serum albumin, calcium (Ca), phosphate, parathyroid hormone (PTH), alkaline phosphatase level (ALP), 1,25OH vitamin D, creatinine and fibroblast growth factor 23 (FGF23) concentrations. PWV is a noninvasive measurement of arterial stiffness, whereby PWV value increases with arterial stiffness. In CKD transplant and dialysis patients, PWV is positively correlated to cardiovascular mortality.17 ABI and ba-PWV measures were conducted using a noninvasive vascular screening device (Omron Colin VP-1000) following the method outlined in Chen et al.18 *In a* subset of patients (due to time and funding constraints), Ca concentration and isotope ratio were measured in urine ($$n = 52$$ for Ca concentration, $$n = 36$$ for Ca isotope ratio) and serum ($$n = 36$$ for Ca concentration, $$n = 54$$ for Ca isotope ratio). For Ca concentration and isotope ratio measurements, urine and serum were freeze dried and then mineralized in nitric acid and hydrogen peroxide by microwave digestion. Samples were then re-dissolved in 2 mol/L HNO3 for ion exchange chromatography. An aliquot was taken and diluted in 0.3 mol/L HNO3 for measurement of calcium concentrations (see following text).
Ion exchange chromatography was performed to isolate Ca from the sample's matrix, a necessary step for isotope ratio measurement. This was performed using a prepFAST-MC automated chromatography system (ESI, Omaha, NE, USA) at the Wollongong Isotope Geochronology Laboratory (WIGL) following the method outlined in Romaniello et al.19 The Ca elution was re-dissolved in 0.05 mol/L HNO3 for isotope ratio measurement.
Calcium concentration determination was performed by quadrupole inductively coupled plasma mass spectrometry (Q ICP–MS) on a ThermoFisher iCAP-Q at WIGL. A calibration curve was produced using a multi-element standard (Inorganic Ventures 71A) with concentrations ranging from 0.5 to 250 ng/g. An internal standard (Inorganic Ventures 71D) was introduced along with the samples and 45Sc was measured to account for instrument drift.
Calcium isotopes were measured on a ThermoFisher Neptune Plus multi-collector ICP–MS at WIGL. A 100 μL/min PFA nebulizer was used with a CETAC Aridus II desolvator as sample introduction system, along with jet sample and X skimmer cones. To account for mass bias, a 1.5 μg/g solution of Alfa Aesar Specpure Ca elemental standard in 0.05 mol/L HNO3 was measured before and after each sample following the standard-sample bracketing method.20 The Ca concentration of samples in 0.05 mol/L HNO3 was adjusted to match that of the primary standard within $10\%$. Instrument blanks were measured before each standard and sample and subtracted from each isotope. 42Ca, 43Ca, and 44Ca were collected in Faraday cups in medium resolution mode, for 40 cycles of 4.194 s each. Mass bias factors was calculated for 42Ca/43Ca, 42Ca/43Ca, and 43Ca/44Ca ratios. Mass 43.5 was also collected to measure 87Sr2+, and assess the contribution of 86Sr2+ and 88Sr2+ isobaric interferences on 43Ca and 44Ca, respectively. The $\frac{43.5}{44}$ ratio was generally <10−5, such that no correction for isobaric interference was necessary (the change imparted by the correction on the 44Ca/42Ca would be within the analytical error). If the $\frac{43.5}{44}$ ratio was greater than 10−5, the analysis was rejected. Furthermore, the analysis was also rejected if either the 44Ca intensity of the sample deviated from that of the standard by more than $10\%$, or the absolute value of the 2sd of the three mass bias factors was greater than 0.1.
The 44Ca/42Ca and 43Ca/42Ca isotope ratios were converted to δ$\frac{44}{42}$CaWIGL and δ$\frac{43}{42}$CaWIGL values, respectively, expressed in per-mille (‰) and defined as (for instance for δ$\frac{44}{42}$CaWIGL): where (44Ca/42Ca)WIGL is the 44Ca/42Ca ratio of the Alfa Aesar Specpure Ca standard solution. In addition to the filters mentioned earlier, the analysis was rejected if the 2sd of δ$\frac{44}{42}$CaWIGL × 0.50667 and δ$\frac{43}{42}$CaWIGL exceeded 0.1 (indicating deviation for mass-dependent isotope fraction).
The δ$\frac{44}{42}$CaWIGL values can be converted to δ$\frac{44}{42}$Ca relative to “ICP Ca Lyon,”8,21,22 “ICP1,”9,16,23 or NIST SRM 915a10,24 standards by adding 0.009‰, 0.277‰, or 0.527‰, respectively. δ$\frac{44}{42}$CaWIGL values were then converted to δ$\frac{44}{42}$CaSRM915a values (i.e. using NIST SRM 915a as reference standard) for ease of comparison across studies. All Ca isotope compositions presented in the following text are δ$\frac{44}{42}$CaSRM915a values (noted δ$\frac{44}{42}$Ca for convenience as in previous studies; e.g. Heuser et al., Shroff et al., and Tanaka et al.10,14,24).
No Ca isotope reference material with a matrix similar to urine or blood exists, thus we used International Association for the Physical Sciences of the Oceans (IAPSO) seawater instead.25 The mean δ$\frac{44}{42}$*Ca is* 0.91‰ ± 0.02‰ [1 standard deviation (1 SD); $$n = 11$$], within error of the value of 0.94‰ ± 0.12‰ reported in Tacailet al.25 Precision was assessed by processing several aliquots of the seawater standard and yielded an uncertainty of 0.025‰ (1 SD; $$n = 11$$). The mean total procedure blank is 31 ng ± 21 ng of Ca (1 SD; $$n = 8$$), ∼$0.1\%$ of the amount of Ca processed for samples and affecting δ$\frac{44}{42}$Ca values by only ∼0.0002‰, well within the analytical uncertainty.
For statistical analysis, continuous variables are expressed as mean (standard deviation) or median (interquartile range) as per distribution. Categorical variables are expressed as number (percentage). Comparisons between groups, according to CKD status, were conducted using one-way ANOVA, χ2 and Kruskal–Wallis test as appropriate. Pearson's correlation test was used to assess association between the Ca isotope ratio of urine or serum, creatinine and other variables (creatinine and FGF-23 were log-transformed for data to be normally distributed). To determine which baseline variables were independently associated with the Ca isotope ratio of urine or serum, we performed linear regression analyses. The stronger determinants for serum Ca isotopes were selected by stepwise backward multivariable linear regression analysis. For inclusion, P-values were set at <0.2. Models were compared with adjusted R2 and the model with the largest value selected. We did not include CKD status in the multivariable linear regression analysis due to collinearity with creatinine. Analysis was conducted in SPSS (version 25) and R, and a P-value <0.05 was considered significant.26 Groups were not age matched, and age correction was not possible due to sample size, but the effect of age, as well as dietary and supplemental Ca, on interpretations are discussed in the following text.
The sensitivity and specificity for using calcium isotope ratios as a method of detecting vascular calcification were examined using the receiver operating characteristic (ROC) curve analysis for evaluating diagnostic tests and predictive models.27 ROC curve analysis was performed using R package, pROC.28 The reference method for detecting vascular calcification was an ABI ≤0.9 or ≥1.3 for PAD (suggesting intimal vascular calcification) and ba-PWV for arteriosclerosis (suggesting medial artery calcification). For arteriosclerosis, it was determined to be present if ba-PWV was ≥1800 cm/s or if ba-PWV was >0.16 × age2 − 4.40 × age + 977.52 cm/s for female subjects or 0.20 × age2 − 12.13 × age + 1341.34 cm/s for male subjects (where age is in years).18 *An area* under the curve (AUC) of 0.9–1 indicates an excellent diagnostic test; 0.8–0.89 a very good diagnostic test; 0.70–0.79 a good diagnostic test; 0.6–0.7 a sufficient diagnostic test; 0.5–0.6 a poor test; and <0.5 not useful test.29
## Clinical data
The median age of all participants was 60 [43–72] years, with participants in the control group being significantly younger (41 [29–56] years) than all other groups ($P \leq 0.01$). Overall, the majority of participants were male ($$n = 45$$, $58\%$). In the control group, the majority of participants were female ($$n = 16$$, $56\%$), whereas those in the other three groups were predominantly male ($P \leq 0.01$).
Body mass index (BMI) is not significantly different between groups ($$P \leq 0.10$$), nor is dietary Ca ($$P \leq 0.98$$; note for dietary *Ca data* were available for the mild–moderate CKD group). Two subjects in the control group took Ca medication ($7\%$ of that group), 6 in the mild–moderate CKD group ($67\%$), 6 in the transplant group ($35\%$), and 16 in the dialysis group ($76\%$). As expected, participants undertaking dialysis show levels of PO43−, PTH, ALP, creatinine, and FGF23 significantly higher than those in other groups ($P \leq 0.01$, Table 1). Additionally, they show reduced vitamin D levels compared to other groups ($P \leq 0.01$, Table 1). Fourteen ($18\%$) participants exhibit an ABI of ≤0.9 indicating the presence of PAD (intimal calcification), but with no significant differences in proportions of participants with ABI ≤ 0.9 across all three CKD groups ($$P \leq 0.54$$, Table 1). For participants with an ABI > 0.9, the control group shows the highest proportion of participants ($$n = 23$$, $49\%$) ($P \leq 0.01$, Table 1).
**Table 1.**
| Characteristic | All | Control | Mild–moderate | Dialysis | Transplant | P-value |
| --- | --- | --- | --- | --- | --- | --- |
| n | 78 | 28 | 9 | 22 | 19 | |
| Age (years) | 59 [43–71] | 38 [28–54] # | 70 [66–72] | 71 [67–74] | 58 [50–70] | <0.01* |
| Male | 45 (58) | 12 (44) # | 5 (56) | 13 (59) | 15 (79) # | 0.09 |
| SBP (mmHg) | 139 [122–159] | 129 [118–141] | 152 [145–179] | 145 [140–166] | 144 [138–161] | <0.01* |
| BMI (kg/m2) | 28.0 [24.5–30.8] | 26.0 [23.5–27.2] | 29.8 [23.5–32.6] | 29.1 [24.0–38.1] | 28.7 [27.4–31.9] | 0.102 |
| eGFR (mL/min/1.73m2) | | 86 (7.2) # | 47 (24) | 6.7 (2.0) # | 57 (21) | |
| PO43− (mmol/L) | | 1.13 (0.16) | 1.12 (0.29) | 1.51 (0.51) # | 1.03 (0.2) | <0.01* |
| Albumin (g/L) | | 40.8 (2.7) | 39 (4.2) | 32.6 (3.0) # | 41 (11) | <0.01* |
| PTH (pmol/L) | | 4.1 (3–5) | 5 (4.3–8.5) | 23 (4.1–54.6) ac | 7.8 (5.5–13.2) ac | <0.01* |
| ALP (U/L) | | 69.5 (58–82.8) | 82.5 (73.8–99) | 120 (66–188) a | 73 (63.5–80) a | <0.01* |
| Creatinine (μmol/L) | | 75 (65–81) # | 110 (87–151) | 666 (519–739) # | 115 (93–149) | <0.01* |
| FGF23 (ng/L) | | 1.65 (1.55–1.76) # | 1.98 (1.79–2.07) | 3.29 (3.14–4.19) # | 2.02 (1.87–2.21) | <0.01* |
| Vitamin D (1,25OH) (nmol/L) | | 120 (33.8) | 112.6 (30.8) | 44.24 (25.3) # | 109 (42.5) | <0.01* |
| PAD (ABI ≤ 0.9) i | | 3 (4.9) | 2 (3.3) | 5 (8.2) | 4 (6.6) | 0.54 |
| No PAD (ABI > 0.9) i | | 23 (37.7) | 5 (8.2) | 8 (13.1) ac | 11 (18.0) a | <0.01* |
| ba-PWV (cm/s) h | | 1227 (229) | 1694 (264) | 2050 (713) a | 1585 (318) a | <0.01* |
| Arteriosclerosis h | | 3 (6.8) | 3 (42.9) | 6 (75) a | 9 (69.23) a | <0.01* |
| Urine Ca (μg/g) | | 119 (86) | 30 (38) | 29 (20) | 31 (22) | |
| Urine δ44/42Ca (‰) | | 0.54 (0.19) | 0.48 (0.15) | 0.69 (0.39) | 0.92 (0.42) | |
| Serum Ca (μg/g) | | 57 (14) | | 59.6 (8.4) | 60.9 (n/a) | |
| Serum δ44/42Ca (‰) g | | −0.70 (0.16) | −0.60 (0.14) | 0.14 (0.24) ac | −0.47 (0.16) a | <0.01* |
The control group has significantly lower ba-PWV (1227 cm/s ± 229 cm/s; 1 SD) when compared to both the transplant (1586 cm/s ± 318 cm/s; 1 SD) and dialysis (2120 cm/s ± 790 cm/s; 1 SD) groups ($P \leq 0.01$), however, not when compared to the mild–moderate CKD group ($$P \leq 0.08$$, Table 2). Furthermore, the control group also has a lower prevalence of arteriosclerosis (medial calcification) as determined by ba-PWV ($$n = 3$$, $5.9\%$) ($P \leq 0.01$) than the dialysis and transplant group, however, not when compared to the mild–moderate CKD group ($$P \leq 0.08$$).
**Table 2.**
| Unnamed: 0 | Pearson correlation coefficient | P-value |
| --- | --- | --- |
| Age (years) | 0.50 | <0.001 |
| log(creatinine) (μmol/L) | 0.83 | <0.001 |
| BMI (kg/m2) | 0.09 | 0.53 |
| SBP (mmHg) | 0.25 | 0.08 |
| ABI | 0.17 | 0.30 |
| ba-PWV (cm/s) | 0.55 | <0.001 |
| Ca (mmol/L) | 0.18 | 0.19 |
| PO43− (mmol/L) | 0.36 | 0.009 |
| Albumin (g/L) | −0.72 | <0.001 |
| PTH (pmol/L) | 0.26 | 0.06 |
| ALP (U/L) | 0.25 | 0.07 |
| Vitamin D (1,25OH) (nmol/L) | −0.63 | <0.001 |
| log(FGF23) (ng/L) | 0.78 | <0.001 |
## Urine calcium concentration and isotope composition
The mean urine Ca concentration (determined by ICP–MS) is 66 μg/g ± 72 μg/g. The mean urine Ca concentration of the control group (119 μg/g ± 86 μg/g) is significantly higher than that of each other group ($P \leq 0.01$ between control and mild–moderate CKD groups; $P \leq 0.001$ between control and dialysis or transplant groups; Fig. 1). There is a strong correlation ($P \leq 0.001$) between urine Ca and log-transformed creatinine (R2 = 0.21), log-transformed FGF23 (R2 = 0.26; Fig. 2), eGFR (R2 = 0.26), age (R2 = 0.19), and albumin (R2 = 0.18), and to a lesser extent with log-transformed PTH (R2 = 0.18, $$P \leq 0.002$$). The mean Ca isotope composition (δ$\frac{44}{42}$Ca) of urine is 0.61‰ ± 0.30‰ and shows no relationship with the CKD groups (not shown). We found poor correlations between urine δ$\frac{44}{42}$Ca and serum creatinine (R2 = 0.2, $$P \leq 0.3$$) and serum δ$\frac{44}{42}$Ca ($r = 0.4$, $$P \leq 0.06$$), and no correlations with other parameters.
**Fig. 1:** *Boxplot of urine calcium concentrations (determined by ICP–MS) across the four groups. All figures were drawn using R package ggplot2.37* **Fig. 2:** *Urine Ca concentration (in μg/g) as a function of FGF23. The curve shows a local polynomial regression and the gray area represents the confidence interval at 0.9 level on the regression. Regressions and their confidence intervals were drawn in this and following figures using R package ggplot237 with a span value (degree of smoothing) of 0.75 for polynomial regressions.*
## Serum ca isotope composition
The mean δ$\frac{44}{42}$Ca of serum is −0.34‰ ± 0.42‰ (1 SD) with a significant difference across the four groups ($P \leq 0.01$). The mean δ$\frac{44}{42}$Ca of serum increases from controls (−0.70‰ ± 0.16‰; 1 SD, $$n = 10$$), to mild–moderate CKD (−0.60‰ ± 0.14‰; 1 SD, $$n = 4$$), to transplant (−0.47‰ ± 0.16‰; 1 SD, $$n = 12$$), and is highest in the dialysis group (0.14‰ ± 0.24‰; 1 SD, $$n = 19$$) (Fig. 3). There is no significant difference between the Ca isotope composition of serum in controls and that in the mild–moderate CKD group ($P \leq 0.05$). Each group was sub-sampled to yield a similar mean age, and the mean δ$\frac{44}{42}$Ca of each age-matched group was calculated. The age-matched control group has a mean δ$\frac{44}{42}$Ca of −0.77‰ ± 0.11‰ (1 SD, $$n = 9$$), not significantly different from that of the whole control group ($$P \leq 0.329$$). Similarly, the age-matched dialysis group has a mean δ$\frac{44}{42}$Ca of 0.17‰ ± 0.22‰ (1 SD, $$n = 8$$), also not significantly different for that of the whole dialysis group ($$P \leq 0.794$$).
**Fig. 3:** *Boxplot of serum δ44/42Ca values across the four groups.*
There is no relationship between serum δ$\frac{44}{42}$Ca and dietary Ca intake ($$P \leq 0.156$$; Fig. 4). There are also no significant differences between the mean serum δ$\frac{44}{42}$Ca of transplant subjects taking Ca supplement, and that of those not taking any Ca supplement ($$P \leq 0.518$$). Similarly, there is no significant differences between the mean serum δ$\frac{44}{42}$Ca of dialysis subjects taking Ca medication as a phosphate binder, and that of those not taking any Ca medication ($$P \leq 0.197$$; Fig. 5).
**Fig. 4:** *Serum δ44/42Ca (in ‰) as a function of dietary Ca intake (in mg), for three groups where both data were available. The lack of relationship suggests that dietary Ca may not have a major influence of serum Ca isotopes in our cohort.* **Fig. 5:** *Serum δ44/42Ca (in ‰) as a function of whether calcium supplement/medication was taken. On the x-axis, “0” indicates no Ca supplement/medication taken, while “1” indicates Ca supplement/medication was taken. Note there was no serum δ44/42Ca data for the control subjects taking Ca supplement (N = 2), nor for the mild–moderate CKD subjects not taking Ca supplement (N = 3). The similarity in Ca isotope values for each transplant or dialysis, whether Ca supplement/medication was taken or not, suggests that Ca supplement/medication does not have a major influence on serum δ44/42Ca values.*
There are strong correlations ($P \leq 0.001$) between serum δ$\frac{44}{42}$Ca and log-transformed creatinine, log-transformed FGF23, PWV, vitamin D (Fig. 6), age, eGFR, and albumin (not shown). Serum δ$\frac{44}{42}$Ca was also significantly correlated with phosphate, ALP and PTH (Table 2). Positive correlation between serum δ$\frac{44}{42}$Ca and ALP is in agreement with the recent study of CKD children14; however, they also observed positive association between serum δ$\frac{44}{42}$Ca and vitamin D, and no correlation or a negative correlation between serum δ$\frac{44}{42}$Ca and PTH; however, here we observe a negative correlation between serum δ$\frac{44}{42}$Ca and vitamin D (Fig. 6), and positive between serum δ$\frac{44}{42}$Ca and PTH (not shown). In multivariate regression analysis, the main covariables strongly associated with serum δ$\frac{44}{42}$Ca are creatinine ($P \leq 0.001$) and PWV ($P \leq 0.001$) (Table 3).
**Fig. 6:** *Serum Ca isotope composition as a function of (a) log-transformed FGF23; (b) log-transformed creatinine; (c) pulse wave velocity; and (d) vitamin D (1,25). The lines show linear regressions and gray areas represent the confidence interval at 0.9 level.* **Fig. 7:** *ROC curve for Ca isotopes in serum as medial calcification predictor.* TABLE_PLACEHOLDER:Table 3.
Serum Ca isotope composition shows a strong association with medial artery calcification: ROC curve analysis shows that a serum δ$\frac{44}{42}$Ca value greater than −0.53‰ predicts presence of medial artery calcification with a sensitivity of $81.8\%$ and specificity of $77.3\%$, with an AUC of 0.818 (Fig. 7). This diagnostic tool performs better than ba-PWV (sensitivity: $64.5\%$, specificity: $65.6\%$ and AUC of 0.662, for a cut-off point at 1564 cm/s30).
## Discussion
Measurement of Ca isotopes in urine has been previously used to investigate and quantify changes in bone mineral balance,9-14,16,23,31 or in serum as a biomarker for multiple myeloma disease.32 Here, the Ca isotope composition of urine shows no systematic changes with CKD progression nor with medial or intimal calcification. This could be because, for logistical reasons, we collected spot urine samples instead of 24 hr urine. Although in our study, the conditions for urine collection were uniform across the cohort (in term of fasting and time of collection during the day), because urine composition is variable across the day, not being able to pool 24 hr urine could have resulted in blurring any association between urine δ$\frac{44}{42}$Ca and other biomarkers.
Tanaka et al.24 applied Ca isotopes in serum and bone to investigate bone mineral balance in CKD and diabetic rats. They found that serum Ca isotopes in rats were positively correlated with bone mineral density (measured on the right femoral bone using the DEXA method). Shroff et al.14 measured Ca isotopes in blood, urine and feces of a cohort of children affected with CKD, and children receiving dialysis therapy. Serum δ$\frac{44}{42}$Ca values in children with CKD and in children receiving dialysis were much lower than those of controls, interpreted as a loss of bone mineral content. Here, in our adult cohort, we observe the opposite, where subjects receiving dialysis display higher serum δ$\frac{44}{42}$Ca values than controls (and transplant subjects and subjects with mild–moderate CKD). The difference between the two studies could perhaps be explained as (i) in children, bone formation actively takes place, but not in adults; thus the study in children highlights the effect of bone loss in CKD subjects, and/or (ii) unlike adults, children with CKD experience vascular calcification to a much lesser extent than their adult counterparts; thus it is possible that the increase in serum Ca isotopes as a result of vascular calcification is minimal in children with CKD. While we note that the different groups are not age matched, since the median age of the dialysis group is greater than that of controls, and because bone resorption increases with age thus decreasing serum δ$\frac{44}{42}$Ca values,11,12 if we were able to correct each group for difference in bone mineral balance, the difference in serum δ$\frac{44}{42}$Ca values between the control and dialysis group could be even greater.
The Ca isotope composition of blood tracks changes in calcium mineral balance in the body, e.g. increasing as a result of bone mineralization.11 Similarly, the formation of Ca deposits in blood vessels could explain the observed increased serum δ$\frac{44}{42}$Ca in subjects showing signs of medial calcification (as defined by a ba-PWV value exceeding the threshold defined by age and sex18). The sensitivity of serum Ca isotopes to vascular calcification is surprisingly high, since arteriosclerosis would represent a small variation in Ca balance compared to bone mineralization.
Vascular calcification is common in CKD and is one of the strongest predictors of cardiovascular events and mortality.33 Its presence contributes to hypertension, increase PWV and left ventricular hypertrophy which all contribute to cardiovascular risk. While many risk factors of vascular calcification are themselves present in patients with CKD, such as older age, diabetes, hypertension, and smoking, it remains more common in the CKD population compared to a similarly aged population. The strong association between vascular calcification and risk of adverse events has generated a strong interest in interventions that can prevent progression or regress these lesions. However, they have been met with limited success. Vascular calcification is not due to a single entity but is rather a common endpoint to multiple pathological processes present in CKD such as hyperphosphatemia, hyperparathyroidism, vitamin D deficiency, hypertension, and disrupted calcium balance. A single drug intervention may therefore yield limited response and to date there is insufficient evidence and conflicting data that any intervention mitigates the risk of vascular calcification.34 Another challenge in targeting vascular calcification in CKD is the tools presently available for diagnosis. Imaging such as computed tomography and PWV are all able to diagnose vascular calcification in CKD and determine risk of cardiovascular events and mortality.35,36 However, they can only diagnose vascular calcification after permanent irreversible changes have set in, making interventions difficult to implement. A need to identify a marker for vascular calcification in the early stages, before clinical detection using traditional tools has occurred, is therefore needed. The physiological processes that disrupt *Ca homeostasis* occur early in CKD, before vascular calcification sets in. We have shown that serum Ca isotope measurements can detect vascular calcification, especially medial artery calcification in patients with CKD, therefore identifying it as a potential marker.
The need for a biomarker that can detect vascular calcification is not limited to being able to identify the disorder prior to irreversible changes occurring. It should also be easily measured for use as a screening population tool. Such a biomarker has the potential to allow for intervention, such as life-style changes or medical therapeutics, which may reverse or prevent the progression of vascular calcification early in the process. It can also be utilized as a tool to screen a large segment of the population who are at risk, such as those with CKD. Our study shows that serum Ca isotopes have the potential to fill this role since it utilizes a blood sample for detection of vascular calcification. However, the main disadvantage is that this method is currently only available in specialized laboratories.
Our main limitations are the small sample size and the cohort design with no prospective component. We recruited limited participants with early CKD compared to controls and KF. Since our study is a cohort analysis, we were unable to demonstrate causality with association in this cohort analysis. Finally, this study was only preformed in a single institution so the results will need to be verified in other settings and countries. Differing dietary practices could notably affect the baseline Ca isotope composition of serum and therefore potentially dim/reduce the sensitivity of serum Ca isotopes to vascular calcification. Culturally related dietary habits primarily linked to dairy products intake contribute interindividual variability (e.g. Tacail et al.22). Interestingly, for each group from the same Australian cohort, our results show a distribution of δ$\frac{44}{42}$Ca values with a typical spread (serum δ$\frac{44}{42}$Ca IQR ∼ 0.25‰) that is comparable to what has been reported in low and high dairy consuming populations (bone δ$\frac{44}{42}$Ca IQR = 0.2 and 0.3‰, respectively22) or in the control and osteoporotic women population from a previous clinical trial (urine δ$\frac{44}{42}$Ca IQR ∼ 0.2‰11). This suggests that the characterization of population baselines for groups with documented and homogeneous dietary habits would allow developing serum Ca isotopes as biomarkers for vascular calcification, in a similar way urine δ$\frac{44}{42}$*Ca is* used to track osteoporosis in women (e.g. Eisenhauer et al.11). Furthermore, at the scale of a cohort of individuals from a given culturally homogeneous group (women from Northern Germany), serum and urine δ$\frac{44}{42}$Ca values were shown to be only moderately influenced by intake of Ca from milk while primarily influenced sensitive to osteoporosis.11. Here, we showed that there is no association between serum δ$\frac{44}{42}$Ca and dietary Ca (Fig. 4), suggesting that diet has little influence on serum Ca isotopes in our cohort.
## Conclusion
This study set out to examine the association between naturally occurring calcium isotopes in serum and urine and its association with vascular calcification in a CKD population. Our results show that Ca isotopes in serum display a strong association to markers of medial calcification in patients with CKD (but not Ca isotopes in spot urine samples). ROC curve analysis indicates that Ca isotopes in serum perform well as a tool to detect medial artery calcification (sensitivity: $81.8\%$, specificity: $77.3\%$, AUC: 0.818). Thus, in the future, medial artery calcification could perhaps be detected in patients with CKD from a small (≤1 mL) blood sample. Confirmation will be required by conducting studies of other cohorts in addition to a prospective study design.
## Authors contributions
AD, TT, and KL designed the study. KL, ML, AF, and AB undertook recruitment. AF and AB undertook sample collection. AD, AF, AB, and FD performed the analyses. AD, KL, and HCH interpreted the results. AD, KL, HCH, TT, and AF wrote the manuscript.
## Funding
IHMRI Clinical Translation Grant and Univeristy of Wollongong, Faculty of SMAH Small Grant.
## Conflicts of interest
The authors declare no conflicts of interest.
## Data availability
All data are incorporated into the article and its online supplementary material.
## References
1. **Global, Regional, and National Burden of Chronic Kidney Disease, 1990-2017: a Systematic Analysis for the Global Burden of Disease Study 2017**. *Lancet* (2020) **395** 709-733. DOI: 10.1016/S0140-6736(20)30045-3
2. Jankowski J., Floege J., Fliser D., Böhm M., Marx N.. **Cardiovascular Disease in Chronic Kidney Disease**. *Circulation* (2021) **143** 1157-1172. DOI: 10.1161/CIRCULATIONAHA.120.050686
3. Palit S., Kendrick J.. **Vascular Calcification in Chronic Kidney Disease: Role of Disordered Mineral Metabolism**. *Curr. Pharm. Des.* (2014) **20** 5829-5833. DOI: 10.2174/1381612820666140212194926
4. Ruderman I., Holt S. G., Hewitson T. D., Smith E. R., Toussaint N. D.. **Current and Potential Therapeutic Strategies for the Management of Vascular Calcification in Patients with Chronic Kidney Disease Including Those on Dialysis**. *Semin. Dial.* (2018) **31** 487-499. DOI: 10.1111/sdi.12710
5. Krishnasamy R., Pedagogos E.. **Should Nephrologists Consider Vascular Calcification Screening?**. *Nephrology* (2017) **22** 31-33. DOI: 10.1111/nep.13019
6. Toepfer E. T., Rott J., Bartosova M., Kolevica A., Machuca-Gayet I., Heuser A., Rabe M., Shroff R., Bacchetta J., Zarogiannis S. G., Eisenhauer A., Schmitt C. P.. **Calcium Isotope Fractionation by Osteoblasts and Osteoclasts, Across Endothelial and Epithelial Cell Barriers and with Binding to Proteins**. *Am. J. Physiol. Regul. Integr. Comp. Physiol.* (2021) **321** R29-R40. DOI: 10.1152/ajpregu.00334.2020
7. Tacail T.. **Physiologie isotopique du calcium chez les mammifères**. *Ecole Normale Supérieure de Lyon* (2017) 344
8. Hassler A., Martin J. E., Ferchaud S., Grivault D., Le Goff S., Albalat E., Hernandez J.-A., Tacail T., Balter V.. **Lactation and Gestation Controls on Calcium Isotopic Compositions in a Mammalian Model**. *Metallomics* (2021) **13**. DOI: 10.1093/mtomcs/mfab019
9. Morgan J. L., Skulan J. L., Gordon G. W., Romaniello S. J., Smith S. M., Anbar A. D.. **Rapidly Assessing Changes in Bone Mineral Balance Using Natural Stable Calcium Isotopes**. *Proc. Natl. Acad. Sci. USA* (2012) **109** 9989-9994. DOI: 10.1073/pnas.1119587109
10. Heuser A., Frings-Meuthen P., Rittweger J., Galer S. J. G.. **Calcium Isotopes in Human Urine as a Diagnostic Tool for Bone Loss: Additional Evidence for Time Delays in Bone Response to Experimental Bed Rest**. *Front Physiol.* (2019) **10** 12. DOI: 10.3389/fphys.2019.00012
11. Eisenhauer A., Müller M., Heuser A., Kolevica A., Glüer C. C., Both M., Laue C., Hehn U. v., Kloth S., Shroff R., Schrezenmeir J.. **Calcium Isotope Ratios in Blood and Urine: a New Biomarker for the Diagnosis of Osteoporosis**. *Bone Rep.* (2019) **10** 100200. DOI: 10.1016/j.bonr.2019.100200
12. Shroff R., Fewtrell M., Heuser A., Kolevica A., Lalayiannis A., McAlister L., Silva S., Goodman N., Schmitt C. P., Biassoni L., Rahn A., Fischer D. C., Eisenhauer A.. **Naturally Occurring Stable Calcium Isotope Ratios in Body Compartments Provide a Novel Biomarker of Bone Mineral Balance in Children and Young Adults**. *J. Bone Miner. Res.* (2021) **36** 133-142. DOI: 10.1002/jbmr.4158
13. Heuser A., Eisenhauer A.. **A Pilot Study on the Use of Natural Calcium Isotope (44Ca/40Ca) Fractionation in Urine as a Proxy for the Human Body Calcium Balance**. *Bone* (2010) **46** 889-896. DOI: 10.1016/j.bone.2009.11.037
14. Shroff R., Lalayiannis A. D., Fewtrell M., Schmitt C. P., Bayazit A., Askiti V., Jankauskiene A., Bacchetta J., Silva S., Goodman N., McAlister L., Biassoni L., Crabtree N., Rahn A., Fischer D.-C., Heuser A., Kolevica A., Eisenhauer A.. **Naturally Occurring Stable Calcium Isotope Ratios Are a Novel Biomarker of Bone Calcium Balance in Chronic Kidney Disease**. *Kidney Int.* (2022) **102** 613-623. DOI: 10.1016/j.kint.2022.04.024
15. Yamada S., Giachelli C. M.. **Vascular Calcification in CKD-MBD: Roles for Phosphate, FGF23, and Klotho**. *Bone* (2017) **100** 87-93. DOI: 10.1016/j.bone.2016.11.012
16. Channon M. B., Gordon G. W., Morgan J. L., Skulan J. L., Smith S. M., Anbar A. D.. **Using Natural, Stable Calcium Isotopes of Human Blood to Detect and Monitor Changes in Bone Mineral Balance**. *Bone* (2015) **77** 69-74. DOI: 10.1016/j.bone.2015.04.023
17. Tomiyama H., Yamashina A.. **The Application of Brachial-Ankle Pulse Wave Velocity as a Clinical Tool for Cardiovascular Risk Assessment**. *Hypertension* (2012) **60** e40. DOI: 10.1161/HYPERTENSIONAHA.112.201806
18. Chen S. C., Chang J.-M., Liu W.-C., Tsai Y.-C., Tsai J. C., Hsu P.-C., Lin T.-H., Lin M.-Y., Su H.-M., Hwang S.-J., Chen H. C.. **Brachial-Ankle Pulse Wave Velocity and Rate of Renal Function Decline and Mortality in Chronic Kidney Disease**. *Clin. J. Am. Soc. Nephrol.* (2011) **6** 724-732. DOI: 10.2215/CJN.07700910
19. Romaniello S. J., Field M. P., Smith H. B., Gordon G. W., Kim M. H., Anbar A. D.. **Fully Automated Chromatographic Purification of Sr and Ca for Isotopic Analysis**. *J. Anal. At. Spectrom.* (2015) **30** 1906-1912. DOI: 10.1039/C5JA00205B
20. Albarède F., Beard B.. **Analytical Methods for Non-Traditional Isotopes**. *Rev. Mineral. Geochem.* (2004) **55** 113-152. DOI: 10.2138/gsrmg.55.1.113
21. Tacail T., Albalat E., Télouk P., Balter V.. **A Simplified Protocol for Measurement of Ca Isotopes in Biological Samples**. *J. Anal. At. Spectrom.* (2014) **29** 529. DOI: 10.1039/c3ja50337b
22. Tacail T., Martin J. E., Herrscher E., Albalat E., Verna C., Ramirez-Rozzi F., Clark G., Valentin F., Balter V.. **Quantifying the Evolution of Animal Dairy Intake in Humans Using Calcium Isotopes**. *Quat. Sci. Rev.* (2021) **256** 106843. DOI: 10.1016/j.quascirev.2021.106843
23. Morgan J. L., Gordon G. W., Arrua R. C., Skulan J. L., Anbar A. D., Bullen T. D.. **High-Precision Measurement of Variations in Calcium Isotope Ratios in Urine by Multiple Collector Inductively Coupled Plasma Mass Spectrometry**. *Anal. Chem.* (2011) **83** 6956-6962. DOI: 10.1021/ac200361t
24. Tanaka Y. K., Yajima N., Higuchi Y., Yamato H., Hirata T.. **Calcium Isotope Signature: New Proxy for Net Change in Bone Volume for Chronic Kidney Disease and Diabetic Rats**. *Metallomics* (2017) **9** 1745-1755. DOI: 10.1039/C7MT00255F
25. Tacail T., Télouk P., Balter V.. **Precise Analysis of Calcium Stable Isotope Variations in Biological Apatites Using Laser Ablation MC-ICPMS**. *J. Anal. At. Spectrom.* (2016) **31** 152-162. DOI: 10.1039/C5JA00239G
26. Fisher R A.. *Statistical Methods for Research Workers* (1926)
27. Zou K. H., o Malley A. J., Mauri L.. **Receiver-Operating Characteristic Analysis for Evaluating Diagnostic Tests and Predictive Models**. *Circulation* (2007) **115** 654-657. DOI: 10.1161/CIRCULATIONAHA.105.594929
28. Robin X., Turck N., Hainard A., Tiberti N., Lisacek F., Sanchez J.-C., Müller M.. **pROC: an Open-Source Package for R and S+ to Analyze and Compare ROC Curves**. *BMC Bioinf.* (2011) **12** 77. DOI: 10.1186/1471-2105-12-77
29. Hanley J. A., McNeil B. J.. **The Meaning and Use of the Area under a Receiver Operating Characteristic (ROC) Curve**. *Radiology* (1982) **143** 29-36. DOI: 10.1148/radiology.143.1.7063747
30. Liu C. S., Li C.-I., Shih C.-M., Lin W.-Y., Lin C.-H., Lai S. W., Li T.-C., Lin C.-C.. **Arterial Stiffness Measured as Pulse Wave Velocity is Highly Correlated with Coronary Atherosclerosis in Asymptomatic Patients**. *J. Atheroscler. Thromb.* (2011) **18** 652-658. DOI: 10.5551/jat.7021
31. Skulan J., DePaolo D. J.. **Calcium Isotope Fractionation between Soft and Mineralized Tissues as a Monitor of Calcium Use in Vertebrates**. *Proc. Natl. Acad. Sci. USA* (1999) **96** 13709-13713. DOI: 10.1073/pnas.96.24.13709
32. Gordon G. W., Monge J., Channon M. B., Wu Q., Skulan J. L., Anbar A. D., Fonseca R.. **Predicting Multiple Myeloma Disease Activity by Analyzing Natural Calcium Isotopic Composition**. *Leukemia* (2014) **28** 2112-2115. DOI: 10.1038/leu.2014.193
33. Liu M., Li X. C., Lu L., Cao Y., Sun R. R., Chen S., Zhang P. Y.. **Cardiovascular Disease and Its Relationship with Chronic Kidney Disease**. *Eur. Rev. Med. Pharmacol. Sci.* (2014) **18** 2918-2926. PMID: 25339487
34. Xu C., Smith E., Tiong M., Ruderman I., Toussaint N.. **Interventions to Attenuate Vascular Calcification Progression in Chronic Kidney Disease: a Systematic Review of Clinical Trials**. *J. Am. Soc. Nephrol.* (2022) **33** 1011-1032. DOI: 10.1681/ASN.2021101327
35. Chen J., Budoff M. J., Reilly M. P., Yang W., Rosas S. E., Rahman M., Zhang X., Roy J. A., Lustigova E., Nessel L., Ford V., Raj D., Porter A. C., Soliman E. Z., Wright J. T., Wolf M., He J.. **Coronary Artery Calcification and Risk of Cardiovascular Disease and Death among Patients with Chronic Kidney Disease**. *JAMA Cardiol* (2017) **2** 635-643. DOI: 10.1001/jamacardio.2017.0363
36. Krishnasamy R., Tan S. J., Hawley C. M., Johnson D. W., Stanton T., Lee K., Mudge D. W., Campbell S., Elder G. J., Toussaint N. D., Isbel N. M.. **Progression of Arterial Stiffness Is Associated with Changes in Bone Mineral Markers in Advanced CKD**. *BMC Nephrol.* (2017) **18** 281. DOI: 10.1186/s12882-017-0705-4
37. Wickham H.. *ggplot2: Elegant Graphics for Data Analysis* (2016)
|
---
title: Impaired ketogenesis is associated with metabolic-associated fatty liver disease
in subjects with type 2 diabetes
authors:
- Sejeong Lee
- Jaehyun Bae
- Doo Ri Jo
- Minyoung Lee
- Yong-ho Lee
- Eun Seok Kang
- Bong-Soo Cha
- Byung-Wan Lee
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9989459
doi: 10.3389/fendo.2023.1124576
license: CC BY 4.0
---
# Impaired ketogenesis is associated with metabolic-associated fatty liver disease in subjects with type 2 diabetes
## Abstract
### Aims
The ketogenic pathway is an effective mechanism by which the liver disposes of fatty acids (FAs) to the peripheral tissues. Impaired ketogenesis is presumed to be related to the pathogenesis of metabolic-associated fatty liver disease (MAFLD), but the results of previous studies have been controversial. Therefore, we investigated the association between ketogenic capacity and MAFLD in subjects with type 2 diabetes (T2D).
### Methods
A total of 435 subjects with newly diagnosed T2D was recruited for the study. They were classified into two groups based on median serum β-hydroxybutyrate (β-HB) level: intact vs. impaired ketogenesis groups. The associations of baseline serum β-HB and MAFLD indices of hepatic steatosis index, NAFLD liver fat score (NLFS), Framingham Steatosis index (FSI), Zhejian University index, and Chinese NAFLD score were investigated.
### Results
Compared to the impaired ketogenesis group, the intact ketogenesis group showed better insulin sensitivity, lower serum triglyceride level, and higher low-density lipoprotein-cholesterol and glycated hemoglobin levels. Serum levels of liver enzymes were not different between the two groups. Of the hepatic steatosis indices, NLFS (0.8 vs. 0.9, $$p \leq 0.045$$) and FSI (39.4 vs. 47.0: $$p \leq 0.041$$) were significantly lower in the intact ketogenesis group. Moreover, intact ketogenesis was significantly associated with lower risk of MAFLD as calculated by FSI after adjusting for potential confounders (adjusted odds ratio 0.48, $95\%$ confidence interval 0.25-0.91, $$p \leq 0.025$$).
### Conclusions
Our study suggests that intact ketogenesis might be associated with decreased risk of MAFLD in T2D.
## Introduction
Non-alcoholic fatty liver disease (NAFLD) is a liver condition ranging from simple steatosis to inflammation or fibrosis in the absence of excessive alcohol consumption [1]. Hepatic triglycerides (TGs) accumulate in the liver mainly through de novo lipogenesis or delivery of fatty acids (FAs) from peripheral tissues in the NAFLD or insulin resistant settings [2, 3]. Recently, metabolic-associated fatty liver disease (MAFLD) was suggested as a term that more accurately reflects the pathogenesis of this type of fatty liver disease [4].
In the normal liver, excess delivered FAs undergo β-oxidation, yielding acetyl-CoA [5]. The majority of acetyl-CoA is condensed in the ketogenic pathway to form ketone bodies, mostly β-hydroxybutyrate (βHB) and acetoacetate, which are exported to the peripheral tissues and are utilized as efficient fuels. Through β-oxidation and the ketogenic pathway, the normal liver has considerable capacity to dispose of delivered FAs [6].
In this context, it can be expected that ketogenesis is relatively downregulated in patients with MAFLD. Studies have reported that ketogenesis is suppressed in patients with fatty liver disease (7–9). One previous study using isotope tracers reported that acetyl-CoA oxidation through the tricarboxyclic acid (TCA) cycle was upregulated in MAFLD patients, whereas the ketogenic pathway from acetyl-CoA was markedly reduced [7]. This suggested that impaired ketogenesis from acetyl-CoA in the liver might be important in the development of MAFLD. However, clinical data on the association between ketogenesis and MAFLD are controversial, as there were also previous studies that showed increased ketone levels in patients with fatty liver disease [10, 11].
In the present study, we aimed to investigate whether impaired ketogenesis is related to MAFLD in patients with type 2 diabetes (T2D). To this end, we recruited newly diagnosed T2D patients and investigated the association between ketogenic capacity based on blood βHB level and MAFLD status using various indices of fatty liver disease.
## Study design and population
For this retrospective cross-sectional study, we registered the study subject on the basis of our previous studies [12, 13]. Since 2009, we have built a cohort, the diabetes registry of Severance Diabetes Center (a tertiary care hospital in Seoul, Korea) with the patients who underwent a standardized mixed-meal stimulation test on their first visit to our diabetes center. The electronic medical records of patients from April 2017 to March 2022 were reviewed. In the registry protocol, we routinely collected blood samples at 0 and 90 minutes (basal and stimulated, respectively) for glucose, insulin, and C-peptide analyses. Inclusion criteria were patients aged ≥19 years with newly diagnosed T2D based on the 2019 Korean Diabetes Association guidelines [14] and measured serum βHB, which has been available at our center since 2017 from the initial visit. We excluded patients who had undergone organ transplantation or chemotherapy, steroid users, patients who had taken antidiabetic drugs prior to initial blood sampling, and those who visited the emergency room due to hyperglycemia.
A total of 435 patients was ultimately included for analysis in this study. Subjects were classified into two groups based on median initial serum βHB level: those with sufficient ketogenic capacity (intact ketogenesis group) and those without (impaired ketogenesis group).
This study was approved by the independent institutional review board of Severance Hospital [4-2022-1101].
## Measurements and variables
Patient demographics and clinical and biochemical measurements were collected during the study period. The variables gathered were age, sex, body mass index (BMI), use of an antidiabetic drug, hypertension, and blood chemistry.
MAFLD was assessed using previous validated liver steatosis prediction models: hepatic steatosis index (HSI), NAFLD liver fat score (NLFS), Framingham Steatosis Index (FSI), Zhejian University (ZJU) index, and Chinese NAFLD score. The equations are described in Supplementary Table 1. HSI and NLFS are well-validated models to detect hepatic steatosis in the general population [15]. FSI is a clinical model that includes metabolic parameters such as BMI, diabetes diagnosis, and good discrimination in the National Health and Nutrition Examination Survey III cohort [16]. The diagnostic performance of the ZJU index and Chinese NAFLD score was verified in several studies, especially in an Asian cohort including Japanese and Chinese subjects [17, 18].
Serum βHB, the most abundant form of ketone body, was measured before initiating diabetes medication to determine the subject’s ketogenic capacity. Fasting serum βHB concentration was assessed by an enzymatic assay using a commercial reagent from Landox Laboratories Ltd. (County Antrim, UK) and an Atellica CH 930 analyzer (Siemens Healthcare Diagnostics, Marburg, Germany). βHB values measured below the lower limit of the assay were expressed as 0. Blood samples for measuring glucometabolic parameters including serum glucose, insulin, and cholesterol in the fasting state were obtained after overnight fasting. Low-density lipoprotein cholesterol (LDL-C) levels were calculated using Friedewald’s equation in cases of subjects without actual LDL-C measurements, if their blood TG levels were below 400 mg/dL [19]. Postprandial insulin level was also measured 90 minutes after the mixed-meal test (Mediwell Diabetic Meal; Meail Dairies Co., Yeongdong-gun, Chungbuk, Korea). The estimated glomerular filtration rate (eGFR) was calculated based on the four-variable Modification of Diet in Renal Disease study equation. Insulin resistance and pancreatic β-cell function were assessed using the homeostasis model assessment of insulin resistance (HOMA-IR) index and HOMA- β.
## Statistical analysis
Glucometabolic parameters and biomarker-based indices to assess hepatic steatosis were compared between the intact ketogenesis and impaired ketogenesis groups. Normally distributed continuous variables were presented as mean (standard deviation), and non-normal continuous variables were presented as median (interquartile range [IQR]). The normality of continuous variables was assessed by the Shapiro-Wilk test. Categorical variables were presented as number with percentage (%). The difference between groups was evaluated using Student’s t-test for continuous variables with normal distribution and Mann-Whitney U-test for continuous variables with non-normal distribution. The frequencies of categorical variables were compared using Pearson’s Chi-square test. The correlation between serum βHB level and each liver steatosis prediction model was assessed by Spearman’s correlation coefficient, which is a statistical method for non-normally distributed variables. In addition, we performed regression analysis to evaluate the clinical significance of initial βHB level for prediction of hepatic steatosis. Logistic regression analysis, a statistical technique used to predict the relationship between independent variables and a binary dependent variable was performed with MAFLD occurrence based on the cut-off values of each MAFLD indices as a dependent variable. The ketogenic capacity used as independent variable in the logistic regression analysis was defined as being divided into intact/impaired based on the median serum βHB level. In the adjusted model, age, sex, BMI, glycated hemoglobin (HbA1c), low-density lipoprotein cholesterol (LDL-C), HOMA-IR, and HOMA-β were adjusted. P-values <0.05 were considered statistically significant. Statistical analyses were performed using R software version 3.6.3 (R Project for Statistical Computing, Vienna, Austria).
## Clinical and laboratory characteristics of patients
The baseline characteristics of 435 newly diagnosed T2D patients categorized according to level of βHB are shown in Table 1. Subjects were divided into intact ($$n = 226$$) and impaired ($$n = 209$$) ketogenesis group. The median age of the study subjects was 54 years (IQR, 44 to 63) and $62.8\%$ were men. The median serum βHB level was 0.11 (0.0-0.2) mmol/L. Compared to the impaired ketogenesis group, patients with intact ketogenesis were significantly younger and showed non-significantly lower BMI. There were no significant differences in aspartate aminotransferase, alanine aminotransferase, and eGFR between the two groups. Fasting blood glucose level and HbA1c level was significantly higher in the intact group than the impaired group. Moreover, the intact ketogenesis group showed markedly lower fasting and postprandial insulin levels as well as lower HOMA-β and HOMA-IR, suggesting decreased insulin secretory function, better insulin sensitivity, and poor glycemic control. The blood level of TG was lower and LDL-C was higher in the intact ketogenesis group.
**Table 1**
| Unnamed: 0 | Intact ketogenesis | Impaired ketogenesis | P-value |
| --- | --- | --- | --- |
| | (N=226) | (N=209) | |
| Demographic | Demographic | Demographic | Demographic |
| Age (years) | 50.5 [41.0–63.0] | 57.0 [48.0–64.0] | 0.001 |
| Sex (Male, n (%)) | 142 (62.8%) | 131 (62.7%) | >0.999 |
| HTN | 82 (36.3%) | 90 (43.1%) | 0.178 |
| BMI (kg/m2) | 25.8 [23.2–28.3] | 26.5 [24.2–28.9] | 0.146 |
| Biochemistry | Biochemistry | Biochemistry | Biochemistry |
| AST (IU/L) | 24.5 [19.0-35.0] | 27.0 [19.0–36.0] | 0.325 |
| ALT (IU/L) | 28.5 [19.0-46.0] | 30.0 [21.0–44.0] | 0.755 |
| Total cholesterol (mg/dL) | 196.0 [157.0–234.0] | 189.0 [156.5–218.0] | 0.174 |
| TG (mg/dL) | 135.5 [96.0–214.0] | 150.0 [109.0–233.0] | 0.039 |
| HDL-C (mg/dL) | 43.0 [37.0–53.0] | 44.0 [39.0–52.0] | 0.435 |
| LDL-C (mg/dL) | 112.3 [83.5–154.0] | 105.8 [77.0–134.4] | 0.021 |
| TG/LDL-C | 1.2 [0.8–1.9] | 1.4 [1.0–2.2] | 0.009 |
| eGFR (ml/min/1.73 m²) | 89.8 [77.0-105.5] | 90.0 [77.0–107.0] | 0.806 |
| uACR (mg/g creatinine) | 10.7 [5.9-25.5] | 11.4 [5.6–29.7] | 0.734 |
| Gluco-metabolic parameters | Gluco-metabolic parameters | Gluco-metabolic parameters | Gluco-metabolic parameters |
| Fasting glucose (mg/dL) | 143.0 [120.0–220.0] | 136.0 [120.0–172.0] | 0.046 |
| HbA1c (%) | 8.9 [7.3-11.0] | 7.6 [7.0–9.2] | <0.001 |
| Fasting C-peptide (ng/mL) | 2.4 [1.9-3.0] | 2.7 [2.2-3.6] | <0.001 |
| Postprandial C-peptide (ng/mL) | 5.3 [3.8-6.9] | 6.4 [4.8-8.4] | <0.001 |
| Fasting insulin (μIU/mL) | 9.3 [6.1-13.7] | 11.7 [7.6–18.4] | <0.001 |
| Postprandial insulin (μIU/mL) | 40.8 [22.9-62.6] | 52.4 [32.0–89.1] | <0.001 |
| HOMA-IR | 3.7 [2.2–5.6] | 4.3 [2.8–6.6] | 0.009 |
| HOMA-β | 36.4 [18.7–59.0] | 49.8 [22.6–85.2] | <0.001 |
| βHB (mmol/L) | 0.2 [0.1–0.4] | 0.0 [0.0–0.0] | <0.001 |
| MAFLD indices | MAFLD indices | MAFLD indices | MAFLD indices |
| Hepatic steatosis index | 38.3 [34.1-42.1] | 38.7 [34.6–42.4] | 0.666 |
| Hepatic steatosis index ≥36 | 147 (65.0%) | 141 (67.5%) | 0.666 |
| NAFLD liver fat score | 0.8 [0.1-1.3] | 0.9 [0.5–1.4] | 0.045 |
| NAFLD liver fat score >-0.64 | 219 (96.9%) | 205 (98.1%) | 0.631 |
| Framingham Steatosis Index | 39.4 [20.4-64.3] | 47.0 [98.6–65.2] | 0.041 |
| Framingham Steatosis Index ≥23 | 165 (73.0%) | 171 (81.8%) | 0.038 |
| Zhejian University index | 41.0 [36.1-46.4] | 40.1 [37.0–43.8] | 0.257 |
| Zhejian University index >38 | 131 (58.0%) | 130 (62.2%) | 0.422 |
| Chinese NAFLD score | -0.6 [-1.1–0.0] | -0.4 [-1.0–0.1] | 0.264 |
| Chinese NAFLD score >-0.79 | 133 (58.8%) | 144 (68.9%) | 0.038 |
The values of the five hepatic steatosis indices of the two groups are described in Figure 1. Overall, the intact ketogenesis group showed a tendency to lower hepatic steatosis indices compared to the impaired ketogenesis group. However, significant differences between the groups were found only for NLFS and FSI. In NLFS, the index values of the intact ketogenesis group and impaired ketogenesis group were 0.8 vs. 0.9 ($$p \leq 0.045$$), and the index value of FSI in groups was 39.4 vs. 47.0 ($$p \leq 0.041$$). Additionally, the incidence of MAFLD, based on the cut-off values, was lower in the intact ketogenesis group than in the impaired ketogenesis group. Based on assessment by FSI, 165 ($73.0\%$) subjects fulfilled the definition of MAFLD in the intact ketogenesis group whereas 171 ($81.8\%$) subjects in the impaired ketogenesis group ($$p \leq 0.038$$). According to the Chinese NAFLD score, 133 ($58.8\%$) in the intact group vs. 144 ($68.9\%$) in the impaired group, indicating a significant difference between the two groups ($$p \leq 0.038$$).
**Figure 1:** *Liver steatosis indices of the intact ketogenesis group and the impaired ketogenesis group: (A) Hepatic steatosis index, (B) NAFLD liver fat score, (C) Framingham Steatosis Index, (D) Zhejian University index, and (E) Chinese NAFLD score. ns, statistically not significant (P-value ≥ 0.05)*, statistically significant (P-value < 0.05).*
## Serum βHB level is correlated with lower hepatic steatosis indices
The correlations between serum βHB level and the hepatic steatosis indices were analyzed using Spearman’s correlation coefficient (Table 2). There was a significant negative correlation between serum βHB level and hepatic steatosis as assessed by NLFS (r= -0.103, $$p \leq 0.032$$) and FSI (r= -0.106, $$p \leq 0.028$$). A negative correlation was observed between βHB level and hepatic steatosis based on the other indices, but the relationship was not statistically significant.
**Table 2**
| Unnamed: 0 | In all subjects (N=435) | In all subjects (N=435).1 |
| --- | --- | --- |
| | β-hydroxybutyrate | β-hydroxybutyrate |
| | r | p-value |
| Hepatic steatosis index | -0.051 | 0.281 |
| NAFLD liver fat score | -0.103 | 0.032 |
| Framingham Steatosis Index | -0.106 | 0.028 |
| Zhejian University index | -0.068 | 0.155 |
| Chinese NAFLD score | -0.058 | 0.226 |
## Ketogenic capacity is associated with risk of MAFLD incidence
Multivariable logistic regression analysis of the relationship between MAFLD and ketogenic capacity is shown in Table 3. To determine the independence of ketogenic capacity as a risk factor for MAFLD, multivariable logistic regression analysis was performed adjusting for covariates of age, sex, BMI, HbA1c, LDL-C, HOMA-IR, and HOMA-β. The analysis showed ketogenic capacity to maintain a significant association with MAFLD as calculated by FSI ([Odds ratio, 0.48; $95\%$ confidence interval 0.25 to 0.91; $$P \leq 0.025$$]. With other hepatic steatosis indices, ketogenic capacity tended to be associated with MAFLD incidence, but the relationship was not statistically significant.
**Table 3**
| In all subjects (N=435) | Hepatic steatosis index | Hepatic steatosis index.1 | NAFLD liver fat score | NAFLD liver fat score.1 | Framingham Steatosis Index | Framingham Steatosis Index.1 |
| --- | --- | --- | --- | --- | --- | --- |
| In all subjects (N=435) | Adjusted OR | P-value | Adjusted OR | P-value | Adjusted OR | P-value |
| In all subjects (N=435) | (95% CI) | P-value | (95% CI) | P-value | (95% CI) | P-value |
| Age (years) | 0.96 (0.94–0.99) | 0.004 | 0.99 (0.94–1.04) | 0.634 | 1.02 (0.99–1.04) | 0.212 |
| Sex (female vs. male) | 2.03 (1.08–3.90) | 0.030 | 1.03 (0.28–4.26) | 0.968 | 0.57 (0.31–1.07) | 0.082 |
| BMI (kg/m2) | 2.30 (1.95–2.79) | <0.001 | 1.08 (0.88–1.31) | 0.473 | 1.93 (1.67–2.27) | <0.001 |
| HbA1c (%) | 1.00 (0.84–1.19) | 0.977 | 0.89 (0.62–1.31) | 0.545 | 1.34 (1.12–1.62) | 0.002 |
| LDL-C (mg/dL) | 1.00 (0.99–1.01) | 0.508 | 1.00 (0.99–1.02) | 0.695 | 1.00 (0.99–1.00) | 0.595 |
| HOMA-IR | 1.02 (0.95–1.10) | 0.524 | 1.10 (0.89–1.58) | 0.536 | 1.00 (0.93–1.09) | 0.941 |
| HOMA-β | 1.00 (0.99–1.00) | 0.160 | 1.02 (0.99–1.06) | 0.225 | 1.00 (0.99–1.01) | 0.925 |
| Ketogenic capacity* | 1.35 (0.77–2.39) | 0.707 | 0.80 (0.19–2.98) | 0.753 | 0.48 (0.25–0.91) | 0.025 |
| | Zhejian University index | Zhejian University index | Chinese NAFLD score | Chinese NAFLD score | | |
| | Adjusted OR | P-value | Adjusted OR | P-value | | |
| | (95% CI) | P-value | (95% CI) | P-value | | |
| Age (years) | 0.98 (0.96–1.01) | 0.146 | 0.99 (0.96–1.02) | 0.301 | | |
| Sex (female vs. male) | 1.38 (0.74–2.58) | 0.315 | 0.75 (0.39–1.44) | 0.389 | | |
| BMI (kg/m2) | 1.81 (1.59–2.08) | <0.001 | 2.59 (2.15–3.21) | <0.001 | | |
| HbA1c (%) | 1.56 (1.29–1.89) | <0.001 | 1.03 (0.86–1.22) | 0.782 | | |
| LDL-C (mg/dL) | 1.01 (1.00–1.01) | 0.119 | 1.00 (0.99-1.01) | 0.406 | | |
| HOMA-IR | 1.43 (1.23–1.69) | <0.001 | 1.04 (0.97-1.13) | 0.268 | | |
| HOMA-β | 0.98 (0.97–0.99) | <0.001 | 0.99 (0.99-1.00) | 0.101 | | |
| Ketogenic capacity | 0.54 (0.29–1.00) | 0.051 | 0.57 (0.29-1.08) | 0.089 | | |
## Discussion
In this cross-sectional study, we focused on ketogenic capacity and MAFLD with glucometabolic disorders in patients with T2D. We found that T2D patients with intact ketogenesis showed better hepatic steatosis indices, especially FSI and NLFS. After multivariable logistic regression analysis, ketogenic capacity showed significance in predicting the extent of MAFLD as calculated by FSI. Most of the hepatic steatosis indices other than FSI tended to be correlated with ketogenic capacity but did not show significance in regression analysis. This may be because our study included the large number of newly diagnosed and drug-naïve T2D patients with mild or early stage of MAFLD, as shown in the normal mean liver enzyme levels and slightly increased hepatic steatosis indices. In addition, since the variables used in each index are different, it is believed that differences in sensitivity may have occurred.
In the current study, subjects with intact ketogenesis showed significantly lower insulin resistance without a significant obesity difference compared to those with impaired ketogenesis. The association between insulin resistance and MAFLD has already been well established, and it is thought to have influenced on the pathogenesis of MAFLD in the impaired ketogenesis group. In addition, relative hyperinsulinemia in the impaired ketogenesis group might be due not only to insulin resistance but also to lower hepatic insulin clearance, which has been previously reported to be associated with MAFLD [20]. With respect to dyslipidemia in intact ketogenic patients with T2D, higher level of LDL-C and lower TG level were observed in the intact ketogenesis group. The most common pattern of dyslipidemia in T2D is hypertriglyceridemia and reduced HDL cholesterol level. T2D itself does not significantly increase level of LDL-C, but the small dense LDL particles are increased in T2D [21]. As a possible explanation for the lipid profile of the intact ketogenesis group, it might be assumed that the pathway in which acetyl-CoA is converted to acetoacetyl-CoA during ketogenesis to result in cholesterol synthesis is more dominant than the pathway in which acetyl-CoA is converted to malonyl-CoA to produce TG in the T2D patients with intact ketogenesis [5, 6].
Ketogenesis or ketone body has been of interest since studies have shown that ketogenesis is associated with better metabolic outcomes. To apply this notion to lifestyle modification, ketogenic, low-carbohydrate diets have been reported to have a significant weight loss effect as well as a glucose-lowering effect (22–24). The ketogenic diet has also demonstrated its beneficial effects on MAFLD [25]. In addition, some researchers have reported that subjects with intact ketogenesis, who can produce sufficient ketone bodies during fasting periods, have better metabolic features or clinical outcomes than those who cannot. A cross-sectional study using health check-up data showed that the presence of ketonuria after fasting was associated with metabolic superiority, such as lower body weight, waist circumstance, blood pressure, and blood glucose level [26]. In a longitudinal prospective study, healthy individuals with spontaneous fasting ketonuria had a lower risk of incident diabetes [27]. However, discrepant reports have been made on the association between ketogenesis and fatty liver disease. A longitudinal study using a cohort of 153,076 nondiabetic Korean subjects reported that fasting ketonuria was associated with reduced risk of incident hepatic steatosis [28]. Similarly, a cross-sectional study reported that fasting ketonuria imparted reduced risk for advanced liver fibrosis in MAFLD patients [29]. In contrast, Dutch researchers analyzed cohort data and reported that subjects with suspected fatty liver disease had higher blood levels of ketone bodies [11]. Several previous studies showed that ketone levels were increased in patients with prediabetes or diabetes [30, 31], which was closely related to fatty liver disease. The result of our study in drug-naïve T2D patients is in line with previous studies that reported lower risk of steatosis or fibrosis in nondiabetic subjects with intact ketogenesis [28, 29].
The pathophysiology or mechanism by which intact ketogenesis in patients with T2D might be protective against MAFLD has not been elucidated, although candidate mechanisms have been proposed. First, in the liver, ketogenesis is an efficient pathway for disposal of FAs from peripheral lipolysis. FAs delivered to the liver are converted to acetyl-CoA through β-oxidation, and ketogenesis is the non-oxidative pathway that converts the acetyl-CoA into energy and disposes of it to the peripheral tissues [5, 6]. When the ketogenic pathway is impaired, acetyl-CoA is oxidized to CO2 in the TCA cycle; researchers in the United States reported that the alternative TCA cycle was upregulated via fasting,while ketogenesis was impaired in patients with fatty liver disease [7]. Unlike the ketogenic pathway, the TCA cycle promotes gluconeogenesis/de novo lipogenesis and increases hepatic oxygen consumption. This would cause steatosis and oxidative stress on the liver. As a result, when the ketogenic pathway is impaired, the effective pathway for disposal of FAs in the liver is impaired, and the materials for de novo lipogenesis increase along with oxidative stress. Second, the ketogenic process induces hepatic peroxisome proliferation-activated receptor α (PPARα) and fibroblast growth factor 21 (FGF21) action that is important for hepatic lipid metabolism (32–34). The PPARα-FGF21 axis plays a critical role in metabolism across a broad spectrum of organs as well as lipid metabolism in the liver [33, 35]. Dysregulation of this axis is associated with pathogenesis of MAFLD [35, 36]. Therefore, induction of the PPARα-FGF21 axis in a ketogenic state could decrease steatosis. FGF21 signaling also activates hepatic autophagy [35], in which reduction is a significant process in the pathogenesis of MAFLD [37].
Despite the supportive results of our study, several limitations should be considered. First, our study was a cross-sectional analysis with a relatively small sample size. Therefore, an association between ketogenesis and MAFLD could not be concluded. In addition, characteristics of the intact ketogenesis group, such as higher blood glucose levels and the reduced insulin secretory function, may have acted as prerequisites for the ketogenic ability. From that point of view, ketogenesis may act as a compensatory action in T2D patients with decreased insulin secretion. Due to the limitations of a cross-sectional study, this study cannot suggest the causal relationships and mechanisms clearly. Further studies with larger populations and longitudinal design are needed to collect more evidence on this issue. Second, the study did not use imaging modalities such as ultrasonography, computed tomography, or transient elastography to evaluate the presence or extent of MAFLD. Further studies using these modalities are needed. Third, we did not have data on acetoacetate or acetone because there is no measurable method for them in our center. Therefore, we evaluated ketogenic potency with serum βHB, which is the most abundant form of ketone bodies. In the pathogenesis of MAFLD, not only the metabolic component but also the genetic component are important, and the genetic aspect is mainly expressed in the form of hepatic redox state [38]. The βHB/acetoacetate ratio is often used as a marker that reflects hepatic mitochondrial redox state but due to the lack of measurement of acetoacetate, we could not use this marker in our study. If this marker can be included in the analysis in the future, it is considered that the mechanistic explanation of our findings will be more clear. In addition, our study included only T2D patients. Therefore, care must be used when applying the results of this study to a nondiabetic population. However, since previous studies showing the association between ketogenesis and MAFLD were conducted on nondiabetic subjects, this limitation also shows the originality and clinical significance of this study. To the best of our knowledge, this study is the first to report the relationship of ketogenic capacity and hepatic steatogenic status in patients with T2D.
## Conclusions
In T2D patients, intact ketogenesis is associated with better MAFLD indices. Previous findings suggesting that ketogenesis is an efficient pathway for the liver to dispose of FAs, induces the PPARα-FGF21 axis, and reduces oxidative stress are considered to support the results. Large-scale prospective studies using imaging modalities are needed in the future.
## 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 Institutional review board of Severance Hospital. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
SL, JB, and B-WL designed the research. DJ contributed to acquisition of data. SL, JB, and B-WL analyzed the data. SL and JB wrote the article. ML, Y-hL, EK, B-SC, and B-WL contributed to revision of 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/fendo.2023.1124576/full#supplementary-material
## References
1. Diehl AM, Day C. **Cause, pathogenesis, and treatment of nonalcoholic steatohepatitis**. *N Engl J Med* (2017) **377**. DOI: 10.1056/NEJMra1503519
2. Lambert JE, Ramos–Roman MA, Browning JD, Parks EJ. **Increased**. *Gastroenterology* (2014) **146**. DOI: 10.1053/j.gastro.2013.11.049
3. Sunny Nishanth E, Parks Elizabeth J, Browning Jeffrey D, Burgess Shawn C. **Excessive hepatic mitochondrial tca cycle and gluconeogenesis in humans with nonalcoholic fatty liver disease**. *Cell Metab* (2011) **14**. DOI: 10.1016/j.cmet.2011.11.004
4. Eslam M, Sanyal AJ, George J. **Mafld: A consensus-driven proposed nomenclature for metabolic associated fatty liver disease**. *Gastroenterology* (2020) **158** 1999-2014.e1. DOI: 10.1053/j.gastro.2019.11.312
5. McGarry JD, Foster DW. **Regulation of ketogenesis and clinical aspects of the ketotic state**. *Metabolism: Clin Exp* (1972) **21**. DOI: 10.1016/0026-0495(72)90059-5
6. Balasse EO. **Kinetics of ketone body metabolism in fasting humans**. *Metabolism: Clin Exp* (1979) **28** 41-50. DOI: 10.1016/0026-0495(79)90166-5
7. Fletcher JA, Deja S, Satapati S, Fu X, Burgess SC, Browning JD. **Impaired ketogenesis and increased acetyl-coa oxidation promote hyperglycemia in human fatty liver**. *JCI Insight* (2019) **5**. DOI: 10.1172/jci.insight.127737
8. Inokuchi T, Orita M, Imamura K, Takao T, Isogai S. **Resistance to ketosis in moderately obese patients: Influence of fatty liver**. *Internal Med (Tokyo Japan)* (1992) **31**. DOI: 10.2169/internalmedicine.31.978
9. Mey JT, Erickson ML, Axelrod CL, King WT, Flask CA, McCullough AJ. **B-hydroxybutyrate is reduced in humans with obesity-related nafld and displays a dose-dependent effect on skeletal muscle mitochondrial respiration in vitro**. *Am J Physiol Endocrinol Metab* (2020) **319**. DOI: 10.1152/ajpendo.00058.2020
10. Bugianesi E, Gastaldelli A, Vanni E, Gambino R, Cassader M, Baldi S. **Insulin resistance in non-diabetic patients with non-alcoholic fatty liver disease: Sites and mechanisms**. *Diabetologia* (2005) **48**. DOI: 10.1007/s00125-005-1682-x
11. Post A, Garcia E, van den Berg EH, Flores-Guerrero JL, Gruppen EG, Groothof D. **Nonalcoholic fatty liver disease, circulating ketone bodies and all-cause mortality in a general population-based cohort**. *Eur J Clin Invest* (2021) **51**. DOI: 10.1111/eci.13627
12. Lee EY, Hwang S, Lee SH, Lee YH, Choi AR, Lee Y. **Postprandial c-peptide to glucose ratio as a predictor of B-cell function and its usefulness for staged management of type 2 diabetes**. *J Diabetes Investig* (2014) **5**. DOI: 10.1111/jdi.12187
13. Lee SH, Lee BW, Won HK, Moon JH, Kim KJ, Kang ES. **Postprandial triglyceride is associated with fasting triglyceride and homa-ir in Korean subjects with type 2 diabetes**. *Diabetes Metab J* (2011) **35**. DOI: 10.4093/dmj.2011.35.4.404
14. Kim MK, Ko SH, Kim BY, Kang ES, Noh J, Kim SK. **2019 Clinical practice guidelines for type 2 diabetes mellitus in Korea**. *Diabetes Metab J* (2019) **43** 398-406. DOI: 10.4093/dmj.2019.0137
15. **Easl-Easd-Easo clinical practice guidelines for the management of non-alcoholic fatty liver disease**. *J Hepatol* (2016) **64**. DOI: 10.1016/j.jhep.2015.11.004
16. Long MT, Pedley A, Colantonio LD, Massaro JM, Hoffmann U, Muntner P. **Development and validation of the framingham steatosis index to identify persons with hepatic steatosis**. *Clin Gastroenterol Hepatol* (2016) **14** 1172-80.e2. DOI: 10.1016/j.cgh.2016.03.034
17. Wang J, Xu C, Xun Y, Lu Z, Shi J, Yu C. **Zju index: A novel model for predicting nonalcoholic fatty liver disease in a Chinese population**. *Sci Rep* (2015) **5**. DOI: 10.1038/srep16494
18. Xia MF, Yki-Järvinen H, Bian H, Lin HD, Yan HM, Chang XX. **Influence of ethnicity on the accuracy of non-invasive scores predicting non-alcoholic fatty liver disease**. *PloS One* (2016) **11**. DOI: 10.1371/journal.pone.0160526
19. Friedewald WT, Levy RI, Fredrickson DS. **Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge**. *Clin Chem* (1972) **18** 499-502. DOI: 10.1093/clinchem/18.6.499
20. Najjar SM, Caprio S, Gastaldelli A. **Insulin clearance in health and disease**. *Annu Rev Physiol* (2022) **85**. DOI: 10.1146/annurev-physiol-031622-043133
21. Arca M, Pigna G, Favoccia C. **Mechanisms of diabetic dyslipidemia: Relevance for atherogenesis**. *Curr Vasc Pharmacol* (2012) **10**. DOI: 10.2174/157016112803520864
22. Hussain TA, Mathew TC, Dashti AA, Asfar S, Al-Zaid N, Dashti HM. **Effect of low-calorie versus low-carbohydrate ketogenic diet in type 2 diabetes**. *Nutr (Burbank Los Angeles County Calif)* (2012) **28**. DOI: 10.1016/j.nut.2012.01.016
23. Saslow LR, Daubenmier JJ, Moskowitz JT, Kim S, Murphy EJ, Phinney SD. **Twelve-month outcomes of a randomized trial of a moderate-carbohydrate versus very low-carbohydrate diet in overweight adults with type 2 diabetes mellitus or prediabetes**. *Nutr Diabetes* (2017) **7** 304. DOI: 10.1038/s41387-017-0006-9
24. Choi YJ, Jeon SM, Shin S. **Impact of a ketogenic diet on metabolic parameters in patients with obesity or overweight and with or without type 2 diabetes: A meta-analysis of randomized controlled trials**. *Nutrients* (2020) **12**. DOI: 10.3390/nu12072005
25. Luukkonen PK, Dufour S, Lyu K, Zhang XM, Hakkarainen A, Lehtimäki TE. **Effect of a ketogenic diet on hepatic steatosis and hepatic mitochondrial metabolism in nonalcoholic fatty liver disease**. *Proc Natl Acad Sci USA* (2020) **117**. DOI: 10.1073/pnas.1922344117
26. Joo NS, Lee DJ, Kim KM, Kim BT, Kim CW, Kim KN. **Ketonuria after fasting may be related to the metabolic superiority**. *J Korean Med Sci* (2010) **25**. DOI: 10.3346/jkms.2010.25.12.1771
27. Kim G, Lee SG, Lee BW, Kang ES, Cha BS, Ferrannini E. **Spontaneous ketonuria and risk of incident diabetes: A 12 year prospective study**. *Diabetologia* (2019) **62**. DOI: 10.1007/s00125-019-4829-x
28. Kim Y, Chang Y, Kwon MJ, Hong YS, Kim MK, Sohn W. **Fasting ketonuria and the risk of incident nonalcoholic fatty liver disease with and without liver fibrosis in nondiabetic adults**. *Am J Gastroenterol* (2021) **116**. DOI: 10.14309/ajg.0000000000001344
29. Lim K, Kang M, Park J. **Association between fasting ketonuria and advanced liver fibrosis in non-alcoholic fatty liver disease patients without prediabetes and diabetes mellitus**. *Nutrients* (2021) **13**. DOI: 10.3390/nu13103400
30. Mahendran Y, Vangipurapu J, Cederberg H, Stancáková A, Pihlajamäki J, Soininen P. **Association of ketone body levels with hyperglycemia and type 2 diabetes in 9,398 Finnish men**. *Diabetes* (2013) **62**. DOI: 10.2337/db12-1363
31. Saasa V, Beukes M, Lemmer Y, Mwakikunga B. **Blood ketone bodies and breath acetone analysis and their correlations in type 2 diabetes mellitus**. *Diagnostics (Basel Switzerland)* (2019) **9**. DOI: 10.3390/diagnostics9040224
32. Badman MK, Pissios P, Kennedy AR, Koukos G, Flier JS, Maratos-Flier E. **Hepatic fibroblast growth factor 21 is regulated by pparalpha and is a key mediator of hepatic lipid metabolism in ketotic states**. *Cell Metab* (2007) **5**. DOI: 10.1016/j.cmet.2007.05.002
33. Piccinin E, Moschetta A. **Hepatic-specific pparα-Fgf21 action in nafld**. *Gut* (2016) **65**. DOI: 10.1136/gutjnl-2016-311408
34. Zhang Y, Lei T, Huang JF, Wang SB, Zhou LL, Yang ZQ. **The link between fibroblast growth factor 21 and sterol regulatory element binding protein 1c during lipogenesis in hepatocytes**. *Mol Cell Endocrinol* (2011) **342**. DOI: 10.1016/j.mce.2011.05.003
35. Byun S, Seok S, Kim YC, Zhang Y, Yau P, Iwamori N. **Fasting-induced Fgf21 signaling activates hepatic autophagy and lipid degradation**. *Nat Commun* (2020) **11** 807. DOI: 10.1038/s41467-020-14384-z
36. Li H, Fang Q, Gao F, Fan J, Zhou J, Wang X. **Fibroblast growth factor 21 levels are increased in nonalcoholic fatty liver disease patients and are correlated with hepatic triglyceride**. *J Hepatol* (2010) **53**. DOI: 10.1016/j.jhep.2010.05.018
37. Czaja MJ. **Function of autophagy in nonalcoholic fatty liver disease**. *Digestive Dis Sci* (2016) **61**. DOI: 10.1007/s10620-015-4025-x
38. Luukkonen PK, Qadri S, Ahlholm N, Porthan K, Männistö V, Sammalkorpi H. **Distinct contributions of metabolic dysfunction and genetic risk factors in the pathogenesis of non-alcoholic fatty liver disease**. *J Hepatol* (2022) **76**. DOI: 10.1016/j.jhep.2021.10.013
|
---
title: Evaluating the influence of sleep quality and quantity on glycemic control
in adults with type 1 diabetes
authors:
- Marta Botella-Serrano
- Jose Manuel Velasco
- Almudena Sánchez-Sánchez
- Oscar Garnica
- J. Ignacio Hidalgo
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9989462
doi: 10.3389/fendo.2023.998881
license: CC BY 4.0
---
# Evaluating the influence of sleep quality and quantity on glycemic control in adults with type 1 diabetes
## Abstract
### Background
Sleep quality disturbances are frequent in adults with type 1 diabetes. However, the possible influence of sleep problems on glycemic variability has yet to be studied in depth. This study aims to assess the influence of sleep quality on glycemic control.
### Materials and methods
An observational study of 25 adults with type 1 diabetes, with simultaneous recording, for 14 days, of continuous glucose monitoring (Abbott FreeStyle Libre system) and a sleep study by wrist actigraphy (Fitbit Ionic device). The study analyzes, using artificial intelligence techniques, the relationship between the quality and structure of sleep with time in normo-, hypo-, and hyperglycemia ranges and with glycemic variability. The patients were also studied as a group, comparing patients with good and poor sleep quality.
### Results
A total of 243 days/nights were analyzed, of which $77\%$ ($$n = 189$$) were categorized as poor quality and $33\%$ ($$n = 54$$) as good quality. Linear regression methods were used to find a correlation ($r = 0.8$) between the variability of sleep efficiency and the variability of mean blood glucose. With clustering techniques, patients were grouped according to their sleep structure (characterizing this structure by the number of transitions between the different sleep phases). These clusters showed a relationship between time in range and sleep structure.
### Conclusions
This study suggests that poor sleep quality is associated with lower time in range and greater glycemic variability, so improving sleep quality in patients with type 1 diabetes could improve their glycemic control.
## Introduction
Poor sleep quality and insufficient amount of sleep are common in the general population and people with type 1 diabetes mellitus (T1DM) [1, 2]. A shorter duration of the deep sleep phase [3, 4], subjective quality of sleep, excessive daytime sleepiness [1, 2], and higher prevalence of obstructive sleep apnea [5] have been demonstrated in both adults and children with T1DM. The impact of these disturbances on glycemic control in patients with T1DM is an area of increasing interest. Previous studies suggest that sleep disturbances decrease insulin sensitivity, worse glycemic control, and increase glycemic variability [6]. Recently, the American Diabetes Association recommended the study of sleep patterns as part of clin-ical evaluation of a patient with T1DM [7]. The main objective of this study is to investigate by machine learning techniques the relationship among sleep structure, sleep quantity and quality, and glycemic control in patients with T1DM.
Griggs et al. [ 2020] [8] found in 38 patients that a higher glucose variability was associated within-person with more sleep disruptions or worse sleep. Our work extends theirs by grouping sleep patterns and analyzing the influence on glucose values during the day. Feupe et al. [ 2013] [9] studied the relation-ship between deep sleep duration and HbA1c level and concluded that they are inversely correlated.
Some previous studies used signal processing techniques to study the influence of physical exercise during the day on the glucose evolution during the following night [10]. To find coherence between the patient’s circadian rhythms, they used the cosinor technique (a technique used in circadian physiology) and wavelets. Another similar study using the wavelet coherence analysis is Griggs et al. [ 2022] [11]. Other studies (12–15) found significant relations between variability in sleep duration and poor glycemic control.
Our study complements these works by including the different sleep states during the night, grouping them into repetitive patterns, and studying their influence on different metrics of the following day. We apply clustering techniques and language processing techniques.
## Materials and methods
The study was approved by the ethics committee of the Príncipe de Asturias Hospital of Alcalá de Henares, Madrid, Spain. The research was compliant with the Declaration of Helsinki guidelines. Written consent was obtained from each participant prior to engagement.
## Inclusion/exclusion criteria
Eligible participants were adults between 18 and 65 years with T1DM with at least one year of duration, being on treatment with an insulin pump or multiple doses of subcutaneous insulin per day (MDI), having the availability of a mobile device capable of reading the sensors of the FreeStyle Libre system, and giving informed consent for inclusion in the study. Pre-screened subjects were excluded if they were diagnosed with a significant psychiatric disorder. Subjects in treatment with corticoids or patients that have required hospitalization or surgery on the last six months were excluded.
## Data gathering and preprocessing
The main objective of this study is to analyze the impact of sleep disturbances on short-term gly-cemic control, glycemic variability, and the frequency of hypoglycemia in a group of patients with T1DM. For this purpose, flash continuous glucose monitoring (performed by Abbott FreeStyle Libre devices) and a sleep study using wrist actigraphy (Fitbit Ionic device on the non-dominant wrist) were carried out simultaneously for 14 days in a group of patients. The CGM data includes interstitial blood glucose levels recorded during the entire time the patient wore the sensor, not only during sleep. Fitbit ionic devices incorporate a light sensor (photoplethysmography, PPG) and an accelerometer to identify sleep stages. From [16], “Fitbit uses proprietary sleep-staging machine learning algorithms applied to mo-tion, heart rate variability, and respiratory rate, with the last two calculated from heartbeat data sensed by PPG”. Twenty-five patients were included, although data from three patients had to be discarded, with a total of 243 nights/days recorded. The study analyzes interindividual and intraindividual differences in glycemic control concerning nights with worse or better quality/quantity of sleep.
Three visits were programmed to complete the collaboration of the participants. The study was explained to the participants in a first visit (Pre-screening Visit), and all patients signed an informed consent form. Participants were also committed to continuing with their usual treatment. In addition, sociodemographic variables, anthropometric data, and clinical data were collected from medical records. In a second visit, Visit 1, participants were given a wristband with wristwatch actigraphy (Fitbit Ionic device), and a FreeStyle Libre sensor (first generation, no alarms) was placed for continuous glucose monitoring for 14 days. The sensor was connected to the Abbot Libre View platform. Patients self-completed the Pittsburgh Sleep Quality Index (PSQI) questionnaire to assess habitual self-perceived sleep quality [1]. Visit 1 took place between 1 and 30 days after the pre-screening visit. Finally, during Visit 2, the glucose sensor and wrist actigraphy were removed. Visit 2 was programmed to be held 15 days after Visit 1. During this period, participants could contact the study’s technical staff to solve any technical concerns.
PSQI examines seven components: sleep quality, latency, habitual sleep efficiency, sleep duration, sleep disturbances, use of sleep medication, and daytime dysfunction. With 19 questions, participants rate the components on a scale of 0 to 3, ranging from 0 to 21, with higher scores indicating worse sleep quality (>5 reveal poor sleepers).
Recording of blood glucose data was performed through the Abbott Libre View application. Time-in-range is estimated directly by the FreeStyle Libre systems. In addition, we calculated it from the microdata generated by the meter using the Rosendaal method [17], which assumes a linear progression between two glucose values and calculates the specific value for each minute (linear interpolation). The same method was used to calculate time in hypo and hyperglycemia. The recording of the wrist activity was also performed automatically and digitized by the Fitbit mobile application. Microdata is not available directly from Fitbit, so we adapted the API (Application Program Interface) for recovering detailed information [1].
Glucose sensors were placed on the arm, and the Fitbit device was worn on the non-dominant wrist. The Fitbit and glucose data were synchronized at the closest multiple of 5 minutes. Once synchronized, sleep times spent in each stage were added to resynchronize with the 15 minutes used by FreeStyle Libre data. Days with gaps higher than one hour and a half in glucose were discarded. Fitbit data presented some outliers in sleeping times, mixing nap and night sleeping times for some days. Those days were eliminated manually. Heart rate, steps, and burned calories were collected and synchronized for future studies. Limitations of Fitbit Ionic devices are discussed in section 5 The glucose monitoring variables analyzed in this study are: time-in-range 70-180 mg/dl (TR) in percentage, mean blood glucose (mg/dl) (Mean_glucose), standard deviation (SD), coefficient of vari-ation (CV), percentage of time spent in level 1 hypoglycemia (55-70 mg/dl) and level 2 hypoglycemia (< 55 mg/dl) (T Hypo), time in hyperglycemia level 1 (180-250 mg/dl) and level 2 (> 250 mg/dl) (T Hyper), number of hypoglycemia/hyperglycemia episodes with at least 15 min of duration, Mean Amplitude of Glycemic Excursions (MAGE) and Mean Daily Glucose Differences (MDGD).
## Methodology
Figure 1 shows the workflow we have used in this study. First, we recorded the 24-hour time series of blood glucose levels (box B) and the sleep state sequences during the corresponding nights of the par-ticipants in the study (box A). After performing the clustering (subsection 2.4) according to the structure of sleep states (box C), we consequently grouped the daily time series of glucose levels corresponding to the nights of each cluster (box D). Then, the glucose time series were averaged by cluster (box E), and we obtained the dynamics of the glucose level that characterize each cluster. As a final phase, the behavior of the clusters is studied in two different ways: on the one hand, a language processing tech-nique is applied to find similarities and dissimilarities (subsection 2.5) in the glucose time series (box F) and, on the other hand, a statistical analysis (subsection 2.6) is performed to compare the glycemic characteristics between clusters (box G).
**Figure 1:** *Sequence of steps for the development of our study: recording of sleep states and glucose levels (A, B), clustering (C), characterizing cluster glucose behavior (D, E), finding specific glucose patterns (F) and comparison of glucose characteristics (G). 1
https://python-fitbit.readthedocs.io/en/latest/.*
## Analysis of the sequence of sleep states
Throughout the night, the person transits between different sleep states (wake, light, rem, deep), forming a time series of states or categories [18]. Figure 1 (box A) presents this time series as a sequence of colors displaying the sleep states. Each sleep state is represented by a color.
In this study, we want to determine whether there are patterns in the sleep time series during nights that correlate to patterns of blood glucose level evolution of the following day. To search for sleep patterns, we tried different time series clustering techniques [19]. Clustering is a group of machine learning techniques that identify clusters in the data. A cluster is a group or subset of elements of a population.
In this work, we applied clustering to group the nights of the participants based on the sequence of sleep states, i.e., each cluster includes those temporal sequences that are most similar to each other [18]. Subsequently, we analyzed the evolution of glucose during the following day for each cluster. Finally, we analyzed specific and expected behaviors in the diurnal evolution of glucose values for the different clusters. This last step is explained in more detail in subsection 2.5.
Figure 1 (box C) illustrates the result of applying clustering to the sleep data in four clusters, i.e., four sleep behavior patterns of the study participants.
## Similarity among glucose time series
Figure 1 (box E) shows each cluster’s average glucose time series. To identify specific behaviors of each cluster, we applied techniques commonly used in language processing [20, 21]. To do this, we trans-form a time series of numerical glucose values into a sequence of symbols. These symbols are obtained after the time series is normalized and reduced by obtaining the average of a number n of glucose values (in this work, $$n = 4$$) (Piecewise Aggregate Approximation, PAA) [22]. A symbol is assigned to each aver-aged point within a dictionary based on the statistical distribution (Symbolic Aggregate approXimation, SAX) [20, 21]. Finally, the symbols are grouped into words of a specific size (12 in this work).
Next, we identify behaviors specific to each cluster and those familiar to all clusters. Based on the number of occurrences of each “word” in all the time series of each cluster, we obtain the weight vectors associated with each word [Term Frequency -Inverse Document Frequency [23]]. Thanks to the weight vectors, we calculate the cosine similarity [24] and use this value to know if a word is specific to a cluster. In the average time series, we show those cluster-specific segments in cool colors (dark and light blue) and warm colors (red and orange) the similar segments across all clusters. The concept is that words with a very high frequency of occurrence in one cluster and a shallow frequency in the others appear in blue, whether the word appears a lot in all the clusters appears in orange or red. If the occurrence in the other clusters is medium, the color is green or yellow. Hence, we can identify the dynamics of glucose that characterizes a cluster. In Figure 1 (box F), we present a summary of the process for words of size
## Statistical methods
In order to find out the possible relationship between the glucose levels and the quality of sleep, several cluster analyses and correlational studies were performed using the R language and related libraries [18, 25]. These types of models, together with language processing techniques taken from the field of artificial intelligence, make it possible to determine possible patterns between the different variables selected and the nocturnal sleep patterns [26].
On the one hand, the K-means algorithm is used in the various cluster analyses among available data [27]: first, considering the variables associated with sleep alone, then the variables associated with glucose levels, and finally, the set of all variables. On the other hand, the correlation analysis takes as a reference Pearson’s correlation coefficient. These values have been obtained after processing these glucose records with the R Package gluvarpro [28].
To compare the glycemic characteristics of clusters, we used Welch’s F-test [29] using the package from Dag et al. [ 2018] [30]. We performed pairwise tests using Bonferroni’s correction [31, 32] for the p-values to calculate pairwise differences for each variable between the scores of each cluster (using the same package as before). In addition, we used Shannon’s entropy [33] for analyzing the results of the clustering. Shannon’s entropy provides an idea of how ordered sleep was. Higher values of entropy indicate higher levels of disorder.
## Participant characteristics
Twenty-five subjects participated in this study, of whom fourteen were female and eleven were male. Table 1 shows the characteristics of the participants identified by a random ID and including gender (M=Male; F=Female), age, BMI, HbA1c, diabetes treatment (MDI: Multiples doses of insulin; CSII: *Continuous subcutaneous* insulin infusion), and years of evolution of T1DM.
**Table 1**
| ID | Gender | Age | BMI | HbA1c | Treatment | Years T1DM |
| --- | --- | --- | --- | --- | --- | --- |
| HUPA001 | F | Q4 | Q2 | Q4 | ISCI | Q3 |
| HUPA002 | M | Q4 | Q2 | Q2 | ISCI | Q4 |
| HUPA003 | F | Q3 | Q1 | Q2 | ISCI | Q2 |
| HUPA004 | M | Q2 | Q4 | Q3 | ISCI | Q1 |
| HUPA005 | F | Q1 | Q2 | Q1 | ISCI | Q4 |
| HUPA006 | M | Q1 | Q3 | Q3 | ISCI | Q2 |
| HUPA007 | M | Q2 | Q4 | Q1 | ISCI | Q1 |
| HUPA008 | F | Q1 | Q4 | Q4 | ISCI | Q1 |
| HUPA009 | F | Q3 | Q2 | Q3 | ISCI | Q4 |
| HUPA010 | F | Q3 | Q1 | Q1 | ISCI | Q2 |
| HUPA011 | F | Q2 | Q2 | Q3 | ISCI | Q4 |
| HUPA014 | F | Q4 | Q3 | Q4 | MDI | Q2 |
| HUPA015 | F | Q3 | Q1 | Q1 | MDI | Q1 |
| HUPA016 | F | Q2 | Q3 | Q1 | ISCI | Q3 |
| HUPA017 | F | Q1 | Q1 | Q4 | MDI | Q3 |
| HUPA018 | F | Q2 | Q1 | Q2 | ISCI | Q4 |
| HUPA019 | M | Q1 | Q3 | Q2 | MDI | Q1 |
| HUPA020 | M | Q3 | Q3 | Q4 | MDI | Q2 |
| HUPA021 | F | Q4 | Q3 | Q3 | MDI | Q1 |
| HUPA022 | M | Q4 | Q2 | Q1 | ISCI | Q2 |
| HUPA023 | M | Q1 | Q1 | Q3 | MDI | Q1 |
| HUPA024 | M | Q3 | Q4 | Q4 | MDI | Q4 |
| HUPA025 | M | Q2 | Q4 | Q1 | ISCI | Q3 |
| HUPA026 | F | Q4 | Q4 | Q2 | MDI | Q3 |
| HUPA027 | M | Q1 | Q1 | Q2 | MDI | Q3 |
| Average | F:14/25 | 38.3 | 24.4 | 7.4 | ISCI:15/25 | 18.1 |
The mean age is 38.3 years, with an age range of 18-60.8 years, while the first and third quartiles are 26.4 and 47.9 years. The mean duration of diabetes is 18.1 years with a range of 0.8-39.5 years, and the first/third quartiles are $\frac{11.2}{24.2}$ years. HbA1c mean is $7.4\%$ (range 6-$9.7\%$) and 1st/3rd quartiles are 7/$7.8\%$. Body Mass Index (BMI) mean is 24.4 (range 18.5-32.2) and 1st/3rd quartiles are $\frac{22.3}{26.3.}$ Fifteen patients were on continuous insulin pump therapy, and ten were on multiple daily insulin doses. The CGM shows a mean blood glucose of 155 mg/dl, high glycemic variability (CV 36), and a time in hypo and hyperglycemia above target.
## Pittsburgh questionnaire results
Table 2 shows the results of the Pittsburgh questionnaire. The results of the PSQI give a poor overall sleep quality in $\frac{12}{23}$ patients, being these results concordant with the objective assessment of the actigraphy. Sleep disturbances contribute most to the high PSQI score (sudden nocturnal awakenings or other reasons like heat, cold, pain, nightmares, snoring, coughing, or the need to urinate). This result is also concordant with the actigraphy results, where the mean duration of objective nocturnal awakenings (WASO) is 52 minutes. Nine patients report significant daytime dysfunction (score >1) regarding drowsiness or poor mood for daily activities. One of the participants was discarded because he/she worked in shifts.
**Table 2**
| ID | Quality | Latency | Duration | Efficiency | Disturbances | Medication | Di dysfunction | Global | Subjective hours sleep |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| HUPA001 | 3.0 | 3 | 1.0 | 0 | 3 | 0.0 | 3.0 | 13 | 6.0 |
| HUPA002 | 0.0 | 1 | 1.0 | 0 | 1 | 1.0 | 2.0 | 6 | 7.0 |
| HUPA003 | 3.0 | 2 | 3.0 | 0 | 2 | 0.0 | 2.0 | 12 | 5.0 |
| HUPA004 | 0.0 | 1 | 0.0 | 0 | 1 | 3.0 | 2.0 | 7 | 8.0 |
| HUPA005 | 1.0 | 1 | 1.0 | 0 | 2 | 0.0 | 0.0 | 5 | 7.0 |
| HUPA006 | 0.0 | 1 | 1.0 | 1 | 1 | 0.0 | 2.0 | 6 | 6.3 |
| HUPA007 | 0.0 | 1 | 1.0 | 0 | 1 | 1.0 | 1.0 | 5 | 7.0 |
| HUPA008 | 0.0 | 1 | 2.0 | 0 | 1 | 1.0 | 1.0 | 6 | 5.0 |
| HUPA009 | 0.0 | 1 | 1.0 | 0 | 1 | 0.0 | 2.0 | 5 | 6.3 |
| HUPA010 | 2.0 | 3 | 3.0 | 3 | 2 | 0.0 | 3.0 | 16 | 4.0 |
| HUPA011 | 1.0 | 1 | 0.0 | 0 | 1 | 0.0 | 1.0 | 4 | 7.3 |
| HUPA014 | 2.0 | 2 | 1.0 | 3 | 2 | 0.0 | 0.0 | 10 | 7.0 |
| HUPA015 | 1.0 | 0 | 0.0 | 0 | 1 | 0.0 | 1.0 | 4 | 6.3 |
| HUPA016 | 1.0 | 1 | 0.0 | 0 | 1 | 0.0 | 0.0 | 3 | 7.3 |
| HUPA017 | 2.0 | 1 | 3.0 | 3 | 2 | 0.0 | 3.0 | 15 | 4.3 |
| HUPA018 | 2.0 | 1 | 1.0 | 1 | 1 | 0.0 | 1.0 | 8 | 5.0 |
| HUPA019 | 1.0 | 1 | 0.0 | 0 | 1 | 0.0 | 0.0 | 3 | 10.0 |
| HUPA020 | 0.0 | 1 | 0.0 | 0 | 1 | 0.0 | 1.0 | 4 | 6.3 |
| HUPA021 | 3.0 | 3 | 1.0 | 1 | 2 | 0.0 | 0.0 | 12 | 6.0 |
| HUPA022 | 0.0 | NV | 0.0 | NV | NV | 0.0 | 1.0 | NV | 6.0 |
| HUPA023 | 0.0 | 2 | 0.0 | 0 | 1 | 0.0 | 0.0 | 3 | 8.3 |
| HUPA024 | 0.0 | 3 | 2.0 | 1 | 1 | 0.0 | 0.0 | 7 | 5.0 |
| HUPA025 | 0.0 | 0 | 1.0 | 2 | 1 | 1.0 | 0.0 | 5 | 6.0 |
| HUPA026 | 1.0 | 1 | 0.0 | 0 | 2 | 3.0 | 1.0 | 10 | 8.0 |
| HUPA027 | 1.0 | 1 | 0.0 | 0 | 1 | 0.0 | 0.0 | 3 | 7.5 |
| Average | 0.91 | 3.8 | 1.38 | 119 | 0.57 | 1.33 | 0.32 | 129 | 7.09 |
## Sleep and glycemic control characteristics
Table 3 presents the values of the sleep monitoring variables for the participants of the study. Al-though most participants have a sleep efficiency higher than $90\%$, there are also 3 cases with a value close to $45\%$ and two others with low-efficiency values ($58\%$ and $68\%$). Sleep data from some parti-cipants, such as HUPA007 or HUPA008, were discarded due to inconsistency in the reported data.
**Table 3**
| ParticipantID | Efficiency% | Asleep(min) | Light(min) | Deep(min) | REM(min) | Awake(min) | Bed(min) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| HUPA001 | 94±2 | 366.08±114.66 | 218.08±49.39 | 53.38±30.51 | 94.62±43.92 | 42.38±13.93 | 419±128 |
| HUPA002 | 97±2 | 390.73±74.93 | 216.27±69.91 | 84.36±19.82 | 90.09±23.34 | 33.36±14.53 | 424±87 |
| HUPA003 | 93±3 | 335.75±98.39 | 228.50±83.39 | 46.75±16.27 | 60.50±20.89 | 45.25±22.80 | 350±98 |
| HUPA004 | 96±2 | 333.00±43.97 | 207.00±29.39 | 67.30±16.93 | 58.70±30.56 | 40.10±7.75 | 376±44 |
| HUPA005 | 58±14 | 350.88±54.08 | 211.38±25.62 | 69.12±24.19 | 70.38±23.13 | 50.38±10.68 | 401±58 |
| HUPA006 | 47±33 | 384.17±49.01 | 233.67±27.35 | 66.17±13.35 | 84.33±20.3 | 68.33±22.44 | 467±57 |
| HUPA007 | 94±2 | 325.46±38.89 | 204.38±38.82 | 54.62±15.08 | 66.46±23.48 | 37.92±7.76 | 100±74 |
| HUPA011 | 91±3 | 386.46±41.92 | 255.23±30.75 | 56.15±10.67 | 75.08±26.24 | 55.92±11.84 | 442±43 |
| HUPA014 | 93±2 | 481.92±98.54 | 293.67±60.12 | 78.58±23.78 | 109.67±44.29 | 86.75±35.94 | 569±127 |
| HUPA015 | 94±1 | 417.46±90.13 | 250.08±74.14 | 72.92±17.64 | 94.46±30.54 | 54.23±21.19 | 476±106 |
| HUPA016 | 97±2 | 416.69±40.41 | 255.85±36.96 | 62.54±14.31 | 98.31±20.66 | 53.08±16.17 | 474±48 |
| HUPA017 | 90±3 | 404.85±54.43 | 245.62±55.28 | 61.85±15.73 | 97.38±24.81 | 66.00±19.94 | 471±66 |
| HUPA018 | 47±5 | 408.82±37.31 | 205.91±33.48 | 76.27±20.58 | 126.64±35.54 | 51.64±9.64 | 403±24 |
| HUPA019 | 93±2 | 363.50±37.44 | 253.50±23.93 | 72.00±30.01 | 38.00±24.98 | 60.33±19.25 | 424±52 |
| HUPA020 | 46±9 | 319.90±80.52 | 202.50±67.91 | 57.00±18.37 | 60.40±19.73 | 48.70±19.97 | 369±94 |
| HUPA021 | 69±2 | 387.00±20.65 | 240.75±26.79 | 61.00±18.23 | 85.25±27.62 | 56.88±13.29 | 444±28 |
| HUPA022 | 94±3 | 303.86±81.22 | 204.07±56.36 | 38.71±17.51 | 61.07±30.62 | 41.86±13.4 | 346±92 |
| HUPA023 | 96±1 | 421.00±72.27 | 275.90±44.03 | 73.40±15.64 | 71.70±28.29 | 55.10±9.35 | 509±43 |
| HUPA024 | 92±4 | 298.00±68.25 | 213.00±55.61 | 34.33±9.00 | 50.67±16.63 | 56.17±16.34 | 170±138 |
| HUPA025 | 92±3 | 383.50±51.79 | 232.00±42.67 | 72.75±13.50 | 78.75±20.05 | 55.75±13.14 | 80±103 |
| HUPA026 | 61±6 | 437.06±36.85 | 241.50±33.85 | 82.81±17.17 | 112.75±26.56 | 55.88±12.10 | 325±219 |
| HUPA027 | 93±2 | 345.31±48.17 | 189.23±29.65 | 77.77±13.74 | 78.31±19.97 | 45.85±11.12 | 449±73 |
| Average | 83±5 | 375.52±60.63 | 230.82±45.25 | 64.54±17.82 | 80.16±26.46 | 52.81±15.57 | 391±119 |
Following the consensus on sleep quality assessment of the National Sleep Foundation [34], three vari-ables were used for evaluating the sleep quality: the number of awakenings during the night, WASO or Wake After Sleep Onset and the sleep efficiency (as the ration of total sleep time to time in bed [35].
Table 4 shows the sleep characteristics of the participant. Sleep quality was categorized as poor if at least two of three mentioned criteria were met, i.e., sleep efficiency < $85\%$ or Wake After Sleep Onset (WASO) > 40 min or a number of awakenings > 4. The sleep characteristics of the participants show large inter-individual differences, and only $48\%$ of the patients have a good overall sleep quality, although the mean sleep duration is not low (mean of 7.15 hours). Of the 243 nights analyzed, $77\%$ ($$n = 189$$) were of poor sleep quality and $33\%$ ($$n = 54$$) of good quality. It should be noted that eight patients had no night with good sleep quality. The factor that most determined poor sleep quality was the duration of nighttime awakenings, with a mean of 52.81 minutes.
**Table 4**
| Participant | Total | Sleep Quality | Sleep Quality.1 | Sleep Time | Sleep Efficiency | WASO | Awakenings |
| --- | --- | --- | --- | --- | --- | --- | --- |
| ID | Nights | Good | Poor | Avg (hours) | Avg (%) | Avg (min) | Avg per night |
| HUPA001P | 13 | 3 | 10 | 6.81±2.11 | 94.08±2.14 | 42.38±13.93 | 5.00±2.24 |
| HUPA002P | 11 | 8 | 3 | 7.07±1.45 | 97.45±1.63 | 33.36±14.53 | 2.73±1.90 |
| HUPA003P | 12 | 7 | 5 | 6.35±1.99 | 93.08±2.81 | 45.25±22.8 | 3.25±1.76 |
| HUPA004P | 10 | 6 | 4 | 6.22±0.75 | 95.80±1.81 | 40.10±7.75 | 3.20±1.40 |
| HUPA005P | 8 | 0 | 8 | 6.69±0.97 | 58.50±14.41 | 50.38±10.68 | 3.50±1.51 |
| HUPA006P | 6 | 0 | 6 | 7.54±0.98 | 47.67±33.68 | 68.33±22.44 | 6.33±3.33 |
| HUPA007P | 13 | 8 | 5 | 6.06±0.68 | 94.08±2.10 | 37.92±7.76 | 2.69±1.49 |
| HUPA011P | 13 | 1 | 12 | 7.37±0.72 | 90.92±3.43 | 55.92±11.84 | 4.23±1.83 |
| HUPA014P | 12 | 1 | 11 | 9.48±2.12 | 92.67±2.39 | 86.75±35.94 | 5.08±2.50 |
| HUPA015P | 13 | 3 | 10 | 7.94±1.76 | 93.69±1.55 | 54.23±21.19 | 3.69±1.49 |
| HUPA016P | 13 | 2 | 11 | 7.91±0.80 | 96.54±2.22 | 53.08±16.17 | 0.15±0.38 |
| ID | Nights | Good | Poor | Avg (hours) | Avg (%) | Avg (min) | Avg per night |
| HUPA017P | 13 | 1 | 12 | 7.85±1.11 | 90.31±3.35 | 66.00±19.94 | 4.85±1.68 |
| HUPA018P | 11 | 0 | 11 | 7.67±0.69 | 46.91±5.49 | 51.64±9.64 | 4.09±1.14 |
| HUPA019P | 6 | 0 | 6 | 7.06±0.87 | 93.00±1.67 | 60.33±19.25 | 6.17±5.00 |
| HUPA020P | 10 | 0 | 10 | 6.14±1.57 | 46.50±9.16 | 48.7±19.97 | 3.70±1.25 |
| HUPA021P | 8 | 0 | 8 | 7.40±0.47 | 69.38±2.20 | 56.88±13.29 | 4.62±1.85 |
| HUPA022P | 14 | 8 | 6 | 5.76±1.54 | 93.93±2.97 | 41.86±13.40 | 3.50±1.56 |
| HUPA023P | 10 | 0 | 10 | 7.93±1.33 | 95.50±1.18 | 55.10±9.35 | 4.80±1.40 |
| HUPA024P | 6 | 1 | 5 | 5.90±1.38 | 92.00±4.47 | 56.17±16.34 | 1.17±2.86 |
| HUPA025P | 12 | 1 | 11 | 7.32±1.01 | 92.08±3.00 | 55.75±13.14 | 4.17±2.52 |
| HUPA026P | 16 | 0 | 16 | 8.22±0.71 | 60.75±5.86 | 55.88±12.10 | 5.31±1.82 |
| HUPA027P | 13 | 4 | 9 | 6.52±0.81 | 93.38±2.33 | 45.85±11.12 | 3.23±1.54 |
| Overall | 243 | 54 | 189 | 7.15±1.17 | 83.1±4.99 | 52.81±15.57 | 3.88±1.93 |
Table 5 shows the percentages of time spent in sleep phases of Light, REM and Deep. Light phases percentage ranges from $45\%$ to $60.8\%$ with an average of $54.14\%$. The average time spent in the Deep phase is $15.08\%$, with a maximum of $20.19\%$ and a minimum of $11.72\%$. Finally, participants spent an average of $18.49\%$ of the time in bed in the REM phase, ranging from $8.62\%$ to $27.24\%$.
**Table 5**
| Participant ID | Light% | Deep% | REM% |
| --- | --- | --- | --- |
| HUPA0001P | 54.85 | 12.49 | 22.23 |
| HUPA0002P | 50.23 | 20.19 | 21.88 |
| HUPA0003P | 59.51 | 12.68 | 16.36 |
| HUPA0004P | 55.71 | 17.98 | 15.47 |
| HUPA0005P | 53.25 | 16.93 | 17.19 |
| HUPA0006P | 51.74 | 14.55 | 18.74 |
| HUPA0007P | 56.18 | 15.01 | 18.33 |
| HUPA0011P | 57.72 | 12.72 | 16.85 |
| HUPA0014P | 52.26 | 14.0 | 18.88 |
| HUPA0015P | 52.07 | 15.76 | 20.0 |
| HUPA0016P | 53.85 | 13.26 | 20.88 |
| HUPA0017P | 51.83 | 13.51 | 20.76 |
| HUPA0018P | 45.0 | 16.56 | 27.24 |
| HUPA0019P | 60.8 | 16.55 | 8.62 |
| HUPA0020P | 54.46 | 15.25 | 17.3 |
| HUPA0021P | 54.44 | 13.65 | 19.17 |
| HUPA0022P | 59.13 | 11.72 | 17.05 |
| HUPA0023P | 58.11 | 15.43 | 14.81 |
| HUPA0024P | 59.81 | 10.19 | 14.2 |
| HUPA0025P | 52.66 | 16.63 | 18.07 |
| HUPA0026P | 49.04 | 16.83 | 22.83 |
| HUPA0027P | 48.32 | 19.87 | 19.94 |
| Average | 54.14 | 15.08 | 18.49 |
Table 6 shows the overnight glycemic characteristics of the participants. Patients presented low values of time in range, with an average of 59.97 ± $14.74\%$, which is an indication of poor glycemic control. The high values of the average CV (36.45 ± 8.76) and standard deviation of the mean glucose (55.85 ± 14.) are also concordant with this appreciation.
**Table 6**
| Participant ID | Nights | Mean glucosemg/dl | SD | CV | TR | T Hyper% | T Hypo% |
| --- | --- | --- | --- | --- | --- | --- | --- |
| HUPA001P | 13 | 181.71±32.27 | 67.12±10.97 | 37.27±5.07 | 54.21±13.77 | 43.54±14.74 | 2.25±3.13 |
| HUPA002P | 11 | 113.68±30.59 | 50.18±15.85 | 44.72±11.52 | 60.94±16.59 | 15.00±17.30 | 24.07±18.09 |
| HUPA003P | 12 | 139.88±22.85 | 55.38±16.57 | 38.98±5.56 | 69.66±12.53 | 22.69±14.68 | 7.65±7.24 |
| HUPA004P | 10 | 178.75±44.75 | 74.15±23.86 | 44.07±14.86 | 44.37±19.36 | 43.70±22.39 | 11.93±13.73 |
| HUPA005P | 8 | 151.06±23.74 | 43.05±14.68 | 29.29±11.12 | 69.09±16.96 | 27.21±18.19 | 3.70±4.45 |
| HUPA006P | 6 | 212.05±97.73 | 62.41±36.48 | 35.97±20.23 | 46.76±28.69 | 48.68±29.74 | 4.56±5.55 |
| HUPA007P | 13 | 173.64±29.53 | 73.14±14.31 | 42.53±8.47 | 46.28±17.35 | 45.07±17.61 | 8.65±8.49 |
| HUPA011P | 13 | 159.30±19.37 | 54.00±8.38 | 34.09±4.97 | 65.47±10.23 | 31.96±11.56 | 2.57±3.52 |
| HUPA014P | 12 | 186.47±19.12 | 68.55±22.54 | 36.68±11.00 | 44.96±10.21 | 50.83±11.34 | 4.20±3.77 |
| HUPA015P | 13 | 165.90±20.72 | 65.03±10.67 | 39.33±5.08 | 57.68±12.01 | 38.60±13.38 | 3.72±3.22 |
| HUPA016P | 13 | 157.32±45.81 | 67.24±20.44 | 43.51±12.62 | 51.16±18.54 | 36.10±22.75 | 12.74±10.33 |
| HUPA017P | 13 | 198.39±27.38 | 62.06±16.51 | 31.67±8.14 | 37.39±16.94 | 60.25±18.30 | 2.35±3.36 |
| HUPA018P | 11 | 144.12±35.97 | 62.64±17.93 | 43.60±5.62 | 49.73±11.83 | 31.77±18.56 | 18.50±14.91 |
| HUPA019P | 6 | 159.97±17.59 | 54.82±5.05 | 34.40±2.64 | 59.14±11.14 | 36.85±11.92 | 4.01±2.24 |
| HUPA020P | 10 | 193.99±29.73 | 72.11±18.31 | 37.23±7.74 | 44.92±14.04 | 51.29±15.31 | 3.78±3.88 |
| HUPA021P | 8 | 141.27±14.84 | 44.83±5.46 | 31.92±4.32 | 74.07±9.15 | 22.78±10.68 | 3.14±5.71 |
| HUPA022P | 14 | 111.12±23.51 | 31.58±7.55 | 29.00±6.96 | 79.57±11.25 | 5.01±8.62 | 15.42±13.42 |
| HUPA023P | 10 | 132.89±21.29 | 38.98±7.22 | 29.53±4.81 | 78.92±12.69 | 18.05±14.39 | 3.03±4.47 |
| HUPA024P | 6 | 157.51±31.58 | 56.53±18.80 | 37.25±14.74 | 57.63±9.82 | 35.17±15.25 | 7.19±10.08 |
| HUPA025P | 12 | 113.51±14.56 | 36.46±9.21 | 32.09±7.44 | 79.58±11.24 | 7.42±6.85 | 12.99±8.94 |
| HUPA026P | 16 | 133.64±24.60 | 57.68±20.78 | 42.86±11.98 | 60.70±21.91 | 22.25±17.37 | 17.05±13.76 |
| HUPA027P | 13 | 121.24±21.16 | 30.73±8.24 | 25.82±7.92 | 87.04±18.09 | 8.36±18.09 | 4.60±4.91 |
| Overall | 243 | 155.79±29.49 | 55.85±14.99 | 36.45±8.76 | 59.97±14.74 | 31.94±15.86 | 8.10±7.60 |
## Association between sleep quality and blood glucose
In Figure 2, the upper triangular view of the correlation matrix for different variables recorded in this study is displayed as a correlogram. Red/blue colors for showing negative/positive correlation, and low/high intensity indicating the absolute value of the correlation. In addition to the expected correlations, we can point out several facts. Both poor and good sleep qualities have no significant correlation with any other variable. A positive correlation of 0.8 was found between the standard deviation of sleep efficiency and the standard deviation of mean blood glucose. In addition, the standard deviation of sleep efficiency has a positive correlation ($\frac{0.66}{0.62}$/$\frac{0.65}{0.61}$) with the standard deviation of glucose, the coefficient of variation, the time in range, and the time in hyperglycemia. There is a positive correlation (0.55) between the coefficient of variation and the mean time spent in hypoglycemia.
**Figure 2:** *Correlation matrix for all the variables recorded in this study.*
As mentioned, we used the clustering techniques to find patterns in sleep behavior. We experi-mented with a different number of clusters, having found $k = 4$ to be the best option, showing a clear relationship between sleep structure and glucose variables. With $k = 3$, we have a cluster with very broad glucose patterns, whereas, with $k = 5$, the relationship between sleep structure and the different glucose variables is confirmed with no additional information, remaining the main clusters the same. With $k = 4$, the number of nights grouped in each cluster was: 33, 78, 41 and 91.
In the left column of Figure 3, we can see the four sleep clusters, while the right column shows the average glucose dynamics in the days corresponding to each cluster. Remind that in the right column, the specificity of the glucose patterns is shown with intense blue color, while the patterns common to all clusters are shown in red. Because we take glucose level samples every 15 minutes, we have 96 samples per day. In the horizontal axis, we mark the main hourly correspondences. In order to correctly compare the sleep clusters, we show a total of 44 possible states per night (horizontal axis, on the left column).
**Figure 3:** *Analysis of specific patterns for each cluster. (A) Sleep States for cluster 1 (33 nights) (B) Average glucose behavior for cluster 1. (C) Sleep States for cluster 2 (78 nights) (D) Average glucose behavior for cluster 2. (E) Sleep States for cluster 3 (41 nights) (F) Average glucose behavior for cluster 3. (G) Sleep States for cluster 4 (91 nights) (H) Average glucose behavior for cluster 4.*
We calculated the Shannon’s entropy for the four clusters resulting in the following order (from highest entropy to lowest): 1,3, 2, and 4. Figure 4 shows the results of the main glucose variables for each cluster. On the one hand, the four clusters have slightly different sample sizes. On the other hand, we cannot assume that the clusters will have the same variance. The result of Welch’s F-test indicates that we can reject the null hypothesis that each cluster has the same mean value.
**Figure 4:** *Main glucose related results for four clusters in
Figure 3
. (A) Time in Range (B) Time in Hypoglycemia. (C) Time in Hyperglycemia (D) Coefficient of variation. (E) Mean level of Glucose (F) Standard Deviation of glucose levels.*
After the statistical tests with Bonferroni’s correction, we obtain several observations. Regarding time in Range, there is statistical significance between Cluster 1 and Cluster 4. We found significant differences for Cluster 1 and Cluster 2, versus Cluster 3 and Cluster 4 in time in hypoglycemia, and for Cluster 3 versus Cluster 2 and Cluster 4 in time in hyperglycemia. Cluster 1 and Cluster 4 are also significantly different in terms of CV. For the mean of glucose values, despite Welch’s F-test, the pairwise tests found no significant differences. One explanation could be that the applied Bonferroni’s adjustment was too severe [36, 37]. Another variable related to glucose variability, the standard deviation of glucose levels, presented significant differences for Cluster 1 and Cluster 3 versus Cluster 2 and Cluster 4. So, from these statistical results and the parallel observation of Figures (3 and 4 several conclusions arose. Clusters 1 and 2 (Figures 3A, C) have the longest sleep state sequences, the highest nocturnal glucose levels and a maximum peak around noon. Clusters 3 and 4 (Figures 3E, G), with the shortest sleep state sequences have the lowest nocturnal glucose levels and a maximum peak around sunset. In addition, these are the two clusters with the lowest time in hypoglycemia (Figure 4B). In the case of Cluster 3, this is mainly because this cluster has the longest time in hyperglycemia (Figure 4C), while Cluster 4 is the cluster with the longest time in range (Figure 4A). Cluster 1, with the highest Shannon’s entropy and the longest sequence of states [40], has as distinctive characteristic a very pronounced drop in glucose levels prior to the night and, at the same time, the most accentuated nocturnal rise (displayed in dark blue in Figure 3B). In addition, it is the cluster with the lowest time in range (Figure 4A), the highest time in hypoglycemia (Figure 4B), coefficient of variation (Figure 4D) and standard deviation (Figure 4F). Cluster 3, with the shortest sleep sequence (16 states), has the unique characteristic of a pronounced drop in glucose level during the night (Figure 3F). It is also the cluster with the highest time in hyperglycemia (Figure 4C) and, therefore, the highest mean level of glucose (Figure 4E). In Figure 4F, we can see that the lowest standard deviation corresponds to clusters 2 and 4. This could be related to the fact that these two clusters have the sleep state sequences with the lowest Shannon’s entropy. The shortest time in hyperglycemia and therefore lower mean glucose level (Figures 4C, E) corresponds to Cluster 2, which has a low level of Shannon’s entropy and a medium length of the sleep state sequence (Figure 3C).
## Discussion
This study shows that in adults with T1DM, subjective and objectively assessed sleep quality is poor, as occurs in $77\%$ nights analyzed with actigraphy and $52\%$ patients report Pittsburgh index > 5. In a previous epidemiological study using the Pittsburgh survey to measure subjective sleep quality [1] in a sample of 222 patients, authors found that $41\%$ have poor sleep quality (Pittsburgh index >5). According to our study’s observational data, sleep quality variability in adults with type 1 diabetes is associated with more significant variability in nocturnal blood glucose levels.
Similar findings have been reported in a group of adolescent patients using actigraphy and CGM, where sleep fragmentation, earlier awakening, and longer duration of WASO are associated with greater glycemic variability and longer time in hypoglycemia [8].
To our best knowledge, this is one of the first studies using machine learning techniques to analyze the relationship between sleep structure and times in normo-, hypo-, and hyperglycemia and to show that better sleep structure is associated with longer time in the glycemic range during that day. Our results confirm those of a previous study in 20 adult patients [38], showing that poor sleep quality is associated with greater glycemic variability. However, they found no association between sleep quality and time in range. This study only analyzes the relationship between sleep quality and glycemia with a linear mixed-effects model. The application of machine learning clustering reveals that nights with a higher disorder of the sleep structure presented lower time in range and a higher percentage of time in hypoglycemia. Increased time in the deep sleep phase was correlated with lower HBA1c and less time in nocturnal hypoglycemia in a previous study [9].
Other previous studies that do not use continuous glucose monitoring also suggest that sleep disturb-ances worsen glucose control. In particular, patients with short sleep duration (<6.5 hours) reported higher HbA1c than patients with longer sleep duration (>6.5 hours) [14]. Social jet lag (major changes in the duration and timing of sleep between weekdays and holidays) was associated with worse chronic metabolic control [39]. Some studies demonstrate the influence of sleep quality or duration on glycemic control in children. However, the findings among the different authors are not the same: it has been reported that a longer duration of the light sleep phase is associated with higher mean daily blood glucose, more episodes of hyperglycemia, and higher HbA1c [4], and that increased nocturnal awakenings [40] correlate with high glycemic variability.
Most of these studies have limitations in that they were conducted on a small number of patients, some only used subjective sleep assessments, and only two studies in adults simultaneously performed continuous glucose monitoring and polysomnography.
There are several possible mechanisms involved in poorer glycemic control [41]. Decreasing the duration of the REM phase would produce lower nocturnal glucose consumption, given that in the cerebral REM phase, glucose consumption is similar to awake. In contrast, in the cerebral non-REM phase, glucose consumption is much lower. In the general population and patients with diabetes, sleep deprivation, fragmentation, and decreased deep sleep are associated with decreased insulin sensitivity, possibly mediated by increased cortisol and Growth hormone (GH) levels. In patients with T1DM, higher nocturnal levels of growth hormone, adrenaline, ACTH, and cortisol than in the control population have been reported [3]. In an experimental study, the partial restriction of a single night of sleep (4 hours) de-creased peripheral insulin sensitivity measured by the hyperinsulinemic-euglycemic clamp in patients with T1DM [42].
Although it has been proposed that continuous glucose monitoring may alter sleep quality due to hyper-or hypoglycemia alarms, in this study, the CGM did not have alarms, so the likelihood of inter-ference of the CGM on sleep quality is very low. Finally, sleep disturbances could worsen glycemic control by an indirect mechanism related to patients’ behavior and cognitive functions. An association has been described in children and teenagers between a shorter duration of sleep and a decrease in the frequency of self-monitoring and insulin bolus administration [43].
## Limitations
The use of Fitbit devices is controversial. The previous generation of Fitbit devices was equipped only with body movement sensors and, therefore, were unsuitable for recording sleep stages. How-ever, new generations incorporated heart rate recording and a light sensor, so the Fitbit Ionic models greatly improved the ability to identify sleep stages. In fact, they are considered sleep-staging models [16]. However, the data is not available directly from the web. Instead, programming an API is necessary to obtain the data. Although the code was tested thoroughly, a deeper validation with other wrist devices of higher complexity would be beneficial. Recently evaluations of several commercial sleep technologies during sleeping concluded that Fitbit ionic measured with greater accuracy and limited bias Total sleep time (TST), total wake time (TWT), and sleep efficiency (SE). Regarding sleep, stages were reported poor for the time spent in REM sleep and with lower error in the other two stages [44]. We did not find other validation studies for Fitbit Ionic. The sample size of $$n = 22$$ should be considered when evaluat-ing the results and conclusions presented in this work. Although all of the participants are adults and sleep recommendations are not different throughout adulthood, it would be necessary to study better the differences on sleep patterns based on age.
This study did not look at possible differences between patients treated with MDI or CSII. Separating patients by treatment would further reduce the number of nights used to analyze each group as we clustered by night rather than by patient. Future work should include larger samples to investigate such possible relationships.
## Conclusion
To our best knowledge, our work is the first study that, using artificial intelligence and statistical techniques, has found a relationship between sleep structure and times in normo-, hypo-, and hypergly-cemia. Our main conclusion is that better sleep structure is associated with a longer time in the glycemic range. Future studies are needed to confirm these findings in a larger patient population and investigate the mechanisms involved in the decreased time in range and increased glycaemic variability caused by poor sleep quality. We believe that sleep disturbances should be a factor to be assessed in the clinical practice of patients with type 1 diabetes and that strategies should be designed to treat these disturbances. As future work, we are considering conducting a study to investigate further the relationship between sleep and glycemia by age group. Intuitively, variables affected by age, such as habits, responsibilities, social influences, etc., may produce significant differences in sleep patterns and diabetes outcomes.
## 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 Hospital Universitario Príncipe de Asturias. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
MB-S: Experimental design. Medical analysis. JV: AI techniques, programming, figures, writting. AS-S: statistical analysis OG: AI analysis, writing JIH: AI analysis, writing and 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. Denic-Roberts H, Costacou T, Orchard TJ. **Subjective sleep disturbances and glycemic control inadults with long-standing type 1 diabetes: The pittsburgh’s epidemiology of diabetes complicationsstudy**. *Diabetes Res Clin Pract* (2016) **119**. DOI: 10.1016/j.diabres.2016.06.013
2. van Dijk M, Donga E, van Dijk JG, Lammers GJ, van Kralingen KW, Dekkers OM. **Disturbedsubjective sleep characteristics in adult patients with long-standing type 1 diabetes mellitus**. *Diabet-ologia* (2011) **54**. DOI: 10.1007/s00125-011-2184-7
3. Jauch-Chara K, Schmid SM, Hallschmid M, Born J, Schultes B. **Altered neuroendocrine sleep archi-tecture in patients with type 1 diabetes**. *Diabetes Care* (2008) **31**. DOI: 10.2337/dc07-1986
4. Perfect MM, Patel PG, Scott RE, Wheeler MD, Patel C, Griffin K. **Sleep, glucose, and daytimefunctioning in youth with type 1 diabetes**. *Sleep* (2012) **35**. DOI: 10.5665/sleep.1590
5. Borel A, Benhamou P, Baguet J, Halimi S, Levy P, Mallion J. **High prevalence of obstruct-ive sleep apnoea syndrome in a type 1 diabetic adult population: A pilot study**. *Diabetes Med* (2010) **27**. DOI: 10.1111/j.1464-5491.2010.03096.x
6. Koren D, O’Sullivan KL, Mokhlesi B. **Metabolic and glycemic sequelae of sleep disturbances inchildren and adults**. *Curr Diabetes Rep* (2015) **15** 562. DOI: 10.1007/s11892-014-0562-5
7. Draznin B, Aroda VR, Bakris G, Benson G, Brown FM, Freeman R. **4. comprehensive MedicalEvaluation and assessment of comorbidities: Standards of medical care in diabetes-2022**. *DiabetesCare* (2022) **45**. DOI: 10.2337/dc22-S004
8. Griggs S, Redeker NS, Jeon S, Grey M. **Daily variations in sleep and glucose in adolescents with type1 diabetes**. *Pediatr Diabetes* (2020) **21**. DOI: 10.1111/pedi.13117
9. Feupe SF, Frias PF, Mednick SC, McDevitt EA, Heintzman ND. **Nocturnal continuous glucose andsleep stage data in adults with type 1 diabetes in real-world conditions**. *J Diabetes Sci Technol* (2013) **7**. DOI: 10.1177/193229681300700525
10. Farabi SS, Carley DW, Quinn L. **Glucose variations and activity are strongly coupled in sleep and wake in young adults with type 1 diabetes**. *Biol Res Nurs* (2017) **19**. DOI: 10.1177/1099800416685177
11. Griggs S, Barbato E, Hernandez E, Gupta D, Margevicius S, Grey M. **Glucose and unstructured physical activity coupling during sleep and wake in young adults with type 1 diabetes**. *Sci Rep* (2022) **12** 5790. DOI: 10.1038/s41598-022-09728-2
12. Patel NJ, Savin KL, Kahanda SN, Malow BA, Williams LA, Lochbihler G. **Sleep habits in adolescents with type 1 diabetes: Variability in sleep duration linked with glycemic control**. *Pediatr Diabetes* (2018). DOI: 10.1111/pedi.12689
13. Chontong S, Saetung S, Reutrakul S. **Higher sleep variability is associated with poorer glycaemic control in patients with type 1 diabetes**. *J Sleep Res* (2016) **25**. DOI: 10.1111/jsr.12393
14. Borel AL, Pépin JL, Nasse L, Baguet JP, Netter S, Benhamou PY. **Short sleep duration measured by wrist actimetry is associated with deteriorated glycemic control in type 1 diabetes**. *Diabetes Care* (2013) **36**. DOI: 10.2337/dc12-2038
15. Griggs S, Grey M, Strohl KP, Crawford SL, Margevicius S, Kashyap SR. **Variations in sleep characteristics and glucose regulation in young adults with type 1 diabetes**. *J Clin Endocrinol Metab* (2022) **107**. DOI: 10.1210/clinem/dgab771
16. Haghayegh S, Khoshnevis S, Smolensky MH, Diller KR, Castriotta RJ. **Accuracy of wristband fit-bit models in assessing sleep: Systematic review and meta-analysis**. *J Med Internet Res* (2019) **21**. DOI: 10.2196/16273
17. Rosendaal FR, Cannegieter SC, van der Meer FJ, Briët E. **A method to determine the optimal intensity of oral anticoagulant therapy**. *Thromb Haemost* (1993) **69**. DOI: 10.1055/s-0038-1651587
18. Gabadinho A, Ritschard G, Müller NS, Studer M. **Analyzing and visualizing state sequences in r with TraMineR**. *J Stat Software* (2011) **40** 1-37. DOI: 10.18637/jss.v040.i04
19. Patel P, Keogh E, Lin J, Lonardi S. **Mining motifs in massive time series databases**. (2002) 370
20. Zan CT, Yamana H. **An improved symbolic aggregate approximation distance measure based on its statistical features**. *In: Proceedings of the 18th international conference on information integ-ration and web-based applications and services. iiWAS ‘16* (2016) 72-80. DOI: 10.1145/3011141.3011146
21. Keogh EJ, Pazzani MJ. *Journal: Knowledge and Information Systems* (2001) **3**. DOI: 10.1007/PL00011669
22. Guo C, Li H, Pan D. **An improved piecewise aggregate approximation based on statistical features for time series mining**. *In: Proceedings of the 4th international conference on knowledge science, engineering and management. KSEM’10* (2010)
23. Beel J, Langer S, Gipp B. **TF-IDuF: A novel term-weighting sheme for user modeling based on users’ personal document collections**. *Proceedings of the iConference 2017* (2017)
24. Novotný V. *Implementation notes for the soft cosine measure* (2018)
25. 25
R Core Team. R: A language and environment for statistical computing. Vienna, Austria (2021). Available at: https://www.R-project.org/.. *R: A language and environment for statistical computing* (2021)
26. Jurafsky D, Martin JH. *Speech and language processing : an introduction to natural language pro-cessing, computational linguistics, and speech recognition* (2009)
27. Lloyd S. **Least squares quantization in PCM**. *IEEE Trans Inf Theory* (1982) **28**. DOI: 10.1109/TIT.1982.1056489
28. Contador S. *Gluvarpro: Glucose variability measures from continuous glucose monitoring data; 2020* (2020)
29. Delacre M, Lakens D, Leys C. **Why psychologists should by default use welch’s f-test instead of student’s t-test**. *Int Rev Soc Psychol* (2017) **30** 92. DOI: 10.5334/irsp.82
30. Dag O, Dolgun A, Konar NM. **Onewaytests: An r package for one-way tests in independent groups designs**. *R J* (2018) **10**. DOI: 10.32614/RJ-2018-022
31. Haynes W, Dubitzky W, Wolkenhauer O, KH C, Yokota H. *Bonferroni correc-tion* (2013) **p**. DOI: 10.1007/978-1-4419-9863-7_1213
32. Armstrong RA. **When to use the bonferroni correction**. *Ophthalmic Physiol Optics* (2014) **34**. DOI: 10.1111/opo.12131
33. Shannon CE. **A mathematical theory of communication**. *Bell System Tech J* (1948) **27**. DOI: 10.1002/j.1538-7305.1948.tb00917.x
34. Ohayon M, Wickwire EM, Hirshkowitz M, Albert SM, Avidan A, Daly FJ. **National sleep foundation’s sleep quality recommendations: first report**. *Sleep Health* (2017) **3** 6-19. DOI: 10.1016/j.sleh.2016.11.006
35. Reed DL, Sacco WP. **Measuring sleep efficiency: What should the denominator be**. *J Clin Sleep Med* (2016) **12**. DOI: 10.5664/jcsm.5498
36. Abdi H, Salkind NJ. **Bonferroni and sidak corrections for multiple comparisons**. *Encyclo-pedia of measurement and statistics* (2007)
37. Lee S, Lee DK. **What is the proper way to apply the multiple comparison test**. *Korean J Anesthesiol* (2020) **73** 572. DOI: 10.4097/kja.d.18.00242.e1
38. Brandt R, Park M, Wroblewski K, Quinn L, Tasali E, Cinar A. **Sleep quality and glycaemic variability in a real-life setting in adults with type 1 diabetes**. *Diabetologia* (2021) **64**. DOI: 10.1007/s00125-021-05500-9
39. Larcher S, Gauchez AS, Lablanche S, Pépin JL, Benhamou PY, Borel AL. **Impact of sleep behavior on glycemic control in type 1 diabetes: the role of social jetlag**. *Eur J Endocrinol* (2016) **175** 411. DOI: 10.1530/EJE-16-0188
40. Pillar G, Schuscheim G, Weiss R, Malhotra A, McCowen KC, Shlitner A. **Interactions between hypoglycemia and sleep architecture in children with type 1 diabetes mellitus**. *J Pediatr* (2003) **142**. DOI: 10.1067/mpd.2003.66
41. Tsuneki H, Sasaoka T, Sakurai T. **Sleep control, GPCRs, and glucose metabolism**. *Trends Endo-crinol Metab* (2016) **27**. DOI: 10.1016/j.tem.2016.06.011
42. Donga E, van Dijk M, van Dijk JG, Biermasz NR, Lammers GJ, van Kralingen K. **Partial sleep restriction decreases insulin sensitivity in type 1 diabetes**. *Diabetes Care* (2010) **33**. DOI: 10.2337/dc09-2317
43. McDonough RJ, Clements MA, DeLurgio SA, Patton SR. **Sleep duration and its impact on adherence in adolescents with type 1 diabetes mellitus**. *Pediatr Diabetes* (2017) **18**. DOI: 10.1111/pedi.12381
44. Stone JD, Rentz LE, Forsey J, Ramadan J, Markwald RR, Finomore VS. **Evaluations of com-mercial sleep technologies for objective monitoring during routine sleeping conditions**. *Nat Sci Sleep* (2020) **12** 821. DOI: 10.2147/NSS.S270705
|
---
title: 'Thyroid hormones and carnitine in the second trimester negatively affect neonate
birth weight: A prospective cohort study'
authors:
- Mengmeng Yang
- Man Sun
- Chenyu Jiang
- Qianqian Wu
- Ying Jiang
- Jian Xu
- Qiong Luo
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9989483
doi: 10.3389/fendo.2023.1080969
license: CC BY 4.0
---
# Thyroid hormones and carnitine in the second trimester negatively affect neonate birth weight: A prospective cohort study
## Abstract
### Background
Maternal thyroid hormones and carnitine are reported to affect neonate birth weight during the second trimester, which is one of the most important markers for fetal growth and perinatal mortality and morbidity. Nevertheless, the effect of thyroid hormone and carnitine in the second trimester on birth weight has yet to be understood.
### Method
This was a prospective cohort study with 844 subjects enrolled during the first trimester. Thyroid hormones, free carnitine (C0), neonate birth weight, as well as other related clinical and metabolic data were collected and assessed.
### Results
Pre-pregnancy weight and body mass index (BMI) as well as neonate birth weight were significantly different among different free thyroxine (FT4) level groups. Maternal weight gain and neonate birth weight varied significantly when grouped by different thyroid-stimulating hormone (TSH) levels. There was a significantly positive correlation between C0 and TSH ($r = 0.31$), free triiodothyronine (FT3) ($r = 0.37$), and FT4 ($r = 0.59$) (all $P \leq 0.001$). In addition, a significantly negative influence was found between birth weight and TSH (r = −0.48, $$P \leq 0.028$$), so as C0 (r = −0.55, $P \leq 0.001$) and FT4 (r = −0.64, $P \leq 0.001$). Further assessment detected a stronger combined effect of C0 and FT4 ($P \leq 0.001$) and of C0 and FT3 ($$P \leq 0.022$$) on birth weight.
### Conclusion
Maternal C0 and thyroid hormones are of great importance in neonate birth weight, and routine examination of C0 and thyroid hormones during the second trimester has a positive effect on the intervention of birth weight.
## Introduction
Adequate thyroid hormones (THs) are crucial for the fetal growth and metabolism and play an important role in neurodevelopment [1, 2]. The fetal thyroid grand starts to develop at 12th week of gestation and is functionally mature around the 18th to 20th week, whereas, for the first half of pregnancy, the fetus relies entirely on the supply of maternal TH through the transplacental passage (2–4). Abnormal maternal thyroid function during pregnancy is associated with adverse obstetrical and offspring outcomes, such as spontaneous abortion, anemia, preeclampsia, placental abruption, congenital anomalies, preterm birth and/or low birth weight, fetal distress in labor, stillbirth and/or perinatal death, and postpartum hemorrhage [3].
Carnitine exists as free carnitine (C0) and acylcarnitine fractions in blood and plays an important role in fatty acid oxidation during the gestational metabolism [4]. Carnitine mainly comes from food, especially red meat, fish, and dairy products [5]. Carnitine is critical for the transfer of activated long-chain fatty acids from the cytoplasm to the mitochondria for β-oxidation, resulting in the esterification of carnitine to form acylcarnitine derivatives [6]. Some evidence suggests that carnitine deficiency is manifested in gestational diabetes mellitus (GDM), leading to the development of macrosomia and small for gestational age (SGA). Clinical studies showed that applying C0 (1 g/day) for a few weeks could relieve hyperthyroidism symptoms in patients. Carnitine was hypothesized to act in the periphery by antagonizing TH action. However, the correlation between C0 and THs in mid-pregnancy remains unknown.
Birth weight is one of the most important markers for fetal growth and development in utero, which reflects fetal adaptations to the intrauterine environment. SGA newborns have an increased risk of prenatal mortality [7], whereas large for gestational age (LGA) newborns have a higher risk of obesity and diabetes mellitus in later life [2]. Various studies have shown that a higher maternal free thyroxine (FT4) level is associated with a lower birth weight [3, 8, 9]. However, studies undertaken so far provide conflicting evidence concerning the impact of C0 on TH and birth weight.
In the present study, we investigated the associations between maternal serum thyroid parameters and carnitine-related metabolites during the second trimester of pregnancy. Furthermore, we examined whether the birth weight was modified by maternal serum TH and carnitine.
## Study subjects
This was a prospective cohort study. Pregnant women who received regular perinatal healthcare in the Outpatient Department of the Women’s Hospital School of Medicine Zhejiang University and delivered in the hospital between June 2017 and April 2019 were recruited. A total of 844 pregnant women with complete demographic and obstetric data were included for analysis. This study was approved by the Ethics Committee of Women’s Hospital School of Medicine Zhejiang University.
All pregnant women were enrolled during the first trimester, all of whom had measured thyroid-stimulating hormone (TSH), FT4, free triiodothyronine (FT3), and total thyroxine (TT4) concentration in the second trimester, and neonatal birth weight data were available. Women with multiple pregnancies, accompanied with pregnancy complications such as abortion, GDM, and hypertensive disorders, using medication known to interfere thyroid hormones, or had a history of thyroid diseases were excluded.
Maternal clinical characteristics—including age, height, pre-pregnancy and prenatal weight and body mass index (BMI), gravity, parity, mode of delivery, educational level, and neonatal birth weight—were obtained from hospital information system and child care system.
The blood samples were collected after overnight fasting, and samples were centrifuged within 6 h. The concentrations of TSH, FT4, FT3, and TT4 were determined according to the measurement instructions.
## Metabolic profiling detection by LC-MS/MS
We aim to investigate the 31-carnitine-related plasma metabolite level of pregnant women at the second trimester. We also obtain their neonate blood plasma sample. The blood samples were stored at −20°. Furthermore, the samples were prepared by tandem mass spectrometry (4000 QTrapTM; AB Sciex, Darmstadt, Germany) to test the concentration. The method used in the present study was essentially a modification of the procedure described elsewhere. Amino acid (AA) and acylarnitine (AC) were quantified using appropriate isotope-labeled standards. Liquid Chromatogram (LC) separation was performed on an Acquity UPLC HSS T3 column (2.1 × 100 mm, 100A°, 1.8-µm particle size; Waters Corporation, MA) using water with $0.1\%$ formic acid ($0.1\%$ methanol and 5 mM ammonium acetate) detected with a Xevo-G2-QTOF MS (Waters Corporation) operating in a positive mode. Raw data were processed using TargetLynx as described previously. Accuracy of quantification was below $6\%$ for all quantified metabolites except glutamic acid ($13.9\%$). Quantitative data were obtained using MetIDQTM software.
## Statistical analysis
The data were expressed as mean ± standard deviation (SD). The baseline characteristics of the subjects were described, and the p-values are indicated. Binary variables were presented as frequency and percentage and were compared using the Chi-squared test. A nonparametric test or a t-test was used to compare the medians of continuous variables. The heatmap was available from the package “ggplot” as an enhanced version or its basic function stats in R. We used the liner regression model, as well as multiple regressions to investigate the association of C0, FT3, FT4, TSH, and TT4 with birth weight. We assessed the combined effects of C0 and FT4 on birth weight by adding a product interaction term of the C0 × FT4 to the model. The same analysis was done on the effect of other hormone on birth weight. A heatmap was constructed to display the differences in birth weight. All statistical analyses were performed using R statistical software version 3.4.1 (package rms, ggplot, visreg, and mass) or Statistical Package of Social Sciences version 20.0 for Windows (IBM Corp., Armonk, NY). In all analyses, $P \leq 0.05$ was considered statistically significant.
## Clinical characteristics of subjects grouped by thyroid hormone
The maternal characteristics grouped by FT4 quartile are shown in Table 1. The different ranges of FT4 are as follows: Q1: 8.33–9.72 pmol/L; Q2: 9.73–10.65 pmol/L; Q3: 10.66–11.57 pmol/L; and Q4: 11.58–15.06 pmol/L. We found that pre-pregnancy weight and BMI as well as weight gain were significantly different among the four groups with a higher level in the lower FT4–level group. We also found that compared with the higher FT4–level group, birth weight was significantly heavier in the lower groups. There were significantly statistically differences in height and gestational week at delivery among different groups, but no difference in maternal age, nulliparous rate, and ART rate among these four groups.
**Table 1**
| FT4 groups (pmol/L) | Q1 (n = 211) | Q2 (n = 211) | Q3 (n = 211) | Q4 (n = 211) | P |
| --- | --- | --- | --- | --- | --- |
| | 8.33–9.72 | 9.73–10.65 | 10.66–11.57 | 11.58–15.06 | |
| Maternal age (years) | 31.48 ± 2.42 | 31.28 ± 2.62 | 31.65 ± 3.05 | 31.07 ± 2.79 | 0.149 |
| Height (cm) | 162.87 ± 5.49 | 159.65 ± 5.14 | 162.63 ± 2.27 | 160.44 ± 1.98 | <0.001 |
| Pre-pregnancy weight (kg) | 65.58 ± 4.83 | 64.52 ± 3.91 | 63.48 ± 3.65 | 64.18 ± 2.97 | <0.001 |
| Pre-pregnancy BMI (kg/m2) | 22.65 ± 4.01 | 21.86 ± 3.74 | 21.24 ± 3.19 | 21.74 ± 3.26 | <0.001 |
| Weight gain (kg) | 22.38 ± 4.21 | 21.65 ± 4.01 | 20.88 ± 2.45 | 19.78 ± 2.67 | <0.001 |
| Gestational week at delivery (weeks) | 38.3 ± 0.67 | 38.4 ± 0.78 | 39.4 ± 0.72 | 38.7 ± 0.87 | <0.001 |
| Nulliparous (%) | 166 (78.7) | 164 (77.7) | 167 (79.1) | 165 (78.2) | 0.987 |
| ART (%) | 13 (6.2) | 12 (5.7) | 9 (4.3) | 12 (5.7) | 0.843 |
| Birth weight (g) | 3633.23 ± 99.78 | 3527.79 ± 86.63 | 3404.56 ± 100.45 | 3217.93 ± 77.94 | <0.001 |
Table 2 presents clinical and biochemical characteristics of participants grouped by TSH. In the group with TSH < 2.5 mIU/L, pre-pregnancy BMI, weight gain, and birth weight were significantly higher; whereas, in the higher TSH–level group, pre-pregnancy weight and gestational week at delivery were higher. There was no difference in maternal age, nulliparous rate, and ART rate between the two groups.
**Table 2**
| TSH groups (mIU/L) | TSH<2.5 (n = 762) | TSH ≥ 2.5 (n = 82) | P |
| --- | --- | --- | --- |
| Maternal age (years) | 31.18 ± 2.42 | 31.65 ± 2.05 | 0.091 |
| Height (m) | 163.27 ± 5.49 | 162.63 ± 3.72 | 0.3032 |
| Pre-pregnancy weight (kg) | 63.58 ± 4.83 | 65.48 ± 3.65 | 0.001 |
| Pre-pregnancy BMI (kg/m2) | 22.65 ± 4.01 | 21.24 ± 3.19 | 0.002 |
| Weight gain (kg) | 21.38 ± 4.21 | 18.23 ± 2.45 | <0.001 |
| Gestational week at delivery (weeks) | 38.3 ± 0.67 | 39.4 ± 0.72 | <0.001 |
| Nulliparous (%) | 674 (88.5) | 73 (89.02) | 0.877 |
| ART (%) | 73 (9.6) | 4 (4.9) | 0.160 |
| Birth weight (g) | 3543.23 ± 99.78 | 3304.56 ± 100.45 | <0.001 |
## Biochemical characteristics of subjects grouped by thyroid hormone
Among 31 carnitine-related metabolites, we selected the metabolites with statistically significant differences and listed them in Table 3. Our results showed that carnitine-related AA [alanine (ALA), tyrosine (TYR), and valine (VAL)], short-chain AC (C2, C3, C3DC+C4OH, and C5:1), medium-chain AC (C6DC and C12), and long-chain AC (C14, C16:1) were significantly different among the four groups. As the FT4 level increased, ALA, TYR, C14, and C16:1 decreased, whereas VAL, C0, C2, C3, C3DC+C4OH, C5, C6DC, and C12 increased. There were also statistical differences in LEU+ILE+PRO-OH, SA, C4, C8, C14OH, C16, C18, and C18OH among these four groups.
**Table 3**
| AA and AC profiles (μmol/L) | Q1 | Q2 | Q3 | Q4 | P |
| --- | --- | --- | --- | --- | --- |
| ALA | 399.64 ± 21.46 | 368.87 ± 35.12 | 343.56 ± 24.14 | 322.77 ± 28.62 | <0.001 |
| ARG | 9.16 ± 2.29 | 8.74 ± 2.94 | 8.76 ± 1.82 | 8.81 ± 2.38 | 0.231 |
| GLY | 225.64 ± 25.11 | 226.66 ± 25.32 | 224.28 ± 18.34 | 227.45 ± 28.40 | 0.580 |
| LEU+ILE+PRO-OH | 135.38 ± 24.42 | 134.16 ± 31.72 | 130.85 ± 23.85 | 139.89 ± 23.96 | 0.005 |
| MET | 10.47 ± 3.32 | 10.06 ± 3.15 | 10.81 ± 2.87 | 10.42 ± 2.85 | 0.095 |
| ORN | 52.63 ± 12.76 | 52.86 ± 12.16 | 52.33 ± 11.44 | 53.37 ± 12.99 | 0.850 |
| PHE | 48.76 ± 8.49 | 48.65 ± 9.84 | 49.03 ± 8.43 | 49.93 ± 7.88 | 0.420 |
| PRO | 85.74 ± 16.32 | 86.21 ± 17.71 | 83.31 ± 14.06 | 86.34 ± 14.82 | 0.166 |
| SA | 0.77 ± 0.14 | 0.79 ± 0.13 | 0.78 ± 0.11 | 0.81 ± 0.12 | 0.009 |
| TYR | 41.73 ± 9.16 | 39.73 ± 7.86 | 37.63 ± 4.65 | 35.63 ± 4.16 | <0.001 |
| VAL | 118.72 ± 17.35 | 123.72 ± 18.15 | 125.15 ± 19.27 | 127.20 ± 29.27 | 0.001 |
| C0 | 16.35 ± 3.39 | 18.00 ± 2.95 | 20.51 ± 2.68 | 22.35 ± 3.62 | <0.001 |
| C2 | 2.67 ± 0.82 | 2.86 ± 0.76 | 3.04 ± 0.78 | 3.14 ± 0.84 | <0.001 |
| C3 | 0.30 ± 0.18 | 0.42 ± 0.14 | 0.50 ± 0.21 | 0.60 ± 0.19 | <0.001 |
| C3DC+C4OH | 0.23 ± 0.12 | 0.28 ± 0.11 | 0.33 ± 0.09 | 0.45 ± 0.14 | <0.001 |
| C4 | 0.09 ± 0.01 | 0.08 ± 0.02 | 0.09 ± 0.02 | 0.10 ± 0.02 | <0.001 |
| C5 | 0.12 ± 0.01 | 0.23 ± 0.09 | 0.27 ± 0.10 | 0.34 ± 0.17 | <0.001 |
| C5DC+C6OH | 0.38 ± 0.12 | 0.36 ± 0.13 | 0.39 ± 0.11 | 0.38 ± 0.13 | 0.085 |
| C6 | 0.018 ± 0.011 | 0.017 ± 0.012 | 0.017 ± 0.011 | 0.018 ± 0.013 | 0.678 |
| C6DC | 0.018 ± 0.012 | 0.21 ± 0.17 | 0.28 ± 0.11 | 0.36 ± 0.13 | <0.001 |
| C8 | 0.50 ± 0.13 | 0.49 ± 0.18 | 0.56 ± 0.19 | 0.54 ± 0.16 | <0.001 |
| C8:1 | 0.063 ± 0.022 | 0.060 ± 0.025 | 0.062 ± 0.028 | 0.062 ± 0.024 | 0.654 |
| C10 | 0.056 ± 0.019 | 0.056 ± 0.022 | 0.058 ± 0.021 | 0.059 ± 0.020 | 0.337 |
| C12 | 0.12 ± 0.182 | 0.14 ± 0.038 | 0.16 ± 0.034 | 0.18 ± 0.056 | <0.001 |
| C14 | 0.28 ± 0.045 | 0.26 ± 0.052 | 0.24 ± 0.048 | 0.23 ± 0.092 | <0.001 |
| C14OH | 0.014 ± 0.0055 | 0.012 ± 0.0052 | 0.013 ± 0.0058 | 0.013 ± 0.0064 | 0.005 |
| C16 | 0.023 ± 0.012 | 0.026 ± 0.013 | 0.026 ± 0.011 | 0.024 ± 0.015 | 0.035 |
| C16:1 | 0.22 ± 0.056 | 0.20 ± 0.034 | 0.18 ± 0.067 | 0.13 ± 0.033 | <0.001 |
| C18 | 0.57 ± 0.20 | 0.53 ± 0.22 | 0.59 ± 0.24 | 0.56 ± 0.21 | 0.040 |
| C18:1 | 0.55 ± 0.17 | 0.54 ± 0.18 | 0.53 ± 0.12 | 0.56 ± 0.22 | 0.335 |
| C18OH | 0.015 ± 0.0047 | 0.016 ± 0.0043 | 0.012 ± 0.0069 | 0.015 ± 0.0026 | <0.001 |
Carnitine-related metabolites grouped according to the TSH level are shown in Table 4. Compared with the lower TSH–level group, ALA and glycine (GLY) significantly decreased in the higher TSH–level group. LEU+ILE+PRO-OH, VAL, C0, C2, C3, C3DC+C4OH, C5, C6DC, C8:1, C10, C12, and C14 were significantly higher when TSH increased. There was a statistical difference in C4 between groups. On the basis of previous studies, C0 has a vital role in metabolism.
**Table 4**
| AA and AC profiles (μmol/L) | TSH<2.5 | TSH ≥ 2.5 | P |
| --- | --- | --- | --- |
| ALA | 378.64 ± 35.52 | 367.89 ± 41.67 | 0.011 |
| ARG | 8.93 ± 2.52 | 8.87 ± 3.01 | 0.841 |
| GLY | 0.81 ± 0.10 | 0.64 ± 0.12 | <0.001 |
| LEU+ILE+PRO-OH | 35.63 ± 4.16 | 41.73 ± 9.16 | <0.001 |
| MET | 10.56 ± 3.44 | 10.72 ± 3.56 | 0.690 |
| ORN | 53.65 ± 11.64 | 54.06 ± 12.08 | 0.763 |
| PHE | 48.78 ± 6.72 | 49.23 ± 7.78 | 0.571 |
| PRO | 86.34 ± 12.29 | 85.89 ± 15.89 | 0.760 |
| SA | 0.76 ± 0.15 | 0.77 ± 0.18 | 0.574 |
| TYR | 38.87 ± 6.87 | 39.12 ± 7.08 | 0.755 |
| VAL | 118.72 ± 17.35 | 127.20 ± 29.27 | <0.001 |
| C0 | 16.35 ± 3.62 | 20.00 ± 3.39 | <0.001 |
| C2 | 2.78 ± 0.62 | 3.41 ± 0.78 | <0.001 |
| C3 | 0.32 ± 0.18 | 0.62 ± 0.21 | <0.001 |
| C3DC+C4OH | 0.03 ± 0.011 | 0.05 ± 0.013 | <0.001 |
| C4 | 0.09 ± 0.02 | 0.08 ± 0.02 | <0.001 |
| C5 | 0.02 ± 0.015 | 0.04 ± 0.014 | <0.001 |
| C5DC+C6OH | 0.39 ± 0.13 | 0.37 ± 0.14 | 0.189 |
| C6 | 0.017 ± 0.011 | 0.018 ± 0.013 | 0.443 |
| C6DC | 0.32 ± 0.12 | 0.44 ± 0.124 | <0.001 |
| C8 | 0.52 ± 0.19 | 0.50 ± 0.16 | 0.359 |
| C8:1 | 0.38 ± 0.027 | 0.45 ± 0.056 | <0.001 |
| C10 | 0.37 ± 0.062 | 0.46 ± 0.041 | <0.001 |
| C12 | 0.31 ± 0.031 | 0.42 ± 0.023 | <0.001 |
| C14 | 0.28 ± 0.025 | 0.33 ± 0.032 | <0.001 |
| C14OH | 0.014 ± 0.0054 | 0.015 ± 0.0048 | 0.108 |
| C16 | 0.026 ± 0.013 | 0.025 ± 0.012 | 0.505 |
| C16:1 | 0.19 ± 0.064 | 0.20 ± 0.058 | 0.175 |
| C18 | 0.58 ± 0.19 | 0.60 ± 0.21 | 0.370 |
| C18:1 | 0.56 ± 0.17 | 0.57 ± 0.15 | 0.609 |
| C18OH | 0.014 ± 0.0045 | 0.014 ± 0.0039 | 0.999 |
## A clustering heatmap illustrating the relationship between thyroid hormones and carnitine metabolites
We used a clustering heatmap to describe the relationship between THs and 31 carnitine-related metabolites (Figure 1). In the row clustering step, pregnant women were grouped according to TH levels. We defined the low TH level and the high TH level as less than 10th centile and as more than 10th centile, respectively, of all participants’ full range. In addition, every biomarker was clustered into different subgroups on the column side according to their color patterns in the center grids of heatmap. Red color indicates a high expression content, and blue color indicates a low expression content. The level of C0 is higher in the high FT4–level and high TSH–level groups. Heatmaps provided a systematic and clustered visualization of the analyzed data, facilitating monitoring of TH levels and carnitine-related metabolites.
**Figure 1:** *A clustering heatmap illustrating the classification between carnitine-related metabolites and thyroid hormones. Both the rows of thyroid hormones and the columns of carnitine-related metabolites have been clustered, respectively.*
## Relationship between C0 and TSH, FT3, FT4, and TT4
The relationships between C0 and TSH, FT3, FT4, and TT4 of all pregnant women are shown in Figure 2. We observed that C0 was positively correlated with TSH ($r = 0.31$, $P \leq 0.001$), FT3 ($r = 0.37$, $P \leq 0.001$), and FT4 ($r = 0.59$, $P \leq 0.001$). There was no significant correlation between C0 and TT4.
**Figure 2:** *The relationship between free carnitine and thyroid hormones (FT4, FT3, TSH, and TT4). The correlation between C0 and FT4 (A), FT3 (B), TSH (C), and TT4 (D) was shown above. There were significant positive correlations between C0 and FT4, FT3, and TSH rather than TT4. r = Spearman’s correlation coefficient.*
## Effects of C0, TSH, FT3, FT4, and TT4 on birth weight
In the second trimester, a higher TSH (r = −0.48, $$P \leq 0.028$$) level was associated with a lower birth weight (Figure 3D). There was also a negative association between C0 (r = −0.55, $P \leq 0.001$) and FT4 (r =−0.64, $P \leq 0.001$) and birth weight (Figures 3A, B). There was no statistically significant difference between TT4 and birth weight (Figures 3C, E).
**Figure 3:** *Association of maternal free carnitine and thyroid hormones in the second trimester pregnancy and neonate birth weight. Linear regression models for C0 (A), FT4 (B), FT3 (C), TSH (D), TT4 (E) with birth weight, as predicted mean with 95% CI, were shown above. C0, FT4, and TSH negatively influenced neonate birth weight (all P < 0.05), whereas there were no statistically association between FT3 and TT4 and neonate birth weight. Analyses were adjusted for maternal age, BMI, parity, and fetal sex.*
Linear multiple regression model demonstrated that TSH, FT4, and C0 were negatively associated with birth weight. The birth weight decreased by 12.079 g ($95\%$ CI: −18.642, −8.532), by 29.203 g ($95\%$ CI: −33.149, −15.511), and by 21.079 g ($95\%$ CI: −29.842, −13.316) for every unit increase in C0, FT4, and TSH concentrations, respectively (Table 5).
**Table 5**
| Variables | β coefficient | P-value |
| --- | --- | --- |
| TSH | −21.079 (−29.842, −13.316) | 0.013 |
| FT4 | −29.203 (−33.149, −15.511) | 0.004 |
| FT3 | −22.728 (−37.836, 11.456) | 0.125 |
| TT4 | −18.185 (−13.947, 36.511) | 0.403 |
| C0 | −12.079 (−18.642, −8.532) | 0.038 |
## Combined effects of C0 and TSH, FT3, FT4, or TT4 on birth weight
We found that the level of TSH, FT3, FT4, and C0 in the second trimester would have greater effects on birth weight. Meanwhile, we found a relationship between C0 and thyroid function parameters. Therefore, we assessed the combined effects of C0 and THs, including TSH, FT3, FT4, and TT4, on birth weight. Figure 4 shows the heatmaps (fulfilled contour plot) for the combined association of C0 (x-axis) and TSH, FT4, or FT3 in the second trimester (y-axis) with birth weight (z-axis: red indicates higher birth weight, and blue indicates lower birth weight). For the low level of C0, the birth weight will be decreased, with a gradual increase in the FT4/FT3 level. For the high level of C0, the effect on birth weight was more obvious. When carnitine is increased to an appropriate level and the thyroid concentration is high (FT4: 18.4 pmol/L; C0: 17.5 µmol/L; FT3: 4.3 pmol/L; and C0: 16.3 µmol/L), the birth weight would be at the average level. When the FT4//FT3 level is high, even if the carnitine level decreases, the birth weight may remain at the present level. Free thyroxine (FT4/FT3) may play a greater role in regulating birth weight than C0. There was a considerable difference according the combination of FT4 and C0 in the second trimester ($P \leq 0.001$, Figure 4A). Lower FT4 and C0 levels were associated with a 0.8 SD higher birth weight. A lower FT4 level but a median C0 level had an influence of a 0.5 SD higher birth weight. A low FT4 level and a higher C0 level had no effect on birth weight. A low FT3 level and a low C0 level were associated with a 0.35 SD higher birth weight. A low FT3 level but a median C0 level were associated with a 0.2 SD higher birth weight. A low FT3 level and a higher C0 level had no effect on birth weight. In line with these analyses, a low FT4 level and a low C0 level were associated with more pronounced effects on birth weight. However, the effects estimate of birth weight did not different when combinations of TSH and C0, TT4, and C0.
**Figure 4:** *Combined effects of C0 and thyroid hormones in the second trimester on birth weight. Heatmap (filled contour plot) for the correlation of birth weight (red indicates increased gestational age–adjusted birth weight, and blue indicates decreased gestational age–adjusted birth weight) according to the interaction of C0 and FT4 (A), FT3 (B), TSH (C), and TT4 (D) in the second trimester. Analyses were adjusted for maternal age, BMI, parity, and fetal sex.*
## Discussion
Our research showed that there was a significantly positive correlation between C0 and TSH, FT3, and FT4. In addition, C0, FT4, and TSH all had significantly negative influence on newborn birth weight. Therefore, we evaluated the co-effects of C0 and THs on birth weight and found that the combinations of FT4 and C0 and of FT3 and C0 were significantly associated with birth weight, whereas that of TSH and C0 and that of TT4 and C0 were not.
Our study showed that FT4 and TSH had a significantly negative effect on neonate birth weight, whereas FT3 and TT4 had not. These results are consistent with that of Zhang et al., who found that higher TSH or FT4 concentrations throughout pregnancy were associated with lower birth weight [2]. Other studies also showed that babies born to mothers with higher serum FT4 levels had an elevated risk for SGA, whereas those with lower serum FT4 levels had a higher risk for LGA [10, 11]. Low FT4 levels, which may lead to an increase in circulating glucose, are associated with an increased risk of GDM, resulting in a higher placental glucose transport to the fetus and subsequent fetal weight gain [10, 12]. Leon et al. also reported that subjects with low FT4 levels were significantly associated with a higher insulin resistance index, thus leading to an increased risk of LGA [12]. Another potential mechanism is that higher TSH and FT4 levels accelerate the degradation of lipids and proteins, resulting in chronic energy deficiency in pregnant women, which has been shown to have a negative impact on neonate birth weight [2]. FT4 is the active component of TT4, the most abundant TH in the body. Because the rate of conversion of TT4 to FT4 in vivo is limited by enzymes, there is no significant relationship between TT4 and birth weight. FT3 is three to five times more active than FT4. It is unclear whether maternal T3 actually crosses the placenta. FT3 is a TH that plays a direct biological role. Activation of deiodinase leads to a high rate of conversion of FT4 to FT3, resulting in low FT4 levels and high FT3 levels. High FT3 levels increase fetal weight directly through anabolic effects on fetal metabolism and stimulation of fetal oxygen consumption [9]. A study showed that FT3 levels were positively associated with gestational weight gain in pregnant women [13]. It has been reported that higher FT3 levels are associated with neonatal obesity, but the mechanism by which T3 affects fetal weight is unclear [9]. Notably, THs are necessary for fetal cell differentiation and triggering organ development events in early pregnancy, and both high and low maternal FT4 levels were associated with adverse effects on birth weight [2]. Korevaar et al. found an inverted U-shaped association of maternal FT4 with child IQ and gray matter volume [14].
The mitochondrial matrix enzyme, carnitine acetyltransferase, catalyzes the conversion of A-CoA and C0 to acetyl-carnitine and free Co-A, which plays a vital role in the production of energy in skeletal muscle, whereas TH is known to regulate several enzymes in this pathway [15]. Maebashi et al. [ 16] reported that urinary carnitine excretion was positively correlated with serum thyroxine concentrations, with a significantly higher mean carnitine excretion in patients with hyperthyroidism and a lower carnitine excretion in patients with hypothyroidism compared with that in the control group. After correction of thyroid status, urinary carnitine excretion returned to normal in both groups [16]. Another study examined total, free, and esterified carnitine levels in the skeletal muscle of patients with hyperthyroidism and hypothyroidism before and after drug treatment [17]. A significant decrease was observed in total muscle carnitine concentrations in patients with hyperthyroidism compared with that in control subjects, which largely attributed to a decrease in the esterified carnitine portion. Total muscle carnitine levels were reduced in patients with hypothyroidism yet did not reach statistical significance, and no significant differences were found in esterified carnitine concentrations compared with control values. Meanwhile, Wong et al. [ 15] found that the level of acylcarnitine was relatively unremarkable in thyroid diseases. Other researchers argued that carnitine impairs the access of TH to the nucleus, thus decreasing the activity of TH [5, 18]. An observational pilot study found that the symptoms of patients with subclinical hyperthyroidism relieved obviously after taking L-carnitine and selenium without any significant changes of their endocrine status [19]. C0 was found to be negative on birth weight, and the combination of C0 and FT4 and of C0 and FT3 would significantly decrease the birth weight in this study, which reflected that C0 had a synergistic effect with THs, and assessment of total serum carnitine and changes in urinary carnitine excretion might help to find underlying mechanisms.
The present study has certain limitations. First, neonate birth weight could be influenced by many internal and external factors, including sex hormones, diet, medicine, and heredity. With so many confounding factors, subjects enrolled in this study could not be matched completely. Second, we did not assess the total serum carnitine and urinary carnitine, which might help explain the changing and transition of carnitine during pregnancy and the underlying mechanisms. We do find the co-effect of C0 and THs in the second trimester and their influence on neonate birth weight.
In conclusion, C0 and TH are of great importance in neonate birth weight, and routine examination of C0 and TH in the second trimester has a positive effect on the intervention of birth weight.
## 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 IRB-20220254-R. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
MY contributed to the collection, analysis, and interpretation of data as well as manuscript preparation. MS and QW contributed to the data collection and analysis, and YJ contributed to the interpretation of data. CJ contributed to the language editing. JX and QL contributed to the study design, data interpretation, and manuscript preparation. QL is the guarantor of this work and, as such, has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. 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. Howe CG, Eckel SP, Habre R, Girguis MS, Gao L, Lurmann FW. **Association of prenatal exposure to ambient and traffic-related air pollution with newborn thyroid function: Findings from the children's health study**. *JAMA Netw Open* (2018) **1**. DOI: 10.1001/jamanetworkopen.2018.2172
2. Zhang C, Yang X, Zhang Y, Guo F, Yang S, Peeters RP. **Association between maternal thyroid hormones and birth weight at early and late pregnancy**. *J Clin Endocrinol Metab* (2019) **104**. DOI: 10.1210/jc.2019-00390
3. Krassas GE, Poppe K, Glinoer D. **Thyroid function and human reproductive health**. *Endocr Rev* (2010) **31**. DOI: 10.1210/er.2009-0041
4. Galland S, Georges B, Le Borgne F, Conductier G, Dias JV, Demarquoy J. **Thyroid hormone controls carnitine status through modifications of gamma-butyrobetaine hydroxylase activity and gene expression**. *Cell Mol Life Sci* (2002) **59**. DOI: 10.1007/s00018-002-8445-3
5. Wang Y, Li X, Yang Q, Wang W, Zhang Y, Liu J. **Granulocyte-Colony-Stimulating factor effectively shortens recovery duration in anti-Thyroid-Drug-Induced agranulocytosis: A systematic review and meta-analysis**. *Front Endocrinol (Lausanne)* (2019) **10**. DOI: 10.3389/fendo.2019.00789
6. Manta-Vogli PD, Schulpis KH, Dotsikas Y, Loukas YL. **The significant role of carnitine and fatty acids during pregnancy, lactation and perinatal period**. *Nutr support specific groups pregnant women Clin Nutr* (2020) **39**. DOI: 10.1016/j.clnu.2019.10.025
7. Chen Y, Liu Y, Zhang Y, Hu R, Qian Z, Xian H. **Gestational weight gain per pre-pregnancy body mass index and birth weight in twin pregnancies: A cohort study in wuhan, China**. *Sci Rep* (2018) **8** 12496. DOI: 10.1038/s41598-018-29774-z
8. Lee SY, Cabral HJ, Aschengrau A, Pearce EN. **Associations between maternal thyroid function in pregnancy and obstetric and perinatal outcomes**. *J Clin Endocrinol Metab* (2020) **105**. DOI: 10.1210/clinem/dgz275
9. Forhead AJ, Fowden AL. **Thyroid hormones in fetal growth and prepartum maturation**. *J Endocrinol* (2014) **221** R87-R103. DOI: 10.1530/JOE-14-0025
10. Zhu YD, Han Y, Huang K, Zhu BB, Yan SQ, Ge X. **The impact of isolated maternal hypothyroxinaemia on the incidence of large-for-gestational-age infants: the ma'anshan birth cohort study**. *BJOG* (2018) **125**. DOI: 10.1111/1471-0528.15107
11. Haddow JE, Craig WY, Neveux LM, Haddow HR, Palomaki GE, Lambert-Messerlian G. **Implications of high free thyroxine (FT4) concentrations in euthyroid pregnancies: the FaSTER trial**. *J Clin Endocrinol Metab* (2014) **99**. DOI: 10.1210/jc.2014-1053
12. Leon G, Murcia M, Rebagliato M, Alvarez-Pedrerol M, Castilla AM, Basterrechea M. **Maternal thyroid dysfunction during gestation, preterm delivery, and birthweight**. *Infancia y Medio Ambiente Cohort Spain Paediatr Perinat Epidemiol* (2015) **29**. DOI: 10.1111/ppe.12172
13. Kahr MK, Antony KM, DelBeccaro M, Hu M, Aagaard KM, Suter MA. **Increasing maternal obesity is associated with alterations in both maternal and neonatal thyroid hormone levels**. *Clin Endocrinol (Oxf)* (2016) **84**. DOI: 10.1111/cen.12974
14. Korevaar TIM, Medici M, Visser TJ, Peeters RP. **Thyroid disease in pregnancy: new insights in diagnosis and clinical management**. *Nat Rev Endocrinol* (2017) **13**. DOI: 10.1038/nrendo.2017.93
15. Wong S, Hannah-Shmouni F, Sinclair G, Sirrs S, Dahl M, Mattman A. **Acylcarnitine profile in thyroid disease**. *Clin Biochem* (2013) **46**. DOI: 10.1016/j.clinbiochem.2012.10.006
16. Maebashi M, Kawamura N, Sato M, Imamura A, Yoshinaga K. **Urinary excretion of carnitine in patients with hyperthyroidism and hypothyroidism: augmentation by thyroid hormone**. *Metabolism* (1977) **26**. DOI: 10.1016/0026-0495(77)90101-9
17. Sinclair C, Gilchrist JM, Hennessey JV, Kandula M. **Muscle carnitine in hypo- and hyperthyroidism**. *Muscle Nerve* (2005) **32**. DOI: 10.1002/mus.20336
18. Benvenga S, Lakshmanan M, Trimarchi F. **Carnitine is a naturally occurring inhibitor of thyroid hormone nuclear uptake**. *Thyroid* (2000) **10**. DOI: 10.1089/thy.2000.10.1043
19. Nordio M. **A novel treatment for subclinical hyperthyroidism: A pilot study on the beneficial effects of l-carnitine and selenium**. *Eur Rev Med Pharmacol Sci* (2017) **21**
|
---
title: What and how do different stakeholders contribute to intervention development?
A mixed methods study.
authors:
- Emmy Racine
- Lauren O Mahony
- Fiona Riordan
- Gráinne Flynn
- Patricia M. Kearney
- Sheena M. McHugh
journal: HRB Open Research
year: 2023
pmcid: PMC9989546
doi: 10.12688/hrbopenres.13544.2
license: CC BY 4.0
---
# What and how do different stakeholders contribute to intervention development? A mixed methods study.
## Abstract
Background: UK Medical Research Council guidelines recommend end-user involvement in intervention development. There is limited evidence on the contributions of different end-users to this process. The aim of this Study Within A Trial (SWAT) was to identify and compare contributions from two groups of end-users - people with diabetes’ (PWD) and healthcare professionals’ (HCPs), during consensus meetings to inform an intervention to improve retinopathy screening uptake.
Methods: A mixed method, explanatory sequential design comprising a survey and three semi-structured consensus meetings was used. PWD were randomly assigned to a PWD only or combined meeting. HCPs attended a HCP only or combined meeting, based on availability. In the survey, participants rated intervention proposals on acceptability and feasibility. Survey results informed the meeting topic guide. Transcripts were analysed deductively to compare feedback on intervention proposals, suggestions for new content, and contributions to the final intervention.
Results: Overall, 13 PWD and 17 HCPs completed the survey, and 16 PWD and 15 HCPs attended meetings. For 31 of the 39 intervention proposals in the survey, there were differences (≥$10\%$) between the proportion of HCPs and PWD who rated proposals as acceptable and/or feasible. End-user groups shared and unique concerns about proposals; both were concerned about informing but not scaring people when communicating risk, while concerns about resources were mostly unique to HCPs and concerns about privacy were mostly unique to PWD. Fewer suggestions for new intervention content from the combined meeting were integrated into the final intervention as they were not feasible for implementation in general practice. Participants contributed four new behaviour change techniques not present in the original proposals: goal setting (outcome), restructuring the physical environment, material incentive (behaviour) and punishment.
Conclusions: Preferences for intervention content may differ across end-user groups, with feedback varying depending on whether end-users are involved simultaneously or separately.
## Amendments from Version 1
We would like to thank the reviewers for their suggestions and comments to improve the academic merit of our research. We have addressed each on a point-by-point basis in the responses section. Main amendments made to the paper include: a. we accept that in most interventions, patients or service users should be considered ‘ key players that everyone else has a stake in’, however, this current intervention was a multilevel intervention which targeted both people with diabetes and healthcare professionals, that is, it had components that targeted people with diabetes (i.e., personal testimonials, reminders, information provision etc.) and professionals working in general practice (i.e., audit, feedback, electronic prompts etc.). It was made clear at the outset of the meetings that the focus was both people with diabetes and HCPs. We agree that this positioning likely influenced how PWD (and HCPs) contributed during the combined meeting. We aim to reflect this in the discussion on our previous analysis of both PWD and HCP experiences of taking part in the consensus meetings. We also agree with the suggestion that the researchers were an ‘invisible power’ in the decision-making process, who influenced the final intervention and have added a paragraph to the discussion section to address this. b. In terms of PPI methodology, we have discussed the reviewer's suggestion to clarify the distinction between research participants and PPI contributors and have re-written the sentences in question. In terms of PPI involvement in this study, GF was involved throughout the research process as has been correctly pointed out. We also consulted with an existing PPI group on the design of the consensus meeting invitation letter, evidence summary and self- completion survey. We have added further information to the methods section to give a more accurate depiction of the role of PPI in this study.
## List of abbreviations
APEASE Affordability, Practicability, Effectiveness, Acceptability, Side effects and Equity BCT Behavioural Change techniques DRS Diabetic Retinopathy Screening GP General Practitioner
HCP Health Care Professional NHS National Health Service PN Practice Nurse PPI Patient and Public Involvement
PWD People With Diabetes SMS Short Message Service SPSS Statistical Package for the Social Sciences SREC Social Research Ethics Committee
SWAT Study Within A Trial
## Introduction
According to the UK Medical Research Council guidance on the development and evaluation of complex interventions, interventions should be developed with user involvement, drawing on existing evidence and appropriate theory 1. User involvement usually includes those who will deliver the intervention (often healthcare professionals [HCPs]) and the intended target population (often patients and the public). It is expected to improve the intervention fit with the target group’s perceived needs enhancing acceptability; feasibility; evaluability and adoption 2, 3.
While some studies have found that different end-users have similar priorities and preferences when making decisions about health research and service delivery 4, 5, other studies have found that different end-users endorse different perspectives 6, 7. In the context of intervention development, limited evidence exists on what different intervention users contribute to the process. Morton et al. have suggested that different stakeholders may have different priorities for intervention content 8. For instance, the cost of a proposed intervention might be more important than feasibility for intervention commissioners, whereas those receiving the intervention may be more concerned with its acceptability. However, more substantive research is needed to empirically examine and compare what different end-users contribute to the intervention development process.
Furthermore, group dynamics are complex, and some user groups may find it more difficult to voice their priorities and perspectives compared with others 9. Studies involving end-users in intervention development tend to treat all end-users (e.g., patients and HCPs) as one homogenous group 10– 12. We previously compared participants’ experiences of taking part in meetings to inform the development of an intervention to increase diabetic retinopathy screening attendance 13. Three meetings were held comprising people with diabetes only; a combined meeting of people with diabetes and HCPs; and a HCP only meeting. We found that involving both people with diabetes and HCPs in the same group led to a perceived lack of common ground where both groups felt undervalued by the other group and were reluctant to express their opinions 13. While these findings might suggest that intervention end-users may find it more acceptable to involve each group separately, we are also keen to know whether their contributions during these meetings differed according to group composition. Understanding whether user contributions differ according to group composition could enable researchers to design and conduct more appropriate and effective user involvement activities which in turn could potentially improve intervention fit with the target group’s perceived needs.
The aim of this Study Within A Trial (SWAT) was to identify and compare people with diabetes’ and HCPs’ contributions during three consensus meetings to inform intervention development, including their feedback on the acceptability and feasibility of intervention content, suggestions for new intervention content, and contributions to the final intervention.
## Methods
This SWAT was embedded in the intervention development phase of the Improving Diabetes Eye-Screening Attendance (IDEAs) pilot trial 14. IDEAs used a systematic three-step process combining theory, user involvement and evidence on intervention effectiveness to develop a multifaceted intervention targeting people with diabetes and HCPs to improve uptake of RetinaScreen, a national Diabetic Retinopathy Screening (DRS) programme 15. As part of the user involvement process, three semi-structured consensus meetings were conducted to review and discuss proposals for intervention content.
## Design
This SWAT is a mixed method study using an explanatory sequential design 16. Quantitative data (self-completion participant survey) were collected and analysed first, followed by the qualitative data (consensus group meetings) which were collected and analysed second in sequence 17. The quantitative results provided an overview of participant ratings of acceptable and feasible intervention content, while the qualitative analysis allowed for further exploration of why participants rated intervention content the way they did by using a topic guide informed by survey findings.
## Recruitment
People with diabetes People with diabetes were recruited using an information flyer developed by the research team including a graphic designer (http://doi.org/10.5281/zenodo.4321202). The flyer was distributed using a range of recruitment strategies including social marketing recruitment, community outreach recruitment, health system recruitment, and partnering with other organisations. All individuals who contacted the study team and returned a short demographic survey (Supplementary File 1 in the *Extended data* 18) were randomly assigned (using an online random number generator) to either the meeting for the people with diabetes only, or the combined meeting.
Health care professionals HCPs were recruited through local professional networks known to the study team. An email invitation was sent to 50 HCPs (practice nurses, diabetes nurse specialists, general practitioners, and specialist physicians). All HCPs were allocated based on their availability to the HCP-only meeting or combined meeting. Further details on the recruitment process have been described in detail elsewhere 13.
## Data collection
Quantitative phase Before each consensus meeting, participants were sent an evidence summary of barriers to and enablers of attendance at diabetic retinopathy screening, and interventions to address non-attendance (Supplementary File 2 in the *Extended data* 18), and a self-completion survey (Supplementary File 3 in the *Extended data* 18). The evidence summary and survey were designed with input from the Irish National Adult Literacy Agency and a Patient and Public Involvement (PPI) group and revised based on their feedback.
The survey outlined 39 proposals for intervention content that were grouped at the practice-level (‘ways to encourage the practice staff to make sure person attends’) and patient-level (‘ways to encourage the person to attend diabetes eye screening’). The proposals contained operationalised behaviour change techniques (BCTs), defined as an “observable, replicable, and irreducible components of an intervention” that have the potential to change behaviour 19. The proposals (operationalised techniques) were short statements/descriptions of how the selected BCT would be put into practice 20, in line with the study focus on increasing diabetic retinopathy screening uptake. The BCTs in the survey were selected to address known barriers to and enablers of screening attendance based on previous formative research conducted by the IDEAs research team 15 and existing evidence of their effectiveness either in interventions to increase retinopathy screening attendance or interventions in other settings 21, 22. A total of 24 unique BCTs were operationalised across the 39 intervention proposals in the survey. Further details on these 24 BCTs has been provided in Supplementary File 4 in the *Extended data* 18.
In the survey, participants were asked to rate the acceptability and feasibility of each proposal. All items were rated on a Likert response scale ranging from 1 to 5 (from ‘strongly disagree’ to ‘strongly agree’) with higher scores indicating greater acceptability or feasibility. These survey questions were adapted from existing measures developed by Weiner et al. to rate implementation acceptability and feasibility 23. Acceptability was defined as the perception among end-users that the intervention proposal was agreeable or satisfactory. Feasibility was defined as the extent to which the intervention proposal could be successfully implemented in general practice. People with diabetes received a paper format of the survey while HCPs received an electronic format.
Qualitative phase Following completion of the surveys, participants took part in one of three consensus group meetings. Each meeting was held for two hours in University College Cork and was facilitated by the same facilitator experienced in consensus group techniques/processes. This facilitator was a male professor of health services research who held no relationship with participants. This individual was a member of the Project Steering Group, acting in an advisory capacity but not actively involved in data collection and analysis beyond the consensus meetings. This individual was invited to facilitate the meetings as they could adopt a neutral position having no vested interest in any of the intervention components.
During the meetings, a summary of the ratings of acceptability/feasibility was presented to participants. This was followed by a series of small group discussions (facilitated by members of the research team) where participants were asked to discuss how each intervention proposal would work in practice (See Supplementary File 5 in the *Extended data* 18 for Facilitator Guide). Facilitators asked participants to discuss and give feedback on both practice-level and patient-level proposals. Prompts about patient-level proposals included 1) who should deliver the message to remind patients to attend diabetes eye screening? 2) how should the message be delivered? 3) when should the message be delivered? and 4) what should the message contain? Participants were asked to focus their discussion on proposals where the consensus on acceptability and feasibility based on the survey was unclear. However, given the semi-structured nature of the meetings, participants also made new suggestions. The small group discussions and the feedback to the larger group were digitally audio recorded with participant consent.
## Data analysis
Participant survey responses were entered into SPSS software (version 26, RRID:SCR_016479) and analysed using descriptive statistics. Consensus meeting transcripts were analysed using NVivo 12 software (RRID:SCR_014802). If this software were unavailable, it would be possible to conduct the analysis using Excel and Word.
Comparing end-users’ feedback on the acceptability and feasibility of intervention content To examine participants’ ratings of the acceptability and feasibility of intervention proposals, the five-point Likert scale used in the survey was collapsed into three categories: ‘disagree’ [1 strongly disagree, 2 disagree], ‘neither disagree or agree’ [3] and ‘agree’ [4 agree, 5 strongly agree]. Contingency tables were generated for each intervention proposal by participant type (HCP or people with diabetes) and Fisher’s exact test was used as appropriate 24. Results were examined to identify proposals which had a difference (≥$10\%$) between the proportion of HCPs and people with diabetes who agreed that intervention proposal was feasible and/or acceptable.
Guided by the survey results, interview transcripts were analysed using deductive content analysis. A codebook (developed a priori by LOM) designed to mirror the self-completion survey to identify and code feedback on specific proposals was used. Participants in the combined meeting were asked to reach group consensus on intervention proposals, therefore it was difficult to attribute feedback exclusively to people with diabetes or HCPs or both. Therefore, the people with diabetes only meeting and the HCP only meeting were analysed before the combined meeting was analysed, to allow the researchers to see whether feedback from the combined meeting echoed that of the people with diabetes only and HCP only meetings.
To compare participants’ feedback on the acceptability and feasibility of intervention proposals, thematic analysis was performed by LOM, guided by joint displays of the survey results and qualitative coding. The joint displays were examined for recurring patterns between survey ratings and discussion during the consensus meetings, to identify reasons for agreement/disagreement e.g., what was or was not acceptable/feasible to whom, and why. An overview of this sequence of mixed methods is provided in Figure 1.
**Figure 1.:** *Overarching sequence of mixed methods.*
Comparing end users’ suggestions for new intervention content To identify and compare end-users’ suggestions for new intervention content, two researchers (ER and FR) conducted a deductive content analysis 25 to identify suggested changes to proposed intervention content and suggestions of additional intervention content. Both researchers read the consensus meeting transcripts multiple times (data familiarisation) and then independently extracted all suggestions made by participants in relation to intervention content and mode of delivery. A suggestion was defined prior to data analysis as any suggestion about intervention content or mode of delivery proposed by a member of the group, at any stage during the meeting, that was agreed with by one or more other members of the group. Agreement or disagreement between participants was ascertained based on explicit verbal expression or sounds or noises which conveyed their agreement or disagreement (e.g., mmm). The two researchers met to discuss the suggestions they had extracted. Any differences were discussed, and agreement was reached by consensus on the list of suggestions put forward by participants. Each new suggestion was then coded (yes/no) according to whether it would be feasible to incorporate into the intervention to be delivered. The scope of the intervention was defined as: To identify how each new suggestion aligned with existing behavioural change techniques, they were mapped to Behaviour Change Technique Taxonomy (BCTTv1) 26. Further information on how this mapping was conducted is provided in Supplementary File 6 in the *Extended data* 18.
Comparing end users’ contributions to the final intervention Using deductive content analysis, one researcher (ER) categorised (yes/no) all recommendations (including feedback on proposals and suggestions for new intervention content) according to whether they were incorporated into the final intervention. Full details about the decision process regarding the final intervention content has been published elsewhere 15. The final decision on the intervention content was made by a subgroup of the IDEAs study research team and a GP collaborator, basing decisions on the APEASE (affordability, practicability, effectiveness, acceptability, side effects, equity, sustainability) criteria. Practicality and acceptability criteria were populated based on findings from the rating survey and the discussions during the consensus meetings. The effectiveness criterion was based on a rapid evidence review of different approaches to improve screening uptake. Remaining criteria (affordability, equity, side-effects (unintended consequences), sustainability) were based on group discussions about what was feasible, bearing in mind previous formative research with patients and healthcare professionals and organisational factors relating to the primary care environment.
## Patient and Public Involvement (PPI)
A PPI contributor (GF) was involved in the SWAT from the outset. GF is a person with diabetes, previously known to the lead author (ER). She contributed to the initial discussions about the study which ultimately informed the SWAT grant application, reviewed the grant application prior to submission and made changes to its content including the addition of disseminating the research amongst people with diabetes. GF was also involved in the development of materials used to recruit people with diabetes and assisted the research team with recruitment by posting recruitment flyers online via social media networks. She contributed to and reviewed each draft of this manuscript and is a co-author on this publication. The lead author also worked with a separate primary care research PPI group to develop and refine the materials that were sent to participants prior to the consensus meeting. PPI contributors in this group were asked to review draft versions of the consensus meeting invitation letter, evidence summary and self-completion questionnaire. Significant changes were made to the wording and layout of the materials as a result of their input. For example, section headings were added to the self-completion questionnaire which reduced its length from five pages to three pages. After the consensus meetings were conducted, the IDEAs study worked with a dedicated PPI group throughout the duration of the trial 15.
## Ethical approval
The study received ethical approval from the Social Research Ethics Committee (SREC) at University College Cork (Log number 2018-122, approval received $\frac{13}{08}$/2018). Written informed consent was obtained from all participants prior to completing the rating survey and taking part in the consensus meetings.
## Comparing end users’ feedback on the acceptability and feasibility of intervention content
In total, 30 participants (13 people with diabetes and 17 HCPs) completed and returned the surveys. Missingness within the data ranged from $3.3\%$ to $6.7\%$, depending on the survey proposal. There was incomplete data for 6 participants (4 people with diabetes, 2 HCPs). Table 1 presents the 31 proposals which had differences (≥$10\%$) between the proportion of HCPs and people with diabetes who agreed the proposal was acceptable and/or feasible 18.
**Table 1.**
| Intervention Component ( embedded BCT) | Proposal ( Operationalised BCT) | Self-completion survey | Self-completion survey.1 | Self-completion survey.2 | Self-completion survey.3 | Self-completion survey.4 | Self-completion survey.5 | Semi-structured consensus meetings |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | Agreed proposal acceptable | Agreed proposal acceptable | Agreed proposal acceptable | Agreed proposal feasible | Agreed proposal feasible | Agreed proposal feasible | Related concern (or preference) where indicated in the data (Joint; HCP; People with diabetes) |
| | | People with diabetes % (n) | HCP % (n) | Diff. % | People with diabetes % (n) | HCP % (n) | Diff. % | Related concern (or preference) where indicated in the data (Joint; HCP; People with diabetes) |
| Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals |
| (i) Using a personal story from someone else with diabetes who delivers the message who… ( 9.1 Credible source) | …is a similar age and profile to People with diabetes and explains how screening was a way for them to take charge of their health. (6.2 Social comparison and 15.1 Verbal persuasion about capability) | 84 (11) | 64.7 (11) | 19.3 | 75 (9) | 58.8 (10) | 16.2 | |
| | …has retinopathy and tells them the benefits of screening. (5.3 Information about social and environmental consequences and 11.2 reduce negative emotions) | 92.3(12) | 82.2 (15) | 10.1 | 100 (12) | 58 (10) | 42 | Balancing act: informing but not scaring people with diabetes (Joint) |
| | …has retinopathy and tells them it is important to go to screening before it is too late, there may be no symptoms and everyone with diabetes is at risk. (5.1 information about health consequences and 5.5 Anticipated regret) | 100 (13) | 94.1 (16) | 5.9 | 83.3 (10) | 64.7 (11) | 18.6 | Balancing act: informing but not scaring people with diabetes (Joint) |
| | …wishes they went to screening sooner and prompts the person to think about the regret they will feel if they do not attend screening. (5.5 Anticipated regret and 6.2 Social comparison) | 84.6 (11) | 58.8 (10) | 15.8 | 75 (9) | 47.1 (8) | 27.9 | Balancing act: informing but not scaring people with diabetes (Joint) |
| | …explains there is no harm from drops used during screening and the overall benefits outweigh the short-term discomfort. (5.1 Information about health consequences) | 92.3 (12) | 82.4 (14) | 9.9 | 83.3 (10) | 64.7 (11) | 18.6 | |
| | …provides an observable example that shows them how to consent or attend. (6.1 Demonstration of behaviour) | 76.9 (10) | 76.5 (12) | 0.4 | 75 (9) | 52.9 (9) | 22.1 | |
| | …delivers a message recognising the anxiety people might feel but emphasizes the positive consequences of attending. (11.2 Reducing negative emotions) | 100 (13) | 94.1 (16) | 5.9 | 83.3 (10) | 64.7 (11) | 18.6 | |
| | …prompts the person to imagine the outcomes of attending vs. not attending. (9.2 Pros and cons) | 84.6 (11) | 52.9 (9) | 31.7 | 66.7 (8) | 47.1 (8) | 19.6 | |
| (ii) Someone in the practice could… (9.1 Credible source) | …Encourage the person to attend screening. (3.1 Social support (unspecified)) | 92.3 (12) | 100 (17) | 7.7 | 75 (9) | 94 (16) | 19 | |
| | …Tell the person that they approve of screening and hope the person will attend. (6.3 Information about other’s approval) | 92.3 (12) | 94.1 (16) | 1.8 | 83.3 (10) | 93.8 (15) | 10.5 | |
| | …Persuade the person they will be able to attend screening (e.g., help them to think about times they successfully managed their diabetes or attended appointments). (15.3 Focus on past success) | 69.2 (9) | 76.5 (13) | 7.3 | 58.3 (7) | 70.6 (12) | 12.3 | |
| | …Explain the difference between routine eye checks and the screening test, what both tests can and cannot tell them, and that routine checks are not a substitute. (5.1 Information about health consequences) | 92.3 (12) | 94.1 (16) | 1.8 | 75 (9) | 94.1 (16) | 19.1 | Some people with diabetes have a limited understanding of the need for and practicalities of screening (Joint preference) |
| | …Advise the person how to consent to screening and to ask for help if they are unable/unsure about how to do this (4.1 Instruction on how to perform the behaviour and 3.2 Social support (practical)) | 92.3 (12) | 100 (17) | 7.7 | 91.7 (11) | 76.5 (13) | 15.2 | |
| | …Tell the person that after their appointment they will be reassured, or they can get treated in time to stop things getting worse. (5.1 Information about health consequences and 5.6 Information about emotional consequences) | 100 (13) | 94.1 (16) | 5.9 | 91 (11) | 76.5 (13) | 14.5 | |
| | …Explain how it’s important to go to screening before it is too late, they personally are at risk and that screening applies to them. (5.1 Information about health consequences and 5.5 Anticipated regret and 5.2 Salience of consequences) | 100 (13) | 88.2 (15) | 11.8 | 91.7 (13) | 82.3 (14) | 9.4 | Balancing act: informing but not scaring people with diabetes (Joint) |
| | …Encourage the person to think of screening not as something extra, but as part of the whole package of self-management. (13.2 Framing/reframing) | 100 (12) | 88.2 (15) | 11.8 | 90.9 (10) | 88.2 (15) | 2.7 | |
| | …Help the person to make a plan about when and where they will consent and how they will attend when they get their appointment. (1.4 Action planning) | 61.5 (8) | 64.7 (11) | 3.2 | 50 (6) | 70.6 (12) | 20.6 | |
| (iii) Other ideas of how to encourage the person to attend | Arrange for support from family/friends (e.g., encouragement to consent/attend). (3.1 Social support (unspecified)) | 69.2 (9) | 58.8 (10) | 10.4 | 41.7 (5) | 47.1 (8) | 5.4 | Risks patient privacy (People with diabetes Strays outside HCPs area of responsibility (HCPs) Some people with diabetes have a limited understanding of the need for and practicalities of screening (Joint preference) |
| | Advise/arrange for practical support like transportation from family/friends. (3.2 Social support (practical)) | 66.7 (8) | 82.4 (14) | 15.7 | 58.3 (7) | 41.2 (7) | 17.1 | Risks patient privacy (People with diabetes Strays outside HCPs area of responsibility (HCPs) Some people with diabetes have a limited understanding of the need for and practicalities of screening (Joint preference) |
| | Draw the person’s attention to the number of people like them who have attended. (6.2 Social comparison) | 53.8 (7) | 58.8 (10) | 5 | 50 (6) | 64.7 (11) | 14.7 | |
| | The person with diabetes ticks off a checklist when they have consented/attended. (2.3 Self-monitoring of behaviour) | 69.2 (9) | 35.3 (6) | 33.9 | 75 (9) | 35.3 (6) | 39.7 | Relying on active participation from people with diabetes (Joint) |
| Practice-level | Practice-level | Practice-level | Practice-level | Practice-level | Practice-level | Practice-level | Practice-level | Practice-level |
| (iv) Ways to encourage the practice staff to make sure the person attends | Provide practice with observable example/ information on how to check and register people with diabetes. (4.1 Instruction on how to perform the behaviour and 6.1 Demonstration of the behaviour) | 100 (13) | 94.1 (16) | 5.9 | 91.7 (11) | 76.5 (13 ) | 15.2 | Resource implications (HCPs) |
| | Prompt practice to check the register during consultation and register person if necessary (7.1 Prompts and cues) | 92.3 (12) | 82.4 (14) | 9.9 | 91.7 (11) | 70.6 (12) | 21.1 | Resource implications (HCPs) |
| | Provide a new resource to the practice (e.g., researcher checks if person registered, consented and/or attended) (12.2 Restructuring the social environment) | 83.3 (10) | 64.7 (11) | 18.6 | 53.8 (7) | 58.8 (10) | 5 | Resource implications (HCPs) Risks patient privacy (People with diabetes) |
| | Provide checklist of ways to encourage consent/ attendance (12.5 Adding objects to the environment) | 76.9 (10) | 52.9 (9) | 24 | 58.3 (7) | 64.7 (11) | 6.4 | Resource implications (HCPs) |
| (v) Telling practices about the benefits/consequences of their patients attending/not attending | The benefits to the practice when their patients attend (e.g., receiving timely results, they have access to local service) (5.3 Information about social and environmental consequences) | 81.8 (9) | 70.6 (12) | 11.2 | 83.3 (10) | 70.6 (12) | 12.7 | Motivating practice staff to make sure the person attends screening (HCPs) |
| | Consequences when their patients do not attend (e.g., eye damage, costs of missed appointments). (5.3 Information about social and environmental consequences) | 66.7 (8) | 76.5 (13) | 9.8 | 90.9 (10) | 70.6 (12) | 20.3 | Motivating practice staff to make sure the person attends screening (HCPs) |
| (vi) Use a personal story from a patient to inform practices… (9.1 Credible Source) | … about the benefits and risks to patients of attending/not attending (5.1 Information about health consequences) | 76.9 (10) | 47.1 (8) | 29.8 | 72.7 (8) | 41.2 (7) | 31.5 | |
| | … that patients are more likely to attend screening if a health professional prompts or encourages them to do so. (9.1 Credible source) | 84.6 (11) | 70.6 (12) | 14 | 90.9 (10) | 64.7 (11) | 26.2 | |
| (vii) Give practices feedback… | …On national or international uptake or targets (2.2 Feedback on behaviour and 1.6 Discrepancy between current behaviour and goal) | 76.9 (10) | 82.4 (14) | 5.5 | 100 (11) | 88.2 (15) | 11.8 | Motivating practice staff to make sure the person attends screening (HCPs) |
| | Use a trusted source to deliver feedback and messages (e.g. colleague) 9.1 Credible Source | 91.7 (11) | 81.3 (13) | 10.4 | 76.9 (10) | 68.8 (11) | 8.1 | |
Concerns about intervention content Following integration of the survey results and qualitative feedback from the consensus meetings, themes related to the preference for and several main concerns about acceptable and feasible intervention content (Figure 2). Table 1 presents where these relate to intervention proposals and whether it was a joint concern, or preference, of both people with diabetes and HCPs, HCPs only or people with diabetes only.
**Figure 2.:** *Concerns about intervention content organised by health care provider (HCP) concerns, joint concerns, or people with diabetes’ concerns.*
The results are organised according to the joint preference, joint concerns, HCP concerns and people with diabetes’ concerns. Examples of intervention proposals that relate to each area of concern are presented, along with the survey results and a short summary of participants’ feedback from the consensus groups.
Joint preference Participants in all three meetings considered several intervention proposals to be acceptable and feasible because they believed some people with diabetes have a limited understanding of the screening process. In the survey, both people with diabetes and HCPs agreed the proposal to use someone in the practice who would explain the difference between routine eye checks and the screening test was acceptable ($92.3\%$ vs. $94.1\%$, respectively), though they differed in agreement with feasibility ($75\%$ vs. $94.1\%$, respectively). Data from the meetings provided no indication as to why people with diabetes rated feasibility lower than HCPs, however both groups flagged that there is confusion among some people with diabetes about the difference between routine eye tests and retinal screening. Participants in the combined meeting agreed that messages delivered to patients should outline the difference between routine eye tests and retinal screening and emphasise that damage can be asymptomatic to dispel the “false sense of security”. Similarly, participants in the people with diabetes only meeting thought messages should aim to increase patient understanding of the screening process. For example, highlighting the possible consequences of non-attendance and “alert you (people with diabetes) to the dangers involved”. Participants in the HCP only meeting agreed messages should emphasize that screening is free.
In the survey, less people with diabetes than HCPs agreed the proposal to arrange practical support was acceptable ($66.7\%$ vs $82.4\%$ respectively), though less HCPs agreed it was feasible ($58.3\%$ vs HCPs $41.2\%$). Participants in the people with diabetes only meeting felt many people with diabetes are not aware of the need to organise transportation for after the screening procedure, and so messages should tell people they would need support rather than arranging it for them. HCPs had concerns about the feasibility of this proposal, which are discussed below under the concern straying outside their area of responsibility.
Joint concerns Some HCPs and people with diabetes had concerns about proposals which might rely on active participation from people with diabetes, for example, the proposal for the person with diabetes to tick off a checklist when they have consented to/attended to screening. In the survey, a larger proportion of people with diabetes than HCPs agreed providing a checklist would be acceptable ($69.5\%$ and $35.3\%$, respectively) and feasible ($75\%$ and $35.3\%$, respectively). In the people with diabetes only and combined meeting, some people with diabetes felt having a checklist would help people be “proactive” in the management of their diabetes, while others thought that this would put too much responsibility on the person who “might lose or forget it”. Some of those in the HCP only meeting thought that only motivated and engaged patients would use the checklist.
Participants from all three meetings were concerned about achieving the balance between communicating the risks of diabetic retinopathy while not scaring people when informing them about screening. This concern related to several proposals to use other people with diabetes or HCPs to deliver messages. In the survey, both HCPs and people with diabetes agreed it would be acceptable to use a message from someone who has retinopathy and tells them it is important to go to screening before it is too late, there may be no symptoms and everyone with diabetes is at risk. However, $83.3\%$ of people with diabetes agreed it would be feasible compared to $64.7\%$ of HCPs. Participants across all meetings believed that “scaremongering” or “shock tactics” would not encourage people to attend. Rather than “shock” people, messages should inform them of the “truth” about the possible consequences of non-attendance and be provided “by the right person, in the right way”. Both people with diabetes and HCPs agreed that the same message (tells them it is important to go to screening before it is too late, there may be no symptoms and everyone with diabetes is at risk) when delivered by HCP rather than another person with diabetes would be acceptable ($100\%$ and $88.2\%$, respectively) and feasible ($91.7\%$ and $82.3\%$, respectively). Participants in the people with diabetes only meeting thought the GP would be the best person to deliver a message to attend screening as people “trust” their GP and are “much more inclined to listen to them”. HCPs in the HCP only and combined meeting had concerns that that delivering these messages during consultations would take a considerable amount of time.
Health care professionals’ concerns Concerns about the resource implications of delivering intervention proposals including time, staff, and money, were raised throughout all three meetings. Resource concerns were often a reason for HCPs’ lesser agreement with proposals, especially those which aimed to encourage practice staff to ensure the person attends. Few people with diabetes and HCPs thought the proposal to provide a new resource to the practice (e.g., researcher checks if person registered, consented and/or attended) was feasible ($53.8\%$ vs $58.8\%$, respectively). While both agreed the proposal to prompt practice to check the (DRS) register during consultation and register person if necessary was acceptable, a slightly lower proportion of HCPs thought it was feasible ($82.4\%$ and $70.6\%$, respectively). They emphasized not having time for multiple prompts and reminders like letters or emails; “we absolutely don't have the time. We can't take anything on, it’s just beyond unbelievable.” HCPs in both meetings had concerns about proposals to tell practices about the benefits/consequences of their patients attending/not attending. This was reflected in the different proportions who agreed such was feasible (people with diabetes $90.9\%$ vs HCP $70.6\%$, respectively). Some HCPs believed financial incentives might be best to motivate GPs to ensure their patients are registered and attend DRS. HCPs in the combined meeting suggested that once practices have a registration uptake at a particular level, they could receive financial remuneration and therefore be “incentivised to do it (register patients)”. There were also concerns about using feedback to motivate HCPs to encourage patients to attend, namely by providing practices with comparison numbers (% people attending in other practices/ nationally). This discussion arose around the proposal to give feedback on national or international uptake or targets. Some participants in the HCP only meeting felt “you would totally tap into [competitive] personalities” but there was a lack of consensus on this proposal in the combined meeting. Some participants in this meeting thought a comparator could be a useful motivator, whereas one GP noted that the differing demographic of patients across practices would make comparisons difficult. HCPs in both meetings argued that feedback needs to be specific and tailored to their practice and their patients, as national averages and practice comparisons are “totally useless” as they “cannot address that on a one-to-one level with a patient”.
As previously mentioned, the proposal to arrange practical support like transportation was not considered feasible by people with diabetes nor HCPs ($58.3\%$ vs $41.2\%$, respectively). HCPs in the HCP only and combined meeting felt this proposal strayed outside of their area of responsibility, as they mostly interpreted it as having to arrange the transportation for the patient themselves, something they felt was “not their (HCP) problem” as patients “need to take ownership and responsibility”.
People with diabetes’ concerns Participants in the people with diabetes only meeting were concerned that some proposals threatened their privacy. For example, arranging practical or social support would make it difficult for those who wish to keep their diabetes private to do so. Both people with diabetes and HCPs thought the proposal to provide a new resource to the practice like a researcher was not feasible ($53.8\%$ and $58.8\%$, respectively). A few participants in the people with diabetes only meeting were concerned about privacy should someone within the practice other than their GP/PN have access to their information. Contrastingly, more people with diabetes than HCPs thought this proposal would be acceptable ($83.3\%$ vs $64.7\%$ respectively). However, this may be explained by HCP concerns about resourcing this proposal.
## Comparing stakeholders’ suggestions for new intervention content
Participants in the people with diabetes only meeting made 26 suggestions for new intervention content, of which 7 were deemed feasible to incorporate into the final intervention ($30\%$). Participants in the combined meeting also made 26 new suggestions, of which 3 were feasible ($15\%$). Participants in the HCP only meeting made 32 new suggestions, of which 7 were feasible ($22\%$). Table 2 shows the suggestions for new intervention content that were deemed feasible to incorporate. New suggestions were deemed unfeasible to incorporate into the intervention if they could not be implemented in the *Irish* general practice setting. For example, participants in all three meetings suggested that the reminder message should be delivered by professionals outside general practice, that the national screening programme could modify their processes to make it easier for people with diabetes to register and attend the service, and that national-level changes (e.g., media campaign to improve attendance, establishing a national diabetes register) should be introduced to increase screening attendance.
**Table 2.**
| Suggestion | People with diabetes only meeting | Combined meeting | HCP only meeting | Behaviour Change Technique | Incorporated into the final intervention |
| --- | --- | --- | --- | --- | --- |
| Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals | Patient-level proposals |
| Visuals should not be gruesome | ✓ | - | ✓ | | ✓ |
| Distinguish the difference between HBA1c and retinal screening | ✓ | ✓ | - | 5.1 Information about health consequences 13.2 Framing/ reframing | □ |
| Outline that GP has noticed that the patient has not attended | ✓ | - | - | 2.2 Feedback on behaviour 6.3 Information about others’ approval | ✓ |
| GP should recommend that the patient talks to another patient at the practice | ✓ | - | - | 6.2 Social comparison 6.3 Information about others approval | □ |
| Do not use scaremongering language | - | - | ✓ | | ✓ |
| Personal story from a celebrity | - | - | ✓ | 9.1 Credible source 6.2 Social comparison 6.3 Information about others’ approval | □ |
| Provide a link to further information online | - | - | ✓ | 5.1 Information about health consequences | □ |
| Ask patients to attend as a favour to the practice to get their numbers up | - | - | ✓ | 6.2 Social Comparison 13.2 Framing/ reframing | □ |
| Tell patients that they need to prioritise their eyes, emphasise how important they are compared to other things | - | - | ✓ | 5.1 Information about health consequences | ✓ |
| Patients should be reminded to attend screening before they come to the practice to collect their next prescription as a ‘subtle threat’ | - | - | ✓ | 10.1 Material incentive (behaviour) * 14.2 Punishment * | □ |
| Practice-level proposals | Practice-level proposals | Practice-level proposals | Practice-level proposals | Practice-level proposals | Practice-level proposals |
| One person at practice dedicated to reminding patients to attend screening | ✓ | ✓/✗ 1 | - | 12.1 Restructuring physical environment * | ✓ |
| Have a chart at practice with the % numbers they want to achieve | ✓ | - | - | 1.3 Goal setting * 12.5 Adding objects to the environment | □ |
| Inform practices that they can market themselves as a practice known for good diabetes care | ✓ | - | - | 5.3 Information about social and environmental consequences | □ |
| Practice staff should be shown how to use the GP software to check screening registration and attendance | - | ✓ | - | 12.1 Restructuring physical environment | ✓ |
New suggestions deemed feasible to incorporate into the intervention mapped to 12 BCTs in the taxonomy (Table 2). There were four additional BCTs identified in the new suggestions that were not present in the intervention proposals outlined in the survey: goal setting (outcome), restructuring the physical environment, material incentive (behaviour) and punishment. Additional information on the BCTs identified is provided in Supplementary File 7 in the *Extended data* 18.
## Comparing end users’ contributions to the final intervention
The final intervention included a practice briefing, audit and feedback with technical support, practice-endorsed reminders (delivered in person, by phone and letter) and an information leaflet targeting key attitudinal and knowledge barriers. The people with diabetes only meeting had $\frac{23}{51}$ ($45\%$) recommendations incorporated into the final intervention, of these 20 were feedback on the intervention proposals and three were new suggestions. The combined meeting had $\frac{19}{49}$ ($39\%$) recommendations incorporated into the final intervention, of these 17 were proposed and two were new suggestions. The HCP only meeting had $\frac{24}{55}$ ($44\%$) recommendations incorporated into the final intervention, of these 21 were proposed and three were new suggestions. Table 2 shows the new suggestions that were incorporated into the final intervention. All three meetings made new suggestions that were deemed feasible but not incorporated into the final intervention. These suggestions, along with the reasons for exclusion (based on the APEASE criteria), are outlined in Supplementary File 8 in the *Extended data* 18.
## Summary of main findings and links to existing literature
Although there is growing awareness in the literature that involving different intervention end-users in the development process may have a different impact on the final intervention developed 8, 11, 27, to our knowledge, this is the first study to examine and compare in detail the contributions of different intervention end-users as part of a consensus approach to inform intervention development.
There were three main findings. Firstly, people with diabetes and HCPs had both shared and unique opinions about the acceptability and feasibility of some aspects of the proposed intervention content. Some opinions were shared by both end-users and were echoed throughout all three consensus meetings, for example that there is a limited understanding of the screening process, or that we should balance informing people without scaring them when communicating about screening. However, HCPs also had unique concerns related to their role as healthcare providers, while people with diabetes had their own concerns about intervention proposals which might risk their privacy. Such differences suggest that while there is a common ground when it comes to preferences for and concerns about intervention content, there are some aspects of the intervention which may be a greater priority for different end-users. Secondly, participants in all three meetings made suggestions for new intervention content which mapped to BCTs that were not present in the proposed intervention content however, participants in the combined meeting made less feasible suggestions as they could not be implemented in the *Irish* general practice setting. Finally, participants in all three meetings made recommendations that were incorporated into the final intervention. However, participants in the combined meeting had fewer recommendations incorporated than the other two meetings.
In the meetings involving people with diabetes only and HCPs only, respective groups had different opinions about the delivery of messages to attend screening e.g., who should deliver the message, when the message should be delivered, and what the message should contain. Those in the meeting of people with diabetes only tended to base their recommendations on what would be most acceptable to the person with diabetes. In contrast, participants in the HCP only meeting focused more on what was feasible from a resource perspective. These concerns are consistent with reports of increased workload and staff burnout in *Irish* general practice 28, 29. In addition, some HCPs perceived that certain intervention proposals would involve straying outside their area of responsibility. They tended to disagree with proposals which they equated to an extra job or responsibility, understandable given the increasing responsibilities in general practice for chronic disease management 30. Future intervention developers should consider these different perspectives of respective end-users so that they may involve them in the development process in the most effective way.
On the other hand, participants in this study also had joint preferences for intervention content. Both HCPs and people with diabetes were conscious that while it was important to outline the seriousness of retinopathy, there is a need to strike a balance between informing but not scaring people about the screening process and potential disease consequences from non-attendance. This aligns with the body of literature on the use, or avoidance, of fear appeals to encourage preventative health behaviours, evidence which has demonstrated that providing information about possible negative consequences may prompt defensive responses 31. For instance, one US study found that avoidance of cancer risk information was associated with lower participation in colorectal cancer screening 32. During the consensus meetings, people with diabetes and HCPs had concerns about intervention content which might scare or frighten people, such as having a message delivered by someone who is visually impaired or prompting the person to feel regret. Intervention developers should select behaviour change techniques that promote adaptive, rather than maladaptive behaviour, as suggested by a qualitative study of fear appeals as a method in behaviour change interventions 33. These joint contributions by participants in our study offer a useful perspective to intervention developers about how end-users will receive communication, but also demonstrates there are instances where end-users can share priorities for intervention content.
Our findings indicate that end-user groups’ contributions to the intervention development process can differ based on whether they are involved separately or simultaneously. Participants in the combined meeting of people with diabetes and HCPs made fewer feasible suggestions for new intervention content and fewer recommendations from this meeting were incorporated into the final intervention. This suggests their contributions may have been influenced by group composition. Our previous analysis of participants’ experiences of taking part in the consensus meetings found that, although members of the combined meeting appeared to work together, during follow-up data collection both end-user groups held different views about what intervention proposals would and would not work 13. Our aim was to elicit feedback on components that would target PWD and HCPs, but both HCPs and PwD that participated in the combined meeting were uncomfortable with asserting what the other end-user group should or should not do. To fill this void, participants went off task and made suggestions that were outside the scope of an intervention intended for primary care 13.
In this study, one skilled facilitator who was partly involved in the wider intervention development process facilitated all consensus meetings. While this was helpful in contributing to consistency, it is also possible that group dynamic and discussion might have been different had a person with diabetes co-facilitated the meetings e.g. this co-facilitator might have supported people in the combined meeting to speak on occasions where participants felt uncomfortable, or it was difficult to reach consensus. As the meeting involved small group discussion, we found this helped people to be forthcoming about their experiences and views, particularly in the meeting with PWD only.
This current study alongside our previous analysis suggests that it may be useful to involve each end-user group, those who will deliver the intervention and the intended target population, separately rather than simultaneously in a consensus process to inform intervention development. When involving different end-users together in a consensus process, researchers should also consider facilitating these groups differently, paying special attention to acknowledge potentially unique views while also reaching consensus. Previous research has recognised the potential complexity of multi-stakeholder involvement, highlighting the need to manage group interactions, potential power imbalances and synthesising the views of different groups 34. One approach which might have been useful in the context of our research and could be relevant to future work in this field, would be hold the separate stakeholder groups first to allow for independent discussion and feedback, followed by a combined group in which consolidated feedback may be compared and discussed.
By comparing different ways of involving end-users, we hope to provide useful consideration for future intervention development. However, our study is just one example; involving a small number of participants. There are many factors which have contributed to final intervention content. We cannot definitively assert that involving different types of end users together will yield different intervention content. The ideas incorporated into the final intervention were not solely influenced by the consensus process, as researchers held power to make these final decisions. Ideally, future studies, involving different interventions and subject matters, would explore and report their experiences with involving end users and how this may have influenced intervention content. This would build a clearer picture of the optimal way to involve different stakeholders in this process.
## Strengths and limitations
This study has several strengths including the use of a mixed methods, explanatory sequential design. Consensus meeting data supported the quantitative analysis by providing explanations, where available, for different participant ratings provided in the survey. By integrating the two, we aimed to draw out new findings beyond the information gained from the separate results 35. Fetters et al. have reported that such qualitative methods are often applied in order to explore reasons why a phenomenon occurs or to describe the nature of an individual's experience 17. The involvement of PPI contributors is a further strength of this research. A PPI partner (GF) was involved in the SWAT throughout the duration of the study and is a named co-author on this publication. A separate PPI group were involved in the development of the materials sent to participants prior to the consensus meetings. Supplementary files 3.2. and 3.3 in the *Extended data* 18 show how the study invitation letter and survey were improved as result of PPI feedback. These improvements helped to ensure that materials were more accessible and acceptable to participants.
This study includes a number of limitations. Firstly, as this was a SWAT, the primary aim of the consensus meetings was to review and discuss proposals for intervention content for the host trial and not to explicitly compare end-user contributions 15. While the semi-structured approach of the meetings allowed participants to discuss proposed intervention content and generate new ideas for such content, it made it difficult to compare end-user contributions as the content and nature of the discussions varied across meetings. For example, some groups did not discuss certain survey ratings and intervention proposals, and some groups discussed particular proposals in more detail than others. This meant that explanations for survey ratings are not present in qualitative form consistently for all intervention proposals. Adopting a more structured approach, for example the nominal group technique or Delphi method 36, during the consensus meetings may have made it easier to compare views on all proposals across groups. The consensus meetings were designed to be semi-structured to elicit participants views on what components may be acceptable and feasible for them. The semi-structured format did necessitate the research team deliberating after the meetings to consider consensus meeting feedback and decide what which components to incorporate into the intervention. During these meetings the research team discussed the feedback alongside other considerations, as mentioned: equity, side effects/safety, effectiveness. The challenges of combining different forms of evidence during the intervention development process has previous been acknowledged 15; that is, integrating stakeholder feedback, with theory and evidence of effectiveness. Although the decisions about intervention components in this study were shaped by the consensus meeting discussions, had we adopted a more structured approach, we recognise PWD and HCP could have engaged in a more deliberate dialogue around final intervention components.
An additional limitation is the absence of some key end-users from the consensus meetings. There were no people with type 2 diabetes available to participate in the combined meeting. Despite using a range of strategies to recruit a representative sample of people with diabetes, we encountered issues with participant availability when arranging the combined meeting. Existing research has established that people with type 1 and type 2 diabetes have different experiences when managing their condition and engaging with HCPs and health services 37– 39. Therefore, the involvement of people with type 2 diabetes in the combined meeting could have potentially changed the nature of the discussion and led to different recommendations. There was also a lack of involvement of practice administrators in the consensus meeting. Participants in the HCP only meeting suggested that practice administrators would be best placed to deliver the intervention. Involving them in the consensus meetings may have led to different recommendations as they play a key role in undertaking clerical duties to support delivery of care, and as gatekeepers, help to preserve boundaries of organisation and controlling access to the practice 40. However, the literature finds they are often overlooked by policymakers, undervalued by GPs and patients and excluded from research 40. Future research in general practice should consider involving practice administrators to ensure that all user voices are heard.
A final limitation was the lack of capture of non-verbal cues such as when participants nod in agreement or disagreement. As this SWAT looked to examine and compare agreement with proposed intervention content, such non-verbal data may have been useful. While non-verbal cues can offer rich data 41 and we may have been able to capture this through video recording of the meeting, it has also been found that the use of video-recording equipment during focus groups can inhibit participants’ interaction 42.
## Implications
The results of this SWAT informed the development of the IDEAs intervention which has been tested as part of a pilot cluster randomised trial with a view to progressing to a definitive trial 14. Involving end-users in decisions about planning and conducting health research, policy and services is gaining increasing momentum and as such, PPI is now required by many health research funders, journals, and research ethics committees 43, 44. However, evidence on the impact of PPI is largely based on anecdotal reflections from researchers and members of the public which are descriptive and selective 45. Numerous studies have called for planned and methodologically rigorous research to evaluate the impact of PPI on the research process 46– 48. In this study, people with diabetes were involved as participants in the consensus meetings and not throughout the design and conduct of the research as PPI contributors. However, their role discussing and making decisions about the intervention content and delivery is not dissimilar to the active role that PPI contributors have in the research process 49– 51. This SWAT provides evidence on the contribution of different end-users to the intervention development process and how different end-users can have different priorities for intervention content. While our study provides useful reflections for future intervention development using consensus processes, results should be interpreted with caution given this is just one example of involving stakeholders, and other factors may have influenced the final intervention content.
Nevertheless, the results of this study, coupled with the results of our analysis of participants’ experiences of taking part in the three separate meetings to inform intervention development 13, suggest that it may potentially be more acceptable and useful to involve patients/members of the public and HCPs separately when conducting PPI activities. When involving stakeholders together in PPI activities, alternative approaches to facilitation may need to be considered. Furthermore, as the process and impact of PPI is heavily dependent on the context in which it is being conducted, further research exploring the experiences and contributions of different end-users is needed, including an exploration of different facilitation models. This would enable all individuals interested in involving patients and members of the public in health research, policy, planning and development of health care to design and conduct more appropriate and effective user involvement 8, 52.
## Conclusion
UK Medical Research Council guidance on the development and evaluation of complex interventions states that interventions should be developed with user involvement, drawing on existing evidence and appropriate theory 1. However, there is limited evidence on what different intervention users contribute to the intervention development process and whether their contributions differ according to group composition. Our findings show that preferences and priorities for intervention content can differ across end-user groups, and that suggestions and recommendations for intervention content and design may also vary depending on whether users are involved simultaneously or separately. Considering these findings, attention should be paid to how end-users are involved in intervention development processes. This will stand to help researchers to design and conduct more appropriate user involvement, which in turn, could potentially improve intervention fit with the end-user’s perceived needs.
## Underlying data
The consensus meeting data are not publicly available due to limitations based on the ethical approval received and participant consent. Participants of the consensus process were not asked for their consent to store their data in a public repository. Participants consented to their anonymised data being made available for further collaborative research purposes outside of the current study upon reasonable request from the corresponding author and provision of a written proposal to the Principal Investigator (Dr Sheena McHugh, [email protected]).
Open Science Framework: What and how do different stakeholders contribute to intervention development? A mixed methods study. https://doi.org/10.17605/OSF.IO/NJS9Y 18.
The project contains the following underlying data: Data are available under the terms of the Creative Commons CC0 1.0 Universal (CC0 1.0) Public Domain Dedication License.
## Extended data
Open Science Framework: What and how do different stakeholders contribute to intervention development? A mixed methods study. https://doi.org/10.17605/OSF.IO/NJS9Y 18.
This project contains the following extended data: Data are available under the terms of the Creative Commons CC0 1.0 Universal (CC0 1.0) Public Domain Dedication License.
## References
1. **MRC Developing and evaluating complex interventions.**. (2006) 1-39
2. Wight D, Wimbush E, Jepson R. **Six steps in quality intervention development (6SQuID).**. (2016) **70** 520-5. DOI: 10.1136/jech-2015-205952
3. Corbett T, Singh K, Payne L. **Understanding acceptability of and engagement with Web-based interventions aiming to improve quality of life in cancer survivors: A synthesis of current research.**. (2018) **27** 22-33. DOI: 10.1002/pon.4566
4. Wensing M, Huntink E, van Lieshout J. **Tailored implementation of evidence-based practice for patients with chronic diseases.**. (2014) **9** e101981. DOI: 10.1371/journal.pone.0101981
5. Huntink E, van Lieshout J, Aakhus E. **Stakeholders’ contributions to tailored implementation programs: an observational study of group interview methods.**. (2014) **9** 185. DOI: 10.1186/s13012-014-0185-x
6. Droog E, Foley C, Healy O. **Perspectives on the underlying drivers of urgent and emergency care reconfiguration in Ireland.**. (2018) **33** 364-79. DOI: 10.1002/hpm.2469
7. Spence H, Baker K, Wharton-Smith A. **Childhood pneumonia diagnostics: Community health workers’ and national stakeholders’ differing perspectives of new and existing aids.**. (2017) **10** 1290340. DOI: 10.1080/16549716.2017.1290340
8. Morton KL, Atkin AJ, Corder K. **Engaging stakeholders and target groups in prioritising a public health intervention: the Creating Active School Environments (CASE) online Delphi study.**. (2017) **7** e013340. DOI: 10.1136/bmjopen-2016-013340
9. Smithson J. **Using and analysing focus groups: Limitations and possibilities.**. (2000) **3** 103-19. DOI: 10.1080/136455700405172
10. O’Hara MC, Hynes L, O’Donnell M. **Strength in Numbers: an international consensus conference to develop a novel approach to care delivery for young adults with type 1 diabetes, the**. (2017) **3** 25. DOI: 10.1186/s40900-017-0076-9
11. Owens C, Farrand P, Darvill R. **Involving service users in intervention design: a participatory approach to developing a text-messaging intervention to reduce repetition of self-harm.**. (2011) **14** 285-95. DOI: 10.1111/j.1369-7625.2010.00623.x
12. Lowes L, Robling MR, Bennert K. **Involving lay and professional stakeholders in the development of a research intervention for the DEPICTED study.**. (2011) **14** 250-60. DOI: 10.1111/j.1369-7625.2010.00625.x
13. Racine E, Riordan F, Phillip E. **'It just wasn't going to be heard': A mixed methods study to compare different ways of involving people with diabetes and health-care professionals in health intervention research.**. (2020) **23** 870-83. DOI: 10.1111/hex.13061
14. Riordan F, Racine E, Smith SM. **Feasibility of an implementation intervention to increase attendance at diabetic retinopathy screening: protocol for a cluster randomised pilot trial.**. (2020) **6** 64. DOI: 10.1186/s40814-020-00608-y
15. Riordan F, Racine E, Phillip ET. **Development of an intervention to facilitate implementation and uptake of diabetic retinopathy screening.**. (2020) **15** 34. DOI: 10.1186/s13012-020-00982-4
16. Doyle L, Brady AM, Byrne G. **An overview of mixed methods research.**. (2009) **14** 175-85. DOI: 10.1177/1744987108093962
17. Fetters MD, Curry LA, Creswell JW. **Achieving integration in mixed methods designs-principles and practices.**. (2013) **48** 2134-56. DOI: 10.1111/1475-6773.12117
18. Mahony LO, Racine E, Flynn G. (2022). DOI: 10.17605/OSF.IO/NJS9Y
19. Michie S, Atkins L, West R. **The Behaviour Change Wheel: A Guide to Designing Interventions**. (2014)
20. Kolehmainen N, Francis JJ. **Specifying content and mechanisms of change in interventions to change professionals’ practice: an illustration from the Good Goals study in occupational therapy.**. (2012) **7** 100. DOI: 10.1186/1748-5908-7-100
21. Lawrenson JG, Graham-Rowe E, Lorencatto F. **What works to increase attendance for diabetic retinopathy screening? An evidence synthesis and economic analysis.**. (2018) **22** 1-160. DOI: 10.3310/hta22290
22. Lawrenson JG, Graham-Rowe E, Lorencatto F. **Interventions to increase attendance for diabetic retinopathy screening.**. (2018) **1** CD012054. DOI: 10.1002/14651858.CD012054.pub2
23. Weiner BJ, Lewis CC, Stanick C. **Psychometric assessment of three newly developed implementation outcome measures.**. (2017) **12** 108. DOI: 10.1186/s13012-017-0635-3
24. Jung SH. **Stratified Fisher’s exact test and its sample size calculation.**. (2014) **56** 129-40. DOI: 10.1002/bimj.201300048
25. Hsieh HF, Shannon SE. **Three approaches to qualitative content analysis.**. (2005) **15** 1277-88. DOI: 10.1177/1049732305276687
26. Michie S, Richardson M, Johnston M. **The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions.**. (2013) **46** 81-95. DOI: 10.1007/s12160-013-9486-6
27. Hall JF, Crocker TF, Clarke DJ. **Supporting carers of stroke survivors to reduce carer burden: Development of the Preparing is Caring intervention using Intervention Mapping.**. (2019) **19** 1408. DOI: 10.1186/s12889-019-7615-2
28. Crosbie B, O’Callaghan ME, O’Flanagan S. **A real-time measurement of general practice workload in the Republic of Ireland: a prospective study.**. (2020) **70** e489-96. DOI: 10.3399/bjgp20X710429
29. O’Dea B, O’Connor P, Lydon S. **Prevalence of burnout among Irish general practitioners: a cross-sectional study.**. (2017) **186** 447-53. DOI: 10.1007/s11845-016-1407-9
30. **National Framework for the Integrated Prevention and Management of Chronic Disease in Ireland 2020-2025**. (2020)
31. Ruiter RA, Kessels LT, Peters GJ. **Sixty years of fear appeal research: current state of the evidence.**. (2014) **49** 63-70. DOI: 10.1002/ijop.12042
32. Emanuel AS, Kiviniemi MT, Howell JL. **Avoiding cancer risk information.**. (2015) **147** 113-20. DOI: 10.1016/j.socscimed.2015.10.058
33. Peters GJ, Ruiter RA, Kok G. **Threatening communication: A qualitative study of fear appeal effectiveness beliefs among intervention developers, policymakers, politicians, scientists, and advertising professionals.**. (2014) **49** 71-9. DOI: 10.1002/ijop.12000
34. Concannon TW, Grant S, Welch V. **Practical Guidance for Involving Stakeholders in Health Research.**. (2019) **34** 458-63. DOI: 10.1007/s11606-018-4738-6
35. Guetterman TC, Fetters MD, Creswell JW. **Integrating Quantitative and Qualitative Results in Health Science Mixed Methods Research Through Joint Displays.**. (2015) **13** 554-61. DOI: 10.1370/afm.1865
36. Rankin NM, McGregor D, Butow PN. **Adapting the nominal group technique for priority setting of evidence-practice gaps in implementation science.**. (2016) **16** 110. DOI: 10.1186/s12874-016-0210-7
37. Vincze G, Barner JC, Lopez D. **Factors associated with adherence to self-monitoring of blood glucose among persons with diabetes.**. (2004) **30** 112-25. DOI: 10.1177/014572170403000119
38. Hortensius J, Kars MC, Wierenga WS. **Perspectives of patients with type 1 or insulin-treated type 2 diabetes on self-monitoring of blood glucose: a qualitative study.**. (2012) **12** 167. DOI: 10.1186/1471-2458-12-167
39. Bolaños E, Sarría-Santamera A. **[Perspective of patients on type-2 diabetes and their relationship with primary care health professionals: a qualitative study].**. (2003) **32** 195-200. DOI: 10.1016/s0212-6567(03)79251-8
40. Litchfield I, Gale N, Burrows M. **The future role of receptionists in primary care.**. (2017) **67** 523-4. DOI: 10.3399/bjgp17X693401
41. Denham MA, Onwuegbuzie AJ. **Beyond words: Using nonverbal communication data in research to enhance thick description and interpretation.**. (2013) **12** 670-96. DOI: 10.1177/160940691301200137
42. Powell RA, Single HM. **Focus Groups.**. (1996) **8** 499-504. DOI: 10.1093/intqhc/8.5.499
43. Dyer S. **Rationalising public participation in the health service: the case of research ethics committees.**. (2004) **10** 339-48. DOI: 10.1016/j.healthplace.2004.08.004
44. Greenhalgh T, Hinton L, Finlay T. **Frameworks for supporting patient and public involvement in research: Systematic review and co-design pilot.**. (2019) **22** 785-801. DOI: 10.1111/hex.12888
45. Staley K, Buckland SA, Hayes H. **'The missing links': understanding how context and mechanism influence the impact of public involvement in research.**. (2014) **17** 755-64. DOI: 10.1111/hex.12017
46. Minogue V, Boness J, Brown A. **The impact of service user involvement in research.**. (2005) **18** 103-12. DOI: 10.1108/09526860510588133
47. Brett J, Staniszewska S, Mockford C. **A systematic review of the impact of patient and public involvement on service users, researchers and communities.**. (2014) **7** 387-95. DOI: 10.1007/s40271-014-0065-0
48. Boivin A, Richards T, Forsythe L. **Evaluating patient and public involvement in research.**. (2018) **363** k5147. DOI: 10.1136/bmj.k5147
49. Gallivan J, Kovacs Burns K, Bellows M. **The many faces of patient engagement.**. (2012) **4** e32
50. Arnstein SR. **A ladder of citizen participation.**. (1969) **35** 216-24
51. Staniszewska S, Brett J, Simera I. **GRIPP2 reporting checklists: Tools to improve reporting of patient and public involvement in research.**. (2017) **3** 13. DOI: 10.1186/s40900-017-0062-2
52. Owens C, Ley A, Aitken P. **Do different stakeholder groups share mental health research priorities? A four‐arm Delphi study.**. (2008) **11** 418-13. DOI: 10.1111/j.1369-7625.2008.00492.x
53. Racine E. **‘It’s a nice thing to do but…’: exploring the methods and impact of patient and public involvement (PPI) in trials.**. (2020)
54. **The Participation Ladder: A Consumer/Survivor Lens**. (2020)
|
---
title: Relationships of BMI, muscle-to-fat ratio, and handgrip strength-to-BMI ratio
to physical fitness in Spanish children and adolescents
authors:
- Samuel Manzano-Carrasco
- Jorge Garcia-Unanue
- Eero A. Haapala
- Jose Luis Felipe
- Leonor Gallardo
- Jorge Lopez-Fernandez
journal: European Journal of Pediatrics
year: 2023
pmcid: PMC9989582
doi: 10.1007/s00431-023-04887-4
license: CC BY 4.0
---
# Relationships of BMI, muscle-to-fat ratio, and handgrip strength-to-BMI ratio to physical fitness in Spanish children and adolescents
## Abstract
This study aimed to determine the relationship of body mass index (BMI), muscle-to-fat ratio (MFR), and handgrip strength-to-BMI ratio to physical fitness parameters in an active young population according to sex across four different time points. A total of 2256 Spanish children and adolescents (aged 5–18) from rural areas participating in an extracurricular sport in different municipal sports schools participated in this study. Participants were divided into children (5–10 years) and adolescents (11–18 years), boys and girls, and across four different time points [2018, 2019, 2020, 2021]. Data on anthropometric measures (BMI, MFR, appendicular skeletal muscle mass) and physical fitness (handgrip strength, cardiorespiratory fitness, and vertical jump) were collected. Boys who were overweight, but especially boys with obesity, had higher absolute handgrip strength in children and adolescents than their normal weight counterparts in 2020 and 2021. Boys and girls with normal weight presented higher cardiorespiratory fitness and vertical jump than their overweight and obese peers over the years. The MFR was directly correlated with the cardiorespiratory fitness and vertical jump variables, but not with handgrip strength, in boys and girls. The handgrip strength-to-BMI ratio in both sexes was positively correlated to the different physical fitness parameters.
Conclusion: BMI, MFR, and handgrip strength-to-BMI can be used as health and physical fitness indicators in this population. What is Known:• BMI is the main indicator commonly used as a proxy for obesity for many years. Nevertheless, it cannot differentiate between fat mass and fat-free mass.• There are other indicators such as MFR and handgrip strength-to-BMI that might be more accurate and can serve to monitor the health and fitness of children and adolescents. What is New:• MFR showed a positive and significant correlation with cardiorespiratory fitness and vertical jump in both sexes. On the other hand, the handgrip strength-to-BMI presented a positive correlation with cardiorespiratory fitness, vertical jump, and handgrip strength.• The use of these indicators obtained through different parameters of body composition and physical fitness can serve as a tool to identify the relationships of the paediatric population with physical fitness.
## Introduction
Scientific evidence suggests a progressive increase in obesity and a decrease in physical fitness among children and adolescents worldwide [1, 2]. Excessive time spent on sedentary behaviours and, in particular, sedentary technology use, unhealthy dietary composition, poor physical fitness, and insufficient sleep are the main factors responsible for these public health issues [3–5]. These unhealthy behaviours have gotten worsen due to the restrictions involving SARS-CoV-2 [6]. In this sense, it is necessary to address a greater capacity in policy and adherence towards daily physical activity practice, making visible an existing problem in society and to raise awareness of physical inactivity through better surveillance and monitoring of different parameters.
The Global Action Plan on Physical Activity 2018–2030 (Action 4.2.) [ 7] and the United Nations’ Sustainable Development Goals [8] suggest monitoring and surveillance of fitness and health to understand the effectiveness of current policies and guide future actions to enhance healthy behaviours among children and adolescents. However, contrary to physical activity surveillance, which is implemented in all countries of the European Union, the monitoring of physical fitness and body composition of children and adolescents is not as extended. Bulgaria, Finland, Portugal, and Slovenia are the main promoters of this type of surveillance [9]. The implementation of field-based fitness test batteries or protocols offers an opportunity to track and record physical fitness parameters such as cardiorespiratory, musculoskeletal, and body composition [10]. But the standardisation of this data into benchmarking health indicators is not that easy. The main indicator commonly used as a proxy for obesity for many years has been the body mass index (BMI) [11]. Nonetheless, it is a relatively poor proxy of body composition in childhood [12] and has been questioned due to its limitations in detecting adiposity in the young population [13]. In addition, it cannot differentiate between fat mass and fat-free mass [14], which may have different effects on health outcomes [15], and it does not inform about the current physical fitness of children and adolescents. Thus, it is important to provide new evidence on other indicators related to body composition and physical fitness that may be more accurate in monitoring and surveillance of health and fitness in children and adolescents.
In recent years, the waist-to-height ratio (WHtR), which combines waist circumference and height, has been used for detecting abdominal obesity [16]. Furthermore, authors have suggested the use of other indicators used less commonly, as they seem to address several of the limitations of BMI, such as the muscle-fat-ratio (MFR) or the handgrip strength-to-BMI ratio. The MFR studies the relationship between skeletal muscle mass and total body fat mass [17, 18]. This indicator relies on precise measurements of body composition and it has been considered the main indicator of low muscle mass [19]. On the other hand, the handgrip strength-to-BMI ratio is calculated as the handgrip strength test result divided by BMI [20, 21]. Due to the fact that handgrip strength can be measured quickly and easily in field-based testing together with BMI, the handgrip strength-to-BMI ratio, which takes into account body composition and fitness parameters, could help to determine the state as well as the evolution of the young population. Additionally, these markers have been correlated with metabolic risk [17, 22], central adiposity [23] in children and adolescents, and arterial hypertension and type 2 diabetes in adults [23–25]. Nonetheless, to the best of our knowledge, the effectiveness of these indicators in relating to body composition and physical fitness among active or partially active children and adolescents is yet to be explored. Therefore, we investigated the relationship of BMI, MFR, and handgrip strength-to-BMI ratio to physical fitness parameters in an active young Spanish population according to sex across four consecutive years.
## Participants
Baseline data of the Active Health project [26, 27] collected from May 2018 to December 2021 were analysed in this cross-sectional study (Fig. 1). All participants were participating in an extracurricular sport activity at least 2 days a week for a minimum of 1 h each day from different municipal sports schools in Castilla-La Mancha (a central and rural region of Spain). Although in this project all participants enrolled in sports schools are invited to participate, only those who completed all the tests were taken into account in the analysis. A total convenience sample of 2256 children and adolescents aged 5 to 18 years old (11.0 ± 2.7 years; 43.6 ± 15.1 kg; 146.4 ± 16.1 cm) participated in this study. The final sample of the study was formed by 1558 boys ($69\%$ of the study population) and 698 girls ($31\%$ of the study population). The sample was divided based on sex (boys and girls), age range (children 5–10 years and adolescents 11–18 years, according to other studies) [17], and academic year (2018, 2019, 2020, and 2021). The main exclusion criterion was the presence of physical disability or any health problem which might influence the performance in the fitness tests. Participants’ parents were informed about the aim and nature of the test in the study and written informed consents was obtained. Table 1 presents the descriptive data of the participants (anthropometric and physical fitness variables).Fig. 1Number of participants in the Active Health project over a four-yearTable 1General characteristics of the sample across 4 yearsVariables2018($$n = 293$$)2019($$n = 887$$)2020($$n = 501$$)2021($$n = 575$$)Age (years)12.6 (2.12)10.9 (2.65)10.9 (2.94)10.4 (2.58)Weight (kg)47.9 (14.07)42.8 (14.70)44.6 (16.76)41.6 (14.21)Height (cm)152.7 (13.43)145.8 (15.77)146.5 (17.64)144.2 (15.72)BMI (kg/m2)20.1 (3.53)19.6 (3.72)20.0 (3.94)19.5 (3.74)Fat mass (kg)11.5 (6.01)10.5 (5.75)11.2 (6.42)10.4 (5.79)Fat mass (%)23.2 (6.79)23.6 (6.62)24.2 (6.70)24.2 (6.85)Muscle mass (kg)34.4 (9.29)30.6 (9.97)31.7 (11.39)29.5 (9.53)Muscle mass (%)72.7 (6.41)72.3 (6.23)71.7 (6.31)71.8 (6.47)ASMM (kg)1.5 (0.58)1.4 (0.53)1.3 (0.49)1.3 (0.54)MFR (kg/kg)3.5 (1.25)3.5 (1.57)3.1 (1.13)3.2 (1.12)Grip strength-to-BMI (kg/kg/m2)1.3 (0.42)1.1 (0.40)1.0 (0.41)1.0 (0.40)Handgrip strength (kg)26.4 (9.20)20.9 (9.27)20.0 (9.49)20.1 (8.70)Handgrip strength (pc)64 (27.81)51 (29.03)46 (29.21)56 (28.43)20-mSRT (stages)6 (2.15)5 (2.32)5 (2.54)4.5 (2.25)20-mSRT (pc)68 (22.05)67 (24.52)62 (25.99)63 (25.06)VO2max (ml/kg/min)48.3 (4.97)48.9 (4.90)47.8 (5.37)47.6 (4.65)Vertical jump (cm)No data22.1 (6.48)23.3 (7.24)22.0 (6.45)Vertical jump (pc)No data47 (26.61)44 (25.14)45 (26.03)Data are presented as mean (SD)kg kilogrammes, cm centimetres, m metres, BMI body mass index, ASMM appendicular skeletal muscle mass, MFR muscle-fat-ratio, pc percentile, 20-mSRT 20-m Shuttle-Run Test, VO2max maximal oxygen uptake This research was carried out in compliance with the standards of the Declaration of Helsinki (2013 revision, Brazil) [28] and following the guidelines of the European Community for Good Clinical Practice ($\frac{111}{3976}$/88 July 1990) as well as the Spanish legal framework for clinical research on humans (Royal Decree $\frac{561}{1993}$ in clinical trials). The Active Health project was approved by the Bioethics Committee for Clinical Research of the Virgen de la Salud Hospital in Toledo and by the supervisors of the University of Castilla-La Mancha (Ref.: $\frac{508}{17042020}$).
## Assessments
Data collection took place in each of the participating sports schools before and during timetabled extracurricular sports activities on different days. Each test was set up at individual stations and the participants rotated between them in groups of 10–12 every hour, except of the cardiorespiratory fitness test which was done as a group. Each station was controlled by an experienced research. The established protocol of tests explained below:
## Anthropometric measurements
Each participant underwent an anthropometric assessment utilising a methodology at 5-min intervals, according to prior research [29]. For this assessment, a portable segmental analyser of multifrequency body composition (Tanita MC-780, Tanita Corp., Tokyo, Japan) was used to measure weight (kg), fat mass (kg and %), and muscle mass (kg and %). Height (cm) was assessed with a height rod (Seca 214, Hamburg, Germany). BMI was calculated with the weight (kg) divided by the squared height (m). The appendicular skeletal muscle mass (ASMM) was calculated by the sum of the muscle mass of four limbs, and muscle-to-fat ratio (MFR = ASMM [kg]/fat mass [kg]) was also calculated [30]. The evaluations were conducted while wearing clothing and without shoes.
## Physical fitness
An adapted version of the extended Assessing Levels of Physical Activity (ALPHA) health-related fitness battery for children and adolescents [10] was used to assess the different parameters of physical fitness. In accordance with earlier studies [31, 32], a percentile (pc) value based on age and sex was used to standardise the findings from all tests. All fitness tests were conducted by researchers, and the order in which they were carried out was as follows: A handgrip strength with hand dynamometer with adjustable grip was used to evaluate upper-body muscular strength (Constant R Model: 14192-709E). Participants in a full-extension elbow position were required to close their hands with a continuous maximum force for three seconds. The test was performed with the dominant hand and the non-dominant hand alternately. It was possible to try again with a 30-s rest period in between. Each participant’s best score from their dominant hand was taken into consideration for an analysis to the nearest 1 g, and the result was recorded in kilogrammes as absolute values [10]. Pc or relative values based on age and sex were used to standardise the test results [31, 32].
A vertical jump test was completed to assess lower-body muscular power. Height was recorded in centimetres and calculated to the nearest 0.1 cm by photoelectric cells. This technological equipment consists of two parallel bars (Optojump, Microgate, Bolzano, Italy) which measure flight time taken as the duration between take-off and landing. Participants were instructed to jump as high as possible, and three attempts were allowed with 30 s of recovery. The test results were used to standardise as a pc based on age and sex [31, 32].
Finally, cardiorespiratory fitness was assessed by performing a maximum incremental field test (20-m Shuttle-Run Test [20-mSRT]). Participants had to run between two lines 20 m apart while keeping a pace emitted by acoustic signals by a portable speakerphone. The initial speed is 8.5 km h−1, which is increased by 0.5 km h−1 each min [33]. The test ended when the participant failed to reach the end of the lines concurrent with the audio signals on two consecutive occasions. Otherwise, the test finished when the participant stopped because of fatigue. The results were transformed in stages of 1-min duration, and the maximal oxygen uptake (VO2max) was estimated using the formula by Leger et al. [ 33]: VO2max (ml·kg−1·min−1) = 31.025 + 3.248·X1 – 3.248·X2 + 0.1536·X1·X2, where the final speed is X1 (km·h−1) and age is X2 (year as the lower rounded integer). The test was performed only once, and it was performed last so that performance and fatigue did not interfere with the participants.
Lastly, the handgrip strength-to-BMI ratio was estimated with the handgrip strength (kg) and BMI (kg/m2).
## Statistical analysis
Data were presented as means ± standard deviations. A Kolmogorov–Smirnov test was used to confirm a normal distribution of the variables. Furthermore, categorical data are presented as absolute and relative frequencies. The dataset is balanced and does not present missing values (except for vertical jump in 2018 data, because this parameter was not evaluated that year). The sample was divided based on sex (boys and girls); age range (children 5–10 years and adolescents 11–18 years, according to other studies) [17]; and year of the analysis (2018, 2019, 2020, and 2021). Differences in physical fitness parameters (i.e. dependent variables) between weight status–based World Health Organization (WHO) BMI-for-age reference (normal weight, overweight, and obese, i.e. independent variable) were evaluated by one-way ANOVA for independent samples due to the categorical format of the factor. In order to evaluate the relationship between physical fitness parameters and anthropometric status based on MFR and handgrip strength-to-BMI, Pearson’s product moment correlation was used due to the scale format of the factor. The level of significance was set at $p \leq 0.05.$
## Differences in physical fitness according to BMI
Tables 2 and 3 show the differences between the three weight status–based WHO BMI-for-age references (normal weight, overweight, and obese) in physical fitness parameters separated by sex (boys and girls) and age group. Boys and girls showed significant differences in all physical fitness parameters in the four analysed time points. Table 2Differences in boys on physical fitness parameters regard weight status–based World Health Organization BMI-for-age referenceBoys2018201920202021Normal weightOverweightObeseNormal weightOverweightObeseNormal weightOverweightObeseNormal weightOverweightObeseHandgrip strength (kg)5–1017.3 (0.99)18.0 (1.22)19.6 (1.51)14.6 (0.28) ¥15.7 (0.38)16.7 (0.42)13.6 (0.38) ¥13.4 (0.57)15.3 (0.52)14.1 (0.48) Ŷ ¥17.4 (0.69)16.3 (0.60)11–1828.6 (8.22) Ŷ32.2 (9.58)31.2 (10.23)25.6 (0.46) Ŷ ¥29.6 (0.63)29.3 (0.87)26.9 (0.74)28.8 (1.08)30.3 (1.36)26.7 (0.58)27.0 (0.96)27.98 (1.21)Handgrip strength (pc)5–1058 (6.14)68 (7.50)80 (9.28)47 (57.41) Ŷ ¥57 (3.24)63 (3.52)37 (3.16) ¥39 (4.70)55 (4.31)45 (3.07) Ŷ ¥68 (4.45)63 (3.88)11–1858 (2.55) Ŷ73 (3.99)71 (5.32)43 (1.93) Ŷ ¥60 (2.64)58 (3.68)44 (2.73) ¥53 (3.97)61 (5.01)56 (2.53)61 (4.20)60 (5.34)VO2max (ml/kg/min)5–1052.0 (0.82) ¥49.5 (0.99)46.7 (1.27)52.4 (0.31) Ŷ ¥49.9 (0.43) ƛ46.9 (0.47)51.5 (0.36) Ŷ ¥49.4 (0.54) ƛ46.4 (0.49)50.6 (0.42) ¥49.3 (0.53) ƛ47.0 (0.53)11–1850.6 (0.39) Ŷ ¥48.2 (0.62) ƛ43.2 (0.85)50.9 (0.31) Ŷ ¥48.1 (0.43) ƛ43.5 (0.60)51.1 (0.51) Ŷ ¥46.9 (0.74) ƛ42.2 (0.93)49.7 (0.40) Ŷ ¥46.9 (0.66) ƛ42.2 (0.85)20-mSRT (pc)5–1085 (4.48) ¥72 (5.39)56 (6.96)80 (1.76) Ŷ ¥70 (2.46) ƛ50 (2.70)75 (2.23) Ŷ ¥65 (3.29) ƛ46 (3.02)73 (2.50) ¥65 (3.61) ƛ50 (3.18)11–1870 (1.74) Ŷ ¥61 (2.74) ƛ38 (3.76)69 (1.52) Ŷ ¥56 (2.09) ƛ33 (2.93)70 (2.26) Ŷ ¥50 (3.31) ƛ31 (4.14)65 (1.99) Ŷ ¥49 (3.31) ƛ26 (4.24)Vertical jump (cm)5–10No data available21.0 (0.44) Ŷ ¥18.9 (0.60) ƛ15.9 (0.64)21.5 (0.39) ¥20.1 (0.59) ƛ17.2 (0.54)21.1 (0.43) ¥19.8 (0.63) ƛ17.1 (0.55)11–1828.2 (0.57) ¥27.1 (0.72) ƛ22.0 (0.91)30.0 (0.48) ¥28.6 (0.70) ƛ25.7 (0.88)28.3 (0.48) Ŷ ¥24.5 (0.80) ƛ21.3 (1.02)Vertical jump (pc)5–10No data available62 (2.68) Ŷ ¥50 (3.64) ƛ35 (3.95)62 (2.58) ¥56 (3.83) ƛ35 (3.51)61 (2.79) ¥50 (4.40) ƛ37(3.53)11–1841 (2.47) ¥37. ( 3.13) ƛ19 (3.94)40 (2.00) ¥34 (2.91)25 (3.67)44 (2.05) Ŷ ¥30 (3.40)19 (4.32)Data are presented as mean (SD). Ŷ significant differences ($p \leq 0.05$) between normal weight and overweight. ¥ significant differences ($p \leq 0.05$) between normal weight and obesity. ƛ significant differences ($p \leq 0.05$) between overweight and obesity. An one-way ANOVA for independent samples was used as statistical analysiskg kilogrammes, cm centimetres, pc percentile, 20-mSRT 20-m Shuttle-Run Test, VO2max maximal oxygen uptakeTable 3Differences in girls on physical fitness parameters regard weight status–based World Health Organization BMI-for-age referenceGirls2018201920202021Normal weightOverweightObeseNormal weightOverweightObeseNormal weightOverweightObeseNormal weightOverweightObeseHandgrip strength (kg)5–1015.2 (0.78)13.7 (1.29)No data13.4 (0.35)14.3 (5.28)15.2 (0.72)12.9 (0.43)14.4 (0.64)12.5 (0.80)12.6 (0.38) Ŷ14.6 (0.61) Ŷ13.4 (0.63)11–1824.9 (0.81)25.0 (1.30)28.7 (1.83)21.3 (0.50) Ŷ24.0 (0.71)24.0 (1.21)21.1 (0.65)22.7 (1.07)12.5 (1.25)23.4 (0.59)24.0 (0.86)17.3 (1.54)Handgrip strength (pc)5–1061 (7.17)43 (11.93)No data51 (3.49)62 (5.23)67 (7.16)44 (4.47)63 (6.68)49 (8.69)49 (3.43)59 (5.51)57 (5.71)11–1866 (4.28)64 (6.87)81 (9.68)42 (61.80) Ŷ62 (4.69)59 (7.94)38 (4.01)50 (6.55)41 (7.69)62 (2.96)66 (4.34)80 (7.52)VO2max (ml/kg/min)5–1046.9 (0.63)46.1 (1.05)No data49.7 (0.35) Ŷ ¥47.6 (0.53)45.8 (0.71)48.1 (0.38) ¥46.6 (0.57)45.3 (0.74)48.3 (0.31) ¥47.1 (0.50)46.3 (0.52)11–1845.9 (0.66) Ŷ ¥42.2 (1.07)39.8 (1.51)46.0 (0.42) Ŷ ¥43.9 (0.63)40.0 (1.11)44.2 (0.62) ¥41.3 (1.02)38.7 (1.20)45.6 (0.54) Ŷ ¥42.3 (0.78)39.7 (1.34)20-mSRT (pc)5–1083 (4.65)67 (7.74)No data83 (2.12) Ŷ ¥73 (3.22)58 (4.34)78 (3.16) Ŷ ¥64 (4.68)50 (6.09)74 (2.76) ¥65 (4.46)60 (4.63)11–1884 (3.19) Ŷ ¥65 (5.13)44 (7.22)75 (2.55) ¥65 (3.82)37 (6.76)67 (3.94) ¥52 (6.44)34 (7.57)72 (2.72) Ŷ ¥54 (3.92)32 (6.76)Vertical jump (cm)5–10No data available20.1 (0.46) Ŷ ¥18.0 (0.64)15.9 (0.89)18.7 (0.56) ¥18.1 (0.34)15.0 (1.09)18.8 (0.44) ¥17.5 (0.71)16.3 (0.74)11–1826.6 (0.67) Ŷ ¥21.9 (0.96)19.1 (1.57)24.6 (0.74) ¥22.4 (1.19)19.7 (1.40)24.3 (0.56) Ŷ ¥21.2 (0.80)19.3 (1.38)Vertical jump (pc)5–10No data available64 (3.22) Ŷ ¥50 (4.45)38 (6.22)53 (3.56) ¥48 (5.33)29 (6.93)56 (3.10)48 (5.01)42 (5.20)11–1856 (2.92) Ŷ ¥35 (4.18)23 (6.80)44 (3.56) ¥33 (5.75)24 (6.75)48 (2.80) Ŷ ¥33 (4.03)25 (6.95)Data are presented as mean (SD). Ŷ significant differences ($p \leq 0.05$) between normal weight and overweight. ¥ significant differences ($p \leq 0.05$) between normal weight and obesity. ƛ significant differences ($p \leq 0.05$) between overweight and obesity. An one-way ANOVA for independent samples was used as statistical analysiskg kilogrammes, cm centimetres, pc percentile, 20-mSRT 20-m Shuttle-Run Test, VO2max maximal oxygen uptake
## Handgrip strength
In the younger group, boys with overweight showed a higher pc than boys with normal weight in 2019 ($$p \leq 0.037$$; ES: 0.24 CI: 0.42 to 19.69) and 2021 ($p \leq 0.001$; ES: 5.65 to 5.83). Furthermore, boys with obesity had higher handgrip strength (kg and pc) in 2020 ($p \leq 0.01$; ES: 3.73 to 4.79) and 2021 ($p \leq 0.01$; ES: 4.15 to 5.13) than boys with normal weight. Finally, in 2021, girls with overweight had higher handgrip strength (kg) than girls with normal weight ($$p \leq 0.018$$; ES: 3.99; CI: 0.26 to 3.79).
On the other hand, in the older group, boys with overweight showed higher handgrip strength (kg and pc) than the boys with normal weight in 2018 ($p \leq 0.01$; ES: 0.39 to 4.36) and 2019 ($p \leq 0.001$; ES: 7.30 to 7.40). Similarly, boys with obesity had a greater pc than boys with normal weight in 2020 ($$p \leq 0.009$$; ES: 4.26; CI: 3.37 to 31.05). In girls, girls with overweight showed higher handgrip strength (kg and pc) than girls with normal weight in 2019 ($$p \leq 0.01$$; ES: 0.44 to 4.40).
Finally, boys with obesity had better handgrip strength (kg and pc) than boys with normal weight in both age groups in 2019 ($p \leq 0.001$; ES: 0.39 to 6.07).
## Cardiorespiratory fitness
In the younger group, boys with normal weight presented higher VO2max and pc compared to those with overweight in 2018 ($p \leq 0.01$; ES: 4.15 to 4.57) and obesity in 2021 ($p \leq 0.001$; ES: 7.57 to 8.14). Likewise, girls with normal weight showed higher VO2max and pc in 2019 ($p \leq 0.001$; ES: 7.00 to 7.33) and 2021 ($p \leq 0.05$; ES: 3.75 to 4.55) than girls with obesity. In 2020, girls with normal weight had a higher pc than girls who were overweight ($$p \leq 0.036$$; ES: 3.64; CI: 0.73 to 28.35).
In the older group, boys with overweight showed a higher pc than boys with obesity in 2018 ($p \leq 0.001$; ES: 6.77 to 7.00). Similarly, boys with normal weight had a greater VO2max and pc in 2021 than boys with overweight ($p \leq 0.001$; ES: 5.13 to 5.74) and obesity ($p \leq 0.001$; ES: 11.32 to 11.81). In contrast, girls with normal weight had higher VO2max and pc than girls who presented overweight in 2018 ($p \leq 0.01$; ES: 4.25 to 4.43) and 2021 ($p \leq 0.001$; ES: 4.91 to 5.73) and obesity in 2018 ($p \leq 0.001$; ES: 5.32 to 7.14) and 2021 ($p \leq 0.001$; ES: 5.43 to 7.78). Moreover, in 2019, girls with normal weight had a higher pc than obese girls ($p \leq 0.001$; ES: 7.42; CI: 20.32 to 55.50) and higher VO2max than girls with overweight ($p \leq 0.05$; ES: 3.63 to 4.67).
Lastly, in both age groups, boys with normal weight showed a higher VO2max and pc than those who were overweight and obese in 2018 ($p \leq 0.001$; ES: 4.89 to 11.20), 2019 ($p \leq 0.01$; ES: 4.58 to 15.47), 2020 ($p \leq 0.05$; ES: 3.61 to 11.75), and 2021 ($p \leq 0.05$; ES: 4.33 to 6.20). Girls with the normal weight status had higher VO2max and pc than the girls with obesity in 2020 ($p \leq 0.001$; ES: 4.82 to 5.95). Finally, boys with overweight had higher VO2max and pc than boys with obesity in 2019 ($p \leq 0.001$; ES: 5.74 to 8.85) and 2020 ($p \leq 0.001$; ES: 5.30 to 5.90).
## Vertical jump
In the younger group, boys with normal weight showed higher vertical jump (cm and pc) than overweight ($p \leq 0.05$; ES: 3.78 to 3.83) and obese in 2019 ($p \leq 0.001$; ES: 1.99 to 9.06) and in 2021 in boys ($p \leq 0.001$; ES: 7.80 to 8.10) and girls ($$p \leq 0.11$$; ES: 4.13; CI: 0.44 to 4.59). In 2021, there were no significant differences in the pc of girls ($p \leq 0.05$).
In the older group, boys with normal weight had higher vertical jump (cm and pc) compared to those with obesity in 2019 ($p \leq 0.001$; ES: 6.67 to 8.24) and 2021 ($p \leq 0.001$; ES: 7.38 to 8.74). Also, in 2021, the normal weight group had higher vertical jump (cm and pc) than boys with overweight ($p \leq 0.01$; ES: 4.96 to 5.74), girls with overweight ($p \leq 0.01$; ES: 4.26 to 4.48), and girls with obesity ($p \leq 0.05$; ES: 4.23 to 4.78).
In both groups, boys with overweight showed positively significant differences (cm and pc) than those with obesity in 2019 ($p \leq 0.01$; ES: 3.89 to 6.26) and 2021($p \leq 0.01$; ES: 3.38 to 4.58), except for the pc in the older group ($$p \leq 0.138$$; ES: 2.83; CI: -2.23 to 24.30). In addition, boys with normal weight showed higher vertical jump (cm and pc) than those with obesity ($p \leq 0.01$; ES: 4.91 to 9.04). Similarly, girls with normal weight displayed higher vertical jump (cm and pc) than girls with overweight in 2019 ($p \leq 0.05$; ES: 3.78 to 5.72) and girls with obesity in 2019 ($p \leq 0.001$; ES: 5.32 to 6.27) and 2020 ($p \leq 0.05$; ES: 3.69 to 4.41). Finally, significant differences (cm and pc) were found between boys with overweight and obesity in both age groups ($p \leq 0.05$; ES: 3.74 to 5.78), except for the pc of the older group in 2020 ($$p \leq 0.172$$; ES: 2.69; CI: − 2.35 to 20.23).
## Relationship of MFR and handgrip strength-to-BMI ratio to physical fitness
Sex and age group correlations of anthropometric indicators and physical fitness parameters in the four different time points are presented in Table 4. In both groups of boys, MFR was significantly correlated with cardiorespiratory fitness ($r = 0.47$ to 0.57, $p \leq 0.001$) and vertical jump ($r = 0.33$ to 0.55, $p \leq 0.001$). Nevertheless, MFR was not significantly correlated with handgrip strength in the different years ($p \leq 0.05$). MFR was positively correlated with cardiorespiratory fitness and vertical jump in 2019, 2020, and 2021 ($r = 0.21$ to 0.64, $p \leq 0.001$) except in the younger group in 2018 ($p \leq 0.05$). Finally, in girls, MFR was significantly correlated with handgrip strength in the youngest group in 2018 ($r = 0.65$, $$p \leq 0.011$$) and in the oldest group in 2019 and 2021 (r = − 0.28 to − 0.23, $p \leq 0.05$). MFR had no significant correlation with cardiorespiratory fitness and vertical jump in the different years ($p \leq 0.05$).Table 4Relationship between physical fitness parameters and anthropometric status based on muscle-fat-ratio and handgrip strength-to-BMIBoysMuscle-fat-ratioGrip strength-to-BMI20182019202020212018201920202021Handgrip strength (kg)5–10−0.036 ($$p \leq 0.808$$)−0.078 ($$p \leq 0.210$$)0.046 ($$p \leq 0.560$$)−0.046 ($$p \leq 0.548$$)0.781 ($p \leq 0.001$)0.713 ($p \leq 0.001$)0.800 ($p \leq 0.001$)0.838 ($p \leq 0.001$)10–18−0.065 ($$p \leq 0.386$$)−0.066 ($$p \leq 0.204$$)−0.002 ($$p \leq 0.977$$)0.142 ($$p \leq 0.056$$)0.772 ($p \leq 0.001$)0.786 ($p \leq 0.001$)0.813 ($p \leq 0.001$)0.767 ($p \leq 0.001$)Handgrip strength (pc)5–10−0.180 ($$p \leq 0.227$$)−0.040 ($$p \leq 0.522$$)−0.005 ($$p \leq 0.951$$)−0.080 ($$p \leq 0.301$$)0.659 ($p \leq 0.001$)0.683 ($p \leq 0.001$)0.667 ($p \leq 0.001$)0.665 ($p \leq 0.001$)10–18−0.085 ($$p \leq 0.255$$)−0.065 ($$p \leq 0.207$$)0.016 ($$p \leq 0.834$$)0.130 ($$p \leq 0.079$$)0.658 ($p \leq 0.001$)0.711 ($p \leq 0.001$)0.714 ($p \leq 0.001$)0.670 ($p \leq 0.001$)VO2max (ml/kg/min)5–100.575 ($p \leq 0.001$)0.518 ($p \leq 0.001$)0.569 ($p \leq 0.001$)0.438 ($p \leq 0.001$)0.419 ($$p \leq 0.004$$)0.442 ($p \leq 0.001$)0.303 ($p \leq 0.001$)0.280 ($p \leq 0.001$)10–180.571 ($p \leq 0.001$)0.477 ($p \leq 0.001$)0.517 ($p \leq 0.001$)0.502 ($p \leq 0.001$)0.503 ($p \leq 0.001$)0.439 ($p \leq 0.001$)0.419 ($p \leq 0.001$)0.419 ($p \leq 0.001$)20-mSRT(pc)5–100.563 ($p \leq 0.001$)0.509 ($p \leq 0.001$)0.532 ($p \leq 0.001$)0.485 ($p \leq 0.001$)0.291 ($$p \leq 0.053$$)0.436 ($p \leq 0.001$)0.330 ($p \leq 0.001$)0.271 ($p \leq 0.001$)10–180.516 ($p \leq 0.001$)0.430 ($p \leq 0.001$)0.533 ($p \leq 0.001$)0.502 ($p \leq 0.001$)0.440 ($p \leq 0.001$)0.415 ($p \leq 0.001$)0.403 ($p \leq 0.001$)0.393 ($p \leq 0.001$)Vertical jump(cm)5–10No data available0.486 ($p \leq 0.001$)0.553 ($p \leq 0.001$)0.333 ($p \leq 0.001$)*No data* available0.487 ($p \leq 0.001$)0.376 ($p \leq 0.001$)0.347 ($p \leq 0.001$)10–180.524 ($p \leq 0.001$)0.448 ($p \leq 0.001$)0.487 ($p \leq 0.001$)0.448 ($p \leq 0.001$)0.452 ($p \leq 0.001$)0.528 ($p \leq 0.001$)Vertical jump(pc)5–100.479 ($p \leq 0.001$)0.531 ($p \leq 0.001$)0.319 ($p \leq 0.001$)0.440 ($p \leq 0.001$)0.329 ($p \leq 0.001$)0.316 ($p \leq 0.001$)10–180.458 ($p \leq 0.001$)0.371 ($p \leq 0.001$)0.411 ($p \leq 0.001$)0.407 ($p \leq 0.001$)0.359 ($p \leq 0.001$)0.427 ($p \leq 0.001$)GirlsMuscle-fat-ratioGrip strength-to-BMI20182019202020212018201920202021Handgrip strength (kg)5–100.437 ($$p \leq 0.118$$)−0.061 ($$p \leq 0.481$$)0.200 ($$p \leq 0.068$$)−0.149 ($$p \leq 0.103$$)0.860 ($p \leq 0.001$)0.736 ($p \leq 0.001$)0.841 ($p \leq 0.001$)0.726 ($p \leq 0.001$)10–18−0.002 ($$p \leq 0.987$$)−0.283 ($$p \leq 0.002$$)−0.171 ($$p \leq 0.153$$)−0.231 ($$p \leq 0.020$$)0.712 ($p \leq 0.001$)0.560 ($p \leq 0.0010.698$ ($p \leq 0.001$)0.670 ($p \leq 0.001$)Handgrip strength (pc)5–100.655 ($$p \leq 0.011$$)−0.075 ($$p \leq 0.383$$)−0.010 ($$p \leq 0.930$$)−0.138 ($$p \leq 0.131$$)0.756 ($$p \leq 0.002$$)0.666 ($p \leq 0.001$)0.664 ($p \leq 0.001$)0.555 ($p \leq 0.001$)10–18−0.043 ($$p \leq 0.772$$)−0.248 ($$p \leq 0.007$$)−0.116 ($$p \leq 0.337$$)−0.235 ($$p \leq 0.018$$)0.574 ($p \leq 0.001$)0.510 ($p \leq 0.001$)0.608 ($p \leq 0.001$)0.534 ($p \leq 0.001$)VO2max (ml/kg/min)5–100.045 ($$p \leq 0.879$$)0.235 ($$p \leq 0.006$$)0.488 ($p \leq 0.001$)0.286 ($$p \leq 0.001$$)−0.206 ($$p \leq 0.479$$)0.142 ($$p \leq 0.155$$)0.520 ($p \leq 0.001$)0.374 ($p \leq 0.001$)10–180.599 ($p \leq 0.001$)0.430 ($p \leq 0.001$)0.444 ($p \leq 0.001$)0.449 ($p \leq 0.001$)0.583 ($p \leq 0.001$)0.281 ($$p \leq 0.013$$)0.178 ($$p \leq 0.143$$)0.415 ($p \leq 0.001$)20-mSRT(pc)5–100.269 ($$p \leq 0.352$$)0.215 ($$p \leq 0.012$$)0.401 ($p \leq 0.001$)0.234 ($$p \leq 0.010$$)−0.004 ($$p \leq 0.990$$)0.181 ($$p \leq 0.071$$)0.443 ($p \leq 0.001$)0.296 ($$p \leq 0.001$$)10–180.622 ($p \leq 0.001$)0.387 ($p \leq 0.001$)0.461 ($p \leq 0.001$)0.418 ($p \leq 0.001$)0.452 ($p \leq 0.001$)0.218 ($$p \leq 0.013$$)0.136 ($$p \leq 0.265$$)0.372 ($p \leq 0.001$)Vertical jump(cm)5–10No data available0.393 ($p \leq 0.001$)0.439 ($p \leq 0.001$)0.292 ($$p \leq 0.001$$)*No data* available0.514 ($p \leq 0.001$)0.500 ($p \leq 0.001$)0.411 ($p \leq 0.001$)10–180.598 ($p \leq 0.001$)0.349 ($p \leq 0.001$)0.501 ($p \leq 0.001$)0.455 ($p \leq 0.001$)0.316 ($$p \leq 0.083$$)0.482 ($p \leq 0.001$)Vertical jump(pc)5–100.388 ($p \leq 0.001$)0.437 ($p \leq 0.001$)0.268 ($$p \leq 0.003$$)0.429 ($p \leq 0.001$)0.446 ($p \leq 0.001$)0.308 ($$p \leq 0.001$$)10–180.645 ($p \leq 0.001$)0.326 ($p \leq 0.001$)0.434 ($p \leq 0.001$)0.428 ($p \leq 0.001$)0.223 ($$p \leq 0.066$$)0.340 ($$p \leq 0.001$$)Values marked in bold are significant. A Pearson product moment correlation was used as statistical analysiskg kilogrammes, cm centimetres, pc percentile, 20-mSRT 20-m Shuttle-Run Test, VO2max maximal oxygen uptake On the other hand, in both groups of boys, the handgrip strength-to-BMI ratio was directly correlated with all physical fitness parameters ($r = 0.27$ to 0.83; $p \leq 0.001$), except with cardiorespiratory fitness (pc) in 2018 ($r = 0.29$, $$p \leq 0.053$$). In girls, the handgrip strength-to-BMI ratio was significantly correlated with overall physical fitness parameters in both age groups ($r = 0.21$ to 0.86, $p \leq 0.001$), except in cardiorespiratory fitness in the younger group in 2018 and in the oldest group in 2020 and in vertical jump in 2020 ($p \leq 0.05$).
## Discussion
This is the first study investigating the relationship of BMI, MFR, and handgrip strength-to-BMI ratio to muscular strength and cardiorespiratory fitness in physically active children and adolescents according to sex across four consecutive years (2018 to 2021). This study evidenced that weight status taken from BMI as well as MFR and handgrip strength-to-BMI ratio have a significant relationship with different fitness parameters and could be used as health indicators for this population. Our findings showed that children and adolescents with normal weight status, regardless of sex, had higher cardiorespiratory fitness and vertical jump than those who were overweight and obese. In contrast, overweight participants, particularly boys with obesity, displayed significantly higher handgrip strength than those with normal weight status. Moreover, both the MFR and handgrip strength-to-BMI show a significant correlation with cardiorespiratory fitness and vertical jump in both sexes, while handgrip strength-to-BMI also displays a positive correlation with handgrip strength. Even though previous research determined that body composition analyses can allow for the identification and diagnosis of weight status based on BMI, the data of the present study show that the use of other body composition measurements—the MFR and handgrip strength-to-BMI indicators—can serve as a tool for identifying relationships of the paediatric population with physical fitness.
## BMI (weight status) and physical fitness
A total of $57\%$ of participants in the present study had healthy weight status, while $26\%$ and $17\%$ of participants were overweight or obese, respectively. This shows that although the participants regularly practised a sporting activity, approximately $40\%$ of participants had a high BMI value. A recent study investigated the prevalence and incidence of overweight and obesity rates in children and adolescents across eight Spanish regions, suggesting that childhood obesity prevalence and incidence rates vary by region in Spain [34]. In this study, the incidence of obesity in a rural young population descriptively increases over four consecutive years studied, with 2020 being the year where $20\%$ of the total sample were obese. This may be due to the period of confinement caused by COVID-19 as well as their decrease in physical activity and possible worse eating habits during this period according to other studies [35].
An overweight and obese status can lead to a higher prevalence of suffering from metabolic syndrome compared with normal weight status in children and adolescents [36]. In addition, this may have an impact on physical fitness, which has been seen as an important marker of health [37]. Our findings indicate that the association between BMI based on weight status and physical fitness is significant, which extends previous findings in the young population [27, 38, 39]. Furthermore, the association of higher muscular strength with an overweight and obesity status was more pronounced especially in boys. These findings are in accordance with the study by Fernandez et al. [ 40], who have consistently reported that greater handgrip strength is strongly associated with a high BMI. This may be because boys with higher fat mass have more handgrip strength than girls despite the fact that there are no differences between the different weight status groups. Additionally, this relationship shows to be stable over the 4 years. Instead, a normal weight status was significantly associated with higher performance in cardiorespiratory fitness and vertical jump. The significance of cardiorespiratory fitness levels for cardiovascular health in the young population has been clearly demonstrated [41, 42]. Nevertheless, some research did not account for an important factor such as weight status in this association [43]. Thus, despite a positive relationship in handgrip strength with a high BMI as well as better cardiorespiratory fitness and muscular strength with an optimal BMI, it is important to consider data about body composition parameters when examining relationships of physical fitness and health outcomes.
## MFR, handgrip strength-to-BMI, and physical fitness
BMI remains one of the most widely used measures of adiposity and weight status in the young population [11, 44]. However, this indicator does not discriminate between fat mass and fat-free mass [14] and also does not reflect fat distribution and accumulation [45]. Accordingly, the MFR and handgrip strength-to-BMI ratio might be more reliable by addressing several of the limitations of BMI. In the elderly population, the handgrip strength-to-BMI ratio has been suggested for diagnosing sarcopenia [46]. In addition, the MFR may be a potential indicator for type 2 diabetes, metabolic syndrome, and hypertension in adults [47–49]. In youth, although the evidence is scare, Steffl et al. have suggested that the handgrip strength-to-BMI ratio can be used to identify children who are at risk of sarcopenic obesity [21]. Similarly, preschool children demonstrated that a greater handgrip strength-to-BMI ratio was associated with lower fat mass and percentage of body fat [50]. To date, there have been few studies that compare how these two indicators are related to fitness and health features in children and adolescents in the same sample. Our findings show that the MFR and handgrip strength-to-BMI ratio in both sexes were significantly correlated with cardiorespiratory fitness and vertical jump, while the handgrip strength-to-BMI ratio also showed a correlation with handgrip strength. Furthermore, these indicators tend to be representative over the 4 years studied.
Although MFR is a more difficult indicator to calculate as it depends on anthropometric measurements and body composition assessments obtained from specific equipment, a decrease in MFR was related to an excessive reduction in muscle strength and power in the lower extremities [51]. In contrast, body fat is strongly and inversely associated with 20-mSRT performance in children [52, 53]. This evidence supports the crucial role of other anthropometric parameters not covered by the BMI that should be taken into account for assessing the performance on physical fitness tests as well as for the health of children and adolescents. Therefore, more studies with large population data on BMI, MFR, and handgrip strength-to-BMI ratio in children and adolescents are warranted to provide further evidence on whether weight status should be considered not only through the BMI but also through other anthropometric indicators.
Major strengths of this study were [1] the novelty of this research, which includes other anthropometric indicators in young people and their relationship with different physical fitness variables; [2] the relatively large sample of rural active children and adolescents ($$n = 2256$$) who were measured using a standardised procedure; and [3] the data obtained from this population in four consecutive years. The current study also has some limitations that need to be considered. Although four different years have been studied, the design of this study is cross-sectional and therefore limits the interference with regard to the casualty of the associations examined. Moreover, the level of daily physical activity and socio-economic status were not controlled for or taken into account and may influence the generalisability of the results. Finally, body composition was not measured with advanced assessments or high precision by imaging techniques or specialised equipment such as DEXA or magnetic resonance. Nevertheless, as it is a large sample, bioimpedance may be a feasible measurement, since it is relatively cheaper, easily applied, and without radiation [54]; therefore, bioimpedance may be a good tool for routine assessment of body composition in this population. Thus, we recognise this limitation and suggest future studies using other specific equipment to confirm our findings.
## Conclusions
This study shows that weight status taken from BMI as well as MFR and handgrip strength-to-BMI is important indicators for health that are significant in different physical fitness parameters. A normal weight status presented significant values in cardiorespiratory fitness and vertical jump regarding those who were overweight and obese. Moreover, both indicators were positively correlated with handgrip strength, cardiorespiratory fitness and vertical jump (handgrip strength-to-BMI), and cardiorespiratory fitness and vertical jump (MFR) in both sexes and over the 4 years. Therefore, even though previous research determined that body composition analyses can allow for the identification and diagnosis of weight status based on BMI, the data of the present study show that the use of other body composition measurements such as MFR or handgrip strength-to-BMI can serve as a tools for identifying relationships of the paediatric population with physical fitness.
## References
1. Guthold R, Stevens GA, Riley LM, Bull FC. **Global trends in insufficient physical activity among adolescents: a pooled analysis of 298 population-based surveys with 1·6 million participants**. *The Lancet Child & Adolescent Health* (2020.0) **4** 23-35. DOI: 10.1016/S2352-4642(19)30323-2
2. Afshin A, Forouzanfar MH, Reitsma MB, Sur P, Estep K, Lee A, Marczak L. **Health effects of overweight and obesity in 195 countries over 25 years**. *N Engl J Med* (2017.0) **377** 13-27. DOI: 10.1056/NEJMoa1614362
3. Pearson N, Biddle SJ. **Sedentary behavior and dietary intake in children, adolescents, and adults: a systematic review**. *Am J Prev Med* (2011.0) **41** 178-188. DOI: 10.1016/j.amepre.2011.05.002
4. Iannotti RJ, Kogan MD, Janssen I, Boyce WF. **Patterns of adolescent physical activity, screen-based media use, and positive and negative health indicators in the US and Canada**. *J Adolesc Health* (2009.0) **44** 493-499. DOI: 10.1016/j.jadohealth.2008.10.142
5. 5.Fatima Y, Doi S, Mamun A (2015) Longitudinal impact of sleep on overweight and obesity in children and adolescents: a systematic review and bias‐adjusted meta‐analysis. Obesity Reviews 16:137–149. 10.1111/obr.12245
6. Rúa-Alonso M, Rial-Vázquez J, Nine I, Lete-Lasa JR, Clavel I, Giráldez-García MA, Rodríguez-Corral M, Dopico-Calvo X, Iglesias-Soler E. **Comparison of physical fitness profiles obtained before and during COVID-19 pandemic in two independent large samples of children and adolescents: DAFIS project**. *Int J Environ Res Public Health* (2022.0) **19** 3963. DOI: 10.3390/ijerph19073963
7. 7.World Health OGlobal action plan on physical activity 2018–2030: more active people for a healthier world2018GenevaWorld Health Organization. *Global action plan on physical activity 2018–2030: more active people for a healthier world* (2018.0)
8. 8.UN DESA (2016) Transforming our world: the 2030 agenda for sustainable development. https://wedocs.unep.org/20.500.11822/11125. Accessed 30 Jan 2023
9. 9.World Health Organization. Regional Office for Europe (2021) 2021 physical activity factsheets for the European Union Member States in the WHO European Region. World Health Organization. Regional Office for Europe, Copenhagen. https://apps.who.int/iris/handle/10665/345335. Accessed 30 Jan 2023
10. Ruiz JR, Castro-Piñero J, España-Romero V, Artero EG, Ortega FB, Cuenca MM, Jimenez-Pavón D, Chillón P, Girela-Rejón MJ, Mora J. **Field-based fitness assessment in young people: the ALPHA health-related fitness test battery for children and adolescents**. *Br J Sports Med* (2011.0) **45** 518-524. DOI: 10.1136/bjsm.2010.075341
11. Cole TJ. **Establishing a standard definition for child overweight and obesity worldwide: international survey**. *BMJ* (2000.0) **320** 1240-1240. DOI: 10.1136/bmj.320.7244.1240
12. Freedman DS, Wang J, Maynard LM, Thornton JC, Mei Z, Pierson RN, Dietz WH, Horlick M. **Relation of BMI to fat and fat-free mass among children and adolescents**. *Int J Obes* (2005.0) **29** 1-8. DOI: 10.1038/sj.ijo.0802735
13. Javed A, Jumean M, Murad MH, Okorodudu D, Kumar S, Somers V, Sochor O, Lopez-Jimenez F. **Diagnostic performance of body mass index to identify obesity as defined by body adiposity in children and adolescents: a systematic review and meta-analysis**. *Pediatr Obes* (2015.0) **10** 234-244. DOI: 10.1111/ijpo.242
14. Kyle UG, Schutz Y, Dupertuis YM, Pichard C. **Body composition interpretation: contributions of the fat-free mass index and the body fat mass index**. *Nutrition* (2003.0) **19** 597-604. DOI: 10.1016/S0899-9007(03)00061-3
15. Bigaard J, Frederiksen K, Tjønneland A, Thomsen BL, Overvad K, Heitmann BL, Sørensen TI. **Body fat and fat-free mass and all-cause mortality**. *Obes Res* (2004.0) **12** 1042-1049. DOI: 10.1038/oby.2004.131
16. Freedman DS, Kahn HS, Mei Z, Grummer-Strawn LM, Dietz WH, Srinivasan SR, Berenson GS. **Relation of body mass index and waist-to-height ratio to cardiovascular disease risk factors in children and adolescents: the Bogalusa Heart Study**. *Am J Clin Nutr* (2007.0) **86** 33-40. DOI: 10.1093/ajcn/86.1.33
17. McCarthy H, Samani-Radia D, Jebb S, Prentice A. **Skeletal muscle mass reference curves for children and adolescents**. *Pediatr Obes* (2014.0) **9** 249-259. DOI: 10.1111/j.2047-6310.2013.00168.x
18. Park BS, Yoon JS. **Relative skeletal muscle mass is associated with development of metabolic syndrome**. *Diabetes Metab J* (2013.0) **37** 458-464. DOI: 10.4093/dmj.2013.37.6.458
19. 19.Kim K, Hong S, Kim EY (2016) Reference values of skeletal muscle mass for Korean children and adolescents using data from the Korean National Health and Nutrition Examination Survey 2009–2011. PLoS One 11:e0153383. 10.1371/journal.pone.0153383
20. Bianco A, Jemni M, Thomas E, Patti A, Paoli A, Ramos Roque J, Palma A, Mammina C, Tabacchi G. **A systematic review to determine reliability and usefulness of the field-based test batteries for the assessment of physical fitness in adolescents—The ASSO Project**. *International Journal of Occupational Medicine Environmental Health* (2015.0) **28** 445-478. DOI: 10.13075/ijomeh.1896.00393
21. 21.Steffl M, Chrudimsky J, Tufano JJ (2017) Using relative handgrip strength to identify children at risk of sarcopenic obesity. PloS One 12:e0177006. 10.1371/journal.pone.0177006
22. López-Gil JF, Weisstaub G, Ramírez-Vélez R, García-Hermoso A. **Handgrip strength cut-off points for early detection of cardiometabolic risk in Chilean children**. *Eur J Pediatr* (2021.0) **180** 3483-3489. DOI: 10.1007/s00431-021-04142-8
23. 23.Hernández-Jaña S, Sanchez-Martinez J, Solis-Urra P, Esteban-Cornejo I, Castro-Piñero J, Sadarangani KP, Aguilar-Farias N, Ferrari G, Cristi-Montero C (2021) Mediation role of physical fitness and its components on the association between distribution-related fat indicators and adolescents’ cognitive performance: exploring the influence of school vulnerability. The Cogni-Action Project. Frontiers in Behavioral Neuroscience 15:746197. 10.3389/fnbeh.2021.746197
24. 24.Moosaie F, Abhari SMF, Deravi N, Behnagh AK, Esteghamati S, Firouzabadi FD, Rabizadeh S, Nakhjavani M, Esteghamati A (2021) Waist-to-height ratio is a more accurate tool for predicting hypertension than waist-to-hip circumference and BMI in patients with type 2 diabetes: a prospective study. Frontiers in Public Health 9:726288. 10.1101/2020.09.29.20203752
25. Ashwell M, Gunn P, Gibson S. **Waist-to-height ratio is a better screening tool than waist circumference and BMI for adult cardiometabolic risk factors: systematic review and meta-analysis**. *Obes Rev* (2012.0) **13** 275-286. DOI: 10.1111/j.1467-789X.2011.00952.x
26. Manzano-Carrasco S, Garcia-Unanue J, Lopez-Fernandez J, Hernandez-Martin A, Sanchez-Sanchez J, Gallardo L, Felipe JL. **Differences in body composition and physical fitness parameters among prepubertal and pubertal children engaged in extracurricular sports: the active health study**. *Eur J Pub Health* (2022.0) **32** i67-i72. DOI: 10.1093/eurpub/ckac075
27. Manzano-Carrasco S, Felipe JL, Sanchez-Sanchez J, Hernandez-Martin A, Gallardo L, Garcia-Unanue J. **Weight status, adherence to the Mediterranean diet, and physical fitness in Spanish children and adolescents: The Active Health Study**. *Nutrients* (2020.0) **12** 1680. DOI: 10.3390/nu12061680
28. 28.Shrestha B, Dunn L (2019) The declaration of helsinki on medical research involving human subjects: a review of seventh revision. J Nepal Health Res Counc 17:548–552. 10.33314/jnhrc.v17i4.1042
29. Thivel D, Verney J, Miguet M, Masurier J, Cardenoux C, Lambert C, Courteix D, Metz L, Pereira B. **The accuracy of bioelectrical impedance to track body composition changes depends on the degree of obesity in adolescents with obesity**. *Nutr Res* (2018.0) **54** 60-68. DOI: 10.1016/j.nutres.2018.04.001
30. Salton N, Kern S, Interator H, Lopez A, Moran-Lev H, Lebenthal Y, Brener A. **Muscle-to-fat ratio for predicting metabolic syndrome components in children with overweight and obesity**. *Child Obes* (2022.0) **18** 132-142. DOI: 10.1089/chi.2021.0157
31. Gulías-González R, Sánchez-López M, Olivas-Bravo Á, Solera-Martínez M, Martínez-Vizcaíno V. **Physical fitness in Spanish schoolchildren aged 6–12 years: reference values of the battery EUROFIT and associated cardiovascular risk**. *J Sch Health* (2014.0) **84** 625-635. DOI: 10.1111/josh.12192
32. Castro-Piñero J, González-Montesinos JL, Mora J, Keating XD, Girela-Rejón MJ, Sjöström M, Ruiz JR. **Percentile values for muscular strength field tests in children aged 6 to 17 years: influence of weight status**. *J Strength Cond Res* (2009.0) **23** 2295-2310. DOI: 10.1519/JSC.0b013e3181b8d5c1
33. Leger LA, Mercier D, Gadoury C, Lambert J. **The multistage 20 metre shuttle run test for aerobic fitness**. *J Sports Sci* (1988.0) **6** 93-101. DOI: 10.1080/02640418808729800
34. de Bont J, Bennett M, León-Muñoz LM, Duarte-Salles T. **The prevalence and incidence rate of overweight and obesity among 2.5 million children and adolescents in Spain**. *Revista Española de Cardiología (English Edition)* (2022.0) **75** 300-307. DOI: 10.1016/j.rec.2021.07.002
35. Ramos-Álvarez O, Arufe-Giráldez V, Cantarero-Prieto D, Ibáñez-García A. **Impact of SARS-CoV-2 lockdown on anthropometric parameters in children 11/12 years old**. *Nutrients* (2021.0) **13** 4174. DOI: 10.3390/nu13114174
36. Friend A, Craig L, Turner S. **The prevalence of metabolic syndrome in children: a systematic review of the literature**. *Metab Syndr Relat Disord* (2013.0) **11** 71-80. DOI: 10.1089/met.2012.0122
37. Ortega FB, Ruiz JR, Castillo MJ, Sjöström M. **Physical fitness in childhood and adolescence: a powerful marker of health**. *Int J Obes* (2008.0) **32** 1-11. DOI: 10.1038/sj.ijo.0803774
38. 38.Valenzuela PL, Pinto‐Escalona T, Lucia A, Martínez‐de‐Quel Ó (2022) Academic performance and psychosocial functioning in European schoolchildren: the role of cardiorespiratory fitness and weight status. Pediatric Obesity 17:e12850. 10.1111/ijpo.12850
39. Dumith SC, Ramires VV, Souza MA, Moraes DS, Petry FG, Oliveira ES, Ramires SV, Hallal PC. **Overweight/obesity and physical fitness among children and adolescents**. *J Phys Act Health* (2010.0) **7** 641-648. DOI: 10.1123/jpah.7.5.641
40. Fernández I, Canet O, Giné-Garriga M. **Assessment of physical activity levels, fitness and perceived barriers to physical activity practice in adolescents: cross-sectional study**. *Eur J Pediatr* (2017.0) **176** 57-65. DOI: 10.1007/s00431-016-2809-4
41. 41.Ortega FB, Cadenas-Sanchez C, Lee DC, Ruiz JR, Blair SN, Sui X (2018) Fitness and fatness as health markers through the lifespan: an overview of current knowledge. Progress in Preventive Medicine (New York, Ny) 3:e0013. 10.1097/pp9.0000000000000013
42. Agostinis-Sobrinho C, Kievišienė J, Rauckienė-Michaelsson A, Dubey VP, Norkiene S, Moreira C, Lopes L, Santos R. **Cardiovascular health behavior and cardiorespiratory fitness in adolescents: a longitudinal study**. *Eur J Pediatr* (2022.0) **181** 4091-4099. DOI: 10.1007/s00431-022-04623-4
43. Artero EG, Ruiz JR, Ortega FB, España-Romero V, Vicente-Rodríguez G, Molnar D, Gottrand F, González-Gross M, Breidenassel C, Moreno LA. **Muscular and cardiorespiratory fitness are independently associated with metabolic risk in adolescents: the HELENA study**. *Pediatr Diabetes* (2011.0) **12** 704-712. DOI: 10.1111/j.1399-5448.2011.00769.x
44. Dietz WH, Bellizzi MC. **Introduction: the use of body mass index to assess obesity in children**. *Am J Clin Nutr* (1999.0) **70** 123S-125S. DOI: 10.1093/ajcn/70.1.123s
45. McCarthy HD. **Body fat measurements in children as predictors for the metabolic syndrome: focus on waist circumference**. *Proceedings of the Nutrition Society* (2006.0) **65** 385-392. DOI: 10.1079/PNS2006514
46. McLean RR, Shardell MD, Alley DE, Cawthon PM, Fragala MS, Harris TB, Kenny AM, Peters KW, Ferrucci L, Guralnik JM, Kritchevsky SB, Kiel DP, Vassileva MT, Xue QL, Perera S, Studenski SA, Dam TTL. **Criteria for clinically relevant weakness and low lean mass and their longitudinal association with incident mobility impairment and mortality: the Foundation for the National Institutes of Health (FNIH) Sarcopenia Project**. *The Journals of Gerontology: Series A* (2014.0) **69** 576-583. DOI: 10.1093/gerona/glu012
47. 47.Chen YY, Fang WH, Wang CC, Kao TW, Yang HF, Wu CJ, Sun YS, Wang YC, Chen WL (2019) Fat-to-muscle ratio is a useful index for cardiometabolic risks: a population-based observational study. PLoS One 14:e0214994. 10.1371/journal.pone.0214994
48. Wang N, Sun Y, Zhang H, Chen C, Wang Y, Zhang J, Xia F, Benedict C, Tan X, Lu Y. **Total and regional fat-to-muscle mass ratio measured by bioelectrical impedance and risk of incident type 2 diabetes**. *J Cachexia Sarcopenia Muscle* (2021.0) **12** 2154-2162. DOI: 10.1002/jcsm.12822
49. Seo YG, Song HJ, Song YR. **Fat-to-muscle ratio as a predictor of insulin resistance and metabolic syndrome in Korean adults**. *J Cachexia Sarcopenia Muscle* (2020.0) **11** 710-725. DOI: 10.1002/jcsm.12548
50. 50.Reisberg K, Riso EM, Jürimäe J (2021) Physical fitness in preschool children in relation to later body composition at first grade in school. Plos One 16:e0244603. 10.1371/journal.pone.0244603
51. Ushio K, Mikami Y, Obayashi H, Fujishita H, Fukuhara K, Sakamitsu T, Hirata K, Ikuta Y, Kimura H, Adachi N. **Decreased muscle-to-fat mass ratio is associated with low muscular fitness and high alanine aminotransferase in children and adolescent boys in organized sports clubs**. *J Clin Med* (2021.0) **10** 2272. DOI: 10.3390/jcm10112272
52. Henriksson P, Sandborg J, Henström M, Delisle Nyström C, Ek E, Ortega FB, Löf M. **Body composition, physical fitness and cardiovascular risk factors in 9-year-old children**. *Sci Rep* (2022.0) **12** 2665. DOI: 10.1038/s41598-022-06578-w
53. Joensuu L, Syväoja H, Kallio J, Kulmala J, Kujala UM, Tammelin TH. **Objectively measured physical activity, body composition and physical fitness: cross-sectional associations in 9-to 15-year-old children**. *Eur J Sport Sci* (2018.0) **18** 882-892. DOI: 10.1080/17461391.2018.1457081
54. Völgyi E, Tylavsky FA, Lyytikäinen A, Suominen H, Alén M, Cheng S. **Assessing body composition with DXA and bioimpedance: effects of obesity, physical activity, and age**. *Obesity* (2008.0) **16** 700-705. DOI: 10.1038/oby.2007.94
|
---
title: 'Indicators of quality of diabetes care in persons with type 2 diabetes with
and without severe mental illness: a Danish nationwide register-based cohort study'
authors:
- Lenette Knudsen
- Stine H. Scheuer
- Lars J. Diaz
- Caroline A. Jackson
- Sarah H. Wild
- Michael E. Benros
- Dorte L. Hansen
- Marit E. Jørgensen
- Gregers S. Andersen
journal: The Lancet Regional Health - Europe
year: 2022
pmcid: PMC9989638
doi: 10.1016/j.lanepe.2022.100565
license: CC BY 4.0
---
# Indicators of quality of diabetes care in persons with type 2 diabetes with and without severe mental illness: a Danish nationwide register-based cohort study
## Body
Research in contextEvidence before this studyIn Medline, we performed a title and abstract search for all previous evidence on quality of diabetes care in persons with type 2 diabetes with and without severe mental illness (published between database inception and July 30, 2022). No language restriction was applied, and we used the following search terms in various combinations; ‘severe mental illness’, ‘schizophrenia’, ‘bipolar’, ‘major depress∗’, ‘severe depress∗’, ‘psychos∗’, ‘mani∗’, ‘type 2 diabetes’, ‘diabetes mellitus’, ‘diabetes’, ‘quality of care’, ‘process of care’, ‘care’, ‘treatment’, and ‘diabetes care’. We included studies conducted in countries with universal health care coverage, including Europe, Canada, and Australia. For studies on persons with depression, we included major or severe depression. A total of ten studies were found. Previous studies from countries with universal health care coverage have found conflicting results, with two studies reporting improved quality of care in persons with severe mental illness, one reporting no difference, and three reporting lower quality of care. Four studies reported diverse findings depending on the indicators, for example one study reported no difference in assessment of hemoglobin A1c, foot and eye screening and a higher likelihood of low-density lipoprotein-cholesterol assessment in persons with compared to persons without severe mental illness. Limitations of the previous studies included limited coverage of study populations, type, and definition of severe mental illness. In summary, studies on quality of diabetes care with all types of severe mental illness collectively and individually are limited. Added value of this studyThis study is a nationwide study providing additional evidence on receipt of diabetes care and achievement of treatment targets in persons with type 2 diabetes with and without severe mental illness. The study addresses previous gaps by providing population-based data for persons with any severe mental illness and additionally for persons with schizophrenia, bipolar disorders, and major depression. Implications of all the available evidenceOur results signify need for a change in clinical practice and health policies to reduce the gap in quality of diabetes care in persons with severe mental illness compared to persons without.
## Summary
### Background
This study aims to examine quality of diabetes care in persons with type 2 diabetes with and without severe mental illness (SMI).
### Methods
In a nationwide prospective register-based study, we followed persons with type 2 diabetes in Denmark with and without SMI including schizophrenia, bipolar disorder, or major depression. Quality of care was measured as receipt of care (hemoglobin A1c, low-density lipoprotein-cholesterol and urine albumin creatinine ratio assessment and eye and foot screening) and achievement of treatment targets between 2015 and 2019. Quality of care was compared in persons with and without SMI using generalized linear mixed models adjusted for key confounders.
### Findings
We included 216,537 persons with type 2 diabetes. At entry 16,874 ($8\%$) had SMI. SMI was associated with lower odds of receiving care, with the most pronounced difference in urine albumin creatinine ratio assessment and eye screening (OR: 0.55, $95\%$ CI: 0.53–0.58 and OR: 0.37 $95\%$ CI: 0.32–0.42, respectively). Among those with an assessment, we found that SMI was associated with higher achievement of recommended hemoglobin A1c levels and lower achievement of recommended low-density lipoprotein-cholesterol levels. Achievement of recommended low-density lipoprotein-cholesterol levels was similar in persons with versus without schizophrenia.
### Interpretation
Compared to persons without SMI, persons with SMI were less likely to receive process of care, with the most pronounced differences in urine albumin creatinine ratio assessment and eye screening.
### Funding
This study was funded by $\frac{10.13039}{100018562}$Steno Diabetes Center Copenhagen through an unrestricted grant from $\frac{10.13039}{501100009708}$Novo Nordisk Foundation.
## Evidence before this study
In Medline, we performed a title and abstract search for all previous evidence on quality of diabetes care in persons with type 2 diabetes with and without severe mental illness (published between database inception and July 30, 2022). No language restriction was applied, and we used the following search terms in various combinations; ‘severe mental illness’, ‘schizophrenia’, ‘bipolar’, ‘major depress∗’, ‘severe depress∗’, ‘psychos∗’, ‘mani∗’, ‘type 2 diabetes’, ‘diabetes mellitus’, ‘diabetes’, ‘quality of care’, ‘process of care’, ‘care’, ‘treatment’, and ‘diabetes care’. We included studies conducted in countries with universal health care coverage, including Europe, Canada, and Australia. For studies on persons with depression, we included major or severe depression. A total of ten studies were found. Previous studies from countries with universal health care coverage have found conflicting results, with two studies reporting improved quality of care in persons with severe mental illness, one reporting no difference, and three reporting lower quality of care. Four studies reported diverse findings depending on the indicators, for example one study reported no difference in assessment of hemoglobin A1c, foot and eye screening and a higher likelihood of low-density lipoprotein-cholesterol assessment in persons with compared to persons without severe mental illness. Limitations of the previous studies included limited coverage of study populations, type, and definition of severe mental illness. In summary, studies on quality of diabetes care with all types of severe mental illness collectively and individually are limited.
## Added value of this study
This study is a nationwide study providing additional evidence on receipt of diabetes care and achievement of treatment targets in persons with type 2 diabetes with and without severe mental illness. The study addresses previous gaps by providing population-based data for persons with any severe mental illness and additionally for persons with schizophrenia, bipolar disorders, and major depression.
## Implications of all the available evidence
Our results signify need for a change in clinical practice and health policies to reduce the gap in quality of diabetes care in persons with severe mental illness compared to persons without.
## Introduction
Compared to the background population, persons with severe mental illness (SMI), such as schizophrenia, bipolar disorder, and major depression have a 10–15-year shorter life expectancy.1 This may partly be due to an excess risk of type 2 diabetes and cardiovascular diseases.1 Persons with SMI have a 2–3 times higher risk of type 2 diabetes than the background population.2 Among persons with type 2 diabetes, comorbid SMI is associated with a higher risk of diabetes complications and mortality compared to persons without SMI.3 Disparity in quality of diabetes care may partly explain these poorer outcomes in persons with SMI.4 International and national diabetes care guidelines have been developed to ensure high quality of diabetes care, including annual assessments of hemoglobin A1c (HbA1c) and low-density lipoprotein (LDL)-cholesterol, and careful monitoring of achievement of treatment targets to prevent diabetes complications and mortality.5,6 However, patient-provider and system-level barriers can result in insufficient care among those with SMI, resulting in inequalities in quality of care.4 Previous studies from countries with universal health care coverage have found conflicting results,7, 8, 9, 10, 11, 12, 13, 14, 15, 16 with three studies reporting worse quality of diabetes care in persons with SMI compared to persons without,9,11,12 while others have found similar or better quality of care in persons with SMI.7,8,10,13, 14, 15, 16 However, most studies were conducted in persons with schizophrenia,9,11,13 or summarised for SMI overall,7,8,10,12,16 with inconsistencies in which SMI diagnoses were included. SMI comprises a heterogeneous group of diagnoses and summarizing overall SMI may underestimate differences within specific SMI diagnoses. Previous studies were also limited in methodology, such as limited data coverage resulting in selected populations7,9,12 or a lack of complete coverage of data on quality indicators.9,11 Most studies examined the quality of diabetes care on receipt of care8,9,11, 12, 13,15,16 and many studies only examined a few indicators.7,8,10, 11, 12, 13, 14 *In a* nationwide study, we aimed to address these gaps by examining the quality of diabetes care measured as receipt of care and achievement of treatment targets in persons with type 2 diabetes with and without SMI. We also examined whether the quality of diabetes care varied by type of SMI, including schizophrenia, bipolar disorder, and major depression.
## Study design and study population
We identified all persons with type 2 diabetes diagnosed before 2015 who were 18 years or older at the time of type 2 diabetes diagnosis and followed them to the end of 2019. The study linked person-level data with a unique personal identification number from the Danish Civil Registration System17 with Danish nationwide healthcare registers.18 Persons with type 2 diabetes were identified in a nationwide diabetes register.19 The register is based on an algorithm that collects data from five health registers containing diabetes-related information.19 Inclusion in the diabetes register includes a diabetes diagnosis in the National Patient Register,20 use of diabetes podiatry in the Danish National Health Service Register,21 purchase of any diabetes medication in the Danish National Prescription Registry,22 diabetes diagnosis in the Danish Adult Diabetes Registry,5 or an eye screening recorded in Danish Registry of Diabetic Retinopathy.23
## Definition of severe mental illness
Persons with SMI were identified in the Danish Psychiatric Research Register. The register contains records of all admissions to psychiatric inpatient facilities since 1969 and visits to outpatient and emergency psychiatric departments since 1995.24 Persons with SMI were defined as all persons with an inpatient, outpatient or emergency contact where the diagnosis included schizophrenia or schizophrenia spectrum disorder (ICD-10: F20-F29, ICD-8: 295.x9, 296.89, 297.x9, 298.29– 298.99, 299.04, 299.05, 299.09, 301.83), bipolar disorder (ICD-10: F30-F31, ICD-8: 296.19, 296.39, 298.19) or major depression (ICD-10: F32-F33, ICD-8: 296.09, 296.29, 298.09, 300.49) from 1969 (when the register started) to 31.12.2019 (end of follow-up). There has been a lack of consensus in research of which diagnosis SMI includes. However, in most research SMI is defined as schizophrenia and schizophrenia spectrum disorder, bipolar disorder, and major depression.25 These diagnoses are also used in previous register-based studies from Denmark.3,26 The date of onset of SMI was defined as the date of first contact (inpatient, outpatient, or emergency department visit). SMI were grouped into any SMI, and each specific SMI diagnosis (schizophrenia, bipolar disorder, or major depression, which were not mutually exclusive).
## Quality of diabetes care
Quality of diabetes care was measured according to Danish National Diabetes Care Guidelines.27 The quality of diabetes care was measured as receipt of care in the entire population and achievement of treatment targets was measured among those who had an assessment. Receipt of care was measured as having had an assessment of HbA1c, LDL-cholesterol, urine albumin creatinine ratio (UACR), and foot- and eye screening. Achievement of recommended treatment targets among those who had an assessment was defined on the basis of HbA1c ≤53 mmol/mol, LDL-cholesterol levels ≤2.5 mmol/l, and HbA1c >70 mmol/mol. Table 1 lists the definitions of the quality of care indicators and the data sources used for each indicator. Danish national guidelines recommended that persons with diabetes should receive an assessment of HbA1c, LDL-cholesterol, UACR, and foot screening at least once every year, and eye screening once every two years in the study period.27 We added three months to the intervals to allow for a buffer in accordance with the national quality database.28 This resulted in four 15-month intervals for HbA1c, LDL-cholesterol, UACR, and foot screening and two 27-month intervals for eye screening during the five-year follow-up. We examined assessment of each indicator in each non-overlapping interval. The end of follow-up was 31.12.2019 for all indicators except for eye screening, where end of follow-up was 30.06.2019.Table 1Definition of quality indicators for diabetes care and data sources. Quality indicatorsDefinition of indicatorsIntervalData sourcesReceipt of careAnnual assessment of HbA1cNumerator: Persons with a HbA1c assessmentDenominator: Persons with type 2 diabetes with and without SMIa15 monthsDADRNLDDNHSRAnnual assessment of LDL-cholesterolNumerator: Persons ≥30 years with a LDL-cholesterol assessmentDenominator: Persons ≥30 years old with type 2 diabetes with and without SMIb15 monthsDADR NLDAnnual assessment of UACRNumerator: Persons with a UACR assessmentDenominator: Persons with type 2 diabetes with and without SMIa15 monthsDADRNLDAnnual foot screeningNumerator: Persons with a foot screeningDenominator: Persons with type 2 diabetes with and without SMI15 monthsDADRDNHSREye screening every second yearNumerator: Persons with an eye screeningDenominator: Persons with type 2 diabetes with and without SMI27 monthsDADRDNHSRDiabaseAchievement of the treatment targetRecommended HbA1c levelsNumerator: Persons with HbA1c levels ≤53 mmol/molDenominator: Persons with an assessment of HbA1c with type 2 diabetes with and without SMIa15 monthsDADRNLDHigh HbA1c levelsNumerator: Persons with HbA1c levels ≥70 mmol/molDenominator: Persons with an assessment of HbA1c with type 2 diabetes with and without SMIa15 monthsDADRNLDRecommended LDL-cholesterol levelsNumerator: Persons ≥30 years with LDL-cholesterol levels ≤2.5 mmol/lDenominator: Persons ≥30 years with an assessmentb15 monthsDADRNLDHbA1c = Hemoglobin A1c; SMI = severe mental illness; LDL-cholesterol = low-density lipoprotein cholesterol; UACR = Urine albumin creatinine ratio; DADR = The Danish Adult Diabetes Registry; NLD = the National Laboratory Database; DNHSR = the Danish National Health Service Registry; Diabase = The Danish Registry of Diabetic Retinopathy.aPopulation excluding the Central Denmark Region.bPopulation ≥30 years excluding the Central Denmark Region.
Persons were followed from 01.01.2015 until the end of follow-up, death, or emigration, whichever came first. We excluded persons who died or emigrated within the first interval.
Data on the quality of diabetes care were obtained from the following four registers: the National Laboratory Database, which contains routine biomarker results since 2015 from all hospitals and general practitioners in all regions except the Central Denmark Region29; the Danish National Health Service Registry,21 which contains information on the use of health care services for all persons living in Denmark since 1990 and from which we used service codes related to HbA1c assessment, foot- and eye screening of persons with diabetes; the Danish Adult Diabetes Registry, containing information on the quality of diabetes care in persons with diabetes treated in outpatient clinics and general practice since 20045; and the Danish Registry of Diabetic Retinopathy containing information on retinopathy screening from all hospital eye departments and private ophthalmological practices since 2013.23 As the National Laboratory Database did not include information on persons living in Central Denmark Region, we excluded that population from the analyses of quality indicators based on information from the National Laboratory Database including HbA1c, LDL-cholesterol, and UACR. A flowchart of the different study populations used for each quality indicator is presented in Fig. 1.Fig. 1Flowchart of study populations for each quality indicator. SMI = severe mental illness; HbA1c = Hemoglobin A1c; LDL-cholesterol = low-density lipoprotein cholesterol; UACR = Urine albumin creatinine ratio.
## Definition of covariates
We used prior evidence and the method of directed acyclic graphs to identify potential confounders and mediators (Supplementary Fig. S1). The identified potential confounders were: Age, sex, calendar time, diabetes duration (as time since date of diagnosis until time of follow-up), level of education, and migrant status. Data on date of birth, sex, and migrant status, including immigrants and refugees, was obtained from the Danish Civil Registration System.17 Migrants were defined as persons born outside Denmark or with parents born outside Denmark and without Danish citizenship and categorized as Danish, Western, or Non-Western.17 Information on the highest level of education was collected from the Danish Education Registry and defined as the highest achieved education at the date of type 2 diabetes diagnosis.30 It was categorized as low (lower secondary and below), medium (upper secondary), and high (tertiary and above) according to the International Standard Classification of Education.
## Statistical analysis
Characteristics of persons at the start of follow-up were presented as mean (± standard deviation [SD]) for continuous variables and as percentages (count) for categorical variables for persons with type 2 diabetes with or without any SMI, and for persons with type 2 diabetes with or without schizophrenia, bipolar disorder, or major depression, respectively.
Mixed logistic regression models were used to examine the association between the quality indicators and SMI. The value of each repeated measure of the quality indicators was included as the outcome ($\frac{0}{1}$). The models were analyzed with a person-specific random intercept to account for the correlation between the repeated measures of the quality indicators from the same person. SMI and covariates were included as fixed effects. The models were adjusted for confounders in two steps. Model 1) included basic demographic factors, age, sex, diabetes duration, and calendar time, and model 2) additionally included socio-demographic factors, education, and migrant status. SMI was included as a time-varying variable, meaning that persons were considered unexposed to SMI until a diagnosis of SMI during follow-up and then considered exposed to SMI afterwards. As the SMI groups were not mutually exclusive, we ran separate models for each SMI (any SMI, schizophrenia, bipolar disorder, and major depression). Results from models with linear versus spline terms for each continuous variable (age and diabetes duration) were compared. The results from the different models were similar, and therefore we included a linear term for each continuous variable in the final models.
The adjusted odds ratio derived from logistic regression analysis may overestimate the risk ratio when the outcome is frequent.31 In our study, several of the outcomes were frequent (e.g., mean HbA1c assessments was $87\%$ in persons without SMI). To compensate for that, we also calculated the absolute risk (defined as the model-derived probability of an event) of each quality indicator for a given set of covariates.
We conducted a complete case analysis, and therefore excluded $9\%$ of our study population due to missing information on education.
Statistical analyses were performed using R, version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria; www.R-project.org).
## Ethics
Register-based studies do not require ethical approval in Denmark. The Danish Data Protection Agency has granted access to, and use of data, and all data were anonymized.
## Data statement
All study data are held at Statistics Denmark's servers and are confidential due to privacy reasons. Access to data requires application and permission from the registries.
## Role of funding source
This study was funded by Steno Diabetes Center Copenhagen through an unrestricted grant from Novo Nordisk Foundation.
## Results
We followed 216,537 persons with type 2 diabetes; of whom 16,874 ($8\%$) had any SMI, 12,155 ($6\%$) major depression, 6080 ($3\%$) schizophrenia, and 2259 ($1\%$) bipolar disorders (flowchart presented in Fig. 1). Of those with any SMI, 15,176 ($90\%$) were diagnosed with any SMI at start of follow-up, while 1698 ($10\%$) were diagnosed with any SMI during follow-up and a total of 11,747 ($70\%$) received the diagnosis before or on the same date as the type 2 diabetes diagnosis.
Of all persons with any SMI, $72\%$ [12,155] were diagnosed with major depression, $36\%$ [6080] with schizophrenia, and $13\%$ [2259] with bipolar disorder.
Persons with any SMI, schizophrenia, or major depression were more likely to be younger, women, have lower education, and be of non-Western descent than persons without any SMI, schizophrenia, or major depression, respectively (Table 2). Persons with bipolar disorder were also more likely to be younger, women, but had similar education levels and migration status, compared to persons without (Table 2).Table 2Characteristics of persons with any SMI, schizophrenia, bipolar disorder, major depression or without any SMI, schizophrenia, bipolar, and major depression at the start of follow-up. Without any SMI($$n = 199$$,663)Any SMI($$n = 16$$,874)Schizophrenia($$n = 6080$$)Bipolar disorder($$n = 2259$$)Major depression($$n = 12$$,155)Age at start of follow-up, mean (±SD) years66.7 (12.2)62.2 (13.5)59.6 (13.2)63.0 (12.0)63.0 (13.6)Women, no. % 88,863 (44.5)9347 (55.4)3171 (52.2)1288 (57.0)7069 (58.2)Diabetes duration at start of follow-up, median (IQR)6.2 [3.2; 11.4]5.9 [3.0; 11.1]6.0 [3.1; 11.2]6.0 [3.2; 11.0]5.9 [3.0; 11.1]Education at type 2 diabetes diagnosis, no. (%) Low76,066 (38.1)7414 (48.6)2983 (49.1)870 (38.6)5149 (42.4) Medium75,454 (37.8)5413 (35.5)1744 (28.7)789 (34.9)4010 (33.0) High29,262 (14.7)2441 (16.0)742 (12.2)434 (19.2)1869 (15.4) Missing, no (%)18,881 (9.5)1606 (9.5)611 (10.0)166 (7.3)1127 (9.2)Migrant status, no. (%) Danish177,237 (88.8)14,463 (85.7)5114 (84.1)2074 (91.8)10,483 (86.3) Western decent4901 (2.5)436 (2.6)151 (2.5)66 (2.9)307 (2.5) Non-Western decent17,525 (8.8)1975 (11.7)815 (13.4)119 (5.3)1365 (11.2)Type of SMI, no. (%) Schizophrenia6080 (36.0)6080 (100.0)883 (39.1)1952 (16.1) Bipolar disorder2259 (13.4)883 (14.5)2259 (100.0)1251 (10.3) Major depression12,155 (72.0)1952 (32.1)1251 (55.4)12,155 (100.0)Receipt of care during the entire follow-up: HbA1c assessment, mean (±SD)a0.87 (0.34)0.85 (0.36)0.84 (0.36)0.85 (0.36)0.85 (0.35) UACR assessment, mean (±SD)a0.55 (0.50)0.46 (0.50)0.42 (0.49)0.43 (0.49)0.48 (0.50) LDL-cholesterol assessment, mean (±SD)b0.81 (0.39)0.78 (0.41)0.78 (0.42)0.79 (0.41)0.79 (0.41) Foot screening, mean (±SD)0.50 (0.50)0.46 (0.50)0.46 (0.50)0.50 (0.50)0.46 (0.50) Eye screening, mean (±SD)0.67 (0.47)0.56 (0.50)0.53 (0.50)0.54 (0.50)0.57 (0.49)Achieving treatment targets in persons with assessments during the entire follow-up HbA1c ≤53 mmol/mol, mean (±SD)c0.59 (0.49)0.60 (0.49)0.61 (0.49)0.65 (0.48)0.60 (0.49) HbA1c ≥70 mmol/mol, mean (±SD)c0.13 (0.33)0.15 (0.36)0.16 (0.36)0.12 (0.33)0.15 (0.36) LDL-cholesterol ≤2.5 mmol/l, mean (±SD)d0.76 (0.43)0.72 (0.45)0.73 (0.44)0.72 (0.45)0.71 (0.46)SMI = Severe mental illness; HbA1c = Hemoglobin A1c; LDL-cholesterol = low-density lipoprotein cholesterol; UACR = Urine albumin creatinine ratio; SD = Standard deviation; IQR = Interquartile range.aPopulation excluding the Central Denmark Region ($$n = 169$$,100).bPopulation ≥30 years excluding the Central Denmark Region ($$n = 168$$,176).cAmong the population with assessments excluding the Central Denmark Region: without any SMI $$n = 135$$,458 ($87\%$ of the population), with any SMI $$n = 10$$,961 ($84\%$ of the population), with schizophrenia $$n = 4066$$ ($83\%$ of the population), with bipolar disorder $$n = 1396$$ ($82\%$ of the population), with major depression $$n = 7837$$ ($85\%$ of the population).dAmong the population with assessments ≥30 years excluding the Central Denmark Region: without any SMI $$n = 133$$,769 ($86\%$ of the population), with any SMI $$n = 10$$,740 ($83\%$ of the population), with schizophrenia $$n = 3954$$ ($82\%$ of the population), with bipolar disorder $$n = 1376$$ ($81\%$ of the population), with major depression $$n = 7693$$ ($84\%$ of the population).
Differences in receipt of care and achievement of treatment targets over the entire follow-up adjusted for confounders are presented in Fig. 2.Fig. 2Odds Ratios ($95\%$ CI) for receipt of care and achievement of treatment targets in persons with any SMI, schizophrenia, bipolar disorder, or major depression compared to persons without any SMI, schizophrenia, bipolar disorder, or major depression, respectively (model 2∗). ∗Model 2 adjusted for age, sex, diabetes duration, calendar time, education, and migrant status. † In persons with assessments. SMI = Severe mental illness; HbA1c = Hemoglobin A1c; LDL-cholesterol = low-density lipoprotein cholesterol; UACR = Urine albumin creatinine ratio; CI = confidence interval.
## Receipt of care
Persons with any SMI, schizophrenia, bipolar disorder, and major depression had lower odds of receiving HbA1c, LDL-cholesterol, UACR assessments, and eye screenings than persons without the specific SMI (Fig. 2). We found the lowest odds for UACR assessments and eye screenings (results for any SMI: OR: 0.55, $95\%$ CI: 0.53–0.58 and OR: 0.37, $95\%$ CI: 0.32–0.42, respectively).
The odds of receipt of assessments of HbA1c and LDL-cholesterol were similar across the different SMI diagnoses, whereas it differed for UACR and eye screening. For UACR assessments and eye screenings, the effect was greater for persons with schizophrenia and bipolar disorder compared to persons with major depression. Persons with any SMI or major depression had lower odds of receiving foot screening than those without. This was also the case with schizophrenia or bipolar disorder, albeit the latter analyses did not reach statistical significance.
The absolute risk for persons with fixed covariates was $45.1\%$ vs. $59.7\%$ for UACR assessment and $69.5\%$ vs. $75.3\%$ for foot screening in persons with vs. without any SMI. The absolute risk for LDL-cholesterol was $92.6\%$ vs. $95.1\%$ in persons with vs. without any SMI, whereas it was close to one for both HbA1c assessment and eye screening (e.g., the absolute risk for eye screening was $99.8\%$ in persons with any SMI and $99.9\%$ in persons without SMI) (absolute risks are presented in Supplementary Table S3).
In line with our results, previous studies have reported a lower receipt of care for assessments of HbA1c, LDL-cholesterol, UACR, and eye screening11,12 and one study found no difference in foot screening for persons with and without schizophrenia.9 Contrary to our findings, other studies found no difference in receipt of assessment of HbA1c,7,9,16 LDL-cholesterol8,9 and no difference8,16 or marginally lower odds of foot- or eye screening and receipt of UACR assessment9 in persons with SMI. However, one study found a higher number of LDL-cholesterol assessments in persons with SMI16 and another study found higher odds of UACR assessment.16 A recent Scottish study found that persons with SMI were more likely to receive HbA1c, LDL-cholesterol, UACR, and foot- and eye screening the first year after type 2 diabetes diagnosis compared to persons without,15 which is contrary to our results. However, when examining the quality of care over 10 years, persons with SMI were less likely to receive eye screening, which was in line with our results.
The difference between our results and previous studies could be due to differences in methodology, such as data sources and the definition of study populations. The definition of the SMI population differed in our study and previous studies.7,8,16 For example, one study defined SMI as schizophrenia or bipolar disorder whereas we also included major depression.16 A Scottish study only based the definition of SMI on inpatient contacts,15 whereas we also included outpatient contacts.
The definition of the diabetes population also differed in our study compared to previous studies. Our study included complete data for all persons with type 2 diabetes from outpatient clinics and primary care. Whereas a Scottish study only included persons with newly diagnosed type 2 diabetes,15 a UK study only included persons with type 2 diabetes treated in selected general practices,16 and a Danish study included persons with type 1 or type 2 diabetes.9 The differences between the Danish and Scottish studies could also be an expression of better quality of diabetes care in persons with SMI in Scotland. In Scotland, the pay-performance scheme for general practitioners offered financial incentives to promote good practice, including assessing cardiometabolic risk factors in persons with SMI.34 In Scotland, foot screening is expected to be performed as part of the annual review of persons with diabetes and invitations to eye screening on a specified date and in a specified place are sent to persons with diabetes, with the opportunity to change the appointment by telephone. In Denmark, general practitioners do not have the same financial incentives to promote care and persons with diabetes are expected to arrange their own foot and eye screening. However, whether the differences are due to differences in methodology or health care should be addressed in future studies.
## Achievement of treatment targets
Among persons who had an assessment, any SMI, schizophrenia, bipolar disorder, or major depression were associated with higher odds of achieving HbA1c targets. Compared to persons without, persons with schizophrenia or bipolar disorders had the highest odds of having HbA1c ≤53 mmol/mol (OR 1.98, $95\%$ CI: 1.77–2.22; OR 1.90, $95\%$ CI: 1.57–2.31, respectively). We found no differences in odds of HbA1c >70 mmol/mol in persons with any SMI or major depression compared to persons without the specific SMI. In contrast, we found lower odds of HbA1c >70 mmol/mol in persons with schizophrenia or bipolar disorders than in those without, however, the confidence intervals included 1 (OR 0.85 [0.72–1.00]; OR 0.79 [0.60–1.04] respectively).
In persons who had an assessment, persons with any SMI or major depression alone had lower odds of LDL-cholesterol ≤2.5 mmol/l (OR 0.84, $95\%$ CI: 0.78–0.91; OR 0.78, $95\%$ CI: 0.71–0.85, respectively) compared to persons without, while we found no difference for persons with bipolar disorder or schizophrenia when compared to persons without the specific SMI.
Adjustment for potential confounders only slightly changed the effect estimates (results of model 1 are shown in Supplementary Table S2, and results of model 2 are shown in Fig. 2).
The absolute risk for the treatment target HbA1c ≤53 mmol/mol was $79.4\%$ vs. $72.1\%$ in persons with vs. without any SMI and for HbA1c >70 mmol/mol it was $0.5\%$ in both persons with and without any SMI. For LDL-cholesterol the absolute risk was $89.9\%$ vs. $91.4\%$ in persons with vs. without any SMI (absolute risks are presented in Supplementary Table S3).
Two previous studies found that SMI was associated with higher proportions of persons achieving good glycemic control,7,16 which was in line with our findings. Opposite this, one study found lower proportions achieving good glycemic control10 and two studies found no difference.13,14 In line with our findings, one previous study found that depression was associated with better achievement of lipid targets,14 while two other studies found no difference between persons with and without SMI.10,16 Two of the previous studies were based on crude data,7,13 whereas we controlled for possible confounders and examined repeated measures over time in mixed-effect models which could explain the differences in findings.
## Main findings
In this nationwide prospective follow-up study, we found that persons with SMI had markedly lower receipt of HbA1c, LDL-cholesterol, UACR assessments, and eye screenings compared to persons without SMI. The difference was most pronounced for UACR assessment and eye screening, where persons with SMI had $45\%$ and $63\%$ lower odds of receiving assessments of UACR or eye screening, respectively.
Among persons with an assessment, we found that persons with SMI had higher achievement of recommended HbA1c levels, while they had a lower achievement of recommended LDL-cholesterol levels compared to persons without SMI. However, some of the results differed when comparing persons with and without schizophrenia or bipolar disorders. For example, persons with schizophrenia had no difference in achieving recommended LDL-cholesterol levels compared to persons without schizophrenia.
For HbA1c assessment and eye screening and to some extent also LDL-cholesterol assessment there was a very high coverage of assessments and screenings both in persons with and without SMI (absolute risks were close to one), suggesting that the lower odds from the logistic regression may exaggerate a risk association.31 Thus, the results related to these indicators may be of limited clinical importance.
The revealed inequalities in receiving care in persons with SMI could be due to patient-provider level barriers. In periods with severe psychiatric symptoms, physical health often comes second, both among professionals and persons with diabetes.4 The treatment of SMI and diabetes in two compartmentalised health systems might contribute to more barriers in offering a routine follow-up to persons with diabetes. In Denmark, $80\%$ of persons with type 2 diabetes have a general practitioner as their primary diabetes health professional, and the remaining persons with more complex treatment courses receive care in endocrinological outpatient clinics.32 The diabetes health professionals are responsible for ensuring annual assessment of HbA1c, LDL-cholesterol, UACR, and foot- and eye screening among persons with diabetes. The diabetes health professionals prescribes an annual assessment of HbA1c, LDL-cholesterol, and UACR at a laboratory. The diabetes health professionals do encourage their patients to book an appointment for foot- and eye screening, but the person with diabetes have to book appointments with the podiatrist and ophthalmologist themselves. The cost of foot screenings is partly subsidized, and ophthalmologists often have long waiting times. Mental health services in Denmark are responsible for annual assessment of HbA1c and LDL-cholesterol among persons receiving active psychiatric treatment who have not already received this in primary care. This is to monitor for side effects of the psychiatric treatment.
More pronounced difference for UACR and eye screening among persons with SMI could therefore be due to the additional barriers in obtaining these assessments. UACR assessment obviously requires the individual to collect a urine sample, which persons often find unpleasant or difficult and needs extra encouragement from the health professionals. Persons with SMI may face more challenges with providing the urine sample or the diabetes health professional may be more reluctant to encourage sample collection in this group. Eye screening is conducted by an ophthalmologist, which could be far away from the persons’ home and the persons will have to book the appointment themselves. Persons with SMI may be less willing to receive care in a less familiar setting and to book and remember to attend the appointment themselves.
We found that among persons with assessments, those with SMI were more likely to have recommended HbA1c levels. These findings could be because a lower proportion with SMI received care in the first place. It is likely, that a smaller proportion receiving care often results in improved achievement of treatment targets, as the persons receiving care may be healthier than persons not receiving care. Another possible explanation could be that both diabetes and psychiatric health professional pay attention to and react to the results of the HbA1c assessments. On the other hand, we found that any SMI and major depression were associated with lower achievement of recommended LDL-cholesterol.
We found a difference in receipt of diabetes care and achievement of treatment targets across SMI diagnoses highlighting the importance of analyzing each diagnosis separately. The difference may indicate diverse awareness or barriers within different diagnoses. However, the reasons need to be explored further and addressed.
## Comparison with previous studies
In this study of persons with type 2 diabetes, we found that $8\%$ had co-existing SMI, $6\%$ major depression, $3\%$ schizophrenia, and $1\%$ bipolar disorder. The prevalence was higher in our study compared to a Scottish study reporting that of all persons with type 2 diabetes, $1\%$, $0.5\%$, and $3\%$ had a hospital admission for respectively schizophrenia, bipolar disorder, or major depression.15 The higher prevalence in our study is likely due to the inclusion of both in and out-patient contacts. Opposite this, a systematic review found that the prevalence of depression was $18\%$ in persons with type 2 diabetes33 However, they included mild, moderate, and major depression, whereas we only included major depression, which can explain the differences in prevalence.
## Strengths and limitations
Our study has several strengths. The use of different nationwide registers made it possible to construct a nationwide prospective study with data on almost all persons in Denmark with type 2 diabetes with and without SMI, with no selection due to health coverage or participation in a survey. This means that the findings are generalizable to Denmark's entire type 2 diabetes population. The cohort of persons with type 2 diabetes is based on the diabetes register, which is constructed using five national registers.19 In Denmark, around $80\%$ of persons with type 2 diabetes are treated in general practice and therefore do not have a diagnosis in the National Patient Register.19 However, these persons are captured in the diabetes register, as it uses diabetes-defining information from other registers such as use of podiatry in the Danish National Health Service Registry, diabetes medication in the Danish National Prescription Registry, and eye examination in the Danish Registry of Diabetic Retinopathy.23 Despite the strength of including persons with type 2 diabetes treated in general practice, we were not able to capture persons with undiagnosed diabetes. In Denmark, no systematic screening for type 2 diabetes exists nor for persons with SMI. Whether or not more person with SMI have undiagnosed diabetes is difficult to predict.
We used complete data on quality indicators from several registers with high coverage and high data validity.17,19,21,24,29,30 For example, this included data on HbA1c, LDL-cholesterol, and UACR from the National Laboratory Database, which provides information on all laboratory tests in the entire study population except for persons living in Denmark Central Region, who was excluded for these analyses. The longitudinal nature of the data allowed us to examine the quality of diabetes care over five years and account for changes over time. Moreover, we examined receipt of care and achievement of treatment targets which provided a more nuanced exploration of the quality of diabetes care, whereas previous studies have primarily focused on receipt of care.8,9,11,12,15,16 Additionally, we examined the quality of diabetes care in persons with type 2 diabetes with and without any SMI and specific diagnoses of SMI, which allowed us to examine differences overall and across different SMI diagnoses. Lastly, we could distinguish between the type of diabetes, thus including persons with type 2 diabetes only. Several former studies have not distinguished between type 1 and type 2 diabetes.9,12,16 Our study also has some limitations. Since SMI was ascertained using in- and outpatient psychiatric hospital records, we did not include persons with SMI who received a diagnosis in primary care or at a private psychologist. However, as most persons with a suspected SMI would be referred to a psychiatric hospital, we do not believe this would exclude a large proportion with SMI. Although we included persons from in- and outpatient psychiatric records, it was impossible to distinguish the ascertainment route, so we could not examine differences in quality indicators in different severity of SMI. Our study only included persons with more severe cases of depression, referred to as major depression, requiring treatment in the secondary health care sector, so the findings might not be generalizable to persons with less severe depression treated in primary care by a general practitioner or a psychologist. Potential confounders and mediators were identified using directed acyclic graphs and based on prior evidence. However, we cannot reject that a different directed acyclic graph would have changed the structure of the analyses. We excluded around $9\%$ of our study population due to missing information on the level of education. When comparing persons with and without missing information on education, we found that persons with missing information were older, had longer duration of diabetes and were more often migrants (Supplementary Table S1). These persons might also receive a lower quality of diabetes care30 and thus this exclusion might have introduced selection bias, which could result in some underestimation of our findings.
There was a large proportion of missingness in achievement of treatment targets, with 13–$19\%$ of persons without any measurements during follow-up. We were only able to examine differences in persons with values of HbA1c and LDL-cholesterol, where a higher proportion with SMI had missing values. This means that we may have introduced selection bias in the results on achieving treatment targets.
Investigation of the role of the well-recognized metabolic effects of treatments for SMI was beyond the scope of this study and requires further research, particularly among persons with diabetes. Data on other important receipt of care and treatment targets including blood pressure and body mass index were not available in this study. Further research is required to address whether more stringent treatment targets for sub-groups of the study population for example persons with a history of cardiovascular disease or albuminuria were met and whether recommended lipid-lowering or diabetes treatments were prescribed appropriately.
## Conclusions
Persons with SMI had a markedly lower receipt of assessment of HbA1c, LDL-cholesterol, UACR, and eye screening, compared to persons without SMI, with the most pronounced differences for UACR and eye screening. Due to high coverage of HbA1c and LDL-cholesterol assessments and eye screening, the finding related to UACR assessments may be of highest clinical importance.
Among persons with assessments, we found that persons with SMI had better achievement of recommended levels of HbA1c and lower achievement of recommended LDL-cholesterol levels. These results may reflect persons with SMI who are healthier and have fewer complications than those who did not receive assessments.
Our findings highlight the need to develop effective interventions to reduce marked inequalities in diabetes care between persons with and without SMI. The pronounced differences could contribute to higher risk of complications and mortality in persons with diabetes and SMI compared to persons with diabetes only.
## Contributors
LK, SHS, MEB, DLH, MEJ and GSA led the conception, design, and planning of the study. LK and SHS lead data management and analyses with support from LJD and GSA. LK led drafting of the work with support from SHS. All authors contributed to the interpretation of the data and revising the manuscript critically for important intellectual content and read and approved the final manuscript. LK and SHS are responsible for the overall content of the manuscript as guarantors. LK, SHS, LJD and GSA had access to the data and LK, SHS and GSA controlled the decision to publish.
## Data sharing statement
The data used in this study are held at Statistics Denmark's servers. The data are confidential for data privacy reasons and therefore, cannot be made publicly available. Access to data requires an application and permission from the different owners of the registers.
## Prior presentation
Parts of this study were presented at the European Diabetes Epidemiology Group Annual meeting in Greece, 2nd – 5th April 2022.
## Declaration of interests
LK: holds shares in Novo Nordisk A/S, SHS: none, LJD: none, CAJ: none, SHW: none, MEB: none, DLH: None, MEJ: holds shares in Novo Nordisk, has received research grants from AMGEN, Astra Zeneca, $\frac{10.13039}{100001003}$Boehringer Ingelheim, Novo Nordisk and Sanofi Aventis, GSA: holds shares in Novo Nordisk A/S.
## Supplementary data
Supplementary Materials STROBE-checklist-v4-cohort
## References
1. Wahlbeck K., Westman J., Nordentoft M., Gissler M., Laursen T.M.. **Outcomes of Nordic mental health systems: life expectancy of patients with mental disorders**. *Br J Psychiatry* (2011) **199** 453-458. PMID: 21593516
2. Lindekilde N., Scheuer S., Rutters F.. **Prevalence of type 2 diabetes in psychiatric disorders: an umbrella review with meta-analysis of 245 observational studies from 32 systematic reviews**. *Diabetologia* (2022) **65** 440-456. PMID: 34841451
3. Scheuer S.H., Kosjerina V., Lindekilde N.. **Severe mental illness and the risk of diabetes complications. a nationwide register-based cohort study**. *J Clin Endocrinol Metab* (2022) **107** e3504-e3514. PMID: 35359003
4. De Hert M., Cohen D., Bobes J.. **Physical illness in patients with severe mental disorders. II. Barriers to care, monitoring and treatment guidelines, plus recommendations at the system and individual level**. *World Psychiatry* (2011) **10** 138-151. PMID: 21633691
5. Jørgensen M.E., Kristensen J.K., Reventlov Husted G., Cerqueira C., Rossing P.. **The Danish Adult diabetes registry**. *Clin Epidemiol* (2016) **8** 429-434. PMID: 27843339
6. **Executive summary: standards of medical care in diabetes--2014**. *Diabetes Care* (2014) **37** S5-S13. PMID: 24357214
7. Das-Munshi J., Schofield P., Ashworth M.. **Inequalities in glycemic management in people living with type 2 diabetes mellitus and severe mental illnesses: cohort study from the UK over 10 years**. *BMJ Open Diabetes Res Care* (2021) **9**
8. Han L., Doran T., Holt R.I.G.. **Impact of severe mental illness on healthcare use and health outcomes for people with type 2 diabetes: a longitudinal observational study in England**. *Br J Gen Pract* (2021) **71** e565-e573. PMID: 33571951
9. Jørgensen M., Mainz J., Carinci F., Thomsen R.W., Johnsen S.P.. **Quality and predictors of diabetes care among patients with schizophrenia: a Danish nationwide study**. *Psychiatr Serv* (2018) **69** 179-185. PMID: 29032706
10. Kristensen F.P., Rohde C., Østergaard S.D., Thomsen R.W.. **Four-year HbA1c and LDL-cholesterol trajectories among individuals with mental disorders and newly developed type 2 diabetes**. *Brain Behav* (2021) **11**
11. Kurdyak P., Vigod S., Duchen R., Jacob B., Stukel T., Kiran T.. **Diabetes quality of care and outcomes: comparison of individuals with and without schizophrenia**. *Gen Hosp Psychiatry* (2017) **46** 7-13. PMID: 28622820
12. Mai Q., Holman C.D., Sanfilippo F.M., Emery J.D., Preen D.B.. **Mental illness related disparities in diabetes prevalence, quality of care and outcomes: a population-based longitudinal study**. *BMC Med* (2011) **9** 118. PMID: 22044777
13. Rathmann W., Pscherer S., Konrad M., Kostev K.. **Diabetes treatment in people with type 2 diabetes and schizophrenia: retrospective primary care database analyses**. *Prim Care Diabetes* (2016) **10** 36-40. PMID: 25937183
14. Rohde C., Knudsen J.S., Schmitz N., Østergaard S.D., Thomsen R.W.. **The impact of hospital-diagnosed depression or use of antidepressants on treatment initiation, adherence and HbA(1c)/LDL target achievement in newly diagnosed type 2 diabetes**. *Diabetologia* (2021) **64** 361-374. PMID: 33073329
15. Scheuer S.H., Fleetwood K.J., Licence K.A.M.. **Severe mental illness and quality of care for type 2 diabetes: a retrospective population-based cohort study**. *Diabetes Res Clin Pract* (2022) **190**
16. Whyte S., Penny C., Phelan M., Hippisley-Cox J., Majeed A.. **Quality of diabetes care in patients with schizophrenia and bipolar disorder: cross-sectional study**. *Diabet Med* (2007) **24** 1442-1448. PMID: 18042084
17. Pedersen C.B.. **The Danish Civil registration system**. *Scand J Public Health* (2011) **39** 22-25. PMID: 21775345
18. Thygesen L.C., Daasnes C., Thaulow I., Brønnum-Hansen H.. **Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving**. *Scand J Public Health* (2011) **39** 12-16. PMID: 21898916
19. Carstensen B., Rønn P.F., Jørgensen M.E.. **Prevalence, incidence and mortality of type 1 and type 2 diabetes in Denmark 1996-2016**. *BMJ Open Diabetes Res Care* (2020) **8**
20. Lynge E., Sandegaard J.L., Rebolj M.. **The Danish national patient register**. *Scand J Public Health* (2011) **39** 30-33. PMID: 21775347
21. Andersen J.S., Olivarius Nde F., Krasnik A.. **The Danish national health service register**. *Scand J Public Health* (2011) **39** 34-37. PMID: 21775348
22. Kildemoes H.W., Sørensen H.T., Hallas J.. **The Danish national prescription Registry**. *Scand J Public Health* (2011) **39** 38-41. PMID: 21775349
23. Andersen N., Hjortdal J., Schielke K.C.. **The Danish Registry of diabetic retinopathy**. *Clin Epidemiol* (2016) **8** 613-619. PMID: 27822108
24. Mors O., Perto G.P., Mortensen P.B.. **The Danish psychiatric central research register**. *Scand J Public Health* (2011) **39** 54-57. PMID: 21775352
25. Suijkerbuijk Y.B., Schaafsma F.G., van Mechelen J.C., Ojajärvi A., Corbière M., Anema J.R.. **Interventions for obtaining and maintaining employment in adults with severe mental illness, a network meta-analysis**. *Cochrane Database Syst Rev* (2017) **9**
26. Hjorthøj C., Østergaard M.L., Benros M.E.. **Association between alcohol and substance use disorders and all-cause and cause-specific mortality in schizophrenia, bipolar disorder, and unipolar depression: a nationwide, prospective, register-based study**. *Lancet Psychiatry* (2015) **2** 801-808. PMID: 26277044
27. **Treatment guidelines for type 2 diabetes**. (2019)
28. 28The Danish Clinical Quality Program (RKKP)Danish Diabetes Database national annual report 2019/20202020Regionernes Kliniske KvalitetsudviklingsprogramAarhus. (2020)
29. Arendt J.F.H., Hansen A.T., Ladefoged S.A., Sørensen H.T., Pedersen L., Adelborg K.. **Existing data sources in clinical Epidemiology: laboratory information system databases in Denmark**. *Clin Epidemiol* (2020) **12** 469-475. PMID: 32547238
30. Jensen V.M., Rasmussen A.W.. **Danish education registers**. *Scand J Public Health* (2011) **39** 91-94. PMID: 21775362
31. Zhang J., Yu K.F.. **What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes**. *JAMA* (1998) **280** 1690-1691. PMID: 9832001
32. Thomsen R.W., Friborg S., Nielsen J.S., Schroll H., Johnsen S.P.. **The Danish Centre for Strategic Research in Type 2 Diabetes (DD2): organization of diabetes care in Denmark and supplementary data sources for data collection among DD2 study participants**. *Clin Epidemiol* (2012) **4** 15-19. PMID: 23071407
33. Ali S., Stone M.A., Peters J.L., Davies M.J., Khunti K.. **The prevalence of co-morbid depression in adults with Type 2 diabetes: a systematic review and meta-analysis**. *Diabet Med* (2006) **23** 1165-1173. PMID: 17054590
34. Roland M.. **Linking physicians' pay to the quality of care--a major experiment in the United Kingdom**. *N Engl J Med* (2004) **351** 1448-1454. PMID: 15459308
|
---
title: Epidemiology of hypertension among adults in Addis Ababa, Ethiopia
authors:
- Mulugeta Mekonene
- Kaleab Baye
- Samson Gebremedhin
journal: Preventive Medicine Reports
year: 2023
pmcid: PMC9989685
doi: 10.1016/j.pmedr.2023.102159
license: CC BY 4.0
---
# Epidemiology of hypertension among adults in Addis Ababa, Ethiopia
## Highlights
•*Hypertension is* highly prevalent among adults in Addis Ababa.•The risk of hypertension is higher in older age, men, and the obese.•*Hypertension is* associated with poor sleep quality.•Regular monitoring of blood pressure and weight-loss interventions are needed.
## Abstract
The public health significance of hypertension is increasing in low- and middle-income countries. However, there is limited epidemiological evidence in Ethiopia. We assessed the prevalence of hypertension and explored its predictors among adults in Addis Ababa, Ethiopia. A community-based cross-sectional study was conducted from April to May 2021 among randomly selected adults aged 18–64 years. A face-to-face interview using an adapted STEPwise Approach to NCD Risk Factor Surveillance (STEPS) questionnaire was conducted. Multilevel mixed-effects logistic regression model was fitted to determine factors associated with hypertension. The sample consisted of a total of 600 adults (mean age: 31.2 ± 11.4 years, $51.7\%$ women). The overall age-standardized prevalence of hypertension was $22.1\%$ and $47.8\%$ according to the Seventh Joint National Commission (JNC7) and the 2017 American Heart Association (AHA) guidelines, respectively. About $25.6\%$ were newly diagnosed with hypertension. The age groups of 40–54 years (AOR = 8.97; $95\%$ CI: 2.35,34.23), and 55–64 years (AOR = 19.28; $95\%$ CI: 3.96,93.83) as compared to the 18–24 age group, male sex (AOR = 2.90; $95\%$ CI: 1.22,6.87), obesity (AOR = 1.92; $95\%$ CI: 1.02,3.59), abdominal obesity (AOR = 4.26; $95\%$ CI: 1.42,12.81), and very poor sleep quality (AOR = 3.35; $95\%$ CI: 1.15,9.78) were independent predictors of hypertension. This study revealed that the burden of hypertension among adults is very high. Hypertension is independently associated with older age group, male sex, obesity, abdominal obesity, and poor sleep quality. Therefore, the study highlights the need to develop regular blood pressure surveillance programs, weight loss intervention, and improvement of sleep quality.
## Background
Non-communicable diseases (NCDs) are the leading cause of death worldwide, causing 41 million deaths every year, which is equivalent to>$71\%$ of all global deaths (WHO, 2021). The figure is expected to rise to 52 million by 2030 (WHO, 2021b). Globally, between 2000 and 2019, the total adult mortality attributable to NCDs increased by $31\%$ (WHO, 2020). More than three-fourths of NCD-related deaths occurred in low- and middle-income countries (LMIC) (WHO, 2021b).
Hypertension is an important determinant of cardiovascular disease (CVD). Known risk factors include genetic, behavioral, and environmental exposure throughout life (NCD Risk Factor Collaboration, 2017). Specific components of the diet (especially sodium and potassium), obesity, alcohol, smoking, physical inactivity, and psychological stress are also linked with hypertension (Li and Shang, 2021, NCD Risk Factor Collaboration, 2017). Hypertension is a major contributor to CVD such as stroke and heart diseases, causing $45\%$ of heart disease-related deaths and $51\%$ of deaths attributed to stroke worldwide (WHO, 2013). Out of 17.9 million total deaths due to CVDs in 2019, more than half (10.8 million) were from hypertension complications (GBD Risk Factors Collaborators, 2020, WHO, 2021a).
The global burden of hypertension is increasing drastically. In 2010, an estimated 1.39 billion adults, equivalent to a prevalence of $31.1\%$ had hypertension worldwide. The magnitude is increasing in LMIC (1.04 billion), largely due to economic growth, dietary change, and an aging population, while remaining stable or decreasing in high-income countries (HIC) (349 million people) (Mills et al., 2020). The swift epidemiological change in LMIC from communicable to NCDs in the past few decades is largely attributable to four major modifiable risk factors like tobacco use, physical inactivity, alcohol, and unhealthy diet (WHO, 2021b).
Similar to other LMICs, *Ethiopia is* also undergoing an epidemiological transition shifting the causes of mortality from infectious diseases to NCDs. In particular, hypertension incidence is increasing at an unprecedented rate (Mills et al., 2020, Tesfa and Demeke, 2021). The fragile health system that is already been overstretched by communicable diseases, is unlikely to withstand the increasing burden of NCDs if timely prevention measures are not taken. A systematic meta-analysis and observational studies including a STEPS survey conducted in Ethiopia, estimated the prevalence of hypertension in Addis Ababa to be between $15\%$ and $30\%$ (Asemu et al., 2021, Ethiopian Public Health Institute, 2016, Kibret and Mesfin, 2015, Tesfa and Demeke, 2021, Tesfaye et al., 2009, Tiruneh et al., 2020). Although studies indicated hypertension as an epidemy in urban areas, surveillance systems on its epidemiology and related risk factors have not been established.
This study used the baseline survey of the SuNCD-AA (Surveillance of Non-Communicable Diseases in Addis Ababa) established for monitoring the epidemiology of NCDs including hypertension. The study will be used as baseline data for monitoring the epidemiology of hypertension in Addis Ababa. Establishing a regular surveillance system in the city with an appropriate study design is vital to capture the timely changes. The evidence is also important to develop specific hypertension preventive strategies and interventions. Therefore, this study aimed to assess the epidemiology and risk factors of hypertension among adults in Addis Ababa, Ethiopia.
## Study design and setting
This study was a part of the baseline survey of SuNCD-AA project conducted from May to June 2021 and employed a community-based cross-sectional design.
The study was conducted in Addis Ababa, the capital and largest city of Ethiopia, which has an estimated population of 4.5 million, of which $68\%$ are adults 18–64 years of age (Population Census Commission [Ethiopia], 2008). Administratively, Addis *Ababa is* divided into 10 sub-cities with 116 districts and has 12 public hospitals, 40 private hospitals, 96 health centers, and >800 clinics.
## Study population
In the SuNCD-AA baseline survey, men and women 18 to 64 years of age, who were permanent residents of Addis Ababa city were eligible for inclusion regardless of their medical history. Women participants were excluded if they had a self-reported pregnancy or gave birth in the past 12 months of the survey.
## Sample size determination and sampling techniques
A total of 600 eligible adults 18–64 years of age were enrolled in the baseline survey at the selected households. The required sample was determined using Cochran's single population proportion formula (Cochran, 1977) by assuming: a $20.6\%$ expected prevalence of hypertension (Tesfa and Demeke, 2021), $95\%$ confidence level, $4\%$ margin of error, and design effect (DEFF) of 1.5. DEFF of 1.5 was determined using the standard DEFF = 1 + δ (n – 1) formula taking cluster size (n) of 20 and intra-cluster correlation of (δ) of $2\%$.
Subjects were selected using a multistage cluster sampling technique. The study included all 10 sub-cities. One district (‘woreda’) was randomly picked from each sub-city, resulting in 2 villages (’ketena’ -the smallest geographical unit of the district) selected from each district randomly, and 20 villages were represented overall. From each village, 30 households were randomly selected using a computer-generated random number from the urban health extension workers' database. One eligible subject was picked randomly from households with multiple options. Those who declined to participate or couldn't be found after repeated attempts were replaced with individuals from nearby households.
## Variables of the study
The primary outcome of interest was hypertension (coded as yes/no). Hypertension was defined as systolic blood pressure (SBP) of 140 mmHg or more, or diastolic blood pressure (DBP) of 90 mmHg or above, or currently on medication based on JNC 7 classification (Chobanian et al., 2003), and SBP of 130 mmHg or more, or DBP of 80 mmHg or above, or currently on medication as per guidelines provided by the 2017 American Heart Association (AHA) (Whelton et al., 2018).
Socio-demographic characteristics: sex, age, educational status, marital status, occupation, religion, household size, and wealth index; Anthropometric and behavioural factors: body mass index, abdominal obesity, waist-to-hip ratio, smoking status, alcohol consumption, khat chewing, physical activity level, and fruit and vegetable intakes; Other factors: diabetes mellitus, stress score, sleep duration, and quality, excessive sleepiness, snoring, and family history of hypertension were the independent variables that were used to explain the dependent variable.
## Data collection tools and procedures
Data were gathered through face-to-face interviews using an adapted WHO ‘STEPwise approach for NCD surveillance’ questionnaire, a standard method for monitoring behavioural, dietary, and metabolic risk factors of NCDs (WHO, 2022). The questionnaire was modified according to the STEPS manual to include locally relevant items such as Khat chewing status. In addition to the STEPS questions, perceived stress and sleep pattern variables were added. The instrument was also translated to the Amharic language, pretested, and contextualized to the local setting. Selected questions extracted from the Ethiopian Demographic and Health Survey (EDHS) questionnaire were used to collect socio-demographic and household economy-related information.
Physical activity level was measured by the Global Physical Activity Questionnaire (GPAQ) tool. The instrument explores three main areas of day-to-day activities: work (including domestic work), transport, and recreational activities. Subsequently, total physical activity level was classified as high or low, based on the metabolic equivalent of task (MET)-minutes per week (WHO, 2022).
Current alcohol consumption and smoking status were assessed. Participants who consumed any amount of alcohol in the past 30 days were considered alcohol consumers. Khat (Catha Edulis Forsk) (green leaf with stimulant effect) chewing habits of participants were assessed based on ever or never chewed khat. Fruit and vegetable consumption was assessed by asking participants the number of days fruit and vegetables were consumed in a typical week.
The stress level was measured using Cohen’s 10-item Perceived Stress Scale (PSS) (Cohen et al., 1983). Response categories were based on a 5-point Likert scale ranging from never [0] to very often [4]. To obtain PSS scores, all the items were summed up after the response of four positively stated items has been reverse coded. Then we categorized the stress level into low (0–13), moderate (14–26), and high (27–40) perceived stress. The sleep duration was categorized as short (<7 h per day), normal (7–9 h per day), and long (>9 h per day).
Data were digitally collected using the Open Data Kit (ODK)® system via KoBo Toolbox® server. Enumerators and supervisors were trained nurses with extensive field experience. Four-day training was offered to the data collectors using a training manual.
## Anthropometric measurements
The weight, height, waist, and hip circumferences of participants were measured following standard procedures. The weight of the participants was measured using SH2003B® digital scale (accuracy ± 100 g) and the scale was tared to zero before each measurement. Participants’ weight was measured without shoes and heavy clothing, and recorded to the nearest 0.1 kg. Height was measured without shoes to the nearest 0.1 cm using a portable Heuer® stadiometer. BMI is calculated as body weight in kilograms divided by height squared in meters (kg/m2) and then classified as underweight (<18.5), normal weight (18.5–24.9), overweight (25–29.9), and obese (≥30) following standard cut-off points.
Waist and hip circumferences were measured as a measure of central obesity by a non-stretchable flexible tape with minimal clothing to the nearest 0.1 cm. Waist circumference (WC) was measured by placing a tape around the bare abdomen at the midpoint between the lower margin of the last palpable rib and the top of the iliac crest of the hip bone, and classified as normal (men < 94 cm and women < 80 cm), an increased risk (men 94–102 cm and women 80–88 cm) or greatly increased risk (men > 102 cm and women > 88 cm) (WHO, 2008). Hip circumference was measured by placing a measuring tape around the hip at the maximum circumference over the buttocks or around the greater trochanter of the femoral bone. Waist-to-hip ratio (WHR) was classified as normal (men < 0.90 and women < 0.85) or substantially increased (men ≥ 0.90 and women ≥ 0.85) (WHO, 2008).
All anthropometric measurements were performed in duplicate and if the difference was within a tolerable range (200 g for weight, 0.5 cm for height, and 1 cm for waist and hip circumferences), the average was used. Otherwise, the measurements were repeated.
## Blood pressure measurement
Blood pressure was measured using a Folee® automated digital monitor system following standard procedures. It was taken in a sitting position from the left arm with feet flat on the floor and arm supported at heart level after 15 min of rest. In individuals with recent exercise, smoking, heavy meal, or caffeine intake, the measurement was delayed for at least 30 min. The measurement was repeated twice and if the difference was within the acceptable limit (10 mmHg in SBP and 5 mmHg in DBP) the average was recorded. Otherwise, a new set of readings was taken.
## Blood glucose measurement
Fasting and random blood glucose level were determined from capillary blood using the Diavue® monitoring system. Based on the American Diabetic Association (ADA), we defined diabetic states by aggregating fasting blood sugar ≥ 126 mg/dl or postprandial blood sugar ≥ 200 mg/dl or on medication for raised blood sugar (American Diabetes Association Professional Practice et al., 2022).
## Statistical analysis
The Statistical Package for the Social Sciences (IBM Corp., SPSS for Windows Version 26: New York, USA) was used for data processing and analysis. Additional analysis was made using STATA/IC 15.0 (College Station, TX: StataCorp LLC). Categorical variables are expressed using frequency distributions. The normality of numeric variables was first assessed using the Kolmogorov-Smirnov test and then appropriate measures of central tendency and dispersion were used to summarize the data. Arithmetic mean (±standard deviation (SD) and median (inter-quartile range (IQR)) were applied for normal and skewed distributions, respectively. For proportions, a $95\%$ confidence interval (CI) was estimated using STATA’s binomial CI calculator.
The association between hypertension and the predictors were measured by comparing normotensive and hypertensive individuals. Bivariable and multivariable multilevel mixed-effects logistic regression were fitted, by taking villages as clusters. Crude (COR) and adjusted (AOR) odds ratios were reported. Explanatory variables with a p-value < 0.25 in the bivariate model were fitted into the multivariate model to compute AOR and a p-value < 0.05 was considered statistically significant. Multicollinearity was assessed using the multicollinearity diagnostics (variance inflation factor (VIF) and tolerance test.
We standardized the prevalence and predictors of hypertension by assigning survey weights estimated using the age and sex profile of the city’s population. The product of design weight and poststratification weight were computed to determine survey weights. Design weight was calculated as the inverse of the sampling fraction. Post-stratification weight was determined from the recent national population census using the reported age and sex composition of the city (Population Census Commission [Ethiopia], 2008).
## Ethical considerations
The SuNCD-AA study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Institutional Review Board of the College of Health Sciences, Addis Ababa University (ref # $\frac{109}{20}$/SPH). Informed written consent was obtained from each subject without inducement or undue influence.
## Basic characteristics of study participants
A total of 600 adults between 18 and 64 years of age, from all sub-cities in Addis Ababa participated in the study. Of the total participants, 310 ($51.7\%$) were female. The mean (±SD) age of the participants was 31.2 (±11.4) years. More than two-thirds of participants were employed, the majority ($93\%$) had formal education, and nearly half ($46.8\%$) were married. The majority ($81.1\%$) were Orthodox Christians, followed by Muslims ($15.3\%$). The median (IQR) household size was 4 (3–5) and the median monthly household income was 4,000 (2500–7000) Ethiopian Birr (equivalent to 95 (59–167) USD). Moreover, about $39.2\%$ of participants were categorized into medium wealth index (Table 1).Table 1Socio-demographic characteristics of the survey participants, Addis Ababa, Ethiopia, June 2021.VariablesFrequency ($$n = 600$$)PercentageSex Men29048.3 Women31051.7Age (years) 18–2421435.6 25–3925642.6 40–549716.2 55–64335.5Educational status No formal schooling427.0 Primary education16928.2 Secondary education24941.5 Higher education14023.3Marital status Married/Cohabiting28146.8 Not ever married26944.8 Widowed203.3 Divorced/separated315.1Occupation Not working (including retired)18230.4 Trade (including petty trade)10717.9 Student10217.0 Professional/technical/managerial9115.1 Manual (skilled or unskilled)7813.0 Others396.6Religion Orthodox Christian48681.1 Muslim9215.3 Protestants193.1 Others40.6Household size 1–438764.6 ≥521335.4Wealth index Poor15427.4 Medium22139.2 Rich18733.3
## Anthropometry and behavioural characteristics
The mean (±SD) BMI was 23.2 kg/m2 (±4.2) among males and 25.0 kg/m2 (±4.5) among females. The majority had a normal BMI, while $29.7\%$ and $9\%$ of the adults were overweight and obese, respectively. The mean WC was 81.8 cm (±13.9) in men and 80.7 cm (±13.9) in women; whereas, the mean WHR was 0.89 (±0.12) and 0.83 (±0.12) in the two genders, respectively. About $35\%$ of adults had increased or greatly increased risk based on the WC classification, and based on the WHR index, $40.9\%$ had substantially increased risk (Table 2).Table 2Anthropometric and behavioural characteristics of the survey participants, Addis Ababa, Ethiopia, June 2021.VariablesFrequencyPercentageBMI Underweight427.0 Normal32654.4 Overweight17829.7 Obese549.0Abdominal obesity Normal39365.5 Increased risk10717.9 Greatly increased risk10016.6WHR Normal35459.1 Substantially increased24640.9Family history of HTN Yes17829.6 No42270.4Fruit consumed in days per week <3 days48480.7 ≥3 days11619.3Vegetable consumed in days per week <3 days42370.5 ≥3 days17729.5Note. BMI = Body Mass Index; HTN = Hypertension; WHR = Waist-to-hip ratio.
Three hundred forty-five ($57.6\%$) consumed fruits, at least once per week with a median of 1 (0–2) days/week, and one-fifth of respondents consumed more than three days/week. Similarly, more than three forth 459 ($76.5\%$) of the participants consumed vegetables at least once per week with a median of 2 (1–3) days/week, and a third of them consumed more than three days/week.
The perceived stress scale estimated that about $2.6\%$ of survey participants had a high-stress level while, $49.6\%$ and $47.8\%$ had low and moderate stress scores, respectively. The median reported sleep duration of the study participants was 8 (7–9) hours/day, $25.3\%$ of the subjects reported a sleep duration of < 7 h/day, $32.9\%$ reported an average sleep duration of > 9 h/day and the rest had a normal duration of sleep (Table 3).Table 3Stress and sleep pattern characteristics of the survey participants, Addis Ababa, Ethiopia, June 2021.VariablesFrequencyPercentageStress score Low29849.6 Moderate28747.8 High162.6Sleep duration Short (<7 h)15225.3 Normal (7–9 h)25141.8 Long (>9 h)19732.9Sleep quality Very good15325.4 Good30951.5 Average9315.5 Poor315.2 Very poor142.4Excessive sleepiness Yes12420.7 No47679.3Snoring No50283.7 Yes9215.4 Don’t know60.9
## Prevalence of hypertension
The mean SBP and DBP of the participants were 119.3 mmHg ($95\%$ CI: 117.9, 120.6) and 79.2 mmHg ($95\%$ CI: 78.3, 80.1), respectively. The mean SBP was 120.7 mmHg ($95\%$ CI: 118.7–122.8) among males and 117.9 mmHg ($95\%$ CI: 116.1, 119.7) among females. Similarly, the mean DBP was 80.2 mmHg ($95\%$ CI: 78.8, 81.6) in males and 78.3 mmHg ($95\%$ CI: 77.1, 79.5) in females.
The overall weighted prevalence of hypertension in Addis Ababa was $22.1\%$ ($95\%$ CI: 18.8, 25.5), higher among men ($25.7\%$) than women ($18.8\%$). Nearly one in every four adults in the city had hypertension. Of the total hypertensive respondents, $25.6\%$ had just been diagnosed in the survey.
The prevalence significantly increased with age ($p \leq 0.001$): 40–54 years ($46.5\%$, $95\%$ CI: 38.2, 54.8) and 55–64 years ($64.9\%$, $95\%$ CI: 56.0, 73.8). The prevalence of hypertension varied statistically between the BMI categories ($p \leq 0.001$); the highest reported was $50.2\%$ ($95\%$ CI: 37.2, 63.1) in the obese group, followed by the overweight group at $31.3\%$ ($95\%$ CI: 22.8, 39.8). The magnitude of hypertension increased steadily from $14.4\%$ among participants in the normal abdominal obesity category compared to $47.5\%$ in the greatly increased risk category (p for trend < 0.001). The prevalence was also significantly higher in individuals with substantially increased than normal WHR ($31.4\%$ vs $15.6\%$, $p \leq 0.001$), and in people with diabetes than non-diabetic ($53.5\%$ vs $18.8\%$, $p \leq 0.001$). However, we found no association between physical activity level and hypertension ($$p \leq 0.96$$).
The proportion of hypertension according to the JNC7 guideline among Addis Ababa adults was $22.1\%$, while the corresponding prevalence was $47.8\%$ in the new guideline of ACC/AHA 2017, with a relative increase of $116\%$ (Fig. 1).Fig. 1Prevalence of hypertension among adults according to JNC7 and ACC/AHA 2017 guidelines, Addis Ababa, Ethiopia, June 2021 ($$n = 600$$). Note. JNC = Joint National Commission; ACC = American College of Cardiology; AHA = American Heart Association.
## Predictors of hypertension
In the bivariate logistic regression model, age groups of 40–54 years, 55–64 years, marital status, BMI, abdominal obesity, DM status, and sleep quality were found significantly associated with hypertension (Table 4).Table 4Bivariable and multivariable multilevel logistic regression analyses of factors associated with hypertension among adults in Addis Ababa, Ethiopia, June 2021.a. VariablesCategorynHypertensive% ($95\%$ CI)COR ($95\%$ CI)AOR ($95\%$ CI)SexWomen5818.8 (13.5, 24.1)11Men7525.7 (18.8, 32.6)1.64 (0.91, 2.93)2.90 (1.22, 6.87) *Age Group18–24157.1 (7.5, 15.0)1125–395119.8 (14.3, 25.3)3.27 (0.94, 11.32)3.57 (0.85, 15.00)40–544546.5 (38.2, 54.8)11.60 (3.81, 35.27) **8.97 (2.35, 34.23) **55–642164.9 (56.0, 73.8)24.41 (6.37, 93.44) **19.28 (3.96, 93.83) **Education levelNo formal schooling1229.1 (17.2, 44.7)1Primary education3721.9 (16.2, 28.8)0.67 (0.30, 1.46)NISecondary education5220.8 (16.1, 26.3)0.63 (0.30, 1.34)NIHigher education3222.6 (16.4, 30.3)0.67 (0.27, 1.66)NIMarital statusSingle, not ever married3914.6 (10.9, 19.4)11Married/Cohabiting7225.6 (20.8, 31.1)2.04 (1.08, 3.86) *0.62 (0.32, 1.18)Widowed1468.9 (45.3, 85.6)13.55 (6.16, 29.79) **2.50 (0.83, 7.55)Divorced/separated825.3 (12.9, 43.8)2.13 (0.74, 6.06)0.71 (0.25, 2.04)BMINormal4614.0 (10.2, 17.7)11Underweight49.9 (0.7, 19.1)0.83 (0.17, 4.01)0.85 (0.16, 4.49)Overweight5631.3 (24.5, 38.2)2.93 (1.77, 4.87) **1.69 (0.89, 3.21)Obese2750.2 (36.7, 63.6)7.15 (4.39, 11.64) **1.92 (1.02, 3.59) *Abdominal obesityNormal5714.4 (9.4, 19.5)11Increased risk2926.5 (15.7, 37.3)2.40 (1.07, 5.37) *2.56 (0.69, 9.46)Greatly increased risk4747.5 (39.0, 56.0)6.79 (3.36, 13.69) **4.26 (1.42, 12.81) *WHRNormal5615.6 (9.6, 21.6)1.001.00Substantially increased7731.4 (25.8, 37.1)2.54 (1.41, 4.58) **0.48 (0.23, 1.01)Current smoking statusNo12821.8 (18.7, 25.4)1Yes530.7 (13.2, 56.2)1.40 (0.50, 3.88)NICurrent alcohol useNo6923.4 (18.9, 28.6)1Yes6420.9 (16.6, 25.8)0.84 (0.59, 1.18)NILevel of physical activityHigh10122.3 (18.7, 26.4)1Low3221.5 (15.6, 28.9)0.98 (0.51, 1.87)NIDM statusNormal8718.8 (14.0, 23.7)11Prediabetes2725.8 (15.8, 36.0)1.53 (0.67, 3.50)1.06 (0.51, 2.18)Diabetes1953.5 (38.0, 69.2)5.05 (2.68, 9.50) **1.78 (0.93, 3.41)Stress scoreLow6521.6 (17.3, 26.7)1Moderate6422.5 (18.0, 27.7)0.94 (0.58, 1.50)NIHigh423.8 (8.6, 51.0)1.07 (0.33, 3.41)NISleep duration7–9 h5421.7 (17.0, 27.2)1<7 h3522.9 (16.8, 30.2)1.10 (0.65, 1.85)NI>9 h4422.0 (16.7, 28.4)1.03 (0.53, 1.99)NISleep qualityVery good2012.8 (8.3, 19.1)11Good8126.0 (21.4, 31.2)2.56 (1.17, 5.59) *2.22 (1.08, 4.54)Average2122.8 (15.3, 32.5)2.00 (0.81, 4.95)1.99 (0.69, 5.75)Poor724.0 (12.0, 42.3)2.37 (0.91, 6.18)1.19 (0.43, 3.30)Very poor427.4 (10.2, 55.7)2.52 (0.64, 9.82)3.35 (1.15, 9.70) *Fruit per week≥3 days2218.6 (12.5, 26.8)1<3 days11122.9 (19.4, 26.9)1.32 (0.50, 3.45)NIVegetable per week≥3 days3821.3 (15.9, 28.0)1<3 days9522.4 (18.7, 26.6)1.08 (0.49, 2.37)NINote.a Result obtained from multilevel mixed-effects regression model considering cluster as level 2; *Significantly associated with P ≤ 0.05; ** Significantly associated with $P \leq 0.01$ on multiple logistic regression; NI, not included because $P \leq 0.25$ in the unadjusted model. Abbreviations: n = number of hypertensive individuals; CI = confidence interval; AOR = adjusted odds ratio; COR = crude odds ratio.
Our multilevel multivariable logistic regression model suggested several independent factors that were identified as hypertension predictors after controlling for multiple covariates. Age and sex were associated with hypertension. The odds of hypertension increased in the age group of 40–54 years (AOR = 8.97; $95\%$ CI: 2.35, 34.23) and 55–64 years (AOR = 19.28; $95\%$ CI: 3.96, 93.83) as compared to participants with 18–24 years of age. The risk of hypertension was greater in men (AOR = 2.90; $95\%$ CI: 1.22, 6.87) than in women.
Obese (BMI ≥ 30) participants were more likely to be hypertensive (AOR = 1.92; $95\%$ CI: 1.02, 3.59) compared to normal BMI. Moreover, participants with abdominal obesity were more likely to develop hypertension (AOR = 4.26; $95\%$ CI: 1.42, 12.81). The study also indicated very poor sleep quality increased the odds of hypertension (AOR = 3.35; $95\%$ CI: 1.15, 9.78) compared to subjects with very good sleep quality. Additionally, diabetes was shown to increase the likelihood of hypertension though with borderline insignificance (AOR = 1.78; $95\%$ CI: 0.93, 3.41, $$p \leq 0.07$$) (Table 4).
However, in this study, important risk factors hypothesized in other studies including smoking, alcohol use, physical activity, stress level, sleep duration, low fruit, and vegetable consumption were not significantly associated with hypertension.
## Discussion
We assessed the prevalence of hypertension and its associated risk factors in Addis Ababa. Hypertension is fairly high in adults from Addis Ababa. Our study found a significant association between hypertension and older age group, male sex, obesity, abdominal obesity, and very poor sleep quality.
Age-standardized prevalence of hypertension was $22.1\%$, which implies nearly one-in-four adults in Addis Ababa are hypertensive. This study also found the prevalence of hypertension in $25.7\%$ of men and $18.8\%$ of women. The overall result is almost similar to the recent studies (Hasan et al., 2018, Mosisa et al., 2021, Tesfa and Demeke, 2021, Tiruneh et al., 2020). However, it is lower as compared to global and sub-Saharan regional prevalence (Mills et al., 2020, NCD Risk Factor Collaboration, 2017). We also estimated the prevalence according to the JNC7 and ACC/AHA 2017 guidelines which were $22.1\%$ and $47.8\%$, respectively- with a relative increase of $116\%$. A study from Iran and India reported similar findings (Gupta et al., 2020, Mirzaei et al., 2020). Currently, *Ethiopia is* using the JNC7 guideline but a study should be conducted on which guideline to be utilized for earlier detection and better blood pressure control.
Our study revealed increasing age and being male as important predictors of hypertension. Supporting evidence was reported from the studies conducted in Ethiopia and elsewhere (Asemu et al., 2021, Dereje et al., 2020, Hasan et al., 2018, Kumma et al., 2021, Omar et al., 2020, Tiruneh et al., 2020, Zaki et al., 2021). Hypertension increases with age due to structural changes in the walls of blood vessels (Pinto, 2007), and men are more likely to be susceptible to behavioral risk factors.
Obesity, diet (salt, fruits, and vegetables), and physical inactivity are major risk factors for hypertension (Mills et al., 2020). In this study, obesity and abdominal obesity exhibited a significant association with hypertension as reported in earlier studies (Asemu et al., 2021, Badego et al., 2020, Tesfa and Demeke, 2021). Obesity can increase blood pressure through a series of mechanisms, including insulin resistance, activation of the sympathetic nervous system, and sodium retention resulting in increased renal reabsorption and activation of the renin-angiotensin system (Rahmouni et al., 2005). Thus, the rise in the proportion of overweight and obese adults in Addis Ababa ($38.7\%$) and low consumption of fruits and vegetables might indicate the increasing incidence of hypertension in the future. This finding suggests the need for weight loss interventions focusing on dietary and physical activity to decrease the rate of hypertension.
The link between sleep, stress, and hypertension has been given due attention in recent studies. We found that a very poor sleep quality increases the odds of hypertension, studies from the US (Li and Shang, 2021) and China (Li et al., 2019) reported similar findings. Poor sleep quality and increased stress influences leptin and ghrelin levels in the body which increase appetite and reduce energy expenditure, this eventually results in obesity and increased risk of hypertension (Spiegel et al., 2004). This study found that sleep, stress, diet, and nutritional status are increasingly related and highlight the importance of integrated approaches for the prevention of hypertension.
Generally, without intervention, the incidence of hypertension is expected to continue to increase, therefore, future large intervention studies and clinical trials are warranted to test integrated strategies based on nutrition-related weight loss, sleep improvement, and stress management for hypertension prevention and control, especially in the urban adult populations.
This study applied survey weights to adjust differences in the probability of selection where no study has been done with a similar design in the study setting so far. To make the study comprehensive, important determinants like sleep quality and psychological stress were included. Though, the study is not without limitations. The current study is cross-sectional, and therefore, causal inference cannot be established. Although the study is comprehensive, it would have been better to include risk factors like hyperlipidemia and salt intake in our model. Several factors were found not associated with hypertension, partly due to the small variations observed but could also -in some cases- be due to low sample size.
## Conclusion
Hypertension prevalence is high in Addis Ababa. This indicates that hypertension become an increasing trend in the population. Older age group, male sex, obesity, abdominal obesity, and very poor sleep quality were significantly associated with hypertension. Therefore, the study suggests a need to develop regular blood pressure surveillance programs for early diagnosis and prevention of complications; applying integrated intervention strategies such as weight loss through proper nutrition augmented with physical activity for the prevention of overall obesity, and improvement of sleep quality to reduce the risk of developing hypertension.
## Funding
The study was funded by Addis Ababa University, Thematic Research Program. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
## CRediT authorship contribution statement
Mulugeta Mekonene: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization. Kaleab Baye: Conceptualization, Methodology, Investigation, Writing – review & editing, Supervision, Funding acquisition. Samson Gebremedhin: Conceptualization, Methodology, Investigation, Data curation, Writing – review & editing, Visualization, 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. American Diabetes Association Professional Practice, C., American Diabetes Association Professional Practice, C., Draznin, B., Aroda, V.R., Bakris, G., Benson, G., Brown, F.M., Freeman, R., Green, J., et al. 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes-2022. Diabetes Care. 2022;45(Supplement_1):S17-S38. Doi: 10.2337/dc22-S002.
2. Asemu M.M., Yalew A.W., Kabeta N.D., Mekonnen D., Kirchmair R.. **Prevalence and risk factors of hypertension among adults: a community based study in Addis Ababa, Ethiopia**. *PLoS One* (2021.0) **16** e0248934. DOI: 10.1371/journal.pone.0248934
3. Badego B., Yoseph A., Astatkie A., Odoi A.. **Prevalence and risk factors of hypertension among civil servants in Sidama Zone, south Ethiopia**. *PLoS One* (2020.0) **15** e0234485. DOI: 10.1371/journal.pone.0234485
4. Chobanian A.V., Bakris G.L., Black H.R., Cushman W.C., Green L.A., Izzo J.L., Jones D.W., Materson B.J., Oparil S., Wright J.T., Roccella E.J.. **Seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure**. *Hypertension* (2003.0) **42** 1206-1252. PMID: 14656957
5. Cochran W.G.. (1977.0)
6. Cohen S., Kamarck T., Mermelstein R.. **A global measure of perceived stress**. *J. Health Soc. Behav.* (1983.0) **24** 385-396. DOI: 10.2307/2136404
7. Dereje N., Earsido A., Temam L., Abebe A.. **Uncovering the high burden of hypertension and its predictors among adult population in Hosanna town, southern Ethiopia: a community-based cross-sectional study**. *BMJ Open* (2020.0) **10** e035823
8. Ethiopian Public Health Institute. (2016.0)
9. GBD Risk Factors Collaborators. **Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet* (2020.0) **396** 1223-1249. PMID: 33069327
10. Gupta K., Ramakrishnan S., Zachariah G., Rao J.S., Mohanan P.P., Venugopal K., Sateesh S., Sethi R., Jain D., Bardolei N., Mani K., Kakar T.S., Jain V., Gupta P., Gupta R., Bansal S., Nath R.K., Tyagi S., Wander G.S., Gupta S., Mandal S., Senguttuvan N.B., Subramanyam G., Roy D., Datta S., Ganguly K., Routray S.N., Mishra S.S., Singh B.P., Bharti B.B., Das M.K., Deb P.K., Deedwania P., Seth A., Shivkumar Rao J., Singh B.P., Bharti B.B., Sinha A.K., Gupta K., Ramakrishnan S., Bhushan S., Verma S.K., Bhargava B., Roy A., Bansal S., Sood S., Isser H.S., Pandit N., Nath R.K., Tyagi S., Trehan V., Gupta M.D., Girish M.P., Ahuja R., Manchanda S.C., Mohanty A., Jain P., Shrivastava S., Kalra I.P.S., Sarang B.S., Ratti H.S., Sahib G.B., Gupta R., Amit S.K.A., Goswami K.C., Bahl V.K., Chopra H.K., Seth A., Zachariah G., Mohanan P.P., Venugopal K., Koshy G., Nair T., Shyam N., Roby A., George R., Kumar S., Kader A., Abraham M., Viswanathan S., Jabir A., Menon J., Unni G., Mathew C., Jayagopal P.B., Sajeev P.K., Ashokan S., Asharaf A.K., Mandal N., Pancholia A.K., Bardolei N., Gupta R., Bardolei D., Das A., Aggarwal S.N., Malviya S.S., Routray S.M., Mishra P., Ali N., Barward Y.S., Singh D., Tomar S., Chaddha C., Dani K., Vyas S., Bhatt G.S., Doshi S., Wander C.B., Gupta S., Meena N.B., Sateesh G., Senguttuvan A.M., Subramanyam R., Subramanyam V., Muruganandam R.K., Sethi D., Narain P., Saran S., Jain P.K., Jain D., Kumar S., Goel K., Roy M.K., Datta S., Ganguly S., Das A., Kumar S., Chandra P.K.. **Impact of the 2017 ACC/AHA guidelines on the prevalence of hypertension among Indian adults: Results from a cross-sectional survey**. *Int J Cardiol Hypertens.* (2020.0) **7** 100055. DOI: 10.1016/j.ijchy.2020.100055
11. Hasan M., Sutradhar I., Akter T., Das Gupta R., Joshi H., Haider M.R., Sarker M., Li Y.. **Prevalence and determinants of hypertension among adult population in Nepal: data from Nepal Demographic and Health Survey 2016**. *PLoS One* (2018.0) **13** e0198028. DOI: 10.1371/journal.pone.0198028
12. Kibret K.T., Mesfin Y.M.. **Prevalence of hypertension in Ethiopia: a systematic meta-analysis**. *Public Health Rev.* (2015.0) **36** 14. DOI: 10.1186/s40985-015-0014-z
13. Kumma W.P., Lindtjørn B., Loha E., Sichieri R.. **Prevalence of hypertension, and related factors among adults in Wolaita, southern Ethiopia: a community-based cross-sectional study**. *PLoS One* (2021.0) **16** e0260403. PMID: 34910760
14. Li C., Shang S.. **Relationship between sleep and hypertension: findings from the NHANES (2007–2014)**. *Int. J. Environ. Res. Public Health* (2021.0) **18** 7867. DOI: 10.3390/ijerph18157867
15. Li M., Yan S., Jiang S., Ma X., Gao T., Li B.. **Relationship between sleep duration and hypertension in northeast China: a cross-sectional study**. *BMJ Open* (2019.0) **9** e023916. DOI: 10.1136/bmjopen-2018-023916
16. Mills K.T., Stefanescu A., He J.. **The global epidemiology of hypertension**. *Nat. Rev. Nephrol.* (2020.0) **16** 223-237. DOI: 10.1038/s41581-019-0244-2
17. Mirzaei M., Mirzaei M., Mirzaei M., Bagheri B.. **Changes in the prevalence of measures associated with hypertension among Iranian adults according to classification by ACC/AHA guideline 2017**. *BMC Cardiovasc. Disord.* (2020.0) **20** 372. DOI: 10.1186/s12872-020-01657-0
18. Mosisa G., Regassa B., Biru B.. **Epidemiology of hypertension in selected towns of Wollega zones, Western Ethiopia, 2019: a community-based cross-sectional study**. *SAGE Open Med.* (2021.0) **9** 1-8. DOI: 10.1177/20503121211024519
19. NCD Risk Factor Collaboration. **Worldwide trends in blood pressure from 1975 to 2015: a pooled analysis of 1479 population-based measurement studies with 19.1 million participants**. *Lancet* (2017.0) **389** 37-55. PMID: 27863813
20. Omar S.M., Musa I.R., Osman O.E., Adam I.. **Prevalence and associated factors of hypertension among adults in Gadarif in eastern Sudan: a community-based study**. *BMC Public Health* (2020.0) **20** 291. DOI: 10.1186/s12889-020-8386-5
21. Pinto E.. **Blood pressure and ageing**. *Postgrad. Med. J.* (2007.0) **83** 109-114. DOI: 10.1136/pgmj.2006.048371
22. Population Census Commission [Ethiopia]. (2008.0)
23. Rahmouni K., Correia M.L., Haynes W.G., Mark A.L.. **Obesity-associated hypertension: new insights into mechanisms**. *Hypertension* (2005.0) **45** 9-14. DOI: 10.1161/01.HYP.0000151325.83008.b4
24. Spiegel K., Tasali E., Penev P., Van Cauter E.. **Brief communication: sleep curtailment in healthy young men is associated with decreased leptin levels, elevated ghrelin levels, and increased hunger and appetite**. *Ann. Intern. Med.* (2004.0) **141** 846-850. PMID: 15583226
25. Tesfa E., Demeke D.. **Prevalence of and risk factors for hypertension in Ethiopia: a systematic review and meta-analysis**. *Health Sci. Rep.* (2021.0) **4** e372. PMID: 34589614
26. Tesfaye F., Byass P., Wall S.. **Population based prevalence of high blood pressure among adults in Addis Ababa: uncovering a silent epidemic**. *BMC Cardiovasc. Disord.* (2009.0) **9** 39. DOI: 10.1186/1471-2261-9-39
27. Tiruneh S.A., Bukayaw Y.A., Yigizaw S.T., Angaw D.A., Widmer R.J.. **Prevalence of hypertension and its determinants in Ethiopia: A systematic review and meta-analysis**. *PLoS One* (2020.0) **15** e0244642. PMID: 33382819
28. Whelton P.K., Carey R.M., Aronow W.S., Casey D.E., Collins K.J., Dennison Himmelfarb C., DePalma S.M., Gidding S., Jamerson K.A., Jones D.W., MacLaughlin E.J., Muntner P., Ovbiagele B., Smith S.C., Spencer C.C., Stafford R.S., Taler S.J., Thomas R.J., Williams K.A., Williamson J.D., Wright J.T.. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: a Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *Hypertension* (2018.0) **71**. DOI: 10.1161/HYP.0000000000000065
29. WHO. (2008.0)
30. WHO. A global brief on hypertension: silent killer, global public health crisis: World Health Day 2013. Accessed from: https://www.who.int/publications/i/item/a-global-brief-on-hypertension-silent-killer-global-public-health-crisis-world-health-day-2013 on Dec 1, 2021.
31. WHO. Global Health Estimates. (2020.0) 2020
32. WHO. Cardiovascular diseases (CVDs): Key facts. 2021a. Accessed from: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds) on Dec 1, 2021.
33. WHO. Noncommunicable diseases: key facts. 2021b. Accessed from: https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases on Dec 1, 2021.
34. WHO. STEPwise Approach to NCD Risk Factor Surveillance 2022. Accessed from: https://www.who.int/teams/noncommunicable-diseases/surveillance/systems-tools/steps on Aug 10, 2022.
35. Zaki N.A.M., Ambak R., Othman F., Wong N.I., Man C.S., Morad M.F.A., He F.J., MacGregor G., Palaniveloo L.. **The prevalence of hypertension among Malaysian adults and its associated risk factors: data from Malaysian Community Salt Study (MyCoSS)**. *J. Health Popul. Nutr.* (2021.0) **40** 8. DOI: 10.1186/s41043-021-00237-y
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.