text
stringlengths 254
1.16M
|
|---|
---
title: Self-organized metabotyping of obese individuals identifies clusters responding
differently to bariatric surgery
authors:
- Dimitra Lappa
- Abraham S. Meijnikman
- Kimberly A. Krautkramer
- Lisa M. Olsson
- Ömrüm Aydin
- Anne-Sophie Van Rijswijk
- Yair I. Z. Acherman
- Maurits L. De Brauw
- Valentina Tremaroli
- Louise E. Olofsson
- Annika Lundqvist
- Siv A. Hjorth
- Boyang Ji
- Victor E. A. Gerdes
- Albert K. Groen
- Thue W. Schwartz
- Max Nieuwdorp
- Fredrik Bäckhed
- Jens Nielsen
journal: PLOS ONE
year: 2023
pmcid: PMC9980823
doi: 10.1371/journal.pone.0279335
license: CC BY 4.0
---
# Self-organized metabotyping of obese individuals identifies clusters responding differently to bariatric surgery
## Abstract
Weight loss through bariatric surgery is efficient for treatment or prevention of obesity related diseases such as type 2 diabetes and cardiovascular disease. Long term weight loss response does, however, vary among patients undergoing surgery. Thus, it is difficult to identify predictive markers while most obese individuals have one or more comorbidities. To overcome such challenges, an in-depth multiple omics analyses including fasting peripheral plasma metabolome, fecal metagenome as well as liver, jejunum, and adipose tissue transcriptome were performed for 106 individuals undergoing bariatric surgery. Machine leaning was applied to explore the metabolic differences in individuals and evaluate if metabolism-based patients’ stratification is related to their weight loss responses to bariatric surgery. Using Self-Organizing Maps (SOMs) to analyze the plasma metabolome, we identified five distinct metabotypes, which were differentially enriched for KEGG pathways related to immune functions, fatty acid metabolism, protein-signaling, and obesity pathogenesis. The gut metagenome of the most heavily medicated metabotypes, treated simultaneously for multiple cardiometabolic comorbidities, was significantly enriched in Prevotella and Lactobacillus species. This unbiased stratification into SOM-defined metabotypes identified signatures for each metabolic phenotype and we found that the different metabotypes respond differently to bariatric surgery in terms of weight loss after 12 months. An integrative framework that utilizes SOMs and omics integration was developed for stratifying a heterogeneous bariatric surgery cohort. The multiple omics datasets described in this study reveal that the metabotypes are characterized by a concrete metabolic status and different responses in weight loss and adipose tissue reduction over time. Our study thus opens a path to enable patient stratification and hereby allow for improved clinical treatments.
## Introduction
Obesity is generally associated with several different comorbidities, with type 2 diabetes (T2D) and cardiovascular diseases among the most common, and cross interaction of metabolic responses from these co-morbidities makes it difficult to study metabolic alterations associated with obesity. Thus, there is an increasing interest to study heterogeneous diseases like obesity through the collection of multiple omics data from various cohorts [1–3]. Due to the heterogeneity of phenotypes within obese individuals it is, however, generally difficult to stratify cohorts into groups, e.g. individuals with or without the metabolic syndrome, that can be compared using traditional statistical methods when omics data are to be analyzed. The use of machine learning methods is therefore gaining more attention for understanding and deconvoluting multifactorial disease [4,5], in particular as it enables stratification of individuals in a given cohort, without a priori knowledge of clinical labels.
Obesity is a growing worldwide epidemic, with an estimated 1.9 billion adults being overweight and another 650 million being obese [6–8], and it is associated with increased risk of multiple comorbidities including T2D, hypertension, dyslipidemia, non-alcoholic fatty liver disease and various types of cancers [9,10]. Numerous clinical approaches have been proposed to model obesity and predict bariatric surgery outcomes, by using clinical parameters, artificial intelligence and comparing predefined patient groups [11–14]. Another clinical definition for describing individuals with multiple dysmetabolic morbidities, including obesity, is the metabolic syndrome, where obese individuals fulfill two out of these four criteria: 1) fasting glucose >100 mg/dl; 2) triacylglycerol > 150 mg/dl; 3) high-density lipoprotein(HDL) cholesterol <40 mg/dl for males and <50 mg/dl for females; 4) blood pressure above 130 systolic or 85 diastolic [15]. The multitude of co-existing metabolic perturbations may also mask associations between metabolic activities in different tissues, including the gut microbiota, hence posing a challenge in systematically studying obesity, its’ implications and the outcome of surgical intervention with higher resolution. A systems biology approach on the other hand could offer detailed phenotypic profiling possibilities using omics analysis. Metabolomics has recently been proposed as an approach to better comprehend obesity and linked comorbidities [16–18] and identify optimal candidate groups for further interventions [19,20]. The gut metagenome is a contributing factor to the complexity of obesity [21–25], although it’s causal role has yet to be established [26]. Recent studies have pinpointed that the production and regulation of metabolites of bacterial origin in humans, play an important role in metabolic diseases [24,27–30]. Given these interactions, there is a clear need to propose a systems biology framework to obesity population-based studies, to improve the identification of distinct sub-populations but also drive the development of personalized interventions [31].
With the objective of getting novel insight into how metabolism in different tissues varies in obese individuals and evaluate if grouping of patients according to metabolism is related to their weight loss response to bariatric surgery, we generated multiple omics datasets from 106 individuals undergoing bariatric surgery. Specifically, we wanted to evaluate if the heterogeneity of a bariatric surgery population can be stratified phenotypically using metabotyping, i.e. grouping according to the individuals fasting plasma metabolome, that captures the functional output of a complex multi-organ system, human hosts and their microbes rather than by traditional clinical classifiers, e.g. the metabolic syndrome. For this we established a novel workflow that first utilizes metabolomics for unlabeled stratification of individuals with several comorbidities and different pharmacological treatment regimens. We then analyzed transcriptome data from liver, jejunum, mesenteric and subcutaneous adipose tissues along with shotgun metagenomic sequencing from fecal samples to produce a discriminatory multi-marker signature of underlying metabolic phenotypes within obesity. The framework is solely based on omics data types representative of various biological molecule classes (metabolome, transcriptome, metagenome) and machine learning, instead of comorbidities, medications, and disease-specific classifiers, thus making it suitable for studying multifactorial metabolic conditions, besides obesity.
## Ethics approval and consent to participate
The study was performed in accordance with the Declaration of Helsinki and was approved by the Academic Medical Center Ethics Committee of the Amsterdam UMC. All participants provided written informed consent.
## BARIA cohort
The recruitment of participants was conducted from the BARIA [32] study with a total of 106 individuals included. The baseline characteristics of BARIA participants in the Self-Organizing Map (SOM)-defined metabotypes are described in Table 1.
**Table 1**
| Clinical Metadata | SOM Cluster 1 | SOM Cluster 2 | SOM Cluster 3 | SOM Cluster 4 | SOM Cluster 5 | BARIApopulation |
| --- | --- | --- | --- | --- | --- | --- |
| Demographic | Demographic | Demographic | Demographic | Demographic | Demographic | |
| Participants (%) | 17(16%) | 29(27.4%) | 25(23.6%) | 18(17%) | 17(16%) | 106(100%) |
| Female (% Total Participants, % of SOM Cluster) | 13(12.2%, 76.5%) | 25(23.6%, 86.2%) | 18(17%, 72%) | 14(13.2%, 77.8%) | 14(13.2%, 82.4%) | 84(79.2%) |
| Male (% Total Participants, % of SOM Cluster) | 4 (3.8%, 23.5%) | 4 (3.8%, 13.78%) | 7 (6.6%, 28%) | 4 (3.8%, 22.22%) | 3 (2.8%, 17.6%) | 22(20.8%) |
| Anthropometric | Anthropometric | Anthropometric | Anthropometric | Anthropometric | Anthropometric | |
| Age (years) | 48(29–60)* | 40(20–57)* | 53(26–64)* | 56(39–64)* | 44(22–62)* | 46(20–14) |
| BMI (kg/m2) | 39.5(34–45.4) | 38.2(32.9–60.9) | 39.8(33–57.5) | 38.3(33.8–47.1) | 39.8(34.7–46.4) | 39.42(32.9–70) |
| Waist circumference (cm) | 125.3 ± 12.6 | 122.6 ± 12.3 | 123.7 ± 11.5 | 125.8 ± 12.2 | 123 ± 9.9 | 84.3 ± 57.7 |
| Upper thigh circumference (cm) | 135(120–149) | 133(116–147) | 130(103–165) | 133(115–139) | 136(123–144) | 122.5(103–165) |
| Total Body Fat (%) | 53.6(41.6–64.7) | 54.1(31.7–94.9) | 51.8(39.3–104.8) | 56.5(40.6–78.9) | 57.6(44–64.5) | 51(31.7–104.8) |
| Fat Free Mass (kg) | 60.9(54.1–93.8) | 59.6(50.3–90.6) | 59.1(47.5–90.2) | 59.8(49.5–85.1) | 60.8(54–83.5) | 58.9(47.5–93.8) |
| Systolic blood Pressure (mmHg) | 131.5(116–156) | 132(102–155) | 133(108–161) | 136(115–193) | 135(115–157) | 132.5(102–193) |
| Diastolic blood Pressure (mmHg) | 84.5(59–91) | 81(54–99) | 82(67–105) | 80(45–121) | 82(65–94) | 81(45–121) |
| Clinical lab values | Clinical lab values | Clinical lab values | Clinical lab values | Clinical lab values | Clinical lab values | |
| Fasting glucose (mmol/l) | 5.8(4.8–11.4) | 5.9(4.6–14.8) | 5.7(5–13.8) | 5.8(4.6–6.8) | 5.6(4.5–8.7) | 5.8(4.5–14.8) |
| HbA1c (mmol/mol) | 5.7(5.3–9.1) | 5.7(4.6–9.8) | 5.6(5–9.3) | 5.8(5.2–6.9) | 5.5(5.2–8.3) | 5.7(4.6–9.8) |
| HOMA-IR | 1.7(0.6–3.4) | 1.6(0.5–6.9) | 2.2(0.5–4.7) | 1.3(0.8–4.8) | 1.5(0.8–4.8) | 1.6(0.6–6.9) |
| HOMA2-β | 108.7(38.3–183.2) | 87.9(29.1–227.8) | 112(52.7–226.2) | 92.1(52.4–357.8) | 104.2(50.8–185.5) | 93.5(29.1–357.8) |
| Total Cholesterol (mmol/l) | 5.4 ± 1.1 | 4.6 ± 1 | 4.9 ± 1.1 | 5.3 ± 1.2 | 4.3 ± 0.9 | 4.9 ± 1.1 |
| Triglycerides (mmol/l) | 1.5(0.8–3.5) | 1.3(0.6–5.8) | 1.4(0.8–6) | 1.4(0.8–5.9) | 1.2(0.6–1.9) | 1.4(0.6–6) |
| HDL Cholesterol (mmol/l) | 1.2(0.8–1.8)* | 1.1(0.6–1.9)* | 1.1(0.7–2.5)* | 1.2(0.7–2.1)* | 1.2(1–2.7)* | 1.6(0.2–2.7) |
| LDL Cholesterol (mmol/l) | 3.6 ± 1.1 | 2.9 ± 0.9 | 3.6 ± 0.9 | 3.4 ± 1.7 | 2.6 ± 0.8 | 3 ±1.1 |
| Creatinine (μmol/l) | 68(55–96) | 63(46–83) | 66(47–112) | 75(56–172) | 65(58–99) | 66(46–172) |
| Glomerular Filtration Rate (kl/1.73m2) | 85(70–91)* | 90(71–91)* | 86(62–91)* | 78(26–90)* | 89(66–91)* | 88.5(26–91) |
Individuals underwent a complete metabolic work-up at the start of their bariatric surgery trajectory. Anthropometric measurements including height, weight and waist and hip circumference were taken. In addition, body fat percentage using bioelectrical impedance and blood pressure were measured. Fasting blood samples were used for the determination of hemoglobin, HbA1c, glucose, lipid profile, alanine aminotransferase, aspartate aminotransferase, insulin, and creatinine levels. Within three months before surgery, a 2-hour mixed meal tolerance test was performed to assess insulin resistance and investigate dynamic alterations in circulating metabolites. Within three months before surgery, a 2-hour mixed meal tolerance test (MMT) was performed to assess insulin resistance and investigate dynamic alterations in circulating metabolites. The MMT consisted of a compact 125ml drink (Nutricia®) containing in total 23.3 grams fat, 74.3 grams carbohydrates (of which 38.5 grams sugar) and 24.0 grams protein. The participants received this meal after fasting for a minimum of nine hours. Time point zero refers to the moment at which the participant had fully consumed the meal. Blood samples were drawn via an intravenous line at baseline, 10, 20, 30, 60, 90 and 120 minutes. All samples were stored at -80°C until further processing.
## Metabolome analysis
EDTA plasma samples under fasting conditions were collected from 106 BARIA participants. Samples were shipped to METABOLON (Morisville, NC, USA) for performing analysis using ultra high-performance liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) untargeted metabolomics, as previously described [27]. The metabolomic counts obtained, underwent significant curation via metabolites’ pre-filtering, imputation for subsets of metabolites’ missing values and data normalization, in order to minimize the effect of artifacts in the downstream analysis. Out of 1345 metabolites measured by METABOLON, 652 metabolites were fully detected across all samples, 640 metabolites were partially detected across all samples, and 53 metabolites were not detected and therefore had a missing value. The mean number of detected peaks (absolute abundance) for the fully detected metabolites in the BARIA cohort was 52 583 199. Whereas the mean absolute abundance for fully detected metabolites was 1 817 049. Metabolomics prefiltering and imputation were performed by utilizing a variation of the Perseus platform [33]. Essentially, data has been pre-filtered so as to have a maximum of $25\%$ missing values for a metabolite across all samples. This was followed by a log transformation of all the measured metabolites’ raw intensities across the entire dataset. Then, we calculated the total data mean and standard deviation (by omitting missing values). Taking into account that the metabolite intensities distribution is approximately following normality, we chose a small distribution 2.5 standard deviations away from the original data mean towards the left tail of the original data distribution, with 0.5 standard deviations width. This new shrunken range corresponds to the actual lowest level of detection by the spectrometer. Here by drawing random values from this mini distribution, we filled the missing prefiltered data of choice.
Normalization was conducted to the total signal for each sample, since each sample is a separate injection on the mass spectrometer. Effective control for changes in sample matrix affects ionization efficiency, hence there can be inevitable differences in how much each sample is loaded onto the column with each injection, etc. Therefore, we summed up the total ion intensity (i.e. total signal) for each of the samples and identified the sample with the lowest total signal. After this we could proceed to calculating the correction factor for each sample by dividing the total signal with the lowest total signal, CorrectionFactori=TotalsignalforeachindividualsampleiLowesttotalsignalintensity. The next step is to divide each individual metabolite within a sample with the respective CorrectionFactori After imputation and normalization, we obtained 986 metabolites. All the calculations for imputing and normalizing the metabolomics dataset have been conducted with MATLAB_R2018b and the standard built-in packages.
Differential analysis was conducted among the five SOM-defined Clusters in R (version 3.6.3) and RStudio (version 1.2.5033). Statistical analysis has been performed for fasting peripheral plasma with two methods: ANOVA(Analysis of Variance) and Kruskal Wallis test, with the use of HybridMTest package [34]. HybridMTest performs hybrid multiple hypothesis testing using empirical Bayes probability. The significance level and cut-off used for the dataset of fasting peripheral plasma was $P \leq 0.05$ and was applied to metabolites that were significantly differential with both ANOVA and Kruskal Wallis methods.
## Clustering metabolome profiles with self- organizing maps
The fasting peripheral metabolomics were then input to the SOM toolbox [35] algorithmic setup in MATLAB_R2014b. SOMs conducted unsupervised competitive learning and produced low-dimensionality visualizations by employing vector quantization [36,37], a topology preserving projection. SOMs are essentially networks consisting of neurons in a lower dimensional space than the initial dataset, visually represented in a 2-dimensional grid. Each neuron has d-dimensions, equal to the number of features of the dataset and acts as a weight vector. During the SOM training phase, the weight vectors are gradually shifted in each iteration of learning, and the map gradually gets organized, so that neurons that are neighbors on the grid get similar weight vectors throughout the iterative training.
In our analysis, SOM took as input a set of prototype vectors representing the data. Every data item, here BARIA subject’s fasting metabolome, was mapped into one point (node or neuron) in the map [38]. Mapping took place throughout the training phase of the SOM. The number of nodes was calculated internally by a heuristic formula, given the number of input vectors and their dimensionality, as ∼5*N2, where N is the number of data items and the number was slightly altered in order to fit hexagonal (instead of rectangular) nodes. The training method deployed in our study was batch training, where instead of taking each input vector separately and assigning a weight vector, the dataset was given to the SOM as a whole and the new weight vectors are weighted averages of the data vectors. In order to assign the prototype vector to the node, the Euclidean distances among prototype vectors and each neuron were calculated and set as the metric for the similarity measure. The “winner” node in the grid, was the one with the smallest Euclidean distance from the input vector. Once the assignment was complete, then the weights of the prototype vector along with the weights of the subset of its spatial neighbors in the array, got updated [39,40]. This entailed that all these local re-arrangements would be propagated along the grid, during the training epochs. As a result, after learning, more similar data items would be associated with nodes that are closer in the array, whereas less similar items would be situated gradually farther away in the array.
When having a very large number of SOM nodes, one cannot easily quantify the results, hence the need for further grouping with a partitive approach. The resulting map was then subjected into k−means clustering, as a built-in function of the SOM toolbox, for obtaining a recommended partition of map nodes. An open question in this case was the number of clusters, since k−means in general takes this as a predefined parameter. Since k−means is sensitive to initialization, we ran a cross validation simulation for 100 times for each k (starting from ∼5*N2, which corresponds to the number of nodes of the neural network to 1 with step of -1) for each with different random initializations. The best partitioning for each number of clusters was selected using error criterion and the minimization of the Davies-Bouldin cluster validity index [41]. Davies-Bouldin index is a metric of the ratio of the within cluster scatter, to the between cluster separation. The index’s value is essentially the average similarity between each cluster and its most similar one, averaged over all the clusters. This implies that the best clustering scheme minimizes the Davies-Bouldin index. Eventually, when all the iteration for the potential values of k were concluded, the minimum overall Davies-Bouldin index was chosen, which resulted in the recommended partition of five clusters.
## Transcriptome analysis
Biopsies from liver (106 samples), jejunum (105 samples), mesenteric adipose fat (104 samples) and subcutaneous adipose fat (105 samples) were collected at the time of the bariatric surgery, as previously described [32]. RNA was extracted from biopsies using TriPure Isolation Reagent (Roche, Basel, Switzerland) and Lysing Matrix D, 2 mL tubes (MP Biomedical, Irvine, CA, USAs) in a FastPrep®-24 Instrument (MP Biomedical, Irvine, CA, USAs) with homogenization for 20 seconds at 4.0 m/sec, with repeated bursts until no tissue was visible; homogenates were kept on ice for 5 minutes between homogenization bursts if multiple cycles were needed. RNA was purified with chloroform (Merck, Darmstadt, Germany) in phase lock gel tubes (5PRIME) with centrifugations at 4°C, and further purified and concentrated using the RNeasy MinElute kit (Qiagen, Venlo, The Netherlands). The quality of RNA was analysed on a BioAnalyzer instrument (Agilent), with quantification on Nanodrop (Thermo Fisher Scientific, Waltham, MA, USA). Due to degradation of the RNA, libraries for RNAseq sequencing were prepared by rRNA depletion; library preparation and sequencing were performed at Novogene (Nanjing, China) on an HiSeq instrument (Illumina Inc., San Diego, CA, USA) with 150 bp paired-end reads and 10G data/sample. The average read count per sample from liver and jejunum tissues were 42 ± 15 million. For mesenteric and subcutaneous fat, the average read count per sample were 43.2 ± 20 million.
The extracted fastq files were analyzed with nf-core/rnaseq [42], a bioinformatics analysis pipeline used for RNA sequencing data. The workflow processed raw data from FastQ inputs (FastQC, TrimGalore!), aligned the reads (STAR) with Homo sapiens GRCh38 as reference genome, generates gene counts (featureCounts, StringTie) and performed extensive quality-control on the results (RSeqQC, dupRadar, Preseq, edgeR, multiQC). The pipeline was built using Nextflow.
*Differential* gene expression analysis for five SOM defined cluster participants has been performed for liver, jejunum, subcutaneous adipose and mesenteric adipose tissues, respectively, in R (version 3.6.3) and RStudio (version 1.2.5033) with DESeq2 [43] package. The statistical analysis method for calculating differential expression rates was the LRT test (log-ratio test). After FDR correction (FDR $5\%$) with multiple hypothesis testing with IHW [44] package, we analyzed genes with $P \leq 0.05$ by DEGreport’s [45] degPatterns function, so as to identify subgroups of co-expressed genes across the SOM clusters and assign a z score to each metabotype. For these differentially significant co-expressed genes we performed gene enrichment with Enrichr platform [46] using KEGG(*Kyoto encyclopedia* of genes and genomes) metabolic pathways [47]. Adjustment for the confounding factors of age and gender was conducted via the built-in function of DESeq2.
## Microbiome analysis
Fecal samples from 106 participants (108 fecal samples due to having two samples from two participants) were collected on the day of surgery and immediately frozen at -80C. Total fecal genomic DNA was extracted from 100 mg of feces using a modification of the IHMS DNA extraction protocol Q [48]. Briefly, fecal samples were extracted in Lysing Matrix E tubes (MP Biomedical, Irvine, CA, USA) containing ASL buffer (Qiagen), and lysis of cells was obtained, after homogenization by vortexing for 2 minutes, by two cycles of heating at 90°C for 10 minutes followed by three bursts of bead beating at 5.5 m/sec for 60 seconds in a FastPrep®-24 Instrument (MP Biomedical, Irvine, CA, USAs). After each bead-beating burst, samples were placed on ice for 5 minutes. The supernatants containing fecal DNA were collected after the two cycles by centrifugation at 4°C. Supernatants from the two centrifugations steps were pooled and a 600 μL aliquot from each sample was purified using the QIAamp DNA Mini kit (Qiagen, Venlo, The Netherlands) in the QIAcube (Qiagen, Venlo, The Netherlands) instrument using the procedure for human DNA analysis. Samples were eluted in 200 μL of AE buffer (10 mmol/L Tris·Cl; 0.5 mmol/L EDTA; pH 9.0). Libraries for shotgun metagenomic sequencing were prepared using a PCR-free method; library preparation and sequencing were performed at Novogene (Nanjing, China) on an HiSeq instrument (Illumina Inc., San Diego, CA, USA) with 150 bp paired-end reads and 6G data/sample.
MEDUSA [49] pipeline was used for pre-processing of raw shotgun metagenomics sequence data. MEDUSA is an integrated pipeline for analysis of short metagenomic reads, which maps reads to reference databases, combines output from several sequencing runs and manipulates tables of read counts. The input number of total reads from the metagenome analysis were on average 23.4 ± 2.2 million reads per sample and the total aligned reads 16.6 ± 1.8 million reads per sample. The sequencing runs had high quality with almost $98\%$ of the reads passing the quality cut-off (~(20 million reads per sample). Out of the high-quality reads, on average $0.04\%$ aligned to the human genome, although the data had been cleaned for human reads. Out of the high quality non-human reads, $78.4\%$ aligned to MEDUSA’s software gene catalogue. Quality filtered reads were mapped to a genome catalogue and gene catalogue using Bowtie2 [50]. Statistical analysis was performed in R (version 3.6.3) and RStudio (version 1.2.5033) on rarefied count, (20 million reads per sample). The taxon ids were input to taxize [51] package, so as to get full taxonomic information and ranking for the species. This dataset was input to DESeq2 [43] and phyloseq [52] packages for conducting downstream differential statistical analysis. Similar to the BARIA transcriptomics counts, log normalization has been conducted based on gene counts geometric distribution. Statistical analysis test for calculating differential expression rates was LRT. The IHW package, as part of DESeq2 suite, is utilized for multiple hypothesis testing and adjusting the respective P values, with alpha significance threshold set at $P \leq 0.05$ and FDR at $5\%$. Adjustment for the confounding factors of age and gender was conducted via the built-in function of DESeq2.
## DIABLO correlation analysis and biomarkers minimal signature
DIABLO [53] stands for Data Integration Analysis for Biomarker discovery using Latent cOmponents and performs supervised multi-omics data integration, by maximizing the correlation between co-expressed elements in the input datasets. DIABLO algorithm extends sparce Generalized Canonical Correlation Analysis [54] and by expanding the Partial Least Squares (PLS) regression, used singular value decomposition for dimensionality reduction and selected co-expressed (correlated) variables that could explain the categorical outcome of interest, in our case the five SOM-derived metabotypes. DIABLO analysis was conducted in R (version 3.6.3) and RStudio (version 1.2.5033) through the package of mixOmics [55] (version6.10.9). DIABLO output a set of latent variables (components) based on the dimensionality and the importance of the input datasets. All the datasets in this study carried the same weight, hence the DIABLO dataset matrix initialization design parameter was diagonal. The original input was 289 metabolites, 119 microbial species and 776 genes, all the differentially identified components from the omics datasets. This chosen number of components could extract sufficient information to discriminate all SOM-defined metabotypes. Then, a set of coefficients was attributed to each variable, that indicated the importance of each variable in DIABLOMulti-omics Datasets. The goal was to have maximization of the covariance between a linear combination of the variables from each input dataset and each categorical outcome. The algorithm was optimized with a 10-fold validation over 10 training epochs. After tuning these two hyperparameters (number of variables from each dataset, choice of variables that maximize co-variance), DIABLO produced as output a minimal signature of total 113 markers that distinguish the given metabotypes.
## Metabolomics based stratification of bariatric surgery population via SOM
To create a multi-omics profile of obesity, a total of 106 individuals from the BARIA [32] bariatric surgery cohort were recruited. The multiple omics analysis included metabolomics on fasting peripheral blood samples, and we employed this dataset for stratification of the heterogenous group of individuals, independent of traditional clinical indexes, such as Body Mass Index (BMI), hypercholesterolemia, hypertension and treatment for T2D. To enable stratification based on the metabolomics data we built an unsupervised artificial neural network that could group individuals based only on the similarity of their metabolome, a SOM. The SOM [36] evaluated metabolomic similarity by calculating the Euclidean distances between complete metabolomic profiles, and projected BARIA individuals with high inter-group similarity onto a “map” of lower dimensionality compared to that of the initial dataset. The SOM was trained with 106 prototype vectors, where each prototype vector corresponded to a BARIA participant’s peripheral plasma metabolite profile, consisting of 986 metabolites (see Methods). Iterative training of the SOM resulted in a map of 48 nodes, all projected onto a hexagonal grid (Fig 1A).
**Fig 1:** *Self-organizing maps reveal five distinct metabotypes within BARIA cohort.(A) Architecture of a competitive artificial neural network. Each individual’s complete metabolomic profile is assigned a weight. The weights are in turn assigned to neurons in the competitive layer of the neural network. In the competitive layer, SOM algorithm calculates the similarity metric (here Euclidean distance) between each metabolomic profile and the neurons and then updates the weights. After training, the network assigns the individual’s metabolomic profile to the “winner” output node, the node that is essentially more similar to the input metabolomic profile. Once this step is complete, all the nodes are comprising the SOM. Finally, all the nodes of the SOM are subjected to k-means clustering resulting in the partitioned topology, the metabotypes (SOM & k-means defined clusters). (B) Clustergram of hierarchical cluster analysis depicting the distribution of medically treated cardiometabolic comorbidities of the individuals in each of the metabotypes (SOM & k-means defined clusters). The treated comorbidities are: hypertension, T2D, GERD and cholesterol. In parallel columns are the gender and metabolic syndrome status of each individual, respectively. (C) Clinical variables associated with obesity and their statistical significance across the metabotypes (SOM & k-means defined clusters): age (C. i), BMI (C. ii), HDL cholesterol (C. iii), LDL cholesterol (C. iv), creatinine and (C. v), glomerular filtration rate (c. vi); statistical significance among metabotypes is calculated with Kruskal-Wallis test; the symbols indicating significance among metabotypes are ‘*’: P< = 0.05, ‘**’: P< = 0.01, ‘***’: P< = 0.001.*
These 48 nodes considerably reduced the dimensionality and further within-cluster variance was minimized using K−means [56], which preserved metabolomic distances and identified centroids of core metabolomes.
## SOM and k-means clustering reveal five distinct metabotypes
Clustering the SOM with k−means identified five clusters (metabotypes), each with different features (Fig 1A, Table 1), including unique distributions of comorbidities and medication usage. Polypharmacy is a notable characteristic within this study population, including use of medication for T2D ($$n = 20$$), hypertension ($$n = 30$$), hypercholesterolemia ($$n = 42$$) and gastroesophageal reflux disease (GERD, $$n = 16$$). Medication usage was distributed across the five metabotypes and is shown in Fig 1B. Clusters 1 and 3 include most individuals simultaneously treated for hypertension and high cholesterol (four and five individuals respectively), whereas cluster 2 includes individuals co-treated for hypertension, high cholesterol and T2D (four individuals). Nevertheless, the distribution of overlapping treated cardiometabolic comorbidities is quite uniform among clusters and not skewed towards a particular metabotype. In order to assess the effect of missing values and data imputation in SOM clustering, a separate mapping analysis was conducted by using the unimputed metabolomics dataset. The final map clustering did not diverge from the original prediction. Hence, the ability of the SOM to assign similar items on the same node was not affected by the imputation of a minimal set of missing metabolite values. We next assessed the biometric features of each metabotype, by performing differential analysis on the clinical variables available. BMI, body fat, and waist circumference did not significantly differ between clusters, however age, HDL cholesterol and glomerular filtration rate varied between clusters (Fig 1C). Given that all BARIA participants are affected by severe obesity, the stratification based on their SOM metabolomic profile reveals that BMI and treatment of cardiometabolic comorbidities are not the clinical features more accurately describing and differentiating the metabotypes, but age, cholesterol and markers associated with kidney function are important features. We also evaluated if there is any association within the clusters and individuals having the metabolic syndrome and found that there was no such association (Fig 1B). Furthermore, if we grouped the individuals according to having or not-having the metabolic syndrome, we also found no separation according to age or other clinical parameters besides those defining the metabolic syndrome (S1 Fig).
## Metabolomic profiles characterized by lipid and amino acid metabolites
Following stratification of the individuals into the five metabotypes we performed differential analysis of the metabolome for the five different metabotypes. Statistical analysis revealed 289 differentially significant metabolites. In comparison we only identified 3 differentially significant metabolites when the cohort was grouped according to presence of the metabolic syndrome or no (S2 Fig), which shows that driving grouping of the cohort based on the metabolomics data enables more detailed insight into metabolic differences among the individuals. KEGG pathway analysis revealed that the most highly enriched metabolite classes among the 289 metabolites were lipids, amino acids and xenobiotics, followed by cofactors and vitamins, nucleotides, carbohydrates, peptides, energy and partially characterized molecules. Clusters 2 and 3 exhibited the highest relative abundance of differentially significant metabolites, mainly lipids and amino acids (Fig 2A).
**Fig 2:** *Differentially abundant metabolites and metabolic pathways among the five defined SOM clusters (metabotypes).(A) Relative abundance and distribution of differentially significant metabolites among SOM and k-means defined clusters. Clusters two and three are most abundant in lipids (especially lysophospholipids and sphingomyelins) and amino acids (urea, arginine and proline metabolism). (B) Distribution of differentially significant metabolic pathways among SOM and k-means defined clusters, where numbers within each dot indicate how many metabolites of that particular specific pathway were differentially abundant across clusters. (C) Top 20 differentially significant metabolites among the SOM and k-means defined clusters, (P<0.05).*
Among the enriched KEGG metabolic pathways that had the highest number of differentially significant metabolites were fatty acids (Fig 2B), specifically 19 lysophospholipids, 16 dicarboxylate fatty acids, 14 sphingomyelins and 12 phosphatidylcholines. The amino acid metabolic pathways with the most significant metabolites were arginine and proline metabolism with 11 compounds, tyrosine metabolism with 8 metabolites, methionine, cysteine SAM and taurine metabolism with 8 metabolites, too, while branched-chain amino acid metabolism for isoleucine and valine had 7 metabolites. The top 20 differentially abundant metabolites are a mixture of lipids, partially characterized molecules, peptides and amino acids, and some of them, despite being the end product of endogenous ketogenesis produced by the liver, also carry the potential of being the result of gut microbial metabolism, such as 3-hydroxybutyrate and acetoacetate (Fig 2C). Our analysis identified lipid metabolites (especially lysophospholipids and sphingomyelins) and amino acid metabolites (urea, arginine and proline metabolism) being significantly altered among the clusters.
## Hepatic and adipose tissue transcriptomes enriched for immune, amino acid and lipid metabolism functions
To better understand the relationship between metabolite levels and gene expression, we next sequenced RNA extracted from biopsies taken during bariatic surgery from liver, jejunum, mesenteric adipose tissue and subcutaneous adipose tissue. We identified differentially expressed genes between clusters and conducted gene set enrichment analysis. This analysis revealed 682 hepatic genes differentially expressed across the five metabotypes. In contrast, only four genes were differentially expressed in jejunum, whereas 45 and 49 genes were differentially expressed in mesenteric and subcutaneous adipose tissue, respectively. These liver, mesenteric and subcutaneous adipose tissue gene sets were subjected to enrichment analysis for retrieving their functional profiles (S1–S4 Tables). Due to the low number of differentially expressed genes in jejunum, we were unable to obtain a gene set enrichment signature for jejunum tissue. The top represented pathways in the liver included fatty acid elongation /saturation reflecting lipids in the plasma, glycan and sphingolipid biosynthesis, cell function regulation (ErbB signaling pathway, protein export) and immune responses (Fig 3A).
**Fig 3:** *Differentially enriched KEGG metabolic pathways among the five defined SOM clusters (metabotypes).(A) Top 15 differentially enriched KEGG metabolic pathways for hepatic transcriptome among the SOM and k-means defined clusters, ranked based on their scores after differential gene expression analysis (DESeq2, P<0.05) and gene set analysis (GSA with EnrichR). (B) Top 15 differentially enriched KEGG metabolic pathways for mesenteric adipose transcriptome among the SOM and k-means defined clusters, ranked based on their scores after differential gene expression analysis (DESeq2, P<0.05) and gene set analysis (GSA) with EnrichR). (C) Top 10 differentially enriched KEGG metabolic pathways for subcutaneous adipose tissue transcriptome among the SOM and k-means defined clusters, ranked based on their scores after differential gene expression analysis (DESeq2, P<0.05) and gene set analysis (GSA with EnrichR). (D) 20 highest scoring KEGG metabolic pathways according to EnrichR GSA score for liver, mesenteric adipose and subcutaneous adipose tissues. Z score indicates different levels of differentially expressed pathways, for each SOM and k-means defined cluster.*
The mesenteric adipose tissue was enriched for amino acid metabolic processes (Fig 3B) reflecting amino acids in the plasma, and subcutaneous adipose tissue was found enriched in many pathways related to pathogens (Fig 3C) and may reflect increased immune activation associated with metabolic disease. To investigate how these pathways are regulated across the five metabotypes, we examined the normalized gene expression levels of differentially expressed genes among the clusters. The metabolic pathways enriched within the hepatic transcriptome exhibited mixed directionality in regulation and were assessed individually, for each metabotype (Fig 3D). Amino acid metabolic pathways in mesenteric adipose tissue exhibited consistent upregulation in clusters 4 and 5 (Fig 3D). Transcriptome analysis from these three tissues showed distinct regulation of lipid, amino acid, immune response and pathogenic pathways amongst the metabotypes.
## Metabotypes exhibit distinct microbial community composition
Since the gut microbiota is known to be correlated with development of comorbidities linked to obesity [57–59], we also generated a gut microbiota profile for the BARIA individuals from shotgun metagenomic sequencing of fecal DNA. Statistical analysis revealed 119 differentially abundant species among the SOM metabotypes, the top 30 of which are shown in Fig 4A and are dominated by Bacteroidetes and Firmicutes, especially Lactobacillus.
**Fig 4:** *Differentially significant microbial species and phyla among the five defined SOM clusters (metabotypes).(A) Top 30 from 119 differentially significant microbial species among the SOM & k-means defined clusters, after differential analysis with DESeq2 (P<0.05). (B) Relative abundance and distribution of differentially significant microbial species for the top 4 most abundant phyla for SOM & k-means defined clusters.*
Out of the 119 differentially abundant species, 70 belonged to Firmicutes phylum, 22 to Bacteroidetes, 11 to Actinobacteria, 11 to Proteobacteria, one to Chloroflexi, one to Cyanobacteria, one to Euryarchaeota, one to Spirochaetes and one to Fusobacteria. Within Firmicutes, Clostridia are more highly abundant for cluster 2 and Weisella for clusters 2, 4 and 5. Within Bacteroidetes, Bacteroides and Prevotella species are significantly more abundant in clusters 1, 2 and 3. For Actinobacteria, Bifidobacterium are considerably more abundant in 1 and 2, whereas species within Enterobacteriaceae family have higher abundance for clusters 4 and 5 (Fig 4B, S5 Table). In order to assess if there is a difference in alpha and beta diversity among metabotypes, we used a series of metrics (Observed, Chao1, ACE, Shannon, Simpson, Inverse Simpson for alpha diversity and Whittaker index along with dispersion analysis for beta diversity), shown in S3 and S4 Figs. Our metagenomics pipeline displayed that none of the different alpha diversity metrics reach statistical significance. The beta dispersion (with centroids) results coupled with a permutational ANOVA (PERMANOVA) analysis (for 999 permutations) showed $F = 0.19$ and $$P \leq 0.9431.$$ As seen in S4 Fig, the SOM-defined clusters largely spatially overlap but appear to have different centroids and different dispersions. Nevertheless, the large inter-individual variation cannot account for the negative PERMANOVA results, either. In such cases, there is a need to have a correct specification of the mean-variance relationship by means of multivariate extensions of GLM with methods such as negative binomials, DESeq2 [43]. The DESeq analysis revealed that despite the non-statistically significant diversities, there are SOM-defined clusters that are enriched in specific genera, such as Bacteroides, Prevotella and Lactobacillus.
As comparison, when the patients were grouped after presence or absence of metabolic syndrome, we only identified 54 differentially significant species. Similarly, none of the alpha diversity or beta diversity metrics or ordination were statistically significant. ( S5 and S6 Figs). Out of the 54 significant gut microbial species 33 species belonged to Firmicutes phylum, 10 to Bacteroidetes, six to Actinobacteria, four to Proteobacteria and one to Fusobacteria (S7 Fig, S6 Table). Within Firmicutes, there is a trend for Lactobacillus species to be two to 8 times significantly less abundant in metabolic syndrome BARIA individuals. In contrast, statistically significant *Streptococcus species* are twice as abundant in metabolic syndrome diagnosed individuals. The majority of the gut microbial species belonging to *Bacteroidetes is* two to three times depleted in metabolic syndrome diagnosed BARIA individuals, whereas Actinobacteria levels are elevated in metabolic syndrome. Differentially significant Proteobacteria tend to be depleted in metabolic syndrome.
Our analysis showed that metabolic syndrome diagnosis can indeed capture a fraction of the microbial variability within obesity. Even so, our suggested metabotyping approach can identify more gut microbial species across the spectrum of obesity and its related comorbidities.
## Individual metabotypes display unique clinical and multiple omics features
Our collective analyses show that the five different metabotypes clearly associate with unique gene expression and microbial community composition patterns and hence represents groups of individuals having distinct differences in their metabolism. To further explore these unique patterns, we next performed a detailed evaluation of the molecular fingerprints of each metabotype using the findings from the multiple omics datasets differential analysis.
17 individuals had metabotype 1 (13 women/four men), and they had the highest fat free mass 60.9 (54.1–93.8) kg and the highest total cholesterol (5.4 ± 1.1 mmol/L). Of these, four participants were treated for hypertension, three for T2D and four for GERD, whereas almost half the cluster’s population (8 participants) was treated for high cholesterol. It is noticeable that three out of four male participants were co-treated for hypertension, GERD/H. pylori infection and cholesterol (Fig 1B). Isobutyrylcarnitine was at a higher level in this metabotype (see S6 Table) compared with the other metabotypes, and the same was observed for the tyrosine metabolic pathway intermediate 4-methoxyphenol. When associating the differentially significant fasting metabolites with anthropometric features (S8 and S9 Figs), we observed negative correlations between sphingomyelins, fasting glucose (r = -0.8, $P \leq 0.001$), HbA1c (r = -0.6, $P \leq 0.01$) and age (r = -0.5, $P \leq 0.01$) specifically for this metabotype. In summary, metabotype 1 was characterized by high cholesterol, males using medication, downregulation of immune response pathways in the liver, lower abundance in Prevotella and higher abundance in Bacteroides (Fig 4B) compared to other metabotypes.
Metabotype 2 was the largest cluster consisting of 29 participants. It was female dominated (25 females/four males) and had the youngest individuals of 40 (20–57) years of average age with a BMI of 38.2 (32.9–60.9) kg/m2. The highest number of T2D individuals was noted here ($$n = 8$$), with the highest mean HbA1c value at 42 ± 12 mmol/mol. The individuals were the most heavily medicated, since it contained 11 individuals with treatment for dyslipidemia, 6 with hypertension along with the individuals affected by T2D, of which four participants were treated for all conditions simultaneously. When considering the metabolome, lysophospholipids, 1-arachidonoyl-GPC*(20:4)*, 1-linoleoyl-GPC(18:3)*, 1-linoleoyl-GPE(18:2)*, 1-oleoyl-GPE(18:1), 1-palmitoyl-GPC(16:0)* and 1-stearoyl-GPC(16:0) were higher in comparison to the majority of the clusters. Similarly, branched-chain amino acid (BCAA) metabolites 1-carboxyethylvaline, 1-carboxyethylisoleucine and valine were all at elevated levels. 3-hydroxyoleoylcarnitine, 3-hydroxydecanoate and 2-hydroxybutyrate-2-hydroxyisobutyrate were positively correlated with both glucose and HbA1c ($r = 0.5$, $P \leq 0.001$) (S8 and S9 Figs). To summarize, metabotype 2 represented the youngest individuals, yet the individuals being most heavily medicated for comorbidities. The individuals have high abundance of BCAAs and hydroxy fatty acids, even though fatty acids biosynthetic pathways were downregulated in mesenteric adipose fat. In the gut microbiome Prevotella, Bacteroides and Lactobacillus species were found to be highly abundant (Fig 4B).
There were 18 individuals having metabotype 3 and this metabotype had the highest percentage of males among the clusters (11 females/7 males). The individuals exhibited the highest HOMA2-IR at 2.2 (0.5–4.7) and the highest HOMA2-β at 112 (52.7–226.2). It included three individuals with T2D (out of 6 in total treated for T2D), 10 hypertensive (out of 13 in total treated for hypertension) and 12 individuals treated for dyslipidemia (out of 15 in total treated for high cholesterol), whereas three were treated for T2D, hypertension and dyslipidemia at the same time. Even though the anthropometrics differed, the individuals of metabotype 3 had similar metabolome and microbiome profiles as metabotype 2, but with varying transcriptome patterns. Similar to metabotype 2, lysophospholipids, 1-arachidonoyl-GPC* (20:4)*, 1-linoleoyl-GPC (18:3)*, 1-linoleoyl-GPE (18:2)*, 1-oleoyl-GPE (18:1), 1-palmitoyl-GPC (16:0)* and 1-stearoyl-GPC (16:0) were detected in equally high levels in the individuals of metabotype 3. Noticeably, all sphingomyelins were elevated for individuals in this metabotype (S6 Table). Cluster 3 appeared to be the most insulin resistant and most treated for dyslipidemia, in spite of the highly abundant metabolome in lysophospholipids and sphingomyelins. In essence, hepatic upregulation of immune responses and subcutaneous adipose tissue upregulation of pathogenic-related pathways (Fig 3D), in conjunction with high Prevotella and Lactobacillus abundance (Fig 4B) completed the cluster’s omics profile.
18 individuals had metabotype 4, including two individuals with T2D, 7 with hypertension and 8 treated for high cholesterol. The median age was the highest in this cluster compared to all others at 56 (39–64) years. Cluster 4 had the highest total body fat at 56.5 (40.6–78.9) kg. BARIA individuals stratified within this metabotype exhibited the highest creatinine at 75 (56–172) μmol/L, and lowest glomelular filtration rate at 78 (26–90) kl/1.73m2. The transcriptomics datasets from liver tissues exhibited a very strong negative regulation of cortisol synthesis, glutamatergic synapse, cGMP-PKG signaling pathway and GABAergic synapse. When focusing on the gut microbial species, individuals of metabotype 4 had many changes in the microbial composition, and the abundance of some of the species correlated with plasma glucose, low-density lipoprotein (LDL) cholesterol and cholesterol (S10 Fig). In outline, the individuals of metabotype 4 had potentially impaired kidney function, high body fat, downregulation of synaptic pathways in the liver, upregulation of fatty acid metabolic process in mesenteric adipose tissue, upregulation of pathogenic-related pathways in subcutaneous adipose tissue and increased levels of Clostridium, *Streptococcus and* Klebsiella in the gut microbial metagenome.
17 individuals had metabotype 5, with dominance of females (14 females/three males) and the median age of the individuals in this cluster was the second youngest, 44 (22–62) years. The participants were relatively treatment naïve, only four were treated for T2D, three for hypertension and three participants for dyslipidemia, with very little overlapping treatments. 1-carboxyethylvaline, 1-carboxyethylisoleucine and valine positively correlated with HbA1c ($r = 0.4$, $P \leq 0.01$) (S8 and S9 Figs). In conclusion, metabotype 5 corresponded to a relatively young cluster, with no striking comorbidity treatment, upregulated fatty acid metabolic pathways and immune response pathways in the liver and highly abundant in Citrobacter.
In order to ensure that this analysis of the multiple omics datasets is not confounded by covariates such as age and gender, we adjusted our statistical analysis to account for these two factors, in addition to our previous analysis. For the metabolome, we observed 174 metabolites being differentially significant among SOM-defined clusters, instead of 226 that we had before. In spite of that, the relative distribution of differentially significant pathways and their % pathway abundance remains identical (S11 Fig). As far as the Metabolite Specific *Pathways is* concerned, the same pathways appear to be most abundant, such as lysophospholipids, phosphatidylcholines, dicarboxylate fatty acids, and branched-chain amino acid metabolites and also maintain their distribution among the metabotypes.
In the hepatic and adipose tissue transcriptome analysis, we noticed a direct effect from the confounders. This is expected since even if the genetic making(genome) of men and women is the same, the transcriptome is distinctly dimorphic with dissimilar disease susceptibilities [60]. Pathogenic pathways along with compound degradation pathways were more dominant in the liver transcriptome enrichment, especially in clusters 1 and 2 (S12 Fig). *In* general, after the confounders removal there is a more distinct pattern of nutrients catabolism, degradation and absorption and less presence of inflammatory pathways in the hepatic transcriptome when comparing to the pathways that we obtained in our previous enrichment analysis. Surprisingly, the mesenteric adipose tissue was once more enriched for amino acid metabolic processes, in fact the same pathways as before including phenylalanine, tyrosine, tryptophan biosynthesis and fatty acids metabolism. In subcutaneous adipose tissue we also obtained a different set of enriched metabolic pathways, this time not including pathogenic pathways. Instead, we observed similar pathways as in the mesenteric adipose tissue, primarily amino acid and fatty acid metabolic pathways.
In the metagenomics dataset analysis 109 gut microbial species being differentially significant among SOM-defined clusters, instead of 288 that we had before. However, removing those confounders does not affect the overall profiling of the dataset, especially in the most dominant Phyla of Bacteroidetes and Firmicutes (S13 Fig). Even after the model adjustment, Prevotella, Bacteroides and Lactobacillus remained the most dominant species with exactly the same distribution among the SOM-defined metabotypes.
## Multi-omics integration elucidates discriminatory signature and associations between datatypes
To reveal key interactions between multi-omics data sets, we used DIABLO [53] to identify how the five metabotypes are associated with altered expression in different tissues and an altered gut microbiota. Initially, we provided the differentially abundant metabolites, genes from liver, jejunum, mesenteric and subcutaneous adipose tissue, and gut microbial species for each BARIA individual, along with their respective metabotype membership, as input to the algorithm. DIABLO simultaneously calculates the correlations among all input multiple omics datasets and selects a minimal set of input variables that differentiate the metabotypes. The computational framework used here for integrating various omics datasets successfully identified a highly correlated discriminatory signature for SOM-defined obesity phenotypes that includes multiple Prevotella species (P. veroralis, P. copri, P. multisaccharivorax, P. oulorum, P. denticola, P. sp. oral taxon 299, P. bryantii, P. melaninogenica), Intestinibacter bartlettii, Anaerococcus prevotii, lipid metabolites (especially phospatidylcholines), hepatic function associated genes, lipid metabolism and cardiomyopathy pathways, subcutaneous adipose tissue IL6 and SELE genes involved in inflammatory and immune system pathways and mesenteric adipose tissue genes enriched in prolactin signaling, T2D and PI3K-Akt signaling pathways (S14 Fig).
## Metabotypes are associated with weight loss response to bariatric surgery
In order to define the clinical value of metabotyping, we had to assess the metabotypes’ response to bariatric surgery. Hence, we performed a longitudinal biometrics post-operative control of the BARIA obese individuals at three time points: three months, six months and 12 months after surgery, where we monitored the weight, waist circumference and upper leg circumference. It is noteworthy that there are no distinct statistically significant responses in the weight loss or waist circumference reduction immediately after bariatric surgery (3 months after surgery), contrary to what would be expected (Fig 5A and 5B).
**Fig 5:** *Weight and fat loss progression at distinct time points after bariatric surgery for the five defined SOM clusters (metabotypes).(A) Weight (kg) of BARIA individuals at baseline, three months, six months and one year after bariatric surgery for each metabotype. (B) Weight loss(kg) of BARIA individuals at baseline, three months, six months and one year after bariatric surgery for each metabotype. (C) Waist circumference (cm) of BARIA individuals at baseline, three months, six months and one year after bariatric surgery for each metabotype. (D) Reduction of waist circumference(cm) of BARIA individuals at baseline, three months, six months and one year after bariatric surgery for each metabotype. (E) Upper leg circumference (cm) of BARIA individuals at baseline, three months, six months and one year after bariatric surgery for each metabotype. (F) Reduction of upper leg circumference(cm)of BARIA individuals at baseline, three months, six months and one year after bariatric surgery for each metabotype. Statistical significance among metabotypes is calculated with t-test and adjusted with FDR; the symbols indicating significance among metabotypes are ‘*’: P< = 0.05, ‘**’: P< = 0.01, ‘***’: P< = 0.001.*
There is a trend that metabotypes 2 and 5 have the highest weight loss one year post-operatively (35kg and 38 kg in average, respectively). Metabotype 2 exhibits the largest waist circumference loss at three months after surgery (12cm) even if this is not deemed statistically significant (Fig 5C and 5D). However, there is a clear pattern in the reduction of adipose tissue in the upper leg circumference. Metabotypes 1 and 5 are the best responders when it comes to upper leg circumference reduction, with the loss being consistent at all three time points. Upper leg circumference loss is significant ($P \leq 0.05$) when compared to the worst responder cluster, metabotype 3 (Fig 5E and 5F). This trend is the same for weight loss, regardless of being confirmed by statistical calculations.
Surprisingly, when the BARIA individuals were grouped according to having or not-having metabolic syndrome, there were no notable statistically significant differences in weight and adiposity loss in none of the three time points.
## Discussion
Here we present a novel unsupervised machine learning framework for stratification of individuals in human volunteer cohorts, with a high prevalence and variance of comorbidities. This framework enables a naïve to clinical labels stratification based on fasting metabolome rather than purely clinical parameters that may fail to accurately encompass the multitude of nuances in human population-based studies. The main findings of our study revealed pronounced changes in lysophospholipids, phosphatidylcholines, dicarboxylate fatty acids, sphingomyelins, and branched-chain amino acid metabolites among the five different metabotypes; KEGG metabolic pathways related to immune functions, fatty acid biosynthesis and elongation, protein signaling and pathogenic pathways were regulated in different ways for each metabotype; the abundance of Prevotella and Lactobacillus species varied the most between the metabotypes, and metabotypes 4 and 5 had a lower abundance compared to metabotypes 2 and 3. Multi-omics integration enabled reducing the dimensionality and identified a concrete biomarker signature able to differentiate between the five distinct metabotypes. The differences in metabolism among the individuals in the five metabotypes are associated with different responses in terms of weight loss and reduction of waist and upper leg circumference to bariatric surgery.
A considerable advantage of our approach is that SOM and k-means clustering effectively reduced the initial omics dimensionality and resulted in a reusable topological projection of the metabolome. Given the lack of an external multimodal multiple omics dataset for validating our results, establishing a metabolome mapping that can recognize or characterize new unknown inputs can be proven useful. New metabolomes can be projected into the same map, without the requirement of further algorithmic training. That way we can compare metabolic distances among new BARIA inclusions or even the potential post-surgical metabolomes of the initial 106 inclusions. Comparing the post-operative metabolome with the baseline pre-operative state could provide further mechanistic comprehension of the pathophysiological mechanisms of obesity and the responses to the bariatric surgery intervention in the future. Also using the multi-parameter metabolic syndrome as a classifier was here shown not to enable new insight into what drives differences in metabolism within the cohort. Metabotyping identified more gut microbial species among BARIA individuals, whereas the metabolic syndrome classification captured a fraction of the microbial variability. It has been previously attempted in animal studies to model interactions between genes, gut microbiota and the molecular mechanisms underlying obesity [61–63], but their clinical application to humans has been limited [64] so far. Increased microbial variability among metabotypes along with the results from the KEGG pathways enrichment in liver and adipose tissues could be the effect of gut microbial species in the hosts’ gene regulation. In the metabotype comparison, the statistically significant anthropometric features of age and glomerular filtration rate along with the differentially significant KEGG pathways could plausibly reflect the process of cellular and biological senescence [65,66]. The detection of senescence in the metabolome by our proposed SOM and k-means methodology, without prior knowledge of biometric characteristics strengthens our claim that the identified metabotypes stand as different representations of human metabolism among the BARIA obese individuals.
Our findings reveal that the overall effect of the metabotyping is still present in both the metabolome and the gut microbial metagenome even after regressing out confounders like age and gender. What has changed after the correction, is that metabolite compounds and gut microbial species that were less abundant in the initial dataset have now been eliminated after confounder adjustment.
On the other hand, confounding factors appear to have a more direct effect on the transcriptome, even if enrichment analysis exhibited the distinct regulation of lipid, amino acid, and pathogenic pathways amongst the metabotypes.
A limiting factor that needs to be considered when interpreting our findings is the selection of the eligible individuals for bariatric surgery. The significant variability within human cohorts is often not possible to capture in a finite number of clinical variables. For example, classifiers for obesity-associated comorbidities such as hypertension, T2D, and dyslipidemia may be treated as binary variables (present vs. absent) [9], however the overall wellness of an individual with any of these disorders can vary significantly as a function of how well managed each of these conditions are, among many other factors. The BARIA exclusion criteria for surgical interventions have to be strictly met for minimizing the risks and complications of such an evasive procedure. As a result, the BARIA inclusions might not fully represent the obesity spectrum. Many of differentially significant metabolites that were identified by our pipelines are directly implicated in inflammatory pathways. A clinical inflammatory marker, such as C-Reactive Protein (CRP) would be very valuable for confirming this observation. However, CRP at baseline did not exhibit any overall statistical significance among the SOM-defined clusters and was not available in the one year follow up (S15 Fig). There is a visible trend in weight loss and leg/waist circumference reduction among the SOM and k-means defined clusters over time. Nonetheless the statistically significant differences among all the identified SOM clusters were not conclusive, probably due to statistical power. Despite the 106 BARIA inclusions and the high quality of the omics dataset, each cluster contains 17–29 individuals, which might account for the values of the statistically significant results.
## Conclusions
The principal contribution of this study is the detailed omics dataset for obese individuals, that includes metabolome, microbiome and especially transcriptome from multiple tissues. Our findings suggested that participants’ stratification based on metabotyping could enhance our ability to get molecular insights into the causes of diseases from multi-omics integrative analysis. The combination of SOM metabotyping and DIABLO correlation analysis highlights the data-driven nature of this approach. DIABLO analysis enabled the identification of an underlying common yet discriminatory minimal multi-omics signature for the SOM-defined metabotypes, that could lead to predictive markers of the bariatric surgery outcome. In this light, use of biologic parameters such as the plasma metabolome, as a direct readout of the overall status of an entire multiorgan system host and its microbiome, to determine grouping of individuals, offers a unique approach that may more accurately classify individuals into distinct disease physiological states [11,12,67]. Rather than traditional clinical disease classifiers, this grouping method may reduce the confounding effects of such clinical metadata [68,69]. The multiple omics dataset’s association framework can be the starting point for selecting candidate compounds for a more thorough examination and provide mechanistic insight into the causality of pathogenicity originating in the tissues, mediated by bacteria and materializing via metabolites and clinical metadata. The multi-omics integrative framework implemented could also be utilized as a hypothesis generating tool for comprehending cardiometabolic disease. Our data suggest that self-organized metabotyping, based only on metabolite distribution, with no other prior knowledge on the individuals’ clinical status in combination with DIABLO integrative analysis, constitute a valuable computational approach studying multifaceted metabolic disorders.
## References
1. McLendon R, Friedman A, Bigner D, Van Meir EG, Brat DJ, Mastrogianakis GM. **Comprehensive genomic characterization defines human glioblastoma genes and core pathways**. *Nature* (2008.0) **455** 1061-8. DOI: 10.1038/nature07385
2. Manichanh C, Borruel N, Casellas F, Guarner F. **The gut microbiota**. *IBD. Nat Rev Gastroenterol Hepatol* (2012.0) **9** 599-608. PMID: 22907164
3. Dao MC, Clément K. **Gut microbiota and obesity: Concepts relevant to clinical care**. *Eur J Intern Med* (2018.0) **48** 18-24. DOI: 10.1016/j.ejim.2017.10.005
4. Vanamala JKP, Knight R, Spector TD. **Can Your Microbiome Tell You What to Eat**. *Cell Metab* (2015.0) **22** 960-1. DOI: 10.1016/j.cmet.2015.11.009
5. Wilmanski T, Rappaport N, Earls JC, Magis AT, Manor O, Lovejoy J. **Blood metabolome predicts gut microbiome α-diversity in humans**. *Nat Biotechnol* (2019.0) **37** 1217-28. DOI: 10.1038/s41587-019-0233-9
6. 6Obesity and overweight. [cited 2020 Jul 29]. Available from: https://www.who.int/en/news-room/fact-sheets/detail/obesity-and-overweight.
7. **WHO. Obesity: preventing and managing the global epidemic**. *WHO Technical Report Series number 894* (2000.0)
8. Bentham J, Di Cesare M, Bilano V, Bixby H, Zhou B, Stevens GA. **Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128·9 million children, adolescents, and adults**. *Lancet* (2017.0) **390** 2627-42. PMID: 29029897
9. Engin A., Engin AB, Engin A. **The Definition and Prevalence of Obesity and Metabolic Syndrome**. *Obesity and Lipotoxicity* (2017.0) 1-17. DOI: 10.1007/978-3-319-48382-5_1
10. Haslam DW, James WPT. **Obesity.**. *Lancet* (2005.0) **366** 1197-209. DOI: 10.1016/S0140-6736(05)67483-1
11. Ahlqvist E, Storm P, Käräjämäki A, Martinell M, Dorkhan M, Carlsson A. **Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables**. *Lancet Diabetes Endocrinol* (2018.0) **6** 361-9. DOI: 10.1016/S2213-8587(18)30051-2
12. Wang TJ. **Risk Prediction in Cardiovascular Medicine Assessing the Role of Circulating, Genetic, and Imaging Biomarkers in Cardiovascular Risk Prediction**. *Circulation* (2011.0) **123** 551-65. PMID: 21300963
13. Huang PL. **A comprehensive definition for metabolic syndrome**. *Dis Model Mech* (2009.0) **2** 231-7. DOI: 10.1242/dmm.001180
14. Bektaş M, Reiber BMM, Costa J, George P, Donald LB. **Artificial Intelligence in Bariatric Surgery: Current Status and Future Perspectives.**. *Obes Surg* (2022.0) **32** 2772-2738. DOI: 10.1007/s11695-022-06146-1
15. Francisco SS, Education C, Adult P. **World Health Organization–Defined Metabolic Syndrome Is a Better Predictor of Coronary Calcium Than the Adult Treatment Panel III Criteria in American**. *Diabetes Care* (2004.0) **12** 2977-9. DOI: 10.2337/diacare.27.12.2977
16. Fiamoncini J, Rundle M, Gibbons H, Thomas EL, Geillinger-k K, Bunzel D. **Plasma metabolome analysis identifies distinct human metabotypes in the postprandial state with different susceptibility to weight loss–mediated metabolic improvements**. *FASEB J* (2018.0) **32** 5447-58. DOI: 10.1096/fj.201800330R
17. Riedl A, Gieger C, Hauner H, Daniel H, Linseisen J, Landstr I. **Metabotyping and its application in targeted nutrition: an overview**. *Br J Nutr* (2017.0) **12** 1631-44. DOI: 10.1017/S0007114517001611
18. Lacruz ME, Kluttig A, Tiller D, Medenwald D, Giegling I, Rujescu D. **Instability of personal human metabotype is linked to all-cause mortality.**. *Sci Rep* (2018.0) **8** 9810. DOI: 10.1038/s41598-018-27958-1
19. Riedl A, Wawro N, Gieger C, Meisinger C, Peters A, Roden M. **Identification of Comprehensive Metabotypes Associated with Cardiometabolic Diseases in the Population-Based KORA Study.**. *Mol Nutr Food Res* (2018.0) **62** 180017. DOI: 10.1002/mnfr.201800117
20. Cirulli ET, Guo L, Leon Swisher C, Shah N, Huang L, Napier LA. **Profound Perturbation of the Metabolome in Obesity Is Associated with Health Risk**. *Cell Metab* (2019.0) **29** 488-500.e2. DOI: 10.1016/j.cmet.2018.09.022
21. Karlsson FH, Tremaroli V, Nookaew I, Bergström G, Behre CJ, Fagerberg B. **Gut metagenome in European women with normal, impaired and diabetic glucose control**. *Nature* (2013.0) **498** 99-103. DOI: 10.1038/nature12198
22. Sommer F, Bäckhed F. **The gut microbiota-masters of host development and physiology**. *Nat Rev Microbiol* (2013.0) **11** 227-38. DOI: 10.1038/nrmicro2974
23. Deschasaux M, Bouter KE, Prodan A, Levin E, Groen AK, Herrema H. **Depicting the composition of gut microbiota in a population with varied ethnic origins but shared geography**. *Nature Medicine* (2018.0) **24** 1526-31. DOI: 10.1038/s41591-018-0160-1
24. Mardinoglu A, Shoaie S, Bergentall M, Ghaffari P, Zhang C, Larsson E. **The gut microbiota modulates host amino acid and glutathione metabolism in mice**. *Mol Syst Biol* (2015.0) **11** 834-834. DOI: 10.15252/msb.20156487
25. Schroeder BO, Bäckhed F. **Signals from the gut microbiota to distant organs in physiology and disease**. *Nat Med* (2016.0) **22** 1079-89. DOI: 10.1038/nm.4185
26. Bouter KE, van Raalte DH, Groen AK, Nieuwdorp M. **Role of the Gut Microbiome in the Pathogenesis of Obesity and Obesity-Related Metabolic Dysfunction**. *Gastroenterology* (2017.0) **152** 1671-8. DOI: 10.1053/j.gastro.2016.12.048
27. Koh A, Molinaro A, Ståhlman M, Khan MT, Schmidt C, Mannerås-Holm L. **Microbially Produced Imidazole Propionate Impairs Insulin Signaling through mTORC1**. *Cell* (2018.0) **175** 947-961.e17. DOI: 10.1016/j.cell.2018.09.055
28. Savolainen O, Lind MV, Bergström G, Fagerberg B, Sandberg AS, Ross A. **Biomarkers of food intake and nutrient status are associated with glucose tolerance status and development of type 2 diabetes in older Swedish women**. *Am J Clin Nutr* (2017.0) **106** 1302-10. DOI: 10.3945/ajcn.117.152850
29. Sun L, Xie C, Wang G, Wu Y, Wu Q, Wang X. **Gut microbiota and intestinal FXR mediate the clinical benefits of metformin**. *Nat Med* (2018.0) **24** 1919-29. DOI: 10.1038/s41591-018-0222-4
30. Schiattarella GG, Sannino A, Toscano E, Giugliano G, Gargiulo G, Franzone A. **Gut microbe-generated metabolite trimethylamine-N-oxide as cardiovascular risk biomarker: A systematic review and dose-response meta-analysis**. *Eur Heart J* (2017.0) **38** 2948-56. DOI: 10.1093/eurheartj/ehx342
31. Nielsen J.. **Systems Biology of Metabolism: A Driver for Developing Personalized and Precision Medicine.**. *Cell Metab* (2017.0) **25** 572-9. DOI: 10.1016/j.cmet.2017.02.002
32. Van Olden C, Van de Laar A, Meijnikman A, Aydin O, Van Olst N, Hoozemans JB. **A Systems Biology approach to understand gut microbiota and host metabolism in morbid obesity: design of the BARIA Longitudinal Cohort Study**. *J Intern Med* (2020.0) 0-2. DOI: 10.1111/joim.13157
33. Tyanova S, Temu T, Sinitcyn P, Carlson A, Hein MY, Geiger T. **The Perseus computational platform for comprehensive analysis of (prote) omics data**. *Nat Methods* (2016.0) **13** 731-40. DOI: 10.1038/nmeth.3901
34. Stan Pounds DF. *HybridMTest: Hybrid Multiple Testing* (2019.0)
35. Vesanto J.. *SOM Toolbox.*
36. Kohonen T.. **Self-organized formation of topologically correct feature maps**. *Biol Cybern* (1982.0) **43** 59-69
37. Kohonen TK. *MATLAB Implementations and Applications of the Self-Organizing Map* (2014.0)
38. Kohonen T.. *Springer Series in Information Sciences* (1995.0) **30**
39. Kohonen T.. *Springer Series in Information Sciences* (1989.0) **8** 119-57
40. Liu Y, Weisberg RH, Mooers CNK. **Performance evaluation of the self-organizing map for feature extraction**. *J Geophys Res Ocean* (2006.0) **111** 1-14
41. Davies D.L.. **BDW. A cluster separation measure**. *IEEE Trans Pattern Anal Mach Intell* **PAMI-1** 224-22
42. Ewels PA, Peltzer A, Fillinger S, Patel H, Alneberg J, Wilm A. **The nf-core framework for community-curated bioinformatics pipelines**. *Nat Biotechnol* (2020.0) **38** 276-8. DOI: 10.1038/s41587-020-0439-x
43. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol* (2014.0) **15** 550. DOI: 10.1186/s13059-014-0550-8
44. Ignatiadis N, Klaus B, Zaugg JB, Huber W. *Data-driven hypothesis weighting increases detection power in genome- scale multiple testing* (2016.0) **13**
45. Pantano L.. *DEGreport: Report of DEG analysis* (2019.0)
46. Chen EY, Tan CM, Kou Y, Duan Q, Wang Z, Meirelles GV. **Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool**. *BMC Bioinformatics* (2013.0) **14** 128. DOI: 10.1186/1471-2105-14-128
47. Kanehisa M, Goto S. **KEGG: Kyoto Encyclopedia of Genes and Genomes**. *Nucleic Acids Res* (2000.0) **28** 27-30. DOI: 10.1093/nar/28.1.27
48. Costea PI, Zeller G, Sunagawa S, Pelletier E, Alberti A, Levenez F. **Towards standards for human fecal sample processing in metagenomic studies**. *Nat Biotechnol* (2017.0) **35** 1069-76. DOI: 10.1038/nbt.3960
49. Karlsson FH, Nookaew I, Nielsen J. **Metagenomic Data Utilization and Analysis (MEDUSA) and Construction of a Global Gut Microbial Gene Catalogue.**. *PLoS Comput Biol* (2014.0) **10**. DOI: 10.1371/journal.pcbi.1003706
50. Langmead B, Salzberg SL. **Fast gapped-read alignment with Bowtie 2**. *Nat Methods* (2012.0) **9** 357-9. DOI: 10.1038/nmeth.1923
51. Chamberlain SA, Szöcs E. **taxize: taxonomic search and retrieval**. *R. F1000Research* (2013.0) **2** 191. PMID: 24555091
52. Mcmurdie PJ, Holmes S. **phyloseq: An R Package for Reproducible Interactive Analysis and Graphics of Microbiome Census Data**. *PLoS One* (2013.0) **8** e61217. DOI: 10.1371/journal.pone.0061217
53. Singh A, Shannon CP, Rohart F, Tebbutt SJ. **Le K anh. Systems biology DIABLO: an integrative approach for identifying key molecular drivers from multi-omics assays**. *Bioinformatics* (2019.0) **35** 3055-62. PMID: 30657866
54. Tenenhaus A, Guillemot V, Grill J, Frouin V. *Variable selection for generalized canonical correlation analysis* (2014.0) 569-83
55. Singh A, Cao K anh L, Rohart F. **mixOmics: An R package for ‘ omics feature selection and multiple data integration**. *PLoS Comput Biol* (2017.0) **13** 1-19. DOI: 10.1371/journal.pcbi.1005752
56. Lloyd SP. **Least Squares Quantization in PCM**. *IEEE Trans Inf Theory* (1982.0) **28** 129-37
57. Wu H, Tremaroli V, Schmidt C, Lundqvist A, Olsson LM, Krämer M. **The Gut Microbiota in Prediabetes and Diabetes: A Population-Based Cross-Sectional Study.**. *Cell Metab* (2020.0) **32** 379-390. DOI: 10.1016/j.cmet.2020.06.011
58. Sze MA, Schloss PD. **Looking for a signal in the noise: Revisiting obesity and the microbiome**. *MBio* (2016.0) **7** 1-10. DOI: 10.1128/mBio.01018-16
59. Thingholm LB, Rühlemann MC, Koch M, Fuqua B, Laucke G, Boehm R. **Obese Individuals with and without Type 2 Diabetes Show Different Gut Microbial Functional Capacity and Composition**. *Cell Host Microbe* (2019.0) **26** 252-264.e10. DOI: 10.1016/j.chom.2019.07.004
60. Gershoni M, Pietrokovski S. **The landscape of sex-differential transcriptome and its consequent selection in human adults.**. *BMC Biol* (2017.0) **15** 1-15. PMID: 28100223
61. Kleinert M, Clemmensen C, Hofmann SM, Moore MC, Renner S, Woods SC. *Nature Reviews Endocrinology* (2018.0) **14** 140-62
62. Renner S, Blutke A, Clauss S, Deeg CA, Kemter E, Merkus D. *Cell and Tissue Research* (2020.0) **380** 341-78. PMID: 31932949
63. Koya D, Kanasaki K. **Biology of obesity: Lessons from animal models of obesity.**. *J Biomed Biotechnol* (2011.0). DOI: 10.1155/2011/197636
64. Yazdi FT, Clee SM, Meyre D. **Obesity genetics in mouse and human: Back and forth, and back again**. *PeerJ* (2015.0) **2015** 1-69
65. Childs BG, Durik M, Baker DJ, Van Deursen JM. **Cellular senescence in aging and age-related disease: From mechanisms to therapy**. *Nat Med* (2015.0) **21** 1424-35. DOI: 10.1038/nm.4000
66. Collado M, Blasco MA, Serrano M. **Review Cellular Senescence in Cancer and Aging**. *Cell* (2007.0) **130** 223-33. PMID: 17662938
67. Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee EP, Mccabe E. **Metabolite profiles and the risk of developing diabetes**. *Nat Med* (2011.0) **17** 448-54. DOI: 10.1038/nm.2307
68. Holmes E, Wilson ID, Nicholson JK. **Forum Metabolic Phenotyping in Health and Disease**. *Cell* (2008.0) **134** 714-7. PMID: 18775301
69. Nicholson JK, Holmes E, Kinross J, Burcelin R, Gibson G, Jia W. **Host-gut Metabolic Interactions.**. *Science (80-).* (2012.0) **336** 1262-8
|
---
title: 'Michigan men’s diabetes project II: Protocol for peer-led diabetes self-management
education and long-term support in Black men'
authors:
- Jaclynn Hawkins
- Srijani Sengupta
- Katherine Kloss
- Katie Kurnick
- Alana Ewen
- Robin Nwawkwo
- Martha Funnell
- Jamie Mitchell
- Lenette Jones
- Gretchen Piatt
journal: PLOS ONE
year: 2023
pmcid: PMC9980828
doi: 10.1371/journal.pone.0277733
license: CC BY 4.0
---
# Michigan men’s diabetes project II: Protocol for peer-led diabetes self-management education and long-term support in Black men
## Abstract
Previous literature has indicated that Black men are twice as likely to develop type 2 diabetes compared to their non-Hispanic White counterparts and are also more likely to have associated complications. Furthermore, Black men have lower access to quality health care, and masculinity norms have been shown to hinder them from seeking the limited care that is available. In this study, we aim to investigate the effect of peer-led diabetes self-management education and long-term ongoing support on glycemic management. The first phase of our study will consist of modification of existing diabetes education content to be more appropriate for the population of interest, Then, in the second phase, we will conduct a randomized controlled trial to test the intervention. Participants randomized to the intervention arm will receive diabetes self-management education, structured diabetes self-management support, and a more flexible ongoing support period. Participants randomized to the control arm will receive diabetes self-management education. Diabetes self-management education will be taught by certified diabetes care and education specialists, while the diabetes self-management support and ongoing support period will be facilitated by fellow Black men with diabetes who will be trained in group facilitation, patient-provider communication strategies, and empowerment techniques. The third phase of this study will consist of post-intervention interviews and dissemination of findings to the academic community. The primary goal of our study is to determine whether long-term peer-led support groups in conjunction with diabetes self-management education are a promising solution to improve self-management behaviors and decrease A1C levels. We will also evaluate the retention of participants throughout the study, which has historically been an issue in clinical studies focused on the Black male population. Finally, the results from this trial will determine whether we can proceed to a fully-powered R01 trial or if other modifications of the intervention are necessary.
Trial registration: Registered at ClinicalTrials.gov with an ID of NCT05370781 on May 12, 2022.
## Background and rationale
In the United States, more than 37.3 million Americans, or $11.3\%$ of the population, are living with diabetes [1]. Disparities in diabetes prevalence and complications also exist among racial/ethnic groups [2]. For instance, Black people are twice as likely to have type 2 diabetes (T2D) compared to non-Hispanic White people [2]. Black men have worsened glycemic management (e.g. checking blood glucose levels daily) compared to non-Hispanic White men, their risk for T2D complications is higher and they are more likely to die earlier from these complications [2, 3].
Extant research in health disparities shows how poverty and race can influence the health and wellbeing trajectories of Black men [2]. For instance, T2D has its most damaging effects within urban low-income communities of color, and yet they continue to lack access to quality health care [2]. While low-income Black men have similar rates of T2D compared to Black women, Black men are disproportionately impacted with respect to severity and mortality [2]. In addition to being less likely to follow diabetes self-management plans and engage in diabetes self-management behaviors (e.g. checking blood glucose levels daily), low-income Black men face a multitude of additional challenges to maintaining a healthy diet, such as limited availability of healthy foods [2–5]. Black men also often experience barriers that interfere with diabetes management behaviors such as a lack of social support, negative patient-provider relationships, cost and long work hours [6, 7]. Though existing studies find that Black men are at elevated risk for suboptimal diabetes management, they are also less likely to participate in research studies that test diabetes self-management interventions [8]. Additionally, Black male participants are more likely to drop out of diabetes self-management intervention trials before their completion, resulting in additional problems [8].
There is a new emphasis in recent research on the critical role of gender in the management of health behaviors. This body of work highlights how male gender norms can conflict with both healthy practices and engagement in healthcare services [2, 9]. For instance, among men, the need to exhibit toughness, confidence, suppress emotions while exuding independence and control can pose a barrier to engagement with care (i.e. following treatment plans) [9]. Existing studies also reveal how male gender norms can impede utilization of critical social support offered by family and community networks, including the acceptance of compassionate environments (i.e. emotional support) [9].
Peer leaders (PLs) are trained lay individuals used to advance diabetes self-management in minority communities through encouraging and helping with goal setting, problem solving, and social, emotional, and peer support [10–13]. They are chosen based on their cultural background and community connections. The use of PLs is an essential component of interventions aimed at improving diabetes health outcomes among low-income Black communities [10–13]. Yet, despite the documented effectiveness of peer-led diabetes interventions on bettering diabetes-related health behaviors and health outcomes, this model has largely not been adapted to meet the needs of Black men. It is also important to note that among studies of peer leader models, participants and peers are largely women [8]. An opportunity exists to tailor messaging and intervention content to better serve Black men living with type 2 diabetes.
While there is an absence of research that utilizes Black male peer leaders (i.e., male lay helpers) in support of Black men with diabetes, male lay helpers have provided Black men with health education and emotional support in other areas of health care; including interventions for the prevention of chronic illness, such as for HIV and hypertension [14, 15]. Gender-specific programming utilizing male leaders to reduce stigma related to health care seeking has also been recommended by Black men [16]. More studies are needed to establish the efficacy of these approaches and to determine whether the use of male PLs can improve health outcomes among Black men with diabetes.
Research also suggests that tailoring diabetes self-management education and support (DSME/S) interventions to address the needs of Black men can play an instrumental role in improving their health outcomes [7, 17]. The National Standards for Diabetes Care and the National Standards for Diabetes Self-Management Education and Support highlight the relevance of providing initial DSME and on-going DSMS that is individualized to assist people with diabetes in maintaining effective self-management, aiming to improve health equity across different populations [18]. DSMS interventions, including those delivered by professionals and peer leaders, bettered A1C and self-management outcomes in comparison to control groups in previous studies [19–22]. Piatt et al. studied the use of a 12 session, 15-month intervention that combined DSME directed by a diabetes educator with DSMS delivered by peer leaders (PL) in 9 Black community churches in metro-Detroit [23]. The intervention completion rate was high ($96.6\%$) [23]. Twelve peer leaders ($$n = 12$$) were trained in how to facilitate goal setting, skills development, and group cohesion [23]. After the professional DSME was provided, 6 monthly DSMS groups were facilitated by 12 PL (mean age: 61.9 years, $100\%$ Black, $25\%$ male), followed by a supplemental 6 months of further support to assess the logistical feasibility of sustaining DSMS efforts [23].
As previously stated, there is a paucity of diabetes intervention research that focuses on the unique needs of Black men. Peer-led education and support interventions have been successful, but Black men have been poorly recruited and retained in these studies [8]. Discourses on gender equity in health often focus on a binary between men and women, which often ignore the structural and systemic issues that sub-groups are required to navigate. This limited framing of gender equity does not acknowledge the complex health and social inequities faced by marginalized groups of men, particularly those relating to race, age, socioeconomic status, geography and disability. Therefore, a more concerted and nuanced focus on health equity in the field of men’s health is needed. In the context of our proposed project, we seek to create T2D programming that addresses the intersection of gender and markers of marginalization (e.g. race/ethnicity; income). It is our expectation to provide evidence for the effectiveness of ongoing peer-led DSME/S intervention for Black men with T2D. The findings will facilitate understanding of barriers and facilitators to implementing this community-based approach and the impact this initiative has on the individual and community levels.
We propose to tailor an existing peer intervention by 1) using male peers as interventionists and 2) modifying the intervention content to focus on messaging appropriate for men. Our study will then assess the effectiveness of the adapted, peer-led DSMS and ongoing support period. Success of this study sets the stage for larger implementation trials and adoption of evidence-based interventions to address foundations of health disparities for Black men living with type 2 diabetes.
## Objectives
This study’s objectives are twofold: 1) in collaboration with Black men with type 2 diabetes, we will adapt an evidence-based peer leader intervention designed to improve diabetes self-management to better fit the needs of this population; and 2) conduct a pilot randomized controlled trial (RCT) to evaluate participant recruitment and retention rates, treatment and intervention satisfaction and estimate intervention effect sizes on our primary outcomes self-management behaviors and glycemic control measured through hemoglobin A1C (A1C) as well as on secondary outcomes such as diabetes social support at baseline, 3-, 9- and 15-months. Our previous study, Michigan Men’s Diabetes Project (MenD) [24], measured health outcomes for Black men with T2D who participated in peer-led support group interventions. Based on preliminary findings from MenD, the present study (hereafter referred to as MenD II) will be modified to include a 6-month period of group based ongoing support in addition to an added emphasis on patient-provider communication strategies. Below we describe the protocol for MenD II.
## Study design
This proposed study, MenD II, includes a developmental phase (adjustment of the intervention, including stakeholder feedback, followed by feasibility testing with Black men), a validation phase (RCT), and a dissemination phase. The dissemination phase includes: post-intervention interviews/focus groups, and sharing of findings with academics and community members.
## Phase I: Intervention adaptation
n this study, we will adapt the existing intervention from previous peer leader research and the MenD study for Black men with type 2 diabetes to add additional components, with the goal to better meet the needs of Black men living with diabetes [23]. During Phase I we will conduct 14 interviews to better understand factors that impact the implementation of a peer-led diabetes self-management education support (PLDSMS) intervention adapted for Black men with T2D in the US. Interviewees will be diabetes educators, community partners, participants from previous peer leader studies, peer leaders from previous studies, and researchers in the fields of diabetes and/or men’s health and physical health. These individuals will be familiar with working with lay health workers, experienced with our target population, and/or knowledgeable about the research topic. Phase I of this project was approved by the University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board (HUM00200496) on June 6, 2021. The interviews in phase I will be guided by the Tailored Implementation in Chronic Disease (TICD) framework, a comprehensive framework describing the determinants of implementation success [25]. Determinants of practice are the barriers and facilitators that might impact on implementation of an intervention. The TICD Checklist includes 57 potential determinants of practice grouped into seven domains [25]. These seven domains include guideline factors, individual health professional factors, patient factors, professional interactions, incentives and resources, capacity for organizational change, and social, political and legal factors [25]. The aim of the interviews is to understand the structural, contextual and cultural factors impacting the implementation of a peer-led diabetes education and support intervention adapted for Black men with T2D. We will use data from these interviews to inform the implementation of the intervention in Phase II, a tailored PLDSMS RCT. At this point, participant recruitment and educational materials will be adapted for gender. It is essential that intervention materials are adapted to account for language, culture, health literacy and urban poverty contexts.
Standardized treatment goals and the DSMS protocol will be adapted to include local references, cultural values and familiar situations. Though many changes cannot be predicted, based on the existing literature, our study team will develop a DSMS module that will emphasize teaching male participants how to engage in clinic appointments with their diabetes and other health care providers (i.e. sharing concerns, being assertive to advocate for needs, and asking questions).
## Phase II: RCT
During Phase II, an RCT will be conducted of the modified intervention to measure participant recruitment and retention rates, treatment and intervention satisfaction and estimate intervention effect sizes on our primary outcomes (i.e., A1C and self-management behaviors) as well as on secondary outcomes (i.e., diabetes social support and diabetes-related distress). Phase II of this study was approved by the University of Michigan Medical School Institutional Review Board (HUM00200469) on November 11, 2021.
The RCT will be conducted among $$n = 64$$ Black male adult residents of southeastern Michigan. Four men will be recruited and trained to function as peer leaders, with sixty men randomized to a control group or to the tailored PLDSMS (See Fig 1). Our study team hypothesizes that 1) participants in the adapted PLDSMS approach will have overall improved outcomes over the control group, and 2) participants in the PLDSMS will learn diabetes self-management skills at greater levels than participants in the control group.
**Fig 1:** *RCT enrollment, intervention, and assessments.*
To measure growth, the primary outcomes will be change in A1C levels and self-management behaviors [26–28]. Secondary outcomes measured will include BMI, blood pressure, quality of life, depressive symptoms, empowerment, social support, barriers and resources, and adherence to gender norms [29–35]. Primary and secondary outcomes will be measured at the baseline visit and at each follow-up visit.
## Phase III: Dissemination
Phase III will consist of three parts; administering post-intervention interviews and focus groups with both participants and stakeholders, statistical analysis, and dissemination of findings. Activities to share results will include presentation of findings to academic and community forums. In academia, research findings will be presented to the annual meeting of the American Diabetes Association, with the preparation of at least 3 manuscripts. To share results in the community, community-based dissemination of findings will include a report and presentation that will be shared electronically with researchers, community-supported, translational health research that are focused on recruiting and working with African American men in diabetes research. A toolkit will be provided to a local senior center in southeast Michigan and will contain community resources, and data collection tools to allow for continuation of learning, and functions staff need to play to sustain improvements in outcomes.
## Study setting
Due to the risks associated with COVID-19, all DSME/S and ongoing support sessions will be held virtually via a Health Insurance Portability and Accountability Act (HIPAA) compliant virtual platform, Zoom for Health at U-M [36]. Post-treatment and follow-up self-report questionnaires will be completed in person, telephonically, or virtually. All biometrics will be collected in person at a local church or at a local senior center. To facilitate privacy, confidentiality, and safety in each location, rooms with doors, and telephones, will be available. In person health assessments will follow up to date local, state, and national COVID-19 protocols and recommendations. The data for this study will be collected at baseline, 3 months (DSME completion), 6 months (DSMS completion), and 15 months (ongoing support completion).
## Inclusion criteria
To be eligible for the study, participants ($$n = 64$$) will be Black/African American men ages 21 years or older who have had a diagnosis of type 2 diabetes for a six-month period or longer. Previous studies done by this team focused on older Black men who were at least 55. The range of participants has been expanded in order to increase recruitment and allow analysis of age differences in our sample population. Additionally, it is essential that participants have transportation to attend program activities, be under the care of a physician addressing their diabetes, and be willing to attend both group delivered virtual sessions and in-person health assessments.
Four of the participants will serve as peer leaders and will attend 30 hours of training to learn skills needed to facilitate DSMS. In addition to the general inclusion criteria listed for participants, the peer leaders must also have at least an 8th grade education, have had type 2 diabetes for over one year, be actively working on their own self-management goals, and be ready to go through the training and be a peer leader.
This study team contemplated restricting eligibility to a higher-risk population of participants with A1C ≥ $8\%$. A majority of the preliminary data suggest that over $50\%$ of the proposed study sample will have an A1C ≥ $8\%$. By focusing on all Black men with diabetes, it allows us to cast a wide net for secondary prevention and public health impact. Potential participants who meet eligibility criteria will be invited to partake in the baseline screening assessment. While the above eligibility criteria have been chosen based on previous studies, adjustments will be made to the future, larger trial, based on results and feedback from our proposed study.
## Exclusion criteria
The exclusion criteria for this study includes 1) non-ambulatory, serious health conditions (i.e., severe symptomatic heart disease, visual impairment, renal failure, and peripheral neuropathy); 2) psychiatric illness (i.e., severity requiring hospitalization); and 3) serious diabetes complications (i.e., blindness) that would interfere with meaningful participation in the groups.
## Intervention
During Phase II of this study, an 18-month randomized control trial (RCT) will take place and include: 1) diabetes self-management virtual education and support (DSME/S) with peer leaders to assess participant recruitment and retention rates, treatment and intervention satisfaction and estimate trial effect sizes on our primary outcomes of self-management behaviors; and 2) the estimate on secondary outcomes such as glycemic control (A1C), diabetes social support, diabetes-related distress, and adherence to gender norms. To measure outcomes, these data will be collected at baseline, 3-months (DSME completion), 6-months (DSMS completion), and 15-months (ongoing support completion) at a senior center in southeast Michigan, a church in southeast Michigan, telephonically, or virtually. In the Phase II RCT, we will individually randomize enrolled participants using a $\frac{50}{50}$ randomization scheme to either the peer led DSMS or a control group. All DSME/S sessions will be held virtually on Zoom. At the time this protocol was submitted, three peer leaders were recruited, enrolled and completed training and participant recruitment had begun.
## Intervention (peer leader DSMS n = 30)
The participants randomized to the peer-led DSMS group will receive 10-hours of virtual diabetes self-management education (DSME) over the course of 3 months. The 10-hours was chosen to align with what is reimbursable through Medicare for initial diabetes self-management training [37]. A certified diabetes care and education specialists (CDCES) will deliver DSME with two peer leaders co-facilitating one 15-person group. A CDCES facilitating DSME classes ensures consistent and accurate delivery of the diabetes content while covering all required and necessary self-management topics. Research shows that Black men with diabetes experience significant barriers to healthcare, so we are offering DSME with the assumption that most participants have not received formal DSME.
These participants will then transition into six 90-min monthly virtual PLDSMS sessions tailored to Black men with T2D. PLs will facilitate DSMS with the oversight of the CDCES. Though the CDCES will not be present in the PLDMS, the participants will meet with the CDCES once each month to answer questions and get support for the next DSMS session(s) as needed. Additionally, the CDCES will be available via telephone to answer any clinical questions the PLs need support with. In previous studies, we observed that PLs were most effective and confident when they had ongoing support and assistance to uphold their efforts in the areas of clinical content, educational methods, group facilitation, and communication skills. The patient-directed, DSMS session content is standardized around 6 core processes: 1) reflecting on relevant self-management experiences, 2) discussing emotions, 3) problem-solving barriers to diabetes management, 4) addressing questions about diabetes, 5) setting behavioral goals and 6) discussing patient provider communication strategies [24]. During DSME, participants will also be given a guidebook titled, “Diabetes Management Guidebook,” which is culturally specific and literacy-appropriate. The guidebook was initially developed for previous projects and was well received [38].
Upon the completion of DSMS, participants in the PLDSMS group will begin a 6-month period of ongoing support. In this stage, participants will be encouraged to foster ongoing DSMS through programs and initiatives that are meaningful to them. On-going support will build on the peer-led component of this study and encourage participants to engage in the activities of their choice (i.e. forming a walking group, discussing self-management topics, cooking classes, etc.) to improve health outcomes. Peer leaders will not be compensated for this period to assess the logistical feasibility of sustaining DSMS efforts after the study is complete.
## Control group (control n = 30)
Participants randomized to the control group will engage in 10 hours of group-delivered virtual DSME provided by a CDCES. These participants will not receive any DSMS or ongoing support from PLs and the PLs will not engage in the DSME sessions. The main purpose of the trial includes evaluation of the recruitment and retention rates, intervention satisfaction, and estimated intervention effect sizes. Taking this into consideration, the enhanced usual care condition was chosen as the control in order to 1) ensure that any intervention effects are not due to provision of diabetes education alone, 2) minimize ethical concerns regarding assignment of underserved populations to receive a no-treatment control and 3) control for improvements due to attention and positive regard and expectancies for improvement due to participation in treatment (i.e., Hawthorne effects).
## Peer leaders (n = 4)
Candidates recruited to be PLs will participate in PL training (detailed below), co-facilitate DSME sessions with a CDCES, lead DSMS sessions, and complete the same health assessments as all other participants in the study. After the peer leader training is complete, quarterly leader review sessions will be held with the PLs and CDCES so PLs can support one another, ask questions, and further hone group facilitation skills. The number of these sessions will be based on the needs of the PLs and determined in collaboration with the CDCES and the PLs. PLs will be offered $10/h for the training, leader review sessions, DSME sessions, and DSMS sessions. PLs will not be paid during the final ongoing support period to simulate the real-world scenario of facilitating support after the research study has concluded.
## Primary outcome measures
Primary outcomes for this study include glycemic control and regimen adherence (See Table 1). The primary measurement used to determine metabolic control will be A1C. We will use a DCA Vantage point-of-care testing instrument to measure A1C. This analyzer is capable of measuring hemoglobin A1C with non-fasting fingerstick in 6 minutes. To measure regimen adherence and self-management behaviors, the Perceived Diabetes Self-Management Scale, the Adherence to Refills and Medicines Scale for Diabetes (ARMS-D), and the Diabetes Care Profile will be utilized [26, 33–34].
**Table 1**
| Anthropometric Data and Clinical Data | Anthropometric Data and Clinical Data.1 | Anthropometric Data and Clinical Data.2 |
| --- | --- | --- |
| Measure | Collection Time | Source of Measure |
| Hemoglobin A1c (%) | Baseline, 3, 9, 15 | Primary data collection |
| Height (inches) and weight (lbs) | Baseline, 3, 9, 15 | Primary data collection |
| BP (mmHg) | Baseline, 3, 9, 15 | Primary data collection |
| Medication use | Baseline, 3, 9, 15 | Primary data collection |
| Survey Data | | |
| Measure | Collection Time | Source of Measure |
| Sociodemographic characteristics, healthcare utilization, comorbidities | Baseline, 3, 9, 15 | Diabetes Care Profile [26] |
| Diabetes quality of life, General quality of lie | Baseline, 3, 9, 15 | Type 2 Diabetes Distress Assessment System [27], 12-Item Short Form Survey (SF-12) [28] |
| Depressive symptom severity | Baseline, 3, 9, 15 | Patient Health Questionnaire (PHQ-9) [29] |
| Empowerment | Baseline, 3, 9, 15 | Diabetes Empowerment Scale Short Form [30] |
| Perceived social support | Baseline, 3, 9, 15 | Social Support Questionnaire [31] |
| Self-care barriers and resources | Baseline, 3, 9, 15 | Chronic Illness Resources Survey [32] |
| Self-care behaviors | Baseline, 3, 9, 15 | Perceived Diabetes Self-Management [33], Adherence to Refills and Medicines Scale for Diabetes (ARMS-D) [34] |
| Masculinity norms scale | Baseline, 3, 9, 15 | Conformity to Masculine Inventory (CMNI-30) [35] |
## Secondary outcome measures
To measure adherence to gender norms, the Conformity to Masculine Inventory (CMNI-30) will be utilized [35]. BMI will be calculated using height and weight, with height being measured using a stadiometer and weight being measured using a calibrated digital scale. Blood pressure will be measured 3 times using the oscillometric technique and a final blood pressure will be calculated by taking the average of these measures.
Perceived social support will be measured using the Social Support questionnaire [31], depressive symptom severity will be measured using the Patient Health Questionnaire (PHQ-9) [29], the Type 2 Diabetes Distress Assessment System [28] will be utilized to determine diabetes-related distress and general quality of life will be measured with the 12-Item Short Form Survey (SF-12) [28]. Empowerment will be measured using the Diabetes Empowerment Scale Short Form [30]. The Chronic Illness Resources Survey will be completed to indicate barriers and resources to self-care [32]. As an additional measurement, participants will complete questionnaires that assess socio-demographic, behavioral, psychosocial, and openness to health services utilization and are supported in diverse populations with diabetes (See Table 1).
## Participant timeline
For this study, data will be collected at baseline, completion of DSME (3-months), completion of DSMS (9-months), and completion of ongoing support period (15-months) (See Table 1).
## Power analysis
While we plan to recruit 64 Black men to be our participants for this study, we expect a final sample size of 48 participants (excluding the four participants who will serve as peer leaders), assuming a $20\%$ attrition rate. There will be 12 participants per group, with 2 groups in the control arm and 2 groups in the PLDSMS arm. Assuming correlations of 0.25 between successive measurements of A1C, this sample size will yield a power of 0.8 to detect a difference of 0.6 standard deviations between average values of A1C in the treatment and control groups.
## Peer leader recruitment, training, and assessment
Peer leaders will be recruited from the interviews in Phase I, from participants in previous studies who indicated interest in becoming a PL, or by staff at the senior center. After the peer leaders are selected, they will receive about 30 hours of training in facilitation skills, coping strategies, and empowerment-based communication skills. The training sessions will be spread over 3 months in order to minimize burden on the PLs, and they will be compensated at $10/hr. The research staff will work with the PLs to ensure the timing of the training sessions is suitable for all parties. The curriculum and materials for the training will be based on Piatt et al. due to the $100\%$ attendance rate of the 15 PLs in the study [23]. The content will be modified to include specific concerns voiced by Phase I interviewees. The training process will model alternative perspectives that allow for healthy behaviors to be framed as competence and strength, rather than as a challenge to their masculinity.
Peer leaders will be trained in a group setting, and the curriculum will include both didactic learning and skill building with a specific focus on providing support to adults with type 2 diabetes. The training will be led by two CDCESs who helped develop and implement similar training curricula in previous projects [19]. During the training, the PLs will have many opportunities to practice their group facilitation skills. The CDCES and fellow peer leaders will provide constructive and specific feedback to support the PL in enhancing their skill set.
At the conclusion of the training period, the PLs will be given written assessments and a process-based evaluation to assess their understanding of the concepts and skills taught. Post training measures will also be in place to further assess goal-setting, communication, and facilitation abilities.
We will hold quarterly meetings with all the peer leaders to offer additional opportunities to hone skills, share experiences, and offer support to one another. The peer leaders will be encouraged to exchange information and support each other outside these sessions as well.
At the time this protocol was submitted, peer leader recruitment and training had been completed.
## Participant recruitment
We plan to take a multifaceted approach to participant recruitment. Our first strategy is to utilize paper and electronic advertisements at community partner sites. We will hang flyers describing the study at the senior center and church where biometric data will be collected. These two places will be sent electronic copies of the flyer and they may choose to include it in newsletters. Additionally, the Healthier Black Elder Center will include a recruitment posting in their newsletter and email listings [39].
The second strategy is to use two research registries to reach out to potential participants that meet our inclusion criteria. The first is the Michigan Center for Urban African American Aging Research Participant Research Pool (MCUAAAR PRP) through the Healthier Black Elder Center, a volunteer registry of Black males who are at least 55 years old that is accessible to researchers to find candidates for their studies. Our previous studies have successfully utilized MCUAAAR PRP as a recruitment tool. There are currently a total of 1080 men on the registry, with 339 of them known to have type 2 diabetes [40]. The second registry is the University of Michigan Data Office for Clinical and Translational Research DataDirect PHI system [41]. Both of these registries will provide our research team with names, phone numbers, and email addresses of men who meet our inclusion criteria. Men from these registries will be emailed and/or called to gauge interest and to sign up for enrollment.
Our third strategy is snowball sampling by providing the flyer to a variety of different people including participants of a previous DSMES study, participants who were interviewed during the Phase I of this study, and other researchers. Those who are given the flyer will be encouraged to share it with people in their personal and professional networks who may benefit from the study.
Men interested in participating in the study will either call a central office phone number or be identified through cold calls and/or recruitment emails. If the men express interest in the study, they will be screened to ensure they meet the inclusion criteria previously stated. Upon completion of the phone screen, the study team member will inform the individual if they are eligible to complete the baseline interview. If the individual is eligible, a baseline assessment will be scheduled, and details will be given to the individual.
At the time this protocol was submitted, participant recruitment had begun but no participants were enrolled.
## Retention
Our team has had high recruitment and retention rates among Black men in our target recruitment region in previous studies. A former diabetes-focused qualitative research study conducted by our team retained 30 out of 32 men. Specific procedures based on our previous work will be implemented to minimize participant attrition. To increase retention: 1) Non-biometric data collection sessions will be completed telephonically or virtually in order to maximize the convenience of data collection for each participant, and 2) Advanced scheduling, multiple reminder letters, and phone reminders will be utilized to encourage participants to attend their in-person appointments.
## Allocation
Participants will be randomized to the peer led diabetes self-management support (PLDSMS) group or a control group using a $\frac{50}{50}$ randomization scheme. A computer will create a random number sequence, and then the randomization feature in Research Electronic Data Capture (REDCap) will be used to assign participants [42].
After an individual provides informed written consent at the baseline visit, a study coordinator without prior knowledge of the sequence will randomize the participant via REDCap, allocate the individual to the selected group, and inform the participant of their assigned group.
## Blinding
Blinding is not possible for this study because the nature of the intervention requires active participation in each phase. Participants will be notified of their assigned group and given earned incentives accordingly.
## Data collection procedures
All research staff will be trained in standardized measurement and data collection techniques. Quantitative and qualitative assessments will occur for all participants at baseline (0-months), completion of DSME (3-months), completion of DSMS (9-months), and completion of ongoing support period (15-months). All participants, including peer leaders will be invited to a 60-minute appointment at a local senior center or church in the metro-*Detroit area* for physiological testing. Self-report questionnaires will be completed telephonically or virtually, depending on each participant’s preference. After completing each assessment, participants will be compensated $50. Participants who withdraw from the study will still be invited to participate in data collection appointments at each timepoint. All data will be stored in REDCap, a HIPAA-compliant, web-based application provided by the University of Michigan [41], and in a Microsoft Access database [43].
## Data management
Participants will be assigned a study identification number after they enroll which will be used on all materials and data for the study’s remainder. Confidentiality is of utmost importance, and our participants will be assured of their anonymity. Participants will not be named in any reports regarding the study data and the data collected will only be used for research purposes. The principal investigator will retain control of all data collected, including questionnaires, audiotapes, and transcribed notes. Data will be stored in a locked file drawer in a locked office, and only the research team will have access.
## Phase I and Phase III
Qualitative analysis will be conducted on the interviews and focus group discussions from Phases I and III. Interviews and discussions will be recorded and transcribed, and we will develop codes utilizing a grounded theory approach. The research team will formulate a coding manual and definitions based on the text which will guide ongoing coding. Pairs of coders will read the transcripts and codes that achieve $80\%$ agreement on code application will be kept. Analyses will be done using Atlas.ti. The findings from Phase I will be used to further refine intervention content as needed. Post-intervention interviews and focus group discussions from Phase III will be evaluated to determine the effectiveness of the intervention and to identify any domain-specific issues.
## Phase II
With a conservative estimate of a $20\%$ overall attrition rate, our final sample would be 48 participants (excluding peer leaders), given our initial target sample size of 64 Black men. This will result in 24 participants in each group, which is similar to other pilot intervention trials. One goal of our intervention is to have a low dropout rate. Our past studies have retained 90–$95\%$ of participants, and for this study, an intervention retention rate of $75\%$ in the PLDSMS arm will be considered sufficient to proceed to the R01 [23, 44]. There will be no interim data analysis. We plan to use the adapted PLDSMS intervention to power a future R01 based on a minimally clinical important difference (i.e., $0.5\%$ decrease in A1C) rather than based on specific effect sizes from this trial. If we find that the PLDSMS intervention resulted in a small effect on the primary and secondary endpoints, this would suggest that further refinement is needed before proceeding to a fully powered trial.
Our goal is to disseminate findings from the current study and detect between group differences. We will assess statistical significance for the intervention group by time interaction using an Individually Randomized Group-Treatment Trial, with intervention group as a between-subjects factor (2 levels), repeated measurements over time as a within subjects factor (3 levels), a within-subject correlation of.5, and an α of.05, and a non-sphericity correction of.75. The effect size used in this design is Cohen’s f (f = σmeans/σ). The effect of the intervention on diabetes management and A1C (primary outcomes) will be assessed using mixed-effects models. Diabetes management and A1C at baseline (0 months), completion of DSME (3 months), completion of DSMS (9 months), and completion of ongoing support period (15 months) will be used as the dependent variables. Independent variables include intervention group, time of assessment, and interaction between time and intervention group. To take into account correlation between observations, random intercepts and slopes will be included into the model. In the analysis we will follow intention-to-treat principles. Analyses will be conducted with adjusting for stratifying variables (i.e. age group).
## Formal committee
A Data Safety and Monitoring Board (DSMB) will be established to protect the safety of all participants, and to ensure the integrity of the data collected. We will follow techniques suggested in previous literature to establish this board, which will consist of the Principal Investigator (PI), research team members, and five other people who are not involved with the study [45]. These outside members will all be University professors with proficiency in clinical diabetes intervention research, and they will serve as voting members of the board. The DSMB will hold at least three conference calls over the award period for approximately two hours per call to discuss the progress of the intervention and review research results, if applicable. At the first meeting, the board will elect a chair of the DSMB. To facilitate these conference calls, the principal investigator will prepare a report on the progress of the project to date. This report will be circulated well in advance of the conference call to allow all members ample time to read it. These calls will be scheduled and organized by the study coordinator and will be held at a time convenient for all members.
## Safety/Harms
Participants will be asked to provide finger stick capillary blood samples during assessments to measure A1C levels so that we can assess changes throughout the study. These blood draws have risks similar to any other routine blood sample collection including: minor discomfort, minor pain, bruising, and bleeding at the puncture site. Skilled personnel will perform all blood draws and sample collection to mitigate associated risks.
Participants will also be asked to complete a series of psychosocial questionnaires at each assessment. All of the questionnaires are standardized measures that have been used in our own trials and in other diabetes research. There are no significant risks anticipated related to their completion. However, answering questions in the PHQ-9, the demographic questionnaire, the Diabetes Quality of Life measure, the SF-12, and others may cause minor discomfort for participants [26–35]. In order to alleviate this, breaks will be given to reduce fatigue, and research assistants will be trained to obtain personal information in a sensitive fashion. All participants will be informed that they may discontinue completion of individual items, questionnaires, or the study protocol at any time. Participants who experience significant discomfort will have the option of meeting with the PI. Although prompting patients to review their diabetes care practices and providing them with feedback about their health may cause some emotional discomfort or anxiety, this discomfort could prime patients and their primary care physician to address any problems identified, thus ultimately benefiting the participants.
Another risk of our study is breach of confidentiality of data. All research staff will be trained in research ethics, confidentiality protection, and HIPAA prior to and throughout the study period through the Human Research Protection Program (HRPP) in the University of Michigan Office of Research (UMOR) [46]. Peer leaders will also be trained in the protection of participant confidentiality and must pass standardized testing before interacting with the other participants. Participants will be assigned a study identification number during enrollment that will be used in all study materials and data. Identifiable information will be kept separately from the data and laboratory values. The master list containing participants’ names and study identification numbers will be kept in a locked filing cabinet in the School of Social Work accessible only to the research team. Audio/visual recordings of interviews and group discussions will be stored securely on the University of Michigan Dropbox [47], a HIPAA compliant data storage system, and will only be accessible to study personnel. After the study is completed, these recordings will promptly be deleted. If a participant’s A1C is over $14\%$, systolic blood pressure value is 200 mm Hg or greater, and/or diastolic blood pressure value is 100 mm Hg or greater, both the patient and their provider will be alerted as this poses a risk to the patient and requires immediate attention. All other reports will not identify individual participants. No persons from the community-based locations utilized in the study will have access to personal health information or participant survey data.
For the purposes of this study, adverse events will be considered any undesirable sign, symptom, or medical condition occurring during the study, whether or not related to the intervention. Adverse events include new events not present during the training period or events that were present during the training period but increased in severity over time. All members of the research team will be trained on identifying an adverse event and instructed to report them to the principal investigator immediately. Additionally, study participants will be provided with the phone number of the project coordinator to report any adverse events that occur during the duration of the study. Each adverse event will be recorded and assessed for its date of onset, duration, severity, seriousness, and relationship to study treatment, and any action/treatment that is required. All adverse events will be collected, analyzed, and monitored using an adverse event form. Our team will verify appropriate reporting of adverse events throughout the study.
## Auditing
To ensure PLs have the skills to facilitate DSMS, the CDCES that was involved in the PL training and also facilitated the DSME classes, will meet with the PLs after each DSME session. During this time the CDCES and PLs will review a skills assessment frequently used during the PL training to assess the PLs skills during that session. This assessment will also cover effective and ineffective interactions during the DSME session to support the PL in increasing their group facilitation skills and confidence before they start DSMS sessions. The DSMS and ongoing support periods are intended to provide a comfortable, confidential environment where the participants can talk to their peers alone. However, to ensure treatment fidelity, three DSMS sessions will be selected at random and recorded and evaluated by our research team.
## Discussion
Successful management of diabetes requires both diabetes education and social support that is appropriate for each patient’s lifestyle and environment [2]. The current set up of our healthcare system is unable to provide Black men living with diabetes the resources and support they need [2]. This project aims to investigate whether the use of gender-matched lay helpers in diabetes self-management education and long-term support is a promising solution to improve the health of this high risk population.
Although literature on community-based diabetes interventions with lay helpers is available, these studies have predominantly focused on non-hispanic white, middle-class populations, and among the studies that have focused on Black populations, women dominate as participants and peer-leaders [8]. The results of these interventions have been compelling, which indicates that a similar use of PLDSMS with Black men may be beneficial. Black men who have participated in previous intervention studies have also voiced that they would be most comfortable with Black males as their lay helpers [16]. Our study will utilize lay helpers during DSME and DSMS who match participants in both gender and race in order to encourage participation and retention throughout the intervention. The intervention proposed also includes an ongoing support period where peer leaders are not compensated and can choose their method of interacting with participants to assess the real-world feasibility of long-term DSMS. We believe using Black male peers as leaders and adapting the intervention content has the potential to improve diabetes outcomes by reinforcing proper self-management behaviors.
The data from this study will enhance diabetes intervention literature by contributing the experience of Black men in an urban setting using an interdisciplinary treatment approach. We will also identify strategies for increasing Black male participation and retention in research studies, as this has proven a challenge in previous literature [8]. Lastly, this pilot RCT will allow us to refine recruitment strategies, training materials, and the implementation protocol to be used in a larger cluster RCT. Data from this study will be disseminated to academics and the sample population community to sustain improvements in diabetes outcomes and advance health equity.
## Ethics
This study protocol outlined in Fig 1 has been reviewed and approved by the University of Michigan Health Sciences Center Institutional Review Board (Phase I: HUM00200496, Phase II and III: HUM00200469). All individuals interested in participation will be required to provide a written informed consent document approved by the University of Michigan Institutional Review Board (UM IRB). At the baseline screening assessment session, eligible participants will complete the Informed Consent form. Consent forms will include all required elements of informed consent, including purpose of the study, duration, voluntary participation, alternatives and right to withdraw. Participants will be told that they will be compensated for each study assessment and that the intervention will be provided to them at no cost. In addition, the consent form will discuss the fact that participants have an equal chance of being randomized to either treatment condition. Participants will be provided with a copy of the informed consent form for their records.
## References
1. 1American Diabetes Association [Internet]. 2022
July
28 [cited 2022 Oct 10]. Arlington: The American Diabetes Association. Available from: https://diabetes.org/about-us/statistics/about-diabetes.. *American Diabetes Association* (2022.0)
2. Hawkins J.. **Type 2 diabetes self-management in non-Hispanic Black men: A current state of the literature**. *Curr Diab Rep* (2019.0) **19** 10. DOI: 10.1007/s11892-019-1131-8
3. Liburd LC, Namageyo-Funa A, Jack L. **Understanding “masculinity” and the challenges of managing type-2 diabetes among African-American men.**. *J Natl Med Assoc* (2007.0) **99** 550-552. PMID: 17534013
4. Lee LT, Jung SE, Bowen PG, Clay OJ, Locher JL, Cherrington AL. **Understanding the dietary habits of black men with diabetes**. *J Nurse Pract* (2019.0) **15** 365-369. DOI: 10.1016/j.nurpra.2018.12.023
5. Thorpe RJ, Kennedy-Hendricks A, Griffith DM, Bruce MA, Coa K, Bell CN. **Race, social and environmental conditions, and health behaviors in men.**. *Fam Community Health* (2015.0) **38** 297-306. DOI: 10.1097/FCH.0000000000000078
6. Hawkins J, Mitchell J, Piatt G, Ellis D. **Older African American Men’s perspectives on factors that influence type 2 diabetes (T2D) self management and peer-led interventions.**. *Geriatrics* (2018.0) **3** 38-47. DOI: 10.3390/geriatrics3030038
7. Crabtree K, Sherrer N, Rushton T, Willig AL, Agne AA, Shelton T RN. **Diabetes connect: African American men’s preferences for a community-based diabetes management program**. *Diabetes Educ* (2015.0) **41** 118-126. DOI: 10.1177/0145721714557043
8. Sherman L, Hawkins J, Bonner T. **An analysis of the recruitment and participation of African American men in type 2 diabetes self-management research: A review of the published literature**. *Soc Work Public Health* (2016.0) **32** 38-48. DOI: 10.1080/19371918.2016.1188742
9. Hawkins J, Watkins DC, Kieffer E, Spencer MS, Nicklett EJ, Piatt G. **An exploratory study of gender identity and its influence on health behavior among African American and Latino men with type 2 diabetes**. *Am J Mens Health* (2016.0) **11** 344-356. DOI: 10.1177/1557988316681125
10. Heisler M.. **Overview of peer support models to improve diabetes self management and clinical outcomes**. *Diabetes Spectr* (2007.0) **20** 214-221. DOI: 10.2337/diaspect.20.4.214
11. Heisler M.. **Different models to mobilize peer support to improve diabetes self management and clinical outcomes: evidence, logistics, evaluation considerations and needs for future research**. *Fam Pract* (2010.0) **27** 23-32. DOI: 10.1093/fampra/cmp003
12. Tang TS, Funnell MM, Gillard M, Nwankwo R, Heisler M. **The development of a pilot training program for peer leaders in diabetes**. *Diabetes Educ* (2011.0) **37** 67-77. PMID: 21220362
13. Tang TS, Nwankwo R, Whiten Y, Oney C. **Training peers to deliver a church based diabetes prevention program**. *Diabetes Educ* (2012.0) **38** 519-525. DOI: 10.1177/0145721712447982
14. Maulsby C, Millett G, Lindsey K, Kelley R, Johnson K, Montoya D. **A systematic review of HIV interventions for black men who have sex with men (MSM).**. *BMC Public Health* (2013.0) **13** 1. PMID: 23280303
15. Hess PL, Reingold JS, Jones J, Fellman MA, Knowles P, Ravenell JE. **Barbershops as hypertension detection, referral, and follow-up centers for black men**. *Hypertension* (2007.0) **49** 1040-1046. DOI: 10.1161/HYPERTENSIONAHA.106.080432
16. Hurt TR, Seawell AH, O’Connor MC. **Developing effective diabetes programming for black men**. *Glob Qual Nurs Res* (2015.0) **2** 233339361561057. DOI: 10.1177/2333393615610576
17. Hawkins J, Kieffer EC, Sinco B, Spencer M, Anderson M, Rosland AM. **Does gender influence participation? Predictors of participation in a community health worker diabetes management intervention with African American and Latino adults**. *Diabetes Educ* (2013.0) **39** 647-654. DOI: 10.1177/0145721713492569
18. Davis J, Fischl AH, Beck J, Browning L, Carter A, Condon JE. **2022 National Standards for Diabetes Self-Management Education and Support**. *Diabetes Care* (2022.0) **45** 484-494. DOI: 10.2337/dc21-2396
19. Funnell MM, Tang TS, Anderson RM. **From DSME to DSMS: developing empowerment based self management support**. *Diabetes Spectr* (2007.0) **20** 221-226. DOI: 10.2337/diaspect.20.4.221
20. Norris SL, Chowdhury FM, Van Le K, Horsley T, Brownstein JN, Zhang X. **Effectiveness of community health Workers in the Care of persons with diabetes**. *Diabet Med* (2006.0) **23** 544-556. DOI: 10.1111/j.1464-5491.2006.01845.x
21. Thompson JR, Horton C, Flores C. **Advancing diabetes self management in the Mexican-American population: a community health worker model in a primary care setting**. *Diabetes Educ* (2007.0) **33** 159S-65S. DOI: 10.1177/0145721707304
22. Brown HS, Wilson KJ, Pagan JA, Arcari CM, Martinez M, Smith K. **Cost effectiveness analysis of a community health worker intervention for low-income hispanic adults with diabetes**. *Prev Chron Dis* (2012.0) **9** 120074. DOI: 10.5888/pcd9.120074
23. Koscielniak NJ, Funnell M, Piatt G. **Building infrastructure for diabetes self management support in church-based settings—results of a 15-month cluster-randomized controlled trial**. *Diabetes* (2018.0) **67** 871. DOI: 10.2337/db18-871-P
24. Hawkins J., Kloss K., Funnell M.. **Michigan Men’s diabetes project (MenD): protocol for a peer leader diabetes self-management education and support intervention.**. *BMC Public Health* (2021.0) **21** 562. DOI: 10.1186/s12889-021-10613-2
25. Flottorp SA, Oxman AD, Krause J, Musila NR, Wensing M, Godycki-Cwirko M. **A checklist for identifying determinants of practice: A systematic review and synthesis of frameworks and taxonomies of factors that prevent or enable improvements in healthcare professional practice**. *Implementation Science* (2013.0) **8** 35. DOI: 10.1186/1748-5908-8-35
26. Wallston KA, Rothman RL, Cherrington A. **Psychometric properties of the perceived diabetes self-management scale (PDSMS).**. *J Behav Med* (2007.0) **30** 395-401. DOI: 10.1007/s10865-007-9110-y
27. Mayberry LS, Gonzalez JS, Wallston KA, Kripalani S, Osborn CY. **The ARMS-D outperforms the SDSCA, but both are reliable, valid, and predict glycemic control**. *Diabetes Research and Clinical Practice* (2013.0) **102** 96-104. DOI: 10.1016/j.diabres.2013.09.010
28. Fitzgerald JT, Davis WK, Connell CM, Hess GE, Funnell MM, Hiss RG. **Development and validation of the diabetes care profile**. *Evaluation & the Health Professions* (1996.0) **19** 208-230. DOI: 10.1177/016327879601900205
29. Polonsky WH, Fisher L, Hessler D, Desai U, King SB, Perez-Nieves M. **Toward a more comprehensive understanding of the emotional side of type 2 diabetes: A re-envisioning of the assessment of diabetes distress**. *Journal of Diabetes and its Complications* (2022.0) **36** 108103. DOI: 10.1016/j.jdiacomp.2021.108103
30. Jenkinson C, Layte R. **Development and testing of the UK SF-12.**. *J Health Serv Res Policy* (1997.0) **2** 14-18. DOI: 10.1177/135581969700200105
31. Kroenke K, Spitzer RL. **The PHQ-9: a new depression diagnostic and severity measure.**. *Psychiatr Ann.* (2002.0) **32** 509-15. DOI: 10.3928/0048-5713-20020901-06
32. Anderson RM, Funnell MM, Fitzgerald JT, Marrero DG. **The diabetes empowerment scale. A measure of psychosocial self-efficacy**. *Diabetes Care* (2000.0) **23** 739-743. DOI: 10.2337/diacare.23.6.739
33. Epino HM, Rich ML, Kaigamba F, Hakizamungu M, Socci AR, Bagiruwigize E. **Reliability and construct validity of three health-related self-report scales in HIV-positive adults in rural Rwanda**. *AIDS Care* (2012.0) **24** 1576-1583. DOI: 10.1080/09540121.2012.661840
34. Glasgow RE, Toobert DJ, Barrera M, Strycker LA. **The Chronic Illness Resources Survey: Cross-validation and sensitivity to intervention**. *Health Education Research* (2005.0) **20** 402-409. DOI: 10.1093/her/cyg140
35. Levant RF, McDermott R, Parent MC, Alshabani N, Mahalik JR, Hammer JH. **Development and evaluation of a new short form of the Conformity to Masculine Norms Inventory (CMNI-30).**. *Journal of Counseling Psychology* (2020.0) **67** 622-636. DOI: 10.1037/cou0000414
36. 36Zoom at U-M [Internet]. (no date) [Cited 2022 Oct 15]. Ann Arbor: Information and Technology Services. Available from: https://its.umich.edu/communication/videoconferencing/zoom.
37. 37Medicare reimbursement guidelines for DSMT. 2021
Feb
4 [Cited 2022 Oct 15]. Atlanta: Centers for Disease Control and Prevention. Available from: https://www.cdc.gov/diabetes/dsmes-toolkit/reimbursement/medicare.html.. *Medicare reimbursement guidelines for DSMT* (2021.0)
38. Tang TS, Gillard ML, Funnell MM, Nwankwo R, Parker E, Spurlock D. **Developing a new generation of ongoing diabetes self-management support interventions**. *The Diabetes Educator* (2005.0) **31** 91-97. DOI: 10.1177/0145721704273231
39. 39Healthier black elders center(Hbec). (no date) [Cited 2022 Oct 15]. Detroit: Institute of Gerontology. Available from: https://iog.wayne.edu/about/healthier-black-elders-center.
40. 40Participant resource pool. (no date) [Cited 2022 Oct 15]. Ann Arbor: Michigan Center for Urban African American Aging Research. Available from: https://mcuaaar.org/cores/community-liaison-and-recruitment-core/participant-resource-pool/.
41. 41Self-serve data tools. (no date) [Cited 2022 Sept 20]. Ann Arbor: Office of Research. Available from: https://research.medicine.umich.edu/our-units/data-office-clinical-translational-research/self-serve-data-tools.
42. 42Redcap access. (no date) [Cited 2022 Sept 20]. Ann Arbor: Michigan Institute for Clinical and Health Research. Available from: https://michr.umich.edu/redcap-access. CTSA: UL1TR002240.
43. 43Database Software and Applications. (no date) [Cited 20 Sep 2022]. Redmond: Microsoft. Available from: https://www.microsoft.com/en-us/microsoft-365/access.
44. Piatt G, Provenzano AM, Nwankwo R, Hall D, Kloss KA, Hawkins JM. **62-OR: Fostering Sustainability through Diabetes Self-Management Support in African-American Churches: Results of the Praise Diabetes Project**. *Diabetes* (2021.0) **70**. DOI: 10.2337/db21-62-OR
45. **A proposed charter for clinical trial data monitoring committees: Helping them to do their job well.**. *Lancet* (2005.0) **365** 711-22. DOI: 10.1016/S0140-6736(05)17965-3
46. 46Human research protection program (Hrpp). 2022
April
20 [Cited 2022 Sept 20]. Ann Arbor: Research Ethics & Compliance Available from https://research-compliance.umich.edu/human-subjects.. *Human research protection program (Hrpp)* (2022.0)
47. 47Dropbox at U-M. (no date) [Cited 2022 Sept 20]. Ann Arbor: Health Information Technology & Services. Available from: https://hits.medicine.umich.edu/collaboration/files-content/dropbox-u-m.
|
---
title: 'Evaluation of the Salivary Matrix Metalloproteinase-9 in Women With Polycystic
Ovaries Syndrome and Gingival Inflammation: A Case-Control Study'
journal: Cureus
year: 2023
pmcid: PMC9980837
doi: 10.7759/cureus.34458
license: CC BY 3.0
---
# Evaluation of the Salivary Matrix Metalloproteinase-9 in Women With Polycystic Ovaries Syndrome and Gingival Inflammation: A Case-Control Study
## Abstract
Background Polycystic ovary syndrome (PCOS) is an endocrine disease of women of reproductive age that impacts their oral and systemic well-being. This study aimed to compare the gingival inflammation indices and matrix metalloproteinase-9 (MMP-9) of non-obese women with PCOS.
Materials and methods *This is* a case-control study in which 78 women were referred to the Babol Clinic Hospital in Northern Iran between 2018 and 2019. They were divided into three groups: 26 women with PCOS and gingivitis, 26 women with PCOS with no gingivitis, and 26 women with no PCOS and no gingivitis as a control group. After recording the anthropometric and demographic variables, fasting saliva samples were taken from all participants before any periodontal intervention. These samples were transferred to Babol Molecular Cell Research Center under highly guaranteed cold-chain conditions to measure the serum levels of MMP-9. Periodontal status was evaluated for Gingival Index (GI), Plaque Index (PI), and Bleeding on Probing (BOP). Analysis of variance was used to compare the mean results for these indices. The significance level was considered when p ≤ 0.05.
Results All the gingival indices were significantly higher for women with PCOS with gingivitis compared to the results for women from the other two groups. Similarly, women with PCOS showed high salivary MMP-9 levels but were within the normal reference ranges.
Conclusion The gingival indices (GI, PI, and BOP) and salivary MMP-9 are higher in women with PCOS, regardless of the gingival status.
## Introduction
Polycystic ovary syndrome (PCOS) is one of the most common women endocrinopathies [1], which is caused by multifactorial etiology based on the interaction between genetic, environmental, and hormonal factors [2,3]. The estimated prevalence of PCOS in women of reproductive age from Iran ranges from $5.8\%$ to $19.5\%$, according to the different diagnostic criteria of PCOS [4,5].
Being a systemic disorder, there are several evidences of the relationship between gingivitis and systemic disorders, including PCOS, diabetes, and cardiovascular disease [6]. Since both gingivitis and PCOS are associated with systemic inflammation and insulin resistance, these two disorders might share a common pathophysiological pathway [7].
Studies suggest a potential link between this periodontal disease and PCOS through activation of different proinflammatory incidents like activation of different reactive oxygen species, myeloperoxidase, oxidative stress, inflammatory cytokines (such as IL-6 and TNF-α), high-sensitivity C-reactive protein (hs-CRP), adhesion molecules, and blood lymphocytes and monocytes [8-11]. However, some studies showed contradictory evidence [12].
Matrix metalloproteinase-9 (MMP-9) is a 92-kilodalton protein with protease activity whose main substrate is extracellular matrix and basement membrane tissue connections. It is the only family member that can attach and digest collagen as the most important component of the basement membrane due to its 3-fibronectin structure [13]. MMPs are found in a variety of environments. They have a very important role in migrating lymphoid and myeloid cells' physiological rearrangement of tissues, including organogenesis, normal growth, embryonic growth, angiogenesis, and ovulation [14]. This study aims to evaluate the possible linkage between different indices of gingival health, including the salivary MMP-9, to the polycystic ovary status in women with PCOS from Babol Northern Iran.
## Materials and methods
The present study is a case-control study conducted on women with PCOS who were referred to the Babol Clinic Hospital-Northern Iran between 2018 and 2019. The inclusion criteria for the case group were women aged 18 to 40 diagnosed with PCOS based on the Rotterdam diagnostic criteria [1]. The inclusion criteria for the women in the control group were those referred to the clinic for gynecological causes other than PCOS, with a normal menstrual cycle, without any clinical or biochemical indicators of hyperandrogenism, and no polycystic ovarian changes in their pelvic ultrasonography.
Exclusion criteria included any recent or current smoking history, women with diabetes mellitus, any history of malignancy, congenital adrenal hyperplasia, Cushing's syndrome, and androgen-secreting tumors, current smoking, intake of oral or injectable contraceptives within the last three months, history of antiepileptics or any drugs which are known to affect the gingiva by any level. Women having < 20 teeth and women with periodontitis were excluded.
Groups of body mass index (BMI) in kg/m2 were considered. If BMI is < 18.5, it falls within the underweight range. If BMI is 18.5 to <25, it falls within the healthy weight range. If BMI is 25.0 to <30, it falls within the overweight range. If BMI is 30.0 or higher, it falls within the obesity range [15].
A specialist gynecologist (assistant professor) examined all enrolled women. The women were diagnosed with PCOS according to the criteria above and then referred to the board-certified dentist after gaining the proper consent to be enrolled in the study. The latter examined the enrolled women for their periodontal state as either periodontally healthy or with gingivitis. The dentist in charge was blind to the PCOS status of women. If according to the depth of probing and clinical attachment loss, had periodontitis, it would be excluded from the study.
The gingival examination includes the Gingival Index (GI) (Silness and Löe), Plaque Index (PI) (Löe Index), Bleeding on Probing (BOP), and the number of existing teeth. To calculate the sample size, using the ratio of means method, the number of samples in each group (PCOS with gingivitis, PCOS with periodontally and systemically healthy control group) was calculated to be 26 persons with a $95\%$ confidence level and $90\%$ power. The enrolled women were divided into three groups: (Group 1) included women with PCOS and gingivitis, (Group 2) included women with PCOS with no gingivitis, and (Group 3) included women without PCOS or gingivitis.
Before the study's beginning, the measurement's reliability was determined by calculating the correlation coefficient for GI, PI, and BOP measurements. For this purpose, each index was measured twice in five women with gingivitis within one week. The results of these two examinations in terms of measurement criteria using correlation coefficient for quantitative indicators and kappa coefficient for quality index (BOP) between $80\%$ and $85\%$ was acquired, which indicated that the measurement had good reliability.
Details of periodontal examination After receiving informed consent from the enrolled women, the periodontal examination was performed in the case and the control group using the mirror and the periodontal probe (William's probe) made by Medisporex-Pakistan. For this purpose, the Ramfjord teeth, including maxillary right first molar (Tooth 16), maxillary left central incisor (Tooth 11), maxillary left first bicuspid (Tooth 24), mandibular left first molar (Tooth 36), mandibular right central incisor (Tooth 41) and mandibular right first bicuspid (Tooth 44) were evaluated for GI, PI, and BOP. The adjacent tooth was evaluated as recommended if any Ramfjord teeth were pulled out in the past [16].
The periodontal disease diagnosis was based on inflammation in the gum tissue and the periodontium. In the case of gingivitis, the gums become erythematous, edematous, and easily bled during any intervention, such as probing with a periodontal probe. While healthy gum is coral pink, firm in consistency, does not bleed easily during probing, had normal gingival sulcus depth, and normal bone height according to the latest classification scheme for periodontal diseases and conditions [17]. The participants' periodontal state was recorded as either healthy or with mild, moderate, or severe gingivitis.
Details of MMP-9 evaluation After the periodontal inspection to ensure no blood contamination, fasting saliva samples were taken from all participants in the morning before any periodontal intervention. The samples were transferred to sterilized micro-tubes and then sent to the Cellular and Molecular Biology Research Center at Babol University of Medical Sciences under the proper cold-chain conditions to measure the saliva levels of human MMP-9 using the quantitative single-wash 90 minutes Sandwich Enzyme-Linked Immunosorbent Assay (ELISA) kits. The sensitivity of this test was 22.17 pg/mL, with a range of (105.47-675 pg/mL).
Statistical analyses *The data* were analyzed using IBM Software Statistical Packages for Social Sciences (SPSS) Version 26.0 (IBM Corporation, Armonk, NY). Analysis of Variance (ANOVA) was used to compare the quantitative variables for the overall groups and for in-between group comparison. We could not have a normal distribution for any test parameter, even with the log transformation of the parameters. A significance level (p≤0.05) was chosen to indicate a significant association.
Ethical principles of the research This study was approved by the Ethics Committee of Mazandaran University of Medical Sciences on October 1, 2018, under the ethical approval code (IR.MAZUMS.REC.1397.2934). The participants were assured that their participation was voluntary, their data would remain confidential, and their autonomy would be respected. Written informed consent was obtained from each one of them.
## Results
A total of 78 women were enrolled in the study, all were never smokers, with 26 women in each group. The ranges of the age, BMI, and teeth count of the enrolled women were (16-40 years old), (18.70-36.60 kg/m2), and (25-32 teeth), respectively. Although the results of the parameters mentioned above were nearly similar and comparable between the three groups, they were significantly higher in women with PCOS and gingivitis. The periodontal indices for lower and upper jaws are significantly higher for women with PCOS and gingivitis than women from the other two groups. Figures 1A-1I represent different periodontal images for the enrolled women from the three groups.
Although all the results of the MMP-9 in all groups are within the normal reference ranges, yet women with PCOS and gingivitis had significantly higher ranges than other groups. Figure 2 illustrates the results of the MMP-9 for women in all groups. Women with PCOS and gingivitis had significantly higher MMP-9 than all groups. This result was evident by the overall mean comparison and the in-between group comparison using ANOVA. Women with PCOS but no gingivitis had significantly higher MMP-9 than women who had neither PCOS nor gingivitis.
**Figure 2:** *Boxplot of the results of MMP-9 in women from the three groups of the study. The overall p-value by ANOVA was < 0.005. The mean of the MMP-9 is illustrated on the left of each boxplot. The significance level of in-between groups comparison was measured. The single faint circle represents a single outlier result in one woman with PCOS and gingivitis.*
## Discussion
Blood and saliva are valid sources for the estimation of many biomarkers like MMP-9, and there is evidence that serum and salivary MMP-9 show significantly increased levels in women with PCOS and gingivitis, which may indicate an exaggerated effect of the gingival inflammatory process in PCOS [12,18]. For the sake of convenience, we chose salivary samples in this study.
Relatively higher levels of MMP-9, the proinflammatory markers in saliva, may be a manifestation of low-grade systemic inflammation associated with PCOS [19]. Per the present study, systematic reviews by Márquez-Arrico et al. and Kellesarians et al. have confirmed that a positive association between periodontal pathologies and PCOS, particularly gingivitis and chronic periodontitis [19,20]. Likewise, Wendland et al. [ 2021] concluded that young women with PCOS and good oral health maintenance do not vary from healthy controls concerning gingival indices [21].
In the study of Dursun et al., clinical periodontal parameters similar to ours were compared in 25 women with PCOS and 27 healthy control women. These indices in tested women with PCOS were higher than in the women control group. They showed the volume of the gingival crevicular fluid, as well as the amount of nitric oxide (NO) and myeloperoxidase (MPO) in the gingival sulcus, was higher among the women with PCOS. Their findings suggested that the local/periodontal oxidant status in women with PCOS was damaged, and their susceptibility to periodontal diseases increased considerably [22].
In the present study, although all the readings of MMP-9 fall within the normal reference range, they were higher in women with PCOS regardless of their gingival status compared to women from the control group. The possible linkage between MMP-9 and PCOS pathogenesis is that the rate of growth and regeneration of ovarian follicles is modified by the balance between MMPs in the extracellular environment [23], which may explain the higher levels seen in women with PCOS compared to women in the control group in some studies [24-26]. Other studies provided contradictory results [27-29].
The small sample size, being a single-center cross-sectional observational study, limited the generalizability of the results, and fail to ensure causality. Additionally, this study did not convey a comparable evaluation of other proinflammatory and inflammatory indicators of PCOS and similar conditions.
## Conclusions
The indicators of gingival inflammation (GI, PI, and BOP) were higher in women with PCOS compared to women with PCOS with healthy gingiva, women in the control group. Although salivary MMP-9 was significantly higher for women with PCOS regardless of the gingival status compared to women in the other two groups, the levels were within the normal reference ranges, which will ameliorate the effect of such finding on the possible linkage to PCOS pathogenesis. Still, this association should be emphasized in the management plan of women with PCOS by regular referral for maintaining optimal dental and gingival care. Gingival and periodontal health status could mirror the ovarian status in women with PCOS to an extent, but this needs to be implicated in large-scale longitudinal studies in multicenter settings are recommended.
## References
1. Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group. **Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome**. *Fertil Steril* (2004) **81** 19-25
2. McCook JG, Reame NE, Thatcher SS. **Health-related quality of life issues in women with polycystic ovary syndrome**. *J Obstet Gynecol Neonatal Nurs* (2005) **34** 12-20
3. DuRant EM, Leslie NS, Critch EA. **Managing polycystic ovary syndrome: a cognitive behavioral strategy**. *Nurs Womens Health* (2009) **13** 292-300. PMID: 19686552
4. Tehrani FR, Simbar M, Tohidi M, Hosseinpanah F, Azizi F. **The prevalence of polycystic ovary syndrome in a community sample of Iranian population: Iranian PCOS prevalence study**. *Reprod Biol Endocrinol* (2011) **9** 39. PMID: 21435276
5. Jalilian A, Kiani F, Sayehmiri F, Sayehmiri K, Khodaee Z, Akbari M. **Prevalence of polycystic ovary syndrome and its associated complications in Iranian women: a meta-analysis**. *Iran J Reprod Med* (2015) **13** 591-604. PMID: 26644787
6. Kinane D, Bouchard P. **Periodontal diseases and health: Consensus Report of the Sixth European Workshop on Periodontology**. *J Clin Periodontol* (2008) **35** 333-337
7. Akcalı A, Bostanci N, Özçaka Ö, Öztürk-Ceyhan B, Gümüş P, Buduneli N, Belibasakis GN. **Association between polycystic ovary syndrome, oral microbiota and systemic antibody responses**. *PLoS One* (2014) **9** 0
8. Porwal S, Tewari S, Sharma RK, Singhal SR, Narula SC. **Periodontal status and high-sensitivity C-reactive protein levels in polycystic ovary syndrome with and without medical treatment**. *J Periodontol* (2014) **85** 1380-1389. PMID: 24592911
9. Souza Dos Santos AC, Soares NP, Costa EC, de Sá JC, Azevedo GD, Lemos TM. **The impact of body mass on inflammatory markers and insulin resistance in polycystic ovary syndrome**. *Gynecol Endocrinol* (2015) **31** 225-228. PMID: 25373529
10. Farook FF, Ng KT, MNM N, Koh WJ, Teoh WY. **Association of periodontal disease and polycystic ovarian syndrome: a systematic review and meta-analysis with trial sequential analysis**. *Open Dentistry J* (2019) **13** 478-487
11. Victor VM, Rovira-Llopis S, Bañuls C. **Insulin resistance in PCOS patients enhances oxidative stress and leukocyte adhesion: role of myeloperoxidase**. *PLoS One* (2016) **11** 0
12. Akcalı A, Bostanci N, Özçaka Ö. **Gingival inflammation and salivary or serum granulocyte-secreted enzymes in patients with polycystic ovary syndrome**. *J Periodontol* (2017) **88** 1145-1152. PMID: 28598286
13. Visse R, Nagase H. **Matrix metalloproteinases and tissue inhibitors of metalloproteinases: structure, function, and biochemistry**. *Circ Res* (2003) **92** 827-839. PMID: 12730128
14. Kessenbrock K, Plaks V, Werb Z. **Matrix metalloproteinases: regulators of the tumor microenvironment**. *Cell* (2010) **141** 52-67. PMID: 20371345
15. **National Heart, Lung, and Blood Institute (NHLBI) Obesity Education Initiative Expert Panel on the Identification, Evaluation, and Treatment of Obesity in Adults (US). Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults. The Evidence Report. Bethesda (MD): National Heart, Lung, and Blood Institute**. (2023)
16. Hajishengallis G. **Periodontitis: from microbial immune subversion to systemic inflammation**. *Nat Rev Immunol* (2015) **15** 30-44. PMID: 25534621
17. Caton JG, Armitage G, Berglundh T. **A new classification scheme for periodontal and peri-implant diseases and conditions - introduction and key changes from the 1999 classification**. *J Clin Periodontol* (2018) **45 Suppl 20** 0-8
18. Randeva HS, Tan BK, Weickert MO, Lois K, Nestler JE, Sattar N, Lehnert H. **Cardiometabolic aspects of the polycystic ovary syndrome**. *Endocr Rev* (2012) **33** 812-841. PMID: 22829562
19. Márquez-Arrico CF, Silvestre-Rangil J, Gutiérrez-Castillo L, Martinez-Herrera M, Silvestre FJ, Rocha M. **Association between periodontal diseases and polycystic ovary syndrome: a systematic review**. *J Clin Med* (2020) **9** 1586. PMID: 32456146
20. Kellesarian SV, Malignaggi VR, Kellesarian TV. **Association between periodontal disease and polycystic ovary syndrome: a systematic review**. *Int J Impot Res* (2017) **29** 89-95. PMID: 28275229
21. Wendland N, Opydo-Szymaczek J, Formanowicz D, Blacha A, Jarząbek-Bielecka G, Mizgier M. **Association between metabolic and hormonal profile, proinflammatory cytokines in saliva and gingival health in adolescent females with polycystic ovary syndrome**. *BMC Oral Health* (2021) **21** 193. PMID: 33849511
22. Dursun E, Akalın FA, Güncü GN. **Periodontal disease in polycystic ovary syndrome**. *Fertil Steril* (2011) **95** 320-323. PMID: 20800834
23. Liu B, Guan YM, Zheng JH. **Elevated serum levels of matrix metalloproteinase-2 in women with polycystic ovarian syndrome**. *Int J Gynaecol Obstet* (2007) **96** 204-205. PMID: 17291506
24. Javed F, Ahmed A. **Proinflammatory cytokines in the saliva, gingival crevicular fluid and serum of diabetic patients with periodontal disease**. *J Res Practice in Dentistry (DENT)* (2013) **2013** 956990
25. Ranjbaran J, Farimani M, Tavilani H, Ghorbani M, Karimi J, Poormonsefi F, Khodadadi I. **Matrix metalloproteinases 2 and 9 and MMP9/NGAL complex activity in women with PCOS**. *Reproduction* (2016) **151** 305-311. PMID: 26733727
26. Dambala K, Vavilis D, Bili E, Goulis DG, Tarlatzis BC. **Serum visfatin, vascular endothelial growth factor and matrix metalloproteinase-9 in women with polycystic ovary syndrome**. *Gynecol Endocrinol* (2017) **33** 529-533. PMID: 28300464
27. Gerlach RF, Uzuelli JA, Souza-Tarla CD, Tanus-Santos JE. **Effect of anticoagulants on the determination of plasma matrix metalloproteinase (MMP)-2 and MMP-9 activities**. *Anal Biochem* (2005) **344** 147-149. PMID: 15950912
28. Gerlach RF, Demacq C, Jung K, Tanus-Santos JE. **Rapid separation of serum does not avoid artificially higher matrix metalloproteinase (MMP)-9 levels in serum versus plasma**. *Clin Biochem* (2007) **40** 119-123. PMID: 17150202
29. Souza-Tarla CD, Uzuelli JA, Machado AA, Gerlach RF, Tanus-Santos JE. **Methodological issues affecting the determination of plasma matrix metalloproteinase (MMP)-2 and MMP-9 activities**. *Clin Biochem* (2005) **38** 410-414. PMID: 15820769
|
---
title: Meal-timing patterns and chronic disease prevalence in two representative Austrian
studies
authors:
- Isabel Santonja
- Leonie H. Bogl
- Jürgen Degenfellner
- Gerhard Klösch
- Stefan Seidel
- Eva Schernhammer
- Kyriaki Papantoniou
journal: European Journal of Nutrition
year: 2023
pmcid: PMC9980854
doi: 10.1007/s00394-023-03113-z
license: CC BY 4.0
---
# Meal-timing patterns and chronic disease prevalence in two representative Austrian studies
## Abstract
### Purpose
This study aimed at describing meal-timing patterns using cluster analysis and explore their association with sleep and chronic diseases, before and during COVID-19 mitigation measures in Austria.
### Methods
Information was collected in two surveys in 2017 ($$n = 1004$$) and 2020 ($$n = 1010$$) in representative samples of the Austrian population. Timing of main meals, nighttime fasting interval, last-meal-to-bed time, breakfast skipping and eating midpoint were calculated using self-reported information. Cluster analysis was applied to identify meal-timing clusters. Multivariable-adjusted logistic regression models were used to study the association of meal-timing clusters with prevalence of chronic insomnia, depression, diabetes, hypertension, obesity and self-rated bad health status.
### Results
In both surveys, median breakfast, lunch and dinner times on weekdays were 7:30, 12:30 and 18:30. One out of four participants skipped breakfast and the median number of eating occasions was 3 in both samples. We observed correlation between the different meal-timing variables. Cluster analysis resulted in the definition of two clusters in each sample (A17 and B17 in 2017, and A20 and B20 in 2020). Clusters A comprised most respondents, with fasting duration of 12–13 h and median eating midpoint between 13:00 and 13:30. Clusters B comprised participants reporting longer fasting intervals and later mealtimes, and a high proportion of breakfast skippers. Chronic insomnia, depression, obesity and self-rated bad health-status were more prevalent in clusters B.
### Conclusions
Austrians reported long fasting intervals and low eating frequency. Meal-timing habits were similar before and during the COVID-19-pandemic. Besides individual characteristics of meal-timing, behavioural patterns need to be evaluated in chrono-nutrition epidemiological studies.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00394-023-03113-z.
## Introduction
Circadian misalignment through night shift work has been associated with various chronic disease outcomes [1–6]. Recently, this research has been expanded to identify other sources of circadian disruption in the general population that may result from mistimed exposures, such as sleeping or eating at the “wrong” time [7–9]. More concretely, the adherence to a more diurnal eating pattern has been associated with a lower cancer risk [10–12] and an improvement of metabolic function and weight loss [13]. Moreover, individual aspects associated with a diurnal eating pattern, such as eating breakfast or abstaining from eating at night, have been associated with better sleep quality and longer sleep duration [8]. On the contrary, individuals experience a reduction of sleep duration and an increase of daytime sleepiness during Ramadan, a period of 29–30 days, during which healthy adult Muslims abstain from eating and drinking from dawn to sunset [14]. These observations stress the importance of evaluating timing (in addition to quality, quantity and frequency) of diet in epidemiological studies.
In terms of exposure assessment, most epidemiological studies have characterized meal-timing through questionnaires and have tended to analyse single meal-timing aspects, e.g. dinner time [10] or breakfast skipping [15–20]. However, different meal-timing aspects, such as nighttime fasting interval, interval between the last eating occasion and bed time or breakfast skipping, are interrelated and their effects could overlap. Because of that, Khanna et al. [ 21] argued that analysing meal-timing behaviours, rather than isolated aspects, using cluster analysis might be a more suitable method of assessing the effects of meal-timing.
Moreover, meal-timing patterns have not been previously described in detail in the Austrian population, and we hypothesized that dietary patterns, especially the eating frequency, could have been modified in Austria as a consequence of the COVID-19 mitigation measures and lockdown, as reported for other populations [22–26]. Furthermore, although evidence starts to emerge that meal-timing might have an influence on health outcomes, little is known in relation to sleep outcomes.
Therefore, the first aim of this study was to describe meal-timing and meal frequency, as well as fasting patterns, and to define predominant meal-timing behaviours in the Austrian population. Secondly, we aimed to compare meal-timing patterns before and during the first COVID-19 mitigation measures and evaluate potential effects of the lockdown on meal-timing habits in Austria. Lastly, we explored the cross-sectional association of meal-timing behaviours and chronic insomnia, chronic disease prevalence and self-rated health-status.
## Population and study design
Two online surveys were implemented by Interrogare GmbH − a market research institute based in Germany −, with the aim of eliciting detailed information on sleep habits and its determinants in representative samples of the *Austrian* general population. We collected information on sleep, daily routines, such as meal-times, and lifestyle characteristics. Participants of both surveys were selected to represent the age (≥ 18 years), sex and county distribution of Austria’s general population. Participation in both surveys was voluntary and anonymous, and informed consent was implied through participation.
The first survey took place in September 2017, included 63 questions and took approximately 30 min to complete. In total, 1004 participants completed the survey. Of those, current night shift workers ($$n = 52$$) and one participant with missing information in all meal-timing variables were excluded from the present analysis ($$n = 951$$ participants included in description of meal-timing). Moreover, 27 further participants with missing information on key meal-timing variables were excluded from the cluster and association analysis ($$n = 924$$ participants included in the cluster analysis).
The second survey was conducted in June 2020, after the first COVID-19 wave and the implementation of mitigation measures in Austria (March 16 to May 1st, 2020). The questionnaire, which included 81 questions and took approximately 20 min to complete, was completed by 1010 participants. Current nightshift workers ($$n = 55$$) and three additional participants reporting gender other than women or men were excluded for the analysis ($$n = 952$$ participants included in description of meal timing). Additionally, 82 participants were excluded for the cluster analysis because of insufficient information on meal-timing ($$n = 870$$ participants included in the cluster analysis).
## Variables
The surveys collected the following meal-timing information during the week and on weekends: timing of breakfast, lunch and dinner (as drop down menu with 1-h intervals, e.g., from 12:00 till 13:00), snacks between meals (“yes”/”no”), snack between breakfast and lunch (“yes”/”no”), snack between lunch and dinner (“yes”/”no”), snack between dinner and breakfast (“yes”/”no”) and timing of last snack of the day (hours, minutes). We used the midpoint of the intervals of the hourly bins (e.g., “12:00–13:00″ was substituted by 12:30) to create pseudocontinuous variables for the time of breakfast, lunch and dinner. The continuous variable nighttime fasting was defined as the time elapsed between the last and the first meal of the day. We created two additional continuous variables: last meal to bed time (hours) and eating midpoint (hours), defined as the midpoint between the first and the last meal of the day. We also generated the discrete variable number of eating occasions (ranging from 0 to 6, the maximum number of eating occasions that could be reported in the surveys), which included main meals and snacks, and a dichotomous variable, breakfast skipping (“skipping” vs. “eating”). Each variable was calculated independently for weekdays and weekends.
The survey collected detailed information on sleep duration, sleep timing and sleep quality. We used the 3rd edition of the International Classification of Sleep Disorders (ICSD-III) [27] to define chronic insomnia (“yes”/”no”), as explained in Weitzer et al. [ 28]. The surveys also collected information on self-rated health status (“In your opinion: *How is* your health status in general?”, one answer possible: “very good”/“good”/“moderate”/“bad”/“very bad”) and diagnosed medical conditions (“During the past 12 months, did you have any of the following diseases or conditions?”; multiple answers possible). With this information, we defined the following dichotomous (“yes”/“no”) outcome variables: depression, diabetes, hypertension and bad or very vad self-rated health status. Participants also reported their height and weight, which were used to calculate BMI and define obesity [“yes” (BMI ≥ 30 kg/m2)/ “no” (BMI < 30 kg/m2)].
The survey also collected information on sex and age (“How old are you?”, with respondents asked to fill in their age in the 2017 survey, and to choose one of the following categories in the 2020 survey: “ < 20” / “20–24” / “25–29” / “30–34” / “35–39” / “40–44” / “45–49” / “50–54” / “55–59” / “60–64” / “65–69” / “ ≥ 70”) and other confounders and effect modifiers of interest, i.e. self-rated chronotype (“One hears about “morning” and “evening” types of people. Which ONE of these types do you consider yourself to be?”, one answer possible: “definitely a morning type”/ “rather more a morning than an evening type”/ “rather more an evening than a morning type”/ “definitely an evening type”), marital status (“*What is* your current marital status?”, one answer possible: “single”/ “married or in a partnership”/ “divorced”/ “widowed”), work status [“*What is* your current work status?”, multiple answers possible: “(self-) employed full-time” / “(self-) employed part-time” / “retired” / “unemployed” / “student, further training, unpaid work experience” / “disabled” / “in compulsory military or community service” / “household”], alcohol consumption [“How much alcohol do you drink per week? ( Please give approximate/average amounts)”, with respondents asked to fill in the number of glasses of beer and wine and shots of liquor/whiskey/gin etc. consumed per week], smoking status (“Do you currently smoke?”, one answer possible: “No, never”/“No, not anymore”/“Yes, I currently smoke”) and history of nightshifts (“Have you ever worked night shifts (schedule including ≥ 3 h of work between 12 pm and 6 am and at least 3 nights/month)?”, one answer possible: “No” / “Yes, in the past” / “Yes, currently”).
## Statistical analysis
Summary statistics [medians and interquartile ranges (IQRs), and frequency (N and %)] were used to describe baseline characteristics and meal-timing patterns, independently for each survey. The correlation between numerical meal-timing variables during the week and during the weekend and between different numerical variables in each survey was analysed graphically (with matrix scatter plots) and the significance of correlation was tested using the Pearson correlation coefficient. The association of these variables with breakfast skipping was assessed using point biserial correlation. The association between breakfast skipping during the week and the weekend was analysed using the χ2 test.
Within surveys, cluster analysis was performed to group individuals with similar meal-timing behaviours. Nighttime fasting, last meal to bed time and eating midpoint during the week were standardized and included as indicators. To establish the cluster groups, a combination of hierarchical and non-hierarchical clustering methods was applied. Firstly, we used Ward’s method (hierarchical method) removing univariate outliers (values > 3 SD above or below the mean) and generated the resulting dendrogram, in order to select the optimum number of clusters (two for each survey). Using the initial cluster centres obtained by hierarchical clustering and including also outliers of the variables, an iterative non-hierarchical K-means clustering procedure was applied. The Cohen’s ҡ coefficient for the solutions obtained by hierarchical methods and by non-hierarchical methods (final cluster solution) was 0.96 for the 2017 survey, indicating almost perfect agreement, and 0.77 for the 2020 survey (substantial agreement).
To describe characteristics or predictors of the different cluster groups, as well as participants’ sociodemographic and lifestyle characteristics, summary statistics [medians and interquartile ranges (IQRs), and frequency (N and %)] were used. Differences on the indicators between cluster groups were analysed using the Wilcoxon rank-sum test. Within surveys and using the largest cluster as reference category, unconditional logistic regression analysis was performed to study the association of meal-timing behaviours and chronic insomnia, depression, obesity, diabetes, hypertension and self-rated health status. Logistic regression models were used and odds ratios (OR) with $95\%$ confidence intervals were calculated. Age and sex-adjusted ORs (AORs) and multivariable-adjusted ORs (MV-ORs) are presented. In addition to age and sex and based on a directed acyclic graph, we considered the following potential confounders for the multivariable adjusted model: self-rated chronotype, marital status, work status, alcohol consumption, smoking status and history of nightshifts.
Risk estimates were compared across strata of sex and chronotype profiles (early/late). In sensitivity analyses, only participants without report of heavy alcohol drinking (drinking ≤ 12 standard glasses of alcohol a week) or those who had no history of nightshift were included.
All statistical analyses were performed using STATA 16.
## Results
The respondents of both surveys, as well as the sub-samples of participants allocated to cluster groups, had similar baseline characteristics, which are shown in Suppl. table 1.
## Meal-timing in the Austrian population in 2017 and 2020
Participants reported similar timing of the main meals during the week (median times of breakfast, lunch and dinner: 7:30, 12:30 and 18:30, respectively) and long nighttime fasting intervals during the week (median: 13 h) in both surveys. Some participants ($8.8\%$ in 2017 and $10.1\%$ in 2020) had a snack after dinner around 21:00. The time elapsed between last meal and bedtime during the week was longer in 2020 (6.0 h in 2020 vs. 4.0 h in 2017) and the number of eating occasions decreased slightly. During the weekend, participants generally reported later breakfast times and longer fasting intervals compared to weekdays; in 2020, participants also reported later lunch and dinner times during the weekend. Furthermore, eating breakfast was more frequently reported during the weekends than on weekdays (2017: $83.9\%$ vs. $75.4\%$; 2020: $81.1\%$ vs. $74.9\%$; Table 1 and Suppl. Figure 1).Table 1Meal timing, fasting intervals and number of eating occasions in two representative samples of the general Austrian populationSurvey 2017 ($$n = 951$$)Survey 2020 ($$n = 952$$)N/n (%)Median (IQR)N/n (%)Median (IQR)Weekdays Breakfast time717 (75.4)7:30 (6:30–8:30)713 (74.9)7:30 (6:30–8:30) Lunch time814 (85.6)12:30 (12:30–13:30)836 (87.8)12:30 (12:30–13:30) Dinner time880 (92.5)18:30 (18:30–19:30)864 (90.8)18:30 (18:30–19:30) Snack after dinner time84 (8.8)21:00 (20:00–21:45)96 (10.1)21:01 (20:01–22:01) Nighttime fasting (h)944 (99.3)13.0 (12.0–15.0)887 (93.2)13.0 (12.0–17.0) Last meal to bed time (h)924 (97.2)4.0 (3.0–5.3)913 (95.9)6.0 (5.0–7.5) Number of eating occasions951 (100.0)3.0 (3.0–4.0)952 (100.0)3.0 (2.0–4.0)Weekend Breakfast time798 (83.9)8:30 (7:30–9:30)772 (81.1)8:30 (7:30–9:30) Lunch time830 (87.3)12:30 (12:30–13:30)831 (87.3)13:30 (12:30–13:30) Dinner time872 (91.7)18:30 (18:30–19:30)847 (89.0)19:30 (18:30–19:30) Snack after dinner time143 (15.0)21:00 (20:15–22:00)135 (14.2)21:01 (20:01–22:01) Nighttime fasting (h)938 (98.6)14.0 (13.0–15.0)887 (93.2)14.0 (13.0–16.0) Last meal to bed time (h)890 (93.6)4.5 (3.5–5.5)885 (93.0)6.5 (5.5–7.5) Number of eating occasions951 (100.0)3.0 (3.0–4.0)952 (100.0)3.0 (3.0–4.0)
## Analysis of collinearity and association among meal-timing variables
We found moderate to strong correlation between individual meal-timing variables (nighttime fasting, last meal to bed time or number of eating occasions) during the week and during the weekend in both surveys (r ≥ 0.58 in 2017 and r ≥ 0.67 in 2020; see Fig. 1). Breakfast skipping during the week was also correlated with breakfast skipping in the weekend ($p \leq 0.001$) in both surveys: the majority of those skipping breakfast during the week also reported skipping it during the weekend (2017: $$n = 130$$/234, $55.6\%$; 2020: $$n = 158$$/239, $66.1\%$). Because of these associations, we used only meal-timing patterns during the week to generate the cluster solutions and in association analyses. Fig. 1Correlation of meal timing variables in a the 2017 survey b the 2020 survey. Pearson’s r correlation coefficient is shown in the upper panels of the matrix Furthermore, we found evidence of correlation among the aforementioned meal-timing variables. For example, we found a negative correlation between nighttime fasting and the number of eating occasions, which was consistent during week and weekend in both surveys, and moderate positive correlation between nighttime fasting and last meal to bed time, which was particularly strong during the weekends in 2017 ($r = 0.60$). ( Fig. 1) In addition, we found a correlation between breakfast patterns and nighttime fasting and number of eating occasions, with participants skipping breakfast reporting longer nighttime fasting periods and eating less frequently (Suppl. Figure 2 and 3 for meal timing patterns during the week; similar results were obtained for the weekend, but are not shown).
## Cluster analysis
In each survey, cluster analysis resulted in the definition of two cluster groups [2017: A17 ($$n = 720$$; $77.9\%$) and B17 ($$n = 204$$; $22.1\%$); 2020: A20 ($$n = 576$$; $66.2\%$) and B20 ($$n = 294$$; $33.8\%$)] with different meal-timing patterns (Table 2). Participants in cluster groups B ate later and less frequently, were more likely to skip breakfast and reported longer nighttime fasting intervals and time elapsed from last meal to bed time compared to this in cluster A.Table 2Distribution of meal timing patterns in cluster groupsSurvey 2017 ($$n = 924$$)Survey 2020 ($$n = 870$$)Cluster A17 ($$n = 720$$)Median (IQR)Cluster B17 ($$n = 204$$)Median (IQR)p (cluster differences)aCluster A20 ($$n = 576$$)Median (IQR)Cluster B20 ($$n = 294$$)Median (IQR)p (cluster differences)aNighttime fasting* (h)12.0 (11.0–13.3)18.0 (17.0–20.0) < 0.00113.0 (12.0–13.0)18.0 (16.0–19.0) < 0.001Last meal to bed time* (h)4.0 (3.0–5.0)4.5 (3.5–5.5) < 0.0015.5 (4.5–6.5)7.4 (5.8–9.0) < 0.001Breakfast skipping (yes); N(%)46 (6.4)184 (90.2) < 0.00110 (1.74)175 (59.5) < 0.001Eating midpoint* (hh:mm)13:00 (12:30–14:00)16:00 (15:00–17:00) < 0.00113:30 (12:30–14:00)15:30 (13:30–16:30) < 0.001Number of eating occasions3.0 (3.0–4.0)2.0 (2.0–2.5) < 0.0013.0 (3.0–4.0)2.0 (2.0–3.0) < 0.001*Variables used to generate cluster solution. a. p-values calculated using Wilcoxon rank-sum test The sociodemographic and lifestyle characteristics of the different groups are shown in Table 3. In both surveys, participants in cluster groups B were more likely to be single, divorced or widowed and current smokers, later chronotypes and not to engage in any type of physical activity. Additionally, in 2017, participants in cluster B had a significantly higher BMI. Besides, in 2020, participants in cluster A were more likely to be employed. Table 3Sociodemographic characteristics across cluster groups in 2017 and 2020Survey 2017 ($$n = 924$$)Survey 2020 ($$n = 870$$)Cluster A17($$n = 720$$)Cluster B17($$n = 204$$)Cluster A20($$n = 576$$)Cluster B20($$n = 294$$)N(%)N(%)N(%)N(%)Age 18–2498 (13.6)21 (10.3)71 (12.3)37 (12.6) 25–34119 (16.5)42 (20.6)117 (20.3)45 (15.3) 35–44164 (22.8)46 (22.5)131 (22.7)65 (22.1) 45–54177 (24.6)61 (29.9)138 (24.0)78 (26.5) ≥ 55162 (22.5)34 (16.7)119 (20.7)69 (23.5) Sex (Women)376 (52.2)97 (47.5)298 (51.7)158 (53.7) BMI median(IQR)*24.5 (21.8–27.7)25.6 (22.2–28.7)24.2 (21.6–27.7)24.9 (21.5–29.0)Education High school or less289 (40.1)82 (40.2)197 (34.2)104 (35.4) Matura260 (36.1)84 (41.2)206 (35.8)116 (39.5)University degree or above171 (23.8)38 (18.6)173 (30.0)74 (25.2)*Marital status* *,** Single207 (28.8)72 (35.3)175 (30.4)113 (38.4) Married/ in a partnership427 (59.3)98 (48.0)352 (61.1)139 (47.3) Divorced75 (10.4)29 (14.2)42 (7.3)35 (11.9) Widowed11 (1.5)5 (2.5)7 (1.2)7 (2.4)*Work status* ** Employed full time363 (50.4)105 (51.5)304 (52.8)145 (49.3) Employed part time85 (11.8)18 (8.8)79 (13.7)32 (10.9) Retired97 (13.5)24 (11.8)61 (10.6)34 (11.6) Unemployed and disabled50 (6.9)19 (9.3)38 (6.6)37 (12.6) Student, further training…81 (11.3)23 (11.3)70 (12.2)38 (12.9) Household44 (6.1)15 (7.4)24 (4.2)8 (2.7)Area of residenceUrban328 (45.6)98 (48.0)286 (49.7)154 (52.4)Rural < 50.000 inhabitants305 (42.4)84 (41.2)211 (36.6)99 (33.7)Rural > 50.000 inhabitants87 (12.1)22 (10.8)79 (13.7)41 (13.9)Drinking alcohol No standard glasses285 (39.6)83 (40.7)226 (39.2)113 (38.4) 1–6 standard glasses/week245 (34.0)56 (27.5)259 (45.0)126 (42.9) 7–12 standard glasses/week95 (13.2)36 (17.6)53 (9.2)32 (10.9) > 12 standard glasses/week95 (13.2)29 (14.2)38 (6.6)23 (7.8)Smoking status*, ** No, never333 (46.3)65 (31.9)274 (47.6)110 (37.4) No, not anymore204 (28.3)43 (21.1)158 (27.4)74 (25.2) Yes, I currently smoke183 (25.4)96 (47.1)144 (25.0)110 (37.4)Time of physical activitya,*,** No physical activity247 (34.3)96 (47.1)132 (22.9)90 (30.6) Before 12 pm90 (12.5)14 (6.9)109 (18.9)38 (12.9) 12.00–18.00176 (24.4)37 (18.1)144 (25.0)70 (23.8) After 18.00207 (28.8)57 (27.9)191 (33.2)96 (32.7)Self-rated chronotype*,** *Definitely a* morning person151 (21.0)33 (16.2)137 (23.8)52 (17.7) *Rather a* morning person216 (30.0)42 (20.6)160 (27.8)75 (25.5) Rather an evening person238 (33.1)65 (31.9)165 (28.7)85 (28.9) Definetely an evening person115 (16.0)64 (31.4)114 (19.8)82 (27.9) Ever worked on nightshifts199 (27.6)51 (25.0)180 (31.3)104 (35.4)Differences in the distribution between cluster groups were evaluated using Pearson’s χ2a. In the survey in 2017 information on moderate and vigorous physical activity is available; in 2020 only on vigorous physical activity*$p \leq 0.05$ in the survey 2017**$p \leq 0.05$ in the survey 2020
## Meal-timing patterns and health status
The self-reported prevalence of common chronic diseases and self-rated health status across cluster groups and survey is shown in Table 4. Subjects in clusters B were more likely to report chronic insomnia [AOR ($95\%$ CI) = 2.43 (1.46–4.05) in 2017; AOR ($95\%$ CI) = 1.70 (1.00–2.90) in 2020], although this effect was reduced after adjusting for lifestyle and sociodemographic confounders [MV-OR ($95\%$ CI) = 2.23 (1.29–3.87) in 2017; MV-OR ($95\%$ CI) = 1.49 (0.84–2.63) in 2020]. Participants in groups B17 and B20 were also more likely to report having been diagnosed with depression than subjects in A17 and A20, respectively, but risk estimates were only significant in 2020 [MV-OR ($95\%$ CI) = 1.55 (1.02–2.36)]. Participants in clusters A were less likely to report obesity, but risk estimates were not significant [MV-OR ($95\%$ CI) = 1.18 (0.75–1.85) in 2017 and MV-OR = 1.25 (0.84–1.86) in 2020]. Prevalence of diabetes was similar across surveys and cluster groups, and we did not found any consistent trend for hypertension. Finally, we observed also a higher risk of reporting a bad or very bad health status among subjects in cluster B [AOR ($95\%$ CI) = 2.21 (1.27–3.84) in 2017; AOR ($95\%$ CI) = 2.86 (1.58–5.20) in 2020], which was reduced after adjusting for confounders [MV-OR ($95\%$ CI) = 1.68 (0.92–3.08) in 2017; MVOR ($95\%$ CI) = 2.48 (1.29–4.74)].Table 4Association of meal timing behaviour and self-rated health status and chronic diseases in 2017 and 2020 (OR: Odds Ratio, $95\%$CI: $95\%$ confidence interval)Survey 2017 ($$n = 924$$)Survey 2020 ($$n = 870$$)Cluster A17($$n = 720$$)N(%)Cluster B17 ($$n = 204$$)N(%)AOR ($95\%$ CI)aMV-OR ($95\%$ CI)bCluster A20($$n = 576$$)N(%)Cluster B20($$n = 294$$)N(%)AOR ($95\%$ CI)aMV-OR ($95\%$ CI)bChronic insomnia43 (6.0)27 (13.2)2.43 (1.46–4.05)2.23 (1.29–3.87)32 (5.6)27 (9.2)1.70 (1.00–2.90)1.49 (0.84–2.63)Depression80 (11.1)29 (14.2)1.32 (0.83–2.08)1.14 (0.70–1.87)65 (11.3)56 (19.0)1.84 (1.25–2.72)1.55 (1.02–2.36)Obesity117 (16.3)35 (17.2)1.11 (0.73–1.70)1.18 (0.75–1.85)91 (15.8)60 (20.4)1.34 (0.92–1.94)1.25 (0.84–1.86)Diabetes33 (4.6)10 (4.9)1.20 (0.56–2.54)1.27 (0.57–2.84)25 (4.3)14 (4.8)1.06 (0.54–2.11)0.99 (0.47–2.07)Hypertension106 (14.7)25 (12.3)0.86 (0.53–1.41)0.81 (0.48–1.36)77 (13.4)49 (16.7)1.26 (0.84–1.89)1.10 (0.71–1.70)Bad or very bad self-rated health status38 (5.3)22 (10.8)2.21 (1.27–3.84)1.68 (0.92–3.08)20 (3.5)28 (9.5)2.86 (1.58–5.20)2.48 (1.29–4.74)ORs calculated using unconditional logistic regressiona. AOR = Adjusted OR. Adjusted for age and sexb. MV-OR = multivariable-adjusted OR. Adjusted for age, sex, self-rated chronotype, marital status, work status, alcohol consumption, smoking status and history of nightshifts Stratification resulted in low numbers and, thus, only a model adjusted for age and sex was calculated (Suppl. table 2 and 3). Some differences by sex categories were observed (Suppl. Table 2). The increased risk of chronic insomnia was stronger among men [AOR ($95\%$ CI) = 2.63 (1.24–5.62) in 2017 and AOR ($95\%$ CI) = 2.78 (1.22–6.38) in 2020] than among women [AOR ($95\%$ CI) = 2.41 (1.20–4.86) in 2017 and AOR ($95\%$ CI) = 1.17 (0.57–2.41) in 2020]. In turn, the effect of meal-timing on depression was only significant among women [AOR ($95\%$ CI) = 1.87 (1.01–3.49) in 2017 and AOR ($95\%$ CI) = 1.91 (1.15–3.19) in 2020] and the effect on self-rated health status was also stronger among them [AOR ($95\%$ CI) = 3.16 (1.44–6.94) in 2017 and AOR ($95\%$ CI) = 3.02 (1.27–7.18) in 2020]. In the analysis stratified by chronotypes (Suppl. Table 3), the higher risk of depression in cluster groups B was stronger among subjects with early chronotypes [AOR ($95\%$ CI) = 2.01 (0.95–4.22) in 2017 and AOR ($95\%$ CI) = 1.98 (1.12–3.50) in 2020], while the effect on health-status was stronger among those with late chronotype [AOR ($95\%$ CI) = 2.51 (1.22–5.17) in 2017 and AOR ($95\%$ CI) = 3.35 (1.51–7.43) in 2020].
Results from analysis restricting to never night shift workers (Suppl. table 4) and to participants without report of heavy alcohol drinking (Suppl. table 5) were similar to the main analysis.
## Discussion
In this study, we provide a description of meal-timing patterns in the Austrian population in 2017 (prepandemic) and in 2020 (during the first COVID-19 mitigation measures), and explored the associations between meal-timing patterns and health outcomes. In both surveys, participants reported eating between 7:30 (median breakfast time) and 18:30 (median dinnertime) or, for those who reported having a snack after dinner, around 21:00 (median snack after dinner time) during the week. Lunch was eaten around 12:30 (median) and breakfast was skipped by about $25\%$ of Austrians on weekdays. In 2020, all main meals were eaten later on weekends than during the week. We observed long nighttime fasting periods and a frequency of 3 eating occasions a day in both surveys. We found correlation between meal-timing variables, and therefore, performed cluster analysis in each survey to group participants according to different meal timing behavioural patterns. The results of the cluster analysis were similar in 2017 and 2020: one group (A17 or A20) was formed by the majority of the participants, who reported long fasting periods and early mealtimes; the remaining participants comprised groups B17 and B20, characterized by even longer fasting intervals, later mealtimes and a high proportion of breakfast skippers.
Huseinovic et al. [ 29] described a north–south gradient of meal-times in Europe, with Scandinavian countries eating earlier and Mediterranean countries later. According to this, meal-times in the Austrian population are similar to those reported for countries like Germany or Denmark. On the other hand, differences in meal-timing between countries could also be explained by differences in longitude between countries within the same time zone. Austria is located in the extreme east of the Central European time zone, with earlier sunrise and sunset throughout the year. This could explain why Austrian meal-times are earlier compared to countries that are in the same latitude but situated in the western extreme of their time zone and, thus, experience later sun time (e.g. France, Spain). Huseinovic et al. [ 29] also showed that countries in Central and Northern Europe tend to consume more calories later in the day than earlier on. This is in line with our finding that around one quarter of the Austrians skip breakfast during the week and around one fifth during weekends. However, this proportion is strikingly higher than the one described in a cross-sectional study in Poland during the COVID-19 mitigation measures [22]. In this study, with a younger study population than ours and an under-representation of men, and in which individuals working on a regular basis during the lockdown were excluded, only $1.2\%$ of the participants reported never eating breakfast.
The number of eating occasions (median = 3 in both surveys) reported in our samples was rather low, compared to the numbers described in other countries. Based on self-reported data, Americans have about 5.6 meals a day [30] and a study carried out in five European countries reported an eating frequency that ranked between 4.3 (France) and 7.1 (The Netherlands) eating occasions/day [31]. According to Huseinovic et al. [ 29], people living in Northern and Central Europe eat more frequently than those in southern, Mediterranean countries do. The low eating frequency reported in our study might be partly explained by the way we assessed exposure. In our survey, information on maximal 6 eating episodes was obtained (3 main meals and 3 snacks) and participants could not report having two or more snacks between two main meals or after dinner, which could have resulted in underreporting of eating occasions.
Suprisingly, meal-timing and frequency during the COVID-19 mitigation measures was very similar to in 2017. The results on eating frequency oppose studies from other countries that have described an increase in snacking frequency during lockdowns [23, 25]. In a survey conducted in a representative sample of the population of Jordan [24], participants were more likely to have breakfast, lunch and dinner during the lockdown than before mitigation measures. Furthermore, a study conducted among students in Peru [26] found delays in meal-times and eating-midpoint and a reduction of the nighttime fasting interval after 12 weeks of lockdown measures. Some of these inconsistent findings might be due to differences in the formulation of the meal-timing questions and the fact that we compared the results of the 2020 survey with prepandemic information collected in a different population sample and, thus, intra-individual comparisons could not be performed. Besides, lockdown measures differed among countries and these differences might also explain the discrepancies with our findings.
We also explored the interrelationship between meal-timing and frequency variables and observed moderate correlation between the different variables considered in our study. The association between longer nighttime fasting and skipping breakfast or having an earlier last meal of the day is evident, and this interdependence of meal-timing aspects has also been described previously [32, 33]. Therefore, there has been a claim for analysis focusing on behaviours rather than isolated aspects [21]. This was our motivation for using cluster analysis to group participants according to their combined patterns of meal-timing. Indeed, by doing so, we identified two distinct meal-timing clusters in both surveys that were associated with different insomnia and chronic disease prevalences.
Our exploratory analysis across clusters revealed that insomnia symptoms were more prevalent among participants in groups B17 and B20. Prevalence of depression was also consistently higher in clusters B than A. Besides, in both surveys, participants with later mealtimes and longer fasting periods were slightly more likely to report being obese. There was also an association of meal-timing behaviours with self-rated health status, which was weaker in 2017 than in 2020. These results suggest that longer fasting intervals are not beneficial under all circumstances and that, in everyday life, longer fasting intervals might be the consequence of skipping mealtimes earlier in the day, such as breakfast. Therefore, some of the beneficial health effects usually attributed to fasting in controlled conditions might be reduced or reverted in real-life settings.
Althoug any study had evaluated meal-timing behavioural clusters in relation to chronic diseases before, several studies have analysed the association between individual aspects of meal-timing and chronic disease outcomes. Traditionally, eating breakfast and having several smaller meals in a day have been considered healthy behaviours [34]. Indeed, recent meta-analyses of observational studies concluded that skipping breakfast might be associated with higher risk of diabetes [35], depression [36], overweight and obesity [37]. However, this last assumption has been challenged by a meta-analysis of randomized trials, in which breakfast skipping was associated with modest weight loss [38]. Moreover, Schwingshackl et al. [ 39] showed that the evidence from randomized control trials did not support the belief that a higher meal frequency contributes to weight control. Concurrently, there has been a shift of paradigm and evidence from experimental studies is emerging, showing prolonged fasting intervals (≥ 12 h), like the ones reported in our samples, might be a successful strategy to reduce body weight [40] or blood pressure [41]. On top of that, not only the duration of the fasting window, but also aligning the eating window with the day and the fasting window with the night might contribute to better health outcomes, through the synchronization of central and peripheral clocks [34]. Eating during the dark phase might also negatively affect sleep, as signalled by a meta-analysis of studies conducted during Ramadan [14]. However, it is unclear how all these interrelated meal-timing aspects interact with each other and affect human health, and studies analysing the combined effects of fasting and meal-frequency and timing are lacking.
To our knowledge, this is the first study to analyse meal-timing as a behaviour, rather than isolated aspects of meal-timing, using cluster analysis to characterize meal-timing behaviours in real-life settings. It is also the first study to describe meal-timing in a representative sample of the Austrian population. Another strength is that our results are consistent in the two samples analysed. The inclusion of two identical surveys administered at two time points in representative samples of Austrians, allowed the comparison of data obtained before and during the pandemic. The analysis was carefully adjusted for a wide range of potential confounders. An additional strength is the definition used for chronic insomnia, which was based on criteria suggested by the ISCD-III [27]. Our analysis also has several limitations. First, as this is a cross sectional study, we are unable to comment on the direction of the described associations. Second, we performed several statistical analysis and, therefore, our results on the association between meal-timing and health outcomes are subject to multiple testing errors. This was an exploratory analysis and results should be interpreted with caution. Third, our surveys did not collect information on food content, liquid intake or diet type. Fourth, we might have underestimated the number of eating occasions, as a maximum of one snack could be reported between two main meals in our surveys. Last, information on meal-times was self-reported. Studies using objective meal-timing measures show that human eating patterns are erratic and not well captured through self-reports [42, 43], so under- or over-reporting of eating occasions cannot be discarded in our study and the long fasting intervals observed in our sample might be a consequence of this. Moreover, the questions used for exposure assessment had not been validated against objective measures, as, to our knowledge, at the time the surveys were conducted, no validated meal-timing questionnaire existed. This lack of validated tools also limits the comparability of our results with other studies. Recently, some studies have been published showing moderate agreement between short meal-timing questionnaires and the Automated Self-Administered 24-h recall (ASA24®) Dietary Assessment tool [44] and prospective food records [45]. Future studies on meal-timing with improved and validated reporting methods are, therefore, necessary to confirm our results.
In conclusion, our results suggest that Austrians have earlier meal-times, longer fasting intervals and lower eating frequency than most European countries, although these results might be partly due to an under-reporting of eating occasions. In Austria, meal-timing habits barely changed during the first COVID-19 mitigation measures. Individual meal-timing aspects were highly correlated between them and cluster analysis revealed two well-differentiated groups with different meal-timing behaviours in both surveys. Future epidemiologic studies with improved reporting (e.g. validated questionnaires and objective measures of meal-timing) and analytical methods (e.g. cluster analysis of behavioural patterns) are warranted to evaluate the impact of meal-timing on chronic disease risk.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 234 KB)
## References
1. Ward EM, Germolec D, Kogevinas M, McCormick D, Vermeulen R, Anisimov VN, Aronson KJ, Bhatti P, Cocco P, Costa G, Dorman DC, Fu L, Garde AH, Guénel P, Hansen J, Härmä MI, Kawai K, Khizkhin EA, Knutsson A, Lévi F, Moreno CRC, Pukkala E, Schernhammer E, Travis R, Waters M, Yakubovskaya M, Zeeb H, Zhu Y, Zienolddiny S, Grosse Y, Hall AL, Benbrahim-Tallaa L, Girschik J, Bouvard V, El Ghissassi F, Turner MC, Diver WR, Herceg Z, Olson N, Rowan EG, Rumgay H, Guyton KZ, Schubauer-Berigan MK. **Carcinogenicity of night shift work**. *Lancet Oncol* (2019.0) **20** 1058-1059. DOI: 10.1016/s1470-2045(19)30455-3
2. Vetter C, Devore EE, Wegrzyn LR, Massa J, Speizer FE, Kawachi I, Rosner B, Stampfer MJ, Schernhammer ES. **Association between rotating night shift work and risk of coronary heart disease among women**. *JAMA* (2016.0) **315** 1726-1734. DOI: 10.1001/jama.2016.4454
3. Wang D, Ruan W, Chen Z, Peng Y, Li W. **Shift work and risk of cardiovascular disease morbidity and mortality: A dose-response meta-analysis of cohort studies**. *Eur J Prev Cardiol* (2018.0) **25** 1293-1302. DOI: 10.1177/2047487318783892
4. Liu Q, Shi J, Duan P, Liu B, Li T, Wang C, Li H, Yang T, Gan Y, Wang X, Cao S, Lu Z. **Is shift work associated with a higher risk of overweight or obesity? A systematic review of observational studies with meta-analysis**. *Int J Epidemiol* (2018.0) **47** 1956-1971. DOI: 10.1093/ije/dyy079
5. Gan Y, Yang C, Tong X, Sun H, Cong Y, Yin X, Li L, Cao S, Dong X, Gong Y, Shi O, Deng J, Bi H, Lu Z. **Shift work and diabetes mellitus: a meta-analysis of observational studies**. *Occup Environ Med* (2015.0) **72** 72-78. DOI: 10.1136/oemed-2014-102150
6. Strohmaier S, Devore EE, Zhang Y, Schernhammer ES. **A review of data of findings on night shift work and the development of DM and CVD Events: a synthesis of the proposed molecular mechanisms**. *Curr Diabetes Rep* (2018.0). DOI: 10.1007/s11892-018-1102-5
7. Bonmati-Carrion M, Arguelles-Prieto R, Martinez-Madrid M, Reiter R, Hardeland R, Rol M, Madrid J. **Protecting the melatonin rhythm through circadian healthy light exposure**. *Int J Mol Sci* (2014.0) **15** 23448-23500. DOI: 10.3390/ijms151223448
8. Pot GK. **Sleep and dietary habits in the urban environment: the role of chrono-nutrition**. *Proc Nutr Soc* (2018.0) **77** 189-198. DOI: 10.1017/S0029665117003974
9. Castro MA, Garcez MR, Pereira JL, Fisberg RM. **Eating behaviours and dietary intake associations with self-reported sleep duration of free-living Brazilian adults**. *Appetite* (2019.0) **137** 207-217. DOI: 10.1016/j.appet.2019.02.020
10. Kogevinas M, Espinosa A, Castello A, Gomez-Acebo I, Guevara M, Martin V, Amiano P, Alguacil J, Peiro R, Moreno V, Costas L, Fernandez-Tardon G, Jimenez JJ, Marcos-Gragera R, Perez-Gomez B, Llorca J, Moreno-Iribas C, Fernandez-Villa T, Oribe M, Aragones N, Papantoniou K, Pollan M, Castano-Vinyals G, Romaguera D. **Effect of mistimed eating patterns on breast and prostate cancer risk (MCC-Spain Study)**. *Int J Cancer* (2018.0) **143** 2380-2389. DOI: 10.1002/ijc.31649
11. Srour B, Plancoulaine S, Andreeva VA, Fassier P, Julia C, Galan P, Hercberg S, Deschasaux M, Latino-Martel P, Touvier M. **Circadian nutritional behaviours and cancer risk: New insights from the NutriNet-santé prospective cohort study: Disclaimers**. *Int J Cancer* (2018.0) **143** 2369-2379. DOI: 10.1002/ijc.31584
12. Palomar-Cros A, Espinosa A, Straif K, Pérez-Gómez B, Papantoniou K, Gómez-Acebo I, Molina-Barceló A, Olmedo-Requena R, Alguacil J, Fernández-Tardón G, Casabonne D, Aragonés N, Castaño-Vinyals G, Pollán M, Romaguera D, Kogevinas M. **The Association of nighttime fasting duration and prostate cancer risk: results from the multicase-control (MCC) Study in Spain**. *Nutrients* (2021.0) **13** 2662. DOI: 10.3390/nu13082662
13. Allison KC, Goel N. **Timing of eating in adults across the weight spectrum: Metabolic factors and potential circadian mechanisms**. *Physiol Behav* (2018.0) **192** 158-166. DOI: 10.1016/j.physbeh.2018.02.047
14. Faris MAE, Jahrami HA, Alhayki FA, Alkhawaja NA, Ali AM, Aljeeb SH, Abdulghani IH, BaHammam AS. **Effect of diurnal fasting on sleep during Ramadan: a systematic review and meta-analysis**. *Sleep Breath* (2020.0) **24** 771-782. DOI: 10.1007/s11325-019-01986-1
15. Lee SA, Park EC, Ju YJ, Lee TH, Han E, Kim TH. **Breakfast consumption and depressive mood: A focus on socioeconomic status**. *Appetite* (2017.0) **114** 313-319. DOI: 10.1016/j.appet.2017.04.007
16. Lee YS, Kim TH. **Household food insecurity and breakfast skipping: Their association with depressive symptoms**. *Psychiatry Res* (2019.0) **271** 83-88. DOI: 10.1016/j.psychres.2018.11.031
17. Miki T, Eguchi M, Kuwahara K, Kochi T, Akter S, Kashino I, Hu H, Kurotani K, Kabe I, Kawakami N, Nanri A, Mizoue T. **Breakfast consumption and the risk of depressive symptoms: The Furukawa Nutrition and Health Study**. *Psychiatry Res* (2019.0) **273** 551-558. DOI: 10.1016/j.psychres.2019.01.069
18. Ren Z, Cao J, Cheng P, Shi D, Cao B, Yang G, Liang S, Du F, Su N, Yu M, Zhang C, Wang Y, Liang R, Guo L, Peng L. **Association between breakfast consumption and depressive symptoms among chinese college students: a cross-sectional and Prospective Cohort Study**. *Int J Environ Res Pub Health* (2020.0) **17** 1571. DOI: 10.3390/ijerph17051571
19. Zhu Z, Cui Y, Gong Q, Huang C, Guo F, Li W, Zhang W, Chen Y, Cheng X, Wang Y. **Frequency of breakfast consumption is inversely associated with the risk of depressive symptoms among Chinese university students: A cross-sectional study**. *PLoS One* (2019.0) **14** e0222014. DOI: 10.1371/journal.pone.0222014
20. Joo HJ, Kim GR, Park EC, Jang SI. **Association between Frequency of breakfast consumption and insulin resistance using triglyceride-glucose index: A Cross-Sectional Study of the Korea National Health and nutrition examination survey (2016–2018)**. *Int J Environ Res Pub Health* (2020.0) **17** 3322. DOI: 10.3390/ijerph17093322
21. Khanna N, Eicher-Miller HA, Boushey CJ, Gelfand SB, Delp EJ. **Temporal Dietary patterns using Kernel k-means clustering**. *ISM* (2011.0) **2011** 375-380. DOI: 10.1109/ISM.2011.68
22. Sidor A, Rzymski P. **Dietary choices and habits during COVID-19 Lockdown: experience from Poland**. *Nutrients* (2020.0) **12** 1657. DOI: 10.3390/nu12061657
23. Zeigler Z. **COVID-19 Self-quarantine and weight gain risk factors in adults**. *Curr Obes Rep* (2021.0) **10** 423-433. DOI: 10.1007/s13679-021-00449-7
24. Al-Domi H, Al-Dalaeen A, Al-Rosan S, Batarseh N, Nawaiseh H. **Healthy nutritional behavior during COVID-19 lockdown: A cross-sectional study**. *Clin Nutr ESPEN* (2021.0) **42** 132-137. DOI: 10.1016/j.clnesp.2021.02.003
25. Kriaucioniene V, Bagdonaviciene L, Rodríguez-Pérez C, Petkeviciene J. **Associations between changes in health behaviours and body weight during the COVID-19 Quarantine in Lithuania: The Lithuanian COVIDiet Study**. *Nutrients* (2020.0). DOI: 10.3390/nu12103119
26. Baquerizo-Sedano L, Chaquila JA, Aguilar L, Ordovás JM, González-Muniesa P, Garaulet M. **Anti-COVID-19 measures threaten our healthy body weight: Changes in sleep and external synchronizers of circadian clocks during confinement**. *Clin Nutr* (2022.0) **41** 2988-2995. DOI: 10.1016/j.clnu.2021.06.019
27. 27.American Academy of Sleep M (2014) International classification of sleep disorders.
28. Weitzer J, Santonja I, Degenfellner J, Yang L, Jordakieva G, Crevenna R, Seidel S, Klösch G, Schernhammer E, Papantoniou K. **Sleep complaints in former and current night shift workers: findings from two cross-sectional studies in Austria**. *Chronobiol Int* (2021.0) **38** 893-906. DOI: 10.1080/07420528.2021.1895200
29. Huseinovic E, Winkvist A, Freisling H, Slimani N, Boeing H, Buckland G, Schwingshackl L, Olsen A, Tjonneland A, Stepien M, Boutron-Ruault MC, Mancini F, Artaud F, Kuhn T, Katzke V, Trichopoulou A, Naska A, Orfanos P, Tumino R, Masala G, Krogh V, Santucci de Magistris M, Ocke MC, Brustad M, Jensen TE, Skeie G, Rodriguez-Barranco M, Huerta JM, Ardanaz E, Quiros JR, Jakszyn P, Sonestedt E, Ericson U, Wennberg M, Key TJ, Aune D, Riboli E, Weiderpass E, Berteus Forslund H. **Timing of eating across ten European countries - results from the European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study**. *Public Health Nutr* (2019.0) **22** 324-335. DOI: 10.1017/S1368980018002288
30. Hunt KJ, St Peter JV, Malek AM, Vrana-Diaz C, Marriott BP, Greenberg D. **Daily eating frequency in US Adults: associations with low-calorie sweeteners, body mass index, and nutrient intake (NHANES 2007–2016)**. *Nutrients* (2020.0) **12** 2566. DOI: 10.3390/nu12092566
31. Park MK, Freisling H, Huseinovic E, Winkvist A, Huybrechts I, Crispim SP, de Vries JHM, Geelen A, Niekerk M, van Rossum C, Slimani N. **Comparison of meal patterns across five European countries using standardized 24-h recall (GloboDiet) data from the EFCOVAL project**. *Eur J Nutr* (2018.0) **57** 1045-1057. DOI: 10.1007/s00394-017-1388-0
32. Ruddick-Collins LC, Morgan PJ, Johnstone AM. **Mealtime: A circadian disruptor and determinant of energy balance?**. *J Neuroendocrinol* (2020.0) **32** e12886. DOI: 10.1111/jne.12886
33. Kant AK, Graubard BI. **Within-person comparison of eating behaviors, time of eating, and dietary intake on days with and without breakfast: NHANES 2005–2010**. *Am J Clin Nutr* (2015.0) **102** 661-670. DOI: 10.3945/ajcn.115.110262
34. Paoli A, Tinsley G, Bianco A, Moro T. **The Influence of Meal Frequency and Timing on Health in Humans: The Role of Fasting**. *Nutrients* (2019.0) **11** 719. DOI: 10.3390/nu11040719
35. Ballon A, Neuenschwander M, Schlesinger S. **Breakfast skipping is associated with increased risk of type 2 Diabetes among Adults: a systematic review and meta-analysis of Prospective Cohort Studies**. *J Nutr* (2018.0) **149** 106-113. DOI: 10.1093/jn/nxy194
36. Zahedi H, Djalalinia S, Sadeghi O, Zare Garizi F, Asayesh H, Payab M, Zarei M, Qorbani M. **Breakfast consumption and mental health: a systematic review and meta-analysis of observational studies**. *Nutr Neurosci* (2022.0) **25** 1250-1264. DOI: 10.1080/1028415x.2020.1853411
37. Ma X, Chen Q, Pu Y, Guo M, Jiang Z, Huang W, Long Y, Xu Y. **Skipping breakfast is associated with overweight and obesity: A systematic review and meta-analysis**. *Obes Res Clin Pract* (2020.0) **14** 1-8. DOI: 10.1016/j.orcp.2019.12.002
38. Bonnet JP, Cardel MI, Cellini J, Hu FB, Guasch-Ferre M. **Breakfast Skipping, body composition, and cardiometabolic risk: a systematic review and meta-analysis of randomized trials**. *Obesity (Silver Spring)* (2020.0) **28** 1098-1109. DOI: 10.1002/oby.22791
39. Schwingshackl L, Nitschke K, Zähringer J, Bischoff K, Lohner S, Torbahn G, Schlesinger S, Schmucker C, Meerpohl JJ. **Impact of meal frequency on anthropometric outcomes: a systematic review and network meta-analysis of randomized controlled trials**. *Adv Nutr* (2020.0) **11** 1108-1122. DOI: 10.1093/advances/nmaa056
40. Liu L, Chen W, Wu D, Hu F. **Metabolic efficacy of time-restricted eating in adults: a systematic review and meta-analysis of randomized controlled trials**. *J Clin Endocrinol Metab* (2022.0) **107** 3428-3441. DOI: 10.1210/clinem/dgac570
41. Moon S, Kang J, Kim SH, Chung HS, Kim YJ, Yu JM, Cho ST, Oh CM, Kim T. **Beneficial effects of time-restricted eating on metabolic diseases: a systemic review and meta-analysis**. *Nutrients* (2020.0) **12** 1267. DOI: 10.3390/nu12051267
42. Gill S, Panda S. **A Smartphone App Reveals Erratic Diurnal Eating Patterns in Humans that Can Be Modulated for Health Benefits**. *Cell Metab* (2015.0) **22** 789-798. DOI: 10.1016/j.cmet.2015.09.005
43. Gupta NJ, Kumar V, Panda S. **A camera-phone based study reveals erratic eating pattern and disrupted daily eating-fasting cycle among adults in India**. *PLoS One* (2017.0) **12** e0172852. DOI: 10.1371/journal.pone.0172852
44. Chakradeo P, Rasmussen HE, Swanson GR, Swanson B, Fogg LF, Bishehsari F, Burgess HJ, Keshavarzian A. **Psychometric testing of a food timing questionnaire and food timing screener**. *Curr Dev Nutr* (2022.0). DOI: 10.1093/cdn/nzab148
45. Gioia SC, Guirette M, Chen A, Tucker C, Gray BE, Vetter C, Garaulet M, Scheer F, Saxena R, Dashti HS. **How accurately can we recall the timing of food intake? a comparison of food times from recall-based survey questions and daily food records**. *Curr Dev Nutr* (2022.0) **6** 60026. DOI: 10.1093/cdn/nzac002
|
---
title: 'The nonlinear correlation between the cardiometabolic index and the risk of
diabetes: A retrospective Japanese cohort study'
authors:
- Fubing Zha
- Changchun Cao
- Mengru Hong
- Huili Hou
- Qionghua Zhang
- Bin Tang
- Haofei Hu
- Yong Han
- Yibing Zan
- Yulong Wang
- Jianwen Xu
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9980900
doi: 10.3389/fendo.2023.1120277
license: CC BY 4.0
---
# The nonlinear correlation between the cardiometabolic index and the risk of diabetes: A retrospective Japanese cohort study
## Abstract
### Background
The cardiometabolic index (CMI) has been proposed as a novel indicator of cardiometabolic status. However, evidence on the relationship between CMI and diabetes mellitus (DM) risk was limited. Our study aimed to explore the relationship between CMI and DM risk among a large cohort of Japanese adults.
### Methods
This retrospective cohort study recruited 15453 Japanese adults without diabetes at baseline who underwent physical examinations at the Murakami Memorial Hospital between 2004 and 2015. Cox proportional-hazards regression was applied to evaluate the independent relationship between CMI and diabetes. Our study performed a generalized smooth curve fitting (penalized spline technique) and an additive model (GAM) to determine the non-linear relationship between CMI and DM risk. In addition, a set of sensitivity analyses and subgroup analyses were employed to evaluate the relationship between CMI and incident DM.
### Results
After adjusting for confounding covariates, CMI was positively related to the DM risk in Japanese adults (HR: 1.65, $95\%$CI: 1.43-1.90, $P \leq 0.0001$). A series of sensitivity analyses were also employed in this study to guarantee the reliability of the findings. In addition, our study discovered a non-linear association between CMI and diabetes risk. CMI’s inflection point was 1.01. A strong positive association between CMI and diabetes incidence was also discovered to the left of the inflection point (HR: 2.96, $95\%$CI: 1.96-4.46, P<<0.0001). However, their association was not significant when CMI was higher than 1.01 (HR: 1.27, $95\%$CI: 0.98-1.64, $$P \leq 0.0702$$). Interaction analysis showed that gender, BMI, habit of exercise, and smoking status interacted with CMI.
### Conclusion
Increased CMI level at baseline is associated with incident DM. The association between CMI and incident DM is also non-linear. A high CMI level is associated with an increased risk for DM when CMI is below 1.01.
## Introduction
Diabetes Mellitus (DM) is a metabolic disorder with chronic hyperglycemia. According to the International Diabetes Federation’s epidemiological data, approximately 463 million people aged 20-79 years were diagnosed with diabetes worldwide in 2019, with a prevalence of $9.3\%$ [1]. Diabetes is one of the most common metabolic diseases and imposes a heavy economic burden on patients and their countries [2]. Although DM is irreversible, it is mainly preventable [3]. In order to prevent and detect DM, it is crucial to completely comprehend the risk factors for the disease.
The distribution of body fat accumulation significantly impacts the onset of metabolic syndrome, diabetes, and insulin resistance (IR) [4, 5]. Previous research found a strong association between type 2 diabetes mellitus (T2DM) and a variety of traditional obesity indicators, including waist circumference (WC) and body mass index (BMI). Additionally, it has been proposed that the waist-to-height ratio (WHtR) is more accurate than BMI and WC for identifying cardiovascular risk, including T2DM [6, 7]. The triglycerides to high-density lipoprotein cholesterol ratio (TG/HDL-C ratio), which has been shown to affect type 2 diabetes, is another reliable and simple insulin resistance measurement (8–10). Wakabayashi I et al. [ 11] first developed the cardiometabolic index (CMI), which was the product of WHtR and TG/HDL-C and could be used to determine the risk of cardiometabolic disease and type 2 diabetes (12–14). CMI is an excellent predictor of type 2 diabetes compared to other obesity and lipid indicators, such as BMI, body mass index (BAI), WC, and triglycerides (TG) [11, 15, 16]. Furthermore, CMI is connected to some obesity-related metabolic disorders, including hyperuricemia, nonalcoholic fatty liver disease, renal disease, and stroke (17–20). Most recent studies on the relationship between CMI and diabetes have cross-sectional designs and small sample numbers. Unfortunately, neither the non-linear relationship between CMI and DM nor subgroup analyses were performed. Therefore, a retrospective cohort study was designed to observe the relationship between CMI and DM in a sizable cohort of Japanese adults. Furthermore, this study directed therapeutic practice by exploring the quantitative association between CMI and DM in different genders.
## Data source
The information was derived from this study: Takuro Okamura et al. [ 21]: Ectopic fat obesity presents the greatest risk for incident type 2 diabetes: a population-based longitudinal study. Dryad Digital Repository (https://doi.org/10.1038/s41366-018- 0076-3). Under the Dryad terms of service, researchers can use the data for secondary analysis without harming the authors following the Dryad terms of service.
## Study participants
Written informed consent was obtained from each participant in the initial study, which was carried out with the Murakami Memorial Hospital’s Clinical Research Ethics Committee [21]. Therefore, this secondary analysis did not need ethical approval. Additionally, the initial study was conducted under the Declaration of Helsinki. All procedures, including the declarations in the Declarations section, were carried out under the relevant norms and laws.
The original study initially enrolled 20944 Japanese individuals who took the physical exam between 2004 and 2015 and completed at least a second exam. Afterward, 5491 individuals were excluded, and 15453 individuals (8419 male and 7034 female) were left for data analysis of our study (Figure 1). If subjects met any of the following criteria at the baseline, they were excluded from the study: [1] type 2 diabetes; [2] known liver disease (such as hepatitis B or hepatitis C at baseline); [3] alcoholic fatty liver disease; [4] fasting plasma glucose (FPG)≥ 6.1 mmol/L; [5] missing data of variables; [6] incomplete HDL-C; [7] taking any medication.
**Figure 1:** *Study Population.*
## Covariates
In our research, covariates were chosen following our clinical expertise and prior research. The following variables were utilized as covariates: [1] categorical variables: smoking status, habit of exercise, and gender; [2] continuous variables: ethanol consumption, diastolic blood pressure (DBP), systolic blood pressure (SBP), age, BMI, alanine aminotransferase (ALT), gamma-glutamyl transferase (GGT), aspartate aminotransferase (AST), total cholesterol (TC), glycosylated hemoglobin (HbA1c), and FPG. Through the use of a standardized self-management questionnaire, information on each individual’s medical history and lifestyle characteristics were gathered for the original study. The professional staff accurately measured the subject’s height, weight, WC, and blood pressure. The original study team used a consistent process to acquire laboratory test data under controlled circumstances. The initial research evaluated ethanol consumption based on subjects’ ethanol consumption during the prior month and then calculated the average weekly ethanol consumption [21].
## Cardiometabolic index
CMI is regarded as a continuous variable. [ WC (cm)/height (cm)]×[TG (mmol/L)/HDL-C (mmol/L)] was the formula used to calculate CMI in detail [11].
## Diagnosis of incident diabetes
Diabetes was defined as glycosylated hemoglobin ≥ $6.5\%$, fasting plasma glucose ≥ 7 mmol/L [22], or self-reported during the follow-up period.
## Statistical analysis
Statistical analysis was employed using the R software package (version 3.3.1) (www.r-project.org, The R Foundation) and Empower-Stats (version 2.16.1) (www.empowerstats.com, XY Solutions, Inc., Boston, MA).
We explored the characteristics of all subjects at baseline according to quartiles of CMI. Skewed and normally distributed continuous variables were described as median (quartile) and mean ± standard deviation, respectively. The differences among the CMI groups were compared by one-way ANOVA test, Kruskal-Wallis H test, or chi-square test. Comparisons of survival and cumulative event rates were conducted using the Kaplan-Meier method. In addition, we compared the Kaplan-Meier hazard ratios (HR) of adverse events by log-rank test [23].
The multivariate Cox regression analysis was also used to explore the association of CMI with the risk of DM. Furthermore, we constructed three models to assess the association between CMI and diabetes risk: crude Model, Model I, and Model II. We only adjusted for these covariances if the hazard ratios changed by ≥$10\%$ when added to the adjusted Model [24].
Since CMI was a continuous variable, we tried to identify the nonlinear relationship between CMI and diabetes through Cox proportional hazards regression model with cubic spline functions and the smooth curve fitting (penalized spline method). If the relationship was nonlinear, we employed the two-piece linear regression model to determine the inflection point [25]. The present study used the log-likelihood ratio to describe the most appropriate Model for the association of CMI with diabetes.
The present study employed a series of sensitivity analyses to assess robust findings. We converted CMI into a categorical variable according to the quartile. Then we calculated P for the trend to verify the results of CMI as the continuous variable and examine the possibility of nonlinearity. Obesity and the elderly were associated with an increased risk of diabetes. Hence, we excluded individuals with BMI≥25kg/m2 or age≥60 years in other sensitivity analyses to assess the relationship between CMI and diabetes risk. Furthermore, the present study employed a generalized additive model (GAM) to incorporate the continuity variables into the equation as a curve to examine the robustness of our findings. To assess the impact of potential unmeasured confounding between CMI and the risk of diabetes, we further calculated E-values [26].
Moreover, we applied the Cox proportional hazard model to the subgroup analysis (ethanol consumption, habit of exercise, smoking status, BMI, age, and gender). Firstly, the interaction test between these variables and CMI was performed before the subgroup analysis. The likelihood ratio test was used to compare models with and without the multiplicative interaction term. Secondly, continuous variables, including BMI and age, were converted into categorical variables based on clinical cut-off points age (<60, ≥60 years) and BMI (<25, ≥25kg/m2). Thirdly, a fully adjusted analysis was performed for each stratum, except for the stratification factor. Ultimately, the likelihood ratio test was used to determine whether interaction terms existed in models with and without interaction terms. STROBE statement was obeyed during the whole research [24, 27]. Statistical significance was determined by $P \leq 0.05$ in two-tailed tests.
## Characteristics of participants
Table 1 presented all eligible individuals’ basic clinical measurements, biochemical tests, and other parameters. The final analysis included 15453 individuals, with a mean age of 43.71 ± 8.90 years and a male participation rate of $54.48\%$. 373 participants eventually got diabetes after a median of 5.39 years of follow-up. The mean ± SD of WC, BMI, TG, HDL-C, and CMI were 76.47 ± 9.11 cm, 22.12 ± 3.13 kg/m2, 0.91 ± 0.66 mmol/L, 1.46 ± 0.40 mmol/L and 0.36 ± 0.39. We assigned participants into subgroups using CMI quartiles (≤0.133, 0.133-0.230, 0.230-0.423, > 0.423). In the highest CMI group, individuals had higher ethanol consumption, DBP, SBP, WC, BMI, age, ALT, AST, GGT, TG, TC, HbA1c, FPG, higher rates of men, smokers, but lower HDL-C and lower rates of the habit of exercise.
**Table 1**
| CMI | Q1 (≤0.133) | Q2 (0.133 to ≤0.233) | Q3 (0.233 to ≤0.437) | Q4 (>0.437) | P-value |
| --- | --- | --- | --- | --- | --- |
| Participants | 3863 | 3863 | 3863 | 3864 | |
| Gender | | | | | <0.001 |
| Female | 2951 (76.39%) | 2136 (55.29%) | 1330 (34.43%) | 617 (15.97%) | |
| Male | 912 (23.61%) | 1727 (44.71%) | 2533 (65.57%) | 3247 (84.03%) | |
| Age(years) | 40.98 ± 8.27 | 43.23 ± 8.82 | 45.10 ± 9.05 | 45.53 ± 8.69 | <0.001 |
| Ethanol consumption(g/week) | 1 (0, 22) | 1 (0, 60) | 2.8 (0, 84) | 12(1, 90) | <0.001 |
| Smoking status | | | | | <0.001 |
| Never-smoker | 3029 (78.41%) | 2513 (65.05%) | 1980 (51.26%) | 1505 (38.95%) | |
| Ex-smoker | 465 (12.04%) | 671 (17.37%) | 849 (21.98%) | 964 (24.95%) | |
| Current-smoker | 369 (9.55%) | 679 (17.58%) | 1034 (26.77%) | 1395 (36.10%) | |
| Habit of exercise | | | | | <0.001 |
| No | 3125 (80.90%) | 3176 (82.22%) | 3155 (81.67%) | 3291 (85.17%) | |
| Yes | 738 (19.10%) | 687 (17.78%) | 708 (18.33%) | 573 (14.83%) | |
| SBP (mmHg) | 107.67 ± 12.81 | 112.10 ± 13.89 | 116.36 ± 14.46 | 121.84 ± 14.86 | <0.001 |
| DBP (mmHg) | 66.55 ± 9.06 | 69.76 ± 9.74 | 72.98 ± 9.95 | 77.03 ± 10.26 | <0.001 |
| BMI (kg/m2) | 19.98 ± 2.13 | 21.27 ± 2.47 | 22.62 ± 2.70 | 24.59 ± 3.08 | <0.001 |
| WC (cm) | 69.54 ± 6.46 | 73.83 ± 7.37 | 78.26 ± 7.48 | 84.24 ± 7.82 | <0.001 |
| ALT (IU/L) | 14 (11, 17) | 15 (12, 19) | 18 (14, 23) | 23 (17, 33) | <0.001 |
| AST (IU/L) | 16 (13, 19) | 17 (14, 20) | 17 (14, 21) | 19 (16,24) | <0.001 |
| GGT(IU/L) | 12 (10, 15) | 13 (11, 18) | 16 (12, 23) | 22 (16, 34) | |
| HDL-C (mmol/L) | 1.84 ± 0.38 | 1.57 ± 0.29 | 1.35 ± 0.25 | 1.09 ± 0.21 | <0.001 |
| TG (mmol/L) | 0.38 (0.30, 0.46) | 0.61 (0.52, 0.71) | 0.88 (0.76, 1.03) | 1.52 (1.24, 1.95) | <0.001 |
| TC (mmol/L) | 4.85 ± 0.80 | 5.00 ± 0.81 | 5.18 ± 0.85 | 5.47 ± 0.87 | <0.001 |
| HbA1c (%) | 5.13 ± 0.30 | 5.15 ± 0.31 | 5.18 ± 0.33 | 5.22 ± 0.34 | <0.001 |
| FPG (mmol/L) | 4.96 ± 0.39 | 5.10 ± 0.39 | 5.22 ± 0.39 | 5.35 ± 0.37 | <0.001 |
| CMI | 0.09 (0.07, 0.11) | 0.18 (0.15, 0.20) | 0.31 (0.27, 0.37) | 0.68 (0.53, 0.95) | <0.001 |
## The results of the relationship between CMI and incident diabetes
Table 2 revealed that 373 participants developed diabetes. The total incidence rate of all persons was 399.14 per 100,000 person-years. In particular, the incidence rate of the four CMI groups were 103.16, 160.79, 331.58, and 956.27 100,000 person-years, respectively. Participants with a high CMI level had higher incidence rates of diabetes than those with the lowest CMI level ($P \leq 0.0001$ for trend).
**Table 2**
| CMI | Participants (n) | DM events (n) | Incidence rate (per 100,000 person-year) | Crude Model (HR, 95% CI, P) | Model I (HR, 95% CI, P) | Model II (HR, 95% CI, P) | Model III (HR, 95% CI, P) | Model IV (HR, 95% CI, P) | Model V (HR, 95% CI, P) |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Total | 15453.0 | 373.0 | 399.14 | 2.22 (2.05, 2.41) <0.0001 | 1.98 (1.78, 2.19) <0.0001 | 1.65 (1.43, 1.90) <0.0001 | 1.59 (1.37, 1.84) <0.0001 | 1.50 (1.18, 1.90) 0.0010 | 1.65 (1.42, 1.91) <0.0001 |
| Q1 | 3863.0 | 103.16 | 103.16 | ref | ref | ref | ref | ref | ref |
| Q2 | 3863.0 | 160.79 | 160.79 | 1.48 (0.88, 2.51) 0.1403 | 1.17 (0.69, 1.99) 0.5711 | 0.92 (0.54, 1.58) 0.7747 | 0.91 (0.53, 1.56) 0.7276 | 0.80 (0.45, 1.43) 0.4594 | 1.09 (0.62, 1.92) 0.7683 |
| Q3 | 3863.0 | 331.58 | 331.58 | 3.04 (1.90, 4.87) <0.0001 | 1.99 (1.22, 3.25) 0.0062 | 1.23 (0.75, 2.02) 0.4065 | 1.15 (0.69, 1.91) 0.5902 | 1.12 (0.65, 1.93) 0.6736 | 1.27 (0.75, 2.17) 0.3714 |
| Q4 | 3864.0 | 956.27 | 956.27 | 8.70 (5.62, 13.48) <0.0001 | 4.79 (2.97, 7.74) <0.0001 | 2.04 (1.26, 3.32) 0.0040 | 1.86 (1.13, 3.07) 0.0153 | 1.40 (0.81, 2.44) 0.2323 | 2.18 (1.30, 3.68) 0.0034 |
| P for trend | | | <0.0001 | <0.0001 | <0.0001 | <0.0001 | 0.0003 | 0.0334 | <0.0001 |
Kaplan-Meier curves for the probability of diabetes-free survival were depicted in Figure 2. A significant difference existed between the four CMI groups regarding diabetes risk ($P \leq 0.0001$). There was a gradual decrease in the probability of diabetes-free survival as CMI levels increased. Therefore, individuals in the top CMI group had the highest risk of diabetes.
**Figure 2:** *Kaplan–Meier event-free survival curve. Kaplan–Meier event-free survival curve. Kaplan–Meier analysis of incident diabetes based on CMI quartiles (log-rank, P < 0.0001).*
Table 2 showed the Cox proportional hazard regression models, which assessed the association between CMI and diabetes risk. Both the three adjusted and unadjusted models were presented in Table 2. In the crude mode, CMI was positively correlated with diabetes (HR: 2.22, $95\%$CI: 2.05-2.41, $P \leq 0.0001$). In Model I (adjusting for gender, ethanol consumption, habit of exercise, smoking status, age, DBP, and SBP), the findings did not have apparent changes (HR: 1.98, $95\%$CI: 1.78-2.19, $P \leq 0.0001$). In Model II (adjusting for gender, ethanol consumption, habit of exercise, smoking status, age, DBP, SBP, AST, GGT, ALT, TC, FPG, and HbA1c), the risk of DM increased by $65\%$ for each unit increase in CMI (HR: 1.65, $95\%$CI: 1.43-1.90, $P \leq 0.0001$).
## Sensitive analysis
To examine the robustness of our conclusions, we employed a series of sensitivity analyses. CMI was transformed into a categorical variable (based on quartile) and reinserted into the models. Compared with the Q1 group, the adjusted HR ($95\%$ CI) for the Q4 group was 2.04 (1.26-3.32). Furthermore, when CMI was transformed into a categorical variable, the trends of HR were not equal, suggesting a possible nonlinear association between CMI and diabetes risk. Additionally, the continuity covariate was inserted into the equation by a GAM. The findings of Model III were consistent with the results of Model II (HR: 1.59, $95\%$CI:1.37-1.84, $P \leq 0.0001$) (Table 2). Besides, to evaluate the impact of potential unmeasured confounding between CMI and the risk of diabetes, this study further generated E-values. The E value for this study was 2.69 ($95\%$CI: 2.65-2.72). Compared to the relative risk and CMI for unmeasured confounding variables, the E value for this study was larger. The results indicated that unknown or unmeasured confounding variables hardly affected the association between CMI and the risk of diabetes.
In addition, individuals with a BMI ≥ 25kg/m2 were excluded from other sensitivity analyses. In Model IV, after adjusting gender, ethanol consumption, habit of exercise, smoking status, age, DBP, and SBP, AST, GGT, ALT, TC, FPG, HbA1c, there was also a positive association between CMI and diabetes risk (HR: 1.50, $95\%$CI: 1.18-1.90) (Table 2). Individuals with age ≥ 60 years were also excluded from other sensitivity analyses. The results suggested that after adjusting for gender, ethanol consumption, habit of exercise, smoking status, DBP, SBP, AST, GGT, ALT, TC, FPG, and HbA1c, CMI was still positively correlated with diabetes risk (HR: 1.65, $95\%$CI: 1.42-1.91) in Model V (Table 2). Based on the sensitivity analyses, our findings were well-robust.
## The analyses of the non-linear relationship
The GAM and the smooth curve fitting (penalty curve method) were applied to verify the nonlinearity in the association between CMI and the risk of diabetes (Figure 3). Table S1 revealed a nonlinear association between CMI and diabetes after adjusting for gender, ethanol consumption, habit of exercise, smoking status, age, DBP, SBP, AST, GGT, ALT, TC, FPG, and HbA1c. According to a two-piecewise linear regression model, the present study observed the inflection point of CMI was 1.01 (P for the log-likelihood ratio test= 0.003). When CMI was lower than 1.01, CMI was positively associated with diabetes risk (HR: 2.96, $95\%$CI: 1.96-4.46, P<<0.0001). In contrast, when CMI was above 1.01, their association was not significant (HR: 1.27, $95\%$CI: 0.98-1.64, $$P \leq 0.0702$$).
**Figure 3:** *The nonlinear relationship between CMI ratio and incident diabetes. A nonlinear relationship was detected after adjusting for gender, age, ethanol consumption, smoking status, habit of exercise, SBP, DBP, ALT, AST, GGT, TC, HbA1c, and FPG.*
## The results of the subgroup analysis
Interaction tests performed before subgroup analyses showed that gender, BMI, habit of exercise, and smoking status interacted with CMI ($P \leq 0.001$). In contrast, age and ethanol consumption did not interact with CMI ($P \leq 0.05$) (Table S2). Therefore, further subgroup analyses with gender, BMI, habit of exercise, and smoking status were performed (Figure 4). Specifically, a stronger relationship between CMI and diabetes risk was observed in participants with female, current-smoker, lack of exercise habits, and BMI ≥ 25kg/m2. In contrast, a weaker relationship between CMI and diabetes risk was observed in participants with male, habit of exercise, never-smoker, ex-smoker, and BMI < 25kg/m2.
**Figure 4:** *Subgroups analysis. Specifically, a stronger relationship between CMI and diabetes risk was observed in participants with female, current-smoker, lack of exercise habits, and BMI≥25kg/m2. In contrast, a weaker relationship between CMI and diabetes risk was observed in participants with male, habit of exercise, never-smoker, ex-smoker, and BMI< 25kg/m2.*
## Discussion
Our retrospective study investigated the association between CMI and the risk of developing diabetes in Japanese individuals. This study showed that higher CMI was associated with a higher risk of diabetes. The association between CMI on diabetes was also examined on the left and right sides of the inflection point. CMI level had a nonlinear association with incident diabetes. When CMI was below 1.01, we discovered a significant positive correlation between CMI and diabetes incidence (HR: 2.96, $95\%$CI: 1.96-4.46, $P \leq 0.0001$). However, their association was not significant when CMI was higher than 1.01 (HR: 1.27, $95\%$CI: 0.98-1.64, $$P \leq 0.0702$$). Interaction analysis showed that gender, BMI, habit of exercise, and smoking status interacted with CMI.
CMI is a newly proposed indicator related to cardiovascular risk factors, confirming its utility in the early detection of associated cardiovascular disorders [12, 14, 28]. CMI could also identify the status of DM more accurately [11]. CMI should be considered a combination of dyslipidemia and obesity, because it combines TG/HDL-C and WHtR and can be used as a valid differential indicator for diabetes. Our results are consistent with those of the following studies. In a cross-sectional study involving 11478 individuals in rural Northeast China, Shi WR et al. [ 29] demonstrated an association between elevated CMI and risk of developing diabetes after adjusting for race, age, income level, education level, medication usage, vegetable intake, meat intake, fatty food after intake, physical activity, family history of DM, hypertension, history of cardiovascular diseases, drinking status, and current smoking. In another cross-sectional study of 10196 subjects, Wakabayashi I et al. [ 11] discovered that after adjusting for age and histories of regular exercise, alcohol drinking, and smoking, there was a stronger association between CMI and diabetes. A longitudinal study of 7347 middle-aged and elderly Chinese showed that individuals with high CMI had a far higher risk of developing type 2 diabetes after adjusting for region, sex, age, marital status, ln(per capita expenditures), education, hypertension, drinking, smoking, low-density lipoprotein cholesterol (LDL-C), and TC [13]. This study involved 15453 Japanese adults, and we found that the incidence of diabetes was higher with increased CMI levels. After adjusting for gender, ethanol consumption, habit of exercise, smoking status, age, DBP, SBP, AST, GGT, ALT, TC, FPG, and HbA1c, the results showed that the risk of DM increased by $65\%$ for each unit increase in CMI. Furthermore, the sensitivity studies showed that this association could still be observed in Japanese adults with BMI < 25 kg/m2 or age < 60 years. Our study included a larger population compared to those of previous studies. In addition, we adjusted for more covariates, such as SBP, DBP, GGT, FPG, and HbA1c, which were all critical risk factors for diabetes. More importantly, we used sensitivity and subgroup analysis methods to validate further the solidity of the association between CMI and diabetes. In short, our results further confirmed the positive association between CMI and diabetes risk in the Japanese population. This study provided supporting evidence for clinical interventions for CMI levels to reduce the risk of diabetes.
To our knowledge, previous studies have not explored a possible curvilinear relationship between CMI and DM. The nonlinear association between CMI and DM in various genders was first examined in the current study. After adjusting for gender, ethanol consumption, habit of exercise, smoking status, age, DBP, SBP, AST, GGT, ALT, TC, FPG, and HbA1c, the smooth curve result revealed that the association between CMI and DM was nonlinear. We determined the CMI inflection point using a two-piecewise linear regression model. When the CMI level was lower than 1.01, the risk of DM increased by $196\%$ for each unit increase in CMI (HR: 2.96, $95\%$CI: 1.96-4.46, P<<0.0001). CMI was not associated with incident DM when CMI was higher than 1.01 (HR: 1.27, $95\%$CI: 0.98-1.64, $$P \leq 0.0702$$). Elevated CMI will indicate that participants have an increased risk of developing diabetes during follow-up, alerting people to make early changes in lifestyle habits to improve outcomes.
The mechanism by which CMI leads to the development of diabetes remains unclear. The aberrant lipid metabolism may account for these results in individuals with assessed CMI. In obese individuals with high WHtR, excess free fatty acids can impair insulin’s ability to play a role in glucose metabolism and lead to insulin resistance [30]. In individuals with abdominal obesity, the number of insulin receptors on target tissues and the binding affinity decrease, resulting in a reduced ability to process glucose [31, 32]. Meanwhile, elevated TG status contributes to the development of diabetes in a manner similar to abdominal obesity. It could be viewed as a crucial transitional stage from obesity to diabetes mellitus. Reduced HDL-C levels may negatively affect β cells’ ability, decreasing insulin sensitivity and output [32, 33].
Several strengths of our study can be found. First, we examined the nonlinear relationship using a GAM and a smooth curve fitting to identify the optimal inflection point for the effect of CMI on diabetes. Second, results were also rigorously adjusted statistically to reduce the influence of confounding factors, ensuring our findings’ reliability. Third, our results were tested for robustness through sensitivity analyses (CMI transformation, employing a GAM to put the continuity covariate into the equation as a curve, employing a GAM to put the continuity covariate into the equation as a curve, estimating E-values to examine the potential for unmeasured confounders, subgroup analysis, and reassessing the relationship between CMI on diabetes after excluding participants with BMI ≥ 25kg/m2 or age ≥ 60 years) to ensure their reliability. Fourth, we employed a subgroup analysis to find other risk factors that might affect the relationship between CMI and diabetes.
There are still some limitations in our study. First, our study includes only the Japanese population. As a result, other geographic and racial groups cannot use the study’s findings. Second, because individuals with excessive drinking habits, viral hepatitis, or drug use were excluded from the study, our findings may not apply to the broader population. In the future, we can consider designing our research or cooperation with other researchers to collect as many people as possible, including lack of data, excessive drinking habits, viral hepatitis, or drug use. Third, similar to the characteristics of all retrospective studies, our study may have had unmeasured or uncontrolled confounding covariates such as education, income, marital status, dietary factors and family history of diabetes. However, E values were calculated to quantify the potential impact of unmeasured confounding covariates. Unmeasured confounding covariates were unlikely to influence the relationship between CMI and the risk of diabetes. Fourth, the initial study only measured baseline WHtR, HDL-C, and TG. In addition, the initial investigation did not address changes in WHtR, HDL-C, and TG over time. In the future, we will consider designing our study to document more confounding factors, including education, income, marital status, dietary factors, family history of diabetes, and fluctuations in TG, WHtR, and HDL-C during follow-up. Therefore, we could explore the impact of changes in CMI on future diabetes risk through a GAM model.
## Conclusion
This cohort study shows a non-linear association between CMI and diabetes in the Japanese population. There is a strong positive association between CMI and the risk of developing diabetes when CMI is less than 1.01. These data provide strong evidence to enhance the value of CMI in evaluating diabetes.
## 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://datadryad.org/stash/dataset/doi:$10.5061\%$2Fdryad.8q0p192.
## Ethics statement
The studies involving human participants were reviewed and approved by the Institutional Review Board of the Murakami Memorial Hospital. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
FZ and CC contributed to the study concept and design, researched and interpreted the data, and drafted the manuscript. MH, HLH, and QZ examined the data and reviewed the manuscript. BT, HFH, YH, and YZ oversaw the project’s progress, contributed to the discussion and reviewed the manuscript. YW and JX are the guarantors of this work and, as such, had full access to all the data in the study and took 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.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1120277/full#supplementary-material
## References
1. Saeedi P, Petersohn I, Salpea P, Malanda B, Karuranga S, Unwin N. **Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: Results from the international diabetes federation diabetes atlas, 9(th) edition**. *Diabetes Res Clin Pract* (2019) **157**. DOI: 10.1016/j.diabres.2019.107843
2. 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** 14790. DOI: 10.1038/s41598-020-71908-9
3. Wu Y, Hu H, Cai J, Chen R, Zuo X, Cheng H. **A prediction nomogram for the 3-year risk of incident diabetes among Chinese adults**. *Sci Rep* (2020) **10** 21716. DOI: 10.1038/s41598-020-78716-1
4. Wang K, Gong M, Xie S, Zhang M, Zheng H, Zhao X. **Nomogram prediction for the 3-year risk of type 2 diabetes in healthy mainland China residents**. *EPMA J* (2019) **10**. DOI: 10.1007/s13167-019-00181-2
5. Golubnitschaja O, Costigliola V. **General report & recommendations in predictive, preventive and personalised medicine 2012: White paper of the European association for predictive, preventive and personalised medicine**. *EPMA J* (2012) **3**. DOI: 10.1186/1878-5085-3-14
6. Correa MM, Thume E, De Oliveira ER, Tomasi E. **Performance of the waist-to-height ratio in identifying obesity and predicting non-communicable diseases in the elderly population: A systematic literature review**. *Arch Gerontol Geriatr* (2016) **65**. DOI: 10.1016/j.archger.2016.03.021
7. 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) **13**. DOI: 10.1111/j.1467-789X.2011.00952.x
8. Chen Z, Hu H, Chen M, Luo X, Yao W, Liang Q. **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
9. Lim TK, Lee HS, Lee YJ. **Triglyceride to HDL-cholesterol ratio and the incidence risk of type 2 diabetes in community dwelling adults: A longitudinal 12-year analysis of the Korean genome and epidemiology study**. *Diabetes Res Clin Pract* (2020) **163**. DOI: 10.1016/j.diabres.2020.108150
10. Zheng D, Li H, Ai F, Sun F, Singh M, Cao X. **Association between the triglyceride to high-density lipoprotein cholesterol ratio and the risk of type 2 diabetes mellitus among Chinese elderly: The Beijing longitudinal study of aging**. *BMJ Open Diabetes Res Care* (2020) **8**. DOI: 10.1136/bmjdrc-2019-000811
11. Wakabayashi I, Daimon T. **The "cardiometabolic index" as a new marker determined by adiposity and blood lipids for discrimination of diabetes mellitus**. *Clin Chim Acta* (2015) **438**. DOI: 10.1016/j.cca.2014.08.042
12. Higashiyama A, Wakabayashi I, Okamura T, Kokubo Y, Watanabe M, Takegami M. **The risk of fasting triglycerides and its related indices for ischemic cardiovascular diseases in Japanese community dwellers: the suita study**. *J Atheroscler Thromb* (2021) **28**. DOI: 10.5551/jat.62730
13. Qiu Y, Yi Q, Li S, Sun W, Ren Z, Shen Y. **Transition of cardiometabolic status and the risk of type 2 diabetes mellitus among middle-aged and older Chinese: A national cohort study**. *J Diabetes Investig* (2022) **13**. DOI: 10.1111/jdi.13805
14. Acosta-Garcia E, Concepcion-Paez M. **[Cardiometabolic index as a predictor of cardiovascular risk factors in adolescents]**. *Rev Salud Publica (Bogota)* (2018) **20**. DOI: 10.15446/rsap.V20n3.61259
15. Liu X, Wu Q, Yan G, Duan J, Chen Z, Yang P. **Cardiometabolic index: a new tool for screening the metabolically obese normal weight phenotype**. *J Endocrinol Invest* (2021) **44**. DOI: 10.1007/s40618-020-01417-z
16. Wang Z, He S, Chen X. **Capacity of different anthropometric measures to predict diabetes in a Chinese population in southwest China: A 15-year prospective study**. *Diabetes Med* (2019) **36**. DOI: 10.1111/dme.14055
17. Liu Y, Wang W. **Sex-specific contribution of lipid accumulation product and cardiometabolic index in the identification of nonalcoholic fatty liver disease among Chinese adults**. *Lipids Health Dis* (2022) **21**. DOI: 10.1186/s12944-021-01617-3
18. Zuo YQ, Gao ZH, Yin YL, Yang X, Feng PY. **Association between the cardiometabolic index and hyperuricemia in an asymptomatic population with normal body mass index**. *Int J Gen Med* (2021) **14**. DOI: 10.2147/IJGM.S340595
19. Li FE, Luo Y, Zhang FL, Zhang P, Liu D, Ta S. **Association between cardiometabolic index and stroke: A population- based cross-sectional study**. *Curr Neurovasc Res* (2021) **18**. DOI: 10.2174/1567202618666211013123557
20. Wang HY, Shi WR, Yi X, Wang SZ, Luan SY, Sun YX. **Value of reduced glomerular filtration rate assessment with cardiometabolic index: insights from a population-based Chinese cohort**. *BMC Nephrol* (2018) **19** 294. DOI: 10.1186/s12882-018-1098-8
21. Okamura T, Hashimoto Y, Hamaguchi M, Obora A, Kojima T, Fukui M. **Ectopic fat obesity presents the greatest risk for incident type 2 diabetes: a population-based longitudinal study**. *Int J Obes (Lond)* (2019) **43**. DOI: 10.1038/s41366-018-0076-3
22. **2. Classification and diagnosis of diabetes: Standards of medical care in diabetes-2021**. *Diabetes Care* (2021) **44**. DOI: 10.2337/dc21-S002
23. Robson ME, Tung N, Conte P, Im SA, Senkus E, Xu B. **OlympiAD final overall survival and tolerability results: Olaparib versus chemotherapy treatment of physician's choice in patients with a germline BRCA mutation and HER2-negative metastatic breast cancer**. *Ann Oncol* (2019) **30**. DOI: 10.1093/annonc/mdz012
24. Skrivankova VW, Richmond RC, Woolf B, Davies NM, Swanson SA, VanderWeele TJ. **Strengthening the reporting of observational studies in epidemiology using mendelian randomisation (STROBE-MR): explanation and elaboration**. *BMJ* (2021) **375**. DOI: 10.1136/bmj.n2233
25. Zhu F, Chen C, Zhang Y, Chen S, Huang X, Li J. **Elevated blood mercury level has a non-linear association with infertility in U.S. women: Data from the NHANES 2013-2016**. *Reprod Toxicol* (2020) **91**. DOI: 10.1016/j.reprotox.2019.11.005
26. Haneuse S, VanderWeele TJ, Arterburn D. **Using the e-value to assess the potential effect of unmeasured confounding in observational studies**. *JAMA* (2019) **321**. DOI: 10.1001/jama.2018.21554
27. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche 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
28. Wang H, Chen Y, Sun G, Jia P, Qian H, Sun Y. **Validity of cardiometabolic index, lipid accumulation product, and body adiposity index in predicting the risk of hypertension in Chinese population**. *Postgrad Med* (2018) **130**. DOI: 10.1080/00325481.2018.1444901
29. Shi WR, Wang HY, Chen S, Guo XF, Li Z, Sun YX. **Estimate of prevalent diabetes from cardiometabolic index in general Chinese population: a community-based study**. *Lipids Health Dis* (2018) **17** 236. DOI: 10.1186/s12944-018-0886-2
30. Kahn SE, Hull RL, Utzschneider KM. **Mechanisms linking obesity to insulin resistance and type 2 diabetes**. *Nature* (2006) **444**. DOI: 10.1038/nature05482
31. Barazzoni R, Gortan CG, Ragni M, Nisoli E. **Insulin resistance in obesity: an overview of fundamental alterations**. *Eat Weight Disord* (2018) **23**. DOI: 10.1007/s40519-018-0481-6
32. Sharma AM, Lau DC. **Obesity and type 2 diabetes mellitus**. *Can J Diabetes* (2013) **37** 63-4. DOI: 10.1016/j.jcjd.2013.03.360
33. Goodpaster BH, Kelley DE. **Skeletal muscle triglyceride: marker or mediator of obesity-induced insulin resistance in type 2 diabetes mellitus**. *Curr Diabetes Rep* (2002) **2**. DOI: 10.1007/s11892-002-0086-2
|
---
title: 'Fatty acids and risk of dilated cardiomyopathy: A two-sample Mendelian randomization
study'
authors:
- Jiexin Zhang
- Qiang Luo
- Jun Hou
- Wenjing Xiao
- Pan Long
- Yonghe Hu
- Xin Chen
- Han Wang
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9980906
doi: 10.3389/fnut.2023.1068050
license: CC BY 4.0
---
# Fatty acids and risk of dilated cardiomyopathy: A two-sample Mendelian randomization study
## Abstract
### Background
Previous observational studies have shown intimate associations between fatty acids (FAs) and dilated cardiomyopathy (DCM). However, due to the confounding factors and reverse causal association found in observational epidemiological studies, the etiological explanation is not credible.
### Objective
To exclude possible confounding factors and reverse causal associations found in observational epidemiological studies, we used the two-sample Mendelian randomization (MR) analysis to verify the causal relationship between FAs and DCM risk.
### Method
All data of 54 FAs were downloaded from the genome-wide association studies (GWAS) catalog, and the summary statistics of DCM were extracted from the HF Molecular Epidemiology for Therapeutic Targets Consortium GWAS. Two-sample MR analysis was conducted to evaluate the causal effect of FAs on DCM risk through several analytical methods, including MR-Egger, inverse variance weighting (IVW), maximum likelihood, weighted median estimator (WME), and the MR pleiotropy residual sum and outlier test (MRPRESSO). Directionality tests using MR-Steiger to assess the possibility of reverse causation.
### Results
Our analysis identified two FAs, oleic acid and fatty acid (18:1)-OH, that may have a significant causal effect on DCM. MR analyses indicated that oleic acid was suggestively associated with a heightened risk of DCM (OR = 1.291, $95\%$CI: 1.044–1.595, $$P \leq 0.018$$). As a probable metabolite of oleic acid, fatty acid (18:1)-OH has a suggestive association with a lower risk of DCM (OR = 0.402, $95\%$CI: 0.167–0.966, $$P \leq 0.041$$). The results of the directionality test suggested that there was no reverse causality between exposure and outcome ($P \leq 0.001$). In contrast, the other 52 available FAs were discovered to have no significant causal relationships with DCM ($P \leq 0.05$).
### Conclusion
Our findings propose that oleic acid and fatty acid (18:1)-OH may have causal relationships with DCM, indicating that the risk of DCM from oleic acid may be decreased by encouraging the conversion of oleic acid to fatty acid (18:1)-OH.
## 1. Introduction
Dilated cardiomyopathy (DCM) is a type of heart muscle disease that causes left ventricular dilation and systolic dysfunction [1]. This abnormal structure or function of the myocardium can develop into complications such as sudden cardiac death and intractable heart failure, and can therefore ultimately be life-threatening [2]. Furthermore, chronically treated patients may occasionally exhibit acute decompensated heart failure [2]. Globally, DCM is the most frequent cardiomyopathy, occurring in people of any age or gender [2]. As early as 2013, the prevalence of DCM increased from $\frac{1}{2}$,500 to $\frac{1}{250}$ [3]. DCM is classified as a serious heart disease by the World Health Organization because of its high morbidity and mortality rates [4]. Due to its high hospital admission rates and the potential need for a heart transplant, DCM imposes significant financial obligations [2]. As a consequence, despite advances in DCM treatment, outcomes still need to be improved. A significant advancement in DCM clinical therapy will be made if an intervention that effectively prevents or treats DCM without imposing a heavy financial burden on patients and with excellent compliance is Id.
DCM is a sophisticated form of ventricular remodeling that alters the morphology of the ventricles and causes myocardial fibrosis. In this condition, all chambers of the heart enlarge, ventricular wall pressure rises, and systolic function decreases [2]. DCM is characterized by an enlarged and weakened heart muscle, and patients often present with weakness, dizziness, pallor, dyspnoea, edema, abdominal distention, and even embolism [5]. Besides, patients with DCM are always accompanied by mitral regurgitation, ventricular arrhythmias, and other rhythm disturbances [2]. An electrocardiogram, cardiac MRI features, history-taking, and clinical findings all play a role in the clinical diagnosis [6]. Endocardial biopsy and the use of biomarkers like B-type natriuretic peptide can also assist with the diagnosis and confirmation of the disease process [7, 8]. DCM has a complex multifactorial etiology that includes both inherited and environmental factors [2]. The most important cause of dilated cardiomyopathy is genetic, with the main mode of inheritance being autosomal dominant [2]. The pathogenic genes of DCM mainly encode cytoskeletal proteins and sarcomeric proteins. Non-genetic triggers of DCM include tachycardia-induced cardiomyopathy, hypertension, alcohol or cocaine abuse, inflammatory disorders, and autoimmune diseases [2]. Evidence of inflammatory cell infiltration and gene expression patterns compatible with immune cell activation is often revealed by pathological examination of myocardial biopsy samples (or autopsies) of patients with DCM [4, 9]. Abnormalities in myocardial fatty acid metabolism play an important role in DCM and can trigger a range of events such as impaired energetics, and oxidative stress and even lead to reduced cardiac function [10]. Neglia et al. showed that patients with DCM have reduced levels of fatty acid oxidation [11]. As DCM pathogenesis research advances, identifying its potential targets can serve as a foundation for new interventions.
Fatty acids (FAs) are essential nutrients for the human body. The diet-heart hypothesis is primarily based on the effect of fatty acids in dietary fat on blood lipids, which in turn modulate cardiovascular outcomes by affecting thrombosis, endothelial function, inflammation, arrhythmias, and the onset of diabetes [12]. FAs are the primary metabolic substrates for myocardial action in the heart under aerobic conditions and have an impact on processes like ion channel function and cell signaling [13]. Yazaki et al. first investigated the metabolic process of fatty acids using 123I-β-methyl-p-iodophenyl pentadecanoic acid (BMIPP) single photon emission computed tomography (SPECT) technology labeling in 1999 and discovered that the severity of abnormal fatty acid metabolism in DCM myocardia reflected the severity of hemodynamic deterioration and histological abnormalities [14]. N-3 polyunsaturated fatty acids (PUFAs) seemed to reduce the risk of arrhythmia in patients with DCM, according to data from the Microvolt T-Wave Alternans (MTWA) test, which has been demonstrated to be one of the best predictors of severe arrhythmias and sudden death [15]. According to the findings of DCM metabolomics studies, linoleic acid could be used as the primary biomarker to measure the effectiveness of DCM treatment, and hydroxyl fatty acid esters were potential biomarkers to identify DCM [16]. Additionally, Nitro-oleic acid (NO2-OA), a signaling mediator produced naturally by reacting oleic acid with nitrogen dioxide (NO2), was created as a therapeutic agent for fibrotic and inflammatory diseases and has been proven effective in numerous animal models of cardiovascular diseases [17]. Controversially, several in vitro and animal investigations have raised concerns that linoleic acid and other n-6 PUFAs may induce inflammation and thrombosis [18]. Clinical trial results also suggest an increase in arrhythmia occurrences in patients with ventricular tachycardia who are treated with n-3 PUFAs [19]. The heterogeneity between trials may be brought about by variations in the research populations and dietary sources of fatty acids. Therefore, it is difficult to be convinced that FAs have a causal effect on DCM and which FA regulation might be most important in DCM.
Previous observational studies have shown that FAs are substantially associated with DCM, but traditional observational studies are prone to confounding. For the confounding factors that can be observed in such studies, covariable correction methods can make comparisons between groups that tend to be balanced. However, for unmeasured confounders, even the best epidemiological design cannot correct for bias. Therefore, we introduced an instrumental variables (IVs) model to simulate sample randomization allocation in order to eliminate the bias effects of any confounding factors, other than target exposure, on causal inference [20]. Mendelian randomization (MR), a research method proposed by Katan in 1986, infers and analyzes causal effects by establishing an instrumental variable model utilizing single-nucleotide polymorphisms (SNPs) [20]. Genetic variation, according t' Mendel's law, is independent of environmental influences and avoids the influence of confounding factors. Simultaneously, genetic variation as a starting point of distinct outcomes might rule out the potential reverse causality possibility [21]. Consequently, we may obtain the causal effect of “exposure outcome” using the MR approach, which cannot be determined using the usual observational research design. In our study, we employed a two-sample MR approach to analyze GWAS data to evaluate the causal relationship between FAs and DCM risk.
## 2.1. Study design
As shown in the schematic diagram in Figure 1, we employed the two-sample MR approach to confirm the causality between FAs and DCM. Each FA was designed as an exposure factor, while DCM served as an outcome indicator.
**Figure 1:** *Schematic diagram of two-sample MR analysis. The three assumptions in MR model: 1. Relevance. IVs must be strongly related to exposure factor. 2. Exclusivity. IVs affect outcome only through exposure factor not through any other pathways. 3. Independence. IVs and confounding factors are independent of each other.*
## 2.2. Data source
All data used are openly accessible. SNPs associated with different FAs were extracted from four major large clinical studies (Table 1). ( a) *Some* genetic associations between SNPs and FAs were obtained from the genome-wide association study (GWAS) by Wu et al. [ 22], which consists of 5 cohorts, namely the Atherosclerosis Risk in Communities (ARIC) Study, the Cardiovascular Health Study (CHS), the Coronary Artery Risk Development in Young Adults (CARDIA) Study, the Invecchiare in Chianti (InCHIANTI) Study, and the Multi-Ethnic Study of Atherosclerosis (MESA). Among them, ARIC and CHS, which are from the Cohort for Heart and Aging Research in Genomic Epidemiology (CHARGE), were the prominent data contributors. Only 8,961 European participants from these cohorts were included in this GWAS data set. ( b) *The* genetic loci of PUFAs were also obtained from the five above mentioned population-based cohorts for GWAS. Differently, Rozenn et al. selected 8,866 subjects of European descent for genetic analysis [23]. ( c). The LURIC Health Study, from 1997 to 2000, involved 3,316 German Caucasians (2,309 men and 1,007 women), and excluded participants who had had an acute illness (except acute coronary syndrome), non-cardiac chronic illness, or malignant neoplasms within the past 5 years [24]. This prospective cohort study assessed genetic and environmental risk factors for cardiovascular disease in patients. Participants were filtered to remove individuals under the age of 18; persons with absent covariates including age, waist–hip ratio, BMI, and sex; and samples with a < $99\%$ sample call rate. Finally, a total of 687,262 SNPs were obtained from 3,061 samples [25]. ( d) *Other* genetic data were acquired from 5,662 controls in the Pakistan Myocardial Infarction Risk Study (PROIS) and 13,814 UK participants in the INTERVAL study [26]. PROIS is a case–control study of first-ever acute myocardial infarction in nine urban centers in Pakistan, and INTERVAL is a prospective cohort study of approximately 50,000 blood donors from the United Kingdom. These final samples were screened strictly with genetic and lipid characteristics. Blood samples were collected for measurement and analysis to obtain levels of each fatty acid. The plasma phospholipids were separated by TLC, and the fatty acids were subsequently quantified by gas chromatography. Plasma total fatty acids in the INCHIANTI cohort were determined using a similar GC technique. Serum lipid levels were quantified using a high-resolution mass spectrometer. In our study, considering that gene loci of these fatty acids rarely reach genome-wide significance in GWAS, SNPs with suggestive genome-wide significance thresholds (i.e., $P \leq 5$ × 10−5) were selected as IVs.
**Table 1**
| Trait | Year | Study | Sample | Subjects | Case | Control |
| --- | --- | --- | --- | --- | --- | --- |
| Palmitic acid palmitoleic acid stearic acid oleic acid | 2013 | ARIC | Plasma phospholipid | 3269 | | |
| | 2013 | CARDIA | Plasma phospholipid | 1507 | | |
| | 2013 | CHS | Plasma phospholipid | 2404 | | |
| | 2013 | InCHIANTI | Total plasma | 1075 | | |
| | 2013 | MESA | Plasma phospholipid | 706 | | |
| Alphalinolenic acid docosahexaenoic acid docosapentaenoic acid eicosapentaenoic acid | 2011 | ARIC | Plasma phospholipid | 3268 | | |
| | 2011 | CARDIA | Plasma phospholipid | 1507 | | |
| | 2011 | CHS | Plasma phospholipid | 2326 | | |
| | 2011 | InCHIANTI | Total plasma | 1075 | | |
| | 2011 | MESA | Plasma phospholipid | 690 | | |
| Arachidonic acid dihomo-gamma-linolenic acid linoleic acid linolenic acid myristic acid trans-palmitoleic acid 9-cis,12-trans octadecanoic acid 9-trans,12-cis octadecanoic acid 9-trans,12-trans octadecanoic acid | 2020 | LURIC | Serum | 3061 | | |
| Fatty acid (22:0) Fatty acid (24:0) | 2015 | CHARGE | Plasma phospholipid | 10129 | | |
| Fatty acids1 | 2021 | INTERVAL | Serum | 13814 | | |
| | 2009 | PROMIS | Serum | 5662 | | |
| Dilated cardiomyopathy | 2021 | HF METTC | Serum | 533543 | 1861.0 | 531682.0 |
Summary statistics for genetic IVs of DCM were extracted from the largest available GWAS meta-analysis with European ancestry performed by the Heart Failure (HF) Molecular Epidemiology for Therapeutic Targets Consortium [27]. For the DCM GWAS, they included 533,543 European individuals (1,861 cases and 531,682 controls), and the age, gender and principal components of patients was adjusted.
## 2.3. Genetic instrumental variables
In MR analysis, IVs must satisfy three core characteristics: [1] Relevance. There needs to be a strong relationship between IVs and exposure factors; otherwise, causal effects will have bias. [ 2] Exclusivity. IVs can only be associated with outcome through exposure factors, not through any other pathways. [ 3] Independence. Random distribution of alleles provides a theoretical basis for the independence of IVs and confounding factors. Hence, all data for which a battery of steps was performed in controlling the quality of selected SNPs were eligible. To ensure a strong correlation between IVs and oleic acid, the filter criteria ($P \leq 5$ × 10−8, r2 < 0.001, genetic distance = 10,000 KB, minor allele frequency >0.01) were set using Plink Software. Importantly, the selected SNPs are not in linkage disequilibrium (LD) [28]. Then, we searched the Catalog and PhenoScanner database to find and exclude SNPs which had relationships with known confounders, to meet the standard of the independence of genetic variation and confounders. Next, we calculated the F statistic, and abandoned SNPs with an F-value < 10 to avoid bias [29]. Those satisfactory SNPs whose F values were significantly >10 after stringent screening were used for the following MR analysis.
## 2.4. Mendelian randomization analysis
The Mendelian randomization method defines genetic variation as IVs and solves the bias effect of confounding factors on causal judgment to a large extent by introducing the IVs model. In the current study, we utilized a two-sample MR design approach and combined different summary statistical methods to effectively infer the causal relationships between exposure factors and outcomes based on different MR hypotheses, including the inverse variance weighted (IVW) method, MR-Egger, the MR pleiotropy residual sum and outlier test (MRPRESSO), maximum likelihood and weighted median estimator (WME). In addition, we used MR-Steiger to verify whether there is a potential reverse causal possibility between FAs and DCM [30].
The IVW method was predominant in our MR analysis. It assumes that all bias is zero and produces a combined estimate of causal effects by using a combination of the Wald ratio and the selective fixed effects model, that is, the inverse variance weighted mean (IVW estimate) [31]. The accuracy of this estimate is higher than that of any causal effect estimate based on a single genetic IV. The outliers in IVW estimates were then detected and corrected with MRPRESSO. Then, using MR-Egger, maximum likelihood, WME, and consistent estimates of causal effects, the relationships between exposure factors and outcome were further verified. MR-*Egger is* a weighted linear regression method based on the InSIDE hypothesis that can give valid tests and consistent estimates of causal effects even if all IVs are invalid [32]. The WME estimate is more significant than the simple median, and it is calculated based on the weighted empirical distribution function for the ratio estimate of a single SNP [33]. The odds ratio (OR) and $95\%$CI were calculated to explain the results of MR analysis. To interpret the multiple testing in this study, we used the Bonferroni correction for significance level. P-value between 9.26 × 10−4 (0.05 divided by 54 risk factors) and 0.05 were considered to be potentially associated. All analyses in our study were carried out using R (Version 4.1.2, https://www.r-project.org/) software package, two-Sample MR and MRPRESSO packages.
## 2.5. Heterogeneity and pleiotropic evaluates
To ensure the assumption of exclusivity, we performed heterogeneity and pleiotropy analyses on the included IVs. The intercept of MR-Egger regression is an important indicator of SNPs' potential pleiotropy. The closer the intercept value is to zero, the smaller the effect of genetic pleiotropy. Heterogeneity was analyzed mainly by MRPRESSO method and Cochran Q test. Meanwhile, we carried out “leave one-out” sensitivity analysis of the remaining SNPs by the IVW method after excluding one SNP, hoping to evaluate the effect of the individual SNP on DCM. If the p-value is not significant, it indicates that there is no bias and heterogeneity in the included IVs. On the contrary, if the p-value is < 0.05, bias and heterogeneity exist.
## 3. Results
All 54 available FAs' data were extracted. The SNPs of each FA were strictly screened in turn, and the included SNPs were selected as IVs for multiple calculation and verification. After calculation, we found that two FAs, oleic acid and fatty acid (18:1)-OH, were suggestively associated with DCM (Figure 2). However, the other 52 FAs had no significant causal relationships with DCM risk (Supplementary material). The detailed MR analysis results of oleic acid and fatty acid (18:1)-OH are as follows.
**Figure 2:** *The odds ratio (OR) results for the causal effects with oleic acid and DCM, fatty acid (18:1)-OH and DCM respectively by using MR methods. The small black square represents the point estimate of the study effect size (OR). The length of the line segment represents the 95% confidence interval (95%CI) for each effect value for each study.*
## 3.1.1. Oleic acid
SNPs with low allele frequencies ≤ 0.01 or no meaningful genome-wide association evidence ($P \leq 5$ × 10−5) were excluded. We identified five SNPs, rs34143286, rs2465604, rs174448, rs2555277, and rs9564082, as LD-independent IVs (after the clumping process). The variance of oleic acid explained by genetic instruments was $29.1\%$. In our study, F statistics were all significantly >10, suggesting that our results were highly trustworthy and largely unaffected by weak IVs (Table 2).
**Table 2**
| Exposure | SNP | EA/OA | EAF | beta | se | P | F |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Oleic acid | rs9564082 | A/G | 0.213 | −0.109 | 0.027 | 4.246 × 10−5 | 33.005 |
| | rs174448 | A/G | 0.643 | −0.140 | 0.019 | 3.026 × 10−13 | 74.938 |
| | rs2555277 | A/G | 0.378 | 0.083 | 0.019 | 9.349 × 10−6 | 26.822 |
| | rs34143286 | T/C | 0.018 | 1.009 | 0.237 | 1.996 × 10−5 | 308.125 |
| | rs2465604 | T/C | 0.821 | −0.106 | 0.025 | 1.84 × 10−5 | 27.346 |
| Fatty acid (18:1)-OH | rs1366632 | G/T | 0.424 | −0.037 | 0.009 | 1.6 × 10−5 | 13.031 |
| | rs72907101 | T/C | 0.011 | −0.175 | 0.04 | 1.2 × 10−5 | 12.985 |
| | rs13208968 | A/G | 0.477 | 0.036 | 0.009 | 2 × 10−5 | 12.601 |
| | rs7941838 | G/A | 0.163 | 0.048 | 0.012 | 3.6 × 10−5 | 12.25 |
| | rs9319312 | G/T | 0.141 | −0.050 | 0.012 | 4.6 × 10−5 | 11.801 |
| | rs12885354 | G/A | 0.139 | 0.052 | 0.012 | 3.1 × 10−5 | 12.612 |
| | rs79758221 | G/T | 0.028 | −0.115 | 0.026 | 8.2 × 10−6 | 14.029 |
| | rs7274662 | G/A | 0.217 | −0.047 | 0.01 | 4.3 × 10−6 | 14.629 |
| | rs6005264 | G/T | 0.256 | −0.046 | 0.01 | 2.5 × 10−6 | 15.71 |
## 3.1.2. Fatty acid (18:1)-OH
We selected nine independent SNP loci as IVs, namely rs72907101, rs1366632, rs79758221, rs7274662, rs6005264, rs7941838, rs12885354, rs9319312, and rs13208968. The explained variance of fatty acid (18:1)-OH was $59.8\%$. F statistics for each instrument–exposure association were all greater than the empirical threshold of 10, demonstrating a small likelihood of weak IVs (Table 2).
## 3.2. MR analysis for potential causal effects of oleic acid and fatty acid (18:1)-OH on DCM
In our study, oleic acid correlated positively with DCM risk. Based on the results obtained from the IVW approach, oleic acid has a suggestive causal association with increased risk of DCM, with a variance measured by IVs of $29.1\%$. ( OR = 1.291, $95\%$CI: 1.044–1.595). The maximum likelihood approach confirmed this finding (OR = 1.296, $95\%$CI: 1.0149–1.595). However, the results of MR-Egger regression (OR = 1.249, $95\%$CI: 0.961–1.622) and the WME approach (OR = 1.271, $95\%$CI: 0.974–1.659) showed no statistical significance.
The IVW results for fatty acid (18:1)-OH showed a suggestive negative correlation with DCM risk, with a variance measured by IVs of $59.8\%$ (OR = 0.402, $95\%$CI: 0.167–0.966). This negative correlation was proved by the maximum likelihood method (OR = 0.409, $95\%$CI: 0.168–0.997) and MRPRESSO method (OR = 0.402, $95\%$CI: 0.204–0.794). However, the results of MR-Egger regression (OR = 0.260, $95\%$CI: 0.052–1.300) and the WME method (OR = 0.324, $95\%$CI: 0.099–1.061) showed no statistical significance.
Ultimately, we employed the MR-Steiger test to determine the direction of the causal effect, confirming that the instrumentation of oleic acid and fatty acid (18:1)-OH both influenced susceptibility to DCM ($P \leq 0.001$), rather than vice versa, and that the orientation of the effect was fairly robust (Table 3).
**Table 3**
| Exposure | Outcome | R2 for exposure | R2 for outcome | Correct_causal_direction | Psteige |
| --- | --- | --- | --- | --- | --- |
| Oleic acid | Dilated cardiomyopathy | 0.015 | 7.34 × 10−6 | True | 6.95 × 10−28 |
| Fatty acid (18:1)-OH | Dilated cardiomyopathy | 0.019 | 3.39 × 10−5 | True | 1.10 × 10−74 |
## 3.3. Heterogeneity and pleiotropic evaluates
The results of Cochran's Q test and the MR-Egger test in the analysis for oleic acid and DCM proved no heterogeneity ($$P \leq 0.535$$) and no sensitive pleiotropy ($$P \leq 0.702$$), respectively. Further, there were no strange outliers in the included IVs' loci, which was mutually confirmed with the IVW results. Additionally, the results of the IVW random effect model and fixed effect model were consistent (OR: 1.290, $95\%$CI: 1.044–1.595, $$P \leq 0.018$$). After the calculations made using the “leave-one-out” method, we found that the included IVs affected these results indistinctively, suggesting our analysis was credible (Figure 3A). Additionally, funnel plots also confirmed that the causality between oleic acid and DCM was practically not affected by potential bias (Figure 3C).
**Figure 3:** *Heterogeneity and pleiotropic evaluates. Sensitivity analysis for SNP of (A) oleic acid and (B) fatty acid (18:1)-OH by “leave-one-out” method (Each black dot represents the odds ratio (OR) of the risk of developing DCM caused by increased oleic acid or fatty acid levels after excluding a specific SNP, red dots represent IVW estimates using all SNP, and horizontal segments represent 95% confidence intervals of the estimated values). Funnel plot to visualize the causal effect of (C) oleic acid and (D) fatty acid (18:1)-OH on DCM.*
Cochran's Q test and the horizontal pleiotropy of fatty acid (18:1)-OH suggested there was an absence of heterogeneity ($$P \leq 0.776$$) and gene pleiotropy ($$P \leq 0.548$$). The MRPRESSO test revealed no significant outliers. The “leave-one-out” sensitivity analysis affirmed the previous results (Figure 3B). The asymmetrical distribution of SNPs' loci in funnel plots also confirmed the absence of potential bias (Figure 3D).
## 4. Discussion
In this study, we used summary statistics from the GWAS database of large-scale clinical trials to execute a two-sample MR approach to assess the causal association between each available FA and DCM. Overall, we identified two FAs that have causal effects on DCM, oleic acid and fatty acid (18:1)-OH. Our results found that oleic acid may have an increased risk of DCM, while fatty acid (18:1)-OH may decrease the risk of DCM contrarily. The causal effects of other FAs on DCM, however, were found to be non-significant.
Oleic acid (C18:1 ω-9), a non-essential FA, is the main component of biofilm and also the highest content of monounsaturated fatty acid (MUFAs) provided by diets (~$90\%$ of all MUFAs) [34]. Several authors have reported that oleic acid can significantly reduce nitric oxide levels in endothelial cells, which may be related to fatty acid-induced overproduction of superoxide and ROS in the mitochondrial electron transport chain [35]. LURIC study for different types of omega-9 MUFAs showed that oleic acid was positively correlated with indicators of inflammation, vascular endothelial cell activation, heart failure, and even an increased risk of all-cause death (HR = 1.080, $95\%$CI:1.010–1.160) [36]. Such risk effects of oleic acid on DCM may be related to its physiological toxicity for human beta cells [37]. However, controversy still exists regarding this theory. In the PREDIMED trial conducted by Guasch-Ferré et al. MUFA intake was found to have a significant correlation with total cardiovascular disease (CVD) risk as a protective factor (HR = 0.630, $95\%$CI: 0.430–0.940) [38]. This protection may be mediated by the synthesis of the endogenous signaling mediator NO2-OA [17] and hydroxylation to hydroxyl fatty acids in response to metabolic disturbances and increased inflammatory stress in the DCM [17]. *In* general, these positive correlation results are consistent with ours, suggesting the causal effect of oleic acid on DCM.
Fatty acid (18:1)-OH is a monounsaturated hydroxyl FA (HFA) that contains 18 carbons. Few studies on HFAs have been reported, and the specific regulatory mechanism is currently unclear. Of these studies, 10-hydroxy-2-decenoic acid, the main FA of royal jelly, has been noted to inhibit endothelial vascular cell growth by inhibiting cell proliferation and migration [39] and to resist inflammation [40]. Our findings propose that fatty acids (18:1)-OH, as a protective factor of DCM, has a significant causal effect on DCM risk. In a previous study, the corresponding HFAs were synthesized by hydroxylation with FAs as the substrate and nicotinamide adenine dinucleotide phosphate as the electron donor under the catalysis of cytochrome P450 [41]. Based on the current evidence for HFAs, we speculate that this protective effect may be related to fatty acid hydroxyl fatty acid esters (FAHFAs). FAHFAs are a recently discovered group of endogenous lipid molecules that were first reported in Cell in 2014 [42]. They are produced by the esterification of C16 and C18 FAs, such as oleic acid, palmitic acid and their corresponding HFAs [42]. In an obese mouse model, FAHAFs have been found to inhibit the secretion of pro-inflammatory factors such as TNF and IL-1β [43]. Research by the Beth Israel Deaconess Medical Center and Harvard Medical School has noted that FAHAFs can inhibit the endoplasmic reticulum stress response by reducing the activation of the JNK/MAPK pathway [42].Hence, FAHAFs are a promising class of anti-inflammatory lipids. To summarize, the protective effect of fatty acids (18:1)-OH on DCM may be related to their anti-inflammatory effects, which work directly or indirectly through esterification to form FAHAFs. Further experimental in vivo and clinical studies are needed to investigate this theory.
Different FAs have different effects on the risk of CVD. Most PUFAs are essential FAs that cannot be synthesized by the human body. Fish oil, which is rich in N-3 fatty acids, has been reported to potentially protect against cardiovascular disease by raising the threshold for arrhythmia, lowering blood pressure, and improving arterial and endothelial function [44]. However, results from another randomized controlled trial suggest that fish oil supplementation may not reduce the risk of ventricular fibrillation and tachycardia, but may even trigger arrhythmias in some patients (HR = 1.76, $95\%$CI: 1.15–2.28). [ 19] Linoleic acid, an omega-6 PUFA, is the precursor of arachidonic acid, and can be converted into prostaglandins in the inflammatory cascade; it has been found to have a pro-inflammatory effect [45]. Cells involved in inflammation usually have higher levels of omega-6 FAs and arachidonic acid. It is, however, interesting to note that linoleic acid and other omega-6 FAs may be risk factors for promoting inflammation or thrombosis [45], although a prospective clinical cohort study in Circulation has shown the cardioprotective effects of omega-6 PUFA intake (RR = 0.850, $95\%$CI: 0.780–0.920) [46]. These conflicting results may be in part due to the different lineages, ages, and genders of people, sources of FAs and inclusion criteria used for different study groups. Our findings showed that the other 52 available FAs had no significant causal correlations with DCM. More studies of the mechanisms involved in the regulation of DCM by fatty acids are needed to verify whether the fatty acids themselves play a role or their metabolites do.
This is the first study to estimate the causal relationship between FAs and DCM. We explained the etiology of this relationship using the MR method and excluded the influence of confounding factors and reverse causal correlation to ensure the results were more reliable. At the same time, we conducted IVW, MR-Egger, MWE, maximum likelihood, MRPRESSO, and other calculation methods to validate the casual effect. In addition, genetic data of FAs and DCM were extracted from GWAS datasets, which aggregate a mass of clinical samples. However, our research has some limitations. First of all, most of the genetic data used were from Europeans, but some came from Asians. Although our data suggest that there may not be heterogeneity between these genetic variants, a small proportion of Asians may still lead to stratification among different ethnic groups. This issue requires further isolation of the raw data for validation. Secondly, the size of genetic effects could not be directly transformed into a clinical intervention effect. Considering that genetic variation is longer lasting, the causal effect estimated by MR is greater than the effect of clinical intervention generally. Thirdly, the non-linear relationship between exposure factors and outcomes could not be evaluated by the MR method. Fourthly, we cannot completely exclude possible diet-gene or gene-environment interactions that could have had an impact on our results. Furthermore, like other MR studies, we are unable to fully address the unobserved pleiotropies. It should be acknowledged that estimates of IVW effects are prone to bias when some instrumental SNPs exhibit horizontal pleiotropy (e.g., when FAs are measured by different measurements).
## 5. Conclusion and perspective
For the first time, we systematically analyzed the potential causal effect of 54 FAs on DCM. We identified a potential risk factor (oleic acid) and a potential protective factor [fatty acid (18:1)-OH] that were found to have suggestive significant causal effects on DCM events. On the one hand, more basic and clinical studies are needed to validate our results. On the other hand, the potential protective mechanism of fatty acid (18:1)-OH in DCM could be explored by detecting variations in oleic acid and the corresponding HFA and FAHFA, so as to provide theoretical references for the clinical prevention and treatment of DCM.
## 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
YH, XC, and HW conceived the study, performed manuscript revision, and took accountability for all aspects of the work. JZ performed the data interpretation, drafted and revised the manuscript. QL designed the methodology and did the software analysis. JH, WX, and PL were in charge of supervision and administration. 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. The reviewer DY declared a shared affiliation with the authors WX and YH to the handling editor 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.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1068050/full#supplementary-material
## References
1. Elliott PM. **Diagnosis and management of dilated cardiomyopathy**. *Heart.* (2000) **84** 106-12. DOI: 10.1136/heart.84.1.106
2. Jefferies JL, Towbin JA. **Dilated cardiomyopathy**. *Lancet.* (2010) **375** 752-62. DOI: 10.1016/S0140-6736(09)62023-7
3. Hershberger RE, Hedges DJ, Morales A. **Dilated cardiomyopathy: the complexity of a diverse genetic architecture**. *Nat Rev Cardiol.* (2013) **10** 531-47. DOI: 10.1038/nrcardio.2013.105
4. Richardson P, McKenna W, Bristow M, Maisch B, Mautner B, O'Connell J. **Report of the 1995 world health organization/international society and federation of cardiology task force on the definition and classification of cardiomyopathies**. *Circulation.* (1996) **93** 841-2. DOI: 10.1161/01.CIR.93.5.841
5. Silver M, Chen P, Li R, Cheng CY, Wong TY, Tai ES. **Pathways-driven sparse regression identifies pathways and genes associated with high-density lipoprotein cholesterol in two Asian cohorts**. *PLoS Genet.* (2013) **9** e1003939. DOI: 10.1371/journal.pgen.1003939
6. Japp AG, Gulati A, Cook SA, Cowie MR, Prasad SK. **The diagnosis and evaluation of dilated cardiomyopathy**. *J Am Coll Cardiol.* (2016) **67** 2996-3010. DOI: 10.1016/j.jacc.2016.03.590
7. Cooper LT, Baughman KL, Feldman AM, Frustaci A, Jessup M, Kuhl U. **the role of endomyocardial biopsy in the management of cardiovascular disease: a scientific statement from the american heart association, the American college of cardiology, and the european society of cardiology. Endorsed by the heart failure society of America and the heart failure association of the European society of cardiology**. *J Am Coll Cardiol.* (2007) **50** 1914-31. DOI: 10.1161/CIRCULATIONAHA.107.186093
8. Omland T, Sabatine MS, Jablonski KA, Rice MM, Hsia J, Wergeland R. **Prognostic value of B-type natriuretic peptides in patients with stable coronary artery disease: the peace trial**. *J Am Coll Cardiol.* (2007) **50** 205-14. DOI: 10.1016/j.jacc.2007.03.038
9. Noutsias M, Rohde M, Göldner K, Block A, Blunert K, Hemaidan L. **Expression of functional T-cell markers and T-cell receptor vbeta repertoire in endomyocardial biopsies from patients presenting with acute myocarditis and dilated cardiomyopathy**. *Eur J Heart Fail.* (2011) **13** 611-8. DOI: 10.1093/eurjhf/hfr014
10. Tu Z, Li S, Sharp TL, Herrero P, Dence CS, Gropler RJ. **Synthesis and evaluation of 15-(4-(2-[18F]Fluoroethoxy)Phenyl) pentadecanoic acid: a potential pet tracer for studying myocardial fatty acid metabolism**. *Bioconjug Chem.* (2010) **21** 2313-9. DOI: 10.1021/bc100343h
11. Karwi QG, Uddin GM, Ho KL, Lopaschuk GD. **Loss of metabolic flexibility in the failing heart**. *Front Cardiovasc Med.* (2018) **5** 68. DOI: 10.3389/fcvm.2018.00068
12. Erkkilä A, de Mello VD, Risérus U, Laaksonen DE. **dietary fatty acids and cardiovascular disease: an epidemiological approach**. *Prog Lipid Res.* (2008) **47** 172-87. DOI: 10.1016/j.plipres.2008.01.004
13. Calder PC. **Functional roles of fatty acids and their effects on human health**. *JPEN J Parenter Enteral Nutr.* (2015). DOI: 10.1177/0148607115595980
14. Yazaki Y, Isobe M, Takahashi W, Kitabayashi H, Nishiyama O, Sekiguchi M. **Assessment of myocardial fatty acid metabolic abnormalities in patients with idiopathic dilated cardiomyopathy using 123i Bmipp spect: correlation with clinicopathological findings and clinical course**. *Heart.* (1999) **81** 153-9. DOI: 10.1136/hrt.81.2.153
15. Nodari S, Metra M, Milesi G, Manerba A, Cesana BM, Gheorghiade M. **The role of N-3 pufas in preventing the arrhythmic risk in patients with idiopathic dilated cardiomyopathy**. *Cardiovasc Drugs Ther.* (2009) **23** 5-15. DOI: 10.1007/s10557-008-6142-7
16. Ampong I. **Metabolic and metabolomics insights into dilated cardiomyopathy**. *Ann Nutr Metab.* (2022) **78** 147-55. DOI: 10.1159/000524722
17. Braumann S, Schumacher W, Im NG, Nettersheim FS, Mehrkens D, Bokredenghel S. **Nitro-Oleic Acid. (No(2)-Oa) improves systolic function in dilated cardiomyopathy by attenuating myocardial fibrosis.**. *Int J Mol Sci* (2021) **22** 9052. DOI: 10.3390/ijms22169052
18. Calder PC. **Polyunsaturated fatty acids, inflammation, and immunity**. *Lipids.* (2001) **36** 1007-24. DOI: 10.1007/s11745-001-0812-7
19. Raitt MH, Connor WE, Morris C, Kron J, Halperin B, Chugh SS. **fish oil supplementation and risk of ventricular tachycardia and ventricular fibrillation in patients with implantable defibrillators: a randomized controlled trial**. *Jama.* (2005) **293** 2884-91. DOI: 10.1001/jama.293.23.2884
20. Newhouse JP, McClellan M. **Econometrics in outcomes research: the use of instrumental variables**. *Annu Rev Public Health.* (1998) **19** 17-34. DOI: 10.1146/annurev.publhealth.19.1.17
21. Smith GD, Lawlor DA, Harbord R, Timpson N, Day I, Ebrahim S. **Clustered environments and randomized genes: a fundamental distinction between conventional and genetic epidemiology**. *PLoS Med.* (2007) **4** e352. DOI: 10.1371/journal.pmed.0040352
22. Wu JH, Lemaitre RN, Manichaikul A, Guan W, Tanaka T, Foy M. **Genome-wide association study identifies novel loci associated with concentrations of four plasma phospholipid fatty acids in the de novo lipogenesis pathway: results from the cohorts for heart and aging research in genomic epidemiology**. *Consortium Circ Cardiovasc Genet.* (2013) **6** 171-83. DOI: 10.1161/CIRCGENETICS.112.964619
23. Lemaitre RN, Tanaka T, Tang W, Manichaikul A, Foy M, Kabagambe EK. **Genetic loci associated with plasma phospholipid N-3 fatty acids: a meta-analysis of genome-wide association studies from the charge consortium**. *PLoS Genet.* (2011) **7** e1002193. DOI: 10.1371/journal.pgen.1002193
24. Silbernagel G, Chapman MJ, Genser B, Kleber ME, Fauler G, Scharnagl H. **High intestinal cholesterol absorption is associated with cardiovascular disease and risk alleles in Abcg8 and Abo: evidence from the luric and yfs cohorts and from a meta-analysis**. *J Am Coll Cardiol.* (2013) **62** 291-9. DOI: 10.1016/j.jacc.2013.01.100
25. Zhou J, Passero K, Palmiero NE, Müller-Myhsok B, Kleber ME, Maerz W. **Investigation of gene-gene interactions in cardiac traits and serum fatty acid levels in the luric health study**. *PLoS ONE.* (2020) **15** e0238304. DOI: 10.1371/journal.pone.0238304
26. Harshfield EL, Fauman EB, Stacey D, Paul DS, Ziemek D, Ong RMY. **Genome-wide analysis of blood lipid metabolites in over 5,000 South Asians reveals biological insights at cardiometabolic disease loci**. *BMC Med.* (2021) **19** 232. DOI: 10.1186/s12916-021-02087-1
27. Sakaue S, Kanai M, Tanigawa Y, Karjalainen J, Kurki M, Koshiba S. **A cross-population atlas of genetic associations for 220 human phenotypes**. *Nat Genet.* (2021) **53** 1415-24. DOI: 10.1038/s41588-021-00931-x
28. Wu F, Huang Y, Hu J, Shao Z. **Mendelian randomization study of inflammatory bowel disease and bone mineral density**. *BMC Med.* (2020) **18** 312. DOI: 10.1186/s12916-020-01778-5
29. Staiger D, Stock JH. **Instrumental variables regression with weak instruments**. *Econometrica.* (1997) **65** 557-86. DOI: 10.2307/2171753
30. Hemani G, Tilling K, Davey Smith G. **Orienting the causal relationship between imprecisely measured traits using gwas summary data**. *PLoS Genet.* (2017) **13** e1007081. DOI: 10.1371/journal.pgen.1007081
31. Burgess S, Butterworth A, Thompson SG. **Mendelian randomization analysis with multiple genetic variants using summarized data**. *Genet Epidemiol.* (2013) **37** 658-65. DOI: 10.1002/gepi.21758
32. 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
33. Bowden J, Davey Smith G, 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
34. Vannice G, Rasmussen H. **Position of the academy of nutrition and dietetics: dietary fatty acids for healthy adults**. *J Acad Nutr Diet.* (2014) **114** 136-53. DOI: 10.1016/j.jand.2013.11.001
35. Storniolo CE, Roselló-Catafau J, Pintó X, Mitjavila MT, Moreno JJ. **Polyphenol fraction of extra virgin olive oil protects against endothelial dysfunction induced by high glucose and free fatty acids through modulation of nitric oxide and endothelin-1**. *Redox Biol.* (2014) **2** 971-7. DOI: 10.1016/j.redox.2014.07.001
36. Delgado GE, Krämer BK, Lorkowski S, März W, von Schacky C, Kleber ME. **Individual omega-9 monounsaturated fatty acids and mortality-the ludwigshafen risk and cardiovascular health study**. *J Clin Lipidol* (2017) **11** 126-35. DOI: 10.1016/j.jacl.2016.10.015
37. Plötz T, Krümmel B, Laporte A, Pingitore A, Persaud SJ, Jörns A. **The monounsaturated fatty acid oleate is the major physiological toxic free fatty acid for human beta cells**. *Nutr Diabetes.* (2017) **7** 305. DOI: 10.1038/s41387-017-0005-x
38. Guasch-Ferré M, Babio N, Martínez-González MA, Corella D, Ros E, Martín-Peláez S. **Dietary fat intake and risk of cardiovascular disease and all-cause mortality in a population at high risk of cardiovascular disease**. *Am J Clin Nutr.* (2015) **102** 1563-73. DOI: 10.3945/ajcn.115.116046
39. Izuta H, Chikaraishi Y, Shimazawa M, Mishima S, Hara H. **10-Hydroxy-2-decenoic acid, a major fatty acid from royal jelly, inhibits vegf-induced angiogenesis in human umbilical vein endothelial cells**. *Evid Based Complement Alternat Med.* (2009) **6** 489-94. DOI: 10.1093/ecam/nem152
40. Fujii A, Kobayashi S, Kuboyama N, Furukawa Y, Kaneko Y, Ishihama S. **Augmentation of wound healing by royal jelly. (Rj) in streptozotocin-diabetic rats.**. *Jpn J Pharmacol.* (1990) **53** 331-7. DOI: 10.1254/jjp.53.331
41. Nelson DR. **World of cytochrome P450s**. *Philos Trans R Soc Lond B Biol Sci.* (2013) **368** 20120430. DOI: 10.1098/rstb.2012.0430
42. Yore MM, Syed I, Moraes-Vieira PM, Zhang T, Herman MA, Homan EA. **Discovery of a class of endogenous mammalian lipids with anti-diabetic and anti-inflammatory effects**. *Cell.* (2014) **159** 318-32. DOI: 10.1016/j.cell.2014.09.035
43. Hamad ARA, Sadasivam M, Rabb H. **hybrid lipids, peptides, and lymphocytes: new era in type 1 diabetes research**. *Journal of Clinical Investigation.* (2019) **129** 3527-9. DOI: 10.1172/JCI130313
44. Kromhout D, Yasuda S, Geleijnse JM, Shimokawa H. **Fish oil and omega-3 fatty acids in cardiovascular disease: do they really work?**. *Eur Heart J.* (2012) **33** 436-43. DOI: 10.1093/eurheartj/ehr362
45. Simopoulos AP. **The importance of the omega-6/omega-3 fatty acid ratio in cardiovascular disease and other chronic diseases**. *Exp Biol Med.* (2008) **233** 674-88. DOI: 10.3181/0711-MR-311
46. Farvid MS, Ding M, Pan A, Sun Q, Chiuve SE, Steffen LM. **Dietary linoleic acid and risk of coronary heart disease: a systematic review and meta-analysis of prospective cohort studies**. *Circulation.* (2014) **130** 1568-78. DOI: 10.1161/CIRCULATIONAHA.114.010236
|
---
title: 'The status quo, contributors, consequences and models of digital overuse/problematic
use in preschoolers: A scoping review'
authors:
- Chenggong Wang
- Haoyue Qian
- Hui Li
- Dandan Wu
journal: Frontiers in Psychology
year: 2023
pmcid: PMC9980908
doi: 10.3389/fpsyg.2023.1049102
license: CC BY 4.0
---
# The status quo, contributors, consequences and models of digital overuse/problematic use in preschoolers: A scoping review
## Abstract
Digital devices play a critical role in preschoolers’ learning and development. Despite the evidence that digital devices use may facilitate preschoolers’ learning and development, their overuse/problematic use has become a global concern as the popularity and widespread use of digital devices. This scoping review aims to synthesize the empirical evidence to identify the status quo, influential factors, developmental outcomes, and models of overuse/problematic use in preschoolers. This search has identified 36 studies published in international peer-reviewed journals during 2001–2021, converging into four common topics: the current situation, the influential factors, the consequences, and the models. First, the average percentages of overuse and problematic use across the studies collected in this research were $48.34\%$, and $26.83\%$, separately. Second, two influential factors were identified: [1] children’s characteristics and [2] parental and family factors. Third, early digital overuse/problematic use was found to have a negative impact on the following domains: [1] physical health, [2] psychosocial health, [3] problematic behaviors, and [4] cognitive development; Fourth, most relevant studies adopted general linear models, while few of them adopted experimental designs. Finally, the implications for future studies and practical improvements are also addressed.
## Introduction
Nowadays, digital technology is advancing at an unprecedented rate and dramatically shaping children’s daily lives and early development (Dong and Mertala, 2021). In particular, the COVID-19 pandemic and the associated lockdowns have forced preschoolers (ages 3–6) to learn online at home with various digital devices (Dong et al., 2020). Preschool years are a critical period of psychosocial and cognitive development and may influence life-long screen habits (Jimenez et al., 2016; Radesky and Christakis, 2016). Additionally, preschool years are also characterized by large amounts of brain plasticity. Therefore, these young minds are very sensitive and vulnerable to the effects of overusing digital devices such as smartphones, iPads, notebooks, etc. Although some studies have shown that digital devices use could promote the development of young children, such as engaging them in collaborative learning, reasoning, and problem-solving (Plowman and Stephen, 2003; Yelland, 2006). However, an increasing volume of evidence suggests that digital overuse/problematic use is evolving into a critical risk factor that threatens preschoolers’ health and well-being (Rocha and Nunes, 2020; Mallawaarachchi et al., 2022). However, there are very few empirical studies on preschoolers’ digital overuse/problematic use; thus, there is little evidence for high-stake policy-making regarding early childhood health and education matters. Therefore, a synthesis or scoping review is needed to depict the whole picture of what has been explored and reported about the topic. To meet this urgent need, this scoping review examines a wide breadth of research resources during the past two decades (2000–2021) related to the exploratory questions regarding early digital overuse/problematic use.
## Digital overuse/problematic use in preschoolers
During this digital era, children are inevitably exposed to digital media earlier in their lives and for a longer time (Dong et al., 2020). For instance, Rideout and Robb [2020] reported that 75–$96\%$ of infants use media daily, and this new generation of children is called ‘digital citizen’ or ‘digital native’ (Dong and Mertala, 2021). However, the increased screen time in preschooler’s daily life has raised concerns among public health organizations, parents, and scholars (Rocha and Nunes, 2020). Screen time refers to the time an individual spends on devices with a screen, including smartphones, tablets, computers, and televisions (Dong et al., 2020). The American Academy of Pediatrics (AAP) recommended that parents avoid digital media use in children younger than 18 to 24 months and use less than an hour per day for children aged 2 to 5 years (American Academy of Pediatrics, 2001). According to the World Health Organization (WHO), children aged below 3 are not recommended to use any digital media, and children aged 3–4 can use digital media in less than 1 h (WHO, 2019).
However, existing studies have found that preschoolers spend more screen time than the WHO standards. For example, a national survey in the United States found that toddlers (under 2 years old) spent an average of one hour of screen media per day, as reported by their parents (Lauricella et al., 2015). Coincidently, researchers also found that preschoolers consumed an average of 2–3 h per day in front of various screens (McNeill et al., 2019). All these findings have jointly confirmed that digital overuse is prevalent among preschoolers. Furthermore, Yalçin et al. ( 2021a) reported that problematic media use could be observed in infancy and toddler (under age 3). Domoff et al. [ 2020] defined ‘problematic media use’ as excessive media use that interferes with a child’s functioning, which captures dysfunctional social, behavioral, and/or academic development. It is due to excessive or maladaptive media use as evidenced by the following behaviors: [1] loss of interest in other activities; [2] preoccupation with media; [3] withdrawal from others; [4] high tolerance for media; and [5] deceptive behaviors surrounding media (Domoff et al., 2020). However, few studies have explored the factors and outcomes of digital overuse/problematic use in preschoolers. This scoping review collects all the existing studies about this topic, focusing on how early digital overuse/problematic use affects children’s development and what lessons we can learn from these studies.
In addition, family environment is an inevitable factor that is highly relevant and influential to early digital overuse/problematic use. Children’s screen time is directly associated with the related practices of their parents, such as parents’ digital addiction, parental depression, and parenting style (Lam, 2015). In particular, the status of parental digital use is also a key factor influencing children’s digital use (Dong and Mertala, 2021). Therefore, an all-around understanding of all these factors will allow us to gain a more comprehensive picture of early digital overuse/problematic use. However, most studies have focused on these factors without investigating the interactions of two or all. Therefore, another aim of this study is to summarize the factors identified in the existing studies and to propose a model for future studies.
## Theoretical framework: Developmental cascade model
This scoping review was guided by the developmental cascade model, demonstrating how risks developed in earlier developmental periods cascade into widespread difficulties later (Gottlieb, 2007; Masten and Cicchetti, 2010). This model suggests that cumulative developmental consequences in developing systems could spread effects across levels, domains, systems, and even generations (Masten and Cicchetti, 2010). In this study, digital overuse/problematic use represents a risk from the behavioral domain. Therefore, it may lead to different developmental issues, which might, in turn, cause some negative outcomes affecting other domains (e.g., later academic achievement) through altering cognitive, neural, physical, and psychosocial development. This model is consistent with Bronfenbrenner’s bioecological framework (Bronfenbrenner, 1977, 1979), suggesting that development is a highly interactive process positioned within concentric circles of mutual influence. Based on this model (Figure 1), we have reviewed the existing literature to address the leading research problem: what are the status quo, influential factors, consequences, and models of digital overuse/problematic use in preschoolers?
**Figure 1:** *Theoretical model.*
## Methods
The scoping review methodology adopted in this study has been widely applied to summarize research findings for policymakers, or practitioners, identify research gaps and establish the areas for future research (Arksey and O'Malley, 2005; Levac et al., 2010). This method enables the current study to explore the width and depth of existing studies on digital overuse/problematic use (including TV, tablet, smartphone, and video game) and its factors/outcomes, identify research gaps, and establish the areas for future research. Specifically, this study has searched, identified, collected, and examined the potential sources for their relevance to the research objectives and mapped them to the key themes and concepts underpinning the research questions. In this article, we followed the 5-step framework suggested by Arksey and O'Malley [2005]: [1] articulating the research questions; [2] identifying relevant studies; [3] studies selection; [4] charting the data; and [5] collating, summarizing and reporting the results.
## Phase 1: Articulating the research questions
The following questions were proposed to guide this study: [1] *What is* the status of digital use among preschoolers; [2] What are the influential factors of digital overuse in the preschool period? [ 3] What are the outcomes of digital overuse on children? [ 4] What statistic models have been adopted to describe the relationship among all the critical variables in this field of studies? [ 5] What are the research gaps in this field of study?
## Phase 2: Identifying relevant studies
The authors conceived the research questions through a series of discussions, and the first author consulted an expert in this field to identify the appropriate search terms and databases. As a result, an extensive automated search of peer-reviewed articles in the three databases (ProQuest, Web of Science, and Google Scholar) was conducted in April 2022. The literature search aimed to thoroughly identify all the research articles on “early digital use and development” published during 2001–2022. However, the studies published in 2022 did not meet the inclusion criteria. Therefore, using a full year as a cut-off point, we included studies from 2001 to 2021. Three different sets of terms with two Boolean operators (AND and OR) were utilized to search for and extract relevant literature from the databases: (screen time OR digital use OR digital overuse OR problematic digital use OR digital addiction OR TV OR smartphone OR tablet OR video game OR internet game) AND (infant OR toddler OR preschool OR prekindergarten OR kindergarten OR preschoolers OR preschoolers OR kindergarteners OR children) AND (cognition OR cognitive development OR mental health OR psychosocial health OR problematic behaviors OR behavioral problem OR physical health OR body mass index). Search terms were created via extensive piloting.
## Phase 3: Selecting studies
A set of criteria was employed to ensure that only full-text, English, peer-reviewed journal articles meeting the objectives of this systematic review were included. The inclusion criteria were as follows: We excluded the articles that: [1] were not an original study, but a case report, review, commentary, erratum, or letter to the editor; [2] were original studies without empirical data, such as only semi-structured interviews or qualitative analysis; [3] studied children aged 6 years or up.
As shown in Figure 2, the final search yielded 2,786 articles, of which 2,665 duplicates were removed. The first and second authors independently reviewed and selected the articles based on the inclusion criteria, and the agreement was $95.12\%$. Next, the authors screened full-text articles and extracted data from those that met the inclusion criteria. Due to the COVID-19 lockdowns, the authors maintained online communication throughout the full-text review process to resolve conflicts and maintain consistency. Of all the studies included for full-text review, 65 articles were excluded by title and abstract. Out of the 56 full-text studies assessed for eligibility, 20 were excluded. The authors discussed the studies that were uncertain whether to be eligible until reaching $100\%$ agreements. Finally, a total of 36 articles were eligible for review.
**Figure 2:** *The selection of studies included in this scoping review.*
## Phase 4: Charting the data
The 36 articles were charted to examine the types of research identified (see Table 1). The aggregate number of preschoolers in this scoping review was 49,126, and the sample size ranged from 38 to 20,324. The samples were recruited from 15 countries across multiple geographical regions, including Europe ($$n = 9$$), Asia ($$n = 14$$), North America ($$n = 6$$), and other countries/regions ($$n = 7$$). Most were cross-sectional studies ($$n = 31$$), and the others were longitudinal ones ($$n = 5$$). Most were general survey studies ($$n = 34$$), and the rest were experimental studies ($$n = 2$$). In particular, four articles focus on the status quo of digital use, 13 on the influential factors, and 25 on the outcomes. Among the 25 outcome-related articles, 10 focus on the influence of digital use on cognitive development, four on its influence on psychosocial development, seven on the influence on children’s behaviors, and 10 on the influence on physical health. Almost all the studies ($$n = 35$$) sampled preschoolers (under Age 6), whereas one study had some participants over age 6.
**Table 1**
| Author/s (Year): Country | Research topic | Sample size | Age range | Research design | Modeling | Major findings |
| --- | --- | --- | --- | --- | --- | --- |
| Kim et al. (2021): South Korea | Factor | 400 | 2–5 years | Cross-sectional | Logistic regression | Mother’s smartphone addiction positively predicts children’s early smartphone exposure. However, no correlation was found between mother’s smartphone addiction and child’s smartphone use time. |
| Konok et al. (2021): Hungary | Outcome | 40 (study1); 56 (study2) | 4–6 years | Cross-sectional | Linear regression | MTSDs use was associated with global precedence in selective attention tasks but an atypical, local precedence in a divided attention task. More importantly, playing with a digital game eliminated the advantage of selective attention over divided attention observed in the non-digital and slow digital game conditions. Besides, MTSD use was not associated with emotion recognition but with the worse theory of mind. |
| Madigan et al. (2020): Canada | Status | 3589 | 2–3 years | Longitudinal | Logistic regression | At ages two and three years, most children did not meet screen time pediatric guidelines (< 7 h per week). Besides, maternal screen time is positively associated with exceeding the screen time guidelines. |
| Velumani et al. (2021): India | Outcome | 280 | 12–36 months | Cross-sectional | Linear regression | The level of screen dependency positively predicts the degree of child nourishment. |
| Coyne et al. (2021): USA | Factor | 269 | 24–36 months | Cross-sectional | Structural Equation Modelling | Higher levels of media emotion regulation were associated with more problematic media use and more extreme emotions when media was removed in toddlers. Toddle’s temperament (precisely the dimensions of negative affect and surgency) influenced problematic media use and extreme emotions, and their relationship was mediated by media emotion regulation. |
| Lehrl et al. (2021): Germany | Factor | 4914 | 0–5 years | Cross-sectional | Multiple regression (including moderation effect) | Toddlers with more analogy home learning activities (e.g., parent–child activities including playing word games, reading, and counting) showed less frequent digital activities. Digital HLE activities resulted in weaker socio-emotional skills for preschoolers. Analog HLE moderated the effect of digital HLE on children’s language skills. |
| Xie et al. (2020): China | Factor & Outcome | 1897 | 3–6 years | Cross-sectional | Multiple regression | Screen time was strongly associated with preschoolers’ socioeconomic status (gender, household location, maternal education). In addition, preschoolers with screen time over 60 min per day had more behavioral problems (total and externalizing behaviors) than those less than 60 min per day. |
| Tay et al. (2021): Singapore | Factor | 3413 | 2–7 years | Cross-sectional | ANOVA | Parents’ guidance toward digital use was positively related to preschoolers’ time spent using digital media. |
| Anitha et al. (2021): India | Outcome | 348 | 1.5–5 years | Cross-sectional | Chi-square test | Children under-five years of age, compared to screen time < 2 h per day, children with screen time > 2 h per day and media addiction showed more clinically developmental disorder problems and ADHD problems. |
| Suherman et al. (2021): Indonesia | Factor | 104 | 3–6 years | Cross-sectional | Spearman Rho Test | Children under authoritative parenting style had less level of gadget addiction. |
| Cho and Lee (2017): South Korea | Factor & Outcome | 303 | 1–6 years | Cross-sectional | Hierarchical regression (including mediation effect) | All addictive tendencies had significant positive effects on problematic behaviors and significant negative effects on emotional intelligence. Parents’ self-evaluative of their smartphone usage mediated the effect of children’s smartphone addiction proneness (such as voluntary isolation and personality distortion) on their problematic behaviors. |
| Özyurt et al. (2018): Turkey | Factor | 76 | 3–6 years | Cross-sectional | Wilcoxon test | After conducting a parental training program (Triple P), parental perceived educational purposes of using digital devices changed, and the duration of their children’s digital device use decreased. |
| Jusienė et al. (2020): Lithuania | Outcome | 190 | 4–5 years | Cross-sectional | Multiple linear regression | Executive functioning measures were not significantly predicted by MTSD use. |
| Poulain et al. (2018): Germany | Outcome | 527 | 2–6 years | Longitudinal | Multiple regression | Baseline use of mobile phones was significantly associated with more conduct problems and hyperactivity or inattention at follow-up. Further, peer relationship problems at baseline were significantly associated with greater mobile phone use at follow-up. No significant associations were present between mobile phone use and emotional problems at baseline/ follow-up |
| Baek et al. (2013): South Korea | Factor | 488 | 0–5 years | Cross-sectional | Chi-square test | Compared to mothers with high cognitive and emotional efficacy, those with low cognitive and emotional efficacy allowed their children to use smartphones more frequently. |
| Keefe-Cooperman (2016): USA | Factor & Outcome | 492 | 3–5 years | Cross-sectional | Bivariate Correlation & ANOVA | Preschoolers with greater usage time of digital device had lower WPPSI-IV Visual Spatial Composite scores and Full-Scale IQ scores, on average. Lower maternal education, lower SES, and being from a historically disadvantaged background were associated with greater usage time of digital device. |
| Yalçin et al. (2021a): Turkey | Factor | 1245 | 2–5 years | Cross-sectional | Chi-square test & Logistic regression | The playing video games were partly predicted by child and family characteristics. |
| Yalçin et al. (2021b): Turkey | Status & Factor | 1245 | 2–6 years | Cross-sectional | Multiple logistic regression | The family, child, and screen use characteristics partly predicted problematic screen exposure. |
| Chang et al. (2018): South Korea | Status | 390 | 2–5 years | Cross-sectional | | TV and smartphones were the most popular digital devices used by toddlers. Most toddlers began using smart devices at 12–24 months. |
| Zhao et al. (2018): China | Outcome | 20324 | 3–4 years | Cross-sectional | Logistic regression (including mediation effect) | Every additional hour of screen time was associated with an increased risk for poor psychosocial well-being. In addition, body mass index, sleep duration, and parent–child interaction mediated the effect of excessive screen time on children’s psychosocial well-being, among which parent–child interaction contributed the most. |
| Beyens and Nathanson (2019): Netherlands | Outcome | 402 | 3–5 years | Cross-sectional | Multiple regression | Heavier television and tablet use were associated with later bedtime and later wake time, but not with fewer hours of sleep. In addition, heavier daily television use and evening smartphone use were associated with increased daytime napping. Moreover, heavier daily television use, daily and evening smartphone use, and evening tablet use were associated with poorer sleep consolidation. |
| Collings et al. (2018): UK | Outcome | 1338 | 1–3 years | Longitudinal | Linear regression | Every 1 h per day of TV viewing could predict a larger waist circumference. |
| Cox et al. (2012): Australian | Outcome | 135 | 2–6 years | Cross-sectional | Hierarchical regression (including mediation effect) | Weekday TV viewing positively impacts children’s BMI z-score, and this effect is mediated by sedentary behavior, not the kilojoule intake during TV viewing. |
| Sijtsma et al. (2015): Netherlands | Outcome | 759 | 3.4–4.4 years | Cross-sectional | Ordinary least square regression (including mediation effect) | A television in the bedroom or more televisions at home gave a higher screen time, which was associated with decreased sleep duration and resulted in higher BMI. The preschool children’s screen time and sleep duration mediated the relationship between home television ownership and BMI. |
| Li et al. (2021): Australia | Outcome | 38 | 4–6.3 years | Cross-sectional | t-test | Compared to the ‘non-digital user’, ‘heavy-digital user’ performed poorer in the Dimensional Change Card Sort task and lower activation of the prefrontal cortex (BA 9) |
| Hutton et al. (2020): USA | Outcome | 69 | 36–63 months | Cross-sectional | Spearman’s ρ | Access to child’s own smartphone and tablet was negatively correlated with Get Ready to Read score of emergent literacy and CTOPP score of processing speed. Access to child’s own smartphone and tablet was only marginally negatively correlated with the other language and literacy measures. |
| Cheung et al. (2017): UK | Outcome | 715 | 6–36 months | Cross-sectional | Path analysis | Tablet use was significantly associated with a reduced overall amount of sleep and delayed sleep onset. However, tablet use was not significantly associated with frequency of night awakenings |
| Gülay Ogelman et al. (2018): Turkey | Outcome | 162 | 5–6 years | Cross-sectional | Linear Regression | the use of mobile technologies was no predictive effect on the children’s social skill levels. Tablet use was not associated with social status. Smartphone use was significantly associated with lower social preferences in children. |
| McDaniel and Radesky (2020): USA | Factor & Outcome | 183 | 1–5 years | Longitudinal | Structural Equation Modelling | Greater child externalizing behavior significantly predicted greater tablet use (not phone use) at follow-up via parenting stress (based on structural equation modeling). However, greater smartphone and tablet use did not significantly predict later externalizing behavior. |
| McNeill et al. (2019): Australia | Outcome | 185 | 3–5 years | Longitudinal | Linear regression | High-dose app users at baseline had a significantly lower inhibition score at follow-up than low-dose app users; App use did not significantly predict other cognitive outcomes at follow-up |
| van den Heuvel et al. (2019): Canada | Outcome | 893 | 18 months | Cross-sectional | Logistic regression | For children who used a smartphone and tablet, each additional 30-min increase in daily smartphone and tablet use was significantly associated with increased odds of parent-reported expressive speech delay. However, use was not significantly associated with other parent-reported communication delays |
| Lin et al. (2020): China | Outcome | 161 | 18–36 months | Cross-sectional | Multiple regression | Smartphone and tablet use were significantly correlated with language development. However, when confounding variables were controlled for, the association was no longer significant, i.e., children who spent more time on smartphone and tablets were not more likely to have language delay. |
| Borajy et al. (2019): Saudi Arabia | Outcome | 74 | 1.5–3 years | Cross-sectional | Linear regression and Logistic regression | Child’s smartphone and tablet use did not significantly influence the odds of having speech delay |
| Lan et al. (2020): China | Outcome | 2903 | 2–6 years | Cross-sectional | Linear regression | Each additional hour spent on smartphones and tablets was independently associated with a reduction in daily sleep duration of 11 and 6 min in boys and girls, respectively. Compared to non-portable devices, use of portable ones was more closely associated with short sleep duration |
| Moon et al. (2019): South Korea | Outcome | 117 | 3–5 years | Cross-sectional | Spearman correlation | Smart device usage frequency positively correlated with three-year-old children’s fine motor skill development. In addition, smart device usage level was positively correlated with social development. However, smart device usage time was negatively correlated with expressive language months. No such correlations were found in children aged four to five years. |
| Nathanson and Beyens (2018): Netherlands | Outcome | 402 | 3–5 years | Cross-sectional | Multiple regression | Heavier evening and daily tablet use (and, to some extent, smartphone use) were related to sleep disturbances. Besides, playing games on MEDs at bedtime was related to compromised sleep duration |
## Phase 5: Collating, summarizing, and reporting results
We extracted and collated the following essential information: author/s (year): country, research topic, sample size, age range of participator, research design, statistic model, and major findings. The first author independently reviewed the included articles and extracted data using a pre-established coding scheme. This coding scheme is used to collate and summarize the sources in four aspects, including [1] the status of children’s digital use; [2] the influential factors of digital use among preschoolers, [3] the outcomes of digital use in preschoolers, and [4] the statistic models used in this field. Any inconsistency was resolved through discussion and consensus with the co-author (s).
## The status quo of digital use
Among all the 36 studies, four have explored the status of children’s digital use, which mainly reported [1] overuse or problematic usage behaviors, [2] the frequency or children’s time spent on digital use, and [3] the types of digital devices being used. First, three studies reported severe problematic usage of digital devices. For example, Madigan et al. [ 2020] found that most of the 2-year-old and 3-year-old children’s screen time exceeded the line set by the WHO guidelines. Yalçin et al. ( 2021a) found that $22.5\%$ of children had a problematic screen exposure score of ≥7 (defined as a high level, total score = 13), while the median score of the problematic screen exposure of the children was 4 (Interquartile Range: 3–6). Second, one study was concerned about children’s time spent on digital use. Chang et al. [ 2018] found that $39.3\%$ of the 390 toddlers watched TV almost every day, while $12.0\%$ of children used smartphones daily. In particular, more children and time had been spent on digital devices on weekends than on weekdays from a very young age (24 months old). Third, one study has focused on the digital devices used. Tay et al. [ 2021] found that children aged 2 to 4 spent 1.19 h per day on digital entertainment, with television and mobile phone being the most popular devices. In summary, the average percentages of overuse and problematic use across the studies collected in this research were 48.34 and $26.83\%$, separately (see Table 2).
**Table 2**
| Citation | Country | Digital overuse/Problematic use rate |
| --- | --- | --- |
| Kim et al. (2021) | South Korea | |
| Konok et al. (2021) | Hungary | |
| Madigan et al. (2020) | Canada | 87.9%† |
| Velumani et al. (2021) | India | 82.2%† |
| Coyne et al. (2021) | USA | |
| Lehrl et al. (2021) | Germany | |
| Xie et al. (2020) | China | 54.8%† |
| Tay et al. (2021) | Singapore | 29.9%† |
| Anitha et al. (2021) | India | 28.1%* |
| Suherman et al. (2021) | Indonesia | 29.9%* |
| Cho and Lee (2017) | South Korea | 12.2%† (smartphone) |
| Özyurt et al. (2018) | Turkey | 39.6%† (TV) |
| Jusienė et al. (2020) | Lithuania | 48.1%† |
| Poulain et al. (2018) | Germany | 20.0%† (TV) |
| Baek et al. (2013) | South Korea | 28.6%† (Smartphone) |
| Keefe-Cooperman (2016) | USA | |
| Yalçin et al. (2021a) | Turkey | 22.5%* |
| Chang et al. (2018) | South Korea | 63.1%† |
| Yalçin et al. (2021b) | Turkey | 56.7%† |
| Zhao et al. (2018) | China | 78.6%† |
| Beyens and Nathanson (2019) | Netherlands | |
| Collings et al. (2018) | UK | 62.0%† |
| Cox et al. (2012) | Australian | |
| Sijtsma et al. (2015) | Netherlands | |
| Li et al. (2021) | Australian | |
| Hutton et al. (2020) | USA | |
| Cheung et al. (2017) | UK | |
| Gülay Ogelman et al. (2018) | Turkey | |
| McDaniel and Radesky (2020) | USA | 47.1%† (TV) |
| McNeill et al. (2019) | Australian | 23.8%† (App) |
| van den Heuvel et al. (2019) | Canada | 22.4%† (Mobile media devices) |
| Lin et al. (2020) | China | |
| Borajy et al. (2019) | Saudi Arabia | |
| Lan et al. (2020) | China | 73.9%† |
| Moon et al. (2019) | South Korea | 39.30%† |
| Nathanson and Beyens (2018) | Netherlands | |
| Average | | 48.34%† (Overuse) |
| Average | | 26.83%* (Problematic use) |
## The influential factors of digital use
Among all the 36 studies, 12 have explored the factors that influenced children’s digital use, mainly focusing on two essential aspects: [1] children’s characteristics and [2] parental and family factors (see Table 3).
**Table 3**
| Factors | Research |
| --- | --- |
| Children’s characteristics | Children’s characteristics |
| Cultural background (Minority background) | Keefe-Cooperman (2016) |
| Gender (Boy) | Paulus et al. (2018); Xie et al. (2020); Yalçin et al. (2021a) |
| Psychological and behavioral problem | McDaniel and Radesky (2020); negative affect and surgency: Coyne et al. (2021) |
| Peer relationship | Poulain et al. (2018) |
| Number of sisters and brothers | Yalçin et al. (2021a,b) |
| Parental and family factors | Parental and family factors |
| Parent media time | Cho and Lee (2017); Madigan et al. (2020); Coyne et al. (2021); Mother: Kim et al. (2021) |
| Parental attitudes toward children’s digital device use | Özyurt et al. (2018); Lehrl et al. (2021); Tay et al. (2021); Yalçin et al. (2021a) |
| Parenting behavior | Baek et al. (2013); Suherman et al. (2021) |
| Socioeconomic Status | Keefe-Cooperman (2016); Xie et al. (2020); Yalçin et al. (2021a,b) |
| Psychological health | Depressed: Kim et al. (2021); Parenting stress: McDaniel and Radesky (2020) |
## Children’s characteristics
Altogether eight studies found that children’s biological and sociocultural status influenced their early digital use. First, gender was found to be related to early digital use. For instance, Paulus et al. [ 2018] found that boys were more often and much easier to encounter computer gaming disorder than girls. Moreover, they also found that children with attention deficit hyperactivity disorder (ADHD) showed significantly higher scores in computer gaming disorder evaluation, and clinically relevant inattention scores predicted longer and more computer gaming. Later, Xie et al. [ 2020] revealed that boys had significantly more time on digital screens than girls. Similarly, Yalçin et al. ( 2021a) also reported that gender could significantly predict digital use. Second, children’s psychological and behavioral problem also affect their digital use. For example, Coyne et al. [ 2021] found individual’s temperament (specifically negative affect and surgency) contributed to problematic media use and extreme emotions, and their relationship was mediated by media emotion regulation. McDaniel and Radesky [2020] found that externalizing behavioral problems significantly predicted greater tablet use (not phone use) at follow-up via parenting stress. Parents of preschoolers with externalizing behavior are more likely to use media as a behavior modifier or babysitter (Nikken and Schols, 2015). In particular, mothers of children with externalizing behavior problems are under more pressure to raise their children and have no reasonable solutions for the externalizing behaviors their children exhibit. Therefore, allowing the child to use digital devices, such as playing games and watching animation, becomes a way to calm their children. Previous studies have verified this interpretation. For example, infants with regulatory problems (such as self-soothing difficulties and impulsive/demanding behaviors) were found to consume more TV and videos and were more likely to be given mobile devices for individual use (Radesky et al., 2016; Levine et al., 2019). Third, Paulus et al. [ 2018] found that peer relationship problems at baseline were significantly associated with greater mobile phone use at follow-up. This finding suggests that children with less social exposure may be prone to electronic product dependence when they grow up, and timely assessment and intervention for problematic digital use in these children is necessary. Fourth, children’s cultural backgrounds also mattered. For example, Keefe-Cooperman [2016] found that younger children with a minority background used the digital device more than European American preschoolers.
## Parental and family factors
Seventeen studies have explored the parental factors related to early digital use, covering a wide range of factors, including parental digital use, parental attitudes toward children’s digital device use, parenting style & efficacy, family socioeconomic status (SES), and psychological health of the parents. First, parents’ digital use/problematic use was an important factor in their children’s digital overuse/problematic use. Moreover, four research studies have addressed this issue. For example, Cho and Lee [2017] found that parental smartphone usage caused smartphone problematic use proneness in their children, further leading to various problems such as interference with daily life and voluntary isolation. Madigan et al. [ 2020] found that maternal screen time use predicted preschoolers’ exceeding digital use guidelines. Kim et al. [ 2021] found that children’s first smartphone exposure was predicted by maternal smartphone addiction, while mothers’ smartphone addiction did not predict the recent smartphone use time spent by children. Recently, Coyne et al. [ 2021] showed that parent media time was related to the children’s problematic media use, with longer parent media time corresponding to severe problematic media use.
Second, four studies have addressed the role that parental beliefs and practices play in digital use. For instance, Özyurt et al. [ 2018] found that after attending a parental training program (Triple P), parents’ perceived educational purposes for digital use changed, and their children’s digital use time also declined. Later, Tay et al. [ 2021] found that parents’ guidance toward digital use (such as limiting screen time to 1 h per day and introducing high-quality educational programs) was positively related to the amount of time spent using digital media by preschoolers. In addition, Lehrl et al. [ 2021] found that even toddlers with more analogy home learning activities (e.g., parent–child activities including playing word games, reading, and counting) showed less frequent digital activities. Finally, Yalçin et al. ( 2021a) found that parents setting rules for preschoolers’ screen use could predict the state of children’s video game play.
Third, parenting style and efficacy also contribute to early digital use. For instance, Baek et al. [ 2013] found the effect of parental efficacy on preschoolers’ smartphone use. Specifically, mothers capable of solving problems and having a positive identity could control their children’s smartphone use and be aware of the positive aspects of smartphones. And the parents restricting children’s digital use helped reduce the frequency and time of smartphone use, reducing problematic levels. Recently, Suherman et al. [ 2021] revealed a significant relationship between parenting style and problematic digital use among preschoolers. Specifically, children under the authoritative parenting style had less digital problematic use.
Fourth, four studies jointly indicated an SES effect in early digital use. For instance, Keefe-Cooperman [2016] found that parents with lower SES were linked to the children’s greater usage time of digital devices. Next, Xie et al. [ 2020] reported that children’s screen time was closely correlated with their household location (urban or rural) and maternal education, as the children who lived in a rural area or with low-educated parents spent more screen time. Recently, Yalçin et al. ( 2021a) found that video game overuse by preschoolers could be predicted by their parental education level and the number of children in the family. Later, Yalçin et al. ( 2021b) confirmed that the frequency of problematic screen exposure varied in parental educational levels, maternal occupation, family type and size, and settlement type (urban or rural).
Fifth, two studies (Yalçin et al., 2021a,b) showed that the number of children in a family was related to preschoolers’ digital use. In particular, they found that if someone else was commonly playing video games at home, it resulted in a higher possibility and earlier start time point of video game playing for the children in this family. The findings suggest that the increased number of children gives caregivers less control over each child’s use of digital devices and, accordingly, high levels of digital use among preschoolers.
Finally, two studies found that parental psychological health affected preschoolers’ digital use. For example, McDaniel and Radesky [2020] confirmed an association between parenting stress and greater tablet use in preschoolers. Later, Kim et al. [ 2021] found that depressed mothers were more likely to have digital addiction, resulting in their children’s earlier digital use.
## The outcomes of children’s digital use
Among all the 36 studies, 26 have investigated the outcomes of children’s digital use, focusing on [1] children’s physical health, [2] psychosocial development, [3] problematic behaviors, and [4] cognitive development (see Table 4).
**Table 4**
| Domain | Factors | Positive association | Negative association | No association | Unnamed: 5 | Unnamed: 6 |
| --- | --- | --- | --- | --- | --- | --- |
| Physical health | Adiposity | Cox et al. (2012); Sijtsma et al. (2015); Collings et al. (2018); Velumani et al. (2021) | | | | |
| Physical health | Physical activity | | Cox et al. (2012); Sijtsma et al. (2015) | | | |
| Physical health | Motor skill | Moon et al. (2019) | | Keefe-Cooperman (2016) | | |
| Physical health | Sleep quality | Daytime napping: Beyens and Nathanson (2019) | Sleep consolidation: Beyens and Nathanson (2019); Sleep onset: Cheung et al. (2017) | Frequency of night awakenings: Cheung et al. (2017) | | |
| Physical health | Sleep duration | | Sijtsma et al. (2015); Cheung et al. (2017); Nathanson and Beyens (2018); Lan et al. (2020) | Beyens and Nathanson (2019) | | |
| Psychosocial development | Psychosocial wellbeing | | Zhao et al. (2018) | | | |
| Psychosocial development | Emotion ability | | Emotional intelligence: Cho and Lee (2017) | Emotion recognition: Konok et al. (2021) | | |
| Psychosocial development | Theory of mind | | Konok et al. (2021) | | | |
| Psychosocial development | Social skill | | Gülay Ogelman et al. (2018) | | | |
| Problematic behaviors | Externalizing behavioral | Hyperactivity/ Inattention: Poulain et al. (2018); Xie et al. (2020); Anitha et al. (2021); Lehrl et al. (2021); Conduct problems: Poulain et al. (2018); Anitha et al. (2021); Lehrl et al. (2021); Aggressive behaviors: Lin et al. (2020) | | McDaniel and Radesky (2020) | | |
| Problematic behaviors | Pervasive developmental disorder | Anitha et al. (2021) | | | Emotional problems | Poulain et al. (2018); Lin et al. (2020) |
| Cognitive development | Attentional control | | Konok et al. (2021) | | | |
| Cognitive development | Executive function | | Li et al. (2021); Inhibition control: McNeill et al. (2019) | Jusienė et al. (2020); Working memory: McNeill et al. (2019) | | |
| Cognitive development | Visual–spatial abilities and Intelligence | | Keefe-Cooperman (2016) | | | |
| Cognitive development | Language and literacy | | Moon et al. (2019); van den Heuvel et al. (2019); Hutton et al. (2020) | Borajy et al. (2019); Lin et al. (2020) | | |
## Physical health
Nine articles addressed the impact of early digital use on children’s physical health, covering adiposity (Cox et al., 2012; Sijtsma et al., 2015; Collings et al., 2018; Velumani et al., 2021), physical activity (Cox et al., 2012; Sijtsma et al., 2015), motor skill (Keefe-Cooperman, 2016; Moon et al., 2019), and sleep duration and quality (Sijtsma et al., 2015; Cheung et al., 2017; Nathanson and Beyens, 2018; Beyens and Nathanson, 2019; Lan et al., 2020). In addition, four studies consistently found that early digital devices use could cause adiposity. For instance, Cox et al. [ 2012] reported that weekday TV viewing was positively correlated with child BMI z-score. More critical, sedentary behavior, not the kilojoule intake during TV viewing, mediated the positive effect of weekday TV viewing on children’s BMI z-score. Later, Sijtsma et al. [ 2015] found that longer screen time was associated with a higher BMI. They further revealed that a TV in the bedroom or more TVs at home gave a higher screen time, decreasing sleep duration and resulting in higher BMI. In addition, Collings et al. [ 2018] confirmed that every 1 h per day of TV viewing significantly resulted in a larger waist circumference. Recently, Velumani et al. [ 2021] revealed that the level of toddlers’ nourishment (such as normal weight, overweight, and obese) was positively predicted by the level of screen dependency.
Additionally, two studies revealed that early digital use would decrease children’s physical activity. Cox et al. [ 2012] found that weekday and weekend TV viewing was positively associated with the minutes spent in sedentary activities. Meanwhile, Sijtsma et al. [ 2015] also found that longer screen time was associated with less outdoor play. Another two articles explored the relationship between early digital use and motor skill development. Keefe-Cooperman [2016] revealed that screen time was not significantly correlated with the fine motor quotient. However, Moon et al. [ 2019] demonstrated that the use frequency of table or smartphone was positively associated with fine motor skill development in three-year-old children.
More importantly, five articles focus on the influences of digital device use on preschoolers’ sleep. Regarding sleep duration, Sijtsma et al. [ 2015] found that a TV in the bedroom or more TVs at home gave a higher screen time and further decreased sleep duration. Cheung et al. [ 2017] also found that tablet use was significantly associated with reduced sleep and delayed onset of sleep. In addition, Nathanson and Beyens [2018] found that heavier evening and daily tablet use (and, to some extent, smartphone use) contributed to sleep disturbances. Later, Beyens and Nathanson [2019] found that heavier TV and tablet use in the evening caused late bedtimes and wake-up times but did not affect sleep duration. In fact, heavier daily TV and evening smartphone use resulted in increased daytime napping. Recently, Lan et al. [ 2020] found that each additional hour spent on smartphones and tablets was independently associated with a reduced daily sleep duration of 11 and 6 min in boys and girls. In addition, compared to non-portable devices, the use of portable ones was more closely associated with short sleep duration.
## Psychosocial development
Altogether four studies explored the impact of early digital use on children’s psychosocial development, which involved psychosocial wellbeing (Zhao et al., 2018), theory of mind (Konok et al., 2021), social skill development (Gülay Ogelman et al., 2018) and emotional development (Cho and Lee, 2017; Konok et al., 2021). For example, Zhao et al. [ 2018] found that excessive screen time led to poor psychosocial wellbeing levels. In addition, BMI, sleep duration, and parent–child interaction mediated the effect of excessive screen time on children’s psychosocial wellbeing, among which parent–child interaction contributed the most. Moreover, Konok et al. [ 2021] found that mobile touch screen devices (MTSD) use was associated with a worse theory of mind. In contrast, Gülay Ogelman et al. [ 2018] revealed that mobile technologies had no predictive effect on children’s social skill levels. Regarding emotional development, Cho and Lee [2017] found that all smartphone addictive tendencies negatively affected emotional intelligence. However, Konok et al. [ 2021] did not find the impact of MTSD use on preschoolers’ emotion recognition. Although most research findings suggest that preschoolers’ digital use negatively affects their psychosocial development, more studies are needed to further examine the relationship between the two.
## Problematic behaviors
Researchers also noticed that early digital use might link with problematic behaviors, including externalizing behavior (Cho and Lee, 2017; Poulain et al., 2018; Lin et al., 2020; McDaniel and Radesky, 2020; Xie et al., 2020; Anitha et al., 2021; Lehrl et al., 2021), emotional problems (Poulain et al., 2018; Lin et al., 2020), and pervasive developmental disorder (Anitha et al., 2021). Seven of the eight studies consistently found that digital device use was associated with children’s externalizing behavior. For instance, Cho and Lee [2017] found that preschoolers’ problematic smartphone use proneness positively predicted their problematic behaviors, such as aggression, hyperactivity, and withdrawal. Laster, Poulain et al. [ 2018] confirmed that baseline mobile phone usage predicted more externalizing behavior at follow-ups, such as total difficulties, conduct problems, and hyperactivity or inattention. Next, Xie et al. [ 2020] revealed that preschoolers with screen time over 60 min per week tended to have more hyperactivity/inattention and actual difficulties. Similarly, Lin et al. [ 2020] found that preschoolers who spent more time on touchscreen devices were more likely to externalize aggressive behaviors. Only McDaniel and Radesky [2020] found that more digital use did not significantly predict later externalizing behavior. Recently, Anitha et al. [ 2021] reported that preschoolers with screen time over 2 h per day and media problematic use showed more clinically developmental disorder problems, attention deficit hyperactivity disorder related problems, and conduct problems. Similarly, Lehrl et al. [ 2021] found that preschoolers with a greater digital HLE experienced more difficulties and hyperactivity/inattention problems.
## Cognitive development
Existing studies have extensively explored the impact of early digital device use on children’s cognitive development, including attention patterns (Konok et al., 2021), visual–spatial ability & intelligence (Keefe-Cooperman, 2016), executive function (McNeill et al., 2019; Jusienė et al., 2020; Li et al., 2021), and language and literacy development (Borajy et al., 2019; Moon et al., 2019; van den Heuvel et al., 2019; Hutton et al., 2020; Lin et al., 2020). For instance, Keefe-Cooperman [2016] found that preschoolers with a greater digital use time performed worse in Wechsler Preschool and Primary Scale of Intelligence-Fourth Edition (WPPSI-IV). Specifically, Konok et al. [ 2021] found that frequent users exhibited more global precedence in selective attention tasks but atypical, local precedence in a divided attention task. However, playing with a fast-digital game eliminated the advantage of selective attention over divided attention observed in the non-digital and slow digital game conditions.
Additionally, three articles focus on the impact of early digital use on children’s executive function. For example, McNeill et al. [ 2019] also revealed that high-dose app users at baseline had a significantly lower inhibition control than working memory score at follow-up than low-dose app users. However, Jusienė et al. [ 2020] revealed that screen use did not predict different executive abilities. Specifically, TV, computer, smartphone, and tablet use were not related to inhibitory control, working memory, and mental set shifting in preschoolers from low-risk backgrounds. Nevertheless, using functional near-infrared spectroscopy technology, Li et al. [ 2021] found that compared to ‘non-digital user’, ‘heavy-digital user’ performed poorer in the Dimensional Change Card Sort task, which means poorer executive function, and showed lower activation of the prefrontal cortex (BA 9). This finding indicates that neuroimaging studies might help to understand the impact of digital overuse on early executive function.
Four of the five studies showed that digital use had a negative impact on early language and literacy development. For instance, van den Heuvel et al. [ 2019] revealed that for children who used a smartphone and tablet, each additional 30-min increase in daily smartphone and tablet use was significantly associated with increased odds of parent-reported expressive speech delay. Additionally, Moon et al. [ 2019] found that digital device usage time was negatively correlated with expressive language months in three-year-old children. Later, Hutton et al. [ 2020] found that access to a child’s smartphone and tablet was negatively correlated with the Get Ready to Read score of emergent literacy and the Comprehensive Test of Phonological Processing score of processing speed. In addition, access to a child’s smartphone and tablet was only marginally negatively correlated with the other language and literacy measures. Similarly, Lin et al. [ 2020] found that digital use was significantly and negatively associated with language development. However, the association was no longer significant when confounding variables were controlled. And a ‘null result’ was reported by Borajy et al. [ 2019], who found child’s smartphone and tablet use did not significantly influence the odds of having speech delay. This inconsistency deserves further studies.
## Modeling the relationship between preschoolers’ digital use and development
All the studies exploring children’s digital use have focused mainly on two relationships: [1] the relationship between influential factors and early digital use; and [2] the relationship between children’s digital use and developmental outcomes. Several statistical models have been developed to model these relationships, including correlation analysis, regression analysis, moderation analysis, mediation analysis, comparison of means between groups (t-test/chi-square/ANOVAs), and General Linear Mixed Models (GLMM).
## Correlation analysis
Four studies conducted correlation analysis to explore the relationship between digital use and developmental factors/outcomes. First, Keefe-Cooperman [2016] revealed a negative correlation between preschoolers’ digital use time and the WPPSI-IV and Full-Scale IQ scores. Second, Suherman et al. [ 2021] revealed a correlation between parenting style and problematic digital use among preschoolers. Third, Hutton et al. [ 2020] revealed the negative correlation between a child’s smartphone and tablet usage and emergent literacy and processing speed score. Finally, Moon et al. [ 2019] demonstrated the correlation between smart device usage and social development.
## Regression analysis
Regression analysis was the most used analysis to address “digital use as a factor” and “digital use as an outcome.” Specifically, 14 studies used regression analysis to investigate the impact of digital use on children’s outcomes. For instance, Cheung et al. [ 2017] conducted a regression analysis on the predictive power of tablet use on sleep time and delayed sleep onset. In 2018, Gülay Ogelman et al. [ 2018], Nathanson and Beyens [2018], and Poulain et al. [ 2018], and conducted regression analyses to explore its impact. In 2019, Beyens and Nathanson [2019], Borajy et al. [ 2019], McNeill et al. [ 2019], and van den Heuvel et al. [ 2019] implemented general linear modeling (GLM) to model the impact of digital use. In 2020, Jusienė et al. [ 2020], Lan et al. [ 2020], Lin et al. [ 2020], and Xie et al. [ 2020] also conducted GLM to do the regression analysis. In 2021, Velumani et al. [ 2021] and Lehrl et al. [ 2021] also implemented GLM regression analysis.
Another seven studies used regression analysis to explore the influential factors of early digital use. For instance, in 2018, Paulus et al. [ 2018] and Poulain et al. [ 2018] employed GLM to analyze the influential factors. Later, Madigan et al. [ 2020] reported the prediction of maternal screen time use on the possibility that their children’s screen time exceeds the WHO guideline. Finally, in 2021, Kim et al. [ 2021] and (Yalçin et al., 2021a,b) have modeled the influences of family characteristics, child characteristics, and screen use characteristics on early digital use. In particular, Konok et al. [ 2021] employed the General Linear Mixed Models (GLMM) to reveal that compared to non-users, frequent MTSD user preschoolers exhibit more global precedence in the selective attention tasks.
## Moderation/Mediation analysis
Six studies employed the mediation analysis to explore the mediator of the relationship concerning digital use. For instance, Cox et al. [ 2012] conducted a mediation analysis to develop a mediation model of the positive effect of weekday TV viewing on children’s BMI z-score. Later, Sijtsma et al. [ 2015], Cho and Lee [2017], and Zhao et al. [ 2018] employed mediation analysis to model the effects of different variables. In 2020, McDaniel and Radesky [2020] developed a structural equation model (SEM) to demonstrate the mediated prediction of child externalizing behavior on tablet use (not phone use). Recently, Coyne et al. [ 2021] and Lehrl et al. [ 2021] adopted SEM and multivariate regressions to develop mediation or moderation models.
## Discussion
This scoping review has provided important insights into the status quo, influential factors, outcomes, and statistical models reflected by the existing studies published during 2001–2021. This section will discuss the major findings and the boundaries of their application, as well as the implications for future research.
## Digital overuse/problematic use making preschoolers at risk
This scoping review identified the status quo of digital overuse/problematic use among preschoolers. First, preschoolers’ screen time has exceeded the World Health Organization guidelines and thus, they are at higher risk for addictive behavior. Second, the change in the type of digital devices varied with the spread of mobile device use and digital education (Madigan et al., 2020; Yalçin et al., 2021a). Last, the status of digital use among preschoolers is not satisfactory, as the existing studies demonstrated a relatively high rates of early digital overuse and problematic use, and the average percentage of the studies collected in this research were 48.34 and $26.83\%$, separately. This finding implies that our children could at risk and that swift action is needed to stop this challenging situation. In particular, the finding that boys spent more time on screens than girls and were more likely to become addicted (Xie et al., 2020) implies that boys are more vulnerable to digital overuse and need more attention, prevention, and intervention from parents and early childhood teachers.
## Parent matters
This scoping review found that parenting factors influenced preschoolers’ digital use. In particular, maternal smartphone addiction (Kim et al., 2021), parent media time (Cho and Lee, 2017; Coyne et al., 2021), parents with lower SES (Keefe-Cooperman, 2016), maternal depression (Kim et al., 2021) were positively correlated with digital overuse in preschoolers. In contrast, mother’s positive parenting style (Lehrl et al., 2021; Suherman et al., 2021; Tay et al., 2021), attitude (Özyurt et al., 2018), and self-efficacy (Baek et al., 2013) were negatively correlated with the digital overuse in preschoolers. All these findings jointly imply that parents’, especially mothers’ influences might help prevent or reduce digital overuse. Therefore, parenting programs are needed to help parents, especially mothers, understand how to cope with the challenges caused by early digital overuse/problematic use.
However, there were still some researchers holding different views toward parental factors. For example, Yalçin et al. ( 2021a) found that the mother’s employment status influenced the child’s digital use, whereas Kim et al. [ 2021] did not confirm such a relationship. In addition, Keefe-Cooperman [2016] found that children whose parents were less educated spent more time using digital devices. Moreover, Xie et al. [ 2020] revealed that children with lower-educated parents had more screen time. Interestingly, Kim et al. [ 2021] found that mothers’ digital addiction led to children’s problematic digital use instead of longer digital time. Two reasons might be associated with these mixed results. The first reason is that culture matters regarding the family structure and environment. For instance, Kim et al. [ 2021] conducted a study in Korea, where many mothers are full-time housewives. Thus, their employment status did not influence the family environment and their children’s digital use. However, Yalçin et al. ( 2021a) conducted a study in Turkey, where working mothers may influence the family environment differently. Nonetheless, future studies are needed to further explore the potential cultural influences. The second reason is relevant to the survey content, which is based on different concepts and definitions such as “digital overuse” or “problematic digital use.” Parents might have other purposes for using digital devices in preschoolers, and there might be some cultural differences. For example, Keefe-Cooperman [2016] and Xie et al. [ 2020] studied digital use for different purposes. For families with highly educated parents, their children who use digital devices may be more likely to meet the need for education, which benefits their development. In contrast, for parents with lower education, the purpose of using digital devices may be for leisure and entertainment or treat it as an e-babysitter, putting their children at risk of overusing digital devices. Therefore, exploration of the impact of digital device use on preschoolers’ development in the future may require distinguishing the specific purposes for which digital devices are used.
## Digital overuse/problematic use hurts early development
This scoping review has synthesized the empirical evidence to demonstrate the possible harm of digital overuse/problematic use on early childhood development. First, most studies showed that digital device overuse/problematic use might harm children’s physical health. For example, researchers found that early digital overuse could cause adiposity (Cox et al., 2012; Sijtsma et al., 2015; Velumani et al., 2021), physical inactivity (Cox et al., 2012; Sijtsma et al., 2015), poorer motor skill development (Moon et al., 2019), lack of sleep duration (Sijtsma et al., 2015; Cheung et al., 2017; Lan et al., 2020) and poor sleep quality (Sijtsma et al., 2015; Cheung et al., 2017; Nathanson and Beyens, 2018; Beyens and Nathanson, 2019; Lan et al., 2020). However, there are still some findings that remain controversial. Although most studies consistently found that digital overuse was positively associated with children’s obesity (BMI as an indicator; Cox et al., 2012; Sijtsma et al., 2015; Velumani et al., 2021), some researchers still hold different opinions. For example, Collings et al. [ 2018] found an insignificant correlation between time spent watching TV and BMI scores. In addition, they only found that every 1 h per day of TV viewing significantly resulted in a larger waist circumference. Given the inconsistent findings drawn from different studies, one possible explanation could be that TV, as a traditional digital device, is being used less frequently. Another possibility is that TV watching time may predict BMI by one or a combination of the following mechanisms: decreased physical activity, increased energy intake, increased sedentary behavior, and reduced sleep time. This means that there might be some mediating or moderating variables warranting further studies in the future.
However, the relationship between digital overuse/problematic use and motor skill development remains uncertain. A possible reason for this result is that in Keefe-Cooperman [2016], digital devices mainly referred to TV, while in Moon et al. [ 2019], digital devices changed to smartphones and tablet computers, which usually require the use of fingers for operation. We suggest that frequent use of the index finger could facilitate fine motor development in preschoolers. A previous study reported a similar finding that fine motor milestone achievements of toddlers (19–36 months) were associated with early touch screen scrolling (Bedford et al., 2016). Notably, the effects of digital overuse on children’s motor skill development still need more exploration due to insufficient empirical articles.
Second, the reviewed studies were concerned about the impact of early digital overuse/problematic use on children’s cognitive development, even though there were no agreements on how it affected each aspect of cognitive skills. For example, many studies found that early digital overuse negatively influenced children’s cognitive development, especially their language and literacy (Moon et al., 2019; van den Heuvel et al., 2019; Hutton et al., 2020). However, Lin et al. [ 2020] and Borajy et al. [ 2019] did not find the impact of digital use on children’s language and the odds of having speech delay. Another argument is surrounding the impact on executive function. For example, both Li et al. [ 2021] and McNeill et al. [ 2019] found that digital use had a negative influence on children’s executive function, whereas Jusienė et al. [ 2020] did not. The different tools and experimental tasks might cause these inconsistencies. Thus, further studies are also needed to settle this argument.
Third, the reviewed studies indicated that digital use for non-education purposes could cause a range of behavioral problems, including conduct problems (Poulain et al., 2018; Lehrl et al., 2021), hyperactivity/inattention (Poulain et al., 2018; Anitha et al., 2021; Lehrl et al., 2021), aggressive behaviors (Lin et al., 2020), emotional problems (Poulain et al., 2018; Lin et al., 2020), and even pervasive developmental disorder (Anitha et al., 2021). Two important factors might help prevent behavioral problems: the purpose of using digital products and the content of children’s viewing of digital products. Previous studies have demonstrated that using digital devices for educational purposes rather than entertainment would reduce behavioral problems (Fang et al., 2022). Similarly, choosing educational content rather than purely entertainment content could promote children’s imitation and learning of positive behaviors (such as prosocial behavior) rather than problematic behaviors. Therefore, using digital products for educational purposes rather than entertainment purposes might be better in reducing the negative impact of digital products on children’s behavior.
Fourth, the reviewed studies have revealed the impact of digital overuse/problematic use on children’s social–emotional development, with contradictory findings caused by using different aspects as the dependent variable. For example, Cho and Lee [2017] found that the increment in digital use reduced children’s emotional intelligence, making them hard to identify their own and others’ emotional states. In contrast, Konok et al. [ 2021] focused on emotional recognition and found that touchscreen use did not affect children’s performance in emotional recognition. Nevertheless, three other studies (Gülay Ogelman et al., 2018; Zhao et al., 2018; Konok et al., 2021) jointly demonstrated a great impact of early digital use on early psychosocial development and suggested that excessive digital use led to a poor level of psychosocial wellbeing, theory of mind and social skills. Therefore, we tend to conclude that the existing evidence demonstrates the negative impact of digital overuse/problematic use on early social–emotional development.
## Early digital overuse/problematic use: The models and measures
This study has reviewed the empirical studies on this topic and found various statistical models adopted in different studies, including correlation analysis, regression analysis, moderation, and mediation analysis. However, the current literature demonstrated several disadvantages regarding the statistic models. First, most of these studies adopted general linear modeling, while few adopted experimental designs. Moreover, the linear regression analysis could only determine different levels of correlation among variables, failing to confirm causality. Therefore, more empirical studies with longitudinal design should be used to explore the causal relationship among possible factors, children’s digital overuse/problematic use, and developmental outcomes in the future. Second, children’s digital overuse/problematic use is inseparable from their family, school, and living community, in which the data are inevitably nested. In this sense, a simple linear model might not be able to depict the whole picture. Therefore, the features and nature of the data should be considered in future studies. Third, the existing studies on measuring children’s digital overuse/problematic use focused on a simple aspect, such as the length of time or screen time. Future studies should consider developing a comprehensive instrument to capture the whole picture of digital use in preschoolers. Last but not least, although a certain number of studies exploring cognitive development, only one study addressed this question using a neuroimaging approach. Therefore, future studies should include behavioral and neuroscientific approaches to advance our understanding of the effects of early digital use on children’s cognitive development.
## Culture and COVID-19: Two important but overlooked factors
This scoping review indicated that two critical factors had been neglected or understudied. First, very few studies have explored the cultural influences or between-culture differences. Although early digital overuse is a global trend, the use and development of information technology in different countries are not synchronized. In addition, parents from different cultural backgrounds may hold different views on digital use and family education. For example, Eastern parents treat children’s education more strictly, while Western education is more individualized and child-centered. There is a need to explore whether culture, particularly *Western versus* Eastern, in itself and interaction with other factors, influences young people’s digital use. Second, the COVID-19 pandemic has thoroughly changed our lifestyle, forcing us to study or work from home. As a result, parents and their children must use digital devices for work, study, and entertainment. This has caused a sharp increase in screen time for children and their parents. However, very few studies have explored digital overuse/problematic use during the COVID-19 lockdowns and its impact on preschoolers’ longitudinal development. This research gap needs to be filled as soon as possible.
## Conclusions, limitations, and implications
As the first scoping review on digital overuse/problematic use in preschoolers, this study has synthesized the empirical evidence during the past 20 years to identify its status quo, influential factors, consequences, and models. First, the average percentage of overuse and problematic use across the studies collected in this research were 48.34 and $26.83\%$, separately. In particular, boys were more vulnerable to digital overuse, and parents, especially mothers, were helpful in reducing or preventing it. However, early digital overuse/problematic use for non-educational purposes may hurt preschoolers’ physical and psychosocial health and cognitive development and lead to problematic behaviors. The common methods and models adopted in this type of studies are also reviewed. In addition, this study also found a significant research gap: lacking longitudinal, neuroimaging, and multidisciplinary studies.
Nevertheless, this study has two limitations. First, it has only searched the three common full-text databases: ProQuest, Web of Science, and Google Scholar. Although very inclusive and comprehensive, the three databases might not include all the relevant studies. Other databases, such as EBSCO, JSTOR, SCOPUS, and ERIC, should also be included in the future. Second, this scoping review has focused on the English articles identified from international peer-reviewed journals. Those journals in Chinese or other languages were not included in this scoping review. Future scoping reviews should consider the language issues and include important and highly relevant journals in other important languages such as Chinese, French, and Spanish.
## 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
CW and HQ contributed to data collection, processing and analysis, and original manuscript drafting. DW contributed to project conceptualization, data collection and analysis, original manuscript drafting, and supervision. HL contributed to constructive discussions and manuscript revision. All authors contributed to the article and approved the submitted version.
## Funding
This study was funded by the National Natural Science Foundation of China (Ref No. 62277037) and the Start-up Research Grant at the Education University of Hong Kong (Ref. No. RG $\frac{48}{2021}$-2022R).
## 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. **American Academy of Pediatrics: children, adolescents, and television**. *Pediatr* (2001) **107** 423-426. DOI: 10.1542/peds.107.2.423
2. Anitha F. S., Narasimhan U., Janakiraman A., Janakarajan N., Tamilselvan P.. **Association of digital media exposure and addiction with child development and behavior: a cross-sectional study**. *Ind. Psychiatry J.* (2021) **30** 265-271. DOI: 10.4103/ipj.ipj_157_20
3. Arksey H., O'Malley L.. **Scoping studies: towards a methodological framework**. *Int. J. Soc. Res. Methodol.* (2005) **8** 19-32. DOI: 10.1080/1364557032000119616
4. Baek Y. -M., Lee J. -M., Kim K. S.. **A study on smart phone use condition of infants and toddlers**. *Int. J. Smart Home* (2013) **7** 123-132. DOI: 10.14257/ijsh.2013.7.6.12
5. Bedford R., Saez de Urabain I. R., Cheung C. H., Karmiloff-Smith A., Smith T. J.. **Toddlers’ fine motor milestone achievement is associated with early touchscreen scrolling**. *Front. Psychol.* (2016) **7** 1108. DOI: 10.3389/fpsyg.2016.01108
6. Beyens I., Nathanson A. I.. **Electronic media use and sleep among preschoolers: evidence for time-shifted and less consolidated sleep**. *Health Commun.* (2019) **34** 537-544. DOI: 10.1080/10410236.2017.1422102
7. Borajy S., Albkhari D., Turkistani H., Altuwairiqi R., Aboalshamat K., Altaib T.. **Relationship of electronic device usage with obesity and speech delay in children**. *Fam. Med. Prim. Care Rev.* (2019) **21** 93-97. DOI: 10.5114/fmpcr.2019.84542
8. Bronfenbrenner U.. **Toward an experimental ecology of human development**. *Am. Psychol.* (1977) **32** 513-531. DOI: 10.1037/0003-066X.32.7.513
9. Bronfenbrenner U.. *The Ecology of Human Development: Experiments by Nature and Design* (1979)
10. Chang H. Y., Park E.-J., Yoo H.-J., won Lee J., Shin Y.. **Electronic media exposure and use among toddlers**. *Psychiatry Investig.* (2018) **15** 568-573. DOI: 10.30773/pi.2017.11.30.2
11. Cheung C. H., Bedford R., Saez De Urabain I. R., Karmiloff-Smith A., Smith T. J.. **Daily touchscreen use in infants and toddlers is associated with reduced sleep and delayed sleep onset**. *Sci. Rep.* (2017) **7** 1-7. DOI: 10.1038/srep46104
12. Cho K.-S., Lee J.-M.. **Influence of smartphone addiction proneness of young children on problematic behaviors and emotional intelligence: mediating self-assessment effects of parents using smartphones**. *Comput. Hum. Behav.* (2017) **66** 303-311. DOI: 10.1016/j.chb.2016.09.063
13. Collings P. J., Kelly B., West J., Wright J.. **Associations of TV viewing duration, meals and snacks eaten when watching TV, and a TV in the bedroom with child adiposity**. *Obesity* (2018) **26** 1619-1628. DOI: 10.1002/oby.22288
14. Cox R., Skouteris H., Rutherford L., Fuller-Tyszkiewicz M., Dell'Aquila D., Hardy L. L.. **Television viewing, television content, food intake, physical activity and body mass index: a cross-sectional study of preschool children aged 2–6 years**. *Health Promot. J. Austr.* (2012) **23** 58-62. DOI: 10.1071/HE12058
15. Coyne S. M., Shawcroft J., Gale M., Gentile D. A., Etherington J. T., Holmgren H.. **Tantrums, toddlers and technology: temperament, media emotion regulation, and problematic media use in early childhood**. *Comput. Hum. Behav.* (2021) **120** 106762. DOI: 10.1016/j.chb.2021.106762
16. Domoff S. E., Borgen A. L., Radesky J. S.. **Interactional theory of childhood problematic media use**. *Hum. Behav. Emerg. Tech.* (2020) **2** 343-353. DOI: 10.1002/hbe2.217
17. Dong C., Cao S., Li H.. **Young children’s online learning during COVID-19 pandemic: Chinese parents’ beliefs and attitudes**. *Child Youth Serv. Rev.* (2020) **118** 105440. DOI: 10.1016/j.childyouth.2020.105440
18. Dong C., Mertala P.. **Preservice teachers’ beliefs about young children’s technology use at home**. *Teach. Teach. Educ.* (2021) **102** 103325. DOI: 10.1016/j.tate.2021.103325
19. Fang M., Tapalova O., Zhiyenbayeva N., Kozlovskaya S.. **Impact of digital game-based learning on the social competence and behavior of preschoolers**. *Educ. Inf. Technol.* (2022) **27** 3065-3078. DOI: 10.1007/s10639-021-10737-3
20. Gottlieb G.. **Probabilistic epigenesis**. *Dev. Sci.* (2007) **10** 1-11. DOI: 10.1111/j.1467-7687.2007.00556.x
21. Gülay Ogelman H., Güngör H., Körükçü Ö., Erten Sarkaya H.. **Examination of the relationship between technology use of 5–6 year-old children and their social skills and social status**. *Early Child Dev. Care* (2018) **188** 168-182. DOI: 10.1080/03004430.2016.1208190
22. Hutton J. S., Huang G., Sahay R. D., DeWitt T., Ittenbach R. F.. **A novel, composite measure of screen-based media use in young children (screen Q) and associations with parenting practices and cognitive abilities**. *Pediatr. Res.* (2020) **87** 1211-1218. DOI: 10.1038/s41390-020-0765-1
23. Jimenez M. E., Wade R., Lin Y., Morrow L. M., Reichman N. E.. **Adverse experiences in early childhood and kindergarten outcomes**. *Pediatrics* (2016) **137** e20151839. DOI: 10.1542/peds.2015-1839
24. Jusienė R., Rakickienė L., Breidokienė R., Laurinaitytė I.. **Executive function and screen-based media use in preschool children**. *Infant. Child Dev.* (2020) **29** e2173. DOI: 10.1002/icd.2173
25. Keefe-Cooperman K.. **Digital media and preschoolers: implications for visual spatial development**. *Res. Prac. J. Early Childhood Field* (2016) **19** 24-42
26. Kim B., Han S., Park E.-J., Yoo H., Suh S., Shin Y.. **The relationship between mother’s smartphone addiction and children’s smartphone usage**. *Psychiatry Investig.* (2021) **18** 126-131. DOI: 10.30773/pi.2020.0338
27. Konok V., Liszkai-Peres K., Bunford N., Ferdinandy B., Jurányi Z., Ujfalussy D. J.. **Mobile use induces local attentional precedence and is associated with limited socio-cognitive skills in preschoolers**. *Comput. Hum. Behav.* (2021) **120** 106758. DOI: 10.1016/j.chb.2021.106758
28. Lam L. T.. **Parental mental health and internet addiction in adolescents**. *Addict. Behav.* (2015) **42** 20-23. DOI: 10.1016/j.addbeh.2014.10.033
29. Lan Q.-Y., Chan K. C., Kwan N. Y., Chan N. Y., Wing Y. K., Li A. M.. **Sleep duration in preschool children and impact of screen time**. *Sleep Med.* (2020) **76** 48-54. DOI: 10.1016/j.sleep.2020.09.024
30. Lauricella A. R., Wartella E., Rideout V. J.. **Young children's screen time: the complex role of parent and child factors**. *J. Appl. Dev. Psychol.* (2015) **36** 11-17. DOI: 10.1016/j.appdev.2014.12.001
31. Lehrl S., Linberg A., Niklas F., Kuger S.. **The home learning environment in the digital age—associations between self-reported “analog” and “digital” home learning environment and Children’s socio-emotional and academic outcomes**. *Front. Psychol.* (2021) **12** 592513. DOI: 10.3389/fpsyg.2021.592513
32. Levac D., Colquhoun H., O'Brien K. K.. **Scoping studies: advancing the methodology**. *Implement. Sci.* (2010) **5** 1-9. DOI: 10.1186/1748-5908-5-69
33. Levine L. E., Waite B. M., Bowman L. L., Kachinsky K.. **Mobile media use by infants and toddlers**. *Comput. Hum. Behav.* (2019) **94** 92-99. DOI: 10.1016/j.chb.2018.12.045
34. Li H., Wu D., Yang J., Luo J., Xie S., Chang C.. **Tablet use affects preschoolers’ executive function: fNIRS evidence from the dimensional change card sort task**. *Brain Sci.* (2021) **11** 567. DOI: 10.3390/brainsci11050567
35. Lin H.-P., Chen K.-L., Chou W., Yuan K.-S., Yen S.-Y., Chen Y.-S.. **Prolonged touch screen device usage is associated with emotional and behavioral problems, but not language delay, in toddlers**. *Infant Behav. Dev.* (2020) **58** 101424. DOI: 10.1016/j.infbeh.2020.101424
36. Madigan S., Racine N., Tough S.. **Prevalence of preschoolers meeting vs exceeding screen time guidelines**. *JAMA Pediatr.* (2020) **174** 93-95. DOI: 10.1001/jamapediatrics.2019.4495
37. Mallawaarachchi S. R., Anglim J., Hooley M., Horwood S.. **Associations of smartphone and tablet use in early childhood with psychosocial, cognitive and sleep factors: a systematic review and meta-analysis**. *Early Child Res. Q.* (2022) **60** 13-33. DOI: 10.1016/j.ecresq.2021.12.008
38. Masten A. S., Cicchetti D.. **Developmental cascades**. *Dev. Psychopathol.* (2010) **22** 491-495. DOI: 10.1017/S0954579410000222
39. McDaniel B. T., Radesky J. S.. **Longitudinal associations between early childhood externalizing behavior, parenting stress, and child media use**. *Cyberpsychol. Behav. Soc. Netw.* (2020) **23** 384-391. DOI: 10.1089/cyber.2019.0478
40. McNeill J., Howard S. J., Vella S. A., Cliff D. P.. **Longitudinal associations of electronic application use and media program viewing with cognitive and psychosocial development in preschoolers**. *Acad. Pediatr.* (2019) **19** 520-528. DOI: 10.1016/j.acap.2019.02.010
41. Moon J. H., Cho S. Y., Lim S. M., Roh J. H., Koh M. S., Kim Y. J.. **Smart device usage in early childhood is differentially associated with fine motor and language development**. *Acta Paediatr.* (2019) **108** 903-910. DOI: 10.1111/apa.14623
42. Nathanson A. I., Beyens I.. **The relation between use of mobile electronic devices and bedtime resistance, sleep duration, and daytime sleepiness among preschoolers**. *Behav. Sleep Med.* (2018) **16** 202-219. DOI: 10.1080/15402002.2016.1188389
43. Nikken P., Schols M.. **How and why parents guide the media use of young children**. *J. Child Fam. Stud.* (2015) **24** 3423-3435. DOI: 10.1007/s10826-015-0144-4
44. Özyurt G., Dinsever Ç., Çalişkan Z., Evgin D.. **Effects of triple P on digital technological device use in preschool children**. *J. Child Fam. Stud.* (2018) **27** 280-289. DOI: 10.1007/s10826-017-0882-6
45. Paulus F. W., Sinzig J., Mayer H., Weber M., von Gontard A.. **Computer gaming disorder and ADHD in young children—a population-based study**. *Int. J. Ment. Health Addict* (2018) **16** 1193-1207. DOI: 10.1007/s11469-017-9841-0
46. Plowman L., Stephen C.. **A ‘benign addition’? Research on ICT and pre-school children**. *J. Comput. Assist. Learn.* (2003) **19** 149-164. DOI: 10.1046/j.0266-4909.2003.00016.x
47. Poulain T., Vogel M., Neef M., Abicht F., Hilbert A., Genuneit J.. **Reciprocal associations between electronic media use and behavioral difficulties in preschoolers**. *Int. J. Environ. Res. Public Health* (2018) **15** 814. DOI: 10.3390/ijerph15040814
48. Radesky J. S., Christakis D. A.. **Increased screen time: implications for early childhood development and behavior**. *Pediatr. Clin.* (2016) **63** 827-839. DOI: 10.1016/j.pcl.2016.06.006
49. Radesky J. S., Eisenberg S., Kistin C. J., Gross J., Block G., Zuckerman B.. **Overstimulated consumers or next-generation learners? Parent tensions about child mobile technology use**. *Ann. Fam. Med.* (2016) **14** 503-508. DOI: 10.1370/afm.1976
50. Rideout V., Robb M. B.. *The Common Sense census: Media use by kids age zero to eight, 2020. San Francisco, CA: Common Sense Media* (2020)
51. Rocha B., Nunes C.. **Benefits and damages of the use of touchscreen devices for the development and behavior of children under 5 years old—a systematic review**. *Psicol. Reflex Crit.* (2020) **33** 24. DOI: 10.1186/s41155-020-00163-8
52. Sijtsma A., Koller M., Sauer P. J., Corpeleijn E.. **Television, sleep, outdoor play and BMI in young children: the GECKO Drenthe cohort**. *Eur. J. Pediatr.* (2015) **174** 631-639. DOI: 10.1007/s00431-014-2443-y
53. Suherman R. N., Saidah Q., Nurhayati C., Susanto T., Huda N.. **The relationship between parenting style and gadget addiction among preschoolers**. *Malays. J. Med. Health Sci.* (2021) **17** 117-122
54. Tay L. Y., Aiyoob T. B., Chua T. B. K., Ramachandran K., Chia M. Y. H.. **Pre-schoolers’ use of technology and digital media in Singapore: entertainment indulgence and/or learning engagement?**. *Educ. Media Int.* (2021) **58** 1-20. DOI: 10.1080/09523987.2021.1908498
55. van den Heuvel M., Ma J., Borkhoff C. M., Koroshegyi C., Dai D. W., Parkin P. C.. **Mobile media device use is associated with expressive language delay in 18-month-old children**. *J. Dev. Behav. Pediatr.* (2019) **40** 99-104. DOI: 10.1097/DBP.0000000000000630
56. Velumani S., Panchal M., Patel B.. **Screen dependency versus child nourishment among toddlers: a correlational study**. *Indian J. Psy. Nurs.* (2021) **18** 100-106. DOI: 10.4103/iopn.iopn_32_21
57. WHO. (2019). New WHO guidance: Very limited daily screen time recommended for children under 5. Available at: https://www.aoa.org/news/clinical-eye-care/public-health/screen-time-for-children-under-5?sso=y. (2019)
58. Xie G., Deng Q., Cao J., Chang Q.. **Digital screen time and its effect on preschoolers’ behavior in China: results from a cross-sectional study**. *Ital. J. Pediatr.* (2020) **46** 9-7. DOI: 10.1186/s13052-020-0776-x
59. Yalçin S. S., Çaylan N., Erat Nergiz M., Oflu A., Yıldız D., Tezol Ö.. **Video game playing among preschoolers: prevalence and home environment in three provinces from Turkey**. *Int. J. Environ. Health Res.* (2021b) **32** 2233-2246. DOI: 10.1080/09603123.2021.1950653
60. Yalçin S. S., Tezol Ö., Çaylan N., Erat Nergiz M., Yildiz D., Çiçek Ş.. **Evaluation of problematic screen exposure in pre-schoolers using a unique tool called “seven-in-seven screen exposure questionnaire”: cross-sectional study**. *BMC Pediatr.* (2021a) **21** 472-411. DOI: 10.1186/s12887-021-02939-y
61. Yelland N.. **New technologies and young children: technology in early childhood education. Teach learn**. *Network* (2006) **13** 10-13. DOI: 10.3316/aeipt.155999
62. Zhao J., Zhang Y., Jiang F., Ip P., Ho F. K. W., Zhang Y.. **Excessive screen time and psychosocial well-being: the mediating role of body mass index, sleep duration, and parent-child interaction**. *J. Pediatr.* (2018) **202** 157-162.e1. DOI: 10.1016/j.jpeds.2018.06.029
|
---
title: Uptake of severe acute respiratory syndrome coronavirus 2 spike protein mediated
by angiotensin converting enzyme 2 and ganglioside in human cerebrovascular cells
authors:
- Conor McQuaid
- Alexander Solorzano
- Ian Dickerson
- Rashid Deane
journal: Frontiers in Neuroscience
year: 2023
pmcid: PMC9980911
doi: 10.3389/fnins.2023.1117845
license: CC BY 4.0
---
# Uptake of severe acute respiratory syndrome coronavirus 2 spike protein mediated by angiotensin converting enzyme 2 and ganglioside in human cerebrovascular cells
## Abstract
### Introduction
There is clinical evidence of neurological manifestations in coronavirus disease-19 (COVID-19). However, it is unclear whether differences in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)/spike protein (SP) uptake by cells of the cerebrovasculature contribute to significant viral uptake to cause these symptoms.
### Methods
Since the initial step in viral invasion is binding/uptake, we used fluorescently labeled wild type and mutant SARS-CoV-2/SP to study this process. Three cerebrovascular cell types were used (endothelial cells, pericytes, and vascular smooth muscle cells), in vitro.
### Results
There was differential SARS-CoV-2/SP uptake by these cell types. Endothelial cells had the least uptake, which may limit SARS-CoV-2 uptake into brain from blood. Uptake was time and concentration dependent, and mediated by angiotensin converting enzyme 2 receptor (ACE2), and ganglioside (mono-sialotetrahexasylganglioside, GM1) that is predominantly expressed in the central nervous system and the cerebrovasculature. SARS-CoV-2/SPs with mutation sites, N501Y, E484K, and D614G, as seen in variants of interest, were also differentially taken up by these cell types. There was greater uptake compared to that of the wild type SARS-CoV-2/SP, but neutralization with anti-ACE2 or anti-GM1 antibodies was less effective.
### Conclusion
The data suggested that in addition to ACE2, gangliosides are also an important entry point of SARS-CoV-2/SP into these cells. Since SARS-CoV-2/SP binding/uptake is the initial step in the viral penetration into cells, a longer exposure and higher titer are required for significant uptake into the normal brain. Gangliosides, including GM1, could be an additional potential SARS-CoV-2 and therapeutic target at the cerebrovasculature.
## Introduction
While severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) primarily elicits respiratory infectious coronavirus disease-19 (COVID-19) (Bchetnia et al., 2020), many non-respiratory organs are also affected, including the brain (Huang et al., 2020; Mao L. et al., 2020; Moriguchi et al., 2020; Saleki et al., 2020; Brady et al., 2021; McQuaid et al., 2021), heart (Dhakal et al., 2020; Puntmann et al., 2020; Perez-Bermejo et al., 2021), kidneys (Fanelli et al., 2020; Martinez-Rojas et al., 2020), and liver (Mao R. et al., 2020; Wang et al., 2020; Zhong et al., 2020; Marjot et al., 2021). This may suggest that there is also viral distribution from the blood during the pathogenesis of this disease. There is continuous evolution of SARS-CoV-2 variants that affect people of all ages, which may lead to short- and long-term symptoms, including neurological manifestations. Several cell types associated with the neurovasculature are likely to interact with SARS-CoV-2, which may modulate/restrict its entry into the parenchyma. Thus, three cell types of the human cerebrovasculature (endothelium, pericytes, and smooth muscle cells) were used to determine if there were differential mechanisms of SARS-CoV-2 uptake. The data will provide a mechanistic basis for further studies, and hopefully contribute to the development of targeted therapeutic approaches, especially for “long COVID-19.” The spike protein (SP) of SARS-CoV-2 is a structural protein, which is assemble as a trimer of the heterodimer (S1-S2), that protrudes from the viral surface to give it the crown-like appearance (Gordon et al., 2020; Li et al., 2020). The S1 unit contains a receptor binding domain (RBD), which promotes attachment to host cells via facilitators, such as to the extracellular peptidase domain on angiotensin converting enzyme 2 receptor (ACE2), the main receptor for SARS-CoV-2 (Hamming et al., 2004; Doobay et al., 2007; Hoffmann et al., 2020; Perrotta et al., 2020; Ma et al., 2021). TMPRESS 2 (transmembrane protease, serine 2) on the host cells cleaves the SP to promote viral entry into cells (Matsuyama et al., 2020). There are other receptors/facilitators on the cell surface that mediate the entry of SARS-CoV-2, including oligosaccharide receptors via sialic acid (Chen et al., 2005; Tortorici et al., 2019; Erickson et al., 2021; Rhea et al., 2021). Thus, there are multiple interaction sites between SARS-CoV-2 and host cells, which may contribute to cell type specific effects and the diverse symptoms of COVID-19 (McQuaid et al., 2021).
SARS-CoV-2 has been detected in brains of severely infected deceased people, however, it is unclear as to how it gets there and if this leads to significant viral neuro-invasion (McQuaid et al., 2021). Recombinant spike proteins have been used to study viral behavior by using in vitro models of brain endothelial cells and in vivo studies (Buzhdygan et al., 2020; Brady et al., 2021; Rhea et al., 2021). While it was reported that SP interacts with the brain endothelial cells, this was independent of ACE2, in mice (Rhea et al., 2021). The entry of SARS-CoV-2 into brain would be influenced by its interaction with several cell types, including the endothelium at the interface between blood and brain, and pericytes and smooth muscles cells. However, SARS-CoV-2/SP uptake by these cell types has not been fully characterized. Thus, this study was necessary to provide fundamental basic scientific data on SARS-CoV-2 interaction at the cerebrovasculature so as to provide a better understanding to the field and for further studied.
Herein, we used fluorescently labeled SP of wild type (WT) and mutants (from variants of concern) to establish the mechanism of SARS-CoV-2/SP uptake by human cerebrovascular cells (endothelial cells, pericytes and smooth muscle cells). We show that there was differential SARS-CoV-2/SP uptake by these three cell types, with the endothelial cells showing the lowest capacity for the uptake, which will limit entry into the parenchyma. SARS-CoV-2/SP uptake was mediated by ACE2 and a ganglioside (mono-sialotetrahexasylganglioside (GM1). SARS-CoV-2/SPs with mutation sites N501Y and E484K and D614G, showed a higher uptake compared to control wild type SARS-CoV-2/SP, except for D614G in pericytes. The striking differences for these mutants were greater binding to sialic acid via wheat germ agglutinin (WGA) and neutralization of the mutant uptake was less effective than that of the wild type SARS-CoV-2/SP using anti-ACE2 or anti-GM1 antibodies. The added value of our findings to the existing evidence is to provide data showing that GM1, which is expressed at the cerebrovasculature, is likely also an entry point for SARS-CoV-2 into these human cells. This also brings SARS-CoV-2 close to ACE2, which likely facilitates its entry into host cells, since both ACE2 and gangliosides are mainly located in the lipid raft/caveolin. Also, our data have shown that even though SARS-CoV-2 uptake mechanisms are similar, there are subtle differences, which could contribute the differences in the infection outcomes.
## Materials
SARS-CoV-2 Spike proteins [recombinant SARS-CoV-2 Spike Protein (SP-RBD, Arg319-Phe 541; cat# RP-87678, HEK293 cell expressed and binds ACE2] was obtained from Life Technologies Corporation, Carlsbad, CA, USA. Mutants SPs and its control wild type protein were obtained from RayBiotech Inc., Peachtree Corners, GA, USA. Recombinant mutants N501Y (cat# 230-30184, expressed region Arg319-Phe541), D614G (Cat# 230-3030186, expressed region Arg319-Gln690), E484K (cat# 230-30188, expressed region Arg319-Phe541) and their control wild type SP (Cat# 230-30162, expressed region Arg319-Phe541) were also HEK 293 expressed. All SPs were labeled separately with Alexa Fluor 555, using a kit (Microscale protein labeling kit; ThermoFisher Scientific; Waltham, MA, USA) and by following the manufacturer instructions. Anti-ACE2 antibody (R&D Systems, Cat# AF933) was labeled with Alexa Fluor 488 by following the manufacturer instructions (Microscale protein labeling kit; ThermoFisher Scientific). In addition, the labeled SPs or anti-ACE2 antibody were purified using 3 kDa molecular weight cut-off ultrafiltration filter (Amicon Ultra Centrifugal Filter, Millipore). There was no detectable dye in the filtrate.
Antibodies raised against the extracellular domain of potential SP binding receptors were used. These include, anti-ACE2 antibody (R&D Systems, Cat# AF933) used at a low (10 μg/ml) and high (60 μg/ml) concentration; anti-TMPRSSE2 antibody (Invitrogen, Cat# PA5-14264) used at 13 μg/ml; anti-CD147 antibody (Invitrogen, Cat# 34-5600) used at 2.5 μg/ml; anti-NP-1 antibody (Invitrogen, Cat# PA5-47027) used at 2 μg/ml and anti-GM1 antibody (Abcam, Cat# Ab23943) used at 5 μg/ml. The concentration used was obtained from the manufacturer guidance. Wheat Germ agglutinin (WGA; Cat# L9640) and heparin (cat# H3393) were obtained from Sigma, and used at 10 and 100 μg/mL, respectively. Transferrin from human serum conjugated to Alexa Fluor 488 (Cat# T13342), BODIPY FL C5-Lactosylceramide complex to BSA (Cat# B34402) and BODIPY FL ganglioside (Cat# B13950) were obtained from ThermoFisher Scientific) and used at 10 μg/ml. Nystatin (Cat# J62486.09), and chlorpromazine (Cat# J63659) were obtained from ThermoFisher Scientific, and used at 25 and 10 μg/ml, respectively. Angiotensin II (cat# $\frac{1158}{5}$) was obtained from R&D Systems and used at 0.1 μg/mL.
## Cell culture
Three cell types were selected to focus more on the vascular mechanisms of SARS-CoV-2/SP uptake since they are also present in the vasculature of other organs. Human Cerebral Microvascular Endothelial Cells (hCMEC/D3) were purchased from Millipore (#SCC066) and expanded in EBM-2 Endothelial Cell Growth Basal Medium (Lonza #00190860) with EGM-2 MV* Microvascular Endothelial Singlequot kit (Lonza #CC-4147) supplemented media. hCMEC/D3 cells were expanded in T25 flask (ThermoFisher Scientific #156367) on a collagen IV (50 μg/ml Sigma-Aldrich #122-20) based growth matrix. * hCMEC/D3 cells were cultured in modified EBM-2 MV medium (Lonza) containing (v/v) $0.025\%$ VEGF, IGF and EGF, $0.1\%$ bFGF, $0.1\%$ rhFGF, $0.1\%$ gentamycin, $0.1\%$ ascorbic acid, $0.04\%$ hydrocortisone, and $1\%$ 100 U/ml penicillin, and 100 μg/ml streptomycin. When hCMEC/D3 cells were grown on glass slides (ibidi u-chamber 12 well glass slides #81201) for experiments, collagen and fibronectin (50 μg/ml Millipore FC014) were used as matrix. Human Brain Vascular Smooth Muscle Cells (HBVSMCs) were purchased from Sciencell Research Laboratories, Inc., Carlsbad, CA, USA (Cat#1100) and expanded in Smooth Muscle Cell Medium (Sciencell #1101) by following manufacturer’s guidelines. HBVSMCs cell were expanded in T25 flasks with a Poly-L-Lysine [PLL ($0.01\%$ Sigma-Aldrich #P4707)] based growth matrix. When HBVSMCs were grown on glass slides (ibidi u-chamber 12 well glass slides) for experiments, fibronectin (50 μg/ml Millipore FC014) was used as matrix. Human Brain Vascular Pericytes (HBVPs) were purchased from Sciencell (#1200) and expanded in Human Pericyte Cell Medium (Sciencell #1201) by following manufacturer’s guidelines. HBVP cell were expanded in T25 flasks with a PLL ($0.01\%$ Sigma-Aldrich #P4707) based growth matrix. When HBVP were grown on glass slides (ibidi u-chamber 12 well glass slides) for experiments, fibronectin (50 μg/ml Millipore FC014) was used as matrix. All cell types were split at $90\%$ confluency into fresh matrix coated flasks or slides for growth and experiments. Medium was changed within 24 h of initial split and every 2–3 days thereafter. Cells were kept in an incubator (37°C, $5\%$ CO2 humidified) during growth and experiments. Studies at 4°C were performed in a fridge.
## Uptake experiments
Cells were grown to confluent on glass-wells slides (ibidi u-chamber 12-well glass slides), media removed, cells washed 3x with Hank’s Balance Salt Solution containing calcium and magnesium HBSS + Ca/Mg, glucose and bicarbonate [Gibco, Waltham, MA, USA [14025-092]] before exposure to SP diluted in HBSS + Ca/Mg at a given concentration (usually 100 nM) and kept in an incubator (37°C, $5\%$ CO2 humidified) for the duration of the experiment. At the end of the experiment, cells were washed with HBSS + Ca/Mg, fixed in $4\%$ paraformaldehyde (PFA) for 10 min and mounted with ProLong™ Glass Antifade Mountant with NucBlue™ Stain (ThermoFisher Scientific P36985).
## Inhibition studies
Cells were pre-incubate with the inhibitor (diluted in HBSS) for 15 min and in the present of SP (100 nM) for 4 h at 37°C and $5\%$ CO2 in the humidified incubator. All cells were then washed, fixed, mounted, and imaged. Values were calculated as percentage of controls, which were SP uptake without the inhibitor but with the vehicle solution.
## Low temperature studies
Cells were pre-incubate in the fridge (4°C) for 30 min to adapt the cells to this temperature before running the experiment for 1 h in the fridge in the presence of SP-555 (100 nM). For comparison, cells were incubated in the incubator for 30 min followed by 1 h in the presence of SP-555 (100 nM). All cells were then washed, fixed, mounted and imaged. The person performing the experiments was blinded to the type of tracer used in the experiments.
## Immunocytochemistry (ICC) and imaging
Cells were grown to confluency on glass slides, medium removed, washed with HBSS + Ca/Mg and fixed in $4\%$ PFA for 15 min. Cells were not permeabilized. The same primary antibodies used for inhibition assay were used in the ICC and at the same concentration. The secondary antibodies, which were conjugated to Alexa Fluor 488, were donkey anti-rabbit (Thermofisher Scientific #A32790), anti-goat (Thermofisher Scientific #A32814) and anti-mouse (Thermofisher Scientific #A32766), and used at 1:200 dilution. Cells were mounted using with ProLong™ Glass Antifade Mountant with NucBlue™ Stain (ThermoFisher Scientific P36985) and imaged on Axiovert 25 Zeiss Inverted microscope (Obrekochen, Germany).
## ACE2 binding to SP
Recombinant human ACE2 (HEC 293 cells; cat# 230-30165), RayBiotech Life, 663276971Inc663276971RCRevathi Chandrasekar663276971-944626917Please note that the bracket is missing in some occurrences in the article, hence either removed or included based on the sentence formation. Kindly check and correct if necessary. ( Peachtree Corners, GA), dissolved in carbonate/bicarbonate buffer, was immobilized (2 μg/ml) on glass slides for 1 h at room temperature (RT), blocked with a non-protein buffer (Pierce Blocking buffer), washed, incubated with SP-555 at different concentrations in HBSS for 1 h at RT, washed, mounted, and imaged. SP-555 intensity at each concentration were analyzed and expressed as intensity/μm2. Values are mean ± SEM.
## Imaging and analysis
Whole well images were taken on an Olympus VS-120 slide scanner. Exposure and gain were fixed for all experimental groups based on pilot experiments. All fluorescence (SP-555 and DAPI) quantification was performed without enhancement of signals. The person performing the analysis were blinded to the experimental design and tracer used. The VSI files were imported into Qupath for analysis. Five square fields with an area of 1 × 10–6 μm were randomly placed on each imaged. Custom pixel classifiers were created to measure the fluorescence intensity in the red (555 nm) channel. The settings for the pixel classifiers were standardized to be: random trees, moderate pixel resolution, and local mean subtraction set at 1. The cell count was performed with Qupath’s cell counter, the settings varied amongst the different cell types. Custom object classifiers for cell detection were created to correct the count of cells in the blue (359 nm) channel. Custom pixel threshold was used for analysis with controls on each slide and for each cell type. The analysis process was automated using scripts generated by Java Groovy in Qupath. Quantification of co-localization was done without enhancement and with custom pixel classifiers trained per slide within Qupath. Training image for the red channel, green channel, and merged channel were created, and subsequently a custom pixel classifier was created and trained per channel to perform the measurements. The five square fields were randomly placed on each imaged. The averaged area measurement was divided by its cell count to get AU per cell and averaged per well for standardization. Each slide had a control non-stained well to ensure proper training and analysis. The analysis process was automated using scripts generated by Java Groovy in Qupath. Data were expressed as fluorescence intensity per cell for standardization. Data for Figures 1, 2 were obtained using images from an Axiovert 25 Zeiss Inverted microscope (Obrekochen, Germany). These were not compared with any of the other data.
**FIGURE 1:** *Progressive increase in SP uptake by cerebrovascular cells over time. (A) Schematic diagram showing the location of the three human cerebrovascular cell types used in this study [micro-vessel/capillaries endothelial cells (hCMEC/D3), pericytes (HBVP) and vascular smooth muscle cells (HBVSMC)]. Created by using BioRender.com. (B–D) Representative images of SARS-CoV-2 spike protein (SP)-555 (red) uptake, at 4 h, counter stained with 4’,6-diamidino-2-phenylindole (DAPI). Images on the right are the white boxed areas. (E) SP-555 uptake over time for the hCMEC/D3 (green), HBVP (red) and HBVSMC (purple). AU = Fluorescence arbitrary unit. Values are mean ± SEM. N = 3 to 6 wells per group. Scale bar = 100 μm.* **FIGURE 2:** *Concentration-dependent SARS-CoV-2/SP uptake by cerebrovascular cells. (A) SP-555 uptake at different SP-555 concentration for the hCMEC/D3 (green), HBVP (red), and HBVSMC (purple) using curve fitting. (B) Binding constant (Kd) for HBVSMC that was best fitted to a saturable uptake. There was no saturation for the other cell types over the concentrations used. Values are mean ± SEM. N = 3–5 wells/group Au = Fluorescence arbitrary unit. Supplementary Figures 1, 2.*
## Cell viability
Unlabeled wild type SP (0–200 nM) was used to measure in vitro cell viability by the MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay [Roche, Cell proliferation Kit 1 (MTT) Cat no. 11465007001]. This is a colorimetric method according to which a tetrazolium-based compound is reduced to formazan by living cells. The amount of formazan produced is directly proportional to the number of living cells in the culture. Each cell type was incubated for 24 h at 37°C in the presence of $5\%$ CO2.
## Statistic
All analysis were performed using Graphpad Prims version 9.2.0. Statistically analyzed was by analysis of variance (ANOVA) followed by Tukey post-hoc test for three or more groups. Unpaired t-test was used to compare two groups. The differences were considered to be significant at $p \leq 0.05.$ For statistical representation, *$P \leq 0.05$, **$P \leq 0.01$, and ***$P \leq 0.001$ and ****$P \leq 0.0001$ are the levels of statistical significance. All values were expressed as mean ± SEM. N is the number of cell culture wells per group. For each well the average intensity of the five fields was used. Outliers were identified and removed using ROUT method with $Q = 10$% in GraphPad Prism version 9.2.0.
## Progressive SARS-CoV-2/SP uptake by cerebrovascular cells over time
The human brain cerebrovascular cells, endothelial cell (hCMEC/D3), primary pericyte (HBVP), and primary vascular smooth muscle cell (HBVSMC) were used (Figure 1A), as a monolayer to characterize the uptake mechanisms of wild type (WT) spike proteins (SP-555). While SP-555 signal was associated with each cell type, it was mostly seen on the cell surface of hCMEC/D3, but for the other two cell types there were more internalization (Figures 1B–D). The time-dependent uptake pattern, determined at 100 nM SP-555 and over 20 h, showed that the cell type approached equilibrium after 6 h with the exception of HBVSMC, which was linear (Figure 1E). The a half-time (t$\frac{1}{2}$) for equilibration was 2–3 h. The endothelial cells had the lowest capacity to take up SP compared to the other two cell types, possibly due to difference in the cell size and restricted uptake. SP-555 intensity values were expressed as per cell.
## Concentration-dependent SARS-CoV-2/SP uptake by cerebrovascular cells
The concentration-dependent SP-555 uptake was determined at 4 h, and this showed a pattern of approaching saturation after about 200 nM for the HBVSMC, while for the hCMEC/D3 and HBVP it was almost linear (Figure 2A). Higher concentrations were not used as previous studies used lower or about 100 nM (Buzhdygan et al., 2020; Rhea et al., 2021) and the relevant of higher concentration maybe questionable. The estimated binding affinity for the HBVMC was 100 nM (Figure 2B). However, this Kd value was greater than that for ACE2/SP-555 binding (protein-protein binding), in vitro, using a non-cellar assay (Supplementary Figure 1), which was similar to that of the manufacturer value that also used a non-cell-based assay. Thus, possible mechanisms of SP uptake are likely similar for these cell types but may be due to differential uptake by multiple facilitators. We confirm that SP was not toxic to these cell types (Supplementary Figure 2). Even though there are limitations in using the 3-(4,5-dimethylthazol-2yl)-2,5-diphenyl tetrazolium bromide (MTT) assay (Ghasemi et al., 2021), we used the same conditions for each cell type and followed the manufacturer instructions. This assay is dependent on the mitochondria in viable cells to metabolism and reduce MTT to formazan, a water-soluble violet-blue compound. Thus, the increased levels seen at the higher SP concentration for the hCMEC/D3 may be due to increased metabolic activity due to the higher mitochondria content of cerebral endothelium (Oldendorf and Brown, 1975; Oldendorf et al., 1977; Andrade Silva et al., 2021; Mullen et al., 2021). This increase in intensity is not due to toxicity since normally, toxicity is associated with a reduction in the intensity (lower mitochondrial activity and less viable cells). SARS-CoV-2 may cause lipid toxicity, as reported for HEK293 cells (Nguyen V. et al., 2022). This should not significantly affect our data since the experiments were performed at or less than 4 h, while the MTT assay was performed after 24 h incubation with the SP.
## ACE2 mediates SARS-CoV-2/SP uptake in cerebrovascular cells
The pattern of SP uptake indicates some receptor binding and these cell types express a number of receptors that are associated with SARS-CoV-2, such as ACE2, a major binding site of SP (Supplementary Figure 3). ACE2 is expressed in human brain endothelial cells (Qiao et al., 2020), and in pericytes and vascular smooth muscle cells (He et al., 2020). We confirmed that ACE2 interacts with SP-555, in vitro, by using a non-cell-based assay (Supplementary Figure 1). The levels of ACE2 detected with an anti-ACE2 antibody (αACE2-488;10 μg/ml) were similar for the hCMEC/D3, HBVP and HBVSMC cells, and this represented about 50 to $80\%$ of SP-555 cellular binding (Figures 3A–C). Thus, ACE2 availability for SP binding was similar for these cell types. However, ACE2 (αACE2-488) co-localization with SP-555 (merged cellular binding-yellow areas) was $50\%$ of SP-555 binding for the hCMEC/D3 cells, compared to that of 15 and $20\%$ for the BHVP and HBVSMC, respectively (Figure 3D). Thus, there were more SP-555 cellular binding to other sites associated with the BHVP and HBVSMC compared to that of the hCMEC/D3 cells (Figures 3A–C). To further elaborate on this, we used excess unlabeled ligands, which will displace the specific binding of the labeled molecule. Excess unlabeled αACE2 (60 μg/ml) reduced SP-555 uptake by about 40–$50\%$ (about 50–$60\%$ SP-cellular binding remaining) for the hCMEC/D3, BHVP and HBVSMC cells (Figure 3E), which confirmed the data on SP-555 colocalization with αACE2-488 (Figure 3D). In the presence of excess unlabeled SP (1 μM) there was about $40\%$ (30–$50\%$) SP-555 cellular binding still associated with these cell types (Figure 3F). Thus, unlabeled SP displaced the bound labeled SP (SP-555) uptake by 50–$70\%$ for these cell types (Figure 3F). SP-555 uptake in the presence of excess unlabeled SP is due to membrane bound and/or non-specific uptake. To establish the levels of extracellular receptor binding, SP-555 uptake was determined at 4°C and compared to that at 37°C. At 37°C, SP-555 uptake was 2.2- to 5.5-fold greater than that at 4°C for these cell type (Figure 3G). Thus, there was likely greater SP interaction with the cells and internalization (including membrane bound) at 37°C. Plasma membrane lipid homeoviscosity is altered by temperature, which could affect protein uptake by the lipid bilayer environment. An illustration diagram of SP-555 uptake via ACE2 is shown in Figure 3H. There was no significant effect of angiotensin II, the endogenous ligand of ACE2, on SP uptake by the cerebrovascular endothelial cells (Supplementary Figure 4).
**FIGURE 3:** *ACE2 mediates SARS-CoV-2/SP uptake in cerebrovascular cells. (A–C) SP-555 and anti-human ACE2 antibody (αACE2-488) co-localized on each of the cell types, hCMEC/D3 (A), HBVP (B), and HBVSMC (C). Lowest row is the magnified image of the white boxed area above. (D) Intensities of αACE2 and merged SP/αACE2 (yellow) were expressed as a percentage of SP-555 AU (controls were the intensities of SP-555 cellular binding (all values were corrected as per cell). (E) Excess unlabeled αACE2 displaced SP-555 binding, which confirms the data in panel (D). (F) Excess unlabeled SP (self-competition) displaced SP-555 uptake. Controls were the intensities of SP-555 cellular binding without excess unlabeled αACE2 (panel E) or unlabeled SP (panel F). (G) SP-555 receptor binding (4°C) considerable less than that at 37°C. (H) Schematic diagram showing that SP-555 mimics SARS-CoV-2 binding to ACE2. Adapted from “Proposed Therapeutic Treatments for COVID-19 Targeting Viral Entry Mechanism by www.biorender.com (2021). Retrieved from https://app.biorender.com/biorender-templates. Red dashed line is the control levels (100%). Values are mean ± SEM. N = number of well use (each data point). Statistically analyzed was by analysis of variance (ANOVA) followed by Tukey post-hoc test. *P < 0.05, **P < 0.01, and ***P < 0.001 and ****P < 0.0001. GraphPad Prism version 9.2.0 was used. Scale bar = 100 μm. Supplementary Figures 1, 3.*
## Sialic acid/GM1-mediates SARS-CoV-2/SP uptake in cerebrovascular cells
Glycans, are carbohydrates based polysaccharides that are attached to molecules on the cell surface, and can bind many toxins and pathogens, including viruses (Lingwood, 2011). Sialic acid containing glycan (N-acetyl D-glucosamine) has been reported to play a role in SP binding (Rhea et al., 2021). Similarly, glycosaminoglycans (GAGs), which are sulfate polysaccharides, are thought to play a role in infections (Dick and Vogt, 2014; Aquino and Park, 2016; Lima et al., 2017; Shi et al., 2021). While wheat germ agglutinin (WGA), a lectin that binds sialic acid, increased SP-555 uptake by all three cell types, it was 2.4- to 3.2-greater for hCMEC/D3 and HBVP and 1.4-fold greater for HBVMC compared to that of controls without WGA (Figure 4A). SP uptake and internalization were increased with WGA (Supplementary Figure 5). In contrast, heparin, a polysaccharide, which belongs to the GAG family, did not affect SP uptake in the hCMEC/D3 cells but reduced its uptake to by 30–$60\%$ for HBVP and HBVSMC (Figure 4A). Further studies are needed to elaborate on the significance of these findings. Anti- GM1 antibody (αGM1) reduced SP-555 uptake by 60–$80\%$ of controls in all of these cell types (Figure 4A). However, in the presence of both antibodies (αACE2 and αGM1) the uptake was similar to that of αGM1 alone (Figure 4B). Since GM1 is present mainly in the lipid raft, SP binding by these cell types could be mediated via the caveolin/lipid raft. Further work is needed. There are reports that SP can be taken up into cells by clathrin-mediated endocytosis and by binding to sialic acid residues on cell membrane bound glycoproteins (Lingwood, 2011; Bayati et al., 2021; Rhea et al., 2021). We confirmed that these cell types, including the endothelial cells, can take up transferrin (Supplementary Figure 6A), a molecule that is known to be transported by clathrin-mediated vesicles (Inoue et al., 2007; Mayle et al., 2012). Bovine serum albumin (BSA) conjugated to lactosylceramide BODIPY, a molecule that is involved in the syntheses of gangliosides, including GM1 but not GM4, and taken up by the lipid raft, was taken up by these cell types and colocalized with SP-555 (Supplementary Figure 6B). Thus, it is possible that SP could be taken up by the lipid raft. GM1-BODIPY was also incorporated within the cell membrane (Supplementary Figure 6C). SP-555 is mainly cell membrane bound before possible internalization. While there is no specific inhibitor for each of these uptake mechanisms, nystatin (an inhibitor of lipid raft-mediated uptake) and chlorpromazine (an inhibitor of clathrin-mediated uptake) were tested (Plummer and Manchester, 2013; Wang et al., 2016; Sui et al., 2017). Both inhibitors were effective in blocking SP-555 uptake in each of these cell types by about $60\%$, except for chlorpromazine in the HBVSMC, which inhibited the uptake by $30\%$ (Figure 4C). It’s tempting to speculate that SP uptake by GM1 assists in its binding to ACE2 (Figure 4B), since both anti-ACE2 and anti-GM1 antibodies elicited the same effect as anti-GM1 alone, and ACE2 and GM1 are located within the caveolin/lipid raft (Glende et al., 2008; Lu et al., 2008; Garofalo et al., 2021). Gangliosides, which are present mostly in the lipid rafts of cell membranes mediate SARS-CoV-2 uptake by facilitating binding to ACE2 (Pirone et al., 2020; Fantini et al., 2020, 2021; Sun, 2021; Nguyen L. et al., 2022). The anti-GM1 antibody (Abcam, cat#Ab 23943), used in our studies was a rabbit polyclonal IgG isoform specific for GM1, with little interaction with the other ganglioside (sialic containing glycosphingolipid), a class of anionic glycosphingolipids (manufacturer information). Thus, there is little, if any, interaction of this anti-GM1 with GM2 and GM3. In addition, there was > $70\%$ inhibition of SP uptake with this anti-GM1 antibody (Figure 4), and thus, < $30\%$ of the SP uptake could be due to other facilitators.
**FIGURE 4:** *Sialic acid-/GM1-mediated SP uptake by cerebrovascular cells. (A) Effect of wheat germ agglutinin (WGA), heparin (HEP) and anti-mono-sialotetra- hexasylganglioside antibody (αGM1) on SP-555 uptake. (B) Effect of αGM1, anti-human ACE2 antibody (αACE2) and both αGM1 and αACE2 together on SP-555 uptake. (C) Effect of nystatin and chlorpromazine on SP-555 uptake. Controls were SP-555 intensities in the absence of the test compounds. (D) Schematic diagram showing a proposed SARS-CoV-2/SP uptake by both GM1 and ACE2. Both GM1 and ACE2 are present within the lipid raft region of cell membranes. Created by using BioRender.com. Red dashed line is the control levels (100%). Values are mean ± SEM. N = each data point is a well. Supplementary Figure 4. Statistically analyzed was by analysis of variance (ANOVA) followed by Tukey post-hoc test. *P < 0.05, **P < 0.01, and ***P < 0.001 and ****P < 0.0001. GraphPad Prism version 9.2.0 was used.*
## Higher mutant SARS-CoV-2/SP uptake by cerebrovascular cells
While SARS-CoV-2 variants harbor many mutation sites, three mutation sites were selected from variants of concern (WHO Coronavirus (COVID-19) Dashboard, n.d.) to determine if their uptake is altered in these cell type (Table 1). These mutations were within the RBD (N501Y and E484K) and one (D614G), a common mutation site, which is outside the RBD and furin cleavage site (Figure 5A). For the hCMEC/D3 cells, the uptake of mutants D614G, N501Y and E484K was significantly increased by 1. 5-, 1. 9-, and 2.8-fold, respectively, compared to control wild type SP (Figure 5B). In contrast, for the HBVP, the uptake of mutant D614G was unchanged, but for N501Y and E484K it was significantly increased by 1.7-fold compared to that of controls (Figure 5C). However, for the HBVSMC, the uptake of mutants D614G, N501Y, and E484K was significantly increased by 3. 2-, 5. 0-, and 3.8-fold, respectively, compared to that of controls (Figure 5D). Thus, again, there was differential mutant SP uptake by the three cell types. Uptakes of D614G, N501Y, and E484K were highest for the HBVSMC, but there was no significant differences for the hCMEC/D3 and HBVP (Supplementary Figure 7). This may reflect the type and distribution of SP receptors. Mutation sites E484K and N501Y confer gain-of-function, and N501Y increases SP affinity (Bayarri-Olmos et al., 2021; Tian et al., 2021; Xie et al., 2021). Mutant E484K also confers immune escape (Weisblum et al., 2020; Harvey et al., 2021). Mutant D614G seem to confers increased infectivity and transmissibility (Volz et al., 2021).
Mutant SP uptake was considerably increased with WGA by 6.9–17.9-fold in the three cell types compared to their respective mutant control without WGA (Figures 5E–M). This was 2.9–7.7-fold greater than that seen for WGA with the wild type SP (Figures 4A–C). Mutant SP uptake was inhibited by anti-GM1 in the three cell types (Figures 5E–M). In contrast, heparin had no effect except for E484K uptake in hCMEC/D3 and HBVSMC, which was decreased and increased, respectively (Figures 5E–M). Heparin inhibits SP binding in non-cell-based assays, as reported (Gupta et al., 2021; Ali et al., 2022). While *It is* unclear on how heparin increased E484K uptake in HBVSMC, it is possible that heparin binds E484K and this complex increased its binding to ACE2, a reported mechanism (Ali et al., 2022). Further work is needed. The neutralization effect of anti-GM1 antibody was less effective (Figures 5B–M) by 1.5–3.1-fold compared to wild type SP (Figures 4A–C). These differences may be due to the effectiveness of the mutants SP binding and the distribution/accessibility of the glycans and ACE2 on the three cell types. Excess αACE2 (60 μg/ml) suppressed SP E484K uptake (Figures 6A, B) but this was also less effective compared to that of wild type SP with the exception of hCMEC/D3 cells (Figures 3A–G).
**FIGURE 6:** *Anti-ACE2 suppressed mutant SARS-CoV-2/SP E484K uptake by cerebrovascular cells. (A,B) Excess unlabeled αACE2 suppressed labeled SP-555 E484K uptake in the hCMEC/D3 cells, HBVP, and HBVSMC compared to controls in the absence of unlabeled αACE2 for each cell type. Red dashed line is control levels in the absence of αACE2 (100%). Values are mean ± SEM. N = number of data points (wells) shown with the histogram. Scale bar = 100 μm. Supplementary Figure 7. **P < 0.01; ***P < 0.001.*
## Discussion
We have shown that there was differential uptake of SARS-CoV-2/SP by three human cerebrovascular cell types, endothelial cells, pericytes, and vascular smooth muscle cells. The endothelial cells, the physical site of the blood brain barrier (BBB), had the lowest capacity to take up SARS-CoV-2/SP compared to that of the other cell types, which may explain the low viral replication by the cerebrovasculature (Constant et al., 2021). In addition, SARS-CoV-2/SP uptake was dependent on the duration of its exposure and on its concentration (titer). Thus, longer exposure to SARS-CoV-2 and a higher titer, which are variable in the pathogenesis of the disease, will contribute to COVID-19 severity and possibly its neurological manifestations. SARS-CoV-2/SP uptake was mediated by both ACE2 and GM1, which contains sialic acid, but these two processes were not independent.
Uptake of SARS-CoV-2/mutant SP, N501Y, E484K, and D614G, were increased by the three cell types compared to that of the wild type SAR-CoV-2/SP, except for D613G in pericyte, which was unchanged. While WGA considerable increased the uptake of the three mutants SARS-CoV-2/SP, anti-ACE2 and anti-GM1 were less effective in neutralizing mutant SARS-CoV-2/SP uptake compared to that of the wild type controls. There was differential SAR-CoV-2/mutant SP uptake by these cell types. Therefore, in addition to ACE2, the established SARS-CoV-2 receptor, glycolipids, especially ganglioside, is a SARS-CoV-2/SP facilitators. While our hypothesis was that ACE2 will be the main facilitator, our data show that SARS-CoV-2 interaction with these host cells was more complex. It appears that multiple facilitators are involved to ensure that the virus access host cell machinery for its survival.
Our data show that ACE2 is a facilitator of wild type and mutant SARS-CoV-2 uptake by cerebrovascular cells, as it is for other host cell. There is an extensive literature on this (Pirone et al., 2020; Qiao et al., 2020; Brady et al., 2021; McQuaid et al., 2021). Even though each cell type expressed ACE2, there is differential effects by the cell type that can be subtle. Almost twice as much of SARS-CoV-2/SP colocalized with ACE2 for the endothelial cells compared to that of pericytes and vascular smooth muscle cells. Despite this the endothelial cells restricted the total uptake of SARS-CoV-2/SP compared to the other two cell types. The higher uptakes of the mutant SARS-CoV-2/SP (E484K, N501Y and D614G) likely associated with increased binding to facilitators, including ACE2 (WHO Coronavirus (COVID-19) Dashboard, n.d.). Further work is needed to explain why SARS-CoV-2 mutant D614G uptake is unchanged in the HBVP (human pericyte).
In contrast to ACE2, the role of gangliosides, sialic acid containing glycans and possibly other glycolipids in SARS-CoV-2 uptake is emerging as another viral facilitator. Glycolipids are a large group of heterogenous compounds, which consist of monosaccharide residues (head group on mainly the outer cell membrane surface) linked by a glycosidic bond to a hydrophobic lipid moiety (within the cell membrane), such as acylglycerol, sphingoid, or ceramide (Lingwood, 2011; Malhotra, 2012; Sipione et al., 2020). The sphingoids or glycosphingolipids are found mainly in animals cell membranes, and include neutral, such as cerebrosides, and acidic molecules, such as gangliosides. Gangliosides consist of a ceremide lipid moiety linked to an oligosaccharide chain of hexoses and sialic acids. Gangliosides have 0,1, 2, or 3 sialic acid moiety, and GM1, which has one sialic acid residue, is a common member of the ganglioside group (Lingwood, 2011; Malhotra, 2012; Sipione et al., 2020). They are mainly present in the lipid raft of plasma membrane (Brown and London, 2000; Degroote et al., 2004; Moreno-Altamirano et al., 2007; Lingwood, 2011; D’Angelo et al., 2013). While these molecules are ubiquitous expressed in cell membranes, studies on the relative abundance of each glycolipid or ganglioside and their functions in different cell types have not been reported (Sipione et al., 2020). This is likely due to the heterogeneity of these diverse molecules, low levels to detect and their varied functions (Svennerholm, 1963; Sipione et al., 2020). While it is believed that glycolipids make up a small fraction of the plasma membrane lipids, they have essential functions, which include plasma membrane fluidity, stabilization of the plasma membrane, protein receptor distribution, protein-protein interaction, ligand-receptor interaction, cell-cell communications, adhesion and release of neurotrophins, due mainly to their amphipathic nature (Battistin et al., 1985; Vorbrodt, 1986; Iwabuchi et al., 1998; Hakomori, 2000; Mitsuda et al., 2002; Jeyakumar et al., 2003; Ngamukote et al., 2007; Furian et al., 2008; Lingwood, 2011; Malhotra, 2012; Sipione et al., 2020).
Gangliosides, including GM1, are expressed in vascular endothelial cells, and involved in endocytosis and signaling (Born and Palinski, 1985; Weigel and Yik, 2002). Brain micro vessels and cultured brain endothelial cells, including cell lines, express sialic acid oligosaccharides (N acetyl-D-glucosamine), which binds WGA (Fatehi et al., 1987; Plattner et al., 2010). SARS-CoV-2/SP binds to WGA and increased its uptake into brain, in mice (Erickson et al., 2021; Rhea et al., 2021). Our data confirm that WGA binds SARS-CoV-2/SP, but it’s mainly located on the endothelial cell surface. However, our data show that anti-GM1 antibody suppressed SARS-CoV-2/SP uptake in all three cerebrovascular cell type, and thus, a facilitator for SARS-CoV-2/SP. Human brain endothelial cells contain low levels of gangliosides, including GM3, GM2 and GM1 (Gillard et al., 1987; Kanda et al., 1994; Duvar et al., 1997, 2000; Müthing et al., 1999). These compounds are structurally and functionally polymorphic and the content varies in different tissues, age, conditions and animals (Ledeen and Wu, 2015; Schengrund, 2015; Aureli et al., 2016).
However, GM1 protects the cerebrovasculature from photochemical-induced (rose Bengal)-induced damage (Frontczak-Baniewicz et al., 2000), blast traumatic brain injury (Rubovitch et al., 2017), oxidative damage (Zhao et al., 2015). neurovascular injury of glutamate and kanate (Favaron et al., 1988), alcohol injury (Hungund et al., 1990), calcium toxicity (Nakamura et al., 1992), injury caused by middle cerebral artery occlusion (MCAO) (Zhang et al., 2019), and diabetic injuries (Figliomeni et al., 1992). GM1 also increases cerebral blood flow (CBF) mediated by nitric oxide (Furian et al., 2008). In addition, GM1 and GM2 increase cell proliferation, DNA synthesis and protects the VSMC (Sachinidis et al., 2000; Gouni-Berthold et al., 2001). Gangliosides (GM2 > GM1) potentiates platelet-derived growth factor (PDGF) induced proliferation of VSMC (Sachinidis et al., 1996; Sasaki and Toyoda, 2019). GM1 functions as a co-receptor for fibroblast growth factor in endothelial cells (Rusnati et al., 2002, 1999). Thus, it is possible that SARS-CoV-2 interaction at the cerebravasculature could contribute to cerebrovascular dysfunction and possibly Alzheimer’s disease (AD)-like symptoms.
The brain contains many forms of gangliosides, but about $95\%$ of the total is made up of GM1, GD1a, GD1b, and GQ1b (Gong et al., 2002; Kaida et al., 2009; Wang and Yu, 2013; Chiricozzi et al., 2020). Levels of GM1 are reduced with aging and in AD, while GM2 is increased (Krauss and Burke, 1982; Kracun et al., 1991; Svennerholm et al., 2002; Svennerholm and Gottfries, 2008; Liu et al., 2015; Ledeen and Wu, 2018). GM1 levels are also reduced in other neurodegenerative diseases, such as Huntington’s and Parkinson’s diseases (Svennerholm et al., 2002; Desplats et al., 2007; Schneider et al., 2010; Magistretti et al., 2019; Sipione et al., 2020). GM1 levels in CSF has been shown to improve day-to-day activity in AD (Augustinsson et al., 1997), and it increases choline acetyl esterase activity (ChAT) (Fong et al., 1995). It potentiates the effects of neurotrophic factor (Facci et al., 1990; Ferrari et al., 1993; Garofalo and Cuello, 1994) and basic fibroblast growth factor (Iwashita et al., 1996). GM1 has been shown to reduced amyloid-β toxicity (Kreutz et al., 2011). Thus, SARS-CoV-2 interacting with gangliosides may lead to AD-like symptoms.
However, glycolipids are known to bind many toxins and pathogens, including viruses (Lingwood, 2011; Sipione et al., 2020; Nicoli et al., 2021). GM1 is a receptor for microbes. It binds toxins, such as cholera and Shiga, and viruses, such as influenza and HIV (Sachinidis et al., 2000; Ilver et al., 2003; Suzuki, 2005; Lehmann et al., 2006; Chinnapen et al., 2007; Varki, 2007; Johannes, 2017; Chiricozzi et al., 2020; Cutillo et al., 2020). Sialic acid on GM1 binds the B subunit of cholera toxin and affluenza A (Wu et al., 2007). Guillain-Barre syndrome (GBS) appears to be associated with the presence of anti-gangliosides antibodies in blood, and there were reports of GBS-like effects in some SARS-CoV-2 and influenza vaccinated subjects (Kaida et al., 2009; Vellozzi et al., 2014; Martín Arias et al., 2015; Lunn et al., 2021; Shapiro Ben David et al., 2021). However, sialic containing glycans, like GM1, offers an additions potential target to be considered and to explore therapeutics for of SARS-CoV-2.
## Conclusion
SARS-CoV-2/SP uptake by three cell types of human cerebrovasculature (endothelial cells, pericytes, and VSMC) was time and concentration/titer dependent. It was the lowest for the endothelial cells, which may limit viral uptake in the normal healthy brains. Wild type SARS-CoV-2/SP, and mutant SARS-CoV-2/SP containing the common mutation sites, D614G, N501Y, and E484K, as seen in variants of interest, were differentially taken up by GM1- and ACE2-mediated mechanisms by the three the cell types. WGA, a lectin that binds sialic acid, increases mutant SARS-CoV-2/SP considerably compared to that of the wild type controls. While the mutant SARS-CoV-2/SPs were more effective in binding, anti-ACE2 and anti-GM1 antibodies were less effective in neutralizing their uptake. *In* general, the uptake mechanisms were similar in each cell type. Since SARS-CoV-2 uptake is the initial step in the viral penetration into cells, the data suggest that GM1, which contains one sialic acid residue, is also an important entry point of SARS-CoV-2 into these cells. GM1 could be an addition potential SARS-CoV-2 and therapeutic target at the cerebrovasculature, and perhaps other cell types. It is likely that in the severely infected patients there is greater uptake by the brain due to the degree of cerebrovasculature dysfunction in the aging brain, prior health conditions, complication with the infections, and the titre and duration of the viral exposure. Cardio-respiratory failure is likely a major contributed to the neurological symptoms since the brain is dependent on an adequate supply oxygenated blood containing nutrients, especially glucose.
## Limitations of the study
This is an in vitro study designed to explore possible mechanisms for SARS-CoV-2/SP uptake by three cell types of the cerebrovasculature exposed to the same titer/concentration for 4 h (usually) in controlled conditions. In human, it is likely rare to be exposed to the same viral titer for 4 h. The RBD of the SARS-CoV-2/SPs was used as a model of SARS-CoV-2, since it is essential for viral entry into host cells and viral binding/uptake was investigated in this study. Thus, only the attachment part of the viral life cycle can be explored with. Also, other cell types of the neurovascular unit, such as astrocytes and microglia need to be explored to confirm SARS-CoV-2/SP uptake mechanisms. However, it is likely that there will be similar mechanisms since these cells also express ACE2 and have sialic containing glycans, including GM1.
## Further studies
There are a number of possible lines for future studies, which include the following:
## 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.
## Author contributions
CM performed experiments, imaging, preparation of figures and table, and contributed to manuscript preparation. AS performed experiments, imaging, data analysis, preparation of figures and table, contributed to manuscript preparation, and prepared the references. ID contributed resources and critical review of the manuscript. RD provided the concept, designed the study, and wrote the manuscript. All authors approved 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/fnins.2023.1117845/full#supplementary-material
## References
1. Ali N., Khan R., AlAsmari A. F., Kumar V.. **In silico investigations of heparin binding to SARS-CoV-2 variants with a focus at the RBD/ACE2 interface.**. (2022) **115** 70-79. DOI: 10.1016/j.procbio.2022.02.012
2. Andrade Silva M., da Silva A. R. P. A., do Amaral M. A., Fragas M. G., Câmara N. O. S.. **Metabolic alterations in SARS-CoV-2 infection and its implication in kidney dysfunction.**. (2021) **12**. DOI: 10.3389/fphys.2021.624698
3. Aquino R. S., Park P. W.. **Glycosaminoglycans and infection.**. (2016) **21** 1260-1277. DOI: 10.2741/4455
4. Augustinsson L.-E., Blennow K., Blomstrand C., Bråne G., Ekman R., Fredman P.. **Intracerebroventricular administration of GM1 ganglioside to presenile Alzheimer patients.**. (1997) **8** 26-33. DOI: 10.1159/000106597
5. Aureli M., Mauri L., Ciampa M. G., Prinetti A., Toffano G., Secchieri C.. **GM1 Ganglioside: past studies and future potential.**. (2016) **53** 1824-1842. DOI: 10.1007/s12035-015-9136-z
6. Battistin L., Cesari A., Galligioni F., Marin G., Massarotti M., Paccagnella D.. **Effects of GM1 ganglioside in cerebrovascular diseases: a double-blind trial in 40 cases.**. (1985) **24** 343-351. DOI: 10.1159/000115823
7. Bayarri-Olmos R., Jarlhelt I., Johnsen L. B., Hansen C. B., Helgstrand C., Rose Bjelke J.. **Functional effects of receptor-binding domain mutations of SARS-CoV-2 B.1.351 and P.1 variants.**. (2021) **12**. DOI: 10.3389/fimmu.2021.757197
8. Bayati A., Kumar R., Francis V., McPherson P. S.. **SARS-CoV-2 infects cells after viral entry via clathrin-mediated endocytosis.**. (2021) **296**. DOI: 10.1016/j.jbc.2021.100306
9. Bchetnia M., Girard C., Duchaine C., Laprise C.. **The outbreak of the novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2): a review of the current global status**. (2020) **13**. DOI: 10.1016/J.JIPH.2020.07.011
10. Born G. V. R., Palinski W.. **Unusually high concentrations of sialic acids on the surface of vascular endothelia.**. (1985) **66** 543-549. PMID: 4063159
11. Brady M., McQuaid C., Solorzano A., Johnson A., Combs A., Venkatraman C.. **Spike protein multiorgan tropism suppressed by antibodies targeting SARS-CoV-2.**. (2021) **4**. DOI: 10.1038/s42003-021-02856-x
12. Brown D. A., London E.. **Structure and function of sphingolipid- and cholesterol-rich membrane rafts.**. (2000) **275** 17221-17224. DOI: 10.1074/jbc.R000005200
13. Buzhdygan T. P., DeOre B. J., Baldwin-Leclair A., Bullock T. A., McGary H. M., Khan J. A.. **The SARS-CoV-2 spike protein alters barrier function in 2D static and 3D microfluidic in-vitro models of the human blood–brain barrier.**. (2020) **146**. DOI: 10.1016/j.nbd.2020.105131
14. Chen Z., Mi L., Xu J., Yu J., Wang X., Jiang J.. **Function of HAb18G/CD147 in invasion of host cells by severe acute respiratory syndrome coronavirus.**. (2005) **191** 755-760. DOI: 10.1086/427811
15. Chinnapen D. J.-F., Chinnapen H., Saslowsky D., Lencer W. I.. **Rafting with cholera toxin: endocytosis and trafficking from plasma membrane to ER.**. (2007) **266** 129-137. DOI: 10.1111/j.1574-6968.2006.00545.x
16. Chiricozzi E., Lunghi G., di Biase E., Fazzari M., Sonnino S., Mauri L.. **GM1 ganglioside is a key factor in maintaining the mammalian neuronal functions avoiding neurodegeneration.**. (2020) **21**. DOI: 10.3390/ijms21030868
17. Constant O., Barthelemy J., Bolloré K., Tuaillon E., Gosselet F., Chable-Bessia C.. **SARS-CoV-2 poorly replicates in cells of the human blood-brain barrier without associated deleterious effects.**. (2021) **12**. DOI: 10.3389/fimmu.2021.697329
18. Cutillo G., Saariaho A.-H., Meri S.. **Physiology of gangliosides and the role of antiganglioside antibodies in human diseases.**. (2020) **17** 313-322. DOI: 10.1038/s41423-020-0388-9
19. D’Angelo G., Capasso S., Sticco L., Russo D.. **Glycosphingolipids: synthesis and functions.**. (2013) **280** 6338-6353. DOI: 10.1111/febs.12559
20. Degroote S., Wolthoorn J., van Meer G.. **The cell biology of glycosphingolipids.**. (2004) **15** 375-387. DOI: 10.1016/j.semcdb.2004.03.007
21. Desplats P. A., Denny C. A., Kass K. E., Gilmartin T., Head S. R., Sutcliffe J. G.. **Glycolipid and ganglioside metabolism imbalances in Huntington’s disease.**. (2007) **27** 265-277. DOI: 10.1016/j.nbd.2007.05.003
22. Dhakal B. P., Sweitzer N. K., Indik J. H., Acharya D., William P.. **SARS-CoV-2 infection and cardiovascular disease: COVID-19 heart.**. (2020) **29** 973-987. DOI: 10.1016/j.hlc.2020.05.101
23. Dick R. A., Vogt V. M.. **Membrane interaction of retroviral Gag proteins.**. (2014) **5**. DOI: 10.3389/FMICB.2014.00187/BIBTEX
24. Doobay M. F., Talman L. S., Obr T. D., Tian X., Davisson R. L., Lazartigues E.. **Differential expression of neuronal ACE2 in transgenic mice with overexpression of the brain renin-angiotensin system.**. (2007) **292** R373-R381. DOI: 10.1152/ajpregu.00292.2006
25. Duvar S., Peter-Katalinic’ J., Hanisch F.-G., Miithing J., Bielefeld U.. **Isolation and structural characterization of glycosphingolipids of in vitro propagated bovine aortic endothelial cells.**. (1997) **7** 1099-1109. DOI: 10.1093/glycob/7.8.1099
26. Duvar S., Suzuki M., Muruganandam A., Yu R. K.. **Glycosphingolipid composition of a new immortalized human cerebromicrovascular endothelial cell line.**. (2000) **75** 1970-1976. DOI: 10.1046/j.1471-4159.2000.0751970.x
27. Erickson M. A., Rhea E. M., Knopp R. C., Banks W. A.. **Interactions of SARS-CoV-2 with the Blood–Brain Barrier.**. (2021) **22**. DOI: 10.3390/ijms22052681
28. Facci L., Leon A., Skaper S. D.. **Hypoglycemic neurotoxicity in vitro: involvement of excitatory amino acid receptors and attenuation by monosialoganglioside GM1.**. (1990) **37** 709-716. DOI: 10.1016/0306-4522(90)90101-9
29. Fanelli V., Fiorentino M., Cantaluppi V., Gesualdo L., Stallone G., Ronco C.. **Acute kidney injury in SARS-CoV-2 infected patients.**. (2020) **24**. DOI: 10.1186/s13054-020-02872-z
30. Fantini J., Chahinian H., Yahi N.. **Leveraging coronavirus binding to gangliosides for innovative vaccine and therapeutic strategies against COVID-19.**. (2021) **538** 132-136. DOI: 10.1016/j.bbrc.2020.10.015
31. Fantini J., di Scala C., Chahinian H., Yahi N.. **Structural and molecular modelling studies reveal a new mechanism of action of chloroquine and hydroxychloroquine against SARS-CoV-2 infection.**. (2020) **55**. DOI: 10.1016/j.ijantimicag.2020.105960
32. Fatehi M. I., Gerhart D. Z., Myers T. G., Drewes L. R.. **Characterization of the blood-brain barrier: glycoconjugate receptors of 14 lectins in canine brain, cultured endothelial cells, and blotted membrane proteins.**. (1987) **415** 30-39. DOI: 10.1016/0006-8993(87)90266-6
33. Favaron M., Manev H., Alho H., Bertolino M., Ferret B., Guidotti A.. **Gangliosides prevent glutamate and kainate neurotoxicity in primary neuronal cultures of neonatal rat cerebellum and cortex.**. (1988) **85** 7351-7355. DOI: 10.1073/pnas.85.19.7351
34. Ferrari G., Batistatou A., Greene L.. **Gangliosides rescue neuronal cells from death after trophic factor deprivation.**. (1993) **13** 1879-1887. DOI: 10.1523/JNEUROSCI.13-05-01879.1993
35. Figliomeni B., Bacci B., Panozzo C., Fogarolo F., Triban C., Fiori M. G.. **Experimental diabetic neuropathy: effect of ganglioside treatment on axonal transport of cytoskeletal proteins.**. (1992) **41** 866-871. DOI: 10.2337/diab.41.7.866
36. Fong T. G., Neff N. H., Hadjiconstantinou M.. **Systemic administration of GM1 ganglioside increases choline acetyltransferase activity in the brain of aged rats.**. (1995) **132** 157-161. DOI: 10.1016/0014-4886(95)90020-9
37. Frontczak-Baniewicz M., Gadamski R., Barskov I., Gajkowska B.. **Beneficial effects of GM1 ganglioside on photochemically-induced microvascular injury in cerebral cortex and hypophysis in rat.**. (2000) **52** 111-118. DOI: 10.1016/S0940-2993(00)80094-9
38. Furian A. F., Oliveira M. S., Magni D. V., Souza M. A., Bortoluzzi V. T., Bueno L. M.. **l-NAME prevents GM1 ganglioside-induced vasodilation in the rat brain.**. (2008) **53** 362-369. DOI: 10.1016/j.neuint.2008.07.011
39. Garofalo L., Cuello A. C.. **Nerve growth factor and the monosialoganglioside GM1: analogous and different in vivo effects on biochemical, morphological, and behavioral parameters of adult cortically lesioned rats.**. (1994) **125** 195-217. DOI: 10.1006/exnr.1994.1024
40. Garofalo T., Misasi R., Preta G.. **Editorial: targeting lipid rafts as a strategy against infection and cancer.**. (2021) **9**. DOI: 10.3389/fcell.2021.748905
41. Ghasemi M., Turnbull T., Sebastian S., Kempson I.. **The MTT assay: utility, limitations, pitfalls, and interpretation in bulk and single-cell analysis.**. (2021) **22**. DOI: 10.3390/ijms222312827
42. Gillard B. K., Jones M. A., Marcus D. M.. **Glycosphingolipids of human umbilical vein endothelial cells and smooth muscle cells.**. (1987) **256** 435-445. DOI: 10.1016/0003-9861(87)90600-X
43. Glende J., Schwegmann-Wessels C., Al-Falah M., Pfefferle S., Qu X., Deng H.. **Importance of cholesterol-rich membrane microdomains in the interaction of the S protein of SARS-coronavirus with the cellular receptor angiotensin-converting enzyme 2.**. (2008) **381** 215-221. DOI: 10.1016/j.virol.2008.08.026
44. Gong Y., Tagawa Y., Lunn M. P. T., Laroy W., Heffer-Lauc M., Li C. Y.. **Localization of major gangliosides in the PNS: implications for immune neuropathies.**. (2002) **125** 2491-2506. DOI: 10.1093/brain/awf258
45. Gordon D. E., Jang G. M., Bouhaddou M., Xu J., Obernier K., White K. M.. **A SARS-CoV-2 protein interaction map reveals targets for drug repurposing.**. (2020) **583** 459-468. DOI: 10.1038/s41586-020-2286-9
46. Gouni-Berthold I., Seul C., Ko Y., Hescheler J., Sachinidis A.. **Gangliosides GM1 and GM2 induce vascular smooth muscle cell proliferation via extracellular signal-regulated kinase 1/2 pathway.**. (2001) **38** 1030-1037. DOI: 10.1161/hy1101.093104
47. Gupta Y., Maciorowski D., Zak S. E., Kulkarni C. V., Herbert A. S., Durvasula R.. **Heparin: a simplistic repurposing to prevent SARS-CoV-2 transmission in light of its in-vitro nanomolar efficacy.**. (2021) **183** 203-212. DOI: 10.1016/j.ijbiomac.2021.04.148
48. Hakomori S. I.. **Cell adhesion/recognition and signal transduction through glycosphingolipid microdomain.**. (2000) **17** 143-151. DOI: 10.1023/a:1026524820177
49. Hamming I., Timens W., Bulthuis M., Lely A., Navis G., van Goor H.. **Tissue distribution of ACE2 protein, the functional receptor for SARS coronavirus. A first step in understanding SARS pathogenesis.**. (2004) **203** 631-637. DOI: 10.1002/path.1570
50. Harvey W. T., Carabelli A. M., Jackson B., Gupta R. K., Thomson E. C., Harrison E. M.. **SARS-CoV-2 variants, spike mutations and immune escape.**. (2021) **19** 409-424. DOI: 10.1038/s41579-021-00573-0
51. He L., Mäe M. A., Muhl L., Sun Y., Pietilä R., Nahar K.. **Pericyte-specific vascular expression of SARS-CoV-2 receptor ACE2 – implications for microvascular inflammation and hypercoagulopathy in COVID-19.**. (2020). DOI: 10.1101/2020.05.11.088500
52. Hoffmann M., Kleine-Weber H., Schroeder S., Krüger N., Herrler T., Erichsen S.. **SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor.**. (2020) **181** 271-280.e8. DOI: 10.1016/j.cell.2020.02.052
53. Huang Y. H., Jiang D., Huang J. T.. **SARS-CoV-2 detected in cerebrospinal fluid by PCR in a case of COVID-19 encephalitis.**. (2020) **87**. DOI: 10.1016/j.bbi.2020.05.012
54. Hungund B. L., Reddy M. V., Bharucha V. A., Mahadik S. P.. **Monosialogangliosides (GM1 and AGF2) reduce acute ethanol intoxication: sleep time mortality, and cerebral cortical Na+, K+-ATPase.**. (1990) **19** 443-451. DOI: 10.1002/ddr.430190409
55. Ilver D., Johansson P., Miller-Podraza H., Nyholm P.-G., Teneberg S., Karlsson K.-A.. **Bacterium-host protein-carbohydrate interactions.**. (2003) **363** 134-157. DOI: 10.1016/S0076-6879(03)01049-8
56. Inoue Y., Tanaka N., Tanaka Y., Inoue S., Morita K., Zhuang M.. **Clathrin-dependent entry of severe acute respiratory syndrome coronavirus into target cells expressing ACE2 with the cytoplasmic tail deleted.**. (2007) **81** 8722-8729. DOI: 10.1128/JVI.00253-07
57. Iwabuchi K., Handa K., Hakomori S.. **Separation of “glycosphingolipid signaling domain” from caveolin-containing membrane fraction in mouse melanoma B16 cells and its role in cell adhesion coupled with signaling.**. (1998) **273** 33766-33773. DOI: 10.1074/jbc.273.50.33766
58. Iwashita A., Hisajima H., Notsu Y., Okuhara M.. **Effects of basic fibroblast growth factor and ganglioside GM1 on neuronal survival in primary cultures and on eight-arm radial maze task in adult rats following partial fimbria transections.**. (1996) **353** 342-348. DOI: 10.1007/BF00168638
59. Jeyakumar M., Thomas R., Elliot-Smith E., Smith D. A., van der Spoel A. C., D’Azzo A.. **Central nervous system inflammation is a hallmark of pathogenesis in mouse models of GM1 and GM2 gangliosidosis.**. (2003) **126** 974-987. DOI: 10.1093/brain/awg089
60. Johannes L.. **Shiga Toxin-A model for glycolipid-dependent and lectin-driven endocytosis.**. (2017) **9**. DOI: 10.3390/toxins9110340
61. Kaida K., Ariga T., Yu R. K.. **Antiganglioside antibodies and their pathophysiological effects on Guillain-Barré syndrome and related disorders–a review.**. (2009) **19** 676-692. DOI: 10.1093/glycob/cwp027
62. Kanda T., Yoshino H., Ariga T., Yamawaki M., Yu R. K.. **Glycosphingolipid antigens in cultured bovine brain microvascular endothelial cells: sulfoglucuronosyl paragloboside as a target of monoclonal IgM in demyelinative neuropathy [corrected]**. (1994) **126** 235-246. DOI: 10.1083/jcb.126.1.235
63. Kracun I., Rosner H., Drnovsek V., Heffer-Lauc M., Cosović C., Lauc G.. **Human brain gangliosides in development, aging and disease.**. (1991) **35** 289-295. PMID: 1814411
64. Krauss R. M., Burke D. J.. **Identification of multiple subclasses of plasma low density lipoproteins in normal humans.**. (1982) **23** 97-104. DOI: 10.1016/S0022-2275(20)38178-5
65. Kreutz F., Frozza R. L., Breier A. C., de Oliveira V. A., Horn A. P., Pettenuzzo L. F.. **Amyloid-β induced toxicity involves ganglioside expression and is sensitive to GM1 neuroprotective action.**. (2011) **59** 648-655. DOI: 10.1016/j.neuint.2011.06.007
66. Ledeen R. W., Wu G.. **The multi-tasked life of GM1 ganglioside, a true factotum of nature.**. (2015) **40** 407-418. DOI: 10.1016/j.tibs.2015.04.005
67. Ledeen R., Wu G.. **Gangliosides of the nervous system.**. (2018) **1804** 19-55. DOI: 10.1007/978-1-4939-8552-4_2/FIGURES/3
68. Lehmann F., Tiralongo E., Tiralongo J.. **Sialic acid-specific lectins: occurrence, specificity and function.**. (2006) **63** 1331-1354. DOI: 10.1007/s00018-005-5589-y
69. Li Y., Zhou W., Yang L., You R.. **Physiological and pathological regulation of ACE2, the SARS-CoV-2 receptor.**. (2020) **157**. DOI: 10.1016/j.phrs.2020.104833
70. Lima M., Rudd T., Yates E.. **New applications of heparin and other glycosaminoglycans.**. (2017) **22**. DOI: 10.3390/molecules22050749
71. Lingwood C. A.. **Glycosphingolipid functions.**. (2011) **3**. DOI: 10.1101/cshperspect.a004788
72. Liu L., Zhang K., Tan L., Chen Y.-H., Cao Y.-P.. **Alterations in cholesterol and ganglioside GM1 content of lipid rafts in platelets from patients with Alzheimer disease.**. (2015) **29** 63-69. DOI: 10.1097/WAD.0000000000000041
73. Lu Y., Liu D. X., Tam J. P.. **Lipid rafts are involved in SARS-CoV entry into Vero E6 cells.**. (2008) **369** 344-349. DOI: 10.1016/j.bbrc.2008.02.023
74. Lunn M. P., Cornblath D. R., Jacobs B. C., Querol L., van Doorn P. A., Hughes R. A.. **COVID-19 vaccine and Guillain-Barré syndrome: let’s not leap to associations.**. (2021) **144** 357-360. DOI: 10.1093/brain/awaa444
75. Ma M., Xu Y., Su Y., Ong S.-B., Hu X., Chai M.. **Single-cell transcriptome analysis decipher new potential regulation mechanism of ACE2 and NPs signaling among heart failure patients infected with SARS-CoV-2.**. (2021) **8**. DOI: 10.3389/fcvm.2021.628885
76. Magistretti P. J., Geisler F. H., Schneider J. S., Andy Li P., Fiumelli H., Sipione S.. **Gangliosides: treatment avenues in neurodegenerative disease.**. (2019) **10**. DOI: 10.3389/FNEUR.2019.00859/XML/NLM
77. Malhotra R.. **Membrane glycolipids: functional heterogeneity: a review.**. (2012) **1**. DOI: 10.4172/2161-1009.1000108
78. Mao L., Jin H., Wang M., Hu Y., Chen S., He Q.. **Neurologic manifestations of hospitalized patients with coronavirus disease 2019 in Wuhan, China.**. (2020) **77**. DOI: 10.1001/jamaneurol.2020.1127
79. Mao R., Qiu Y., He J.-S., Tan J.-Y., Li X.-H., Liang J.. **Manifestations and prognosis of gastrointestinal and liver involvement in patients with COVID-19: a systematic review and meta-analysis.**. (2020) **5** 667-678. DOI: 10.1016/S2468-1253(20)30126-6
80. Marjot T., Webb G. J., Barritt A. S., Moon A. M., Stamataki Z., Wong V. W.. **COVID-19 and liver disease: mechanistic and clinical perspectives.**. (2021) **18** 348-364. DOI: 10.1038/s41575-021-00426-4
81. Martín Arias L. H., Sanz R., Sáinz M., Treceño C., Carvajal A.. **Guillain-Barré syndrome and influenza vaccines: a meta-analysis.**. (2015) **33** 3773-3778. DOI: 10.1016/j.vaccine.2015.05.013
82. Martinez-Rojas M. A., Vega-Vega O., Bobadilla N. A.. **Is the kidney a target of SARS-CoV-2?**. (2020) **318** F1454-F1462. DOI: 10.1152/ajprenal.00160.2020
83. Matsuyama S., Nao N., Shirato K., Kawase M., Saito S., Takayama I.. **Enhanced isolation of SARS-CoV-2 by TMPRSS2-expressing cells.**. (2020) **117** 7001-7003. DOI: 10.1073/pnas.2002589117
84. Mayle K. M., Le A. M., Kamei D. T.. **The intracellular trafficking pathway of transferrin.**. (2012) **1820** 264-281. DOI: 10.1016/j.bbagen.2011.09.009
85. McQuaid C., Brady M., Deane R.. **SARS-CoV-2: is there neuroinvasion?**. (2021) **18**. DOI: 10.1186/s12987-021-00267-y
86. Mitsuda T., Furukawa K., Fukumoto S., Miyazaki H., Urano T., Furukawa K.. **Overexpression of ganglioside GM1 results in the dispersion of platelet-derived growth factor receptor from glycolipid-enriched microdomains and in the suppression of cell growth signals.**. (2002) **277** 11239-11246. DOI: 10.1074/jbc.M107756200
87. Moreno-Altamirano M. M. B., Aguilar-Carmona I., Sánchez-García F. J.. **Expression of GM1, a marker of lipid rafts, defines two subsets of human monocytes with differential endocytic capacity and lipopolysaccharide responsiveness.**. (2007) **120** 536-543. DOI: 10.1111/j.1365-2567.2006.02531.x
88. Moriguchi T., Harii N., Goto J., Harada D., Sugawara H., Takamino J.. **A first case of meningitis/encephalitis associated with SARS-Coronavirus-2.**. (2020) **94** 55-58. DOI: 10.1016/j.ijid.2020.03.062
89. Mullen P. J., Garcia G., Purkayastha A., Matulionis N., Schmid E. W., Momcilovic M.. **SARS-CoV-2 infection rewires host cell metabolism and is potentially susceptible to mTORC1 inhibition.**. (2021) **12**. DOI: 10.1038/s41467-021-22166-4
90. Müthing J., Duvar S., Heitmann D., Hanisch F. G., Neumann U., Lochnit G.. **Isolation and structural characterization of glycosphingolipids of in vitro propagated human umbilical vein endothelial cells.**. (1999) **9** 459-468. DOI: 10.1093/glycob/9.5.459
91. Nakamura K., Wu G., Ledeen R. W.. **Protection of neuro-2a cells against calcium ionophore cytotoxicity by gangliosides.**. (1992) **31** 245-253. DOI: 10.1002/jnr.490310205
92. Ngamukote S., Yanagisawa M., Ariga T., Ando S., Yu R. K.. **Developmental changes of glycosphingolipids and expression of glycogenes in mouse brains.**. (2007) **103** 2327-2341. DOI: 10.1111/j.1471-4159.2007.04910.x
93. Nguyen L., McCord K. A., Bui D. T., Bouwman K. M., Kitova E. N., Elaish M.. **Sialic acid-containing glycolipids mediate binding and viral entry of SARS-CoV-2.**. (2022) **18** 81-90. DOI: 10.1038/s41589-021-00924-1
94. Nguyen V., Zhang Y., Gao C., Cao X., Tian Y., Carver W.. **The spike protein of SARS-CoV-2 impairs lipid metabolism and increases susceptibility to lipotoxicity: implication for a role of Nrf2.**. (2022) **11**. DOI: 10.3390/cells11121916
95. Nicoli E.-R., Annunziata I., D’Azzo A., Platt F. M., Tifft C. J., Stepien K. M.. **GM1 gangliosidosis—A mini-review.**. (2021) **12**. DOI: 10.3389/fgene.2021.734878
96. Oldendorf W. H., Brown W. J.. **Greater number of capillary endothelial cell mitochondria in brain than in muscle.**. (1975) **149** 736-738. DOI: 10.3181/00379727-149-38889
97. Oldendorf W. H., Cornford M. E., Brown W. J.. **The large apparent work capability of the blood-brain barrier: a study of the mitochondrial content of capillary endothelial cells in brain and other tissues of the rat.**. (1977) **1** 409-417. DOI: 10.1002/ana.410010502
98. Perez-Bermejo J. A., Kang S., Rockwood S. J., Simoneau C. R., Joy D. A., Silva A. C.. **SARS-CoV-2 infection of human iPSC–derived cardiac cells reflects cytopathic features in hearts of patients with COVID-19.**. (2021) **13**. DOI: 10.1126/scitranslmed.abf7872
99. Perrotta F., Matera M. G., Cazzola M., Bianco A.. **Severe respiratory SARS-CoV2 infection: does ACE2 receptor matter?**. (2020) **168**. DOI: 10.1016/j.rmed.2020.105996
100. Pirone L., del Gatto A., di Gaetano S., Saviano M., Capasso D., Zaccaro L.. **A multi-targeting approach to fight sars-cov-2 attachment.**. (2020) **7**. DOI: 10.3389/FMOLB.2020.00186/BIBTEX
101. Plattner V. E., Germann B., Neuhaus W., Noe C. R., Gabor F., Wirth M.. **Characterization of two blood–brain barrier mimicking cell lines: distribution of lectin-binding sites and perspectives for drug delivery.**. (2010) **387** 34-41. DOI: 10.1016/j.ijpharm.2009.11.030
102. Plummer E. M., Manchester M.. **Endocytic uptake pathways utilized by CPMV nanoparticles.**. (2013) **10** 26-32. DOI: 10.1021/mp300238w
103. Puntmann V. O., Carerj M. L., Wieters I., Fahim M., Arendt C., Hoffmann J.. **Outcomes of cardiovascular magnetic resonance imaging in patients recently recovered from coronavirus disease 2019 (COVID-19).**. (2020) **5**. DOI: 10.1001/jamacardio.2020.3557
104. Qiao J., Li W., Bao J., Peng Q., Wen D., Wang J.. **The expression of SARS-CoV-2 receptor ACE2 and CD147, and protease TMPRSS2 in human and mouse brain cells and mouse brain tissues.**. (2020) **533** 867-871. DOI: 10.1016/j.bbrc.2020.09.042
105. Rhea E. M., Logsdon A. F., Hansen K. M., Williams L. M., Reed M. J., Baumann K. K.. **The S1 protein of SARS-CoV-2 crosses the blood–brain barrier in mice.**. (2021) **24** 368-378. DOI: 10.1038/s41593-020-00771-8
106. Rubovitch V., Zilberstein Y., Chapman J., Schreiber S., Pick C. G.. **Restoring GM1 ganglioside expression ameliorates axonal outgrowth inhibition and cognitive impairments induced by blast traumatic brain injury.**. (2017) **7**. DOI: 10.1038/srep41269
107. Rusnati M., Tanghetti E., Urbinati C., Tulipano G., Marchesini S., Ziche M.. **Interaction of fibroblast growth factor-2 (FGF-2) with free gangliosides: biochemical characterization and biological consequences in endothelial cell cultures.**. (1999) **10**. DOI: 10.1091/MBC.10.2.313
108. Rusnati M., Urbinati C., Tanghetti E., Dell’Era P., Lortat-Jacob H., Presta M.. **Cell membrane GM 1 ganglioside is a functional coreceptor for fibroblast growth factor 2.**. (2002) **99** 4367-4372. DOI: 10.1073/pnas.072651899
109. Sachinidis A., Kraus R., Seul C., Meyer zu Brickwedde M. K., Schulte K., Ko Y.. **Gangliosides GM1, GM2 and GM3 inhibit the platelet-derived growth factor-induced signalling transduction pathway in vascular smooth muscle cells by different mechanisms.**. (1996) **71** 79-88. PMID: 8884181
110. Sachinidis A., Seul C., Gouni-Berthold I., Seewald S., Ko Y., Vetter H.. **Cholera toxin treatment of vascular smooth muscle cells decreases smooth muscle α-actin content and abolishes the platelet-derived growth factor-BB-stimulated DNA synthesis.**. (2000) **130** 1561-1570. DOI: 10.1038/sj.bjp.0703480
111. Saleki K., Banazadeh M., Saghazadeh A., Rezaei N.. **The involvement of the central nervous system in patients with COVID-19.**. (2020) **31** 453-456. DOI: 10.1515/revneuro-2020-0026
112. Sasaki N., Toyoda M.. **Vascular diseases and gangliosides.**. (2019) **20**. DOI: 10.3390/IJMS20246362
113. Schengrund C.-L.. **Gangliosides: glycosphingolipids essential for normal neural development and function.**. (2015) **40** 397-406. DOI: 10.1016/j.tibs.2015.03.007
114. Schneider J. S., Sendek S., Daskalakis C., Cambi F.. **GM1 ganglioside in Parkinson’s disease: results of a five year open study.**. (2010) **292** 45-51. DOI: 10.1016/j.jns.2010.02.009
115. Shapiro Ben David S., Potasman I., Rahamim-Cohen D.. **Rate of recurrent Guillain-Barré syndrome after mRNA COVID-19 vaccine BNT162b2.**. (2021) **78** 1409-1411. DOI: 10.1001/jamaneurol.2021.3287
116. Shi D., Sheng A., Chi L.. **Glycosaminoglycan-protein interactions and their roles in human disease.**. (2021) **8**. DOI: 10.3389/fmolb.2021.639666
117. Sipione S., Monyror J., Galleguillos D., Steinberg N., Kadam V.. **Gangliosides in the brain: physiology, pathophysiology and therapeutic applications.**. (2020) **14**. DOI: 10.3389/fnins.2020.572965
118. Sui Z.-H., Xu H., Wang H., Jiang S., Chi H., Sun L.. **Intracellular trafficking pathways of edwardsiella tarda: from Clathrin- and Caveolin-mediated endocytosis to endosome and lysosome.**. (2017) **7**. DOI: 10.3389/fcimb.2017.00400
119. Sun X.-L.. **The role of cell surface sialic acids for SARS-CoV-2 infection.**. (2021) **31** 1245-1253. DOI: 10.1093/glycob/cwab032
120. Suzuki Y.. **Sialobiology of influenza: molecular mechanism of host range variation of influenza viruses.**. (2005) **28** 399-408. DOI: 10.1248/bpb.28.399
121. Svennerholm L.. **Chromatographlc separation of human brain gangliosides.**. (1963) **10** 613-623. DOI: 10.1111/j.1471-4159.1963.tb08933.x
122. Svennerholm L., Gottfries C.-G.. **Membrane lipids, selectively diminished in Alzheimer brains, suggest synapse loss as a primary event in early-onset form (Type I) and demyelination in late-onset form (Type II).**. (2008) **62** 1039-1047. DOI: 10.1046/j.1471-4159.1994.62031039.x
123. Svennerholm L., Bråne G., Karlsson I., Lekman A., Ramström I., Wikkelsö C.. **Alzheimer Disease – effect of continuous intracerebroventricular treatment with GM1 ganglioside and a systematic activation programme.**. (2002) **14** 128-136. DOI: 10.1159/000063604
124. Tian F., Tong B., Sun L., Shi S., Zheng B., Wang Z.. **N501Y mutation of spike protein in SARS-CoV-2 strengthens its binding to receptor ACE2.**. (2021) **10**. DOI: 10.7554/eLife.69091
125. Tortorici M. A., Walls A. C., Lang Y., Wang C., Li Z., Koerhuis D.. **Structural basis for human coronavirus attachment to sialic acid receptors.**. (2019) **26** 481-489. DOI: 10.1038/s41594-019-0233-y
126. Varki A.. **Glycan-based interactions involving vertebrate sialic-acid-recognizing proteins.**. (2007) **446** 1023-1029. DOI: 10.1038/nature05816
127. Vellozzi C., Iqbal S., Broder K.. **Guillain-Barre syndrome, influenza, and influenza vaccination: the epidemiologic evidence.**. (2014) **58** 1149-1155. DOI: 10.1093/cid/ciu005
128. Volz E., Hill V., McCrone J. T., Price A., Jorgensen D., O’Toole Á. **Evaluating the effects of SARS-CoV-2 spike mutation D614G on transmissibility and pathogenicity.**. (2021) **184** 64-75.e11. DOI: 10.1016/j.cell.2020.11.020
129. Vorbrodt A. W.. **Changes in the distribution of endothelial surface glycoconjugates associated with altered permeability of brain micro-blood vessels.**. (1986) **70** 103-111. DOI: 10.1007/BF00691427
130. Wang H., Liu W., Yu F., Lu L.. **Disruption of clathrin-dependent trafficking results in the failure of grass carp reovirus cellular entry.**. (2016) **13**. DOI: 10.1186/s12985-016-0485-7
131. Wang J., Yu R. K.. **Interaction of ganglioside GD3 with an EGF receptor sustains the self-renewal ability of mouse neural stem cells in vitro.**. (2013) **110** 19137-19142. DOI: 10.1073/PNAS.1307224110/SUPPL_FILE/PNAS.201307224SI.PDF
132. Wang Y., Liu S., Liu H., Li W., Lin F., Jiang L.. **SARS-CoV-2 infection of the liver directly contributes to hepatic impairment in patients with COVID-19.**. (2020) **73** 807-816. DOI: 10.1016/j.jhep.2020.05.002
133. Weigel P. H., Yik J. H. N.. **Glycans as endocytosis signals: the cases of the asialoglycoprotein and hyaluronan/chondroitin sulfate receptors.**. (2002) **1572** 341-363. DOI: 10.1016/s0304-4165(02)00318-5
134. Weisblum Y., Schmidt F., Zhang F., DaSilva J., Poston D., Lorenzi J. C. C.. **Escape from neutralizing antibodies by SARS-CoV-2 spike protein variants.**. (2020) **9**. DOI: 10.7554/eLife.61312
135. WHO Coronavirus (COVID-19) Dashboard (n.d.). WHO Coronavirus (COVID-19) Dashboard | WHO Coronavirus (COVID-19) Dashboard With Vaccination Data. Available online at: https://covid19.who.int/
(accessed February 3, 2022).. (n.d.)
136. Wu G., Lu Z.-H., Obukhov A. G., Nowycky M. C., Ledeen R. W.. **Induction of calcium influx through TRPC5 channels by cross-linking of GM1 ganglioside associated with 5 1 integrin initiates neurite outgrowth.**. (2007) **27** 7447-7458. DOI: 10.1523/JNEUROSCI.4266-06.2007
137. Xie X., Liu Y., Liu J., Zhang X., Zou J., Fontes-Garfias C. R.. **Neutralization of SARS-CoV-2 spike 69/70 deletion, E484K and N501Y variants by BNT162b2 vaccine-elicited sera.**. (2021) **27** 620-621. DOI: 10.1038/s41591-021-01270-4
138. Zhang W., Krafft P. R., Wang T., Zhang J. H., Li L., Tang J.. **Pathophysiology of ganglioside GM1 in ischemic stroke: ganglioside GM1: a critical review.**. (2019) **28** 657-661. DOI: 10.1177/0963689718822782
139. Zhao H., Li X., Li G., Sun B. O., Ren L., Zhao C.. **Protective effects of monosialotetrahexosylganglioside sodium on H2O2-induced human vascular endothelial cells.**. (2015) **10** 947-953. DOI: 10.3892/etm.2015.2603
140. Zhong P., Xu J., Yang D., Shen Y., Wang L., Feng Y.. **COVID-19-associated gastrointestinal and liver injury: clinical features and potential mechanisms.**. (2020) **5**. DOI: 10.1038/s41392-020-00373-7
|
---
title: In Search of the Hyperglycemic Threshold Required to Induce Growth Hormone
(GH) Suppression
journal: Cureus
year: 2023
pmcid: PMC9980919
doi: 10.7759/cureus.34463
license: CC BY 3.0
---
# In Search of the Hyperglycemic Threshold Required to Induce Growth Hormone (GH) Suppression
## Abstract
Introduction According to the 2014 Endocrine Society Clinical Practice Guideline on acromegaly, the confirmation of acromegaly diagnosis is established by finding a lack of suppression of growth hormone (GH) to < 1 ug/L following documented hyperglycemia during an oral glucose tolerance test. However, in this setting, the concept of hyperglycemia has never been clearly defined.
Objective This study aimed to define the hyperglycemic threshold required to induce GH suppression.
Methods We retrieved the glycemia profile of 44 individuals after a standard 2-h 75g oral glucose tolerance test prescribed to assess GH suppression and performed a comprehensive analysis of two subgroups of individuals (28 reaching GH suppression and 16 in whom GH suppression was not observed). All of the data were analyzed with the program Graph Pad Prism. Differences between means were assessed by Student’s unpaired t-test or Mann-Whitney U test as deemed appropriate. Fisher’s exact test was used for categorical variables.
Results Individuals in G1 and G2 were different only for the median basal GH and median IGF-1.
No significant differences in terms of the prevalence of diabetes and prediabetes were found. The glucose peak was achieved earlier in the group that reached GH suppression.
The median of the highest glucose values of both subgroups was not different. A correlation between peak and baseline glucose value was found only among those in whom GH suppression was reached. Among these, the median glucose peak (P50) was 177 mg/dl, whereas the 75th percentile (P75) and 25th percentile (P25) were 199 mg/dl and 120 mg/dl, respectively.
Conclusion Considering that $75\%$ of those in whom GH suppression was observed after an oral glucose overload test reached blood glucose values above 120 mg/dl, we propose to use this value as the blood glucose threshold for inducing GH suppression. In light of our results, whenever GH suppression is not observed; and the highest glycemic value is below 120 mg/dl, it might be useful to repeat the test prior to any conclusion.
## Introduction
Acromegaly is a disorder caused by chronic growth hormone (GH) hypersecretion leading to IGF-1 over-production. Screening of this condition is performed by an IGF-1 test since GH is normally released in pulses and random levels vary widely. Investigation of GH suppression during an Oral Glucose Tolerance Test (OGTT) 75 g for two hours is recommended as a confirmatory test for those cases with elevated age- and sex-adjusted IGF-1 levels. In acromegaly, suppression failure occurs, and there may be a paradoxical rise in GH in response to the glucose challenge [1].
Recommendations propose confirmation of the diagnosis by finding a lack of suppression of GH to < 1 ug/L following documented hyperglycemia during an OGTT [2].
In a known acromegalic, cure reassessment after pituitary surgery may be indicated for those with a random GH at 12 weeks or later greater than 1 µg/L. The value to define disease control is less well established, but cut-offs of 1 ug/L and 0,4 ug/L have been suggested [3,4].
Recent evidence suggests that GH nadir concentrations, as measured by a modern GH assay, are much lower than the current cut-offs mentioned in guidelines for the diagnosis and follow-up of acromegaly [5]. Moreover, several studies have shown that patients with mild acromegaly may have nadir GH lower than 1 μg/L [6-8].
Independently of the GH nadir considered, hyperglycemia is the necessary condition to reach GH suppression. A consensus has never been achieved in defining at which glycemic threshold occurs GH suppression.
Aiming to address this issue, we reviewed the results of OGTT tests performed from 2016 to 2021 and analyzed the glycemic profile of those individuals in whom GH suppression was reached.
## Materials and methods
We retrieved all GH suppression tests performed in the period encompassing the years 2016 to 2021. In parallel, the IGF-1 levels preceding the tests were also collected.
Standard 2-h 75g OGTT was performed in each subject after overnight fasting. Blood was drawn through an indwelling venous cannula at -15, 0, 30, 60, 90, and 120 minutes, after oral glucose administration, for plasma glucose and GH levels. The patients remained at rest throughout the test. The diagnosis of diabetes or pre-diabetes was not known before OGTT was performed.
Basal glucose and GH were defined as the mean of values registered at -15 minutes and 0 corresponding the latter to the time immediately before glucose administration.
Tests in whom the glucose load elicited a paradoxical increase in GH were excluded.
Impaired fasting glucose is defined as fasting plasma glucose of 100-125 mg/dL. Impaired glucose tolerance is defined as 2-h plasma glucose 140-199 mg/dL after OGTT. Diabetes Mellitus was defined as having either fasting plasma glucose ≥ 126 mg/dL or 2-h OGTT plasma glucose ≥ 200 mg/dL.
Serum GH was measured by an electrochemiluminescence immunoassay (ROCHE - Elecsys hGH, lower detection limit 0,030 ng/ml, according to the manufacturer's indications).
IGF-1 was measured by an immunoassay (ROCHE - Elecsys IGF-1, lower detection limit 7 ng/ml, highest measure value without dilution 1600 ng/ml, according to the manufacturer's indications).
Tests were divided into 2 subgroups: G1 - GH nadir more than 1 ug/L following oral glucose administration and G2 - GH nadir less than 1 ug/L following an oral glucose load. Thereafter, glycemia percentiles of individuals within this latter group were determined.
All of the data were analyzed with the program Graph Pad Prism version 9.3.1 (GraphPad Software). Quantitative variables were tested for Gaussian distribution with the Shapiro-Wilk test. Differences between means were assessed by Student’s unpaired t-test or Mann-Whitney U test as deemed appropriate. Fisher’s exact test was used for categorical variables. Receiver Operator Characteristic (ROC) analysis was used to evaluate the diagnostic performance of glycemia as a predictor of the outcome in terms of GH suppression. The area under the ROC curve (AUC) was used to assess overall diagnostic accuracy.
A value of $p \leq 0.05$ was accepted as denoting statistical significance.
## Results
We identified 46 tests. Two tests in which a paradoxical increase was observed were excluded since we were searching for the necessary glycemic threshold to induce GH suppression.
G1 included 16 tests performed on 10 women and 6 men. G2 included 28 tests performed on 20 women and 8 men. The mean age in G1 and G2 was 50±18.6 and 54±12 years, respectively.
The review of clinical files of individuals in each group allowed the integration of clinical, laboratory, and semiological data conducting to the conclusion that all but one patient in G1 had a diagnosis of acromegaly (10 already submitted to surgery, five awaiting surgery); the exception was a patient with a macro prolactinoma submitted to the suppression test because of an abnormal basal GH despite a normal IGF-1.
G2 included 14 patients with past surgery for acromegaly and 14 individuals in whom the clinical and laboratory diagnosis of acromegaly was doubtful.
Individuals in G1 and G2 were different only for the median basal GH (4.83 ng/ml versus 1 ng/ml) and median IGF-1 (300 ng/ml versus 227 ng/ml).
As summarized in Table 1, the median glucose peak was not significantly different between groups. Moreover, a ROC analysis revealed an AUC of 0.5859.
**Table 1**
| Unnamed: 0 | G1 (GH nadir >1 ug/L n = 16 | G2(GH nadir < 1 ug/L n = 28 | p |
| --- | --- | --- | --- |
| Gender | | | ns |
| Female | 10 | 20 | |
| Male | 6 | 8 | |
| Mean Age (years) | 50±18.6 | 54±12 | ns |
| Baseline Glucose (mg/dl/mmol/L) | | | |
| Mean | 95±12.8/5.3±0.71 | 95±15/5.3±0.83 | ns |
| Minimum | 78/4.3 | 63/3.5 | |
| Maximum | 113/6.3 | 130/7.2 | |
| Glucose peak after 75 g oral glucose (mg/dl/mmol/L) | | | |
| Median | 184 | 177 | ns |
| Ratio between the glucose peak and the baseline value | | | |
| Mean | 2.03±0.43 | 1.9±0.49 | ns |
| Minimum | 1.57 | 1.01 | |
| Maximum | 2.88 | 3.2 | |
| Oral Glucose Test | | | ns |
| Diabetes | 3 | 4 | |
| Pre Diabetes | 6 | 8 | |
| Baseline GH (ng/ml = ug/L) | | | |
| Median | 4.83 | 1.05 | 0.0001 |
| IGF1 (ng/ml/nmol/L) | | | |
| Median | 300/29.74 | 227/39.3 | 0.0072 |
No significant differences in terms of the prevalence of diabetes and prediabetes were found. Prediabetes was observed in 37,$5\%$ and 28,$6\%$ in G1 and G2, respectively, and diabetes mellitus in 18,$8\%$ and 14,$3\%$, respectively.
The glucose peak was achieved earlier in group G2 (Figure 1): $75\%$ of the patients reached the highest value at or < 60 minutes.
**Figure 1:** *Time to glucose peak.*
Moreover, only in G2 a correlation ($r = 0.7185$, $P \leq 0.0001$) between the highest value and the basal value of glucose was observed (Figure 2).
**Figure 2:** *Correlation between glucose peak and baseline glucose observed in G2.*
Among individuals that reached GH suppression, the median glucose value (P50) was 177 mg/dl (9.8 mmol/L), whereas the 75th percentile (P75) and 25th percentile (P25) were 199 mg/dl (11 mmol/L) and 120.25 mg/dl (6.7 mmol/L), respectively.
Based on the suppression test, among the 24 acromegalic patients already submitted to surgery, 14 ($58.3\%$) were considered in remission.
## Discussion
The main metabolic consequence of acromegaly is insulin resistance, which may progress to diabetes mellitus. Alexopoulou O et al. [ 9] reported a prevalence of impaired fasting glycemia or glucose tolerance of $26\%$ and a prevalence of diabetes mellitus of $28\%$ at diagnosis of acromegaly. In the current study, the prevalence of diabetes mellitus in acromegalic patients was 18,$8\%$, and the prevalence of prediabetes was 37,$8\%$. No difference was found in the prevalence of diabetes and prediabetes among patients with active and inactive acromegaly. However, time to glucose peak occurred in the majority of individuals from G2 (patients in remission of acromegaly or normal individuals) at or before 60 minutes, whereas in patients from G1 (with active acromegaly), tended to occur later. The latter pattern is likely to correspond to insulin resistance [10].
Growth hormone secretion is regulated by two hypothalamic hormones: GH-releasing hormone (GHRH) and somatotropin release-inhibiting factor (SRIF). Ghrelin, a gut-derived peptide, has been considered the third regulator of GH secretion with a stimulatory action. Hyperglycemia suppresses ghrelin [11] and is associated with a somatostatin release into the hypophyseal portal blood suppressing GH levels [5].
Oral glucose administration is the standard method for assessing inhibitory control of GH release. What lacks clarification is the level of glucose that should be reached to confidently conclude that there is no GH suppression.
The median glucose peak was not different between groups with or without GH suppression. However, whereas a correlation between the peak and the baseline value of glucose was documented in those in whom the outcome was GH suppression, it was not observed in those who failed to suppress GH. The variable individual degree of insulin resistance among those in the latter group might have contributed to this difference.
There was concordance between the response to OGTT and clinical data. Except for the case of a patient presenting a pituitary macro adenoma without acromegalic phenotype and high levels of prolactin consistent with the diagnosis of macro prolactinoma. Moreover, the patient had a normal IGF-1 and elevated GH that was not suppressed by hyperglycemia. Although we cannot exclude a mixed somato-lactotroph tumor, alternatively, considering that venepuncture has been regarded as a psychological and physiological stressor on the one hand and on the other that stress can elicit GH secretion, one might speculate to what extent this fact can counteract the inhibitory effect of hyperglycemia.
Glycemia was not a predictor of outcome in terms of GH suppression. The absence of GH suppression observed in G1 was independent of blood glucose, reinforcing the hypothesis that it is an intrinsic feature of autonomous GH secretion.
Furthermore, we observed that the median glucose peak (P50) among individuals who reached GH suppression was 177 mg/dl, whereas the 75th percentile (P75) and 25th percentile (P25) were 199 mg/dl and 120 mg/dl, respectively. Based on this observation and taking into account that 126 mg/dl is the fasting glycemic value accepted to establish the diagnosis of diabetes mellitus, we consider that the value of 120 mg/dl accomplishes the condition of hyperglycemia and may be proposed as the threshold for GH suppression. Further studies involving a larger number of cases are desirable to corroborate current results.
The exact mechanisms underlying GH suppression are not deeply understood. A nadir GH concentration below 1 ug/L was observed in normal individuals either during an OGTT test or saline infusion if GH secretion was evaluated over 180 minutes [12]. Thus, the observation of a GH nadir below 1 ug/L among normal individuals may result from a spontaneous variability and not as a consequence of a glucose overload. On the other hand, other mechanisms, including the release of neuropeptides associated with food ingestion, cannot be ruled out as potential contributors to GH suppression, justifying that in some cases, GH suppression may occur for values below 120 mg/dl as observed in a quarter of the cases of the present series.
## Conclusions
Failure of suppression following documented hyperglycemia has been regarded as diagnostic for acromegaly. However, the term hyperglycemia has never been accurately defined. As a consequence, the interpretation of GH suppression tests is not always straightforward because there are doubts as to whether the theoretical assumption of hyperglycemia has been reached.
In the present study, $75\%$ of the patients in whom GH suppression was achieved had a maximum glycemic value of 120 mg/dl or more. Only in a minority GH suppression was observed despite a maximum blood glucose value of less than 120 mg/dl. Therefore, we propose to use this glycemic threshold to evaluate GH suppression following an oral glucose tolerance test. In light of our results, whenever GH suppression is not observed; and the highest glycemic value is below 120 mg/dl, it might be useful to repeat the test before any conclusion. Due to the small number of cases included, further studies are needed to confirm our findings.
## References
1. Scaroni C, Albiger N, Daniele A. **Paradoxical GH increase during OGTT is associated with first-generation somatostatin analog responsiveness in acromegaly**. *J Clin Endocrinol Metab* (2019) **104** 856-862. PMID: 30285115
2. Katznelson L, Laws ER Jr, Melmed S, Molitch ME, Murad MH, Utz A, Wass JA. **Acromegaly: an endocrine society clinical practice guideline**. *J Clin Endocrinol Metab* (2014) **99** 3933-3951. PMID: 25356808
3. Giustina A, Chanson P, Bronstein MD. **A consensus on criteria for cure of acromegaly**. *J Clin Endocrinol Metab* (2010) **95** 3141-3148. PMID: 20410227
4. Melmed S, Bronstein MD, Chanson P. **A consensus statement on acromegaly therapeutic outcomes**. *Nat Rev Endocrinol* (2018) **14** 552-561. PMID: 30050156
5. Hage M, Kamenický P, Chanson P. **Growth hormone response to oral glucose load: from normal to pathological conditions**. *Neuroendocrinology* (2019) **108** 244-255. PMID: 30685760
6. Dimaraki EV, Jaffe CA, DeMott-Friberg R, Chandler WF & Barkan AL. **Acromegaly with apparently normal GH secretion: implications for diagnosis and follow-up**. *J Clin Endocrinol* (2002) **87** 3537-3542
7. Ribeiro-Oliveira A Jr, Faje AT, Barkan AL. **Limited utility of oral glucose tolerance test in biochemically active acromegaly**. *Eur J Endocrinol* (2011) **164** 17-22. PMID: 20926592
8. Zahr R, Fleseriu M. **Updates in diagnosis and treatment of acromegaly**. *Eur Endocrinol* (2018) **14** 57-61. PMID: 30349595
9. Alexopoulou O, Bex M, Kamenicky P, Mvoula AB, Chanson P, Maiter D. **Prevalence and risk factors of impaired glucose tolerance and diabetes mellitus at diagnosis of acromegaly: a study in 148 patients**. *Pituitary* (2014) **17** 81-89. PMID: 23446424
10. Chung ST, Ha J, Onuzuruike AU. **Time to glucose peak during an oral glucose tolerance test identifies prediabetes risk**. *Clin Endocrinol (Oxf)* (2017) **87** 484-491. PMID: 28681942
11. Nakagawa E, Nagaya N, Okumura H. **Hyperglycaemia suppresses the secretion of ghrelin, a novel growth-hormone-releasing peptide: responses to the intravenous and oral administration of glucose**. *Clin Sci (Lond)* (2002) **103** 325-328. PMID: 12193159
12. Grottoli S, Razzore P, Gaia D. **Three-hour spontaneous GH secretion profile is as reliable as oral glucose tolerance test for the diagnosis of acromegaly**. *J Endocrinol Invest* (2003) **26** 123-127. PMID: 12739738
|
---
title: 'International multicenter study comparing COVID-19 in patients with cancer
to patients without cancer: Impact of risk factors and treatment modalities on survivorship'
authors:
- Issam I Raad
- Ray Hachem
- Nigo Masayuki
- Tarcila Datoguia
- Hiba Dagher
- Ying Jiang
- Vivek Subbiah
- Bilal Siddiqui
- Arnaud Bayle
- Robert Somer
- Ana Fernández Cruz
- Edward Gorak
- Arvinder Bhinder
- Nobuyoshi Mori
- Nelson Hamerschlak
- Samuel Shelanski
- Tomislav Dragovich
- Yee Elise Vong Kiat
- Suha Fakhreddine
- Abi Hanna Pierre
- Roy F Chemaly
- Victor Mulanovich
- Javier Adachi
- Jovan Borjan
- Fareed Khawaja
- Bruno Granwehr
- Teny John
- Eduardo Yepez Yepez
- Harrys A Torres
- Natraj Reddy Ammakkanavar
- Marcel Yibirin
- Cielito C Reyes-Gibby
- Mala Pande
- Noman Ali
- Raniv Dawey Rojo
- Shahnoor M Ali
- Rita E Deeba
- Patrick Chaftari
- Takahiro Matsuo
- Kazuhiro Ishikawa
- Ryo Hasegawa
- Ramón Aguado-Noya
- Alvaro Garcia García
- Cristina Traseira Puchol
- Dong Gun Lee
- Monica Slavin
- Benjamin Teh
- Cesar A Arias
- Dimitrios P Kontoyiannis
- Alexandre E Malek
- Anne-Marie Chaftari
journal: eLife
year: 2023
pmcid: PMC9981148
doi: 10.7554/eLife.81127
license: CC BY 4.0
---
# International multicenter study comparing COVID-19 in patients with cancer to patients without cancer: Impact of risk factors and treatment modalities on survivorship
## Abstract
### Background:
In this international multicenter study, we aimed to determine the independent risk factors associated with increased 30 day mortality and the impact of cancer and novel treatment modalities in a large group of patients with and without cancer with COVID-19 from multiple countries.
### Methods:
We retrospectively collected de-identified data on a cohort of patients with and without cancer diagnosed with COVID-19 between January and November 2020 from 16 international centers.
### Results:
We analyzed 3966 COVID-19 confirmed patients, 1115 with cancer and 2851 without cancer patients. Patients with cancer were more likely to be pancytopenic and have a smoking history, pulmonary disorders, hypertension, diabetes mellitus, and corticosteroid use in the preceding 2 wk (p≤0.01). In addition, they were more likely to present with higher inflammatory biomarkers (D-dimer, ferritin, and procalcitonin) but were less likely to present with clinical symptoms (p≤0.01). By country-adjusted multivariable logistic regression analyses, cancer was not found to be an independent risk factor for 30 day mortality ($$p \leq 0.18$$), whereas lymphopenia was independently associated with increased mortality in all patients and in patients with cancer. Older age (≥65y) was the strongest predictor of 30 day mortality in all patients (OR = 4.47, $p \leq 0.0001$). Remdesivir was the only therapeutic agent independently associated with decreased 30 day mortality (OR = 0.64, $$p \leq 0.036$$). Among patients on low-flow oxygen at admission, patients who received remdesivir had a lower 30 day mortality rate than those who did not (5.9 vs $17.6\%$; $$p \leq 0.03$$).
### Conclusions:
Increased 30 day all-cause mortality from COVID-19 was not independently associated with cancer but was independently associated with lymphopenia often observed in hematolgic malignancy. Remdesivir, particularly in patients with cancer receiving low-flow oxygen, can reduce 30 day all-cause mortality.
### Funding:
National Cancer Institute and National Institutes of Health.
## Introduction
The COVID-19 pandemic has challenged the health care system worldwide and has spread to more than 200 countries, causing hundreds of millions confirmed cases and several million deaths (Johns Hopkins Coronavirus Resource Center, 2021).
Data from multiple studies have shown consistently that older age and comorbidities such as cardiovascular disease, diabetes mellitus (DM), hypertension, and chronic obstructive pulmonary disease (COPD) have been associated with severe illness and increased mortality (Zhou et al., 2020).
Several studies on COVID-19 mortality suggested that patients with cancer had poor outcomes (Kuderer et al., 2020; Rivera et al., 2020; He et al., 2020; Albiges et al., 2020; Robilotti et al., 2020; Lee et al., 2020; Mehta et al., 2020; Tian et al., 2020; Lunski et al., 2021; Sun et al., 2020). Many of these studies included only patients with cancer and did not have a comparator group of patients without cancer (Kuderer et al., 2020; Rivera et al., 2020; He et al., 2020; Albiges et al., 2020; Robilotti et al., 2020). Other studies compared COVID-19 mortality between patients with and without cancer and found that patients with cancer had worse outcomes (Mehta et al., 2020; Tian et al., 2020; Lunski et al., 2021; Sun et al., 2020; Rüthrich et al., 2021). However, all of these comparative studies included a relatively small number of patients with cancer and were restricted to a particular country, which limits their generalizability to patients with cancer worldwide.
Given that many of the therapeutic studies on COVID-19 were conducted in patients without cancer and given the poor outcomes of COVID-19 reported in patients with cancer (Mehta et al., 2020; Tian et al., 2020; Lunski et al., 2021; Sun et al., 2020; Rüthrich et al., 2021; Beigel et al., 2020; Salazar et al., 2021; Libster et al., 2021; Janiaud et al., 2021; Horby et al., 2021), we aimed to conduct a large multicenter study to compare the impact of these various treatments on the outcome in patients with cancer vs patients without cancer.
Therefore, given the global spread of the COVID-19 pandemic and the worldwide prevalence of cancer, we undertook this international initiative that included 16 centers from five continents to study and compare the clinical course, risk factors, and treatment modalities impact on outcomes of COVID-19 in patients with cancer vs patients without cancer on a worldwide basis.
## Study design and participants
This was a retrospective international multicenter study that included all patients diagnosed with COVID-19 by RT-PCR for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) at the site or an outside facility between January 4, 2020 and November 15, 2020.
The study involved 16 centers from 9 countries, 8 centers in the United States and 1 each in Australia, Brazil, France, Japan, Lebanon, Singapore, South Korea, and Spain. Patients were divided into two groups: patients without cancer and those with cancer diagnosed or treated within a year before the diagnosis of COVID-19.
## Multicenter collaboration and data collection
The University of Texas MD Anderson Cancer Center was the coordinating center that designed the study, built the electronic case report form, and collected de-identified patient information from all participating centers using the secure Research Electronic Data Capture platform. We reviewed each patient’s electronic hospital record and collected all needed data. This study was approved by the institutional review board at MD Anderson Cancer Center (Protocol# 2020–0437) and the institutional review boards of the collaborating centers. A patient waiver of informed consent was obtained.
The follow-up period was defined as 30 days after a diagnosis of COVID-19.
Neutropenia was defined as an absolute neutrophil count <500 cells/mL. Lymphopenia was defined as an absolute lymphocyte count (ALC) <500 cells/mL.
## Treatment modalities
For each patient, data were collected on COVID-19 treatments, including antimicrobial therapy and potential antiviral therapy. We evaluated patient’s oxygen requirement: low flow was defined as oxygen supplementation of ≤6 l/min through a nasal cannula or facemask and high flow included all other modalities of oxygen supplementation, including mechanical ventilation.
## Outcomes measures
The primary outcome of interest was 30 day mortality. Any death that occurred within 30 days after COVID-19 diagnosis was considered to be COVID-19-related, irrespective of other comorbidities that could have contributed to death. The secondary outcomes included mechanical ventilation, progression to lower respiratory tract infection, co-infection, and hospital readmissions within 30 days after COVID-19 diagnosis.
## Statistical analysis
Patient characteristics and outcomes were compared between COVID-19 patients with and without cancer. Categorical variables were compared using χ2 or Fisher’s exact test, and continuous variables were compared using Wilcoxon rank sum test. Logistic regression model was used to identify the factors that were independently associated with 30 day mortality, and the following factors were included in the analysis in addition to cancer status: patients’ demographic (including country) and clinical characteristics, medical history, and laboratory findings at diagnosis and treatment. First, univariable analysis of each factor was performed, but factors were not considered if more than $40\%$ of the data were missing. Next all the factors with p-values≤0.15 on their univariable analyses were included in a full logistic regression model and then the full model was reduced to the final model by backward elimination procedure so that all the factors remaining in the final model had p-values≤0.05 except cancer and country. Cancer was kept in the final model despite its p-value in order to evaluate its independent impact on mortality. The final model was country-adjusted to account for the treatment differences among the countries. For the above complete cases analysis, only patients with no missing data in any of the variables retained for the final regression model were included in the final model analysis. To investigate the impact of missing data on our primary data analysis, we performed a sensitivity analysis. We first estimated missing values using multiple imputation technique and then performed a similar multivariable analysis based on the imputed datasets. Lastly, the analysis results were compared between the complete-case and multiple imputation analyses. Logistic regression model was also used to identify the independent predictors of mortality among patients with and without cancer, respectively, and similar sensitivity analysis was performed for each group as well. In cancer patients, the mortality rates of four different cancers were estimated and compared: hematological malignancy, lung cancer, and non-lung solid tumors with and without metastasis. χ2 or Fisher’s exact test was used for comparisons. If a significant result ($p \leq 0.05$) was detected, pairwise comparisons were performed with α levels adjusted using Holm’s sequential Bonferroni adjustment to control type I error. In addition, the associations between mortality and certain therapeutic agents including remdesivir, steroids, and convalescent plasma, as well as the impact of their treatment timing on mortality, were also evaluated using χ2 or Fisher’s exact test. All statistical tests were two-sided with a significance level of 0.05, except the pairwise comparisons with the α adjustment. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC, USA).
## Results
A total of 4015patients diagnosed with COVID-19 by PCR were included in the study. After excluding 18 patients who had missing demographics and 31 patients younger than 18y, we evaluated 3966 COVID-19 patients: 2851 without cancer and 1115 with cancer (see Figure 1 for a consort flow diagram of patient attrition in analyses).
**Figure 1.:** *Consort diagram of patient attrition.*
## Demographics and clinical characteristics of patients with and without cancer
Patient characteristics are presented in Table 1. Compared to patients without cancer, patients with cancer were older (median age, 61 vs 50 y; $p \leq 0.0001$); more likely to have a smoking history (38 vs $17\%$; $p \leq 0.0001$), pulmonary disorder (27 vs $21\%$; $p \leq 0.001$), hypertension (49 vs $36\%$, $p \leq 0.0001$), DM (27 vs $23\%$, $$p \leq 0.01$$), or corticosteroid use in the 2 wk preceding COVID-19 diagnosis (17 vs $4\%$, $p \leq 0.0001$); and less likely to present with clinical symptoms, including cough (46 vs $65\%$; $p \leq 0.0001$), fever (45 vs $66\%$; $p \leq 0.0001$), and shortness of breath (35 vs $48\%$; $p \leq 0.0001$).
**Table 1.**
| Characteristic | Without cancer (n=2851)N (%) | With cancer(n=1115)N (%) | p-value |
| --- | --- | --- | --- |
| Demographic and baseline clinical characteristics | | | |
| Age (years), median (range) | 50 (18–100) | 61 (18–100) | <0.0001 |
| Sex, male | 1335 (47) | 506 (45) | 0.41 |
| Race/ethnicity | | | <0.0001 |
| White | 720/2674 (27) | 534/916 (58) | |
| Black | 517/2674 (19) | 155/916 (17) | |
| Hispanic | 547/2674 (20) | 178/916 (19) | |
| Asian | 275/2674 (10) | 33/916 (4) | |
| Middle Eastern | 63/2674 (2) | 3/916 (0.3) | |
| Other | 552/2674 (21) | 13/916 (1) | |
| Prior pulmonary disorders | 431/2064 (21) | 275/1017 (27) | <0.001 |
| COPD/bronchiolitis obliterans | 175/2054 (9) | 75 (7) | 0.07 |
| Asthma | 178 (6) | 106 (10) | <0.001 |
| Obstructive sleep apnea | 98/2054 (5) | 89/948 (9) | <0.0001 |
| History of heart failure | 240/2036 (12) | 85/1098 (8) | <0.001 |
| History of ischemic heart disease | 226/2830 (8) | 94/1101 (9) | 0.57 |
| History of hypertension | 1020/2837 (36) | 546/1110 (49) | <0.0001 |
| History of diabetes mellitus | 659/2837 (23) | 299/1104 (27) | 0.01 |
| Current or previous smoker | 348/2055 (17) | 409/1066 (38) | <0.0001 |
| Corticosteroid treatment within 2wk prior to COVID-19 diagnosis | 40/1041 (4) | 189/1097 (17) | <0.0001 |
| Presenting symptoms | 981/1061 (92) | 813/1102 (74) | <0.0001 |
| Cough | 685/1061 (65) | 505/1102 (46) | <0.0001 |
| Fever | 698/1061 (66) | 498/1102 (45) | <0.0001 |
| Shortness of breath | 508/1061 (48) | 387/1102 (35) | <0.0001 |
| Chest pain | 100/1061 (9) | 66/1102 (6) | 0.003 |
| Headache | 133/1061 (13) | 80/1102 (7) | <0.0001 |
| Gastrointestinal symptoms | 148/1061 (14) | 104/1102 (9) | 0.001 |
| Loss of smell | 84/1061 (8) | 58/1102 (5) | 0.013 |
| Loss of taste | 78/1061 (7) | 55/1102 (5) | 0.022 |
| ICU admission | 225/1846 (12) | 141/1100 (13) | 0.62 |
| Abnormal laboratory values | | | |
| ANC <0.5K/μl | 2/1434 (0.1) | 30/419 (7) | <0.0001 |
| ALC <0.5K/μl | 487/1537 (32) | 225/463 (49) | <0.0001 |
| Platelet count<100K/μl | 187/1446 (13) | 151/387 (39) | <0.0001 |
| Hemoglobin<10g/dL | 437/1533 (29) | 283/601 (47) | <0.0001 |
| D-dimer, median (range), μg/ml | 1.51 (0.04–735.0) | 1.95 (0.25–93.24) | 0.013 |
| Ferritin, median (range), ng/ml | 823 (1.10–89672) | 1015 (1.5–100001) | <0.0001 |
| Procalcitonin, median (range), ng/ml | 0.21 (0.009–163.5) | 0.25 (0–101.8) | 0.008 |
| IL-6, median (range), pg/ml | 61 (0–5001) | 44 (0–7525) | 0.13 |
| Imaging findings | | | |
| New infiltrates | 151/622 (24) | 138/424 (33) | 0.003 |
| Ground-glass opacities | 487/622 (78) | 287/425 (68) | <0.0001 |
| Peripheral distribution of infiltrates | 253/622 (41) | 60/347 (17) | <0.0001 |
| Treatment | | | |
| Hydroxychloroquine | 475/2054 (23) | 196/1114 (18) | <0.001 |
| Azithromycin | 972/2054 (47) | 201/1114 (18) | <0.0001 |
| Remdesivir | 338 (12) | 97 (9) | 0.004 |
| Tocilizumab | 75/2054 (4) | 67 (6) | 0.002 |
| Convalescent plasma | 253/2054 (12) | 61/948 (6) | <0.0001 |
| Steroids | 879/2054 (43) | 192/1114 (17) | <0.0001 |
| Others* | 166/2054 (8) | 124/953 (13) | <0.0001 |
| Outcomes | | | |
| Co-infection after COVID-19 diagnosis | 158/1994 (8) | 116/1066 (11) | 0.006 |
| Multi-organ failure | 130/1052 (12) | 135/1096 (12) | 0.98 |
| Thrombotic complication | 56/1048 (5) | 48/1072 (5) | 0.36 |
| Discharged on supplemental oxygen among hospitalized | 90/776 (12) | 75/438 (17) | 0.007 |
| patients | | | |
| Hospital re-admission within 30 days of COVID-19 diagnosis | | | <0.0001 |
| No | 667/806 (83) | 318/481 (66) | |
| Yes | 60/806 (7) | 63/481 (13) | |
| Stayed in hospital (throughout 30 days) | 79/806 (10) | 100/481 (21) | |
| Death within 30 days of COVID-19 diagnosis | 226 (8) | 122 (11) | 0.003 |
Compared to patients without cancer, patients with cancer were more likely to present with neutropenia (7 vs $0.1\%$; $p \leq 0.0001$), lymphocytopenia (49 vs $32\%$; $p \leq 0.0001$), thrombocytopenia (39 vs $13\%$; $p \leq 0.0001$), and anemia (47 vs $29\%$; $p \leq 0.0001$) and had higher median levels of inflammatory biomarkers, including D-dimer (1.95 vs 1.51 μg/ml, $$p \leq 0.013$$), ferritin (1015 vs 823 ng/ml, $p \leq 0.0001$), and procalcitonin (0.25 vs 0.21 ng/ml, $$p \leq 0.008$$). On imaging studies (CT), patients with cancer were less likely than patients without cancer to be present with ground-glass opacities (68 vs $78\%$; $p \leq 0.0001$) or peripheral distribution of the infiltrates (17 vs $41\%$; $p \leq 0.0001$).
## Treatment and outcomes of patients with and without cancer
For most treatments for COVID-19 infection, patients without cancer were more likely to receive them than patients with cancer, including hydroxychloroquine (23 vs $18\%$, $p \leq 0.001$), azithromycin (47 vs $18\%$, $p \leq 0.0001$), remdesivir (12 vs $9\%$, $$p \leq 0.004$$), convalescent plasma (12 vs $6\%$, <0.0001), and steroids (43 vs $17\%$, $p \leq 0.0001$). However, patients with cancer were more likely to receive tocilizumab (6 vs $4\%$, $$p \leq 0.002$$). In addition, the treatments patients received for COVID-19 infection significantly differed among the countries (data not shown). The rate of COVID-related hospital admission was higher in the patients without cancer (62 vs $45\%$; $p \leq 0.0001$). Co-infections occurred more frequently in patients with cancer (11 vs $8\%$; $$p \leq 0.006$$). Among hospitalized patients, patients with cancer were more likely than patients without cancer to be discharged on supplemental oxygen (17 vs $12\%$; $$p \leq 0.007$$). Likewise, the rate of hospital readmission within 30 days was higher in patients with cancer (13 vs $7\%$; $p \leq 0.0001$). Furthermore, the mortality rate within 30 days was also significantly higher in patients with cancer by univariable analyses (11 vs $8\%$; $$p \leq 0.003$$; Table 1).
## Risk factors for death within 30 days after COVID-19 diagnosis
The independent predictors of 30 day mortality among all patients identified by the multivariable analysis are shown in Table 2. The multivariable complete-case analysis also showed that cancer was not an independent risk factor for 30 day mortality ($$p \leq 0.18$$; Table 2). Older age (≥65y) was the strongest predictor of 30 day mortality in all patients (OR = 4.47; $95\%$CI=3.27–6.11; $p \leq 0.0001$; Table 2) and in both patients with (OR = 6.64, $p \leq 0.0001$) and without cancer (OR = 4.91, $p \leq 0.0001$; Table 3).
Other independent risk factors for 30 day mortality in all patients included hypoxia at diagnosis (OR = 4.58; $p \leq 0.0001$), need for mechanical ventilation (OR = 2.20; $$p \leq 0.008$$), and presence of co-infection (OR = 1.83; $$p \leq 0.002$$; Table 2).
In patients with cancer, lower respiratory tract infection manifested by the presence of pulmonary infiltrates either at diagnosis or during the course of infection was a strong independent predictor of 30 day mortality (OR = 3.70; $95\%$ CI = 1.94–7.08; $p \leq 0.0001$; Table 3).
Among patients with cancer, the 30 day mortality rate was significantly higher in patients with lung cancer ($22\%$) than in patients with non-lung cancer solid tumors ($6\%$, $p \leq 0.0001$), including those with lung metastases ($7\%$, $p \leq 0.001$; Table 4). Patients with hematological malignancies had a significantly higher 30 day mortality than patients with non-lung cancer solid tumors (13 vs $6\%$, $p \leq 0.001$) but tended to have a lower mortality rate than patients with lung cancer (13 vs $22\%$, $$p \leq 0.07$$; Table 4). However, we did not find a significant difference in mortality between hematological malignancies and solid tumors by multivariable analysis ($$p \leq 0.30$$).
**Table 4.**
| Patient group | No. of patients | No. of patients.1 | No. (%) who died within 30 days† | No. (%) who died within 30 days†.1 |
| --- | --- | --- | --- | --- |
| Hematological malignancy | 283 | 283 | 37 (13) | 37 (13) |
| Transplant within 1y of COVID-19 diagnosis | 15 | 15 | 2 (13) | 2 (13) |
| Lymphoma or myeloma | 164 | 164 | 19 (12) | 19 (12) |
| Lymphocytic leukemia (ALL/CLL) | 44 | 44 | 6 (14) | 6 (14) |
| Myelocytic leukemia | 62 | 62 | 8 (13) | 8 (13) |
| Solid tumor * | 632 | 632 | 47 (7) | 47 (7) |
| Lung cancer | 64 | 64 | 14 (22) | 14 (22) |
| Metastatic non-lung cancer solid tumor | 261 | 261 | 17 (7) | 17 (7) |
| Non-metastatic non-lung cancer solid tumor | 307 | 307 | 16 (5) | 16 (5) |
By multivariable analysis, remdesivir was the only therapeutic agent independently associated with decreased 30 day all-cause mortality in all patients (OR = 0.64; $95\%$ CI = 0.42–0.97; $$p \leq 0.036$$; Table 2), and in patients with cancer (OR = 0.44; $95\%$ CI = 0.20–0.96; $$p \leq 0.04$$) as well (Table 3). However, in patients without cancer, remdesivir was not among the factors independently associated with mortality by multivariable analysis (Table 3).
Among patients on low-flow oxygen at admission, the mortality rate was lower among those who received remdesivir than among those who did not (5.9 [3 of 51] vs $17.6\%$ [68 of 387]; $$p \leq 0.03$$). However, among patients on high-flow oxygen at admission, there was no difference in the mortality rate between patients who received remdesivir and those who did not (29.7 [11 of 37] and $34.4\%$ [53 of 154], respectively; $$p \leq 0.59$$).
Since $85\%$ of patients treated with remdesivir also received corticosteroids, we evaluated the impact of their combination therapy on 30 day mortality. Among patients on low-flow oxygen at admission, the mortality rates among those who received remdesivir alone or in combination with corticosteroids ($6\%$ [3 of 51]) was significantly lower than the mortality rate of those who received corticosteroids alone ($18.3\%$ [21 of 115]; $$p \leq 0.036$$). However, mortality rates were similar for remdesivir alone and combination therapy among various patients’ groups (Supplementary file 1).
Giving convalescent plasma to patients later than 3d after diagnosis did not make any difference in mortality (data not shown). In contrast to what was observed for convalescent plasma, the later corticosteroids were administered, the greater the benefit. There was a trend toward a greater reduction in 30 day mortality in patients who received corticosteroids later (>5d after diagnosis), whereas no difference was observed in patients who received corticosteroids earlier (Supplementary file 2).
## Sensitivity analysis
Lastly, sensitivity analyses were performed to evaluate the impact of missing data on our primary analyses. The multivariable logistic regression models of mortality predictors based on the complete-case analysis and multiple imputation analysis were compared for all patients (Table 2) and for patients with and without cancer (Table 3). Among all patients, the two models were similar except that multiple imputation analyses did not show remdesivir having a significant impact on mortality while the complete-case analysis did (Table 2). However, among patients with cancer, both models showed that remdesivir was independently associated with decreased mortality with similar effects. On the other hand, the two models showed some differences in identifying a few risk factors for 30 day mortality. Among patients without cancer, the two models were similar except that multiple imputation analysis identified one more risk factor – ALC<0.5K/µl at diagnosis (Table 3).
## Discussion
Unlike the previously published studies that compared patients with COVID-19 with and without cancer, our study included a large number of patients with cancer from five different continents. Our findings demonstrate that cancer is not an independent risk factor for increased 30 day all-cause mortality in a multivariable logistic regression analyses that accounted for the treatment differences among the countries, which is inconsistent with a recent large study in the US that showed that patients recently receiving cancer treatment had a worse outcome (Chavez-MacGregor et al., 2022). The higher mortality rate observed in patients with cancer by univariable analysis seems to be driven by patients with lung cancer and patients with hematological malignancies, a finding that is consistent with prior literature (Lee et al., 2020; Mehta et al., 2020). Furthermore, lymphopenia which is observed frequently in patients with hematological malignancy was independently associated with higher COVID-19 mortality according to our country-adjusted multivariable analyses. Two large studies in COVID-19 patients with cancer have shown that lymphopenia was independently associated with increased 30 day mortality (Lunski et al., 2021; Schmidt et al., 2022). In addition, the mortality rate in the patients with solid tumors other than lung cancer was not different from the mortality rate in the patients without cancer.
Most of the other independent risk factors that we identified for 30 day mortality after a diagnosis of COVID-19 have also been commonly reported as risk factors for mortality in previous studies of COVID-19 in patients with and without cancer with older age (≥65y) being the strongest independent predictor of 30 day mortality in our study (Zhou et al., 2020; Kuderer et al., 2020; Rivera et al., 2020; He et al., 2020; Albiges et al., 2020; Robilotti et al., 2020; Lee et al., 2020; Mehta et al., 2020; Tian et al., 2020; Lunski et al., 2021; Sun et al., 2020; Rüthrich et al., 2021). In our study, the patients with cancer had a more complicated course after discharge from the hospital; specifically, they more frequently required supplemental oxygen and readmission within 30 days after COVID-19 diagnosis.
The presentation pattern of COVID-19 in patients with cancer was different to that of patients without cancer. Patients with cancer seemed to be less symptomatic. This could be related to the fact that patients with cancer tended to be older, with fewer inflammatory cells (neutrophils and lymphocytes), and more often on corticosteroids. Levels of inflammatory biomarkers were also higher in patients with cancer, particularly D-dimer, ferritin, and procalcitonin.
On multivariable analysis, the only therapeutic agent that independently decreased 30 day all-cause mortality among all patients and in patients with cancer was remdesivir, even after accounting for treatment diffrences among countries. Upon further analysis, remdesivir was found to further decrease mortality in patients with pneumonia and mild hypoxia who were receiving low-flow oxygen (≤6l/min) and not in patients with severe advanced pneumonia who were receiving high-flow oxygen and/or ventilatory support. This is consistent with a large multicenter prospective randomized placebo-controlled trial - Adaptive COVID-19 Treatment Trial-1 (ACTT-1), that found that remdesivir significantly reduced the time to recovery and 28d mortality in patients who were on low-flow oxygen at baseline but not in patients who were on mechanical ventilation or extracorporeal membrane oxygenation at baseline (Beigel et al., 2020).
In a meta-analysis that examined four large prospective randomized trials, remdesivir was shown to be associated with reduced 14 d mortality and reduced need for mechanical ventilation (Shrestha et al., 2021), which also supports our analysis.
In contrast, the WHO-sponsored multinational Solidarity Trial of COVID-19 hospitalized patients who were randomly assigned to either remdesivir (2750patients) or standard of care (2708patients) showed no difference in overall 28d mortality (Pan et al., 2021). However, in two large meta-analyses that each examined more than 13,000 COVID-19 patients from randomized and non-randomized studies, including the WHO randomized trial, remdesivir was associated with a significant improvement in the 28d recovery rate (Rezagholizadeh et al., 2021; Lai et al., 2021). Furthermore, in a large study evaluating treatments and outcomes of COVID-19 among patients with cancer, remdesivir alone was significantly associated with a lower 30 day all-cause mortality rate than other treatments (including high-dose corticosteroids and tocilizumab; Rivera et al., 2020). In addition, in a large multicenter matched controlled study involving mostly patients without cancer, Mozaffari et al. demonstrated that remdesivir was significantly effective in reducing mortality in patients on low-flow oxygen but not patients with advanced disease on high-flow oxygen and ventilator support (Mozaffari et al., 2022). More recently, a prospective randomized, double blind, placebo-controlled study (involving mainly patients without cancer) showed that early initiation of remdesivir prevents progression to severe COVID-19 (Gottlieb et al., 2022).
Therefore, the cumulative data in the literature do support our findings that remdesivir in all patients particularly in patients with cancer improves outcome and may reduce mortality especially if started early in COVID-19 patients with moderate pneumonia who are receiving low-flow oxygen who fit into the stage II of the three-stage COVID-19 classification previously proposed by Siddiqi and Mehra, 2020.
The use of corticosteroids (dexamethasone 6mg/d) was shown in a large randomized open-label study conducted in the United Kingdom to be associated with reduced 28d mortality, particularly in patients requiring oxygen supplementation and invasive ventilation who receive corticosteroids after 7d from the onset of symptoms (Horby et al., 2021). Subsequently, two meta-analyses of several prospective randomized trials demonstrated that corticosteroid use was significantly associated with a decrease in COVID-19 mortality (Sterne et al., 2020; Siemieniuk et al., 2020). In our study, we found that if corticosteroids were started more than 5d after the PCR diagnosis of COVID-19, there was a trend toward a reduction in 30 day COVID-19 mortality compared to starting corticosteroids earlier. 5 d after diagnosis by PCR testing would possibly be equivalent to 7d after the onset of symptoms since the average time from symptom onset to PCR diagnosis has been estimated to be 2–3d (Kucirka et al., 2020). Furthermore, by multivariable analysis and upon further subanalysis (Supplementary file 1), remdesivir was found to improve outcome independently of the effect of steroids.
Our study has several limitations. First, the retrospective design precluded complete assessment of disease progression in the outpatients, which limited their input into the general study. Second, some non-cancer centers contributed data only on their hospitalized patients who were likely sicker. This may have biased the data toward a higher rate of hospitalization among patients without cancer and made the data more heterogeneous. In addition, the predominance of symptomatic patients ($87\%$) might also limit our evaluation of the impact of cancer in the whole picture of the disease. Third, our data contained a lot of missing values due to the nature of data collection. However, we performed sensitivity analyses to evaluate its impact on our primary analyses, and it showed that overall this impact was limited. Most independent predictors were identified by both complete-case and multiple imputation analyses with similar effects. The differences between the two analysis methods in identifying a few independent mortality predictors were probably due to the sample size changes from complete-case analyses to multiple imputation analyses. Last, this study was conducted prior to the introduction of COVID vaccines and included patients who were not vaccinated and who were infected by early variants which limits the generalizability of our results to contemporary COVID-19 patients. A multinational European registry showed that severity of COVID-19 and mortality in cancer patients have improved since 2020. This could be multifactorial related to earlier diagnosis, improved management including current antivirals, monoclonal antibodies, vaccination, as well as different circulating variants of the virus that could be associated with less severe disease than the earlier strains (Pinato et al., 2022). However, a recent study comparing vaccinated and unvaccinated patients with cancer showed that despite the protective role of vaccination, vulnerable patients with cancer, particularly those with risk factors such as lymphopenia, active and progressing cancer, and advanced age, can develop severe and fatal breakthrough infections (Schmidt et al., 2022).
In conclusion, this is the largest multicenter worldwide study comparing COVID-19 in patients with cancer to those without. In this study, although the limited effect size, underlying malignancy was not found to be an independent risk factor for a higher 30 day all-cause COVID-19 mortality. However, lymphopenia and anemia were frequently observed in patients with hematological malignancies, and patients with lung cancer were associated with the highest risks for poor outcome. Finally, remdesivir stood out as the only therapeutic agent independently associated with decreased 30 day mortality, particularly in patients with cancer on low-flow oxygen. Corticosteroids tended to be most useful if given more than 5d after COVID-19 diagnosis. The role of these therapeutics and their timing of administration should be verified in larger studies, especially in patients with cancer, who tend to have a higher degree of immunosuppression, which may lead to prolongation of the viral phase.
## Funding Information
This paper was supported by the following grants:
## Data availability
We are unable to share the data given our restriction policy and the fact that this study includes data from 16 centers and from the five continents. We do not have an agreement or the permission to share our data and other centers’ data. All the analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC).
## References
1. Albiges L, Foulon S, Bayle A, Gachot B, Pommeret F, Willekens C, Stoclin A, Merad M, Griscelli F, Lacroix L, Netzer F, Hueso T, Balleyguier C, Ammari S, Colomba E, Baciarello G, Perret A, Hollebecque A, Hadoux J, Michot J-M, Chaput N, Saada V, Hauchecorne M, Micol J-B, Sun R, Valteau-Couanet D, André F, Scotte F, Besse B, Soria J-C, Barlesi F. **Determinants of the outcomes of patients with cancer infected with SARS-cov-2: results from the Gustave Roussy cohort**. *Nature Cancer* (2020) **1** 965-975. DOI: 10.1038/s43018-020-00120-5
2. Beigel JH, Tomashek KM, Dodd LE, Mehta AK, Zingman BS, Kalil AC, Hohmann E, Chu HY, Luetkemeyer A, Kline S, Lopez de Castilla D, Finberg RW, Dierberg K, Tapson V, Hsieh L, Patterson TF, Paredes R, Sweeney DA, Short WR, Touloumi G, Lye DC, Ohmagari N, Oh M-D, Ruiz-Palacios GM, Benfield T, Fätkenheuer G, Kortepeter MG, Atmar RL, Creech CB, Lundgren J, Babiker AG, Pett S, Neaton JD, Burgess TH, Bonnett T, Green M, Makowski M, Osinusi A, Nayak S, Lane HC. **Remdesivir for the treatment of Covid-19-final report**. *The New England Journal of Medicine* (2020) **383** 1813-1826. DOI: 10.1056/NEJMoa2007764
3. Chavez-MacGregor M, Lei X, Zhao H, Scheet P, Giordano SH. **Evaluation of COVID-19 mortality and adverse outcomes in US patients with or without cancer**. *JAMA Oncology* (2022) **8** 69-78. DOI: 10.1001/jamaoncol.2021.5148
4. Gottlieb RL, Vaca CE, Paredes R, Mera J, Webb BJ, Perez G, Oguchi G, Ryan P, Nielsen BU, Brown M, Hidalgo A, Sachdeva Y, Mittal S, Osiyemi O, Skarbinski J, Juneja K, Hyland RH, Osinusi A, Chen S, Camus G, Abdelghany M, Davies S, Behenna-Renton N, Duff F, Marty FM, Katz MJ, Ginde AA, Brown SM, Schiffer JT, Hill JA. **Early remdesivir to prevent progression to severe Covid-19 in outpatients**. *New England Journal of Medicine* (2022) **386** 305-315. DOI: 10.1056/NEJMoa2116846
5. He W, Chen L, Chen L, Yuan G, Fang Y, Chen W, Wu D, Liang B, Lu X, Ma Y, Li L, Wang H, Chen Z, Li Q, Gale RP. **COVID-19 in persons with haematological cancers**. *Leukemia* (2020) **34** 1637-1645. DOI: 10.1038/s41375-020-0836-7
6. Horby P, Lim WS, Emberson JR, Mafham M, Bell JL, Linsell L, Staplin N, Brightling C, Ustianowski A, Elmahi E, Prudon B, Green C, Felton T, Chadwick D, Rege K, Fegan C, Chappell LC, Faust SN, Jaki T, Jeffery K, Montgomery A, Rowan K, Juszczak E, Baillie JK, Haynes R, Landray MJ. **Dexamethasone in hospitalized patients with covid-19**. *The New England Journal of Medicine* (2021) **384** 693-704. DOI: 10.1056/NEJMoa2021436
7. Janiaud P, Axfors C, Schmitt AM, Gloy V, Ebrahimi F, Hepprich M, Smith ER, Haber NA, Khanna N, Moher D, Goodman SN, Ioannidis JPA, Hemkens LG. **Association of convalescent plasma treatment with clinical outcomes in patients with COVID-19: a systematic review and meta-analysis**. *JAMA* (2021) **325** 1185-1195. DOI: 10.1001/jama.2021.2747
8. **COVID-19 Map**. (2021)
9. Kucirka LM, Lauer SA, Laeyendecker O, Boon D, Lessler J. **Variation in false-negative rate of reverse transcriptase polymerase chain reaction–based SARS-cov-2 tests by time since exposure**. *Annals of Internal Medicine* (2020) **173** 262-267. DOI: 10.7326/M20-1495
10. Kuderer NM, Choueiri TK, Shah DP, Shyr Y, Rubinstein SM, Rivera DR, Shete S, Hsu CY, Desai A, de Lima Lopes G, Grivas P, Painter CA, Peters S, Thompson MA, Bakouny Z, Batist G, Bekaii-Saab T, Bilen MA, Bouganim N, Larroya MB, Castellano D, Del Prete SA, Doroshow DB, Egan PC, Elkrief A, Farmakiotis D, Flora D, Galsky MD, Glover MJ, Griffiths EA, Gulati AP, Gupta S, Hafez N, Halfdanarson TR, Hawley JE, Hsu E, Kasi A, Khaki AR, Lemmon CA, Lewis C, Logan B, Masters T, McKay RR, Mesa RA, Morgans AK, Mulcahy MF, Panagiotou OA, Peddi P, Pennell NA, Reynolds K, Rosen LR, Rosovsky R, Salazar M, Schmidt A, Shah SA, Shaya JA, Steinharter J, Stockerl-Goldstein KE, Subbiah S, Vinh DC, Wehbe FH, Weissmann LB, Wu JTY, Wulff-Burchfield E, Xie Z, Yeh A, Yu PP, Zhou AY, Zubiri L, Mishra S, Lyman GH, Rini BI, Warner JL. **Clinical impact of COVID-19 on patients with cancer (CCC19): a cohort study**. *Lancet* (2020) **395** 1907-1918. DOI: 10.1016/S0140-6736(20)31187-9
11. Lai CC, Chen CH, Wang CY, Chen KH, Wang YH, Hsueh PR. **Clinical efficacy and safety of remdesivir in patients with COVID-19: a systematic review and network meta-analysis of randomized controlled trials**. *The Journal of Antimicrobial Chemotherapy* (2021) **76** 1962-1968. DOI: 10.1093/jac/dkab093
12. Lee LY, Cazier JB, Angelis V, Arnold R, Bisht V, Campton NA, Chackathayil J, Cheng VW, Curley HM, Fittall MW, Freeman-Mills L, Gennatas S, Goel A, Hartley S, Hughes DJ, Kerr D, Lee AJ, Lee RJ, McGrath SE, Middleton CP, Murugaesu N, Newsom-Davis T, Okines AF, Olsson-Brown AC, Palles C, Pan Y, Pettengell R, Powles T, Protheroe EA, Purshouse K, Sharma-Oates A, Sivakumar S, Smith AJ, Starkey T, Turnbull CD, Várnai C, Yousaf N, Kerr R, Middleton G. **COVID-19 mortality in patients with cancer on chemotherapy or other anticancer treatments: a prospective cohort study**. *Lancet* (2020) **395** 1919-1926. DOI: 10.1016/S0140-6736(20)31173-9
13. Libster R, Pérez Marc G, Wappner D, Coviello S, Bianchi A, Braem V, Esteban I, Caballero MT, Wood C, Berrueta M, Rondan A, Lescano G, Cruz P, Ritou Y, Fernández Viña V, Álvarez Paggi D, Esperante S, Ferreti A, Ofman G, Ciganda Á, Rodriguez R, Lantos J, Valentini R, Itcovici N, Hintze A, Oyarvide ML, Etchegaray C, Neira A, Name I, Alfonso J, López Castelo R, Caruso G, Rapelius S, Alvez F, Etchenique F, Dimase F, Alvarez D, Aranda SS, Sánchez Yanotti C, De Luca J, Jares Baglivo S, Laudanno S, Nowogrodzki F, Larrea R, Silveyra M, Leberzstein G, Debonis A, Molinos J, González M, Perez E, Kreplak N, Pastor Argüello S, Gibbons L, Althabe F, Bergel E, Polack FP. **Early high-titer plasma therapy to prevent severe covid-19 in older adults**. *The New England Journal of Medicine* (2021) **384** 610-618. DOI: 10.1056/NEJMoa2033700
14. Lunski MJ, Burton J, Tawagi K, Maslov D, Simenson V, Barr D, Yuan H, Johnson D, Matrana M, Cole J, Larned Z, Moore B. **Multivariate mortality analyses in COVID-19: comparing patients with cancer and patients without cancer in Louisiana**. *Cancer* (2021) **127** 266-274. DOI: 10.1002/cncr.33243
15. Mehta V, Goel S, Kabarriti R, Cole D, Goldfinger M, Acuna-Villaorduna A, Pradhan K, Thota R, Reissman S, Sparano JA, Gartrell BA, Smith RV, Ohri N, Garg M, Racine AD, Kalnicki S, Perez-Soler R, Halmos B, Verma A. **Case fatality rate of cancer patients with COVID-19 in a new York hospital system**. *Cancer Discovery* (2020) **10** 935-941. DOI: 10.1158/2159-8290.CD-20-0516
16. Mozaffari E, Chandak A, Zhang Z, Liang S, Thrun M, Gottlieb RL, Kuritzkes DR, Sax PE, Wohl DA, Casciano R, Hodgkins P, Haubrich R. **Remdesivir treatment in hospitalized patients with coronavirus disease 2019 (COVID-19): a comparative analysis of in-hospital all-cause mortality in a large multicenter observational cohort**. *Clinical Infectious Diseases* (2022) **75** e450-e458. DOI: 10.1093/cid/ciab875
17. Pan H, Peto R, Henao-Restrepo AM, Preziosi MP, Sathiyamoorthy V, Abdool Karim Q, Alejandria MM, Hernández García C, Kieny MP, Malekzadeh R, Murthy S, Reddy KS, Roses Periago M, Abi Hanna P, Ader F, Al-Bader AM, Alhasawi A, Allum E, Alotaibi A, Alvarez-Moreno CA, Appadoo S, Asiri A, Aukrust P, Barratt-Due A, Bellani S, Branca M, Cappel-Porter HBC, Cerrato N, Chow TS, Como N, Eustace J, García PJ, Godbole S, Gotuzzo E, Griskevicius L, Hamra R, Hassan M, Hassany M, Hutton D, Irmansyah I, Jancoriene L, Kirwan J, Kumar S, Lennon P, Lopardo G, Lydon P, Magrini N, Maguire T, Manevska S, Manuel O, McGinty S, Medina MT, Mesa Rubio ML, Miranda-Montoya MC, Nel J, Nunes EP, Perola M, Portolés A, Rasmin MR, Raza A, Rees H, Reges PPS, Rogers CA, Salami K, Salvadori MI, Sinani N, Sterne JAC, Stevanovikj M, Tacconelli E, Tikkinen KAO, Trelle S, Zaid H, Røttingen JA, Swaminathan S. **Repurposed antiviral drugs for covid-19-interim who solidarity trial results**. *The New England Journal of Medicine* (2021) **384** 497-511. DOI: 10.1056/NEJMoa2023184
18. Pinato DJ, Patel M, Scotti L, Colomba E, Dolly S, Loizidou A, Chester J, Mukherjee U, Zambelli A, Dalla Pria A, Aguilar-Company J, Bower M, Salazar R, Bertuzzi A, Brunet J, Lambertini M, Tagliamento M, Pous A, Sita-Lumsden A, Srikandarajah K, Colomba J, Pommeret F, Seguí E, Generali D, Grisanti S, Pedrazzoli P, Rizzo G, Libertini M, Moss C, Evans JS, Russell B, Harbeck N, Vincenzi B, Biello F, Bertulli R, Ottaviani D, Liñan R, Rossi S, Carmona-García MC, Tondini C, Fox L, Baggi A, Fotia V, Parisi A, Porzio G, Queirolo P, Cruz CA, Saoudi-Gonzalez N, Felip E, Roqué Lloveras A, Newsom-Davis T, Sharkey R, Roldán E, Reyes R, Zoratto F, Earnshaw I, Ferrante D, Marco-Hernández J, Ruiz-Camps I, Gaidano G, Patriarca A, Bruna R, Sureda A, Martinez-Vila C, Sanchez de Torre A, Berardi R, Giusti R, Mazzoni F, Guida A, Rimassa L, Chiudinelli L, Franchi M, Krengli M, Santoro A, Prat A, Tabernero J, Van Hemelrijck M, Diamantis N, Gennari A, Cortellini A. **Time-dependent COVID-19 mortality in patients with cancer: an updated analysis of the oncovid registry**. *JAMA Oncology* (2022) **8** 114-122. DOI: 10.1001/jamaoncol.2021.6199
19. Rezagholizadeh A, Khiali S, Sarbakhsh P, Entezari-Maleki T. **Remdesivir for treatment of COVID-19; an updated systematic review and meta-analysis**. *European Journal of Pharmacology* (2021) **897**. DOI: 10.1016/j.ejphar.2021.173926
20. Rivera DR, Peters S, Panagiotou OA, Shah DP, Kuderer NM, Hsu C-Y, Rubinstein SM, Lee BJ, Choueiri TK, de Lima Lopes G, Grivas P, Painter CA, Rini BI, Thompson MA, Arcobello J, Bakouny Z, Doroshow DB, Egan PC, Farmakiotis D, Fecher LA, Friese CR, Galsky MD, Goel S, Gupta S, Halfdanarson TR, Halmos B, Hawley JE, Khaki AR, Lemmon CA, Mishra S, Olszewski AJ, Pennell NA, Puc MM, Revankar SG, Schapira L, Schmidt A, Schwartz GK, Shah SA, Wu JT, Xie Z, Yeh AC, Zhu H, Shyr Y, Lyman GH, Warner JL. **Utilization of COVID-19 treatments and clinical outcomes among patients with cancer: a COVID-19 and cancer Consortium (CCC19) cohort study**. *Cancer Discovery* (2020) **10** 1514-1527. DOI: 10.1158/2159-8290.CD-20-0941
21. Robilotti EV, Babady NE, Mead PA, Rolling T, Perez-Johnston R, Bernardes M, Bogler Y, Caldararo M, Figueroa CJ, Glickman MS, Joanow A, Kaltsas A, Lee YJ, Lucca A, Mariano A, Morjaria S, Nawar T, Papanicolaou GA, Predmore J, Redelman-Sidi G, Schmidt E, Seo SK, Sepkowitz K, Shah MK, Wolchok JD, Hohl TM, Taur Y, Kamboj M. **Determinants of COVID-19 disease severity in patients with cancer**. *Nature Medicine* (2020) **26** 1218-1223. DOI: 10.1038/s41591-020-0979-0
22. Rüthrich MM, Giessen-Jung C, Borgmann S, Classen AY, Dolff S, Grüner B, Hanses F, Isberner N, Köhler P, Lanznaster J, Merle U, Nadalin S, Piepel C, Schneider J, Schons M, Strauss R, Tometten L, Vehreschild JJ, von Lilienfeld-Toal M, Beutel G, Wille K. **COVID-19 in cancer patients: clinical characteristics and outcome-an analysis of the LEOSS registry**. *Annals of Hematology* (2021) **100** 383-393. DOI: 10.1007/s00277-020-04328-4
23. Salazar E, Christensen PA, Graviss EA, Nguyen DT, Castillo B, Chen J, Lopez BV, Eagar TN, Yi X, Zhao P, Rogers J, Shehabeldin A, Joseph D, Masud F, Leveque C, Olsen RJ, Bernard DW, Gollihar J, Musser JM. **Significantly decreased mortality in a large cohort of coronavirus disease 2019 (COVID-19) patients transfused early with convalescent plasma containing high-titer anti-severe acute respiratory syndrome coronavirus 2 (SARS-cov-2) spike protein IgG**. *The American Journal of Pathology* (2021) **191** 90-107. DOI: 10.1016/j.ajpath.2020.10.008
24. Schmidt AL, Labaki C, Hsu C-Y, Bakouny Z, Balanchivadze N, Berg SA, Blau S, Daher A, El Zarif T, Friese CR, Griffiths EA, Hawley JE, Hayes-Lattin B, Karivedu V, Latif T, Mavromatis BH, McKay RR, Nagaraj G, Nguyen RH, Panagiotou OA, Portuguese AJ, Puc M, Santos Dutra M, Schroeder BA, Thakkar A, Wulff-Burchfield EM, Mishra S, Farmakiotis D, Shyr Y, Warner JL, Choueiri TK. **COVID-19 vaccination and breakthrough infections in patients with cancer**. *Annals of Oncology* (2022) **33** 340-346. DOI: 10.1016/j.annonc.2021.12.006
25. Shrestha DB, Budhathoki P, Syed NIH, Rawal E, Raut S, Khadka S. **Remdesivir: A potential game-changer or just A myth? A systematic review and meta-analysis**. *Life Sciences* (2021) **264**. DOI: 10.1016/j.lfs.2020.118663
26. Siddiqi HK, Mehra MR. **COVID-19 illness in native and immunosuppressed states: a clinical–therapeutic staging proposal**. *The Journal of Heart and Lung Transplantation* (2020) **39** 405-407. DOI: 10.1016/j.healun.2020.03.012
27. Siemieniuk RA, Bartoszko JJ, Zeraatkar D, Kum E, Qasim A, Díaz Martinez JP, Izcovich A, Rochwerg B, Lamontagne F, Han MA, Agarwal A, Agoritsas T, Azab M, Bravo G, Chu DK, Couban R, Cusano E, Devji T, Escamilla Z, Foroutan F, Gao Y, Ge L, Ghadimi M, Heels-Ansdell D, Honarmand K, Hou L, Ibrahim S, Khamis A, Lam B, Mansilla C, Loeb M, Miroshnychenko A, Marcucci M, McLeod SL, Motaghi S, Murthy S, Mustafa RA, Pardo-Hernandez H, Rada G, Rizwan Y, Saadat P, Switzer C, Thabane L, Tomlinson G, Vandvik PO, Vernooij RW, Viteri-García A, Wang Y, Yao L, Zhao Y, Guyatt GH, Brignardello-Petersen R. **Drug treatments for covid-19: living systematic review and network meta-analysis**. *BMJ* (2020) **370**. DOI: 10.1136/bmj.m2980
28. Sterne JAC, Murthy S, Diaz JV, Slutsky AS, Villar J, Angus DC, Annane D, Azevedo LCP, Berwanger O, Cavalcanti AB, Dequin PF, Du B, Emberson J, Fisher D, Giraudeau B, Gordon AC, Granholm A, Green C, Haynes R, Heming N, Higgins JPT, Horby P, Jüni P, Landray MJ, Le Gouge A, Leclerc M, Lim WS, Machado FR, McArthur C, Meziani F, Møller MH, Perner A, Petersen MW, Savovic J, Tomazini B, Veiga VC, Webb S, Marshall JC. **Association between administration of systemic corticosteroids and mortality among critically ill patients with COVID-19: A meta-analysis**. *JAMA* (2020) **324** 1330-1341. DOI: 10.1001/jama.2020.17023
29. Sun L, Surya S, Le AN, Desai H, Doucette A, Gabriel P, Ritchie M, Rader D, Maillard I, Bange E, Huang A, Vonderheide RH, DeMichele A, Verma A, Mamtani R, Maxwell KN. **Rates of COVID-19-Related Outcomes in Cancer Compared to Non-Cancer Patients**. *medRxiv* (2020). DOI: 10.1101/2020.08.14.20174961
30. Tian J, Yuan X, Xiao J, Zhong Q, Yang C, Liu B, Cai Y, Lu Z, Wang J, Wang Y, Liu S, Cheng B, Wang J, Zhang M, Wang L, Niu S, Yao Z, Deng X, Zhou F, Wei W, Li Q, Chen X, Chen W, Yang Q, Wu S, Fan J, Shu B, Hu Z, Wang S, Yang X-P, Liu W, Miao X, Wang Z. **Clinical characteristics and risk factors associated with COVID-19 disease severity in patients with cancer in Wuhan, China: a multicentre, retrospective, cohort study**. *The Lancet. Oncology* (2020) **21** 893-903. DOI: 10.1016/S1470-2045(20)30309-0
31. Zhou F, Yu T, Du R, Fan G, Liu Y, Liu Z, Xiang J, Wang Y, Song B, Gu X, Guan L, Wei Y, Li H, Wu X, Xu J, Tu S, Zhang Y, Chen H, Cao B. **Clinical course and risk factors for mortality of adult inpatients with COVID-19 in wuhan, china: a retrospective cohort study**. *Lancet* (2020) **395** 1054-1062. DOI: 10.1016/S0140-6736(20)30566-3
|
---
title: Molecular signatures distinguish senescent cells from inflammatory cells in
aged mouse callus stromal cells
authors:
- Jiatong Liu
- Xi Lin
- Andrew McDavid
- Yutiancheng Yang
- Hengwei Zhang
- Brendan F. Boyce
- Lianping Xing
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9981154
doi: 10.3389/fendo.2023.1090049
license: CC BY 4.0
---
# Molecular signatures distinguish senescent cells from inflammatory cells in aged mouse callus stromal cells
## Abstract
Cellular senescence plays important roles in age-related diseases, including musculoskeletal disorders. Senescent cells (SCs) exert a senescence-associated secretory phenotype (SASP) by producing SASP factors, some of which overlap with factors produced by inflammatory cells (Inf-Cs). However, the differences between SCs and Inf-Cs and how they interact with each other during fracture repair have not been well studied. Here, we analyzed single cell RNA sequencing data of aged mouse fracture callus stromal cells. We defined Inf-Cs as cells that express NF-κB Rela/Relb, SCs as cells that express the senescence genes, Cdkn1a, Cdkn2a or Cdkn2c, and inflammatory SCs (Inf-SCs) as cells that express both NF-κB and senescence genes. Differentially expressed genes and pathway analyses revealed that Inf-SCs and SCs had a similar gene expression profile and upregulated pathways that are related to DNA damage/oxidation-reduction and cellular senescence, while Inf-Cs expressed different gene signatures and pathways from SCs and Inf-SCs, mainly related to inflammation. Cellchat software analysis indicated that SCs and Inf-SCs are potential ligand-producing cells that affect Inf-Cs as target cells. Cell culture experiments demonstrated that SC conditioned medium promoted inflammatory gene expression by callus-derived mesenchymal progenitor cells, and Inf-Cs had reduced osteoblast differentiation capacity. In summary, we have identified three cell subclusters associated with inflammation and senescence in callus stromal cells, predicted potential effects of Inf-SCs and SCs on Inf-Cs by production of active ligands, and demonstrated that when mesenchymal progenitors acquire inflammatory phenotypes their osteogenic potential is reduced.
## Introduction
Cellular senescence plays important roles in human diseases, including musculoskeletal disorders [1, 2]. Recently we found that aged mice have markedly increased senescent cells (SCs) in fracture callus that exert the senescence-associated secretory phenotype (SASP). Clearance of SCs with senolytic drugs promotes fracture healing in aged mice [3] and young mice [4]. However, fractures are also associated with inflammation. Inflammatory cells (Inf-Cs) produce pro-inflammatory factors, some of which overlap with SASP factors [5, 6].
Pre-clinic studies have revealed potential difference between SCs and Inf-Cs in fracture healing. For instance, the acute inflammatory response in the callus following fracture peaks at 24-48 hours and quickly disappears within a week, and deficiency or inhibition of inflammation at this stage causes delayed fracture healing by inhibiting endochondral ossification [7, 8]. In contrast, SCs accumulate in the callus gradually and peaks around 10-14 days post-fracture and removing SCs by senolytic drugs, given at 3-7 days post-fracture before the peak of SC accumulation, improves fracture healing [3].
Distinguishing SCs from Inf-Cs in aged callus is very important, not only because both of them are increased and produce inflammatory factors, but also because they require different drug treatments, e.g. senolytic drugs for SCs, and non-steroidal anti-inflammation drugs (NSAIDs) for Inf-Cs. NSAIDs improve fracture healing in mice [9], but the drugs have been reported to significantly increase the risk of a second hip fracture, especially in old patients [10]. The meta-analysis of the NSAID treatment in clinical trials reports a negative effect of NSAID, which is highly dose and time dependent because long term NSAID administration increase the rates of non-union fracture in the elder patients [11, 12]. These studies indicate that anti-inflammatory drugs may not be a good choice for treating fractures in the elderly.
Thus, identification of any differences between the behavior of Inf-Cs and SCs for treatment of aging fracture is an important unmet need. Here, we analyzed our recently published single cell RNA-sequencing (scRNA-seq) data set of aged mouse callus stromal cells [13]. We compared cells that express inflammatory genes and cells that express senescence-associated genes to study molecular signatures of Inf-Cs and SCs and their interactions using bioinformatic analyses. We validated our findings with cell culture experiments and callus cells from NF-κB-GFP reporter mice.
## Analysis of single cell RNA-sequencing dataset
scRNA-seq data that we published recently were reanalyzed using cells from aged callus [13] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE199755, GEO, GSE199755). In our previous study, we collected CD45-CD31-Ter119- stromal cells from callus of 4-month-old young (equivalent to a 26-year-old human) and 21-month-old (equivalent to a 62-year-old human) aged C57BL/6J male mice at 10-day post-fracture (dpf) by fluorescence-activated cell scoring (FACS), the time point when the expression levels of senescence-associated genes reach the peak [3]. In the current study, we analyzed data from 6,834 aged stromal cells using a Seurat (version 4.0.6) R package. The rationale of using the data from aged mice is that the callus in aged mice significantly higher levels of inflammation and cellular senescence than those in young mice [3]. In brief, the top 2,000 variable genes were identified and ranked by coefficient of variation. The reason why only top 2,000 variable genes were identified is that we used the same threshold described in the original study for Seurat packages [14], where Stuart et al. used only top 2,000 variable genes in their analysis and concluded that focusing on these high variable genes in downstream analysis helps to highlight biological signal in single-cell datasets. Dimensionality reduction of datasets was performed by “RunPCA” function with 10 principal components (npcs = 25) at a resolution of 0.1. Find Neighbors function was used to compute the shared nearest-neighbors for a given dataset with parameter $k = 25.$ Clusters of cells were identified based on SNN modularity optimization with “FindCluster” function. “ RunUMAP” function was further used to perform Uniform Manifold Approximation and Projection (UMAP) dimensional reduction. Cell clusters were visualized on reduced UMAP dimensions using “DimPlot” function. Differentially expressed genes (DEGs) of each cluster were identified with “FindAllMarkers” function. The top 10 DEGs with the highest average log 2-fold-change were presented in a heatmap using “DoHeatmap” function for cluster functional annotation.
## Functional enrichment analysis
To examine the biological processes and signaling pathways in various cell subsets, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. DEGs identified in Seurat/R and with an average log 2-fold-change>0.414, p value<0.05 were used. The top 6 upregulated genes corresponding to biological process or pathways with the highest –log 10-fold-change were presented and used for cluster functional annotation.
## Cell-cell commination analyses
To identify and illustrate intercellular signaling communication, we used an open-source R package iTALK (http://github.com/Coolgenome/iTALK) that is designed to profile and visualize the ligand-receptor mediated inter-cellular cross-talk signal using scRNA-seq data. In brief, we converted Seurat object to Cellchat/R object, a publicly available database of 2021 with validated molecular interactions for Mus Musculus for ligand-receptor identification [15]. We inferred the communication probability/strength between interaction groups using computeCommunProb/CellChat/R function with “type=truncateMean” as the gene expression average method and “trim=0.1” as the threshold. We set Inf-SCs and SCs as the sender/source cells and Inf-Cs as receiver cells, based on Cellchat predicted result to predict the potential ligand-receptor pairs. We also applied the Nichenet packages, a method that predicts ligand- downstream target genes between interacting cells by combining their gene expression data with prior knowledge on signaling and gene regulatory networks [16]. Similar to the Cellchat analysis, Inf-SCs and SCs were set as sender/niche cells and Inf-Cs were set as receiver/target cells.
## Animals and tibial fracture procedure
NF-κB-GFP reporter mice (Strain#: 027529) on a C57BL/6 background were purchased from the Jackson Laboratory [17], in which a transgenic construct contains an enhanced GFP-luciferase fusion gene under the control of tandem copies of a 36-base enhancer (containing two NF-κB binding sites) upstream of a herpes simplex virus minimal thymidine kinase promoter. Mice were housed in micro-isolator technique rodent rooms. We used 4-month-old male mice in the current study to exclude the effects of variations in levels of female sex hormones. Open tibial fractures were performed according to the standard operating procedure established in the URMC Center for Musculoskeletal Research [18]. In brief, a 5 mm long incision was made in the skin on the anterior side of the tibia after anesthesia. A sterile 27 G needle was inserted into the marrow cavity of the tibia from the proximal end, temporarily withdrawn to facilitate midshaft transection of the tibia using a scalpel, and then reinserted to stabilize the fracture. The incision was closed with 5-0 nylon sutures. Mice received buprenorphine SR, 0.5 mg/kg to control pain. Fractures were confirmed by radiography using a Faxitron device (Hologic, Marlborough, MA). All animal procedures were conducted in accordance with approved guidelines of the University of Rochester Committee for Animal Resources (protocol number: 2001-121R).
## Callus-derived mesenchymal progenitor cell preparation, cell growth, and osteoblast differentiation assays
For CaMPC preparation, mice were euthanized by CO2 inhalation and secondary cervical dislocation at 10 dpf. Surrounding soft tissue was dissected from callus, which was cut into pieces. Callus pieces were washed thoroughly with cold PBS and then digested with ACCUMAX cell detachment solution (Stem cell Tech) for 30 minutes at room temperature and cultured in basal medium (alpha-MEM medium containing $15\%$ FBS). Cells that migrated from callus pieces were cultured to confluence in basal medium and named CaMPCs. For cell growth assays, CaMPCs from NF-κB-GFP mice were subjected to FACS to collect GFP+ and GFP- cells. GFP+ and GFP- cells were cultured in basal medium for 2 days and stained with a cell counting kit 8 (CCK8) (Abcam, cat#: ab228554) following the manufacturer’s instructions. For osteoblast differentiation assays, GFP+ and GFP- cells were cultured in basal medium containing $10\%$ FBS with 50 μg/ml ascorbic acid and 10 mM β-glycerophosphate for 14 days and stained for alkaline phosphatase with 1-step NBT/BCIP reagent (Thermo Scientific, cat#: 34042).
## RT-qPCR
RNA was extracted in TRIzol solution and cDNA was synthesized using the iSCRIPT cDNA synthesis kit (BioRad). qPCR was performed with iQ SYBR Green Supermix using an iCycler PCR machine (BioRad). β-actin was amplified on the same plates and used to normalize the data. Each sample was prepared in triplicate and each experiment was repeated at least once. The relative abundance of each gene was calculated by subtracting the CT value of each sample for an individual gene from the corresponding CT value of β-actin (ΔCT). ΔΔCT was obtained by subtracting the ΔCT from the reference point. These values were then raised to the power 2 (2ΔΔCT) to yield fold-expression relative to the reference point. Representative data are presented as means ± SD of the triplicates or of 3 wells of cell cultures. The sequences of primers and qPCR conditions used in current study are shown in Supplemental Table 1.
## Beta-galactosidase activity assay
Senescent cells (SCs) were detected used the fluorescent senescence-associated-β-galactosidase assay [19, 20]. Callus cells or cultured CaMPCs were treated with 100 μM 9H-(1,3-dichloro-9,9-dimethylacridin-2-one-7-yl) β-D-galactopyranoside (DDAOG) Thermo Fisher Scientific, cat#: D-6488) for 50 minutes and subjected to flow cytometry for 7-hydroxy-9H(I,3-dichloro-9,9-dimethylacridin-2-one) signal.
## Statistical analysis
Statistical analysis was performed using GraphPad Prism 9 software (GraphPad Software Inc., San Diego, CA, USA). Data are presented as mean ± SD. Levene’s test was used to evaluate the homogeneity of variance and Shapiro Wilk test was used to assess the normality of the data. The in vitro data presented in this study are homogenous and fit into normal distribution. Comparisons among 3 or more groups were analyzed using One-way (for groups with one variable) or Two-way (for groups with two variables) ANOVA following Tukey post-hoc test. p values <0.05 were considered statistically significant.
## Results
Callus stromal cells contain distinct cell sub-populations that express genes associated with inflammation, senescence, or combined inflammation and senescence. To investigate the differences between callus inflammatory cells (Inf-Cs) and senescent cells (SCs), we analyzed our recently published scRNA-seq dataset of CD45-CD31-Ter119- callus stromal cells [13]. We used data from aged callus cells because inflammation and cellular senescence are increased with aging [13, 21]. We identified 4 major cell clusters by graph-based clustering (Figure 1A). Cluster1 and cluster2 expressed known osteogenic genes (Runx2, Acta2, Col5a1), cluster3 expressed osteogenic genes with a relative lower level than cluster1 and 3, and cluster4 expressed known adipogenic genes (Apoe, Lpl) (Figure 1B). Thus, we named cluster1, cluster2, cluster3 as osteogenic-1, osteogenic-2, osteogenic-3, and cluster4 as adipogenic cluster, respectively (Figure 1A). Heatmap of top differentially expressed genes (DEGs) showed that similar to our previous report that aged callus cells have a high inflammatory signature [13], cells in clusters1, 2 and 4 expressed inflammatory (S100a11, Retnlg, Mmp9, Cxcr2, and Ccr2) genes. Interestingly, cells in cluster3 expressed genes associated with chromosome modification and cellular senescence (Tuba4a and Dtymk) (Figure 1C).
**Figure 1:** *Callus stromal cells contain cell subsets that express genes related to inflammation, senescence, and inflammation/senescence. scRNA-seq dataset from aged (21-month-old) callus CD45-CD31-Ter119- stromal cells were analyzed. (A) A total of 6,834 cells were subjected to unsupervised SNN clustering using Seurat/R and identified 4 major clusters: cluster1: osteogenic cells-1, cluster2: osteogenic cells-2, cluster3: osteogenic cells-3, and cluster4: adipogenic cells. (B) Expression of putative osteogenic and adipogenic marker genes: Runx2 and Col5a1 (osteoblast-lineage), Acta2 (skeleton stromal/stem cell-lineage), and Apoe and Lpl (adipocyte-lineage). (C) Heatmap of the top 10 DEGs in clusters1-4 showing cluster1 and cluster2 expressed inflammatory and matrix related genes, cluster3 expressed histone genes, and cluster4 expressed inflammatory and oncogenes. (D) Feature plot of expression of inflammatory genes (Rela and Relb) or senescence-associated genes (Cdkn1a, Cdkn2a, Cdkn2c). (E) Venn plot of cell numbers in Inf-C (inflammatory+senescent-cells), Inf-SC (inflammatory+senescent+ cells) and SC (inflammatory-senescent+ cells). (F) Violin plot of the expression level of inflammatory genes (Rela and Relb) or senescence-associated genes (Cdkn1a, Cdkn2a, Cdkn2c) in three subclusters.*
To label inflammatory cells and SCs for further analysis, we defined cells that express the inflammation-related genes, Rela or Relb [22], as the inflammatory cell (Inf-C) subset and cells that express senescence-associated genes as the SC subset. We examined the expression of 10 genes that have been used to detect SCs of various cell types in the literature [23, 24] and decided to use Cdkn1a, Cdkn2c and Cdkn2a as senescence-associated genes because Cdkn1a and Cdkn2c are expressed mainly by cluster 3 (Supplemental Figure 1) and Cdkn2a has been used in our previous study to label callus SCs [3]. Thus, we defined cells that express Cdkn1a, Cdkn2a or Cdkn2c as SCs. Here Inf-Cs did not express Cdkn1a/Cdkn2a/Cdkn2c, while SCs did not express Rela/Relb. Inf-Cs were present in all clusters, while SCs were mainly localized in cluster3, the osteogenic-3, and cluster4, the adipogenic cluster. Some cells in the adipogenic cluster co-expressed both inflammation and senescence-associated genes and were named as the inflammatory SC (Inf-SC) subset (Figure 1D). Inf-Cs comprised $34\%$ [2321], Inf-SCs $8\%$ [579], and SCs $7\%$ [452] of the total 6,834 cells analyzed (Figure 1E). The expression pattern of these inflammatory and senescence genes in the 3 subsets was illustrated in a violin plot (Figure 1F).
To investigate functional differences among Inf-C, Inf-SC, SC subsets, we compared their top 10 DEGs. A half of the top 10 DEGs in Inf-SCs and SCs overlapped (Figure 2A), while Inf-Cs expressed a different set of genes. The heatmap revealed that Inf-Cs highly expressed the matrix degrading gene, Mmp9, and other inflammation-associated genes (S100a6, S100a11). Inf-SCs and SCs highly expressed Cdkn2c, a gene associated with cellular senescence, and Tuba1b gene-associated with histone modification (Figure 2B). Although ~half of the genes expressed by Inf-SCs and SCs were similar, Tuba1b, Rrm1, Cdk1, Asf1b and Snrpd1 were distinct in Inf-SCs, while Kif11, Hist1hib, Cit, Fbxo5, and Nusap1 were expressed only by SCs (Figures 2A, B). Gene ontology (GO) analysis revealed that the top 6 GO terms corresponding to biological process in the Inf-C subset were mainly related to chemokine and immune cell-related signaling pathways, while those in the Inf-SC and SC subsets were mainly related to DNA damage and repair, oxidation-reduction and cellular senescence (Figure 2C). KEGG analysis of 6 upregulated pathways also showed a trend similar to GO analysis (Figure 2D).
**Figure 2:** *Inflammatory, inflammatory senescent, and senescent cells have distinct signature genes and up-regulated pathways. (A) List of top 10 DEGs in Inf-SCs and SCs. Red indicates same signature genes between Inf-SC and SC. (B) Heatmap of the top 10 DEGs in Inf-Cs, Inf-SCs and SCs. Top 6 upregulated pathways by GO (C) and KEGG (D) analysis using the DEGs with Log2Fold changes greater than 25%.*
SCs acquire senescence-associated secretory phenotype (SASP) [25]. We reported higher expression of 12 SASP factors in callus tissue from aged mice than those from young mice, including Tgfβ1, Tnfα, Il1a, Il1b, Il4, Il6, Csf1, Cxcl1, Cxcl2, Ccl3, Ccl5, and Ccl8 [3]. Among these SASP factors, only Tgfβ1 and Tnfα showed differential expression in Inf-C and Inf-SC, SC subsets. Tgfβ1 and Tnfα were expressed mainly by Inf-SC and SC subsets (Figure 3A). In addition, we examined the expression of 111 cytokine genes listed in a commercial mouse cytokine array (http://www.rndsystems.com/products/proteome-profiler-mouse-xl-cytokine-array_ary028) and found that Ccl6 and Vegfa had different expression profiles among the 3 cell subsets. Ccl6 was expressed mainly by the Inf-C and Inf-SC subsets, while Vegfa was expressed mainly by the Inf-SC and SC subsets (Figure 3B). As the heatmap shows in Figure 2B, Mmp9 was expressed by all 3 subsets, but the expression level was the highest in Inf-Cs (Figure 3C).
**Figure 3:** *Differentially expressed inflammatory factors among inflammatory, inflammatory senescent and senescent cells. (A) Differential expression of SASP factors identified in aged callus tissues. (B) Differential expression of cytokines based on the gene list in a mouse cytokine array. (C) Differential expression of the pro-inflammatory protease, Mmp9, among top 10 DEGs in
Figure 2B
.*
## Cell-cell communication analyses reveals that inflammatory senescent and senescent cells regulate inflammatory cells
To investigate how Inf-Cs, Inf-SCs and SCs interact with each other, we first applied Cellchat analysis, a software that can predict potential ligand-receptor interactions, based on differential expression levels of ligand/receptors pairs and numbers of interactions [15]. Cellchat analysis indicated that the strength of interaction was much greater when Inf-SCs and SCs were set as sender/source cells and Inf-Cs were set as recipient cells than other source-recipient pairs (Figure 4A), indicating that Inf-Cs are likely the target cells of Inf-SCs and SCs via ligand-receptor interaction. The chord diagram showed ligand-receptor pairs expressed by SCs and Inf-SCs to target Inf-Cs, and the major pairs included Anxa-Fpr1/Fpr2 and C3-(Itgam+Itgb2) (Figure 4B). We found that 31 ligand-receptor pairs were expressed by Inf-SC or SC subsets to target the Inf-C subset (Figure 4C). Some ligands interacted with more than one receptor, including Tgfβ1 (Transforming Growth *Factor beta* 1), Spp1 (Secreted Phosphoprotein 1), Ncam1 (Neural Cell Adhesion Molecule 1), Icam1 and 2 (Intercellular adhesion molecule 1 and 2), and Fn1 (Fibronectin-1). Interestingly, apart from TGFβ1, which we reported to contribute to delayed fracture healing in aged mice [3], most elevated ligand genes encode cell adhesion proteins (Figure 4D).
**Figure 4:** *Inflammatory senescent and senescent cells regulate inflammatory cells via ligand/receptor interaction. Cellchat R software was used for predicting ligand-receptor interaction among Inf-Cs, Inf-SCs and SCs. (A) Circle plot showing the interaction strength predicted by Cellchat among three subclusters with interaction numbers. (B) Chord plot showing main ligand-receptors by cell-cell communication targeting Inf-Cs. The thickness of arrows is proportional to the interaction strength between ligand-receptor pairs. (C) Potential ligand-receptor interaction between Inf-SCs or SCs and Inf-Cs. Red indicates the ligand-receptor pairs detected only in Inf-SCs vs. Inf-Cs. Green indicates the ligand-receptor pairs detected only in SCs vs. Inf-Cs. Black indicates the ligand-receptor pairs detected in both Inf-SCs vs. Inf-Cs and SCs vs. Inf-Cs. Communication probability and p-values are indicated by circle size and color. (D) Violin plot showing the expression level of predicted ligands among Inf-Cs, Inf-SCs and SCs.*
We also performed Nichenet analysis, a software that can both predict ligands from sender cells and target genes from receiver cells with an intercellular communication process of interest [20]. We set the Inf-SCs and SCs as sender/niche cells and Inf-Cs as receiver cells, based on the results from Cellchat (Figure 4A). Nichenet analysis predicted 62 ligands (Supplemental Figure 2). The top 18 ligands expressed by Inf-SCs and SCs are listed in Figures 5A, B, some of which were similar to those detected by Cellchat in Figure 4C, including adhesion molecules, Itgβ1 (Integrin Subunit Beta 1), C3 (Complement component 3), Itgαm (the integrin alpha M chain), Sell (Selectin L), and TNF family members, Tnfsf13 (TNF Superfamily Member 13). Ligand-target gene matrix analysis detected numerous target genes under the regulation of ligands from Inf-SCs and SCs (Figure 5C). Some of the target genes are reported to be highly relevant to inflammatory responses, such as Cebpβ (CCAAT/enhancer-binding protein beta) [26, 27]. GO analysis of predicted target genes from all 62 ligands revealed several biological processes related to cell adhesion, immunity and inflammation, indicating that the ligands expressed by Inf-SCs and SCs could drive inflammatory response in Inf-Cs (Figure 5D).
**Figure 5:** *Potential ligands from inflammatory senescent and senescent cells and potential target genes in inflammatory cells. Nichenet software was used for predicting ligand expression by SCs and Inf-SCs, and target genes in Inf-Cs. (A) Top 18 ligands expressed by SCs and Inf-SCs. (B) Expression of top 18 ligands in SCs and Inf-SCs. (C) Ligand-target prediction between ligands from Inf-SCs and SCs and target genes in Inf-Cs. (D) Top 6 upregulated pathways by gene ontology analysis of predicted target genes.*
## Senescent cells promote an inflammatory phenotype in callus-derived mesenchymal progenitors
Based on the results of cell-cell commination analyses in Figures 4 and 5, we hypothesized that SCs and Inf-SCs produce secretory factors that have paracrine effects on callus-derived mesenchymal progenitors (CaMPCs) to induce them to develop an inflammatory phenotype, e.g. express inflammatory genes, which can be blocked by senolytic drugs. We collected conditioned medium (CM) from callus pieces that were isolated from young and aged mice in the presence and absence of the senolytic drugs, dasatinib and quercetin. The rationale for using CM from young and aged callus pieces is that we previously reported significantly increased SC numbers in callus of aged mice [3]. We treated CaMPCs with CM and examined the expression of the inflammatory genes, Rela and Mmp9 (Figure 6A). In comparison with CM from young mice, CM from aged mice increased the expression of Rela and Mmp9, which was prevented by dasatinib+quercetin (Figure 6B), indicating that CM from aged mice induces an inflammatory phenotype in CaMPCs by a SC-mediated mechanism.
**Figure 6:** *Senescent cells promote inflammatory gene expression of callus-derived mesenchymal progenitors. Young and aged mice were sacrificed at 10 dpf, and callus tissues were used to generate senescent CM. (A) Callus pieces were cultured in the presence of senolytic drugs, 200 nM dasatinib (D) + 20 μM quercetin (Q) for 2 d to generate CM. CaMPCs were treated with 30% of CM for 2 d. (B) Expression levels of inflammation-associated gene, Rela, and SASP factor, Mmp9, in CaMPCs examined by qPCR. Relative mRNA expression is the fold-change versus young vehicle-treated cells as 1. n=3 wells. Repeated once. Two-way ANOVA followed by Tukey post-hoc test. Data represent mean ± SD. *p<0.05 vehicle- versus D+Q-treated cells.*
To determine if Inf-Cs induced by SC and Inf-SC have decreased osteoblast differentiation capacity, we used NF-κB-GFP reporter mice that enable us to isolate cells with high NF-κB (=GFP+ cells) as inflammatory cells [17]. We collected CM from H2O2-induced SCs, treated CaMPCs from NF-κB-GFP reporter mice, and sorted GFP+ and GFP- cells (Figure 7A; Supplemental Figure 3). Compared to cells that were treated with CM from PBS-treated calluses, about $12\%$ cells that were treated with CM from H2O2-treated calluses became GFP+ cells, indicating that NF-κB signaling was activated by factors produced by SCs (data not shown). The GFP+ cells had a small, but significantly reduced cell growth compared to GFP- cells (Figure 7B). After cells were cultured in osteoblast differentiation medium for 2 weeks, GFP+ cells had significantly lower alkaline phosphatase (ALP) staining (Figure 7C) and expression levels of genes associated to both osteoblast differentiation (Osx, Runx2) and mineralization (Osteocalin-Bglap, Ostepontin-Spp1) than GFP- cells (Figure 7D). These data indicate that inflammatory stromal cells induced by SC CM have impaired osteoblast differentiation and mineralization ability.
**Figure 7:** *NF-κB-GFP+ callus-derived mesenchymal progenitors have reduced osteoblast function. Young NF-κB-GFP mice were sacrificed at 10 dpf, and CaMPCs were generated from callus cultures. (A) CaMPCs isolated from NF-κB-GFP mice were treated with senescent CM and subjected to FACS as GFP+ and GFP- cells. Control (ctrl) group is the untreated/unsorted CaMPCs isolated from NF-κB-GFP mice, which we used as GFP negative cells for sorting GFP- and GFP+ cells following treatment. (B) Cell growth assessed by a CCK8 kit. n=3 wells. Repeated once. (C) Osteoblast differentiation assessed by ALP staining. n=3 wells. Repeated once. (D) Expression levels of genes associated to osteoblast differentiation and mineralization determined by qPCR. Relative mRNA expression is the fold-change vs. control as 1. n=3 wells. Repeated once. Data represent mean ± SD. One-way ANOVA followed by Tukey post-hoc test. *p<0.05 ctrl versus GFP+ cells; #p<0.05 GFP+ versus GFP- cells.*
## Discussion
SCs produce SASPs, which include some pro-inflammatory factors produced by Inf-Cs. Thus, *It is* important to distinguish SCs from Inf-Cs because different drugs would be used to inhibit their adverse effects in some clinical conditions. The big discovery of this study is that Inf-Cs are transcriptionally different from SCs. In the current study, we analyzed scRNA-seq data from aged callus stromal cells and demonstrated that the cells include Inf-C, Inf-SC and SC subsets, based on their gene expression profiles. Cell-cell communication analyses predicted that SCs are the predominant cells that affect Inf-Cs. Further, our in vitro cell culture experiments indicated that SCs promote an inflammatory phenotype in callus mesenchymal progenitor cells, and these inflammatory callus stromal cells have reduced intrinsic osteoblast differentiation capacity. Thus, the big discovery of our study is that Inf-Cs are transcriptionally different from SCs. Our data suggest that prevention of inflammatory phenotypes in mesenchymal progenitor cells by blocking SCs may serve as an additional mechanism for senolytic therapy-enhanced fracture healing.
Why is it important to distinguish Inf-Cs from SCs? In the elderly, inflammation and senescence often co-exist due to age-associated chronic inflammation [28, 29] and age-associated cellular senescence [30, 31]. However, inflammation and senescence have different causes, cellular processes, and therapies. Inflammatory processes are associated with immune responses and activation of NF-κB signaling in living organisms, including short-term acute inflammation and long-term chronic inflammation. Chronic inflammation is a low-grade unresolved immune response that occurs during aging, known as inflammaging, and is a risk factor for many aging-related diseases [32]. Different from inflammation, cellular senescence is stimulated by multiple stress signals, including replicative senescence, DNA damage, programmed development senescence, and oncogene activation [33]. Senescence-associated pathways are associated with activation of the cell cycle regulators, p53/p21WAF1/CIP1 and p16INK4A/pRB, and NF-κB and c/EBPβ are important downstream transcription factors [24, 34, 35]. While both processes produce pro-inflammatory cytokines, Coleman et al. reported that senescent endothelial cells may have a protective role to limit local inflammatory responses [36]. Indicative of the mixed nature of the cellular processes in inflammation and senescence, anti-inflammation treatment was reported to decrease senescence and increase osteogenic function in skeletal stem progenitor cells [9], and senolytic drugs had diminished anti-inflammatory effects on atherosclerotic lesions [37].
Senolytic drugs that kill SC cells have been used in 12 clinical trials for age-associated conditions, including skeletal diseases. For example, Khosla et al. are conducting a phase-2 clinical trial in healthy women over 70 years of age, in which subjects will take the senolytic drugs, dasatinib+quercetin or fisetin in a 3 day-on and 28 day-off cycle for 5 cycles. Serum bone turnover makers, C-terminal telopeptide of type I collagen (bone resorption) and A procollagen type I intact N terminal propeptide (bone formation), will be measured before and after the treatment as primary outcome measures (ClinicalTrials.gov Identifier: NCT04313634). This trial will be closed by April 2023. However, although there is no clinical trial examining the effects of senolytics in human fracture repair, recent studies reported that dasatinib+quercetin promote fracture healing in young [17] and aged mice [3].
Unlike consistent reports of beneficial effects of senolytic drugs in fracture healing [38, 39], conflicting findings have been reported for inflammatory drugs (NSAIDs) in fracture healing, despite their analgesic potency being well established [40]. A systematic review of all available literature, including animal and clinical studies, reported a great diversity in the data presented in the studies [41], leaving clinicians puzzled over potential safety issues. Thus, well-randomized prospective clinical trials are warranted.
Chronic inflammation has been shown to induce senescence of mesenchymal stem/progenitor cells in fracture healing and rheumatoid arthritis [9, 42]. Josephson et al. reported that mesenchymal stem/progenitor cells in middle-aged mice have an increased senescent phenotype and that NSAID treatment can improve fracture healing in middle-aged mice [9]. However, emerging evidence also indicates that senescence is an upstream source of inflammatory factors [37, 43]. Biavasco et al. reported that oncogene-activated senescence of human hematopoietic stem and progenitor cells promotes systemic inflammation by secreting TNFα [37, 43]. In the present study, we observed that treatment of CaMPCs with SC CM only induced their inflammatory phenotype, but not cellular senescence (Supplemental Figure 4). We suspect that either the SASP induced by H2O2 alone was not enough to trigger senescence of CaMPCs or the concentration of SASPs in the CM was too low to induce senescence of CaMPCs.
Can we distinguish Inf-Cs from SCs since both cell types express the same SASP factors? We initially attempted to use SASP expression to distinguish Inf-Cs from SCs in our data set, such as Tgfβ1, Tnfα, Il1a, Il1b, Il4, and Il6, but failed because there is no difference in the expression between Inf-Cs and SCs. We selected cells expressing NF-κB p65 as Inf-Cs because NF-κB is a well-known inducer of inflammation [22]. Under our experimental conditions, most of the NF-κB+ cells did not stain positively for β-galactosidase, a marker of SCs (Supplemental Figure 4). The finding that callus SCs also express genes associated with ROS and DNA damage can help to distinguish them from Inf-Cs.
Although it is generally accepted that inflammatory stimuli trigger cellular senescence [25, 44], the relationship between senescence and inflammation still be the “chicken or egg conundrum”. Our study demonstrated or partially demonstrated that SCs are different from Inf-Cs and SCs inflamed callus stromal cells, which caused the decreased osteogenic ability of these Inf-stromal cells. Our *Cellchat analysis* predicts that callus SCs and Inf-SCs affect Inf-Cs, based on the finding that SCs and Inf-SCs produce more ligands/factors than Inf-Cs (Figure 4A). Thus, it is possible that within a combined aging and injury micro-environment where the number of both Inf-Cs and SCs are increased, SCs likely function as effector cells to influence Inf-Cs. In addition, *Nichenet analysis* identified several target genes in Inf-Cs that may be regulated by ligands expressed by SCs and Inf-SCs, such as Itgβ1/Cebpβ, and Gip1/Mcl1 (Figure 5C). C/EBPβ is a transcription factor often involved in inflammation [45, 46]. Mcl1 (Myeloid Cell Leukemia 1), a member of the Bcl-2 family for maintaining cell viability [47], can act as a molecular switch for double-strand break DNA repair [48] perhaps by having an anti-senescence effect.
We also found that NF-κB+ CaMPCs have reduced osteoblast differentiation capacity. This is not surprising because several mechanisms by which NF-κB activation inhibits mesenchymal progenitor cell differentiation to osteoblasts have been reported. For example, conditional deletion of TRAF3 (TNF Receptor Associated Factor 3), a negative regulatory protein of NF-κB in Prx1-expressing mesenchymal progenitor cells, inhibited osteoblast formation by promoting degradation of β-catenin [49]. Activation of the Wnt-β-catenin signaling pathway is negatively regulated by formation of a β-catenin destruction complex composed of the proteins, APC (*Adenomatous polyposis* coli), Annexin and GSK3β (Glycogen synthase kinase 3 β) [50]. Interestingly, we found high expression of Anxa1, Anxa2, Gsk3β, Apc in Inf-Cs (Supplemental Figure 5), suggesting that the down-regulation of β-catenin signaling in inflammatory CaMPCs may be related to high expression of these negative regulators.
Limitations of the current study include the following. Since we used Rela/Relb-expressing cells to define Inf-Cs, it is possible that Inf-Cs are heterogenous with different biologic functions. More study is needed to investigate interactions between Inf-Cs and SCs and further sub-cluster analysis of Inf-Cs with a pro-senescent phenotype, such as pro-inflammatory factor-expressing Inf-Cs. We need to determine if these cell subsets exist in vivo. We currently do not have a set of surface markers for flow cytometry or immunostaining to identify these potential subsets, which also limits further determination of their function. We found Inf-SCs, cells that express genes related to both inflammation and cellular senescence. These Inf-SCs appear to have upregulated pathways similar to SCs. The origin of Inf-SCs is not clear and it is not known if they represent an intermediary state between Inf-Cs and SCs. Our trajectory analysis suggests that Inf-SCs do not originate from Inf-Cs (Supplemental Figure 6). Functional differences among these subsets need to be investigated. Another limitation is that we are unable to confirm our findings in patients with bone fracture, which will require high quality of scRNA-seq data from human callus cells. In our scRNAseq dataset, we did not detect the expression of genes related to chondrogenesis (Sox9, Col2a1, Col10a1). We suspect that the digestion method used in this study does not allow us to isolate cells with chondrogenic property from callus cells. In the future, we should modify our isolation protocol to collect cells with chondrogenic property from callus cells and investigate cellular senescence in this cell population.
In summary, callus SCs not only express SASP factors, but also genes related to ROS and DNA damage, which distinguishes them from Inf-Cs. Bioinformatic analyses predict strong interactions between Inf-Cs and SCs and potential influence of SCs on Inf-Cs by producing active ligands, although further analysis of Inf-C sub-populations is needed to elucidate the contribution of Inf-Cs to the increased cellular senescence in aged bone fracture. Despite the limitations, this study has initially answered our original question: what is the difference between inflammatory cells and SCs in aged callus: they have different gene signatures, SCs produce more factors than Inf-Cs, and SCs play a dominant role in driving inflammation and decreased osteogenic capacity in callus stromal cells.
## 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 University of Rochester Committee for Animal Resources (protocol number: 2001-121R).
## Author contributions
JL, XL, AM, HZ, BB, and LX designed the study. JL, YY, and HZ performed experiments. JL, XL, AM, and YY performed bioinformatic analyses. JL and LX wrote the original manuscript. JL, XL, LX wrote the original rebuttal for resubmission. 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.2023.1090049/full#supplementary-material
## References
1. Berry DC JY, Arpke RW, Close EL, Uchida A, Reading D, Berglund ED. **Cellular aging contributes to failure of cold-induced beige adipocyte formation in old mice and humans**. *Cell Metab* (2017) **25**. DOI: 10.1016/j.cmet.2016.10.023
2. Liu X, Wan M. **A tale of the good and bad: Cell senescence in bone homeostasis and disease**. *Int Rev Cell Mol Biol* (2019) **346** 97-128. DOI: 10.1016/bs.ircmb.2019.03.005
3. Liu J, Zhang J, Lin X, Boyce BF, Zhang H, Xing L. **Age-associated callus senescent cells produce tgf-Beta1 that inhibits fracture healing in aged mice**. *J Clin Invest* (2022) **132**. DOI: 10.1172/JCI148073
4. Saul D, Monroe DG, Rowsey JL, Kosinsky RL, Vos SJ, Doolittle ML. **Modulation of fracture healing by the transient accumulation of senescent cells**. *Elife* (2021) **10**. DOI: 10.1101/2021.05.18.444618
5. Franceschi C, Campisi J. **Chronic inflammation (Inflammaging) and its potential contribution to age-associated diseases**. *J Gerontol A Biol Sci Med Sci* (2014) **69**. DOI: 10.1093/gerona/glu057
6. Ferrucci L, Fabbri E. **Inflammageing: Chronic inflammation in ageing, cardiovascular disease, and frailty**. *Nat Rev Cardiol* (2018) **15**. DOI: 10.1038/s41569-018-0064-2
7. Rapp AE, Bindl R, Recknagel S, Erbacher A, Muller I, Schrezenmeier H. **Fracture healing is delayed in immunodeficient Nod/Scid−Il2rgamma cnull mice**. *PloS One* (2016) **11**. DOI: 10.1371/journal.pone.0147465
8. Maruyama M, Rhee C, Utsunomiya T, Zhang N, Ueno M, Yao Z. **Modulation of the inflammatory response and bone healing**. *Front Endocrinol (Lausanne)* (2020) **11**. DOI: 10.3389/fendo.2020.00386
9. Josephson AM, Bradaschia-Correa V, Lee S, Leclerc K, Patel KS, Muinos Lopez E. **Age-related inflammation triggers skeletal Stem/Progenitor cell dysfunction**. *Proc Natl Acad Sci U.S.A.* (2019) **116** 6995-7004. DOI: 10.1073/pnas.1810692116
10. Chuang PY, Shen SH, Yang TY, Huang TW, Huang KC. **Non-steroidal anti-inflammatory drugs and the risk of a second hip fracture: A propensity-score matching study**. *BMC Musculoskelet Disord* (2016) **17** 201. DOI: 10.1186/s12891-016-1047-2
11. Wheatley BM, Nappo KE, Christensen DL, Holman AM, Brooks DI, Potter BK. **Effect of nsaids on bone healing rates: A meta-analysis**. *J Am Acad Orthop Surg* (2019) **27**. DOI: 10.5435/JAAOS-D-17-00727
12. Al Farii H, Farahdel L, Frazer A, Salimi A, Bernstein M. **The effect of nsaids on postfracture bone healing: A meta-analysis of randomized controlled trials**. *Ota Int* (2021) **4**. DOI: 10.1097/OI9.0000000000000092
13. Lin X, Zhang H, Liu J, Wu CL, Mcdavid A, Boyce BF. **Aged callus skeletal Stem/Progenitor cells contain an inflammatory osteogenic population with increased irf and nf-kappab pathways and reduced osteogenic potential**. *Front Mol Biosci* (2022) **9**. DOI: 10.3389/fmolb.2022.806528
14. Stuart T, Butler A, Hoffman P, Hafemeister C, Papalexi E, Mauck WM. **Comprehensive integration of single-cell data**. *Cell* (2019) **177** 1888-1902.E21. DOI: 10.1016/j.cell.2019.05.031
15. Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan CH. **Inference and analysis of cell-cell communication using cellchat**. *Nat Commun* (2021) **12** 1088. DOI: 10.1038/s41467-021-21246-9
16. Browaeys R, Saelens W, Saeys Y. **Nichenet: Modeling intercellular communication by linking ligands to target genes**. *Nat Methods* (2020) **17**. DOI: 10.1038/s41592-019-0667-5
17. Everhart MB, Han W, Sherrill TP, Arutiunov M, Polosukhin VV, Burke JR. **Duration and intensity of nf-kappab activity determine the severity of endotoxin-induced acute lung injury**. *J Immunol* (2006) **176** 4995-5005. DOI: 10.4049/jimmunol.176.8.4995
18. Brown ML, Yukata K, Farnsworth CW, Chen DG, Awad H, Hilton MJ. **Delayed fracture healing and increased callus adiposity in a C57bl/6j murine model of obesity-associated type 2 diabetes mellitus**. *PloS One* (2014) **9**. DOI: 10.1371/journal.pone.0099656
19. Flor AC, Doshi AP, Kron SJ. **Modulation of therapy-induced senescence by reactive lipid aldehydes**. *Cell Death Discovery* (2016) **2** 16045. DOI: 10.1038/cddiscovery.2016.45
20. Flor A, Pagacz J, Thompson D, Kron S. **Far-red fluorescent senescence-associated beta-galactosidase probe for identification and enrichment of senescent tumor cells by flow cytometry**. *J Vis Exp* (2022) **187**. DOI: 10.3791/64176
21. Ding Q, Liu H, Liu L, Ma C, Qin H, Wei Y. **Deletion of P16 accelerates fracture healing in geriatric mice**. *Am J Transl Res* (2021) **13**
22. Liu T, Zhang L, Joo D, Sun SC. **Nf-kappab signaling in inflammation**. *Signal Transduct Target Ther* (2017) **2** 17023. DOI: 10.1038/sigtrans.2017.23
23. Mademtzoglou D, Asakura Y, Borok MJ, Alonso-Martin S, Mourikis P, Kodaka Y. **Cellular localization of the cell cycle inhibitor Cdkn1c controls growth arrest of adult skeletal muscle stem cells**. *Elife* (2018) **7**. DOI: 10.7554/eLife.33337
24. Kumari R, Jat P. **Mechanisms of cellular senescence: Cell cycle arrest and senescence associated secretory phenotype**. *Front Cell Dev Biol* (2021) **9**. DOI: 10.3389/fcell.2021.645593
25. Coppe JP, Desprez PY, Krtolica A, Campisi J. **The senescence-associated secretory phenotype: The dark side of tumor suppression**. *Annu Rev Pathol* (2010) **5** 99-118. DOI: 10.1146/annurev-pathol-121808-102144
26. Meng X, Lu P, Bai H, Xiao P, Fan Q. **Transcriptional regulatory networks in human lung adenocarcinoma**. *Mol Med Rep* (2012) **6**. DOI: 10.3892/mmr.2012.1034
27. Du C, Pan P, Jiang Y, Zhang Q, Bao J, Liu C. **Microarray data analysis to identify crucial genes regulated by cebpb in human Snb19 glioma cells**. *World J Surg Oncol* (2016) **14** 258. DOI: 10.1186/s12957-016-0997-z
28. Sanada F, Taniyama Y, Muratsu J, Otsu R, Shimizu H, Rakugi H. **Source of chronic inflammation in aging**. *Front Cardiovasc Med* (2018) **5**. DOI: 10.3389/fcvm.2018.00012
29. Chung HY, Kim DH, Lee EK, Chung KW, Chung S, Lee B. **Redefining chronic inflammation in aging and age-related diseases: Proposal of the senoinflammation concept**. *Aging Dis* (2019) **10**. DOI: 10.14336/AD.2018.0324
30. Mchugh D, Gil J. **Senescence and aging: Causes, consequences, and therapeutic avenues**. *J Cell Biol* (2018) **217** 65-77. DOI: 10.1083/jcb.201708092
31. Di Micco R, Krizhanovsky V, Baker D, D'adda Di Fagagna F. **Cellular senescence in ageing: From mechanisms to therapeutic opportunities**. *Nat Rev Mol Cell Biol* (2021) **22** 75-95. DOI: 10.1038/s41580-020-00314-w
32. Straub RH, Schradin C. **Chronic inflammatory systemic diseases: An evolutionary trade-off between acutely beneficial but chronically harmful programs**. *Evol Med Public Health* (2016) **2016** 37-51. DOI: 10.1093/emph/eow001
33. Dodig S, Cepelak I, Pavic I. **Hallmarks of senescence and aging**. *Biochem Med (Zagreb)* (2019) **29** 030501. DOI: 10.11613/BM.2019.030501
34. Hardy K, Mansfield L, Mackay A, Benvenuti S, Ismail S, Arora P. **Transcriptional networks and cellular senescence in human mammary fibroblasts**. *Mol Biol Cell* (2005) **16**. DOI: 10.1091/mbc.e04-05-0392
35. Kuilman T, Michaloglou C, Vredeveld LC, Douma S, Van Doorn R, Desmet CJ. **Oncogene-induced senescence relayed by an interleukin-dependent inflammatory network**. *Cell* (2008) **133**. DOI: 10.1016/j.cell.2008.03.039
36. Coleman PR, Chang G, Hutas G, Grimshaw M, Vadas MA, Gamble JR. **Age-associated stresses induce an anti-inflammatory senescent phenotype in endothelial cells**. *Aging (Albany Ny)* (2013) **5**. DOI: 10.18632/aging.100622
37. Childs BG, Baker DJ, Wijshake T, Conover CA, Campisi J, Van Deursen JM. **Senescent intimal foam cells are deleterious At all stages of atherosclerosis**. *Science* (2016) **354**. DOI: 10.1126/science.aaf6659
38. Jeon OH, Kim C, Laberge RM, Demaria M, Rathod S, Vasserot AP. **Local clearance of senescent cells attenuates the development of post-traumatic osteoarthritis and creates a pro-regenerative environment**. *Nat Med* (2017) **23**. DOI: 10.1038/nm.4324
39. Novais EJ, Tran VA, Johnston SN, Darris KR, Roupas AJ, Sessions GA. **Long-term treatment with senolytic drugs dasatinib and quercetin ameliorates age-dependent intervertebral disc degeneration in mice**. *Nat Commun* (2021) **12** 5213. DOI: 10.1038/s41467-021-25453-2
40. Geusens P, Emans PJ, De Jong JJ, Van Den Bergh J. **Nsaids and fracture healing**. *Curr Opin Rheumatol* (2013) **25**. DOI: 10.1097/BOR.0b013e32836200b8
41. Pountos I, Georgouli T, Calori GM, Giannoudis PV. **Do nonsteroidal anti-inflammatory drugs affect bone healing? a critical analysis**. *Scientificworldjournal* (2012) **2012** 606404. DOI: 10.1100/2012/606404
42. Lee HJ, Lee WJ, Hwang SC, Choe Y, Kim S, Bok E. **Chronic inflammation-induced senescence impairs immunomodulatory properties of synovial fluid mesenchymal stem cells in rheumatoid arthritis**. *Stem Cell Res Ther* (2021) **12** 502. DOI: 10.1186/s13287-021-02453-z
43. Biavasco R, Lettera E, Giannetti K, Gilioli D, Beretta S, Conti A. **Oncogene-induced senescence in hematopoietic progenitors features myeloid restricted hematopoiesis, chronic inflammation and histiocytosis**. *Nat Commun* (2021) **12** 4559. DOI: 10.1038/s41467-021-24876-1
44. Zhu X, Chen Z, Shen W, Huang G, Sedivy JM, Wang H. **Inflammation, epigenetics, and metabolism converge to cell senescence and ageing: The regulation and intervention**. *Signal Transduct Target Ther* (2021) **6** 245. DOI: 10.1038/s41392-021-00646-9
45. Zahid MDK, Rogowski M, Ponce C, Choudhury M, Moustaid-Moussa N, Rahman SM. **Ccaat/Enhancer-binding protein beta (C/Ebpbeta) knockdown reduces inflammation, er stress, and apoptosis, and promotes autophagy in oxldl-treated Raw264.7 macrophage cells**. *Mol Cell Biochem* (2020) **463**. DOI: 10.1007/s11010-019-03642-4
46. Chen C, Zhou Y, Wang H, Alam A, Kang SS, Ahn EH. **Gut inflammation triggers C/Ebpbeta/Delta-Secretase-Dependent gut-To-Brain propagation of abeta and tau fibrils in alzheimer's disease**. *EMBO J* (2021) **40**. DOI: 10.15252/embj.2020106320
47. Mandelin AM, Pope RM. **Myeloid cell leukemia-1 as a therapeutic target**. *Expert Opin Ther Targets* (2007) **11**. DOI: 10.1517/14728222.11.3.363
48. Widden H, Placzek WJ. **The multiple mechanisms of Mcl1 in the regulation of cell fate**. *Commun Biol* (2021) **4** 1029. DOI: 10.1038/s42003-021-02564-6
49. Li J, Ayoub A, Xiu Y, Yin X, Sanders JO, Mesfin A. **Tgfbeta-induced degradation of Traf3 in mesenchymal progenitor cells causes age-related osteoporosis**. *Nat Commun* (2019) **10** 2795. DOI: 10.1038/s41467-019-10677-0
50. Wu D, Pan W. **Gsk3: A multifaceted kinase in wnt signaling**. *Trends Biochem Sci* (2010) **35**. DOI: 10.1016/j.tibs.2009.10.002
|
---
title: The Tankyrase Inhibitor OM-153 Demonstrates Antitumor Efficacy and a Therapeutic
Window in Mouse Models
authors:
- Shoshy A. Brinch
- Enya Amundsen-Isaksen
- Sandra Espada
- Clara Hammarström
- Aleksandra Aizenshtadt
- Petter A. Olsen
- Lone Holmen
- Merete Høyem
- Hanne Scholz
- Gunnveig Grødeland
- Sven T. Sowa
- Albert Galera-Prat
- Lari Lehtiö
- Ilonka A.T.M. Meerts
- Ruben G.G. Leenders
- Anita Wegert
- Stefan Krauss
- Jo Waaler
journal: Cancer Research Communications
year: 2022
pmcid: PMC9981206
doi: 10.1158/2767-9764.CRC-22-0027
license: CC BY 4.0
---
# The Tankyrase Inhibitor OM-153 Demonstrates Antitumor Efficacy and a Therapeutic Window in Mouse Models
## Abstract
The catalytic enzymes tankyrase 1 and 2 (TNKS$\frac{1}{2}$) alter protein turnover by poly-ADP-ribosylating target proteins, which earmark them for degradation by the ubiquitin–proteasomal system. Prominent targets of the catalytic activity of TNKS$\frac{1}{2}$ include AXIN proteins, resulting in TNKS$\frac{1}{2}$ being attractive biotargets for addressing of oncogenic WNT/β-catenin signaling. Although several potent small molecules have been developed to inhibit TNKS$\frac{1}{2}$, there are currently no TNKS$\frac{1}{2}$ inhibitors available in clinical practice. The development of tankyrase inhibitors has mainly been disadvantaged by concerns over biotarget-dependent intestinal toxicity and a deficient therapeutic window. Here we show that the novel, potent, and selective 1,2,4-triazole–based TNKS$\frac{1}{2}$ inhibitor OM-153 reduces WNT/β-catenin signaling and tumor progression in COLO 320DM colon carcinoma xenografts upon oral administration of 0.33–10 mg/kg twice daily. In addition, OM-153 potentiates anti–programmed cell death protein 1 (anti–PD-1) immune checkpoint inhibition and antitumor effect in a B16-F10 mouse melanoma model. A 28-day repeated dose mouse toxicity study documents body weight loss, intestinal damage, and tubular damage in the kidney after oral–twice daily administration of 100 mg/kg. In contrast, mice treated oral–twice daily with 10 mg/kg show an intact intestinal architecture and no atypical histopathologic changes in other organs. In addition, clinical biochemistry and hematologic analyses do not identify changes indicating substantial toxicity. The results demonstrate OM-153–mediated antitumor effects and a therapeutic window in a colon carcinoma mouse model ranging from 0.33 to at least 10 mg/kg, and provide a framework for using OM-153 for further preclinical evaluations.
### Significance:
This study uncovers the effectiveness and therapeutic window for a novel tankyrase inhibitor in mouse tumor models.
## Introduction
The ADP-ribosyltransferases family members tankyrase 1 and 2 (TNKS$\frac{1}{2}$) are protein-modifying enzymes at the crossroad of multiple cellular pathways (1–7). Using the redox metabolite NAD+, TNKS$\frac{1}{2}$ catalyze a posttranslational modification termed poly(ADP-ribosyl)ation which serves as a recognition signal for E3 ligase–mediated polyubiquitination followed by proteasomal degradation (1, 8–10). TNKS$\frac{1}{2}$ poly(ADP-ribosyl)ate several target proteins (1–7, 11–13), of which the proteins AXIN1, AXIN2 (AXIN$\frac{1}{2}$), and angiomotin (AMOT), angiomotin like 1 and 2 (AMOTL$\frac{1}{2}$) have received the largest attention as they are central in regulating the activity of the cellular wingless-type mammary tumor virus integration site (WNT)/β-catenin and Hippo signaling pathways, respectively [14, 15]. TNKS$\frac{1}{2}$ inhibitor (TNKSi) treatment leads to stabilization of AXIN$\frac{1}{2}$, and hence stabilized β-catenin degradasomes, which in turn enhances degradation of the transcriptional regulator β-catenin resulting in inhibition of WNT/β-catenin signaling [14, 16, 17]. In Hippo signaling, TNKSi-mediated AMOT protein stabilization facilitates changes in the subcellular location of the transcription cofactors yes-associated protein 1 YAP (YAP) and WW domain containing transcription regulator 1 (WWTR1/TAZ), leading to reduced YAP signaling [7, 15, 18]. The WNT/β-catenin and Hippo signaling pathways are involved in a multitude of disease conditions, including tumorigenesis and tumor immune evasion (19–21). Hence, significant efforts have been made to develop selective TNKS$\frac{1}{2}$ inhibitors (14, 22–29). In particular, inhibitors based on the 1,2,4-triazole scaffold, such as JW74, OD336, G007-LK, OM-1700, and OM-153, target the adenosine binding pocket of the TNKS$\frac{1}{2}$ catalytic domains with high selectivity, and are therefore able to display selectivity over other PARP enzymes (22, 27, 30–32).
Clinical tankyrase inhibitor development has so far been hampered by concerns about target-specific and signaling pathway–specific side effects, in particular intestinal toxicity [5, 33, 34]. Despite this, the tankyrase inhibitors E7449 and STP1002 have entered clinical trials in the cancer arena (35–38). This supports the potential of TNKS$\frac{1}{2}$ inhibitors and justifies the continued development of drugs directed toward TNKS$\frac{1}{2}$ inhibition. Recently, we developed the small-molecule 1,2,4-triazole-based TNKSi OM-153 through iterative design–make–test cycles possessing favorable drug properties. These properties include efficacy, off-target liabilities, absorption/distribution/metabolism/excretion (ADME) parameters, and pharmacokinetic profile in mice [32]. Here, we evaluate the biological properties and potential toxicity issues of OM-153 in detail. First, we tested the efficacy of OM-153 in a panel of tumor cell lines showing low-nanomolar range biotarget engagement and tumor cell growth inhibition. Next, we carried out a standardized dose-escalating experiment using the COLO 320DM colon carcinoma xenograft model, showing significant multidose tumor growth inhibition, and also in an isogenic B16-F10 mouse melanoma model. Subsequently, the oral single-dose maximum tolerated dose (MTD) for OM-153 was established. The MTD was followed by a two-dose (10 mg/kg and 100 mg/kg), oral and twice daily 28-day repeated dose mouse toxicity evaluation, including multiorgan histopathology, biochemistry, and hematology in mice. With the applied methodology used in mice, we show no significant adverse toxicity in mice dosed oral–twice daily with 10 mg/kg, and anti–colon carcinoma efficacy when dosed oral–twice daily at 0.33–10 mg/kg. Hence, in contrast to the earlier studies [33, 34], our data propose the existence of a therapeutic window for OM-153 ranging from 0.33 to at least 10 mg/kg in a mouse colon carcinoma model.
## Cell Culture and In Vitro Experiments
The cell lines where obtained from ATCC or the Japanese Collection of Research Bioresources (JCRB) in 2014 and culturing was performed as described previously [7]. The cell cultures were kept below 20 passages (∼10 weeks) and routinely monitored (upon thawing and monthly) for Mycoplasma infections with MycoAlert Mycoplasma Detection Kit (Lonza). All cell lines were authenticated by short tandem repeat profiling to confirm their identity (Eurofins). TNKS$\frac{1}{2}$ and PARP1 biochemical assays [27, 32], luciferase-based WNT/β-catenin signaling pathway reporter assay [22, 27, 32], anti-proliferative assays, and NCI-60 tumor cell line panel screen [7], immunoblot analyses [7, 32, 39], RNA isolation, and real-time qRT-PCR [7, 27, 39, 40] were performed as previously described. Structured illumination microscopy (SIM) was performed as previously described [7, 39] using the following additional primary antibody: AXIN1 (2087, RRID: AB_2274550, Cell Signaling Technology) and displayed as maximum intensity projections rendered from 20 Z planes spanning 3.68 μm in total.
## Animal Experiments
The pharmacokinetic analyses were performed according to the standard protocols of Medicilon as previously described [22, 27, 32]. Three CD-1 mice (999M-018, Sino-British SIPPR/BK Lab Animal) were treated by intravenous (i.v) (0.4 or 2 mg/kg) or oral (10 or 100 mg/kg) administrations of OM-153 (Symeres), once or twice daily (dosing at experiment start and after 6 hours), using $5\%$ DMSO (D2650, Sigma-Aldrich), $50\%$ PEG400 (06855, Sigma-Aldrich), and $45\%$ saline as vehicle. The same vehicle and oral–twice daily treatment regime were used for all other experiments. Samples were collected after 15 and 30 minutes plus 1, 2, 4, and 6 hours for group 1, and after 6.5, 8, 10, and 24 hours for group 2.
The experiment using COLO 320DM (CCL-220, RRID: CVCL_0219, ATCC) xenografts was carried out as previously described [40] at Reaction Biology. On day 3 after subcutanous (s.c.) tumor challenge in CB17-SCID mice (CRL: 561, RRID: IMSR_CRL: 561, Charles River), tumor-bearing animals (mean tumor volume: 12 mm3) were randomized into six groups (all $$n = 10$$). The day after, the animals were treated oral–twice daily with 10, 3.3, 1, 0.33, or 0.1 mg/kg OM-153 or vehicle until day 37. Eight mice, evenly distributed between the treatment groups, were euthanized for ethical reasons before experiment termination, due to skin ulcerations in the tumor area (6 mice) or body weight loss (>$20\%$, two mice). Next, the mice were sacrificed, approximately 4 hours after first dose of the day. Protein and RNA extracts were prepared from four moderately sized tumors within each treatment group. Mass spectrometry analyses of OM-153 in plasma and tumors were performed according to the standard protocols of Pharmacelsus.
The experiment using B16-F10 tumors was carried out as previously described [39] at Reaction Biology. On day six after tumor challenge (s.c.) in C57BL/6N mice (CRL: 493, RRID: IMSR_CRL: 493, Charles River), tumor-bearing animals (mean tumor volume: 31 mm3) were randomized into six groups (all $$n = 15$$). On the same day, the animals were treated oral–twice daily with vehicle, anti-programmed cell death 1 (anti–PD-1; 10 mg/kg i.p. on day 6, 9, and 12, BE0033–2, RRID: AB_1107747, Bio X Cell), 10 mg/kg OM-153, or combined treatment with anti–PD-1 and 10, 1, and 0.1 mg/kg OM-153 until day 20. Fifteen mice, distributed between the treatment groups, were euthanized for ethical reasons before experiment termination, due to skin ulcerations in the tumor area. Next, the mice were sacrificed, approximately 8 hours after last dosing. Protein and RNA extracts were prepared from 8 tumors within each treatment group.
The dose escalation experiment was performed in single male CD-1 mice (CRL: 022, RRID: IMSR_CRL: 022, Charles River) at BSL BIOSERVICE/Eurofins using their standard protocol. The animals were treated oral–twice daily with escalating doses of 500, 1,000, 1,500, and 2,000 mg/kg OM-153, and observed for 48 hours for clinical signs (7 days for the verification animal, 2,000 mg/kg). Body weight was recorded daily.
The 28-day oral repeated dose toxicity study was performed at Reaction Biology (10 mg/kg) and at Oslo University Hospital (100 mg/kg) using male CD-1 (CRL: 022, RRID: IMSR_CRL: 022, Charles River) and C57BL/6J (JAX: 000664, RRID: IMSR_JAX: 000664, Jackson) mice, respectively. The mice were treated oral–twice daily with 10 ($$n = 7$$) or 100 ($$n = 4$$) mg/kg OM-153 or vehicle control (both experiments, $$n = 4$$) until day 28, or study termination (for 100 mg/kg). Body weight and food consumption were measured thrice weekly. For the experiment using 10 mg/kg oral–twice daily dosing, blood was collected at the end of the experiment and hematologic and clinical biochemistry analyses were performed at IDEXX according to their standard protocols. Duodenum, jejunum, ileum, colon–rectum, lymph nodes, pancreas, heart, kidney, liver, spleen, lung, prostate, and testis were collected, formalin-fixed ($10\%$), paraffin-embedded, stained with hematoxylin and eosin (H&E), and imaged as previously described [39]. Staining of ileums using immunofluorescence was performed as previously described [29, 39]. Donkey anti-rabbit IgG cyanine Cy3 was used as secondary antibody [(550 nm), 1:500, 711165152, RRID: AB_2307443, Jackson ImmunoResearch] for 60 minutes at 37°C. DAPI nuclear dye [(340 nm), 1 μg/μL, D9542, Sigma Aldrich] was added to the final washing solution. RNAscope multiplex fluorescent reagent kit v2 detection (323100, Advanced Cell Diagnostics) was used to detect expression of mouse leucine-rich repeat-containing G-coupled receptor 5 (Lgr5) mRNA (312171, Advanced Cell Diagnostics) according to the manufacturer's instructions. Imaging was performed using confocal microscopy as described previously [39]. All animal experiments were approved by local animal experiment authorities (ethics committee of the Chinese Association for Laboratory Animal Sciences, German Centre for the Protection of Laboratory Animals, and Norwegian Food Safety Authority), and were carried out in compliance with FELASA guidelines and recommendations.
## Quantification and Statistical Analysis
Statistical analyses were performed as previously described using t tests and Mann–Whitney rank sum tests [7].
## Data Availability Statement
The data generated in this study are available within the article and its Supplementary Data files. *Materials* generated in this study can be made available upon request to the corresponding author.
## OM-153 Specifically Inhibits TNKS1/2, WNT/β-Catenin Signaling, and Cell Growth of an APC-Mutated Colon Carcinoma Cell Line
OM-153 was recently developed and shows an overall favorable drug property profile (ref. 32; Fig. 1A). To evaluate potency and specificity, OM-153 was tested in biochemical TNKS$\frac{1}{2}$ and PARP1 assays, and in a luciferase-based WNT/β-catenin signaling reporter assay. OM-153 was compared with a selected TNKSi panel consisting of NVP-TNKS565 [24], AZ6102 [25], compound 40 [23], IWR-1 [28], XAV939 [14], E7449 [35], and previously identified 1,2,4-triazole-based TNKS$\frac{1}{2}$ inhibitors (Supplementary Fig. S1; refs. 22, 27, 31). OM-153 that binds to the adenosine site of the TNKS$\frac{1}{2}$ catalytic domains [32], decreased the activities of TNKS1 and TNKS2 with IC50 values (concentrations resulting in $50\%$ inhibition) of 13 and 2.0 nmol/L, respectively, approaching the technical limit of the assay due to required protein concentration (Fig. 1B; Supplementary Fig. S2A and S2B). The IC50 value for PARP1 inhibition was >100,000 nmol/L (Fig. 1B; Supplementary Fig. S2C). In contrast, compounds that bind to the nicotinamide pocket also inhibited PARP1 activity (Fig. 1B; Supplementary Fig. S2C). OM-153 inhibited luciferase-based WNT/β-catenin signaling reporter activity with an IC50 value of 0.63 nmol/L (Fig. 1B; Supplementary Fig. S2D).
**FIGURE 1:** *OM-153 specifically inhibits TNKS1/2, WNT/β-catenin signaling and cell growth of an APC-mutated colon cancer cell line. A, Chemical structure of OM-153. B, IC50 values for inhibition of TNKS1, TNKS2, and PARP1 (biochemical assays) and luciferase-based WNT/β-catenin signaling reporter assay in HEK293 cells (WNT reporter). The table is sorted in ascending order according to the WNT reporter IC50 values. Mean ± SD values from a minimum of three independent experiments are shown. For B and C, numbers in superscript indicate data from reference 22, 27, 31, and 32. As indicated, compounds inhibit TNKS1/2 by binding to the catalytic domain, either in the nicotinamide pocket or in adenosine pocket, or in both (dual binder). C, MTS colorimetric cell growth assay for various concentrations of tankyrase inhibitors in treated COLO 320DM and RKO colon cancer cells. GI50 and GI25 values were calculated relative to controls (100%, 0.01% DMSO) and experiment time 0 values (t0, set to 0%) after 5 days of cultivation. Data from one representative experiment of more than three repeated assays, each with six replicates, are shown.*
Next, in comparison with the selected TNKSi panel, the potential of OM-153 as an antiproliferative agent was tested in a cell growth assay in the adenomatous polyposis coli (APC)-mutated, WNT/β-catenin signaling–dependent, and TNKSi-sensitive colon carcinoma cell line COLO 320DM [7, 32, 33, 40]. To evaluate the specificity of the cell growth inhibition, the APC–wild-type, WNT/β-catenin signaling–independent and TNKSi-insensitive colon carcinoma cell line RKO was used as a control [7, 29, 32]. Treatment with OM-153 decreased cell growth in COLO 320DM cells with a GI50 value of 10 nmol/L and a GI25 value of 2.5 nmol/L (concentrations resulting in $50\%$ and $25\%$ growth inhibition, respectively), while cell growth in RKO cells was insubstantially affected by the treatment (GI50 and GI25 values >10,000 nmol/L; Fig. 1C; Supplementary Fig. S3).
In conclusion, in addition to our recent report [32], the results demonstrate that OM-153 is a highly potent and specific inhibitor of TNKS$\frac{1}{2}$, WNT/β-catenin signaling and cell growth in the APC-mutated colon carcinoma cell line COLO 320DM.
## OM-153 Inhibits WNT/β-Catenin, YAP, and MYC Signaling and Shows an Antiproliferative Effect in Human Cancer Cell Lines
Recently, we showed that TNKS$\frac{1}{2}$ inhibition could attenuate cell growth in a wide range of cancer types in vitro [7]. To evaluate cancer cell line growth inhibition by OM-153, the NCI-60 tumor cell line panel was screened. Of the 60 tested cancer cell lines, 16 cell lines showed >$25\%$ relative growth inhibition upon treatment with 10 nmol/L OM-153 (Fig. 2A; Supplementary Fig. S4). These cell lines originated from lung, brain, ovary, and kidney, and included the previously identified TNKSi-sensitive cell lines OVCAR-4 (ovarian cancer; ref. 7) and UO-31 (renal cancer; ref. 7). These two cell lines along with the highly TNKSi-sensitive cell line ABC-1 (non–small cell lung cancer; ref. 7), were treated with various doses of OM-153 for five days. Treatment with OM-153 showed a cytotoxic effect in ABC-1 cells with a GI50 value of 2.0 nmol/L and a GI25 value of 1.5 nmol/L (Fig. 2B). In OVCAR-4 and UO-31 cells, cytostatic effects were observed with GI25 values of 2.5 and 3.5 nmol/L, respectively (Fig. 2B). In conclusion, low nanomolar concentrations of OM-153 can reduce cell growth in a subset of cancer cell lines in vitro.
**FIGURE 2:** *OM-153 inhibits WNT/β-catenin, YAP, and MYC signaling and shows an antiproliferative effect in human cancer cell lines. A, NCI-60 human cancer cell line proliferation/viability screen using OM-153 (10 nmol/L) for 48 hours and ≥ 25% relative growth inhibition is highlighted by gray bars (control, 0.01% DMSO = 0%). Prostate (Pro) and central nervous system (CNS). B, MTS colorimetric cell growth assay for various concentrations of OM-153 in treated ABC-1, OVCAR-4, and UO-31 cells. GI50 and GI25 values (nmol/L) were calculated relative to control (100%, 0.01% DMSO) and experiment time 0 values (Abs 490 t0, set to 0%) after 5 days of cultivation. Mean ± SD values for data from one representative experiment of more than three repeated assays, each with six replicates, are shown. C, Evaluation of WNT/β-catenin signaling. Left, immunoblots of cytoplasmic AXIN1 and nuclear active form of β-catenin [non-phospho (Ser33/37/Thr41)] and total β-catenin. Lamin B1 loading controls are duplicated for OVCAR-4 immunoblots. Right, real-time RT-qPCR analysis of the WNT/β-catenin signaling target gene AXIN2. For C–E, after 24 hours of treatment with OM-153 (10 nmol/L) or controls (0.0001% DMSO) in COLO 320DM, UO-31, OVCAR-4, ABC-1, and RKO. Immunoblots show representative data from two or more independent experiments. Actin (cytoplasm) and lamin B1 (nucleus) were used as loading controls. Boxplots show median, first, and third quartiles and maximum and minimum whiskers for combined data from three independent experiments with three replicates each. Stippled lines (black) depict control mean values = 1. Two-tailed t tests: ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Mann–Whitney rank sum tests: ‡, P < 0.01. ns., not significant. D, Evaluation of YAP signaling. Left, immunoblots of cytoplasmic AMOTL2 (top) and nuclear YAP and TAZ (bottom). Right, real-time RT-qPCR analysis of the YAP signaling target genes AMOTL2, CCN1, and CNN2. For D and E, lamin B1 loading controls are duplicated for RKO immunoblots (YAP/TAZ and CCND1). E, Evaluation of MYC signaling. Left, immunoblots of nuclear MYC and CCND1. Lamin B1 loading controls are duplicated for ABC-1 immunoblots. Right, real-time RT-qPCR analysis of the MYC signaling target genes MYC and CCND1.*
Recently, we showed that TNKS$\frac{1}{2}$ inhibition cell-type dependently can block WNT/β-catenin and/or YAP signaling, and consequently MYC proto-oncogene bHLH transcription factor (MYC) signaling, resulting in attenuated cancer cell growth [7]. Hence, immunoblot and real-time qRT-PCR analyses were conducted to examine the effect of treatment with 10 nmol/L OM-153 against these signaling pathways in a panel of cells consisting of the TNKSi-sensitive cell lines COLO 320DM (Fig. 1C; Supplementary Fig. S3), UO-31, OVCAR-4, and ABC-1 [7]. The TNKSi-insensitive colon carcinoma cell line RKO was included as a control (Fig. 1C; Supplementary Fig. S3; refs. 7, 29). First, the effects of OM-153 treatment on WNT/β-catenin signaling components were tested. OM-153 stabilized the β-catenin degradasome–forming protein AXIN1 in all cell lines (Fig. 2C). A reduction of the transcriptionally active form of β-catenin in the nuclei was only documented in COLO 320DM and ABC-1 cells (Fig. 2C; Supplementary Fig. S5A). To investigate OM-153–mediated control of WNT/β-catenin signaling, real-time qRT-PCR analysis of the cell type–independent and universal target gene AXIN2 was performed. The analysis showed significantly reduced transcription of AXIN2, demonstrating attenuation of WNT/β-catenin signaling in COLO 320DM, OVCAR-4 and ABC-1 cells (Fig. 2C). Moreover, SIM microscopy revealed the formation of cytoplasmic puncta containing AXIN1 and colocalizing β-catenin, indicating accumulation of β-catenin degradasomes [7, 17] in COLO 320DM cells (Fig. 3).
**FIGURE 3:** *OM-153 induces formation of β-catenin and AXIN1-containing puncta and reduces nuclear β-catenin in COLO 320DM cells. Representative SIM images of β-catenin and AXIN1 after 24 hours of treatment with OM-153 (10 nmol/L) or controls (0.0001% DMSO) in COLO 320DM. Left, single-channel images in grayscale. Right, merged images of β-catenin (green), AXIN1 (magenta), and nuclear DAPI staining (blue). Arrows highlight colocalization of β-catenin and AXIN1. Scale bars, 10 μm and 2 μm (magnification).*
Next, the effect of OM-153 treatment on YAP signaling was examined. Consistent with earlier research [7], an immunoblot analysis showed cytoplasmic stabilization of the TNKS$\frac{1}{2}$ target protein AMOTL2 in all cell lines upon OM-153 treatment, while nuclear YAP accumulation was only seen in UO-31, OVCAR-4, and ABC-1 cells (Fig. 2D; Supplementary Fig. S5B). Moreover, the real-time qRT-PCR analysis showed reduced transcription of the YAP signaling target genes AMOTL2, cellular communication network factor 1 and 2 (CCN1 and CCN2) in all TNKSi-sensitive cell lines (Fig. 2D).
Finally, the effect of OM-153 treatment on MYC expression was assessed. Immunoblot and real-time qRT-PCR analyses showed that MYC and cyclin D1 (CCND1) protein, as well as transcription of MYC and the MYC signaling target gene CCND1, were decreased in all TNKSi-sensitive cell lines (Fig. 2E; Supplementary Fig. S5C).
In conclusion, OM-153 functions as a potent cell type–dependent inhibitor of WNT/β-catenin, YAP and MYC signaling and can reduce cell growth in a subset of human cancer cell lines in cell culture.
## OM-153 Inhibits WNT/β-Catenin Signaling and Shows Antitumor Effect in a Human Colon Carcinoma Xenograft Model
A mouse pharmacokinetic analysis was designed to evaluate drug exposure for the mouse efficacy and toxicity studies using oral administration of 10 or 100 mg/kg OM-153, dosed twice daily at the start of the experiment and after 6 hours. After the second dose of 10 mg/kg OM-153, a plasma concentration (Cmax2) of 2,700 ng/mL (5.3 μmol/L) was observed. Although OM-153 was below the detection threshold after 24 hours, these results suggest a satisfactory timeframe with an efficacious concentration in vivo relative to the COLO 320DM in vitro GI50 value (5.1 ng/mL = 10 nmol/L; Fig. 4A; Supplementary Fig S6A). No corrections for plasma- and tissue-protein binding were performed for the in vitro experiments. At a 100 mg/kg oral–twice daily dosing, a Cmax2 of 20,000 ng/mL (39.2 μmol/L) was observed, and 1,900 ng/mL (3.7 μmol/L) were still detected after 24 hours (Fig. 4A; Supplementary Fig. S6A).
**FIGURE 4:** *OM-153 inhibits WNT/β-catenin signaling and shows antitumor effect in a human colon carcinoma xenograft model. A, Mouse pharmacokinetic profile upon intravenous (i.v., 0.4 or 2 mg/kg) or oral (p.o., 10 or 100 mg/kg) administration of OM-153 either once (group 1) or twice daily (BID, group 2) in CD-1 mice. Half-life (t1/2#, calculated by setting the terminal phase from 8 hours to 24 hours for oral dosing, and 6.5 hours to 24 hours for intravenous dosing), time to achieve first and second dose Cmax (tmax1 and 2), maximum concentration reached (Cmax1 and 2), AUC, mean residence time (MRT), volume of distribution (Vz or Vz/F), clearance (Cl or Cl/F), and fraction absorbed/bioavailability (F). B, Tumor end volume (mm3), left. Relative body weights, right. Mean body weight at experiment initiation is set at 0%. ** indicates statistically significant body weight reductions after treatment with 3.3 and 1 mg/kg OM-153. Mean ± SD values are shown. For B–E, from COLO 320DM-challenged (s.c.) CB17-SCID mice upon treatment with 10 (n = 8), 3.3 (n = 10), 1 (n = 9), 0.33 (n = 9), or 0.1 (n = 9) mg/kg OM-153 and vehicle control (n = 7) from day 4 through 37 (all oral–twice daily). Boxplots show median, first, and third quartiles and maximum and minimum whiskers. One-tailed t tests: **, P < 0.01; *, P < 0.05. One-tailed Mann–Whitney rank sum tests: ‡, P < 0.01; and †, P < 0.05. For C–E, n = 4 tumors, collected 4 hours post last dosing, were analyzed for each treatment group. For D and E, stippled lines depict control mean values = 1. C, Mass spectrometry analyses of OM-153 in plasma (nmol/L, left) and tumor (nmol/kg, right). D, Representative and quantified protein immunoblot ratios (protein vs. loading control) showing altered expression of AXIN1, AXIN2 and β-catenin (total). E, Real-time RT-qPCR analyses of WNT/β-catenin signaling target genes DKK1, APCDD1, and NKD1.*
To evaluate whether the in vitro anti-proliferative effect of OM-153 could be translated to an in vivo setting, COLO 320DM tumor-challenged CB17-SCID mice were treated oral–twice daily with five doses of OM-153. Doses in the range of 0.33–10 mg/kg resulted in $74\%$–$85\%$ tumor end volume inhibition when compared with the control. Importantly, no changes in body weights that would be suggestive for toxicity were observed among mice receiving 10 mg/kg oral–twice daily (Fig. 4B; Supplementary Fig. S6B–S6E).
Next, mass spectrometry analyses were performed at the experimental endpoint to measure the concentrations of OM-153 in plasma and tumors. The analyses revealed a dose-dependent detection of OM-153 above 10 nmol/L for doses in the range of 0.33–10 mg/kg in both plasma and tumors (Fig. 4C). To evaluate WNT/β-catenin signaling biomarkers tumors were analyzed by immunoblot and real-time qRT-PCR. OM-153 treatment stabilized AXIN1 and AXIN2 proteins in the tumor specimen, while β-catenin protein and the WNT/β-catenin signaling target genes dickkopf WNT signaling pathway inhibitor 1 (DKK1), APC downregulated 1 (APCDD1), and NKD inhibitor of WNT signaling pathway 1 (NKD1) were dose-dependently reduced (Fig. 4D and E; Supplementary Fig. S6E).
In summary, the results show that oral–twice daily dosing of 0.33–10 mg/kg OM-153 can attenuate WNT/β-catenin signaling, showing a pronounced antitumor efficacy in the COLO 320DM colon carcinoma mouse xenograft model.
## Combined OM-153 and Anti–PD-1 Treatment Confers Antitumor Effect in Mouse Melanoma
Recently, we reported that TNKS$\frac{1}{2}$ inhibition potentiated the effect of PD-1 antibody–based immunotherapy against murine B16-F10 melanoma tumors, an anti–PD-1–resistant model showing β-catenin–induced immune evasion [39]. The study documented that the combined treatment effect was dependent on loss of β-catenin in the tumor cells and induction of a CD8+ T cell–mediated adaptive antitumor immune response [39].
To examine the effect of combining OM-153 with anti-PD-1 treatment, B16-F10 tumors were established in immunocompetent C57BL/6N mice, followed by oral–twice daily treatment using various doses of OM-153, combined with three treatments of anti–PD-1 delivered three days apart. Similar to our previous report, OM-153 and anti–PD–1 monotherapy controls did not exert any significant antitumor effects (ref. 39; Fig. 5A; Supplementary Fig. S7A–S7C). However, when OM-153 was dosed in the range of 0.1–10 mg/kg in combination with anti–PD-1 treatment, $51\%$–$65\%$ tumor end volume inhibition was documented when compared with the vehicle control (Fig. 5A; Supplementary Fig. S7A–S7C). No alterations in body weights were observed in any of the treatment groups (Fig. 5A).
**FIGURE 5:** *Combined OM-153 and anti–PD-1 treatment confers antitumor effect in mouse melanoma. A, Tumor end volume (mm3), left. Relative body weights, right. Mean body weight at experiment initiation is set at 0%. Mean ± SD values are shown. For A–C, from B16-F10-challenged (s.c.) C57BL/6N mice upon treatment from day 6 through 20. Vehicle control (n = 12), 10 mg/kg anti-(α)PD-1 dosed intraperitoneally on day 6, 9, and 12 (n = 13), 10 mg/kg OM-153 (n = 11), or combined treatment with αPD-1 and 10 (n = 13), 1 (n = 13), and 0.1 (n = 13) mg/kg OM-153. All OM-153 treatments, oral–twice daily. Boxplots show median, first and third quartiles and maximum and minimum whiskers. One-tailed t tests: ***, P < 0.001; **, P < 0.01; and *, P < 0.05. One-tailed Mann–Whitney rank sum test: †, P < 0.05. B, Representative and quantified protein immunoblot ratios (protein vs. loading control) showing altered expression of AXIN1 and the active form of β-catenin [non-phospho (Ser33/37/Thr41)]. For B and C, tumor analysis upon treatment with vehicle control and 10 mg/kg OM-153 (both n = 8), collected 8 hours post last dosing, and stippled lines depict control mean values = 1. C, Real-time RT-qPCR analyses of the WNT/β-catenin signaling target gene Axin2.*
Tumors treated oral–twice daily with 10 mg/kg OM-153 exhibited significant stabilization of AXIN1 protein, but showed insignificant downregulation of β-catenin in an immunoblot analysis (Fig. 5B; Supplementary Fig. S7D). The real-time qRT-PCR analysis showed significantly decreased transcription of the target gene Axin2, indicating attenuated WNT/β-catenin signaling (Fig. 5C).
In summary, OM-153 can decrease WNT/β-catenin signaling in B16-F10 tumors and sensitize B16-F10 melanoma tumors to anti-PD-1 immunotherapy.
## Treatment with 10 mg/kg OM-153 Twice Daily Is Tolerated in Mice
TNKS$\frac{1}{2}$ inhibition has been associated with the induction of intestinal toxicity linked to biotarget and WNT/β-catenin signaling pathway-specific effects [33, 34]. Hence, we performed a dose escalation experiment to evaluate acute toxicity and the MTD of OM-153 in mice. No clinical signs of systemic toxicity were observed in any animals after oral–twice daily dosing with 500, 1,000, 1,500, or 2,000 mg/kg of OM-153 (Supplementary Fig. S8A).
To evaluate the long-term effects of OM-153 treatment, a 28-day oral repeated dose toxicity study was carried out using oral–twice daily dosing of 100 and 10 mg/kg. In mice dosed with 100 mg/kg, reduced activity and body weight loss in 3 of 4 mice were documented, and the mice were euthanized for ethical reasons on day 7 (Fig. 6A; Supplementary Fig. S8B). In contrast, in mice dosed with 10 mg/kg, no clinical signs, body weight loss, or reduced food consumption were observed (Fig. 6B; Supplementary Fig. S8C and S8D). Next, duodenum, jejunum and ileum, colon-rectum, lymph nodes, pancreas, heart, kidney, liver, spleen, lung, prostate and testis were collected, stained with H&E and the histopathology was evaluated. In the small intestines of 2 out of three 100 mg/kg treated mice, loss of villi height, width, crypt, and as well as surface epithelium, accompanied by inflammation, was documented (Fig. 6C; Supplementary Fig. S8E). In the same mice, signs of acute tubular damage in the kidney, presumably caused by intestinal damage, were observed (Supplementary Fig. S8E). No abnormal histopathologic changes in the other organs were detected (Supplementary Fig. S8E). On the contrary, the overall intestinal architecture was intact in the 10 mg/kg group (Fig. 6D). Persistent proliferation in the crypt compartments was indicated by visualization of the proliferation and intestinal stem cell (ISC)-like marker antigen identified by mAb Ki 67 (MKI67; Fig. 6E). Keratin 20 (KRT20) staining specified presence of terminally differentiated epithelial cells in the villus termini (Fig. 6E). Similar to previous observations [33, 41], expression of the ISC marker and WNT/β-catenin signaling target gene Lgr5 was reduced (Fig. 6F). No atypical histopathologic changes in other organs were observed (Supplementary Fig. S8E). Blood was collected at the end of the experiment using 10 mg/kg dosing to perform clinical biochemistry and hematologic analyses. The clinical biochemistry analysis did not suggest any signs of liver damage as indicated by insignificant alteration of aspartate transaminase (AST), alanine transaminase (ALT), and glutamate dehydrogenase (GLDH) levels (Fig. 7A). However, significant decreases in magnesium, potassium, and triglycerides could indicate a poorer gastrointestinal absorption or altered kidney function (Fig. 7A). The hematologic analysis suggested no signs of inflammation, although a tendency toward reduced leukocytes and reticulocytes, as well as a moderate but significant reduction in erythrocytes and hemoglobin was noted, which may indicate moderately altered hematopoiesis (Fig. 7B). These alterations are possibly caused by changes in WNT/β-catenin and YAP signaling activities, as both pathways are known regulators of hematopoiesis [42, 43].
**FIGURE 6:** *Oral–twice daily treatment with 10 mg/kg OM-153 does not induce intestinal toxicity in mice. A, Relative body weights upon 7 days of treatment with 100 mg/kg OM-153 (n = 4) and vehicle control (n = 4) in C57BL/6J mice. For A and B, mean body weight at experiment initiation is set at 0%. All treatments, oral–twice daily (p.o. BID). Means ± SDs are shown and two-tailed t test: *, P < 0.05. n.s., not significant. B, Relative body weights upon 28 days of treatment with 10 mg/kg OM-153 (n = 7) and vehicle control (n = 4) in CD-1 mice. C, Treated as described in A and stained with H&E. For zoom images, arrowheads highlight intact intestinal villi (gray) and crypts (black), while asterixes highlight loss of villi (gray) and crypts (black). For C–F, representative pictures, taken from multiple sections of ileums. For C–E, scale bars: 100 μm (original magnification × 200). D, Stained with H&E. For D–F, treated as described in B. E, MKI67 (magenta) and KRT20 (red) visualized using immunofluorescence. F,Lgr5 (magenta) visualized using RNAscope. Scale bars, 100 μm (original magnification ×100).* **FIGURE 7:** *Clinical biochemistry and hematologic analyses of mice administered oral–twice daily dosing with 10 mg/kg OM-153 in a 28-day toxicity evaluation. A, Clinical biochemistry analysis. Aspartate aminotransferase (AST), alanine transaminase (ALT), gamma-glutamyl transferase (GGT), glutamate dehydrogenase (GLDH), blood urea nitrogen (BUN), grams/liter (g/L), units/liter (U/L), micromolar (μmol/L), and millimolar (mmol/L). For A and B, upon 28 days of treatment with 10 mg/kg OM-153 (n = 7) or vehicle control (n = 4), both ora–twice daily (p.o. BID) dosing in CD-1 mice. Boxplots show median, first, and third quartiles and maximum and minimum whiskers. Two-tailed t tests: ***, P < 0.001; **, P < 0.01; *, P < 0.05. Mann–Whitney rank sum tests: ‡, P < 0.01; †, P < 0.05. B, Hematologic analysis. Mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), MCH concentration (MCHC), 109 (giga)/liter (G/L), 1012 (tera)/liter (T/L), grams/deciliter (g/dL), femtoliter (fL), pictogram (pg), and per microliter (/μL). Basophils, band neutrophils, anisocytosis, or atypical cells were not detected.*
In conclusion, toxicity in mice at 100 mg/kg OM-153 oral–twice daily dosing is, at least in partially, caused by intestinal damage. In contrast, a 10 mg/kg OM-153 oral–twice daily dosing was overall tolerated in mice, and the intestinal architecture was intact despite a diminished Lgr5 expression in ISCs at the crypt the base.
## Discussion
Here we show that OM-153 is a highly potent and specific inhibitor of TNKS$\frac{1}{2}$, WNT/β-catenin, YAP and MYC signaling capable of reducing cell growth in a subset of human cancer cell lines. Oral–twice daily dosing of 0.33–10 mg/kg OM-153 in COLO 320DM mouse colon carcinoma xenografts attenuated WNT/β-catenin signaling and resulted in a $74\%$–$85\%$ tumor growth inhibition. Moreover, in the isogenic and immunocompetent B16-F10 mouse melanoma model, oral–twice daily dosing of 0.1–10 mg/kg OM-153 potentiated efficacy of anti–PD-1 treatment and resulted in a $51\%$–$65\%$ tumor growth inhibition.
Deregulated WNT/β-catenin signaling contributes to aberrant cell growth and carcinogenesis, but is in parallel also critical for stem cell renewal, proliferation, and differentiation, during both embryogenesis and tissue homeostasis [20, 44]. In the last decade, two studies using the early-generation TNKS$\frac{1}{2}$ inhibitors G007-LK and G-631, observed TNKSi-induced anti–colon carcinoma efficacies [33, 34]. These reports also describe severe side-effects at therapeutic doses including intestinal toxicity limiting the therapeutic window [33, 34, 45].
In the 28-day repeated dose mouse toxicity study, when dosing mice oral–twice daily with 100 mg/kg OM-153, we indeed documented rapid loss of body weight, intestinal epithelial degeneration, and inflammation, as well as tubular damage in the kidney. However, we were not able to detect substantial side-effects and toxicity in mice treated with oral–twice daily dosing using 10 mg/kg, and importantly, their intestinal architecture was intact. In the ileum of these mice, MKI67 staining documented continual proliferation in the crypts and KRT20 staining showed differentiated epithelial cells. Similar to our previous report, showing TNKSi-induced elimination of differentiated cells traced from WNT/β-catenin signaling–dependent LGR5+ ISCs [41], Lgr5 expression was lost and the result may propose the presence of a TNKSi-resistant ISC population. Hence, one may hypothesize that intestinal proliferation from dispensable LGR5+ ISCs [46] can be maintained by activated proliferation of an alternative source of stem cells. Quiescent, WNT/β-catenin signaling-independent, LGR5−/BMI1 proto-oncogene, polycomb ring finger (BMI1)+ and crypt +4-positioned ISCs have been described in that role earlier [41, 47, 48]. Comprehensive follow-up studies of possible TNKSi-resistant ISC populations are evidently required.
We note that the primary pharmacologic on-target effect for OM-153 is seen in the low nanomolar range, as indicated by the TNKS$\frac{1}{2}$, WNT/β-catenin signaling reporter IC50-values, and COLO 320DM GI50 value, while the micromolar range Cmax values for oral–twice daily dosing with 10 and 100 mg/kg OM-153 are more than 2 logs higher (5.3 and 39.2 μmol/L, respectively). The results propose that the documented toxicity is likely not coupled to on-target inhibition of TNKS$\frac{1}{2}$ and WNT/β-catenin signaling, but rather caused by concentration-dependent interactions with unknown off-targets, or optionally related to the physicochemical properties of OM-153 [45]. Further detailed mechanistic investigation of plausible off-target toxicity is considered necessary. The possible compatibility of the tankyrase biotarget, with advancing tankyrase-specific inhibitors toward clinical trials in the cancer arena, has recently been demonstrated by the initiation of clinical trials with the TNKSi STP1002 [37].
The therapeutic index is normally measured as the ratio of the highest exposure to the drug that results in no toxicity (i.e., toxic dose in $50\%$ of subjects, TD50) to the exposure that produces the desired effect (i.e., efficacious dose in $50\%$ of subjects, ED50; ref. 45). The therapeutic window for a drug is commonly known as the dose range that can treat a disease effectively without having toxic effects. When using monotherapy treatment, we document no adverse toxicity in mice treated with oral–twice daily dosing using 10 mg/kg OM-153, and a significant antitumor efficacy in the COLO 320DM colon carcinoma xenograft model when dosing mice oral–twice daily with 0.33–10 mg/kg OM-153. Together, these data indicate a therapeutic window ranging from 0.33 mg/kg to at least 10 mg/kg.
In conclusion, in our experimental setting, we show that OM-153 exhibits antitumor efficacy and a viable therapeutic window in mouse models, which rationalize further preclinical evaluations to translate TNKS$\frac{1}{2}$ inhibition to a human therapeutic setting.
## Authors’ Disclosures
G. Grødeland reports personal fees from AstraZeneca, Sanofi, Bayer, Thermo Fisher, and personal fees from Norwegian System for compensation to patients outside the submitted work. S.T. Sowa reports grants from Jane and Aatos Erkko Foundation during the conduct of the study. L. Lehtiö reports a patent to WO$\frac{2019}{243822}$ issued and a patent to WO$\frac{2022}{008896}$ issued. R.G. Leenders reports a patent to WO$\frac{2019}{243822}$ issued and a patent to WO$\frac{2022}{008896}$ issued (the drug development project is $100\%$ owned by the Oslo University Hospital and administered by the TTO Inven2; Symeres Inc. has a scientific and BD collaboration agreement with Inven2). A. Wegert reports a patent to WO$\frac{2019}{243822}$ issued and a patent to WO$\frac{2022}{008896}$ issued (the drug development project is $100\%$ owned by the Oslo University hospital and administered by the TTO Inven2; Symeres B.V. has a scientific and BD collaboration agreement with Inven2). S. Krauss reports a patent to WO$\frac{2019}{243822}$ issued and a patent to WO$\frac{2022}{008896}$ issued (the drug development program is $100\%$ owned by the Oslo University Hospital and administered by the TTO Inven2; Symeres Inc. has a scientific and BD collaboration agreement with Inven2 on the drug development program). J. Waaler reports a patent to WO$\frac{2019}{243822}$ issued and a patent to WO$\frac{2022}{008896}$ issued (the drug development project is $100\%$ owned by the Oslo University Hospital and administered by the TTO Inven2). Our project collaborator Symeres Inc. has a scientific and BD collaboration agreement with Inven2. No disclosures were reported by the other authors.
## Authors’ Contributions
S.A. Brinch: Conceptualization, data curation, software, formal analysis, supervison, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. E. Amundsen-Isaksen: Data curation, formal analysis, validation, investigation, visualization, methodology, writing-review and editing. S. Espada: Data curation, formal analysis, validation, investigation, visualization, writing-review and editing. C. Hammarström: Data curation, formal analysis, validation, investigation, visualization, methodology, writing-review and editing. A. Aizenshtadt: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-review and editing. P.A. Olsen: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-review and editing. L. Holmen: Data curation, formal analysis, validation, investigation, visualization, writing-review and editing. M. Høyem: Data curation, formal analysis, validation, investigation, visualization, writing-review and editing. H. Scholz: Resources, supervision, funding acquisition, writing-review and editing. G. Grødeland: Resources, supervision, funding acquisition, writing-review and editing. S.T. Sowa: Data curation, software, formal analysis, validation, investigation, methodology, writing-review and editing. A. Galera-Prat: Data curation, software, formal analysis, validation, investigation, methodology, writing-review and editing. L. Lehtiö: Resources, data curation, software, formal analysis, supervision, funding acquisition, validation, methodology, writing-review and editing. I.A.T.M. Meerts: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-review and editing. R.G. G. Leenders: Conceptualization, data curation, software, formal analysis, validation, investigation, methodology, writing-review and editing. A. Wegert: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing; S. Krauss: Conceptualization, resources, supervision, funding acquisition, methodology, writing-original draft, project administration, writing-review and editing. J. Waaler: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing.
## References
1. Haikarainen T, Krauss S, Lehtio L. **Tankyrases: structure, function and therapeutic implications in cancer**. *Curr Pharm Des* (2014) **20** 6472-88. PMID: 24975604
2. Wang H, Kuusela S, Rinnankoski-Tuikka R, Dumont V, Bouslama R, Ramadan UA. **Tankyrase inhibition ameliorates lipid disorder via suppression of PGC-1α PARylation in db/db mice**. *Int J Obes (Lond)* (2020) **44** 1691-702. PMID: 32317752
3. Li N, Zhang Y, Han X, Liang K, Wang J, Feng L. **Poly-ADP ribosylation of PTEN by tankyrases promotes PTEN degradation and tumor growth**. *Genes Dev* (2015) **29** 157-70. PMID: 25547115
4. Kim S, Han S, Kim Y, Kim H-S, Gu Y-R, Kang D. **Tankyrase inhibition preserves osteoarthritic cartilage by coordinating cartilage matrix anabolism via effects on SOX9 PARylation**. *Nat Commun* (2019) **10** 4898. PMID: 31653858
5. Fujita S, Mukai T, Mito T, Kodama S, Nagasu A, Kittaka M. **Pharmacological inhibition of tankyrase induces bone loss in mice by increasing osteoclastogenesis**. *Bone* (2018) **106** 156-66. PMID: 29055830
6. Bhardwaj A, Yang Y, Ueberheide B, Smith S. **Whole proteome analysis of human tankyrase knockout cells reveals targets of tankyrase-mediated degradation**. *Nat Commun* (2017) **8** 2214. PMID: 29263426
7. Mygland L, Brinch SA, Strand MF, Olsen PA, Aizenshtadt A, Lund K. **Identification of response signatures for tankyrase inhibitor treatment in tumor cell lines**. *iScience* (2021) **24** 102807. PMID: 34337362
8. Nie L, Wang C, Li N, Feng X, Lee N, Su D. **Proteome-wide analysis reveals substrates of E3 ligase RNF146 targeted for degradation**. *Mol Cell Proteomics* (2020) **19** 2015-30. PMID: 32958691
9. Zhang Y, Liu S, Mickanin C, Feng Y, Charlat O, Michaud GA. **RNF146 is a poly(ADP-ribose)-directed E3 ligase that regulates axin degradation and Wnt signalling**. *Nat Cell Biol* (2011) **13** 623-9. PMID: 21478859
10. Callow MG, Tran H, Phu L, Lau T, Lee J, Sandoval WN. **Ubiquitin ligase RNF146 regulates tankyrase and Axin to promote Wnt signaling**. *PLoS One* (2011) **6** e22595. PMID: 21799911
11. Kim MK. **Novel insight into the function of tankyrase**. *Oncol Lett* (2018) **16** 6895-902. PMID: 30546421
12. Zimmerlin L, Zambidis ET.. **Pleiotropic roles of tankyrase/PARP proteins in the establishment and maintenance of human naive pluripotency**. *Exp Cell Res* (2020) **390** 111935. PMID: 32151493
13. Li N, Wang Y, Neri S, Zhen Y, Fong LWR, Qiao Y. **Tankyrase disrupts metabolic homeostasis and promotes tumorigenesis by inhibiting LKB1-AMPK signalling**. *Nat Commun* (2019) **10** 4363. PMID: 31554794
14. Huang S-MA, Mishina YM, Liu S, Cheung A, Stegmeier F, Michaud GA. **Tankyrase inhibition stabilizes axin and antagonizes Wnt signalling**. *Nature* (2009) **461** 614-20. PMID: 19759537
15. Wang W, Li N, Li X, Tran MK, Han X, Chen J. **Tankyrase Inhibitors Target YAP by Stabilizing Angiomotin Family Proteins**. *Cell Rep* (2015) **13** 524-32. PMID: 26456820
16. Lee E, Salic A, Krüger R, Heinrich R, Kirschner MW. **The roles of APC and Axin derived from experimental and theoretical analysis of the Wnt pathway**. *PLoS Biol* (2003) **1** E10. PMID: 14551908
17. Thorvaldsen TE, Pedersen NM, Wenzel EM, Schultz SW, Brech A, Liestøl K. **Structure, Dynamics, and Functionality of Tankyrase Inhibitor-Induced Degradasomes**. *Mol Cancer Res* (2015) **13** 1487-501. PMID: 26124443
18. Troilo A, Benson EK, Esposito D, Garibsingh R-AA, Reddy EP, Mungamuri SK. **Angiomotin stabilization by tankyrase inhibitors antagonizes constitutive TEAD-dependent transcription and proliferation of human tumor cells with Hippo pathway core component mutations**. *Oncotarget* (2016) **7** 28765-82. PMID: 27144834
19. Zheng Y, Pan D. **The hippo signaling pathway in development and disease**. *Dev Cell* (2019) **50** 264-82. PMID: 31386861
20. Nusse R, Clevers H. **Wnt/β-catenin signaling, disease, and emerging therapeutic modalities**. *Cell* (2017) **169** 985-99. PMID: 28575679
21. Luke JJ, Bao R, Sweis RF, Spranger S, Gajewski TF.. **WNT/β-catenin pathway activation correlates with immune exclusion across human cancers**. *Clin Cancer Res* (2019) **25** 3074-83. PMID: 30635339
22. Voronkov A, Holsworth DD, Waaler J, Wilson SR, Ekblad B, Perdreau-Dahl H. **Structural basis and SAR for G007-LK, a lead stage 1,2,4-triazole based specific tankyrase 1/2 inhibitor**. *J Med Chem* (2013) **56** 3012-23. PMID: 23473363
23. Bregman H, Chakka N, Guzman-Perez A, Gunaydin H, Gu Y, Huang X. **Discovery of novel, induced-pocket binding oxazolidinones as potent, selective, and orally bioavailable tankyrase inhibitors**. *J Med Chem* (2013) **56** 4320-42. PMID: 23701517
24. Shultz MD, Cheung AK, Kirby CA, Firestone B, Fan J, Chen CH-T. **Identification of NVP-TNKS656: the use of structure-efficiency relationships to generate a highly potent, selective, and orally active tankyrase inhibitor**. *J Med Chem* (2013) **56** 6495-511. PMID: 23844574
25. Johannes JW, Almeida L, Barlaam B, Boriack-Sjodin PA, Casella R, Croft RA. **Pyrimidinone nicotinamide mimetics as selective tankyrase and wnt pathway inhibitors suitable for in vivo pharmacology**. *ACS Med Chem Lett* (2015) **6** 254-9. PMID: 25815142
26. Mizutani A, Yashiroda Y, Muramatsu Y, Yoshida H, Chikada T, Tsumura T. **RK-287107, a potent and specific tankyrase inhibitor, blocks colorectal cancer cell growth in a preclinical model**. *Cancer Sci* (2018) **109** 4003-14. PMID: 30238564
27. Waaler J, Leenders RGG, Sowa ST, Alam Brinch S, Lycke M, Nieczypor P. **Preclinical lead optimization of a 1,2,4-triazole based tankyrase inhibitor**. *J Med Chem* (2020) **63** 6834-46. PMID: 32511917
28. Chen B, Dodge ME, Tang W, Lu J, Ma Z, Fan CW. **Small molecule-mediated disruption of Wnt-dependent signaling in tissue regeneration and cancer**. *Nat Chem Biol* (2009) **5** 100-7. PMID: 19125156
29. Waaler J, Machon O, Tumova L, Dinh H, Korinek V, Wilson SR. **A novel tankyrase inhibitor decreases canonical Wnt signaling in colon carcinoma cells and reduces tumor growth in conditional APC mutant mice**. *Cancer Res* (2012) **72** 2822-32. PMID: 22440753
30. Waaler J, Machon O, Von Kries JP, Wilson SR, Lundenes E, Wedlich D. **Novel synthetic antagonists of canonical Wnt signaling inhibit colorectal cancer cell growth**. *Cancer Res* (2011) **71** 197-205. PMID: 21199802
31. Anumala UR, Waaler J, Nkizinkiko Y, Ignatev A, Lazarow K, Lindemann P. **Discovery of a Novel series of tankyrase inhibitors by a hybridization approach**. *J Med Chem* (2017) **60** 10013-25. PMID: 29155568
32. Leenders RGG, Brinch SA, Sowa ST, Amundsen-Isaksen E, Galera-Prat A, Murthy S. **Development of a 1,2,4-triazole-based lead tankyrase inhibitor: part II**. *J Med Chem* (2021) **64** 17936-49. PMID: 34878777
33. Lau T, Chan E, Callow M, Waaler J, Boggs J, Blake RA. **A novel tankyrase small-molecule inhibitor suppresses APC mutation-driven colorectal tumor growth**. *Cancer Res* (2013) **73** 3132-44. PMID: 23539443
34. Zhong Y, Katavolos P, Nguyen T, Lau T, Boggs J, Sambrone A. **Tankyrase inhibition causes reversible intestinal toxicity in mice with a therapeutic index < 1**. *Toxicol Pathol* (2016;) **44** 267-78. PMID: 26692561
35. Mcgonigle S, Chen Z, Wu J, Chang P, Kolber-Simonds D, Ackermann K. **E7449: A dual inhibitor of PARP1/2 and tankyrase1/2 inhibits growth of DNA repair deficient tumors and antagonizes Wnt signaling**. *Oncotarget* (2015) **6** 41307-23. PMID: 26513298
36. Plummer R, Dua D, Cresti N, Drew Y, Stephens P, Foegh M. **First-in-human study of the PARP/tankyrase inhibitor E7449 in patients with advanced solid tumours and evaluation of a novel drug-response predictor**. *Br J Cancer* (2020) **123** 525-33. PMID: 32523090
37. **First-in-human dose-escalation study of STP1002 in patients with advanced-stage solid tumors**
38. **An open-label, multicenter, phase 1/2 study of Poly(ADP-Ribose) polymerase (PARP) Inhibitor E7449 as single agent in subjects with advanced solid tumors or with B-cell malignancies and in combination with Temozolomide (TMZ) or with carboplatin and paclitaxel in subjects with advanced solid tumors**
39. Waaler J, Mygland L, Tveita A, Strand MF, Solberg NT, Olsen PA. **Tankyrase inhibition sensitizes melanoma to PD-1 immune checkpoint blockade in syngeneic mouse models**. *Commun Biol* (2020) **3** 196. PMID: 32332858
40. Solberg NT, Waaler J, Lund K, Mygland L, Olsen PA, Krauss S. **TANKYRASE inhibition enhances the antiproliferative effect of PI3K and EGFR inhibition, mutually affecting β-CATENIN and AKT signaling in colorectal cancer**. *Mol Cancer Res* (2018) **16** 543-53. PMID: 29222171
41. Norum JH, Skarpen E, Brech A, Kuiper R, Waaler J, Krauss S. **The tankyrase inhibitor G007-LK inhibits small intestine LGR5**. *Biol Res* (2018) **51** 3. PMID: 29316982
42. Donato E, Biagioni F, Bisso A, Caganova M, Amati B, Campaner S. **YAP and TAZ are dispensable for physiological and malignant haematopoiesis**. *Leukemia* (2018) **32** 2037-40. PMID: 29654273
43. Richter J, Traver D, Willert K. **The role of Wnt signaling in hematopoietic stem cell development**. *Crit Rev Biochem Mol Biol* (2017) **52** 414-24. PMID: 28508727
44. Steinhart Z, Angers S. **Wnt signaling in development and tissue homeostasis**. *Development* (2018) **145** dev146589. PMID: 29884654
45. Muller PY, Milton MN.. **The determination and interpretation of the therapeutic index in drug development**. *Nat Rev Drug Discov* (2012) **11** 751-61. PMID: 22935759
46. Tian H, Biehs B, Warming S, Leong KG, Rangell L, Klein OD. **A reserve stem cell population in small intestine renders Lgr5-positive cells dispensable**. *Nature* (2011) **478** 255-59. PMID: 21927002
47. Sangiorgi E, Capecchi MR. **Bmi1 is expressed in vivo in intestinal stem cells**. *Nat Genet* (2008) **40** 915-20. PMID: 18536716
48. Yan KS, Chia LA, Li X, Ootani A, Su J, Lee JY. **The intestinal stem cell markers Bmi1 and Lgr5 identify two functionally distinct populations**. *Proc Natl Acad Sci U S A* (2012) **109** 466-71. PMID: 22190486
|
---
title: Re-Tear Rates Following Rotator Cuff Repair Surgery
journal: Cureus
year: 2023
pmcid: PMC9981227
doi: 10.7759/cureus.34426
license: CC BY 3.0
---
# Re-Tear Rates Following Rotator Cuff Repair Surgery
## Abstract
Aim Re-tears following rotator cuff repair surgery are a common occurrence. Previous studies have identified several factors that have been shown to increase the risk of re-tears. The purpose of this study was to evaluate the rate of re-tear following primary rotator cuff repair and to identify the factors that may contribute to this rate.
Method The authors performed a retrospective review, looking at rotator cuff repair surgeries performed between May 2017 and July 2019 performed in a hospital by three specialist surgeons. All methods of repair were included. All patients' medical data, including imaging and operation records, were reviewed.
Results A total of 148 patients were identified. Ninety-three males and 55 females with a mean age of 58 years (range 33-79 years). Thirty-four patients ($23\%$) had post-operative imaging with either magnetic resonance imaging or ultrasound, where it was found that 20 ($14\%$) had a confirmed re-tear. Of these patients, nine went on to have further repair surgery. The average age of the re-tear patients was 59 (age range 39-73) and $55\%$ were female. The majority of the re-tears were from chronic rotator cuff injuries. This paper did not identify any correlation between smoking status or diabetes mellitus and re-tear rates.
Conclusions This study indicates that re-tear after rotator cuff repair surgery is a common complication. The majority of studies find increasing age to be the biggest risk factor; however, this was not the case in our study which found females in their 50s to have the highest rate of re-tear. Additional research is required to understand what factors can contribute towards rotator cuff re-rupture rates.
## Introduction
Rotator cuff tears are one of the most common injuries seen by orthopedic surgeons with many suggesting the incidence to be as high as one in five people [1]. Rotator cuff tears in elderly patients are typically the result of age-related deterioration as opposed to younger patients where tears are more likely to be caused by trauma [2].
Despite rotator cuff tears being able to be defined by certain characteristics, there is no clear consensus on the classification used for this injury [3]. The main characteristics that are often used to classify the tear include size, depth, and location. Due to the high prevalence of rotator cuff tears, repair surgery is one of the most widely performed orthopedic surgeries, with arthroscopic repair being the preferred method [4]. Rotator cuff repair can decrease pain and increase function, thus improving a patient’s quality of life. Studies have shown positive post-operative satisfaction from patients [5].
Tears in the rotator cuff can be treated in one of two ways: conservatively or surgically [6,7]. The surgical procedures that can be performed include subacromial decompression, arthroscopic rotator cuff repair, and rotator cuff debridement (arthroscopic or mini-open). It has been demonstrated that there is no significant difference in the clinical benefits achieved by either of the arthroscopic approaches to rotator cuff repair [8,9].
Despite the advancements in procedures for repair, re-tear of the repaired tendon is one of the most common complications encountered post-operatively [10]. Re-tear rates range between $13\%$ and $94\%$ of cases [11]. Some studies have suggested that patients who have a re-tear following repair surgery frequently still have a significant recovery in comparison with their preoperative state. Furthermore, the re-rupture is typically smaller than the original tear [12]. Studies have identified a number of risk variables that have been demonstrated to increase the likelihood of a rotator cuff re-tear occurring. These variables include increasing age, significant tear size, female gender, poor muscle quality, and thicker tears [12-16]. The purpose of this retrospective study was to evaluate the rate of re-rupture following primary rotator cuff repair and to identify the factors that may contribute to this.
## Materials and methods
For this review, authors identified 148 patients who had been admitted to either Huddersfield Royal Infirmary Hospital or Calderdale Royal Hospital from May 2017 to June 2019. Patients who had undergone surgical fixation for a rotator cuff repair were eligible for inclusion in this study. All arthroscopic repairs performed by our orthopedic surgeons specializing in the treatment of upper limbs were included. Patients who were managed with a conservative approach were excluded from the study.
This paper accounted for acute and chronic rotator cuff tears of any size. We classified any injury over three months as chronic. We did not enquire as to the mechanism of injury. Patients’ smoking status and whether they had diabetes mellitus were considered during the data collection. For the re-tear patients, where able, we classified the grade of their original rotator cuff tear using the Goutallier classification which is used for the assessment of muscle degeneration.
## Results
Of the 148 patients identified, our cohort consisted of 93 males and 55 females with a mean age of 58 years (range 33-79 years). The injuries were right-sided in 97 patients and left-sided in the remaining 51 patients.
Surgery to repair the rotator cuff was performed on 148 individuals. Arthroscopic rotator cuff repair was performed on 121 of these patients. Subacromial decompression was performed on 25 patients, and three individuals underwent both procedures in the operation. A total of 145 patients had a chronic tear in their rotator cuff, whereas three patients had an acute tear. Only 34 ($23\%$) of the 148 patients who had arthroscopic surgery of their rotator cuff tear had a post-operative magnetic resonance image (MRI) performed. Following surgery, the average period of time prior to imaging being performed was 217 days.
The data showed that MRIs and ultrasounds were performed on 34 patients with 27 undergoing MRI, six undergoing an ultrasound, and one patient receiving imaging from both modalities. The imaging showed that 14 ($9\%$) patients did not have a re-tear, while 20 ($14\%$) did (Figure 1). Of the 20 patients that had a confirmed re-tear, 18 ($90\%$) had undergone arthroscopic repair whilst two ($10\%$) had a subacromial decompression. Following the results of the imaging, nine out of the 20 patients who had a verified re-tear went on to have revision surgery. One patient did not wish to proceed with further surgery and the patients remaining who had a confirmed re-tear were conservatively managed.
**Figure 1:** *Graph to show our percentage of patients that had post-operative imaging*
Overall, of the 148 patients that were operated on, nine went on to receive a second repair surgery. The age range of these 20 patients who had a confirmed re-tear ranged from 39 to 73 years old, and $55\%$ of them were female. Table 1 illustrates the re-tear rates in relation to the age groups of the patients. The median age of these patients was $59.65\%$ of the re-tears on the right side, whilst the remaining $45\%$ were on the left side.
**Table 1**
| Patients' age | Re-tear rates |
| --- | --- |
| Under 50 | 5% |
| 50-59 | 40% |
| 60-69 | 35% |
| 79-79 | 20% |
| Over 80 | 0% |
For the nine patients that went on to receive revision surgery, alternate operative methods were used. Four patients had a rotator cuff re-repair, two patients underwent a reverse shoulder arthroplasty (RSA) and two underwent a superior capsule reconstruction (SCR). Percentages are illustrated in Figure 2. One patient was found not to have a re-tear, therefore, no repair was needed. Out of the 20 patients who had a confirmed re-tear two were current smokers, two were previous smokers, and 16 were non-smokers. Two of the patients had a diagnosis of diabetes mellitus. Eighteen patients had chronic rotator cuff tears and only two had acute ones. Of the 20 patients who had a confirmed re-tear, 11 of these patients received pre-operative MRIs. We were then able to apply the Goutallier classification to grade their initial rotator cuff tear. Six patients had a grade zero, two had a grade one, one had a grade two, and two had a grade three.
**Figure 2:** *Repair methods used during further surgeryRSA: reverse shoulder arthroplasty, SCR: superior capsule reconstruction*
A total of 115 patients had no confirmed or suspected rotator cuff re-tear. The majority of these patients were seen at least twice in the clinic post-operatively to monitor their progression and post-operative symptoms. During these consultations that were led by either a consultant or registrar, there were no concerns or suspicions of any complications including a re-tear. It was not uncommon for these patients to still have some degree of discomfort. However, it was felt by the clinicians that for these patients the degree of pain they were experiencing was to be expected for the stage that they were postoperative. Patients were discharged from the clinic up to one year following their rotator cuff repair if they had no post-operative concerns.
Despite a confirmed re-tear, 11 patients did not receive an operative repair and were instead managed conservatively. There were several factors influencing this decision. Firstly, each consultant’s individual view regarding operative vs. conservative management of re-tears played a role in the decision. Often the decision to manage conservatively was made in conjunction with offering alternatives for pain management such as shoulder injections or longer rehabilitation programs. Some patients simply did not want to have surgery and therefore declined an operation. For others, despite an MRI-proven re-tear, their range of movement had improved to a satisfactory level by the time the patients were seen in the clinic and therefore operative management was deemed not necessary. It has been shown in previous studies that conservative management should be trialed in low-demand patients [17].
## Discussion
Re-tear after a rotator cuff repair is a commonly encountered complication and has been shown to be caused by a combination of various factors. Previous studies have identified that the size of the rotator cuff tear and age are the greatest predictors of outcomes [18]. In addition, the quality of the tissue, limb dominance, and smoking status are all factors that can influence the likelihood of a re-tear occurring after the initial repair [19]. Although it has been shown that increasing age can increase the risk of re-tear, the number of re-tears that occurred in patients in their 50s was shown to be the highest of any age group [19,13]. Table 2 illustrates findings from other studies comparing re-tear rates and mean age.
**Table 2**
| Cohort | Re-tear rate % | Mean age, year |
| --- | --- | --- |
| Our study | 13.5 | 59.0 |
| Klepps et al. | 31.3 | 64.0 |
| Lapner et al. | 27.6 | 56.8 |
| Gallagher et al. | 17.4 | 65.7 |
| Sheean et al. | 13.3 | 65.0 |
| Ma et al. | 30.2 | 61.2 |
In our research, we found that the female gender was associated with a higher prevalence of rotator cuff re-tears when compared to males. We found that $55\%$ of those who had a re-tear were female, despite only $37\%$ of the overall cohort of patients undergoing rotator cuff repairs being female. This varies from the findings of other studies that have shown that gender does not play a significant role in the development of rotator cuff injuries [24].
The management of re-tears in patients might vary based on a variety of criteria, such as the age of the patient, the patient's functional state, the magnitude of the re-tear, and the length of the tendon that is still intact. There are several different approaches that can be taken to address tears in the rotator cuff. These include conservative therapy, revision rotator cuff repairs, superior capsular reconstruction, tendon transfers, and reverse shoulder arthroplasty. Only $45\%$ of patients in this study who had a rotator cuff re-tear confirmed on imaging proceeded to have surgical repair, while the remaining patients had conservative therapy instead.
The majority of patients are seen in the clinic post-operatively after the original rotator cuff repair surgery. During this consultation, following history and examination, the clinician decides whether to suspect a re-tear. There is often concern that a re-tear has developed if the patient has post-operative pain or decreased function following the repair [25]. In the post-operative reviews, both pain and reduced range of movement in the shoulder were all indications for the patient to have post-operative imaging. However, not all tears generate symptoms. Therefore, some patients may have tears that are not identified due to not receiving any form of imaging. MRI is the favored modality of imaging as although ultrasound is useful, it gives a limited comprehensive view of the shoulder [25]. The presence of atrophy and fatty infiltration on an MRI scan has been shown to correlate with failed rotator cuff repair. Failed repairs often show a progression of fatty infiltration and muscle atrophy [25].
In this study, postoperative imaging was found to be rather uncommon; with just $22\%$ of patients receiving any form of imaging after their procedure. As a result, it is challenging to make an accurate assessment of the number of patients who sustained rotator cuff re-tears, as well as the potential causes of these tears and the mechanism through which they happened. The most common imaging modality chosen by clinicians to identify re-rupture rates was found to be MRI, with only a few clinicians choosing ultrasound to identify re-rupture rates [25]. When it comes to identifying partial thickness tears, several studies have demonstrated that ultrasound is highly reliable [26]. Studies have even shown it to be more specific than MRI. In these studies, it was found that ultrasound had a specificity of $66.7\%$, whilst MRI had a specificity of just $63.6\%$ [27]. In light of the significant financial disparity between the two imaging modalities, it is essential to investigate the feasibility of performing ultrasounds for the diagnosis of re-rupture before resorting to MRI.
A limitation of our study is the lack of post-operative imaging that our patients received. As only $23\%$ of the patients had post-operative imaging, we are unable to determine the true incidence of re-tear that may have occurred following the surgery.
## Conclusions
Our results suggest that arthroscopic rotator cuff repair is an effective surgery resulting in a positive outcome for the majority of patients. From our data, we can conclude the risk of a post-operative re-tear was $14\%$. However, we acknowledge not all patients had post-operative imaging. Our outcomes of rotator cuff repair appear consistent with the outcomes in the literature. Additional research is required to understand what factors can contribute towards rotator cuff re-rupture rates.
## References
1. Kucirek NK, Hung NJ, Wong SE. **Treatment options for massive irreparable rotator cuff tears**. *Curr Rev Musculoskelet Med* (2021) **14** 304-315. PMID: 34581991
2. Keener JD, Patterson BM, Orvets N, Chamberlain AM. **Degenerative rotator cuff tears: refining surgical indications based on natural history data**. *J Am Acad Orthop Surg* (2019) **27** 156-165. PMID: 30335631
3. Lädermann A, Burkhart SS, Hoffmeyer P, Neyton L, Collin P, Yates E, Denard PJ. **Classification of full-thickness rotator cuff lesions: a review**. *EFORT Open Rev* (2016) **1** 420-430. PMID: 28461921
4. Gallagher BP, Bishop ME, Tjoumakaris FP, Freedman KB. **Early versus delayed rehabilitation following arthroscopic rotator cuff repair: a systematic review**. *Phys Sportsmed* (2015) **43** 178-187. PMID: 25797067
5. Novoa-Boldo A, Gulotta LV. **Expectations following rotator cuff surgery**. *Curr Rev Musculoskelet Med* (2018) **11** 162-166. PMID: 29435813
6. De Carli A, Fabbri M, Lanzetti RM. **Functional treatment in rotator cuff tears: is it safe and effective? A retrospective comparison with surgical treatment**. *Muscles Ligaments Tendons J* (2017) **7** 40-45. PMID: 28717610
7. Chalmers PN, Ross H, Granger E, Presson AP, Zhang C, Tashjian RZ. **The effect of rotator cuff repair on natural history: a systematic review of intermediate to long-term outcomes**. *JB JS Open Access* (2018) **3** 0
8. MacDermid JC, Bryant D, Holtby R. **Arthroscopic versus mini-open rotator cuff repair: a randomized trial and meta-analysis**. *Am J Sports Med* (2021) **49** 3184-3195. PMID: 34524031
9. Oliva F, Piccirilli E, Bossa M. **I.S.Mu.L.T - Rotator cuff tears guidelines**. *Muscles Ligaments Tendons J* (2015) **5** 227-263. PMID: 26958532
10. Jeong HY, Kim HJ, Jeon YS, Rhee YG. **Factors predictive of healing in large rotator cuff tears: is it possible to predict retear preoperatively?**. *Am J Sports Med* (2018) **46** 1693-1700. PMID: 29595993
11. Mandaleson A. **Re-tears after rotator cuff repair: current concepts review**. *J Clin Orthop Trauma* (2021) **19** 168-174. PMID: 34123722
12. Galanopoulos I, Ilias A, Karliaftis K, Papadopoulos D, Ashwood N. **The impact of re-tear on the clinical outcome after rotator cuff repair using open or arthroscopic techniques - a systematic review**. *Open Orthop J* (2017) **11** 95-107. PMID: 28400878
13. Bedeir YH, Jimenez AE, Grawe BM. **Recurrent tears of the rotator cuff: effect of repair technique and management options**. *Orthop Rev (Pavia)* (2018) **10** 7593. PMID: 30057724
14. Lädermann A, Denard PJ, Burkhart SS. **Management of failed rotator cuff repair: a systematic review**. *J ISAKOS* (2016) **1** 32-37. PMID: 27134759
15. Diebold G, Lam P, Walton J, Murrell GA. **Relationship between age and rotator cuff retear: a study of 1,600 consecutive rotator cuff repairs**. *J Bone Joint Surg Am* (2017) **99** 1198-1205. PMID: 28719559
16. Longo UG, Carnevale A, Piergentili I, Berton A, Candela V, Schena E, Denaro V. **Retear rates after rotator cuff surgery: a systematic review and meta-analysis**. *BMC Musculoskelet Disord* (2021) **22** 749. PMID: 34465332
17. Geary MB, Elfar JC. **Rotator cuff tears in the elderly patients**. *Geriatr Orthop Surg Rehabil* (2015) **6** 220-224. PMID: 26328240
18. Le BT, Wu XL, Lam PH, Murrell GA. **Factors predicting rotator cuff retears: an analysis of 1000 consecutive rotator cuff repairs**. *Am J Sports Med* (2014) **42** 1134-1142. PMID: 24748610
19. Sambandam SN, Khanna V, Gul A, Mounasamy V. **Rotator cuff tears: an evidence based approach**. *World J Orthop* (2015) **6** 902-918. PMID: 26716086
20. Klepps S, Bishop J, Lin J, Cahlon O, Strauss A, Hayes P, Flatow EL. **Prospective evaluation of the effect of rotator cuff integrity on the outcome of open rotator cuff repairs**. *Am J Sports Med* (2004) **32** 1716-1722. PMID: 15494338
21. Lapner PL, Sabri E, Rakhra K, McRae S, Leiter J, Bell K, Macdonald P. **A multicenter randomized controlled trial comparing single-row with double-row fixation in arthroscopic rotator cuff repair**. *J Bone Joint Surg Am* (2012) **94** 1249-1257. PMID: 22810395
22. Sheean AJ, de Sa D, Woolnough T, Cognetti DJ, Kay J, Burkhart SS. **Does an increased critical shoulder angle affect re-tear rates and clinical outcomes following primary rotator cuff repair? A systematic review**. *Arthroscopy* (2019) **35** 2938-2947. PMID: 31515108
23. Ma HL, Chiang ER, Wu HT, Hung SC, Wang ST, Liu CL, Chen TH. **Clinical outcome and imaging of arthroscopic single-row and double-row rotator cuff repair: a prospective randomized trial**. *Arthroscopy* (2012) **28** 16-24. PMID: 21982391
24. Sabo MT, LeBlanc J, Hildebrand KA. **Patient gender and rotator cuff surgery: are there differences in outcome?**. *BMC Musculoskelet Disord* (2021) **22** 838. PMID: 34592991
25. Thakkar RS, Thakkar SC, Srikumaran U, McFarland EG, Fayad LM. **Complications of rotator cuff surgery-the role of post-operative imaging in patient care**. *Br J Radiol* (2014) **87** 20130630. PMID: 24734935
26. Gilat R, Atoun E, Cohen O, Tsvieli O, Rath E, Lakstein D, Levy O. **Recurrent rotator cuff tear: is ultrasound imaging reliable?**. *J Shoulder Elbow Surg* (2018) **27** 1263-1267. PMID: 29398398
27. Nazarian LN, Jacobson JA, Benson CB. **Imaging algorithms for evaluating suspected rotator cuff disease: Society of Radiologists in Ultrasound consensus conference statement**. *Radiology* (2013) **267** 589-595. PMID: 23401583
|
---
title: 'Protective Effects of Crocin Against Methotrexate-Induced Hepatotoxicity in
Adult Male Albino Rats: Histological, Immunohistochemical, and Biochemical Study'
journal: Cureus
year: 2023
pmcid: PMC9981239
doi: 10.7759/cureus.34468
license: CC BY 3.0
---
# Protective Effects of Crocin Against Methotrexate-Induced Hepatotoxicity in Adult Male Albino Rats: Histological, Immunohistochemical, and Biochemical Study
## Abstract
Background: Among the many known adverse effects of methotrexate (MTX), hepatotoxicity stands out as a major drawback that limits its therapeutic applicability. There is growing evidence that crocin has antioxidant, anti-hyperglycemic, cardioprotective, and anti-inflammatory effects. This study's aim is to evaluate the potential protective effect of crocin against MTX-induced liver damage in rats using biochemical, histological, and immunohistochemical analyses.
Methods: Twenty-four adult male albino rats were split into four groups at random (six rats/group) as follows: normal control (saline, intraperitoneal (i.p.) injections), crocin-treated (100 mg/kg daily for 14 days, i.p.), MTX-treated (20 mg/kg single i.p. injection on day 15), and crocin/MTX-treated groups (crocin 100 mg/kg/day for 14 days, i.p. + MTX 20 mg/kg single i.p. injection on day 15). On day 16 of the experiment, blood and tissue specimens were used to assess the liver functions, oxidative stress markers, transforming growth factor beta 1 (TGF-β1), caspase-3, BCL-2-associated X protein (BAX), and B-cell lymphoma 2 (BCL-2) expression.
Results: The results of the current research revealed the protective actions of crocin against MTX-induced hepatotoxicity. Our results showed that crocin possesses antioxidants (decrease malondialdehyde (MDA), increase glutathione (GSH) levels, and enhance catalase (CAT) and superoxide dismutase (SOD) enzymatic activity), anti-fibrotic (decrease TGF-β1), and anti-apoptotic (decrease BAX and caspase-3 expression while increase BCL-2) actions in liver. Moreover, crocin administration along with MTX restores the normal histological structure of hepatic tissues.
Conclusion: *The data* presented in the current study using an in vivo animal model support the notion that crocin should be further studied in humans to assess its potential hepatoprotective effects against MTX-induced liver damage.
## Introduction
Methotrexate (MTX) is an anti-folic acid medication and an aminopterin stable derivative that inhibits deoxyribonucleic acid (DNA) synthesis and repair. MTX is a very potent cytotoxic drug that alters cellular metabolism and so suppresses cell growth [1]. It was initially prescribed for children with acute leukemia, and it was later used in the treatment of psoriasis and rheumatoid arthritis. The MTX-cytotoxic effect, however, is not limited to cancer cells and affects many other normal tissues, including the stomach mucosa, gall bladder, hematopoietic cells of the bone marrow, and liver [2].
Hepatotoxicity is one of the key recognized toxicity profiles of MTX, which limits its therapeutic applicability [3]. The hepatotoxic mechanism of MTX is yet unclear, nevertheless, conceivable explanations might be proposed [1,4]. One possible mechanism of MTX-induced hepatotoxicity is its action on the intestinal mucosa, which disrupts intestinal barrier functions, allowing bacteria to translocate to the liver and cause hepatotoxicity [5]. Furthermore, MTX-enhanced intestinal permeability has been associated with hepatic inflammation, increased hepatic transaminases, and abnormal liver histological architecture such as increased liver fibrosis, cirrhosis, and hepatocyte apoptosis [5]. It has previously been reported that MTX's cytotoxic action on the liver is mediated in part by its ability to inhibit the conversion of homocysteine to methionine, resulting in an increase in homocysteine, which damages the endoplasmic reticulum and stimulates fat accumulation in hepatocytes, the proinflammatory cytokines, and hepatic stellate cells, all of which lead to liver fibrosis [4].
With adequate monitoring and medication regimens of MTX, the frequency of high transaminases has decreased to $22\%$ in the past few years [6]. However, around $5\%$ of patients receiving continuous low-dose MTX might develop advanced hepatic fibrosis or cirrhosis. MTX prescription, even at modest doses, is contraindicated in individuals with risk factors such as pre-existing liver disease, alcohol usage, obesity, and diabetes mellitus [3]. Many adjuvants with cytoprotective properties, such as ursodeoxycholic acid and carotene, have been recommended for users to reduce the occurrence of MTX-induced adverse effects [4,6]. Likewise, saffron, a perennial stemless plant used as a food additive, has recently been the focus of many research studies because of its recognized pharmacological characteristics like anti-inflammatory, anti-cancer, anti-hyperlipidemic, and cardioprotective effects [7-12]. The primary active ingredient in saffron is picrocrocin, along with its derivatives such as safranal, flavonoid compounds, and crocin.
Crocin is an active ingredient in saffron that is responsible for its red color [12,13]. Crocin is water soluble and heat stable, and is composed of gentiobiose and crocetin, which are disaccharides and carboxylic acids, respectively. A growing body of evidence showed that crocin has antitumor, antioxidant, anti-hyperglycemic, cardioprotective, anti-inflammatory, and DNA-protective effects [14,15]. Crocin pretreatment protects the stomach mucosa from ischemia-reperfusion damage by increasing messenger ribonucleic acid (mRNA) expression and the activity of certain antioxidant enzymes [16]. Although crocin has been demonstrated to have hepatoprotective properties against oxidative stress, other putative mechanisms involved in safeguarding the liver from MTX-induced damage have been less investigated. Hence, this research was carried out to examine the potential protective effect of crocin against MTX-induced liver damage in rats, putting a focus on the antioxidant and antiapoptotic effects. To accomplish this aim, liver functions, oxidative stress markers, transforming growth factor beta 1 (TGF-β1), caspase-3, BCL-2-associated X protein (BAX), and B-cell lymphoma 2 (BCL-2) expression were all assessed in a rat model.
## Materials and methods
Animals Twenty-four male albino rats (120-150 g) were employed in the current experiment. The rats were acquired from the Ophthalmology Research Institute in Giza, Egypt. They were housed in standard rat cages and allowed a week to acclimate before the experiment began. They were maintained on a regular chow diet, water, and reversed cycles of darkness and light. All the guidelines of the Ethics Committee, Faculty of Medicine, Suez Canal University were followed during the experiment (Research# 4331). Every attempt was made to minimize the total number of animals utilized and their anguish.
Experimental procedure Rats were randomly designated into four groups (six rats/group) as follows: group I (normal control group): rats were given intraperitoneal (i.p.) injections of an equal volume of normal physiological saline (vehicle) for two weeks; group II (crocin-treated group): crocin (Sigma-Aldrich, London, UK) dissolved in saline was administered to rats (100 mg/kg/day, i.p.) once daily started on the first day of the experiment and continued for 14 days [17]; group III (MTX-treated group): on day 15 of the experiment, a single dose of MTX (20 mg/kg, i.p., with normal saline as solvent) was administered to the rats [17]; group IV (crocin/MTX-treated group): starting from the first day of the study, the rats were given crocin (100 mg/kg/day, i.p.) for 14 days followed by a single MTX injection (20 mg/kg, i.p.) on day 15.
On day 16, a day following MTX injection, by i.p. injection of 50 mg/kg of ketamine and 5 mg/kg of xylazine, rats were anesthetized [18]. Blood samples were drawn from the abdominal aorta and left to stand for four hours at ambient temperature before being spun in a centrifuge to separate the serum from the blood cells, which were then used for various biochemical assays. For the tissue-based biochemical and histopathological analyses, samples of the liver from the right lobe were taken.
Serum biochemical analyses Liver Function Tests To assess liver function, the serum activities of alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), and lactate dehydrogenase (LDH) were measured in all blood samples. These enzymes were tested by the enzymatic assay kits for colorimetry (Bio-diagnostics, Giza, Egypt) utilizing a spectrophotometer for ultraviolet (UV) visibility (UV-1601-PC; Shimadzu, Kyoto, Japan) [19]. Total protein and serum albumin were also measured.
C-Reactive Protein (CRP) *The serum* level of CRP was quantified using automated COBAS 6000 (module 501, Roche Diagnostics, Basel, Switzerland) according to the manufacturer's guidelines. Every sample was carried out in duplicate.
Assessment of oxidative stress markers In all experimental groups, liver tissues were minced and homogenized. In ice-cold saline, a homogenate of $10\%$ w/v was prepared. The prepared homogenates were then centrifuged for 15 minutes at 18,000g (14°C). The supernatants of liver homogenate were collected to assess hepatic malondialdehyde (MDA) level as a lipid peroxidation indicator using the colorimetric assay kits (catalog # MD 2529, Bio-diagnostics) according to the manufacturer's recommendations. Levels of glutathione (GSH), as well as activity of catalase (CAT) and superoxide dismutase (SOD), were all assessed as indicators of liver antioxidant status using the colorimetric assay kits (Bio-diagnostics, catalog # GR 2511, CA 2517, and SD 2521, respectively).
Quantitative real-time polymerase chain reaction (qRT-PCR) of TGF-β1 TGF-β1 expression levels were determined using qRT-PCR. Frozen liver tissues from each rat were used to extract the total ribonucleic acid (RNA) of different study groups following the manufacturer's directions with the Qiagen tissue extraction kit (Qiagen, Germantown, Maryland). Spectrophotometry was used to assess the retrieved RNA's purity and concentration (dual-wavelength spectrophotometer, Beckman, Irvine, California). The isolated RNA was utilized in the preparation of complementary DNA (cDNA) considering the manufacturer's instructions with a high-fidelity reverse transcription kit (Fermentas, Waltham, Massachusetts). Using the Applied *Biosystems apparatus* and StepOneTM software version 3.1 (Thermo Fisher Scientific, Waltham, Massachusetts), the reaction was incubated at 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for one minute. The TGF-β1 levels were standardized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA levels in each sample [19]. The relative gene expression was calculated using the 2−ΔΔCT method in comparison to the controls [20]. The TGF-β1 primer sequences used were as follows: sense: 5'-TTGCCCTCTACAA CCAACACAA - 3'; and antisense: 5'-GCTTGCGACCCACGTAGT A-3'. All experiments were performed in triplicate.
Histopathological study The liver samples were taken and preserved for 24 hours in $10\%$ neutral buffered formalin. The liver samples were then washed, dehydrated in ascending alcoholic gradients, and paraffin-embedded for preparation of paraffin wax blocks. Sections of 4 um thickness were prepared and processed for hematoxylin and eosin (H&E) staining, as detailed by Bancroft et al. [ 20]. To overall assess the severity of hepatic tissue lesions in different study groups, a semiquantitative scoring method was employed as described by Ragab et al. [ 21]. Five high-power fields from each animal were examined by two independent histopathologists. A score ranging from 0 to 3 was used, with 0 indicating no histopathological alterations, 1+ denoting pathological changes in less than $20\%$ of the studied fields, 2+ indicating histopathologic changes in between $20\%$ and $60\%$ of the investigated fields, and 3+ indicating histopathologic changes in more than $60\%$ of the fields examined.
Immunohistochemical (IHC) study and morphometric analysis To determine the transcription of BAX and caspase-3, and BCL-2 proteins in the liver sections, IHC staining was carried out in accordance with Torlakovic et al. 's protocol [22]. The liver sections from various experimental groups were incubated with monoclonal antibodies of BAX, BCL-2, and caspase-3 (Dako, Carpinteria, California) at a dilution of 1:200. BAX, BCL-2, and caspase-3 were considered positive in cells that exhibited brown precipitate. Negative control was prepared by incubating the tissue specimens with the antibody diluent while omitting the primary antibody. The IHC profiler plugin in the ImageJ software (National Institutes of Health, Bethesda, Maryland) has been used to assess the expression of three markers [23]. Using the 3,30-diaminobenzidine (DAB)-stained cytoplasmic option of the IHC plugin, 10 non-overlapping fields were assessed from each experimental group using immune-stained sections (200x).
Statistical analysis The collected data were analyzed using the Statistical Package for Social Sciences (SPSS, version 22, IBM Corp., Armonk, NY), which were given as mean ± standard deviation (SD). The difference in mean values among groups was calculated using the one-way analysis of variance (ANOVA), which was then confirmed by the Bonferroni multiple comparison test. Two-tailed p-values were given, and a $p \leq 0.05$ was considered to be significant. Every possible comparison was conducted between the study groups.
## Results
Biochemical results The present data showed that liver functions were significantly deteriorated with MTX administration. As depicted in Table 1, rats treated with a single injection of MTX demonstrated a highly significant increase ($p \leq 0.001$) in the serum levels of AST, ALT, and ALP enzymes in comparison to control and crocin-treated groups. On the other hand, administration of crocin before MTX injection significantly reduced the AST level when compared with the MTX-treated group. Although, both ALT and ALP levels were reduced in rats treated with crocin and MTX combined, no significant difference was detected when compared to the MTX-treated group. It is worth noting that rats treated with MTX had non-significant differences in LDH, serum albumin, and total protein levels when compared to the normal control group (Table 1).
**Table 1**
| Liver function | Normal control | Crocin | MTX | Crocin/MTX | P-value |
| --- | --- | --- | --- | --- | --- |
| AST | 156.8 ± 38.2 | 173.7 ± 69.4 | 322.7 ± 15#@ | 237 ± 25.3#$ | <0.0011 |
| ALT | 24.3 ± 0.8 | 29 ± 10.6 | 88.3 ± 2.6#@ | 77.5 ± 8.2#@ | <0.0012 |
| ALP | 77.7 ± 12.9 | 93.8 ± 41.5 | 500.8 ± 8.1#@ | 326.33 ± 7.9#@ | <0.0012 |
| LDH | 1209.7 ± 45.7 | 1716.7 ± 593.8 | 1921 ± 83.9 | 1795.5 ± 1005.4 | 0.0792 |
| Serum albumin | 3.4 ± 0.37 | 3.1 ± 0.19 | 3.1 ± 0.25 | 3.1 ± 0.33 | 0.1621 |
| Total protein | 7.6 ± 0.5 | 7.4 ± 0.9 | 6.8 ± 0.4 | 7 ± 0.3 | 0.1451 |
Next, the liver inflammation status in all experimental groups was evaluated using the CRP, a marker of inflammation. Nonetheless, as seen in Figure 1, no statistically significant difference could be detected between the research groups.
**Figure 1:** *CRP level among studied groupsData are expressed as mean ± SD and were analyzed using 1one-way ANOVA and Bonferroni post-hoc test (n = 6 for each group).MTX: methotrexate: CRP: C-reactive protein.*
Markers of lipid peroxidation and oxidative stress *In this* study, our results revealed that a single injection of MTX significantly raised ($p \leq 0.001$) the lipid peroxidation marker (MDA) level in liver homogenates when compared to the normal control and the crocin-treated groups. Furthermore, when comparing the group treated with MTX to both the control and the crocin-treated groups, our results revealed a significant reduction ($p \leq 0.001$) of GSH level and decreased SOD and CAT enzymatic activity in liver tissue homogenates. Pretreatment with crocin in the crocin/MTX-treated group mitigated the impact of MTX on lipid peroxidation and oxidative stress markers. Crocin administration in rats injected with MTX reduced MDA while increasing GSH levels and enhancing the activity of CAT and SOD in liver tissues, as demonstrated in Table 2.
**Table 2**
| Oxidative stress markers | Normal control | Crocin | MTX | Crocin/MTX | P-value |
| --- | --- | --- | --- | --- | --- |
| MDA (nmol/g tissue) | 0.26 ± 0.04 | 0.36 ± 0.18 | 0.90 ± 0.33#@ | 0.51 ± 0.02$ | <0.0011 |
| GSH (μmol/g tissue) | 0.239 ± 0.026 | 0.238 ± 0.023 | 0.078 ± 0.003#@ | 0.138 ± 0.023#@$ | <0.0011 |
| CAT (μ/g tissue) | 1.08 ± 0.051 | 1.07 ± 0.009 | 0.5 ± 0.011#@ | 1.03 ± 0.005#@$ | <0.0021 |
| SOD (μ/g tissue) | 0.037 ± 0.051 | 0.034 ± 0.003 | 0.014 ± 0.003#@ | 0.027 ± 0.006$ | <0.0012 |
TGF-β1 mRNA expression in liver tissue using qRT-PCR Administration of MTX induced a marked increase in the mRNA expression level of TGF-β1, a profibrogenic marker, when compared with both control and crocin-treated groups ($p \leq 0.05$). However, the administration of crocin together with MTX induced a significant downregulation of TGF-β1 when compared with the MTX-treated group. However, TGF-β1 expression remained significantly higher among rats treated with crocin/MTX in comparison to that of the normal control group (Figure 2).
**Figure 2:** *TGF-β1 mRNA relative expression level among studied groups using qRT-PCRData are expressed as mean ± SD and were analyzed using one-way ANOVA and Bonferroni post-hoc test. # Compared to the normal group at p < 0.05. @ Compared to the crocin group at p < 0.05. $ Compared to the MTX-treated group at p < 0.05 (n = 6 for each group).TGF-β1: transforming growth factor beta 1; MTX: methotrexate; qRT-PCR: quantitative real-time polymerase chain reaction.*
Histopathological results The histological framework of liver tissue in both the normal control and crocin-treated groups was comparable (Figures 3A, 3B). H&E-stained sections of the liver from both groups demonstrated the classical view of the liver lobules, with the central vein in the core and the portal triad branches at the lobules' periphery. The central vein and peripherally located portal triad (hepatic artery, portal vein, and bile duct) were connected with columns of radially distributed hepatocytes with prominent, centrally placed, and spherical nuclei. The single-cell width hepatocytic columns were separated by blood sinusoids with scattered Kupffer cells.
Sections from the MTX-treated group stained with H&E disclosed congestion of blood vessels with dilated and congested sinusoids, which indicates hepatoportal vascular congestion. Hepatocytes demonstrated cytoplasmic vacuolation, hydropic degeneration, and variable stages of necrosis, including nuclear fading, shrinkage, and fragmentation. Whole areas of cellular absence were present with variable-sized areas of necrotic foci (Figure 3C). Notably, there was no cholestasis or inflammatory cellular infiltration in the group receiving MTX (Table 3). Crocin administration in combination with MTX, on the other hand, improved the histological architecture of the liver tissue in comparison to the MTX-treated group. Based on the histopathological assessment and no vascular congestion, few hepatocytes (<$20\%$) demonstrated necrotic or hydropic degenerative changes (Figure 3D and Table 3). These findings underline the crocin's hepatoprotective effects against MTX-induced liver damage.
**Figure 3:** *Sections in the rat liver stained with H&E from different study groups(A) Normal control group with normal histological structure of the liver showing a central vein (CV), a branch of the hepatic artery (HA), a branch of the portal vein (PV), a bile duct (BD), sinusoids (S), hepatocytes (arrow), and von Kupffer cells (arrowheads). (B) Crocin-treated group showing a central vein (CV), a branch of the portal vein (PV), a bile duct (BD), sinusoids (S), hepatocytes (arrow), and von Kupffer cells (arrowheads). (C) MTX-treated group showing congested blood vessels (CBV), dilated and congested sinusoids (S), hepatocytes in different stages of necrosis (arrow), hydropic degeneration (arrowheads), areas of cellular necrosis (circles), and necrotic foci (N). (D) Crocin/MTX-treated group with no congested blood vessels (CV), fewer dilated sinusoids (S), fewer necrotic hepatocytes (arrow), and fewer hepatocytes with hydropic degeneration (arrowheads) (H&E 400x).H&E: hematoxylin and eosin; MTX: methotrexate.* TABLE_PLACEHOLDER:Table 3 Immunohistochemical and morphometric results To evaluate the possible protective effects of crocin against MTX-induced hepatotoxicity, IHC labeling was used to examine the expression of pro-apoptotic protein BAX, anti-apoptotic protein BCL-2, and caspase-3 (apoptosis coordinating enzyme) in liver tissues from various research groups. Both the untreated control group and the crocin-treated group showed low positive BAX expression, with only a few hepatocytes demonstrating strong positive cytoplasmic immunostaining (Figures 4A, 4B, 5A, 5B). MTX administration increased the cytoplasmic expression of the pro-apoptotic marker BAX, with almost all hepatocytes showing strong BAX immunostaining (Figures 4C, 5C). BAX immunostained sections from crocin/MTX-treated groups revealed a marked decrease in BAX expression in the majority of hepatocytes (Figures 4D, 5D). Similarly, caspase-3 immunostaining represented low positive and positive expression in normal and crocin-treated groups, respectively (Figures 4E, 4F, 5E, 5F). MTX caused marked upregulation of caspase-3 immunostaining, which was detected in the cytoplasm and nuclei or hepatocytes (Figures 4G, 5G). This upregulated expression of caspase-3 detected in MTX-intoxicated rats was almost normalized after concomitant administration of crocin and MTX (Figures 4H, 5H).
On the other hand, the IHC assessment of anti-apoptotic marker BCL-2 revealed considerable expression in both normal untreated and crocin-treated groups (Figures 4I, 4J, 5I, 5J). The level of BCL-2 protein expression was moderately reduced in liver tissues of the MTX-treated group (Figures 4K, 5K). Concomitant administration of crocin and MTX significantly increased BCL-2 expression to levels equivalent to the normal control group (Figures 4L, 5L). Hence, the current study's findings indicate that one potential mechanism of crocin's hepatoprotective activities against MTX-induced hepatotoxicity is through its action on BAX, caspase-3, and BCL-2.
**Figure 4:** *Liver immunostaining for BAX (A-D), caspase-3 (E-H), and BCL-2 (I-L) in hepatic tissue of all studied groups (immunostaining 200x)BCL-2: B-cell lymphoma 2; BAX: BCL-2-associated X protein; MTX: methotrexate.* **Figure 5:** *Histogram profiles of IHC scoring of BAX (A-D), caspase-3 (E-H), and BCL-2 (I-L) created by the ImageJ program (IHC profiler plugin)Each histogram shows the percentage contribution of high positive, positive, low positive, and negatively stained cells, as well as the final score of each protein expression.IHC: immunohistochemical; BCL-2: B-cell lymphoma 2; BAX: BCL-2-associated X protein; MTX: methotrexate.*
## Discussion
MTX is a primary medication for treating some diseases such as specific subtypes of leukemia and lymphoma; however, its administration is accompanied by several adverse reactions that restrict its clinical application. Hepatotoxicity is one of the most prevalent and harmful adverse reactions of MTX. Numerous studies have demonstrated that MTX and its metabolites induce inflammatory processes, oxidative stress, fibrosis, and apoptosis in hepatocytes [17,21]. The primary aim of the present study was to determine whether crocin might attenuate the cytotoxic effect of MTX on liver cells. To reach our objective, we evaluated the liver function, oxidative stress, inflammatory marker, TGF-β1 (the master regulator of fibrosis), apoptosis, and histological alterations in liver tissue.
In this experiment, a single injection of MTX (20 mg/kg) resulted in a substantial increase in the levels of hepatic enzymes, including ALT, AST, and ALP, when compared to the normal control group ($p \leq 0.001$), which is consistent with the findings of preceding research [6,22,23]. Furthermore, crocin administration in combination with MTX considerably decreased the serum levels of three assessed liver enzymes compared to the MTX-treated group; however, except for the AST level, none of these differences were statistically significant. These findings were partially consistent with those of Akbari et al., who demonstrated that crocin pretreatment of rats with ischemia-reperfusion hepatic injury lowered the level of liver enzymes in blood [24]. The effect of crocin on liver function tests can be justified by its membrane-stabilizing activities that prevent the intracellular enzymes from leakage into the blood [22]. Furthermore, another study reported the effect of carotenoids, which include crocin in enhanced liver regeneration and repair, and consequently the liver aminase enzyme levels return to normal levels [25].
Concerning the oxidative stress markers, our finding revealed a significantly higher level of MDA and a significantly lower level of GSH as well as reduced enzymatic activity of SOD and CAT compared to the normal untreated group and the crocin-treated group ($p \leq 0.001$). Our results confirm the oxidative stress induced by the administration of MTX in hepatic tissues reported in previous studies, which reported an increase in MDA and reactive oxygen metabolites and a decrease in GSH, SOD, and glutathione peroxidase (GP-x) in MTX-treated rats when compared with the control group [26,27]. In agreement with two previous studies [14,28], the results of the current study indicated that pretreatment with crocin combined with MTX significantly increases the GSH level in liver homogenate, as well as enhances the activity of fundamental antioxidant enzymes (SOD and CAT), while substantially lowering MDA level in hepatic tissue when compared with MTX-treated group. In light of this, the observed effectiveness of crocin to counteract the hepatotoxic effects of MTX can be attributed to an increase in GSH synthesis, enhancement of SOD and CAT enzymatic activity, and a decrease in reactive oxygen species production (MDA).
Our results from biochemical analysis demonstrated a non-significant decrease in the level of CRP in both MTX-treated and crocin/MTX-treated groups when compared with normal control. This finding can be explained by the well-known anti-inflammatory and immunosuppressant effects of MTX [29,30]. In addition, these data are supported by the histological analysis of H&E-stained sections, which revealed no inflammatory reaction associated with the detected necrosis of hepatocytes in the MTX-treated group. This can be attributed to the fact that necrotic cell death, which occurs first, stimulates inflammatory responses, which might take a longer time to be detected [31]. However, the MTX dose administered in the current study has been reported previously to induce marked elevation of pro-inflammatory cytokines such as interleukin 6, interleukin 12, and TNF-α [17], which is contradicting our results. This contradiction can be justified by the difference in the time between MTX administration and rat scarification in both experiments. Further research is needed to resolve this issue and to better understand how MTX and crocin influence the inflammatory process.
In the present study, the expression of TGF-β1 mRNA was increased in MTX-treated rats, although no fibrosis was detected in the histological study of liver tissue from the same group. TGF-β1, a cytokine that promotes fibrosis, enhances collagen synthesis and deposition of extracellular matrix and can stimulate the transformation of hepatic stellate cells to myofibroblast [32]. It has been reported that TGF-β1 regulates the mitogen-activated protein kinase (MAPK) signaling and suppressor of mothers against decapentaplegic homolog $\frac{2}{3}$ (SMAD$\frac{2}{3}$) pathways to drive stellate cell activation [33]. Additionally, inhibiting the TGF-β1 signaling pathway can possibly interrupt the advancement of liver disease [34]. In our study, we detected that crocin's anti-fibrotic effects might be attributed to its ameliorative action on TGF-β1; hence, crocin helps to protect the liver against MTX-induced hepatotoxicity. These findings were in line with that of Algandaby [35], who discovered that crocin reduced TGF-β1 expression, which in turn prevented collagen deposition in the treated group's liver.
For more confirmation of our biochemical findings, we studied the putative protective action of crocin on the liver by undertaking extensive histopathological and immunochemical investigations. Compared to the negative control group, administration of MTX resulted in a significant distortion of the histological architecture of the hepatic tissue. Hepatocytes showed hydropic degeneration with variable stages of necrosis, as well as whole areas of cellular absence were detected with variable-sized areas of necrotic foci. Alternatively, pretreatment with crocin along with MTX improved liver architecture when compared to the MTX-treated group, which is consistent with Chhimwal et al. who demonstrated the protective effect of crocin on the histological structure of liver [36].
We further assessed the effect of crocin on hepatocyte apoptosis by assessing BAX (pro-apoptotic), BCL-2 (anti-apoptotic), and caspase-3 (apoptosis coordinating enzyme) protein expression in hepatic tissue of different study groups using IHC. Our results revealed that MTX administration enhanced the expression of the proteins BAX and caspase-3 while decreasing the expression of BCL-2, which led to an increase in hepatocyte apoptosis. However, the current study's findings indicated that concurrent administration of crocin and MTX reversed MTX-induced hepatotoxic effects by decreasing BAX and caspase-3 on one hand while increasing the expression of the BCL-2 protein on the other. These IHC findings were in agreement with several previous studies [22,28]. As a result, the current IHC data suggest that one possible mechanism of crocin's hepatoprotective actions against MTX-induced hepatotoxicity is through its influence on BAX, caspase-3, and BCL-2. Taken together, our findings strongly emphasize that crocin protects the liver from the damage caused by MTX, which can be attributed in part to its potent antioxidant and anti-fibrotic actions.
One important limitation of the current study is the short duration of crocin administration, whereas long-term use of crocin therapy may have had a more effective protective effect. Similarly, a single dose of MTX has been assessed; however, MTX is usually used as a part of regimens to treat chronic diseases. Though further research should concentrate on using different doses of MTX for longer durations. In addition, due to a limitation of resources, we were unable to investigate further the influence of MTX and crocin on the inflammatory process. More investigations are needed to validate our results and to further understand how crocin affects other molecular mechanisms implicated by MTX administration in the liver.
## Conclusions
In conclusion, our findings confirmed the hepatotoxic effect of MTX as its administration induced deterioration of liver enzymes (AST, ALT, and ALP), an increase in oxidative stress damage, lipid peroxidation, and apoptosis in the liver, as well as an increase in TGF-β1 expression. Significantly, the results of the current study demonstrated how crocin protects against MTX-induced hepatotoxicity. Our findings indicate that this effect is a result of the compound's antioxidant (reduced MDA and elevated GSH, CAT, and SOD), anti-apoptotic (upregulation of BCL-2 and downregulation of caspase-3 and BAX), and anti-fibrotic (downregulation of TGF-β1 gene expression) properties of crocin, as well as its ability to regulate the liver enzymes AST, ALT, and ALP. Furthermore, crocin treatment also restored the normal histological architecture of liver tissue in rats treated with MTX. As a result, the findings presented in this work utilizing an in vivo animal model support the idea that crocin should be explored further in humans to evaluate its putative hepatoprotective effects against MTX-induced hepatotoxicity.
## References
1. Demirci T, Gedikli S, Ozturk N, Aydemir Celep N. **The protective effect of N-acetylcysteine against methotrexate-induced hepatotoxicity in rat**. *Eurasian J Med Invest* (2019) **3** 219-226
2. Jalili C, Abdolmaleki A, Roshankhah S, Salahshoor MR. **Histopathological and biomedical parameters determination in the protective effect of crocin on hepatotoxicity induced by methotrexate in rats**. *J Herbmed Pharmacol* (2022) **9** 48-54
3. Bath RK, Brar NK, Forouhar FA, Wu GY. **A review of methotrexate-associated hepatotoxicity**. *J Dig Dis* (2014) **15** 517-524. PMID: 25139707
4. Bedoui Y, Guillot X, Sélambarom J. **Methotrexate an old drug with new tricks**. *Int J Mol Sci* (2019) **20** 5023. PMID: 31658782
5. Cao Y, Shi H, Sun Z. **Protective effects of magnesium glycyrrhizinate on methotrexate-induced hepatotoxicity and intestinal toxicity may be by reducing COX-2**. *Front Pharmacol* (2019) **10** 119. PMID: 30971913
6. Vaidya B, Bhochhibhoya M, Nakarmi S. **Efficacy of vitamin E in methotrexate-induced hepatotoxicity in rheumatoid arthritis: an open-label case-control study**. *Int J Rheumatol* (2020) **2020** 5723485. PMID: 32411250
7. Erfanparast A, Tamaddonfard E, Taati M, Dabbaghi M. **Effects of crocin and safranal, saffron constituents, on the formalin-induced orofacial pain in rats**. *Avicenna J Phytomed* (2015) **5** 392-402. PMID: 26468458
8. Hosseini A, Mousavi SH, Ghanbari A, Homaee Shandiz F, Raziee HR, Pezeshki Rad M, Mousavi SH. **Effect of saffron on liver metastases in patients suffering from cancers with liver metastases: a randomized, double blind, placebo-controlled clinical trial**. *Avicenna J Phytomed* (2015) **5** 434-440. PMID: 26468463
9. Chahine N, Makhlouf H, Duca L, Martiny L, Chahine R. **Cardioprotective effect of saffron extracts against acute doxorubicin toxicity in isolated rabbit hearts submitted to ischemia-reperfusion injury**. *Z Naturforsch C J Biosci* (2014) **69** 459-470. PMID: 25854766
10. Mehdizadeh R, Parizadeh M, Khooei AR, Mehri S, Hosseinzadeh H. **Cardioprotective effect of saffron extract and safranal in isoproterenol-induced myocardial infarction in Wistar rats**. *Iran J Basic Med Sci* (2013) **16** 56-63. PMID: 23638293
11. Razavi BM, Amanloo MA, Imenshahidi M, Hosseinzadeh H. **The relaxant activity of safranal in isolated rat aortas is mediated predominantly via an endothelium-independent mechanism - vasodilatory mechanism of safranal**. *J Pharmacopuncture* (2016) **19** 329-335. PMID: 28097042
12. Hussein SA, Ali AH, Ahmed TE. **The potential protective effect of crocin against hyperhomocysteinemia induced oxidative stress in rats**. *Benha Vet Med J* (2017) **33** 271-282
13. Mohammadi E, Mehri S, Badie Bostan H, Hosseinzadeh H. **Protective effect of crocin against d-galactose-induced aging in mice**. *Avicenna J Phytomed* (2018) **8** 14-23. PMID: 29387570
14. Bandegi AR, Rashidy-Pour A, Vafaei AA, Ghadrdoost B. **Protective effects of Crocus sativus L. extract and crocin against chronic-stress induced oxidative damage of brain, liver and kidneys in rats**. *Adv Pharm Bull* (2014) **4** 493-499. PMID: 25671180
15. Rajaei Z, Hadjzadeh MA, Nemati H, Hosseini M, Ahmadi M, Shafiee S. **Antihyperglycemic and antioxidant activity of crocin in streptozotocin-induced diabetic rats**. *J Med Food* (2013) **16** 206-210. PMID: 23437790
16. Mard SA, Azad SM, Ahangarpoor A. **Protective effect of crocin on gastric mucosal lesions induced by ischemia-reperfusion injury in rats**. *Iran J Pharm Res* (2016) **15** 93-99. PMID: 28228808
17. Chauhan P, Sharma H, Kumar U, Mayachari A, Sangli G, Singh S. **Protective effects of Glycyrrhiza glabra supplementation against methotrexate-induced hepato-renal damage in rats: an experimental approach**. *J Ethnopharmacol* (2020) **263** 113209. PMID: 32738390
18. Solaiman A, Sawires SKS. **The ameliorative potential of Alda-1 on experimentally induced liver fibrosis in adult male mice. A histological, immunohistochemical and biochemical study**. *Egypt J Histol* (2022) **45** 949-968
19. Amin SN, Hassan SS, Rashed LA. **Effects of chronic aspartame consumption on MPTP-induced Parkinsonism in male and female mice**. *Arch Physiol Biochem* (2018) **124** 292-299. PMID: 29096532
20. Livak KJ, Schmittgen TD. **Analysis of relative gene expression data using real-time quantitative PCR and the 2−ΔΔCT method**. *Methods* (2001) **25** 402-408. PMID: 11846609
21. Ragab AR, Elkablawy MA, Sheik BY, Baraka HN. **Antioxidant and tissue-protective studies on Ajwa extract: dates from Al Madinah Al-monwarah, Saudia Arabia**. *J Environ Anal Toxicol* (2013) **3** 163
22. Torlakovic EE, Nielsen S, Vyberg M, Taylor CR. **Getting controls under control: the time is now for immunohistochemistry**. *J Clin Pathol* (2015) **68** 879-882. PMID: 26286753
23. Karlsson Sundbaum J, Eriksson N, Hallberg P, Lehto N, Wadelius M, Baecklund E. **Methotrexate treatment in rheumatoid arthritis and elevated liver enzymes: a long-term follow-up of predictors, surveillance, and outcome in clinical practice**. *Int J Rheum Dis* (2019) **22** 1226-1232. PMID: 31012257
24. Akbari G, Mard SA, Dianat M, Mansouri E. **The hepatoprotective and microRNAs downregulatory effects of crocin following hepatic ischemia-reperfusion injury in rats**. *Oxid Med Cell Longev* (2017) **2017** 1702967. PMID: 28367266
25. Yucel Y, Oguz E, Kocarslan S. **The effects of lycopene on methotrexate-induced liver injury in rats**. *Bratisl Lek Listy* (2017) **118** 212-216. PMID: 28471231
26. Uraz S, Tahan V, Aygun C. **Role of ursodeoxycholic acid in prevention of methotrexate-induced liver toxicity**. *Dig Dis Sci* (2008) **53** 1071-1077. PMID: 17934844
27. Vardi N, Parlakpinar H, Cetin A, Erdogan A, Cetin Ozturk I. **Protective effect of beta-carotene on methotrexate-induced oxidative liver damage**. *Toxicol Pathol* (2010) **38** 592-597. PMID: 20448084
28. Al-Abkal F, Abdel-Wahab BA, El-Kareem HF, Moustafa YM, Khodeer DM. **Protective effect of pycnogenol against methotrexate-induced hepatic, renal, and cardiac toxicity: an in vivo study**. *Pharmaceuticals (Basel)* (2022) **15** 674. PMID: 35745592
29. Nedelcu RI, Balaban M, Turcu G. **Efficacy of methotrexate as anti-inflammatory and anti-proliferative drug in dermatology: three case reports**. *Exp Ther Med* (2019) **18** 905-910. PMID: 31384322
30. Sramek M, Neradil J, Veselska R. **Much more than you expected: the non-DHFR-mediated effects of methotrexate**. *Biochim Biophys Acta Gen Subj* (2017) **1861** 499-503. PMID: 27993660
31. Rock KL, Kono H. **The inflammatory response to cell death**. *Annu Rev Pathol* (2008) **3** 99-126. PMID: 18039143
32. Dewidar B, Meyer C, Dooley S, Meindl-Beinker AN. **TGF-β in hepatic stellate cell activation and liver fibrogenesis—updated 2019**. *Cells* (2019) **8** 1419. PMID: 31718044
33. Schmidt S, Messner CJ, Gaiser C, Hämmerli C, Suter-Dick L. **Methotrexate-induced liver injury is associated with oxidative stress, impaired mitochondrial respiration, and endoplasmic reticulum stress in vitro**. *Int J Mol Sci* (2022) **23** 15116. PMID: 36499436
34. Mohamed M, El Sheikh AK, Mohammed HH. **Modulation of liver P-glycoprotien expression may contribute to gossypin protection against methotrexate-induced hepatotoxicity**. *Indian J Pharmacol* (2021) **53** 25-30. PMID: 33975996
35. Algandaby MM. **Antifibrotic effects of crocin on thioacetamide-induced liver fibrosis in mice**. *Saudi J Biol Sci* (2018) **25** 747-754. PMID: 29740240
36. Chhimwal J, Sharma S, Kulurkar P, Patial V. **Crocin attenuates CCl4-induced liver fibrosis via PPAR-γ mediated modulation of inflammation and fibrogenesis in rats**. *Hum Exp Toxicol* (2020) **39** 1639-1649. PMID: 32633567
|
---
title: Senior Health Clinic for 75-year-old home-dwelling Finns – study design, clinic
protocol and non-response analysis
authors:
- Marika Salminen
- Sari Stenholm
- Jaana Koskenniemi
- Päivi Korhonen
- Tiina Pitkänen
- Paula Viikari
- Maarit Wuorela
- Matti Viitanen
- Laura Viikari
journal: BMC Health Services Research
year: 2023
pmcid: PMC9981251
doi: 10.1186/s12913-023-09199-9
license: CC BY 4.0
---
# Senior Health Clinic for 75-year-old home-dwelling Finns – study design, clinic protocol and non-response analysis
## Abstract
### Background
In the Finnish policy on older people preventive activities, which maintain functional capacity and independent living, are emphasized. The Turku Senior Health Clinic, aimed at maintaining independent coping of all home-dwelling 75-year-old citizens in the city of Turku, was founded in the beginning of 2020. The aim of this paper is to describe design and protocol of the Turku Senior Health Clinic Study (TSHeC) and provide results of the non-response analysis.
### Methods
The non-response analysis used data from 1296 participants ($71\%$ of those eligible) and 164 non-participants of the study. Sociodemographic, health status, psychosocial and physical functional ability indicators were included in the analysis. Participants and non-participants were also compared in respect to their neighborhood socioeconomic disadvantage. Differences between participants and non-participants were tested using the Chi squared or Fisher´s exact test for categorical variables and t-test for continuous variable.
### Results
The proportions of women ($43\%$ vs. $61\%$) and of those with only satisfying, poor or very poor self-rated financial status ($38\%$ vs. $49\%$) were significantly lower in non-participants than in participants. Comparison of the non-participants and participants in respect to their neighborhood socioeconomic disadvantage showed no differences. The prevalence of hypertension ($66\%$ vs. $54\%$), chronic lung disease ($20\%$ vs. $11\%$), and kidney failure ($6\%$ vs. $3\%$) were higher among non-participants compared to participants. Feelings of loneliness were less frequent among non-participants ($14\%$) compared to participants ($32\%$). The proportions of those using assistive mobility devices ($18\%$ vs. $8\%$) as well as those having previous falls ($12\%$ vs. $5\%$) were higher in non-participants than in participants.
### Conclusions
The participation rate of TSHeC was high. No neighborhood differences in participation were found. Health status and physical functioning of non-participants seemed to be slightly worse than those of the participants, and more women than men participated. These differences may weaken the generalizability of the findings of the study. The differences have to be taken into account when recommendation for the content and implementation of preventive nurse-managed health clinic in primary health care in *Finland is* going to be given.
### Trial registration
ClinicalTrials.gov Identifier: NCT05634239; registration date; 1st of December 2022. Retrospectively registered.
## Background
The Finnish policy on older people emphasizes the priority of living at home. Because the number of older people continues to grow, the number of those with morbidity, comorbidity and/or frailty is also growing. Thus, preventive activities, which maintain functional capacity and independent living, are highly emphasized [1].
Both cardiovascular diseases (CVDs) and dementia are highly prevalent among older people, and they share several modifiable risk factors supporting the possibility of preventive interventions [2, 3]. In people aged 75 or older, leading vascular metabolic risk factors are high systolic blood pressure, high fasting plasma glucose, diabetes, high body mass index and high LDL cholesterol [3, 4]. Metabolic risks are increasing, on average, every year, which means that no real progress in reducing behavioral risks has been achieved. Combination of aging population and increasing metabolic risks most likely maintains the increasing trends in non-communicable diseases [5]. Moreover, CVDs, impaired cognitive function and dementia are associated with a considerably increased risk of disability [3, 6–8], hospitalization [9, 10], institutionalization [11–13], and/or mortality [6, 9, 14].
Frailty is common among older people, especially among older women, even though the prevalence of frailty varies according to the measurement used [15–18]. Frailty has shown to be highly common in older people with CVDs [16, 19], and it worsens prognosis of the CVD patients [6, 9, 16]. Frailty is also strongly associated with dementia, cognitive impairment [8, 20] and multimorbidity [21]. Frailty increases the risk of hospitalization [22], institutionalization [23] and mortality [13, 15, 22]. In addition to metabolic risk factors, CVDs, cognitive impairment, dementia, and frailty, also pain [24], musculoskeletal conditions [25, 26], depression [26] and loneliness [27] may threaten functional ability and independent coping of older people and, thus, should be screened for preventive actions.
The key themes in the Finnish national recommendation to guarantee a good quality of life and improved services for older persons include promoting the functional capacity of older people, increasing voluntary work, utilizing digitalization and technologies, organizing and providing services, arranging guidance and service coordination, ensuring skilled personnel the quality of services [1]. During the past 10 years, various preventive health clinics for older people have already been implemented in municipalities and cities in Finland. However, to the best of our knowledge, no systematic assessment of the findings, applicability and/or effects of these procedures have been implemented so far.
The main purpose of this paper is to describe the study design and protocol of the Turku Senior Health Clinic Study (TSHeC). Results of the non-response analysis are also provided.
## Study population
TSHeC population consisted of all Finnish and Swedish speaking home-dwelling citizens born in 1945 in the city of Turku, in southwestern Finland in the beginning of 2020 ($$n = 2044$$). Those with municipal home care ($$n = 196$$) were excluded from the study population, 33 deceased before invitation, 382 refused to participate in the clinic´s health check, and 128 were not reached. Of those 1305 examined at the clinic, nine subjects declined to participate in the study, leaving 1296 study participants ($71\%$ of those eligible). The flow chart of the study is shown in Fig. 1.Fig. 1Flow chart of the study The study sample in non-response analysis included 1296 study participants examined at the clinic and 164 subjects who refused to participate ($43\%$ of 382 subjects who refused to participate) in the health clinic check but were willing to answers a short, structured telephone interview.
## Turku Senior Health Clinic Study
TSHeC was targeted at 75-year-old citizens with an underlying idea that at that age it would be almost the last moment to nudge people towards healthy lifestyle and taking care of their health and functional ability in order to maintain living independently.
The short-term aims of TSHeC are to survey health and functional statuses, and prevalence of specified risk factors for CVDs, dementia, frailty, and functional decline of 75-year-old independently home-dwelling citizens of the city of Turku. The aims also include assessment of the frequency of follow-up treatments needed, and recommendations given for lifestyle changes and evidence-based use of medication, as well as enforcement of these recommendations. Also, participants´ feedback on TSHeC will be assessed. In addition, based on the results, recommendations considering the protocol and implementation of preventive health clinics targeted at older people, will be provided. The long-term aim of the research project is to assess the effects of the TSHeC on the need of institutional care and home care provided by the city of Turku as well as the cost-effectiveness of the clinic during the 10-year follow-up. For this purpose, participants of TSHeC will be compared to non-participants and earlier cohorts of 75-year-olds in terms of the use of home care and institutional care.
## Recruitment
Before the clinic was founded, a couple of media articles about the upcoming, free of charge, health check were published in the local newspaper. Contact information of all home-dwelling citizens of the city of Turku born in 1945 was requested from the Finnish Digital and Population Data Services Agency. A clinic nurse contacted eligible subjects by phone. During the phone call, subjects were given information on TSHeC. A written invitation was sent to those who were not reached by phone. After receiving the written invitation, they were reached again by phone, twice, if needed. Those who declined to participate in the health check were encouraged to at least participate in a structured telephone interview. Those who refused were not contacted again. The personnel of the health clinic was bilingual, which eased the participation of Swedish-speaking subjects.
## Clinic protocol and data collection
TSHeC for 75-year-old independent home dwellers was implemented between January 2020 and June 2021 in the Turku City Hospital by three trained clinic nurses, two physiotherapists, and a consultative geriatrician. Appointments to health checks were scheduled to those willing to participate. They were sent written information and postal questionnaire concerning their sociodemographic, health behavior, health status, psychosocial and physical functional ability (Table 1). They were advised to take the filled questionnaire along to the appointments. Blood samples were drawn one week before the clinic appointments at the units of the Turku University Hospital laboratory (Tykslab) and analyzed at Tykslab. Clinic protocol is demonstrated in Fig. 2.Table 1Content of the Turku Senior Health Clinic checkPostal questionnaireNurse´s appointmentPhysiotherapist´s appointmentSociodemographicsgendermarital statusliving situationeducationeconomic statusHealth behavioursmokingalcohol usecircadian rhythmfrequency of exerciseHealth statusself-rated healthdiagnosed diseasescontinenceuse of medicationInterview- pain- fatigue/tiredness- mood- cognition (6CITa)- fracture risk (FRAXb)- risk for diabetes- oral healthClinical examination- height- weight- waist and hip circumferences- vision- hearing- blood pressure- pulse- orthostatic hypotensionPsychosocial functional abilityquality of lifesatisfactionlonelinesssocial participationhobbiesPhysical functional abilityneed for help in everyday livinginstrumental activities of everyday living physical activityrecent health related changes in physical activityfear of fallingInterview- use of assistive mobility device- managing in everyday living (walking 400 m, climbing stairs, doing housework, using public transportation vehicle, cutting toenails)- fear of falling (FES-Ic)- falls risk (FROP-Com Screend)Physical examination- hand grip strength- balance, walking speed, sit to stand (SPPBe)- one-leg stand- 30 s sit-to-stand- bening paroxysmal positional vertigoLaboratory tests (1 week before nurse´s appointment)- electrocardiogram- complete blood count- creatinine- alanine transaminase- glucose- thyreotropin- vitamin B12- folate- total, HDL and LDL cholesterol, triglyserides- calcium- vitamin DaSix Item Cognitive Impairment Test [28]bFracture Risk Assessment Tool [29]cFalls Efficacy Scale—International [30]dThree items of the Falls Risk for Older People in the Community tool [31]eShort Physical Performance Battery [32]Fig. 2Senior Health Clinic protocol During the 60–90-min appointment with the clinic nurse, results of the laboratory tests and the questionnaire filled beforehand were reviewed together with the participant. More information on the participant´s health status was gathered with an interview, and the participant was clinically examined by the clinic nurse. Participant´s health issues were discussed. If there was a need for and/or a possibility to lifestyle changes (e.g., in diet, exercising, social activation, weight control), the clinic nurse encouraged the participant to make those changes. After the appointment, the clinic nurse made a summary of each participant´s health status, which was reviewed together with the consultative geriatrician. The clinical nurse contacted the participant again if changes in medication and/or an appointment with health center physician was suggested by the geriatrician.
Appointment with the physiotherapist lasted from 30 to 45 min and included the assessment of physical functioning, the use of assistive mobility devices, and managing in everyday living. During the appointment, participants got individualized information on physical training and nutrition, especially protein intake, to maintain and/or improve their physical functioning. The content of the appointment was highly preventive and supportive, and information and suggestions given were based on participant´s level of physical functioning, motivation, and own goals. For those interested, group exercises suitable for their needs, e.g., muscle strengthening and/or balance training, were suggested. Participants were encouraged to maintain independence in mobility without mobility devices by improving muscle strength and balance. However, if there was a need for mobility devices to ensure safety, an appointment to the Assistive Technology Services of the city of Turku was scheduled. In case of musculoskeletal conditions, participants got self-care advice and/or an appointment with a health center physician, if needed.
If participant had symptoms or diseases that needed urgent medical care, geriatrician of Urgent Geriatric Outpatient Clinic was immediately consulted. In non-urgent cases, an appointment with health center physician, dentist of Oral and Dental Care, coordinator of the Memory Clinic, psychiatric nurse and/or dietician was scheduled. In addition to municipal services, services of local voluntary third-sector organizations and expert institutions were also recommended, and appointments scheduled if there was a need for rehabilitation, social activities, housing services, care, and supportive services as well as health-promoting activities. All municipal follow-up treatment facilities, personnel, and collaborators as well as voluntary third-sector collaborators are shown in Table 2.Table 2Follow-up treatment facilities, personnel, and collaborators of the Senior Health ClinicMunicipal services Urgent Geriatric Outpatient Clinic Local Health Centers/Stations Oral and Dental Care Memory Clinic Psychiatric nurse Dietician Service guidance for older people Welfare centers for older people Sport Services Centre Strength in Old Age Program Assistive Technology ServicesVoluntary third-sector collaborators Nine local registered voluntary organizations and expert institutions aimed to enhance psychosocial, mental, cognitive and physical functioning and quality of life of older people by organizing rehabilitation, social activities, housing services, care and supportive services as well as health-promoting activities for older people and their care givers Six months after the TSHeC check, the clinic nurse contacted participants for a structured follow-up telephone interview. During the 10–20-min interview, participants were asked how they experienced the clinic and the content of the clinic protocol, and about the responsiveness of the clinic protocol to their needs, and adherence to different changes (concerning medication, diet, exercising, weight control, and/or social activity) they were encouraged to make, if there were any.
## Ethics
The study was conducted according to the guidelines of the Declaration of Helsinki. The Ethics Committee of the Hospital District of Southwest Finland approved the study protocol (Diary number $\frac{87}{1801}$/2019). Participants provided written informed consent for the study.
## Statistical analyses
The present analyses used data of 1296 participants of the TSHeC check and 164 non-participants who were willing to participate in a short, structured telephone interview. For the non-response analysis, we included sociodemographic, health status, psychosocial and physical functional ability indicators. In addition, we compared participants and non-participants in respect to their neighborhood socioeconomic disadvantage. By using postal code information from the Statistics of Finland, standardized index of neighborhood socioeconomic disadvantage was calculated by using median household income in 2020 (coded as additive inverse), low educational attainment in 2020 (percentage of people over 18 years old with low education) and unemployment rate in 2019 (unemployed people belonging to the labor force/total labor force) [33]. For each of the three variables, we derived a standardized z score (national mean = 0, Standard deviation = 1). A total disadvantage score was then calculated by taking the mean value across all z scores; the mean of the score in the study population was 0.05 (range − 1.54 to 2.16), with a higher score indicating a higher disadvantage.
Differences between participants and non-participants were tested using the Chi squared or Fisher´s exact test for categorical variables and two-sample t-test for a continuous variable. P values less than 0.05 were considered statistically significant. All statistical analyses were performed using SAS System for Windows, version 9.4 (SAS Institute INC., Cary, NC, USA).
## Results
The proportions of women ($43\%$ vs. $61\%$) and of those with only satisfying, poor or very poor self-rated financial status ($38\%$ vs. $49\%$) were significantly lower in non-participants than those of participants were. Comparison of the non-participants and participants in respect to their neighborhood socioeconomic disadvantage, by using the data from the Statistics of Finland, showed no differences between the two groups. Of 13 diseases or chronic conditions that had possibly been previously diagnosed, significant differences between the non-participants and participants were found only in the prevalence of hypertension ($66\%$ vs. $55\%$), chronic lung disease ($22\%$ vs. $11\%$), and kidney failure ($6\%$ vs. $3\%$), all being higher among the non-participants. Feelings of loneliness were significantly less frequent among the non-participants ($14\%$) compared to the participants ($32\%$). The proportions of those using assistive mobility devices ($18\%$ vs. $8\%$) as well as those having falls during the previous 12 months ($12\%$ vs. $5\%$) were significantly higher in non-participants compared to the participants (Table 3).Table 3Characteristics of participants and non-participants of the Turku Senior Health Clinic StudyParticipants($$n = 1296$$)n (%)Non-participants($$n = 164$$)n (%)P-valueaFemale789 [61]71 [43] < 0.001Living alone483 [37]60 [37]0.865Neighborhood socioeconomic disadvantage index, mean (SD)0.27 (0.83)0.29 (0.82)0.831Education0.124 University248 [19]23 [14] Post-secondary level or university of applied sciences284 [22]30 [18] Vocational upper or general secondary education329 [25]53 [32] Basic education or none434 [34]58 [35]Self-rated financial status0.013 Very good or good668 [52]101 [62] Satisfying556 [44]60 [37] Poor or very poor62 [5]2 [1]Having someone who helps when needed0.054 Yes574 [45]67 [41] No45 [4]1 [1] No need for help656 [51]96 [59]Self-rated health0.311 Very good or good666 [51]77 [47] Moderate528 [41]69 [42] Poor or very poor102 [8]18 [11]Diabetes247 [20]38 [23]0.304Coronary artery disease143 [12]21 [13]0.641Myocardial infarction71 [6]10 [6]0.848Heart failure86 [7]14 [9]0.453Hypertension666 [54]108 [66]0.002Stroke or transient ischemic attack112 [9]16 [10]0.770Cancer270 [22]36 [22]0.904Chronic lung disease135 [11]32 [20]0.001Rheumatoid arthritis or osteoarthritis288 [24]30 [18]0.121Kidney failure33 [3]10 [6]0.016Parkinson´s disease17 [1]1 [1]0.712Mental disease52 [4]6 [4]0.743Other chronic disease326 [28]52 [32]0.366Feelings of loneliness < 0.001 Not at all890 [69]142 [87] Sometimes384 [30]21 [13] Often or always21 [2]1 [1]Depressive symptoms during the previous month143 [11]19 [12]0.840Feelings of fatigue during the previous month0.109 Not at all837 [65]101 [62] Sometimes332 [26]53 [32] Often–all the time122 [9]10 [6]Assistive mobility device109 [8]29 [18] < 0.001Self-rated ability to walk 400 m1254 [97]157 [96]0.644Self-rated ability to climb stair one floor at one go1268 [98]156 [96]0.070Number of falls during the previous 12 months0.003 None1221 [95]144 [88] 1–241 [3]18 [8] ≥ 329 [2]7 [4]Reduction of daily exercising during the previous 12 months299 [23]46 [28]0.167aX2-test or Fisher´s exact test for categorical and T-test for continuous variables Reasons for non-participation were not systematically documented, and only a part of the non-participants explained their reasons for non-participation. The most frequently mentioned reasons for non-participation were regular medical controls due to a chronic condition, regular health checks in private health care, and ongoing care of a severe illness. Fear of COVID-19 was mentioned only a few times.
## Discussion
All 75-year-old home-dwelling citizens in the city of Turku, who did not have municipal home care, were invited to the TSHeC. Intensive efforts were made to increase the response rate before and during the recruitment phase. During the survey preparation, media articles were published in the local newspaper and eligible subjects were contacted several times, if needed. Bilingual personnel of the health clinic eased the participation of Swedish-speaking subjects. With these evidence-based recruitment strategies [34], participation rate of $71\%$, that is consistent with those of other population-based Finnish studies among older people [12, 26], was achieved.
The protocol of TSHeC was implemented between January 2020 and June 2021. Due to the COVID-19 pandemic, a four and a half month´s break in health clinic appointments was held between March and August 2020. Although not all non-participants indicated reasons for non-participation, fear of COVID-19 was mentioned only a few times. However, non-participants of the Turku Senior Health Clinic Study were more likely to be male, less likely to suffer from feelings of loneliness and/or they had better self-rated financial status than the participants of the study. Factors associated with non-participation were earlier studied in Japanese population-based cohort study aimed to prevent lifestyle-related diseases [35], in a survey of Norwegian coronary heart disease patients [36], in a randomized controlled trial of multidomain lifestyle intervention for prevention of cognitive decline among French dementia-free subjects [37], and in a review article exploring characteristics of those who do and do not engage with preventive health checks [38]. In most earlier studies, non-participants were more likely women [35–37] and had lower socioeconomical statuses, which are inconsistent with the results of the present study [35, 37, 38]. Only according to the review article [38], non-participants were more likely men, as in the present study. It is to be noted, that most of the earlier studies were conducted among subjects of all-ages, also including subjects under the age of 65 years [35, 36, 38].
Because of the existing evidence of the association of socioeconomic status and health behavior, functional ability, health, as well as mortality in older people [39–41], the index of neighborhood socioeconomic disadvantage was added in non-response analyses. The aim was to examine if there are challenges in participation and, due to this, a need for targeted specific strategies for preventive actions in certain neighborhoods. The results of this study showed no difference in the neighborhood socioeconomic disadvantage level between the participants and non-participants, which suggests that people were willing and able to participate across the city of Turku. However, consistent with earlier studies [36, 42, 43], health status and physical functional ability of the non-participants seemed to be slightly worse than those of the participants. Non-participants have shown to value health less strongly, have lower interest in obtaining personal health information, have low self-efficacy, feel less in control of their health, prefer less time-demanding studies, and are less likely to believe in the efficacy of health checks [38, 43]. Found differences between participants and non-participants may deteriorate the generalizability of the findings of our study, and, have to be taken into account when recommendation for the content and implementation of preventive nurse-managed health clinic in primary health care in *Finland is* going to be given.
Because of the prevalence of frailty [15–18] and its associations to morbidity [6, 8, 9, 16, 19, 20], hospitalization [22], institutionalization [23] and mortality [15, 18, 22] among older people, preventive and supportive appointment with a physiotherapist was also included in the content of the TSHeC. All participants got the assessment of physical functional ability and individualized information to maintain and/or improve physical functioning and independent mobility by the clinic physiotherapist, not only those with a clear need for it.
Several chronic diseases and geriatric syndromes have overlapping risks and protective factors [44]. The aims of the nurse-managed TSHeC are to examine the health, functional ability, and risk factors of home-dwelling 75-year-olds and by tackling these risk factors, diminish adverse outcomes such as decline in functional ability, institutionalization, and death. Evidence exists that nurse-managed preventive health clinics for older adults with medical, functional and health behavior components improve the access to care, use of preventive services, support in the promotion of health, management of chronic diseases, adherence to treatment and patient satisfaction, patient outcomes, and reduce hospitalization [45–47]. By the structured telephone interview six months after the health check, the adherence to different activities suggested can be assessed.
The strength of this study is a large sample size which e.g., enables us to determine subgroups with best adherence for different suggested health behavior changes. Also, some limitations must be mentioned. Reasons for non-participation were not systematically recorded. Another limitation is the low response rate of the non-participants; less than half of the non-participants were willing to participate a short, structured telephone interview. It is possible that non-participants who refused to answer even a short telephone interview had worse health status and physical functional ability compared to those who were willing to answer. This may have biased the data of non-participants.
Both by the nurturing and supportive health check and by the six-month follow-up telephone interview, participants were nudged towards a healthy lifestyle, taking care of their health and functional ability to maintain independent coping. TSHeC will add knowledge about the health and functional ability of community-dwelling older adults as well as adherence to recommended health behavior changes. Based on these results of the TSHeC, recommendations for the protocol and implementation of preventive nurse-managed health clinics in primary health care in *Finland is* aimed to be given to support more homogeneous preventive service for older people in Finland.
## References
1. 1.Ministry of Social Affairs and Health. Quality recommendation to guarantee a good quality of life and improved services for older persons 2020–2023. Publications of the Ministry of Social Affairs and Health 2020:37. Helsinki, 2020. https://julkaisut.valtioneuvosto.fi/bitstream/handle/10024/162595/STM_2020_37_J.pdf?sequence=1&isAllowed=y. Accessed 10 Jan 2022.
2. Leszek J, Mikhaylenko EV, Belousov DM. **The links between cardiovascular diseases and Alzheimer’s disease**. *Curr Neuropharmacol* (2021.0) **19** 152-169. DOI: 10.2174/1570159X18666200729093724
3. Lisko I, Kulmala J, Annetorp M. **How can dementia and disability be prevented in older adults: where are we today and where are we going?**. *J Intern Med* (2021.0) **289** 807-830. DOI: 10.1111/joim.13227
4. **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** 1223-49. DOI: 10.1016/S0140-6736(20)30752-2
5. **Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017**. *Lancet.* (2018.0) **392** 1923-94. DOI: 10.1016/S0140-6736(18)32225-6
6. Forman DE. **The importance of physical function as a clinical outcome: assessment and enhancement**. *Clin Cardiol.* (2020.0) **43** 108-17. DOI: 10.1002/clc.23311
7. **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.* (2020.0) **396** 1204-22. DOI: 10.1016/S0140-6736(20)30925-9
8. Petermann-Rocha F, Lyall DM, Gray SR. **Associations between physical frailty and dementia incidence: a prospective study from UK Biobank**. *Lancet* (2020.0) **1** e58-68. DOI: 10.1016/s2666-7568(20)30007-6
9. Fuentes-Abolafio IJ, Stubbs B, Pérez-Belmonte LM. **Physical functional performance and prognosis in patients with heart failure: a systematic review and meta-analysis**. *BMC Cardiovasc Disord* (2020.0) **20** 512. DOI: 10.1186/s12872-020-01725-5
10. Sharar R, Lutski M, Zucker I, Weinstein G. **Risk for hospitalization surrounding dementia diagnosis: a national registry-based study**. *Alzheimer Dement.* (2021.0) **17** e055394. DOI: 10.1002/alz.055394
11. Hajek A, Brettschneider C, Lange C. **Longitudinal Predictors of Institutionalization in Old Age**. *Plos One.* (2015.0) **10** e0144203. DOI: 10.1371/journal.pone.014420
12. Salminen M, Vire J, Viikari L. **Predictors of institutionalization among home-dwelling older Finnish people: a 22-year follow-up study**. *Aging Clin Exp Res* (2017.0) **29** 499-505. DOI: 10.1007/s40520-016-0722-3
13. Salminen M, Laine J, Vahlberg T. **Factors associated with institutionalization among home-dwelling patients of Urgent Geriatric Outpatient Clinic: a 3-year follow-up study**. *Eur Ger Med* (2020.0) **11** 745-751. DOI: 10.1007/s41999-020-00338-7
14. **Global mortality from dementia: Application of a new method and results from the Global Burden of Disease Study 2019**. *Alzheimers Dement.* (2021.0) **7** e12200. DOI: 10.1002/trc2.12200
15. Koivukangas MM, Hietikko E, Strandberg T. **The prevalence of frailty using three different frailty measurements in two Finnish cohorts born before and after the Second World War**. *J Nutr Health Aging* (2021.0) **25** 611-617. DOI: 10.1007/s12603-021-1586-6
16. Marinus N, Vigorito C, Giallauria F. **Frailty is highly prevalent in specific cardiovascular diseases and females, but significantly worsens prognosis in all affected patients: A systematic review**. *Ageing Res Rev.* (2021.0) **66** 101233. DOI: 10.1016/j.arr.2020.101233
17. Sezgin D. **Prevalence of frailty in 62 countries across the world: a systematic review and meta-analysis of population-level studies**. *Age Ageing.* (2021.0) **50** 96-104. DOI: 10.1093/ageing/afaa219
18. Salminen M, Viljanen A, Eloranta S. **Frailty and mortality: an 18-year follow-up study among Finnish Community-dwelling older people**. *Aging Clin Exp Res.* (2020.0) **32** 2013-19. DOI: 10.1007/s40520-019-01383-4
19. Kleipool EEF, Hoogendijk EO, Trappenburg MC. **Frailty in older adults with cardiovascular disease: cause, effect or both?**. *Aging Dis.* (2018.0) **9** 489-97. DOI: 10.14336/AD.2017.1125
20. Kulmala J, Nykänen I, Mänty M, Hartikainen S. **Association between frailty and dementia: a population-based study**. *Gerontology* (2014.0) **60** 16-21. DOI: 10.1159/000353859
21. Vetrano DL, Palmer K, Marengoni A. **Frailty and Multimorbidity: A Systematic Review and Meta-analysis**. *J Gerontol A Biol Sci Med Sci* (2019.0) **74** 659-666. DOI: 10.1093/gerona/gly110
22. Landré B, Aegerter P, Zins M, Goldberg M, Ankri J, Herr M. **Association between hospitalization and change of frailty status in the GAZEL cohort**. *J Nutr Health Aging* (2019.0) **23** 466-473. DOI: 10.1007/s12603-019-1186-x
23. Viljanen A, Salminen M, Irjala K. **Frailty, walking ability and self-rated health in predicting institutionalization: an 18-year follow-up study among Finnish community-dwelling older people**. *Aging Clin Exp Res* (2021.0) **33** 547-554. DOI: 10.1007/s40520-020-01551-x
24. Karjalainen M, Saltevo J, Tiihonen M. **Frequent pain in older people with and without diabetes – Finnish community-based study**. *BMC Geriatr* (2018.0) **18** 73. DOI: 10.1186/s12877-018-0762-y
25. 25.Kaila-Kangas L. Musculoskeletal disorders and diseases in Finland. Results of the Health 2000 Survey. Publications of the National Public Health Institute B25/2007. Helsinki, 2007. https://www.julkari.fi/bitstream/handle/10024/78197/2007b25.pdf?sequence=1&isAllowed=y. Accessed 11 Feb 2022.
26. 26.Koponen P, Borodulin K, Lundqvist A, Sääksjärvi K and Koskinen S, eds. Health, functional capacity and welfare in Finland – FinHealth 2017 study. National Institute for Health and Welfare, Report 4/2018. Helsinki 2018. https://www.julkari.fi/bitstream/handle/10024/136223/Rap_4_2018_FinTerveys_verkko.pdf?sequence=1&isAllowed=y. Accessed 21 Nov 2022.
27. Nyqvist F, Näsman M, Hemberg J, Nygård M. **Risk factors for loneliness among older people in a Nordic regional context – a longitudinal study**. *Ageing Soc* (2021.0). DOI: 10.1017/S0144686X21001707
28. Brooke P, Bullock R. **Validation of a 6 item cognitive impairment test with a view to primary care usage**. *Int J Geriatr Psychiatry* (1999.0) **14** 936-940. DOI: 10.1002/(SICI)1099-1166(199911)14:11<936::AID-GPS39>3.0.CO;2-1
29. 29.WHO Scientific Group Technical Report. Assessment of osteoporosis at the primary health care level. https://www.sheffield.ac.uk/FRAX/pdfs/WHO_Technical_Report.pdf. Accessed 13 March 2022.
30. Yardley L, Beyer N, Hauer K. **Development and initial validation of the Falls Efficacy Scale-International (FES-I)**. *Age Ageing.* (2005.0) **34** 614-9. DOI: 10.1093/ageing/afi196
31. Russell MA, Hill KD, Day LM. **Development of the Falls Risk for Older People in the Community (FROP-Com) screening tool**. *Age Ageing* (2009.0) **38** 40-46. DOI: 10.1093/ageing/afn196
32. Guralnik JM, Simonsick EM, Ferrucci L. **A Short Physical Performance Battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission**. *J Gerontol* (1994.0) **49** M85-M94. DOI: 10.1093/geronj/49.2.m85
33. Halonen JI, Pulakka A, Stenholm S. **Change in neighborhood disadvantage and change in smoking behaviors in adults: a longitudinal, within-individual study**. *Epidemiol* (2016.0) **27** 803-809. DOI: 10.1097/EDE.0000000000000530
34. Lacey RJ, Wilkie R, Wynne-Jones G. **Evidence for strategies that improve recruitment and retention of older adults aged 65 years and over in randomized trials and observational studies: a systematic review**. *Age Ageing* (2017.0) **46** 895-903. DOI: 10.1093/ageing/afx057
35. Hara M, Shimanoe C, Otsuka Y. **Factors associated with non-participation in face-to-face second survey conducted 5 years after the baseline survey**. *J Epidemiol* (2015.0) **25** 117-125. DOI: 10.2188/jea.JE20140116
36. Munkhaugen J, Sverre E, Peersen K. **Patient characteristics and risk factors of participants and non-participants in the NOR-COR study**. *Scand Cardiovasc J* (2016.0) **50** 317-322. DOI: 10.1080/14017431.2016.1202445
37. Coley N, Coniasse-Brioude D, Igier V. **Disparities in the participation and adherence of older adults in lifestyle-based multidomain dementia prevention and the motivational role of perceived disease risk and intervention benefits: an observational ancillary study to a randomized controlled study**. *Alzheimers Res Ther* (2021.0) **13** 157. DOI: 10.1186/s13195-021-00904-6
38. Dryden R, Williams B, McCowan C. **What do we know about who does and who does not attend to general health checks? Findings from a narrative scoping review**. *BMC Public Health* (2012.0) **12** 723. DOI: 10.1186/1471-2458-12-723
39. Rautio N, Heikkinen E, Ebrahim S. **Socio-economic position and its relationship to physical capacity among elderly people living in Jyväskylä, Finland: five- and ten-year follow-up studies**. *Soc Sci Med.* (2005.0) **60** 2405-16. DOI: 10.1016/j.socscimed.2004.11.029
40. Mäki N, Martikainen P, Eikemo T. **Educational differences in disability-free life expectancy: a comparative study of long-standing activity limitation in eight European countries**. *Soc Sci Med* (2013.0) **94** 1-8. DOI: 10.1016/j.socscimed.2013.06.009
41. Khalatbari-Soltani S, Blyth FM, Naganathan V. **Socioeconomic status, health-related behaviours, and death among older people: the Concord health and aging in men project prospective cohort study**. *BMC Geriatr* (2020.0) **20** 261. DOI: 10.1186/s12877-020-01648-y
42. van Heuvelen MJG, Hochstenbach JBM, Brouwer WH. **Differences between participants and non-participants in an RCT on physical activity and psychological interventions for older persons**. *Aging Clin Exp Res* (2005.0) **17** 45. DOI: 10.1007/BF03324603
43. Akmatov MK, Jentsch L, Riese P. **Motivations for (non)participation in population-based health studies among the elderly comparison of participants and non-participants of a prospective study on influenza vaccination**. *BMC Med Res Methodol* (2017.0) **17** 18. DOI: 10.1186/s12874-017-0302-z
44. Kivipelto M, Solomon A, Ahtiluoto S. **The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER): study design and progress**. *Alzheimers Dement* (2013.0) **9** 657-665. DOI: 10.1016/j.jalz.2012.09.012
45. Coddington JA, Sands LP. **Cost of health care and quality outcomes of patients at nurse-managed clinics**. *Nurs Economics* (2008.0) **26** 75-83
46. Link D, Perry D, Cesarotti E. **Meeting new health care challenges with a proven innovation. Nurse-managed health care clinics**. *Nurs Adm Q.* (2014.0) **38** 128-32. DOI: 10.1097/NAQ.0000000000000004
47. 47.Randall S, Crawford T, 2 , Currie J, River J, Betihavas V. Impact of community based nurse-led clinics on patient outcomes, patient satisfaction, patient access and cost effectiveness: A systematic review. Int J Nurs Stud 2017;73:24–33. 10.1016/j.ijnurstu.2017.05.008
|
---
title: 'Participant recruitment, baseline characteristics and at-home-measurements
of cardiometabolic risk markers: insights from the Supreme Nudge parallel cluster-randomised
controlled supermarket trial'
authors:
- Josine M. Stuber
- Beryl A. C. E. van Hoek
- Anne L. Vos
- Edith G. Smit
- Jeroen Lakerveld
- Joreintje D. Mackenbach
- Joline W. J. Beulens
- Jody C. Hoenink
- Jody C. Hoenink
- Femke Rutters
- Wilma E. Waterlander
- Denise T. D. de Ridder
- Marleen Gillebaart
- Stephanie Blom
- Femke E. de Boer
- Gert-Jan de Bruijn
- Michel C. A. Klein
- Jacqueline E. W. Broerse
- Tjerk-Jan Schuitmaker-Warnaar
- Cédric N. H. Middel
- Yvonne T. van der Schouw
- Ivonne Sluijs
- Marjolein C. Harbers
- Elizabeth Velema
journal: Trials
year: 2023
pmcid: PMC9981252
doi: 10.1186/s13063-023-07157-8
license: CC BY 4.0
---
# Participant recruitment, baseline characteristics and at-home-measurements of cardiometabolic risk markers: insights from the Supreme Nudge parallel cluster-randomised controlled supermarket trial
## Abstract
### Background
Recruiting participants for lifestyle programmes is known to be challenging. Insights into recruitment strategies, enrolment rates and costs are valuable but rarely reported. We provide insight into the costs and results of used recruitment strategies, baseline characteristics and feasibility of at-home cardiometabolic measurements as part of the Supreme Nudge trial investigating healthy lifestyle behaviours. This trial was conducted during the COVID-19 pandemic, requiring a largely remote data collection approach. Potential sociodemographic differences were explored between participants recruited through various strategies and for at-home measurement completion rates.
### Methods
Participants were recruited from socially disadvantaged areas around participating study supermarkets ($$n = 12$$ supermarkets) across the Netherlands, aged 30–80 years, and regular shoppers of the participating supermarkets. Recruitment strategies, costs and yields were logged, together with completion rates of at-home measurements of cardiometabolic markers. Descriptive statistics are reported on recruitment yield per used method and baseline characteristics. We used linear and logistic multilevel models to assess the potential sociodemographic differences.
### Results
Of 783 recruited, 602 were eligible to participate, and 421 completed informed consent. Most included participants were recruited via letters/flyers at home ($75\%$), but this strategy was very costly per included participant (89 Euros). Of paid strategies, supermarket flyers were the cheapest (12 Euros) and the least time-invasive (< 1 h). Participants who completed baseline measurements ($$n = 391$$) were on average 57.6 (SD 11.0) years, $72\%$ were female and $41\%$ had high educational attainment, and they often completed the at-home measurements successfully (lipid profile $88\%$, HbA1c $94\%$, waist circumference $99\%$). Multilevel models suggested that males tended to be recruited more often via word-of-mouth (ORfemales 0.51 ($95\%$CI 0.22; 1.21)). Those who failed the first attempt at completing the at-home blood measurement were older (β 3.89 years ($95\%$ CI 1.28; 6.49), whilst the non-completers of the HbA1c (β − 8.92 years ($95\%$ CI − 13.62; − 4.28)) and LDL (β − 3.19 years ($95\%$ CI − 6.53; 0.09)) were younger.
### Conclusions
Supermarket flyers were the most cost-effective paid strategy, whereas mailings to home addresses recruited the most participants but were very costly. At-home cardiometabolic measurements were feasible and may be useful in geographically widespread groups or when face to face contact is not possible.
### Trial registration
Dutch Trial Register ID NL7064, 30 May 2018, https://trialsearch.who.int/Trial2.aspx?TrialID=NTR7302
### Supplementary Information
The online version contains supplementary material available at 10.1186/s13063-023-07157-8.
## Background
Cardiometabolic diseases (CMD), including type 2 diabetes and cardiovascular diseases, pose a major individual and societal burden [1, 2]. Differences in prevalence of CMD and lifestyle behaviours have been associated with socioeconomic position (SEP), with a higher CMD burden for individuals with a low SEP [3]. It is well established that maintaining healthy lifestyle behaviours reduces CMD risk, but this is not straightforward for most individuals. For example, population surveys on dietary patterns show that at least half of Dutch adults do not adhere to dietary guidelines and even less than $25\%$ consume enough fruit and vegetables [4], despite the widespread availability of information on the composition of a healthy diet.
It is clear that solely providing information on a healthy lifestyle is not sufficient to change behaviour, as many other individual-level factors than knowledge affect lifestyle behaviour, including motivations, beliefs, personal resources and preferences [5, 6]. In turn, these factors are driven by more ‘upstream determinants’ in the social-cultural environment (e.g. cultural beliefs on what is healthy), the built environment (e.g. availability of healthy food outlets), the economic environment (e.g. food prices) and the policy environment (e.g. health-promotion campaigns) [6–12]. Healthy lifestyle programmes that are designed to address these individual-level and upstream determinants could contribute to a CMD risk reduction and reduce health inequalities.
The effectiveness of lifestyle programmes is usually evaluated by randomised controlled trials, for which the inclusion of sufficient study participants is paramount. Low inclusion rates can result in a lack of statistical power to evaluate study findings [13, 14], leading to a waste of research funding. However, including a sufficient number of participants is known to be challenging. Well-designed recruitment strategies which are tailored to the target group are, therefore, crucial. Tailoring can be based on sociodemographic, sociocultural and socioeconomic characteristics and the developed recruitment strategy should match the study type and its design [15].
The combination of active and passive recruitment strategies is often used. Examples are the use of flyers, email invites or social media posts (all passive strategies) and face-to-face recruitment at the location of the target population (active strategy) [16–19]. Passive recruitment strategies are reported to be more cost-effective in the recruitment of participants for healthy lifestyle programmes than active strategies [18–20]. Active recruitment strategies, however, seem most effective in reaching those who would be likely to benefit most from a lifestyle programme [19, 20]. For instance, Smit et al. [ 2021] showed that of all participants enrolled in their trial, the healthiest participants in terms of physical activity guideline adherence were recruited via passive strategies such as personal letters/flyers, social media and word-of-mouth. In contrast, those with lower physical activity guideline adherence were more often recruited via health professional referrals (active strategy) [19].
Previous studies show that adults with a low SEP are a diverse group for whom various barriers exist to participate in healthy lifestyle programmes. Also, response rates are often lower than for those with a high SEP [21–24]. Examples of barriers to participation are psychosocial barriers, such as having other priorities or being sceptical, study-related challenges such as overly complex study materials and other reasons such as lack of motivation [21], family issues or financial constraints [25]. As such, it is important to account for potential barriers among the target group prior to recruitment (e.g. by providing easy-to-understand recruitment materials). In addition, when aiming to reach a diverse study population, it is likely important to use a variety of recruitment strategies [21, 25]. Yet, it is unclear which strategies can be most (cost)effective to reach a study sample consisting of a variety of individuals, including for example individuals with a low SEP.
Insights from projects into recruitment strategies and costs, participant enrolment rates and characteristics of the included study sample are valuable to learn from and for planning of new projects. However, detailed recruitment information and especially insights into costs and yield in terms of participant characteristics are rarely reported. We conducted a real-life trial—the Supreme Nudge trial—that aimed to promote healthy lifestyle behaviours within socially disadvantaged neighbourhoods in the Netherlands [26, 27]. We focused on healthy food choices via the implementation of nudging and pricing strategies in real-life supermarkets and on healthy walking behaviours via a smartphone coaching app. This trial was conducted during the COVID-19 pandemic, which required a predominantly remote recruitment and data collection approach. This paper aims to provide insight into the costs, feasibility and results of used recruitment strategies, the baseline characteristics of the included study sample and the feasibility of at-home measurement of cardiometabolic risk markers. Furthermore, we explored potential sociodemographic differences in the most effective recruitment strategies and completion of the at-home measurements of cardiometabolic risk makers.
## The Supreme Nudge trial
Data presented in this paper are based on the recruitment and baseline measurement of the Supreme Nudge parallel cluster-randomised controlled trial, of which the study protocol was previously published [26]. Briefly, in the Supreme Nudge trial, we implemented nudging and pricing strategies in supermarkets to promote healthy food purchases. In addition, a smartphone physical activity coaching app and a step counting app were randomised on the individual level across all supermarket clusters with the aim to increase daily step counts by providing this dynamically tailored coaching content. Participants in the physical activity intervention group received the step counter app plus a mobile coaching app sending messages, and participants in the control group received only the step counter app. Our target sample size was to include at least 360 participants to secure sufficient power for evaluation of the primary trial endpoint [26].
Twelve supermarkets of a Dutch supermarket chain within socially disadvantaged neighbourhoods participated in the trial. Supermarkets were located in relatively rural areas across different towns in the Netherlands. A socially disadvantaged neighbourhood was defined based on a below national average socioeconomic status score as calculated by The Netherlands Institute for Social Research [28]. The status scores were based on average levels of education, income and employment per postal code. The intervention implementation duration was 6 months to 1 year, depending on the supermarket enrolment date (further explained below). Participants were recruited from socially disadvantaged areas surrounding the participating supermarkets, and further eligibility criteria included: being between 30 and 80 years of age, being a regular shopper at a participating supermarket (> $50\%$ of weekly household groceries purchased at participating supermarket), having the intention to continue visiting the supermarket for the next (half) year and being able to communicate in the Dutch language adequately. Additional individual-level randomisation in the smartphone physical activity intervention required participants to use a smartphone with a mobile data plan and Android 8, iOS 13 or more recent software versions installed, to have experience with mobile text messaging and to have no contra-indication for light physical activity. Follow-up measurements took place after 3 months, 6 months, and, depending on supermarket enrolment date, 12 months.
Participant recruitment and data collection for the trial were initially planned to take place in face-to-face appointments with participants. However, the COVID-19 pandemic forced us to adapt our protocols to a largely remote recruitment and data collection approach [26].
## Recruitment strategies
A stepwise recruitment strategy was applied, starting with a set of passive strategies followed by active strategies. The recruitment period ran from mid-January until November 2021 (Additional file 1: Supplementary Table 1). Passive strategies used first were news articles in local (online) media and flyers distributed in the supermarket and via mail. Next, Facebook posts were placed on the participating supermarket pages, posters were displayed in-store and at some other relevant locations in the neighbourhood (e.g. physiotherapy practices) and postal invitation letters were sent to every household of the municipality around the included supermarkets. Thereafter, the supermarket’s customer panels were sent an email invitation, a municipality-targeted social media campaign was launched (Facebook and Instagram) and an advertisement was placed on the website of the Dutch Heart Foundation (study funder). Moreover, two active recruitment approaches were used in addition to the passive strategies. The first was to encourage included participants via phone to ask their partner or neighbours to register for screening (promoting word-of-mouth), and the second was recruitment in the supermarket by researchers. As recruitment phases took place during 10 weeks in the spring and autumn of 2021, COVID-19-related restrictions were in place with varying intensity; a national lockdown was in effect at the start of the recruitment period, which was loosened by late spring. Therefore, the active in-store recruitment by the research team was conducted only when the restrictions allowed this—approximately halfway through the recruitment period. The recruitment materials mentioned that participants would receive a free grocery box after successful study completion.
## Data collection
The incurred costs and total staff hours needed for each recruitment strategy were logged by the research team in spread sheets. Total material cost (Euros) consisted of the sum of material costs spent on paper, envelopes and mailing costs of the recruitment letters or for example travel costs for in-store recruitment. Our supermarket partner helped with the design of the flyers, which was free of cost for the research team. Total time spent (hours) by research staff was calculated as the sum of all staff hours, which consisted of a combination of (unpaid) research interns and (paid) research assistants and pre-doctoral research fellows. All costs were made during the recruitment period. Information on participant registrations, eligibility, inclusion rates and measurement completion were centralised via a cloud-based clinical data management platform (Castor Electronic Data Capture). The screening questionnaire included an item asking participants how they had heard about the trial. Participants could provide multiple answers based on a pre-defined list describing all applied recruitment strategies, plus an open-ended answering option.
Data were self-reported and self-assessed by participants at home, considering the COVID-19-related restrictions. Data collection and study procedures are reported in a protocol paper [26]. This paper also reports in detail the exact measurements at the different time points during follow-up of the trial. Briefly, via the screening and baseline questionnaires, data were collected on participant characteristics including sex (male, female), age, highest completed educational attainment (here categorised as follows: low educational attainment: no education and primary education, medium educational attainment: secondary educational attainments, high educational attainment: tertiary educational attainments), household size (number of adults, number of children) and smoking status (current, irregular, former, never). In addition, questionnaire items on medical history asked participants about prevalent type 2 diabetes, hypertension, hyperlipidaemia and cardiovascular diseases and about medication use for these health conditions.
The primary trial outcome was adherence to the Dutch dietary guidelines, measured via the Dutch Healthy Diet 2015 food frequency questionnaire. Adherence to the Dutch dietary guidelines could range from 0, reflecting no adherence, to 150, reflecting full adherence (DHD15-index scores) [29]. Secondary outcomes were cardiometabolic risk markers, daily step counts, questionnaire items related to lifestyle behaviours and healthy food purchases, which are all further detailed below.
Cardiometabolic risk markers comprised of waist circumference (cm), HbA1c (mmol/mol), LDL-cholesterol (mmol/L), HDL-cholesterol (mmol/L), total cholesterol (mmol/L), total cholesterol/HDL-ratio and triglycerides (mmol/L). At-home measurement of cardiometabolic markers required participants to self-assess their waist circumference and perform a finger prick to collect capillary blood into two small capillary tubes. Instructional materials for the at-home measurements included a step-by-step instruction letter and a web-based video. The blood samples were sent back to the hospital lab via a medical return envelope in the mail, which took on average one day before the sample arrived in the lab. Results were analysed by enzymatic colorimetric test via the Roche/Hitachi Cobas C systems. In contrast to HbA1c, blood lipids remain less stable at room temperature (i.e. during mailing of the sample). As such, more invalid measurements of lipid profiles were expected. When the lab indicated the blood sample was invalid, participants were invited to collect another blood sample up to a maximum of three attempts. Research staff conducted home visits for participants requesting assistance with the cardiometabolic measurements as soon as COVID-19 restrictions allowed.
Data collection regarding the healthy food purchases were based on a supermarket loyalty card which participants were instructed to use during each supermarket visit for the complete study period. A baseline measurement for this outcome was defined as an average total percentage of healthy food and beverage purchases during a 4-week pre-intervention period, for those participants who started using a loyalty card during this period.
Daily step counts were measured via a step counter app that was developed as part of the Supreme Nudge app-based physical activity coaching system (SNapp). The development process of SNapp is described in greater detail elsewhere [30]. In brief, the app continuously quantifies step count using the smartphone’s built-in pedometer or accelerometer. Every hour, the app synchronises with a database server to store the user’s current step count level, date, time and type of sensor used for step count tracking (i.e. either a pedometer or accelerometer). The app is compatible with Android and iOS devices. Participants were asked to install the app based on textual and video instructions and to run the app on their personal smartphones during the Supreme Nudge trial. Participants received written instructions on how to use the app and were encouraged to carry their phones on their bodies as much as possible throughout the day. For the purpose of this study, the number of steps taken per day was averaged over a 7-day baseline period. We conducted a small validation study to explore whether the step counts derived from our app corresponded to step counts derived from default smartphone step counters (e.g. Samsung Health or the iOS Health app) and to step counts derived from accelerometers. We asked a convenience sample of 20 participants to wear GT3X ActiGraphs for at least 7 days and to report their wear time (start- and end time) in a diary. We also asked them to install our step counter app and, if available, report daily step count derived from their smartphones’ default step counters. Eleven participants had sufficient data available (≥ 4 days of GT3X ActiGraph accelerometer wear time and ≥ 4 days of reported step counts based on their default smartphone step counter). The Spearman’s correlation coefficient comparing the step counts derived from our app with those derived from default smartphone step counters was 0.62. The Spearman’s correlation coefficient comparing the step counts derived from our app with those derived from the accelerometers was 0.66. Based on these results, we considered our app to perform acceptable in counting steps in order to serve our research purposes.
Questionnaire data related to lifestyle behaviours were food decision styles, including reflective [31], habitual [32, 33] and impulsive [31, 34, 35] decision styles, nudges and social cognitive factors which measured health goals [36, 37], healthy shopping [38, 39], perceived social norm [40] and attractiveness of healthy foods [41], customer satisfaction measured via one general satisfaction score and eight sub-components and walking behaviours and social cognitive factors [42, 43]. All these questionnaire items were measured on 7-point Likert scales, ranging from fully disagree [1] to fully agree [7].
## Study outcomes and statistical analyses
The material costs, feasibility and yield of used recruitment strategies were assessed via descriptive statistics. Descriptive statistics report the total material cost (Euros) and total time spent (hours) by research staff for each applied recruitment strategy. In addition, the number and percentage (n(%)) of participants, the material costs, and hours spent per recruitment strategy are presented for (a) registered participants who completed eligibility assessment, (b) eligible participants, (c) included participants in the cluster-randomised trial, and (d) participants who completed the baseline measurement for the primary study outcome. Calculations of costs and hours spent per strategy are based on the total costs/hours per strategy divided by the number of participants (a) registered, (b) eligible, (c) included and (d) completed the baseline measurement for the primary trial outcome per recruitment strategy. All hours are calculated including travel time starting from the research office. Moreover, percentages are presented for the proportion of participants who indicated to have been recruited via one or two or more strategies.
Participant baseline characteristics and baseline measurements on trial outcomes are presented for the total sample. Continuous variables are described by the mean and standard deviation (SD) for normally distributed variables or by the median and interquartile range (IQR) for non-normally distributed variables. Categorical variables are presented by their number and percentage. Participant characteristics were explored stratified by supermarket location for visual inspection of potential differences in participant characteristics. Considering the physical activity intervention was randomised and implemented at the individual level, baseline characteristics and step count data for the control versus intervention participants of the physical activity coaching app intervention are presented in a separate baseline table.
To explore potential sociodemographic differences (age, sex, and educational attainment) within the top five most effective recruitment strategies, we assessed associations between the recruitment strategies (independent variables) and the sociodemographic variables (dependent variables) using linear multilevel models for age and logistic multilevel models for sex and educational attainment. All models included a random intercept on the supermarket level. Males were used as reference category in the analyses of sex, and the combination of low and medium educational attainment was used as reference category for the analyses of educational attainment. Participants with low and medium education were combined in one reference category due to very low participant numbers in the low educated group (< 10 participants for five out of the ten investigated outcomes). Moreover, we provided insight into the feasibility and costs of at-home measurement of cardiometabolic risk by reporting the related costs in terms of travel costs and staff hours for home visiting. Travel costs (Euros) and staff hours were calculated as the sum of all costs and staff hours related to the home visits. Furthermore, descriptive statistics are presented for the number and percentage (n(%)) of participants who (a) had a failed first attempt at the blood measurement and were invited for a second attempt, (b) ultimately did not complete the HbA1 and LDL measurement after a maximum of two attempts and (c) indicated to drop-out of the study due to experienced difficulties with the at-home measurements. Last, using the same analytical approach as for exploring sociodemographic differences in recruitment strategies, we explored potential sociodemographic differences in participants who required assistance via a home visit to complete the at-home measurement of cardiometabolic risk; who had a failed first attempt at the blood measurements or who were ultimately non-completers of the at-home measurement of HbA1c, LDL-cholesterol and/or waist circumference. Successful completion of cardiometabolic measurements was defined as completing a valid measurement result, after one or multiple attempts.
We presented regression coefficients and $95\%$ confidence intervals to demonstrate the variation in recruitment effectiveness and at-home measurements across sociodemographic subgroups. We do not describe whether associations were statistically significant, because of the low numbers in some of the analysed subgroups. All analyses were conducted in R (version 4.0.3) using the packages table1 [44] and lme4 [45].
## Recruitment costs, yield and differential effectiveness per strategy
Figure 1 presents the participant flowchart up to baseline measurement completion of the primary trial outcome. In total, 783 individuals registered for participation, 602 ($77\%$) were eligible and 421 ($54\%$) were included as study participants after completing informed consent. The majority of participants who completed the eligibility assessment indicated to have been recruited via a single strategy ($83\%$). Additionally, $16\%$ indicated to have been recruited via two methods and $1\%$ reported to have been exposed to three or more recruitment strategies. Most of the included participants ($$n = 421$$) were recruited via letters/flyers received at home ($$n = 314$$; $75\%$) or flyers received at the supermarket ($$n = 52$$; $12\%$) (Table 1). Less frequent reported strategies were in-store recruitment by a researcher ($$n = 32$$; $8\%$), news items ($$n = 31$$; $7\%$), word-of-mouth ($$n = 24$$; $6\%$) and social media ($$n = 16$$; $4\%$). Of paid strategies, flyers in the supermarket were the cheapest per participant included through this method (12 Euros) and the least time-invasive for research staff (< 1 h), whilst the most expensive strategies were the social media campaigns (194 Euros) and mailing of letters/flyers (89 Euros). In-store recruitment was high in terms of staff hours per included participant (9 h) as well as costs (50 Euros) (Table 1). A detailed breakdown of the costs per recruitment strategy is provided in Supplementary Table 2 (Additional file 1).Fig. 1Participant flowchart of the Supreme Nudge trial Table 1Overview of recruitment investments and yields in the Supreme Nudge trialCompleted eligibly assessment ($$n = 664$$)Eligible ($$n = 602$$)Included ($$n = 421$$)Baseline measurement completion of primary trial outcome ($$n = 391$$)Total material cost (€)Total time (hours)n(%)Material cost per participant (€)Time per participant (hours)n(%)Material cost per participant (€)Time per participant (hours)n(%)Material cost per participant (€)Time per participant (hours)n(%)Material cost per participant (€)Time per participant (hours)Media news article01049(7.4)0.00.244(7.3)0.00.231(7.4)0.00.331(7.9)0.00.3Flyers in the supermarket6402279(11.9)8.10.371(11.8)9.00.352(12.4)12.30.452(13.3)12.30.4Posters2902218(2.7)16.11.215(2.5)19.31.511(2.6)26.42.010(2.6)29.02.2Mailing of flyers and recruitment letters27,9811120477(71.8)58.72.4433(71.9)64.62.6314(74.6)89.13.6288(73.7)96.53.9Social media advertisements31001027(4.1)114.80.424(4.0)129.20.416(3.8)193.80.615(3.8)206.70.7Email to supermarket customer panel0216(2.4)0.00.114(2.3)0.00.111(2.6)0.00.211(2.8)0.00.2Advertisement on website of the Dutch Heart Foundation0217(2.6)0.00.116(2.7)0.00.112(2.9)0.00.211(2.8)0.00.2Word of mouth0834(5.1)0.00.232(5.3)0.00.324(5.7)0.00.324(6.1)0.00.3In-store recruitment158327668(10.2)23.34.167(11.1)23.64.132(7.6)49.58.630(7.7)52.89.2Total33,5941472
## Baseline characteristics of the included study sample
We enrolled 421 participants out of which 18 dropped out of the study before completing any baseline measurements and two never performed baseline measurements (Fig. 1). As such, baseline data were collected from 401 study participants, of which ten did not provide data for the primary study outcome and were therefore excluded, leaving 391 participants in the baseline sample. A total of 326 participants were randomised in the physical activity app intervention. Of those participants, 64 were excluded before the baseline period because they [1] never installed the step counter app, despite having received three reminders ($$n = 36$$; $56\%$), [2] used a smartphone that was incompatible with the step counter app ($$n = 20$$; $31\%$) or [3] were unwilling to install the step counter app ($$n = 8$$; $13\%$).
Trial participants who completed the measurement of the primary trial outcome ($$n = 391$$) had a mean age of 57.6 (SD 11.0) years, $71\%$ ($$n = 282$$) were female and $41\%$ ($$n = 162$$) were highly educated (Table 2). The population characteristics were approximately equally distributed across supermarket clusters, although some clusters included relatively more participants with high educational attainment than others (Additional file 1: Supplementary Table 3). On average, participants scored 104.7 (SD18.6) points on the DHD15-index (Table 2).Table 2Participant characteristics of the Supreme Nudge trial and baseline measures of outcomes, for the total sample ($$n = 391$$)Total sample ($$n = 391$$)Population characteristicsAge, years, mean (SD)57.7(11.0)Femalea, n (%)282(72.1)Educational attainment, n (%) Low96(24.6) Medium133(34.0) High162(41.4)Household size, n adultsa (median [IQR])2[1]Household size, n childrena (median [IQR])0[1]Smoking statusa, n (%) Current smoker22(5.6) Irregular smoker10(2.5) Former smoker184(46.8) Never smoked173(44.0)Prevalent type 2 diabetesa, n (%)24(6.1)Medication for type 2 diabetesa, n (%)23(5.9)Prevalent hypertensiona, n (%)60(15.3)Medication for hypertensiona, n (%)77(19.6)Prevalent hyperlipidaemiaa, n (%)55(14.0)Medication for hyperlipidaemiaa, n (%)55(14.0)Prevalent cardiovascular diseasea, n (%)46(11.7)Primary outcome Total dietary guideline adherence, scored 0 (low adherence) to 150 (high adherence), mean (SD)104.7(18.6)Secondary outcomesDietary guideline adherence on sub-components, scored 0 (low adherence) to 10 (high adherence), mean (SD) or median [IQR]: Vegetables6.1(3.2) Fruits6.7(3.4) Whole grains7.3(2.8) Legumes8.7[10.0] Nuts5.8[8.1] Dairy5.8[6.5] Fish6.2(3.5) Cooking fats and butter10.0[9.7] Tea5.0[9.8] Coffee7.8(2.7) Red meat8.9(2.6) Processed meat5.5[6.3] Sugar sweetened beverages8.0(3.2) Alcoholic beverages8.1(3.3) Salt8.3(2.1)Cardiometabolic measures, mean (SD): HbA1cb, mmol/mol37.4(7.3) LDL-cholesterolc, mmol/L3.0(1.0) HDL-cholesterol d, mmol/L1.6(0.5) Total cholesterole, mmol/L5.4(1.1) Total cholesterol/HDL-ratioe3.7(1.2) Triglyceridesf, mmol/L1.8(1.0)Waist circumferenceg, cm, mean (SD): Females94.0(13.7) Males103.0(10.9)Healthy food purchasesh Total percentage healthy purchases46.8(22.3)Customer satisfactioni, scored 1 (low) to 7 (high), mean (SD): Total customer satisfaction5.6(1.1) Environment and atmosphere5.6(1.2) Layout and routing5.5(1.1) Supermarket tidiness5.8(1.1) Assortment of food products5.2(1.2) General product prices5.0(1.2) Discount prices5.6(1.0) Fruit and vegetable prices5.0(1.3) Bread prices5.2(1.2)Food decision style for vegetablesj, scored 1 (low) to 7 (high), mean (SD): Reflective5.2(1.1) Habitual5.0(0.9) Impulsive3.5(1.2)Food decision style for snacksk, scored 1 (low) to 7 (high), mean (SD): Reflective4.0(1.4) Habitual3.1(1.4) Impulsive3.6(1.6)Nudges and social cognitive factorsi, scored 1 (low) to 7 (high), mean (SD): Health goals6.3(0.9) Healthy shopping6.0(1.0) Perceived social norm4.7(1.0) Attractiveness healthy foods5.9(1.1)Low educational attainment: no education and primary education; medium educational attainment: secondary educational attainments; high educational attainment: tertiary educational attainments.an = 2 missing values; bn = 22 missing values; cn = 49 missing values; dn = 36 missing values; en = 48 missing values; fn = 43 missing values; gn = 6 missing values; hn = 161 missing values; in = 3 missing values; jn = 30 missing values; kn = 139 missing values Table 3 presents the baseline and demographic characteristics of the participants in the physical activity app intervention. A total of 137 participants were randomly assigned to the control group and 125 participants to the intervention group. All these participants completed the baseline measurement of step counts. Participants (mean age 56.6 years (SD 11.0); $$n = 195$$ ($74\%$) females; $$n = 113$$ ($43\%$) high educated) mostly used Android smartphones ($$n = 184$$/262, $70\%$) and walked a median of 2710 (IQR 3696) steps per day during the 7-day baseline period. Daily step counts at baseline were equally balanced between the control and intervention groups. Table 3Participant characteristics and baseline measurement of step counts for Supreme Nudge trial participants additionally included in the physical activity coaching app intervention, for the total sample ($$n = 262$$)Total sample physical activity intervention ($$n = 262$$)Age, years (mean (SD))56.5(11.0)Female (n (%))195(74.4)Educational attainment (n (%)) Low50(19.1) Medium99(37.8) High113(43.1)Smartphone type: Android, n (%)184(70.2) iOS, n (%)78(29.8)Daily step count, median [IQR]: Averaged over 7-day baseline period2710[3696]Walking behaviour and social cognitive factorsa, scored 1–7, mean (SD) Consequences of behaviour3.1(2.0) Social comparison2.6(1.7) Action planning4.0(1.9) Self-monitoring3.3(1.9) Social support2.5(1.7) Goal setting3.4(1.9) Barrier identification2.2(1.6) Self-evaluation3.7(1.9) Encouragement2.5(1.8) Others’ approval1.7(1.3)eHealth literacya, scored 1–7, mean (SD)4.9(1.5)Low educational attainment: no education and primary education; medium educational attainment: secondary educational attainments; high educational attainment: tertiary educational attainments.an = 19 missing values
## Sociodemographic differences in effectiveness of recruitment strategies
Flyers in the supermarket (β 2.56 years ($95\%$ CI − 0.63; 5.85)) and media news articles (β 2.33 years ($95\%$ CI − 1.69; 6.44)) recruited somewhat older participants and relatively more females (flyers: ORfemales 1.91 ($95\%$ CI 0.93; 4.31) and news articles: ORfemales 1.64 ($95\%$ CI: 0.69; 4.51)). Word-of-mouth tended to recruit relatively more males (ORfemales 0.51 ($95\%$ CI 0.22; 1.21)). Those with low and medium educational attainment tended to be more often recruited via flyers in the supermarket (ORhigh education 0.72 ($95\%$ CI 0.38; 1.33)) and news articles (ORhigh education 0.67 ($95\%$ CI 0.29; 1.45)), whilst those with high educational attainment via in-store recruitment (ORhigh education 1.75 ($95\%$ CI 0.81; 3.81)) (Table 4). Descriptive statistics and the absolute numbers analysed per group are presented in Supplementary Table 4 (Additional file 1).Table 4Associations between the top five most successful recruitment strategies and the sociodemographic variables age, sex (female) and educational attainment (high education) ($$n = 391$$)Agea, yearsSexEducational attainmentβ($95\%$ CI)ORfemales($95\%$ CI)ORhigh education($95\%$ CI)*Recruited via* mailing of recruitment flyers and letters ($$n = 288$$)− 1.76(− 4.29; 0.70)1.00(0.60; 1.64)1.22(0.76; 1.98)*Recruited via* flyers in the supermarket ($$n = 52$$)2.56(− 0.63; 5.85)1.91(0.93; 4.31)0.72(0.38; 1.33)*Recruited via* media news article ($$n = 31$$)2.33(− 1.69; 6.44)1.64(0.69; 4.51)0.67(0.29; 1.45)*Recruited via* in-store recruitment ($$n = 30$$)− 0.11(− 4.19; 4.06)1.21(0.52; 3.14)1.75(0.81; 3.81)*Recruited via* word-of-mouth ($$n = 24$$)− 0.40(− 4.94; 4.14)0.51(0.22; 1.21)1.01(0.42; 2.35)Statistical significant outcomes ($p \leq 0.05$) are displayed in bold text. Associations are based on linear multilevel models (age) and logistic multilevel models (sex and education), including a random intercept on the supermarket level. Males are used as reference category in the analyses of sex, and the combination of low and medium educational attainment was used as reference category for the analyses of educational attainment. OR odds ratio, CI confidence interval; an = 2 missing values
## Feasibility and sociodemographic differences in at-home measurement of cardiometabolic risk
At-home measurements were often completed successfully, as for lipid profile $88\%$ ($$n = 342$$) of participants completed the measurement, $94\%$ ($$n = 369$$) for HbA1c and $99\%$ ($$n = 385$$) for waist circumference (Table 1). Researcher assistance with the at-home measurements was requested by $17\%$ ($$n = 68$$) of the total sample. On average, the time spent per participant by research staff was almost 2 h (including preparation time and travel time), and the accompanying travel costs via a shared rental car were 24 Euros, resulting in 129 staff hours and 1672 Euros travel costs in total during the baseline measurements attributable to conducting home visits. A first attempt at the blood measurements failed for 86 participants ($22\%$), who were invited for a second attempt. Ultimately, the HbA1c measurement was not completed by 20 participants ($5\%$) and the LDL measurement by 47 participants ($12\%$). Three participants (< $1\%$) dropped out of the study due to persisting difficulties with the at-home measurements.
Those with a failed first attempt at the blood measurement were relatively more often lower educated (ORhigh education 0.64 ($95\%$ CI 0.38; 1.06)) and relatively older (β 3.89 years ($95\%$ CI 1.28; 6.49), whilst the non-completers of the HbA1c (β − 8.92 years ($95\%$ CI − 13.62; − 4.28)) and LDL (β − 3.19 years ($95\%$ CI − 6.53; 0.09)) were relatively younger. Among non-completers of the waist circumference measurement were relatively more males than females (ORfemales 0.18 ($95\%$ CI 0.03; 0.95) (Table 5). Descriptive statistics and the absolute numbers analysed per group are presented in Supplementary Table 5 (Additional file 1).Table 5Associations between elements of at-home measurement of cardiometabolic risk markers and the sociodemographic variables age, sex (female) and educational attainment (high education) ($$n = 391$$)Agea, yearsSexEducational attainmentβ($95\%$ CI)ORfemales($95\%$ CI)ORhigh education($95\%$ CI)Participants requesting a home-visit ($$n = 68$$)0.96(− 1.93; 3.84)0.97(0.56; 1.78)1.11(0.64; 1.92)Failed first attempt of blood measurement ($$n = 86$$)3.89(1.28; 6.49)1.13(0.67; 1.99)0.64(0.38; 1.06)Non-completers HbA1c measurement ($$n = 22$$)− 8.92(− 13.62; − 4.28)1.15(0.43; 3.60)1.46(0.59; 3.58)Non-completers LDL-cholesterol measurement ($$n = 49$$)− 3.19(− 6.53; 0.09)0.79(0.41; 1.55)0.91(0.48; 1.71)Non-completers waist circumference measurement ($$n = 8$$)− 0.61(− 8.38; 7.04)0.18(0.03; 0.95)1.45(0.33; 6.37)Statistical significant outcomes ($p \leq 0.05$) are displayed in bold text. Associations are based on linear multilevel models (age) and logistic multilevel models (sex and education), including a random intercept on the supermarket level. Males are used as reference category in the analyses of sex, and the combination of low and medium educational attainment was used as reference category for the analyses of educational attainment. OR odds ratio, CI confidence interval; an = 2 missing values
## Discussion
This paper reports on costs, feasibility and results of used recruitment strategies, baseline characteristics of the study sample, potential sociodemographic differences in successful recruitment strategies and feasibility of cardiometabolic measurements at home. The results indicate that letters/flyers sent to home addresses and distributed in the supermarkets recruited most participants. Supermarket flyers were the cheapest and least time-invasive of all paid recruitment strategies. The social media campaign was the most expensive, and mailings of recruitment letters and in-store-recruitment were the most time-invasive and therefore costly in terms of recruited staff hours. Flyers in the supermarket and media news articles seemed to recruit relatively more females and word-of-mouth seemed to recruit relatively more males. At-home measurement of cardiometabolic markers was feasible, as they were successfully completed by most participants. Older participants had more often a failed first attempt of blood measurement, whilst younger participants were more often non-completers of the blood measurements.
The success of mailing letters and flyers to home addresses and the distribution of flyers in the supermarket is likely due to their wide reach and the fact that these strategies were used early in the recruitment process. The fact that our social media campaign, which was used later, appeared very costly in terms of recruitment yield is in contrast to what other studies showed. In studies where social media were used as the primary recruitment method, it appeared (cost)-effective (e.g. ~6–14 Euros per included participant [46, 47]). It is possible that many of our study participants had already registered early in the recruitment process before exposure to the social media campaign. The social media campaign may have persuaded those who were initially doubtful or forgot about the study after having first received a letter or flyer. The same could apply to in-store recruitment, as many shoppers had mentioned that they had seen a flyer but had not acted upon it. Although it is to be expected that participants registered after having been exposed to multiple recruitment strategies, only a few participants indicated to have been recruited via multiple strategies. This could be the result of underreporting due to recall bias or unconscious processing of exposure to recruitment strategies [48, 49].
The included study sample was comparable with the general Dutch population in terms of educational attainment, but females were overrepresented and participants scored relatively high on the DHD15-index. Overrepresentation of females in epidemiological studies is a well-known phenomenon [50]. This may be due to higher levels of health consciousness among females, a relatively higher interest in food-related topics, or having primary responsibility for grocery shopping within a household. Regarding the DHD15-index score, included participants scored relatively high with 105 points on average compared to, for example, a cross-sectional Dutch population sample in the Eet & Leef study (average of 96 points) [51]. This suggests that our study sample may be more health-conscious than the general Dutch population. Furthermore, our sample underrepresented smokers (~$5\%$ versus ~$19\%$ nationally) [52] but was representative in terms of the prevalence of hypertension [53].
Our results suggested that passive recruitment strategies yielded more females whilst active strategies more males, which is also seen in other studies [20, 46]. Specifically, our supermarket flyers and the media news articles seemed to have recruited relatively more females and also relatively older participants. This may partly be explained by a larger responsibility for grocery shopping or a higher frequency of supermarket visits, leading to larger exposure to the supermarket flyers. Also, that the active strategy word-of-mouth seemed to recruit more males is likely the result of encouraging included participants (mainly female) to ask their partner if they are interested to participate. Furthermore, our findings suggested that passive recruitment strategies recruited relatively more participants with low and medium educational attainment compared to active in-store recruitment, via which relatively more participants with high educational attainment were recruited. This is in contrast to another study which indicated that passive strategies often recruit relatively more individuals with high educational attainment compared to active strategies [20]. It could be that individuals with low and medium educational attainment were relatively more susceptible to the content of the flyers mentioning the free grocery box as an incentive.
The COVID-19 pandemic had a considerable impact on recruitment and data collection procedures. At-home measurement of cardiometabolic makers and self-installation of the coaching app may have been challenging for some participants, since participants were required to understand and be able to follow instructions from a written manual or video. Participants only received support from research staff when actively asking for it, which requires a certain level of self-efficacy. This may have led to some selection bias, resulting in a study sample consisting of participants with relatively high levels of self-efficacy. The at-home measurements may have introduced self-assessment bias and reduced measurement reliability for waist circumference measurement. Previous research has shown that especially those with lower educational attainment had relatively more often an overestimation of their waist circumference measurement, as well as younger versus older participants [54]. Another challenge was that the COVID-19 pandemic itself likely interfered with inclusion and the completion of measurements, as individuals could have been hesitant to register, unable to participate or complete measurements. By investing in sending out multiple at-home measurement packages and conducting home visits, we reached a relatively high completion rate of the blood measurements. Yet, blood measurements were not successfully completed for all participants, and reasons for non-completion of blood measurements were often practical (e.g. insufficient blood flow). Logistically, at-home cardiometabolic measurements were to some extent efficient, since this saved staff hours and costs due to limited commute and work at study locations. On the other hand, conducting home visits to aid with the cardiometabolic measurements was also costly and time-invasive.
We recruited participants in socially disadvantaged neighbourhoods via various passive and active recruitment approaches with the aim to recruit a relative high proportion of those with a lower and medium educational attainment [26]. Yet, despite these efforts, our sample consisted of $41\%$ highly educated participants, and insights from this paper demonstrate that recruitment is challenging and costly. Another important insight was that the use of flyers/letters sent to home addresses can be an effective recruitment strategy, although it is very costly. Thus, the use of complementary recruitment strategies is recommended to boost inclusion rates, combined with strategies which could for example be promoting word-of-mouth to reach a specific group. Researchers may consider at-home cardiometabolic measurements to simplify logistics and reach potential participants from widespread geographical locations. Future studies with larger sample sizes could focus on further determining the most effective recruitment strategies to reach individuals with a low SEP and could investigate the importance of the order in which recruitment strategies are used on their (cost)-effectiveness. Since recruitment is time invasive and costly, it is recommended that future projects consider comprehensive recruitment plans and allocate sufficient time and financial resources for the recruitment process.
## Strengths and limitations
Strengths of this study are the utilisation of a wide range of recruitment strategies and the monitoring of used strategies and recruitment yields to explore their effects. Nevertheless, determination of the most decisive strategy for participants to register was not possible since participants were not asked for the most decisive method and had the option to register multiple strategies instead of just one. Also, the relatively small sample size when exploring sociodemographic differences in recruitment strategy effectiveness and cardiometabolic measurements likely led to wide confidence intervals. For the interpretation of the results, we therefore focused on interpretation of effect sizes rather than statistical significance. Moreover, this study did not investigate the influence of the order in which recruitment strategies were used and time intervals between recruitment strategies on their (cost)-effectiveness. Furthermore, since this study was performed during the COVID-19 pandemic, this may have affected recruitment yield, but we were unable to quantify this impact. Last, we lack sufficient insight into reasons why 181 eligible participants did not proceed with study enrolment, since most of them did not respond after having received the information letter. Some replied that participation appeared more comprehensive (e.g. with at-home blood measurements) than anticipated which may have been a barrier for others as well.
## Conclusion
In the Supreme Nudge trial, supermarket flyers were the most cost-effective in recruiting participants of all paid strategies. Mailings to home addresses recruited the most participants but were very costly. Conducting at-home cardiometabolic measurements was feasible with relatively high completion rates. These insights are valuable for future research projects that aim to recruit a diverse study population and to study cardiometabolic risk markers in, for example, geographically large and widespread groups or when face to face contact is not possible.
## Supplementary Information
Additional file 1: Supplementary Table 1. Recruitment strategy utilisation, per month in 2021. Supplementary Table 2. Detailed breakdown of material costs and of time per recruitment strategy. Supplementary Table 3. Population characteristics of the Supreme Nudge trial ($$n = 391$$) presented by supermarket clusters ($$n = 12$$). Supplementary Table 4. Absolute numbers analyse per group for the top five most successful recruitment strategies and the sociodemographic variables age, sex and educational attainment ($$n = 391$$). Supplementary Table 5. Absolute numbers analysed for elements of at-home measurement of cardiometabolic risk markers and the sociodemographic variables age, sex and educational attainment ($$n = 391$$).
## References
1. **Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017**. *Lancet.* (2018.0) **392** 1923-1994. DOI: 10.1016/S0140-6736(18)32225-6
2. Roth GA, Mensah GA, Fuster V. **The global burden of cardiovascular diseases and risks a compass for global action**. *J Am Coll Cardiol* (2020.0) **76** 2980-2981. DOI: 10.1016/j.jacc.2020.11.021
3. Shavers VL. **Measurement of socioeconomic status in health disparities research**. *J Natl Med Assoc* (2007.0) **99** 1013-1023. PMID: 17913111
4. van Rossum CTM, Buurma-Rethans EJM, Dinnissen CS, Beukers MH, Brants HAM, Dekkers ALM. *The diet of the Dutch: results of the Dutch National Food Consumption Survey 2012-2016* (2020.0)
5. Marijn Stok F, Renner B, Allan J, Boeing H, Ensenauer R, Issanchou S. **Dietary behavior: an interdisciplinary conceptual analysis and taxonomy**. *Front Psychol* (2018.0) **9** 1689. DOI: 10.3389/fpsyg.2018.01689
6. Anderson ES, Winett RA, Wojcik JR. **Self-regulation, self-efficacy, outcome expectations, and social support: social cognitive theory and nutrition behavior**. *Ann Behav Med* (2007.0) **34** 304-312. DOI: 10.1007/BF02874555
7. Friel S, Hattersley L, Ford L, O’Rourke K. **Addressing inequities in healthy eating**. *Health Promot Int* (2015.0) **30** 77-88. DOI: 10.1093/heapro/dav073
8. Hermstad A, Honeycutt S, Flemming SS, Carvalho ML, Hodge T, Escoffery C. **Social environmental correlates of health behaviors in a faith-based policy and environmental change intervention**. *Health Educ Behav* (2018.0) **45** 672-681. DOI: 10.1177/1090198118757826
9. 9.Adams J, Mytton O, White M, Monsivais P. Why are some population interventions for diet and obesity more equitable and effective than others? The Role of Individual Agency. PLoS Med. 2016;13(4):1–7.
10. 10.Dollman J. Social and environmental influences on physical activity behaviours. Int J Environ Res Public Health. 2018;15(1):1–3.
11. Lakerveld J, Mackenbach J. **The upstream determinants of adult obesity**. *Obes Facts* (2017.0) **10** 216-222. DOI: 10.1159/000471489
12. Swinburn B, Egger G, Raza F. **Dissecting obesogenic environments: the development and application of a framework for identifying and prioritizing environmental interventions for obesity**. *Prev Med* (1999.0) **29** 563-570. DOI: 10.1006/pmed.1999.0585
13. 13.Gillies K, Kearney A, Keenan C, Treweek S, Hudson J, Brueton VC, et al. Strategies to improve retention in randomised trials. Cochrane Database Syst Rev. 2021;3:1–2.
14. Kang H. **The prevention and handling of the missing data**. *Korean J Anesthesiol* (2013.0) **64** 402-406. DOI: 10.4097/kjae.2013.64.5.402
15. Burroughs AR, Visscher WA, Haney TL, Efland JR, Barefoot JC, Williams RB. **Community recruitment process by race, gender, and SES gradient: lessons learned from the community health and stress evaluation (CHASE) study experience**. *J Community Health* (2003.0) **28** 421-437. DOI: 10.1023/A:1026029723762
16. Whatnall MC, Hutchesson MJ, Sharkey T, Haslam RL, Bezzina A, Collins CE. **Recruiting and retaining young adults: what can we learn from behavioural interventions targeting nutrition, physical activity and/or obesity? A systematic review of the literature**. *Public Health Nutr* (2021.0) **24** 5686-5703. DOI: 10.1017/S1368980021001129
17. 17.Maher CA, Davis CR, Curtis RG, Short CE, Murphy KJ. A physical activity and diet program delivered by artificially intelligent virtual health coach: proof-of-concept study. JMIR Mhealth Uhealth. 2020;8(7):1–12.
18. Lee RE, McGinnis KA, Sallis JF, Castro CM, Chen AH, Hickmann SA. **Active vs. passive methods of recruiting ethnic minority women to a health promotion program**. *Ann Behav Med* (1997.0) **19** 378-384. DOI: 10.1007/BF02895157
19. Smit E, Leenaars K, Wagemakers A, van der Velden K, Molleman G. **How to recruit inactive residents for lifestyle interventions: participants’ characteristics based on various recruitment strategies**. *Health Promot Int* (2021.0) **36** 989-999. DOI: 10.1093/heapro/daaa134
20. 20.Estabrooks P, You W, Hedrick V, Reinholt M, Dohm E, Zoellner J. A pragmatic examination of active and passive recruitment methods to improve the reach of community lifestyle programs: The Talking Health Trial. Int J Behav Nutr Phy. 2017;14:1–10.
21. 21.Stuber JM, Middel CNH, Mackenbach JD, Beulens JWJ, Lakerveld J. Successfully recruiting adults with a low socioeconomic position into community-based lifestyle programs: a qualitative study on expert opinions. Int J Environ Res Public Health. 2020;17(8):1–15.
22. Turrell G, Patterson C, Oldenburg B, Gould T, Roy MA. **The socio-economic patterning of survey participation and non-response error in a multilevel study of food purchasing behaviour: area- and individual-level characteristics**. *Public Health Nutr* (2003.0) **6** 181-189. DOI: 10.1079/PHN2002415
23. Pescud M, Pettigrew S, Wood L, Henley N. **Insights and recommendations for recruitment and retention of low socio-economic parents with overweight children**. *Int J Soc Res Methodol* (2015.0) **18** 617-633. DOI: 10.1080/13645579.2014.931201
24. Chinn DJ, White M, Howel D, Harland JOE, Drinkwater CK. **Factors associated with non-participation in a physical activity promotion trial**. *Public Health* (2006.0) **120** 309-319. DOI: 10.1016/j.puhe.2005.11.003
25. Coday M, Boutin-Foster C, Sher TG, Tennant J, Greaney ML, Saunders SD. **Strategies for retaining study participants in behavioral intervention trials: Retention experiences of the NIH behavior change consortium**. *Ann Behav Med* (2005.0) **29** 55-65. DOI: 10.1207/s15324796abm2902s_9
26. 26.Stuber JM, Mackenbach JD, de Boer FE, de Bruijn GJ, Gillebaart M, Harbers MC, et al. Reducing cardiometabolic risk in adults with a low socioeconomic position: protocol of the Supreme Nudge parallel cluster-randomised controlled supermarket trial. Nutr J. 2020;19(1):1–19.
27. Lakerveld J, Mackenbach JD, de Boer F, Brandhorst B, Broerse JEW, de Bruijn GJ. **Improving cardiometabolic health through nudging dietary behaviours and physical activity in low SES adults: design of the Supreme Nudge project**. *BMC Public Health* (2018.0) **18** 899. DOI: 10.1186/s12889-018-5839-1
28. 28.SCPStatusscores 20172019Den HaagSociaal en Cultureel Planbureau. *Statusscores 2017* (2019.0)
29. van Lee L, Feskens EJ, Meijboom S, Hooft van Huysduynen EJ, van’t Veer P, de Vries JH. **Evaluation of a screener to assess diet quality in the Netherlands**. *Br J Nutr* (2016.0) **115** 517-526. DOI: 10.1017/S0007114515004705
30. 30.Vos AL, de Bruijn GJ, Klein MCA, Lakerveld J, Boerman SC, Smit EG. SNapp, a tailored smartphone app intervention to promote walking in adults of low socioeconomic position: development and qualitative pilot study [not peer-reviewed or edited preprint]. JMIR Format Res Open Peer Rev Period. 2022. https://preprints.jmir.org/preprint/40851.
31. Scott SG, Bruce RA. **Decision-making style - the development and assessment of a new measure**. *Educ Psychol Meas* (1995.0) **55** 818-831. DOI: 10.1177/0013164495055005017
32. Verplanken B, Orbell S. **Reflections on past behavior: a self-report index of habit strength**. *J Appl Soc Psychol* (2003.0) **33** 1313-1330. DOI: 10.1111/j.1559-1816.2003.tb01951.x
33. Ersche KD, Lim TV, Ward LHE, Robbins TW, Stochl J. **Creature of Habit: a self-report measure of habitual routines and automatic tendencies in everyday life**. *Personal Individ Differ* (2017.0) **116** 73-85. DOI: 10.1016/j.paid.2017.04.024
34. Verplanken B, Herabadi A. **Individual differences in impulse buying tendency: feeling and no thinking**. *Eur J Personal* (2001.0) **15** S71-S83. DOI: 10.1002/per.423
35. Spinella M. **Normative data and a short form of the Barratt Impulsiveness Scale**. *Int J Neurosci* (2007.0) **117** 359-368. DOI: 10.1080/00207450600588881
36. de Ridder D, Gillebaart M. **What’s in a nudge**. *Tijdschrift voor gezondheidswetenschappen* (2016.0) **94** 261-265
37. Forwood SE, Ahern AL, Hollands GJ, Ng YL, Marteau TM. **Priming healthy eating. You can’t prime all the people all of the time**. *Appetite.* (2015.0) **89** 93-102. DOI: 10.1016/j.appet.2015.01.018
38. GiskeS K, Van Lenthe FJ, Brug J, Mackenbach JP, Turrell G. **Socioeconomic inequalities in food purchasing: the contribution of respondent-perceived and actual (objectively measured) price and availability of foods**. *Prev Med* (2007.0) **45** 41-48. DOI: 10.1016/j.ypmed.2007.04.007
39. van Ansem WJC, Schrijvers CTM, Rodenburg G, van de Mheen D. **Is there an association between the home food environment, the local food shopping environment and children’s fruit and vegetable intake? Results from the Dutch INPACT study**. *Public Health Nutr* (2013.0) **16** 1206-1214. DOI: 10.1017/S1368980012003461
40. Higgs S, Liu J, Collins EIM, Thomas JM. **Using social norms to encourage healthier eating**. *Nutr Bull* (2019.0) **44** 43-52. DOI: 10.1111/nbu.12371
41. Dijkstra SC, Neter JE, van Stralen MM, Knol DL, Brouwer IA, Huisman M. **The role of perceived barriers in explaining socio-economic status differences in adherence to the fruit, vegetable and fish guidelines in older adults: a mediation study**. *Public Health Nutr* (2015.0) **18** 797-808. DOI: 10.1017/S1368980014001487
42. Abraham C, Michie S. **A taxonomy of behavior change techniques used in interventions**. *Health Psychol* (2008.0) **27** 379-387. DOI: 10.1037/0278-6133.27.3.379
43. Michie S, Ashford S, Sniehotta FF, Dombrowski SU, Bishop A, French DP. **A refined taxonomy of behaviour change techniques to help people change their physical activity and healthy eating behaviours: the CALO-RE taxonomy**. *Psychol Health* (2011.0) **26** 1479-1498. DOI: 10.1080/08870446.2010.540664
44. Rich B. *Using the table1 package to create HTML tables of descriptive statistics: the comprehensive R archive network* (2021.0)
45. 45.Bates D, Maechler M, Bolker B, Walker S, Christensen RHB, Singmann H, et al. Package ‘lme4’: linear mixed-effects models using ‘Eigen’ and S4 (Version 1.1-31): The Comprehensive R Archive Network; 2022 [Available from: https://cran.r-project.org/web/packages/lme4/lme4.pdf].
46. Hoenink JC, Mackenbach JD, van der Laan LN, Lakerveld J, Waterlander W, Beulens JWJ. **Recruitment of participants for a 3D virtual supermarket: cross-sectional observational study**. *JMIR Form Res* (2021.0) **5** e19234. DOI: 10.2196/19234
47. 47.Whitaker C, Stevelink S, Fear N. The use of facebook in recruiting participants for health research purposes: a systematic review. J Med Internet Res. 2017;19(8):1–11.
48. Yoo CY. **Unconscious processing of Web advertising: effects on implicit memory, attitude toward the brand, and consideration set**. *J Interact Mark* (2008.0) **22** 2-18. DOI: 10.1002/dir.20110
49. Althubaiti A. **Information bias in health research: definition, pitfalls, and adjustment methods**. *J Multidiscip Healthc* (2016.0) **9** 211-217. DOI: 10.2147/JMDH.S104807
50. Galea S, Tracy M. **Participation rates in epidemiologic studies**. *Ann Epidemiol* (2007.0) **17** 643-653. DOI: 10.1016/j.annepidem.2007.03.013
51. Hoenink JC, Waterlander W, Beulens JWJ, Mackenbach JD. **The role of material and psychosocial resources in explaining socioeconomic inequalities in diet: a structural equation modelling approach**. *SSM Popul Health* (2022.0) **17** 101025. DOI: 10.1016/j.ssmph.2022.101025
52. 52.CBS. StatLine: Gezondheid, leefstijl, zorggebruik en -aanbod, doodsoorzaken; kerncijfers 2021 [updated 22 december 2021. Available from: https://opendata.cbs.nl/#/CBS/nl/dataset/81628NED/table?ts=1650030421018.
53. 53.CBSStatline: Gezondheid en zorggebruik; persoonskenmerken2022Den HaagCentral Bureau Statistics Netherlands. *Statline: Gezondheid en zorggebruik; persoonskenmerken* (2022.0)
54. 54.Ayala AMC, Nijpels G, Lakerveld J. Validity of self-measured waist circumference in adults at risk of type 2 diabetes and cardiovascular disease. BMC Med. 2014;12.
|
---
title: Association of Hematological and Biochemical Parameters with Serological Markers
of Acute Dengue Infection during the 2022 Dengue Outbreak in Nepal
authors:
- Bibek Raj Bhattarai
- Abhishek Mishra
- Suraj Aryal
- Mandira Chhusyabaga
- Rajshree Bhujel
journal: Journal of Tropical Medicine
year: 2023
pmcid: PMC9981284
doi: 10.1155/2023/2904422
license: CC BY 4.0
---
# Association of Hematological and Biochemical Parameters with Serological Markers of Acute Dengue Infection during the 2022 Dengue Outbreak in Nepal
## Abstract
### Background
Nepal faced a major dengue outbreak in 2022. The majority of hospitals and laboratories had limited resources for dengue confirmation and had to rely on rapid dengue diagnostic tests. The purpose of the study is to find the predictive hematological and biochemical parameters in each serological phase of dengue infection (NS1 and IgM) that may assist in dengue diagnosis, severity assessment, and patient management via the use of rapid serological tests.
### Method
A laboratory-based cross-sectional study was conducted among dengue patients. Rapid antigen (NS1) and serological test (IgM/IgG) was performed to diagnose positive dengue cases. Furthermore, hematological and biochemical investigations were carried out and compared between NS1 and/or IgM-positive participants. A logistic regression analysis was used to identify the validity of the hematological and biochemical characteristics for dengue diagnosis as well as patient management. Receiver-operating characteristic (ROC) curve analysis was used to define the best cut-off, sensitivity, and specificity.
### Result
Multiple logistic regression showed thrombocytopenia (ORA = 1.000; $$p \leq 0.006$$), leukopenia (ORA = 0.999; $p \leq 0.001$), glucose level (ORA = 1.028; $$p \leq 0.029$$), aspartate aminotransferase (ORA = 1.131; $$p \leq 0.001$$), and monocytosis (ORA = 2.332; $$p \leq 0.020$$) as significant parameters in the NS1-only positive group. Similarly, thrombocytopenia (ORA = 1.000; $$p \leq 0.001$$), glucose level (ORA = 1.037; $$p \leq 0.004$$), and aspartate aminotransferase (ORA = 1.141; $p \leq 0.001$) were significant in IgM-only positive patients. Moreover, thrombocytopenia (ORA = 1.000; $p \leq 0.001$), leukopenia (ORA = 0.999; $p \leq 0.001$), glucose (ORA = 1.031; $$p \leq 0.017$$), aspartate aminotransferase (ORA = 1.136; $p \leq 0.001$), and lymphopenia (ORA = 0.520; $$p \leq 0.067$$) were independent predictors in both NS1 + IgM positive groups. Platelets consistently demonstrated a higher area under the curve with increased sensitivity and specificity throughout all models, while aspartate aminotransferase (AUC = 0.811) and glucose (AUC = 0.712) demonstrated better results when single IgM positivity was observed. The total leukocyte count performed better when both NS1 + IgM were positive (AUC = 0.814).
### Conclusion
Hence, thrombocytopenia, elevated AST, high glucose level, leukopenia with monocytosis, and leukopenia with lymphopenia may predict dengue diagnosis and its severity during an active infection. Therefore, these laboratory parameters can be used to complement less sensitive rapid tests, improve dengue diagnosis, and help with proper patient management.
## 1. Introduction
Dengue virus (DENV) is a 50 nm, single-stranded RNA virus with a genome approximately 11 kb in length [1]. The virus contains three structural genes encoding capsid protein (C), membrane protein (M), and envelope protein (E), as well as seven nonstructural (NS) genes encoding NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5 proteins [2]. Dengue virus is transmitted primarily by the vectors Aedes aegypti and *Aedes albopictus* and is most prevalent in tropical and subtropical areas. Dengue infection is usually asymptomatic and self-curable [3]. The World Health Organization (WHO) classified symptomatic dengue as dengue with or without warning signs and severe dengue [4]. The incubation period of the virus ranges from 3 to 10 days, typically 5–7 days, and follows a clinical course as a biphasic febrile phase lasting 2–7 days, a critical phase which lasts 24–48 hours, and a convalescent phase [5]. With an estimated infection of about 400 million people annually, the disease now affects more than 100 countries, most of them in Asia, with a disease burden of $70\%$ [3]. Nepal recently had a dengue outbreak in 2022, and as of December 11, 2022, the total cases of dengue had reached 54,232, with 67 deaths [6].
Various diagnostic methods, such as virus-specific serological tests, molecular detection, and virus isolation, are used for the definitive diagnosis of DENV detection [7]. The three markers most commonly used in serological tests are NS1-Ag and IgM for acute infections and IgG for previous infections. NS1-Ag can be detected from the first 0–9 days of symptoms onset, while IgM is detected 4–5 days after symptoms onset, and production may continue approximately for 3 months or more postonset; IgG levels can be detected throughout the life, starting from 10 to 14 days of postinfection [8, 9].
The gold standard test, like nucleic acid amplification tests, is not readily available in hospitals and clinics in a resource-limited country like Nepal. Thus, many of these facilities rely on lateral flow assays (LFA) or immunochromatography (ICT)-based detection methods for dengue diagnosis. Lateral flow assays are friendly to use and have a rapid turnaround time. Detection of NS1-Ag can be as sensitive as a molecular test during the first 0–7 days of onset of symptoms; however, detection can be compromised in secondary infection due to IgG antibodies from a previous infection [9, 10].
Active dengue detection via ICT, while user-friendly, easy to use, and with rapid turnover time, has low sensitivity and low specificity, as well as higher crossreactivity leading to more false positives. Thus, hematological and biochemical parameters can be beneficial as a supportive test for dengue diagnosis in addition to rapid dengue tests via ICT methods. Furthermore, only a few studies have compared and associated the laboratory parameters with the serological markers of dengue, analyzing just the surface. Therefore, this study attempts to provide an in-depth analysis of different hematological and biochemical tests and associate each parameter with a serological marker (NS1 and/or IgM) of acute dengue infection. Thus, the incorporation of biochemical and hematological parameters may act as a supportive parameter for its diagnosis and would be essential for proper patient management to prevent the life-threatening consequences of dengue.
## 2.1. Inclusion and Exclusion Criteria
After obtaining written informed consent, participants with fever/body pain along with a positive dengue profile test were included in the study. DENV-infected patients were categorized into NS1-only, IgM-only, and dual positive/both NS1 + IgM-positive groups. Study populations with negative dengue profile tests and abnormal hematological and biochemical profiles were excluded from the study. Patients with positive IgG in the dengue profile test were also excluded. Participants showing no symptoms and further tested negative for dengue profiles along with normal hematological and biochemical parameters were taken as a control group.
## 2.2. Specimen Collection and Processing
Following standard operating procedures, venous blood samples were collected. Whole blood was collected in a K3 EDTA vacuum tube and a gel and clot activator tube. A complete blood profile (hemoglobin, RBC and RBC indices, hematocrit, total leukocyte count, differential leukocyte count, and platelets) was performed from blood samples collected in a K3 EDTA tube with a hematology analyzer (Beckman Coulter DxH 520, USA). Similarly, a biochemistry analyzer (Selectra Pro S, ELITech Group, Netherlands) was used to perform biochemical analyses on enzymes (ALP, ALT, AST), bilirubin (total and direct), proteins (total protein and albumin), and nonprotein nitrogenous compounds (urea and creatinine) via a serum sample. Neutrophil:lymphocyte ratio (NLR), lymphocyte:monocyte ratio (LMR), and AST/ALT ratio were calculated based on data.
A serum sample was used to detect dengue infection. Qualitative dengue detection was based on the principle of the rapid chromatographic immunoassay (Dengue NS1 + IgM/IgG Combo Rapid Test, Healgen®). Patients with positive dengue cases were tested for either NS1 or IgM positivity or both NS1 and IgM positivity. Any result that was negative on any one of these profiles was treated as a dengue-negative case. All results were verified by a medical laboratory technologist and a microbiologist.
## 2.3. Statistical Analysis
The data were analyzed using IBM SPSS version 25. Shapiro–Wilk normality test was applied to analyze the data for normal distribution. Categorical variable were described as in numbers and percentage. Continuous variables were shown as the median (Q3−Q1). Univariate analysis was performed appropriately using the Mann–Whitney U test, which was used for overall analysis between dengue positive and dengue negative groups. Likewise, Kruskal Wallis H test, an omnibus test statistic was used to compare > 2 groups. Furthermore, in the case of statistical association, pairwise analysis was performed via Dunn's post hoc test with Bonferroni adjustment, and a p value <0.05 was considered significant. Parameters that were mutually significant in both the univariate analysis and the comparative analysis were included in a univariate logistic regression analysis where a p value <0.25 was considered significant. In the multivariate logistic regression, a few parameters were added despite insignificant results in univariate logistics due to their clinical relevance. Binary logistic regression (in a dichotomous outcome) and multinomial logistic regression (more than 2 outcomes) were performed as required. Results were presented as crude and adjusted odds ratios with a $95\%$ confidence interval ($95\%$ CI). Those variables that yielded the lowest p value <0.05 in multivariate logistics have been considered statistically significant. The covariates, which are common in both binary and multinomial logistic regression, were considered true supportive parameters in dengue diagnosis; thus, they were further analyzed for optimum cut-off via maximizing both sensitivity and specificity using the ROC curve.
## 3.1. Characteristics and Demographics of the Study Population
Out of 348 total study populations, $50.9\%$ ($$n = 177$$) were dengue positive and $49.1\%$ ($$n = 171$$) were dengue negative. The median age of overall participants was 33 years (Q3−Q1 = 50 years −23 years); participants were predominately male, $58.6\%$ ($$n = 204$$) vs. female participants, $41.4\%$ ($$n = 144$$). Among 177 dengue-positive subjects, $23.9\%$ ($$n = 83$$) were NS1-only positive, $13.8\%$ ($$n = 48$$) were IgM-only positive, and $13.2\%$ ($$n = 46$$) were both NS1 + IgM positive. Similarly, $62.7\%$ ($$n = 111$$) men and $37.3\%$ ($$n = 66$$) women were dengue positive. Furthermore, the age group 20–29 years was found to have more positive cases (Supplementary Figure 1).
The Mann–Whitney test revealed that the age in the dengue positive group (median = 37 years) was significantly higher than in the dengue negative group (median = 30 years), $U = 17719.5$, $p \leq 0.006.$ Likewise, the Kruskal-Wallis test also showed the significance of age, $H = 7.949$, $p \leq 0.047.$ On the contrary, gender showed no statistical significance in both Mann–Whitney tests, $U = 13873.5$, $p \leq 0.115$, and the Kruskal-Wallis test, $H = 3.761$, $p \leq 0.288$, between the dengue positive and dengue negative groups.
## 3.2. Association of Dengue Infection with Laboratory Findings
The overall association of laboratory findings between the dengue positive and negative groups is presented in Table 1. Likewise, the comparison of laboratory findings between NS1 only, IgM only, and both NS1 + IgM positive dengue patients was shown as median (Q3−Q1) in Table 2 and the subsequent significance of the Kruskal–Wallis H test and Dunn's post hoc test for pairwise comparison is shown in Table 3. Briefly, in both the Mann–Whitney U test and the Kruskal-Wallis H test, low MCV, high MCHC, decreased platelet count, decrease in TLC, high monocyte count, low LMR, increased glucose level, increased total protein, decreased albumin, increased liver enzymes (AST, ALT, and ALP), and an increased AST/ALT ratio were observed in the dengue positive group.
## 3.3. Logistic Regression and Predictive Markers
A logistic regression analysis was performed, with all the significant variables included in the univariate analysis to adjust for confounders and assess the association between the predictors and outcome. Despite the significance of MCV and MCH, they were not further used for logistic regression due to the high correlation (≥0.7) between each other. Neutrophils and lymphocytes were included in multivariate logistics because of their clinical relevance.
Binary logistic regression was used to assess the association between continuous independent laboratory parameters and dichotomous outcomes (dengue positive and dengue negative). Analysis of the overall model showed statistical significance (χ2 = 363.93, $p \leq 0.05$) and a percentage accuracy classification (PAC) of $93.7\%$. Independent variables, TLC ($$p \leq 0.002$$, OR: 1.000, $95\%$ CI: 0.999–1.000), platelets ($p \leq 0.001$, OR: 1.000, $95\%$ CI: 1.000–1.000), AST ($$p \leq 0.001$$, OR: 1.129, $95\%$ CI: 1.050–1.214), and glucose ($$p \leq 0.008$$, OR: 1.034, $95\%$ CI: 1.009–1.061) were added significantly to the model (Table 4).
Multinomial logistic regression was used to assess the association between continuous independent laboratory parameters and polychotomous outcomes (NS1, NS1 + IgM, IgM, and dengue negative), with dengue negative as the reference group. In all three outcomes (NS1, both NS1 + IgM, and IgM), thrombocytopenia ($$p \leq 0.006$$, $p \leq 0.001$, and $$p \leq 0.001$$), blood glucose level ($$p \leq 0.029$$, $$p \leq 0.017$$, and $$p \leq 0.004$$), and increased AST ($$p \leq 0.001$$, $p \leq 0.001$, and $p \leq 0.001$) remained common significant parameters while leukopenia was statistically significant only in the NS1 ($p \leq 0.001$) and both NS1 + IgM ($p \leq 0.001$) positive groups. In addition, monocytosis ($$p \leq 0.020$$) was significant in NS1 only positive group while lymphopenia ($$p \leq 0.016$$) was significant in both NS1 + IgM positive groups (Tables 5 and 6).
## 3.4. Area under the Receiver Operating Characteristics (AUROC) Analysis
Because platelets, glucose, and AST were common in both binary and multinomial logistic regressions, they were regarded as true supportive laboratory parameters; therefore, each parameter was further analyzed for cut-off, sensitivity, and specificity using a ROC curve. Additionally, TLC was also included due to its clinical relevance (Supplementary Figures 2 and 3).
AUC of platelets, TLC, AST, and glucose along with optimal cut-off values that maximize both sensitivity and specificity were analyzed. In short, in the overall analysis, platelets (AUC: 0.924) and AST (AUC: 0.882) demonstrated both sensitivity and specificity above $80\%$. Among NS1 only, both NS1 + IgM, and IgM only positive participants, both NS1 + IgM positive parameters, i.e., platelets (AUC: 0.832), AST (AUC: 0.803), and TLC (AUC: 0.814), performed better than the other positive subjects, with sensitivity and specificity each above $70\%$. Although glucose was significant in multivariate logistics, it underperformed TLC in overall (AUC: 0.686 vs. 0.817), in NS1 only (AUC: 0.533 vs. 0.685), and in both NS1 + IgM (AUC: 0.632 vs. 0.814) positive study participants but exceeded it in IgM only positive participants (AUC: 0.712 vs. 0.581) (Table 7).
## 4. Discussion
The major periodic dengue outbreaks in 2010, 2013, 2016, 2019, and now 2022 show a 2–3 year cyclical pattern in Nepal. More sensitive serological tests such as ELISA and dengue confirmatory tests like PCR may not be widely available in developing countries like Nepal. As a result, most settings resort to less sensitive and less specific lateral flow assays for dengue diagnosis. Furthermore, during an outbreak condition, the rapid diagnostic kit is the method of choice due to its feasibility, ease of use, and economic value, as all patients cannot afford many expensive tests. Additionally, in settings where confirmatory tests are not easily accessible, positive rapid tests along with abnormal hematological and biochemical parameters would be essential for proper patient management to prevent life-threatening concerns of dengue.
Our study showed that dengue infection is significantly associated with age, which is also observed in other studies [11, 12]. Moreover, this study showed people in the age group of 20–29 years are more susceptible to infection, which is supported by other studies [13, 14]. Our study also presented the insignificant finding of dengue virus infection with sex, which is in contrast to that reported by Pun et al. [ 12].
In our study, we analyzed routine hematological and biochemical parameters that may be associated with dengue cases. Among the parameters analyzed, our study demonstrated thrombocytopenia, elevated AST, and increased blood glucose levels to be significantly associated with dengue-positive cases. Likewise, leukopenia, a low lymphocyte count, and monocytosis were significantly associated with certain serological courses of disease.
As per the WHO, hematocrit and thrombocytopenia are the most important laboratory parameters measured during dengue infection [15]. But, our study showed no significant association between hematocrit during the serological course of NS1 and IgM. Few studies report similar findings of insignificance; this may be because our study included only dengue fever patients with mild primary active infections. Thus, there is less chance of plasma leakage, which does not indicate abnormal hematocrit results [16, 17]. Thrombocytopenia, which is well correlated with dengue severity as shown by various studies, also remained significant in our study [18–21]. This decrease in platelets may be due to low production or increased destruction of platelets via activation of the complement factor C3 and further binding of the C5b-9 complex to the platelet surface [22].
In our study, the median increase in ALT, AST, and ALP was significantly associated with the dengue-positive group; nonetheless, only AST was found to be independently associated. This finding corresponds to studies supporting higher transaminase levels during dengue positivity [23, 24]. ALT is primarily of hepatic origin, while AST is of both hepatic and nonhepatic origin; hence, damage to nonhepatic tissues can also elevate AST as compared to ALT [25]. As a result, despite the significant results obtained in this study, a higher level of AST may not correctly represent hepatic involvement in the dengue-positive group [26]. Furthermore, the recommended drug to minimize dengue symptoms is acetaminophen, which even at therapeutic dosage can cause a temporary elevation in transaminase levels [27, 28].
One study demonstrated that hyperglycemic stress facilitates dengue virus translation and increases protein expression [29]. Our study also showed patients with dengue virus infection had higher glucose levels than the dengue-negative group. A similar outcome has also been presented in a study conducted by Hasanat et al. [ 30]. Another study suggested prioritizing patients with diabetes mellitus in the diagnosis of dengue but not using it as a factor in assessing dengue severity [31].
In addition to decreased platelets, increased AST levels, and high glucose levels as independent predictors of dengue virus infection, leukopenia with monocytosis (in NS1 only) and leukopenia with lymphopenia (in both NS1 + IgM) were also observed in certain serological durations of illness. Leukopenia with monocytosis and leukopenia with lymphopenia have also been observed in other studies [17, 32, 33]. Virus-induced destruction of WBC and inhibition of myeloid progenitor cells causes leukopenia, while monocytes phagocytose and present the antigen to T- helper cells, causing monocytosis; this explains leukopenia with monocytosis [34, 35].
## 5. Limitation
More sensitive tests such as ELISA and RT-PCR were not available for confirmation of dengue infection. Among the various disadvantages of ICT-based rapid tests is increased crossreactivity, which can lead to false positive outcomes. Furthermore, additional clinical features and disease severity of the patients were not evaluated.
## 6. Conclusion
The study found that certain hematological and biochemical parameters can predict the outcome of dengue infection, which can assist physicians in the diagnosis and proper patient management. Parameters, such as thrombocytopenia, AST, hyperglycemia, and leukopenia with monocytosis (in the NS1-only phase); thrombocytopenia, elevated AST, and high blood glucose (in the IgM-only phase); and thrombocytopenia, elevated AST, high blood glucose, and leukopenia with lymphopenia (in the dual-positive/both NS1 + IgM phase), can provide insight into dengue positivity and help with patient management.
## Data Availability
The datasets of the current study are available from the corresponding author upon reasonable request.
## Ethical Approval
The research has complied with all the relevant national regulations and institutional policies and has been approved by the Nepal Health Research Council (NHRC Registration No. $\frac{494}{2022}$).
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## Authors' Contributions
Conceptualization was done by BRB, AM, and RB; Methodology was provided by BRB and RB; Investigation was done by BRB, AM, and SA; *Formal analysis* was performed by BRB; BRB and MC wrote the original draft; BRB, RB, SA, AM, and MC reviewed and edited the manuscript; Validation, Supervision, and Project administration were done by RB.
## References
1. Holbrook M. R.. **Historical Perspectives on Flavivirus Research**. (2005.0) **11** 13-51. PMID: 32287584
2. Murugesan A., Manoharan M.. (2020.0)
3. Organization W. H.. **Dengue and severe dengue**. (2022.0)
4. Srikiatkhachorn A., Rothman A. L., Gibbons R. V.. **Dengue—how best to classify it**. (2011.0) **53** 563-567. DOI: 10.1093/cid/cir451
5. Tyler M., Sharp J. P.-P., Stephen H.. (2017.0)
6. EaDC D.. **Situation update of dengue 2022**. (2022.0)
7. De Paula S. O., Fonseca B. A. Ld. **Dengue: a review of the laboratory tests a clinician must know to achieve a correct diagnosis**. (2004.0) **8** 390-398. DOI: 10.1590/s1413-86702004000600002
8. Bilal Habib M., Sher Akbar N., Saleem A.. **A comparative study of serological diagnosis of Dengue outbreak 2019**. (2021.0) **21** 1117-1123. DOI: 10.4314/ahs.v21i3.20
9. 9Centers for Disease Control and Prevention NCfEaZIDN
Division of Vector-Borne Diseases (DVBD). Dengue
2022Georgia, USACDC. (2022.0)
10. Peeling R. W., Artsob H., Pelegrino J. L.. **Evaluation of diagnostic tests: dengue**. (2010.0) **8** S30-S37. DOI: 10.1038/nrmicro2459
11. Egger J. R., Coleman P. G.. **Age and clinical dengue illness**. (2007.0) **13** 924-927. DOI: 10.3201/eid1306.070008
12. Pun R., Pant K. P., Bhatta D. R., Pandey B. D.. **Acute dengue infection in the western terai region of Nepal**. (2011.0) **51**. DOI: 10.31729/jnma.25
13. Fukusumi M., Arashiro T., Arima Y.. **Dengue sentinel traveler surveillance: monthly and yearly notification trends among Japanese travelers, 2006–2014**. (2016.0) **10**. DOI: 10.1371/journal.pntd.0004924
14. Munir M. A., Alam S. E., Khan Z. U.. **Dengue fever in patients admitted in tertiary care hospitals in Pakistan**. (2014.0) **64** 553-559. PMID: 25272543
15. World Health O.. (2009.0)
16. Azin F. R. F. G., Gonçalves R. P., Pitombeira M. H. d. S., Lima D. M., Branco I. C.. **Dengue: profile of hematological and biochemical dynamics**. (2011.0) **34** 36-41. DOI: 10.5581/1516-8484.20120012
17. Salvatory Kalabamu F., Maliki S.. **Use of haematological changes as a predictor of dengue infection among suspected cases at kairuki hospital in dar Es salaam, Tanzania: a retrospective cross sectional study**. (2021.0) **5** 91-98. DOI: 10.24248/eahrj.v5i1.655
18. Agrawal V. K., Prusty B. S. K., Reddy C. S., Mohan Reddy G. K., Agrawal R. K., Sekher Srinivasarao Bandaru V. C.. **Clinical profile and predictors of Severe Dengue disease: a study from South India**. (2018.0) **9** 334-340. DOI: 10.22088/cjim.9.4.334
19. Hasan Khan M. I., Anwar E., Agha A.. **Factors predicting severe dengue in patients with dengue Fever**. (2013.0) **5**. DOI: 10.4084/mjhid.2013.014
20. Chao C.-H., Wu W.-C., Lai Y.-C.. **Dengue virus nonstructural protein 1 activates platelets via Toll-like receptor 4, leading to thrombocytopenia and hemorrhage**. (2019.0) **15**. DOI: 10.1371/journal.ppat.1007625
21. Khetan R. P., Stein D. A., Chaudhary S. K.. **Profile of the 2016 dengue outbreak in Nepal**. (2018.0) **11** p. 423. DOI: 10.1186/s13104-018-3514-3
22. Ojha A., Nandi D., Batra H.. **Platelet activation determines the severity of thrombocytopenia in dengue infection**. (2017.0) **7**. DOI: 10.1038/srep41697
23. Trung D. T., Thao L. T. T., Vinh N. N.. **Liver involvement associated with dengue infection in adults in Vietnam**. (2010.0) **83** 774-780. DOI: 10.4269/ajtmh.2010.10-0090
24. Sedhain A., Bhattarai G. R., Adhikari S., Shrestha B., Sapkota A.. **Liver involvement associated with dengue infection during A major outbreak in Central Nepal**. (2013.0) **2** 42-46. DOI: 10.3126/jaim.v2i2.8775
25. Green R. M., Flamm S.. **AGA technical review on the evaluation of liver chemistry tests**. (2002.0) **123** 1367-1384. DOI: 10.1053/gast.2002.36061
26. Samanta J., Sharma V.. **Dengue and its effects on liver**. (2015.0) **3** 125-131. DOI: 10.12998/wjcc.v3.i2.125
27. 27LiverTox
Clinical and Research Information on Drug-Induced Liver Injury
2016, https://www.ncbi.nlm.nih.gov/books/NBK548162/
28. Pandejpong D., Saengsuri P., Rattarittamrong R., Rujipattanakul T., Chouriyagune C.. **Is excessive acetaminophen intake associated with transaminitis in adult patients with dengue fever?**. (2015.0) **45** 653-658. DOI: 10.1111/imj.12756
29. Shen T. J., Chen C. L., Tsai T. T.. **Hyperglycemia exacerbates dengue virus infection by facilitating poly(A)-binding protein-mediated viral translation**. (2022.0) **7**. DOI: 10.1172/jci.insight.142805
30. Hasanat M. A., Ananna M. A., Ahmed M. U., Alam M. N.. **Testing blood glucose may be useful in the management of dengue**. (2010.0) **19** 382-385
31. Latt K. Z., Poovorawan K., Sriboonvorakul N., Pan-ngum W., Townamchai N., Muangnoicharoen S.. **Diabetes mellitus as a prognostic factor for dengue severity: retrospective study from Hospital for Tropical Diseases, Bangkok**. (2020.0) **7-8**. DOI: 10.1016/j.clinpr.2020.100028
32. Tsai J. J., Chang J. S., Chang K.. **Transient monocytosis subjugates low platelet count in adult dengue patients**. (2017.0) **2** 1-16. DOI: 10.1159/000457785
33. Oliveira É C. L. d., Pontes E. R. J. C., Cunha R. V. d., Fróes I. B., Nascimento D. d.. **Alterações hematológicas em pacientes com dengue**. (2009.0) **42** 682-685. DOI: 10.1590/s0037-86822009000600014
34. Lin S. F., Liu H. W., Chang C. S., Yen J. H., Chen T. P.. **[Hematological aspects of dengue fever]. Gaoxiong yi xue ke xue za zhi = the**. (1989.0) **5** 12-16
35. Chaloemwong J., Tantiworawit A., Rattanathammethee T.. **Useful clinical features and hematological parameters for the diagnosis of dengue infection in patients with acute febrile illness: a retrospective study**. (2018.0) **18** p. 20. DOI: 10.1186/s12878-018-0116-1
|
---
title: Construction of a Novel Diagnostic Model Based on Ferroptosis-Related Genes
for Hepatocellular Carcinoma Using Machine and Deep Learning Methods
authors:
- Shiming Yi
- Chunlei Zhang
- Ming Li
- Jiafeng Wang
journal: Journal of Oncology
year: 2023
pmcid: PMC9981290
doi: 10.1155/2023/1624580
license: CC BY 4.0
---
# Construction of a Novel Diagnostic Model Based on Ferroptosis-Related Genes for Hepatocellular Carcinoma Using Machine and Deep Learning Methods
## Abstract
Hepatocellular carcinoma (HCC) is one of the most general malignant tumors. Ferroptosis, a type of necrotic cell death that is oxidative and iron-dependent, has a strong correlation with the development of tumors and the progression of cancer. The present study was designed to identify potential diagnostic Ferroptosis-related genes (FRGs) using machine learning. From GEO datasets, two publicly available gene expression profiles (GSE65372 and GSE84402) from HCC and nontumor tissues were retrieved. The GSE65372 database was used to screen for FRGs with differential expression between HCC cases and nontumor specimens. Following this, a pathway enrichment analysis of FRGs was carried out. In order to locate potential biomarkers, an analysis using the support vector machine recursive feature elimination (SVM-RFE) model and the LASSO regression model were carried out. The levels of the novel biomarkers were validated further using data from the GSE84402 dataset and the TCGA datasets. In this study, 40 of 237 FRGs exhibited a dysregulated level between HCC specimens and nontumor specimens from GSE65372, including 27 increased and 13 decreased genes. The results of KEGG assays indicated that the 40 differential expressed FRGs were mainly enriched in the longevity regulating pathway, AMPK signaling pathway, the mTOR signaling pathway, and hepatocellular carcinoma. Subsequently, HSPB1, CDKN2A, LPIN1, MTDH, DCAF7, TRIM26, PIR, BCAT2, EZH2, and ADAMTS13 were identified as potential diagnostic biomarkers. ROC assays confirmed the diagnostic value of the new model. The expression of some FRGs among 11 FRGs was further confirmed by the GSE84402 dataset and TCGA datasets. Overall, our findings provided a novel diagnostic model using FRGs. Prior to its application in a clinical context, there is a need for additional research to evaluate the diagnostic value for HCC.
## 1. Introduction
According to the findings of the Global Cancer Statistics 2018, there were around 841,000 newly diagnosed cases of liver cancer and 782,000 deaths caused by liver cancer around the world, with China alone accounting for about $50\%$ of the total number of cases and deaths [1–3]. It is estimated that between 75 and 80 percent of all occurrences of liver cancer are caused by hepatocellular carcinoma (HCC), which is an aggressive kind of malignant tumor that is typically discovered at a later stage when treatment is no longer effective [4, 5]. Although there have been significant progresses and advancements in the treatment of HCC in recent years, in terms of surgical procedures, chemotherapeutic medications, and targeted drugs, HCC continues to have a very high incidence and mortality rate, which poses a serious threat to human health [6, 7]. The most popular blood biomarker for HCC, alpha-fetoprotein (AFP), demonstrates subpar performance as a serological test in HCC surveillance due to its low sensitivity being only $10\%$–$20\%$ in early-stage HCC and its labile levels during hepatitis flares [8, 9]. It is due to the fact that AFP levels fluctuate during hepatitis flares. Therefore, patients diagnosed with HCC at an early stage who have a high chance of experiencing recurrence need to be identified as quickly as possible so that tailored therapeutic options can be optimized and patient survival can be improved.
In recent years, the technology of microarrays has been employed in conjunction with integrated bioinformatics analysis in order to locate novel genes that have been linked to a range of diseases [10, 11]. *These* genes have the potential to function as diagnostic and prognostic biological markers. For instance, Lan et al. reported that the expressions of KIAA1429 were distinctly increased in HCC specimens. In individuals with HCC, having a high expression of KIAA1429 was related with having a bad prognosis. The knockdown of KIAA1429 resulted in a reduction in cell proliferation and metastasis both in vitro and in vivo. This was accomplished through a post-transcriptional alteration of GATA3 that was dependent on N6-methyladenosine [12]. Zhang et al. showed that DDX39 expression was positively connected with advanced clinical stages, and survival assays confirmed that patients with high-DDX39 levels exhibited a poor outcome. DDX39 was increased in HCC tissues and cells. According to the findings of a functional analysis, increased levels of DDX39 in HCC cells facilitated motility, migration, growth, and invasion via regulating the Wnt/-catenin pathway [13]. In addition, several genes in the blood of HCC patients were also reported to show important diagnostic values, such as serum IL27, HMMR, NXPH4, PITX1, and THBS4 [14, 15].
Ferroptosis is a sort of regulated cell death (RCD) that is triggered by the accumulation of harmful lipid peroxidation and is dependent on the presence of iron [16]. In recent years, the induction of ferroptosis has emerged as a promising therapeutic alternative to suppress tumor proliferation and growth, especially for advanced tumors that are resistant to surgical treatment, radiotherapy, and chemotherapy [17, 18]. It has been shown that ferroptosis plays an important role in the regulation of metabolism and redox biology, which has implications for the development of cancer and its treatment, including HCC [19–21]. Shan et al. showed that UBA1 contributed to the progression of HCC by elevating the activity of the Nrf2 signaling pathway and lowering the concentration of ferric ions, which triggered ferroptosis-inhibiting bioactivities [22]. In addition, several studies have reported the prognostic value of many ferroptosis-related genes (FRGs). However, the diagnostic model based on ferroptosis-related genes has not been investigated. In this study, we aimed to develop a diagnostic model based on ferroptosis-related genes using machine and deep learning methods.
## 2.1. Microarray Data Source
The GEO database was searched using the following keywords in order to retrieve the mRNA expression datasets of HCC: “hepatocellular carcinoma,” “homo sapiens” (porgn: txid9606),” and “expression profiling by array.” Following an in-depth analysis, two GSE profiles (GSE65372 and GSE84402) were chosen, and their respective downloads were initiated. GSE65372 and GSE84402 were based on GPL14951 and GPL570, respectively. The array data for GSE65372 were composed of 39 HCC specimens and 15 nontumor specimens, respectively. For GSE84402, the array data also included 14 HCC specimens and 14 nontumor specimens. All data were freely accessible, and the present study did not involve any human or animal experimentation.
## 2.2. Differential Expression Analysis
We began by retrieving the expression data of 237 FRGs from the GSE65372 database. Within this dataset, only 237 FRGs were found to be expressed. These data were then applied to normal samples and HCC samples. Following that, the Student's t-test was carried out in R in order to identify the FRGs that exhibited different levels of expression in the two distinct samples. Genes that had a p value of less than 0.001 were determined to be significant.
## 2.3. Pathway Analysis
The “clusterProfiler,” “enrichplot,” and “ggplot2” programs were used to conduct GO and KEGG pathway enrichment analyses in order to determine the biological characteristics of differently expressed genes (DEGs) linked to ferroptosis. These analyses were carried out in order to identify the biological features of DEGs. Enrichment results with an FDR (false discovery rate) of <0.05 were recognized as significant functional categories.
## 2.4. Candidate Diagnostic Biomarker Screening
Two different machine learning methods were employed to make predictions about the disease's progression in order to find meaningful prognostic variables. The least absolute shrinkage and selection operator (LASSO) is an approach for regression analysis that makes use of regularization in order to increase the accuracy of prediction. In order to determine the genes that are significantly connected with the differentiation of HCC samples from normal samples, the LASSO regression algorithm was implemented in R and carried out with the “glmnet” package. Support vector machine (SVM) is a popular type of supervised machine learning approach that may be used for both classification and regression. As a result, support vector machine recursive feature elimination (SVM-RFE) was utilized in order to choose the pertinent characteristics in order to find the group of genes that had the capacity to differentiate across groups the most effectively.
## 2.5. Diagnostic Value of Feature Biomarkers in HCC
An ROC curve was constructed by using the mRNA expression data of 39 HCC samples and 15 nontumor samples. It was done so that the predictive value of the selected biomarkers could be evaluated. The value of the area under the ROC curve was used to measure the diagnostic efficiency in distinguishing HCC samples from nontumor specimens, which was further confirmed using the GSE65372 dataset. Assessing AUC, sensitivity, and specificity were all parts of the process that were used to evaluate the diagnostic potential of the best gene biomarkers. In addition, the predict function included within the “glm” package of the R programming language was utilized to build a logistic regression model that was based on 11 novel genes. Our model was then used to make predictions regarding the sample types found within the GSE65372 dataset. In a similar manner, ROC curves were utilized in order to assess the diagnostic capability of the logistic regression model. In addition to this, the expressions of the essential genes were verified even further using the GSE84402 and TCGA datasets.
## 2.6. Statistical Analysis
All statistical analyses were conducted using R (version 3.6.3). $p \leq 0.05$ was considered as statistically significant.
## 3.1. Identification of Differential Expressed FRGs in the GSE65372 Datasets
40 of the 237 FRGs exhibited a dysregulated level between HCC specimens and nontumor specimens, including 27 increased and 13 decreased genes, which were identified from the GSE65372 dataset. The clustering heatmap displayed the expression pattern of FRGs that were differentially expressed between the samples (Figure 1(a)). Figure 1(b) illustrates the correlation between these genes.
## 3.2. Functional Analyses for the Differential Expressed FRGs
To explore the functional effects of differential expressed FRGs, we performed GO and KEGG assays. As shown in Figures 2(a) and 2(b), we found that the 40 differential expressed FRGs were mainly associated with responses to oxidative stress, cellular response to oxidative stress, regulation of autophagy, cellular response to chemical stress, mitochondrial outer membrane, organelle outer membrane, outer membrane, TOR complex, transcription coregulator activity, DNA-binding transcription factor bindin, and antioxidant activity. The results of KEGG assays indicated that the 40 differential expressed FRGs were mainly enriched in the longevity regulating pathway, AMPK signaling pathway, the mTOR signaling pathway, and hepatocellular carcinoma (Figure 2(c)).
## 3.3. Differential Expressed FRGs Were Identified as Diagnostic Genes for HCC
Estimating the diagnostic capability of differentially expressed FRGs was our goal in order to take into account the differences that exist between patients with HCC and healthy individuals. Subsequently, we carried out two separate machine learning algorithms in the GSE65372 datasets for the identification of the distinct differentially expressed FRGs in order to differentiate HCC from normal specimens. These algorithms were used to identify the FRGs that was significantly different between the two groups. In order to choose HCC-related features, the LASSO logistic regression algorithm was utilized, and the penalty parameter tuning process was carried out using 10-foldcross-validation (Figures 3(a) and 3(b)). After that, we sorted through the 17 differentially expressed FRGs using the SVM-RFE algorithm in order to locate the best possible combination of feature genes. In the end, seven genes were selected as the best candidates for feature genes (Figures 3(c) and 3(d)). Following the intersection of the marker genes generated from the LASSO and SVM-RFE models, 11 new markers (HSPB1, CDKN2A, LPIN1, MTDH, DCAF7, TRIM26, PIR, BCAT2, EZH2, and ADAMTS13) were identified for further investigation (Figure 3(e)).
## 3.4. The Identification of the Diagnostic Value of the New Model for HCC
With the use of the glm R package, we developed a logistic regression model. Subsequent ROC curves demonstrated that the 11 marker gene-based logistic regression model correctly differentiated normal samples from HCC samples with an area under the curve (AUC) value of 1.000. This model was based on the 11 marker genes mentioned earlier (Figure 4(a)). In addition, ROC curves were constructed for each of the 11 marker genes in order to provide light on the ability of individual genes to differentiate normal samples from those containing HCC. AUC was higher than 0.7 for every gene, as shown in Figure 4(b). Based on the information shown above, it appears that the logistic regression model provides a higher level of accuracy and specificity when compared to the individual marker genes when it comes to discriminating HCC samples from normal samples.
## 3.5. Expressions of Novel Diagnostic Genes in the GSE84402 and TCGA Datasets
In the final step of this process, we checked the expression of marker genes using the GSE84402 dataset. We found that the GSE20680 dataset was consistent with the patterns of expression for ADAMTS13, DCAF7, EZH2, HSPB1, and CDKN2A (Figure 5). Among them, the expressions of DCAF7, EZH2, HSPB1, and CDKN2A in HCC specimens were distinctly increased compared with normal specimens, while the expressions of ADAMTS13 were distinctly decreased in HCC samples. In addition, in TCGA datasets, we found that the expression of 10 genes showed a dysregulated level in HCC (Figure 6).
## 4. Discussion
HCC is the most prevalent primary malignancy of the liver, accounting for about $90\%$ of all malignant cases. It is also the most curable form of primary liver cancer [23, 24]. The fact that the formation of HCC is a multistep process, as well as a multigene alteration-induced malignancy with a high level of heterogeneity, has been established via extensive research and documentation [25, 26]. It has been determined that hepatitis B, hepatitis C, alcoholism, steatohepatitis, and obesity are all etiologic factors that contribute to the disease [27, 28]. Recent studies at the molecular levels have indicated that specific gene mutations play an important part in the progression of HCC. By controlling iron metabolism, amino acid and glutathione metabolism, and reactive oxygen species (ROS) metabolism, ferroptosis has shown promising results in inducing cancer cell death in recent years, especially in the elimination of aggressive malignancies that are resistant to conventional therapies [29, 30]. Therefore, ferroptosis can be a potential and powerful target for cancer therapy. However, the relationship between ferroptosis-related genes and HCC progression is still vastly unknown, making it a challenge to develop ferroptosis therapy for HCC.
Thanks to the development of high-throughput technologies, gene microarray analysis has emerged as a powerful tool for detecting DEGs and, by extension, putative biomarkers in a wide range of disorders. Gene microarray analysis has been used in a number of studies to discover crucial genes in the etiology of HCC. There is hope that integrated multiple gene microarray analysis will help find more reliable gene biomarkers. Machine learning algorithms have been shown to offer great potential for screening sensitive diagnostic biomarkers in a variety of diseases, and this research has only increased in the last few years [31, 32]. In this study, we screened differential expressed FRGs, and 40 of 237 FRGs exhibited a dysregulated level between HCC specimens and nontumor samples, including 27 increased and 13 decreased genes. By eliminating cells from the environment that lack vital nutrients, ferroptosis has been shown to play a crucial role in suppressing carcinogenesis, as demonstrated by recent scientific studies. Functional studies of FRGs as tumor promoters or inhibitors have increased in the field of HCC. The results of KEGG indicated that the 40 differential expressed FRGs were manly enriched in the longevity regulating pathway, AMPK signaling pathway, the mTOR signaling pathway, and hepatocellular carcinoma, highlighting their roles in HCC progression. Our finding suggested the 44 differential expressed FRGs may play an important role in the progression of HCC.
Based on the 40 differential expressed FRGs, we carried out LASSO and SVM and confirmed 11 novel marker genes (HSPB1, CDKN2A, LPIN1, MTDH, DCAF7, TRIM26, PIR, BCAT2, EZH2, and ADAMTS13). The AUC for all 11 genes are more than 0.75, indicating that they can reliably and accurately separate HCC specimens from nontumor specimens. Among the 11 genes, some genes have been functionally studied in HCC. For instance, He et al. reported that the expressions of MTDH were found to be distinctly elevated in HCC specimens. In HCC patients, the expressions of MTDH were predictive of a short overall survival without any heterogeneity. In addition, high-grade histological differentiation, nonvascular invasion, and HCC metastases were all found to be linked with MTDH expression. The results of in vitro investigations showed that MTDH has the ability to limit cell growth in all four HCC cell lines, in addition to activating caspase-$\frac{3}{7}$ activity and death [33]. Wang et al. showed that, when compared with normal liver tissue, the level of TRIM26 expression was much lower in HCC tissue; this was found to be associated with an advanced T stage and a bad prognosis. In vitro studies with HCC cells showed that inhibiting TRIM26 led to increased cancer cell proliferation and metastasis [34]. These findings were consistent with our findings. Our ROC curves showed that the logistic regression model based on these 11 marker genes successfully distinguished between normal and HCC samples (AUC = 1.000) using the R package glm. Our findings suggested the novel diagnostic model based on 11 marker genes had great clinical reference values. Finally, we demonstrated the expression of 11 marker genes in other GSE84402 and TCGA datasets. The expression of several genes was on track. However, more samples were needed to further confirm our findings.
Several limitations could also be found in our study. First, the sample size was low; despite the fact that our findings were constructed using and validated using two separate datasets. Validation of this model in larger prospective clinical studies is required in the future. Second, to further understand the molecular functions of the 11 critical genes, additional biological research is required.
## 5. Conclusion
We developed a novel diagnostic model based on 11 FRGs for HCC. These efforts may also serve to further promote patient compliance, assist healthcare providers in better managing patients, and eventually improve their overall health status and quality of life.
## Data Availability
The data used to support this study are available from the corresponding author upon request.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## Authors' Contributions
Shiming Yi and Jiafeng Wang designed the study and supervised the data collection. Shiming Yi and Chunlei Zhang analyzed the data and interpreted the data. Chunlei Zhang and Ming Li prepared the manuscript for publication and reviewed the draft of the manuscript. All authors have read and approved the manuscript.
## 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. Feng R. M., Zong Y. N., Cao S. M., Xu R. H.. **Current cancer situation in China: good or bad news from the 2018 Global Cancer Statistics?**. (2019) **39** p. 22. DOI: 10.1186/s40880-019-0368-6
3. Villanueva A., Carcinoma H.. **Hepatocellular carcinoma**. (2019) **380** 1450-1462. DOI: 10.1056/nejmra1713263
4. Hartke J., Johnson M., Ghabril M.. **The diagnosis and treatment of hepatocellular carcinoma**. (2017) **34** 153-159. DOI: 10.1053/j.semdp.2016.12.011
5. Dika I. E., Abou-Alfa G. K.. **Treatment options after sorafenib failure in patients with hepatocellular carcinoma**. (2017) **23** 273-279. DOI: 10.3350/cmh.2017.0108
6. Juárez-Hernández E., Motola-Kuba D., Chávez-Tapia N. C., Uribe M., Barbero Becerra V.. **Biomarkers in hepatocellular carcinoma: an overview**. (2017) **11** 549-558. DOI: 10.1080/17474124.2017.1311785
7. Schlachterman A., Craft W. W., Hilgenfeldt E., Mitra A., Cabrera R.. **Current and future treatments for hepatocellular carcinoma**. (2015) **21** 8478-8491. DOI: 10.3748/wjg.v21.i28.8478
8. Lencioni R.. **Loco-regional treatment of hepatocellular carcinoma in the era of molecular targeted therapies**. (2010) **78** 107-112. DOI: 10.1159/000315238
9. Nakakura E. K., Choti M. A.. **Management of hepatocellular carcinoma**. (2000) **14** 1085-1098. PMID: 10929592
10. Krizkova S., Kepinska M., Emri G.. **Microarray analysis of metallothioneins in human diseases--A review**. (2016) **117** 464-473. DOI: 10.1016/j.jpba.2015.09.031
11. Stahl F., Hitzmann B., Mutz K.. **Transcriptome analysis**. (2012) **127** 1-25. DOI: 10.1007/10_2011_102
12. Lan T., Li H., Zhang D.. **KIAA1429 contributes to liver cancer progression through N6-methyladenosine-dependentpost-transcriptional modification of GATA3**. (2019) **18** p. 186. DOI: 10.1186/s12943-019-1106-z
13. Zhang T., Ma Z., Liu L.. **DDX39 promotes hepatocellular carcinoma growth and metastasis through activating Wnt/**. (2018) **9** p. 675. DOI: 10.1038/s41419-018-0591-0
14. Eun J. W., Jang J. W., Yang H. D.. **Serum proteins, HMMR, NXPH4, PITX1 and THBS4; A panel of biomarkers for early diagnosis of hepatocellular carcinoma**. (2022) **11** p. 2128. DOI: 10.3390/jcm11082128
15. Kao J. T., Feng C. L., Yu C. J.. **IL-6, through p-STAT3 rather than p-STAT1, activates hepatocarcinogenesis and affects survival of hepatocellular carcinoma patients: a cohort study**. (2015) **15** p. 50. DOI: 10.1186/s12876-015-0283-5
16. Chen X., Li J., Kang R., Klionsky D. J., Tang D.. **Ferroptosis: machinery and regulation**. (2021) **17** 2054-2081. DOI: 10.1080/15548627.2020.1810918
17. Jiang X., Stockwell B. R., Conrad M.. **Ferroptosis: mechanisms, biology and role in disease**. (2021) **22** 266-282. DOI: 10.1038/s41580-020-00324-8
18. Xu T., Ding W., Ji X.. **Molecular mechanisms of ferroptosis and its role in cancer therapy**. (2019) **23** 4900-4912. DOI: 10.1111/jcmm.14511
19. Capelletti M. M., Manceau H., Puy H., Peoc’h K.. **Ferroptosis in liver diseases: an overview**. (2020) **21** p. 4908. DOI: 10.3390/ijms21144908
20. Gan B.. **Mitochondrial regulation of ferroptosis**. (2021) **220**. DOI: 10.1083/jcb.202105043
21. Mou Y., Wang J., Wu J.. **Ferroptosis, a new form of cell death: opportunities and challenges in cancer**. (2019) **12** p. 34. DOI: 10.1186/s13045-019-0720-y
22. Shan Y., Yang G., Huang H.. **Ubiquitin-like modifier activating enzyme 1 as a novel diagnostic and prognostic indicator that correlates with ferroptosis and the malignant phenotypes of liver cancer cells**. (2020) **10**. DOI: 10.1016/j.bbadis.2022.166528
23. Kim D. W., Talati C., Kim R.. **Hepatocellular carcinoma (HCC): beyond sorafenib-chemotherapy**. (2017) **8** 256-265. DOI: 10.21037/jgo.2016.09.07
24. Jabbour T. E., Lagana S. M., Lee H.. **Update on hepatocellular carcinoma: pathologists’ review**. (2019) **25** 1653-1665. DOI: 10.3748/wjg.v25.i14.1653
25. Zhang Y., Wang C., Xu H., Xiao P., Gao Y.. **Hepatocellular carcinoma in the noncirrhotic liver: a literature review**. (2019) **31** 743-748. DOI: 10.1097/meg.0000000000001419
26. Foerster F., Galle P. R.. **The current landscape of clinical trials for systemic treatment of HCC**. (2021) **13** p. 1962. DOI: 10.3390/cancers13081962
27. Cheung T. T., Ma K. W., She W. H.. **A review on radiofrequency, microwave and high-intensity focused ultrasound ablations for hepatocellular carcinoma with cirrhosis**. (2021) **10** 193-209. DOI: 10.21037/hbsn.2020.03.11
28. Wang W. T., Jin W. L., Li X.. **Intercellular communication in the tumour microecosystem: mediators and therapeutic approaches for hepatocellular carcinoma**. (2022) **1868**. DOI: 10.1016/j.bbadis.2022.166528
29. Liang C., Zhang X., Yang M., Dong X.. **Recent progress in ferroptosis inducers for cancer therapy**. (2019) **31**. DOI: 10.1002/adma.201904197
30. Mao H., Zhao Y., Li H., Lei L.. **Ferroptosis as an emerging target in inflammatory diseases**. (2020) **155** 20-28. DOI: 10.1016/j.pbiomolbio.2020.04.001
31. DeGregory K. W., Kuiper P., DeSilvio T.. **A review of machine learning in obesity**. (2018) **19** 668-685. DOI: 10.1111/obr.12667
32. Deo R. C.. **Machine learning in medicine**. (2015) **132** 1920-1930. DOI: 10.1161/circulationaha.115.001593
33. He R., Gao L., Ma J.. **The essential role of MTDH in the progression of HCC: a study with immunohistochemistry, TCGA, meta-analysis and in vitro investigation**. (2017) **9** 1561-1579
34. Wang Y., He D., Yang L.. **TRIM26 functions as a novel tumor suppressor of hepatocellular carcinoma and its downregulation contributes to worse prognosis**. (2015) **463** 458-465. DOI: 10.1016/j.bbrc.2015.05.117
|
---
title: Exploring Myocardial Ischemia-Reperfusion Injury Mechanism of Cinnamon by Network
Pharmacology, Molecular Docking, and Experiment Validation
authors:
- Tao Xue
- Yan Xue
- Yangyue Fang
- Chuanghong Lu
- Yu Fu
- Zefeng Lai
- Xiaojun Qin
- Feng Huang
- Zhiyu Zeng
- Jianping Huang
journal: Computational and Mathematical Methods in Medicine
year: 2023
pmcid: PMC9981296
doi: 10.1155/2023/1066057
license: CC BY 4.0
---
# Exploring Myocardial Ischemia-Reperfusion Injury Mechanism of Cinnamon by Network Pharmacology, Molecular Docking, and Experiment Validation
## Abstract
Myocardial ischemia-reperfusion injury (MIRI) is a common complication of acute myocardial infarction that seriously endangers human health. Cinnamon, a traditional Chinese medicine, has been used to counteract MIRI as it has been shown to possess anti-inflammatory and antioxidant properties. To investigate the mechanisms of action of cinnamon in the treatment of MIRI, a deep learning-based network pharmacology method was established to predict potential active compounds and targets. The results of the network pharmacology showed that oleic acid, palmitic acid, beta-sitosterol, eugenol, taxifolin, and cinnamaldehyde were the main active compounds, and phosphatidylinositol-3 kinase (PI3K)/protein kinase B (Akt), mitogen-activated protein kinase (MAPK), interleukin (IL)-7, and hypoxia-inducible factor 1 (HIF-1) are promising signaling pathways. Further molecular docking tests revealed that these active compounds and targets exhibited good binding abilities. Finally, experimental validation using a zebrafish model demonstrated that taxifolin, the active compound of cinnamon, has a potential protective effect against MIRI.
## 1. Introduction
Acute myocardial infarction (AMI) is a life-threatening cardiovascular condition that has significant global health and economic effects [1]. For example, in China, the mortality rate of AMI has been steadily rising since 2002, reaching 62.33 per 100,000 in urban regions and 78.47 per 100,000 in rural areas in 2018 [2]. In the United States, more than 11 million people were hospitalized with AMI in 2010, which resulted in a direct economic cost of more than $450 billion [3].
AMI is frequently accompanied by myocardial ischemia-reperfusion injury (MIRI), which can exacerbate cardiac dysfunction and increase the likelihood of a poor prognosis for AMI [4]. It is known that various pathophysiological factors are involved in MIRI [5], such as oxidative stress, apoptosis of cardiomyocytes, calcium overload, endothelial dysfunction, mitochondrial dysfunction, and intramyocardial inflammation [6]. Although the understanding of the mechanism of MIRI has made great progress over the past decade, the clinical therapeutic effect appears to be limited. Therefore, there is an urgent need to identify new targets and transformative therapeutic measures. In recent years, traditional Chinese medicine (TCM) has been used to counteract MIRI [7–9].
Cinnamon, a commonly used spice worldwide, is considered to have important pharmacological value and was included in the Chinese Pharmacopoeia in 2020. Cinnamon has antioxidant, anti-inflammatory, endothelium protecting, and immune response regulating properties and has been used in preclinical and clinical research to prevent and cure cardiovascular disorders [10]. Moreover, cinnamon and its extracts have been shown to have the potential to aid in the treatment of MIRI by reducing the area of myocardial infarction via various mechanisms [11, 12]. These studies highlight the possible relevance of cinnamon in MIRI therapy, but its mechanism of action remains unexplained.
Network pharmacology is a new technology that integrates chemistry, pharmacology, and bioinformatics to provide new insights into the complex mechanisms of action of herbal medicines [13]. However, some challenges with the network pharmacology approach remain, such as the lack of comprehensive data on various drugs, genes, and proteins [14]. Nevertheless, drug-target interaction prediction (DTI), as one of the most direct and successful methods for discovering new drugs and targets, can help to some extent solve the problems of network pharmacology. And several new DTI methods have been developed in recent years, all of which have shown promising results [15, 16]. In addition, molecular docking, as a tool for predicting binding patterns between small molecules and target proteins, is an important bridge between structural chemistry and the life sciences.
In this study, to better examine the primary components and probable mechanisms of cinnamon in the treatment of MIRI, a deep learning-based network pharmacology technique, molecular docking, and experiment were provided (Figure 1).
## 2.1. The Collection and Screening of the Active Compounds of Cinnamon
The chemical compounds of cinnamon were taken from the TCM System Pharmacology (TCMSP) Database (http://tcmspw.com/tcmsp.php) [17], following a screening process based on an oral bioavailability (OB) of ≥$30\%$ and drug-likeness (DL) of ≥0.18. As a supplement, additional active compounds were added by reviewing the literature.
## 2.2. Collection of Targets of Cinnamon for the Treatment of MIRI
The targets of cinnamon were obtained from the TCMSP and SwissTargetPrediction [18], and the targets for MIRI were found by scanning the Online Mendelian Inheritance in Man (OMIM) [19], the Therapeutic Target Database (TTD) (http://db.idrblab.net/ttd/) [20], the GeneCards database (http://www.genecards.org) [21], and the DisGeNET database (http://www.disgenet.org/web/DisGeNET/) [22] using the phrase “ischemia-reperfusion injury.” The active compounds of cinnamon and the targets of MIRI were then imported into the online tool Venny (http://bioinfogp.cnb.csic.es/tools/venny/), from which the intersection targets could be outputted.
## 2.3. Predictions of Targets Based on DTI
DTI has recently been shown to be effective in identifying potential targets. In 2022, Zhao et al. [ 23] proposed HyperAttentionDTI, which is a DTI-based model that integrates sequence-based deep learning and attention mechanisms. In order to display the interactions between amino acids and atoms and to manage how features are represented on the channel, HyperAttentionDTI calculates an attention vector for each amino acid and atom pair. In this study, HyperAttentionDTI was used for prediction due to its greater performance when compared to other models.
Based on the active compounds and targets collected, compounds were transformed into the simplified molecular input line entry system (SMILES) strings, the targets were converted into protein amino acid sequences, and each SMILES string was matched to each amino acid sequence one by one. We collected 29063 items of data as a test set. And the Davis dataset is a drug-target interaction dataset composed primarily of the SMILES structures of compounds, amino acid sequences of proteins, and interaction labels (0 or 1). In this study, the model was trained using five-fold cross-validation on the Davis dataset. Finally, based on the results of the HyperAttentionDTI predictions, the number of intersection targets was increased, providing a rich database for the subsequent analysis step of the network pharmacology method.
## 2.4. Construction of the Compound-Disease-Target Network
The active compounds and potential therapeutic targets were imported into Cytoscape (version 3.9.1) [24] to create a compound-disease-target network. Then, Network Analyzer was used to identify the potential main compounds of cinnamon for the treatment of MIRI during network visualization.
## 2.5. Construction of the Protein-Protein Interaction (PPI) Network
The intersection targets were loaded into the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org) [25] to obtain the PPI network, with the species set to “Homo sapiens” and the confidence score set to ≥0.7. The network was imported into Cytoscape for visualization and key target screening. Degree unDir, betweenness unDir, and closeness unDir were three crucial metrics determined by Centiscape (version 2.2) that were chosen as screening criteria based on the PPI network.
## 2.6. Enrichment Analysis
Gene Ontology (GO) functional enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathway enrichment analysis ($P \leq 0.01$) were performed on core targets using Metascape (https://metascape.org/) [26]. The species was set as H. sapiens, and the Bioinformatics website (http://www.bioinformatics.com.cn/) and the R toolkit (version 4.1.2) were used to draw the graph. Then, a cinnamon-active compound-target-pathway network was built, which was required for further therapeutic target screening.
## 2.7. Molecular Docking Verification
To further verify the main compounds and targets outlined above, molecular docking was used to calculate their binding affinities. First, the crystal structures of the protein targets and related information were obtained from the UniProt [27] (https://www.uniprot.org/) and RCSB Protein Data Bank (PDB) [28] (https://www.rcsb.org/) databases. For proteins that could not be queried, a PDB format file was created using a homology modeling method, and a three-dimensional (3D) structure was predicted using Aphafold2 [29]. Then, using AutoDockTools (version 1.5.7), operations, such as hydrogenation, charge addition, removal of water molecules, and removal of metal ions, were carried out, and the files were ultimately converted into pdbqt format. Second, the 3D structures of the active compounds were received from the PubChem [30] database (https://pubchem.ncbi.nlm.nih.gov/). Compounds in which the 3D structure could not be found were transformed into a 3D structure according to their two-dimensional structure. Subsequently, qvina-w was used to perform blind docking while AutoDockTools was used to create the global docking box [31]. The binding score was used to evaluate the ability of a natural compound to bind to the target. Finally, Python (version 3.9.1) and Pymol (version 2.4) were used to create heat maps and 3D docking maps of the docking results.
## 2.8. Animal Experiment
We designed a zebrafish model to confirm the protective effect of the obtained composition on MIRI. Due to equipment and financial constraints, only oleic acid, eugenol, taxifolin, and cinnamaldehyde were selected for the experiment. We randomly divided 84 zebrafish larvae into 7 groups and pretreated them with different microinjections 12 hours before hypoxia and reoxygenation (H/R). We then assessed their cardiac function and arrhythmia.
## 2.8.1. Chemicals
We obtained 3,3′-dimethoxybenzidine from Aladdin (Shanghai, China), and 1-phenyl 2-thiourea (PTU) was purchased from Sigma-Aldrich (St. Louis, MO).
## 2.8.2. Animal Care and Use
We used the AB wild-type strain and the transgenic line, Tg (cmlc:EGFP), of zebrafish (purchased from Guangxi Yisheng Biotechnology Co., Ltd. Nanning, GX, China). Zebrafish were housed in the zebrafish husbandry center of Guangxi Medical University, where they were fed with live brine shrimp twice daily and maintained at 28.5°C on a 14-hour light/10-hour dark cycle. As described in a previous study [32], $0.3\%$ PTU was added to the egg water of the Tg (cmlc:EGFP) fish 24 hours postfertilization (hpf), and fluorescent microscopy was used to select healthy embryonic zebrafish with fluorescent hearts at 48 hpf. All procedures were approved by the Guangxi Medical University Animal Care and Use Committee.
We randomly assigned 84 larvae to 7 groups. They were pretreated with Phosphate Buffer Solution (PBS, purchased from Sigma-Aldrich, St. Louis, MO), $1\%$ DMSO (purchased from Sigma-Aldrich, St. Louis, MO), eugenol (0.1 μg/g body weight, purchased from Sigma-Aldrich, St. Louis, MO), cinnamaldehyde (0.5 μg/g body weight, purchased from Sigma-Aldrich, St. Louis, MO), oleic acid (1.5 μg/g body weight, purchased from Sigma-Aldrich, St. Louis, MO), and taxifolin (0.5 μg/g body weight, purchased from Sigma-Aldrich, St. Louis, MO) by microinjection 12 hours before H/R or normoxic water starting at 2 days post-fertilization (dpf), respectively.
## 2.8.3. Optical Imaging and Heart Function Analysis
The environmental temperature was maintained at 23°C ± 2°C. Hearts were examined between 2 and 6 dpf, when a distinct ventricle and atrium were present. To evaluate cardiac function, the Tg (cmlc:EGFP) zebrafish embryos were embedded in $4\%$ methylcellulose (warmed to room temperature) and subjected to video capturing, maintaining the anterior orientation to the left and the dorsal orientation to the top of the field. Direct immersion optics were used in conjunction with a digital high-speed camera (100 frames/second; C13440; Hamamatsu Digital Camera) mounted on a Leica microscope (DMi 8; McBain Instruments) to record 15-second movies of beating hearts. Images were captured using the HC Image software (Hamamatsu). Cardiac function was analyzed from the high-speed movies using a semiautomatic optical heartbeat analysis software (freely available for research purposes at http://www.sohasoftware.com), which quantifies heart rate (HR), diastolic area (DA)/systolic area (SA), and fractional area change (FAC). HR variability (HRV) was analyzed from the recorded videos. Image analysis software Ethovision XT (Noldus Information Technology, Inc., Leesburg, VA) was used to automate the procedure for counting the HR. Data from each record were exported and archived in text format for further HRV analysis. Nonlinear analysis provided two consecutive RR intervals (RRn and RRn +1), which were projected on the Poincaré graph and used to adjust to an elliptical function.
## 2.8.4. Statistical Analysis
All results are shown as means ± standard deviations. Comparisons between groups were evaluated using one-way analyses of variance and Student's t-tests. Statistical analyses were performed using Prism version 6.01 (GraphPad Software, San Diego, California, CA). Data were considered statistically significant at $P \leq 0.05.$
## 3.1. Active Compounds and Potential Targets of Cinnamon
To investigate the therapeutic mechanism, we first collected the active ingredients of cinnamon. We retrieved 220 cinnamon compounds from the TCMSP database. After screening with an OB of $30\%$ and DL of 0.18, 7 compounds remained. As a supplement, 12 additional active compounds were incorporated by analyzing the literature [33, 34] (Table 1).
The corresponding targets of the active compounds of cinnamon were obtained from the TCMSP and SwissTargetPrediction databases, and 271 targets were identified after merging the UniProt database entries and deleting duplicate values. In addition, 1844 targets of MIRI were obtained by screening and de-duplicating the GeneCards database, the TTD database, the OMIM database, and the DisGeNET database. After screening 271 active compound targets and 1844 MIRI targets using the Venn diagram, 130 intersection targets remained.
## 3.2. Predictions of the DTI Model
Since DTI is a binary classification task, the accuracy, precision, recall, and area under the curve (AUC) metrics were used to assess the performance of the model. The results are shown in Table 2. HyperAttentionDTI revealed 4214 active compound-target pairs with interaction relationships. The number of targets that the model predicted for each compound is displayed in Table 3. These targets were then combined with the intersecting targets obtained in Section 3.1 and deweighted to produce 1144 targets (Supplementary Table S1).
## 3.3. Construction of the Compound-Disease-Target Network
The active compounds and potential targets identified in the preceding steps were imported into Cytoscape to create the active compound-disease-target network diagram (Figure 2). Only the subnetwork with a high degree value was selected for visualization considering that it was impossible to depict the enormous number of targets. The analysis of the network revealed that the average degree of the 19 active compounds was 241.89, and oleic acid, stearic acid, palmitic acid, linoleic acid, beta-sitosterol, and eugenol were the top active compounds.
## 3.4. PPI Network Analysis
To explore the mechanism underlying the therapeutic effects of cinnamon against MIRI, 1144 targets were imported to the STRING database to construct a PPI network. And the degree, closeness, and betweenness were calculated to be 51.64, 0.00038, and 1580.54, respectively. After screening according to these thresholds, we obtained 216 key targets (Supplementary Table S2) (Figure 3).
## 3.5. The Results of GO and KEGG Enrichment Analyses
To determine the molecular mechanisms underlying cinnamon treatment of MIRI, we used Metascape to carry out GO biofunctional annotation and KEGG pathway enrichment analysis of the key targets. We obtained 3093 GO terms, which comprised 2675 biological process (BP) terms, 185 cellular component (CC) terms, and 233 molecular function (MF) terms. The top 10 considerably enriched terms for BP, CC, and MF are visualized in Figure 4(a). Results showed that the main BP terms were positive regulation of cell migration, cell death, cell motility, cellular component movement, and locomotion; the main CC terms were vesicle lumen, secretory granule lumen, cytoplasmic vesicle lumen, and membrane raft; the main MF terms were kinase binding, signaling receptor activator activity, protein kinase binding, and integrin binding.
We then identified 208 KEGG signaling pathways. The top 20 paths with the highest level of enrichment were chosen for visualization (Figure 4(b)). Results showed that the targets were enriched mainly in the lipid and atherosclerosis, PI3K-Akt, MAPK, and IL-17 signaling pathways. In addition to these signaling pathways, the HIF-1 signaling pathway was shown to be involved in the reprogramming of cellular energy metabolism, which suggested that cinnamon affects MIRI by interfering with HIF-1.
Furthermore, to better understand the mechanism of action of cinnamon in the treatment of MIRI, an active compound-target-pathway relationship network was built based on the enriched targets of the KEGG pathway (Figure 5). Results revealed an interaction between the active compound and the target as well as the related pathways of cinnamon for the treatment of MIRI.
## 3.6. Molecular Docking Results
To validate our findings, we used molecular docking to evaluate the interaction between the core active compounds and the targets. The binding affinity was less than −5.0 kcal/mol, which indicated a good interaction. Numerous significant targets, including prostaglandin-endoperoxide synthase 2 (PTGS2), glycogen synthase kinase 3 (GSK3B), and mitogen-activated protein kinase 14 (MAPK14), were docked to six primary active compounds in cinnamon. The binding affinity results are shown in Table 4. Oleic acid, palmitic acid, eugenol, and taxifolin have good binding affinity with PTGS2, beta-sitosterol has a good binding affinity with GSK3B, and cinnamaldehyde has a good binding affinity with MAP2K1. These results demonstrated that therapeutic benefit of the six compounds could have been achieved by PTGS2, GSK3B, and MAP2K1. The conformation of the core active compounds and targets is shown in Figure 6.
## 3.7. Hypoxia/Reoxygenation-Induced Cardiac Contractility Dysfunction
To simulate the development of MIRI, embryos were transferred to a beaker of hypoxic water for 48 hours and subsequently normoxic water for 2 hours. Heart functions, including FAC, DA, and SA were measured using a digital high-speed camera. We observed a decrease in FAC and an increase in DA and SA, implying cardiac dysfunction and consequent functional compensation. We next evaluated the heart function of the zebrafish larvae after reoxygenation and observed the further deterioration after the recovery of oxygen, which indicated that the myocardium suffered from ischemia-reperfusion injury (Figures 7(a) and 7(b)). Notable, both hypoxia and hypoxia/reoxygenation increased the risk of arrhythmia (Figures 8(a)–8(d)).
## 3.8. Taxifolin from Cinnamon Alleviated Myocardial Ischemia-Reperfusion Injury
We screened four natural compounds from cinnamon including taxifolin, eugenol, oleic acid, and cinnamaldehyde and evaluated their potential effects on MIRI through the above network pharmacology analysis and previous reports. Two dpf, the zebrafish were initially treated with hypoxia for 48 hours and then transferred into normoxic water for 2 hours. Taxifolin, an active ingredient of cinnamon, has been shown to elicit anti-inflammatory and antioxidant stress pharmacological activity; however, its potential efficacy in MIRI prevention remains unknown. In the present study, we found that exposure to taxifolin attenuated hypoxia- and reoxygenation-induced loss of FAC and pathological cardiac compensation (Figures 7(a)–7(d)) in zebrafish and protects them from arrhythmias (Figures 8(e)–8(h)). Meanwhile, we failed to find the protective effects of eugenol, oleic acid and cinnamaldehyde on myocardial ischemia-reperfusion injury (Figures 7(a)–7(d)).
## 4. Discussion
Our study combining the DTI approach with network pharmacology revealed that the potential anti-MIRI compounds in cinnamon were oleic acid, palmitic acid, beta-sitosterol, eugenol, taxifolin, and cinnamaldehyde, which acted on important targets, including PTGS2, GSK3B, and MAPK14, as well as the signaling pathways of PI3K-Akt, MAPK, IL-7, and HIF-1.
Previous studies have demonstrated that oleic acid protects against cadmium-induced oxidative damage to the heart and liver tissues of male rats via the regulation of the antioxidant defense system, inflammatory response, and metabolic enzyme function [35]. Palmitic acid is a saturated fatty acid that plays a role in metabolic syndromes, cardiovascular diseases, cancer, neurodegenerative diseases, and inflammation by influencing molecular signaling [36–41]. Beta-sitosterol elicits anti-inflammatory [42], antioxidant [43], and immunomodulatory activity [44]. Eugenol is a phenolic aromatic molecule with antioxidant, anti-inflammatory, and antiapoptotic properties [45]. Eugenol has been studied as a therapy for ischemia-reperfusion damage. For example, eugenol protects against ischemia-reperfusion injury in heart transplant recipients by reducing the inflammatory response, apoptosis [46]. Taxifolin is involved in a variety of pharmacological processes, including antioxidant processes, mitochondrial protection, and advanced glycation end-product formation and suppression [47]. It has become increasingly valuable in the treatment of cancer, cardiovascular diseases, and chronic hepatitis, among other diseases [48]. Cinnamaldehyde is an aldehyde organic compound with anti-inflammatory, antioxidant, and vascular protective properties [49, 50] and has shown considerable promise in the prevention and treatment of cardiovascular diseases. Taken together, these studies demonstrate the potential of these compounds in the treatment of MIRI.
Furthermore, we built a PPI network and obtained the top 10 target proteins of cinnamon against MIRI-mediated cardiac dysfunction (TNF, IL6, GAPDH, IL1B, TP53, INS, VEGFA, EGFR, SRC, and CTNNB1) by degree value. According to the core target proteins, the GO enrichment analysis demonstrated that target proteins of cinnamon core components are mainly involved in the positive regulation of cell death or migration, and these cellular biological behaviors play a crucial role in aggravating MIRI and subsequent cardiac dysfunction. In addition, KEGG enrichment analysis showed that cinnamon could interfere with a variety of MIRI signal pathways, especially the PI3K-Akt, MAPK, and IL-17 signaling pathways, which are involved in inflammation and oxidative stress. The PI3K-Akt signaling pathway has been detected in numerous different types of cells and is involved in the physiological processes of various diseases, including cancer, myocardial infarction, and heart failure. Icariside II is a bioflavonoid compound that has a positive effect on MIRI by activating the PI3K-Akt signaling pathway and inhibiting inflammation and cardiomyocyte apoptosis [51]. The MAPK signaling pathway, which is involved in a cascade of multilevel protein kinases, such as JNK and p38 MAPK, has been implicated in apoptosis and inflammation [52]. By silencing integrin β3 (ITGB3), the MAPK signaling pathway becomes activated, which increases the phosphorylation of downstream GSK-3 and Cx43, promotes mouse cardiomyocyte proliferation, inhibits apoptosis and the inflammatory response in cardiomyocytes, and provides protection against MIRI [53]. In turn, the IL-7 signaling pathway is important for a variety of inflammatory responses [54]. Macrophages play an important role in MIRI development, and IL-7 can improve MIRI by promoting cardiomyocyte apoptosis via macrophage infiltration and polarization [55]. Surprisingly, the HIF-1 signaling pathway was discovered to be linked to cellular energy metabolism reprogramming. HIF-1 is a transcription factor that comprises two subunits: HIF-1α and HIF-1β. HIF-1 is an oxygen-sensitive transcription factor that drives the adaptive metabolic response to hypoxia and is important in MIRI [56]. Through the HIF-1/BNIP3 pathway, berberine is thought to enhance mitochondrial autophagy, lower cardiac enzyme activity, induce cardiomyocyte proliferation, block cardiomyocyte apoptosis, and protect the heart against MIRI [57].
PTGS2 is a subtype of prostaglandin-endoperoxide synthase and plays a specific role in the inflammatory response. It has been demonstrated that inhibiting PTGS2 activates the MAPK pathway, thereby reducing the inflammatory response and improving myocardial remodeling in mice with myocardial infarction [58]. GSK3B plays a significant role in the treatment of MIRI and is involved in energy metabolism, inflammation, apoptosis, and the oxidative stress response [59]. MAPK14, also known as p38a, is involved in cell proliferation, apoptosis, and the inflammatory response. An increasing number of studies have highlighted the significance of MAPK14 in the treatment of myocardial inflammatory diseases [60, 61]. Therefore, PTGS2, GSK3B, and MAPK14 may serve as target proteins related to MIRI.
Therefore, we conducted animal experiments to validate the active compounds. Due to the limitations of funding and experimental equipment, only four compounds were selected for verification. Fortunately, the potential protective effect of taxifolin on MIRI was observed in the zebrafish experiment. Notably, the solvent DMSO used in our experiments had nonnegligible zebrafish cardiotoxicity, which may lead to an underestimation of the MIRI-protective effects of other main components of cinnamon and underscore the need for process optimization for the extraction and preservation of active ingredients in Chinese herbal medicines. Nonetheless, considering that zebrafish shares high homology to human MIRI pathology mechanisms, our experimental results may provide new insights for further research.
The network pharmacology approach proposed in this study may achieve promising prediction results in terms of active compounds, corresponding targets, and pathways. However, because of the inherent nature of network pharmacology, it was not possible to obtain all compounds and targets via databases and analysis software. Moreover, even if the DTI model was used to expand the number of targets, the problem could not be entirely solved. Furthermore, relevant data could not be used to train the model to improve performance because of the difficulty in obtaining negative samples. Therefore, there is still room for improvement in future work in terms of the quality and quantity of training data.
## 5. Conclusion
In this study, we proposed a deep learning-based network pharmacology approach. With the aid of the DTI method, we predicted six potential active compounds and three targets, and their validity was assessed using molecular docking and an animal experiment. Although only one active compound, taxifolin, was finally demonstrated to have a protective effect against MIRI by the zebrafish model, we believe that more promising findings can be obtained if further experiments are conducted. Anyway, the aforementioned analysis of the prediction compounds, targets, and pathways according to other studies suggested that cinnamon may be considered as a potential for effective treatment of MIRI.
## Data Availability
The datasets analyzed during the current study are available in the https://github.com/2020-xuetao/MIRI.
## Conflicts of Interest
The authors declare that there are no conflicts of interest.
## Authors' Contributions
J.P.H., F.H., and Z.Y. conceived the idea. T.X. used network pharmacology and deep learning methods to predict the core components and targets and drafted the manuscript. Y.X. wrote the experimental part of the manuscript and analyzed the data. Y.Y.F. performed the molecular docking analysis. Y.F. screened the key targets of protein-protein interactions. C.H.L., Z.F.L., and X.J.Q. designed and conducted the zebrafish experiment. J.P.H., T.X., and Y.X. reviewed the manuscript. All authors approved the final version. Tao Xue and Yan Xue contributed equally to this work.
## References
1. Reed G. W., Rossi J. E., Cannon C. P.. **Acute myocardial infarction**. (2017) **389** 197-210. DOI: 10.1016/S0140-6736(16)30677-8
2. **Report on cardiovascular health and diseases burden in China: an updated summary of 2020**. (2021) **36** 521-545
3. Weintraub W. S., Daniels S. R., Burke L. E.. **Value of primordial and primary prevention for cardiovascular disease: a policy statement from the American Heart Association**. (2011) **124** 967-990. DOI: 10.1161/CIR.0b013e3182285a81
4. Ge X., Meng Q., Wei L.. **Myocardial ischemia-reperfusion induced cardiac extracellular vesicles harbour proinflammatory features and aggravate heart injury**. (2021) **10**. DOI: 10.1002/jev2.12072
5. Yellon D. M., Hausenloy D. J.. **Myocardial reperfusion injury**. (2007) **357** 1121-1135. DOI: 10.1056/NEJMra071667
6. Hausenloy D. J., Yellon D. M.. **Myocardial ischemia-reperfusion injury: a neglected therapeutic target**. (2013) **123** 92-100. DOI: 10.1172/JCI62874
7. Xue Y., Fu W., Liu Y.. **Ginsenoside Rb2 alleviates myocardial ischemia/reperfusion injury in rats through SIRT1 activation**. (2020) **85** 4039-4049. DOI: 10.1111/1750-3841.15505
8. Xu M., Li X., Song L.. **Baicalin regulates macrophages polarization and alleviates myocardial ischaemia/reperfusion injury via inhibiting JAK/STAT pathway**. (2020) **58** 655-663. DOI: 10.1080/13880209.2020.1779318
9. Liu H., Liu W., Qiu H.. **Salvianolic acid B protects against myocardial ischaemia-reperfusion injury in rats via inhibiting high mobility group box 1 protein expression through the PI3K/Akt signalling pathway**. (2020) **393** 1527-1539. DOI: 10.1007/s00210-019-01755-7
10. Shang C., Lin H., Fang X.. **Beneficial effects of cinnamon and its extracts in the management of cardiovascular diseases and diabetes**. (2021) **12** 12194-12220. DOI: 10.1039/D1FO01935J
11. Hwa J. S., Jin Y. C., Lee Y. S.. **2-Methoxycinnamaldehyde from**. (2012) **139** 605-615. DOI: 10.1016/j.jep.2011.12.001
12. Sedighi M., Nazari A., Faghihi M.. **Protective effects of cinnamon bark extract against ischemia–reperfusion injury and arrhythmias in rat**. (2018) **32** 1983-1991. DOI: 10.1002/ptr.6127
13. Hopkins A. L.. **Network pharmacology**. (2007) **25** 1110-1111. DOI: 10.1038/nbt1007-1110
14. Zhou Z., Chen B., Chen S.. **Applications of network pharmacology in traditional Chinese medicine research**. (2020) **2020** 7. DOI: 10.1155/2020/1646905
15. Torng W., Altman R. B.. **Graph convolutional neural networks for predicting drug-target interactions**. (2019) **59** 4131-4149. DOI: 10.1021/acs.jcim.9b00628
16. Nguyen T., Le H., Quinn T. P., Nguyen T., Le T. D., Venkatesh S.. **GraphDTA: predicting drug-target binding affinity with graph neural networks**. (2021) **37** 1140-1147. DOI: 10.1093/bioinformatics/btaa921
17. Ru J., Li P., Wang J.. **TCMSP: a database of systems pharmacology for drug discovery from herbal medicines**. (2014) **6** p. 13. DOI: 10.1186/1758-2946-6-13
18. Daina A., Michielin O., Zoete V.. **SwissTargetPrediction: updated data and new features for efficient prediction of protein targets of small molecules**. (2019) **47** W357-W364. DOI: 10.1093/nar/gkz382
19. Amberger J. S., Hamosh A.. **Searching online mendelian inheritance in man (OMIM): A knowledgebase of human genes and genetic phenotypes**. (2017) **58** 1.2.1-1.2.12. DOI: 10.1002/cpbi.27
20. Wang Y., Zhang S., Li F.. **Therapeutic target database 2020: enriched resource for facilitating research and early development of targeted therapeutics**. (2020) **48** D1031-D1041. DOI: 10.1093/nar/gkz981
21. Safran M., Dalah I., Alexander J.. **GeneCards Version 3: the human gene integrator**. (2010) **2010, article baq020**. DOI: 10.1093/database/baq020
22. Piñero J., Bravo À., Queralt-Rosinach N.. **DisGeNET: a comprehensive platform integrating information on human disease-associated genes and variants**. (2017) **45** D833-D839. DOI: 10.1093/nar/gkw943
23. Zhao Q., Zhao H., Zheng K., Wang J.. **HyperAttentionDTI: improving drug-protein interaction prediction by sequence-based deep learning with attention mechanism**. (2022) **38** 655-662. DOI: 10.1093/bioinformatics/btab715
24. Shannon P., Markiel A., Ozier O.. **Cytoscape: a software environment for integrated models of biomolecular interaction networks**. (2003) **13** 2498-2504. DOI: 10.1101/gr.1239303
25. Szklarczyk D., Morris J. H., Cook H.. **The STRING database in 2017: quality-controlled protein-protein association networks, made broadly accessible**. (2017) **45** D362-D368. DOI: 10.1093/nar/gkw937
26. Zhou Y., Zhou B., Pache L.. **Metascape provides a biologist-oriented resource for the analysis of systems-level datasets**. (2019) **10** p. 1523. DOI: 10.1038/s41467-019-09234-6
27. **UniProt: the universal protein knowledgebase in 2021**. (2021) **49** D480-D489. DOI: 10.1093/nar/gkaa1100
28. Burley S. K., Berman H. M., Kleywegt G. J., Markley J. L., Nakamura H., Velankar S.. **Protein data Bank (PDB): the single global macromolecular structure archive**. (2017) **1607** 627-641. DOI: 10.1007/978-1-4939-7000-1_26
29. Jumper J., Evans R., Pritzel A.. **Highly accurate protein structure prediction with AlphaFold**. (2021) **596** 583-589. DOI: 10.1038/s41586-021-03819-2
30. Kim S., Chen J., Cheng T.. **PubChem 2019 update: improved access to chemical data**. (2019) **47** D1102-D1109. DOI: 10.1093/nar/gky1033
31. Hassan N. M., Alhossary A. A., Mu Y., Kwoh C. K.. **Protein-ligand blind docking using quickvina-w with inter-process spatio- temporal integration**. (2017) **7** p. 15451. DOI: 10.1038/s41598-017-15571-7
32. Zou X., Liu Q., Guo S.. **A novel zebrafish larvae hypoxia/reoxygenation model for assessing myocardial ischemia/reperfusion injury**. (2019) **16** 434-442. DOI: 10.1089/zeb.2018.1722
33. Moreno E. K. G., de Macêdo I. Y. L., Batista E. A.. **Evaluation of antioxidant potential of commercial cinnamon samples and its vasculature effects**. (2022) **2022** 13. DOI: 10.1155/2022/1992039
34. Gao Z. Y., Chen T. Y., Yu T. T.. **Cinnamaldehyde prevents intergenerational effect of paternal depression in mice via regulating GR/miR-190b/BDNF pathway**. (2022) **43** 1955-1969. DOI: 10.1038/s41401-021-00831-0
35. Bhattacharjee B., Pal P. K., Chattopadhyay A., Bandyopadhyay D.. **Oleic acid protects against cadmium induced cardiac and hepatic tissue injury in male Wistar rats: a mechanistic study**. (2020) **244, article 117324**. DOI: 10.1016/j.lfs.2020.117324
36. Fatima S., Hu X., Gong R. H.. **Palmitic acid is an intracellular signaling molecule involved in disease development**. (2019) **76** 2547-2557. DOI: 10.1007/s00018-019-03092-7
37. Cheng L., Yu Y., Szabo A.. **Palmitic acid induces central leptin resistance and impairs hepatic glucose and lipid metabolism in male mice**. (2015) **26** 541-548. DOI: 10.1016/j.jnutbio.2014.12.011
38. Yuan L., Mao Y., Luo W.. **Palmitic acid dysregulates the Hippo-YAP pathway and inhibits angiogenesis by inducing mitochondrial damage and activating the cytosolic DNA sensor cGAS-STING-IRF3 signaling mechanism**. (2017) **292** 15002-15015. DOI: 10.1074/jbc.M117.804005
39. Binker-Cosen M. J., Richards D., Oliver B., Gaisano H. Y., Binker M. G., Cosen-Binker L. I.. **Palmitic acid increases invasiveness of pancreatic cancer cells AsPC-1 through TLR4/ROS/NF-**. (2017) **484** 152-158. DOI: 10.1016/j.bbrc.2017.01.051
40. Marwarha G., Rostad S., Lilek J., Kleinjan M., Schommer J., Ghribi O.. **Palmitate increases**. (2017) **57** 907-925. DOI: 10.3233/JAD-161130
41. Tian D., Qiu Y., Zhan Y.. **Overexpression of steroidogenic acute regulatory protein in rat aortic endothelial cells attenuates palmitic acid-induced inflammation and reduction in nitric oxide bioavailability**. (2012) **11** p. 144. DOI: 10.1186/1475-2840-11-144
42. Paniagua-Pérez R., Flores-Mondragón G., Reyes-Legorreta C.. **Evaluation of the anti-inflammatory capacity of beta-Sitosterol in rodent assays**. (2016) **14** 123-130. DOI: 10.21010/ajtcam.v14i1.13
43. Ponnulakshmi R., Shyamaladevi B., Vijayalakshmi P., Selvaraj J.. (2019) **29** 276-290. DOI: 10.1080/15376516.2018.1545815
44. Fraile L., Crisci E., Córdoba L., Navarro M. A., Osada J., Montoya M.. **Immunomodulatory properties of beta-sitosterol in pig immune responses**. (2012) **13** 316-321. DOI: 10.1016/j.intimp.2012.04.017
45. Ulanowska M., Olas B.. **Biological properties and prospects for the application of eugenol-a review**. (2021) **22** p. 3671. DOI: 10.3390/ijms22073671
46. Fen W., Jin L., Xie Q.. **Eugenol protects the transplanted heart against ischemia/reperfusion injury in rats by inhibiting the inflammatory response and apoptosis**. (2018) **16** 3464-3470. DOI: 10.3892/etm.2018.6598
47. Feng E., Wang J., Wang X.. **Inhibition of HMGB1 might enhance the protective effect of taxifolin in cardiomyocytes via PI3K/AKT signaling pathway**. (2021) **20** 316-332. DOI: 10.22037/ijpr.2020.113584.14384
48. Saito S., Yamamoto Y., Maki T.. **Taxifolin inhibits amyloid-**. (2017) **5** p. 26. DOI: 10.1186/s40478-017-0429-5
49. Mateen S., Rehman M. T., Shahzad S.. **Anti-oxidant and anti-inflammatory effects of cinnamaldehyde and eugenol on mononuclear cells of rheumatoid arthritis patients**. (2019) **852** 14-24. DOI: 10.1016/j.ejphar.2019.02.031
50. Nour O. A. A., Shehatou G. S. G., Rahim M. A., El-Awady M. S., Suddek G. M.. **Cinnamaldehyde exerts vasculoprotective effects in hypercholestrolemic rabbits**. (2018) **391** 1203-1219. DOI: 10.1007/s00210-018-1547-8
51. Guan B.-F., Dai X.-F., Huang Q.-B.. **Icariside II ameliorates myocardial ischemia and reperfusion injury by attenuating inflammation and apoptosis through the regulation of the PI3K/AKT signaling pathway**. (2020) **22** 3151-3160. DOI: 10.3892/mmr.2020.11396
52. Yue J., Lopez J. M.. **Understanding MAPK signaling pathways in apoptosis**. (2020) **21** p. 2346. DOI: 10.3390/ijms21072346
53. Wang M. C., Wang D., Lu Y. H., Li Z. H., Jing H. Y.. **Protective effect of MAPK signaling pathway mediated by ITGB3 gene silencing on myocardial ischemia-reperfusion injury in mice and its mechanism**. (2021) **25** 820-836. DOI: 10.26355/eurrev_202101_24647
54. Zhao H., Wu L., Yan G.. **Inflammation and tumor progression: signaling pathways and targeted intervention**. (2021) **6** p. 263. DOI: 10.1038/s41392-021-00658-5
55. Yan M., Yang Y., Zhou Y.. **Interleukin-7 aggravates myocardial ischaemia/reperfusion injury by regulating macrophage infiltration and polarization**. (2021) **25** 9939-9952. DOI: 10.1111/jcmm.16335
56. Zheng J., Chen P., Zhong J.. **HIF-1**. (2021) **23**. DOI: 10.3892/mmr.2021.11991
57. Zhu N., Li J., Li Y.. **Berberine protects against simulated ischemia/reperfusion injury-induced H9C2 cardiomyocytes apoptosis in vitro and myocardial ischemia/reperfusion-induced apoptosis in vivo by regulating the mitophagy-mediated HIF-1**. (2020) **11** p. 367. DOI: 10.3389/fphar.2020.00367
58. Ge Z. W., Zhu X. L., Wang B. C.. **MicroRNA-26b relieves inflammatory response and myocardial remodeling of mice with myocardial infarction by suppression of MAPK pathway through binding to PTGS2**. (2019) **280** 152-159. DOI: 10.1016/j.ijcard.2018.12.077
59. Wen C., Lan M., Tan X.. **GSK3**. (2022) **2022** 23. DOI: 10.1155/2022/2588891
60. Qin S. Q., Zhang Z. S., Wang X. Y., Shi J. Z., Yang X. B.. **MiR-24 protects cardiomyocytes against hypoxia/reoxygenation-induced injury through regulating mitogen-activated protein kinase 14**. (2020) **61** 806-814. DOI: 10.1536/ihj.19-496
61. Zhang L., Han B., Liu H.. **Circular RNA circACSL1 aggravated myocardial inflammation and myocardial injury by sponging miR-8055 and regulating MAPK14 expression**. (2021) **12** p. 487. DOI: 10.1038/s41419-021-03777-7
|
---
title: 'Traditional Chinese Medicine Compound Preparations Are Associated with Low
Disease-Related Complication Rates in Patients with Rheumatoid Arthritis: A Retrospective
Cohort Study of 11,074 Patients'
authors:
- Yanyan Fang
- Jian Liu
- Ling Xin
- Xiaolu Chen
- Xiang Ding
- Qi Han
- Mingyu He
- Xu Li
- Yanqiu Sun
- Fanfan Wang
- Jie Wang
- Xin Wang
- Jianting Wen
- Xianheng Zhang
- Qin Zhou
- Junru Zhang
journal: BioMed Research International
year: 2023
pmcid: PMC9981299
doi: 10.1155/2023/1019290
license: CC BY 4.0
---
# Traditional Chinese Medicine Compound Preparations Are Associated with Low Disease-Related Complication Rates in Patients with Rheumatoid Arthritis: A Retrospective Cohort Study of 11,074 Patients
## Abstract
### Objective
To evaluate whether traditional Chinese medicine compound preparations (TCMCPs) are associated with rheumatoid arthritis- (RA-) related complications (including readmission, Sjogren's syndrome, surgical treatment, and all-cause death) in patients with RA.
### Methods
Clinical outcome data were retrospectively collected from patients with RA discharged from the Department of Rheumatology and Immunology of the First Affiliated Hospital of Anhui University of Chinese Medicine from January 2009 to June 2021. The propensity score matching method was used to match baseline data. Multivariate analysis was conducted to analyze sex, age, the incidence of hypertension, diabetes, and hyperlipidemia and identify the risk of readmission, Sjogren's syndrome, surgical treatment, and all-cause death. Users of TCMCP and nonusers of TCMCP were defined as the TCMCP and non-TCMCP groups, respectively.
### Results
A total of 11,074 patients with RA were included in the study. The median follow-up time was 54.85 months. After propensity score matching, the baseline data of TCMCP users corresponded with those of non-TCMCP users, with 3517 cases in each group. Retrospective analysis revealed that TCMCP significantly reduced clinical, immune, and inflammatory indices in patients with RA, and these indices were highly correlated. Notably, the composite endpoint prognosis for treatment failure in TCMCP users was better than that in non-TCMCP users (HR = 0.75 (0.71-0.80)). The risk of RA-related complications in TCMCP users with high-exposure intensity (HR = 0.669 (0.650-0.751)) and medium-exposure intensity (HR = 0.796 (0.691-0.918)) was significantly lower than those in non-TCMCP users. An increase in exposure intensity was associated with a concomitant decrease in the risk of RA-related complications.
### Conclusion
The use of TCMCPs, as well as long-term exposure to TCMCPs, may lower RA-related complications, including readmission, Sjogren's syndrome, surgical treatment, and all-cause death, in patients with RA.
## 1. Introduction
Rheumatoid arthritis (RA) is a chronic inflammatory disease that mainly causes gradual joint damage and affects other body systems [1, 2]. The worldwide incidence rate of RA is approximately $1\%$, and although this condition affects people of all ages, it is more prevalent in women than in men [3, 4]. Currently, the etiology of RA is unclear, which poses a challenge to the effective treatment of RA and increases rehospitalization rates [5]. Although synovitis is a primary pathological marker of RA, many extra-articular manifestations may occur because of RA's complex, chronic, inflammatory, and autoimmune characteristics [6–8]. Extra-articular manifestations and complications are common in RA, contributing to higher incidence rates and premature mortality [6]. A hallmark clinical feature of RA is the symmetrical polyarthritis that manifests as redness and pain in the joints, especially smaller joints, and long-term morning stiffness [9, 10], with the potential to progress to serious joint injury and disability [11]. Progressive and severe joint injury, chronic pain, loss of function, and insufficient response to treatment regimens are indications for final joint replacement surgery [12]. Cohort studies based on national data from several countries have shown that RA is associated with high mortality [13, 14]. Therefore, readmission, extra-articular manifestations, surgical treatment, and all-cause death are considered potential RA-related complications.
Modern pharmacological treatments for RA mainly include nonsteroidal anti-inflammatory drugs, glucocorticoids, conventional disease-modifying antirheumatic drugs (cDMARDs), and biologic DMARDs that are used to alleviate chronic pain in patients by reducing the local inflammatory response [15]. However, RA treatment is complex and requires the combined application of multiple drugs, some of which have significant side effects and high treatment costs, resulting in poor patient compliance. Traditional Chinese medicine (TCM) might have many therapeutic advantages for RA [16–18]. Xin'an Jianpi Tongbi prescription, including Xinfeng capsule (XFC), Huangqin Qingre Chubi capsule (HQC), and Wuwei Wentong Chubi capsule (WWT), is a routinely used TCM compound preparation (TCMCP), which contains Astragalus membranaceus, Semen coicis, Tripterygium wilfordii, Scolopendra spp., Scutellaria baicalensis, Gardenia jasminoides, Prunus persica, Clematis chinensis, Poria cocos, Epimedium brevicornu, Cinnamomum cassia, Curcumae Longae, and other drugs. Many studies have shown that this TCMCP has high efficacy against RA [18–20]. A randomized, double-blind, multicenter, and placebo-controlled trial showed high efficacy and safety of XFC in the treatment of patients with RA [21, 22]. Animal experiments have demonstrated that HQC improves the baseline severity of arthritis in a collagen-induced arthritis mouse model [23, 24]. WWT has also been reported to have a good pharmacological effect on RA [25]. However, although the TCMCPs have favorable therapeutic effects on RA, their specific effect on the incidence of RA-related complications is still unclear.
In this study, we retrospectively analyzed the effect of Xin'an Jianpi Tongbi prescription on immune inflammation in RA and the risk of four RA-related complications, including readmission, Sjogren's syndrome, surgical treatment, and all-cause death.
## Study Cohort (Figure 1)
Clinical data of discharged patients with RA from the Department of Rheumatology and Immunology of the First Affiliated Hospital of Anhui University of Chinese Medicine were retrospectively collected from January 2009 to June 2021. The diagnostic criteria for RA by the American College of Rheumatology were adopted in this study [26]. Telephonic follow-up time was calculated from the time of discharge until February 28, 2022. Based on the history of TCMCP usage, the risk of RA-related complications, including readmissions, Sjogren's syndrome, surgical treatments, and all-cause death, was evaluated. This study was approved by the Medical Ethics Committee of the First Affiliated Hospital of Anhui University of Chinese Medicine (approval number: 2022MCZQ01).
## 2.2. Data Collection
Demographic information, including age and sex; clinical data including baseline complications, baseline cDMARD, and corticosteroid treatment; and data on TCMCPs were collected and evaluated retrospectively.
## 2.3. Treatment
In the First Affiliated Hospital of Anhui University of Chinese Medicine, the basic drugs for treating RA consist of cDMARDs (including methotrexate, leflunomide, sulfasalazine, and hydroxychloroquine sulfate), nonsteroidal anti-inflammatory drugs (including celecoxib, meloxicam, and lornoxicam), and glucocorticoids (methylprednisolone). It should be noted that TCM is a commonly used treatment means in TCM hospitals. We gradually withdrew the use of biologics by increasing the use of TCM.
## 2.4. Inflammatory and Immune Indices
Inflammatory and immune indices, including erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), anti-cyclic citrullinated peptide (anti-CCP), rheumatoid factor (RF), immunoglobulin A (IgA), immunoglobulin G (IgG), immunoglobulin M (IgM), complement component 3 (C3), and complement component 4 (C4) levels, were evaluated after TCMCP treatment.
## 2.5.1. Xin'an Jianpi Tongbi Prescription
Xin'an Jianpi Tongbi prescription is a compound preparation of TCM based on the Xin'an medical theory. It contains Xinfeng capsule [22] (Z20050062 from Wanyao Pharmaceutical Co., Ltd., patent number: ZL 2013 1 0011369.8), Huangqin Qingre Chubi capsule [24] (Z20200001 from Wanyao Pharmaceutical Co., Ltd., patent number: ZL 2011 1 0095718.X), and Wuwei Wentong Chubi capsule [25] (patent number: ZL 2020 10714863.0). The Xinfeng capsule is composed of Astragalus membranaceus, Semen coicis, Tripterygium wilfordii, and Scolopendra spp. These four medicinal materials were extracted by refluxing twice with $75\%$ ethanol. In the first step, ten times the amount of ethanol was added for extraction for 2 h, after which eight times the amount of ethanol was added and allowed to extract for 1.5 h. The drug residues were boiled with eight times the amount of water and extracted for 1.5 h. This was then filtered and allowed to stand. The supernatant was collected and combined with the alcohol extract under pressure to concentrate, and the paste was collected. The sample was dried, crushed, mixed with dextrin, and granulated with ethanol. This was followed by drying, whole granulating, sterilizing, filling, and outsourcing. The Huangqin Qingre Chubi capsule is composed of Scutellaria baicalensis, Prunus persica, Gardenia jasminoides, Semen coicis, and Clematis chinensis. These five medicinal materials were decocted and extracted three times as follows: ten times the amount of water was added for the first time and extracted for 1.5 h; eight times the amount of water was added for the second and third times and extracted for 1 h. The mixture was strained and allowed to stand. Then, the supernatant was absorbed and concentrated under pressure, and the paste was collected; this was then vacuum-dried, the dry extract was crushed, and dextrin was added. Ethanol was used to soften the materials, which were screened, granulated, dried, whole-grained, and filled into capsules. The Wuwei Wentong Chubi capsule is composed of Poria cocos, Epimedium brevicornu, Cinnamomum cassia, Curcumae Longae, and Scutellaria baicalensis. These five medicinal materials were decocted and extracted three times as follows: ten times the amount of water was added for the first time and extracted for 1.5 h; eight times the amount of water was added for the second and third times and extracted for 1 h. This mixture was strained and allowed to settle. Then, the supernatant was absorbed and concentrated under reduced pressure, and the paste was collected. This was then vacuum-dried, the dry extract was crushed, and dextrin was added. Ethanol was used to soften the material, which was then sieved using no. 12 mesh, granulated, dried, whole-grained, filled into capsules, and outsourced. All capsules were produced by the preparation center of the First Affiliated Hospital of Anhui University of Chinese medicine, and the variation range of each capsule was ±$10\%$.
## 2.5.2. RA-Related Complications
Readmission refers to RA patients who have been hospitalized twice or more. Sjogren's syndrome refers to an RA-related complication with a frequency of $10.41\%$ ($\frac{1022}{9813}$). Surgical treatment refers to RA patients with severe joint deformities requiring surgical treatment. All-cause death refers to death caused by long-term RA distress.
## 2.5.3. Classification of Quantitative Variables
The usage of TCMCPs was defined as “1,” and nonusage was defined as “0.” After TCMCP treatment, a decrease in ESR, CRP, IgA, IgG, IgM, C3, C4, RF, and anti-CCP levels was recorded as “1,” whereas an increase or no change in the level was recorded as “0.” The decrease in inflammatory and immune index values indicated the effectiveness of TCMCP treatment.
## 2.5.4. Exposure Intensity
According to exposure intensity, patients who received TCMCPs for less than 1 month, 1–3 months, 3–6 months, and ≥6 months after discharge were defined as the nonexposure, low-exposure, medium-exposure, and high-exposure groups, respectively.
## 2.6. Statistical Analysis
Continuous variables are reported as medians with interquartile ranges (IQR), whereas categorical variables are reported as frequencies and percentages. Categorical variables were compared using Fisher's exact test, whereas continuous variables were compared using the Wilcoxon signed-rank test. Univariate and multivariate COX proportional hazards models were developed to evaluate risk factors for the occurrence of endpoint events and are presented as hazard ratios (HR with $95\%$ confidence intervals (CIs)). Univariate models contained a single predictor for calculating different baseline risks for each site. Multivariate models included age, sex, comorbidities at baseline, and TCMCP as model covariates. All analyses were performed using SPSS V.22 (Armonk, NY, USA) software. Differences were considered statistically significant when the p value was less than 0.05.
## 3.1. Baseline Characteristics of TCMCP and Non-TCMCP Patients (Table 1)
The baseline data, including sex, age, hypertension, diabetes, hyperlipidemia, and cDMARD and corticosteroid treatment, of 9813 patients with RA who were successfully followed up were recorded. The median follow-up time was 54.85 months. Before propensity score matching, there were significant differences between TCMCP users and non-TCMCP users in terms of sex, age, hypertension, diabetes, cDMARDs, and corticosteroid treatment ($p \leq 0.05$). However, after matching, no significant difference was found between TCMCP users and non-TCMCP users in the same aspects ($p \leq 0.05$).
## 3.2. Changes in the RA-Related Inflammatory and Immune Indices after TCMCP Treatment (Table 2)
Hospitalization data of 3517 patients in the matched TCMCP group who received Xin'an Jianpi Tongbi prescription during hospitalization were collected and analyzed. Their posttreatment inflammatory and immune indices were lower than those before treatment ($p \leq 0.05$).
## 3.3. Association Analysis of TCMCPs with RA-Related Inflammatory and Immune Indices (Table 3)
We further analyzed the association between Xin'an Jianpi Tongbi prescription and RA-related inflammatory and immune indices. The results indicated that XFC was positively correlated with a decrease in CRP ($$p \leq 0.039$$, OR = 1.216), ESR ($$p \leq 0.003$$, OR = 1.298), and C4 ($$p \leq 0.028$$, OR = 1.258) levels. Similarly, HQC was positively correlated with a decrease in CRP ($p \leq 0.001$, OR = 1.641), ESR ($$p \leq 0.002$$, OR = 1.324), C4 ($$p \leq 0.024$$, OR = 1.272), IgG ($$p \leq 0.019$$, OR = 1.247), and IgA ($$p \leq 0.022$$, OR = 1.237) levels.
## Kaplan-Meier Curves for a Composite Endpoint for Treatment Failure for TCMCP Users versus Non-TCMCP Users (Figure 2)
The results of the log-rank test showed that TCMCP users had better composite endpoint prognoses for treatment failure (HR = 0.75 (0.71-0.80), $p \leq 0.001$) than non-TCMCP users.
## COX Regression Model for Analysis of Risk Factors for Four RA-Related Complications (Table 4) and Visualization of the Analysis Results (Figure 3)
Further, we used univariate and multivariate COX regression to analyze risk factors for the four RA-related complications, namely, readmission, Sjogren's syndrome, surgical treatment, and all-cause death. The results showed that TCMCPs reduced the risk of readmission, Sjogren's syndrome, surgical treatment, and risk of all-cause death by $20.7\%$, $32.6\%$, $29.0\%$, and $27.4\%$, respectively.
Advancing age increased the risk of Sjogren's syndrome, surgical treatment, and all-cause death by $60.5\%$, $94.7\%$, and $106.0\%$, respectively. The comorbidity hypertension increased the risk of readmission, Sjogren's syndrome, and all-cause death by $51.9\%$, $56.6\%$, and $50.6\%$, respectively. The male sex and the presence of comorbidity diabetes increased the risk of all-cause death by $51.7\%$ and $59.6\%$, respectively. Hyperlipidemia had a $34.0\%$ increased risk of readmission.
## 3.6. Risk of RA-Related Complications at Different Exposure Times (Table 5)
We found that the use of TCMCPs was associated with a lower risk of RA-related complications. In addition, the risk of RA-related complications varied according to the exposure time. Notably, the risk of RA-related complications in TCMCP users with high-exposure intensity (adjusted HR = 0.699, $95\%$CI = 0.650-0.751, $p \leq 0.001$) and medium-exposure intensity (adjusted HR = 0.796, $95\%$CI =0.691-0.918, $$p \leq 0.002$$) was significantly lower than that in non-TCMCP patients.
## 4. Discussion
In this population-based cohort study, a large amount of data on RA patients from the First Affiliated Hospital of Anhui University of Chinese Medicine were used to evaluate the effects of TCMCPs on clinical immunological and inflammatory indicators and RA-related complications. We found that RA patients treated with Xin'an Jianpi Tongbi Preparation not only exhibited lower immune and inflammatory indices than non-TCMCP users but also were associated with a low risk of RA-related complications.
TCM has a multicomponent, multitargeted synergistic anti-inflammatory and anti-immune effect. Previously, we found that TCMCPs significantly improved the RA-related immunological and inflammatory effects [16]. Modern pharmacological studies have also reported that Xin'an Jianpi Tongbi preparation drugs, i.e., Astragalus membranaceus, Semen coicis, Tripterygium wilfordii, Scolopendra spp., Scutellaria baicalensis, Gardenia jasminoides, Poria cocos, Epimedium brevicornu, Cinnamomum cassia, and Curcumae Longae, can improve the RA-related immunological and inflammatory response. Among them, active agents in *Astragalus membranaceus* have been shown to improve RA-induced synovial and joint injury [27, 28]. Semen coicis extract, including polyphenols and polysaccharides, has immunological, antioxidant, and anti-inflammatory effects [29]. Tripterygium wilfordii lactone, the active ingredient of Tripterygium wilfordii[30, 31], inhibited cell growth and inflammatory response of RA-associated fibroblasts, such as synovial cells, by regulating the expression of the hsa-circ-0003353/microRNA-31-5p/cyclin-dependent kinase 1 axis [32]. Scolopendra spp. combined with TCM has shown significant clinical efficacy in patients with RA [33]. Baicalin had an anti-inflammatory effect in a collagen-induced arthritis rat model, possibly by inhibiting the toll-like receptor 2/myeloid differentiation factor 88/NF-kappa B p65 signaling pathway [34]. Geniposide exhibited anti-inflammatory and antiangiogenesis pharmacological effects through the inhibition of vascular endothelial growth factor-induced angiogenesis in vascular endothelial cells by reducing the translocation of sphingosine kinase 1 [35]. Poria cocos polysaccharide enhanced the secretion of immune stimulants but inhibited the secretion of immune inhibitors, enhancing the host immune response [36]. Icariin inhibited cell proliferation by interfering with the cell cycle in RA fibroblasts, including synovial cells, promoting mitochondria-dependent apoptosis and intracellular reactive oxygen species production, which potentially improves RA outcomes [37]. Cinnamomum cassia extract had a therapeutic effect on RA, which was attributed to its antiproliferation and antimigration effects on synovial fibroblasts [38]. A systematic review showed that curcumin had a significant effect on the clinical and inflammatory parameters of RA and significantly improved morning stiffness, walking time, and joint swelling [39]. Thus, the pharmacological effects of TCM support the use of TCMCP for reducing the risk of RA-related complications. These results demonstrated that TCMCPs could act as a protective factor against RA-related complications (readmission, Sjogren's syndrome, surgical treatment, and all-cause death).
However, we also found that RA patients with comorbidities such as hypertension or hyperlipidemia had a significantly high risk of readmission. A study showed that hypertension and dyslipidemia were the most common complications of RA [40]. Consistent with our results, these classic complications increased the risk of recurrence of RA inflammation [41], potentially contributing to increased readmission of RA patients. Advanced age and hypertension were shown to be significantly associated with the extra-articular manifestations of RA [42, 43], which is consistent with our findings. An analysis based on British electronic medical records showed that the incidence of joint replacement increased with age [12]. Our results revealed a $94.7\%$ increased risk of surgical treatment in patients with RA aged 57 years and older, which corroborates findings from previous studies. Consistent with other studies, our results also showed that older patients with RA, men, and those with hypertension and diabetes had a higher risk of death [44–46]. Collectively, these results show that advanced age is a significant risk factor for extra-articular diseases, surgical treatment, and all-cause death. In addition, comorbidity hypertension is a risk factor for admission, extra-articular diseases, and all-cause death, whereas hyperlipidemia and diabetes are risk factors for recurrent admission and all-cause death, respectively. Among patients with RA, the risk of all-cause death is higher in men than in women.
Our study further found that medium- and high-exposure intensity, especially high-exposure intensity, were significantly associated with a reduced risk of RA-related complications. This indicates that long-term treatment with TCM could decrease the frequency of RA-related complications, which is consistent with the results of previous clinical data mining studies [16, 47]. Our results also suggest that long-term exposure to Xin'an Jianpi Tongbi preparation reduces RA-related complications.
This study had some notable limitations. First, there were no radiological data in our research to measure the severity of RA disease. Although, early on, we retrieved radiological data from the hospital information system, these data were textual, and we lacked models and algorithms to process textual data. Second, biologic DMARDs were not included in this study owing to insufficient data, which constitutes a major difference from the common practice in RA treatment and prevents appropriate comparisons with most of the literature on RA. Third, the recurrence frequency per unit time was not calculated for the frequently hospitalized patients, which differs from the common practice for RA treatment and also hinders proper comparison with most of the literature on RA. Fourth, the lack of data on adverse events of TCMCPs in this study did not allow a comprehensive analysis of the role of the drugs. Finally, we only studied Sjogren's syndrome and lacked data on other extra-articular manifestations of RA, which makes our findings one-sided. We intend to address these limitations in our future research. Nevertheless, our study has two significant strengths: the clinical advantage of using TCM and the statistical advantage of using large samples. This was a population-based cohort study, which included the clinical administration of medication to a population, making our results more clinically acceptable. The large sample size provides sufficient statistical ability to study the improvement effect of TCMCP on RA-related clinical indicators.
## 5. Conclusion
This population-based cohort study showed that TCMCP use, as well as long-term exposure to TCMCP in patients with RA, decreased the risk of RA-related complications, including readmission, Sjogren's syndrome, surgical treatment, and all-cause death. These findings are expected to inform clinical decisions regarding the use of TCMCP in RA management.
## Data Availability
All relevant data are included in the manuscript.
## Conflicts of Interest
The authors declare that they have no competing interests.
## Authors' Contributions
FYY and LJ contributed to the study design. FYY contributed to data analysis and wrote the first draft. CXL, DX, HQ, HMY, LX, SYQ, WJ, WX, WFF, WJT, ZXH, and ZQ performed telephone follow-ups and data collection. LJ, XL, and ZJR supervised the project and revised the manuscript. All authors reviewed and agree to be accountable for the content of the final manuscript.
## References
1. Maden M.. **RA signaling in limb development and regeneration in different species**. (2020) **95** 87-117. DOI: 10.1007/978-3-030-42282-0_4
2. Giannini D., Antonucci M., Petrelli F., Bilia S., Alunno A., Puxeddu I.. **One year in review 2020: pathogenesis of rheumatoid arthritis**. (2020) **38** 387-397. PMID: 32324123
3. van der Woude D., van der Helm-van Mil A. H. M.. **Update on the epidemiology, risk factors, and disease outcomes of rheumatoid arthritis**. (2018) **32** 174-187. DOI: 10.1016/j.berh.2018.10.005
4. Huang J., Fu X., Chen X., Li Z., Huang Y., Liang C.. **Promising therapeutic targets for treatment of rheumatoid arthritis**. (2021) **12, article 686155**. DOI: 10.3389/fimmu.2021.686155
5. Liu Z. C., Gao L., Zhang W. H., Wang J., Liu R. R., Cao B. H.. **Effects of a 4-week Omaha System transitional care programme on rheumatoid arthritis patients' self-efficacy, health status, and readmission in mainland China: a randomized controlled trial**. (2020) **26**. DOI: 10.1111/ijn.12817
6. Figus F. A., Piga M., Azzolin I., McConnell R., Iagnocco A.. **Rheumatoid arthritis: extra-articular manifestations and comorbidities**. (2021) **20**. DOI: 10.1016/j.autrev.2021.102776
7. England B. R., Thiele G. M., Anderson D. R., Mikuls T. R.. **Increased cardiovascular risk in rheumatoid arthritis: mechanisms and implications**. (2018) **23**. DOI: 10.1136/bmj.k1036
8. Wasserman A.. **Rheumatoid arthritis: common questions about diagnosis and management**. (2018) **97** 455-462. PMID: 29671563
9. Brzustewicz E., Henc I., Daca A.. **Autoantibodies, C-reactive protein, erythrocyte sedimentation rate and serum cytokine profiling in monitoring of early treatment**. (2017) **3** 259-268. DOI: 10.5114/ceji.2017.70968
10. Pirmardvand Chegini S., Varshosaz J., Taymouri S.. **Recent approaches for targeted drug delivery in rheumatoid arthritis diagnosis and treatment**. (2018) **46** 502-514. DOI: 10.1080/21691401.2018.1460373
11. Lin Y. J., Anzaghe M., Schülke S.. **Update on the pathomechanism, diagnosis, and treatment options for rheumatoid arthritis**. (2020) **9** p. 880. DOI: 10.3390/cells9040880
12. Hawley S., Edwards C. J., Arden N. K.. **Descriptive epidemiology of hip and knee replacement in rheumatoid arthritis: an analysis of UK electronic medical records**. (2020) **50** 237-244. DOI: 10.1016/j.semarthrit.2019.08.008
13. Turkiewicz A., Neogi T., Björk J., Peat G., Englund M.. **All-cause mortality in knee and hip osteoarthritis and rheumatoid arthritis**. (2016) **27** 479-485. DOI: 10.1097/EDE.0000000000000477
14. Kerola A. M., Kazemi A., Rollefstad S.. **All-cause and cause-specific mortality in rheumatoid arthritis, psoriatic arthritis and axial spondyloarthritis: a nationwide registry study**. (2022) **61** 4656-4666. DOI: 10.1093/rheumatology/keac210
15. Littlejohn E. A., Monrad S. U.. **Early diagnosis and treatment of rheumatoid arthritis**. (2018) **45** 237-255. DOI: 10.1016/j.pop.2018.02.010
16. Fang Y., Liu J., Xin L.. **Identifying compound effect of drugs on rheumatoid arthritis treatment based on the association rule and a random walking-based model**. (2020) **2020** 10. DOI: 10.1155/2020/4031015
17. Lv Q. W., Zhang W., Shi Q.. **Comparison of**. (2015) **74** 1078-1086. DOI: 10.1136/annrheumdis-2013-204807
18. Liu J., Huang C. B., Wang Y.. **Chinese herbal medicine Xinfeng Capsule in treatment of rheumatoid arthritis: study protocol of a multicenter randomized controlled trial**. (2013) **11** 428-434. DOI: 10.3736/jintegrmed2013059
19. Li S., Wan L., Zhao L.. **Clinical observation of Huangqin Qingre Chubi capsule in treating rheumatoid arthritis and its effect on serum M1 and M2 inflammatory factors**. (2022) **38** 190-194. DOI: 10.13412/j.cnki.zyyl.2022.02.025
20. Zhang Y., Liu J., Jiang H.. **Study on the effect of Wuwei Wentong Chubi Capsule on rheumatoid arthritis patients with cold-dampness syndrome based on association rules**. (2020) **9** 7-11
21. Liu J., Wang Y., Huang C.. **Efficacy and safety of Xinfeng capsule in patients with rheumatoid arthritis: a multi-center parallel-group double-blind randomized controlled trial**. (2015) **35** 487-498. DOI: 10.1016/s0254-6272(15)30130-8
22. Meng M., Wu X., Wang X. Y., Liu J., Du D., Ge Z.. **Study on the quality evaluation index method of Xinfeng Capsules**. (2012) **24** 456-458. DOI: 10.16448/j.cjtcm.2012.05.037
23. Wang X., Chang J., Zhou G.. **The traditional Chinese medicine compound Huangqin Qingre Chubi capsule inhibits the pathogenesis of rheumatoid arthritis through the CUL4B/Wnt pathway**. (2021) **12, article 750233**. DOI: 10.3389/fphar.2021.750233
24. Liu J. Q., Liu X. C., Liu J., Zhang Y. Y., Wang T. J., Zhou A.. **Preliminary study on HPLC fingerprint of Huangqin Qingre Chubi Capsules and determination of three components**. (2022) **17** 56-58
25. Jiang H., Liu J., Wang Y.. **Screening the Q-markers of TCMs from RA rat plasma using UHPLC-QTOF/MS technique for the comprehensive evaluation of Wu-Wei-Wen-Tong Capsule**. (2021) **56**. DOI: 10.1002/jms.4711
26. Arnett F. C., Edworthy S. M., Bloch D. A.. **The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis**. (1988) **31** 315-324. DOI: 10.1002/art.1780310302
27. Jiang H., Wu F. R., Liu J., Qin X. J., Jiang N. N., Li W. P.. **Effect of astragalosides on long non-coding RNA expression profiles in rats with adjuvant-induced arthritis**. (2019) **44** 1344-1356. DOI: 10.3892/ijmm.2019.4281
28. Liu X. Y., Xu L., Wang Y.. **Protective effects of total flavonoids of**. (2017) **44** 105-114. DOI: 10.1016/j.intimp.2017.01.010
29. Zhang C., Zhang W., Shi R., Tang B., Xie S.. **Coix lachryma-jobi extract ameliorates inflammation and oxidative stress in a complete Freund's adjuvant-induced rheumatoid arthritis model**. (2019) **57** 792-798. DOI: 10.1080/13880209.2019.1687526
30. Wang J., Liu J., Wen J., Wang X.. **Triptolide inhibits inflammatory response and migration of fibroblast like synovial cells in rheumatoid arthritis through the circRNA 0003353/JAK2/STAT3 signaling pathway**. (2022) **42** 367-374. DOI: 10.12122/j.issn.1673-4254.2022.03.08
31. Lin N., Zhang Y. Q., Jiang Q.. **Clinical practice guideline for**. (2021) **11, article 608703**. DOI: 10.3389/fphar.2020.608703
32. Wen J. T., Liu J., Wan L.. **Triptolide inhibits cell growth and inflammatory response of fibroblast-like synoviocytes by modulating hsa-circ-0003353/microRNA-31-5p/CDK1 axis in rheumatoid arthritis**. (2022) **106, article 108616**. DOI: 10.1016/j.intimp.2022.108616
33. Zhong H., Zhao J.. **Clinical application of insect drugs**. (2003) **23** 257-259. PMID: 14719290
34. Bai L., Bai Y., Yang Y.. **Baicalin alleviates collagen-induced arthritis and suppresses TLR2/MYD88/NF-**. (2020) **22** 2833-2841. DOI: 10.3892/mmr.2020.11369
35. Wang Y., Wu H., Gui B. J.. **Geniposide alleviates VEGF-induced angiogenesis by inhibiting VEGFR2/PKC/ERK1/2-mediated SphK1 translocation**. (2022) **100, article 154068**. DOI: 10.1016/j.phymed.2022.154068
36. Ríos J. L.. **Chemical constituents and pharmacological properties of Poria cocos**. (2011) **77** 681-691. DOI: 10.1055/s-0030-1270823
37. Pu L., Meng Q., Li S., Liu B., Li F.. **Icariin arrests cell cycle progression and induces cell apoptosis through the mitochondrial pathway in human fibroblast-like synoviocytes**. (2021) **912, article 174585**. DOI: 10.1016/j.ejphar.2021.174585
38. Liu J., Zhang Q., Li R. L.. **Anti-proliferation and anti-migration effects of an aqueous extract of**. (2020) **58** 863-877. DOI: 10.1080/13880209.2020.1810287
39. Pourhabibi-Zarandi F., Shojaei-Zarghani S., Rafraf M.. **Curcumin and rheumatoid arthritis: a systematic review of literature**. (2021) **75**. DOI: 10.1111/ijcp.14280
40. Namas R., Joshi A., Ali Z., Al Saleh J., Abuzakouk M.. **Demographic and clinical patterns of rheumatoid arthritis in an Emirati cohort from United Arab Emirates**. (2019) **2019** 10. DOI: 10.1155/2019/3057578
41. Fragoulis G. E., Panayotidis I., Nikiphorou E.. **Cardiovascular risk in rheumatoid arthritis and mechanistic links: from pathophysiology to treatment**. (2020) **18** 431-446. DOI: 10.2174/1570161117666190619143842
42. Chen X., Zhang M., Wang T., Li Y., Wei M.. **Influence factors of extra-articular manifestations in rheumatoid arthritis**. (2020) **15** 787-795. DOI: 10.1515/med-2020-0217
43. Ortega-Hernandez O. D., Pineda-Tamayo R., Pardo A. L., Rojas-Villarraga A., Anaya J. M.. **Cardiovascular disease is associated with extra-articular manifestations in patients with rheumatoid arthritis**. (2009) **28** 767-775. DOI: 10.1007/s10067-009-1145-8
44. Kim S. U., Kim B. K., Park J. Y.. **Fibrosis-4 index at diagnosis can predict all-cause mortality in patients with rheumatoid arthritis: a retrospective monocentric study**. (2020) **30** 70-77. DOI: 10.1080/14397595.2018.1558760
45. Lee E. E., Shin A., Lee J.. **All-cause and cause-specific mortality of patients with rheumatoid arthritis in Korea: a nation-wide population-based study**. (2022) **89**. DOI: 10.1016/j.jbspin.2021.105269
46. Kuo C. F., Luo S. F., See L. C., Chou I. J., Chang H. C., Yu K. H.. **Rheumatoid arthritis prevalence, incidence, and mortality rates: a nationwide population study in Taiwan**. (2013) **33** 355-360. DOI: 10.1007/s00296-012-2411-7
47. Zhou Q., Liu J., Xin L.. **Exploratory compatibility regularity of traditional Chinese medicine on osteoarthritis treatment: a data mining and random walk-based identification**. (2021) **2021** 12. DOI: 10.1155/2021/2361512
|
---
title: Evaluation of a visual acuity eHealth tool in patients with cataract
authors:
- Joukje C. Wanten
- Noël J.C. Bauer
- Janneau L.J. Claessens
- Thomas van Amelsfort
- Tos T.J.M. Berendschot
- Robert P.L. Wisse
- Rudy M.M.A. Nuijts
journal: Journal of Cataract and Refractive Surgery
year: 2022
pmcid: PMC9981317
doi: 10.1097/j.jcrs.0000000000001108
license: CC BY 4.0
---
# Evaluation of a visual acuity eHealth tool in patients with cataract
## Body
Cataract is the world's leading cause of age-related vision loss.1 During the past few decades, it has become one of the most performed surgeries worldwide and the number of procedures is likely to increase.2 The corresponding postoperative care includes frequent and rather time-consuming routine check-up appointments. In combination with the low incidence of serious sight-threatening complications, optimizing the postoperative cataract care pathway through eHealth technology is a logical next step in improving the patient journey.
The efficiency of the postoperative care could be enhanced by using remote care using teleconsultation and (online) remote measurements. Over the past few years, organizing remote care has accelerated, partly because of the COVID-19 pandemic.3 Several clinics have already implemented this kind of care and replaced 1 or more regular clinical follow-up examinations by telephone consultations.4 However, these teleconsultations are only partly applicable in ophthalmologic care because they lack objective outcome parameters for visual acuity and refractive state. Upcoming eHealth applications which provide the opportunity of self-monitoring and collecting objective outcome parameters may offer a solution. Utilization of these applications will lower the burden on patients after cataract surgery by saving follow-up visits at the outpatient clinic, which may improve efficiency and lower costs. The increased use of digital tools in general supports the implementation of eHealth solutions.
One of these eHealth applications is the Easee web-based tool that allows patients to individually assess their visual acuity and corresponding refraction using a smartphone and computer screen. Recently, noninferiority was shown for refraction measurements of this tool compared with manifest refraction obtained from standard measurements at the outpatient clinic in a healthy study population. Besides, the web-based tool and ETDRS chart showed similar results for the uncorrected distance visual acuity (UDVA) with mean values of 0.33 ± 0.30 logMAR and 0.39 ± 0.39 logMAR ($$P \leq .21$$), respectively.5 The aim of this study was to validate the web-based tool for assessment of the visual acuity in patients who underwent cataract surgery. We hypothesize agreement between the visual acuity measurements performed by the web-based tool as compared with the conventional assessments.
## Abstract
The Easee eHealth tool was validated for the assessment of visual acuity in patients who underwent cataract surgery and showed clinically acceptable outcomes in up to $88\%$ of patients.
## Purpose:
To validate the Easee web-based tool for the assessment of visual acuity in patients who underwent cataract surgery.
### Setting:
University Eye Clinic Maastricht, Maastricht, the Netherlands.
### Design:
Prospective method comparison study.
### Methods:
Subjects aged between 18 and 69 years who underwent cataract surgery on 1 or both eyes at the Maastricht University Medical Center+ were eligible to participate in this study. The uncorrected (UDVA) and corrected distance visual acuity (CDVA) assessments were performed using the web-based tool (index test) and conventional ETDRS and Snellen charts (reference tests). The outcomes of the different tests were expressed in logMAR, and a difference of <0.15 logMAR was considered clinically acceptable.
### Results:
46 subjects with 75 operated eyes were included in this study. The difference of the UDVA between the web-based tool and ETDRS or Snellen was −0.05 ± 0.10 logMAR ($P \leq .001$ [0.15; −0.26]) and −0.04 ± 0.15 logMAR ($$P \leq .018$$ [0.24; −0.33]), respectively. For the CDVA, these differences were −0.04 ± 0.08 logMAR ($P \leq .001$ [0.13; −0.21]) and −0.07 ± 0.10 logMAR ($P \leq .001$ [0.13; −0.27]), respectively. The Pearson correlation coefficients between the web-based tool and ETDRS were maximally 0.94 and compared with Snellen 0.92. In total, $73\%$ to $88\%$ of the visual acuity measurement differences were within 0.15 logMAR.
### Conclusions:
The web-based tool was validated for the assessment of visual acuity in patients who underwent cataract surgery and showed clinically acceptable outcomes in up to $88\%$ of patients. Most of the participants had a positive attitude toward the web-based tool, which requires basic digital skills.
## Test–Retest
Firstly, a test–retest analysis was performed among 5 healthy volunteers by measuring the right eye UDVA using the Snellen, ETDRS visual acuity charts, and the web-based tool. The measurements were performed at 3 different dates by the same individual under the same controlled and optimized circumstances, providing an indication of intraindividual variability of these tests.
## Study Design and Recruitment
From November 2020 to March 2021, a total of 46 participants were recruited from the University Eye Clinic of Maastricht University Medical Center (MUMC+). Subjects were eligible if they were aged between 18 and 69 years, underwent cataract surgery on 1 or both eyes, and were able to perform the web-based tool in Dutch, German, or English. The age limit of 69 years was selected based on the data of European statistics concerning digital skills to minimize the effects of digital proficiency on study outcomes.6 All participants were informed about the study in advance and signed an informed consent before enrolment. This hospital-based validation study was approved by the local medical ethics committee and institutional review board of the MUMC+ (Maastricht, the Netherlands). The study was executed in accordance with the tenets of the Declaration of Helsinki.
## Conventional (Reference Tests) and Web-Based (Index Test) Assessments
Both UDVA and corrected distance visual acuity (CDVA) were assessed using the Snellen and ETDRS charts as reference tests. The Snellen chart was routinely assessed by an optometrist at the postoperative visit before study enrolment. For the Snellen chart, the line assessment method was used. After study enrolment, visual acuity was assessed using the ETDRS chart by the researcher. The chart was placed 4 m from the subject, and the last attempted line on the ETDRS chart was determined until no optotypes could be distinguished. The total number of correctly identified optotypes was added to the score of the last attempted line to determine a logMAR score. Monocular CDVA measurements were performed using trial frames with the subjective manifest refraction as routinely measured by the optometrist.
The web-based tool (Easee B.V.) is an online visual function test using a computer screen and a smartphone (Figure 1). The smartphone is used as a remote controller to submit the input of the user from a distance of 3 m from the computer screen. All participants performed the web-based tool at the outpatient clinic after their (regular) postoperative visit, under controlled and optimized conditions, using a commonly used smartphone (Samsung Galaxy S6) and a laptop (Dell Latitude 5501, 15.6 inch). The test consisted of 3 parts with audio and visually instructed guidance: intake, arrangement of the test (calibrating and connecting the screen, placing a chair at 3 m distance), and performing the test. A short demo video illustrated the purpose of the web-based tool. During the test, a sequence of optotypes (tumbling-E and proprietary optotypes) was displayed that had to be identified correctly by the subject. The application used built-in algorithms to check the consistency of the input.
**Figure 1.:** *The web-based tool is performed by the patient by using a computer, 3 m distance, and a smartphone used as a remote controller.*
All participants performed the web-based test twice. Firstly, monocular UDVA measurements were performed and secondly monocular CDVA measurements. CDVA was assessed using trial frames with the subjective manifest refraction as used during the reference tests. Subjects could get assistance in using the smartphone and were reminded to cover up the appropriate eye during the tests. The amount of time the participants needed to perform the web-based tool was collected. Online refractive measurements were not performed in this study.
All assessments were performed using a fixed sequence: All operated eyes were routinely examined performing the Snellen UDVA and CDVA, first left and then the right eye. Subsequently, the researcher conducted the ETDRS measurements followed by measurements using the web-based tool in the same abovementioned sequence. All participants were unaware of their results.
## Questionnaires
An exploratory questionnaire was performed to assess pretest and posttest confidence of subjects toward the web-based tool. Questions pertained to the recommendation of the web-based tool to other patients, to the level of confidence of the subjects in their results, and to the amount of assistance during the tests. Outcomes were scored for every individual subject using a Likert scale (ranging from “strongly agree” to “strongly disagree”). In addition, the digital skills indicator survey, derived from the Eurostat survey on ICT usage by individuals, was performed.7
## Sample Size
The sample size calculation of 46 participants was based on a desired limit of agreement (LoA) of 0.01 logMAR and an assumed SD of 0.02 logMAR.8
## Statistical Analysis
Analyses were performed using SPSS (v. 25, IBM Corp.). An outcome was considered statistically significant when the P value was ≤0.05. Frequencies and descriptive statistics were used to summarize the baseline criteria and analyze the distribution of the variables. Left and right eyes were analyzed both separately and combined.
When using the ETDRS and web-based tool, UDVA and CDVA were expressed in logMAR. Each optotype of the ETDRS chart had a score of 0.02 log units, and 1 line represented 0.10 logMAR. The Snellen test was measured in decimals and afterward converted to logMAR. The analysis of variance between the web-based tool, Snellen, and ETDRS charts was performed using general linear model–repeated measures. Differences between visual acuity outcomes of the individual tests were compared using paired t tests. Data were displayed in scatterplots, and the related Pearson correlation coefficients were calculated. Bland Altman plots were used to visualize the agreement between the web-based tool and the reference tests.9 A difference ≥0.15 logMAR was considered clinically relevant because this is the usual intraindividual variability in repeated visual acuity measurements.10–12
## RESULTS
Firstly, a small test–retest sample was performed among 5 healthy volunteers and showed SDs for the ETDRS of 0.05 logMAR, Snellen 0.04 logMAR, and web-based tool of 0.08 logMAR. The study population consisted of 22 women ($48\%$) and 24 men ($52\%$) with a mean age of 62.8 ± 7.1 years (ranging from 26 to 69 years). Bilateral cataract surgery was performed in 29 patients ($63\%$). The manifest refraction spherical equivalent (MRSE) was −0.41 ± 0.84 diopters (D) for 41 operated right eyes and −0.64 ± 1.33 D for 34 operated left eyes. Most of the 44 patients ($96\%$) had basic digital skills.
A total of 75 operated eyes completed the assessments using the web-based tool and the conventional ETDRS chart. Outcomes of the web-based tool, Snellen, and ETDRS chart showed a significant visual acuity underestimation of the index test, when compared with the reference tests for right UDVA and CDVA, and left CDVA (Table 1). The differences between the visual acuity outcomes of the web-based tool were maximally −0.07 ± 0.10 logMAR ($P \leq .001$) compared with the ETDRS chart and had a maximal value of −0.08 ± 0.12 logMAR ($P \leq .001$) compared with the Snellen chart. The correlation ranged from 0.70 to 0.94 and up to $88.2\%$ of the web-based outcomes was within the clinically significant difference cutoff value of ±0.15 logMAR (Table 2).
The correlation coefficients between the web-based tool and ETDRS chart of both eyes combined for UDVA and CDVA were 0.94 and 0.89, respectively (both $P \leq .001$) (Figures 2, A and 3, A). The corresponding Bland Altman plots showed $95\%$ LoAs ranging from 0.15 to −0.26 logMAR and 0.13 to −0.21 logMAR, respectively (Figures 2, B and 3, B). For the comparison between the scores of the web-based tool and Snellen chart, scatterplots and corresponding Bland Altman plots can be found in the Appendix (Supplemental Figures 1 and 2, http://links.lww.com/JRS/A761, http://links.lww.com/JRS/A762). The Snellen CDVA score had a statistically significant mean difference of maximally −0.08 ± 0.12 logMAR with the web-based outcomes. The UDVA and CDVA between Snellen and the web-based tool had a correlation coefficient of 0.89 and 0.71 (both $P \leq .001$), respectively. The $95\%$ LoA ranged from 0.24 to −0.33 logMAR for the UDVA and from 0.13 to −0.27 logMAR for the CDVA.
**Figure 2.:** *A: Scatterplot UDVA of the web-based tool and ETDRS chart for the right and left eyes. The line presents the line of equality. B: Bland-Altman plot of UDVA determined by the web-based and ETDRS chart. The blue line represents the mean value, and the red dashed lines represent the ±1.96 SD (95% limits of agreement).* **Figure 3.:** *A: Scatterplot CDVA of the web-based tool and ETDRS chart for the right and left eyes. The line presents the line of equality. B: Bland-Altman plot of CDVA determined by the web-based and ETDRS chart. The blue line represents the mean value, and the red dashed lines represent the ±1.96 SD (95% limits of agreement).*
The mean time for performing the web-based tool was for the UDVA of the first and second tested eye 98 ± 45 and 88 ± 48 seconds, respectively. The CDVA assessment was completed in 73 ± 43 seconds for the first and 65 ± 22 for the second tested eye. Questionnaire outcomes are shown in Figure 4. The distribution of these outcomes was skewed and therefore not suitable for additional analyses.
**Figure 4.:** *Outcomes of the questionnaire about the attitude and experiences of subjects using the web-based tool. Every questionnaire item was scored using a 5-level Likert scale. The outcomes are given in percentages.*
## DISCUSSION
The aim of this research was to validate the web-based tool for visual acuity assessment among patients who underwent cataract surgery. This study demonstrates statistically significant differences for both UDVA and CDVA scores between the web-based tool and the gold standard ETDRS chart of maximally −0.07 ± 0.10 logMAR and −0.05 ± 0.08 logMAR, respectively. Apparently, the web-based tool underestimates visual acuity, falling within the clinically acceptable cutoff of 0.15 logMAR. The Pearson correlation coefficients show a good reliability. However, it must be noted that this correlation cannot be defined as agreement because it only measures association.13 The Bland Altman plots show a wide distribution between these tests, with a $95\%$ LoA maximum variation between 0.15 and −0.26 logMAR. However, up to $88\%$ of the patients' visual acuity outcome differences were within the range of ±0.15 logMAR. Patients who were out of this range had either higher or worse visual acuity scores. Based on these results, we believe the web-based tool has an acceptable accuracy for clinical application.
Since the Snellen chart is the most commonly used test in daily clinical practice, the web-based outcomes were also compared with those obtained using a Snellen chart. Only the CDVA score showed a statistically significant mean difference of maximally −0.08 ± 0.12 logMAR. The Pearson correlation coefficients showed a reduced reliability compared with the correlation of the web-based tool and ETDRS chart. In total, $82\%$ of the patients had a visual acuity outcome difference within ±0.15 logMAR. Furthermore, explorative analyses did not reveal any consistent or useful relationships between questionnaire results and visual acuity outcomes.
Questionnaire outcomes showed that most of the participants had a positive attitude toward the web-based tool. The net promotor score for the confidence toward the web-based outcomes was 86.9 before and 91.1 after performing the test. Generally, the amount of time the participants needed to perform the web-based assessment declined over the course of measurements. We observed a learning curve for completing the test in which the last performed measurement was completed the fastest.
A study using the web-based tool indicated this test as a valid and safe method for measuring visual acuity and refraction in healthy eyes. They found no difference between UDVA assessed by the web-based tool and an ETDRS chart, with mean values of 0.33 ± 0.30 and 0.39 ± 0.39 logMAR ($$P \leq .21$$), respectively.5 A study among patients with keratoconus showed an UDVA mean difference of −0.01 logMAR ($$P \leq .76$$), comparing ETDRS and the web-based tool, with a broad distribution including a LoA of −0.63 to 0.60 logMAR, albeit in subjects with a lower visual acuity.14 *Several previous* studies compared digital tools with conventional visual acuity charts, including the “Eye Chart,” “Peek Acuity,” and “Vision at Home” tool. These tools showed maximal mean differences of −0.01 logMAR (LoA of −0.21 to 0.19), 0.01 logMAR (LoA of −0.40 to 0.42), and 0.06 logMAR (LoA of −0.23 to 0.35) when compared with the ETDRS chart, respectively. The “Eye Chart” and “Peek Acuity” were tested among healthy adults, with a mean age of 64 and 65 years, respectively. The “Vision at Home” tool was performed by adolescents, adults, and elderly. Only the “Peek Acuity” tool was tested in the home and clinical environment, the other tools were tested in the controlled clinical environment. The tests were all performed using (habitual) spectacle correction. In comparison with the web-based tool, these visual acuity tests have an equivalent or better performance but were tested with a different methodology: The digital tests were all smartphone-based and were performed at different testing distances (the “Peek Acuity” and “Vision at Home” tool at 2 m testing distance and the “Eye Chart” at 4 feet [1.20 m] distance), and the visual acuities were scored using different methods (including letter-by-letter and line assignment) and not specifically assessed among patients who previously underwent cataract surgery.15–17 This might be an explanation for the discrepancies in outcomes compared with the web-based tool in this study. According to a systematic review, digital tools were in general less accurate in measuring visual acuity compared with conventional charts and showed wide distributions.18 *There is* consensus that outcomes of different visual acuity assessments vary.11 This variability is partly due to the psychophysical origin of the tests. Other reasons can be the design structure of the charts (decimal or logMAR), the optotypes used, the scoring methods, and the conditions under which the test is administered. Previous studies demonstrated that a decimal chart overestimates visual acuity compared with a logMAR chart.19 Concerning the scoring methods, the letter-by-letter method is more accurate compared with the line assignment method.20 *For this* study, the difference in scoring methods is presumably the primary factor that has caused some bias. The ETDRS chart is scored by the letter-by-letter method, the Snellen chart by the line assignment method, and the web-based tool by using an algorithm with a customized letter-by-letter method. Furthermore, the web-based tool has 7 optotypes in each line instead of 5. The abovementioned characteristics contribute to the general variability between visual acuity tests. If the postcataract pathway will represent a combination of both in-hospital and at home visual acuity tests, this should be taken into account. Nevertheless, a combination of these testing procedures can be very helpful because it offers flexibility for both patients and clinicians. Besides, the main aim of visual acuity testing after cataract surgery is a safety check for postoperative complications. In the case of a nonsignificant complication, an underestimation of the visual acuity up to 1.5 lines would be of lesser clinical relevance. Therefore, the variability of this visual acuity tool does not have a negative influence on the patient pathway but offers an additional screening opportunity.
In addition, the usage of different optotypes in assessments may affect the outcomes as well. Previous studies compared Landolt rings with numbers and showed higher visual acuity outcomes (0.13 ± 0.14 logMAR) using number optotypes.21 Other confirmed these lower outcomes when using Landolt rings compared with the Snellen (tumbling-)E chart or LEA symbols.21 In yet another study, there was no significant difference observed between visual acuity outcomes assessed by the Landolt and ETDRS chart among healthy and cataract eyes.22 These mentioned observations may have contributed to some discrepancies in our results. The web-based tool used tumbling-E optotypes, and the conventional charts had letter optotypes. Finally, the outcomes strongly depend on the achieved visual acuity of the tested subjects. Patients with the better scores tend to have more accurate visual acuity outcomes using the web-based tool.5 However, our outcomes did not confirm these findings, presumably due to the high overall visual acuity in our study population.
The limitation of this study was the fact that the web-based tool was only performed once. As a consequence, no test–retest or intraindividual consistency results were obtained from patients, which could have (partly) explained the variance between the gold standard and the web-based tool. Nevertheless, the web-based tool is considered to have a high test–retest reproducibility because of the nonvariable interpretation of patient responses by the tool.5 However, our additional test–retest analysis in healthy volunteers indicates that the variability of the web-based tool is up to twice as high compared with the conventional charts. Previous research showed a mean test–retest variability of the ETDRS and Snellen charts of 0.10 logMAR (LoA of −0.18 to 0.18) and −0.02 logMAR (LoA of −0.35 to 0.31), respectively.12 For assessing the intraindividual consistency among postcataract users of the web-based tool, further research is necessary. We can conclude that the differences among all 3 test outcomes in this study confirm the great variability generally observed.
Since the visual acuity assessments were performed under ideal conditions, it is expected that when patients perform the web-based tool in their home environment, outcomes may have both a lower reliability and greater variability. Other aspects, which may have influenced the outcomes, are the nonrandomized test sequence and duration of testing. The web-based tool was performed after the ETDRS chart assessment, which could have resulted in less accurate visual acuity outcomes using the web-based tool due to fatigue. In our results, no learning curve pattern could be demonstrated between first and second examined eyes. The subjects were blinded to the results to limit the performance bias. However, observation bias could not completely be prevented.
For implementation of a digital tool, practical issues must be taken into account. In this study, both UDVA and CDVA were evaluated. When performing an online visual acuity test in the home environment, the CDVA can only be measured after the patient has received his/her newly prescribed spectacles. Hence, directly after cataract surgery, only the UDVA measurements are applicable at home. Furthermore, elderly patients who undergo cataract surgery may not be able to perform digital tests unsupervised in their home environment. The introduction and usage of eHealth must always be in concordance with the patient. In addition, it should be noted that remote visual acuity testing will not completely replace ophthalmologic examination at the outpatient clinic but can enhance the efficiency of cataract care. Our results suggest that the web-based tool is useful in detecting larger changes in visual acuity but is probably not sensitive enough to reliably detect subtle changes.
Based on the results of this study, the web-based tool is validated for assessment of the visual acuity in patients who underwent cataract surgery. The web-based tool showed different outcomes compared with the conventional tests for both the UDVA and the CDVA, but most of these differences were within the established clinically acceptable limit of ±0.15 logMAR. These results are sufficient to introduce the web-based tool as a reliable screening method for detecting significant deterioration or lack of improvement of visual acuity in post-cataract patients. Our results suggest that the test can function as an interim assessment during the postoperative cataract care pathway. However, patients need to have basic digital skills to perform this web-based visual acuity assessment. Future research into this digital tool with a larger study population is necessary. WHAT WAS KNOWNThe Easee test has shown promising results for refraction and visual acuity measurements among healthy volunteers and patients with keratoconus. This web-based tool was not tested among patients who underwent cataract surgery before. WHAT THIS PAPER ADDSThe web-based tool is validated for assessment of the visual acuity in patients who underwent cataract surgery. The test is useful in detecting larger changes in visual acuity but is probably not sensitive enough to detect subtle changes.
## WHAT WAS KNOWN
The Easee test has shown promising results for refraction and visual acuity measurements among healthy volunteers and patients with keratoconus. This web-based tool was not tested among patients who underwent cataract surgery before.
## WHAT THIS PAPER ADDS
The web-based tool is validated for assessment of the visual acuity in patients who underwent cataract surgery. The test is useful in detecting larger changes in visual acuity but is probably not sensitive enough to detect subtle changes.
## References
1. Liu YC, Wilkins M, Kim T, Malyugin B, Mehta JS. **Cataracts**. *Lancet* (2017.0) **390** 600-612. PMID: 28242111
2. Behndig A, Cochener B, Güell JL, Kodjikian L, Mencucci R, Nuijts RM, Pleyer U, Rosen P, Szaflik JP, Tassignon MJ. **Endophthalmitis prophylaxis in cataract surgery: overview of current practice patterns in 9 European countries**. *J Cataract Refract Surg* (2013.0) **39** 1421-1431. PMID: 23988244
3. Kang S, Thomas PBM, Sim DA, Parker RT, Daniel C, Uddin JM. **Oculoplastic video-based telemedicine consultations: COVID-19 and beyond**. *Eye (Lond)* (2020.0) **34** 1193-1195. PMID: 32398851
4. Thompson-Coon J, Abdul-Rahman AK, Whear R, Bethel A, Vaidya B, Gericke CA, Stein K. **Telephone consultations in place of face to face out-patient consultations for patients discharged from hospital following surgery: a systematic review**. *BMC Health Serv Res* (2013.0) **13** 128. PMID: 23561005
5. Wisse RPL, Muijzer MB, Cassano F, Godefrooij DA, Prevoo YFDM, Soeters N. **Validation of an independent web-based tool for measuring visual acuity and refractive error (the manifest versus online refractive evaluation trial): prospective open-label noninferiority clinical trial**. *J Med Internet Res* (2019.0) **21** e14808. PMID: 31702560
6. **Individuals' level of digital skills**
7. **Digital Skills Indicator: derived from Eurostat survey on ICT usage by individuals**
8. McAlinden C, Khadka J, Pesudovs K. **Statistical methods for conducting agreement (comparison of clinical tests) and precision (repeatability or reproducibility) studies in optometry and ophthalmology**. *Ophthalmic Physiol Opt* (2011.0) **31** 330-338. PMID: 21615445
9. Martin Bland J, Altman DG. **Statistical methods for assessing agreement between two methods of clinical measurement**. *Lancet* (1986.0) **327** 307-310
10. Kaiser PK. **Prospective evaluation of visual acuity assessment: a comparison of snellen versus ETDRS charts in clinical practice (an AOS thesis)**. *Trans Am Ophthalmol Soc* (2009.0) **107** 311-324. PMID: 20126505
11. Siderov J, Tiu AL. **Variability of measurements of visual acuity in a large eye clinic**. *Acta Ophthalmol Scand* (1999.0) **77** 673-676. PMID: 10634561
12. Rosser DA, Laidlaw DA, Murdoch IE. **The development of a “reduced logMAR” visual acuity chart for use in routine clinical practice**. *Br J Ophthalmol* (2001.0) **85** 432-436. PMID: 11264133
13. Patton N, Aslam T, Murray G. **Statistical strategies to assess reliability in ophthalmology**. *Eye (Lond)* (2006.0) **20** 749-754. PMID: 16327799
14. Muijzer MB, Claessens JLJ, Cassano F, Godefrooij DA, Prevoo YFDM, Wisse RPL. **The evaluation of a web-based tool for measuring the uncorrected visual acuity and refractive error in keratoconus eyes: a method comparison study**. *PLoS One* (2021.0) **16** e0256087. PMID: 34407131
15. Tiraset N, Poonyathalang A, Padungkiatsagul T, Deeyai M, Vichitkunakorn P, Vanikieti K. **Comparison of visual acuity measurement using three methods: standard ETDRS chart, near chart and a smartphone-based eye chart application**. *Clin Ophthalmol* (2021.0) **15** 859-869. PMID: 33664563
16. Bastawrous A, Rono HK, Livingstone IAT, Weiss HA, Jordan S, Kuper H, Burton MJ. **Development and validation of a smartphone-based visual acuity test (peek acuity) for clinical practice and community-based fieldwork**. *JAMA Ophthalmol* (2015.0) **133** 930-937. PMID: 26022921
17. Han X, Scheetz J, Keel S, Liao C, Liu C, Jiang Y, Muller A, Meng W, He M. **Development and validation of a smartphone-based visual acuity test (vision at home)**. *Transl Vis Sci Technol* (2019.0) **8** 27
18. Claessens JLJ, Geuvers JR, Imhof SM, Wisse RPL. **Digital tools for the self-assessment of visual acuity: a systematic review**. *Ophthalmol Ther* (2021.0) **10** 715-730. PMID: 34169468
19. Mataftsi A, Koutsimpogeorgos D, Brazitikos P, Ziakas N, Haidich AB. **Is conversion of decimal visual acuity measurements to logMAR values reliable?**. *Graefes Arch Clin Exp Ophthalmol* (2019.0) **257** 1513-1517. PMID: 31069515
20. Vanden Bosch ME, Wall M. **Visual acuity scored by the letter-by-letter or probit methods has lower retest variability than the line assignment method**. *Eye (Lond)* (1997.0) **11** 411-417. PMID: 9373488
21. Rohrschneider K, Spittler AR, Bach M. **Comparison of visual acuity measurement with Landolt rings versus numbers [in German]**. *Ophthalmologe* (2019.0) **116** 1058-1063. PMID: 30927070
22. Kuo HK, Kuo MT, Tiong IS, Wu PC, Chen YJ, Chen CH. **Visual acuity as measured with Landolt C chart and Early Treatment of Diabetic Retinopathy Study (ETDRS) chart**. *Graefes Arch Clin Exp Ophthalmol* (2011.0) **249** 601-605. PMID: 20658145
|
---
title: The impact of maternal obesity on in vivo uterine contractile activity during
parturition in the rat
authors:
- Ronan Muir
- Raheela Khan
- Anatoly Shmygol
- Siobhan Quenby
- Matthew Elmes
journal: Physiological Reports
year: 2023
pmcid: PMC9981334
doi: 10.14814/phy2.15610
license: CC BY 4.0
---
# The impact of maternal obesity on in vivo uterine contractile activity during parturition in the rat
## Abstract
Maternal obesity is associated with increased risk of prolonged and dysfunctional labor and emergency caesarean section. To elucidate the mechanisms behind the associated uterine dystocia, a translational animal model is required. Our previous work identified that exposure to a high‐fat, high‐cholesterol (HFHC) diet to induce obesity down‐regulates uterine contractile associated protein expression and causes asynchronous contractions ex vivo. This study aims to investigate the impact of maternal obesity on uterine contractile function in vivo using intrauterine telemetry surgery. Virgin female Wistar rats were fed either a control (CON, $$n = 6$$) or HFHC ($$n = 6$$) diet for 6 weeks prior to conception, and throughout pregnancy. On Day 9 of gestation, a pressure‐sensitive catheter was surgically implanted aseptically within the gravid uterus. Following 5 days recovery, intrauterine pressure (IUP) was recorded continuously until delivery of the 5th pup (Day 22). HFHC induced obesity led to a significant 1.5‐fold increase in IUP ($$p \leq 0.026$$) and fivefold increase in frequency of contractions ($$p \leq 0.013$$) relative to CON. Determination of the time of labor onset identified that HFHC rats IUP ($$p \leq 0.046$$) increased significantly 8 h prior to 5th pup delivery, which contrasts to CON with no significant increase. Myometrial contractile frequency in HFHC rats significantly increased 12 h prior to delivery of the 5th pup ($$p \leq 0.023$$) compared to only 3 h in CON, providing evidence that labor in HFHC rats was prolonged by 9 h. In conclusion, we have established a translational rat model that will allow us to unravel the mechanism behind uterine dystocia associated with maternal obesity.
This is the first study to investigate the effect of dietary induced adiposity upon uterine contractile activity in vivo. Uterine contractile activity was measured by surgically implanting a pressure sensitive telemeter probe into the gravid uterus using aseptic techniques. Using this surgical technique, it was identified that feeding a high‐fat, high‐cholesterol diet to induce adiposity significantly prolonged labor compared to lean control fed animals.
## INTRODUCTION
Obesity rates in women of reproductive age are estimated to reach $50\%$ by 2050 (Butland et al., 2007) and the National Health Service (NHS) is on the cusp of a maternity care crisis as pregnancy complications will rise significantly. Maternal obesity substantially increases the risk of caesarean delivery, postpartum hemorrhage, longer duration of hospital stay, requirement for neonatal intensive care and stillbirth (CMACE, 2010; Heselhurst et al., 2008). Emergency caesarean delivery rates are significantly increased within the obese population (Poobalan et al., 2009), because of poor uterine contractile activity, and prolonged labor (Bogaerts et al., 2013; Kominiarek et al., 2011). With a single caesarean delivery costing £1530 more than a spontaneous vaginal delivery (NICE, 2011), and the significant health risks to the parturient during and after the procedure, the need to elucidate the etiology and mechanisms behind maternal obesity induced uterine dystocia is critical.
To unravel the mechanism behind prolonged and dysfunctional labor, we established a rat model of maternal obesity. We identified that dietary high‐fat, high‐cholesterol (HFHC) induced adiposity leads to asynchronous myometrial contractions in laboring uterine strips ex vivo (Muir et al., 2016). Furthermore, the un‐coordinated and dysfunctional myometrial contractions with maternal obesity were associated with adverse effects on contractile associated protein (CAP) expression. Obese rat dams exhibited reduced protein expression of the gap junction protein connexin 43 (cx43) that synchronizes myometrial contractions during labor (Elmes et al., 2011; Muir et al., 2016) and increased expression of phosphorylated cx43 (pcx43) (phosphorylation at serine 368) compared to lean controls. Phosphorylation of cx43 (pcx43) is negatively correlated with gap junction assembly and reduces cell‐to‐cell communication (Su & Lau, 2014). Increased cx43 phosphorylation in cardiomyocytes causes gap junction disassembly and reduced contractile activity in vitro (Huang et al., 2004). These observed differences in expression of cx43 and pcx43 in the laboring uterus could explain why our maternally obese rats exhibit asynchronous contractions. Steroids hormones estrogen and progesterone regulate myometrial expression of cx43, where progesterone suppresses myometrial gap junctions during pregnancy (Garfield et al., 1980; Hendrix et al., 1992) and estrogen administration during pregnancy induces premature gap junction formation and labor in the rat (MacKenzie & Garfield, 1986). The shift in protein expression of cx43 with maternal obesity may result from higher progesterone concentrations (Elmes et al., 2011) that limits the change to a contractile uterine milieu, resulting in compromised uterine contractility.
Despite the research in this field, our current knowledge is based entirely on in vitro and ex vivo studies, no study has yet investigated the effect of maternal obesity on term‐laboring uterine contractile activity in vivo. Research published by Pierce et al. [ 2010] illustrated that it is possible to utilize blood pressure telemetry systems to record the change in intrauterine pressure (IUP) during term and premature labor as an indirect measure of myometrial contractile activity. The approach has the potential to determine whether maternal obesity compromises uterine function in vivo. The aim of the study was to investigate the effects of dietary HFHC induced adiposity upon uterine contractility in vivo, through surgical implantation of telemetry pressure probes to measure IUP, to test the hypothesis that maternal obesity adversely affects uterine contractile function causing prolonged and dysfunctional labor. Establishing that our rat model of maternal obesity exhibits asynchronous myometrial contractions in vivo and exhibits prolonged labor can help unravel the mechanism behind prolonged and dysfunctional labor and identify potential dietary intervention or drug therapies to improve pregnancy and labor outcomes with maternal obesity.
## Ethics statement
All animal work was approved by the University of Nottingham Animal Welfare Ethical Review Board (AWERB) Approval Ref No 000055 and Home Office (PPL $\frac{40}{3598}$). All licensed procedures were carried out by licensed researchers under the Animals Scientific Procedures Act (ASPA) of 1986 within the animal facilities of the University of Nottingham. All investigators understand the ethical principles under which Physiological Reports operate, and all animal work complies with the ethics checklist.
## Animals
Twelve weanling virgin female Wistar rats (Rattus rattus) weighing 60 g (Charles River, UK) were pair housed under normal conditions (12 h light: dark photoperiod, 21 ± 5°C room temperature, $55\%$ ± $5\%$ relative humidity, food and water access ad libitum) and randomly assigned to be fed either a standard control laboratory chow (CON, $$n = 6$$) (Harlan Laboratories, UK) or HFHC ($$n = 6$$) diet as previously published (Elmes et al., 2011). Each rat was maintained on their respective diets for 6 weeks prior to mating with stud Wistar males (Charles River, UK) and throughout pregnancy. Pregnancy was confirmed via a successful mating with a stud male, and the time of appearance of a semen plug was recorded as gestational day 0.
## Surgical procedure
At Day 9 of gestation, a chronic use TA11‐C40 pressure sensitive transmitter (Data Science International, Mn, USA) was surgically implanted into the gravid uterus using aseptic techniques. Animals were placed under general anesthesia ($2\%$ Isoflurane $95\%$ Oxy/CO2), with pedal reflex and respiratory rate monitored to determine and maintain depth of anesthesia. Animals were administered preoperative buprenorphine (0.0168 mL/100 g) and meloxicam ($\frac{0.004}{100}$ g) subcutaneously to provide postoperative pain amelioration. The abdomen was prepared by shaving with clippers and swabbed with chlorhexidine and Viruscan (Cairn Technology, Sheffield, UK). Aseptically, a vertical incision was made from the xiphoid sternum to the bladder exposing the muscle layer; the muscle layer was opened following the linea alba. The gravid uterus was exteriorized to count the number of embryos within each uterine horn; the uterine horn with the greatest gravidity was selected for cannulation. Microscopically, a 2.5 mm incision was made into the uterus between the ovary and the first embryo. The pressure sensitive tip of the catheter was fed into the gravid uterus until it resided between the third and the fourth embryo. The catheter was secured using a few drops of Vet Bond (3 M, MN, USA). The cannulized horn was then interiorized back into the abdominal cavity, and the transmitter anchored to the interior of the muscle layer near the linea alba about 1 cm below the liver. The muscle layer was closed using a simple continuous suture pattern and finished with a muscle layer knot. The skin was closed using a simple continuous subcuticular stitch, with an Aberdeen knot to finish. The area around the wound was cleaned using chlorhexidine and dressed with Opsite (Opsite, London, UK). Animals were allowed to return to consciousness and provided with mash (water softened chow) for 24 h postoperatively, then maintained on their respective CON or HFHC diet for the rest of the trial. All rats recovered fully within 5 days of surgery, weight, food, and water intake daily, along with symptoms of pain or change in behavior recorded. Postoperative buprenorphine (0.0168 mL/100 g) and meloxicam($\frac{0.004}{100}$ g) were provided subcutaneously for up to 5 days to alleviate any postoperative pain or discomfort. Following a full recovery, the transmitted data were collected every second via radio‐telemetry over an AM frequency at 500 Hz to a DSI PhysioTel receiver pad (model RPC‐1). The DSI data exchange matrix would collect and store data via DSI Dataquest A.R.T Analysis software (version 4.33); changes in the local atmospheric pressure were accounted for by the DSI Ambient Pressure reference monitor (model APR‐1). Recording of intrauterine pressure continued until term delivery of the 5th pup where rat dams were euthanized using CO2 asphyxiation and cervical dislocation and pups by overdose of pentobarbitone. Blood was collected via cardiac puncture into EDTA coated tubes (Sarstedt, Nümbrecht, Germany), and subsequently centrifuged at 13,000 rpm to isolate the plasma which was flash frozen and stored at −80°C. The uterus, liver, kidneys, gonadal and perineal fat depots were dissected out, weighed, and snap frozen and stored at −80°C for future research. Euthanasia of rat dams after delivery of their 5th pup was an important component of the study protocol to standardize the point at which accurate and confident comparisons could be made between the CON and HFHC rats in established labor. It also meant that if the HFHC diet did compromise uterine contractile activity and length of labor as hypothesized there was appropriate time for it to be observed experimentally.
## Data acquisition and statistical analysis
All data were extracted using DSI Dataquest A.R.T Analysis software (version 4.33). Intrauterine pressure, integral activity, and data frequency counts calculated from extracted data using DSI Dataquest A.R.T Analysis software, Microsoft Excel (Microsoft, USA), and LabChart Reader version 8 (AD instruments, New Zealand). Frequency of contractions was determined by manually counting every single contractile event within each 1 h period from day 20 of gestation until term delivery of the 5th pup. All data were analyzed using the Statistical Package for Social Science version 21.0 (SPSS, Chicago, IL, USA) and expressed as the mean value ± SEM, with statistical significance determined by p ≤ 0.05. Excluding the frequency array counts for the number and strength of contractions, all data were analyzed using mixed between‐within ANOVA to determine the effect and interaction of maternal diet (CON/HFHC) and time to delivery of the 5th pup for fold‐change in intrauterine pressure and contraction frequency. As the data frequency counts were not normally distributed, they were analyzed by the nonparametric Mann–Whitney U‐test. Data frequency array counts determined the effect of maternal diet on the number of contractions within assigned pressure range bins (0–70 mmHg) at 48, 24, 12, 6, 3, and 1 h before birth of the 5th pup. All graphs were produced using GraphPad Prism version 6.0 (GraphPad, San Diego, CA, USA).
## Body and fat depot weight changes
Feeding the HFHC diet 6 weeks prior to conception to induce maternal obesity significantly increased pregestational weight gain compared to controls ($$p \leq 0.003$$). Controls gained 96.02 ± 4.8 g, and HFHC gained 122.74 ± 4.9 g (see Table 1). Weight gain during pregnancy was similar between the dietary treatment groups as was the weight at the end of pregnancy. Individual perirenal and gonadal fat depot weights were higher following exposure to the HFHC diet relative to controls, and the total visceral fat mass in HFHC rats was twice the level of controls at 10.4 ± 1.82 g compared to 5.87 ± 1.01 g ($$p \leq 0.054$$, see Table 1). Importantly, exposure to the HFHC diet and the effects of intrauterine telemetry surgery did not significantly affect litter size ($$p \leq 0.132$$), pup sex ratio ($$p \leq 0.804$$), pup weights ($$p \leq 0.126$$) or lead to significant differences in the number of fetal losses ($$p \leq 0.365$$) Table 1.
**TABLE 1**
| Category | Control (n = 6) | HFHC (n = 6) | p‐value |
| --- | --- | --- | --- |
| Pre‐gestational weight gain (g) | 96.02 ± 4.82 | 122.74 ± 4.91 | 0.003 |
| Gestational weight gain (g) | 131.3 ± 6.9 | 131.9 ± 10.82 | 0.963 |
| Final weight (g) | 351.82 ± 8.27 | 384.62 ± 16.4 | 0.105 |
| Perirenal fat (g) | 2.97 ± 0.53 | 5.3 ± 1.26 | 0.118 |
| Gonadal fat (g) | 2.9 ± 0.74 | 5.1 ± 0.91 | 0.09 |
| Total visceral fat (g) | 5.87 ± 1.01 | 10.4 ± 1.82 | 0.054 |
| Litter size | 11 ± 1.21 | 13.5 ± 0.92 | 0.132 |
| Offspring male: female ratio | 1.22 ± 0.38 | 1.37 ± 0.64 | 0.804 |
| Average pup weight (g) | 6.45 ± 0.11 | 6.06 ± 0.19 | 0.126 |
| Fetal losses | 4.33 ± 1.41 | 2.33 ± 0.365 | 0.365 |
## The effect of HFHC induced obesity on contractile activity in vivo
Visual analysis of raw intrauterine pressure (IUP) traces from representative CON and HFHC rats shows very clearly that HFHC induced obesity leads to greater uterine contractile activity during parturition (Figure 1). More detailed analysis from Day 20 of gestation right through to delivery of the 5th pup shows that animals fed the HFHC diet displayed significant increases in both IUP (Figure 2a) and contraction frequency (Figure 3a) relative to CON animals. The HFHC fed rats exhibited a 1.5‐fold increase in IUP ($$p \leq 0.026$$) and fivefold increase in frequency of uterine contractions ($$p \leq 0.013$$) from Day 20 of gestation until delivery of the 5th pup, relative to the CON animals that showed a 0.31 and 4.0‐fold increase in IUP and contraction frequency, respectively (Figures 2b and 3b). Regardless of which diet was fed, the frequency of contractions increased significantly as labor progressed towards delivery of the 5th pup and peaked during the final hours of labor (CON ($$p \leq 0.012$$) and HFHC ($$p \leq 0.016$$)). Maternal diet did not significantly affect integral activity ($$p \leq 0.760$$). However, integral activity did increase significantly as CON ($$p \leq 0.036$$) and HFHC rats ($$p \leq 0.05$$) progressed towards delivery (Figure 4).
**FIGURE 1:** *Representative traces of the change in intrauterine pressure (IUP) during the final 4 h approaching term delivery of the 5th pup at time point zero within (a) CON & (b) HFHC rats.* **FIGURE 2:** *Effect of feeding a HFHC diet upon (a) intrauterine pressure & (b) fold‐change in intrauterine pressure; end of −1 h marks birth of 5th pup. Group sizes for CON (n = 6) and HFHC (n = 6) diet animals. Values are means ± SEM. Black arrow indicates the point at which there is a significant and sustained increase in intrauterine pressure above baseline.* **FIGURE 3:** *Effect of feeding a HFHC diet upon (a) contraction frequency & (b) fold‐change in contraction frequency; end of −1 h marks birth of 5th pup. Group sizes for CON (n = 6) and HFHC (n = 6) diet animals. Values are means ± SEM. Arrows; [White (CON) and black (HFHC)] identify the point at which there is a significant and sustained increase above baseline.* **FIGURE 4:** *Effect of feeding a HFHC diet upon integral activity. Group sizes for CON (n = 6) and HFHC (n = 6) diet animals; end of −1 h marks birth of 5th pup. Values are means ± SEM.*
## Data frequency array counts
Data frequency array counts were carried out on IUP data collected at 48, 24, 12, 6, 3, and 1 h prior to delivery of the 5th pup and assigned into different 4 mmHg incremental bins ranging from 0 to 70 mmHg. Statistical analysis did not reveal a significant effect of diet or time on frequency counts because of high variability within the dietary groups. Despite this, the data still highlighted increased contractile activity and prolonged labor. Forty‐eight hours prior to delivery of the 5th pup, CON and HFHC animals displayed relatively low uterine contractile activity, with a high frequency of counts below an IUP of 18 mmHg (Figure 5). Twenty‐four hours later, the HFHC animals displayed an increased frequency of IUP counts above 18 mmHg relative to CON rats, this trend continued with HFHC animals displaying a rightward shift characterized by an increasing frequency of counts in higher IUP bins (>18 mmhg) at 12, 6, 3 and 1 h prior to delivery of the 5th pup. This shift continued in the HFHC rats where maximal contractions of 70 mmHg were observed 1‐h prior to delivery of the 5th pup. Unlike the HFHC animals, CON animals displayed relatively high frequency of counts below 18 mmHg throughout the same 48 h. Only at 1 h prior to delivery of the 5th pup did the frequency counts of CON rats exceed 18 mmHg where they reached the maximal mean peak of 58 mmHg.
**FIGURE 5:** *Effect of feeding a CON or HFHC diet upon data frequency array counts at (a) 48 h, (b) 24 h, (c) 12 h, (d) 6 h, (e) 3 h, and (f) 1 h prior to delivery of the 5th pup.*
## The effect of HFHC induced obesity on the duration of labor
With evidence that HFHC induced obesity caused a 1.5‐fold increase in IUP and a fivefold increase in frequency of contractions, we wanted to investigate the effect of maternal diet on the timing and length of labor. Statistical analysis by one‐way ANOVA repeated measures determined the point at which IUP and contractile frequency significantly increased and remained above baseline activity in both CON and HFHC fed rats. It is evident that HFHC rats displayed a significant increase in IUP 8 h prior to delivery of the 5th pup ($$p \leq 0.046$$), which was in complete contrast to CON rats that showed no significant increase in IUP (Figure 2b). Analysis of fold change in frequency of myometrial contractions identified that HFHC rats exhibited a significant and consistent increase above baseline 12 h prior to delivery of the 5th pup ($$p \leq 0.023$$). In comparison, CON rats only experienced a significant increase in contractile frequency 3 h prior to delivery (Figure 3b) suggesting that HFHC rats labored 9 h longer than CON. These data provide evidence that animals sustained on the HFHC diet display increased contractile activity for a longer duration relative to lean CON rats. This increase in contractile activity and duration of labor is symptomatic of a prolonged duration of labor.
## DISCUSSION
It is well established that obese parturients are at a significantly greater risk of exhibiting prolonged and dysfunctional labor and requiring emergency caesarean delivery (Bogaerts et al., 2013; Kominiarek et al., 2011; Poobalan et al., 2009); however, the etiology and mechanisms remain unresolved. Our previous research confirmed that exposure to a HFHC diet to induce adiposity had a significant negative effect on term‐laboring uterine contractility ex vivo, leading to un‐coordinated contractions and suppressed response to uterotonic oxytocin that is commonly used to augment labor (Muir et al., 2016). The key aim of the current study was to ascertain whether diet induced obesity also compromises in vivo uterine contractile activity during labor through surgical implantation of an intrauterine telemetry device as established by Pierce et al. [ 2010].
It is important to highlight that exposure to the HFHC diet to increase visceral white adipose tissue in the current study did not quite reach significance with a p‐value of 0.054, and differs to what we have shown previously (Muir et al., 2016). This is likely a result of the postoperative recovery period, when the rats utilized their fat as energy reserves during a period of reduced appetite and recuperation. An additional contributing factor that needs to be considered is the number of pups delivered before tissue collection, as animals in our previous trials were euthanized after delivery of the 1st pup, which may have a potential impact on their fat depot weights.
Using the novel surgical technique, we highlight a clear shift from relative uterine quiescence from gestational Day 20 to a contractile state by Day 22 of gestation. Relative to CON, the HFHC rats displayed significant increases in IUP and the number of uterine contractions. They also exhibited a higher frequency of stronger IUP data counts up to 70 mmHg from Day 20 of gestation to delivery of the 5th pup. The significant increase in uterine contractile activity provides evidence that the HFHC rats were in labor for 12 h before delivering their 5th pup compared to only 3 h in CON. This finding clearly indicates that diet induced obesity‐prolonged duration of labor and translates nicely to obese human pregnancies where prolonged labor is commonly observed (Bogaerts et al., 2013; Kominiarek et al., 2011).
Human clinical studies have established that increasing body mass is associated with prolonged duration of the first stage of labor, and in some cases a complete failure for labor to progress (Chin et al., 2012; Vahratian et al., 2005; Zhang et al., 2007).
Failure of labor to progress often stems from poor uterine contractile activity resulting in emergency caesarean delivery (Cedergren, 2009; Vahratian et al., 2005). Research has identified if labor is allowed to progress in obese parturients, all be it slowly to the 2nd stage of labor, they will produce a contraction profile similar to normal weight women and deliver vaginally (Buhimschi et al., 2004). This is similar to what was observed in the current study, where rats exposed to the HFHC diet suffered a prolonged duration of labor, but would in most instances be able to deliver to the 5th pup. Consistent with the data in the current study is the evidence from a recent clinical trial that maternal obesity causes higher basal tone of the uterine muscle and stronger contractions than leaner parturients during labor when measured by intrauterine pressure catheters, (Hautakangas et al., 2022). The suggested mechanism behind the increased basal uterine tone and stronger contractions with maternal obesity was potentially a difference in cervical dilation. Cervical tissue is connected to the muscle fibers of the uterine wall, and the status of the cervix can have an influence on the muscle tension of the uterus. A noncompliant or firmer cervix that has undergone less dilation, which may be the case with nulliparous obese pregnancies, may increase the tension of the uterus causing higher intrauterine pressure and strength of contractions (Hautakangas et al., 2022). Our previous work on the effects of diet induced obesity on ex vivo myometrial contractile activity, also identified an increased basal uterine tone compared to CON rats (Muir et al., 2016), again showing nice translation of our model to obese human pregnancy. It is pertinent to consider that the increased basal tone and contractile activity with obesity could be caused by physiological and molecular differences in the longitudinal or circular muscle layers of the myometrium. Intercellular communication through gap junctions is more intense in the circular with greater gene and protein expression of cx43 than the longitudinal layer of the bovine myometrium (Doualla‐Bell et al., 1995). Interestingly, treatment of the circular layer with ant‐estrogen inhibited expression of cx43; however, the effect was absent in the longitudinal layer, demonstrating that cx43 expression is differentially regulated in mycocytes from the circular and longitudinal layers (Doualla‐Bell et al., 1995). Future work could involve isolating the longitudinal and circular layers of the myometrium in our model to ascertain whether the increased basal tone, contractile strength and prolonged labor results from adverse changes to either muscle layer.
With a well‐established rat model of maternal obesity that exhibits prolonged labor, the potential mechanisms can now be unraveled to improve labor outcomes in obese human pregnancies. Ablation of murine uterine gene expression of cx43 by Tamoxifen has been observed to prolong labor by compromising synchronization of myometrial contractions (Döring et al., 2006). This finding matches the significant decrease in uterine expression of cx43 in the laboring uteri of HFHC rats (Elmes et al., 2011; Muir et al., 2016) that also exhibit asynchronous myometrial contractions ex vivo (Muir et al., 2016). It would be interesting to use the same intrauterine pressure approach to investigate whether inhibition of cx43 causes the same contractile phenotype as observed in model of maternal obesity. High circulating plasma concentrations of progesterone impede activation of uterine contractile activity by decreasing uterine expression of cx43, leading to poor synchronization of contractions during labor (Challis et al., 2000). Increased progesterone concentrations also negatively impact expression of key CAP's within the uterus including Cav‐1, OXTR and COX‐2 (Elmes et al., 2011) and limit expression of ion channels (Shmygol et al., 2007; Smith et al., 2005; Zhang et al., 2007), receptors (Gimpl & Fahrenholz, 2000, 2002; Klein et al., 1995), and uterotonin signaling (Myatt & Lye, 2004). Exposure to a HFHC diet significantly increases plasma cholesterol concentrations, that can inhibit myometrial contractile strength directly (Smith et al., 2005) or increase membrane fluidity compromising integral protein expression and down‐stream contractile signaling mechanisms (Gimpl & Fahrenholz, 2000; Pucadyil & Chattopadhyay, 2006; Shmygol et al., 2007). Disruption of the human ether‐a‐go‐go related gene (hERG) ion channel for example increases the duration of individual uterine contractions, promoting poor uterine contractility (Parkington et al., 2014). A further potential mechanism that could be detrimental to uterine contractile activity during parturition include increased synthesis of utero‐relaxant adipokines such as leptin by excess visceral white adipose tissue (Hehir & Morrison, 2012; Moynihan et al., 2006; Mumtaz et al., 2015). Despite all this knowledge, further research is required to elucidate the mechanisms behind uterine dystocia associated with maternal obesity. The rat model of maternal obesity presented in the current study would help to achieve this.
In summary, surgical implantation of the pressure sensitive telemetry system is a novel approach to determine whether chronic exposure to a HFHC diet had a negative effect upon IUP and uterine contractility during term labor in vivo. This study provides clear evidence that HFHC induced obesity‐prolonged labor duration significantly by 9 h compared to lean CON rats. These findings translate nicely to obese human pregnancies meeting the aims and hypothesis of the study. The data presented highlights the versatility of this dietary animal model and surgical procedure in investigating the mechanism(s) behind prolonged and dysfunctional labor associated with maternal obesity. The methods and model presented here could be applied to investigate any number of hypotheses governing uterine dystocia or the multifactorial interaction of the body systems to trigger and sustain parturition with maternal obesity.
## CONCLUSION
This is the first known study to investigate the effect of exposure to a HFHC diet to induce maternal obesity on uterine contractility in vivo, by surgically implanting a pressure‐sensitive catheter into the gravid uterus. Through this novel approach, we identified that exposure to the HFHC diet leads to a 9‐h prolongation of labor relative to lean CON rats. This study in combination with our previous published research now provides an established translational rat model of maternal obesity that can be utilized to investigate and identify the mechanism(s) through which maternal obesity compromises uterine contractility during labor using both ex vivo and now in vivo methods.
## AUTHOR CONTRIBUTIONS
Study design, management, and theater assistance was performed by Matthew Elmes. Daily running of the animal trial and all the surgical procedures carried out by Ronan Muir. The manuscript was written and reviewed by Ronan Muir and Matthew Elmes. The manuscript was reviewed by Ronan Muir, Raheela Khan, Anatoly Shmygol, Siobhan Quenby, and Matthew Elmes.
## FUNDING INFORMATION
This work was funded by the University of Nottingham BBSRC‐DTP PhD studentship, University Hospitals Coventry & Warwickshire NHS Trust studentship [grant number BB/J$\frac{014508}{1}$] and Rosetrees Trust [grant number M449].
## CONFLICT OF INTEREST STATEMENT
The authors have no conflicts of interest to report.
## AUTHOR'S TRANSLATIONAL PERSPECTIVE
This study and previous research by our group have highlighted the clear efficacy and translatability of this dietary animal model in the study of maternal obesity associated uterine dystocia. This dietary animal model could be used to investigate and identify the aberrant mechanism by which obesity compromises uterine contractility during parturition. This could lead to the development and understanding of pharmacological or dietary interventions to help curb the significantly high rates of caesarean delivery resulting from prolonged and dysfunctional labor associated with maternal obesity.
## References
1. Bogaerts A., Witters I., Van den Bergh B. R. H., Jans G., Devlieger R.. **Obesity in pregnancy: Altered onset and progression of labour**. *Midwifery* (2013) **29** 1303-1313. PMID: 23427851
2. Buhimschi C., Buhimschi I., Malinow A., Weiner C.. **Intrauterine pressure during the second stage of labour in obese women**. *Obstetrics and Gynecology* (2004) **103** 225-230. PMID: 14754688
3. Butland B., Jebb S., Kopelman P., McPherson K., Thomas S., Mardell J., Parry V.. **Foresight report—tackling obesities: future choices—project report**. (2007)
4. Cedergren M. I.. **Non‐elective caesarean delivery due to ineffective uterine contractility or due to obstructed labour in relation to maternal body mass index**. *European Journal of Obstetrics, Gynecology, and Reproductive Biology* (2009) **145** 163-166. PMID: 19525054
5. **Maternal obesity in the UK: Findings from a national project, pediatric diabetes**. (2010)
6. Challis J., Matthews S. G., Gibb W., Lye S. J.. **Endocrine and paracrine regulation of birth at term and preterm**. *Endocrine Reviews* (2000) **21** 514-550. PMID: 11041447
7. Chin J. R., Henry E., Holmgren C. M., Varner M. W., Branch D. W.. **Maternal obesity and contraction strength in the first stage of labor**. *American Journal of Obstetrics and Gynecology* (2012) **207** 129.e1-129.e6
8. Döring B., Shynlova O., Tsui P., Eckardt D., Janssen‐Bienhold U., Hofmann F., Feil S., Feil R., Lye S. J., Willecke K.. **Ablation of connexin43 in uterine smooth muscle cells of the mouse causes delayed parturition**. *Journal of Cell Science* (2006) **119** 1715-1722. PMID: 16595547
9. Doualla‐Bell F., Lye S. J., Fabrie F., Fortier M. A.. **Differential expression and regulation of connexin‐43 and cell‐cell coupling in myocytes from the circular and longitudinal layers of the bovine myometrium**. *Endocrinology* (1995) **136** 5322-5328. PMID: 7588277
10. Elmes M. J., Tan D. S., Cheng Z., Wathes D. C., Mcmullen S.. **The effects of a high‐fat, high‐cholesterol diet on markers of uterine contractility during parturition in the rat**. *Reproduction* (2011) **141** 283-290. PMID: 21078880
11. Garfield R. E., Kannan M. S., Daniel E. E.. **Gap junction formation in myometrium: Control by estrogens, progesterone, and prostaglandins**. *The American Journal of Physiology* (1980) **238** C81-C89. PMID: 7369350
12. Gimpl G., Fahrenholz F.. **Human oxytocin receptors in cholesterol‐rich vs. cholesterol‐poor microdomains of the plasma membrane**. *European Journal of Biochemistry* (2000) **267** 2483-2497. PMID: 10785367
13. Gimpl G., Fahrenholz F.. **Cholesterol as stabilizer of the oxytocin receptor**. *Biochimica et Biophysica Acta* (2002) **1564** 384-392. PMID: 12175921
14. Hautakangas T., Uotila J., Kontiainen J., Huhtala H., i Palomäki O.. **Impact of obesity on uterine contractile activity during labour: A blinded analysis of a randomised controlled trial cohort**. *BJOG: An International Journal of Obstetrics and Gynaecology* (2022) **129** 1790-1797. PMID: 35195337
15. Hehir M. P., Morrison J. J.. **The adipokine apelin and human uterine contractility**. *American Journal of Obstetrics and Gynecology* (2012) **206** 359.e1-359.e5
16. Hendrix E. M., Mao S. J., Everson W., Larsen W. J.. **Myometrial connexin 43 trafficking and gap junction assembly at term and in preterm labor**. *Molecular Reproduction and Development* (1992) **33** 27-38. PMID: 1324698
17. Heselhurst N., Simpson H., Ells L. J., Rankin J., Wilkinson J., Lang R., Brown T. J., Summerbell C. D.. **The impact of maternal BMI status on pregnancy outcomes with immediate short‐term obstetric resource implications: A meta‐analysis**. *Obesity Reviews* (2008) **9** 635-683. PMID: 18673307
18. Huang Y. S., Tseng Y. Z., Wu J. C., Wang S. M.. **Mechanism of oleic acid‐induced gap junctional disassembly in rat cardiomyocytes**. *Journal of Molecular and Cellular Cardiology* (2004) **37** 755-766. PMID: 15350848
19. Klein U., Gimpl G., Fahrenholz F.. **Alteration of the myometrial plasma membrane cholesterol content with beta‐cyclodextrin modulates the binding affinity of the oxytocin receptor**. *Biochemistry* (1995) **34** 13784-13793. PMID: 7577971
20. Kominiarek M. A., Zhang J., Vanveldhuisen P., Troendle J., Beaver J., Hibbard J. U.. **Contemporary labor patterns: The impact of maternal body mass index**. *American Journal of Obstetrics and Gynecology* (2011) **205** 244.e1-244.e8
21. MacKenzie L. W., Garfield R. E.. **Effects of 17 beta‐estradiol on myometrial gap junctions and pregnancy in the rat**. *Canadian Journal of Physiology and Pharmacology* (1986) **64** 462-466. PMID: 3730929
22. Moynihan A. T., Hehir M. P., Glavey S. V., Smith T. J., Morrison J. J.. **Inhibitory effect of leptin on human uterine contractility in vitro**. *American Journal of Obstetrics and Gynecology* (2006) **195** 504-509. PMID: 16647683
23. Muir R., Ballan J., Clifford B., McMullen S., Khan R., Shmygol A., Quenby S., Elmes M.. **Modelling maternal obesity: The effects of a chronic high–fat, high‐ cholesterol diet on uterine expression of contractile associated proteins and ex‐vivo contractile activity during labour in the rat**. *Clinical Science* (2016) **130** 183-192. PMID: 26543049
24. Mumtaz S., AlSaif S., Wray S., Noble K.. **Inhibitory effect of visfatin and leptin on human and rat myometrial contractility**. *Life Sciences* (2015) **25** 57-62
25. Myatt L., Lye S. J.. **Expression, localization and function of prostaglandin receptors in myometrium**. *Prostaglandins, Leukotrienes & Essential Fatty Acids* (2004) **70** 137-148. PMID: 14683689
26. **Caesarean section**. (2011)
27. Parkington H., Stevenson J., Tonta M., Paul J., Butler T., Maiti K., Chan E.‐C., Sheehan P., Brennecke S., Coleman H., Smith R.. **Diminished hERG K**. *Nature Communications* (2014) **5** 1-8
28. Pierce S. L., Kutschke W., Cabeza R., England S. K., Skarra D. V., Cornwell T., Solodushko V., Brown A., Taylor M. S.. **In vivo measurement of intrauterine pressure by telemetry: A new approach for studying parturition in mouse models in vivo measurement of intrauterine pressure by telemetry: A new approach for studying parturition in mouse models**. *Innovative Methodology* (2010) **42** 310-316
29. Poobalan a. S., Aucott L. S., Gurung T., Smith W. C. S., Bhattacharya S.. **Obesity as an independent risk factor for elective and emergency caesarean delivery in nulliparous women—systematic review and meta‐analysis of cohort studies**. *Obesity Reviews* (2009) **10** 28-35. PMID: 19021871
30. Pucadyil T. J., Chattopadhyay A.. **Role of cholesterol in the function and organization of G‐protein coupled receptors**. *Progress in Lipid Research* (2006) **45** 295-333. PMID: 16616960
31. Shmygol A., Noble K., Wray S.. **Depletion of membrane cholesterol eliminates the Ca**. *The Journal of Physiology* (2007) **581** 445-456. PMID: 17331986
32. Smith R. D., Babiychuk E. B., Noble K., Draeger A., Wray S.. **Increased cholesterol decreases uterine activity: Functional effects of cholesterol alteration in pregnant rat myometrium increased cholesterol decreases uterine activity: Functional effects of cholesterol alteration in pregnant rat myometrium**. *American Journal of Physiology‐Cell Physiology* (2005) **288** C982-C988. PMID: 15613497
33. Su V., Lau A. F.. **Connexins: Mechanisms regulating protein levels and intercellular communication**. *FEBS Letters* (2014) **588** 1212-1220. PMID: 24457202
34. Vahratian A., Siega‐Riz A. M., Savitz D. a., Zhang J.. **Maternal pre‐pregnancy overweight and obesity and the risk of cesarean delivery in nulliparous women**. *Annals of Epidemiology* (2005) **15** 467-474. PMID: 15921926
35. Zhang J., Kendrick A., Quenby S., Wray S.. **Contractility and calcium signaling of human myometrium are profoundly affected by cholesterol manipulation: Implications for labor?**. *Reproductive Sciences* (2007) **114** 456-466
|
---
title: Case report and literature review on a large MBC with ulceration
authors:
- Qin-Qin Luo
- Na-Na Luo
journal: Medicine
year: 2023
pmcid: PMC9981360
doi: 10.1097/MD.0000000000033131
license: CC BY 4.0
---
# Case report and literature review on a large MBC with ulceration
## Rationale:
Metaplastic breast cancer (MBC) is a rare tumor of the breast, and skin ulceration of breast tumors is a difficult clinical problem that reduces the patient’s quality of life.
### Patient concerns:
There are currently no Standard Treatment Guidelines for MBC at present, and the treatment for the skin ulceration of breast tumors is limited in clinics.
### Diagnosis:
Here, we report the case of a patient with a large MBC and skin ulceration, accompanied by exudation and odor.
### Intervention:
The combined treatment of albumin paclitaxel and carrelizumab (anti-PD-1 immunotherapy) was effective in reducing the tumor, but it increased the severity of the skin ulceration. After taking traditional Chinese medicine, the skin ulceration healed completely. Then the patient underwent a mastectomy and radiotherapy.
### Outcomes:
The patient has a high quality of life and was in good condition after the comprehensive treatment.
### Lessons:
This suggests that traditional Chinese medicine may have a good auxiliary therapeutic effect on the skin ulceration of MBC.
## 1. Introduction
At present, breast cancer is a malignant tumor with the highest incidence in the world, and is also the main cause of cancer-related death in women.[1] Metaplastic breast cancer (MBC) is a rare pathological type of breast cancer with obvious heterogeneity, and its morbidity is <$1\%$ of all breast cancers.[2,3] The most common clinical symptom of MBC is a painless large breast lump. When breast tumor cells proliferate and invade the skin, skin ulceration can occur, which is often accompanied by bleeding, pain, exudation, and odor. For serious skin ulcerations, skin grafting is necessary. Skin ulceration is a difficult clinical problem, which not only reduces patients’ quality of life (QoL) but also creates barriers to radical mastectomy.
In this study, we observed the perfect healing of skin ulceration in a patient with a large MBC following the adjuvant therapy of traditional Chinese medicine (TCM).
## 2. Case report
A young woman in her 30s visited a previous institute in April 2022 because she had noticed a right breast lump growing rapidly within a few months (Fig. 1A and B). Breast biopsy revealed a mixed tissue of invasive glandular carcinoma and multifocally distributed metaplastic carcinoma. Immunohistochemistry of the tumor confirmed the following: Glandular carcinoma: HER2 1+, estrogen receptor-negative, progesterone receptor (+,$2\%$), and Ki-67 = $60\%$; Metaplastic carcinoma: HER2-negative, estrogen receptor-negative, progesterone receptor-negative, and Ki-67 = $40\%$. Enhanced magnetic resonance imaging of the breast revealed a right breast tumor (87 × 66 × 63 mm) (Fig. 1C). Computed tomography of the head, chest, and abdomen, and whole-body bone scan indicated no distant metastasis. The results of the blood tests were normal, and the levels of carcinoembryonic antigen reached 46.0 ng/mL. The large breast tumor was painless and the small skin ulceration was accompanied by exudation and odor, which decreased her QoL. And her Eastern Cooperative Oncology Group performance score was 0.
**Figure 1.:** *(A and B) Appearance of the breast mass before the treatment. (C) Enhanced MRI of the breast before the treatment. MRI = magnetic resonance imaging.*
Initially, she received 3 cycles of epirubicin and cyclophosphamide chemotherapy, but the tumor did not reduce. Albumin paclitaxel and carrelizumab were administered on June 13, 2022. The tumor size was remarkablely reduced after the first cycle of chemotherapy (Fig. 2A). Therefore, the patient completed the other 5 cycles, and the exact times of the 6 cycles were (June 13, 2022, June 28, 2022, July 7, 2022, July 27, 2022, August 10, 2022, August 23, 2022). After the second cycle, the large neoplasm shrank remarkably, while the skin ulcers fused and became more serious, which was bleeding and odoriferous (Fig. 2B). She visited our institute and sought help from the TCM (July 15, 2022). According to her signs and symptoms, we provided a Chinese herbal medicine decoction (200 ml, twice a day), which had the effect of reinforcing qi and nourishing blood, the main Chinese herbs are described in Table 1. The skin ulceration healed gradually after the patient took the decoction (Fig. 2C). During the next 3 cycles of chemotherapy, she insisted on taking a Chinese herbal medicine decoction, which was adjusted slightly according to her symptoms, and the skin ulceration was gradually healed (Fig. 2D and E). After the last cycle of chemotherapy, the skin ulceration healed completely without any scarring (Fig. 2F). The primary lump disappeared and skin retraction was observed (Fig. 3A and B). Enhanced magnetic resonance imaging of the breast revealed that the breast lump had completely faded away (Fig. 3C). The patient’s QoL was improved and she was confident in accepting the following treatment since her skin ulceration began to heal. After the chemotherapy, the patient underwent mastectomy on September 12, 2022. Pathological examination revealed axillary sentinel lymph nodes metastasis ($\frac{2}{4}$) (ypT0N1M0). The surgical wound healed well and radiotherapy was completed. The patient was in good condition and was receiving treatment with carrelizumab (once every 3 weeks).
## 3. Discussion and review
MBC is a rare and unique type of breast cancer. According to the latest World Health Organization classification criteria for breast and female reproductive system tumors of World Health Organization, MBC can be divided into 6 subtypes: squamous cell carcinoma (SCC), spindle cell carcinoma, low-grade SCC, mesenchymal differentiated metaplasia carcinoma, fibromatosis metaplasia carcinoma, and mixed metaplasia carcinoma. Currently, the Vogts and Norris classification methods are widely used in the classification of MBC, which can be divided into 5 subtypes: spindle cell carcinoma, SCC, carcinosarcoma, stromal carcinoma, and osteoclast giant cell carcinoma.[3] The most common subtypes in the western world are spindle cell carcinoma and SCC in the eastern world. The pathology of this patient’s tumor was a heterogeneous tumor mixed with metaplastic carcinoma and adenocarcinoma, which is extremely rare in the clinic. The pathology of metaplastic cancer is SCC. The most common sign of MBC is painless breast lumps, but the incidence rate of lymph node metastasis is relatively low.[4–6] The pathological features of MBC are usually triple negative.[7–9] *In this* case, the breast tumor was large but painless and the pathological features were consistent with those reported in the literature.
Regarding the prognosis of metaplasia of breast cancer, it is generally believed that compared with triple-negative breast cancer, it has a worse prognosis because it is more invasive and more likely to have recurrence and distant metastasis. As most patients with MBC are triple negative, they cannot benefit from endocrine therapy or postoperative targeted therapy. Moreover, MBC has low sensitivity to chemotherapy. Although anthracyclines have been recommended to treat this special cancer in a previous study,[9] there is still no standard systematic treatment for MBC. In this case, there was no change in the breast lump after 3 cycles of epirubicin and cyclophosphamide chemotherapy (epirubicin + cyclophosphamide), which may be partly due to the heterogeneity of the patient’s tumor. The combination of albumin paclitaxel and immunotherapy had performed a positive effect, suggesting that this treatment may be an effective method for heterogeneous breast cancer with metastatic carcinoma.
Skin ulceration is a difficult clinical problem because of the continuous enlargement of breast cancer. Cancerous skin ulceration is often accompanied by bleeding, exudation, pain, and odor, which have a negative effect on the patients’ mental health and QoL. In severe cases, the skin ulceration cannot heal for a long time. Eventually, the necrotic tissue easily leads to sepsis due to bacteria or other infections, as well as death due to the rapid growth of tumors. Skin ulceration of breast cancer not only seriously affects patients’ mental health and QoL seriously, but also makes the treatment difficult. At present, anti-tumor therapies, such as chemotherapy, endocrine therapy, immunotherapy, targeted therapy, and surgery are the main treatment for skin ulceration. However, some patients with large tumors and severe ulcerations must undergo autologous skin grafting after surgical resection.[10] In addition to tumor cell proliferation, long-term chemotherapy, radiation therapy, laser therapy, and other treatments can also lead to or aggravate skin ulceration in breast tumors. For example, paclitaxel has been reported to cause skin ulceration.[11] Moreover, experimental studies have shown that paclitaxel can lead to skin ulcers in mice, and the severity of skin ulcers is positively correlated with the dose of paclitaxel.[12] At present, there are a few clinical studies on breast cancer with skin ulceration, and some case reports have been published. For example, a 31-year-old patient, who was diagnosed with advanced breast cancer with a skin ulcer (T4bN3bM0), received 4 cycles of cyclophosphamide, epirubicin and 5-FU chemotherapy and 4 cycles of paclitaxel combined with Target of Rapamycin. The patient achieved partial remission, and her QoL significantly improved. There was no recurrence 1 year after radical mastectomy.[13] A 68-year-old patient with advanced breast cancer and a huge skin ulcer received long-term combined treatment of trastuzumab (>400 times), aromatase inhibitors, and anti-cancer drugs. She achieved a progression-free survival of >9 years.[14] The QoL of an elderly patient with a huge breast tumor is seriously affected by a severe skin ulcer. After comprehensive treatment including chemotherapy, radiotherapy, surgery, and skin grafting, the breast tumor was removed and the serious skin ulcer was repaired. Although the tumor recurred 1 year later, the patient’s QoL was significantly improved.[15] Some scholars have used paclitaxel combined with bevacizumab as neoadjuvant chemotherapy to treat 2 cases of large breast tumors with skin ulceration. The skin ulcer was reduced and their QoL was improved after prolonged chemotherapy. They have successfully performed radical surgery at last.[16] In foreign countries, systemic anti-cancer therapy (such as chemotherapy) is the main treatment for breast cancer with skin ulceration, whereas, there are few local treatments for skin ulceration.
In China, systemic anti-cancer therapy remains the main treatment for breast cancer patients with skin ulceration, and local treatment for skin ulceration has also been explored. For example, a scholar applied a $25\%$ 5-FU (750 mg/day) solution wet dressing to treat a patient with a skin ulcer in a huge breast tumor, accompanied by systemic chemotherapy treatment. The skin ulcer was significantly improved after 2 cycles. The patient underwent radical surgery, and there was no recurrence or metastasis during the follow-up period of >2 years.[17] Han Yu[18] used local tissue implant radiotherapy to treat a huge skin ulcer in a breast cancer patient who had not received a therapeutic effect from the comprehensive treatment. Local radiotherapy reduced the tumor load in a relatively short time and improved the patient’s QoL. Wang Haimei[19] treated 30 breast cancer patients with skin ulceration using exterior coating nano-realgar powder, which was made from a Chinese herbal medicine realgar. Compared with the control group ($5\%$ cyclophosphamide solution wet dressing), the treatment group (nano-realgar exterior coating) promoted the healing of skin ulcers more effectively after 3 months of treatment, suggesting that the local application of new type of Chinese medicine may be an efficient method for the treatment of breast cancer with skin ulceration.
In this case, the tumor was significantly reduced after the treatment with albumin paclitaxel and carilizumab. However, the skin ulcers gradually fused into a larger wound with bleeding and odor, which reduced the patient’s QoL. Almost all the doctors told the patient that the skin ulcer wound not only could not heal but would become more serious during chemotherapy. The skin ulcer wound gradually healed after taking Chinese medicine, which greatly encouraged the patient. She told us that TCM was wonderful and that the magical efficacy increased her confidence in receiving the following treatment.
In this case, the prescription is based on the Danggui Buxue Decoction and Sijunzi Decoction. Danggui Buxue *Decoction is* composed of Astragalus and Angelicae, which is a classic formula for nourishing qi and blood. It has been confirmed that the main active ingredients of Astragalus and Angelicae can repair the oxidative damage of vascular endothelial cells and play a role in inflammation.[20,21] It has been found that Astragalus and Angelicae are the 2 most frequently used herbs in the treatment of ecthyma and diabetic foot ulcers by the method of data mining.[22,23] In the diabetic foot ulcer wound healing rat model, Danggui Buxue Decoction may promote vascular regulatory molecules expression of NO by increasing inducible NOS expression, which may promote vascular endothelial growth factor (VEGF) expression and accelerate wound healing simultaneously.[24] Danggui Buxue Decoction decoction can reduce the inflammatory response in diabetic foot ulcer rat models and promote neovascularization to accelerate wound healing.[25] Sijunzi *Decoction is* a classic prescription for replenishing qi in clinics and is composed of Radix Codonopsis, Atractylodes macrocephala, Poria cocos, and Licorice. Studies have shown that Sijunzi Decoction can promote the expression of the *Klotho* gene to inhibit the atrophy of the skin and subcutaneous adipose tissue, and improve the free radical metabolism by increasing the activity of superoxide dismutase, thereby protecting the skin.[26,27] In vivo, Sijunzi Decoction can accelerate angiogenesis and promote the growth of granulation tissue and wound Eepithelialization, thereby promoting wound healing by accelerating the proliferation of human fibroblasts and the expression of EGF, TGF-β, and VEGF.[28] *In a* mouse model of refractory ulcers, the Sijunzi Decoction may promote the growth of granulation tissue, and increase the expression of VEGF to accelerate wound angiogenesis, thus promoting wound healing.[29,30] Because it is rare and highly invasive, prospective research on MBC is rare. Most of the published data are derived from retrospective studies, and there are currently no Standard Treatment Guidelines for MBC at present. Therefore, it is still necessary to formulate the most appropriate treatment plan according to patients’ own conditions in clinical practice. In this case, the patient’s breast tumor was significantly reduced after the combination of albumin paclitaxel and immunotherapy, and the skin ulcer healed well after the adjuvant administration of TCM, suggesting that TCM may have a good adjuvant effect on patients with breast cancer accompanied by skin ulceration.
## Acknowledgments
We would like to thank all the staff and nurses for their kind cooperation. We would also like to thank the patient.
## Author contributions
Resources: Qin Qin Luo, Na Na Luo.
Supervision: Qin Qin Luo, Na Na Luo.
## References
1. Siegel RL, Miller KD, Fuchs HE. **Cancer statistics, 2022.**. *CA Cancer J Clin* (2022.0) **72** 7-33. PMID: 35020204
2. Vranic S, Stafford P, Palazzo J. **Molecular profiling of the metaplastic spindle cell carcinoma of the breast reveals potentially targetable biomarkers.**. *Clin Breast Cancer* (2020.0) **20** 326-331.e1. PMID: 32197944
3. El Zein D, Hughes M, Kumar S. **Metaplastic carcinoma of the breast is more aggressive than triple-negative breast cancer: a study from a single institution and review of literature.**. *Clin Breast Cancer* (2017.0) **17** 382-91. PMID: 28529029
4. Donato H, Candelária I, Oliveira P. **Imaging findings of metaplastic carcinoma of the breast with pathologic correlation.**. *J Belg Soc Radiol* (2018.0) **102** 46. PMID: 30039058
5. Alaoui M’hamdi H, Abbad F, Rais H. **Rare variant of metaplastic carcinoma of the breast: a case report and review of the literature.**. *J Med Case Rep* (2018.0) **12** 43. PMID: 29463294
6. Xiao M, Yang Z, Tang X. **Clinicopathological characteristics and prognosis of metaplastic carcinoma of the breast.**. *Oncol Lett* (2017.0) **14** 1971-8. PMID: 28781640
7. Bian T, Lin Q, Wu Z. **Metaplastic carcinoma of the breast: imaging and pathological features.**. *Oncol Lett* (2016.0) **12** 3975-80. PMID: 27895758
8. Ng CKY, Piscuoglio S, Geyer FC. **The landscape of somatic genetic alterations in metaplastic breast carcinomas.**. *Clin Cancer Res* (2017.0) **23** 3859-70. PMID: 28153863
9. Xu D, Hou L. **Clinicopathologic characteristics of mixed epithelial/mesenchymal metaplastic breast carcinoma (carcinosarcoma): a meta-analysis of Chinese patients.**. *Pol J Pathol* (2019.0) **70** 174-82. PMID: 31820860
10. Song D, Li Z, Zhou X. **Reconstruction of large complex wound after local advanced breast cancer resection.**. *Chin J Plast Surg* (2018.0) **34** 630-5
11. Bicher A, Levenback C, Burke TW. **Infusion site soft-tissue injury after paclitaxel administration.**. *Cancer* (1995.0) **76** 116-20. PMID: 8630862
12. Dorr RT, Snead K, Liddil JD. **Skin ulceration potential of paclitaxel in a mouse skin model in vivo.**. *Cancer* (1996.0) **78** 152-6. PMID: 8646711
13. Sakurai K, Fujisaki S, Matsuo S. **A case of advanced breast cancer with skin ulceration successfully treated with paclitaxel and toremifene therapy.**. *Gan To Kagaku Ryoho* (2009.0) **36** 2484-6. PMID: 20037463
14. Takeda Y, Tanaka N, Konishi J. **Combination of trastuzumab, aromatase inhibitor and anti-cancer drugs obtained a good prognosis for an inoperable stage III B breast cancer patient with giant skin ulceration.**. *Gan To Kagaku Ryoho* (2012.0) **39** 629-32. PMID: 22504690
15. Yamaguchi K, Matsunuma R, Hayami R. **Large breast tumor ulceration and QoL in an 80-year-old woman.**. *Case Rep Oncol* (2021.0) **14** 580-4. PMID: 33976637
16. Shinoda C, Mori R, Nagao Y. **Two cases of mastectomy after Paclitaxel + bevacizumab therapy for locally advanced breast cancer.**. *Case Rep Oncol* (2014.0) **7** 323-9. PMID: 24932175
17. Zhang X, Chen Q. **Topical treatment with 5-fluorouracil (5-FU) in breast tumor ulceration: a case report.**. *Mod Health Care Medical Innov Res* (2007.0) **4** 186-186
18. Han Y, Kang Tian-hua, Ren X-J. **A case of large breast tumor ulceration with liver cancer.**. *Chin J Exp Diagn* (2017.0) **21** 1845-7
19. Wang H, Liu Z. **Clinical observation on the treatment of breast tumor ulceration with nano realgar.**. *Emerg Chin Tradit Med* (2017.0) **26** 1427-9
20. Tan Y, Liu C, Zhu Q. **Protective effects of main active components combination of astragalus membranaceus and angelica sinensis on oxidative injury of human umbilical vein endothelial cells.**. *Chin J Integr Med Cardio-cerebrovasc Dis* (2020.0) **18** 4141-8
21. Wang W, Liu S, Qin L. **Research progress of pharmacological effect of astragalus and angelica on invigorating qi and activating blood circulation.**. *Chin J Exp Formulae* (2021.0) **27** 207-16
22. Shang A, Lin W, Yu C. **A study on the law of TCM medicine in the treatment of ecthyma based on data mining.**. *Clin J Chin Med* (2021.0) **13** 66-9
23. Xie F. *Study on the Medication Rule of Traditional Chinese Medicine in the Treatment of Diabetic Foot Based on Data Mining* (2022.0)
24. Huang Q, Chang J-X, Zheng S. **An experimental study on mechanism of danggui buxue decoction in promoting wound healing of diabetic foot ulcer in vitro.**. *Med Pharm J Chin PLA* (2022.0) **34** 12-5
25. Huang Q, Chang J-X, Zheng S. **Effect of Danggui Buxue decoction on angiogenesis and inflammatory reaction in rat models with diabetic foot.**. *Clin Misdiagnosis Mistherapy* (2021.0) **34** 101-6
26. Lei C. *Experimental Study on Anti-aging of “Jiawei Sijunzi Decoction”* (2013.0)
27. Li H, Gao J, Li J. **Effect of Sijunzi decoction on free radical metabolism in rats with ischemia-reperfusion injury.**. *Chin Arch Rraditional Chin Med* **2007** 293-4
28. Kong LZ. *Effects of Chinese Herbal Medicine for Supplementing Qi and Blood on Proliferation and Growth Factor Expression of HumanVSkin Fibroblasts in vitro* (2019.0)
29. Xing J, Que H. **Effects of qi- blood- enriching and spleen-nourishing herbs on intractable wounds in rats.**. *J Shanghai Univ Tradit Chin Med* **2008** 64-6
30. Que H, Xing J. **Effects of supplementing qi and blood on vascular endothelial growth factor and angiogenesis of chronic refractory wounds in rats.**. *J Integr Tradit Chin West Med* (2014.0) **33** 89-2
|
---
title: Epicardial fat density, coronary artery disease and inflammation in people
living with HIV
authors:
- Manel Sadouni
- Marie Duquet-Armand
- Mohamed Ghaiss Alkeddeh
- Mohamed El-Far
- Etienne Larouche-Anctil
- Cécile Tremblay
- Jean-Guy Baril
- Benoit Trottier
- Carl Chartrand-Lefebvre
- Madeleine Durand
journal: Medicine
year: 2023
pmcid: PMC9981370
doi: 10.1097/MD.0000000000032980
license: CC BY 4.0
---
# Epicardial fat density, coronary artery disease and inflammation in people living with HIV
## Abstract
Studies have shown an increased risk of coronary artery disease (CAD) in the human immunodeficiency virus (HIV) population. Epicardial fat (EF) quality may be linked to this increased risk. In our study, we evaluated the associations between EF density, a qualitative characteristic of fat, and inflammatory markers, cardiovascular risk factors, HIV-related parameters, and CAD. Our study was cross-sectional, nested in the Canadian HIV and Aging Cohort Study, a large prospective cohort that includes participants living with HIV (PLHIV) and healthy controls. Participants underwent cardiac computed tomography angiography to measure volume and density of EF, coronary artery calcium score, coronary plaque, and low attenuation plaque volume. Association between EF density, cardiovascular risk factors, HIV parameters, and CAD were evaluated using adjusted regression analysis. A total of 177 PLHIV and 83 healthy controls were included in this study. EF density was similar between the two groups (−77.4 ± 5.6 HU for PLHIV and −77.0 ± 5.6 HU for uninfected controls, $$P \leq .162$$). Multivariable models showed positive association between EF density and coronary calcium score (odds ratio, 1.07, $$P \leq .023$$). Among the soluble biomarkers measured in our study, adjusted analyses showed that IL2Rα, tumor necrosis factor alpha and luteizing hormone were significantly associated with EF density. Our study showed that an increase in EF density was associated with a higher coronary calcium score and with inflammatory markers in a population that includes PLHIV.
## 1. Introduction
The introduction of highly active antiretroviral therapy in the mid-90s has led to a decrease in the mortality related to human immunodeficiency virus (HIV) infection accompanied by an increase in life expectancy.[1,2] Individuals living with HIV have a higher risk to develop age-related diseases such as coronary artery disease (CAD) and several cohort studies have shown an increased risk of CAD in people living with HIV compared with the general population.[3–5] The pathophysiological processes underlying this increased risk are not well understood but may be linked to metabolic changes such as adipose tissue disturbances or inflammation.[6,7] Epicardial fat (EF), which is the visceral fat of the heart, can be considered as an active organ capable of secreting inflammatory cytokines that may have local and systemic effects on cardiovascular disease (CVD).[8,9] Several studies have shown that the amount of EF is increased in participants living with HIV (PLHIV), and that it is associated with a greater risk of CAD.[10–12] On the other hand, only a limited number of studies have explored how EF quality, beyond its quantity, can be related to inflammation and cardiovascular health among PLHIV.[13–15] Fat quality can be measured using fat computed tomography (CT) density (or CT attenuation, in Hounsfield units [HU]). This qualitative characteristic of fat can vary due to alterations in lipid content, tissue vascularity, and inflammation.[13–17] Studies performed in the general population have shown that changes in fat density may impact CAD risk independently of fat quantity.[13,18,19] *In this* study, we aimed to measure the associations between EF CT-measured density in both PLHIV and uninfected controls, and levels of inflammatory markers, cardiovascular risk factors, HIV-related parameters, and coronary artery plaque burden.
## 2.1. Study design and study population
We conducted a cross-sectional study nested within the Canadian HIV and Aging Cohort Study (CHACS), an ongoing multicenter prospective cohort following participants with and without HIV, with the aim of identifying the determinants of premature aging in people with HIV. Inclusion and exclusion criteria were previously reported.[20] We report results from the CHACS cardiovascular imaging sub-study, in which consecutive CHACS participants, aged 40 years or older, with a 10-year Framingham risk score between $5\%$ and $20\%$, without symptoms or history of CAD, with a preserved renal function and without allergy to contrast media were invited to undergo non-contrast cardiac CT and coronary CT angiography. This study was approved by the Institutional Review Board of the Medical Center of the University of Montreal and participating centers. All participants gave written informed consent.
## 2.2. Study procedures
Non-contrast cardiac CT and coronary CT angiography were performed using a 256-slice CT scanner (Brilliance iCT; Philips Healthcare, Best, The Netherlands). Prior to CT, participants were given 50 to 75 mg of metoprolol orally if heart rate was >60 beats/min, and 0.4 mg of nitroglycerin sublingually unless contraindicated.
Non-contrast cardiac CT was performed for coronary calcium scoring and EF evaluation using the following parameters: slice thickness 2.5 mm, matrix 512 × 512, field-of-view 250 mm, scan voltage 120 kV and prospective electrocardiographic-gating. CT angiography was performed for coronary plaque analysis using 370 mg/mL of iopamidol (Bracco Imaging, Milan, Italy) at a flow rate of 5 mL/s. Images were reconstructed using a hybrid iterative reconstruction algorithm (Philips iDose; Philips Healthcare, level 3).
## 2.3. Exposure of interest: EF density
EF is the fat deposit located between the visceral pericardium and the myocardium. EF density decreases with fat gain and increases with fat loss.[16,21] In addition, EF density increases with inflammatory cell infiltration and metabolic dysfunction.[14,16,22] Therefore, the higher the density, the more metabolically active the fat deposition might be.[13,23] EF density and volume were both assessed using non-contrast CT images and a semi-automated software (Aquarius Intuition version 4.4.11; TeraRecon Headquarters, Forster City, CA). The pericardium was manually traced on axial slices from the pulmonary artery bifurcation to the apex of the heart. A CT density between −190 and −30 HU was used to select the EF and exclude any other tissue. Mean EF density and volume were calculated using the software based on the adipose tissue area, the number of slices, slice thickness, and intersection gaps, and reported as a continuous value, expressed in HU and cm3.
Two observers both blinded to HIV status and clinical data measured EF volume and density using the semi-automated software. Inter-observer and intra-observer agreement for EF volume and density measurement were highly reproducible (intraclass correlation coefficient for inter-observer agreement 0.75 for EF volume and 0.99 for density; Intra-observer agreement 0.97 for EF volume and 0.97 for density).[24]
## 2.4. Outcome of interest: coronary plaque
Coronary artery calcium (CAC) measurement was performed using the method by Agatston et al[25] on non-contrast CT images. Coronary plaque analysis was performed using CT angiography images as previously described.[26] Plaque volume was quantified in multiplanar reformat. Total plaque, calcified, non-calcified and mixed plaque volumes per participant were defined as the sum of volumes of all plaques, calcified, non-calcified and mixed plaques per participant. Low-attenuation plaque, a marker of plaque vulnerability, was defined as the sum of volumes of all portions of 30 HU or less of each plaque per participant.
Inter-observer and intra-observer agreement for plaque volume analysis was excellent as previously described.[27]
## 2.5. Metabolic and inflammatory biomarkers
Eighty-eight inflammatory, anti-inflammatory and metabolic markers were measured in the CHACS cohort. These biomarkers were measured in plasma collected from the study participants using regular ELISA assays and Meso Scale technology as previously described.[28] *In this* study, we present results for 11 markers as these were the only one that were significantly associated to EF density in univariable analysis.
## 2.6. Statistical analysis
Continuous variables are reported as mean ± standard deviation or median [25th–75th interquartile range], as appropriate. Categorical variables are reported in frequency and percentages. Student t test or Mann–Whitney U test was used to compare normally and nonnormally distributed variables between participants with HIV and controls and chi-squared test was used to compare categorical variables.
The association of EF density with cardiovascular risk factors, as well as with HIV related parameters, was assessed using univariable and multivariable linear regression models, with and without inclusion of EF volume into the models. Multivariable model included all HIV-related parameters and cardiovascular risk factors risk factors significantly associated with EF density.
The association of EF density with coronary plaque (CAC, total, calcified, noncalcified and mixed plaque volume) was evaluated using zero-inflated Poisson regression, as previously described.[10] The use of zero-inflated model is made necessary by the excess of 0 in the distribution of total coronary plaque volumes. Coronary plaque volume variables were natural-log transformed prior to statistical analysis. Multivariable models were adjusted for CVD risk factors.
In addition and as an exploratory analysis, the association of EF with inflammatory biomarkers was evaluated using univariable and multivariable linear regression models. Multivariable models included inflammatory biomarkers and CVD risk factors if they showed a univariable association with EF. As this analysis was hypothesis-generating, we did multiple tests and did not correct for multiple comparisons.
Effect modification by HIV of each association was assessed by inclusion of an interaction term to the fully adjusted models, and significance of the interaction term was assessed by a likelihood ratio test.
For patients with incomplete data, the mean or median value was used to impute the missing data. Values for the following number of participants were missing and imputed: Body mass index (BMI),[5] smoking exposure,[14] high density lipoprotein-cholesterol,[4] low density lipoprotein-cholesterol,[9] antiretroviral therapy (ART) exposure duration,[7] non-nucleoside reverse transcriptase inhibitors exposure duration.[7] A P value <.05 was considered statistically significant. Statistical analyses were performed using R (version 3.3.2; R Foundation for Statistical Computing, Vienna, Austria).
## 3.1. Participant’s characteristics
Overall, 260 participants from the CHACS cohort were included in this study (177 ($68.1\%$) PLHIV and 83 ($31.9\%$) uninfected controls). Participant’s characteristics stratified by HIV status are listed in Table 1. PLHIV had a similar 10-year Framingham risk score (median of $10\%$ [7–16] for PLHIV and $9\%$ [7–15] for controls, $$P \leq .414$$). PLHIV were more likely to be male ($92.0\%$ for PLHIV vs $77.1\%$ for controls, $$P \leq .002$$), to smoke more (median smoking exposure of 6 [0–24.4] pack/years for PLHIV vs 0 [0–8.4] for controls, $$P \leq .001$$) and had a lower cholesterol level (mean total cholesterol of 4.9 ± 1.1 for PLHIV vs 5.3 ± 1.0 for controls, $$P \leq .003$$). PLHIV had a lower BMI compared to uninfected controls (mean of 25.3 ± 4.1 kg/m2 for PLHIV vs 27.1 ± 4.3 kg/m2 for controls, $$P \leq .002$$) while waist circumference was similar between the 2 groups (94.7 ± 11.5 for PLHIV vs 95.3 ± 10.6 for controls, $$P \leq .675$$).
**Table 1**
| Variables | HIV+ = 177 | HIV− = 83 | P value |
| --- | --- | --- | --- |
| CVD risk factors | | | |
| Age (yr) | 55.8 ± 7.0 | 56.5 ± 8.0 | .470 |
| Male sex, n (%) | 163 (92%) | 64 (77.1%) | .002 |
| Race, n (%) | Race, n (%) | Race, n (%) | Race, n (%) |
| Asian | 3 (1.7%) | 0 (0%) | .227 |
| Black | 17 (9.6%) | 2 (2.4%) | .227 |
| Caucasian | 147 (83%) | 77 (92.8%) | .227 |
| Hispanic | 9 (5.1%) | 4 (4.8%) | .227 |
| Diabetes, n (%) | 17 (9.6%) | 3 (3.6%) | .150 |
| High blood pressure, n (%) | 56 (31.6%) | 19 (22.9%) | .192 |
| Family history of premature CVD, n (%) | 37 (20.9%) | 18 (21.7%) | 1 |
| Smoking status, n (%) | Smoking status, n (%) | Smoking status, n (%) | Smoking status, n (%) |
| Current | 53 (29.9%) | 9 (10.8%) | .001 |
| Ex | 65 (36.7%) | 34 (41.0%) | .001 |
| Never | 55 (31.1%) | 39 (47.0%) | .001 |
| Smoking exposure (pack/year) | 6 [0–24.4] | 0 [0–8.4] | .001 |
| Total-Cholesterol (mmol/L) | 4.9 ± 1.1 | 5.3 ± 1.0 | .003 |
| HDL-cholesterol (mmol/L) | 1.2 ± 0.4 | 1.4 ± 0.4 | .001 |
| LDL-cholesterol (mmol/L) | 2.9 ± 0.9 | 3.2 ± 0.9 | .003 |
| Statin use, n (%) | 52 (29.4%) | 15 (18.1%) | .073 |
| BMI (kg/m2) | 25.3 ± 4.1 | 27.1 ± 4.3 | .002 |
| BMI categories | BMI categories | BMI categories | BMI categories |
| Underweight | 7 (3.9%) | 0 (0%) | .006 |
| Normal | 77 (43.5) | 23 (27.7%) | .006 |
| Overweight | 72 (40.7) | 41 (49.4%) | .006 |
| Obese | 21 (11.9) | 19 (22.9%) | .006 |
| Waist circumference (cm) | 94.7 ± 11.5 | 95.3 ± 10.6 | .675 |
| 10-year Framingham risk score (%) | 10 [7–16] | 9 [7–15] | .414 |
| HIV-specific variables | HIV+ = 177 | | P value |
| HIV infection duration (yr) | 18.3 ± 7.7 | | |
| Participants exposed to ART, n (%) | 166 (93.8%) | | |
| ART exposure duration (yr)* | 13.5 ± 6.5 | | |
| Participants exposed to PIs, n (%) | 132 (74.6%) | | |
| PIs exposure duration (yr)* | 9.5 ± 5.1 | | |
| Participants exposed to NRTIs, n (%) | 166 (93.8%) | | |
| NRTIs exposure duration (yr)* | 12.4 ± 6.2 | | |
| Participants exposed to NNRTIs, n (%) | 112 (62.3%) | | |
| NNRTIs exposure duration (yr)* | 4.5 [2.0–7.5] | | |
| Participants exposed to INSTIs, n (%) | 80 (45.2%) | | |
| IIs exposure duration (yr)* | 1.8 [0.9–3.8] | | |
| Detectable viral load,† n (%) | 15 (8.5%) | | |
| Viral load among detectable | 1135.0 [71.5–24,909.5] | | |
| Nadir CD4 cell count (cells/mm3) | 200 [100–300] | | |
| Current CD4 cell count (cells/mm3) | 611.7 ± 304.0 | | |
| Current CD8 cell count (cells/mm3) | 831.9 ± 417.2 | | |
| Inflammatory markers (N = 123) | HIV+ = 76 | HIV− = 47 | P value |
| IL2Ra | 711.5 [542.0–966.6] | 604.09 [470.23–812.01] | .080 |
| IL7 | 24.8 [19.9–30.0] | 19.4 [14.9–21.8] | <.001 |
| Leptin | 3510.8 [861.5–8770.2] | 3348.7 [979.1–7992.1] | .904 |
| Peptide C | 2115 [1399–3017] | 1757.5 [1220.6–3141.3] | .306 |
| Insulin | 17.3 [11.0–37.1] | 13.4 [9.2–27.8] | .194 |
| TNFα | 1.4 [1.1–2.0] | 1.2 [0.7–1.5] | .005 |
| ENA78 | 4272 [2965–5854] | 3667 [2387–5119] | .159 |
| GROα | 1644.5 [1257.1–1955.3] | 1299.2 [943.3–1628.9] | <.001 |
| FSH | 3581.0 [1855.6–6043.3] | 2393.1 [1604.0–4377.9] | .046 |
| LH | 2344.3 [1644.5–3062.4] | 1761.3 [1431.8–2746.6] | .089 |
| PP | 171.6 [103.7–303.4] | 200.6 [111.2–598.6] | .055 |
Subjects with HIV had a mean duration of HIV infection of 18.3 ± 7.7 years. More than $93\%$ of PLHIV were exposed to ART with a mean duration of 13.5 ± 6.5 years. $74.6\%$ were exposed to proteas inhibitors, $93.8\%$ where exposed to nucleoside reverse transcriptase inhibitors, $62.3\%$ were exposed to NNRTIs and $45.2\%$ were exposed to integrase strand transfer inhibitors. Viral load was detectable in $8.5\%$ of PLHIV and mean CD4 count was 611.7 ± 304.0 cells/mm3.
All levels of circulating inflammatory markers were similar between the 2 groups except for interleukine-7 (IL7), tumor necrosis factor alpha (TNFα), growth-regulated oncogene alpha and follicular stimulating hormone which were higher in PLHIV compared to controls (all $P \leq .05$).
PLHIV had a greater EF volume (113.4 ± 49.2 cm2) than uninfected controls (101.7 ± 43.7 cm2) ($$P \leq .046$$), whereas EF density was similar between the 2 groups (−77.4 ± 5.6 HU for PLHIV and −77.0 ± 5.6 HU for uninfected controls, $$P \leq .162$$). There was no significant difference in presence of coronary calcium ($66.1\%$ in PLHIV and $57.8\%$ in controls, $$P \leq .326$$) and in CAC score between the 2 groups (14.5 [0–127.8] for PLHIV and 8.3 [0–68.0] for controls, $$P \leq .196$$). At coronary CT angiography, total plaque volume showed no significant difference between PLHIV and controls (81.0 [0–357.9] for PLHIV and 40.7 [0–172.4] for controls, $$P \leq .052$$), however, PLHIV had a higher prevalence and volume of the non-calcified and mixed plaques compared to uninfected controls (prevalence of non calcified plaques for PLHIV $21.1\%$ vs $7.8\%$ for controls, $$P \leq .018$$, prevalence of mixed plaques for PLHIV $48.7\%$ vs $31.2\%$ for controls, $$P \leq .017$$). All cardiac CT results are presented in Table 2.
**Table 2**
| Non-contrast cardiac CT (N = 257) | HIV+ = 177 | HIV− = 83 | P value |
| --- | --- | --- | --- |
| Presence of coronary calcium | 117 (66.1%) | 48 (57.8%) | .326 |
| Coronary artery calcium score | 14.5 [0–127.8] | 8.3 [0–68.0] | .196 |
| Epicardial fat volume (cm2) | 113.8 ± 49.2 | 101.7 ± 43.7 | .046 |
| Epicardial fat density (HU) | −77.4 ± 5.6 | −77.0 ± 5.6 | .582 |
| Coronary CT angiography (N = 226) | HIV+ = 152 | HIV− = 77 | P value |
| Plaque presence | 106 (69.7%) | 46 (59.7%) | .172 |
| Total plaque volume (mm3) | 81.0 [0–357.9] | 40.7 [0–172.4] | .052 |
| Calcified plaque presence | 69 (45.5%) | 38 (49.3%) | .670 |
| Calcified plaque volume (mm3) | 0 [0–79.5] | 0 [0–95.7] | .795 |
| Non calcified plaque presence | 32 (21.1%) | 6 (7.8%) | .018 |
| Non calcified plaque volume (mm3) | 0 [0–0] | 0 [0–0] | .014 |
| Mixed plaque presence | 74 (48.7%) | 24 (31.2%) | .017 |
| Mixed plaque volume (mm3) | 0 [0–136.9] | 0 [0–51.9] | .010 |
| Low attenuation plaque volume (mm3) | 17.9 [0.0–97.7] | 6.4 [0.0–47.4] | .056 |
## 3.2. EF density and CVD risk factors
Univariable analysis showed that EF density was negatively associated with age, with fat density decreasing with advancing age (β = −0.13 per 1 year increase in age, $$P \leq .004$$), smoking exposure (β = −0.04 per each additional pack-years of exposure, $$P \leq .026$$), triglyceride (β = −0.80 per 1 mmol/L increase in triglyceride, $$P \leq .015$$), statin use (β = −2.25 for use vs non use, $$P \leq .004$$), BMI (β = −0.32 per 1 unit increase in BMI, $P \leq .001$) and EF volume (β = −0.09 for each increase in 1 cm3, $P \leq .001$) while a positive association was observed with male sex (β = 3.03, $$P \leq .003$$). In a multivariable regression model including all the traditional cardiovascular risk factors and EF volume, only male sex, diabetes and EF volume were independently associated with EF density (Table 3).
**Table 3**
| Unnamed: 0 | Univariable analysis | Univariable analysis.1 | Multivariable analysis* | Multivariable analysis*.1 | Multivariable analysis† | Multivariable analysis†.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | Beta‡ (95% CI) | P value | Beta‡ (95% CI) | P value | Beta‡ (95% CI) | P value |
| Age (per 1 year increase) | −0.13 (−0.23 to 0.04) | .004 | −0.13 (−0.23 to 0.04) | .006 | −0.03 (−0.10 to 0.03) | .294 |
| Male sex | 3.03 (1.02–5.05) | .003 | 2.60 (0.46–4.75) | .017 | 3.11 (1.64–4.58) | <.001 |
| HIV status | −0.41 (−1.87 to 1.05) | .582 | −1.04 (−2.55 to 0.47) | .174 | 0.16 (−0.88 to 1.21) | .756 |
| Diabetes | −1.73(−4.28 to 0.82) | .182 | 0.68 (−1.93 to 3.29) | .608 | 2.34 (0.54–4.15) | .011 |
| High blood pressure | 0.10 (−1.61 to 1.40) | .894 | 1.17 (−0.35 to 2.70) | .131 | 0.80 (−0.35 to 2.70) | .135 |
| Family history of premature CVD | −0.25 (−1.92 to 1.42) | .768 | 0.002 (−1.60 to 1.61) | .997 | 0.58 (−1.69 to 0.52) | .298 |
| Smoking exposure (per 1 pack/year increase) | −0.04 (−0.08 to 0.01) | .026 | −0.03 (−0.07 to 0.002) | .064 | 0.01 (−0.02 to 0.03) | .671 |
| HDL-cholesterol (per 1 mmol/L increase) | 0.17 (−1.68 to 2.02) | .857 | 0.32 (−1.64 to 2.27) | .748 | −0.02 (−1.36 to 1.32) | .976 |
| LDL-cholesterol (per 1 mmol/L increase) | 0.06 (−0.69 to 0.82) | .867 | −0.05 (−0.82 to 0.72) | .899 | −0.21 (−0.74 to 0.32) | .433 |
| Triglycerides (per 1 mmol/L increase) | −0.80 (−1.46 to 0.15) | .015 | −0.42 (−1.09 to 0.26) | .223 | −0.10 (−0.57 to 0.36) | .669 |
| Statin use | −2.25 (−3.78 to 0.71) | .004 | −1.46 (−3.05 to 0.13) | .071 | −0.48 (−1.58 to 0.61) | .382 |
| BMI (per 1 kg/m2 increase) | −0.32 (−0.48 to 0.16) | <.001 | −0.33 (−0.49 to 0.16) | <.001 | 0.11 (0.02–0.24) | .081 |
| Epicardial fat volume (per 1 cm3 increase) | −0.09 (−0.10 to 0.08) | <.001 | – | – | −0.09 (−0.10 to 0.08) | <.001 |
## 3.3. EF density and HIV-related parameters
In univariable models, EF density was only significantly associated with NNRTI exposure duration (β = −2.12, per 10 years increase of exposure, $$P \leq .031$$) while no association was observed with other ART classes, HIV duration, CD4 and CD8 count. In a multivariable regression model including all HIV-related parameters, CVD risk factors, and EF density, none of the HIV related parameters was significantly associated to EF density. Results are presented in Tables 4 and 5.
## 3.4. EF density and coronary plaque burden
Univariable analysis showed no significant association between EF density and coronary calcium score, total plaque volume, subtypes of plaque volume and low-attenuation plaque volume (all P ˃.05) (Table 6). Multivariable models adjusting for CVD risk factors showed positive association between EF density and calcium score (logit component odds ratio [OR], 1.07 per 1 HU increase in EF density, $$P \leq .023$$ and OR for Poisson regression = 1.00 per 1 HU increase in EF density, $$P \leq .635$$) while no association was observed with other measures of plaque. Additional adjustment for EF volume showed that the association between EF density and calcium score remained statistically significant (logit component OR, 1.09 per 1 HU increase in EF density, $$P \leq .039$$ and Poisson component OR 1.00 per 1 HU increase in EF density, $$P \leq .877$$) (Table 6).
**Table 6**
| Unnamed: 0 | Unadjusted analysis | Unadjusted analysis.1 | Adjusted analysis* | Adjusted analysis*.1 | Adjusted analysis† | Adjusted analysis†.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | Estimate (95% CI) | P value | Estimate (95% CI) | P value | Estimate (95% CI) | P value |
| Coronary artery calcium score‡ | L 1.02 (0.97–1.07) | .390 | L 1.07 (1.00–1.13) | .023 | L 1.09 (1.00–1.18) | .039 |
| Coronary artery calcium score‡ | P 0.99 (0.98–1.00) | .208 | P 1.00 (0.98–1.01) | .635 | P 1.00 (0.98–1.02) | .877 |
| Total plaque volume (cm3)§ | L 1.00 (0.95–1.05) | .948 | L 1.04 (0.98–1.10) | .202 | L 1.06 (0.95–1.15) | .201 |
| Total plaque volume (cm3)§ | P 0.99 (0.98–1.01) | .358 | P 1.00 (0.98–1.01) | .706 | P 1.01 (0.99–1.03) | .618 |
| Calcified plaque volume (cm3)∥ | L 1.00 (0.96–1.06) | .842 | L 1.03 (0.98–1.09) | .254 | L 1.05 (0.97–1.14) | .242 |
| Calcified plaque volume (cm3)∥ | P 0.99 (0.98–1.01) | .297 | P 0.99 (0.98–1.01) | .573 | P 1.00 (0.97–1.03) | .899 |
| Non calcified plaque volume (cm3)¶ | L 0.96 (0.90–1.03) | .252 | L 0.98 (0.91–1.05) | .492 | L 1.01 (0.91–1.12) | .842 |
| Non calcified plaque volume (cm3)¶ | P 1.01 (0.98–1.03) | .668 | P 1.01 (0.98–1.04) | .661 | P 1.02 (0.97–1.07) | .438 |
| Mixed plaque volume (cm3)# | L 1.00 (0.95–1.05) | .990 | L 1.05 (0.99–1.11) | .096 | L 1.07 (0.98–1.16) | .121 |
| Mixed plaque volume (cm3)# | P 1.00 (0.98–1.01) | .604 | P 1.00 (0.98–1.02) | .942 | P 1.01 (0.98–1.03) | .663 |
| Low attenuation plaque volume (cm3)** | L 0.99 (0.94–1.05) | .820 | L 1.03 (0.97–1.10) | .287 | L 1.05 (0.96–1.15) | .253 |
| Low attenuation plaque volume (cm3)** | P 0.99 (0.98–1.01) | .249 | P 1.00 (0.98–1.01) | .742 | P 1.01 (0.99–1.03) | .389 |
## 3.5. Association of EF density with metabolic and inflammatory biomarkers
Table 7 summarizes the results of the association of EF density with metabolic and inflammatory markers among 123 participants with available values (76 HIV+, 47 HIV−). Among the soluble biomarkers measured in our study, univariable regression analyses showed that only IL2Rα, IL7, leptin, peptid C, insulin, TNFα, EAN78, growth-regulated oncogene alpha, follicular stimulating hormone, luteizing hormone (LH) and PP were associated with EF density. After adjustment for CVD risk factors and EF volume, only IL2Rα, TNFα and LH remained significantly associated with EF density.
**Table 7**
| Unnamed: 0 | Unadjusted analysis | Unadjusted analysis.1 | Adjusted analysis* | Adjusted analysis*.1 | Adjusted analysis† | Adjusted analysis†.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | Beta (95% CI) | P value | Beta (95% CI) | P value | Beta (95% CI) | P value |
| IL2RA‡ | 0. 24 (0.03–0.45) | .024 | 0.28 (0.08–0.48) | .006 | 0.19 (0.05–0.34) | .010 |
| IL7 | −0.12 (−0.22 to 0.01) | .025 | −0.08 (−0.19 to 0.02) | .099 | −0.07 (−0.15 to 0.01) | .075 |
| Leptine‡ | −0.02 (−0.03 to 0.01) | <.001 | −0.02 (−0.03 to 0.01) | .004 | −0.01 (−0.02 to 0.00) | .078 |
| Peptid C‡ | −0.07 (−0.12 to 0.03) | .002 | −0.04 (−0.09 to 0.01) | .080 | −0.01 (−0.05 to 0.02) | .522 |
| Insuline | −0.04 (−0.07 to 0.01) | .003 | −0.02 (−0.05 to 0.01) | .146 | −0.01 (−0.03 to 0.01) | .395 |
| TNF alpha | 1.11 (0.12–2.10) | .029 | 1.48 (0.52–2.44) | .002 | 0.92 (0.23–1.63) | .010 |
| ENA78‡ | −0.03 (−0.07 to 0.001) | .045 | −0.02 (−0.06 to 0.01) | .162 | −0.01 (−0.04 to 0.01) | .207 |
| GRO alpha‡ | −0.19 (−0.34 to 0.03) | .016 | −0.16 (−0.31 to 0.00) | .045 | −0.08 (−0.20 to 0.03) | .165 |
| FSH‡ | 0.02 (0.00–0.04) | .049 | 0.02 (0.00–0.05) | .019 | 0.01 (0.00–0.03) | .055 |
| LH‡ | 0.04 (0.00–0.08) | .038 | 0.04 (0.00–0.08) | .038 | 0.03 (0.00–0.06) | .029 |
| PP‡ | 0.35 (0.05–0.65) | .021 | 0.27 (−0.03 to 0.59) | .075 | 0.15 (−0.08 to 0.38) | .197 |
## 4.1. Main findings of the study
In the current cross-sectional study, we showed that EF density did not differ by HIV status. EF density was positively associated with male sex, diabetes, and coronary calcium score, independent of cardiovascular risk factors and EF volume, but no association was found with HIV specific factors. Finally, we found that EF density was independently associated with some metabolic and inflammatory markers, mostly IL2Rα, TNFα, and LH.
## 4.2. EF density
PLHIV have a higher risk to develop CAD with subsequent complications, as shown in large cohort studies.[3–5] Mechanisms underlying this increased risk are not yet fully understood but several factors such as ART, abnormal fat deposition, inflammation or immune activation may play a role.
HIV infection and use of ART can lead to alterations in fat distribution.[29,30] We and others have previously showed that quantity of EF was increased in PLHIV compared to uninfected controls.[10,11,31] Studies in the general population as well as in PLHIV have demonstrated that greater quantity of EF was associated with a greater coronary atherosclerotic plaque burden.[11,32,33] EF secretes bioactive cytokines that are known to promote atherosclerosis.[8,34] Recently, assessment of EF quality has gained interest. However, only a few studies reported results specific to the HIV population.
## 4.3. EF density and HIV infection
Unexpectedly, PLHIV in our cohort did not have a significantly different EF density compared to healthy controls, despite having a higher volume of EF. As in our study, Buggey et al,[35] did not find a significant difference between HIV-positive and HIV-negative participants when comparing EF density measured in the left atrium roof and peri-right coronary artery. In another study evaluating density of abdominal visceral fat, fat density did not differ by HIV status.[36] EF inflammation and fibrosis may had alleviated the expected decrease in fat density and therefore explain the absence of difference in fat density between HIV-positive and HIV-negative participants despite the difference in EF volume.
In our study, we observed an inverse correlation between EF density and NNRTI exposure duration and positive association with detectable viral load in univariable analyses. However, these associations did not persist after adjustment for CVD risk factors and EF volume. In their study, Longenecker et al[37] found that baseline EF density and changes over time were inversely correlated with ART and duration of protease inhibitor exposure. Similarly, studies evaluating abdominal visceral fat demonstrated that exposure to ART was associated with a lower density of fat.[16,38,39] We and others have demonstrated that a longer exposure to ART and to some specific ART classes were associated with an increased EF volume.[10,11]
## 4.4. EF density and coronary plaque burden
To our knowledge, our study is the first to evaluate the association of EF density with coronary artery plaque burden in PLHIV. We found that EF density was positively associated with coronary calcium score in multivariable models adjusting for CVD risk factors and EF volume.
Our results are consistent with findings in the general population. In their studies, Pracon et al[40] and Liu et al[18] showed that EF density was positively correlated to coronary calcium score. Other studies demonstrated a positive association of EF density with coronary atherosclerosis severity as well as with myocardial infarction.[13,18,41] Higher EF radiodensity may reflect fat inflammation and fibrosis and represent a form of unfavorable metabolic activity which may directly affect coronary atherosclerosis. The association of EF density with coronary plaque burden was independent of EF volume suggesting that different mechanisms may account for clinical effect. Of note, some groups have reported opposite association between EF density and coronary plaque and showed that decreased density was associated with impaired CAD profile.[42–44] This may be explained by the inclusion of patients with different degrees of inflammation and fibrosis in EF.
Future mechanistic studies will need to be carefully designed to unravel the complex relationship between fat quality and CAD.
## 4.5. EF density and metabolic and inflammatory biomarkers
Fat abnormalities in PLHIV have been associated with metabolic dysregulation, inflammation, and immune activation. In our study, we evaluated a selected panel of metabolic and inflammatory markers and showed that only IL2Rα, TNFα and LH were positively associated with EF density independent of CVD risk factors and EF volume. IL2 receptor subunit alpha plays an important role in the proliferation and differentiation of T cells.[45,46] High levels of IL2Rα have been associated with multiple inflammatory diseases such as rheumatoid arthritis,[47] as well as with CAD.[48] TNFα represents one of the most potent pro-inflammatory cytokines. Studies have demonstrated that TNF-α has roles in the development of atherosclerosis.[49,50] Finally, recent findings demonstrated that adipose tissue regulates the LH secretion via leptin production.[51] Elevated levels of LH are associated with abnormalities predisposing to CVD risk such as insulin resistance, dyslipidemia, and systemic inflammation.[52,53] Our findings support the hypothesis that an abnormal EF quality may be linked to subclinical atherosclerosis through its underlying source of inflammatory mediators potentially favoring atherosclerosis. Consequently, CT measurements of quality of EF might add valuable information in the cardio-vascular risk assessment of HIV individuals.
Only few studies evaluated the correlation of fat density with inflammatory markers in PLHIV and found results comparable to ours. Longenecker et al,[37] and Chen et al,[54] showed that density of EF was negatively correlated to hs-CRP, IL6 and positively correlated T-cell activation. Similarly, other studies evaluating abdominal visceral fat density showed that it was negatively correlated to hs-CRP, IL6 and leptin level and positively correlated to adiponectin.[36,38]
## 4.6. Strengths and limitations
Our study has many strengths including its robust design, the use of CT that allowed us to characterize fat radiodensity in addition to volume and coronary plaque burden without additional exposition to radiation. We evaluated EF density in a population of individuals living with HIV and healthy controls, male and female and demonstrated for the first time that EF density was positively correlated to calcium coronary score in a population that includes PLHIV and to metabolic and inflammatory markers independent of EF volume. However, our study did not have sufficient power to assess whether this link was specific to PLHIV.
There are several limitations of our study: First, there may be significant regional variation of adipose tissue characteristics, particularly around atherosclerotic plaques. Future studies in the HIV population should consider more granular characterization of radiodensity by location. As we have previously reported, our study was conducted in participants who were predominantly male. It remains unknown whether these findings can be generalized to women. Although this does not bias our results, it limits their generalizability. Finally, this was a cross-sectional analysis which does not allow for determination of causality. Additionally, as a small group of participants had calcified, noncalcified or mixed plaques, the analyses on plaque sub-types may have lacked statistical power. This was also true for antiretroviral therapy, where a small group of participants were in each sub-class of ART. We performed a fair number of statistical tests, especially for the association of EF density and metabolic and inflammatory markers and we chose not to adjust p values for multiple comparisons which may have increase the likelihood of obtaining false positive results. But this was a hypothesis-generating study and further studies are needed to confirm our findings.
As PLHIV live longer, understanding the relationship between HIV, ART, and contributors to CAD risk is important. Our study shows that an increase in EF density is associated with a higher coronary calcium score in a population including PLHIV. An increase in EF density is also associated to metabolic and inflammatory dysfunction. Future measurement of EF density may provide additional insight into adipose tissue function beyond measurement of quantity alone and may have implications for assessment of long-term cardiovascular risk.
## Acknowledgements
We would like to thank all the staff and participants for their contribution in the actual study including: Stéphanie Matte, Annie Chamberland and Nathalie Bellavance.
## Author contributions
Conceptualization: Manel Sadouni, Cécile Tremblay, Carl Chartrand-Lefebvre, Madeleine Durand.
Data curation: Marie Duquet-Armand, Mohamed Ghaiss Alkeddeh, Etienne Larouche-Anctil, Jean-Guy Baril, Benoit Trottier, Carl Chartrand-Lefebvre, Madeleine Durand.
Formal analysis: Manel Sadouni.
Funding acquisition: Cécile Tremblay, Carl Chartrand-Lefebvre, Madeleine Durand.
Investigation: Madeleine Durand.
Methodology: Manel Sadouni, Cécile Tremblay, Carl Chartrand-Lefebvre, Madeleine Durand.
Supervision: Madeleine Durand.
Writing – original draft: Manel Sadouni.
Writing – review & editing: Manel Sadouni, Cécile Tremblay, Jean-Guy Baril, Carl Chartrand-Lefebvre, Madeleine Durand.
## References
1. Bhaskaran K, Hamouda O, Sannes M. **Changes in the risk of death after HIV seroconversion compared with mortality in the general population.**. *JAMA* (2008) **300** 51-9. PMID: 18594040
2. **Life expectancy of individuals on combination antiretroviral therapy in high-income countries: a collaborative analysis of 14 cohort studies.**. *Lancet* (2008) **372** 293-9. PMID: 18657708
3. Triant VA, Lee H, Hadigan C. **Increased acute myocardial infarction rates and cardiovascular risk factors among patients with human immunodeficiency virus disease.**. *J Clin Endocrinol Metab* (2007) **92** 2506-12. PMID: 17456578
4. Durand M, Sheehy O, Baril JG. **Association between HIV infection, antiretroviral therapy, and risk of acute myocardial infarction: a cohort and nested case-control study using Québec’s public health insurance database.**. *J Acquir Immune Defic Syndr* (2011) **57** 245-53. PMID: 21499115
5. Islam FM, Wu J, Jansson J. **Relative risk of cardiovascular disease among people living with HIV: a systematic review and meta-analysis.**. *HIV Med* (2012) **13** 453-68. PMID: 22413967
6. Vos AG, Idris NS, Barth RE. **Pro-inflammatory markers in relation to cardiovascular disease in HIV infection. A systematic review.**. *PLoS One* (2016) **11** e0147484. PMID: 26808540
7. Palella FJ, McKibben R, Post WS. **Anatomic fat depots and coronary plaque among human immunodeficiency virus-infected and uninfected men in the multicenter AIDS cohort study.**. *Open Forum Infect Dis* (2016) **3** ofw098. PMID: 27419170
8. Mazurek T, Zhang L, Zalewski A. **Human epicardial adipose tissue is a source of inflammatory mediators.**. *Circulation* (2003) **108** 2460-6. PMID: 14581396
9. Baker AR, Silva NF, Quinn DW. **Human epicardial adipose tissue expresses a pathogenic profile of adipocytokines in patients with cardiovascular disease.**. *Cardiovasc Diabetol* (2006) **5** 1. PMID: 16412224
10. Sadouni M, Durand M, Boldeanu I. **Association of epicardial fat with non-calcified coronary plaque volume and with low attenuation plaque in people living with HIV.**. *AIDS* (2021) **35** 1575-84. PMID: 33831908
11. Brener M, Ketlogetswe K, Budoff M. **Epicardial fat is associated with duration of antiretroviral therapy and coronary atherosclerosis.**. *AIDS* (2014) **28** 1635-44. PMID: 24809732
12. Raggi P, Zona S, Scaglioni R. **Epicardial adipose tissue and coronary artery calcium predict incident myocardial infarction and death in HIV-infected patients.**. *J Cardiovasc Comput Tomogr* (2015) **9** 553-8. PMID: 26310588
13. Konishi M, Sugiyama S, Sato Y. **Pericardial fat inflammation correlates with coronary artery disease.**. *Atherosclerosis* (2010) **213** 649-55. PMID: 21040916
14. Baba S, Jacene HA, Engles JM. **CT Hounsfield units of brown adipose tissue increase with activation: preclinical and clinical studies.**. *J Nucl Med* (2010) **51** 246-50. PMID: 20124047
15. Hu HH, Chung SA, Nayak KS. **Differential computed tomographic attenuation of metabolically active and inactive adipose tissues: preliminary findings.**. *J Comput Assist Tomogr* (2011) **35** 65-71. PMID: 21245691
16. Lake JE, Moser C, Johnston L. **CT fat density accurately reflects histologic fat quality in adults with HIV on and off antiretroviral therapy.**. *J Clin Endocrinol Metab* (2019) **104** 4857-64. PMID: 31329901
17. Gifford A, Towse TF, Walker RC. **Characterizing active and inactive brown adipose tissue in adult humans using PET-CT and MR imaging.**. *Am J Physiol Endocrinol Metab* (2016) **311** E95-E104. PMID: 27166284
18. Liu Z, Wang S, Wang Y. **Association of epicardial adipose tissue attenuation with coronary atherosclerosis in patients with a high risk of coronary artery disease.**. *Atherosclerosis* (2019) **284** 230-6. PMID: 30777338
19. Goeller M, Achenbach S, Cadet S. **Pericoronary adipose tissue computed tomography attenuation and high-risk plaque characteristics in acute coronary syndrome compared with stable coronary artery disease.**. *JAMA Cardiol* (2018) **3** 858-63. PMID: 30027285
20. Durand M, Chartrand-Lefebvre C, Baril JG. **The Canadian HIV and aging cohort study – determinants of increased risk of cardio-vascular diseases in HIV-infected individuals: rationale and study protocol.**. *BMC Infect Dis* (2017) **17** 611. PMID: 28893184
21. Murphy RA, Register TC, Shively CA. **Adipose tissue density, a novel biomarker predicting mortality risk in older adults.**. *J Gerontol A Biol Sci Med Sci* (2014) **69** 109-17. PMID: 23707956
22. **Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016.**. *Lancet* (2017) **390** 1211-59. PMID: 28919117
23. Divoux A, Tordjman J, Lacasa D. **Fibrosis in human adipose tissue: composition, distribution, and link with lipid metabolism and fat mass loss.**. *Diabetes* (2010) **59** 2817-25. PMID: 20713683
24. Sadouni M, Boldeanu I, Durand M. **Quantification of epicardial fat using non contrast cardiac CT in an HIV population: reproducibility and association with other body fat indices.**. *Eur J Radiol Open* (2021) **8** 100317. PMID: 33490311
25. Agatston AS, Janowitz WR, Hildner FJ. **Quantification of coronary artery calcium using ultrafast computed tomography.**. *J Am Coll Cardiol* (1990) **15** 827-32. PMID: 2407762
26. Boldeanu I, Sadouni M, Mansour S. **Prevalence and characterization of subclinical coronary atherosclerotic plaque with CT among Individuals with HIV: results from the Canadian HIV and aging cohort study.**. *Radiology* (2021) **299** 571-80. PMID: 33876969
27. Chen Z, Boldeanu I, Nepveu S. **In vivo coronary artery plaque assessment with computed tomography angiography: is there an impact of iterative reconstruction on plaque volume and attenuation metrics?**. *Acta Radiol* (2017) **58** 660-9. PMID: 27650033
28. El-Far M, Durand M, Turcotte I. **Upregulated IL-32 expression and reduced gut short chain fatty acid caproic acid in people living with HIV with subclinical atherosclerosis.**. *Front Immunol* (2021) **12** 664371. PMID: 33936102
29. Alikhani A, Morin H, Matte S. **Association between lipodystrophy and length of exposure to ARTs in adult HIV-1 infected patients in Montreal.**. *BMC Infect Dis* (2019) **19** 820. PMID: 31533648
30. Giralt M, Domingo P, Guallar JP. **HIV-1 infection alters gene expression in adipose tissue, which contributes to HIV-1/HAART-associated lipodystrophy.**. *Antivir Ther* (2006) **11** 729-40. PMID: 17310817
31. Lo J, Abbara S, Rocha-Filho JA. **Increased epicardial adipose tissue volume in HIV-infected men and relationships to body composition and metabolic parameters.**. *AIDS* (2010) **24** 2127-30. PMID: 20588167
32. Mahabadi AA, Berg MH, Lehmann N. **Association of epicardial fat with cardiovascular risk factors and incident myocardial infarction in the general population: the Heinz Nixdorf Recall Study.**. *J Am Coll Cardiol* (2013) **61** 1388-95. PMID: 23433560
33. Alexopoulos N, McLean DS, Janik M. **Epicardial adipose tissue and coronary artery plaque characteristics.**. *Atherosclerosis* (2010) **210** 150-4. PMID: 20031133
34. Margaritis M, Antonopoulos AS, Digby J. **Interactions between vascular wall and perivascular adipose tissue reveal novel roles for adiponectin in the regulation of endothelial nitric oxide synthase function in human vessels.**. *Circulation* (2013) **127** 2209-21. PMID: 23625959
35. Buggey J, Yun L, Hung CL. **HIV and pericardial fat are associated with abnormal cardiac structure and function among Ugandans.**. *Heart* (2020) **106** 147-53. PMID: 31537637
36. Lake JE, Debroy P, Ng D. **Associations between subcutaneous fat density and systemic inflammation differ by HIV serostatus and are independent of fat quantity.**. *Eur J Endocrinol* (2019) **181** 451-9. PMID: 31430720
37. Longenecker CT, Margevicius S, Liu Y. **Effect of pericardial fat volume and density on markers of insulin resistance and inflammation in patients with human immunodeficiency virus infection.**. *Am J Cardiol* (2017) **120** 1427-33. PMID: 28822563
38. Debroy P, Lake JE, Moser C. **Antiretroviral therapy initiation is associated with decreased visceral and subcutaneous adipose tissue density in people living with human immunodeficiency virus.**. *Clin Infect Dis* (2021) **72** 979-86. PMID: 32107532
39. Gelpi M, Knudsen AD, Larsen KB. **Long-lasting alterations in adipose tissue density and adiponectin production in people living with HIV after thymidine analogues exposure.**. *BMC Infect Dis* (2019) **19** 708. PMID: 31399063
40. Pracon R, Kruk M, Kepka C. **Epicardial adipose tissue radiodensity is independently related to coronary atherosclerosis. A multidetector computed tomography study.**. *Circ J* (2011) **75** 391-7. PMID: 21178296
41. Mahabadi AA, Balcer B, Dykun I. **Cardiac computed tomography-derived epicardial fat volume and attenuation independently distinguish patients with and without myocardial infarction.**. *PLoS One* (2017) **12** e0183514. PMID: 28837682
42. Franssens BT, Nathoe HM, Visseren FL. **Relation of epicardial adipose tissue radiodensity to coronary artery calcium on cardiac computed tomography in patients at high risk for cardiovascular disease.**. *Am J Cardiol* (2017) **119** 1359-65. PMID: 28279438
43. Abazid RM, Smettei OA, Kattea MO. **Relation between epicardial fat and subclinical atherosclerosis in asymptomatic individuals.**. *J Thorac Imag* (2017) **32** 378-82
44. Goeller M, Achenbach S, Marwan M. **Epicardial adipose tissue density and volume are related to subclinical atherosclerosis, inflammation and major adverse cardiac events in asymptomatic subjects.**. *J Cardiovasc Comput Tomogr* (2018) **12** 67-73. PMID: 29233634
45. Malek TR, Castro I. **Interleukin-2 receptor signaling: at the interface between tolerance and immunity.**. *Immunity* (2010) **33** 153-65. PMID: 20732639
46. Gillis S, Smith KA. **Long term culture of tumour-specific cytotoxic T cells.**. *Nature* (1977) **268** 154-6. PMID: 145543
47. Pettersson T, Söderblom T, Nyberg P. **Pleural fluid soluble interleukin 2 receptor in rheumatoid arthritis and systemic lupus erythematosus.**. *J Rheumatol* (1994) **21** 1820-4. PMID: 7837144
48. Durda P, Sabourin J, Lange EM. **Plasma levels of soluble interleukin-2 receptor α: associations with clinical cardiovascular events and genome-wide association scan.**. *Arterioscler Thromb Vasc Biol* (2015) **35** 2246-53. PMID: 26293465
49. Zhang Y, Yang X, Bian F. **TNF-α promotes early atherosclerosis by increasing transcytosis of LDL across endothelial cells: crosstalk between NF-κB and PPAR-γ.**. *J Mol Cell Cardiol* (2014) **72** 85-94. PMID: 24594319
50. Monaco C, Nanchahal J, Taylor P. **Anti-TNF therapy: past, present and future.**. *Int Immunol* (2015) **27** 55-62. PMID: 25411043
51. Childs GV, Odle AK, MacNicol MC. **The importance of leptin to reproduction.**. *Endocrinology* (2021) **162**
52. Moran LJ, Misso ML, Wild RA. **Impaired glucose tolerance, type 2 diabetes and metabolic syndrome in polycystic ovary syndrome: a systematic review and meta-analysis.**. *Hum Reprod Update* (2010) **16** 347-63. PMID: 20159883
53. Toulis KA, Goulis DG, Mintziori G. **Meta-analysis of cardiovascular disease risk markers in women with polycystic ovary syndrome.**. *Hum Reprod Update* (2011) **17** 741-60. PMID: 21628302
54. Chen M, Hung CL, Yun CH. **Sex differences in the association of fat and inflammation among people with treated HIV infection.**. *Pathog Immun* (2019) **4** 163-79. PMID: 31508536
|
---
title: 'Antiepileptic therapy in a patient with star fruit intoxication: A case report'
authors:
- Aixun Li
- Baoxin Chen
- Xianglan Jin
- Yu Bai
- Jingfeng Zhang
- Chengcheng Zhang
- Miaomiao Cheng
- Chunyan Guo
- Yu Zhang
- Jing Zhou
journal: Medicine
year: 2023
pmcid: PMC9981371
doi: 10.1097/MD.0000000000032969
license: CC BY 4.0
---
# Antiepileptic therapy in a patient with star fruit intoxication: A case report
## Rationale:
It is rare for uremia patients to have epileptic seizures after eating star fruit, only a dozen cases are reported worldwide. Such patients usually have poor prognoses. Few patients had good prognoses, all of them were treated with expensive renal replacement therapy. At present, there is no report on the addition of drug therapy to these patients based on the initial renal replacement therapy.
### Patient concerns:
A 67-year-old male patient with star fruit intoxication who had a history of diabetic nephropathy, hypertension, polycystic kidney, and chronic kidney disease in the uremic phase, and regular hemodialysis 3 times a week for 2 years. Initial clinical manifestations include hiccups, vomiting, speech disturbances, delayed reactions, and dizziness, which gradually progress to hearing and visual impairment, seizures, confusion, and coma.
### Diagnoses:
This patient was diagnosed with seizures caused by star fruit intoxication. The experience of eating star fruit and the electroencephalograms can prove our diagnosis.
### Interventions:
We performed intensive renal replacement therapy according to the recommendations in the literature. However, his symptoms did not improve significantly until he received an extra dose of levetiracetam and resumed his previous dialysis schedule.
### Outcomes:
The patient was discharged after 21 days without neurologic sequelae. Five months after discharge, he was readmitted due to poor seizure control.
### Lessons:
To improve the prognosis of these patients and to reduce their financial burden, the use of antiepileptic drugs should be emphasized.
## 1. Introduction
Most people know star fruit as delicious fruit. However, few of them know that one of the star fruit ingredients, caramboxin, and has neurotoxicity.[1] Due to impaired excretion of certain drugs and toxins, eating star fruit can be dangerous to patients with renal dysfunction. Seizures or status epilepticus typically indicate a bad prognosis for patients with star fruit intoxication. Most physicians believe that aggressive additional renal replacement therapy (RRT) is the only way to cure patients with star fruit intoxication and that intense treatments for seizures or status epilepticus have a poor prognosis. In this particular case, the patient had multiple underlying diseases and presented with manifestations of neurological damage including seizures. After receiving medication for 21 days, he was discharged but was readmitted 5 months later with poorly controlled epilepsy.
## 2. Case presentation
A 67-year-old male patient with diabetic nephropathy, hypertension, and polycystic kidney disease and uremic stage of chronic renal failure (CRF) had been on regular RRT 3 times a week for 2 years. He developed symptoms such as hiccups, vomiting, speech disorders, slowed reaction, and dizziness after eating a star fruit 6 days ago. The patient was admitted via the emergency department on January 19, 2022. At the time of admission, his blood pressure was $\frac{198}{76}$ mm Hg, pulse rate was 62/minute, respiratory rate was 16/minute and body temperature was 36.6°C. Physical examinations were not specific. His biochemical data revealed the following: creatinine levels = 866.6 μmol/L, blood urea nitrogen = 19.71 μmol/L, sodium = 140.9 mmol/L, potassium = 4.68 mmol/L, pH = 7.436, PO2 = 186 mm Hg, PCO2 = 37.4 mm Hg, HCO3 = 24.8 mmol/L, glucose = 9.58 mmol/L, hemoglobin = 126 g/L and white blood cells = 5.62 *10^9. The brain computed tomography scan (CT), and magnetic resonance images (MRI) (Fig. 1A) did not reveal any specific abnormalities. On the night of January 19, his consciousness level declined and he presented aphasia (Glasgow Coma Scale [GCS] E3V3M5). The neurologist considered acute ischemic stroke and administered thrombolysis after a CT scan to exclude cerebral hemorrhage, but there was no significant improvement in his symptoms. On the morning of January 20, he presented with a seizure of binocular upward gaze and limb twitching, which resolved with intramuscular midazolam, but still with a progressive decrease in the level of consciousness (GCS E2V2M4). It was confirmed that he had no family history of seizures or epilepsies after asking family members. The diffusion-weighted image showed slight hyperintensities in the bilateral cerebellum and occipital lobe (Fig. 1B). Physical examinations revealed bilateral pupils of equal size and roundness, pupil diameter of 3 mm, sensitive reflex to light, right-sided gaze in both eyes, normal muscle tone, bilateral Babinski (+), bilateral Chaddock (+), and the rest of the physical examinations were uncooperative. Finally, the patient was diagnosed with seizure and toxic encephalopathy caused by eating star fruits. He was admitted to the neurology department for continued treatments on January 21.
**Figure 1.:** *Magnetic resonance images during the first hospitalization. (A) DWI on January 19. No obvious hyperintensities were seen. (B) DWI on January 20. The bilateral cerebellar and occipital lobe have patchy slight hyperintensities. (C) DWI on January 28. The bilateral parietal lobe, occipital lobe, and temporal lobe had multiple patchy hyperintensities, which were larger than the last time, and some of them appeared new. (D) DWI on February 4. Hyperintensities are lower than last time. (E) DWI on April 25. No obvious hyperintensities. (F) T2 FLAIR on April 25. No obvious lesions left. DWI = diffusion-weighted image, FLAIR = fluid-attenuated inversion recovery.*
The course of treatment in the neurology department is shown in Table 1. An electroencephalogram (EEG) was performed on January 21 (Fig. 2A). Our original plans were to give him RRT at the previous frequency level, 3 times a week, starting on January 21 (next on January 22), and to give levetiracetam 500mg/day followed by levetiracetam 250 mg each time after RRT to prevent future seizures. As we reviewed the literature, we found that the additional RRT is the key to improving outcomes rather than intensifying treatments of seizures or epilepsy. Therefore, we performed RRT for 6 consecutive days from January 24 to January 29 and reduced the dosage of levetiracetam to 250 mg/day. The brain MRI on January 28 (Fig. 1C) showed that the hyperintensities were larger than the last time, and some appeared new.
On January 30, after intensive RRT for 6 consecutive days, he still had occasional right-sided gaze, hearing and vision impairments, confusion, and agitation (GCS E2V3M4). The EEG (Fig. 2B and C) on that day confirmed that his condition had not improved significantly. On January 31, we tentatively increased the dosage of levetiracetam again to 500mg/day, restored the frequency of RRT to what he received before (3 times a week), and gave levetiracetam 250 mg after every time of RRT. The patient condition improved rapidly after the above treatments. Four days later, on February 3, he regained consciousness (GCS E4V5M6) and could recall the onset of the disease. He was also able to walk and to complete simple activities of daily life independently. The brain MRI on February 4 (Fig. 1D) and the EEG on February 8 (Fig. 2D and E) both indicated that his condition was better than before.
The patient was discharged on February 11, 2022, 21 days after admission, with complete resolution of symptoms and no neurological sequelae, but physicians still recommended him taking levetiracetam for at least 6 months. At the same time, he continued to receive regular hemodialyses at the out-patient hemodialysis unit after discharge. The brain MRI on April 25, 2022 was unremarkable and had no residual lesions (Fig. 1E and F).
He was readmitted to the hospital again on July 8, 2022, with symptoms of confused speech and abnormal behavior. On MRI and CT, no appreciable abnormalities were discovered. A large number of spike-waves and slow-waves periodic discharged on the EEG on the day of admission (Fig. 3A). We consequently assumed that his poorly controlled epilepsy was to blame for his symptoms this time. In terms of therapy, we just upped the levetiracetam dosage to 500 mg Bid while maintaining the frequency of his RRT. The second dose was given after the RRT on the day of the RRT. Epileptiform discharges gradually disappeared on the EEG on July 11 and July 20 (Fig. 3B and C). This patient was discharged on the 12th day after the second admission, also with no neurological sequelae. He said that he did not feel any discomfort after being discharged from the hospital.
**Figure 3.:** *Electroencephalograms during the second hospitalization. (A) July 8. Slow-wave and sharp-wave periodic emission. (B) July 11. Scattered or rhythmic medium-high amplitude sharp waves. (C) July 20. Most of the abnormal waves disappear.*
## 3. Discussion
Star fruit is a tropical fruit popular in many countries and regions. However, only some people are familiar with its nephrotoxic and neurotoxic effect. Oxalate causes nephrotoxicity, which is most common in patients with normal renal function.[2–4] Acute tubular necrosis and interstitial nephritis can be caused by oxalic acid crystals deposited in the renal tubules.[5] *Caramboxin is* the cause of neurotoxicity, which is primarily seen in patients with abnormal renal function.[1] *It is* impossible to estimate the incidence of neurotoxicity in individuals with normal renal function because so few of them are reported in the previous literature.
Martin was the first to report star fruit intoxication in patients on regular dialysis.[6] Eight patients developed persistent hiccups after ingesting star fruit, but no other symptoms were described. Since then, there have been more reports of hiccups, vomiting, and even seizures and comas after ingesting star fruit, but the root cause of these symptoms was unclear. Until 2013, Garcia-Cairasco extracted this neurotoxic substance from star fruit, named it caramboxin, and demonstrated its neuroexcitatory properties.[1] Caramboxin has decreased clearance in patients with abnormal renal function and produces excitatory effects on the central nervous system. Neto classified the neurotoxic effects of star fruit into 3 levels of intoxication: mild, moderate and severe intoxication.[7] Coma, seizures or status epilepticus and shock are signs of severe intoxication. This result suggests that seizure is a poor prognostic factor for star fruit intoxication and deserves the attention of physicians. In the case we report here, the symptoms from initial hiccups and vomiting to impaired consciousness were largely consistent with those described in previous reports. This patient also presented seizures and coma, which were classified as severe intoxication according to Neto classification. It may suggest a poor prognosis.
The differences between the case we reported this time and the previous ones are the following.
First, in terms of treatment. In previously published cases of this disease, the choice of renal replacement therapy modality is the focus of treatment for most physicians. This is the first report of a treatment focus on antiepileptic therapy. As recommended in previous literature, we performed RRT for 6 consecutive days while administering 250 mg of levetiracetam daily. However, the outcomes after 6 days were unsatisfactory. There was no significant improvement in the patient symptoms, which MRI and EEG confirmed. Then, we considered whether the consecutive RRT resulted in an excessive clearance of levetiracetam and an insufficient blood level to control epilepsy. Because of this, we experimented with decreasing the frequency of RRT and increasing the levetiracetam dosage. His condition significantly improved after 4 days. The optimal treatment for patients with star fruit intoxication has been controversial. In the Neto classification, it was proved that aggressive RRT is effective for patients with mild and moderate intoxication. In contrast, there are still significant differences in outcomes for patients with severe intoxication, even they received timely supportive therapy and aggressive RRT. Despite this, it is undeniable that aggressive RRT is currently the only way to improve patient prognosis. This is why we decided to increase the frequency of RRT in the first place. However, the subsequent change in this patient condition differed from what had previously been reported in the literature.
Second, on the prognosis. The patient we reported had several diseases and was in the uremic phase of CRF. Furthermore, he had seizures and was severe intoxication according to Neto grading criteria, with a likely poor prognosis. The short-term prognosis of the patient we report is favorable. But the star fruit intoxication seems to lower the threshold of seizures from our follow-up results. The outcomes of patients with star fruit intoxication vary widely. Some patients recovered in a short time without any sequela and did not need other therapy, some patients recovered but still needed to continue dialysis or take antiepileptic drugs, and others eventually died. According to previous reports, $26\%$ of patients with star fruit intoxication had seizures. The mortality rate of these patients who developed seizures was $61\%$, which was significantly higher than those without seizures.[8] The reasons for these individual differences are still controversial. What is certain is that there is no association between the type of star fruit and the form in which it was ingested and the severity of intoxication symptoms or mortality.[7,9] A paper reviewed the characteristics of patients with star fruit intoxication in the last decade: $63.2\%$ of patients were male and $69.1\%$ of patients had previous renal abnormalities ($63.8\%$ of these patients were on hemodialysis or peritoneal dialysis). Meanwhile, patients with abnormal kidney function had a worse prognosis compared with healthy individuals.[10] However, it was not verified that the degree of decline in the renal function directly predicted the severity of caramboxin-induced neurological impairment.
Lastly, there is no published information about the follow-up of individuals who have previously suffered from star fruit intoxication. For this patient, we followed up for 6 months. In the fifth month after discharge, and he was readmitted for suspected poorly controlled epilepsy. It cannot be excluded that a history of star fruit intoxication decreased the threshold for seizures.
These are some of the ideas that have emerged through caring for this patient. We cannot explain why the patient symptoms did not improve significantly after consecutive RRT and why the seizures were induced during the second hospitalization because there is no way to detect the amount of caramboxin in the body or measure the blood concentration of levetiracetam on time. Most physicians focus on the choice of RRT when caring for these kinds of patients. We think using antiepileptic drugs is just as important as choosing the right RRT for people with seizures or persistent status epilepticus. At the same time, the relationship between the dose of antiepileptic drugs and the choice of RRT needs to be explored. Clinicians who see a patient with CRF and feel they may be experiencing symptoms should also consider whether the patient has ever consumed star fruit because some of the symptoms are similar to those of metabolic encephalopathy and stroke.
## Author contributions
Methodology: Yu Bai, Jingfeng Zhang, Chengcheng Zhang, Miaomiao Cheng.
Writing – original draft: Aixun Li, Chunyan Guo, Yu Zhang.
Writing – review & editing: Baoxin Chen, Xianglan Jin, Jing Zhou.
## References
1. Garcia-Cairasco N, Moyses-Neto M, Del Vecchio F. **Elucidating the neurotoxicity of the star fruit.**. *Angew Chem Int Ed Engl* (2013) **52** 13067-70. PMID: 24281890
2. Barman AK, Goel R, Sharma M. **Acute kidney injury associated with ingestion of star fruit: acute oxalate nephropathy.**. *Indian J Nephrol* (2016) **26** 446-8. PMID: 27942177
3. Herath N, Kodithuwakku G, Dissanayake T. **Acute kidney injury following star fruit ingestion: a case series.**. *Wilderness Environ Med* (2021) **32** 98-101. PMID: 33518496
4. Stumpf MAM, Schuinski AFM, Baroni G. **Acute kidney injury with neurological features: beware of the star fruit and its caramboxin.**. *Indian J Nephrol* (2020) **30** 42-6. PMID: 32015601
5. Yasawardene P, Jayarajah U, De Zoysa I. **Mechanisms of star fruit (Averrhoa carambola) toxicity: a mini-review.**. *Toxicon* (2020) **187** 198-202. PMID: 32966829
6. Martin LC, Caramori JST, Barreti P. **Intractable hiccups induced by carambola (Averrhoa carambola) ingestion in patients with end-stage renal failure.**. *J Bras Nefrol* (1993) **15** 92-4
7. Neto MM, da Costa JAC, Garcia-Cairasco N. **Intoxication by star fruit (Averrhoa carambola) in 32 uraemic patients: treatment and outcome.**. *Nephrol Dial Transplant* (2003) **18** 120-5. PMID: 12480969
8. Chua CB, Sun CK, Tsui HW. **Association of renal function and symptoms with mortality in star fruit (Averrhoa carambola) intoxication.**. *Clin Toxicol (Phila)* (2017) **55** 624-8. PMID: 28443386
9. Auxiliadora-Martins M, Alkmin Teixeira GC, da Silva GS. **Severe encephalopathy after ingestion of star fruit juice in a patient with chronic renal failure admitted to the intensive care unit.**. *Heart Lung* (2010) **39** 448-52. PMID: 20561840
10. Yasawardene P, Jayarajah U, De Zoysa I. **Nephrotoxicity and neurotoxicity following star fruit (Averrhoa carambola) ingestion: a narrative review.**. *Trans R Soc Trop Med Hyg* (2021) **115** 947-55. PMID: 33693950
|
---
title: Handgrip strength is correlated with activities of daily living, balance, and
body composition in patients with thoracolumbar compression fracture
authors:
- Hirokazu Inoue
- Yukinori Hayashi
- Hideaki Watanabe
- Hideaki Sawamura
- Yasuyuki Shiraishi
- Ryo Sugawara
- Atsushi Kimura
- Masaaki Masubuchi
- Katsushi Takeshita
journal: Medicine
year: 2023
pmcid: PMC9981377
doi: 10.1097/MD.0000000000033141
license: CC BY 4.0
---
# Handgrip strength is correlated with activities of daily living, balance, and body composition in patients with thoracolumbar compression fracture
## Abstract
This study assessed the relationship between handgrip strength (HGS) and activities of daily living, balance, walking speed, calf circumference, body muscle, and body composition in elderly patients with thoracolumbar vertebral compression fracture (VCF). A cross-sectional study in a single hospital was performed with elderly patients diagnosed with VCF. After admission, we evaluated HGS, 10-meter walk test (speed), Barthel Index, Berg Balance Scale (BBS), numerical rating scale of body pain, and calf circumference. We examined skeletal muscle mass, skeletal muscle mass index, total body water (TBW), intracellular water, extracellular water (ECW), and phase angle (PhA) in patients with VCF using multi-frequency direct segmental bioelectrical impedance analysis after admission. A total of 112 patients admitted for VCF were enrolled (26 males, 86 females; mean age 83.3 years). The prevalence of sarcopenia according to the 2019 Asian Working Group for Sarcopenia guideline was $61.6\%$. HGS was significantly correlated with walking speed ($P \leq .001$, $R = 0.485$), Barthel Index ($P \leq .001$, $R = 0.430$), BBS ($P \leq .001$, $R = 0.511$), calf circumference ($P \leq .001$, $R = 0.491$), skeletal muscle mass index ($P \leq .001$, $R = 0.629$), ECW/TBW ($P \leq .001$, r = −0.498), and PhA ($P \leq .001$, $R = 0.550$). HGS was more strongly correlated with walking speed, Barthel Index, BBS, ECW/TBW ratio, and PhA in men than women. In patients with thoracolumbar VCF, HGS is associated with walking speed, muscle mass, activities of daily living measured using the Barthel Index, and balance measured using BBS. The findings suggest that HGS is an important indicator of activities of daily living, balance, and whole-body muscle strength. Furthermore, HGS is related to PhA and ECW/TBW.
## 1. Introduction
The prevalence of osteoporosis is increasing in line with the aging worldwide population, leading to a greater prevalence of vertebral compression fractures (VCFs) and hip fractures caused by falls. Population studies have indicated that $8\%$ of adults over 50 years of age have osteoporosis of the lumbar spine, and the prevalence of VCF adjusted for age and sex is estimated at 117 per 100,000 person-years.[1] VCFs are more common in elderly populations, causing gait disturbances and decreased activities of daily living (ADL). The Asian Working Group for Sarcopenia guideline reported that the prevalence of sarcopenia was $4.1\%$ to $11.5\%$ in the general elderly population.[2] Loss of lower leg muscle mass and grip strength are also risk factors for the development of VCF, and may be closely related to the pathogenesis of sarcopenia.[3] The reported prevalence of sarcopenia in patients with VCF ranges from $33.9\%$ to $65\%$.[3–5] Handgrip strength (HGS) is associated with body musculature. Handgrip dynamometers are used in many studies for various purposes because they are easy to use and provide useful data. In oncology, HGS is associated with cancer-related fatigue,[6] quality of life,[7] postoperative complications,[8] and mortality risk.[8] We previously demonstrated that HGS is correlated with walking ability in lumbar spinal stenosis.[9] Moreover, lower HGS may be associated with the onset of vertebral fracture.[3] Women with distal radial fracture exhibit lower HGS and lower dynamic body balance capacity. In women with distal radial fracture who are at risk of future fragility fractures, HGS and poor dynamic body balance ability were found to be significant risk factors.[10] A previous study demonstrated that HGS was the most important factor in the association between sarcopenia and osteoporosis, falls, and fractures.[11] HGS in healthy elderly people was found to be greater than that in VCF patients.[12] These results suggest that HGS may be a useful screening tool in VCF patients for assessing patient status and fracture risk.
Bioelectrical impedance analysis (BIA) is a useful, simple, and noninvasive tool for examining body composition. BIA is also used together with dual-energy X-ray absorptiometry to diagnose sarcopenia. Furthermore, BIA can be used to measure phase angle (PhA) and the distribution of body water between the extracellular water (ECW) and intracellular water (ICW) positions. Recent clinical studies have shown that PhA is a prognostic indicator in many clinical conditions (including nutritional risk, cancer, kidney disease, and human immunodeficiency virus infection) and in surgical patients.[13–16] Segmental multi-frequency BIA can be used to precisely measure ICW and ECW and generate an “edema index.” This index is calculated as the ratio of extracellular water to total body water (ECW/TBW). A previous study demonstrated that ECW/TBW, as a combination of overhydration and protein-energy wasting, may be a significant predictor of poorer outcomes.[17] However, the relationships between PhA and ECW/TBW in VCF patients are currently unclear.
A small number of studies have suggested that sarcopenia is associated with ADL in patients with VCF.[18] However, no studies have reported a correlation between HGS and ADL in patients with VCF. The purpose of the current study was to investigate the relationship between HGS and ADL, balance, walking speed, calf circumference, body muscle, and body composition in patients with thoracolumbar VCF.
## 2.1. Patients
We retrospectively enrolled patients who were admitted at a single hospital for thoracolumbar VCF between October 2016 and March 2021. The study protocol was approved by the Ethics Review Board of Shiobara Spring Hospital. The present study was performed in accordance with the World Medical Association Declaration of Helsinki principles. We calculated the necessary sample size to achieve an alpha value of 0.05 and power of 0.80 using G*Power 3 statistical software (v 3.1.9.7, Heinrich-Heine-University, Düsseldorf, Germany),[19,20] yielding an estimated sample size of 26. The inclusion criteria were as follows: patients with ≥ 1 recent symptomatic VCF (T5–L5); 60 years or older; and a low-energy injury (simple fall) or an injury without trauma. We defined osteoporotic VCF as axial compression of a vertebral body with an intact posterior restraining element including wedge, biconcave, and compression deformities as described by Eastell et al[21] We assumed that the patient had acute VCF when they had tenderness at the fracture site without a callus on spinal radiography. The exclusion criteria were as follows: pathological fracture, including fractures related to malignancy, infection, or other medical conditions; burst fracture with a retro-pulsed bony fragment into the spinal canal; neurological deficit; use of steroids or medications for severe liver or kidney disease; no available BIA instrument; and no measures of fitness.
## 2.2. Measurements
HGS of both upper limbs were measured using a handheld dynamometer. The patients squeezed the dynamometer as hard as possible for 3 seconds. Two attempts were performed with each hand, with a brief rest between trials. We used the best performance for the analysis.[22] Gait speed was measured using the timed 10-meter walk test. The 10-meter walk test measures the time that it takes a patient to walk 10 meters. We did not perform the tests in patients for whom it was difficult to perform them safely, such as those who needed walking aids, those who had recently had a lower limb fracture or surgery, or those who had neurological conditions. Trained physical therapists conducted the tests. For patients who could not perform the test, gait speed was recorded as 0 m/s. The Barthel *Index is* 1 of the most common rating scales for measuring activity limitations in patients with neuromuscular and musculoskeletal conditions. The Barthel Index measures 10 items of functioning in daily life, including feeding, bathing, grooming, dressing, toilet uses, transfers, mobility, and stair use. The Berg Balance Scale (BBS) was applied to evaluate standing and sitting balance, and each item was evaluated on a 5-point scale (a total of 56 points). Patients evaluated their average body pain using an 11-point numerical rating scale (0: no pain; 10: worst pain imaginable) on admission. Body mass index was calculated as weight/height2 (kg/m2). Calf circumference was measured at the thickest point on both sides. The mean of the right and left leg measurements was calculated.
## 2.3. Definition of sarcopenia
We used low HGS (< 28 kg for males and < 18 kg for females) and walking speed < 1.0 m/s to define sarcopenia in this study, according to the 2019 Asian Working Group for Sarcopenia.[23] A multi-frequency validated BIA instrument, the Inbody S10 (Biospace, Seoul, Korea), was used to examine the patient in a supine position. This device was also used to measure skeletal muscle mass and fat mass. Skeletal muscle mass index (SMI) was the measured skeletal muscle mass divided by the square of the height in meters. Sarcopenia for men and women was defined as SMI values of < 7.0 kg/m2 and < 5.7 kg/m2, respectively. PhA, ECW, ICW, and TBW values were obtained using BIA. The ratio of ECW to TBW (ECW/TBW) and the ratio of ICW to TBW (ICW/TBW) were then calculated to compare the distribution of body water.
## 2.4. Statistical analysis
Continuous variables are presented as mean ±standard deviation. Correlations between HGS and the continuous variables were assessed using Pearson correlation coefficients. A P value of < 0.05 was considered to indicate statistical significance. All statistical analyses were performed using SPSS for Windows version 17.0 (SPSS, Chicago, IL).
## 3. Results
A total of 164 patients were initially assessed. We excluded 52 patients because of missing measures of fitness or because BIA was not available. We enrolled 112 patients (average age: 83.3 years; 26 males, 86 females). The total prevalence of sarcopenia was $61.6\%$ ($50\%$ in male patients, $65.1\%$ in female patients). We compared demographic parameters with HGS (Table 1). Female patients had significantly weaker HGS than male patients (r = −0.568, $P \leq .001$). In the cohort, there were significant correlations between HGS and age (r = −0.452, $P \leq .001$), height ($R = 0.600$, $P \leq .001$), weight ($R = 0.258$, $$P \leq .006$$), walking speed ($R = 0.485$, $P \leq .001$), and calf circumference ($R = 0.491$, $P \leq .001$). Fifty-nine patients were not able to walk 10 meters on admission. HGS was significantly correlated with Barthel Index ($R = 0.430$, $P \leq .001$) and BBS ($R = 0.511$, $P \leq .001$). HGS was not correlated with numerical rating scale of body pain. In male patients, HGS was significantly correlated with age (r = −0.585, $$P \leq .002$$), height ($R = 0.486$, $$P \leq .012$$), walking speed ($R = 0.611$, $$P \leq .001$$), calf circumference ($R = 0.584$, $$P \leq .003$$), Barthel Index ($R = 0.582$, $$P \leq .002$$), and BBS ($R = 0.646$, $$P \leq .005$$). In female patients, HGS was significantly correlated with age (r = −0.296, $$P \leq .006$$), height ($R = 0.294$, $$P \leq .006$$), weight ($R = 0.248$, $$P \leq .021$$), walking speed ($R = 0.293$, $$P \leq .006$$), calf circumference ($R = 0.467$, $P \leq .001$), Barthel Index ($R = 0.338$, $$P \leq .001$$), and BBS ($R = 0.374$, $$P \leq .004$$).
**Table 1**
| Unnamed: 0 | Total (n = 112) | P | r | Male (n = 26) | P.1 | r.1 | Female (n = 86) | P.2 | r.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Age (yr) | 83.3 ± 7.8 | <.001 | −0.452 | 80.3 ± 9.2 | .002 | −0.585 | 84.3 ± 7.1 | .006 | −0.296 |
| Gender (Male: Female) | 26: 86 | <.001 | −0.568 | - | - | - | - | - | - |
| Height (cm) | 150.1 ± 8.7 | <.001 | 0.600 | 161.2 ± 6.0 | .012 | 0.486 | 146.8 ± 6.3 | .006 | 0.294 |
| Weight (kg) | 50.7 ± 16.8 | .006 | 0.258 | 60.4 ± 21.5 | .678 | −0.085 | 47.8 ± 14.0 | .021 | 0.248 |
| BMI (kg/m2) | 22.4 ± 6.9 | .607 | 0.049 | 23.4 ± 9.1 | .359 | −0.188 | 22.2 ± 6.2 | .140 | 0.161 |
| Grip (kg) | 15.3 ± 7.3 | - | - | 22.8 ± 8.8 | - | - | 13.1 ± 4.9 | - | - |
| Walking speed (m/s) | 0.38 ± 0.48 | <.001 | 0.485 | 0.59 ± 0.62 | .001 | 0.611 | 0.31 ± 0.41 | .006 | 0.293 |
| Barthel index | 58.3 ± 24.2 | <.001 | 0.430 | 65.4 ± 26.1 | .002 | 0.582 | 56.2 ± 23.3 | .001 | 0.338 |
| BBS | 24.3 ± 18.0 | <.001 | 0.511 | 31.8 ± 21.3 | .005 | 0.646 | 22.1 ± 16.5 | .004 | 0.374 |
| NRS | 6.0 ± 2.4 | .346 | −0.107 | 5.6 ± 2.7 | .638 | 0.116 | 6.1 ± 2.3 | .163 | −0.181 |
| Calf circumference (cm) | 28.9 ± 3.9 | <.001 | 0.491 | 30.1 ± 4.2 | .003 | 0.584 | 28.6 ± 3.8 | <.001 | 0.467 |
HGS was significantly correlated with the muscle mass of the right arm ($R = 0.684$, $P \leq .001$), left arm ($R = 0.622$, $P \leq .001$), trunk ($R = 0.655$, $P \leq .001$), right leg ($R = 0.692$, $P \leq .001$), and left leg ($R = 0.673$, $P \leq .001$) (Table 2). HGS exhibited a significant negative correlation with the ECW/TBW ratio (r = −0.498, $P \leq .001$). However, HGS was significantly positively correlated with the ICW/TBW ratio ($R = 0.492$, $P \leq .001$) and PhA ($R = 0.550$, $P \leq .001$). In male patients, HGS was significantly correlated with the muscle mass of the right arm ($R = 0.467$, $$P \leq .016$$), trunk ($R = 0.391$, $$P \leq .048$$), right leg ($R = 0.625$, $$P \leq .001$$), and left leg ($R = 0.639$, $P \leq .001$), and the ECW/TBW ratio (r = −0.457, $$P \leq .019$$), ICW/TBW ratio ($R = 0.467$, $$P \leq .016$$), and PhA ($R = 0.653$, $P \leq .001$). In female patients, HGS was significantly correlated with the muscle mass of the right arm ($R = 0.544$, $P \leq .001$), left arm ($R = 0.497$, $P \leq .001$), trunk ($R = 0.507$, $P \leq .001$), right leg ($R = 0.429$, $P \leq .001$), and left leg ($R = 0.378$, $P \leq .001$), and the ECW/TBW ratio (r = −0.419, $P \leq .001$), ICW/TBW ratio ($R = 0.432$, $P \leq .001$), and PhA ($R = 0.426$, $P \leq .001$).
**Table 2**
| Unnamed: 0 | Total (n = 112) | P | r | Male (n = 26) | P.1 | r.1 | Female (n = 86) | P.2 | r.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| SMI (cm2/m2) | 5.30 ± 1.15 | <.001 | 0.629 | 6.39 ± 1.08 | .003 | 0.556 | 4.97 ± 0.95 | <.001 | 0.443 |
| Muscle mas | Muscle mas | Muscle mas | Muscle mas | Muscle mas | Muscle mas | Muscle mas | Muscle mas | Muscle mas | Muscle mas |
| Right arm (kg) | 1.5 ± 0.5 | <.001 | 0.684 | 2.1 ± 0.4 | .016 | 0.467 | 1.3 ± 0.4 | <.001 | 0.544 |
| Left arm (kg) | 1.5 ± 0.6 | <.001 | 0.622 | 2.1 ± 0.6 | .123 | 0.311 | 1.3 ± 0.4 | <.001 | 0.497 |
| Trunk (kg) | 14.3 ± 3.5 | <.001 | 0.655 | 18.4 ± 3.0 | .048 | 0.391 | 13.0 ± 2.5 | <.001 | 0.507 |
| Right leg (kg) | 4.6 ± 1.4 | <.001 | 0.692 | 6.3 ± 1.5 | .001 | 0.625 | 4.1 ± 1.0 | <.001 | 0.429 |
| Left leg (kg) | 4.6 ± 1.4 | <.001 | 0.673 | 6.3 ± 1.4 | <.001 | 0.639 | 4.1 ± 1.0 | <.001 | 0.378 |
| ICW (kg) | 14.2 ± 3.5 | <.001 | 0.569 | 18.1 ± 3.4 | .096 | 0.333 | 13.1 ± 2.6 | .001 | 0.349 |
| ECW (kg) | 9.8 ± 2.3 | <.001 | 0.500 | 12.3 ± 2.3 | .275 | 0.222 | 9.0 ± 1.7 | .012 | 0.271 |
| TBW (kg) | 24.0 ± 5.8 | <.001 | 0.543 | 30.4 ± 5.7 | .151 | 0.290 | 22.1 ± 4.3 | .003 | 0.319 |
| ECW/TBW ratio | 0.408 ± 0.009 | <.001 | −0.498 | 0.403 ± 0.010 | .019 | −0.457 | 0.410 ± 0.009 | <.001 | −0.419 |
| ICW/TBW ratio | 0.592 ± 0.009 | <.001 | 0.492 | 0.596 ± 0.010 | .016 | 0.467 | 0.591 ± 0.009 | <.001 | 0.432 |
| Phase angle (˚) | 3.71 ± 0.83 | <.001 | 0.550 | 4.15 ± 0.81 | <.001 | 0.653 | 3.58 ± 0.79 | <.001 | 0.426 |
## 4. Discussion
In this study, we investigated the prevalence of sarcopenia in patients with VCF and examined the relationships between HGS and a range of factors. The findings revealed that among 112 enrolled patients, sarcopenia was present in 69 patients ($61.6\%$). Moreover, HGS was correlated with sex, age, height, weight, walking speed, Barthel Index, BBS, calf circumference, SMI, muscle mass, ECW/TBW ratio, ICW/TBW ratio, and PhA.
Sarcopenia is a risk factor for osteoporotic VCF and has recently become a major issue in medical care for the elderly.[24] The prevalence of sarcopenia ranges from $5.8\%$ to $14.9\%$ in men and from $4.1\%$ to $16.6\%$ in women, as reported using the relative appendicular skeletal muscle index or skeletal muscle index, according to the International Working Group on Sarcopenia or the European Working Group on Sarcopenia in Older People criteria.[25] Thus, the prevalence of sarcopenia appears to vary depending on the diagnostic criteria used. Hida et al[5] demonstrated that the prevalence rates of sarcopenia were $42\%$ in 70-year-old patients with acute VCF and $25\%$ in patients without acute VCF. The researchers reported that sarcopenia and lower leg muscle mass were risk factors for VCF.[5] Eguchi et al[3] found that decreased leg muscle mass and decreased HGS were risk factors for VCF in elderly women. Previous studies have reported a sarcopenia prevalence of $33.9\%$ to $65\%$ in VCF patients.[3–5,26] In the present study, the prevalence was $61.6\%$, similar to that reported in previous studies.
The current results revealed that HGS was correlated with age, sex, height, weight, walking speed, and calf circumference in patients with VCF. Our previous study demonstrated that HGS was correlated with lower extension power, height, weight, and age in patients with lumbar spinal stenosis.[9] HGS decreases considerably with age.[27] Additionally, low muscle strength and power are strongly associated with 2 complementary definitions of poor athletic performance, regardless of age or sex. HGS is related to age, height, and weight, but not to body mass index. Moreover, positive correlations have been reported between calf circumference and HGS in geriatric patients ($R = 0.422$)[28] and young Japanese women ($R = 0.377$).[29] The present findings reveal that HGS is associated with the Barthel Index and BBS. Barthel *Index is* a scale used to measure ADL, and BBS is used to measure balance abilities. Thus, our findings suggest that HGS is associated with ADL and balance. In previous studies, HGS was significantly correlated with Barthel Index scores in frail elderly people ($R = 0.214$)[30] and patients with dysphagia ($R = 0.38$).[31] Moreover, HGS was significantly correlated with BBS in elderly people ($R = 0.576$).[28] *As a* consequence, high HGS in elderly people is an indicator of good balance, and decreased HGS is an important risk factor for falls in postmenopausal women.[32] These findings suggest that HGS is a good indicator of an individual’s ADL and balance.
In the current study, HGS was correlated with muscle mass measured by BIA. Our previous study revealed that HGS in patients with lumbar spinal stenosis was correlated with psoas muscle mass and skeletal muscle mass at the L3 level.[9] Multiple studies have reported a close relationship between HGS and lifespan, whole-body muscle volume, and physical activity.[33] The areas of the psoas and paraspinal muscles are important for grading the vitality of the patient. Low psoas muscle area is correlated with low HGS and short physical performance battery scores, indicating physical frailty.[34] Low psoas muscle area is also related to prolonged hospital stays in elderly cardiac surgery patients. In frail patients classified on the basis of HGS, the total area of the psoas muscle was smaller than that of nonfrail patients.[35] In the current study, HGS was related to PhA, ECW/TBW ratio, and ICW/TBW ratio in patients with VCF. Previously, the ECW/TBW ratio was reported to be significantly higher in patients with sepsis than healthy individuals, while PhA and the ICW/TBW ratio were significantly lower in patients with sepsis.[36] In hepatic, pancreatic, and biliary surgery, an increased ECW/TBW ratio in patients with fluid imbalance suggests a possible causal relationship with the development of ascites and fluid retention in the postoperative period.[37] A high ECW/TBW ratio was correlated with lower Subjective Global Assessment scores not only in patients receiving renal replacement therapy[38] but also in patients with autosomal dominant polycystic kidney disease.[39] Moreover, a high ECW/TBW ratio was associated with malnutrition according to the Subjective Global Assessment questionnaire. One study of peritoneal dialysis patients reported a cutoff value of ECW/TBW for 1-year mortality of > 0.371 for men and > 0.372 for women,[40] and another study reported a cutoff value of 0.400.[17] A study of acute heart failure patients revealed a cutoff value of ECW/TBW of 0.390[41] for a higher incidence of rehospitalization. In this study, the mean ECW/TBW ratio was 0.408, suggesting that patients with VCF are in worse general condition. PhA was previously reported to be significantly and positively associated with somatic protein and muscle function in cancer patients.[42] Another study reported that PhA was positively correlated with survival in patients undergoing hemodialysis.[43] The ECW/TBW ratio and PhA have been used as indicators of poor systemic status and survival. To our knowledge, this is the first study to demonstrate the relationship between HGS, ECW/TBW ratio, and PhA.
We evaluated male and female patients across many variables in the current study. Many previous studies of VCF only evaluated female patients or grouped male and female patients together.[3–5,12,18,44] HGS was more strongly correlated with walking speed, Barthel Index, BBS, ECW/TBW ratio, and PhA in male than female patients. This finding suggests that HGS is more critically related to ADL and body condition in male than female patients. Thus, male patients with VCF may be in worse condition and face more severe difficulties than female VCF patients.
The current study involved several limitations. First, patients were retrospectively surveyed at a single institution. Thus, the sample population may have been biased. Second, the cross-sectional design of the current study limited our ability to draw conclusions regarding causal relationships. In future, we plan to conduct a cohort study of VCF patients to enable the detection of causal factors. Third, the number of male patients with VCF in the current study was small. Including male VCF patients will be important in future studies of sarcopenia and frailty because they appear to have greater problems with sarcopenia and frailty than female VCF patients.
In conclusion, HGS was correlated with the Barthel Index in patients with thoracolumbar VCF. Thus, for thoracolumbar VCF patients, HGS provides an indicator of not only whole-body muscle strength but also ADL and balance. Moreover, PhA, ECW/TBW, and ICW/TBW were significantly correlated with HGS. Lastly, HGS was more strongly correlated with gait speed, Barthel Index, BBS, ECW/TBW ratio, and PhA in male than female patients.
## Acknowledgments
The authors thank Mr. Kabasawa and Mr. Sakaguchi, as well as the physical and occupational therapists at Shiobara Spring Hospital, for providing support in carrying out this study. We thank Benjamin Knight, MSc., and Carol Wilson, PhD, from Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript.
## Author contributions
Conceptualization: Hirokazu Inoue, Hideaki Sawamura, Yasuyuki Shiraishi, Ryo Sugawara, Atsushi Kimura, Katsushi Takeshita.
Data curation: Hirokazu Inoue, Yukinori Hayashi, Hideaki Sawamura.
Formal analysis: Hirokazu Inoue, Hideaki Watanabe.
Supervision: Masaaki Masubuchi, Katsushi Takeshita.
Validation: Hirokazu Inoue, Yukinori Hayashi, Hideaki Watanabe.
Visualization: Hirokazu Inoue, Yukinori Hayashi.
Writing – original draft: Hirokazu Inoue, Hideaki Watanabe.
## References
1. Cooper C, Atkinson EJ, O’Fallon WM. **Incidence of clinically diagnosed vertebral fractures: a population-based study in Rochester, Minnesota, 1985-1989.**. *J Bone Miner Res* (1992) **7** 221-7. PMID: 1570766
2. Chen LK, Lee WJ, Peng LN. **Recent advances in Sarcopenia research in Asia: 2016 update from the Asian Working Group for Sarcopenia.**. *J Am Med Dir Assoc* (2016) **17** 767.e1767 e761-767.e7
3. Eguchi Y, Toyoguchi T, Orita S. **Reduced leg muscle mass and lower grip strength in women are associated with osteoporotic vertebral compression fractures.**. *Arch Osteoporos* (2019) **14** 112. PMID: 31760559
4. Ohyama S, Hoshino M, Takahashi S. **Presence of sarcopenia does not affect the clinical results of balloon kyphoplasty for acute osteoporotic vertebral fracture.**. *Sci Rep* (2021) **11** 122. PMID: 33420234
5. Hida T, Shimokata H, Sakai Y. **Sarcopenia and sarcopenic leg as potential risk factors for acute osteoporotic vertebral fracture among older women.**. *Eur Spine J* (2016) **25** 3424-31. PMID: 25690348
6. Kilgour RD, Vigano A, Trutschnigg B. **Cancer-related fatigue: the impact of skeletal muscle mass and strength in patients with advanced cancer.**. *J Cachexia Sarcopenia Muscle* (2010) **1** 177-85. PMID: 21475694
7. Norman K, Stobaus N, Smoliner C. **Determinants of hand grip strength, knee extension strength and functional status in cancer patients.**. *Clin Nutr* (2010) **29** 586-91. PMID: 20299136
8. Chen C-H, Ho C, Huang Y-Z. **Hand-grip strength is a simple and effective outcome predictor in esophageal cancer following esophagectomy with reconstruction: a prospective study.**. *J Cardiothorac Surg* (2011) **6** 98. PMID: 21843340
9. Inoue H, Watanabe H, Okami H. **Handgrip strength correlates with walking in lumbar spinal stenosis.**. *Eur Spine J* (2020) **29** 2198-204. PMID: 32651633
10. Fujita K, Kaburagi H, Nimura A. **Lower grip strength and dynamic body balance in women with distal radial fractures.**. *Osteoporos Int* (2019) **30** 949-56. PMID: 30607458
11. Sjoblom S, Suuronen J, Rikkonen T. **Relationship between postmenopausal osteoporosis and the components of clinical sarcopenia.**. *Maturitas* (2013) **75** 175-80. PMID: 23628279
12. Anand A, Shetty AP, Renjith KR. **Does sarcopenia increase the risk for fresh vertebral fragility fractures?: A case-control study.**. *Asian Spine J* (2020) **14** 17-24. PMID: 31575110
13. Pirlich M, Schutz T, Spachos T. **Bioelectrical impedance analysis is a useful bedside technique to assess malnutrition in cirrhotic patients with and without ascites.**. *Hepatology* (2000) **32** 1208-15. PMID: 11093726
14. Mushnick R, Fein PA, Mittman N. **Relationship of bioelectrical impedance parameters to nutrition and survival in peritoneal dialysis patients.**. *Kidney Int Suppl* (2003) **87** S53-56
15. Schwenk A, Beisenherz A, Romer K. **Phase angle from bioelectrical impedance analysis remains an independent predictive marker in HIV-infected patients in the era of highly active antiretroviral treatment.**. *Am J Clin Nutr* (2000) **72** 496-501. PMID: 10919947
16. Barbosa-Silva MC, Barros AJ. **Bioelectric impedance and individual characteristics as prognostic factors for post-operative complications.**. *Clin Nutr* (2005) **24** 830-8. PMID: 15975694
17. Guo Q, Yi C, Li J. **Prevalence and risk factors of fluid overload in Southern Chinese continuous ambulatory peritoneal dialysis patients.**. *PLoS One* (2013) **8** e53294. PMID: 23341936
18. Takahashi K, Kubo A, Ishimura K. **Correlation among sarcopenia, malnutrition and activities of daily living in patients with vertebral compression fractures: a comparison based on admission and discharge parameters evaluating these conditions.**. *J Phys Ther Sci* (2018) **30** 1401-7. PMID: 30568324
19. Faul F, Erdfelder E, Buchner A. **Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses.**. *Behav Res Methods* (2009) **41** 1149-60. PMID: 19897823
20. Faul F, Erdfelder E, Lang AG. **G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences.**. *Behav Res Methods* (2007) **39** 175-91. PMID: 17695343
21. Eastell R, Cedel SL, Wahner HW. **Classification of vertebral fractures.**. *J Bone Miner Res* (1991) **6** 207-15. PMID: 2035348
22. Park S, Kim HJ, Ko BG. **The prevalence and impact of sarcopenia on degenerative lumbar spinal stenosis.**. *Bone Joint J* (2016) **98-B** 1093-8. PMID: 27482023
23. Chen LK, Woo J, Assantachai P. **Asian Working Group for Sarcopenia: 2019 consensus update on Sarcopenia diagnosis and treatment.**. *J Am Med Dir Assoc* (2020) **21** 300-307.e2. PMID: 32033882
24. Fielding RA, Vellas B, Evans WJ. **Sarcopenia: an undiagnosed condition in older adults. Current consensus definition: prevalence, etiology, and consequences. International working group on sarcopenia.**. *J Am Med Dir Assoc* (2011) **12** 249-56. PMID: 21527165
25. Lee WJ, Liu LK, Peng LN. **Comparisons of sarcopenia defined by IWGS and EWGSOP criteria among older people: results from the I-Lan longitudinal aging study.**. *J Am Med Dir Assoc* (2013) **14** 528.e1528 e521-528.e7
26. Hong W, Cheng Q, Zhu X. **Prevalence of sarcopenia and its relationship with sites of fragility fractures in elderly Chinese men and women.**. *PLoS One* (2015) **10** e0138102. PMID: 26367872
27. Lauretani F, Russo CR, Bandinelli S. **Age-associated changes in skeletal muscles and their effect on mobility: an operational diagnosis of sarcopenia.**. *J Appl Physiol (1985)* (2003) **95** 1851-60. PMID: 14555665
28. Krause KE, McIntosh EI, Vallis LA. **Sarcopenia and predictors of the fat free mass index in community-dwelling and assisted-living older men and women.**. *Gait Posture* (2012) **35** 180-5. PMID: 21982745
29. Yasuda T. **Simplified morphological evaluation of skeletal muscle mass and maximum muscle strength in healthy young women: comparison between thigh and calf.**. *Womens Health (Lond)* (2020) **16** 1745506520962009. PMID: 33063630
30. Kitamura K, Nakamura K, Nishiwaki T. **Determination of whether the association between serum albumin and activities of daily living in frail elderly people is causal.**. *Environ Health Prev Med* (2012) **17** 164-8. PMID: 21861116
31. Matsuo H, Yoshimura Y, Ishizaki N. **Dysphagia is associated with functional decline during acute-care hospitalization of older patients.**. *Geriatr Gerontol Int* (2017) **17** 1610-6. PMID: 27910255
32. Rouzi AA, Ardawi MS, Qari MH. **Risk factors for falls in a longitudinal cohort study of Saudi postmenopausal women: the Center of Excellence for Osteoporosis Research Study.**. *Menopause* (2015) **22** 1012-20. PMID: 25608272
33. Aibar-Almazan A, Martinez-Amat A, Cruz-Diaz D. **Sarcopenia and sarcopenic obesity in Spanish community-dwelling middle-aged and older women: association with balance confidence, fear of falling and fall risk.**. *Maturitas* (2018) **107** 26-32. PMID: 29169576
34. Zuckerman J, Ades M, Mullie L. **Psoas muscle area and length of stay in older adults undergoing cardiac operations.**. *Ann Thorac Surg* (2017) **103** 1498-504. PMID: 27863730
35. Reeve TE, Ur R, Craven TE. **Grip strength measurement for frailty assessment in patients with vascular disease and associations with comorbidity, cardiac risk, and sarcopenia.**. *J Vasc Surg* (2018) **67** 1512-20. PMID: 29276105
36. Shin J, Park I, Lee JH. **Comparison of body water status and its distribution in patients with non-septic infection, patients with sepsis, and healthy controls.**. *Clin Exp Emerg Med* (2021) **8** 173-81. PMID: 34649405
37. Chong JU, Nam S, Kim HJ. **Exploration of fluid dynamics in perioperative patients using bioimpedance analysis.**. *J Gastrointest Surg* (2016) **20** 1020-7. PMID: 26715560
38. Demirci MS, Demirci C, Ozdogan O. **Relations between malnutrition-inflammation-atherosclerosis and volume status. The usefulness of bioimpedance analysis in peritoneal dialysis patients.**. *Nephrol Dial Transplant* (2011) **26** 1708-16. PMID: 20921295
39. Ryu H, Park HC, Kim H. **Bioelectrical impedance analysis as a nutritional assessment tool in autosomal dominant polycystic kidney disease.**. *PLoS One* (2019) **14** e0214912. PMID: 30947248
40. Kang SH, Choi EW, Park JW. **Clinical significance of the edema index in incident peritoneal dialysis patients.**. *PLoS One* (2016) **11** e0147070. PMID: 26785259
41. Park CS, Lee SE, Cho HJ. **Body fluid status assessment by bio-impedance analysis in patients presenting to the emergency department with dyspnea.**. *Korean J Intern Med* (2018) **33** 911-21. PMID: 29241303
42. Norman K, Stobaus N, Zocher D. **Cutoff percentiles of bioelectrical phase angle predict functionality, quality of life, and mortality in patients with cancer.**. *Am J Clin Nutr* (2010) **92** 612-9. PMID: 20631202
43. Abad S, Sotomayor G, Vega A. **The phase angle of the electrical impedance is a predictor of long-term survival in dialysis patients.**. *Nefrologia* (2011) **31** 670-6. PMID: 22130282
44. Hoshino M, Takahashi S, Yasuda H. **Balloon kyphoplasty versus conservative treatment for acute osteoporotic vertebral fractures with poor prognostic factors: propensity score matched analysis using data from two prospective multicenter studies.**. *Spine (Phila Pa 1976)* (2019) **44** 110-7. PMID: 29958202
|
---
title: Transcriptome profiling of skeletal muscles from Korean patients with Bethlem
myopathy
authors:
- Seung-Ah Lee
- Ji-Man Hong
- Jung Hwan Lee
- Young-Chul Choi
- Hyung Jun Park
journal: Medicine
year: 2023
pmcid: PMC9981387
doi: 10.1097/MD.0000000000033122
license: CC BY 4.0
---
# Transcriptome profiling of skeletal muscles from Korean patients with Bethlem myopathy
## Abstract
Bethlem myopathy is one of the collagens VI-related muscular dystrophies caused by mutations in the collagen VI genes. The study was designed to analyze the gene expression profiles in the skeletal muscle of patients with Bethlem myopathy. Six skeletal muscle samples from 3 patients with Bethlem myopathy and 3 control subjects were analyzed by RNA-sequencing. 187 transcripts were significantly differentially expressed, with 157 upregulated and 30 downregulated transcripts in the Bethlem group. Particularly, 1 (microRNA-133b) was considerably upregulated, and 4 long intergenic non-protein coding RNAs, LINC01854, MBNL1-AS1, LINC02609, and LOC728975, were significantly downregulated. We categorized differentially expressed gene using Gene Ontology and showed that Bethlem myopathy is strongly associated with the organization of extracellular matrix (ECM). Kyoto Encyclopedia of Genes and Genomes pathway enrichment reflected themes with significant enrichment of the ECM-receptor interaction (hsa04512), complement and coagulation cascades (hsa04610), and focal adhesion (hsa04510). We confirmed that Bethlem myopathy is strongly associated with the organization of ECM and the wound healing process. Our results demonstrate transcriptome profiling of Bethlem myopathy, and provide new insights into the path mechanism of Bethlem myopathy associated with non-protein coding RNAs.
## 1. Introduction
Collagen VI-related myopathy is a group of genetic disorders that affects skeletal muscles and connective tissues. It is clinically characterized by proximal muscle weakness, joint contractures, and distal joint hyperlaxity.[1,2] Inheritance patterns can also be autosomal dominant or recessive.[1] The severity of the clinical symptoms varies from the milder Bethlem myopathy to the more severe Ullrich congenital muscular dystrophy. Ullrich congenital muscular dystrophy showed congenital muscle weakness, hypotonia, striking joint hyperlaxity, particularly of the distal joints, in conjunction with contractures of the proximal joints, including the hips, elbows, and spine.[3] In Bethlem myopathy, the symptom onset may also be congenital; however, hypotonia at birth is usually rare.[4] Children with Bethlem myopathy have mild muscle weakness and contractures, including in the ankle dorsiflexors and long finger flexors. Collagen VI-related myopathy is caused by the pathogenic variants of COL6A1, COL6A2, and COL6A3. *Three* genes encode the α1, α2, and α3 chains of collagen VI, which are major components of the muscle extracellular matrix (ECM) forming a microfibrillar network and basement membrane.[5,6] The 3 collagen VI alpha chains assemble into tetramers before being secreted into the ECM. Pathogenic variants in any one of the 3 collagen VI genes result in either the loss or misfolding of collagen VI in the muscle ECM.[1] However, the exact mechanism by which collagen VI deficiency or misfolding in the muscle ECM leads to myopathy is not fully understood.
Microarray analysis and RNA-sequencing (RNA-seq) are useful gene expression platforms, which show molecular changes in biological samples and define the biological pathways involved in disease pathogenesis.[7,8] Several studies have analyzed gene expression profiles using microarray analysis and RNA-seq in muscles or cells of patients and mouse models with collagen VI-related myopathy.[9–11] Collagen VI-related myopathy has been reported several times in Korea.[12–16] Bethlem myopathy, a mild phenotype of collagen VI-related myopathy, is the most common cause of congenital myopathy in Korea.[13] To identify the pathogenic mechanism of Bethlem myopathy, we comprehensively analyzed the gene expression profiles in the skeletal muscles of patients with Bethlem myopathy using RNA-seq.
## 2.1. Study subjects
We reviewed the medical records of the myopathy database from January 2002 to August 2021. We selected 6 muscle samples from 3 patients with Bethlem myopathy (MF226, MF1014, and MF1474) and 3 control subjects. Table 1 summarizes the clinical and genetic spectra of the study participants. Three patients with Bethlem myopathy showed typical clinical presentations, including proximal muscle weakness and multiple contractures, and were genetically confirmed. The variants carried by these 3 patients were pathogenic or likely pathogenic according to the American College of Medical Genetics and Genomics and the Association for Molecular Pathology guidelines,[17] and have been previously reported as pathogenic variants.[13,18] Three control subjects were selected from patients based on the following criteria: Psychogenic weakness; Normal muscle pathology; *Normal serum* creatine kinase level, and; Boys or girls under 20 years of age. The study was conducted following the declaration of Helsinki and institutional criteria. The research protocol was approved by the institutional review board of Gangnam Severance Hospital, Korea (IRB No: 3-2021-0300). Written informed consents were obtained from all participants. For participants below the age of 16, written informed consent were obtained from their parents.
**Table 1**
| Subject | Sex | Variant | Age at diagnosis, yr | Age at onset, yr | Inheritance | Clinical presentation | CK, IU/l | Muscle pathology |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| MF226 | F | COL6A2: c.856-1G > C | 10 | 6 | Autosomal dominant | Prominent contractures of ankle and interphalangeal joints, and mild proximal muscle weakness | 488 | Increased internal nuclei, increased size variability, atrophic fibers, and a few degenerating fibers |
| MF1014 | M | COL6A1: c.1056 + 2dup | 18 | 12 | Sporadic | Proximal muscle weakness and contractures of ankle and interphalangeal joints | 432 | Myopathic changes with degenerating fibers and interstitial fibrosis |
| MF1474 | F | COL6A1: c.868G > C (p.G290R) | 16 | 2 | Autosomal dominant | Proximal muscle weakness, multiple contractures, and scoliosis | 76 | Increased fiber size variation, hypertrophic fibers, and type I predominance |
| Control 1 | M | - | 18 | - | - | Psychogenic weakness | 84 | Normal finding |
| Control 2 | M | - | 14 | - | - | Psychogenic weakness | 137 | Normal finding |
| Control 3 | F | - | 16 | - | - | Psychogenic weakness | 54 | Normal finding |
## 2.2. RNA-seq
The total RNA concentration was calculated using the Quant-iTTM Ribogreen RNA assay kit (Thermo Fisher Scientific, MA). To assess the integrity of the total RNA, 6 samples were run on a TapeStation RNA ScreenTape device (Agilent, Santa Clara, CA). Only high quality RNA preparations, with RNA integrity number values > 7.0, were used for RNA library construction. A library was prepared with 1 µg of total RNA from each sample using the Illumina TruSeq Stranded mRNA Sample Prep kit (Illumina, Inc., San Diego, CA). The first step in the workflow involves purifying the poly-A-containing mRNA molecules using poly-T oligo-attached magnetic beads. Following purification, mRNA was fragmented using divalent cations at elevated temperatures. The cleaved RNA fragments were copied into the first-strand complementary DNA (cDNA) using Superscript II reverse transcriptase (Thermo Fisher Scientific, MA), and random primers. This was followed by second-strand cDNA synthesis using DNA polymerase I, RNase H, and dUTP. These cDNA fragments then went through an end repair process, the addition of a single “A” base, and ligation of the indexing adapters. The products were purified and enriched using polymerase chain reaction (PCR) to create the final cDNA library. The libraries were quantified using quantitative PCR (qPCR) according to the qPCR Quantification Protocol Guide (KAPA Library Quantification kits for Illumina Sequencing platforms) and analyzed using the D1000 Screen Tape assay (Agilent Technologies, Waldbronn, Germany). Indexed libraries were sequenced on the NovaSeq 6000 platform (Illumina, San Diego, CA).
## 2.3. Statistical analysis of gene expression levels
The quality of the RNA-seq data was evaluated using Fast QC, and all were determined to be of high quality. The total number of reads was 112390858, 103286098, 121022280, 101810524, 109786544, and 118204882 in the skeletal muscles from MF226, MF1014, MF1474, control 1, control 2, and control 3, respectively. There were no significant differences in the sequencing depth or mapping efficiency between the 2 groups. We analyzed similarities and patterns among samples using hierarchical clustering and principal component analysis. Gene expression levels and volume plots were analyzed as previously described.[19] Gene enrichment and functional annotation analysis for the significant genes was performed using gene ontology (GO) (www.geneontology.org/), and pathway analysis for the differentially expressed genes was performed based on the kyoto encyclopedia of genes and genomes (KEGG) pathways (http://www.genome.jp/kegg/pathway.html).
## 2.4. Availability of data and materials
The datasets generated and/or analyzed during the current study are deposited in the NCBI sequence read archive database (http://www.ncbi.nlm.nih.gov/bioproject/796623), under the accession numbers PRJNA796623.
## 3. Results
The heat map for hierarchical clustering showed separation between the myopathy and control groups (Fig. 1A). Principal component analysis also showed a separate clustering of the samples by group (Fig. 1B). Among the 35,993 identified transcripts, 17,931 transcripts with 0 fragments per kilobase of transcripts per million fragments mapped were excluded. Of the remaining 18,062 transcripts with nonzero fragments per kilobase of transcripts per million, 187 transcripts were significantly differentially expressed (|fold change| ≥2, $P \leq .05$), with 157 upregulated and 30 downregulated transcripts in the Bethlem group (see Table S1, Supplemental Digital Content, http://links.lww.com/MD/I549, which represented the list of genes). A flow diagram for the identification of candidate genes is shown in Figure S1, Supplemental Digital Content, http://links.lww.com/MD/I550. The top 20 upregulated and downregulated genes are summarized in Figure 2. Among the noncoding RNAs, microRNA (miR)-133b was significantly upregulated, and 4 long intergenic non protein coding RNAs (lncRNAs), LINC01854, MBNL1-AS1, LINC02609, and LOC728975, were significantly downregulated.
**Figure 1.:** *Heatmap for hierarchical clustering and principal component analysis (PCA) of the myopathy and control groups. (A) The heatmap for hierarchical clustering was generated using the R package “plots” using the expression for each gene (rows) and sample (columns). The expression levels of each gene across samples are shown as Z-scores scaled by their fragments per kilobase of transcript per million mapped reads (FPKMS) values from RNA-seq. The scaled expression values are color-coded according to the legend. The dendrogram depicting hierarchical clustering is based on the expression levels of all genes. (B) PCA of gene expression in skeletal muscles from the myopathy and control groups. Samples are color-coded according to the legend (blue circles: the Bethlem group; red circles: the control group). The numbers in blankets correspond to the proportion of variance explained by the respective principal component. RNA-seq = RNA-sequencing.* **Figure 2.:** *Top 20 genes that are significantly upregulated and downregulated with > 2-fold change.*
The differentially expressed genes were categorized using GO. We identified 404 enriched GO terms in the Bethlem group versus the control group (false discovery rate-adjusted $P \leq .05$). The top 20 enriched GO terms came from categories centered on the ECM and are listed in Supplemental Digital Content (see Table S2, Supplemental Digital Content, http://links.lww.com/MD/I551). The KEGG pathway enrichment reflected themes with significant enrichment of the ECM-receptor interaction (has04512), complement and coagulation cascadehashsa04610), and focal adhehasn (hsa04510) in the Bethlem group (Table 2 and see Figure S2, Supplemental Digital Content, http://links.lww.com/MD/I552). Among the genes that mapped the complement and coagulation chasades (hsa04610), ten transcripts (SERPING1, C1QA, C1QB, C1QC, C1R, C1S, C3, F3, F13A1, and CFH) were significantly upregulated in the Bethlem group. Among the genes that mapped the ECM-receptor ihasraction (hsa04512), 7 transcripts (CD44, COL6A1, COL6A3, FN1, HSPG2, THBS3, and TNXB) were significantly upregulated in the Bethlem group (Table 3). COL6A2 was also upregulated in the Bethlem group, but the difference was not statistically significant ($$P \leq .06$$). Three other collagen genes, COL12A1, COL14A01, and COL16A1 were also significantly upregulated in the Bethlem group (Table 3).
## 4. Discussion
Our results showed the gene expression profiles of skeletal muscles in patients with Bethlem myopathy. Bethlem myopathy was genetically confirmed, and the control subjects exhibited no clinical, pathological, or electrophysiological neuromuscular disorders. Although the sample size was small, skeletal muscles were obtained from both the myopathy and control group patients were teenagers. Therefore, our study was able to demonstrate the molecular changes in the skeletal muscles of patients with Bethlem myopathy.
Our study showed 157 upregulated and 30 downregulated transcripts in the Bethlem group. Among these, the upregulated transcripts were involved in the differentiation and regeneration of skeletal muscles, stabilizing muscle membranes, and major ECM components. In our study, the most significantly upregulated gene was TNNT2, which was associated with muscle regeneration. This is consistent with previous study findings.[10] nicotinamide N-methyltransferase (NNMT) was the second most highly upregulated gene. The NNMT, encoded by NNMT, is an enzyme that is a major consumer of NAD+, involved in metabolic derangement in various tissues.[20] Under normal conditions, NNMT is expressed mainly in the liver and skeletal muscles. Dysregulation of NNMT leads to reduced expression of nicotinamide phosphoribosyl transferase, a major enzyme involved in NAD+ biosynthesis. Since the cofactor NAD+ plays an important role in stabilizing muscles from metabolic and structural degeneration, NNMT is upregulated in Duchene muscular dystrophy, other muscle diseases, and mdx mice.[21] Among the noncoding RNAs, miR-133b was significantly upregulated in the Bethlem group. miR-133b promotes fibrosis in renal cells and is associated with chronic chagas disease and dilated cardiomyopathy.[22,23] One study previously suggested that noncoding RNAs, miR-181a and miR-30c, may play an important role in regulating gene expression in patients with Ullrich congenital muscular dystrophy; however, there has been no evaluation of miR-133b so far.[9] In the present study, miR-181a and miR-30c were not significantly expressed.
Our results also revealed several downregulated transcripts. NANOS1 was the most significantly downregulated protein coding gene. NANOS1 controls cell cycle progression and affects the SMAD Family Member 3/transforming growth factor (TGF)-β fibroblast maturation pathway.[24,25] One group previously identified significant GO categories of downregulated genes associated with posttranscriptional regulation of gene expression in patients with Ullrich congenital muscular dystrophy compared with controls, among which NANOS1 was identified.[10] The second most downregulated protein coding gene was LRRC3B, which is a tumor suppressor.[26] Indoleamine 2, 3-dioxygenase, encoded by IDO1, catalyzes the first and rate-limiting step in tryptophan catabolism to kynurenine, and plays an important role in tumor cell evasion of the immune system.[27] lncRNAs are known to play essential roles in the proliferation and apoptosis of cells, particularly cancer cells.[28] Our study showed that these 4 lncRNAs were significantly downregulated. LINC02609 was previously reported as a key lncRNA associated with distant metastasis and poor prognosis in patients with clear cell renal cell carcinoma.[29] MBNL1-AS1 is also known to repress proliferation and enhance apoptosis in bladder cancer cells.[30] However, our results could not be confirmed by previous studies.
The KEGG pathway analysis showed significant enrichment of the ECM-receptor interaction (hsa04512), complement and coagulation cascades (hsa04610), and focal adhesion (hsa04510) in the Bethlem group. The ECM-receptor interaction pathway has been shown to be significantly differentially expressed in many studies.[7,9–11] However, the 3 collagen-6-related genes (COL6A1, COL6A2, and COL6A3) were highly elevated in the Bethlem group, regardless of whether the mechanism of variants had a loss of function or a dominant-negative effect. This is contrary to the results of a previous study that showed that the expression of COL6A1, COL6A2, and COL6A3 depends on the inheritance mechanism.[31] *It is* known that the complement and coagulation cascades (hsa04610) and focal adhesion (hsa04510) are linked to wound healing, and that collagen VI plays an important role in this process.[32,33] The relationship between collagen VI-related myopathy and wound healing has been previously reported in dermal fibroblasts from patients with Ullrich congenital muscular dystrophy.[9] However, our results did not identify differences in several previously reported pathways, including the cell cycle (hsa04110), DNA replication (hsa03030), nitrogen metabolism (hsa00910), TGF-beta signaling pathway (hsa04350), arrhythmogenic right ventricular cardiomyopathy (hsa05412), hypertrophic cardiomyopathy (hsa05410), viral myocarditis (hsa05416), hematopoietic cell lineage (hsa04640), renin-angiotensin system (hsa04614), or circadian rhythm (hsa04710).[9,11,31,34] Among them, the TGF-β signaling pathway (hsa04350) showed significant changes in several studies.[9,11,31,34] The difference between the results of this study and those of previous studies can be attributed to a variety of reasons. First, most studies, including ours, were performed with a small number of samples. Second, the subject specimens were diverse, such as the patients muscles, cultured fibroblasts, and mouse model muscles. Third, patient age and disease severity may have influenced the results.
Our study has several limitations. The major limitation was the small number of muscle samples from the patients with Bethlem myopathy and controls. Second, we could not use age- or sex-matched controls. If the number of patients with Bethlem myopathy is large to allow studies to subgroup them according to age, sex, clinical severity, mutation type, and we could further reinforce and complement our results.
## 5. Conclusion
In conclusion, we confirmed that Bethlem myopathy is strongly associated with the production and organization of ECM and the wound healing process. Additionally, our results suggest that the pathogenesis of Bethlem myopathy is influenced by several nonprotein-coding genes.
## Acknowledgments
The authors would like to thank the patients for their help and involvement in this study.
## Author contributions
Conceptualization: Seung-Ah Lee, Hyung Jun Park, Young-Chul Choi.
Data curation: Ji-Man Hong.
Formal analysis: Jung Hwan Lee.
Software: Hyung Jun Park.
Supervision: Young-Chul Choi.
Validation: Seung-Ah Lee, Ji-Man Hong.
Writing – original draft: Seung-Ah Lee, Jung Hwan Lee.
Writing – review & editing: Seung-Ah Lee, Hyung Jun Park.
## References
1. Bönnemann CG. **The collagen VI-related myopathies: muscle meets its matrix.**. *Nat Rev Neurol* (2011) **7** 379-90. PMID: 21691338
2. Vanegas OC, Bertini E, Zhang R-Z. **Ullrich scleroatonic muscular dystrophy is caused by recessive mutations in collagen type VI.**. *Proc Natl Acad Sci USA* (2001) **98** 7516-21. PMID: 11381124
3. Ullrich O. **Kongenitale, atonisch-sklerotische Muskeldystrophie.**. *Monatsschr Kinderheilkd* (1930) **47** 502-10
4. Bethlem J, Wijngaarden GK. **Benign myopathy, with autosomal dominant inheritance. a report on three pedigrees.**. *Brain* (1976) **99** 91-100. PMID: 963533
5. Lampe A, Dunn D, Von Niederhausern A. **Automated genomic sequence analysis of the three collagen VI genes: applications to Ullrich congenital muscular dystrophy and Bethlem myopathy.**. *J Med Genet* (2005) **42** 108-20. PMID: 15689448
6. Doriguzzi C, Palmucci L, Mongini T. **Congenital muscular dystrophy associated with familial junctional epidermolysis bullosa letalis.**. *Eur Neurol* (1993) **33** 454-60. PMID: 8307068
7. Mantione KJ, Kream RM, Kuzelova H. **Comparing bioinformatic gene expression profiling methods: microarray and RNA-Seq.**. *Med Sci Monit Basic Res* (2014) **20** 138-42. PMID: 25149683
8. Hänzelmann S, Castelo R, Guinney J. **GSVA: gene set variation analysis for microarray and RNA-seq data.**. *BMC Bioinf* (2013) **14** 1-15
9. Paco S, Casserras T, Rodríguez MA. **Transcriptome analysis of Ullrich congenital muscular dystrophy fibroblasts reveals a disease extracellular matrix signature and key molecular regulators.**. *PLoS One* (2015) **10** e0145107. PMID: 26670220
10. Paco S, Kalko SG, Jou C. **Gene expression profiling identifies molecular pathways associated with collagen VI deficiency and provides novel therapeutic targets.**. *PLoS One* (2013) **8** e77430. PMID: 24223098
11. Scotton C, Bovolenta M, Schwartz E. **Deep RNA profiling identified CLOCK and molecular clock genes as pathophysiological signatures in collagen VI myopathy.**. *J Cell Sci* (2016) **129** 1671-84. PMID: 26945058
12. Lee JH, Shin HY, Park HJ. **Clinical, pathologic, and genetic features of collagen VI-related myopathy in Korea.**. *J Clin Neurol* (2017) **13** 331-9. PMID: 28831785
13. Park H, Jang H, Kim J. **Discovery of pathogenic variants in a large Korean cohort of inherited muscular disorders.**. *Clin Genet* (2017) **91** 403-10. PMID: 27363342
14. Park HJ, Choi YC, Kim SM. **Molecular genetic diagnosis of a bethlem myopathy family with an autosomal-dominant COL6A1 Mutation, as evidenced by exome sequencing.**. *J Clin Neurol* (2015) **11** 183-7. PMID: 25749816
15. Kim SY, Kim WJ, Kim H. **Collagen VI-related myopathy: expanding the clinical and genetic spectrum.**. *Muscle Nerve* (2018) **58** 381-8. PMID: 29406609
16. Choi E, Shin S, Lee S. **Coexistence of digenic mutations in the collagen VI genes (COL6A1 and COL6A3) leads to Bethlem myopathy.**. *Clin Chim Acta* (2020) **508** 28-32. PMID: 32389683
17. Richards S, Aziz N, Bale S. **Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.**. *Genet Med* (2015) **17** 405-24. PMID: 25741868
18. Giusti B, Lucarini L, Pietroni V. **Dominant and recessive COL6A1 mutations in Ullrich scleroatonic muscular dystrophy.**. *Ann Neurol* (2005) **58** 400-10. PMID: 16130093
19. Park HJ, Hong J-M, Lee JH. **Comparative transcriptome analysis of skeletal muscle in ADSSL1 myopathy.**. *Neuromuscul Disord* (2019) **29** 274-81. PMID: 30853170
20. Komatsu M, Kanda T, Urai H. **NNMT activation can contribute to the development of fatty liver disease by modulating the NAD (+) metabolism.**. *Sci Rep* (2018) **8** 8637. PMID: 29872122
21. Ryu D, Zhang H, Ropelle ER. **NAD+ repletion improves muscle function in muscular dystrophy and counters global PARylation.**. *Sci Transl Med* (2016) **8** 361ra139
22. Sun Z, Ma Y, Chen F. **miR-133b and miR-199b knockdown attenuate TGF-β1-induced epithelial to mesenchymal transition and renal fibrosis by targeting SIRT1 in diabetic nephropathy.**. *Eur J Pharmacol* (2018) **837** 96-104. PMID: 30125566
23. Ferreira LRP, Frade AF, Santos RHB. **MicroRNAs miR-1, miR-133a, miR-133b, miR-208a and miR-208b are dysregulated in chronic chagas disease cardiomyopathy.**. *Int J Cardiol* (2014) **175** 409-17. PMID: 24910366
24. Li B, Chen H, Yang X. **Knockdown of eIF3a ameliorates cardiac fibrosis by inhibiting the TGF-β1/Smad3 signaling pathway.**. *Cell Mol Biol (Noisy-le-grand)* (2016) **62** 97-101
25. Dong Z, Liu Z, Cui P. **Role of eIF3a in regulating cell cycle progression.**. *Exp Cell Res* (2009) **315** 1889-94. PMID: 19327350
26. Li GS, Kong GY, Zou Y. **Protective role of LRRC3B in preventing breast cancer metastasis and recurrence post-bupivacaine.**. *Oncol Lett* (2017) **14** 5013-7. PMID: 29085514
27. Cheong JE, Sun L. **Targeting the IDO1/TDO2–KYN–AhR pathway for cancer immunotherapy–challenges and opportunities.**. *Trends Pharmacol Sci* (2018) **39** 307-25. PMID: 29254698
28. Jin J, Wang H, Zheng X. **Inhibition of LncRNA MALAT1 attenuates cerebral ischemic reperfusion injury via regulating AQP4 expression.**. *Eur Neurol* (2020) **83** 581-90. PMID: 33130678
29. Su Y, Zhang T, Tang J. **Construction of competitive endogenous RNA network and verification of 3-Key LncRNA signature associated with distant metastasis and poor prognosis in patients with clear cell renal cell carcinoma.**. *Front Oncol* (2021) **11** 513
30. Wei X, Yang X, Wang B. **Lnc RNA MBNL 1-AS 1 represses cell proliferation and enhances cell apoptosis via targeting miR-135a-5p/PHLPP 2/FOXO 1 axis in bladder cancer.**. *Cancer Med* (2020) **9** 724-36. PMID: 31769229
31. Butterfield RJ, Dunn DM, Hu Y. **Transcriptome profiling identifies regulators of pathogenesis in collagen VI related muscular dystrophy.**. *PLoS One* (2017) **12** e0189664. PMID: 29244830
32. Theocharidis G, Drymoussi Z, Kao AP. **Type VI collagen regulates dermal matrix assembly and fibroblast motility.**. *J Investig Dermatol* (2016) **136** 74-83. PMID: 26763426
33. Oono T, Specks U, Eckes B. **Expression of type VI collagen mRNA during wound healing.**. *J Investig Dermatol* (1993) **100** 329-34. PMID: 8440917
34. Guadagnin E, Mohassel P, Johnson KR. **Transcriptome analysis of collagen VI-related muscular dystrophy muscle biopsies.**. *Ann Clin Transl Neurol* (2021) **8** 2184-98. PMID: 34729958
|
---
title: A new immune checkpoint-associated nine-gene signature for prognostic prediction
of glioblastoma
authors:
- Xiao Jin
- Xiang Zhao
journal: Medicine
year: 2023
pmcid: PMC9981394
doi: 10.1097/MD.0000000000033150
license: CC BY 4.0
---
# A new immune checkpoint-associated nine-gene signature for prognostic prediction of glioblastoma
## Abstract
Glioblastoma (GBM) is a highly malignant neurological tumor that has a poor prognosis. While pyroptosis affects cancer cell proliferation, invasion and migration, function of pyroptosis-related genes (PRGs) in GBM as well as the prognostic significance of PRGs remain obscure. By analyzing the mechanisms involved in the association between pyroptosis and GBM, our study hopes to provide new insights into the treatment of GBM. Here, 32 out of 52 PRGs were identified as the differentially expressed genes between GBM tumor versus normal tissues. And all GBM cases were assigned to 2 groups according to the expression of the differentially expressed genes using comprehensive bioinformatics analysis. The least absolute shrinkage and selection operator analysis led to the construction of a 9-gene signature, and the cancer genome atlas cohort of GBM patients were categorized into high risk and low risk subgroups. A significant increase in the survival possibility was found in low risk patients in comparison with the high risk ones. Consistently, low risk patients of a gene expression omnibus cohort displayed a markedly longer overall survival than the high risk counterparts. The risk score calculated using the gene signature was found to be an independent predictor of survival of GBM cases. Besides, we observed significant differences in the expression levels of immune checkpoints between the high risk versus low risk GBM cases, providing instructive suggestions for immunotherapy of GBM. Overall, the present study developed a new multigene signature for prognostic prediction of GBM.
## 1. Introduction
Glioblastoma (GBM) is a highly aggressive central nervous system malignant tumor with high mortality and bad prognosis.[1] Worse still, GBM represents around $57\%$ of all gliomas as well as $48\%$ of all primary brain malignancies and its median survival is <2 years.[2] The conventional strategy for treatment of the newly diagnosed GBM patients involves surgery combined with concurrent radiotherapy using temozolomide followed by adjuvant temozolomide. Alternatively, adjuvant temozolomide can be administered concurrently with cancert-treating fields, delivering low intensity alternating electric fields.[2] *While a* number of biomarkers and gene signatures show the promising potential for prognostic prediction of GBM, there still exists a gap between research and clinical practice. Hence, it is of clinical significance to identify gene signatures for predicting the prognosis of GBM. Given that significant alterations of the neurovascular unit occur in central nervous system malignancies including GBM, some potential immunotherapies for GBM may be available.[3] Pyroptosis, a type of programmed lytic cell death, exhibits characteristic features including swelling and rupture of cells, cellular content release, as well as significant pro-inflammation effects. During pyroptosis, inflammasomes perceive danger signaling and cellular events, eliciting caspase activation, Gasdermin D cleavage, and release of IL-18 and IL-1β.[4] And it can suppress tumor occurrence and progression, while forming a microenvironment for providing nutrients to cancer tissues and promoting tumor growth.[5] Emerging evidence shows that it functions in tumor cell proliferation, invasion, and migration, influencing cancer prognosis.[6,7] For instances, pyroptosis induces apoptosis of tumor cells in digestive tract. And NLRs3, AIM24, and GSDM5 family are critically involved in pyroptosis-related signaling in digestive cancer, including esophageal cancer, gastric cancer, and colitis-associated colorectal cancer.[8] Recently, a study identified a new gene signature for prognostic prediction of ovarian cancer which is related to pyroptosis.[9] Meanwhile, a prognostic signature for lung adenocarcinoma was found to contain 5 pyroptosis-related genes (PRGs) including NLRP1, NLRP2, NLRP7, NOD1, and CASP6. And lncRNA KCNQ1OT1, miR-335-5p, NLRP1, and NLRP7 were shown to form a regulatory axis that may be critically implicated in lung cancer progression.[10] To date, research reports on relationship between pyroptosis and GBM are still lacking.
Herein, we examined the expression profiles of PRGs using bioinformatics approaches and divided the GBM cases of the the cancer genome atlas (TCGA) or gene expression omnibus (GEO) cohort into 2 subtypes based on the risk score. Strikingly, we found a novel gene signature with a prognostic significance which was related to immune checkpoint expression in GBM. This study may facilitate further investigation on prognostic prediction and treatment of GBM.
## 2.1. Datasets
RNA sequencing and clinical data of 169 GBM cases as well as 5 normal individuals from TCGA database were downloaded (portal.gdc.cancer.gov/repository). A GEO cohort of GBM patients (GSE83300) were used to validate the risk model. Patients in this cohort underwent a shorter duration of follow up (up to 3 years) than those in the TCGA cohort.
## 2.2. Identification of the differentially expressed genes (DEGs)
As presented in Table S1, Supplemental Digital Content, http://links.lww.com/MD/I576, a total of 52 PRGs were chosen for this study.[9,11–13] Out of the 52 PRGs, 32 were identified as the DEGs using the TCGA cohort (Table S2, Supplemental Digital Content, http://links.lww.com/MD/I577). Normalization of the expression data to fragment per kilobase million values was carried out prior to comparative analysis. Identification of DEGs was carried out using the package Limma at a value of $P \leq .05.$ The DEGs were notated as below: *** $P \leq .001$, ** $P \leq .01$, and * $P \leq .05.$ search tool for the Retrieval of Interacting Genes version 11.5 (https://cn.string-db.org/) was employed to construct a protein–protein interaction (PPI) network.
## 2.3. Generation and external validation of the novel gene signature
To determine prognostic relevance of DEGs, Cox regression analysis was conducted to investigate association of the PRGs with survival condition of GBM cases in the TCGA cohort. A cutoff P value was set at 0.2 to avoid omissions. Twenty survival-associated genes were screened out for subsequent studies (Table S3, Supplemental Digital Content, http://links.lww.com/MD/I578). Thereafter, least absolute shrinkage and selection operator (LASSO) analysis was carried out with the package glmnet to further screen the candidates and develop a gene model for prognostic prediction. Finally, 9 genes and the relevant coefficients were selected, and the penalty parameter (λ) was chosen based on the minimum criteria. *The* gene expression data were subjected to standardization and centralization by implementing R function scale, and risk score calculation was then performed as follows: Risk score=∑7i Xi×Yi (X and Y represent the coefficient and expression level, respectively). GBM cases of the TCGA cohort were divided into low risk and high risk subgroups using the median risk score. And Kaplan–*Meier analysis* was carried out to comparatively analyze the overall survival (OS) between the 2 subgroups. Nine-gene signature based principal component analysis (PCA) was conducted by the R Stats Package function prcomp. Three-year receiver operating characteristic (ROC) curve analysis was performed by using the R packages survival, survminer, and time-ROC. A GEO cohort of GBM cases was used to validate the prognostic model. DEG expression level in the GEO cohort was subjected to normalization using the function scale, and the same formula as that for the TCGA cohort was used to calculate the risk score. Like the TCGA cohort, the GEO cohort of patients were assigned to the 2 different groups according to the median risk score, and the multigene signature was validated by a comparative analysis between the 2 groups. Besides, decision curve analysis (DCA) was conducted to assess the novel signature.
## 2.4. Assessment of prognostic significance of the risk score
Clinical data of GBM cases in the TCGA cohort were extracted. And we undertook the univariate and multivariable regression analyses to determine correlations of clinical characteristics with the risk score.
## 2.5. Predictive nomogram
A nomogram represents a graphical statistical tool for quantitative assessment of risk of subjects in a clinical setting integrated with several risk factors. We developed a nomogram by integrating the prognostic signatures to evaluate 1-year, 2-year, and 3-year OS of GBM cases.
## 2.6. Functional characterization of DEGs
The GBM cases were categorized into the low risk and high risk group using the median risk score. Identification of DEGs between the 2 groups was performed using the criteria of FDR < 0.05 and |log2FC| ≥ 1, and DEGs were then subjected to gene ontology and *Kyoto encyclopaedia* of genes and genomes analyses using package cluster Profiler.
## 2.7. Immune activity analysis
The enrichment scores of immune cell subpopulations as well as the activities of immune associated pathways were examined by performing single-sample gene set enrichment analysis with the package GSVA. Besides, potential immune checkpoints were retrieved from previous studies.
## 2.8. Statistics
Comparison of the gene expression level between GBM versus normal brain tissues was made using single factor analysis of variance, and Pearson chi-square test was conducted to comparatively analyze categorical variables. Both Kaplan–Meier method and the log-rank test were employed to carry out inter group comparison of the OS. The univariate and multivariable analyses were used to assess prognostic significance of the gene signature. Significance of the model in prognostic prediction of GBM was determined by using DCA. Besides, comparison of immune cell infiltration, activation of immune associated pathways, and the expression level of immune checkpoints between the 2 subgroups was made using the Mann–Whitney test. Data were statistically analyzed with R software v4.1.0 (PMID: 36155484). Figure 1 presents the work-flow diagram in this research.
**Figure 1.:** *The work-flow diagram.*
## 3.1. Screening of DEGs between tumor versus normal tissues
We comparatively analyzed the expression of 52 PRGs among 169 GBM samples and 5 normal tissue samples from the TCGA. Among them, 32 were identified as the DEGs, of which 4 (NLRP2, NLRP7, NLRP1, and PRKACA) were downregulated and the remaining 28 (CHMP6, CASP9, SCAF11, HMP4A, HMGB1, CHMP2A, IRF2, BAK1, NOD1, NLRC4, CASP8, GSDME, IRF1, BAX, CASP3, GZMB, IL18, PYCARD, gasdermin D (GSDMD), CASP6, CASP5, NOD2, TP53, AIM2, GSDMA, CASP1, CASP4, and GZMA) were up-regulated in the tumor samples (Fig. 2A). We further carried out PPI analysis to investigate the interactions among the 32 DEGs. As depicted in Figure 2B, the PPI analysis identified CASP1, GSDMD, NLRP1, AIM2, PYCARD, CASP8, CASP5, TP53, and CASP3 as hub genes. Moreover, we constructed a correlation network for the DEGs (Fig. 2C).
**Figure 2.:** *The expression and interaction analyses of 32 pyroptosis-associated DEGs. (A) A heatmap showing the relative expression levels of the DEGs between tumor versus normal tissues. Color scale: red, high expression; blue, low expression. **P < .01 and ***P < .001. (B) Interaction analysis of DEGs based on a PPI network. The highest confidence of the minimum required interaction score was 0.9. (C) Correlation analysis of DEGs. The positive and negative correlations were indicated by red and blue lines, respectively. Color depth corresponded to the correlation degree. DEGs = differentially expressed genes, PPI = protein–protein interaction.*
## 3.2. PRG-based tumor clustering
To determine the correlation of the PRG expression with GBM subgrouping, we conducted the consensus clustering analysis on the TCGA cohort of GBM cases. As depicted in Figure 3A, the highest intragroup correlations as well as low intergroup correlations were observed at the clustering variable (k) of 2, showing that all 169 GBM cases can be divided into 2 clusters according to PRG expression levels. Meanwhile, we identified 484 DEGs between the 2 clusters (Table S4, Supplemental Digital Content, http://links.lww.com/MD/I579). Notably, no marked differences in clinical characteristics including survival status, gender and age were detected between the 2 clusters (Fig. 3B). By contrast, a marked difference in the OS was detected between them (Fig. 3C).
**Figure 3.:** *PRG-based clustering of GBM cases. (A) 169 GBM cases were assigned to 2 groups according to consensus clustering matrix for k = 2. (B) A heatmap showing clinical characteristics of the GBM cases. (C) Kaplan–Meier curves of OS. GBM = glioblastoma, OS = overall survival, PRGs = pyroptosis-related genes.*
## 3.3. Establishment of a gene model for prognostic prediction
We next carried out univariate Cox regression analysis on the 159 GBM patients with complete survival data to screen for survival associated genes. As illustrated in Figure 4A, 9 genes (GZMB, AREG, LOXL1, MSTN, PTX3, IGFBP6, STC1, POM121L9P, and TGM2) were chosen for further investigation according to the criterion of $P \leq .2.$ Among the 9 genes, 8 (GZMB, AREG, LOXL1, PTX3, IGFBP6, STC1, POM121L9P, and TGM2) were found to be correlated with an increased risk (hazard ratios, HRs > 1), and the remaining 1 (MSTN) was identified as a protective gene (HRs < 1). LASSO analysis using the optimum λ value led to the construction of a 9-gene signature (Fig. 4B and C). We calculated the risk score by the following formula: Risk score = (0.181*GZMB exp.) + (0.104* AREG exp.) + (0.053*LOXL1 exp.) + (−0.115*MSTN exp.) + (0.016* PTX3 exp.) + (0.036*IGFBP6 exp.) + (0.003*STC1 exp.) + (0.247*POM121L9P exp.) + (0.113*TGM2 exp.). One hundred-nine GBM cases were equally assigned to the high risk or low risk group based on the median risk score (Fig. 4D). Strikingly, PCA analysis revealed a clear separation of the 2 groups with a distinct risk score (Fig. 4E). In comparison with low risk cases, the high risk ones displayed more deaths as well as shortened survival time (Fig. 4F). Moreover, a marked difference in OS was observed between the 2 groups of patients (Fig. 4G). To further assess the prognostic model, we conducted ROC curve analysis. As depicted in Figure 4H, the area under the ROC curve was 0.732, 0.726, and 0.761 for 1-year, 2-year, and 3-year survival, respectively.
**Figure 4.:** *Construction of a multigene signature based on the TCGA dataset. (A) Identification of 9 survival-related PRGs by the univariate regression method (P < .2). (B) Construction of a 9-gene signature using LASSO regression. (C) Tuning parameter selection by cross validation in the LASSO model. (D) Risk score-based distribution of the GBM cases. (E) Score plot of PCA on the risk of the GBM patients. (F) Comparative analysis on the survival status between the 2 groups of patients. (G) Kaplan–Meier curves of the GBM cases. (H) ROC curve analysis of the gene model. GBM = glioblastoma, LASSO = least absolute shrinkage and selection operator, PCA = principal component analysis, PRGs = pyroptosis-related genes, ROC = receiver operating characteristic, TCGA = the cancer genome atlas.*
## 3.4. Validation of the prognostic multigene signature using a different cohort of GBM cases
A GEO cohort of 50 GBM cases (GSE83300) was used to conduct external validation of the prognostic signature. The expression profiles of PRGs were subjected to normalization using the “Scale” function prior to analysis. Among the 50 GEO patients, 26 and 24 were assigned to the high risk and low risk group, respectively, according to the median risk score derived from the TCGA cases (Fig. 5A). PCA analysis revealed a good separation between the 2 groups (Fig. 5B). As shown in Figure 5C, the low risk cases exhibited increased survival time as well as reduced mortality rate in comparison with the high risk cases. Besides, a marked difference in OS was identified between the 2 categories of patients (Fig. 5D). Notably, the area under the ROC curve was found to be 0.683, 0.721, and 0.648 for 1-year, 2-year, and 3-year survival, respectively, indicating that the model has good predictive efficacy for the GEO cohort (Fig. 5E).
**Figure 5.:** *Validation of the gene model. (A) GBM cases of a GEO cohort were divided into the low risk and high risk groups. (B) Score plot of PCA on the risk of GBM patients. (C) Comparative analysis on survival status between the 2 groups. (D) Kaplan–Meier curves of the 2 groups of cases. (E) The ROC analyses of the model. GBM = glioblastoma, GEO = gene expression omnibus, PCA = principal component analysis, ROC = receiver operating characteristic.*
## 3.5. The value of the signature model in prognostic prediction
To further examine the role of the signature model derived risk score in prognostic prediction, we conducted both univariate and multivariable regression analyses. As depicted in Figure 6A, the univariate regression analysis identified the risk score as an independent predictor for poor survival of GBM patients (HR = 3.439, $95\%$ confidence interval: 2.331 − 5.072). Meanwhile, the multivariable analysis revealed that the risk score serves as a prognostic predictor of the patients (HR = 3.239, $95\%$ confidence interval: 2.166 − 4.844) (Fig. 6B). No marked differences in clinical features were detected between the 2 groups of patients, as indicated in the heatmap (Fig. 6C). We further performed DCA analysis to determine the sensitivity and specificity of the multigene model by comparing with clinicopathological characteristics of the GBM cases. As illustrated in Figure 6D, the 9-gene signature exhibited a better performance in the prognostic prediction than the clinicopathological characteristics. Besides, the nomogram combining clinical features with the 9-gene signature produced a stable and accurate prediction of prognosis (Fig. 6E), indictive of a significant clinical value.
**Figure 6.:** *The univariate and multivariable regression analyses of the risk score. (A) The univariate analysis. (B) The multivariable analysis. (C) Heatmap showing the differences in clinicopathologic characteristics between the 2 groups of patients. *P < .05. (D) DCA analysis of the signature. (E) Nomogram combining clinical features with the signature. DCA = decision curve analysis.*
## 3.6. Functional characterization of the DEGs
We next extracted DEGs between the 2 groups of GBM cases with package Limma and undertook functional analysis. As shown in Table S5, Supplemental Digital Content, http://links.lww.com/MD/I580, 97 DEGs were identified according to the criteria of |log2FC | ≥ 1 and FDR < 0.05; out of them, 7 were downregulated and the remaining 90 were highly expressed in the high risk patients. Furthermore, gene ontology and *Kyoto encyclopaedia* of genes and genomes analysis revealed a predominant enrichment of the DEGs in chemokine signaling pathway, the immune response, TNF signaling, NF − kappa B signaling, IL − 17 signaling, PI3K − Akt signaling, and AGE − RAGE signaling pathway (Fig. 7A and B).
**Figure 7.:** *Functional characterization of the DEGs. (A) The bubble diagram showing GO annotations of the DEGs. Note that the bigger bubbles represent more enriched DEGs in the categories, while the increasing intensity of redness indicates the greater differences. The q value is an adjusted P value. (B) Barplot of enriched pathways. Note that the longer bars represent more enriched DEGs in the pathways, while the increasing intensity of redness indicates the greater differences. DEGs = differentially expressed genes, GO = gene ontology.*
## 3.7. Inter group comparison of the immune activities
We further carried out single-sample gene set enrichment analysis to comparatively analyzed enrichment scores of 16 immune cell subpopulations as well as the activities of 13 immune-associated pathways between the 2 groups of GBM cases. As depicted in Figure 8A, a significant increase in the enrichment of immune cell subpopulations, including dendritic cells, macrophages, natural killer cells, neutrophils, T helper cells, Tfh cells, regulatory T cells, and tumor-infiltrating lymphocytes, was observed in the high risk cases in comparison with low risk ones. Among the 13 pathways, 12 except for MHC class-I pathway were found to be more active in high risk cases than in low risk ones (Fig. 8B). Moreover, a marked difference in the expression level of 35 immune checkpoints was observed between the 2 clusters of patients. Among the 35 checkpoints, CD200 and VTCN1 were up-regulated in low risk patients, and the remaining 33 were present at a higher level in high risk patients (Fig. 9).
**Figure 8.:** *ssGSEA-based comparison of immune activities. (A and B) Comparative analysis on the enrichment scores of 16 immune cell subpopulations as well as the activities of 13 immune-associated pathways between the 2 clusters of GBM cases. *P < .05, **P < .01, and ***P < .001. GBM = glioblastoma, ssGSEA = single-sample gene set enrichment analysis.* **Figure 9.:** *Comparative analysis on the expression level of 35 immune checkpoints between the 2 clusters of GBM cases. *P < .05, **P < .01, and ***P < .001. GBM = glioblastoma.*
## 4. Discussion
Here, we presented data showing that 32 out of 52 PRGs display differential expression between GBM versus normal tissue samples. Moreover, a PPI analysis revealed that CASP1, GSDMD, NLRP1, AIM2, PYCARD, CASP8, CASP5, TP53, and CASP3 were hub genes. Recent researches have shown that activated CASP1 could cleave C1q binding protein and boost aerobic glycolysis in tumor cells.[14] Upon activation, CASP3 can cleave GSDME to induce necrosis, providing novel insights into malignant tumor chemotherapy.[15] CASP8 has been defined as a molecular switch that regulates apoptosis, necroptosis, and pyroptosis; its activation participated in inflammatory responses of COVID-19 patients, probably causing lung injury.[16,17] NLRP1 is abundant in epithelial barrier tissues. And it could serve as a direct sensor for infection of dsRNA and RNA viruses, while being related to immune response.[18,19] AIM2 inflammasome contributes to surveillance of DNA damage and regulation of neurodevelopment.[20] PYCARD is implicated in both pyroptosis and apoptosis.[21] Pyroptosis plays a dual function in tumor progression and treatment. And it can lead to the release of inflammatory factors while transforming normal cells into malignant ones.[22] Pyroptosis may potentially be used for prognostic prediction and treatment of malignant tumors.[23] Herein, we divided a TCGA cohort of GBM cases into 2 groups based on the expression level of 52 PRGs and identified 32 DEGs between them. Further investigation revealed no marked differences in clinicopathologic characteristics between the 2 groups. Cox univariate analysis and LASSO analysis led to the development of a 9-gene signature (GZMB, AREG, LOXL1, MSTN, PTX3, IGFBP6, STC1, POM121L9P, TGM2) that was subsequently verified by external validation using a GEO dataset. GZMB was proven to function in pyroptosis of various cancers.[24] AREG can diminish tumor resistance while averting immunosuppression induced by programmed cell death 1 ligand.[25] LOX family is associated with glioma progression, and LOXL1 can confer antiapoptotic activity and promote gliomagenesis through stabilizing BAG2.[26] *As a* myogenesis inhibitor, MSTN regulates developmental maturation of skeletal myocytes and inhibits excessive cardiac autophagy to reduce cardiac hypertrophy.[27] PTX3 acts as a key player in humoral innate immunity that is implicated in resistance to certain pathogens as well as inflammation regulation, while it is associated with COVID-19, breast cancer, melanoma, and ischemia-reperfusion injury.[28–31] A recent study shows that IGFBP6, an IGF-II inhibitor, is critically involved in suppressing cancer cell survival and migration, while regulating apoptosis and cell migration in glioma.[32] STC1 is a biomarker of cellular senescence, and the senescence-associated secretory phenotype has been identified as a promising therapeutic target for, and driver of, age related disorders ranging from neurodegenerative conditions to malignant tumor.[33,34] POM121L9P is related to poor prognosis of patients with epithelial ovarian cancer.[35] Besides, TGM2 expression is correlated with development of colorectal cancer and endometrial cancer.[36,37] The current study found that the DEGs were enriched in a number of pathways, including immune-associated pathways, chemokine signaling pathway, IL − 17 pathway, TNF pathway, NF − kappa B pathway, PI3K − Akt signaling, AGE − RAGE pathway, and so forth. It has been reported that both chemokines and the corresponding receptors regulate cell migration, affecting numerous biological and cellular processes as well as the pathogenesis of diseases including inflammatory disorders and malignant tumors.[38] TNF signaling pathway is implicated in cancer metastasis, while participating in the occurrence of autoimmune and neuroinflammatory disorders.[39,40] As the founding member of a new inflammatory cytokine family, IL-17 plays a host-protective role mainly due to its pro-inflammatory properties, and unrestrained IL-17 signaling contributes to immunopathology, autoimmune disorders, and tumor development.[41] Increasing evidence shows that NF-κB signaling is critical for generating pro-inflammatory cytokine and chemokine cascades in response to acute respiratory virus infections.[42] Recent studies demonstrated that AGE-RAGE signaling not only participates in chronic obstructive lung disease and diabetic kidney disease, but also regulates apoptotic signaling to promote tumor progression.[43–45] PI3K − Akt signaling is implicated in regulating numerous cellular and physiological processes, such as cell division, differentiation, survival, and autophagy.[46] Herein, we observed greater enrichment of the immune cell subpopulations and increased activities of the immune associated pathways in high risk GBM cases as compared to the low risk cases. Moreover, marked differences in the expression levels of the immune checkpoints were detected between the 2 groups of patients. In this case, CD200 and VTCN1 were highly expressed in low risk cases, and the remaining 32 checkpoints were present at a higher level in high risk cases. It has been reported that while CD200 can promote immunosuppression in tumor microenvironment,[47] VTCN1 expression is related to reduced inflammatory CD4 + T-cell response within the microenvironment.[48] Our findings indicate that the expression level of CD200 and VTCN1 may be relevant to a better OS of the low risk patients. To date, specific regulatory mechanism of CD200 and VTCN1 in GBM remains unclear. Besides, we presented a nomogram combining clinical features with the 9-gene signature which may have a clinical relevance for GBM. However, there are some limitations to our study, such as the fact that it has not been applied in clinical practice for a long period of time, and further exploration is needed on the mechanisms associated with the 9-gene signature in the development of GBM.
The current research identified certain PRGs that were differently expressed between GBM tumor versus normal tissues, indicative of an association of GBM with pyroptosis. Furthermore, we showed that the risk score calculated using the 9-gene signature potentially serves as an independent predictor for the OS of GBM patients. Further studies revealed that the DEGs between high risk versus low risk GBM cases were related to tumor immunity. Overall, we developed a new multigene signature for prognostic prediction of GBM patients, thereby laying a solid foundation of GBM immunotherapy.
## Acknowledgements
We appreciate the accessible data from the TCGA and GEO (GSE83300) network.
## Author contributions
Conceptualization: Xiao Jin.
Data curation: Xiang Zhao.
Formal analysis: Xiang Zhao.
Visualization: Xiang Zhao.
Writing – original draft: Xiao Jin.
Writing – review & editing: Xiao Jin.
## References
1. Silantyev AS, Falzone L, Libra M. **Current and future trends on diagnosis and prognosis of glioblastoma: from molecular biology to proteomics.**. *Cells* (2019) **8** 863. PMID: 31405017
2. Tan AC, Ashley DM, Lopez GY. **Management of glioblastoma: state of the art and future directions.**. *CA Cancer J Clin* (2020) **70** 299-312. PMID: 32478924
3. Daubon T, Hemadou A, Romero Garmendia I. **Glioblastoma immune landscape and the potential of new immunotherapies.**. *Front Immunol* (2020) **11** 585616. PMID: 33154756
4. Lin J, Cheng A, Cheng K. **New insights into the mechanisms of pyroptosis and implications for diabetic kidney disease.**. *Int J Mol Sci* (2020) **21** 7057. PMID: 32992874
5. Xia X, Wang X, Cheng Z. **The role of pyroptosis in cancer: pro-cancer or pro “host?”.**. *Cell Death Dis* (2019) **10** 650. PMID: 31501419
6. Hou J, Zhao R, Xia W. **PD-L1-mediated gasdermin C expression switches apoptosis to pyroptosis in cancer cells and facilitates tumour necrosis.**. *Nat Cell Biol* (2020) **22** 13961264-1396
7. Jiang M, Qi L, Li L. **The caspase-3/GSDME signal pathway as a switch between apoptosis and pyroptosis in cancer.**. *Cell Death Discov* (2020) **6** 112. PMID: 33133646
8. Zhou C-B, Fang J-Y. **The role of pyroptosis in gastrointestinal cancer and immune responses to intestinal microbial infection.**. *Biochim Biophys Acta Rev Cancer* (2019) **1872** 1-10. PMID: 31059737
9. Ye Y, Dai Q, Qi H. **A novel defined pyroptosis-related gene signature for predicting the prognosis of ovarian cancer.**. *Cell Death Discov* (2021) **7** 71. PMID: 33828074
10. Lin W, Chen Y, Wu B. **Identification of the pyroptosis-related prognostic gene signature and the associated regulation axis in lung adenocarcinoma.**. *Cell Death Discov* (2021) **7** 161. PMID: 34226539
11. Wu D, Wang S, Yu G. **Cell death mediated by the pyroptosis pathway with the aid of nanotechnology: prospects for cancer therapy.**. *Angew Chem Int Ed Engl* (2021) **60** 8018-34. PMID: 32894628
12. Al Mamun A, Wu Y, Monalisa I. **Role of pyroptosis in spinal cord injury and its therapeutic implications.**. *J Adv Res* (2021) **28** 97-109. PMID: 33364048
13. An S, Hu H, Li Y. **Pyroptosis plays a role in osteoarthritis aging dis.**. *Aging Dis* (2020) **11** 1146-57. PMID: 33014529
14. Sünderhauf A, Raschdorf A, Hicken M. **GC1qR cleavage by caspase-1 drives aerobic glycolysis in tumor cells.**. *Front Oncol* (2020) **10** 575854. PMID: 33102234
15. Wang Y, Gao W, Shi X. **Chemotherapy drugs induce pyroptosis through caspase-3 cleavage of a gasdermin.**. *Nature* (2017) **547** 99-103. PMID: 28459430
16. Li S, Zhang Y, Guan Z. **SARS-CoV-2 triggers inflammatory responses and cell death through caspase-8 activation.**. *Signal Transduct Target Ther* (2020) **5** 235. PMID: 33037188
17. Fritsch M, Günther SD, Schwarzer R. **Caspase-8 is the molecular switch for apoptosis, necroptosis and pyroptosis.**. *Nature* (2019) **575** 683-7. PMID: 31748744
18. Bauernfried S, Scherr MJ, Pichlmair A. **Human NLRP1 is a sensor for double-stranded RNA.**. *Science* (2021) **371** eabd0811. PMID: 33243852
19. Taabazuing CY, Griswold AR, Bachovchin DA. **The NLRP1 and CARD8 inflammasomes.**. *Immunol Rev* (2020) **297** 13-25. PMID: 32558991
20. Lammert CR, Frost EL, Bellinger CE. **AIM2 inflammasome surveillance of DNA damage shapes neurodevelopment.**. *Nature* (2020) **580** 647-52. PMID: 32350463
21. Miao H, Wang L, Zhan H. **A long noncoding RNA distributed in both nucleus and cytoplasm operates in the PYCARD-regulated apoptosis by coordinating the epigenetic and translational regulation.**. *PLoS Genet* (2019) **15** e1008144. PMID: 31086376
22. Karki R, Kanneganti T-D. **Diverging inflammasome signals in tumorigenesis and potential targeting.**. *Nat Rev Cancer* (2019) **19** 197-214. PMID: 30842595
23. Wang YY, Liu XL, Zhao R. **Induction of pyroptosis and its implications in cancer management.**. *Front Oncol* (2019) **9** 971. PMID: 31616642
24. Tang R, Xu J, Zhang B. **Ferroptosis, necroptosis, and pyroptosis in anticancer immunity.**. *J Hematol Oncol* (2020) **13** 110. PMID: 32778143
25. Xu Q, Long Q, Zhu D. **Targeting amphiregulin (AREG) derived from senescent stromal cells diminishes cancer resistance and averts programmed cell death 1 ligand (PD-L1)-mediated immunosuppression.**. *Aging Cell* (2019) **18** e13027. PMID: 31493351
26. Yu H, Ding J, Zhu H. **LOXL1 confers antiapoptosis and promotes gliomagenesis through stabilizing BAG2.**. *Cell Death Differ* (2020) **27** 3021-36. PMID: 32424143
27. Qi H, Ren J, Ba L. **MSTN attenuates cardiac hypertrophy through inhibition of excessive cardiac autophagy by blocking AMPK/mTOR and miR-128/PPARγ/NF-κB.**. *Mol Ther Nucleic Acids* (2020) **19** 507-22. PMID: 31923740
28. Brunetta E, Folci M, Bottazzi B. **Macrophage expression and prognostic significance of the long pentraxin PTX3 in COVID-19.**. *Nat Immunol* (2021) **22** 19-24. PMID: 33208929
29. Zhang P, Liu Y, Lian C. **SH3RF3 promotes breast cancer stem-like properties via JNK activation and PTX3 upregulation.**. *Nat Commun* (2020) **11** 2487. PMID: 32427938
30. Rathore M, Girard C, Ohanna M. **Cancer cell-derived long pentraxin 3 (PTX3) promotes melanoma migration through a toll-like receptor 4 (TLR4)/NF-κB signaling pathway.**. *Oncogene* (2019) **38** 5873-89. PMID: 31253871
31. de Oliveira THC, Souza DG, Teixeira MM. **Tissue dependent role of PTX3 during ischemia-reperfusion injury.**. *Front Immunol* (2019) **10** 1461. PMID: 31354697
32. Bei Y, Huang Q, Shen J. **IGFBP6 regulates cell apoptosis and migration in glioma.**. *Cell Mol Neurobiol* (2017) **37** 889-98. PMID: 27650075
33. Basisty N, Kale A, Jeon OH. **A proteomic atlas of senescence-associated secretomes for aging biomarker development.**. *PLoS Biol* (2020) **18** e3000599. PMID: 31945054
34. Peng F, Xu J, Cui B. **Oncogenic AURKA-enhanced N-methyladenosine modification increases DROSHA mRNA stability to transactivate STC1 in breast cancer stem-like cells.**. *Cell Res* (2021) **31** 345-61. PMID: 32859993
35. Oliveira DVNP, Prahm KP, Christensen IJ. **Gene expression profile association with poor prognosis in epithelial ovarian cancer patients.**. *Sci Rep* (2021) **11** 5438. PMID: 33686173
36. Gu C, Cai J, Xu Z. **MiR-532-3p suppresses colorectal cancer progression by disrupting the ETS1/TGM2 axis-mediated Wnt/β-catenin signaling.**. *Cell Death Dis* (2019) **10** 739. PMID: 31570702
37. Lan T, Mu C, Wang Z. **Diagnostic and prognostic values of serum EpCAM, TGM2, and HE4 levels in endometrial cancer.**. *Front Oncol* (2020) **10** 1697. PMID: 33014844
38. Liu K, Wu L, Yuan S. **Structural basis of CXC chemokine receptor 2 activation and signalling.**. *Nature* (2020) **585** 135-40. PMID: 32610344
39. Zhou Q, Wu X, Wang X. **The reciprocal interaction between tumor cells and activated fibroblasts mediated by TNF-α/IL-33/ST2L signaling promotes gastric cancer metastasis.**. *Oncogene* (2020) **39** 1414-28. PMID: 31659258
40. Atretkhany K-SN, Gogoleva VS, Drutskaya MS. **Distinct modes of TNF signaling through its two receptors in health and disease.**. *J Leukoc Biol* (2020) **107** 893-905. PMID: 32083339
41. Amatya N, Garg AV, Gaffen SL. **IL-17 signaling: the Yin and the Yang.**. *Trends Immunol* (2017) **38** 310-22. PMID: 28254169
42. Kircheis R, Haasbach E, Lueftenegger D. **NF-κB pathway as a potential target for treatment of critical stage COVID-19 patients.**. *Front Immunol* (2020) **11** 598444. PMID: 33362782
43. Wu XQ, Zhang DD, Wang YN. **AGE/RAGE in diabetic kidney disease and ageing kidney.**. *Free Radic Biol Med* (2021) **171** 260-71. PMID: 34019934
44. Sharma A, Kaur S, Sarkar M. **The AGE-RAGE axis and RAGE genetics in chronic obstructive pulmonary disease.**. *Clin Rev Allergy Immunol* (2021) **60** 244-58. PMID: 33170477
45. Waghela BN, Vaidya FU, Ranjan K. **AGE-RAGE synergy influences programmed cell death signaling to promote cancer.**. *Mol Cell Biochem* (2021) **476** 585-98. PMID: 33025314
46. Jafari M, Ghadami E, Dadkhah T. **PI3k/AKT signaling pathway: erythropoiesis and beyond.**. *J Cell Physiol* (2019) **234** 2373-85. PMID: 30192008
47. Choueiry F, Torok M, Shakya R. **CD200 promotes immunosuppression in the pancreatic tumor microenvironment.**. *J ImmunoTher Cancer* (2020) **8** e000189. PMID: 32581043
48. Podojil JR, Miller SD. **Potential targeting of B7-H4 for the treatment of cancer.**. *Immunol Rev* (2017) **276** 40-51. PMID: 28258701
|
---
title: 'Astragalus membranaceus formula for moderate-high risk idiopathic membranous
nephropathy: A meta-analysis'
authors:
- Dan Wang
- Lijuan Wang
- Mingrui Zhang
- Ping Li
- Qinghua Zhang
- Kun Bao
journal: Medicine
year: 2023
pmcid: PMC9981402
doi: 10.1097/MD.0000000000032918
license: CC BY 4.0
---
# Astragalus membranaceus formula for moderate-high risk idiopathic membranous nephropathy: A meta-analysis
## Background:
Idiopathic membranous nephropathy (IMN) is a noninflammatory autoimmune glomerulonephropathy. Based on the risk stratification for disease progression, conservative nonimmunosuppressive and immunosuppressive therapy strategies have been recommended. However, there remains challenges. Therefore, novel approaches to treat IMN are needed. We evaluated the efficacy of *Astragalus membranaceus* (A membranaceus) combined with supportive care or immunosuppressive therapy in the treatment of moderate-high risk IMN.
### Methods:
We comprehensively searched PubMed, Embase, the Cochrane Library, the China National Knowledge Infrastructure, the Database for Chinese Technical Periodicals, Wanfang Knowledge Service Platform, and SinoMed. We then performed a systematic review and cumulative meta-analysis of all randomized controlled trials assessing the two therapy methods.
### Results:
The meta-analysis included 50 studies involving 3423 participants. The effect of A membranaceus combined with supportive care or immunosuppressive therapy is better than that of supportive care or immunosuppressive therapy along in regulating for improving 24 hours urinary total protein (MD = −1.05, $95\%$ CI [−1.21, −0.89], $$P \leq .000$$), serum albumin (MD = 3.75, $95\%$ CI [3.01, 4.49], $$P \leq .000$$), serum creatinine (MD = −6.24, $95\%$ CI [−9.85, −2.63], $$P \leq .0007$$), complete remission rate (RR = 1.63, $95\%$ CI [1.46, 1.81], $$P \leq .000$$), partial remission rate (RR = 1.13, $95\%$ CI [1.05, 1.20], $$P \leq .0004$$).
### Conclusions:
Adjunctive use of A membranaceus preparations combined with supportive care or immunosuppressive therapy have a promising treatment for improving complete response rate, partial response rate, serum albumin, and reducing proteinuria, serum creatinine levels compared to immunosuppressive therapy in people with MN being at moderate-high risk for disease progression. Given the inherent limitations of the included studies, future well-designed randomized controlled trials are required to confirm and update the findings of this analysis.
## 1. Introduction
In China, membranous nephropathy (MN) emerged as the leading type of biopsy finding in patients aged > 40 years, which would soon surpass IgA nephropathy.[1] *Idiopathic membranous* nephropathy (IMN) is a noninflammatory autoimmune glomerulonephropathy, characterized by the formation of immune complex deposits on the subepithelial in the kidney.[2] The most patients with IMN present nephrotic range proteinuria.[3–5] Approximately $40\%$ of patients will undergo spontaneous remission,[6] while another $30\%$ will have a poor response to immunosuppressive therapy and then progress to end-stage renal disease (ESRD).[7] If follow-up is extended to 10 to 20 years, progression to ESRD may occur in $50\%$ to $60\%$ of patients without treatment.[8] Based on the risk stratification for disease progression, conservative nonimmunosuppressive and immunosuppressive therapy strategies have been recommended by the Kidney Disease: Improving Global Outcomes (KDIGO) Clinical Practice Guideline to treat patients with IMN.[9] Unfortunately, there remains challenges of compromised clinical response, high cost, serious adverse effects and high recurrences.[10–16] Therefore, novel approaches to treat IMN are needed.
Over thousands years, traditional Chinese medicine (TCM) has been extensively used in East Asia and developed a unique theoretical system, containing many different therapeutic and preventive methods (such as Chinese herbal medicine, acupuncture and moxibustion).[17] Astragalus (Radix Astragali) is one of the most widely prescribed herbs in traditional Chinese medicine. Astragalus (huang qi in Chinese), is the dry root of *Astragalus membranaceus* (A membranaceus; Fisch.) Bge. var. Mongholicus (Bge.) Hsiao or A membranaceus (Fisch.) Bge. Hitherto, over 100 chemical constitutions have been isolated and identified from A.membranaceus, including flavonoids, saponins, polysaccharides, and amino acids.[18–20] To date, a number of studies in animal and cellular models have proven that Astragalus possesses potent protective effects in kidney.[21–23] Many clinical studies have demonstrated that Astragalus can improve kidney function and reduce proteinuria.[24–26] However, the quality of these studies had not been assessed systematically. It is imperative to assess the efficacy and safety of A membranaceus as adjunctive therapy to Western medicine therapy for IMN. The aim of the current study is to evaluate the effificacy of A membranaceus combined with supportive care or immunosuppressive therapy in the treatment of moderate-high risk IMN.
## 2. Methods
This systematic review followed the methods of the Cochrane Handbook for Systematic Reviews of Interventions (version 6.3) and complied with the 2009 Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement guidelines.[27,28] We registered the review protocol with PROSPERO at the beginning (CRD: 42021232472).
## 2.1.1. Types of studies.
Randomized controlled trials (RCTs) on the treatment of adults with IMN using oral Astragalus preparation were eligible. Quasi-RCTS (RCTs in which allocation to treatment was obtained by alternate medical records, alternation, use of date of birth, or other divinable methods) were excluded. There was no restriction on languages or publication status.
## 2.1.2.1. Inclusion criteria.
We included adults (aged 18 years and older) histologically diagnosed with IMN. Patients were classified as being at moderate-high risk for disease progression when recruited. According to the KDIGO classification: moderate risk is defined as normal eGFR, proteinuria of more than 3.5 g/d and no decrease of more than $50\%$ after 6 months of conservative therapy with angiotensin converting enzyme inhibitor (ACEi) or angiotensin II receptor blocker (ARB), and not fulfilling high-risk criteria. While high risk is defined as eGFR of less than 60 mL/min/1.73 m2 and/or proteinuria of more than 8 g/d for more than 6 months; or normal eGFR, proteinuria of more than 3.5 g/d and no decrease of more than $50\%$ after 6 months of conservative therapy with ACEi or ARB, and at least one of the following: serum albumin of less than 25 g/L; anti-phospholipase A2 receptor antibody of more than 50 RU/mL. Therapeutic regimen complied with KDIGO guideline. Oral forms of A membranaceus preparation, including boiled decoction, extracts and granules were eligible.
## 2.1.2.2. Exclusion criteria.
Studies with the following conditions will be excluded: Trials included patients with secondary forms of membranous nephropathy; *Study data* could not be available from the report or by contacting the authors; Patients who were received renal replacement therapy; Studies assessed A membranaceus combined with other complementary therapies (such as moxibustion, acupuncture); Therapies related to TCM were used in the control group; Western medicine was contained in both groups, but they were differented from each other; Huangqi doesn‘t serves as a principal medicine.
## 2.1.3. Types of interventions.
All participants received routine therapies according to clinical practice guidelines. Conventional therapy includes more rest, low salt, low fat, high quality and protein diet; drugs aimed to correct dyslipidaemia (e.g., statins), antialdosterone drugs (e.g., spironolactone), antihypertensive (e.g., ACEi or ARB), antithrombotic agents (e.g., dipyridamole).
Treatment group participants needed to have received oral *Astragalus formula* (decoction, pill, powder, or capsule) in combination with immunosuppressive therapy.
Control group participants received immunosuppressive therapy.
Huangqi serves as a principal medicine, defined as follows: the properties of Huangqi are consistent with the main aims of the formula.
## 2.1.4.1. Primary outcomes.
Complete response rate was assessed according to the definition provided in each single study.
Partial response rate was assessed according to the definition provided in each single study.
Proteinuria measured by 24 hours urinary total protein (UTP), Urine protein/creatinine ratio.
Kidney function measured by Serum creatinine concentration (SCr).
Adverse events.
## 2.1.4.2. Secondary outcomes.
Disease activity assessed by serum albumin.
Primary outcome or secondary outcome measurements were collected immediately after treatment and at the end of follow-up period.
## 2.2. Search strategy
We searched seven databases from their inception to May 2022. Three English databases include EMBASE, MEDLINE, Cochrane Central Register of Controlled Trials. The following Chinese databases were also searched: China National Knowledge Infrastructure, China Biomedical Literature Database, Chongqing VIP, and Wanfang. Reference lists of significant reviews on similar topics and relevant studies were examined.
The search strategies applied for this review were shown (Appendix 1, Supplemental Digital Content, http://links.lww.com/MD/I444).
## 2.3. Data collection and analysis
Two authors independently screened the titles, abstract, and full-text and discarded studies which were not satisfied the eligibility criteria. Then, they in parallel extracted data from eligible studies, using a pre-designed data extraction form. Any published versions discrepancies was highlighted. If more than one publication about one study existed, we grouped them together and used the most complete data. When necessary, original authors were contacted by email to clarify details or to acquire further information about their trials. Information about study characteristics (the first author’s name, participants gender, age, histological type, proteinuria severity, sample size, follow-up period.), intervention protocol of A membranaceus (dosage forms, dosage, frequency, and duration), concurrent supportive care or immunosuppressive therapy, and outcome data was collected. Two reviewers assessed the methodological quality of included studies independently, using the Cochrane risk of bias tool (Cochrane Handbook V.5.1.0)[29] (See Table 1). Individual studies were graded as high, unclear, or low risk of bias. Heterogeneity was assessed using the Cochrane Q statistic and I2 test. A P value less than.05 was used for statistical significance.[30] I2 values more than $50\%$ were correspond to high levels of heterogeneity. Publication bias were examined by using funnel plots for asymmetry and Egger’s linear regression analysis, when one outcome included 10 or more studies. A third author examined the consistency check. If the consensus was not reached, methodologists were consulted to resolve it.
**Table 1**
| First author, year [Ref] | Intervention (ingredients of Huangqi formula) | Control participant (Intervention/control) | Participant (Intervention/control) | Age mean ± SD (yr) | Treatment duration (mo) | Complete response | Partial response | Baseline eGFR mean ± SD (mL/min) | Baseline proteinuria mean ± SD (g/24 h) |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Caiz 2016 | Shenqi Dihuang Decoction (tangshen 15 g, milkvetch root 20 g, prepared rehmannia root 15 g, Chinese yam 15 g, indian bread 20 g, oriental waterplantain rhizome 10 g, tree peony root bark 10 g, asiatic cornelian cherry fruit 15 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 22/23 | 46.36 vs 44.42 | 6 | 24 h urine protein of less than 0.5 g, serum albumin of more than 35 g/L, nephritic syndrome disappeared completely. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 4.63 ± 0.814.47 ± 0.7 |
| Caiz 2019 | Shenqi Dihuang Decoction (tangshen 15 g, milkvetch root 20 g, prepared rehmannia root 15 g, Chinese yam 15 g, indian bread 20 g, oriental waterplantain rhizome 10 g, tree peony root bark 10 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 32/26 | 42.26 ± 13. 26 | 6 | 24 h urine protein of less than 0.5 g, serum albumin of more than 35 g/L, nephritic syndrome disappeared completely. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 4.99 ± 1.514.90 ± 1.03 |
| Caiz 2019 | Shenqi Dihuang Decoction (tangshen 15 g, milkvetch root 20 g, prepared rehmannia root 15 g, Chinese yam 15 g, indian bread 20 g, oriental waterplantain rhizome 10 g, tree peony root bark 10 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 32/26 | 44.32 ± 13. 32 | 6 | 24 h urine protein of less than 0.5 g, serum albumin of more than 35 g/L, nephritic syndrome disappeared completely. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 4.99 ± 1.514.90 ± 1.03 |
| Duohl 2020 | Self-formulated Jianpi Lishi Tongluo Prescription (milkvetch root 20 g, indian bread 15 g, oriental waterplantain rhizome 10 g, largehead atractylodes rhizome 12 g, glabrous greenbrier rhizome 10 g, plantain seed 15 g, Chinese waxgourd peel 15 g, cicada slough 10 g, earthworm 10 g, black-tail snake 10 g, tortoise carapace and plastron 9 g, danshen root 10 g, safflower 10 g, figwort root 12 g, dwarf lilyturf tuber 12 g) | Angiotensin converting enzyme inhibitor | 36/36 | 46.2 ± 10.45 | 3 | Proteinuria remained negative or normal 24 h urine protein, normal eGFR. | Sustained 25–50% reduction in urinary protein, normal eGFR. | – | 5.64 ± 3.115.6 ± 2.4 |
| Duohl 2020 | Self-formulated Jianpi Lishi Tongluo Prescription (milkvetch root 20 g, indian bread 15 g, oriental waterplantain rhizome 10 g, largehead atractylodes rhizome 12 g, glabrous greenbrier rhizome 10 g, plantain seed 15 g, Chinese waxgourd peel 15 g, cicada slough 10 g, earthworm 10 g, black-tail snake 10 g, tortoise carapace and plastron 9 g, danshen root 10 g, safflower 10 g, figwort root 12 g, dwarf lilyturf tuber 12 g) | Angiotensin converting enzyme inhibitor | 36/36 | 47.97 ± 8.87 | 3 | Proteinuria remained negative or normal 24 h urine protein, normal eGFR. | Sustained 25–50% reduction in urinary protein, normal eGFR. | – | 5.64 ± 3.115.6 ± 2.4 |
| Lix 2014 | Therapy of Invigorating Spleen and Kidney, Activating Blood and Dispelling Wind (milkvetch root 30 g, cherokee rose fruit 30 g, gordon euryale seed 30 g, coix seed 30 g, largehead atractylodes rhizome 15 g, eucommia 15 g, danshen root 15 g, dodder seed 15 g, stiff silkworm 10 g, tangshen 10 g, Chinese angelica 10 g, peach seed 10 g, divaricate saposhnikovia root 6 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 32/31 | 28.1 ± 7.26 | 6 | Proteinuria remained negative or 24 h urine protein of less than 0.2 g, normal eGFR. | Sustained 25–50% reduction in urinary protein, normal eGFR. | – | 4.23 ± 1.224.08 ± 1.35 |
| Lix 2014 | Therapy of Invigorating Spleen and Kidney, Activating Blood and Dispelling Wind (milkvetch root 30 g, cherokee rose fruit 30 g, gordon euryale seed 30 g, coix seed 30 g, largehead atractylodes rhizome 15 g, eucommia 15 g, danshen root 15 g, dodder seed 15 g, stiff silkworm 10 g, tangshen 10 g, Chinese angelica 10 g, peach seed 10 g, divaricate saposhnikovia root 6 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 32/31 | 26.23 ± 6.71 | 6 | Proteinuria remained negative or 24 h urine protein of less than 0.2 g, normal eGFR. | Sustained 25–50% reduction in urinary protein, normal eGFR. | – | 4.23 ± 1.224.08 ± 1.35 |
| Mazw 2011 | Jia-wei-bu-yang-huan-wu Powder (milkvetch root 60 g, Chinese angelica 12 g, Sichuan lovage rhizome 9 g, earthworm 9 g, peach seed 10 g, safflower 10 g, peony root 12 g, Chinese yam 15 g, asiatic cornelian cherry fruit 12 g, cowherb seed 15 g, plantain seed 15 g, barbary wolfberry fruit 12 g, tangshen 15 g.) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 30/30 | 46.07 ± 11.06 | 1 | Proteinuria remained negative or 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 4.96 ± 2.534.93 ± 2.25 |
| Mazw 2011 | Jia-wei-bu-yang-huan-wu Powder (milkvetch root 60 g, Chinese angelica 12 g, Sichuan lovage rhizome 9 g, earthworm 9 g, peach seed 10 g, safflower 10 g, peony root 12 g, Chinese yam 15 g, asiatic cornelian cherry fruit 12 g, cowherb seed 15 g, plantain seed 15 g, barbary wolfberry fruit 12 g, tangshen 15 g.) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 30/30 | 48.37 ± 9.70 | 1 | Proteinuria remained negative or 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 4.96 ± 2.534.93 ± 2.25 |
| Pangzx 2019 | Jianpi Bushen Decoction (milkvetch root 30 g, plantain seed 12 g, Chinese yam 12 g, tangshen 12 g, cowherb seed 12 g, Chinese angelica 9 g, peach seed 9 g, asiatic cornelian cherry fruit 9 g, safflower 9 g, barbary wolfberry fruit 9 g, peony root 9 g, earthworm 6 g, Sichuan lovage rhizome 6 g) | Angiotensin converting enzyme inhibitor | 52/51 | 44.10 ± 4.29 | 1 | nephritic syndrome disappeared completely. | Symptoms and signs improved. | – | 4.99 ± 2.364.98 ± 2.33 |
| Pangzx 2019 | Jianpi Bushen Decoction (milkvetch root 30 g, plantain seed 12 g, Chinese yam 12 g, tangshen 12 g, cowherb seed 12 g, Chinese angelica 9 g, peach seed 9 g, asiatic cornelian cherry fruit 9 g, safflower 9 g, barbary wolfberry fruit 9 g, peony root 9 g, earthworm 6 g, Sichuan lovage rhizome 6 g) | Angiotensin converting enzyme inhibitor | 52/51 | 43.91 ± 4.33 | 1 | nephritic syndrome disappeared completely. | Symptoms and signs improved. | – | 4.99 ± 2.364.98 ± 2.33 |
| Pinggh 2021 | Qiqi Yishen capsule (milkvetch root, asiatic cornelian cherry fruit, Indian bread, coix seed, barbary wolfberry fruit, chinese angelica, Radix Salviae Miltiorrhizae, peony root, twotoothed achyranthes root, rhubarb root and rhizome, yerbadetajo herb) | Angiotensin II receptor blocker | 40/40 | 41.3 ± 10.2 | 4 | Normal 24 h urine protein. | 40% reduction in urinary protein. | 81.47 ± 15.04283.24 ± 15.651 | 5.49 ± 1.135.93 ± 0.88 |
| Pinggh 2021 | Qiqi Yishen capsule (milkvetch root, asiatic cornelian cherry fruit, Indian bread, coix seed, barbary wolfberry fruit, chinese angelica, Radix Salviae Miltiorrhizae, peony root, twotoothed achyranthes root, rhubarb root and rhizome, yerbadetajo herb) | Angiotensin II receptor blocker | 40/40 | 40.3 ± 11.1 | 4 | Normal 24 h urine protein. | 40% reduction in urinary protein. | 81.47 ± 15.04283.24 ± 15.651 | 5.49 ± 1.135.93 ± 0.88 |
| Panz 2020 | Shen Zhi HuoXue Decoction (milkvetch root 50 g, leech 15 g, tangshen 20 g, peach seed 15 g, safflower 15 g, peony root 15 g, dwarf lilyturf tuber 10 g, Chinese angelica 15 g, unprocessed rehmannia root 15 g, lotus seed 10 g, Chinese wolfberry root-bark 10 g, liquorice root 10 g) | Angiotensin II receptor blocker | 15/15 | 48.13 ± 12.12 | 2 | Proteinuria remained negative or normal 24 h urine protein, serum albumin of more than 35 g/L, normal eGFR. | 40% reduction in urinary protein. | – | 4.11 ± 1.4554.3 ± 1.355 |
| Panz 2020 | Shen Zhi HuoXue Decoction (milkvetch root 50 g, leech 15 g, tangshen 20 g, peach seed 15 g, safflower 15 g, peony root 15 g, dwarf lilyturf tuber 10 g, Chinese angelica 15 g, unprocessed rehmannia root 15 g, lotus seed 10 g, Chinese wolfberry root-bark 10 g, liquorice root 10 g) | Angiotensin II receptor blocker | 15/15 | 41.26 ± 10.79 | 2 | Proteinuria remained negative or normal 24 h urine protein, serum albumin of more than 35 g/L, normal eGFR. | 40% reduction in urinary protein. | – | 4.11 ± 1.4554.3 ± 1.355 |
| Qiaoln 2020 | Yiqi Huashi Tongluo Decoction (milkvetch root 30 g, tangshen 15 g, leech 6 g, danshen root 30 g, indian bread 20 g, two toothed achyranthes root 20 g, unprocessed rehmannia root 24 g, zhuling 20 g, oriental waterplantain rhizome 15 g, earthworm 12 g, coix seed 15 g, hedyotis 15 g) | Angiotensin II receptor blocker | 30/29 | 47.27 ± 9.938 | 3 | Proteinuria remained negative or normal 24 h urine protein, normal eGFR. | 40% reduction in urinary protein. | – | 5.22 ± 1.824.91 ± 1.63 |
| Qiaoln 2020 | Yiqi Huashi Tongluo Decoction (milkvetch root 30 g, tangshen 15 g, leech 6 g, danshen root 30 g, indian bread 20 g, two toothed achyranthes root 20 g, unprocessed rehmannia root 24 g, zhuling 20 g, oriental waterplantain rhizome 15 g, earthworm 12 g, coix seed 15 g, hedyotis 15 g) | Angiotensin II receptor blocker | 30/29 | 43.07 ± 9.083 | 3 | Proteinuria remained negative or normal 24 h urine protein, normal eGFR. | 40% reduction in urinary protein. | – | 5.22 ± 1.824.91 ± 1.63 |
| Wangl 2020 | Shenqi Dihuang Decoction (milkvetch root 20 g, tangshen 15 g, prepared rehmannia root 15 g, Chinese yam 15 g, asiatic cornelian cherry fruit 15 g, indian bread 20 g, tree peony root bark 10 g, oriental waterplantain rhizome 10 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 28/26 | 47.6 ± 13.2 | 6 | 24 h urine protein of less than 0.5 g, serum albumin of more than 35 g/L. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 5.87 ± 1.216.26 ± 1.12 |
| Wangl 2020 | Shenqi Dihuang Decoction (milkvetch root 20 g, tangshen 15 g, prepared rehmannia root 15 g, Chinese yam 15 g, asiatic cornelian cherry fruit 15 g, indian bread 20 g, tree peony root bark 10 g, oriental waterplantain rhizome 10 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 28/26 | 46.6 ± 12. 9 | 6 | 24 h urine protein of less than 0.5 g, serum albumin of more than 35 g/L. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 5.87 ± 1.216.26 ± 1.12 |
| Yangfw 2021 | Yishen Tongluo Recipe (milkvetch root 20 g, largehead atractylodes rhizome 15 g, tangshen 15 g, epimedium herb 15 g, chinese angelica 15 g, zedoray rhizome 12 g, earthworm 12 g, leech 3 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 23/22 | 46.82 ± 9.43 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 50% reduction in urinary protein. | – | 5.47 ± 1.875.32 ± 2.31 |
| Yangfw 2021 | Yishen Tongluo Recipe (milkvetch root 20 g, largehead atractylodes rhizome 15 g, tangshen 15 g, epimedium herb 15 g, chinese angelica 15 g, zedoray rhizome 12 g, earthworm 12 g, leech 3 g) | Angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 23/22 | 45.63 ± 9.14 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 50% reduction in urinary protein. | – | 5.47 ± 1.875.32 ± 2.31 |
| Aiy 2017 | Dispersing three jiao-activating blood-dredging collateral Recipe (milkvetch root, cassia twig, epimedium herb, cablin patchouli herb, cardamon fruit, largehead atractylodes rhizome, indian bread, Sichuan lovage rhizome, danshen root, safflower, earthworm, leech) | Prednisone and cyclophosphamide | 30/29 | 45 vs 42 | 6 | Proteinuria remained negative or 24 h urine protein of less than 0.2 g, normal eGFR. | Sustained 25–50% reduction in urinary protein, normal eGFR. | – | 5.86 ± 1.55.26 ± 0.92 |
| Daim 2018 | Qingrehuoxue Hushen Decoction (milkvetch root 30 g, largehead atractylodes rhizome 15 g, atractylodes rhizome 10 g, Chinese yam 20 g, zhuling 12 g, indian bread 12 g, Chinese angelica 15 g, Sichuan lovage rhizome 10 g, barbated skullcup herb 15 g, stiff silkworm 15 g, hedyotis 30 g, coix seed 30 g, earthworm 10 g) | Prednisone and tacrolimus | 30/30 | -- | 6 | Proteinuria remained negative or 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 5.46 ± 2.025.12 ± 1.32 |
| Guowg 2015 | Self-formulated Prescription (milkvetch root 60 g, dodder seed 20 g, barbary wolfberry fruit 20 g, ginger processed pinellia 10 g, stiff silkworm 15 g, common buried rubber 15 g, zedoray rhizome 15 g, centipede 2 g) | Methylprednisolone and prednisone and cyclophosphamide | 37/37 | 38.2 vs 37.6 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 7.35 ± 2.036.86 ± 1.89 |
| Guoyp 2021 | Qidi Taozhi Erchan Recipe (milkvetch root 50 g, unprocessed rehmannia root 30 g, peach seed 12 g, earthworm 12 g, leech 9 g, cicada slough 10 g, amur cork-tree 10 g, heterophylly falsestarwort root 15 g, atractylodes rhizome 15 g, indian bread 15 g, epimedium herb 15 g, coix seed 20 g) | Tacrolimus | 42/42 | 46.96 ± 8.63 | 3 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 4.38 ± 2.794.53 ± 2.46 |
| Guoyp 2021 | Qidi Taozhi Erchan Recipe (milkvetch root 50 g, unprocessed rehmannia root 30 g, peach seed 12 g, earthworm 12 g, leech 9 g, cicada slough 10 g, amur cork-tree 10 g, heterophylly falsestarwort root 15 g, atractylodes rhizome 15 g, indian bread 15 g, epimedium herb 15 g, coix seed 20 g) | Tacrolimus | 42/42 | 48.21 ± 7.57 | 3 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 4.38 ± 2.794.53 ± 2.46 |
| Haoj 2017 | Huangzhiyishen capsule (milkvetch root, leech, barbary wolfberry fruit, Chinese yam, coix seed, sanqi) | Glucocorticoid and cyclophosphamide | 32/32 | 47.72 ± 6.84 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 5.72 ± 1.565.85 ± 1.84 |
| Hexc 2016 | Self-formulated Qingxue Xiaobai Prescription (Chinese yam 15 g, indian bread 15 g, common anemarrhena rhizome 20 g, kudzuvine root 15 g, plantain seed 15 g, milkvetch root 30 g, Chinese magnoliavine fruit 20 g, sanqi 10 g, snakegourd root 15 g, finger citron 20 g, coix seed 20 g) | Prednisone and cyclosporin A | 50/50 | 42.22 ± 3.65 | 6 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 5.32 ± 0.475.43 ± 0.54 |
| Hug 2020 | Yishen Huashi Granules (ginseng, milkvetch root, largehead atractylodes rhizome, indian bread, oriental waterplantain rhizome, pinellia tuber, incised notopterygium rhizome and root, doubleteeth pubescent angelica root, divaricate saposhnikovia root, Chinese thorowax root, golden thread, debark peony root, dried tangerine peel, liquorice root, fresh ginger, Chinese date) | Prednisone and tacrolimus/cyclosporin A | 21/20 | – | 9 | 24 h urine protein of less than 0.5 g, serum albumin of more than 35 g/L, normal eGFR. | Sustained 50% reduction in urinary protein, normal eGFR. | – | – |
| Jiaozs 2018 | ShenQi ZhiLong Decoction (tangshen 20 g, milkvetch root 20 g, leech 6 g, earthworm 10 g, turmeric yellow 10 g, giant knotweed rhizome 15 g, glabrous greenbrier rhizome 15 g, hedyotis 30 g) | Tacrolimus | 30/30 | 45.43 ± 7.25 | 3 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 4.3 ± 0.394.19 ± 0.24 |
| Jiaozs 2018 | ShenQi ZhiLong Decoction (tangshen 20 g, milkvetch root 20 g, leech 6 g, earthworm 10 g, turmeric yellow 10 g, giant knotweed rhizome 15 g, glabrous greenbrier rhizome 15 g, hedyotis 30 g) | Tacrolimus | 30/30 | 46.86 ± 6.79 | 3 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 4.3 ± 0.394.19 ± 0.24 |
| Ladh 2018 | The Decoction of Benefiting Kidney Qi and Promoting Blood Circulation (milkvetch root 30 g, tangshen 15 g, Chinese yam 30 g, indian bread 30 g, asiatic cornelian cherry fruit 15 g, prepared rehmannia root 15 g, dodder seed 15 g, largehead atractylodes rhizome 10 g, danshen root 15 g, Chinese angelica 15 g, gordon euryale seed 30 g, cherokee rose fruit 15 g, dandelion 30 g, Radix Arnebiae 10 g) | Cyclosporin A | 46/52 | 48.15 ± 2.75 | 2 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 5.93 ± 1.555.8 ± 1.69 |
| Ladh 2018 | The Decoction of Benefiting Kidney Qi and Promoting Blood Circulation (milkvetch root 30 g, tangshen 15 g, Chinese yam 30 g, indian bread 30 g, asiatic cornelian cherry fruit 15 g, prepared rehmannia root 15 g, dodder seed 15 g, largehead atractylodes rhizome 10 g, danshen root 15 g, Chinese angelica 15 g, gordon euryale seed 30 g, cherokee rose fruit 15 g, dandelion 30 g, Radix Arnebiae 10 g) | Cyclosporin A | 46/52 | 47.75 ± 2.82 | 2 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 5.93 ± 1.555.8 ± 1.69 |
| Leigp 2016 | Qidi gushen Decoction (milkvetch root 30–90 g, unprocessed rehmannia root 15–30 g, gordon euryale seed 30–45 g, hedyotis 30–60 g, danshen root 15–20 g, fineleaf schizonepeta herb 10 g) | Prednisone and cyclophosphamide | 48/22 | 38.1 ± 19.7 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 6.89 ± 3.316.74 ± 2.79 |
| Leigp 2016 | Qidi gushen Decoction (milkvetch root 30–90 g, unprocessed rehmannia root 15–30 g, gordon euryale seed 30–45 g, hedyotis 30–60 g, danshen root 15–20 g, fineleaf schizonepeta herb 10 g) | Prednisone and cyclophosphamide | 48/22 | 40.6 ± 22.3 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 6.89 ± 3.316.74 ± 2.79 |
| Leigp 2019 | Qidi gushen Tablet (milkvetch root, rehmannia root, gordon euryale seed, hedyotis, danshen root, fineleaf schizonepeta herb) | Angiotensin converting enzyme inhibitor/(prednisone and cyclophosphamide) | 15/15 | 51.69 ± 11.31 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 3.87 ± 2.024.22 ± 1.75 |
| Leigp 2019 | Qidi gushen Tablet (milkvetch root, rehmannia root, gordon euryale seed, hedyotis, danshen root, fineleaf schizonepeta herb) | Angiotensin converting enzyme inhibitor/(prednisone and cyclophosphamide) | 15/15 | 48.97 ± 11.61 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 3.87 ± 2.024.22 ± 1.75 |
| Leisb 2020 | Huangqi Chifeng Decoction (milkvetch root 30 g, cherokee rose fruit 20 g, gordon euryale seed 20 g, peony root 10 g, divaricate saposhnikovia root 10 g, earthworm 10 g, hedyotis 10 g) | Tacrolimus | 33/32 | 68.35 ± 4.2268.21 ± 4.01 | 6 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 2 g, serum albumin improved. | – | 6.53 ± 0.316.48 ± 0.35 |
| Liangj 2017 | Reinforcing spleen and kidney and clearing heat and activating blood-based treatment (milkvetch root, Chinese yam, largehead atractylodes rhizome, gordon euryale seed, Sichuan lovage rhizome, indian bread, unprocessed rehmannia root, cherokee rose fruit, tortoise carapace and plastron, Chinese angelica, earthworm, baical skullcap root) | Prednisone and cyclophosphamide | 25/23 | 48.76 ± 11.7847.08 ± 14.77 | 6 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 6.05 ± 2.057.15 ± 2.51 |
| Lidy 2018 | Yishen Jianpi Tongluo Decoction (milkvetch root 30 g, prepared rehmannia root 15 g, asiatic cornelian cherry fruit 15 g, indian bread 15 g, largehead atractylodes rhizome 15 g, Chinese angelica 15 g, earthworm 15 g, danshen root 15 g, Sichuan lovage rhizome 12 g, cassia twig 10 g, liquorice root 3 g) | Prednisone and cyclophosphamide | 42/42 | 45.2 ± 6.546.4 ± 6.3 | 3 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 4.32 ± 0.514.26 ± 0.53 |
| Lij 2020 | Wenyang Qushi Tongluo Recipe (milkvetch root 30 g, Chinese angelica 15 g, Sichuan lovage rhizome 12 g, safflower 10 g, leech 3 g, cablin patchouli herb 10 g, dried tangerine peel 15 g, cardamon fruit 10 g, largehead atractylodes rhizome 15 g, indian bread 15 g, epimedium herb 15 g) | Prednisone and cyclophosphamide | 60/60 | 44.8 ± 10.5645.23 ± 10.07 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 7.97 ± 0.817.93 ± 0.83 |
| Liuhx 2019 | Tonifying Kidney and Removing Blood Stasis and Clearing Away Heat (milkvetch root, tangshen, unprocessed rehmannia root, largehead atractylodes rhizome, Chinese yam, indian bread, Chinese angelica, Sichuan lovage rhizome, danshen root, hedyotis, rhubarb root and rhizome) | Cyclosporin A/tacrolimus and angiotensin converting enzyme inhibitor/angiotensin II receptor blocker | 29/29 | 47.41 ± 13.646.34 ± 14.38 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | 127.68 ± 35.53138.65 ± 41.88 | 7.65 ± 2.478.04 ± 2.58 |
| Liuxy 2018 | Huatan Quyu Decoction (bile arisaema 10 g, mustard 10 g, tangle 10 g, danshen root 15 g, twotoothed achyranthes root 15 g, motherwort herb 10 g, peony root 10 g, Chinese angelica 15 g, Sichuan lovage rhizome 15 g, largehead atractylodes rhizome 15 g, stiff silkworm 10 g, milkvetch root 30 g) | Prednisone and cyclophosphamide | 40/40 | 44.87 ± 10.1645.21 ± 9.42 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 5.44 ± 2.335.1 ± 2.1 |
| Liyg 2020 | Shenqizhilong Decoction (hedyotis 30 g, glabrous greenbrier rhizome 15 g, giant knotweed rhizome 15 g, turmeric 10 g, earthworm 10 g, leech 6 g, milkvetch root 20 g, tangshen 20 g) | Tacrolimus | 42/42 | 49.75 ± 6.2949.82 ± 6.34 | 3 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 4.25 ± 0.514.18 ± 0.53 |
| Loucl 2019 (cyclophosphamide) | Jianpi Yiqi Qingre Huoxue decoction (milkvetch root, tangshen, largehead atractylodes rhizome, radix salviae miltiorrhizae, chinese angelica, motherwort herb, hedyotis, baical skullcap root, plantain herb, coix seed, zhuling) | Prednisone and cyclophosphamide | 20/20 | 54.3 ± 13.3 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 7.17 ± 2.236.20 ± 2.82 |
| Loucl 2019 (cyclosporin A) | Jianpi Yiqi Qingre Huoxue decoction (milkvetch root, tangshen, largehead atractylodes rhizome, radix salviae miltiorrhizae, chinese angelica, motherwort herb, hedyotis, baical skullcap root, plantain herb, coix seed, zhuling) | Prednisone and cyclosporin A | 20/20 | 49.5 ± 13.6 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 6.37 ± 2.436.52 ± 2.41 |
| Maxg 2017 | Qidi gushen Decoction (milkvetch root 50–90 g, gordon euryale seed 30–45 g, hedyotis 30–60 g, danshen root 15–20 g, fineleaf schizonepeta herb 10 g) | Prednisone/methylprednisolone and cyclophosphamide/tacrolimus | 15/14 | 52.555 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 5.47 ± 1.25.76 ± 1.08 |
| Shenxm 2017 | Jianpi Bushen Prescription (milkvetch root 30 g, epimedium herb 15 g, morinda root 15 g, dodder seed 15 g, indian bread 15 g, largehead atractylodes rhizome 12 g, Chinese yam 15 g, liquorice root 6 g) | Prednisone and cyclophosphamide | 30/30 | 41.00 ± 6.1042.63 ± 6.31 | 3 | 24 h urine protein of less than 0.4 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 4.5 ± 0.5534.61 ± 0.605 |
| Wangql 2016 | Shenqi Dihuang Decoction加减 (tangshen 15 g, milkvetch root 45 g, prepared rehmannia root 15 g, Chinese yam 12 g, asiatic cornelian cherry fruit 15 g, indian bread 15 g, oriental waterplantain rhizome 12 g, tree peony root bark 15 g, golden thread 6 g, danshen root 12 g, Sichuan lovage rhizome 15 g, Chinese magnoliavine fruit 15 g, liquorice root 6 g) | Prednisone and tacrolimus | 30/30 | 47.67 ± 14.81248.9 ± 16.164 | 6 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 5.26 ± 2.15.01 ± 1.42 |
| Wangt 2017 | Yishen Tongluo Decoction (milkvetch root 30–60 g, prepared rehmannia root 15 g, asiatic cornelian cherry fruit 15 g, cassia twig 10 g, largehead atractylodes rhizome 15 g, indian bread 15 g, Chinese angelica 15 g, Sichuan lovage rhizome 12 g, danshen root 15 g, earthworm 15 g, scorpion 5 g, liquorice root 3 g) | Prednisone and cyclophosphamide | 39/36 | 45.38 ± 6.7443.59 ± 8.62 | 3 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 1 g, serum albumin improved. | – | 4.96 ± 0.314.91 ± 0.24 |
| Weiyj 2019 | Qilong Tongshen Recipe (milkvetch root 30 g, earthworm 10 g, Chinese angelica 12 g, Sichuan lovage rhizome 12 g, peony root 12 g, peach seed 12 g, safflower 12 g, hawthorn 30 g, leech 3 g) | Prednisone and cyclophosphamide | 57/57 | 52.81 ± 9.2252.84 ± 10.47 | 3 | 24 h urine protein of less than 0.3 g, serum albumin of more than 40 g/L, normal eGFR. | Sustained 25–50% reduction in urinary protein. | – | 4.76 ± 1.254.87 ± 1.48 |
| Wuj 2021 | Qidi Gushen recipe (milkvetch root 60 g, raw land 30 g, gordon euryale seed 30 g, hedyotis 30 g, radix salviae miltiorrhizae 20 g, fineleaf schizonepeta herb 10 g) | Prednisone and cyclophosphamide | 49/49 | 52.38 ± 3.7653.69 ± 3.63 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 3.8 ± 1.133.78 ± 1.06 |
| Wuqf 2017 | Buqi Qufeng Method (milkvetch root 30 g, largehead atractylodes rhizome 10 g, Chinese yam 10 g, asiatic cornelian cherry fruit 10 g, scorpion 4 g, stiff silkworm 10 g, orientvine vine 15 g, cicada slough 10 g, hedyotis 30 g, plantain herb 30 g, peach seed 10 g, safflower 10 g, ground beetle 10 g, Sichuan lovage rhizome 15 g) | Prednisone and cyclophosphamide and tacrolimus | 15/15 | 45.47 ± 15.1347.53 ± 14.46 | 6 | Proteinuria remained negative. | Sustained 40% reduction in urinary protein, normal eGFR. | – | 6.82 ± 2.065.91 ± 2.1 |
| Xiex 2018 | Tripterygium wilfordii polyglycosides, Yiqi Huoxue Lishui Therapy (milkvetch root 30 g, peony root 20 g, Chinese angelica 20 g, Sichuan lovage rhizome 20 g, peach seed 20 g, largehead atractylodes rhizome 20 g, indian bread 20 g, liquorice root 10 g) | Prednisone and cyclophosphamide | 90/90 | 46.1 ± 1.245.4 ± 1.5 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | Sustained 25% reduction in urinary protein, normal eGFR. | – | 6.33 ± 1.566.35 ± 1.47 |
| Yangy 2021 | Huangtu Yishen Granules (milkvetch root 30–60 g, dodder 15–25 g, cherokee rose fruit 15–30 g, gordon euryale seed 30–50 g, largehead atractylodes rhizome 10–15 g, peony root 10–15 g, chinese angelica 10–15 g, glabrous greenbrier rhizome 20–30 g, oriental waterplantain rhizome 10–15 g) | Prednisone and tacrolimus | 21/21 | 51.10 ± 9.9452.71 ± 10.45 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | 80.99 ± 4.9281.17 ± 7.06 | 7.45 ± 2.257.68 ± 2.14 |
| Yangyc 2016 | Yishen Xiaobai Recipe (milkvetch root 45 g, unprocessed rehmannia root 24 g, largehead atractylodes rhizome 9 g, scorpion 9 g, danshen root 30 g, indian bread 20 g, cicada slough 20 g, stiff silkworm 20 g, earthworm 20 g, leech 5 g, Chinese angelica 15 g, safflower 15 g, peach seed 15 g, Sichuan lovage rhizome 15 g) | Prednisone and cyclophosphamide | 33/33 | 44 vs 42 | 12 | 24 h urine protein of less than 0.3 g, serum albumin of more than 30 g/L, normal eGFR. | Sustained 50% reduction in urinary protein, normal eGFR. | – | 5.61 ± 3.665.49 ± 3.51 |
| Yangzh 2019 | Nephrotic Prescription No.1 (milkvetch root 30 g, prepared rehmannia root 20 g, largehead atractylodes rhizome 10 g, divaricate saposhnikovia root 6 g, asiatic cornelian cherry fruit 10 g, dodder seed 15 g, leech 3 g, coix seed 20 g) | Prednisone and cyclophosphamide | 39/39 | 56.92 ± 7.0456.75 ± 7.12 | 12 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 7.93 ± 0.727.85 ± 0.76 |
| Yux 2018 | Yiqi Huoxue Huayu method (milkvetch root 30 g, Chinese yam 30 g, indian bread 20 g, platycodon root 20 g, danshen root 20 g, tree peony root bark 10 g, rehmannia root 20 g, yerbadetajo herb 20 g, gordon euryale seed30 g, cherokee rose fruit 20 g, oyster shell 20 g, Chinese thorowax root 6 g) | Prednisone/methylprednisolone and cyclophosphamide/tacrolimus | 15/15 | 42.33 ± 17.5947.13 ± 14.11 | 6 | Proteinuria remained negative. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 6.06 ± 1.826.71 ± 2.08 |
| Yuy 2015 | ShenQi Decoction (milkvetch root 30 g, tangshen 20 g, Chinese yam 20 g, largehead atractylodes rhizome 15 g, prepared rehmannia root 15 g, indian bread 20 g, cassia twig 15 g, danshen root 20 g, leech 5 g, hedyotis 20 g, Chinese magnoliavine fruit 15 g, liquorice root 15 g) | Tacrolimus | 28/26 | 45.5 ± 9.4246.47 ± 9.12 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | 96.31 ± 24.6393.8 ± 25.1 | 6.31 ± 2.556.36 ± 2.67 |
| Zhangff 2020 | Xuantong Sanjiao and Huoxue Tongluo Formula (milkvetch root 45 g, cassia twig 10 g, epimedium herb 10 g, cablin patchouli herb 10 g, dried tangerine peel 12 g, cardamon fruit 10 g, largehead atractylodes rhizome 12 g, indian bread 15 g, Sichuan lovage rhizome 12 g, danshen root 10 g, safflower 12 g, earthworm 12 g, leech 6 g) | Prednisone and cyclophosphamide | 30/29 | 45 vs 42 | 6 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 5.86 ± 1.55.26 ± 0.92 |
| Zhangyt 2021 | Shenqi Dihuang Decoction (heterophylly falsestarwort root, milkvetch root, raw land, chinese yam, asiatic cornelian cherry fruit, indian bread, tree peony root bark, golden thread, radix salviae miltiorrhizae, chinese angelica, sichuan lovage rhizome, hedyotis, liquorice root) | Prednisone and tacrolimus | 39/39 | 50.97 ± 10.4252.53 ± 11.69 | 6 | 24 h urine protein of less than 0.2 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 7.85 ± 6.097.99 ± 6.25 |
| Zhaoc 2019 | Yishen Jianpi Huoluo Decoction (milkvetch root 30 g, Chinese yam 30 g, prepared rehmannia root 20 g, largehead atractylodes rhizome 20 g, indian bread 20 g, barbated skullcup herb 15 g, Sichuan lovage rhizome 20 g, danshen root 20 g, leech 10 g) | Prednisone and tacrolimus | 40/40 | 16-6616-67 | 6 | 24 h urine protein of less than 0.4 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 8.91 ± 1.678.78 ± 1.98 |
| Zhaoc 2020 | Jianpi Yishen Huoxue Prescription (milkvetch root 30 g, prepared rehmannia root 20 g, Chinese yam 30 g, largehead atractylodes rhizome 30 g, indian bread 30 g, coix seed 30 g, asiatic cornelian cherry fruit 15 g, oriental waterplantain rhizome 10 g, mealy fangji 10 g, Chinese angelica 20 g, peony root 10 g, danshen root 20 g, leech 10 g) | Prednisone and tacrolimus | 40/40 | 45.8 ± 9.845.9 ± 9.8 | 6 | 24 h urine protein of less than 0.4 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3 g, serum albumin improved. | – | 9.84 ± 2.899.92 ± 2.66 |
| Zhaor 2021 | Huangtu Yishen Granules (milkvetch root 30–60 g, dodder 15–25 g, gordon euryale seed 30–50 g, cherokee rose fruit 15–30 g, peony root 10–15 g, chinese angelica 10–15 g, glabrous greenbrier rhizome 20–30 g, largehead atractylodes rhizome 10–15 g, oriental waterplantain rhizome 10–15 g) | Prednisone and tacrolimus | 21/20 | 50.37 ± 11.9950.07 ± 14.17 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | 106.77 ± 16.75101.97 ± 16.96 | 6.9269 ± 2.410895.78924 ± 1.71299 |
| Zhuyq 2021 | Qizhi dihuang decoction (milkvetch root 30 g, leech 6 g, cultivated land 15 g, tangshen 15 g, asiatic cornelian cherry fruit 15 g, chinese yam 15 g, indian bread 15 g, oriental waterplantain rhizome 12 g, tree peony root bark 9 g, debark peony root 12 g, sichuan lovage rhizome 12 g, radix salviae miltiorrhizae 12 g) | Prednisone and cyclophosphamide | 31/30 | 47.27 ± 12.8149.87 ± 13.32 | 2 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | 24 h urine protein of less than 3.5 g, serum albumin improved. | – | 7.2 ± 4.46.67 ± 2.79 |
| Zuojj 2018 | Qiling Tongluo Formula (milkvetch root 20–50 g, largehead atractylodes rhizome 15 g, indian bread 15–30 g, tortoise carapace and plastron 15 g, cicada slough 10 g, black-tail snake 10 g, stiff silkworm 10 g, earthworm 12 g, ground beetle 10 g, leech 3 g, scorpion 3 g, orientvine vine 15 g, glossy ganoderma 15 g) | Methylprednisolone and cyclophosphamide | 40/40 | 48 | 6 | 24 h urine protein of less than 0.3 g, serum albumin of more than 35 g/L, normal eGFR. | Sustained 50% reduction in urinary protein, normal eGFR. | – | 6.06 ± 1.635.96 ± 1.76 |
## 2.4. Data synthesis and statistical analysis
Data were pooled in Cochrane systematic review software Review Manager (Version 5.4.1) and Stata (Version 12.0) by using a random-effects model when I2 was greater than $50\%$. Otherwise, a fixed-effects model was analyzed.[31] *Dichotomous data* were expressed as odds ratio with $95\%$ confidence intervals (CI). For continuous scales of measurement, results were accessed as mean difference (MD) or standardized mean difference (SMD) with $95\%$ CI. We carried out outcomes analysis based on randomized participants. Where multiple intervention groups in one study existed, pair-wise comparisons related to our study were made.
## 3.1. Study selection
Eleven thousand one hundred and ninety-two citations were retrieved from the databases. We excluded 4074 duplicates and 3227 irrelevant records. Finally, after screening, 50 eligible studies were identified. The major reason for exclusion were classified as no TCM formulas or the principal medicine is not Astragalus, non-randomized controlled study design, or not IMN. ( Fig. 1).
**Figure 1.:** *The study flow diagram. CBM = China Biomedical Literature Database, CNKI = China National Knowledge Infrastructure, CQVIP = Chongqing VIP, IMN = Idiopathic membranous nephropathy, RCTs = randomized controlled trials, TCM = Traditional Chinese Medicine.*
## 3.2. Study characteristics
The characteristics of study were presented in Table 1. All included studies were carried out in China. A total of 3423 participants with IMN were enrolled in the 50 eligible studies.[32–81] Forty eight studies[32–47,49–60,62–81] reported participants’ gender (male 1729 vs female 1279, respectively). Participants’ age ranged from 16 to 68, with a mean age of 48 years old. Ehrenreich-Churg Stage I to IV lesions of MN varied among participants. Fifty studies recruited the two groups of patients. Treatment duration ranged from 1 to 18 months, most of which was 6 months ($\frac{31}{50}$, $62\%$). Fifty studies were parallel arm studies, including three study arms. A membranaceus preparation was administered with the form of decoction in all studies (50 RCTs). The dose of A membranaceus varied from 20 to 90 g/d, and 30 g was the most common dose. A membranaceus preparations also included tablet form granule and capsule. Most of the studies reported at least one laboratory outcome of complete remission rate, partial remission rate, proteinuria, or kidney function. None of them measured the endpoint event of mortality or number of patients progressing to ESKD. In all studies, we regarded Huangqi as the principal medicine for its properties consistent with the main aims of the prescription.
## 3.3. Quality assessment
The methodological quality of literature was present in Figures 2 and 3. All of the included studies mentioned “randomized,” just 20 trials[37,38,40,43,46,51,54,56,58,59,62,64–66,70,76,78–81] provided the details of randomization methods. No eligible studies used blinding methods to decrease performance bias, so all of them were judged as “high risk.” None of them reported allocation concealment. All of them were evaluated at low risk of bias for blinding of outcome assessment in consideration of laboratory test outcomes. All the studies were evaluated as a “low risk” about reporting bias for all the outcome data were reported. Over $90\%$ of studies were judged at low risk of attrition bias because of no withdrawals or losses to follow-up during the study period. One trial[35] reported dropouts, and all provided detailed and reasonable explanations.
**Figure 2.:** *Risk of bias summary of included studies. “?” = unclear risk of bias; “–” = low risk of bias; “+” = high risk of bias.* **Figure 3.:** *Evaluation for bias risk of included studies.*
## 3.4. Effects on complete response rate
Complete renal remission was measured in 48 studies[32,33,35–40,42–81] ($$n = 3324$$). For moderate-high risk patients, the complete remission of using an oral A membranaceus preparation were significantly higher than those without the oral A membranaceus preparation (48 trials, $$n = 3324$$, RR 1.63 [$95\%$ CI 1.46, 1.81]). Further, we performed the subgroup analyses. The pooled estimation indicated that the rates of overall remission in the oral A membranaceus preparation combined with nonimmunosuppressive therapy group were significantly higher than those in the nonimmunosuppressive therapy group (11 trials,[71–81] $$n = 669$$, RR 1.71 [$95\%$ CI 1.30, 2.25]). And the rates of complete remission in the oral A membranaceus formula combined with immunosuppressive therapy group were significantly higher than those in the immunosuppressive therapy group alone (37 trials,[32,33,35–40,42–70] $$n = 2655$$, RR 1.61 [$95\%$ CI 1.43, 1.81]) (Fig. 4).
**Figure 4.:** *Forest plot of complete response rate. ACEi = angiotensin converting enzyme inhibitor, ARB = angiotensin receptor blocker, CI = confidence intervals.*
## 3.5. Effects on partial response rate
Forty eight eligible trials[32,33,35–40,42–81] reported partial response rate data ($$n = 3324$$). For moderate-high risk patients, the partial remission of using an oral A membranaceus preparation were slightly higher than those without the oral A membranaceus preparation (48 trials, $$n = 3324$$, RR 1.13 [$95\%$ CI 1.05, 1.20]). We performed the subgroup analyses. Pooled results indicated that compared with nonimmunosuppressive therapy group alone, the partial remission rate of using *Astragalus formula* combined were no significantly high (11 trials,[71–81] $$n = 669$$, RR 1.08 [$95\%$ CI 0.96, 1.23]). The partial remission rate of using *Astragalus formula* combined with immunosuppressive therapy group were significantly higher than those with immunosuppressive therapy alone (37 trials,[32,33,35–40,42–70] $$n = 2655$$, RR 1.14 [$95\%$ CI 1.05, 1.23]) (Fig. 5).
**Figure 5.:** *Forest plot of partial response rate. ACEi = angiotensin converting enzyme inhibitor, ARB = angiotensin receptor blocker, CI = confidence intervals.*
## 3.6. Effects on proteinuria
For moderate-high risk patients, Astragalus preparation significantly decreased 24 hours proteinuria at end of treatment (46 trials,[32,33,35–40,42–53,55–60,62–81] $$n = 3253$$, MD −1.05 g/24 h, $95\%$ CI −1.21, −0.89; I2 = $93\%$). We performed the subgroup analyses. Pooled results indicated that compared with nonimmunosuppressive therapy group alone, the using *Astragalus formula* combined suggested lower end-of-treatment proteinuria (11 trials,[71–81] $$n = 669$$, MD −0.99 g/24 h, $95\%$ CI −1.29, −0.70; I2 = $59\%$). The 24 hours proteinuria level of using *Astragalus formula* combined with immunosuppressive therapy group were significantly lower than those without the oral A membranaceus preparation group (35 trials,[32,33,35–40,42–53,55–60,62–70] $$n = 2584$$, MD −1.07 g/24 h, $95\%$ CI −1.25, −0.88; I2 = $95\%$), albeit with substantial heterogeneity (Fig. 6).
**Figure 6.:** *Forest plot of proteinuria. ACEi = angiotensin converting enzyme inhibitor, ARB = angiotensin receptor blocker, CI = confidence intervals, SD = standard deviation.*
## 3.7. Effects on kidney function
For moderate-high risk patients, combining with Astragalus preparation slightly decreased SCr compared with control (31 studies,[32,33,36–39,42–49,51,52,55–58,64–66,68,69,73–76,78,80] $$n = 2184$$, MD −6.24 µmol/L, $95\%$ CI −9.85, −2.63; I2 = $94\%$). We performed the subgroup analyses. Pooled results indicated that compared with nonimmunosuppressive therapy group alone, the using *Astragalus formula* combined suggested lower SCr (6 trials,[73–76,78,80] $$n = 437$$, MD −1.64 µmol/L, $95\%$ CI −6.60, 3.31; I2 = $87\%$). The SCr level of using *Astragalus formula* combined with immunosuppressive therapy group were significantly lower than those with immunosuppressive therapy alone (25 trials,[32,33,36–39,42–49,51,52,55–58,64–66,68,69] $$n = 1747$$, MD −7.53 µmol/L, $95\%$ CI −12.37, −2.70; I2 = $95\%$), albeit with substantial heterogeneity (Fig. 7).
**Figure 7.:** *Forest plot of serum creatinine. ACEi = angiotensin converting enzyme inhibitor, ARB = angiotensin receptor blocker, CI = confidence intervals, SD = standard deviation.*
## 3.8. Effects on serum albumin
Forty four trials[32,33,35–40,43–49,51–53,55–60,62–81] measured serum albumin. Immunosuppressive therapy combining with Astragalus preparation significantly improved serum albumin compared with control group (44 studies, $$n = 3043$$, MD 3.75, $95\%$ CI 3.01, 4.49; I2 = $91\%$). We performed the subgroup analyses. Pooled results indicated that compared with nonimmunosuppressive therapy group alone, the using *Astragalus formula* combined suggested higher serum albumin (11 studies,[71–81] $$n = 669$$, MD 3.02, $95\%$ CI 1.84, 4.21; I2 = $89\%$). The serum albumin level of using *Astragalus formula* combined with immunosuppressive therapy group were significantly higher than those with immunosuppressive therapy alone (33 studies,[32,33,35–40,43–49,51–53,55–60,62–70] $$n = 2374$$, MD 4.45, $95\%$ CI 3.08, 4.86; I2 = $90\%$), albeit with substantial heterogeneity (Fig. 8).
**Figure 8.:** *Forest plot of serum albumin. ACEi = angiotensin converting enzyme inhibitor, ARB = angiotensin receptor blocker, CI = confidence intervals, SD = standard deviation.*
## 3.9. Safety outcomes
Safety outcomes were reported in 29 trials[32,33,35,37–41,43–46,48–50,52–59,61–64,69,70] out of 50 trials included studies. Ten studies[35,37,38,44,45,52–54,57,59] reported no adverse events occurred during the treatment period. The most common adverse events observed in 19 studies[32,33,39–41,43,46,48–50,55,56,58,61–64,69,70] were gastrointestinal discomfort (74 cases), infect (50 cases), elevated blood sugar (40 cases), abnormal liver function (36 cases), hypertension (26 cases), aleucocytosis (21 cases), Cushing syndrome (17 cases), Insomnia (16 cases). A case of myelosuppression were reported in one studies.[43] All adverse events abated spontaneously and there was no difference for the frequency of adverse events of incidence among the groups.
## 3.10. Publication bias
In point of publication bias, there was suspect in outcomes of complete response rate (Egger’s test $$P \leq .006$$), proteinuria (Egger’s test $$P \leq .002$$), and SCr (Egger’s test $$P \leq .015$$), but not in serum albumin (Egger’s test $$P \leq .116$$) and partial response rate (Egger’s test $$P \leq .986$$), suggesting a lack of studies with negative results.
## 4. Discussion
This systematic review and meta-analysis was to provide an overview of use of A membranaceus preparations in the treatment of moderate-high risk IMN. As antibodies against phospholipase A2 receptor were detected in $70\%$ to $80\%$ of patients, IMN is confirmed as an autoimmune disease now.[82] And then immunosuppressive therapies have been applied to treat IMN patients widely. Among them, rituximab or cyclophosphamide combined with corticosteroids is recommended as a first-line regimen for IMN by the KDIGO guideline.[83] Many studies showed cyclophosphamide combined with corticosteroids has favorable effects in preventing progressing to ESRD.[84] However, the limitation of cyclophosphamide is serious adverse effects associated with accumulated dose.[85] Moreover, cyclosporine and tacrolimus have been effective in promoting remission in $70\%$ of MN patients,[86] but the high rate of relapse of these drugs should not be ignored. More recently, the administration of rituximab has encouraging results. Unfortunately, these biologic agents are too much to afford, especially in less developed Countries. Therefore, we urgently need to find an effective, safe and economic therapeutic method to treat moderate-high risk IMN.
A membranaceus is a traditional Chinese medicine with a wide range of active components, mainly including Astragaloside IV, A membranaceus polysaccharide and A membranaceus isoflavone. It is widely used in clinical practice. There are many extraction preparations of A. membranaceus: A membranaceus injection, A membranaceus oral liquid, A membranaceus capsule, A membranaceus polysaccharide for injection and compound A membranaceus nasal spray.
More and more evidence show that A membranaceus has obvious advantages in the treatment of IMN.[25,87] Some studies on the pharmacological efficacy of A membranaceus in treating nephropathy have suggested that it plays an important role in improving renal perfusion, managing blood pressure and delaying renal function progression.[88] But the efficacy of A membranaceus-containing formula for moderate-high risk IMN remains to be further reviewed and analyzed.
This meta-analysis included 50 RCTs and involved 3423 patients to evaluating the relationship between A membranaceus formula in combination with immunosuppressive therapy and the use of immunosuppressive therapy alone in the treatment of IMN.
Based on the analysis of available data, we found that the efficacy of A membranaceus formula combined with immunosuppressive therapy is better than immunosuppressive therapy used alone in the treatment of IMN in improving 24 h UTP, serum albumin, SCr.
## 4.1. Limitations
Evidence on the potential benefits or harms of oral A membranaceus preparations for IMN is limited. None of the included studies provided data on calculation of sample size, blindness, allocation concealment and a placebo for the TCM in the control. Most of studies were small and conducted in single hospitals, showing methodological weaknesses especially in terms of randomization. This may have impact on results. The efficacy of oral A membranaceus preparations might be overestimated for deficiency of negative studies. A membranaceus has been associated with some side effects. However, only 30 studies reported a few adverse effects. Fourth, use of A membranaceus preparations may be confined outside East Asia. This poses challenges on the widespread applicability of A membranaceus for patients with IMN globally.
## 5. Conclusion
The present review demonstrated that adjunctive use of A membranaceus preparations combined with immunosuppressive therapy have a promising treatment for improving complete response rate, partial response rate, serum albumin and reducing proteinuria, serum creatinine levels compared to immunosuppressive therapy in people with MN being at moderate-high risk for disease progression. The overall quality of evidence was low, so the conclusion should be interpreted with caution. There is a need to confirm by high-quality evidence later.
## Author contributions
Conceptualization: Lijuan Wang.
Data curation: Mingrui Zhang.
Investigation: Ping Li.
Resources: Dan Wang.
Supervision: Qinghua Zhang.
Visualization: Dan Wang.
Writing – original draft: Dan Wang, Lijuan Wang.
Writing – review & editing: Kun Bao.
## References
1. Xu X, Wang G, Chen N. **Long-term exposure to air pollution and increased risk of membranous nephropathy in China.**. *J Am Soc Nephrol* (2016) **27** 3739-46. PMID: 27365535
2. Ronco P, Debiec H. **Molecular pathogenesis of membranous nephropathy.**. *Annu Rev Pathol* (2020) **15** 287-313. PMID: 31622560
3. Troyanov S, Wall CA, Miller JA. **Idiopathic membranous nephropathy: definition and relevance of a partial remission.**. *Kidney Int* (2004) **66** 1199-205. PMID: 15327418
4. Schieppati A, Mosconi L, Perna A. **Prognosis of untreated patients with idiopathic membranous nephropathy.**. *N Engl J Med* (1993) **329** 85-9. PMID: 8510707
5. MacTier R, Boulton Jones JM, Payton CD. **The natural history of membranous nephropathy in the West of Scotland.**. *Q J Med* (1986) **60** 793-802. PMID: 3774962
6. Polanco N, Gutiérrez E, Rivera F. **Spontaneous remission of nephrotic syndrome in membranous nephropathy with chronic renal impairment.**. *Nephrol Dial Transplant* (2012) **27** 231-4. PMID: 21624942
7. Glassock RJ. **Diagnosis and natural course of membranous nephropathy.**. *Semin Nephrol* (2003) **23** 324-32. PMID: 12923720
8. Ponticelli C, Passerini P. **Management of idiopathic membranous nephropathy.**. *Expert Opin Pharmacother* (2010) **11** 2163-75. PMID: 20707756
9. Alsharhan L, Beck LH. **Membranous nephropathy: core curriculum 2021.**. *Am J Kidney Dis* (2021) **77** 440-53. PMID: 33487481
10. Chen Y, Schieppati A, Chen X. **Immunosuppressive treatment for idiopathic membranous nephropathy in adults with nephrotic syndrome.**. *Cochrane Database Syst Rev* (2014) **2014** Cd004293. PMID: 25318831
11. Xie G, Xu J, Ye C. **Immunosuppressive treatment for nephrotic idiopathic membranous nephropathy: a meta-analysis based on Chinese adults.**. *PLoS One* (2012) **7** e44330. PMID: 22957065
12. Chen Y, Schieppati A, Cai G. **Immunosuppression for membranous nephropathy: a systematic review and meta-analysis of 36 clinical trials.**. *Clin J Am Soc Nephrol* (2013) **8** 787-96. PMID: 23449768
13. Cattran D, Brenchley P. **Membranous nephropathy: thinking through the therapeutic options.**. *Nephrol Dial Transplant* (2017) **32** i22-9. PMID: 28391348
14. Hofstra JM, Fervenza FC, Wetzels JF. **Treatment of idiopathic membranous nephropathy.**. *Nat Rev Nephrol* (2013) **9** 443-58. PMID: 23820815
15. Ahlmann M, Hempel G. **The effect of cyclophosphamide on the immune system: implications for clinical cancer therapy.**. *Cancer Chemother Pharmacol* (2016) **78** 661-71. PMID: 27646791
16. Shah SR, Altaf A, Arshad MH. **Use of cyclosporine therapy in Steroid Resistant Nephrotic Syndrome (SRNS): a review.**. *Glob J Health Sci* (2015) **8** 136-41. PMID: 26573045
17. Wusiman A, Abula S, Shayibuzhati M. **Traditional Uyghur medicine: concepts, historical perspective, and modernization.**. *Altern Ther Health Med* (2017) **23** 34-41
18. Auyeung KK, Han QB, Ko JK. *Am J Chin Med* (2016) **44** 1-22. PMID: 26916911
19. Fu J, Wang Z, Huang L. **Review of the botanical characteristics, phytochemistry, and pharmacology of**. *Phytother Res* (2014) **28** 1275-83. PMID: 25087616
20. Liu P, Zhao H, Luo Y. **Anti-aging implications of**. *Aging Dis* (2017) **8** 868-86. PMID: 29344421
21. Lei X, Zhang L, Li Z. **Astragaloside IV/lncRNA-TUG1/TRAF5 signaling pathway participates in podocyte apoptosis of diabetic nephropathy rats.**. *Drug Des Devel Ther* (2018) **12** 2785-93
22. Lei X, Zhang BD, Ren JG. **Astragaloside suppresses apoptosis of the podocytes in rats with diabetic nephropathy via miR-378/TRAF5 signaling pathway.**. *Life Sci* (2018) **206** 77-83. PMID: 29792879
23. Guo H, Wang Y, Zhang X. **Astragaloside IV protects against podocyte injury via SERCA2-dependent ER stress reduction and AMPKα-regulated autophagy induction in streptozotocin-induced diabetic nephropathy.**. *Sci Rep* (2017) **7** 6852. PMID: 28761152
24. Leehey DJ, Casini T, Massey D. **Remission of membranous nephropathy after therapy with**. *Am J Kidney Dis* (2010) **55** 772. PMID: 20338467
25. Ahmed MS, Hou SH, Battaglia MC. **Treatment of idiopathic membranous nephropathy with the herb**. *Am J Kidney Dis* (2007) **50** 1028-32. PMID: 18037104
26. Lang R, Wang X, Liang Y. **Research progress in the treatment of idiopathic membranous nephropathy using Traditional Chinese Medicine.**. *J Transl Int Med* (2020) **8** 3-8. PMID: 32435606
27. Moher D, Liberati A, Tetzlaff J. **Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement.**. *BMJ* (2009) **339** b2535-b2535. PMID: 19622551
28. Moher D, Shamseer L, Clarke M. **Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement.**. *Syst Rev* (2015) **4** 1. PMID: 25554246
29. Higgins J, Green S. *Cochrane Handbook for Systematic Reviews of Interventions, Version 5.1.0* (2013)
30. Higgins JP, Thompson SG, Deeks JJ. **Measuring inconsistency in meta-analyses.**. *BMJ* (2003) **327** 557-60. PMID: 12958120
31. Higgins J, Green S. *Cochrane Handbook for Systematic Reviews for Interventions, Version 5.1.0* (2011)
32. Guo WG, Wang YL, Yi BQ. **Treatment of idiopathic membranous nephropathy by combination of Chinese and western medicine.**. *J Clin Rat Drug Use* (2015) **8** 61-2
33. Yu Y. *Observation the Effect of ShenQi Decoction Combined with Tacrolimus in the Treatment of Spleen and Kindey Deficiency Type I-II Membranous Nephropathy [thesis]* (2015)
34. Lei GP, Li XH, Gao BF. **The 65 cases of clinical observation of Qidigushen formula on the treatment of idiopathic membranous nephropathy.**. *Chin J Int Trad Western Nephro* (2016) **17** 913-4
35. Wang QL. *Clinical Observation of the Codonopsis, Astragalus and Rehmannia Decoction Combined with Tacrolimus and Low-dose Corticosteroids in Treatment of Nephrotic Idiopathic Membranous Nephropathy [thesis]* (2016)
36. He XC, Zhao FB, Li JW. **Observation on the curative effect of Self-formulated Qingxue Xiaobai prescription on membranous nephrotic syndrome.**. *Mod J Int Trad Chin Western Med* (2016) **25** 3453-4
37. Yang YC, Wang CL, Li J. **Observation on the curative effect of Yishen Xiaobai Recipe on idiopathic membranous nephropathy.**. *Shaanxi J TCM* (2016) **37** 1140-1
38. Ai Y. *Clinical Curative Effects and Safety Analysis on Dispersing Three Jiao-Activating Blood-dredging Collateral Recipe in Treating Idiopathic Membranous Nephropathy [thesis]* (2017)
39. Hao J, Yu WM. **Clinical effect of combination with Huangzhiyishen capsule in the treatment of Stage I -II Idiopathic Membranous Nephropathy.**. *Chin J Int Trad Western Nephro* (2017) **18** 620-1
40. Liang J. *Reinforcing Spleen and Kidney and Clearing Heat and Activating Blood-based Treatment of Idiopathic Membranous Nephropathy Clinical Observation [thesis]* (2017)
41. Ma XG. *The Therapy Effect of Qi Di Gu Shen Fang for IMN Patients and Index Changing on Anti-PLA2R Antibo [thesis]* (2017)
42. Wu QF. *Clinical Observation and Theoretical Exploration of Buqi Qufeng Method in the Treatment of Membranous Nephropathy [thesis]* (2017)
43. Wang T. *The Clinical Research of Idiopathic Membranous Nephropathy with Spleen and Kidney Deficiency Syndrome and Blood Stagnation in Kidney Meridian Treated by Yishen Tongluo Decoction [thesis]* (2017)
44. Li DY, Zheng DY, Tang J. **Effect of Yishen Jianpi Tongluo Decoction on urinary protein and coagulation in patients with idiopathic membranous nephropathy.**. *Mod J Int Trad Chin Western Med* (2018) **27** 4020-7
45. Jiao ZS. *Observation on Curative Effect of ShenQi ZhiLong Decoction in Treating Membranous Nephropathy of Qi Deficiency and Blood Stasis Type [thesis]* (2018)
46. Zuo JJ, Zhao Z, Sun R. **Clinical observation of combination of Qiling Tongluo formula, Methylprednisolone and Cyclophosphamide treating high-risk patients of idiopathic membranous nephropathy.**. *Chin Arch TCM* (2018) **36** 2723-5
47. La DH. **Effect and mechanism of the decoction of benefiting kidney Qi and promoting blood circulation on primary membranous nephropathy Chinese high altitude medicine and biology.**. *Chin High Altit Med Bio* (2018) **39** 179-83
48. Dai M, Zhu Y. **Discussion with the effect of Qingrehuoxue Hushen Decoction in adjuvant treatment of membranous nephropathy.**. *Contemp Med Symp* (2018) **16** 186-8
49. Shen XN. *Clinical Observation on Treatment of Stage I and II Idiopathic Membranous Nephropathy (Spleen Kidney Yang Deficiency Type) with Jianpi Bushen Prescription [thesis]* (2018)
50. Xu X, Zhan JH, Guo YX. **Study on the effect of tripterygium wilfordii polyglycosides, Yiqi Huoxue Lishui Therapy combined with western medicine on the treatment of idiopathic membranous nephropathy and the expression of PCX.**. *Chin J Inf Contr* (2018) **33** 63-4
51. Liu XY, Che YJ, Li JH. **Efficacy evaluation of Huatan Quyu decoction for idiopathic membranous nephropathy with spleen-kidney Yang Deficiency Type.**. *InnerMongolia J TCM* (2018) **37** 28-9
52. Yu X. *Clinical Study on the Treatment of Idiopathic Membranous Nephropathy by Yiqi Huoxue Huayu Method [thesis]* (2018)
53. Zhao C. **Clinical observation on the treatment of StageⅠand Ⅱ membranous nephropathy by Yishen Jianpi Huoluo Decoction Combined with Western Medicine.**. *World Latest Med Inf* (2019) **19** 216-8
54. Lei GP, Zhu KR, Hu LF. **The effect of Qidigushen Pills on the treatment of PLA2R of idiopathic membranous nephropathy patients with the TCM syndrome differentiation qi-yin deficiency – a prospective randomized controlled clinical study.**. *Chin J Int Trad Western Nephro* (2019) **20** 508-10
55. Lou CL. **Clinical efficacy and safety evaluation of Jianpi Yiqi Qingre Huoxue decoction combined with different immunosuppressive agents in the treatment of IMN.**. *Modern Chin Doctor* (2019) **57** 124-30
56. Liu HX. *Clinical Study on Treatment of Idiopathic Membranous Nephropathy by Tonifying Kidney and Removing Blood Stasis and Clearing away Heat and its Effect on Serum PLA2R Antibody [thesis]* (2019)
57. Wei YJ. *Clinical Study of Qilong Tongshen Recipe in Treating Membranous Nephropathy with qi Deficiency and Blood Stasis Syndrome [thesis]* (2019)
58. Yang ZH, Lang XJ, Cheng D. **Clinical observation of Nephrotic Prescription No.1 combined with western medicine in treating membranous nephropathy.**. *Chin J Trad Med Sci Technol* (2019) **26** 55-7
59. Zhao C, Jia LM, Chen BC. **Observation on the curative effect of the prescription of Jianpi Yishen Huoxue Prescription in the treatment of atypical membranous nephropathy and the change of thrombus elastogram.**. *Chin J Trad Med Sci Technol* (2020) **27** 62-4
60. Zhang FF, Ai Y, Zhao BW. **Clinical observation on Xuantong Sanjiao and Huoxue Tongluo Formula in treating idiopathic membranous nephropathy.**. *China J TCM Pharm* (2020) **35** 1596-8
61. Hu G, Jiang QY, Zhang MT. **Effect of Yishen Huashi Granules combined with immunosuppressive drugs on idiopathic membranous nephropathy.**. *Int J Transpl Hemopurif* (2020) **18** 16-9
62. Li J, Ma Y, Zhu XT. **Clinical effect of Wenyang Qushi Tongluo recipe on idiopathic membranous nephropathy.**. *J Hebei TCM Pharmacol* (2020) **35** 26-8
63. Lei SB. **Tacrolimus combined with Huangqi Chifeng decoction in the treatment of elderly idiopathic membranous nephropathy.**. *Henan Med Res* (2020) **29** 476-7
64. Li YG. *Efficacy of Shenqizhilong Decoction in Idiopathic Membranous Nephropathy: A Randomised Controlled Trial [thesis]* (2020)
65. Wu J. **Effect of Qidi Gushen recipe in the treatment of patients with idiopathicmembranous nephropathy and its influence on laboratory indexes.**. *Clin Med Res Pract* (2021) **6** 133-8
66. Yang Y. *Differential Expression of MiRNA in Urinary Exosomes in Idiopathic Membranous Nephropathy and Intervention Study of Huangtu Yishen Granules [thesis]* (2021)
67. Guo YP, Guo BL, Liu XQ. **Effects of Qidi Taozhi Erchan recipe combined with tacrolimus on CD4+, CD8+ and urine protein quantification in patients with membranous nephropathy.**. *Clin Res Pract* (2021) **6** 129-31
68. Zhao R. *Huangtu Yishen Granules on Patients with Idiopathic Membranous Nephropathy Clinical Study on the Effect of Th17/Treg Immune Imbalance [thesis]* (2021)
69. Zhang YT. *Clinical Observation of Shenqi Dihuang Decoction Combined with Tacrolimus in the Treatment of Membranous Nephropathy with Deficiency of qi and Yin and Blood Stasis [thesis]* (2021)
70. Zhu YQ. *Clinical Observation of Qizhi Dihuang Decoction in Improving Hypercoagulable State of Idiopathic Membranous Nephropathy [thesis]* (2021)
71. Ma ZW. *Clinical Study on Idiopathic Membranous Nephropathy of Blood Stasis due to qi Deficiency with Jia-wei-bu-yang-huan-wu Powder [thesis]* (2011)
72. Li X, Lu B. **Clinical observation of therapy of invigorating spleen and kidney, activating blood and dispelling wind for idiopathic membranous nephropathy.**. *J New Chin Med* (2014) **46** 73-5
73. Cai Z, Wang L, Duan YF. **The clinical observatian of shenqidihuang soup on the treatment of the TCM syndrome differentiation qi-yin deficiency of idiopathic membranous nephropathy.**. *Chin J Med* (2016) **51** 82-5
74. Pang ZX. **Observation on the effect of combination of traditional Chinese and western medicine on the type of qi deficiency and blood stasis in idiopathic membranous nephropathy.**. *J Pract TCM* (2019) **35** 944-5
75. Cai Z, Wang L, Zhao WJ. **Clinical observation of Shenqi Dihuang Decoction on 32 cases of idiopathic membrane nephropathy with deficiency of Qi and Yin.**. *Beijing J TCM* (2019) **38** 1029-32
76. Duo HL, Wei HJ, Liang Y. **Clinical study on the treatment of 36 cases of idiopathic membranous nephropathy with spleen deficiency and kidney collateral stasis by Self-formulated Jianpi Lishi Tongluo prescription combined with western medicine.**. *Jiangsu J TCM* (2020) **52** 27-30
77. Pan Z. *Clinical Observation on Treatment of Qi Deficiency and Blood Stasis in II Stage Membranous Nephropathy with Shen Zhi HuoXue Decoction [thesis]* (2020)
78. Qiao LN. *Clinical Study of Yiqi Huashi Tongluo Decoction in Treating Idiopathic Membrane Nephropathy with Spleen and Kidney Qi Deficiency and Damp-Heat Stasis Syndrome [thesis]* (2020)
79. Wang L. **Clinical observation of Shenqi Dihuang decoction in the treatment of idiopathic membranous nephropathy with positive antibody of M - type phospholipase A2 receptor.**. *Shandong Univ CM* (2020) **42** 522-7
80. Ping GH. **Clinical observation of Qiqi Yishen capsule on idiopathic membranous nephropathy with spleen-kidney-qi deficiency and blood stasis syndrome.**. *Shanxi J TCM* (2021) **37** 22-4
81. Yang FW. **Clinical studies of Yishen Tongluo Recipe on intervention of low and medium risk of PLA2R-Ab positive membranous nephropathy.**. *Chin J Int Nephrop* (2021) **22** 1053-6
82. van de Logt AE, Fresquet M, Wetzels JF. **The anti-PLA2R antibody in membranous nephropathy: what we know and what remains a decade after its discovery.**. *Kidney Int* (2019) **96** 1292-302. PMID: 31611068
83. **KDIGO 2021 clinical practice guideline for the management of glomerular diseases.**. *Kidney Int* (2021) **100** S1-276. PMID: 34556256
84. Jha V, Ganguli A, Saha TK. **A randomized, controlled trial of steroids and cyclophosphamide in adults with nephrotic syndrome caused by idiopathic membranous nephropathy.**. *J Am Soc Nephrol* (2007) **18** 1899-904. PMID: 17494881
85. Caravaca-Fontán F, Fernández-Juárez GM, Floege J. **The management of membranous nephropathy-an update.**. *Nephrol Dial Transplant* (2022) **37** 1033-42. PMID: 34748001
86. Praga M, Barrio V, Juárez GF. **Tacrolimus monotherapy in membranous nephropathy: a randomized controlled trial.**. *Kidney Int* (2007) **71** 924-30. PMID: 17377504
87. Chen Y, Deng Y, Ni Z. **Efficacy and safety of traditional chinese medicine (Shenqi particle) for patients with idiopathic membranous nephropathy: a multicenter randomized controlled clinical trial.**. *Am J Kidney Dis* (2013) **62** 1068-76. PMID: 23810688
88. Zhu Y, Tang Y, Liqun HE. **Antagonistic effect of Astragalus Saponin I on renal interstitial fibrosis.**. *Liaoning J Trad Chin Med* (2014) **41** 2700-2
|
---
title: The risk of classical galactosaemia in newborns with borderline galactose metabolites
on newborn screening
authors:
- Isaac Bernhardt
- Emma Glamuzina
- Bryony Ryder
- Detlef Knoll
- Natasha Heather
- Mark De Hora
- Dianne Webster
- Callum Wilson
journal: JIMD Reports
year: 2022
pmcid: PMC9981414
doi: 10.1002/jmd2.12339
license: CC BY 4.0
---
# The risk of classical galactosaemia in newborns with borderline galactose metabolites on newborn screening
## Abstract
Newborn screening (NBS) for classical galactosaemia (CG) facilitates early diagnosis and treatment to prevent life‐threatening complications, but remains controversial, and screening protocols vary widely between programmes. False‐negatives associated with first‐tier screening of total galactose metabolites (TGAL) are infrequently reported; however, newborns with TGAL below the screening threshold have not been systematically studied. Following the diagnosis of CG in two siblings missed by NBS, a retrospective cohort study of infants with TGAL just below the cut‐off (1.5 mmol/L blood) was conducted. Children born in New Zealand (NZ) from 2011 to 2019, with TGAL 1.0–1.49 mmol/L on NBS were identified from the national metabolic screening programme (NMSP) database, and clinical coding data and medical records were reviewed. GALT sequencing was performed if CG could not be excluded following review of medical records. 328 infants with TGAL 1.0–1.49 mmol/L on NBS were identified, of whom 35 had ICD‐10 codes relevant to CG including vomiting, poor feeding, weight loss, failure to thrive, jaundice, hepatitis, *Escherichia coli* urinary tract infection, sepsis, intracranial hypertension and death. CG could be excluded in $\frac{34}{35}$, due to documentation of clinical improvement with continued dietary galactose intake, or a clear alternative aetiology. GALT sequencing in the remaining individual confirmed Duarte‐variant galactosaemia (DG). In conclusion, undiagnosed CG appears to be rare in those with TGAL 1.0–1.49 mmol/L on NBS; however, our recent experience with missed cases is nevertheless concerning. Further work is required to establish the optimum screening strategy, to maximize the early detection of CG without excess false‐positives.
## INTRODUCTION
Classical galactosaemia (CG) caused by galactose‐1‐phosphate uridyltransferase (GALT) deficiency occurs secondary to bi‐allelic GALT variants (OMIM#230400), at an incidence of ~$\frac{1}{50}$ 000 live‐births in Caucasians. 1 GALT catalyzes the conversion of galactose‐1‐phosphate to UDP‐galactose in the Leloir pathway, and deficiency leads to accumulation of galactose‐1‐phosphate and galactose among other metabolites. 2 CG presents in neonates with hepatopathy, cataract, encephalopathy and sepsis, and has high mortality unless recognized and treated with dietary galactose‐restriction. 2 Early detection by newborn screening (NBS) allows prompt initiation of galactose‐restricted diet, and prevention of life‐threatening disease complications and cataract. 3, 4 However NBS for galactosaemia remains controversial, due to the view that it is readily diagnosed clinically, and that neuro‐developmental complications occur despite treatment. 2, 5, 6 Newborn screening approaches for CG vary widely between screening programmes. 7, 8 CG screening protocols are influenced by regional variation in the incidence of CG and non‐classical galactosaemia, service capacity to manage false‐positives, and varied approaches to reporting Duarte‐variant galactosaemia (DG). 7, 9 NBS for galactosaemia was introduced in New Zealand (NZ) in the 1970's. The dried blood spot (DBS) sample is collected at 48–72 h, and is sent to the centralized national metabolic screening programme (NMSP). Total galactose metabolites (TGAL) comprising the sum of galactose and galactose‐1‐phosphate in whole‐blood, is used as the primary marker. Second‐tier analyses are performed if TGAL is above the cut‐off, including measurement of galactose‐1‐phosphate, galactose and GALT activity (quantitative assay). 10, 11 This approach is preferred due to the goal of detecting CG, galactokinase (OMIM#230200) and epimerase (OMIM#230350) deficiency, and to avoid detection of DG.
The NMSP detected 10 patients with CG from 2000 to 2008, and all had significantly elevated TGAL (range 2.24–16.6 mmol/L). The TGAL cut‐off during this period was 0.8 mmol/L, which was associated with a positive predictive value (PPV) well below $10\%$. Therefore, the cut‐off was increased to 1.5 mmol/L, but remained 0.8 mmol/L for infants in neonatal intensive care, due to potentially inadequate milk intake in this population. In 2019, two siblings with TGAL <1.5 mmol/L were missed by NBS then subsequently diagnosed with CG after presenting clinically with liver failure, raising concern that other CG patients had been missed by the NMSP. A retrospective cohort study of infants with TGAL 1.0–1.49 mmol/L was conducted, to estimate the incidence of false‐negatives associated with a cut‐off of 1.5 mmol/L. It was hypothesized that missed cases would become progressively unwell unless CG was diagnosed clinically, liver‐transplantation was performed, or low‐galactose feeds were started as treatment for an alternative disorder mimicking CG.
## MATERIALS AND METHODS
In NZ all infants are assigned a unique national health identifier (NHI) at birth, which is recorded on the NBS sample and all subsequent healthcare episodes. All NBS results are recorded in the NMSP database. Participation in NBS in NZ is voluntary, with high uptake (>$99.5\%$ births). The vast majority of secondary and tertiary healthcare episodes for children in NZ occur at public hospitals, and each healthcare episode is coded according to the International Classification of Disease‐10 (ICD‐10). Therefore significant childhood illnesses requiring hospital care are reliably captured, and ICD‐10 coding data can be readily matched to NBS results as previously described. 12 TGAL was measured in whole‐blood using a fluorescent galactose oxidase method. NHI numbers were obtained for all NBS samples from July 2011 to December 2019 with TGAL 1.0–1.49 mmol/L. ICD‐10 codes linked to these NHI were obtained from the Ministry of Health (MOH). Coding criteria considered consistent with a missed diagnosis of CG included: galactosaemia, poor feeding, poor growth, weight loss, vomiting, lactose intolerance, neonatal jaundice, liver disease/failure, hepatomegaly, cholestasis, hepatitis, coagulopathy, liver transplantation, sepsis, Escherichia Coli infection, cataract, renal tubulopathy, developmental delay, intracranial hypertension, encephalopathy, seizures, coma and death. If relevant coding criteria were present, review of clinical records was conducted to determine if the clinical course was potentially consistent with CG. Documented resolution of symptoms or normal growth and development, in the presence of ongoing dietary galactose intake, was considered to be inconsistent with CG.
If clinical records were potentially consistent with CG, GALT sequencing was performed on DNA extracted from the DBS sample. The coding regions of the GALT gene, flanking ±20 base‐pairs (encompassing the splice sites) and the 3′ end of the promoter region were sequenced by Sanger‐based sequencing. Analysis of sequence data was performed using Variant Reporter Software v2.
Data were also collected regarding total NBS samples during the study period, and those with TGAL >1.5 mmol/L including true‐positives and false‐positives. A positive screen was defined as an NBS result triggering additional actions including request for repeat DBS samples or clinical referral. The clinical presentation, NBS data and GALT sequencing results for known CG patients missed by NBS in the study period are also described.
## RESULTS
503 938 infants underwent NBS in the study period. Of these, 27 positive screens for CG were detected. CG was confirmed in only $\frac{2}{27}$ (PPV $7.4\%$). No patients were diagnosed with galactokinase, epimerase or mutarotase deficiency. 328 infants were identified with TGAL 1.0–1.49 mmol/L (Figure 1). Of these, 35 had relevant ICD‐10 codes including: poor feeding [9], neonatal jaundice [5], abnormal weight loss [5], poor weight gain [4], nausea and vomiting [3], sepsis [3], E. Coli urinary tract infection [2], coma [2], seizures [1], developmental delay [1], hepatitis [1], intracranial hypertension [1] and death [1]. One individual developed intracranial hypertension, and died at 3 years of age, and was found to have a large posterior fossa tumour at post‐mortem. Of the other cases, all but one were noted to have clinical improvement or normal growth and development, with evidence of ongoing regular dietary lactose intake.
**FIGURE 1:** *Summary of inclusion and exclusion criteria for cohort selection. CG, classical galactosaemia; ICD‐10, International Classification of Diesease‐10; TGAL, total galactose metabolites. *CG symptoms determined from ICD‐10 codes as described in ‘Methods’. # Excludes clinically‐presenting false‐negative CG cases (Patients A&B)*
One individual was identified for whom CG could not be excluded following review of clinical records. This individual presented to hospital at 6 days of age with $15\%$ weight loss since birth, jaundice, and apnoea. Formula feeds were given by nasogastric tube during this admission. However note was made of a change of formula prior to discharge; unfortunately the new formula preparation was not documented. Subsequent records at 3 years of age documented hearing loss, hypotonia, global developmental delay and normal liver enzymes, but dietary history was not recorded. GALT sequencing showed that this individual was heterozygous for the pathogenic c.563A > G, p.(Gln188Arg) variant and the Duarte (D2) variant c.‐119_‐116delGTCA, consistent with DG. This result was not considered relevant to the clinical presentation.
During the study period, two individuals presented clinically with CG, and were missed by NBS. These siblings were born at term, and were breastfed prior to NBS at 2 days of age (Table 1). TGAL was 0.62 mmol/L in the older male sibling (Patient A), and 0.81 mmol/L in the younger female sibling (Patient B).
**TABLE 1**
| Unnamed: 0 | Patient A Missed on NBS and clinically as repeat testing was taken on intravenous fluids and after red cell transfusion | Patient B (Sibling of Patient A who was 13 months at the time of her birth, well and had no diagnosis) |
| --- | --- | --- |
| Sex | Male | Female |
| Gestational age at birth | 41 weeks | 38 weeks + 6 days |
| Birthweight | 4.2 kg | 3.6 kg |
| Feeds prior to NBS | Breastmilk | Breastmilk |
| NBS TGAL (2 days of age; cut off 1.5 mmol/L) | 0.62 mmol/L | 0.81 mmol/L |
| Feeds prior to presentation | Standard cow's milk infant formula | Breastmilk |
| Age at presentation | 4 weeks | 10 days |
| Presenting symptoms | Poor growth, liver failure with jaundice and severe ascites | Weight loss, liver failure with coagulopathy and hyperammonaemia |
| TGAL at presentation | 1.28 mmol/L a | 9.2 mmol/L b |
| Galactose (normal < 1.35 mmol/L) | 0.9 mmol/L a | 9.2 mmol/L b |
| Gal‐1‐P (normal < 1.35 mmol/L) | 0.62 mmol/L a | 0.0 mmol/L b |
| GALT activity (quantitative assay) 8 , 9 | 43% a | 6% b |
| Initial treatment | Extensively hydrolysed formula (high MCT, low galactose) for unexplained conjugated jaundice | Extensively hydrolysed formula (high MCT, low galactose) |
| Progress following initial treatment | All presenting symptoms resolved | Coagulation and ammonia normalized, growth and liver enzymes improved |
| Age at diagnosis | 14 months (diagnosed following diagnosis of younger sibling, Patient B) | 6 weeks |
| GALT genotype | Homozygous c.563A > G, p.Gln188Arg | Homozygous c.563A > G, p.Gln188Arg |
| Treatment following CG diagnosis | Galactose‐restricted diet including soy formula | Galactose‐restricted diet including soy formula |
| Repeat GALT activity at diagnosis c | <5% | <5% |
| Gal‐1‐P on low‐galactose feeds following diagnosis | 0.17 mmol/L | 0.24 mmol/L |
| Outcome | 4 years of age: liver disease resolved, mild bilateral cataracts, subtle issues with behavior and speech development | 3 years of age; liver disease resolved, normal growth and development |
Patient A was fed with standard infant formula containing lactose prior to presentation with liver failure at 4 weeks of age. TGAL was 1.28 mmol/L at 35 days of age, with galactose 0.9 mmol/L, galactose‐1‐phosphate 0.62 mmol/L and GALT activity $43\%$ of the mean normal value. He had received intravenous fluids and red cell transfusion in the preceding 4 days, thus explaining the normal GALT activity. Extensively hydrolysed formula feeds were started for management of cholestasis, due to their high medium‐chain triglyceride (MCT) content; however, this preparation was also low in galactose. He made a full clinical recovery with no evidence of liver disease or cataract at 14 months of age.
Patient B was born 13 months after Patient A, and presented with liver failure at 10 days of age. TGAL was 9.2 mmol/L at 17 days with GALT activity <$10\%$ of the mean normal value. Whole exome sequencing in Patient B identified the homozygous GALT variant c.563A > G, p.(Gln188Arg), which was then detected in Patient A by Sanger sequencing.
## DISCUSSION
Population NBS for CG has been feasible since 1964, 6 and is now undertaken by many screening programmes. Screening protocols vary significantly between centres, and include primary measurement of either GALT activity or galactose metabolites (TGAL), with variable screening thresholds. 7, 8, 13 First‐tier TGAL measurement enables detection of galactokinase, epimerase and the recently described mutarotase deficiency (OMIM#618881), in addition to CG. 14, 15 However TGAL may also be raised in DG, heterozygous CG carriers, neonatal liver disease and other rare inherited metabolic disorders, and therefore second‐tier analysis of galactose metabolites and GALT activity is used to increase specificity. 15 The laboratory burden of second‐tier testing is subject to the TGAL cut‐off used; however, false‐positives occur at a significant rate despite second‐tier testing. 4 False‐negatives associated with primary TGAL screening are thought to be rare, but have been reported in association with hypomorphic GALT alleles, or when galactose intake prior to sampling is significantly limited. 8, 16, 17 To our knowledge, this is the first systematic study of infants with TGAL below the screening threshold.
In this cohort of 328 newborns with TGAL in the range of 1.0–1.49 mmol/L, no missed cases of CG were identified, and therefore the risk of undiagnosed CG in this group appears to be low. It is unlikely but plausible that further individuals with CG in this cohort escaped detection by this review; for example, if low‐galactose feeds were initiated in primary care for another indication, in a newborn with very mild symptoms not requiring hospital care.
The primary limitation of this study is that GALT sequencing was only performed in one individual, who was found to have DG. It is likely that many others also had DG, or were heterozygous CG carriers. However neither DG or CG‐heterozygosity is associated with short‐ or long‐term medical complications. 9, 18, 19, 20 Therefore their detection by NBS is undesirable, and genetic testing for these entities was not considered to be clinically relevant or ethically justifiable. While no individuals with cataract were identified in this cohort, testing for variants in GALK1 or GALM may also have been informative; however, detection of galactokinase or mutarotase deficiency was not the purpose of this study.
Newborns with TGAL just below the screening threshold of 1.5 mmol/L were selected for inclusion, due to the presumed higher risk of false‐negatives in this group. However it is noteworthy that the two index cases missed on NBS had TGAL well below the cut‐off and were not captured by the study criteria. It is also noteworthy that while both siblings were homozygous for the common pathogenic c.563A > G, p.(Gln188Arg) variant associated with severe disease, 5 Patient A had an unusually late presentation despite ongoing dietary galactose intake. While no known risk factors for false‐negative screening were detected in these patients, it is peculiar that they were siblings, and it is therefore possible that other shared genetic or environmental factors contributed to an attenuated elevation of TGAL at the time of NBS. Unfortunately, diagnostic delay in Patient A was further compounded by falsely normal repeat biochemistry, due to preceding fluid and transfusion therapy.
The prevalence of CG diagnosed on NBS in the study period was unexpectedly low. 19 individuals were diagnosed with CG in NZ over 20 years (incidence $\frac{1}{63}$ 000 live‐births), consistent with the expected population incidence. 1 However during the study period the observed incidence was $\frac{4}{503}$ 938 ($\frac{1}{125}$ 000), and $\frac{2}{4}$ were missed by NBS (sensitivity $50\%$). While no additional missed cases were identified in this study, the low detection rate of CG by NBS over this period is cause for concern. A survey of other Australasian screening programmes, of which $\frac{4}{5}$ undertake NBS for CG, revealed five additional missed cases (TGAL range 0.71–1.09 mmol/L). Of these patients, only $\frac{2}{5}$ were not receiving galactose feeds prior to NBS sample collection at 2–3 days of age. These programmes have all lowered their cut‐offs, with three now using 0.5 mmol/L and one using 1.0 mmol/L. While a screening threshold of 0.5 mmol/L would detect the index cases presented here, it would significantly increase second‐tier testing and false‐positives (including DG). Reporting of false positive NBS results may cause significant distress for the family of an affected infant, and this recognized harm of NBS needs to be balanced against the benefits of increased detection of CG. 21 Beyond Australasia, the GALT assay is more commonly used as the primary screening test, and false‐negatives using this approach appear to be rare. 6, 7, 13 Although false‐negative GALT assays are known to occur following neonatal transfusion, 6 this is likely to be an increasingly uncommon scenario. 22 Use of a primary GALT approach is associated with a significant burden of false‐positives, especially during summer months due to heat degradation of the DBS sample. 23 Additionally, it does not detect non‐classical galactosaemia due to galactokinase, epimerase or mutarotase deficiency, which are secondary targets of NBS using a primary TGAL approach. Both approaches merit consideration, and the NMSP has implemented an interim TGAL cut‐off of 0.8 mmol/L.
Finally, the authors acknowledge that large‐scale prospective studies confirming the benefit of NBS for CG are lacking. 8, 24 However early treatment is associated with reduced mortality in uncontrolled studies. 3, 25 In the experience of the NMSP, NBS facilitates timely detection and treatment, particularly with regards to early identification of coagulopathy, which if unrecognized, has resulted in fatal outcome.
## CONCLUSION
Clinically presenting patients with CG, with TGAL well below cut‐off, have been missed by the NMSP and other screening programmes that utilize a primary TGAL method. In this retrospective cohort study of newborns with TGAL 1.0–1.49 mmol/L on NBS, no additional missed cases of CG were identified. This suggests that false‐negatives associated with TGAL in this range are not common, and such patients are likely to present clinically. The occurrence of clinically presenting false‐negative cases with TGAL between 0.5–1.0 mmol/L is nevertheless concerning, and warrants consideration of a significantly lower threshold or alternative screening strategy. NBS for CG may significantly reduce morbidity and mortality in affected infants; however, the optimum screening strategy and thresholds remain to be clarified.
## CONFLICT OF INTEREST
Isaac Bernhardt declares no conflict of interest. Emma Glamuzina declares no conflict of interest. Bryony Ryder declares no conflict of interest. Detlef Knoll declares no conflict of interest. Natasha Heather declares no conflict of interest. Mark De Hora declares no conflict of interest. Dianne Webster declares no conflict of interest. Callum Wilson declares no conflict of interest.
## ETHICAL STATEMENT/INFORMED CONSENT
Participation for Newborn Screening (NBS) in New *Zealand is* voluntary. Since July 2011, routine informed consent for NBS has included consent for additional uses of the dried blood‐spot sample, for activities including quality control and audit by the National Metabolic Screening Programme, as well as for research approved by an ethics committee and the Ministry of Health. Therefore, only samples obtained subsequent to the implementation of this policy were included. This study was approved by the Health and Disability Ethics Committee. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000 [5]. Written consent for publication of de‐identified information has been obtained from the parent/guardian of Case A and Case B.
## DATA AVAILABILITY STATEMENT
Data archiving is not mandated but data will be made available on reasonable request.
## References
1. Fridovich‐Keil JL, Walter JH, Valle D, Beaudet AL, Vogelstein B, Kinzler KW, Antonarakis SE, Ballabio A. *The Online Metabolic and Molecular Bases of Inherited Disease* (2008) 1-92
2. Coelho AI, Rubio‐Gozalbo ME, Vicente JB, Rivera I. **Sweet and sour: an update on classic galactosemia**. *J Inherit Metab Dis* (2017) **40** 325-342. PMID: 28281081
3. Ohlsson A, Guthenberg C, von Döbeln U. **Galactosemia screening with low false‐positive recall rate: the Swedish experience**. *JIMD Rep* (2012) **2** 113-117. PMID: 23430863
4. Welling L, Boelen A, Derks TG. **Nine years of newborn screening for classical galactosemia in The Netherlands: effectiveness of screening methods, and identification of patients with previously unreported phenotypes**. *Mol Genet Metab* (2017) **120** 223-228. PMID: 28065439
5. Rubio‐Gozalbo ME, Haskovic M, Bosch AM. **The natural history of classic galactosemia: lessons from the GalNet registry**. *Orphanet J Rare Dis* (2019) **14** 86. PMID: 31029175
6. Schweitzer‐Krantz S. **Early diagnosis of inherited metabolic disorders towards improving outcome: the controversial issue of galactosaemia**. *Eur J Pediatr* (2003) **162** S50-S53. PMID: 14614623
7. Pyhtila BM, Shaw KA, Neumann SE, Fridovich‐Keil JL. **Newborn screening for galactosemia in the United States: looking back, looking around, and looking ahead**. *JIMD Rep* (2015) **15** 79-93. PMID: 24718839
8. Varela‐Lema L, Paz‐Valinas L, Atienza‐Merino G, Zubizarreta‐Alberdi R, Villares RV, López‐García M. **Appropriateness of newborn screening for classic galactosaemia: a systematic review**. *J Inherit Metab Dis* (2016) **39** 633-664. PMID: 27116003
9. Carlock G, Fischer ST, Lynch ME. **Developmental outcomes in Duarte galactosemia**. *Pediatrics* (2019) **143**. PMID: 30593450
10. Beutler F, Baluda M. **A simple spot screening test for galactosemia**. *J Lab Clin Med* (1966) **68** 137-141. PMID: 4380286
11. Fujimoto A, Okano Y, Miyagi T, Isshiki G, Oura T. **Quantitative Beutler test for newborn mass screening of galactosemia using a fluorometric microplate reader**. *Clin Chem* (2000) **46** 806-810. PMID: 10839768
12. Wilson C, Knoll D, de Hora M, Kyle C, Glamuzina E, Webster D. **The risk of fatty acid oxidation disorders and organic Acidemias in children with Normal newborn screening**. *JIMD Rep* (2017) **35** 53-58. PMID: 27928776
13. Jumbo‐Lucioni PP, Garber K, Kiel J. **Diversity of approaches to classic galactosemia around the world: a comparison of diagnosis, intervention, and outcomes**. *J Inherit Metab Dis* (2012) **35** 1037-1049. PMID: 22450714
14. Wada Y, Kikuchi A, Arai‐Ichinoi N. **Biallelic GALM pathogenic variants cause a novel type of galactosemia**. *Genet Med* (2019) **21** 1286-1294. PMID: 30451973
15. Berry GT. **Galactosemia: when is it a newborn screening emergency?**. *Mol Genet Metab* (2012) **106** 7-11. PMID: 22483615
16. Crushell E, Chukwu J, Mayne P, Blatny J, Treacy EP. **Negative screening tests in classical galactosaemia caused by S135L homozygosity**. *J Inherit Metab Dis* (2009) **32** 412-415. PMID: 19418241
17. Bosch AM. **Galactosaemia ‐ should it be screened in newborns?**. *Dev Period Med* (2018) **22** 221-224. PMID: 30281516
18. Welling L, Bernstein LE, Berry GT. **International clinical guideline for the management of classical galactosemia: diagnosis, treatment, and follow‐up**. *J Inherit Metab Dis* (2017) **40** 171-176. PMID: 27858262
19. Fridovich‐Keil JL, Carlock G, Patel S, Potter NL, Coles CD. **Acute and early developmental outcomes of children with Duarte galactosemia**. *JIMD Rep* (2022) **63** 101-106. PMID: 35028275
20. Ficicioglu C, Thomas N, Yager C. **Duarte (DG) galactosemia: a pilot study of biochemical and neurodevelopmental assessment in children detected by newborn screening**. *Mol Genet Metab* (2008) **95** 206-212. PMID: 18976948
21. Wilcken B. **Expanded newborn screening: reducing harm, assessing benefit**. *J Inherit Metab Dis* (2010) **33** 205-210
22. Steiner LA, Bizzarro MJ, Ehrenkranz RA, Gallagher PG. **A decline in the frequency of neonatal exchange transfusions and its effect on exchange‐related morbidity and mortality**. *Pediatrics* (2007) **120** 27-32. PMID: 17606558
23. Thibodeau D, Andrews W, Meyer J. **Comparison of the effects of season and prematurity on the enzymatic newborn screening tests for galactosemia and biotinidase deficiency**. *Screening* (1993) **2** 19-27
24. Lak R, Yazdizadeh B, Davari M, Nouhi M, Kelishadi R. **Newborn screening for galactosaemia**. *Cochrane Database Syst Rev* (2020) **22**
25. Padilla CD, Lam ST. **Issues on universal screening for galactosemia**. *Ann Acad Med Singapore* (2008) **3** 39-33
|
---
title: The role of ferroptosis-related genes in airway epithelial cells of asthmatic
patients based on bioinformatics
authors:
- Ye Zheng
- Jingyao Fan
- Xiaofeng Jiang
journal: Medicine
year: 2023
pmcid: PMC9981416
doi: 10.1097/MD.0000000000033119
license: CC BY 4.0
---
# The role of ferroptosis-related genes in airway epithelial cells of asthmatic patients based on bioinformatics
## Abstract
It has been reported that airway epithelial cells and ferroptosis have certain effect on asthma. However, the action mechanism of ferroptosis-related genes in airway epithelial cells of asthmatic patients is still unclear. Firstly, the study downloaded the GSE43696 training set, GSE63142 validation set and GSE164119 (miRNA) dataset from the gene expression omnibus database. 342 ferroptosis-related genes were downloaded from the ferroptosis database. Moreover, differentially expressed genes (DEGs) between asthma and control samples in the GSE43696 dataset were screened by differential analysis. Consensus clustering analysis was performed on asthma patients to classify clusters, and differential analysis was performed on clusters to obtain inter-cluster DEGs. Asthma-related module was screened by weighted gene co-expression network analysis. Then, DEGs between asthma and control samples, inter-cluster DEGs and asthma-related module were subjected to venn analysis for obtaining candidate genes. The last absolute shrinkage and selection operator and support vector machines were respectively applied to the candidate genes to screen for feature genes, and functional enrichment analysis was performed. Finally, a competition endogenetic RNA network was constructed and drug sensitivity analysis was conducted. There were 438 DEGs (183 up-regulated and 255 down-regulated) between asthma and control samples. 359 inter-cluster DEGs (158 up-regulated and 201 down-regulated) were obtained by screening. Then, the black module was significantly and strongly correlated with asthma. The venn analysis yielded 88 candidate genes. 9 feature genes (NAV3, ITGA10, SYT4, NOX1, SNTG2, RNF182, UPK1B, POSTN, SHISA2) were screened and they were involved in proteasome, dopaminergic synapse etc. Besides, 4 mRNAs, 5 miRNAs, and 2 lncRNAs collectively formed competition endogenetic RNA regulatory network, which included RNF182-hsa-miR-455-3p-LINC00319 and so on. The predicted therapeutic drug network map contained NAV3-bisphenol A and other relationship pairs. The study investigated the potential molecular mechanisms of NAV3, ITGA10, SYT4, NOX1, SNTG2, RNF182, UPK1B, POSTN, SHISA2 in airway epithelial cells of asthmatic patients through bioinformatics analysis, providing a reference for the research of asthma and ferroptosis.
## 1. Introduction
Bronchial asthma is a common chronic inflammatory disease worldwide, which affects more than 300 million people and brings a heavy burden to the public healthcare system.[1,2] *It is* induced by allergens, and is accompanied by cough, shortness of breath, and wheezing. Repeated inflammatory stimulation of bronchial asthma can lead to airway hyperresponsiveness, airway remodeling, and mucus secretion.[3] Airway epithelial cells are the primary barrier to allergens, and involved in the development of asthma. Firstly, airway epithelial cells can clear allergens through mucociliary clearance mechanisms.[3] Mucin dysregulation has been observed in asthmatic patients. Secondly, airway epithelial cells can recognize pathogen/danger-associated molecular patterns through a variety of pattern recognition receptors, and then they can secrete alarmin cytokines (IL25, IL33, and TSLP) and chemokines (CCL5, CCL7, and CCL22), causing immune cell infiltration via chemotactic effects and resulting in local immune response.[4–8] Thirdly, they also act as antigen-presenting cells, present antigen peptides, and induce the differentiation of naive T cells into CD4+ T cells, which plays a pivotal role in type II immune response.[9,10] To date, focusing on adaptive immune resistant components in asthmatic patients has not fully cured asthma. Thus, we hypothesize that airway epithelial cell defects may also contribute to asthma.
Ferroptosis, a non-apoptosis regulated cell death characterized by iron-dependent lipid peroxidation and reactive oxygen species (ROS) accumulation, has significantly different morphological characteristics, biochemical indicators, and genetics than other known cell death patterns.[11–13] In ferroptosis, the cell membrane is intact, and the nuclear morphology is not altered. However, the mitochondrial function is significantly impaired, the raw materials for intracellular synthesis of glutathione (GSH) are reduced, the glutathione peroxidase 4 is inactivated, and ROS in the cytoplasm is increased.[14] In the lung tissue of mice with house dust mite induced asthma, the glutathione peroxidase 4 and solute carrier family 7 member 11 (SLC7A11) which is an important negative regulator of ferroptosis were significantly decreased, whereas ROS levels were significantly increased, compared to normal controls. Additionally, IL4, IL13, and IL33 produced by airway epithelial cells were increased in the bronchoalveolar lavage fluid of asthmatic mice.[15] Treatment with ferroptosis inhibitors ferrostatin-1 (Fer-1) could alleviate airway inflammation and reduce airway epithelial cell death, indicating that airway epithelial cell ferroptosis is closely related to asthma.[16] However, in human airway epithelial cells, the genes significantly related to ferroptosis of airway epithelial cells are unknown, and whether ferroptosis genes can be used as biomarkers to discriminate normal and asthmatic patients has not yet been reported.
Therefore, this study identified 9 feature genes significantly related to ferroptosis and asthma in human airway epithelial cells, and their ability to diagnose asthma, regulatory networks, biological functions, and therapeutic drug network were analyzed. Our findings may provide new direction for the pathogenesis and treatment of asthma.
## 2.1. Data source
This analysis downloaded the GSE43696 training set (Asthma: Control = 88: 20), GSE63142 validation set (Asthma: Control = 128: 27), and GSE164119 (miRNA) (Asthma: Control = 9: 7) datasets from gene expression omnibus database (https://www.ncbi.nlm.nih.gov/gds). Among them, GSE164119 (miRNA) and GSE43696 datasets were respectively used for the identification of differentially expressed miRNA (DE-miRNA) and differentially expressed lncRNA (DE-lncRNA). 342 ferroptosis-related genes (FRGs) were downloaded from ferroptosis database (FerrDb; http://www.zhounan.org/ferrdb/current/).
## 2.2. Acquisition of differentially expressed FRGs (DE-FRGs)
Based on GSE43696 dataset, differentially expressed genes (DEGs) between asthma samples and control samples were screened by limma package setting the condition of |log2FC| > 0.5 and $P \leq .05.$[17] The volcano map was plotted based on the obtained results. DEGs of Top100 were displayed by drawing a heat map with the pheatmap package.[18] Then, venn analysis of DEGs and FRGs was performed to obtain DE-FRGs.
## 2.3. Enrichment analysis of DE-FRGs
*The* gene ontology (GO) system included biological process (BP), molecular functions (MF), and cellular components (CC). Kyoto encyclopedia of genes and genomes (KEGG) database comprehensively integrated genomic, proteomic, chemical components, and other systematic functional information. This study used clusterprofiler package to perform GO and KEGG enrichment analysis on DE-FRGs ($P \leq .05$). The enrichment results were presented using the ggplot2 package.[19,20]
## 2.4. Construction of the protein-protein interaction (PPI) network
In order to explore whether there was a reciprocal relationship among DE-FRGs, we used STRING (https://string-db.org) website to set confidence = 0.4 to remove discrete proteins and obtain a PPI network. Moreover, based on GSE43696 dataset, the receiver operating characteristic (ROC) curve was drawn through pROC package to explore diagnostic power of key genes.[21] Then, the assessment results were validated by GSE63142 dataset.
## 2.5. Consensus clustering analysis
The study used consensusclusterplus package to perform consensus clustering analysis on 88 asthma patients according to DE-FRGs.[22] The number of clusters k was chosen to classify asthma patients, and principal component analysis was performed on the clusters to clarify their distribution. The differential analysis was performed on clusters by limma package setting |log2fold change| > 0.5 and $P \leq .05$ to screen out inter-cluster DEGs.[17] In addition, the box plot was drawn to show the expression trends of DE-FRGs in different clusters.
## 2.6. Functional enrichment analysis of inter-cluster DEGs
Firstly, we respectively used c2.cp.kegg.v7.2.symbols.gmt and c5.go.bp.v7.5.1.symbols.gmt as the reference gene sets. Based on the gene expression profile files, the GO and KEGG pathways was scored in each sample by GSVA package.[23] Then, GO and KEGG entries with inter-cluster differences were obtained by limma package setting the |log2(fold change)| > 0.1 and $P \leq .05.$ The expression heat map was plotted by pheatmap package to display the results. In addition, GO and KEGG pathways on inter-cluster DEGs were performed using the clusterprofiler package. The enrichment results were visualized by ggplot2 package.
## 2.7. Weighted gene co-expression network analysis (WGCNA)
In this study, all gene expression matrices of 108 samples (Asthma: Control = 88: 20) in the GSE43696 dataset were treated as input data by WGCNA package.[24] Asthma and control were used as trait to construct a co-expression network. Next, in order to ensure the accuracy of the analysis, the samples were clustered for excluding the outliers. Sample clustering and clinical trait heat map were constructed to visualize the results. A soft threshold was determined for the data to ensure that the interactions among genes were maximally consistent with the scale-free distribution. The modules were segmented by dynamic tree cutting algorithm setting minModuleSize = 300. Finally, we performed correlation analysis between gene modules and asthma, and selected the module which was most related to the disease as the asthma-related module.
## 2.8. Delineation of gene network clusters
Gene pairs in modules which was significantly strongly correlated with asthma were found using the STRING (https://string-db.org) website, and they were imported into cytoscape software. Gene network clusters were classified by molecular complex detection (MCODE) with the Degree Cutoff = 2, Node score cutoff = 2, k-score = 2 and Max. Depth = 100. *The* gene clusters of MCODE score top5 were selected for display.
## 2.9. Screening out feature genes by machine learning method
First, we performed venn analysis on DEGs between asthma and control samples, inter-cluster DEGs and asthma-related module to obtain candidate genes. Then, least absolute shrinkage and selection operator (LASSO) regression analysis was performed on the candidate genes by glmnet package to obtain gene coefficient map and cross-validation error map.[25] The support vector machines (SVM)-recursive feature elimination (RFE) algorithm was acted on the candidate genes to narrowed down the range of feature genes. Finally, the intersection of the genes screened by LASSO and SVM algorithms was taken to obtain the feature genes. The GSE43696 dataset contained 50 mild to moderate asthma samples, 38 severe asthma samples and 20 healthy samples. We divided the GSE43696 dataset into 2 small datasets of 50 mild to moderate asthma samples versus 20 healthy samples and 38 severe asthma samples versus 20 healthy samples. In order to test whether the feature genes had ability to identify different levels of asthma, the ROC analysis was made for two small datasets, and the GSE63142 dataset was used to verify the results.
## 2.10. Construction of a nomogram and gene set enrichment analysis (GSEA) of feature genes
Based on the GSE43696 dataset, we used rms package to construct the diagnostic nomogram of feature genes and clinical factors.[26] Scores for age and gender clinical factors were calculated, and the scores for each factor were summed to obtain total points, then, the survival rate of patients was predicted based on the total points. Calibration curve was plotted to verify the validity of nomogram. Besides, the feature genes were correlated with all genes of the asthma samples in turn and ranked. The screening conditions for GSEA were SIZE > 20 and NOM.$P \leq .05.$
## 2.11. Screening for DE-miRNA and DE-lncRNA
Differential analysis was performed on 7 control samples and 9 asthma samples in the GSE164119 dataset using the edgeR package.[27] *Differential analysis* of 20 control samples and 88 asthma samples in the GSE43696 dataset was carried out by limma package. The screening criteria for DE-miRNA and DE-lncRNA were |log2fold change| > 0.5 and $P \leq .05.$ Volcano plot and heat map were drawn to visualize the results.
## 2.12. Construction of a competing endogenous RNA (ceRNA) regulatory network
ceRNA is a role element that can compete for binding RNA. We used miRWalk website (http://mirwalk.umm.uni-heidelberg.de/) to predict the miRNAs of up-regulated feature genes with default parameters score = 0.95 and position = 3UTR. The miRNAs were intersected with the down-regulated miRNAs to obtain the miRNA1. lncRNA1 which interacted with miRNA1 was forecasted using the starBase database. *The* genes with regulatory relationships were extracted based on the miRNA1. Then, miRWalk was used to predict the miRNAs of down-regulated feature genes, and the miRNAs were intersected with the up-regulated miRNAs to obtain the miRNA 2. lncRNA2 that interacted with miRNA2 was predicted through starBase database. The shared lncRNAs were obtained by taking intersection of lncRNA 2 and down-regulated lncRNAs, then, the miRNAs and genes with regulatory relationships were extracted from the shared lncRNAs. Finally, mRNA-miRNA-lncRNA interaction pairs were obtained by integrating the prediction results of up- and down-regulated feature gene. The interaction relationships were visualized using the cytoscape software.
## 2.13. Construction of the mRNA-transcription factor (TF) regulatory network
In this study, we used NetworkAnalyst (https://www.networkanalyst.ca/) database to retrieve the TFs which regulated feature genes. A mRNA-TF regulatory network was constructed using cytoscape software on feature genes and TFs. To explore the correlation between the feature genes and TF, their correlation coefficients and P value were calculated by spearman. The results were displayed by drawing correlation heat map.
## 2.14. Prediction of potential therapeutic drugs
The feature genes were entered in CTD database (http://ctdbase.com/) to predict potential therapeutic agents for the treatment of asthma patients. The screening condition was Interaction Count ≥ 2.
## 3.1. Identification and functional annotation analysis of DE-FRGs
There were 438 DEGs (183 up-regulated DEGs and 255 down-regulated DEGs) between asthma and control samples. The results were visualized in heat and volcano plots (Fig. 1 A and B). 9 DE-FRGs were differentially expressed between asthma and control samples (Fig. 1C). DE-FRGs were involved in 575 BP, 20 MF, and 13 KEGG signaling pathways, of which BP included the lipoxygenase pathway, response to oxygen levels, regulation of peptidyl-serine phosphorylation and others, KEGG pathways contained AGE-RAGE signaling pathway in diabetic complications, fluid shear stress and atherosclerosis, HIF-1 signaling pathway, etc. The bubble diagram and bar graph displayed a part of enriched GO and KEGG pathways (Fig. 1D and E).
**Figure 1.:** *Identification and functional annotation analysis of DE-FRGs. (A) Volcano plot of differential genes of asthma and normal control subjects. (B) Heat map of up and down Top100 differential genes of asthma and normal control subjects. Each small square represents each gene, and its color represents the expression level of the gene. The dendrogram on the left represents the cluster analysis results of different genes from different samples. (C) Venn diagram of DEGs and ferroptosis-related genes (FRGs). (D) GO function annotation results (Top10). (E) KEGG pathway enrichment results (Top10). DE-FRGs = differentially expressed FRGs, DEGs = differentially expressed genes, GO = gene ontology, KEGG = Kyoto encyclopedia of genes and genomes.*
## 3.2. Identification and evaluation of the key genes
A total of 6 nodes and 8 edges constituted the PPI network, which included IL6-NOX1, VEGFA-DDIT4, etc. ( Fig. 2A). The 6 obtained genes namely IL6, NOX1, VEGFA, DDIT4, MEF2C, NOS2 were taken as key genes. From the ROC curve of the GSE43696 dataset, we could see that area under the curve (AUC) of DDIT4 = 0.783, NOX1 AUC = 0.82, MEF2C AUC = 0.777, NOS2 AUC = 0.715, VEGFA AUC = 0.741, and IL6 AUC = 0.715, their AUC values were all greater than 0.7, indicating that the key genes had a better ability to distinguish asthma and control. In the ROC curve of the GSE63142 validation set, the AUC of the key genes were all exceeded 0.6, similarly validating a good diagnostic ability of their (Fig. 2B).
**Figure 2.:** *Identification and evaluation of the key genes. (A) The PPI network downloaded from the STRING database indicated the interactions among the 6 DE-FRGs. (B) ROC curves of 6 DE-FRGs in training set (GSE43696) and validation set (GSE63142). DE-FRGs = differentially expressed FRGs, PPI = protein-protein interaction, ROC = receiver operating characteristic.*
## 3.3. Classification of asthma patient clusters and identification of inter-cluster DEGs
Asthma patients were divided into cluster 1 ($$n = 42$$) and cluster 2 ($$n = 46$$) according to 9 DE-FRGs by consensus clustering analysis (Fig. 3A). The results of principal component analysis showed that 2 clusters were clearly distributed in different regions (Fig. 3B). A total of 359 inter-cluster DEGs (158 were up-regulated and 201 were down-regulated) were obtained after the screening. It could be seen from the box plot that the expression of ALOXE3, BEX1, IL6, MEF2C, NOS2, and NOX1 were extremely significant different between cluster 1 and cluster 2 (Fig. 3C).
**Figure 3.:** *Classification of asthma patient clusters and identification of inter-cluster DEGs. (A) Consensus Cluster analysis identified two distinct subtypes of asthma subjects according to DE-FRGs within the training set. (B) Graph of PCA score plot for cluster 1 and cluster 2 of asthma subjects. (C) Expression of 9 DE-FRGs in cluster 1 and cluster 2. CDF= cumulative distribution function, DE-FRGs = differentially expressed FRGs, DEGs = differentially expressed genes, GSVA= gene set variation analysis, PCA = principal component analysis.*
## 3.4. Functional annotation analysis of inter-cluster DEGs
There were 14 GO entries and 33 KEGG pathways between cluster 1 and cluster 2. GO entries included hydrolase activity acting on ether bonds, midbody, oxidoreductase activity, etc., and KEGG signaling pathways contained ppar signaling pathway, histidine metabolism, drug metabolism cytochrome P450 and others (Fig. 4A and B). In addition, a total of 63 BP, 10 CC, 12 MF, and 4 KEGG enrichment pathways were enriched by inter-cluster DEGs, in which the BP pathways covered regulation of neural precursor cell proliferation, myeloid leukocyte migration and others. KEGG signaling pathways consisted of salivary secretion, cytokine-cytokine receptor interaction, neuroactive ligand-receptor interaction and IL-17 signaling pathway. The bar and bubble diagrams showed the GO and KEGG enrichment pathways (Fig. 4C and D).
**Figure 4.:** *Functional annotation analysis of inter-cluster DEGs. (A) Heatmap of GSVA score for GO function annotation of cluster 1 and cluster 2. (B) Heatmap of GSVA score for KEGG enrichment of cluster 1 and cluster 2. (C) GO function annotation in differential genes of cluster 1 and cluster 2. (D) KEGG pathway enrichment results in differential genes of cluster 1 and cluster 2. DEGs = differentially expressed genes, GO = gene ontology, KEGG = Kyoto encyclopedia of genes and genomes.*
## 3.5. Screening of the gene module
From Figure 5A, it could be seen that the overall clustering of the dataset samples was good and no sample elimination was required. The sample clustering and clinical trait heat maps were presented for the results (Fig. 5B). The optimal soft threshold was determined to be 7 by WGCNA, and the R2 was around 0.87 at this time, indicating that the network approximated a scale-free network (Fig. 5C). A series of 10 gene modules were screened by constructing co-expression matrices (Fig. 5D). The heat map showed that the black module (1404 genes) was significantly and strongly correlated with asthma (Fig. 5E). The black module was divided into 42 gene network clusters, including SNRPG-HNRNPD, GJA1-CGH2 and other gene relationship pairs (Fig. 5F).
**Figure 5.:** *Screening of the gene module by WGCNA. (A) Clustering information of GSE43696 samples. (B) Clustering and phenotypic information of GSE43696 samples, the upper part of the figure is clustering and the lower part is group. (C) Determination of the soft threshold. (D) Merging of the similar modules analyzed by the dynamic cutting tree algorithm (minModuleSize = 300). (E) The heat map of correlation between different modules and clinical characters. The ordinate represents different modules and the abscissa represents different groups. Each block represents the correlation coefficient and significance P value of a module and group. (F) The five top score clusters of genes/proteins related to asthma based on the black module. WGCNA = Weighted gene co-expression network analysis.*
## 3.6. Identification and evaluation of feature genes
A total of 88 candidate genes were obtained through venn analysis (Fig. 6A). 13 genes (NAV3, ITGA10, CSH1, SYT4, NOX1, SNTG2, LRP2, RNF182, UPK1B, LRRC31, POSTN, FAM155A, SHISA2) were screened by LASSO (Fig. 6B). A series of 51 genes were obtained after SVM algorithm (Fig. 6C). The 9 feature genes namely NAV3, ITGA10, SYT4, NOX1, SNTG2, RNF182, UPK1B, POSTN, and SHISA2 were yielded after the final screening (Fig. 6D). As seen from the ROC curve of 38 severe asthma samples and 20 healthy samples, the AUC of the 9 feature genes were greater than 0.7 in both GSE43696 dataset and GSE63142 validation set, indicating that the feature genes had a great ability to distinguish severe asthma samples and healthy samples. In the ROC curve of 50 mild to moderate asthma samples and 20 healthy samples, the AUC of 9 feature genes were all over 0.6 in the 2 datesets, suggesting that the feature genes had a good ability to differentiate moderate asthma samples and healthy samples (Fig. 6E). The box plot showed that the expression trends of the feature genes were consistent between asthma and control groups in both datasets (Fig. 6F).
**Figure 6.:** *Identification feature genes by machine learning method and feature genes evaluation. (A) Venn plot exhibiting the reliable gene features about asthma and ferroptosis among WGCNA (MEblack), DEGs, and two clusters’ DEGs. (B) LASSO feature selection method to narrowed down the 88 gene features. The left picture is convergence graph of 88 gene features. Each curve represents the coefficient change trajectory of each independent variable. The right picture is adjusting the different parameters to achieve the minimum binomial deviation of the model based on the penalty parameter that was determined by 10-fold cross validation following the 1-SE criterion, the best-performing feature set was selected. (C) SVM-RFE algorithm to narrowed down the 88 gene features. The pictures show the average accuracy/error as a function of the number of selected gene features. (D) Venn diagram of SVM_RFE and LASSO. (E) ROC curves of 9 feature genes related ferroptosis to distinguish between normal control and asthmatic subjects in GSE43696 and GSE63142. (F) Expression of 9 feature genes in normal control and asthmatic subjects. DEGs = differentially expressed genes, LASSO = least absolute shrinkage and selection operator, NC = normal control, ROC = receiver operating characteristic, RFE = recursive feature elimination, SVM = support vector machines, WGCNA = Weighted gene co-expression network analysis.*
## 3.7. Construction of a nomogram and functional annotation analysis of feature genes
9 feature genes (NAV3, ITGA10, SYT4, NOX1, SNTG2, RNF182, UPK1B, POSTN, and SHISA2), age and gender formed the nomogram, and the calibration curve indicated that the prediction ability of nomogram was excellent (Fig. 7A and B). NAV3 was involved in adaptive immune response, cytokine-mediated signaling pathway (GO) and epstein-barr virus infection, herpes simplex virus 1 infection (KEGG). ITGA10 enriched for GO including cytokine-mediated signaling pathway, G protein-coupled receptor activity, etc. and KEGG containing cytokine-cytokine receptor interaction and others. SYT4 enriched for significant GO and KEGG pathways such as cytoplasmic translation, large ribosomal subunit, NOD-like receptor signaling pathway. NOX1 was involved in GO pathways such as endoplasmic reticulum to Golgi vesicle-mediated transport, Golgi vesicle transport, and KEGG signaling pathways contained antigen processing and presentation, intestinal immune network for IgA production, etc. SNTG2 gene enrichment results showed an up-regulation trend for TOP5 GO entries covered chromosome segregation, etc, an up-regulation trend for TOP5 KEGG included proteasome, parkinson disease, huntington disease and a down-regulation trend contained coronavirus disease – COVID-19, herpes simplex virus 1 infection, etc. RNF182 participated in cytoplasmic translation, cytosolic ribosome (GO) and dopaminergic synapse (KEGG) etc. UPK1B enriched in GO and KEGG pathways concluding protein glycosylation (GO), mucin type o-glycan biosynthesis (KEGG) and others. The TOP5 GO entries and KEGG enrichment pathways of POSTN showed an up-regulation trend for cyplasmic translation (GO) and ribosome (KEGG). The GO and KEGG entries that SHISA2 was attended containing DNA-binding transcription activator activity (GO), amino sugar and nucleotide sugar metabolism (KEGG) (see Supplemental Figure S1, Supplemental Digital Content, http://links.lww.com/MD/I556, which demonstrates gene set enrichment analysis for 9 feature genes).
**Figure 7.:** *Construction of a nomogram and functional annotation analysis of nine feature genes. (A) Nomogram for predicting probabilities of asthma. For each subjects, eleven lines are drawn upward to determine the value received from the eleven predictors in the nomogram. The sum of these numbers is located on the “Total Points” axis. In addition, a line is drawn downward to determine the possibility of asthma. (B) The calibration curves for predicting the probability of asthma.*
## 3.8. Identification of DE-miRNA and DE-lncRNA and construction of a ceRNA regulatory network
There were 68 DE-miRNAs (42 up-regulated DE-miRNAs and 26 down-regulated DE-miRNAs) and 9 DE-lncRNAs (1 up-regulated lncRNA and 8 down-regulated lncRNAs) between asthma and control comparison groups (Fig. 8A–D). A total of 4 mRNAs, 5 miRNAs and 2 lncRNAs collectively formed a ceRNA regulatory network, and which included RNF182-hsa-miR-455-3p-LINC00319, SHISA2-hsa-miR-330-3p-LINC00261 and other mRNA-miRNA-lncRNA interactions pairs (Fig. 9A–D).
**Figure 8.:** *Identification of differentially expressed miRNA and differentially expressed lncRNA. (A) Volcano plot of differentially expressed miRNAs of asthma and normal control subjects. (B) Heat map of up and down differentially expressed miRNAs of asthma and normal control subjects. (C) Volcano plot of differentially expressed lncRNAs of asthma and normal control subjects. (D) Heat map of up and down differentially expressed lncRNAs of asthma and normal control subjects.* **Figure 9.:** *Construction of ceRNA regulatory network. (A) Venn diagram of miRWalk prediction results and down-regulated miRNAs. (B) Venn diagram of miRWalk prediction results and up-regulated miRNAs. (C) Venn diagram of StarBase prediction results and down-regulated lncRNAs. (D) ceRNA network for UPK1B, ITGA10, RNF182, SHISA2. Circles in the figure represent key genes (orange: up-regulated; golden yellow: down-regulated), diamonds represent miRNAs (red: up-regulated; green: down-regulated), triangles represent lncRNAs, and blue represent down-regulated lncRNAs. ceRNA = competing endogenous RNA.*
## 3.9. Construction of a mRNA-TF regulatory network
A sum of 58 TFs and 5 feature genes (RNF182, ITGA10, SNTG2, SHISA2, NAV3) constituted the mRNA-TF regulatory network which included SNTG2-EZH2, RNF182-FOXA2, and so on (Fig. 10A). The heat map showed that there were differences in the correlations between feature genes and TFs, for example, the SNTG2 was positively a correlated with EZH2, while NAV3 was negatively correlated with EZH2 (Fig. 10B).
**Figure 10.:** *Construction of mRNA-TF regulatory network. (A) mRNA-Tissue factors regulatory network. The red circles in the figure represent key genes, and the green represent tissue factors. (B) Heat map of correlation between key genes and tissue factors. The relationship pairs with asterisks in the figure represent P < .05, |cor|>0.3. TF = transcription factor.*
## 3.10. Drug sensitivity analysis
NAV3, ITGA10, SYT4, NOX1, SNTG2, RNF182, UPK1B, POSTN, and SHISA2 were predicted to 28, 14, 15, 60, 13, 10, 11, 41, and 23 drugs in order. The therapeutic drug network diagram included NAV3-bisphenol A, ITGA10-bisphenol A, RNF182-valproic acid and other relationship pairs (Fig. 11, and see Supplemental Table S1, Supplemental Digital Content, http://links.lww.com/MD/I557, which illustrates key genes and predicted potential therapeutics).
**Figure 11.:** *Therapeutic drug network. Network diagram of 9 feature genes and potential therapeutic drugs. Red circles represent 9 feature genes, green diamonds represent predicted therapeutic drugs.*
## 4. Discussion
This study obtained 438 DEGs in the training set. Then, 9 DE-FRGs were obtained after intersecting with the FerrDb. These 9 DE-FRGs were significantly enriched in BP such as lipoxygenase pathway, response to oxygen levels, and HIF-1 signaling pathway, according to GO and KEGG results. Interestingly, ferroptosis is caused by lipid peroxidation, and lipid metabolism is associated with the onset of asthma. For example, the 15-LO1/PEBP1 complex, which can produce hydroperoxyl-phospholipids, drive ferroptosis and enhance type 2 inflammation-related signaling pathways, is found in the airway epithelial cells of asthmatic patients.[28] In addition, the HIF-1 signaling pathway is the main regulatory pathway for oxygen homeostasis.[29] Under inflammatory conditions, HIF-1α in goblet cells of the airway epithelium can promote the expression of mucin 5AC, which interferes with the mucociliary clearance mechanisms of the airway.[30,31] Additionally, ferroptosis may cause renal tubular damage via the HIF-1α pathway in diabetic kidney disease.[32] Thus, ferroptosis may be involved in the functional changes of asthma airway epithelial cells through multiple ways and may contribute to the development of type 2 inflammation.
After that, we used the STRING database to evaluate the interactions among these 9 DE-FRGs. We obtained 6 DE-FRGs with significant interactions. To verify the ability of these 6 DE-FRGs to discriminate asthma and normal subjects, we performed ROC analysis. According to the AUC value, NOX1 had the best performance (AUC ≥ 0.82), followed by DDIT4, IL-6 and NOS2. NOX1 is an enzyme that produces ROS during ferroptosis and initiates lipid peroxidation. In mouse macrophages and the human lung cancer cell line Calu-1, increased NOX1 leads to lipid peroxidation and subsequent ferroptosis.[33,34] NOX1 has also been mentioned as a marker of moderate to severe asthma. It is involved in the ion transport regulatory pathway and oxidative stress in airway epithelial cells of asthma.[35–37] DDIT4 (also known as Rtp801) is a developmental glucocorticoid gene, and may be involved in the susceptibility to asthma and lung development.[38] DDIT4 can inhibit the mTOR signaling in mouse fibroblasts, it is necessary for amplifying oxidative stress caused by cigarette smoke.[39] IL-6 also can promote lipid peroxidation and induce ferroptosis, which can be reversed by Fer-1 treatment.[40] NOS2, a nitric oxide synthase gene, is strongly correlated with the levels of fractional exhaled nitric oxide which is a marker of airway inflammation in asthma.[41,42] *And ferroptosis* can induce the expression of NOS2 in macrophages, Th2 cytokines can induce the expression of NOS2 either.[43] In some studies,[44] MEF2C knockout enhanced the effect of ferroptosis inducer Erastin on meningioma cell ferroptosis and lipid peroxidation in vitro, and strengthened the meningioma growth inhibition mediated by ferroptosis in mouse models, this may be related to the downregulation of NF2 and E-cadherin. MEF2C is also proposed as a biomarker for asthma, which decreases with the increase of asthma severity.[45] The number increase of airway blood vessels is a feature of airway remodeling in asthma. It is shown that the levels of VEGFA in alveolar lavage fluid in asthmatic patients correlate with the number of blood vessels.[46] Further more, ferroptosis in endometrial stromal cells triggers the production of VEGFA and IL-8, thus promoting angiogenesis in adjacent lesions and accelerating disease progression.[47] These findings suggest that these 6 interactions genes may play a vital role in ferroptosis in asthma. Whether they can jointly promote the ferroptosis in airway epithelial cells, and how they interact, needs to be further studied.
Asthma is a heterogeneous disease.[48] To further analyze the function of ferroptosis in airway epithelial cells, we divided asthmatic patients into two subtypes according to 9 DE-FRGs. Through GSVA analysis, we found that there were obvious differences in biological functions between the two subtypes. And DEGs between the two subtypes were significantly enriched in the inflammatory signaling pathways, such as cytokine-cytokine receptor interaction. In IL-13 stimulated bronchial epithelial cell line (BEAS-2B), Fer-1 treatment reduced ferroptosis and oxidative damage, and inhibited the production of inflammatory cytokines.[49] In another study,[28] IL-13 induced the production of hydroperoxyl-phospholipids by 15-LO1, reduced intracellular GSH, and increased extracellular oxidative GSH. This redox imbalance makes airway epithelial cells more sensitive to ferroptosis. Cellular GSH is further reduced by inhibition of SLC7A11, promoting ferroptosis and expression of type 2 inflammatory cytokines like CCL26 and POSTN. Moreover, IL-13 induced mucin 5AC is regulated by the 15-LO1 pathway in human bronchial epithelial cells, indicating that inflammatory cytokines-induced mucus secretion is regulated by the ferroptosis-related pathway.[50] In sum, ferroptosis and inflammatory cytokines promote each other and interfere with the function of airway epithelial cells. In this study, we obtained consistent results, demonstrating that the inflammatory response is related to ferroptosis in the asthmatic airway epithelium. A variety of inflammatory cells and cytokines participate in asthma, which inflammatory cytokines (such as IL-13) could cause ferroptosis of airway epithelial cells and which ones do not, a further study is required.
To further narrow down the genes significantly associated with ferroptosis in airway epithelial cells in asthmatic patients, we intersected DEGs between asthma and control samples, inter-cluster DEGs and asthma-related module. We obtained 88 asthma-related ferroptosis associated genes. Then, LASSO and SVM-RFE were used to identify the critical ferroptosis associated genes involved in asthma. Finally, 9 feature genes (NAV3, ITGA10, SYT4, NOX1, SNTG2, RNF182, UPK1B, POSTN, and SHISA2) were identified as the essential genes. Next, we also analyzed the ability of these 9 feature genes to distinguish between asthma and normal subjects. Our results showed that these 9 feature genes could clearly distinguish between normal and mild to moderate asthma, as well as between normal and severe asthma in two datasets. As we can see, they had better diagnostic ability for severe asthma, this indicated that ferroptosis may be associated with asthma severity. Among these 9 feature genes, RNF182 had the best ability in distinguishing severe asthma, with AUC of 0.889 (GSE43696) and 0.849 (GSE63142), respectively. But NOX1 had the best ability to differentiate between mild and moderate asthma, with AUC of 0.798 (GSE43696) and 0.826 (GSE63142), respectively. This indicates that there are specific differences in airway epithelial cell physiological functions between mild to moderate asthmatic patients and severe asthmatic patients. Notably, NOX1 was screened from the ferroptosis database and our subsequent machine learning analysis, demonstrating its importance in ferroptosis of asthma airway epithelial cells. To evaluate the comprehensive diagnostic ability of these 9 feature genes for asthma, we performed nomogram analysis. The results showed that these 9 feature genes had good discrimination ability for asthma, with age, RNF182 and ITGA10 being the top three contributing factors. However, in addition to age and gender, more clinical indicators need to be added in the future.
After GSEA analysis, we found that POSTN-associated genes were involved in adaptive immune responses; NOX1-associated genes were involved in antigen processing and presentation; RNF182-associated genes were involved in viral infection and innate immunodeficiency; and, SYT4-associated genes were involved in the NOD-like receptor signaling pathway and TGF-β signaling pathway. They are all associated with immune response, may be crucial in the development of asthma. It has been shown that POSTN upregulation in vascular smooth muscle cells can increase cell sensitivity to ferroptosis by inhibiting SLC7A11 expression, suppressing P53, and reducing GSH synthesis.[51] POSTN is also thought to be involved in airway remodeling of asthma,[52] and IL4 and IL13 can promote the expression of POSTN in airway epithelial cells. Furthermore, POSTN is considered as a marker of Th2-high asthma.[53] RNF182 increased expression in hepatocellular carcinoma can mediate p65 ubiquitination, thus accelerating the degradation of p65 protein, blocking the binding of p65 to the SLC7A11 promoter, and promoting ferroptosis.[54] In the LNCaP prostate cancer cell line, the ferroptosis inducer Erastin can increase SYT4 expression. In contrast, the ferroptosis inhibitor Fer-1 can reduce SYT4 expression, suggesting that SYT4 is associated with ferroptosis.[55] However, the mechanisms underlying the role of RNF182 and SYT4 in ferroptosis of asthmatic airway epithelial cells have not been reported. UPK1B is significantly associated with DNA methylation sites in whole blood of asthma patients, but the biological mechanism is unclear.[56] SHISA2, SNTG2, NAV3, and ITGA10 are new genes that have not been previously reported in ferroptosis and asthma, and their role in ferroptosis of asthmatic airway epithelial cells remains to be explored. In summary, there are few studies on the ferroptosis mechanism of these 9 feature genes in asthmatic airway epithelial cells. Fortunately, we found miRNAs, lncRNAs, transcription factors, and therapeutic drug associated with these genes in the network analysis, which may provide reference for mechanistic study of ferroptosis.
## 5. Conclusion
This study firstly found that the ferroptosis genes could distinguish asthma patients from normal subjects. Secondly, through the methods of WGCNA and machine learning, 9 feature genes (NAV3, ITGA10, SYT4, NOX1, SNTG2, RNF182, UPK1B, POSTN, SHISA2) related to ferroptosis were identified. The diagnostic ability of these 9 feature genes for asthma was verified. The regulatory network and intervention drugs were analyzed. However, the causal relationship between ferroptosis and these 9 feature genes could not be determined, which is one of the limitations of this study. In the future, we will continue to investigate the specific molecular mechanisms of these 9 feature genes in ferroptosis of airway epithelial cells, so as to provide theoretical basis for understanding asthma and ferroptosis of airway epithelial cells.
## Acknowledgments
We would like to thank TopEdit (www.topeditsci.com) for its linguistic assistance during the preparation of this manuscript.
## Author contributions
Conceptualization: Ye Zheng, Xiaofeng Jiang.
Data curation: Ye Zheng.
Formal analysis: Ye Zheng.
Investigation: Ye Zheng.
Methodology: Ye Zheng.
Project administration: Ye Zheng.
Resources: Ye Zheng.
Software: Ye Zheng.
Supervision: Jingyao Fan, Xiaofeng Jiang.
Validation: Ye Zheng, Jingyao Fan.
Visualization: Ye Zheng.
Writing – original draft: Ye Zheng.
Writing – review & editing: Jingyao Fan, Xiaofeng Jiang.
## References
1. **The Global Asthma Report 2022.**. *Int J Tuberc Lung Dis* (2022) **26** 1-104
2. Eder W, Ege MJ, von Mutius E. **The asthma epidemic.**. *N Engl J Med* (2006) **355** 2226-35. PMID: 17124020
3. Taylor D, Bateman E, Boulet L. **A new perspective on concepts of asthma severity and control.**. *Eur Respir J* (2008) **32** 545-54. PMID: 18757695
4. Lambrecht BN, Hammad H. **The airway epithelium in asthma.**. *Nat Med* (2012) **18** 684-92. PMID: 22561832
5. Kitajima M, Lee HC, Nakayama T. **TSLP enhances the function of helper type 2 cells.**. *Eur J Immunol* (2011) **41** 1862-71. PMID: 21484783
6. Savinko T, Matikainen S, Saarialho-Kere U. **IL-33 and ST2 in atopic dermatitis: expression profiles and modulation by triggering factors.**. *J Invest Dermatol* (2012) **132** 1392-400. PMID: 22277940
7. Shin H-W, Kim D-K, Park M-H. **IL-25 as a novel therapeutic target in nasal polyps of patients with chronic rhinosinusitis.**. *J Allergy Clin Immunol* (2015) **135** 1476-85.e7. PMID: 25725991
8. Ziegler SF, Artis D. **Sensing the outside world: TSLP regulates barrier immunity.**. *Nat Immunol* (2010) **11** 289-93. PMID: 20300138
9. Oei E, Kalb T, Beuria P. **Accessory cell function of airway epithelial cells.**. *Am J Physiol Lung Cell Mol Physiol* (2004) **287** L318-31. PMID: 15246982
10. Salik E, Tyorkin M, Mohan S. **Antigen trafficking and accessory cell function in respiratory epithelial cells.**. *Am J Respir Cell Mol Biol* (1999) **21** 365-79. PMID: 10460754
11. Dixon SJ, Lemberg KM, Lamprecht MR. **Ferroptosis: an iron-dependent form of nonapoptotic cell death.**. *Cell* (2012) **149** 1060-72. PMID: 22632970
12. Doll S, Proneth B, Tyurina YY. **ACSL4 dictates ferroptosis sensitivity by shaping cellular lipid composition.**. *Nat Chem Biol* (2017) **13** 91-8. PMID: 27842070
13. Stockwell BR, Angeli JPF, Bayir H. **Ferroptosis: a regulated cell death nexus linking metabolism, redox biology, and disease.**. *Cell* (2017) **171** 273-85. PMID: 28985560
14. Friedmann Angeli JP, Schneider M, Proneth B. **Inactivation of the ferroptosis regulator Gpx4 triggers acute renal failure in mice.**. *Nat Cell Biol* (2014) **16** 1180-91. PMID: 25402683
15. Tang W, Dong M, Teng F. **Environmental allergens house dust mite-induced asthma is associated with ferroptosis in the lungs.**. *Exp Ther Med* (2021) **22** 1-10
16. Zeng Z, Huang H, Zhang J. **HDM induce airway epithelial cell ferroptosis and promote inflammation by activating ferritinophagy in asthma.**. *FASEB J* (2022) **36** e22359. PMID: 35621121
17. Wang Y, Wang Z, Sun J. **Identification of HCC subtypes with different prognosis and metabolic patterns based on mitophagy.**. *Front Cell Dev Biol* (2021) **9** 16
18. Zhang M-Y, Huo C, Liu J-Y. **Identification of a five autophagy subtype-related gene expression pattern for improving the prognosis of lung adenocarcinoma.**. *Front Cell Dev Biol* (2021) **9** 20
19. Yu G, Wang L-G, Han Y. **clusterProfiler: an R package for comparing biological themes among gene clusters.**. *OMICS* (2012) **16** 284-7. PMID: 22455463
20. Min SH, Zhou J. **smplot: an R package for easy and elegant data visualization.**. *Front Genet* (2021) **12** 10
21. Robin X, Turck N, Hainard A. **pROC: an open-source package for R and S+ to analyze and compare ROC curves.**. *BMC Bioinf* (2011) **12** 1-8
22. Dong C, Dang D, Zhao X. **Integrative characterization of the role of IL27 in melanoma using bioinformatics analysis.**. *Front Immunol* (2021) **4253** 14
23. Liang L, Yu J, Li J. **Integration of scRNA-Seq and bulk RNA-Seq to analyse the heterogeneity of ovarian cancer immune cells and establish a molecular risk model.**. *Front Oncol* (2021) **3734** 13
24. Langfelder P, Horvath S. **WGCNA: an R package for weighted correlation network analysis.**. *BMC Bioinf* (2008) **9** 1-13
25. Zhang M, Zhu K, Pu H. **An immune-related signature predicts survival in patients with lung adenocarcinoma.**. *Front Oncol* (2019) **9** 1314. PMID: 31921619
26. Liu T-T, Li R, Huo C. **Identification of CDK2-related immune forecast model and ceRNA in lung adenocarcinoma, a pan-cancer analysis.**. *Front Cell Dev Biol* (2021) **9** 23
27. Liu S, Wang Z, Zhu R. **Three differential expression analysis methods for RNA sequencing: limma, EdgeR, DESeq2.**. *J Vis Exp* (2021) e62528
28. Nagasaki T, Schuyler AJ, Zhao J. **15LO1 dictates glutathione redox changes in asthmatic airway epithelium to worsen type 2 inflammation.**. *J Clin Invest* (2022) **132** e151685. PMID: 34762602
29. Vanderhaeghen T, Beyaert R, Libert C. **Bidirectional crosstalk between hypoxia inducible factors and glucocorticoid signalling in health and disease.**. *Front Immunol* (2021) **12** 684085. PMID: 34149725
30. Wu S, Li H, Yu L. **IL-1β upregulates Muc5ac expression via NF-κB-induced HIF-1α in asthma.**. *Immunol Lett* (2017) **192** 20-6. PMID: 29031476
31. Ridley C, Thornton DJ. **Mucins: the frontline defence of the lung.**. *Biochem Soc Trans* (2018) **46** 1099-106. PMID: 30154090
32. Feng X, Wang S, Sun Z. **Ferroptosis enhanced diabetic renal tubular injury via HIF-1α/HO-1 pathway in db/db mice.**. *Front Endocrinol (Lausanne)* (2021) **12** 626390. PMID: 33679620
33. Yun MR, Park HM, Seo KW. **5-Lipoxygenase plays an essential role in 4-HNE-enhanced ROS production in murine macrophages via activation of NADPH oxidase.**. *Free Radic Res* (2010) **44** 742-50. PMID: 20370567
34. Chen X, Huang J, Yu C. **A noncanonical function of EIF4E limits ALDH1B1 activity and increases susceptibility to ferroptosis.**. *Nat Commun* (2022) **13** 1-16. PMID: 34983933
35. Zayed H. **Novel comprehensive bioinformatics approaches to determine the molecular genetic susceptibility profile of moderate and severe asthma.**. *Int J Mol Sci* (2020) **21** 4022. PMID: 32512817
36. Sethi GS, Dharwal V, Naura AS. **Immunological basis of Oxidative stress-induced lung inflammation in asthma and COPD.**. *Oxidative Stress in Lung Diseases* (2019) 195-223
37. Ho WE, Cheng C, Peh HY. **Anti-malarial drug artesunate ameliorates oxidative lung damage in experimental allergic asthma.**. *Free Radic Biol Med* (2012) **53** 498-507. PMID: 22634146
38. Sharma S, Kho AT, Chhabra D. **Glucocorticoid genes and the developmental origins of asthma susceptibility and treatment response.**. *Am J Respir Cell Mol Biol* (2015) **52** 543-53. PMID: 25192440
39. Hernández-Saavedra D, Sanders L, Perez MJ. **RTP801 amplifies nicotinamide adenine dinucleotide phosphate oxidase-4–dependent oxidative stress induced by cigarette smoke.**. *Am J Respir Cell Mol Biol* (2017) **56** 62-73. PMID: 27556956
40. Han F, Li S, Yang Y. **Interleukin-6 promotes ferroptosis in bronchial epithelial cells by inducing reactive oxygen species-dependent lipid peroxidation and disrupting iron homeostasis.**. *Bioengineered* (2021) **12** 5279-88. PMID: 34402724
41. Modena BD, Tedrow JR, Milosevic J. **Gene expression in relation to exhaled nitric oxide identifies novel asthma phenotypes with unique biomolecular pathways.**. *Am J Respir Crit Care Med* (2014) **190** 1363-72. PMID: 25338189
42. Yamada M, Motoike IN, Kojima K. **Genetic loci for lung function in Japanese adults with adjustment for exhaled nitric oxide levels as airway inflammation indicator.**. *Commun Biol* (2021) **4** 1-15. PMID: 33398033
43. Hirai K, Shirai T, Suzuki M. **Association between (CCTTT) n repeat polymorphism in NOS2 promoter and asthma exacerbations.**. *J Allergy Clin Immunol* (2018) **142** 663-665.e3. PMID: 29518423
44. Bao Z, Hua L, Ye Y. **MEF2C silencing downregulates NF2 and E-cadherin and enhances Erastin-induced ferroptosis in meningioma.**. *Neuro Oncol* (2021) **23** 2014-27. PMID: 33984142
45. Sheu C-C, Tsai M-J, Chen F-W. **Identification of novel genetic regulations associated with airway epithelial homeostasis using next-generation sequencing data and bioinformatics approaches.**. *Oncotarget* (2017) **8** 82674-82688. PMID: 29137293
46. Feltis BN, Wignarajah D, Zheng L. **Increased vascular endothelial growth factor and receptors: relationship to angiogenesis in asthma.**. *Am J Respir Crit Care Med* (2006) **173** 1201-7. PMID: 16528018
47. Li G, Lin Y, Zhang Y. **Endometrial stromal cell ferroptosis promotes angiogenesis in endometriosis.**. *Cell Death Discovery* (2022) **8** 1-12. PMID: 35013145
48. Kaur R, Chupp G. **Phenotypes and endotypes of adult asthma: moving toward precision medicine.**. *J Allergy Clin Immunol* (2019) **144** 1-12. PMID: 31277742
49. Yang N, Shang Y. **Ferrostatin-1 and 3-Methyladenine ameliorate ferroptosis in OVA-induced asthma model and in IL-13-challenged BEAS-2B cells.**. *Oxid Med Cell Longev* (2022) **2022** 16
50. Zhao J, Maskrey B, Balzar S. **Interleukin-13–induced MUC5AC is regulated by 15-lipoxygenase 1 pathway in human bronchial epithelial cells.**. *Am J Respir Crit Care Med* (2009) **179** 782-90. PMID: 19218191
51. Ma W-Q, Sun X-J, Zhu Y. **Metformin attenuates hyperlipidaemia-associated vascular calcification through anti-ferroptotic effects.**. *Free Radic Biol Med* (2021) **165** 229-42. PMID: 33513420
52. Hur GY, Broide DH. **Genes and pathways regulating decline in lung function and airway remodeling in asthma.**. *Allergy Asthma Immunol Res* (2019) **11** 604-21. PMID: 31332973
53. Sonnenberg-Riethmacher E, Miehe M, Riethmacher D. **Periostin in allergy and inflammation.**. *Front Immunol* (2021) **12** 11
54. Liu Y, Ouyang L, Mao C. **PCDHB14 promotes ferroptosis and is a novel tumor suppressor in hepatocellular carcinoma.**. *Oncogene* (2022) **41** 3570-83. PMID: 35688944
55. Wo Q, Liu Z, Hu L. **Identification of ferroptosis-associated genes in prostate cancer by bioinformatics analysis.**. *Front Genet* (2022) **13** 12
56. Kogan V, Millstein J, London SJ. **Genetic-epigenetic interactions in asthma revealed by a genome-wide gene-centric search.**. *Hum Hered* (2018) **83** 130-52. PMID: 30669148
|
---
title: 'Diagnostic value of magnetic resonance imaging and magnetic resonance arthrography
for assessing acetabular labral tears: A systematic review and meta-analysis'
authors:
- Zhihao Huang
- Wenyu Liu
- Tianyu Li
- Zhihao Liu
- Pengfei Zhao
journal: Medicine
year: 2023
pmcid: PMC9981430
doi: 10.1097/MD.0000000000032963
license: CC BY 4.0
---
# Diagnostic value of magnetic resonance imaging and magnetic resonance arthrography for assessing acetabular labral tears: A systematic review and meta-analysis
## Background:
This study aimed to systematically evaluate the value of magnetic resonance imaging (MRI) and magnetic resonance arthrography (MRA) in the diagnosis of acetabular labral tears.
### Methods:
Databases including PubMed, Embase, Cochrane Library, Web of Science, CBM, CNKI, WanFang Data, and VIP were electronically searched to collect relevant studies on magnetic resonance in the diagnosis of acetabular labral tears from inception to September 1, 2021. Two reviewers independently screened the literature, extracted data, and assessed the risk of bias in the included studies by using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. RevMan 5.3, Meta Disc 1.4, and Stata SE 15.0 were used to investigate the diagnostic value of magnetic resonance in patients with acetabular labral tears.
### Results:
A total of 29 articles were included, involving 1385 participants and 1367 hips. The results of the meta-analysis showed that the pooled sensitivity, pooled specificity, pooled positive likelihood ratio, pooled negative likelihood ratio, pooled diagnostic odds ratio, area under the curve of the summary receiver operating characteristic, and Q* of MRI for diagnosing acetabular labral tears were 0.77 ($95\%$ confidence interval [CI], 0.75–0.80), 0.74 ($95\%$ CI, 0.68–0.80), 2.19 ($95\%$ CI, 1.76–2.73), 0.48 ($95\%$ CI, 0.36–0.65), 4.86 ($95\%$ CI, 3.44–6.86), 0.75, and 0.69, respectively. The pooled sensitivity, pooled specificity, pooled positive likelihood ratio, pooled negative likelihood ratio, pooled diagnostic odds ratio, area under the curve of the summary receiver operating characteristic, and Q* of MRA for diagnosing acetabular labral tears were 0.87 ($95\%$ CI, 0.84–0.89), 0.64 ($95\%$ CI, 0.57–0.71), 2.23 ($95\%$ CI, 1.57–3.16), 0.21 ($95\%$ CI, 0.16–0.27), 10.47 ($95\%$ CI, 7.09–15.48), 0.89, and 0.82, respectively.
### Conclusion:
MRI has high diagnostic efficacy for acetabular labral tears, and MRA has even higher diagnostic efficacy. Due to the limited quality and quantity of the included studies, the above results should be further validated.
## 1. Introduction
The acetabular labrum is a fibrocartilaginous ring attached to the edge of the acetabulum. It plays an important physiological role in ensuring wider coverage of the femoral head,[1] reducing femoroacetabular joint contact pressure,[2] and increasing the stability of the hip joint.[3] The acetabular labrum increases the articular surface area by $22\%$ and acetabular volume by $33\%$ and is believed to create a seal in the hip joint.[4] However, acetabular labral tear (ALT) destroys its physiological function, resulting in clinical symptoms such as hip pain and limited movement.[5] ALT was first recognized as a pathological entity in 1957 when a bucket handle labral tear was discovered after an attempted reduction of a posterior hip dislocation.[6] ALT can be associated with a variety of pathological conditions of the hip,[7] and it is one of the most common causes of hip joint pain.[8] It has been shown that hip and groin pain is caused by a labral tear in about $22\%$ to $55\%$ of patients.[6] If not diagnosed and treated in time, the range of ALT increases and causes trauma,[9] classic hip dysplasia,[10,11] Legg–Calve–Perthes disease,[12] and hip osteoarthritis.[13] At present, the diagnostic methods of ALT mainly include magnetic resonance imaging (MRI), magnetic resonance arthrography (MRA), and arthroscopy. Arthroscopy is an invasive examination, which has the disadvantages of possible complications and high examination costs. Arthroscopy is generally carried out in the operation.[14] Therefore, MR has become the first choice for the diagnosis of ALT. Because it is difficult to directly display the acetabular labral with computed tomography and X-ray, the detection rate of ALT was not high before MR examination is widely applied, and many patients were delayed the optimal treatment time due to lack of timely and correct diagnosis. With the widespread use of MR, the sensitivity (Sen) and specificity (Spe) of the diagnosis of ALT have been significantly improved. Many original studies have explored the value of MR in the diagnosis of ALT, but most of them were single diagnostic tests. In this study, a meta-analysis was conducted to comprehensively evaluate the value of MR in diagnosing ALT, to provide a basis for clinical diagnosis and scientific decision-making.
## 2. Methods
We adhered to the Preferred Reporting in Systematic Reviews and Meta-Analysis 2020 guidelines,[15] and this review was registered in International Prospective Register of Systematic Reviews (registration number is CRD42021281868).
## 2.1. Eligibility criteria
The inclusion criteria were as follows: participants with suspected ALT who underwent MR before arthroscopy or surgery (not limited by age, race, and nationality); prospective or retrospective study design; direct or indirect availability of the results—true positive, false positive, false negative, and true negative.
The exclusion criteria were as follows: duplicate articles; articles with inconsistent research contents; non-English and non-Chinese articles; conference abstracts; case reports; and animal test.
Effect sizes included the pooled Sen, Spe, positive likelihood ratio (+LR), negative likelihood ratio (–LR), diagnosis odds ratio (DOR), summary receiver operating characteristics, area under the curve (AUC) of the summary receiver operating characteristic, and Q*.
## 2.2. Search strategy
A literature search was carried out by 2 independent reviewers. PubMed, Embase, The Cochrane Library, Web of Science, CBM, CNKI, WanFang Data, and VIP were explored from inception date to September 1, 2021.
## 2.3. Study selection and data extraction
Literature screening and data extraction were carried out independently by 2 reviewers. Different opinions were solved through discussion. Excel 2021 was used to extract data, mainly recording the first author, publication time, national research type, magnetic field intensity and examination method of MR, reference standard, age, gender, number of hips, and 4-fold data (true positive, false positive, false negative, and true negative).
## 2.4. Risk of bias assessment of the included studies
Two reviewers used the Quality Assessment of Diagnostic Accuracy Studies-2 tool to independently assess the risk of bias in the included studies.[16] Each item was rated as “yes” (low bias or good applicability), “no” (high bias or poor applicability), or “unclear” (lack of relevant information or uncertain bias).
## 2.5. Statistical analysis
Review Management version 5.3 was used to assess the risk of bias in the included studies. Meta disc version 1.4 and Stata SE version 15.0 were used for meta-analysis. The correlation coefficient of Sen logarithm and (1 − Spe) logarithm was used to analyze whether there was a threshold effect. If the P value of the Spearman correlation coefficient was less than.05, it indicated that there was no threshold effect; otherwise, it indicated that there was a threshold effect. The heterogeneity of the meta-analysis results was tested by χ2 and I2. χ2 statistic with $P \leq .1$ or I2 > $50\%$ indicated significant heterogeneity among the studies,[17] which needed to be pooled by a random-effect model if the significant heterogeneity was not solved by meta-regression or subgroup analysis. Otherwise, a fixed-effect model was adopted. Sensitivity analysis was carried out by excluding the included studies one by one.[18] The publication bias was detected by Deek funnel plot.[19]
## 3.1. Literature search
A total of 622 articles were identified by searching the databases. Additional 44 articles were identified during the screening of the reference sections of the included articles. The detailed information is shown in Method S1, Supplemental Digital Content, http://links.lww.com/MD/I463. After screening layer by layer, 29 articles were finally included.[20–48] The process and the results of literature screening are shown in Figure 1.
**Figure 1.:** *Flow diagram of the literature search and selection processes.*
## 3.2. Detailed information and risk of bias results
The detailed information of the included studies is shown in Table 1. The risk of biased results of the included studies is shown in Figures 2 and 3 and Table S1, Supplemental Digital Content, http://links.lww.com/MD/I464.
## 3.3. Meta-analysis of MRI
A total of 14 articles with 20 studies on 1246 hips were included (Table 2).
**Table 2**
| Study name | Country | Magnetic field intensity | Hips | TP | FP | FN | TN |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Byrd et al 2004 | USA | 1.5T | 40 | 8 | 4 | 24 | 4 |
| Crespo-Rodríguez et al 2017 | Spain | 3.0T | 50 | 42 | 0 | 1 | 7 |
| Czerny et al 1996 | Austria | 0.5T and 1.0T | 22 | 6 | 0 | 14 | 2 |
| Edwards et al 1995 | England | 1.5T | 23 | 0 | 1 | 1 | 21 |
| Gao et al 2019 | China | 3.0T | 195 | 156 | 4 | 28 | 7 |
| Linda et al 2016 | Canada | 3.0T | 42 | 40 | 1 | 0 | 1 |
| Magee 2015 | USA | 3.0T | 43 | 38 | 1 | 4 | 0 |
| Magee 2015 | USA | 3.0T | 43 | 37 | 1 | 5 | 0 |
| Mintz et al 2005 | USA | 1.5T | 92 | 86 | 2 | 3 | 1 |
| Mintz et al 2005 | USA | 1.5T | 92 | 85 | 2 | 4 | 1 |
| Sundberg et al 2006 | USA | 3.0T | 8 | 5 | 2 | 0 | 1 |
| Sutter et al 2014 | Switzerland | 1.5T | 28 | 20 | 1 | 6 | 1 |
| Sutter et al 2014 | Switzerland | 1.5T | 28 | 23 | 1 | 3 | 1 |
| Tian et al 2014 | China | 3.0T | 90 | 36 | 7 | 23 | 24 |
| Tian et al 2014 | China | 3.0T | 90 | 39 | 8 | 20 | 23 |
| Tian et al 2016 | China | 3.0T | 122 | 53 | 9 | 34 | 26 |
| Tian et al 2016 | China | 3.0T | 122 | 56 | 9 | 31 | 26 |
| Toomayan et al 2006 | USA | 1.5T | 51 | 1 | 0 | 3 | 3 |
| Toomayan et al 2006 | USA | 1.5T | 51 | 1 | 0 | 11 | 2 |
| Zlatkin et al 2010 | USA | 1.5T | 14 | 11 | 0 | 2 | 1 |
## 3.4. Heterogeneity test
Spearman correlation coefficient of Sen logarithm and (1 − Spe) logarithm was 0.570 ($$P \leq .009$$), indicating that there was a threshold effect in this study. The I2 of Sen and −LR was greater than $50\%$, and the effect sizes were pooled by the random-effect model. The I2 of Spe, +LR, and DOR was less than $50\%$, and the effect sizes were pooled by the fixed-effect model (Fig. 4).
**Figure 4.:** *Forest plot of MRI for the diagnosis of ALT. Note: The subgraph of (A–F) refer to Sen, Spe, +LR, −LR, DOR, AUC, and Q*, respectively. ALT = acetabular labral tears, AUC = area under the curve, DOR = diagnosis odds ratio, −LR = negative likelihood ratio, +LR = positive likelihood ratio, MRI = magnetic resonance imaging, Sen = sensitivity, Spe = specificity.*
## 3.5. Pooled effect sizes
The pooled effects sizes were as follows: Sen(pooled) = 0.77 ($95\%$ confidence interval [CI], 0.75–0.80), Spe(pooled) = 0.74 ($95\%$ CI, 0.68–0.80), +LR(pooled) = 2.19 ($95\%$ CI, 1.76–2.73), −LR(pooled) = 0.48 ($95\%$ CI, 0.36–0.65), DOR(pooled) = 4.86($95\%$ CI, 3.44–6.86), AUC = 0.75, and Q* = 0.69 (Fig. 4).
## 3.6. Meta-regression analysis
According to the study time, study country, and MRI magnetic field intensity, a meta-regression analysis was carried out. The results showed that study time was the main source of heterogeneity ($P \leq .05$).
## 3.7. Subgroup analysis
The variable with statistical significance in the meta-regression (study time) was analyzed in subgroups. The results were as follows:
## 3.8. Meta-analysis of subgroup(Year: 1995–2009)
Six articles with 8 studies on 379 hips were included (Table 2).
## 3.9. Heterogeneity test
Spearman correlation coefficient of Sen logarithm and (1 − Spe) logarithm was 0.739 ($$P \leq .036$$), indicating that there was a threshold effect in this subgroup study. The I2 of Sen and Spe was greater than $50\%$, and the effect sizes were pooled by the random-effect model. The I2 of +LR, −LR, and DOR was less than $50\%$, and the effect sizes were pooled by the fixed-effect model (Table 3).
**Table 3**
| Group | n | Sen(pooled) | Sen(pooled).1 | Spe(pooled) | Spe(pooled).1 | +LR(pooled) | +LR(pooled).1 | −LR(pooled) | −LR(pooled).1 | DOR(pooled) | DOR(pooled).1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Group | n | I 2 | P value | I 2 | P value | I 2 | P value | I 2 | P value | I 2 | P value |
| Year | Year | Year | Year | Year | Year | Year | Year | Year | Year | Year | Year |
| 1995–2009 | 8 | 95.2% | .000 | 64.9% | .006 | 0.0% | .629 | 42.5% | .095 | 28.9% | .198 |
| 2010–2019 | 12 | 86.9% | .000 | 12.2% | .325 | 0.0% | .589 | 47.7% | .033 | 0.0% | .487 |
## 3.10. Pooled effect sizes
The pooled effect sizes were as follows: Sen(pooled) = 0.76 ($95\%$ CI, 0.70–0.81), Spe(pooled) = 0.76 ($95\%$ CI, 0.61–0.87), +LR(pooled) = 1.18 ($95\%$ CI, 0.79–1.77), −LR(pooled) = 0.87 ($95\%$ CI, 0.65–1.16), DOR(pooled) = 1.49 ($95\%$ CI, 0.63–3.56), AUC = 0.64, and Q* = 0.61 (Table 4).
**Table 4**
| Subgroup | n | Sen(pooled) (95% CI) | Spe(pooled) (95% CI) | +LR(pooled) (95% CI) | −LR(pooled) (95% CI) | DOR(pooled) (95% CI) | AUC | Q* |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Year | Year | Year | Year | Year | Year | Year | Year | Year |
| 1995–2009 | 8 | 0.76 (0.70–0.81) | 0.76 (0.61–0.87) | 1.18 (0.79–1.77) | 0.87 (0.65–1.16) | 1.49 (0.63–3.56) | 0.64 | 0.61 |
| 2010–2019 | 12 | 0.78 (0.75–0.81) | 0.74 (0.66–0.80) | 2.49 (1.93–3.21) | 0.42 (0.36–0.50) | 6.12 (4.22–8.88) | 0.77 | 0.71 |
## 3.11. Meta-analysis of subgroup(Years: 2010–2019)
Eight articles with 12 studies on 867 hips were included (Table 2).
## 3.12. Heterogeneity test
Spearman correlation coefficient of Sen logarithm and (1 − Spe) logarithm was 0.410 ($$P \leq .186$$), indicating that there was no threshold effect in this subgroup study. The I2 of Sen was greater than $50\%$, and the effect size was pooled by the random-effect model. The I2 of Spe, +LR, −LR, and DOR was less than $50\%$, and the effect sizes were pooled by the fixed-effect model (Table 3).
## 3.13. Pooled effect sizes
The pooled effect sizes were as follows: Sen(pooled) = 0.78 ($95\%$ CI, 0.75–0.81), Spe(pooled) = 0.74 ($95\%$ CI, 0.66–0.80), +LR(pooled) = 2.49 ($95\%$ CI, 1.93–3.21), −LR(pooled) = 0.42 ($95\%$ CI, 0.36–0.50), DOR(pooled) = 6.12 ($95\%$ CI, 4.22–8.88), AUC = 0.77, and Q* = 0.71 (Table 4).
## 3.14. Sen analysis
After excluding individual studies one by one, the remaining studies were pooled and analyzed again. The results showed that each study eliminated had little impact on the amount of pooled effect sizes, indicating that the results of this study were relatively stable and the reliability of the analysis results was high (Fig. 5).
**Figure 5.:** *The sensitivity analysis of MRI. MRI = magnetic resonance imaging.*
## 3.15. Publication bias analysis
Taking the inverse of the square root of effective sample size [1/root (ESS)] as the ordinate and DOR as the abscissa, the results of Deeks test showed that the P value of slope coefficient was 0.89, suggesting that there was no publication bias in the MRI examination method (Fig. 6).
**Figure 6.:** *Funnel plot of MRI for the diagnosis of ALT. ALT = acetabular labral tears, MRI = magnetic resonance imaging.*
## 3.16. Meta-analysis of MRA
A total of 24 articles with 27 studies on 942 hips were included (Table 5).
**Table 5**
| Study name | Country | Magnetic field intensity | Hips | TP | FP | FN | TN |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Aprato et al 2013 | Italy | 1.5T | 41 | 31 | 1 | 3 | 6 |
| Banks et al 2012 | UK | 1.5T | 69 | 13 | 26 | 3 | 27 |
| Byrd et al 2004 | USA | 1.5T | 40 | 23 | 7 | 9 | 1 |
| Chan et al 2005 | Taiwan | 1.5T | 17 | 16 | 1 | 0 | 0 |
| Crespo-Rodríguez et al 2017 | Spain | 1.5T | 50 | 43 | 1 | 0 | 6 |
| Czerny et al 1996 | Austria | 0.5T and 1.0T | 22 | 18 | 0 | 2 | 2 |
| Czerny et al 1999 | Austria | 0.5T and 1.0T | 40 | 30 | 2 | 3 | 5 |
| El-Liethy et al 2019 | Egypt | 1.5T | 31 | 21 | 2 | 3 | 5 |
| Freedman et al 2006 | USA | 1.5T | 24 | 22 | 1 | 1 | 0 |
| Hong et al 2010 | China | 1.5T | 14 | 13 | 0 | 0 | 1 |
| Jin et al 2012 | South Korea | 3.0T | 16 | 10 | 1 | 1 | 4 |
| Keeney et al 2004 | USA | 1.5T | 102 | 66 | 5 | 27 | 4 |
| Leunig et al 1997 | Switzerland | 1.5T | 23 | 10 | 2 | 6 | 5 |
| Magee 2015 | USA | 3.0T | 43 | 39 | 1 | 3 | 0 |
| Magee 2015 | USA | 3.0T | 43 | 38 | 1 | 4 | 0 |
| McCarthy et al 2013 | USA | 1.5T | 70 | 49 | 3 | 11 | 7 |
| Nishii et al 1996 | Japan | 1.5T | 19 | 9 | 0 | 2 | 8 |
| Petersilge et al 1996 | USA | 1.5T | 10 | 8 | 0 | 0 | 1 |
| Sahin et al 2014 | Turkey | 1.5T | 14 | 10 | 2 | 0 | 2 |
| Studler et al 2003 | Switzerland | 1.5T | 57 | 43 | 6 | 1 | 7 |
| Sundberg et al 2006 | USA | 1.5T | 8 | 4 | 2 | 1 | 1 |
| Sutter et al 2014 | Switzerland | 1.5T | 28 | 22 | 0 | 4 | 2 |
| Sutter et al 2014 | Switzerland | 1.5T | 28 | 23 | 1 | 3 | 1 |
| Tian et al 2014 | China | 3.0T | 34 | 19 | 2 | 2 | 11 |
| Tian et al 2014 | China | 3.0T | 34 | 20 | 2 | 1 | 11 |
| Toomayan et al 2006 | USA | 1.5T | 51 | 22 | 0 | 2 | 6 |
| Zlatkin et al 2010 | USA | 1.5T | 14 | 13 | 1 | 0 | 0 |
## 3.17. Heterogeneity test
Spearman correlation coefficient of Sen logarithm and (1 − Spe) logarithm was − 0.153 ($$P \leq .465$$), indicating that there was no threshold effect in this study. The I2 of Sen, Spe, and +LR was greater than $50\%$, and the effect sizes were pooled by the random-effect model. The I2 of −LR and DOR was less than $50\%$, and the effect sizes were pooled by the fixed-effect model (Fig. 7).
**Figure 7.:** *Forest plot of MRA for the diagnosis of ALT. Note: The subgraph of (A–F) refer to Sen, Spe, +LR, −LR, DOR, AUC, and Q*, respectively. ALT = acetabular labral tears, AUC = area under the curve, DOR = diagnosis odds ratio, −LR = negative likelihood ratio, +LR = positive likelihood ratio, MRA = magnetic resonance arthrography, Sen = sensitivity, Spe = specificity.*
## 3.18. Pooled effect sizes
The pooled effect sizes were as follows: Sen(pooled) = 0.87 ($95\%$ CI, 0.84–0.89), Spe(pooled) = 0.64 ($95\%$ CI, 0.57–0.71), +LR(pooled) = 2.23 ($95\%$ CI, 1.57–3.16), −LR(pooled) = 0.21 ($95\%$ CI, 0.16–0.27), DOR(pooled) = 10.47 ($95\%$ CI, 7.09–15.48), AUC = 0.89, and Q* = 0.82 (Fig. 7).
## 3.19. Meta-regression analysis
According to the study time, study country, and MRI magnetic field intensity, a meta-regression analysis was carried out. The cause of heterogeneity was not found.
## 3.20. Sen analysis
After excluding individual studies one by one, the remaining studies were pooled and analyzed again. The results showed that each study eliminated had little impact on the amount of pooled effect sizes, indicating that the results of this study were relatively stable and the reliability of the analysis results was high (Fig. 8).
**Figure 8.:** *The sensitivity analysis of MRA. MRA = magnetic resonance arthrography.*
## 3.21. Publication bias analysis
Taking the inverse of the square root of effective sample size [1/root (ESS)] as the ordinate and DOR as the abscissa, the results of Deeks test showed that the P value of slope coefficient was 0.79, suggesting that there was no publication bias in the MRI examination method (Fig. 9).
**Figure 9.:** *Funnel plot of MRA for the diagnosis of ALT. ALT = acetabular labral tears, MRA = magnetic resonance arthrography.*
## 4. Discussion
The research quality of the included studies was assessed by the Quality Assessment of Diagnostic Accuracy Studies-2 tool. The results showed that the quality of the Applicability Concerns in 3 aspects, including Patient Selection, Index Test, and Reference Standard, was good. However, the Risk of Bias assessment in terms of Patient Selection, Reference Standard, and Flow and Timing was not satisfactory. The main reason is that the included studies did not provide answers to the following questions: “*Was a* consecutive or random sample of patients enrolled?”; “ Were the reference standard results interpreted without knowledge of the results of the index tests?”; and “Was there an appropriate interval between index test and reference standard?” Additionally, some studies did not provide clear information about the following aspects: “The reference standard results interpreted with the knowledge of the results of the index tests”; “Not all patients receive the same reference standard”; and “Not all patients included in the analysis.” By referring to the effect sizes of MRI and MRA, we found that MRI and MRA had high accuracy in diagnosing ALT, but the effect sizes Sen(pooled) and DOR(pooled) of MRA were higher than those of MRI, while the effect size −LR(pooled) of MRA was lower than that of MRI. Considering that MRA has a higher diagnostic value, it has become the examination of choice for the evaluation of the acetabular labrum because of its excellent soft-tissue contrast and spatial resolution.[49] When MRA is used, the injection of contrast media allows the joint capsule to expand and distinguish between the acetabular labral and the surrounding capsule tissue. The contrast media inserted into the acetabular labral also make the ALT more clearly displayed. Therefore, MRA has become the preferred imaging examination for the diagnosis of ALT. To explore the source of heterogeneity, this study also conducted meta-regression and subgroup analyses. The results of the subgroup analysis for different research years of MRI showed that the effect sizes +LR(pooled) and DOR(pooled) for the research years from 2010 to 2019 were higher than those for the research years from 1995 to 2009. The effect size −LR(pooled) for the research years 2010 to 2019 was lower than that for the research years 1995 to 2009. This shows that in recent years, the rapid development of biotechnology has improved the diagnostic efficiency of MRI. According to the subgroup analysis of different MRI research year, we found that there was no threshold effect in the study after 2009, indicating that the diagnostic methods and evaluation criteria of MRI tend to be uniform after 2009.[50] To improve the stability and reliability of the research results, during the implementation of this meta-analysis, 2 reviewers independently extracted the data and assessed the risk of bias. Strict inclusion and exclusion criteria were formulated during literature screening. Considering the differences between studies, meta-regression and subgroup analysis were carried out to find the source of heterogeneity. When the source of heterogeneity could not be found, the random-effect model was used to make the final results more reliable. This also makes our research more comprehensive than previous studies in terms of study time, study country and magnetic field intensity.[51–53] Although meta-regression and subgroup analysis were carried out for the included studies, the reports of patients’ age, condition, and course of disease were incomplete, and there were certain differences in testing equipment and image analyst information, which might have also led to certain heterogeneity among the included studies. Moreover, the sample size of some studies was small, and the quality of some of the included studies was not very high. Finally, some studies regarded patients as research objects, while others regarded acetabular labrum as research objects, which also affect the results of this meta-analysis.
## 5. Conclusion
In this study, it was found that MR had a certain value in the diagnosis of ALT; in particular, MRA had higher diagnostic efficiency, and its application in the diagnosis of ALT was feasible in a clinical setting. However, due to the limitations of this study, the above conclusions still need to be further verified.
## Author contributions
Conceptualization: Zhihao Huang, Wenyu Liu, Tianyu Li.
Formal analysis: Zhihao Huang, Wenyu Liu, Pengfei Zhao.
Investigation: Wenyu Liu.
Methodology: Zhihao Huang, Zhihao Liu.
Project administration: Zhihao Huang.
Supervision: Zhihao Huang, Wenyu Liu.
Visualization: Zhihao Huang, Tianyu Li.
Writing – original draft: Zhihao Huang.
Writing – review & editing: Zhihao Huang, Wenyu Liu, Tianyu Li, Zhihao Liu, Pengfei Zhao.
## References
1. Ferguson SJ, Bryant JT, Ganz R. **The acetabular labrum seal: a poroelastic finite element model.**. *Clin Biomech (Bristol, Avon)* (2000) **15** 463-8. PMID: 10771126
2. Ferguson SJ, Bryant JT, Ganz R. **An in vitro investigation of the acetabular labral seal in hip joint mechanics.**. *J Biomech* (2003) **36** 171-8. PMID: 12547354
3. Philippon MJ. **The role of arthroscopic thermal capsulorrhaphy in the hip.**. *Clin Sports Med* (2001) **20** 817-30. PMID: 11675889
4. Seldes RM, Tan V, Hunt J. **Anatomy, histologic features, and vascularity of the adult acetabular labrum.**. *Clin Orthop Relat Res* (2001) **382** 232-40
5. Dwyer MK, Lewis CL, Hanmer AW. **Do neuromuscular alterations exist for patients with acetabular labral tears during function?**. *Arthroscopy* (2016) **32** 1045-52. PMID: 27129378
6. Groh MM, Herrera J. **A comprehensive review of hip labral tears.**. *Curr Rev Musculoskelet Med* (2009) **2** 105-17. PMID: 19468871
7. Amber I, Mohan S. **Preventing overdiagnosis of acetabular labral “Tears” in 40-plus-year-old patients: shouldn’t these be called labral “Fissures” Instead?**. *Acad Radiol* (2018) **25** 387-90. PMID: 29199059
8. Schmitz MR, Campbell SE, Fajardo RS. **Identification of acetabular labral pathological changes in asymptomatic volunteers using optimized, noncontrast 1.5-T magnetic resonance imaging.**. *Am J Sports Med* (2012) **40** 1337-41. PMID: 22422932
9. Ikeda T, Awaya G, Suzuki S. **Torn acetabular labrum in young patients. Arthroscopic diagnosis and management.**. *J Bone Joint Surg Br* (1988) **70** 13-6. PMID: 3339044
10. Dorrell JH, Catterall A. **The torn acetabular labrum.**. *J Bone Joint Surg Br* (1986) **68** 400-3. PMID: 3733805
11. Klaue K, Durnin CW, Ganz R. **The acetabular rim syndrome. A clinical presentation of dysplasia of the hip.**. *J Bone Joint Surg Br* (1991) **73** 423-9. PMID: 1670443
12. Suzuki S, Kasahara Y, Seto Y. **Arthroscopy in 19 children with Perthes’ disease. Pathologic changes of the synovium and the joint surface.**. *Acta Orthop Scand* (1994) **65** 581-4. PMID: 7839839
13. McCarthy JC, Noble PC, Schuck MR. **The role of labral lesions to development of early degenerative hip disease.**. *Clin Orthop Relat Res* (2001) **393** 25-37
14. Weber AE, Harris JD, Nho SJ. **Complications in hip arthroscopy: a systematic review and strategies for prevention.**. *Sports Med Arthrosc Rev* (2015) **23** 187-93. PMID: 26524553
15. Page MJ, McKenzie JE, Bossuyt PM. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.**. *BMJ* (2021) **372** n71. PMID: 33782057
16. Whiting PF, Rutjes AW, Westwood ME. **QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies.**. *Ann Intern Med* (2011) **155** 529-36. PMID: 22007046
17. Guo Z, Wang K, Kadeer K. **The efficacy and safety of flow-diverting device and coil embolization for intracranial aneurysms: a meta-analysis.**. *Eur Rev Med Pharmacol Sci* (2021) **25** 5383-91. PMID: 34533790
18. Gao Q, Zhang W, Li T. **The efficacy and safety of glucokinase activators for the treatment of type-2 diabetes mellitus: a meta-analysis.**. *Medicine (Baltim)* (2021) **100** e27476
19. Egger M, Davey Smith G, Schneider M. **Bias in meta-analysis detected by a simple, graphical test.**. *BMJ* (1997) **315** 629-34. PMID: 9310563
20. Aprato A, Massè A, Faletti C. **Magnetic resonance arthrography for femoroacetabular impingement surgery: is it reliable?**. *J Orthop Traumatol* (2013) **14** 201-6. PMID: 23397418
21. Banks DB, Boden RA, Mehan R. **Magnetic resonance arthrography for labral tears and chondral wear in femoroacetabular impingement.**. *Hip Int* (2012) **2** 387-90
22. Byrd JW, Jones KS. **Diagnostic accuracy of clinical assessment, magnetic resonance imaging, magnetic resonance arthrography, and intra-articular injection in hip arthroscopy patients.**. *Am J Sports Med* (2004) **32** 1668-74. PMID: 15494331
23. Chan YS, Lien LC, Hsu HL. **Evaluating hip labral tears using magnetic resonance arthrography: a prospective study comparing hip arthroscopy and magnetic resonance arthrography diagnosis.**. *Arthroscopy* (2005) **21** 1250. PMID: 16226655
24. Crespo-Rodríguez AM, De Lucas-Villarrubia JC, Pastrana-Ledesma M. **The diagnostic performance of non-contrast 3-Tesla magnetic resonance imaging (3-T MRI) versus 1.5-Tesla magnetic resonance arthrography (1.5-T MRA) in femoro-acetabular impingement.**. *Eur J Radiol* (2017) **88** 109-16. PMID: 28189195
25. Czerny C, Hofmann S, Neuhold A. **Lesions of the acetabular labrum: accuracy of MR imaging and MR arthrography in detection and staging.**. *Radiology* (1996) **200** 225-30. PMID: 8657916
26. Czerny C, Hofmann S, Urban M. **MR arthrography of the adult acetabular capsular-labral complex: correlation with surgery and anatomy.**. *AJR Am J Roentgenol* (1999) **173** 345-9. PMID: 10430132
27. Edwards DJ, Lomas D, Villar RN. **Diagnosis of the painful hip by magnetic resonance imaging and arthroscopy.**. *J Bone Joint Surg Br* (1995) **77** 374-6. PMID: 7744918
28. El-Liethy NE, Zeitoun R, Kamal HA. **Magnetic resonance arthrography, a valuable pre-operative imaging modality infemoro-acetabular impingement.**. *Egypt J Radiol Nucl Med* (2019) **50** 79
29. Freedman BA, Potter BK, Dinauer PA. **Prognostic value of magnetic resonance arthrography for Czerny stage II and III acetabular labral tears.**. *Arthroscopy* (2006) **22** 742-7. PMID: 16843810
30. Gao G, Fu Q, Cui L. **The diagnostic value of ultrasound in anterosuperior acetabular labral tear.**. *Arthroscopy* (2019) **35** 2591-7. PMID: 31416655
31. Hong W, Zhang X, Wang W. **The preliminary application of magnetic resonance arthrography in the diagnosis of acetabular labral tears.**. *Chin J Radiol* (2010) **44** 1140-3
32. Jin W, Kim KI, Rhyu KH. **Sonographic evaluation of anterosuperior hip labral tears with magnetic resonance arthrographic and surgical correlation.**. *J Ultrasound Med* (2012) **31** 439-47. PMID: 22368134
33. Keeney JA, Peelle MW, Jackson J. **Magnetic resonance arthrography versus arthroscopy in the evaluation of articular hip pathology.**. *Clin Orthop Relat Res* (2004) **429** 163-9
34. Leunig M, Werlen S, Ungersböck A. **Evaluation of the acetabular labrum by MR arthrography.**. *J Bone Joint Surg Br* (1997) **79** 230-4. PMID: 9119848
35. Linda DD, Naraghi A, Murnaghan L. **Accuracy of non-arthrographic 3T MR imaging in evaluation of intra-articular pathology of the hip in femoroacetabular impingement.**. *Skeletal Radiol* (2017) **46** 299-308. PMID: 27975135
36. Magee T. **Comparison of 3.0-T MR vs 3.0-T MR arthrography of the hip for detection of acetabular labral tears and chondral defects in the same patient population.**. *Br J Radiol* (2015) **88** 20140817. PMID: 26090824
37. McCarthy JC, Glassner PJ. **Correlation of magnetic resonance arthrography with revision hip arthroscopy.**. *Clin Orthop Relat Res* (2013) **471** 4006-11. PMID: 23904247
38. Mintz DN, Hooper T, Connell D. **Magnetic resonance imaging of the hip: detection of labral and chondral abnormalities using noncontrast imaging.**. *Arthroscopy* (2005) **21** 385-93. PMID: 15800516
39. Nishii T, Nakanishi K, Sugano N. **Acetabular labral tears: contrast-enhanced MR imaging under continuous leg traction.**. *Skeletal Radiol* (1996) **25** 349-56. PMID: 8738000
40. Petersilge CA, Haque MA, Petersilge WJ. **Acetabular labral tears: evaluation with MR arthrography.**. *Radiology* (1996) **200** 231-5. PMID: 8657917
41. Sahin M, Calisir C, Omeroglu H. **Evaluation of labral pathology and hip articular cartilage in patients with Femoroacetabular Impingement (FAI): comparison of multidetector CT arthrography and MR arthrography.**. *Pol J Radiol* (2014) **79** 374-80. PMID: 25352941
42. Studler U, Kalberer F, Leunig M. **MR arthrography of the hip: differentiation between an anterior sublabral recess as a normal variant and a labral tear.**. *Radiology* (2008) **249** 947-54. PMID: 18840790
43. Sundberg TP, Toomayan GA, Major NM. **Evaluation of the acetabular labrum at 3.0-T MR imaging compared with 1.5-T MR arthrography: preliminary experience.**. *Radiology* (2006) **238** 706-11. PMID: 16436825
44. Sutter R, Zubler V, Hoffmann A. **Hip MRI: how useful is intraarticular contrast material for evaluating surgically proven lesions of the labrum and articular cartilage?**. *AJR Am J Roentgenol* (2014) **202** 160-9. PMID: 24370140
45. Tian CY, Wang JQ, Zheng ZZ. **3.0 T conventional hip MR and hip MR arthrography for the acetabular labral tears confirmed by arthroscopy.**. *Eur J Radiol* (2014) **83** 1822-7. PMID: 25022979
46. Tian C, Yuan H, Wang J. **3.0 T high-resolution MRI of acetabular labrum tear.**. *Diagn Imaging Interv Radiol* (2016) **25** 138-41
47. Toomayan GA, Holman WR, Major NM. **Sensitivity of MR arthrography in the evaluation of acetabular labral tears.**. *AJR Am J Roentgenol* (2006) **186** 449-53. PMID: 16423951
48. Zlatkin MB, Pevsner D, Sanders TG. **Acetabular labral tears and cartilage lesions of the hip: indirect MR arthrographic correlation with arthroscopy-a preliminary study.**. *AJR Am J Roentgenol* (2010) **194** 709-14. PMID: 20173149
49. Ha YC, Choi JA, Lee YK. **The diagnostic value of direct CT arthrography using MDCT in the evaluation of acetabular labral tear: with arthroscopic correlation.**. *Skeletal Radiol* (2013) **42** 681-8. PMID: 23073899
50. Zhang J, Xu Z, Li K. **Evaluation on the effect index of diagnostic test.**. *Chin J Evid-based Med* (2013) **13** 890-5
51. Smith TO, Hilton G, Toms AP. **The diagnostic accuracy of acetabular labral tears using magnetic resonance imaging and magnetic resonance arthrography: a meta-analysis.**. *Eur Radiol* (2011) **21** 863-74. PMID: 20859632
52. Reiman MP, Thorborg K, Goode AP. **Diagnostic accuracy of imaging modalities and injection techniques for the diagnosis of femoroacetabular impingement/labral tear: a systematic review with meta-analysis.**. *Am J Sports Med* (2017) **45** 2665-77. PMID: 28129509
53. Zhang P, Li C, Wang W. **3.0 T MRI is more recommended to detect acetabular labral tears than MR Arthrography: an updated meta-analysis of diagnostic accuracy.**. *J Orthop Surg Res* (2022) **17** 126. PMID: 35232459
|
---
title: 'Effect of summer acupoint application treatment (SAAT) on gut microbiota in
healthy Asian adults: A randomized controlled trial'
authors:
- Jie Zhou
- Bangmin Zhou
- Xiaoyue Kou
- Tao Jian
- Limei Chen
- Xinghua Lei
- Shijian Jia
- Xiaoying Xie
- Xianbo Wu
journal: Medicine
year: 2023
pmcid: PMC9981433
doi: 10.1097/MD.0000000000032951
license: CC BY 4.0
---
# Effect of summer acupoint application treatment (SAAT) on gut microbiota in healthy Asian adults: A randomized controlled trial
## Abstract
Acupoint application has served as an important complementary and adjunctive therapy in China. The purpose of this study is to explore the impact of summer acupoint application treatment (SAAT) on the abundance and biological structure of gut microbiota in healthy Asian adults. Based on the CONSORT guidelines, 72 healthy adults were included in this study, randomly divided into 2 groups, receiving either traditional (acupoint application within known relevant meridians, Group A) or sham (treated with placebo prepared by mixing the equal amount of starch and water, Group B) SAAT. SAAT stickers include extracts from Rhizoma Corydalis, Sinapis alba, Euphorbia kansui, Asari Herba, and the treatment group received 3 sessions of SAAT for 24 months, administered to BL13 (Feishu), BL17 (Geshu), BL20 (Pishu), and BL23 (Shenshu) acupoints. Fecal microbial analyses via ribosomal ribonucleic acid (rRNA) sequencing were performed on donor stool samples before and after 2 years of SAAT or placebo treatment to analyze the abundances, diversity, and structure of gut microbiota. No significant baseline differences were present between groups. At the phylum level, the baseline relative abundance of Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, and Fusobacteria was identified in fecal samples collected from each group. After treatment, the relative abundance of Firmicutes was significantly increased in both groups ($P \leq .05$). Notably, a significant decrease in the relative abundance of Fusobacteria was observed in the SAAT treatment group ($P \leq .001$), while the abundance of Bacteroidetes was decreased significantly in the placebo group ($P \leq .05$). At the genus level, the relative abundance of Faecalibacterium and Subdoligranulum species in the 2 groups were all significantly increased ($P \leq .05$). In addition, a significant reduction in the relative abundance of Blautia, Bacteroides, and Dorea in Group A ($P \leq .05$) and *Eubacterium hallii* group and Anaerostipes ($P \leq .05$) in Group B was observed after treatment. Our findings indicated SAAT substantially influenced the bacterial community structure in the gut microbiota of healthy Asian adults, which might serve as potential therapeutic targets for related diseases, and provided a foundation for future studies aimed at elucidating the microbial mechanisms underlying SAAT for the treatment of various conditions such as obesity, insulin resistance, irritable bowel syndrome.
## 1. Introduction
Acupoint application has served as an important complementary and adjunctive therapy in East Asia, and gains wide attention worldwide recently, which has been widely applied in 183 countries according to a 2013 survey.[1] Traditionally, acupuncture contains the rich and profound scientific connotation of Chinese medicine theories that are compatible with Confucianism and Taoism.[2] Currently, there have been various forms of acupuncture applied in several different scenarios, which included electroacupuncture, laser acupuncture, acupressure, auricular needle, knife needle, and moxibustion, etc.[3] Among these therapies, acupoint application treatment (AAT), a modality of traditional acupuncture, as the promising noninvasive procedure combining points and meridians with Chinese herbal medicine has been further refined and generalized in recent years.[4] An experimental study suggested that point application with Angong Niuhuang stickers could improve cognitive function in cerebral ischemic rat models with equivalent efficacy to conventional invasive acupuncture, which is expected to become an economic method for ischemic stroke.[4] Another research indicated AAT with Angong Niuhuang stickers could promote the expression of Bcl-2 and suppress the expression of pro-apoptotic proteins, including Bax and p53 in the hippocampal CA1 area of the cerebral ischemic rat models.[5] Until recently, however, prospective randomized controlled studies regarding the utility of AAT in healthy adults for chronic disease prevention are lacking.
At present, the AAT approach as a route of administration is commonly applied in Asian countries, specifically China and Korea.[2] In Korea, the AAT treatment is often practiced by applying capsicum plaster or other single ingredients on the acupuncture point.[6–9] In contrast, the AAT approach in *China is* performed by plastering Chinese medicine compounds on an acupoint.[10–12] However, it is noteworthy that the choice of herbal formula and points depends on the heterogeneous disease and syndrome.[13] Indeed, the AAT approach has been applied to treat severe chronic and allergic diseases.[14–16] For example, the basic drugs in the prescription for AAT in summer to treat the pulmonary diseases attacking in winter include Rhizoma Corydalis, Semen Sinapis Albae, Radix Euphorbiae Kansui, and Herba Asari.[13,17] Semen Sinapis *Albae is* pungent and hot in nature and has the effects of warming the lung, resolving phlegm, eliminating swelling, unblocking the ligaments, and relieving pain, while Herba Asari, is pungent in taste and warm in nature, which could warm the lung, dissolve phlegm, dispel wind, and disperse cold.[18,19] Therefore, AAT with this formula may contribute to the prevention of inflammatory and metabolic diseases.[20] Additionally, Cortex Cinnamomi and Flos Caryophylli are often applied to alleviate painful chronic diseases.[21,22] Correspondingly, the main acupuncture points used in combination include the Bladder meridian (BL), Ren meridian, and Du meridian according to clinical syndrome differentiation.[13,15] Specifically, BL13 (Feishu) is the most fundamental point for AAT, followed by BL17 (Geshu), BL20 (Pishu), and BL23 (Shenshu) acupoints.[19,23,24] In the Chinese medical system, a season-based AAT approach has also been applied to improve or prevent recurrent seasonal diseases.[25] The summer acupoint application treatment (SAAT, also known as “Sanfujiu”) is a way in which herbal compounds are applied in summer at special points (generally in dog days) to prevent active periods or modify diseases attacking in winter, including bronchial asthma,[11] allergic rhinitis,[14,26] and chronic obstructive pulmonary disease,[16] and so forth,[12] and SAAT has been widely used in many provinces in China since the 1950s.[13] A 2-year follow-up study revealed that SAAT on the BL13 and BL12 (Fengmen) acupoints could reduce the frequency and severity of asthma [27] and SSAT is efficacious in treating seasonal *Allergic rhinitis* and the treatment efficiency was positively correlated with the length of treatment course.[25] In the realm of traditional Chinese medicine (TCM) theory, SAAT was deemed to be applicable to healthy adults for the purpose of prevention. However, there is limited research focused on the mechanisms of SAAT in the prevention of seasonal diseases.
The gut microorganisms play crucial roles in human health and the imbalance of the gut microbiome is involved in the development of disease. The development of culture-independent, high-throughput sequencing technology of microbial metagenomes has enabled the identification of previously unknown members of the microbiota, thereby providing a powerful new perspective for the composition and function of fecal microorganisms.[28] Recently, studies have found that the occurrence of diseases might be connected to the effects of gut dysbiosis, such as inflammatory bowel disease, rheumatoid arthritis, type 2 diabetes, and obesity.[29] Therefore, gut microorganisms as the potentially modifiable causative factor in the initiation and development of diseases have gained much attention.[30] Considering the importance of gut microbiota, and the pivotal role of gut microbiota in the therapeutic effects of TCM, an increasing number of researches have focused on the interactions between TCM and gut microbiota.[31] Until now, few studies have investigated the effects of SAAT on redressing the disturbance of intestinal microbiota.[32] For these reasons, we designed a randomized controlled study to explore the potential biological effects of SAAT compared to placebo on the composition of fecal microbiota in healthy adults as assessed by 16S ribosomal deoxyribonucleic acid (16S rDNA) sequencing.
## 2.1. Study design
The aim of this study was to evaluate the efficacy of SAAT on the abundance and biological structure of gut microbiota in healthy Asian adults. This is a randomized controlled trial that participants will be randomly divided into 2 groups (A receiving the SAAT with herbal compound and B receiving placebo). The protocol was approved by the Institutional Review Board of XinDu Hospital of Traditional Chinese Medicine and Chengdu Hanhe Traditional Chinese Medicine Hospital for Preventive Treatment (Approval No. 201801). All participants provided verbal and written consent, and this study conformed to the standards set by the latest revision of the Declaration of Helsinki. The study adheres to the CONSORT guidelines for reporting randomized trials. After participants were enrolled in the study, they were assigned either to study group A or group B on the basis of random numbers generated by an independent investigator. The acupuncturist would give the participant the corresponding intervention according to the random number. All acupuncturists in the study have Chinese medicine practitioner licenses with at least 5 years of clinical experience. The subjects received 3 treatment sessions per year on the hottest days in summer (“Sanfu” Days), according to the lunar calendar, for 24 consecutive months, which was followed by a follow-up period of 6 months. After enrollment, no additional drugs or interventions were allowed during the treatment period. The obligatory acupoints included BL13 (Feishu), BL17 (Geshu), BL20 (Pishu), and BL23 (Shenshu). Fecal samples were collected twice before and after the treatment in both groups and then frozen at −80°C within half an hour after sampling. Specifically, the first sample was collected in the morning before administering the first SAAT or placebo and the second sample was collected in the morning 6 months after the end of treatments. The specimens underwent DNA extraction and 16S rDNA sequencing, which are described below.
## 2.2. Participants
From February 1, 2018, to November 30, 2019, a total of 72 healthy adults were recruited after receiving systematic examinations and randomly distributed into Group A ($$n = 34$$) and Group B ($$n = 38$$). The inclusion criteria for subjects recruitment included several factors: age 18 to 65 years; no history of acupuncture therapy; medical screening finding good physical health and no meaningful laboratory abnormalities; and providing written informed assent and cooperating with the preventive treatment. The exclusion criteria were as follows: pregnant or lactating women, and those desirous of conceiving in the near future; subjects with organic diseases; subjects with mental disorders; subjects who underwent recent acupuncture or surgical treatments; subjects with skin diseases or defects at the sites of acupoint that prevents the application of SAAT; subjects participating in other clinical trials; and subjects receiving antibiotics, probiotics, prebiotics, TCM, or other drugs within 3 months.
## 2.3. Study intervention and control
For Group A, the subjects were treated with SAAT at each center. The herbal pastes comprised of Rhizoma Corydalis, Semen Sinapis Albae, Radix Euphorbiae Kansui, Herba Asari, and starch, at a ratio of 7:7:4:4:7. In parallel, subjects randomized to Group B were treated with placebo prepared by mixing the equal amount of starch and water. Notably, all herbs were purchased from Sichuan Traditional Chinese Medicine Co. LTD (Chengdu, China). Initially, the herbs were ground into powder (average size of 125 ± 5.8 µm) by WFM Ultra-Grinding Vibration Mill (Chengdu Saierte Machinery Co. LTD, Chengdu, China). Further, the mixture of herbs (10 g) was transformed into an ointment-like semi-solid dosage form using 2 mL of ginger juice. Then, the herbal semi-solid-like ointment was transferred inside a rubber ring on square tape. The rubber ring, with an approximate inner diameter of 2 cm and height of 1 mm, was placed in the middle of the square tape. The herbal ointment or placebo was plastered on 8 acupoints of BL, including bilateral BL13, BL17, BL20, and BL23. All subjects received the SAAT or sham treatment thrice per year on the hottest days in summer for 2 consecutive years (as shown in Table 1). When the subjects had potential adverse events or reported unbearable burning sensations, stinging, pain, and itching during treatment, they were allowed to peel off the ointment immediately and our acupuncturists would deal with the adverse events accordingly. Under normal conditions, the SAAT or sham treatments were performed for 4 hours each time.
**Table 1**
| Yr | Date (Dog day) | Date (Dog day).1 | Date (Dog day).2 |
| --- | --- | --- | --- |
| 2018 | July 17 | July 27 | August 16 |
| 2019 | July 12 | July 22 | August 11 |
| 2020 | July 16 | July 26 | August 15 |
## 2.4. Sample collection and gut microbiota analysis
Fecal samples were collected twice before and after SAAT or sham treatments for participants, shipped on dry ice within half an hour after collection, and then frozen at −80°C freezer for analysis. Microbial DNA was extracted from human samples using the HiPure Soil DNA Kit B (Magen Biotech., Guangzhou, China) according to the manufacturer's protocol. The extracted DNA was quantified and the V3 and V4 hypervariable regions were amplified by polymerase chain reaction (PCR) system (ABI GeneAmp® 9700, Applied Biosystems, Waltham, MA) with the primers containing the forward sequence of “CCTACGGRRBGCASCAGKVRVGAAT” and the reverse sequence of “GGACTACNVGGGTWTCTAATCC.” The PCR reactions were conducted using the following program: 3 minutes of denaturation at 95°C, 27 cycles of 30 seconds at 95°C, 45 seconds for annealing at 55°C, and 45 seconds for elongation at 72°C, and final extension at 72°C for 10 minutes. PCR products were purified and quantified by the agarose gel electrophoresis with $1.5\%$ agarose gel. Then the indexed adapters were added to the ends of the 16S rDNA amplicons to generate the indexed libraries ready for downstream next-generation sequencing sequencing on lllumina Miseq (Illumina, San Diego, CA). The concentrations of DNA libraries were validated by Qubit3.0 Fluorometer (Invitrogen, Waltham, MA, USA). After quantifying the library to 10 nM, the DNA libraries were multiplexed and loaded on an Illumina MiSeq instrument according to the manufacturer’s instructions (TruSeq™ DNA Sample Prep Kit, Illumina, San Diego, CA). Afterward, the sequencing was performed using the paired-end approach. Further, the image analysis and base calling were conducted by the MiSeq Control Software embedded in the MiSeq instrument. The positive and negative reads were joined together to filter the results contained in the sequence of N and the sequence length larger than the 200 bp sequence was retained. The retained sequences were qualitatively filtered and the chimeric sequences were deleted, and the sequence operational taxonomic unit (OTU) cluster analysis was then performed using VSEARCH (1.9.6) with the similarity set to $97\%$. The reference taxonomy was SILVA release 132. The Ribosomal Database Program Classifier Bayesian Algorithm of the OTU species taxonomy was then used to analyze representative sequences and under different species classification levels, statistical community compositions of each sample were performed. Microbiota alpha diversity was assessed by using the abundance-based coverage estimator and Chao1 indices of species richness and the Shannon and Simpson indices of diversity. Next, we assessed the β diversity of gut microbiota in diverse groups using principal component analysis (PCA), unweighted principal coordinate analysis (PCoA), and weighted distance matrices (nonmetric multidimensional scaling, NMDS). Finally, Linear discriminant analysis coupled with effect size (LEfSe) was performed with LEFSE software (LEfSe 1.0) to identify potential microbial biomarkers between groups.
## 2.5. Statistical analysis
Statistical analyses for clinical data were performed by SPSS software version 25 for Windows (SPSS, Chicago, IL). The categorical variables were analyzed using chi-squared test or Fisher exact test and expressed as frequency (percentage), while the continuous variables were expressed as mean ± standard deviation and compared with the Student t test or Mann–Whitney test as appropriate. The intestinal microorganisms were analyzed with the programming languages R (version 3.6.1). The alpha-diversity and relative abundance of the genera and species of gut microbiota were analyzed using Wilcoxon signed-rank/rank-sum test and expressed as median (interquartile range). For beta diversity analysis, the Bray–Curtis algorithm with PCOA, PCA, and NMDS was used to analyze the intergroup differences of the microbiota. An analysis of similarity was used for testing the significance of dissimilarity between the 2 groups by applying the read abundance. In addition, linear discriminant analysis of the effect size (LEfSe) and MetaStat were used to determine differences among the groups. The cladogram diagram showed the hierarchy of evolution between the group differences of microbial community structure and species. The difference of 0.05 was considered statistically significant when the P value < 0.05.
## 3. Results
A total of 72 subjects participated in the study and 60 (29 from Group A and 31 from Group B) participants completed the clinical trials. The remaining 12 patients dropped out for the following reasons: 2 participants from Group A withdrew due to adverse events including hyperpigmentation, allergic to herbs, or burning sensation in the skin, 5 subjects (1 from Group A and 4 from Group B) withdrew due to pregnancy, and 5 subjects (2 from Group A and 3 from Group B) withdrew before the end of the treatments due to personal reasons. In total, 130 stool samples were obtained from participants, of which 60 subjects provided 2 specimens before and after treatment, 10 subjects provided only pretreatment specimens, and the rest 2 subjects did not provide any samples.
## 3.1. Baseline clinical characteristics
Of the 72 participants enrolled in this analysis, there were 11 men and 23 women in group A (average age 43.00 ± 10.79 years) and 10 men and 28 women in group B (average age 37.24 ± 13.67 years). As shown in Table 2, there were no other significant differences in baseline characteristics between these 2 groups.
**Table 2**
| Group | Gender | Gender.1 | Average age (yr) | Height (cm) | Weight (kg) | Heart rate (beats/min) | Blood pressure (mm Hg) | Blood pressure (mm Hg).1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Group | Men | Women | Average age (yr) | Height (cm) | Weight (kg) | Heart rate (beats/min) | Systolic | Diastolic |
| Group A(n = 34) | 11 | 23 | 43.00 ± 10.79 | 162.06 ± 7.57 | 59.12 ± 9.23 | 74.88 ± 9.90 | 116.68 ± 10.88 | 75.12 ± 8.82 |
| Group B(n = 38) | 10 | 28 | 37.24 ± 13.67 | 159.42 ± 8.84 | 56.67 ± 8.91 | 73.97 ± 9.87 | 112.63 ± 11.42 | 73.13 ± 9.44 |
| P value | .57 | .57 | .05 | .18 | .26 | .70 | .13 | .36 |
## 3.2. Sequencing analysis and OTU clustering
In total, 6994,402 sequences were obtained from the 130 samples through 16S rDNA high-throughput sequencing analysis. Furthermore, 510 bacterial OTUs were obtained, with a limited number at the $97\%$ similarity cutoff level (as shown in Table 3). Raw reads were submitted to the National Center for Biotechnology Information Genbank, under project accession number PRJNA766254.
**Table 3**
| Index | Group | Before intervention | Before intervention.1 | Before intervention.2 | After intervention | After intervention.1 | After intervention.2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Index | Group | N | Mean ± SD | P value | N | Mean ± SD | P value |
| Effective sequences | A | 32 | 60042.72 ± 12772.94 | .695 | 29 | 50403.83 ± 12991.60 | .061 |
| Effective sequences | B | 38 | 58576.47 ± 17519.82 | .695 | 31 | 44690.90 ± 10055.65 | |
| Average length (bp) | A | 32 | 446.62 ± 5.17 | .998 | 29 | 444.93 ± 2.99 | .382 |
| Average length (bp) | B | 38 | 446.61 ± 5.22 | .998 | 31 | 444.21 ± 3.34 | |
## 3.3. Alpha diversity analysis
The values of Good coverage of all libraries were over $99.9\%$, indicating that the sample sequencing was relatively complete. Moreover, no significant differences were observed in the evaluated alpha diversity indexes (Shannon, Simpson, Chao1) between the groups. Rarefaction curves indicated that the sequencing depth was enough since the samples reached the plateau phase (Fig. 1). In addition, the number of sequenced reads of each sample was adequate to detect most species in the sample.
**Figure 1.:** *The graphical representation shows the rarefaction curve of each sample. The X-axis shows the effective sequences per sample. Y-axis shows the observed OTUs per sample. Aa12–Aa61 represent the samples’ names before intervention. Ab12–Ab61 indicate the samples’ names after the intervention. In the legend, Aa and Ac represent the total samples of Group A before and after the intervention, respectively. Ba and Bc represent the samples of Group B before and after the intervention. The legends of the later figures indicating the curve names are the same, respectively. OTU = operational taxonomic unit.*
## 3.4. Beta diversity analysis
Structural similarity was explored with the PCoA of the beta diversity analysis, which contained the first 2 principal coordinate axes. There was no appreciable change in the 2 groups before and after intervention (Fig. 2). Subsequently, the PCA analysis also revealed differences between the 2 groups based on the first 2 principal component scores. The data of the 2 groups also showed no segregation along axes regardless of before or after intervention (Fig. 3). The NMDS analysis (non-metric multidimensional scaling) depicted the intra-group dispersion of samples, indicating no significant difference between the 2 groups before or after intervention (Fig. 4).
**Figure 2.:** *The plots indicate the PCoA analysis of 2 groups before and after the intervention. X-axis and Y-axis represent the first 2-component scores. PCoA = principal coordinates analysis.* **Figure 3.:** *The plots present the PCA analysis of 2 groups before and after the intervention. X-axis and Y-axis represent the first 2-component scores. PCA = principal component analysis.* **Figure 4.:** *The plots indicate the NMDS analysis of 2 groups before and after the intervention. NMDS = non-metric multi-dimensional scaling.*
## 3.5.1. LEfSe analysis.
*In* general, the LEfSe analysis provides a phylogenetic tree diagram of clustered species, presenting the changes in the important microorganisms between the different treatment groups. Remarkably, 20 microbiota species were found to have a changed relative abundance before and after SAAT therapy in Group A (Aa vs Ac, Fig. 5A). Furthermore, we also observed an alteration of 10 microbiota species in Group B after 2 years of sham treatment (Ba vs Bc, Fig. 5B). Accordingly, we could speculate that alteration of intestinal microflora was more common in SAAT treatment group as compared to sham treatment group.
**Figure 5.:** *The graphical representations indicate the changes in gut microorganisms by SAAT and sham SAAT. In the phylogenetic tree diagrams of species clustering, green nodes represent microorganisms that play essential roles in the green group. The red nodes represent microorganisms that play significant roles in the red group. Contrarily, the yellow nodes represent microorganisms that do not play essential roles in both groups. SAAT = summer acupoint application treatment.*
## 3.5.2. Analysis of the structure of bacterial phyla.
At the phylum level, the baseline relative abundance of Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, and Fusobacteria was identified in fecal samples collected from each group (Fig. 6). The statistical difference for the abundance of gut microbiota at the phylum level was insignificant between the 2 groups before treatment. Among various bacterial phyla, Firmicutes was the most abundant phylum before treatment in both groups, which accounted for $62.68\%$ in Group A and $65.28\%$ in Group B. After 2 years of treatment, the relative abundance of Firmicutes increased to $73.99\%$ in Group A ($$P \leq .042$$) and $74.82\%$ in Group B ($$P \leq .016$$). Furthermore, at the completion of treatment, the relative abundance of Fusobacteria in Group A was reduced from 0.0078 ± 0.0006 to 0.0005 ± 0.0000 ($P \leq .001$), while the relative abundance of Bacteroidetes was decreased from 0.2395 ± 0.0750 to 0.1300 ± 0.0182 ($$P \leq .037$$) in Group B. Whereas there were no significant differences in the relative abundances of other bacterial phyla before and after treatment ($P \leq .05$).
**Figure 6.:** *The heatmap shows the 5 main bacterial phyla of 2 groups before and after the intervention.*
## 3.5.3. Analysis of the structure of bacterial genera.
At the genus level, As depicted in Figure 7, the relative abundances of 2 genera including Faecalibacterium and Subdoligranulum were significantly increased from 0.085 ± 0.0142 to 0.1622 ± 0.0256 ($$P \leq .014$$) and from 0.085 ± 0.0142 to 0.1622 ± 0.0256 ($$P \leq .014$$) in Group A after treatment completion, respectively. Conversely, the relative abundances of 3 genera including Dorea, Bacteroides, and Blautia significantly reduced from 0.0245 ± 0.0048 to 0.0115 ± 0.0019 ($$P \leq .012$$), 0.2004 ± 0.0391 to 0.1102 ± 0.0205 ($$P \leq .039$$) and 0.1021 ± 0.0163 to 0.0653 ± 0.0096 ($$P \leq .049$$) in Group A, respectively. As for Group B after completion of treatment, the relative abundances of genera Faecalibacterium and Subdoligranulum were substantially increased from 0.0828 ± 0.011 to 0.2001 ± 0.0217 ($$P \leq .001$$) and from 0.0152 ± 0.0036 to 0.0451 ± 0.0102 ($$P \leq .003$$), respectively. Contrarily, the relative abundances of the *Eubacterium hallii* and *Anaerostipes* genera were significantly decreased from 0.0297 ± 0.0045 to 0.0139 ± 0.0039 ($$P \leq .011$$) and from 0.0293 ± 0.007 to 0.0111 ± 0.0037 ($$P \leq .022$$), respectively.
**Figure 7.:** *The graphs indicate the difference between the bacterial genera in the 2 groups before and after intervention (relative abundance values >0.01).*
## 4. Discussion
In this randomized, placebo-controlled study, we evaluated the potential effect of SAAT on fecal microbiota composition in healthy adults by 16S rDNA sequencing. Generally, our present investigation first revealed the relative abundance of the *Fusobacteria phylum* declined significantly after 2 years of treatment with SAAT. To our knowledge, the relative increase of bacterial taxa in the Phylum *Fusobacteria is* associated with periodontal disease[33] and inflammatory bowel disease[34] in humans, which implies that SAAT might change the composition of the gut microbiota to potentially ameliorate colitis-associated diseases.[35] Concomitantly, we observed the proportion of Bacteroidetes was decreased in Group B, resulting in an increasing trend in the Firmicutes-Bacteroidetes ratio (F/B ratio). The F/B ratio has been widely used to indicate microbial dysbiosis[36] and the ratio was lower in Group A as compared to Group B at the completion of treatment (4.59 vs 6.03). The increased F/B ratio, caused by an expansion of Firmicutes and/or a contraction of Bacteroidetes, has been considered a promising modifiable risk factor for obesity and type 2 diabetes,[37] which was well validated in rodent and human samples.[38] Our results with regard to phylum-level changes indicated SAAT might serve as a precautionary measure for metabolic diseases and the specific mechanism needs further exploration.
After the 2-year SAAT treatment, the relative abundance of Blautia, Bacteroides, and Dorea decreased, while the relative abundance of Faecalibacterium and Subdoligranulum increased at the genus level, and the proportion of the *Eubacterium hallii* and Anaerostipes remained relatively constant. Blautia and Faecalibacterium are short-chain fatty acid-producing bacteria (acetate and butyrate, respectively).[39] Data from animal studies show that increasing acetate production via regulation of gut microbiota including Blautia and Dorea could promote ghrelin secretion, hyperphagia, obesity, and its related sequelae.[40] Alternatively, this class of bacteria has been strongly linked to the pathogenesis of acute graft-versus-host disease.[41] Remarkably, less abundant butyrate-producing bacteria including Faecalibacterium along with reduced butyrate formation were observed in inflammatory bowel disease and type 2 diabetes individuals.[42] Consistent with previous studies that demonstrated positive effects of acupuncture on gut microbiota,[28] we hypothesized that SAAT treatment could increase the abundance of beneficial bacterium Faecalibacterium and reduce harmful bacteria, modulating intestinal-immune system and ameliorating inflammation, thereby achieving the purpose of preventing chronic diseases in healthy population. More recently, a study by Bao et. al suggested that acupuncture may increase the abundance of short-chain fatty acids producing bacteria and anti-inflammatory bacteria including Faecalibacterium, thereby enhancing intestinal barrier function, and inhibiting Th1/Th17 cells related proinflammatory cytokines, which provided a safe, effective therapeutic manner for patients with mild to moderate Crohn disease.[28] Research by Wang et al found electroacupuncture could decrease the body weight, waist circumference, and visceral adipose tissues of obese rats by regulating Firmicutes/Bacteroidetes ratio, thereby improving insulin sensitivity, glucose homeostasis, and lipid metabolism.[43] Xu et al explored the impact of acupuncture on intestinal bacteria in osteosarcoma tumor-burdened mice, revealing acupuncture treatment delayed the decrease of Bacteroidetes and the increase of Firmicutes, and the tumor growth in mice-burdened osteosarcoma.[44] The results of Yu et al also suggested that warm acupuncture could regulate a variety of microbial genera and metabolites related to insomnia, including Blautia, reversing the butyrate-mediated upregulation of the cAMP signaling pathway and GAT-1 expression.[45] Collectively, this study is among the very few suggesting that SAAT could modulate the intestinal microbiota and serve as a measure to prevent inflammation and metabolism-associated diseases, whereas the mechanisms underlying SAAT regulating the intestinal microflora remain to be elucidated.
According to the TCM theory, the health of a person depends on the dynamic balance between the physiological state and the surrounding environment. Recently, the gut microbiota has also been identified as an important link for balance.[18] In disease states, however, this holistic balance was disrupted, and SAAT combining points and meridians with drugs were used for triggering a comprehensive and systemic adjustment to balance the disruption.[43] Chan et al performed a meta-analysis and found that san fu tian (a form of SAAT) therapy has positive effects on adult asthma and this technology is relatively safe because of its noninvasive nature.[15] The results of Shi et al indicated SAAT with Shenhuang plaster reshaped the composition of the microbiota, especially butyrate-producing gut bacterium, and promotes intestinal peristalsis, and the combination of paclitaxel with Shenhuang SAAT exerted potent immunostimulatory effects.[23] At present, there have been several studies applying SAAT or other forms of acupuncture to the remission of coronavirus disease 2019-related symptoms, including headache,[46] diarrhea,[47] and fatigue.[48] Whereas, few studies have focused on the mechanisms underlying the prophylactic effect of SAAT.[49] *In this* study, we addressed the potential role of SAAT in maintaining intestinal microbiota homeostasis and the prevention of inflammatory and metabolic disease. It is necessary to carry out long-term follow-up studies in the future to explore the effect of SAAT on the prevention and rehabilitation of chronic conditions, including the sequelae of COVID-19.[50] There are several limitations that should be acknowledged in this study. First, we had a limited number of participants, which might lead to subject selection bias. Second, the extended follow-up has not been assessed and the long-term effect of acupuncture on these indices could not be determined. Third, we only included healthy individuals without enrolling in disease groups, which limits our ability to infer the modulation effect of SAAT on gut microbiome disorders. Finally, we did not perform the metabolomics study of feces and the combination of fecal metabolomics and 16S rDNA gene sequencing would be the future direction of our research. Nonetheless, the present trial is the first randomized, controlled study to evaluate the impact of SAAT on the abundance and biological structure of gut microbiota in healthy Asian adults, which needs further validation in larger cohorts.
## 5. Conclusion
In summary, this study has illustrated that the SAAT substantially influenced the bacterial community structure in the gut microbiota of healthy adults, which might serve as potential therapeutic targets for related diseases, and provided a foundation for future studies aimed at elucidating the microbial mechanisms underlying SAAT for the treatment of inflammatory and metabolic disease. However, our findings should be interpreted with some caution, and the practitioner must reconsider the indications and specifications of SAAT while applying it to healthy people despite its noninvasive nature. In the long term, our findings are important for understanding the positive effect of SAAT on the host intestinal microbiota, and further research should combine microbiome and metabolome data to further assess the molecular effects of SAAT on the prevention of inflammatory and metabolic diseases.
## Acknowledgments
We thank the Science and Technology Department of Sichuan Province, Sichuan Provincial Administration of Traditional Chinese Medicine for the funding.
## Author contributions
Conceptualization: Jie Zhou, Bangmin Zhou, Xianbo Wu.
Data curation: Bangmin Zhou, Xiaoyue Kou, Shijian Jia.
Investigation: Xiaoyue Kou, Tao Jian.
Methodology: Xiaoyue Kou, Tao Jian, Xiaoying Xie.
Project administration: Tao Jian, Xiaoying Xie.
Software: Jie Zhou, Bangmin Zhou.
Supervision: Limei Chen, Xinghua Lei, Xianbo Wu.
Validation: Limei Chen, Xinghua Lei.
Visualization: Jie Zhou, Bangmin Zhou.
Writing – original draft: Jie Zhou, Bangmin Zhou.
Writing – review & editing: Xianbo Wu.
## References
1. Zhang YQ, Jiao RM, Witt CM. **How to design high quality acupuncture trials-a consensus informed by evidence.**. *BMJ* (2022) **376** e067476. PMID: 35354583
2. Kaptchuk TJ. **Acupuncture: theory, efficacy, and practice.**. *Ann Intern Med* (2002) **136** 374-83. PMID: 11874310
3. Du Y, Zhang L, Liu W. **Effect of acupuncture treatment on post-stroke cognitive impairment: a randomized controlled trial.**. *Medicine (Baltim)* (2020) **99** e23803
4. Zhang DS, Liu YL, Zhu DQ. **Point application with Angong Niuhuang sticker protects hippocampal and cortical neurons in rats with cerebral ischemia.**. *Neural Regen Res* (2015) **10** 286-91. PMID: 25883629
5. Zhang D, Fu M, Song C. **Expressions of apoptosis-related proteins in rats with focal cerebral ischemia after Angong Niuhuang sticker point application.**. *Neural Regen Res* (2012) **7** 2347-53. PMID: 25538759
6. Kim KS, Koo MS, Jeon JW. **Capsicum plaster at the korean hand acupuncture point reduces postoperative nausea and vomiting after abdominal hysterectomy.**. *Anesth Analg* (2002) **95** 1103-7. PMID: 12351304
7. Kim KS, Kim DW, Yu YK. **The effect of capsicum plaster in pain after inguinal hernia repair in children.**. *Paediatr Anaesth* (2006) **16** 1036-41. PMID: 16972832
8. Koo MS, Kim KS, Lee HJ. **Antiemetic efficacy of capsicum plaster on acupuncture points in patients undergoing thyroid operation.**. *Korean J Anesthesiol* (2013) **65** 539-43. PMID: 24427460
9. Yun Y, Kim HN, Kim SE. **Comparative analysis of gut microbiota associated with body mass index in a large Korean cohort.**. *BMC Microbiol* (2017) **17** 151. PMID: 28676106
10. Sun SL, Sun ZR, Guo ZX. **Twenty-seven cases of urinary retention after stroke treated by point-application combined with acupuncture.**. *Zhongguo Zhen Jiu* (2012) **32** 933-4. PMID: 23259278
11. Wang XL, Zhang XJ, Huang HY. **Sixty-three cases of pediatric asthma treated with Zhuang medicine herb line moxibustion with point application.**. *Zhongguo Zhen Jiu* (2013) **33** 350. PMID: 23819245
12. Wang L, Pang L, Bai X. **Clinical efficacy on pediatric recurrent pneumonia treated with point application in summer for the prevention in winter.**. *Zhongguo Zhen Jiu* (2016) **36** 261-5. PMID: 27344831
13. Fang YG, Zhou XZ, Liu BY. **A study on the basic drugs and points for point application in summer to treat the diseases with attacks in winter.**. *J Tradit Chin Med* (2010) **30** 180-4. PMID: 21053623
14. Wen CY, Liu YF, Zhou L. **A systematic and narrative review of acupuncture point application therapies in the treatment of allergic rhinitis and asthma during Dog Days.**. *Evid Based Complement Alternat Med* (2015) **2015** 846851. PMID: 26543488
15. Chan CW, Lee SC, Lo KC. **Tian jiu therapy for the treatment of asthma in adult patients: a meta-analysis.**. *J Altern Complement Med* (2015) **21** 200-7. PMID: 25759906
16. Li F, Gao Z, Jing J. **Effect of point application on chronic obstructive pulmonary disease in stationary phase and effects on pulmonary function: a systematic evaluation of randomized controlled trials.**. *J Tradit Chin Med* (2012) **32** 502-14. PMID: 23427380
17. Wang Z, Han D, Qie L. **Acupoints selecting and medication rules analysis based on data mining technique for bronchial asthma treated with acupoint application.**. *Zhongguo Zhen Jiu* (2015) **35** 591-3. PMID: 26480562
18. Yue SJ, Wang WX, Yu JG. **Gut microbiota modulation with traditional Chinese medicine: a system biology-driven approach.**. *Pharmacol Res* (2019) **148** 104453. PMID: 31541688
19. Yu S, Yang J, Yang M. **Application of acupoints and meridians for the treatment of primary dysmenorrhea: a data mining-based literature study.**. *Evid Based Complement Alternat Med* (2015) **2015** 752194. PMID: 25802545
20. Chen S, Jin YT, Zhu ZY. **In vivo study on site of action of sinapine thiocyanate following acupoint herbal patching.**. *Evid Based Complement Alternat Med* (2018) **2018** 9502902. PMID: 29725357
21. Berman BM, Langevin HM, Witt CM. **Acupuncture for chronic low back pain.**. *N Engl J Med* (2010) **363** 454-61. PMID: 20818865
22. Wang Y, Jafari M, Tang Y. **Predicting meridian in Chinese traditional medicine using machine learning approaches.**. *PLoS Comput Biol* (2019) **15** e1007249. PMID: 31765369
23. Shi Y, Xu J, Ding B. **Gastrointestinal motility and improvement efficacy of Shenhuang plaster application on shenque: identification, evaluation, and mechanism.**. *J Immunol Res* (2020) **2020** 2383970. PMID: 32733972
24. Chen HM, Chen CH. **Effects of acupressure at the Sanyinjiao point on primary dysmenorrhoea.**. *J Adv Nurs* (2004) **48** 380-7. PMID: 15500532
25. Peng J, Wu X, Hu J. **Influencing factors on efficacy of summer acupoint application treatment for allergic rhinitis: a retrospective study.**. *J Tradit Chin Med* (2012) **32** 377-81. PMID: 23297559
26. Han D, Liu C, Qie L. **Acupoint selection and medication rules analysis for allergic rhinitis treated with acupoint application-based on data mining technology.**. *Zhongguo Zhen Jiu* (2015) **35** 1177-80. PMID: 26939343
27. Wu X, Peng J, Li G. **Efficacy evaluation of summer acupoint application treatment on asthma patients: a two-year follow-up clinical study.**. *J Tradit Chin Med* (2015) **35** 21-7. PMID: 25842724
28. Bao C, Wu L, Wang D. **Acupuncture improves the symptoms, intestinal microbiota, and inflammation of patients with mild to moderate Crohn’s disease: a randomized controlled trial.**. *EClinicalMedicine* (2022) **45** 101300. PMID: 35198926
29. Feng W, Ao H, Peng C. **Gut microbiota, a new frontier to understand traditional Chinese medicines.**. *Pharmacol Res* (2019) **142** 176-91. PMID: 30818043
30. Kristensen NB, Bryrup T, Allin KH. **Alterations in fecal microbiota composition by probiotic supplementation in healthy adults: a systematic review of randomized controlled trials.**. *Genome Med* (2016) **8** 52. PMID: 27159972
31. Gong X, Li X, Bo A. **The interactions between gut microbiota and bioactive ingredients of traditional Chinese medicines: a review.**. *Pharmacol Res* (2020) **157** 104824. PMID: 32344049
32. Zhang J, Kuang X, Tang C. **Acupuncture for amnestic mild cognitive impairment: a pilot multicenter, randomized, parallel controlled trial.**. *Medicine (Baltim)* (2021) **100** e27686
33. Bragg M, Freeman EW, Lim HC. **Gut microbiomes differ among dietary types and stool consistency in the Captive Red Wolf (Canis rufus).**. *Front Microbiol* (2020) **11** 590212. PMID: 33304337
34. Wei D, Xie L, Zhuang Z. **Gut microbiota: a new strategy to study the mechanism of electroacupuncture and moxibustion in treating ulcerative colitis.**. *Evid Based Complement Alternat Med* (2019) **2019** 9730176. PMID: 31354859
35. Olvera-Rosales LB, Cruz-Guerrero AE, Ramirez-Moreno E. **Impact of the gut microbiota balance on the health-disease relationship: the importance of consuming probiotics and prebiotics.**. *Foods* (2021) **10** 1261. PMID: 34199351
36. Mariat D, Firmesse O, Levenez F. **The Firmicutes/Bacteroidetes ratio of the human microbiota changes with age.**. *BMC Microbiol* (2009) **9** 123. PMID: 19508720
37. Everard A, Cani PD. **Diabetes, obesity and gut microbiota.**. *Best Pract Res Clin Gastroenterol* (2013) **27** 73-83. PMID: 23768554
38. Sanz Y, Moya-Perez A. **Microbiota, inflammation and obesity.**. *Adv Exp Med Biol* (2014) **817** 291-317. PMID: 24997040
39. Sokol H, Leducq V, Aschard H. **Fungal microbiota dysbiosis in IBD.**. *Gut* (2017) **66** 1039-48. PMID: 26843508
40. Bose S, Ramesh V, Locasale JW. **Acetate metabolism in physiology, cancer, and beyond.**. *Trends Cell Biol* (2019) **29** 695-703. PMID: 31160120
41. Wexler AG, Goodman AL. **An insider’s perspective: Bacteroides as a window into the microbiome.**. *Nat Microbiol* (2017) **2** 17026. PMID: 28440278
42. Fan Y, Pedersen O. **Gut microbiota in human metabolic health and disease.**. *Nat Rev Microbiol* (2021) **19** 55-71. PMID: 32887946
43. Wang H, Wang Q, Liang C. **Acupuncture regulating gut microbiota in abdominal obese rats induced by high-fat diet.**. *Evid Based Complement Alternat Med* (2019) **2019** 4958294. PMID: 31275411
44. Xu X, Feng X, He M. **The effect of acupuncture on tumor growth and gut microbiota in mice inoculated with osteosarcoma cells.**. *Chin Med* (2020) **15** 33. PMID: 32292489
45. Yu H, Yu H, Si L. **Influence of warm acupuncture on gut microbiota and metabolites in rats with insomnia induced by PCPA.**. *PLoS One* (2022) **17** e0267843. PMID: 35482778
46. Sun M, Jin X, Zang M. **Acupuncture for headache in COVID-19: a protocol for systematic review and meta-analysis.**. *Medicine (Baltim)* (2021) **100** e28174
47. Liu N, Xu Y, Zhang D. **Moxibustion for diarrhea in COVID-19: a protocol for systematic review and meta-analysis.**. *Medicine (Baltim)* (2022) **101** e28777
48. Wang T, Xu C, Pan K. **Acupuncture and moxibustion for chronic fatigue syndrome in traditional Chinese medicine: a systematic review and meta-analysis.**. *BMC Complement Altern Med* (2017) **17** 163. PMID: 28335756
49. Ren M, Liu Y, Ni X. **The role of acupuncture and moxibustion in the treatment, prevention, and rehabilitation of patients with COVID-19: a scoping review.**. *Integr Med Res* (2022) **11** 100886. PMID: 35967901
50. Luo W, Zhai Y, Sun M. **Clinical study on acupuncture treatment of COVID-19: a protocol for a systematic review and meta-analysis.**. *Medicine (Baltim)* (2022) **101** e28296
|
---
title: Genetic mapping of microbial and host traits reveals production of immunomodulatory
lipids by Akkermansia muciniphila in the murine gut
authors:
- Qijun Zhang
- Vanessa Linke
- Katherine A. Overmyer
- Lindsay L. Traeger
- Kazuyuki Kasahara
- Ian J. Miller
- Daniel E. Manson
- Thomas J. Polaske
- Robert L. Kerby
- Julia H. Kemis
- Edna A. Trujillo
- Thiru R. Reddy
- Jason D. Russell
- Kathryn L. Schueler
- Donald S. Stapleton
- Mary E. Rabaglia
- Marcus Seldin
- Daniel M. Gatti
- Gregory R. Keele
- Duy T. Pham
- Joseph P. Gerdt
- Eugenio I. Vivas
- Aldons J. Lusis
- Mark P. Keller
- Gary A. Churchill
- Helen E. Blackwell
- Karl W. Broman
- Alan D. Attie
- Joshua J. Coon
- Federico E. Rey
journal: Nature Microbiology
year: 2023
pmcid: PMC9981464
doi: 10.1038/s41564-023-01326-w
license: CC BY 4.0
---
# Genetic mapping of microbial and host traits reveals production of immunomodulatory lipids by Akkermansia muciniphila in the murine gut
## Abstract
The molecular bases of how host genetic variation impacts the gut microbiome remain largely unknown. Here we used a genetically diverse mouse population and applied systems genetics strategies to identify interactions between host and microbe phenotypes including microbial functions, using faecal metagenomics, small intestinal transcripts and caecal lipids that influence microbe–host dynamics. Quantitative trait locus (QTL) mapping identified murine genomic regions associated with variations in bacterial taxa; bacterial functions including motility, sporulation and lipopolysaccharide production and levels of bacterial- and host-derived lipids. We found overlapping QTL for the abundance of *Akkermansia muciniphila* and caecal levels of ornithine lipids. Follow-up in vitro and in vivo studies revealed that A. muciniphila is a major source of these lipids in the gut, provided evidence that ornithine lipids have immunomodulatory effects and identified intestinal transcripts co-regulated with these traits including Atf3, which encodes for a transcription factor that plays vital roles in modulating metabolism and immunity. Collectively, these results suggest that ornithine lipids are potentially important for A. muciniphila–host interactions and support the role of host genetics as a determinant of responses to gut microbes.
*Systems* genetics reveals interactions between host and microbial phenotypes in the murine gut, including a role for *Akkermansia muciniphila* in the production of immunomodulatory ornithine lipids.
## Main
The gut microbiome plays fundamental roles in mammalian physiology and human health1–3. Environmental exposures and host genetic variation modulate gut microbiota composition4–6 and contribute to the large degree of interpersonal variation observed in human gut microbial communities. Recent advances in sequencing technologies and analytical pipelines have fuelled progress in our understanding of the impact of host genetics and the gut microbiome on health. Population studies have revealed host genetic-gut microbial trait associations in human7–11 and mouse cohorts12,13. Additionally, studies leveraging host genetic information and Mendelian randomization have highlighted connections between the gut microbiome and other molecular complex traits including faecal levels of short-chain fatty acids14, plasma proteins15 and ABO histo-blood group type16 in humans. However, most of these studies have focused on microbial organismal composition and there is currently a major gap in our understanding of the impact of host genetic variation on the functional capacity of the gut microbiome.
Microbial metabolites are critical nodes of communication between microbes and the host. These include small molecules derived from dietary components (for example, Trimethylamine N-oxide)17 or de novo synthesized by microbes such as vitamins18 and lipids19. Lipids including eicosanoids, phospholipids, sphingolipids and fatty acids act as signalling molecules to control many cellular processes20–22. Gut microbes not only modulate absorption of dietary lipids via regulation of bile acid production and metabolism but are also a major source of lipids and precursor metabolites for lipids produced by the host23,24. Bacterial cell membrane-associated lipids are also important for microbe–host interactions19,25, although our understanding of their roles in these dynamics is only emerging for gut bacteria.
Defining the general principles that govern microbe–host interactions in the gut ecosystem is a daunting task. *Systems* genetic studies can generate hypotheses that invoke processes and molecules that have no precedent, which can be used for the identification of genes, pathways and networks underlying these interactions. To investigate the connections between gut microbes, intestinal lipids and host genetic variation, we leveraged the Diversity Outbred (DO) mouse cohort, a genetically diverse population derived from eight founder strains: C57BL/6J (B6), A/J (A/J), 129S1/SvImJ [129], NOD/ShiLtJ (NOD), NZO/HLtJ (NZO), CAST/EiJ (CAST), PWK/PhJ (PWK) and WSB/EiJ (WSB)26,27. These eight strains harbour distinct gut microbial communities and exhibit disparate metabolic responses to diet-induced metabolic disease28. The DO population is maintained by an outbreeding strategy aimed at maximizing the power and resolution of genetic mapping. We characterized the faecal metagenome, intestinal transcriptome and caecal lipidome in DO mice and performed quantitative trait locus (QTL) analysis to identify host genetic loci associated with these traits. We integrated microbiome QTL (mbQTL) and caecal lipidome QTL (clQTL) to uncover microbe–lipid associations and identified candidate genes expressed in the distal small intestine associated with these co-mapping traits. These datasets represent a valuable resource for interrogating the molecular mechanisms underpinning interactions between the host and the gut microbiome.
## Gut microbial features are associated with host genetics
We characterized the faecal microbiome from 264 DO mice fed a high-fat high-sucrose (HF/HS) diet for ~22 weeks (Extended Data Fig. 1). We and others previously showed that this diet elicits a wide range of metabolic responses in the eight founder strains that are associated with microbiome changes, and identified loci associated with variation in abundance of bacterial taxa in the gut28,29; here we examine the role of host genetics in influencing gut microbiome traits with a focus on gut bacterial functions. Metagenomic analysis revealed ~1.9 million unique predicted microbial open reading frames (that is, metagenes), 2,803 bacterial functions (KEGG orthologues, KOs) and 187 bacterial taxa across all mice. We also performed metagenomic binning to obtain metagenome-assembled genomes (MAGs), corresponding to species-level bacterial genomes (Extended Data Fig. 2, Supplementary Tables 1–4 and Supplementary Note 1).
We next used QTL analysis to identify regions of the mouse genome associated with the abundance of these traits. We detected 760 associations for KOs (logarithm of odds (LOD) > 6.87, Pgenome-wide-adj < 0.2), 200 of which were genome-wide significant (LOD > 7.72, Pgenome-wide-adj < 0.05) and 45 associations for bacterial taxa (LOD > 6.87, Pgenome-wide-adj < 0.2), 15 of which were genome-wide significant (LOD > 7.72, Pgenome-wide-adj < 0.05) (Fig. 1a and Supplementary Tables 5 and 6). We identified a QTL hotspot on chromosome 15 at 63–64 Mbp; this genomic region was associated with 154 microbial traits with LOD score > 6 (Supplementary Table 7). We estimated DO founder allele effects as best linear unbiased predictors for the traits that mapped to this locus. Among these, we detected two clear groups of traits that exhibited opposite allele effects: a group of KOs and taxa showing positive association with the 129 allele, and another group of KOs and taxa that were negatively associated with the 129 allele (Extended Data Fig. 3). As detailed below, the two most abundant gut bacterial phyla, Firmicutes and Bacteroidetes, mapped to this locus with opposite allele effects. Fig. 1Genetic architecture of QTL for microbial traits in the DO mouse cohort.a, QTL mapping results for 2,803 gut microbial KO function traits (top panel) and 187 bacterial taxa traits (bottom panel) using sex, days on diet and cohort as covariates. Each dot represents a QTL on the mouse genome for a given trait. Dashed lines represent significance thresholds for QTL determined by permutation tests (LOD > 9.19, Pstudy-wide-adj < 0.05; LOD > 7.72, Pgenome-wide-adj < 0.05; LOD > 6.87, Pgenome-wide-adj < 0.2). QTL hotspot at Chromosome 15 is highlighted by grey shading and orange colour text. b, Gut microbiome QTL hotspot on Chr15 has multiple bacterial sporulation and motility functions mapping to it. Protein coding genes are displayed for Chr15: 61–65 Mbp region, *Gasdermin* genes are highlighted in blue. c, *Enrichment analysis* (Fisher’s exact test) for functions mapping at hotspot on Chr15. d, QTL for microbial functions that mapped to Chromosome 15 hotspot had negative 129S1/SvImJ allele effects. QTL for Firmicutes mapping to Chromosome 15 hotspot had negative 129S1/SvImJ allele effects, whereas QTL for Bacteroidetes mapping to this locus had positive 129S1/SvImJ allele effects. e, Spearman correlation analysis between the number of sporulation KOs detected in Firmicutes MAGs mapping at Chromosome 15 QTL hotspot and the LOD scores for these MAGs ($$P \leq 3.87$$ × 10−3, Spearman’s ρ = 0.346).*Source data* Pathway enrichment analysis showed that bacterial ‘motility proteins’ and ‘cell growth’ functional categories were significantly enriched in the group of KOs associated most strongly with 129 alleles (Fig. 1b,c). More specifically, abundances of 14 sporulation functions were negatively associated with 129 alleles (Fig. 1d). Further investigation of the KO distribution across all MAGs revealed that all bacterial sporulation KOs were only present in MAGs belonging to Firmicutes, whereas most of KOs that showed positive 129 allele effects were present in MAGs belonging to Bacteroidetes (Extended Data Fig. 4a). To assess whether the allele effects observed from QTL mapping corresponded to the trait patterns in the DO founder strains, we examined previously published 16S ribosomal RNA gene data from age-matched mice from the eight founder strains, also fed an HF/HS diet13. Consistent with these findings, we found that the 129 mouse strain had higher levels of Bacteroidetes and the highest Bacteroidetes/Firmicutes ratio (Extended Data Fig. 4b). Interestingly, we detected a significant positive correlation between the number of sporulation KOs in Firmicutes MAGs mapping at this locus and the LOD scores for these MAGs (Fig. 1e). Importantly, Firmicutes MAGs commonly detected in our dataset that do not contain sporulation KOs (for example, Lactobacillus, Lactococcus) did not exhibit significant association to this QTL. These results support the notion that host genetic variation affects gut community structure in part by modulating the abundance of sporulating bacteria.
Single nucleotide polymorphism (SNP) association analysis within the Chr15 QTL hotspot identified six significant SNPs: two intron variants, SNP rs582880514 in the *Gsdmc* gene and SNP rs31810445 in the *Gsdmc2* gene, both with LOD scores of 8.0; four SNPs that were intergenic variants (Extended Data Fig. 4c). Gasdermins (Gsdm) are a family of pore-forming proteins that cause membrane permeabilization and pyroptosis30, an inflammatory form of programmed cell death that is triggered by intra- and extracellular pathogens31. These results indicate that host genetic variation in Gsdmc/Gsdmc2 is associated with abundance of gut bacterial functions and raises the hypothesis that these host proteins could modulate the abundance of bacterial taxa harbouring motility and/or sporulation functions.
## Caecal lipids are associated with gut microbes and host genetics
We employed a broad discovery strategy to agnostically detect lipid actors potentially relevant to gut microbiome–host interactions. We used liquid chromatography coupled with tandem mass spectrometry (LC–MS/MS) to characterize the caecal lipidome of 381 DO mice, including all mice used for the metagenomic analysis. We identified 1,048 lipid species representing 35 lipid classes (Fig. 2a,b) and the four major lipid categories: [1] fatty acyls, [2] phospholipids, [3] sphingolipids and [4] glycerolipids. The highest numbers of lipids were recorded for the classes of triglycerides (TG) and phosphatidylcholines (PC), species known to be abundant in the mammalian host32. Of the 3,384 lipid species detected in DO caecum, 547 ($16.2\%$) were detected at higher levels in the caecum of conventionally raised mice compared with caecum of germ-free animals (fold-change >10-fold, adjusted $P \leq 0.05$). Phosphatidylglycerols (PG), for example, which represent the second largest phospholipid class in our data, are known to be a major component of the bacterial lipidome33. In mammals, on the other hand, PG are only a minor component. Similarly, among glycerolipids, monogalactosyldiacylglycerols (MGDG) account for the second highest number of lipids detected in this class. While they are found at high levels in bacteria and plants, these lipids are only minor components of animal tissue34. These findings suggest that our analysis of the caecal lipidome captures components of the host and the gut microbiome. Correlation analysis between MAGs and caecal lipids abundance, plus comparison of the caecal lipidome of conventionally raised vs germ-free mice identified taxa that potentially modulate the abundance of lipids in the gut (Extended Data Fig. 5a,b, Supplementary Tables 8–10 and Supplementary Note 2). Furthermore, QTL mapping identified 399 significant QTL associations for caecal lipid features (LOD > 7.60, Pgenome-wide-adj < 0.05) (Fig. 2c, Supplementary Table 11 and Supplementary Note 3). Altogether these associations provide a wealth of information offering potential molecular descriptors of the genetic regulation of the microbiome. Fig. 2Genetic architecture of the caecal lipidome in DO mice.a, A total of 3,384 caecal lipid features were quantified across 381 DO mice, 1,048 of which were identified as lipids from four major classes. Each dot represents a caecal lipid feature. Features of each class occupied characteristic regions in the m/z – RT space. b, Identified lipids belonged to 35 lipid subclasses, with bacteria-associated PG and MGDG as common subclasses. c, A total of 3,964 suggestive caecal lipid QTL (LOD > 6, Pgenome-wide-adj < 0.2) and 12 QTL hotspots were identified. Hotspots are marked with arrows and the corresponding genomic locus indicated. Dashed lines represent significance thresholds for QTL as determined by permutation tests (LOD > 7.60, Pgenome-wide-adj < 0.05). Of the identified lipids, $68.2\%$ showed a total of 1,162 QTL (top panel), while a similar portion of $70.1\%$ of unidentified features contributed 2,802 QTL (bottom panel). RT, retention time. For lipid class abbreviations, see Supplementary Table 16.Source data
## Mediation analysis reveals bacteria–caecal lipids connections
To identify causal links between gut microbial traits and caecal lipid traits, we performed mediation analysis between individual gut microbial metagenes and lipid features that co-map (Methods). Mediation analysis seeks to determine whether a QTL has separate effects on two traits, or if it affects one trait through its effect on another trait, in which case the intermediate trait is called a mediator. Figure 3a shows gut microbial metagenes mediating the QTL effect on a caecal lipid trait. We reasoned that if a microbial trait influenced a caecal lipid that was independent from the caecal lipid’s QTL, its inclusion as a covariate would be unlikely to affect the caecal lipid QTL signal significantly. However, for microbial traits that mediate the QTL effect on the caecal lipid, there would be a large drop in the original caecal lipid QTL LOD score. Interestingly, we found three caecal lipid features with QTL that were mediated by microbial metagenes. Most of these mediating microbial traits were genes belonging to the bacterium Akkermansia muciniphila. It is important to note that the direction of the causal effect between microbial trait and caecal lipid cannot be directly inferred from the data. These results suggest that A. muciniphila levels and the abundance of these lipid species in the gut are modulated by the same loci and that the two traits are potentially connected (Fig. 3b,c).Fig. 3Mediation analysis revealed potential causal relationship between A. muciniphila and OL.a, Illustration of Mediation effect model and Null model. Each dot in the scatterplot represents the result of the mediation test for a gut microbial metagene–caecal lipid feature pair; x axis shows the drop in QTL LOD score for caecal lipid features when adding gut microbial metagenes as covariates to the caecal lipid QTL model; y axis shows the original QTL LOD score for each caecal lipid. Dots with the same y axis value represent the mediation test of individual metagenes with one caecal lipid feature. A high QTL LOD score drop represents a significant mediation effect of the gut microbial feature to the caecal lipid feature. Association of three unknown caecal lipid features with the host genome was impacted by A. muciniphila genes. This is depicted as multiple red dots (many dots appear as lines) showing significant mediation effects. b, Three lipid features mediated by A. muciniphila genes were identified as ornithine lipids. The dashed lines connecting a and b point to the fragmentation patterns of identified ornithine lipids, as shown by the m/z values; key fragments are shown in red colour together with their respective chemical structures. c, Distribution of LOD score drop when adding individual A. muciniphila genes as covariates (Mediation model) or adding individual genes not from A. muciniphila as covariates (Null model) for three identified ornithine lipids. d, Three ornithine lipids species QTL co-mapped at five loci (Chromosome 1, Chromosome 2, Chromosome 7, Chromosome 12, Chromosome 15) with A. muciniphila MAGs QTL.QTL with LOD > 5.5 are highlighted by red colour. e, Founder allele effects for A. muciniphila MAGs and caecal OL were estimated in the DO population from the founder strain coefficients observed for the corresponding QTL at each locus from d.*Source data* We further tested whether these caecal lipids and A. muciniphila mapped to the same loci. Mapping of the 46 reconstructed A. muciniphila MAGs to the host genome revealed multiple QTL including Chr1: 92.9 Mbp, Chr2: 79.4 Mbp, Chr7: 129.8 Mbp, Chr12: 59.4 Mbp, and Chr15: 75.9 Mbp (Fig. 3d). Interestingly, the three caecal lipids also showed QTL at the same loci and exhibited similar founder allele effect patterns (Fig. 3e). These founder allele effects on A. muciniphila abundance are consistent with a previous study of gut bacterial abundance in the DO founder strains13. Although these lipid features were not initially identified by our lipidomic analysis pipeline, they appeared to be closely related to each other. Further analysis of their fragmentation spectra suggested that these unidentified features were ornithine lipids (OL) (Fig. 3b and Supplementary Note 4). This was confirmed with a synthetic OL (see below). The three features would have the sum compositions of OL 30:0, OL 31:0 and OL 32:0, detected as [M+H]+ ions. In OL, a 3-hydroxy fatty acid is connected via an amide linkage to the ornithine amino acid that serves as the headgroup. A second fatty acid is then connected to the first via an ester linkage35. OL are bacteria-specific non-phosphorus glycolipids that are found in the outer membranes of selected Gram-negative bacteria36,37.
## A. muciniphila produces OL in the mouse and human gut
A. muciniphila is a Gram-negative bacterium that has been associated with many beneficial effects on host metabolic health38,39. While previous research suggests that OL are important for microbe–host interactions25,40, the occurrence of these lipids in gut bacteria was not known. To test whether A. muciniphila produces OL, we first profiled lipids in A. muciniphila and two other Gram-negative species, Bacteroides thetaiotaomicron and *Escherichia coli* grown under anaerobic conditions. We found similarly high levels of all three targeted OL species in extracts from A. muciniphila but not in the other species, which were indistinguishable from the solvent blank (Fig. 4a). Since phosphate limitation triggers production of OL in some bacterial species25, in follow-up experiments we tested whether phosphate levels modulated abundance of OL in A. muciniphila grown in vitro. We examined three different levels of phosphate (0.02 mM (growth limiting), 0.2 mM (adequate) and 2 mM (excess)). LC–MS/MS analysis confirmed that OL are a dominant lipid species detected in A. muciniphila cell extracts regardless of the phosphate levels included in the growth media (Extended Data Fig. 6a,b). Furthermore, OL were detected in extracellular vesicles isolated from A. muciniphila grown in vitro (Extended Data Fig. 6c and Supplementary Note 6). These results suggest that OL are probably localized in the A. muciniphila outer membranes and provide insights into how these lipids may interact with the host. Fig. 4A. muciniphila produces OL in the mouse and human gut.a, OL abundance for the three major species detected in mice in cell pellets collected from A. muciniphila (A. m), B. thetalotamicron (B. t) and E. coli (E. c) grown in vitro ($$n = 3$$ biologically independent samples per organism). b, OL detected in caecal contents from gnotobiotic mice colonized with A. muciniphila, B. thetaiotaomicron, E. coli and A. muciniphila plus E. coli for two weeks ($$n = 3$$–4 mice per treatment). c, Detection of prominent OL species in human faecal samples is significantly correlated with A. muciniphila abundance as determined by two-sided Spearman correlation ($$n = 16$$ independent faecal samples). Box and whisker plots denote the interquartile range, median and spread of points within 1.5 times the interquartile range; data beyond the end of the whiskers are plotted individually. Statistical difference between treatment groups was tested by unpaired two-sided Welch’s t- test. Source data We further profiled lipids produced by A. muciniphila colonizing the gut of gnotobiotic mice. Five groups of adult germ-free B6 mice were mono-colonized with each of the species mentioned above, bi-associated with E. coli and A. muciniphila or kept germ-free ($$n = 3$$–5 per group). Mice were maintained in the same HF/HS diet used for the DO study for two weeks after inoculation. LC–MS/MS analysis of caecal contents from these mice showed that only mice colonized with A. muciniphila had detectable levels of OL in their caecum (Fig. 4b). Altogether, these results confirm that A. muciniphila gut colonization is causally linked with high levels of OL.
We examined whether A. muciniphila colonization is associated with the presence of OL in the human gut. We analysed lipid content in a subset of faecal samples from a previously characterized cohort of old adults41 spanning a wide range of A. muciniphila relative abundances (not detectable to $39.8\%$). LC–MS/MS analysis of these human faecal samples detected a broader range of OL species than axenic cultures or mice colonized with A. muciniphila, but the levels of the three previously identified OL 15:0_15:0, OL 16:0_15:0 and OL 17:0_15:0 were all significantly correlated with A. muciniphila levels (Fig. 4c). Together, these results suggest that A. muciniphila is a major producer of OL in the mouse and human gut.
## OL modulate lipopolysaccharide (LPS)-induced cytokine responses
To test whether A. muciniphila-derived OL elicit immune responses on the host, we first chemically synthesized the most abundant OL detected in the DO mouse gut, that is, OL_15:0_15:0. Because of the generally beneficial effects of A. muciniphila on host health as previously documented in both human and mouse studies, and on the basis of the structural similarity between OL and lipid A from LPS, we speculated that the OL might function as antagonists of lipid A. We examined the effects of the OL preparation in the absence and presence of LPS on cytokine production by bone-marrow-derived-macrophages (BMDM). Treatment with LPS induced a significant increase in the production of TNF-α and IL-6 by BMDM obtained from B6 and 129 mice (Extended Data Fig. 7a). In contrast, treatment with OL preparation did not stimulate significant production of TNF-α and IL-6 by these cells (Extended Data Fig. 7b), except for a modest increase at 500 ng ml−1 and 1,000 ng ml−1. However, we observed that pretreatment of macrophages with OL had an inhibitory effect on LPS-induced TNF-α and IL-6 in both B6 and 129 mice without causing significant changes in cell viability (Extended Data Fig. 7c,d). These results suggest that A. muciniphila-derived OL can prevent LPS-induced inflammation response. Furthermore, we measured other cytokines secreted by LPS-treated BMDM and observed that the OL preparation inhibited the production of IL-1β, MCP-1, MIP-1α, GM-CSF, IL-12 and RANTES (Fig. 5), although there were differences in the responses to LPS and OL as a function of BMDM genetic background. In addition, OL increased the levels of anti-inflammatory cytokine IL-10 in these cells (Fig. 5), suggesting that OL may modulate inflammation by altering the levels of both pro-inflammatory and anti-inflammatory cytokines. Interestingly, production of IL-12 in the presence of LPS was more than ten times higher in 129 mice than in B6 mice, and OL had a larger inhibitory effect in these mice (Fig. 5). These results indicate that A. muciniphila-derived OL may influence host innate immune responses and their effects may vary as a function of host genetics. Fig. 5OL modulate LPS-induced production of cytokines from BMDM.Levels of IL-1β, IL-6, IL-10, IL-12, TNF-α, MCP-1, MIP-1α, GM-CSF and RANTES detected in supernatants from B6 and 129 mice BMDM stimulated with LPS (10 ng ml−1) and different concentrations of OL. Box and whisker plots denote the interquartile range, median and spread of points within 1.5 times the interquartile range; data beyond the end of the whiskers are plotted individually. Source data
## Intestinal genes co-map with A. muciniphila and OL QTL
We sought to generate regulatory maps of gene expression regulation in the small intestine and to identify overlapping SNPs associated with gut microbiome. We reasoned that identifying genes whose expression demonstrate shared genetic architecture with bacterial taxa/genes/lipids would not only narrow the list of candidate genes at each locus but would also provide invaluable insights into the biology underlying the microbe–host interactions. Furthermore, the power of expression QTL (eQTL) mapping to connect genetic polymorphism and complex traits has been well documented by others42,43. We profiled transcript levels in the distal small intestines of 234 DO mice using RNA-seq. We detected 8,137 transcripts with a minimum of ten counts per million (CPM) in at least $10\%$ of DO mice. We identified 4,462 local eQTL with an average LOD score of 21.2 and 10,894 distal eQTL with an average LOD score of 7.1 (Supplementary Table 12). By comparing eQTL allele effects with those for the co-mapping mbQTL and clQTL, we identified gut microbial features and caecal lipids that were potentially co-regulated with intestinal transcripts (Extended Data Fig. 8 and Supplementary Note 7).
We searched the support intervals for the five co-mapping QTL regions for A. muciniphila and OL (Chr1, Chr2, Chr7, Chr12 and Chr15) for candidate host genes of interest using the eQTL data. By comparing the allele effects between co-mapping eQTL and the A. muciniphila/OL QTL, we identified several candidate host genes whose eQTL allele effects were correlated with A. muciniphila/OL (Fig. 6, Extended Data Fig. 9 and Supplementary Table 13). At the Chr1 QTL region, there were four candidate genes: [1] Gene Activating transcription factor 3 (Atf3) had a distal eQTL at Chr1: 92.96 Mbp with QTL LOD score of 6.55. ATF3 plays an important role during host immune response events by negatively regulating the transcription of pro-inflammatory cytokines induced by the activation of toll-like receptor 444. [ 2] *The* gene TRAF-interacting protein with a forkhead-associated domain (Tifa) had a distal eQTL at Chr1: 90.95 Mbp with LOD score of 6.19. TIFA has been reported to sense bacterial-derived heptose-1,7-bisphosphate—an intermediate in the synthesis of LPS—via a cytosolic surveillance pathway triggering the NF-kB response45,46. Additionally, TIFA interacts with TRAF6 to mediate host innate immune responses. [ 3] *The* gene Jumonji domain-containing protein 8 (Jmjd8) had a distal eQTL at Chr1: 92.14 Mbp with LOD score of 6.72. JMJD8 functions as a positive regulator of TNF-induced NF-kB signalling47. A recent study showed that JMJD8 is required for LPS-mediated inflammation and insulin resistance in adipocytes48. [ 4] *The* gene Gcg had a distal eQTL at Chr1: 92.36 Mbp with LOD score of 7.11. Gcg encodes multiple peptides including glucagon, glucagon-like peptide-1(GLP-1). GLP-1 levels are induced by a variety of inflammatory stimuli, including endotoxin, IL-1β and IL-649. The finding that these genes with distal eQTL that co-map with A. muciniphila and OL QTL on Chr1 are involved in host immune responses to microbial-associated molecular patterns (MAMPs) such as LPS suggests that expression of these genes contributes to the regulation of host responses to OL and/or potentially modulates the abundance of A. muciniphila. Fig. 6eQTL for distal small intestine (ileum) genes that co-map with A. muciniphila and caecal OL at Chromosome 1.a, QTL of A. muciniphila, caecal OL and eQTL for Tifa, Atf3, Jmjd8 and Gcg co-map at Chr1: 90–95 Mbp. LOD score in y axis represents significance of QTL for each trait. b, Spearman correlation of allele effects between Tifa, Atf3, Jmjd8 and *Gcg* gene eQTL and A. muciniphila/OL QTL.Source data
## Dissecting the link between A. muciniphila and Atf3
We investigated whether the co-mapping between A. muciniphila/OL QTL and *Atf3* gene eQTL could be explained by ATF3 impacting the abundance of these traits. To address this question, we measured the abundance of this taxon in wild-type (WT) mice and animals lacking the *Atf3* gene consuming HF/HS diet for four weeks. We observed that Atf3−/− and WT mice had comparable levels of A. muciniphila in faeces as detected by qPCR. Abundance of A. muciniphila was ~$15\%$ lower in faecal samples from Atf3−/− mice compared with wild type ($$n = 7$$ per genotype), yet the differences did not reach significance (Extended Data Fig. 10a). These results suggest that Atf3 does not play a major role in A. muciniphila fitness. It might also act in combination with other factors, which would align with the observation that the abundance of gut A. muciniphila is a polygenic trait.
An alternative explanation for the observed co-mapping is that A. muciniphila/OL modulate expression of Atf3. To examine this idea, we assessed expression profiles of B6 and 129 BMDM stimulated with LPS or a combination of the OL preparation and LPS. DESeq2 analysis identified 674 genes differentially expressed in cells from B6 mice treated with OL (420 genes were upregulated and 254 genes downregulated), whereas 384 genes (304 genes were upregulated and 80 genes downregulated) were impacted by OL in BMDM derived from 129 mice. While differences in gene expression of some of the cytokines discussed above (Extended Data Fig. 10b) were consistent between genotypes, the overall overlap of differentially expressed genes between genotypes was relatively low (Extended Data Fig. 10c) and the responses to the OL varied significantly by genotype (Extended Data Fig. 10e). As mentioned above, ATF3 is a negative regulator of TLR4 signalling. We observed that OL upregulated Atf3 expression for both B6 and 129 BMDMs (Extended Data Fig. 10d). Furthermore, a previous study50 identified 30 genes downregulated by ATF3 in BMDMs (B6 background). Consistent with this result, we found that OL downregulated the expression of these genes in BMDM derived from B6 mice. In contrast, we found that 18 out of these 30 genes were upregulated by OL in BMDM from 129 mice (Extended Data Fig. 10f). These results suggest that the observed co-mapping between A. muciniphila/OL QTL and Atf3 eQTL could be explained by the effect of OL on *Atf3* gene expression and that increased expression of this gene may trigger distinct programmes as a function of host genotype potentially impacting immune and metabolic responses differently.
Altogether, the work supports the notion that A. muciniphila is the major producer of caecal OL in the distal gut and that A. muciniphila-produced OL can negatively regulate host LPS-induced inflammation by upregulating Atf3 expression.
## Discussion
We applied a systems genetics approach to identify relationships between gut microbes, their encoded functions, caecal lipids and host intestinal gene expression. We identified bacterial functions influenced by host genetic variation and discovered that the bacterium A. muciniphila produces immunoactive OL that are detected in faecal samples from humans and mice colonized with this bacterium. A. muciniphila has been previously associated with host genetic variation at several loci in both mice and humans11,12,51,52; however, environmental conditions including diet, which is a major known determinant of microbiome composition, differ dramatically among these studies. The associations described in the present study differ from the ones previously reported in other mouse studies using different diets12,51. We also examined whether gut microbiome traits acted as mediator to previously published metabolic traits for the same cohort of DO mice53; however, no significant mediation was detected, possibly due to the limited statistical power of our study to infer the influence of the gut microbiome on complex metabolic traits.
Previous work suggested that some Gram-negative bacteria produce OL under phosphate-limiting conditions54–56. In contrast, we observed that OL levels were consistently high across a 100-fold phosphate level range, suggesting that phosphate is not a major driver of OL synthesis in A. muciniphila. Notably, a recent study showed that increased OL production by the bacterial pathogen *Pseudomonas aeruginosa* makes its cellular surface more hydrophobic, and resulted in lower virulence and higher resistance to antimicrobials and host immune defences25. A. muciniphila consumes host glycans present in the mucus layer, which is in proximity to the host epithelium. While mucin carbohydrates and amino acids serve as substrates for A. muciniphila, there are also soluble host defence molecules trapped in this layer that prevent invasion of microbes to the underlying mucosal epithelial cells. We speculate that membrane OL impact interactions of A. muciniphila with the intestinal milieu and may represent an adaptation critical to its niche and important for its interactions with the host. Development of tools to genetically manipulate A. muciniphila will be needed to test these hypotheses.
The inhibitory effects of OL on LPS-induced cytokines that we and others have observed57,58 may represent an important aspect of how A. muciniphila impact host physiology. Previous studies identified both natural and synthetic molecules that can inhibit TLR4-mediated LPS signalling—compounds that prevent septic shock, and have anti-inflammatory and anti-neuropathic pain activities in vivo59. One group of LPS antagonist molecules targeting CD14 shares structural features with A. muciniphila OL including a glucose unit linked to two hydrophobic chains and a basic nitrogen on C-660, supporting the potential anti-inflammatory effects of OL. Although the precise mechanisms of how OL inhibit LPS signalling are unknown, our study suggests that A. muciniphila-derived OL may modulate inflammatory responses.
Remarkably, three host innate immunity genes—Atf3, Tifa and Jmjd8—were co-regulated with A. muciniphila. Tifa is located in the ‘cytokine-dependent colitis susceptibility locus’ (Cdcs1) region, a critical genetic determinant of colitis susceptibility in 129 and B6 strains61. TIFA is an important modifier of innate immune signalling through its regulation of TRAF proteins, leading to the activation of NF‐κB and inflammation. Considering the importance of TIFA-dependent immunity to Gram-negative bacteria45, and the differential effects of OL on LPS-treated BMDM from 129 and B6 strains, our results suggest that this gene could be a key player in A. muciniphila-OL–host interactions. Previous studies suggested that ATF3 modulates inflammatory responses by suppressing the expression of TLR4 or CCL4 in macrophages44,62 and revealed a critical role of microbiota in ATF3-mediated gut homoeostasis63. These studies showed that ATF3 negatively regulates Il6 and *Il12* gene expression levels44. In line with this, we found that OL negatively influence these cytokines in LPS-treated BMDM, and their abundance is associated with the same locus that influences Atf3 expression. Previous studies also showed that ATF3 positively regulates host expression of antimicrobial peptides64 and suggested that the production of OL makes the bacterium P. aeruginosa more hydrophobic and resistant to cationic antimicrobial peptides25. However, we observe neither co-mapping of A. muciniphila with expression of antimicrobial peptides nor pronounced differences in A. muciniphila colonization levels between Atf3−/− mice and WT littermates. Instead, the co-mapping of A. muciniphila and Atf3 could be explained by our findings suggesting that [1] A. muciniphila is a major producer of OL in the gut and [2] OL upregulate expression of this key regulator. Although the molecular mechanisms underlying these observations warrant further investigation, these results suggest that A. muciniphila and OL levels are linked to central players of the host immune defence system and support the predominant role of host genetics as a determinant of responses to gut microbes, in particular to A. muciniphila.
In summary, the work presented here links the presence of OL in the human and mouse gut with A. muciniphila and suggests that these lipids are key players in A. muciniphila–host interactions. Our work highlights the importance of bacterial functions and lipids as mediators of the influence of host genetics on the gut microbiome.
## Animal studies
Animal care and study protocols were approved by the AAALAC-accredited Institutional Animal Care and Use Committee of the College of Agricultural Life Sciences at the University of Wisconsin-Madison (UW-Madison). All experiments with mice were performed under protocols approved by the UW-Madison Animal Care and Use Committee (Protocol number A005821 for the DO mice, Protocol number M00559 for gnotobiotic and Atf3 KO mice).
## DO mouse model
DO mice were obtained from the Jackson Laboratory at ~four weeks of age and maintained in the Department of *Biochemistry vivarium* at the UW-Madison. DO mice were allocated in waves of 100 animals, each with an equal number of males and females. All mice were maintained in a temperature (22.2 °C) and humidity ($60\%$) controlled environment under a 12 h light/dark cycle (lights on at 6:00 and off at 18:00). All mice were fed an HF/HS diet (TD.08811, Envigo Teklad, $44.6\%$ kcal fat, $34\%$ carbohydrate and $17.3\%$ protein) and received sterilized water ad libitum upon arrival at the facility. Mice were kept in the same vivarium room and were individually housed to monitor food intake and prevent cross-inoculation via coprophagy. DO mice were killed at 22–25 weeks of age. Faecal samples were collected immediately before euthanasia after a four h fast. Caecal contents and additional tissues were collected promptly after killing and all samples were immediately flash frozen in liquid nitrogen and stored at −80 °C until further processing. Other studies have been published with these mice13,53,65,66.
## Gnotobiotic studies
C57BL/6J germ-free mice were bred and housed in the gnotobiotic mouse facility at the UW-Madison. Male mice were used for the ornithine lipid study. All mice were maintained in a controlled environment (22.2 °C, $60\%$ humidity) in plastic flexible film gnotobiotic isolators under a strict 12 h light/dark cycle (lights on at 6:00 and off at 18:00) on standard chow diet (LabDiet 5021). At eight weeks of age, mice were switched to a western-style HF/HS diet ($44.6\%$ kcal fat, $34\%$ carbohydrate and $17.3\%$ protein) from Envigo Teklad (TD.08811) and orally gavaged with 200 µl of bacterial cultures. At two weeks after colonization, mice were euthanized and caecal contents collected.
## DO founder mice
C57BL6J (B6) and 129S1/SvImJ [129] male mice (five weeks old) were obtained from the Jackson Laboratory. All mice were maintained in a controlled environment (22.2 °C, $60\%$ humidity) under a strict 12 h light/dark cycle (lights on at 6:00 and off at 18:00). All mice were fed a standard chow diet (LabDiet 5021) and received sterilized water ad libitum for 1 week. At six weeks of age, all mice were euthanized to collect bone marrow cells.
## Atf3 mouse studies
Atf3 heterozygous mice (B6.129X1-Atf3tm1Dron/HaiMmnc) were obtained from the Mutant Mouse Resource and Research Center at University of North Carolina. Age- and sex-matched littermates of Atf3-deficient whole body knockout mice (Atf3−/−) and WT mice were generated by crossing Atf3 heterozygous mice. Mice were maintained in a controlled environment under a strict 12 h light/dark cycle (lights on at 6:00 and off at 18:00) at 22.2 °C and $60\%$ humidity. Animals were fed an HF/HS diet (TD.08811, Envigo Teklad, $44.6\%$ kcal fat, $34\%$ carbohydrate and $17.3\%$ protein) and received sterilized water ad libitum after weaning. Faecal samples were collected at seven weeks of age.
## Metagenomic shotgun DNA sequencing
Faecal DNA was extracted from individual pellets collected from DO mice using previously described methods28,67. Following DNA extraction, Illumina paired-end (PE) libraries were constructed using a previously described protocol68, with a modification of gel selecting DNA fragments at ~450 bp length. PE reads (2 × 125) were generated using a combination of MiSeq and HiSeq 2500 platforms.
## Metagenomic reads processing
Raw reads were preprocessed using Fastx Toolkit (v0.0.13) as follows: [1] for demultiplexing raw samples, fastx_barcode_splitter.pl with –partial 2, mismatch 2 was used; [2] when more than one forward and reverse read file existed for a single sample (due to being run on more than one lane, more than one platform or at more than one time), read files were concatenated into one forward and one reverse read file; [3] barcodes were trimmed to form reads (fastx_trimmer -f 9 -Q 33) and [4] reads were trimmed to remove low-quality sequences (fastq_quality_trimmer -t 20 -l 30 -Q33). Following trimming, unpaired reads were eliminated from the analysis using custom Python scripts. To identify and eliminate host sequences, reads were aligned against the mouse genome (mm10/GRCm38) using bowtie269 (v2.3.4) with default settings, and microbial DNA reads that did not align with the mouse genome were identified using samtools (v1.3) (samtools view -b -f 4 -f 8).
## Metagenomic de novo assembly and gene prediction
After removing low-quality sequences and host contaminating DNA sequences, each metagenomic sample was de novo assembled into longer DNA fragments (contigs) using metaSPAdes70 (v3.11.1) with multiple k-mer sizes (metaspades.py -k 21, 33, 55, 77). Contigs shorter than 500 bp were discarded from further processing. Open reading frames (ORFs) (that is, microbial genes, also called metagenes) were predicted from assembled contigs via Prodigal71 (v2.6.3) using Hidden Markov Model (HMM) with default parameters. All predicted genes shorter than 100 bp were discarded from further processing. To remove redundant genes, all predicted ORFs were compared pairwise using the criterion of $95\%$ identity at the nucleotide level over $90\%$ of the length of the shorter ORFs via CD-HIT72 (v4.6.8). In each CD-HIT cluster, the longest ORF was selected as representative. This final non-redundant (NR) microbial gene set was defined as the DO gut microbiome NR gene catalogue.
## Metagenomic annotation
Gene taxonomic annotation was performed using DIAMOND73 (v0.9.23) by aligning genes in the DO gut microbiome NR gene catalogue with the NCBI NR database (downloaded 21 December 2018) using default cutoffs: e-value <1 × 10−3 and bit score >50. Taxonomic assignment used the following parameters: ‘–taxonmap prot.accession2taxid.gz–taxonnodes nodes.dmp’ in DIAMOND command and was determined by the lowest common ancestor (LCA) algorithm when there were multiple alignments. Gene functional annotation was done using the KEGG orthology and links annotation (KOALA) method via the KEGG server (https://www.kegg.jp/ghostkoala/), using 2,698,820 prokaryote genus pan-genomes as reference. The bit score cut-off for K-number assignment was 60.
## Microbiome trait quantification
Quantification of microbial genes was done by aligning clean PE reads from each sample with the DO gut microbiome NR gene catalogue using Bowtie2 (v2.3.4) and default parameters. RSEM74 (v1.3.1) was used to estimate microbial gene abundance. Relative abundances of microbial gene CPM were calculated using microbial gene expected counts divided by gene effective length, then normalized by the total sum. We focused the taxonomic analysis on bacteria since these represented the vast majority of annotated metagenes. We detected 1,927,034 total metagenes and from these, 1,636,209 were annotated as bacterial genes, 195 as archaeal genes, 17,372 as eukaryotic genes and 946 as viruses. There were also 272,312 genes that were unclassified. To obtain abundance information for microbial functions, the CPM of genes with the same KO annotation were summed together. In case there were multiple KO annotations for a single gene, we used all KO annotations. To obtain taxonomic abundance, the CPM of genes with the same NCBI taxa annotation were summed together at phylum, order, class, family and genus levels, with a minimum of ten genes per taxon.
## MAGs reconstruction
To reconstruct bacterial genomes, we clustered assembled contigs with the density-based algorithm DBSCAN using features of two reduced dimensions of contigs 5-mer frequency and contig coverage. The binning process was performed by the pipeline Autometa75 (docker image: ijmiller2/autometa:docker_patch) and allowed deconvolution of taxonomically distinct microbial genomes from metagenomic sequences. The quality of reconstructed metagenomes was evaluated using CheckM76 (v1.1.3); genome completeness >$90\%$ and genome contamination <$5\%$ were required to assign high-quality MAGs. MAGs quantification was done by aligning all clean PE reads from each sample with MAGs from the same sample. Genome coverage was calculated using the bedtools (v2.29.2) ‘genomecov’ command, followed by normalization by library size across all samples. To further remove redundant MAGs, we clustered high-quality MAGs on the basis of whole-genome nucleotide similarity estimation (pairwise average nucleotide identity (ANI)) using Mash software77 (v2.2) with $90\%$ ANI. From high-quality MAGs, we also annotated predicted ORFs from each MAG against the KEGG database and compared the functional potential encoded among different taxa. A. muciniphila MAG IDs are included in Supplementary Table 14.
## Sample preparation for caecal lipidomic analysis
Caecal contents (30 ± 7.5 mg) along with 10 μl SPLASH Lipidomix internal standard mixture were aliquoted into a tube with a metal bead and 270 μl methanol (MeOH) were added for protein precipitation. Control samples comprised 30 ± 7.5 mg of bead beat-combined DO founder strain caecum (NZO, PWK, NOD, B6, 129, AJ) extracted with each batch. To each tube, 900 μl methyl tert-butyl ether (MTBE) and 225 μl of water were added as extraction solvents. All steps were performed at 4 °C on ice. The mixture was homogenized by bead beating for eight min at 25 Hz. Finally, the mixture was centrifuged for eight min at 11,000 × g at 4 °C, after which 240 μl of the lipophilic upper layer were transferred to glass vials and dried by vacuum centrifuge for 60 min.
The dried lipophilic extracts were re-suspended in 200 μl MeOH:toluene (9:1 v/v) per 10 mg dry weight (minimum of 100 μl) to account for varying water content in the samples. The dry weight was determined by drying down the remaining mixture including all solid parts.
## LC–MS/MS analysis of DO mouse caecal samples
Sample analysis by LC–MS/MS was performed in randomized order on an Acquity CSH C18 column held at 50 °C (2.1 mm × 100 mm × 1.7 μm particle diameter; Waters) using an Ultimate 3000 RSLC binary pump (400 μl min−1 flow rate; Thermo Fisher) or a Vanquish binary pump for validation experiments. Mobile phase A consisted of 10 mM ammonium acetate in acetonitrile/H2O (70:30 v/v) containing 250 μl l−1 acetic acid. Mobile phase B consisted of 10 mM ammonium acetate in isopropanol/acetonitrile (90:10 v/v) with the same additives. Mobile phase B was initially held at $2\%$ for two min and then increased to $30\%$ over three min; further increased to $50\%$ over one min and $85\%$ over 14 min; and then raised to $95\%$ over one min and held for seven min. The column was re-equilibrated for two min before the next injection.
DO lipid extracts (20 μl) were injected by an Ultimate 3000 RSLC autosampler (Thermo Fisher) coupled to a Q Exactive Focus mass spectrometer by a HESI II heated electrospray ionization (ESI) source. Both source and inlet capillary were kept at 300 °C. Sheath gas was set to 25 units, auxiliary gas to ten units and the spray voltage was set to 5,000 V (+) and 4,000 V (−), respectively. The MS was operated in polarity switching mode, acquiring positive and negative mode MS1 and MS2 spectra (Top2) during the same separation. MS acquisition parameters were 17,500 resolving power, 1 × 106 automatic gain control (AGC) target for MS1 and 1 × 105 AGC target for MS2 scans, 100 ms MS1 and 50 ms MS2 ion accumulation time, 200- to 1,600 Th MS1 and 200- to 2,000 Th MS2 scan range, 1 Th isolation width for fragmentation, stepped HCD collision energy (20, 30, 40 units), $1.0\%$ under fill ratio and ten s dynamic exclusion.
## QTL mapping
Genetic QTL mapping was performed using the R/qtl2 (v0.24) package78 which fit a linear mixed effect model that included accounting for overall genetic relationship with a random effect, that is, kinship effect. The leave one chromosome out (LOCO) method was used, which accounts for population structure without reducing QTL mapping power. For each gut microbiome trait and caecal lipidome traits, sex, days on diet and mouse cohort (wave) were used as additive covariates as described previously13. For gut microbiome traits and caecal lipidome traits, normalized abundance/coverage was transformed to normal quantiles. The mapping statistic reported was the log10 likelihood ratio (LOD score). The QTL support interval was defined using the $95\%$ Bayesian confidence interval78. Significance thresholds for QTL were determined by permutation analysis ($$n = 1$$,000). We included 2,803 gut microbiome function traits, 197 gut microbiome taxon traits and 3,384 caecal lipid feature traits for our QTL mapping. The reported genome-wide P values were not adjusted for the multiple phenotypes to avoid overly declaring QTL in the initial analysis. We used genome-wide $P \leq 0.05$ for significant QTL and used genome-wide $P \leq 0.2$ to find concordant QTL mapping and hotspots.
## Mediation analysis
Mediation analysis was carried out as previously described79. Mediation analysis was used to relate individual gut microbial metagenes and lipid features by scanning all 136,200 identified metagenes with at least ten CPM in $20\%$ of the samples to all 3,963 caecal lipid features. We used the subset of animals for which both gut metagenomic and caecal lipid data were available ($$n = 221$$). We first defined gut microbial traits with suggestive QTL as the outcome variable; we then included candidate caecal lipid mediators as additive covariates in the suggestive mbQTL mapping model and re-ran the QTL analysis. We performed the same analysis with caecal lipid features as the outcome and gut microbial features as candidate mediators. A mediatory role was supported by a significant decrease in LOD score from the original outcome QTL. Significance of the LOD score drop for a given candidate gut microbial metagene mediator on a given caecal lipid QTL was estimated by z-score scaled by LOD score drop, and a conservative z-score ≤ −6 was recorded as a potential causal mediator. The mean of fitted distributions for a given gut bacterial taxon, for example all metagenes from A. muciniphila gut, was scaled to the corresponding z-score to evaluate the mediation significance for this gut bacterial taxon.
## Bacterial culturing and bacterial extracellular vesicle isolation
A. muciniphila was grown anaerobically in defined medium (Supplementary Table 15). To test for the effects of phosphate condition, the concentration of phosphate in the medium was adjusted to 0.02, 0.2 or 2 mM. E. coli MS200-1 strain was grown in LC medium (10 g l−1 bacto-tryptone, 5 g l−1 bacto-yeast extract, 5 g l−1 NaCl). B. thetaiotaomicron strain VPI-5482 was grown in CMM medium. All bacterial strains were grown at 37 °C. Cells for lipid analyses from the three strains were obtained by centrifugation. Isolation of A. muciniphila extracellular vesicles used a previously described method80.
## Human faecal samples
Stool samples were obtained from a previous study41. Samples were collected from participants of the Wisconsin Longitudinal Study. Briefly, participants collected stool samples directly into sterile containers, then samples were kept at ~4 °C until arrival (48 h or less) at the processing laboratory. Upon arrival, sterile straws were filled with the faecal material and stored at −80 °C as previously described41. 16S rRNA gene sequencing data for these samples were previously published. The use of the Wisconsin Longitudinal Study faecal samples was approved by the Institutional Review Board at UW-Madison. Consent from participants was obtained via a process involving both verbal and written components by trained interviewers, and records were archived both digitally and physically at UW-Madison. This effort did not include collection of samples from vulnerable populations or from minors.
## Sample preparation for OL validation experiments
For caecal contents, 30 ± 6 mg caecal contents were aliquoted into a tube with a metal bead and 280 μl MeOH were added for protein precipitation. To each tube, 900 μl MTBE and 225 μl of water were added as extraction solvents. All steps were performed at 4 °C on ice. The mixture was homogenized by bead beating for eight min at 25 Hz. For bacterial cultures, ~75 μl of bacterial culture were aliquoted into a tube and 280 μl MeOH were added for protein precipitation. After the mixture was vortexed for 10 s, 900 μl MTBE were added as extraction solvent and the mixture was vortexed for ten s and mixed on an orbital shaker for six min. Phase separation was induced by adding 225 μl of water followed by 20 s of vortexing. All steps were performed at 4 °C on ice. Finally, each mixture was centrifuged for eight min at 11,000 × g at 4 °C, after which 240 μl of the lipophilic upper layer were transferred to glass vials and dried in a vacuum centrifuge for 60 min. The dried lipophilic extracts were re-suspended in 200 μl MeOH:toluene (9:1 v/v).
## LC–MS/MS analysis of OL validation experiments
Sample analysis by LC–MS/MS was performed in randomized order on an Acquity CSH C18 column held at 50 °C (2.1 mm × 100 mm × 1.7 μm particle diameter; Waters) using an Ultimate 3000 RSLC binary pump (400 μl min−1 flow rate; Thermo Fisher) or a Vanquish binary pump. The same mobile phase and gradient as for the DO samples were used.
For the validation experiments, 10 μl of caecal or culture extract were injected by a Vanquish Split Sampler HT autosampler (Thermo Fisher) coupled to a Q Exactive HF mass spectrometer by a HESI II heated ESI source. Both source and inlet capillary were kept at 350 °C (Thermo Fisher). Sheath gas was set to 25 units, auxiliary gas to 15 units and spare gas to five units, while the spray voltage was set to 3,500 V and the S-lens RF level to 90. The MS was operated in polarity switching dd-MS2 mode (Top2), acquiring positive and negative mode MS1 and MS2 spectra during the same separation. MS acquisition parameters were 30,000 resolution, 1 × 106 AGC target for MS1 and 5 × 105 AGC target for MS2 scans, 100 ms MS1 and 50 ms MS2 ion accumulation time, 200 to 2,000 Th MS1 scan range, 1.0 Th isolation width for fragmentation and stepped HCD collision energy (20, 30, 40 units).
## Lipidomic analysis
All resulting LC–MS lipidomics raw files were converted to mgf files via MSConvertGUI (ProteoWizard, Dr Parag Mallick, Stanford University) and processed using LipiDex81 and Compound Discoverer 2.0 or 2.1.0.398 (Thermo Fisher) for DO and validation experiments, respectively. All raw files were loaded into Compound Discoverer with blanks marked as such to generate two result files using the following workflow processing nodes: Input Files, Select Spectra, Align Retention Times, Detect Unknown Compounds, Group Unknown Compounds, Fill Gaps and Mark Background Compounds for the so called ‘Aligned’ result and solely Input Files, Select Spectra and Detect Unknown Compounds for an ‘Unaligned’ Result. Under Select Spectra, the retention time limits were set between 0.4 and 21 min, MS order as well as unrecognized MS order replacements were set to MS1. Further replacements were set to FTMS Mass Analyzer and HCD Activation Type. Under Align Retention Times, the mass tolerance was set to ten ppm and the maximum shift according to the data set to 0.6 min for the DO and 0.5 min for the validation experiments. Under Detect Unknown Compounds, the mass tolerance was also set to ten ppm, with an S/N threshold of five (DO) or three (validation), and a minimum peak intensity of 5 × 106 (DO) or 1 × 105 (validation).
For the DO samples, [M+H]+1 and [M−H]−1 were selected as ions and a maximum peak width of 0.75 min as well as a minimum number of scans per peak equalling seven were set. For the validation samples, [M+H]+1 and [M−H+TFA]−1 were selected as ions and a maximum peak width of 0.75 min as well as a minimum number of scans per peak equalling five were set. Lastly, for Group Unknown Compounds as well as Fill Gaps, mass tolerance was set to ten ppm and retention time tolerance to 0.2 min. For best compound selection, rules #1 and #2 were set to unspecified, while MS1 was selected for preferred MS order and [M+H]+1 as the preferred ion. For everything else, the default settings were used. Resulting peak tables were exported as Excel files in three levels of Compounds, Compound per File and Features (just Features for the ‘Unaligned’) and later saved as csv. In LipiDex’ Spectrum Searcher ‘LipiDex_HCD_Acetate’, ‘LipiDex_HCD_Plants’, ‘LipiDex_Splash_ISTD_Acetate’, LipiDex_HCD_ULCFA’ and ‘Ganglioside_20171205’ were selected as libraries for the DO, and ‘Coon_Lab_HCD_Acetate_20171229’, ‘Ganglioside_20171205’ and ‘Ornithine-Lipids_20180404’ for the validation experiments. For all searches, the defaults of 0.01 Th for MS1 and MS2 search tolerances, a maximum of one returned search result and an MS2 low mass cut-off of 61 Th were kept. Under the Peak Finder tab, Compound Discoverer was chosen as peak table type, and its ‘Aligned’ and ‘Unaligned’ results, as well as the MS/MS results from Spectrum Researcher were uploaded. Features had to be identified in a minimum of one file while keeping the defaults of a minimum of $75\%$ of lipid spectral purity, an MS2 search dot product of at least 500 and reverse dot product of at least 700, as well as a multiplier of 2.0 for FWHM window, a maximum of 15 ppm mass difference, adduct/dimer and in-source fragment (and adduct and dimer) filtering and a maximum RT M.A.D Factor of 3.5. As post-processing in the DO, all features that were only found in one file and had no ID were deleted, and duplicates were also deleted. Peak areas of the three targeted ornithine lipid species were obtained via TraceFinder v3.3.350.0 (Thermo Fisher). Details of the lipid classes searched for in these databases with their respective adducts are shown in Supplementary Table 15. Lipids ID matching was performed at <±5 ppm between runs.
## Chemicals and methods
All chemicals were obtained from Chem-Impex, Sigma-Aldrich, Agros Organics or TCI America. All reagents and solvents were used without further purification except for hexane, ethyl acetate and dichloromethane, which were distilled before use. Analytical thin-layer chromatography (TLC) was performed on 250 µm glass-backed silica plates with F-254 fluorescent indicator from Silicycle. Visualization was performed using UV light and iodine.
## General instrumentation information
Nuclear magnetic resonance (NMR) spectra were recorded in deuterated solvents at 400 MHz on a Bruker-Avance spectrometer equipped with a BFO probe, and at 500 MHz on a Bruker-Avance spectrometer equipped with a DCH cryoprobe. Chemical shifts are reported in parts per million using residual solvent peaks or tetramethylsilane (TMS) as a reference. Couplings are reported in hertz (Hz). ESI–exact mass measurement (ESI–EMM) mass spectrometry data were collected on a Waters LCT instrument.
## OL synthesis
Tridecanoic acid (compound 1, 3.2 g, 15 mmol) was dissolved in dichloromethane (150 ml, 0.1 M) in a round-bottom flask equipped with a stir bar. 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (EDC-HCl) (4.3 g, 22.5 mmol), 4-dimethylaminopyridine (DMAP) (273 mg, 2.25 mmol) and Meldrum’s acid (3.2 g, 22.5 mmol) were added to the flask, and the reaction was stirred overnight at room temperature. The next day, the reaction mixture was washed with 1 M HCl (3 × 75 ml), saturated NaHCO3 (3 × 75 ml) and brine (3 × 75 ml). The mixture was then dried over magnesium sulfate and concentrated under reduced pressure. The resultant oil was then dissolved in benzene (19 ml) in a round-bottom flask with a stir bar, and benzyl alcohol (45 mmol, 4.7 ml) was added. The reaction was heated to 95 °C for three hours and then concentrated under reduced pressure. The crude reaction mixture was purified by silica gel flash chromatography (5–$10\%$ ethyl acetate in hexane as eluent), yielding 3.6 g of compound 2 as an oil ($69\%$ yield over two steps).
Compound 2 (3.6 g, 10.4 mmol) was added to a round-bottom flask equipped with a stir bar and dissolved in a 2:1 mixture of tetrahydrofuran (16 ml) and ethanol (8 ml). The round-bottom flask was cooled in an ice bath, and sodium cyanoborohydride (1.6 g, 26 mmol) was added to the mixture. One M aqueous HCl (26 ml, 26 mmol) was added via addition funnel, and the reaction was allowed to stir to room temperature and monitored by TLC. Upon consumption of starting material, the aqueous portion of the reaction was extracted with dichloromethane (3 × 20 ml) and combined with the organic portion. The combined organic portions were washed with brine (3 × 20 ml), dried over MgSO4 and concentrated under reduced pressure to yield 3.26 g of compound 3 ($93\%$ crude). The material was used without further purification.
Pentadecanoic acid (1.93 g, 9 mmol) was added to a round-bottom flask equipped with a stir bar and dissolved in dichloromethane (80 ml). To the flask was added EDC-HCl (2.68 g, 14 mmol), DMAP (974 mg, 8 mmol) and compound 3 (2.78 g, 8 mmol). The reaction mixture was allowed to stir overnight at room temperature. The next day, the mixture was washed with 1 M HCl (3 × 50 ml), saturated NaHCO3 (3 × 50 ml) and saturated brine (3 × 50 ml). The mixture was then dried over magnesium sulfate and concentrated under reduced pressure. The crude material was purified by silica gel flash chromatography (5–$10\%$ ethyl acetate in hexane as eluent), yielding 4.3 g of compound 4 ($94\%$ isolated yield).
To a flame-dried round-bottom flask equipped with a stir bar was added Pd/C (798 mg, 0.75 mmol Pd). Dry dichloromethane was added to the flask to make a slurry, and the atmosphere was exchanged for nitrogen. Compound 4 (4.3 g, 7.5 mmol) was dissolved in anhydrous methanol and added to the reaction vessel. The atmosphere was then exchanged for hydrogen (balloon pressure), and the reaction was allowed to proceed overnight. The next day, the reaction was diluted with ethyl acetate and filtered over celite. The mixture was concentrated under reduced pressure to yield compound 5 as a white solid (3.5 g, $97\%$ crude yield). The material was used without further purification.
Compound 5 (256 mg 0.5 mmol) was added to a round-bottom flask equipped with a stir bar and dissolved in dimethylformamide (DMF) (5 ml). To the flask was added N,N-Diisopropylethylamine (DIPEA) (277 μl, 1.6 mmol) and hexafluorophosphate azabenzotriazole tetramethyl uronium (HATU) (216 mg, 5.5 mmol), and the mixture was stirred for 15 min. Protected ornithine (250 mg, 0.6 mmol) was added to the mixture, which was stirred at room temperature and monitored by TLC. When starting material was no longer observed by TLC, the mixture was diluted in diethyl ether (20 ml) and washed with 1 M HCl (3 × 20 ml), saturated NaHCO3 (3 × 20 ml) and brine (3 × 20 ml). The mixture was dried over magnesium sulfate and concentrated under reduced pressure to yield a white solid (376 mg crude). This sample was combined with an additional sample of the same crude material that appeared identical by 1H NMR analysis and was then purified by silica gel flash chromatography ($25\%$ ethyl acetate in hexane as eluent) to yield 131 mg of compound 6.
To a flame-dried round-bottom flask equipped with a stir bar was added Pd/Cn (17.0 mg, 0.16 mmol Pd). Dry dichloromethane was added to the flask to make a slurry, and the atmosphere was exchanged for nitrogen. The protected ornithine lipid (compound 6, 131 mg, 0.160 mmol) was dissolved in a mixture of 4 ml anhydrous methanol/dichloromethane (DCM) (1:1) and added to the reaction vessel. The atmosphere was then exchanged for hydrogen (balloon pressure), and the reaction was allowed to proceed overnight. The next day, the reaction was filtered over celite. The mixture was concentrated under reduced pressure to yield OL as an off-white solid (82.2 mg, $86\%$ crude yield). Deprotected OL was identified using LC and ESI-EMM ([M]+ calculated 597.5207, measured 597.5188, 0.002 ppm) in the resultant mixture and the material was used without further purification in the experiments described herein.
## RNA-seq and eQTL analysis
Samples of flash-frozen distal ileum from DO mice were homogenized with Qiagen Tissuelyser (two step two min at 25 Hz, with flipping plate homogenization with five min ice incubation). Total RNA was extracted from homogenized samples using Qiagen 96 universal kit (Qiagen). RNA clean-up was performed using Qiagen RNeasy mini kit (Qiagen). DNA was removed by on-column DNase digestion (Qiagen). Purified RNA was quantified using a Nanodrop 2000 spectrophotometer and RNA fragment analyzer (Agilent). Library preparation was performed using the TruSeq Stranded mRNA sample preparation guide (Illumina). IDT unique dual indexes (UDIs), Illumina UDIs or NEXTflex UDIs were used as barcodes for each library sample. RNA sequencing was performed on an Illumina NovaSeq 6000 platform. Raw RNA-seq reads quality control was performed using Trimmomatic82 (v0.39) with default parameters. Genotype-free genome reconstruction and allele specific expression quantification were performed using the GBRS tool (http://churchill-lab.github.io/gbrs/). Genes with at least ten transcripts per million in at least $10\%$ of DO mice were used for downstream analyses. For eQTL mapping, sex, RNA-seq index, RNA-seq wave and mouse cohort (wave) were used as additive covariates. eQTL analysis was otherwise the same as previously described53.
## BMDM assay and cell viability measurement
Bone marrow was isolated from femur and tibia from ~six-week-old B6 and 129 mice fed with chow diet. Bone marrow cells were re-suspended into single-cell suspensions and cultured in complete DMEM medium supplemented with $10\%$ fetal calf serum (FCS), 2 mM l-glutamine, $1\%$ penicillin/streptomycin and 20 ng ml−1 mouse macrophage colony stimulating factor (M-CSF) (BioLegend) for the purpose of differentiation. BMDM cells were randomly allocated into treatment groups. BMDMs were collected at day seven and treated with LPS, OL or LPS + OL for 6 hours in media supplemented with $1\%$ fetal bovine serum (FBS), then supernatants were collected for measurement of cytokines. For optimization, cytokine (TNF-α and IL-6) production from LPS- or OL-treated BMDM was performed using mouse TNF-α ELISA MAX Deluxe kit and mouse IL-6 ELISA MAX Deluxe kit (BioLegend), respectively. Follow-up cytokine (IL-1β, IL-6, IL-10, IL-12, MCP-1, TNF-α, MIP-1α, GM-CSF and RANTES) production assays in response to LPS + OL co-cultured BMDM were performed using Q-Plex Mouse Cytokine Screen 16-Plex (Quansys). Cell viability was determined by flow cytometry (Thermo Fisher Attune NxT) after staining with 7-amino-actinomycin D (eBioscience).
## RNA-seq of BMDM
Frozen BMDM were homogenized with Qiagen Tissuelyser (two min at 20 Hz) and total RNA was extracted using Qiagen 96 universal kit (Qiagen). RNA clean-up was performed using Qiagen RNeasy mini kit (Qiagen). DNA was removed by on-column DNase digestion (Qiagen). Library preparation was performed using the TruSeq Stranded mRNA sample preparation guide (Illumina). RNA sequencing was performed on an Illumina NovaSeq 6000 platform. Raw RNA-seq reads quality control was performed using Trimmomatic82 (v0.39) with default parameters. Gene quantification was performed using RSEM74 (v1.3.1). DESeq283 (v1.26.0) was used to identify differentially expressed genes between groups.
## Akkermansia-specific qPCR for mouse faecal samples
To quantify Akkermansia abundance in mouse faecal samples, previously validated primers specific for A. muciniphila were used (forward CAGCACGTGAAGGTGGGGAC and reverse CTTGCGGTTGGCTTCAGAT)84. A. muciniphila genomic DNA isolated from a pure culture was used to generate a standard curve encompasing seven points (range: 1 ng μl−1–0.015625 ng μl−1). The PCR reaction contained SsoAdvanced Universal SYBR Green Supermix (Bio-Rad). Faecal A. muciniphila abundance was normalized by faecal weight.
## Data analysis and statistical analysis
All data integration and statistical analysis were performed in R (v3.6.3). Data collection and analysis were not performed blind to the conditions of the experiments. No data were excluded from the analysis. No statistical methods were used to pre-determine sample sizes, but our sample sizes are similar to those reported in previous publications13. Differences between groups were evaluated using unpaired two-tailed Welch’s t-test. Enrichment analysis was performed with Fisher’s exact test using a custom R function. Correlation analysis was performed with two-sided Spearman’s correlation using the R function ‘cor.test()’. For multiple testing, Benjamini-Hochberg false discovery rate (FDR) procedure was used to adjust P values. Data integration was performed using R packages dplyr (v1.0.6), tidyr (v1.1.3), reshape2 (v1.4.4) and data.table (v1.14.0). Heat maps were plotted using the R package pheatmap (v1.0.12). Other plots were created using the R packages ggplot2 (v3.3.3), gridExtra (v2.3), RcolorBrewer (v1.1-2) and ggsci (v2.9).
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary InformationSupplementary Notes 1–7. Reporting Summary Supplementary Tables 1–16.
The online version contains supplementary material available at 10.1038/s41564-023-01326-w.
## Extended data
Extended Data Fig. 1Overview of the study. Fecal metagenomes ($$n = 264$$ animals), caecal lipidomes ($$n = 381$$ animals) and distal small intestine transcriptomes ($$n = 234$$ animals) were generated from Diversity Outbred mice. Quantitative trait loci (QTL) analysis identified genomic regions associated with variations in bacterial taxa, bacterial functions, levels of bacterial- and host-derived lipids and small intestine transcript levels. Mediation analysis and co-mapping comparisons were used to identify causal links between traits.
Extended Data Fig. 2DO metagenomic analysis.a, Average percent of assembled reads across all samples. b, Comparison of percent of reads mapping to our generated assembly vs. public database ($$n = 297$$ animals). c, Microbial functions detected for KEGG pathways across all metagenomes. KEGG Orthology (KO) numbers were identified by annotating predicted ORFs to the KEGG database. d, Top 20 gut microbial genera detected across all DO mice ($$n = 264$$ animals). e, Quality of metagenome-assembled genomes. f, Two variants of A. muciniphila MAGs detected in the DO mice. Box and whisker plots denote the interquartile range, median and spread of points within 1.5 times the interquartile range, data beyond the end of the whiskers are plotted individually. Source data Extended Data Fig. 3DO gut microbiome QTL hotspot at Chr15: 61–65Mbp. Founder allele effects of KO and taxa trait QTL at Chr15 hotspot (LOD > 6). Source data Extended Data Fig. 4DO gut microbiome QTL hotspot and SNP associations.a, Presence/absence of KOs that mapped to Chr15 hotspot across all MAGs. Sporulation functions were not detected in Bacteroidetes. b, Estimated founder allele effects for Bacteroidetes and Firmicutes, and Bacteroidetes/Firmicutes ratio (left panel). Observed abundance of Bacteroidetes Firmicutes and Bacteroidetes/Firmicutes ratio in founder strains as determined by Kemis et al. ( right panel, $$n = 9$$-12 animals/founder strain). c, SNPs significantly associated with these traits in Chr15 hotspot include two intron SNPs in Gsdmc and *Gsdmc2* genes. Box and whisker plots denote the interquartile range, median and spread of points within 1.5 times the interquartile range, data beyond the end of the whiskers are plotted individually. Source data Extended Data Fig. 5Correlation between gut bacterial MAGs and caecal lipids.a, Heatmap showing two-sided Spearman correlation coefficients between the abundance of MAGs and caecal lipid levels across DO mice. Bacterial MAGs were clustered into five groups whereas caecal lipids were clustered into six groups. b, Enrichment of the lipid classes for each caecal lipid clusters. Fisher’s exact test was used and Benjamini-Hochberg for multiple tests correction. Source data
Extended Data Fig. 6Detection of ornithine lipids (OL) in Akkermansia muciniphila.a, Heatmap showing relative abundance of all OL species detected in cell pellets from A. muciniphila grown in vitro in defined media supplemented with different levels of phosphate: 20 µM, 200 µM and 2000µM. b, Relative abundance of lipid features detected in cell pellets from A. muciniphila grown in defined media with different levels of phosphate. Top 200 most abundant lipids features are shown. c, Relative abundance of OL features detected extracellular vesicles (AmEVs) purified from A. muciniphila grown in defined medium with the comparison to A. muciniphila cells. Source data Extended Data Fig. 7Cytokine production by BMDM.a,b, (a) TNF-α and (b) IL-6 levels detected in supernatants from BMDM cells in B6 and 129 mice treated for six hours with different concentrations of LPS or OL. c, Cell viability of BMDM cells in B6 and 129 mice treated for six hours with 10 ng/mL LPS and different concentrations of OL. d, Flow cytometry gating strategy for BMDM cell viability assays. $$n = 3$$ biological replicates/treatment group. Box and whisker plots denote the interquartile range, median and spread of points within 1.5 times the interquartile range, data beyond the end of the whiskers are plotted individually. Source data Extended Data Fig. 8Examples of co-mapping QTL.a, At Chr8: 10.5-14.5 Mbp, co-mapping of gut bacterial lipopolysaccharide cholinephosphotransferase function with Pglyrp1 eQTL was observed. b, At Chr4: 50 Mbp, co-mapping of an unidentified caecal feature and a local Acnat1 eQTL was observed. c, The knowledge of Acnat1 conjugating taurine to fatty acids guided the identification of the feature as an N-acyl taurine. d, Fragmentation pattern of identified N-acyl taurine. e, At Chr17: 30-34 Mbp, several unidentified features co-mapped which subsequently could be identified as tocopherols and exemplarily shown for the most significant feature alpha-tocopherol glucuronide. f, Fragmentation pattern of identified alpha-tocopherol glucuronide. Source data Extended Data Fig. 9Founder allele effects on co-mapping traits associated with A. muciniphila levels. A. muciniphila, caecal OL and eQTL genes co-mapping at Chr1: 90-95 Mbp, Chr2: 77-81 Mbp, Chr7: 126-131 Mbp, Chr12: 55-63 Mbp and Chr15: 75-79 Mbp. Source data
Extended Data Fig. 10Expression of genes in BMDM treated with OL/LPS.a, Abundance of A. muciniphila in faecal pellets from Atf3-/- mice and WT mice ($$n = 7$$ mice/genotype; four females, three males for both genotypes). b, Gene expression level of Il1b, Il6 and Il12a from BMDM cells derived from B6 and 129 mice treated for six hours with LPS (10 ng/ml) or with LPS (10 ng/mL) and OL (1 µg/mL). $$n = 3$$ biological replicates/treatment group. c, Number of differentially expressed genes in BMDM derived from B6 and 129 mice. d, Gene expression levels of Atf3 in BMDM from B6 and 129 mice treated for six hours with LPS (10 ng/mL) or LPS (10 ng/mL) and OL (1 µg/mL). $$n = 3$$ biological replicates/genotype/treatment group. e, Differentially expressed genes in BMDM from B6 and 129 mice. f, Previously reported ATF3 regulated genes in BMDM50. Impact of OL on these genes in B6 and 129 mice. Box and whisker plots denote the interquartile range, median and spread of points within 1.5 times the interquartile range; data beyond the end of the whiskers are plotted individually. Statistical difference between treatment groups was tested by two-sided Welch’s t- test. Source data is available for this paper at 10.1038/s41564-023-01326-w.
## Source data
Source Data Fig. 1Statistical source data. Source Data Fig. 2Statistical source data. Source Data Fig. 3Statistical source data. Source Data Fig. 4Statistical source data. Source Data Fig. 5Statistical source data. Source Data Fig. 6Statistical source data. Source Data Extended Data Fig. 2Statistical source data. Source Data Extended Data Fig. 3Statistical source data. Source Data Extended Data Fig. 4Statistical source data. Source Data Extended Data Fig. 5Statistical source data. Source Data Extended Data Fig. 6Statistical source data. Source Data Extended Data Fig. 7Statistical source data. Source Data Extended Data Fig. 8Statistical source data. Source Data Extended Data Fig. 9Statistical source data. Source Data Extended Data Fig. 10Statistical source data.
## Peer review information
Nature Microbiology thanks Ran Blekhman and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
## References
1. Turnbaugh PJ. **An obesity-associated gut microbiome with increased capacity for energy harvest**. *Nature* (2006.0) **444** 1027-1031. DOI: 10.1038/nature05414
2. Wen L. **Innate immunity and intestinal microbiota in the development of Type 1 diabetes**. *Nature* (2008.0) **455** 1109-1113. DOI: 10.1038/nature07336
3. Tremaroli V, Bäckhed F. **Functional interactions between the gut microbiota and host metabolism**. *Nature* (2012.0) **489** 242-249. DOI: 10.1038/nature11552
4. Rey FE. **Metabolic niche of a prominent sulfate-reducing human gut bacterium**. *Proc. Natl Acad. Sci. USA* (2013.0) **110** 13582-13587. DOI: 10.1073/pnas.1312524110
5. Ley RE, Turnbaugh PJ, Klein S, Gordon JI. **Human gut microbes associated with obesity**. *Nature* (2006.0) **444** 1022-1023. DOI: 10.1038/4441022a
6. Yatsunenko T. **Human gut microbiome viewed across age and geography**. *Nature* (2012.0) **486** 222-227. DOI: 10.1038/nature11053
7. Bonder MJ. **The effect of host genetics on the gut microbiome**. *Nat. Genet.* (2016.0) **48** 1407-1412. DOI: 10.1038/ng.3663
8. Wang J. **Genome-wide association analysis identifies variation in vitamin D receptor and other host factors influencing the gut microbiota**. *Nat. Genet.* (2016.0) **48** 1396-1406. DOI: 10.1038/ng.3695
9. **Association of host genome with intestinal microbial composition in a large healthy cohort**. *Nat. Genet.* (2016.0) **48** 1413-1417. DOI: 10.1038/ng.3693
10. Hughes DA. **Genome-wide associations of human gut microbiome variation and implications for causal inference analyses**. *Nat. Microbiol.* (2020.0) **5** 1079-1087. DOI: 10.1038/s41564-020-0743-8
11. Kurilshikov A. **Large-scale association analyses identify host factors influencing human gut microbiome composition**. *Nat. Genet.* (2021.0) **53** 156-165. DOI: 10.1038/s41588-020-00763-1
12. Org E. **Genetic and environmental control of host-gut microbiota interactions**. *Genome Res.* (2015.0) **25** 1558-1569. DOI: 10.1101/gr.194118.115
13. Kemis JH. **Genetic determinants of gut microbiota composition and bile acid profiles in mice**. *PLoS Genet.* (2019.0) **15** e1008073. DOI: 10.1371/journal.pgen.1008073
14. Sanna S. **Causal relationships among the gut microbiome, short-chain fatty acids and metabolic diseases**. *Nat. Genet.* (2019.0) **51** 600-605. DOI: 10.1038/s41588-019-0350-x
15. **Individual variations in cardiovascular-disease-related protein levels are driven by genetics and gut microbiome**. *Nat. Genet.* (2018.0) **50** 1524-1532. DOI: 10.1038/s41588-018-0224-7
16. Rühlemann MC. **Genome-wide association study in 8,956 German individuals identifies influence of ABO histo-blood groups on gut microbiome**. *Nat. Genet.* (2021.0) **53** 147-155. DOI: 10.1038/s41588-020-00747-1
17. Wang Z. **Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease**. *Nature* (2011.0) **472** 57-63. DOI: 10.1038/nature09922
18. Kjer-Nielsen L. **MR1 presents microbial vitamin B metabolites to MAIT cells**. *Nature* (2012.0) **491** 717-723. DOI: 10.1038/nature11605
19. Brown EM. **Bacteroides-derived sphingolipids are critical for maintaining intestinal homeostasis and symbiosis**. *Cell Host Microbe* (2019.0) **25** 668-680.e7. DOI: 10.1016/j.chom.2019.04.002
20. Dennis EA, Norris PC. **Eicosanoid storm in infection and inflammation**. *Nat. Rev. Immunol.* (2015.0) **15** 511-523. DOI: 10.1038/nri3859
21. Baxter AA, Hulett MD, Poon IK. **The phospholipid code: a key component of dying cell recognition, tumor progression and host–microbe interactions**. *Cell Death Differ.* (2015.0) **22** 1893-1905. DOI: 10.1038/cdd.2015.122
22. de Carvalho C, Caramujo M. **The various roles of fatty acids**. *Molecules* (2018.0) **23** 2583. DOI: 10.3390/molecules23102583
23. Schoeler M, Caesar R. **Dietary lipids, gut microbiota and lipid metabolism**. *Rev. Endocr. Metab. Disord.* (2019.0) **20** 461-472. DOI: 10.1007/s11154-019-09512-0
24. Kindt A. **The gut microbiota promotes hepatic fatty acid desaturation and elongation in mice**. *Nat. Commun.* (2018.0) **9** 3760. DOI: 10.1038/s41467-018-05767-4
25. Kim S-K. **Bacterial ornithine lipid, a surrogate membrane lipid under phosphate-limiting conditions, plays important roles in bacterial persistence and interaction with host: role of ornithine lipid in chronic adaptation**. *Environ. Microbiol.* (2018.0) **20** 3992-4008. DOI: 10.1111/1462-2920.14430
26. Svenson KL. **High-resolution genetic mapping using the mouse diversity outbred population**. *Genetics* (2012.0) **190** 437-447. DOI: 10.1534/genetics.111.132597
27. Churchill GA, Gatti DM, Munger SC, Svenson KL. **The diversity outbred mouse population**. *Mamm. Genome* (2012.0) **23** 713-718. DOI: 10.1007/s00335-012-9414-2
28. Kreznar JH. **Host genotype and gut microbiome modulate insulin secretion and diet-induced metabolic phenotypes**. *Cell Rep.* (2017.0) **18** 1739-1750. DOI: 10.1016/j.celrep.2017.01.062
29. O’Connor A, Quizon PM, Albright JE, Lin FT, Bennett BJ. **Responsiveness of cardiometabolic-related microbiota to diet is influenced by host genetics**. *Mamm. Genome* (2014.0) **25** 583-599. DOI: 10.1007/s00335-014-9540-0
30. Shi J. **Cleavage of GSDMD by inflammatory caspases determines pyroptotic cell death**. *Nature* (2015.0) **526** 660-665. DOI: 10.1038/nature15514
31. Liu X, Xia S, Zhang Z, Wu H, Lieberman J. **Channelling inflammation: gasdermins in physiology and disease**. *Nat. Rev. Drug Discov.* (2021.0) **20** 384-405. DOI: 10.1038/s41573-021-00154-z
32. Jain M. **A systematic survey of lipids across mouse tissues**. *Am. J. Physiol. Endocrinol. Metab.* (2014.0) **306** E854-E868. DOI: 10.1152/ajpendo.00371.2013
33. Sohlenkamp C, Geiger O. **Bacterial membrane lipids: diversity in structures and pathways**. *FEMS Microbiol. Rev.* (2016.0) **40** 133-159. DOI: 10.1093/femsre/fuv008
34. Parsons JB, Rock CO. **Bacterial lipids: metabolism and membrane homeostasis**. *Prog. Lipid Res.* (2013.0) **52** 249-276. DOI: 10.1016/j.plipres.2013.02.002
35. Vences-Guzmán MÁ, Geiger O, Sohlenkamp C. **Ornithine lipids and their structural modifications: from A to E and beyond**. *FEMS Microbiol. Lett.* (2012.0) **335** 1-10. DOI: 10.1111/j.1574-6968.2012.02623.x
36. López-Lara IM, Sohlenkamp C, Geiger O. **Membrane lipids in plant-associated bacteria: their biosyntheses and possible functions**. *Mol. Plant Microbe Interact.* (2003.0) **16** 567-579. DOI: 10.1094/MPMI.2003.16.7.567
37. Geiger O, González-Silva N, López-Lara IM, Sohlenkamp C. **Amino acid-containing membrane lipids in bacteria**. *Prog. Lipid Res.* (2010.0) **49** 46-60. DOI: 10.1016/j.plipres.2009.08.002
38. Everard A. **Cross-talk between**. *Proc. Natl Acad. Sci. USA* (2013.0) **110** 9066-9071. DOI: 10.1073/pnas.1219451110
39. Depommier C. **Supplementation with**. *Nat. Med.* (2019.0) **25** 16. DOI: 10.1038/s41591-019-0495-2
40. Diercks H. **Accumulation of novel glycolipids and ornithine lipids in**. *J. Bacteriol.* (2015.0) **197** 497-509. DOI: 10.1128/JB.02004-14
41. Dill-McFarland KA. **Close social relationships correlate with human gut microbiota composition**. *Sci. Rep.* (2019.0) **9** 703. DOI: 10.1038/s41598-018-37298-9
42. Gusev A. **Integrative approaches for large-scale transcriptome-wide association studies**. *Nat. Genet.* (2016.0) **48** 245-252. DOI: 10.1038/ng.3506
43. Tian J. **Identification of the bile acid transporter**. *Genetics* (2015.0) **201** 1253-1262. DOI: 10.1534/genetics.115.179432
44. Gilchrist M. **Systems biology approaches identify ATF3 as a negative regulator of Toll-like receptor 4**. *Nature* (2006.0) **441** 173-178. DOI: 10.1038/nature04768
45. Gaudet RG. **Cytosolic detection of the bacterial metabolite HBP activates TIFA-dependent innate immunity**. *Science* (2015.0) **348** 1251-1255. DOI: 10.1126/science.aaa4921
46. Zhou P. **Alpha-kinase 1 is a cytosolic innate immune receptor for bacterial ADP-heptose**. *Nature* (2018.0) **561** 122-126. DOI: 10.1038/s41586-018-0433-3
47. Yeo KS. **JMJD8 is a positive regulator of TNF-induced NF-κB signaling**. *Sci. Rep.* (2016.0) **6** 34125. DOI: 10.1038/srep34125
48. You D, Jung BC, Villivalam SD, Lim H-W, Kang S. **JMJD8 is a novel molecular nexus between adipocyte-intrinsic inflammation and insulin resistance**. *Diabetes* (2021.0) **71** 43-59. DOI: 10.2337/db21-0596
49. Kahles F. **GLP-1 secretion is increased by inflammatory stimuli in an IL-6-dependent manner, leading to hyperinsulinemia and blood glucose lowering**. *Diabetes* (2014.0) **63** 3221-3229. DOI: 10.2337/db14-0100
50. Labzin LI. **ATF3 is a key regulator of macrophage IFN responses**. *J. Immunol.* (2015.0) **195** 4446-4455. DOI: 10.4049/jimmunol.1500204
51. Benson AK. **Individuality in gut microbiota composition is a complex polygenic trait shaped by multiple environmental and host genetic factors**. *Proc. Natl Acad. Sci. USA* (2010.0) **107** 18933-18938. DOI: 10.1073/pnas.1007028107
52. 52.Leamy, L. J. et al. Host genetics and diet, but not immunoglobulin A expression, converge to shape compositional features of the gut microbiome in an advanced intercross population of mice. Genome Biol.15, 552 (2014).
53. Keller MP. **Genetic drivers of pancreatic islet function**. *Genetics* (2018.0) **209** 335-356. DOI: 10.1534/genetics.118.300864
54. Dees C, Shively JM. **Localization of quantitation of the ornithine lipid of**. *J. Bacteriol.* (1982.0) **149** 798-799. DOI: 10.1128/jb.149.2.798-799.1982
55. Vences-Guzmán MÁ. **Discovery of a bifunctional acyltransferase responsible for ornithine lipid synthesis in**. *Environ. Microbiol.* (2015.0) **17** 1487-1496. DOI: 10.1111/1462-2920.12562
56. Kawai Y, Yano I, Kaneda K. **Various kinds of lipoamino acids including a novel serine-containing lipid in an opportunistic pathogen**. *Eur. J. Biochem.* (1988.0) **171** 73-80. DOI: 10.1111/j.1432-1033.1988.tb13760.x
57. Kawai Y, Kaneda K, Morisawa Y, Akagawa K. **Protection of mice from lethal endotoxemia by use of an ornithine-containing lipid or a serine-containing lipid**. *Infect. Immun.* (1991.0) **59** 2560-2566. DOI: 10.1128/iai.59.8.2560-2566.1991
58. Kawai Y, Akagawa K. **Macrophage activation by an ornithine-containing lipid or a serine-containing lipid**. *I* (1989.0) **57** 2086-2091
59. Peri F, Piazza M, Calabrese V, Damore G, Cighetti R. **Exploring the LPS/TLR4 signal pathway with small molecules**. *Biochem. Soc. Trans.* (2010.0) **38** 1390-1395. DOI: 10.1042/BST0381390
60. Piazza M. **Glycolipids and benzylammonium lipids as novel antisepsis agents: synthesis and biological characterization**. *J. Med. Chem.* (2009.0) **52** 1209-1213. DOI: 10.1021/jm801333m
61. Ryzhakov G. **Alpha kinase 1 controls intestinal inflammation by suppressing the IL-12/Th1 axis**. *Nat. Commun.* (2018.0) **9** 3797. DOI: 10.1038/s41467-018-06085-5
62. Khuu CH, Barrozo RM, Hai T, Weinstein SL. **Activating transcription factor 3 (ATF3) represses the expression of CCL4 in murine macrophages**. *Mol. Immunol.* (2007.0) **44** 1598-1605. DOI: 10.1016/j.molimm.2006.08.006
63. Cao Y. **Critical role of intestinal microbiota in ATF3-mediated gut immune homeostasis**. *J. Immunol.* (2020.0) **205** 842-852. DOI: 10.4049/jimmunol.1901000
64. Du Y. **ATF3 positively regulates antibacterial immunity by modulating macrophage killing and migration functions**. *Front. Immunol.* (2022.0) **13** 839502. DOI: 10.3389/fimmu.2022.839502
65. Keller MP. **Gene loci associated with insulin secretion in islets from non-diabetic mice**. *J. Clin. Invest.* (2019.0) **129** 4419-4432. DOI: 10.1172/JCI129143
66. Linke V. **A large-scale genome–lipid association map guides lipid identification**. *Nat. Metab.* (2020.0) **2** 1149-1162. DOI: 10.1038/s42255-020-00278-3
67. Turnbaugh PJ. **A core gut microbiome in obese and lean twins**. *Nature* (2009.0) **457** 480-484. DOI: 10.1038/nature07540
68. Faith JJ, McNulty NP, Rey FE, Gordon JI. **Predicting a human gut microbiota’s response to diet in gnotobiotic mice**. *Science* (2011.0) **333** 101-104. DOI: 10.1126/science.1206025
69. Langmead B, Salzberg SL. **Fast gapped-read alignment with Bowtie 2**. *Nat. Methods* (2012.0) **9** 357-359. DOI: 10.1038/nmeth.1923
70. Nurk S, Meleshko D, Korobeynikov A, Pevzner P. **metaSPAdes: a new versatile metagenomic assembler**. *Genome Res.* (2017.0) **27** 824-834. DOI: 10.1101/gr.213959.116
71. Hyatt D. **Prodigal: prokaryotic gene recognition and translation initiation site identification**. *BMC Bioinformatics* (2010.0) **11** 119. DOI: 10.1186/1471-2105-11-119
72. Li W, Godzik A. **Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences**. *Bioinformatics* (2006.0) **22** 1658-1659. DOI: 10.1093/bioinformatics/btl158
73. Buchfink B, Xie C, Huson DH. **Fast and sensitive protein alignment using DIAMOND**. *Nat. Methods* (2015.0) **12** 59-60. DOI: 10.1038/nmeth.3176
74. 74.Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-seq data with or without a reference genome. BMC Bioinformatics12, 323 (2011).
75. Miller IJ. **Autometa: automated extraction of microbial genomes from individual shotgun metagenomes**. *Nucleic Acids Res.* (2019.0) **47** e57. DOI: 10.1093/nar/gkz148
76. Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW. **CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes**. *Genome Res.* (2015.0) **25** 1043-1055. DOI: 10.1101/gr.186072.114
77. Ondov BD. **Mash: fast genome and metagenome distance estimation using MinHash**. *Genome Biol.* (2016.0) **17** 132. DOI: 10.1186/s13059-016-0997-x
78. Broman KW. **R/qtl2: software for mapping quantitative trait loci with high-dimensional data and multiparent populations**. *Genetics* (2019.0) **211** 495-502. DOI: 10.1534/genetics.118.301595
79. Chick JM. **Defining the consequences of genetic variation on a proteome-wide scale**. *Nature* (2016.0) **534** 500-505. DOI: 10.1038/nature18270
80. Ashrafian F. *Front. Microbiol.* (2019.0) **10** 2155. DOI: 10.3389/fmicb.2019.02155
81. Hutchins PD, Russell JD, Coon JJ. **LipiDex: an integrated software package for high-confidence lipid identification**. *Cell Syst.* (2018.0) **6** 621-625.e5. DOI: 10.1016/j.cels.2018.03.011
82. Bolger AM, Lohse M, Usadel B. **Trimmomatic: a flexible trimmer for Illumina sequence data**. *Bioinformatics* (2014.0) **30** 2114-2120. DOI: 10.1093/bioinformatics/btu170
83. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol.* (2014.0) **15** 550. DOI: 10.1186/s13059-014-0550-8
84. Collado MC, Derrien M, Isolauri E, de Vos WM, Salminen S. **Intestinal integrity and**. *Appl. Environ. Microbiol.* (2007.0) **73** 7767-7770. DOI: 10.1128/AEM.01477-07
|
---
title: 'Early skeletal muscle mass decline is a prognostic factor in patients receiving
gemcitabine plus nab-paclitaxel for unresectable pancreatic cancer: a retrospective
observational study'
authors:
- Yukari Suzuki
- Kei Saito
- Yousuke Nakai
- Hiroki Oyama
- Sachiko Kanai
- Tatsunori Suzuki
- Tatsuya Sato
- Ryunosuke Hakuta
- Kazunaga Ishigaki
- Tomotaka Saito
- Tsuyoshi Hamada
- Naminatsu Takahara
- Ryosuke Tateishi
- Mitsuhiro Fujishiro
journal: Supportive Care in Cancer
year: 2023
pmcid: PMC9981495
doi: 10.1007/s00520-023-07659-w
license: CC BY 4.0
---
# Early skeletal muscle mass decline is a prognostic factor in patients receiving gemcitabine plus nab-paclitaxel for unresectable pancreatic cancer: a retrospective observational study
## Abstract
### Purpose
Patients with pancreatic cancer often have cancer cachexia at diagnosis. Recent studies suggested that loss of skeletal muscle mass was related to cancer cachexia, which hindered continuance of chemotherapy and could be one of prognostic factors in pancreatic cancer, however the association remains unclear in patients receiving gemcitabine and nab-paclitaxel (GnP).
### Methods
We retrospectively studied 138 patients with unresectable pancreatic cancer receiving first-line GnP at the University of Tokyo from January 2015 to September 2020. We calculated body composition in CT images before chemotherapy and at initial evaluation, and evaluated the association of both body composition before chemotherapy and its changes at initial evaluation.
### Results
Compared by skeletal muscle mass index (SMI) change rate between pre-chemotherapy and initial evaluation, there were statistically significantly differences in the median OS: 16.3 months ($95\%$CI 12.3–22.7) and 10.3 months ($95\%$CI 8.3–18.1) between SMI change rate ≥ -$3.5\%$ and < -$3.5\%$ groups ($$P \leq 0.01$$). By multivariate analysis for OS, CA19-9 (HR 3.34, $95\%$CI 2.00–5.57, $P \leq 0.01$), PLR (HR 1.68, $95\%$CI 1.01–2.78, $$P \leq 0.04$$), mGPS (HR 2.32, $95\%$CI 1.47–3.65, $P \leq 0.01$) and relative dose intensity (HR 2.21, $95\%$CI 1.42–3.46, $P \leq 0.01$) were significantly poor prognostic factors. SMI change rate (HR 1.47, $95\%$CI 0.95–2.28, $$P \leq 0.08$$) showed a trend to poor prognosis. Sarcopenia before chemotherapy was not significantly associated with PFS or OS.
### Conclusion
Early skeletal muscle mass decline was associated with poor OS. Further investigation is warranted whether the maintenance of skeletal muscle mass by nutritional support would improve prognosis.
## Introduction
The incidence of pancreatic cancer (PC) is increasing worldwide [1]. Despite surgery being the only curative treatment, 80–$85\%$ of patients present with an advanced stage [2, 3]. Immunotherapy has been investigated as one of treatment options, but systemic cytotoxic chemotherapy is still the standard of care for locally advanced or metastatic PC, including gemcitabine plus nab-paclitaxel (GnP) [4], and FOLFIRINOX (5-fluorouracil, leucovorin, irinotecan, and oxaliplatin) [5]. Despite the improvement of survival by those intense combination regimens, they are associated with adverse effects (AEs) and require appropriate patient selection.
Patients with PC, especially elderly patients, are often underweight and undernourished at diagnosis, with $50\%$ reported to have cancer cachexia at diagnosis [6, 7]. Cancer cachexia is defined as a multifactorial syndrome defined by an ongoing loss of skeletal muscle mass that cannot be fully reversed by conventional nutritional support and leads to progressive functional impairment [8]. Recent studies suggested that loss of skeletal muscle mass was associated with cancer cachexia, which hindered continuance of chemotherapy, and can be one of prognostic factors of survival in PC [9–13]. However, it remains unclear whether sarcopenia at diagnosis or decline in skeletal muscle mass during chemotherapy is more prognostic of survival in PC, with various regimens such as FOLFIRINOX [14, 15] and GnP [16] being evaluated.
In this retrospective study, we investigated the association of both body composition before chemotherapy and its changes at initial evaluation of chemotherapy in patients receiving first-line GnP for unresectable PC.
## Patients
Data on patients with unresectable PC who started GnP as first-line chemotherapy at the Department of Gastroenterology, the University of Tokyo from January 2015 to September 2020 were retrospectively studied. The analysis was based on follow-up information, which was received until April 2022. This study was approved by the ethics committee of the University of Tokyo Hospital.
All patients were histologically or cytologically diagnosed as pancreatic ductal adenocarcinoma and were diagnosed as locally advanced or metastatic diseases on CT. Chemotherapy was administered on days 1, 8 and 15 of a 28-day cycle, combined gemcitabine at 1000 mg/m2 and nab-paclitaxel at 125 mg/m2 [17].
## Data collection
We extracted data, including age, sex, height, weight, Eastern Cooperative Oncology Group performance status (ECOG PS), laboratory data (white blood cell with differential, albumin, C-reactive protein, carcinoembryonic antigen and carbohydrate antigen 19–9 [CA19-9]) from our prospectively maintained pancreatic cancer database and electric medical records in our hospital.
Body Mass Index (BMI) was calculated by dividing the weight (kg) by the square of the height (m), and the cutoff value was set at 22, the standard value in Japan. Neutrophil/lymphocyte ratio (NLR), Platelet/lymphocyte ratio (PLR) and modified Glasgow Prognostic Score (mGPS) were calculated from the above-mentioned data. The cutoff value of NLR and PLR was set by creating receiver operating characteristic (ROC) curve with a dichotomous variable divided by median overall survival (353.5 days) as the dependent variable.
In addition, we evaluated relative dose intensity (RDI) up to first 2 cycles, early tumor shrinkage (ETS) and presence of dose reduction at 1st cycle to analyze prognostic factors. RDI was calculated by dividing the actual dose by the standard dose of gemcitabine and nab-paclitaxel up to first 2 cycles, the cutoff value was set at the median. The standard dose was set at 125 mg/m2 for nab-paclitaxel and 1000 mg/m2 for gemcitabine, based on the results of the phase 3 study with metastatic pancreatic cancer [4, 17]. ETS was calculated from the maximum tumor diameter before chemotherapy and at initial evaluation according to RECIST 1.1, the cutoff value was set at $20\%$ [18–20].
## Body composition assessment
We calculated the skeletal muscle mass area (cm2), subcutaneous fat area (cm2) and visceral fat area (cm2) at the level of the third lumbar vertebra in CT images before chemotherapy introduction and at initial evaluation by using SliceOmatic medical imaging software (Tomovision, Canada) [21]. The ranges of tissue Hounsfield unit (HU) thresholds were within -29 to 150 HU for skeletal muscle mass area, -190 to -30 HU for subcutaneous fat area, and -150 to -50 HU for visceral fat area, as shown Fig. 1 [22]. Skeletal muscle area, subcutaneous fat area, and visceral fat area were normalized for height in meters squared (m2) and reported as skeletal muscle mass index (SMI) (cm2/m2), subcutaneous adipose tissue index (SATI) (cm2/m2), and visceral adipose tissue index (VATI) (cm2/m2). Visceral-to-subcutaneous fat area ratio (VSR) was calculated by dividing visceral fat area by subcutaneous fat area to assess for the presence of visceral obesity. SMI change rate (%) was calculated by subtracting SMI before chemotherapy from SMI at initial evaluation, and dividing by SMI before chemotherapy, and standardizing at 60 days. Fig.1Assessment of body composition. The image illustrates the different proportions of skeletal muscle area (red), subcutaneous fat area (turquoise), and visceral fat area (yellow). Skeletal muscle area (cm2), subcutaneous fat area (cm2), and visceral fat area (cm2) at the level of the third lumbar vertebra in CT scan were quantified by using SliceOmatic medical imaging software. Skeletal muscle area highlighted red was quantified within -29 to 150 HU, subcutaneous fat area highlighted turquoise was quantified within -190 to -30 HU, and visceral fat area highlighted yellow was quantified within -150 to -50 HU Sarcopenia was defined as male SMI < 42 cm2/m2 and female SMI < 38 cm2/m2 based on the criteria proposed by the Hepatology Society of Japan [23, 24]. The cutoff values for VSR and SMI change rate were set by creating the ROC curve with a dichotomous variable divided by median overall survival as the dependent variable.
## Statistical analysis
We investigated the association of sarcopenia and changes in body composition during chemotherapy with progression free survival (PFS), overall survival (OS), AEs, and tumor response including response rate (RR) and disease control rate (DCR).
Both PFS and OS were calculated starting from the CT date of initial evaluation. PFS and OS were estimated using Kaplan–Meier method and survival curves were compared using log-rank test. Comparisons between two groups were evaluated using the Mann–Whitney U test for continuous variables and using the Fisher’s exact test for categorical data. AEs were evaluated according to CTCAE ver 4.0. Hazard ratios (HRs) with $95\%$ confidence intervals (CIs) for OS and PFS were estimated by a Cox proportional hazards model to determine the independent prognostic factors. Factors with p-values < 0.20 in the univariable analyses were evaluated in the multivariable analyses. All tests were 2-sided, and p-value < 0.05 was considered statistically significant. Statistical analyses were performed using JMP version 16 software (SAS Institute Inc., Cary, NC).
## Patient characteristics
Between January 2015 and September 2020, 152 patients started GnP as first-line chemotherapy, and pre-chemotherapy and initial evaluation CT scans were available in 138 patients. Fourteen patients who did not receive follow up CT evaluation were excluded from the analysis (Fig. 2). The median interval between pre-chemotherapy and initial evaluation CT scan was 61.5 days (interquartile range (IQR), 54–70). The median follow-up period was 13.1 months (IQR, 8.9–22.5). Baseline characteristics at chemotherapy introduction are summarized in Table 1. Median age was 67.5 years old (IQR, 59.7–74) and 80 patients ($58.0\%$) were male. Distant metastasis was present in 97 patients ($70.3\%$); liver in 62 ($44.9\%$), lung in 20 ($14.5\%$), lymph nodes in 38 ($27.5\%$) and peritoneal dissemination in 21 ($15.2\%$). ECOG PS was 0 in 76 patients ($55.1\%$). The median SMI, VATI, SATI, VSR were 40.9cm2/m2 (IQR, 35.8–46.9), 31.6 cm2/m2 (IQR, 14.2–48.2), 33.7cm2/m2 (IQR, 21.8–47.1) and 0.91 (IQR, 0.49–1.39), respectively (Table 2). Sixty-one patients ($44.2\%$) were diagnosed as sarcopenia. The median PFS was 6.5 months ($95\%$CI 5.1–8.2) and the median OS was 15.2 months ($95\%$CI 11.2–19.0).Fig. 2Patient flowchart. GnP; gemcitabine and nab-paclitaxelTable 1Patient CharacteristicsTotal cohort ($$n = 138$$)SMI change rate ≥ -$3.5\%$ ($$n = 84$$)SMI change rate < -$3.5\%$ ($$n = 54$$)p-valueAge67.5 (59.7–74)69 (61.2–74.7)66 (59–72.2)0.12Age ≥ 75 years old29 (21.0)21 (25.0)8 (14.8)0.14Male sex80 (58.0)43 (51.2)37 (68.5)0.04ECOG *Performance status* $\frac{0}{1}$/$\frac{276}{61}$/1 ($\frac{55.1}{44.2}$/0.7)$\frac{47}{36}$/1 ($\frac{56.0}{42.9}$/1.1)$\frac{29}{25}$/0 ($\frac{53.7}{46.3}$/0)0.57BMI, kg/m221.5 (19.5–23.6)21.3 (19.2–23.6)21.8 (20.5–23.7)0.51Metastasis97 (70.3)62 (73.8)35 (64.8)0.26Liver62 (44.9)41 (48.8)21 (38.9)0.25Lung20 (14.5)14 (16.7)6 (11.1)0.35Lymph node38 (27.5)28 (33.3)10 (18.5)0.05Peritoneal dissemination21 (15.2)10 (11.9)11 (20.4)0.18Biliary drainage before chemotherapy31 (22.5)11 (13.1)20 (37.0) < 0.01CA19-9, U/ml678.5 (105.2–4098)531 (55.7–3868.7)1064.5 (200.2–4546.2)0.33NLR2.9 (2.1–4.1)2.8 (2–3.9)3 (2.4–4.7)0.31PLR176.9 (128.9–246.6)164.2 (128.3–229.0)192.8 (134.9–256.5)0.40mGPS, 061 (44.9)36 (43.9)25 (46.3)0.78CCI, ≥ 315 (10.9)8 (9.5)7 (13.0)0.52SMI, cm2/m240.9 (35.8–46.9)39.8 (35.2–45.2)43.8 (36.3–50.0)0.02VATI, cm2/m231.6 (14.2–48.2)31.6 (13.4–46.1)31.3 (14.6–55.4)0.08SATI, cm2/m233.7 (21.8–47.1)33.5 (21.1–45.2)34.3 (24.9–48.6)0.93VSR0.91 (0.49–1.39)0.81 (0.48–1.36)0.97 (0.51–1.45)0.40Sarcopenia*61 (44.2)40 (47.6)21 (38.9)0.31Interval between pretreatment and initial evaluation CT61.5 (24–109)62 (55–71)60 (52–69.2)0.11Numbers are shown in n (%) or median (interquartile range [IQR]). * Defined as male SMI < 42 cm2/m2 and female SMI < 38 cm2/m2 based on the criteria proposed by the Hepatology Society of JapanBMI Body mass index, CCI Charlson comorbidity index, ECOG Eastern Cooperative Oncology Group, mGPS modified Glasgow Prognostic Score, NLR Neutrophil/lymphocyte ratio, PLR Platelet/lymphocyte ratio, SATI subcutaneous adipose tissue index, SMI skeletal muscle mass index, VATI visceral adipose tissue index, VSR visceral-to-subcutaneous fat area ratioTable 2Body composition according to the SMI change rateTotal ($$n = 138$$)SMI change rate ≥ -$3.5\%$ ($$n = 84$$)SMI change rate < -$3.5\%$ ($$n = 54$$)p-valueSMI change rate, %-2.1 (-6.5–2.1)0.55 (-1.7–4.9)-8.3 (-12.5- -5.6) < 0.01SMI before chemotherapy, cm2/m240.9 (35.8–46.9)39.8 (35.2–45.2)43.8 (36.3–50.0)0.02SMI at initial evaluation, cm2/m240.4 (35.2–45.2)41.0 (36.1–45.3)39 (32.5–44.1)0.11VATI before chemotherapy, cm2/m231.6 (14.2–48.2)31.6 (13.4–46.1)31.3 (14.6–55.4)0.08VATI at initial evaluation, cm2/m223.1 (11.2–43.2)26.6 (10.7–43.8)21.9 (11.2–43.2)0.76SATI before chemotherapy, cm2/m233.7 (21.8–47.1)33.5 (21.1–45.2)34.3 (24.9–48.6)0.93SATI at initial evaluation, cm2/m228.7 (17.3–41.3)28.3 (18.8–43.1)29.3 (15.2–40.8)0.11VSR before chemotherapy0.91 (0.49–1.39)0.81 (0.48–1.36)0.97 (0.51–1.45)0.40VSR at initial evaluation0.85 (0.56–1.30)0.83 (0.53–1.27)0.87 (0.56–1.35)0.81Sarcopenia* before chemotherapy61 (44.2)40 (47.6)21 (38.9)0.31Sarcopenia* at initial evaluation66 (47.8)34 (40.5)32 (59.3)0.03Numbers are shown in n (%) or median (interquartile range [IQR]). * Defined as male SMI < 42 cm2/m2 and female SMI < 38 cm2/m2 based on the criteria proposed by the Hepatology Society of JapanBMI Body mass index, CCI Charlson comorbidity index, ECOG Eastern Cooperative Oncology Group, mGPS modified Glasgow Prognostic Score, SATI subcutaneous adipose tissue index, SMI Skeletal muscle mass index, VATI visceral adipose tissue index, VSR Visceral-to-subcutaneous fat area ratio
## SMI change rate and clinical outcomes
By creating ROC curve with a dichotomous variable divided by the median OS as the dependent variable, the cutoff value for SMI change rate was set at -$3.5\%$ (Fig. 3). The median OS in the total cohort was 16.3 months ($95\%$CI 12.3–22.7) and 10.3 months ($95\%$CI 8.3–18.1) in SMI change rate ≥ -$3.5\%$ and < -$3.5\%$ groups ($$P \leq 0.01$$, Fig. 4A).Fig. 3Receiver operating characteristic curve with SMI change rateFig. 4Overall survival according to the SMI change rate. Solid lines indicate SMI change rate < -$3.5\%$ and broken lines indicate SMI change rate ≥ -$3.5\%$. A. Overall survival of the total cohort. The median overall survival was 10.3 months ($95\%$CI, 8.3–18.1) for SMI change rate < -$3.5\%$ and 16.3 months ($95\%$CI, 12.3–22.7) for SMI change rate ≥ -$3.5\%$ ($$P \leq 0.01$$). B. Overall survival in non-elderly (< 75 years old) patients. The median overall survival was 11.8 months ($95\%$CI, 8.2–19.0) for SMI change rate < -$3.5\%$ and 15.8 months ($95\%$CI, 11.2–22.7) for SMI change rate ≥ -$3.5\%$ ($$P \leq 0.07$$). C. Overall survival in elderly (≥ 75 years old) patients. The median overall survival was 9.5 months ($95\%$CI, 5.5–30.2) for SMI change rate < -$3.5\%$ and 16.5 months ($95\%$CI, 10.2–40.4) for SMI change rate ≥ -$3.5\%$ ($$P \leq 0.11$$) Patient characteristics divided by SMI change rate are shown in Table 1. The rates of male sex and biliary drainage were significantly higher in SMI change rate < -$3.5\%$ group. Body composition before chemotherapy and at the initial evaluation is shown in Table 2. The median SMI before chemotherapy was higher in SMI change rate < -$3.5\%$ group: 39.8 and 43.8 cm2/m2 ($$P \leq 0.02$$), but the difference was not significant at the initial evaluation. The rate of sarcopenia at the initial evaluation was significantly higher in SMI change rate < -$3.5\%$ group: $40.5\%$ and $59.3\%$ ($$P \leq 0.03$$).
There were no significant differences in objective response ($$P \leq 0.55$$): RR was $23.8\%$ and $16.7\%$ and DCR was $89.3\%$ and $81.5\%$ in SMI change rate ≥ -$3.5\%$ and < -$3.5\%$ groups (Table 3). The median PFS by SMI change rate in the total cohort was not significantly different: 7.2 months ($95\%$CI 5.3–9.1) and 5.4 months ($95\%$CI 3.7–8.7) in SMI change rate ≥ -$3.5\%$ and < -$3.5\%$ groups, respectively ($$P \leq 0.24$$, Fig. 5A).Table 3Treatment outcomes according to the SMI change rateTotal ($$n = 138$$)SMI change rate ≥ -$3.5\%$ ($$n = 84$$)SMI change rate < -$3.5\%$ ($$n = 54$$)p-valueNumber of cycles7.5 (4–12)8 (5–12.7)6 (2.7–10.2)0.01RDI for the first two cycles, %69.2 (56.7–83.3)66.7 (57.5–83.3)70 (56.7–83.3)0.45Dose reduction at 1st cycle90 (65.2)54 (64.3)36 (66.7)0.77ETS, %12 (0–22.7)13.6 (1.8–23.1)9.2 (0–21.4)0.44Best response0.55 Complete response1 (0.7)1 (1.2)0 Partial response28 (20.3)19 (22.6)9 (16.7) Stable disease90 (65.2)55 (65.5)35 (64.8) Progression disease13 (9.4)6 (7.1)7 (13.0) Not evaluable6 (4.4)3 (3.6)3 (5.5) Response rate, %21.023.816.70.30 Disease control rate, %86.289.381.50.19Reasons for discontinuation Disease progression95 (68.9)60 (71.4)35 (64.8)0.41 Serious adverse event17 (12.3)11 (13.1)6 (11.1)0.72 *Poor* general condition13 (9.4)6 (7.1)7 (13.0)0.25 Discontinue at initial evaluation21 (15.2)8 (9.5)13 (24.1)0.02 Introduction of 2nd line treatment105 (79.0)66 (80.5)39 (76.5)0.58Numbers are shown in n (%) or median (interquartile range [IQR]). ETS Early tumor shrinkage, RDI relative dose intensityFig. 5Progression free survival according to the SMI change rate. Solid lines indicate SMI change rate < -$3.5\%$ and broken lines indicate SMI change rate ≥ -$3.5\%$. A. Progression free survival of the total cohort. The median PFS was 5.4 months ($95\%$CI, 3.7–8.7) for SMI change rate < -$3.5\%$ and 7.2 months ($95\%$CI, 5.3–9.1) for SMI change rate ≥ -$3.5\%$ ($$p \leq 0.24$$). B. Progression free survival in non-elderly (< 75 years old) patients. The median progression free survival was 5.1 months ($95\%$CI, 3.5–8.7) for SMI change rate < -$3.5\%$ and 7.2 months ($95\%$CI, 5.1–9.8) for SMI change rate ≥ -$3.5\%$ ($$P \leq 0.16$$). C. Progression free survival in elderly (≥ 75 years old) patients. The median progression free survival was 6.3 months ($95\%$CI, 3.5-NA) for SMI change rate < -$3.5\%$ and 8.0 months ($95\%$CI, 4.0–9.6) for SMI change rate ≥ -$3.5\%$ ($$P \leq 0.66$$). CI; confidence interval, SMI; Skeletal muscle mass index In terms of safety, the incidences of AEs were comparable between two groups, other than all grades neutropenia (Table 4). However, SMI change rate < -$3.5\%$ group had experienced more discontinuations at initial evaluation ($$P \leq 0.02$$), and fewer total cycles of chemotherapy ($$P \leq 0.01$$) compared to SMI change rate ≥ -$3.5\%$ group. Table 4Adverse effects according to the SMI change rateAll gradesGrade ≥ 3SMI change rate ≥ -$3.5\%$SMI change rate < -$3.5\%$p-valueSMI change rate ≥ -$3.5\%$SMI change rate < -$3.5\%$p-valueHematologic Neutropenia75 (89.3)39 (72.2)0.0153 (63.1)30 (55.6)0.37 Thrombocytopenia49 (58.3)25 (46.3)0.204 (4.8)00.07 Anemia72 (85.7)45 (83.3)0.5212 (14.3)3 (5.6)0.14Non-hematologic Vomiting5 (6.0)1 (1.9)0.2800 Nausea18 (21.4)5 (9.3)0.082 (2.4)00.22 Anorexia32 (38.1)21 (38.9)0.602 (2.4)1 (1.9)0.59 Fatigue18 (21.4)13 (24.1)0.573 (3.6)00.13 Diarrhea12 (14.3)7 (13.0)0.581 (1.2)1 (1.9)0.57 Constipation35 (41.7)19 (35.2)0.4300 Peripheral neuropathy46 (54.8)21 (38.9)0.103 (3.6)3 (5.6)0.52Numbers are shown in n (%)
## Prognostic factors for PFS and OS
The results of univariable and multivariable analyses of PFS and OS are shown in Tables 5A, B. In the multivariable analysis, CA19-9 (HR 2.12, $95\%$ CI 1.34–3.36, $P \leq 0.01$) and mGPS (HR 1.58, $95\%$ CI 1.02–2.44, $$P \leq 0.03$$) were significant prognostic factors for PFS. Meanwhile, CA19-9 (HR 3.34, $95\%$ CI 2.00–5.57, $P \leq 0.01$), PLR (HR 1.68, $95\%$CI 1.01–2.78, $$P \leq 0.04$$), mGPS (HR 2.32, $95\%$CI 1.47–3.65, $P \leq 0.01$) and RDI up to 2 cycles (HR 2.21, $95\%$CI 1.42–3.46, $P \leq 0.01$) were significantly prognostic factors for OS. SMI change rate (HR 1.47, $95\%$CI 0.95–2.28, $$P \leq 0.08$$) and ETS (HR 1.53, $95\%$CI 0.94–2.49, $$P \leq 0.08$$) was also associated with OS, though statistically not significant. Neither sarcopenia before chemotherapy nor sarcopenia at initial evaluation was significantly associated with PFS or OS.Table 5Prognostic factors for progression free survival and overall survival Univariable analysisMultivariable analysisHR ($95\%$ CI)p-valueHR ($95\%$ CI)p-value5A. Progression free survival Age ≥ 75y0.93 (0.56–1.54)0.79 Male Sex0.86 (0.57–1.30)0.49 ECOG *Performance status* 1, 21.00 (0.66–1.51)0.98 BMI < 221.10 (0.73–1.65)0.64 Metastatic disease1.52 (0.96–2.40)0.071.20 (0.74–1.96)0.45 CA19-9 ≥ 500 U/ml2.37 (1.54–3.65) < 0.012.12 (1.34–3.36) < 0.01 NLR ≥ 3.21.41 (0.93–2.13)0.091.20 (0.75–1.92)0.43 PLR ≥ 1951.61 (1.05–2.48)0.021.25 (0.77–2.04)0.36 mGPS 1, 21.72 (1.12–2.63)0.011.58 (1.02–2.44)0.03 CCI ≥ 31.52 (0.77–2.98)0.21 Biliary drainage before chemotherapy, Yes1.01 (0.61–1.67)0.94 Sarcopenia* before chemotherapy, Yes1.12 (0.74–1.69)0.59 VSR before chemotherapy, Male ≥ 1.26, Female ≥ 0.520.86 (0.57–1.30)0.48 Sarcopenia at initial evaluation, Yes1.08 (0.72–1.63)0.68 VSR at initial evaluation, Male ≥ 1.29, Female ≥ 0.561.23 (0.81–1.85)0.32 SMI change rate < -$3.5\%$1.28 (0.84–1.96)0.24 RDI < $69.2\%$1.20 (0.79–1.81)0.37 Dose reduction at 1st cycle, Yes0.85 (0.55–1.30)0.45 ETS < $20\%$1.22 (0.78–1.92)0.365B. Overall survival Age ≥ 75y0.92 (0.55–1.55)0.77 Male Sex1.06 (0.71–1.60)0.74 ECOG *Performance status* 1, 21.02 (0.68–1.54)0.89 BMI < 221.05 (0.70–1.57)0.79 Metastatic disease1.69 (1.07–2.68)0.021.24 (0.76–2.02)0.36 CA19-9 ≥ 500 U/ml3.62 (2.22–5.88) < 0.013.34 (2.00–5.57) < 0.01 NLR ≥ 3.21.58 (1.05–2.37)0.021.25 (0.76–2.05)0.36 PLR ≥ 1952.22 (1.44–3.41) < 0.011.68 (1.01–2.78)0.04 mGPS 1, 21.98 (1.30–3.02) < 0.012.32 (1.47–3.65) < 0.01 CCI ≥ 31.14 (0.57–2.30)0.69 Biliary drainage before chemotherapy, Yes1.27 (0.78–2.04)0.32 Sarcopenia* before chemotherapy, Yes1.13 (0.75–1.69)0.54 VSR before chemotherapy, Male ≥ 1.26, Female ≥ 0.520.81 (0.54–1.22)0.33 Sarcopenia at initial evaluation, Yes1.29 (0.85–1.94)0.22 VSR at initial evaluation, Male ≥ 1.29, Female ≥ 0.560.95 (0.64–1.42)0.83 SMI change rate < -$3.5\%$1.64 (1.08–2.52)0.021.47 (0.95–2.28)0.08 RDI < $69.2\%$1.52 (1.00–2.31)0.042.21 (1.42–3.46) < 0.01 Dose reduction at 1st cycle, Yes1.19 (0.77–1.84)0.41 ETS < $20\%$1.42 (0.90–2.23)0.121.53 (0.94–2.49)0.08BMI Body mass index, CCI Charlson comorbidity index, CI confidence interval, ECOG Eastern Cooperative Oncology Group, ETS early tumor shrinkage, mGPS modified Glasgow Prognostic Score, HR hazard ratio, NLR Neutrophil/lymphocyte ratio, PLR Platelet/lymphocyte ratio, RDI Relative dose intensity, SMI skeletal muscle mass index, VSR visceral-to-subcutaneous fat area ratio*Defined as male SMI < 42 cm2/m2 and female SMI < 38 cm2/m2 based on the criteria proposed by the Hepatology Society of Japan
## Exploratory analyses of body composition by age
Twenty-nine patients ($21.0\%$) were ≥ 75 years old in our cohort. There were no significant differences in RR ($19.3\%$ vs. $27.6\%$, $$P \leq 0.34$$), the median PFS (6.3 vs. 7.1 months, $$P \leq 0.79$$) and the median OS (14.1 vs. 16.3 months, $$P \leq 0.77$$) between non-elderly (< 75 years old) and elderly (≥ 75 years old) patients. When body composition was compared between non-elderly and elderly patients (Table 6), VATI both before chemotherapy and at initial evaluation was significantly higher in elderly patients. There were no significant differences in sarcopenia ($44.0\%$ and $44.8\%$, $$P \leq 0.93$$) or SMI change rates (-$2.4\%$ and -$1.8\%$, $$P \leq 0.23$$) in non-elderly and elderly patients. The median PFS was 5.1 and 7.2 months in SMI change rate < -$3.5\%$ and SMI change rate ≥ -$3.5\%$ groups in non-elderly patients ($$P \leq 0.16$$, Fig. 5B), while it was 6.3 and 8.0 months in SMI change rate < -$3.5\%$ and SMI change rate ≥ -$3.5\%$ groups in elderly patients ($$P \leq 0.66$$, Fig. 5C). SMI change rate was associated with OS, though not statistically significant. While the median OS was 11.8 and 15.8 months in SMI change rate < -$3.5\%$ and SMI change rate ≥ -$3.5\%$ groups in non-elderly patients ($$P \leq 0.07$$, Fig. 4B), it was 9.5 and 16.5 months for SMI change rate < -$3.5\%$ and SMI change rate ≥ -$3.5\%$ groups in elderly patients ($$P \leq 0.11$$, Fig. 4C).Table 6Body composition according to the age < 75 years old ($$n = 109$$) ≥ 75 years old ($$n = 29$$)p-valueSMI change rate, %-2.4 (-6.6–1.9)-1.8 (-6.4–7.1)0.23SMI before chemotherapy, cm2/m240.8 (35.5–47.5)42.3 (35.9–45.8)0.44SMI at initial evaluation, cm2/m240.0 (25.1–45.2)42.0 (35.2–45.1)0.85VATI before chemotherapy, cm2/m229.4 (12.1–43.6)45.0 (27.9–57.8) < 0.01VATI at initial evaluation, cm2/m221.5 (10.2–41.3)35.2 (18–52.5)0.03SATI before chemotherapy, cm2/m230.9 (19.6–49.3)37.4 (29.4–44.5)0.66SATI at initial evaluation, cm2/m226.5 (15.5–40.7)34.0 (26.9–43.4)0.49VSR before chemotherapy0.84 (0.48–1.32)1.13 (0.76–1.61)0.16VSR at initial evaluation0.81 (0.48–1.23)0.93 (0.71–1.44)0.27Sarcopenia* before chemotherapy48 (44.0)13 (44.8)0.93Sarcopenia* at initial evaluation54 (49.5)12 (41.4)0.43Numbers are shown in n (%) or median (interquartile range [IQR]). * Defined as male SMI < 42 cm2/m2 and female SMI < 38 cm2/m2 based on the criteria proposed by the Hepatology Society of JapanBMI Body mass index, CCI Charlson comorbidity index, ECOG Eastern Cooperative Oncology Group, mGPS modified Glasgow Prognostic Score, SATI subcutaneous adipose tissue index, SMI Skeletal muscle mass index, VATI visceral adipose tissue index, VSR Visceral-to-subcutaneous fat area ratio
## Discussion
In this retrospective study, we found that early skeletal muscle mass decline was associated with shorter OS in patients receiving first-line GnP for unresectable PC. On the other hands, sarcopenia before chemotherapy was not associated with OS. Our study results suggested early decline of SMI after introduction of chemotherapy rather than the value of SMI before chemotherapy might be prognostic of survival in patients with unresectable PC.
Sarcopenia as one of prognostic factors in patients with cancer is increasingly reported in various cancers. Recent studies suggested the role of sarcopenia in patients receiving palliative chemotherapy for PC. In our cohort, sarcopenia was observed in $44.2\%$ at the time of diagnosis, which was similar to that of previous reports [7, 16]. While some studies suggested association of sarcopenia at diagnosis with prognosis [12, 14, 25], others reported change in body composition was associated with survival [10–12, 26]. Sarcopenia before chemotherapy was not associated with PFS or OS in our cohort, by using the criteria developed by the Hepatology Society of Japan based on the AWGS criteria (male SMI < 42 cm2/m2 and female SMI < 38 cm2/m2) [23]. However, SMI change up to initial evaluation was associated with OS, suggesting body composition change can be predictive of prognosis in patients receiving palliative chemotherapy for PC. Interestingly, our definition of SMI decline > -$3.5\%$ was not associated with tumor response or PFS. The reason for discrepancy between PFS and OS is unclear but the similar outcomes were also observed in elderly patients receiving GnP chemotherapy [16].
In terms of safety, it was suggested that SMI change was not significantly associated with either AEs, other than all grades neutropenia, or RDI up to first 2 cycles of chemotherapy. Since reduced RDI was associated with poor survival, the maintenance of RDI is as important as the control of severe AEs, as previous studies reported the association of RDI with efficacy of FOLFIRINOX for PC [27, 28]. In our study, though 2-cycle RDI was comparable, discontinuation of chemotherapy after initial evaluation ($24.1\%$ and $9.5\%$) and discontinuation due to poor condition ($13.0\%$ and $7.1\%$) were more often encountered in SMI change rate ≥ -$3.5\%$ group compared to SMI change rate < -$3.5\%$ group. As a result, the number of cycles was higher in SMI change rate ≥ -$3.5\%$ group. Thus, sarcopenia during chemotherapy can lead to cessation of chemotherapy due to the deteriorated patient condition and non-chemotherapeutic support to prevent sarcopenia might improve clinical outcomes of palliative chemotherapy in PC.
Nutritional support has been increasingly investigated in the field of oncology. Anamorelin, an oral ghrelin-like agent, reportedly improved body weight and anorexia-related symptoms in cancer patients [29] and we also reported that insufficient protein intake was a poor prognostic factor in patients with unresectable PC receiving chemotherapy [30]. Nutritional interventions such as nutritional supplements [31] or pancreatic exocrine replacement treatment [32, 33] might also affect body composition. Thus, we should further investigate whether those nutritional interventions would improve sarcopenia during chemotherapy and lead to the improved prognosis or not.
Age itself can affect body composition and its impact on chemotherapy might differ by age. However, in our exploratory analyses, the associations of SMI change were comparable between elderly and non-elderly patients. The median OS tended to be longer in cases with SMI decline ≥ -$3.5\%$, regardless of age. Meanwhile, a previous study of pancreatic cancer receiving GnP chemotherapy reported that sarcopenia at diagnosis was associated with poor OS only in elderly (> 70 years old) patients [16]. We previously reported comorbidity, rather than age, was an important prognostic factors in gemcitabine-based chemotherapy [34]. Recently, the importance of cognitive assessment is also reported in elderly patients [35, 36]. The relation of age, comorbidity and body composition can be multifactorial and more comprehensive evaluation in a large prospective cohort is warranted.
Our study had several limitations. Firstly, this was a retrospective study at a single academic center and the selection bias was inevitable. For example, the rate of sarcopenia at diagnosis of PC was similar between elderly and non-elderly patients. Elderly patients who could receive GnP might be a selected population in a good clinical condition. Thus, our study results need to be validated in the external cohort. Secondly, definition of sarcopenia using CT scan have not been established. The AWGS 2019 definition uses grip strength, physical function (walking speed, 5 times stand up, short physical performance battery) and skeletal muscle mass measurement by dual energy X-ray absorptiometry or bioelectrical impedance analysis to determine sarcopenia [37]. We applied the criteria for sarcopenia by the Hepatology Society of Japan since this was retrospective study. Definition of sarcopenia in cases with malignancy including PC receiving palliative chemotherapy needs further investigation.
In conclusion, short-term decline of skeletal muscle mass was associated with poor OS in patients receiving GnP for unresectable PC. Further investigation is warranted whether the maintenance of skeletal muscle mass by nutritional support or medications would improve prognosis or not.
## References
1. Siegel RL, Miller KD, Fuchs HE, Jemal A. **Cancer Statistics, 2021**. *CA Cancer J Clin* (2021.0) **71** 7-33. DOI: 10.3322/caac.21654
2. Mizrahi JD, Surana R, Valle JW, Shroff RT. **Pancreatic cancer**. *Lancet* (2020.0) **395** 2008-2020. DOI: 10.1016/s0140-6736(20)30974-0
3. Siegel RL, Miller KD, Jemal A. **Cancer statistics, 2020**. *CA Cancer J Clin* (2020.0) **70** 7-30. DOI: 10.3322/caac.21590
4. Von Hoff DD, Ervin T, Arena FP, Chiorean EG, Infante J, Moore M. **Increased survival in pancreatic cancer with nab-paclitaxel plus gemcitabine**. *N Engl J Med* (2013.0) **369** 1691-1703. DOI: 10.1056/NEJMoa1304369
5. Conroy T, Desseigne F, Ychou M, Bouché O, Guimbaud R, Bécouarn Y. **FOLFIRINOX versus gemcitabine for metastatic pancreatic cancer**. *N Engl J Med* (2011.0) **364** 1817-1825. DOI: 10.1056/NEJMoa1011923
6. Mitsunaga S, Kasamatsu E, Machii K. **Incidence and frequency of cancer cachexia during chemotherapy for advanced pancreatic ductal adenocarcinoma**. *Support Care Cancer* (2020.0) **28** 5271-5279. DOI: 10.1007/s00520-020-05346-8
7. Takeda T, Sasaki T, Suzumori C, Mie T, Furukawa T, Yamada Y. **The impact of cachexia and sarcopenia in elderly pancreatic cancer patients receiving palliative chemotherapy**. *Int J Clin Oncol* (2021.0) **26** 1293-1303. DOI: 10.1007/s10147-021-01912-0
8. Fearon K, Strasser F, Anker SD, Bosaeus I, Bruera E, Fainsinger RL. **Definition and classification of cancer cachexia: an international consensus**. *Lancet Oncol* (2011.0) **12** 489-495. DOI: 10.1016/s1470-2045(10)70218-7
9. Martin L, Birdsell L, Macdonald N, Reiman T, Clandinin MT, McCargar LJ. **Cancer cachexia in the age of obesity: skeletal muscle depletion is a powerful prognostic factor, independent of body mass index**. *J Clin Oncol* (2013.0) **31** 1539-1547. DOI: 10.1200/jco.2012.45.2722
10. Salinas-Miranda E, Deniffel D, Dong X, Healy GM, Khalvati F, O'Kane GM. **Prognostic value of early changes in CT-measured body composition in patients receiving chemotherapy for unresectable pancreatic cancer**. *Eur Radiol* (2021.0) **31** 8662-8670. DOI: 10.1007/s00330-021-07899-6
11. Nakano O, Kawai H, Kobayashi T, Kohisa J, Ikarashi S, Hayashi K. **Rapid decline in visceral adipose tissue over 1 month is associated with poor prognosis in patients with unresectable pancreatic cancer**. *Cancer Med* (2021.0) **10** 4291-4301. DOI: 10.1002/cam4.3964
12. 12.Choi Y, Oh DY, Kim TY, Lee KH, Han SW, Im SA et al (2015) Skeletal Muscle Depletion Predicts the Prognosis of Patients with Advanced Pancreatic Cancer Undergoing Palliative Chemotherapy, Independent of Body Mass Index. PLoS One 10(10):e0139749. 10.1371/journal.pone.0139749
13. Griffin OM, Duggan SN, Ryan R, McDermott R, Geoghegan J, Conlon KC. **Characterising the impact of body composition change during neoadjuvant chemotherapy for pancreatic cancer**. *Pancreatology* (2019.0) **19** 850-857. DOI: 10.1016/j.pan.2019.07.039
14. Kurita Y, Kobayashi N, Tokuhisa M, Goto A, Kubota K, Endo I. **Sarcopenia is a reliable prognostic factor in patients with advanced pancreatic cancer receiving FOLFIRINOX chemotherapy**. *Pancreatology* (2019.0) **19** 127-135. DOI: 10.1016/j.pan.2018.11.001
15. Uemura S, Iwashita T, Ichikawa H, Iwasa Y, Mita N, Shiraki M. **The impact of sarcopenia and decrease in skeletal muscle mass in patients with advanced pancreatic cancer during FOLFIRINOX therapy**. *Br J Nutr* (2021.0) **125** 1140-1147. DOI: 10.1017/S0007114520003463
16. Asama H, Ueno M, Kobayashi S, Fukushima T, Kawano K, Sano Y. **Sarcopenia: prognostic value for unresectable pancreatic ductal adenocarcinoma patients treated with gemcitabine plus Nab-Paclitaxel**. *Pancreas* (2022.0) **51** 148-152. DOI: 10.1097/mpa.0000000000001985
17. Ueno H, Ikeda M, Ueno M, Mizuno N, Ioka T, Omuro Y. **Phase I/II study of nab-paclitaxel plus gemcitabine for chemotherapy-naive Japanese patients with metastatic pancreatic cancer**. *Cancer Chemother Pharmacol* (2016.0) **77** 595-603. DOI: 10.1007/s00280-016-2972-3
18. Cremolini C, Loupakis F, Antoniotti C, Lonardi S, Masi G, Salvatore L. **Early tumor shrinkage and depth of response predict long-term outcome in metastatic colorectal cancer patients treated with first-line chemotherapy plus bevacizumab: results from phase III TRIBE trial by the Gruppo Oncologico del Nord Ovest**. *Ann Oncol* (2015.0) **26** 1188-1194. DOI: 10.1093/annonc/mdv112
19. Heinemann V, Stintzing S, Modest DP, Giessen-Jung C, Michl M, Mansmann UR. **Early tumour shrinkage (ETS) and depth of response (DpR) in the treatment of patients with metastatic colorectal cancer (mCRC)**. *Eur J Cancer* (2015.0) **51** 1927-1936. DOI: 10.1016/j.ejca.2015.06.116
20. 20.Vivaldi C, Fornaro L, Cappelli C, Pecora I, Catanese S, Salani F et al (2019) Early Tumor Shrinkage and Depth of Response Evaluation in Metastatic Pancreatic Cancer Treated with First Line Chemotherapy: An Observational Retrospective Cohort Study. Cancers (Basel) 11(7). 10.3390/cancers11070939
21. Fujiwara N, Nakagawa H, Kudo Y, Tateishi R, Taguri M, Watadani T. **Sarcopenia, intramuscular fat deposition, and visceral adiposity independently predict the outcomes of hepatocellular carcinoma**. *J Hepatol* (2015.0) **63** 131-140. DOI: 10.1016/j.jhep.2015.02.031
22. 22.Mitsiopoulos N, Baumgartner RN, Heymsfield SB, Lyons W, Gallagher D, Ross R (1998) Cadaver validation of skeletal muscle measurement by magnetic resonance imaging and computerized tomography. J Appl Physiol (1985) 85(1):115–22. 10.1152/jappl.1998.85.1.115
23. Nishikawa H, Shiraki M, Hiramatsu A, Moriya K, Hino K, Nishiguchi S. **Japan Society of Hepatology guidelines for sarcopenia in liver disease (1st edition): Recommendation from the working group for creation of sarcopenia assessment criteria**. *Hepatol Res* (2016.0) **46** 951-963. DOI: 10.1111/hepr.12774
24. Chen LK, Liu LK, Woo J, Assantachai P, Auyeung TW, Bahyah KS. **Sarcopenia in Asia: consensus report of the Asian Working Group for Sarcopenia**. *J Am Med Dir Assoc* (2014.0) **15** 95-101. DOI: 10.1016/j.jamda.2013.11.025
25. 25.Naumann P, Eberlein J, Farnia B, Hackert T, Debus J, Combs SE (2019) Continued Weight Loss and Sarcopenia Predict Poor Outcomes in Locally Advanced Pancreatic Cancer Treated with Chemoradiation. Cancers (Basel) 11(5). 10.3390/cancers11050709
26. Basile D, Parnofiello A, Vitale MG, Cortiula F, Gerratana L, Fanotto V. **The IMPACT study: early loss of skeletal muscle mass in advanced pancreatic cancer patients**. *J Cachexia Sarcopenia Muscle* (2019.0) **10** 368-377. DOI: 10.1002/jcsm.12368
27. Lee JC, Kim JW, Ahn S, Kim HW, Lee J, Kim YH. **Optimal dose reduction of FOLFIRINOX for preserving tumour response in advanced pancreatic cancer: Using cumulative relative dose intensity**. *Eur J Cancer* (2017.0) **76** 125-133. DOI: 10.1016/j.ejca.2017.02.010
28. Vary A, Lebellec L, Di Fiore F, Penel N, Cheymol C, Rad E. **FOLFIRINOX relative dose intensity and disease control in advanced pancreatic adenocarcinoma**. *Ther Adv Med Oncol* (2021.0) **13** 17588359211029825. DOI: 10.1177/17588359211029825
29. Hamauchi S, Furuse J, Takano T, Munemoto Y, Furuya K, Baba H. **A multicenter, open-label, single-arm study of anamorelin (ONO-7643) in advanced gastrointestinal cancer patients with cancer cachexia**. *Cancer* (2019.0) **125** 4294-4302. DOI: 10.1002/cncr.32406
30. Hasegawa Y, Ijichi H, Saito K, Ishigaki K, Takami M, Sekine R. **Protein intake after the initiation of chemotherapy is an independent prognostic factor for overall survival in patients with unresectable pancreatic cancer: A prospective cohort study**. *Clin Nutr* (2021.0) **40** 4792-4798. DOI: 10.1016/j.clnu.2021.06.011
31. 31.Kim SH, Lee SM, Jeung HC, Lee IJ, Park JS, Song M et al (2019) The Effect of Nutrition Intervention with Oral Nutritional Supplements on Pancreatic and Bile Duct Cancer Patients Undergoing Chemotherapy. Nutrients 11(5). 10.3390/nu11051145
32. Saito T, Hirano K, Isayama H, Nakai Y, Saito K, Umefune G. **The Role of Pancreatic Enzyme Replacement Therapy in Unresectable Pancreatic Cancer: A Prospective Cohort Study**. *Pancreas* (2017.0) **46** 341-346. DOI: 10.1097/mpa.0000000000000767
33. Saito T, Nakai Y, Isayama H, Hirano K, Ishigaki K, Hakuta R. **A Multicenter Open-Label Randomized Controlled Trial of Pancreatic Enzyme Replacement Therapy in Unresectable Pancreatic Cancer**. *Pancreas* (2018.0) **47** 800-806. DOI: 10.1097/mpa.0000000000001079
34. Nakai Y, Isayama H, Sasaki T, Sasahira N, Tsujino T, Kogure H. **Comorbidity, not age, is prognostic in patients with advanced pancreatic cancer receiving gemcitabine-based chemotherapy**. *Crit Rev Oncol Hematol* (2011.0) **78** 252-259. DOI: 10.1016/j.critrevonc.2010.05.007
35. Wildiers H, Heeren P, Puts M, Topinkova E, Janssen-Heijnen ML, Extermann M. **International Society of Geriatric Oncology consensus on geriatric assessment in older patients with cancer**. *J Clin Oncol* (2014.0) **32** 2595-2603. DOI: 10.1200/jco.2013.54.8347
36. Hamaker ME, Te Molder M, Thielen N, van Munster BC, Schiphorst AH, van Huis LH. **The effect of a geriatric evaluation on treatment decisions and outcome for older cancer patients - A systematic review**. *J Geriatr Oncol* (2018.0) **9** 430-440. DOI: 10.1016/j.jgo.2018.03.014
37. Chen LK, Woo J, Assantachai P, Auyeung TW, Chou MY, Iijima K. **Asian Working Group for Sarcopenia: 2019 Consensus Update on Sarcopenia Diagnosis and Treatment**. *J Am Med Dir Assoc* (2020.0) **21** 300-7.e2. DOI: 10.1016/j.jamda.2019.12.012
|
---
title: A phase I first-in-man study to investigate the pharmacokinetics and safety
of liposomal dexamethasone in patients with progressive multiple myeloma
authors:
- Josbert Metselaar
- Twan Lammers
- Amelie Boquoi
- Roland Fenk
- Fabio Testaquadra
- Mirle Schemionek
- Fabian Kiessling
- Susanne Isfort
- Stefan Wilop
- Martina Crysandt
journal: Drug Delivery and Translational Research
year: 2023
pmcid: PMC9981510
doi: 10.1007/s13346-022-01268-6
license: CC BY 4.0
---
# A phase I first-in-man study to investigate the pharmacokinetics and safety of liposomal dexamethasone in patients with progressive multiple myeloma
## Abstract
Despite the introduction of multiple new drugs and combination therapies, conventional dexamethasone remains a cornerstone in the treatment of multiple myeloma (MM). Its application is, however, limited by frequent adverse effects of which the increased infection rate may have the strongest clinical impact. The efficacy-safety ratio of dexamethasone in MM may be increased by encapsulation in long-circulating PEG-liposomes, thereby both enhancing drug delivery to MM lesions and reducing systemic corticosteroid exposure. We evaluated the preliminary safety and feasibility of a single intravenous (i.v.) infusion of pegylated liposomal dexamethasone phosphate (Dex-PL) in heavily pretreated relapsing or progressive symptomatic MM patients within a phase I open-label non-comparative interventional trial at two dose levels. In the 7 patients that were enrolled (prior to having to close the study prematurely due to slow recruitment), Dex-PL was found to be well tolerated and, as compared to conventional dexamethasone, no new or unexpected adverse events were detected. Pharmacokinetic analysis showed high and persisting concentrations of dexamethasone in the circulation for over a week after i.v. administration, likely caused by the long-circulation half-life of the liposomes that retain dexamethasone as the inactive phosphate prodrug form, something which could significantly limit systemic exposure to the active parent drug. Thus, despite the limitations of this small first-in-man trial, Dex-PL seems safe and well tolerated without severe side effects. Follow-up studies are needed to confirm this in a larger patient cohort and to evaluate if i.v. Dex-PL can provide a safer and more efficacious dexamethasone treatment option for MM.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s13346-022-01268-6.
## Introduction
Multiple myeloma (MM) is a plasma cell malignancy that is characterized by the accumulation of clonal plasma cells, mainly in the bone marrow, leading to clinical manifestations such as anemia, bone destruction, and renal insufficiency [1]. Despite progress in our understanding of disease’s biology and the improvement in therapeutic options, therapy remains palliative [2, 3]. Almost all patients with multiple myeloma (MM), who respond to initial treatment, will eventually relapse and require further therapy [4]. Standard therapy comprises a combination of high dose chemotherapy, local radiation therapy, and, in the younger and fitter patients, autologous stem cell transplantation. Proteasome inhibitors, immunomodulatory drugs, monoclonal antibodies, corticosteroids, and alkylating agents are the most active agents [5]. Additionally, a variety of new drugs and drug classes and strategies (e.g., CAR T cell therapy) have changed the landscape significantly [6, 7].
Corticosteroids, and in particular dexamethasone, are an important mainstay in the treatment of MM [8–10]. In the last decades due to the previously rather limited number of alternative treatment options, dexamethasone was quite often used as a monotherapy for MM. A high dose oral treatment regimen of up to 160 mg per week divided over 4 daily doses has been popular albeit at the cost of side effects, including psychiatric, metabolic and diabetes-related complications, gastrointestinal disorders, hypertension, and especially life-threatening infections [11]. Considering the unfavorable side effect profile and the lower efficacy compared to the newer targeted treatment options available, dexamethasone monotherapy is nowadays rarely given. Rather, the drug is co-administered with the majority of current MM combination therapies. The standard dexamethasone dosage in combination settings involves a lower oral dose of weekly 20–40 mg [12], although doses of 320 mg within 4 weeks are still recommended in some treatment regimens [13]. Even though a number of new drugs have been approved for the treatment of MM especially in the last decade (such as daratumumab, carfilzomib, and pomalidomide), dexamethasone remains the backbone of almost every combination regimen [7].
Dexamethasone exerts its beneficial effect in MM in several ways. Besides acting on plasma cells directly at the level of growth inhibition and induction of apoptosis [14, 15], it also reduces the production of tumor-promoting interleukins (most notably IL-6) by osteoclasts and macrophages in the bone marrow, where plasma cells home and form the tumor lesion [16]. While it is obvious that these bone marrow lesions are the target site for agents like dexamethasone, a small amount of drug actually fully infiltrates these sites upon systemic administration, and high levels accumulate off-site in healthy tissues due to suboptimal pharmacokinetic properties of glucocorticoids, particularly the large volume of distribution and rapid clearance [17].
We postulate that the therapeutic value of dexamethasone in MM may be increased by utilizing the concept of nanomedicine-based drug delivery to MM bone marrow lesions. A graphical comparison of the advantage of this concept in comparison with free dexamethasone administration is given in Fig. 1. Presented in this figure is the most obvious drug delivery formulation for this purpose, namely, the pegylated liposome, which is a small-sized (100 nm) phospholipid bilayer vesicle enriched with cholesterol and coated with a layer of short-chained poly(ethylene glycol) (PEG). Cholesterol and PEG prevent premature removal of the phospholipid vesicles by the mononuclear phagocyte system (MPS) and impart so-called “long-circulating behavior” after i.v. infusion [18]. PEG-liposomes have three well-established properties that render them valuable in the treatment of MM: [1] a natural propensity for the bone marrow (besides liver and spleen) [19], which is also the site where plasma cells localize, accumulate and build their own supportive microenvironment; [2] a proven ability to target tumors in general, making use of enhanced vascular permeability as a result of neovascularization and inflammation [20, 21]; and [3] after extravasation at pathological sites, PEG-liposomes show strong uptake by macrophages, a cell type which has gained a reputation as key regulator of MM progression [22, 23].Fig. 1Schematic representation of the difference between conventional oral dexamethasone treatment and intravenously infused PEG-liposomal dexamethasone, emphasizing particularly the striking difference in PK profiles and target site accumulation obtained after liposomal treatment [24] An important advantage of using PEG-liposomes over other drug delivery systems is that they are well established in the clinic, with proven therapeutic value and a fully investigated safety profile. Several products based on PEG-liposomes are clinically used or under investigation [24]. The therapeutic value of PEG-liposomes is more specifically validated in MM by the successful addition of PEG-liposomal doxorubicin to the chemotherapeutic arsenal in relapsed or refractory MM, producing an increase in progression-free survival [25].
We here report the results of an exploratory phase 1 clinical study (ClinicalTrials.gov Identifier NCT03033316) with the primary aim of evaluating safety and pharmacokinetics and secondary aim of evaluating the efficacy of the abovementioned PEG-liposomal dexamethasone (Dex-PL) formulation in patients with heavily pretreated, symptomatic MM.
## Study design
We conducted an 8-week open-label, multi-center, dose-escalating phase I study in patients with pretreated progressive MM. The study was registered under EUdraCT number 2014–005137-32 and approved by both the Independent Ethics Committee at the Medical Faculty of RWTH Aachen University and the Federal Institute for Drugs and Medical Devices (BfArM) and conducted in accordance with the ethical principles of ICH’s Good Clinical Practice guidelines and the Declaration of Helsinki (https://database.ich.org/sites/default/files/ICH_E6-R3_GCP-Principles_Draft_2021_0419.pdf). Written informed consent for study participation was obtained from each patient prior to screening and the performance of any study-specific procedures.
## Patient inclusion criteria
All patients older than 18 years old with a relapse or a progression of previously diagnosed symptomatic MM according to International Myeloma Working Group (IMWG) criteria and who were previously treated with at least two lines of therapy including at least one proteasome inhibitor and at least one of the immunomodulatory imide drugs (IMiDs) were included. Patients were to have a measurable disease (M-protein and/or free light chains) in serum and/or urine.
## Patient exclusion criteria
Key exclusion criteria included the following: previously documented irresponsiveness to dexamethasone monotherapy, a diagnosis of plasma cell leukemia, detectable signs of hepatitis infection, other infections requiring systemic treatment, and treatment with oral or injectable (including intra-articular) corticosteroids (CS) within 4 weeks prior to the screening visit.
## Liposomal dexamethasone study drug
Pegylated liposomal dexamethasone (Dex-PL) was provided by Enceladus Pharmaceuticals, Naarden, the Netherlands. Dex-PL is composed of dipalmitoyl phosphatidyl choline, pegylated distearoyl phosphatidyl ethanolamine, and cholesterol, which constitute the lipid bilayer of carefully sized 100 nm vesicles that stably encapsulate dexamethasone sodium phosphate in the aqueous interior. Before treatment, the required dose of this composition was diluted in $0.9\%$ sodium chloride solution and administered intravenously during 1 to 2 h depending on the dose, using a stepwise protocol of increased infusion rates so as to minimize the chance of infusion-related side effects.
## Study procedures
The design of the study was based on the standard “3 + 3”-algorithm as described by Storer [26]. In this setting, 3 patients per dose level were to be enrolled. If one patient reached a dose limiting toxicity (DLT), another 3 patients would be enrolled in the same dose level. DLT was defined as any toxicity of grade 3 or higher and judged to be related to the study drug by the investigator. For laboratory results, only a number of explicitly defined laboratory adverse events were considered to be DLT (details can be found in the Supplementary Table 2). If two or more patients reached DLT in one dose level, maximum tolerated dose (MTD) was defined at the last dose level beneath DLT. A maximum of five dose levels were originally planned ranging from a single dose of 10 mg to four times a weekly dose of 40 mg.
## Primary and secondary endpoints
To assess safety as a primary outcome, patients were evaluated for presence of and changes in adverse events during each visit and were asked to report symptoms if those had occurred between visits. Adverse events (AEs) and serious adverse events (SAE) were registered and graded in accordance with the National Cancer Institute Common Terminology Criteria for AEs (CTCAE Version 4.03). On pre-defined time points around each dosing, safety laboratory assessments were done (blood chemistry, urine analysis, and hematology). The full schedule of assessments can be found in Supplementary Table 1. Preliminary efficacy was a secondary endpoint and was assessed as response according to IMWG Criteria [27] at week 4 and 8 after the first dose in serum (protein electrophoresis (M-gradient), involved immunoglobulin heavy chain, free kappa, and lambda light chains) and in urine (free kappa and lambda light chains). For assessment of quality of life (QoL), the NCCN distress thermometer as well as a pain visual analog scale (VAS) were used. For pharmacokinetic evaluation, blood samples were collected at day 0 (before treatment administration, immediately after the end of the infusion at $t = 0$, and 2 h after end of infusion), as well as on days 1, 3, 7, 14, 21, 28, 42, and 56. In these samples, both liposomal dexamethasone sodium phosphate and free (released) dexamethasone were assessed, taking into account the fact that phosphatases extremely efficiently convert the phosphate form in plasma into the parent drug and thus all dexamethasone phosphate is liposomal while free dexamethasone represents the released drug [28].
## Statistical analysis
Data listings were provided for PK and safety data. Summaries were presented overall and by cohort. Summary statistics for continuous data contained number of subjects, mean, median, standard deviation, and range (minimum, maximum); summary statistics for categorical data contained number and percentage. The Safety Population includes all subjects who received all or part of the study drug who have at least one post-dose safety assessment. The pharmacokinetic population (PP) includes all subjects who received all or part of the study drug and have at least one valid PK parameter.
## Patient characteristics
In total, seven patients were enrolled in this study and all gave informed consent. Due to rising number of competing trials and increasing number of successfully proven new treatment alternatives during the time of study initiation, enrollment had to be prematurely stopped after two dose levels due to slow recruitment. One patient in the second dose level did not reach the week 8 assessment due to progressive disease with development of a plasma cell leukemia, leading to the recruitment of an additional patient. All patients had progressive multiple myeloma according to IMWG criteria, but no immediate clinical need of a line of treatment. Of the 7 patients, 4 were male and 3 were female. At baseline, the median age was 81 (range 52–87 years) and the mean weight was 70 kg (range 57–91 kg). Further details are given in Table 1.Table 1Characteristics of patients enrolled in the studyDose level 10 mg ($$n = 3$$)Dose level 40 mg ($$n = 4$$)Median age (years)81 (range 52–86)75.5 (range 52–87)Male patients$66.7\%$$50\%$Median height (cm)172 (range: 163–178)166 (range: 162–170)Median weight (kg)65 (range: 63–91)68 (range: 57–80)Median number of concomitant diseases3 (range: 2–5)3 (range: 1–4)Median number of previous lines of chemotherapy3 (range: 2–4)4 (range: 2–6)Median number of concomitant medications7 (range: 4–8)3 (range 2–4)Median M-protein at baseline (g/dl)1.6 (range: 1.05–2.1)1.5 (range: 0.1–2.0)Median creatinine level at baseline (mg/dl)1.32 (range: 0.81–1.56)1.20 (range: 0.81–1.73)Median baseline leukocytes (count/nl)3.7 (range: 2.1–11.8)7.5 (range: 4.3–10.1)Median baseline erythrocytes (count/pl)3.6 (range: 3.3–4.2)3.7 (range: 3.0–4.0)Median baseline platelets (count/nl)249 (range: 117–326)121 (range: 62–185)
## Safety evaluation
The safety results are summarized in Table 2. The occurrence and severity of adverse events (AEs), the occurrence of systemic AEs as measured by potentially clinically significant changes in ECG, vital signs, physical examinations, and laboratory tests, as well as the occurrence of injection site reactions were part of the main endpoints. AEs were graded in line with the Common Terminology Criteria for Adverse Events (CTCAE) version 4.03 (https://ctep.cancer.gov/protocoldevelopment/electronic_applications/ctc.htm).
Table 2Adverse events (AEs) observed in the patients, graded for relatedness and severity. Light amber shaded lines indicate “possibly related”, light orange “probably related” as rated by the investigator, with related Grade 2 adverse events shown in bold* Indicates 3 identical eventsobserved consecutively in the same person. It needs to be mentioned that thepsychiatric and nervous system-related AEs in this table were observed in onesingle patient There were no deaths reported in this study. In total 28 AEs were reported, and they were seen in all 7 patients. Eleven out of 28 AEs in total were categorized as ‘not related’ to the study drug, one was rated ‘unlikely related’ while 8 were judged ‘possibly related’ and 7 ‘probably related’. The most frequent AEs fell in the category of blood and lymphatic system disorders, with 9 events recorded in 5 out of 7 patients. *In* general, no apparent clinically significant changes or abnormalities were observed during the study period, which would have led to termination of the study. No infusion reactions were observed. One patient experienced a neutropenia that was graded of life-threatening severity. This subject had a long-standing fluctuating leuko- and neutropenia, due to impaired bone marrow function as a result of pre-treatment. It was, therefore, judged to be unlikely related to the study drug. The leuko- and neutropenia were not accompanied by clinical symptoms and lasted only one day. Knowing that i.v. administered liposome products tend to be taken up by liver macrophages [18], it is relevant to note that no liver toxicity was seen in these patients.
No dose-limiting toxicities occurred in both dose levels. There were no serious adverse events (SAEs) and no hospitalizations due to AEs. In the second cohort, one patient suffered from premature progression of disease (development of plasma cell leukemia) and could not be kept in the study cohort until week 8. Therefore, an additional patient was added to this cohort.
## Pharmacokinetics
While a complete picture of the pharmacokinetic behavior of PEG-liposomal dexamethasone could not be obtained with only limited data from 7 patients, the results shown in Fig. 2 confirm observations with PEG-liposomal drug products reported by others [29]. In line with those publications, the very high plasma levels and long circulation half-lives observed with liposomal dexamethasone are striking. They indicate that the volume of distribution of the liposomes is not much larger than the plasma volume itself. Fig. 2Pharmacokinetic profiles of encapsulated dexamethasone phosphate and free dexamethasone measured simultaneously upon i.v. infusion of two different doses of PEG-liposomal dexamethasone The curves also show that with the 40 mg liposomal dexamethasone dose some free dexamethasone enters the circulation upon liposome administration, presumably as a result of liposomal clearance by liver macrophages and lymphoid organs, knowing that the liposomes are stable and are not prematurely leaking drug in the circulation [28]. However, compared to the encapsulated dexamethasone phosphate concentrations, the systemic exposure to free dexamethasone is proportionally very low, and most marked only in the first week after liposome administration. In Table 3, the calculated pharmacokinetic parameters for liposomal dexamethasone at the two dose levels are shown and compared to literature data obtained with i.v. and oral free dexamethasone. Table 3Pharmacokinetic parameters of liposomal dexamethasone phosphate vs. free dexamethasone at different dosing levels upon i.v. infusioni.v. liposomal Dex 10 mgi.v. liposomal Dex 40 mgi.v. free Dex 150 mg* [30]i.v. free Dex 4 mg* [31]Oral free Dex 6 mg* [31]UnitPlasma conc at $t = 0$ (s0)2.716.43.60.10.068 (Cmax)mcg/mLDistribution volume (Vd)0.0500.0351.00.941.09L/kgPlasma half-life (t$\frac{1}{2}$)83.4113.44.09.06.9hrsClearance (CL)0.0310.01511.66.47.7L/hrsArea under the curve (AUC)326,1362,685,826114626774mcg*hrs/L* Indicates free dexamethasone data obtained from literature
## Preliminary efficacy results
As only two dose levels could be fully recruited and therefore no patient received more than one dose of the drug, efficacy results are limited. Clinical parameters are indicating that the activity of MM stayed stable in the 40 mg dose group during the 8 weeks of assessment while it tended to rise in the 10 mg treatment group: mean M-protein levels were 1.26 g/dl at baseline in the 40 mg cohort and 1.15 g/dl at week 8 after treatment, while in the 10 mg cohort, the baseline level was 1.58 g/dl and climbed to 2.91 g/dl at week 8.
This picture was confirmed by the immunoglobulin assessments. Figure 3 shows changes in the serum concentrations of immunoglobulin heavy chain or—if the patient that had light chain MM—free kappa and lambda light chains over the 8 weeks of assessment. With progression defined as more than $25\%$ increase in these serum levels, stable disease as between $25\%$ increase and $50\%$ reduction, and a therapeutic response as more than $50\%$ reduction, all patients in the 10 mg dose group showed disease progression, while all patients in the 40 mg group showed stable disease. At week 8, the picture is less clear with stable disease shown with 1 out of 3 patients in the 10 mg group and in 2 out of 3 patients in the 40 mg group (one patient in the 40 mg group was missing due to premature disease progression).Fig. 3IMWG response calculated as percentage change from baseline of serum levels of immunoglobulin heavy chain or—in case of light chain myeloma—the free light chain, at week 4 and 8 after the i.v. infusion of PEG-liposomal dexamethasone phosphate. Bars show individual patients. Blue are patients in the 10 mg and orange in the 40 mg dose group
## Discussion
We conducted a combined phase I/IIa open label interventional trial designed to evaluate the safety and efficacy of a pegylated liposome formulation of dexamethasone (Dex-PL). We found that a single intravenous infusion of liposomal dexamethasone was well tolerated, both during the infusion and the follow-up study period. Regarding pharmacokinetics, while the type of patients treated IV with Dex-PL was clearly different from the patients in the referenced study on free dexamethasone, the multifold lower Cl and Vd with liposomal dexamethasone are hard to miss, as is the more than tenfold longer half-life (Table 3). Indeed, liposomal encapsulation seems to increase the AUC of the drug with more than a factor 200, taking into account the difference in dose level.
The increasing number of treatment options developed during the past decade have turned MM from a cancer with a poor survival outcome into a more chronic disease mostly treated in an outpatient setting in which clinically quieter stages (e.g., under oral maintenance treatment) alternate with relapses that require changes in the (now mostly targeted) treatment strategy. Almost all conventional drug therapies in MM are combinations in which dexamethasone represents an important backbone. Also, in the new targeted combination treatment approaches that involve proteasome inhibitors, immunomodulatory drugs, and/or monoclonal antibodies, dexamethasone retains its place as standard-of-care [32]. The strong immunosuppressive activity of glucocorticoids (GC) and in particular dexamethasone are thought to amplify the immunomodulatory activity of these more targeted drugs. On the other hand, however, GC are characterized by a significant side effect profile that includes a substantially increased chance of infections particularly in combination with immune-modulatory drugs in MM [33]. Considering the fact that infections are one of the main risks in the long-term management of MM especially during chemotherapeutic regimens, the search for new effective MM drugs has also been a quest to reduce infection-related morbidity and mortality.
Outside the field of cancer, several lines of investigation have been pursued to improve the therapeutic index of systemically administered GC [34], notable examples being the more selective glucocorticoid receptor agonists. Another widely explored strategy resides in the selective delivery of GC to pathological sites using nanomedicines, with several GC nanodrugs currently in (pre)clinical development [35] and with a good amount of preclinical proof-of-concept available in experimental tumor models [36–38].
The most advanced GC nanodrug is pegylated liposomal prednisolone, which was recently shown to improve efficacy over GC standard-of-care in a large cohort of patients with active rheumatoid arthritis [39]. The same product was also evaluated in patients with severe orbital inflammation due to Graves’ disease and in patients on dialysis with arteriovenous fistula failure [40, 41]. The use of the same liposome system encapsulating the 6.5-fold more potent dexamethasone, and thus theoretically requiring less liposomes for the same effect, seems a logical next step especially for indications in which dexamethasone is the clinically preferred GC. Indeed, dexamethasone tends to be the GC of choice in several types of cancer, of which MM is the disease in which there is the highest unmet medical need for improving the efficacy-safety ratio of the drug. Recently, a small clinical trial was completed with PEG-liposomal dexamethasone in cancer patients with castrate-resistant prostate carcinoma in which the safety of repeated doses of up to 18.5 mg was assessed [42]. This was based on preclinical studies with liposomal dexamethasone revealing that there may be additional therapeutic benefit of enhanced local delivery to tumor sites. First, the drug may impact the extracellular matrix density in and around tumors, thereby enhancing drug delivery to and into tumors [43]. Furthermore, liposomal dexamethasone has been investigated in preclinical models of prostate cancer, where it appears to block the growth of bone metastases, and does so more effectively than free dexamethasone [44].
Although the current dataset is limited, we believe the results presented here support the hypothesis that long-circulating liposomes are safe and (pending confirmation in follow up trials) may provide for a favorable tolerability of dexamethasone therapy in MM. Among the most interesting findings are the changes in pharmacokinetic behavior of the GC conferred by liposomal encapsulation. Interestingly, the enormously increased area under the curve (AUC) does not seem translate in high systemic activity, suggested by the limited number of adverse effects observed. Possible steroid-related psychiatric and nervous system-related side effects were seen in one patient. These side effects mostly resolved in the weeks after treatment. With regard to infections, it is noteworthy that although a moderate lung infection and a mild influenza were reported, these resolved and were judged as not related by the investigator.
The observation that a multifold larger AUC does not lead to marked and substantial systemic dexamethasone effects corroborates the assumption that dexamethasone largely remains in its inactive phosphate prodrug form, safely encapsulated in the liposomes, limiting systemic exposure. At the same time, the high plasma concentrations may result in high drug concentrations at target sites, because MM bone marrow lesions are characterized by increased micro-vessel density, permeability, and by ongoing angiogenesis, which could render the target tissue directly accessible to the liposomes [43–45]. Here, populations of phagocytosing immune cells can take up and digest the liposomes, liberating the dexamethasone phosphate prodrug and converting it into active dexamethasone that can either exert its pharmacological activity in these immune cells or upon release in other key cells in the pathological lesions, including the myeloma plasma cells.
An obvious limitation of this study is the small number of patients due to the premature closing of the trial. The study protocol intended a higher number of patients with higher and also repeated doses of dexamethasone. Unfortunately, recruitment into the study became increasingly difficult. When the study was planned, only lenalidomide and bortezomib were widely available. Finally, when the study was running, several more and well-tolerated drugs had been approved for the treatment of multiple myeloma, making inclusion into a study with dexamethasone monotherapy difficult.
The administration of liposomal dexamethasone up to a single dose of 40 mg seems safe and led to extended circulating drug levels, and no dose-limiting toxicities were observed yet. The levels of circulating free active drug were low. There appeared to be a dose-dependent pharmacological effect, as 40 mg of Dex-PL lead to disease stabilization in 4 out of 4 patients at week 4, while 10 mg of Dex-PL resulted in progressive disease in 3 out of 3 patients. Further studies are needed to investigate higher doses, to characterize side effects as well as to identify of the dose-limiting toxicity, and above all to obtain evidence for improved efficacy. Possibly, liposomal dexamethasone can serve as a combination formulation in many drugs regimens in the future, with higher drug concentrations in the bone marrow and fewer side effects due to lower systemic exposure. All in all, this first-in-man MM patient study may provide a stepping stone to future studies in which this potentially safer form of dexamethasone is explored alongside other and newer MM drug combinations.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (PDF 94 KB)
## References
1. Palumbo A, Anderson K. **Multiple myeloma**. *N Engl J Med* (2011.0) **364** 1046-1060. DOI: 10.1056/NEJMra1011442
2. Mahindra A. **Latest advances and current challenges in the treatment of multiple myeloma**. *Nat Rev Clin Oncol* (2012.0) **9** 135-143. DOI: 10.1038/nrclinonc.2012.15
3. Ocio EM. **New drugs and novel mechanisms of action in multiple myeloma in 2013: a report from the International Myeloma Working Group (IMWG)**. *Leukemia* (2014.0) **28** 525-542. DOI: 10.1038/leu.2013.350
4. Kyle RA, Rajkumar SV. **Treatment of multiple myeloma: a comprehensive review**. *Clin Lymphoma Myeloma* (2009.0) **9** 278-288. DOI: 10.3816/CLM.2009.n.056
5. Palumbo A. **International Myeloma Working Group consensus statement for the management, treatment, and supportive care of patients with myeloma not eligible for standard autologous stem-cell transplantation**. *J Clin Oncol* (2014.0) **32** 587-600. DOI: 10.1200/JCO.2013.48.7934
6. Rodríguez-Lobato LG. **G CAR T-cells in multiple myeloma: state of the art and future directions**. *Front Oncol* (2020.0) **10** 1243. DOI: 10.3389/fonc.2020.01243
7. Chim CS. **Management of relapsed and refractory multiple myeloma: novel agents, antibodies, immunotherapies and beyond**. *Leukemia* (2018.0) **32** 252-262. DOI: 10.1038/leu.2017.329
8. Kumar S. **Single agent dexamethasone for pre-stem cell transplant induction therapy for multiple myeloma**. *Bone Marrow Transplant* (2004.0) **34** 485-490. DOI: 10.1038/sj.bmt.1704633
9. von Lilienfeld-Toal M. **A systematic review of phase II trials of thalidomide/dexamethasone combination therapy in patients with relapsed or refractory multiple myeloma**. *Eur J Haematol* (2008.0) **81** 247-252. DOI: 10.1111/j.1600-0609.2008.01121.x
10. Salem K. **Combination chemotherapy increases cytotoxicity of multiple myeloma cells by modification of nuclear factor (NF)-kappaB activity**. *Exp Hematol* (2013.0) **41** 209-218. DOI: 10.1016/j.exphem.2012.10.002
11. Friedenberg WR. **High-dose dexamethasone for refractory or relapsing multiple myeloma**. *Am J Hematol* (1991.0) **36** 171-175. DOI: 10.1002/ajh.2830360303
12. Moreau P. **Bortezomib, thalidomide, and dexamethasone with or without daratumumab before and after autologous stem-cell transplantation for newly diagnosed multiple myeloma (CASSIOPEIA): a randomised, open-label, phase 3 study**. *Lancet* (2019.0) **394** 31240-31241. DOI: 10.1016/S0140-6736(19)31240-1
13. Rajkumar SV. **Lenalidomide plus high-dose dexamethasone versus lenalidomide plus low-dose dexamethasone as initial therapy for newly diagnosed multiple myeloma: an open-label randomised controlled trial**. *Lancet Oncol* (2010.0) **11** 29-37. DOI: 10.1016/S1470-2045(09)70284-0
14. Sharma S, Lichtenstein A. **Dexamethasone-induced apoptotic mechanisms in myeloma cells investigated by analysis of mutant glucocorticoid receptors**. *Blood* (2008.0) **112** 1338-1345. DOI: 10.1182/blood-2007-11-124156
15. Murray MY. **Attenuation of dexamethasone-induced cell death in multiple myeloma is mediated by miR-125b expression**. *Cell Cycle* (2013.0) **12** 2144-2153. DOI: 10.4161/cc.25251
16. Ishikawa H. **Effect of glucocorticoids on the biologic activities of myeloma cells: inhibition of interleukin-1 beta osteoclast activating factor-induced bone resorption**. *Blood* (1990.0) **75** 715-720. DOI: 10.1182/blood.V75.3.715.715
17. Alessi J. **Dexamethasone in the era of COVID-19: friend or foe? An essay on the effects of dexamethasone and the potential risks of its inadvertent use in patients with diabetes**. *Diabetol Metab Syndr* (2020.0) **12** 80. DOI: 10.1186/s13098-020-00583-7
18. Allen TM. **Liposomes containing synthetic lipid derivatives of poly(ethylene glycol) show prolonged circulation half-lives in vivo**. *Biochim Biophys Acta* (1991.0) **1066** 29-36. DOI: 10.1016/0005-2736(91)90246-5
19. Huang SK. **Microscopic localization of sterically stabilized liposomes in colon carcinoma-bearing mice**. *Cancer Res* (1992.0) **52** 5135-5143. PMID: 1394121
20. Abraham SA. **The liposomal formulation of doxorubicin**. *Methods Enzymol* (2005.0) **391** 71-97. DOI: 10.1016/S0076-6879(05)91004-5
21. Jakob C. **Angiogenesis in multiple myeloma**. *Eur J Cancer* (2006.0) **42** 1581-1590. DOI: 10.1016/j.ejca.2006.02.017
22. Ribatti D. **Macrophages in multiple myeloma**. *Immunol Lett* (2014.0) **161** 241-244. DOI: 10.1016/j.imlet.2013.12.010
23. Hofkens W. **Intravenously delivered glucocorticoid liposomes inhibit osteoclast activity and bone erosion in murine antigen-induced arthritis**. *J Control Release* (2011.0) **152** 363-369. DOI: 10.1016/j.jconrel.2011.03.001
24. Immordino ML, Dosio F, Cattel L. **Stealth liposomes: review of the basic science, rationale, and clinical applications, existing and potential**. *Int J Nanomedicine* (2006.0) **1** 297-315. PMID: 17717971
25. Ning YM. **Liposomal doxorubicin in combination with bortezomib for relapsed or refractory multiple myeloma**. *Oncology (Williston Park)* (2007.0) **21** 1503-1508. PMID: 18077994
26. Storer BE. **Design and analysis of phase I clinical trials**. *Biometrics* (1989.0) **45** 925-937. DOI: 10.2307/2531693
27. Durie BG. **International uniform response criteria for multiple myeloma**. *Leukemia* (2006.0) **20** 1467-1473. DOI: 10.1038/sj.leu.2404284
28. Metselaar JM. **Complete remission of experimental arthritis by joint targeting of glucocorticoids with long-circulating liposomes**. *Arthritis Rheum* (2003.0) **48** 2059-2066. DOI: 10.1002/art.11140
29. Gabizon A. **Pharmacokinetics of pegylated liposomal doxorubicin: review of animal and human studies**. *Clin Pharmacokinet* (2003.0) **42** 419-436. DOI: 10.2165/00003088-200342050-00002
30. Brady ME. **The pharmacokinetics of single high doses of dexamethasone in cancer patients**. *Eur J Clin Pharmacol* (1987.0) **32** 593-596. DOI: 10.1007/BF02455994
31. 31.Spoorenberg SMC, et al. Pharmacokinetics of oral vs. intravenous dexamethasone in patients hospitalized with community-acquired pneumonia. Br J Clin Pharmacol. 2014;78(1):78–83.
32. Tacchetti P. **Emerging and current treatment combinations for transplant-ineligible multiple myeloma patients**. *Expert Rev Hematol* (2021.0) **14** 1085-1098. DOI: 10.1080/17474086.2021.1983426
33. Nucci M. **Infections in patients with multiple myeloma in the era of high-dose therapy and novel agents**. *Clin Infect Dis* (2009.0) **49** 1211-1225. DOI: 10.1086/605664
34. Buttgereit F. **Optimised glucocorticoid therapy: the sharpening of an old spear**. *Lancet* (2005.0) **365** 801-803. DOI: 10.1016/S0140-6736(05)17989-6
35. Luehder F. **Novel drug delivery systems tailored for improved administration of glucocorticoids**. *Int J Mol Sci* (2017.0) **18** 1836. DOI: 10.3390/ijms18091836
36. Banciu M. **Utility of targeted glucocorticoids in cancer therapy**. *J Liposome Res* (2008.0) **18** 47-57. DOI: 10.1080/08982100801893978
37. Schiffelers RM. **Liposome-encapsulated prednisolone phosphate inhibits growth of established tumors in mice**. *Neoplasia* (2005.0) **7** 118-127. DOI: 10.1593/neo.04340
38. Ozbakir B. **Liposomal corticosteroids for the treatment of inflammatory disorders and cancer**. *J Control Release* (2014.0) **19** 624-636. DOI: 10.1016/j.jconrel.2014.05.039
39. 39.Metselaar JM, et al. Intravenous pegylated liposomal prednisolone outperforms intramuscular methylprednisolone in treating rheumatoid arthritis flares: a randomized controlled clinical trial. J Control Release. 2022;341:548–54.
40. Detiger SE. **A pilot study on the use of prednisolone-encapsulated liposomes for the treatment of moderate-to-severe Graves' orbitopathy with reduced systemic steroid exposure**. *Acta Ophthalmol* (2021.0) **99** 797-804. DOI: 10.1111/aos.14751
41. Voorzaat BM. **A randomized trial of liposomal prednisolone (LIPMAT) to enhance radiocephalic fistula maturation: a pilot study**. *Kidney Int Rep* (2020.0) **5** 1327-1332. DOI: 10.1016/j.ekir.2020.05.030
42. Vrouwe JPM. **An exploratory first-in-man study to investigate the pharmacokinetics and safety of liposomal dexamethasone at a 2- and 1-week interval in patients with metastatic castration resistant prostate cancer**. *Pharmacol Res Perspect* (2021.0) **9** e00845. DOI: 10.1002/prp2.845
43. Martin JD. **Dexamethasone increases cisplatin-loaded nanocarrier delivery and efficacy in metastatic breast cancer by normalizing the tumor microenvironment**. *ACS Nano* (2019.0) **13** 6396-6408. DOI: 10.1021/acsnano.8b07865
44. Kroon J. **Liposomal delivery of dexamethasone attenuates prostate cancer bone metastatic tumor growth in vivo**. *Prostate* (2015.0) **75** 815-824. DOI: 10.1002/pros.22963
45. Giuliani N. **Angiogenesis and multiple myeloma**. *Cancer Microenviron* (2011.0) **4** 325-337. DOI: 10.1007/s12307-011-0072-9
|
---
title: In vitro–in vivo correlation of drug release profiles from medicated contact
lenses using an in vitro eye blink model
authors:
- Ana F. Pereira-da-Mota
- Maria Vivero-Lopez
- Piyush Garg
- Chau-Minh Phan
- Angel Concheiro
- Lyndon Jones
- Carmen Alvarez-Lorenzo
journal: Drug Delivery and Translational Research
year: 2022
pmcid: PMC9981533
doi: 10.1007/s13346-022-01276-6
license: CC BY 4.0
---
# In vitro–in vivo correlation of drug release profiles from medicated contact lenses using an in vitro eye blink model
## Abstract
There is still a paucity of information on how in vitro release profiles from drug-loaded contact lenses (CLs) recorded in 3D printed eye models correlate with in vivo profiles. This work aims to evaluate the release profiles of two drug-loaded CLs in a 3D in vitro eye blink model and compare the obtained results with the release in a vial and the drug levels in tear fluid previously obtained from an animal in vivo study. In vitro release in the eye model was tested at two different flow rates (5 and 10 µL/min) and a blink speed of 1 blink/10 s. Model CLs were loaded with two different drugs, hydrophilic pravastatin and hydrophobic resveratrol. The release of both drugs was more sustained and lower in the 3D eye model compared to the in vitro release in vials. Interestingly, both drugs presented similar release patterns in the eye model and in vivo, although the total amount of drugs released in the eye model was significantly lower, especially for resveratrol. Strong correlations between percentages of pravastatin released in the eye model and in vivo were found. These findings suggest that the current 3D printed eye blink model could be a useful tool to measure the release of ophthalmic drugs from medicated CLs. Nevertheless, physiological parameters such as the composition of the tear fluid and eyeball surface, tear flow rates, and temperature should be optimized in further studies.
## Introduction
The use of contact lenses (CLs) as platforms for controlled delivery of ophthalmic drugs was envisioned in 1961 by Otto Wichterle and co-workers [1]. After 60 years, the first commercial drug-delivering CL has become available in Japan, Canada, and the USA [2]. Compared to eye drops, CLs may significantly extend drug residence time and increase ocular bioavailability, while unproductive drug absorption is minimized [3].
CLs are one of the most successfully commercialized biomedical devices, with nearly 150 million users worldwide [4]. However, several obstacles oppose their use as platforms for ocular drug delivery [5]. Most polymers used to prepare CLs, e.g., 2-hydroxyethyl methacrylate (HEMA) and 3-(methacryloyloxy)propyl tris(trimethylsiloxy)silane (TRIS), lack affinity for drugs; thus, typically, CLs do not uptake the required dose or release it too rapidly. A wide variety of strategies to overcome this issue is under development [6–9]. Additionally, there are no standardized methods for testing in vitro the drug release profiles from CLs. In vitro methods for mimicking the composition and dynamics of tear fluid, and the frequency and pressure of blinking are still a challenge. Thus, in most reports focused on CLs, the in vitro drug release profiles are recorded in small beakers using a variety of medium composition, volume, stirring, and replacement conditions [10, 11]. As a consequence, most in vitro results are not predictive of in vivo performance [12]. This means that in vivo testing in animal models and human preclinical studies are still needed, even to evaluate early-stage drug-CL combination products, which makes the development very costly in time and resources.
In vitro models that mimic the in vivo scenario and key ocular parameters are highly explored. Microfluidic devices have been designed to regulate the flow and volume where CLs are immersed, but other physiological conditions were not reproduced by these devices, including factors such as corneal and eyelid shape and format, tear film thickness, or blinking [13, 14]. 3D printed in vitro eye models to evaluate the in vitro performance of CLs have recently been undertaken to overcome some challenges faced by using microfluidic devices and more appropriately simulate the effects of tear flow rate, tear volume, air exposure, and eyelid blinking frequency [15, 16]. Not surprisingly, under dynamic conditions of low tear fluid flow, CLs showed slower drug release profiles compared to static release in a beaker, prolonging the release for days or weeks for some specific compounds and drugs such as red food dye [16], polyethylene glycol, hydroxypropyl methylcellulose [17], moxifloxacin [18], and fluconazole [19]. Moreover, in vitro models can provide relevant insights in the development process of drug-loaded CLs and prioritize successful materials that may go forward to in vivo testing in animal preclinical models or human clinical studies.
There is still a paucity of information on how in vitro release profiles recorded in 3D printed eye models correlate with in vivo profiles. Comparison of the behavior of the same drug-loaded CLs in both the in vitro model and the common rabbit eye model is, therefore, required for the validation of the information gathered in vitro. To gain an insight into the in vitro–in vivo correlations, the aim of this work was to analyze the release profiles of drug-loaded CLs recorded in a 3D printed in vitro eye blink model and compare the obtained results with the release in a small beaker and the tear levels previously obtained in vivo. For the sake of robustness, CLs loaded with drugs differing in physicochemical properties were tested, namely, CLs designed to uptake pravastatin (a hydrophilic statin, log P = − 0.23 [20]) and resveratrol (a highly hydrophobic antioxidant, log $$P \leq 3.09$$ [21]) were prepared and evaluated [22, 23].
Pravastatin sodium and resveratrol may be useful for the treatment of a wide range of anterior and posterior ocular diseases. Prolonged oral therapy for hypercholesterolemia with statins has been shown to promote corneal healing, prevent cataract formation, reduce glaucoma severity, and reduce the appearance of hard exudates and microaneurysms in patients diagnosed with diabetic macular edema; topical ocular treatment has the advantage of avoiding systemic adverse reactions [24–27]. Resveratrol is an antioxidant agent that aids the management of oxidative-stress-related eye diseases and improves the healing of corneal epithelial cells [28]. In previous studies, both drugs were incorporated in model CLs, and the in vivo performance was evaluated in New Zealand white rabbits [22, 29]. Both drug-loaded CLs provided significantly higher and more prolonged drug levels in the rabbits’ tear fluid compared to eye drops with the same dose, which favored ocular biodistribution in the anterior and posterior structures of the eye, including cornea, sclera, lens, aqueous and vitreous humors, and retina. To carry out the present work, HEMA-based CLs were copolymerized with specific functional monomers that enhance drug affinity. In the case of pravastatin, HEMA was copolymerized with ethylene glycol phenyl ether methacrylate (EGPEM) and N-(3-aminopropyl) methacrylamide hydrochloride (APMA) (Fig. 1). For resveratrol, methacryloyloxyethyl phosphorylcholine (MPC) was added as an antifouling comonomer. The developed CLs, coded as AECLs and MCLs, demonstrated adequate solvent uptake, light transmission, mechanical properties, and ocular safety. In vitro release experiments were carried out with sterile CLs loaded under specific conditions for each drug depending on their physicochemical properties. In vitro release in the 3D eye blink model was tested at two different tears fluid flow rates (5 and 10 µl/min of fluid) and a blink speed of 1 blink/10 s. The amount of drug released from the CLs was collected and quantified by high-performance liquid chromatography (HPLC). To the best of our knowledge, this is the first time that drug-loaded CL release profiles have been evaluated in an in vitro 3D eye model and in vitro–in vivo correlations (IVIVC) are attempted. Fig. 1Chemical structures of the drugs and monomers. a Pravastatin sodium, b N-(3-aminopropyl) methacrylamide hydrochloride (APMA), c ethylene glycol phenyl ether methacrylate (EGPEM), d trans-resveratrol, and e 2-methacryloyloxyethyl phosphorylcholine (MPC)
## Materials
Pravastatin sodium was supplied by Biocon Limited (Bengaluru, Karnataka, India). Resveratrol was from ChemCruz, Santa Cruz Biotechnology Inc. (Dallas, TX, USA). Di-sodium hydrogen phosphate anhydrous (NaH2PO4) and 2-hydroxyethyl methacrylate (HEMA) were from Merck (Darmstadt, Germany). N-(3-aminopropyl) methacrylamide hydrochloride (APMA) was from PolySciences Inc. (Warrington, PA, USA). Ethylene glycol dimethacrylate (EGDMA), ethylene glycol phenyl ether methacrylate (EGPEM), 2,2′-azobis(isobutyronitrile) (AIBN), 2-methacryloyloxyethyl phosphorylcholine (MPC), polyvinyl alcohol (PVA, 89–98 kDa, $99\%$ hydrolyzed), and dimethyl sulfoxide (DMSO) were from Sigma-Aldrich (Steinheim, Germany). Sodium chloride (NaCl) was from Labkem (Barcelona, Spain), and sodium hydroxide (NaOH) was from VWR Chemicals (Leuven, Belgium). The 3D printing UV-sensitive resin was from Anycubic Technology Co. (Shenzhen, Guangdong, China). Methanol $99.9\%$ for LC–MS grade was from Thermo Fisher Scientific (Loughborough, UK). Simulated lachrymal fluid (SLF) was prepared as previously reported [22]. Ultrapure water (resistivity > 18.2 MΩ cm; Milli-Q®; Millipore Ibérica, Madrid, Spain) was obtained by reverse osmosis.
## Contact lens preparation
Two different types of HEMA-based CLs were prepared as previously described [22, 29]. Briefly, AECLs for pravastatin were prepared by mixing HEMA (3 mL) with APMA (21.45 mg), EGPEM (112.50 µL), and EGDMA (12.10 µL). The monomer solutions were magnetically stirred (200 rpm at room temperature) for 120 min, and the initiator (AIBN, 14.79 mg) was then added and solubilized by magnetic stirring for a further 30 min.
To prepare MCLs for resveratrol, HEMA (3 mL) was mixed with MPC (337.5 mg) and EGDMA (12.10 µL) under magnetic stirring (150 rpm at room temperature) for 60 min. The mixture was kept under magnetic stirring for 30 min more to ensure the complete dissolution of AIBN (32.85 mg).
Both AECLs and MCLs were synthesized by adding 60 µL of monomer solution into curved polypropylene moulds typically used for daily disposable CL preparation. The moulds were kept at 50 °C for 12 h and then at 70 °C for 24 h to complete thermal polymerization. Then, the moulds were immersed in MilliQ® water to facilitate CL separation. The obtained CLs were washed under magnetic stirring (200 rpm) in 1 L of MilliQ® water and NaCl $0.9\%$ until complete removal of unreacted monomers occurred; the solvent was replaced at least three times per day. The absence of unreacted monomers was verified by UV–Vis spectrophotometry (Agilent 8453, Waldbronn, Germany).
The final dimensions of hydrated CLs (immersed in phosphate buffer, pH 7.4) were approximately 12 mm diameter, 7.8 mm curvature, and 0.1 mm thickness for AECLs and approximately 14 mm diameter, 8.8 mm curvature, and 0.1 mm thickness for MCLs.
## Pravastatin sodium
Dried AECLs (average mass 16.91 ± 1.28 mg) were packaged and sealed in polyamide/polyethylene vacuum bags filled with 10 mL of an aqueous pravastatin solution (0.1 mg/mL) for at least 48 h and sterilized by high hydrostatic pressure (HHP, 70 °C and 600 MPa for 10 min) [30]. The CLs were stored in sealed bags at room temperature and protected from light until release experiments were performed. All the experiments were carried out in quadruplicate. The amount of pravastatin loaded was quantified by HPLC, as explained in “Drug quantification methods” section.
## Resveratrol
Sterile MCLs (average mass 16.8 ± 1.86 mg, sterilized by steam heat at 121 °C, 20 min) were placed in tubes containing 7 mL of a resveratrol solution (0.1 mg/mL in ethanol:water 10:90 v/v) previously filtered (Filter-Lab® polyethersulphone (PES) syringe filter 0.22 μm; Barcelona, Spain). The loading solution was added to the tubes under sterile conditions in a biological safety cabinet. The tubes were maintained protected from light to avoid resveratrol degradation at 37 °C, 180 rpm for 72 h, after which the release tests were performed ($$n = 4$$). The amount of resveratrol loaded was quantified by HPLC, as described in “Drug quantification methods” section.
For resveratrol release experiments, a two-step sterilization protocol was implemented as resveratrol may degrade at high temperature [36, 37]. The MCLs were sterilized by steam heat (121 °C, 20 min) in empty Falcon® tubes, and then the tubes were filled with a previous filtered resveratrol solution (0.1 mg/mL in ethanol:water 10:90 v/v). MCLs loaded, on average, approximately 13.20 ± 0.90 mg of resveratrol per g of dried hydrogel after being immersed in the drug solution for 72 h.
In the in vitro vial conditions, the MCLs released 54.43 ± $7.29\%$ resveratrol in the first 10 h (Fig. 4a). As to what happened with pravastatin, the amount of resveratrol released in the eye blink model was significantly lower: 0.37 ± $0.22\%$ at 5 µL/min and 0.47 ± $0.26\%$ at 10 µL/min (ANOVA $p \leq 0.001$) than the amount released in the vial. The difference between the vial results and eye blink model might be related to the diffusion resistance associated with the hydrophobic nature of resveratrol [38] since resveratrol solubility in NaCl $0.9\%$ was quantified to be approximately 27.4 μg/mL [23]. In the case of the vial tests, the volume of the release medium was increased up to 12 mL to avoid medium saturation and false plateaus.
Fig. 4a Resveratrol release profiles from MCLs experimentally recorded in a vial filled with 6 mL of NaCl $0.9\%$ over 10 h and on the eye blink model (flow rate of 5 and 10 µL/min) and b amount of resveratrol retained in the PVA eyelid and CLs after 10 h on the eye blink model ($$n = 4$$, mean values and standard deviations). * Statistically different between the amount of resveratrol retained in the model eyelid with the flow rate of 5 and 10 µL/min, $p \leq 0.05$ The amount of resveratrol absorbed by the PVA eyelid was about tenfold higher compared to pravastatin, showing a higher affinity between resveratrol and the PVA eyelid (Fig. 4b). This higher affinity could also contribute to the decrease in resveratrol detected in the fluid collected from the eye blink model. The higher fluid flow rate (10 µL/min) induced a higher amount of resveratrol absorbed by the eyelid (ANOVA, $p \leq 0.001$). As a consequence of the slow release from the MCLs, the amount of resveratrol remaining in the CLs after 10 h of the release tests was approximately $40\%$ of the drug loaded (91.29 ± 13.85 µg for 5 µL/min and 71.37 ± 8.77 µg for 10 µL/min). No statistical differences were detected between both flow rates (ANOVA, $$p \leq 0.10$$).
Resveratrol-loaded CLs presented a sustained release in the eye blink model similar to what occurred in vivo (Fig. 7). However, the concentration of resveratrol detected in the fluid collected in the eye blink model was approximately 150-fold lower than in vivo (ANOVA, $p \leq 0.001$), likely associated with several factors that also conditioned the release of the hydrophilic statin in the eye blink model, plus the low solubility of resveratrol in an aqueous medium, as previously mentioned. No statistical differences were detected between the two different flow rates (ANOVA, $p \leq 0.05$).
Fig. 7a In vivo tear fluid levels of resveratrol were recorded during wear of resveratrol-loaded MCLs for 8 and 10 h ($$n = 6$$ for 8 h and $$n = 3$$ for 9 and 10 h), data taken from Vivero-Lopez et al. [ 29]. b Normalized resveratrol released concentrations in NaCl $0.9\%$ from resveratrol-loaded MCLs over 10 h on the eye model (flow rate of 5 and 10 µL/min) ($$n = 4$$, mean values and standard deviations) For resveratrol experiments, plots of in vivo release versus in vitro release in the eye blink model presented slopes that were remarkably high: 333.96 and 245.16 for the flow rate of 5 and 10 µL/min, respectively (Fig. 8), which supported that the percentage of resveratrol released in vivo was significantly higher than that recorded in the eye blink model. In comparison, no differences were obtained for in vitro vial release and the eye blink model. This finding highlights the difficulties in developing in vitro release models that can predict the in vivo performance of CLs loaded with hydrophobic drugs, which represent about $40\%$ of current pharmaceutical treatments [44].Fig. 8Correlations for in vivo (a) or in vitro (b) versus eye blink model percentage of resveratrol released. The eye blink model experiments were carried out with 5 and 10 µL/min of flow rate (eye blink model 5 and eye blink model 10, respectively)
## Eye blink model
The fabrication and assembly of the 3D eye model were similar to those reported in previous publications by some authors of this paper [15, 16], with minor changes (Fig. 2).Fig. 2In vitro eye blink model (Ocublink) setup. The eyelid movement spreads the tear solution, which is supplied through the tubing that is attached to the eyelid support, over the eyeball, and the contact lens (fitted on the eyeball, shown in the close-up image). The out-flow solution is collected in the collection unit located below the eyeball
## Eyeball and collection unit
The eyeball, lower eyelid, and collection unit were fabricated using a combination of 3D printing and moulding techniques, as previously described [31]. The components were printed using a hydrophobic UV-polymerizable resin on an SLA (stereolithography) 3D printer (Photon S; Anycubic, Shenzhen, China) and an FDM (fused deposition modelling) 3D printer (Prusa i3 MK3S +; Prusa Prague, Czech Republic) to ensure water-sealed parts. All printing parameters were set to the manufacturer’s default settings. The eyeball was composed of two curvatures, 11.25 mm for the larger globe and 8.6 mm for the smaller globe containing the cornea, and 8.6 mm was chosen to match the most common base curve of CLs. Compared to previous models [16], the model used in the current study did not have any coatings for the front corneal surface but instead had a 300-µm groove at the center to allow for a CL to be mounted. In preliminary tests, it was found that the resin materials for the eyeball did not absorb pravastatin and resveratrol. The collection unit of the model was designed to allow the tear film to flow from the eyeball into the wells via gravity.
## Eyelid
The eyelids were designed to have a curvature of 8.8 mm, which leaves a gap of approximately 200 µm between the eyeball and the eyelid. Once a contact lens was applied on the eyeball, this gap was reduced by the thickness of the lens (75–150 µm). The eyelid was designed to rotate around the smaller globe and flexes over the larger globe.
PVA eyelids were prepared by dissolving PVA (89–98 kDa, $20\%$ w/v) in dimethyl sulfoxide:ultrapure water 80:20 v/v mixtures following a previously described protocol [16]. Briefly, the mixture was gently stirred for 5 min and heated at 120 °C for 2 h. After heating, the mixture was stirred again to ensure proper mixing of PVA. The obtained viscous solution was cast in 3D printed moulds (allowing for the preparation of four eyelids in the same mould) and then frozen at − 30 °C for 12 h. The resulting gels were thawed at room temperature for 1 h, removed from the moulds, and immersed in ultrapure water for 3 days, replacing the medium daily to remove the dimethyl sulfoxide used to prepare the PVA solution. After the washing process, the eyelids were immersed in ultrapure water to maintain the hydration of the eyelid until being used in the release experiments.
## Flow rate and blinking
A commercial syringe pump (PHD ULTRA; Harvard Apparatus, Holliston, MA) was used to simulate the dynamic tear flow in the eye blink model. Simulated lachrymal fluid (SLF) or NaCl $0.9\%$ was delivered through a hole at the top of the eyelid and spread over the eyeball surface/CL through blinking. In this study, the blink speed was set to 1 blink/10 s, and the flow rate was adjusted to 5 and 10 µL/min to ensure enough volume for sample collection at the predetermined time points. The blink velocity used in this model was 50 rpm or 45 mm/s for both closing and opening speeds; the average physiological speeds for closing and opening of the eyelid have been reported to be approximately 134 ± 4 and 26 ± 2 mm/s, respectively [32].
## Temperature and humidity
The entire system was covered with an acrylic chamber to maintain stable humidity and temperature (20 ± 1.5 °C) during the experiment. Humidity levels were maintained close to approximately $80\%$ using a humidifier and controlled through a hygrometer.
## Release sampling
At predetermined time points (5, 15, 30 min, and every hour until 10 h), the out-flow solution was pipetted from the collection unit, stored in 300 or 600 µL Eppendorf® tubes, and frozen at − 30 °C until HPLC analysis. The amount of resveratrol and pravastatin released from the CLs was quantified by HPLC previous dilution of the samples in ethanol:water 50:50 v/v and SLF, respectively. All the experiments were carried out in quadruplicate.
## Drug extraction from the eyelid and CLs
After 10 h of experimentation, each PVA eyelid was removed from the system, cut into small pieces, and immersed in 1 ml of SLF or 3 ml of ethanol:water 50:50 v/v for AECLs and MCLs. The eyelids were maintained at 37 °C and 180 rpm for at least 12 h to extract the amount of drug absorbed. The same procedure was applied for the CLs after 10 h on the eye model (without cutting) to determine the remaining amount of drug in the CLs at the end of the test.
## Release in a vial
Sterilized pravastatin-loaded AECLs ($$n = 4$$) were rinsed with SLF to remove excess drug from the CL surface and immersed in 2 or 10 mL of SLF (pH = 7.4) to evaluate if the release volume could have an impact on the drug release profile. The in vitro release experiments were performed at 37 °C, under oscillatory movement (180 rpm), and at predefined time points, 150 µL was removed and replaced by the same volume of fresh medium. The amount of pravastatin released from the CLs was quantified by HPLC, as described in “Drug quantification methods” section.
Resveratrol-loaded MCLs ($$n = 4$$) were rinsed with NaCl $0.9\%$ and placed into 15 mL Falcon® tubes filled with 6 mL of NaCl $0.9\%$. At each predetermined timepoint, aliquots of 200 μL were removed and replaced with the same volume of fresh medium. After 8 h of release, 6 mL of fresh NaCl $0.9\%$ was added, increasing the total volume of release medium to 12 mL to avoid medium saturation. The amount of resveratrol released from the CLs was quantified by HPLC, as described in “Drug quantification methods” section.
## Drug quantification methods
The amount of pravastatin and resveratrol loaded and released from the CLs was quantified by HPLC using previously developed methods [22, 23]. In vitro release studies in vials were quantified by JASCO HPLC (AS-4140 autosampler, PU-4180 pump, LC-NetII/ADC interface box, CO-4060 column oven, MD-4010 photodiode array detector; JASCO, Tokyo, Japan) operated with the ChromNAV software v.2. In vitro release experiments using the eye blink model were quantified by Waters HPLC (Autosampler Waters 2690, Photodiode Detector 2996; Milford, MA, USA), operated with the Empower2 software.
In the case of pravastatin, the mobile phase consisted of methanol:0.02 M sodium phosphate (NaH2PO4) buffer (50:50 v/v, pH adjusted to 7.0 with NaOH) at 1.00 mL/min and 25 °C. For pravastatin analysis, both HPLC equipments were fitted with a Waters Symmetry C18 column (5 µm, 3.9 × 150 mm). The injection volume was 80 µL, and the total run time of each sample was 10 min. Pravastatin was quantified at 238 nm (retention time 5.15 min). The HPLC method was validated using pravastatin solutions in simulated lachrymal fluid between 1 and 40 µg/mL.
For resveratrol, the analysis was carried out under isocratic elution using a mobile phase of methanol:water 50:50 v/v at a flow rate of 1 mL/min, 35 °C, and with 8 min of run time. The injection volume was 80 μL, and the UV detector was set at 305 nm. The retention time was 4.6 min. Both HPLC equipment were fitted with a Waters Symmetry C18 column (5 μm, 4.6 × 250 mm). Validation of the method was performed using a calibration curve of resveratrol in ethanol:water 50:50 v/v in the 0.05–6 μg/mL range.
## Statistical analysis
Statistical analysis was performed using Statgraphics Centurion 18 v. 18.1.13 (Statgraphics Technologies, Inc., Warrenton, VA, USA). One-way analysis of variance (ANOVA) followed by multiple range test was carried out. The descriptive data were presented as mean ± standard deviation. In all cases, statistical significance was considered significant for a value of $p \leq 0.05.$
## Three-dimensional eye blink model
In this work, a 3D printed eye blink model was used to mimic some physiological ocular parameters, and in vivo comparisons were carried out. To mimic the tear fluid flow in the eye, fluid flow rates of 5 and 10 µL/min were chosen to ensure there was enough volume for sample collection at every predetermined time point. Preliminary trials carried out with flow rates closer to those reported for physiological tear flow values in humans (1.4–4.3 μL/min, [33]) or in rabbits (0.47–0.66 μL/min [34]) demonstrated that these rates were insufficient to maintain a reliable, constant flow rate of fluid on the eye model required for subsequent sample collections. It should be noted that the experiments with the eye blink model were carried out at room temperature as the system still lacks internal heating.
## Pravastatin
The amount of pravastatin loaded by the AECLs was approximately 3.50 ± 0.84 mg/g of dried hydrogel. Pravastatin stability against HHP sterilization was previously verified, and the chosen conditions did not trigger degradation compared to non-sterilized pravastatin CLs, which showed that this method is compatible with the sterilization of pravastatin-loaded CLs [30].
Pravastatin release was investigated in vitro by recording in parallel the amount of pravastatin released when the AECLs were placed in the eye blink model and in test tubes (Fig. 3a). In the in vitro eye blink model, the pravastatin release profiles were more sustained, and the percentage of drug released after 10 h of the experiment was lower compared to in vitro vial release. Statistically significant differences were detected between the percentage of drug released in a vial filled with 2 and 10 mL and the eye blink model (flow rate of 5 and 10 µL/min) in the first 3 h of the experiment (ANOVA, $p \leq 0.007$). The influence of the flow rate in the eye blink model on the drug released was tested with two different flow rates (5 and 10 µL/min), and a slight decrease in the percentage of pravastatin released was observed with the flow rate of 5 µL/min compared to a flow rate of 10 µL/min, but no statistical differences were detected (ANOVA, $p \leq 0.05$).
After 10 h of experimentation on the eye blink model, the CLs and eyelids were immersed in SLF in order to measure the amount of pravastatin retained in the materials (Fig. 3b). A significantly higher amount of pravastatin was detected in the PVA eyelid at 5 µL/min compared to the flow rate of 10 µL/min (ANOVA, $p \leq 0.001$). However, no significant differences were detected in the amount of drug that remained in the CLs for the different flow rates (ANOVA, $$p \leq 0.72$$).
Fig. 3a Pravastatin release profiles from AECLs experimentally recorded in a vial filled with 2 or 10 mL of SLF over 10 h and using the eye blink model (flow rate of 5 and 10 µL/min) and b amount of pravastatin retained in the PVA eyelid and CLs after 10 h on the eye blink model ($$n = 4$$, mean values and standard deviations). * Statistically significant differences in the amount of pravastatin released in vitro in 2 and 10 mL and in the eye blink model (flow rate of 5 and 10 µL/min); ** statistically different between the amount of pravastatin retained in the model eyelid with the flow rate of 5 and 10 µL/min, $p \leq 0.05$ For the in vitro vial release, two different volumes were evaluated, 10 and 2 mL of SLF. No statistically significant differences were detected between pravastatin release profiles in 2 and 10 mL of SLF (ANOVA, $p \leq 0.05$). This minor effect of the volume on the percentage of pravastatin released may be explained by the free-water solubility of pravastatin (40 mg/mL, [35]); therefore, the volume decrease did not induce a false plateau or delayed the release.
The differences between the percentages of pravastatin released in the vial and in the eye blink model could be attributed to several reasons such as the release volume and the absorption of the drug by the eye model materials, among others. Firstly, in the in vitro vial system, the CLs were immediately immersed into vials containing a volume 33 times higher than the volume delivered in the eye model (300 and 600 µL/h). This increase in fluid volume could influence the concentration gradient between the inside of the CL and the release medium, promoting a faster release in the first hours of the release. Secondly, in theory, the CLs in the eye blink model were exposed to a total fluid of 3.0 and 6.0 mL of SLF after 10 h. However, the amounts of fluid collected were 2.23 ± 0.21 mL and 5.47 ± 0.65 mL (for a flow rate of 5 and 10 µL/min, respectively), corresponding to a fluid loss of 25 and $9\%$. This nonspecific fluid loss due to evaporation, absorption by the eyelid, or dead volume could also contribute to a decrease in the fluid that reached the CL and consequently a decrease in the volume available for drug release. Thirdly, a portion of pravastatin was absorbed by the PVA-eyelid over 10 h of experiment, 2.42 ± 0.26 µg at 5 µL/min and 0.88 ± 0.49 µg at 10 µL/min. As a result, the drug release was slower in the eye blink model, especially in the first few hours.
Pravastatin release profiles from the AECLs during the in vivo experiment and in the eye blink model for both flow rates are compared in Fig. 5. The maximum concentration for both experiments was obtained after 30 min of CL application. Pravastatin maximum levels were 177.5 ± 116.8 μg/mL for in vivo and 28.39 ± 3.00 μg/mL and 39.13 ± 20.48 μg/mL for the eye blink model with a flow rate of 5 μL/min and 10 μL/min, respectively. The peak of maximum concentration was followed by a smooth decrease in drug concentration in the tear fluid and in the fluid collected from the eye blink model. No burst release was observed.
Fig. 5a In vivo tear fluid levels of pravastatin were recorded during wear of pravastatin-loaded AECLs for 8 and 10 h ($$n = 6$$ for 8 h and $$n = 3$$ for 9 and 10 h), data taken from Pereira-da-Mota et al. [ 22]. b Normalized released pravastatin concentration in SLF from pravastatin-loaded AECLs over 10 h on the eye model (flow rate of 5 and 10 µL/min) ($$n = 4$$, mean values and standard deviations) Despite the similar release patterns, the amount of pravastatin released from the CLs in the eye blink model was about fivefold lower than that recorded in the in vivo studies (ANOVA, $p \leq 0.05$). This finding might be related to several factors: (i) the drug absorption into the PVA eyelid in the model, (ii) the composition of the release medium, (iii) the temperature, and (iv) the complexity of the eyeball piece. Firstly, the PVA eyelid absorbed approximately 2.42 ± 0.26 µg pravastatin when tested under 5 µL/min and 0.89 ± 0.49 µg for 10 µL/min. Secondly, the physiological tear fluid contains proteins, lipids, and mucin that can promote drug release from the CLs [39]. In previous studies, the influence of BSA and lysozyme on pravastatin release rate from CLs was evaluated, and an increase in the amount of pravastatin released was observed with the incorporation of both proteins in 2 mL of SLF [22]. The effect of proteins, lipids, and other components of the tear fluid on the release kinetics from CLs is an important aspect to take into account in further studies. Thirdly, in this work, a stable room temperature of 20 ± 1.5 °C was maintained during the release experiments in the eye model, but an increase in temperature from 20 to 34 °C was previously shown to enhance $20\%$ of the fractional mass released from pHEMA hydrogels after 48 h in vitro in vials [40, 41]. This phenomenon could be related to higher kinetic energy of the drug molecules when the temperature increases, leading to faster diffusional transport [42]. The effect of temperature on pravastatin solubility could also contribute to a higher amount of drug released from the CLs in vivo [43]. Fourthly, another limitation that could compromise the release from CLs in the eye blink model pertains to the composition of the eyeball piece. In the present study, the eyeball piece was printed with a hydrophobic UV-polymerizable resin, which does not represent corneal surface properties. Ninety percent of the cornea consists of the stroma, which contains a high percentage of water and consists of collagen fibril lamellae oriented parallel to each other, which is more comparable to a hydrophilic hydrogel. In in vivo conditions, corneal osmosis adds water to the post-lens tear film (the tear film between the back surface of the CL and the corneal epithelium), diluting the concentration of the drug in the post-lens tear film, which could increase the drug release from the drug-loaded CL.
Comparing the release profiles from the different flow rates in the eye blink model, a slightly faster release was observed for the flow rate of 10 µL/min, achieving a higher maximum concentration at 30 min, followed by a decrease in the concentration of drug present in the fluid collected (ANOVA, $p \leq 0.05$).
Correlations for in vivo-in vitro in the eye blink model or in vitro–in vitro (in a vial versus eye blink model) were investigated using Levy plots (Fig. 6). Release tests in the eye blink model with a flow rate of 10 µL/min led to a correlation coefficient (r2) closer to 1 (0.993) compared to the Levy plot obtained for 5 µL/min (Fig. 6a). The intercepts at the origin were + 6.25 and + 1.93; the slopes were 1.83 and 1.62 for the flow rate of 5 and 10 µL/min, respectively. Thus, a stronger correlation was observed with the 10 µL/min flow rate that favored the in vivo-in vitro correlations. Also, a higher correlation coefficient was obtained for the Levy plot comparing the in vitro tests in a vial (2 and 10 mL) and eye blink model with the flow rate of 10 µL/min (Fig. 6b).Fig. 6Levy plots for in vivo (a) or in vitro (b) vs. eye blink model percentage of pravastatin released. The eye blink model experiments were carried out with 5 and 10 µL/min of flow rate (eye blink model 5/model 5 and eye blink model 10/model 10, respectively). The in vitro tests in a vial were carried out in 2 and 10 mL of SLF
## In vivo release eye blink model comparisons
The release profiles from the eye blink model for both pravastatin and resveratrol (concentration versus time) obtained were also compared to the in vivo rabbit data already reported for these same CLs [22, 29]. Briefly, six healthy male New Zealand white rabbits were selected for the in vivo release studies wearing drug-loaded CLs. Drug-loaded CLs were removed from the loading solutions, rinsed with sterile saline solution for CLs, and carefully placed on the rabbits’ right eye below the nictitating membrane and without local anesthesia. Samples of the tear fluid were collected using Schirmer test strips before and after CL wearing ($t = 5$, 15, 30 min, and every hour until 8 or 10 h). The drug concentration in the tear fluid was quantified by immersing the Schirmer strips in SLF or ethanol:water (50:50 v/v), and the resulting solutions were quantified by HPLC [22, 29].
## Conclusions
In the current study, the usefulness of the developed 3D printed eye blink model for the evaluation of drug release profiles of medicated CLs has been explored in detail. To our knowledge, this was one of the first studies that have attempted to correlate the release of drugs from lenses from an in vivo rabbit study with a blink model. It may serve as an important starting point to understand areas that need to be further developed and investigated to create better simulations. There are numerous factors in the eye that can affect drug release kinetics, and understanding the contribution/effect of each factor individually to drug release kinetics is not obvious. For instance, in this study, increasing the flow rate by 2 times was expected to increase the drug release rate by 2 times. However, this was not the case: increasing the flow by 2 times only resulted in a marginal increase in the amounts of drugs released, which was not obvious. The release profile of both drugs was more sustained and lower in the in vitro eye blink model compared to the in vitro release in vials, especially for the hydrophobic drug resveratrol. Both drug-loaded CLs showed similar release patterns in the eye blink model as in in vivo studies. However, the amount of drug released in the eye blink model was significantly lower compared to previously obtained in vivo data. More linear Levy plots were recorded for pravastatin release in the eye model (flow rate of 10 µL/min) and in vivo data. The information gathered in the present study may serve to gain an insight into relevant physiological parameters that influence the in vivo release from CLs, such as the composition of the tear fluid, tear flow rates, temperature of the system, and composition of the eyeball surface. The obtained results may serve as a guide for further improvements of the 3D printed eye blink model.
## References
1. 1.Wichterle O. U.S. Patents. 3,660,545; 3,408,429; 3,496,254; 3,499,862.
2. 2.https://www.jjvision.com/press-release/johnson-johnson-vision-care-receives-fda-approval-acuvuer-theravisiontm-ketotifen. Accessed Oct 2022.
3. Lanier OL, Christopher KG, Macoon RM, Yu Y, Sekar P, Chauhan A. **Commercialization challenges for drug eluting contact lenses**. *Expert Opin Drug Deliv* (2020.0) **17** 1133-1149. DOI: 10.1080/17425247.2020.1787983
4. 4.https://www.reviewofcontactlenses.com/article/material-gains-50-years-of-the-soft-contact-lens. Accessed Oct 2022.
5. Jones L, Hui A, Phan CM, Read ML, Azar D, Buch J, Ciolino JB, Naroo SA, Pall B, Romond K, Sankaridurg P, Schnider CM, Terry L, Willcox M. **CLEAR - contact lens technologies of the future**. *Cont Lens Anterior Eye* (2021.0) **44** 398-430. DOI: 10.1016/j.clae.2021.02.007
6. Minami T, Ishida W, Kishimoto T, Nakajima I, Hino S. **In vitro and in vivo performance of epinastine hydrochloride-releasing contact lenses**. *PLoS ONE* (2019.0) **14** e0210362. DOI: 10.1371/journal.pone.0210362
7. Sekar P, Chauhan A. **Effect of vitamin-E integration on delivery of prostaglandin analogs from therapeutic lenses**. *J Colloid Interface Sci* (2019.0) **539** 457-467. DOI: 10.1016/j.jcis.2018.12.036
8. Alvarez-Lorenzo C, Anguiano-Igea S, Varela-Garcia A, Vivero-Lopez M, Concheiro A. **Bioinspired hydrogels for drug-eluting contact lenses**. *Acta Biomater* (2019.0) **84** 49-62. DOI: 10.1016/j.actbio.2018.11.020
9. DiPasquale SA, Wuchte LD, Mosley RJ, Demarest RM, Voyles ML, Byrne ME. **One week sustained in vivo therapeutic release and safety of novel extended-wear silicone hydrogel contact lenses**. *Adv Healthc Mater* (2022.0) **11** e2101263. DOI: 10.1002/adhm.202101263
10. Hui A, Willcox M. **In vivo studies evaluating the use of contact lenses for drug delivery**. *Optom Vis Sci* (2016.0) **93** 367-376. DOI: 10.1097/OPX.0000000000000809
11. Pereira-da-Mota AF, Phan CM, Concheiro A, Jones L, Alvarez-Lorenzo C. **Testing drug release from medicated contact lenses: the missing link to predict in vivo performance**. *J Control Release* (2022.0) **343** 672-702. DOI: 10.1016/j.jconrel.2022.02.014
12. Minami T, Ishida W, Kishimoto T, Nakajima I, Hino S, Arai R, Matsunaga T, Fukushima A, Yamagami S. **In vitro and in vivo performance of epinastine hydrochloride-releasing contact lenses**. *PLoS ONE* (2019.0) **14** e0210362. DOI: 10.1371/journal.pone.0210362
13. Tieppo A, Pate KM, Byrne ME. **In vitro controlled release of an anti-inflammatory from daily disposable therapeutic contact lenses under physiological ocular tear flow**. *Eur J Pharm Biopharm* (2012.0) **81** 170-177. DOI: 10.1016/j.ejpb.2012.01.015
14. Pimenta AF, Valente A, Pereira JM, Pereira JC, Filipe HP, Mata JL, Colaço R, Saramago B, Serro AP. **Simulation of the hydrodynamic conditions of the eye to better reproduce the drug release from hydrogel contact lenses: experiments and modeling**. *Drug Deliv Transl Res* (2016.0) **6** 755-762. DOI: 10.1007/s13346-016-0303-1
15. Phan CM, Walther H, Qiao H, Shinde R, Jones L. **Development of an eye model with a physiological blink mechanism**. *Trans Vis Sci Tech* (2019.0) **8** 1. DOI: 10.1167/tvst.8.5.1
16. Phan CM, Shukla M, Walther H, Heynen M, Suh D, Jones L. **Development of an in vitro blink model for ophthalmic drug delivery**. *Pharmaceutics* (2021.0) **13** 300. DOI: 10.3390/pharmaceutics13030300
17. Phan CM, Walther H, Smith RW, Riederer D, Lau C, Osborn Lorenz K, Subbaraman LN, Jones L. **Determination of the release of PEG and HPMC from nelfilcon A daily disposable contact lenses using a novel in vitro eye model**. *J Biomater Sci Polym Ed* (2018.0) **29** 2124-2136. DOI: 10.1080/09205063.2018.1514192
18. Phan CM, Bajgrowicz-Cieslak M, Subbaraman LN, Jones L. **Release of moxifloxacin from contact lenses using an in vitro eye model: impact of artificial tear fluid composition and mechanical rubbing**. *Transl Vis Sci Technol* (2016.0) **5** 3. DOI: 10.1167/tvst.5.6.3
19. Phan CM, Bajgrowicz M, Gao H, Subbaraman LN, Jones LW. **Release of fluconazole from contact lenses using a novel in vitro eye model**. *Optom Vis Sci* (2016.0) **93** 387-394. DOI: 10.1097/OPX.0000000000000760
20. Murphy C, Deplazes E, Cranfield CG, Garcia A. **The role of structure and biophysical properties in the pleiotropic effects of statins**. *Int J Mol Sci* (2020.0) **21** 8745. DOI: 10.3390/ijms21228745
21. Yang SC, Tseng CH, Wang PW, Lu PL, Weng YH, Yen FL, Fang JY. **Pterostilbene, a methoxylated resveratrol derivative, efficiently eradicates planktonic, biofilm, and intracellular MRSA by topical application**. *Front Microbiol* (2017.0) **8** 1103. DOI: 10.3389/fmicb.2017.01103
22. 22.Pereira-da-Mota AF, Vivero-Lopez M, Serramito M, Diaz-Gomez L, Serro AP, Carracedo G, Huete-Toral `F, Concheiro A, Alvarez-Lorenzo C. Contact lenses for pravastatin delivery to eye segments: design and in vitro-in vivo correlations. J Control Release. 2022;348:431–43. 10.1016/j.jconrel.2022.06.001.
23. Vivero-Lopez M, Muras A, Silva D, Serro AP, Otero A, Concheiro A, Alvarez-Lorenzo C. **Resveratrol-loaded hydrogel contact lenses with antioxidant and antibiofilm performance**. *Pharmaceutics* (2021.0) **13** 532. DOI: 10.3390/pharmaceutics13040532
24. Ooi KG, Khoo P, Vaclavik V, Watson SL. **Statins in ophthalmology**. *Surv Ophthalmol* (2019.0) **64** 401-432. DOI: 10.1016/j.survophthal.2019.01.013
25. Gupta A, Gupta V, Thapar S, Bhansali A. **Lipid-lowering drug atorvastatin as an adjunct in the management of diabetic macular edema**. *Am J Ophthalmol* (2004.0) **137** 675-682. DOI: 10.1016/j.ajo.2003.11.017
26. Ozkiris A, Erkiliç K, Koç A, Mistik S. **Effect of atorvastatin on ocular blood flow velocities in patients with diabetic retinopathy**. *Br J Ophthalmol* (2007.0) **91** 69-73. DOI: 10.1136/bjo.2006.098285
27. 27.Nielsen SF, Nordestgaard BG. Statin use before diabetes diagnosis and risk of microvascular disease: a nationwide nested matched study. Lancet Diabetes Endocrinol. 2014;2(11):894–900. 10.1016/S2213-8587(14)70173-1.
28. Tsai TY, Chen TC, Wang IJ, Yeh CY, Su MJ, Chen RH, Tsai TH, Hu FR. **The effect of resveratrol on protecting corneal epithelial cells from cytotoxicity caused by moxifloxacin and benzalkonium chloride**. *Invest Ophthalmol Vis Sci* (2015.0) **56** 1575-1584. DOI: 10.1167/iovs.14-15708
29. 29.Vivero-Lopez M, Pereira-da-Mota AF, Carracedo G, Huete-Toral F, Parga A, Otero A, Concheiro A, Alvarez-Lorenzo C. Phosphorylcholine-based contact lenses for sustained release of resveratrol: design, antioxidant and antimicrobial performances, and in vivo behavior. ACS Appl Mater Interf. 2022. 10.1021/acsami.2c18217
30. Pereira-da-Mota AF, Vivero-Lopez M, Topete A, Serro AP, Concheiro A, Alvarez-Lorenzo C. **Atorvastatin-eluting contact lenses: effects of molecular imprinting and sterilization on drug loading and release**. *Pharmaceutics* (2021.0) **13** 606. DOI: 10.3390/pharmaceutics13050606
31. 31.Phan CM, Walther H, Gao H, Rossy J, Subbaraman LN, Jones L. Development of an in vitro ocular platform to test contact lenses. J Vis Exp. 2016;6(110):e53907. 10.3791/53907.
32. 32.Kwon KA, Shipley RJ, Edirisinghe M, Ezra DG, Rose G, Best SM, Cameron RE. High-speed camera characterization of voluntary eye blinking kinematics. J R Soc Interface. 2013;10(85):20130227. 10.1098/rsif.2013.0227.
33. Glasson MJ, Stapleton F, Keay L, Willcox MD. **The effect of short term contact lens wear on the tear film and ocular surface characteristics of tolerant and intolerant wearers**. *Cont Lens Anterior Eye* (2006.0) **29** 41-47. DOI: 10.1016/j.clae.2005.12.006
34. 34.Chrai SS, Patton TF, Mehta A, Robinson JR. Lacrimal and instilled fluid dynamics in rabbit eyes. J Pharm Sci. 1973;62(7):1112–21. 10.1002/jps.2600620712.
35. 35.European Directorate for the Quality of Medicines & HealthCare. European Pharmacopeia. 6th ed. Council of Europe. 2008.
36. Zupančič Š, Lavrič Z, Kristl J. **Stability and solubility of trans-resveratrol are strongly influenced by pH and temperature**. *Eur J Pharm Biopharm* (2015.0) **93** 196-204. DOI: 10.1016/j.ejpb.2015.04.002
37. Silva RCD, Teixeira JA, Nunes WDG, Zangaro GAC, Pivatto M, Caires FJ, Ionashiro M. **Resveratrol: a thermoanalytical study**. *Food Chem* (2017.0) **237** 561-565. DOI: 10.1016/j.foodchem.2017.05.146
38. Salehi B, Mishra AP, Nigam M, Sener B, Kilic M, Sharifi-Rad M, Fokou PVT, Martins N, Sharifi-Rad J. **Resveratrol: a double-edged sword in health benefits**. *Biomedicines* (2018.0) **6** 91. DOI: 10.3390/biomedicines6030091
39. Mahomed A, Wolffsohn JS, Tighe BJ. **Structural design of contact lens-based drug delivery systems; in vitro and in vivo studies of ocular triggering mechanisms**. *Cont Lens Anterior Eye* (2016.0) **39** 97-105. DOI: 10.1016/j.clae.2015.07.007
40. Tieppo A, Boggs AC, Pourjavad P, Byrne ME. **Analysis of release kinetics of ocular therapeutics from drug releasing contact lenses: best methods and practices to advance the field**. *Cont Lens Anterior Eye* (2014.0) **37** 305-313. DOI: 10.1016/j.clae.2014.04.005
41. Miyazaki S, Suzuki S, Kawasaki N, Endo K, Takahashi A, Attwood D. **In situ gelling xyloglucan formulations for sustained release ocular delivery of pilocarpine hydrochloride**. *Int J Pharm* (2001.0) **229** 29-36. DOI: 10.1016/s0378-5173(01)00825-0
42. Ferreira JA, Oliveira P, Silveira E. **Drug release enhanced by temperature: an accurate discrete model for solutions in H3**. *Comput Math Appl* (2020.0) **79** 852-875. DOI: 10.1016/j.camwa.2019.08.002
43. Jia C, Yin Q, Song J, Hou G, Zhang M. **Solubility of pravastatin sodium in water, methanol, ethanol, 2-propanol, 1-propanol, and 1-butanol from (278 to 333) K**. *J Chem Eng Data* (2008.0) **53** 2466-2468. DOI: 10.1021/je800196k
44. Torres-Luna C, Fan X, Domszy R, Hu N, Wang NS, Yang A. **Hydrogel-based ocular drug delivery systems for hydrophobic drugs**. *Eur J Pharm Sci* (2020.0) **154** 105503. DOI: 10.1016/j.ejps.2020.105503
|
---
title: Variant spectrum of PIEZO1 and KCNN4 in Japanese patients with dehydrated hereditary
stomatocytosis
authors:
- Erina Nakahara
- Keiko Shimojima Yamamoto
- Hiromi Ogura
- Takako Aoki
- Taiju Utsugisawa
- Kenko Azuma
- Hiroyuki Akagawa
- Kenichiro Watanabe
- Michiko Muraoka
- Fumihiko Nakamura
- Michi Kamei
- Koji Tatebayashi
- Jun Shinozuka
- Takahisa Yamane
- Makoto Hibino
- Yoshiya Katsura
- Sonoko Nakano-Akamatsu
- Norimitsu Kadowaki
- Yoshiro Maru
- Etsuro Ito
- Shouichi Ohga
- Hiroshi Yagasaki
- Ichiro Morioka
- Toshiyuki Yamamoto
- Hitoshi Kanno
journal: Human Genome Variation
year: 2023
pmcid: PMC9981561
doi: 10.1038/s41439-023-00235-y
license: CC BY 4.0
---
# Variant spectrum of PIEZO1 and KCNN4 in Japanese patients with dehydrated hereditary stomatocytosis
## Abstract
Hereditary stomatocytosis (HSt) is a type of congenital hemolytic anemia caused by abnormally increased cation permeability of erythrocyte membranes. Dehydrated HSt (DHSt) is the most common subtype of HSt and is diagnosed based on clinical and laboratory findings related to erythrocytes. PIEZO1 and KCNN4 have been recognized as causative genes, and many related variants have been reported. We analyzed the genomic background of 23 patients from 20 Japanese families suspected of having DHSt using a target capture sequence and identified pathogenic/likely pathogenic variants of PIEZO1 or KCNN4 in 12 families.
## Hemolytic anemia: new variants in hereditary subtypes
Genetic variants identified in 12 Japanese families are likely to be responsible for a rare hereditary condition causing red blood cells (RBCs) to rupture. Keiko Shimojima Yamamoto of Tokyo Women’s Medical University, Japan, and colleagues conducted genetic analyses on blood samples from 23 people in 20 Japanese families with suspected dehydrated hereditary spherocytosis (DHSt), caused by RBC ion imbalances leading to dehydration and rupture. In 12 families, the analyses revealed probable disease-causing variants in two genes, PIEZO1 and KCNN4, some of which were identified for the first time. *Both* genes are related to RBC ion channels and the mutations lead to two slightly different forms of DHSt. Further understanding could improve treatment strategies. Comprehensive genomic analysis is a powerful tool for understanding the genetic cause of congenital hemolytic anemia.
## Introduction
Hereditary stomatocytosis (HSt) is a type of congenital hemolytic anemia caused by abnormally increased cation permeability of erythrocyte membranes1,2. The most common subtype of HSt (dehydrated HSt [DHSt] or hereditary xerocytosis [HX]) is diagnosed by screening tests, such as evaluation of erythrocyte morphology, measurement of the cation concentration inside and outside the erythrocyte membrane, or osmotic gradient ektacytometry3.
DHSt is an autosomal dominant hemolytic anemia characterized by abnormally shaped red blood cells (RBCs) and associated with primary erythrocyte dehydration4. DHSt is thought to be rare, and a prevalence estimate of 1:50,000 has been suggested1. This condition is characterized by mild to moderate hemolysis with varying numbers of stomatocytes on peripheral blood smears, which are sometimes rare, ill-formed, and likely overlooked. The reticulocyte count is elevated, and the mean cellular hemoglobin concentration (MCHC) and mean cell hemoglobin content (MCH) are increased. Paradoxically, the red cell mean corpuscular volume (MCV) is slightly increased5. Patients may also present with a history of perinatal edema and show pseudohyperkalemia due to the loss of potassium ions from RBCs stored at room temperature. Complications such as splenomegaly and cholelithiasis may occur due to increased trapping of RBCs in the spleen and elevated bilirubin levels, respectively. Furthermore, DHSt is frequently associated with iron overload, which may lead to hepatosiderosis6, diabetes mellitus, failure of the pituitary gland, and heart failure.
In 2012, DHSt was first identified as being related to alterations in the piezo-type mechanosensitive ion channel component 1 gene (PIEZO1; MIM* 611184)7. PIEZO1 encodes a mechanosensitive ion channel that translates a mechanic stimulus into calcium influx7. The identified missense variants showed the gain-of-function PIEZO1 phenotype, providing insight to help explain the increased permeability of cations in RBCs of patients with DHSt8. Dehydrated hereditary stomatocytosis 1 with or without pseudohyperkalemia and/or perinatal edema (DHS1: OMIM#194380) is a dominantly inherited red cell membrane disorder caused by gain-of-function mutations of PIEZO1 in most cases.
In 2015, another gene, the potassium intermediate/small conductance calcium-activated channel, subfamily N, member 4 gene (KCNN4; MIM*602754), encoding the calcium ion-dependent potassium selective Gardos channel, was identified as being associated with DHSt9. Dehydrated hereditary stomatocytosis-2 (DHS2, OMIM#616689) is caused by a heterozygous mutation in KCNN4.
We participated in clinical research on patients with hereditary red cell membrane disorders. For this purpose, we developed a target capture sequencing (TCS) system for precise and comprehensive diagnosis of suspected hereditary red cell membrane disorders in patients10. Previously, we reported the genetic background of hereditary spherocytosis, which can be distinguished by morphological characteristics of RBCs and osmotic abnormalities of RBC membranes10. Here, we report the genomic variants identified in Japanese patients with HSt, especially those related to DHSt.
## Materials and methods
From April 2015 to June 2021, 20 Japanese families with suspected DHSt were enrolled in this study. This study was performed in accordance with the principles of the Declaration of Helsinki and approved by the ethics committee of the institution. After obtaining written informed consent, blood samples were collected from all patients. In addition, we collected detailed clinical information from the attending doctors, including family histories, clinical courses, and physical findings.
In most patients, when possible, we first performed additional red cell membrane functional examinations, including the acidified glycerol hemolysis time (AGLT) test, flow-cytometric osmotic fragility (FCM-OF) test, and eosin-5’-maleimide (EMA) binding test with a negative direct antiglobulin test as per previously reported methods10. DHSt was suspected when clinical findings such as hemolytic anemia with stomatocytosis and hemochromatosis not due to transfusion, positive family history, and past history of perinatal edema were observed, and laboratory tests revealed elevated MCV, increased % residual red cells (%RRC) in the FCM-OF test, and normal or increased EMA binding.
Genomic DNA was extracted from the patient’s peripheral blood using a QIAamp DNA extraction kit according to the manufacturer’s instructions (QIAGEN, Hilden, Germany). The Haloplex HS target enrichment system (Agilent Technologies, Santa Clara, CA, USA) was used for TCS. Using SureDesign (https://earray.chem.agilent.com/suredesign/home.htm), the target panel was designed to include all coding exons and intron‒exon boundaries of the 74 possible candidate genes10. Massive parallel sequencing was performed using the Illumina MiSeq platform (Illumina Inc., San Diego, CA, USA). Raw data were aligned to the human genome sequence GRCh37/hg19. *The* generated FASTQ files were imported into SureCall v3.5 (Agilent Technologies) for variant calling. Analysis following the filtering of the obtained variants was described previously10. The obtained variants were filtered according to the following strategy: [1] variant frequencies were below $1\%$ in 1000G_EAS and ALL (1000 Genomes), HGDV, and dbSNP; [2] synonymous variants were excluded (nonsynonymous variants, variants associated with frameshift, insertion/deletion variants, and variants in splicing donor/acceptor sites were included); [3] variants with allele frequencies less than $30\%$ of the total read depth were excluded; and [4] the CADD_phred was higher than 20 if obtained. Variant information obtained using wANNOVAR (http://wannovar.wglab.org/) was used for curation. Integrative Genomics Viewer (https://software.broadinstitute.org/software/igv/) was used for visual evaluation. All variants were evaluated using the guidelines proposed by the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP)11.
The existence of the identified variants in the probands of the enrolled patients was confirmed using conventional PCR-Sanger sequencing. Genotyping for αLELY (low expression allele of SPTA1), UGT1A1, and Memphis I and II (SLC4A1) was also performed using conventional PCR-Sanger sequencing for all patients12–15.
## Results
Among the 20 examined families, 12 were shown to have pathogenic or likely pathogenic variants of PIEZO1 or KCNN4 (diagnosis ratio of $60\%$) in accordance with the ACMG/AMP guidelines11. All the variants were confirmed by Sanger sequencing (Supplemental Figs. S1 and S2). The patients’ clinical and genetic information is summarized in Table 1.Table 1Clinical information and the results of this study. Patient numberPatient 1Patient 2Son of patient 2Patient 3Patient 4Daugther of patient 4Mother of patient 4Patient 5Patient 6Patient 7Patient 8Patient 9Patient 10Patient 11Patient 12GenderFMMFMFFMFFMFFFMAge at examination47 y39 y8 m2 m68 y31 y89 y69 y19 y16 y28 y15 y68 y14 y8 yFamily history−++−+++++−+−NA+−Clinical features Splenomegaly−−−NANA+−+−−+−+−NA Splenectomy−+−−NANANA−−−−−−+− GallstoneNA+NA−NANANA+NANANA−+NA+ Previous blood transfusion−+−+−NA++−++++++ Other findingsDiabetes mellitus, failure of the pituitary glandFailure of the pituitary gland, Infertility, Heart failure, Cerebral infarction, Renal infarctionFetal edemaFetal edemaDiabetes mellitusExtramedullary hematopoiesisDiabetes mellitusRBC morphology Target cell+++−+NANA++++++++ Stomatocyte+++−+NANA+++++++− Acanthocyte−+−+−NANA−−−−−−+− Elliptocyte−−−−−NANA−+−−−−−+ Poikilocyte−−−−−NANA−−−−−−−+ Schistocyte−−+−−NANA−−+−−−−− Anisocyte−−−++NANA−−−++−−− Polychromasia−−−+−NANA−−−−−−−− Nucleated erythrocyte−+−−−NANA−−−−−−−−Laboratory findings Hb (g/dL)[13 <: male, 12 <: female]12.38.410.010.514.612.76.913.515.611.010.19.78.510.710.1 MCV (fL)[86–98]112.3115.688.085.093.391.790.8107.192.7103.6107.4107.1108.395.386.6 MCHC (%)[31–35]34.634.436.034.937.637.039.035.836.038.234.833.034.437.735.6 Reticulocytes (‰)[0.2–2.7]10.412.68.42.58.312.78.57.113.86.711.58.016.723.96.4 LDH (U/L)[240–490]186284236308202173187157144142175139156191239 Haptoglobin (mg/dL)[19–170]97.53.0NANA86.0NANA16.056.013.02.018.0<10NA<10 Total bilirubin (mg/dL)[0.2–1.2]1.737.90.50.62.11.81.43.41.36.97.32.96.66.55.3 Indirect Bilirubin (mg/dL)[0.2–1.0]1.315.70.2NA1.81.20.52.60.96.76.82.85.46.34.8 Ferritin[20–250: male, 10–80: female]16122537.3277.3NA3895172489649.5108.7305.11663NA2583.0NA87.1Membrane examination AGLT[30 min <]NANANANANANANA30 min<30 min<NANANANA30 min<30 min< EMA (% of control)114NANA89103NANA113971121111131129993 FCM-OF (% of control)193NANA219122NANA132143112113186179142125Identified variants GenesPIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1PIEZO1KCNN4KCNN4 RefSeq IDNM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_001142864.3NM_002250.2NM_002250.2 Affected exonsExon 11Exon 14Exon 14Exon 32Exon 42Exon 42Exon 42Exon 48Exon 51Exon 51Exon 51Exon 51Exon 51Exon 5Exon 7 cDNA changec.1281_1282ins[GG;1257_1281]c.1792G > Ac.1792G > Ac.4370 C > Tc.6041 C > Tc.6041 C > Tc.6041 C > Tc.6968 A > Cc.7463 G > Ac.7483_7488dupc.7483_7488dupc.7483_7488dupc.7483_7488dupc.835 G > Ac.1055 G > A Protein changep. A427_L428insGMDQSYVCAp. V598Mp. V598Mp. A1457Vp. T2014Ip. T2014Ip. T2014Ip. K2323Tp. R2488Qp. L2495_E2496dupp. L2495_E2496dupp. L2495_E2496dupp. L2495_E2496dupp. A279Tp. R352H TypeInsertionMissenseMissenseMissenseMissenseMissenseMissenseMissenseMissenseDuplicationDuplicationDuplicationDuplicationMissenseMissense StatusHeteroHeteroHeteroHeteroHeteroHeteroHeteroHeteroHeteroHeteroHeteroHeteroHeteroHeteroHetero dbSNP ID−−−rs532444891−−−−rs749288233rs1064793545rs1064793546rs1064793547rs1064793548−rs774455945 SIFT (score)NA0.0040.0040.0190.0170.0170.0170.0830NANANANA0.2370 Polyphen2 (score)NA110.060.0470.0470.0470.430.999NANANANA0.9990.999 CADD_phredNA313124.328.628.628.626.735NANANANA2635 ClinvarNot reportedPathogenic/Likely pathogenicPathogenic/Likely pathogenicNot reportedNot reportedNot reportedNot reportedNot reportedNot reportedNot reportedNot reportedNot reportedNot reportedNot reportedPathogenic ACMG criteriaPM2, PM4, PM6, PP4PS1, PS3, PM2, PP3, PP4PS1, PS3, PM2, PP3, PP4PM2, PM6, PP3, PP4PS4, PM2, PP3, PP4PS4, PM2, PP3, PP4PS4, PM2, PP3, PP4PS4, PM2, PP3, PP4PS1,PS3, PM2, PP1, PP3, PP4PS1, PM2, PM4, PP1, PP3, PP4PS1, PM2, PM4, PP1, PP3, PP4PS1, PM2, PM4, PP1, PP3, PP4PS1, PM2, PM4, PP1, PP3, PP4PS1, PM5, PP2, PP3, PP4PS1, PS3, PM6, PP3, PP4 InterpretationLikely pathogenicPathogenicPathogenicLikely pathogenicLikely pathogenicLikely pathogenicLikely pathogenicLikely pathogenicPathogenicPathogenicPathogenicPathogenicPathogenicLikely pathogenicPathogenic Reported/novelNovelReportedReportedNovelReportedReportedReportedNovelReportedReportedReportedReportedReportedNovelReportedStatus of other polymorphic variants αLELY variants (c.5572 C > G)−HeteroHomoHetero−NANA−−−HomoHetero−Hetero− αLELY variants (c.6531–12 C > T)−HeteroHomoHetero−NANA−−−HomoHetero−Hetero− UGT1A1 variants (*6)−−−−−NANA−−−−Hetero−−− UGT1A1 variants (*28)−−−Hetero−NANA−−−−−Hetero−Homo Memphis I (SLC4A1: c.166 A > G)−−−−−NANAHeteroHomoHetero−−Hetero−− Memphis II (SLC4A1: c.2561 C > T)−−−−−NANA−Hetero−−−−−HeteroPt patient, F female, M male, y years, m months, NA not available, [] normal range, _ underbar indicates abnormal finding.
Ten families showed seven types of heterozygous PIEZO1 variants (one insertion, five missense, and one in-frame duplication). Among them, two variants (p.A427_L428insGMDQSYVCA and p.K2323T) were novel and not included in the databases of ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) and gnomAD (http://www.gnomad-sg.org/). The p.A1457V variant identified in Patient 3 was included in the dbSNP database (https://www.ncbi.nlm.nih.gov/snp/) with ID = rs532444891. The minor allele frequency in the Japanese population was 0.00155, indicating that it is very rare in the general population. It is also identified in the Human Genetic Variation Database (https://www.hgvd.genome.med.kyoto-u.ac.jp/index.html), the variation database of the Japanese population, with a very low incidence16. Because no reports suggesting the p.A1457V variant as the disease-causing variant exists, we first report this as a novel disease-causing variant in this study.
Four variants (p.V598M, p.T2014I, p.R2488Q, p.L2495_E2496dup) were previously reported6,8,17–19. The p.V598M and p.T2014I were segregated in the families of Patients 2 and 4, respectively. The p.R2488Q variant is often reported as a disease-causing variant by many researchers and is registered in the dbSNP database with ID = rs7492882336,18. However, this has not been reported in ClinVar.
Heterozygous KCNN4 variants were identified in two families. The variant p.R352H has been previously reported9,20. In KCNN4, p.R352H erythrocytes, preliminary data also suggested that altered channel activation kinetics led to erythrocyte dehydration9. Thus, we first report p.A279T as a novel disease-causing variant in this study.
## Discussion
In this study, 12 Japanese families suspected of having DHSt were genetically diagnosed with causative variants related to DHSt. Among them, 10 families ($83\%$) had PIEZO1 variants. Most PIEZO1 variants were first identified in this study in Japanese patients.
Four of the ten families ($40\%$) shared the recurrent variant (p.L2495_E2496dup). This variant was previously reported in a Japanese family associated with hereditary high phosphatidylcholine, hemolytic anemia, and hemochromatosis-induced diabetes mellitus21. The p.L2495_E2496dup is located at the junction of the α2 and α3 intracellular COOH-terminal domains, which is predicted to be involved in pore formation of the ion channel22. This variant causes changes in the hydropathicity profile in the regions where it is located, suggesting possible structural change23. Picard et al. described the clinical, hematologic and genetic characteristics of a retrospective series of 126 subjects from 64 families with DHSt20. Among them, 19 families showed PIEZO1 variants, and 10 of 19 families ($53\%$) had p.L2495_E2496dup. Although this variant has been reported frequently worldwide8,18,20–24, it is not included in ClinVar.
The other recurrent PIEZO1 variants (p.V598M, p.T2014I, p.R2488Q) were first identified in the Japanese population, and two of them (p.V598M and p.T2014I) were confirmed to be segregated in the families; however, it is still unclear whether they came from the same founder. Further analyses would be needed.
PIEZO1 encodes a mechanosensitive ion channel that translates a mechanical stimulus into calcium influx and is related to DHS1, which is a dominantly inherited red cell membrane disorder. No PIEZO1 variant associated with loss-of-function (LoF) was found, and the pLI score of PIEZO1 was “0” in gnomAD. Therefore, PIEZO1 is tolerant to LoF, and the gain-of-function mechanism is considered the mechanism rather than LoF, as mentioned above. Indeed, functional studies of DHS1-associated PIEZO1 variants exhibited a partial gain-of-function phenotype, with many mutants demonstrating delayed channel inactivation8.
Although patients with DHSt often have fully or partially compensated hemolysis with few symptoms, iron overload is a universal finding, even in patients without transfusions or with only sporadic blood transfusions, and this causes progressive organ damage25. All patients with PIEZO1 variations in this study showed elevated levels of ferritin, except a young 8-year-old patient. It was previously reported that ferritin levels at diagnosis were correlated with the age of patients20. The ferritin level of our patients also reflects this trend. Ma et al. showed that constitutive or macrophage expression of a gain-of-function Piezo1 allele in mice disrupts levels of the iron regulator hepcidin and causes iron overload26. They further show that PIEZO1 is a key regulator of macrophage phagocytic activity and subsequent erythrocyte turnover26. Their discovery may be a new seed to treat hyperferritinemia in DHS1 patients.
In contrast, KCNN4 encodes a Ca2 +-activated K+ channel. Although KCNN4 has been reported to be associated with some diseases, including inflammatory bowel disease, Crohn’s disease, and Alzheimer’s disease, germline pathogenic variants in KCNN4 have only been shown to be associated with DHS227–30. To date, ten KCNN4 variants (p.V222L, p.V282M, p.V282E, p.S314P, p.A322V, p.H328R, p.H340R, p.H340N, p.R352H, p.V369_Lys373del) have been reported in patients with DHS2 9,18,20,24,31,32. Among them, p.R352H identified in patient 12 in our study has been recurrently identified20. The novel p.A279T, first identified in our study, is located near p.V282M and p.V282E. The 231-289th amino acids of KCNN4 form a two-pore potassium channel domain, and the 304-377th amino acids form a calmodulin-binding domain24,33. These regions are highly conserved and have an important role in KCNN4.
In patients’ RBCs that carry KCNN4 variations, the channel conductance is considered to be increased. Despite leading to a more active channel, the gain-of-function in KCNN4 is not systematically linked to RBC dehydration, and routine hematological tests have failed to clearly diagnose DHSt20. Rivera et al. analyzed the characteristics of two de novo KCNN4 variants (p.V222L and p.H340N)32. However, the data did not correlate with RBC dehydration caused by KCNN4 gain-of-function, raising the question of whether this pathology should be classified as a DHSt. Moreover, it emphasized the difficulty of diagnosing altered RBC permeability facing KCNN4 variants and the great variability in RBC phenotypes associated with the KCNN4 gain-of-function mechanism32.
It is important to distinguish DHSt from hereditary spherocytosis (HS). Splenectomy should be avoided in patients with DHSt because it seems to aggravate the risk of thrombosis25. Especially in DHS1, postsplenectomy thrombotic events are called a major risk20. In our study, thrombotic events were frequent in splenectomized Patient 2 with DHS1. On the other hand, no thrombotic events occurred in splenectomized Patient 14 with DHS2. Picard et al. also reported that none of the four DHS2 splenectomized patients experienced thrombosis20. To date, the number of reported cases of DHS2 is lower than that for DHS1. However, we should not allow definitive conclusions to be drawn on this issue. Further evaluation of case information is needed, and patients splenectomized before being genetically diagnosed should be carefully monitored for a long time.
In this study, two patients (Patient 2 and Patient 3) with DHS1 had a history of fetal edema. A previous report showed that perinatal edema was observed in DHS1 but not in DHS2 patients20. The severity of perinatal edema is heterogeneous, so careful pregnancy follow-up with ultrasound monitoring is needed in both genotypes34.
We performed functional examinations of the RBC membrane. It is known that the EMA test and FCM-OF test are good combinations for the diagnosis of HS35. Recently, the results of these tests for DHSt patients were reported. Zama et al. showed that the result of the EMA test of DHSt patients is normal or increased36. Our data are consistent with their results.
In DHSt, a dysfunctional membrane protein eventually leads to potassium leakage out of the RBCs that exceeds the inward flux of sodium, and the accompanying net loss of water results in RBC dehydration, shrinkage, fragility, and hemolysis25. The results of the FCM-OF test in our patients reflect this pathology.
Although we genotyped known polymorphisms of αLELY, UGT1A1, and Memphis I and II, we could not identify any correlation between clinical findings and severity.
Recently, ABCB6 has been reported to be responsible for familial pseudohyperkalemia, a disorder related to DHSt37. However, this gene was not included in the gene panel used in this study. Therefore, it is the subject of our future project to investigate whether ABCB6 is related to patients with DHSt without disease-causing variants of PIEZO1 and KCNN4.
In conclusion, here, we first report the hematological, clinical, and genetic features of DHSt in Japan. Comprehensive genomic analysis is a powerful tool for understanding the genetic cause of congenital hemolytic anemia and would be beneficial for the molecular diagnosis and clinical management of DHSt.
## Supplementary information
Image of the elecropherogram of Sanger sequencing Images of the elecropherograms of Sanger sequencing The online version contains supplementary material available at 10.1038/s41439-023-00235-y.
## References
1. Andolfo I, Russo R, Gambale A, Iolascon A. **New insights on hereditary erythrocyte membrane defects**. *Haematologica* (2016.0) **101** 1284-1294. DOI: 10.3324/haematol.2016.142463
2. Narla J, Mohandas N. **Red cell membrane disorders**. *Int. J. Lab. Hematol.* (2017.0) **39** 47-52. DOI: 10.1111/ijlh.12657
3. Vives-Corrons JL, Krishnevskaya E, Rodriguez IH, Ancochea A. **Characterization of hereditary red blood cell membranopathies using combined targeted next-generation sequencing and osmotic gradient ektacytometry**. *Int. J. Hematol.* (2021.0) **113** 163-174. DOI: 10.1007/s12185-020-03010-9
4. Mohandas N. **Inherited hemolytic anemia: a possessive beginner’s guide**. *Hematol. Am. Soc. Hematol. Educ. Program.* (2018.0) **2018** 377-381. DOI: 10.1182/asheducation-2018.1.377
5. Iolascon A, Andolfo I, Russo R. **Advances in understanding the pathogenesis of red cell membrane disorders**. *Br. J. Haematol.* (2019.0) **187** 13-24. DOI: 10.1111/bjh.16126
6. Andolfo I. **Multiple clinical forms of dehydrated hereditary stomatocytosis arise from mutations in PIEZO1**. *Blood* (2013.0) **121** 3925-3935. DOI: 10.1182/blood-2013-02-482489
7. Zarychanski R. **Mutations in the mechanotransduction protein PIEZO1 are associated with hereditary xerocytosis**. *Blood* (2012.0) **120** 1908-1915. DOI: 10.1182/blood-2012-04-422253
8. Albuisson J. **Dehydrated hereditary stomatocytosis linked to gain-of-function mutations in mechanically activated PIEZO1 ion channels**. *Nat. Commun.* (2013.0) **4** 1884. DOI: 10.1038/ncomms2899
9. Rapetti-Mauss R. **A mutation in the Gardos channel is associated with hereditary xerocytosis**. *Blood* (2015.0) **126** 1273-1280. DOI: 10.1182/blood-2015-04-642496
10. Shimojima Yamamoto K. **Clinical and genetic diagnosis of thirteen Japanese patients with hereditary spherocytosis**. *Hum. Genome Var.* (2022.0) **9** 1. DOI: 10.1038/s41439-021-00179-1
11. Richards S. **Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology**. *Genet Med.* (2015.0) **17** 405-424. DOI: 10.1038/gim.2015.30
12. Delaunay J. **Different impacts of alleles alphaLEPRA and alphaLELY as assessed versus a novel, virtually null allele of the SPTA1 gene in trans**. *Br. J. Haematol.* (2004.0) **127** 118-122. DOI: 10.1111/j.1365-2141.2004.05160.x
13. 13.Kaliniczenko, A. et al. Frequency of the DI*A, DI*B and Band 3 Memphis polymorphism among distinct groups in Brazil. Hematol. Transfus. Cell. Ther.S2531–1379, 00052–00059 (2022).
14. Rets A, Clayton AL, Christensen RD, Agarwal AM. **Molecular diagnostic update in hereditary hemolytic anemia and neonatal hyperbilirubinemia**. *Int. J. Lab. Hematol.* (2019.0) **41** 95-101. DOI: 10.1111/ijlh.13014
15. Wilmotte R, Marechal J, Delaunay J. **Mutation at position -12 of intron 45 (c–>t) plays a prevalent role in the partial skipping of exon 46 from the transcript of allele alphaLELY in erythroid cells**. *Br. J. Haematol.* (1999.0) **104** 855-859. DOI: 10.1046/j.1365-2141.1999.01271.x
16. Higasa K. **Human genetic variation database, a reference database of genetic variations in the Japanese population**. *J. Hum. Genet.* (2016.0) **61** 547-553. DOI: 10.1038/jhg.2016.12
17. Andolfo I. **PIEZO1-R1864H rare variant accounts for a genetic phenotype-modifier role in dehydrated hereditary stomatocytosis**. *Haematologica* (2018.0) **103** e94-e97. DOI: 10.3324/haematol.2017.180687
18. Glogowska E. **Novel mechanisms of PIEZO1 dysfunction in hereditary xerocytosis**. *Blood* (2017.0) **130** 1845-1856. DOI: 10.1182/blood-2017-05-786004
19. Gnanasambandam R. **Increased red cell KCNN4 activity in sporadic hereditary xerocytosis associated with enhanced single channel pressure sensitivity of PIEZO1 mutant V598M**. *Hemasphere* (2018.0) **2** e55. DOI: 10.1097/HS9.0000000000000055
20. Picard V. **Clinical and biological features in PIEZO1-hereditary xerocytosis and Gardos channelopathy: a retrospective series of 126 patients**. *Haematologica* (2019.0) **104** 1554-1564. DOI: 10.3324/haematol.2018.205328
21. Imashuku S. **PIEZO1 gene mutation in a Japanese family with hereditary high phosphatidylcholine hemolytic anemia and hemochromatosis-induced diabetes mellitus**. *Int. J. Hematol.* (2016.0) **104** 125-129. DOI: 10.1007/s12185-016-1970-x
22. More TA, Dongerdiye R, Devendra R, Warang PP, Kedar PS. **Mechanosensitive Piezo1 ion channel protein (PIEZO1 gene): update and extended mutation analysis of hereditary xerocytosis in India**. *Ann. Hematol.* (2020.0) **99** 715-727. DOI: 10.1007/s00277-020-03955-1
23. de Meira Oliveira P. **Heterogeneous phenotype of Hereditary Xerocytosis in association with PIEZO1 variants**. *Blood Cells Mol. Dis.* (2020.0) **82** 102413. DOI: 10.1016/j.bcmd.2020.102413
24. Fermo E. **Hereditary xerocytosis due to mutations in PIEZO1 gene associated with heterozygous pyruvate kinase deficiency and beta-thalassemia trait in two unrelated families**. *Case Rep. Hematol.* (2017.0) **2017** 2769570. PMID: 28367341
25. Frederiksen H. **Dehydrated hereditary stomatocytosis: clinical perspectives**. *J. Blood Med.* (2019.0) **10** 183-191. DOI: 10.2147/JBM.S179764
26. Ma S. **A role of PIEZO1 in iron metabolism in mice and humans**. *Cell* (2021.0) **184** 969-982.e913. DOI: 10.1016/j.cell.2021.01.024
27. Süss C. **KCNN4 expression is elevated in inflammatory bowel disease: this might be a novel marker and therapeutic option targeting potassium channels**. *J. Gastrointestin. Liver Dis.* (2020.0) **29** 539-547. DOI: 10.15403/jgld-903
28. Simms LA. **KCNN4 gene variant is associated with ileal Crohn’s disease in the Australian and New Zealand population**. *Am. J. Gastroenterol.* (2010.0) **105** 2209-2217. DOI: 10.1038/ajg.2010.161
29. Kosoy R. **Genetics of the human microglia regulome refines Alzheimer’s disease risk loci**. *Nat. Genet.* (2022.0) **54** 1145-1154. DOI: 10.1038/s41588-022-01149-1
30. Philp AR. **Kcnn4 is a modifier gene of intestinal cystic fibrosis preventing lethality in the Cftr-F508del mouse**. *Sci. Rep.* (2018.0) **8** 9320. DOI: 10.1038/s41598-018-27465-3
31. Mansour-Hendili L. **Multiple thrombosis in a patient with Gardos channelopathy and a new KCNN4 mutation**. *Am. J. Hematol.* (2021.0) **96** E318-e321. DOI: 10.1002/ajh.26245
32. Allegrini B. **New KCNN4 variants associated with anemia: stomatocytosis without erythrocyte dehydration**. *Front. Physiol.* (2022.0) **13** 918620. DOI: 10.3389/fphys.2022.918620
33. Rivera A. **The erythroid K-Cl cotransport inhibitor [(dihydroindenyl)oxy]acetic acid blocks erythroid Ca(2+)-activated K(+) channel KCNN4**. *Am. J. Physiol. Cell Physiol.* (2022.0) **323** C694-c705. DOI: 10.1152/ajpcell.00240.2022
34. Beneteau C. **Recurrent mutation in the PIEZO1 gene in two families of hereditary xerocytosis with fetal hydrops**. *Clin. Genet.* (2014.0) **85** 293-295. DOI: 10.1111/cge.12147
35. Arora RD. **Flow cytometric osmotic fragility test and eosin-5’-maleimide dye-binding tests are better than conventional osmotic fragility tests for the diagnosis of hereditary spherocytosis**. *Int. J. Lab. Hematol.* (2018.0) **40** 335-342. DOI: 10.1111/ijlh.12794
36. Zama D. **A novel PIEZO1 mutation in a patient with dehydrated hereditary stomatocytosis: a case report and a brief review of literature**. *Ital. J. Pediatr.* (2020.0) **46** 102. DOI: 10.1186/s13052-020-00864-x
37. Andolfo I. **Missense mutations in the ABCB6 transporter cause dominant familial pseudohyperkalemia**. *Am. J. Hematol.* (2013.0) **88** 66-72. DOI: 10.1002/ajh.23357
|
---
title: ERdj5 protects goblet cells from endoplasmic reticulum stress-mediated apoptosis
under inflammatory conditions
authors:
- Hyunjin Jeong
- Eun-Hye Hong
- Jae-Hee Ahn
- Jaewon Cho
- Jae-Hyeon Jeong
- Chae-Won Kim
- Byung-Il Yoon
- Ja Hyun Koo
- Yun-Yong Park
- Yoon Mee Yang
- Takao Iwawaki
- Bruce A. Vallance
- Sun-Young Chang
- Hyun-Jeong Ko
journal: Experimental & Molecular Medicine
year: 2023
pmcid: PMC9981579
doi: 10.1038/s12276-023-00945-x
license: CC BY 4.0
---
# ERdj5 protects goblet cells from endoplasmic reticulum stress-mediated apoptosis under inflammatory conditions
## Abstract
Endoplasmic reticulum stress is closely associated with the onset and progression of inflammatory bowel disease. ERdj5 is an endoplasmic reticulum-resident protein disulfide reductase that mediates the cleavage and degradation of misfolded proteins. Although ERdj5 expression is significantly higher in the colonic tissues of patients with inflammatory bowel disease than in healthy controls, its role in inflammatory bowel disease has not yet been reported. In the current study, we used ERdj5-knockout mice to investigate the potential roles of ERdj5 in inflammatory bowel disease. ERdj5 deficiency causes severe inflammation in mouse colitis models and weakens gut barrier function by increasing NF-κB-mediated inflammation. ERdj5 may not be indispensable for goblet cell function under steady-state conditions, but its deficiency induces goblet cell apoptosis under inflammatory conditions. Treatment of ERdj5-knockout mice with the chemical chaperone ursodeoxycholic acid ameliorated severe colitis by reducing endoplasmic reticulum stress. These findings highlight the important role of ERdj5 in preserving goblet cell viability and function by resolving endoplasmic reticulum stress.
## Inflammatory bowel disease: A protective enzyme in intestinal goblet cells
Studies of inflammatory bowel disease (IBD) in mice reveal the role of an enzyme that assists the degradation of mis-folded proteins in the ‘goblet’ cells involved in producing the mucus barrier that lines and protects the interior of the gut. The most common forms of IBD are ulcerative colitis and Crohn’s disease. Researchers in South Korea led by Hyun-Jeong Ko at Kangwon National University, Chuncheon, and Sun-Young Chang at Ajou University, Suwon, compared goblet cell biology in normal mice with mice in which the gene encoding the protein-degrading enzyme ERdj5 had been disabled. This showed that ERdj5 protects goblet cells from a form of stress-mediated cell death that occurs during gut inflammation. The results suggest that drugs modifying the molecular mechanisms underlying ERdj5’s actions could open new avenues towards preventing and treating IBD.
## Introduction
The biological barrier of the gut consists of a monolayer of epithelial cells, immune cells, and the microbiota1. The outer surface of the epithelium, which is covered with a sticky layer of mucus, acts as a physical barrier, preventing the entry of noxious substances and enteric pathogens into the body2. In the colon, two layers of mucus constitute a thick inner layer that is firmly adherent to the epithelium and impermeable to commensal bacteria and a loose outer layer where the microbiota resides3. There are two forms of mucin in the gut: secreted and transmembrane mucins. Secreted mucins, such as mucin 2 (MUC2), form the mucus barrier, whereas transmembrane mucins, such as MUC1, MUC3, MUC4, MUC12, MUC13, and MUC17, are involved in signaling events and barrier defense4. Goblet cells produce highly glycosylated mucins to protect the vulnerable intestinal tract5. Additionally, colonic mucins provide a niche and energy source for the gut microbiome and act as a physical barrier against pathogens3,6. A defective mucus layer with reduced goblet cells is a common characteristic feature of ulcerative colitis (UC)7. However, the mechanism underlying goblet cell dysfunction or reduction in patients with UC has not been clearly defined.
The accumulation of misfolded or unfolded proteins within the endoplasmic reticulum (ER) induces the unfolded protein response (UPR) to relieve ER stress8. The UPR is a highly conserved mechanism in metazoans and consists of three ER-associated pathways that initiate adaptive transcriptional programs within the nucleus: PKR-like ER-resident kinase (PERK), activated transcription factor 6 (ATF6), and inositol-requiring enzyme 1 (IRE1)/X-box binding protein 1 (XBP-1)9,10. Several components of the UPR, including XBP-1, are essential for the biogenesis of the cellular secretory machinery of the exocrine glands11. Recently, it was reported that unabated ER stress resulting from the genetic deletion of XBP-1 or misfolded MUC2 within intestinal epithelial cells (IECs), leads to inflammation in the intestinal tract in mice, which is similar to IBD12,13. Moreover, it has been shown that polymorphisms in several genes, including XBP-1 and anterior gradient protein-2 (AGR2), increase the risk of both forms of IBD: Crohn’s disease (CD) and UC13,14. Elevated ER stress-induced localized intestinal inflammation has been suggested to be associated with the dysfunction of Paneth cells and goblet cells15.
Protein disulfide isomerases (PDIs), including endoplasmic reticulum-resident protein 57 (ERp57), endoplasmic reticulum DnaJ domain-containing protein 5 (ERdj5), and AGR2, are a family of ER foldases that catalyze disulfide reduction, oxidation, and isomerization16. These factors reduce ER stress by inhibiting the aggregation of unfolded/misfolded proteins in the ER, suggesting a potential role for PDIs as chaperones17. PDIs are important for the proper folding and quality control of secreted proteins18. MUC2 is a highly glycosylated glycoprotein that acts as a primary defender in IECs and is produced by goblet cells. MUC2-knockout (KO) mice develop spontaneous inflammation due to increased exposure to microbial products from the normal flora19,20. Considering that MUC2 protein synthesis is regulated by PDIs, defects in AGR2 or ERdj5 may result in misfolded mucin production. Thus, AGR2 deficiency results in the loss of intestinal gel-forming mucus and induces spontaneous inflammation of the intestines in humans and mice21.
ERdj5 has disulfide reductase activity that is required for disulfide reduction of proteins to be translocated into the cytoplasm for ER-associated degradation22. ERdj5 is also required for the proper folding of several proteins, including the low-density lipoprotein receptor, as well as for the degradation of misfolded proteins23. Functional defects in ERdj5 are associated with well-known ER stress-related diseases, such as Parkinson’s and Alzheimer’s disease24. Additionally, ERdj5 expression is increased in the salivary glands of patients with Sjögren’s syndrome25. Likewise, ERdj5-KO mice exhibit increased levels of ER stress in the salivary gland, exhibiting a Sjögren’s syndrome-like phenotype with spontaneous periductal inflammation in the salivary glands25. Thus, the ablation of ERdj5 in mice is likely to induce ER stress in highly secretory cells. However, little is known regarding the role of ERdj5 in mucin production and its involvement in IBD development.
Here, we used ERdj5-KO mice to investigate the role of ERdj5 in MUC2 production in colonic goblet cells. In the steady state, ERdj5-KO mice did not show abnormalities in MUC2 production. However, in the context of dextran sulfate sodium (DSS)-induced colitis, goblet cells in ERdj5-KO mice underwent apoptosis due to increased ER stress, resulting in reduced MUC2 production. Our findings revealed that ERdj5 deficiency further exacerbated inflammation in DSS-induced colitis by enhancing barrier breakdown and increasing NF-κB-mediated inflammation. These results suggest that ERdj5 is crucial for maintaining proper goblet cell function.
## Animals and treatments
Female C57BL/6 mice were purchased from Koatech (Pyeongtaek, South Korea). ERdj5-KO mice (RBRC05508) were purchased from RIKEN BioResource Center (Ibaraki, Japan)26. All experiments were approved by the Institutional Animal Care and Use Committee of Kangwon University (IACUC, admission number KW-190319-1). The mice were bred at the Animal Laboratory Center of Kangwon National University. For acute colitis induction, the mice were administered $2\%$ DSS (molecular weight 36–50 kDa; MP Biomedicals, Solon, OH, USA) in drinking water for 5 days, followed by three consecutive days of tap water. Changes in body weight were monitored daily to assess the degree of colitis. On Day 8, colon length was measured, and histological analysis was performed. The mice were administered 500 mg/kg ursodeoxycholic acid (UDCA; Sigma‒Aldrich, St. Louis, MO, USA) once per day via oral gavage beginning on the first day of DSS treatment for a total of 8 days. Additionally, 1 mg/kg SB225002 (Cayman, Ann Arbor, MI, USA), a CXCR2 antagonist, was intraperitoneally (i.p.) injected once daily into DSS-treated mice. Recombinant mouse IL-22 (50 µg/kg, BioLegend, San Diego, CA, USA) was injected i.p. into DSS-treated mice every other day.
## Quantitative real-time PCR and RNA sequencing
Total RNA was extracted from colon tissues using TRIzol reagent (Ambion, Austin, TX, USA) and a tissue-homogenizing matrix kit (Biostar Korea, Daejeon, Korea). The yield and purity of RNA were confirmed using a UV/Vis Nano spectrophotometer (MicroDigital, Seongnam, Korea), and the RNA was reverse-transcribed into cDNA according to the GoScript Reverse Transcription System Technical Manual (Promega, Madison, WI, USA). Quantitative real-time PCR (qRT‒PCR) was performed on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA) using a GoTaq qPCR system. All reactions were performed under the following conditions: 39 cycles of predenaturation at 95 °C for 3 min, denaturation at 95 °C for 10 s, and annealing at 60 °C for 30 s and 72 °C for 30 s. The primers used for qPCR were as follows: m-GAPDH forward (5′-TGT GTC CGT GGA TCT GA-3′), m-GAPDH reverse (5′-CCT GCT TCA CCA CCT TCT TGA T-3′), m-MUC2 forward (5′-ATG CCC ACC TCC TCA AAG AC-3′), m-MUC2 reverse (5′-GTA GTT TCC GTT GGA ACA GTG AA-3′), m-Zo-1 forward (5′-GAC CAA TAG CTG ATG TTG CCA GAG-3′), m-Zo-1 reverse (5′-TAT GAA GGC GAA TGA TGC CAG A-3′), m-Cldn1 (Claudin-1) forward (5′-CTG GAA GAT GAG GTG CAG AAG A-3′), m-Cldn1 reverse (5′-CCA CTA ATG TCG CCA GAC CTG AA-3′), m-GRP78 (BiP) forward (5′-TGT GGT ACC CAC CAA GAA GTC-3′), m-GRP78 (BiP) reverse (5′-TTC AGC TGT CAC TCG GAG AAT-3′), m-XBP-1s forward (5′-GAG TCC GCA GGT G-3′), m-XBP-1s reverse (5′-GTG TCA GAG TCC ATG GGA-3′), m-ATF4 forward (5′-CCT GAA CAG CGA AGT GTT GG-3′), and m-ATF4 reverse (5′-TGG AGA ACC CAT GAG GTT TCA A-3′). All primers were synthesized by Macrogen (Seoul, Korea). RNA quality checks, bioinformatics data analysis, and RNA sequencing were performed by Macrogen. The candidate genes were screened based on P values < 0.01 and fold-change > 3. *The* genes of interest were categorized into multiple biological pathways using KEGG27. GO clustering analysis was performed using DAVID 6.7 software28. The heatmap was visualized using the R package. Microarray data (GSE16879, GSE36807, and GSE47908) were downloaded from the NCBI gene expression omnibus (GEO). The RNA sequencing data are available to access GEO Series number GSE224087. To compare ERdj5 expression in tissue-derived cells, we reanalyzed a previously published dataset29 at the EMBL European Bioinformatics Institute (https://www.ebi.ac.uk) under accession number ENSG00000077232.
## ELISA
Interleukin (IL)-6, IL-1β, IL-10, IL-22, CXCL1 and IFN- γ were measured using mouse uncoated ELISA kits (Invitrogen, Vienna, Austria) and mouse IL-22 and CXCL1 Duosets (R&D systems, Minneapolis, MN, USA). Briefly, murine colon tissues were homogenized using a Minilys personal homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France). Tissue homogenates were centrifuged, and the supernatants were analyzed according to the manufacturer’s instructions. The absorbance was measured at 450 nm using a SpectraMax i3 (Molecular Devices, San Jose, CA, USA).
## Cell isolation and flow cytometry
To analyze immune cells in the lamina propria (LP) of the colon, the mouse colon was inverted in a polyethylene tube30. To remove mucus and the epithelium, the colon was treated with 1 mM 1,4-dithiothreitol (DTT; Biosesang, Seongnam, Korea) in phosphate-buffered saline (PBS; Corning Inc., Corning, NY, USA) for 10 min at 20 °C and then with 30 mM ethylene-diamine-tetraacetic acid (EDTA; Invitrogen) for 8 min at 20 °C. To isolate immune cells from the LP, colon tissues were incubated with 108 U/ml collagenase IV (Sigma‒Aldrich) at 37 °C for 90 min, and the cells were harvested by shaking for 12 min at 20 °C. After gradient centrifugation with 44–$66\%$ Percoll (GE Healthcare Life Sciences, Uppsala, Sweden), mononuclear immune cells were enriched for further analysis. The cells were stained with a combination of the following antibodies: FITC-conjugated anti-CD45 (30-F11; BioLegend), BV421-conjugated anti-CD11b (M$\frac{1}{70}$; BioLegend), APC-conjugated anti-Ly6C (HK1.4; BD Biosciences), PE-Cy7-conjugated anti-Ly6G (1A8; BD Biosciences), and 7-AAD (BioLegend). For flow cytometry, cells were acquired using a FACS Verse (BD Biosciences, Bergen County, NJ, USA), and the data were analyzed using FlowJo software V10 (BD Biosciences) and 7-AAD-negative and CD45-positive cells. Intestinal epithelial cells (IECs) and nonepithelial cells (non-IECs) were separately obtained on Day 2 following DSS treatment as described previously31 and were analyzed to assess ERdj5 expression levels.
## Histology
Colon tissues were rolled and fixed in $10\%$ neutral formalin, and 3 μm paraffin sections were stained with hematoxylin and eosin (H&E). The degree of inflammation was measured by a pathologist using i-Solution Lite (IMT i-Solution Inc., Daejeon, Korea). The pathological grade is expressed as a percentage32: 0, normal; g1, intestinal gland loss ≤ $25\%$ of the LP mucosae and slightly increased inflammatory cell infiltration of the LP; g2, intestinal gland loss ≥ $25\%$ or ≤ $50\%$ and significantly increased inflammatory cell infiltration; and g3, intestinal gland loss ≥ $50\%$ with severe erosion of inflammatory cells. To visualize mucin and goblet cells, colon sections were stained with a Periodic acid Schiff (PAS) stain kit (Abcam, Cambridge, UK) according to the manufacturer’s instructions. The number of goblet cells per crypt was determined using a Nikon Eclipse Ts2 (Nikon, Minato, Tokyo, Japan).
## Immunofluorescence staining
Colon tissue Section (5 µm thick) were deparaffinized, and antigens were retrieved using a microwave and 10 mM sodium citrate buffer (Donginbiotech Co., Seoul, Korea). After being blocked with $1\%$ bovine serum albumin (BSA; MP Biomedicals) in PBS, the tissues were stained with primary antibodies overnight at 4 °C. After being washed with PBS, the tissues were incubated with secondary antibodies for 2 h at 4 °C. The slides were mounted with a mounting solution containing 4′,6-diamidino-2-phenylindole (DAPI; Vector, Burlingame, CA, USA) for nuclear staining.
For organoid staining, the colonoids were fixed with $4\%$ paraformaldehyde (Invitrogen) for 1 h at 20 °C and permeabilized with $2\%$ Triton X-100 (Sigma‒Aldrich) in PBS for 2 h at 37 °C. After being blocked with $3\%$ BSA in PBS containing $1\%$ Triton X-100 for 2 h at 20 °C, the colonoids were incubated with primary antibodies overnight at 4 °C. The colonoids were incubated with secondary antibodies for 2 h at 37 °C and mounted with antifade mounting medium containing DAPI. To evaluate cell apoptosis, a TUNEL assay was performed using an ApopTag Fluorescein In Situ Apoptosis Detection Kit (Merck, Kenilworth, NJ, USA) according to the manufacturer’s instructions. Images were obtained using a confocal laser scanning microscope (LSM880; Carl Zeiss, Göttingen, Germany) at the Central Laboratory of Kangwon National University and analyzed with Zen software (Carl Zeiss).
The samples were stained with primary antibodies, including rabbit anti-mouse MUC2 (PA5-79702, 1:200; Invitrogen), rabbit anti-mouse CLCA1 (clone EPR12254-88, 1:200; Abcam), Alexa Fluor 647-conjugated rat anti-mouse E-cadherin (clone DECMA-1, 1:200; BioLegend), rabbit anti-mouse ZO-1 (61-7300, 1:250; Invitrogen), mouse anti-mouse Claudin-1 (clone 2H10D10, 1:250; Invitrogen), rabbit anti-mouse NF-κB (clone D14E12, 1:500; Cell Signaling Technology), rabbit anti-mouse myeloperoxidase (MPO; PA5-16672, 1:200; Invitrogen), APC-conjugated Ly6G (clone 1A8, 1:200; BD Biosciences), and goat anti-mouse lysozyme C (sc-27958, 1:200; Santa Cruz Biotechnology, Dallas, TX, USA). The secondary antibodies used were anti-mouse IgG H&L Alexa Fluor 488 (ab150105, 1:500; Abcam), anti-rabbit IgG (H + L) Alexa Fluor 594 (A-11012, 1:400; Invitrogen), anti-rabbit IgG H&L Alexa Fluor 647 (ab150079, 1:400; Abcam), and anti-goat IgG (H + L) Alexa Fluor 555 (A32816, 1:400; Invitrogen).
## Colon organoid culture
Colon organoids were cultured as previously described33. Briefly, the colon was opened longitudinally and treated with penicillin‒streptomycin (Gibco, Carlsbad, CA, USA) and gentamycin (Gibco) at 4 °C for 30 min to remove microbiota, and a cell recovery solution (Corning) was added and incubated at 4 °C for 30 min. Using curved forceps, villi were scratched and centrifuged at 800 rpm and 4 °C. After being washed with advanced DMEM/F12 (Gibco), the harvested crypts were counted. After being mixed 1:1 with Matrigel (Corning) and advanced DMEM/F12 containing penicillin‒streptomycin, Glutamax (Gibco), and HEPES (Gibco), the media containing the crypts was plated in 24-well plates and filled with $50\%$ L-WRN growth media containing DMEM/F12, $20\%$ fetal bovine serum (FBS, Gibco), penicillin‒streptomycin, B27 (Invitrogen), N2 (Invitrogen), 10 mM nicotinamide (Sigma), 1.25 mM N-acetylcysteine (Sigma), 16.7 pM mEGF (Invitrogen), 10 μM P38i (Sigma), 0.5 μM A83-01 (Tocris, Bristol, UK), and 10 μM Y-27632 (Abmole, Houston, TX, USA). In 24-well plates, the crypt seeding density was 100 crypts/well. Colonoids were passaged to a maximum of passage 3 and used for experiments when they stopped growing and the inner necrotic core started to darken34, which usually occurred on Day 5 after Passage 3. We validated the results using three replicate experiments and proceeded with 200 colonoids per experiment. To mimic inflammatory conditions, colonoids were treated with Pam3CSK4 (InvivoGen, San Diego, CA, USA), an agonist of Toll-like receptor (TLR)2/TLR1, for 48 h at 37 °C on Day 5.
## Citrobacter rodentium infection
C. rodentium (DBS100) was cultured overnight in a shaking incubator at 200 rpm and 37 °C in Luria-Bertani (LB) broth one day before infection. The mice were infected by oral gavage with 0.1 ml of LB broth containing 2.5 × 108 CFU of streptomycin-resistant C. rodentium35. The mice were monitored and weighed daily for symptoms of distress or illness. On Day 14 postinfection, CFUs were measured in the colon and cecum. Tissue homogenates were serially diluted and cultured overnight at 37 °C in LB agar containing streptomycin. On Day 10 postinfection, mouse colon tissue was collected for immunofluorescence staining of goblet cell apoptosis.
## Western blotting
Colon tissues were homogenized in PRO-PREP™ Protein Extraction Solution (Intron, Seongnam, Korea) and mixed with 5× sample buffer (Genscript, Piscataway, NJ, USA) containing a protease inhibitor (Gendepot, Katy, TX, USA) and phosphatase inhibitor (Gendepot). The protein concentration was quantified using the BCA protein assay (Abbkine, Wuhan, China). The antibodies used for western blotting were PERK (clone C33E10), p-PERK (clone 16F8), eIF-2α (9722 S), p-eIF-2α (9721 S), GRP78/BiP (clone C50B12), CHOP (clone L63F7), IKK (2682 S), p-IKK (clone 16A6), IκB (9242 S), NF-κB p65 (clone D14E12), p-NF-κB p65 (clone 93H1), and anti-rabbit IgG-HRP (7074P2), all of which were purchased from Cell Signaling Technology (Danvers, MA, USA); ATF4 (clone H-290), β-actin (clone C4), and ERdj5 (clone 66.7), which were purchased from Santa Cruz Biotechnology; and anti-mouse IgG-HRP (ADI-SAB-100-J), which was purchased from Enzo Life Sciences (NY, USA). Protein bands were visualized using a chemiluminescence detection system for horseradish peroxidase (GBioscience, St. Louis, MO, USA).
The lysates of MODE-K cells were used to determine the expression of ER stress marker proteins, including PERK, p-PERK, eIF-2α, and p-eIF-2α, using western blotting. MODE-K cells were activated with Pam3CSK4 for a specified time, and NF-κB pathway signaling was analyzed. MODE-K cells were treated with 5 µM GSK 2606414 (Tocris), a PERK inhibitor, for 2 hr before Pam3CSK4 treatment.
## Construction of ERdj5-deficient MODE-K cells
The mouse intestinal epithelial cell line MODE-K36 was cultured in RPMI-1640 (Corning) supplemented with $10\%$ FBS, Glutamax, sodium pyruvate, and MEM NEAA (all from Gibco). LentiCRISPRv2, which was a gift from Feng Zhang, was purchased from Addgene (Addgene plasmid #52961; http://n2t.net/addgene:52961; RRID: Addgene_52961)37. The vector was digested using the restriction enzyme BsmBI (Enzynomics, Daejeon, Korea). After the vector was cut, gel extraction (Real Biotech, Taipei, Taiwan) was performed. The cleaved vector and annealed oligo were ligated (Real Biotech, Beijing, China). sgRNA (genome target site: 5′-GTA CAG TGG GGG CTA AAT CA-3′) targeting the mouse ERdj5 gene was cloned downstream of the U6 promoter, which was cut by the BsmB1 enzyme to generate a lentivirus vector. After ligation, the cells were transformed with Stbl3. The next day, a single colony was chosen, placed in LB broth and grown overnight in a shaking incubator. Plasmid DNA was extracted from liquid bacterial cultures using a Miniprep kit (Qiagen, Hilden, Germany). HEK 293FT cells were transfected with Lipofectamine 2000 (Life Technologies, Carlsbad, CA, USA) using purified plasmid DNA and packaging vectors. psPAX2 (Addgene plasmid #12260; http://n2t.net/addgene:12260; RRID: Addgene_12260) and pMD2.G (Addgene plasmid #12259; http://n2t.net/addgene:12259; RRID: Addgene_12259) were gifts from Didier Trono. The medium was changed after 6 h, and the virus was harvested after 72 h. MODE-K cells were infected with virus in medium containing 6 mg/ml polybrene (Sigma). After 8 h, the medium was replaced with complete medium, and cells were cultured with 6 mg/ml puromycin (Gibco) to select virus-infected cells. The deletion of ERdj5 was confirmed by western blot analysis.
## Statistical analysis
Data analysis was performed using GraphPad Prism 9 (San Diego, California, USA) and FlowJo V10 software. Differences between the two groups were determined using Student’s t test. One-way analysis of variance (ANOVA) followed by post hoc tests (Tukey’s multiple comparisons test) and two-way ANOVA followed by post hoc tests (Bonferroni’s multiple comparisons test) were used to compare data from more than two groups. A P value < 0.05 was considered significant. The data are representative of three independent experiments.
## ERdj5 expression is upregulated in UC colons
Although ERdj5 is highly expressed in the colon in humans and mice38, few reports have suggested a role for ERdj5 in gut physiology and pathogenic colitis. Gene expression profile analysis was performed using BRB-array tools with GEO data (GSE16879, GSE36807, and GSE47908). To examine the expression levels of the indicated genes, a comparative analysis was performed with R software. ERdj5 mRNA expression levels were significantly higher in the colon tissues of patients with UC than in those of normal controls (Fig. 1a). ERdj5 mRNA levels were positively correlated with MUC2 mRNA levels in patients with UC (Supplementary Fig. 1a). In addition, oral DSS treatment dramatically increased the protein expression of ERdj5 in murine colon tissue (Supplementary Fig. 1b). On Day 2 of DSS treatment, ERdj5 expression increased significantly in IECs but not in non-IECs in WT mice (Supplementary Fig. 1c-f). In addition, MUC2 expression was reduced in the colonic IECs of ERdj5-KO mice on Day 2 of DSS treatment, and a similar positive correlation between MUC2 and ERdj5 expression was observed in the WT CON/WT DSS groups as in the patient data (Supplementary Fig. 1c, g, h). We used single-cell transcriptome analysis to further confirm that ERdj5 was primarily expressed in the goblet cells of the human large intestine29.Fig. 1ERdj5 deficiency exacerbates DSS-induced colitis in a murine model.a ERdj5 (DNAJC10) mRNA expression levels in normal controls or patients with ulcerative colitis (UC) were analyzed in the public GEO database. b–e Mice were administered $2\%$ DSS in drinking water for 5 days and subsequently switched to normal water ($$n = 8$$–10 per group). b Body weight. c Colon length on Day 8. d Representative histopathological features of the colon in H&E-stained sections (×100, scale bar: 100 µm); m, mucosa; sm, submucosa; mm, muscular layer; pathological grade (g1, g2, g3). e Measurement of the length with different grading scores of inflammation in the colon. The data are representative of three independent experiments, and the values are expressed as the mean ± SEM; **$P \leq 0.01$ and ***$P \leq 0.001$; ns, not significant. Two-way ANOVA followed by Bonferroni’s test for (b) and one-way ANOVA followed by Tukey’s test for (c).
## ERdj5 deficiency induces more severe inflammation in DSS-induced colitis
To determine whether ERdj5 plays a protective or detrimental role in colonic inflammation, DSS-induced colitis was evaluated in wild-type (WT) and ERdj5-KO mice. Although no apparent colonic inflammation was observed in ERdj5-KO mice under steady-state conditions, the administration of $2\%$ DSS for 5 days significantly decreased body weight and colon length compared to those of WT mice (Fig. 1b, c). The level of tissue damage in the entire colon was assessed by H&E staining, and DSS-treated ERdj5-KO mice had more severe intestinal gland loss and inflammatory cell infiltration than WT mice (Fig. 1d). Histological analysis was performed, and the degree of inflammation is expressed as a percentage of the length combined with the degree of severity of the entire colon. A total of $18.7\%$ of WT and $43.5\%$ of KO mice had Grade 3 colonic inflammation, with significant loss of intestinal glands and severe deposition of inflammatory cells on the mucous membrane plate (Fig. 1e).
To further investigate the role of ERdj5 in colitis, colon samples from DSS-treated mice were obtained, and RNA sequencing was performed. Numerous changes were observed in the differentially expressed genes (DEGs) in DSS-treated ERdj5-KO mice compared to DSS-treated WT mice (Fig. 2a, b). An extensive analysis of GO terms using the DAVID bioinformatics database showed that many genes with particular ontologies, including the response to cytokines, surface receptor signaling, and the inflammatory response, were significantly altered in the colons of DSS-treated ERdj5-KO mice compared to WT mice (Fig. 2b, c). A heatmap of the DEGs associated with inflammation indicated that several cytokines and chemokines associated with severe inflammation, including Cxcl1, Cxcl2, Il1b, and Il6, were more highly expressed in the colon samples of DSS-treated ERdj5-KO mice than in those of WT mice (Fig. 2d, Supplementary Fig. 2a). We also confirmed elevated protein expression levels of IL-1β, IL-6, and CXCL1 in the distal colon tissues of DSS-treated ERdj5-KO mice (Fig. 2e, f). Moreover, neutrophil-related genes, including S100a8, S100a9, and Saa339,40, were increased in the inflamed colon tissues of DSS-treated ERdj5-KO mice compared to those of WT mice (Fig. 2d). Furthermore, a significant correlation between ERdj5 and S100A8/A9 was found in UC patients following mRNA analysis of data from a public database (Supplementary Fig. 1a). We found that Ly6G+ neutrophil infiltration was elevated in the colons of DSS-treated ERdj5-KO mice (Fig. 2g, Supplementary Fig. 3a, b). The transcription levels of Arg1 and Nos2, which are increased in neutrophils41,42, were also increased in DSS-treated ERdj5-KO mice (Supplementary Fig. 2b). However, the inhibition of neutrophil migration by a CXCR2 antagonist did not alleviate DSS-induced colitis in ERdj5-KO mice (Supplementary Fig. 3c). Collectively, these results demonstrate that ERdj5-deficient mice exhibited more severe inflammation with profound neutrophil infiltration than WT mice under DSS-induced colitis conditions; however, the inhibition of neutrophil infiltration alone did not attenuate severe colitis in these mice. Fig. 2RNA-seq reveals an altered gene expression profile for inflammation in the absence of ERdj5.a Venn diagram. The number of differentially expressed genes (DEGs) from colon samples extracted from 8-week-old female WT or ERdj5-KO mice ($$n = 3$$ per group) treated with or without $2\%$ DSS. b KEGG pathway analysis of DEGs in colon samples extracted from 8-week-old female WT or ERdj5-KO mice ($$n = 3$$ per group) treated with or without $2\%$ DSS. Pathways with statistical significance ($P \leq 0.01$) are shown. c GO analysis of major signaling pathways identified by RNA sequencing using the DAVID bioinformatics database. The top 10 pathways based on P values are shown, along with the number of genes with the respective ontology. d Heatmap of the DEGs associated with inflammation. The log2 ratios of ERdj5 KO DSS/WT DSS are presented (blue, underexpression; red, overexpression). Values over 5 or under −5 were rounded. e Levels of IL-1β and IL-6 in colon homogenates ($$n = 6$$–8 per group). f Level of CXCL1 in colon homogenates ($$n = 6$$–8 per group). g Representative flow cytometric analysis of CD11b+Ly6G+ neutrophils and CD11b+Ly6C+ monocytes among pregated live CD45+IA-IE-CD11c- cells in the LP of the colon. The percentage of CD11b+Ly6G+ neutrophils is expressed as the mean ± SEM ($$n = 3$$–4 per group). * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; one-way ANOVA followed by Tukey’s test.
## ERdj5 deficiency enhances goblet cell death in response to stress
In association with severe inflammation, intestinal goblet cell density was markedly reduced in the colons of DSS-treated ERdj5-KO mice, as shown by PAS staining of heavily glycosylated proteins (Fig. 3a, b). We further examined MUC2 protein expression in the colon by immunofluorescence staining using MUC2-specific antibodies (Fig. 3c). PAS-positive cells and MUC2 expression were comparable in the colons of ERdj5 KO and WT mice without DSS administration; however, these levels were significantly decreased in the colons of DSS-treated ERdj5-KO mice compared to DSS-treated WT mice (Fig. 3b–d). These results suggest that severe inflammation caused by ERdj5 deficiency leads to the loss of goblet cells. Fig. 3ERdj5 deficiency decreases mucin secretion and goblet cells in response to inflammatory signals.a Periodic acid Schiff (PAS) staining of the colons of WT and ERdj5-KO mice (×100, scale bar 500 µm; $$n = 6$$ per group). b Goblet cell count per crypt in PAS images ($$n = 15$$–25 per group). c Representative immunofluorescence images of MUC2 (green) in colon tissues, with DAPI (blue) for nuclear staining (×200, scale bar 200 µm). d Level of MUC2 mRNA expression ($$n = 6$$–8 per group). e Representative immunofluorescence images of colonoids originating from WT or ERdj5-KO mice treated with vehicle or Pam3CSK4. MUC2 (green), TUNEL (red), E-cadherin (magenta), and DAPI (blue) (left panel, ×200 and right panel, ×800). f MUC2+ cell counts per high-power field (HPF) and the percentage of TUNEL-positive cells among MUC2+ and MUC2- cells determined in colonoid images ($$n = 5$$ per group). g Representative immunofluorescence images of colon tissues from WT or ERdj5-KO mice on Day 2 following DSS treatment. CLCA1 (red), TUNEL (green), and DAPI (blue) (left panel, ×200 and right panel, ×630). h CLCA1+ cell count per HPF and TUNEL-positive percentages among CLCA1+ cells from colon tissue images ($$n = 4$$ per group). e, g Scale bars correspond to 100 μm and 20 μm, respectively. The data are representative of three independent experiments, and the values are expressed as the mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; one-way ANOVA followed by Tukey’s test.
To assess whether the increased inflammation in the colons of ERdj5-KO mice was directly associated with the increased apoptosis of goblet cells, we evaluated apoptosis in MUC2-producing cells using colon organoids derived from WT and ERdj5-KO mice. There was no remarkable structural differences in colonoids obtained from the WT group and ERdj5 KO groups without stimulation (Supplementary Fig. 4a). However, treating mature colonoids with Pam3CSK4, a TLR2 ligand, partially ruptured the structure of those in the ERdj5 KO group, while colonoids in the WT group did not rupture (Supplementary Fig. 4b). In addition, an increase in CXCL1 production was observed in the culture supernatant of ERdj5 KO organoids compared to WT organoids following Pam3CSK4 treatment (Supplementary Fig. 4c). Immunofluorescence staining of colonoids confirmed that the number of MUC2-producing goblet cells was lower in ERdj5-KO mice than in WT mice following stimulation with Pam3CSK4 (Fig. 3e, f). Intriguingly, MUC2+TUNEL+ cells were significantly increased in Pam3CSK4-treated ERdj5 KO colonoids compared with Pam3CSK4-treated WT colonoids, suggesting that TLR2 stimulation induced apoptosis in ERdj5-deficient goblet cells (Fig. 3f). Most distal colon cells obtained from ERdj5-KO mice on Day 8 after DSS treatment were severely damaged (Fig. 3a–d); therefore, it was difficult to evaluate sections with similar levels of inflammation. Thus, we compared colon tissues obtained on Day 2 after DSS treatment. An increased proportion of goblet cells expressing CLCA1 and MUC2 underwent apoptosis in the distal colons of DSS-treated ERdj5-KO mice, even though there was no evident inflammation (Fig. 3g, h, and Supplementary Fig. 5a, b). Next, we analyzed UPR gene expression in isolated mouse IECs on Day 2 of DSS treatment and observed a significant increase in GRP78 and CHOP expression in ERdj5-KO IECs but not in WT IECs (Supplementary Fig. 5c). Thus, we hypothesize that early upregulation of UPR gene expression on Day 2 of DSS treatment drives goblet cell apoptosis. However, lysozyme+ Paneth cells, which are highly secretory cells in the the small intestine, were not affected in ERdj5-KO mice, even under DSS insult (Supplementary Fig. 5d, e). Collectively, these results suggest that ERdj5 deficiency may trigger apoptosis in secretory epithelial cells, especially goblet cells, and further exacerbate intestinal inflammation.
## ERdj5 helps maintain gut barrier function by enhancing tight junctions
Tight junctions act as barriers in the intestinal epithelium and prevent microbial intrusion43. In patients with IBD, intestinal barrier destruction is common, which worsens inflammatory pathology44. Thus, we postulated that ERdj5 plays a role in the maintenance of gut barrier integrity. The levels of typical tight junction proteins, such as zonula occludens-1 (ZO-1) and Claudin-1, were decreased in the colons of ERdj5-KO mice compared to WT mice after DSS administration. However, no changes were detected in Zo-1 or Claudin-1 expression in ERdj5-KO colons compared to WT colons under normal conditions (Fig. 4a). The immunofluorescence results for ZO-1 and Claudin-1 were similar to the mRNA results (Fig. 4b). In addition, we found that the expression of IL-10 and IL-22, which are critical for the maintenance of intestinal homeostasis45, was significantly lower in ERdj5-KO mice than in WT mice after DSS administration (Fig. 4c). Treatment with recombinant IL-22 ameliorated DSS-induced colitis in WT mice46 but not in ERdj5-KO mice (Supplementary Fig. 3d). Thus, we infer that IL-22 treatment does not rescue goblet cell apoptosis in ERdj5-KO mice. Fig. 4ERdj5 contributes to the maintenance of intestinal barrier integrity.a Zo-1 and Cldn1 mRNA expression in the colon tissues of WT or ERdj5-KO mice treated with DSS ($$n = 6$$ per group). b Immunofluorescence images of colon tissue stained for the tight junction proteins ZO-1 (red) and Claudin-1 (green) (×200, scale bar 100 µm). c IL-10 and IL-22 expression in colon tissue homogenates ($$n = 5$$–8 per group). d Mice were infected with 2.5 × 108 CFU of C. rodentium. On Day 14 postinfection, CFUs in the colon and cecum were measured ($$n = 9$$–12 per group). e Representative immunofluorescence images of colon tissues from WT or ERdj5-KO mice 10 days after C. rodentium infection. CLCA1 (red), TUNEL (red), and DAPI (blue) (×200, scale bar 100 µm). f CLCA1+ cell counts per crypt of WT and ERdj5-KO mice in e ($$n = 10$$–12 per group). The data are representative of three independent experiments, and the values are expressed as the mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; one-way ANOVA followed by Tukey’s test for a, c, f and Student’s t test for d.
We next assessed whether the reduction in gut barrier function and mucin production due to ERdj5 deficiency impacted protection against enteric bacteria. ERdj5-KO mice were orally infected with 2.5 × 108 CFU of C. rodentium and monitored for 2 weeks to assess morbidity and mortality. No infection-related deaths were recorded in either group; however, ERdj5-KO mice exhibited delayed clearance of C. rodentium in the cecum and colon 14 days postinfection (Fig. 4d). The levels of inflammatory cytokines, including IFN-γ, IL-1β, and IL-6, were increased in ERdj5-KO mice following C. rodentium infection (Supplementary Fig. 6). FITC-dextran levels were significantly increased in DSS-treated ERdj5-KO mouse serum compared to DSS-treated WT mice, as assessed by a FITC-dextran permeability assay47, suggesting that ERdj5-KO mice had impaired gut barrier function (Supplementary Fig. 7). We also analyzed the number of goblet cells, since C. rodentium infection reportedly decreases the goblet cell count48. Although there was no difference in CLCA1+ cells in the colons of uninfected mice in the two groups, the number of CLCA1+ goblet cells was significantly reduced in the colons of ERdj5-KO mice on Day 10 postinfection (Fig. 4e, f). Taken together, these results suggest that ERdj5 is required for proper mucin secretion from goblet cells and the maintenance of gut barrier integrity under stressful conditions.
## ERdj5 deficiency activates NF-κB-dependent inflammatory signals by enhancing the UPR, which leads to epithelial cell apoptosis
DSS administration increased the transcription of several genes associated with protein processing in the ER, including Sec31b, Prkn, Cryab, and Tram1l1, in the colon tissues of ERdj5-KO mice (Supplementary Fig. 2b), suggesting that the ablation of ERdj5 altered abnormal protein processing. Considering that it was previously reported that ERdj5 regulates the UPR by inhibiting the phosphorylation of eukaryotic initiation factor 2α (eIF2α)49, we next assessed the levels of phosphorylated eIF2α and PERK and found that they were higher in the colon tissues of ERdj5-KO mice than in those of WT mice (Fig. 5a, b). In addition, a significant increase was observed in the levels of GRP78 and other PERK pathway-related proteins, including ATF4 and CHOP, in ERdj5-KO mice. The mRNA expression levels of ATF4, XBP1s, and GRP78 were also higher in DSS-treated ERdj5-KO mice than in DSS-treated WT mice (Supplementary Fig. 8a–c). XBP1 mRNA levels were positively correlated with ERdj5 mRNA levels in UC patients (Supplementary Fig. 1a). To further assess changes in the molecular expression of UPR factors induced by ERdj5 ablation, we generated ERdj5-KO MODE-K cells using CRISPR/Cas9 (Supplementary Fig. 9a). The loss of ERdj5 led to a marked increase in the levels of phosphorylated eIF2α, indicating an increase in the PERK/eIF2α branch of the UPR (Fig. 5c).Fig. 5ERdj5 deficiency causes ER stress and inflammation mediated by NF-κB activation.a Expression profile of ER stress-related proteins in the colons of WT or ERdj5-KO mice treated with DSS ($$n = 6$$ per group). b Relative band intensities are expressed as the mean ± SEM. c ER stress proteins in ERdj5-KO MODE-K cells altered by the CRISPR/Cas9 system. d The level of CXCL1 in ERdj5-KO MODE-K cells after Pam3CSK4 treatment (100, 200, 500, and 1000 ng/ml) for 6 h ($$n = 3$$ per group). e The protein expression of NF-κB pathway factors in MODE-K cells treated with Pam3CSK4 (1 µg/ml) for 0, 5, 15, 30, 60, 180, and 360 min. f Nuclear translocation of NF-κB/p65 in ERdj5-KO MODE-K cells at the indicated time points after Pam3CSK4 treatment (1 µg/ml). f Immunofluorescence image (×200, scale bar 10 µm). g The mean fluorescence intensity of NF-κB/p65 (green) is shown ($$n = 9$$~27 per group). The data are representative of three independent experiments, and the values are expressed as the mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; Student’s t test.
Previous studies have suggested that an elevated UPR can increase NF-κB-mediated signaling via stimulation with TLR ligands50. Increased intestinal ER stress may increase inflammation by elevating NF-κB-dependent chemokine expression. To test our hypothesis, we assessed the expression level of CXCL1 in ERdj5-KO MODE-K cells following stimulation with Pam3CSK4. ERdj5-KO MODE-K cells secreted higher levels of CXCL1 than WT MODE-K cells in a dose-dependent manner (Fig. 5d). These results suggest that ERdj5 deletion in IECs increases CXCL1 secretion in response to TLR2 stimulation.
To clarify the mechanism of TLR2-induced CXCL1 production in ERdj5-deficient cells, we examined the nuclear translocation of NF-κB. Enhanced phosphorylation of IκB kinase (IKK) and early degradation of inhibitor of NF-κB alpha (IκBα) in ERdj5-KO MODE-K cells led to an increase in the level of phosphorylated NF-κB p65 (Fig. 5e). Accordingly, nuclear localization of NF-κB p65 in ERdj5-KO MODE-K cells was significantly increased after Pam3CSK4 treatment at early time points (Fig. 5f, g), suggesting that ERdj5 deficiency enhanced NF-κB activation. In addition, treatment of ERdj5-KO MODE-K cells with a PERK inhibitor inhibited NF-κB signaling following TLR2 stimulation (Supplementary Fig. 9b). Collectively, these results suggest that enhanced ER stress in the absence of ERdj5 sensitizes CXCL1 production via TLR2 stimulation via early and enhanced NF-κB activation.
## The chemical chaperone UDCA ameliorates DSS-induced colitis in the absence of ERdj5
It has been previously reported that oral administration of UDCA protects against acute DSS-induced colitis51. Since ERdj5 deficiency increased ER stress and epithelial cell death, we postulated that treatment with chemical chaperones, including UDCA, could alleviate severe DSS-induced colitis in ERdj5-KO mice. Daily oral administration of UDCA to ERdj5-KO mice restored the lost body weight and shortened colon length induced by DSS administration (Fig. 6a, b). UDCA attenuated the severe infiltration of immune cells into the intestinal glands and tissue damage in DSS-treated ERdj5-KO mice (Fig. 6c, d). In addition, the transcription levels of GRP78, XBP1s, and ATF4, which were highly increased after DSS administration in ERdj5-KO mice, were reduced by UDCA treatment, relieving excessive ER stress (Fig. 6e, Supplementary Fig. 8d, e). Although DSS-administered ERdj5-KO mice treated with UDCA had restored IL-22 levels (Supplementary Fig. 8f), we believe that this UDCA-driven amelioration of DSS-induced colitis in ERdj5-KO mice was independent of the restoration of IL-22 levels (Supplementary Fig. 3d). In addition, the levels of CXCL1, IL-1β, and IL-6 were markedly reduced in DSS-administered ERdj5-KO mice following UDCA treatment (Fig. 6f, g). UDCA treatment also decreased neutrophil infiltration and restored tight junction proteins (Fig. 6h, Supplementary Fig. 3a). Finally, we observed that the reduced levels of MUC2 in DSS-administered ERdj5-KO mice were restored by UDCA treatment (Fig. 6i, j). Collectively, these results show that goblet cell dysfunction and subsequent severe inflammation in DSS-administered ERdj5-KO mice could be rescued by chemical chaperone-mediated ER stress reduction. Fig. 6UDCA treatment ameliorates DSS-induced colitis in ERdj5-KO mice. WT and ERdj5-KO mice were administered $2\%$ DSS in drinking water for 5 days and subsequently provided normal water. The mice were orally administered 500 mg/kg UDCA daily ($$n = 3$$–5 per group). a Bodyweight. b Colon length. c Representative H&E-stained colon tissue images (×100, scale bar 100 µm). Thin arrows indicate hyperplasia in affected areas. Thick arrows indicate multifocal inflammatory cell infiltration; m, mucosa; sm, submucosa; mm, muscular layer. d Histological scoring. e GRP78 mRNA expression in colon tissues. f CXCL1, g IL-1 and IL-6 levels in colon homogenates. h Zo-1 and Cldn1 mRNA levels in colon tissues. i Representative PAS staining (×100) and MUC2 immunofluorescence (×200) images of the colon. Scale bar corresponds to 200 µm. j MUC2 mRNA expression. The data are representative of three independent experiments, and the values are expressed as the mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$; ns, not significant; one-way ANOVA was followed by Tukey’s test.
## Discussion
ERdj5 is highly expressed in the colon of patients with UC and in murine colitis. To investigate the role of ERdj5 in gut physiology and pathogenic colitis, we used ERdj5-KO mice. ERdj5-KO mice have a relatively normal gut physiology in the steady state. However, ERdj5-KO mice were more susceptible to DSS-induced colitis and enteric pathogens, such as C. rodentium, than WT mice. Although ablation of ERdj5 itself did not affect mucin production or goblet cell viability under homeotic conditions, there was a defect in mucin production under inflammatory conditions, presumably due to goblet cell apoptosis. In addition, ERdj5 deficiency weakened gut barrier integrity by reducing tight junction proteins and IL-22 production. Thus, reduced mucin production, increased goblet cell apoptosis, and damaged barrier integrity were observed in the colons of ERdj5-deficient mice following DSS administration or TLR stimulation. These findings suggest that ERdj5 is important for maintaining intestinal homeostasis by mucosal barriers under colitis conditions (Supplementary Fig. 10).
MUC2 has many cysteine residues linked by intra- and intermolecular disulfides for assembly in the ER and Golgi apparatus52. Disulfide bonds between the cysteine residues of mucin are important for proper folding and structural integrity53,54. MUC2 can serve as a substrate for ERdj5, which mediates the efficient cytosolic degradation of misfolded proteins via the ER-associated degradation pathway55. We determined that ERdj5 was positively correlated with MUC2, MUC1, MUC5AC, and MUC5B (Supplementary Fig. 1a). In the colon of DSS-treated ERdj5-KO mice, the expression of Galnt5 and St3gal2, which are involved in mucin-type O-glycan biosynthesis56, was downregulated, followed by reduced synthesis of mature mucin. Furthermore, the expression of Hsph1, Hyou1, and Ero1l, which are associated with protein folding in the ER57, was increased in the colons of ERdj5-KO mice, suggesting a critical need for chaperone activity during protein folding. In response to demand for MUC2 biosynthesis, ERdj5 deficiency disrupts the misfolded protein disposal system, leads to the accumulation of misfolded MUC2 and enhancing ER stress and the UPR. Based on the results from DSS-treated ERdj5-KO mice and ERdj5-KO MODE-K cells, ERdj5 deficiency increased ER stress, as evidenced by increased phosphorylation levels of PERK and eIF-2α, as well as GRP78, ATF4, and CHOP. CHOP is a transcription factor induced by ATF4 and cleaved ATF6 that induces several proapoptotic factors, including death receptor 5 (DR5) and growth arrest and DNA damage 34 (GADD34)58. Thus, the increased ER stress in ERdj5-KO mice with elevated CHOP expression might be responsible for the increased apoptosis of goblet cells induced by DSS administration and TLR stimulation. In addition, this finding suggests that ERdj5 regulates ER stress responses by correcting the misfolded MUC2 protein.
Goblet cells are easily exposed to ER stress under inflammatory and infectious conditions. Excessive ER stress can lead to the death of goblet cells, which is driven by ER stress-triggered reactive oxygen species (ROS)59,60. Inflammatory cytokines, including IL-1β, IL-6, and TNF-α, can stimulate mucin production61. In addition, some microbial products deliver innate signals that induce mucin production by IECs through NF-κB signaling62. DSS treatment also induces the hypersecretion of mucin59. When ERdj5-KO mice were exposed to environmental insults, such as DSS, TLR ligands, and infection, MUC2 production and goblet cells were markedly decreased in the colons of ERdj5-KO mice. Thus, ERdj5 is essential for maintaining goblet cell homeostasis under inflammatory conditions, although ERdj5 deficiency itself does not directly cause ER stress in vivo.
The importance of MUC2 production in gut barrier homeostasis has been revealed in several previous studies. Mice that were genetically deficient in MUC2 spontaneously developed severe intestinal inflammation, which frequently progressed to invasive adenocarcinoma and rectal tumors at a later age63. Similarly, Winnie and Eeyore mice, which contain missense mutations in MUC2, exhibit early dysbiosis, the loss of goblet cells, mucus barrier impairment, and spontaneous development of colitis64. The AGR2 and ER to nucleus signaling 2 (ERN2), which is also known as IRE1β14,65, are goblet cell-specific ER proteins involved in normal MUC2 biosynthesis. Ire1β-/- mice have decreased production of intestinal MUC2 and increased susceptibility to DSS-induced colitis66. AGR2 variants have been identified as risk factors for IBD, including CD and UC67. AGR2 mediates the formation of disulfide bonds and proper folding of MUC2 through its thioredoxin-like domain14. Mice that are deficient in AGR2 are highly susceptible to DSS-induced colitis because of decreased levels of intestinal mucin, as well as abnormal Paneth cells in association with increased levels of ER stress68. In addition to genetic defects in MUC2 production, common food additive polysaccharides, such as maltodextrin, can trigger ER stress in goblet cells, mediate mucus depletion in association with alterations in MUC2 folding, and exacerbate chemically induced intestinal inflammation69. In infectious colitis and IBD, where MUC2 production is significantly increased, misfolded MUC2 accumulates, leading to ER stress-mediated goblet cell death59. These reports highlight the importance of proper MUC2 production for the maintenance of goblet cell health and overall mucosal homeostasis. Interestingly, although MUC2 protein levels were significantly reduced in AGR2-/- mice and MUC2-/- mice under steady-state conditions, ERdj5-KO mice exhibited normal levels of MUC2 in the colon20,68. In addition, there were no signs of defects in mucus-filled goblet cells of the colonic epithelium and colonoids from ERdj5-KO mice under steady-state conditions. These results suggest that ERdj5 is dispensable for MUC2 production in healthy gut environments. Thus, in the absence of ERdj5, other ER-resident PDI molecules, including AGR2 and PDIA3, may play compensatory roles to ensure proper ER function and proteostasis under steady-state conditions. Intriguingly, ERdj5 plays a crucial role in normal MUC2 biosynthesis. However, further investigations are necessary to elucidate the role of ERdj5 under inflammatory conditions.
ER stress and the UPR are important risk factors for IBD onset. As a consequence of the specific deletion of XBP-1 in murine IECs, aberrant ER stress leads to gut-specific inflammation13,36. Consistent with this finding, multiple single-nucleotide polymorphisms (SNPs) in XBP-1 have been observed in patients with IBD13. As common genetic factors in human IBD patients, SNPs recognize the UPR pathways, the PERK-ATF4-CHOP pathway, and GRP78, which is an important regulator of UPR initiation13. Activation of the NF-κB signaling pathway is a hallmark of intestinal inflammation in patients with IBD70. Gut microbiome-dependent TLR activation or an exacerbated ER stress pathway in IECs activates NF-κB signaling36,71. The depletion of XBP-1 in IECs results in IRE1-dependent NF-κB activation36, which might be triggered by TLR-dependent signaling in response to microbial substances71. Therefore, we postulated that ER stress in IECs was closely associated with intestinal inflammation. Accordingly, several previous reports have suggested that ER stress can reinforce the NF-κB signaling pathway. Chemically induced ER stress with tunicamycin or thapsigargin induces NF-κB activation in an IRE1-dependent manner72, and TRAF2 plays a critical role in ER stress-induced IκBα degradation and NF-kB activation73. In addition, the PERK/eIF2α pathway, another UPR axis, is associated with NF-κB activation under ER stress conditions74. Phosphorylation of eIF2α activates the NF-κB pathway by suppressing the translation of IκBα without inducing IκB degradation75. Finally, the ATF6 branch of the UPR is associated with NF-κB activation in response to ER stress. Inhibition of ATF6 suppresses NF-κB activation by Shiga-toxigenic E. coli toxin76. In the present study, ERdj5 deficiency promoted ER stress and reinforced TLR2-induced NF-κB activation, followed by enhanced chemokine production. Collectively, these results suggest that the attenuation of NF-κB could be a successful therapeutic approach for ER stress-associated inflammatory diseases, including IBD.
Conversely, inflammation can also induce an ER stress response. ER stress is known to be associated with inflammatory conditions, such as atherosclerosis, cystic fibrosis, and IBD15,77,78. As an example of inflammation-induced ER stress, proinflammatory cytokines, including IL-6, IL-1β, and TNF-α, can trigger the UPR74. Furthermore, inflammation-induced ROS and nitric oxide (NO) production, as well as inhibition of sarco/endoplasmic reticulum Ca2+ ATPase (SERCA) and ER chaperones, are also associated with the UPR79. Thus, inflammation itself can induce the ER stress response, and in turn, ER stress can enhance the inflammatory response to TLR stimulation. To develop preventive and therapeutic drugs against IBD, several strategies that reduce ER stress have been effective in preclinical studies. Recently, chemical chaperones, including UDCA and 4-phenylbutyric acid, were shown to suppress gut inflammation in a mouse model80. We confirmed that UDCA treatment significantly improved DSS-induced colitis in ERdj5-KO mice, suggesting that UDCA is a potential therapeutic candidate for the treatment of ER stress-related colitis.
Defective degradation of misfolded MUC2 in the context of ERdj5 deficiency leads to defective mucus barrier integrity and inflammation, which increases ER stress. The increased ER stress caused by ERdj5 deficiency affects goblet cells more severely, leading to their selective depletion by apoptosis. Our findings highlight the crucial role of ERdj5 in preserving goblet cell survival and function, as well as the importance of the mucus barrier in maintaining gut homeostasis. UDCA treatment of DSS-administered ERdj5-KO mice ameliorated colitis by restoring mucin production and gut barrier molecules, suggesting the potential of UDCA in the treatment of patients with IBD. These results suggest that the amelioration of excessive ER stress could be a prospective strategy for the development of preventive and therapeutic agents for the treatment of patients with IBD.
## Supplementary information
Supplementary_Figure The online version contains supplementary material available at 10.1038/s12276-023-00945-x.
## References
1. McCracken VJ, Lorenz RG. **The gastrointestinal ecosystem: a precarious alliance among epithelium, immunity and microbiota**. *Cell. Microbiol.* (2001) **3** 1-11. DOI: 10.1046/j.1462-5822.2001.00090.x
2. Liévin-Le Moal V, Servin AL. **The front line of enteric host defense against unwelcome intrusion of harmful microorganisms: mucins, antimicrobial peptides, and microbiota**. *Clin. Microbiol. Rev.* (2006) **19** 315-337. DOI: 10.1128/CMR.19.2.315-337.2006
3. Johansson ME, Hansson GC. **Immunological aspects of intestinal mucus and mucins**. *Nat. Rev. Immunol.* (2016) **16** 639-649. DOI: 10.1038/nri.2016.88
4. van Putten JPM, Strijbis K. **Transmembrane Mucins: Signaling Receptors at the Intersection of Inflammation and Cancer**. *J. Innate Immun.* (2017) **9** 281-299. DOI: 10.1159/000453594
5. Pelaseyed T. **The mucus and mucins of the goblet cells and enterocytes provide the first defense line of the gastrointestinal tract and interact with the immune system**. *Immunol. Rev.* (2014) **260** 8-20. DOI: 10.1111/imr.12182
6. Fang J. **Slimy partners: the mucus barrier and gut microbiome in ulcerative colitis**. *Exp. Mol. Med.* (2021) **53** 772-787. DOI: 10.1038/s12276-021-00617-8
7. Singh V. **Chronic Inflammation in Ulcerative Colitis Causes Long-Term Changes in Goblet Cell Function**. *Cell. Mol. Gastroenterol. Hepatol.* (2022) **13** 219-232. DOI: 10.1016/j.jcmgh.2021.08.010
8. Depaepe T. **At the Crossroads of Survival and Death: The Reactive Oxygen Species-Ethylene-Sugar Triad and the Unfolded Protein Response**. *Trends Plant Sci.* (2021) **26** 338-351. DOI: 10.1016/j.tplants.2020.12.007
9. Kaser A, Blumberg RS. **Survive an innate immune response through XBP1**. *Cell Res.* (2010) **20** 506-507. DOI: 10.1038/cr.2010.61
10. Gardner BM, Pincus D, Gotthardt K, Gallagher CM, Walter P. **Endoplasmic reticulum stress sensing in the unfolded protein response**. *Cold Spring Harb. Perspect. Biol.* (2013) **5** a013169. DOI: 10.1101/cshperspect.a013169
11. Lee AH, Chu GC, Iwakoshi NN, Glimcher LH. **XBP-1 is required for biogenesis of cellular secretory machinery of exocrine glands**. *Embo J.* (2005) **24** 4368-4380. DOI: 10.1038/sj.emboj.7600903
12. Kaser A, Zeissig S, Blumberg RS. **Inflammatory bowel disease**. *Annu. Rev. Immunol.* (2010) **28** 573-621. DOI: 10.1146/annurev-immunol-030409-101225
13. Kaser A. **XBP1 links ER stress to intestinal inflammation and confers genetic risk for human inflammatory bowel disease**. *Cell* (2008) **134** 743-756. DOI: 10.1016/j.cell.2008.07.021
14. Park SW. **The protein disulfide isomerase AGR2 is essential for production of intestinal mucus**. *Proc. Natl Acad. Sci. U. S. A.* (2009) **106** 6950-6955. DOI: 10.1073/pnas.0808722106
15. Kaser A, Martínez-Naves E, Blumberg RS. **Endoplasmic reticulum stress: implications for inflammatory bowel disease pathogenesis**. *Curr. Opin. Gastroenterol.* (2010) **26** 318-326. DOI: 10.1097/MOG.0b013e32833a9ff1
16. Meng Y. **The protein disulfide isomerase 1 of Phytophthora parasitica (PpPDI1) is associated with the haustoria-like structures and contributes to plant infection**. *Front. Plant Sci.* (2015) **6** 632. DOI: 10.3389/fpls.2015.00632
17. Haeri M, Knox BE. **Endoplasmic Reticulum Stress and Unfolded Protein Response Pathways: Potential for Treating Age-related Retinal Degeneration**. *J. Ophthalmic Vis. Res.* (2012) **7** 45-59. PMID: 22737387
18. Sato Y. **Synergistic cooperation of PDI family members in peroxiredoxin 4-driven oxidative protein folding**. *Sci. Rep.* (2013) **3** 2456. DOI: 10.1038/srep02456
19. Linden SK, Sutton P, Karlsson NG, Korolik V, McGuckin MA. **Mucins in the mucosal barrier to infection**. *Mucosal Immunol.* (2008) **1** 183-197. DOI: 10.1038/mi.2008.5
20. Van der Sluis M. **Muc2-deficient mice spontaneously develop colitis, indicating that MUC2 is critical for colonic protection**. *Gastroenterology* (2006) **131** 117-129. DOI: 10.1053/j.gastro.2006.04.020
21. Al-Shaibi AA. **Human AGR2 Deficiency Causes Mucus Barrier Dysfunction and Infantile Inflammatory Bowel Disease**. *Cell. Mol. Gastroenterol. Hepatol.* (2021) **12** 1809-1830. DOI: 10.1016/j.jcmgh.2021.07.001
22. Ushioda R. **Redox-assisted regulation of Ca2+ homeostasis in the endoplasmic reticulum by disulfide reductase ERdj5**. *Proc. Natl Acad. Sci. U. S. A.* (2016) **113** E6055-e6063. DOI: 10.1073/pnas.1605818113
23. Oka OB, Pringle MA, Schopp IM, Braakman I, Bulleid NJ. **ERdj5 is the ER reductase that catalyzes the removal of non-native disulfides and correct folding of the LDL receptor**. *Mol. Cell* (2013) **50** 793-804. DOI: 10.1016/j.molcel.2013.05.014
24. Muñoz-Lobato F. **Protective role of DNJ-27/ERdj5 in Caenorhabditis elegans models of human neurodegenerative diseases**. *Antioxid. Redox Signal.* (2014) **20** 217-235. DOI: 10.1089/ars.2012.5051
25. Apostolou E, Moustardas P, Iwawaki T, Tzioufas AG, Spyrou G. **Ablation of the Chaperone Protein ERdj5 Results in a Sjögren’s Syndrome-Like Phenotype in Mice, Consistent With an Upregulated Unfolded Protein Response in Human Patients**. *Front. Immunol.* (2019) **10** 506. DOI: 10.3389/fimmu.2019.00506
26. Hosoda A, Tokuda M, Akai R, Kohno K, Iwawaki T. **Positive contribution of ERdj5/JPDI to endoplasmic reticulum protein quality control in the salivary gland**. *Biochem. J.* (2009) **425** 117-125. DOI: 10.1042/BJ20091269
27. Huang da W, Sherman BT, Lempicki RA. **Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists**. *Nucleic Acids Res.* (2009) **37** 1-13. DOI: 10.1093/nar/gkn923
28. Huang da W, Sherman BT, Lempicki RA. **Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources**. *Nat. Protoc.* (2009) **4** 44-57. DOI: 10.1038/nprot.2008.211
29. Wang Y. **Single-cell transcriptome analysis reveals differential nutrient absorption functions in human intestine**. *J. Exp. Med.* (2020) **217** e20191130. DOI: 10.1084/jem.20191130
30. Kim YI. **CX(3)CR1(+) Macrophages and CD8(+) T Cells Control Intestinal IgA Production**. *J. Immunol.* (2018) **201** 1287-1294. DOI: 10.4049/jimmunol.1701459
31. Graves CL. **A method for high purity intestinal epithelial cell culture from adult human and murine tissues for the investigation of innate immune function**. *J. Immunol. Methods* (2014) **414** 20-31. DOI: 10.1016/j.jim.2014.08.002
32. Erben U. **A guide to histomorphological evaluation of intestinal inflammation in mouse models**. *Int. J. Clin. Exp. Pathol.* (2014) **7** 4557-4576. PMID: 25197329
33. Allaire JM. **Interleukin-37 regulates innate immune signaling in human and mouse colonic organoids**. *Sci. Rep.* (2021) **11** 8206. DOI: 10.1038/s41598-021-87592-2
34. Grebenyuk S, Ranga A. **Engineering Organoid Vascularization**. *Front. Bioeng. Biotechnol.* (2019) **7** 39. DOI: 10.3389/fbioe.2019.00039
35. Taylor GA. **Irgm1-deficiency leads to myeloid dysfunction in colon lamina propria and susceptibility to the intestinal pathogen Citrobacter rodentium**. *PLoS Pathog.* (2020) **16** e1008553. DOI: 10.1371/journal.ppat.1008553
36. Adolph TE. **Paneth cells as a site of origin for intestinal inflammation**. *Nature* (2013) **503** 272-276. DOI: 10.1038/nature12599
37. Sanjana NE, Shalem O, Zhang F. **Improved vectors and genome-wide libraries for CRISPR screening**. *Nat. Methods* (2014) **11** 783-784. DOI: 10.1038/nmeth.3047
38. Cunnea PM. **ERdj5, an endoplasmic reticulum (ER)-resident protein containing DnaJ and thioredoxin domains, is expressed in secretory cells or following ER stress**. *J. Biol. Chem.* (2003) **278** 1059-1066. DOI: 10.1074/jbc.M206995200
39. Sroussi HY, Lu Y, Zhang QL, Villines D, Marucha PT. **S100A8 and S100A9 inhibit neutrophil oxidative metabolism in-vitro: involvement of adenosine metabolites**. *Free Radic. Res.* (2010) **44** 389-396. DOI: 10.3109/10715760903431434
40. Zhang G. **Elevated Expression of Serum Amyloid A 3 Protects Colon Epithelium Against Acute Injury Through TLR2-Dependent Induction of Neutrophil IL-22 Expression in a Mouse Model of Colitis**. *Front. Immunol.* (2018) **9** 1503. DOI: 10.3389/fimmu.2018.01503
41. Zhang X, Xu W. **Neutrophils diminish T-cell immunity to foster gastric cancer progression: the role of GM-CSF/PD-L1/PD-1 signalling pathway**. *Gut* (2017) **66** 1878-1880. DOI: 10.1136/gutjnl-2017-313923
42. Yadav S. **Nitric oxide synthase 2 enhances the survival of mice during Salmonella Typhimurium infection-induced sepsis by increasing reactive oxygen species, inflammatory cytokines and recruitment of neutrophils to the peritoneal cavity**. *Free Radic. Biol. Med.* (2018) **116** 73-87. DOI: 10.1016/j.freeradbiomed.2017.12.032
43. Lee B, Moon KM, Kim CY. **Tight Junction in the Intestinal Epithelium: Its Association with Diseases and Regulation by Phytochemicals**. *J. Immunol. Res.* (2018) **2018** 2645465. DOI: 10.1155/2018/2645465
44. Groschwitz KR, Hogan SP. **Intestinal barrier function: molecular regulation and disease pathogenesis**. *J. Allergy Clin. Immunol.* (2009) **124** 3-20. DOI: 10.1016/j.jaci.2009.05.038
45. Wei HX, Wang B, Li B. **IL-10 and IL-22 in Mucosal Immunity: Driving Protection and Pathology**. *Front. Immunol.* (2020) **11** 1315. DOI: 10.3389/fimmu.2020.01315
46. Kim S. **Amelioration of DSS-Induced Acute Colitis in Mice by Recombinant Monomeric Human Interleukin-22**. *Immune Netw.* (2022) **22** e26. DOI: 10.4110/in.2022.22.e26
47. Sham HP. **Immune Stimulation Using a Gut Microbe-Based Immunotherapy Reduces Disease Pathology and Improves Barrier Function in Ulcerative Colitis**. *Front. Immunol.* (2018) **9** 2211. DOI: 10.3389/fimmu.2018.02211
48. Bergstrom KS. **Modulation of intestinal goblet cell function during infection by an attaching and effacing bacterial pathogen**. *Infect. Immun.* (2008) **76** 796-811. DOI: 10.1128/IAI.00093-07
49. Thomas CG, Spyrou G. **ERdj5 sensitizes neuroblastoma cells to endoplasmic reticulum stress-induced apoptosis**. *J. Biol. Chem.* (2009) **284** 6282-6290. DOI: 10.1074/jbc.M806189200
50. Kim S, Joe Y, Surh YJ, Chung HT. **Differential Regulation of Toll-Like Receptor-Mediated Cytokine Production by Unfolded Protein Response**. *Oxid. Med. Cell. Longev.* (2018) **2018** 9827312. DOI: 10.1155/2018/9827312
51. Van den Bossche L. **Ursodeoxycholic Acid and Its Taurine- or Glycine-Conjugated Species Reduce Colitogenic Dysbiosis and Equally Suppress Experimental Colitis in Mice**. *Appl. Environ. Microbiol.* (2017) **83** e02766-16. PMID: 28115375
52. Javitt G. **Assembly Mechanism of Mucin and von Willebrand Factor Polymers**. *Cell* (2020) **183** 717-729.e716. DOI: 10.1016/j.cell.2020.09.021
53. Mueller P. **High level in vivo mucin-type glycosylation in Escherichia coli**. *Microb. Cell Fact.* (2018) **17** 168. DOI: 10.1186/s12934-018-1013-9
54. Cornick S, Tawiah A, Chadee K. **Roles and regulation of the mucus barrier in the gut**. *Tissue Barriers* (2015) **3** e982426. DOI: 10.4161/21688370.2014.982426
55. Ushioda R. **ERdj5 is required as a disulfide reductase for degradation of misfolded proteins in the ER**. *Science* (2008) **321** 569-572. DOI: 10.1126/science.1159293
56. Xi D. **The glycosyltransferase ST3GAL2 is regulated by miR-615-3p in the intestinal tract of Campylobacter jejuni infected mice**. *Gut Pathog.* (2021) **13** 42. DOI: 10.1186/s13099-021-00437-1
57. Piróg KA. **XBP1 signalling is essential for alleviating mutant protein aggregation in ER-stress related skeletal disease**. *PLoS Genet.* (2019) **15** e1008215. DOI: 10.1371/journal.pgen.1008215
58. Hu H, Tian M, Ding C, Yu S. **The C/EBP Homologous Protein (CHOP) Transcription Factor Functions in Endoplasmic Reticulum Stress-Induced Apoptosis and Microbial Infection**. *Front. Immunol.* (2018) **9** 3083. DOI: 10.3389/fimmu.2018.03083
59. Tawiah A. **High MUC2 Mucin Expression and Misfolding Induce Cellular Stress, Reactive Oxygen Production, and Apoptosis in Goblet Cells**. *Am. J. Pathol.* (2018) **188** 1354-1373. DOI: 10.1016/j.ajpath.2018.02.007
60. Tiwari S, Begum S, Moreau F, Gorman H, Chadee K. **Autophagy is required during high MUC2 mucin biosynthesis in colonic goblet cells to contend metabolic stress**. *Am. J. Physiol. Gastrointest. Liver Physiol.* (2021) **321** G489-g499. DOI: 10.1152/ajpgi.00221.2021
61. Enss ML. **Proinflammatory cytokines trigger MUC gene expression and mucin release in the intestinal cancer cell line LS180**. *Inflamm. Res.* (2000) **49** 162-169. DOI: 10.1007/s000110050576
62. Lemjabbar H, Basbaum C. **Platelet-activating factor receptor and ADAM10 mediate responses to Staphylococcus aureus in epithelial cells**. *Nat. Med.* (2002) **8** 41-46. DOI: 10.1038/nm0102-41
63. Velcich A. **Colorectal cancer in mice genetically deficient in the mucin Muc2**. *Science* (2002) **295** 1726-1729. DOI: 10.1126/science.1069094
64. Heazlewood CK. **Aberrant mucin assembly in mice causes endoplasmic reticulum stress and spontaneous inflammation resembling ulcerative colitis**. *PLoS Med.* (2008) **5** e54. DOI: 10.1371/journal.pmed.0050054
65. Tsuru A. **Negative feedback by IRE1β optimizes mucin production in goblet cells**. *Proc. Natl Acad. Sci. U. S. A.* (2013) **110** 2864-2869. DOI: 10.1073/pnas.1212484110
66. Bertolotti A. **Increased sensitivity to dextran sodium sulfate colitis in IRE1beta-deficient mice**. *J. Clin. Invest.* (2001) **107** 585-593. DOI: 10.1172/JCI11476
67. Zheng W. **Evaluation of AGR2 and AGR3 as candidate genes for inflammatory bowel disease**. *Genes Immun.* (2006) **7** 11-18. DOI: 10.1038/sj.gene.6364263
68. Zhao F. **Disruption of Paneth and goblet cell homeostasis and increased endoplasmic reticulum stress in Agr2-/- mice**. *Dev. Biol.* (2010) **338** 270-279. DOI: 10.1016/j.ydbio.2009.12.008
69. Laudisi F. **The Food Additive Maltodextrin Promotes Endoplasmic Reticulum Stress-Driven Mucus Depletion and Exacerbates Intestinal Inflammation**. *Cell. Mol. Gastroenterol. Hepatol.* (2019) **7** 457-473. DOI: 10.1016/j.jcmgh.2018.09.002
70. Liu T, Zhang L, Joo D, Sun SC. **NF-κB signaling in inflammation**. *Signal Transduct. Target. Ther.* (2017) **2** 17023. DOI: 10.1038/sigtrans.2017.23
71. Martinon F, Chen X, Lee AH, Glimcher LH. **TLR activation of the transcription factor XBP1 regulates innate immune responses in macrophages**. *Nat. Immunol.* (2010) **11** 411-418. DOI: 10.1038/ni.1857
72. Hu P, Han Z, Couvillon AD, Kaufman RJ, Exton JH. **Autocrine tumor necrosis factor alpha links endoplasmic reticulum stress to the membrane death receptor pathway through IRE1alpha-mediated NF-kappaB activation and down-regulation of TRAF2 expression**. *Mol. Cell. Biol.* (2006) **26** 3071-3084. DOI: 10.1128/MCB.26.8.3071-3084.2006
73. Kaneko M, Niinuma Y, Nomura Y. **Activation signal of nuclear factor-kappa B in response to endoplasmic reticulum stress is transduced via IRE1 and tumor necrosis factor receptor-associated factor 2**. *Biol. Pharm. Bull.* (2003) **26** 931-935. DOI: 10.1248/bpb.26.931
74. Chipurupalli S, Samavedam U, Robinson N. **Crosstalk Between ER Stress, Autophagy and Inflammation**. *Front. Med. (Lausanne)* (2021) **8** 758311. DOI: 10.3389/fmed.2021.758311
75. Deng J. **Translational repression mediates activation of nuclear factor kappa B by phosphorylated translation initiation factor 2**. *Mol. Cell. Biol.* (2004) **24** 10161-10168. DOI: 10.1128/MCB.24.23.10161-10168.2004
76. Yamazaki H. **Activation of the Akt-NF-kappaB pathway by subtilase cytotoxin through the ATF6 branch of the unfolded protein response**. *J. Immunol.* (2009) **183** 1480-1487. DOI: 10.4049/jimmunol.0900017
77. Hotamisligil GS. **Endoplasmic reticulum stress and atherosclerosis**. *Nat. Med.* (2010) **16** 396-399. DOI: 10.1038/nm0410-396
78. Ribeiro CM, Boucher RC. **Role of endoplasmic reticulum stress in cystic fibrosis-related airway inflammatory responses**. *Proc. Am. Thorac. Soc.* (2010) **7** 387-394. DOI: 10.1513/pats.201001-017AW
79. Ly LD. **Oxidative stress and calcium dysregulation by palmitate in type 2 diabetes**. *Exp. Mol. Med.* (2017) **49** e291. DOI: 10.1038/emm.2016.157
80. Cao SS. **The unfolded protein response and chemical chaperones reduce protein misfolding and colitis in mice**. *Gastroenterology* (2013) **144** 989-1000.e1006. DOI: 10.1053/j.gastro.2013.01.023
|
---
title: Activating transcription factor-2 supports the antioxidant capacity and ability
of human mesenchymal stem cells to prevent asthmatic airway inflammation
authors:
- Hyein Ju
- HongDuck Yun
- YongHwan Kim
- Yun Ji Nam
- Seungun Lee
- Jinwon Lee
- Seon Min Jeong
- Jinbeom Heo
- Hyungu Kwon
- You Sook Cho
- Gowun Jeong
- Chae-Min Ryu
- Dong-Myung Shin
journal: Experimental & Molecular Medicine
year: 2023
pmcid: PMC9981582
doi: 10.1038/s12276-023-00943-z
license: CC BY 4.0
---
# Activating transcription factor-2 supports the antioxidant capacity and ability of human mesenchymal stem cells to prevent asthmatic airway inflammation
## Abstract
Glutathione (GSH), an abundant nonprotein thiol antioxidant, participates in several biological processes and determines the functionality of stem cells. A detailed understanding of the molecular network mediating GSH dynamics is still lacking. Here, we show that activating transcription factor-2 (ATF2), a cAMP-response element binding protein (CREB), plays a crucial role in maintaining the level and activity of GSH in human mesenchymal stem cells (MSCs) by crosstalking with nuclear factor erythroid-2 like-2 (NRF2), a well-known master regulator of cellular redox homeostasis. Priming with ascorbic acid 2-glucoside (AA2G), a stable vitamin C derivative, increased the expression and activity of ATF2 in MSCs derived from human embryonic stem cells and umbilical cord. Subsequently, activated ATF2 crosstalked with the CREB1-NRF2 pathway to preserve the GSH dynamics of MSCs through the induction of genes involved in GSH synthesis (GCLC and GCLM) and redox cycling (GSR and PRDX1). Accordingly, shRNA-mediated silencing of ATF2 significantly impaired the self-renewal, migratory, proangiogenic, and anti-inflammatory capacities of MSCs, and these defects were rescued by supplementation of the cells with GSH. In addition, silencing ATF2 attenuated the ability of MSCs to alleviate airway inflammatory responses in an ovalbumin-induced mouse model of allergic asthma. Consistently, activation of ATF2 by overexpression or the AA2G-based priming procedure enhanced the core functions of MSCs, improving the in vivo therapeutic efficacy of MSCs for treating asthma. Collectively, our findings suggest that ATF2 is a novel modulator of GSH dynamics that determines the core functionality and therapeutic potency of MSCs used to treat allergic asthma.
## Asthma: Antioxidant-boosting protein improves stem cell treatment
A cellular protein that promotes a key antioxidant will be a crucial component in stem cell therapies for allergic asthma. Stem cells derived from umbilical cords have been proposed as treatments for incurable allergic asthma, due to their ability to combat inflammation and regenerate damaged cells. Now, Dong-Myung Shin at University of Ulsan College of Medicine in Seoul, South Korea, and co-workers have shown that the activating transcription factor 2 (ATF2) acts to maintain healthy levels of the antioxidant glutathione, which is essential for the effectiveness of stem cell therapy. Specifically, ATF2 interplays with a specific nuclear protein to activate genes involved in glutathione synthesis. The researchers showed that the ability of MSC treatments to reduce airway inflammation in asthmatic mouse models was greatly reduced by silencing ATF2, and enhanced by its over-expression.
## Introduction
Asthma is the most common chronic disease of the lungs in children and adults. The prevalence of asthma has doubled in the past decade, leading to a substantial global health and economic burden. Asthma is an allergic disease that is characterized by a combination of inflammation and structural remodeling in the airways1. The resulting airway obstruction causes breathing difficulties, wheezing, shortness of breath, and coughing. Immune responses mediated by innate lymphoid cells and T helper 2 (Th2) cells contribute to allergic airway inflammation and fibrosis, causing permanent deterioration in pulmonary function2–4. Asthma patients are grouped into one of four or five categories and are treated in a stepwise manner, depending on symptom severity or extent of disease. Inhaled corticosteroids, long-acting β2-adrenergic receptor agonists, long‑acting muscarinic antagonists, and leukotriene receptor antagonists are used as asthma-control drugs, and an IgE-specific monoclonal antibody is used to treat the most severe form of the disease5. Although this stepwise approach has improved the management of asthma and reduced dependency on inhaled short-acting bronchodilators for symptom relief, none of the currently available treatments can alter the progression of the disease; hence, there is an urgent need to develop novel therapies.
Preclinical and clinical studies have suggested beneficial effects of mesenchymal stem cells (MSCs) in treating incurable allergic asthma6–10. These progenitor cells are typically derived from adult tissues, such as the bone marrow, adipose tissue, and umbilical cord (UC) or UC blood, and can also be established by differentiation from pluripotent stem cells (PSCs), including embryonic stem cells (ESCs) and induced PSCs (iPSCs)11–14. The therapeutic effects of MSCs are thought to be attributable to their multipotency and ability to directly regenerate damaged cells in target tissues. In addition, MSCs can have indirect effects by providing growth factors, mediating cell‒cell interactions, and supplying matrix proteins to modulate the microenvironment of damaged target tissues and facilitate regeneration15,16. In particular, the anti-inflammatory and immunomodulatory functions of MSCs are achieved by inhibiting the activation, proliferation, and function of immune cells, including T cells, B cells, innate lymphoid cells, natural killer cells, and antigen-presenting cells17,18.
Despite the multifactorial benefits of MSCs, their clinical application has been hindered by limited therapeutic efficacy and a lack of knowledge of their precise mode of action. The high therapeutic potency of primitive MSCs is reportedly lost after large-scale ex vivo expansion, which is required to obtain a sufficient number for therapeutic purposes, due to an accumulation of epigenetic abnormalities and oxidative stress provoked by supraphysiological stimulations19. We have described several ex vivo expansion methods to preserve the primitiveness of MSCs, including i) enriching and preserving small-sized cells20, ii) enhancing the antioxidant capacity by real-time monitoring of glutathione (GSH) dynamics9,21, and iii) enhancing cell migration and engraftment activity by priming with small molecules22. In addition, we recently reported that supplementation with small compounds without genetic manipulation enables the enrichment and expansion of small primitive MSCs with a high antioxidant and engraftment capacity, which was termed the Primed/Fresh/OCT4 (PFO) enrichment procedure23. All of these procedures enhance the levels and dynamics of GSH, which is essential to maintain the stemness and therapeutic efficacy of human MSCs9,21,23.
Mechanistically, MSCs with high GSH dynamics display activation of the cyclic adenosine monophosphate (cAMP)-response element (CRE) binding protein-1 (CREB1) and nuclear factor erythroid-2 like-2 (NRF2) pathway, leading to the induction of genes involved in GSH synthesis (GCLC and GCLM) and redox cycling (GSR and PRDX1)10,21,23. The intracellular levels and dynamics of GSH in MSCs can be improved by pretreatment/priming with forskolin (FSK), a CREB1 activator, or by priming with ascorbic acid 2-glucoside (AA2G), a stable vitamin-C (VitC) derivative that activates CREB1 and in turn upregulates NRF2 target genes responsible for GSH synthesis and redox cycling10,21. The biological effects of these GSH-enhancing conditions stimulate the core functions of MSCs derived from various sources, including human ESCs and adult tissues such as the UC and bone marrow. Notably, in previous studies, the in vivo therapeutic effects of MSCs were enhanced by improving GSH dynamics in an experimental asthma animal model and a humanized graft-versus-host disease mouse model10,23.
In this study, we demonstrate that activating transcription factor-2 (ATF2), a member of the leucine zipper domain-containing CREB/ATF transcription factor family, plays a key role in modulating GSH dynamics and determining the core functionality and therapeutic potency of MSCs used to treat allergic asthma.
## Study approval
Human UC samples were obtained from healthy full-term newborns after obtaining written informed consent. All procedures were performed in accordance with the guidelines of the Ethics Committee on the Use of Human Subjects at Asan Medical Center (IRB#: 2015-0303). All animal experiments were approved by and performed in accordance with the guidelines and regulations of the Institutional Animal Care and Use Committee of the University of Ulsan College of Medicine (IACUC-2019-12-221 and IACUC-2019-12-325).
## Culture of MSCs
Human ESC-derived MSCs (hES-MSCs) were established by differentiation from H9 hESCs11,12 and were maintained in EGM2-MV medium (Lonza, San Diego, CA, USA) on plates coated with rat tail collagen type I (Sigma-Aldrich, St. Louis, MO, USA), as described previously9,10,13. Human UC MSCs (hUC-MSCs) were isolated from UCs, as described previously24, and were grown in low-glucose DMEM containing $10\%$ heat-inactivated fetal bovine serum (HyClone, Pittsburgh, PA, USA), 5 ng/mL human epidermal growth factor (Sigma-Aldrich), 10 ng/mL basic fibroblast growth factor, and 50 ng/mL long-R3 insulin-like growth factor-1 (ProSpec, Rehovot, Israel), as described previously10,21,22. All MSCs used in this study were expanded for fewer than seven passages to ensure their functionality and were maintained at 37 °C in a humidified atmosphere containing $5\%$ CO2.
For GSH-enhancing priming, MSCs were plated at a density of 7 × 104 cells/mL in culture medium with the indicated concentration of AA2G (Sigma-Aldrich) for 3 days or 2 µM FSK (Sigma-Aldrich) for the indicated number of hours. The intracellular GSH level was rescued by supplementation with 0.125 µM GSH ethyl ester (GSH-EE; Sigma-Aldrich) for 4 h. The PFO procedure was performed by supplementation with AA2G, followed by treatment with low concentrations of sphingosine-1-phosphate (S1P) and valproic acid (VPA), as previously described23. In brief, MSCs were plated at a density of 7 × 104 cells/mL and maintained in culture medium with 0.74 mM (Sigma-Aldrich) for two days. One day before the functional evaluation, 50 nM S1P and 0.5 mM VPA (Sigma-Aldrich) were added to the culture medium containing 0.74 mM AA2G.
## RNA interference and ectopic expression of ATF2
For knockdown (KD) of ATF2, three independent shRNAs targeting human ATF2 were cloned into the pLenti6/Block-iT lentiviral vector (Invitrogen/Thermo Fisher Scientific, Waltham, MA, USA). For ectopic expression, the open reading frame (ORF) of human ATF2 in the pDONR223 plasmid (Addgene plasmid # 82889) was cloned into the pEZ-Lv235 (#EZ016, GeneCopoeia, Rockville, MD, USA) plasmid using the Gateway Technology reaction in accordance with the manufacturer’s instructions (Invitrogen/Thermo Fisher Scientific). Lentiviruses carrying each ATF2 shRNA or human ATF2 ORF were produced and used to infect hES-MSCs or hUC-MSCs, as described previously25. The sequences of the shRNAs are shown in Supplementary Table 1. The ORF of human ATF2 was kindly provided to us by Jesse Boehm, Matthew Meyerson, and David Root.
## In vitro cell proliferation, self-renewal, multipotency, and migration of MSCs
Several in vitro assays were performed to assess the cellular activities of MSCs. An MTT assay (Sigma-Aldrich) was used to assess cell proliferation, and a colony forming unit-fibroblast (CFU-F) assay was used to assess self-renewal. Multipotency (in vitro differentiation into chondrogenic, osteogenic, or adipogenic lineages) and transwell migration in response to platelet-derived growth factor (PDGF; 10 ng/mL PDGF-AA, R&D Systems, Minneapolis, MN, USA) were also assessed. Angiogenesis was quantified using Matrigel, and in vitro anti-inflammation was analyzed as described previously8–10,20,21. The digital images generated in these assays were assessed quantitatively using Image-Pro 5.0 software (Media Cybernetics, Rockville, MD, USA).
## Real-time monitoring of the GSH-recovery capacity of living MSCs
Real-time monitoring of the GSH-recovering capacity (GRC) of every living cell under different culture conditions was achieved using an Operetta High-Content Imaging System (HH12000000; PerkinElmer, Waltham, MA, USA) at ×200 or ×400 magnification, as described previously21. This system provides a nondestructive, integrated, and image-based high-throughput method for analyzing the qualitative and quantitative aspects of GSH dynamics in living MSCs. The GRC assay was based on the unique properties of FreSHtracer (Fluorescent real-time thiol tracer; Cell2in, Inc., Seoul, Korea), a reversible chemical probe for GSH9,26. Upon reacting with GSH, FreSHtracer shows a spectral shift in the λmax of its ultraviolet‒visible absorption from 520 nm to 430 nm, resulting in decreased fluorescence emission intensity at 580 nm (F580, λex 520 nm) and increased fluorescence intensity at 510 nm (F510, λex 430 nm)9,26. Thus, to determine the fluorescence ratios of FreSHtracer, fluorescence emissions were measured at 510 and 580 nm after excitation at 430 and 520 nm, respectively. These fluorescence signals were analyzed using Harmony High-Content Imaging and Analysis Software 3.1 (PerkinElmer) in confocal mode. The GSH dynamics indices, related initial fluorescence ratios (representing baseline total GSH), and slopes after diamide treatment (representing the GRC) of each plot are presented in Supplementary Dataset 1.
## Asthma animal model
Asthma was induced in 6-week-old female BALB/c mice (JA Bio, Suwon, Korea) by sensitization with intraperitoneal injections of 100 µg of ovalbumin (OVA, Sigma-Aldrich) and 2 mg of aluminum hydroxide (Sigma-Aldrich) on Days 0 and 7, followed by allergen challenge via intranasal injection of 50 μg of OVA on Days 14, 15, 16, 21, 22, and 23, as reported previously27. After 17 days, 3 × 105 hUC-MSCs stably expressing a control (shCTR) or ATF2-specific (shATF2) shRNA construct were suspended in 100 µL of phosphate-buffered saline (PBS) and injected via the tail vein. The same procedure was applied for the administration of hUC-MSCs or hES-MSCs, which were expanded under normal (naïve) culture conditions or using the PFO procedure. PBS alone was injected as a control (sham and asthma groups). Mice were randomly allocated to treatment groups, and the order of allergen sensitization or challenge and injection of MSCs or vehicle was randomized. Treatment groups were masked to investigators who participated in the therapeutic evaluation procedures.
## Analysis of airway inflammation
For mechanistic insights into MSC therapy, airway inflammation was evaluated by histological examination and bronchoalveolar lavage fluid (BALF) analysis, as well as via analyses of the expression levels of cytokine genes and proteins in the lung, as reported previously9,10. Therapeutic outcomes were analyzed using two independent sets of five animals per group. All histological, BALF, and cytokine analyses were performed by blinded investigators.
For analysis of engraftment of the hUC-MSCs, human β2-microglobulin (hB2M) was detected using a mouse monoclonal antibody (SC80668; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and an Alexa Fluor 488-labeled anti-mouse secondary antibody (Invitrogen).
Differentiation lineage was determined by costaining for hB2M and prosurfactant protein C (SFTPC) rabbit polyclonal antibody (ab90716; Abcam, Cambridge, UK) and visualization with an Alexa Fluor 546-labeled anti-rabbit secondary antibody (Invitrogen). Nuclei were counterstained using DAPI (Sigma-Aldrich). Digital images were selected at random from each slide and used for quantification using Image-Pro 5.0 software.
*For* gene expression analyses, total RNA was isolated from frozen lung tissues using the RNeasy Mini Kit (Qiagen, Hilden, Germany) and treated with DNase I (Qiagen). Total RNA (800 ng) was reverse-transcribed with TaqMan Reverse Transcription Reagent (Applied Biosystems, Foster City, CA), and the threshold cycle (Ct) was subsequently determined via real-time quantitative PCR (RQ-PCR), as described previously28. The relative expression level of each target gene was determined using the 2-ΔΔCt method, with Gapdh as the endogenous control gene. All primers used in the RQ-PCR assay are listed in Supplementary Table 2.
## Statistical analysis
Statistical significance was evaluated by the nonparametric Mann‒Whitney test and one-way or two-way ANOVA with the Bonferroni post hoc test using GraphPad Prism 7.0 software (GraphPad Software, La Jolla, CA, USA); $p \leq 0.05$ was considered statistically significant.
## Priming to enhance GSH levels activates ATF2 in MSCs
In a previous transcriptome analysis, a population of hES-MSCs with a high level of GSH (GSHHigh) was characterized by upregulation of the genes encoding ATF2 and other structurally related activating protein-1 (AP1) proteins, such as JUN, JUNB, and FRA121,29. This finding was validated by increased levels of the proteins encoded by these genes, as well as their phosphorylated active counterparts, in GSHHigh hES-MSCs21. ATF2 forms a heterodimer with several AP1 proteins, binds to the CRE to regulate gene expression and is activated by several extracellular stimuli, such as hypoxia, oxidative stress, and DNA damage29–31. Therefore, we examined whether priming to enhance GSH levels could activate cAMP-dependent ATF2 and affect the CREB1-NRF2 signaling cascade in MSCs derived from different sources (Fig. 1a). To this end, we examined the expression level and activity of ATF2 in hES-MSCs and hUC-MSCs cultured in medium with different concentrations (0, 0.37, 0.74, and 1.48 mM) of AA2G for 72 h. RQ-PCR and western blot analyses showed that the transcript and protein levels of ATF2 were increased by AA2G priming in both hES-MSCs and hUC-MSCs (Fig. 1b–d), peaking at the 0.74 mM AA2G concentration. This upregulation was accompanied by an increase in the level of active ATF2 protein, which is phosphorylated at threonine 69 (Thr69) and 71 (Thr71) via mitogen-activated protein kinases such as p38, JNK, and ERK32. Consistent with these results, the expression levels of a subset of ATF2 target genes, including PDGFRA, MMP2, and PLAU, were increased following AA2G priming, and this effect was greater in hUC-MSCs than in hES-MSCs (Fig. 1e and Supplementary Fig. 1a).Fig. 1AA2G priming activates ATF2 and NRF2 in human MSCs.a A schematic overview of the ATF2 and CREB1-NRF2 cascades involved in GSH dynamics in human MSCs. b RQ-PCR analysis ($$n = 4$$) of the ATF2 transcript in the AA2G-treated hES-MSCs and hUC-MSCs. c Western blot analyses ($$n = 3$$) of total (t-ATF2) and phosphorylated ATF2 (p-ATF2) proteins in the AA2G-treated hES-MSCs and hUC-MSCs. The expression level of β-actin was used as a loading control. Molecular weight marker sizes (kD) are shown on the left. d Quantification of the western blotting data described in c. e RQ-PCR analyses ($$n = 4$$) of ATF2 target genes following treatment of hES-MSCs and hUC-MSCs with AA2G for 72 h. f Schematic summary of the PFO procedure, which included supplementation with 0.74 mM AA2G for two days, followed by further stimulation with 50 nM sphingosine 1-phosphate (S1P) and 0.5 mM valproic acid (VPA) one day before functional evaluation. RQ-PCR ($$n = 4$$; g) and western blot analyses ($$n = 3$$; h) for the expression of ATF2 in hES-MSCs and hUC-MSCs under normal (naïve) or PFO culture conditions. i RQ-PCR analyses ($$n = 4$$) of NRF2 and NQO1 in the hES-MSCs and hUC-MSCs primed with 0.74 mM AA2G for 72 h. j Representative confocal microscopy images of the NRF2 protein (green) in the hES-MSCs and hUC-MSCs treated with or without AA2G. Magnification, ×1000. Scale bar, 10 µm. Nuclei were stained with DAPI (blue). b, d–i Data are represented as ratios relative to the nontreated cells (-AA2G). All quantification results are shown as the mean ± SEM (*$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ compared with nontreated cells, via two-way ANOVA).
We previously reported that further stimulation of AA2G-primed MSCs with a low concentration of S1P and VPA was beneficial for preserving primitive MSCs, characterized morphologically by their small size and high GSH dynamics23. In this regard, we examined whether the expression of ATF2 could be affected by the PFO procedure based on the combination of three small molecules, AA2G, S1P, and VPA (Fig. 1f). The PFO procedure increased ATF2 transcript and protein expression levels in both hES- and hUC-MSCs (Fig. 1g, h), resulting in the upregulation of ATF2 target genes (Supplementary Fig. 1b).
Next, we examined the effect of FSK priming on ATF2 expression. FSK treatment of hUC-MSCs increased the ATF2 transcript level only minimally (Supplementary Fig. 1c) but significantly increased the levels of total and phosphorylated ATF2 proteins, with peaks occurring 4 and 2 h after FSK priming, respectively (Supplementary Fig. 1d). In hES-MSCs, FSK priming had slight effects on the levels of the ATF2 transcript and protein (Supplementary Fig. 1e, f). Overall, these data demonstrate that GSH-enhancing priming conditions can regulate the expression and activity of ATF2 in human MSCs, depending on the particular cell context and priming factors used. Since AA2G treatment stably activated ATF2 in both hES-MSCs and hUC-MSCs, we used the optimal dose of 0.74 mM AA2G in subsequent studies.
## ATF2 regulates redox homeostasis in MSCs by crosstalking with the NRF2 pathway
We then investigated whether ATF2 can directly modulate the CREB1-NRF2 signaling cascade in MSCs. In both hES-MSCs and hUC-MSCs, AA2G priming increased the mRNA expression levels of NRF2 and NQO1, a well-known NRF2 target gene (Fig. 1i), and stimulated translocation of the NRF2 protein into the nucleus, indicating activation of the NRF2 pathway (Fig. 1j). Consistent with these results, AA2G priming also increased the expression levels of genes related to GSH synthesis (GCLC and GCLM) and redox cycling (GSR and PRDX1) (Fig. 2a), which have been reported as targets of the CREB1-NRF2 pathway that play a role in the maintenance of redox homeostasis in MSCs21. NRF2 and its targets (GCLC, GCLM, GSR, and PRDX1) were decreased at the mRNA and protein levels in MSCs harboring shRNAs targeting ATF2 (shATF2) (Fig. 2b, c, and Supplementary Fig. 2a–d). Notably, KD of ATF2 also impaired the AA2G-mediated nuclear translocation of the NRF2 protein (Fig. 2d and Supplementary Fig. 2e) and induction of the GSH-related genes targeted by CREB1-NRF2 (GCLC, GCLM, PRDX1, and GSR) (Fig. 2e and Supplementary Fig. 2f), suggesting an interplay between the ATF2 and CREB1-NRF2 signaling cascades in MSCs. The majority of ATF2 target genes activated by AA2G priming were downregulated by KD of ATF2 (Supplementary Fig. 2g).Fig. 2Crosstalk between the ATF2 and NRF2 pathways in AA2G-primed MSCs.a RQ-PCR analyses ($$n = 4$$) of CREB1-NRF2-dependent GSH synthesis (GCLC and GCLM) and redox cycling (GSR and PRDX1) genes in AA2G-primed MSCs. Expression levels were calculated as the ratio of the value of AA2G-primed MSCs to the nontreated cells (set to 1; see the red dotted line). RQ-PCR analysis ($$n = 4$$) of the NRF2 transcript (b) and western blot analyses of NRF2 protein (c) in the MSCs expressing a scrambled control shRNA (shCTR) or an ATF2-specific shRNA (shATF2). Two independent shATF2 constructs were used. d Representative confocal microscopy images of the NRF2 protein (green) in the hUC-MSCs treated with or without AA2G and expressing shCTR or shATF2. Nuclei were stained with DAPI (blue). Magnification, ×1000. Scale bar, 10 µm. e Western blot analyses of NRF2 target genes in the AA2G-primed hES-MSCs and hUC-MSCs expressing shCTR or shATF2. Molecular weight marker sizes (kD) are shown on the left of the blots. f F510/F580 fluorescence ratio (FR) plots of the hES-MSCs and hUC-MSCs carrying the indicated shCTR or shATF2 constructs. The GSH dynamics index (GI) for each sample ($$n = 3$$) was quantified based on both the initial FR (representing the baseline of total GSH) and the slope after 0.1 or 0.2 mM diamide treatment (representing the GRC). Representative images of F510 (GSH bound) and F580 (GSH free) fluorescence are shown in Supplementary Fig. 3b. All quantification results are shown as the mean ± SEM. Statistical analyses were performed via one-way (b) or two-way (a and f) ANOVA with Bonferroni post hoc tests (*$p \leq 0.05$, ***$p \leq 0.001$ compared with nontreated or shCTR cells).
To investigate the biological relevance of these findings, we used a high-throughput GRC assay (Supplementary Fig. 3), which enables real-time monitoring of the qualitative and quantitative aspects of GSH dynamics in living cells using a reversible chemical probe21. Changes in the intracellular GSH level were monitored for approximately 1 h after exposure to diamide, a thiol-specific oxidant. KD of ATF2 decreased the basal level of GSH and impaired the GRC following diamide treatment in both hES-MSCs and hUC-MSCs (Fig. 2f), indicating a crucial role of ATF2 in maintaining GSH dynamics in MSCs. Collectively, these results demonstrate that ATF2 acts as a novel mediator of redox homeostasis in MSCs by crosstalking with the NRF2 signaling cascade.
## ATF2 regulates the core functions of MSCs
To explore its role in maintaining the characteristics of MSCs that influence their therapeutic potency, we silenced ATF2 in hES-MSCs by infecting the cells with lentiviruses harboring two independent shATF2 constructs. KD of ATF2 had little effect on the expression of surface markers characteristic of MSCs, including CD29, CD73, and CD105 (Supplementary Fig. 4a). Furthermore, KD of ATF2 had little effect on the in vitro differentiation of hES-MSCs into the chondrogenic, adipogenic, and osteogenic lineages, which were evaluated by an increased level of cartilage proteoglycans (Alcian Blue staining), accumulation of lipid droplets (Oil Red O staining), and mineral deposition (Alizarin Red S staining), respectively (Supplementary Fig. 4b).
KD of ATF2 in hES-MSCs decreased the potency of CFU-F, which represents the presence of true clonogenic progenitor cells (Fig. 3a), but did not significantly affect cell proliferation (Supplementary Fig. 4c). A transwell chemotactic assay revealed that ATF2-KD hES-MSCs exhibited a severe defect in PDGF-stimulated cell migration (Fig. 3b). Furthermore, the ability of conditioned medium (CM) from ATF2-KD hES-MSCs to induce angiogenesis in a Matrigel tube formation assay was lower than that of CM from MSCs harboring a control shRNA (shCTR) construct (Fig. 3c).Fig. 3ATF2 is critical for maintaining the core functions of hES-MSCs. The effects of silencing ATF2 in hES-MSCs on colony forming unit-fibroblast (CFU-F) potency ($$n = 3$$; a), chemotaxis ($$n = 7$$; b) in response to treatment with 10 ng/mL PDGF-AA, and angiogenesis in an in vitro Matrigel tube formation assay ($$n = 4$$; c). Cells expressed a scrambled control (shCTR) or ATF2-specific (shATF2) shRNA. Two independent shATF2 constructs were used. Representative results for each assay are shown on the left (b: magnification, ×200; scale bar, 100 µm; c: magnification, ×40; scale bar, 200 µm). For the Matrigel tube formation assay, conditioned medium was prepared from the indicated hES-MSCs, and saline and recombinant human VEGF-A were used as negative and positive controls, respectively. Quantitative data are presented as ratios relative to the shCTR cells and are expressed as the mean ± SEM. Statistical analyses were performed via one-way ANOVA (*$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ compared with shCTR cells).
As observed in hES-MSCs, KD of ATF2 had minimal effects on the basic characteristics of MSCs, including surface marker phenotypes, multipotency, and cell proliferation (Fig. 4a, Supplementary Fig. 5a, b), but significantly impaired the core functions of hUC-MSCs, including the potency of CFU-F (self-renewal) and PDGF-responsive chemotaxis capacity (Fig. 4b, c, and Supplementary Fig. 5c).Fig. 4ATF2-silenced hUC-MSCs display defective therapeutic functions. Analyses of cell proliferation ($$n = 6$$; a), CFU-F potency ($$n = 5$$; b), and chemotaxis in response to treatment with 10 ng/mL PDGF-AA ($$n = 7$$; c) in hUC-MSCs harboring a scrambled (shCTR) or ATF2-specific (shATF2) shRNA. c Representative examples of chemotactic assays (magnification, ×200; scale bar, 100 µm) are shown next to the corresponding quantitative data. d, e Anti-inflammation assays using conditioned medium (CM) prepared from the AA2G-treated hUC-MSCs. d Quantification of TNF-α and IL-6 proteins secreted from MH-S cells that were stimulated with LPS for 8 h in the absence or presence of CM harvested from the indicated cells ($$n = 6$$). CM from IMR90 normal primary fibroblasts was used as a control. e RQ-PCR analyses of the expression levels of selected murine inflammatory genes in the LPS-stimulated MH-S cells ($$n = 6$$). All quantitative data are represented as fold changes relative to the shCTR group and are displayed as the mean ± SEM. Statistical analyses were performed via one-way ANOVA (*$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ compared with shCTR cells).
To examine the anti-inflammatory response of MSCs, we collected CM from hUC-MSCs and applied it to MH-S murine alveolar macrophages that were pretreated with lipopolysaccharide (LPS). As reported previously10,21,23, CM from hUC-MSCs expressing shCTR significantly reduced secretion of the proinflammatory cytokines tumor necrosis factor-α (TNF-α) and interleukin-6 (IL-6) by LPS-stimulated MH-S cells, whereas CM collected from control cells (IMR90 human lung fibroblasts) did not (Fig. 4d). Notably, CM from ATF2-KD hUC-MSCs was less able to repress the secretion of TNF-α and IL-6 by LPS-stimulated MH-S cells than CM from control hUC-MSCs (Fig. 4d), indicating impairment of the anti-inflammatory potency of ATF2-KD cells. These results were validated further by examining the repressive effects of CM from the control and ATF2-KD cells on the expression levels of various proinflammatory genes in the LPS-treated MH-S cells (Fig. 4e and Supplementary Fig. 6).
Importantly, the ectopic expression of ATF2 enhanced the PDGF-responsive chemotaxis activities in both hES- and hUC-MSCs (Supplementary Fig. 7a–e). In addition, the proangiogenic or anti-inflammatory capacities were increased by the overexpression of ATF2 in hES- or hUC-MSCs, respectively (Supplementary Fig. 7c, f, and g). Collectively, the results of these in vitro functional assays indicate that ATF2 plays a critical role in preserving the primitive state of MSCs, as evidenced by their improved self-renewal, migratory, proangiogenic, anti-inflammatory, and immunomodulatory activities, all of which are related to their therapeutic potency.
## The role of GSH dynamics in the ATF2-dependent functionality of MSCs
Next, we investigated whether the ATF2-mediated functionality of MSCs is dependent on the intracellular GSH level. To this end, ATF2-KD MSCs, which showed a reduced basal level of GSH and GRC (Fig. 2f), as well as impaired clonogenic and migratory capacities (Figs. 3 and 4), were treated with GSH-EE, a cell-permeable form of GSH. Notably, the defects in the potency of CFU-F (self-renewal) and PDGF-responsive chemotaxis capacity observed in ATF2-KD MSCs were rescued by treatment with GSH-EE (Fig. 5a−c).Fig. 5The role of GSH dynamics in the ATF2-dependent functionality of MSCs. Chemotaxis (a and b) in response to treatment with 10 ng/mL PDGF-AA ($$n = 7$$; b) and CFU-F potency ($$n = 3$$; c) in the control (shCTR) and ATF2-silenced (shATF2) MSCs with 0.125 mM GSH-EE. Chemotaxis (d and e) in response to treatment with 10 ng/mL PDGF-AA ($$n = 7$$; e) and CFU-F potency ($$n = 3$$; f) in the control (shCTR) and ATF2-silenced (shATF2) MSCs treated with 0.74 mM AA2G for 72 h. Representative images of the chemotaxis assays are presented on the left (magnification, ×200; scale bar, 100 µm). Quantitative data are represented as fold changes relative to the shCTR group and are displayed as the mean ± SEM. Statistical analyses were performed via two-way ANOVA (*$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ compared with shCTR cells; #$p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$ compared with nontreated cells).
This result led us to explore whether ATF2 is involved in the beneficial effects of GSH-enhancing priming by supplementation with AA2G. Consistent with previous reports10,23, AA2G priming increased PDGF-responsive migration and self-renewal (CFU-F potency) in both hES-MSCs and hUC-MSCs (Fig. 5d−f and Supplementary Fig. 8). These beneficial effects of AA2G were impaired significantly by silencing ATF2, indicating the critical role of ATF2 in the mode of action of AA2G priming of MSCs. As previously reported23, the PFO procedure based on AA2G supplementation increased the core functions of MSCs, including self-renewal (CFU-F), PDGF-responsive cell migration, and the proangiogenic, anti-inflammatory, and immunomodulatory capacities of MSCs. Importantly, the silencing of ATF2 significantly impaired these beneficial effects of the PFO procedure in both hES-MSCs and hUC-MSCs (Supplementary Fig. 9), further demonstrating the importance of ATF2 in AA2G-based GSH-enhancing priming conditions. Overall, these results demonstrate the importance of ATF2 as a novel mediator of GSH dynamics and related primitiveness during ex vivo expansion and priming of MSCs.
## The importance of ATF2 for the use of MSCs in asthma therapy in vivo
To confirm the findings described above in vivo, we employed an OVA mouse model of allergenic asthma, which represents a Th2 immune cell-driven inflammatory airway allergic response27, and compared the therapeutic potencies of a single intravenous injection of 3 × 105 hUC-MSCs expressing a control (shCTR) or ATF2-specific (shATF2) shRNA (Fig. 6a). As reported previously33, severe inflammation in the bronchial and vascular areas of the lung tissues was observed in the OVA-sensitized asthmatic mice administered PBS vehicle (Fig. 6b). Examination of the BALF revealed that these OVA-induced asthmatic mice displayed a significant increase in the overall cellularity and abundance of inflammatory cells (Fig. 6c). Single administration of hUC-MSCs expressing shCTR attenuated the inflammation of the lung tissue and infiltration of inflammatory cells, including macrophages, neutrophils, and lymphocytes, in the BALF (Fig. 6b, c). However, these beneficial effects were significantly defective in asthmatic mice injected with hUC-MSCs expressing shATF2.Fig. 6Silencing of ATF2 impairs the therapeutic effects of MSCs in a murine model of allergic asthma.a Schematic overview of the experimental protocols for the induction of asthma and the intravenous (i.v.) injection of 3 × 105 hUC-MSCs harboring a control (shCTR) or ATF2-specific shRNA (shATF2). OVA ovalbumin, alumn aluminum hydroxide, i.p. intraperitoneal injection, i.n. intranasal injection. b Hematoxylin and eosin staining of lung tissues (magnification, ×40, scale bar, 250 µm) from the sham mice and the OVA-induced mice injected with vehicle control (PBS) or hUC-MSCs expressing shCTR or shATF2. Higher magnification images (×200) are shown in the lower panels. Scale bar, 100 µm. c The numbers of total cells, macrophages, neutrophils, lymphocytes, and eosinophils identified via cytospin staining of BALF from mice ($$n = 20$$) in the indicated groups. Representative images of cytospin staining (magnification, ×400) are also shown. Scale bar, 50 µm. d Quantification of IL-4, IL-5, and IL-13 proteins in BALF from mice ($$n = 10$$) in the indicated groups. e RQ-PCR analyses of the indicated cytokines in lung tissues from mice ($$n = 20$$) in the indicated groups. Quantitative data are represented as the mean ± SEM. Statistical significance was examined via one-way ANOVA with Bonferroni post hoc tests (*$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ compared with the shCTR group; #$p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$ compared with the PBS group). The exact p values and numbers of replicates are available in Supplementary Dataset 2.
Consistent with these findings, the levels of the IL-4, IL-5, and IL-13 proteins, all of which are Th2 immune response mediators, were lower in the BALF of asthmatic mice administered hUC-MSCs expressing shCTR than in that of asthmatic mice administered the vehicle control; however, this was not the case for mice administered hUC-MSCs expressing shATF2 (Fig. 6d). The asthmatic mice were characterized by upregulation of genes related to the Th2-mediated immune response (e.g., Il-4, Il-5, and Il-13) and proinflammatory cytokines (e.g., Ccl7, Il6, Tnf-α, and Inf-γ); these increases were attenuated by single administration of hUC-MSCs expressing shCTR but not by administration of those expressing shATF2 (Fig. 6e, Supplementary Fig. 10a, b). Overall, this preclinical study demonstrates that ATF2 plays a crucial role in MSCs to alleviate airway inflammatory responses.
Next, we examined whether ATF2 could modulate the in vivo engraftment of hUC-MSCs in the lung tissues of OVA-induced asthmatic mice. Staining of the lung tissues with an hB2M-specific antibody revealed that the frequencies of hB2M+ engrafted cells were comparable in the mice administered hUC-MSCs expressing shCTR and those administered hUC-MSCs expressing shATF2 (Fig. 7a, b, and Supplementary Fig. 10c), indicating that ATF2 had a minimal effect on in vivo engraftment of the MSCs. Confocal microscopy analyses of the lung tissues revealed that the hB2M+ engrafted cells were not stained with an antibody targeting SFTPC, a type 2 alveolar epithelial cell marker. Instead, the majority of hB2M+ cells were located in proximity to SFTPC+ cells in both the shCTR and shATF2 groups (Fig. 7c and Supplementary Fig. 10d). Therefore, these results indicate that the engrafted cells protected against the airway inflammation response via a paracrine effect rather than by directly contributing to tissue-resident cells. Fig. 7Immunostaining analysis of the engraftment and cellular properties of transplanted hUC-MSCs.a Immunostaining to detect cells expressing hB2M (green) in the lung tissues of OVA-stimulated asthmatic mice 1 week after injection of PBS vehicle or hUC-MSCs harboring a control (shCTR) or ATF2-specific (shATF2) shRNA (magnification, ×1000; scale bar, 200 µm). Lower magnification images (×200) are shown in Supplementary Fig. 10c. b Quantification of the engrafted hB2M+ cells in lung tissues from mice ($$n = 5$$) in the indicated groups. Quantitative data are shown as the mean ± SEM. Statistical significance was examined via one-way ANOVA with Bonferroni post hoc tests (***$p \leq 0.001$ compared with the shCTR group; ###$p \leq 0.001$ compared with the PBS group). c Representative confocal micrographs showing the immunohistochemical detection of hB2M (green) and the alveolar epithelial cell marker SFTPC (red) in lung tissues from mice in the indicated groups. Nuclei were stained with DAPI (blue). Two independent images are shown (magnification, ×1000; scale bar, 200 µm). Negative control experiments using mouse and rabbit IgG control antibodies are shown in Supplementary Fig. 10d.
We next investigated whether MSC priming with ATF2 activation could be beneficial for treating allergic asthma. To address this issue, we treated OVA-sensitized asthmatic mice with hES- or hUC-MSCs, which were expanded ex vivo by normal culture (naïve) or the PFO procedure. Compared with naïve MSCs, both MSCs by the PFO procedure showed superior therapeutic efficacy, based on the findings of the better attenuation of lung inflammation and the infiltration of inflammatory cells, particularly macrophages and neutrophils (Supplementary Figs. 11a–c and 12a–c). The frequencies of hB2M+ engrafted cells were increased in the lungs of mice administered PFO-primed MSCs (Supplementary Figs. 11d and 12d), indicating that the superior in vivo engraftment of PFO-MSCs could be responsible for their improved therapeutic potency compared with that of naïve MSCs.
## Discussion
GSH dynamics are important for preserving both the primitive state of MSCs and their therapeutic efficacy toward several intractable disorders9,21,23,34. The results presented here demonstrate that ATF2 is a novel mediator of GSH dynamics in MSCs that acts via interplay with the NRF2 signaling cascade, thereby affecting the functionality and therapeutic potency of MSCs toward allergic asthma.
Several reports have described protocols for the stable preservation and ex vivo expansion of primitive MSCs with high therapeutic potency, for example, by enriching small-sized cells or enhancing the antioxidant capacity9,10,20,21,23. In these previous reports, transcriptome analyses revealed that MSCs with high levels of GSH displayed common molecular features, including activation of the CREB1-NRF2 pathway, resulting in the upregulation of genes related to GSH synthesis and redox cycling, as well as high expression levels of AP1 family transcription factors such as ATF2, JUN, JUNB, and FRA1. In our recent study, we found that the FOS proto-oncogene AP1 protein played a central role in maintaining both the core functions of hES-MSCs in vitro and the in vivo engraftment of transplanted hESC-MSCs, thus affecting their therapeutic potency in a preclinical study of interstitial cystitis/bladder pain syndrome16. AP1 activity is responsive to extracellular signals29, and the functions of AP1 complexes are diverse due to their ability to form distinct heterodimers. For example, ATF2 reportedly forms eight different complexes with other members of the ATF, JUN, and FOS families, whereas JUN can form 15 different dimeric complexes. Therefore, investigating the role of the interplay between these AP proteins in controlling GSH dynamics could not only advance our understanding of the molecular signature of the primitiveness of MSCs but also overcome the technical limitations of current MSC therapies.
Here, we found that ATF2 is a crucial mechanomediator of redox homeostasis in hES-MSCs and hUC-MSCs that acts by collaborating with the CREB1-NRF2 pathway. This finding is consistent with a previous report demonstrating that the ATF2 protein forms complexes with NRF2 and other multiple basic-leucine zipper proteins and is recruited to promote the heme oxygenase-1 gene following arsenite treatment35. In addition, another study found that ectopic expression of ATF2 inhibited ferroptosis induced by a bromodomain and extraterminal domain protein inhibitor in human breast cancer cells by upregulating NRF2. The clinical relevance of the positive correlation between ATF2 and NRF2 was demonstrated via The Cancer Genome Atlas Program (TCGA) dataset analysis of breast, lung, and cervical tissues36. ATF2 is a stress-response protein that is upregulated by oxidative, inflammatory, and DNA damage stresses29. In esophageal squamous epithelial cancer cells, activation of ATF2 via phosphorylation of threonine residues 69 and 71 reduces oxidative stress-induced apoptosis and consequently reinforces cell cycle arrest by upregulating p21WAF1 and JUN37. In addition, resveratrol (3,5,4’-trihydroxystilbene), a naturally occurring polyphenol with antioxidant activity, reportedly increases the transcriptional activation potentials of CREB and ATF2 to mediate cytoprotective and tumor suppressive outcomes38. Therefore, the activation of ATF2 by diverse extracellular stimuli could affect the intracellular level and dynamics of GSH to modulate various biological processes, including inflammation, aging, tumorigenesis, and the primitiveness of stem cells.
In our current study, the expression level and activity of ATF2 were stimulated by priming human MSCs with AA2G, a VitC derivative, to enhance GSH dynamics. In our previous study10, we found that AA2B stably promoted the primitive state of MSCs and the naïve pluripotency of murine ESCs and overcame the critical drawbacks of VitC, which is extremely unstable in aqueous solution because it readily oxidizes to dehydroascorbate, leading to cellular toxicity. We also found that AA2G reproduced the known biological effects of VitC, including TET-dependent DNA demethylation in murine ESCs and suppression of p53 during the generation of murine iPSCs, and that activation of the CREB1 pathway accounted for the beneficial effects of AA2G in ESCs and MSCs10. Furthermore, priming with AA2G promoted the core functions of MSCs, including self-renewal (based on CFU-F activity), PDGF-responsive cell migration, and anti-inflammatory potency. The in vivo importance of these findings was demonstrated by using a polyinosinic:polycytidylic acid (poly-I:C)-induced murine asthma model representing viral infection pathogenesis10. In this study, ATF2 played a crucial role in the beneficial effects of the PFO procedure based on the combination of three small molecules, AA2G, S1P, and VPA, which significantly improved the therapeutic potency of MSCs from different sources for treating allergic asthma. In this regard, the present study elucidates a novel mode of action of various AA2G-based priming procedures, namely, the role of ATF2 in preserving GSH dynamics and the related primitiveness of MSCs. Therefore, we postulate that ATF2 might be responsible for the effects of AA2G priming in other populations of stem cells, such as human PSCs, neural stem cells, and hematopoietic stem cells. Further studies are required to verify this hypothesis. In addition, the direct target(s) or critical mediator(s) of ATF2 in controlling GSH dynamics should be identified in different cellular or microenvironmental contexts.
GSH-enhancing priming conditions, such as AA2G priming and the PFO procedure, could connect cellular redox signaling via numerous common pathways, such as the receptor tyrosine kinase and G protein-coupled receptor (GPCR) pathways39. GPCRs activate heterotrimeric G proteins in the plasma membrane; unlike the Gi (Gαi/o) subunit, the *Gs alpha* subunit protein (Gαs) is responsible for stimulating cAMP- and PKA-dependent pathways by activating adenylyl cyclase39. Notably, we previously found that the beneficial effects of AA2G priming in murine ESCs and human MSCs were prevented by treatment with melittin, which inhibits Gαs and stimulates Gαi/o, underlining the critical role of Gαs in AA2G priming-mediated effects. Furthermore, the GPCR-related genes GNAI1 and HTR2B were identified as components of the CREB1-associated gene networks generated to characterize the transcriptome of AA2G-treated murine ESCs and human MSCs10. Therefore, GPCR signaling could play a crucial upstream role in modulating ATF2-mediated GSH dynamics and the related functionality of MSCs following AA2G priming. Additional studies are required to investigate the importance of GPCRs in the ATF2-NRF2 pathway, focusing on the specific roles of Gα subunit proteins in MSCs under different ex vivo expansion conditions.
Since the burden of uncontrolled asthma is substantial and growing continuously40, the identification of pathway-specific approaches for the prevention and treatment of this disease is required to reduce costs and improve the quality of life of patients. Due to their strong anti-inflammatory and immunomodulatory effects on innate and adaptive immune cells, MSCs have been used to treat intractable asthma, which is a major cause of morbidity and mortality worldwide1,5. In preclinical studies using animal models representing different pathogeneses, including asthma caused by house dust mites, poly-I:C, or OVA stimulation, MSC therapy was effective in alleviating airway inflammatory responses, hyperresponsiveness, and remodeling7. In our current study, we used an OVA-stimulated murine asthma model to demonstrate the in vivo importance of ATF2 in the therapeutic potency of MSCs, particularly toward airway inflammation, which is an important pathophysiological feature of asthma. The results of this preclinical study are consistent with our previous studies showing that AA2G-primed MSCs or those with high levels of GSH exhibit enhanced therapeutic potency in a mouse model of virus-associated asthma9,10.
In these previous preclinical studies, the in vivo engraftment capacity of MSCs with high GSH dynamics was superior to that of control MSCs. In contrast, ATF2-silenced MSCs were engrafted into the lung at considerably higher levels than control MSCs. To further examine this unexpected finding, we investigated the properties and locations of the engrafted cells by costaining the hB2M and SFTPC proteins. When hES-MSCs were sorted based on the intracellular level of GSH, the MSCs that survived in the lungs expressed the SFTPC protein, indicating their direct contribution to the alveolar epithelium9. Although MSCs reportedly take on the gene expression profile of lung epithelial cells both in vitro and in vivo41–47 and are stimulated by tissue injury42, the transdifferentiation of mesodermal MSCs into surfactant protein-producing cells is rare in a normal physiological environment. In this regard, our previous study showed that AA2G-primed hES-MSCs engrafted into mouse lungs showed little expression of SFTPC10. Similarly, in our current study, we found that the hB2M+ engrafted cells were negative for SFTPC expression but were located in the proximity of SFTPC+ type 2 alveolar epithelial cells. Importantly, the anti-inflammatory capacity of hUC-MSCs was severely impaired by ATF2 silencing but enhanced by ATF2 overexpression. Taken together, these findings indicate that the MSCs engrafted into mouse lungs induced an anti-inflammatory response via a paracrine effect rather than by directly transdifferentiating into tissue-resident cells. The anti-inflammatory and immunosuppressive activities of MSCs are mediated by cell contact-dependent mechanisms involving B7-H148 and by the secretion of soluble factors such as IL-10, transforming growth factor-β, nitric oxide, prostaglandin E2, and indoleamine 2,3-dioxygenase49,50. Therefore, for determination of the mode of action of MSC therapies in asthma, further studies are required to investigate which mediators could be affected by ATF2.
In summary, this study demonstrates that ATF2 mediates GSH dynamics and the related functional and therapeutic ability of MSCs to alleviate inflammatory responses in an experimental asthma model. Moreover, this study provides an in vivo proof of concept that the expression or activity of ATF2 can be used as a biomarker for predicting and evaluating the functions of MSCs for their ex vivo expansion and therapeutic applications.
## Supplementary information
Supplementary Information Source dataset The online version contains supplementary material available at 10.1038/s12276-023-00943-z.
## References
1. Holgate ST. **Asthma**. *Nat. Rev. Dis. Prim.* (2015) **1** 15025. DOI: 10.1038/nrdp.2015.25
2. Yang J, Kim EK, Park HJ, McDowell A, Kim YK. **The impact of bacteria-derived ultrafine dust particles on pulmonary diseases**. *Exp. Mol. Med.* (2020) **52** 338-347. DOI: 10.1038/s12276-019-0367-3
3. Wilhelm C, Stockinger B. **Innate lymphoid cells and type 2 (th2) mediated immune responses - pathogenic or beneficial?**. *Front. Immunol.* (2011) **2** 68. DOI: 10.3389/fimmu.2011.00068
4. Lambrecht BN, Hammad H. **The immunology of asthma**. *Nat. Immunol.* (2015) **16** 45-56. DOI: 10.1038/ni.3049
5. Holgate ST, Polosa R. **Treatment strategies for allergy and asthma**. *Nat. Rev. Immunol.* (2008) **8** 218-230. DOI: 10.1038/nri2262
6. Zhang LB, He M. **Effect of mesenchymal stromal (stem) cell (MSC) transplantation in asthmatic animal models: a systematic review and meta-analysis**. *Pulm. Pharmacol. Ther.* (2019) **54** 39-52. DOI: 10.1016/j.pupt.2018.11.007
7. Srour N, Thebaud B. **Stem cells in animal asthma models: a systematic review**. *Cytotherapy* (2014) **16** 1629-1642. DOI: 10.1016/j.jcyt.2014.08.008
8. Jin HJ. **Senescence-associated MCP-1 secretion is dependent on a decline in BMI1 in human mesenchymal stromal cells**. *Antioxid. Redox Signal.* (2016) **24** 471-485. DOI: 10.1089/ars.2015.6359
9. Jeong EM. **Real-time monitoring of glutathione in living cells reveals that high glutathione levels are required to maintain stem cell function**. *Stem Cell Rep.* (2018) **10** 600-614. DOI: 10.1016/j.stemcr.2017.12.007
10. Lee S. **Ascorbic acid 2-glucoside stably promotes the primitiveness of embryonic and mesenchymal stem cells through ten-eleven translocation- and cAMP-responsive element-binding protein-1-dependent mechanisms**. *Antioxid. Redox Signal.* (2020) **32** 35-59. DOI: 10.1089/ars.2019.7743
11. Hong KS. **A porous membrane-mediated isolation of mesenchymal stem cells from human embryonic stem cells**. *Tissue Eng. Part C. Methods* (2015) **21** 322-329. DOI: 10.1089/ten.tec.2014.0171
12. Kim JM. **Perivascular progenitor cells derived from human embryonic stem cells exhibit functional characteristics of pericytes and improve the retinal vasculature in a rodent model of diabetic retinopathy**. *Stem Cells Transl. Med.* (2016) **5** 1268-1276. DOI: 10.5966/sctm.2015-0342
13. Ryu CM. **Longitudinal intravital imaging of transplanted mesenchymal stem cells elucidates their functional integration and therapeutic potency in an animal model of interstitial cystitis/bladder pain syndrome**. *Theranostics* (2018) **8** 5610-5624. DOI: 10.7150/thno.27559
14. Shin JH. **Safety of human embryonic stem cell-derived mesenchymal stem cells for treating interstitial cystitis: a phase I study**. *Stem Cells Transl. Med.* (2022) **11** 1010-1020. DOI: 10.1093/stcltm/szac065
15. Huang Y. **miR-19b enhances osteogenic differentiation of mesenchymal stem cells and promotes fracture healing through the WWP1/Smurf2-mediated KLF5/β-catenin signaling pathway**. *Exp. Mol. Med.* (2021) **53** 973-985. DOI: 10.1038/s12276-021-00631-w
16. Yu HY. **Intravital imaging and single cell transcriptomic analysis for engraftment of mesenchymal stem cells in an animal model of interstitial cystitis/bladder pain syndrome**. *Biomaterials* (2022) **280** 121277121277. DOI: 10.1016/j.biomaterials.2021.121277
17. ArefNezhad R, Motedayyen H, Mohammadi A. **Therapeutic aspects of mesenchymal stem cell-based cell therapy with a focus on human amniotic epithelial cells in multiple sclerosis: a mechanistic review**. *Int. J. Stem Cells* (2021) **14** 241-251. DOI: 10.15283/ijsc21032
18. Kiaie N, Ghanavati SPM, Miremadi SS, Hadipour A, Aghdam RM. **Mesenchymal stem cell-derived exosomes for COVID-19 therapy, preclinical and clinical evidence**. *Int. J. Stem Cells* (2021) **14** 252-261. DOI: 10.15283/ijsc20182
19. Heo J. **Sirt1 regulates DNA methylation and differentiation potential of embryonic stem cells by antagonizing Dnmt3l**. *Cell Rep.* (2017) **18** 1930-1945. DOI: 10.1016/j.celrep.2017.01.074
20. Kim Y. **Small hypoxia-primed mesenchymal stem cells attenuate graft-versus-host disease**. *Leukemia* (2018) **32** 2672-2684. DOI: 10.1038/s41375-018-0151-8
21. Lim J. **Glutathione dynamics determine the therapeutic efficacy of mesenchymal stem cells for graft-versus-host disease via CREB1-NRF2 pathway**. *Sci. Adv.* (2020) **6** eaba1334. DOI: 10.1126/sciadv.aba1334
22. Lim J. **Valproic acid enforces the priming effect of sphingosine-1 phosphate on human mesenchymal stem cells**. *Int. J. Mol. Med.* (2017) **40** 739-747. DOI: 10.3892/ijmm.2017.3053
23. Lim J. **Small‐sized mesenchymal stem cells with high glutathione dynamics show improved therapeutic potency in graft‐versus‐host disease**. *Clin. Transl. Med.* (2021) **11** e476. DOI: 10.1002/ctm2.476
24. Mushahary D, Spittler A, Kasper C, Weber V, Charwat V. **Isolation, cultivation, and characterization of human mesenchymal stem cells**. *Cytom. A* (2018) **93** 19-31. DOI: 10.1002/cyto.a.23242
25. Heo J. **The CDK1/TFCP2L1/ID2 cascade offers a novel combination therapy strategy in a preclinical model of bladder cancer**. *Exp. Mol. Med.* (2022) **54** 801-811. DOI: 10.1038/s12276-022-00786-0
26. Jeong EM. **Monitoring glutathione dynamics and heterogeneity in living stem cells**. *Int. J. Stem Cells* (2019) **12** 367-379. DOI: 10.15283/ijsc18151
27. Kang H. **Effect of Acinetobacter lwoffii on the modulation of macrophage activation and asthmatic inflammation**. *Clin. Exp. Allergy* (2021) **52** 518-529. DOI: 10.1111/cea.14077
28. Heo J. **Phosphorylation of TFCP2L1 by CDK1 is required for stem cell pluripotency and bladder carcinogenesis**. *EMBO Mol. Med.* (2020) **12** e10880. DOI: 10.15252/emmm.201910880
29. Lopez-Bergami P, Lau E, Ronai Z. **Emerging roles of ATF2 and the dynamic AP1 network in cancer**. *Nat. Rev. Cancer* (2010) **10** 65-76. DOI: 10.1038/nrc2681
30. Kim J, Wong PK. **Loss of ATM impairs proliferation of neural stem cells through oxidative stress‐mediated p38 MAPK signaling**. *Stem cells* (2009) **27** 1987-1998. DOI: 10.1002/stem.125
31. Yu T. **The regulatory role of activating transcription factor 2 in inflammation**. *Mediators Inflamm.* (2014) **2014** 950472. DOI: 10.1155/2014/950472
32. Huebner K, Prochazka J, Monteiro AC, Mahadevan V, Schneider-Stock R. **The activating transcription factor 2: an influencer of cancer progression**. *Mutagenesis* (2019) **34** 375-389. DOI: 10.1093/mutage/gez041
33. Ha EH. **Endothelial Sox17 promotes allergic airway inflammation**. *J. Allergy Clin. Immunol.* (2019) **144** 561-573 e566. DOI: 10.1016/j.jaci.2019.02.034
34. Watanabe J. **Preconditioning of bone marrow-derived mesenchymal stem cells with N-acetyl-L-cysteine enhances bone regeneration via reinforced resistance to oxidative stress**. *Biomaterials* (2018) **185** 25-38. DOI: 10.1016/j.biomaterials.2018.08.055
35. Gong P, Stewart D, Hu B, Vinson C, Alam J. **Multiple basic-leucine zipper proteins regulate induction of the mouse heme oxygenase-1 gene by arsenite**. *Arch. Biochem. Biophys.* (2002) **405** 265-274. DOI: 10.1016/S0003-9861(02)00404-6
36. Wang L. **ATF2 inhibits ani-tumor effects of BET inhibitor in a negative feedback manner by attenuating ferroptosis**. *Biochem. Biophys. Res. Commun.* (2021) **558** 216-223. DOI: 10.1016/j.bbrc.2020.08.113
37. Walluscheck D. **ATF2 knockdown reinforces oxidative stress-induced apoptosis in TE7 cancer cells**. *J. Cell Mol. Med.* (2013) **17** 976-988. DOI: 10.1111/jcmm.12071
38. Thiel G, Rössler OG. **Resveratrol stimulates cyclic AMP response element mediated gene transcription**. *Mol. Nutr. Food Res.* (2016) **60** 256-265. DOI: 10.1002/mnfr.201500607
39. Petry A, Görlach A. **Regulation of NADPH oxidases by G protein-coupled receptors**. *Antioxid. Redox Signal.* (2019) **30** 74-94. DOI: 10.1089/ars.2018.7525
40. Yaghoubi M, Adibi A, Safari A, FitzGerald JM, Sadatsafavi M. **The projected economic and health burden of uncontrolled asthma in the United States**. *Am. J. Respir. Crit. Care Med.* (2019) **200** 1102-1112. DOI: 10.1164/rccm.201901-0016OC
41. Krause DS. **Multi-organ, multi-lineage engraftment by a single bone marrow-derived stem cell**. *Cell* (2001) **105** 369-377. DOI: 10.1016/S0092-8674(01)00328-2
42. Rojas M. **Bone marrow-derived mesenchymal stem cells in repair of the injured lung**. *Am. J. Respir. Cell Mol. Biol.* (2005) **33** 145-152. DOI: 10.1165/rcmb.2004-0330OC
43. Krause DS. **Bone marrow–derived cells and stem cells in lung repair**. *Proc. Am. Thorac. Soc.* (2008) **5** 323-327. DOI: 10.1513/pats.200712-169DR
44. Kassmer SH, Bruscia EM, Zhang P-X, Krause DS. **Nonhematopoietic cells are the primary source of bone marrow-derived lung epithelial cells**. *Stem Cells* (2012) **30** 491-499. DOI: 10.1002/stem.1003
45. Cerrada A. **Human decidua-derived mesenchymal stem cells differentiate into functional alveolar type II-like cells that synthesize and secrete pulmonary surfactant complexes**. *PLoS One* (2014) **9** e110195. DOI: 10.1371/journal.pone.0110195
46. Carraro G. **Human amniotic fluid stem cells can integrate and differentiate into epithelial lung lineages**. *Stem Cells* (2008) **26** 2902-2911. DOI: 10.1634/stemcells.2008-0090
47. Liu A. **Wnt5a through noncanonical Wnt/JNK or Wnt/PKC signaling contributes to the differentiation of mesenchymal stem cells into type II alveolar epithelial cells in vitro**. *PLoS One* (2014) **9** e90229. DOI: 10.1371/journal.pone.0090229
48. Sheng H. **A critical role of IFNgamma in priming MSC-mediated suppression of T cell proliferation through up-regulation of B7-H1**. *Cell Res.* (2008) **18** 846-857. DOI: 10.1038/cr.2008.80
49. Krampera M. **Role for interferon-γ in the immunomodulatory activity of human bone marrow mesenchymal stem cells**. *Stem Cells* (2006) **24** 386-398. DOI: 10.1634/stemcells.2005-0008
50. Wang Y, Chen X, Cao W, Shi Y. **Plasticity of mesenchymal stem cells in immunomodulation: pathological and therapeutic implications**. *Nat. Immunol.* (2014) **15** 1009-1016. DOI: 10.1038/ni.3002
|
---
title: In vitro and in silico studies of 7′′,8′′-buddlenol D anti-inflammatory lignans
from Carallia brachiata as p38 MAP kinase inhibitors
authors:
- Nonthaneth Nalinratana
- Utid Suriya
- Chanyanuch Laprasert
- Nakuntwalai Wisidsri
- Preeyaporn Poldorn
- Thanyada Rungrotmongkol
- Wacharee Limpanasitthikul
- Ho-Cheng Wu
- Hsun-Shuo Chang
- Chaisak Chansriniyom
journal: Scientific Reports
year: 2023
pmcid: PMC9981598
doi: 10.1038/s41598-023-30475-5
license: CC BY 4.0
---
# In vitro and in silico studies of 7′′,8′′-buddlenol D anti-inflammatory lignans from Carallia brachiata as p38 MAP kinase inhibitors
## Abstract
Excessive macrophage activation induces the release of high levels of inflammatory mediators which not only amplify chronic inflammation and degenerative diseases but also exacerbate fever and retard wound healing. To identify anti-inflammatory molecules, we examined Carallia brachiata—a medicinal terrestrial plant from Rhizophoraceae. Furofuran lignans [(−)-(7′′R,8′′S)-buddlenol D [1] and (−)-(7′′S,8′′S)-buddlenol D [2]] isolated from the stem and bark inhibited nitric oxide (half maximal inhibitory concentration (IC50): 9.25 ± 2.69 and 8.43 ± 1.20 micromolar for 1 and 2, respectively) and prostaglandin E2 (IC50: 6.15 ± 0.39 and 5.70 ± 0.97 micromolar for 1 and 2, respectively) productions in lipopolysaccharide-induced RAW264.7 cells. From western blotting, 1 and 2 suppressed LPS-induced inducible nitric oxide synthase and cyclooxygenase-2 expression in a dose-dependent manner (0.3–30 micromolar). Moreover, analysis of the mitogen-activated protein kinase (MAPK) signaling pathway showed decreased p38 phosphorylation levels in 1- and 2-treated cells, while phosphorylated ERK$\frac{1}{2}$ and JNK levels were unaffected. This discovery agreed with in silico studies which suggested 1 and 2 bound to the ATP-binding site in p38-alpha MAPK based on predicted binding affinity and intermolecular interaction docking. In summary, 7′′,8′′-buddlenol D epimers demonstrated anti-inflammatory activities via p38 MAPK inhibition and may be used as viable anti-inflammatory therapies.
## Introduction
Inflammation is a cascade event generated by innate immunity responses against microbial infection1. Additionally, non-infectious stimuli such as cell damage and tissue injury also trigger these responses and lead to local inflammation. Uncontrolled acute inflammation gradually destroys tissues and organs, diminishes their functions, and evolves into chronic inflammation and degenerative diseases such as osteoarthritis, rheumatoid arthritis, and even cancer2. When exposed to chemoattractants, tissue monocytes evolve to macrophages which engulf and destroy microbes in phagolysosomes using various hydrolytic enzymes and substances, including nitric oxide (NO). Moreover, NO is released into the extracellular fluid and functions as an inflammatory mediator which enhances phagocyte migration to injured or infected sites to amplify immune responses. However, excessive NO production limits recovery processes and destroys surrounding tissues3,4. Toll-like receptors (TLRs), especially TLR-4 which is a pathogen recognition receptor, and mitogen-activated protein kinase (MAPK) signaling pathways have important roles during inflammatory responses. Their deactivation mitigates acute sepsis, chronic inflammation, and degenerative diseases5,6. Eritoran, a structural lipopolysaccharide (LPS) lipid A mimic, and SB203580, a pyridinylimidazole derivative, both target TLR-4 and p38 MAPK, respectively. Additionally, curcumin and its analog L48H37 are effective in inhibiting TLR-47,8. Furthermore, p38 MAPK inhibition is beneficial for Alzheimer’s disease treatment9. Natural compounds such as icariin, apigenin, and astaxanthin also decrease amyloid-β-induced neurotoxicity via p38 MAPK inhibition9.
Natural products are vital sources of medicines and have been used extensively by different populations; in Thailand, traditional medicines have been used for centuries. Based on Thai traditional medicine, *Carallia brachiata* (Lour.) Merr. ( Rhizophoraceae) is one such natural source whose stem and bark are used for antipyretic remedies. Similarly, in Ayurveda (traditional Hindu medicine system) the stem bark is also used to treat oral ulcers and stomatitis10. In a previous study, ethyl acetate (EtOAc) and methanol (MeOH) bark extracts showed significant wound healing properties in incision and excision wound rat models11. Furthermore, MeOH and hydro-ethanolic leaf extracts exhibited analgesic, anti-inflammatory, and anti-diabetic activities in in vivo models12,13. Based on these traditional uses and pharmacological activities, the anti-inflammatory activity of this plant appears to underlie its mechanism of action.
In terms of chemical investigations, two pyrrolidine alkaloids [hygroline and (+)-pseudohygroline], a megastigmane diglycoside [3-hydroxy-5,6-epoxy-β-ionol-3-O-β-apiofuranosyl-(1 → 6)-β-glucopyranoside], and flavonoid glycosides such as apigenin-7-O-α-rhamnoside-(1 → 2)-β-glucopyranoside were isolated from leaves of C. brachiata14–16. Also, p-hydroxy benzoic acid and two proanthocyanidins [carallidin and mahuanin A], which possessed anti-radical scavenging activities against 2,2-diphenyl-1-picrylhydrazyl (DPPH) and superoxide radicals and inhibited xanthine oxidase, were isolated from bark17. To the best of our knowledge, no evidence exists in the literature to support the compounds contributing to the activities addressed in traditional uses of this plants, especially for anti-inflammatory activity which is linked antipyretic and wound healing properties. In this work, we identified compounds which exerted anti-inflammatory activities, examined their effects on p38 MAPK signaling pathway, and developed in silico models to identify ligand binding targets.
## Compounds isolated from C. brachiata stem and bark
Eleven compounds were isolated from stem and bark. Of these, two sesquineolignans [(−)-(7′′R,8′′S)-buddlenol D [1] and (−)-(7′′S,8′′S)- buddlenol D [2]], together with (+)-7″R,8′′S:7″′R,8′′′S-hedyotisol A [3], (−)-syringaresinol [4], two modified stilbene dimers [(+)-diptoindonesin D [5] and (+)-parviflorol [6]], a proanthrocyanidin [(−)-mahuanin A [7]], 4-hydroxy-2-methoxyphenyl-6-O-syringoyl-β-D-glucopyranoside [8], and three benzoic acid derivatives [vanillic acid [9], protocatechuic acid [10], and syringaldehyde [11]] were identified by spectroscopic analysis and compared with the literature (Fig. 1). On silica gel GF254 thin-layer chromatography (TLC) plate, 1 and 2 were observed at retardation factor (Rf) values of 0.44 and 0.48, respectively, when triply developed in $20\%$ acetone/dichloromethane (Me2CO/CH2Cl2) (Supplementary Information: Figure S1).Figure 1Compounds isolated from EtOAc extract of C. brachiata stem and bark. ( −)-(7′′R,8′′S)-buddlenol D [1], (−)-(7′′S,8′′S)-buddlenol D [2], (+)-7″R,8′′S:7″′R,8′′′S-hedyotisol A [3], (−)-syringaresinol [4], (+)-diptoindonesin D [5], (+)-parviflorol [6], (−)-mahuanin A [7], 6-O-syringoyl-β-d-glucopyranoside derivative [8], vanillin [9], protocatechuic acid [10], and syringaldehyde [11].
## (−)-(7′′R,8′′S)-Buddlenol D (1) and (−)-(7′′S,8′′S)-buddlenol D (2) inhibit LPS-stimulated NO and PEG2 expression in RAW264.7 cells
To determine the anti-inflammatory activity of C. brachiata stem and bark, the crude MeOH and its partitioned extracts were preliminary tested for NO inhibition in murine macrophages (RAW264.7). The EtOAc extract was selected for further isolation. All isolates were screened for NO inhibition at a final concentration of 30 μM. Only buddlenols D 1 and 2, which exhibited > $50\%$ inhibition, underwent half maximal inhibitory concentrations (IC50) analysis. In pre-treated RAW264.7 cells with 1 and 2 prior to stimulation with 500 ng/mL LPS, 1 and 2 inhibited NO production in a concentration-dependent manner (0.3–30 µM) and exhibited no cytotoxicity (Supplementary Information: Figure S5), giving calculated IC50 values of 9.25 ± 2.69 and 8.43 ± 1.20 µM, respectively. Dexamethasone (10 µM) was used as a positive control showed percentage of NO inhibition at 78.69 ± $1.48\%$ (Fig. 2).Figure 2The inhibitory effects of 1 and 2 on (A) nitric oxide (NO) and (B) prostaglandin E2 (PGE2) production in lipopolysaccharide (LPS)-induced RAW264.7 macrophages. Dexamethasone was used as a positive control. Percentage inhibition was calculated using the LPS-induced group as a control. Data were represented as the mean ± standard error of the mean from three independent experiments ($$n = 3$$).
LPS also stimulates prostaglandin E2 (PGE2) production via the arachidonic acid pathway. Buddlenols D 1 and 2 were evaluated for PGE2 inhibition as a putative fever suppressant. Cell pretreatment with 1 and 2 inhibited LPS-induced PGE2 production in a concentration-dependent manner; IC50 values for 1 and 2 were 6.15 ± 0.39 and 5.70 ± 0.97 µM, respectively, whereas dexamethasone (10 µM) exhibited an inhibition percentage of 91.21 ± $1.17\%$ (Fig. 2).
## (−)-(7′′R,8′′S)-Buddlenol D (1) and (−)-(7′′S,8′′S)-buddlenol D (2) inhibit LPS-induced inducible nitric oxide synthase (iNOS) and cyclooxygenase-2 (COX2) expression
Since buddlenols D 1 and 2 suppressed NO and PGE2 production, their effects on iNOS and COX2 protein expression were evaluated by western blotting. RAW264.7 cells treated with 500 ng/mL LPS significantly increased iNOS and COX2 expression when compared with untreated cells. Cells pre-incubated with 1 and 2 markedly decreased iNOS and COX2 expression in LPS-stimulated cells in a concentration-dependent manner. Both iNOS and COX2 expression were more reduced in 1-treated cells than 2-treated cells, although the IC50 values for NO and PGE2 inhibition were not significantly different between groups. Dexamethasone (10 µM) almost completely ablated iNOS and COX2 expression (Fig. 3).Figure 3The effects of 1 and 2 on inducible nitric oxide synthase (iNOS) and cyclooxygenase 2 (COX2) protein expression. ( A) Representative immunoblots showing LPS-induced RAW264.7 macrophages treated with 1 and 2. The cropped blots were used in the figure. The protein lysates of each treatment group from the same experiment were performed simultaneously on the same gel for each protein detection. Full-length uncropped blots were shown in supplementary information data (Figure S59). Densitometric histograms showing (B) iNOS and (C) COX2 in LPS-induced RAW264.7 macrophages. Dexamethasone was used as a positive control. Data were expressed as relative intensity when compared with LPS-induced groups and were represented as the mean ± standard error of the mean from three independent experiments ($$n = 3$$). * $p \leq 0.05$ vs. the LPS-induced group.
## (−)-(7′′R,8′′S)-Buddlenol D (1) and (−)-(7′′S,8′′S)-buddlenol D (2) suppress LPS-induced p38 but not ERK1/2 and JNK phosphorylation
To investigate the possible intracellular mechanisms underpinning iNOS and COX2 inhibition by buddlenols D 1 and 2, we evaluated their effects on p38, extracelluar signal-regulated protein kinase 1 and 2 (ERK$\frac{1}{2}$), and c-Jun N-terminal kinase (JNK) phosphorylation. Treatment with 500 ng/mL LPS significantly increased p-p38, p-ERK$\frac{1}{2}$, and p-JNK levels detected at 24-h incubation when compared with non-stimulated cells. Dexamethasone (10 µM) also significantly suppressed these levels when induced by LPS and suggested its anti-inflammatory activity may be related to the inhibition of p38, ERK$\frac{1}{2}$, and JNK phosphorylation. In cells pre-treated with buddlenols 1 and 2, p38 phosphorylation was significantly reduced when compared with LPS-treated cells. However, ERK$\frac{1}{2}$ and JNK phosphorylation levels were not significantly changed. Therefore, the effects of 1 and 2 on NO and PGE2 suppression may be due to inhibited p38 phosphorylation (Fig. 4). In addition, the effect of a specific p38 inhibitor SB203580 on p38 kinase and NO inhibition was investigated. The results showed that SB203580 at 1 µM completely inhibited LPS-induced p38 phosphorylation and reduced NO production with an IC50 of 0.82 ± 0.05 µM, compared to IC50 values of 9.25 ± 2.69 and 8.43 ± 1.20 µM for 1 and 2, respectively (Supplementary Information: Figure S6).Figure 4The effects of 1 and 2 on mitogen-activated protein kinase (MAPK) signaling elements. ( A) Representative immunoblots showing LPS-induced RAW264.7 macrophages treated with 1 and 2. The cropped blots were used in the figure. The protein lysates of each treatment group from the same experiment were performed simultaneously on the same gel for each protein detection. Full-length uncropped blots were shown in supplementary information data (Figure S60). Densitometric histograms showing (B) phosphorylated ERK (p-ERK), (C) phosphorylated JNK (p-JNK), and (D) phosphorylated p38 (p-38) in LPS-induced RAW264.7 macrophages. Dexamethasone was used as a positive control. Data were expressed as relative intensity compared with LPS-induced groups and represented as the mean ± standard error of the mean from three independent experiments ($$n = 3$$). * $p \leq 0.05$ vs. LPS-induced groups.
## The in silico prediction of ligand binding targets and ADMET calculation
Since buddlenols D 1 and 2 suppressed NO and PGE2 production, molecular docking approaches were used to identify possible binding targets to understand inhibitory actions at the atomic level. The binding affinity with respect to predicted fitness scores of 1 and 2 and targets [TLR-4 and p38-α MAPK proteins] were examined and compared to reference ligands (Table 1).Table 1Predicted binding affinity termed fitness scores toward focused receptor targets of 1, 2, and reference ligands. CompoundsFitness score*/receptor targetsP38-α (ATP-binding site)P38-α (non ATP-binding site)TLR4BIRB-796–56.64–SB-20358046.68––ZINC25778142––25.55137.9728.2219.99235.2829.0120.18*Higher fitness scores indicated better binding affinity.
Molecular docking showed that 1 and 2 preferably bound to the ATP-binding pocket of p38-α rather than to the non-ATP (allosteric) site as their fitness scores for the allosteric pocket were much lower than BIRB-796, a p38-α non-ATP site inhibitor. For TLR-4, 1 and 2 exhibited approximately a 0.8-fold decrease in fitness score when compared with ZINC25778142, a TLR-4 inhibitor. Additionally, we investgatged protein–ligand intermolecular interactions in the ATP-binding site of p38-α MAPK and TLR4-MD2 (myeloid differentiation protien 2) interface using the most likely occurring conformation (highest fitness score) (Fig. 5A,B). These data showed 1 underwent hydrogen bonding interactions with H107, M109, and L171 and hydrophobic interactions with V30, Y35, V38, A51, and F169, whereas the amino acid residues K53, M109, and G170, and V30, Y35, V38, F169, and L171 interacted with 2 via hydrogen bonding and hydrophobic forces, respectively. For TLR-4, both buddlenols D 1 and 2 tended to interrupt the TLR4 interface via a transient construction of hydrogen bonds with D209, E230, and T235 for 1, and D181, D209, E230, and T259 for 2. Also, two hydrophobic interactions with H179 and W256 were observed in 1.Figure 5Focused ligand binding sites and ligand intermolecular interactions with surrounding amino acids of (A) p38-α MAPK at the ATP-binding site and (B) the TLR4-MD2 interface in complex with docked conformations of 1, 2, and reference compounds (SB203580 for P38-α MAPK at the ATP-binding site and ZINC25778142 for TLR-4); pink, green, and light blue denote hydrophobic, conventional hydrogen bond, and halogen interactions, respectively. This molecular recognition analysis was based on the rigid docked conformation predictions.
Regarding to the selectivity of 1 and 2 toward other kinases in MAPK signaling pathway, the binding affinity ratios of 1 and 2 compared to corresponding crystallized ligand were evaluated. The results showed that binding affinity ratios of the 1 and 2 were 0.81 and 0.76 for p38α MAPK, respectively, while their binding affinity ratios against other kinases in MAPK pathway were ranged from 0.44 to 0.64 (Table 2). The computation study indicated that 1 and 2 selectively inhibited the p38α MAPK, which was corresponded well to the western blot results (Fig. 4).Table 2Binding affinity of 1 and 2 toward kinase targets involved in the MAPK signaling cascade compared to the original crystallized ligand. Targets (PDB ID:)Fitness scoreBinding affinity ratio compared to corresponding crystallized ligandCrystallized ligand1212P38α MAPK (3ZSH)46.6837.9735.280.810.76ERK1 (4QTB)54.2830.5729.110.560.54ERK2 (1PME)45.3923.0823.180.510.51JNK1(2NO3)36.5422.9723.280.630.64JNK2 (3NPC)67.3530.5731.410.450.47JNK3 (4W4Y)47.9821.1927.870.440.58IκB kinase β (4KIK)50.6323.6525.860.470.51* The equation for binding affinity ratio was fitness score of the compound divided by fitness score of crystallized ligand.
The ADMET data (Table 3) showed that 1 and 2 were practically non-toxic and not irritant compounds, but they exhibited low bioavailability and violated the Lipinski’s rule-of-five (drug-likeness assessment).Table 3Calculated ADMET parameters of 1 and 2.Parameter *compound12I. Physicochemical properties and pharmacokinetics MW (g/mol)644.66644.66 iLOGP4.214.49 HBD44 HBA1313 TPSA (Å2)163.99163.99 RB1313 nHA4646 MR163.79163.79 Log S (ESOL)− 4.76 (moderately soluble)− 4.76 (moderately soluble) GI absorptionLowLow BBB permeantNoNo P-gp substrateNoNo CYP inhibitor (CYP1A2, 2C19, 2C9, 2D6, 3A4)NoNo Lipinski’s rule-of-fiveNo; 2 violations: MW > 500, nitrogens or oxygens > 10No; 2 violations: MW > 500, nitrogens or oxygens > 10 Bioavailability score0.170.17II. Toxicity LD501500 mg/kg1500 mg/kg Toxicity class44 MutagenicityNoNo TumorigenicityLowLow IrritantNoNo Reproductive effectNoNo*The acronyms were referred to their full terms as follows: MW molecular weight, iLOGP octanol/water partition coefficient, HBD number of H-bond donors, HBA number of H-bond acceptors, TPSA topological polar surface area, RB rotatable bonds, nHA number of heavy atoms, MR molar refractivity, Log S (ESOL) decimal logarithm of the molar solubility in water (ESOL model), GI absorption gastrointestinal absorption, BBB permeant blood–brain barrier permeant, P-gp substrate permeability glycoprotein substrate, CYP inhibitor cytochrome P450 inhibitor, LD50 median lethal dose. Lipinski (rule-of-five) and Abbot bioavailability score are the assessment for the oral drug-candidates.
## Discussion
For the structural determination of buddlenols D 1 and 2, adduct ions [M+Cl]− at m/z of 679.2166 and 679.2156 were observed by HR-ESI–MS, and calculated for their molecular formula, C33H40O13 of 1 and 2, respectively. From Heteronuclear Multiple Bond Correlation (HMBC) spectra, hydroxy protons at δH 7.05 and 7.15 of 1 were placed on C-4″ (δC 135.9, syringylglycerol subunit) and C-4′ (δC 136.3, a syringyl part of syringaresinol subunit), respectively. Additionally, a C8″-O-C4 ether linkage of both substructures was suggested by carbon chemical shifts at C-8″ (δC 88.0) and C-4 (δC 135.9)18. From NMR analysis, HMBC correlations [δH 7.06 (OH-4″)/C-4″ (δC 136.1), δH 7.15 (OH-4′)/C-4′ (δC 136.3)], and 13C-NMR signals of C-8″ (δC 89.4) and C-4 (δC 136.2) were also observed in 2. Relative configurations on C-8″ and C-7″ of 1 and 2 were guided by NOESY correlations. The Nuclear Overhauser Effect Spectroscopy (NOESY) spectrum of 1 showed a prominent correlation of δH 4.18 (H-8″) and δH 4.98 (H-7″) and a relatively weak correlation of δH 4.18 (H-8″)/δH 4.38 (OH-7″), while a strong correlation of δH 4.00 (H-8″) and δH 4.34 (OH-7″) and a relatively pale correlation of δH 4.00 (H-8″)/δH 4.96 (H-7″) were observed in NOESY spectrum of 2. Moreover, J7′′,8′′ of 1 and 2 were observed at 4 and 6.8 Hz, respectively, and corresponded with erythro- and threo- forms of an acyl glycerol moiety18,19. Thus, 7′′,8′′-erythro and 7′′,8′′-threo buddlenols D were deduced for 1 and 2, respectively, and supported chemical shift differences for 7″ and 8′′ positions (ΔδC8′′-C7′′) as suggested by Xiong et al.19; [ΔδC8′′-C7′′: 14.3 and 15.3 for erythro and threo isomers, respectively]. From circular dichroism (CD) spectra of 1 and 2, positive Cotton effects were observed at 239 and 233 nm, respectively, indicating the 8″S configuration of both compounds19. Additionally, negative benzene 1Lb-band Cotton effects observed at 283 and 282 nm for 1 and 2, respectively, were consistent with 4 [Δε − 3.75, 281.9 nm] and revealed 7R,7′R,8S,8′S configurations in these molecules20. The absolute configuration of benzylic carbon (C-7″) bearing a hydroxy group in 1 and 2 was suggested by Cotton effects in Rh2(OCOCF3)4-induced CD experiments, based on the secondary alcohol bulkiness rule. E band signs (≈ 350 nm) were observed as negative and positive Δε for 1 and 2, respectively, suggesting the 7″R configuration for 1 and 7″S configuration for 221. Moreover, the assigned configurations of 1 (7′′R,8′′S-buddlenol D) and 2 (7′′S,8′′S-buddlenol D) were consistent with the electronic circular dichroism (ECD) spectrum calculated based on Density Functional Theory (DFT) (Supplementary Information: Figure S2–S42).
Buddlenol D was first isolated from *Buddleja davidii* (Scrophulariaceae)22. Later, the absolute configuration of two buddlenol D isomers (7′′R,8′′S and 7′′S,8′′S) were thoroughly investigated in work by Xiong et al.19. Both 7′′R,8′′S- and 7′′S,8′′S-buddlenols D were isolated from *Sinocalamus affinis* (Poaceae) and *Melodinus cochinchinesis* (Apocynaceae) using reverse phase chromatography19,23. In our study these 7″,8″-buddlenol D epimers were separated by silica gel column chromatography eluted in $20\%$ Me2CO/CH2Cl2. Buddlenol D and syringaresinol are furofuran lignans comprised of synapyl alcohol units, while hedyotisol A comes from coniferyl alcohol monomers. In our study, phenolic compounds (1–4, 8, 9, and 11) which possessed substructural guaiacyl and syringyl moieties, oligostilbenes [5, 6], and a proanthocyanidin [7] demonstrated the influence of shikimate pathway on the biosynthesis of compounds stored in C. brachiata stem and bark. Although 7″,8″-buddlenol D epimers were isolated from the aforementioned plants, their biological activity and medicinal use were not previously addressed.
We also reported the anti-inflammatory activity of 7″,8″-buddlenol D epimers. The inhibition of PGE2, a pyrogenic mediator, via suppressed COX2 expression by 1 and 2 supported the antipyretic properties of C. brachiata stem and bark for traditional use, and was putatively related to the finding that a higher core temperature correlated with higher PGE2 production24. Additionally, 1 and 2 significantly inhibited iNOS expression induced by LPS treatment. As LPS-induced iNOS and COX2 expression, these results suggested 1 and 2 exerted anti-inflammatory activities by suppressing these proinflammatory enzymes and reducing NO and PGE2 levels. MAPK signal transduction, including ERK$\frac{1}{2}$, JNK, and p38 MAPK, modulates inflammatory processes in response to stimuli via different receptors25. ERK$\frac{1}{2}$, JNK, and p38 MAPK phosphorylation activities are the major downstream signaling phenomena after TLR4 activation. Therefore, molecules regulating TLR4-mediated ERK$\frac{1}{2}$, JNK, and p38 MAPK phosphorylation may exhibit anti-inflammatory potential. From our data, we hypothesized the underlying mechanisms of 1 and 2 may occurred via modulated p38 MAPK phosphorylation. In contrast, dexamethasone, which is an anti-inflammatory drug and used as a positive control inhibited all ERK$\frac{1}{2}$, JNK, and p38MAPK phosphorylation26 and exerted potent anti-inflammatory activity by suppressing NO and PGE2 production and down-regulating iNOS and COX2 expression27.
To examine if 1 and 2 inhibited p38 MAPK phosphorylation, a specific p38 inhibitor SB203580 was used to identify anti-inflammatory activity. In RAW264.7 cells pre-treated with SB203580 prior to LPS stimulation, p38 MAPK phosphorylation suppression was clearly observed and resulted in NO inhibition (Supplementary Information: Figure S6)28,29. The Western blot experiments showed that the compounds 1 and 2 inhibited p38 MAP kinase up to 24 h after LPS stimulation, suggesting that their inhibitory effects on proinflammatory signaling induced by LPS could last long. This last long effect was also observed in pre-treatment with a specific p38 inhibitor, SB203580. Although the activation of NF-κB or AP-1 was not investigated in this study, the decrease in expression of downstream inflammatory effectors, iNOS and COX2, could help explain that these suppressions were influenced by p38 kinase inhibition30 (Figs. 3, 4). Thus, aided by SB203580 inhibitory observations, we suggest that p38 phosphorylation inhibition by 1 and 2 may regulate anti-inflammatory processes31. Additionally, the anti-inflammatory and antioxidant (DPPH radical scavenging) activities of 1 and 2 support the use of C. brachiata extracts for oral stomatitis and wound healing32 (Supplementary Information: Table S1).
Next, we used in silico methods to investigate the anti-inflammatory mechanisms of 1 and 2 at the atomic level. Our binding affinity model data suggested 1 and 2, as expected, bound TLR4 and p38-α MAPK; however, they both demonstrated a 0.8-fold lower binding capability when compared with reference ligands [ZINC25778142 and SB203580]. This may have been due to reduced ligand intermolecular interactions in the binding sites of targeted proteins. Based on our docking evidence, we hypothesize that 1 and 2 bound to both TLR-4 and p38-α MAPK at the ATP-binding site. However, immunoblot analyses suggested 1 and 2 suppressed a p38 MAPK downstream element while suppression of TLR4-mediated ERK$\frac{1}{2}$ and JNK was not observed. This implied that 1 and 2 may not bind effectively to TLR-4 since they did not inhibit the phosphorylation of all downstream elements8. Also, they did recognize smaller numbers of non-covalent interactions (Fig. 5B) when compared with the TLR-4 inhibitor reference (ZINC25778142). Thus, p38 MAPK may be a key player in compound binding, interrupting proinflammatory responses, and inhibiting the target as supported by computational and experimental studies. In addition to our binding analysis of p38-α MAPK, SB203580 exhibited higher intermolecular interactions when compared with 1 and 2 (Fig. 5A), which included hydrogen bonding with M109 and hydrophobic interactions with V30, Y35, V38, A51, T106, F169, and L171. Also, two halogen interactions with L104 and V105 were identified. All residues interacting with 1 and 2 were identified as contact residues in a previous report33, supporting the notion that 1 and 2 docked at preferable sites and had the potential to inhibit p38-α MAPK. In addition to TLR4, ZINC25778142 showed more intermolecular interactions when compared with 1 and 2 (Fig. 5B). These included five hydrogen bonds with D209, S211, L212, D234, and N236, hydrophobic interactions with A232, and W256, and two additional halogen bonds with L208 and E230. We also observed similar ZINC25778142 interactions with previously reported amino acid residues (D209, S211, and D234)34,35, thus verifying our docking protocol.
When combined, we suggest 1 and 2 bind to the ATP-binding pocket of p38-α MAPK and inhibit p38 phosphorylation, a downstream element of signaling pathway, rather than upstream TLR-4 inhibition. These results agree with the key roles for p38-α MAPK in regulating inflammatory responses, including iNOS and COX2 regulation, in macrophages36. However, future in-depth studies on 1 and 2 during LPS-TLR-4/MD-2 transduction are warranted.
Concerning the sharing features of ATP-binding pocket among kinases, we thus observed the binding capability toward other kinases involved in the focused signaling cascade. As listed in Table 2, the binding affinity based on the fitness score suggested that 1 and 2 were most favorable to bind to p38-α MAPK at the ATP-binding site. The binding affinity ratio for p38-α MAPK at the ATP binding site compared to corresponding crystallized ligand was 0.81 and 0.76 for 1 and 2, respectively, while the other kinases were falling into the smaller values (~ 0.4 to 0.6). Hence, molecular docking studies suggested that 1 and 2 could selectively impede p38-α MAPK at the ATP-binding site.
Although, 1 and 2 might possess low GI absorption profile, based on in silico calculation 1 and 2 exhibited good metabolic profiles. Compounds 1 and 2 could be potential compounds for further structural development for better pharmacokinetic profiles and suitable drug delivery system. Moreover, after oral administration of 1 and 2, they might lose the syringyl unit during the digestion process and could be transformed to enterodiol derivatives by gut microbiome37. There was evidence that enterodiol exhibited p38 down regulation as one of the mechanisms inhibiting the growth of colorectal cancer cells38.
In our previous work, we reported a natural p38 MAPK inhibitor (+)-S-deoxydihydroglyparvin which is a sulfur-containing propanamide derivative isolated from *Glycosmis parva* (Rutaceae) leaves39. As furofuran lignans such as epimagnolin B from *Magnolia fargesii* (Magnoliaceae), (−)-sesamin-2,2′-diol from *Isodon japonicus* (Lamiaceae), and zanthpodocarpins A and B from *Zanthoxylum podocarpum* (Rutaceae) displayed anti-inflammatory activity via NO inhibition in cultured cells40–42, we suggest this compound class may exhibit anti-inflammatory activity. From our study, we propose 7″,8″-buddlenol D epimers function as p38 MAPK inhibitors.
## Plant materials
Stems and barks from C. brachiata were collected in January 2018 from the botanical gardens of the Faculty of Pharmaceutical Sciences, Chulalongkorn University, Bangkok. The samples were identified by a study author (C. Chansriniyom). The herbarium specimen (CC-CB-010118) was deposited at the Department of Pharmacognosy and Pharmaceutical Botany, Chulalongkorn University. In addition, the handling and notification of stems and barks of C. brachiata were carried out in accordance with the Plant Variety Protection Act B.E. 2542 [1999]: Section 53, the Kingdom of Thailand.
## Extraction and isolation
Dried and coarsely powdered C. brachiata stems and barks (4.0 kg) were successively macerated in MeOH (3 × 15 L) to yield a 203.80 g MeOH extract which was partitioned with hexane to generate a hexane extract (7.34 g). The remaining organic layer was added to water and further partitioned with EtOAc and n-butanol (n-BuOH) to obtain EtOAc (27.39 g), n-BuOH (50.39 g), and aqueous (113.89 g) extracts.
The EtOAc extract (27.39 g) underwent silica gel column chromatography (Flash column chromatography (Flash CC), PuriFlash® XS 420, Advion Inc., NY, USA) and was eluted in a gradient solvent of MeOH, EtOAc and hexane at an initial ratio of 0:40:60 (respectively), then progressed to 30:70:0 (respectively), and terminated in $100\%$ MeOH. Based on chromatographic patterns fractions were combined to generate seven fractions (CBE1–CBE7).
CBE7 (6.58 g) was separated into seven sub-fractions using Flash CC [Si (Silica gel 70–230 or 230–400 mesh, Merck, Darmstadt, Germany), a mobile phase of MeOH-CH2Cl2 system ($8\%$, $12\%$, $20\%$ MeOH/CH2Cl2, gradient manner)]. CBE7-3 (874 mg) was separated into 48 fractions using Medium Pressure Liquid Chromatography (MPLC, Eyela® medium pressure pump VSP 3050, Kyoto, Japan) (Si, $4\%$MeOH/CH2Cl2). CBE7-3-12 (178.2 mg) was purified using MPLC (Si, $20\%$Me2CO/CH2Cl2) to obtain 1 (6 mg) and 2 (2.4 mg). CBE7-3-20 (88.9 mg) was separated by MPLC (Si, $20\%$Me2CO/CH2Cl2) to afford 3 (6 mg). CBE7-4 (466 mg) was purified using two MPLC steps (Si, $10\%$MeOH/ MeOH/CH2Cl2 and $6\%$MeOH/CH2Cl2) to achieve 8 (6.3 mg).
CBE6 (2.23 g) was separated by MPLC (Si, $80\%$EtOAc/hexane) into seven sub-fractions. CBE6-2 (453.8 mg) underwent MPLC (Si, $50\%$Me2CO/hexane) to generate 12 sub-fractions. Sub-fractions 5–6 (CBE6-2-(5–6), 110 mg) were combined and underwent two purification steps on an C18-reverse phase silica gel (RP-18, LiChroprep®, 25–40 μm, Merck, Darmstadt, Germany) column using $40\%$MeCN(acetonitrile)/water and Si, $4\%$MeOH/CH2Cl2 to obtain 18 mg of 4. CBE6-2-10 (167 mg) was separated by MPLC (RP-18, $30\%$MeCN/water) to yield 7 (73.5 mg).
CBE5 (830 mg) was purified by MPLC (Si, EtOAc/CH2Cl2, gradient) to give 18 sub-fractions. CBE5-12 (36.6 mg) was separated by MPLC (Si, $2\%$MeOH/CH2Cl2) into 10 sub-fractions. Sub-fractions 8–10 (3.2 mg) were combined and subjected to Sephadex LH-20 column chromatography ($50\%$MeOH/CH2Cl2) to provide 9 (2.0 mg). CBE5-13 (41.3 mg) was purified by MPLC (Si, $4\%$MeOH/CH2Cl2) to obtain 5 (4.1 mg). CBE5-14 (32.7 mg) was separated by MPLC (Si, $4\%$MeOH/CH2Cl2) to yield 10 (1.0 mg). CBE5-15 (44.0 mg) was purified by MPLC (RP-18, $30\%$ MeCN/water) to generate CBE5-15-3 (10 mg) which was purified on a Sephadex™ LH-20 (GE Healthcare, Amersham, UK) column (MeOH) to yield 6 (6.8 mg).
CBE3-7 (222.9 mg), a lower polar fraction, was separated by MPLC (Si, EtOAc/CH2Cl2/hexane, 1:1:3) to yield 11 (3.6 mg).
## General procedures for structure determination
Ultraviolet (UV) spectra were recorded on a Jasco V-530 UV/VIS spectrophotometer (Jasco, Kyoto, Japan). Infrared spectra were recorded on a FTIR-4200 spectrophotometer (Attenuated total reflectance) (Jasco, Kyoto, Japan). Optical rotation data were recorded on a Jasco P-2000 polarimeter (Jasco, Kyoto, Japan). One- and two-dimensional nuclear magnetic resonance (1D- and 2D-NMR) spectra were obtained from Varian Unity AS400 (Varian, Inc. Vacuum Technologies, MA, USA), Bruker Ascend 400 NMR (Bruker, Karlsruhe, Germany), Bruker Advance NEO 400 MHz, or JEOL 500 NMR (JEOL USA, Inc. MA, USA) instruments. Electron spray ionization-mass spectrometry (ESI-MS) data were recorded on a VG-Biotech Quatro-5022 mass spectrometer (VG Biotech, Altrincham, UK). High resolution-electron spray ionization-mass spectrometry (HR-ESI–MS) data were recorded on an Agilent 6540 UHD Accurate-Mass Q-TOF mass spectrometer (Agilent Technologies, CA, USA). Circular dichroism (CD) experiments were performed using a Jasco J-815 circular dichroism spectrophotometer (Jasco, Kyoto, Japan). Rh2 (OCOCF3)4-induced CD studies were conducted in CH2Cl2 at a molar ratio of 1:0.4 compound to ligand43.
The spectroscopic data of 1 and 2 as followed: (−)-(7R,7′R,7′′R,8S,8′S,8′′S)-4′,4′′-dihydroxy-3,3′,3′′,5,5′,5′′-hexamethoxy-7,9′:7′,9-diepoxy-4,8′′-oxy-8,8′-sesquineolignan-7′′,9′′-diol (7″R,8″S-erythro buddlenol D, 1); white solid. 1H-NMR (acetone-d6, 400 MHz) δ: 7.15 (1H, s, OH-4′), 7.05 (1H, s, OH-4″), 6.78 (2H, s, H-2, -6), 6.71 (2H, s, H-2″, -6″), 6.68 (2H, s, H-2′, -6′), 4.98 (1H, q, $J = 4$ Hz, H-7″), 4.74 (1H, d, $J = 4.4$ Hz, H-7), 4.68 (1H, d, $J = 4.4$ Hz, H-7′), 4.38 (1H, d, $J = 4.4$ Hz, OH-7″), 4.26 (2H, m, H-9a, 9′a), 4.18 (1H, m, H-8″), 3.88 (6H, s, OCH3-3, 5), 3.86 (2H, m, H-9b, -9′b), 3.84 (1H, m, H-9″a), 3.82 (6H, s, OCH3-3′, 5′), 3.80 (6H, s, OCH3-3″, 5″), 3.43 (1H, m, H-9″b), 3.12 (2H, m, H-8, -8′); 13C-NMR (acetone-d6, 100 MHz) δ: 154.3 (C-3, -5), 148.8 (C-3′, -5′), 148.5 (C-3″, -5″), 139.2 (C-1), 136.3 (C-4′), 135.9 (C-4, -4″), 133.2 (C-1′), 132.8 (C-1″), 104.9 (C-2″, -6″), 104.6 (C-2′, -6′), 104.2 (C-2, -6), 88.0 (C-8″), 86.8 (C-7′), 86.7 (C-7), 73.7 (C-7″), 72.6 (C-9, -9′), 61.1 (C-9″), 56.7 (OCH3-3, 5, 3′, 5′,3″, 5″), 55.4 (C-8, -8′); IR (ATR) cm-1: 3434 (broad), 2936, 2874, 1612, 1593, 1516, 1461, 1214, 1113; UV λmax (MeOH) nm (log ε): 206 (4.58), 238 (3.82), 272 (2.99). ESI–MS m/z: 667 [M+Na]+. HR-ESI–MS m/z: 679.2166 [M+Cl]−, calcd. for C33H40O13Cl 679.2163. [ α]20D: − 12.96° (c 1.25 × 10–4 g/mL, MeOH). CD nm (MeCN): 204.8 (Δε − 13.88), 208.7 (Δε − 1.50), 213.1 (Δε − 5.35), 217.9 (Δε + 3.23), 225.7 (Δε − 4.14), 234.3 (Δε + 2.06), 239.0 (Δε + 0.50), 244.6 (Δε − 0.73), 283.3 (Δε − 0.09). Rh2(OCOCF3)4-induced CD nm (CH2Cl2): 355.4 (Δε − 0.83) (Supplementary Information: Figure S7–S17).
(−)-(7R,7′R,7′′S,8S,8′S,8′′S)-4′,4′′-Dihydroxy-3,3′,3′′,5,5′,5′′-hexamethoxy-7,9′:7′,9-diepoxy-4,8′′-oxy-8,8′-sesquineolignan-7′′,9′′-diol (7″S,8″S-threo buddlenol D, 2): white solid. 1H-NMR (acetone-d6, 400 MHz) δ: 7.15 (1H, s, OH-4′), 7.06 (1H, s, OH-4″), 6.76 (2H, s, H-2, -6), 6.74 (2H, s, H-2″, -6″), 6.68 (2H, s, H-2′, -6′), 4.96 (1H, dd, $J = 6.8$, 3.6 Hz, H-7″), 4.73 (1H, d, $J = 4.0$ Hz, H-7), 4.68 (1H, d, $J = 4.0$ Hz, H-7′), 4.34 (1H, d, $J = 3.2$ Hz, OH-7″), 4.25 (2H, m, H-9a, 9′a), 4.00 (1H, m, H-8″), 3.90 (6H, s, OCH3-3, 5), 3.87 (2H, m, H-9b, -9′b), 3.82 (6H, s, OCH3-3′, 5′), 3.79 (6H, s, OCH3-3″, 5″), 3.67 (1H, m, H-9″a), 3.36 (1H, m, H-9″b), 3.11 (2H, m, H-8, -8′); 13C-NMR (acetone-d6, 100 MHz) δ: 153.9 (C-3, -5), 148.8 (C-3′, -5′), 148.4 (C-3″, -5″), 139.3 (C-1), 136.3 (C-4′), 136.2 (C-4), 136.1 (C-4″), 133.2 (C-1′), 132.8 (C-1″), 105.4 (C-2″, -6″), 104.5 (C-2′, -6′), 104.0 (C-2, -6), 89.4 (C-8″), 86.8 (C-7′), 86.6 (C-7), 74.1 (C-7″), 72.6 (C-9, -9′), 61.6 (C-9″), 56.7 (OCH3-3, 5, 3′, 5′,3″, 5″), 55.4 (C-8, -8′); IR (ATR) cm−1: 3466 (broad), 2939, 2844, 1610, 1594, 1520, 1459, 1216, 1116; UV λmax (MeOH) nm (log ε): 206 (4.64), 236 sh (3.93), 272 (3.24). ESI–MS m/z: 667 [M+Na]+. HR-ESI–MS m/z: 679.2156 [M+Cl]−, calcd. for C33H40O13Cl 679.2163. [ α]20D: − 9.78° (c 4.6 × 10–5 g/mL, MeOH). CD nm (MeCN): 203.9 (Δε − 12.58), 207.7 (Δε + 3.04), 210.1 (Δε + 1.19), 214.8 (Δε + 6.30), 219.0 (Δε − 0.24), 221.4 (Δε + 0.42), 225.5 (Δε − 0.98), 233.1 (Δε + 0.61), 237.8 (Δε − 0.62), 282.2 (Δε − 0.04). Rh2(OCOCF3)4-induced CD nm (CH2Cl2): 347.2 (Δε + 0.81) (Supplementary Information: Figure S18–S28).
For compounds 3–11, spectroscopic data and references were available (Supplementary Figure S29–S58).
## Cell culture and treatment
Mouse macrophage cells RAW264.7 (Catalog no. TIB-71, ATCC, Manassas VA, USA) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with $10\%$ heat-inactivated fetal bovine serum in $5\%$ CO2 at 37ºC. Cells were routinely passaged every 3 days. All compounds were dissolved in DMSO and diluted in DMEM (final DMSO concentration = $0.5\%$).
## NO production
Cells (5 × 105 cells/mL) in 100 µL were seeded in a 96‑well plate and incubated for 24 h at 37 °C. After this, cells were treated with different compound concentrations and incubated for 24 h. Then, medium containing compound was replaced by phenol red-free medium containing LPS (Sigma-Aldrich, MO, USA) (final concentration = 500 µg/mL) and the cells were further incubated for 24 h. NO levels were determined by collecting 100 µL supernatants and mixing with 100 µL modified Griess reagent (Sigma-Aldrich). Reaction mixtures were incubated for 15 min at room temperature and absorbance measured at 560 nm using a microplate reader. NO levels were quantified by interpolation using a nitrite standard curve. Cells treated with dexamethasone (10 µM) were used as a positive control group.
## Prostaglandin E2 production
Cells (5 × 105 cells/mL) in 100 µL were seeded in a 96‑well plate and incubated for 24 h at 37 °C. Then, cells were treated with different compound concentrations and incubated for 24 h. After that, medium containing compound was replaced by medium containing LPS (final concentration = 500 µg/mL) and the cells were incubated for another 24 h. Supernatants were collected and subjected to enzyme-linked immunosorbent assay for PGE2 (Enzo Life Sciences, NY, USA) following manufacturer’s instructions.
## Western blotting
Protein expression was determined in cell lysates from each treatment group at the same treatment manner with previous experiments. A 30-µg protein aliquot underwent sodium dodecyl sulfate–polyacrylamide gel electrophoresis and separated proteins were transferred to polyvinylidene fluoride membranes. The blots were blocked in $5\%$ bovine serum albumin and probed with primary antibodies: anti-iNOS (ab15323), anti-COX2 (CST#2282), anti-phospho-ERK (CST#4376), anti-ERK (CST#4695), anti-phospho-JNK (CST#4668), anti-JNK (CST#9252), anti-phospho-p38 (CST#4511), anti-p38 (CST#9212), and followed by incubation with a horse radish peroxidase-conjugated secondary antibody. Protein bands were detected using a chemiluminescent substrate and images acquired using ImageQuant LAS 4000 (GE Healthcare, Buckinghamshire, UK). Relative protein band intensity was normalized to β-actin or GAPDH expression using ImageJ. SB203580 (Tokyo Chemical Industry, Japan) is a specific p38-α inhibitor and was used as a positive control for p38 phosphorylation inhibition.
## Molecular docking
Protein–ligand docking was conducted for eight targeted proteins under nine systems-TLR4, p38-α MAPK (ATP-binding site), p38-α MAPK (non-ATP-binding site), ERK1, ERK2, JNK1, JNK2, JNK3 and IκB kinase β using crystallized structures (retrieved from RCSB Protein Data Bank), listed as follows: 2Z6544, 3ZSH33, 1KV245, 4QTB46, 1PME47, 2NO348, 3NPC49, 4W4Y50, 4KIK51, respectively. In addition to protein preparation, the 3-dimensional structures of 1 and 2 were manually constructed using the Gaussian 09 program and optimized using the HF/6-31d basis set implemented in the program52. For docking studies, the Gold docking program was used, and protocols were as follows: [1] the ligand binding site was defined as 6 Å for sphere docking and [2] ChemScore was used for the scoring function. Binding between proteins and compounds was visualized in Accelrys Discovery Studio 2.5 (Accelrys Inc.).
For all systems, docking was validated using aforementioned protocols to redock the corresponding crystallized ligand (Table S2), followed by aligning redocked pose with its original conformation to compare its similarity termed root-mean-squared-deviation (RMSD) of structure coordinates. For TLR4, since no crystallized small molecule inhibitors targeting the TLR-MD2 protein–protein interface were available, docking was performed by rational validation. Briefly, we manually docked a previously studied compound (ZINC25778142)34 and the docked pose was then observed its orientation and intermolecular interactions with the key reported residues including D209, S211, and D234. Our docking protocols were validated for all systems (Supplementary Information: Figure S61).
## ADMET prediction
Pharmacokinetic properties (Absorption, Distribution, Metabolism and Excretion) of compounds 1 and 2 were calculated using SwissADME platform (http://www.swissadme.ch/)53 while their toxicity profiles were predicted using Protox-II platform (https://tox-new.charite.de/protox_II/)54 and the DataWarrior software package55.
## Statistical analysis
All data were expressed as the mean ± standard error of the mean from at least three independent experiments. Differences among groups were evaluated using one-way analysis of variance, followed by post-hoc tests with Bonferroni correction. Statistical significance was accepted at $p \leq 0.05.$
## Conclusions
For the first time, we characterized the anti-inflammatory activities of 7″,8″-buddlenol D epimers from C. brachiata stem and bark. Their action mechanisms were mediated by inhibited p38 MAPK phosphorylation which attenuated iNOS and COX2 expression and inhibited NO and PGE2 production in LPS-induced RAW264.7 macrophages. These findings support the use of C. brachiata for antipyretic, oral stomatitis, and wound healing purposes. Therefore, 7″,8″-buddlenol D epimers can be used as biological markers for C. brachiata extract.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30475-5.
## References
1. Newton K, Dixit VM. **Signaling in innate immunity and inflammation**. *Cold Spring Harbor Perspect. Biol.* (2012.0) **4** a006049. DOI: 10.1101/cshperspect.a006049
2. Chen L. **Inflammatory responses and inflammation-associated diseases in organs**. *Oncotarget* (2017.0) **9** 7204-7218. DOI: 10.18632/oncotarget.23208
3. Coleman JW. **Nitric oxide in immunity and inflammation**. *Int. Immunopharmacol.* (2001.0) **1** 1397-1406. DOI: 10.1016/S1567-5769(01)00086-8
4. Sharma JN, Al-Omran A, Parvathy SS. **Role of nitric oxide in inflammatory diseases**. *Inflammopharmacology* (2007.0) **15** 252-259. DOI: 10.1007/s10787-007-0013-x
5. Peri F, Calabrese V. **Toll-like Receptor 4 (TLR4) modulation by synthetic and natural compounds: An update**. *J. Med. Chem.* (2014.0) **57** 3612-3622. DOI: 10.1021/jm401006s
6. Nedeva C, Menassa J, Puthalakath H. **Sepsis: Inflammation is a necessary evil**. *Front. Cell Dev. Biol.* (2019.0). DOI: 10.3389/fcell.2019.00108
7. Youn HS, Saitoh SI, Miyake K, Hwang DH. **Inhibition of homodimerization of Toll-like receptor 4 by curcumin**. *Biochem. Pharmacol.* (2006.0) **72** 62-69. DOI: 10.1016/j.bcp.2006.03.022
8. Wang Y. **Curcumin analog L48H37 prevents lipopolysaccharide-induced TLR4 signaling pathway activation and sepsis via targeting MD2**. *J. Pharmacol. Exp. Therap.* (2015.0) **353** 539-550. DOI: 10.1124/jpet.115.222570
9. Lee JK, Kim NJ. **Recent advances in the inhibition of p38 MAPK as a potential strategy for the treatment of Alzheimer's disease**. *Molecules* (2017.0) **22** 1287. DOI: 10.3390/molecules22081287
10. Arya V, Arya ML. **A review on anti-inflammatory plant barks**. *Int. J. PharmTech Res.* (2011.0) **3** 899-908
11. Krishnaveni B, Neeharika V, Venkatesh S, Padmavathy R, Reddy B. **Wound healing activity of**. *Indian J. Pharm. Sci.* (2009.0) **71** 576-578. DOI: 10.4103/0250-474X.58184
12. Junejo JA, Rudrapal M, Zaman MK. **Antidiabetic activity of**. *Indian J. Nat. Prod. Resourc.* (2020.0) **11** 18-29
13. Islam MA, Hossain MS, Azad M, Rashid MH-O, Mofizur M. **In vivo evaluation of analgesic, antiinflammatory and antidiabetic activities of methanol extract of**. *In Vivo* (2020.0) **1** 38-46
14. Fitzgerald J. **(+)-Hygroline, the major alkaloid of**. *Aust. J. Chem.* (1965.0) **18** 589-590. DOI: 10.1071/CH9650589
15. Yadav JS, Narasimhulu G, Mallikarjuna Reddy N, Subba Reddy BV. **Total synthesis of (+)-pseudohygroline**. *Tetrahedron Lett.* (2010.0) **51** 1574-1577. DOI: 10.1016/j.tetlet.2010.01.060
16. Ling S. **A new diglycosyl megastigmane from**. *Fitoterapia* (2004.0) **75** 785-788. DOI: 10.1016/j.fitote.2004.09.019
17. Phuwapraisirisan P, Sowanthip P, Miles D, Tip-pyang S. **Reactive radical scavenging and xanthine oxidase inhibition of proanthocyanidins from**. *Phytother. Res.* (2006.0) **20** 458-461. DOI: 10.1002/ptr.1877
18. Gan M. **Glycosides from the root of**. *J. Nat. Prod.* (2008.0) **71** 647-654. DOI: 10.1021/np7007329
19. Xiong L. **Lignans and neolignans from**. *J. Nat. Prod.* (2011.0) **74** 1188-1200. DOI: 10.1021/np200117y
20. Snatzke G, Kajitar M, Werner-Zamojska F. **Circular dichroism—XLVII: Influence of substitution pattern on the benzene**. *Tetrahedron* (1972.0) **28** 281-288. DOI: 10.1016/0040-4020(72)80134-0
21. Gerards M, Snatzke G. **Circular dichroism, XCIII determination of the absolute configuration of alcohols, olefins, epoxides, and ethers from the CD of their “in situ” complexes with [Rh**. *Tetrahedron Asymmetry* (1990.0) **1** 221-236. DOI: 10.1016/S0957-4166(00)86328-4
22. Houghton J. **P. Lignans and neolignans from**. *Phytochemistry* (1985.0) **24** 819-826. DOI: 10.1016/S0031-9422(00)84901-8
23. Yang L, He J-J, Cui X-Y, Liu Y-P, Wang B. **Chemical constituents from**. *Biochem. Syst. Ecol.* (2021.0) **95** 104245. DOI: 10.1016/j.bse.2021.104245
24. Esh CJ. **The influence of environmental and core temperature on cyclooxygenase and PGE2 in healthy humans**. *Sci. Rep.* (2021.0) **11** 6531. DOI: 10.1038/s41598-021-84563-5
25. Hommes DW, Peppelenbosch MP, van Deventer SJH. **Mitogen activated protein (MAP) kinase signal transduction pathways and novel anti-inflammatory targets**. *Gut* (2003.0) **52** 144-151. DOI: 10.1136/gut.52.1.144
26. Zhu QY, Liu Q, Chen JX, Lan K, Ge BX. **MicroRNA-101 targets MAPK phosphatase-1 to regulate the activation of MAPKs in macrophages**. *J. Immunol.* (2010.0) **185** 7435-7442. DOI: 10.4049/jimmunol.1000798
27. Al-Harbi NO. **Dexamethasone attenuates LPS-induced acute lung injury through inhibition of NF-κB, COX-2, and pro-inflammatory mediators**. *Immunol. Invest.* (2016.0) **45** 349-369. DOI: 10.3109/08820139.2016.1157814
28. Ajizian SJ, English BK, Meals EA. **Specific inhibitors of p38 and extracellular signal regulated kinase mitogen-activated protein kinase pathways block inducible nitric oxide synthase and tumor necrosis factor accumulation in murine macrophages stimulated with lipopolysaccharide and interferon-gamma**. *J. Infect. Dis.* (1999.0) **179** 939-944. DOI: 10.1086/314659
29. Haddad EB. **Role of p38 MAP kinase in LPS-induced airway inflammation in the rat**. *Brit. J. Pharmacol.* (2001.0) **132** 1715-1724. DOI: 10.1038/sj.bjp.0704022
30. Kwon D-J, Ju SM, Youn GS, Choi SY, Park J. **Suppression of iNOS and COX-2 expression by flavokawain A via blockade of NF-κB and AP-1 activation in RAW 264.7 macrophages**. *Food Chem. Toxicol.* (2013.0) **58** 479-486. DOI: 10.1016/j.fct.2013.05.031
31. Chen C, Chen YH, Lin WW. **Involvement of p38 mitogen-activated protein kinase in lipopolysaccharide-induced iNOS and COX-2 expression in J774 macrophages**. *Immunology* (1999.0) **97** 124-129. DOI: 10.1046/j.1365-2567.1999.00747.x
32. Comino-Sanz IM, López-Franco MD, Castro B, Pancorbo-Hidalgo PL. **The role of antioxidants on wound healing: A review of the current evidence**. *J. Clin. Med.* (2021.0) **10** 3558. DOI: 10.3390/jcm10163558
33. Azevedo R. **X-ray structure of p38α bound to TAK-715: Comparison with three classic inhibitors**. *Acta Crystallogr. Sect. D Biol. Crystallogr.* (2012.0) **68** 1041-1050. DOI: 10.1107/s090744491201997x
34. Švajger U. **Novel toll-like receptor 4 (TLR4) antagonists identified by structure- and ligand-based virtual screening**. *Eur. J. Med. Chem.* (2013.0) **70** 393-399. DOI: 10.1016/j.ejmech.2013.10.019
35. Han J. **Structure-based rational design of a Toll-like receptor 4 (TLR4) decoy receptor with high binding affinity for a target protein**. *PLoS One* (2012.0) **7** e30929. DOI: 10.1371/journal.pone.0030929
36. Ono K, Han J. **The p38 signal transduction pathway: activation and function**. *Cell. Signal.* (2000.0) **12** 1-13. DOI: 10.1016/s0898-6568(99)00071-6
37. Senizza A. **Lignans and gut microbiota: An interplay revealing potential health implications**. *Molecules* (2020.0) **25** 5709. DOI: 10.3390/molecules25235709
38. Shin MK, Jeon YD, Jin JS. **Apoptotic effect of enterodiol, the final metabolite of edible lignans, in colorectal cancer cells**. *J. Sci. Food Agric.* (2019.0) **99** 2411-2419. DOI: 10.1002/jsfa.9448
39. Laprasert C, Chansriniyom C, Limpanasithikul W. *J. Adv. Pharma. Technol. Res.* (2021.0) **12** 32-39. DOI: 10.4103/japtr.JAPTR_64_20
40. Kim JY. **In vitro anti-inflammatory activity of lignans isolated from**. *Bioorg. Med. Chem. Lett.* (2009.0) **19** 937-940. DOI: 10.1016/j.bmcl.2008.11.103
41. Hong SS. **A new furofuran lignan from**. *Arch. Pharmacal Res.* (2009.0) **32** 501-504. DOI: 10.1007/s12272-009-1404-x
42. Zhou X-J. **Two dimeric lignans with an unusual α, β-unsaturated ketone motif from**. *Bioorg. Med. Chem. Lett.* (2011.0) **21** 373-376. DOI: 10.1016/j.bmcl.2010.10.135
43. Frelek J, Szczepek WJ. **[Rh**. *Tetrahedron Asymmetry* (1999.0) **10** 1507-1520. DOI: 10.1016/S0957-4166(99)00115-9
44. Kim HM. **Crystal structure of the TLR4-MD-2 complex with bound endotoxin antagonist Eritoran**. *Cell* (2007.0) **130** 906-917. DOI: 10.1016/j.cell.2007.08.002
45. Pargellis C. **Inhibition of p38 MAP kinase by utilizing a novel allosteric binding site**. *Nat. Struct. Biol.* (2002.0) **9** 268-272. DOI: 10.1038/nsb770
46. Chaikuad A. **A unique inhibitor binding site in ERK1/2 is associated with slow binding kinetics**. *Nat. Chem. Biol.* (2014.0) **10** 853-860. DOI: 10.1038/nchembio.1629
47. Fox T. **A single amino acid substitution makes ERK2 susceptible to pyridinyl imidazole inhibitors of p38 MAP kinase**. *Protein Sci.* (1998.0) **7** 2249-2255. DOI: 10.1002/pro.5560071102
48. Liu M. **Discovery of a new class of 4-anilinopyrimidines as potent c-Jun N-terminal kinase inhibitors: Synthesis and SAR studies**. *Bioorg. Med. Chem. Lett.* (2007.0) **17** 668-672. DOI: 10.1016/j.bmcl.2006.10.093
49. Kuglstatter A. **X-ray crystal structure of JNK2 complexed with the p38alpha inhibitor BIRB796: Insights into the rational design of DFG-out binding MAP kinase inhibitors**. *Bioorg. Med. Chem. Lett.* (2010.0) **20** 5217-5220. DOI: 10.1016/j.bmcl.2010.06.157
50. Park H. **Structural basis and biological consequences for JNK2/3 isoform selective aminopyrazoles**. *Sci. Rep.* (2015.0) **5** 8047. DOI: 10.1038/srep08047
51. Liu S. **Crystal structure of a human IκB kinase β asymmetric dimer**. *J. Biol. Chem.* (2013.0) **288** 22758-22767. DOI: 10.1074/jbc.M113.482596
52. 52.Frisch, M. et al. Gaussian 09, Revision D. 01, Gaussian, Inc., Wallingford CT. See also: http://www.gaussian.com (2009).
53. Daina A, Michielin O, Zoete V. **SwissADME: A free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules**. *Sci. Rep.* (2017.0) **7** 42717. DOI: 10.1038/srep42717
54. Banerjee P, Eckert AO, Schrey AK, Preissner R. **ProTox-II: A webserver for the prediction of toxicity of chemicals**. *Nucleic Acids Res.* (2018.0) **46** W257-w263. DOI: 10.1093/nar/gky318
55. Sander T, Freyss J, von Korff M, Rufener C. **DataWarrior: An open-source program for chemistry aware data visualization and analysis**. *J. Chem. Inf. Model.* (2015.0) **55** 460-473. DOI: 10.1021/ci500588j
|
---
title: An anthropometric evidence against the use of age-based estimation of bodyweight
in pediatric patients admitted to intensive care units
authors:
- Nobuyuki Nosaka
- Tatsuhiko Anzai
- Ryo Uchimido
- Yuka Mishima
- Kunihiko Takahashi
- Kenji Wakabayashi
journal: Scientific Reports
year: 2023
pmcid: PMC9981604
doi: 10.1038/s41598-023-30566-3
license: CC BY 4.0
---
# An anthropometric evidence against the use of age-based estimation of bodyweight in pediatric patients admitted to intensive care units
## Abstract
Age-based bodyweight estimation is commonly used in pediatric settings, but pediatric ICU patients often have preexisting comorbidity and resulting failure to thrive, hence their anthropometric measures may be small-for-age. Accordingly, age-based methods could overestimate bodyweight in such settings, resulting in iatrogenic complications. We performed a retrospective cohort study using pediatric data (aged < 16 years) registered in the Japanese Intensive Care Patient Database from April 2015 to March 2020. All the anthropometric data were overlaid on the growth charts. The estimation accuracy of 4 age-based and 2 height-based bodyweight estimations was evaluated by the Bland–Altman plot analysis and the proportion of estimates within $10\%$ of the measured weight (ρ$10\%$). We analyzed 6616 records. The distributions of both bodyweight and height were drifted to the lower values throughout the childhood while the distribution of BMI was similar to the general healthy children. The accuracy in bodyweight estimation with age-based formulae was inferior to that with height-based methods. These data demonstrated that the pediatric patients in the Japanese ICU were proportionally small-for-age, suggesting a special risk of using the conventional age-based estimation but supporting the use of height-based estimation of the bodyweight in the pediatric ICU.
## Introduction
Anthropometric measurements (e.g. bodyweight, height, and head circumferences) are important to determine the dosage of medications and the equipment size for pediatric patients1. These anthropometric indices undergo dramatic changes as children grow during their entire childhood2, and numerous age-based and height-based methods have been proposed to guide appropriate medical interventions. These estimation methods are particularly important in pediatric emergency and intensive care settings, where immediate medical interventions are often required before measuring bodyweight on site, thus various age-based estimation formulas for estimating bodyweight3–7 have been proposed because precise age information is readily available in most cases.
However, it is notable that these age-based estimation formulae were developed based on general populations5–10, and pediatric patients admitted to intensive care units (ICUs) may not follow the anthropometric archetype of the general population because pediatric patients in the ICU often have preexisting comorbidity and resulting failure to thrive11–13. A couple of studies have provided anthropometric characteristics of the pediatric population admitted to ICUs. In a prospective British single-center study14, the pediatric population in the ICU had significantly lower weight-for-age compared to the general British children, with the increased proportion of extremely low weight-for-age ($18\%$ of the study population were less than − 2.5 SD below the UK reference population mean bodyweight). Ross et al.15, using a large retrospective analysis of prospectively collected data from multiple pediatric ICUs in the United States, also showed that pediatric ICU patients had lower weight-for-age compared to the general US population. From the perspective of medical safety, this evidence collectively implicated that the use of age-based estimation of anthropometric values may pose a risk to the pediatric ICU population because of drug dosage errors5. However, the performance comparison of different bodyweight estimation methods has not yet been well explored for the pediatric ICU population.
In this study, we aimed to characterize anthropometric data of the pediatric ICU population in Japan, and evaluate the validity of age-based bodyweight estimation methods for the pediatric ICU population. We hypothesized that the pediatric population in ICU is proportionally small-for-age, hence height-based bodyweight estimation should be used for the pediatric patients in the ICU.
## Methods
In this study, we aimed to investigate the distribution of anthropometric indices (bodyweight, height, and body mass index [BMI, identical to the Kaup index]) of Japanese children in ICU on the growth charts. We also aimed to evaluate the performance of established age-based bodyweight calculation tools compared with height-based estimation methods for the pediatric population in Japanese ICU.
## Study design and cohort
We performed a retrospective cohort study using the data derived from the Japanese Intensive Care Patient Database (JIPAD), a national intensive care unit registry in Japan16. We obtained the 5-year JIPAD data of consecutive patients aged less than 16 years who were admitted to ICU from April 2015 to March 2020. The database provides patient demographics and anthropometric data including bodyweight and height16. This study was reviewed and approved, and the need for informed consent was waived considering the retrospective design and complete anonymization, by Tokyo Medical and Dental University Review Board (M2020-245). All methods in our study were performed in accordance with the relevant guidelines and regulations.
## Data plotting on growth charts and standard deviation score calculation
All height and bodyweight data were plotted on the growth charts for Japanese children17 officially provided online by the Japanese Society for Pediatric Endocrinology (JSPE; http://jspe.umin.jp/medical/chart_dl.html, Accessed on April 2021). Percentile data of bodyweight and height for each age were calculated by using R software, version 4.1.2 (The R Foundation for Statistical Computing, Vienna, Austria).
To quantitatively compare the anthropometric indices of pediatric ICU patients with the above JSPE reference-standard, we used standard deviation scores (SDS) for bodyweight, height, and BMI as previously described15,18,19. The SDS for each anthropometric index was calculated using the Excel-based Clinical Tool for Growth Evaluation of Children provided by the JSPE (A general version can be downloaded at http://jspe.umin.jp/medical/chart_dl.html, Accessed on April 2021. A special version for big data analysis was kindly provided by Dr. Yoshiya Ito on behalf of JSPE). Each index required age-in-month to calculate, although the JIPAD database provides age-in-year for subjects aged more than three years. Therefore, for subjects aged three years or older, we calculated these indices using a surrogate age-in-month of “12 × (age) + 6” (e.g. 126 months-old for 10-year-old subjects). Patients were classified into the “extremely low” category for each index when the index was less than − 2.5 SD of the general Japanese population mean14.
Statistical analysis for the distribution of anthropometric data was performed using PRISM 7 (GraphPad) and R software (The R Foundation for Statistical Computing).
## Validity assessment of bodyweight estimation tools
We evaluated the validity of a total of six bodyweight estimation methods (Supplementary material 1): four age-based formulae (the original APLS formula20, the new APLS formula21, the Best Guess formula10, and the JAPAN formulae5) and two height-based methods (Broselow Pediatric Emergency Reference Tape 2019 edition [BT22; Vyaire Medical, Inc., Mettawa, IL, USA], and the JAPAN scale23). We chose the above 6 methods because we have recently developed and validated the age-based JAPAN formulae and the height-based JAPAN scale for bodyweight estimation for children using a Japanese large nationwide longitudinal survey5,23, and the other selected formulae have been commonly applied for bodyweight estimation and widely evaluated internationally3 although the covered age range varies according to the formulae (Supplementary material 1). Instead of fitting Broselow “Tape” to the actual patients, height data were cross-referenced to the BT scale and the JAPAN scale upon height-based bodyweight estimation. Notably, the covered height range varies according to the scales (Supplementary material 1).
The Bland–Altman approach and the proportions of the estimates within $10\%$ of the recorded weight (ρ$10\%$) were used to evaluate the accuracy and precision of the estimation methods as previously described3,4,24,25. *We* generated Bland–Altman plots to visually evaluate the agreement between the recorded and estimated bodyweight and calculated the bias and $95\%$ limits of agreement (LOA)26. The resulting graph describes the difference of the two values (recorded and estimated bodyweight) plotted (the Y axis) against the mean of the two values (the X axis). The bias represents the difference between the recorded and estimated bodyweight where positive and negative values indicate under- and over-estimation of the bodyweight on average, respectively. The $95\%$ LOA shows the interval in which $95\%$ of the differences between the recorded and estimated bodyweight will fall.
While smaller bias and narrow $95\%$ LOA interval mean a better estimation method, the ρ$10\%$ should be as large as possible to be a reliable bodyweight estimation method3,4,24,25. In this study, we also assessed ρ$15\%$ and ρ$20\%$ to reinforce the findings. In addition, we also evaluated the proportions of estimates within absolute difference (2 kg and 4 kg) of the recorded weight, because the percentage difference would carry different impacts depending on the recorded weight in pediatric patients (e.g. The $10\%$ difference for a 10-kg child is 1 kg while it becomes 5 kg for a 50-kg child).
## Ethical approval and consent to participate
The study was approved by the Tokyo Medical and Dental University Review Board (M2020-245) as well as the steering committee of JIPAD, and anonymized data were provided for analysis by the JIPAD.
## Pediatric ICU patients are proportionally small
A total of 7433 admission records from 60 facilities in the JIPAD database were identified in the study period: We excluded 113 records due to missing or improbable data and 704 readmissions within the same hospital stay. We analyzed 6616 admission records with complete data for age, sex, height, and bodyweight. The characteristics of the overall study cohort was presented in Table 1. Overall, the distributions of both bodyweight and height were shifted to the lower side (Fig. 1; the detailed data were shown in the Supplementary material 2 and 3) with approximate mean SDS of − 1.2 and around $20\%$ of patients categorized in the extremely low category (Table 1). The distribution of BMI is almost bell-shaped (Fig. 2) and had higher mean SDS of − 0.52 ($95\%$ CI − 0.57 to − 0.48) with less subjects in the extremely low category ($10.7\%$, Table 1). The disease category subgroup analysis revealed that subjects admitted due to “cardiovascular”, “respiratory”, or “gastrointestinal” diseases had lower mean SDSs than the other categories, mainly contributing to expansion of the population in extremely low categories (Table 2).Table 1Characteristics of study population. N6616Age (year), median [IQR]2.0 [0.0, 8.0]Female [numbers (%)]3029 (45.8)Elective admission [numbers (%)]4184 (63.2)Post-operation [numbers (%)]4653 (70.3)Bodyweight SDS [mean (SD)]− 1.27 (2.36)Bodyweight category [numbers (%)] SDS < − 2.51308 (19.8) − 2.5 ≦ SDS < 2.55266 (79.6) SDS ≧ 2.542 (0.6) Height SDS [mean (SD)]− 1.21 (2.13)Height category [numbers (%)] SDS < − 2.51391 (21.0) − 2.5 ≦ SDS < 2.55138 (77.7) SDS ≧ 2.587 (1.3) BMI SDS [mean (SD)]− 0.52 (1.91)BMI category [numbers (%)] SDS < − 2.5706 (10.7) − 2.5 ≦ SDS < 2.55714 (86.4) SDS ≧ 2.5196 (3.0)BMI, body mass index; IQR, interquartile range; SD(S), standard deviation (score).Patients with SDS < − 2.5 were classified into the “extremely low” category for each index. Figure 1Percentile distribution of height and bodyweight of pediatric patients admitted to intensive care unit. Percentile distribution of height (red lines) and bodyweight (green lines) were overdrawn on the growth charts for Japanese children (black lines, reference #2. The charts were reproduced with official permission of the Japanese Society for Pediatric Endocrinology from Isojima et al. Growth standard charts for Japanese children with mean and standard deviation (SD) values based on the year 2000 national survey. Clin Pediatr Endocrinol 25: 71–76, 2016. ©JSPE). The 1st (fine), 2nd (middle), 3rd (bold), 4th (middle), and 5th (fine) lines from the top indicate 97.5 percentile, 75.0 percentile, 50 percentile, 25 percentile and 2.5 percentile, respectively. Crude data is shown in Supplementary material 2.Figure 2Distributions of SD score of BMI of pediatric population in intensive care units. The bin width of each bar is 0.2. There are 278 data points outside the axis limit. SD, standard deviation; BMI, body mass index. Table 2Characteristics of study population including details of distribution across demographic and anthropometrical indices. NCardiovascular*RespiratoryNeurologicalGastrointestinalTraumaOthers**184115391423786134893Age (year), median [IQR]0.0[0.0, 3.0]3.0[1.0, 8.0]4.0[1.0, 9.0]2.0[0.0, 6.0]6.5[1.0, 10.0]7.0[2.0, 13.0]Female [numbers (%)]877(47.6)595(38.7)661(46.5)383(48.7)48(35.8)465(52.1)Elective admission [numbers (%)]1509(82.0)822(53.4)791(55.6)539(68.6)0(0.0)523(58.6)Post-operation [numbers (%)]1524(82.8)862(56.0)962(67.6)689(87.7)0(0.0)616(69.0)SDS [mean (SD)]Bodyweight− 1.51(2.12)− 1.56(2.83)− 0.57(1.77)− 1.58(2.67)− 0.37(1.34)− 1.26(2.29)Height− 1.30(1.88)− 1.47(2.54)− 0.74(1.94)− 1.58(2.22)− 0.16(1.88)− 1.14(1.94)BMI− 0.84(1.67)− 0.53(2.16)− 0.04(1.63)− 0.58(2.30)− 0.34(1.49)− 0.61(1.85)Extremely low category (SDS < − 2.5) [numbers (%)]Bodyweight422(22.9)362(23.5)149(10.5)193(24.6)3(2.2)179(20.0)Height393(21.3)411(26.7)181(12.7)225(28.6)8(6.0)173(19.4)BMI246(13.4)187(12.2)71(5.0)89(11.3)8(6.0)105(11.8)BMI, body mass index; IQR, interquartile range; SD(S), standard deviation (score).*Disease category “Cardiovascular” diseases includes “Post-cardiopulmonary resuscitation”.**Disease category “Others” includes “Sepsis”.
## The accuracy of age-based bodyweight estimation for pediatric ICU patients is low
The performance of the bodyweight estimation methods was visually summarized in Bland–Altman plots (BA-plots; Fig. 3). The BA-plots for the four age-based methods were more widely distributed than those of the height-based formulae. The BA analysis also provides quantitative assessment of the performance where the best estimation formula should give small bias and narrow $95\%$ LOA interval. The bias of age-based formulae was farther to zero with the wider $95\%$ LOA when compared with that of height-based methods, which indicated that age-based formulae had lower accuracy and precision than height-based methods. We also calculated the ρ$10\%$ which should be a larger value when the estimation formula performs better, and the overall accuracy of age-based formulae indicated by ρ$10\%$ was lower, compared to the two height-based methods (Table 3).Figure 3Bland–Altman plots for estimated bodyweight and measured bodyweight. Bland–Altman plots were drawn for 4 age-based formulas (A) and 2 height-based formulas (B). For the Bland–Altman plots, the long-dashed line indicates the bias, and the area between the short-dashed lines denotes the $95\%$ limits of agreement. There is one data point plotted outside the axis limit in New APLS, Best Guess, JAPAN Formulae, and Broselow Tape, respectively. Table 3Accuracy of each bodyweight estimation formula. Age-based formulaHeight-based formulaOriginal APLSNew APLSBest GuessJAPAN FormulaeBroselow TapeJAPAN ScaleAssessed number*349658876374392555753391Values of Bland–*Altman analysis* Bias− 0.8285− 2.365− 4.286− 1.312− 0.4335− 0.4270 SD of Bias4.2225.1476.4985.4862.9333.123 $95\%$ LOA− 9.103–7.446− 12.45–7.723− 17.02–8.450− 12.06–9.440− 6.183–5.316− 6.548–5.694Proportions of estimates ρ$10\%$ (%, $95\%$CI)38.4 (36.8–40.1)30.7 (29.5–31.9)18.5 (17.6–19.5)38.2 (36.7–39.8)49.3 (48.0–50.6)53.8 (52.1–55.5) ρ$15\%$ (%, $95\%$CI)52.7 (51.0–54.3)43.4 (42.2–44.7)28.1 (27.0–29.2)51.6 (50.0–53.1)66.5 (65.2–67.7)70.6 (69.0–72.1) ρ$20\%$ (%, $95\%$CI)64.9 (63.3–66.5)53.9 (52.6–55.2)37.6 (36.4–38.8)63.4 (61.8–64.9)79.1 (78.0–80.1)82.5 (81.2–83.7) ρ2kg (%, $95\%$CI)52.5 (50.8–54.1)57.7 (56.5–59.0)37.7 (36.5–38.9)48.3 (46.8–49.9)76.6 (75.5–77.7)68.0 (66.4–69.6) ρ4kg (%, $95\%$CI)77.8 (76.4–79.2)75.9 (74.8–77.0)62.8 (61.6–64.0)72.5 (71.1–73.9)91.0 (90.3–91.8)87.9 (86.8–89.0)LOA; limits of agreement; SD: standard deviation;ρ: proportion of estimates within indicated fraction of the measured weight, or within indicated bodyweight range; CI: confidence interval.*The assessed numbers are different depending on the formulae because the covered age/height range varies (see Supplemental material 1).
## Discussion
Rapid and precise estimation of anthropometric values of children is important in an emergency room and ICU because they are key determinants for drug dosage and size of equipment. Several age-based bodyweight estimation formulae have been proposed because age is the most readily available information hence allowing us to immediately work out the answers even in urgent settings such as cardiopulmonary resuscitation, however, the accuracy of these formulae has been questioned4. Height-based estimation formulae such as Broselow Tape are also widely used methods, but their accuracy has also been challenged3. Importantly, these estimation methods were derived from general pediatric populations, therefore whether it is applicable to critically ill children has not been well elucidated. This is particularly important in the pediatric ICU where significant proportions of patients have preexisting comorbidities and resulting failure to thrive. In this study, we demonstrated detailed visual data on anthropometric characteristics of the pediatric ICU population in Japan where both the ICU database and the national pediatric anthropometric references have been long established17.
We have demonstrated that the distributions of bodyweight and height of pediatric ICU population are shifted to the lower side, in line with the previous studies14,15. We have also demonstrated that the proportion of extremely low weight-for-age/height-for-age reaches to around $20\%$ of pediatric patients in ICU while the general prevalence of childhood stunting in developed countries including *Japan is* around $6\%$27. On the other hand, we have also described that BMI-for-age had a balanced bell-shaped distribution, which suggests that the body shape is maintained conformable to bodyweight and height for each age in this population. Considering that BMI is a practical assessment index for nutritional status19,28, we speculate that the possible major explanation for the distribution dissociation between weight-/height-for-age and BMI-for-age is the patients’ morbidity rather than the poor nutritional status. Indeed, most disease groups had lower SDS of weight- or height-for-age than that of BMI-for-age, whereas the subjects with “trauma”, which is an acutely acquired condition, had comparable values of these indexes which were closer to zero. In response, most disease group have more subjects in the extremely low categories of weight- or height-for-age than that of BMI-for-age, whereas few subjects with “trauma” belonged to the extremely low categories of these indexes (Table 2).
These “proportionally small-for-age” anthropometric characteristics explain why the height-based bodyweight estimation methods had superior validity over age-based methods for children in ICU. In line with this, Flannigan et al. have described that the age-based new APLS formulae can overestimate the bodyweight of PICU patients in the UK29 by approximately $20\%$. Moreover, as shown in the Supplementary material 3, the distribution in bodyweight has a wide range in each age, suggesting that mean-for-age bodyweight alone carries a high risk of misestimation of actual bodyweight. This evidence collectively agrees with the recent SCCM guideline for safe medication in ICU where BT was recommended to reduce medication errors30.
Importantly, the “proportionally small-for-age” anthropometric characteristics of pediatric ICU population could influence the safety upon device size selection; i.e. age-based methods could overestimate device size in this population, contributing to undesirable outcomes. For example, overestimating endotracheal tube size does matter for pediatric ICU population because this may result in multiple unrequired attempts at intubation and upper airway injury due to excessive pressures on the mucosa, leading to post-extubation sore throat31,32, or subglottic stenosis at worst33,34. Indeed, there are several studies which demonstrated the inferior ability of the age-based device size estimation over the other approaches35–39. Therefore, from the viewpoint of patient safety, we suggest avoiding the age-based device size estimation and choose alternative way (e.g. height-based estimation) given the pediatric ICU population has such “proportionally small-for-age” anthropometric trends.
This study was inherently subject to some limitations. First, we calculated SDS for subjects aged 3 years or older using surrogate age of “12 × (age) + 6” as explained in the Methods section. We performed the sensitivity analysis, for confirmation, with the SDS calculated using the most conservative surrogate age of “12 × (age)” for these subjects, in which the results produced the same conclusion as the original (data not shown). However, we did not perform the analysis with the SDS calculated using the surrogate age of “12 × (age) + 11” because it was evident that the calculated SDS becomes smaller as the reference age gets older. Second, similar to other databases14,40, the JIPAD database allows guardian-reported or estimated values in case measured values of bodyweight and height are not available. However, we have reported that the accuracy of mother-reported anthropometric values are extremely high in Japan (ρ$10\%$: $94.9\%$, ρ$20\%$: $98.7\%$)24. Third, our data confirmed the superiority of height-based methods over age-based methods for bodyweight estimation of pediatric ICU patients, however, the ρ$10\%$ of height-based methods in this study were still lower than those reported previously4,41. Accordingly, we recommend avoiding age-based methods, and using height-based methods until obtaining patients’ actual bodyweight information in these population.
## Conclusion
We demonstrated that the distributions of bodyweight and height of pediatric population in intensive care units are skewed toward small-for-age using prospectively collected database from 60 ICUs in Japan. Our results suggest a special risk of using age-based methods, and support relative but clear advantages of using height-based methods for patient safety, especially in pediatric ICU settings.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30566-3.
## References
1. 1.Tofovic, S. P. & Kharasch, E. in Smith’s Anesthesia for Infants and Children (eds P.J. Davis & F.P. Cladis) Ch. 7, (Elsevier, 2017).
2. Isojima T, Kato N, Ito Y, Kanzaki S, Murata M. **Growth standard charts for Japanese children with mean and standard deviation (SD) values based on the year 2000 national survey**. *Clin. Pediatr. Endocrinol.* (2016.0) **25** 71-76. DOI: 10.1297/cpe.25.71
3. Wells M, Goldstein LN, Bentley A, Basnett S, Monteith I. **The accuracy of the Broselow tape as a weight estimation tool and a drug-dosing guide - A systematic review and meta-analysis**. *Resuscitation* (2017.0) **121** 9-33. DOI: 10.1016/j.resuscitation.2017.09.026
4. Wells M, Goldstein LN. **How and why paediatric weight estimation systems fail - a body composition study**. *Cureus* (2020.0) **12** e7198. DOI: 10.7759/cureus.7198
5. Nosaka N. **New age-based weight estimation formulae for Japanese children**. *Pediatr. Int.* (2017.0) **59** 727-732. DOI: 10.1111/ped.13259
6. Theron L, Adams A, Jansen K, Robinson E. **Emergency weight estimation in Pacific Island and Maori children who are large-for-age**. *Emerg. Med. Australas.* (2005.0) **17** 238-243. DOI: 10.1111/j.1742-6723.2005.00729.x
7. Leffler S, Hayes M. **Analysis of parental estimates of children's weights in the ED**. *Ann. Emerg. Med.* (1997.0) **30** 167-170. DOI: 10.1016/s0196-0644(97)70137-9
8. Park J. **A new age-based formula for estimating weight of Korean children**. *Resuscitation* (2012.0) **83** 1129-1134. DOI: 10.1016/j.resuscitation.2012.01.023
9. Luscombe M, Owens B. **Weight estimation in resuscitation: Is the current formula still valid?**. *Arch. Dis. Child.* (2007.0) **92** 412-415. DOI: 10.1136/adc.2006.107284
10. Tinning K, Acworth J. **Make your Best Guess: An updated method for paediatric weight estimation in emergencies**. *Emerg. Med .Australas.* (2007.0) **19** 528-534. DOI: 10.1111/j.1742-6723.2007.01026.x
11. Mehta NM. **Nutritional practices and their relationship to clinical outcomes in critically ill children–an international multicenter cohort study**. *Crit. Care Med.* (2012.0) **40** 2204-2211. DOI: 10.1097/CCM.0b013e31824e18a8
12. Grippa RB. **Nutritional status as a predictor of duration of mechanical ventilation in critically ill children**. *Nutrition* (2017.0) **33** 91-95. DOI: 10.1016/j.nut.2016.05.002
13. Campos-Mino S, Figueiredo-Delgado A. **Failure to thrive in the PICU: An overlooked real problem**. *Pediatr. Crit. Care Med.* (2019.0) **20** 776-777. DOI: 10.1097/PCC.0000000000001986
14. Prince NJ, Brown KL, Mebrahtu TF, Parslow RC, Peters MJ. **Weight-for-age distribution and case-mix adjusted outcomes of 14,307 paediatric intensive care admissions**. *Intensive Care Med* (2014.0) **40** 1132-1139. DOI: 10.1007/s00134-014-3381-x
15. Ross PA, Newth CJ, Leung D, Wetzel RC, Khemani RG. **Obesity and mortality risk in critically ill children**. *Pediatrics* (2016.0) **137** e20152035. DOI: 10.1542/peds.2015-2035
16. Irie H. **The Japanese Intensive care PAtient Database (JIPAD): A national intensive care unit registry in Japan**. *J. Crit. Care* (2020.0) **55** 86-94. DOI: 10.1016/j.jcrc.2019.09.004
17. Isojima T. **Growth references for Japanese individuals with Noonan syndrome**. *Pediatr. Res.* (2016.0) **79** 543-548. DOI: 10.1038/pr.2015.254
18. Must A, Anderson SE. **Body mass index in children and adolescents: considerations for population-based applications**. *Int. J. Obes. (Lond)* (2006.0) **30** 590-594. DOI: 10.1038/sj.ijo.0803300
19. Bechard LJ. **Nutritional status based on body mass index is associated with morbidity and mortality in mechanically ventilated critically ill children in the PICU**. *Crit. Care Med.* (2016.0) **44** 1530-1537. DOI: 10.1097/CCM.0000000000001713
20. Mackway-Jones K, Molyneux E, Phillips B. *Advanced Paediatric Life Support: The Practical Approach* (2005.0)
21. Samuels M, Wieteska S. *Advanced Paediatric Life Support: The Practical Approach* (2011.0)
22. Deboer S, Seaver M, Broselow J. **Color coding to reduce errors**. *Am. J. Nurs.* (2005.0) **105** 68-71. DOI: 10.1097/00000446-200508000-00031
23. Nosaka N. **Development of a Japanese scale for assessment of paediatric normal weight**. *Resuscitation* (2016.0) **105** e11-12. DOI: 10.1016/j.resuscitation.2016.04.025
24. Nosaka N, Fujiwara T, Knaup E, Okada A, Tsukahara H. **Validity of mothers' reports of children's weight in Japan**. *Acta Med. Okayama* (2016.0) **70** 255-259. DOI: 10.18926/AMO/54500
25. Loo PY, Chong SL, Lek N, Bautista D, Ng KC. **Evaluation of three paediatric weight estimation methods in Singapore**. *J. Paediatr Child Health* (2013.0) **49** E311-316. DOI: 10.1111/jpc.12141
26. Bland JM, Altman DG. **Statistical methods for assessing agreement between two methods of clinical measurement**. *Lancet* (1986.0) **1** 307-310. DOI: 10.1016/S0140-6736(86)90837-8
27. de Onis M, Blossner M, Borghi E. **Prevalence and trends of stunting among pre-school children, 1990–2020**. *Public Health Nutr.* (2012.0) **15** 142-148. DOI: 10.1017/S1368980011001315
28. Valla FV. **Nutritional management of young infants presenting with acute bronchiolitis in Belgium, France and Switzerland: Survey of current practices and documentary search of national guidelines worldwide**. *Eur. J. Pediatr.* (2019.0) **178** 331-340. DOI: 10.1007/s00431-018-3300-1
29. Flannigan C, Bourke TW, Sproule A, Stevenson M, Terris M. **Are APLS formulae for estimating weight appropriate for use in children admitted to PICU?**. *Resuscitation* (2014.0) **85** 927-931. DOI: 10.1016/j.resuscitation.2014.03.313
30. Kane-Gill SL. **Clinical practice guideline: Safe medication use in the ICU**. *Crit. Care Med.* (2017.0) **45** e877-e915. DOI: 10.1097/CCM.0000000000002533
31. McHardy FE, Chung F. **Postoperative sore throat: Cause, prevention and treatment**. *Anaesthesia* (1999.0) **54** 444-453. DOI: 10.1046/j.1365-2044.1999.00780.x
32. Stout DM, Bishop MJ, Dwersteg JF, Cullen BF. **Correlation of endotracheal tube size with sore throat and hoarseness following general anesthesia**. *Anesthesiology* (1987.0) **67** 419-421. DOI: 10.1097/00000542-198709000-00025
33. Sherman JM, Nelson H. **Decreased incidence of subglottic stenosis using an "appropriate-sized" endotracheal tube in neonates**. *Pediatr. Pulmonol.* (1989.0) **6** 183-185. DOI: 10.1002/ppul.1950060311
34. Contencin P, Narcy P. **Size of endotracheal tube and neonatal acquired subglottic stenosis. Study Group for Neonatology and Pediatric Emergencies in the Parisian Area**. *Arch. Otolaryngol. Head Neck Surg.* (1993.0) **119** 815-819. DOI: 10.1001/archotol.1993.01880200015002
35. Davis D, Barbee L, Ririe D. **Pediatric endotracheal tube selection: A comparison of age-based and height-based criteria**. *AANA J.* (1998.0) **66** 299-303. PMID: 9830856
36. Shibasaki M. **Prediction of pediatric endotracheal tube size by ultrasonography**. *Anesthesiology* (2010.0) **113** 819-824. DOI: 10.1097/ALN.0b013e3181ef6757
37. Park S. **Prediction of endotracheal tube size using a printed three-dimensional airway model in pediatric patients with congenital heart disease: A prospective, single-center, single-group study**. *Korean J. Anesthesiol.* (2021.0) **74** 333-341. DOI: 10.4097/kja.21114
38. Good RJ. **accuracy of bedside ultrasound femoral vein diameter measurement by PICU providers**. *Pediatr. Crit. Care Med.* (2020.0) **21** e1148-e1151. DOI: 10.1097/PCC.0000000000002439
39. Tsukamoto M, Yamanaka H, Yokoyama T. **Predicting the appropriate size of the uncuffed nasotracheal tube for pediatric patients: A retrospective study**. *Clin. Oral Investig.* (2019.0) **23** 493-495. DOI: 10.1007/s00784-018-2774-6
40. Numa A, McAweeney J, Williams G, Awad J, Ravindranathan H. **Extremes of weight centile are associated with increased risk of mortality in pediatric intensive care**. *Crit. Care* (2011.0) **15** R106. DOI: 10.1186/cc10127
41. Wells M, Goldstein LN, Bentley A. **A systematic review and meta-analysis of the accuracy of weight estimation systems used in paediatric emergency care in developing countries**. *Afr. J. Emerg. Med* (2017.0) **7** S36-S54. DOI: 10.1016/j.afjem.2017.06.001
|
---
title: ANGPTL4 stabilizes atherosclerotic plaques and modulates the phenotypic transition
of vascular smooth muscle cells through KLF4 downregulation
authors:
- Dong Im Cho
- Min Joo Ahn
- Hyang Hee Cho
- Meeyoung Cho
- Ju Hee Jun
- Bo Gyeong Kang
- Soo Yeon Lim
- Soo Ji Yoo
- Mi Ra Kim
- Hyung-Seok Kim
- Su-Jin Lee
- Le Thanh Dat
- Changho Lee
- Yong Sook Kim
- Youngkeun Ahn
journal: Experimental & Molecular Medicine
year: 2023
pmcid: PMC9981608
doi: 10.1038/s12276-023-00937-x
license: CC BY 4.0
---
# ANGPTL4 stabilizes atherosclerotic plaques and modulates the phenotypic transition of vascular smooth muscle cells through KLF4 downregulation
## Abstract
Atherosclerosis, the leading cause of death, is a vascular disease of chronic inflammation. We recently showed that angiopoietin-like 4 (ANGPTL4) promotes cardiac repair by suppressing pathological inflammation. Given the fundamental contribution of inflammation to atherosclerosis, we assessed the role of ANGPTL4 in the development of atherosclerosis and determined whether ANGPTL4 regulates atherosclerotic plaque stability. We injected ANGPTL4 protein twice a week into atherosclerotic Apoe−/− mice and analyzed the atherosclerotic lesion size, inflammation, and plaque stability. In atherosclerotic mice, ANGPTL4 reduced atherosclerotic plaque size and vascular inflammation. In the atherosclerotic lesions and fibrous caps, the number of α-SMA(+), SM22α(+), and SM-MHC(+) cells was higher, while the number of CD68(+) and Mac2(+) cells was lower in the ANGPTL4 group. Most importantly, the fibrous cap was significantly thicker in the ANGPTL4 group than in the control group. Smooth muscle cells (SMCs) isolated from atherosclerotic aortas showed significantly increased expression of CD68 and Krüppel-like factor 4 (KLF4), a modulator of the vascular SMC phenotype, along with downregulation of α-SMA, and these changes were attenuated by ANGPTL4 treatment. Furthermore, ANGPTL4 reduced TNFα-induced NADPH oxidase 1 (NOX1), a major source of reactive oxygen species, resulting in the attenuation of KLF4-mediated SMC phenotypic changes. We showed that acute myocardial infarction (AMI) patients with higher levels of ANGPTL4 had fewer vascular events than AMI patients with lower levels of ANGPTL4 ($p \leq 0.05$). Our results reveal that ANGPTL4 treatment inhibits atherogenesis and suggest that targeting vascular stability and inflammation may serve as a novel therapeutic strategy to prevent and treat atherosclerosis. Even more importantly, ANGPTL4 treatment inhibited the phenotypic changes of SMCs into macrophage-like cells by downregulating NOX1 activation of KLF4, leading to the formation of more stable plaques.
## Cardiovascular disease: Stopping and stabilizing atherosclerotic plaques
Treatment with a protein that stabilizes existing plaques within blood vessels could help reduce the risk of future cardiovascular events in patients with atherosclerosis. These plaques arise in part from a change in the behavior of the muscle cells within the walls of the blood vessels, which leads to the accumulation of lipids and other biomolecules and creates conditions that can ultimately result in a heart attack or stroke. Researchers led by Youngkeun Ahn and Yong Sook Kim at Chonnam National University Hospital, Gwangju, South Korea, have shown that they can counter this process in a mouse model of atherosclerosis by treatment with a protein called ANGPTL4. This molecule keeps vascular muscle cells in a state that prevents further plaque formation, while stabilizing existing plaques and countering the inflammatory processes that can further accelerate the cardiovascular disease.
## Introduction
Atherosclerosis, a complex vascular disorder, develops in the context of dyslipidemia and chronic inflammation and is the main contributor to cardiovascular mortality. In response to injury or inflammatory stimuli, circulating monocytes are recruited to activated endothelial cells by adhesion molecules such as monocyte chemoattractant protein-1 (MCP-1) and vascular cell adhesion molecule-1 (VCAM-1). These monocytes differentiate into macrophages to take up excessive lipids through scavenger receptors and form fatty streaks. In the lesion, activated vascular smooth muscle cells (VSMCs) become highly proliferative and migrate to the subendothelial space to form the tunica intima, and the phenotype of the VSMCs is shifted to the synthetic type from the contractile type to contribute to atheroma formation. This dynamic interplay between endothelial cells, VSMCs, and macrophages is altered in vascular pathologies such as atherosclerosis.
Thin-cap fibroatheroma, which has an increased risk of thrombosis, is a kind of vulnerable plaque1,2. Vulnerable plaques are at high risk of cardiovascular events such as acute myocardial infarction (AMI) and stroke3. Despite recent advances in the study of plaque biology, the mechanisms and contributing factors controlling the stability of atherosclerotic lesions remain unclear.
Clinical interventions for atherosclerosis include mostly lipid-lowering therapies, of which statins are the most widely used. Statins decrease C-reactive protein (CRP) concentrations, increase the collagen content of atherosclerotic plaques, change endothelial function, and reduce monocyte recruitment and macrophage accumulation in plaques4–6.
The results of many clinical trials attest to the diverse roles of statins7. Moderate or intensive statin therapy decreases low-density lipoprotein (LDL) cholesterol8, increases atheroma stabilization9, and enhances the regression of coronary plaque volume10. The results of the Study of Coronary Atheroma by Intravascular Ultrasound (SATURN) trial suggest that plaque regression occurs in more than $60\%$ of patients treated with rosuvastatin or atorvastatin11. Unfortunately, atherosclerosis continues to progress in up to one-third of patients despite intensive statin therapy12.
In 2017, the Canakinumab Anti-inflammatory Thrombosis Outcomes Study (CANTOS) showed that inhibition of inflammatory interleukin-1β (IL-1β) can significantly reduce the number of cardiovascular events independently of lipid-lowering, but the mechanisms underlying this anti-inflammatory therapy remain unclear13. The Cardiovascular Inflammation Reduction Trial (CIRT) showed that treatment with low-dose methotrexate fails to lower cardiovascular event rates14. Then, in 2019, The Colchicine Cardiovascular Outcomes Trial (COLCOT) suggested that treatment with colchicine reduces the incidence of adverse coronary and cerebral atherothrombotic events15.
Since we have shown the cardioprotective effects of anti-inflammatory angiopoietin-like 4 (ANGPTL4) in a mouse model of MI16, we were interested in whether ANGPTL4 could exert a protective function during the development of atherosclerosis. ANGPTL proteins are structurally similar to angiopoietins that play a role in a wide array of biological functions, including the regulation of lipid and glucose metabolism, hematopoietic stem cell expansion, chronic inflammation, angiogenesis, and wound healing. Therefore, to investigate the effect of ANGPTL4 on inflammatory vascular disease, we performed a comprehensive evaluation of atherogenesis, including inflammatory phenotype and plaque stability, in an Apoe-/- mouse model of high-fat diet-induced atherosclerosis. Furthermore, to understand the clinical implications of our studies, we measured circulating levels of ANGPTL4 in patients with cardiovascular disease to determine the association of ANGPTL4 with clinical outcomes.
Recent data indicate that VSMCs in atherosclerotic lesions undergo a phenotypic transition to macrophage-like cells that express both macrophage and SMC markers and promote inflammation and enhanced atherogenesis17. In addition, activation of the transcription factor KLF4 by oxidized phospholipids promotes phenotypic modulation of VSMCs in mouse and human atherosclerotic plaques18. Other functions of VSMCs include sustaining vascular wall reactive oxygen species (ROS) generation, inflammation, and matrix remodeling. Thus, VSMCs are believed to play a major role in atherogenesis and the evolution of atherosclerotic lesions19.
In this study, we demonstrated that ANGPTL4 administration significantly reduced atherosclerotic lesion size, macrophage content, vascular inflammation, and phenotypic transition of smooth muscle cells, contributing to plaque stabilization by downregulating NADPH oxidase 1 (NOX1) activation of KLF4 in an atherosclerosis mouse model.
## Human study population
Study subjects provided written informed consent before enrollment. For immunohistochemical studies, human left anterior descending artery (LAD) specimens were obtained from the Department of Forensic Medicine, Chonnam National University Medical School, after approval of the research protocol used in this study. The Institutional Review Board of Chonnam National University Medical School and Hospital waived the requirement to obtain informed consent for the use of the forensic samples because of the inaccessibility to personally identifiable information and the fact that there were no tests for heritable traits in accordance with Article 33 of the Enforcement Regulations provided by the Korean government.
A total of 253 AMI patients who were admitted to Chonnam National University Hospital between January 2016 and July 2019 and who consented to the use of their blood samples for scientific purposes were screened. Forty-four patients were excluded due to a previous history of AMI or angina treated with percutaneous coronary intervention (PCI). Two more patients who died during the index hospitalization were also excluded. All patients had a successful PCI. Blood samples were obtained from all patients during PCI, and plasma ANGPTL4 levels were measured using an enzyme-linked immunosorbent assay (ELISA) kit (EHANGPTL4, Thermo Fisher Scientific). Patients were divided into two groups with high or low ANGPTL4 according to the median ANGPTL4 level. We compared clinical outcomes between the two groups during the follow-up period. The median duration of follow-up was 1.9 years (interquartile range, 1.3–3.3 years). This study was a single-center study, and the study protocol was approved by the Chonnam National University Hospital Institutional Review Board (BTMP-2020-330).
## Study definitions and endpoints
AMI was defined as the presence of acute myocardial injury detected by abnormal levels of cardiac biomarkers and angiographically proven atherothrombotic coronary artery disease (CAD). Cardiology specialists collected the patients’ medication lists and medical histories, such as the presence of hypertension, diabetes mellitus, and dyslipidemia. All laboratory variables were measured upon admission, except for lipid profiles, which were obtained after at least 9 h of fasting within 24 h of hospitalization. The baseline left ventricular ejection fraction (LVEF) was determined by two-dimensional echocardiography performed before or immediately after PCI. Coronary blood flow before and after PCI was classified by the thrombolysis in myocardial infarction (TIMI) score, and coronary lesion complexity was based on the American College of Cardiology (ACC)/American Heart Association (AHA) definitions. Patients who underwent PCI received 300 mg of aspirin and 600 mg of clopidogrel, 60 mg of prasugrel, or 180 mg of ticagrelor as a loading dose before PCI. Doses of 50 to 70 U/kg of unfractionated heparin were used before or during PCI to maintain an activated clotting time of 250 to 300 s. After PCI, 100 to 300 mg of aspirin and 75 mg of clopidogrel, 5 to 10 mg of prasugrel, or 45 to 90 mg of ticagrelor were prescribed daily. All patients had coronary lesions with at least $50\%$ stenosis by quantitative coronary analysis. The study endpoint was vascular events associated with plaque instability, which consisted of recurrent AMI, stent thrombosis, and ischemic cerebral infarction. Recurrent MI was defined as the development of recurrent angina symptoms with new 12-lead electrocardiographic changes or increased cardiac-specific biomarkers.
## Mouse lines and atherosclerosis models
C57Bl/6 Apoe−/− mice were purchased from Jackson Laboratories (Bar Harbor, ME). Male mice were used for experiments beginning at 8 to 10 weeks of age. Accelerated atherosclerosis was induced by feeding the mice for 8 weeks with a Paigen diet containing $16\%$ fat, $1.25\%$ cholesterol, and $0.5\%$ cholate (D12336, Research Diets) and with a Western diet containing $0.15\%$ cholesterol (D12079B, Research Diets). Apoe−/− mice receiving a high-fat diet were intraperitoneally injected with 2 μg of ANGPTL4 (4880-AN, R&D Systems) two times per week for 8 weeks. The mice were bred and maintained under pathogen-free conditions at Chonnam National University Medical School animal facilities. All experiments were performed after approval by our local ethics committee at Chonnam National University Medical School (CNU IACUC-H-2018-55). Ldlr−/− mice were provided by the National Institute of Food and Drug Safety Evaluation (Chungcheongbukdo, Korea). The mouse genotyping primers are described in Supplementary Table 2.
## Cell culture
Human aortic SMCs (CC-2571, Lonza) were used between passages 4 and 10. Human aortic SMCs were maintained in SmGM 2 Smooth Muscle Cell Growth Medium 2 (CC-3182, Lonza). Human induced pluripotent stem cells (hiPSCs) were differentiated into hiPSC-derived cardiac progenitor cells (hiPSC-CPCs) using a STEMdiff™ Cardiomyocyte Differentiation kit (05010, STEMCELL). First, hiPSCs were incubated for 2 days in STEMdiff™ Cardiomyocyte Differentiation Medium A, for 2 days in STEMdiff™ Cardiomyocyte Differentiation Medium B, and for 4 days in STEMdiff™ Cardiomyocyte Differentiation Medium C. On Day 8, hiPSC-CPCs were cultured in advanced DMEM/F-12 medium (12634010, Gibco®, Life Technologies) with 5 μM CHIR99021 (S2924, Selleck Chemicals) and 2 μM retinoic acid (R2625, Sigma‒Aldrich) for 3 days and recovered in advanced DMEM/F-12 medium for 4 days. Human iPSC-epicardial cells (hiPSC-EPCs) were cultured in SMC growth medium with PDGF-BB (10 ng/ml, MBS142119, MyBioSource) and TGFβ1 (2 ng/ml, 7754-BH, R&D Systems) for 12 days. Human iPSC-derived smooth muscle cells (hiPSC-SMCs) (αSMA+/SM22α+) were identified by qPCR and immunofluorescence staining. For mouse bone marrow-derived macrophages (BMDMs), mononuclear cells were isolated from mouse bone marrow and cultured for 7 days in macrophage differentiation medium (supplemented with $30\%$ L929 cell-conditioned medium, $20\%$ fetal bovine serum [FBS; 16000044, Gibco], and $50\%$ RPMI-1640 [11875-093, Gibco]). L929 cell-conditioned medium was prepared by growing L929 cells in RPMI-1640 medium containing $10\%$ FBS for 10 days. The medium containing macrophage colony-stimulating factor secreted by the L929 cells was harvested and passed through a 0.22-mm filter. BMDMs were cultured in RPMI supplemented with $10\%$ FBS for 3 days. Conditioned medium was collected and used immediately or stored at −80 °C.
## RNA isolation and real-time PCR
Tissue and cell RNAs were extracted using TRIzol (15596018, Life Technologies) and converted to cDNA using an Applied Biosystems High-Capacity cDNA reverse transcription kit (4368814, Life Technologies) according to the manufacturer’s instructions. Real-time PCR was performed using a QuantiTect SYBR Green PCR kit (204143, Qiagen) and Corbett Research Rotor-Gene RG-3000 Real-Time PCR System. The mouse primers are listed in Supplementary Table 1.
## Analysis of atherosclerotic lesions
For the analysis of mouse atherosclerotic lesions, aortas were harvested, cleaned of the adventitia, dissected longitudinally along the greater and lesser curvature for bilateral presentation, pinned, and en face-stained with Oil red O (O1391, Sigma‒Aldrich) for lipid measurement at the surface of the vascular wall. The images were captured using a digital camera (Samsung, Korea). The aortas and aortic roots were stained for lipid deposition with Oil red O. In brief, hearts with aortic roots were embedded in optimal cutting temperature (OCT) compound (3801480, Leica) for cryosectioning (Leica CM1850). Atherosclerotic lesions were quantified in 10-μm transverse sections, and the averages were calculated from 3 to 5 sections. The slides were stained using Oil red O for lipid deposition, hematoxylin-eosin (H&E, ab245880, Abcam) for aortic plaque necrosis, Masson’s trichrome staining (ab150686, Abcam) for the aortic fibrous cap, and picrosirius red staining (150681, Abcam) for analysis of collagen content. For histological analysis, the images were quantified as the average lesion area, which was measured using a color image analysis system (NIS-Elements Imaging Software, Nikon, Japan). For analysis of the cellular composition or inflammation of atherosclerotic lesions, sections were stained with an antibody against CD68 (ab125212, Abcam), Mac2 (CL8942AP, Cedarlane), α-SMA (A2547, Sigma), transgelin (SM22α) (Ab14106, Abcam), or SM-MHC (TA323338, OriGene). Foam cells were stained with 10 μg/ml BODIPY $\frac{493}{503}$ (D3922, Invitrogen) at the same time as secondary antibody incubations. The nuclei were counterstained using 4’,6-diamidino-2-phenylindole (DAPI), and the positive areas were quantified with a color image analysis system.
## Isolation of murine aortic vascular smooth muscle cells
Aortic VSMCs were isolated by procedures described in Dr. Gary Owens’s Laboratory. Briefly, aortas from the root to the iliac bifurcation were dissected from adult Apoe−/− mice and opened longitudinally. The aortas were sequentially washed with $1\%$ penicillin/streptomycin in HBSS (14025-092, Gibco), and then the endothelial cells and the adventitia were gently removed. Aortas were cut into small pieces and incubated at 37 °C in $5\%$ CO2 in an enzyme solution (collagenase type II [LS004176, Worthington], elastase [E1250, Sigma], $1\%$ penicillin/streptomycin, $20\%$ FBS in DMEM/F-12 medium) for 1 h. Aortas were cultured in a medium with $20\%$ FBS and $1\%$ penicillin/streptomycin. After 7 days, the aortic medium of the SMCs was replaced with a fresh medium. Aortic SMCs were transferred to a fresh medium containing $10\%$ FBS after passage 3 depending on their viability.
## Loading with water-soluble cholesterol
Aortic VSMCs were serum starved in DMEM with $0.2\%$ BSA for 24 h and then loaded with cholesterol. SMCs were incubated with cholesterol-methyl-β-cyclodextrin complexes (Chol:MβCD, 10 μg/mL) (C4951, Sigma) and $0.2\%$ BSA (A9418, Sigma) in DMEM for 3 days and then prepared for experiments.
## Flow cytometry
Freshly isolated aortas were resuspended in FACS buffer (PBS containing $1\%$ FBS and 2 mM EDTA) and stained with conjugated antibody for 20 to 30 min at 4 °C. Cells were washed and resuspended in FACS buffer for flow cytometric analyses in which inflammatory cell populations were designated, following gating/stratification of their marker profile. The aortas were cut into small pieces and digested in an enzyme mixture containing 450 U/mL collagenase I (C0130, Sigma‒Aldrich), 125 U/mL collagenase XI (C7657, Sigma‒Aldrich), 60 U DNase I (DN25, Sigma‒Aldrich), and 60 U/mL hyaluronidase (2592, Worthington Laboratories) in PBS with Ca2+/Mg2+ for 60 min at 37 °C with gentle shaking. After incubation, the digestion mixture was homogenized through a 70-μm nylon mesh. The digestion mixture was centrifuged at 2000 rpm for 15 min at 4 °C, and the cells were simultaneously stained with antibodies at 4 °C for 15 min and then washed and resuspended in a staining buffer. All antibodies are listed in Supplementary Table 3. The cell suspensions were analyzed with a BD FACSCANTO II flow cytometry system (BD Biosciences), and postacquisition analysis was performed with FlowJo7 software (Tree Star).
## Foam cell formation assay
BMDMs from Apoe−/− mice were fixed with $4\%$ paraformaldehyde (DN-4310, DANA Korea) for 20 min and then stained with Oil red O for 20 min at 37 °C. The BMDMs were washed with PBS, rinsed in $60\%$ isopropanol (109634, Merck Millipore) for 15 s, and observed using a fluorescence microscope (Nikon, Japan). The Oil red O content was quantified using a color image analysis system (NIS-Elements Imaging Software, Nikon).
## Oxidized LDL uptake assay
The assay was performed using an oxidized LDL (oxLDL) uptake assay kit (601180, Cayman) or DiI-oxLDL (L34358, Invitrogen). The BMDMs from Apoe−/− mice were incubated with oxLDL-DyLight 488 or Dil-oxLDL (10 μg/mL) for 4 h. After unbound LDL was washed out, the cells were fixed with $4\%$ paraformaldehyde. The fluorescent cells were observed using an Eclipse Ti2 microscope (Nikon, Japan).
## Immunoblot analysis
Cells were washed with ice-cold phosphate-buffered saline and lysed in PRO-PREP Protein Extraction Solution (17081, iNtRON Biotechnology) on a rotation wheel for 20 min at 4 °C. After centrifugation at 10,000×g for 10 min, the supernatant was prepared as a protein extract. Equal amounts of proteins were fractionated by electrophoresis on 8 or $10\%$ acrylamide gels and were transferred onto a polyvinylidene fluoride membrane (IPVH00010, Millipore), followed by blotting with primary antibody and horseradish peroxidase-conjugated secondary IgG antibodies. Protein expression was detected using an Image Reader (LAS-3000 Imaging System, Fuji Photo Film).
## Human aorta smooth muscle cell proliferation assay
To stain proliferating cells with Ki67, HAoSMCs were fixed in $4\%$ paraformaldehyde for 15 min. Then, the cell membrane was penetrated by $0.25\%$ Triton X-100 for 10 min and blocked with normal goat serum for 1 h, followed by Ki67 antibody (ab15580, Abcam) staining at 4 °C overnight. Subsequently, the cells were incubated with secondary antibodies conjugated with Alexa-488 (A11034, Molecular Probes) or Alexa-594 (A11037, Molecular Probes) for 1 h, followed by mounting with 4,6-diamidino-2-phenylindole (D1306, Molecular Probes). The stained cells were visualized using a fluorescence microscope. Additionally, HAoSMC proliferation was analyzed using a colorimetric bromodeoxyuridine (BrdU) ELISA kit (6813, Cell Signaling Technology) according to the manufacturer’s manual. Cells were seeded into 96-well plates at a density of 1 × 104 cells/well. The cells were then pretreated with or without different concentrations of recombinant human ANGPTL4 (4487-AN, R&D Systems) without serum for 3 h and stimulated with platelet-derived growth factor-BB (PDGF-BB, 20 ng/mL, MBS142119, MyBioSource) for 24 h. The cells were then labeled with BrdU labeling reagent for 3 h. After fixation, the cells were incubated with an anti-BrdU antibody for 1 h. After washing, HRP-conjugated secondary antibody substrate TMB was added to each well, and the plates were incubated at room temperature for 30 min. The absorbance was measured at a dual wavelength of $\frac{450}{550}$ nm.
## Blood lipid and cytokine analyses
Pooled plasma samples from mice were assayed for leptin (MOB00B, R&D systems), IL-6 (BMS603-2, Invitrogen) IL-1β (BMS6002, Invitrogen), IL-18 (BMS618-3, Invitrogen), and ANGPTL4 (EHANGPTL4, Invitrogen) using ELISA kits.
## Luciferase reporter assay
For the luciferase assay, transfections were performed with Lipofectamine 2000 transfection reagent (Invitrogen, 11668-019) in human aortic SMCs. The Firefly luciferase vector was cotransfected with the KLF4 promoter construct as a control for transfection efficiency. The cells were then pretreated with or without different concentrations of recombinant human ANGPTL4 (4487-AN, R&D Systems) and stimulated with oxLDL (10 μg/mL) (L34357, Invitrogen) or recombinant human tumor necrosis factor α (TNF-α; 10 ng/mL) (210-TA, R&D Systems) for 24 h. The promoter activity was measured using the Pierce Gaussia-Firefly Luciferase Dual Assay kit (16182, Thermo Fisher Scientific) and a Sirius Luminometer (Berthold Technologies) according to the manufacturer’s instructions.
## ROS detection
In cultured VSMCs treated with TNFα (100 ng/ml), ROS levels were determined immediately after sample collection. Cellular ROS levels were assessed by measuring CM-H2DCFDA (C-6827, Invitrogen) fluorescence. CM-H2DCFDA was added to the culture medium at a final concentration of 13 μM, and the cells were incubated for 10 min at 37 °C. Cells were then washed three times with prewarmed PBS, and fluorescence images were taken using an Eclipse Ti2 microscope (Nikon, Japan). *Superoxide* generation in aortic root sections from Apoe−/− mice was evaluated by measuring DHE (dihydroethidium, D11347, Invitrogen) fluorescence. The aortic root sections were incubated with 5 μM DHE for 10 min in the dark and observed using a fluorescence microscope (Nikon, Japan). Fluorescent and grayscale images were analyzed with NIH ImageJ 1.53 software to determine the mean fluorescence density.
## Scanning electron microscopy
Aortas isolated from Apoe−/− mice for scanning electron microscope analysis were fixed in situ with $3\%$ glutaraldehyde (G5882, Sigma) and harvested. Fixed aortas were postfixed with $2\%$ osmium tetroxide (OsO4, pH 7.4) (201030, Sigma) and dehydrated with graded ethanol (30–$100\%$). Chemical dehydration was achieved by incubation of the samples with $50\%$ hexamethyldisilazane (440191, Sigma) for 20 min and an additional 20 min with fresh $100\%$ hexamethyldisilazane. Aortic samples were prepared using a standard procedure for scanning electron microscopy, and photographs were taken in a routine manner.
## High-resolution optical resolution photoacoustic microscopy (OR-PAM)
A nanosecond laser system (SPOT-10-200-532, Elforlight, Daventry, UK) was used at the primary wavelength of 532 nm with a duration of 6 ns. A single-mode optical fiber (P1-405BPM-FC-1, Thorlabs, NJ, USA) transferred the laser beam, which was collimated to 2 mm in diameter. The laser beam was focused by a doublet lens (AC254-060-A, Thorlabs, NJ, USA), reflected 45° by a custom aluminum-coated prism in the beam combiner. The focused laser beam conducted three-dimensional scanning with a MEMs scanner (OptichoMS-001, Opticho Inc., Ltd., Pohang, Korea) on the x-axis and two linear stages (L-509, Physik Instruments (PI), Karlsruhe, Germany) on the x-y axis. The laser beam was used to irradiate the sample, and an acoustic wave was generated and passed through the beam combiner. The acoustic signal was detected using a high-frequency transducer (V214-BC-RM, 50 MHz center frequency, Olympus) as the analog signal. Then, an RF amplifier (ZX60-3018G-S+, Mini-Circuit, Brooklyn, NY, USA) and a low-pass crystal filter (CLPFL-0050, 50 MHz, CRYSTEK, Fort Myers, Florida, USA) were used as preprocessing steps to clean the analog signal. A high-speed digitizer (ATS9371, AlazarTech, Pointe-Claire, QC, Canada) converted the analog signal to a digital signal. The OR-PAM data were recorded, showing the x-y location and depth information of the digital acoustic signal. The control and preprocessing steps were stepped by a LabVIEW program (National Instruments, Austin, TX, USA). The measured lateral resolution was 12 µm, and the axial resolution was 27 µm.
## Statistical analysis
For the human study, continuous variables are presented as the means ± standard deviations or medians and interquartile ranges and were compared using the unpaired t-test or the Mann–Whitney rank-sum test. Discrete variables are expressed as counts and percentages and were analyzed with Pearson’s chi-square test or Fisher’s exact test. Kaplan–Meier curves were constructed for comparison of the study endpoints between the two groups, and differences were assessed with the log-rank test. Cox proportional hazards regression was used to identify factors associated with an increased risk of vascular events. Factors associated with mortality with a p value of less than 0.20 in the univariate analysis were entered into the multivariate model, and nonsignificant factors were removed using a backward-selection procedure. All analyses were two-tailed, and all variables were considered significant at the level of $p \leq 0.05.$ All statistical analyses were performed using SPSS for Windows ver. 26.0 (IBM SPSS Inc., Chicago, IL, USA).
Differences between the experimental and control groups in the animal study were analyzed by Student’s t-test or one-way ANOVA with Bonferroni’s multiple-comparisons test.
All statistical analyses were performed using SPSS for Windows ver. 26.0 (IBM SPSS Inc., Chicago, IL, USA). A P value less than 0.05 was considered significant.
## ANGPTL4 administration attenuates the progression of atherosclerotic plaques in atherosclerotic mice
To determine the effect of ANGPTL4 administration on atherosclerotic lesion development, Apoe−/− mice fed a high-fat diet were injected with 2 μg of ANGPTL4 two times per week for 8 weeks and compared with mice injected with PBS alone. The experimental procedures are summarized in Fig. 1a. Atherosclerosis was induced by feeding a high-fat diet for 8 weeks. From the beginning of the high-fat diet, mice were intraperitoneally injected with PBS or ANGPTL4 protein twice per week. The entire aorta, aortic root, blood, BMDMs, and VSMCs were collected for further analyses. Fig. 1Effects of ANGPTL4 administration on atherosclerotic progression in the atherosclerosis Apoe−/− mouse model.a Experimental scheme. Apoe−/− mice were fed a high-fat diet (HFD) with an injection of PBS or recombinant ANGPTL4 protein (2 μg per mouse, intraperitoneally twice a week) for 8 weeks. Blood, BMDMs, aortic roots, and aortas were collected for further analyses. b Representative Oil red O-stained aortas from PBS-injected and ANGPTL4-injected Apoe−/− mice fed an HFD for 8 weeks. The relative atherosclerotic plaque area was quantified. c, *Lesion area* measured in Oil red O-stained cross sections of the aortic roots from the PBS and ANGPTL4 groups. The plaque area was measured and quantified as the relative size of the plaque to the aortic area. Scale bar, 500 μm. d Representative images are shown for H&E-stained aortic roots from the PBS and ANGPTL4 groups. The necrotic core area, qualified by the anucleated area, was measured and quantified as the relative size of the necrotic core to the plaque or aortic root. Scale bar, 500 μm. e Representative images of atherosclerotic aortas were obtained using high-resolution optical resolution photoacoustic microscopy (OR-PAM). Blue arrows indicate atherosclerotic plaques. f Scanning electron microscope images showing the aorta surface isolated from the PBS and ANGPTL4 groups. Boxed areas on the aortic surface depict atherosclerotic plaques at higher magnification. Scale bar, 20 μm. Data were presented as the mean ± SEM. # $p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$ (by Student’s t-test).
En face analysis of aortic lesions in the entire aorta and quantification of atherosclerotic lesions showed that the relative plaque size was smaller in the ANGPTL4 group than in the PBS group (Fig. 1b). Oil red O-stained aortic roots showed that both the plaque area (0.408 ± 0.087 vs. 0.315 ± 0.095 mm2, $p \leq 0.05$) and relative plaque size (30.860 ± $4.031\%$ vs. 21.989 ± $5.236\%$, $p \leq 0.001$) were smaller in the ANGPTL4 group (Fig. 1c). The necrotic core was measured through H&E staining, and the area of the necrotic core and percentage of the necrotic core of the aortic root were significantly smaller in the ANGPTL4 group than in the PBS group (11.680 ± $2.899\%$ vs. 7.622 ± $3.604\%$, $p \leq 0.05$, Fig. 1d).
The atheroprotective effect of ANGPTL4 was additively validated in an Ldlr−/− mouse atherosclerosis model. In Ldlr−/− mice fed a high-fat diet for 16 weeks, atherosclerotic lesions and necrotic core areas were smaller in the ANGPTL4 group than in the PBS group (Supplementary Fig. 1a–c). In addition to aortic roots, aortas also showed smaller atherosclerotic lesions in the ANGPTL4 group than in the PBS group (Supplementary Fig. 1d).
The plaque within the aorta was visualized using an optical resolution photoacoustic microscopy system, which showed less plaque in the ANGPTL4 group than in the PBS group (Fig. 1e). Scanning electron microscopy was used to visualize the luminal surfaces of mouse aortas. After 8 weeks of the high-fat diet, endothelial deposits and atherosclerotic plaque on the aortic surface were decreased in the ANGPTL4 group (Fig. 1f). These results indicated that ANGPTL4 administration attenuates atherosclerotic plaque progression in Apoe−/− mice.
## ANGPTL4 administration relieves inflammatory and atherogenic phenotypes
Macrophages are an important component of atherosclerotic lesions and are associated with the progression of plaques. To determine whether ANGPTL4 contributes to macrophage function and gene expression in atherosclerosis, BMDMs from PBS- or ANGPTL4-injected atherosclerotic mice were isolated, and the mRNA expression of inflammation-related markers was compared. The expression levels of inflammatory markers such as Tnfrsf11b, Tlr4, Ccl2, and Nos2 were lower, whereas that of anti-inflammatory IL-10 was higher in BMDMs isolated from the ANGPTL4 group (Fig. 2a, b). Next, the atherogenic features of macrophages were examined using Oil red O staining and a fluorescence oxidized LDL uptake assay (Fig. 2c, d). Foam cell formation and oxidized LDL uptake were reduced in BMDMs of the ANGPTL4 group. Fig. 2Effects of ANGPTL4 on macrophage function in atherosclerosis.a, b Proinflammatory and anti-inflammatory gene expression were analyzed by real-time PCR in BMDMs isolated from Apoe−/− mice treated with PBS or ANGPTL4. c BMDMs from the PBS and ANGPTL4 groups were analyzed using Oil red O staining. Scale bar, 200 μm. d BMDMs from the PBS and ANGPTL4 groups were treated with oxLDL-DyLight 488 for 24 h, and then oxidized LDL uptake was analyzed. Scale bar, 200 μm. e Flow cytometry analyses of single-cell aortic suspensions isolated from the PBS and ANGPTL4 groups. Inflammatory macrophages were quantified by the number of CD80+ cells among the CD45+F$\frac{4}{80}$+CD11b+ population. f, g The content of macrophages in aortic root sections from the two groups was determined by immunohistochemical staining with anti-CD68 antibody (f) and anti-Mac2 antibodies (g). Representative images are shown, and CD68-positive areas (f) and Mac2-positive areas (g) were measured and quantified as a percentage of the plaque area. Data were presented as the mean ± SEM. # $p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$, ####$p \leq 0.0001$ (by Student’s t-test).
To assess the macrophage phenotype, we isolated single cells from the entire aorta. We observed that the numbers of CD11b+F$\frac{4}{80}$+ macrophages and CD80+ proinflammatory macrophages isolated from the ANGPTL4 group were significantly reduced compared with those isolated from the PBS group (Fig. 2e). Next, we assessed the distribution of macrophages in the plaque lesions by immunohistochemical staining. The area of CD68+ macrophages (0.173 ± 0.044 vs. 0.093 ± 0.039 mm2, $p \leq 0.01$) and Mac2+ macrophages (0.15 ± 0.041 vs. 0.088 ± 0.03 mm2, $p \leq 0.01$) and the percentage of CD68+ macrophages (40.311 ± $5.382\%$ vs. 29.122 ± $5.767\%$, $p \leq 0.01$) and Mac2+ macrophages (32.59 ± $5.575\%$ vs. 24.92 ± $6.722\%$, $p \leq 0.05$) within plaques were decreased in the ANGPTL4 group (Fig. 2f, g). Taken together, these data demonstrate that ANGPTL4 significantly reduced the macrophage inflammatory response and lipid accumulation.
## ANGPTL4 administration attenuates atherogenic mediators in Apoe−/− mice
To investigate how ANGPTL4 affects the atherosclerotic response in cells, we studied the mRNA expression in the entire aorta using real-time PCR. The modulation of VSMCs from the contractile phenotype to the synthetic/proliferative phenotype is a critical step in the pathogenesis of vascular diseases20,21. Contractile VSMC markers, such as Myh9, Myh11, Smtn, Tagln, Acta2, and Cnn1, were significantly upregulated in the ANGPTL4 group (Fig. 3a). SMCs express cell transition markers22,23 and lose the expression of contractile markers in pathological lesions18. We found that the mRNA expression of proinflammatory markers such as Icam-1, Vcam-1, Tnfrsf11b, and Tlr4 was significantly lower in the ANGPTL4 group than in the PBS group (Fig. 3b). The expression levels of macrophage markers such as CD68 and Lgals3 were decreased by ANGPTL4 administration (Fig. 3c). Moreover, we compared the ANGPTL4 mRNA levels in aortic tissues and aortic SMCs isolated from atherosclerotic mice administered PBS or ANGPTL4 (Fig. 3d, e). Additionally, the effect of ANGPTL4 treatment on mouse SMCs and human aortic SMCs was analyzed (Fig. 3f, g). Collectively, ANGPTL4 mRNA was markedly increased by ANGPTL4 treatment in aortic tissues and aortic SMCs (Fig. 3d–g). Similarly, we examined the systemic effects of ANGPTL4 administration on circulating inflammatory cytokines in the plasma of Apoe−/− mice. Elevated levels of circulating leptin, IL-6, IL-1β, and IL-18 were profoundly reduced in the ANGPTL4 group (Fig. 3h). The circulating ANGPTL4 levels were not significantly different between the two groups (Supplementary Fig. 2). These results demonstrate that ANGPTL4 reduces mediators of vascular inflammation and supports a contractile phenotype in SMCs. Fig. 3The expression patterns of atherogenic mediators from the aortas and plasma of Apoe−/− mice. Relative expression of genes related to contractility (a), proinflammation (b), and macrophage markers (c) in aortas isolated from the PBS and ANGPTL4 groups. Apoe−/− mice fed a high-fat diet (HFD) were injected with PBS or ANGPTL4 twice a week for 8 weeks (2 μg, i.p.). ANGPTL4 expression in aortas (d) and aortic SMCs (e, f) of Apoe−/− mice and human aortic SMCs (g). Aortic SMCs were stimulated with cholesterol (10 μg/ml) or TNFα (100 ng/ml) and oxLDL (10 μg/ml) with or without ANGPTL4. h Circulating leptin, IL-6, IL-1β, and IL-18 were quantified in the PBS and ANGPTL4 groups. Data were presented as the mean ± SEM. # $p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$, ####$p \leq 0.0001$ (by Student’s t-test or one-way ANOVA with Bonferroni’s multiple-comparisons test).
## ANGPTL4 treatment attenuates the proliferation of VSMCs
Because the migration and proliferation of VSMCs contribute to atherosclerosis and cardiovascular disease24, we examined the effect of ANGPTL4 treatment on the phenotypes of stimulated VSMCs. We assayed cell proliferation by the BrdU incorporation assay and showed that ANGPTL4 treatment substantially inhibited proliferation in human aortic SMCs stimulated with either FBS or platelet-derived growth factor-BB (PDGF-BB) (Supplementary Fig. 3a). Ki67 staining also showed that ANGPTL4 treatment inhibited proliferation in human aortic SMCs. The numbers of Ki67-positive cells were lower in both PDGF-BB- and FBS-stimulated human aortic SMCs after ANGPTL4 treatment (Supplementary Fig. 3b).
## ANGPTL4 administration stabilizes atherosclerotic plaques
Because systemic ANGPTL4 injection may affect more than the inflammatory status, we further examined whether plaque stability, a key factor during the pathogenesis of atherosclerosis, may be causally involved in the alleviation of atherosclerosis. Most importantly, the measurement of fibrous cap thickness revealed that the atherosclerotic lesions of the ANGPTL4-treated Apoe−/− mice had smaller necrotic cores (127.260 ± 23.882 vs. 102.371 ± 23.348 μm, $p \leq 0.05$) and thicker fibrous caps (28.052 ± 7.164 vs. 49.60 ± 11.506 μm, $p \leq 0.001$), indicating advanced signs of plaque stability compared with those of the PBS-treated mice (Fig. 4a). Importantly, the fibrous cap thickness was 1.77-fold greater in the lesions of ANGPTL4-treated mice than in those of PBS-treated mice. In an Ldlr−/− atherosclerosis model, the fibrous cap of the aortic roots and aorta was also thicker in the ANGPTL4 group than in the PBS group (Supplementary Fig. 4a, b). Because collagen plays a key role in determining plaque stability25, we also analyzed collagen deposition in the atherosclerotic plaques using picrosirius red staining. Intraplaque collagen levels were significantly higher in the ANGPTL4 group than in the PBS group in aortic root lesions (31.71 ± 5.297 vs. 23.04 ± 5.886, $p \leq 0.001$), suggesting that ANGPTL4 can increase the stability of atherosclerotic plaques (Fig. 4b).Fig. 4Effects of ANGPTL4 on the stability of atherosclerotic plaques.a–e Representative histologic analysis of the aortic root from the PBS and ANGPTL4 groups. a Representative images of Masson trichrome staining are shown. Boxed areas in the aortic root depict the fibrous cap and necrotic core at higher magnification, and the fibrous cap and necrotic core were measured as the lesion thickness. Scale bar, 100 μm. b Representative images of picrosirius red staining for collagen, and quantification of the collagen content presented as a percentage of the plaque area. Scale bar, 100 μm. Immunofluorescence staining of atherosclerotic plaques showed α-SMA and CD68 (c), SM22α (d), and SM-MHC (e) in the fibrous caps. Quantification of the fibrous cap thickness is presented in the right panels. Scale bar, 20 μm. f H&E (left panels) and SM-MHC and SM22α-stained confocal images of lesions representing preatheromatous plaques and complicated lesions of the human LAD. Scale bar, 10 μm. Data were presented as the mean ± SEM. # $p \leq 0.05$, ###$p \leq 0.001$ (by Student’s t-test). LAD left anterior descending artery.
Because plaque rupture is inversely correlated with the number of contractile VSMCs26, we assessed the VSMC marker proteins α-SMA, Sm22α (transgelin), and SM-MHC in atherosclerotic plaques by immunohistochemical staining (Fig. 4c–e). Compared with that in the PBS group, the fibrous cap of aortic roots with ANGPTL4 had a marked increase in the α-SMA+ plaque area (1.43-fold), Sm22α+ plaque area (1.49-fold), and SM-MHC+ plaque area (1.45-fold), which indicated enhanced features of stability. We next examined SMC (SM-MHC and SM22α) expression in lesions with different degrees of human atherosclerotic disease burden. Immunostaining of serial left anterior descending arteries (LADs) for SM-MHC and SM22α revealed decreased expression of contractile SMC markers in the media and intima of complicated lesions compared with preatheromatous plaques of the human LAD (Fig. 4f). Collectively, these data demonstrate that ANGPTL4 increased the thickness of the fibrous cap in atherosclerotic lesions, contributing to atherosclerotic plaque stability.
## ANGPTL4 regulates SMC phenotypic changes in atherosclerotic lesions
To assess the characteristics of the plaque, we stained the atherosclerotic lesions for macrophages and SMCs. In atherosclerotic Apoe−/− mice, ANGPTL4 administration reduced the expression of the macrophage marker CD68 in many α-SMA+ cells in the plaque and adjacent media of the atherosclerotic aortic roots compared with PBS administration (Fig. 5a). To test whether ANGPTL4 is directly involved in VSMC-derived foam cell formation, we stained plaques with α-SMA and performed lipid staining with fluorescent BODIPY, which is commonly used to fluorescently stain neutral lipids. The ANGPTL4 group had less lipid deposition in SMC+ cells in the plaque relative to the PBS group, suggesting that ANGPTL4 effectively reduces the size of the lipid core and VSMC-derived foam cells in advanced atherosclerosis (Fig. 5b).Fig. 5ANGPTL4 regulates SMC phenotypic changes in atherosclerosis.a Immunofluorescence staining of atherosclerotic plaques showing CD68 (green) and α-SMA (red) in the aortic root. Boxed areas show close-up images of CD68+α-SMA+ cells (arrowheads) in atherosclerotic plaques. Quantification of the frequency of double-positive (CD68+α-SMA+) cells among the total α-SMA+ cells within the whole lesion and the fibrous cap ($$n = 16$$). Scale bar, 20 μm. b Representative images of atherosclerotic plaques in the aortic root showing lipid droplets stained by BODIPY (green) and α-SMA (red). Arrowheads indicate BODIPY+αSMA+ cells. The percentage of BODIPY+α-SMA+ cells within the plaque area ($$n = 11$$). Scale bar, 20 μm. c Aortic SMCs were isolated from atherosclerotic Apoe−/− mice treated with PBS or ANGPTL4 and then stained with α-SMA as an SMC marker and CD68 as a macrophage marker. Quantification of α-SMA/CD68 fluorescence intensity. Scale bar, 100 μm. d Aortic SMCs were stimulated with cholesterol (10 μg/ml) for 72 h with or without ANGPTL4 and stained with α-SMA and CD68. Scale bar, 100 μm. e In human LAD, the atherosclerotic lesion displayed cells double positive for α-SMA and CD68. Scale bar, 10 μm. Data were presented as the mean ± SEM. # $p \leq 0.05$, ###$p \leq 0.001$, ####$p \leq 0.0001$ (by Student’s t-test or one-way ANOVA with Bonferroni’s multiple-comparisons test).
In SMCs isolated from atherosclerotic Apoe-/- mice, the level of α-SMA was conserved, whereas CD68 was not induced in the ANGPTL4 group (Fig. 5c). SMCs isolated from mouse aortas were stimulated with cyclodextrin–cholesterol complexes with or without ANGPTL4 and then stained with α−SMA and CD68 to confirm their cell type identity (Fig. 5d). Cholesterol-loaded SMCs also showed a significant decrease in α-SMA expression, which was significantly recovered by ANGPTL4 treatment. In contrast, CD68 expression induced by cholesterol was attenuated by ANGPTL4 treatment. Furthermore, ANGPTL4 treatment significantly decreased oxidized LDL uptake and the number of cholesterol-overloaded foam cells among SMCs, indicating that ANGPTL4 reduces lipid accumulation by inhibiting oxidized LDL uptake (Supplementary Fig. 5a, b). As shown in Supplementary Fig. 5c, ANGPTL4 rescued the attenuation of mRNA levels for contractile markers caused by cholesterol in SMCs isolated from mouse aortas. In contrast, the elevated expression of macrophage markers mediated by cholesterol was significantly decreased in ANGPTL4-treated SMCs (Supplementary Fig. 5d). More importantly, cells staining positive for both α-SMA and CD68 were frequently observed in advanced atherosclerotic plaques of human LAD (Fig. 5e). These data indicate that phenotypic changes in SMCs are an ongoing process throughout plaque pathogenesis and that ANGPTL4 plays a key role in regulating VSMC phenotypic modulation.
## ANGPTL4 downregulates KLF4 expression in SMCs in advanced atherosclerotic plaques
KLF4 is a critical transcription factor for the phenotypic changes that occur as VSMCs transition into macrophage-like cells and is upregulated during plaque instability18. We investigated whether ANGPTL4 administration downregulates KLF4. We found that KLF4 expression was significantly decreased in both the plaque area and the media layer of the aortic roots in the ANGPTL4 group relative to the PBS group (Fig. 6a). Additionally, KLF4 mRNA and protein in the aorta were also decreased in the ANGPTL4 group compared with the PBS group (Fig. 6b, c). As shown in Fig. 6d and Supplementary Fig. 6a, KLF4 expression was markedly inhibited by ANGPTL4 treatment in cholesterol-stimulated SMCs isolated from Apoe−/− mice. Similarly, KLF4 expression was suppressed by treatment with ANGPTL4 in human aortic SMCs and human iPSC-SMCs (Supplementary Fig. 6b). In human aortic SMCs stimulated by TNFα and oxLDL, KLF4 promoter activity was also induced but inhibited by ANGPTL4 treatment (Fig. 6e). Interestingly, KLF4 was substantially upregulated in α-SMA+ cells from complicated lesions of human LAD (Fig. 6f). These results indicate that ANGPTL4 downregulates KLF4 expression and plays an important role in regulating VSMC phenotypic changes and plaque instability. Taken together, these data show that ANGPTL4 interferes with plaque development possibly through a reduction in inflammatory factors and cellular changes related to plaque dynamics. Fig. 6Attenuation of KLF4 upregulation by ANGPTL4 in atherosclerosis.a Expression of KLF4 within the plaque of the aortic root from the two groups was determined by immunofluorescent staining. Data were the mean fluorescence intensity (MFI) of KLF4 (mean ± SEM, $$n = 8$$). Scale bar, 20 μm. b, c Aortic SMCs were isolated from atherosclerotic Apoe−/− mice treated with PBS or ANGPTL4, and the level of KLF4 was measured. The levels of KLF4 mRNA (b) and protein (c) were lower in the ANGPTL4 group than in the PBS group. d Western blotting of KLF4 expression in SMCs from Apoe−/− mice pretreated with ANGPTL4 for 24 h and then stimulated with cholesterol (10 μg/ml) for 72 h. e KLF4 promoter activity was increased in human aortic SMCs stimulated by oxLDL (10 μg/ml) or oxLDL and TNFα (100 ng/ml) for 24 h but was inhibited by ANGPTL4 (1, 5 μg/ml) treatment. f Representative images of α-SMA+KLF4+ staining in preatheromatous plaques and complicated lesions of the human LAD. Scale bar, 10 μm. Data were presented as the mean ± SEM. # $p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$ (by Student’s t-test or one-way ANOVA with Bonferroni’s multiple-comparisons test).
## ANGPTL4 modulates SMC phenotypic changes through KLF4 induction by NOX1
NOX1-dependent ROS generation is required for VSMC proliferation and migration after vascular injury. Furthermore, the phenotypic change of SMCs to macrophage-like cells is induced by NOXA1-dependent NOX1 activation of KLF4 in atherosclerotic lesions27. Thus, we studied whether ANGPTL4 regulates NOX1 activation in KLF4-induced SMC phenotypic changes. The *Nox1* gene, but not Nox4 (Supplementary Fig. 7), was induced by cholesterol administration in VSMCs, whereas ANGPTL4 treatment blocked Nox1 induction along with KLF4 (Fig. 7a). Moreover, in an atherosclerosis mouse model, Nox1 and KLF4 were significantly downregulated in the ANGPTL4 group (Fig. 7b). To determine whether the protective effect of ANGPTL4 contributes to reduced ROS, intracellular ROS levels were determined by measuring dihydroethidium (DHE) fluorescence (Fig. 7c) and that of the chloromethyl derivative of dichlorodihydrofluorescein diacetate (CM-H2DCFDA), an oxidant-sensitive dye (Fig. 7d). The ROS levels in the aortic sinus atherosclerotic lesions were significantly lower in the ANGPTL4 group than in the PBS group (Fig. 7c). ROS levels determined by CM-H2DCFDA fluorescence were increased in aortic SMCs treated with TNF-α. Interestingly, ROS generation was markedly reduced in response to ANGPTL4 (Fig. 7d). Human aortic SMCs were stimulated with TNFα and oxLDL with or without ML171, a pharmacological inhibitor of NOX1, and then stained with KLF4, αSMA, and CD68 to confirm their cell type identity. Notably, KLF4 expression was markedly inhibited by ML171 treatment in human aortic SMCs (Fig. 7e). SMCs stimulated with TNFα and oxLDL showed a significant decrease in α-SMA expression, which was significantly recovered by ML171 treatment. In contrast, the expression of CD68 induced by TNFα and oxLDL was attenuated by ML171 treatment (Fig. 7f). Collectively, these findings suggest that ANGPTL4 is a key regulator of NOX1 activation of KLF4-induced SMC phenotypic changes in atherosclerotic plaques (Fig. 7g).Fig. 7ANGPTL4 modulates SMC phenotypic changes through KLF4 induction by NOX1.a Aortic SMCs pretreated with ANGPTL4 were stimulated with cholesterol (10 μg/ml) or oxLDL (10 μg/ml) and TNFα (100 ng/ml), and the levels of Nox1 and KLF4 were measured. b Aortas were isolated from atherosclerotic Apoe−/− mice treated with PBS or ANGPTL4, and the levels of Nox1 and KLF4 were measured. c Representative images and quantification of dihydroethidium (DHE) fluorescence in aortic root sections. The aortic root sections were incubated with 5 μM DHE for 10 min in the dark. The data presented are the MFI of DHE (mean ± SEM, $$n = 9$$). Scale bar, 100 μm. d Representative images of CM-H2DCFDA staining. Aortic SMCs from atherosclerotic Apoe−/− mice were pretreated with ANGPTL4 and TNFα and exposed to 13 μM CM-H2DCFDA. Quantification of ROS levels by MFI. Scale bar, 100 μm. e, f Human aortic SMCs were stimulated with oxLDL (10 μg/ml) and TNFα (100 ng/ml) with or without ML171, a NOX1 inhibitor (0.5 μg/ml, 5 μg/ml), and were stained with KLF4 (e) and α-SMA and CD68 (f). Scale bar, 100 μm. g Proposed mechanism of action for ANGPTL4 administration in atherosclerosis and plaque stabilization. Data were presented as the mean ± SEM. # $p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$, ####$p \leq 0.0001$ (by Student’s t-test or one-way ANOVA with Bonferroni’s multiple-comparisons test).
## Clinical implications of ANGPTL4
Using ELISA, we analyzed the circulating levels of ANGPTL4 in patients with MI (Supplementary Fig. 8a). The clinical characteristics of the two groups are summarized in Table 1. The patients’ mean ages were 63.4 years and 64.9 years in the low and high ANGPTL4 groups, respectively. A total of 79.6 and $76.9\%$ of patients were male in the low and high ANGPTL4 groups, respectively. In the low ANGPTL4 group, $99\%$ of patients were prescribed statins as a discharge medication, and $40.8\%$ of these patients were prescribed ticagrelor or prasugrel. In the high ANGPTL4 group, $95.2\%$ of patients were prescribed a statin, and $51\%$ were prescribed one of the newer P2Y12 inhibitors. The mean ANGPTL4 levels of the low and high ANGPTL4 groups were 0.925 ± 0.42 and 3.09 ± 2.28 ng/mL, respectively. Compared with patients with a lower level of ANGPTL4, patients with a higher level of ANGPTL4 had lower levels of triglycerides (120.28 ± 67.93 vs. 144.00 ± 83.78 mg/dL) and a lower LVEF (53.86 ± $11.19\%$ vs. 59.48 ± $9.79\%$). The sex proportions and the prevalence of atherosclerotic risk factors, such as hypertension, diabetes mellitus, dyslipidemia, and familial history of coronary artery disease or smoking, were comparable between the two groups. Evidence-based medications for MI, such as antiplatelets, beta-blockers, renin-angiotensin-aldosterone system blockers, and statins, were also similarly prescribed to patients in both groups. Given that normal triglyceride levels are less than 150 mg/mL, the average range of triglycerides in this study was not considered pathological. Reduced TG levels may be a secondary effect of the beneficial changes related to ANGPTL4.Table 1Baseline clinical characteristics of patients with ANGPTL4 levels above or below the median. FactorLow ANGPTL4 ($$n = 103$$)High ANGPTL4 ($$n = 104$$)P valueAge (years)63.4 ± 12.664.9 ± 12.30.379Men, n (%)82 (79.6)80 (76.9)0.639Systolic BP (mmHg)126.0 ± 22.6127.0 ± 24.50.614Heart rate (/min)76.0 ± 17.580.9 ± 19.30.056Current or ex-smoker, n (%)69 (67.0)60 (57.7)0.168Hypertension, n (%)55 (53.4)47 (45.2)0.238Diabetes mellitus, n (%)37 (35.9)27 (26.0)0.758Dyslipidemia, n (%)25 (24.3)18 (17.3)0.217Obesity, n (%)36 (35.0)37 (35.6)0.925Familial history of CAD, n (%)11 (10.7)7 (6.7)0.313ANGPTL4 levels (ng/mL)0.925 ± 0.423.09 ± 2.28<0.001Serum creatinine (mg/dL)1.96 ± 6.171.12 ± 0.960.517Peak troponin-I (mg/dL)29.11 ± 46.3341.65 ± 57.090.101Peak CK-MB (mg/dL)38.97 ± 69.3061.72 ± 89.530.067Total cholesterol (mg/dL)176.59 ± 39.74178.79 ± 43.320.712Triglyceride (mg/dL)144.00 ± 83.78120.28 ± 67.930.043HDL-cholesterol (mg/dL)43.00 ± 10.0742.65 ± 10.9080.830LDL-cholesterol (mg/dL)112.22 ± 31.52111.86 ± 30.910.940Serum glucose (mg/dL)145.01 ± 51.81163.40 ± 90.7530.547N-terminal pro BNP (pg/mL)2,130 ± 3,798.773,309.63 ± 7,075.200.340High-sensitivity CRP (mg/L)1.38 ± 3.311.43 ± 3.190.924Platelets (×109/L)239.74 ± 65.52273.86 ± 259.5820.199Left ventricular EF (%)59.48 ± 9.7953.86 ± 11.19<0.001Medications at discharge, n (%) Aspirin103 [100]104 [100]1.000 Clopidogrel61 (59.2)51 (49.0)0.141 Prasugrel or ticagrelor42 (40.8)53 (51.0)0.141 Beta-blocker86 (83.5)85 (81.7)0.738 ACEi or ARB96 (93.2)89 (85.6)0.075 Statin102 (99.0)99 (95.2)0.100Values are presented as the mean ± SD or number (percentage).ACEi angiotensin-converting enzyme inhibitor, ARB angiotensin-II receptor blocker, BNP brain-type natriuretic peptide, BP blood pressure, CAD coronary artery disease, CK-MB creatine kinase-myocardial band isoenzyme, CRP C-reactive protein, EF ejection fraction, HDL high-density lipoprotein, LDL low-density lipoprotein. Bold values indicates statistically significant p values less than 0.05 (<0.05).
The distribution of plasma levels of ANGPTL4 ranged from 0.065 to 17.491 ng/mL, and the median value was 1.581 ng/mL (Fig. 8a). Table 2 shows the angiographic and procedural characteristics of patients in the low and high ANGPTL4 groups. There were no significant differences in the proportion of patients with multivessel disease, ACC/AHA lesion complexity, frequencies of pre-PCI TIMI flow 0, post-PCI TIMI flow 3, or symptom-to-balloon time between the groups. Fig. 8Circulating levels of ANGPTL4 in patients with cardiovascular disease.a Distribution of plasma levels of ANGPTL4 measured in all patients ($$n = 207$$). b Plasma levels of ANGPTL4 in the low ANGPTL4 ($$n = 103$$) and high ANGPTL4 ($$n = 104$$) groups. c Kaplan‒Meier curve illustrating the vascular event incidence of patients during the follow-up period after surgery based on plasma ANGPTL4 levels above (red) and below (blue) the median value. Data were presented as the mean ± SEM. #### $p \leq 0.0001$ (by Student’s t-test).Table 2Coronary angiographic and procedural characteristics of patients with ANGPTL4 levels above or below the median. FactorLow ANGPTL4 ($$n = 103$$)High ANGPTL4 ($$n = 104$$)P valueMVD, n (%)48 (46.6)51 (49.0)0.726ACC/AHA B2/C lesion, n (%)100 (97.1)103 (99.0)0.369Pre-PCI TIMI flow grade 0, n (%)39 (37.9)46 (42.7)0.352Symptom-to-balloon time (min)7,420 ± 18,8743,956 ± 13,2260.150Post-PCI TIMI flow grade 3, n (%)101 (98.1)101 (97.1)1.000Values are presented as the mean ± SD or number (percentage).ACC American College of Cardiology, AHA American Heart Association, MVD multivessel disease, PCI percutaneous coronary intervention, TIMI thrombolysis in myocardial infarction.
The median values of plasma ANGPTL4 in the study population were 0.922 (0.065-1.569) ng/mL in the low ANGPTL4 group and 2.548 (1.581–17.491) ng/mL in the high ANGPTL4 group (Fig. 8b). Patients in the high ANGPTL4 group had a higher rate of event-free survival than those in the low ANGPTL4 group ($$p \leq 0.034$$ by the log-rank test) (Fig. 8c). Seven patients ($6.8\%$) had a vascular event during follow-up in the low ANGPTL4 group. The rate of vascular events was 9.26 per 100,000 person-years. In the high ANGPTL4 group, only one patient had a vascular event during follow-up, and the rate was 1.22 per 100,000 person-years.
In the multivariate analysis, plasma ANGPTL4 levels were independently associated with an increased risk of future vascular events, and the hazard ratio was 0.185 ($95\%$ CI, 0.044 to 0.783; $$p \leq 0.022$$). Other factors independently associated with an increased risk of vascular events were a positive family history of CAD and hs-CRP levels at admission (Table 3).Table 3Factors associated with events of plaque instability. FactorCrude hazard ratio on univariate analysis ($95\%$ CI)P valueAdjusted hazard ratio on multivariate analysis ($95\%$ CI)P valueAge (per year increase)1.015 (0.955–1.079)0.630Female1.360 (0.274–6.750)0.707Hypertension1.421 (0.339–5.959)0.631Diabetes mellitus1.104 (0.222–5.488)0.904Dyslipidemia28.650 (0.017–48008.246)0.376Obesity1.086 (0.259–4.556)0.910Smoking1.960 (0.448–7.867)0.343Family history of CAD0.199 (0.047–0.840)0.0280.002 (0.000–0.766)0.041Systolic blood pressure1.017 (0.988–1.046)0.253Heart rate0.997 (0.959–1.037)0.885Pain to balloon time0.941 (0.729–1.213)0.639MVD0.879 (0.219–3.527)0.856Culprit LAD3.657 (0.736–18.162)0.113512.950 (0.502-524347.409)0.078Creatinine0.785 (0.194–3.183)0.735ANGPTL4 levels0.422 (0.162–1.101)0.0780.199 (0.049-0.808)0.024Peak CK-MB0.996 (0.984–1.009)0.584Peak troponin-I0.973 (0.936–1.012)0.1670.863 (0.736-1.013)0.072hsCRP1.195 (1.050–1.359)0.0071.178 (1.023-1.356)0.023Initial LVEF0.999 (0.992–1.006)0.830Ticagrelor or prasugrel at discharge1.017 (0.944–1.095)0.662Beta-blocker at discharge1.878 (0.469–7.517)0.373ACEi or ARB at discharge1.294 (0.261–6.421)0.758Aldosterone antagonist at discharge0.041 (0.000–562.583)0.511Use of statin at discharge0.047 (0.000–3631445.951)0.742ACEi angiotensin-converting enzyme inhibitor, ARB angiotensin-II receptor blocker, CAD coronary artery disease, CI confidence interval, CK-MB creatine kinase-myocardial band isozyme, hsCRP high-sensitivity C-reactive protein, LAD left anterior descending, LVEF left ventricular ejection fraction, MVD multivessel disease. Bold values indicates statistically significant p values less than 0.05 (<0.05).
Furthermore, we analyzed the correlation between ANGPTL4, heart failure, and the incidence of recurrent heart failure (Supplementary Tables 4, 5). The distribution of plasma ANGPTL4 is shown in Supplementary Fig. 8b. The median values of plasma ANGPTL4 in the study population were 1.547 (0.392–2.620) ng/mL in the low ANGPTL4 group and 4.923 (2.651–13.746) ng/mL in the high ANGPTL4 group (Supplementary Fig. 8c). Similarly, plasma levels of ANGPTL4 showed a negative correlation with the incidence of recurrent heart failure (Supplementary Fig. 8d and Supplementary Table 6).
## Discussion
Current guidelines include no other drugs except statins for broad use in atherosclerotic diseases to target plaque vulnerability. However, even with high-dose statin therapy, some patients experience plaque destabilization and subsequent cardiovascular events, such as stroke and AMI. There is indeed an unmet need for antiatherosclerotic therapy beyond lipid lowering.
*Human* genetics studies of carriers of a missense E40K variant of ANGPTL4 or other inactivating ANGPTL4 mutations show that these carriers have lower levels of triglycerides28,29. ANGPTL4, which is regulated by nutritional and other metabolic states in a tissue-dependent manner, regulates many metabolic and nonmetabolic processes30–32. The most well-recognized action of ANGPTL4 is the posttranscriptional regulation of lipoprotein lipase, a function that is shared with ANGPTL3 and ANGPTL833,34.
In an atherosclerotic mouse model, transgenic overexpression of ANGPTL4 in Apoe−/− mice attenuates atherosclerosis primarily by suppressing lipid uptake in macrophages without changing the plasma levels of cholesterol or triglycerides35. Conversely, global ANGPTL4 deficiency in Ldlr−/− mice reduces atherosclerosis, whereas hematopoietic cell-specific ANGPTL4 deletion results in larger atherosclerotic plaques with enhanced foam cell formation36.
We recently identified that ANGPTL4 is copiously released from mesenchymal stem cells in response to inflammatory stimuli such as IL-1β, TNF-α, and inflammatory macrophages. In a mouse MI model, ANGPTL4 administration significantly reduces cardiac injury as well as systemic and cardiac inflammation16. To further develop clinical-translational approaches, we investigated the clinical characteristics of ANGPTL4 and validated the therapeutic efficacy of ANGPTL4 administration in an atherosclerosis mouse model. Chronic hyperlipidemia may initiate endothelial injury, resulting in endothelial permeability, enhanced leukocyte adhesion, and alterations in the expression of adhesive molecules. Vascular inflammation and plaque stability are determinative factors that contribute to the accelerated development of atherosclerotic lesions. VSMCs are key cellular components of arteries, exhibit phenotypic plasticity, and can switch from a contractile phenotype to a synthetic or proliferating phenotype in response to extracellular stimuli. In particular, the contractile phenotype of VSMCs with high expression of transgelin actively participates in suppressing NF-κB inducing kinase (NIK)-involved inflammation in VSMCs37. In our study, VSMCs underwent phenotypic modulation, and ANGPTL4-treated Apoe−/− mice showed a remarkable increase in the expression of genes related to the contractile phenotype in aortas.
Vascular stability is determined by macrophage-secreted proteases, VSMC phenotype, and the synthesis of an elastin-rich extracellular matrix. We found that plaques in ANGPTL4-treated mice had a thicker fibrous cap and a markedly smaller necrotic core. Notably, the latter traits are primary features of stable atherosclerotic plaques, which may suggest that ANGPTL4 reduces plaque size and promotes plaque stability. The fibrous cap, which comprises mostly contractile VSMCs and fibroblasts, is critical for stabilizing and protecting atherosclerotic plaques from rupture, which is a major cause of the clinical sequelae of atherosclerosis. Identification of the mediators governing lesion characteristics would allow the development of specific therapies to stabilize the atherosclerotic plaque, other than those therapies that generally suppress inflammation.
Indeed, VSMCs contribute to 30 to $70\%$ of cells expressing macrophage markers38 and lose the expression of contractile proteins such as α-SMA while expressing markers of other cell types such as CD6839. Finding a new therapy that attenuates SMC transition and maintains a contractile phenotype could be promising for treating atherosclerotic vascular diseases. SMC phenotypic switching is a key phenomenon underlying several vessel-narrowing diseases, such as atherosclerosis, and our study showed that ANGPTL4 could be a potential regulator of SMC differentiation by blocking the transition to macrophage-like cell types by downregulating KLF4.
In response to pathological conditions associated with inflammation, NOX1 plays a critical role in VSMC function40,41. As a supporting mechanism, NOXA1-dependent NADPH oxidase activity in SMCs has been shown to be highly correlated with SMC proliferation and migration, KLF4-mediated transition to a macrophage-like phenotype and plaque inflammation27. In cancer, ANGPTL4 is associated with the upregulation of NOX442 and NOX243 expression. In particular, one study described the stimulation of oncogenic ROS and anoikis resistance by ANGPTL4 through activation of NADPH oxidase44. We found that ANGPTL4 was responsible for the reduction in TNFα-induced NOX1, the major source of ROS. The anti-inflammatory effect of ANGPTL4 appears to be involved in suppressing the inflammatory responses of NOX-derived ROS that contribute to accelerated atherosclerosis. In summary, our results provide a novel role for ANGPTL4 as a strong negative regulator of NOX1/KLF4-mediated VSMC phenotypic switching.
To support our findings, we then analyzed the relationship between ANGPTL4 levels and clinical outcomes in patients with AMI and heart failure. Patients with high levels of ANGPTL4 showed a lower incidence of recurrent heart failure during follow-up. In the multivariate analysis, a family history of CAD was found to be a protective factor for future vascular events, as previously reported45. We also confirmed that the inflammation biomarker hs-CRP is an independent predictor of future vascular events. This is in line with previous studies46,47. In addition, MI patients with no-reflow had lower levels of ANGPTL4 than did patients without no-reflow (3.13 ± 4.85 vs. 11.09 ± 9.94 ng/mL; $$p \leq 0.03$$)48.
*The* genetically inactive ANGPTL4 variant, which does not inhibit lipoprotein lipase activity, was associated with lower levels of triglycerides and a lower risk of CAD in large-scale DNA sequencing and genetics studies29,46. In contrast, ANGPTL4 levels showed no correlation with plasma triglycerides. Plasma ANGPTL4 levels were significantly correlated with age and negatively correlated with HDL-cholesterol but showed no correlation with plasma triglyceride levels. The median ANGPTL4 level was 7.7 (3.2 to 232.4) ng/mL, and the triglyceride and total cholesterol levels were 158.54 ± 84.14 ng/mL and 221.58 ± 39.44 ng/mL, respectively49. Interestingly, the triglyceride-lowering E40K variant of ANGPTL4 did not influence plasma ANGPTL4 levels49,50. In patients with MI, the serum levels of ANGPTL4 on admission were 7.2 ± 8.8 ng/mL and were not significantly associated with hypercholesterolemia, diabetes, or angiographic variables48. Therefore, beneficial effects such as triglyceride lowering may be associated with ANGPTL4 activity, not with the plasma level of ANGPTL4.
Therapeutic approaches targeting inflammation are tasked with improving the clinical outcomes of cardiovascular disease. Despite the beneficial effects of corticosteroids in clinical trials, the risk-benefit ratio seems inconclusive owing to a higher incidence of cardiac rupture with impaired wound healing51. Treatment with nonsteroidal anti-inflammatory drugs is also not recommended after MI, as it could result in an increased risk of bleeding and excess thrombotic events52. The TNF-α antagonist etanercept appears to be damaging to patients with AMI because it increases platelet activation, although it reduces systemic inflammation to some extent53. Etanercept also failed to improve clinical outcomes in patients with congestive heart failure54. Compelling evidence of a role for IL-1β signaling in cardiovascular disease has also been presented. The MRC-ILA Heart Study showed that administration of anakinra, a recombinant IL-1 receptor antagonist, increases major adverse cardiovascular events, including myocardial infarction, and suggested that inhibition of IL-1 signaling increases the risk of cardiovascular events55. Moreover, IL-1β antibody treatment decreases plaque stability indices in Apoe−/− mice with advanced atherosclerosis, indicating an unexpected role of IL-1β in regulating fibrous cap formation56.
In our study, ANGPTL4 administration profoundly decreased vascular inflammation and the progression of atherosclerotic plaques and regulated the phenotypic fate of smooth muscle cells into macrophage-like cells, suggesting a protective role in stabilizing atherosclerotic plaques by downregulating the NOX1 activation of KLF4. More strikingly, we showed that the ANGPTL4 plasma levels of AMI patients during PCI were associated with future events caused by plaque instability. Altogether, therapeutic and preventive interventions capable of inducing ANGPTL4 expression may lessen plaque progression in the context of post-MI remodeling and its complications.
## Supplementary information
Supplementary Information The online version contains supplementary material available at 10.1038/s12276-023-00937-x.
## References
1. Muller JE, Tofler GH, Stone PH. **Circadian variation and triggers of onset of acute cardiovascular disease**. *Circulation* (1989) **79** 733-743. DOI: 10.1161/01.CIR.79.4.733
2. Arbab-Zadeh A, Fuster V. **From detecting the vulnerable plaque to managing the vulnerable patient: JACC state-of-the-art review**. *J. Am. Coll. Cardiol.* (2019) **74** 1582-1593. DOI: 10.1016/j.jacc.2019.07.062
3. Bentzon JF, Otsuka F, Virmani R, Falk E. **Mechanisms of plaque formation and rupture**. *Circ. Res.* (2014) **114** 1852-1866. DOI: 10.1161/CIRCRESAHA.114.302721
4. Karthikeyan VJ, Lip GY. **Statins and intra-plaque angiogenesis in carotid artery disease**. *Atherosclerosis* (2007) **192** 455-456. DOI: 10.1016/j.atherosclerosis.2007.01.018
5. Nie P. **Atorvastatin improves plaque stability in ApoE-knockout mice by regulating chemokines and chemokine receptors**. *PLoS ONE* (2014) **9** e97009. DOI: 10.1371/journal.pone.0097009
6. Hafiane A. **Vulnerable plaque, characteristics, detection, and potential therapies**. *J. Cardiovasc. Dev. Dis.* (2019) **6** 26. DOI: 10.3390/jcdd6030026
7. Nissen SE. **Effect of very high-intensity statin therapy on regression of coronary atherosclerosis: the ASTEROID trial**. *JAMA* (2006) **295** 1556-1565. DOI: 10.1001/jama.295.13.jpc60002
8. Nicholls SJ. **Statins, high-density lipoprotein cholesterol, and regression of coronary atherosclerosis**. *JAMA* (2007) **297** 499-508. DOI: 10.1001/jama.297.5.499
9. Puri R. **Non-HDL cholesterol and triglycerides: implications for coronary atheroma progression and clinical events**. *Arterioscler. Thromb. Vasc. Biol.* (2016) **36** 2220-2228. DOI: 10.1161/ATVBAHA.116.307601
10. Tang X. **The effect of statin therapy on plaque regression following acute coronary syndrome: a meta-analysis of prospective trials**. *Coron. Artery Dis.* (2016) **27** 636-649. DOI: 10.1097/MCA.0000000000000403
11. Puri R. **Long-term effects of maximally intensive statin therapy on changes in coronary atheroma composition: insights from SATURN**. *Eur. Heart J. Cardiovasc. Imaging* (2014) **15** 380-388. DOI: 10.1093/ehjci/jet251
12. Bayturan O. **Clinical predictors of plaque progression despite very low levels of low-density lipoprotein cholesterol**. *J. Am. Coll. Cardiol.* (2010) **55** 2736-2742. DOI: 10.1016/j.jacc.2010.01.050
13. Ridker PM. **Antiinflammatory therapy with canakinumab for atherosclerotic disease**. *N. Engl. J. Med.* (2017) **377** 1119-1131. DOI: 10.1056/NEJMoa1707914
14. Ridker PM. **Low-dose methotrexate for the prevention of atherosclerotic events**. *N. Engl. J. Med.* (2019) **380** 752-762. DOI: 10.1056/NEJMoa1809798
15. Tardif JC. **Efficacy and safety of low-dose colchicine after myocardial infarction**. *N. Engl. J. Med.* (2019) **381** 2497-2505. DOI: 10.1056/NEJMoa1912388
16. Cho DI. **Antiinflammatory activity of ANGPTL4 facilitates macrophage polarization to induce cardiac repair**. *JCI Insight* (2019) **4** e125437. DOI: 10.1172/jci.insight.125437
17. Bennett MR, Sinha S, Owens GK. **Vascular smooth muscle cells in atherosclerosis**. *Circ. Res.* (2016) **118** 692-702. DOI: 10.1161/CIRCRESAHA.115.306361
18. Shankman LS. **KLF4-dependent phenotypic modulation of smooth muscle cells has a key role in atherosclerotic plaque pathogenesis**. *Nat. Med.* (2015) **21** 628-637. DOI: 10.1038/nm.3866
19. Doran AC, Meller N, McNamara CA. **Role of smooth muscle cells in the initiation and early progression of atherosclerosis**. *Arterioscler. Thromb. Vasc. Biol.* (2008) **28** 812-819. DOI: 10.1161/ATVBAHA.107.159327
20. Ross R. **The pathogenesis of atherosclerosis: a perspective for the 1990s**. *Nature* (1993) **362** 801-809. DOI: 10.1038/362801a0
21. Owens GK, Kumar MS, Wamhoff BR. **Molecular regulation of vascular smooth muscle cell differentiation in development and disease**. *Physiol. Rev.* (2004) **84** 767-801. DOI: 10.1152/physrev.00041.2003
22. Alencar GF. **Stem cell pluripotency genes Klf4 and Oct4 regulate complex SMC phenotypic changes critical in late-stage atherosclerotic lesion pathogenesis**. *Circulation* (2020) **142** 2045-2059. DOI: 10.1161/CIRCULATIONAHA.120.046672
23. Pan H. **Single-cell genomics reveals a novel cell state during smooth muscle cell phenotypic switching and potential therapeutic targets for atherosclerosis in mouse and human**. *Circulation* (2020) **142** 2060-2075. DOI: 10.1161/CIRCULATIONAHA.120.048378
24. Mulvihill ER. **Atherosclerotic plaque smooth muscle cells have a distinct phenotype**. *Arterioscler. Thromb. Vasc. Biol.* (2004) **24** 1283-1289. DOI: 10.1161/01.ATV.0000132401.12275.0c
25. Nadkarni SK, Bouma BE, de Boer J, Tearney GJ. **Evaluation of collagen in atherosclerotic plaques: the use of two coherent laser-based imaging methods**. *Lasers Med. Sci.* (2009) **24** 439-445. DOI: 10.1007/s10103-007-0535-x
26. Davies MJ, Richardson PD, Woolf N, Katz DR, Mann J. **Risk of thrombosis in human atherosclerotic plaques: role of extracellular lipid, macrophage, and smooth muscle cell content**. *Br. Heart J.* (1993) **69** 377-381. DOI: 10.1136/hrt.69.5.377
27. Vendrov AE. **NOXA1-dependent NADPH oxidase regulates redox signaling and phenotype of vascular smooth muscle cell during atherogenesis**. *Redox Biol.* (2019) **21** 101063. DOI: 10.1016/j.redox.2018.11.021
28. Stitziel NO. **Coding variation in ANGPTL4, LPL, and SVEP1 and the risk of coronary disease**. *N. Engl. J. Med.* (2016) **374** 1134-1144. DOI: 10.1056/NEJMoa1507652
29. Dewey FE. **Inactivating variants in ANGPTL4 and risk of coronary artery disease**. *N. Engl. J. Med.* (2016) **374** 1123-1133. DOI: 10.1056/NEJMoa1510926
30. Aryal B, Price NL, Suarez Y, Fernandez-Hernando C. **ANGPTL4 in metabolic and cardiovascular disease**. *Trends Mol. Med.* (2019) **25** 723-734. DOI: 10.1016/j.molmed.2019.05.010
31. Hato T, Tabata M, Oike Y. **The role of angiopoietin-like proteins in angiogenesis and metabolism**. *Trends Cardiovasc. Med.* (2008) **18** 6-14. DOI: 10.1016/j.tcm.2007.10.003
32. Santulli G. **Angiopoietin-like proteins: a comprehensive look**. *Front. Endocrinol.* (2014) **5** 4. DOI: 10.3389/fendo.2014.00004
33. Koishi R. **Angptl3 regulates lipid metabolism in mice**. *Nat. Genet.* (2002) **30** 151-157. DOI: 10.1038/ng814
34. Shimizugawa T. **ANGPTL3 decreases very low density lipoprotein triglyceride clearance by inhibition of lipoprotein lipase**. *J. Biol. Chem.* (2002) **277** 33742-33748. DOI: 10.1074/jbc.M203215200
35. Georgiadi A. **Overexpression of angiopoietin-like protein 4 protects against atherosclerosis development**. *Arterioscler. Thromb. Vasc. Biol.* (2013) **33** 1529-1537. DOI: 10.1161/ATVBAHA.113.301698
36. Aryal B. **ANGPTL4 deficiency in haematopoietic cells promotes monocyte expansion and atherosclerosis progression**. *Nat. Commun.* (2016) **7** 12313. DOI: 10.1038/ncomms12313
37. Dai X. **SM22alpha suppresses cytokine-induced inflammation and the transcription of NF-kappaB inducing kinase (Nik) by modulating SRF transcriptional activity in vascular smooth muscle cells**. *PLoS ONE* (2017) **12** e0190191. DOI: 10.1371/journal.pone.0190191
38. Chappell J. **Extensive proliferation of a subset of differentiated, yet plastic, medial vascular smooth muscle cells contributes to neointimal formation in mouse injury and atherosclerosis models**. *Circ. Res.* (2016) **119** 1313-1323. DOI: 10.1161/CIRCRESAHA.116.309799
39. Feil S. **Transdifferentiation of vascular smooth muscle cells to macrophage-like cells during atherogenesis**. *Circ. Res.* (2014) **115** 662-667. DOI: 10.1161/CIRCRESAHA.115.304634
40. Szöcs K. **Upregulation of Nox-based NAD (P) H oxidases in restenosis after carotid injury**. *Arterioscler. Thromb. Vasc. Biol.* (2002) **22** 21-27. DOI: 10.1161/hq0102.102189
41. Xu S. **Increased expression of Nox1 in neointimal smooth muscle cells promotes activation of matrix metalloproteinase-9**. *J. Vasc. Res.* (2012) **49** 242-248. DOI: 10.1159/000332958
42. Shen C-J. **Oleic acid-induced NOX4 is dependent on ANGPTL4 expression to promote human colorectal cancer metastasis**. *Theranostics* (2020) **10** 7083. DOI: 10.7150/thno.44744
43. Yang W-H. **A TAZ–ANGPTL4–NOX2 axis regulates ferroptotic cell death and chemoresistance in epithelial ovarian cancerTAZ promotes ferroptosis in OvCa**. *Mol. Cancer Res.* (2020) **18** 79-90. DOI: 10.1158/1541-7786.MCR-19-0691
44. Zhu P. **Angiopoietin-like 4 protein elevates the prosurvival intracellular O2−: H2O2 ratio and confers anoikis resistance to tumors**. *Cancer Cell* (2011) **19** 401-415. DOI: 10.1016/j.ccr.2011.01.018
45. Harpaz D, Behar S, Rozenman Y, Boyko V, Gottlieb S. **Family history of coronary artery disease and prognosis after first acute myocardial infarction in a national survey**. *Cardiology* (2004) **102** 140-146. DOI: 10.1159/000080481
46. Ridker PM, Cushman M, Stampfer MJ, Tracy RP, Hennekens CH. **Inflammation, aspirin, and the risk of cardiovascular disease in apparently healthy men**. *N. Engl. J. Med.* (1997) **336** 973-979. DOI: 10.1056/NEJM199704033361401
47. Ridker PM, Hennekens CH, Buring JE, Rifai N. **C-reactive protein and other markers of inflammation in the prediction of cardiovascular disease in women**. *N. Engl. J. Med.* (2000) **342** 836-843. DOI: 10.1056/NEJM200003233421202
48. Bouleti C. **Angiopoietin-like 4 serum levels on admission for acute myocardial infarction are associated with no-reflow**. *Int. J. Cardiol.* (2015) **187** 511-516. DOI: 10.1016/j.ijcard.2015.03.263
49. Smart-Halajko MC. **The relationship between plasma angiopoietin-like protein 4 levels, angiopoietin-like protein 4 genotype, and coronary heart disease risk**. *Arterioscler. Thromb. Vasc. Biol.* (2010) **30** 2277-2282. DOI: 10.1161/ATVBAHA.110.212209
50. Gusarova V. **Genetic inactivation of ANGPTL4 improves glucose homeostasis and is associated with reduced risk of diabetes**. *Nat. Commun.* (2018) **9** 2252. DOI: 10.1038/s41467-018-04611-z
51. Giugliano GR, Giugliano RP, Gibson CM, Kuntz RE. **Meta-analysis of corticosteroid treatment in acute myocardial infarction**. *Am. J. Cardiol.* (2003) **91** 1055-1059. DOI: 10.1016/S0002-9149(03)00148-6
52. Schjerning Olsen AM. **Association of NSAID use with risk of bleeding and cardiovascular events in patients receiving antithrombotic therapy after myocardial infarction**. *JAMA* (2015) **313** 805-814. DOI: 10.1001/jama.2015.0809
53. Padfield GJ. **Cardiovascular effects of tumour necrosis factor alpha antagonism in patients with acute myocardial infarction: a first in human study**. *Heart* (2013) **99** 1330-1335. DOI: 10.1136/heartjnl-2013-303648
54. Mann DL. **Targeted anticytokine therapy in patients with chronic heart failure: results of the Randomized Etanercept Worldwide Evaluation (RENEWAL)**. *Circulation* (2004) **109** 1594-1602. DOI: 10.1161/01.CIR.0000124490.27666.B2
55. Morton AC. **The effect of interleukin-1 receptor antagonist therapy on markers of inflammation in non-ST elevation acute coronary syndromes: the MRC-ILA Heart Study**. *Eur. Heart J.* (2015) **36** 377-384. DOI: 10.1093/eurheartj/ehu272
56. Gomez D. **Interleukin-1beta has atheroprotective effects in advanced atherosclerotic lesions of mice**. *Nat. Med.* (2018) **24** 1418-1429. DOI: 10.1038/s41591-018-0124-5
|
---
title: Multidimensional variability in ecological assessments predicts two clusters
of suicidal patients
authors:
- Pablo Bonilla-Escribano
- David Ramírez
- Enrique Baca-García
- Philippe Courtet
- Antonio Artés-Rodríguez
- Jorge López-Castromán
journal: Scientific Reports
year: 2023
pmcid: PMC9981613
doi: 10.1038/s41598-023-30085-1
license: CC BY 4.0
---
# Multidimensional variability in ecological assessments predicts two clusters of suicidal patients
## Abstract
The variability of suicidal thoughts and other clinical factors during follow-up has emerged as a promising phenotype to identify vulnerable patients through Ecological Momentary Assessment (EMA). In this study, we aimed to [1] identify clusters of clinical variability, and [2] examine the features associated with high variability. We studied a set of 275 adult patients treated for a suicidal crisis in the outpatient and emergency psychiatric departments of five clinical centers across Spain and France. Data included a total of 48,489 answers to 32 EMA questions, as well as baseline and follow-up validated data from clinical assessments. A Gaussian Mixture Model (GMM) was used to cluster the patients according to EMA variability during follow-up along six clinical domains. We then used a random forest algorithm to identify the clinical features that can be used to predict the level of variability. The GMM confirmed that suicidal patients are best clustered in two groups with EMA data: low- and high-variability. The high-variability group showed more instability in all dimensions, particularly in social withdrawal, sleep measures, wish to live, and social support. Both clusters were separated by ten clinical features (AUC = 0.74), including depressive symptoms, cognitive instability, the intensity and frequency of passive suicidal ideation, and the occurrence of clinical events, such as suicide attempts or emergency visits during follow-up. Initiatives to follow up suicidal patients with ecological measures should take into account the existence of a high variability cluster, which could be identified before the follow-up begins.
## Introduction
Implementing secondary prevention methods with suicidal patients is one of the best public health tools we have to actually avert suicidal outcomes1,2. Suicidal patients are often assessed in the Emergency Room (ER) or outpatient clinics, and most of them are not hospitalized. Psychiatrists and psychologists modulate clinical care depending on the assessments made in clinical settings, for instance, increasing or reducing the frequency of consultations or adapting the pharmacological treatment, but we have only indirect information about the situation of the patients in their daily life. This situation has changed in the last decade with the widespread diffusion of smartphones, which allows the use of Ecological Momentary Assessment (EMA) to monitor suicidal patients routinely. EMA follow-up could eventually lead to timely Ecological Momentary Interventions (EMI) and boost prevention efforts. A good example was provided by Wang et al. in a prognostic study with 83 inpatients that completed EMA surveys of Suicidal Ideation (SI) several times per day during their hospitalization3. The real-time data of SI allowed them to predict with great accuracy the risk of post-discharge suicide attempts. However, EMA studies in suicidology have been neglected until recently4 and those that exist are limited in their sample size or duration5, mainly due to the fatigue of the patients with the use of the assessment apps and the sensitivity of the questions they answer6.
Although ecological studies of suicidal patients focused first on the intensity and duration of suicidal ideas, the finding of sharp changes in very short periods pushed the interest in measuring their variability7. One of the first studies in this area used paper-and-pencil methods and asked university students to fill up a daily battery of questionnaires for 4 weeks7,8. Their results showed an association between SI variability and a previous history of suicide attempts and suggested that SI variability in multiple attempters was independent of their mood. More recent work, based on EMA, suggests that SI variability could be a marker of risk during the follow-up of suicidal patients3,9,10. For instance, the predictive accuracy of Wang et al. ’s model, which is described above, was improved with dynamic data on SI changes3. In another paper, Oquendo et al. examined 6 weeks of discontinued EMA data of 51 depressed patients and found stable levels of SI variability for each patient over 2 years9. They also found that high variability could increase the propensity to experience SI when exposed to stressful life events. In line with this finding, impulsivity traits were associated with SI variability, but not SI intensity, in another EMA study that followed up 84 depressed patients for 10 days10.
SI variability may be associated with a particular patient profile characterized by mood or affective instability, which is defined by sudden and recurrent changes in affect. For instance, the level of affective instability predicted SI variability, independently of depression severity, in a sample of female patients with borderline personality disorder11. Affective instability has also been associated with peaks of SI in other samples such as working women12, and patients suffering from depression and anxiety13, bipolar disorder14, or psychosis15.
In this paper, we have analyzed the data of the Smartcrises study16, the largest and longest EMA study with suicidal patients to date. The EMA questions in the Smartcrises study focus on three factors that have been related to the emergence of SI (namely sleep, appetite, and social connectedness), as well as SI and suicidal risk. The choice of sleep, appetite, and social connectedness was based on their potential use as behavioral markers of depression and suicide risk, independently of cultural determinants. Sleep disturbances, particularly insomnia and nightmares, are associated with an increased risk of suicidal behavior (including SI, attempts, and suicide) across diagnoses, cultures, and age groups17. Changes in social connectedness caused by acute psychosocial stress, such as the experience of social exclusion, can act as a trigger for suicidal behavior and induce specific biological and psychological changes18. Appetite variations have been also associated with depressive symptomatology and suicidal behavior19,20.
The aim of this paper was to study the clinical profile of suicidal patients according to the variability of their EMA responses. Consequently, we clustered them according to EMA variability during the follow-up and then compared the demographic and clinical features between the clusters. We hypothesized that high variability would be associated with features of clinical severity, such as depression severity or clinical events during follow-up. The results could help to improve the design of future EMI for suicide prevention. An overview of the study is provided in Fig. 1.Figure 1Overview of the study. Notice that the main aim of this paper is to find and analyze which clinical and demographic features could be associated with each variability profile. The candidate set of clinical and demographic features (whose possible association to the variability profiles was analyzed) was obtained upon inclusion and at the end of the follow-up. Hence, some longitudinal features were considered by computing the change at those discrete time instances.
## Sample
This study analyzes a set of 275 out of 419 patients from the Smartcrises study16 that complies with all of the following requirements: [1] a complete diagnostic assessment with Mini International Neuropsychiatric Interview (MINI) version 7.0.2 is available; [2] the age and gender of the patients are known; and [3] the EMA variability can be computed in at least one domain, i.e. the participant has answered at least to three prompts in a domain (notice that EMA variability and domains are formally defined in the “Statistical analysis” section, see also Details of the patient selection process in Supplementary Material). Compared to included patients, non-included patients were more likely to be older and retired. They also showed smaller decreases in depressive symptomatology measured by the Inventory of Depressive Symptomatology (IDS), from baseline to the end of follow-up, and higher scores of psychological pain at inclusion (P values < 0.05). None of these differences persisted after correction for multiple comparisons. Refer to the Supplementary Material for the details about the patient selection process; for instance, Fig. S3 shows a Venn diagram with numbers and reasons for exclusion.
Patients were recruited from outpatient and emergency psychiatric departments in five clinical centers across Spain and France. EMA was performed with the MeMind Wellness Tracker app (available at the App Store and Google Play) as software, and the participants’ smartphones as hardware. The language of the questionnaires and the mobile application, which included all assessment scales and EMA questions, was adapted to the country. Relevant inclusion criteria include being 18 years old or older, and a clinical assessment due to a suicidal crisis in the last 7 days. Patients diagnosed with a current manic, hypomanic or mixed affective episode, or any lifetime psychotic disorder were excluded. Patients with bipolar disorder could be included only in the depressive phases of the illness and in the absence of any manic or mixed symptomatology. Signed written consent (i.e., informed word) was obtained from all participants. Ethics and privacy regulation, including, but not limited to the Declaration of Helsinki21, was followed, and all methods were carried out in accordance with relevant guidelines and regulations. The study protocol was approved by the Institutional Review Board (IRB) of Fundación Jiménez Díaz Hospital, Spain, on the 25th of June 2017 (LSRG-1-005-16), and by the Comité de Protection des Personnes Ouest IV, France, on the 3rd of July 2018 (20187-A02634-49). The study is registered as a clinical trial since October 2018 (ClinicalTrials.gov Identifier: NCT03720730).
## Study design
The Smartcrises study combines validated clinical scales with EMA to assess the health condition of the patients. The battery of self-reported and clinician-reported questionnaires was administered in clinical visits upon inclusion and after 6 months at the end of follow-up. EMA questions were presented longitudinally throughout the study and comprise several clinical domains: [1] the assessment of suicide risk; [2] the wish to live/die; the level of social connectedness, measured in two parts: [3] social support and [4] social withdrawal; [5] sleep disturbances; and [6] the level of appetite. Notice that one of the main deterrents of EMA are fatigue effects, which consist in a decrease in the response rate as time goes by due to the repetitiveness of the questions, and this may result in withdrawal from the study22,23. To tackle this problem and obtain EMA data beyond 1–2 weeks, both the frequency and the set of asked questions changed during the study. Up to 32 EMA questions were defined. Importantly, questions on wish to live/die, instead of direct questions on SI, were chosen to minimize the potentially adverse effects of repeated prompts over a long period. The wish to live/die captures passive ideas of suicide, the main outcome of the Smartcrises study, and is a proxy measure of SI24,25. Indeed, passive SI is strongly correlated with active SI and other suicidal outcomes, including suicide death, according to a recent metanalytic review26.
The EMA assessment is inspired by the Salzburg Suicide Process (SSP) questionnaire, which was designed to assess dynamic changes in suicidal risk27. Most of the questions of the SSP questionnaire are extracted from validated clinical questionnaires. Hence, as shown in Table 1, our EMA assessment compiles questions relevant for a dynamic follow-up of suicidal patients along different domains. Each domain is evaluated with questions inspired from a given validated clinical questionnaire. In particular, EMA questions include items from the Suicidal Status Form (SSF)28, the Perceived Social Support Questionnaire (PSSQ)29, the Interpersonal Needs Questionnaire (INQ)30, the Insomnia Severity Index (ISI)31, and the Council of Nutrition Appetite Questionnaire (CNAQ)32. However, only one to five of the 32 questions were randomly chosen every day to be presented to the patients. The frequency of the questions was progressively reduced based on the evolution of the suicide risk after an attempt: 4–5 prompts in the first month, 3–4 in the next 2 months and finally 1–2 prompts in the last 3 months33. It must be mentioned that the EMA questions were not selected completely at random; instead, those related to sleep and suicide were given much higher probabilities of being asked, due to their importance, and a turn-over system was used in order to avoid repetitions. This approach reduces EMA fatigue, at the expense of every set of responses for the same question being sampled non-equidistantly (non-uniformly).Table 1EMA questions. EMAQuestionDomain1I feel psychological pain (not including physical pain)Suicide risk2I feel stress (overwhelmed)3I am agitated (restlessness)4I am full of hope5I feel hate or anger toward myself6I feel hate or anger toward other people7My wish to live isWish to live8My wish to die is9I wish there were a trusted person with whom I could talk about my personal problemsSocial support10I feel like an outsider11I have the impression that important people around me want to decide what I should think and do12I wish I received more appreciation and affection from other people13I think that I contribute to the well-being of my family/friendsSocial withdrawal14I think that I contribute to the well-being of the people who are close to me15I feel disconnected from other people16Last night I had problems falling asleepSleep17Last night I had problems staying asleep18This morning I had problems because I woke up too early19Currently other people think that my sleep problems affect my quality of life20When I woke up I felt21Last night the quality of my sleep was22Today I am satisfied with my quality of sleep23I am currently worried or stressed about my sleep problems24Currently my sleep problems interfere with my daily activity25Today I feel daytime fatigue due to my sleep problems26During the last days my appetite isAppetite27During the last days I feel full after eating28During the last days I feel hungry29During the last days food tastes30Compared to when I was younger, food tastes31During the last days I eat32During the last days I feel sick or nauseated when I eatRefer to Supplementary Table S1 for the details of the domains and EMA questions.
## Data management
A total of 48,489 answers to the 32 EMA questions from the 275 patients were analyzed in this study. Those answers were appropriately transformed so that they were all expressed in the range 0 to 100; 100 indicating the worst possible mental condition of the patient, regardless of the fact that some of them were reversed worded34. This transformation was done for clarity and displaying purposes to allow visual comparison. Parenthetically, the scores (but not individual items) of the clinical questionnaires were transformed in like manner.
Besides EMA data, information from regular assessments in the Smartcrises study includes demographic and clinical data, as well as validated questionnaires. Demographic data comprised: the number of years studying since the beginning of primary school, the employment situation, the marital status, and the country of recruitment. Clinically meaningful events during follow-up (including self-harm, suicide attempts, emergency visits, or hospitalizations in psychiatry that were obtained from clinical records) were computed in a binary variable. Depressive symptoms were assessed with the clinician-reported version of the IDS35. The Columbia-suicide severity rating scale (C-SSRS)36 was used to assess recent suicidal ideas and lifetime suicidal behavior at inclusion.
In addition, other features were computed from the following questionnaires, which were all self-reported. The overall score of the following scales was obtained: the medication adherence rating scale (MARS)37, the CAGE questionnaire for alcohol addiction38, the Fagerström Test for Nicotine Dependence (FTND)39, the List of Threatening Experiences (LTE)40, and the insomnia severity index (ISI)31. The dimensions of emotional, physical, and sexual abuse or neglect were obtained from the childhood trauma questionnaire (CTQ)41. The 11th version of the Barratt Impulsiveness Scale (BIS)42 was used to measure an overall score of impulsivity, as well as the first-order factors of the scale. Further, moral and physical pain were obtained via visual analog scales (VAS)43 that were administered upon inclusion. All these instruments were administered at baseline and at the end of follow-up, except the LTE, BIS, VAS, and CTQ, which were used only at the baseline assessment.
The mean of available values for each patient, or ipsative mean44, was used to impute the missing values after accounting for reversed-scored questions. This technique is based on the fact that the items in each questionnaire (or its dimension) are related to a single psychometric construct and should be highly correlated. Unlike simply removing the incomplete data (a “complete case analysis”), this technique can help reduce potential biases45. Nonetheless, if the missing rate of any questionnaire was over $20\%$ for a given patient, the data was discarded46. Score changes, instead of last point values, were used when follow-up questionnaires were available to extract longitudinal features. Finally, marginal diagnoses representing less than $5\%$ of the patients and overall scores of questionnaires with different dimensions were not considered in order to reduce the number of features in the dataset. All data management and statistical analysis were performed using the MATLAB software47.
## Statistical analysis
EMA variability was assessed by computing the Median Absolute Deviation (MAD) of the absolute value of the successive slopes48. Refer to Supplementary Material, and Fig. S2, for a discussion about the choice of such a metric. Notice that variability was not computed for each individual EMA question directly, since some items (like those related to sleep quality) were intentionally presented more often than others. Thus, the EMA questions were grouped into six domains, namely: suicide risk, wish to live/die, social support, social withdrawal, sleep, and appetite. They correspond to the EMA questions 1 to 6, 7 and 8, 9 to 12, 13 to 15, 16 to 25, and 26 to 32, respectively, as shown in Table 1. Each domain was defined by grouping the EMA questions from a validated clinical questionnaire (for instance, sleep with the ISI or appetite with the CNAQ), with the exception of the two questions on death wish and wish to live. In this way, EMA questions from the same domain were treated as the same question, but expressed differently, and the variability was computed for each domain, containing a larger number of observations, as opposed to analyzing the questions separately. To validate the domains, the Kendall correlation coefficients49 of the EMA questions within each domain were computed to make sure that they were positively correlated.
A Gaussian Mixture Model (GMM) was used to cluster the patients based on their multidimensional variabilities (i.e., the EMA variabilities along the six domains). This provided a comprehensive analysis of the variability of the patients, which is not limited to a single clinical aspect (e.g., SI). The GMM’s inference was performed via the Expectation–Maximization (EM) algorithm, which allows to infer missing variability domains by taking the conditional expectations of the missing values given the current parameter estimates and the observed values at every maximization (a.k.a. M) step50. The optimal number of groups was determined by the Bayesian Information Criterion (BIC), which produces accurate results even in the inimical scenario of not missing-at-random (NMAR) data51. The analysis of the clinical differences between the groups is twofold. On the one hand, phenotypic profiles were compared using the two-sample Student’s t test computing the pooled estimate of the standard deviation for the numeric variables, and the Pearson’s \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\chi }^{2}$$\end{document}χ2 test of homogeneity for the categorical ones. All tests were two-tailed. P values were adjusted for multiple comparisons applying the Holm–Bonferroni method52. On the other hand, a forward selection was performed to find the optimal set of clinical features that can be used by a random forest to predict the variability group each patient belongs to.
A random forest is a machine learning algorithm that uses bagging (i.e., a two-step process that stands for “bootstrap aggregating”). In the first or “bootstrap” step, several decision trees53,54, a.k.a. weak learners, are trained over different bootstrap samples of the dataset, each one omitting roughly $36.8\%$ of the patients. Omitted patients in each tree are referred to as “out of bag”. In the second or “aggregating” step, the final classification is made taking a democratic (non-weighted) average vote amongst all the trees, thus reducing the risk of overfitting. To further prevent overfitting, diversity (decorrelation among the constituents of the random forest) was enforced by randomly selecting the clinical features that could be used for each decision split55. However, some features have significantly more missing data than others (e.g., those measuring follow-up change) and this could bias the forward selection procedure. Hence, surrogate splits were used56. In this way, if the optimal feature is missing when classifying the patients, the best surrogate feature will be used to take a decision split and to keep the movement from the root to the leaves of the trees, eventually classifying the patients without discarding partial observations. Since there are both numeric and categorical features, the selected one for each split minimized the P value of the Pearson’s \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\chi }^{2}$$\end{document}χ2 tests of independence between each proposed clinical feature and the variability groups. Unlike the standard Classification and Regression Trees (CART) algorithm, this procedure is not biased toward those features that have many levels, not underestimating the importance of the categorical ones57.
The hyperparameters of the random forest were adjusted by employing Bayesian optimization58 upon each candidate feature set analyzed during the forward selection. The performance was assessed computing the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve59 of the predictions of the out of bag patients. Notice that when making predictions for the out of bag patients, only the subset of decision trees that have not been trained using a particular set of patients is used, yielding an unbiased estimator of the true random forest performance that approaches that of leave-one-out cross-validation, but which greatly reduces the computational burden60. Details of the statistical methods are provided in the Supplementary Material.
## Sample description
The study sample of 275 patients had a mean age of 40 years with a Standard Deviation (SD) of 14 years, approximately. It was mainly composed by females ($$n = 185$$, $67.27\%$), where “n” is the number of patients. Almost half of the patients were single ($$n = 123$$), and roughly one-third ($$n = 98$$) were married or had been cohabiting for more than 6 months. Study participants reported intermediate to high educational levels, with a mean number of years studying since the beginning of primary school of 13 with a SD of 3, approximately. There was a balanced proportion of around one-third of employed and unemployed patients, and patients with an incapacity to work. The main psychiatric diagnoses were post-traumatic stress disorder ($$n = 74$$, $26.91\%$), major depressive disorder ($$n = 40$$, $14.55\%$), and binge eating disorder ($$n = 23$$, $8.36\%$). The base rates of clinically meaningful events during follow-up in the total sample ($$n = 374$$) are $6.15\%$ for SI, $1.60\%$ for self-harm, $10.43\%$ for suicide attempts, and $5.08\%$ for emergency hospitalizations.
The participants answered the EMA questions for a mean time period of approximately 130 days with a SD of 104 days. The average length of follow-up was 148 days (SD = 116) in low-variability patients and 98 days (SD = 66) in high-variability patients. The participants answered a mean number of 176 EMA questions with a SD of 161. The mean number of answered questions was 190 (SD = 183) in low-variability patients and 150 (SD = 104) in high-variability patients. The differences between the groups were statistically significant in both cases: t273 = 3.91, $P \leq 0.001$ and t273 = 2.00, $$P \leq 0.046$$, respectively. The EMA response percentage was $43.25\%$ for all the patients (maximum: $72.64\%$, minimum: $16.97\%$), $45.48\%$ (maximum: $70.53\%$, minimum: $20.48\%$) for the low-variability group, and $39.11\%$ (maximum: $83.07\%$, minimum: $7.64\%$), for the high-variability group. A detailed description of response rates through follow-up can be found in Fig. S1.
## Validation of the domains
Figure 2 succinctly shows the correlations among the EMA questions after rescaling them from 0 to 100, accounting for reversed worded questions. There is high correlation amongst them, even if they do not belong to the same domain. The only exception corresponds to the items of the CNAQ assessing appetite that correlate strongly between them but not with items in other domains. Only positive correlations are statistically significant. All questions within the same domain are significantly correlated, with the exception of EMA3, “I am agitated (restlessness)”, which does not correlate with neither EMA4, “I am full of hope”, nor EMA7, “My wish to live is”. Therefore, EMA3 is removed from the “Suicide risk” domain. EMA3 does not correlate with many other questions, but the more independent one is EMA28, “During the last days I feel hungry”, since questions about appetite have the weakest inter-domain correlations. Figure 2Correlation coefficients among the EMA questions. Correlation of the EMA questions after rescaling them from 0 to 100, where 100 represent worse health condition. Lower triangular portion: Kendall correlation coefficients; “–” for negative values. The main diagonal is intentionally empty for clarity. Upper triangular portion: P values of the hypothesis that the corresponding Kendall correlation value is 0; “*” for P values strictly lower than 0.05. The colormap on the right is used for both Kendall correlation coefficients and P values, and ranges from the overall minimum and maximum values of those quantities.
## Cluster analysis
The BIC determined two variability groups. The first one, or low-variability group, has lower mean variability values and comprises $65\%$ of the sample according to the GMM. The second group, or high-variability group, comprises the rest of the sample. The rest of the parameters are depicted in Fig. 3. Patients from both groups attain the highest mean variability values in the “Social withdrawal” domain. The highest increase in mean variability values when comparing the high- with the low-variability group are found in the “Sleep” and “Suicide risk” domains, with a 4.53- and 2.31-fold increase, respectively. With respect to the covariance in variability (not in mean values), all values are positive in the low-variability group, while there are negative variability interactions in the high-variability group (i.e., variability increases in one domain when the variability of other domain decreases and vice versa) as it happens with the “Sleep” and the rest of the domains. The analysis of the main diagonal elements of Fig. 3a,b reveals that the low-variability group is more homogenous, since it has lower variance in variability for all the domains, especially in “Suicide risk” and “Sleep”. Figure 3The variability covariances and means of the low-variability group are shown in insets (a,c), respectively. The variability covariances and means of the high-variability group are shown in insets (b,d), respectively. Recall that they measure variability, not absolute values. For simplicity, only the lower triangular and main diagonal region of the variability covariance matrices is shown. Negative covariances indicate that increased variability in one domain is associated with lower variability in the other. The colormaps are the same across the groups to allow visual comparison, and they range from the overall maximum and minimum values of those qualities. Colormaps are shown at the top of each inset.
## Cluster comparison
Table 2 shows the phenotypic profiles of the two variability groups, where no statistical difference is found after correcting for multiple comparisons. Figure 4 illustrates the ROC curve of the automatic classification of the high-variability from the low-variability group, using a random forest. The AUC is 0.74, with $95\%$ Confidence Interval (CI) [0.68, 0.78], which is considered acceptable61. The hyperparameters obtained by the Bayesian optimization are listed in Table S2 in the Supplementary Material, and the importance of the ten clinical features chosen by the forward selection to make the classification with the random forest are shown in Fig. 5. According to this selection, the most relevant features to separate the groups (i.e., the features that are most useful for the random forest classification) are the depressive symptoms upon inclusion (IDS), followed by the impulsivity factor of cognitive instability (BIS), the marital status, and the frequency of SI at inclusion (C-SSRS), as well as the change in SI frequency, intensity and control from baseline to follow-up and other clinical factors such as nicotine dependence, binge eating disorders, or the occurrence of clinically meaningful events during follow-up. Single marital status and clinically meaningful events during follow-up were more common in the high-variability group, and all the other features showed also higher scores or higher prevalence in that group (Table 2).Table 2Demographics and clinical features by cluster according to variability level. Demographics and clinical featuresLow variability groupHigh variability groupNominal P valuebNumber (% in group)aMeanSDNumber (% in group)aMeanSDAge17941.5913.559637.4315.000.020Female gender117 (65.36)68 (70.83)0.36Country of recruitment Spain149 (83.24)79 (82.29)0.84 France30 (16.76)17 (17.71)*Marital status* Married/cohabitation > 6 months69 (38.76)29 (30.53)0.0047 Separated/divorced37 (20.79)9 (9.47) Widowed5 (2.81)1 (1.05) Single67 (37.64)56 (58.95)Years of study16512.883.518612.573.370.49Employment situation Incapacity58 (32.77)26 (27.37)0.15 Unemployed52 (29.38)41 (43.16) Retired8 (4.52)4 (4.21) Employed59 (33.33)24 (25.26)Psychiatric diagnoses (MINI)c Agoraphobia8 (4.47)10 (10.42)0.057 Alcohol use disorderd6 (3.35)3 (3.13)0.92 Binge eating disorder9 (5.03)14 (14.58)0.0064 Bipolar disorderd5 (2.79)1 (1.04)0.34 Major depressive disorder30 (16.76)10 (10.42)0.15 Obsessive–compulsive disorderd5 (2.79)2 (2.08)0.72 Panic disorder16 (8.94)10 (10.42)0.69 Posttraumatic stress disorder39 (21.79)35 (36.46)0.0089 Social anxiety disorderd4 (2.23)0 (0.00)0.14 Substance use disorder (non-alcohol)d3 (1.68)2 (2.08)0.81Depression severity at inclusion (IDS)17234.5314.448938.9816.660.026Depression severity change (IDS)125− 4.6811.2855 − 6.4112.680.36Suicidal ideation rating (C-SSRS) At inclusion Controllability1182.771.83663.241.540.079 Deterrents1142.201.60652.261.610.81 Duration1152.901.40633.101.440.39 Frequency1183.081.37633.331.510.26 Intensity1223.371.45653.341.460.89 Reasons1194.031.48673.811.670.36 Change from baseline to end of follow-up Controllability530.232.0931 − 0.231.500.29 Deterrents49 − 0.161.6933 − 0.611.870.27 Duration50 − 0.441.6131 − 0.262.050.66 Frequency52 − 0.331.1329 − 0.591.960.45 Intensity59 − 0.611.4931 − 0.481.480.70 Reasons520.351.49330.212.000.72Number of lifetime suicide attempts1782.184.18961.801.770.40Psychological pain at inclusion14660.9629.118762.5327.800.69Physical pain at inclusion14546.9726.808344.9430.580.60Alcohol misuse screening at inclusion (CAGE)13716.7930.647415.5427.970.77Alcohol misuse screening change (CAGE)85 − 2.0624.1637 − 2.0327.880.99Childhood trauma (CTQ) Emotional abuse13444.5430.497442.9731.340.73 Emotional neglect13941.5626.567741.9025.530.93 Physical neglect13118.3218.197319.3521.160.72 Physical abuse12920.7825.627019.9826.400.83 Sexual abuse13319.7029.967320.0225.950.94 Score of denial16413.8223.03959.8219.370.16 Overall scored12722.4215.247122.6916.560.91Nicotine dependence at inclusion (FTND)13017.9527.107020.9429.930.47Nicotine dependence change (FTND)77 − 1.9316.1134 − 7.7820.210.11Impulsivity levels (BIS) Attention14048.6321.927851.8821.820.29 Cognitive complexity13251.0917.467449.6819.100.59 Cognitive instability13548.9723.566751.5824.130.46 Motor13239.1220.877442.3418.440.27 Perseverance12733.0115.486235.8919.000.27 Self-control impulsivity13842.3821.267447.8821.740.08 Overall scored13643.4513.317246.2513.380.15Insomnia severity at inclusion (ISI)13746.5322.308049.7624.210.32Insomnia severity change (ISI)86 − 7.6424.0441 − 9.6326.420.67Life events at inclusion (LTE)13827.8419.797823.1820.490.10Medication adherence at inclusion (MARS)13834.0219.167939.6619.510.039Medication adherence change (MARS)860.8220.4840 − 6.0123.510.10Clinically meaningful events during follow-up38 (22.22)29 (31.52)0.10aThe sum of the number of patients in the two groups may not add up to 275 due to missing data.bSince no statistical difference is found after correcting for multiple comparisons, only the nominal P values are shown.cDiagnoses representing < $1\%$ of the sample are not shown.dItems shown for completeness, but not considered neither for the classification with random forest nor for the multiple comparison correction. Nominal P values strictly smaller than 0.05 are marked in bold. Figure 4ROC curve. The AUC is 0.74, with $95\%$ CI [0.68, 0.78], estimated by taking 2000 bootstrap samples. Dashed blue line: random guessing reference. Solid red line: mean ROC curve of the automatic prediction of the high-variability group using the selected random forest and clinical features. Green area: AUC. Dashed red lines: $95\%$ confidence intervals of the ROC curve. Figure 5Importance of the ten clinical and demographic features used by the random forest to automatically discriminate patients in the low- and high-variability groups. For clarity, the y-axis is sorted in increasing importance order according to the random forest. The importance is computed by summing all changes in the impurity of the nodes from the parent to the two children thanks to a given clinical feature and its corresponding surrogate splits. Impurity is a measure of how the decisions of a node can separate patients in the low- and high-variability groups and it is measured by the Gini’s diversity index. The sum of impurity changes is normalized by the number of branch nodes.
## Discussion
EMA and, by extension, EMI are very promising methods that could transform the field of suicide prevention. In this paper, we have analyzed a large EMA sample of suicidal patients to ascertain if symptom variability was associated with a higher risk of suicidal behavior and/or with other features of clinical severity. Symptom variability defined two clear-cut groups. The high-variability group is characterized by frequent changes not only in SI as previously reported, but also in domains such as social withdrawal, social support, sleep, or appetite. Furthermore, although conventional statistics found few differences between the groups when comparing cross-sectional data (Table 2), machine learning methods allowed us to obtain a good classification performance in this complex dataset with a large number of explanatory variables. Building on prior studies3,9, clinical variables such as the severity of depressive symptoms or the frequency of SI separate patients with high and low symptom variability. Importantly, although the occurrence of clinically meaningful events during follow-up, such as suicide attempts or hospitalizations, was not different between the groups according to standard statistics ($$p \leq 0.10$$), those events were almost 10 percentual points more frequent in the high-variability group and the variable was selected by random forest amongst the best features to differentiate high-variability patients.
Extant EMA studies with suicidal patients have shown so far that: [1] SI fluctuates widely, and [2] some clinical features, such as negative affect and disturbed sleep, could be used to predict SI fluctuations in the short-term5. However, fluctuations are not restricted to SI and sleep. In our study, social withdrawal was the clinical factor that showed the largest variability in both groups, a finding that could be related to the social sensitivity of suicidal patients. Among low-variability patients, social withdrawal was the only dimension with substantial variability, suggesting that in that group mood variations are mainly externalized through social interactions. Suicidal patients are sensitive to social cues and tend to interpret them negatively62, which may lead the patients to minimize their social interactions. An example of this sensitivity can be found in a recent study in which EMA-measured psychological pain correlated with orbitofrontal activation during a social exclusion paradigm in suicide attempters, but not in affective controls18. In the same vein, a small EMA study found that SI variability correlated strongly with SI intensity, but also with the intensity and variability of depressive symptoms and changes in social connectedness63. The pattern of variability extends thus well beyond SI and the instability of highly variable patients affects their social interactions, sleep and appetite, which is consistent with the fact that almost all EMA items correlated significantly with each other across clinical dimensions (Fig. 2). Interestingly, when high-variability patients fluctuate in one clinical dimension (such as suicide risk or sleep), they can be more stable in other dimensions. This is reflected by the negative correlations in Fig. 3 and suggests that some changes occur sequentially, rather than simultaneously, in the high-variability group. In contrast, fluctuations in the low-variability group tended to occur at the same time across the six clinical dimensions.
Some of the features that were selected by the machine learning algorithm to separate high- and low-variability groups have been related to impulsivity. They include one first-order factor of the BIS (i.e., cognitive instability) which reflects intruding and racing thoughts42, but also nicotine dependence64, the diagnosis of binge eating disorder65, and being single66. Since sleep disturbances predict the onset of SI67, it is interesting to note that the construct of cognitive instability has been recently identified as a transdiagnostic symptom associated with insomnia severity68. The role of impulsivity may also be related to affective instability since they are closely related, and partially overlapping, constructs69. A recent study associated SI variability with affective instability in Borderline Personality Disorder (BPD)11. We could not verify if BPD was overrepresented in the high-variability cluster because the disorder is not included in the MINI diagnostic assessment, but since childhood trauma and gender did not separate the clusters, the possibility of BPD diagnoses being concentrated in one of them is unlikely. We also lacked data on Attention Deficit and Hyperactivity Disorder (ADHD), another diagnosis that is frequently associated with cognitive impulsivity and emotional dysregulation. Impulsive reactions may also explain why high-variability participants stopped their follow-up earlier, having answered fewer EMA questions, than low-variability ones. Importantly, the relationship between SI and impulsivity is accentuated during suicidal crises70 but impulsivity can be the target of both pharmacological and psychotherapeutic interventions.
This study is based on a large sample of suicidal patients that were followed for several months using EMA methods. It is partially limited by missing follow-up data. According to a recent systematic review, the compliance in our sample was in the low range of EMA studies with suicidal patients (44 to $90\%$), but this could be expected given the long follow-up (prior studies ranging from 4 to 60 days) and the decline of compliance rates over time71. The use of a turnover pool of questions and a decreasing number of prompts through the follow-up seems to have reduced fatigue effects since most participants were still responding EMA questions after 4 months. Traditional methods are biased toward features that have fewer missing values. The methods applied in this study, like the EM algorithm used for the inference of the GMM, or the surrogate splits of the random forest were implemented to mitigate such a bias by exploiting the inherent structure of the data. This potential bias has also been tackled by following an agnostic approach for the feature selection, in which the algorithm was free to choose the set of features used for the classification.
A second limitation concerns the complexity of the database, which included a large number of variables and time-lagged information. We used a potent classifier based on random forest methods since off-the-shelf methods failed to provide meaningful results in this high-dimensional dataset with NMAR and correlated data. For example, in EMA, high variability patients respond more at the beginning of the follow-up, but drop out earlier than low-variability ones (Fig. S1). The surrogate splits of the random forest leveraged the natural correlations of the features to deal with the missing values, making those correlations beneficial, and the random feature selection and boosting procedure aimed to obtain robust results and to prevent overfitting72. Further, the random forest approach provides, as a byproduct, an objective tool to automatically identify patients with high variability only using demographic and clinical information, providing a probability estimate of the group each patient belongs to. The ability of the classifier to distinguish between groups was good, attaining an AUC of 0.74. Finally, two points should be noted regarding the clinical severity of the sample. First, some patients were recruited after being discharged from the hospital in outpatient consultations, which could select less severely suicidal patients. Second, although the patients excluded from the analyses were fairly similar to those included, higher basal depressive symptomatology and psychological pain suggest that they could represent more severe cases.
In summary, approximately one third of suicidal patients present high SI variability and a general pattern of instability in several domains during their follow-up, which is generally shorter. This pattern can be easily detected in early stages by assessing the severity of SI and depression, as well as impulsivity traits and other factors, and it might be associated with a higher risk of clinical events during follow-up. EMA protocols should be adapted to optimize suicidal risk assessment and preventive interventions in high- and low-variability groups.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30085-1.
## References
1. Zalsman G. **Suicide prevention strategies revisited: 10-year systematic review**. *Lancet Psychiatry* (2016.0) **3** 646-659. DOI: 10.1016/S2215-0366(16)30030-X
2. Turecki G. **Suicide and suicide risk**. *Nat. Rev. Dis. Primer* (2019.0) **5** 1-22. DOI: 10.1038/s41572-019-0121-0
3. Wang SB. **A pilot study using frequent inpatient assessments of suicidal thinking to predict short-term postdischarge suicidal behavior**. *JAMA Netw. Open* (2021.0) **4** e210591. DOI: 10.1001/jamanetworkopen.2021.0591
4. Davidson CL, Anestis MD, Gutierrez PM. **Ecological momentary assessment is a neglected methodology in suicidology**. *Arch. Suicide Res.* (2017.0) **21** 1-11. DOI: 10.1080/13811118.2015.1004482
5. Sedano-Capdevila A, Porras-Segovia A, Bello HJ, Baca-García E, Barrigon ML. **Use of ecological momentary assessment to study suicidal thoughts and behavior: A systematic review**. *Curr. Psychiatry Rep.* (2021.0) **23** 41. DOI: 10.1007/s11920-021-01255-7
6. Porras-Segovia A. **Smartphone-based ecological momentary assessment (EMA) in psychiatric patients and student controls: A real-world feasibility study**. *J. Affect. Disord.* (2020.0) **274** 733-741. DOI: 10.1016/j.jad.2020.05.067
7. Witte T, Fitzpatrick K, Joinerjr T, Schmidt N. **Variability in suicidal ideation: A better predictor of suicide attempts than intensity or duration of ideation?**. *J. Affect. Disord.* (2005.0) **88** 131-136. DOI: 10.1016/j.jad.2005.05.019
8. Witte TK, Fitzpatrick KK, Warren KL, Schatschneider C, Schmidt NB. **Naturalistic evaluation of suicidal ideation: Variability and relation to attempt status**. *Behav. Res. Ther.* (2006.0) **44** 1029-1040. DOI: 10.1016/j.brat.2005.08.004
9. Oquendo MA. **Highly variable suicidal ideation: A phenotypic marker for stress induced suicide risk**. *Mol. Psychiatry.* (2020.0). DOI: 10.1038/s41380-020-0819-0
10. Hadzic A. **The association of trait impulsivity and suicidal ideation and its fluctuation in the context of the interpersonal theory of suicide**. *Compr. Psychiatry* (2020.0) **98** 152158. DOI: 10.1016/j.comppsych.2019.152158
11. Rizk MM. **Variability in suicidal ideation is associated with affective instability in suicide attempters with borderline personality disorder**. *Psychiatry* (2019.0) **82** 173-178. DOI: 10.1080/00332747.2019.1600219
12. Tian L, Yang Y, Yang H, Huebner ES. **Prevalence of suicidal ideation and its association with positive affect in working women: A day reconstruction study**. *Front. Psychol.* (2017.0) **8** 285. DOI: 10.3389/fpsyg.2017.00285
13. Bowen R, Balbuena L, Peters EM, Leuschen-Mewis C, Baetz M. **The relationship between mood instability and suicidal thoughts**. *Arch. Suicide Res.* (2015.0) **19** 161-171. DOI: 10.1080/13811118.2015.1004474
14. Ducasse D. **Affect lability predicts occurrence of suicidal ideation in bipolar patients: A two-year prospective study**. *Acta Psychiatr. Scand.* (2017.0) **135** 460-469. DOI: 10.1111/acps.12710
15. Palmier-Claus JE. **Affective instability prior to and after thoughts about self-injury in individuals with and at-risk of psychosis: A mobile phone based study**. *Arch. Suicide Res.* (2013.0) **17** 275-287. DOI: 10.1080/13811118.2013.805647
16. Berrouiguet S. **Combining mobile-health (mHealth) and artificial intelligence (AI) methods to avoid suicide attempts: The Smartcrises study protocol**. *BMC Psychiatry* (2019.0) **19** 277. DOI: 10.1186/s12888-019-2260-y
17. Lopez-Castroman J, Jaussent I, Baca-Garcia E. **Sleep disturbances and suicidal behavior**. *Behavioral Neurobiology of Suicide and Self Harm* (2020.0) 211-228
18. Olié E. **Prefrontal activity during experimental ostracism and daily psychache in suicide attempters**. *J. Affect. Disord.* (2021.0) **285** 63-68. DOI: 10.1016/j.jad.2021.01.087
19. van Velzen LS. **Risk factors for suicide attempt during outpatient care in adolescents with severe and complex depression**. *Crisis.* (2022.0). DOI: 10.1027/0227-5910/a000860
20. Trivedi MH. **Concise health risk tracking scale: A brief self-report and clinician rating of suicidal risk**. *J. Clin. Psychiatry* (2011.0) **72** 757-764. DOI: 10.4088/JCP.11m06837
21. **World Medical Association Declaration of Helsinki. Ethical principles for medical research involving human subjects**. *Bull. World Health Organ.* (2001.0) **79** 373-374. PMID: 11357217
22. Moitra E, Gaudiano BA, Davis CH, Ben-Zeev D. **Feasibility and acceptability of post-hospitalization ecological momentary assessment in patients with psychotic-spectrum disorders**. *Compr. Psychiatry* (2017.0) **74** 204-213. DOI: 10.1016/j.comppsych.2017.01.018
23. Glenn CR. **Feasibility and acceptability of ecological momentary assessment with high-risk suicidal adolescents following acute psychiatric care**. *J. Clin. Child Adolesc. Psychol.* (2022.0) **51** 32-48. DOI: 10.1080/15374416.2020.1741377
24. Baca-Garcia E. **Estimating risk for suicide attempt: Are we asking the right questions?: Passive suicidal ideation as a marker for suicidal behavior**. *J. Affect. Disord.* (2011.0) **134** 327-332. DOI: 10.1016/j.jad.2011.06.026
25. Porras-Segovia A. **Disturbed sleep as a clinical marker of wish to die: A smartphone monitoring study over three months of observation**. *J. Affect. Disord.* (2021.0) **286** 330-337. DOI: 10.1016/j.jad.2021.02.059
26. Liu RT, Bettis AH, Burke TA. **Characterizing the phenomenology of passive suicidal ideation: A meta-analysis of its prevalence, psychiatric comorbidity, correlates, and comparisons with active suicidal ideation**. *Psychol. Med.* (2020.0) **50** 367-383. DOI: 10.1017/S003329171900391X
27. Fartacek C, Schiepek G, Kunrath S, Fartacek R, Plöderl M. **Real-time monitoring of non-linear suicidal dynamics: Methodology and a demonstrative case report**. *Front. Psychol.* (2016.0) **7** 130. DOI: 10.3389/fpsyg.2016.00130
28. Jobes DA, Linehan MM. *Managing Suicidal Risk: A Collaborative Approach* (2016.0)
29. Zimet GD, Dahlem NW, Zimet SG, Farley GK. **The multidimensional scale of perceived social support**. *J. Pers. Assess.* (1988.0) **52** 30-41. DOI: 10.1207/s15327752jpa5201_2
30. Van Orden KA, Cukrowicz KC, Witte TK, Joiner TE. **Thwarted belongingness and perceived burdensomeness: Construct validity and psychometric properties of the interpersonal needs questionnaire**. *Psychol. Assess.* (2012.0) **24** 197-215. DOI: 10.1037/a0025358
31. Bastien CH, Vallières A, Morin CM. **Validation of the insomnia severity index as an outcome measure for insomnia research**. *Sleep Med.* (2001.0) **2** 297-307. DOI: 10.1016/S1389-9457(00)00065-4
32. Wilson M-MG. **Appetite assessment: Simple appetite questionnaire predicts weight loss in community-dwelling adults and nursing home residents**. *Am. J. Clin. Nutr.* (2005.0) **82** 1074-1081. DOI: 10.1093/ajcn/82.5.1074
33. Cedereke M, Öjehagen A. **Prediction of repeated parasuicide after 1–12 months**. *Eur. Psychiatry* (2005.0) **20** 101-109. DOI: 10.1016/j.eurpsy.2004.09.008
34. Zhang X, Noor R, Savalei V. **Examining the effect of reverse worded items on the factor structure of the need for cognition scale**. *PLoS ONE* (2016.0) **11** e0157795. DOI: 10.1371/journal.pone.0157795
35. Rush AJ, Carmody T, Reimitz P-E. **The inventory of depressive symptomatology (IDS): Clinician (IDS-C) and self-report (IDS-SR) ratings of depressive symptoms**. *Int. J. Methods Psychiatr. Res.* (2000.0) **9** 45-59. DOI: 10.1002/mpr.79
36. Posner K. **The Columbia-suicide severity rating scale: Initial validity and internal consistency findings from three multisite studies with adolescents and adults**. *Am. J. Psychiatry* (2011.0) **168** 1266-1277. DOI: 10.1176/appi.ajp.2011.10111704
37. Fialko L. **A large-scale validation study of the medication adherence rating scale (MARS)**. *Schizophr. Res.* (2008.0) **100** 53-59. DOI: 10.1016/j.schres.2007.10.029
38. Ewing JA. **Detecting alcoholism: The CAGE questionnaire**. *JAMA* (1984.0) **252** 1905-1907. DOI: 10.1001/jama.1984.03350140051025
39. Meneses-Gaya IC, Zuardi AW, Loureiro SR, de Crippa JAS. **Psychometric properties of the Fagerström test for nicotine dependence**. *J. Bras. Pneumol.* (2009.0) **35** 73-82. DOI: 10.1590/S1806-37132009000100011
40. Brugha TS, Cragg D. **The list of threatening experiences: The reliability and validity of a brief life events questionnaire**. *Acta Psychiatr. Scand.* (1990.0) **82** 77-81. DOI: 10.1111/j.1600-0447.1990.tb01360.x
41. Bernstein DP. **Initial reliability and validity of a new retrospective measure of child abuse and neglect**. *Am. J. Psychiatry* (1994.0) **151** 1132-1136. DOI: 10.1176/ajp.151.8.1132
42. Patton JH, Stanford MS, Barratt ES. **Factor structure of the Barratt Impulsiveness Scale**. *J. Clin. Psychol.* (1995.0) **51** 768-774. DOI: 10.1002/1097-4679(199511)51:6<768::AID-JCLP2270510607>3.0.CO;2-1
43. Reips U-D, Funke F. **Interval-level measurement with visual analogue scales in internet-based research: VAS Generator**. *Behav. Res. Methods* (2008.0) **40** 699-704. DOI: 10.3758/BRM.40.3.699
44. Shrive FM, Stuart H, Quan H, Ghali WA. **Dealing with missing data in a multi-question depression scale: A comparison of imputation methods**. *BMC Med. Res. Methodol.* (2006.0) **6** 57. DOI: 10.1186/1471-2288-6-57
45. Bono C, Ried LD, Kimberlin C, Vogel B. **Missing data on the center for epidemiologic studies depression scale: A comparison of 4 imputation techniques**. *Res. Soc. Adm. Pharm.* (2007.0) **3** 1-27. DOI: 10.1016/j.sapharm.2006.04.001
46. Imai H. **Ipsative imputation for a 15-item geriatric depression scale in community-dwelling elderly people**. *Psychogeriatrics* (2014.0) **14** 182-187. DOI: 10.1111/psyg.12060
47. 47.MATLABVersion 9.7.0.1319299 (R2019b) Update 52019The MathWorks Inc.. *Version 9.7.0.1319299 (R2019b) Update 5* (2019.0)
48. Bonilla-Escribano P, Ramírez D, Porras-Segovia A, Artés-Rodríguez A. **Assessment of variability in irregularly sampled time series: Applications to mental healthcare**. *Mathematics* (2021.0) **9** 71. DOI: 10.3390/math9010071
49. Puth M-T, Neuhäuser M, Ruxton GD. **Effective use of Spearman’s and Kendall’s correlation coefficients for association between two measured traits**. *Anim. Behav.* (2015.0) **102** 77-84. DOI: 10.1016/j.anbehav.2015.01.010
50. Murphy KP, Murphy KP. **Mixture models and the EM algorithm**. *Machine Learning: A Probabilistic Perspective* (2012.0) 337-380
51. Ibrahim JG, Zhu H, Tang N. **Model selection criteria for missing-data problems using the EM algorithm**. *J. Am. Stat. Assoc.* (2008.0) **103** 1648-1658. DOI: 10.1198/016214508000001057
52. Wright SP. **Adjusted p-values for simultaneous inference**. *Biometrics* (1992.0) **48** 1005-1013. DOI: 10.2307/2532694
53. Delgado-Gómez D, Baca-García E, Aguado D, Courtet P, López-Castromán J. **Computerized adaptive test vs decision trees: Development of a support decision system to identify suicidal behavior**. *J. Affect. Disord.* (2016.0) **206** 204-209. DOI: 10.1016/j.jad.2016.07.032
54. Delgado-Gómez D, Laria JC, Ruiz-Hernández D. **Computerized adaptive test and decision trees: A unifying approach**. *Expert Syst. Appl.* (2019.0) **117** 358-366. DOI: 10.1016/j.eswa.2018.09.052
55. Breiman L. **Random forests**. *Mach. Learn.* (2001.0) **45** 5-32. DOI: 10.1023/A:1010933404324
56. 56.Springer, C. & Kegelmeyer, W. P. Feature selection via decision tree surrogate splits. In 2008 19th International Conference on Pattern Recognition 1–5. 10.1109/ICPR.2008.4761257 (2008).
57. 57.Loh, W.-Y. & Shih, Y.-S. Split Selection Methods for Classification Trees 26 (1997).
58. Gao L, Ding Y. **Disease prediction via Bayesian hyperparameter optimization and ensemble learning**. *BMC Res. Notes* (2020.0) **13** 205. DOI: 10.1186/s13104-020-05050-0
59. Provost F, Fawcett T. **Robust classification for imprecise environments**. *Mach. Learn.* (2001.0) **42** 203-231. DOI: 10.1023/A:1007601015854
60. James G, Witten D, Hastie T, Tibshirani R, James G, Witten D, Hastie T, Tibshirani R. **Tree-based methods**. *An Introduction to Statistical Learning* (2013.0) 303-335
61. Hosmer DW, Lemeshow S, Sturdivant RX, Hosmer DW, Lemeshow S, Sturdivant RX. **Assessing the fit of the model**. *Applied Logistic Regression* (2013.0)
62. Hagen J, Knizek BL, Hjelmeland H. **Mental health nurses’ experiences of caring for suicidal patients in psychiatric wards: An emotional endeavor**. *Arch. Psychiatr. Nurs.* (2017.0) **31** 31-37. DOI: 10.1016/j.apnu.2016.07.018
63. Peters EM. **Instability of suicidal ideation in patients hospitalized for depression: An exploratory study using smartphone ecological momentary assessment**. *Arch. Suicide Res.* (2020.0) **26** 1-14. PMID: 32669055
64. Martinez S. **The acute and repeated effects of cigarette smoking and smoking-related cues on impulsivity**. *Drug Alcohol Rev.* (2021.0) **40** 864-868. DOI: 10.1111/dar.13206
65. Waltmann M, Herzog N, Horstmann A, Deserno L. **Loss of control over eating: A systematic review of task based research into impulsive and compulsive processes in binge eating**. *Neurosci. Biobehav. Rev.* (2021.0) **129** 330-350. DOI: 10.1016/j.neubiorev.2021.07.016
66. Lim M, Lee S, Park J-I. **Differences between impulsive and non-impulsive suicide attempts among individuals treated in emergency rooms of South Korea**. *Psychiatry Investig.* (2016.0) **13** 389-396. DOI: 10.4306/pi.2016.13.4.389
67. Chu C, Nota JA, Silverman AL, Beard C, Björgvinsson T. **Pathways among sleep onset latency, relationship functioning, and negative affect differentiate patients with suicide attempt history from patients with suicidal ideation**. *Psychiatry Res.* (2019.0) **273** 788-797. DOI: 10.1016/j.psychres.2018.11.014
68. Weiner L. **Investigating racing thoughts in insomnia: A neglected piece of the mood-sleep puzzle?**. *Compr. Psychiatry* (2021.0) **111** 152271. DOI: 10.1016/j.comppsych.2021.152271
69. Peters EM, Baetz M, Marwaha S, Balbuena L, Bowen R. **Affective instability and impulsivity predict nonsuicidal self-injury in the general population: A longitudinal analysis**. *Borderline Personal. Disord. Emot. Dysregul.* (2016.0) **3** 17. DOI: 10.1186/s40479-016-0051-3
70. Liu RT, Trout ZM, Hernandez EM, Cheek SM, Gerlus N. **A behavioral and cognitive neuroscience perspective on impulsivity, suicide, and non-suicidal self-injury: Meta-analysis and recommendations for future research**. *Neurosci. Biobehav. Rev.* (2017.0) **83** 440-450. DOI: 10.1016/j.neubiorev.2017.09.019
71. Kivelä L, van der Does WAJ, Riese H, Antypa N. **Don’t miss the moment: A systematic review of ecological momentary assessment in suicide research**. *Front. Digit. Health* (2022.0) **4** 876595. DOI: 10.3389/fdgth.2022.876595
72. Altman N, Krzywinski M. **Ensemble methods: Bagging and random forests**. *Nat. Methods* (2017.0) **14** 933-934. DOI: 10.1038/nmeth.4438
|
---
title: Brief overview of dietary intake, some types of gut microbiota, metabolic markers
and research opportunities in sample of Egyptian women
authors:
- Nayera E. Hassan
- Salwa M. El Shebini
- Sahar A. El-Masry
- Nihad H. Ahmed
- Ayat N. Kamal
- Ahmed S. Ismail
- Khadija M. Alian
- Mohammed I. Mostafa
- Mohamed Selim
- Mahmoud A. S. Afify
journal: Scientific Reports
year: 2022
pmcid: PMC9981617
doi: 10.1038/s41598-022-21056-z
license: CC BY 4.0
---
# Brief overview of dietary intake, some types of gut microbiota, metabolic markers and research opportunities in sample of Egyptian women
## Abstract
Metabolic syndrome (MetS) is a phenotype caused by the interaction of host intrinsic factors such as genetics and gut microbiome, and extrinsic factors such as diet and lifestyle. To demonstrate the interplay of intestinal microbiota with obesity, MetS markers, and some dietary ingredients among samples of Egyptian women. This study was a cross-sectional one that included 115 Egyptian women; 82 were obese (59 without MetS and 23 with MetS) and 33 were normal weight. All participants were subjected to anthropometric assessment, 24 h dietary recall, laboratory evaluation of liver enzymes (AST and ALT), leptin, short chain fatty acids (SCFA), C-reactive protein, fasting blood glucose, insulin, and lipid profile, in addition to fecal microbiota analysis for Lactobacillus, Bifidobacteria, Firmicutes, and Bacteroid. Data showed that the obese women with MetS had the highest significant values of the anthropometric and the biochemical parameters. Obese MetS women consumed a diet high in calories, protein, fat, and carbohydrate, and low in fiber and micronutrients. The Bacteroidetes and Firmicutes were the abundant bacteria among the different gut microbiota, with low Firmicutes/Bacteroidetes ratio, and insignificant differences between the obese with and without MetS and normal weight women were reported. Firmicutes/Bacteroidetes ratio significantly correlated positively with total cholesterol and LDL-C and negatively with SCFA among obese women with MetS. Findings of this study revealed that dietary factors, dysbiosis, and the metabolic product short chain fatty acids have been implicated in causing metabolic defects.
## Introduction
Metabolic syndrome (MetS), differently known as insulin resistance, syndrome X, etc.., is characterized by WHO as a pathologic condition characterized by abdominal obesity, hyperleptinemia, hypertension, and insulin resistance1. In spite of the fact that there's some variety within the definition by other wellbeing care organization, the contrasts are minor. With the effective success in approximately elimination of communicable infectious diseases in most of the world, this modern non-communicable disease (NCD) has gotten to be the major health hazard of advanced world. The two fundamental forces spreading this ailment are the increment intake of high calorie and low fiber fast food and diminish in physical activity due to mechanized transportations and sedentary frame of leisure time exercises2.
Recent clinical and experimental research revealed that the gut microbiota is one of the foremost vital pathogenic variables in MetS3. Metabolic syndrome itself could be a phenotype caused by the interaction of host intrinsic variables such as genetics and the gut microbiome, and extrinsic components such as diets and the way of life. Metabolic syndrome is frequently accompanied by an imbalance of the gut microbiota, causes a low-grade inflammatory reaction within the body by wrecking the intestine barrier, creating insulin resistance through metabolites influencing host metabolism and hormone secretion, shaping a vicious circle that advances the persistent progress of MetS. Subsequently, intestinal microbiota may be a potential target for the treatment of MetS4.
Trillions of microorganisms live in symbiosis within the human body, and are primarily found within the gastrointestinal system, oral mucosa, saliva, skin, conjunctiva, and vagina5. The number of the microorganisms; that occupy the gastrointestinal tract (i.e., gut microbiota); be around 1 × 10146 and play a basic role in intestinal homeostasis, development, and protection against pathogens. In addition, their presence within the intestine is related to immunomodulatory and metabolic responses7.
Gut microbiota comprises of microbes, yeasts, and viruses. Intestine bacteria are more than 1000 species that have related to six overwhelming phyla: Firmicutes, Bacteroidetes, Actinobacteria, Proteobacteria, Fusobacteria, and Verrucomicrobia. The phyla Firmicutes and Bacteroidetes are the foremost common bacteria representing $90\%$ of the gut microbiota8.
Firmicutes bacteria are Gram-positive one. It plays a key role in the nutrition and metabolism of the host through SCFA synthesis. Through their metabolic products, *Firmicutes bacteria* are indirectly connected with other tissues and organs and regulate hunger and satiety. In contrast, *Bacteroidetes bacteria* are Gram-negative and associated with immunomodulation. Their components, lipo-poly-saccharides and flagellin, interact with cell receptors and enhance immune reactions through cytokine synthesis. Increased or decreased Firmicutes/Bacteroidetes (F/B) ratio are associated with the development of obesity or irritable bowel disease (IBD)9.
Firmicutes, due to their negative influence on glucose and fat metabolism, are commonly referred to as bad gut microbes10. Accumulating evidence proved that short chain fatty acids (SCFA) play a critical role in supporting the intestine and metabolic health. The SCFA acetic acid, propionate and butyrate are the main metabolites produced by the microbiota in the large intestine through the anaerobic fermentation of indigestible carbohydrates; they play a crucial role to gastrointestinal health11.
Microbial SCFA generation is basic for intestine integrity by controlling the luminal pH and mucus production. They are the main energy source for epithelial cells and speculated to play a key role in mucosal immune function. SCFA moreover straight forwardly modulate the metabolic health of the host through multiple neurochemical pathways related to energy expenditure, glucose homeostasis, appetite regulation, and immune-modulation12.
This study aimed to identify the interplay of the intestinal microbiota with obesity, metabolic markers and some dietary ingredient; especially the fat content; among a sample of Egyptian women.
## Subjects and methods
A cross-sectional study included 115 Egyptian women, with ages ranged between 25 and 60 years; mean age 41.62 ± 10.70 years. They were recruited and randomly chosen, from all employees and workers; of all categories; of the “National Research Centre”, Egypt. A written informed consent was obtained from all participants after being informed about the purpose of the study. This research paper was derived from a cross-sectional survey of a project funded by National Research Centre, Egypt, 2019–2022 entitled ‘‘Gut Microbiota in Obesity and Metabolic syndrome among obese women: Interactions of the Microbiome, Epigenetic, Nutrition and Probiotic Intervention.” ( 12th Research Plan of the National Research Centre), which was approved from “Ethics Committee of National Research Centre” (Registration Number is$\frac{19}{236}$). All used methods were performed in accordance with the relevant guidelines and regulations.
## Methods
For each participated woman, blood pressure, anthropometric measurements, 24 h dietary recall, laboratory investigations and microbiota analysis were done.
## Blood pressure
Blood pressure was measured using the standardized mercury sphygmomanometer with a suitable cuff size. It was measured on the left arm while the participated women were sitting relaxed for 5 min. Two readings were obtained and the average was recorded. Systolic blood pressure (SBP); determined by the onset of the “tapping” korotkoff sounds (K1), while the fifth korotkoff sound (K5), or the disappearance of korotkoff sounds, as the definition of diastolic blood pressure (DBP) were recorded.
## Anthropometric measurements
Body weight, height, neck, hip and waist circumferences were measured, following the recommendations of the “International Biological Program”13. Body weight (Wt) was determined to the nearest 0.01 kg using a Seca Scale Balance, with the woman wearing minimal clothes and with no shoes. Body height (Ht) was measured to the nearest 0.1 cm using a Holtain portable anthropometer. Circumferences was measured using non-stretchable plastic tape; approximated to the nearest 0.1 cm. Neck Circumference was measured at a point mid-way between the collarbone and chin in the middle of the neck while Standing or sitting with a straight back. Waist circumference (WC) was measured at the midpoint between the lower curvature of the last fixed rib and the superior curvature of the iliac crest, with the woman in an upright standing position and their arms alongside the body, feet together, and abdomen relaxed. Hip circumference was measured at the maximum extension of the buttocks measuring the largest diameter above the symphisis pubis overlapping the apex of the buttocks. Waist/hip ratio [WC/Hip C in cm] and Body mass index (BMI) [BMI: weight (in kilograms) divided by height (in meters squared)] were calculated. The participated women were all chosen as obese; as their BMI ≥ 30 kg/m2.The participant women were classified according to their BMI into 2 groups: 30 women with normal BMI (BMI = 18– < 25 kg/m2) and 82 obese women (BMI ≥ 30 kg/m2).
## Dietary recalls
Information from each participant about her usual pattern of food intake was obtained. Data was collected by means of dietary interview consisting of 24 h recall that repeated for 3 days, and a food frequency questionnaire. Analysis of food items was done using World Food Dietary Assessment System, (WFDAS), USA, University of California14.
## Blood sampling and laboratory investigations
In the morning, venous blood samples (after 12-h fasting) were drawn from the participated women, using venipuncture. Biochemical parameters were performed on fasting sera that were stored at – 70 °C until used for assessment of liver enzymes: Aspartate amino-transferase (AST) and Alanine amino-transferase (ALT), leptin, Short Chain Fatty Acids (SCFA), C-reactive protein (CRP), fasting blood glucose (FBG), insulin, and lipid profile. All were done in the laboratory of “Medical Excellence Research Center” which is a part of the “National Research Centre”, Egypt.
Serum concentrations of AST and ALT were determined using the automated clinical chemistry analyzer Olympus AU 400 analyzer (https://www.mybiosource.com).
The assay of human Leptin in serum was performed by ELISA method, using kits of BioLegend, Inc., (San Diego – USA), according to the method of Considine et al.15.
Human Short Chain Fatty Acids (SCFA) were assessed in serum using Enzyme Linked Immuno-sorbent Assay (ELISA) kits; Catalog Number: MBS7269061 according to the method described by den Besten et al16.
Fasting blood glucose (FBG) level was measured using the automated clinical chemistry analyzer Olympus AU 400 analyzer. Serum insulin was assessed using Enzyme Immunoassay Test Kit Catalog No. E29-072(Immunospec Corporation).
The assay of the serum CRP was performed by Enzyme Linked Immuno-sorbent Assay (ELISA) kits, Cat No.: RAP00217, (https://www.mybiosource.com.) Estimation of lipid profile: Serum levels of total cholesterol (TC), triglycerides (TG), high density lipoprotein cholesterol (HDL-C) were measured by standardized enzymatic procedures; using kits supplied by Roche Diagnostics (Mannheim, Germany) on the Olympus AU 400 automated clinical chemistry analyzer. Low density lipoprotein cholesterol (LDL-C) was calculated according to formula of Friedewald et al.18 as follows: LDL – C = Total cholesterol – Triglycerides/5 + HDL-C.
Clinically, a patient is considered to have MetS when three or more of the following five conditions exist, which are [1] waist circumference ≥ 88 cm in women, [2] blood pressure ≥ $\frac{135}{85}$ mmHg, [3] triglycerides ≥ 150 mg/dl, [4] HDL-C < 50 mg/dl in women, and [5] fasting glucose ≥ 100 mg/dl19.
## Microbiota analysis
The proportion of Lactobacillus and Bifidobacteria; and Firmicutes/Bacteroidetes ratio strains were assessed in the stool of all participants by using the real time PCR (Polymerase Chain Reaction). Specimen collection and preparation: Stool was collected by defecation in a plain sterilized container allowed to be frozen. Specimen Storage and Preparation: stool was frozen on at − 20°. The primers and probes were used to detect Bifidobacterium spp. and Lactobacillus spp; and Firmicutes spp. and Bacteroidetes spp., where based on 16S rRNA gene sequences retrieved from the National Center for Biotechnology Information databases by means of the Entrez program20. Primer sets used in this study. Target organism Primer Set Sequence (5′ to 3′) Product Size (bp) Ta (°C), time (s) Reference Lactobacillus Lacto-16S-F GGA ATC TTC CAC AAT GGA CG. genus Lacto-16S-R CGC TTT ACG CCC AAT AAA TCC GG 216 56, 10 s Bifidobacterium g-Bifid-F CTC CTG GAA ACG GGT GG Matsuki et al. [ 2004] genus g-Bifid-R GGT GTT CTT CCC GAT ATC TAC A 562 (549–563) 61, 20s Matsuki et al. [ 2004]. Bacteroidetes: 798cfbF AAACTCAAAKGAATTGACGG (Forward), and cfb967R GGTAAGGTTCCTCGCGCTAT (Reverse). Firmicutes: 928F–firm TGAAACTYAAGGAATTGACG(Forward),and1040FirmR CCATGCACCACCTGTC (Reverse), and universal bacterial 16S rRNA sequences: 926F AAACTCAAAKGAATTGACGG(Forward), and 1062R CTCACRRCACGAGCTGAC (Reverse).
Reagents provided by kits: DNA extraction Kit. Assay procedure: DNA extraction: The QIAamp DNA Stool Minikit (Qiagen) was used to extract DNA from one gram of fresh or frozen stool sample according to the manufacturer's instructions. Bacterial quantification by real-time PCR was done.
## Statistical analysis
Data were analyzed using the Statistical Package for Social Sciences (SPSS/Windows Version 18, SPSS Inc., Chicago, IL, USA). Normality of data was tested using the Kolmogorov–Smirnov test. The data were normally distributed. So, the parametric tests were used.
The participated women were classified into: 33 with normal BMI (18– < 25 kg/m2) and 82 obese with BMI ≥ 30 kg/m2. They obese women were classified according to the presence of MetS markers into two groups: 59 obese without MetS (have no or less than 2 markers of MetS), and 23 obese with MetS (have 3 or more markers of MetS).
The parametric data were expressed as mean ± SE. The various parametric variables of the two groups were analyzed and compared using independent t test. Pearson’s correlation test was used to assess the relations between Firmicutes/Bacteroid ratio and the clinical and metabolic parameters, and between gut microbiota and daily intake of total fat, carbohydrate and fiber among the three groups. $p \leq 0.05$ was regarded as statistically significant for all tests.
## Results
Table 1 showed the mean ± SE of the age, blood pressure, anthropometric and body composition parameters of the studied sample. Data revealed highly significant difference between the three groups in most of the parameters at p ≤ 0.01; where the obese women with MetS had the highest values. While the obese women without MetS had the significant highest values regarding WHR, NC and FFM ($p \leq 0.05$).Table 1Characteristic anthropometric parameters and blood pressure of the studied women ANOVA.ParametersObeseControlN = 33pObese without MetSN = 59Obese with MetsN = 23Mean ± SEMAge (year)48.65 ± 0.8543.47 ± 0.5931.70 ± 0.61b,c0.000**Blood pressureSBP (mmHg)115.38 ± 0.45137.50 ± 0.53105.00 ± 0.39b,c0.000**DBP (mmHg)72.92 ± 0.3685.68 ± 0.4063.75 ± 0.31b,c0.000**AnthropometryHeight (cm)158.12 ± 0.81159.11 ± 0.60157.59 ± 0.450.542Weight (Kg)91.82 ± 0.63100.10 ± 0.9449.04 ± 0.83b,c0.000**BMI (Kg/m2)36.19 ± 0.4840.01 ± 0.9319.55 ± 0.53b,c0.000**MWC (cm)102.66 ± 0.31113.96 ± 0.3573.01 ± 0.33b,c0.000**Hip C (cm)123.17 ± 0.83119.19 ± 0.5485.30 ± 0.32b,c0.000**WHR (cm/cm)0.66 ± 0.020.653 ± 0.030.57 ± 0.07b,c0.045*Neck cir. ( cm)39.59 ± 0.7938.01 ± 0.6732.62 ± 0.53b,c0.011**Skin fold thicknessTSF (mm)31.71 ± 0.1233.00 ± 0.1717.30 ± 0.15b,c0.000***BSF (mm)26.24 ± 0.13a30.39 ± 0.1713.72 ± 0.10b,c0.000***Subscap.(mm)31.41 ± 0.1634.32 ± 0.1914.40 ± 0.11b,c0.000***Supra-iliac (mm)27.71 ± 0.26a32.57 ± 0.3515.80 ± 0.23b,c0.000***Abd. SF (mm)31.40 ± 0.1537.67 ± 0.1621.10 ± 0.13b,c0.000***Body compositionBody fat (%)44.27 ± 0.4145.96 ± 0.4323.52 ± 0.33b,c0.000***FFM (kg)53.67 ± 0.4949.93 ± 0.5237.43 ± 0.37b,c0.000***Body water (%)36.55 ± 0.6139.29 ± 0.6627.41 ± 0.72b,c0.000**BMR (KJ)6538.21 ± 0.187034.13 ± 0.1748,014.31 ± 0.14b,c0.000**BMI, Body Mass Index; BMR, Basal Metabolic Rate; MWC, Minimal Waist Circumference; TSF, Triceps slin fold; BSF; Biceps skin fold; SBP, Systolic Blood Pressure; DBP, Diastolic Blood Pressure.*Significant at p ≤ 0.05, **highly Significant at p ≤ 0.01. *** very highly Significant at p ≤ 0.001.aObese without MetS V Obese with Mets, bObese without MetS V Control, cObese with Mets V Control.
Table 2 showed the mean ± SE of biochemical parameters of the three studied groups. The obese women with MetS had the significant highest values for the liver enzymes [Aspartate aminotransferase (AST) and the Alanine aminotransferase (ALT)], fasting blood glucose, insulin, and serum lipid profile; except the HDL-C which showed the insignificant lowest value compared to both the obese women without MetS and the control. The mean value of the C-reactive protein had the highly significant highest value among the obese women without MetS. The mean serum concentration of leptin hormone in the obese women with Mets was the significantly lowest compared to the other two groups. SCFA was insignificantly the highest among the obese women with MetS.Table 2Mean ± SE of biochemical parameters of the three studied groups. ParametersObeseNo:59Obese with MetsNo:23ControlNo:33pMean ± SEALT (U/L)18.58 ± 0.6223.26 ± 0.9315.480.79c0.005**AST(U/L)21.03 ± 0.4425.69 ± 0.3519.12 ± 0.87c0.001***Leptin216.82 ± 0.42191.18 ± 0.31a253.30 ± 0.47b.c0.017*SCFA14.84 ± 0.0417.47 ± 0.0612.21 ± 0.01c0.063CRP723.42 ± 0.65693.35 ± 0.74a390.85 ± 0.79b,c0.000***FBG (mg/dL)114.66 ± 0.03130.13 ± 0.34a86.33 ±.27b,c0.000***Insulin (mIU/L)12.21 ± 0.0716.73 ± 0.047.93 ± 0.03b,c0.001***Lipid profileT C (mg/dL)195.57 ± 0.45208.52 ± 0.39a177.82 ± 0.41b,c0.020*TG (mg/dL)98.50 ± 0.42154.39 ± 0.91a73.12 ± 0.71b,c0.000***HDL-C (mg/dL)57.01 ± 0.4054.34 ± 0.6459.12 ± 0.360.332LDL-C (mg/dL)117.55 ± 0.51123.26 ± 0.43100.73 ± 0.31b,c0.033*Significant between groups *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.FBG, Fasting blood glucose; TC, Total Cholesterol; TG, Triglyceride; HDL-C, High density lipoprotein –Cholesterol LDL-C, Low density lipoprotein-Cholesterol; ALT,Alanine aminotransferase; AST, Aspartate aminotransferase; SCFA, Short chain fatty acid; CRP, C-reactive protein.aObese without MetS V Obese with Mets, bObese without MetS V Control, cObese with Mets V Control.
Table 3 showed the mean ± SE of the log number and types of Microbiota among the three studied groups. Bacteroidetes bacteria were the most prevalent type among them, followed by the bad gut microbes Firmicutes, while Lactobacillus and Bifidobacteria were the least prevalent. There were significant differences between the log numbers of the 4 types of studied microbiota in the three groups at p ≤ 0.021, 0.020, 0.031 respectively. However, insignificant difference was found between the three studied groups, as regard the Firmicutes/Bacteroidetes Ratio (0.72 ± 0.02, 0.69 ± 0.03and 0.73 ± 0.02 respectively) and the two beneficial; the Lactobacillus and Bifidobacteria. Table 3Mean ± SE of the number and types of Microbiota among the three groups. Type of microbiotaObeseNo:59Obese with Mets No:23ControlNo:33pMean ± SELog Lactobasillus6.1096 ± 0.0965.8462 ± 0.1745.9163 ± 0.0680.207Log Bifidobacterium6.1352 ± 0.0916.1200 ± 0.1215.9403 ± 0.0880.337Log Bacteroidetes13.2459 ± 0.19613.0548 ± 0.20512.7845 ± 0.1670.256Log Firmicutes9.4530 ± 0.1948.9127 ± 0.2989.2845 ± 0.1850.281P0.021*0.020*0.031*Firmicutes/Bacteroid Ratio0.7197 ± 0.0160.6890 ± 0.0280.7313 ± 0.0180.456*Significant at $p \leq 0.05.$
Table 4 showed the mean ± SE and % of the recommended daily allowances (RDAs) of nutrients intake of the studied women. The obese women with MetS consumed the significant highest percentage of calories represented by the high intake of proteins, total fats and carbohydrates with the lowest fiber intake compared to the other two groups with significant difference at p ≤ 0.05–000. The intake of vitamin A and D, potassium, calcium, zinc and iron was low in all groups compared to the RDAs, and it was the lowest among the obese women with MetS with significant difference; except for iron and zinc where there were insignificant differences. The intake of sodium was within the limits of the RDAs. For the daily calcium intake, significant difference was found between the obese women with and without Mets and the control at p ≤ 0.021, while insignificant differences were detected for the other two minerals. Consumption of saturate fatty acid (SFAs) was high compared to the RDAs in all the groups, while the daily intake both the monounsaturated fatty acids (MUFAs) and polyunsaturated fatty acids (PUFAs) was the lowest among the obese women with MetS but with insignificant differences. Level of cholesterol intake was high among obese with and without Mets compared to the control with significant difference at p ≤ 0.033.The obese women with and without MetS consumed high fat diet which contributed to 42.29–$42.89\%$ of the total caloric intake, significant difference at p ≤ 0.05 compared to the control was recorded, with insignificant differences in the protein or carbohydrate intake (Table 5).Table 4Mean ± SE and % of the RDA of nutrients intake of the studied women. Nutrient intakeObeseNO:59Obese with MetsNo:23ControlNo:33RDAspMean ± SEM and % of RDAsEnergy (Cal)2253.45 ± 9.41102.432520.13 ± 8.32114.551906.89 ± 6.74c86.6822000.001***Protein (g)74.76 ± 3.69149.5286.04 ± 3.52a172.0862.26 ± 4.33b,c124.52500.000***Fat (g)105.89 ± 6.39137.52120.09 ± 5.73a155.9682.21 ± 8.27b,c106.77770.000***Carbohydrate (g)250.35 ± 4.5183.45273.79 ± 6.1091.26229.49 ± 8.60c76.503000.026*Dietary fiber (g)17.59 ± 0.3470.3615.86 ± 0.4163.4421.37 ± 0.58c85.48250.042*Vit. A (µg)498.65 ± 0.0762.33476.92 ± 0.1359.62638.25 ± 0.02c79.788000.031*Vit. D (µg)2.89 ± 0.0957.802.11 ± 0.0542.203.62 ± 0.07c72.4050.037*Sodium (mg)1506.12 ± 21.11100.411510.30 ± 24.01100.691101.26 ± 31.04b,c73.4215000.024*Potassium (mg)1520.80 ± 16.2576.041357.19 ± 12.3067.861610.69 ± 11.73c80.5320000.031*Calcium (mg)381.06 ± 8.1747.63338.14 ± 7.1142.27565.04 ± 9.10b,c70.638000.021*Iron (mg)5.22 ± 0.5065.254.87 ± 0.3060.885.93 ± 0.20c74.1380.051Zinc (mg)5.43 ± 1.0745.255.10 ± 1.0342.506.29 ± 1.01c52.42120.067Sat. FA (g)33.71 ± 1.1313.4637.89 ± 1.0913.5324.55 ± 1.20c11.59No more than $7\%$ of Total Calories intake0.065MUFs (g)29.02 ± 0.0411.5924.07 ± 0.058.6029.88 ± 1.40c14.1012–$14\%$ of Total Calories intake0.079PUFAs (g)18.04 ± 0.057.2016.11 ± 0.035.7520.15 ± 0.08c9.516–$8\%$ of Total Calories intake0.063Cholesterol (mg)238.30 ± 3.17119.15264.73 ± 6.72132.37192.42 ± 9.15b,c96.212000.033*Significant between groups *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.aObese without MetS V Obese with Mets, bObese without MetS V Control, cObese with Mets V Control. Table 5The percent of Calories delivered from fat, protein, and carbohydrates of total daily calories intakeof the studied women. ParameterObese without MetS No:59Obese with MetSNo:23Control No:33pPercentFat42.2942.8938.800.047Protein13.2713.6513.060.218Carbohydrates44.4443.4648.140.054Significance difference p ≤ 0.05.
Table 6 showed the Pearson’s correlation coefficient between Firmicutes/Bacteroid Ratio and markers of MetS and other Biochemical parameters of the studied sample. Among the control group, Firmicutes/Bacteroid Ratio had significant positive correlations with SBP, WC, leptin and HDL-C, and significant negative correlations with AST, ALT, insulin, triglycerides and total cholesterol. These correlations became significantly negative with the waist circumference (WC), while significant positive correlation appeared with CRP among the obese group without MetS. Among obese group with MetS, Firmicutes/Bacteroid Ratio had significant positive correlations with total cholesterol and LDL-C and significant negative correlation with SCFA.Table 6Pearson’s correlation coefficient between Firmicutes/Bacteroid ratio and markers of MetS and other biochemical parameters of the studied sample. ParametersFirmicutes/bacteroid ratioObese without MetSNo:59Obese with MetSNo:23ControlNo:33RprprpSBP − 0.1510.284 − 0.3910.0720.5110.011*DBP − 0.0790.579 − 0.1030.6470.2130.318BMI0.0020.990 − 0.2310.2890.1010.595MWC − 0.2720.037* − 0.1450.510.4430.014*% Fat0.0090.948 − 0.1580.471 − 0.2950.114AST − 0.0380.788 − 0.0050.981 − 0.6060.000**ALT − 0.0280.840 − 0.2960.17 − 0.6190.000**Leptin − 0.0610.663 − 0.0880.6910.530**0.001**SCFA0.0290.837 − 0.4110.050* − 0.2530.155BCAA0.0270.851 − 0.1030.640.0020.993CRP0.2730.048* − 0.3010.163 − 0.0550.759FBG − 0.2180.116 − 0.0970.658 − 0.3260.064Insulin − 0.019 − 0.892 − 0.3740.079 − 0.4390.011*TG0.0070.9630.30.164 − 0.3620.039*HDL-C0.2330.093 − 0.2220.3080.4610.007**LDL-C − 0.0480.7340.4850.019* − 0.3420.051TC − 0.0710.6140.5620.005** − 0.4260.014*Significance difference *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.000.FBG, Fasting blood glucose; TC, Total Cholesterol; TG, Triglyceride; HDL-C, High density lipoprotein –Cholesterol LDL-C, Low density lipoprotein-Cholesterol; ALT, Alanine aminotransferase; AST, Aspartate aminotransferase; SCFA, Short chain fatty acid; CRP, C-reactive protein.
Table 7 showed correlation coefficient between gut microbiota and daily intake of total fat, carbohydrate and fiber among the studied groups. Log Bacteroidetes showed significant positive correlation with the daily fat intake and highly significant negative correlations with the carbohydrate and fiber intake in the control group. Log Lactobacillus had significant positive correlations with the carbohydrate and fiber intake among obese women without MetS. While, Log Bifidobacteria, Log Firmicutes and log Firmicutes/Bacteroid Ratio had insignificant correlations with all the daily intake of total fat, carbohydrate and fiber among the three groups. Table 7Pearson’s correlation coefficient between gut microbiota and daily intake of total fat, carbohydrate and fiber of the studied groups. Log LactobasillusLog BifidobacteriumLog BacteroidetesLog FirmicutesLog Firmicutes/Bacteroid ratiorpRpRprprpObese without MetsFat0.0640.631 − 0.0020.9900.0900.499 − 0.1000.449 − 0.1630.217Carb0.2630.045* − 0.0800.5470.0980.4610.0020.991 − 0.0720.585Fiber0.2820.031* − 0.0290.8260.0990.4540.0810.540 − 0.0080.951Obese with MetsFat0.0250.908 − 0.2340.282 − 0.0950.6660.0040.9870.0370.887Carb0.1120.609 − 0.2350.2790.0150.9450.0930.6720.0880.680Fiber0.1400.525 − 0.1480.5000.0880.756 − 0.1680.442 − 0.1670.447ControlFat0.0450.402 − 0.2790.0580.3120.038* − 0.1320.231 − 0.2650.068Carb − 0.1080.275 − 0.1920.142 − 0.4620.003** − 0.1380.2230.0970.295Fiber − 0.2280.101 − 0.1790.159 − 0.4800.002** − 0.1800.1580.0720.345Significance difference *p ≤ 0.05, **p ≤ 0.01.
## Discussion
Obesity has been a motivating force behind the raised medical interest in identification of MetS pathogenesis in western communities and, increasingly, in eastern countries21.
Current data of this study reported by dietary history revealed that the obese participated women with MetS consumed the highest percentage of calories represented by the high daily intake of total fats and carbohydrates with lower fiber intake compared to the other two groups. This style of feeding was reflected in the anthropological scales and biochemical parameters, as all of their measures were the highest among the three groups, which is evident in body mass index, waist circumference, as well as all the skin fold thicknesses, in addition to the obvious hyperglycemia, dyslipidemia and higher insulin and CRP concentrations. Oda22, and Battault et al.23 reported that MetS involved a set of risk factors including obesity, hypertension, hyperglycemia, dyslipideamia, hyperuricemia and others. Uncontrolled MetS will eventually lead to non-alcoholic fatty liver (NAFLD), obstructive sleep apnea-hypo apnea syndrome (OSAHS), and other diseases. The pathogenesis of MetS is correlated with multiple factors, such as chronic inflammation, insulin resistance, oxidative stress and autonomic dysfunction.
## Gut microbiota count
Recently, the disturbance of gut microbiota has been discovered as a risk factor for MetS development1. The gut microbiota has developed a symbiotic relationship with the host involving the control of gene expression, gut barrier function, metabolism, nutrition and the general immunological function of the host24. Obesity is connected with changes in the relative abundance of the two dominant bacterial divisions; the Bacteroidetes and the Firmicutes; according to Xiao and Kang25.
The results of this study are in agreement with what was previously reported where significant increase in Bacteroidetes and the *Firmicutes bacteria* over the other types (Lactobacillus and Bifido bacteria) was detected. However, interestingly, the data in the current study detected insignificant difference between the three groups for each type of the studied of microbiota. It is likely that the eating pattern of the control lean group was relatively high in fibers which are important for the growth of the beneficial microbiota. Holscher26 stated that dietary fibers promote a healthy gut microbome, and that the consumption of dietary fibers and probiotic can modulate the microbiota in the gastrointestinal tract.
## Firmicutes/bacteroidetes ratio and markers of MetS and diet
Obese animals and humans have been found by de Wit et al.27 to possess a higher Firmicutes/Bacteroidetes ratio in their gut microbiota than normal-weight people, proposing that this ratio could be used as an obesity biomarker. As a result, the Firmicutes/Bacteroidetes ratio is recently acknowledged as a hallmark of obesity in the scientific literature of Magne et al.28. Furthermore, the F/B ratio has been proposed as an important sign of the gut microbiota health29. This ratio has been connected to multiple clinical conditions, including those related to ageing30 and others associated with obesity and metabolic syndrome31. Several studies have explored the link between diet and the gut microbiota because of the potential of dietary interventions to shape the composition of the gut microbiota. Each type of macronutrients (proteins, dietary fibers, fat) influences the gut microbiota specifically. Changes are observed more at a metabolic level than at a taxonomic level with a quick change in gene expression depending on the macronutrients32. Gut dysbiosis is associated with various pathologic conditions affecting the gastrointestinal tract (diarrhea, irritable bowel syndrome)33, inflammatory bowel diseases34, metabolism of the host (obesity, type 2 diabetes, atherosclerosis)35.
However, there are some controversies regarding the composition of the gut microbial communities in obese individuals in different populations where several studies have shown contradictory results (for example, reduced F/B ratios in obese individuals36,37. There are various reasons for these inconsistent results. Perhaps the association between F/B ratios and obesity depends on specific population, age group, gender, environmental and genetic factors38, and as already mentioned, other phyla (Proteobacteria, etc.) has an important role. In addition, the number of specific bacterial species from the Firmicutes and *Bactericides phylum* associated with obesity is limited. Firmicutes R. bromii is associated with obesity, utilizes and degrades more resistant starch than Eubacteriumrectale, B taiotaomicron, and Bifidobacterium adolescentis38. Interestingly, certain Bacteroides species also carry various genes for carbohydrate-degrading enzymes39.
In this context, the current study showed low Firmicutes/Bacteroidetes Ratio with only numerical difference between the three studied groups, the detected values were 0.72 and 0.69 inthe obese without and with MetS respectively compared to the control, 0.73. The lack to detect modest differences between healthy and obese subjects was in agreement with Magne et al.28 who suggested that the Firmicutes/Bacteroidetes ratio is not a robust marker of micro-biomedysbiosis associated with obesity. Because all the participants in this study belong to the same human race, to similar socio economic level and live in somewhat similar environment, accordingly there was a similarity in their dietary habits. Rinninella et al.40 reported that dietary habits can strongly influence gut microbiota composition. Food components have a key impact on the gut microbiota, influencing its composition in terms of richness and diversity41 However, data revealed significant correlation between the F/B ratio and some markers of the metabolic syndrome represented by negative correlation with the waist circumference and positive correlation with C-reactive protein in the obese women without MetS, while significant positive correlations were detected with LDL-C and total cholesterol among obese women with MetS. Among the control group, F/B ratio had significant positive correlations with SBP, WC, leptin and HDL-C, and significant negative correlations with AST, ALT, insulin, triglycerides and total cholesterol. This may clarified the presence of changes in the relations between F/B ratio and the different markers of MetS among the obese and normal weight women. This has been shown in results that revealed the difference in the significant response between the biochemical parameters and the Firmicutes/Bacteroid Ratio in the control group. In the same time only few significant relations were detected among obese and MetS women. On this basis, this study indicates the importance of this ratio among Egyptian obese women with and without MetS.
The Present study showed significant positive correlation between Log Bacteroidetes with the daily fat intake, and highly significant negative correlations with the carbohydrate and fiber intake in the control group. However, Log Lactobacillus had significant positive correlations with the carbohydrate and fiber intake among obese women without MetS. On the other hand, Log Bifidobacteria, Log Firmicutes and log Firmicutes/Bacteroidetes ratio had insignificant correlations with the entire daily intake of total fat, carbohydrate and fiber among the three groups.
Diet is one of the provocative factors in progression of obesity and is greatly linked to gut microbiota composition42. Nutrient intake and eating habits directly impact the composition, diversity, and metabolism of gut microbiota42,43. As previously mentioned, dietary intake of both obese groups revealed high caloric intake with high fat intake and low fiber content. Many dietary patterns such as Western diet, Mediterranean diet, vegetarian diet, and the gluten-free diet have been shown to affect the discrete diversity of the gut microbiota which can disturb host metabolism43. The Western diet involves high intake of saturated fats, sugar, salt, refined grains and high fructose corn syrup with a low intake of fibers; it is highly related to obesity and metabolic diseases. The Western diet was found to promote inflammation and changes the profile of the gut microbiota from healthy to the obese pattern44. It also has been shown to decrease the total bacteria amount as well as the beneficial Lactobacillus species (sp.) and Bifidobacterium sp. in the gut45.
## Short chain fatty acids
The gut microbiota is involved in the development of obesity by direct interactions with proximal organs or indirect interactions with distant organs through metabolic products (mainly SCFAs) including communication with the liver, adipose tissue, and brain46. SCFAs can play a crucial role in the pathogenesis of obesity. They interact with adipose tissue via two G-protein-coupled receptors expressed in adipocytes (Gpr41 and Gpr43); this promotes adipocytes formation and inhibits lipolysis47. Furthermore, SCFAs down regulate the synthesis of the hunger-suppressing hormones leptin, peptide YY, and glucagon-like peptide 148. SCFA influence lipid and glucose metabolism in order to provide energy for the host, proposing that they may have an impact on the occurrence of metabolic risk factors49.
Data of this study showed that the serum concentration of SCFA was mostly elevated among the obese women with MetS and in the same time showed negative association with the Firmicutes/Bacteroidetes ratio. Also it was evident that the levels of liver enzymes and the serum insulin concentration were also the highest among obese MetS participants, while the level of the leptin hormone was the lowest, which agree with what was mentioned in the previous research.
## Conclusion
It was concluded from this study that both Bacteroidetes and *Firmicutes bacteria* are the most abundant bacteria in the gut among the studied sample, whether in the obese with and without MetS or normal weights women. In addition the results confirm the low Firmicutes/Bacteroid Ratio among all groups. No clear relations were found between this ratio and the fat, carbohydrate, fiber content of the diets. It was one of the most important findings of this research: the links between the short chain fatty acids which is the metabolic product of the gut microbiota and the promotion of obesity and the ensuing metabolic disorders.
## References
1. Agus A, Clément K, Sokol H. **Gutmicrobiota-derived metabolites as central regulators in metabolic disorders**. *Gut* (2021) **70** 1174-1182. DOI: 10.1136/gutjnl-2020-323071
2. Saklayen MG. **The global epidemic of the metabolic syndrome**. *Curr. Hypertens. Rep.* (2018) **20** 12. DOI: 10.1007/s11906-018-0812-z
3. Kim MH, Yun KE, Kim J, Park E, Chang Y, Ryu S, Kim HL, Kim HN. **Gut microbiota and metabolic health among overweight and obese individuals**. *Sci. Rep.* (2020) **10** 19417. DOI: 10.1038/s41598-020-76474-8
4. Wang PX, Deng XR, Zhang C, Yuan HJ. **Gut microbiota and metabolic syndrome**. *Chin. Med. J. (Engl.)* (2020) **133** 808-816. DOI: 10.1097/CM9.0000000000000696
5. Sender R, Fuchs S, Milo R. **Revised estimates for the number of human and bacteria cells in the body**. *PLoS Biol.* (2016) **14** e1002533. DOI: 10.1371/journal.pbio.1002533
6. Thursby E, Juge N. **Introduction to the human gut microbiota**. *Biochem. J.* (2017) **474** 1823-1836. DOI: 10.1042/BCJ20160510
7. Zheng D, Liwinski T, Elinav E. **Interaction between microbiota and immunity in health and disease**. *Cell Res.* (2020) **30** 492-506. DOI: 10.1038/s41422-020-0332-7
8. Rinninella E, Raoul P, Cintoni M, Franceschi F, Miggiano GAD, Gasbarrini A, Mele MC. **What is the healthy gut microbiota composition? A changing ecosystem across age, environment, diet, and diseases**. *Microorganisms* (2019) **7** 14. DOI: 10.3390/microorganisms7010014
9. Stojanov S, Berlec A, Štrukelj B. **The influence of probiotics on the firmicutes/bacteroidetes ratio in the treatment of obesity and inflammatory bowel disease**. *Microorganisms* (2020) **8** 1715. DOI: 10.3390/microorganisms8111715
10. Zaky A, Glastras SJ, Wong MYW, Pollock CA, Saad S. **The role of the gut microbiome in diabetes and obesity-related kidney disease**. *Int. J. Mol. Sci.* (2021) **22** 9641. DOI: 10.3390/ijms22179641
11. Rowland I, Gibson G, Heinken A, Scott K, Swann J, Thiele I, Tuohy K. **Gut microbiota functions: Metabolism of nutrients and other food components**. *Eur. J. Nutr.* (2018) **57** 1-24. DOI: 10.1007/s00394-017-1445-8
12. Blaak EE, Canfora EE, Theis S, Frost G, Groen AK, Mithieux G, Nauta A, Scott K, Stahl B, van Harsselaar J, van Tol R, Vaughan EE, Verbeke K. **Short chain fatty acids in human gut and metabolic health**. *Benef. Microbes* (2020) **11** 411-455. DOI: 10.3920/BM2020.0057
13. Hiernaux J, Tanner J, Weiner JS, Lourie SA. **Growth and physical studies**. *Human Biology: Aguide to Field Methods* (1969)
14. 14.World Food Dietary Assessment SystemWFDAS1995University of California. *WFDAS* (1995)
15. Considine RV, Sinha MK, Heiman ML, Kriauciunas A, Stephens TW, Nyce MR, Ohannesian JP, Marco CC, McKee LJ, Bauer TL. **Serum immunoreactive-leptin concentrations in normal-weight and obese humans**. *N. Engl. J. Med.* (1996) **334** 292-295. DOI: 10.1056/NEJM199602013340503
16. denBesten G, van Eunen K, Groen AK, Venema K, Reijngoud DJ, Bakker BM. **The role of short-chain fatty acids in the interplay between diet, gut microbiota, and host energy metabolism**. *J. Lipid. Res.* (2013) **54** 2325-2340. DOI: 10.1194/jlr.R036012
17. Mitra B, Panja M. **High sensitive C-reactive protein: A novel biochemical markers and its role in coronary artery disease**. *J. Assoc. Phys. India* (2005) **53** 25-32
18. 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
19. Alberti KG, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA, Fruchart JC, James WP, Loria CM, Smith SC. **International Diabetes Federation Task Force on Epidemiology and Prevention; Hational Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; International Association for the Study of Obesity. Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity**. *Circulation* (2009) **120** 1640-1645. DOI: 10.1161/CIRCULATIONAHA.109.192644
20. Wheeler HE, Shah KP, Brenner J, Garcia T, Aquino-Michaels K, Cox NJ, Nicolae DL, Im HK. **Survey of the heritability and sparse architecture of gene expression traits across human tissues**. *PLoS Genet.* (2016) **12** e1006423. DOI: 10.1371/journal.pgen.1006423
21. Belete R, Ataro Z, Abdu A, Sheleme M. **Global prevalence of metabolic syndrome among patients with type I diabetes mellitus: A systematic review and meta-analysis**. *Diabetol. Metab. Syndr.* (2021) **13** 25. DOI: 10.1186/s13098-021-00641-8
22. Oda E. **Historical perspectives of the metabolic syndrome**. *Clin. Dermatol.* (2018) **36** 3-8. DOI: 10.1016/j.clindermatol.2017.09.002
23. Battault S, Meziat C, Nascimento A, Braud L, Gayrard S, Legros C. **Vascular endothelial function masks increased sympathetic vasopressor activity in rats with metabolic syndrome**. *Am. J. Physiol. Heart Circ. Physiol.* (2018) **314** H497-H507. DOI: 10.1152/ajpheart.00217.2017
24. Liu R, Hong J, Xu X, Feng Q, Zhang D, Gu Y. **Gut microbiome and serum metabolome alterations in obesity and after weight-loss intervention**. *Nat. Med.* (2017) **23** 859-868. DOI: 10.1038/nm.4358
25. Xiao H, Kang S. **The role of the gut microbiome in energy balance with a focus on the gut-adipose tissue axis**. *Front. Genet.* (2020) **7** 297. DOI: 10.3389/fgene.2020.00297
26. Holscher HD. **Dietary fiber and prebiotics and the gastrointestinal microbiota**. *Gut Microbes* (2017) **8** 172-184. DOI: 10.1080/19490976.2017.1290756
27. de Wit N, Derrien M, Bosch-Vermeulen H, Oosterink E, Keshtkar S, Duval C, de Vogel-van den Bosch J, Kleerebezem M, Müller M, van der Meer R. **Saturated fat stimulates obesity and hepatic steatosis and affects gut microbiota composition by an enhanced overflow of dietary fat to the distal intestine**. *Am. J. Physiol. Liver Physiol.* (2012) **303** G589-G599. DOI: 10.1152/ajpgi.00488.2011
28. Magne F, Gotteland M, Gauthier L, Zazueta A, Pesoa S, Navarrete P, Balamurugan R. **The Firmicutes/Bacteroidetes ratio: A relevant marker of gut dysbiosis in obese patients?**. *Nutrients* (2020) **12** 1474. DOI: 10.3390/nu12051474
29. Li W, Ma ZS. **FBA ecological guild: Trio of Firmicutes-Bacteroidetes alliance against Actinobacteria in human oral microbiome**. *Sci. Rep.* (2020) **10** 287. DOI: 10.1038/s41598-019-56561-1
30. Liang D, Leung RK, Guan W. **Involvement of gut microbiome in human health and disease: Brief overview, knowledge gaps and research opportunities**. *Gut Pathog.* (2018) **10** 3. DOI: 10.1186/s13099-018-0230-4
31. Woting A, Blaut M. **The intestinal microbiota in metabolic disease**. *Nutrients* (2016) **8** 202. DOI: 10.3390/nu8040202
32. Aguirre M, Eck A, Koenen ME, Savelkoul PHM, Budding AE, Venema K. **Diet drives quick changes in the metabolic activity and composition of human gut microbiota in a validated in vitro gut model**. *Res. Microbiol.* (2016) **167** 114-125. DOI: 10.1016/j.resmic.2015.09.006
33. Carding S, Verbeke K, Vipond DT, Corfe BM, Owen LJ. **Dysbiosis of the gut microbiota in disease**. *Microb. Ecol. Health Dis.* (2015) **26** 26191. DOI: 10.3402/mehd.v26.26191
34. Hills RD, Pontefract BA, Mishcon HR, Black CA, Sutton SC, Theberge CR. **Gut microbiome profound. Implications for diet and disease**. *Nutrients* (2019) **11** 1613. DOI: 10.3390/nu11071613
35. Yong VB. **The role of the microbiome in human health and disease. An introduction for clinicians**. *BMJ* (2017). DOI: 10.1136/bmj.j831
36. Castaner O, Goday A, Park YM, Lee SH, Magkos F, Shiow STE, Schröder H. **The gut microbiome profile in obesity: A systematic review**. *Int J Endocrinol.* (2018) **22** 4095789
37. Vaiserman A, Romanenko M, Piven L, Moseiko V, Lushchak O, Kryzhanovska N, Guryanov V, Koliada A. **Differences in the gut Firmicutes to Bacteroidetes ratio across age groups in healthy Ukrainian population**. *BMC Microbiol.* (2020) **20** 221. DOI: 10.1186/s12866-020-01903-7
38. Ze X, Duncan SH, Louis P, Flint HJ. **Ruminococcusbromii is a keystone species for the degradation of resistant starch in the human colon**. *ISME J.* (2012) **6** 1535-1543. DOI: 10.1038/ismej.2012.4
39. Flint HJ, Scott KP, Duncan SH, Louis P, Forano E. **Microbial degradation of complex carbohydrates in the gut**. *Gut Microbes* (2012) **3** 289-306. DOI: 10.4161/gmic.19897
40. Rinninella E, Cintoni M, Raoul P, Lopetuso LR, Scaldaferri F, Pulcini G, Miggiano GAD, Gasbarrini A, Mele MC. **Food components and dietary habits: Keys for a healthy gut microbiota composition**. *Nutrients* (2019) **11** 2393. DOI: 10.3390/nu11102393
41. Mayer EA, Tillisch K, Gupta A. **Gut/brain axis and the microbiota**. *J. Clin. Investig.* (2015) **125** 926-938. DOI: 10.1172/JCI76304
42. Brahe LK, Astrup A, Larsen LH. **Can we prevent obesity-related metabolic diseases by dietary modulation of the gut microbiota?**. *Adv. Nutr.* (2016) **7** 90-101. DOI: 10.3945/an.115.010587
43. Lazar V, Ditu LM, Pircalabioru GG, Picu A, Petcu L, Cucu N. **Gut microbiota, host organism, and diet trialogue in diabetes and obesity**. *Front. Nutr.* (2019) **6** 21. DOI: 10.3389/fnut.2019.00021
44. Statovci D, Aguilera M, MacSharry J, Melgar S. **The impact of Western diet and nutrients on the microbiota and immune response at mucosal interfaces**. *Front. Immunol.* (2017) **8** 838. DOI: 10.3389/fimmu.2017.00838
45. Bell DS. **Changes seen in gut bacteria content and distribution with obesity: Causation or association?**. *Postgrad. Med.* (2015) **127** 863-868. DOI: 10.1080/00325481.2015.1098519
46. Mitev K, Taleski V. **Association between the gut microbiota and obesity**. *Open Access Maced. J. Med. Sci.* (2019) **7** 2050-2056. DOI: 10.3889/oamjms.2019.586
47. Kimura I, Ozawa K, Inoue D, Imamura T, Kimura K, Maeda T, Terasawa K, Kashihara D, Hirano K, Tani T, Takahashi T, Miyauchi S, Shioi G, Inoue H, Tsujimoto G. **The gut microbiota suppresses insulin-mediated fat accumulation via the short-chain fatty acid receptor GPR43**. *Nat. Commun.* (2013) **4** 1829. DOI: 10.1038/ncomms2852
48. Tseng CH, Wu CY. **The gut microbiome in obesity**. *J. Formos. Med. Assoc.* (2019) **118** S3-S9. DOI: 10.1016/j.jfma.2018.07.009
49. Teixeira TF, Grześkowiak Ł, Franceschini SC, Bressan J, Ferreira CL, Peluzio MC. **Higher level of faecal SCFA in women correlates with metabolic syndrome risk factors**. *Br. J. Nutr.* (2013) **109** 914-919. DOI: 10.1017/S0007114512002723
|
---
title: Risk factors for totally implantable access ports-associated thrombosis in
pediatric oncology patients
authors:
- Yingxia Lan
- Liuhong Wu
- Jin Guo
- Juan Wang
- Huijie Guan
- Baihui Li
- Longzhen Liu
- Lian Zhang
- Ye Hong
- Jun Deng
- Jia Zhu
- Suying Lu
- Feifei Sun
- Junting Huang
- Xiaofei Sun
- Yizhuo Zhang
- Jian Wang
- Ruiqing Cai
journal: Scientific Reports
year: 2023
pmcid: PMC9981621
doi: 10.1038/s41598-023-30763-0
license: CC BY 4.0
---
# Risk factors for totally implantable access ports-associated thrombosis in pediatric oncology patients
## Abstract
The application of totally implantable access ports (TIAPs) reduces treatment-related discomfort; however, the existence of catheter may cause side effects, with the most common one being the occurrence of TIAPs-associated thrombosis. The risk factors for TIAPs-associated thrombosis in pediatric oncology patients have not been fully described. A total of 587 pediatric oncology patients undergoing TIAPs implantation at a single center over a 5-year period were retrospectively analyzed in the present study. We investigated the risk factors for thrombosis, emphasizing the internal jugular vein distance, by measuring the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae on chest X-ray images. Among 587 patients, 143 ($24.4\%$) had thrombosis. Platelet count, C-reactive protein, and the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae were demonstrated to be the main risk factors for the development of TIAPs-associated thrombosis. TIAPs-associated thrombosis, especially asymptomatic events, is common in pediatric cancer patients. The vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae was a risk factor for TIAPs-associated thrombosis, which deserved additional attention.
## Introduction
Cancer patients are prone to thromboembolism. Thrombosis results from coagulation dysfunction, which is caused by various factors such as individual patient factors, disease status, and treatment regimens1. Among these factors, previous studies have identified central venous catheter (CVC) as an independent factor associated with an increased risk of VTE in cancer patients2. In recent years, the use of TIAPs in cancer patients has gradually increased due to its long indwelling time, long maintenance period, and improved quality of life over other CVCs3,4. TIAP has become especially popular among pediatric oncology patients, as it can reduce unintentional pullout of exposed catheters commonly occurring in children and is more likely to produce positive body images than other CVCs.
The currently reported incidence of catheter-related thrombosis (CRT) varies widely, depending on symptomatic events or whether patients are screened for asymptomatic thrombosis5. CRT incidence in pediatric cancer population may differ from that in adults6. And in a large study7, risk factors for CRT differed from those for noncatheter-related venous thromboembolism. Although some studies have analyzed the risk factors for CVC-related thrombosis8,9, most of them adopt peripherally inserted central catheter (PICC), and the risk factors for TIAPs-associated thrombosis have not been fully described. Therefore, the aim of the present study was to elucidate the risk factors for TIAPs-associated thrombosis in pediatric oncology patients.
## Study population
A total of 587 pediatric oncology patients who underwent TIAPs implantation at Sun Yat-sen University cancer center between January 2016 and October 2020 were retrospectively included in our study. Only newly diagnosed with malignancy children younger than 18 years at the time of port insertion were included. Patients who lost follow-up after initial examination and treatment were excluded. TIAPs placement was performed by a professional surgeon and anesthesia team, and all port implantation was guided by Doppler ultrasound. Catheter depth is determined by anatomical landmark technique. Each patient underwent a chest X-ray after port insertion to detect possible misplacements and to rule out pleura-pulmonary complications such as pneumothorax or hemothorax. As for the choice of catheter, we keep the outer diameter of the catheter no more than $\frac{1}{3}$ of the internal diameter of the vein. Blood routine examination, coagulation function, and infectious disease screening were completed for each patient before port placement. Patients’ data were followed up by telephone and access to outpatient and inpatient data. The endpoint was the port removal or the last day of follow-up (December 31, 2021). Our study was approved by the Ethic Committee of Sun Yat-sen University Cancer Center (Approval Number: B2022-197-01) and conducted in accordance with the latest version of the Declaration of Helsinki. The written informed consent was obtained from all participants and/or their legal guardians.
## Analysis of TIAPs-associated thrombosis
The TIAPs-associated thrombosis was confirmed by ultrasound imaging, and no radiographic imaging was used due to its toxicity in children. We explored the clinical presentations of thrombosis, including insertion location and outcome. In addition, the association of thrombosis with age, gender, body mass index (BMI), tumor type, disease stage, albumin, C-reactive protein (CRP), white blood cell (WBC) count, hemoglobin level, platelet count, d-Dimer, international normalized ratio (INR), activated partial prothrombin time (APTT), and the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae in chest X-ray images were analyzed.
## Statistical analysis
All analyses were done by IBM SPSS Statistics version 19.0 (SPSS, Chicago, IL, USA) software. The Chi-square test was used for comparisons of qualitative variables between the groups, if appropriate. Continuous variables were compared using the Student’s test, as appropriate. All variables that affected thrombosis (level of significance, $P \leq 0.1$) in the univariate analysis were included in the multivariate logistic regression analysis. A P-value of < 0.05 was considered statistically significant.
## Patient characteristic
Over the 5-year study period, 587 consecutive newly diagnosed pediatric oncology patients were included. The patient characteristics are summarized in Table 1. Among these patients, 143 ($24.4\%$) developed TIAPs-associated thrombosis. The median age was 4.0 (range 0.1–16) years, and the male to female ratio was 1.4. For patients with thrombosis, the duration of port insertion was defined as the time from TIAPs placement to port removal or the last follow-up time, with an average of 29.5 (range 1.3–72.4) months. At the end of follow-up, $39.2\%$ ($\frac{56}{143}$) of patients with thrombosis had completed treatment and the port had been removed. The mean time from thrombus onset to port insertion was 24.3 days, with a median of 18.0 (range 3–488) days. The median interval between the first ultrasound imaging for thrombosis was 13.0 (range 3–1209) days. Also, the median interval between the second ultrasound imaging was 21 (range 4–890) days. Table 1Patient characteristics ($$n = 587$$).CharacteristicsVTE (+) n (%)VTE (−) n (%)Total n (%)P-valueGender0.277 Male77 (53.8)262 (59.0)339 (57.8) Female66 (46.2)182 (41.0)248 (42.2)Age (years)0.945 Mean (min–max)4.7 (0.1–16)4.7 (0.3–13)4.7 Median4.04.04.0BMI0.426 Mean (min–max)15.6 (10.2–25.6)15.6 (10.1–27.6)15.6 Median15.315.315.3Stage0.969 Early39 (27.3)117 (26.4)156 (26.6) Advanced98 (68.5)307 (69.1)405 (69.0) Recurrence6 (4.2)20 (4.5)26 (4.4)WBC (× 109/L)0.617 Mean (min–max)8.6 (1.28–23.01)8.4 (0.79–63.09)8.4 Median8.147.47.6Hemoglobin (g/L)0.073 Mean (min–max)114.3 (55–155)110.7 (28–158)111.6 Median118.0115.0115.0Platelet count (× 109/L)0.001 Mean (min–max)381.3 (37–1156)336.9 (9–1020)347.7 Median377.0322.0332.5Albumin (g/L)0.745 Mean (min–max)43.5 (30.2–114.8)43.7 (23.2–197.1)43.6 Median43.543.343.3d-Dimer (mg/L)0.411 Mean (min–max)8.5 (0.09–121.19)10.1 (0.05–129.09)9.7 Median0.91.21.1APTT (s)0.469 Mean(min–max)24.4 (0.1–42.2)25.4 (0.12–121)25.1 Median27.327.827.6INR0.239 Mean (min–max)1.01 (0.79–1.31)1.06 (0.82–1.77)1.06 Median1.021.041.03CRP (mg/L)0.002 Mean (min–max)8.2 (0.03–88.12)14.9 (0.02–300.9)13.3 Median1.21.61.5Distancea (mm)0.05 Mean (min–max)17.1 (7–41)15.8 (5–58)16.1 Median15.515.015.0Drugb0.273 a106 (74.1)353 (79.5)459 (78.2) b18 (12.6)37 (8.3)55 (9.4) c19 (13.3)54 (12.2)73 (12.4)aDistance defines the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae.ba: Not use dexamethasone and asparaginase, b: Use dexamethasone and asparaginase, c: Only use dexamethasone.
We classified all pediatric oncology patients into five major groups: leukemia, lymphoma, neuroblastoma, sarcoma, and other types of tumor (Fig. 1). In terms of tumor types, leukemia and other types had a higher incidence of thrombosis, while neuroblastoma and sarcoma had a lower incidence. Leukemia was statistically significant compared with neuroblastoma (Odds ratio [OR] 2.31, $$P \leq 0.043$$), while other types were more prone to thrombosis than neuroblastoma (OR 2.17, $$P \leq 0.008$$) and sarcoma (OR 1.72, $$P \leq 0.05$$).Figure 1Flow chart of study populations.
## Site features of TIAPs-associated thrombosis
Among 143 patients with thrombosis (Table 2), in $67.8\%$ ($\frac{97}{143}$) patients, it occurred in the right internal jugular vein. In 26 patients with thrombosis in the right subclavian vein, the port insertion was the right internal jugular vein. Two patients with thrombosis in the right brachiocephalic vein also had an insertion vessel in the right internal jugular vein, and one patient with thrombosis in the left subclavian vein had a port insertion site in the left internal jugular vein. The internal jugular vein was more prone to thrombosis than the subclavian vein; yet, the difference was not statistically significant (OR 1.56, $$P \leq 0.163$$).Table 2Sites of TIAPs-associated thrombosis in pediatric oncology patients. Insertion position of thrombosis patientsSite of thrombusInsertion position of all patientsRight internal jugular, n (%)125 (87.4)97 (67.8)488 (83.1)Right subclavian, n (%)12 (8.4)38 (26.6)70 (11.9)Left internal jugular, n (%)6 (4.2)5 (3.5)26 (4.4)Left subclavian, n (%)0 [0]1 (0.7)3 (0.5)Right brachipcephalic vein, n (%)0 [0]2 (1.4)0 [0]
## Risk factors for developing TIAPs-associated thrombosis
Hemoglobin ($$P \leq 0.073$$), platelet count ($$P \leq 0.001$$), CRP ($$P \leq 0.002$$), and the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae ($$P \leq 0.05$$) were identified as risk factors for developing TIAPs-associated thrombosis. Multivariate analysis revealed that platelet count, CRP, and the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae remained significant risk factors (Table 3).Table 3Risk factors by multivariate logistic regression analysis. Odds ratio$95\%$ CI (lower–upper)P-valueHemoglobin1.0040.993–1.0160.481Platelet count1.0021.001–1.0030.004CRP0.9900.980–1.0000.054Distance1.0311.001–1.0620.044
## Discussion
CVC is a well-known risk factor for thrombosis; however, only a few studies have investigated the risk factors for thrombosis associated with TIAP use in pediatric cancer patients10,11. Our study focused on risk factors for TIAPs-associated thrombosis, especially the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae. This has not been reported in previous studies and could help early identification of populations at high risk of thrombosis. Identification of risk factors for CRT in childhood cancer patients before port implantation may greatly contribute to prognosis improvement.
In a prospective multicenter French cohort study (ONCOCIP)7, the only independent risk factor for catheter-related thrombosis in adult cancer patients with port implants was the use of cephalic vein for catheter insertion. Specific factors may be related to the occurrence of TIAPs-associated thrombosis. Since previous studies have shown that right internal jugular vein insertion is linked with the lowest incidence of thrombosis12,13, $83.1\%$ ($\frac{488}{587}$) of patients in our center have chosen right internal jugular vein insertion. The remaining patients were unable to choose that site mostly due to the influence of diseases such as right mediastinal mass. Therefore, our findings suggested a higher incidence of thrombosis in the right internal jugular vein insertion, probably because most children chose this site.
Previous studies have shown that the use of dexamethasone and asparaginase are risk factors for thrombosis14; yet, we did not observe similar findings, which might be related to the fact that our pediatric patients had all types of childhood tumors and received different treatment regimens and medication. While cancer patients are prone to thrombosis, especially for those with hematological tumors such as leukemia15. Some cancer types are associated with a lower incidence of thrombosis, such as central nervous system (CNS) tumors that have been previously reported16. In our study, leukemia was significantly prone to thrombosis; neuroblastoma and sarcoma had the lowest thrombosis incidence of $17.2\%$ ($\frac{22}{128}$) and $20.9\%$ ($\frac{28}{134}$), respectively, which were similar to previous findings10.
In previous studies4, many CRT risk factors were discovered, including PICC instead of port, but our study only included TIAP patients. In a chemotherapy-related VTE model17, a platelet count > 350 × 109/L was one of the five predictors for VTE. The Compass-CAT risk prediction model18 still included a platelet count > 350 × 109/L as a risk factor. Some studies7 have also shown that antiplatelet therapy has a protective effect against CRT. Similar to other studies, platelet count was found to be a risk factor for CRT in our study. Current evidence is conflicting as to whether inflammation levels or CRP are risk factors for VTE19. Studies of CRP in cancer populations have not focused on CRT20,21. In a multicenter prospective study22, CRP levels were found to predict venous thromboembolism recurrence after discontinuation of anticoagulation for cancer-associated thrombosis. Overweight or obesity is a prothrombotic state in which risk factors for venous thrombosis are inflammatory responses, decreased fibrinolysis, increased thrombin production, and platelet hyperactivity23. In our study, few children were obese, but CRP was still a risk factor for CRT, which might be related to the different characteristics of catheter-related thrombosis versus non-catheter-related thrombosis. It was also possible that our patients all had malignant tumors, which were risk factors for thrombosis, and that thrombosis could be regulated by inflammatory markers24.
Due to slow blood flow velocity of the internal jugular vein and contralateral compensation, the blood flow velocity will be further reduced after port insertion, so patients with long internal jugular vein catheter may be prone to thrombosis. At present, the accurate judgment of the distance of the internal jugular vein needs to be determined by computed tomograph (CT). In order to reduce radiation doses, we generally chose chest X-ray to determine the correct placement of the catheter after port implantation. In addition, because of the error in the measurement of the distance from the puncture point to the extremitas sternalis claviculae during the operation, the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae in radiographic images was regarded as the distance of the internal jugular vein. We found it to be one of the risk factors for TIAPs-associated thrombosis by univariate and multivariate analysis. The distance was longer in patients with thrombosis, averaging 17.1 (range 7–41) mm, compared with 15.8 (range 5–58) mm in patients without thrombosis. Our study identified a new risk factor for port patients with internal jugular vein insertion that was easily observed. Therefore, we recommended the puncture site as low as possible to reduce the length of the internal jugular vein catheter. However, because low site puncture could easily penetrate the subclavian artery, blind puncture was not advisable. Doppler ultrasound-guided puncture would be recommended. However, since this was a retrospective study, more large clinical trials were needed to further verify our results.
In our center, in order to detect thrombus in time, ultrasound imaging of the upper extremity was routinely performed on the 7th and 21st days after port insertion. Consistent with previous reports25, most thrombotic events in our study occurred during the first 3 months of follow-up. Our data showed that in $28.7\%$ ($\frac{41}{143}$) patients, thrombus was found on the 7th day of routine ultrasound imaging; in $22.4\%$ ($\frac{32}{143}$) patients, it was found on the 21st day, and in the remaining patients, some was found after the appearance of symptoms. Nevertheless, since this was a retrospective study, the time of the first ultrasound imaging was not available for some patients because the examination was performed in other hospitals, and the report could not be obtained. At present, there are no guidelines on when to perform an examination after TIAP placement. PICC-related thrombosis is reported to be a common and almost always asymptomatic complication in children26. Children with CVC-associated thrombosis often have recurrent catheter-related complications. Therefore, asymptomatic thrombosis may be clinically important. In a prospective study27, the incidence of asymptomatic DVT and fibrin sheath monitored by Doppler ultrasound at 1-, 6-, and 12-month after implantation in cancer patients with long-term CVC implantation was 0.10 events per 1000 catheter days (< $1.5\%$). Therefore, asymptomatic patients were more likely to be detected using early ultrasound. Our study found $24.4\%$ of patients with thrombosis, most of whom were asymptomatic, and half of them were detected on day 7 and 21 after TIAP implantation. At our center, which is a large tumor research center in China, we accumulated a lot of experience in the management of thrombus in tumor patients, which we could recommend for thrombus screening in the future.
This study had some limitations. On the one hand, this was a retrospective study with patient selection bias. On the other hand, our patients had all types of childhood tumors and received different treatment regimens and medication, so it is difficult to determine the drug-related effect on thrombosis.
In conclusion, platelet count, CRP, and the vertical distance from the highest point of the catheter to the upper border of the left and right extremitas sternalis claviculae might be the risk factors for TIAPs-associated thrombosis. We recommend ultrasound imaging on day 7 and 21 after port insertion in pediatric oncology patients for early detection of asymptomatic thrombotic events. Moreover, the puncture position should be as low as possible to reduce the length of catheter in the internal jugular vein, which might reduce the occurrence of TIAPs-associated thrombosis.
## References
1. Fernandes CJ, Morinaga LTK, Alves JL. **Cancer-associated thrombosis: The when, how and why**. *Eur. Respir. Rev.* (2019) **28** 180119. DOI: 10.1183/16000617.0119-2018
2. Ashrani AA, Gullerud RE, Petterson TM. **Risk factors for incident venous thromboembolism in active cancer patients: A population based case-control study**. *Thromb. Res.* (2016) **139** 29-37. DOI: 10.1016/j.thromres.2016.01.002
3. Ignatov A, Hoffman O, Smith B. **An 11-year retrospective study of totally implanted central venous access ports: Complications and patient satisfaction**. *Eur. J. Surg. Oncol.* (2009) **35** 241-246. DOI: 10.1016/j.ejso.2008.01.020
4. Fang S, Yang J, Song L, Jiang Y, Liu Y. **Comparison of three types of central venous catheters in patients with malignant tumor receiving chemotherapy**. *Patient Prefer. Adherence.* (2017) **11** 1197-1204. DOI: 10.2147/PPA.S142556
5. Marin A, Bull L, Kinzie M, Andresen M. **Central catheter-associated deep vein thrombosis in cancer: Clinical course, prophylaxis, treatment**. *BMJ Support Palliat. Care.* (2021) **11** 371-380. DOI: 10.1136/bmjspcare-2019-002106
6. Albisetti M, Kellenberger CJ, Bergsträsser E. **Port-a-cath-related thrombosis and postthrombotic syndrome in pediatric oncology patients**. *J. Pediatr.* (2013) **163** 1340-1346. DOI: 10.1016/j.jpeds.2013.06.076
7. Decousus H, Bourmaud A, Fournel P. **Cancer-associated thrombosis in patients with implanted ports: A prospective multicenter French cohort study (ONCOCIP)**. *Blood* (2018) **132** 707-716. DOI: 10.1182/blood-2018-03-837153
8. Tabatabaie O, Kasumova GG, Kent TS. **Upper extremity deep venous thrombosis after port insertion: What are the risk factors?**. *Surgery.* (2017) **162** 437-444. DOI: 10.1016/j.surg.2017.02.020
9. Revel-Vilk S, Yacobovich J, Tamary H. **Risk factors for central venous catheter thrombotic complications in children and adolescents with cancer**. *Cancer* (2010) **116** 4197-4205. DOI: 10.1002/cncr.25199
10. Wiegering V, Schmid S, Andres O. **Thrombosis as a complication of central venous access in pediatric patients with malignancies: A 5-year single-center experience**. *BMC Hematol.* (2014) **14** 18. DOI: 10.1186/2052-1839-14-18
11. Citla Sridhar D, Abou-Ismail MY, Ahuja SP. **Central venous catheter-related thrombosis in children and adults**. *Thromb. Res.* (2020) **187** 103-112. DOI: 10.1016/j.thromres.2020.01.017
12. Verso M, Agnelli G, Kamphuisen PW. **Risk factors for upper limb deep vein thrombosis associated with the use of central vein catheter in cancer patients**. *Intern. Emerg. Med.* (2008) **3** 117-122. DOI: 10.1007/s11739-008-0125-3
13. Debourdeau P, Farge D, Beckers M. **International clinical practice guidelines for the treatment and prophylaxis of thrombosis associated with central venous catheters in patients with cancer**. *J. Thromb. Haemost.* (2013) **11** 71-80. DOI: 10.1111/jth.12071
14. Chen K, Agarwal A, Tassone MC. **Risk factors for central venous catheter-related thrombosis in children: A retrospective analysis**. *Blood Coagul Fibrinolysis.* (2016) **27** 384-388. DOI: 10.1097/MBC.0000000000000557
15. Kekre N, Connors JM. **Venous thromboembolism incidence in hematologic malignancies**. *Blood Rev.* (2019) **33** 24-32. DOI: 10.1016/j.blre.2018.06.002
16. Howie C, Erker C, Crooks B, Moorehead P, Kulkarni K. **Incidence and risk factors of venous thrombotic events in pediatric patients with CNS tumors compared with non-CNS cancer: A population-based cohort study**. *Thromb. Res.* (2021) **200** 51-55. DOI: 10.1016/j.thromres.2021.01.014
17. Khorana AA, Kuderer NM, Culakova E, Lyman GH, Francis CW. **Development and validation of a predictive model for chemotherapy-associated thrombosis**. *Blood* (2008) **111** 4902-4907. DOI: 10.1182/blood-2007-10-116327
18. Gerotziafas GT, Taher A, Abdel-Razeq H. **A predictive score for thrombosis associated with breast, colorectal, lung, or ovarian cancer: The prospective COMPASS-cancer-associated thrombosis study**. *Oncologist.* (2017) **22** 1222-1231. DOI: 10.1634/theoncologist.2016-0414
19. Galeano-Valle F, Ordieres-Ortega L, Oblitas CM. **Inflammatory biomarkers in the short-term prognosis of venous thromboembolism: A narrative review**. *Int. J. Mol. Sci.* (2021) **22** 2627. DOI: 10.3390/ijms22052627
20. Kanz R, Vukovich T, Vormittag R. **Thrombosis risk and survival in cancer patients with elevated C-reactive protein**. *J. Thromb. Haemost.* (2011) **9** 57-63. DOI: 10.1111/j.1538-7836.2010.04069.x
21. Kröger K, Weiland D, Ose C. **Risk factors for venous thromboembolic events in cancer patients**. *Ann. Oncol.* (2006) **17** 297-303. DOI: 10.1093/annonc/mdj068
22. Jara-Palomares L, Solier-Lopez A, Elias-Hernandez T. **D-dimer and high-sensitivity C-reactive protein levels to predict venous thromboembolism recurrence after discontinuation of anticoagulation for cancer-associated thrombosis**. *Br. J. Cancer.* (2018) **119** 915-921. DOI: 10.1038/s41416-018-0269-5
23. Samad F, Ruf W. **Inflammation, obesity, and thrombosis**. *Blood* (2013) **122** 3415-3422. DOI: 10.1182/blood-2013-05-427708
24. Saghazadeh A, Hafizi S, Rezaei N. **Inflammation in venous thromboembolism: Cause or consequence?**. *Int. Immunopharmacol.* (2015) **28** 655-665. DOI: 10.1016/j.intimp.2015.07.044
25. Hohl Moinat C, Périard D, Grueber A. **Predictors of venous thromboembolic events associated with central venous port insertion in cancer patients**. *J. Oncol.* (2014) **2014** 743181. DOI: 10.1155/2014/743181
26. Menéndez JJ, Verdú C, Calderón B. **Incidence and risk factors of superficial and deep vein thrombosis associated with peripherally inserted central catheters in children**. *J. Thromb. Haemost.* (2016) **14** 2158-2168. DOI: 10.1111/jth.13478
27. Boddi M, Villa G, Chiostri M. **Incidence of ultrasound-detected asymptomatic long-term central vein catheter-related thrombosis and fibrin sheath in cancer patients**. *Eur. J. Haematol.* (2015) **95** 472-479. DOI: 10.1111/ejh.12519
|
---
title: Effect of exposure to radiation caused by an atomic bomb on endothelial function
in atomic bomb survivors
authors:
- Shinji Kishimoto
- Nozomu Oda
- Tatsuya Maruhashi
- Shunsuke Tanigawa
- Aya Mizobuchi
- Farina Mohamad Yusoff
- Asuka Fujita
- Toshio Uchiki
- Masato Kajikawa
- Kenichi Yoshimura
- Takayuki Yamaji
- Takahiro Harada
- Yu Hashimoto
- Yukiko Nakano
- Seiko Hirota
- Shinji Yoshinaga
- Chikara Goto
- Ayumu Nakashima
- Yukihito Higashi
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC9981625
doi: 10.3389/fcvm.2023.1122794
license: CC BY 4.0
---
# Effect of exposure to radiation caused by an atomic bomb on endothelial function in atomic bomb survivors
## Abstract
### Background
The purpose of this study was to evaluate the effects of exposure to radiation caused by an atomic bomb in atomic bomb survivors on vascular function and vascular structure and to evaluate the relationships of radiation dose from the atomic bomb with vascular function and vascular structure in atomic bomb survivors.
### Methods
Flow-mediated vasodilation (FMD) and nitroglycerine-induced vasodilation (NID) as indices of vascular function, brachial-ankle pulse wave velocity (baPWV) as an index of vascular function and vascular structure, and brachial artery intima-media thickness (IMT) as an index of vascular structure were measured in 131 atomic bomb survivors and 1,153 control subjects who were not exposed to the atomic bomb. Ten of the 131 atomic bomb survivors with estimated radiation dose in a cohort study of Atomic Bomb Survivors in Hiroshima were enrolled in the study to evaluate the relationships of radiation dose from the atomic bomb with vascular function and vascular structure.
### Results
There was no significant difference in FMD, NID, baPWV, or brachial artery IMT between control subjects and atomic bomb survivors. After adjustment of confounding factors, there was still no significant difference in FMD, NID, baPWV, or brachial artery IMT between control subjects and atomic bomb survivors. Radiation dose from the atomic bomb was negatively correlated with FMD (ρ = −0.73, $$P \leq 0.02$$), whereas radiation dose was not correlated with NID, baPWV or brachial artery IMT.
### Conclusion
There were no significant differences in vascular function and vascular structure between control subjects and atomic bomb survivors. Radiation dose from the atomic bomb might be negatively correlated with endothelial function.
## Introduction
It is well-known that high-dose radiation increases the risk of cardiovascular disease [1], whereas the relationship between low-dose or middle-dose radiation and risk of cardiovascular disease is controversial (1–4). Shimizu et al. [ 1] showed that a dose of more than 0.5 Gy was correlated with an increased risk of cardiovascular disease but that there was no relationship between a radiation dose of less than 0.5 Gy and cardiovascular disease in atomic bomb survivors who had been followed up for 53 years. Tran et al. [ 4] showed that a radiation dose of less than 0.5 Gy was associated with mortality of cardiovascular disease in patients with tuberculosis. Some studies have shown that atomic bomb survivors have persistent inflammation that is positively correlated with radiation dose from the atomic bomb [5, 6]. Inflammation plays a crucial role in the progression of atherosclerosis [7]. Long-term observational studies in atomic bomb survivors showed that radiation to which they were exposed from the atomic bomb is associated with increased blood pressure, serum cholesterol level and incidence of diabetes, which are risk factors for atherosclerosis (8–10). While radiation therapy has improved the prognosis of cancer patients, cardiac exposure from radiation therapy causes an increase in the risk of coronary heart disease, which continues for a long time [11, 12]. In addition, long-term low-dose radiation exposure in healthcare workers was shown to be associated with early atherosclerosis [13]. Therefore, it is important to know the long-term effects of radiation from an atomic bomb on the vasculature function.
It is recognized that endothelial dysfunction occurs from the early stages of atherosclerosis development. Endothelial dysfunction plays an important role in cardiovascular complications [14, 15]. Flow-mediated vasodilation (FMD) was measured as an indicator of endothelial function, and nitroglycerine-induced vasodilation (NID) was measured as an indicator of vascular smooth muscle function [16, 17]. Several investigators have reported that vascular dysfunction predicts cardiovascular events [18, 19]. Measurement of brachial-ankle pulse wave velocity (baPWV) as an index of arterial stiffness, which is an indicator of both vascular function and vascular structure, and measurement of brachial artery intima-media thickness (IMT) as an index of vascular structure have been shown to be significantly correlated with cardiovascular risk factors [20, 21].
The purpose of this study was to evaluate the effects of radiation caused by the atomic bomb in atomic bomb survivors 65 years or longer after exposure to the radiation on vascular function and vascular structure and to evaluate the relationships of the dose of radiation from the atomic bomb to which the survivors were exposed with vascular function and vascular structure in atomic bomb survivors.
## Study protocol 1: Vascular function in atomic bomb survivors
Between August 2010 and June 2021, a total of 131 atomic bomb survivors were recruited for vascular function measurement from subjects who attended the outpatient clinic at Hiroshima University Hospital, and 3,966 control subjects who were not exposed to the atomic bomb were recruited from the Hiroshima University Hospital Vascular Registry. All of the atomic bomb survivors who participated in the study consented to the measurement and study participation, and those who did not consent were excluded. An atomic bomb survivor was defined as an individual who was formally issued an Atomic Bomb Health Handbook based on the law concerning assistance for atomic bomb survivors and who met one or more of the following conditions: having been directly exposed within a few kilometers of the hypocenter of the atomic bomb, having entered within two kilometers of the hypocenter within 2 weeks after the bombing, having been engaged in rescue or other related activities, and having been exposed in utero. The cohort of Atomic Bomb Survivors in Hiroshima (ABS) included about 290,000 atomic bomb survivors in Hiroshima who were issued an Atomic Bomb Health Handbook. ABS was a cohort study that was started in 1971 by the Research Institute for Radiation Biology and Medicine of Hiroshima University. Details of the ABS study methods were described previously [22]. Of the 3,966 control subjects, 2,813 subjects who were less than 65 years old were excluded because this study was started 65 years after the atomic bombing and the youngest age of the atomic bomb survivors was 65 years. Finally, 1,153 control subjects were enrolled in this study. Diabetes mellitus was defined according to the criteria provided by the American Diabetes Association or a previous diagnosis of diabetes [23, 24]. The definition of dyslipidemia was based on the third report of the National Cholesterol Education Program [25].
All measurements were done in the morning, after overnight fasting, in a quiet, dark, air-conditioned room (constant temperature of 22–25°C) during the study. FMD, NID, baPWV, and brachial artery IMT were measured, after maintaining the supine position for 30 min. The observers masked the clinical characteristics of the subjects and the aim of the study. All methods were carried out according with the Declaration of Helsinki and relevant guidelines and regulations. The study was approved by the Ethics Review Board of Hiroshima University. Written informed consent was obtained from all subjects.
## Study protocol 2: Relationship between vascular function and dose of radiation from the atomic bomb to hich atomic bomb survivors were exposed
After checking all members of Protocol 1 against the ABS cohort, 121 of the 131 atomic bomb survivors without estimated dose of radiation to which they were exposed from the atomic bomb in a cohort study of ABS in Hiroshima were excluded. Finally, 10 atomic bomb survivors whose radiation dose was accurately assessed were enrolled in study protocol 2.
## Measurements of FMD and NID
Flow-mediated vasodilation was measured as endothelium-dependent vasodilation of vascular response to reactive hyperemia in the brachial artery by using an automated edge detection system (UNEXEF18G, UNEX Co, Nagoya, Japan) [26]. NID was measured in vascular response to nitroglycerine as endothelium-independent vasodilation, as previously reported [26]. This study was conducted with a methodological approach to FMD, following the recommendations proposed by Thijssen et al. [ 27]. Additional details can be found in the Supplementary material.
## Measurement of brachial IMT and baPWV
Details on the measurement of Brachial IMT and baPWV can be found in the Supplementary material.
## Radiation dosimetry
We used the dose of exposure to radiation from neutrons and gamma rays estimated by using the Atomic Bomb Survivor 1993 Dose (ABS93D). ABS93D was described in detail in a previous report [28]. Briefly, radiation dose calculated by ABS93D is based on individual exposure status such as distance from the hypocenter, shielding and age at time of bombing. Hoshi et al. [ 28] showed that the dose evaluation of ABS93D was close to that of the Dosimetry system 1986 (DS86) by Radiation Effects Research Foundation. We used the weighted radiation dose of the colon, which is often chosen as the whole-body irradiation exemplary organ, by calculating the sum of the gamma ray dose and 10 times the neutron dose considering the biological effectiveness of neutrons.
## Statistical analysis
Results are summarized as means ± SD for continuous variables and as percentages for categorical variables. A 2-sided probability value of <0.05 was considered to indicate statistical significance. The FMD value in subjects over 60 years old was determined to be 2.7 ± $2.5\%$ in a previous study [29]. The number of subjects needed to detect a difference of $1.0\%$ FMD and a standard deviation (SD) of $2.5\%$ between two groups with a probability of 0.05 and a power of 0.80 was 100 per group. Continuous variables were compared by using ANOVA. Categorical variables were compared by using chi-square test. Relationships between variables were determined using Spearman’s correlation coefficients. To create a matched cohort of control subjects and atomic bomb survivors, a propensity score was calculated using logistic regression analysis of the probability of baseline clinical variables in two models: model 1 including age and sex and model 2 including age, sex, body mass index, heart rate, hypertension, dyslipidemia, diabetes mellitus, and current smokers. To create matched pairs to investigate the associations of exposure to radiation with vascular function and vascular structure, matched pairs were using one-to-one propensity-score matching analyses. The caliper size of propensity scores was used a quarter of a standard deviation of the sample estimated propensity scores for comparison of vascular function. The data were processed using JMP pro version 15 (SAS Institute, Cary, NC, USA).
## Study protocol 1: Baseline clinical characteristics
The baseline clinical characteristics of the 1,153 control subjects and 131 atomic bomb survivors are summarized in Table 1. There were significant differences in sex, heart rate and estimated glomerular filtration rate between control subjects and atomic bomb survivors.
**TABLE 1**
| Variables | Control subjects (n = 1,153) | Atomic bomb survivors (n = 131) | P-value |
| --- | --- | --- | --- |
| Age, year | 76 ± 5 | 76 ± 5 | 0.92 |
| Age at atomic bomb exposure, year | | 5 ± 4 | |
| Sex, men/women | 623/530 | 92/39 | <0.01 |
| Body mass index, kg/m2 | 23.0 ± 3.4 | 23.5 ± 3.0 | 0.09 |
| Systolic blood pressure, mmHg | 130 ± 18 | 128 ± 19 | 0.21 |
| Diastolic blood pressure, mmHg | 74 ± 11 | 73 ± 11 | 0.26 |
| Heart rate, bpm | 70 ± 12 | 66 ± 11 | <0.01 |
| Total cholesterol, mg/dL | 184 ± 36 | 178 ± 36 | 0.07 |
| Triglycerides, mg/dL | 122 ± 70 | 120 ± 70 | 0.80 |
| HDL cholesterol, mg/dL | 60 ± 17 | 59 ± 18 | 0.62 |
| LDL cholesterol, mg/dL | 104 ± 30 | 99 ± 29 | 0.14 |
| Glucose, mg/dL | 118 ± 38 | 113 ± 27 | 0.24 |
| HbA1c, % | 6.1 ± 0.8 | 6.2 ± 0.8 | 0.22 |
| eGFR, mL/min per 1.73 m2 | 60.1 ± 18.8 | 56.0 ± 18.8 | 0.03 |
| High-sensitivity CRP, mg/dL | 0.46 ± 1.12 | 0.59 ± 1.32 | 0.58 |
| Medical history, n (%) | Medical history, n (%) | Medical history, n (%) | Medical history, n (%) |
| Hypertension | 927 (80.4) | 102 (77.9) | 0.50 |
| Dyslipidemia | 794 (68.9) | 81 (61.8) | 0.11 |
| Diabetes mellitus | 442 (38.3) | 49 (37.4) | 0.84 |
| Previous coronary heart disease | 219 (19.0) | 34 (26.0) | 0.07 |
| Previous stroke | 128 (11.1) | 17 (13.0) | 0.54 |
| Current smoker | 87 (7.6) | 10 (7.6) | 0.97 |
| Medication, n (%) | Medication, n (%) | Medication, n (%) | Medication, n (%) |
| Antihypertensive drugs | 901 (78.1) | 107 (81.7) | 0.34 |
| Lipid-lowering drugs | 629 (40.6) | 72 (55.0) | 0.93 |
| Antidiabetic drugs | 341 (29.6) | 36 (27.5) | 0.62 |
Moreover, we assessed vascular function in atomic bomb survivors using propensity score matching to create matched pairs between control subjects and atomic bomb survivors. In propensity score-matched pairs of control subjects and atomic bomb survivors in model 1, the clinical characteristics of matched pairs of 131 control subjects and 131 atomic bomb survivors are summarized in Supplementary Table 1. In propensity score-matched pairs of control subjects and atomic bomb survivors in model 2, the clinical characteristics of matched pairs of 119 control subjects and 119 atomic bomb survivors are summarized in Supplementary Table 2.
## Study protocol 1: Vascular function and vascular structure in atomic bomb survivors
There was no significant difference in FMD, NID, baPWV, or brachial artery IMT between control subjects and atomic bomb survivors (2.9 ± 2.6 vs. 3.1 ± $2.6\%$, 9.9 ± 5.6 vs. 10.4 ± $5.5\%$, 1819 ± 391 vs. 1782 ± 366 cm/s, and 0.34 ± 0.07 vs. 0.34 ± 0.07 mm, $$P \leq 0.51$$, $$P \leq 0.38$$, $$P \leq 0.37$$, and $$P \leq 0.94$$, respectively) (Figure 1).
**FIGURE 1:** *Bar graphs show flow-mediated vasodilation (A), nitroglycerine-induced vasodilation (B), brachial-ankle pulse wave velocity (C), and brachial artery intima-media thickness (D) in control subjects and atomic bomb survivors.*
In propensity score-matched pairs of control subjects and atomic bomb survivors in model 1, there was no significant difference in FMD, NID, baPWV, or brachial artery IMT between control subjects and atomic bomb survivors (2.8 ± 2.4 vs. 3.1 ± $2.6\%$, 10.5 ± 5.8 vs. 10.4 ± $5.5\%$, 1853 ± 379 vs. 1782 ± 366 cm/s, and 0.35 ± 0.07 vs. 0.34 ± 0.07 mm, $$P \leq 0.44$$, $$P \leq 0.90$$, $$P \leq 0.16$$, and $$P \leq 0.16$$, respectively) (Supplementary Figure 1). In propensity score-matched pairs of control subjects and atomic bomb survivors in model 2, there was no significant difference in FMD, NID, baPWV, or brachial artery IMT between control subjects and atomic bomb survivors (2.6 ± 2.3 vs. 3.1 ± $2.7\%$, 9.5 ± 5.6 vs. 10.5 ± $5.6\%$, 1753 ± 358 vs. 1783 ± 354 cm/s, and 0.34 ± 0.07 vs. 0.34 ± 0.06 mm $$P \leq 0.11$$, $$P \leq 0.19$$, $$P \leq 0.57$$, and $$P \leq 0.71$$, respectively) (Supplementary Figure 2).
## Study protocol 2: Baseline clinical characteristics of atomic bomb survivors whose radiation dose was accurately assessed
The baseline clinical characteristics of the 10 atomic bomb survivors who radiation dose was accurately assessed are summarized in Supplementary Table 3. Of the 10 atomic bomb survivors, nine ($90.0\%$) had hypertension, eight ($80.0\%$) had dyslipidemia, three ($30.0\%$) had diabetes mellitus, four ($40.0\%$) had previous coronary heart disease, three ($30.0\%$) had previous stroke and one ($10.0\%$) was a current smoker. Mean values were 3.9 ± $1.5\%$ for FMD, 11.8 ± $4.4\%$ for NID, 1630 ± 234 cm/s for baPWV, and 0.32 ± 0.03 mm for brachial artery IMT.
## Study protocol 2: Effects of radiation dose on vascular function and vascular structure in atomic bomb survivors whose radiation dose was accurately assessed
Radiation dose from the atomic bomb was negatively correlated with FMD (ρ = −0.73, $$P \leq 0.02$$) (Figure 2A), whereas radiation dose was not correlated with NID, baPWV, or brachial artery IMT (ρ = 0.25, $$P \leq 0.52$$; ρ = −0.02, $$P \leq 0.97$$; and ρ = −0.23, $$P \leq 0.52$$; respectively) (Figures 2B–D).
**FIGURE 2:** *Scatter plots show relationships of radiation dose with flow-mediated vasodilation (A), nitroglycerine-induced vasodilation (B), brachial-ankle pulse wave velocity (C), and brachial artery intima-media thickness (D) in atomic bomb survivors.*
## Discussion
In the present study, we demonstrated that there was no significant difference in FMD, NID, baPWV, or brachial artery IMT between control subjects and atomic bomb survivors even after adjustment for confounding factors. Radiation dose from the atomic bomb was negatively correlated with FMD, whereas radiation dose was not correlated with NID, baPWV, or brachial artery IMT. As far as we know, this is the first study to assess vascular function and vascular structure in atomic bomb survivors 65 years or longer after exposure to radiation from the atomic bomb.
Previous studies showed a relationship between radiation exposure and cardiovascular disease in atomic bomb survivors (1–4). Shimizu et al. [ 1] showed that a dose of more than 0.5 Gy was correlated with an increased risk of cardiovascular disease but that there was no relationship between a dose of less than 0.5 Gy and cardiovascular disease in atomic bomb survivors who had been followed up for 53 years. Nakamizo et al. [ 30] showed that low-dose or mild-dose radiation exposure increased aorta calcification measured by X-ray films and carotid artery plaque measured by carotid ultrasound, whereas there were no significant relationship of dose of radiation exposed by the atomic bomb with augmentation index or baPWV in atomic bomb survivors examined from 2010 to 2014. However, there has been no information on vascular function in atomic bomb survivors 65 years or longer after radiation exposure. In the present study, there were no significant differences in vascular function and arterial structure between control subjects and atomic bomb survivors. We also performed analyses using two models of adjustment for cardiovascular risk factors. In both adjustment models, there were no significant differences in vascular function and vascular structure between control subjects and atomic bomb survivors. These findings suggest that radiation exposed by the atomic bomb has no specific effects on vascular function and vascular structure in atomic bomb survivors 65 years or longer after exposure to radiation from the atomic bomb.
It is well-known that high-dose radiation increases the risk of cardiovascular disease, whereas the relationship between low-dose or middle-dose radiation and the risk of cardiovascular disease is controversial (1–4). Tran et al. [ 4] showed that a dose of less than 0.5 Gy was associated with mortality of cardiovascular disease in patients with tuberculosis. In the present study, endothelial function did not differ between atomic bomb survivors and control subjects. Atomic bomb survivors were over 75 years of age and had several cardiovascular risk factors other than aging. Those risk factors may mask the effects of low to medium doses of radiation on vascular function. Interestingly, the dose of radiation exposed to was negatively correlated with FMD in atomic bomb survivors who were exposed to a dose of less than 0.6 Gy, suggesting that endothelial function might be impaired in relation to dose-response within the range of low to medium doses of radiation.
The pathogenesis of cardiovascular disease due to a high radiation dose is damage to endothelial cells and inflammation [5, 31], although the pathogenesis of cardiovascular disease due to low-dose or middle-dose radiation is unclear. In an in vitro study, Cervelli et al. [ 32] showed that exposure of human umbilical vein endothelial cells to single doses of less than 0.5 Gy resulted in increased intercellular adhesion molecule-1 and oxygen species generation. Mitchel et al. [ 33] showed that exposure to single doses of 0.025–0.5 Gy decreased the frequency of atherosclerosis lesion in ApoE–/– mice, whereas those doses of radiation increased inflammation, total serum cholesterol levels and severity of the atherosclerosis lesions. On the other hand, some investigators showed that low-dose radiation has an anti-inflammatory effect [34, 35]. Although those in vitro and in vivo studies showed short-term effects of radiation on the vasculature, the long-term effects of radiation on the vasculature are unclear. Long-term observational studies for atomic bomb survivors showed that the dose of radiation exposed by the atomic bomb was associated with increases in blood pressure, serum cholesterol level and incidence of diabetes, which are risk factors for atherosclerosis (8–10). Hayashi et al. [ 36] showed that atomic bomb survivors have persistent inflammation that is positively correlated with to the dose of radiation exposed by the atomic bomb. Kusunoki et al. [ 6] showed that T-cell immunity was attenuated in atomic bomb survivors and that there was a significant negative correlation between inflammatory markers and the number of naïve CD4 T cells. These findings suggest that T-cell immunosenescence may be partly responsible for the prolonged inflammation in atomic bomb survivors. In the present study, a radiation dose of less than 0.6 Gy was negatively correlated with FMD in atomic bomb survivors 65 years or longer after radiation exposure. A low or medium dose of radiation may impair endothelial function via immunological effects, inflammation, and metabolic changes in relation to radiation dose.
This study has some limitations. First, the number of atomic bomb survivors whose radiation dose was accurately assessed was relatively small. The reason for the lack of exposure data for atomic bomb survivors was that it was not known whether they entered within 2 km of the hypocenter after the atomic bomb explosion or the exact location of the exposure was not known. ABS93D provided dose estimates for 33,173 individuals since dose estimates for atomic bomb survivors by ABS93D were only data for direct exposure. Those individuals accounted for only $11.4\%$ of the total population of the ABS cohort. After checking all members of Protocol 1 against the ABS cohort, only 10 atomic bomb survivors were provided dose estimates. Those 10 atomic bomb survivors accounted for $7.6\%$ of the total population of Protocol 1. The proportion of atomic bomb survivors with dose estimates in the present study was slightly less than that of the overall ABS cohort. The present study may have included a large percentage of atomic bomb survivors who entered an area within 2 km of the hypocenter less than 2 weeks after the bombing, who were engaged in rescue or other related activities, or who had been directly exposed without the exact location of the exposure known. However, we found that the dose of radiation exposed by the atomic bomb was negatively correlated with FMD even in a small number of subjects. In the present study, the estimated glomerular filtration rate in atomic bomb survivors was significantly lower than that in control subjects. Sera et al. [ 37] showed that chronic kidney disease was significantly associated with radiation dose in atomic bomb survivors. The subjects in Study Protocol 1 might be a good representation of the overall ABS cohort. Second, atomic bomb survivors have received free medical care and may have received earlier detection and treatment of cardiovascular disease than subjects in the control group under the government care system. Further studies are needed to confirm the results of this study in trials by adjusting for the duration of cardiovascular diseases and treatment duration. Third, in the present study, biochemical oxidative stress markers and inflammatory markers were not measured in all of the subjects. Measurements of biochemical oxidative stress markers and inflammatory markers would allow more specific conclusions to be drown concerning the effect of radiation on vascular function. Fourth, the present study may have some selection bias. Atomic bomb survivors who came to the outpatient clinic at Hiroshima University Hospital were included, but those who were too frail to come to the outpatient clinic or had severe cognitive dysfunction that prevented them from consenting to this study were excluded. It is possible that only atomic bomb survivors in good physical or/and cognitive condition were included in this study. Therefore, the results of this study can be adapted to subjects who maintain activities of daily living that allow them to visit an outpatient clinic. Fifth, assessment of vascular structure in the present study by using brachial IMT and the use of small vessels may have made it difficult to detect differences. However, there was no significant difference in baPWV, as an index of vascular function and vascular structure. These findings suggest that there was no significant difference in vascular structure.
## Conclusion
There was no significant difference in vascular function or vascular structure between control subjects and atomic bomb survivors even after adjustment for cardiovascular risk factors. The data suggested that endothelial function might decrease in relation to increase in radiation dose within a relatively low range of doses of radiation from the atomic bomb. Further studies are needed to confirm the findings of this study in a large and longitudinal trial.
## 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 Review Board of Hiroshima University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
SK and YHi contributed to the study design and writing of the manuscript. SK, NO, TM, ST, AM, FY, AF, TU, MK, TY, TH, YHa, YN, SH, SY, CG, and AN performed the data collection. SK and KY performed statistical analyses after discussion with all authors. YN revised the manuscript critically for important intellectual content. All authors contributed to interpretation of data and review of the manuscript and 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/fcvm.2023.1122794/full#supplementary-material
## References
1. Shimizu Y, Kodama K, Nishi N, Kasagi F, Suyama A, Soda M. **Radiation exposure and circulatory disease risk: Hiroshima and Nagasaki atomic bomb survivor data, 1950-2003.**. (2010) **340**. DOI: 10.1136/bmj.b5349
2. Yamada M, Wong F, Fujiwara S, Akahoshi M, Suzuki G. **Noncancer disease incidence in atomic bomb survivors, 1958-1998.**. (2004) **161** 622-32. PMID: 15161358
3. Takahashi I, Abbott R, Ohshita T, Takahashi T, Ozasa K, Akahoshi M. **A prospective follow-up study of the association of radiation exposure with fatal and non-fatal stroke among atomic bomb survivors in Hiroshima and Nagasaki (1980-2003).**. (2012) **2**. DOI: 10.1136/bmjopen-2011-000654
4. Tran V, Zablotska L, Brenner A, Little M. **Radiation-associated circulatory disease mortality in a pooled analysis of 77,275 patients from the Massachusetts and Canadian tuberculosis fluoroscopy Cohorts.**. (2017) **7**. DOI: 10.1038/srep44147
5. Hayashi T, Morishita Y, Khattree R, Misumi M, Sasaki K, Hayashi I. **Evaluation of systemic markers of inflammation in atomic-bomb survivors with special reference to radiation and age effects.**. (2012) **26** 4765-73. DOI: 10.1096/fj.12-215228
6. Kusunoki Y, Yamaoka M, Kubo Y, Hayashi T, Kasagi F, Douple E. **T-Cell immunosenescence and inflammatory response in atomic bomb survivors.**. (2010) **174** 870-6. DOI: 10.1667/rr1847.1
7. Libby P. **Inflammation in atherosclerosis.**. (2002) **420** 868-74. DOI: 10.1038/nature01323
8. Sasaki H, Wong F, Yamada M, Kodama K. **The effects of aging and radiation exposure on blood pressure levels of atomic bomb survivors.**. (2002) **55** 974-81. DOI: 10.1016/s0895-435600439-0
9. Wong F, Yamada M, Sasaki H, Kodama K, Hosoda Y. **Effects of radiation on the longitudinal trends of total serum cholesterol levels in the atomic bomb survivors.**. (1999) **151** 736-46. PMID: 10360794
10. Tatsukawa Y, Cordova K, Yamada M, Ohishi W, Imaizumi M, Hida A. **Incidence of diabetes in the atomic bomb survivors: 1969-2015.**. (2022) **107** e2148-55. DOI: 10.1210/clinem/dgab902
11. Darby S, Ewertz M, McGale P, Bennet A, Blom-Goldman U, Brønnum D. **Risk of ischemic heart disease in women after radiotherapy for breast cancer.**. (2013) **368** 987-98. DOI: 10.1056/NEJMoa1209825
12. Clarke M, Collins R, Darby S, Davies C, Elphinstone P, Evans V. **Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: an overview of the randomised trials.**. (2005) **366** 2087-106. DOI: 10.1016/s0140-673667887-7
13. Andreassi M, Piccaluga E, Gargani L, Sabatino L, Borghini A, Faita F. **Subclinical carotid atherosclerosis and early vascular aging from long-term low-dose ionizing radiation exposure: a genetic, telomere, and vascular ultrasound study in cardiac catheterization laboratory staff.**. (2015) **8** 616-27. DOI: 10.1016/j.jcin.2014.12.233
14. Ross R. **Atherosclerosis–an inflammatory disease.**. (1999) **340** 115-26. DOI: 10.1056/nejm199901143400207
15. Higashi Y, Noma K, Yoshizumi M, Kihara Y. **Endothelial function and oxidative stress in cardiovascular diseases.**. (2009) **73** 411-8. DOI: 10.1253/circj.cj-08-1102
16. Benjamin E, Larson M, Keyes M, Mitchell G, Vasan R, Keaney J. **Clinical correlates and heritability of flow-mediated dilation in the community: the Framingham heart study.**. (2004) **109** 613-9. DOI: 10.1161/01.cir.0000112565.60887.1e
17. Kishimoto S, Maruhashi T, Kajikawa M, Harada T, Yamaji T, Han Y. **White blood cell count is not associated with flow-mediated vasodilation or nitroglycerine-induced vasodilation.**. (2022) **12**. DOI: 10.1038/s41598-022-12205-5
18. Celermajer D, Sorensen K, Gooch V, Spiegelhalter D, Miller O, Sullivan I. **Non-invasive detection of endothelial dysfunction in children and adults at risk of atherosclerosis.**. (1992) **340** 1111-5. PMID: 1359209
19. Corretti M, Anderson T, Benjamin E, Celermajer D, Charbonneau F, Creager M. **Guidelines for the ultrasound assessment of endothelial-dependent flow-mediated vasodilation of the brachial artery: a report of the international brachial artery reactivity task force.**. (2002) **39** 257-65. DOI: 10.1016/s0735-109701746-6
20. Iwamoto Y, Maruhashi T, Fujii Y, Idei N, Fujimura N, Mikami S. **Intima-media thickness of brachial artery, vascular function, and cardiovascular risk factors.**. (2012) **32** 2295-303. DOI: 10.1161/atvbaha.112.249680
21. Maruhashi T, Soga J, Fujimura N, Idei N, Mikami S, Iwamoto Y. **Endothelial dysfunction, increased arterial stiffness, and cardiovascular risk prediction in patients with coronary artery disease: Fmd-J (flow-mediated dilation Japan) study A.**. (2018) **7**. DOI: 10.1161/jaha.118.008588
22. Kurihara M, Munaka M, Hayakawa N, Yamamoto H, Ueoka H, Ohtaki M. **Mortality statistics among atomic bomb survivors in Hiroshima prefecture, 1968-1972.**. (1981) **22** 456-71. DOI: 10.1269/jrr.22.456
23. **American Diabetes Association: clinical practice recommendations 1999.**. (1999) **22** S1-114. PMID: 10333970
24. **2. Classification and diagnosis of diabetes.**. (2017) **40** S11-24. DOI: 10.2337/dc17-S005
25. **Executive summary of the third report of the national cholesterol education program (Ncep) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (adult treatment panel III).**. (2001) **285** 2486-97. PMID: 11368702
26. Maruhashi T, Soga J, Fujimura N, Idei N, Mikami S, Iwamoto Y. **Nitroglycerine-induced vasodilation for assessment of vascular function: a comparison with flow-mediated vasodilation.**. (2013) **33** 1401-8. DOI: 10.1161/atvbaha.112.300934
27. Thijssen D, Bruno R, van Mil A, Holder S, Faita F, Greyling A. **Expert consensus and evidence-based recommendations for the assessment of flow-mediated dilation in humans.**. (2019) **40** 2534-47. DOI: 10.1093/eurheartj/ehz350
28. Hoshi M, Matsuura M, Hayakawa N, Ito C, Kamada N. **Estimation of radiation doses for atomic-bomb survivors in the hiroshima university registry.**. (1996) **70** 735-40. DOI: 10.1097/00004032-199605000-00017
29. Maruhashi T, Kajikawa M, Kishimoto S, Hashimoto H, Takaeko Y, Yamaji T. **Vascular function is further impaired in subjects aged 80 years or older.**. (2020) **43** 914-21. DOI: 10.1038/s41440-020-0435-z
30. Nakamizo T, Cologne J, Cordova K, Yamada M, Takahashi T, Misumi M. **Radiation effects on atherosclerosis in atomic bomb survivors: a cross-sectional study using structural equation modeling.**. (2021) **36** 401-14. DOI: 10.1007/s10654-021-00731-x
31. Kamiya K, Ozasa K, Akiba S, Niwa O, Kodama K, Takamura N. **Long-term effects of radiation exposure on health.**. (2015) **386** 469-78. DOI: 10.1016/s0140-673661167-9
32. Cervelli T, Panetta D, Navarra T, Andreassi M, Basta G, Galli A. **Effects of single and fractionated low-dose irradiation on vascular endothelial cells.**. (2014) **235** 510-8. DOI: 10.1016/j.atherosclerosis.2014.05.932
33. Mitchel R, Hasu M, Bugden M, Wyatt H, Little M, Gola A. **Low-dose radiation exposure and atherosclerosis in ApoE**. (2011) **175** 665-76. DOI: 10.1667/rr2176.1
34. Large M, Hehlgans S, Reichert S, Gaipl U, Fournier C, Rödel C. **Study of the anti-inflammatory effects of low-dose radiation: the contribution of biphasic regulation of the antioxidative system in endothelial cells.**. (2015) **191** 742-9. DOI: 10.1007/s00066-015-0848-9
35. Ebrahimian T, Beugnies L, Surette J, Priest N, Gueguen Y, Gloaguen C. **Chronic exposure to external low-dose gamma radiation induces an increase in anti-inflammatory and anti-oxidative parameters resulting in atherosclerotic plaque size reduction in Apoe(-/-) mice.**. (2018) **189** 187-96. DOI: 10.1667/rr14823.1
36. Hayashi T, Morishita Y, Kubo Y, Kusunoki Y, Hayashi I, Kasagi F. **Long-term effects of radiation dose on inflammatory markers in atomic bomb survivors.**. (2005) **118** 83-6. DOI: 10.1016/j.amjmed.2004.06.045
37. Sera N, Hida A, Imaizumi M, Nakashima E, Akahoshi M. **The association between chronic kidney disease and cardiovascular disease risk factors in atomic bomb survivors.**. (2013) **179** 46-52. DOI: 10.1667/rr2863.1
|
---
title: An integrated co-expression network analysis reveals novel genetic biomarkers
for immune cell infiltration in chronic kidney disease
authors:
- Jia Xia
- Yutong Hou
- Anxiang Cai
- Yingjie Xu
- Wen Yang
- Masha Huang
- Shan Mou
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9981626
doi: 10.3389/fimmu.2023.1129524
license: CC BY 4.0
---
# An integrated co-expression network analysis reveals novel genetic biomarkers for immune cell infiltration in chronic kidney disease
## Abstract
### Background
Chronic kidney disease (CKD) is characterized by persistent damage to kidney function or structure. Progression to end-stage leads to adverse effects on multiple systems. However, owing to its complex etiology and long-term cause, the molecular basis of CKD is not completely known.
### Methods
To dissect the potential important molecules during the progression, based on CKD databases from Gene Expression Omnibus, we used weighted gene co-expression network analysis (WGCNA) to identify the key genes in kidney tissues and peripheral blood mononuclear cells (PBMC). Correlation analysis of these genes with clinical relevance was evaluated based on Nephroseq. Combined with a validation cohort and receiver operating characteristic curve (ROC), we found the candidate biomarkers. The immune cell infiltration of these biomarkers was evaluated. The expression of these biomarkers was further detected in folic acid-induced nephropathy (FAN) murine model and immunohistochemical staining.
### Results
In total, eight genes (CDCP1, CORO1C, DACH1, GSTA4, MAFB, TCF21, TGFBR3, and TGIF1) in kidney tissue and six genes (DDX17, KLF11, MAN1C1, POLR2K, ST14, and TRIM66) in PBMC were screened from co-expression network. Correlation analysis of these genes with serum creatinine levels and estimated glomerular filtration rate from Nephroseq showed a well clinical relevance. Validation cohort and ROC identified TCF21, DACH1 in kidney tissue and DDX17 in PBMC as biomarkers for the progression of CKD. Immune cell infiltration analysis revealed that DACH1 and TCF21 were correlated with eosinophil, activated CD8 T cell, activated CD4 T cell, while the DDX17 was correlated with neutrophil, type-2 T helper cell, type-1 T helper cell, mast cell, etc. FAN murine model and immunohistochemical staining confirmed that these three molecules can be used as genetic biomarkers to distinguish CKD patients from healthy people. Moreover, the increase of TCF21 in kidney tubules might play important role in the CKD progression.
### Discussion
We identified three promising genetic biomarkers which could play important roles in the progression of CKD.
## Introduction
Chronic kidney disease (CKD) is a global public health issue with a prevalence of $13.4\%$ [1]. It is clinically defined as renal structure abnormalities or dysfunction (estimated glomerular filtration rate, eGFR < 60 ml/min/1.73m2) that has persisted for more than 3 months [2]. CKD has a poor prognosis and can easily progress to end-stage renal disease (ESRD). Renal biopsy is an essential tool for diagnosing the pathology of CKD. Owing to the complex etiology and long-term cause, the molecular basis of CKD is not completely known, and it is difficult to predict patient responses to treatment [3, 4]. With the development of omics technologies in the past decade, transcriptional bioinformatics analyses of chronic diseases have improved our understanding of molecular processes involved in CKD and have identified some novel biomarkers (5–7). In addition, integrated analyses using appropriate methods have identified some candidate genes based on the transcriptome data from such single-gene studies [8, 9]. Kidney often suffers pathogenic immune responses against autoantigens in kidney or renal complication of systemic autoimmune diseases, which drive the renal disease, such as lupus nephropathy, membranous nephropathy and glomerulonephritis [10]. The immune cell infiltration in kidney worsens the renal function and aggravates renal fibrosis [11]. On the other hand, uremic retention solutes in blood have toxic effect on immune cells and contribute to systemic inflammatory and immune dysfunction [12]. It is valuable to give new insight into the immune cell infiltration and find novel genetic biomarker for elucidating the molecular mechanism of CKD.
However, due to the invasion, frequent renal biopsy greatly increases the risk of complications. In end-stage CKD, many uremic toxins remain in the blood. Peripheral blood mononuclear cells (PBMC) respond to such environmental stimuli and undergo functional changes [13]. Therefore, as a non-invasive method, blood tests can help identify key genes or pathways involved in CKD progression [14].
For bioinformatics analysis, the algorithm and dataset sample size have a major impact on analysis results. To our knowledge, no network analysis using multiple methods has been performed on CKD kidney tissue and PBMC samples to find key genes related with immune cell infiltration. *Weighted* gene co-expression network analysis (WGCNA) is an algorithm that can analyze gene expression profiles to filter out disease-related modules and find hub genes. It is the most commonly used algorithm when searching for genes associated with clinical characteristics [15]. In this study, we aimed to use the WGCNA algorithm in combination with differentially expressed genes (DEGs) and least absolute shrinkage and selection operator (LASSO) analysis, integrating three Gene Expression Omnibus (GEO) databases comprising PBMC and kidney tissue samples, to obtain reliable key genes related with immune cell infiltration evaluation for distinguishing the CKD patients from healthy people. In addition, we used online kidney disease databases (www.nephroseq.org), immunohistochemical (IHC) staining detection and murine CKD model for verification.
## Flowchart of bioinformatics analysis
The steps of our bioinformatics analysis are shown in the flowchart (Figure 1). Briefly, the CKD biomarker identification study was divided into the following main steps: [1] Peripheral blood mononuclear cell and renal tissue (TISSUE) data extraction and identification of DEGs; [2] Construction of co-expression network to identify hub genes; [3] *Integrated analysis* to extract key genes from DEGs and hub genes; [4] Expression validation to test credibility using validation group and external online cohort; [5] Multiple evaluation of candidate genes by online clinical database and murine experimental CKD model.
**Figure 1:** *Flowchart to identify chronic kidney disease (CKD) biomarkers, including data extraction, processing and analysis.*
## Identification of DEGs in PBMC and renal tissue from CKD datasets
First, we extracted transcriptome data from GEO datasets of CKD patients. In the clinic, blood tests and renal biopsies reveal various clinical features including abnormal kidney function, morphological changes and immunological dysfunction. In this study, GEO datasets of both PBMC and TISSUE samples were included and analyzed to explore the novel gene biomarkers in tissue and PBMC. The PBMC controls were from kidney-disease-free patients or healthy individuals and the TISSUE controls were from adjacent normal tissues of patients treated with tumor nephrectomy. To remove the batch effect, the R package ComBat was used. Principal component analysis (PCA) showed the normalized GEO samples (GSMs) from different GEO datasets (Figure 2A). GSMs were randomly divided into discovery and validation cohort (approximately a 4:1 ratio, Table S1). The DEGs had to meet the selection criteria (adjusted $P \leq 0.05$). Preliminarily, we found 30 DEGs from PBMC and 142 DEGs from TISSUE samples (Table S2 and S3). Figure 2B shows the fold-change threshold (|log10(FC)| > 0.3 for PBMC and |Log10(FC)| > 0.52 for TISSUE), and Figure 2C showed the ranking of DEGs. Overall, we identified many DEGs in PBMC and kidney tissues from CKD patients, as compared with levels in controls.
**Figure 2:** *Identification of differentially expressed genes (DEGs) in peripheral blood mononuclear cell (PBMC) and TISSUE samples (A) Principal component analysis (PCA) of the GSM datasets. The samples were visualized by scatter plots based on two principal components (PC1 and PC2) of gene expression profiles without (left) or with (right) batch effect removal. Top, PBMC; bottom, TISSUE. (B) Volcano plots of the DEGs. The orange dots meant significantly upregulated genes, and the green dots represented significantly downregulated genes. The grey dots represented non-significantly changed genes. Top, PBMC; bottom, TISSUE. (C) Heatmap showing DEGs in different samples. Left PBMC, Right TISSUE.*
## Kyoto encyclopedia of genes and genomics and gene ontology analysis of DEGs
To interpret the general biological properties of DEGs, we used STRING to create the protein–protein-interaction (PPI) network. We found a close interaction among DEGs from PMBC ($\frac{7}{30}$) and TISSUE samples ($\frac{22}{142}$) (Figures 3A, B). Furthermore, we analyzed the DEGs via Kyoto Encyclopedia of Genes and Genomics (KEGG) pathway and Gene Ontology (GO) biological process analysis. The top KEGG terms associated with PBMC and TISSUE samples are shown in Figures 3C, D. PBMC DEGs were enriched in transcriptional misregulation in cancer, ubiquinone and other terpenoid-quinone biosynthesis and other glycan degradation, indicating the possible effect of the CKD status on energy metabolism and protein glycosylation. TISSUE DEGs were mainly involved in cell–ECM interactions, such as focal adhesion and ECM-receptor interaction. In addition, several signaling pathways were enriched, including TGF-beta signal pathway, AGE-RAGE signal pathway in diabetic complications, PI3K-Akt signal pathway and Hippo signaling pathway, which have been reported to play a role in CKD progression [16, 17]. GO analysis results (Figures 3E, F) were similar to KEGG results. It can be seen that PBMC DEGs were involved in ketogenesis (cellular ketone metabolic process, ketone biosynthetic process), glycosylation processes (protein deglycosylation, hydrolase activity, hydrolyzing O-glycosyl compounds, etc.). In addition, TISSUE DEGs were enriched in ECM interactions and formation terms, such as collagen-containing extracellular matrix and complex of collagen trimers.
**Figure 3:** *Functional enrichment analysis of DEGs (A) Protein–protein interaction (PPI) network of total DEGs from PBMC. Different colors of dots in the circle plot represented different proteins. The connectivity degree was represented by dot size. The edge width was proportional to combined score. (B) PPI network of total DEGs from TISSUE samples. (C) Enriched KEGG pathways among PBMC DEGs. The gene ratio was represented on the horizontal axis. The vertical axis indicated the KEGG signaling pathway terms, and the purple-to-blue gradually changing color indicated the change of significance from low to high. (D) Circular enrichment of KEGG pathways among TISSUE DEGs (hsa04510: Focal adhesion; hsa04512: ECM-receptor interaction; hsa05146: Amoebiasis; hsa05222: Small cell lung cancer; hsa05205: Proteoglycans in cancer; hsa04350: TGF-beta signaling pathway; hsa04933: AGE-RAGE signaling pathway in diabetic complications; hsa04810: Regulation of actin cytoskeleton; hsa04151: PI3K-Akt signaling pathway; hsa01200: Carbon metabolism; hsa00650: Butanoate metabolism; hsa04971: Gastric acid secretion; hsa04710: Circadian rhythm; hsa04270: Vascular smooth muscle contraction; hsa04610: Complement and coagulation cascades; hsa05410: Hypertrophic cardiomyopathy; hsa04390: Hippo signaling pathway). (E) Enriched GO terms among PBMC DEGs. (F) Circular enrichment of GO terms among TISSUE DEGs (GO:0055006: cardiac cell development; GO:0060537: muscle tissue development; GO:0014706: striated muscle tissue development; GO:0035051: cardiocyte differentiation; GO:0048738: cardiac muscle tissue development; GO:0055013: cardiac muscle cell development; GO:0062023: collagen-containing extracellular matrix; GO:0043202: lysosomal lumen; GO:0044291: cell-cell contact zone; GO:0031252: cell leading edge; GO:0043292: contractile fiber; GO:0098644: complex of collagen trimers; GO:0008307: structural constituent of muscle; GO:0030021: extracellular matrix structural constituent conferring compression resistance; GO:0005201: extracellular matrix structural constituent; GO:0003779:actin binding; GO:0043027: cysteine-type endopeptidase inhibitor activity involved in apoptotic process; GO:1901681: sulfur compound binding).*
## WGCNA highlights and functional analysis of CKD-associated gene co-expression modules
DEGs included only the most significantly regulated genes and other regulated genes may be missing from the transcripts. Here, we use WGCNA to robustly construct multiple gene co-expression modules. Hierarchical clustering was used to define branches of the cluster dendrogram in multiple randomly color-coded modules (Figure 4A left, PBMC; Figure 4B, left, TISSUE). The heatmap of correlations between module eigengenes and clinical traits (CKD or not) is shown in Figure 4A (right, PBMC) and Figure 4B (right, TISSUE). The darkolivegreen1 module (Corresponding Correlation, CC = −0.31, $$P \leq 0.01$$) showed the highest negative correlation with CKD trait in PBMC (Figure 4A, right). The lightgreen module (CC = −0.25, $$P \leq 0.002$$) showed a high negative correlation with CKD trait in TISSUE samples (Figure 4B, right). The hub genes of each sample type were filtered from these modules that met the selection module member (MM) and gene significance (GS) criteria described in the methods (Table S4 and S5). *Hub* genes of the PBMC darkolivegreen1 module included DDX17, KLF11, MAN1C1, POLR2K, ST14, TRIM66 and 121 other genes. *Hub* genes of the TISSUE lightgreen module include CDCP1, CLIC5, CORO1C, DACH1, DENND2D, DPP6, GSTA4, MAFB, MAPK10, MYLIP, NEBL, NES, PDLIM2, PFKP, PLA2R1, TCF21, TGFBR3 and GIF1. The GO analysis of hub genes was performed on Metascape (http://metascape.org/). The top three GO terms of TISSUE hub genes were mesenchymal cell differentiation, actin cytoskeleton and DNA-binding transcription repressor activity, RNA polymerase II-specific (Figure 5A). DisNET analysis revealed that hub genes were closely related with glomerular disease (focal glomerulosclerosis and membranous glomerulonephritis) and renal carcinoma (Figure 5B). As for the hub genes of PBMC, the top three GO terms were leukocyte activation, positive regulation of cytokine production and positive regulation of leukocyte cell-cell adhesion (Figure 5C). Among them, The Molecular Complex Detection algorithm (MCODE) analysis enriched ribosome biogenesis and lymphocyte activation (Figure 5D). Above, the hub genes of TISSUE are closely correalted with DNA transcriptional dysregulation in kidney disease, while the hub genes of PBMC may actively affect the lymphocyte activation.
**Figure 4:** *Weighted gene co-expression network analysis (WGCNA) revealing gene co-expression networks in samples from CKD patients (A) WGCNA analysis of PBMC samples. The left dendrogram represented the clusters of differentially expressed genes based on different metrics. Each branch represented one gene, and each color below branches represented one co-expression module. The right heatmap showed the correlation between gene modules and CKD. The correlation coefficient in each cube represented the correlation between gene modules and traits, which decreased from red to blue. (B) WGCNA analysis of TISSUE samples.* **Figure 5:** *Functional enrichment analysis of hub genes in disease-related module (A) Enriched GO terms among TISSUE hub genes. The horizontal axis represented P-value of GO terms in log10 calculated on Metascape by default parameter. (B) Enriched DisGeNET terms among TISSUE hub genes. The horizontal axis represented P-value of GO terms in log10 calculated on Metascape. (C) Network of representative GO terms among PBMC hub genes. The clusters were calculated and visualized with Cytoscape using Metascape online platform by default parameter. The color of the node represented its cluster identity. One GO term from each cluster was selected to be shown as label. (D) Top MCODE terms of PBMC hub genes. All PPI among PBMC hub genes formed a network. The Molecular Complex Detection algorithm (MCODE) was used to identify the connected network components. The network was analyzed by GO enrichment to extract “biological meanings”. One GO term was labelled to represent the MCODE (GO: 0042254: Ribosome biogenesis; GO: 0046649: lymphocyte activation).*
## Identification and validation of common hub genes
Functional analysis based on hub genes did not exactly match the analysis based on DEGs. Therefore, to identify the key genes involved in CKD progression, we tried to define the common hub genes belonging to both DEGs and hub genes from WGCNA module (Figure 6A). 7 genes in PBMC and 14 genes in TISSUE were screened out. Using the LASSO regression algorithm, the common hub genes were reduced to six in PBMC (DDX17, KLF11, MAN1C1, POLR2K, ST14 and TRIM66) and eight in TISSUE (CDCP1, CORO1C, DACH1, GSTA4, MAFB, TCF21, TGFBR3 and TGIF1), which were identified as key genes (Figure 6B). Interestingly, DDX17 and MAN1C1 from PBMC hub genes and CORO1C, TCF21 and TGFBR3 from TISSUE hub genes were among the top-ranked genes in the PPI network. To examine whether the expression of these genes influenced CKD progression, we obtained clinical parameters (eGFR and serum creatinine, SCr) of these 8 tissue genes from www.nephroseq.org. Woroniecka Diabetes Glom database was used to evaluate the eGFR status, while the Ju CKD Glom was used to evaluate the SCr level. Consistently, the low-expressed genes, GSTA4, MAFB, TGFBR3, DACH1 and TCF21 expression was positively related with eGFR and negatively related with SCr. The high-expressed genes, CDCP1, CORO1C and TGIF1 expression was negatively related with eGFR and positively related with SCr (Figures 7A–H). The expression and significance of these hub genes were highlighted in Figure 7I.
**Figure 6:** *Common hub gene selection and least absolute shrinkage and selection operator (LASSO) analysis (A) The common hub genes shared between DEGs and hub genes were visualized in a Venn diagram. Left, PBMC; right, TISSUE. (B) The number of factors was determined by LASSO analysis. The procedure of LASSO Cox model fitting was shown in left panel. One curve represented a gene. The coefficient of each gene against the LC-norm was plotted with the lambda change. L1-norm represented the total absolute value of non-zero coefficients. A coefficient profile generated against the log (lambda) sequence was shown in the right panel. Continuous upright lines were the partial likelihood deviance ± SE; The optimal values and gene symbols were depicted, based on the minimum criteria (lambda.min, left vertical dotted line) and 1-SE criteria (lambda.1se, right vertical dotted line). Top, PBMC; bottom, TISSUE. SE, standard error.* **Figure 7:** *Correlation analysis of TISSUE common hub genes and CKD clinical parameters (A–H) The correlation of TISSUE genetic biomarker mRNA levels with estimated glomerular filtration rate (eGFR) in the Woroniecka Diabetes Glom Cohort (Top) or serum creatinine (SCr) level in the Ju CKD Glom Cohort (Bottom). (A)
CDCP1; (B)
CORO1C; (C)
TGIF1; (D)
GSTA4; (E)
MAFB; (F)
TGFBR3; (G)
DACH1; (H)
TCF21. Data were extracted from www.nephroseq.org. (I) The gene expression level of common hub genes in the discovery cohort.*
To determine the association between these genes and CKD status, we examined these 14 key genes in the validation cohort. Among them, the expression of DDX17 in PBMC, DACH1 and TCF21 in TISSUE samples showed similar characteristics with those in the discovery cohort (Figures 8A–C). Consistently, DACH1 and TCF21 mRNA level were also decreased in the Woroniecka diabetic nephropathy cohort (Figure 8D, DACH1, $P \leq 0.0001$; Figure 8E, TCF21, $P \leq 0.0001$). Moreover, as shown in Figures 8F–H, receiver operating characteristic curve (ROC) was used to investigate whether these three key genes could discriminate between healthy and CKD samples. The classification accuracy (area under the ROC curve, AUC) of these three key genes (DDX17, DACH1, TCF21) was 0.828, 0.825 and 0.981 in the discovery cohort and 0.885, 0.838 and 0.949 in the validation cohort, respectively, showing strong ability to discriminate between CKD and healthy individuals. By reason of the foregoing, we screened out DDX17, DACH1 and TCF21 as genetic biomarkers. The correlation between parameters of renal function and the expression of genetic biomarkers suggested them may play a renoprotective role. Briefly, the renal DACH1 and TCF21 expression was positively correlated with eGFR in CKD patients (DACH1, $$P \leq 7.15$$e-5, R2 = 0.554; TCF21, $$P \leq 5.34$$e-5, R2 = 0.566), whereas it was negatively correlated with SCr levels (DACH1, $$P \leq 0.0022$$, R2 = 0.366; TCF21, $$P \leq 5.04$$e-7, R2 = 0.707).
**Figure 8:** *Validation and immune infiltration evaluation of CKD biomarkers from different samples (A–C) Box plots representing expression level of CKD genetic biomarkers in the discovery and validation cohorts. (D) mRNA level of DACH1 in the Woroniecka Diabetes Glom Cohort. (E) mRNA level of TCF21 in the Woroniecka Diabetes Glom Cohort. (F–H) Receiver operating characteristic (ROC) curve for the discovery and validation cohorts. (A, F)
DDX17 (PBMC); (B, G)
DACH1 (TISSUE); (C, H)
TCF21 (TISSUE). (I) Correlation heatmap demonstrating the relationship between DDX17 (PBMC) and immune cells infiltration. (J) Correlation heatmap demonstrating the relationship between DACH1 and TCF21 (TISSUE) and immune cells infiltration. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.*
## Immune cell infiltration analysis of genetic biomarkers
The TISSUE genetic biomarkers DACH1 and TCF21 are tumor suppressor genes. They are low expressed in the Kidney Renal Clear Cell Carcinoma (KIRC) (Figures S1A, B). The KIRC patients with high DACH1 and TCF21 expression has a better prognosis (Figures S1C, D). On the other hands, PBMC hub genes are closely related with lymphocyte activation (leukocyte activation, positive regulation of cytokine production, positive regulation of leukocyte cell-cell adhesion, etc.) ( Figure S1E). The chronic kidney diseases and kidney malignant tumors have been demonstrated to be linked to a more severe immune cell infiltration. We evaluate the immune cell infiltration of these genetic biomarkers. Immune cell infiltration analysis revealed that DDX17 in PBMC was found to be correlated with neutrophil, type 1 helper cell, type 2 helper cell and mast cell (Figure 8I). The TISSUE DACH1 and TCF21 were both found to be correlated with eosinophil, activated CD8 T cell and activated CD4 T cell (Figure 8J).
## CKD murine model and immunohistochemical validation of genetic biomarker expression
To mimic renal dysfunction and tubulointerstitial fibrosis status during CKD, we developed a folic acid (FA)-induced CKD murine model. The results showed that SCr levels were increased significantly after a 3-day FA treatment (Figure 9A). At day 28, dach1 mRNA levels in the kidneys of mice treated with FA were significantly decreased (Figure 9B, $P \leq 0.001$) and negatively correlated with SCr levels (Figure 9C, $$P \leq 0.0264$$, R2 = 0.4800). Conversely, tcf21 mRNA levels were significantly increased (Figure 9D, $P \leq 0.0001$) and positively correlated with SCr levels (Figure 9E, $$P \leq 0.0128$$, R2 = 0.5595). In isolated PBMC, ddx17 mRNA levels were significantly downregulated in the FA-treated kidney (Figure 9F, $P \leq 0.01$). These results showed that in the FA-induced CKD mouse model, dach1 levels in tissue and ddx17 levels in PBMC were in agreement with the changes in clinical samples. Interestingly, tcf21 was significantly increased in the FA-induced CKD mouse model, opposite to the transcriptome change in clinical samples. Furtherly, IHC staining showed the protein expression of DACH1 and TCF21 in kidney of CKD patients and normal control. Both DACH1 and TCF21 were mainly expressed in the nucleus of glomerulus and renal tubule cells. DACH1 expression decreased in glomerulus and renal tubule of membranous nephropathy (MN) (Figures 9G, H, $P \leq 0.05$), while TCF21 expression increased in MN, especially in the renal tubule cells (Figures 9I, J, $P \leq 0.05$).
**Figure 9:** *Expression of genetic biomarkers in folic acid (FA)-induced CKD murine model and membranous nephropathy (MN) patient’s biopsy (A) SCr level in FA-induced CKD murine model. Blood serum was harvested 3 days after FA injection. (B) mRNA level of dach1 in kidneys from FA-induced CKD murine model. (C) Correlation between dach1 mRNA level in kidney tissue and SCr from FA-induced CKD murine model. (D) mRNA level of tcf21 in kidney tissues from FA-induced CKD murine model. (E) Correlation between tcf21 mRNA level in kidney tissues and SCr from FA-induced CKD murine model. (F) mRNA level of ddx17 in PBMC from FA-induced CKD murine model. (G) IHC staining of DACH1 in normal kidney tissues and CKD tissues. The representative pictures of CKD were from the membranous nephropathy (MN) patient’s biopsy. (H) Quantitative results of DACH1 expression in panel (G). The expression level was calculated by average integrated density (Intden) of positive area. (I) IHC staining of TCF21 in normal kidney tissues and CKD tissues. (J) Quantitative results of TCF21 expression in (I). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.*
## Discussion
CKD is a type of kidney disease in which kidney function deteriorates and/or the structure is abnormal. Owing to its heterogeneity in etiology, the mechanism underlying its occurrence and progression is still not well understood. It is known that the TGF-β/Smad pathway is usually activated in CKD [18]. However, damage to kidney function causes uremic retention solutes to remain in the body, resulting in adverse effects such as inflammation, immune dysfunction and oxidative stress (19–21), but the biological processes involved are not yet fully understood. In this study, we analyzed the gene expression profile of kidney tissues and PBMC from CKD patients, using various bioinformatics methods to find novel common genes or signaling pathways.
In the first-round screening, based on pathway enrichment analysis, signal molecules such as TGF-β, AGE-RAGE, PI3K-AKT and HIPPO play important roles in CKD progression and cross-talk with other molecules [16, 17, 22, 23]. Based on GO analysis, the major functional terms are mainly enriched in cell adhesion and cell-extracellular matrix interactions, among others, which are involved in the critical pathological processes associated with CKD [24].
It is currently believed that WGCNA is better at identifying internal functional connections among regulated genes than DEG analysis. Wang et al. discovered that high expression levels of CEBPZ, IFI16, LYAR, BRIX1, BMS1 and DDX18 in the kidneys are potential key markers of CKD occurrence and progression [8]. However, few studies combining WGCNA and clinical parameters using both kidney tissues and PBMC have been performed. In our study, 6 key genes (KLF11, MAN1C1, POLR2K, ST14, TRIM66, DDX17) in PBMC samples and 8 key genes (CDCP1, CORO1C, GSTA4, MAFB, TGFBR3, TGIF1, TCF21 and DACH1) in TISSUE samples were identified by WGCNA. Several of these genes in tissue have been reported to contribute to the pathology and molecular changes in CKD. MAFB is a transcription factor that mediates renal tubule development and macrophage maturation [25]. TGIF1 can bind to the MH1 domain of SMAD to inhibit TGF-β pathway activation [26]. Most of them have a well correlation with SCr and eGFR in the CKD database. Among the PBMC common hub genes, several genes have been reported to be associated with kidney disease or hematologic disorder. KLF11 is a Krüppel-type zinc finger protein whose deficiency enhances chemokine generation and fibrosis in murine unilateral ureteral obstruction [27] and highly induced by TGF-β [28]. It has been reported that KLF11 inhibits the activity of SMAD7 and enhances TGF-β pathway activation [29], through which KLF11 may influence the lymphocyte function. Besides, the highly upregulation of ST14 promotes cancer cell invasion via imbalanced matriptase pericellular proteolysis [30, 31]. The upregulation of ST14 in PBMC may promote the inflammatory activation of endothelial cells in blood vessel, which is a common uremia-related complication. Thus, these molecules should be studied intensively. Finally, by examining the validation cohort and ROC curve, we identified TCF21 and DACH1 in TISSUE samples and DDX17 in PBMC as potential biomarkers for CKD. The correlation of DACH1 and TCF21 with clinical parameters (eGFR and SCr) in CKD patients suggested that these molecules are potential renoprotective biomarkers in kidney tissues.
Dachshund family transcription factor 1 (DACH1) has been previously described as a tumor suppressor that can inhibit breast cancer invasion and metastasis [32]. In the past several years, many studies have indicated that DACH1 is a renal-protective molecule. GWAS analysis showed that loss of DACH1 function was a susceptibility factor for renal fibrosis [33], and DACH1 can protect against podocyte damage in diabetic nephropathy model mice [34]. Moreover, tubule-specific DACH1-knockout mice were more susceptible to renal damage and fibrosis in a FA-induced nephropathy model [33]. In our study, this gene was identified through transcriptome bioinformatics analysis and confirmed based on clinical parameters, FA-induced nephropathy model and the IHC staining in kidney tissues of CKD patient. This agreement with reported studies showed that our bioinformatics analysis was reliable.
Transcription factor 21 (TCF21) is a transcription factor that plays an important role in the differentiation of mesenchymal cells and the development and maturation of gonads, muscle, kidney and other organs [35]. TCF21-knockout mice developed kidney dysplasia at the embryonic stage and die after birth [36]. Mice with podocyte-specific TCF21-knockout spontaneously developed proteinuria and exhibit FSGS (focal segmental glomerulosclerosis) lesions [37]. Our bioinformatic analysis showed that TCF21 mRNA levels were decreased in CKD samples and positively correlated with eGFR. However, in our FA-induced nephropathy murine model, TCF21 mRNA levels were elevated compared to control group. The TCF21 staining signal was mainly in the nucleus in the healthy kidney, while in our mild case of membranous nephropathy, the signal increased and appeared in the cytoplasm and brush border of renal tubules. Such inconsistency may be caused by the following reasons. First, the protective role of TCF21 was mainly reported in the podocytes, not in the kidney tubules. This finding is supported by analyzing the tissue datasets of CKD cohort mainly from glomerular transcriptome (GSE47183 with 100 glomerular transcriptome samples and GSE66494 with 41 whole kidney transcriptome samples). Combining the two datasets may bias the display of key genes in the glomeruli during CKD progression. Recent study identified TCF21 as a deactivation factor of fibrogenic HSCs in liver fibrosis [38]. It might be a nephroprotective in tubulointerstitial fibrosis. In our mouse model, high dose of folic acid mainly destroys the tubules and leads to the consequent tubulointerstitial fibrosis. Therefore, in the acute injury stage and early CKD stage of our mouse model, the upregulated TCF21 in tubule could play a protective role against kidney injury. In fact, TCF21 protein levels were also elevated in the early stage of diabetic nephropathy in model mice [39]. Second, TCF21 was normally expressed in nuclei of podocytes and highly accumulate in both nuclei and cytoplasma of the injured podocytes in glomerular diseases, even was detected in urine [40]. The severe injury of podocytes in CKD might lead to the loss of this cell population which lead to the reduction of total TCF21 mRNA in kidney tissue. It is in line with our bioinformatics analysis and previous reports.
Among key genes from PBMC, we identified DDX17, which an ATP-dependent RNA helicase that is a coactivator of DNA-regulated transcription factors and is involved in mRNA transcription, splicing and maturation [41]. DDX17 is an immune-related gene defined in immunology database and analysis portal (ImmPort). In our study, it was expressed at low levels in the PBMC of CKD patients from the GEO database and in the PBMC of FA-induced CKD mouse model. In PBMC, DDX17 plays an important role in innate immunity against virus invasion (42–44). DDX17 is an essential mediator of sterile NLRC4 inflammasome activation [45]. Given the fact that the uremia-associated immune deficiency is a well-known complication of CKD and it increase the risk of virus-infection and virus-associated cancers [46], the low DDX17 level in PBMC might associated with CKD progression.
As for the interrelationship of these three biomarkers. DDX5/DDX17 complex can co-activate or co-repress transcription factor transcription. In kidney tissue, the expression of DDX17 was also decreased in the Woroniecka Diabetes Glom database, which was in line with the DACH1 and TCF21 expression. There is no report on the relationship between these three genes in the occurrence and development of CKD. Mechanically, we speculated that the low expression of DDX17 might further down-regulate the activity of TCF21 and DACH1, worsen the glomerular sclerosis and renal interstitial fibrosis. Given its critical role in innate immunity against virus invasion, the low DDX17 level in PBMC may aggravate immune deficiency in ESRD (42–45), and may result in chronic inflammation and increased oxidative stress to exacerbate kidney injury and loss of renal function [47].
During the course of CKD, in addition to the abnormal activation of some signaling pathways such as TGF-β/Smad and PI3K-Akt, some central molecules are also lacking. The low activity of some key molecules that regulate kidney differentiation and development might lead to dysfunction. DACH1 and TCF21 are both important transcription factors for kidney development and maturation. Some studies have demonstrated the association between their expression and CKD and illuminated their glomerular-specific roles in kidney disease [33, 34, 36, 37, 48], but the exact mechanisms underlying their dysregulation in CKD are still elusive. In our study, GO analysis of DEGs and hub genes highlighted DNA transcriptional activity in CKD kidney tissues. Besides, the evaluation of immune cell infiltration showed a positive correlation of DACH1, TCF21 expression with eosinophil, activated CD8 T cell and activated CD4 T cell. This analysis suggested that low expression of DACH1 and TCF21 may lead to pathological dysregulation through aberrant immune cell infiltration. Folic acid induced experimental nephropathy models undergo progression from acute kidney injury (AKI) to CKD with initial damage to proximal tubules, significant alterations to these two transcription factors suggest their overall impact on glomeruli and tubules.
Our study also had some limitations. For example, the included database did not distinguish between the different pathological types of CKD to find pathology-specific biomarkers. In addition, we mainly analyzed the mRNA levels of genes, not the protein level. In addition, many renal proteins can be detected noninvasively in urine. In a subsequent study, we hope to detect the protein levels of key genes under investigation.
## Clinical samples
In this study, we adopt three samples from healthy adjacent kidney tissues of individuals who were performed tumor nephrectomy, whereas other three samples as CKD group whose tissues were taken from CKD patients’ biopsy. The pathologist confirmed these biopsy samples as membranous nephropathy (MN). The patients in this study consented under the ethics committee review of Renji Hospital affiliated to Shanghai Jiao Tong University, School of Medicine.
## Folic acid-induced nephropathy murine model
C57bl/6J mice were obtained from Shanghai Jiao Tong University, School of Medicine. All mice were maintained with a 12-hour light-dark cycle and given food and water ad libitum. Procedures were performed in accordance with guidelines approved by the ethical committee of animal experiments, Shanghai Jiao Tong University, School of Medicine. Folic acid was purchased from Sangon Biotech and dissolved in 300 mM NaHCO3. 12-week-old mice were intraperitoneally injected with FA (250 mg/kg) or NaHCO3. Mice were euthanized and sacrificed 28 days after FA injection. Whole blood was collected in White’s buffer (pH 6.4). Ficoll-Paque (GE healthcare) was used to separate the PBMC according to the manufacturer’s instructions. Kidneys were collected and stored at −80°C for further study.
## Microarray data collection from GEO database
We downloaded 6 gene expression datasets (GSE142153, GSE15072, GSE70528, GSE47183, GSE6280, GSE66494) from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). According to labeling information in the platform, probes were converted into their corresponding gene symbols. The GSE142153 dataset contained 7 CKD PBMC samples and 10 control PBMC samples. The GSE15072 dataset contained 26 CKD PBMC samples and 8 control PBMC samples. The GSE70528 dataset contained 11 CKD PBMC samples. The GSE47183 dataset contained 122 CKD kidney samples. The GSE6280 dataset contained 12 control kidney samples. The GSE66494 dataset contained 53 CKD kidney samples and 8 control kidney samples. In our study, the discovery cohorts from the GSE142153, GSE15072 and GSE70528 datasets were used to build co-expression networks and identify the predominant genes in the CKD PBMC samples. The discovery cohorts from the GSE47183, GSE6280 and GSE66494 datasets were used to build co-expression networks and identify the main genes associated with CKD kidney samples. The assignment of each sample to discovery or validation cohorts was shown in Table S1. The procurement and application for all data were in accordance with the guidelines and principles of the GEO databases.
## Data preprocessing
For both sample types (PBMC and TISSUE), each dataset was combined into a data matrix. ComBat from the R package sva was used to account for batch effects [49, 50]. Whether the batch effect was removed was evaluated by PCA. The prcomp function in R was used to perform PCA. Data visualization was performed using the R packages ggplot2 (https://ggplot2.tidyverse.org) and RColorBrewer (https://cran.r-project.org/web/packages/RColorBrewer/index.html) unless otherwise noted.
## DEGs identification
The R packages limma and edgeR were used to identify the DEGs between CKD and normal samples, respectively [51, 52]. Genes with an adjusted $P \leq 0.05$ and |log10(FC)| > 0.3 were selected as DEGs in PBMC. Genes with an adjusted $P \leq 0.05$ and |log10(FC)| > 0.52 were selected as DEGs in TISSUE. The R package pheatmap was used to generate heatmaps [53].
## Protein–protein-interaction network construction
STRING database (http://string-db.org) was used to build the PPI network, where a combined score > 0.4 was considered statistically significant [54]. The CytoHubba Cytoscape plugin was used to calculate the nodes using the connectivity degree method [55]. We then used the netVisual_circle function in the R package CellChat to visualize the PPI network [56].
## Functional enrichment analysis
We used the R packages Clusterprofiler and org.Hs.eg.db for Kyoto Encyclopedia of Genes and Genomics (KEGG) and Gene Ontology (GO) analysis [57]. KEGG pathways or GO function terms with $P \leq 0.05$ were considered statistically significant. The R package circlize was used to create circos plots [58].
## Weighted gene co-expression network analysis
The R package WGCNA was used to constructed the co-expression network based on discovery cohorts [15]. To merge modules that might be similar, 0.25 was defined as the cut-off height threshold. The phenotypes (CKD) were inputted into the co-expression network and the parameters modulus characteristic gene (ME), MM and GS were calculated. ME represented the important part in the PCA of each gene module and MM represented the connection between modules and genes. Correlation coefficients ≥ 0.50 and P-values < 0.05 were considered indicative of key modules for PBMC. Correlation coefficients ≥ 0.60 and P-values < 0.05 were considered indicative of key modules for kidney tissue. In the modular-trait correlation analysis, genes with high hub modularity were considered as hub genes. *Hub* genes of PBMC met the absolute values of GS > 0.20 and MM > 0.50. *Hub* genes of kidney tissue met the absolute values of GS > 0.20 and MM > 0.60.
## Common hub gene selection and LASSO analysis
Common hub genes were defined as the overlap between DEGs and WGCNA hub genes. Venn diagrams were prepared using the R package venn. LASSO was a regression analysis algorithm that performs both gene selection and classification [59]. To select hub genes which were credibly associated with CKD, a logistic LASSO regression model was constructed based on common hub genes by R package glmnet. 10-fold cross-validation was performed for tuning parameter selection, and the partial likelihood deviance met the minimum criteria.
## Evaluation of immune cell infiltration
A set of genes that mark each infiltrating immune cell type was obtained [60]. The correlation between gene expressions and immune cells infiltration was calculated using Pearson correlation analysis.
## The Cancer Genome Atlas verification of genetic biomarkers in TISSUE
GEPIA (http://gepia.cancer-pku.cn/index.html) is a customizable functionalities website for interactive analysis and visualization based on The Cancer Genome Atlas database [61]. To further verify the two biomarkers of CKD kidney tissue, the GEPIA web server was used to plot gene expression level box plots between kidney renal clear cell carcinoma (KRIC) and normal tissues in the TCGA database. The patient data were grouped according to the transcripts per million (TPM) value. Log2 (TPM+1) was used for log-scale, and four-way analysis of variance (ANOVA) was applied. Overall survival analyses of biomarkers of kidney tissue were also performed using GEPIA.
## Validation of CKD biomarkers
The R package ggpubr was used to generate the gene expression box plot for hub biomarkers. The R package pROC was used to plot the ROC curves [62]. The AUC values were calculated to evaluate the sensitivity and specificity of model [63].
## Kidney function
Serum creatinine (SCr) levels were evaluated through colorimetric assays based on Jaffe’s reaction using deproteinized serum samples (Nanjing Jiancheng). Absorbance was measured at OD510 nm (BioTek) and analyzed accordingly.
## Immunohistochemistry staining
5 um slides cut from $4\%$ paraformaldehyde fixed and paraffin embedded kidney tissues as were obtained from pathology department of Renji Hospital. All sections were de-paraffinized followed by heat-induced antigen retrieval on a heating block in Tris-EDTA buffer, PH = 9.0 for 15 min. Primary Rabbit DACH1 antibody (Proteintech) and Rabbit TCF21 antibody (Sigma Aldrich) was used in a final solution of 1:200 overnight at 4°C. Secondary antibody was applied for 30 min at 37°C, and the color was developed using a diaminobenzidine peroxidase substrate kit (Dako REAL™ EnVision™, DAKO). Sections were then counterstained with hematoxylin, dehydrated and mounted. The expression of DACH1 and TCF21 were imaged by Leica Microscope X20.
## Real-time PCR
Mouse kidney tissues were homogenized in Trizol reagent (TianGen). Total RNA was extracted and reverse transcribed into cDNA (HiScriptIII RT SuperMix, Vazyme). Real-time PCR was performed on LightCycler480 apparatus (Roche) using SYBR Green Mix (Yeasen). Mouse gapdh was used as internal control gene. The relative gene expression was analyzed using 2−ΔΔCT method. Primers were listed: tcf21-F, cgctcacttaaggcagatcc; tcf21-R, gtcaccacttccttcaggtca; dach1-F, cctgggaaacccgtgtactc; dach1-R, agatccaccattttgcactcatt; ddx17-F, gatcgggatcgtgacaggga; ddx17-R, agtcagtcttgctacttctggat; gapdh-F, tggccttccgtgttcctac; gapdh-R, gagttgctgttgaagtcgca.
## Statistical analyses
Data were shown as the Mean ± SEM. A two-tailed independent student’s test was conducted to assess statistical significance. The significance was denoted as follows: **** $p \leq 0.0001$; *** $p \leq 0.001$; ** $p \leq 0.01$; * $p \leq 0.05$; N.S., not significant.
## 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 Ethics Committee of Renji Hospital, School of Medicine, Shanghai Jiao Tong University. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by ethical committee of animal experiments, Shanghai Jiao Tong University, School of Medicine.
## Author contributions
Conceptualization, SM and MH. Methodology, MH. Software, YH. Validation, YH and JX. Formal analysis, JX, YH and YX. Resources, AC. Writing—original draft preparation, JX. Writing—review and editing, WY, YX and SM. Visualization, JX. Supervision, SM and MH. Project administration, SM and MH. Funding acquisition, SM. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1129524/full#supplementary-material
## References
1. Romagnani P, Remuzzi G, Glassock R, Levin A, Jager KJ, Tonelli M. **Chronic kidney disease**. *Nat Rev Dis Primers* (2017) **3** 17088. DOI: 10.1038/nrdp.2017.88
2. Webster AC, Nagler EV, Morton RL, Masson P. **Chronic kidney disease**. *Lancet (Lond Engl)* (2017) **389**. DOI: 10.1016/S0140-6736(16)32064-5
3. Inrig JK, Califf RM, Tasneem A, Vegunta RK, Molina C, Stanifer JW. **The landscape of clinical trials in nephrology: a systematic review of clinicaltrials.gov**. *Am J Kidney Dis* (2014) **63**. DOI: 10.1053/j.ajkd.2013.10.043
4. O’Seaghdha CM, Fox CS. **Genome-wide association studies of chronic kidney disease: what have we learned**. *Nat Rev Nephrol* (2011) **8** 89-99. DOI: 10.1038/nrneph.2011.189
5. Ou SM, Tsai MT, Chen HY, Li FA, Tseng WC, Lee KH. **Identification of galectin-3 as potential biomarkers for renal fibrosis by RNA-sequencing and clinicopathologic findings of kidney biopsy**. *Front Med* (2021) **8**. DOI: 10.3389/fmed.2021.748225
6. Schena FP, Nistor I, Curci C. **Transcriptomics in kidney biopsy is an untapped resource for precision therapy in nephrology: A systematic review**. *Nephrol Dialysis Transplant* (2018) **33**. DOI: 10.1093/ndt/gfx211
7. Ju W, Nair V, Smith S, Zhu L, Shedden K, Song PXK. **Tissue transcriptome-driven identification of epidermal growth factor as a chronic kidney disease biomarker**. *Sci Trans Med* (2015) **7** 316ra193. DOI: 10.1126/scitranslmed.aac7071
8. Wang J, Yin Y, Lu Q, Zhao YR, Hu YJ, Hu YZ. **Identification of important modules and hub gene in chronic kidney disease based on WGCNA**. *J Immunol Res* (2022) **2022** 4615292. DOI: 10.1155/2022/4615292
9. Zhou G, Zhang X, Wang W, Zhang W, Wang H, Xin G. **Both peripheral blood and urinary miR-195-5p, miR-192-3p, miR-328-5p and their target genes PPM1A, RAB1A and BRSK1 may be potential biomarkers for membranous nephropathy**. *Med Sci Monitor* (2019) **25**. DOI: 10.12659/MSM.913057
10. Kant S, Kronbichler A, Sharma P, Geetha D. **Advances in understanding of pathogenesis and treatment of immune-mediated kidney disease: A review**. *Am J Kidney Dis* (2022) **79** 582-600. DOI: 10.1053/j.ajkd.2021.07.019
11. Chung KW, Dhillon P, Huang S, Sheng X, Shrestha R, Qiu C. **Mitochondrial damage and activation of the STING pathway lead to renal inflammation and fibrosis**. *Cell Metab* (2019) **30** 784-99.e5. DOI: 10.1016/j.cmet.2019.08.003
12. Cohen G. **Immune dysfunction in uremia 2020**. *Toxins* (2020) **12** 439. DOI: 10.3390/toxins12070439
13. Yang J, Fang P, Yu D, Zhang L, Zhang D, Jiang X. **Chronic kidney disease induces inflammatory CD40+ monocyte differentiation**. *Circ Res* (2016) **119**. DOI: 10.1161/CIRCRESAHA.116.308750
14. Bernardor J, Alioli C, Meaux MN, Peyruchaud O, Machuca-Gayet I, Bacchetta J. **Peripheral blood mononuclear cells (PBMCs) to dissect the underlying mechanisms of bone disease in chronic kidney disease and rare renal diseases**. *Curr Osteoporos Rep* (2021) **19**. DOI: 10.1007/s11914-021-00707-6
15. Langfelder P, Horvath S. **WGCNA: an r package for weighted correlation network analysis**. *BMC Bioinf* (2008) **9** 559. DOI: 10.1186/1471-2105-9-559
16. Yuan Q, Tang B, Zhang C. **Signaling pathways of chronic kidney diseases, implications for therapeutics**. *Signal Transduct Targeted* (2022) **7** 182. DOI: 10.1038/s41392-022-01036-5
17. Xu C, Wang L, Zhang Y, Li W, Li J, Wang Y. **Tubule-specific Mst1/2 deficiency induces CKD**. *J Am Soc Nephrol* (2020) **31**. DOI: 10.1681/ASN.2019101052
18. Wu W, Wang X, Yu X, Lan HY. **Smad3 signatures in renal inflammation and fibrosis**. *Int J Biol Sci* (2022) **18**. DOI: 10.7150/ijbs.71595
19. Espi M, Koppe L, Fouque D, Thaunat O. **Chronic kidney disease-associated immune dysfunctions: Impact of protein-bound uremic retention solutes on immune cells**. *Toxins* (2020) **12** 300. DOI: 10.3390/toxins12050300
20. Pletinck A, Glorieux G, Schepers E, Cohen G, Gondouin B, Van Landschoot M. **Protein-bound uremic toxins stimulate crosstalk between leukocytes and vessel wall**. *J Am Soc Nephrol* (2013) **24**. DOI: 10.1681/ASN.2012030281
21. Vanholder R, Pletinck A, Schepers E, Glorieux G. **Biochemical and clinical impact of organic uremic retention solutes: A comprehensive update**. *Toxins* (2018) **10** 33. DOI: 10.3390/toxins10010033
22. Stinghen AE, Massy ZA, Vlassara H, Striker GE, Boullier A. **Uremic toxicity of advanced glycation end products in CKD**. *J Am Soc Nephrol* (2016) **27**. DOI: 10.1681/ASN.2014101047
23. Meng XM, Nikolic-Paterson DJ, Lan HY. **TGF-β: the master regulator of fibrosis**. *Nat Rev Nephrol* (2016) **12**. DOI: 10.1038/nrneph.2016.48
24. Remuzzi G, Bertani T. **Pathophysiology of progressive nephropathies**. *New Engl J Med* (1998) **339**. DOI: 10.1056/NEJM199811123392007
25. Moriguchi T, Hamada M, Morito N, Terunuma T, Hasegawa K, Zhang C. **MafB is essential for renal development and F4/80 expression in macrophages**. *Mol Cell Biol* (2006) **26**. DOI: 10.1128/MCB.00001-06
26. Guca E, Suñol D, Ruiz L, Konkol A, Cordero J, Torner C. **TGIF1 homeodomain interacts with smad MH1 domain and represses TGF-β signaling**. *Nucleic Acids Res* (2018) **46**. DOI: 10.1093/nar/gky680
27. De Lorenzo SB, Vrieze AM, Johnson RA, Lien KR, Nath KA, Garovic VD. **KLF11 deficiency enhances chemokine generation and fibrosis in murine unilateral ureteral obstruction**. *PloS One* (2022) **17** e0266454. DOI: 10.1371/journal.pone.0266454
28. Spittau B, Wang Z, Boinska D, Krieglstein K. **Functional domains of the TGF-beta-inducible transcription factor Tieg3 and detection of two putative nuclear localization signals within the zinc finger DNA-binding domain**. *J Cell Biochem* (2007) **101**. DOI: 10.1002/jcb.21228
29. Gohla G, Krieglstein K, Spittau B. **Tieg3/Klf11 induces apoptosis in OLI-neu cells and enhances the TGF-beta signaling pathway by transcriptional repression of Smad7**. *J Cell Biochem* (2008) **104**. DOI: 10.1002/jcb.21669
30. Gao L, Liu M, Dong N, Jiang Y, Lin CY, Huang M. **Matriptase is highly upregulated in chronic lymphocytic leukemia and promotes cancer cell invasion**. *Leukemia* (2013) **27**. DOI: 10.1038/leu.2012.289
31. Chou FP, Chen YW, Zhao XF, Xu-Monette ZY, Young KH, Gartenhaus RB. **Imbalanced matriptase pericellular proteolysis contributes to the pathogenesis of malignant b-cell lymphomas**. *Am J Pathol* (2013) **183**. DOI: 10.1016/j.ajpath.2013.06.024
32. Wu K, Katiyar S, Li A, Liu M, Ju X, Popov VM. **Dachshund inhibits oncogene-induced breast cancer cellular migration and invasion through suppression of interleukin-8**. *Proc Natl Acad Sci USA* (2008) **105**. DOI: 10.1073/pnas.0802085105
33. Doke T, Huang S, Qiu C, Liu H, Guan Y, Hu H. **Transcriptome-wide association analysis identifies DACH1 as a kidney disease risk gene that contributes to fibrosis**. *J Clin Invest* (2021) **131**. DOI: 10.1172/JCI141801
34. Cao A, Li J, Asadi M, Basgen JM, Zhu B, Yi Z. **DACH1 protects podocytes from experimental diabetic injury and modulates PTIP-H3K4Me3 activity**. *J Clin Invest* (2021) **131**. DOI: 10.1172/JCI141279
35. Ao X, Ding W, Zhang Y, Ding D, Liu Y. **TCF21: a critical transcription factor in health and cancer**. *J Mol Med (Berlin Germany)* (2020) **98**. DOI: 10.1007/s00109-020-01934-7
36. Ide S, Finer G, Maezawa Y, Onay T, Souma T, Scott R. **Transcription factor 21 is required for branching morphogenesis and regulates the gdnf-axis in kidney development**. *J Am Soc Nephrol* (2018) **29**. DOI: 10.1681/ASN.2017121278
37. Maezawa Y, Onay T, Scott RP, Keir LS, Dimke H, Li C. **Loss of the podocyte-expressed transcription factor Tcf21/Pod1 results in podocyte differentiation defects and FSGS**. *J Am Soc Nephrol* (2014) **25**. DOI: 10.1681/ASN.2013121307
38. Nakano Y, Kamiya A, Sumiyoshi H, Tsuruya K, Kagawa T, Inagaki Y. **A deactivation factor of fibrogenic hepatic stellate cells induces regression of liver fibrosis in mice**. *Hepatology* (2020) **71**. DOI: 10.1002/hep.30965
39. Makino H, Miyamoto Y, Sawai K, Mori K, Mukoyama M, Nakao K. **Altered gene expression related to glomerulogenesis and podocyte structure in early diabetic nephropathy of db/db mice and its restoration by pioglitazone**. *Diabetes* (2006) **55**. DOI: 10.2337/db05-1683
40. Usui J, Yaguchi M, Yamazaki S, Takahashi-Kobayashi M, Kawamura T, Kaneko S. **Transcription factor 21 expression in injured podocytes of glomerular diseases**. *Sci Rep* (2020) **10** 11516. DOI: 10.1038/s41598-020-68422-3
41. Lamm GM, Nicol SM, Fuller-Pace FV, Lamond AI. **p72: a human nuclear DEAD box protein highly related to p68**. *Nucleic Acids Res* (1996) **24**. DOI: 10.1093/nar/24.19.3739
42. Moy RH, Cole BS, Yasunaga A, Gold B, Shankarling G, Varble A. **Stem-loop recognition by DDX17 facilitates miRNA processing and antiviral defense**. *Cell* (2014) **158**. DOI: 10.1016/j.cell.2014.06.023
43. Zhang X, An T, Zhang X, Shen T, Li H, Dou L. **DDX17 protects hepatocytes against oleic acid/palmitic acid-induced lipid accumulation**. *Biochem Biophys Res Commun* (2022) **612**. DOI: 10.1016/j.bbrc.2022.04.129
44. Bonaventure B, Goujon C. **DExH/D-box helicases at the frontline of intrinsic and innate immunity against viral infections**. *J Gen Virol* (2022) **103**. DOI: 10.1099/jgv.0.001766
45. Wang SB, Narendran S, Hirahara S, Varshney A, Pereira F, Apicella I. **DDX17 is an essential mediator of sterile NLRC4 inflammasome activation by retrotransposon RNAs**. *Sci Immunol* (2021) **6**. DOI: 10.1126/sciimmunol.abi4493
46. Betjes MG. **Immune cell dysfunction and inflammation in end-stage renal disease**. *Nat Rev Nephrol* (2013) **9**. DOI: 10.1038/nrneph.2013.44
47. Vaziri ND. **Oxidative stress in uremia: nature, mechanisms, and potential consequences**. *Semin Nephrol* (2004) **24**. DOI: 10.1016/j.semnephrol.2004.06.026
48. Takahashi-Kobayashi M, Usui J, Yaguchi M, Yamazaki S, Kawamura T, Kaneko S. **Immunohistological score of transcription factor 21 had a positive correlation with its urinary excretion and proteinuria in immunoglobulin a nephropathy**. *Histol Histopathol* (2021) **36**. DOI: 10.14670/HH-18-367
49. Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. **The sva package for removing batch effects and other unwanted variation in high-throughput experiments**. *Bioinformatics* (2012) **28**. DOI: 10.1093/bioinformatics/bts034
50. Chen C, Grennan K, Badner J, Zhang D, Gershon E, Jin L. **Removing batch effects in analysis of expression microarray data: an evaluation of six batch adjustment methods**. *PloS One* (2011) **6** e17238. DOI: 10.1371/journal.pone.0017238
51. 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** e47. DOI: 10.1093/nar/gkv007
52. Robinson MD, McCarthy DJ, Smyth GK. **edgeR: a bioconductor package for differential expression analysis of digital gene expression data**. *Bioinformatics* (2010) **26**. DOI: 10.1093/bioinformatics/btp616
53. Gu Z, Eils R, Schlesner M. **Complex heatmaps reveal patterns and correlations in multidimensional genomic data**. *Bioinformatics* (2016) **32**. DOI: 10.1093/bioinformatics/btw313
54. Szklarczyk D, Gable AL, Nastou KC, Lyon D, Kirsch R, Pyysalo S. **The STRING database in 2021: customizable protein-protein networks, and functional characterization of user-uploaded gene/measurement sets**. *Nucleic Acids Res* (2021) **49**. DOI: 10.1093/nar/gkaa1074
55. Chin CH, Chen SH, Wu HH, Ho CW, Ko MT, Lin CY. **cytoHubba: identifying hub objects and sub-networks from complex interactome**. *BMC Syst Biol* (2014) **8 Suppl 4** S11. DOI: 10.1186/1752-0509-8-S4-S11
56. Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan CH. **Inference and analysis of cell-cell communication using CellChat**. *Nat Commun* (2021) **12** 1088. DOI: 10.1038/s41467-021-21246-9
57. Yu G, Wang LG, Han Y, He QY. **clusterProfiler: an r package for comparing biological themes among gene clusters**. *OMICS* (2012) **16**. DOI: 10.1089/omi.2011.0118
58. Gu Z, Gu L, Eils R, Schlesner M, Brors B. **Circlize implements and enhances circular visualization in r**. *Bioinformatics* (2014) **30**. DOI: 10.1093/bioinformatics/btu393
59. Tibshirani R. **The lasso method for variable selection in the cox model**. *Stat Med* (1997) **16**. DOI: 10.1002/(SICI)1097-0258(19970228)16:4<385::AID-SIM380>3.0.CO;2-3
60. Charoentong P, Finotello F, Angelova M, Mayer C, Efremova M, Rieder D. **Pan-cancer immunogenomic analyses reveal genotype-immunophenotype relationships and predictors of response to checkpoint blockade**. *Cell Rep* (2017) **18**. DOI: 10.1016/j.celrep.2016.12.019
61. Tang Z, Li C, Kang B, Gao G, Li C, Zhang Z. **GEPIA: a web server for cancer and normal gene expression profiling and interactive analyses**. *Nucleic Acids Res* (2017) **45** W98-W102. DOI: 10.1093/nar/gkx247
62. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC. **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
63. Kamarudin AN, Cox T, Kolamunnage-Dona R. **Time-dependent ROC curve analysis in medical research: current methods and applications**. *BMC Med Res Methodol* (2017) **17** 53. DOI: 10.1186/s12874-017-0332-6
|
---
title: Adiponectin/leptin ratio as a predictor of acute rejection in early post-transplant
period in patients after kidney transplantation
authors:
- Karol Graňák
- Matej Vnučák
- Monika Beliančinová
- Patrícia Kleinová
- Margaréta Pytliaková
- Marián Mokáň
- Ivana Dedinská
journal: Frontiers in Medicine
year: 2023
pmcid: PMC9981627
doi: 10.3389/fmed.2023.1117819
license: CC BY 4.0
---
# Adiponectin/leptin ratio as a predictor of acute rejection in early post-transplant period in patients after kidney transplantation
## Abstract
### Introduction
Adipokines are largely involved in the regulation of immune system activity. While leptin is the main pro-inflammatory marker of adipose tissue, adiponectin is characterized by anti-inflammatory effects. The aim of our study was to determine the risk of acute graft rejection in protocol biopsy depending on the adiponectin/leptin (A/L) ratio in patients after kidney transplantation (KT).
### Materials and methods
A total of 104 patients were included in the prospective analysis, in whom the levels of adipokines were examined pre-transplant, in the 3rd month after KT and the A/L ratio was calculated. In the 3rd month after KT, all patients underwent protocol biopsy of the graft and examination of donor-specific antibodies (DSA) using the Luminex method.
### Results
After adjusting for differences in the basic characteristics of the donor and recipient, we identified a subgroup with A/L ratio < 0.5 pre-transplant [HR 1.6126, ($$P \leq 0.0133$$)] and 3 months after KT [HR 1.3150, ($$P \leq 0.0172$$)] as independent risk factor for acute graft rejection. In the subsequent specification of the rejection episode, we identified the risk ratio A/L < 0.5 before KT [HR 2.2353, ($$P \leq 0.0357$$)] and 3 months after KT [HR 3.0954, ($$P \leq 0.0237$$)] as independent risk factor for the development of acute humoral rejection with DSA positivity.
### Conclusion
This is the first study to investigate the relationship between A/L ratio and immunological risk in terms of the development of rejection changes in patients after KT. In our study, we found that A/L ratio < 0.5 is an independent risk factor for the development of acute humoral rejection and de novo DSA production in the third month after KT.
## Introduction
It is well known today that adipose tissue is involved in the production and secretion of a wide range of bioactive peptides, known as adipokines, which have a local (paracrine) but also a systemic (endocrine) effect. In addition to these efferent signals, adipose tissue receives signals from hormonal systems or the central nervous system through many receptors. Thanks to this interactive network, adipose tissue is directly involved in the coordination of biological processes, including energy metabolism and immune functions [1]. Most molecules, especially those secreted by the non-adipocyte fraction of adipose tissue, have a dominant paracrine effect. Leptin and adiponectin are today generally accepted as the only endocrine hormones of adipose tissue with a defined effect on target organs [2]. Their presence in the human body correlates with the amount of adipose tissue. Adipokines can be classified in different ways, but from the point of view of their impact on the immune system, we divide them into two groups - pro-inflammatory and anti-inflammatory [3].
Leptin is considered a major pro-inflammatory marker and shares structural homology with the interleukin (IL)-6 receptor. It participates in the activation and proliferation of granulocytes, monocytes, macrophages, dendritic cells, natural killer cells and leads to increased production of pro-inflammatory cytokines (IL-6, IL-12, tumor necrosis factor—TNF) [4]. Leptin is directly involved in the regulation of activation and differentiation of T and B cells. It supports the proliferation of naïve and memory T cells and increases the secretion of Th1 and Th17 lymphocytes. Mechanism studies have shown that leptin activates the mTOR pathway, thereby having a positive effect on CD4 + CD25 + FOXP3 + effector T cells. In addition, it stimulates the formation, maturation and survival of thymic T cells by reducing their apoptosis [5]. Its important role has also been demonstrated in the regulation of the development and function of B cells. Since the receptor for leptin is expressed on B cells, its direct effect is assumed. Deficiency of leptin signaling led to a reduction of B cells in the bone marrow and peripheral blood, and its reduced level was associated with a lower representation of pro-B and immature B cells in the bone marrow [6].
Adiponectin, as the main representative of the group of anti-inflammatory markers of adipose tissue, acts through two receptors, AdipoR1 and AdipoR2, which, among many other tissues, are also found on most cells of the immune system [7]. Functionally, adiponectin reduces the ability of macrophages to phagocytose and secrete pro-inflammatory cytokines, while increasing the production of anti-inflammatory IL-10 [3]. In endothelial cells, it blocks the expression of adhesive molecules, which results in a reduction of the diapedesis of circulating monocytes [4]. Several studies have shown that adiponectin is a negative regulator of T cell activity. It has been shown to inhibit proliferation and cytokine production of T cells and promote their apoptosis. Recent data also suggest that it is involved in the inhibition of Th1 and Th17 lymphocyte differentiation [8]. Even though the immunomodulatory effect of adiponectin on B lymphocytes is not completely clear, it has been shown to inhibit B lymphopoiesis in long-term bone marrow cultures. In addition, adiponectin stimulates B cells to secrete the peptide PEPITEM, which specifically inhibits the migration of CD4+ and CD8+ memory T cells [9].
Recent studies suggest that patients with adipose tissue dysfunction, characterized by lower adiponectin secretion compared to leptin levels, have an increased cardiometabolic risk. This fact results from an increase in systemic inflammation and oxidative stress, the occurrence of which is negatively correlated with the A/L ratio [10]. A/L ratio can be a practical marker characterizing adipose tissue dysfunction. From the results on the general population, it follows that an A/L ratio >1 can be considered normal, an A/L ratio of 0.5–1 indicates a moderate risk and an A/L ratio <0.5 a high risk [11]. Due to the significant involvement of adipokines in the regulation of immune system processes, we assume that the A/L ratio could be correlated with the occurrence of immune-related reactions in the transplanted population. If confirmed, it would be possible to identify recipients at increased risk of developing acute graft rejection. The aim of our study was to determine the risk of acute graft rejection in protocol biopsy depending on the A/L ratio in patients after KT.
## Materials and methods
Adult patients who underwent primary KT at the Martin Transplantation Center in 2018–2019 were included in our prospective monocenter study. Patients with a history of diabetes mellitus (DM) type I or II, those who died during the study period, patients who suffered from infectious complications and those who did not undergo a protocol graft biopsy (poor anatomical conditions, recurrent infections) were excluded from the follow-up. A total of 104 patients completed our prospective follow-up.
All patients in the studied sample were set on the same immunosuppressive protocol. As part of the induction protocol, they were administered antithymocyte globulin in a cumulative dose of 3.5 mg/kg of body weight, which was divided into three doses (pre-transplantation, day 1 and day 2 after KT). A triple combination was used in the prophylactic immunosuppressive regimen: tacrolimus, mycophenolic acid and corticosteroids. Methylprednisolone was applied at a dose of 500 mg intravenously pretransplantation and on the first posttransplantation day followed by a change to oral prednisone (prednisone 20 mg until the second week after KT, prednisone 15 mg until the fourth week after KT, prednisone 10 mg until 16 weeks after KT and prednisone 7.5 mg up to 12 months after KT). Mycophenolic acid was used in a total daily dose of 1,440 mg until the first month after KT, in a daily dose of 1,080 mg until the third month after KT, and then continued with a daily dose of 720 mg.
We examined the basic level of leptin, adiponectin, IL-6, and IL-10 in KT recipients at the time of flow cytometry crossmatch (FXCM), i.e., approximately 4–5 h before the transplantation. In the post-transplantation period, we examined their levels at 3 months, i.e., at the time of the protocol biopsy of the graft. We used Human Leptin Quantikine ELISA Kit, Human Total Adiponectin ELISA Kit, LEGEND MAX Human IL-6 ELISA Kit and LEGEND MAX Human IL-10 Kit to investigate the levels of adipokines and interleukins. The A/L ratio was calculated from the measured values. We consider an A/L ratio above 1.0 to be normal, an A/L ratio of 0.5–1.0 indicates a moderate risk, and an A/L ratio <0.5 a high metabolic risk [12].
At the time of KT, we recorded: basic characteristics of the donor (donor with extended criteria, cold ischemia time) and characteristics of the recipient (age, sex, length of dialysis treatment, underlying cause of kidney failure, delayed onset of graft function, panel of reactive antibodies, number of mismatches in class A, B, DR, and DQ). At 3 months after KT, we also determined anthropometric parameters (waist circumference, body mass index—BMI), glucose metabolism parameters (c-peptide and immunoreactive insulin levels), lipid profile (total cholesterol, low-density lipoprotein—LDL, high-density lipoprotein—HDL, triglycerides), vitamin D, tacrolimus level and parameters reflecting graft function as glomerular filtration rate determined using the CKD-EPI (Chronic Kidney Disease—Epidemiology Collaboration Index) formula and quantitative proteinuria from 24-h urine collection.
Protocol biopsy of the graft and examination of DSA was performed during a short hospitalization at 3 months after KT in all patients included in our study. Biopsy was performed under ultrasonographic control using an 18-gauge puncture needle. All samples were histologically evaluated by the same pathologist. We then divided the studied sample according to the result of the histological examination (based on the Banff classification from 2019) into a group with a negative result, with findings of interstitial fibrosis and tubular atrophy (IFTA), acute tubular necrosis (ATN), acute cellular rejection (ACR) including borderline changes and antibody-mediated rejection (AMR) with DSA positivity. The examination of DSA was carried out using the LUMINEX methodology, when a value of ≥500 mean fluorescence intensity (MFI) was considered a positive result.
In our study, we used a certified statistical program, MedCalc version 13.1.2. ( MedCalc Software VAT registration number BE 0809 344,640, Member of International Association of Statistical Computing, Ostend, Belgium). Using parametric (Student’s t-test) or non-parametric tests we compared continuous variables between groups; the χ2 test and Fisher’s exact test were used to analyze associations between categorical variables, as appropriate. For non-parametric tests, we used the Wilcoxon test in the first step (Table 1) to compare the group of patients before KT and 3 months after KT. In subsequent analysis (Tables 2, 3), we used the Mann–Whitney test to compare independent groups according to A/L ratio. To perform multivariate analysis, we used Cox regression Hazard model. A P-value <0.05 was considered to be statistically significant.
## Results
A total of 170 patients after primary deceased donor KT were primarily included in the study. 66 patients were excluded from the follow-up meeting the exclusion criteria that we stated in the materials and methods section. Thus, 104 patients completed our prospective follow-up.
The serum level of tacrolimus was maintained during the monitored period in the range of 10–15 ng/L during the first month after KT, then in the range of 8.0–10 ng/L until the third month after KT. We did not observe significant differences in the level of tacrolimus between the individual studied subgroups. Likewise, there was no significant difference in the daily dose of prednisone.
Table 1 summarizes the basic characteristics of the investigated file. Of the total number of patients, $63.5\%$ were men and the average age was 45 ± 11 years. When comparing the average serum levels of adipokines and interleukins at the beginning and in the third month of follow-up, we found that in the third month after KT there was a significant increase in inflammatory markers (leptin, IL-6, IL-10) and, conversely, a significant decrease in anti-inflammatory marker (adiponectin). The A/L ratio decreased significantly during this period (Table 1).
We primarily divided the studied group into three subgroups based on the A/L ratio before transplantation and in the third month after KT: 1. A/L ratio <0.5, 2. A/L ratio 0.5–1.0, 3. A/L ratio >1.0. We compared the subgroups among themselves according to parameters related to the eventual development of graft rejection. Table 2 shows the comparison before KT. We found that there were significantly more men in the subgroup with A/L ratio > 1, on the other hand, there was no difference in the age structure between the individual subgroups. In the subgroup with high metabolic risk (A/L < 0.5), we found a significantly higher BMI value compared to subgroup with A/L ratio > 1, but also with A/L ratio 0.5–1.0. In the mentioned subgroup, patients also had a significantly higher panel reactive antibodies (PRA) value. In the subgroup with a normal A/L ratio, patients showed significantly shorter cold ischemia time and shorter time spent in the dialysis program. The level of IL-6 was significantly higher in the high-risk subgroup (A/L < 0.5), and the level of IL-10, on the other hand, was significantly lower in this subgroup compared to the subgroup with A/L ratio > 1.0. We did not notice a difference in the occurrence of delayed onset of graft function (DGF) or in the representation of donors with extended criteria (Table 2).
Table 3 summarizes a comparison of these subgroups in the third month after KT. As before transplantation, there were significantly more men in the subgroup with A/L ratio > 1.0 in the third month after KT and the age structure also did not change. Patients with an A/L ratio < 0.5 had a significantly higher BMI value, but also a waist circumference value compared to other subgroups. In this high-risk subgroup, patients had a higher PRA value. Patients in the subgroup with A/L ratio > 1.0 spent a significantly shorter time in the hemodialysis program and, compared to the subgroup with A/L ratio < 0.5, had a significantly shorter time of cold ischemia. The level of IL-6 was also significantly higher in the subgroup with A/L ratio < 0.5, on the other hand, the level of IL-10 did not show any significant differences between the subgroups. Using the eGFR value, a significantly better graft function was detected at 3 months after KT in the subgroup with a normal A/L ratio compared to the other subgroups. At the same time, we identified a significantly lower incidence of AMR in this subgroup (A/L > 1.0) compared to the high-risk (A/L < 0.5) and medium-risk (A/L ratio 0.5 – 1.0) subgroups. There were no differences in the incidence of ACR (Table 3).
In the multivariate analysis, we used the Cox regression hazard model. After adjusting for differences in the basic characteristics of the donor and recipient, we identified the length of dialysis treatment more than 24 months as an independent risk factor for A/L ratio < 0.5 pre-transplant [HR 2.2727, ($$P \leq 0.0386$$)] and BMI value > 30 kg/m2 as independent risk factor for A/L ratio < 0.5 in the third month after KT [HR 3.8235, ($$P \leq 0.0386$$)] (Tables 4, 5).
In the next step, we used the Cox regression Hazard model to determine independent risk factors for the occurrence of acute rejection in protocol graft biopsy. After adjusting for differences in the baseline characteristics of the donor and recipient, we identified a subgroup with an A/L ratio < 0.5 pretransplant [HR 1.6126, ($$P \leq 0.0133$$)] and 3 months after KT [HR 1.3150, ($$P \leq 0.0172$$)] as an independent risk factor for the occurrence of acute graft rejection (ACR + AMR) (Table 6). In the subsequent specification of the rejection episode, we identified the risk ratio A/L < 0.5 before transplantation [HR 2.2353, ($$P \leq 0.0357$$)] and 3 months after KT [HR 3.0954, ($$P \leq 0.0237$$)] as an independent risk factor for the development of AMR with DSA positivity (Table 7). On the contrary, the investigated A/L ratios were not detected as independent risk or protective factors for the development of ACR (Table 8).
We evaluated the development of the A/L ratio (according to defined subgroups) during the three months of follow-up in all groups according to the histological findings in the protocol biopsy of the graft. In the individual groups, we did not find significant change in the A/L ratio before transplantation and 3 months after KT (Figure 1).
**FIGURE 1:** *Development of the A/L ratio during 3 months of follow-up in all subgroups (according to histological findings in protocol graft biopsy). IFTA, interstitial fibrosis and tubular atrophy; ATN, acute tubular necrosis; ACR, acute cellular rejection; AMR, antibody-mediated rejection; M, month.*
Finally, we performed the ROC curve analysis for A/L ratio month 3 as a predictor for AMR 3 months after KT with sensitivity 100, specificity 65.2 and criterion ≤ 0.89 (Figure 2).
**FIGURE 2:** *ROC curve analysis (A/L ratio in month 3 as a predictor for AMR 3 months after kidney transplantation). ROC, receiver operating characteristic; AUC, area under the curve.*
## Discussion
To our knowledge, this is the first study that investigated the A/L ratio in patients after KT in the context of the risk of developing acute graft rejection. In our work, we found that the A/L ratio < 0.5 pre-transplantation and 3 months after KT represents an independent risk factor for the finding of acute graft rejection in protocol biopsy. At the same time, we specified that the risk ratio A/L < 0.5 was significantly correlated with the development of AMR in protocol biopsy with de novo DSA production. This finding was clearly supported by the result of the ROC curve analysis with AUR 0.898, which confirmed significantly larger probability of developing AMR in the group with high-risk A/L ratio.
The A/L ratio can be considered as an indicator of adipose tissue dysfunction and the balance between these adipokines may very likely play an important role in the clinical outcome in this group of patients. Monitoring of adipose tissue hormones in the transplant population has only a short history, and until now they have been investigated as separate variables in correlation with cardiometabolic risk factors. In our recently published work, we found that a high-risk A/L ratio (<0.5) was significantly associated with the occurrence of post-transplant diabetes mellitus (PTDM) and pre-diabetic conditions 1 year after KT [13].
As a pro-inflammatory marker, leptin largely alters the adaptive immune system by activating CD4+ T lymphocytes and negative signaling for CD25+ T regulatory cells [14]. Joffre et al. in their work, they assume that the inactivity of regulatory T cells can lead to graft loss, as stimulated CD4 + CD25 + Foxp3 regulatory T cells prevent rejection of the transplanted organ [12]. However, this claim has not yet been confirmed by clinical studies. In previous years, worse graft survival was detected in obese patients with hyperleptinemia. Authors Moraes-Vieira et al. they searched for a possible immunological basis in mouse models. An interesting finding was that obese mice that were leptin-deficient showed better survival of skin grafts, which indicated that the transplant outcome observed in obese patients may not be directly related to obesity but to hyperleptinemia. The authors therefore focused on the immunological background and found that CD4+ T cells differentiated more efficiently into T regulatory lymphocytes and showed a lower degree of proliferation in vivo, which was ultimately highly likely the cause of better graft survival in these mice [15]. No association between leptin and systemic inflammation in patients after KT has been found in observational studies conducted so far [16]. On the contrary, in a study on the general population, the conclusions of which were published in 2017 by Fruhbeck et al. a strong negative correlation was found between c-reactive protein level, serum amyloid A and A/L ratio. These findings may indicate that the A/L ratio reflects adipose tissue dysfunction-induced systemic inflammation [10]. Work on the anti-inflammatory adiponectin suggests that it is involved in the activation of nuclear factor κB (NF-κB) transcription. NF-κB is a protein kinase that regulates the immune system through the activity of T cells and plays an important role in acute rejection of a vascularized organ or in the etiopathogenesis of several autoimmune diseases. The authors of Vu et al. evaluated the association between NF-κB gene polymorphisms and the outcome of the transplant itself in a sample of 607 Hispanics after KT. Recipients with the NF-κB1 polymorphism had significantly fewer biopsy-verified acute graft rejections [17]. Alam et al. in 2013 investigated graft survival in 987 patients after KT based on the adiponectin level. However, an increased level of the protective acting adiponectin was not associated with better graft survival [18].
Fonseca et al. in 2015 were the first to investigate the clinical significance of adipokines in the context of graft dysfunction in patients after KT and not in the context of cardiometabolic complications. In a sample of 40 adult patients who underwent KT, the relationship of leptin, adiponectin levels with DGF and acute rejection were evaluated. Serum levels of adipokines were measured before transplantation and subsequently in the first 7 days after KT. Leptin level was significantly higher in the group of patients who developed DGF compared to those who had prompt onset of graft function. Even serum leptin on the first day after KT predicted DGF slightly better than serum creatinine. Conversely, adiponectinemia was not significantly higher in graft dysfunction and was not a predictor of DGF. In the mentioned study, the authors also monitored the possible prediction of the development of acute graft rejection and the formation of anti-human leukocyte antigen (HLA) antibodies based on leptinemia, but no significant predictive value was found. A possible reason was also the minimal number of patients with acute rejection in the study [19]. The importance of adiponectinemia in predicting the development of graft function after KT was presented in the study published by Roos et al. In 206 patients, they examined the level of total adiponectin and the high molecular weight multimer, as its main active form, before transplantation. At a 36-month follow-up, both forms were significantly associated with markers of endothelial dysfunction, arteriosclerosis, and at the same time significantly predicted graft survival. This inverse association between adiponectin and graft survival may be explained, at least in part, by its protective effects on endothelial cells and vascular inflammation [20].
A secondary finding in our study was that recipients who showed a pre-transplant high-risk A/L ratio < 0.5 spent a significantly longer time in a chronic hemodialysis program. The cause is probably chronic inflammation with oxidative stress, which are one of the basic determinants of cardiovascular morbidity and mortality in long-term dialysis patients [21]. An unsurprising finding was a significantly higher representation of obese patients (BMI > 30 kg/m2) in the group with A/L ratio < 0.5 three months after KT.
Based on our findings, we assume the importance of the level of adipokines in transplant patients, not only for the occurrence of already known cardiometabolic complications, but especially in the context of the development of rejection changes and DSA production. The results of our study indicate that the evaluation of the ratio of these hormones with the opposite effect has a greater clinical significance than monitoring them separately. In clinical practice, the monitoring of the A/L ratio can be an important early predictor of risk groups of recipients for the development of rejection changes, DSA production and, probably, resulting worse function or survival of the graft. However, these claims will require further studies with longer follow-up and a larger sample of patients.
The limitation of this study is the low number of patients included in the individual monitored subgroups. On the other hand, this is the first study that deals with this issue and therefore we consider the sample size for this pilot project to be acceptable. Another limitation may be the absence of previous works on a transplanted sample of patients, so we designed our study based on the findings in the general population cohorts.
## Conclusion
This is the first study to investigate the relationship between A/L ratio and immunological risk in terms of the development of rejection changes in patients after KT. In our study, we found that A/L ratio < 0.5 is an independent risk factor for the development of AMR and de novo DSA production in the third month after KT. Based on our findings, we attribute an importance to adipokines not only in the occurrence of metabolic, but especially immunological complications with a possible impact on the survival of grafts. A/L ratio can be an important early indicator of risk groups of patients undergoing KT. Further studies in a larger sample of patients will be needed to confirm our findings.
## 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
Informed consent for included participants was checked and approved by University Hospital’s and Jessenius Faculty of Medicine’s in Martin, Slovakia, Ethical Committees (EK $\frac{33}{2018}$). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
KG and ID participated in writing the manuscript, performing of the research, and data analysis. MV, MB, PK, and MM participated in data collection. MP participated in performing the research. 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. Kershaw E, Flier J. **Adipose tissue as an endocrine organ.**. (2004) **89** 2548-56. PMID: 15181022
2. Scheja L, Heeren J. **The endocrine function of adipose tissues in health and cardiometabolic disease.**. (2019) **15** 507-24. PMID: 31296970
3. Wensveen F, Valentiæ S, Šestan M, Wensveen T, Poliæ B. **Interactions between adipose tissue and the immune system in health and malnutrition.**. (2015) **27** 322-33. DOI: 10.1016/j.smim.2015.10.006
4. Trzeciak-Ryczek A, Tokarz-Deptula B, Deptula W. **Adipocytokines affecting the immune system – selected data.**. (2011) **36** 92-4
5. Song J, Deng T. **The adipocyte and adaptive immunity.**. (2020) **11**. DOI: 10.3389/fimmu.2020.593058
6. Fujita Y, Yanagida H, Mimori T, Jin Z, Sakai T, Kawanami T. **Prevention of fasting-mediated bone marrow atrophy by leptin administration.**. (2012) **273** 52-8. DOI: 10.1016/j.cellimm.2011.11.007
7. Barnes M, Carson M, Nair M. **Non-traditional cytokines: how catecholamines and adipokines influence macrophages in immunity, metabolism and the central nervous system.**. (2015) **72** 210-9. DOI: 10.1016/j.cyto.2015.01.008
8. Zhang K, Guo Y, Ge Z, Zhang Z, Da Y, Li W. **Adiponectin suppresses T helper 17 cell differentiation and limits autoimmune CNS inflammation via the SIRT1/PPARgamma/RORgammat pathway.**. (2017) **54** 4908-20. DOI: 10.1007/s12035-016-0036-7
9. Chimen M, McGettrick H, Apta B, Kuravi S, Yates C, Kennedy A. **Homeostatic regulation of T cell trafficking by a B cell-derived peptide is impaired in autoimmune and chronic inflammatory disease.**. (2015) **21** 467-75. DOI: 10.1038/nm.3842
10. Frühbeck G, Catalán V, Rodríguez A, Ramírez B, Becerril S, Salvador J. **Involvement of the leptin-adiponectin axis in inflammation and oxidative stress in the metabolic syndrome.**. (2017) **7**. DOI: 10.1038/s41598-017-06997-0
11. Frühbeck G, Catalán V, Rodríguez A, Gómez-Ambrosi J. **Adiponectin-leptin ratio: a promising index to estimate adipose tissue dysfunction. Relation with obesity-associated cardiometabolic risk.**. (2018) **7** 57-62. DOI: 10.1080/21623945.2017.1402151
12. Joffre O, Santolaria T, Calise D, Al Saati T, Hudrisier D, Romagnoli P. **Prevention of acute and chronic allograft rejection with CD4+CD25+Foxp3+ regulatory T lymphocytes.**. (2008) **14** 88-92. PMID: 18066074
13. Graòák K, Vnuèák M, Belianèinová M, Kleinová P, Pytliaková M, Miklušica J. **Adiponectin/Leptin ratio as an index to determine metabolic risk in patients after kidney transplantation.**. (2022) **58**. DOI: 10.3390/medicina58111656
14. De Rosa V, Procaccini C, Calì G, Pirozzi G, Fontana S, Zappacosta S. **A key role of leptin in the control of regulatory T cell proliferation.**. (2007) **26** 241-55. DOI: 10.1016/j.immuni.2007.01.011
15. Moraes-Vieira P, Bassi E, Larocca R, Castoldi A, Burghos M, Lepique A. **Leptin deficiency modulates allograft survival by favoring a Th2 and a regulatory immune profile. [corrected].**. (2013) **13** 36-44. PMID: 23016759
16. Nagy K, Nagaraju S, Rhee C, Mathe Z, Molnar M. **Adipocytokines in renal transplant recipients.**. (2016) **9** 359-73. PMID: 27274819
17. Vu D, Tellez-Corrales E, Sakharkar P, Kissen M, Shah T, Hutchinson I. **Impact of NF-κB gene polymorphism on allograft outcome in Hispanic renal transplant recipients.**. (2013) **28** 18-23. DOI: 10.1016/j.trim.2012.11.001
18. Alam A, Molnar M, Czira M, Rudas A, Ujszaszi A, Kalantar-Zadeh K. **Serum adiponectin levels and mortality after kidney transplantation.**. (2013) **8** 460-7. PMID: 23220424
19. Fonseca I, Oliveira J, Santos J, Malheiro J, Martins L, Almeida M. **Leptin and adiponectin during the first week after kidney transplantation: biomarkers of graft dysfunction?**. (2015) **64** 202-7. DOI: 10.1016/j.metabol.2014.10.003
20. Roos M, Baumann M, Liu D, Heinemann F, Lindemann M, Horn P. **Low pre-transplant adiponectin multimers are associated with adverse allograft outcomes in kidney transplant recipients a 3-year prospective study.**. (2012) **178** 11-5. DOI: 10.1016/j.regpep.2012.06.001
21. Cobo G, Lindholm B, Stenvinkel P. **Chronic inflammation in end-stage renal disease and dialysis.**. (2018) **33** iii35-40. PMID: 30281126
|
---
title: Contribution of preoperative gut microbiota in postoperative neurocognitive
dysfunction in elderly patients undergoing orthopedic surgery
authors:
- Jiangjiang Bi
- Yifan Xu
- Shiyong Li
- Gaofeng Zhan
- Dongyu Hua
- Juan Tan
- Xiaohui Chi
- Hongbing Xiang
- Fengjing Guo
- Ailin Luo
journal: Frontiers in Aging Neuroscience
year: 2023
pmcid: PMC9981628
doi: 10.3389/fnagi.2023.1108205
license: CC BY 4.0
---
# Contribution of preoperative gut microbiota in postoperative neurocognitive dysfunction in elderly patients undergoing orthopedic surgery
## Abstract
### Objective
To investigate the role of gut microbiota and metabolites in POCD in elderly orthopedic patients, and screen the preoperative diagnostic indicators of gut microbiota in elderly POCD.
### Method
40 elderly patients undergoing orthopedic surgery were enrolled and divided into Control group and POCD group following neuropsychological assessments. Gut microbiota was determined by 16S rRNA MiSeq sequencing, and metabolomics of GC–MS and LC–MS was used to screen the differential metabolites. We then analyzed the pathways enriched by metabolites.
### Result
There was no difference in alpha or beta diversity between Control group and POCD group. There were significant differences in 39 ASV and 20 genera bacterium in the relative abundance. Significant diagnostic efficiency analyzed by the ROC curves were found in 6 genera bacterium. Differential metabolites in the two groups including acetic acid, arachidic acid, pyrophosphate etc. were screened out and enriched to certain metabolic pathways which impacted the cognition function profoundly.
### Conclusion
Gut microbiota disorders exist preoperatively in the elderly POCD patients, by which there could be a chance to predict the susceptible population.
### Clinical Trial Registration
[http://www.chictr.org.cn/edit.aspx?pid=133843&htm=4], identifier [ChiCTR2100051162].
## Introduction
Studies have found that the elderly are more prone to suffer from postoperative neurocognitive dysfunction (POCD; Moller et al., 1998), and our country is irreversibly entering an ageing society. It is predicted that by 2050, approximately $50\%$ of patients undergoing general anaesthetic surgery will be aged 65 years or older, and POCD seriously affects patients’ postoperative quality of life, prolongs hospital stays, and even increases the morbidity and mortality rate, putting enormous pressure on families and society (Skvarc et al., 2018). Therefore, it is urgent to study the pathogenesis of POCD in the elderly and explore its treatment.
Postoperative cognitive dysfunction (POCD) which has received widespread attention in recent years, refers to the development of central nervous system complications in the elderly after surgery, manifesting as confusion, anxiety, personality changes and memory impairment (Le et al., 2014; Tian et al., 2017). Some studies suggest that the incidence of POCD in elderly patients undergoing elective surgery under general anesthesia is two to ten times higher than in younger patients. Thus, advanced age is an important risk factor in its pathogenesis.
POCD is known to be similar to AD in terms of pathogenesis, clinical presentation, and morphopathological changes within the central nervous system (Evered et al., 2016; Fang et al., 2019). However, the diagnosis of POCD relies on the anesthesiologist’s own experience and cognitive scales, and its sensitivity for early diagnosis of POCD is poor. Therefore, the development of objective diagnostic indicators for POCD is an important issue that needs to be addressed in clinical anesthesia research.
Naseer et al. [ 2014] showed that the composition and numbers of gut microbiota in AD patients were significantly different from those of healthy controls. In a clinical study of Parkinson’s, Mulak and Bonaz [2015] concluded that the alpha and beta distribution of gut microbiota was different from that of their control group, suggesting the composition and function of the gut microbiota in Parkinson’s patients is different from that of healthy controls. Zhan et al. [ 2018] found that the alpha and beta distributions of the gut microbiota of SAMP8 mice were significantly different from those of control mice. In addition, they found a significant improvements in behavioral scores of cognitive function in this group of mice after gastrointestinal colonisation of pseudo-sterile mice with SAMR1 mouse fecal bacteria. These results all suggest an association between cognitive dysfunction and gut microbiota disorders. Even so, the relationships between gut microbiota and POCD in human have not been studied extensively. In this study, we investigated the role of gut microbiota and metabolites in POCD in elderly orthopedic patients, and tried to seek for the preoperative diagnostic indicators of gut microbiota in elderly POCD patients.
## Clinical research design
This was a prospective case–control study to collect fecal samples from elderly patients undergoing orthopedic surgery, to compare changes in gut microbiota and metabolites between POCD and control patients. The study was approved by the Medical Ethics Committee of Tongji Hospital affiliated to Tongji Medical College, Huazhong University of Science and Technology. All patients in the research signed informed consent before sample collection.
Inclusion criteria: Patients undergoing elective internal fixation of lower limb fractures, knee replacement or hip replacement by general anesthesia, American Society of Anesthesiologists (ASA) classification I – II, age 66–84 years old, and conscious in perioperative period were enrolled.
Exclusion Criteria: Patients suffered from one of following terms were excluded: central nervous system diseases or psychological disorders, long-term use of sedatives or antidepressants within the last year, Parkinson’s disease, severe hearing or visual impairment, drug dependence, alcoholism, inability to communicate with a physician.
The PASS 15 software was applied to calculate the sample size. The study is a diagnostic test and test for One Receiver Operating Characteristic (ROC) curve was applied. Existing studies show that the incidence of POCD in the elderly is about $40\%$, the sample size of the positive group was set at $40\%$, the AUC was 0.8, the type I error α was 0.05, the type II error β was 0.1 and the drop out rate was $20\%$, resulting in a positive sample size of 16 cases and a negative sample size of 24 cases for the study. The total sample size for this study was 40 cases.
## General information collection
The necessarily general information and medical history were collected with permission of each patient before surgery. Information about vital signs during the operation period, postoperative pain score and antibiotics treatment were also collected.
## Neuropsychological assessment
Neuropsychological assessments were performed to each patient 1 day before surgery, 1 day after surgery, 3 days after surgery and 28 days after surgery. The cognitive function of patients were assessed using the Mini-mental state examination (MMSE), Montreal cognitive assessment-basic (MoCA-B) and the Brief mental status assessment (Mini-Cog) and all the participant were divided into POCD and Control groups based on the assessment results. All the assessments above were performed by the same anesthetist. Adapted questions of MoCA-B facilitate the detection of mild cognitive impairment in subjects who are illiterate or possess a low education level. MoCA-B assesses executive function, language, orientation, calculation, etc., and is available for clinical research from official website.1
## DNA extraction and amplification of fecal samples
Fecal samples (about weight 2 g) were collected 1 day before surgery, suspended in fecal storage solution (Langfu Biotechnology Corporation, Shanghai, China) and then snap frozen and stored at −80°C. Bacterial DNA was isolated from the fecal samples using a MagPure Soil DNA LQ Kit (Magen, Guangdong, China) following the manufacturer’s instructions. DNA concentration and integrity were measured by a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States) and agarose gel electrophoresis, respectively. PCR amplification of the V3-V4 hypervariable regions of the bacterial 16S rRNA gene was carried out in a 25 μl reaction using universal primer pairs (343F: 5′TACGGRAGGCAGCAG-3′; 798R: 5′- AGGGTATCTAATCCT-3′). The reverse primer contained a sample barcode and both primers were connected with an Illumina sequencing adapter.
## 16S rRNA gene sequencing
The Amplicon quality was visualized using gel electrophoresis. The PCR products were purified with Agencourt AMPure XP beads (Beckman Coulter Co., USA) and quantified using Qubit dsDNA assay kit. The concentrations were then adjusted for sequencing which was performed on an Illumina NovaSeq6000with two paired-end read cycles of 250 bases each (Illumina Inc., San Diego, CA; OE Biotech Company, Shanghai, China).
Raw sequencing data were in FASTQ format. Paired-end reads were then preprocessed using cut adapt software to detect and cut off the adapter. After trimming, paired-end reads were filtering low quality sequences, denoised, merged and detect and cut off the chimera reads using DADA2 (Callahan et al., 2016) with the default parameters of QIIME2 (Bolyen et al., 2019). At last, the software output the representative reads and the ASV abundance table.
## GC–MS
The samples were analyzed on an Agilent 7890B gas chromatography system coupled to an Agilent 5977AMSD system (Agilent TechnologiesInc., CA, United States). ADB-5MS fused silica capillary column (30 m × 0.25 mm × 0.25 μm, Agilent J & W Scientific, Folsom, CA, United States) was utilized to separate the derivatives. Helium (>$99.999\%$) was used as the carrier gas at a constant flow rate of 1 ml/min through the column.
## LC–MS
A Dionex Ultimate 3,000 RS UHPLC fitted with Q-*Exactive plus* quadrupole-Orbitrap mass spectrometer equipped with heated electrospray ionization (ESI) source (Thermo Fisher Scientific, Waltham, MA, United States) was used to analyze the metabolic profiling in both ESI positive and ESI negative ion modes. An ACQUITY UPLC HSS T3 column (1.8 μm, 2.1 × 100 mm) were employed in both positive and negative modes.
## Statistical analysis and bioinformatic analysis
Statistical analysis was performed using SPSS 17.0 (SPSS Inc., Armonk, New York, United States). Values presented were expressed as mean ± standard error of the mean (S.E.M.), comparisons between groups were performed using one-way analysis of variance (ANOVA) followed by post hoc Tukey tests or Fisher’s exact tests. Normal distribution data were analyzed using one-way ANOVA, whereas non-normal distribution data were analyzed using Fisher’s exact test. Comparisons of numbers of variables in the general information between groups were performed by Chi-square test and Fisher’s exact test. Diagnostic cut-offs, AUC, sensitivity and specificity were determined by ROC curve analysis. p-value <0.05 is considered to be a significant difference.
The representative read of each ASV was selected using QIIME 2 package. All representative reads were annotated and blasted against Silva database Version 138 (16 s rDNA) using q2-feature-classifier with the default parameters. The microbial diversity in fecal samples was estimated using the alpha diversity that include Chao1 index, Shannon index and ACE index. The Unifrac distance matrix performed by QIIME2 software was used for unweighted Unifrac Principal coordinates analysis (PCoA) and phylogenetic tree construction. The 16S rRNA gene amplicon sequencing and analysis were conducted by OE Biotech Co., Ltd. (Shanghai, China).
Multiple statistical analyses were performed using principal component analysis (PCA), partial least squares analysis (PLS-DA) and orthogonal partial least squares analysis (OPLS-DA) to find the differential metabolites between groups. Student’s T test and a Fold change analysis were used to compare metabolites between the two groups. A combination of multidimensional and unidimensional analyses was used to screen for differential metabolites. Metabolic pathways enrichment analysis of differential metabolites were performed based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. The screening criteria was VIp value >1 for the first principal component of the OPLS-DA model and value of $p \leq 0.05$ for the T-test.
## General information and perioperative clinical data
General information and medical history were listed in Table 1. Neuropsychological assessment outcomes including all time points were listed in Table 2. Postoperative pain numeric rating scales (NRS) were listed in Table 3.
## Microbiological composition in POCD patients
The total data volume of raw reads after sequencing ranged from 78,268 to 81,928, clean tags after quality control ranged from 59,473 to 73,165, clean tags after removing chimeras to obtain valid tags ranged from 42,922 to 70,784. The number of ASVs in each sample ranged from 103 to 533. Comparisons between the two groups using Wilcoxon rank sum test, significant differences were observed in 39 ASVs and 20 genus.
The statistical result showed the top relative abundance at 6 levels, including Bacteroidota, Firmicutes, Proteobacteria, Actinobacteriota, Fusobacteria were the top abundance at the phylum level (Figure 1A). It showed significant differences between the two groups using Wilcoxon rank sum test in the relative abundance at the genus level, including Alistipes ($$p \leq 0.008$$), Helicobacter ($$p \leq 0.007$$), Lysobacter ($$p \leq 0.011$$), Barnesiella ($$p \leq 0.017$$), Macrococcus ($$p \leq 0.049$$), etc. ( Figure 1B). Relative abundance of top 10 differential genera bacterium were listed in Figures 1C–L.
**Figure1:** *Comparisons of relative abundance of gut microbiota. (A) Histogram results for the top 20 in order of relative abundance at phylum level, which shows Bacteroidota, Firmicutes, Proteobacteria etc. contribute the main components of gut microbiota, (B) Comparisons between the two groups using Wilcoxon rank sum test, significant differences in relative abundance were observed in 20 genus (*p < 0.05), and (C–L) Relative abundance of differential genera bacterium, analyzed by two-tailed Wilcoxon rank-sum test, data represent the means ± SEM, n = 16 in each group, *p < 0.05 compared with Control group.*
There was no significant difference of alpha diversity as measured by the Chao1 diversity index ($$p \leq 0.098$$). The Shannon index ($$p \leq 0.065$$) and ACE index ($$p \leq 0.110$$) indicated species diversity and evenness, did not show significant differences either (Figures 2A–C). There was no difference between the two groups measured by PCA and PCoA which indicated beta diversity based on the Bray-Curtis distance algorithm (Figures 2D,E).
**Figure 2:** *Alpha diversity and beta diversity in gut microbiota between the two groups. (A–C) There was no significant difference of alpha diversity as measured by the Chao1 diversity index (p = 0.098). The Shannon index (p = 0.065) and ACE index (p = 0.110) indicated species diversity and evenness, did not show significant differences. (D,E) There was no difference between the two groups measured by PCA and PCoA.*
## Evaluation of gut bacteria for diagnosis of POCD using ROC curve analysis
ROC curve analysis was performed to evaluate the diagnostic efficiency of the 20 differential genera gut bacteria in POCD (Figure 3). The diagnostic efficiency of genera Helicobacter, Alistipes, Barnesiella, Lachnospiraceae_NK4A136_group, Prevotellaceae_NK3B31_group, and Pediococcus were significantly higher in POCD group than those in Control group. The area under curve (AUC), p value, sensitivity, specificity of gut bacteria for the diagnosis of POCD are listed in Table 4.
**Figure 3:** *ROC curves of the gut genus bacterium relative abundance for the diagnosis of POCD. Vertical coordinate indicated the sensitivity of diagnosis, horizontal coordinate indicated the 1-specificity of diagnosis, AUC > 0.5 indicated a diagnosis efficiency of the gut bacterium.* TABLE_PLACEHOLDER:Table 4
## Differential metabolites and related metabolic pathways
Untargeted metabolomics analysis including GC–MS and LC–MS platforms were performed, and significant differences were found between two groups at the metabolic profiling level using PCA, PLS-DA and OPLS-DA (R2 = 0.961, Q2 = −0.268; Figures 4A,B). A total of 27 differential metabolites were screened out by T-test and visualized by volcano plots (Figures 4C,D) and Hierarchical Clustering in the GC–MS platform, including Acetic acid, Leucine, Pyrophosphate, etc. ( Figure 5). While 125 differential metabolites were screened out in the LC–MS platform. The differential metabolites were listed in Supplementary materials Data sheet 1 and 2.
**Figure 4:** *(A,B) It showed significant differences between two groups at the metabolic profiling level both in PCA and OPLS-DA (R2 = 0.961, Q2 = −0.268), (C,D) Differential metabolites between the two groups performed by T-test, and visualization by their value of ps and Fold change values using volcano plots. The red dots represent metabolites that were significantly upregulated in the POCD group, the blue dots represent metabolites that were significantly downregulated and the grey dots represent metabolites that were not significant.* **Figure 5:** *Heat map of differential metabolites. We performed hierarchical clustering of all significantly differential metabolites expressions, including 27 metabolites in GC–MS (A), and 125 metabolites in LC–MS (B). The horizontal coordinates indicate sample names and the vertical coordinates indicate differential metabolites. The colour ranges from blue to red, a redder colour indicates a higher abundance of expression of the differential metabolite.*
Metabolic pathways highly enriched by the differential metabolites were then found out using Hypergeometric test based on the KEGG database. The pathways with top richness metabolites in the GC–MS platform were: Protein digestion and absorption, Glyoxylate and dicarboxylate metabolism, mTOR signaling pathway, Glycosaminoglycan biosynthesis-heparan sulfate/heparin, Alcoholic liver disease, Cholinergic synapse, Shigellosis, Oxidative phosphorylation, Valine, leucine and isoleucine biosynthesis, Taurine and hypotaurine metabolism ($p \leq 0.05$; Figure 6A). And the top enriched metabolic pathways in LC–MS platform were: Glycerophospholipid metabolism, Choline metabolism in cancer, Cholinergic synapse, Regulation of actin cytoskeleton, Biosynthesis of unsaturated fatty acids, Nicotine addiction, Synaptic vesicle cycle, Insulin secretion, Gastric acid secretion, Pancreatic secretion, Salivary secretion ($p \leq 0.05$; Figure 6B).
**Figure 6:** *Metabolic pathway enrichment bubble chart. The value of p of the metabolic pathway is the significance of the enrichment of the metabolic pathway, and the significant enrichment pathway was selected for bubble plotting. The vertical coordinate is the name of the metabolic pathway; the horizontal coordinate is the enrichment factor (Rich factor = number of significantly differential metabolites/total number of metabolites in the pathway), the larger the Rich factor, the greater the enrichment; the colour from green to red indicates that the value of p decreases; the larger the bubble, the greater the number of metabolites enriched to that pathway. (A) Metabolic pathway enrichment of GC-MS platform. (B) Metabolic pathway enrichment of LC-MS platform.*
## Correlation between microbiome and metabolites
Spearman correlation was applied to analyze the correlation between gut microbiota and metabolites and the results indicated significant correlations between them. Bacteroides was significantly negative correlated to Methanephosphonothioic acid ($$p \leq 0.005$$) and Ciliatine ($$p \leq 0.002$$), and positive correlated to L-lysine ($$p \leq 0.025$$) and L-alanine ($$p \leq 0.040$$). While Sutterella was positive correlated to Methanephosphonothioic acid ($$p \leq 0.003$$) and Ciliatine ($$p \leq 0.002$$). [ Eubacterium]_coprostanoligenes_group was negative correlated to 5-aminovaleric acid ($$p \leq 0.0003$$), Talose ($$p \leq 0.0008$$), etc. ( Figure 7).
**Figure 7:** *Heat map of correlation between microbiome and metabolites. Each column is for a different species and each row corresponds to a metabolite. The graph shows positive correlations in orange and negative correlations in blue, with darker colours representing greater correlations and colours closer to white representing correlations closer to zero. A *** in the graph represents a correlation p value less than 0.001, a ** in the graph represents a correlation p value less than 0.01 and a * in the graph represents a correlation p value less than 0.05.*
## Discussion
In this study we investigated the gut microbiota from elderly patients undergoing orthopedic surgery. Considering confounding factors, the general information of participants including relevant previous history, laboratory test results and postoperative pain NRS were collected, and there were no differences in variables age, gender, type of surgery, blood glucose, BUN, INR, LVEF, NRS, number of hypertension, diabetes, HLP, stroke, ECG ST segment ischemia between the two groups. The variable age was divided into 4 sections with a 5-year interval. The majority is from 66 to 75 years old, and there was no difference in 4 sections between groups. Which indicated the morbidity of POCD in the elderly patients was equal in the 66–85 years interval. The patients’ gender, cardiac function, renal function, stroke history, blood glucose, lipid level and postoperative pain did not show influences in the POCD morbidity in this study, which excluded the confounding factor and provided us a focus on the role of gut microbiota.
Although alpha diversity and beta diversity of gut microbiota were similar in two groups, but at the genus level, 20 gut bacterium were significantly altered in POCD patients compared with those in control patients. And among the 20 genus, the relative abundance of the genera Helicobacter, Alistipes, Barnesiella, Lachnospiraceae_NK4A136_group, Prevotellaceae_NK3B31_group and Pediococcus were higher in POCD group than those in Control group, and all have got significantly diagnostic efficiency analyzed by the ROC curves, which might be potential target bacterium to diagnose or predict elderly POCD. The alterations above were consistent with the recent research. As is well known, *Helicobacter pylori* (H. pylori) is a Gram-negative bacterium resides in the gastrointestinal tract starting from early age. H. pylori infection might be a risk factor for cognitive decline in the elderly (Han et al., 2018; Cárdenas et al., 2019). The biological mechanisms include reduced absorption of folate and vitamin B-12 and increased homocysteine (Berrett et al., 2018). Ren et al. [ 2020] claimed that the abundance of genera Barnesiella and Alistipes negatively correlated with cognition ability. Tran et al. [ 2019] detected an association of Prevotellaceae and several butyrate-producing genera with Apolipoprotein E genotypes, which is the strongest genetic risk factor for Alzheimer’s disease. Nevertheless, it was reported that decreased abundance of the Lachnospiraceae family was associated with cognitive decline (Liu et al., 2019). These results indicated alterations of abundance of gut microbiota play an important role in the pathogenesis of POCD in elderly patients.
Metabonomics analysis was performed to clarify whether there were associations between the metabolites produced by gut microbiota and POCD in elderly. 27 differential metabolites were screened out in the GC–MS platform, including acetic acid, arachidic acid, leucine, pyrophosphate, etc. In this study, acetic acid from gut microbiota was highly enriched to pathways of protein digestion and absorption, alcoholic liver disease, glyoxylate and dicarboxylate metabolism, glycosaminoglycan biosynthesis – heparan sulfate/heparin.
Acetic acid is the smallest short-chain fatty acid (SCFA) which can form acetylCoA and participate the metabolism of carbohydrates and fats. It is produced and excreted by certain acetic acid bacteria, notably the *Acetobacter genus* and Clostridium acetobutylicum. Zheng et al. [ 2021] built the model of decreased gut acetate level by vancomycin exposure in mice, which lead to reduction of hippocampal synaptophysin level and impaired learning and memory. This alteration might be mediated by the vagus nerve stimulation. Consistent with the recent study, we found the expression level of acetic acid was lower in POCD group than that in Control group, which indicated acetic acid was positively correlated with cognition ability, probably via regulating hippocampal synaptophysin level in the elderly patients undergoing orthopedic surgery.
Arachidic acid is an important component of neuronal membranes, and regulate neurotransmission, neuroinflammation, cell survival and cognition function (Martin et al., 2016). The oxidative arachidic acid metabolic pathway is the core of inflammation, and the main cause of working memory impairment leading to AD pathogenesis (Chen et al., 2022). In this study, arachidic acid was highly enriched in biosynthesis of unsaturated fatty acids pathway, and expression level was lower in POCD group in the LC–MS platform. The result inferred that arachidic acid was probably metabolized into prostaglandins and leukotrienes which mediated neuroinflammation, and finally led to cognition impairment in the elderly POCD.
Pyrophosphate was highly enriched to pathways of oxidative phosphorylation, parkinson disease, neurodegeneration – multiple diseases, and expression level was higher in POCD group. Bisphosphonates interact and regulate calcium ions which play a role in neurodegenerative diseases (Zameer et al., 2018), the result was consistent with recent study.
## Conclusion
Gut microbiota altered in the elderly POCD and some of the genera bacterium might be effective diagnosis indicators to predict POCD incidence before surgery. The differential metabolites including acetic acid, arachidic acid and pyrophosphate etc. played important roles in the cognition impairment in POCD patients.
The limitation of this study is that although we have screened out the differential microbiota and metabolites which might be indicators of the POCD, a verification work has not been done yet. We planned to complete the work by fecal microbiota transplantation in future to investigate the therapeutic effect of the indicators.
## 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 Medical Ethics Committee of Tongji Hospital affiliated to Tongji Medical College, Huazhong University of Science and Technology. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
JB designed and supervised the project and obtained funding. JB, YX, DH, and XC collected samples and performed neurological assessments. FG provided samples. JB, SL, GZ, and JT extracted data and performed statistical analysis. JB drafted the manuscript. AL and HX revised the manuscript for important intellectual content. All authors contributed to the article and approved the submitted version.
## Funding
This study was supported by National Natural Science Foundation of China [82001161] and Braun Foundation for Scientific Research in Anesthesia (BBDF-2019-011).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnagi.2023.1108205/full#supplementary-material
## References
1. Berrett A. N., Gale S. D., Erickson L. D., Brown B. L., Hedges D. W.. **Helicobacter pylori moderates the association between 5-MTHF concentration and cognitive function in older adults**. *PLoS One* (2018) **13** e0190475. DOI: 10.1371/journal.pone.0190475
2. Bolyen E., Rideout J. R., Dillon M. R., Bokulich N. A., Abnet C. C., al-Ghalith G. A.. **Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2**. *Nat. Biotechnol.* (2019) **37** 852-857. DOI: 10.1038/s41587-019-0209-9
3. Callahan B. J., McMurdie P. J., Rosen M. J., Han A. W., Amy Jo A., Johnson S. P.. **DADA2: high-resolution sample inference from Illumina amplicon data**. *Nat. Methods* (2016) **13** 581-583. DOI: 10.1038/nmeth.3869
4. Cárdenas V. M., Boller F., Román G. C.. **Helicobacter pylori, vascular risk factors and cognition in U.S. older adults**. *Brain Sci.* (2019) **9** 370. DOI: 10.3390/brainsci9120370
5. Chen C., Liao J., Xia Y., Liu X., Jones R., Haran J.. **Gut microbiota regulate Alzheimer's disease pathologies and cognitive disorders via PUFA-associated neuroinflammation**. *Gut* (2022) **71** 2233-2252. DOI: 10.1136/gutjnl-2021-326269
6. Evered L., Silbert B., Scott D. A., Ames D., Maruff P., Blennow K.. **Cerebrospinal fluid biomarker for Alzheimer disease predicts postoperative cognitive dysfunction**. *Anesthesiology* (2016) **124** 353-361. DOI: 10.1097/ALN.0000000000000953
7. Fang E. F., Hou Y., Palikaras K., Adriaanse B. A., Kerr J. S., Yang B.. **Mitophagy inhibits amyloid-β and tau pathology and reverses cognitive deficits in models of Alzheimer's disease**. *Nat. Neurosci.* (2019) **22** 401-412. DOI: 10.1038/s41593-018-0332-9
8. Han M.-L., Chen J.-H., Tsai M.-K., Liou J.-M., Chiou J.-M., Chiu M.-J.. **Association between helicobacter pylori infection and cognitive impairment in the elderly**. *J. Formos. Med. Assoc.* (2018) **117** 994-1002. DOI: 10.1016/j.jfma.2017.11.005
9. Le Y., Liu S., Peng M., Tan C., Liao Q., Duan K.. **Aging differentially affects the loss of neuronal dendritic spine, neuroinflammation and memory impairment at rats after surgery**. *PLoS One* (2014) **9** e106837. DOI: 10.1371/journal.pone.0106837
10. Liu P., Wu L., Peng G., Han Y., Tang R., Ge J.. **Altered microbiomes distinguish Alzheimer's disease from amnestic mild cognitive impairment and health in a Chinese cohort**. *Brain Behav. Immun.* (2019) **80** 633-643. DOI: 10.1016/j.bbi.2019.05.008
11. Martin S. A., Brash A. R., Murphy R. C.. **The discovery and early structural studies of arachidonic acid**. *J. Lipid Res.* (2016) **57** 1126-1132. DOI: 10.1194/jlr.R068072
12. Moller J. T., Cluitmans P., Rasmussen L. S., Houx P., Rasmussen H., Canet J.. **Long-term postoperative cognitive dysfunction in the elderly ISPOCD1 study. ISPOCD investigators. International study of post-operative cognitive dysfunction**. *Lancet* (1998) **351** 857-861. DOI: 10.1016/S0140-6736(97)07382-0
13. Mulak A., Bonaz B.. **Brain-gut-microbiota axis in Parkinson's disease**. *World J. Gastroenterol.* (2015) **21** 10609-10620. DOI: 10.3748/wjg.v21.i37.10609
14. Naseer M. I., Bibi F., Alqahtani M. H., Chaudhary A., Azhar E., Kamal M.. **Role of gut microbiota in obesity, type 2 diabetes and Alzheimer's disease**. *CNS Neurol. Disord. Drug Targets* (2014) **13** 305-311. DOI: 10.2174/18715273113126660147
15. Ren T., Gao Y., Qiu Y., Jiang S., Zhang Q., Zhang J.. **Gut microbiota altered in mild cognitive impairment compared with Normal cognition in sporadic Parkinson's disease**. *Front. Neurol.* (2020) **11** 2020. DOI: 10.3389/fneur.2020.00137.eCollection
16. Skvarc D. R., Berk M., Byrne L. K., Dean O. M., Dodd S., Lewis M.. **Post-operative cognitive dysfunction: an exploration of the inflammatory hypothesis and novel therapies**. *Neurosci. Biobehav. Rev.* (2018) **84** 116-133. DOI: 10.1016/j.neubiorev.2017.11.011
17. Tian Y., Guo S., Zhang Y., Ying X., Zhao P., Zhao X.. **Effects of hydrogen-rich saline on hepatectomy-induced postoperative cognitive dysfunction in old mice**. *Mol. Neurobiol.* (2017) **54** 2579-2584. DOI: 10.1007/s12035-016-9825-2
18. Tran T. T. T., Corsini S., Kellingray L., Hegarty C., Le Gall G., Narbad A.. **APOE genotype influences the gut microbiome structure and function in humans and mice: relevance for Alzheimer's disease pathophysiology**. *FASEB J.* (2019) **33** 8221-8231. DOI: 10.1096/fj.201900071R
19. Zameer S., Najmi A. K., Vohora D., Akhtar M.. **Bisphosphonates: future perspective for neurological disorders. Bisphosphonates: future perspective for neurological disorders**. *Pharmacol. Rep.* (2018) **70** 900-907. DOI: 10.1016/j.pharep.2018.03.011
20. Zhan G., Yang N., Li S., Huang N., Fang X., Zhang J.. **Abnormal gut microbiota composition contributes to cognitive dysfunction in SAMP8 mice**. *Aging (Albany NY)* (2018) **10** 1257-1267. DOI: 10.18632/aging.101464
21. Zheng H., Pengtao X., Jiang Q., Qingqing X., Zheng Y., Yan J.. **Depletion of acetate-producing bacteria from the gut microbiota facilitates cognitive impairment through the gut-brain neural mechanism in diabetic mice**. *Microbiome* (2021) **9** 145. DOI: 10.1186/s40168-021-01088-9
|
---
title: Metabolic profiling identifies the significance of caffeine metabolism in CKD
authors:
- Xinghua Guo
- Hongquan Peng
- Peijia Liu
- Leile Tang
- Jia Fang
- Chiwa Aoieong
- Tou Tou
- Tsungyang Tsai
- Xun Liu
journal: Frontiers in Bioengineering and Biotechnology
year: 2023
pmcid: PMC9981652
doi: 10.3389/fbioe.2023.1006246
license: CC BY 4.0
---
# Metabolic profiling identifies the significance of caffeine metabolism in CKD
## Abstract
Background: With the development of chronic kidney disease (CKD), there are various changes in metabolites. However, the effect of these metabolites on the etiology, progression and prognosis of CKD remains unclear.
Objective: We aimed to identify significant metabolic pathways in CKD progression by screening metabolites through metabolic profiling, thus identifying potential targets for CKD treatment.
Methods: *Clinical data* were collected from 145 CKD participants. GFR (mGFR) was measured by the iohexol method and participants were divided into four groups according to their mGFR. Untargeted metabolomics analysis was performed via UPLC-MS/MSUPLC–MSMS/MS assays. Metabolomic data were analyzed by MetaboAnalyst 5.0, one-way ANOVA, principal component analysis (PCA), and partial least squares discriminant analysis (PLS-DA) to identify differential metabolites for further analysis. The open database sources of MBRole2.0, including KEGG and HMDB, were used to identify significant metabolic pathways in CKD progression.
Results: Four metabolic pathways were classified as important in CKD progression, among which the most significant was caffeine metabolism. A total of 12 differential metabolites were enriched in caffeine metabolism, four of which decreased with the deterioration of the CKD stage, and two of which increased with the deterioration of the CKD stage. Of the four decreased metabolites, the most important was caffeine.
Conclusion: Caffeine metabolism appears to be the most important pathway in the progression of CKD as identified by metabolic profiling. Caffeine is the most important metabolite that decreases with the deterioration of the CKD stage.
## 1 Introduction
Chronic kidney disease (CKD) is a chronic disorder involving structural and functional organ changes. There are multiple underlying causes of this disorder which is characterized by its irreversibility and progressive development. CKD is a growing public health concern worldwide. As the disease progresses, the kidney’s ability to remove nitrogenous wastes, exogenous molecules, and metabolism of low molecular weight proteins decrease, resulting in multiple clinical sequelae. It is recognized that metabolites change with the development of kidney injury (Rysz et al., 2021). However, the effect of these metabolites on the etiology, progression, and prognosis of CKD remains unclear. The main clinical challenge is that current medical interventions aimed at delaying kidney decline are very limited. As such, the discovery of potential metabolic pathways could help us identify new therapeutic targets and address this serious problem (Chen et al., 2019a).
Several attempts have been made to identify new CKD-associated biomarkers and deepen our understanding of the pathological mechanisms underlying CKD (Dou et al., 2018; Xu et al., 2018). Metabolomics can be used to uncover metabolites associated with critical kidney functions such as the glomerular filtration rate (GFR), allowing for important clinical information to be obtained from biological samples (Earl et al., 2018; Lai et al., 2018) Rhee et al. reported that ten plasma metabolites are nominally associated with CKD progression (Rhee et al., 2016). In addition, Chen et al. found that five metabolites, including 5-methoxytryptophan (5-MTP), canavaninosuccinate (CSA), acetylcarnitine, tiglylcarnitine, and taurine, can be used to accurately identify the clinical stages of CKD, and that tryptophan hydroxylase-1 (TPH-1) presents a potential therapeutic target in CKD (Chen et al., 2019b). Feng et al. found that CKD rats can be differentiated from sham rats by metabolites involved in the pathways of gut microbial metabolism. They also found that improvement of gut dysbiosis retarded the progression of kidney disease in a rat model of CKD (Feng et al., 2019). Zhao et al. utilized an adenine-induced CKD model to identify perturbations in fatty acid metabolism, purine metabolism, and amino acid metabolism (Zhao et al., 2014). Brunetto et al. compared the serum metabolic profile of healthy and CKD dogs and found decreased urea, creatinine, creatine, citrate, lipids, lactate, branched-chain amino acids (BCAAs), and glutamine in CKD; a specific diet was able to maintain and retard the progression of CKD (Brunetto et al., 2021). Lanzon et al. analyzed the serum and urine of patients with severe obesity and CKD before and after undergoing bariatric surgery (BS) and found that isoleucine and tyrosine were increased in CKD patients compared to those without CKD (Lanzon et al., 2021). Gordin et al. ( Gordin et al., 2019) used metabolic pathway analysis and reported that hexose, mitochondrial, amino acid, and purine pathways are associated with preserved kidney function.
Factors found to accelerate kidney decline have been explored in many studies. Most research involving metabolic profiling has focused on metabolites that increase with deteriorating renal function. However, in this study, we have focused on identifying the metabolic pathways associated with improved kidney function. By utilizing metabolic profiling, we attempted to identify metabolites that decrease as CKD deteriorates, ascertain the most relevant metabolic pathways, and identify potential therapeutic targets to improve kidney function.
## 2.1 Participants and mGFR measurement
All participants provided informed consent before participating in the study according to a protocol approved by the Kiang Wu Hospitalethics committee. All participants were recruited in August 2019 from the Kiang Wu Hospital (Santo Antonio, Macau) and via outpatient clinics. The cohort included 145 patients who met the study inclusion criteria and were diagnosed with CKD based on the NKF-KDOI guidelines. Peripheral venous blood (4 mL) was collected from each patient; plasma samples were used for metabolomic analysis. Renal function was evaluated using GFR (mGFR) by utilizing the plasma clearance of iohexol (Shah et al., 2013). After blood was collected for the aforementioned tests, Iohexol was injected over 2 min (300 mg/mL, GE Healthcare, Shanghai, China) and plasma (6 mL) was collected from the contralateral upper extremity to detect the Iohexol concentration (by HPLC) at 120 and 240 min after Iohexol administration. For participants with eGFR<30 min/mL/1.73 m (Chen et al., 2019a), the blood collection times were changed to 120 and 300 min. All blood samples were centrifuged at 2000 g for 10 min at room temperature to extract plasma and stored at −80°C until analysis.
## 2.2 UPLC–MS/MS assays
All of the untargeted metabolomics analyses were conducted at the Dian Calibra-Metabolon Joint Metabolomics Laboratory (Hangzhou, China). Four different UPLC–MS/MS assays of small molecule metabolites were performed on each sample (Shen et al., 2020). Automatic liquid transfer during sample preparation was handled on a Hamilton automated MicroLab STAR® system (Hamilton, Switzerland). A methanol-based sample extraction solution was added to each sample and mixed using a GeneGrinder 2010 (Spex SamplePrep, United States of America) mixer. After 2 minutes of vigorous shaking and centrifugation to precipitate proteins and other debris, the extracted metabolites in the supernatant were collected and divided into four fractions: two fractions were analyzed by reversed-phase (RP) UPLC-MS/MS under positive electrospray ionization (ESI) mode. The two UPLC methods were slightly different using the same column (BEH C18 2.1 × 100 mm, 1.7 μm column, Waters). The mobile solutions for the two positive ESI UPLC-MS/MS were water and methanol containing $0.05\%$ perfluoropentanoic acid (PFPA) and $0.1\%$ formic acid (FA). The third fraction was used for reversed-phase UPLC-MS/MS in negative ion ESI mode (BEH C18 2.1 × 100 mm, 1.7 μm column, Waters), and the mobile solutions were methanol and water in 6.5 mM ammonium bicarbonate at pH 8. The last fraction was used for hydrophilic interaction liquid chromatography (HILIC)/UPLC-MS/MS in negative ESI mode (BEH Amide 2.1 × 150 mm, 1.7 μm column, Waters), and the mobile solutions consisted of water and acetonitrile with 10 mM ammonium formate at pH 10.8. Each fraction was dried under nitrogen gas flow and then dissolved in reconstitution solutions before being injected into each of the four UPLC-MS/MS systems. The QE mass spectrometer was alternated between full MS and data-dependent MS2 scans using dynamic exclusion for data collection. The scan range was 70–1,000 m/z. Processing, extraction, and peak identification of the raw mass spectrometry data were carried out using in-house developed software, and metabolites were identified by comparing the experimental ion characteristics to entries in an in-house library which was constructed using pure reference standards. The entries in the library included retention time/retention index (RI), mass to charge ratio (m/z), and MS/MS spectral data of each reference standard.
## 2.3 Metabolomics statistical analysis
All participants were divided into four groups according to mGFR: group A (mGFR<30 mL/min/1.73 m (Chen et al., 2019a)), group B (30 mL/min/1.73 m (Chen et al., 2019a)≤mGFR<60 mL/min/1.73 m (Chen et al., 2019a)), group C (60 mL/min/1.73 m (Chen et al., 2019a)≤mGFR <90 mL/min/1.73 m (Chen et al., 2019a)), and group D (mGFR ≥90 mL/min/1.73 m (Chen et al., 2019a)). Statistical analysis of patient data was carried out using SPSS 26.0. Statistical significance was determined using a threshold of $$p \leq 0.05.$$ *Metabolomic data* analyses were carried out using MetaboAnalyst 5.0 (https://www.metaboanalyst.ca). The mass spectrometry data which were acquired by untargeted metabolomics analysis were uploaded as comma separated values (.csv). The uploaded data file contains a data matrix of 145 (samples) × 1,094 (compounds). Before data analysis, a data integrity check was performed to ensure that all the necessary information had been collected. To minimize bias associated with the omission of censored data, all missing and zero values were replaced by half of the minimum positive values across samples in the original data. Normalization was done via log transformation and Pareto scaling.
One-way analysis of variance (ANOVA), principal component analysis (PCA), and partial least squares-discriminant analysis (PLS-DA) were used to screen out the differential metabolites. MetaboAnalyst 5.0 provided one-way ANOVA test results to determine whether the overall comparison among each group was significant. Univariate analyses provided a preliminary overview of features that are potentially significant in discriminating the conditions under investigation. Statistical significance was determined using a threshold of $$p \leq 0.05.$$
PCA and PLS-DA were also performed using MetaboAnalyst 5.0. PCA is an unsupervised method aiming to find the directions that best explain the variance in a data set X) without referring to class labels Y). The data are summarized into much fewer variables called scores, which are weighted averages of the original variables. PLS is a supervised method that uses multivariate regression techniques to extract, via a linear combination of the original variables X), information that can predict class membership Y). To assess the significance of class discrimination, a permutation test was performed. In each permutation, a PLS-DA model was built between the data X) and the permuted class labels Y) using the optimal number of components determined by cross-validation for the model based on the original class assignment. MetaboAnalyst supports two types of test statistics for measuring class discrimination. The first one is based on prediction accuracy during training. The second is the separation distance based on the ratio of the between-group sum of the squares and the within-group sum of squares (B/Wratio). If the observed test statistic is part of the distribution based on the permuted class assignments, the class discrimination cannot be considered statistically significant. Variable Importance in Projection (VIP), which is an important variable in PLS-DA, is a weighted sum of squares of the PLS loadings taking into account the amount of explained Y-variation in each dimension. VIP scores are calculated for each component. When more components are used to calculate the feature importance, the average of the VIP scores is used. A VIP threshold >1.0 was considered statistically significant.
## 2.4 Pathway analysis
Metabolic pathways were identified by utilizing open database sources of MBRole2.0 (http://csbg.cnb.csic.es/mbrole2/), including KEGG and HMDB. Compound names of the differential metabolites were first converted to KEGG IDs using MetaboAnalyst 5.0 and the KEGG IDs were submitted to MBRole 2.0 for KEGG pathway analysis.
## 3.1 Characteristics of the study populations
We recruited 145 individuals aged 20 to 96 years, 68 of whom were male. Based on mGFR, 22 were assigned to group A, 47 to group B, 39 to group C, and 37 to group D. Table 1 presents the summary statistics for each group. Age, weight, body mass index (BMI), systolic blood pressure, diastolic blood pressure, creatinine, eGFR, and mGFR were symmetrically distributed. Arithmetic means and standard deviations are provided. The p-values of one-way ANOVA tests are also presented. Creatinine, mGFR and eGFR were used a trend test, and the Ptrend values were presented. The height followed an asymmetric distribution, and thus median and interquartile ranges are shown. The p-values of the Kruskal–Wallis test are presented. Sex, current smoking, current drinking, diabetes, hypertension, coronary heart disease, stroke, hyperuricemia, use of antiplatelet drugs, antilipemic agents, antihypertensive agents, hypoglycemic agents, immunosuppressors, and uric acid reduction medicine were dichotomous variables. For these, quantities and frequencies are shown as appropriate. The p-values of Chi-square tests are presented. There were no cases of current drinking, stroke, and immunosuppressive drug use.
**TABLE 1**
| Variables | Group | Group.1 | Group.2 | Group.3 | p |
| --- | --- | --- | --- | --- | --- |
| Variables | A | B | C | D | p |
| Age (y) | 74.9 ± 16.5 | 72.7 ± 13.5 | 60.6 ± 14.1 | 41.8 ± 10.6 | <0.01 |
| Sex (M/F) | 9/13 | 25/22 | 18/21 | 16/21 | 0.66 |
| Height (cm) | 154.0 (151.0, 166.5) | 159.0 (155.0, 167.0) | 157.0 (153.0–172.0) | 163.0 (157.0–171.0) | 0.43 |
| Weight (kg) | 62.3 ± 11.2 | 62.5 ± 17.0 | 66.2 ± 15.1 | 64.6 ± 14.1 | 0.85 |
| BMI (cm/kg^2) | 25.1 ± 3.2 | 24.6 ± 5.7 | 25.2 ± 4.7 | 23.9 ± 4.0 | 0.64 |
| Systolic blood pressure (mmHg) | 130.9 ± 14.2 | 132.6 ± 16.9 | 133.0 ± 16.8 | 126.7 ± 12.8 | 0.27 |
| Diastolic blood pressure (mmHg) | 70.7 ± 13.7 | 72.8 ± 13.0 | 77.6 ± 12.1 | 77.5 ± 12.1 | 0.08 |
| Creatinine (mg per 100 mL) | 241.3 ± 144.2 | 105.8 ± 26.5 | 80.9 ± 18.0 | 70.2 ± 18.9 | <0.01* |
| eGFR (ml/(min.1.73 m (Chen et al., 2019a))) | 23.2 ± 11.5 | 48.8 ± 15.5 | 75.5 ± 12.0 | 103.1 ± 15.9 | <0.01* |
| mGFR (ml/(min.1.73 m (Chen et al., 2019a))) | 21.9 ± 6.4 | 44.6 ± 7.9 | 73.2 ± 8.4 | 106.9 ± 12.3 | <0.01* |
| current smoking | 0 | 0 | 0 | 1 (2.7%) | 0.40 |
| current drinking | 0 | 0 | 0 | 0 | - |
| diabetes | 10 (45.5%) | 11 (23.4%) | 7 (17.9%) | 4 (10.8%) | 0.02 |
| hypertension | 15 (68.2%) | 28 (59.6%) | 15 (38.5%) | 4 (10.8%) | <0.01 |
| coronary heart disease | 9 (40.9%) | 11 (23.4%) | 5 (12.8%) | 2 (5.4%) | 0.004 |
| stroke | 0 | 0 | 0 | 0 | - |
| hyperuricemia | 0 | 8 (17.0%) | 11 (28.2%) | 2 (5.4%) | <0.01 |
| antiplatelet drugs | 9 (40.9%) | 11 (23.4%) | 5 (12.8%) | 2 (5.4%) | <0.01 |
| antilipemic agent | 2 (9.1%) | 5 (10.6%) | 6 (15.4%) | 0 | 0.13 |
| anti-hypertensive agent | 15 (68.2%) | 28 (59.6%) | 15 (38.5%) | 4 (10.8%) | <0.01 |
| hypoglycemic agent | 10 (45.5%) | 11 (23.4%) | 7 (17.9%) | 4 (10.8%) | 0.02 |
| immunosuppressor | 0 | 0 | 0 | 0 | - |
| Uric acid reduction medicine | 0 | 8 (17.0%) | 11 (28.2%) | 2 (5.4%) | <0.01 |
## 3.2 Univariate analysis
Before data analysis, a data integrity check was performed to make sure that all the necessary information had been collected. The data normalization result implemented by MetaboAnalyst5.0 provided in Figure 1 supplement. After one-way ANOVA for multigroup analysis, Table 2 supplement shows all significant metabolites selected by ANOVA with p-value threshold 0.05 and Table 2 details these findings for the top 50 metabolites. ANOVA only tells whether the overall comparison is significant or not, it is followed by post hoc analyses in order to identify which two levels are different (Table 2 supplement).
**FIGURE 1:** *Scores plot between the selected PCs. The explained variances are shown in brackets.* TABLE_PLACEHOLDER:TABLE 2
## 3.3 Principal component analysis (PCA)
Figure 1 shows the results of the PCA; the separation trend among each group is shown, indicating that each group had a unique metabolic spectrum.
## 3.4 Partial least squares-discriminant analysis (PLS-DA)
PLS-DA was used to better distinguish the overall difference in the metabolic spectrum among each group and determine the metabolites that were most characteristic of each group. Figure 2 shows the 2D score plot between selected components; Figure 2 supplement shows the classification performance with different number of components; and Figure 3 shows important features identified by PLS-DA. To assess the significance of class discrimination, a permutation test was performed. In each permutation, a PLS-DA model was built between the data and the permuted class labels using the optimal number of components determined by cross validation for the model based on the original class assignment (Figure 3 supplement). VIP scores are calculated for each component. When more than components are used to calculate the feature importance, the average of the VIP scores was used. The average VIP value of caffeine was the highest among the metabolites that decreased with mGFR deterioration.
**FIGURE 2:** *Scores plot between the selected PCs. The explained variances are shown in brackets.* **FIGURE 3:** *Important features identified by PLS-DA. The colored boxes on the right indicate the relative concentrations of the corresponding metabolite in each group under study.*
## 3.5 Pathway analysis
Differential metabolites were selected by one-way ANOVA, PCA, and PLS-DA. The differential metabolites were $p \leq 0.05$ in one one-way ANOVA and VIP>1 in PLS-DA. The metabolites of VIP>1 are presented in Table 3 plsda_vip supplement. The KEGG ID of each differential metabolite was found through MetaboAnalyst 5.0. KEGG IDs of differential metabolites were uploaded to MBRole2.0 and the background was set as the full database for KEGG pathway analysis. MBRole performs an overrepresented (enrichment) analysis of categorical annotations for a set of compounds of interest. These categorical annotations correspond to biological and chemical information available in several public databases and software. Caffeine metabolism, metabolic pathways, pyrimidine metabolism, and histidine metabolism were classified as important in Table 3. Caffeine metabolism was classified as the most important pathway compared with the three other identified pathways. Metabolic pathways included the other three metabolic pathways, so further analysis was directed at the metabolites enriched in these three metabolic pathways (Figure 4).
## 3.6 Caffeine metabolism accumulation of multiple metabolites decreased with the deterioration of CKD
The caffeine metabolism pathway mapped by MBRole2.0 enriched 12 metabolites (Table 3), including Xanthine, Xanthosine, Theophylline, Theobromine, Caffeine, Paraxanthine, 7-Methylxanthine, 3-Methylxanthine, 1-Methylxanthine, 1,3,7-Trimethyluric acid, 5-Acetylamino-6-formylamino-3-methyluracil, and 5-Acetylamino-6-amino-3-methyluracil, which were circled in red in Figure 4A. The levels of paraxanthine, theobromine, caffeine, and theophylline decreased with renal deterioration (metabolites in red boxes in Figure 4A). Xanthosine and xanthine increased with renal deterioration (metabolites in green boxes in Figure 4A).
## 4 Discussion
In this study, we enrolled 145 CKD individuals who were divided into four groups based on mGFR. Many differential metabolites were screened out by metabolic profiling. Through pathway analysis of differential metabolites, we report for the first that the caffeine metabolism pathway is critical in CKD. We showed that in this pathway, paraxanthine, theobromine, caffeine, and theophylline decreased with poorer renal function, while xanthosine and xanthine increased. Caffeine was the most important metabolite, decreasing with deteriorating renal function.
Previous studies have found that coffee consumption may reduce the risk of CKD (Li et al., 2021), showing a negative association between caffeine intake and all-cause mortality in patients with CKD (Bigotte et al., 2019). However, other studies have shown that caffeine exacerbates hypertension in rats with polycystic kidney disease (Tanner and Tanner, 2001), potentiates the development of more severe tubulointerstitial changes, and increases focal glomerulosclerosis (Tofovic et al., 2002). High caffeine-sugar content increases the incidence of cardiovascular disease and tissue inflammation by altering lipid profiles and blood glucose (Eltahir et al., 2020). Yu et al. ( Yu et al., 2016) found that acute caffeine intake causes an acute increase in blood pressure, while chronic caffeine intake decreases blood pressure; the latter may be related to a diuretic effect. Chronic caffeine consumption also reduces sodium absorption, contributing to its antihypertensive effects in salt-sensitive rats (Wei et al., 2018). This study found that caffeine metabolism was the most important pathway and that caffeine was the most important metabolite.
In our study, paraxanthine, theobromine, and theophylline decreased with poorer renal function. It is known that theobromine activates sirtuin one to reduce extracellular matrix accumulation in the kidneys of diabetic rats (Papadimitriou et al., 2015); a single dose of prophylactic theophylline has been shown to prevent acute kidney injury (AKI)/severe kidney dysfunction in term neonates with severe birth asphyxia (Bhatt et al., 2019). This study corroborated that theobromine and theophylline improve the progression of CKD. The effect of paraxanthine on CKD is still unclear. In this study, paraxanthine decreased with decreasing mGFR, which may also improve the progression of CKD.
We also found that xanthosine and xanthine increased with poorer renal function. Chen et al. ( Chen et al., 2020) demonstrated that xanthosine is associated with significantly greater risks of CKD progression. Xanthine is an intermediate metabolite of uric acid (UA), converted by xanthine oxidase (XO). Previous research suggests a pathogenic role of hyperuricemia in the development of CKD (Uedono et al., 2015; Mallat et al., 2016). XO inhibitors have been suggested to slow the progression of kidney disease (Pisano et al., 2017) but this remains controversial (Kimura et al., 2018). The present study supports that xanthosine and xanthine are associated with CKD progression.
We found that metabolites that are associated with CKD progression can be converted to more ‘desirable’ metabolites, as shown in Figure 4A. This conversion involves many enzymes that are shown in Figure 4A. Thus, the regulation of these enzymes offers a potential target for CKD treatment. In subsequent research, we plan to focus on these enzymes. One limitation was that we did not record the baseline caffeine ingestion in our participants; the influence of caffeine ingestion on renal function in CKD should be observed in subsequent studies.
In this study, we found that in pyrimidine metabolism and histidine metabolism (Figures 4B,C), orotidine-5P, pseudouridine, 3-Ureidopropionate, urea, 1-Methyl-L-histidine, Hydantoin-5-propionate, N-Formimino-L-glutamate and imidazole lactate rise as mGFR decreases. Consistent with previously reported results, pseudouridine is extremely correlated with mGFR and might be an ideal biomarker for CKD (Peng et al., 2022). Urea is an endogenous marker in CKD (Weiner et al., 2015). The significance of the accumulation of the other six metabolites in CKD has not been reported. These eight metabolites have limited significance in the search for targets to protect kidney function. Moreover, in histidine metabolism, ergothioneine which is absorbed from the intestine through food intake may function as a major antioxidant (Cheah et al., 2017), and the protective effect on the kidney can be further investigated.
This study has potential limitations. Based on the metabolomic approach, we identified the significance of caffeine metabolism in CKD, but the mechanisms linking caffeine metabolism to CKD have yet to be clarified, and the exact intricate mechanism needs additional animal experiments and prospective studies. Future investigations will need to include more animal experiments and prospective studies to elucidate the mechanism and the significance of caffeine metabolism in CKD identified in this study.
## 5 Conclusion
In conclusion, metabolic profiling identified caffeine metabolism as the most important pathway in CKD progression. Decreased renal function was associated with decreased paraxanthine, theobromine, caffeine, and theophylline, and with increased xanthosine and xanthine. Caffeine was the most important metabolite associated with CKD deterioration.
## 6 Summary at a glance
Metabolic profiling was used to identify significant metabolic pathways in CKD progression. Four metabolic pathways were classified as important in the progression of CKD. Caffeine metabolism appears to be the most important pathway. Caffeine is the most important metabolite that decreases with the deterioration of CKD stage.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by the Institutional Review Board of Ethics Commission of Kiang Wu Hospital (KWH 2018-001). 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
All authors have read and approved the manuscript. XG and HP: Conceptualization, methodology, and writing - original draft preparation. HP: Funding acquisition. PL: Visualization. LT: Investigation. JF: Software and Validation. CA: Supervision. TT and TsT: Data curation. HP and XL: Writing-review, funding and editing.
## 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. Bhatt G. C., Gogia P., Bitzan M., Das R. R.. **Theophylline and aminophylline for prevention of acute kidney injury in neonates and children: A systematic review**. *Arch. Dis. Child.* (2019) **104** 670-679. DOI: 10.1136/archdischild-2018-315805
2. Bigotte V. M., Magrico R., Viegas D. C., Leitao L., Neves J. S.. **Caffeine consumption and mortality in chronic kidney disease: A nationally representative analysis**. *Nephrol. Dial. Transpl.* (2019) **34** 974-980. DOI: 10.1093/ndt/gfy234
3. Brunetto M. A., Ruberti B., Halfen D. P., Caragelasco D. S., Vendramini T. H. A., Pedrinelli V.. **Healthy and chronic kidney disease (CKD) dogs have differences in serum metabolomics and renal diet may have slowed disease progression**. *Metabolites* (2021) **11** 782. DOI: 10.3390/metabo11110782
4. Cheah I. K., Tang R. M., Yew T. S., Lim K. H., Halliwell B.. **Administration of pure ergothioneine to healthy human subjects: Uptake, metabolism, and effects on biomarkers of oxidative damage and inflammation**. *Antioxid. Redox Signal* (2017) **26** 193-206. DOI: 10.1089/ars.2016.6778
5. Chen D. Q., Cao G., Chen H., Argyopoulos C. P., Yu H., Su W.. **Identification of serum metabolites associating with chronic kidney disease progression and anti-fibrotic effect of 5-methoxytryptophan**. *Nat. Commun.* (2019) **10** 1476. DOI: 10.1038/s41467-019-09329-0
6. Chen Y. Y., Chen D. Q., Chen L., Liu J. R., Vaziri N. D., Guo Y.. **Microbiome-metabolome reveals the contribution of gut-kidney axis on kidney disease**. *J. Transl. Med.* (2019) **17** 5. DOI: 10.1186/s12967-018-1756-4
7. Chen Y., Zelnick L. R., Wang K., Hoofnagle A. N., Becker J. O., Hsu C. y.. **Kidney clearance of secretory solutes is associated with progression of CKD: The CRIC study**. *J. Am. Soc. Nephrol.* (2020) **31** 817-827. DOI: 10.1681/asn.2019080811
8. Dou F., Miao H., Wang J. W., Chen L., Wang M., Chen H.. **An integrated lipidomics and phenotype study reveals protective effect and biochemical mechanism of traditionally used alisma orientale juzepzuk in chronic kidney disease**. *Front. Pharmacol.* (2018) **9** 53. DOI: 10.3389/fphar.2018.00053
9. Earl D. C., Ferrell P. J., Leelatian N., Froese J. T., Reisman B. J., Irish J. M.. **Discovery of human cell selective effector molecules using single cell multiplexed activity metabolomics**. *Nat. Commun.* (2018) **9** 39. DOI: 10.1038/s41467-017-02470-8
10. Eltahir H. M., Alamri G., Alamri A., Aloufi A., Nazmy M., Elbadawy H.. **The metabolic disorders associated with chronic consumption of soft and energy drinks in rats**. *Acta Biochim. Pol.* (2020) **67** 79-84. DOI: 10.18388/abp.2020_2914
11. Feng Y. L., Cao G., Chen D. Q., Vaziri N. D., Chen L., Zhang J.. **Microbiome-metabolomics reveals gut microbiota associated with glycine-conjugated metabolites and polyamine metabolism in chronic kidney disease**. *Cell Mol. Life Sci.* (2019) **76** 4961-4978. DOI: 10.1007/s00018-019-03155-9
12. Gordin D., Shah H., Shinjo T., St-Louis R., Qi W., Park K.. **Characterization of glycolytic enzymes and pyruvate kinase M2 in type 1 and 2 diabetic nephropathy**. *Diabetes Care* (2019) **42** 1263-1273. DOI: 10.2337/dc18-2585
13. Kimura K., Hosoya T., Uchida S., Inaba M., Makino H., Maruyama S.. **Febuxostat therapy for patients with stage 3 CKD and asymptomatic hyperuricemia: A randomized trial**. *Am. J. Kidney Dis.* (2018) **72** 798-810. DOI: 10.1053/j.ajkd.2018.06.028
14. Lai Z., Tsugawa H., Wohlgemuth G., Mehta S., Mueller M., Zheng Y.. **Identifying metabolites by integrating metabolome databases with mass spectrometry cheminformatics**. *Nat. Methods* (2018) **15** 53-56. DOI: 10.1038/nmeth.4512
15. Lanzon B., Martin-Taboada M., Castro-Alves V., Vila-Bedmar R., Gonzalez de Pablos I., Duberg D.. **Lipidomic and metabolomic signature of progression of chronic kidney disease in patients with severe obesity**. *Metabolites* (2021) **11** 836. DOI: 10.3390/metabo11120836
16. Li Y., Li W., Lu Y., Zhang J.. **Coffee consumption is associated with a decreased risk of incident chronic kidney disease: A protocol for systematic review and meta-analysis**. *Med. Baltim.* (2021) **100** e27149. DOI: 10.1097/md.0000000000027149
17. Mallat S. G., Al K. S., Tanios B. Y., Jurjus A.. **Hyperuricemia, hypertension, and chronic kidney disease: An emerging association**. *Curr. Hypertens. Rep.* (2016) **18** 74. DOI: 10.1007/s11906-016-0684-z
18. Papadimitriou A., Silva K. C., Peixoto E. B., Borges C. M., Lopes de Faria J. M., Lopes de Faria J. B.. **Theobromine increases NAD(+)/Sirt-1 activity and protects the kidney under diabetic conditions**. *Am. J. Physiol. Ren. Physiol.* (2015) **308** F209-F225. DOI: 10.1152/ajprenal.00252.2014
19. Peng H., Liu X., Aoieong C., Tou T., Tsai T., Ngai K.. **Identification of metabolite markers associated with kidney function**. *J. Immunol. Res.* (2022) **2022** 1-9. DOI: 10.1155/2022/6190333
20. Pisano A., Cernaro V., Gembillo G., D’Arrigo G., Buemi M., Bolignano D.. **Xanthine oxidase inhibitors for improving renal function in chronic kidney disease patients: An updated systematic review and meta-analysis**. *Int. J. Mol. Sci.* (2017) **18** 2283. DOI: 10.3390/ijms18112283
21. Rhee E. P., Clish C. B., Wenger J., Roy J., Elmariah S., Pierce K. A.. **Metabolomics of chronic kidney disease progression: A case-control analysis in the chronic renal insufficiency cohort study**. *Am. J. Nephrol.* (2016) **43** 366-374. DOI: 10.1159/000446484
22. Rysz J., Franczyk B., Lawinski J., Olszewski R., Cialkowska-Rysz A., Gluba-Brzozka A.. **The impact of CKD on uremic toxins and gut microbiota**. *Toxins (Basel).* (2021) **13** 252. DOI: 10.3390/toxins13040252
23. Shah V. O., Townsend R. R., Feldman H. I., Pappan K. L., Kensicki E., Vander Jagt D. L.. **Plasma metabolomic profiles in different stages of CKD**. *Clin. J. Am. Soc. Nephrol.* (2013) **8** 363-370. DOI: 10.2215/cjn.05540512
24. Shen B., Yi X., Sun Y., Bi X., Du J., Zhang C.. **Proteomic and metabolomic characterization of COVID-19 patient sera**. *Cell* (2020) **182** 59-72.e15. DOI: 10.1016/j.cell.2020.05.032
25. Tanner G. A., Tanner J. A.. **Chronic caffeine consumption exacerbates hypertension in rats with polycystic kidney disease**. *Am. J. Kidney Dis.* (2001) **38** 1089-1095. DOI: 10.1053/ajkd.2001.28614
26. Tofovic S. P., Kost C. J., Jackson E. K., Bastacky S. I.. **Long-term caffeine consumption exacerbates renal failure in obese, diabetic, ZSF1 (fa-fa(cp)) rats**. *Kidney Int.* (2002) **61** 1433-1444. DOI: 10.1046/j.1523-1755.2002.00278.x
27. Uedono H., Tsuda A., Ishimura E., Yasumoto M., Ichii M., Ochi A.. **Relationship between serum uric acid levels and intrarenal hemodynamic parameters**. *Kidney Blood Press Res.* (2015) **40** 315-322. DOI: 10.1159/000368507
28. Wei X., Lu Z., Yang T., Gao P., Chen S., Liu D.. **Stimulation of intestinal Cl- secretion through CFTR by caffeine intake in salt-sensitive hypertensive rats**. *Kidney Blood Press Res.* (2018) **43** 439-448. DOI: 10.1159/000488256
29. Weiner I. D., Mitch W. E., Sands J. M.. **Urea and ammonia metabolism and the control of renal nitrogen excretion**. *Clin. J. Am. Soc. Nephrol.* (2015) **10** 1444-1458. DOI: 10.2215/cjn.10311013
30. Xu X., Eales J. M., Akbarov A., Guo H., Becker L., Talavera D.. **Molecular insights into genome-wide association studies of chronic kidney disease-defining traits**. *Nat. Commun.* (2018) **9** 4800. DOI: 10.1038/s41467-018-07260-4
31. Yu H., Yang T., Gao P., Wei X., Zhang H., Xiong S.. **Caffeine intake antagonizes salt sensitive hypertension through improvement of renal sodium handling**. *Sci. Rep.* (2016) **6** 25746. DOI: 10.1038/srep25746
32. Zhao Y. Y., Chen H., Tian T., Chen D. Q., Bai X., Wei F.. **A pharmaco-metabonomic study on chronic kidney disease and therapeutic effect of ergone by UPLC-QTOF/HDMS**. *PLoS One* (2014) **9** e115467. DOI: 10.1371/journal.pone.0115467
|
---
title: 'Endogenous relapse and exogenous reinfection in recurrent pulmonary tuberculosis:
A retrospective study revealed by whole genome sequencing'
authors:
- Wencong He
- Yunhong Tan
- Zexuan Song
- Binbin Liu
- Yiting Wang
- Ping He
- Hui Xia
- Fei Huang
- Chunfa Liu
- Huiwen Zheng
- Shaojun Pei
- Dongxin Liu
- Aijing Ma
- Xiaolong Cao
- Bing Zhao
- Xichao Ou
- Shengfen Wang
- Yanlin Zhao
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC9981662
doi: 10.3389/fmicb.2023.1115295
license: CC BY 4.0
---
# Endogenous relapse and exogenous reinfection in recurrent pulmonary tuberculosis: A retrospective study revealed by whole genome sequencing
## Abstract
### Background
Tuberculosis may reoccur due to reinfection or relapse after initially successful treatment. Distinguishing the cause of TB recurrence is crucial to guide TB control and treatment. This study aimed to investigate the source of TB recurrence and risk factors related to relapse in Hunan province, a high TB burden region in southern China.
### Methods
A population-based retrospective study was conducted on all culture-positive TB cases in Hunan province, China from 2013 to 2020. Phenotypic drug susceptibility testing and whole-genome sequencing were used to detect drug resistance and distinguish between relapse and reinfection. Pearson chi-square test and Fisher exact test were applied to compare differences in categorical variables between relapse and reinfection. The Kaplan–Meier curve was generated in R studio (4.0.4) to describe and compare the time to recurrence between different groups. $p \leq 0.05$ was considered statistically significant.
### Results
Of 36 recurrent events, 27 ($75.0\%$, $\frac{27}{36}$) paired isolates were caused by relapse, and reinfection accounted for $25.0\%$ ($\frac{9}{36}$) of recurrent cases. No significant difference in characteristics was observed between relapse and reinfection (all $p \leq 0.05$). In addition, TB relapse occurs earlier in patients of Tu ethnicity compared to patients of Han ethnicity ($p \leq 0.0001$), whereas no significant differences in the time interval to relapse were noted in other groups. Moreover, $83.3\%$ ($\frac{30}{36}$) of TB recurrence occurred within 3 years. Overall, these recurrent TB isolates were predominantly pan-susceptible strains ($71.0\%$, $\frac{49}{69}$), followed by DR-TB ($17.4\%$, $\frac{12}{69}$) and MDR-TB ($11.6\%$, $\frac{8}{69}$), with mutations mainly in codon 450 of the rpoB gene and codon 315 of the katG gene. $11.1\%$ ($\frac{3}{27}$) of relapse cases had acquired new resistance during treatment, with fluoroquinolone resistance occurring most frequently ($7.4\%$, $\frac{2}{27}$), both with mutations in codon 94 of gyrA.
### Conclusion
Endogenous relapse is the main mechanism leading to TB recurrences in Hunan province. Given that TB recurrences can occur more than 4 years after treatment completion, it is necessary to extend the post-treatment follow-up period to achieve better management of TB patients. Moreover, the relatively high frequency of fluoroquinolone resistance in the second episode of relapse suggests that fluoroquinolones should be used with caution when treating TB cases with relapse, preferably guided by DST results.
## Introduction
Tuberculosis (TB) remains a major global public health issue, with an estimated 10.0 million new cases and more than 1.2 million deaths from TB worldwide in 2019 [World Health Organization (WHO), 2020]. Although most TB patients can be cured after the introduction of a standard combination of chemotherapy, some patients who complete an appropriate course of treatment still experience a subsequent episode, or TB recurrence (Zong et al., 2018). Patients with recurrent TB often require longer rounds of treatment with more toxic drugs, which reduces the success of treatment, leads to further transmission of *Mycobacterium tuberculosis* (MTB), and increases the burden of TB (Liu et al., 2020).
Recurrence of TB can be caused by relapse, also known as endogenous reactivation of the initial infection, or by exogenous reinfection with new MTB strains (Ruan et al., 2022). The proper discrimination between relapse and reinfection is essential for adjusting TB control measures. High relapse rates indicate inadequate TB treatment, whereas high rates of reinfection reveal poor TB cases management with many missed TB cases circulating in the community (Folkvardsen et al., 2020; Du et al., 2021).
The advent of molecular genotyping techniques for MTB has made it possible to assess the magnitude of endogenous relapse versus exogenous reinfection (Bandera et al., 2001; Lambert et al., 2003). These genomic-based typing methods include IS6110 fingerprinting, mycobacterial interspersed repetitive unit-variable number of tandem repeat (MIRU-VNTR), spoligotyping, and whole-genome-sequencing (WGS) (Barbier and Wirth, 2016). However, different genotyping methods often affect the reinfection rate due to different resolutions (Jagielski et al., 2016). Compared to traditional genotyping methods, WGS based on the full-genome of MTB strains has the distinct advantage by allowing the discrimination of MTBC strains at the highest resolution and simultaneously enabling detailed resistance predictions for almost all drugs (Roetzer et al., 2013; Walker et al., 2015).
Despite tremendous progress in TB control, China still has the second-highest TB burden worldwide [World Health Organization (WHO), 2020]. In addition, the presence of TB recurrence can further increase the burden of TB. A better understanding of the sources of recurrent TB and its related risk factors is essential for targeted interventions and for reducing the frequency of TB (Shen et al., 2017). However, limited efforts have been made to identify the major cause of TB recurrence in China, particularly in Hunan province, which has one of the highest TB burdens in China, with an estimated annual TB incidence of 94 cases per 100,000 population (He et al., 2022). To address this concern, we conducted a retrospective study among recurrent TB cases from five counties in Hunan province. We used WGS to determine whether TB recurrence was mainly caused by reinfection or relapse. We performed phenotypic drug susceptibility testing (DST) to compare in vitro DST results between the first and second TB episodes. We also collected demographic information and clinical characteristics of recurrent TB cases to analyze risk factors associated with reinfection and relapse.
## Study population
This retrospective study was conducted based on five DRS (drug resistance surveillance) sites (5 counties: Hecheng, Yongshun, Qidong, Taojiang, and Leiyang) in Hunan province, which were established according to the first national survey of drug resistance in China (Zhao et al., 2012). In these five counties, all suspected pulmonary TB cases from general hospitals or health centers are referred to local designated TB hospitals for confirmed diagnosis and treatment. All TB cases aged 15 years or older with bacteriologically confirmed (sputum-smear positive or culture positive) by local designated TB hospitals or clinics between January 1, 2013 to December 31, 2020 were included in this study. Positive sputum samples were cultured and isolated on Lowenstein-*Jensen medium* at the county-level and then sent to National Tuberculosis Reference Laboratory (NTRL). Information on these TB cases, including demographic characteristics and medical records, is collected at the time of patients’ visits and stored electronically in the National Tuberculosis Information Management System (TBIMS). To identify recurrent TB cases, the medical records of TB patients diagnosed between 2013 and 2020 were extracted from TBIMS on June 30, 2022 and collated using the method described previously (Shen et al., 2017). TB cases with any of the followings were excluded from the further study: [1] unsuccessful treatment outcomes of their initial TB episode (e.g., lose to follow-up, death, treatment failure, etc.); [ 2] less than 6 months of the recurrent interval (the time interval between the recorded end date of the treatment and the date of the re-diagnosis of active TB); [3] strains with subculture failure or contamination; [4] failed extraction of DNA or WGS errors.
## Drug susceptibility testing
All MTB strains isolated from recurrent TB cases were previously stored in 7H9 medium containing $25\%$ glycerin at–80°C refrigerator, and then were thawed and re-cultured on L-J medium for further study. MTB isolates in the logarithmic phase were subjected to drug susceptibility testing against rifampicin, isoniazid, ethambutol, streptomycin, ofloxacin, moxifloxacin, kanamycin, and amikacin using MYCOTB plate (Thermo Fisher Scientific, United States). Previous studies have demonstrated the good accuracy and reproducibility of the MYCOTB plate, which can be used as an alternative method for DST (Xia et al., 2017; Wu et al., 2019). All procedures were performed by trained staff at the national TB reference laboratory of China, as described elsewhere (He et al., 2022). H37Rv (ATCC 27294) was used as pan-susceptible control in each batch of DST. The concentration ranges and cut-off values for determining resistance or sensitivity for each drug used in this study were depicted previously (He et al., 2022). All DSTs were conducted twice to ensure the accuracy of DST results.
## DNA extraction and sequencing
MTB strains were scraped from L-J solid slants, and genomic DNA was obtained from isolates with the cetyltrimethylammonium bromide (CTAB) method as described previously (Shao et al., 2021). The quality and concentration of genomic DNA were assessed by NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific, USA) and Qubit 2.0 fluorometer (Invitrogen, Thermo Fisher Scientific, USA), respectively. Whole genome sequencing was performed by Annoroad Gene Technology company (Beijing, China) using Illumina Hiseq X10 (Illumina, Inc.) with 2 × 150 paired-end (PE) strategies.
## Phylogenetic analysis
In brief, the quality control of raw reads was examined by FastQC (v0.11.9),1 and reads were filtered with Trimmomatic (v 0.38) using default values and minimum Phred Quality score of 20 (Bolger et al., 2014). Retained paired-end reads were mapped to the reference genome H37Rv (GenBank accession NC_000962.3) using BWA-MEM software (v. 0.7.17) (Freschi et al., 2021). Variants information including single nucleotide polymorphisms (SNPs) and small insertion/deletions (Indel) were detected using SAMtools (v1.3.1) and GATK (v.3.8.0) (He et al., 2022). The variants that met the following criteria were kept for further analysis: minimum coverage depth of 10X, Q20 minimum quality score for each variant, and more than $75\%$ allele frequency (He et al., 2022).
SNPs located in repeating regions of the genome such as PE/PPE/PGRs genes, phage sequence, insertions, and mobile elements were excluded. The remaining SNPs in each isolate were pooled into a sequence based on the position, and SNP positions present in at least $95\%$ of isolates were integrated into a sequence alignment (Liu et al., 2022). The maximum likelihood trees were constructed using a general time reversible model in MEGA-X (v.10.1.8) with bootstraps of 1,000 replicates (He et al., 2022). The phylogenetic tree was visualized and annotated using iTOL.2 Snp-dists (v.0.8.2) was used to calculate the SNP distance between pairs of isolates. QuantTB (v. 1.01)3 was used to identify mixed infection of MTB (Anyansi et al., 2020).
## Lineage and genotypic drug resistance prediction
Fast-lineage-caller package (v.3.2)4 was used to call lineage and sub-lineage information of M. tuberculosis. TB Profiler (v.3.0.8)5 was used to predict genotypic drug susceptibility.
## Statistical analysis
Pearson chi-square test and Fisher exact test were used to compare differences in categorical variables between relapse and reinfection. The Kaplan–Meier curve was generated in R studio (4.0.4) to describe and compare the time to recurrence between different groups. All statistical analysis was performed in the SPSS version 18.0 software (SPSS Inc., Chicago, Illinois). $p \leq 0.05$ was considered statistically significant.
## Definition
TB recurrence was defined as a patient who was cured or completed treatment during the most recent course of treatment and then was re-diagnosed with a new TB episode [World Health Organization (WHO), 2013]. Reinfection was defined as a recurrent disease episode caused by a new TB strain with a genetic distance of more than 12 SNPs compared with the strain that caused the original episode. Relapse was defined as a genetic distance of 12 or fewer SNPs between paired strains isolated from two episodes in TB recurrence (Li et al., 2022). The recurrent interval was defined as the time interval between the recorded end date of the initial TB treatment and the date of the re-diagnosis of active TB (Ruan et al., 2022). Based on the phenotypic drug susceptibility testing, Pan-Susceptible was defined as MTB strains that were susceptible to all anti-TB drugs tested in this study (including rifampicin, isoniazid, ethambutol, streptomycin, moxifloxacin, ofloxacin, kanamycin and amikacin), whereas Drug-resistant was defined as MTB strains that were resistant to at least one of these anti-TB drugs but not include the concurrent resistance to rifampicin and isoniazid. MDR-TB was defined as MTB resistance to at least isoniazid and rifampicin. Pre-XDR-TB was defined as MDR-TB with additional resistance to any fluoroquinolones (moxifloxacin or ofloxacin) or any second-line injectable drugs (amikacin or kanamycin), but not both. XDR-TB was defined as MDR-TB with additional resistance to any fluoroquinolones and any second-line injectable drugs.
## Description of the study population
A total of 2,416 bacteriologically confirmed TB cases aged 15 years or older were collected between Jan. 2013 and Dec. 2020. Of which, $88.6\%$ ($\frac{2141}{2416}$) cases were successfully treated, while 275 ($11.4\%$) patients experienced treatment failure, loss to follow-up, treatment interruption, adverse reactions, or death. Overall, 117 ($5.5\%$, $\frac{117}{2141}$) successfully treated cases that experienced TB recurrences, 25 recurrent TB cases were excluded due to their recurrent interval being less than 6 months, and finally, 92 recurrent TB cases were included in further analysis. Among them, $56.5\%$ ($\frac{52}{92}$) had recurrent strains with both episodes. After excluding subculture failure or contamination of any paired isolates ($$n = 7$$) and failure to extract DNA or WGS ($$n = 6$$). Finally, 39 recurrent TB patients with paired strains were enrolled in the final analysis. Of these, one patient had a third episode during the study period, for a total of 79 MTB isolates and 41 recurrent events (Figure 1).
**Figure 1:** *Flowchart of recurrent TB cases included and excluded from the study.*
## Patient characteristics
The characteristics of recurrent TB cases at their primary episode were described below. Among the 39 recurrent TB cases included in this study, the median and mean age of patients was 54.0 [interquartile range (IQR), 45.0–65.0] and 54.4 ± 12.6 years old. The majority of patients were male ($84.6\%$, $\frac{33}{39}$) while $15.4\%$ ($\frac{6}{39}$) were female. Almost $90\%$ ($89.7\%$, $\frac{35}{39}$) of patients were farmers. In terms of treatment history, new cases accounted for $94.9\%$ ($\frac{37}{39}$) of the total. $7.7\%$ ($\frac{3}{39}$) and $12.8\%$ ($\frac{5}{39}$) of recurrent TB cases had complications of hepatitis B and diabetes, respectively. The chest X-ray showed that $30.8\%$ ($\frac{12}{39}$) of patients had cavitation in the first episode of TB. As for HIV status, 12 ($30.8\%$, $\frac{12}{39}$) patients were HIV-negative, while 27 ($69.2\%$, $\frac{27}{39}$) patients had unknown HIV infection status.
## TB relapse and reinfection identified by SNP distance
The whole genome sequencing data of 79 MTB strains collected from 39 recurrent TB cases were first analyzed to determine the presence of mixed infections. Five recurrent TB cases (7 strains of mixed infections in total) were excluded from further analysis because any of their paired strains were identified as having at least two different strains. The paired SNP distances were calculated on the remaining 34 recurrent TB cases, including one patient (patient 15) with three TB episodes (patient 15–1, 15–2, 15–3), thus involving a total of 69 MTB strains and 36 recurrent events (Supplementary Figure S1). Identical genotypes were defined as strains that differed by no more than 12 SNPs (Li et al., 2022). Of 36 recurrent events, 27 ($75.0\%$, $\frac{27}{36}$) paired isolates (patient 1, 2, 3, 4, 5, 7, 8, 9, 11, 12, 13, 14, 15[1–3], 16, 20, 21, 22, 23, 24, 25, 26, 27, 28, 30, 31, 32 and 33) had 5 or fewer SNP differences (Supplementary Figure S1), indicating relapse. Nine ($25.0\%$, $\frac{9}{36}$) paired isolates (patient 6, 10, 15[1–2], 15[2–3], 17, 18, 19, 29, and 34) differed by >12 SNPs, suggesting exogenous reinfection with a new strain of MTB (Supplementary Figure S1). Comparison with SNP differences was also shown in Figure 2, identifying two major groups: paired isolates from relapsed cases had five or fewer SNP differences, whereas paired strains from reinfected cases had a dramatically higher number of SNP differences (range 185–1,074) except for 2 paired isolates (patient 17 and 18) with SNP differences of 14 (Supplementary Figure S1; Figure 2).
**Figure 2:** *The distribution of SNP differences between paired isolates. Reinfection was defined as a recurrent disease episode caused by a new TB strain with a genetic distance of more than 12 SNPs compared with the strain that caused the original episode. Relapse was defined as a genetic distance of 12 or fewer SNPs between paired strains isolated from two episodes in TB recurrence. The SNP differences between paired isolates were calculated by using Snp-dists (v.0.8.2).*
## Phylogenetic reconstructions and drug-resistant profile
The phylogenetic tree was constructed based on 6,847 high-quality SNPs (Figure 3). Fast-lineage-caller analysis showed that the majority ($59.4\%$, $\frac{41}{69}$) of recurrent TB isolates were lineage 2, and $40.6\%$ ($\frac{28}{69}$) were lineage 4. All the isolate pairs from relapse cases were close together on the tree, whereas almost the reinfected isolate pairs appeared quite divergent on the tree (marked in different colors) (Figure 3). We also analyzed the community transmission of these recurrent TB cases, as demonstrated in Figure 3, TB strains collected from different individuals did not show high sequence similarity. Among these recurrent TB isolates, pan-susceptible predominated, accounting for $71.0\%$ ($\frac{49}{69}$), with only 8 ($11.6\%$, $\frac{8}{69}$) and 12 ($17.4\%$, $\frac{12}{69}$) were identified as MDR-TB and DR-TB, respectively. To rationalize these phenotypic drug-resistance, genetic mutations were predicted based on WGS data. A total of 9 recurrent TB strains were identified as genomic MDR-TB, mainly with mutations in codon 450 of the rpoB gene and codon 315 of the katG gene (Table 1). Moreover, 2 strains had detectable drug-resistant mutations to four first-line anti-TB drugs (rifampicin, isoniazid, pyrazinamide, and ethambutol) simultaneously. We further compared the drug-resistant profiles between paired isolates to clarify the development of acquired resistance during treatment. As shown in Figure 3, three relapse cases had acquired new resistance during treatment: two (patient 21 and 26) to fluoroquinolones were both due to mutations in codon 94 of gyrA and one (patient 27) to ethambutol due to a mutation in codon 406 of embB, resulting in amino acid substitution from Gly to Asp (Table 1). Of note, the strain from patient 21, which was MDR with additional resistance to ethambutol and pyrazinamide in the first isolate, had progressed to pre-XDR in their second relapsed isolate. Interestingly, one patient (patient 15) had three TB episodes between 2015 and 2019, of which the first and third episodes isolated the identical TB strain, both MDR-TB, while the new strain isolated from the second was pan-susceptible (Figure 3, Table 1). In addition, one patient (patient 10) was initially infected with a pan-susceptible strain and subsequently reinfected with a new strain that harbored gene mutations related to rifampin and isoniazid resistance (Table 1).
**Figure 3:** *Phylogenetic tree and drug-resistant profile of 69 MTB strains from 34 recurrent patients. Inner band indicates TB recurrence classification (reinfection represents strain pairs differences >12 SNPs, whereas relapse represents strain pairs differences ≤12 SNPs) and the outer band suggests phenotypic drug-resistant type (see legend). Solid circles indicate genetic drug resistance detected by TB-profiler. Reinfected patients are highlighted with different colors and curves connecting patients’ samples in the phylogeny indicating paired strains isolated from the same patient.* TABLE_PLACEHOLDER:Table 1
## Comparison of the characteristics between relapse and reinfection
We analyzed the differences in the characteristics between TB relapse and reinfection. As summarized in Table 2, all these demographic factors and clinical characteristics of patients, such as gender, age, occupation, and comorbidities et al., as well as genetic background and drug-resistant type of strains, had no significant effect on the proportion of TB relapse (all $p \leq 0.05$). In addition, more than $80\%$ of TB recurrence occurred within three years after completion of treatment for the index episode (Table 2). The median of the recurrent time interval to relapse was 17.6 months (IQR, 12.9–28.3 months) compared with 24.3 months (IQR, 12.9–31.5 months) for reinfection cases, and there was no significant association between relapse and earlier recurrence ($$p \leq 0.51$$) (Figure 4A). We further assessed the time interval to relapse stratified by gender, nationality, pulmonary cavity, strain drug-resistant type, and genetic background. As shown in Figure 4, TB relapse occurs earlier in patients of Tu ethnicity compared to patients of Han ethnicity ($p \leq 0.0001$), whereas no significant differences in the time interval to relapse were noted in other groups.
## SNPs in relapse isolate
Of 39 SNPs and small indels (insertion–deletion) identified between the relapse pairs, 23 were non-synonymous polymorphisms. These mutations are located in genes encoding proteins with various functions, such as cell wall and cellular process, lipid metabolism, and information pathways (Supplementary Table S1). In three and four cases, these mutations were involved in drug-resistant related genes and growth advantage regions, respectively. We also identified 6 indels differences that result in frame-shifts within protein-coding regions, but all of these indels were located in non-essential regions (Supplementary Table S1).
## Discussion
To our knowledge, this is the first longitudinal population-based study of sufficient duration to investigate TB recurrence using WGS in Hunan, China. The current study found a relatively high frequency of mixed infections among recurrent TB cases. After excluding patients with mixed infections, our study demonstrated that TB recurrence in Hunan province is mainly caused by endogenous reactivation of the initial infection (relapse), and reinfection accounted for a quarter of recurrent cases. In addition, our study found TB recurrence can occur even more than 4 years after treatment completion of the most recent episode, mainly within 3 years. Evidence of acquired resistance during treatment was also observed in this study, with fluoroquinolone resistance occurring most frequently.
Understanding the proportion of reinfection and relapse will help to implement better post-treatment follow-up and reduce TB burden. Unexpectedly high rates of reinfection suggest that reducing the risk of TB transmission is fundamental, while higher rates of relapse suggest that TB control should focus on improving the efficacy of the first-episode treatment regimen (Folkvardsen et al., 2020; Du et al., 2021). Numerous studies have shown that the proportion of TB recurrence due to exogenous reinfection varies by regions (Bandera et al., 2001; Verver et al., 2005; Zong et al., 2018; Liu et al., 2022). It is generally accepted that the proportion of reinfection in TB recurrence is higher in settings with a high prevalence of TB (Vega et al., 2021; Li et al., 2022), but there are exceptions (Shamputa et al., 2007). Studies of countries with low to moderate TB incidence found that the percentage of reinfection ranging from $10\%$ in Switzerland to $33\%$ in Spain (Schiroli et al., 2015), while reinfection was common in studies of high-burden countries, ranging from $23\%$ in India to 68–$77\%$ in South Africa (Sahadevan et al., 1995; van Rie et al., 1999; Charalambous et al., 2008). Our study found that $25\%$ of TB recurrence were attributed to reinfection, which was comparable to the proportion reported in Jiangsu ($28.9\%$) (Liu et al., 2022), but much higher than that reported in Beijing ($8.8\%$) (Du et al., 2021). Several reasons could be responsible for such variation of the percentage of reinfection. Firstly, the varied duration of follow-up would potentially affect the proportion of recurrence due to reactivation and reinfection (Liu et al., 2022). *In* general, relapse occurs earlier than reinfection, and if cases were followed up for an insufficient period, reinfections would not be captured (Vega et al., 2021), leading to a relatively lower proportion of reinfection. Secondly, different genomic-based typing methods, such as MIRU-VNTR, IS6110 fingerprinting, and whole genome sequencing, have different discriminatory power that can make a difference in the classification of TB recurrence (Shao et al., 2021). In addition, some of the patients’ complications could increase the risk from infection to disease, resulting in more reinfection cases (Lieberman et al., 2016). Moreover, transmission dynamics were also analyzed in our study and community transmission was not observed among these recurrent TB cases, which might due to transmission occurring in a broader population that was not included in our study population.
Mixed infections can complicate TB diagnosis and treatment, and it is also one of the potential confounders in distinguishing relapse from reinfection (Witney et al., 2017). To reduce the misclassification of recurrent TB cases, detection of mixed infection based on whole genome sequencing before determining the main source of TB recurrence is very essential. By using QuantTB, a method for identifying and quantifying individual MTB strains at high resolution (Anyansi et al., 2020), 7 of 79 ($8.9\%$) isolates in this study were identified as mixed infections. Although the sampling and culture methods used in this study may lower the diversity of strains (Liu et al., 2020), a relatively high proportion of mixed infections were still detected, which warned of the urgent need for further studies to determine the prevalence of mixed infections in different settings and its impact on heterogeneous drug-resistance.
Of note, two patients’ pair (patient 17 and patient 18) of isolates in this study displayed 14 SNPs (SNP > 12) between two episodes and were therefore initially classified as reinfection. However, further analysis showed that the strain pairs were located next to each other on the phylogenetic tree and shared the same drug-resistant profile (Figure 2), suggesting that these two recurrent cases were likely caused by relapse. This would leave 7 recurrent cases with paired isolates differing by more than 180 SNPs that were clearly identified as the result of reinfection, indicating that $80.6\%$ of TB recurrences were caused by relapse. The data here was supported by the findings of Walker and colleagues that the diversity between the initial and later isolates from relapsed patients does not generally exceed 14 SNPs, with most cases differing by less than five (Walker et al., 2013). Based on these results, it is reasonable to assume that strains with SNP differences slightly exceeding the thresholds (commonly 6 or 12 SNPs) used to define a cluster may occasionally belong to the same transmission chain and should be taken into account during the epidemiological investigation (Liu et al., 2020). More importantly, similar to previous studies, our study only found reinfections with large phylogenetic distances (range 185–1,074), but nothing at an intermediary level (Witney et al., 2017). This suggests that primary infection does not provide sufficient immune protection against genetically distant strains, which has important implications for future vaccine design (Bryant et al., 2013).
The emergence of drug resistance in relapsed TB weakens the effectiveness of subsequent treatment. In the present study, we found that the acquisition of resistance to fluoroquinolones was the most common during treatment, and this observation was further rationalized by genotypic resistance prediction based on whole-genome sequencing. Similar results have been reported elsewhere (Zong et al., 2018; Du et al., 2021). Although the exact cause of this phenomenon remains unclear, it can be partially explained by the abuse and misuse of fluoroquinolones. In China, because of their broad-spectrum antimicrobial activity, fluoroquinolones are always used as empirical treatment for suspected TB patients and various other types of infections (Du et al., 2021). Consequently, the selection pressure on MTB generated by residual drugs in the host allows the survival and accumulation of drug-resistant strains, resulting in strains with drug-resistance becoming the dominant population. Consistent with our findings, numerous previous studies have confirmed significantly increased prevalence of fluoroquinolones resistance in recent years in China (Xia et al., 2021; Mave et al., 2022). In addition, experimental data showed that fluoroquinolones activate the SOS response, which is likely to be associated with an elevated mutation rate. This may be another important factor contributing to the high frequency of fluoroquinolones resistance (Iacobino et al., 2021).
TB relapse was determined by a wide range of factors, such as socio-demographic and clinical features of TB cases, drug resistance and genetic background of the bacteria, and the disease burden of the study settings (Romanowski et al., 2019). Previous studies have shown that patients infected with Beijing genotype or isoniazid resistant strains were more susceptible to relapse (Hang et al., 2015; Thai et al., 2018). Besides, Romanowski et al. already found that despite poor predictive ability, cavitary disease and 2-month smear positivity could be used as markers for higher risk of relapse (Romanowski et al., 2019). However, in our current study, the relatively small sample size of recurrent TB cases limits our ability to detect significant difference between relapse and reinfection. To make follow-up for TB relapse more practical, future studies could identify socio-environmental and bio-medical factors associated with relapse by using modeling studies or genome-wide association analysis (GWAS), so these can be addressed or guide care after cure. Understanding the time interval distribution of recurrence is important for developing post-treatment control strategies and designing clinical trial studies (Marx et al., 2014). A meta-analysis reported that relapse occurred mainly in the first year after the end of treatment, while late recurrences tended to be reinfections (Romanowski et al., 2019). However, in our study, there was no significant difference in the time interval between relapse and reinfection. TB recurrences, whether caused by relapse or reinfection, occur predominantly within 3 years after completion of therapy. Therefore, for better management of TB patients in this region, we recommend that patients should be followed-up for at least 3 years after completion of therapy. Moreover, we further assessed the time interval to relapse stratified by gender, nationality, pulmonary cavity, et al. Despite the small sample size, a correlation was observed in the present study between Tu nationality and earlier relapse. Further study with an expanded sample size is needed to explore whether there is a genuine correlation between nationality and time interval to relapse.
A major strength of this study is that we conducted a retrospective study of sufficient duration by using whole-genome sequencing data of serial strains from recurrent TB patients, which allowed us to get a more accurate picture of the proportion of recurrence caused by reinfection after excluding mixed infections, as well as to understand the drug resistance acquired during treatment. We must acknowledge several limitations of this study. First, this study was based on routinely collected information and specimens. Some TB recurrent cases might be lost due to death or moving out of the region, which would reduce the accuracy of our results. Second, recurrent TB cases who were excluded from the final analysis due to subculture failure and contamination of any paired isolates may introduce selection bias into this study. Third, the relatively small sample size of drug-resistant TB strains restricted us from exploring the underlying mechanism of acquired drug resistance during treatment. Lastly, the HIV status of most recurrent TB cases in this study is unknown, but given the low prevalence of HIV in this area, we believe this is unlikely to introduce bias to the results of our study.
In conclusion, our data demonstrate that endogenous relapse is the main mechanism leading to TB recurrences in Hunan province. Additionally, our study found TB recurrence can occur even more than 4 years after treatment completion of the most recent episode, mainly within 3 years. Therefore, it is necessary to extend the post-treatment follow-up period to achieve better management of TB patients. Moreover, the relatively high frequency of fluoroquinolone-resistance in the second episode of relapse suggests that fluoroquinolones should be used with caution when treating TB cases with relapse, preferably guided by DST results.
## 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
Ethical approval was not provided for this study on human participants because National TB drug-resistant surveillance (DRS) was ethically approved by the Ethics Committee of Chinese Center for Disease Control and Prevention since the first national survey in 2007 (Zhao et al., 2012). Ethics approval of the present study was waived because all TB isolates used in this study were obtained from previous DRS routine work, and patient information was extracted from the previous database, no additional data and specimens were collected. Patients/participants provided their written informed consent at the time of their first visit to the designated TB clinics or centers.
## Author contributions
WH and YZ contributed to study design, data analysis, and manuscript writing. YT, ZS, BL, CL, HZ, DL, SP, and FH participated in study design, data collection, and analysis. YW, PH, AM, XC, and BZ conducted laboratory testing. HX, SW, and XO revised and polished the manuscript. All the authors have read the final version of the manuscript and have approved it.
## Funding
This work was supported by the National Key R&D Program (No. 2022YFC2305200) and Natural Science Foundation of Xinjiang Uygur Autonomous Region (No. 2022D01A115).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1115295/full#supplementary-material
## References
1. Anyansi C., Keo A., Walker B. J., Straub T. J., Manson A. L., Earl A. M.. **QuantTB – a method to classify mixed**. *BMC Genomics* (2020) **21** 80. DOI: 10.1186/s12864-020-6486-3
2. Bandera A., Gori A., Catozzi L., Degli Esposti A., Marchetti G., Molteni C.. **Molecular epidemiology study of exogenous reinfection in an area with a low incidence of tuberculosis**. *J. Clin. Microbiol.* (2001) **39** 2213-2218. DOI: 10.1128/jcm.39.6.2213-2218.2001
3. Barbier M., Wirth T.. **The evolutionary history, demography, and spread of the**. *Microbiol. Spectr.* (2016) **4** 4. DOI: 10.1128/microbiolspec.TBTB2-0008-2016
4. Bolger A. M., Lohse M., Usadel B.. **Trimmomatic: a flexible trimmer for Illumina sequence data**. *Bioinformatics* (2014) **30** 2114-2120. DOI: 10.1093/bioinformatics/btu170
5. Bryant J. M., Harris S. R., Parkhill J., Dawson R., Diacon A. H., van Helden P.. **Whole-genome sequencing to establish relapse or re-infection with mycobacterium tuberculosis: a retrospective observational study**. *Lancet Respir. Med.* (2013) **1** 786-792. DOI: 10.1016/s2213-2600(13)70231-5
6. Charalambous S., Grant A. D., Moloi V., Warren R., Day J. H., van Helden P.. **Contribution of reinfection to recurrent tuberculosis in south African gold miners**. *Int. J. Tuberc. Lung Dis.* (2008) **12** 942-948. PMID: 18647455
7. Du J., Li Q., Liu M., Wang Y., Xue Z., Huo F.. **Distinguishing relapse from reinfection with whole-genome sequencing in recurrent pulmonary tuberculosis: a retrospective cohort study in Beijing, China**. *Front. Virol.* (2021) **12** 754352. DOI: 10.3389/fmicb.2021.754352
8. Folkvardsen D. B., Norman A., Rasmussen E. M., Lillebaek T., Jelsbak L., Andersen Å.. **Recurrent tuberculosis in patients infected with the predominant mycobacterium tuberculosis outbreak strain in Denmark. New insights gained through whole genome sequencing**. *Infect. Genet. Evol.* (2020) **80** 104169. DOI: 10.1016/j.meegid.2020.104169
9. Freschi L., Vargas R., Husain A., Kamal S. M. M., Skrahina A., Tahseen S.. **Population structure, biogeography and transmissibility of**. *Nat. Commun.* (2021) **12** 6099. DOI: 10.1038/s41467-021-26248-1
10. Hang N. T., Maeda S., Keicho N., Thuong P. H., Endo H.. **Sublineages of**. *Tuberculosis* (2015) **95** 336-342. DOI: 10.1016/j.tube.2015.02.040
11. He W., Tan Y., Liu C., Wang Y., He P., Song Z.. **Drug-resistant characteristics, genetic diversity, and transmission dynamics of rifampicin-resistant mycobacterium tuberculosis in Hunan, China, revealed by whole-genome sequencing**. *Microbiol. Spectr.* (2022) **10** e0154321. DOI: 10.1128/spectrum.01543-21
12. Iacobino A., Piccaro G., Pardini M., Fattorini L., Giannoni F.. **Moxifloxacin activates the SOS response in**. *Microorganisms* (2021) **9** 255. DOI: 10.3390/microorganisms9020255
13. Jagielski T., Minias A., van Ingen J., Rastogi N., Brzostek A., Żaczek A.. **Methodological and clinical aspects of the molecular epidemiology of**. *Clin. Microbiol. Rev.* (2016) **29** 239-290. DOI: 10.1128/cmr.00055-15
14. Lambert M. L., Hasker E., Van Deun A., Roberfroid D., Boelaert M., Van der Stuyft P.. **Recurrence in tuberculosis: relapse or reinfection?**. *Lancet Infect. Dis.* (2003) **3** 282-287. DOI: 10.1016/s1473-3099(03)00607-8
15. Li M., Qiu Y., Guo M., Zhang S., Wang G., Wang Y.. **Investigation on the cause of recurrent tuberculosis in a rural area in China using whole-genome sequencing: a retrospective cohort study**. *Tuberculosis* (2022) **133** 102174. DOI: 10.1016/j.tube.2022.102174
16. Lieberman T. D., Wilson D., Misra R., Xiong L. L., Moodley P., Cohen T.. **Genomic diversity in autopsy samples reveals within-host dissemination of HIV-associated**. *Nat. Med.* (2016) **22** 1470-1474. DOI: 10.1038/nm.4205
17. Liu Q., Qiu B., Li G., Yang T., Tao B., Martinez L.. **Tuberculosis reinfection and relapse in eastern China: a prospective study using whole-genome sequencing**. *Clin. Microbiol. Infect.* (2022) **28** 1458-1464. DOI: 10.1016/j.cmi.2022.05.019
18. Liu Q., Wei J., Li Y., Wang M., Su J., Lu Y.. **Mycobacterium tuberculosis clinical isolates carry mutational signatures of host immune environments**. *Sci. Adv.* (2020) **6** eaba4901. DOI: 10.1126/sciadv.aba4901
19. Marx F. M., Dunbar R., Enarson D. A., Williams B. G., Warren R. M., van der Spuy G. D.. **The temporal dynamics of relapse and reinfection tuberculosis after successful treatment: a retrospective cohort study**. *Clin. Infect. Dis.* (2014) **58** 1676-1683. DOI: 10.1093/cid/ciu186
20. Mave V., Chen L., Ranganathan U. D., Kadam D., Vishwanathan V., Lokhande R.. **Whole genome sequencing assessing impact of diabetes mellitus on tuberculosis mutations and type of recurrence in India**. *Clin. Infect. Dis.* (2022) **75** 768-776. DOI: 10.1093/cid/ciab1067
21. Roetzer A., Diel R., Kohl T. A., Rückert C., Nübel U., Blom J.. **Whole genome sequencing versus traditional genotyping for investigation of a**. *PLoS Med.* (2013) **10** e1001387. DOI: 10.1371/journal.pmed.1001387
22. Romanowski K., Balshaw R. F., Benedetti A., Campbell J. R., Menzies D., Ahmad Khan F.. **Predicting tuberculosis relapse in patients treated with the standard 6-month regimen: an individual patient data meta-analysis**. *Thorax* (2019) **74** 291-297. DOI: 10.1136/thoraxjnl-2017-211120
23. Ruan Q. L., Yang Q. L., Sun F., Liu W., Shen Y. J., Wu J.. **Recurrent pulmonary tuberculosis after treatment success: a population-based retrospective study in China**. *Clin. Microbiol. Infect.* (2022) **28** 684-689. DOI: 10.1016/j.cmi.2021.09.022
24. Sahadevan R., Narayanan S., Paramasivan C. N., Prabhakar R., Narayanan P. R.. **Restriction fragment length polymorphism typing of clinical isolates of**. *J. Clin. Microbiol.* (1995) **33** 3037-3039. DOI: 10.1128/jcm.33.11.3037-3039.1995
25. Schiroli C., Carugati M., Zanini F., Bandera A., Di Nardo Stuppino S., Monge E.. **Exogenous reinfection of tuberculosis in a low-burden area**. *Infection* (2015) **43** 647-653. DOI: 10.1007/s15010-015-0759-9
26. Shamputa I. C., Van Deun A., Salim M. A., Hossain M. A., Fissette K., de Rijk P.. **Endogenous reactivation and true treatment failure as causes of recurrent tuberculosis in a high incidence setting with a low HIV infection**. *Tropical Med. Int. Health* (2007) **12** 700-708. DOI: 10.1111/j.1365-3156.2007.01840.x
27. Shao Y., Song H., Li G., Li Y., Li Y., Zhu L.. **Relapse or re-infection, the situation of recurrent tuberculosis in eastern China**. *Front. Cell. Infect. Microbiol.* (2021) **11** 638990. DOI: 10.3389/fcimb.2021.638990
28. Shen X., Yang C., Wu J., Lin S., Gao X., Wu Z.. **Recurrent tuberculosis in an urban area in China: relapse or exogenous reinfection?**. *Tuberculosis* (2017) **103** 97-104. DOI: 10.1016/j.tube.2017.01.007
29. Thai P. V. K., Ha D. T. M., Hanh N. T., Day J., Dunstan S., Nhu N. T. Q.. **Bacterial risk factors for treatment failure and relapse among patients with isoniazid resistant tuberculosis**. *BMC Infect. Dis.* (2018) **18** 112. DOI: 10.1186/s12879-018-3033-9
30. van Rie A., Warren R., Richardson M., Victor T. C., Gie R. P., Enarson D. A.. **Exogenous reinfection as a cause of recurrent tuberculosis after curative treatment**. *N. Engl. J. Med.* (1999) **341** 1174-1179. DOI: 10.1056/nejm199910143411602
31. Vega V., Rodríguez S., Van der Stuyft P., Seas C., Otero L.. **Recurrent TB: a systematic review and meta-analysis of the incidence rates and the proportions of relapses and reinfections**. *Thorax* (2021) **76** 494-502. DOI: 10.1136/thoraxjnl-2020-215449
32. Verver S., Warren R. M., Beyers N., Richardson M., van der Spuy G. D., Borgdorff M. W.. **Rate of reinfection tuberculosis after successful treatment is higher than rate of new tuberculosis**. *Am. J. Respir. Crit. Care Med.* (2005) **171** 1430-1435. DOI: 10.1164/rccm.200409-1200OC
33. Walker T. M., Ip C. L., Harrell R. H., Evans J. T., Kapatai G., Dedicoat M. J.. **Whole-genome sequencing to delineate**. *Lancet Infect. Dis.* (2013) **13** 137-146. DOI: 10.1016/s1473-3099(12)70277-3
34. Walker T. M., Kohl T. A., Omar S. V., Hedge J., Del Ojo Elias C., Bradley P.. **Whole-genome sequencing for prediction of**. *Lancet Infect. Dis.* (2015) **15** 1193-1202. DOI: 10.1016/s1473-3099(15)00062-6
35. Witney A. A., Bateson A. L., Jindani A., Phillips P. P., Coleman D., Stoker N. G.. **Use of whole-genome sequencing to distinguish relapse from reinfection in a completed tuberculosis clinical trial**. *BMC Med.* (2017) **15** 71. DOI: 10.1186/s12916-017-0834-4
36. World Health Organization (WHO) (2020). Global Tuberculosis Report 2020. Geneva: World Health Organization.. *Global Tuberculosis Report 2020* (2020)
37. World Health Organization (WHO) (2013). Definitions and reporting framework for tuberculosis–2013 revision: Updated December 2014 and January 2020. Geneva: World Health Organization.. *Definitions and reporting framework for tuberculosis–2013 revision: Updated December 2014 and January 2020* (2013)
38. Wu X., Yang J., Tan G., Liu H., Liu Y., Guo Y.. **Drug resistance characteristics of**. *Front. Cell. Infect. Microbiol.* (2019) **9** 345. DOI: 10.3389/fcimb.2019.00345
39. Xia H., Zheng Y., Liu D., Wang S., He W., Zhao B.. **Strong increase in moxifloxacin resistance rate among multidrug-resistant mycobacterium tuberculosis isolates in China, 2007 to 2013**. *Microbiol. Spectr.* (2021) **9** e0040921. DOI: 10.1128/Spectrum.00409-21
40. Xia H., Zheng Y., Zhao B., van den Hof S., Cobelens F., Zhao Y.. **Assessment of a 96-well plate assay of quantitative drug susceptibility testing for**. *PLoS One* (2017) **12** e0169413. DOI: 10.1371/journal.pone.0169413
41. Zhao Y., Xu S., Wang L., Chin D. P., Wang S., Jiang G.. **National survey of drug-resistant tuberculosis in China**. *N. Engl. J. Med.* (2012) **366** 2161-2170. DOI: 10.1056/NEJMoa1108789
42. Zong Z., Huo F., Shi J., Jing W., Ma Y., Liang Q.. **Relapse versus reinfection of recurrent tuberculosis patients in a national tuberculosis specialized hospital in Beijing, China**. *Front. Microbiol.* (2018) **9** 1858. DOI: 10.3389/fmicb.2018.01858
|
---
title: Thiazolidinediones lower the risk of pneumonia in patients with type 2 diabetes
authors:
- Fu-Shun Yen
- James Cheng-Chung Wei
- Yu-Tung Hung
- Chung Y. Hsu
- Chii-Min Hwu
- Chih-Cheng Hsu
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC9981669
doi: 10.3389/fmicb.2023.1118000
license: CC BY 4.0
---
# Thiazolidinediones lower the risk of pneumonia in patients with type 2 diabetes
## Abstract
### Introduction
We conducted this study to compare the risk of pneumonia between thiazolidinedione (TZD) use and nonuse in persons with type 2 diabetes (T2D).
### Methods
We identified 46,763 propensity-score matched TZD users and nonusers from Taiwan’s National Health Insurance Research Database between January 1, 2000, and December 31, 2017. The Cox proportional hazards models were used for comparing the risk of morbidity and mortality associated with pneumonias.
### Results
Compared with the nonuse of TZDs, the adjusted hazard ratios ($95\%$ CI) for TZD use in hospitalization for all-cause pneumonia, bacterial pneumonia, invasive mechanical ventilation, and death due to pneumonia were 0.92 (0.88–0.95), 0.95 (0.91–0.99), 0.80 (0.77–0.83), and 0.73 (0.64–0.82), respectively. The subgroup analysis revealed that pioglitazone, not rosiglitazone, was associated with a significantly lower risk of hospitalization for all-cause pneumonia [0.85 (0.82–0.89)]. Longer cumulative duration and higher cumulative dose of pioglitazone were associated with further lower adjusted hazard ratios in these outcomes compared to no-use of TZDs.
### Discussion
This cohort study demonstrated that TZD use was associated with significantly lower risks of hospitalization for pneumonia, invasive mechanical ventilation, and death due to pneumonia in patients with T2D. Higher cumulative duration and dose of pioglitazone were associated with a further lower risk of outcomes.
## Introduction
The Institute for Health Metrics and Evaluation showed that cases of lower respiratory tract infections worldwide increased from 414.3 million to 488.9 million between 1990 and 2019s (The Institute for Health Metrics and Evaluation (IHME), Global *Health data* exchange, GBD results tool, 2019). Due to the potential impact of accumulated hyperglycemia and oxidative stress, persons with diabetes showed reduced lung function and impaired neutrophil capability (Kornum et al., 2008; Gan, 2013). Studies have shown that persons with diabetes have a 1.2-to 2.6-fold higher risk of pneumonia than those without diabetes (Kornum et al., 2008; Harding et al., 2020). Recently, the incidence of macrovascular and microvascular complications in persons with type 2 diabetes (T2D) has decreased in many countries, possibly attributable to the aggressive control of blood pressure, lipids, and glucose levels. However, the occurrence of pneumonia is still on the rise (Wang et al., 2019; Pearson-Stuttard et al., 2022). The Taiwan Diabetes Atlas reported that the risks of hospitalization and mortality from pneumonia significantly increased from 2005 to 2014 in persons with T2D (Li et al., 2019; Wang et al., 2019). However, the diabetes guidelines for pneumonia management are limited (American Diabetes Association, 2021).
Peroxisome proliferator-activated receptors (PPARs) belong to a large superfamily of nuclear hormone receptors for retinoid, glucocorticoid, and thyroid hormones. PPARs are ligand-activated transcription factors crucial for regulating glucose homeostasis, adipocyte proliferation, atherosclerosis, cell cycle control, and inflammation (Zingarelli and Cook, 2005). Thiazolidinediones (TZDs) are the synthetic ligands of PPARs. Studies have demonstrated that in addition to improving insulin resistance, TZDs have anti-inflammatory and immunomodulatory properties. Preclinical studies have shown that TZDs can decrease neutrophil recruitment, downregulate inflammatory cytokines, and attenuate inflammation in acute lung injury (Zingarelli and Cook, 2005; Grommes et al., 2012). Thus, TZDs can influence the development or progression of pneumonia. One meta-analysis of 13 randomized clinical trials revealed that TZDs could moderately increase the risk of pneumonia in patients with T2D (Singh et al., 2011). However, pneumonias were adverse events of these trials, and most trials had low event rates. Without individual patient data, the pooled results may be worrying. Therefore, we conducted this nationwide cohort study to compare the risk of pneumonia between TZD users and nonusers to assess the impact of TZDs on pneumonia development or progression in persons with T2D.
## Study population
The Bureau of National Health Insurance implemented Taiwan’s National Health Insurance (NHI) program in 1995. The NHI program is a compulsory insurance system. The government and customers pay most of the premium, and the public only pays a small percentage. Approximately $99\%$ of Taiwan’s 23 million persons joined the NHI program in 2000 (Cheng, 2003). All personal information of the insured, including sex, age, area of residence, insurance premium, diagnoses, medical procedures, and prescriptions, are recorded in the NHI Research Database (NHIRD). The diagnosis was based on the International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-$\frac{9}{10}$-CM). The NHIRD linked to the National Death Registry to verify mortality information. This study was approved by the Research Ethics Committee of China Medical University and Hospital [CMUH110-REC1-038 (CR-1)]. The identifiable information of the participants and caregivers was scrambled and encrypted before release to protect individual privacy. Informed consent was waived by the Research Ethics Committee.
## Study design
We identified participants who were newly diagnosed with T2D between January 1, 2000, and December 31, 2017, and followed them until December 31, 2018. The diagnosis of T2D was based on ICD codings (ICD-9-CM codes: 250, except 250.1x; ICD-10-CM: E11) for at least 2 outpatient visits or one hospitalization. The algorithm for using ICD codes to define T2D was validated by a study in Taiwan with an accuracy of $74.6\%$ (Lin et al., 2005). Participants were excluded (Supplementary Figure 1) under the following conditions: (The Institute for Health Metrics and Evaluation (IHME), Global *Health data* exchange, GBD results tool, 2019) age, below 20 or above 80 years; (Gan, 2013) missing age or sex information; (Kornum et al., 2008) diagnosis of type 1 diabetes (Supplementary Table 1), heart failure, or hepatic failure; (Harding et al., 2020) diagnosis of T2D established before January 1, 2000, to exclude prevalent cases.
## Procedures
We defined the first date of TZD use as the index date. Participants who never received TZD treatment served as controls. We recorded the same period from the diagnosis of T2D to the use of TZDs as the index date for the control cases. Some related variables, checked and matched between TZD users and nonusers, were as follows: age (20–40, 41–60, 60–79 years), sex, obesity, smoking status; comorbidities, including alcohol-related disorders, hypertension, dyslipidemia, coronary artery disease (CAD), stroke, peripheral arterial occlusive disease (PAOD), chronic kidney disease (CKD), pneumonia, chronic obstructive pulmonary disease (COPD), liver cirrhosis, psychosis, depression, diagnosed within 1 year before the index date; medications, including oral antidiabetic drugs (OAD), insulin, statin, aspirin, corticosteroid, and immunosuppressants, used during the follow-up period. We calculated the Charlson Comorbidity Index (CCI), Diabetes Complication Severity Index (DCSI) score (Meduru et al., 2007; Young et al., 2008), and the number of oral antidiabetic drugs to evaluate the severity of T2D.
## Main outcomes
The observed main outcomes of this study were hospitalization for all-cause pneumonia, hospitalization for bacterial pneumonia, invasive mechanical ventilation (IMV) use, and death due to pneumonia. One study in Taiwan validated the algorithm of using ICD codes to define pneumonia, with a sensitivity of 92.3–$94.7\%$ (Su et al., 2014). We calculated the events, person-years, and incidence rates for these outcomes during the follow-up period. We compared the cumulative incidences of the main outcomes between TZD users and nonusers.
In the matched cohorts (Table 2), 7,753 ($16.57\%$) TZD users and 5,220 ($11.16\%$) nonusers were hospitalized for all-cause pneumonia during the follow-up time (incidence rate: 21.79 vs. 21.82 per 1,000 person-years). In the multivariable model, TZD users showed a significantly lower risk of hospitalization for all-cause pneumonia than nonusers (aHR = 0.92, $95\%$ CI = 0.88–0.95). Compared with nonusers, TZD users also showed significantly lower risks of hospitalization for bacterial pneumonia (aHR 0.95, $95\%$CI 0.91–0.99), IMV (aHR 0.80, $95\%$ CI 0.77–0.83), and death due to pneumonia (aHR 0.73, $95\%$ CI 0.64–0.82).
**Table 2**
| Outcomes | Non-TZD | Non-TZD.1 | Non-TZD.2 | TZD | TZD.1 | TZD.2 | cHR | (95% CI) | p-value | aHR a | (95% CI).1 | p-value.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Outcomes | N | PY | IR | N | PY | IR | cHR | (95% CI) | p-value | aHR a | (95% CI) | p-value |
| Hospitalization for all-cause pneumonia | | 239195 | 21.82 | 7753 | 355791 | 21.79 | 0.93 | (0.90, 0.97) | <0.001 | 0.92 | (0.88, 0.95) | <0.001 |
| Hospitalization for bacterial pneumonia | 3738 | 241514 | 15.48 | 5448 | 360389 | 15.12 | 0.96 | (0.92, 1.00) | 0.065 | 0.95 | (0.91, 0.99) | 0.014 |
| Invasive mechanical ventilation | 4405 | 244067 | 18.05 | 5468 | 367239 | 14.89 | 0.81 | (0.78, 0.85) | <0.001 | 0.80 | (0.77, 0.83) | <0.001 |
| Death due to pneumonia | 469 | 248600 | 1.89 | 638 | 376022 | 1.70 | 0.74 | (0.65, 0.83) | <0.001 | 0.73 | (0.64, 0.82) | <0.001 |
The Kaplan–*Meier analysis* showed that the cumulative incidences of hospitalization for all-cause pneumonia, IMV use, and death due to pneumonia were significantly lower in TZD users than nonusers (Log-rank test value of $p \leq 0.001$). However, the cumulative incidence of hospitalization for bacterial pneumonia was non-significantly lower in TZD users than in nonusers (Log-rank test value of $$p \leq 0.066$$) (Supplementary Figure 2).
## Statistical analysis
We used propensity-score matching to optimize the relevant covariates between TZD users and nonusers (D’Agostino, 1998). The propensity score for each participant was estimated using non-parsimonious multivariable logistic regression, with TZD use as the dependent variable. We included 35 clinically related covariates as independent variables (Table 1). The nearest-neighbor algorithm was adopted to construct matched pairs, assuming the standardized mean difference (SMD) value <0.1 to be a negligible difference between the study and comparison cohorts.
**Table 1**
| Variables | Before PSM | Before PSM.1 | Before PSM.2 | Before PSM.3 | Before PSM.4 | After PSM | After PSM.1 | After PSM.2 | After PSM.3 | After PSM.4 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Variables | Non-TZD | Non-TZD | TZD | TZD | p-value | Non-TZD | Non-TZD | TZD | TZD | SMD |
| Variables | (N = 159,103) | (N = 159,103) | (N = 52,147) | (N = 52,147) | (N = 46,763) | (N = 46,763) | (N = 46,763) | (N = 46,763) | | |
| Variables | n | % | n | % | n | % | n | % | | |
| Sex | | | | | 0.066 | | | | | 0.018 |
| Female | 76944 | 48.36 | 24977 | 47.90 | | 21458 | 45.89 | 21877 | 46.78 | |
| Male | 82159 | 51.64 | 27170 | 52.10 | | 25305 | 54.11 | 24886 | 53.22 | |
| Age | | | | | <0.001 | | | | | |
| 20–40 | 13712 | 8.62 | 3297 | 6.32 | | 2982 | 6.38 | 2980 | 6.37 | <0.001 |
| 41–60 | 70402 | 44.25 | 25677 | 49.24 | | 22407 | 47.92 | 22391 | 47.88 | 0.001 |
| 60–79 | 74989 | 47.13 | 23173 | 44.44 | | 21374 | 45.71 | 21392 | 45.75 | 0.001 |
| Mean, (SD)a | 58.75 | (12.4) | 58.54 | (11.33) | 0.0003 | 58.8 | (11.47) | 58.81 | (11.45) | <0.001 |
| Obesity | 5043 | 3.17 | 1551 | 2.97 | 0.026 | 1333 | 2.85 | 1368 | 2.93 | 0.004 |
| Smoking status | 4322 | 2.72 | 1133 | 2.17 | <0.001 | 983 | 2.10 | 1074 | 2.30 | 0.013 |
| Comorbidities | | | | | | | | | | |
| Alcohol-related disorders | 5711 | 3.59 | 1674 | 3.21 | <0.001 | 1443 | 3.09 | 1556 | 3.33 | 0.014 |
| Hypertension | 99983 | 62.84 | 36511 | 70.02 | <0.001 | 32518 | 69.54 | 32527 | 69.56 | 0.000 |
| Dyslipidemia | 102868 | 64.66 | 36847 | 70.66 | <0.001 | 32197 | 68.85 | 32589 | 69.69 | 0.018 |
| Coronary artery disease | 42226 | 26.54 | 15247 | 29.24 | <0.001 | 13194 | 28.21 | 13323 | 28.49 | 0.006 |
| Stroke | 25722 | 16.17 | 9101 | 17.45 | <0.001 | 7921 | 16.94 | 8170 | 17.47 | 0.014 |
| PAOD | 7208 | 4.53 | 2587 | 4.96 | <0.001 | 2081 | 4.45 | 2268 | 4.85 | 0.019 |
| Chronic kidney disease | 24415 | 15.35 | 8598 | 16.49 | <0.001 | 7151 | 15.29 | 7515 | 16.07 | 0.021 |
| Pneumonia | 11050 | 6.95 | 2891 | 5.54 | <0.001 | 2860 | 6.12 | 2572 | 5.50 | 0.026 |
| COPD | 28464 | 17.89 | 8322 | 15.96 | <0.001 | 7125 | 15.24 | 7473 | 15.98 | 0.021 |
| Liver cirrhosis | 11620 | 7.30 | 3705 | 7.10 | 0.129 | 3194 | 6.83 | 3394 | 7.26 | 0.017 |
| Psychosis | 15496 | 9.74 | 3947 | 7.57 | <0.001 | 3466 | 7.41 | 3658 | 7.82 | 0.015 |
| Depression | 10012 | 6.29 | 2449 | 4.70 | <0.001 | 2143 | 4.58 | 2293 | 4.90 | 0.015 |
| CCI | | | | | <0.001 | | | | | |
| 0 | 106918 | 67.20 | 30950 | 59.35 | | 28447 | 60.83 | 27970 | 59.81 | 0.021 |
| 1 | 22496 | 14.14 | 9778 | 18.75 | | 8356 | 17.87 | 8457 | 18.08 | 0.006 |
| ≥2 | 29689 | 18.66 | 11419 | 21.90 | | 9960 | 21.30 | 10336 | 22.10 | 0.020 |
| DCSI | | | | | <0.001 | | | | | |
| 0 | 53217 | 33.45 | 12882 | 24.70 | | 12738 | 27.24 | 12161 | 26.01 | 0.028 |
| 1 | 32721 | 20.57 | 10181 | 19.52 | | 9187 | 19.65 | 9221 | 19.72 | 0.002 |
| ≥2 | 73165 | 45.99 | 29084 | 55.77 | | 24838 | 53.11 | 25381 | 54.28 | 0.023 |
| Medication | | | | | | | | | | |
| Metformin | 80512 | 50.60 | 46852 | 89.85 | <0.001 | 41426 | 88.59 | 41468 | 88.68 | 0.003 |
| Sulfonylurea | 65753 | 41.33 | 46520 | 89.21 | <0.001 | 41136 | 87.97 | 41136 | 87.97 | <0.001 |
| DPP-4 inhibitor | 13848 | 8.70 | 8191 | 15.71 | <0.001 | 7311 | 15.63 | 7305 | 15.62 | <0.001 |
| AGI | 12992 | 8.17 | 15344 | 29.42 | <0.001 | 10061 | 21.51 | 10097 | 21.59 | 0.002 |
| SGLT2 inhibitors | 473 | 0.30 | 311 | 0.60 | <0.001 | 275 | 0.59 | 275 | 0.59 | <0.001 |
| OAD numbers | | | | | <0.001 | | | | | |
| 0–1 | 98921 | 62.17 | 6790 | 13.02 | | 7081 | 15.14 | 6790 | 14.52 | 0.018 |
| 2–3 | 56985 | 35.82 | 41848 | 80.25 | | 36634 | 78.34 | 37354 | 79.88 | 0.038 |
| >3 | 3197 | 2.01 | 3509 | 6.73 | | 3048 | 6.52 | 2619 | 5.60 | 0.038 |
| Insulin | 66868 | 42.03 | 25883 | 49.63 | <0.001 | 22240 | 47.56 | 22674 | 48.49 | 0.019 |
| Statin | 66777 | 41.97 | 27411 | 52.56 | <0.001 | 23951 | 51.22 | 24196 | 51.74 | 0.010 |
| Aspirin | 73517 | 46.21 | 28309 | 54.29 | <0.001 | 24474 | 52.34 | 24889 | 53.22 | 0.018 |
| Corticosteroid | 36234 | 22.77 | 10480 | 20.10 | <0.001 | 9632 | 20.60 | 9735 | 20.82 | 0.005 |
| Immunosuppressants | 474 | 0.30 | 109 | 0.21 | <0.001 | 111 | 0.24 | 105 | 0.22 | 0.003 |
We used crude and multivariable-adjusted Cox proportional hazards models to compare outcomes between TZD users and nonusers. The results were presented as hazard ratios (HRs) and $95\%$ confidence intervals (CIs) for TZD users compared with nonusers. This study is based on the intention-to-treat hypothesis. To calculate the observed risks, we censored the participants until the date of respective outcomes, death, or at the end of follow-up on December 31, 2018, whichever came first. The Kaplan–Meier method and log-rank tests were used to compare the cumulative incidences of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia during the follow-up time between TZD users and nonusers. We compared the risk of hospitalization for all-cause pneumonia among different subgroups of age, sex, comorbidities, medications (rosiglitazone, pioglitazone, and others) for clinical applicability of results. We also assessed the cumulative duration (<153, 153–549, ≧550 days) and dose (<2,940, 2,940–10,009, ≧10,110 mg) of pioglitazone for the risks of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia compared with no-use of TZDs to explore the dose relationship. We performed a stratified analysis to see the effect of TZD vs. non-TZD in the risk of all-cause pneumonia stratified by the subgroups of metformin use vs. no-use, SU use vs. no-use, DPP-4 inhibitor use-vs. no-use trying to determine whether other hypoglycemic agents have effect on pneumonia risk; stratified by patient’s resident areas of the Northern, Central, Southern, and Eastern Taiwan trying to see whether the different environmental exposures have different effect on pneumonia risk.
A two-tailed value of $p \leq 0.05$ was considered significant. SAS (version 9.4; SAS Institute, Cary, NC, United States) was used for statistical analysis.
## Participants
From January 1, 2000, to December 31, 2017, we identified 338,361 participants with newly diagnosed T2D. Of these, 52,147 were TZD users, and 159,103 were nonusers (Supplementary Figure 1). After excluding unsuitable participants, 1: 1 propensity-score matching was used to construct 46,763 pairs of TZD users and nonusers. In the matched cohorts (Table 1), $46.34\%$ of the participants were female; the mean (SD) age was 58.81 (11.46) years. The mean follow-up time for TZD users and nonusers was 7.80 (4.65) years and 5.21 (3.79) years, respectively.
## Subgroup analysis
We assessed the variables associated with the risk of hospitalization for all-cause pneumonia and found a significantly lower risk among participants using pioglitazone and statin. However, males, older age, participants with alcohol-related disorders, chronic kidney disease, COPD, depression, higher numbers of oral antidiabetic drugs, insulin, and aspirin use had a significantly higher risk of hospitalization for all-cause pneumonia (Table 3).
**Table 3**
| Variables | n | PY | IR | cHR | 95% CI | p-value | aHR a | 95% CI.1 | p-value.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| TZD nonusers | 5220 | 239195 | 21.82 | 1.0 | (Reference) | – | 1.0 | (Reference) | – |
| Rosiglitazone users | 4254 | 170791 | 24.91 | 1.02 | (0.97, 1.06) | 0.4463 | 1.01 | (0.96, 1.05) | 0.7614 |
| Pioglitazone users | 3499 | 185000 | 18.91 | 0.86 | (0.82, 0.9) | <0.001 | 0.85 | (0.82, 0.89) | <0.001 |
| Sex | . | . | . | | | – | | | |
| Female | 5591 | 289164 | 19.34 | 1.0 | (Reference) | – | 1.0 | (Reference) | – |
| Male | 7382 | 305823 | 24.14 | 1.27 | (1.22, 1.31) | <0.001 | 1.5 | (1.45, 1.55) | <0.001 |
| Age | . | . | . | | | – | | | |
| 20–40 | 322 | 40716 | 7.91 | 1.0 | (Reference) | – | 1.0 | (Reference) | – |
| 41–60 | 3830 | 307283 | 12.46 | 1.58 | (1.41, 1.77) | <0.001 | 1.49 | (1.33, 1.67) | <0.001 |
| 60–79 | 8821 | 246988 | 35.71 | 4.7 | (4.2, 5.25) | <0.001 | 3.98 | (3.55, 4.46) | <0.001 |
| Mean, (SD) | | | | 1.07 | (1.06,1.07) | <0.001 | 1.06 | (1.05,1.06) | <0.001 |
| Comorbidities | | | | | | | | | |
| Alcohol-related disorders | 427 | 14985 | 28.50 | 1.38 | (1.25, 1.52) | <0.001 | 1.32 | (1.19, 1.45) | <0.001 |
| Chronic kidney disease | 2891 | 78044 | 37.04 | 1.97 | (1.89, 2.05) | <0.001 | 1.5 | (1.43, 1.56) | <0.001 |
| COPD | 3062 | 86663 | 35.33 | 1.85 | (1.77, 1.92) | <0.001 | 1.33 | (1.27, 1.38) | <0.001 |
| Depression | 825 | 22424 | 36.79 | 1.81 | (1.68, 1.94) | <0.001 | 1.54 | (1.43, 1.65) | <0.001 |
| Medication | | | | | | | | | |
| OAD numbers | | | | | | | | | |
| 0–1 | 1590 | 98589 | 16.13 | 1.0 | (Reference) | – | 1.0 | (Reference) | – |
| 2–3 | 10976 | 477738 | 22.97 | 1.46 | (1.38, 1.54) | <0.001 | 1.3 | (1.24, 1.38) | <0.001 |
| >3 | 407 | 18660 | 21.81 | 1.58 | (1.41, 1.76) | <0.001 | 1.23 | (1.10, 1.37) | <0.001 |
| Insulin | 7432 | 261655 | 28.40 | 1.76 | (1.70, 1.82) | <0.001 | 1.49 | (1.44, 1.55) | <0.001 |
| Statin | 5574 | 274591 | 20.30 | 0.92 | (0.89, 0.95) | <0.001 | 0.77 | (0.75, 0.80) | <0.001 |
| Aspirin | 8482 | 297047 | 28.55 | 1.95 | (1.88, 2.03) | <0.001 | 1.36 | (1.31, 1.41) | <0.001 |
## Cumulative duration and dose of pioglitazone
We investigated the association between the cumulative duration of pioglitazone use and the risks of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia (Table 4). A longer cumulative duration of pioglitazone use was associated with further lower risks of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia compared with no-use of TZDs. The value of ps for the trend were all significant (Table 4).
**Table 4**
| Variables | Event | PY | IR | cHR | (95% CI) | p value | aHR a | (95% CI).1 | p-value |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | | | | |
| No-use of TZDs | 5220 | 239195 | 21.82 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative days of pioglitazone use (days) | | | | | | | | | |
| <153 | 1417 | 53467 | 26.50 | 1.22 | (1.15, 1.3) | <0.001 | 1.18 | (1.11, 1.25) | <0.001 |
| 153–549 | 1228 | 60978 | 20.14 | 0.93 | (0.87, 0.99) | 0.02 | 0.96 | (0.90, 1.02) | 0.19 |
| ≧550 | 854 | 70555 | 12.10 | 0.54 | (0.5, 0.58) | <0.001 | 0.56 | (0.52, 0.60) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
| Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | | | | |
| no-use of TZDs | 3738 | 241514 | 15.48 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative days of pioglitazone use (days) | | | | | | | | | |
| <153 | 913 | 54273 | 16.82 | 1.09 | (1.01, 1.17) | 0.027 | 1.09 | (1.02, 1.17) | 0.02 |
| 153–549 | 766 | 61815 | 12.40 | 0.8 | (0.74, 0.86) | <0.001 | 0.88 | (0.81, 0.95) | 0.001 |
| ≧550 | 502 | 71142 | 7.06 | 0.45 | (0.41, 0.5) | <0.001 | 0.49 | (0.44, 0.54) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
| Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | | | | |
| No-use of TZDs | 4405 | 244067 | 18.05 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative days of pioglitazone use (days) | | | | | | | | | |
| <153 | 987 | 55155 | 17.90 | 0.99 | (0.93, 1.07) | 0.88 | 1.01 | (0.93, 1.07) | 0.98 |
| 153–549 | 749 | 62609 | 11.96 | 0.67 | (0.62, 0.72) | <0.001 | 0.71 | (0.66, 0.77) | <0.001 |
| ≧550 | 463 | 71979 | 6.43 | 0.36 | (0.32, 0.39) | <0.001 | 0.38 | (0.35, 0.42) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
| Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | | | | |
| No-use of TZDs | 469 | 248600 | 1.89 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative days of pioglitazone use (days) | | | | | | | | | |
| <153 | 113 | 56537 | 2.00 | 1.09 | (0.89, 1.34) | 0.41 | 1.06 | (0.86, 1.31) | 0.57 |
| 153–550 | 90 | 63878 | 1.41 | 0.76 | (0.61, 0.96) | 0.02 | 0.83 | (0.66, 1.04) | 0.11 |
| >550 | 59 | 72706 | 0.81 | 0.4 | (0.3, 0.52) | <0.001 | 0.44 | (0.33, 0.57) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
We also observed an association between the cumulative dose of pioglitazone and the risks of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia (Table 4). The higher cumulative dose of pioglitazone use was associated with further lower risks of hospitalization for all-cause pneumonia, bacterial pneumonia, IMV, and death due to pneumonia compared with no-use of TZDs; the value of ps for the trend were all significant (Table 5).
**Table 5**
| Variables | Event | PY | IR | cHR | (95% CI) | p value | aHR a | (95% CI).1 | p-value |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | Hospitalization for all-cause pneumonia | | | | |
| No-use of TZDs | 5220 | 239195 | 21.82 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative dose of pioglitazone (mg) | Cumulative dose of pioglitazone (mg) | | | | | | | | |
| <2,940 | 1627 | 50547 | 32.19 | 1.49 | (1.41, 1.58) | <0.001 | 1.39 | (1.31, 1.47) | <0.001 |
| 2,940–10,009 | 1091 | 54383 | 20.06 | 0.93 | (0.87, 0.99) | 0.024 | 0.94 | (0.88, 1) | 0.07 |
| ≧10,110 | 781 | 80071 | 9.75 | 0.43 | (0.4, 0.47) | <0.001 | 0.46 | (0.43, 0.5) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
| Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | Hospitalization for bacterial pneumonia | | | | |
| No-use of tzds | 3738 | 241514 | 15.48 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative dose of pioglitazone (mg) | Cumulative dose of pioglitazone (mg) | | | | | | | | |
| <2,940 | 1057 | 51329 | 20.59 | 1.33 | (1.24, 1.42) | <0.001 | 1.3 | (1.21, 1.39) | <0.001 |
| 2,940–10,009 | 671 | 55094 | 12.18 | 0.79 | (0.72, 0.85) | <0.001 | 0.85 | (0.78, 0.92) | <0.001 |
| ≧10,110 | 453 | 80808 | 5.61 | 0.36 | (0.32, 0.39) | <0.001 | 0.4 | (0.36, 0.44) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
| Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | Invasive mechanical ventilation | | | | |
| No-use of TZDs | 4405 | 244067 | 18.05 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative dose of pioglitazone (mg) | Cumulative dose of pioglitazone (mg) | | | | | | | | |
| <2,940 | 1050 | 52109 | 20.15 | 1.12 | (1.05, 1.2) | <0.001 | 1.1 | (1.02, 1.17) | 0.009 |
| 2,940–10,009 | 654 | 55845 | 11.71 | 0.65 | (0.6, 0.71) | <0.001 | 0.69 | (0.64, 0.75) | <0.001 |
| ≧10,110 | 495 | 81789 | 6.05 | 0.34 | (0.31, 0.37) | <0.001 | 0.37 | (0.34, 0.4) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
| Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | Death due to pneumonia | | | | |
| No-use of TZDs | 469 | 248600 | 1.89 | 1 | (Reference) | – | 1.0 | (Reference) | – |
| Cumulative dose of pioglitazone (mg) | Cumulative dose of pioglitazone (mg) | | | | | | | | |
| <2,940 | 114 | 53431 | 2.13 | 1.18 | (0.96, 1.44) | 0.12 | 1.1 | (0.89, 1.35) | 0.37 |
| 2,940–10,009 | 93 | 56936 | 1.63 | 0.89 | (0.71, 1.11) | 0.30 | 0.94 | (0.75, 1.17) | 0.58 |
| ≧10,110 | 55 | 82752 | 0.66 | 0.33 | (0.25, 0.43) | <0.001 | 0.38 | (0.28, 0.5) | <0.001 |
| P for trend | | | | | | | | | <0.001 |
## Additional analyses
The stratified analysis of different 4 regions of Taiwan on all-cause pneumonia risk between TZD users vs. nonusers seems to be consistent (Supplementary Table 2). But among other hypoglycemic agents, combined use of DPP-4 inhibitor and TZD seems to have a higher risk of hospitalization for all-cause pneumonia (Supplementary Table 2).
## Discussion
This study demonstrated that TZD use was associated with significantly lower risks of hospitalization for pneumonia, IMV, and death due to pneumonia than TZD no-use in persons with T2D. The subgroup analysis revealed that the reduced risk of pneumonia by TZDs could be due to pioglitazone use. The dose–response analysis showed that longer cumulative duration and a higher cumulative dose of pioglitazone were associated with further lower risks of these outcomes.
Studies have shown that patients with diabetes have a higher risk of pneumonia than those without diabetes (Kornum et al., 2008; Harding et al., 2020). Patients with suboptimal glycemic control showed a higher risk of pneumonia (Kornum et al., 2008). Our study also revealed that older people, persons with comorbidities, using more oral antidiabetic drugs, insulin, and aspirin had a higher risk of pneumonia. However, patients using TZDs, especially pioglitazone and statin, had a lower risk of hospitalization for all-cause pneumonia. Reports show that statin use is associated with a lower risk of pneumonia due to its potential anti-inflammatory effect (Macedo et al., 2014). However, a systemic review and meta-analysis by Sigh et al. revealed that TZD use was associated with a modestly elevated risk of pneumonia [relative risk (RR) 1.4(1.08–1.82)] (Singh et al., 2011). Gorricho et al. conducted a nested case–control study comparing the use of oral antidiabetic drugs and the risk of community-acquired pneumonia. They showed that TZDs combined with other antidiabetic drugs were associated with an increased risk of pneumonia compared to metformin plus sulfonylureas (Gorricho et al., 2017). The different results obtained from the three studies could be due to differences in the methodology and the study population. Moreover, Shih et al. conducted a case–control study and showed that TZD use was associated with a modest reduction of sepsis risk compared to TZD no-use in persons with T2D (Shih et al., 2015). To our knowledge, our research is the first study designed to compare the risk of pneumonia between TZD users and nonusers and suggest that TZD use may attenuate the risk of hospitalization for all-cause [aHR 0.92 (0.88, 0.95)] and bacterial pneumonia [aHR 0.95 (0.91, 0.99)]. This result may not be affected by the different resident environment of Taiwan. Bu if the patient is combined use of DPP-4 inhibitor and TZD seems to make the TZD lose their protective effect against all-cause pneumonia for reasons that are unclear. This study also showed that pioglitazone, not rosiglitazone, could reduce the risk of pneumonia. Rosiglitazone is a PPARγ agonist, but pioglitazone has both α and γ effects. Each TZD has different patterns of effects on the regulation of gene transcription (Kung and Henry, 2012). Previous studies have shown that the impact of pioglitazone and rosiglitazone on cardiovascular diseases was different (Kung and Henry, 2012). Preclinical studies have also found that the effect of pioglitazone and rosiglitazone on inflammation may be dissimilar (Zingarelli and Cook, 2005; Singh et al., 2011). More research is needed to determine any difference in the effectiveness of pioglitazone and rosiglitazone in the risk of pneumonia.
Diabetes may reduce lung function and pulmonary diffusion capacity due to microangiopathic changes in the lungs (Pitocco et al., 2012). Animal studies have demonstrated that pioglitazone can attenuate endotoxin-induced acute lung injury and pulmonary edema (Grommes et al., 2012). Kim et al. have shown that insulin sensitizers (metformin or TZDs) were independently associated with improvements in forced vital capacity (FVC) in persons with T2D and COPD (Kim et al., 2010). Our study demonstrated that TZDs were significantly associated with a lower risk of invasive mechanical ventilation than non-TZDs in persons with T2D [aHR 0.80 (0.77, 0.83)]. More studies are needed to explore the effect of TZDs on respiratory function, pulmonary microangiopathy, and inflammation.
Although the availability of excellent antibiotics has resulted in a significant reduction in mortality from pneumonia, the reduction in mortality within 7 days of the onset of pneumonia is not prominent. This finding may be due to the inability of antibiotics to rapidly reduce inflammatory events in the lungs (Corrales-Medina and Musher, 2011). Pneumonia may also be an important factor in accelerating premature death in persons with T2D and multimorbidity (Fine et al., 1996; Li et al., 2019; Pearson-Stuttard et al., 2022). Notably, this study showed that TZDs were significantly associated with a lower risk of death due to pneumonia [aHR 0.73 (0.64, 0.82)]. This finding may be attributable to the reduced risk of hospitalization for pneumonia and IMV support by TZD use. This study also showed that TZDs were more effective in protecting against hospitalization for all-cause pneumonia, IMV use, and death due to pneumonia than against hospitalization for bacterial pneumonia (Table 2), which may indicate that the anti-inflammatory effect of TZDs on protection against pneumonia may be greater than their antibacterial effect.
The possible grounds for TZDs to decrease the development and progression of pneumonia in persons with T2D are as follows: (The Institute for Health Metrics and Evaluation (IHME), Global *Health data* exchange, GBD results tool, 2019) pharmacological activation of PPARγ by TZDs can inhibit proinflammatory gene expression and reduce the production of C-reactive protein (CRP), tumor necrosis factor (TNF)-α, interleukin (IL)- 1β, IL-6, inducible nitric oxide synthase (iNOS), inducible cyclooxygenase (COX)-2, matrix metalloproteinase (MMP)-9, macrophage chemoattractant protein (MCP)-1, and plasminogen activator inhibitor (PAI)-1 (Zingarelli and Cook, 2005; Hanefeld et al., 2007; Gan, 2013). The induction of heat shock proteins by PPARγ ligands may alter the activation of nuclear factor-kB (NF-kB) and regulate inflammation (Zingarelli and Cook, 2005; Kornum et al., 2008). TZDs may augment CD36 expression, tether apoptotic cells to macrophages to promote efferocytosis, and evoke an anti-inflammatory response with the resolution of tissue injury (Zingarelli and Cook, 2005; Lea et al., 2014; Harding et al., 2020). Studies also showed that TZDs could increase IL-10 levels, enhance neutrophil recruitment to the infection foci, raise fibroblast growth factor (FGF) 21 levels, and improve survival in animals with sepsis (Trevelin et al., 2017; Pearson-Stuttard et al., 2022). PPARγ may play a role in the differentiation of naive T cells to effector T cells and improve adaptive immunity (Daynes and Jones, 2002; Wang et al., 2019). Preclinical studies demonstrated that TZDs could have direct antibacterial activity (Stegenga et al., 2009; Masadeh et al., 2011; Li et al., 2019). In animal models of lung injury, TZDs decreased pulmonary edema, fibrosis, inflammation, and mortality (Belvisi and Mitchell, 2009; Grommes et al., 2012).
This study has some limitations. First, this NHI dataset lacks complete information on dietary patterns, smoking habits, alcohol drinking, nutritional state, physical activity, vaccination status, and family history. It does not contain data on glucose, hemoglobin A1C, biochemical and microbiological tests, immune condition, pulmonary function tests, and imaging studies, which prevents a better understanding of the patient’s health status and the severity of diabetes. However, we matched the demographic information on sex and age to achieve a balance between the study and control groups. We also matched the items and number of oral antidiabetic drugs, insulin use, and DCSI scores to balance the severity of diabetes and increase the comparability between the study and comparison groups. Second, we had information on prescriptions, but patient compliance with medications was unknown. We could not obtain clues to the doctor’s preference for prescribing and the patient’s choice of medications from this database. Third, almost all the participants in this study were Chinese, and hence, the results may not be generalizable to other races. Fourth, cohort studies are usually associated with few unknown or unobserved confounding factors; therefore, randomized controlled trials are needed to verify our results. Finally, because TZDs have the concern of heart failure risk, if we want to use TZDs to reduce the risk of hospitalization for pneumonia, we must pay close attention to patients for signs and symptoms of heart failure to avoid unexpected harm to patients.
Persons with diabetes are more likely to contract and die from pneumonia than those without diabetes. Although the incidence of vascular complications has decreased, the occurrence of pneumonia is still rising. In addition to recommending persons with diabetes to receive influenza and pneumococcal vaccinations, perhaps TZD use may be an option to attenuate the morbidity and mortality of pneumonia. Additional studies are warranted to clarify all the effects of TZDs potentially linked to pneumonia.
## Data availability statement
The data analyzed in this study is subject to the following licenses/restrictions: Data of this study are available from the National Health Insurance Research Database (NHIRD) published by Taiwan National Health Insurance (NHI) Administration. The data utilized in this study cannot be made available in the paper, the supplemental files, or in a public repository due to the “Personal Information Protection Act” executed by Taiwan government starting from 2012. Requests for data can be sent as a formal proposal to the NHIRD Office (https://dep.mohw.gov.tw/DOS/cp-2516-3591-113.html) or by email to [email protected]. Requests to access these datasets should be directed to [email protected]
## Ethics statement
The studies involving human participants were reviewed and approved by Research Ethics Committee of China Medical University and Hospital. The ethics committee waived the requirement of written informed consent for participation.
## Author contributions
F-SY: study concept, design, and drafting the manuscript, critical revision of the manuscript for important intellectual content, and study supervision. JW: critical revision of the manuscript for important intellectual content, technical and material support, and study supervision. Y-TH: data acquisition, analysis, interpretation, and statistical analysis. CH: data acquisition, interpretation, funding acquisition, technical or material support, and critical revision of the manuscript for important intellectual content. C-MH: study concept and design, data acquisition, analysis, interpretation, drafting the manuscript, critical revision of the manuscript for important intellectual content, statistical analysis, funding acquisition, technical or material support, and study supervision. C-CH: analysis and interpretation of the data, drafting the manuscript, critical revision of the manuscript for important intellectual content, and study supervision. All authors contributed to the article and approved the submitted version.
## Funding
This study is supported in part by the Ministry of Science and Technology (MOHW109-TDU-B-212-114004) and China Medical University Hospital (DMR-111-105). This work received grant support from Taipei Veterans General Hospital (V101C-156, V108C-172, V109C-189). These funding agencies had no role in the study design, data collection and analysis, decision to publish, or manuscript preparation. No organization provided funds to assist with the preparation of this paper, and data analysis was not performed by employees of funders or any author who received funding. The funders had no role in study design, data collection, data analysis, the decision to publish, or manuscript preparation. This study received no additional external funding.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2022.1118000/full#supplementary-material
## References
1. **Comprehensive medical evaluation and assessment of comorbidities: standards of medical care in diabetes-2021**. *Diabetes Care* (2021) **44** S40-S52. DOI: 10.2337/dc21-S004
2. Belvisi M. G., Mitchell J. A.. **Targeting PPAR receptors in the airway for the treatment of inflammatory lung disease**. *Br. J. Pharmacol.* (2009) **158** 994-1003. DOI: 10.1111/j.14765381.2009.00373.x
3. Cheng T. M.. **Taiwan’s new national health insurance program: genesis and experience so far**. *Health Aff (Millwood)* (2003) **22** 61-76. DOI: 10.1377/hlthaff.22.3.61
4. Corrales-Medina V. F., Musher D. M.. **Immunomodulatory agents in the treatment of community-acquired pneumonia: a systematic review**. *J. Infect.* (2011) **63** 187-199. DOI: 10.1016/j.jinf.2011.06.009
5. D’Agostino R. B.. **Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group**. *Stat. Med.* (1998) **17** 2265-2281. DOI: 10.1002/(sici)1097-0258(19981015)17:19<2265::aid-sim918>3.0.co;2-b
6. Daynes R. A., Jones D. C.. **Emerging roles of PPARs in inflammation and immunity**. *Nat. Rev. Immunol.* (2002) **2** 748-759. DOI: 10.1038/nri912
7. Fine M. J., Smith M. A., Carson C. A., Mutha S. S., Sankey S. S., Weissfeld L. A.. **Prognosis and outcomes of patients with community-acquired pneumonia. A meta-analysis**. *JAMA* (1996) **275** 134-141. DOI: 10.1001/jama.275.2.134
8. Gan Y. H.. **Host susceptibility factors to bacterial infections in type 2 diabetes**. *PLoS Pathog.* (2013) **9** e1003794. DOI: 10.1371/journal.ppat.1003794
9. Gorricho J., Garjón J., Alonso A., Celaya M. C., Saiz L. C., Erviti J.. **Use of oral antidiabetic agents and risk of community-acquired pneumonia: a nested case-control study**. *Br. J. Clin. Pharmacol.* (2017) **83** 2034-2044. DOI: 10.1111/bcp.13288
10. Grommes J., Mörgelin M., Soehnlein O.. **Pioglitazone attenuates endotoxin-induced acute lung injury by reducing neutrophil recruitment**. *Eur. Respir. J.* (2012) **40** 416-423. DOI: 10.1183/09031936.00091011
11. Hanefeld M., Marx N., Pfützner A., Baurecht W., Lübben G., Karagiannis E.. **Anti-inflammatory effects of pioglitazone and/or simvastatin in high cardiovascular risk patients with elevated high sensitivity C-reactive protein: the PIOSTAT study**. *J. Am. Coll. Cardiol.* (2007) **49** 290-297. DOI: 10.1016/j.jacc.2006.08.054
12. Harding J. L., Benoit S. R., Gregg E. W., Pavkov M. E., Perreault L.. **Trends in rates of infections requiring hospitalisation among adults with versus without diabetes in the US, 2000-2015**. *Diabetes Care* (2020) **43** 106-116. DOI: 10.2337/dc19-0653
13. Kim H. J., Lee J. Y., Jung H. S., Kim D. K., Lee S. M., Yim J. J.. **The impact of insulin sensitisers on lung function in patients with chronic obstructive pulmonary disease and diabetes**. *Int. J. Tuberc. Lung Dis.* (2010) **14** 362-367. PMID: 20132629
14. Kornum J. B., Thomsen R. W., Riis A., Lervang H. H., Schønheyder H. C., Sørensen H. T.. **Diabetes, glycemic control, and risk of hospitalisation with pneumonia: a population-based case-control study**. *Diabetes Care* (2008) **31** 1541-1545. DOI: 10.2337/dc08-0138
15. Kung J., Henry R. R.. **Thiazolidinedione safety**. *Expert Opin. Drug Saf.* (2012) **11** 565-579. DOI: 10.1517/14740338.2012.691963
16. Lea S., Plumb J., Metcalfe H., Spicer D., Woodman P., Fox J. C.. **The effect of peroxisome proliferator-activated receptor-γ ligands on in vitro and in vivo models of COPD**. *Eur. Respir. J.* (2014) **43** 409-420. DOI: 10.1183/09031936.00187812
17. Li H. Y., Wu Y. L., Tu S. T., Hwu C. M., Liu J. S., Chuang L. M.. **Trends of mortality in diabetic patients in Taiwan: a nationwide survey in 2005-2014**. *J. Formos. Med. Assoc.* (2019) **118** S83-S89. DOI: 10.1016/j.jfma.2019.07.008
18. Lin C. C., Lai M. S., Syu C. Y., Chang S. Y., Teng F. Y.. **Accuracy of diabetes diagnosis in health insurance claims data in Taiwan**. *J. Formos. Med. Assoc.* (2005) **104** 157-163. PMID: 15818428
19. Macedo A. F., Taylor F. C., Casas J. P., Adler A., Prieto-Merino D., Ebrahim S.. **Unintended effects of statins from observational studies in the general population: systematic review and meta-analysis**. *BMC Med.* (2014) **12** 51. DOI: 10.1186/1741-7015-12-51
20. Masadeh M. M., Mhaidat N. M., Al-Azzam S. I., Alzoubi K. H.. **Investigation of the antibacterial activity of pioglitazone**. *Drug Des. Devel. Ther.* (2011) **5** 421-425. DOI: 10.2147/DDDT.S24126
21. Meduru P., Helmer D., Rajan M., Tseng C. L., Pogach L., Sambamoorthi U.. **Chronic illness with complexity: implications for performance measurement of optimal glycemic control**. *J. Gen. Intern. Med.* (2007) **22** 408-418. DOI: 10.1007/s11606-007-0310-5
22. Pearson-Stuttard J., Cheng Y. J., Bennett J., Vamos E. P., Zhou B., Valabhji J.. **Trends in leading causes of hospitalisation of adults with diabetes in England from 2003 to 2018: an epidemiological analysis of linked primary care records**. *Lancet Diabetes Endocrinol.* (2022) **10** 46-57. DOI: 10.1016/S2213-8587(21)00288-6
23. Pitocco D., Fuso L., Conte E. G., Zaccardi F., Condoluci C., Scavone G.. **The diabetic lung--a new target organ?**. *Rev. Diabet. Stud.* (2012) **9** 23-35. DOI: 10.1900/RDS.2012.9.23
24. Shih C. J., Wu Y. L., Chao P. W., Kuo S. C., Yang C. Y., Li S. Y.. **Association between use of oral anti-diabetic drugs and the risk of sepsis: a nested case-control study**. *Sci. Rep.* (2015) **5** 15260. DOI: 10.1038/srep15260
25. Singh S., Loke Y. K., Furberg C. D.. **Long-term use of thiazolidinediones and the associated risk of pneumonia or lower respiratory tract infection: systematic review and meta-analysis**. *Thorax* (2011) **66** 383-388. DOI: 10.1136/thx.2010.152777
26. Stegenga M. E., Florquin S., de Vos A. F., van der Poll T.. **The thiazolidinedione ciglitazone reduces bacterial outgrowth and early inflammation during Streptococcus pneumoniae pneumonia in mice**. *Crit. Care Med.* (2009) **37** 614-618. DOI: 10.1097/CCM.0b013e31819599b6
27. Su V. Y., Liu C. J., Wang H. K., Wu L. A., Chang S. C., Perng D. W.. **Sleep apnea and risk of pneumonia: a nationwide population-based study**. *CMAJ* (2014) **186** 415-421. DOI: 10.1503/cmaj.131547
28. The Institute for Health Metrics and Evaluation (IHME), Global Health data exchange, GBD results tool (2019). Available at: https://ghdx.healthdata.org/gbd-results-tool (accessed April 8, 2022).. (2019)
29. Trevelin S. C., Carlos D., Beretta M., da Silva J. S., Cunha F. Q.. **Diabetes mellitus and sepsis: a challenging association**. *Shock* (2017) **47** 276-287. DOI: 10.1097/SHK.0000000000000778
30. Wang J. S., Wu Y. L., Shin S. J., Tien K. J., Chin M. C., Hwu C. M.. **Hospitalization in patients with type 2 diabetes mellitus in Taiwan: a nationwide population-based observational study**. *J. Formos. Med. Assoc.* (2019) **118** S90-S95. DOI: 10.1016/j.jfma.2019.06.017
31. Young B. A., Lin E., Von Korff M., Simon G., Ciechanowski P., Ludman E. J.. **Diabetes complications severity index and risk of mortality, hospitalization, and health care utilization**. *Am. J. Manag. Care* (2008) **14** 15-23. PMID: 18197741
32. Zingarelli B., Cook J. A.. **Peroxisome proliferator-activated receptor-gamma is a new therapeutic target in sepsis and inflammation**. *Shock* (2005) **23** 393-399. DOI: 10.1097/01.shk.0000160521.91363.88
|
---
title: Effect of tissue-grouped regulatory variants associated to type 2 diabetes
in related secondary outcomes
authors:
- Daiane Hemerich
- Roelof A. J. Smit
- Michael Preuss
- Lauren Stalbow
- Sander W. van der Laan
- Folkert W. Asselbergs
- Jessica van Setten
- Vinicius Tragante
journal: Scientific Reports
year: 2023
pmcid: PMC9981672
doi: 10.1038/s41598-023-30369-6
license: CC BY 4.0
---
# Effect of tissue-grouped regulatory variants associated to type 2 diabetes in related secondary outcomes
## Abstract
Genome-wide association studies have identified over five hundred loci that contribute to variation in type 2 diabetes (T2D), an established risk factor for many diseases. However, the mechanisms and extent through which these loci contribute to subsequent outcomes remain elusive. We hypothesized that combinations of T2D-associated variants acting on tissue-specific regulatory elements might account for greater risk for tissue-specific outcomes, leading to diversity in T2D disease progression. We searched for T2D-associated variants acting on regulatory elements and expression quantitative trait loci (eQTLs) in nine tissues. We used T2D tissue-grouped variant sets as genetic instruments to conduct 2-Sample Mendelian Randomization (MR) in ten related outcomes whose risk is increased by T2D using the FinnGen cohort. We performed PheWAS analysis to investigate whether the T2D tissue-grouped variant sets had specific predicted disease signatures. We identified an average of 176 variants acting in nine tissues implicated in T2D, and an average of 30 variants acting on regulatory elements that are unique to the nine tissues of interest. In 2-Sample MR analyses, all subsets of regulatory variants acting in different tissues were associated with increased risk of the ten secondary outcomes studied on similar levels. No tissue-grouped variant set was associated with an outcome significantly more than other tissue-grouped variant sets. We did not identify different disease progression profiles based on tissue-specific regulatory and transcriptome information. Bigger sample sizes and other layers of regulatory information in critical tissues may help identify subsets of T2D variants that are implicated in certain secondary outcomes, uncovering system-specific disease progression.
## Introduction
Type 2 diabetes mellitus (T2D) has an estimated prevalence of $10\%$ in the United States and is on the rise1,2, leading to increased risk of premature death3, prolonged hyperglycemia from insulin resistance and relative insulin deficiency4, and numerous micro- and macrovascular complications5. These complications affect several organs and tissues, causing e.g. heart damage, eye problems, and nerve disease. Even though there are over 500 independent genetic variants associated with T2D6–8, there is little understanding of their pathophysiology leading to T2D itself and to secondary outcomes. As with other complex traits, most T2D-associated variants are located within non-coding regions of the genome, and might interrupt the action of regulatory elements crucial in relevant tissues9. Several studies point to an enrichment of T2D-associated variants in tissues such as pancreas, adipose, skeletal muscle, liver, arteries, kidney and the heart8,10,11. These are also the tissues affected by secondary outcomes related to diabetes, such as heart disease, nephropathy or peripheral artery disease. Genes controlled by regulatory elements affected by DNA variation act in different pathways in these tissues, and disturbance in gene expression is often reflected in a variety of outcomes for which T2D is a risk factor. Thus, T2D-associated variants altering expression of genes in the heart may more likely affect disease progression through heart-mediated processes rather than kidney-mediated processes. It follows that, while all patients are initially diagnosed with T2D, some patients may develop coronary artery disease while others may suffer kidney failure (Fig. 1).Figure 1Hypothetical tissue-specific consequences of T2D variants. A number of bona fide SNPs have been associated with T2D risk. A subset of these SNPs overlaps enhancers, promoters and eQTLs in different tissues. The variant might affect the regulatory element it overlaps, and consequently the expression of the gene affected by this regulatory element. Carriers of particular sets of tissue-specific regulatory SNPs might manifest consequences of T2D in a different way, with different progression and outcome profiles.
We aimed at identifying T2D progression profiles and their impact in different secondary outcomes whose risk is increased by T2D. To this end, we studied the effect of the combination of sets of T2D-associated variants that show tissue-specificity based on epigenetic markers and expression quantitative trait loci (eQTL) in secondary outcomes related to T2D (Fig. 2). We identified subsets of T2D-associated variants or single-nucleotide polymorphisms (SNPs) acting in regulatory elements and expression quantitative trait loci (eQTL) of relevant tissues, and assessed whether each subset had an increased risk to develop secondary outcomes related to T2D, in order to unravel tissue-specific genetic profiles that could increase risk of an outcome affecting tissues of interest. Figure 2Overview of the approach. We selected variants from 425 T2D-associated loci acting on enhancers, promoters and eQTLs of nine tissues relevant for T2D and its secondary outcomes. We used the nine tissue-grouped variant sets as input for downstream analyses to test their effects on ten secondary outcomes related to T2D: 2-sample MR and PheWAS.
## Overview of the approach
We identified subsets of T2D-associated variants acting in regulatory elements or influencing expression of genes in tissues relevant to T2D and secondary outcomes related to T2D (Fig. 2). We then associated these variant sets to T2D-related outcomes through both 2-sample Mendelian randomization (MR) and PheWAS, in order to investigate for differential (causal) effects..
## T2D-associated variants are enriched in pancreas, heart, eye, liver and kidney
We applied LD score regression (LDSC) in epigenetic data from the Epigenome Integration across Multiple Annotation Projects (Epimap), replicating the cell-type enrichment of T2D-associated variants previously observed in8,10 (“Methods”). LDSC can be used to test whether a particular genome annotation, such as histone marks, capture more heritability than expected by chance. The strongest enrichment was in pancreas ($$p \leq 8.7$$ × 10−4), followed by heart ($$p \leq 0.01$$), endometrial adenocarcinoma ($$p \leq 0.012$$), Multi-Potent Progenitor (MPP) cells ($$p \leq 0.014$$), eye ($$p \leq 0.016$$), liver ($$p \leq 0.017$$) and kidney ($$p \leq 0.032$$) (Supplemental Table 1). Previous studies also highlight the involvement of adipose tissue8 and skeletal muscle12. Due to the involvement of pancreas, heart, eye, liver, kidney, skeletal muscle and adipose tissue in secondary outcomes related to T2D, we selected these tissues for downstream analysis. Despite arteries, nerve and adrenal gland not showing significant results in our enrichment analyses, we included these tissues due to their known involvement in subsequent outcomes of T2D13–16. We used esophagus as a control tissue for downstream analysis, a tissue ranked one of the least significant p-values in the LDSC analysis and is not known to be involved in secondary outcomes related to T2D.
## Identification of sets of variants acting on regulatory elements in relevant tissues
We used three methods to identify subsets of T2D variants acting in tissues involved in secondary outcomes related to T2D: overlap with enhancers and promoters, Summary-based Mendelian Randomization (SMR)17 and Fast Enrichment Estimation Aided Colocalization Analysis FastENLOC18,19.
## Overlap with enhancers and promoters
We identified high confidence enhancers in nine tissues of interest, namely adipose, adrenal gland, arteries, heart, kidney, liver, muscle, nerve, pancreas and esophagus (control) (“Methods”). Briefly, we used as criteria for a confidence enhancer to be present in at least $\frac{2}{3}$ of all datasets of its category in adult tissue present on Epimap database20. According to these criteria, no datasets of adult kidney were available on Epimap20, so we excluded kidney from downstream analysis. We identified an average of 28,047 enhancers and 19,483 promoters in the datasets included (Supplemental Table 2). We further identified enhancers and promoters that are unique to each of the nine tissues of interest. An average of 2371 unique enhancers and 603 unique promoters were identified.
Next, we overlapped all 425 T2D-associated SNPs and their non-coding proxies in high linkage disequilibrium (LD) ($$n = 14$$,007) with these high confidence enhancers and promoters, as well as the unique enhancers and promoters (“Methods”).
## Summary-based Mendelian Randomization (SMR)
We ran SMR17 to identify variants that affect both gene expression and T2D risk. An average of 75 non-independent variants per tissue passed the SMR and HEIDI tests thresholds (“Methods”). We selected significant SMR results unique to each of the nine tissues of interest, in comparison with all datasets from the Genotype-Tissue Expression (GTEx) project.
## FastENLOC
We used fastENLOC18,19 to identify T2D-associated variants that colocalize with eQTLs in the tissues of interest. An average of six non-independent variants per tissue passed the colocalization threshold (“Methods”). We selected unique fastENLOC results as those significant results unique to each of the nine tissues of interest, in relation to all datasets from GTEx.
After performing all three analyses for identification of subsets of variants acting on tissues relevant for T2D and its secondary outcomes, we further narrowed down these subsets to only independent variants (“Methods”). The final subsets of T2D variants had on average 176 SNPs, while subsets of T2D variants acting on unique regulatory elements had on average 30 SNPs (Table 1). These were used as input/instruments on 2-sample MR analyses and PheWAS analyses. Table 1Number of SNPs and their F-statistics (median, 25th and 75th percentiles) in each tissue-grouped variant set, using all regulatory elements and tissue-specific (unique) regulatory elements. TissueAll regulatory elementsTissue-specific (unique) regulatory elementsNF-statistics median (IQR)NF-statistics median (IQR)Adipose19639.8 (31.4–64.6)4551.9 (36.3–71.9)Adrenal gland14040 (31.7–60.6)1350.6 (38.8–71.2)Aorta/arteries20939.3(31–59.2)6534.6 (31.2–56.9)Esophagus19539.8 (32–60.4)6443.4 (33.8–56.8)Heart18639.2 (31.6–63.7)5237.4 (26.7–69.6)Liver16739.8 (31.5–64.9)4436.9 (29.7–64)Muscle16539.4 (30.9–59)6039.7 (29.7–53.7)Nerve17238.9 (31.3–58.6)1736.5 (31.7–64.7)Pancreas15039.4 (31.3–58.9)2435.4 (29.7–44.8)IQR interquartile range.
## No causal relation between tissue-grouped variant sets and T2D secondary outcome
We ran MR analyses to assess the association of each tissue-grouped variant set as genetic instruments to secondary outcomes related to T2D (“Methods”). Briefly, MR is a method that uses genetic variants to estimate causal effects between the exposure and outcome under a set of assumptions, such as independence of confounding factors21. MR-analysis was performed using an inverse-variance weighted (IVW) linear regression, with instrument-outcome associations as dependent variable, instrument-exposure associations as independent variable, and with the intercept constrained to zero (“Methods”). Considering all T2D genetic instruments (425 lead variants identified in7), an increase in T2D risk was associated with an increased risk of all outcomes tested, apart from the control esophagitis (Supplemental Fig. 1). In the tissue-grouped MR analyses, for all outcomes tested except chronic kidney disease (CKD) and the control outcome esophagitis, all tissue-grouped variant sets showed to increase risk of secondary outcomes related to T2D, including the set of our control tissue, esophagus (Supplemental Fig. 1). These results were consistent both when using tissue-grouped variant sets of variants overlapping all regulatory elements in tissues of interest, as well as regulatory elements unique to each of the tissues of interest. Due to the lower power of tissue-grouped sets of variants acting in unique regulatory elements, their confidence intervals were much wider than the much bigger sets of SNPs overlapping all regulatory elements in tissues of interest. In the analysis including variants overlapping all regulatory elements, as expected, the outcome with most risk increase was T2D itself (highest T2D odds ratio (OR) 2.44, $95\%$ (confidence interval (CI) 2.29–2.60) in adrenal gland, 2.44 ($95\%$ CI 2.30–2.59) in nerve). The other outcomes (apart from the control esophagitis) had ORs between 1.04 ($95\%$ CI 0.93–1.16) for CKD in the pancreas subset, and 2.07 ($95\%$ CI 1.68–2.53) for diabetic neuropathy in the adipose subset (Supplemental Fig. 1, Supplemental Table 3). However, we did not observe an instance in which a tissue-grouped variant set increased risk of a secondary outcome of T2D more than others. We also performed three complementary analyses which relax the assumption of no horizontal pleiotropy amongst the genetic variants. First, MR-Egger regression, of which the intercept formally tests for the presence of unbalanced horizontal pleiotropy, and the slope reflects the causal effect estimate after adjusting for this pleiotropy by adding an intercept to the IVW method22. We also applied weighted median-based estimator23 and the weighted mode-based estimator24, which respectively use the weighted median of, and the highest density of, the ratio estimates across the individual instruments as estimate of the true causal effect. Sensitivity analyses using MR Egger, Weighted Median and Weighted Mode were largely consistent with IVW results, and similarly did not provide evidence for heterogeneity across variant sets (Supplemental Table 3).
## Phenome-wide analyses of tissue-specific genetic risk scores (GRSs)
Finally, we took a disease-agnostic approach and tested the association of the GRS of each tissue-grouped variant set (i.e., SNPs overlapping with all regulatory elements in the nine tissues of interest) with phenotypes in a PheWAS analysis31,32 (“Methods”). The only phenotype to pass Bonferroni correction was T2D and its variations, such as T2D with renal manifestations or T2D with peripheral circulatory disorders (Supplemental Fig. 2). When assessing results at nominal significance ($p \leq 0.05$), no tissue-grouped variant set was associated to diseases linked to both T2D and the tissue itself, such as cardiovascular diseases associated to heart-grouped or artery-grouped variants, or obesity-related diseases associated to adipose-grouped variants (Supplemental Fig. 2).
## Discussion
We hypothesized that subsets of variants associated with T2D could have tissue-specific effects, and therefore would influence the emergence of specific secondary outcomes. Similar to previous analyses, we observed an enrichment of T2D variants in pancreas, heart, eye, liver and kidney8,10. We obtained subsets of T2D-associated regulatory variants in these tissues when available, and also others previously implicated in T2D etiology (adipose, adrenal gland, skeletal muscle, arteries and nerve).
We used the selected tissue-grouped variant sets in two analyses to assess their association with secondary outcomes related to T2D: 2-sample MR and PheWAS. T2D loci found through GWAS might speak primarily to the development of T2D (and the tissues important to developing T2D) rather than tissue-specific downstream consequences of T2D. We did not observe, in any of the analyses carried out, that a tissue-grouped variant set increases risk of any particular outcome more than other tissue-grouped variant sets, or has a specific disease signature.
The identification of T2D-associated variants acting in regulatory elements and gene expression is limited to the databases of regulatory elements and gene expression available. Despite the great number of regulatory elements identified by Epimap20, we still do not have the full catalog of regulatory elements in all human tissues and cells. Recent large-scale common and rare GWAS suggest that substantially larger association studies are needed to identify most T2D loci in the population25. Similarly, larger datasets capturing the regulatory landscape of the human genome in relevant tissues are needed to help explain T2D-associated loci, and this work can be extended and applied to more high powered eQTL databases26,27 and more recently published atlas of relevant single-cell epigenomes. Currently available single-cell epigenomic data include T2D-relevant tissue such as coronary artery 28, heart 29, and pancreatic islets30. An extension of this work could also benefit from a more refined selection of input variants associated to T2D, such as the fine-mapped credible sets of potential causal variants for each T2D risk locus made available by31. Such efforts will increase the potential of identification of causal variants acting on gene regulation, and might identify groups of T2D-associated variants that have specific effects in disease risk.
Moreover, despite being assumed that the majority of GWAS-associated loci, which are non-coding, exert small regulatory effects on the expression of genes, the majority of disease-associated genetic variants have not yet been clearly explained by current eQTL data32–35. Studies have found that 5–$40\%$ of trait associations co-localize with eQTLs in relevant tissues36–40. In fact, a study designed specifically to investigate the link between genetic association and regulatory function has failed to capture it. The authors observed that for the majority of putatively causative genes considered, no fine-mapped variants were associated with regulatory regions in relevant tissues40. The authors speculate that lack of statistical power could be one of the reasons, as well as the biological context – causative eQTLs may only manifest in certain developmental windows, under specific conditions, or in a crucial cell subpopulation40. The above may explain why another effort to identify tissue-grouped variant sets based on tissue-expression profiles has similarly failed to identify different disease risks for their tissue-grouped sets41,42. However, a more recent study utilizing larger gene expression datasets for brain and subcutaneous adipose tissue showed evidence that BMI-associated variants colocalizing with gene expression in brain tissue might be driving the genetically predicted effects of BMI on cardiovascular-disease endpoints, whereas adipose tissue variants might predominantly explain the effects of BMI on measures of cardiac function43.
Another limitation of this study is that GWAS summary statistics for the secondary outcomes related to T2D studied where all control individuals have diabetes were not available. Thus, while for example in the case of outcomes such as diabetic nephropathy it is possible that many controls had diabetes, we could not filter for those individuals specifically, and the control group may include individuals without diabetes, and a mix of type 1 and type 2 diabetes.
To conclude, our novel approach for the identification and assessment of tissue-grouped T2D-associated variants did not find evidence for significantly different causal effects in any tissue-grouped variant set that could be used for prediction of secondary outcomes related to T2D. Increasing sample sizes, both in the number of participants as the number of regulatory variants identified in each specific tissue, may overcome the limitations faced in this study. Moreover, the use of datasets at the single-cell resolution could help capture effects not observed in analysis of RNA sequencing performed on bulk tissue. As more experiments on the investigation of the regulatory landscapes in a variety of tissues are performed, the more data will be available for such integration, and the more our knowledge will increase on how regulatory variants act on specific tissues and the interplay of regulatory elements. Further investigation on tissue-specific genetic risk profiles can not only help us understand the disease mechanisms, but also build a basis for tissue-specific, genetic profile-driven therapeutics.
## Ethical approval
All research was performed in accordance with relevant guidelines/regulations, and informed consent for sequencing, phenotype assessment, and publication of results was obtained at time of enrollment for BioMe biobank and FinnGen participants. The Coordinating Ethics Committee of the Helsinki and Uusimaa Hospital District has evaluated FinnGen, and the EU Data Protection Regulation that came into force in May 2018 has been taken into account when planning the project. Further details can be found in https://www.finngen.fi/en/code_of_conduct. BioMe Biobank was approved by the Program for the Protection of Human Subjects. Further details are located in the BioMe researcher FAQ (https://icahn.mssm.edu/research/ipm/programs/biome-biobank/researcher-faqs) and https://icahn.mssm.edu/research/pphs.
## Description of the cohorts
The Mount Sinai BioMe *Biobank is* an ongoing electronic health record (EHR)-linked biorepository that enrolls participants non-selectively from the Mount Sinai Health System44. It has included 60,000 participants from the greater New York *City area* since its inception in 2007. Participants are between 18 and 89 years of age and represent a broad spectrum of racial and ethnic diversity (African ($24\%$), European ($32\%$), Hispanic-Latino ($35\%$) and other ($9\%$) ancestries). At enrollment, participants consent to link their DNA and plasma samples to their de-identified EHRs. Clinical and EHR information are complemented by a detailed questionnaire that gathers demographic and lifestyle information. The median number of clinical encounters for BioMe participants is 21.
The FinnGen study utilizes samples collected by a nationwide network of Finnish biobanks. The study is based on combining genome information with digital health care data from national health registries. The R5 freeze used in this study consists of > 218,700 individuals, up to 17 M variants and > 2800 phenotypes45.
## Variant selection and tissue enrichment
We obtained summary statistics including 425 loci identified by a GWAS meta-analysis, including 21 independent ($p \leq 5$e − 8, > 500 kb and LD r2 < 0.05) variants identified in Europeans only, 153 novel independent SNPs identified in the transethnic meta-analysis, and 251 independent established T2D variants8. Full summary statistics comprise SNP, chromosome, position, effect and non-effect allele frequencies, beta, standard deviation, p-value and N. The full meta-analysis included 1.4 million participants and identified a total of 568 associations across all ancestries. We used LDSC46 to perform tissue enrichment analysis, integrating the full summary statistics from Vujkovic et al.8 with 806 datasets of predicted enhancers from the Epimap project20.
## Tissue-grouped variant sets
We selected variants acting on regulatory elements in each tissue of interest by overlapping variants with enhancers or promoters and bona fide eQTLs. In a secondary analysis, we identified enhancers, promoters and eQTLs that are unique to the tissues of interest, and overlapped those with T2D-associated loci. Variants passing these criteria were then grouped by tissue, and each tissue group was narrowed down to independent variants using function clump from PLINK v1.947, parameters --clump-p1 1e-5 --clump-kb 500 --clump-r2 0.001, using 1000 Genomes phase3 Europeans as reference panel48. GRS was calculated using function –score from PLINK v1.947, weighted by European-specific effect sizes from Vujkovic et al.8.
## T2D variants acting on regulatory elements
We used enhancers and promoters predicted within the scope of Epimap20. We downloaded chromHMM tracks on 35 tissues and cell-types from https://personal.broadinstitute.org/cboix/epimap/ChromHMM/observed_aux_18_hg$\frac{19}{.}$ A list of samples used in this analysis and grouped by tissue is available on Supplemental Table 2. We included only tissues that had at least three replicates generated from adult tissues. We selected enhancer regions classified by ChromHMM as EnhA1, EnhA2, EnhG1 and EnhG2. We selected as promoters regions classified by ChromHMM as TssA, TssFlnk, TssFlnkU or TssFlnkD. We assessed how many times each enhancer and promoter appear across all datasets of each tissue, using bedtools multiinter49. We then retrieved only enhancers or promoters that appear in at least two thirds of the total number of datasets available for each tissue (Supplemental Table 2).
After building a database of high confidence enhancers and promoters in 35 tissues, we also identified regulatory elements that are unique to each tissue, using function bedtools intersect -v. We then overlapped T2D variants and their proxies in high LD with the full set of confidence enhancers/promoters identified in the nine tissues of interest, and the set of enhancers/promoters unique to each tissue of interest, using function bedtools intersect. High LD was defined as r2 > 0.8, retrieved using FUMA v1.3.6b50 and their built-in database of 10,000 randomly selected unrelated Europeans from the UKBiobank51,52 as reference panel. A total of 14,007 variants were intersected.
## T2D variants acting on eQTLs
We used two methods to identify T2D-associated variants influencing gene expression: SMR and colocalization with fastEnloc. SMR integrates gene expression information to pinpoint candidate causal variants by determining whether the association between an associated SNP and the phenotype is mediated through an eQTL17. fastENLOC is a Bayesian hierarchical colocalization method that prioritizes candidate causal variants by colocalizing associated variants and eQTLs18. For both analyses, we used as input data from GTEx v836. Datasets included eQTL information on the tissues relevant for T2D and its secondary outcomes: adipose subcutaneous ($$n = 663$$), adipose visceral ($$n = 541$$), adrenal gland ($$n = 258$$), artery aorta ($$n = 432$$), artery coronary ($$n = 240$$), artery tibial ($$n = 663$$), heart atrial appendage ($$n = 429$$), heart left ventricle ($$n = 432$$), liver ($$n = 226$$), muscle skeletal ($$n = 803$$), nerve tibial ($$n = 619$$) and pancreas ($$n = 328$$). We used esophagus–gastroesophageal junction ($$n = 375$$), mucosa ($$n = 555$$) and esophagus muscularis ($$n = 515$$) as control tissues.
Briefly, the SMR & HEIDI approach integrates summary-level data from GWAS and eQTL studies to test if a transcript and phenotype are associated because of a shared causal variant (i.e., pleiotropy)17. We retrieved variants that simultaneously affect gene expression and T2D risk that passed a Bonferroni corrected p-SMR and a p-HEIDI > 0.05, as in similar studies17. LD data required for the HEIDI test were estimated from genotyped data from the UK Biobank (UKB) study, including 10,000 randomly selected European participants.
We applied fastENLOC18,19, a Bayesian hierarchical colocalization method, to assess which T2D-associated variants colocalize with eQTLs in tissues of interest. We used pre-computed GTEx v8 multi-tissue eQTL annotation available on https://github.com/xqwen/fastenloc. Variants that passed the threshold for SNP colocalization probability (SCP) > 0.1 were considered “colocalizing”.
In order to retrieve eQTLs unique to each tissue of interest, we first ran SMR and fastENLOC on all tissues available on GTEx. We then selected eQTLs passing SMR and fastENLOC significance thresholds that are unique to each of the nine tissues of interest in this study.
## Two-sample Mendelian Randomization
Previous works have described the methods for MR analysis of summary data based on two studies53,54. Here, we used all variants in the tissue-grouped variant sets as proposed instruments to measure their associations to ten outcomes with summary data available from the FinnGen study45: T2D (n total = 215,654; n cases = 35,607), diabetic nephropathy (n total = 213,746; n cases = 3,283), CKD (n total = 216,743; n cases = 3902), peripheral artery disease (PAD) (n total = 213,639; n cases = 7098), heart failure (HF) (n total = 218,208; n cases = 13,087), stroke (n total = 180,862; n cases = 18,661), myocardial infarction (MI) (n total = 200,641; n cases = 11,622), diabetic retinopathy (n total = 216,666; n cases = 14,584), diabetic neuropathy (n total = 163,616; n cases = 1415) and esophagitis (n total = 190,442; n cases = 747) as a control (case definition for all outcomes can be found on Supplementary Table 4). SNP-exposure associations were retrieved from the summary statistics from Vujkovic et al.8, and SNP-outcome associations come from summary statistics from Finngen45. All tissue-grouped variant sets were considered composed of sufficiently strong instruments based on their F-statistics, considering an F-statistic > 10 as strong enough instrument to avoid weak instrument bias55 (Table 1). Using fixed effects IVW analyses, we combined the effects of the individual genetic instruments to obtain a genetically determined association between exposure and outcome under the assumption of the absence of horizontal pleiotropy. Some variants from the tissue-grouped variant sets were removed from MR analyses due to being palindromic genetic instruments with intermediate allele frequencies. Estimates from the IVW analyses can be interpreted as the odds ratio for the outcome trait(s) per 2.72-fold increase in the odds of T2D (i.e., a one unit change in genetic liability to T2D on the log odds scale). We also run sensitivity analyses using methods MR Egger22, Weighted Median23 and Weighted Mode24. Analyses were run using the R-based package ‘TwoSampleMR’56.
## Phenome-Wide Association Study (PheWAS)
We used the PheWAS package in R57 using default settings to test for associations between our tissue-grouped variant sets and a wide range of phenotypes. We included 1039 disease outcomes in 8370 individuals of self-reported European ancestry from BioMe biobank44, using age, sex, body mass index (BMI) and 10 first principal components as covariates. We report the ten most significant associations. Results were corrected for multiple testing by Bonferroni test.
## Supplementary Information
Supplementary Figures. Supplementary Tables. The online version contains supplementary material available at 10.1038/s41598-023-30369-6.
## References
1. Almgren P. **Heritability and familiality of type 2 diabetes and related quantitative traits in the Botnia Study**. *Diabetologia* (2011.0) **54** 2811-2819. DOI: 10.1007/s00125-011-2267-5
2. Xu G. **Prevalence of diagnosed type 1 and type 2 diabetes among US adults in 2016 and 2017: Population based study**. *BMJ* (2018.0) **362** k1497. DOI: 10.1136/bmj.k1497
3. Tancredi M. **Excess mortality among persons with type 2 diabetes**. *N. Engl. J. Med.* (2015.0) **373** 1720-1732. DOI: 10.1056/NEJMoa1504347
4. Kahn SE, Cooper ME, Del Prato S. **Pathophysiology and treatment of type 2 diabetes: Perspectives on the past, present, and future**. *Lancet* (2014.0) **383** 1068-1083. DOI: 10.1016/S0140-6736(13)62154-6
5. Fowler GC, Vasudevan DA. **Type 2 diabetes mellitus: managing hemoglobin A(1c) and beyond**. *South Med. J.* (2010.0) **103** 911-916. DOI: 10.1097/SMJ.0b013e3181eb34b2
6. Xue A. **Genome-wide association analyses identify 143 risk variants and putative regulatory mechanisms for type 2 diabetes**. *Nat. Commun.* (2018.0) **9** 2941. DOI: 10.1038/s41467-018-04951-w
7. Scott RA. **An expanded genome-wide association study of type 2 diabetes in Europeans**. *Diabetes* (2017.0) **66** 2888-2902. DOI: 10.2337/db16-1253
8. Vujkovic M. **Discovery of 318 new risk loci for type 2 diabetes and related vascular outcomes among 1.4 million participants in a multi-ancestry meta-analysis**. *Nat. Genet.* (2020.0) **52** 680-691. DOI: 10.1038/s41588-020-0637-y
9. Visscher PM. **10 years of GWAS discovery: Biology, function, and translation**. *Am. J. Hum. Genet.* (2017.0) **101** 5-22. DOI: 10.1016/j.ajhg.2017.06.005
10. Torres JM. **A multi-omic integrative scheme characterizes tissues of action at loci associated with type 2 diabetes**. *Am. J. Hum. Genet.* (2020.0) **107** 1011-1028. DOI: 10.1016/j.ajhg.2020.10.009
11. Parker SC. **Chromatin stretch enhancer states drive cell-specific gene regulation and harbor human disease risk variants**. *Proc. Natl. Acad. Sci. USA* (2013.0) **110** 17921-17926. DOI: 10.1073/pnas.1317023110
12. Scott LJ. **The genetic regulatory signature of type 2 diabetes in human skeletal muscle**. *Nat. Commun.* (2016.0) **7** 11764. DOI: 10.1038/ncomms11764
13. Kenny HC, Abel ED. **Heart failure in type 2 diabetes mellitus**. *Circ. Res.* (2019.0) **124** 121-141. DOI: 10.1161/CIRCRESAHA.118.311371
14. Dal Canto E. **Diabetes as a cardiovascular risk factor: An overview of global trends of macro and micro vascular complications**. *Eur. J. Prev. Cardiol.* (2019.0) **26** 25-32. DOI: 10.1177/2047487319878371
15. Younger DS. **Diabetic neuropathy: A clinical and neuropathological study of 107 patients**. *Neurol. Res. Int.* (2010.0) **2010** 140379. DOI: 10.1155/2010/140379
16. Deng Y. **Global, regional, and national burden of diabetes-related chronic kidney disease from 1990 to 2019**. *Front. Endocrinol. (Lausanne)* (2021.0) **12** 672350. DOI: 10.3389/fendo.2021.672350
17. Zhu Z. **Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets**. *Nat. Genet.* (2016.0) **48** 481-487. DOI: 10.1038/ng.3538
18. Pividori M. **PhenomeXcan: Mapping the genome to the phenome through the transcriptome**. *Sci. Adv.* (2020.0) **6** eaba2083. DOI: 10.1126/sciadv.aba2083
19. Wen X, Pique-Regi R, Luca F. **Integrating molecular QTL data into genome-wide genetic association analysis: Probabilistic assessment of enrichment and colocalization**. *PLoS Genet.* (2017.0) **13** e1006646. DOI: 10.1371/journal.pgen.1006646
20. Boix CA. **Regulatory genomic circuitry of human disease loci by integrative epigenomics**. *Nature* (2021.0) **590** 300-307. DOI: 10.1038/s41586-020-03145-z
21. Davies NM, Holmes MV, Davey Smith G. **Reading Mendelian randomisation studies: A guide, glossary, and checklist for clinicians**. *BMJ* (2018.0) **362** k601. DOI: 10.1136/bmj.k601
22. Bowden J, Davey Smith G, Burgess S. **Mendelian randomization with invalid instruments: Effect estimation and bias detection through Egger regression**. *Int. J. Epidemiol.* (2015.0) **44** 512-525. DOI: 10.1093/ije/dyv080
23. Bowden J. **Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator**. *Genet. Epidemiol.* (2016.0) **40** 304-314. DOI: 10.1002/gepi.21965
24. Hartwig FP, Davey Smith G, Bowden J. **Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption**. *Int. J. Epidemiol.* (2017.0) **46** 1985-1998. DOI: 10.1093/ije/dyx102
25. Flannick J, Florez JC. **Type 2 diabetes: Genetic data sharing to advance complex disease research**. *Nat. Rev. Genet.* (2016.0) **17** 535-549. DOI: 10.1038/nrg.2016.56
26. Taylor DL. **Integrative analysis of gene expression, DNA methylation, physiological traits, and genetic variation in human skeletal muscle**. *Proc. Natl. Acad. Sci. USA* (2019.0) **116** 10883-10888. DOI: 10.1073/pnas.1814263116
27. Vinuela A. **Genetic variant effects on gene expression in human pancreatic islets and their implications for T2D**. *Nat. Commun.* (2020.0) **11** 4912. DOI: 10.1038/s41467-020-18581-8
28. Turner AW. **Author Correction: Single-nucleus chromatin accessibility profiling highlights regulatory mechanisms of coronary artery disease risk**. *Nat. Genet.* (2022.0) **54** 1259. DOI: 10.1038/s41588-022-01142-8
29. Hocker JD. **Cardiac cell type-specific gene regulatory programs and disease risk association**. *Sci. Adv.* (2021.0) **7** eabf1444. DOI: 10.1126/sciadv.abf1444
30. Chiou J. **Single-cell chromatin accessibility identifies pancreatic islet cell type- and state-specific regulatory programs of diabetes risk**. *Nat. Genet.* (2021.0) **53** 455-466. DOI: 10.1038/s41588-021-00823-0
31. Mahajan A. **Multi-ancestry genetic study of type 2 diabetes highlights the power of diverse populations for discovery and translation**. *Nat. Genet.* (2022.0) **54** 560-572. DOI: 10.1038/s41588-022-01058-3
32. Arvanitis M. **Redefining tissue specificity of genetic regulation of gene expression in the presence of allelic heterogeneity**. *Am. J. Hum. Genet.* (2022.0) **109** 223-239. DOI: 10.1016/j.ajhg.2022.01.002
33. Mu Z. **The impact of cell type and context-dependent regulatory variants on human immune traits**. *Genome Biol.* (2021.0) **22** 122. DOI: 10.1186/s13059-021-02334-x
34. Yao DW. **Quantifying genetic effects on disease mediated by assayed gene expression levels**. *Nat. Genet.* (2020.0) **52** 626-633. DOI: 10.1038/s41588-020-0625-2
35. Chun S. **Limited statistical evidence for shared genetic effects of eQTLs and autoimmune-disease-associated loci in three major immune-cell types**. *Nat. Genet.* (2017.0) **49** 600-605. DOI: 10.1038/ng.3795
36. **The GTEx Consortium atlas of genetic regulatory effects across human tissues**. *Science* (2020.0) **369** 1318-1330. DOI: 10.1126/science.aaz1776
37. Stranger BE. **Population genomics of human gene expression**. *Nat. Genet.* (2007.0) **39** 1217-1224. DOI: 10.1038/ng2142
38. Vuckovic D. **The polygenic and monogenic basis of blood traits and diseases**. *Cell* (2020.0) **182** 1214-1231.e11. DOI: 10.1016/j.cell.2020.08.008
39. Giambartolomei C. **Bayesian test for colocalisation between pairs of genetic association studies using summary statistics**. *PLoS Genet.* (2014.0) **10** e1004383. DOI: 10.1371/journal.pgen.1004383
40. Connally N. **The missing link between genetic association and regulatory function**. *eLife* (2021.0) **11** e74970. DOI: 10.7554/eLife.74970
41. Verkouter I. **The contribution of tissue-grouped BMI-associated gene sets to cardiometabolic-disease risk: A Mendelian randomization study**. *Int. J. Epidemiol.* (2020.0) **49** 1246-1256. DOI: 10.1093/ije/dyaa070
42. Kutalik Z. **Commentary on: "The contribution of tissue-specific BMI-associated gene sets to cardiometabolic disease risk: A Mendelian randomization study"**. *Int. J. Epidemiol.* (2020.0) **49** 1257-1258. DOI: 10.1093/ije/dyaa062
43. Leyden GM. **Harnessing tissue-specific genetic variation to dissect putative causal pathways between body mass index and cardiometabolic phenotypes**. *Am. J. Hum. Genet.* (2022.0) **109** 240-252. DOI: 10.1016/j.ajhg.2021.12.013
44. Belbin GM. **Toward a fine-scale population health monitoring system**. *Cell* (2021.0) **184** 2068-2083.e11. DOI: 10.1016/j.cell.2021.03.034
45. 45.FinnGen, FinnGen documentation R5 release. 2020.
46. Finucane HK. **Partitioning heritability by functional annotation using genome-wide association summary statistics**. *Nat. Genet.* (2015.0) **47** 1228-1235. DOI: 10.1038/ng.3404
47. Purcell S. **PLINK: A tool set for whole-genome association and population-based linkage analyses**. *Am. J. Hum. Genet.* (2007.0) **81** 559-575. DOI: 10.1086/519795
48. Auton A. **A global reference for human genetic variation**. *Nature* (2015.0) **526** 68-74. DOI: 10.1038/nature15393
49. Quinlan AR, Hall IM. **BEDTools: A flexible suite of utilities for comparing genomic features**. *Bioinformatics* (2010.0) **26** 841-842. DOI: 10.1093/bioinformatics/btq033
50. Watanabe K. **Functional mapping and annotation of genetic associations with FUMA**. *Nat. Commun.* (2017.0) **8** 1826. DOI: 10.1038/s41467-017-01261-5
51. Bycroft C. **The UK Biobank resource with deep phenotyping and genomic data**. *Nature* (2018.0) **562** 203-209. DOI: 10.1038/s41586-018-0579-z
52. Sudlow C. **UK biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age**. *PLoS Med.* (2015.0) **12** e1001779. DOI: 10.1371/journal.pmed.1001779
53. Burgess S. **Using published data in Mendelian randomization: A blueprint for efficient identification of causal risk factors**. *Eur. J. Epidemiol.* (2015.0) **30** 543-552. DOI: 10.1007/s10654-015-0011-z
54. Noordam R. **Assessment of causality between serum gamma-glutamyltransferase and type 2 diabetes mellitus using publicly available data: a. Mendelian randomization study**. *Int. J. Epidemiol.* (2016.0) **45** 1953-1960. PMID: 28031309
55. Staiger D, Stock JH. **Instrumental variables regression with weak instruments**. *Econometrica* (1997.0) **65** 557-586. DOI: 10.2307/2171753
56. Hemani G. **The MR-Base platform supports systematic causal inference across the human phenome**. *Elife.* (2018.0) **7** e34408. DOI: 10.7554/eLife.34408
57. Carroll RJ, Bastarache L, Denny JC. **R PheWAS: Data analysis and plotting tools for phenome-wide association studies in the R environment**. *Bioinformatics* (2014.0) **30** 2375-2376. DOI: 10.1093/bioinformatics/btu197
|
---
title: Bacterial PncA improves diet-induced NAFLD in mice by enabling the transition
from nicotinamide to nicotinic acid
authors:
- Shengyu Feng
- Liuling Guo
- Hao Wang
- Shanshan Yang
- Hailiang Liu
journal: Communications Biology
year: 2023
pmcid: PMC9981684
doi: 10.1038/s42003-023-04613-8
license: CC BY 4.0
---
# Bacterial PncA improves diet-induced NAFLD in mice by enabling the transition from nicotinamide to nicotinic acid
## Abstract
Nicotinamide adenine dinucleotide (NAD+) is crucial for energy metabolism, oxidative stress, DNA damage repair, longevity regulation, and several signaling processes. To date, several NAD+ synthesis pathways have been found in microbiota and mammals, but the potential relationship between gut microbiota and their hosts in regulating NAD+ homeostasis remains largely unknown. Here, we showed that an analog of the first-line tuberculosis drug pyrazinamide, which is converted by nicotinamidase/pyrazinamidase (PncA) to its active form, affected NAD+ level in the intestines and liver of mice and disrupted the homeostasis of gut microbiota. Furthermore, by overexpressing modified PncA of Escherichia coli, NAD+ levels in mouse liver were significantly increased, and diet-induced non-alcoholic fatty liver disease (NAFLD) was ameliorated in mice. Overall, the PncA gene in microbiota plays an important role in regulating NAD+ synthesis in the host, thereby providing a potential target for modulating host NAD+ level.
Bacterial nicotinamidase/pyrazinamidase PncA increases host nicotinamide adenine dinucleotide (NAD+) levels and ameliorates diet-induced non-alcoholic fatty liver disease (NAFLD) in mice.
## Introduction
As early as 110 years ago, NAD+ was discovered by the British biochemists Arthur Harden and William John Young as a coenzyme involved in yeast-mediated alcohol fermentation1. Over the next ~30 years, the chemical composition of this coenzyme was determined, and it was found to participate in several redox reactions together with NADH. Although the role of NAD+ in oxidation–reduction reactions is well understood, it was not until the last 10 years that the function of NAD+ was completely elucidated. The sirtuin (SIRT) family, poly (ADP-ribose) polymerases (PARPs), and cyclic ADP-ribose synthases (cADPRSs) are NAD+-dependent enzymes; so NAD+ regulates downstream metabolic pathways by influencing the activity of these enzymes and acting as a metabolic sensor in cells2–5. In addition, these NAD+-dependent proteins play an important role in various biological processes, such as metabolism, signal transduction, oxidative stress, cognitive decline, and other aging-related physiological processes. In recent years, important studies have revealed that NAD+ binds to the 5′ end of mRNA, thereby regulating the transcription initiation of genes. However, there is still no definitive conclusion on how this process is regulated or its significance for organisms6,7.
NAD+ is a small molecule required by almost all organisms and is one of the most abundant molecules in the human body, participating in more than 500 different enzymatic reactions8. In mammals, there are three main synthesis pathways for NAD+: the de novo synthesis pathway using tryptophan, the salvage pathway using nicotinamide (NAM), and the Preiss–Handler pathway using nicotinic acid (NA). Because there are several mechanisms to synthesize NAD+ in mammals, the specific pathways used for different tissues and the most effective pathway remain unclear9,10. In the intestinal flora, there is an additional important pathway that converts NAM into NA, combining the salvage and Preiss–Handler pathways, which is known as deamidation and is catalyzed by PncA11. This process has recently been shown to be an important step for the intestinal flora to regulate host NAD+ level12. This pathway catalyzed by PncA is considered to have been abandoned during biological evolution because no homolog of this gene has been identified in higher mammals. As a result, no enzyme catalyzing the conversion of NAM to NA has been found in mammals to date9. Based on these characteristics, pyrazinamide, a drug that requires PncA to convert into its active form, was developed to treat Mycobacterium tuberculosis13.
NAFLD is the most common chronic liver disease and is closely related to metabolic syndrome. With substantial changes in diet and lifestyle, the prevalence of NAFLD has increased significantly worldwide14. However, there are currently no effective FDA-approved drugs available for clinical use. Therefore, exploring promising therapeutic targets or strategies remains a priority15. Several studies have shown that supplementation with the NAD+ precursors nicotinamide ribonucleotide (NR), nicotinamide mononucleotide (NMN), and ACMSD inhibition (an enzyme that blocks the de novo NAD+ synthesis pathway) significantly improve NAFLD in mice16–18, suggesting NAD+ regulation in the diseased liver as a potential therapeutic target. During the development of NAFLD, the intestinal flora undergoes significant changes19, and PncA in the intestinal flora plays an important role in NAD+ synthesis in mouse liver. Therefore, this study aimed to explore the significance of PncA in regulating mouse NAD+ level and its potential application value in improving NAFLD in mice.
Our research indicated that pyrazinecarbonitrile (PCN) inhibited PncA activity and affected NAD+ levels in the intestines and liver of mice, and disrupted the homeostasis of gut microbiota. Overexpression of modified PncA in mice significantly increased NAD+ level in the liver and improved diet-induced NAFLD.
## PncA inhibitor PCN disrupts gut microbiome homeostasis and reduces host NAD+ level
We summarized the NAD+ synthesis and consumption pathways in mammals and human intestinal flora (Fig. 1). We could see that PncA is absent in mammals but is highly conserved in different bacterial phyla (Supplementary Fig. S1), indicating its importance in the bacterial synthesis of NAD+. Supplementary Fig. S1 also shows other genes related to NAD+ synthesis in different types of bacteria. Fig. 1The synthesis and consumption pathways of NAD+ in mammals and intestinal flora. Two main pathways involved in mammalian NAD synthesis: de novo and salvage pathway. The former pathway converts Trp to QA through the kynurenine pathway and synthesizes NAD+. The main precursors of NAD+ are NAM, NA, NR, and NMN. NAD+-consuming enzymes mainly include SIRTs, PARPs, and CD38, and the metabolites after they use up NAD all contain NAM. In the intestinal flora, the de novo synthesis of NAD+ mainly depends on Asp rather than Trp. Although different in the de novo pathway, they share the same precursors for NAD+ synthesis. Dependence on the same precursors must lead to competition or cooperation in NAD+ metabolism between mammals and their microbiota. The synthetic pathway marked by the red line is specific to gut flora. Trp tryptophan, ACMS α-amino-β-carboxymuconate-ε-semialdehyde, QA quinolinic acid, AMS α-amino-β-muconate-ε-semialdehyde, NaAD nicotinic acid adenine dinucleotide, MNAM methylation of NAM, Asp aspartic acid, ImminoAsp immino-aspartate, QPRT nicotinate-nucleotide pyrophosphorylase, NMNAT Nicotinamide/nicotinic acid mononucleotide adenylyltransferase, NADS glutamine-dependent NAD+ synthetase, NNMT nicotinamide N-methyltransferase, NRK nicotinamide riboside kinase, NADP nicotinamide adenine dinucleotide phosphate. Figure 1 was created by Adobe Illustrator.
To verify whether the PncA in microbiota affects NAD+ synthesis in the host, we used the PncA inhibitor PCN, which is an analog of the commonly used antibiotic pyrazinamide to treat tuberculosis. PCN has been reported to have a strong inhibitory effect on the PncA enzyme activity of tuberculosis20. The protein structure alignment showed the active catalytic sites of PncA are conserved in bacteria (Supplementary Fig. S2a), so we speculated that PCN would also function in other bacteria. An enzyme assay of purified PncA of E. coli also showed that PCN strongly inhibited its deamination activity (Supplementary Fig. S2b and c). To verify the influence of PCN on bacterial growth in vitro, we selected several bacteria with different NAD+ synthesis pathways, including Bifidobacterium longum, which has both de novo synthesis and Preiss–Handler pathways; Akkermansia muciniphila, which only has the de novo synthesis pathway; and *Lactobacillus salivarius* and Streptococcus gordonii, which only have the Preiss–Handler pathway (Supplementary Fig. S1). These three types of bacteria are common gut flora in mammals. The growth rates of A. muciniphila and B. longum were almost unaffected after PCN treatment (Fig. 2a), while the growth of L. salivarius and S. gordonii, which depended only on the Preiss–Handler pathway, was strongly inhibited. This demonstrates that PCN has a strong inhibitory effect on bacteria that depend on the Preiss–Handler pathway. Although B. longum has the PncA gene, PCN had a minimal effect on its growth, indicating that B. longum mainly synthesizes NAD+ through the de novo pathway. Fig. 2PncA inhibitor disrupts intestinal microbiome homeostasis and reduces host NAD+ level.a Effects of PCN on the growth of bacteria with different NAD+ synthesis pathways. b PCA diagram of the intestinal flora in mice treated with PCN ($$n = 7$$). c Intestinal microbial species and diversity after PCN treatment in mice ($$n = 7$$, *$p \leq 0.05$). d–f NAD+ level in the liver, intestine, and hippocampus of mice ($$n = 6$$, *$p \leq 0.05$, n.s = no significance).
In the in vivo experiment, we treated mice with PCN and collected the feces of each group on the last day of the experiment. We then used high-throughput sequencing technology to analyze the changes in the intestinal flora of mice. Principal component analysis (PCA) indicated that PCN treatment had a significant impact on the gut microbes in mice (Fig. 2b). Surprisingly, mice treated with PCN exhibited a significant increase in bacterial richness compared with control mice (Fig. 2c). As a potential explanation, we speculate that PCN disrupts the original balance in the intestinal flora, resulting in the massive expansion of some bacteria. At the species level, PCN significantly increased the abundance of Helicobacter hepaticus, Clostridium cocleatum, and B. pseudolongum (Supplementary Fig. S3a). Besides, the abundance of Bacteroidales increased significantly, whereas the abundance of Clostridiales decreased significantly (Supplementary Fig. S3b). However, we did not find any relationship between those bacteria and their NAD+ synthesis pathway, owing to the complex composition and interdependence of gut microbiota. Moreover, PCN significantly inhibited electron transfer, respiration, and other pathways closely related to NAD+ in the intestinal flora (Supplementary Fig. S3c). Supplementary Fig. S3d shows that NAM increased while NA decreased in the feces of mice after PCN treatment. Together, these findings show that PCN has a significant impact on intestinal flora.
Because PncA in the intestinal flora plays an important role in the synthesis of NAD+ in the host intestine and liver, we explored the effects of PCN treatment on the level of NAD+ in the liver and intestine of host mice. PCN reduced the utilization efficiency of NAM in the liver and intestine of the host, resulting in lower NAD+ level (Fig. 2d and e). Given that PncA catalyzes the conversion of NAM to NA, we speculated that NAD+ synthesis using NA is more efficient than that using NAM in the liver and intestine. It is well known that NAM is produced when NAD+-dependent enzymes consume NAD+, and additional supplementation does not enhance NAD+ synthesis. Surprisingly, in the case of NAM supplementation, PCN also reduced the level of NAD+ in the hippocampus compared with the PBS group (Fig. 2f). According to RNA sequencing (RNA-seq) results, PCN affected the expression of numerous genes in the liver (Supplementary Fig. S4a) and significantly affected host immune processes, such as the intestinal immune network for IgA production and antigen processing (Supplementary Fig. S4b). To demonstrate that PCN regulates NAD+ level in the host through its effect on the intestinal flora rather than directly affecting the host, 293T, HepG2 cells, and antibiotic-treated mice were treated with PCN. The results showed that PCN had no effect on the growth and NAD+ level of human cells as well as antibiotic-treated mouse colon and liver (Supplementary Fig. S5a–d).
## Escherichia coli overexpressing PncA affects host liver NAD metabolism
Our previous experiments and the research of Shats et al. showed that the PncA gene in the intestinal flora plays an important role in regulating host NAD+. Therefore, we attempted to promote the synthesis of mammalian NAD+ using a new approach. We first wanted to supplement bacteria that specifically depend on the deamidation NAD+ synthesis pathway. However, controlling univariate variables for this experiment was challenging, so we constructed E. coli with overexpression of PncA (PncA-OE) and used normal E. coli strains with only vector (PncA-WT) as the control. Based on the gene expression, protein quantification, and enzyme assay of PncA in different E. coli (Supplementary Fig. S6a–c), PncA was functionally induced in the PncA-OE group.
First, we performed an in vitro experiment. NAM (0.5 mM) was added to the culture medium of E. coli, which was cultured until optical density (OD) reached 1.0. Then bacteria were centrifuged, and the supernatant was collected for metabolome analysis (Fig. 3a). As expected, PncA-OE E. coli released more NA in the bacterial culture medium than PncA-WT E. coli did (Fig. 3b). To facilitate the colonization of exogenous E. coli in the intestines of mice in vivo, we first treated mice with a cocktail of antibiotics for 5 days to reduce endogenous bacteria. Next, mice were gavaged with PncA-OE and PncA-WT E. coli and divided into NAM-supplemented and non-NAM-supplemented groups (Fig. 3c). Previous studies showed that the efficiency of NA utilization for NAD+ synthesis in the liver is higher than that of NAM21. We found that the NAD+ level in mouse liver in the PncA-OE E. coli group was higher than the control in the condition of NAM supplement (Fig. 3d).Fig. 3PncA-overexpressing E. coli affect host liver NAD metabolism.a Schematic diagram of the in vitro experiment (created by Adobe Illustrator). b Heatmap of the metabolites in the bacterial culture medium. c Schematic diagram of the mouse experiment. d Liver NAD+ level after colonizing mice with different genotypes of E. coli ($$n = 6$$, *$p \leq 0.05$, n.s = no significance).
Enhancing the synthesis of NAD+ in liver significantly improves NAFLD. Therefore, we constructed NAFLD model mice using a methionine- and choline-deficient (MCD) diet to verify whether PncA-OE E. coli could ameliorate NAFLD. MCD diet rapidly decreased the body weight of the mice (Fig. 4a) and caused liver physiological lesions (Fig. 4b). MCD diet also significantly reduced liver NAD+ and ATP levels (Fig. 4c and d) and increased triglycerides in the liver (Fig. 4e). In addition, after supplementation with PncA-OE E. coli, NAD+ and ATP level in the liver were slightly increased compared with the control group but not significantly (Fig. 4c and d). PncA-OE E. coli did not resolve the lesions in the liver (Fig. 4b). That might have been caused by the low colonization rate of E. coli in mice and low transformation efficiency of NAM to NA by PncA-OE E.coli because we did not observe accumulation of NA in feces during supplementation with PncA-OE E. coli (Supplementary Fig. S6d).Fig. 4PncA-overexpressing E. coli did not significantly ameliorate mouse NAFLD.a Body weight of MCD-induced NAFLD model mice ($$n = 6$$). b Representative Oil Red O staining (top) and hematoxylin and eosin (H&E) staining (bottom) of mouse liver sections. c Liver NAD+ level in NAFLD model mice treated with different bacteria. d Relative content of ATP in mouse liver. e Content of triglyceride in mouse liver. ( $$n = 6$$, ***$p \leq 0.001$, ****$p \leq 0.0001$, n.s = no significance). Scale bars 100 µm.
## PncA overexpression in the liver by adeno-associated virus improves hepatic lesions in mice
We found that supplementation with PncA-OE E. coli was not effective in relieving NAFLD. However, we speculate that this may have been because of the limited colonization ability of bacteria to promote liver NAD+. Therefore, instead of using bacteria, we optimized the sequence of PncA of E. coli for direct expression in mammals. A liver-specific adeno-associated virus (AAV) was constructed to carry PncA into mice via tail vein injection. PncA was highly expressed in mouse liver (Fig. 5a). Intriguingly, the NAD+ level in the AAV-PncA group was significantly higher than that in the vector group and even approximately five times higher than that in the normal diet group (Fig. 5b). This was particularly surprising because the widely used and highly efficient NAD+ precursors NR and NMN only increased the level of NAD+ in the liver by 1.5–2 fold10. However, previous studies have not shown that NA had such a strong NAD+ booster effect on mouse liver22,23. By comparing our results with others, we speculate that the effect of NA on NAD+ synthesis may be limited by the absorption efficiency of NA by liver cells. PncA expression in the liver promotes the transformation of NAM to NA in cells, maximizing the synthesis of NA to NAD+. This suggests that the nicotinamidase PncA, which appears to have been abandoned during the evolution of the NAD+ synthesis pathway, has the potential for NAD+ synthesis in mammals. Fig. 5Improvement of NAD+ in mouse liver by PncA overexpression via liver-specific AAV.a PncA expression in mouse liver in the control and PncA groups ($$n = 3$$, ****$p \leq 0.0001$). b Relative content of NAD+ in mouse liver. c Relative content of ATP in mouse liver. d Content of triglyceride in mouse liver. ( $$n = 6$$, **$p \leq 0.01$, ****$p \leq 0.0001$).
In addition, compared with the vector group, mice expressing PncA showed a significant increase in ATP level and a decreased triglyceride content after supplementation with the MCD diet (Fig. 5c and d). Furthermore, the results of Oil Red O and H&E staining showed that overexpression of PncA in mouse liver improved NAFLD-induced pathological changes (Fig. 6a). RNA-seq analysis showed that the expression of some genes involved in lipid metabolism was significantly changed (Supplementary Fig. S7a), and gene set enrichment analysis revealed that PncA increased the expression of genes mainly involved in the PPAR signaling pathway, fatty acid degradation, and other pathways that promote fat metabolism (Fig. 6b). Based on the metabolome analysis, PCA revealed significant differences between the PncA and vector groups (Supplementary Fig. S7b), and the volcano diagram demonstrated a large number of differential metabolites between the two groups (Supplementary Fig. S7c). The NA concentration in the liver of the PncA group was significantly increased (Fig. 6c), and the levels of other small molecules involved in nicotinate and nicotinamide metabolism were also significantly altered (Supplementary Fig. S7d). Metabolites that differed between the two groups were mainly enriched in the nicotinate and nicotinamide metabolic pathways, followed by phenylalanine, tyrosine, and tryptophan biosynthesis and other pathways related to lipid metabolism (Supplementary Fig. S7e). The raw metabolome analysis data are shown in Supplementary Data 1. To our surprise, the highly expressed genes in the PncA group were also significantly enriched in the Th1 and Th2 cell differentiation pathways, indicating that PncA genes may play an important role in T cell differentiation (Fig. 6d).Fig. 6Improvement in mouse liver lesions by PncA overexpression via AAV.a Representative Oil Red O staining (top) and H&E staining (bottom) of mouse liver sections. b Gene set enrichment analysis (GSEA) of RNA-seq results. c Heatmaps of metabolites with significant differences between the PncA and vector groups. d GSEA of RNA-seq results. Scale bars 100 µm.
To verify that direct NA supplementation in mice did not recapitulate the effect of expressing PncA in the liver, we treated mice with the MCD diet and NA. The body weight of mice decreased throughout the experiment (Supplementary Fig. S8a), and NA had no obvious effect on the NAD+ level in the liver (Supplementary Fig. S8b). Therefore, direct NA supplementation does not increase the level of mouse NAD+.
## Discussion
Coenzyme NAD+ plays a key role in cellular biology and adaptive stress responses. Its depletion is a basic feature of aging and may lead to various chronic diseases24. NAD+ supplementation alleviates several aging-related diseases and even prolongs the lifespan of mice25. Maintaining NAD+ level is essential for the function of high-energy-demanding cells and mature neurons26. NAD+ levels are substantially decreased in major neurodegenerative diseases, such as Alzheimer’s disease, Parkinson’s disease, and muscle atrophy27. Increasing evidence has shown that NAD+ is significantly reduced in various tissues during aging, and determining how to efficiently increase cellular NAD+ levels through physiological and pharmacological methods and prevent age-related diseases has become a hot research topic in recent years. Similarly, identifying new and efficient methods to promote NAD+ synthesis is also vital.
Currently, the most efficient NAD+ synthesis pathway in mammals is a controversial issue. In addition, different organs depend on different NAD+ precursors23. Studies have reported that the liver and kidneys use all three NAD+ synthesis pathways; the spleen, small intestine, and pancreas mainly rely on the salvage of NA and NAM, while the heart, lungs, brain, muscles, and white adipose tissue primarily use the NAM pathway28. Regarding the most efficient precursor, some articles have reported that the efficiency of NAD+ synthesis with NA is higher than that with NAM, but there is still no clear conclusion29. However, the most widely used NAD+ precursors are NR and NMN. After these two precursors enter the cell, NAD+ is synthesized through one or two enzymatic reactions, and it avoids being affected by rate-limiting enzymes, such as nicotinate phosphoribosyltransferase (NAPRT) and nicotinamide phosphoribosyltransferase (NAMPT). These two precursors have been used for various interventions, such as alleviating neurodegenerative diseases, improving hearing, treating diabetes and NAFLD, delaying aging, and extending lifespan30–34.
Cells have different absorption efficiencies for various NAD+ precursors. NAM directly enters the cell through free diffusion, whereas NA, NR, NMN, and other precursors require the assistance of membrane proteins, resulting in a lower absorption efficiency compared with NAM35. However, NAM is believed to be present in sufficient amounts in the body because NAD+ is decomposed to produce NAM, which re-enters the NAD+ synthesis pathway. In addition, NAM is an inhibitor of the SIRTs family. Excessive concentrations of NAM inhibit the activity of SIRT2 and shorten the lifespan of Saccharomyces cerevisiae36. However, PncA overexpression in Drosophila protects neurons and extends their lifespan37. PncA also increases the lifespan of Caenorhabditis elegans38. In this study, we used liver-specific AAV to express PncA in the liver, an organ that relies on multiple NAD+ precursors. In this way, we replenished the liver with the nicotinamidase that was discarded during evolution. The results showed that the Preiss–Handler pathway was highly active in mammalian cells, and this approach increased the level of NAD+ to a greater extent than NR or NMN, which indicated the high efficiency of NA for NAD+ synthesis.
Furthermore, the efficiency of using NA for NAD+ synthesis appears to be highest in the liver. In addition, PncA can process excessive NAM in the liver, thereby relieving the inhibition of SIRT1 by NAM and increasing the activity of SIRT1, which needs further verification. Based on the above results, we speculate that PncA also substantially increases NAD+ levels in other organs that rely on NA for NAD+ synthesis. Moreover, we demonstrated that PncA significantly improved NAFLD in mice, indicating that PncA provides a promising potential target for treating various diseases related to NAD+ deficiency. There are some limitations in our work; germ-free mice would be a more ideal research model for bacteria supplementary experiment, and a constitutive bacterial expression vector would be a high-efficiency method for protein overexpression in vivo which may explain why we did not observe the increase of NA in feces. Besides, further research should be done to evaluate the potential side effects caused by excessive elevation of NAD+ after PncA overexpression.
## Materials
NAM (HY-B0150), NA (HY-B0143), and NR (HY-123033) were obtained from MedChemExpress (Monmouth Junction, NJ, USA). PCN [243-369-5] was obtained from Sigma-Aldrich (St. Louis, MO, USA), and the MCD diet (TD.90262) was obtained from Harlan Teklad. Brain heart infusion (BHI) (HB8297-5), MRS (HB0384-1), and TPY (HB8570) were obtained from Hopebio-Technology (Qingdao, China).
## Cell lines
293T female embryonic kidney cells and HepG2 liver cancer cells were obtained from ATCC and cultured in Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with $10\%$ fetal bovine serum (FBS) and 100 mg/ml penicillin/streptomycin. Each cell line was maintained in a $5\%$ CO2 atmosphere at 37 °C. The mycoplasma contamination status of all cultures was monitored monthly by PCR.
## Bacterial strains
Akkermansia muciniphila (ATCC BAA-835) and L. salivarius (ATCC 1174) were cultured anaerobically in BHI and MRS medium, respectively. Bifidobacterium longum (1.2186) and S. gordonii (1.2496) were obtained from the China General Microbiological Culture Collection Center (Beijing, China) and cultured anaerobically in TPY and BHI medium, respectively.
## Mice
Wild-type C57BL/6J mice were purchased from Cyagen Biosciences Inc (Suzhou, China). Mice were maintained under a 12-h light/dark cycle and fed a standard chow diet in the specific pathogen-free facility at the Laboratory Animal Research Center, Tongji University. For bacterial and AAV infection, 8- to 10-week-old female C57BL/6J mice were used. All experiments were carried out following the national guidelines for the housing and care of laboratory animals (Ministry of Health, China), and the protocol complied with institutional regulations after review and approval by the Institutional Animal Ethics Committee at the Laboratory Animal Research Center, Tongji University (TJAA09220102). Mice were acclimatized for a period of 7 days before the initiation of the experiment.
## Mouse infection
Eight-week-old C57BL/6 female mice were fed with either regular water or autoclaved water containing an antibiotic cocktail (1 g/L ampicillin, 1 g/L neomycin, 1 g/L metronidazole, and 500 mg/L vancomycin) for 5 days and then given regular water. Escherichia coli strains were cultured as described above and quantified prior to infection. For bacterial colonization assays, the mice were infected intragastrically with 1 × 109 cfu PncA-OE or PncA-WT E. coli strains (in 0.2 ml PBS) every 3 days until the end of the experiment. For the group with NAD+ precursors, NAM (400 mg/kg/day) and NA 400 mg/kg/day were delivered via gavage. AAV (1 × 1012 pfu/ml in 100 μl PBS) expressing PncA (AAV-PncA) or vector (AAV-vector) were injected into mice via their tail vein. Animals were sacrificed 60 days after injection. The liver was removed and stored at −80 °C until use.
## MCD diet-induced NAFLD
Eight-week-old C57BL/6J female mice were used. After acclimatization, the mice were randomized based on their body weight and separated into three groups; two receiving the MCD diet and one receiving a normal control diet. For the bacterial experiment, the MCD diet was supplemented 30 days after the first bacterial colonization. For the AAV experiment, the MCD diet was supplemented 40 days after the AAV injection. Then, 2.5 weeks after the beginning of the special diets, animals fasted for 4 h and were anesthetized with isoflurane. Isolated tissues were snap-frozen in liquid nitrogen and stored at −80 °C for later experiments. The experiment was performed twice.
## Purification of PncA and enzyme activity assay
His-tag Protein Purification Kit (P2226) from Beyotime (Beijing, China) was used to purify PncA. PncA activity was determined by the ammonia production detected by the ammonia assay kit (MAK310) from Sigma-Aldrich. In a typical assay, 100 mM HEPES, pH 7.4, containing 500 μM nicotinamide and 37.5 nM PncA were combined at 27 °C. The addition of enzyme initiated the reaction, and ammonia was detected after 15 min. The amount of conversion of NAM by PncA was calculated by ammonia production. As for the bacterial enzyme assay, lysis of E. coli was used.
## PCN experiment
Eight-week-old C57BL/6 female mice were gavaged with either PBS or PCN (150 mg/kg) every day for 2 weeks. Fresh feces were collected rapidly on day 15 and stored at −80 °C until processing for DNA extraction and other processes. Animals were killed at the end of the experiment. Tissues were removed and stored at −80 °C until use. PCN concentration of 100 μg/ml was used for the cell and bacterial experiments.
## Construction of PncA-OE E. coli
For PncA-OE E. coli, the PCR product of a His-tagged PncA was cloned into the pET-28a vector to construct the expression plasmid, which was validated by Sanger sequencing. The plasmid pET-28a-PncA was transformed into E. coli BL21 (DE3), which was cultured in LB broth with kanamycin (50 µg/ml) at 37 °C and shaking at 220 rpm. Isopropyl β-D-1-thiogalactopyranoside (0.5 mM) was added to the LB broth to induce PncA expression during the logarithmic growth phase (OD600 ~0.6). The bacteria were harvested when OD600 was >1. Cells were pelleted by centrifuging for 10 min at 5000×g and resuspended in PBS. The bacteria were counted and stored in $30\%$ glycerol at −80 °C. The overexpression of PncA in E. coli was validated by qPCR and western blotting.
## Cell transfection
Transient transfection was performed using LipoFiter 3.0 (Hanbio Biotechnology Co. Ltd, Shanghai, China).
## Preparation of AAV
293T cells were used to package adenoviruses using a three-plasmid system, including pAAV-RC, pHelper, and shuttle plasmids (with or without the target gene). 293T cells were subcultured into 100-mm plates for transfection. Transfection was performed when the cell density reached 80–$90\%$ with the following transfection complex reagents: pAAV-RC 10 µg, pHelper 20 µg, shuttle plasmid 10 µg, and Lipofiter™ (HB-TRCF-1000, Hanbio Biotechnology Co. Ltd, Shanghai, China) 120 µl. Fresh complete medium containing $10\%$ FBS was replaced 6 h after transfection. Seventy-two hours after transfection, cells containing AAV particles were gently removed using a cell scraper, collected into a 15-ml centrifuge tube, and centrifuged at 150×g for 3 min. Cells were collected, and the culture supernatant was removed. Cells were washed with PBS once and resuspended with 300 μl PBS. The cells were frozen and thawed in liquid nitrogen and 37 °C three times and centrifuged at 2000×g for 5 min at 4 °C to remove cell debris. The lysis supernatant containing AAV particles was collected, and 0.1 μl Benonase (9025-65-4, Merck, Darmstadt, Germany) was added to each 1 ml of crude virus extract to remove the cell genome and plasmid DNA. The cells were centrifuged at 600×g for 10 min at 4 °C, and the supernatant was collected for column purification (V1469-01, Biomiga, San Diego, CA, USA). The 4-ml AAV liquid samples purified by the column were added to the ultrafiltration tube and centrifuged at 1400×g for 30 min to obtain approximately 1 ml of purified AAV, which was stored at −80 °C until use.
## Identification of homologs of NAD+-related genes in different bacteria
We selected some common mammalian microbial flora and pathogenic bacteria from different classifications and explored the genes associated with NAD+ synthesis in their genomes. Quinolinate synthetase (nadA) catalyzes the second step of the de novo biosynthetic pathway of pyridine nucleotide formation, which contains a nadA domain. A hidden Markov model (HMM) of nadA domain was downloaded from pfam39 and used to identify nadA homologous proteins using the hmmsearch function from the HMMer 3.1 package40 against the bacterial proteome that we selected. Matches with E ≤ 10−5 and annotation including quinolinate synthetase were recognized as candidates of nadA. BlastP (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE = Proteins) was used to screen the candidates using the threshold of E ≤ 10−5, and the proteins that passed these two thresholds were recognized as nadA. The identification of nadC, nadD, nadE, PncA, PncB, and Sirt2 was similar to that of nadA. However, nadB, nadR, and nadN had no signature domain, so the HMM method was not appropriate for identifying those genes. Protein sequences of those genes of E. coli were used as seeds submitted to BlastP against the bacteria proteome we selected; proteins with an E value less than 10−5 and appropriate annotation were recognized as corresponding genes.
## Untargeted and targeted vitamin B metabolomics and LC-MS/MS
For bacterial supernatants, the samples were thawed in an ice-water bath and vortexed for 30 s. A 100-μl aliquot of each individual sample was transferred to an Eppendorf tube. After adding 400 μl extract solvent (precooled at −40 °C, acetonitrile–water, 5:3, containing $0.125\%$ formic acid and isotopically labeled internal standard), the samples were vortexed for 30 s. The samples were sonicated for 15 min in an ice-water bath, followed by incubation at −20 °C for 1 h and centrifugation at 12,000 rpm (Relative centrifugal force (RCF) = 13,800×g) and 4 °C for 15 min. A 400-μl aliquot of the supernatant was evaporated to dryness under a gentle stream of nitrogen and reconstituted in 80 μl $1\%$ methanol/water (v/v). After the samples were centrifuged at 12,000 rpm (RCF = 13,800×g, $R = 8.6$ cm) for 15 min at 4 °C, the clear supernatant was subjected to LC-MS/MS analysis.
For mouse feces, an aliquot of each individual sample was weighed and transferred to an Eppendorf tube. After adding two small steel balls and 1000 μl extraction solution (precooled at −40 °C, $50\%$ acetonitrile containing $0.1\%$ formic acid and isotopically labeled internal standard), the samples were vortexed for 30 s, homogenized at 40 Hz for 4 min, and sonicated for 5 min in an ice-water bath. The homogenization and sonication cycles were repeated three times, followed by incubation at −20 °C for 1 h and centrifugation at 12,000 rpm (RCF = 13,800×g, $R = 8.6$ cm) for 15 min at 4 °C. An 800-μl aliquot of the supernatant was evaporated to dryness under a gentle stream of nitrogen and reconstituted in 80 μl $1\%$ methanol/water (v/v). After the samples were centrifuged at 12,000 rpm (RCF = 13,800×g, $R = 8.6$ cm) for 15 min at 4 °C, the clear supernatant was subjected to LC-MS/MS analysis.
For mouse liver, 50-mg samples were weighed and placed in an Eppendorf tube, and 1000 μl extract solution (methanol: acetonitrile: water = 2:2:1, with an isotopically labeled internal standard mixture) was added. The samples were homogenized at 35 Hz for 4 min and sonicated for 5 min in an ice-water bath. The homogenization and sonication cycles were repeated three times. The samples were incubated for 1 h at −40 °C and centrifuged at 12,000 rpm for 15 min at 4 °C. The resulting supernatant was transferred to a fresh glass vial for analysis.
LC-MS/MS analyses were performed using a UHPLC system (Vanquish; Thermo Fisher Scientific) with a UPLC BEH Amide column (2.1 mm × 100 mm, 1.7 μm) coupled to a Q Exactive HFX mass spectrometer (Orbitrap MS; Thermo Fisher Scientific). The mobile phase consisted of 25 mmol/L ammonium acetate and 25 mmol/L ammonia hydroxide in water (pH 9.75) (A) and acetonitrile (B). The auto-sampler temperature was 4 °C, and the injection volume was 2 μl. The QE HFX mass spectrometer was applied to acquire MS/MS spectra using the information-dependent acquisition mode with acquisition software (Xcalibur; Thermo Fisher Scientific). In this mode, the acquisition software continuously evaluated the full MS spectrum. The emergency severity index source conditions were set as follows: sheath gas flow rate 30 Arb, Aux gas flow rate 25 Arb, capillary temperature 350 °C, full MS resolution 60,000, MS/MS resolution = 7500, collision energy $\frac{10}{30}$/60 in normalized collisional energy mode, and spray voltage 3.6 kV (positive) or −3.2 kV (negative).
The raw data were converted to the mzXML format using ProteoWizard and processed with an in-house program (developed using R and based on XCMS) for peak detection, extraction, alignment, and integration. Then, an in-house MS2 database (BiotreeDB) was applied for metabolite annotation. The cutoff for annotation was set at 0.3.
## RNA isolation, RNA-seq, and data processing
Total RNA was extracted from tissues using Trizol reagent (Invitrogen, Carlsbad, CA, USA). Oligo (dT)-attached magnetic beads were used to purify mRNA. Purified mRNA was fragmented into small pieces with fragment buffer at the appropriate temperature. Then, first-strand cDNA was generated using random hexamer-primed reverse transcription, followed by second-strand cDNA synthesis. Afterward, A-Tailing Mix and RNA Index Adapters were added by incubating to end repair. The cDNA fragments obtained in the previous step were amplified by PCR, and products were purified using Ampure XP beads and then dissolved in an elution buffer solution. The product was validated on the Agilent Technologies 2100 bioanalyzer for quality control. The double-stranded PCR products from the previous step were heated, denatured, and circularized by the splint oligo sequence to obtain the final library. Single-strand circular DNA was formatted as the final library. The final library was amplified with phi29 to generate DNA nanoballs containing more than 300 copies of one molecule. DNA nanoballs were loaded into the patterned nanoarray, and paired-end 50-base-pair reads were generated on the BGIseq500 platform (BGI, Shenzhen, China).
The sequencing data were filtered with SOAPnuke (v1.5.2)41 by: [1] removing reads containing sequencing adapters, [2] removing reads with a low-quality base ratio (base quality ≤ 5) >20, and [3] removing reads with an unknown base (N’ base) ratio >$5\%$. Clean reads were obtained and stored in FASTQ format and were mapped to the reference genome using HISAT2 (v2.0.4)42. Bowtie2 (v2.2.5)43 was applied to align the clean reads to the reference coding gene set, and then the gene expression level was calculated by RSEM (v1.2.12)44. The heatmap was drawn by pheatmap (v1.0.8) according to the gene expression in different samples. Differential expression analysis was performed using the DESeq2(v1.4.5)45 with Q < 0.05. To gain insight into phenotypic changes, GO (http://www.geneontology.org/) and KEGG (https://www.kegg.jp/) enrichment analysis of annotated differentially expressed genes were performed by Phyper (https://en.wikipedia.org/wiki/Hypergeometric distribution) based on the hypergeometric test. The significant level of terms and pathways were corrected using the Q value with a rigorous threshold (<0.05) by the Bonferroni test.
## Reverse transcription-quantitative PCR
RNA was treated with DNase, and 1 μg RNA was used for reverse transcription. cDNA diluted 10× was used for reverse transcription-quantitative PCR (RT-qPCR). RT-qPCR was performed using the Light-Cycler system (Roche Diagnostics GmbH, Rotkreuz, Switzerland) and a qPCR Supermix (Vzayme, Nanjing, China) with the indicated primers. An average of at least three technical repeats was used for each biological data point. The ropB gene was used as the reference gene, and the following primers were used: ropB-F, CTGCGCGAAGAAATCGAAGG, ropB-R: TTTCGCCAACGGAACGGATA and PncA-F: TGATCGCCAGCCAAGACT, PncA-R: AGCATCCAGCACCGTGAA.
## Western blotting
Bacteria (109 cfu) were lysed in 200 μl lysis buffer (from the protein purification kit). Samples of 20 μl were loaded onto $12.5\%$ Bis-Tris polyacrylamide gels and transferred onto PVDF membrane (Millipore, Billerica, MA, USA) by electroblotting. The membranes were blocked in Tris-buffered saline and $0.5\%$ Tween containing $5\%$ skimmed milk powder (OXOID, Basingstoke, UK) for 1 h at room temperature and incubated with primary antibody, anti-his (M20001; Abmart, Shanghai, China) with 1:1000 dilution (500 µg/ml) overnight at 4 °C. The membranes were incubated with peroxidase-conjugated secondary antibody (Bio-Rad, Hercules, CA, USA) for 1 h at room temperature. The bands were visualized using Millipore’s enhanced chemiluminescence with the Amersham Imager 600 detection system (GE, Boston, USA).
## Genomic DNA extraction
Microbial community DNA was extracted using a MagPure Stool DNA KF kit B (Magen, Guangzhou, China). DNA was quantified with a Qubit Fluorometer using a Qubit dsDNA BR Assay kit (Invitrogen), and the quality was assessed by running an aliquot on a $1\%$ agarose gel.
## Library construction
Variable regions V3–V4 of the bacterial 16S rRNA gene were amplified with the degenerate PCR primers 341F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′- GGACTACHVGGGTWTCTAAT-3′). Both forward and reverse primers were tagged with Illumina adapter, pad, and linker sequences. PCR enrichment was performed in a 50-µl reaction containing 30 ng template, fusion PCR primer, and PCR master mix. PCR cycling conditions were as follows: 94 °C for 3 min, 30 cycles of 94 °C for 30 s, 56 °C for 45 s, 72 °C for 45 s, and a final extension for 10 min at 72 °C. The PCR products were purified with Ampure XP beads and eluted in Elution buffer. Libraries were qualified with the Agilent 2100 bioanalyzer (Agilent, Santa Clara, CA, USA). The validated libraries were used for sequencing on an Illumina MiSeq platform (BGI) following the standard pipelines of Illumina, and 2 × 300 bp paired-end reads were generated.
## Data processing
First, low-quality data were removed from the original sequencing data by the window method with Readfq v8 (https://github.com/cjfields/readfq). Joint pollution reads and n-containing reads were removed, then low-complexity reads were processed. Samples were distinguished based on barcode and primer. FLASH software46 (fast length adjustment of short reads, v1.2.11) was used for assembling. Using overlapping relationships, pairs of double-end sequencing reads were assembled into a sequence with high area tags. Effective tags were produced by the UCHIME algorithm and clustered into operational taxonomic units (OTUs) using USEARCH47 (v7.0.1090) software. According to the mothur method and Greengenes database, taxonomic information was annotated with representative sequences from OTUs. Phy tools48 and R software (v3.4.1) were used to perform an unweighted pair group method with arithmetic mean clustering analysis based on Bray–Curtis weighted Unifrac and unweighted Unifrac distance matrices. The ade449 package of R (v3.4.1) was used to perform an OTU PCA analysis. The RDP classifier Bayesian algorithm was used to classify the OTU representative sequences. The community composition of individual samples was counted at the species level of the phylum, order, family, and genus, and the histogram of species abundance was performed using the ggplot2 package of R. Alpha diversity statistics were analyzed using the software motor (v1.31.2). Beta diversity was analyzed by QIIME (v1.80)50. Finally, the ggplot2 package of R was used for box plots of alpha and beta diversity.
## Oil Red staining and H&E staining
Hepatic tissue was cut into small pieces, fixed in $4\%$ paraformaldehyde for 4 h, and embedded in OCT (Leica Camera AG, Wetzlar, Germany). Frozen sections (8 µm thickness) were made using a cryostat, and the samples were fixed with $4\%$ paraformaldehyde for an additional 30 min. The slides were washed in distilled water and stained with Oil Red O for 15 min. Next, the slides were counterstained with hematoxylin for 10 s to identify the nuclei. For H&E staining, the slides were first stained in hematoxylin for 3–8 min and then counterstained with eosin for 1–3 min. The histological images were acquired with a light microscope (Olympus, Tokyo, Japan).
## NAD+ detection
NAD+ detection was carried out using the EnzyChrom TM NAD+/NADH+ Assay Kit (E2ND-100) from BioAssay Systems (Hayward, CA, USA). Protein concentration was used to normalize the NAD+ content.
## ATP detection
For the measurement of ATP level, 100-mg liver samples were lysed in 1 ml lysis buffer provided by the ATP Assay Kit (S0026) from Beyotime (Jiangsu, China). Liver ATP levels were evaluated by luciferase activity, as shown in the standard protocol provided by the ATP Assay Kit.
## Triglyceride detection
Liver triglycerides were assayed using a triglyceride assay kit (E1025; Applygen Technologies, Beijing, China).
## Statistics and reproducibility
All data were analyzed using GraphPad Prism 8 (Graphpad Software Inc.). Differences between the two groups were evaluated using Student’s t-test with a two-tailed distribution. All data are presented as the mean ± standard deviation (SD) from at least three independent experiments performed in triplicate. Sample sizes were selected without performing statistical tests. No data were excluded when conducting the final statistical analysis. * $p \leq 0.05$, **p ≤ 0.01, ***p ≤ 0.001, and ****$p \leq 0.0001$ unless stated otherwise.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Peer Review File Supplementary Information-New Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Reporting summary The online version contains supplementary material available at 10.1038/s42003-023-04613-8.
## Peer review information
Communications Biology thanks Leonardo Sorci and the other anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Kevin Theis and Zhijuan Qiu. Peer reviewer reports are available.
## References
1. Harden A, Young WJ, Martin CJ. **The alcoholic ferment of yeast-juice. Part II—The coferment of yeast-juice**. *Proc. R. Soc. Lond.* (1906) **78** 369-375
2. Camacho-Pereira J. **CD38 dictates age-related NAD decline and mitochondrial dysfunction through an SIRT3-dependent mechanism**. *Cell Metab.* (2016) **23** 1127-1139. DOI: 10.1016/j.cmet.2016.05.006
3. Imai S, Guarente L. **NAD**. *Trends Cell Biol.* (2014) **24** 464-471. DOI: 10.1016/j.tcb.2014.04.002
4. Kim MY, Mauro S, Gevry N, Lis JT, Kraus WL. **NAD**. *Cell* (2004) **119** 803-814. DOI: 10.1016/j.cell.2004.11.002
5. Tarragó MG. **A potent and specific CD38 inhibitor ameliorates age-related metabolic dysfunction by reversing tissue NAD**. *Cell Metab.* (2018) **27** 1081-1095.e1010. DOI: 10.1016/j.cmet.2018.03.016
6. Bird JG. **The mechanism of RNA 5′ capping with NAD**. *Nature* (2016) **535** 444-447. DOI: 10.1038/nature18622
7. Walters RW. **Identification of NAD**. *Proc. Natl Acad. Sci. USA* (2017) **114** 480-485. DOI: 10.1073/pnas.1619369114
8. Ansari HR, Raghava GP. **Identification of NAD interacting residues in proteins**. *BMC Bioinformatics* (2010) **11** 160. DOI: 10.1186/1471-2105-11-160
9. Rongvaux A, Andris F, Van Gool F, Leo O. **Reconstructing eukaryotic NAD metabolism**. *Bioessays* (2003) **25** 683-690. DOI: 10.1002/bies.10297
10. Rajman L, Chwalek K, Sinclair DA. **Therapeutic potential of NAD-boosting molecules: the in vivo evidence**. *Cell Metab.* (2018) **27** 529-547. DOI: 10.1016/j.cmet.2018.02.011
11. Zhang H. **Characterization of**. *FEBS J.* (2008) **275** 753-762. DOI: 10.1111/j.1742-4658.2007.06241.x
12. Shats I. **Bacteria boost mammalian host NAD metabolism by engaging the deamidated biosynthesis pathway**. *Cell Metab.* (2020) **31** 564-579.e567. DOI: 10.1016/j.cmet.2020.02.001
13. Steele MA, Des Prez RM. **The role of pyrazinamide in tuberculosis chemotherapy**. *Chest* (1988) **94** 845-850. DOI: 10.1378/chest.94.4.845
14. Younossi ZM. **Non-alcoholic fatty liver disease: a global public health perspective**. *J. Hepatol.* (2019) **70** 531-544. DOI: 10.1016/j.jhep.2018.10.033
15. Cai J, Zhang XJ, Li H. **Progress and challenges in the prevention and control of nonalcoholic fatty liver disease**. *Med. Res. Rev.* (2019) **39** 328-348. DOI: 10.1002/med.21515
16. Katsyuba E. **De novo NAD**. *Nature* (2018) **563** 354-359. DOI: 10.1038/s41586-018-0645-6
17. Pham TX. **Nicotinamide riboside, an NAD**. *Biochim. Biophys. Acta Mol. Basis Dis.* (2019) **1865** 2451-2463. DOI: 10.1016/j.bbadis.2019.06.009
18. Yoshino J, Mills KF, Yoon MJ, Imai S. **Nicotinamide mononucleotide, a key NAD**. *Cell Metab.* (2011) **14** 528-536. DOI: 10.1016/j.cmet.2011.08.014
19. Lau LHS, Wong SH. **Microbiota, obesity and NAFLD**. *Adv. Exp. Med. Biol.* (2018) **1061** 111-125. DOI: 10.1007/978-981-10-8684-7_9
20. Seiner DR, Hegde SS, Blanchard JS. **Kinetics and inhibition of nicotinamidase from**. *Biochemistry* (2010) **49** 9613-9619. DOI: 10.1021/bi1011157
21. Hara N. **Elevation of cellular NAD levels by nicotinic acid and involvement of nicotinic acid phosphoribosyltransferase in human cells**. *J. Biol. Chem.* (2007) **282** 24574-24582. DOI: 10.1074/jbc.M610357200
22. Canto C. **The NAD**. *Cell Metab.* (2012) **15** 838-847. DOI: 10.1016/j.cmet.2012.04.022
23. Zou Y. **Illuminating NAD**. *Dev. Cell.* (2020) **53** 240-252.e247. DOI: 10.1016/j.devcel.2020.02.017
24. Fang EF. **NAD**. *Trends Mol. Med.* (2017) **23** 899-916. DOI: 10.1016/j.molmed.2017.08.001
25. Zhang H. **NAD**. *Science* (2016) **352** 1436-1443. DOI: 10.1126/science.aaf2693
26. Lautrup S, Sinclair DA, Mattson MP, Fang EF. **NAD**. *Cell Metab.* (2019) **30** 630-655. DOI: 10.1016/j.cmet.2019.09.001
27. Verdin E. **NAD**. *Science* (2015) **350** 1208-1213. DOI: 10.1126/science.aac4854
28. Liu L. **Quantitative analysis of NAD synthesis-breakdown fluxes**. *Cell Metab.* (2018) **27** 1067-1080.e1065. DOI: 10.1016/j.cmet.2018.03.018
29. Mori V. **Metabolic profiling of alternative NAD biosynthetic routes in mouse tissues**. *PLoS ONE* (2014) **9** e113939. DOI: 10.1371/journal.pone.0113939
30. Diguet N. **Nicotinamide riboside preserves cardiac function in a mouse model of dilated cardiomyopathy**. *Circulation* (2018) **137** 2256-2273. DOI: 10.1161/CIRCULATIONAHA.116.026099
31. Gong B. **Nicotinamide riboside restores cognition through an upregulation of proliferator-activated receptor-γ coactivator 1α regulated β-secretase 1 degradation and mitochondrial gene expression in Alzheimer’s mouse models**. *Neurobiol. Aging* (2013) **34** 1581-1588. DOI: 10.1016/j.neurobiolaging.2012.12.005
32. Lee HJ, Yang SJ. **Supplementation with nicotinamide riboside reduces brain inflammation and improves cognitive function in diabetic mice**. *Int. J. Mol. Sci.* (2019) **20** 4196. DOI: 10.3390/ijms20174196
33. Mills KF. **Long-term administration of nicotinamide mononucleotide mitigates age-associated physiological decline in mice**. *Cell Metab.* (2016) **24** 795-806. DOI: 10.1016/j.cmet.2016.09.013
34. Wang X, Hu X, Yang Y, Takata T, Sakurai T. **Nicotinamide mononucleotide protects against β-amyloid oligomer-induced cognitive impairment and neuronal death**. *Brain Res.* (2016) **1643** 1-9. DOI: 10.1016/j.brainres.2016.04.060
35. Covarrubias AJ, Perrone R, Grozio A, Verdin E. **NAD**. *Nat. Rev. Mol. Cell Biol.* (2021) **22** 119-141. DOI: 10.1038/s41580-020-00313-x
36. Bitterman KJ, Anderson RM, Cohen HY, Latorre-Esteves M, Sinclair DA. **Inhibition of silencing and accelerated aging by nicotinamide, a putative negative regulator of yeast sir2 and human SIRT1**. *J. Biol. Chem.* (2002) **277** 45099-45107. DOI: 10.1074/jbc.M205670200
37. Balan V. **Life span extension and neuronal cell protection by**. *J. Biol. Chem.* (2008) **283** 27810-27819. DOI: 10.1074/jbc.M804681200
38. van der Horst A, Schavemaker JM, Pellis-van Berkel W, Burgering BM. **The**. *Mech. Ageing Dev.* (2007) **128** 346-349. DOI: 10.1016/j.mad.2007.01.004
39. Mistry J. **Pfam: the protein families database in 2021**. *Nucleic Acids Res.* (2021) **49** D412-d419. DOI: 10.1093/nar/gkaa913
40. Potter SC. **HMMER web server: 2018 update**. *Nucleic Acids Res.* (2018) **46** W200-w204. DOI: 10.1093/nar/gky448
41. Li R, Li Y, Kristiansen K, Wang J. **SOAP: short oligonucleotide alignment program**. *Bioinformatics* (2008) **24** 713-714. DOI: 10.1093/bioinformatics/btn025
42. Kim D, Langmead B, Salzberg SL. **HISAT: a fast spliced aligner with low memory requirements**. *Nat. Methods* (2015) **12** 357-360. DOI: 10.1038/nmeth.3317
43. Langmead B, Salzberg SL. **Fast gapped-read alignment with Bowtie 2**. *Nat. Methods* (2012) **9** 357-359. DOI: 10.1038/nmeth.1923
44. Li B, Dewey CN. **RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome**. *BMC Bioinformatics* (2011) **12** 323. DOI: 10.1186/1471-2105-12-323
45. 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
46. Magoč T, Salzberg SL. **FLASH: fast length adjustment of short reads to improve genome assemblies**. *Bioinformatics* (2011) **27** 2957-2963. DOI: 10.1093/bioinformatics/btr507
47. Edgar RC. **UPARSE: highly accurate OTU sequences from microbial amplicon reads**. *Nat. Methods* (2013) **10** 996-998. DOI: 10.1038/nmeth.2604
48. Revell LJ. **phytools: an R package for phylogenetic comparative biology (and other things)**. *Methods Ecol. Evol.* (2012) **3** 217-223. DOI: 10.1111/j.2041-210X.2011.00169.x
49. Dray S, Dufour A-B. **The ade4 Package: implementing the duality diagram for ecologists**. *J. Stat. Softw.* (2007) **22** 1-20. DOI: 10.18637/jss.v022.i04
50. Caporaso JG. **QIIME allows analysis of high-throughput community sequencing data**. *Nat. Methods* (2010) **7** 335-336. DOI: 10.1038/nmeth.f.303
|
---
title: Retinol dehydrogenase 10 reduction mediated retinol metabolism disorder promotes
diabetic cardiomyopathy in male mice
authors:
- Yandi Wu
- Tongsheng Huang
- Xinghui Li
- Conghui Shen
- Honglin Ren
- Haiping Wang
- Teng Wu
- Xinlu Fu
- Shijie Deng
- Ziqi Feng
- Shijie Xiong
- Hui Li
- Saifei Gao
- Zhenyu Yang
- Fei Gao
- Lele Dong
- Jianding Cheng
- Weibin Cai
journal: Nature Communications
year: 2023
pmcid: PMC9981688
doi: 10.1038/s41467-023-36837-x
license: CC BY 4.0
---
# Retinol dehydrogenase 10 reduction mediated retinol metabolism disorder promotes diabetic cardiomyopathy in male mice
## Abstract
Diabetic cardiomyopathy is a primary myocardial injury induced by diabetes with complex pathogenesis. In this study, we identify disordered cardiac retinol metabolism in type 2 diabetic male mice and patients characterized by retinol overload, all-trans retinoic acid deficiency. By supplementing type 2 diabetic male mice with retinol or all-trans retinoic acid, we demonstrate that both cardiac retinol overload and all-trans retinoic acid deficiency promote diabetic cardiomyopathy. Mechanistically, by constructing cardiomyocyte-specific conditional retinol dehydrogenase 10-knockout male mice and overexpressing retinol dehydrogenase 10 in male type 2 diabetic mice via adeno-associated virus, we verify that the reduction in cardiac retinol dehydrogenase 10 is the initiating factor for cardiac retinol metabolism disorder and results in diabetic cardiomyopathy through lipotoxicity and ferroptosis. Therefore, we suggest that the reduction of cardiac retinol dehydrogenase 10 and its mediated disorder of cardiac retinol metabolism is a new mechanism underlying diabetic cardiomyopathy.
The current challenges for diabetic cardiomyopathy (DCM) are unclear mechanisms and no effective therapy in clinics. Here, the authors found that the decrease of cardiac retinol dehydrogenase 10 in type 2 diabetes leads to retinol metabolism disorder, cardiac lipid toxicity and cardiomyopathy development, suggesting that correcting the imbalance of cardiac retinol metabolism may be an effective strategy for the treatment of DCM.
## Introduction
Diabetes mellitus (DM) is an independent risk factor for cardiovascular diseases1 and can independently induce structural and functional disruptions in the heart, leading to diabetic cardiomyopathy (DCM)1. With the application of new therapies, most complications of DM have been effectively controlled, and the life expectancy of DM patients has been extended, while hidden myocardial injury, which causes 50–$80\%$ of deaths, has become the leading cause of death in DM patients2–4. Mechanistic studies are key to improving the prevention and treatment of DCM. DCM has complex pathogenesis in which lipotoxicity and myocardial cell death should not be underestimated5,6.
Retinol (vitamin A, Rol) and all-trans retinoic acid (atRA) are metabolites of retinol metabolism that have been shown to have altered levels in diseases7,8. The study demonstrated that retinoic acid receptors (RARs) were reduced in the hearts of diabetic rats and activating these receptors by atRA or other activators prevented myocardial injury9; however, it remains unknown whether cardiac retinol metabolite levels are altered and whether these alterations are involved in DCM.
In retinol metabolism, Rol, the metabolic substrate, undergoes a two-step dehydrogenation reaction to generate the active metabolite atRA, which then primarily binds and activates RARs (ligand-induced transcription factors) to exert biological effects10. Retinol dehydrogenase 10 (RDH10) is the rate-limiting enzyme in retinol metabolism that has an important role in the conversion of Rol to atRA and affects organ development during embryonic development11,12, but the role of RDH10 in retinol metabolism and disease in adulthood, particularly in the heart, remains poorly understood.
In this study, we demonstrated that in type 2 diabetes mellitus (T2DM), impaired cardiac retinol metabolism caused by cardiac RDH10 reduction results in DCM through Rol overload-induced cardiotoxicity and atRA deficiency-induced lipotoxicity and ferroptosis.
## Retinol metabolism disorder in the hearts of mice and patients with T2DM
Our preliminary study showed that the db/db mouse is not only a T2DM model but also a good DCM model with a typical pattern of heart function changes from a compensatory stage to a decompensatory stage6. We selected db/db mice at 4, 24, and 32 weeks of age and performed RNA-seq analysis of their hearts. The selected mice were identified as prehyperglycemic, hyperglycemic with compensated heart function, and hyperglycemic with decompensated heart function (Supplementary Fig. 1).
There were 1695, 2584, and 2147 differentially expressed genes (DEGs, fold change ≥1.5) identified in the hearts of db/db mice at 4, 24, and 32 weeks of age, respectively, of which 643, 1293, and 1153 were upregulated while 1052, 1291 and 994 were downregulated (Fig. 1a). A total of 215 DEGs were changed in all 3 groups (Fig. 1b) and were significantly enriched ($p \leq 0.001$, q < 0.001) in 14 pathways classified by Kyoto Encyclopedia of Genes and Genomes (KEGG), with metabolic pathways (mmu01100) containing the most genes (Fig. 1c). Gene ontology (GO) classification of the DEGs in metabolic pathways was then performed and found that most of these DEGs were significantly enriched ($p \leq 0.001$, q < 0.001) in the response to vitamin A (GO:0033189), retinoic acid metabolic process (GO:0042573) and retinol metabolic process (GO:0042572) (Fig. 1d).Fig. 1Altered cardiac retinol metabolism in type 2 diabetes mellitus (T2DM) (n means biologically independent animals).a Summary of differentially expressed genes (DEGs). b Venn diagram of DEGs. c Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of DEGs. d Gene ontology (GO) terms analysis of DEGs in metabolic pathways. e Heat map of DEGs in retinol metabolism pathway. f Cardiac retinol (Rol) and all-trans retinoic acid (atRA) levels in db/db mice, $$n = 3$$, * vs db/m. g Western blotting (WB) of cardiac (retinoic acid receptors) RARs, $$n = 3$$, * vs db/m. h Cardiac Immunohistochemistry (IHC) staining of T2DM patients, n (retinoic acid receptor a, RARa) = 5 (T2DM patient and healthy population), n (retinoic acid receptor b, RARb) = 7 (T2DM patient and healthy population) and n (retinoic acid receptor g, RARg) = 8 (T2DM patient and healthy population), * vs healthy population. Data are expressed as means ± SD. Two-tailed unpaired t-test was used for the analysis of statistical significance. Source data are provided as a Source Data file. ( Black arrows: Typical IHC stained positive cells).
Further analysis of DEGs in retinol metabolism (mmu00830) showed that retinoic acid synthesis-related genes, ALDH1A213 and ALDH1A714,15, were gradually downregulated, while retinoic acid degradation-related genes, CYP3A1116, CYP26B113,17, UGT1A6B18 and UGT1A6A18,19, were gradually upregulated (Fig. 1e). More importantly, we found a nearly 1-fold increase in Rol, a halving of atRA, and significant decreases in retinoic acid receptor α (RARa) and retinoic acid receptor β (RARb) in the hearts of 32-week-old db/db mice (Fig. 1f and g), indicating a retinol metabolism disorder characterized by Rol overload, atRA deficiency, and RARs reduction was present in the hearts of T2DM mice. The reduction of RARa and RARb was also found in the hearts of T2DM patients, suggesting the presence of cardiac retinol metabolism disorder in T2DM patients (Fig. 1h).
## Rol aggravated cardiac Rol overload and exacerbated myocardial injury by promoting myocardial fibrosis and apoptosis in T2DM mice
To demonstrate the role of Rol overload in DCM, we supplemented db/db mice with Rol. We defined an experimental baseline and started Rol supplementation at 8 weeks of age when db/db mice developed stable hyperglycemia and measured body weight, blood glucose, and heart function over the following period until week 24 after baseline, when Rol-supplemented db/db mice developed heart failure manifested by a significant decrease in left ventricular ejection fraction (LVEF) and left ventricular fractional shortening (LVFS) (Fig. 2a–c). Cardiac Rol and atRA levels and cardiac structural remodeling were also measured at week 24 and the results showed that in db/db mice, Rol significantly increased cardiac Rol but not atRA (Fig. 2d) or retinyl ester levels (Supplementary Fig. 2) and exacerbated cardiac structural remodeling including myocardial fibrosis, apoptosis, myofibrillar fragmentation and increased mitochondria in cardiomyocytes (Fig. 2e–g). Additionally, we found that Rol reduced hyperglycemia, hyperinsulinemia, insulin resistance, and to some extent dyslipidemia in db/db mice (Fig. 2h–l), suggesting that the effects of Rol on myocardial injury in db/db mice were not by exacerbating diabetes. At the same time, considering the increase in cardiac Rol levels, we suggest that it is the overload of cardiac Rol that promotes myocardial injury in T2DM mice. Fig. 2Cardiac retinol metabolic status, structure, and function as well as serum insulin and lipids in T2DM mice supplemented with Rol (n means biologically independent animals).a Schematic diagram of Rol supplementary. b Echocardiography. c left ventricular ejection fraction (LVEF) and left ventricular fractional shortening (LVFS), n (baseline) = 5 (db/m), 6 (db/db) and 7 (db/db + Rol); n (4w) = 5 (db/m), 5 (db/db) and 5 (db/db + Rol); n (16w) = 5 (db/m), 6 (db/db) and 5 (db/db + Rol); n (24w) = 3 (db/m), 3 (db/db) and 5 (db/db + Rol); * db/db + Rol vs db/m; # db/db + Rol vs db/db. d Levels of cardiac Rol and atRA, $$n = 3$$, * vs db/m. e Heart image and heart/tibia ratio of db/db mice with Rol, $$n = 5$$ (db/m), 5 (db/db) and 4 (db/db + Rol), * vs db/m. f Cardiac Wheat germ agglutinin (WGA), Masson, Terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL), and Transmission electron microscopy (TEM) staining. g Analysis of f, n (cardiomyocytes area) = 4 (db/m), 5 (db/db) and 5 (db/db + Rol), n (collagen area and Tunel+ cell) = 3, * vs db/m; # vs db/db. h Body weight, n (baseline) = 10 (db/m), 10 (db/db) and 10 (db/db + Rol); n (4w) = 10 (db/m), 8 (db/db) and 9 (db/db + Rol); n (16w) = 10 (db/m), 8 (db/db) and 6 (db/db + Rol); n (24w) = 10 (db/m), 4 (db/db) and 6 (db/db + Rol); * vs db/m. i Blood glucose, n (baseline)=5 (db/m), 10 (db/db) and 10 (db/db + Rol); n (4w) = 5 (db/m), 6 (db/db) and 7 (db/db + Rol); n (16w) = 5 (db/m), 6 (db/db) and 5 (db/db + Rol); n (24w) = 5 (db/m), 5 (db/db) and 4 (db/db + Rol); * vs db/m; # vs db/db. j Serum insulin and homeostasis model assessment of the insulin resistance index (HOMA-IR), $$n = 3$$ (db/m), 5 (db/db) and 4 (db/db + Rol); * vs db/m; # vs db/db. k Insulin tolerance test (ITT), $$n = 4$$ (db/m), 7 (db/db) and 7 (db/db + Rol); * vs db/m; # vs db/db. l Serum lipids, n (triglyceride, triacylglycerol, triacylglyceride, TG and total cholesterol, TC) = 6 (db/m), 5 (db/db) and 6 (db/db + Rol), n (high density liptein cholesterol, HDL) = 7 (db/m), 4 (db/db) and 6 (db/db + Rol), n (low-density lipoprotein, LDL) = 5 (db/m), 4 (db/db) and 5 (db/db + Rol); * vs db/m; # vs db/db. Data are expressed as means ± SD. Two-tailed unpaired t-test was used for the analysis of statistical significance in d while one-way ANOVA with Tukey post hoc test was used for the analysis of statistical significance in c, e, g h, j, k and l. Source data are provided as a Source Data file. ( Black arrows: Typical Tunel stained positive cells; white boxes: lipid droplets).
## atRA restored cardiac atRA levels and alleviated myocardial injury in T2DM mice
To demonstrate the role of atRA deficiency in DCM, we supplemented atRA with db/db mice. We defined the experimental baseline and started atRA supplementation in db/db mice at 8 weeks of age and measured body weight, blood glucose, and heart function over the following period until week 28 after baseline when db/db mice without atRA supplementation developed heart failure (Fig. 3a–c). Cardiac atRA levels and structural remodeling were also measured at week 28 after baseline and the results showed that in db/db mice, atRA supplementation restored cardiac atRA levels (Fig. 3d) and attenuated cardiac structural remodeling including myocardial hypertrophy, myocardial fibrosis, apoptosis, lipid deposition and increased mitochondria in cardiomyocytes (Fig. 3e–g). Additionally, atRA reduced hyperinsulinemia, insulin resistance, and to some extent dyslipidemia in db/db mice (Fig. 3h–l).Fig. 3Cardiac retinol metabolic status, structure, and function as well as serum insulin and lipids in T2DM mice supplemented with atRA (n means biologically independent animals).a Schematic diagram of atRA supplementary. b Echocardiography. c LVEF and LVFS, n (baseline) = 3 (db/m), 3 (db/db) and 5 (db/db + atRA); n (4w) = 3 (db/m), 3 (db/db) and 5 (db/db + atRA); n (16w) = 3 (db/m), 4 (db/db) and 3 (db/db + atRA); n (24w) = 5 (db/m), 3 (db/db) and 3 (db/db + atRA); n (28w) = 4 (db/m), 3 (db/db) and 3 (db/db + atRA); * vs db/m; # vs db/db. d Levels of cardiac atRA, $$n = 3$$, * vs db/m; # vs db/db. e Heart image and heart/tibia ratio, $$n = 5$$ (db/m and db/db) and 4 (db/db + atRA), * vs db/m, # vs db/db. f Cardiac WGA, Masson, Tunel, and TEM staining. g Analysis of f, $$n = 3$$, * vs db/m; # vs db/db. h Body weight, n (baseline) = 5 (db/m), 5 (db/db) and 5 (db/db + atRA); n (4w) = 5 (db/m), 5 (db/db) and 5 (db/db + atRA); n (16w) = 4 (db/m), 5 (db/db) and 4 (db/db + atRA); n (24w) = 4 (db/m), 4 (db/db) and 3 (db/db + atRA); n (28w) = 4 (db/m), 3 (db/db) and 3 (db/db + atRA); * vs db/m. i Blood glucose, n (baseline) = 5 (db/m), 5 (db/db) and 5 (db/db + atRA); n (4w) = 5 (db/m), 6 (db/db) and 5 (db/db + atRA); n (16w) = 5 (db/m), 4 (db/db) and 4 (db/db + atRA); n (24w) = 5 (db/m), 4 (db/db) and 5 (db/db + atRA); n (28w) = 5 (db/m), 4 (db/db) and 4 (db/db + atRA); * vs db/m. j Serum insulin and HOMA-IR, $$n = 4$$(db/m), 5 (db/db) and 3 (db/db + atRA), * vs db/m; # vs db/db. k ITT, $$n = 4$$ (db/m), 6 (db/db) and 5 (db/db + atRA),* vs db/m; # vs db/db. l Serum lipids, n (TG and TC) = 4 (db/m), 5 (db/db) and 3 (db/db + atRA), n (HDL and LDL) = 5 (db/m), 4 (db/db) and 3 (db/db + atRA), * vs db/m, # vs db/db. Data are expressed as means ± SD. One-way ANOVA with Tukey post hoc test was used for the analysis of statistical significance. Source data are provided as a Source Data file. ( Black arrows: Typical Tunel stained positive cells; white boxes: lipid droplets).
These results suggest that atRA prevented myocardial injury by restoring cardiac atRA levels and improving systemic metabolic disorder in T2DM mice.
## RDH10 expression was reduced in the hearts of mice and patients with T2DM
The levels of cardiac Rol and atRA in untreated and Rol-supplemented db/db mice (Fig. 1f and Fig. 2d) suggested that the conversion of Rol to atRA is disrupted in the heart in T2DM. Therefore, we measured the expression of cardiac RDH10, a rate-limiting enzyme in retinol metabolism that plays an important role in converting Rol to atRA, and found that as heart function shifted from compensated to decompensated, cardiac RDH10 expression increased and then decreased in db/db mice (Supplementary Fig. 3a, b and d) while we found a significant decrease in the activity of RDHs in the myocardium of db/db mice (Supplementary Fig. 3c)A severe reduction in cardiac RDH10 was also found in T2DM patients (Supplementary Fig. 3d). These results suggest that RDH10 may be closely associated with disturbed cardiac retinol metabolism and its resulting DCM in T2DM.
## The loss of cardiac RDH10 led to disordered cardiac retinol metabolism and severe myocardial injury in mice
To further confirm the relationship between RDH10, cardiac retinol metabolism, and myocardial injury, we bred mice with cardiomyocyte-specific RDH10 deletion. Global RDH10 knockout is embryonically lethal11 and causes abnormal development in several organs, including the heart20. To exclude possible effects of abnormal heart development and to investigate the relationship between RDH10, retinol metabolism, and myocardial injury in adult heart more specifically, we used the myh6 system to generate cardiomyocyte-specific conditional RDH10-knockout (RDH10-cKO) mice by tamoxifen (TMX) treatment at 5 weeks of age to induce iCre protein specifically in cardiomyocytes (Fig. 4a) and verified the RDH10 deletion in cardiomyocytes of these mice after 4 weeks (Supplementary Fig. 4b and c). Our further results showed that at week 15 after TMX treatment, RDH10-cKO mice exhibited a halving of cardiac atRA (Fig. 4b), heart failure (Fig. 4c and d), and severe cardiac structural remodeling including myocardial hypertrophy, myocardial fibrosis, apoptosis, lipid deposition and increased mitochondria in cardiomyocytes (Fig. 4e–g).Fig. 4Cardiac retinol metabolic status, structure, and function in cardiomyocyte-specific conditional Retinol Dehydrogenase 10-knockout (RDH10-cKO) mice (n means biologically independent animals).a Schematic diagram of the construction of RDH10-cKO mice. b Cardiac Rol and atRA levels, $$n = 3$$ (RDH10fl/fl) and 4 (RDH10-cKO), * vs RDH10fl/fl. c Echocardiography. d LVEF and LVFS, n (5w) = 5 (WT), 4 (RDH10fl/fl), 5 (RDH10-cKO); n (10w) = 5 (WT), 3 (RDH10fl/fl), 7 (RDH10-cKO); n (15w) = 4 (WT), 4 (RDH10fl/fl), 5 (RDH10-cKO); * vs WT; # vs RDH10fl/fl. e Heart image and heart/tibia ratio of RDH10-cKO mice, $$n = 4$$ (WT), 6 (RDH10fl/fl) and 5 (RDH10-cKO). f Cardiac WGA, Masson, Tunel and TEM staining of RDH10-cKO mice. g Analysis of f, $$n = 3$$, * vs WT; # vs RDH10fl/fl. Data are expressed as means ± SD. Two-tailed unpaired t-test was used for the analysis of statistical significance in (b) while one-way ANOVA with Tukey post hoc test was used for the analysis of statistical significance in (d), (e) and (g). Source data are provided as a Source Data file. ( Black arrows: Typical Tunel stained positive cells; white boxes: lipid droplets).
These results suggest that the absence of cardiac RDH10 and its resultant cardiac retinol disorder manifested mainly as atRA deficiency can independently induce myocardial injury.
## RDH10 overexpression ameliorated cardiac retinol metabolism disorder and myocardial injury in T2DM mice
To further validate the role of RDH10 and cardiac retinol metabolism in DCM, we overexpressed RDH10 in the hearts of db/db mice via adeno-associated virus 9 (AAV9)-RDH10 injection at week 8 after experimental baseline (Fig. 5a) and verified that the expression of cardiac RDH10 of these mice restored to a level comparable to that of db/m mice after 5 weeks (Supplementary Fig. 5a and b). At 28 weeks of hyperglycemia in db/db mice, AAV9-RDH10 alleviated cardiac retinol metabolism disorder (Fig. 5b), prevented heart failure (Fig. 5c and d), attenuated myocardial injury including myocardial hypertrophy, fibrosis, apoptosis, and increases in lipid deposition and mitochondria in the cardiomyocytes (Fig. 5e–g).Fig. 5Cardiac retinol metabolic status, structure, and function in T2DM mice injected with via adeno-associated virus 9 (AAV9)-RDH10 (n means biologically independent animals).a Schematic diagram of AAV9-RDH10 injection. b Cardiac Rol and atRA levels, $$n = 3$$, * vs db/db. c Echocardiography. d LVEF and LVFS, n (baseline) = 5 (db/m), 5 (db/db), 4 (db/db-GFP) and 5 (db/db-RDH10); n (4w) = 5 (db/m), 5 (db/db), 4 (db/db-GFP) and 5 (db/db-RDH10); n (16w) = 3 (db/m), 5 (db/db), 4 (db/db-GFP) and 3 (db/db-RDH10); n (28w) = 3 (db/m), 3 (db/db), 3 (db/db-GFP) and 3 (db/db-RDH10); * vs db/db; # vs db/db-GFP; & vs db/db-RDH10. e Heart image and heart/tibia ratio, $$n = 4$$ (db/m and db/db-GFP) and 5(db/db and db/db-RDH10), * vs db/m; # vs db/db; & vs db/db-GFP. f Cardiac WGA, Masson, Tunel, and TEM staining. g Analysis of f, $$n = 3$$, * vs db/m; # vs db/db; & vs db/db-GFP. Data are expressed as means ± SD. Two-tailed unpaired t-test was used for the analysis of statistical significance in (b) while one-way ANOVA with Tukey post hoc test was used for the analysis of statistical significance in (d), (e) and (g). Source data are provided as a Source Data file. ( White arrows: Typical Tunel stained positive cells; white boxes: lipid droplets).
These results suggest that RDH10 reduction and cardiac retinol metabolism disorder is the major cause of DCM in T2DM.
## Reduction in RDH10 promoted lipid deposition and free fatty acids (FFAs) uptake mediated by cardiac atRA deficiency in the hearts of T2DM mice
In DM, cardiomyocytes that do not store lipids accumulate large amounts of lipids, which is a process called myocardial steatosis and is a hallmark of cardiac lipotoxicity and DCM21,22. Cardiac lipid deposition in DCM is closely associated with excessive cardiac uptake of FFAs23, which is mainly mediated by Fatty acid translocase (FAT/CD36)24,25.
In the observation of the ultrastructure of cardiomyocytes, we noticed that altered retinol metabolism status and RDH10 expression were associated with lipid deposition (Fig. 2f, Fig. 3f, Fig. 4f, and Fig. 5f), suggesting a possible link between RDH10, disturbed retinol metabolism and myocardial lipotoxicity in DCM. We performed oil red O staining and cardiac triacylglyceride (TG) levels measurement and found that RDH10 deletion increased myocardial lipid deposition and TG levels whereas AAV9-RDH10 reduced myocardial lipid deposition and TG levels in db/db mice (Fig. 6a, b, e, and f), atRA but not Rol reduced myocardial lipid deposition and TG levels in db/db mice (Fig. 6c, d, and f), suggesting RDH10 reduction and its leading retinol metabolism disorder promotes lipotoxicity via atRA deficiency in the heart in T2DM. We verified this conclusion again by measuring lipid deposition in Neonatal mouse primary cardiomyocytes (NMPCs), which showed that the silent RDH10-induced increase in lipid deposition could be attenuated by atRA but not Rol when exposed to 200 mmol/L palmitic acid (PA) and that AGN193109 acted as an inhibitor of RARs to prevent the effect of atRA (Fig. 6g). CD36 is a key factor in the abnormal uptake of cardiac FFAs in DCM and has been found to be suppressed by synthetic retinoic acid in atherosclerosis26. We found in NMPCs that RDH10 deletion increased both mRNA and protein levels of CD36 by decreasing atRA (Fig. 6h), whereas the increase in cardiac CD36 in db/db mice could be restored by supplementing atRA (Fig. 6i), suggesting that in the heart in T2DM, the decrease in RDH10 and its resulting retinol metabolism disorder promotes FFAs uptake. We further performed a cardiac FFAs uptake capacity assay, which showed that cardiac RDH10 deficiency in mice increases cardiac FFAs uptake, while atRA mitigated the increased cardiac FFAs uptake in db/db mice (Fig. 6j).Fig. 6RDH10 reduction-induced retinol metabolism disorder promoted cardiac lipotoxicity via atRA deficiency-mediated increases in lipid accumulation and FFAs uptake in T2DM mice (n means biologically independent animals in e, f, i, and j while n means independent experiments in h).a Cardiac oil red O staining of RDH10-cKO mice. b Cardiac oil red O staining of db/db mice with AAV9-RDH10. c Cardiac oil red O staining of db/db mice with Rol. d Cardiac oil red O staining of db/db mice with atRA. e Cardiac TG levels of RDH10-cKO mice, $$n = 3$$, * vs RDH10fl/fl. f Cardiac TG levels of db/db mice with AAV9-RDH10, atRA, or Rol, $$n = 4$$ (db/m) and 3 (db/db, db/db-GFP, db/db-RDH10 and db/db + atRA), * vs db/m; # vs db/db; & vs db/db-GFP. g Oil red O staining of Neonatal mouse primary cardiomyocytes (NMPCs). h WB of CD36 in NMPCs, $$n = 4$$, * vs 1; # vs 2; & vs 3. i WB of cardiac CD36, $$n = 3$$, * vs db/m; # vs db/db. j Cardiac FFAs uptake capacity detection of RDH10-cKO mice and db/db mice with atRA, $$n = 3$$, * vs RDH10fl/fl (RDH10-cKO) or db/m (db/db). Data are expressed as means ± SD. Two-tailed unpaired t-test was used for the analysis of statistical significance in (e), (f) (right) and (j) (left) while one-way ANOVA with Tukey post hoc test was used for the analysis of statistical significance in (f) (left), (h), (I) and (j) (right). Source data are provided as a Source Data file.
These results suggest that a decrease in cardiac RDH10 increases cardiac lipid deposition and FFAs uptake by reducing cardiac atRA levels, thereby promoting myocardial lipotoxicity in T2DM.
## Ferroptosis is involved in myocardial injury in T2DM
Ferroptosis was identified in 2012 as a form of cell death that relies on iron-catalyzed lipid peroxidation27. Recently, Wang Xiang et al. and Ni Tingjuan et al. found that ferroptosis is an important mechanism of diabetic cardiomyopathy in different T2DM mouse models, respectively28,29. In our RNA-seq data, we also found that in the hearts of db/db mice, ferroptosis-promoting genes HMOX1, MAP2LCB, ACSL1, TRF, CP, NCOA4, SLC39A8, and ALOX15 were gradually upregulated whereas ferroptosis-inhibiting gene SLC7A11 was gradually downregulated (Supplementary Fig. 7a), along with increased levels of 4-Hydroxynonenal (4-HNE), malondialdehyde (MDA) iron, and non-heme iron (indicators of iron atrophy) (Supplementary Fig. 7b-e), further confirmed the presence of ferroptosis in the heart of T2DM mice. In addition, we also confirmed increased iron and 4-HNE levels in the hearts of T2DM patients (Supplementary Fig. 6f and g), demonstrating that ferroptosis also occurs in the hearts of T2DM patients.
## Reduction in RDH10 promoted ferroptosis mediated by suppression of glutathione peroxidase 4 (GPX4), ferroptosis suppressor protein 1 (FSP1) and ferroportin 1 (FPN1) in the hearts of T2DM mice
The study has demonstrated that atRA prevented iron overload-induced liver injury30, suggesting a link between retinol metabolism and ferroptosis. We measured the levels of cardiac 4-HNE, MDA, iron, and non-heme iron in RDH10-cKO mice and found increased 4-HNE and MDA levels (Fig. 7a–d), which only provided evidence of lipid peroxidation rather than ferroptosis; thus, we further treated RDH10-cKO mice with the ferroptosis inhibitor ferrostatin-1 (Fer-1) to confirm the presence of ferroptosis and found that Fer-1 rescued heart failure and inhibited cardiac 4-HNE accumulation in RDH10-cKO mice (Fig. 7e–h), which suggested that ferroptosis, mainly caused by lipid peroxidation, is involved in myocardial injury in RDH10-cKO mice. To further investigate the molecules that mediate cardiac retinol metabolism disorder leading to ferroptosis, we validated molecules that have been shown to regulate ferroptosis through the regulation of lipid peroxidation, GPX4, FSP1, DHODH, and SLC7A1131–34. Our results showed cardiac RDH10 deficiency reduced cardiac GPX4 and FSP1 (Fig. 7i) but not DHODH and SLC7A11 (Supplementary Fig. 7) in RDH10-cKO mice, while in NMPCs, atRA but not Rol reversed silent RDH10-induced GPX4 and FSP1 reduction, which were blocked by AGN193109 (Fig. 7j and k), suggesting that cardiac retinol metabolism disorder-induced atRA deficiency can lead to cardiac ferroptosis by reducing GPX4 and FSP1-mediated increases in lipid peroxidation. Further, we measured the levels of GPX4, FSP1, 4-HNE, and MDA in db/db mice with RDH10-AAV9 virus and atRA supplementation and found that reversal of RDH10 and atRA restored GPX4 and FSP1 expression and significantly reduced lipid peroxidation in the hearts of db/db mice (Fig. 7l and m), suggesting retinol metabolism disorder-induced atRA deficiency leads to ferroptosis mediated by GPX4 and FSP1 reduction in the heart in T2DM.Fig. 7RDH10 reduction-induced retinol metabolism disorder promoted ferroptosis via atRA deficiency-mediated glutathione peroxidase 4 (GPX4) reduction in the heart in T2DM (n means biologically independent animals).a Cardiac 4-Hydroxynonenal (4-HNE) staining of RDH10-cKO mice, $$n = 3$$, * vs RDH10fl/fl. b Cardiac Malondialdehyde (MDA) levels of RDH10-cKO mice, $$n = 6$$ (RDH10fl/fl) and 11 (RDH10-cKO), * vs RDH10fl/fl. c Cardiac iron levels of RDH10-cKO mice, $$n = 6$$ (RDH10fl/fl) and 10 (RDH10-cKO), * vs RDH10fl/fl. d Cardiac non-heme iron levels of RDH10-cKO mice, $$n = 6$$ (RDH10fl/fl) and 10 (RDH10-cKO). e Schematic diagram of Fer-1 treatment in RDH10-cKO mice. f Echocardiography. g LVEF and LVFS, n (5w) = 3 (RDH10fl/fl), 4 (RDH10-cKO), 5 (RDH10-cKO+Fer-1); *vs RDH10fl/fl; # vs RDH10-cKO+Fer-1. h Cardiac 4-HNE staining of RDH10-cKO mice with Fer-1, $$n = 3$$, * vs RDH10-cKO. i WB of cardiac GPX4 and FSP1 in RDH10-cKO mice, $$n = 3$$, * vs RDH10fl/fl. j WB of GPX4 amd FSP1 in NMPCs. k WB of cardiac GPX4 and FSP1 in db/db mice with atRA, $$n = 3$$, * vs db/m; # vs db/db. l Cardiac 4-HNE staining of db/db mice with AAV9-RDH10 or atRA, $$n = 4$$ (db/db), 3 (db/db + RDH10) and 3 (db/db + atRA), * vs db/db. m Cardiac MDA levels of db/db mice with AAV9-RDH10 or atRA, $$n = 3$$, * vs db/db. Data are expressed as means ± SD. Two-tailed unpaired t-test was used for the analysis of statistical significance in (a), (b), (c), (d), (h) and (i) while one-way ANOVA with Tukey post hoc test was used for the analysis of statistical significance in (g), (k) and (l). Source data are provided as a Source Data file.
The ferroptosis in the heart in T2DM also exhibited iron accumulation (Supplementary Fig. 6d and e), but we did not observe iron accumulation in the hearts of RDH10-cKO mice (Fig. 7c and d). Wang X, et al. demonstrated that there is an increase in serum iron in T2DM35, suggesting to us that we should measure iron accumulation in RDH10-cKO mice in the presence of disturbing systemic iron status. We, therefore, fed RDH10-cKO mice a high iron diet (HID) as described36 and found that RDH10-cKO mice developed heart failure and cardiac iron accumulation after 3 weeks of HID feeding (Fig. 8a–e). More importantly, AAV-RDH10 and atRA could significantly inhibit cardiac iron accumulation in db/db mice (Fig. 8f and g), suggesting that RDH10 deficiency-induced retinol metabolism disorder promotes cardiac iron accumulation via atRA reduction in the heart in T2DM. We measured the levels of transferrin Receptor (TFRC) and FPN1 which regulate iron uptake and output, respectively, and found that alterations in cardiac retinol metabolism altered only the expression of FPN1 (Fig. 8h) but not TFRC (Supplementary Fig. 7), as in the experimental results of other investigators30.Fig. 8RDH10 reduction-induced retinol metabolism disorder promoted iron accumulation via atRA deficiency in the heart in T2DM (n means biologically independent animals).a Schematic diagram of high iron diet (HID) treatment in RDH10-cKO mice. b Echocardiography. c LVEF and LVFS, n (2w) = 3 (RDH10fl/fl), 4 (RDH10-cKO), 3 (RDH10fl/fl + HID) and 3 (RDH10-cKO+HID); n (3w) = 3 (RDH10fl/fl), 3 (RDH10-cKO), 4 (RDH10fl/fl + HID) and 3 (RDH10-cKO+HID); n (4w) = 5 (RDH10fl/fl), 4 (RDH10-cKO), 5 (RDH10fl/fl + HID) and 3 (RDH10-cKO+HID); * vs RDH10fl/fl; # vs RDH10-cKO; & vs RDH10fl/fl + HID. d Cardiac iron levels, $$n = 4$$ (RDH10fl/fl) and 3 (RDH10-cKO), * vs RDH10fl/fl + HID. e Cardiac non-heme iron levels, $$n = 55$$ (RDH10fl/fl) and 3 (RDH10-cKO). f Cardiac iron levels of db/db mice with AAV9-RDH10, Rol or atRA, $$n = 3$$, * vs db/db. g Cardiac non-heme iron levels of db/db mice with AAV9-RDH10, Rol or atRA, $$n = 3$$, * vs db/db. h WB of FPN1 in NMPCs, these results were independently repeated 3 times with similar results. i WB of cardiac FPN1 in db/db mice with atRA, these results were independently repeated 3 times with similar results. Data are expressed as means ± SD. Two-tailed unpaired t-test was used for the analysis of statistical significance in (d), (e), (f) (right) and (g) (right) while one-way ANOVA with Tukey post hoc test was used for the analysis of statistical significance in (c), (f) (left) and (g) (left). Source data are provided as a Source Data file.
AtRA function as a ligand for nuclear RARs, RA-RARs can activate or repress transcription of target genes. We predicted that multiple RARs binding sites exist on the promoter sequences of GPX4, FSP1 and FPN1 and verified the effects of altered retinol metabolism on the transcript levels of GPX4, FSP1 and FPN in NMPCs (Supplementary Fig. 8a and b).
These results suggest that a reduction in RDH10 promoted ferroptosis by cardiac atRA-RARs deficiency-induced GPX4, FSP1, and FPN1 reduction in the heart in T2DM.
## Discussion
In this study, we found by RNA-seq analysis that retinol metabolism disorder gradually emerged in the hearts of db/db mice with the development of DCM and concluded after a series of experiments that retinol metabolism disorder characterized by Rol overload and atRA deficiency is induced by a decrease in RDH10 and promotes myocardial injury mainly through Rol overload-induced cardiotoxicity and atRA deficiency-induced cardiac lipotoxicity and ferroptosis in the heart in T2DM (Fig. 9a).Fig. 9Graphical abstract.a Graphical summary of the mechanistic study of this study. b Graphical summary of clinical implications of this study. Graphics were created with Biorender.com.
We assessed cardiac retinol metabolic status in T2DM mice by measuring the levels of Rol, atRA, and RARs (substrates, products, and major biological effectors of retinol metabolism) and verified retinol metabolism disorder characterized by Rol overload, atRA deficiency, and reduced RARa and RARb in the hearts of T2DM mice. Considering the differences between mice and humans, we also evaluated cardiac retinol metabolic status in T2DM patients. Since human heart tissue is difficult to collect and we could not obtain enough fresh heart tissue to measure the levels of Rol and atRA, we used forensically collected heart sections from T2DM patients to perform immunohistochemistry (IHC) staining of RARs and found that the expression of RARa and RARb were reduced, suggesting that cardiac retinol metabolism disorder may also be present in the hearts of T2DM patients. Additionally, because the samples we used were forensically sourced, we were unable to collect much information about the donors of these samples, which greatly limited our validation in humans. Fortunately, a study by Ni Yang et al. published in early 2021 confirmed the increase in Rol and decrease in atRA in the hearts of patients with heart failure, and this study has filled in the gaps in our validation of humans, although that study focused only on atRA8.
After demonstrating disturbed cardiac retinol metabolism in T2DM, we assessed the effects of altered Rol and atRA levels on myocardial injury in T2DM by supplementing db/db mice with Rol or atRA. Our results showed that Rol, although attenuated systemic metabolic disorder, promoted myocardial injury by exacerbating myocardial fibrosis, apoptosis, and ultrastructural disruption of cardiomyocytes in db/db mice, which, combined with the significant upregulation of myocardial Rol levels, suggested that it is cardiotoxicity caused by cardiac Rol accumulation promotes myocardial injury in T2DM. In contrast, atRA attenuated systemic metabolic disorders and myocardial injury and restored cardiac atRA levels in db/db mice, which, combined with a related study on the importance of atRA on the heart8, suggests that atRA deficiency is a risk factor for myocardial injury in T2DM. Based on these findings, we suggest that cardiac retinol metabolism disorder promotes DCM, atRA is beneficial in the treatment of myocardial injury in T2DM whereas Rol should be avoided in T2DM patients because its cardiotoxicity is overloaded.
As a metabolic substrate in retinol metabolism, Rol has not attracted much attention from researchers in the cardiovascular field, the limited studies available have yielded conflicting conclusions about the roles of Rol in cardiovascular disease, and all of these studies have focused on Rol levels in serum37. In this study, we confirmed the beneficial effects of Rol on the systemic metabolic disorder and the cardiotoxicity of Rol overload in the heart.
The retinol metabolic status in the hearts of db/db mice indicates the impaired conversion of Rol to atRA. Therefore, we hypothesized and validated that RDH10 deficiency is the initiating factor for cardiac retinol metabolism disorder and myocardial injury in T2DM by overexpressing RDH10 in db/db mice and constructing RDH10-cKO mice. RDH10 is a rate-limiting enzyme in retinol metabolism that limits the conversion of Rol to atRA and affects organ development in embryos11,20, but its function in the adult heart is unknown. In this study, we demonstrated for the first time the roles of RDH10 on cardiac retinol metabolism and heart disease in adulthood. Additionally, the findings associated with RDH10-cKO mice exclude possible effects of abnormalities in blood glucose, lipids, and insulin, indicating the importance of cardiac retinol metabolism in the heart.
Our study verified that cardiac atRA deficiency leads to myocardial injury in T2DM through lipotoxicity and ferroptosis. Lipotoxicity marked by lipid accumulation caused by abnormal FFAs uptake, primarily mediated by CD36, is a key contributor to DCM21–24. We verified that in T2DM, alterations in retinol metabolism, particularly atRA levels, regulate cardiac lipid accumulation. In a previous study, researchers suggested that FFAs β-oxidation promoted by atRA/RARs plays the most important role in the regulation of lipid metabolism by retinol metabolism38. However, increased FFAs uptake contributes more to cardiac lipotoxicity than other factors in DCM39. We considered FFAs uptake as a mechanism by which retinol metabolism affects cardiac lipotoxicity in T2DM and succeeded in preliminarily demonstrating this effect in vitro and in vivo. Based on these results, we suggest that in T2DM, RDH10-mediated retinol metabolism disorder promotes cardiac lipotoxicity by increasing cardiac lipid accumulation and FFAs uptake through reduced atRA.
Ferroptosis is a form of cell death that relies on iron-catalyzed lipid peroxidation and is distinct from necrosis, apoptosis, and autophagy27. Relevant studies have shown the involvement of ferroptosis in myocardial injury in T2DM mice28,29. We further measured and confirmed the occurrence of ferroptosis in the heart in T2DM mice and patients. Research showed that atRA inhibits iron overload-induced liver injury in mice30, suggesting a link between retinol metabolism and ferroptosis. We verified that disordered retinol metabolism was associated with cardiac ferroptosis and that atRA supplementation and RDH10 overexpression inhibited cardiac ferroptosis in T2DM mice by reducing iron accumulation and lipid peroxidation. We also found in RDH10-cKO mice that cardiac retinol metabolism disorder mediated by RDH10 deficiency caused ferroptosis by reducing atRA leading to decreases in GPX4 and FSP1 mediating increased lipid peroxidation and decrease in FPN1 mediating cardiac iron accumulation. Based on these results, we conclude that in T2DM, cardiac RDH10 reduction-mediated retinol metabolism disorder leads to ferroptosis, which promotes DCM.
In summary, in this study, we report for the first time that in T2DM, RDH10 reduction leads to cardiac retinol metabolism disorder characterized by Rol overload, atRA deficiency, and RARs reduction, and promotes DCM through Rol overload-induced cardiotoxicity and atRA deficiency-induced lipotoxicity and ferroptosis, as shown in Fig. 9A. We also suggest that atRA and RDH10 could be potential targets for the prevention and treatment of DCM by correcting disordered retinol metabolism, whereas Rol, as known as vitamin A, should be avoided in patients with T2DM because of its deleterious effects on the heart in excess, as shown in Fig. 9B.
## Animal studies
Db/m, db/db mice (C57BLKS/J background), and MYH6-iCre mice (C57BL/6 background) were purchased from GemPharmatech (Nanjing, Jiangsu, China). RDH10fl/fl mice (C57BL/6 background) were generously provided by Prof. Jianxing Ma (Health Sciences Center, University of Oklahoma). All enrolled mice were male and aged 6–8 weeks. Mice maintained at the Center for Disease Model Animals of Sun Yat-sen University. Mice were housed on a 12 h light-dark cycle at 22–25 °C with 40–$70\%$ humidity and allowed free access to food (chow diet, purchased by Guangdong Medical Laboratory Animal Center, consisting of fat [$4.8\%$], protein [$18.6\%$], and carbohydrate [$61\%$]) and water except as noted. Mice were euthanized by intraperitoneal injection of 150 mg/kg sodium pentobarbital when they reached the experimental time endpoint or when any of the following criteria were met: [1] persistent lethargy and failure to clean hair; [2] failure to respond to physical interventions or behavioral signs of human touch, including marked inactivity, dyspnea, sunken eyes, and hunched posture; and [3] abnormal central nervous responses (convulsions, tremors, paralysis, head tilt, etc.). Mice were fasted for 12 h before sampling and testing. All animal experiments were approved by the Animal Care and Ethics Committee of Zhongshan School of Medicine, Sun Yat-sen University, and followed the National Institutes of Health Guidelines on the Care and Use of Animals (the protocol number is SYSU-IACUC-2019-B027).
## T2DM mice
85 male db/db mice were used as T2DM mice and were divided into RNA-seq (15 mice), control (30 mice), Rol treatment (10 mice), atRA treatment (10 mice), AAV9-GFP (10 mice), and AAV9-RDH10 (10 mice) groups. 45 age-matched male db/m mice were used as normal controls for db/db mice.
## RDH10-cKO mice
RDH10-cKO mice, which contain both RDH10fl/fl and MYH6-iCre, were bred by RDH10fl/fl mice and MYH6-iCre mice and were injected tamoxifen intraperitoneally (50 mg/kg, T2859, Sigma -Aldrich, St. Louis, MO) for 5 consecutive days from 5 weeks of age. 55 male RDH10-cKO mice were divided into control (35 mice), Fer-1 treatment (10 mice), and HID (10 mice) groups. 45 age-matched male RDH10fl/fl mice that also received TMX injections served as normal controls for RDH10-cKO mice.
## Animal treatments
Mice in the Rol treatment group received Rol gavage (800 IU/each, 17772, Sigma-Aldrich, St. Louis, MO) every two days from 8 weeks of age. Mice in the atRA treatment group received atRA intraperitoneal injection (5 mg/kg body weight, R2625, Sigma-Aldrich, St. Louis, MO) daily from 8 weeks of age. Mice in the Fer-1 treatment group received Fer-1 intraperitoneal injection (SML0583, 1 mg/kg body weight, Sigma-Aldrich, St. Louis, MO) daily from 6 weeks of age. Mice in HID group were fed HID (8.3 g carbonyl iron/kg, RD17082801, ReadyDietech, Shenzhen, China) from 6 weeks of age.
## Drugs preparation
Rol, atRA, and Fer-1 were accurately weighed and placed in brown light-proof tubes and dissolved with dimethylsulfoxide (DMSO) (ST2335, Beyotime, Shanghai, China) followed by diluting 100-fold with corn oil (C116023, Aladdin, Shanghai, China) to prepare working solutions with the final concentrations of 5 mg/mL, 2.5 mg/mL and 0.4 mg/mL, respectively.
ROL, atRA, and AGN193109 were first accurately weighed and placed in brown light-proof tubes, then dissolved with DMSO (D8418, Sigma-Aldrich, St. Louis, MO) to prepare working solutions with final concentrations of 5 mg/L, 5 mmol/L, and 2 mmol/L, respectively.
## Gene therapy
A recombinant AAV9 vector carrying the mouse RDH10 sequence (AAV9-RDH10, DZ-AAV-Rdh10-OE, Dongze, Hanbio Inc, Shanghai, China) was used to overexpress RDH10. AAV9-GFP (DZ-AAV-Rdh10-NC, Dongze, Hanbio Inc, Shanghai, China) was used as a negative control. 0.8 * 10 ^ 11 vg/per animal of AA9-RDH10 or AAV9-GFP was transferred into T2DM mice, respectively, by tail vein injection at the age of 16 weeks.
## Echocardiography
Mice were anesthetized with $1.5\%$ isoflurane and placed on a thermostat at 37 °C immediately, $0.5\%$ isoflurane is inhaled continuously to prevent them from waking up. For image acquisition, Vevo 3100 Imaging System (VisualSonics, Canada) with a 400 MHz probe was used to detect cardiac motion in the long-axis view, then the probe was rotated 90 degrees to detect cardiac motion in the short-axis view, and graphs were acquired in M-Mode near the papillary muscles. The heart rates of the mice were controlled at 450–600 bpm.
## RNA-seq
Total RNA was extracted from the hearts of db/m mice and db/db mice at the age of 4, 24, and 32 weeks using the RNeasy Mini Kit (74104, Qiagen, Duesseldorf, Germany). Then, the strand-specific library was prepared after rRNA depleted and RNA fragmentation, First Strand cDNA synthesizing, Second Strand cDNA synthesizing, 3’ ends Adenylating, adapter Ligation, and PCR amplification. RNA and the library preparation integrity were inspected with Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). And the cluster and first dimension sequencing primer hybridization were accomplished on the cBot of the Illumina sequencing machine. Finally, performed sequencing at SHBIO Corporation.
The analysis of the KEGG pathway and GO terms were completed using KOBAS (http://kobas.cbi.pku.edu.cn/) online40.
## Cardiac Rol, atRA, and retinyl ester content measurements
Cardiac Rol, atRA, and retinyl ester measurements were performed as described by others41, and the details of the experiment were as follows.
## Sample preparation
Homogenized the heart tissue in 200 μL of cold NaCl solution ($0.9\%$) for 5 s, add 100 μL retinol-D4 (IR-23012, IsoSciences, Ambler, PA, USA) as internal standard for Rol and atRA, and retinyl acetate (46958, Sigma-Aldrich, St. Louis, MO) as internal standard for retinyl esters, then homogenize for another 5 seconds. For Rol and at RA, 1 mL formic acid n-hexane solution ($1\%$) was added to the homogenizing to facilitate phase separation by a 5 min vortex and 5 min centrifugation (16,200 g). The supernatants were collected in a new tube and blown dry with nitrogen followed by re-dilution with 100 μL $70\%$ methanol. For retinyl esters, 24 μL 0.5 M NaOH solution was added to tissue homogenate, followed by 1 mL n-hexane. The mixture was votexed 5 min and centrifuged at 16,200 g for 5 min. The supernatants were collected in a new tube and blown dry with nitrogen followed by re-dilution with 100 μL $100\%$ methanol. The processes above should be protected from the light.
## LC-MS/MS analysis
Samples were carried out on a Sciex Jasper TM high performance liquid chromatography (HPLC) system (Sciex, M.A., U.S.A) which consisted of a solvent degasser, a binary pump, an autosampler, and a column oven. The LC system was coupled with a Sciex Triple Quad TM 4500MD MS (Sciex, M.A., U.S.A) in ESI ionization mode. Data were acquired and analyzed with Analyst® MD version 1.6.3 and Multi Quant TM MD version 3.0.2 (Sciex, M.A., U.S.A). For Rol and at RA, LC was performed on a Phenomenex Kinetex C18 (50 * 2.1 mm, 2.6 μm, Phenomenex, CA, USA). Mobile phase A consisted of acetonitrile/methanol/water (4:3:3, v/v/v) with 0.1 % formic acid, and mobile phase B containing acetonitrile/methanol/water (5.5:3:1.5, v/v/v) with $0.1\%$ formic acid at a column oven temperature of 25 °C. The flow rate was set to 0.2 ml/min and the gradient elution procedure was as follows: initial conditions $75\%$ B; from 0 to 3.5 min linear increase to $95\%$ B; between 3.5 and 4.5 min $95\%$ B was retained; at 4.6 min back to initial conditions with $75\%$ B; finally, $75\%$ B was held from 4.6 to 5.9 min. For retinyl esters, LC was performed on a Phenomenex Kinetex C18 (100 * 2.1 mm, 1.7 μm, Phenomenex, CA, USA). Mobile phase A consisted of acetonitrile containing $0.1\%$ formic acid and 5 mM ammonium acetate, and mobile phase B containing pure water with $0.1\%$ formic acid and 5 mM ammonium acetate at a column oven temperature of 40 °C. The flow rate was set to 0.3 ml/min with the gradient elution as follows: initial conditions $85\%$ B for 1.5 min, from 0 to 9 min linearly increased to $100\%$ B, and then decreased to $85\%$ at 10 min, held for another 1 min. The autosampler was set at 10 °C. The injection volume was 10 μL.
## Mass spectrometry
The MS conditions were as follows: electrospray ionization (ESI) under positive mode; nebulizer gas: nitrogen; curtain gas,30 psi; ion spray voltage, 4000 V; temperature, 400 °C; gas 1 and gas 2, 35 and 40 psi, respectively; collision gas, 10 psi. The parameters of the mass spectrometer were optimized, and the multiple reaction monitoring (MRM) transitions of retinol, all-trans-retinoic acid, as well as D4-retinol were chosen as 269.2 > 93.1, 273.1 > 94.0, and 301.2 > 123.1, respectively. The MRM transitions for retinyl esters were chosen as 329.3 > 269.3 for retinyl acetate, 524.4 > 268.1 for retinyl palmitate (16:0), 552.5 > 268.2 for retinyl stearate (18:0), 522.4 > 268, retinyl palmitoleate (16:1), and retinyl oleate (18:1). The quantification was performed using the calibration curve.
## RDH activity assay
Cardiac RDHs activity assay was performed as described by others41, and the details of the experiment were as follows.
100 μg of cardiac microsomal fractions protein was incubated with 3 μM all-trans-retinol (17772, Sigma-Aldrich, St. Louis, MO) solubilized with bovine serum albumin and 1 mM NAD + (NAD98-RO, Sigma-Aldrich, St. Louis, MO) in 0.5 ml of the reaction buffer for 20 min at 37 °C. Reactions were stopped by the addition of an equal volume of ice-cold methanol, and Retinaldehyde were extracted twice with 2 ml of hexane. Hexane layers were dried, and the dry residue was reconstituted in 0.2 ml of acetonitrile. Retinaldehyde were separated by normal-phase HPLC using Spherisorb S3W column (4.6 mm × 100 mm; Waters) and isocratic mobile phase consisting of acetonitrile at 1 ml/min and analyzed.
## Western blotting analysis
The heart tissues or neonatal mouse primary cardiomyocytes (NMPCs) were lysed with RIPA buffer (P0013B, Beyotime, Shanghai, China) for total protein extraction. The protein concentration was determined using the BCA protein assay kit (71285, Millipore, Bedford, MA, USA) according to the manufacturer’s protocol. 35 μg protein was subjected to SDS-PAGE for electrophoresis, transferred to 0.45 μm PVDF membranes (IPVH00010, Millipore, Bedford, MA, USA), and immunoblotted with antibodies. The bands were quantified using the ImageJ software program (v.1.45, National Institutes of Health, Bethesda, MD, USA). The antibodies are listed in Supplementary Table 1.
## Serum measurement
Serum was obtained by centrifugation (4 °C, 1,500 g, 20 min) of blood collected from the eye sockets of the mice and stored at −80 °C. The serum fasting insulin (FINS) levels were examined using an enzyme-linked immunosorbent assay (CSB-E05071m, Cusabio, Wuhan, Hubei, China). The serum TC, TG, LDL-c, and HDL-c levels were examined using commercial reagent kits (A111-1-1, A110-1-1, A113-1-1 and A1122-1-1, Jiancheng, Nanjing, Jiangsu, China).
## Homeostasis model assessment of the insulin resistance index (HOMA-IR) and Insulin tolerance test (ITT)
The fasting blood glucose (FBG) levels were examined after the mice fasted for 12–16 h, and the HOMA-IR was calculated with the equation (FBG (mM/L) × FINS (mIU/L)) / 22.5. To perform the ITT, 1 U/kg body weight of insulin (Novolin R, Novo Nordisk, Bagsvaerd, Denmark) was intraperitoneally injected into the mice, then the blood glucose levels were examined at 0, 15, 30, 60, 90, and 120 min after the injection. The change curve of the blood glucose level was drawn, and the AUC was calculated.
## Human samples
Heart samples from 11 healthy populations and 11 age-matched T2DM patients (age difference ≤5) used in this study were collected from the National Center for Medico-legal Expertise of Sun Yat-sen University. All sample donors were diagnosed without coronary heart disease and hypertension. The use of human heart samples was approved by the ethics committee of Zhongshan School of Medicine, Sun Yat-sen University with the protocol number (2019-B027) and all data and sample collection were in strict accordance with ethics guidelines of Zhongshan School of Medicine, Sun Yat-sen University. Informed consent was obtained from the legal representatives of the victims. The principles outlined in the Declaration of Helsinki were followed. The information details of the donors were provided in Supplementary Table 3.
## Immunohistochemistry (IHC)
Heart samples were collected and fixed overnight in $4\%$ paraformaldehyde (BL539A, Biosharp, Hefei, Anhui, China), followed by routine dehydration and sectioning (5 μm). Sections were blocked with $3\%$ hydrogen peroxide, antigen repaired with citrate buffer (P0083, Beyotime, Shanghai, China), permeabilized with $0.3\%$ Triton-100 (ST795, Beyotime, Shanghai, China), blocked with $5\%$ BSA (A1933, Sigma-Aldrich, St. Louis, MO), incubated with primary antibody overnight at 4 °C and incubated with HRP-labeled secondary antibody for 30 min at 37 °C. Visualization was performed under the microscope (DFC700T, Leica, Germany) with DAB Horseradish Peroxidase Color Development Kit (P0203, Beyotime, Shanghai, China). Finally, the sections were sealed with neutral balsam fixative (G8590, Solarbio, Beijing, China). The positive cell number was counted from 4–5 fields per sample with the ImageJ software program (v.1.45, National Institutes of Health, Bethesda, MD, USA) and the mean density was quantified from 4–5 fields per sample with the Image-pro plus software program (v.6.0, Media Cybernetics, Rockville, MD, USA). The antibodies are listed in Supplementary Table 1.
## Immunofluorescence
Heart samples were collected and fixed in $4\%$ paraformaldehyde (BL539A, Biosharp, Hefei, China) overnight followed by conventional dehydration and slicing (5 μm). The heart samples sections were blocked with $3\%$ hydrogen peroxide and then performed at 95 °C for 10 min using citrate buffer (P0083, Beyotime, Shanghai, China), $0.3\%$ Triton-100 (ST795, Beyotime, Shanghai, China) was used for permeabilization and then the blocking step was carried out using the $5\%$ BSA (A1933, Sigma-Aldrich, St. Louis, MO). After overnight incubation of the primary antibody at 4 °C, a secondary antibody was applied to the sections at 37 °C for 90 min. All the immune-fluorescence images were captured by a fluorescence microscope (DFC700T, Leica, Germany). The antibody is listed in Supplementary Table 1.
## WGA-FITC staining
The cTNT was staining first following the method of Immunofluorescence, and then the sections were incubated with WGA-FITC (W11261, Sigma-Aldrich, St. Louis, MO) at 37 °C for 30 min. All the immune-fluorescence images were captured by a fluorescence microscope (DFC700T, Leica, Germany). The cardiomyocyte area was calculated from 4–5 fields per sample with the ImageJ software program (v.1.45, National Institutes of Health, Bethesda, MD, USA).
## Masson staining
Standard Masson staining was performed using the Modified Masson’s Trichrome Stain Kit (G1345, Solarbio, Beijing, China) according to the manufacturer’s protocol. The positive area was quantified from several fields per sample with the Image-pro plus software program (v.6.0, Media Cybernetics, Rockville, MD, USA).
## Terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) staining
TUNEL staining was performed using the Colorimetric TUNEL Apoptosis Assay Kit (C1091, Beyotime, Shanghai, China). The positive cell numbers were counted from 4–5 fields per sample with the ImageJ software program (v.1.45, National Institutes of Health, Bethesda, MD, USA).
## Transmission electron microscopy (TEM)
Cardiac ultrastructure was examined under a transmission electron microscope (Tecnai G2 Spirit Twin +GATAN 832.10 W; FEI; Czech Republic) using conventional methods. In brief, heart tissues were fixed with $2.5\%$ glutaraldehyde in 0.1 mol/L phosphate buffer (pH 7.4), followed by $1\%$ OsO4. After dehydration, thin sections were stained with uranyl acetate and lead citrate for observation, images were acquired digitally.
## Oil red O staining
The heart tissues were fixed in $4\%$ paraformaldehyde (BL539A, Biosharp, Hefei, China) overnight then dehydrated with a sucrose gradient, and embedded in the Tissue-Tek OCT compound (4583, Sakura Finetek, Tokyo, Japan). Then, the sections (8 μm) were stained with oil red O (O0625, Sigma-Aldrich, St. Louis, MO, USA) for 15 min.
Cells for oil red O staining were cultured in Millicell EZ SLIDE 8 well glass slides (PEZGS0816, Merck Millipore, Billerica, MA, USA) after isolation, and fixed in $4\%$ paraformaldehyde overnight after treatments. Oil Red O staining was then performed in the same manner as in sections of heart samples.
## Tissues FFAs uptake fluorescence imaging
As described by others42, mice were injected with 1 μg/g body weight BODIPY™ $\frac{558}{568}$ C12 (D3835, Thermo Fisher, MA, USA) via the tail vein, 50 min later, the hearts were collected after removing fat, blood, and auricles and rinsed in PBS. X-ray and fluorescence imaging was performed using a small animal living fluorescence imaging system (IVIS Spectrum, PerkinElmer, USA).
## Cardiac MDA measurement
Cardiac MDA measurement was performed using the Lipid Peroxidation MDA Assay Kit (S0131, Beyotime, Shanghai, China) according to the manufacturer’s protocol.
## Cardiac iron measurement
Cardiac iron measurement was performed using the Tissue Iron Content Assay Kit (BC4355, Solarbio, Beijing, China) according to the manufacturer’s protocol.
## Cardiac non-home iron measurement
Cardiac Non-home iron measurement was performed as described by others43 as follows. Weighed and digested the heart tissues in NHI acid ($10\%$ trichloroacetic acid in 3 M HCl) for 48 h at 65–70 °C. An equal volume of samples, iron standard (500 µg/dl, Aladdin, Shanghai, China) or NHI acid were incubated with 200 µl BAT buffer ($0.2\%$ thioglycolic acid and $0.02\%$ disodium 4,7 diphenyl 1,10 phenanthroline disulfonate in $50\%$ saturated NaAc solution) for 10 min at room temperature. The absorbance of the mixtures was read at 535 nm and the absorbance of the standard was used to scale the unknown sample concentration.
## Perl’s Prussian blue staining
Perl’s Prussian blue staining was performed using the Prussian Blue Iron Stain Kit (with Eosin solution) (G1424, Solarbio, Beijing, China) according to the manufacturer’s protocol.
## Quantitative real-time polymerase chain reaction (Q-PCR)
Total RNA was isolated using the RNAiso Plus (9109, Takara, Tokyo, Japan). 0.5 mg of total RNA was reverse transcribed into complementary DNA using the HiScript III RT SuperMix for qPCR (+gDNA wiper) kit (R323, Vazyme, Nanjing, China). Then the Q-PCR was performed on the ABI Q6 Flex Real-Time PCR machine (Applied Biosystems, Foster City, CA, USA) using the 2x SYBR Green qPCR Master Mix kit (B21203, Bimake, Houston, Texas USA). The relative gene expression levels were analyzed using the 2(–ΔΔCt) method and normalized against 18 S expression. Primer sequences are in Supplementary Table 2.
## Neonatal mouse primary cardiomyocytes (NMPCs) isolation and culture
1–3-day-old neonatal mice were skin disinfected with $75\%$ ethanol, then removed their hearts by cutting with sharp forceps and quickly minced in ice-cold D-hanks. The shredded tissue was digested with $0.1\%$ trypsin for 6–10 h at 4 °C, followed by 3–5 digestions with $0.08\%$ collagenase type II (17101015, Gibco, Grand Island, NY, USA) for 10 min each (37 °C, 80 rpm). The precipitate was transferred to a new tube and neutralized with a volume of DMEM (C11995500BT, Gibco, Grand Island, NY, USA) containing $10\%$ FBS (Gibco, Grand Island, NY, USA) and the digestion was continued by adding new type II collagenase until no precipitate was evident. The collected supernatant was centrifuged (160 g, 5 min), and the precipitate was resuspended in D-hanks and filtered through a 70 µm cell strainer (352350, Falcon, Corning, NY, USA). The filtered cytosol was centrifuged again (160 g, 5 min) and the precipitate was resuspended in DMEM containing $10\%$FBS. The resuspended cells were grown in 15 cm2 cell culture plates for 1.5 h to remove non-cardiomyocytes. Finally, purified cardiomyocytes were cultured in DMEM containing $10\%$ FBS and BrdU (19–160, Sigma-Aldrich, St. Louis, MO, USA) in cell culture dishes or Millicell EZ SLIDE 8 well glass slides (PEZGS0816, Merck Millipore, Billerica, MA, USA) coated with attachment factor (S006100, ThermoFisher, Waltham, MA, USA), and maintained in an incubator at 37 °C and $5\%$ CO2.
## In vitro treatments
Cells were cultured till the morphological expansion and autonomic rhythmic contractions were observed, then changed the culture medium to serum-free medium 12 h before the subsequent experimental treatments.
## Small interfering RNA (si-RNA) transfection
Mouse RDH10 stealth siRNA (1320001, Invitrogen, Carlsbad, CA, USA) was transfected with NMPCs using the Lipo3000 Transfection Kit (L3000015, Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol.
## Drugs treatment
After a 36 h si-RNA transfection, cells were cultures in a serum-free medium for 12 h followed by culturing with mediums containing Rol (17772, 5 μg/L, Sigma-Aldrich, St. Louis, MO), atRA (R2625, 5 μmol/L, Sigma-Aldrich, St. Louis, MO), atRA and AGN193109 (SML2034, 1 μmol/L, Sigma-Aldrich, St. Louis, MO) mixture, or equivalent amounts of DMSO (Sigma -Aldrich, St. Louis, MO), respectively. Then, cells were fixed in $4\%$ paraformaldehyde or collected for subsequent analysis 24 h later.
## Statistics
Statistical differences between groups were evaluated using GraphPad Prism 7 and R 4.2.2. Data were considered statistically significant if P was less than 0.05. All experiments were performed at least in triplicate, and quantitative data are presented as the mean ± SD. Statistical significance for samples was identified using the independent-sample t-test and One-way ANOVA.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Peer Review File Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36837-x.
## Source data
Source Data
## Peer review information
Nature Communications thanks Rajesh Katare, Boyi Gan, Lu Cai and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
## References
1. Kannel WB, Hjortland M, Castelli WP. **Role of diabetes in congestive heart failure: the Framingham study**. *Am. J. Cardiol.* (1974) **34** 29-34. DOI: 10.1016/0002-9149(74)90089-7
2. Wang ZV, Hill JA. **Diabetic cardiomyopathy: catabolism driving metabolism**. *Circulation* (2015) **131** 771-773. DOI: 10.1161/CIRCULATIONAHA.115.015357
3. Jia G, DeMarco VG, Sowers JR. **Insulin resistance and hyperinsulinaemia in diabetic cardiomyopathy**. *Nat. Rev. Endocrinol.* (2016) **12** 144-153. DOI: 10.1038/nrendo.2015.216
4. Kenny HC, Abel ED. **Heart failure in Type 2 diabetes mellitus**. *Circ. Res.* (2019) **124** 121-141. DOI: 10.1161/CIRCRESAHA.118.311371
5. Ritchie RH, Abel ED. **Basic mechanisms of diabetic heart disease**. *Circ. Res.* (2020) **126** 1501-1525. DOI: 10.1161/CIRCRESAHA.120.315913
6. Li X. **Distinct cardiac energy metabolism and oxidative stress adaptations between obese and non-obese type 2 diabetes mellitus**. *Theranostics* (2020) **10** 2675-2695. DOI: 10.7150/thno.40735
7. Lima IOL, Peres WAF, Cruz S, Ramalho A. **Association of ischemic cardiovascular disease with inadequacy of liver store of retinol in elderly individuals**. *Oxid. Med. Cell. Longev.* (2018) **2018** 9785231. DOI: 10.1155/2018/9785231
8. Yang N. **Cardiac retinoic acid levels decline in heart failure**. *JCI insight* (2021) **6** e137593. DOI: 10.1172/jci.insight.137593
9. Guleria RS, Choudhary R, Tanaka T, Baker KM, Pan J. **Retinoic acid receptor-mediated signaling protects cardiomyocytes from hyperglycemia induced apoptosis: role of the renin-angiotensin system**. *J. Cell. Physiol.* (2011) **226** 1292-1307. DOI: 10.1002/jcp.22457
10. Tang XH, Gudas LJ. **Retinoids, retinoic acid receptors, and cancer**. *Annu. Rev. Pathol.* (2011) **6** 345-364. DOI: 10.1146/annurev-pathol-011110-130303
11. Sandell LL. **RDH10 is essential for synthesis of embryonic retinoic acid and is required for limb, craniofacial, and organ development**. *Genes Dev.* (2007) **21** 1113-1124. DOI: 10.1101/gad.1533407
12. Rhinn M, Schuhbaur B, Niederreither K, Dollé P. **Involvement of retinol dehydrogenase 10 in embryonic patterning and rescue of its loss of function by maternal retinaldehyde treatment**. *Proc. Natl. Acad. Sci. USA* (2011) **108** 16687-16692. DOI: 10.1073/pnas.1103877108
13. Ono K, Sandell LL, Trainor PA, Wu DK. **Retinoic acid synthesis and autoregulation mediate zonal patterning of vestibular organs and inner ear morphogenesis**. *Dev. (Camb. Engl.)* (2020) **147** dev192070. DOI: 10.1242/dev.192070
14. Yan J. **Anti-liver fibrosis effects of the total flavonoids of litchi semen on CCl(4)-induced liver fibrosis in rats associated with the upregulation of retinol metabolism**. *Pharm. Biol.* (2022) **60** 1264-1277. DOI: 10.1080/13880209.2022.2086584
15. Kawai T, Yanaka N, Richards JS, Shimada M. **De novo-synthesized retinoic acid in ovarian antral follicles enhances fsh-mediated ovarian follicular cell differentiation and female fertility**. *Endocrinology* (2016) **157** 2160-2172. DOI: 10.1210/en.2015-2064
16. Hamamura K. **Alterations of hepatic metabolism in chronic kidney disease via d-box-binding protein aggravate the renal dysfunction**. *J. Biol. Chem.* (2016) **291** 4913-4927. DOI: 10.1074/jbc.M115.696930
17. Snyder JM. **Knockout of Cyp26a1 and Cyp26b1 during postnatal life causes reduced lifespan, dermatitis, splenomegaly, and systemic inflammation in mice**. *FASEB J. Off. Publ. Fed. Am. Soc. Exp. Biol.* (2020) **34** 15788-15804
18. Rowbotham SE, Illingworth NA, Daly AK, Veal GJ, Boddy AV. **Role of UDP-glucuronosyltransferase isoforms in 13-cis retinoic acid metabolism in humans**. *Drug Metab. disposition: Biol. Fate Chem.* (2010) **38** 1211-1217. DOI: 10.1124/dmd.109.031625
19. Narbonne JF. **A time course investigation of vitamin A level and lipid composition of the liver endoplasmic reticulum in rats following treatment with congeneric polychlorobiphenyls**. *Toxicology* (1990) **60** 253-261. DOI: 10.1016/0300-483X(90)90148-A
20. D’Aniello E, Ravisankar P, Waxman JS. **Rdh10a provides a conserved critical step in the synthesis of retinoic acid during zebrafish embryogenesis**. *PloS one* (2015) **10** e0138588. DOI: 10.1371/journal.pone.0138588
21. Rijzewijk LJ. **Myocardial steatosis is an independent predictor of diastolic dysfunction in type 2 diabetes mellitus**. *J. Am. Coll. Cardiol.* (2008) **52** 1793-1799. DOI: 10.1016/j.jacc.2008.07.062
22. McGavock JM. **Cardiac steatosis in diabetes mellitus: a 1H-magnetic resonance spectroscopy study**. *Circulation* (2007) **116** 1170-1175. DOI: 10.1161/CIRCULATIONAHA.106.645614
23. Sowton AP, Griffin JL, Murray AJ. **Metabolic profiling of the diabetic heart: toward a richer picture**. *Front. Physiol.* (2019) **10** 639. DOI: 10.3389/fphys.2019.00639
24. Glatz JF, Luiken JJ, Bonen A. **Membrane fatty acid transporters as regulators of lipid metabolism: implications for metabolic disease**. *Physiol. Rev.* (2010) **90** 367-417. DOI: 10.1152/physrev.00003.2009
25. Luiken JJ. **Contraction-induced fatty acid translocase/CD36 translocation in rat cardiac myocytes is mediated through AMP-activated protein kinase signaling**. *Diabetes* (2003) **52** 1627-1634. DOI: 10.2337/diabetes.52.7.1627
26. Takeda N. **Synthetic retinoid Am80 reduces scavenger receptor expression and atherosclerosis in mice by inhibiting IL-6**. *Arterioscler. Thromb. Vasc. Biol.* (2006) **26** 1177-1183. DOI: 10.1161/01.ATV.0000214296.94849.1c
27. Dixon SJ. **Ferroptosis: an iron-dependent form of nonapoptotic cell death**. *Cell* (2012) **149** 1060-1072. DOI: 10.1016/j.cell.2012.03.042
28. Ni T, Huang X, Pan S, Lu Z. **Inhibition of the long non-coding RNA ZFAS1 attenuates ferroptosis by sponging miR-150-5p and activates CCND2 against diabetic cardiomyopathy**. *J. Cell. Mol. Med.* (2021) **25** 9995-10007. DOI: 10.1111/jcmm.16890
29. Wang X. **Ferroptosis is essential for diabetic cardiomyopathy and is prevented by sulforaphane via AMPK/NRF2 pathways**. *Acta Pharm. Sin. B.* (2022) **12** 708-722. DOI: 10.1016/j.apsb.2021.10.005
30. Tsuchiya H. **Suppressive effects of retinoids on iron-induced oxidative stress in the liver**. *Gastroenterology* (2009) **136** 341-350.e348. DOI: 10.1053/j.gastro.2008.09.027
31. Yang WS. **Regulation of ferroptotic cancer cell death by GPX4**. *Cell* (2014) **156** 317-331. DOI: 10.1016/j.cell.2013.12.010
32. Doll S. **FSP1 is a glutathione-independent ferroptosis suppressor**. *Nature* (2019) **575** 693-698. DOI: 10.1038/s41586-019-1707-0
33. Mao C. **DHODH-mediated ferroptosis defence is a targetable vulnerability in cancer**. *Nature* (2021) **593** 586-590. DOI: 10.1038/s41586-021-03539-7
34. Jiang L. **Ferroptosis as a p53-mediated activity during tumour suppression**. *Nature* (2015) **520** 57-62. DOI: 10.1038/nature14344
35. Wang X. **Genetic support of A causal relationship between iron status and type 2 diabetes: a mendelian randomization study**. *J. Clin. Endocrinol. Metab.* (2021) **106** e4641-e4651. DOI: 10.1210/clinem/dgab454
36. Fang X. **Loss of cardiac ferritin H facilitates cardiomyopathy via Slc7a11-mediated ferroptosis**. *Circ. Res.* (2020) **127** 486-501. DOI: 10.1161/CIRCRESAHA.120.316509
37. Olsen T, Blomhoff R. **Retinol, retinoic acid, and retinol-binding protein 4 are differentially associated with cardiovascular disease, type 2 diabetes, and obesity: an overview of human studies**. *Adv. Nutr. (Bethesda, Md)* (2020) **11** 644-666. DOI: 10.1093/advances/nmz131
38. Yang D. **Modest decreases in endogenous all-trans-retinoic acid produced by a mouse Rdh10 Heterozygote provoke major abnormalities in adipogenesis and lipid metabolism**. *Diabetes* (2018) **67** 662-673. DOI: 10.2337/db17-0946
39. Nakamura M. **Glycogen synthase kinase-3α promotes fatty acid uptake and lipotoxic cardiomyopathy**. *Cell Metab.* (2019) **29** 1119-1134.e1112. DOI: 10.1016/j.cmet.2019.01.005
40. Bu D. **KOBAS-i: intelligent prioritization and exploratory visualization of biological functions for gene enrichment analysis**. *Nucleic Acids Res.* (2021) **49** W317-W325. DOI: 10.1093/nar/gkab447
41. Klyuyeva AV. **Changes in retinoid metabolism and signaling associated with metabolic remodeling during fasting and in type I diabetes**. *J. Biol. Chem.* (2021) **296** 100323. DOI: 10.1016/j.jbc.2021.100323
42. Huang TS. **Long-term statins administration exacerbates diabetic nephropathy via ectopic fat deposition in diabetic mice**. *Nat. Commun.* (2023) **14** 390. DOI: 10.1038/s41467-023-35944-z
43. Fang X. **Ferroptosis as a target for protection against cardiomyopathy**. *Proc. Natl. Acad. Sci. USA* (2019) **116** 2672-2680. DOI: 10.1073/pnas.1821022116
|
---
title: 'Relationship between postpartum depression and plasma vasopressin level at
6–8 weeks postpartum: a cross-sectional study'
authors:
- Masoumeh Kashkouli
- Shahideh Jahanian Sadatmahalleh
- Saeideh Ziaei
- Anoshirvan Kazemnejad
- Ashraf Saber
- Hamid Darvishnia
- Khadijeh Azarbayjani
journal: Scientific Reports
year: 2023
pmcid: PMC9981712
doi: 10.1038/s41598-022-27223-6
license: CC BY 4.0
---
# Relationship between postpartum depression and plasma vasopressin level at 6–8 weeks postpartum: a cross-sectional study
## Abstract
Postpartum depression (PPD) is the most important postpartum mood disorder due to its significant effect on both the infant and family health. Arginine vasopressin (AVP) has been suggested as a hormonal agent involved in the development of depression. The purpose of this study was to investigate the relationship between the plasma concentrations of AVP and the score of Edinburgh Postnatal Depression Scale (EPDS). This cross-sectional study was conducted in 2016–2017 in Darehshahr Township, Ilam Province, Iran. In the first phase, 303 pregnant women, who were at 38 weeks, met the inclusion criteria, and were not depressed (according to their EPDS scores) were included in the study. In the 6–8 week postpartum follow-up, using the EPDS, 31 depressed individuals were diagnosed and referred to a psychiatrist for confirmation. The maternal venous blood samples of 24 depressed individuals still meeting the inclusion criteria and 66 randomly selected non-depressed subjects were obtained to measure their AVP plasma concentrations with ELISA assay. There was a significant positive relationship between plasma AVP levels and the EPDS score ($$P \leq 0.000$$, $r = 0.658$). Also the mean plasma concentration of AVP was significantly higher in the depressed group (41.35 ± 13.75 ng/ml) than in the non-depressed group (26.01 ± 7.83 ng/ml) ($P \leq 0.001$). In a multiple logistic regression model for various parameters, increased vasopressin levels were associated with increased odds of PPD (OR = 1.15, $95\%$ CI = 1.07–1.24, $$P \leq 0.000$$). Furthermore, multiparity (OR = 5.45, $95\%$ CI = 1.21–24.43, $$P \leq 0.027$$) and non-exclusive breastfeeding (OR = 13.06, $95\%$ CI = 1.36–125, $$P \leq 0.026$$) were associated with increased odds of PPD. Maternal gender preference (having a baby of desired and desired sex) decreased the odds of PPD (OR = 0.13, $95\%$ CI = 0.02–0.79, $$P \leq 0.027$$ and OR = 0.08, $95\%$ CI = 0.01–0.5, $$P \leq 0.007$$). AVP seems to be a contributor to clinical PPD by affecting the hypothalamic–pituitary–adrenal (HPA) axis activity. Furthermore, primiparous women had significantly lower EPDS scores.
## Introduction
Postpartum depression (PPD) is defined as an episode of major depression that is temporally related to the birth of a child1. In 2013, the American Psychiatric Association renamed the disorder to "with peripartum onset", stating that the onset of the mood disorder may be during pregnancy or within the first 4 weeks after birth2. PPD is different from postpartum blues, which is a mild mood disturbance often occurs within the first 3 to 5 days after delivery3. PPD has many negative effects on the mother and newborn; problems that may be brought for the child include developmental disorders, verbal, cognitive and social problems, and the subsequent appearance of behavioral disturbances in the child4–7. The causes of PPD are not fully understood8,9.
According to The American College of Obstetricians and Gynecologists, perinatal depression affects one out of every seven women10. The prevalence of PPD in *Iran is* reportedly $22\%$ at 6–8 weeks postpartum and $18\%$ at 12–14 weeks postpartum11. Many studies support the hypothesis that PPD is related to changes in hypothalamic–pituitary–adrenal axis12–15. Vasopressin plays an important role in the stress response and has been identified as an integral part of the hypothalamic–pituitary–adrenal (HPA) axis as a potential factor in stress-related disorders such as anxiety and depression, but the reason why the AVP system is involved in the regulation of the stress response in PPD is yet to be known16,17. Many studies have investigated the role of vasopressin in the occurrence of major depression16–18. *Vasopressin* gene is located on chromosome 2019 and the effects of vasopressin are mediated by two types of receptors (V1 and V2)20,21. Increased activity of the HPA axis in patients with major depression is one of the well-known factors that increase the secretion of corticotropin-releasing hormone (CRH), leading to an increase in secretion of corticotropin and cortisol22,23. On the other hand, PPD is a disorder caused by stress conditions24, in response to which vasopressin plays a major role in the modulation of HPA axis13,17. This hormone, in combination with CRH, can stimulate the release of adrenocorticotropic hormone (ACTH) from the anterior lobe of pituitary, and eventually, cause the release of cortisol/corticosterone from the adrenal gland25,26. Animal studies have shown that in depressed cases, vasopressin, instead of CRH, has a major stimulatory role on ACTH secretion26. There is evidence indicating that such a role for vasopressin is also conceivable in humans. For example, in depressed individuals, the number of AVP-immunoreactive neurons increases in the paraventricular nucleus (PVN), which can potentiate the effects of CRH27.
Given the possibility of a relationship between AVP and depression16–18, it seems necessary to investigate its relationship with PPD. To our knowledge, no original research has directly examined the relationship yet. The present study aimed to seek for [1] a possible relationship between plasma AVP levels and the Edinburgh Postnatal Depression Scale (EPDS) score at 6–8 weeks postpartum, and [2] a possible relationship between PPD and factors like parity, BMI, maternal gender preference, type of delivery, gender of the baby, maternal age, history of abortion, women’s education level, husband’s education level, and breastfeeding status at 6–8 weeks postpartum.
The hypotheses of the study were as follows: [1] *There is* a positive relationship between plasma AVP levels and EPDS score at 6–8 weeks postpartum, so depressed women have higher plasma levels than non-depressed women at 6–8 weeks postpartum. [ 2] *There is* a significant association between PPD and factors like parity, BMI, maternal gender preference, type of delivery, gender of the baby, maternal age, history of abortion, women’s education level, husband’s education level, and breastfeeding status at 6–8 weeks postpartum.
## Material and methods
The present cross-sectional study was conducted within 2016–2017.
## Participants
The required sample size to study on pregnant women at 38 weeks of gestation was estimated to be 303 individuals (CI = $95\%$, P-value = $5\%$). For assessing the relationship between plasma vasopressin level and EPDS score at 6–8 weeks postpartum, the sample size was calculated to be 95 persons. All the 303 subjects were selected from among those who were referred to Urban and Rural Health Care Centers for prenatal care and were not depressed according to their EPDS score (< 13). They were first briefed about the study objectives and confidentiality of maternal and neonatal information, and then their informed consent was obtained.
The study inclusion criteria were singleton pregnancy, no systemic diseases such as lupus and diabetes mellitus, no pregnancy complications like diabetes, pre-eclampsia, etc., no previous history of psychological problems, Iranian nationality, no use of antidepressants, hormonal contraceptive pills, or sleeping pills within 2 weeks prior to venous blood sampling, good marital relationship with the spouse, no expressed significant economic problems, and no family history of depression or other mental illnesses. The study exclusion criteria were experiencing stressful conditions or using alcohol within 12 h before sampling, insufficient sleep and heavy physical activity the night before sampling, abnormal blood pressure during sampling or at postpartum period, instrumental vaginal delivery, congenital malformations of the newborn, and complications during childbirth (vaginal delivery or cesarean section) leading to treatments such as blood transfusion, resuscitation, hospitalization in the ICU or CCU, or transfer to the operating room. The mothers were controlled according to the routine prenatal care program until delivery. All participants ($$n = 303$$) were once again assessed with the Edinburgh Questionnaire during 6 to 8 weeks after delivery, and if they received a score of 13 or higher, they were referred to a psychiatrist to confirm their depression. Thirty-one of them scored 13 or more; of which, PPD of 29 subjects was confirmed by the psychiatrist. Sixteen non-depressed and five depressed subjects did not meet one of the inclusion criteria or were excluded from the study. Finally, the number of subjects in the depressed and non-depressed groups were 24 and 66, respectively.
Methods of data collection included observation, examination (weight, height, BMI, and other criteria in prenatal care forms such as blood pressure, fetal heart rate, fundal height, and warning signs during pregnancy), and patient interview. Gestational age was calculated from the first day of the last menstrual period (LMP), or the first trimester ultrasound (if uncertain about LMP). Weight, blood pressure, and heart rate of the fetus were measured by the same person using a digital scale, digital barometer, and fetal heart detector (Sonicaid), respectively.
## Questionnaires
The patient interview was conducted using a questionnaire including three sections of personal information, obstetrics and medical history, and laboratory results. All the data were finally recorded in a checklist. Another questionnaire used in this study was the 10-item EPDS, designed by Cox et al.28, to evaluate maternal depression at 38 weeks of gestation and 6–8 weeks postpartum. Each question is given a score of 0 to 3. We used the Persian (Iranian language) version of the EPDS. The acceptability, reliability, and validity of the EPDS have been verified by Montazeri et al.11. Cut-off point was defined as 13 or greater.
## Determination of vasopressin level
Venous blood samples were taking from all the 24 depressed and 66 randomly selected non-depressed individuals 6 to 8 weeks after delivery at 9–9:30 h. Blood samples were taken after 15-min rest. The samples were added to the EDTA-containing chilled plastic tubes. Then they were immediately kept at 4 °C within 30 min, and plasma separation was carried out. The samples were centrifuged at 3000 rpm for 10 min at 4 °C. The produced plasma was frozen at a temperature of − 80 °C until analysis. After transferring to the Endocrinology and Metabolism Laboratories of Shahid Beheshti University of Medical Sciences, Tehran/Iran, the plasma vasopressin level was measured by the ELISA method using Human Anti-Diuretic Hormone (ADH) ELISA kit (ZellBio GmbH, Ulm Germany) with a sensitivity of 0.5 ng/ml. In our study, plasma osmolality was not measured.
The relationship between plasma vasopressin levels and the EPDS score was assessed. Then the participating mothers were divided into depressed and non-depressed groups according to the EPDS score and confirmation by psychiatrist. Next, the mean plasma AVP level was compared between the two groups. Finally, a binary logistic regression analysis was carried out to further understand the association of the independent variables, including AVP and clinical-anamnestic factors with PPD. Odds ratio with $95\%$ confidence interval was used.
## Statistic
The normal distribution of variables in each group was assessed using the Kolmogorov–Smirnov test. A Chi-squared test was used for qualitative variables when comparing the groups. Mann–Whitney test and Independent-sample T-test were applied for non-normally and normally distributed variables, respectively. Binary logistic regression was employed to assess the possible association of PPD with the variables like plasma vasopressin levels, parity, BMI, maternal gender preference, type of delivery, parity, maternal age, history of abortion, women’s education level, husband’s education level and breastfeeding status. For all statistical tests, the level of significance was considered as $P \leq 0.05.$ *Data analysis* was done using the SPSS software (ver. 21).
## Ethics approval and consent to participate
The study protocol was approved by the Institutional Review Board and the Ethics Committee of Tarbiat Modares University of Medical Sciences (IR.TMU.REC.1394.182). All procedures were in accordance with the ethical standards of the Regional Research Committee and with the Declaration of Helsinki 1964 and its later amendments. After explaining the research purposes, a written consent and a verbal assent were collected from all participants. They were also informed that their participation was voluntary, confidential and anonymous, and that they had the right to withdraw from the research at any time.
## Results
Demographic and background data for the two groups are presented in Table 1. Overall, 90 participants were included in the present study at 6–8 weeks postpartum. The results showed that 24 women had depression and 66 women had no depression. There was not a significant difference in the mean of maternal age and BMI between the depressed and non-depressed groups ($P \leq 0.05$).Table 1Demographic and clinical characteristics of the subjects ($$n = 90$$).CharacteristicsIndividual statusP-valueDepressed($$n = 24$$)Non-depressed($$n = 66$$)(N) %(N)%Women’s educationLess than high school diploma4 (16.7)8 (12.1)0.37High school diploma12 [50]25 (37..9)University8 (33.3)33 [50]Husband’s educationLess than high school diploma4 (16.7)9 (13.6)0.79High school diploma11 (45.5)27 (40.9)University9 (37.5)30 (45.5)History of abortionYes1 (4.2)8 (12.1)0.43No23 (95.8)58 (87.9)Maternal gender preferenceDesired sex4(33.3)20(41.7)0.63Non-desired sex8 (66.7)28 (58.3)Baby genderMale11 (45.8)34 (51.5)0.63Female13 (54.2)32 (48.5)Mode of deliveryNVD9 [25]27 [75]0.77CS15 (27.8)39 (72.2)ParityNulliparous9 (37.5)39 (59.1)0.06Multiparous15 (62.5)27 (40.9)*Breastfeeding status* at 6–8 week postpartumNo exclusive breastfeeding6 [25]2 [3]0.004Exclusive breastfeeding18 [75]64 [97]Living areaCity17 (70.8)51 (77.3)0.53Village7 (29.2)15 (22.7)
## Prevalence of PPD
A total of 303 women at 38 weeks of pregnancy from the Urban and Rural Health Care Centers met the inclusion criteria. The mean score of EPDS scale was higher in the depressed women (16.46 ± 3.10) as compared with the non-depressed women (6.12 ± 13.32) ($P \leq 0.001$). Finally, after confirmation of PPD by the psychiatrist, the prevalence of PPD was determined as $9.57\%$.
## Association of the plasma concentrations of AVP and clinical- anamnestic factors with PPD
The Pearson’s correlation test showed a positive significant relationship between the plasma concentrations of AVP and EPDS score in the depressed group ($P \leq 0.001$, $r = 0.65$). According to the Independent-sample T test results, the mean plasma concentrations of AVP was significantly higher in the depressed group (41.35 ± 13.75 ng/ml) than in the non-depressed group (26.01 ± 7.83 ng/ml) (t (28.6) = 5.16, $P \leq 0.001$). We performed binary logistic regression to investigate the association of plasma vasopressin levels and clinical-anamnestic factors, (including parity, BMI, maternal gender preference, type of delivery, gender of the baby, parity, maternal age, history of abortion, women’s education level, husband’s education level and breastfeeding status) with PPD. After analysis, plasma vasopressin levels, parity, maternal gender preference and breastfeeding status showed statically significant relationship with PPD. In the final prediction model, the increasing values of plasma vasopressin corresponded with increasing of the odds of PPD (OR = 1.15, $95\%$ CI = 1.07–1.24, $P \leq 0.001$).
The association between the clinical-anamnestic factors, including mode of delivery ($$P \leq 0.77$$), gender of the baby ($$P \leq 0.63$$), living area ($$p \leq 0.53$$), women’s education level ($$P \leq 0.37$$), husband’s education level ($$P \leq 0.79$$), maternal gender preference ($$P \leq 0.63$$), parity ($$P \leq 0.06$$), gender of the baby ($$P \leq 0.63$$), history of abortion ($$P \leq 0.43$$), breastfeeding status ($$P \leq 0.004$$) and PPD was quantified by Chi-squared test. Among the above variables, only breastfeeding status was related to PPD ($$P \leq 0.004$$) (Table 1).
Using binary logistic regression analysis, the non-exclusively breastfeeding women were found to experience PPD 13 times more than the exclusively breastfeeding (OR = 13.06, $95\%$ CI = 1.36–125, $$P \leq 0.026$$). The exclusively breastfeeding women were divided into depressed and non-depressed groups. We ran a Mann–Whitney’s U test because of the abnormal distribution of AVP levels in the non-depressed group to evaluate the difference in the concentrations of AVP in the exclusively breastfeeding depressed and non-depressed women. The mean plasma AVP concentrations in the depressed and non-depressed groups were 66.73 and 37.78 ng/ml, respectively, showing a significant difference ($U = 282$, Z = − 4.64, $P \leq 0.001$).
An independent T-test was conducted to compare the mean of maternal age between the two groups. There was not a significant difference in the mean of maternal age between the depressed and non-depressed groups.
Multiparous (OR = 5.45, $95\%$ CI = 1.21–24.43, $$P \leq 0.027$$) were more likely to be depressed. Women with no gender preference were more likely to be depressed than ones who had a baby of the non-desired (OR = 0.13, $95\%$ CI = 0.02–0.79, $$P \leq 0.027$$) and desired (OR = 0.08, $$P \leq 0.007$$, $95\%$ CI = 0.01–0.5) sex (Table 2).Table 2Logistic regression analysis of factors affecting postpartum depression. PredictorsAdjusted ORCI$95\%$P-value*Vasopressin1.151.07–1.24 < 0.001ParityPrimiparous1Multiparous5.451.21–24.430.027Breastfeeding statusExclusive1Non-exclusive13.061.36–1250.026Maternal gender preferenceNo preference1Desired sex of the baby0.080.01–0.0.50.007Non-desired sex of the baby0.130.02–0.790.027R2 = 0.41.*P values are adjusted for education level, history of abortion, baby gender, mode of delivery, and living area.
## Discussion
The main aim of this study was to investigate the association between PPD and the plasma concentrations of AVP. Based on our findings, the prevalence of PPD (as confirmed by the psychiatrist) was $9.57\%$ at 6–8 weeks postpartum.
In the present study, there was a significant positive correlation between the mean plasma concentration of AVP and EPDS score; the mean plasma concentration of AVP was significantly higher in the depressed group than in the non-depressed group. Worldwide epidemiological studies show that depression occurs in women twice as much as in men and peaks during the first year postpartum. A few previous studies on the relationship between AVP and depression in non-pregnant populations are as follows. In a 2006 review article, Keck concluded that a dysfunction in AVP and corticotropin-releasing factor (CRF) systems is involved in the pathogenesis and etiology of depression and anxiety29. Increased plasma levels of AVP were suggested in depressed patients30. In the study of Van Londen et al., the mean plasma concentration of AVP was higher in the depressed patients than in the control group18. In chronic stress, vasopressin level is steadily elevated in CRH neurons so that even a slight irritation may lead to a severe AVP/CRH secretion. This can contribute to the development of dysphoric symptoms and may even be a factor responsible for the onset and permanent characteristics of major depression31. The number of CRH neurons, co-express vasopressin, increases in depressed human PVN32. Elevated AVP mRNA level in the supraoptic nucleus (SON) of depressed individuals33 also leads to an increase in plasma vasopressin level26, which is related to suicide risk34 and altered attention and arousal in memory processes in these patients35. In contrast, Wang et al. found no significant difference in vasopressin mRNA expression in the SON or PVN of depressed individuals comparing to the control group, probably due to the mere presence of melancholic depressed individuals in this study36. Brunner et al. could not find any association between vasopressin concentration in plasma and cerebrospinal fluid (CSF) among the depressed and non-depressed individuals. The two were also not different in those who attempted suicide compared to other people37. This finding is not in line with the results of the present study. It might be due to low sample size and diagnostic heterogeneity in the patient and control groups in the above study, which did not allow them to investigate the pathogenetic role of vasopressin.
Griebel et al. suggested that although CRF has been recognized as a key regulator of the stress system, there is evidence that the vasopressinergic system may play an equal role in regulating the stress responses; so vasopressin V (1b) receptor antagonists may be potential treatment for depression. This finding confirms the efficacy of using SSR149415, which was the first selective, orally active vasopressin V (1b) receptor antagonist in the treatment of depression and anxiety disorders caused by traumatic events38. Vasopressin V (1b) receptor is required for the normal response of the HPA axis during chronic stress39. Muller et al. ’s study on mice showed a selective compensatory activation of the hypothalamic vasopressinergic system for maintaining basal ACTH secretion and HPA system activity in heterozygous and homozygous Crhr1−/− mutants in basal and stress conditions. Deficiency of CRH receptor 1 (CRHR1) severely impairs the stress response of the HPA system and reduces stress-related behavior in mice40.
The research findings revealed a significant positive relationship between the plasma concentrations of AVP and the EPDS score at 6 to 8 weeks postpartum. The mean plasma concentrations of AVP were significantly higher in the depressed group (41.35 ± 13.75 ng/ml) than in the non-depressed group (26.01 ± 7.83 ng/ml) (t (28.6) = 5.16, $P \leq 0.001$). In this study, all subjects were controlled for the involvement of other factors affecting plasma vasopressin concentration. The obtained results support the hypothesis of a relationship between AVP and PPD. Since PPD is classified as a sub-type of major depressive disorder, our findings also support the possible role of vasopressin in the development of major depression.
According to the logistic regression results, primiparous status was a significant negative predictor of depression. Some studies show that primiparous women are more anxious during pregnancy comparing to the postpartum period. Our study further confirms that since most of primiparous women are happy with the birth of a newborn and addition of a new member to the family, their EPDS score lower comparing to other women. In a study by Righetti-Veltema et al. in Geneva on the risk factors and predictive signs of PPD at 3 months after delivery, the multiparas reported more difficult pregnancies and higher anxiety and were less involved in the perinatal preparation program in comparison to the control group41. Figueiredo and Conde reported that parity has a significant impact on postpartum anxiety and PPD. According to their study, second-time parents showed more anxiety and depression symptoms than first-time parents in the second and third trimesters and also 3 months after childbirth, but not at birth42. These results are in line with our results. In contrast, O'Hara et al. reported that pregnant and primiparous women experienced more depressive symptoms, marital distress, and psychological problems than non-pregnant women43. In Kaij et al. ’s study, symptoms of depression decreased (using the EDPS scale) with the increasing number of children44. Gotlib et al. showed that multiparity was a risk factor for depression during pregnancy, but not in the postpartum period45. Wenzel et al. claimed that there was typically no relationship between parity and depression during pregnancy and postpartum46, which is consistent with the findings of Lashkaripour et al.47. None of the latter four studies could make a clear and well-established conclusion regarding the relationship between primiparous status and PPD. These conflicting results can be due to different tools used for PPD assessment or different times of evaluation. There are also unclear results about the role of other factors affecting the relationship between parity and PPD such as parents’ education level, socioeconomic factors, and maternal age. We excluded people who were economically disadvantaged to control this variable.
In the present study, 60 women ($66.6\%$ of all participating women) had gender preference; of which, $40\%$ had a baby of the non-desired sex and only $20\%$ of them developed PPD. *In* general, only $16.6\%$ of all PPD women had a baby of the non-desired sex. The majority of the PPD women had not gender preference. Depression in the two groups of women whether their baby had the same gender as they desired before birth (or not) was less than in the group of women who did not have preference about the sex of the baby. Preference of boy baby is still common in the countries like India, China and some other parts of the world, which is deeply rooted in their cultural issues. In these societies, the girl is considered merely an economic consumer of the family as most girls get married, are paid dowry, and then leave the family48,49. In one study, having a baby of the non-desired sex increased the risk of developing PPD50. Although the results of these studies contradict our findings (Table 2), similar to ours, Dhillon and MacArthur identified that maternal gender preference in Asian women living in the UK had no association with PPD51.
Within the total research sample at 6–8 weeks postpartum, 82 women breastfed exclusively ($91.1\%$) and 8 women did not initiate or ceased exclusive breastfeeding. In our study, exclusive breastfeeding included Labbock and Krasovec’s levels 1 and 252. A number of studies have reported no relationship between breastfeeding and PPD53 while some others have shown that breastfeeding may protect against PPD52,53; the latter is consistent with the present study results, suggesting that exclusive breastfeeding may help to reduce mother’s PPD. In our study, women who exclusively breastfed their neonates at 6–8 weeks postpartum were less likely to be depressed.
## Study limitations
This study provides useful data on the relationship between plasma AVP and PPD. However, we had some limitations in conducting the study. For example, it would be better to evaluate plasma concentrations of AVP by Radioimmunoassay (RIA). However, we could not do it because of lack of access to Human Vasopressin Radioimmunoassay (RIA) kits in Iran due to sanctions.
## Conclusion
To the best of our knowledge this is the first paper on the relationship between maternal plasma AVP levels and PPD. The results revealed that the mean plasma AVP level was higher in the depressed group comparing to the non-depressed group. Also the mean plasma concentration of AVP was higher in the exclusively breastfed depressed women comparing to their non-depressed counterparts. Additionally, being a primiparous, exclusive breast-feeding and maternal gender preference were negatively related to PPD. Vasopressin appears to be related to PPD by contributing to the dysregulation of the HPA axis. It is recommended that future studies be conducted with larger sample sizes and longer follow-up periods.
## References
1. Goodman SH, Rouse MH, Connell AM, Broth MR, Hall CM, Heyward D. **Maternal depression and child psychopathology: A meta-analytic review**. *Clin. Child. Fam. Psychol. Rev.* (2011) **14** 1-27. DOI: 10.1007/s10567-010-0080-1
2. 2.American Psychiatric Association ADiagnostic and Statistical Manual of Mental Disorders1980American Psychiatric Association. *Diagnostic and Statistical Manual of Mental Disorders* (1980)
3. O'Hara MW, McCabe JE. **Postpartum depression: Current status and future directions**. *Annu. Rev. Clin. Psychol.* (2013) **9** 379-407. DOI: 10.1146/annurev-clinpsy-050212-185612
4. Poobalan AS, Aucott LS, Ross L, Smith WC, Helms PJ, Williams JH. **Effects of treating postnatal depression on mother-infant interaction and child development: Systematic review**. *Br. J. Psychiatry* (2007) **191** 378-386. DOI: 10.1192/bjp.bp.106.032789
5. Dave S, Sherr L, Senior R, Nazareth I. **Associations between paternal depression and behaviour problems in children of 4–6 years**. *Eur. Child Adolesc. Psychiatry* (2008) **17** 306-315. DOI: 10.1007/s00787-007-0672-6
6. Kompier NF, Keysers C, Gazzola V, Lucassen PJ, Krugers HJ. **Early life adversity and adult social behavior: Focus on arginine vasopressin and oxytocin as potential mediators**. *Front. Behav. Neurosci.* (2019) **13** 143. DOI: 10.3389/fnbeh.2019.00143
7. Brummelte S, Galea LA. **Depression during pregnancy and postpartum: Contribution of stress and ovarian hormones**. *Prog. Neuropsychopharmacol. Biol. Psychiatry* (2010) **34** 766-776. DOI: 10.1016/j.pnpbp.2009.09.006
8. McCoy SJB, Beal JM, Shipman SBM, Payton ME, Watson GH. **Risk factors for postpartum depression: A retrospective investigation at 4-weeks postnatal and a review of the literature**. *J. Am. Osteopath. Assoc.* (2006) **106** 193-198. PMID: 16627773
9. Beck CT. **Predictors of postpartum depression: An update**. *Nurs. Res.* (2001) **50** 275-285. DOI: 10.1097/00006199-200109000-00004
10. **Screening for perinatal depression**. *Obstet. Gynecol.* (2015) **125** 1268-1271. DOI: 10.1097/01.AOG.0000465192.34779.dc
11. Montazeri A, Torkan B, Omidvari S. **The Edinburgh Postnatal Depression Scale (EPDS): Translation and validation study of the Iranian version**. *BMC Psychiatry* (2007) **7** 1-6. DOI: 10.1186/1471-244X-7-11
12. Jolley SN, Elmore S, Barnard KE, Carr DB. **Dysregulation of the Hypothalamic-Pituitry-Adrenal Axis in postpartum deprssion**. *Biol. Res. Nurs.* (2007) **8** 210-222. DOI: 10.1177/1099800406294598
13. Solomonova E, Lee YEA, Robins S, King L, Feeley N, Gold I. **Sleep quality is associated with vasopressin methylation in pregnant and postpartum women with a history of psychosocial stress**. *Psychoneuroendocrinology* (2019) **107** 160-168. DOI: 10.1016/j.psyneuen.2019.05.010
14. Hendrick V, Altshuler LL, Suri R. **Hormonal changes in the postpartum and implications for postpartum depression**. *Psychosomatics* (1998) **39** 93-101. DOI: 10.1016/S0033-3182(98)71355-6
15. Meltzer-Brody S. **New insights into perinatal depression: Pathogenesis and treatment during pregnancy and postpartum**. *Dialogues Clin. Neurosci.* (2011) **13** 89. DOI: 10.31887/DCNS.2011.13.1/smbrody
16. Goekoop JG, De Winter RP, de Rijk R, Zwinderman KH, Frankhuijzen-Sierevogel A, Wiegant VM. **Depression with above-normal plasma vasopressin: Validation by relations with family history of depression and mixed anxiety and retardation**. *Psychiatry Res.* (2006) **141** 201-211. DOI: 10.1016/j.psychres.2005.09.003
17. Morales-Medina JC, Witchey SK, Caldwell HK, López-Muñoz F, Srinivasan V, de Berardis D, Álamo C, Kato TA. **The role of vasopressin in anxiety and depression**. *Melatonin, Neuroprotective Agents and Antidepressant Therapy* (2016) 667-685
18. van Londen L, Goekoop JG, van Kempen GM, Frankhuijzen-Sierevogel AC, Wiegant VM, van der Velde EA. **Plasma levels of arginine vasopressin elevated in patients with major depression**. *Neuropsychopharmacology* (1997) **17** 284-292. DOI: 10.1016/S0893-133X(97)00054-7
19. Alescio-Lautier B, Paban V, Soumireu-Mourat B. **Neuromodulation of memory in the hippocampus by vasopressin**. *Eur. J. Pharmacol.* (2000) **405** 63-72. DOI: 10.1016/S0014-2999(00)00542-2
20. Caffe A, Van Leeuwen F, Luiten P. **Vasopressin cells in the medial amygdala of the rat project to the lateral septum and ventral hippocampus**. *J. Comp. Neurol.* (1987) **261** 237-252. DOI: 10.1002/cne.902610206
21. Ueta Y, Taniguchi S, Yoshida A, Murakami I, Mitani Y, Hisatome I. **A new type of familial central diabetes insipidus caused by a single base substitution in the neurophysin II coding region of the vasopressin gene**. *J. Clin. Endocrinol. Metab.* (1996) **81** 1787-1790. PMID: 8626836
22. Rein T, Ambrée O, Fries GR, Rappeneau V, Schmidt U, Touma C. *The Hypothalamic-Pituitary-Adrenal Axis in Depression: Molecular Regulation, Pathophysiological Role, and Translational Implications* (2019) 89-96
23. Naert G, Ixart G, Maurice T, Tapia-Arancibia L, Givalois L. **Brain-derived neurotrophic factor and hypothalamic-pituitary-adrenal axis adaptation processes in a depressive-like state induced by chronic restraint stress**. *Mol. Cell. Neurosci.* (2011) **46** 55-66. DOI: 10.1016/j.mcn.2010.08.006
24. Seth S, Lewis AJ, Galbally M. **Perinatal maternal depression and cortisol function in pregnancy and the postpartum period: A systematic literature review**. *BMC Pregnancy Childbirth.* (2016) **16** 1-19. PMID: 26728010
25. Antoni FA. **Vasopressinergic control of pituitary adrenocorticotropin secretion comes of age**. *Front. Neuroendocrinol.* (1993) **14** 76-122. DOI: 10.1006/frne.1993.1004
26. Katz DA, Locke C, Greco N, Liu W, Tracy KA. **Hypothalamic-pituitary-adrenal axis and depression symptom effects of an arginine vasopressin type 1B receptor antagonist in a one-week randomized Phase 1b trial**. *Brain Behav.* (2017) **7** e00628. DOI: 10.1002/brb3.628
27. Purba JS, Hoongendijk WJ, Hofman MA, Swaab DF. **Increased number of vasopressin- and oxytocin-expressing neurons in the paraventricular nucleus of the hypothalamus in depression**. *Arch. Gen. Psychiatry.* (1996) **53** 137-143. DOI: 10.1001/archpsyc.1996.01830020055007
28. Cox JL, Holden JM, Sagovsky R. **Detection of postnatal depression. Development of the 10-item Edinburgh Postnatal Depression Scale**. *Br. J. Psychiatry* (1987) **150** 782-786. DOI: 10.1192/bjp.150.6.782
29. Keck M. **Corticotropin-releasing factor, vasopressin and receptor systems in depression and anxiety**. *Amino Acids* (2006) **31** 241-250. DOI: 10.1007/s00726-006-0333-y
30. Abdel Mawella SM. **Psychoendocrinology: Argenine vasopressin and resilience in patients with major depressive disorder**. *CNS Spectr.* (2021) **28** 41-45. DOI: 10.1017/S1092852921000730
31. Griffiths J, Ravindran A, Merali Z, Anisman H. **Dysthymia: A review of pharmacological and behavioral factors**. *Mol. Psychiatry* (2000) **5** 242-261. DOI: 10.1038/sj.mp.4000697
32. Bao A-M, Swaab DF. **Gender difference in age-related number of corticotropin-releasing hormone-expressing neurons in the human hypothalamic paraventricular nucleus and the role of sex hormones**. *Neuroendocrinology* (2007) **85** 27-36. DOI: 10.1159/000099832
33. Meynen G, Unmehopa UA, van Heerikhuize JJ, Hofman MA, Swaab DF, Hoogendijk WJ. **Increased arginine vasopressin mRNA expression in the human hypothalamus in depression: A preliminary report**. *Biol. Psychiatry.* (2006) **60** 892-895. DOI: 10.1016/j.biopsych.2005.12.010
34. Pitchot W, Scantamburlo G, Pinto E, Hansenne M, Reggers J, Ansseau M. **Vasopressin–neurophysin and DST in major depression: Relationship with suicidal behavior**. *J. Psychiatr. Res.* (2008) **42** 684-688. DOI: 10.1016/j.jpsychires.2007.07.007
35. van Londen L, Kerkhof GA, van den Berg F, Goekoop JG, Zwinderman KH, Frankhuijzen-Sierevogel AC. **Plasma arginine vasopressin and motor activity in major depression**. *Biol. Psychiatry.* (1998) **43** 196-204. DOI: 10.1016/S0006-3223(97)80433-7
36. Wang S, Kamphuis W, Huitinga I, Zhou J, Swaab DF. **Gene expression analysis in the human hypothalamus in depression by laser microdissection and real-time PCR: The presence of multiple receptor imbalances**. *Mol. Psychiatry* (2008) **13** 786-799. DOI: 10.1038/mp.2008.38
37. Brunner J, Keck ME, Landgraf R, Uhr M, Namendorf C, Bronisch T. **Vasopressin in CSF and plasma in depressed suicide attempters: Preliminary results**. *Eur. Neuropsychopharmacol.* (2002) **12** 489-494. DOI: 10.1016/S0924-977X(02)00071-8
38. Griebel G, Simiand J, Stemmelin J, Gal C, Steinberg R. **The vasopressin V1b receptor as a therapeutic target in stress-related disorders**. *Curr. Drug Targets-CNS Neurol. Disord.* (2003) **2** 191-200. DOI: 10.2174/1568007033482850
39. Lolait SJ, Stewart LQ, Jessop DS, Young WS, O’Carroll A-M. **The hypothalamic-pituitary-adrenal axis response to stress in mice lacking functional vasopressin V1b receptors**. *Endocrinology* (2007) **148** 849-856. DOI: 10.1210/en.2006-1309
40. Müller MB, Landgraf R, Preil J, Sillaber I, Kresse AE, Zimmermann S, Holsboer F, Wurst W. **Selective activation of the hupothalamic vasopressinergic system in mice deficient for the corticotropin-relesing hormone receptor 1 is dependent on glucocorticoids**. *Endocrinology* (2000) **141** 4262-4269. DOI: 10.1210/endo.141.11.7767
41. Righetti-Veltema M, Conne-Perréard E, Bousquet A, Manzano J. **Risk factors and predictive signs of postpartum depression**. *J. Affect. Disord.* (1998) **49** 167-180. DOI: 10.1016/S0165-0327(97)00110-9
42. Figueiredo B, Conde A. **Anxiety and depression symptoms in women and men from early pregnancy to 3-months postpartum: Parity differences and effects**. *J. Affect. Disord.* (2011) **132** 146-157. DOI: 10.1016/j.jad.2011.02.007
43. O'Hara MW, Rehm LP, Campbell SB. **Predicting depressive symptomatology: Cognitive-behavioral models and postpartum depression**. *J. Abnorm. Psychol.* (1982) **91** 457. DOI: 10.1037/0021-843X.91.6.457
44. Kaij L, Jacobson L, Nilsson Å. **Post-partum mental disorder in an unselected sample: The influence of parity**. *J. Psychosom. Res.* (1967) **10** 317-325. DOI: 10.1016/0022-3999(67)90068-2
45. Gotlib IH, Whiffen VE, Mount JH, Milne K, Cordy NI. **Prevalence rates and demographic characteristics associated with depression in pregnancy and the postpartum**. *J. Consult. Clin. Psychol.* (1989) **57** 269. DOI: 10.1037/0022-006X.57.2.269
46. Wenzel A, Haugen EN, Jackson LC, Brendle JR. **Anxiety symptoms and disorders at eight weeks postpartum**. *J. Anxiety Disord.* (2005) **19** 295-311. DOI: 10.1016/j.janxdis.2004.04.001
47. Lashkaripour K, Bakhshani NM, Hokmabadi S, Sajjadi SAR, Safarzadeh SA. **Postpartum depression and related factors: A 4.5 months study**. *J. Fundam. Ment. Health.* (2012) **4** 404-412
48. Xie R-H, He G, Liu A, Bradwejn J, Walker M, Wen SW. **Fetal gender and postpartum depression in a cohort of Chinese women**. *Soc. Sci. Med.* (2007) **65** 680-684. DOI: 10.1016/j.socscimed.2007.04.003
49. Rodrigues M, Patel V, Jaswal S, De Souza N. **Listening to mothers: Qualitative studies on motherhood and depression from Goa, India**. *Soc. Sci. Med.* (2003) **57** 1797-1806. DOI: 10.1016/S0277-9536(03)00062-5
50. Boyce P, Hickey A. **Psychosocial risk factors to major depression after childbirth**. *Soc. Psychiatry Psychiatr. Epidemiol.* (2005) **40** 605-612. DOI: 10.1007/s00127-005-0931-0
51. Dhillon N, MacArthur C. **Antenatal depression and male gender preference in Asian women in the UK**. *Midwifery* (2010) **26** 286-293. DOI: 10.1016/j.midw.2008.09.001
52. Labbok MH, Coffin CJ. **A call for consistency in definition of breastfeeding behaviors**. *Soc. Sci. Med.* (1997) **44** 1931-1932. DOI: 10.1016/S0277-9536(97)00013-0
53. Chaudron LH, Klein MH, Remington P, Palta M, Allen C, Essex MJ. **Predictors, prodromes and incidence of postpartum depression**. *J. Psychosom. Obstet. Gynaecol.* (2001) **22** 103-112. DOI: 10.3109/01674820109049960
|
---
title: 'Low-temperature Mössbauer spectroscopy of organs from 57Fe-enriched HFE(−/−)
hemochromatosis mice: an iron-dependent threshold for generating hemosiderin'
authors:
- Shaik Waseem Vali
- Paul A. Lindahl
journal: Journal of Biological Inorganic Chemistry
year: 2022
pmcid: PMC9981716
doi: 10.1007/s00775-022-01975-y
license: CC BY 4.0
---
# Low-temperature Mössbauer spectroscopy of organs from 57Fe-enriched HFE(−/−) hemochromatosis mice: an iron-dependent threshold for generating hemosiderin
## Abstract
Hereditary hemochromatosis is an iron-overload disease most often arising from a mutation in the Homeostatic Fe regulator (HFE) gene. HFE organs become overloaded with iron which causes damage. Iron-overload is commonly detected by NMR imaging, but the spectroscopic technique is insensitive to diamagnetic iron. Here, we used Mössbauer spectroscopy to examine the iron content of liver, spleen, kidney, heart, and brain of 57Fe-enriched HFE(−/−) mice of ages 3–52 wk. Overall, the iron contents of all investigated HFE organs were similar to the same healthy organ but from an older mouse. Livers and spleens were majorly overloaded, followed by kidneys. Excess iron was generally present as ferritin. Iron–sulfur clusters and low-spin FeII hemes (combined into the central quadrupole doublet) and nonheme high-spin FeII species were also observed. Spectra of young and middle-aged HFE kidneys were dominated by the central quadrupole doublet and were largely devoid of ferritin. Collecting and comparing spectra at 5 and 60 K allowed the presence of hemosiderin, a decomposition product of ferritin, to be quantified, and it also allowed the diamagnetic central doublet to be distinguished from ferritin. Hemosiderin was observed in spleens and livers from HFE mice, and in spleens from controls, but only when iron concentrations exceeded 2–3 mM. Even in those cases, hemosiderin represented only 10–$20\%$ of the iron in the sample. NMR imaging can identify iron-overload under non-invasive room-temperature conditions, but Mössbauer spectroscopy of 57Fe-enriched mice can detect all forms of iron and perhaps allow the process of iron-overloading to be probed in greater detail.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00775-022-01975-y.
## Introduction
Hereditary hemochromatosis is an iron-overload disease affecting 1 in 200 humans of northwestern European ancestry [1, 2]. Its most common form arises from a C282Y mutation of the Homeostatic Fe regulator (HFE) gene. HFE is involved in the biosynthesis of hepcidin, a peptide generated by hepatocytes. Hepcidin controls iron export from enterocytes, which line the basolateral surface of the duodenum, and from reticuloendothelial macrophages, which are found in the spleen, liver, and elsewhere. Hepcidin regulates iron import into the blood by binding and inactivating ferroportin (FPN), a membrane-bound iron export protein. Binding promotes the translocation of FPN to lysosomes where it is hydrolyzed; this halts iron export into plasma.
In healthy individuals, iron that enters the blood binds transferrin (TFN). Transferrin-bound iron is distributed to cells of the body by binding a receptor on the plasma membrane. TFN enters cells via receptor-mediated endocytosis. In healthy individuals, ~ $30\%$ of TFN is holo, while most of the remainder is apo. Because both forms are present at significant concentrations, TFN serves as an iron buffer that can absorb a bolus of nutrient iron and also release iron to cells. Hepatocytes in individuals with hemochromatosis produce insufficient hepcidin. This causes excessive nutrient iron to enter the blood and TFN to saturate. Once saturated, additional iron enters plasma as Non-Transferrin-Bound Iron or NTBI which enters organs and overloads them with iron [3].
In individuals with hemochromatosis, the liver is the first organ to become overloaded [4]. Once in cells, the excess iron is converted into either ferritin or hemosiderin [5]. Ferritin is an approximately spherically shaped iron-storage protein with a hollow core that can be filled with FeIII nanoparticles. Hemosiderin is an insoluble, amorphous, and heterogeneous degradation product of ferritin located mainly in secondary lysosomes [6]. Excess iron in the liver can cause fibrosis, cirrhosis, and cancer. Other organs also accumulate iron but generally have a lower iron-binding capacity than the liver. However, they may be more easily damaged by iron accumulation [7, 8].
Cells of the reticuloendothelial system, including Kupffer macrophages in the liver, red-pulp macrophages in the spleen, and central nurse macrophages in the bone marrow, are especially sensitive to iron-overload. These cells recycle most bodily iron for erythropoiesis [9]. Red-pulp macrophages extract iron from senescent erythrocytes by degrading hemoglobin. Excessive iron increases hepatic hepcidin expression which causes macrophages to sequester iron [10]. Iron can deposit in splenic macrophages in the form of hemosiderin [11, 12].
Iron in the HFE kidney accumulates mainly in the medulla. Iron is removed from plasma at the glomerulus associated with Bowman’s capsule and is then reabsorbed in the collecting tubules associated with the Loop of Henle [13–15].
Excessive iron in the heart accumulates in cardiomyocytes where it can cause arrhythmias and congestive heart failure [16, 17]. The heart imports less ferric citrate (injected into rodents as an NTBI mimic) than the liver, likely due to the lower expression levels of the divalent metal importer Zip14 [18, 19]. However, it also imports NTBI through calcium channels [6, 18–22].
The brain accumulates a modest amount of iron in iron-overloaded mice [23–25]. NTBI is imported into the brain of hypotransferrinemic mice [26–28]. Individuals with β-thalassemia, another iron-overload disease, have iron deposits in the anterior pituitary [29]. These deposits can alter the production of hormones and lead to hypogonadism. Radioactive 59Fe injected into WT mice was detected in the ventricles within 2 h [30]. Within 24 h, such iron spread throughout the brain, with especially high concentrations localized in the choroid plexus.
Iron-overload in organs is most popularly detected in vivo by an NMR method that is sensitive to superparamagnetic materials such as ferritin [31]. This method, which involves monitoring proton transverse relaxation times (T2*), is widely used, because it is non-invasive [32]. Hocq et al. [ 33] used NMR spectroscopy to estimate the iron content of the liver, spleen, and brain in four human donors (without iron-overload). Ferritin iron darkens T2-weighted NMR images. Relaxation times on brain, liver, and spleen samples can be measured at different magnetic fields. However, the correlation with total iron content is imperfect. T2* values are affected by the size and density of ferritin iron cores and, more importantly, are insensitive to diamagnetic forms of iron such as [Fe4S4]2+ and [Fe2S2]2+ clusters or as low-spin FeII hemes.
Low-temperature AC magnetic susceptibility has also been used to investigate iron-overloaded organs. Gutierrez et al. [ 34] characterized liver, spleen, and heart tissues of DBA/2 HFE knockout mice using this method together with transmission electron microscopy and Selected Area Electron Diffraction to investigate the chemical iron speciation in mice with overload diseases. In that study, iron accumulated in the liver in 9 wk HFE vs WT mice, but no differences between HFE and WT mice spleens and hearts in terms of iron-overload were observed. There was some evidence of ferritin degrading and hemosiderin forming. The same group examined 12 wk HFE and WT mice; HFE mice accumulated ferritin in livers, but the amount of ferritin in kidney and heart was minor and similar for HFE and WT mice.
We have previously used Mössbauer (MB) spectroscopy to examine the iron content of brain [35], liver [36], and heart [37] from 57Fe-enriched healthy control mice of various ages. We also examined hearts and livers from 12 wk 57Fe-enriched HFE mice (the natural lifespan of mice is ~ 100 wk). For healthy organs, the two dominant iron species observed by MB include ferritin and a combination of [Fe4S4]2+ clusters and low-spin FeII hemes. The latter two species collectively yield the “central quadrupole doublet”, called the CD (δ = 0.45 mm/s; ΔEQ = 1.15 mm/s) in MB spectra. Viewed simplistically, ferritin reflects iron that is stored, whereas the CD reflects iron that is used—mainly but not exclusively in mitochondrial respiration. The absolute and relative amounts of ferritin vs. the CD change with the organ and with the age of the mouse. A small contribution due to nonheme high-spin (NHHS) FeII has also been observed as a quadrupole doublet in MB spectra. NHHS FeII likely reflects the labile FeII pool in cells, though other FeII species in the sample must also contribute.
Healthy newborn livers contain a high concentration of iron, mainly in the form of ferritin [36]. Within the first few weeks of life, much of this iron exits the liver and is delivered to other organs as needed for normal development. At 3–4 wk of age, most iron in the liver is present as the CD. As animals age, ferritin accumulates. After the initial exodus, the iron concentration within liver cells (with blood contributions removed) is only ~ 300 µM. Diseased livers (HFE and IRP2-deficient) contain significantly higher concentrations of iron, in the form of ferritin.
In contrast, the iron content of the heart is dominated by the CD [37]. In fact, young hearts contain little if any ferritin. Ferritin levels in the heart increase with age. Brains from fetuses contain ~ 270 µM Fe, mostly in the form of ferritin [35]. The concentration of iron declines in newborn and young brains (to ~ 120 µM Fe), due to the expanding volume of the developing brain. The iron concentration in adult brains is only ~ 200 µM. With age, more ferritin iron accumulates.
MB spectroscopy has also been used by other groups to characterize the iron content of iron-overloaded organs. Hemosiderin was found to be present and even dominate the iron in iron-overloaded organs. This form of iron can be distinguished from ferritin by comparing spectra collected at 5 K and $\frac{60}{70}$ K. At the higher temperatures, magnetic interactions due to ferritin iron collapse, whereas those due to hemosiderin do not [38, 39]. Ferritin exhibits a sextet at ~ 5 K, but by 60–70 K, it exhibits a quadrupole doublet [38, 40]. Seldon et al. [ 41] concluded that hemosiderin is the dominant form of iron in iron-overload diseases and that it is responsible for organ damage. St. Pierre et al. [ 42] concurred that hemosiderin (rather than ferritin) damages iron-overloaded organs. The same group collected MB spectra of spleens, pancreas, heart, and livers of β-thalassemia patients [40]. They observed a sextet at 12 K and a broad singlet at 78 K, with parameters of high-spin FeIII. They detected hemosiderin in spleens and pancreas, and concluded that there are at least three forms of hemosiderin. Spectra of whole tissues were composed of a superposition of features from hemosiderin and ferritin. Ward et al. [ 43] found that hemosiderin was the major iron-storage protein in tissues of iron-overloaded individuals; it accounted for three-quarters of the iron in iron-loaded human livers. Webb et al. [ 44] used MB to examine the pancreas of humans with β-thalassemia, and isolated ferritin and hemosiderin from tissues. Using MB spectroscopy, St. Pierre et al. [ 45, 46] found that hemosiderin was the major form of iron in spleens from human β-thalassemia patients and as well as iron-overloaded spleens and livers from rodents. Other studies reported a closer balance between the two. Chua-anusorn [39] used 12 K and 60 K MB to examine iron deposits in human thalassemic heart tissue. About $40\%$ of the iron was hemosiderin and $35\%$ was ferritin ($20\%$ was unassigned). Gutierrez et al. [ 47] found that ferritin rather than hemosiderin dominated the iron content of HFE mice.
In our previous studies, we did not observe hemosiderin in any healthy or diseased organ. Rather, we exclusively observed ferritin, ISCs, hemes, and small contributions of NHHS FeII. However, we only examined one liver and one heart from HFE mice, and then only at a relatively young age (12 wk) [37]. Here, we used variable temperature MB spectroscopy to examine livers, hearts, spleens, kidneys, and brains from 57Fe-enriched HFE mice of different ages. Organs were perfused with buffer to remove excess blood, dissected under anaerobic conditions, loaded immediately into MB cups, frozen, and evaluated using MB spectroscopy. We characterized the type of iron that accumulated as well as the age-dependence of that accumulation. We provide evidence that hemosiderin is generated when iron in organs exceeds a threshold concentration.
## Experimental procedures
All procedures involving mice were approved by the Animal Use Committee at TAMU (Animal Use Protocol 2018–0204). HFE(−/−) mice, henceforth called HFE mice (stock number 017784, B6.129S6-Hfe < tm2Nca > /J), were purchased, along with control mice (C57BL/6 J) from The Jackson Laboratory (www.jax.org). Animals were housed in the LAAR facility in the School of Veterinary Medicine at TAMU. Mice were raised in disposable all-plastic cages (Innovive model MVX1) containing synthetic bedding (Alpha-Dri Irradiated; Lab Supply, Houston) and all-plastic water bottles. Room temperature was 28 ± 1 °C and lighting was on a $\frac{12}{12}$ h cycle. Mice were bred on an iron-deficient mouse diet (TD.80396.PWD; www.envigo.com) spiked with 50 mg 57Fe (Cambridge Isotope Laboratories; $95.5\%$ enriched oxide powder) per kg chow. The diet was prepared by spraying 50 mL of each of 4 stock solutions onto the chow powder while mixing in a glass bowl. Each stock solution contained the required concentration of 57FeIII citrate (2 × excess of citrate relative to iron) plus sodium ascorbate (5 × relative to iron). The resulting moistened material was pelleted using a plastic pipe and a snug-fitting glass rod. Pellets were pushed out of the pipe onto a glass pan, then baked at 80 °C for 2–4 h. Once cooled, they were sealed in plastic bags and refrigerated until use. Mice were offered food and distilled water ad libitum.
Animals ranging in age from 3 to 52 wk were transported to the Chemistry department at TAMU where they were sacrificed. Mice were anaesthetized by injecting ketamine (5 mg/20 gm mouse) and xylazine (1 mg/20 gm mouse) subcutaneously. Exsanguination was by cardiac puncture once tests for pain (foot-pad squeeze) showed no response. Between 0.3 and 1.2 mL of blood was removed from each animal. Carcasses were imported into a refrigerated N2-atmosphere glove box (Mbraun Labmaster 130) containing 1–20 ppm O2 where they were perfused by passing Ringers’ buffer into the heart (0.1 ml/min for 5–10 min). Brain, liver, heart, spleen, and kidneys were then removed by dissection. After organ masses were determined, they were placed in Mössbauer cups. The liver from 1 mouse was used per cup, whereas 2 hearts, 1 brain, 2–3 spleens, and 3–4 kidneys were used in single MB cups. Samples were frozen in the box by importing an aluminum block that had been pre-chilled in LN2, placing the samples on the block surface, and waiting 1–2 min until they had frozen. Frozen samples were removed from the box and stored in LN2. MB Spectroscopy was performed as described [35–37] using WMOSS software for simulations.
Iron concentrations were determined as described [48], with minor modifications. Organs were thawed after collecting MB spectra. Samples were transferred into pre-weighed falcon tubes, and masses were determined. One mL of $70\%$ Trace Metal Grade HNO3 (Fisher Chemical) was added per 0.1 g of organ mass. Tubes were sealed using electrical tape and incubated at 80 °C overnight. After cooling, solutions were diluted 250 × using high-purity water in replicates of 3 to a final volume of 5 mL and $2\%$ (v/v) HNO3 concentration. A series of calibration standards were prepared with the TEXAM15 stock solution (Inorganic Venture, Christiansburg Virginia, USA), also affording $2\%$ HNO3. Fe concentrations of samples and standards were measured by ICP-MS (Agilent 7700x) in He collision mode.
## Results
Our objective was to better define the iron content of iron-overloaded and diseased organs using MB spectroscopy, and to evaluate the age-dependence of overloading. The raw spectrum of every sample collected exhibited a quadrupole doublet arising from deoxy FeII hemoglobin. Table S1 includes parameters used for all simulations. The blood doublet was simulated as the red line in Fig. 1A. Blood contributed to all spectra despite perfusing animals extensively with buffer prior to dissection and rinsing the dissected organs with buffer prior to loading them into MB cups. A second blood doublet was evident exclusively in spectra of spleens from young mice. It had the same isomer shift but slightly smaller ΔEQ (2.22 mm/s) than the primary blood doublet (2.32 mm/s). We suspect that it arose from fetal hemoglobin [49].Fig. 1Mössbauer spectrum of spleen from 52 wk WT mice before (A) and after (D) removing blood contribution. The red line is a simulation of the standard blood quadrupole doublet, B green line is a simulation of the CD, and C blue line is a simulation of ferritin. Spectra were collected at 5–6 K and with 0.05 T field applied parallel to the gamma radiation unless mentioned otherwise The percentage of raw spectral intensity due to the blood doublet varied from 5 to $90\%$. The absolute concentration of iron due to the blood was less than this range implies; an invariant concentration of blood iron affords a smaller percent of iron in spectra of overloaded organs relative to non-overloaded ones.
All other spectra presented below have had their hemoglobin contributions removed, and indicated percentages refer to hemoglobin-free difference spectra. With blood contributions removed, the dominant spectral feature for each organ was typically a magnetic sextet (Fig. 1C, blue line) arising either exclusively from ferritin or from ferritin and hemosiderin combined (see below). Also evident in most spectra was a central quadrupole doublet (called the CD) (Fig. 1B, green line) arising from a combination of [Fe4S4]2+ clusters and LS FeII hemes. Some samples exhibited a tiny-intensity quadrupole doublet due to non-heme-high-spin (NHHS) FeII species (not evident in Fig. 1).
## Spleen
The average mass of HFE spleens increased over the 52 wk duration of the study, from 55 → 180 mg (Table S2). The average mass of WT spleens at a similar age was slightly less. We removed the iron concentration due to blood from the overall HFE splenic iron concentrations, affording the values listed in Table S3. These concentrations were age-dependent, ranging from 850 µM in the spleen from a 6 wk mouse to 8300 µM in the spleen from a 52 wk mouse ($$n = 1$$ for each determination). These are likely to be the first absolute concentrations of splenic iron reported in which blood contributions were removed.
Most of the (non-hemoglobin) iron in spleens of young (3–4 wk) HFE mice was in the form of ferritin, with the CD representing just 15–$30\%$ of spectral intensity (Fig. 2, A and B). Starting at 4–10 wk of age, HFE spleens began accumulating even more ferritin (Fig. 2C–F), causing the CD to decline percentage-wise (all percentages are given in Table S4). By 32 and 52 wk, the concentration of iron in HFE spleens surpassed that in the liver while the CD declined to ~ $5\%$ of spectral intensity (Fig. 2G and H). Whether this percentage-wise decline in the CD reflects an absolute decline in [Fe4S4]2+ clusters or LS FeII hemes is uncertain, but our limited data suggest that it does not. Fig. 2Mössbauer spectra of spleens isolated from HFE mice of different ages (in wk). A 3, B 4, C 10, D 14, E 18, F 20, G 32, and H 52 We identified four 57Fe-enriched spleens from WT mice in an earlier study conducted a decade ago. Spleens and kidneys were not the focus of those studies, but some specimens had been preserved in liquid N2. Spectra of spleens from WT mice labeled “middle-aged” and “old” (Fig. 3A–D) were also dominated by ferritin. These labels approximately translate to 6–35 and 35–96 wk, respectively; see [35–37]. *In* general, less iron accumulated in the spleens of control mice vs HFE spleens of approximately the same age. The “old” control spleen contained an unusual quadrupole doublet in the center of the spectrum that was absent in spectra of younger WT spleens. The parameters associated with it (δ = 0.38 mm/s and ΔEQ = 0.9 – 1.0 mm/s) differed from those of the standard CD, and we did not assign it. Fig. 3Mössbauer spectra of control spleens (A–D) and kidneys (E–H) isolated from WT control mice of different approximate ages. A, B, and C, “middle-aged”, D “old”. E “pups”, F and G “middle-aged”, H “old”. Arrows in E, F, and G may indicate a trace of NHHS FeII To investigate the presence of hemosiderin, we collected 60 K spectra of HFE and control spleens of different ages (Fig. 4). Approximately 15- $20\%$ of the iron in 10, 20, and 52 wk HFE mouse spleens exhibited magnetic sextet features originating from hemosiderin (Fig. 4A–C). Hemosiderin was also detected in control spleens (Fig. 4E) but not in HFE spleens at 6 wk (Fig. 4D). Some spleen spectra exhibited tiny absorption consistent with an NHHS FeII doublet (Fig. 4A, B, and E), but they represented only $1\%$—$3\%$ of total spectral intensity. Charitou et al. [ 50] reported NHHS FeII in the spleens of thalassemic (th3/ +) mice but none in controls. Fig. 460 K Mössbauer spectra of organs. A 52 wk HFE spleen, B 20 wk HFE spleen, C 10 wk HFE spleen, D 6 wk HFE spleen, E “old” WT spleen, F “old” WT liver, G 52 wk HFE liver, H 32 wk HFE liver, I 20 wk HFE liver, J 52 wk HFE kidney, K 52 wk HFE heart, and L 52 wk HFE brain. Red lines are simulations of hemosiderin and blue lines are simulations of NHHS FeII, both assuming parameters in Table S1. Arrows in A, B, E, and G indicate trace intensity due to NHHS FeII
## Liver
Newborn livers from healthy control mice contain a high concentration of ferritin, but by 3 wk of age, most of the iron from ferritin exits the liver (Fig. 4C of [36]), leaving the CD dominating. At the same age, most iron in 3 wk HFE livers was ferritin, with ~ $40\%$ CD (Fig. 5A). With each additional week of age, the spectral percentage due to ferritin increased, while that due to the CD declined (Table S4)—consistent with iron accumulating in HFE livers as ferritin. In terms of absolute concentration of iron, we estimated the concentration of CD iron in 20 wk HFE liver to be a few 100 µM, whereas that for ferritin was over 1500 µM. We previously estimated CD concentrations of ~ 180 µM in healthy livers (Table S2 of [47]) but only 30 µM in 12 wk HFE livers [37]. The discrepancy in CD concentrations may be due to uncertainties caused by spectral features of ferritin overlapping the CD.Fig. 5Mössbauer spectra of livers isolated from HFE mice of different ages (in wk). A 3, B 4, C 5, D 6, E 8, F 10, G 14, H 18, I 32, and J 52. Arrows in B, C, F, and G may be due to NHHS FeII doublets The overall concentration of iron in HFE livers increased with age (Table S4). Based on spectral intensities, the loading of ferritin iron was roughly proportional to the animal’s age up until ~ 18 wk. Then, between 18 and 32 wk, there was a disproportionate increase in liver iron which was maintained to 52 wk, the end of the study. With age, the fraction of spectral intensity due to the CD declined, reaching only ~ $3\%$ in the spectrum of the 52 wk old liver. The percent absorption is approximately proportional to the concentration of iron in the samples (since all MB cups were filled with liver tissue). Consistent with this, older HFE livers exhibited bronzing, whereas control livers did not. The proportion of liver iron due to ferritin also increased in control livers, but at a slower rate. Impressively, the spectrum of the HFE liver at 3 wk was most like that of a 96 wk control liver; see Fig. 1C of [36]. In summary, the iron contents of HFE livers at any age were qualitatively like those of healthy livers; both were dominated by ferritin followed by the CD. However, the percentages of ferritin vs CD in HFE livers were more characteristic of healthy livers at an older age.
Difference spectra exhibited minor errors due to subtracting the blood quadrupole doublet using simulation lines generated assuming a Voigt lineshape (which is approximately but not precisely correct). Subtraction errors became more noticeable as the percentage spectral intensity of the blood doublet increased. The subtraction error was most obvious near the high-energy line of a nonheme high-spin FeII doublet. This made it challenging to establish the presence of such a species, especially for the spectra of Fig. 5A–E. A different challenge arose when minor spectral features became overwhelmed by the majority species; e.g., attempting to identify NHHS FeII features in the presence of intense ferritin features, as in Fig. 5I–J. Fortunately, there was a “window” associated with the spectra of Fig. 5F–H. Those spectra exhibited a high-energy line of a NHHS FeII doublet that was largely free from subtraction error and not overwhelmed by ferritin spectral features (Fig. 5, arrows).
We previously reported the absence of hemosiderin in a 12 wk liver from an HFE mouse, as judged from the absence of a sextet feature in the MB spectra at 70 K [37]. In the current study, we collected 60 K MB spectra of HFE livers of three different ages. About $20\%$ of the spectral intensity from of a 52 wk old HFE liver was due to hemosiderin (Fig. 4G), whereas spectra of a 32 wk liver and 20 wk HFE liver were devoid of hemosiderin (Fig. 4H and I), as was the liver from a 96 wk control mouse [36]. The high-energy line of the NHHS FeII doublet was clearly present in the 32 and 20 wk spectra. A similar feature was present in the spectrum of the 52 wk liver (Fig. 4G) but is less evident, because the collapsed ferritin features dominated. Livers examined from healthy control mice were devoid of hemosiderin [37]. Charitou et al. [ 50] previously reported the absence of hemosiderin in livers from th3/ + mice.
## Kidneys
Mössbauer spectra of young and middle-aged HFE kidneys (3–14 wk) were dominated by the CD and were virtually devoid of ferritin iron (Fig. 6A–C). At 18 wk (Fig. 6D), ferritin iron began developing intensity, and by 32 and 52 wk (Fig. 6E and F), it dominated the spectra. Unlike spectra of other HFE organs, the CD intensity remained intense even for older animals. The spectrum of control kidneys labeled “pups” was also dominated by the CD and was devoid of ferritin (Fig. 3E). The spectra of “middle aged” control kidneys (Fig. 3F and G) were indistinguishable from their HFE counterparts; all contained significant spectral percentages of both CD and ferritin. The spectrum of “old” WT control kidney (Fig. 3H) was similar to HFE kidneys from 32 wk mice, whereas kidneys from 52 wk HFE mice contained a higher percentage of ferritin. Unlike other organs, the iron contents of kidneys from HFE mice at all ages were indistinguishable (at this level of analysis and between 3 to 52 wk of age) from WT controls. This suggests that the ability to store iron as ferritin is more limited in kidneys than in livers or spleens. The non-hemoglobin iron concentrations of HFE kidneys increased slightly with age, from 160 µM (for a 4 wk sample) to 560 µM (at 32 and 52 wk) (Table S4). A spectrum of the 52 wk HFE kidney collected at 60 K (Fig. 4J) was devoid of a sextet that would have indicated hemosiderin; nor did it exhibit an NHHS FeII doublet. In contrast, Charitou et al. [ 50] reported $500\%$ more NHHS FeII in the kidneys of thalassemic (th3/ +) mice compared to controls. Fig. 6Mössbauer spectra of kidneys isolated from HFE mice of different ages (in wk). A 3, B 10, C 14, D 18, E 32, and F 52 wk. Arrows indicate NHHS FeII doublet
## Heart and brain
HFE hearts did not accumulate nearly as much ferritin as did livers and spleens (Fig. 7A–C). On the other hand, they accumulated more iron than WT control hearts in the same age group. The spectrum of a 3 wk HFE heart (Fig. 7A) exhibited more ferritin-based features than those of control WT hearts between the age of newborn and 4 wk (See Fig. 2 of [37]). That of an HFE heart at 10 wk (Fig. 7B) also exhibited a greater percentage of ferritin than adult controls (see Fig. 3 of [37]). The spectrum of a 52 wk HFE heart (Fig. 7C) was similar to that of a 60 wk control heart (Fig. 1D of [37]). MB spectra of hearts from young healthy controls (up to 4 wk of age) were dominated by the CD and contained only ~ $18\%$ ferritin on average (Fig. 2 of [37]). Corresponding spectra from adult control mice (6–28 wk) were also dominated by the CD but contained ~ $46\%$ ferritin. Older control mice contained ~ $70\%$ ferritin with the remainder in the form of the CD. In contrast, hearts from 3, 18, and 52 wk old HFE mice contained 20, 65, and $80\%$ ferritin, respectively (Table S4). Thus, using the fraction of ferritin loading in the heart, hearts of HFE mice appeared “older” than their chronological age relative to controls. The 60 K MB spectrum of 52 wk hearts (Fig. 4K) was devoid of any magnetic feature, indicating the absence of hemosiderin. Fig. 7Mössbauer spectra of hearts (A–C) and brains (D–G) isolated from HFE mice of different ages (in wk). A 3, B 10, C 52, D 3, E 4, F 10, and G 52. Arrows may indicate NHHS FeII doublets The brain contains little iron, and so, MB spectral intensities were weak (Fig. 7D–G). Within that constraint, there were no major differences evident between the HFE and control brains in similarly aged mice (Fig. 1 of [35]). There was slightly more ferritin in HFE brains at the same age. For control brains, there was no trend from 3, 4, 24, and 58 wk, with an average of $53\%$ ferritin and $34\%$ CD [35]. The 60 K MB spectrum of 52 wk HFE brains (Fig. 4L) was devoid of any magnetic feature, indicating the absence of hemosiderin. There was perhaps a hint of NHHS FeII. Charitou et al. [ 50] reported that brains from 9 month thalassemic mice contained around $10\%$ more ferritin than WT mice, comparable to what we observed in HFE brains.
## Age-dependence of iron-overload in HFE organs
The iron content of all investigated HFE organs, as determined by low-temperature MB spectroscopy, was similar to the same healthy organ but from an older mouse. HFE organs accumulated more ferritin than controls, and they did so starting at an earlier age. The liver and spleen were the first to be overloaded, followed by the kidney, and then, to a lesser extent, heart and brain. Although these latter two organs were less overloaded, even minor overloading might still affect their physiological function.
Previous studies have not observed iron-overloaded spleens in HFE mice. Zhou et al. [ 51] reported that spleens of 10 wk HFE mice were not iron-overloaded. Albalat et al. [ 52] reported a similar result with 12 month old HFE mice. Cavey et al. [ 53] saw splenic iron-overload but only in WT mice in which iron-overload was induced dietarily; no overload was observed in HFE mice. We cannot explain this apparent discrepancy with previous studies. However, there is no doubt that we have observed iron-overloaded spleens. Although we only have one Mössbauer spectrum for each age reported in Fig. 2, each sample contained 2–3 spleens, and so, the spectra represent an average.
*Our* general understanding of iron-overload in spleen and other organs is illustrated in Fig. 8. The rate of iron import (Rin), which is mainly associated with red-cell recycling in spleen, is counterbalanced by the rates of iron export Rout (to bone marrow for erythropoiesis) and dilution due to cell/organ growth (Rdil). For iron to accumulate in the spleen, Rin might increase relative to its rate under WT conditions, Rout might decrease, or Rdil might decrease (or some combination of these). Cavey et al. [ 54] found that the concentration of erythrocytes in HFE mice is ~ $6\%$ higher than in controls, suggesting an increased Rin. Organ growth declines as animals’ age, and so, Rdil might have also decreased with age. In principle, Rout should increase in HFE mice, relative to the WT state (due to less hepcidin and more FPN), causing iron depletion. Since iron-overload was observed, this effect may not dominate under the conditions examined. The excess iron in the spleen first accumulates as ferritin, indicating an increased rate of ferritin metallation (Rferritin). The iron used for this process is likely obtained from the degradation of heme as catalyzed by heme oxygenase [55]. The mechanism explaining the accumulation of ferritin iron in other organs (Fig. 8, bottom) is assumed to be similar except that these organs import either transferrin or NTBI rather than senescent erythrocytes, and they may not import or export iron as quickly as in the spleen. Fig. 8Model of iron-overloading in the spleen (top) and other organs (bottom). Cell growth is indicated by the dashed arrows
## Iron-dependent threshold activation of ferritin degradation
This was our first study in which we detected hemosiderin in any sample; here, we detected it in spleens and livers of older HFE mice, as well as in the spleen of an older control mouse. In all cases, the concentration of ferritin in the organ was high (2–3 mM Fe or higher). The percent of iron in the form of hemosiderin never exceeded ~ $20\%$—i.e., ferritin still dominated. Consistent with this, Charitou et al. [ 50] also observed hemosiderin in iron-overloaded spleens, in their case from th3/ + (heterozygous β-thalessemia) mice. They also reported that ferritin rather than hemosiderin dominated the iron content of iron-overloaded spleens; hemosiderin represented ~ 19 and $23\%$ of total splenic iron at 6 and 9 months of age, respectively (our estimates), similar to our observations.
Earlier studies have reported hemosiderin in diseased iron-overloaded organs—but they also reported that it exceeded ferritin (see Introduction). One possible explanation for this discrepancy is that earlier samples might have been more overloaded and thus had a higher Fe concentration. Such samples were routinely not enriched in 57Fe, so perhaps samples with exceedingly high iron concentrations were selected to achieve the highest possible signal/noise ratios in Mössbauer spectra. Also, the mice we used (and those used by Charitou et al. [ 50]) were fed only 50 mg 57Fe per kg chow—which is ca. $\frac{1}{5}$th of the iron in normal chow. This could have moderated the extent of hemosiderin generation. Freeze-thawing had no effect on the ferritin/hemosiderin ratio [56]. Another concern with those earlier studies is that typically only high-temperature spectra were collected which complicates interpretations.
We propose that the degradation of ferritin → hemosiderin becomes activated only after a threshold concentration of iron (or perhaps only of ferritin) has been is exceeded. We estimate the threshold to be ca. 2–3 mM total (non-hemoglobin) iron. In our samples, that threshold was exceeded in both spleen and liver, and especially (but not exclusively) in older HFE samples. The concentration of ferritin in spleen was actually greater than in liver (Table S3). This difference is even more extreme than reflected in % spectral absorption, because the liver samples completely filled the MB cup, whereas spleen samples occupied only ~ half of the cup volume.
## “Surface-bound ferritin” spectral feature originates from [Fe4S4]2+ clusters and LS FeII hemes
Charitou et al. [ 50] enriched th3/ + and WT mice with 57Fe, and collected hearts, livers, kidneys, brains, and spleens at 1, 3, 6, and 9 months (with 1–2 mice of each strain per time-point). MB spectra at 80 K exhibited a blood doublet, a nonheme high-spin FeII doublet, and two additional doublets. Following Bou-Abdalla et al. [ 57], one of the additional doublets (δ = 0.46 mm/s and ΔEQ = 0.59 mm/s) was assigned to the inner core of ferritin, and the other (δ = 0.46 mm/s and ΔEQ = 1.05 mm/s) was assigned to the surface irons of the ferritin core. The parameters of the surface-ferritin doublet are remarkably similar to those of the CD in our spectra. We hypothesize that the two doublets arise from the same species, namely diamagnetic [Fe4S4]2+ clusters and low-spin ferrous heme centers. At 80 K, MB spectra cannot distinguish superparamagnetic ferritin from these diamagnetic centers as both would yield quadrupole doublets. However, the distinction is clear in our spectra collected at ~ 5 K, because ferritin and hemosiderin exhibit magnetic sextets at that temperature, whereas diamagnetic $S = 0$ [Fe4S4]2+ clusters, [Fe2S2]2+ clusters, and $S = 0$ FeII hemes exhibit quadrupole doublets devoid of magnetic hyperfine interactions.
This observed behavior unambiguously supports our assignment of the disputed doublet. Reinforcing our assignment, many other results of Charitou et al. would agree fully with ours if the “surface-ferritin” doublet were reassigned to the CD and “inner-ferritin” were assigned to both surface and core ferritin iron combined. For example, Charitou et al. [ 50] found that the ratio of the inner/surface-ferritin doublets increased with age and differed between organs, in agreement with our observations. In WT hearts, kidneys, and brain, “surface ferritin” (i.e., the CD) dominated, whereas in the spleen and liver, “inner ferritin” (i.e., ferritin) dominated, as we observed. With age, the proportion of iron due to ferritin (or inner ferritin) increased, also as we observed. We both observed modest spectral differences between Th3/ + and WT hearts. We both observed major iron accumulation in the liver (4–9.5 × more ferritin in Th3/ + adults than in controls). For kidneys, Charitou et al. observed an increase in ferritin at 12 wk, similar to the increase we observed at 15 wk. We both observed slight ferritin accumulation in the brain.
## MB spectroscopy is complementary to NMR
NMR has the huge advantage of being non-invasive and performable at room-temperature and on live patients [58]. The major advantage of MB spectroscopy is that it detects all forms of iron, including diamagnetic iron. Moreover, relative MB spectral intensities are approximately proportional to the concentrations of each species in the sample. NMR primarily detects ferritin and cannot detect diamagnetic iron centers, including oxidized ISCs and low-spin FeII hemes; it is sensitive only to the overall effect of iron-associated paramagnetism on proton relaxation rates. Thus, NMR seems unable to easily distinguish ferritin from hemosiderin. Magnetic susceptibility suffers from similar problems.
All problems considered, low-temperature MB studies of iron-overloaded 57Fe-enriched mouse organs provide the most rigorous description for decomposing iron (overloaded or not) in mammals. Such studies could be performed on any genetic strain of mice, and in an age-dependent manner, to evaluate how iron contents are changing. Much remains to be learned as to the mechanism of hemosiderin formation. Performing NMR on matched 57Fe-enriched mouse organs could allow MB to calibrate and interpret the iron content of NMR images of comparable organs from human patients. This combination of spectroscopic methods might offer a distinct advantage in understanding the process of iron-overloading and in treating iron-overload diseases.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file 1 (DOCX 50 KB)
## References
1. Anderson GJ, Bardou-Jacquet E. **Revisiting hemochromatosis: genetic vs. phenotypic manifestations**. *Ann Trans Med.* (2021) **9** 731. DOI: 10.21037/atm-20-5512
2. Kowdley KV, Gochanour EM, Sundaram V, Shad RA, Handa P. **Hepcidin signaling in health and disease: ironing out the details**. *Hepatol Comm* (2021) **5** 723-735. DOI: 10.1002/hep4.1717
3. Knutson MD. **Non-transferrin-bound iron transporters**. *Free Rad Biol Med* (2019) **133** 101-111. DOI: 10.1016/j.freeradbiomed.2018.10.413
4. Brissot P, Pietrangelo A, Adams PC, de Graaff B, McLaren CE, Loreal O. **Haemochromatosis**. *Nat Rev Dis Primers* (2018). DOI: 10.1038/nrdp.2018.16
5. Chua-anusorn W, Webb J, Macey DJ, Hall PDM, St Pierre TG. **The effect of prolonged iron loading on the chemical form of iron oxide deposits in rat liver and spleen**. *Biochim Biophys Acta Mol Basis Dis* (1999) **1454** 191-200. DOI: 10.1016/S0925-4439(99)00036-8
6. Iancu TC, Shiloh H, Bauminger ER, Pinson A, Hershko C. **Ultrastructural pathology of iron-loaded rat myocardial cells in culture**. *Br J Exp Pathol* (1987) **68** 53-65. PMID: 3814501
7. Subramaniam VN, McDonald CJ, Ostini L, Lusby PE, Wockner LF, Ramm GA, Wallace DF. **Hepatic iron deposition does not predict extrahepatic iron loading in mouse models of hereditary hemochromatosis**. *Am J Pathol* (2012) **181** 1173-1179. DOI: 10.1016/j.ajpath.2012.06.025
8. Noetzli LJ, Papudesi J, Coates TD, Wood JC. **Pancreatic iron loading predicts cardiac iron loading in thalassemia major**. *Blood* (2009) **114** 4021-4026. DOI: 10.1182/blood-2009-06-225615
9. Sukhbaatar N, Weichhart T. **Iron regulation: macrophages in control**. *Pharmaceuticals* (2018) **11** 137. DOI: 10.3390/ph11040137
10. Ganz T. **Hepcidin and iron regulation, 10 years later**. *Blood* (2011) **117** 4425-4433. DOI: 10.1182/blood-2011-01-258467
11. Delaby C, Pilard N, Hetet G, Driss F, Grandchamp B, Beaumont C, Canonne-Hergaux F. **A physiological model to study iron recycling in macrophages**. *Exp Cell Res* (2005) **310** 43-53. DOI: 10.1016/j.yexcr.2005.07.002
12. Ferreira C, Santambrogio P, Martin ME, Andrieu V, Feldmann G, Hénin D, Beaumont C. **H ferritin knockout mice: a model of hyperferritinemia in the absence of iron overload**. *Blood* (2001) **98** 525-532. DOI: 10.1182/blood.V98.3.525
13. Raaij SV, Swelm RV, Bouman K, Cliteur M, van den Heuvel MC, Pertijs J, Patel D, Bass P, van Goor H, Unwin R, Sria SK, Swinkels D. **Tubular iron deposition and iron handling proteins in human healthy kidney and chronic kidney disease**. *Sci Rep* (2018) **8** 9353. DOI: 10.1038/s41598-018-27107-8
14. Zhang D, Meyron-Holtz E, Rouault TA. **Renal iron metabolism: transferrin iron delivery and the role of iron regulatory proteins**. *J Am Soc Nephrol* (2007) **18** 401-406. DOI: 10.1681/ASN.2006080908
15. Chaudhary K, Chilakala A, Ananth S, Mandala A, Veeranan-Karmegam R, Powell FL, Ganapathy V, Gnana-Prakasam JP. **Renal iron accelerates the progression of diabetic nephropathy in the HFE gene knockout mouse model of iron overload**. *Am. J. Phys. Renal Phys.* (2019) **317** 512-517
16. Piga A, Longo F, Duca L, Roggero S, Vinciguerra T, Calabrese R, Hershko C, Cappellini MD. **High nontransferrin bound iron levels and heart disease in thalassemia major**. *Am J Hematol* (2009) **84** 29-33. DOI: 10.1002/ajh.21317
17. Link G, Athias P, Grynberg A, Pinson A, Hershko C. **Effect of iron loading on transmembrane potential, contraction, and automaticity of rat ventribular muscle cells in culture**. *J Lab Clin Med* (1989) **113** 103-111. PMID: 2909644
18. Nam HY, Wang CY, Zhang L, Zhang W, Hojyo S, Fukada T, Knutson MD. **ZIP14 and DMT1 in the liver, pancreas, and heart are differentially regulated by iron deficiency and overload: implications for tissue iron uptake in iron-related disorders**. *Haematologica* (2013) **98** 1049-1057. DOI: 10.3324/haematol.2012.072314
19. Jenkitkasemwong S, Wang CY, Coffey R, Wei Z, Chan A, Biel T, Kim JS, Hojyo S, Fukada T, Knutson MD. **SLC39A14 is required for the development of hepatocellular iron overload in murine models of hereditary hemochromatosis**. *Cell Metab* (2015) **22** 138-150. DOI: 10.1016/j.cmet.2015.05.002
20. Liu Y, Parkes JG, Templeton DM. **Differential accumulation of non-transferrin-bound iron by cardiac myocytes and fibroblasts**. *J Mol Cell Cardiol* (2003) **35** 505-514. DOI: 10.1016/S0022-2828(03)00072-5
21. Ke Y, Chen YY, Chang YZ, Duan XL, Ho KP, Jiang DH, Wang K, Qian ZM. **Post-transcriptional expression of DMT1 I the heart of rat**. *J Cell Physiol* (2003) **196** 124-130. DOI: 10.1002/jcp.10284
22. Kumfu S, Chattipakorn S, Chinda K, Fucharoen S, Chattipakorn N. **T-type calcium channel blockade improves survival and cardiovascular function in thalassemic mice**. *Eur J Haematol* (2012) **88** 535-548. DOI: 10.1111/j.1600-0609.2012.01779.x
23. Heidari M, Johnstone DM, Bassett B, Graham RM, Chua ACG, House MJ, Collingwood JF, Bettencourt C, Houlden H, Ryten M, Olynyk JK, Trinder D, Milward EA. **Brain iron accumulation affects myelin-related molecular systems implicated in a rare neurogenetic disease family with neuropsychiatric features**. *Mol Psych* (2016) **21** 1599-1607. DOI: 10.1038/mp.2015.192
24. Metafratzi Z, Argyropoulou MI, Kiortsis DN, Tsampoulas C, Chaliassos N, Efremidis SC. **T-2 relaxation rate of basal ganglia and cortex in patients with beta-thalassaemia major**. *Br J Radiol* (2001) **74** 407-410. DOI: 10.1259/bjr.74.881.740407
25. Qiu D, Chan GCF, Chu J, Chan Q, Ha SY, Moseley ME, Khong PL. **MR quantitative susceptibility imaging for the evaluation of iron loading in the brains of patients with beta-thalassemia major**. *Am J Neuroradiol* (2014) **35** 1085-1090. DOI: 10.3174/ajnr.A3849
26. Beard JL, Wiesinger JA, Li N, Connor JR. **Brain iron uptake in hypotransferrinemic mice: Influence of systemic iron status**. *J Neurosci Res* (2005) **79** 254-261. DOI: 10.1002/jnr.20324
27. Dickinson TK, Devenyi AG, Connor JR. **Distribution of injected iron 59 and manganese 54 in hypotransferrinemic mice**. *J Lab Clin Med* (1996) **128** 270-278. DOI: 10.1016/S0022-2143(96)90028-1
28. Ueda F, Raja KB, Simpson RJ, Trowbridge IS, Bradbury MWB. **Rate of Fe-59 uptake into brain and cerebrospinal fluid and the influence thereon of antibodies against the transferrin receptor**. *J Neurochem* (1993) **60** 106-113. DOI: 10.1111/j.1471-4159.1993.tb05828.x
29. Noetzli LJ, Panigrahy A, Mittelman SD, Hyderi A, Dongelyan A, Coates TD, Wood JC. **Pituitary iron and volume predict hypogonadism in transfusional iron overload**. *Am J Hematology* (2012) **87** 167-171. DOI: 10.1002/ajh.22247
30. Takeda A, Takatsuka K, Sotogaku N, Oku N. **Influence of iron-saturation of plasma transferrin in iron distribution in the brain**. *Neurochem Int* (2002) **41** 223-228. DOI: 10.1016/S0197-0186(02)00023-2
31. Aslan E, Luo JW, Lesage A, Paquin P, Cerny M, Chin ASL, Olivié D, Gilbert G, Soulières D, Tang A. **MRI-based R2* mapping in patients with suspected or known iron overload**. *Abdominal Radiology* (2021) **46** 2505-2515. DOI: 10.1007/s00261-020-02912-w
32. Wood JC. **Magnetic resonance imaging measurement of iron overload**. *Curr Opin Hematol* (2007) **14** 183-190. DOI: 10.1097/MOH.0b013e3280d2b76b
33. Hocq A, Luhmer M, Saussez S, Louryan S, Gillis P, Gossuin Y. **Effect of magnetic field and iron content on NMR proton relaxation of liver, spleen and brain tissues**. *Contrast Med Mol Imag* (2015) **10** 144-152. DOI: 10.1002/cmmi.1610
34. Gutierrez L, Quintana C, Patino C, Bueno J, Coppin H, Roth MP, Lazaro FJ. **Iron speciation study in Hfe knockout mice tissues: Magnetic and ultrastructural characterisation Biochim**. *Biophys Acta- Mol Basis Dis* (2009) **1792** 541-547. DOI: 10.1016/j.bbadis.2009.03.007
35. Holmes-Hampton GP, Chakrabarti M, Cockrell AL, McCormick SP, Abbott LC, Lindahl LS, Lindahl PA. **Changing iron content of the mouse brain during development**. *Metallomics* (2012) **4** 761-770. DOI: 10.1039/c2mt20086d
36. Chakrabarti M, Cockrell AL, Park JK, McCormick SP, Lindahl LS, Lindahl PA. **Speciation of iron in mouse liver during development, iron deficiency, IRP2 deletion and inflammatory hepatitis**. *Metallomics* (2014) **7** 93-101. DOI: 10.1039/C4MT00215F
37. Wofford JD, Chakrabarti M, Lindahl PA. **Mössbauer spectra of mouse hearts reveal age-dependent changes in mitochondrial and ferritin iron levels**. *J Biol Chem* (2017) **292** 5546-5554. DOI: 10.1074/jbc.M117.777201
38. Bell SH, Weir MP, Dickson DP, Gibson JF, Sharp GA, Peters TJ. **Mössbauer spectroscopic studies of human haemosiderin and ferritin**. *Biochim Biophys Acta* (1984) **787** 227-236. DOI: 10.1016/0167-4838(84)90313-3
39. Chua-anusorn W, Trans KC, Webb J, Macey DJ. **Chemical speciation of iron deposits in thalassemic heart tissue**. *Inorg Chim Acta* (2000) **300–302** 932-936. DOI: 10.1016/S0020-1693(00)00006-2
40. St. Pierre TG, Tran KC, Webb J, Macey DJ, Pootrakul P, Dickson DPE,. **Core structures of haemosiderins deposited in various organs in beta-thalassemia hemoglobin-E disease**. *Hyperfine Interact* (1992) **71** 1279-1282. DOI: 10.1007/BF02397317
41. Selden C, Owen M, Hopkins JMP, Peters TJ. **Studies on the concentration and intracellular localization of iron proteins in liver biopsy specimens from patients with iron overload with special reference to their role in lysosomal disruption**. *Br J Haematol* (1980) **44** 593-603. DOI: 10.1111/j.1365-2141.1980.tb08714.x
42. St Pierre TG, Dickson DPE, Kirkwood JK, Ward RJ, Peters TJ. **A Mössbauer spectroscopic study of the form of iron in iron overload**. *Biochim Biophys Acta* (1987) **924** 447-451. DOI: 10.1016/0304-4165(87)90159-0
43. Ward RJ, Ramsey M, Dickson DPE, Hunt C, Douglas T, Mann S, Aouad F, Peters TJ, Crichton RR. **Further characterization of forms of hemosiderin in iron overloaded tissues**. *Eur J Biochem* (1994) **225** 187-194. DOI: 10.1111/j.1432-1033.1994.00187.x
44. Webb J, St Pierre TG, Tran KC, Chuaanusorn W, Macey DJ, Pootrakul P. **Biologically significant iron(III) oxyhydroxy polymers: mössbauer spectroscopic study of ferritin and hemosiderin in pancreas tissue of beta-thalassemia hemoglobin E disease**. *Inorg Chim Acta* (1996) **243** 121-125. DOI: 10.1016/0020-1693(96)04898-0
45. St Pierre TG, Chua-anusorn W, Webb J, Macey D, Pootrakul P. **The form of iron oxide deposits in thalassemic tissues varies between different groups of patients: a comparison between Thai beta-thalassemia/hemoglobin E patients and Australian beta-thalassemia patients**. *Biochimica Biophy Acta- Mol. Basis Dis* (1998) **1407** 51-60. DOI: 10.1016/S0925-4439(98)00026-X
46. St Pierre TG, Chua-anusorn W, Webb J, Macey DJ. **Iron overload diseases: the chemical speciation of non-heme iron deposits in iron loaded mammalian tissues**. *Hyperfine Interact* (2000) **126** 75-81. DOI: 10.1023/A:1012636510472
47. Gutierrez L, Spasic MV, Muckenthaler MU, Lazaro FJ. **Quantitative magnetic analysis reveals ferritin-like iron as the most predominant iron-containing species in the murine Hfe-haemochromatosis**. *Biochimica Biophysica Acta Mol Basis Disease* (2012) **1822** 1147-1153. DOI: 10.1016/j.bbadis.2012.03.008
48. Dziuba N, Hardy J, Lindahl PA. **Low-molecular-mass iron in healthy blood plasma is not predominantly ferric citrate**. *Metallomics* (2018) **10** 802-817. DOI: 10.1039/c8mt00055g
49. Oshtrakh MI. **The features of Mössbauer spectra of hemoglobin in relation to the quadrupole splitting and heme iron stereochemistry**. *Z Naturforsch A* (1998) **53** 608-614. DOI: 10.1515/zna-1998-6-755
50. Charitou G, Tsertos C, Parpottas Y, Kleanthous M, Lederer CW, Phylactides M. **Study of iron complexes in visceral organs and brain from a**. *J. Mol. Structure.* (2020) **1215** 128251. DOI: 10.1016/j.molstruc.2020.128251
51. Zhou XY, Tomatsu S, Fleming RE, Parkkila S, Waheed A, Jiang JX, Fei Y, Brunt EM, Ruddy DA, Prass CE, Schatzman RC, O’Neill R, Britton RS, Bacon BR, Sly WS. **HFE gene knockout produces mouse model of hereditary hemochromatosis**. *Proc Natl Acad Sci USA* (1998) **95** 2492-2497. DOI: 10.1073/pnas.95.5.2492
52. Albalat E, Cavey T, Leroyer P, Ropert M, Balter V, Loreal O. **Hfe gene knock-out in a mouse model of hereditary hemochromatosis affects bodily iron isotope compositions**. *Front Med* (2021) **8** 711822. DOI: 10.3389/fmed.2021.711822
53. Cavey T, Latour C, Island ML, Leroyer P, Guggenbuhl P, Coppin H, Roth MP, Bendavid C, Brissot P, Ropert M, Loreal O. **Spleen iron, molybdenum, and manganese concentrations are coregulated in hepcidin-deficient and secondary iron overload models in mice**. *FASEB J* (2019) **22** 11072-11081. DOI: 10.1096/fj.201801381RR
54. Cavey T, Ropert M, de Tayrac M, Bardou-Jacquet E, Island ML, Leroyer P, Bendavid C, Brissot P, Loreal O. **Mouse genetic background impacts both on iron and non-iron metals parameters and on their relationships**. *Biometals* (2015) **28** 733-743. DOI: 10.1007/s10534-015-9862-8
55. Poss KD, Tonegawa S. **Heme oxygenase 1 is required for mammalian iron reutilization**. *Proc Natl Acad Sci USA* (1997) **94** 10919-10924. DOI: 10.1073/pnas.94.20.10919
56. Meyrick D, Webb J, Cole C. **Iron and iron proteins found in the genetic disease, hereditary spherocytosis**. *Inorg Chim Acta* (2002) **339** 481-487. DOI: 10.1016/S0020-1693(02)01049-6
57. Bou-Abdallah F, Carney E, Chasteen ND, Arosio P, Viescas AJ, Papaefthymiou GC. **A comparative Mössbauer study of the mineral cores of human H-chain ferritin employing dioxygen and hydrogen peroxide as iron oxidants**. *Biophys Chem* (2007) **130** 114-121. DOI: 10.1016/j.bpc.2007.08.003
58. Papakonstantinou O, Alexopoulou E, Economopoulos N, Benekos O, Kattamis A, Kostaridou S, Ladis V, Efstathopoulos E, Gouliamos A, Kelekis NL. **Assessment of iron distribution between liver, spleen, pancreas, bone marrow, and myocardium by means of R2 relaxometry with MRI in patients with beta-thalassemia major**. *J Magn Reson Imaging* (2009) **29** 853-859. DOI: 10.1002/jmri.21707
|
---
title: BET inhibitors synergize with sunitinib in melanoma through GDF15 suppression
authors:
- Furong Zeng
- Yayun Li
- Yu Meng
- Huiyan Sun
- Yi He
- Mingzhu Yin
- Xiang Chen
- Guangtong Deng
journal: Experimental & Molecular Medicine
year: 2023
pmcid: PMC9981764
doi: 10.1038/s12276-023-00936-y
license: CC BY 4.0
---
# BET inhibitors synergize with sunitinib in melanoma through GDF15 suppression
## Abstract
Targeting bromodomain and extra-terminal domain (BET) proteins has shown a promising therapeutic effect on melanoma. The development of strategies to better kill melanoma cells with BET inhibitor treatment may provide new clinical applications. Here, we used a drug synergy screening approach to combine JQ1 with 240 antitumor drugs from the Food and Drug Administration (FDA)-approved drug library and found that sunitinib synergizes with BET inhibitors in melanoma cells. We further demonstrated that BET inhibitors synergize with sunitinib in melanoma by inducing apoptosis and cell cycle arrest. Mechanistically, BET inhibitors sensitize melanoma cells to sunitinib by inhibiting GDF15 expression. Strikingly, GDF15 is transcriptionally regulated directly by BRD4 or indirectly by the BRD4/IL6/STAT3 axis. Xenograft assays revealed that the combination of BET inhibitors with sunitinib causes melanoma suppression in vivo. Altogether, these findings suggest that BET inhibitor-mediated GDF15 inhibition plays a critical role in enhancing sunitinib sensitivity in melanoma, indicating that BET inhibitors synergize with sunitinib in melanoma.
## Cancer: a drug combination for combating melanoma
Drugs that target proteins involved in issuing stop and start commands to cancer-related genes help to sensitize melanoma cells to a widely used anti-cancer therapy, leading to tumor shrinkage in mice. Furong Zeng from Central South University in Changsha, China, and colleagues screened 240 approved anti-cancer agents in search of molecules that work synergistically with a BET inhibitor, an inhibitor of these proteins, to induce the death of melanoma cells. They found that a protein targeted by BET inhibitors (JQ1 and NHWD-870) regulates a growth factor protein, whose altered expression then helps to sensitize melanoma cells to sunitinib. Sunitinib is approved to treat several types of cancer and works by targeting many different kinds of kinase enzymes. The combination of BET inhibitors and sunitinib showed synergistic antitumor effects in a mouse model of melanoma.
## Introduction
The introduction of targeted therapies and immunotherapy has improved survival outcomes in the majority of melanoma patients1. However, there are still a large number of patients whose outcomes are dismal because of dose-limiting toxicity and innate or acquired resistance2. Multiple causes of resistance to targeted therapies have been clarified, including reactivation of the MAPK pathway and activation of alternate receptor kinase pathways3,4. Similarly, resistance to immunotherapy can occur through the Janus kinase (JAK$\frac{1}{2}$) cytokine signaling pathway, with reduced PD-1 ligand (PD-L1) expression or decreased beta2-microglobulin expression causing the loss of major histocompatibility complex class 1 expression5–7. Therefore, it is still of great significance to explore new individualized targeted therapies for melanoma.
Targeting bromodomain and extra-terminal domain (BET) proteins has become an attractive antitumor strategy due to their abnormal expression and promotion of melanoma carcinogenesis8. Our team previously developed a novel oral BET inhibitor, NHWD-870, which could suppress cancer cell-macrophage interactions through the BRD4/HIF1a/CSF1 axis9. We further demonstrated that BET inhibitors could inhibit melanoma progression through the noncanonical NF-κB/SPP1 pathway10. In addition, many excellent studies have shown that BET inhibitors can modulate sensitivity to other drugs. For example, Yang et al. demonstrated that BET inhibitors sensitized homologous recombination-proficient cancers to poly(adenosine diphosphate–ribose) polymerase (PARP) inhibitors11. Kanojia et al. showed that BET inhibitors induced βIII-tubulin expression and sensitized metastatic breast cancer in the brain to vinorelbine12. Thus, the development of strategies to better kill melanoma cells with BET inhibitor treatment may provide new clinical applications.
Combination drug therapy is an effective strategy to improve drug efficacy in melanoma. Researchers found that BET inhibitors synergized with RAF/BRAF/MEK inhibitors in melanoma13–18. Moreover, cotargeting BET and CDK9 had synergistic effects against melanoma cells in vitro and in vivo19. However, these combination strategies are far from clinical use, highlighting the importance of screening the Food and Drug Administration (FDA)-approved drug library to identify drugs that synergize with BET inhibitors in melanoma cells.
Sunitinib, a novel oral multitargeted tyrosine kinase inhibitor approved by the FDA in 2006, is used to treat patients with clear cell renal cell carcinoma or gastrointestinal stromal tumors20. Our team previously demonstrated that sunitinib had a therapeutic effect on melanoma that was further enhanced by propranolol21. However, whether BET inhibitors are involved in the regulation of sunitinib sensitivity is completely unknown. Here, we identified sunitinib among 240 FDA-approved antitumor drugs to synergize with BET inhibitors in melanoma cells. Mechanistically, BET inhibitors sensitize melanoma cells to sunitinib by inhibiting GDF15 expression. Strikingly, GDF15 is transcriptionally regulated directly by BRD4 or indirectly by the BRD4/IL6/STAT3 axis. Xenograft assays revealed a reduction in tumor volume with combined BET inhibitor and sunitinib treatment. Altogether, these findings suggest that BET inhibitor-mediated GDF15 inhibition plays a critical role in enhancing sunitinib sensitivity, indicating that BET inhibitors synergize with sunitinib in melanoma.
## Cell culture
A375, SK-MEL-28, and HEK293T cells were obtained from the American Type Culture Collection (ATCC, VA, USA). All cell lines were cultured in Dulbecco’s modified Eagle’s medium (Biological Industries) supplemented with $10\%$ fetal bovine serum (Biological Industries) and $1\%$ penicillin–streptomycin solution (Beyotime Biotechnology) in a humidified 37 °C incubator with $5\%$ CO2.
## Reagents and antibodies
FDA-approved antitumor drugs were purchased from Selleck Chemicals (Houston, TX, USA). Antibodies specific for the following proteins were used: STAT3 (#9139, Cell Signaling Technology, MA, USA), phospho-Stat3 (Tyr705) (#9145, Cell Signaling Technology, MA, USA), phospho-Stat3 (Ser727) (#49081, Cell Signaling Technology, MA, USA), BRD2 (#ab139690, Abcam, Wales, UK), BRD3 (#ab50818, Abcam, Wales, UK), BRD4 (#ab128874, Abcam, Wales, UK), GDF15 (27455-1-AP, Proteintech, Wuhan, Hubei, China), Ki67 (#ab15580, Abcam, Wales, UK) and ACTIN (#sc-8432, Santa Cruz Biotechnology, TX, USA).
## Cell cycle and cell apoptosis
Cell cycle analysis was performed by flow cytometry using a cell cycle kit (Beyotime, C1052) as previously described2. The cell cycle distribution was assessed by FlowJo. Apoptosis was assessed with an Annexin V-AF647/PI kit (4 A Biotech, FXP023-050) by flow cytometry.
## Lentiviral transduction and RNA interference
Stable cell lines were generated as described previously22. Transfection with shRNA, siRNA, or cDNA was performed with TurboFect (Thermo Fisher Scientific, R0531) according to the manufacturer’s instructions. STAT3 shRNA was purchased from GeneChem. The wild-type and mutant STAT3 sequences were inserted in the pCDH-3xFLAG-GFP-puroR vector obtained from Youze Biotechnology, and the siRNA sequences were obtained from RiboBio (siSTAT3#1: GCAACAGATTGCCTGCATT; siSTAT3#2: CAACATGTCATTTGCTGAA.). The knockdown efficiency was quantified by real-time PCR and western blotting. The guide RNA sequences constituted a pool of two different sgRNA plasmids to target human GDF15, and the sgRNA sequences were as follows: GAAACTTGCGCGGCTCGCCT, TTCGAACACCGACCTCGTCC.
## RNA extraction and real-time PCR
Total RNA was extracted using MagZol (Magen, R4801). cDNA was generated using a HiScript Q RT SuperMix kit (Vazyme, R223-01). Real-time PCR was performed with SYBR Green Master Mix (Bimake, B21703). GAPDH was used as an internal control. The following primers were used: CDK1-F: GGAAACCAGGAAGCCTAGCATC; CDK1-R: GGATGATTCAGTGCCATTTTGCC; CDC6-F: GGAGATGTTCGCAAAGCACTGG;
CDC6-R: GGAATCAGAGGCTCAGAAGGTG; IL6-F: AGACAGCCACTCACCTCTTCAG; IL6-R: TTCTGCCAGTGCCTCTTTGCTG; GDF15-F: CAACCAGAGCTGGGAAGATTCG;
GDF15-R: CCCGAGAGATACGCAGGTGCA; STAT3-F: CTTTGAGACCGAGGTGTATCACC; STAT3-R: GGTCAGCATGTTGTACCACAGG; GAPDH-F: AATCCCATCACCATCTTCCA
GAPDH-R: GTCATCATATTTGGCAGGTT
## Western blotting
Cell lysates were prepared in NP-40 buffer (Beyotime, P0013F). Protein extracts were analyzed by western blotting according to the manufacturer’s protocol, as previously described22.
## Cell viability
Cell viability was measured using the Cell Counting Kit-8 (CCK-8) assay (Bimake, B34302) as previously described2. The combination index (CI) was calculated using CompuSyn software based on the Chou-Talalay method, and a CI of less than 1 indicated synergy.
## Chromatin immunoprecipitation (ChIP)-qPCR and sequencing
ChIP-qPCR and sequencing were performed as previously described2. Sonicated samples were immunoprecipitated with antibodies against STAT3 (Cell Signaling Technology, 9139 S) and BRD4 (Cell Signaling Technology, 13440). The primers used were as follows: GDF15-Chip-F, GGCAAGAACTCAGGACGGTG; GDF15-Chip-R, TCGTAGCGTTTCCGCAACT. IL6-Chip-F, GACATGCCAAAGTGCTGAGTC; IL6-Chip-R, ACTAGGGGGAAAAGTGCAGC.
## RNA-seq
A375 cells were treated with DMSO, 4 μM sunitinib, 1 μM JQ-1, or 4 μM sunitinib + 1 μM JQ-1 for 24 h. Total RNA was extracted with MagZol and used for RNA-seq analysis. Libraries were constructed and sequenced on a BGISEQ-500RS sequencer. *Only* genes with at least 1 read in each of the six samples and at least 50 reads in total among all samples were retained for subsequent analyses. Differentially expressed genes were defined as those with a | log2-fold-change | > 1 and q value < 0.05.
## Bioinformatics analysis
*The* gene expression profile datasets GSE78864, GSE122819 and GSE122821 were downloaded from the Gene Expression Omnibus (GEO) database. Genes were ranked according to the shrunken limma log2-fold changes, and the GSEA tool was used in ‘preranked’ mode with all default parameters. GSEA was performed using Java desktop software (http://software.broadinstitute.org/gsea/index.jsp). For our own RNA-seq data, GSVA and GSEA were performed using R software (3.6.3). The normalized counts were fit to a negative binomial GLM for differential expression analysis using edgeR.
## TUNEL assay
Apoptotic cells in tumor tissue were evaluated by a TUNEL assay (Beyotime Biotechnology, C1089) as previously described2. The slides were then counterstained with DAPI (Servicebio, Wuhan, China). Positively stained cells were examined by microscopy.
## ELISA for measurement of the IL6 concentration
Melanoma cells were seeded in 6-well plates and cultured overnight. After treatment, the supernatants were collected, and the IL6 concentration was measured using a human IL6 Valukine ELISA Kit (#VAL102, Novus, CO, USA) according to the manufacturer’s protocols.
## Animal study
All animal experiments were performed in accordance with protocols approved by the Ethical Review of Experimental Animals committee at Central South University. A375 cells (2 × 106) were suspended in 100 μl of PBS and inoculated into nude mice (Shanghai SLAC). Tumor-bearing mice were randomly allocated into groups. When the tumor volume reached 50-100 mm3, the mice were treated with vehicle (corn oil + citrate buffer (0.1 mol/L, pH = 3.5), orally), NHWD-870 (in corn oil, 0.75 mg/kg, orally), sunitinib (in citrate buffer, 40 mg/kg, orally), or a combination of both drugs for two days, and treatment was then stopped for one day. This process was repeated for the duration of the treatment period. The tumor size was recorded every three days, and the volume was calculated as [(length × width × width) / 2].
## Statistical analyses
All the data are presented as the means ± SDs and were analyzed with GraphPad Prism 8. Two-tailed unpaired Student’s t test was employed for comparisons between two groups. ANOVA was performed for comparisons among multiple groups. Nonparametric tests were applied if the data were nonnormally distributed. A P value of <0.05 was considered statistically significant.
## Identification of synergy between sunitinib and BET inhibitors in melanoma
To explore potential drugs from the FDA-approved drug library that synergize with BET inhibitors in melanoma, we performed a screen of 240 antitumor drugs combined with JQ1 using an in vitro drug combination assay (Fig. 1a). The coefficient of drug interaction (CDI) was used to assess the effect of the combination treatments23. Sunitinib was identified as one of the most promising drugs in both A375 and SK-MEL-28 cells, in addition to CDK$\frac{4}{6}$ inhibitors, which have been reported to synergize with BET inhibitors in cancer cells, thus supporting the validity of our screen (Fig. 1b, c). To further clarify whether JQ1 synergizes with sunitinib, we conducted a dose-response experiment with increasing drug concentrations in both A375 and SK-MEL-28 cells. The combination index (CI) was used to evaluate the interactions between BET inhibitors and sunitinib in CompuSyn software using the Chou-Talalay method. The CI values in both melanoma cell lines tended to be less than 1 and thus suggested a synergistic effect (Fig. 1d, e). Consistent with this finding, NHWD-870, another BET inhibitor we developed9, also showed strong synergistic effects with sunitinib in both A375 and SK-MEL-28 cells (Fig. 1f, g). In addition, to investigate the antiproliferative effect of cotreatment with BET inhibitors and sunitinib, a colony formation assay was performed, and the results showed that compared with sunitinib or BET inhibitors alone, combination treatment with sunitinib and BET inhibitors significantly suppressed melanoma cell proliferation (Fig. 1h, i). These findings suggested that BET inhibitors synergize with sunitinib in melanoma. Fig. 1Identification of synergy between sunitinib and BET inhibitors in melanoma cells.a Schematic of the screening process for identifying clinically applicable drugs from the FDA-approved drug library that synergize with BET inhibitors in melanoma. b Summary scatter plot of CDI values in A375 and SK-MEL-28 cells. c Targets and targeted pathways of the drugs identified through the screen. d–g Dose-response curves of melanoma cells treated with sunitinib or JQ1/NHWD-870 either alone or in combination for 36 h (JQ1 and sunitinib at a fixed ratio of 1:1, NHWD-870 and sunitinib at a fixed ratio of 1:1000). Synergy was assessed by the Chou-Talalay combination index (CI) for sunitinib and BET inhibitors across the indicated cell lines. The x-axis on the CI plots shows the percentage of cells affected. h, i Colony formation assay of A375 and SK-MEL-28 cells after the indicated treatment. P values were calculated using one-way ANOVA in (h) and (i). * $P \leq 0.05$; ***$P \leq 0.001.$
## Apoptosis and cell cycle arrest mediate the synergy induced by the combination treatment
To further clarify the underlying mechanism of the synergy induced by combination treatment with BET inhibitors and sunitinib, we performed RNA-seq analysis on the following groups of A375 melanoma cells treated for 24 h as indicated: control; JQ1, 1 µM; sunitinib, 4 µM; combination, JQ1, 1 µM and sunitinib, 4 µM. Gene set enrichment analysis (GSEA) was performed and showed that cell cycle progression was significantly inhibited in the combination group compared with the other groups (Fig. 2a). To determine which genes were suppressed in the cell cycle gene set, we drew a Venn diagram and heatmap and found that 17 genes associated with the cell cycle were dramatically inhibited by cotreatment with BET inhibitors and sunitinib (Fig. 2b, c). Among these genes, Cyclin Dependent Kinase 1 (CDK1) is best known for its function as a master regulator of the cell cycle24. Cell division cycle 6 (CDC6) is an essential regulator of DNA replication in eukaryotic cells25. The changes in the expression of these genes were confirmed by quantitative real-time PCR analysis (Supplementary Fig. 1a, b). We further evaluated cell cycle progression in both A375 and SK-MEL-28 cells using flow cytometry. We found that cotreatment with BET inhibitors and sunitinib significantly induced G1 arrest compared with sunitinib or BET inhibitors alone (Fig. 2d, e). We also used gene set variation analysis (GSVA) to explore the effects of the drug treatments on cell death-related pathways, including apoptosis, autophagy, necrosis and ferroptosis. We found that apoptosis and autophagy pathways were dramatically activated in the combination group compared with the other groups (Fig. 2f). Furthermore, the toxic effect of the combination therapy was partially negated by an inhibitor of apoptosis (Z-VAD-FMK) but not by inhibitors of ferroptosis (ferrostatin-1), necroptosis (Nec-1s), or autophagy (CQ) (Fig. 2g). Consistent with these results, Z-VAD-FMK still partially inhibit the death of A375 cells induced by cotreatment with JQ1 and different concentrations of sunitinib (Fig. 2h). Flow cytometry further demonstrated that treatment with BET inhibitors in combination with sunitinib resulted in a much higher proportion of apoptotic cells than sunitinib or BET inhibitors alone (Fig. 2i, j; Supplementary Fig. 1c, d), suggesting that the combination of BET inhibitors and sunitinib drives melanoma cell apoptosis. Altogether, these results indicated that cotreatment with BET inhibitors and sunitinib inhibits melanoma progression by significantly promoting cell cycle arrest and apoptosis. Fig. 2Apoptosis and cell cycle arrest mediate the synergy induced by the combination treatment.a Gene set enrichment analysis (GSEA) showing that cell cycle progression was significantly inhibited in the combination group compared with the DMSO, sunitinib, and JQ1 groups. b Venn diagram of the overlapping genes in the indicated groups. c Heatmap of the 17 genes identified by the Venn diagram that were associated with the cell cycle and inhibited by the combination treatment. d, e Cell cycle distribution of A375 (d) and SK-MEL-28 (e) cells after treatment with sunitinib (1 μM) or JQ1 (1 μM)/NHWD-870 (10 nM) either alone or in combination for 36 h. f Gene set variation analysis (GSVA) of apoptosis-, autophagy-, necrosis-, and ferroptosis-related pathways in the indicated groups. g SK-MEL-28 cells were treated with JQ1 (1 μM), sunitinib (1 μM), or a combination of both drugs with or without cell death inhibitors (CQ, 10 μM; Fer-1, 2 μM; necrostatin-1s, 10 μM; ZVAD-FMK, 5 μM) for 24 h, and cell viability was assessed. h Dose response of sunitinib-induced death of SK-MEL-28 cells treated with JQ1 in the absence or presence of ZVAD-FAK. i, j Apoptosis of A375 (i) and SK-MEL-28 (j) cells after treatment with sunitinib (1 μM) or JQ1 (1 μM)/NHWD-870 (10 nM) either alone or in combination for 36 h. P values were calculated using one-way ANOVA in (f, g, i and j). * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ns nonsignificant.
## STAT3 regulates sunitinib sensitivity in melanoma cells through its phosphorylation
STAT3 inhibition has been implicated in cell cycle arrest and apoptosis26–28. Additionally, the STAT3 signaling pathway is associated with drug sensitivity in melanoma29–31. Thus, we focused our attention on the association of STAT3 signaling with sunitinib sensitivity. By analyzing the associations between sunitinib sensitivity and the activity of STAT3 signaling using the Cancer Therapeutics Response Portal, we found that STAT3 signaling was significantly activated in cells with a higher IC50 of sunitinib (Fig. 3a). Consistent with these results, GSEA demonstrated that the IL6/JAK/STAT3 pathway was markedly activated in STAT3-resistant cells (Fig. 3b). We also generated sunitinib-resistant melanoma cells (Supplementary Fig. 2) and found that the phosphorylation of STAT3 was increased in sunitinib-resistant melanoma cells (Fig. 3c). Furthermore, we generated STAT3 knockdown melanoma cells by transduction of two different shRNAs (Fig. 3d; Supplementary Fig. 3a) and found that STAT3 inhibition significantly enhanced melanoma cell sensitivity to sunitinib (Fig. 3e). Stattic, an inhibitor of STAT3, which is reported to impede the phosphorylation of STAT3 (Fig. 3f), sensitized both A375 and SK-MEL-28 melanoma cells to sunitinib (Fig. 3g). STAT3 functions mainly through its phosphorylation on tyrosine (Y705) and serine (S727) residues, which is important for its dimerization, translocation and transcriptional activation2. Thus, we sought to determine the site of STAT3 phosphorylation that mediates melanoma cell sensitivity to sunitinib. In STAT3 knockdown cells, we ectopically expressed wild-type STAT3 or STAT3 phosphorylation mutants (Fig. 3h). Notably, coexpression of wild-type STAT3 but not the STAT3 phosphorylation mutants, attenuated the shSTAT3-induced sensitization of melanoma cells to sunitinib (Fig. 3i). Taken together, these results demonstrated that STAT3 inhibition enhances melanoma cell sensitivity to sunitinib via suppression of STAT3 phosphorylation. Fig. 3STAT3 regulates sunitinib sensitivity in melanoma cells through its phosphorylation.a GSVA scores of STAT3 targets in cells with a high IC50 of sunitinib or a low IC50 of sunitinib from the Cancer Therapeutics Response Portal dataset. b GSEA of the hallmark IL6/JAK/STAT3 signaling pathway in the indicated parental and sunitinib-resistant cells. c Western blot analysis of the indicated proteins in parental and sunitinib-resistant A375 cells. d Western blot analysis of the indicated proteins in shCtrl and shSTAT3 melanoma cells. e Dose response of sunitinib-induced death in shCtrl and shSTAT3 melanoma cells over a 24 h period. f Western blot analysis of the indicated proteins in A375 and SK-MEL-28 cells after treatment with DMSO, 0.5 μM stattic, or 1 μM stattic for 24 h. g Dose response of sunitinib-induced death in A375 and SK-MEL-28 cells in the presence of DMSO, 0.5 μM stattic, or 1 μM stattic for 24 h. h Western blot analysis of the indicated proteins in shSTAT3 A375 and SK-MEL-28 cells after the expression of the empty vector, wild-type STAT3 plasmid, or STAT3 phosphorylation mutant plasmids. i Viability of the indicated cells after treatment with sunitinib for 24 h. Two-tailed unpaired Student’s t test was performed in (a). Nonlinear regression was applied in (e and g). P values were calculated using one-way ANOVA in (i). *** $P \leq 0.001.$
## BET inhibitors suppress STAT3 signaling via the BRD4/IL6 axis
BET inhibitors have been reported to modulate drug sensitivity in cancer cells by specifically regulating the expression of certain genes or the activity of signaling pathways11,12. Thus, we sought to determine whether BET inhibitors can suppress STAT3 signaling to sensitize melanoma cells to sunitinib. We reanalyzed the RNA-seq data of JQ1-treated vs. control cells. As expected, the expression of c-Myc targets was dramatically inhibited after JQ1 treatment (Fig. 4a). However, the targets with STAT3-induced downregulation were significantly upregulated in the JQ1 group, suggesting that JQ1 can markedly inhibit IL6/STAT3 signaling activity (Fig. 4a). Subsequently, we measured the IL6 concentration in the supernatant using an ELISA kit and found that the IL6 concentration was markedly reduced in melanoma cells after BET inhibitor treatment (Fig. 4b). Consistent with these results, the phosphorylation of STAT3 was also dramatically decreased after BET inhibitor treatment (Fig. 4c, d).Fig. 4BET inhibitors repress STAT3 signaling via the BRD4/IL6 axis.a GSEA of MYC targets and targets with STAT3-induced downregulation in the JQ1 and DMSO groups. b IL6 concentrations in supernatants were measured by ELISA after treatment with DMSO, 1 μM JQ1, or 10 nM NHWD-870 for 24 h. c, d Western blotting and qualitative analysis of the indicated proteins in A375 and SK-MEL-28 cells after treatment with DMSO, 1 μM JQ1, or 10 nM NHWD-870 for 24 h. e Knockdown efficiency of BRD4 quantified by western blotting. f IL6 concentrations in supernatants were measured by ELISA after BRD4 silencing. g, h Western blotting and qualitative analysis of the indicated proteins in A375 and SK-MEL-28 cells after BRD4 silencing. i GSEA of targets with STAT3-induced downregulation identified by RNA-seq of siNC vs. siBRD4 A375 cells. j Heatmap of the top five differentially expressed genes in the gene set. k Assessment of the IL6 mRNA level by RT-PCR in A375 cells after BRD4 silencing. l Assessment of IL6 mRNA levels by RT-PCR in A375 and SK-MEL-28 cells after treatment with DMSO, 1 μM JQ1, or 10 nM NHWD-870 for 24 h. m BRD4 binding peaks in the IL6 promoter in DMSO-treated, NHWD-870-treated, siNC-treated, and siBRD4-treated A375 cells. n ChIP-qPCR analysis of the IL6 promoter in A375 cells with an anti-BRD4 antibody or IgG after treatment with DMSO or 1 μM JQ1. P values were calculated using one-way ANOVA in (b, d, f, h, k and l). Two-way ANOVA was performed in (n). * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ns nonsignificant.
BET inhibitors function primarily by competitively binding to BET proteins and disrupting the interaction between BET proteins and acetylated lysine residues32. To determine which BET proteins are responsible for IL6/STAT3 pathway signaling, we silenced BRD$\frac{2}{3}$/4 individually using shRNA (Fig. 4e; Supplementary Fig. 3b, c). The results showed that knockdown of only the BRD4 protein markedly reduced the IL6 concentration and the phosphorylation of STAT3 (Fig. 4f–h; Supplementary Fig. 3d). We also silenced BRD4 using siRNA and performed RNA-seq, finding that STAT3 signaling was markedly inhibited after BRD4 silencing (Fig. 4i). The top five differentially expressed genes in the gene set were visualized using a heatmap, which suggested that the mRNA level of IL6 was significantly reduced after BRD4 silencing (Fig. 4j). Consistent with the RNA-seq results, BRD4 knockdown and BET inhibitor treatment markedly reduced the mRNA level of IL6, according to quantitative real-time PCR analysis (Fig. 4k, l). Reanalysis of an existing BRD4 ChIP-seq dataset suggested that there is a striking BRD4 binding peak in the IL6 gene promoter, while the amplitude of the binding peak was diminished with NHWD-870 treatment or BRD4 knockdown (Fig. 4m). The results were also validated by ChIP-qPCR (Fig. 4n; Supplementary Fig. 3e). These findings suggested that BET inhibitors suppress STAT3 signaling via the BRD4/IL6 axis.
## BET inhibitors regulate sunitinib sensitivity by inhibiting STAT3 activity and GDF15 expression
We next sought to determine whether STAT3 signaling mediates BET inhibitor-induced sensitization of melanoma cells to sunitinib. Control and STAT3-silenced melanoma cells were incubated with sunitinib or cotreated with sunitinib and BET inhibitors. A marked degree of sensitization to sunitinib was observed after BET inhibitor treatment in control cells but not in STAT3-silenced cells (Fig. 5a). Pharmacologically, stattic sensitized melanoma cells to sunitinib but failed to further enhance the sensitization of melanoma cells to sunitinib in the presence of BET inhibitors (Fig. 5b). These results suggested that STAT3 signaling is the mediator of BET inhibitor-induced sensitization of melanoma cells to sunitinib. To further elucidate the specific protein that mediates BET inhibitor-induced sensitization of melanoma cells to sunitinib, we overlapped the downregulated differentially expressed genes after BET inhibitor treatment and after BRD4 silencing (Fig. 5c). A Venn diagram identified 35 DEGs, among which GDF15 and IL1A exhibited markedly increased mRNA levels in sunitinib-resistant cells and comparable mRNA levels after long-term withdrawal of sunitinib (Fig. 5d, e). We further found that the expression of GDF15 but not IL1A was positively associated with the logIC50 of sunitinib using Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Genome Project (CGP) datasets (Fig. 5f; Supplementary Fig. 4a, b). Subsequently, quantitative real-time PCR analysis validated that the mRNA level of GDF15 was markedly reduced after BET inhibitor treatment (Supplementary Fig. 4c). GDF15 expression was significantly increased in sunitinib-resistant melanoma cells (Fig. 5g, h; Supplementary Fig. 4d, e). We further generated GDF15 knockout cells using CRISPR/Cas9 technology (Fig. 5i; Supplementary Fig. 4f) and found that GDF15 knockout sensitized melanoma cells to sunitinib (Fig. 5j; Supplementary Fig. 4g). In contrast, GDF15 overexpression reduced cell sensitivity to sunitinib (Fig. 5k, l; Supplementary Fig. 4h, i). These findings suggested that BET inhibitors regulate melanoma cell sensitivity to sunitinib by inhibiting STAT3 activity and GDF15 expression. Fig. 5BET inhibitors regulate sunitinib sensitivity by inhibiting STAT3 activity and GDF15 expression.a The viability of shCtrl and shSTAT3 melanoma cells was evaluated after treatment with sunitinib (1 μM) alone or in combination with JQ1 (1 μM)/NHWD-870 (10 nM). b Dose response of sunitinib-induced death in A375 and SK-MEL-28 cells in the presence of DMSO, 1 μM stattic, or 1 μM stattic + JQ1 (1 μM)/NHWD-870 (10 nM). c Venn diagram of the overlapping genes in the indicated groups. d Heatmap of the overlapping genes in parental cells, sunitinib-resistant cells, and cells with long-term withdrawal of sunitinib from the GSE122821 dataset. If the mRNA level of a gene was markedly increased in sunitinib-resistant cells and comparable to that in parental cells after long-term withdrawal of sunitinib, the gene is marked with an asterisk and shown in red. e Fold changes in GDF15 and IL1A expression in parental cells, sunitinib-resistant cells, and cells with long-term withdrawal of sunitinib from the GSE122821 dataset. f Association between the expression of GDF15 and the logIC50 of sunitinib in Genomics of Drug Sensitivity in Cancer and Cancer Genome Project datasets. g GDF15 mRNA levels in parental and sunitinib-resistant A375 cells. h Western blotting and qualitative analysis of GDF15 expression in parental and sunitinib-resistant A375 cells. i GDF15 protein levels were quantified by western blotting in control (sgCtrl) and GDF15-deficient (sgGDF15) cells. j Dose response of sunitinib-induced death in sgCtrl and sgGDF15 A375 cells over a 24 h period. k GDF15 protein levels were quantified by western blotting in control (Flag vector) and GDF15 overexpression (Flag-GDF15) cells. l Dose response of sunitinib-induced death in vector and GDF15-overexpressing A375 cells over a 24 h period. Two-way ANOVA was performed in (a). Nonlinear regression was applied in (b, j and l). One-way ANOVA was performed in (e). Two-tailed unpaired Student’s t test was performed in (g and h). * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001.$
## BRD4 regulates GDF15 expression by directly targeting its promoter or indirectly targeting the IL6/STAT3 axis
To further elucidate the regulatory mechanism of GDF15 in melanoma cells, we first measured the expression of GDF15 in STAT3-silenced cells and found that the expression of GDF15 in STAT3-silenced cells was significantly reduced compared with that in control cells (Fig. 6a, b). The expression of GDF15 was also dramatically decreased after stattic treatment (Fig. 6c, d). Analysis of published ChIP-seq data demonstrated that there was a strikingly enhanced STAT3‐binding peak in the GDF15 promoter in both OCI-Ly10 and OCI-Ly19 cells (Fig. 6e). Based on the identified peak, we designed primers for ChIP‐qPCR analysis, confirming that STAT3 binds to the GDF15 promoter in melanoma cells (Fig. 6f; Supplementary Fig. 4j). It is worth noting that BET inhibitor treatment inhibited GDF15 expression more strongly than STAT3 silencing or stattic treatment (Supplementary Fig. 4c, k), suggesting that BET inhibitors suppress GDF15 expression in another manner in addition to through the BRD4/IL6/STAT3 axis. We previously showed that BRD4 could function as a transcription factor to regulate the expression of certain genes10–12. Therefore, we sought to determine whether BRD4 can directly regulate the expression of GDF15. We incubated STAT3-silenced cells with BET inhibitors and found that the expression of GDF15 was significantly reduced (Fig. 6g, h), which is in line with the findings in stattic-treated cells (Fig. 6i, j). Reanalysis of ChIP-seq data suggested that BRD4 binding to the GDF15 promoter was diminished by BRD4 inhibition and enhanced in BRD4-overexpressing cells (Fig. 6k). These findings were validated by ChIP-qPCR (Fig. 6l; Supplementary Fig. 4l). To further visualize the binding sites of BRD4 and STAT3 in the GDF15 promoter, the ENCODE database and UCSC Genome Browser were used and showed that there were strikingly enhanced peaks in the GDF15 promoter and that the peaks of BRD4 and STAT3 in the GDF15 promoter did not overlap (Supplementary Fig. 5). These results indicated that GDF15 is transcriptionally regulated directly by BRD4 or indirectly by the BRD4/IL6/STAT3 axis. Fig. 6BRD4 regulates GDF15 expression by directly targeting its promoter or indirectly targeting the IL6/STAT3 axis.a GDF15 mRNA levels in A375 and SK-MEL-28 cells after STAT3 silencing. b Western blotting and qualitative analysis of GDF15 expression in control and STAT3-silenced cells. c GDF15 mRNA levels in A375 and SK-MEL-28 cells after stattic treatment. d Western blotting and qualitative analysis of GDF15 expression in DMSO- and stattic-treated cells. e STAT3 binding peak in the GDF15 promoter in OCI-Ly10 and OCI-Ly19 cells (GSE50723). f Validation of STAT3 binding to the promoter of GDF15 in A375 cells by ChIP-qPCR. g GDF15 mRNA levels in STAT3-silenced cells after treatment with DMSO, 1 μM JQ1, or 10 nM NHWD-870 for 24 h. h Western blotting and qualitative analysis of GDF15 expression in STAT3-silenced cells after treatment with DMSO, 1 μM JQ1, or 10 nM NHWD-870 for 24 h. i GDF15 mRNA levels in stattic-treated cells after treatment with DMSO, 1 μM JQ1, or 10 nM NHWD-870 for 24 h. j Western blotting and qualitative analysis of GDF15 expression in stattic-treated cells after treatment with DMSO, 1 μM JQ1, or 10 nM NHWD-870 for 24 h. k BRD4 binding peak in the GDF15 promoter in siNC-treated and siBRD4-treated A375 cells (upper); BRD4 binding peak in the GDF15 promoter in DMSO-treated, JQ1-treated and BRD4-overexpressing cells (lower). l ChIP-qPCR analysis of the GDF15 promoter in A375 cells with an anti-BRD4 antibody or IgG after treatment with DMSO or 1 μM JQ1. One-way ANOVA was performed in (a, b, g, h, i and j). Two-tailed unpaired Student’s t test was performed in (c and d). T test with Welch’s correction was performed in (f). Two-way ANOVA was performed in (l). * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ns nonsignificant.
## Combination of BET inhibitors with sunitinib causes melanoma suppression in vivo
To evaluate the therapeutic potential of combining BET inhibitors with sunitinib in melanoma, A375 cells were inoculated into the right flanks of nude mice to establish subcutaneous xenograft models. When the tumor volume reached 50-100 mm3, the tumor-bearing mice were randomly allocated into groups as follows: vehicle, sunitinib, NHWD-870 or combination treatment using sunitinib and NHWD-870. All treatments were administered for two successive days and stopped for 1 day (Fig. 7a). The results showed that the tumor volume and weight were further markedly decreased in the combination therapy group compared with the other groups, without significant changes in body weight (Fig. 7b, c; Supplementary Fig. 6a). Real-time PCR analysis of tumors showed that GDF15 and IL6 mRNA expression was significantly decreased in the NHWD-870 and combination therapy groups (Supplementary Fig. 6b, c). IHC staining also showed that the protein levels of GDF15, p-STAT3 (Y705) and p-STAT3 (S727) were significantly decreased in the NHWD-870 and combination therapy groups (Fig. 7d; Supplementary Fig. 6d). IHC staining for the proliferation marker Ki67 revealed fewer proliferative cells in the combination therapy group (Fig. 7e). The TUNEL assay indicated a significantly increased number of apoptotic cells in the combination therapy group (Fig. 7f). These results further supported the synergistic effect of the combination of BET inhibitors and sunitinib on inducing tumor suppression in vivo. Taken together, our results strongly demonstrated that combination therapy using BET inhibitors and sunitinib has immense therapeutic potential in melanoma. Mechanistically, BET inhibitors sensitize melanoma cells to sunitinib by inhibiting the BRD4/GDF15 axis and the BRD4/IL6/STAT3/GDF15 axis (Fig. 7g).Fig. 7Combination treatment with BET inhibitors and sunitinib causes melanoma suppression in vivo.a Schedule for administration of sunitinib (40 mg/kg) and NHWD-870 (0.75 mg/kg) in tumor-bearing mice. b, c Tumor weight (b) and tumor volume (c) in the vehicle, NHWD-870, sunitinib, and combination groups. d Quantification of IHC staining of GDF15, p-STAT3 (Y705), and p-STAT3 (S727) in the sectioned tumors. e Ki67 staining of the sectioned tumors was performed to identify tumor cell proliferation in the vehicle, sunitinib, NHWD-870 and combination groups. f A TUNEL assay was performed to quantify apoptotic cells in xenograft tumors in the vehicle, sunitinib, NHWD-870 and combination groups. g A proposed working model. BET inhibitors synergize with sunitinib in melanoma cells by further inducing apoptosis and cell cycle arrest. Mechanistically, BET inhibitors sensitize melanoma cells to sunitinib by inhibiting the IL6/STAT3 signaling pathway and GDF15 expression. GDF15 is transcriptionally directly regulated by BRD4 or indirectly regulated by the BRD4/IL6/STAT3 axis. One-way ANOVA was performed in (b, d, e and f). Brown-Forsythe and Welch ANOVA was performed in (c). * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ns nonsignificant.
## Discussion
BET inhibitors, including NHWD-870 developed by our team, have shown profound efficacy against hematologic and solid tumors in preclinical studies9,10,33. However, several resistance mechanisms limit the efficacy of BET inhibitors in melanoma. For example, Ambrosini et al. demonstrated that NF-κB signaling was significantly activated in BETi-resistant melanoma cells and that inhibition of NF-κB signaling enhanced BET inhibitor sensitivity in melanoma34. Moreover, Chua et al. showed that FGF2 secreted from hepatic stellate cells (HSCs) protected melanoma cells against growth inhibition induced by BET inhibitors, identifying a potential mechanism underlying the resistance of melanoma liver metastases to BET inhibitors35. Therefore, it is of great significance to develop novel strategies to better kill melanoma cells with BET inhibitor treatment.
In the present study, we identified sunitinib as a clinically applicable drug that synergizes with BET inhibitors in melanoma cells through a screen of 240 FDA-approved antitumor drugs. Sunitinib is an indolinone derivative and tyrosine kinase inhibitor approved for clear cell renal cell carcinoma and gastrointestinal stromal tumor treatment. We previously demonstrated that sunitinib had a therapeutic effect on melanoma21. Moreover, Minor et al. showed that melanoma patients with KIT mutations had a good response to sunitinib36. Sunitinib is well known to exert antiangiogenic activity by inhibiting CSF1R, CSF3R, FLT1, FLT3, FLT4, KDR, KIT, PDGFRA, PDGFRB and RET. By analyzing the associations of these targets with immunosuppressive cells and inhibitory immune checkpoints, we previously found that sunitinib treatment was associated with T-cell infiltration and activity and that sunitinib showed a synergistic antitumor effect with an anti-CTLA-4 monoclonal antibody in melanoma through the P62/PD-L1 axis20. Additionally, sunitinib was reported to play a critical role in immune surveillance by inducing a Th-1 immune response or reducing the population of MDSCs37,38. Here, we further revealed that BET inhibitors synergize with sunitinib in melanoma. These findings highlight the importance of sunitinib in melanoma treatment.
We further showed that BET inhibitors sensitize melanoma cells to sunitinib by repressing GDF15 expression in a direct way by targeting its promoter or indirectly by targeting the IL6/STAT3 axis. GDF15, also called NSAID-activated gene-1 (NAG-1), is associated with multiple biological processes and diseases, including cancer39. GDF15 expression was found to be increased in melanoma metastases compared with benign nevi or primary melanomas and was positively correlated with stage in melanoma patients40. High GDF15 serum levels were also correlated with poorer overall survival in melanoma patients41. In addition, GDF15 has been extensively reported to act as an antiapoptotic protein42–46. Mechanistically, GDF15 inhibits apoptosis by promoting rapid and transient Ser473 phosphorylation (activation) of Akt and a subsequent increase in Ser136 phosphorylation (inactivation) of Bad, which is well known to promote apoptosis through the intrinsic mitochondrial pathway43. This finding was consistent with Liu et al. ’s study showing that inhibition of GDF15 stabilized PTEN, in turn inactivating the PI3K/AKT pathway and finally inducing cancer cell apoptosis44. As expected, increased expression of GDF15 in B16F1 melanoma cells promoted tumor growth in the B16F1 melanoma mouse model47. Therefore, GDF15 might be a potential therapeutic target in melanoma.
The association between GDF15 expression and sunitinib sensitivity has never been reported. Sunitinib was reported to induce cell apoptosis in renal cell carcinoma via STAT3 inhibition48. We further demonstrated that STAT3 signaling was significantly activated in sunitinib-resistant melanoma cells and that inhibition of STAT3 enhanced the sensitivity of melanoma to sunitinib. Furthermore, inhibition of STAT3 decreased the expression of GDF15, and GDF15 silencing increased the sensitivity of melanoma cells to sunitinib. These results implied that there is a negative association between GDF15 expression and sunitinib sensitivity. Considering the antiapoptotic role of GDF15, it is reasonable to speculate that GDP15 leads to sunitinib resistance in melanoma by inhibiting apoptosis.
Additionally, BET inhibitors can directly regulate GDF15 expression by competitively binding to BRD4 and reducing its distribution in the promoter of GDF15. These results were validated in STAT3 knockdown and stattic-treated cells. The direct regulation of GDF15 expression by BRD4 has never been clarified. Guo et al. recently showed that BRD4 upregulated the expression of GDF15 by inducing NR5A2 transcriptional activation49. However, this group failed to evaluate the direct association of BRD4 and GDF15. Through ChIP‒qPCR, we found that BRD4 directly bound to the promoter of GDF15 and that the binding was decreased by BRD4 silencing or inhibition and promoted by BRD4 overexpression. Consistent with these results, the xenograft assays further validated that GDF15 expression was significantly decreased in the BET inhibitor-treated group and combination therapy group compared to the DMSO and sunitinib groups.
In summary, we identified an FDA-approved drug, sunitinib, that synergizes with BET inhibitors in melanoma. Mechanistically, BET inhibitors sensitize melanoma cells to sunitinib by inhibiting the BRD4/GDF15 axis and BRD4/IL6/STAT3/GDF15 axis. These findings are potentially translatable toward novel therapies for melanoma and other diseases that can be cotreated with BET inhibitors and sunitinib.
## Supplementary information
Supplementary figures The online version contains supplementary material available at 10.1038/s12276-023-00936-y.
## References
1. Schadendorf D. **Melanoma**. *Lancet* (2018.0) **392** 971-984. DOI: 10.1016/S0140-6736(18)31559-9
2. Deng G. **EEF2K silencing inhibits tumour progression through repressing SPP1 and synergises with BET inhibitors in melanoma**. *Clin. Transl. Med.* (2022.0) **12** e722. DOI: 10.1002/ctm2.722
3. Luebker SA, Koepsell SA. **Diverse mechanisms of BRAF inhibitor resistance in melanoma identified in clinical and preclinical studies**. *Front. Oncol.* (2019.0) **9** 268. DOI: 10.3389/fonc.2019.00268
4. Song C. **Recurrent tumor cell-intrinsic and -extrinsic alterations during MAPKi-induced melanoma regression and early adaptation**. *Cancer Discov.* (2017.0) **7** 1248-1265. DOI: 10.1158/2159-8290.CD-17-0401
5. Curti BD, Faries MB. **Recent advances in the treatment of melanoma**. *N. Engl. J. Med.* (2021.0) **384** 2229-2240. DOI: 10.1056/NEJMra2034861
6. Shin DS. **Primary resistance to PD-1 blockade mediated by JAK1/2 mutations**. *Cancer Discov.* (2017.0) **7** 188-201. DOI: 10.1158/2159-8290.CD-16-1223
7. Zaretsky JM. **Mutations associated with acquired resistance to PD-1 blockade in melanoma**. *N. Engl. J. Med.* (2016.0) **375** 819-829. DOI: 10.1056/NEJMoa1604958
8. Segura MF. **BRD4 sustains melanoma proliferation and represents a new target for epigenetic therapy**. *Cancer Res.* (2013.0) **73** 6264-6276. DOI: 10.1158/0008-5472.CAN-13-0122-T
9. Yin M. **Potent BRD4 inhibitor suppresses cancer cell-macrophage interaction**. *Nat. Commun.* (2020.0) **11** 1833. DOI: 10.1038/s41467-020-15290-0
10. Deng G. **BET inhibitor suppresses melanoma progression via the noncanonical NF-kappaB/SPP1 pathway**. *Theranostics* (2020.0) **10** 11428-11443. DOI: 10.7150/thno.47432
11. Yang L. **Repression of BET activity sensitizes homologous recombination-proficient cancers to PARP inhibition**. *Sci. Transl. Med.* (2017.0) **9** eaal1645. DOI: 10.1126/scitranslmed.aal1645
12. Kanojia D. **BET inhibition increases betaIII-tubulin expression and sensitizes metastatic breast cancer in the brain to vinorelbine**. *Sci. Transl. Med.* (2020.0) **12** eaax2879. DOI: 10.1126/scitranslmed.aax2879
13. Echevarria-Vargas IM. **Co-targeting BET and MEK as salvage therapy for MAPK and checkpoint inhibitor-resistant melanoma**. *Embo. Mol. Med.* (2018.0) **10** e8446. DOI: 10.15252/emmm.201708446
14. Fallahi-Sichani M. **Adaptive resistance of melanoma cells to RAF inhibition via reversible induction of a slowly dividing de-differentiated state**. *Mol. Syst. Biol.* (2017.0) **13** 905. DOI: 10.15252/msb.20166796
15. Korkut A. **Perturbation biology nominates upstream-downstream drug combinations in RAF inhibitor resistant melanoma cells**. *Elife* (2015.0) **4** e04640. DOI: 10.7554/eLife.04640
16. Paoluzzi L. **BET and BRAF inhibitors act synergistically against BRAF-mutant melanoma**. *Cancer Med.* (2016.0) **5** 1183-1193. DOI: 10.1002/cam4.667
17. Tiago M. **Targeting BRD/BET proteins inhibits adaptive kinome upregulation and enhances the effects of BRAF/MEK inhibitors in melanoma**. *Br. J. Cancer* (2020.0) **122** 789-800. DOI: 10.1038/s41416-019-0724-y
18. Zhao B, Cheng X, Zhou X. **The BET-bromodomain inhibitor JQ1 mitigates vemurafenib drug resistance in melanoma**. *Melanoma Res.* (2018.0) **28** 521-526. DOI: 10.1097/CMR.0000000000000497
19. Emran AA. **A combination of epigenetic BET and CDK9 inhibitors for treatment of human melanoma**. *J. Invest. Dermatol.* (2021.0) **141** 2238-2249.e2212. DOI: 10.1016/j.jid.2020.12.038
20. Li H. **The beneficial role of sunitinib in tumor immune surveillance by regulating tumor PD-L1**. *Adv. Sci. (Weinh.)* (2021.0) **8** 2001596. PMID: 33510997
21. Kuang X. **Propranolol enhanced the anti-tumor effect of sunitinib by inhibiting proliferation and inducing G0/G1/S phase arrest in malignant melanoma**. *Oncotarget* (2018.0) **9** 802-811. DOI: 10.18632/oncotarget.22696
22. Deng G. **BECN2 (beclin 2) negatively regulates inflammasome sensors through ATG9A-dependent but ATG16L1- and LC3-independent non-canonical autophagy**. *Autophagy* (2022.0) **18** 340-356. DOI: 10.1080/15548627.2021.1934270
23. Simpkins F. **Dual Src and MEK inhibition decreases ovarian cancer growth and targets tumor initiating stem-like cells**. *Clin. Cancer Res.* (2018.0) **24** 4874-4886. DOI: 10.1158/1078-0432.CCR-17-3697
24. Enserink JM, Chymkowitch P. **Cell cycle-dependent transcription: the cyclin dependent kinase Cdk1 is a direct regulator of basal transcription machineries**. *Int. J. Mol. Sci.* (2022.0) **23** 1293. DOI: 10.3390/ijms23031293
25. Borlado LR, Mendez J. **CDC6: from DNA replication to cell cycle checkpoints and oncogenesis**. *Carcinogenesis* (2008.0) **29** 237-243. DOI: 10.1093/carcin/bgm268
26. Bai L. **A potent and selective small-molecule degrader of STAT3 achieves complete tumor regression in vivo**. *Cancer Cell* (2019.0) **36** 498-511.e417. DOI: 10.1016/j.ccell.2019.10.002
27. Zhao Y. **Induction of cell cycle arrest and apoptosis by CPUC002 through stabilization of p53 and suppression of STAT3 signaling pathway in multiple myeloma**. *Cell Biol. Toxicol.* (2021.0) **37** 97-111. DOI: 10.1007/s10565-020-09565-x
28. Zhou C. **Down-regulation of STAT3 induces the apoptosis and G1 cell cycle arrest in esophageal carcinoma ECA109 cells**. *Cancer Cell Int.* (2018.0) **18** 53. DOI: 10.1186/s12935-018-0549-4
29. Carson R. **HDAC inhibition overcomes acute resistance to MEK inhibition in BRAF-mutant colorectal cancer by downregulation of c-FLIPL**. *Clin. Cancer Res.* (2015.0) **21** 3230-3240. DOI: 10.1158/1078-0432.CCR-14-2701
30. Jiang X, Zhou J, Giobbie-Hurder A, Wargo J, Hodi FS. **The activation of MAPK in melanoma cells resistant to BRAF inhibition promotes PD-L1 expression that is reversible by MEK and PI3K inhibition**. *Clin. Cancer Res.* (2013.0) **19** 598-609. DOI: 10.1158/1078-0432.CCR-12-2731
31. Zhao K. **Morusin enhances the antitumor activity of MAPK pathway inhibitors in BRAF-mutant melanoma by inhibiting the feedback activation of STAT3**. *Eur. J. Cancer* (2022.0) **165** 58-70. DOI: 10.1016/j.ejca.2022.01.004
32. Sun HY, Du ST, Li YY, Deng GT, Zeng FR. **Bromodomain and extra-terminal inhibitors emerge as potential therapeutic avenues for gastrointestinal cancers**. *World J. Gastrointest. Oncol.* (2022.0) **14** 75-89. DOI: 10.4251/wjgo.v14.i1.75
33. Zhang Z. **BET bromodomain inhibition as a therapeutic strategy in ovarian cancer by downregulating FoxM1**. *Theranostics* (2016.0) **6** 219-230. DOI: 10.7150/thno.13178
34. Ambrosini G. **Inhibition of NF-kappaB-dependent signaling enhances sensitivity and overcomes resistance to BET inhibition in uveal melanoma**. *Cancer Res.* (2019.0) **79** 2415-2425. DOI: 10.1158/0008-5472.CAN-18-3177
35. Chua V. **Stromal fibroblast growth factor 2 reduces the efficacy of bromodomain inhibitors in uveal melanoma**. *Embo. Mol. Med.* (2019.0) **11** e9081. DOI: 10.15252/emmm.201809081
36. Minor DR. **Sunitinib therapy for melanoma patients with KIT mutations**. *Clin. Cancer Res.* (2012.0) **18** 1457-1463. DOI: 10.1158/1078-0432.CCR-11-1987
37. Finke JH. **Sunitinib reverses type-1 immune suppression and decreases T-regulatory cells in renal cell carcinoma patients**. *Clin. Cancer Res.* (2008.0) **14** 6674-6682. DOI: 10.1158/1078-0432.CCR-07-5212
38. Ko JS. **Sunitinib mediates reversal of myeloid-derived suppressor cell accumulation in renal cell carcinoma patients**. *Clin. Cancer Res.* (2009.0) **15** 2148-2157. DOI: 10.1158/1078-0432.CCR-08-1332
39. Wang D. **GDF15: emerging biology and therapeutic applications for obesity and cardiometabolic disease**. *Nat. Rev. Endocrinol.* (2021.0) **17** 592-607. DOI: 10.1038/s41574-021-00529-7
40. Kluger HM. **Plasma markers for identifying patients with metastatic melanoma**. *Clin. Cancer Res.* (2011.0) **17** 2417-2425. DOI: 10.1158/1078-0432.CCR-10-2402
41. Weide B. **High GDF-15 serum levels independently correlate with poorer overall survival of patients with tumor-free stage III and unresectable stage IV melanoma**. *J. Invest. Dermatol.* (2016.0) **136** 2444-2452. DOI: 10.1016/j.jid.2016.07.016
42. Nakayasu ES. **Comprehensive proteomics analysis of stressed human islets identifies GDF15 as a target for type 1 diabetes intervention**. *Cell Metab.* (2020.0) **31** 363-374 e366. DOI: 10.1016/j.cmet.2019.12.005
43. Kempf T. **The transforming growth factor-beta superfamily member growth-differentiation factor-15 protects the heart from ischemia/reperfusion injury**. *Circ. Res.* (2006.0) **98** 351-360. DOI: 10.1161/01.RES.0000202805.73038.48
44. 44.Liu, Y. et al. Knockdown of growth differentiation factor-15 inhibited nonsmall cell lung cancer through inactivating PTEN/PI3K/AKT signaling pathway. Genes Genomics10.1007/s13258-022-01328-8 (2022).
45. Chen L, Yin Y, Liu G. **Metformin alleviates bevacizumab-induced vascular endothelial injury by up-regulating GDF15 and activating the PI3K/AKT/FOXO/PPARgamma signaling pathway**. *Ann. Transl. Med.* (2021.0) **9** 1547. DOI: 10.21037/atm-21-4764
46. Guo LL, Wang SF. **Downregulated long noncoding RNA GAS5 fails to function as decoy of CEBPB, resulting in increased GDF15 expression and rapid ovarian cancer cell proliferation**. *Cancer Biother. Radiopharm.* (2019.0) **34** 537-546. PMID: 31314588
47. Lee J, Jin YJ, Lee MS, Lee H. **Macrophage inhibitory cytokine-1 produced by melanoma cells contributes to melanoma tumor growth and metastasis in vivo by enhancing tumor vascularization**. *Melanoma Res.* (2022.0) **32** 1-10. DOI: 10.1097/CMR.0000000000000790
48. Xin H. **Sunitinib inhibition of Stat3 induces renal cell carcinoma tumor cell apoptosis and reduces immunosuppressive cells**. *Cancer Res.* (2009.0) **69** 2506-2513. DOI: 10.1158/0008-5472.CAN-08-4323
49. Guo F. **NR5A2 transcriptional activation by BRD4 promotes pancreatic cancer progression by upregulating GDF15**. *Cell Death Discov.* (2021.0) **7** 78. DOI: 10.1038/s41420-021-00462-8
|
---
title: Tissue transglutaminase exacerbates renal fibrosis via alternative activation
of monocyte-derived macrophages
authors:
- Yoshiki Shinoda
- Hideki Tatsukawa
- Atsushi Yonaga
- Ryosuke Wakita
- Taishu Takeuchi
- Tokuji Tsuji
- Miyako Tanaka
- Takayoshi Suganami
- Kiyotaka Hitomi
journal: Cell Death & Disease
year: 2023
pmcid: PMC9981766
doi: 10.1038/s41419-023-05622-5
license: CC BY 4.0
---
# Tissue transglutaminase exacerbates renal fibrosis via alternative activation of monocyte-derived macrophages
## Abstract
Macrophages are important components in modulating homeostatic and inflammatory responses and are generally categorized into two broad but distinct subsets: classical activated (M1) and alternatively activated (M2) depending on the microenvironment. Fibrosis is a chronic inflammatory disease exacerbated by M2 macrophages, although the detailed mechanism by which M2 macrophage polarization is regulated remains unclear. These polarization mechanisms have little in common between mice and humans, making it difficult to adapt research results obtained in mice to human diseases. Tissue transglutaminase (TG2) is a known marker common to mouse and human M2 macrophages and is a multifunctional enzyme responsible for crosslinking reactions. Here we sought to identify the role of TG2 in macrophage polarization and fibrosis. In IL-4-treated macrophages derived from mouse bone marrow and human monocyte cells, the expression of TG2 was increased with enhancement of M2 macrophage markers, whereas knockout or inhibitor treatment of TG2 markedly suppressed M2 macrophage polarization. In the renal fibrosis model, accumulation of M2 macrophages in fibrotic kidney was significantly reduced in TG2 knockout or inhibitor-administrated mice, along with the resolution of fibrosis. Bone marrow transplantation using TG2-knockout mice revealed that TG2 is involved in M2 polarization of infiltrating macrophages derived from circulating monocytes and exacerbates renal fibrosis. Furthermore, the suppression of renal fibrosis in TG2-knockout mice was abolished by transplantation of wild-type bone marrow or by renal subcapsular injection of IL4-treated macrophages derived from bone marrow of wild-type, but not TG2 knockout. Transcriptome analysis of downstream targets involved in M2 macrophages polarization revealed that ALOX15 expression was enhanced by TG2 activation and promoted M2 macrophage polarization. Furthermore, the increase in the abundance of ALOX15-expressing macrophages in fibrotic kidney was dramatically suppressed in TG2-knockout mice. These findings demonstrated that TG2 activity exacerbates renal fibrosis by polarization of M2 macrophages from monocytes via ALOX15.
## Introduction
Macrophages can play beneficial or detrimental roles in several diseases, depending on their activation status in the pathological tissue microenvironment [1]. Macrophages can polarize into at least two major subtypes, classically activated (M1) and alternatively activated (M2), each of which plays an important role in the opposing regulation of inflammatory progression and suppression in chronic inflammation diseases [2–4]. The surrounding environment that governs macrophage function is closely related to the specific function of macrophages: M1 is activated by lipopolysaccharide/interferon γ and exhibits proinflammatory features, whereas M2 is activated by IL-4/IL-13 stimulation and displays anti-inflammatory properties. However, if injury is uncontrolled and M2 macrophage activity persists, these cells can be detrimental to tissue homeostasis. Excessive activation of M2 macrophages regulates the continuous production of growth factors such as TGF-β, which promotes myofibroblast proliferation and activation, resulting in extracellular matrix deposition as seen in tissue fibrosis diseases [2–4].
Chronic kidney disease (CKD) affects more than 840 million people globally and is characterized by structural abnormalities and dysfunction of the kidneys that last for more than three months [5, 6]. This persistent renal damage causes excessive activation of myofibroblasts and multiple immune cells, especially macrophages, leading to tubulointerstitial fibrosis. Renal fibrosis is a common pathway for pathological deterioration from CKD to end-stage renal failure, but because the detailed pathogenesis mechanism has not been elucidated, there are currently few effective therapeutic agents. M2 macrophage activation correlates with the progression of renal fibrosis. Deletion of macrophages in a unilateral ureteral obstruction (UUO)-treated mouse tubulointerstitial fibrosis model suppressed renal fibrosis, suggesting that M2 macrophages are involved in the development of fibrosis [7, 8]. Indeed, M2 macrophages secrete high levels of TGF-β and promote epithelial-to-mesenchymal transition and subsequent tubulointerstitial fibrosis [9–12]. Furthermore, the existence of fibrosis-specific monocyte/macrophage has been reported [13], increasing the importance of analyzing the pathogenesis of fibrosis with a focus on macrophages. However, studies on M2 macrophages in mice cannot still not be directly applied to humans because mouse M2 macrophages have different basic characteristics, such as cell surface antigen markers.
Martinez et al. previously identified tissue transglutaminase (TG2) as the only reproducible M2 macrophage marker common to both humans and mice by both transcriptomic and proteomic analyses [14]. TG2 is a Ca2+-dependent protein crosslinking enzyme that catalyzes the formation of covalent bond between the γ-carboxamide groups of glutamine residues in peptide bonds and various primary amines, including the ε-amino group of lysine residues in target proteins [15–17]. TG2 expression and crosslinking activity may be associated with differentiation of monocytes and functional maturation of macrophages [18–24]. Additionally, TG2 is involved in cell adhesion and migration of monocytes [23, 25]. However, the detailed role of TG2 in the induction of M2 macrophage and the relevance to fibrosis remains unclear.
In renal fibrosis, TG2 is involved in the accumulation of fibrous proteins through crosslinking and stabilization of extracellular matrix proteins such as collagen and fibronectin, and pathogenesis of renal fibrosis is suppressed in TG2-knockout (TG2KO) and TG2 inhibitor-treated mice [26–28]. This study supports our hypothesis that TG2 promotes renal fibrosis via induction of M2 macrophage polarization infiltrated into kidney. Therefore, the current study sought to elucidate the mechanism by which TG2 regulates the polarization of M2 macrophages and the subsequent function of TG2-induced M2 macrophages in the pathogenesis of renal fibrosis and to better understand the pathogenesis of renal fibrosis, which still lacks useful therapeutic strategies.
Here, we reveal that macrophage polarization is a major contributor to the mechanism whereby TG2 induces fibrosis and that TG2 has a crucial role for the polarization of M2 macrophages of mouse and human origin. TG2-dependent M2 macrophage polarization was found to be derived from bone marrow cells and caused renal fibrosis. Furthermore, we found that TG2 is required for induction of an arachidonate lipoxygenase, leading to polarization of M2 macrophages. These studies may help develop new therapeutic targets not only for renal fibrosis, but also for diseases involving macrophage such as atherosclerosis and osteoporosis, as well as various inflammatory, neurodegenerative, and autoimmune diseases.
## Materials
Chemical reagents were mainly purchased from WAKO chemicals (Osaka, Japan) and Nacalai Tesque (Kyoto, Japan). Primary and fluorescein-conjugated secondary antibodies were listed in Suppl. Table S1. Polyclonal anti-TG2 antibody was produced in our laboratory [29]. HRP-conjugated secondary antibodies were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Cystamine was obtained from Sigma-Aldrich (St. Louis, USA). Z-DON and Boc-DON were obtained from Zedira (Darmstadt, Germany). PD146176 and 15S-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid, 15(S)-HETE, were purchased from Cayman Chemical (Ann Arbor, MI, USA).
## Ethics statement
Animal experiments were conducted at Nagoya University, complying with the national guidelines for the care and use of laboratory animal. All animal experiments were approved by the animal care and use committee of Nagoya University (No. P220002). All animal experiments were performed under anesthesia and all efforts were made to minimize suffering.
## Animal experiments
C57BL/6J male mice (8–12 weeks) were purchased from Japan SLC Inc (Shizuoka, Japan) and group-housed with food and water available ad libitum. TG2 knockout and enhanced GFP-transgenic mice were kindly provided by Dr. Robert M. Graham (Victor Chang Cardiac Research Institute, Australia) [30] and Dr. Masaru Okabe (Osaka University, Osaka, Japan) [31], respectively.
## Unilateral ureteral obstruction
The unilateral ureteral obstruction (UUO) was performed according to the method described by Shweke et al. [ 26]. Briefly, under the anesthesia with $2\%$ isoflurane, the left ureter was ligated at two separated points. Sham-operated mice had their ureter exposed but not ligated. Mice after UUO surgery were perfused with PBS to remove the blood in kidney, and pieces of the kidney were either fixed in $4\%$ paraformaldehyde for histological examination. Cystamine was orally administrated at 1.86 mg/kg/day two days before UUO surgery.
## Histological analysis
Cryosections from the kidney (5 μm) were fixed with $4\%$ paraformaldehyde and reacted with anti-TG2, F$\frac{4}{80}$, α-SMA, and ALOX15 antibodies. The specific signal was detected by the fluorescent-dye-conjugated secondary antibody. As a negative control, the primary antibody was replaced with the same amount of non-immune IgG (NI-IgG) from rabbit or rat (Sigma-Aldrich). Collagen fibers were detected using picrosirius red (Wako chemicals). Briefly, kidney sections (10 μm) were fixed in a saturated solution of picric acid with formalin and acetic acid for 15 min and then stained with $0.05\%$ sirius red reagent. In the sections from each animal, more than 5 randomly selected microscopic fields were captured by a Keyence BZ-9000 microscope. All images were quantitatively estimated for collagen fibers in picrosirius red staining within the respective kidney area according to the tutorial about “quantifying stained tissue” in image analyzer (Image J software, National Institute of Health, Bethesda, MD, USA). Each red color image was split as grayscale images and thresholded optimally. The positive areas above threshold level were measured and an average of at least 3 field from four replicates in each sample group was determined.
## Flow cytometric analysis
Kidneys were cut and digested in Hanks’ buffered saline solution (HBSS) containing 1 mg/ml collagenase (Wako) and 50 µg/ml DNase I (Roche). After filtering through a 70 µm mesh, cells were washed, incubated with the antibodies listed in Suppl. Table S1, and analyzed using Attune Acoustic Focusing Cytometer and Attune Cytometric Software v2.1.0.8626 (Life technologies).
## Quantitative real-time PCR
Total RNA was extracted from cultured cells using the Sepasol-RNA Super Reagent (Nacalai Tesque). Corresponding cDNA were prepared using ReverTra Ace qPCR RT Master Mix with gDNA Remover kit (TOYOBO, Osaka, Japan) and Real-time PCR analysis was performed using THUNDERBIRD SYBR qPCR Mix (TOYOBO) in a LightCycler 96 (Roche Diagnostics, Mannheim, Germany). Used specific primer pairs were summarized in Suppl. Table S2.
## Western blotting analysis
The cell lysates were homogenized in lysis buffer containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA, $1\%$ NP-40, 1 mM NaF, phosphatase inhibitor, and protease inhibitor cocktail (Merck Millipore). After centrifugation, supernatants were collected, and their protein concentrations were measured by Bradford assay (Bio-rad). Then, these samples were mixed with SDS-containing buffer, boiled, subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and transferred to polyvinylidene difluoride membrane (Merck Millipore). After blocking with PBS containing $5\%$ skim milk or BSA, the membrane was reacted with primary antibody listed in Suppl. Table S1, and the specific signal was detected by the peroxidase-conjugated secondary antibody and chemiluminescence reagent (Thermo Scientific, IL, USA). Each experiment was conducted in triplicate.
## Bone marrow transplantation experiments
Bone marrow transplantation experiments were performed as reported [32]. In brief, bone marrow cells obtained from donor mice were washed three times with cold PBS and injected intravenously (3 × 106 cells) into 8.5 Gy-irradiated 8-week-old male recipient mice. After 4 weeks, the substitution rate of bone marrow cells was determined by counting EGFP-positive cells in the peripheral blood, and then the mice were subjected to UUO experiments. WT and TG2KO mice were also transplanted with each bone marrow from WT and TG2KO mice, and then subjected to UUO experiments.
## Macrophage cell culture
Bone marrow-derived macrophages (BMDMs) were prepared according to the method reported previously [33]. Briefly, bone marrow cells were isolated from femur and tibia of 6–10 weeks male mice and differentiated for 6 days in RPMI medium containing $10\%$ FBS and conditioned medium from L929 fibroblasts. M2 macrophage polarization was induced by 20 ng/ml murine recombinant IL-4 (PeproTech, Rocky Hill, NJ, USA) in $5\%$ FBS containing RPMI medium after serum starvation. Human monocytic leukemia cell line (THP-1) was differentiated into macrophages by incubation with 150 nM phorbol 12-myristate-13-acetate (PMA; AdipoGen Life Sciences) for 24 h. M2 macrophage polarization was induced by 20 ng/ml recombinant human IL-4 (PeproTech) in $5\%$ FBS containing RPMI after starvation. For TG2 knockdown, targeting siRNA (sense 5′-cccugaucguugggcugaatt-3′ and antisense 5′-uucagcccaacgaucagggtt-3′) and MISSON siRNA universal negative control #1 (SIC-001) purchased from Sigma-Aldrich were used as reported previously [34].
## RNA sequence analysis
Total RNA from THP-1 cells was lysed and extracted using tissue total RNA mini kit according to manufacturer’s instructions. After the QC procedures, total RNAs were deposited for transcriptome analysis (Filgen biosciences and nanoscience, Nagoya Japan). Briefly, mRNA was enriched using oligo(dT) beads and rRNA was removed. First, the mRNA was fragmented randomly by adding fragmentation buffer, then the cDNA was synthesized by using mRNA template and random hexamers primer, after which a custom second-strand synthesis buffer (Illumina), dNTPs, RNase H, and DNA polymerase I were added to initiate the second-strand synthesis. Second, after a series of terminal repair, a ligation, and sequencing adaptor ligation, the double-stranded cDNA library was completed through size selection and PCR enrichment. Sequences were performed on NovaSeq6000 (Illumina), (6GB/PE150). Raw reads were aligned to the human genome (hg38) using the RNA-Seq Alignment App on Basespace (Illumina, CA). The data reported in this paper have been deposited in the Gene Expression Omnibus database (accession no. GSE222284).
Perseus software (version 1.6.14.0) was used to determine the genes differentially and significantly identified in four sample groups treated with vehicle, IL-4, Z-DON, and IL-4 plus Z-DON. A total of 38,552 genes with more than 50 of read counts detected between groups ($$n = 3$$) were analyzed. *Differential* gene expressions (DGEs) were determined using the threshold (FDR < 0.01). *The* gene list of both 2-fold filtered significant DGEs between vehicle- vs IL-4-treated samples and 1.5-fold filtered significant DGEs between IL-4 vs IL-4 plus Z-DON-treated samples were selected (90 genes). Among them, 55 and 35 genes with TG2-dependent increase and decrease were identified, respectively. The heat maps and hierarchical clustering were generated using Morpheus (https://clue.io/morpheus).
## Statistical analyses
Quantitative data are expressed as the means plus the standard deviation of three replicates from at least three independent experiments. The statistical significance of differences was assessed using Student’s t-test and the values of $P \leq 0.05$ were considered to indicate statistical significance. The one-way ANOVA with post hoc Tukey’s multiple comparisons test was performed with EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) [35]. No randomization blinding was used in animal experiments.
## Macrophage infiltration was reduced in the renal fibrosis model of TG2KO mice
We initially evaluated the correlation between the TG2 expression level and macrophage infiltration into the kidney in a renal fibrosis model using UUO surgery. Fluorescent immunostaining revealed that TG2 expression was markedly enhanced in the interstitial area 3 days after UUO and remained strongly increased until day 12 (Fig. 1A). The expression of macrophage marker F$\frac{4}{80}$ was highest on day 7 and partly colocalized with TG2, suggesting that TG2 is expressed in some macrophages during renal fibrosis progression (Fig. 1B). TG2KO mice had fewer F$\frac{4}{80}$-positive areas compared to WT mice at 12 days after UUO, and a similar tendency was observed at 7 days after UUO, but not significantly different (Fig. 1C). These results suggest that TG2 plays an important role in macrophage infiltration and accumulation at the relatively late stages of renal fibrosis (>day 7) and may account for the increase in M2 macrophages in the late response rather than affecting M1 macrophages in the early response. Fig. 1Evaluation of macrophage infiltration into the kidney after UUO surgery in TG2KO mice. Mice were conducted to UUO surgery. Kidney sections from WT and TG2KO mice after UUO surgery were fixed in $4\%$ paraformaldehyde and immunostained using anti-F$\frac{4}{80}$ plus Alexa Fluor 594 anti-rat antibodies and anti-TG2 plus Alexa Fluor 488 anti-rabbit antibodies (A). The nuclei were counterstained with DAPI. Scale bar = 100 μm. The merged image from WT on day 7 after UUO was enlarged (B). Arrowheads indicate the similar distributions between F$\frac{4}{80}$ and TG2. Scale bar = 50 μm. The percentages of F$\frac{4}{80}$-positive area are presented (C). Data are presented as the mean ± SD ($$n = 3$$) (*$P \leq 0.05$, Student’s t test).
## TG2 is required for M2 macrophages infiltrating fibrotic kidney
We next used flow cytometry to investigate the infiltrated macrophage subtypes that decreased in abundance in the kidney of TG2KO mice after UUO. CD45-positive cells were gated from whole kidney cells (Fig. 2A) and then evaluated for CD11b and F$\frac{4}{80}$ expressions to yield two main population groups (Fig. 2B). CD45-positive CD11b+ F$\frac{4}{80}$low (R1) and CD11b+ F$\frac{4}{80}$hi (R2) cells were classically defined as M1 and M2 macrophages infiltrating the UUO-treated kidney, respectively (Fig. 2B) [36]. This is consistent with flow cytometry analysis of CD206+ cells that were backgated to the magenta-colored R2 group in Fig. 2C, D. The number of cells in the R2 group was significantly decreased in TG2KO mice comparison to WT mice, although that cell number in the R1 group was no significant different (Fig. 2B, E). Furthermore, when CD45+ cells were assessed for CD11b and Ly6C expressions, three population groups, CD45+ CD11b+ Ly6Chi (R3), Ly6Cint (R4), and Ly6Clow (R5), were observed (Fig. 2F). CD206-positive M2 macrophages were classified in both the R4 and R5 groups but not in the R3 group (Fig. 2G). The comparison of each macrophage subtype revealed that the abundance of cells in the R4 group gated by both F$\frac{4}{80}$hi and CD206+ was significantly decreased in TG2KO mice (Fig. 2H), suggesting that TG2 was involved in the polarization of the CD45+ CD11b+ F$\frac{4}{80}$hi CD206+ Ly6Cint M2 macrophage subtype during renal fibrosis. Consistent with these results, the number of α-SMA-positive myofibroblasts and the level of collagen deposition detected by picrosirius red staining were also significantly suppressed in TG2KO mice (Fig. 2I, J).Fig. 2Characterization of infiltrated macrophage into fibrotic kidney after UUO surgery. CD45-positive cells from fibrotic kidney in WT and TG2KO mice were divided (A) into the CD11b+ F$\frac{4}{80}$low (R1) and CD11b+ F$\frac{4}{80}$hi (R2) groups (B). SSC side scatter. CD206-positive cells from WT (black) and TG2KO (gray) mice were selected in histogram plot (C) and colored with magenta in the dot plots shown in B (D). The relative cell counts of two groups were indicated (E). CD45-positive cells were also divided into the CD11b+ Ly6Chi (R3), CD11b+ Ly6Cint (R4), and CD11b+ Ly6Clow (R5) groups (F). CD206-positive cells were colored with magenta in the dot plots shown in F (G). The relative cell counts of R4 and R5 groups gated by both F$\frac{4}{80}$hi and CD206+ were indicated (H). Myofibroblasts in kidney sections were stained using anti-α-SMA antibody plus AlexaFluor 488 goat anti-rabbit IgG and the percentages of their positive area are presented (I). The nuclei were counterstained with DAPI. The collagen fibers in kidney sections were detected by picrosirius red staining and the percentages of their positive area are presented (J). Scale bars = 100 μm. Representative results in at least three independent samples were shown. (* $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ by one-way ANOVA with post hoc Tukey’s multiple comparisons test).
## TG2 expression in bone marrow cells contributes to renal fibrosis after UUO
Monocyte-derived and resident macrophages have been reported to be involved in CKD [37, 38], and therefore we next performed bone marrow transplantation experiments to distinguish immune cell origin and confirm the correlation with TG2 expression. Mice were exposed to a lethal dosage of X-rays and then transplanted with bone marrow cells from GFP-transgenic mice via tail vein injection. After 4 weeks of recovery, UUO was performed on the mice, and cell counts analyzed at each indicated time after surgery (Fig. 3A). The percentage of GFP-positive cells in peripheral blood cells indicated the successful replacement in this transplantation model. This showed that >$98\%$ of CD45+ cells in blood cells were GFP+-cells derived from donor GFP-transgenic mice (Suppl. Fig. S1). We next examined whether the increased M2 macrophages in fibrotic kidney expressed GFP. The M2 macrophage population (CD45+ CD11b+ F$\frac{4}{80}$hi; R2) was confirmed by CD206 expression and almost exclusively expressed GFP (Fig. 3B). The abundance of these GFP-positive R2 populations markedly increased on days 7 and 12 after UUO compared with that on day 3 and were mostly derived from bone marrow cells (Fig. 3C). We next evaluated the levels of renal fibrosis using four experimental groups of bone marrow-chimeric mice transplanted in WT and TG2KO mice as indicated in Fig. 3D. WT and TG2KO mice were exposed to X-rays and transplanted with bone marrow cells from WT or TG2KO mice. In WT-recipient mice, mice harboring TG2KO bone marrow cells had significantly decreased numbers of myofibroblasts and the levels of renal fibrosis compared with levels in mice harboring WT bone marrow cells (Fig. 3E, G, upper column; Fig. 3F, H, lanes 3 vs. 4). Conversely, in TG2KO-recipient mice, mice harboring WT bone marrow cells had significantly increased numbers of myofibroblasts and the levels of renal fibrosis compared with levels in mice harboring TG2KO bone marrow cells (Fig. 3E, G, lower column; Fig. 3F, H, Lanes 5 vs. 6). The mice transplanted with TG2KO bone marrow cells had markedly suppressed fibrosis regardless of whether the recipient was WT or TG2KO (Fig. 3F, H, lanes 4 and 5), indicating that TG2-dependent induction of bone marrow-derived CD11b+ F$\frac{4}{80}$hi CD206+ M2 macrophage subtypes may promote the pathogenesis of renal fibrosis. Fig. 3Role of TG2 in bone marrow-derived cells in renal fibrosis after UUO surgery. Mice were irradiated at lethal dose (8.5 Gy) of X-rays and transplanted with bone marrow cells isolated from GFP-transgenic mice by tail vein injection. After 4 weeks recovery period, mice were subjected to UUO surgery and analyzed on indicated days (A). The CD45-positive cells in fibrotic kidney were divided into F$\frac{4}{80}$ and CD11b, and these population classified by CD206 were colored with magenta (B). Then, the counts and percentages of GFP-positive cells in R2 group were analyzed and plotted (C). WT and TG2KO mice were lethally irradiated by X-rays and transplanted with bone marrow cells isolated from WT and TG2KO mice by tail vein injection (D). After 4 weeks recovery period, mice were conducted to UUO and analyzed on 14 days after UUO surgery. The myofibroblasts and collagen fibers in kidney sections were detected by immunofluorescence staining using anti-α-SMA antibody (E) and picrosirius red staining (G), respectively, and the percentages of their positive area are presented (F, H). The nuclei were counterstained with DAPI. Scale bars = 100 μm. Representative results in at least three independent samples were shown. (*** $P \leq 0.001$ by one-way ANOVA with post hoc Tukey’s multiple comparisons test).
## TG2 activity contributes to IL-4-induced M2 macrophage polarization and exacerbates renal fibrosis
Based on the above results, we speculated that TG2 expression in macrophages may contribute to polarization of M2 macrophages. In in vitro studies using bone marrow-derived macrophages (BMDMs), IL-4 treatment markedly increased in mRNA expression of TG2 and mouse M2 macrophage markers such as CD206, arginase (Arg)-1, transferrin receptor (TFR; Fig. 4A, B), whereas treatment with TG2 inhibitor such as cystamine and Z-DON significantly suppressed the induction of these M2 markers (Fig. 4A, B). Similar results were obtained in the protein levels of TG2 and M2 macrophage markers (Fig. 4C, D and Suppl. Figs. S2 and S3). Then, we next examined whether TG2-dependent M2 polarization in BMDMs was involved in the renal fibrosis. IL-4-treated BMDMs from WT or TG2KO mice were injected into renal subcapsule of TG2KO mice after UUO (Fig. 4E). Interestingly, the number of myofibroblasts and the levels of renal fibrosis in TG2KO mice were significantly increased by injection of BMDMs from WT mice, although there was no effect on renal fibrosis following injection of BMDMs from TG2KO mice (Fig. 4F, G). This was similar to the more severe renal fibrosis effect produced with TG2KO-recipient mice harboring WT bone marrow cells than that in mice harboring TG2KO bone marrow cells (Fig. 3E, G, lower column; Fig. 3F, H, Lane 5 vs. 6).Fig. 4Role of TG2 in M2 macrophage polarization induced by IL-4 and renal fibrosis. BMDMs were prepared using bone marrow cells isolated from mice and cultured with L929 fibroblast conditioned medium. M2 macrophage polarization was induced by treatment of 20 ng/ml recombinant mouse IL-4 in the presence or absence of 0.4 mM cystamine and 50 μM Z-DON for 24–48 h. mRNA expression levels of TG2 and indicated mouse M2 macrophage markers were analyzed (A, B). Data were normalized against mRNA expression of TATA-binding protein (TBP) and relative values (a ratio of the control sample) were presented as the mean ± SD ($$n = 3$$) (**$P \leq 0.01$, *$P \leq 0.05$, Student’s t test). Veh Vehicle, Cys cystamine. The protein levels of these samples were analyzed by immunoblotting using the indicated antibodies (C, D). Anti-β-actin antibody was used as a loading control for each sample. Total intensities of all the bands in each sample were presented after normalizing the results to the expression levels in β-actin. The full-length blots with molecular mass markers are presented in Suppl. Figs. S2 and S3. BMDMs prepared from WT or TG2KO mice were treated by IL-4 for 2 h and transferred into renal subcapsule (4.75 × 105 cells/mouse) of TG2KO mice on day 9 after UUO. These mice were sacrificed on day 12 after UUO and analyzed (E). The myofibroblasts and collagen fibers in kidney sections were detected by immunofluorescence staining using anti-α-SMA antibody and picrosirius red staining, respectively (F), and the percentages of their positive area are presented (G). The nuclei were counterstained with DAPI. Scale bars = 100 μm. Representative results in at least three independent samples were shown (***$P \leq 0.001$ by one-way ANOVA with post hoc Tukey’s multiple comparisons test).
## Cystamine suppressed the M2 macrophage infiltrating fibrotic kidney
We next investigated whether the inhibition of TG2 activity could suppressed M2 macrophage polarization in an in vivo study. Oral administration of cystamine significantly decreased the abundance of M2 macrophages (CD45+ CD11b+ F$\frac{4}{80}$hi; R2) but not that of M1 macrophages (CD45+ CD11b+ F$\frac{4}{80}$low; R1; Fig. 5A–C). Cystamine administration also significantly decreased the abundance of CD11b+ CD206+ Ly6Cint M2 macrophages (R4), but not that of CD11b+ CD206+ Ly6Clow M2 macrophages (R5) (Fig. 5D). The comparison of each macrophage subtype revealed that the abundance of the R4 group gated by both F$\frac{4}{80}$hi and CD206+ was significantly decreased in cystamine-treated mice after UUO (Fig. 2E), suggesting that cystamine inhibits the polarization of CD45+ CD11b+ F$\frac{4}{80}$hi CD206+ Ly6Cint M2 macrophage subtype during renal fibrosis. Consistently, the number of myofibroblasts and the levels of renal fibrosis were also significantly suppressed by cystamine-treated mice (Fig. 5F, G).Fig. 5Effect of cystamine administration in M2 macrophage polarization and renal fibrosis. Mice were conducted to UUO surgery and orally administrated with cystamine (1.86 mg/kg/day). Renal CD45-positive cells were classified by CD206 (A) and colored with magenta in the dot plots divided into the CD11b+ F$\frac{4}{80}$low (R1) and CD11b+ F$\frac{4}{80}$hi (R2) groups (B). The relative cell counts of R1 and R2 groups were indicated (C). These cells also divided into CD11b+ Ly6Chi (R3), CD11b+ Ly6Cint (R4), and CD11b+ Ly6Clow (R5) groups (D). The relative cell counts of R4 and R5 groups gated by F$\frac{4}{80}$hi and CD206+ were indicated (E). The myofibroblasts and collagen fibers in kidney sections were detected by immunofluorescence staining using anti-α-SMA antibody (F) and picrosirius red staining (G), respectively, and the percentages of their positive area are presented in the graph on the right. The nuclei were counterstained with DAPI. Scale bars = 100 μm. Representative results in at least three independent samples were shown. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ by one-way ANOVA with post hoc Tukey’s multiple comparisons test.
## TG2 promotes the polarization of human M2 macrophages via intracellular crosslinking activity
TG2 was identified as the only marker common to mouse and human M2 macrophages based on both transcriptomics and proteomics [14], and we therefore, next examined the contribution of TG2 expression and activity in human M2 macrophages. The human monocyte leukemia cell line, THP-1, was differentiated into macrophages using PMA and then polarized into M2 macrophages by IL-4 treatment. Similar to the results using mouse BMDMs, IL-4 treatment markedly increased in mRNA expression of TG2 and human M2 macrophage markers such as C-C motif chemokine ligand 22 (CCL22), peroxisome proliferator-activated receptor γ (PPARγ), and CD209 (Fig. 6). The induction of these M2 markers were significantly inhibited by treatment with TG2 siRNAs (Fig. 6A and Suppl. Fig. S4) and cell-permeable inhibitors such as cystamine (Fig. 6B) and Z-DON (Fig. 6C), but not by the cell-impermeable inhibitor Boc-DON (Fig. 6D). These results suggest that the crosslinking activity of intracellular TG2 in macrophages plays a common role in promoting polarization of both human and mouse M2 macrophages. Fig. 6Role of TG2 in human M2 macrophage polarization induced by IL-4.Human monocytic leukemia cell line, THP-1, was treated with 150 nM PMA for 24 h. M2 macrophage polarization was induced by 20 ng/ml recombinant human IL-4. mRNA expression levels of TG2 and indicated human M2 macrophage markers were analyzed. For knockdown experiment, PMA-treated macrophages were transfected with siRNA against TG2 (A). As a negative control, scrambled siRNA (Control) was replaced with the same amount of TG2 siRNA. For the TG2 inactivation, cell-permeable (0.4 mM Cystamine, B; 50 μM Z-DON, C) and impermeable inhibitors (100 μM Boc-DON; D) were used. Data were normalized against mRNA expression of TBP and relative values were presented as the mean ± SD ($$n = 3$$) (***$P \leq 0.001$, **$P \leq 0.01$, *$P \leq 0.05$, Student’s t test).
## TG2 regulates M2 macrophage polarization via ALOX15 induction
To elucidate the reason why the TG2 knockdown and inactivation suppressed M2 macrophage polarization, we next investigated the molecular mechanism underlying this. We first checked the signal transducer and activator of transcription 6 (STAT6), upstream regulator in the signaling pathway induced by IL-4. IL4-treated human macrophages had significantly increased levels of phosphorylated STAT6 but were not affected by combined treatment with Z-DON (Fig. 7A and Suppl. Fig. S5), suggesting that STAT6 is phosphorylated upstream of TG2. We next performed global expression analysis via RNA sequencing to explore genes whose expression was regulated by TG2 activity. We selected 55 genes whose expression was altered by IL-4 treatment in a TG2-dependent manner (i.e., both significantly increased by more than 2-fold in IL-4 treatment and decreased by more than one-third by treatment in combination with Z-DON) (Fig. 7B). Expression of the arachidonate 15-lipoxygenase (ALOX15) gene was remarkably increased by IL-4 treatment (Fig. 7C) and was reduced by $40\%$ by inhibition of TG2 activity (data not shown). As ALOX15 has been reported as a regulator of M2 macrophage polarization [39, 40], we next confirmed whether the ALOX15 expression was regulated by TG2 expression and activity and promoted TG2-dependent M2 macrophage polarization. The mRNA expression levels of ALOX15 were decreased in the IL-4-treated macrophages in combination with TG2 siRNA, cystamine, or Z-DON, but not with Boc-DON treatment (Fig. 7D and Suppl. Fig. S4). In addition, treatment with an inhibitor of ALOX15 activity, PD146176, decreased the mRNA expression of human M2 macrophage markers (Fig. 7E). Furthermore, treatment with the ALOX15 metabolite, 15(S)-HETE, significantly increased mRNA expression of M2 macrophage markers including CD36 in IL-4-treated macrophages (Fig. 7F). These results suggested that TG2 promoted M2 macrophage polarization via the expression and activity of ALOX15.Fig. 7Analysis of molecular mechanism of M2 macrophage polarization that is regulated by TG2.Human macrophages differentiated from THP-1 were treated with 20 ng/ml IL-4 in the presence or absence of 50 μM Z-DON. Cells were extracted and analyzed by immunoblotting using the indicated antibodies (A). The full-length blots are presented in Suppl. Fig. S5. Total RNAs were conducted to transcriptome analysis as described in “Materials and methods” section. DGEs between vehicle- vs IL-4-treated samples (2-fold, FDR < 0.01) and between IL-4- vs IL-4 plus Z-DON samples (1.5-fold, FDR < 0.01) were determined. Among them, 55 genes that increase in TG2-dependent manner were identified and indicated as heat map (B). The relative means of the three groups of mRNAs whose expression was enhanced by IL-4 treatment among the 55 genes are plotted in ascending order (C). mRNA expression levels of ALOX15 were evaluated in the IL-4-treated human macrophages transfected with control and TG2 siRNA and treated with cystamine, Z-DON, and Boc-DON (D). mRNA expression levels of TG2 and indicated M2 macrophage markers were analyzed in IL-4-treated human macrophage combined treatment with 5 μM PD146176 (E) or 24 μM 15(S)-HETE (F). Data were normalized against mRNA expression of TBP and relative value were presented as the mean ± SD ($$n = 3$$) (***$P \leq 0.001$, **$P \leq 0.01$, *$P \leq 0.05$, Student’s t test).
## ALOX15 expression was enhanced in macrophages infiltrating fibrotic kidney
We finally evaluated whether ALOX15 expression was observed in macrophages infiltrating fibrotic kidney after UUO. Fluorescent immunostaining revealed that ALOX15 expression was enhanced in the F$\frac{4}{80}$+ macrophages in fibrotic kidney (Fig. 8A). Unexpectedly, expression of ALOX15 occurred not only in macrophages but also in renal tubules and glomeruli. Surprisingly, the number of ALOX15-positive F$\frac{4}{80}$+ macrophages was drastically reduced by approximately $90\%$ in TG2KO mice (Fig. 8B, left panel) although the overall number of infiltrating F$\frac{4}{80}$+ macrophages was only reduced by approximately $40\%$ in TG2KO mice as in Fig. 1C (Fig. 8B, right panel). Given that mice deficient in ALOX15 or treated with an ALOX15 inhibitor significantly suppressed the renal fibrosis by UUO [41], these results suggest that ALOX15 expression is enhanced by intracellular TG2 activity and induces M2 macrophage polarization, leading to pathological exacerbation of renal fibrosis. Fig. 8Evaluation of ALOX15 distribution in fibrotic kidney and its association with renal fibrosis. Kidney section on day 12 after UUO surgery were fixed in $4\%$ paraformaldehyde and immunostained using anti-F$\frac{4}{80}$ plus Alexa Fluor 594 anti-rat antibodies and anti-ALOX15 plus Alexa Fluor 488 anti-rabbit antibodies (A). As an isotype control, each primary antibody was replaced with the same amount of rat and rabbit NI-IgG. The nuclei were counterstained with DAPI. Arrowheads indicate the similar distributions between ALOX15 and F$\frac{4}{80.}$ Scale bar = 50 μm. Representative results in at least three independent samples were shown. The numbers of F$\frac{4}{80}$+ ALOX15+ cells and F$\frac{4}{80}$+ cells infiltrating kidney of WT and TG2KO mice after UUO were presented (B). Data are presented as the mean ± SD ($$n = 3$$) (**$P \leq 0.01$, *$P \leq 0.05$, Student’s t test). Schematic showing the molecular mechanism by which TG2 causes M2 polarization via ALOX15 expression and its metabolism in mouse and human macrophages, leading to renal fibrosis (C).
## Discussion
In this study, we found that TG2 promotes both human and mouse M2 macrophage polarization via its intracellular crosslinking activity. In addition, the monocyte-derived TG2-induced M2 macrophages contribute to the pathogenesis of mouse renal fibrosis. Furthermore, we found that, at least in part by mechanism, TG2 markedly exacerbates renal fibrosis though the enhanced expression of ALOX15 and a metabolite derived from ALOX15 activity. First, we found that in TG2KO mice, there was only a significant decrease in the number of macrophages in the late stage of fibrotic kidney after UUO surgery (Fig. 1). This was interesting as TG2 had been previously thought to be mainly involved in the stabilization and accumulation of fibrous protein through crosslinking activity in renal fibrosis [42, 43], suggesting a new molecular mechanism whereby TG2 is involved in renal fibrosis.
The increased abundance of macrophages in fibrotic kidney was mainly derived from bone marrow but not from proliferative renal resident macrophages (Fig. 3C). The majority of macrophages are distributed throughout the body before birth and are capable of self-renewal and there is consequently little need for monocytes in adults under normal circumstances [44]. However, during a rapid response to inflammation, circulating monocytes must adhere to and infiltrate vascular endothelial cells adjacent to the injured tissue and then differentiate into macrophages [45]. TG2 is present in monocytes [20] and has been reported to contribute to adhesion and migration of monocytes on fibronectin [23]. Our results clearly indicated that TG2 expression in monocytes derived from bone marrow cells increased the abundance of M2 macrophage in fibrotic kidney, exacerbating renal fibrosis (Figs. 3D–H and 4E–G). Furthermore, both F$\frac{4}{80}$ and ALOX15 positive cells in fibrotic kidney were reduced by $90\%$ in TG2KO mice although F$\frac{4}{80}$ positive cells were reduced by about half (Fig. 8). These suggested that loss of TG2 was associated with polarization of M2 macrophages rather than macrophage infiltration (Fig. 8B). However, it cannot be completely ruled out that TG2 is involved in the monocyte infiltration into the fibrotic kidney. This is because fibrosis was more severe in the renal subcapsular injection of BMDMs from WT mice into TG2KO mice than in the transplantation of bone marrow from WT mice into TG2KO mice, whereas the renal subcapsular injection itself may have contributed to worsening fibrosis. Similar to the results obtained in TG2KO mice, inhibitors for TG2 crosslinking activity predominantly reduced the amount of M2 macrophage in fibrotic kidney and suppressed renal fibrosis, despite the limitation of in vivo experiments that cystamine is not a specific inhibitor of TG2 activity alone (Fig. 5). In addition, in vitro studies using both BMDMs and PMA-treated THP-1 demonstrated that TG2 is a critical regulator for human and mouse M2 macrophage polarization (Figs. 4 and 6, and Suppl. Fig. S4). Although TG2 extracellularly contributes to activation of TGF-β [26, 46–48], which is a major regulator for cell differentiation, it was interesting to note that only intracellular, but not extracellular, TG2 regulated M2 macrophage polarization. Active inhibitors targeting TG2 are expected to contribute to the development of useful drugs for the regulation of pathogenesis associated with M2 macrophages [49].
Transcriptome analysis further revealed that ALOX15 expression was strongly induced by IL-4 in a TG2 activity dependent manner. Although ALOX15 was not identified as a gene/protein commonly induced in both mouse and human M2 macrophages [14], but has been reported as a highly inducible IL-4/IL-13 target gene in both mouse [40] and human [39]. ALOX15 belongs to a family of dioxygenases that convert unsaturated fatty acids, preferably arachidonic acid, into monoxide derivatives such as 15(S)-HETE [50, 51]. ALOX15 is implicated in many pathological processes and was recently reported to induce inflammation and fibrogenesis in mice after UUO [41]. Macrophage infiltration and renal fibrosis were demonstrated to decrease in both ALOX15-knockout and inhibitor-treated mice but were increased in transgenic ALOX15-overexpressing mice [41], and although that study did not include an in vitro analysis and the results from human macrophages, these results are consistent with our results obtained here. Additionally, studies of mice undergoing remnant nephrectomy, diabetic nephropathy, and sepsis-induced acute kidney injury had a similar profibrotic role of ALOX15 [52–54]. Consistent with these studies, the ALOX15 expression levels were found to be significantly higher in the kidneys of patients with advanced diabetic nephropathy, one of the main complications of diabetes [55]. This particular study focused on the glomeruli in patients, where ALOX15 levels were elevated in all intrinsic cells of glomerulus and contributed to ferroptosis induction mainly through the ALOX15-mediated lipid metabolism pathway. Compared to the consistent in vivo results in mice, the role of ALOX15 in vitro studies in mouse macrophages is a bit more complicated. Indeed, we have not been able to detect both mRNA and protein expressions of ALOX15 in BMDMs. This result is consistent with previous reports [56, 57] and may be one reason why ALOX15 was not determined as a common M2 macrophage marker for both mice and humans in the Martinez’s report [14]. Since ALOX15 expression in macrophage is detected in vivo (Fig. 8), a possible problem is that in vitro experiments using BMDMs do not adequately reflect the in vivo situation of renal fibrosis. The resident (peritoneal) macrophages have been reported to have a larger population of ALOX15-positive cells compared to BMDMs [58], suggesting that TG2 in BMDMs may involves in not only its own M2 polarization but also that of resident macrophages during renal fibrosis.
In this study, we found that TG2 promoted M2 macrophage polarization via induction of ALOX15 expression, although the regulatory mechanism of TG2 and ALOX15 expression remains to be clarified. Phenformin, a biguanide antidiabetic drug, was reported to prevent IL-4-induced M2 macrophage polarization through decreased STAT6 association and Lys-9 acetylation of histone H3 at the ALOX15 promoter [59]. Since STAT6 is also implicated to associate at the TG2 promoter [60], we demonstrated that an inhibitor of STAT6 phosphorylation predominantly suppressed TG2 expression (Suppl. Fig. S6), whereas TG2 inhibitors did not interfere the phosphorylation and expression of STAT6 (Fig. 7A). Mitogen-activated protein kinase kinase (MEK) is required for M2 macrophage polarization by promoting PPARγ-induced retinoic acid signaling [49]. PPARγ and retinoic acid signaling have also been reported as upstream regulators of TG2 [61] although PPARγ is also reported to be inactivated via crosslinking by TG2 [62, 63]. Moreover, in IL-4-induced M2 macrophages, ALOX15 generates an endogenous ligand for the transcription factor PPARγ, thereby suppressing inflammatory responses [40]. Knockdown and inactivation of TG2 reduced the expression of ALOX15 and vice versa, suggesting that TG2 and ALOX15 expressions were synergistically upregulated for M2 macrophage polarization. In addition to MEK, histone deacetylase activity is also critical for M2 macrophage polarization [49]. Demethylation of histone H3 trimethyl-lysine 27 (H3K27me3) at the promoter is required for IL-4-mediated ALOX15 induction [64]. Furthermore, TG2 serotonylates the glutamine at position 5 (Q5ser) on nucleosomes marked with tri-methylated lysine 4 of histone H3 (H3K4me3) and is responsible for enhanced recruitment of transcription factor, TFIID [65]. Since we detected the increased level of nuclear TG2 in IL-4-treated human macrophage (Suppl. Fig. 7), TG2-mediated serotonylation and regulations of enzyme involved in methylation/demethylation of histone H3 may contribute to the various gene expressions. In addition, we attempted to detect the direct molecular target of TG2 in IL-4-treated human macrophage based on the incorporation of a biotin-pentylamine (BPA) probe into proteins via crosslinking activity. However, the crosslinking activity is weaker than the background signal (due to nonspecific adsorption of streptavidin-peroxidase and the several endogenous biotin-binding proteins) and cannot be detected (data not shown). Based on these results, we speculated that it may be difficult to detect TG2 activity using BPA in cultured macrophages because BPA does not penetrate human macrophage cell membranes, is quickly excreted, or the BPA-incorporated proteins are rapidly degraded. However, the experiments using cell lysates of IL-4-treated human macrophages showed an increase in BPA-incorporated proteins (Suppl. Fig. 8). The detailed relationship between TG2 and ALOX15 needs to be clarified in future studies.
In summary, we demonstrated that a TG2-dependent M2 macrophage polarization mechanism was commonly induced in both mouse and human cells and was involved in renal fibrosis in cellular and animal models. Furthermore, we found that ALOX15 was an important factor acting in the downstream of intracellular TG2 activity in the polarization of M2 macrophages and exacerbated renal fibrosis (Fig. 8C). Our findings provide information about novel pathological mechanisms and new therapeutic targets associated with renal fibrosis that could be widely adopted to several research fields related to TG2 and macrophage polarization.
## Supplementary information
Supplemental materials and original data files Supplementary Tables The online version contains supplementary material available at 10.1038/s41419-023-05622-5.
## References
1. Gordon S, Taylor PR. **Monocyte and macrophage heterogeneity**. *Nat Rev Immunol* (2005.0) **5** 953-64. DOI: 10.1038/nri1733
2. Martinez FO, Sica A, Mantovani A, Locati M. **Macrophage activation and polarization**. *Front Biosci* (2008.0) **13** 453-61. DOI: 10.2741/2692
3. Mills CD. **M1 and M2 macrophages: oracles of health and disease**. *Crit Rev Immunol* (2012.0) **32** 463-88. DOI: 10.1615/CritRevImmunol.v32.i6.10
4. 4.Braga TT, Agudelo JSH, Camara NOS. Macrophages during the fibrotic process: M2 as friend and foe. Front Immunol. 2015;6:602.
5. Jager KJ, Kovesdy C, Langham R, Rosenberg M, Jha V, Zoccali C. **A single number for advocacy and communication-worldwide more than 850 million individuals have kidney diseases**. *Kidney Int* (2019.0) **96** 1048-50. DOI: 10.1016/j.kint.2019.07.012
6. Cheung AK, Chang TI, Cushman WC, Furth SL, Hou FF, Ix JH. **KDIGO 2021 Clinical Practice Guideline for the management of blood pressure in chronic kidney disease**. *Kidney Int* (2021.0) **99** S1-87. DOI: 10.1016/j.kint.2020.11.003
7. Shen B, Liu X, Fan Y, Qiu J. **Macrophages regulate renal fibrosis through modulating TGFβ superfamily signaling**. *Inflammation* (2014.0) **37** 2076-84. DOI: 10.1007/s10753-014-9941-y
8. Kitamoto K, Machida Y, Uchida J, Izumi Y, Shiota M, Nakao T. **Effects of liposome clodronate on renal leukocyte populations and renal fibrosis in murine obstructive nephropathy**. *J Pharm Sci* (2009.0) **111** 285-92. DOI: 10.1254/jphs.09227FP
9. Yeh YC, Wei WC, Wang YK, Lin SC, Sung JM, Tang MJ. **Transforming growth factor-{beta}1 induces Smad3-dependent {beta}1 integrin gene expression in epithelial-to-mesenchymal transition during chronic tubulointerstitial fibrosis**. *Am J Pathol* (2010.0) **177** 1743-54. DOI: 10.2353/ajpath.2010.091183
10. Wang S, Meng X-M, Ng Y-Y, Ma FY, Zhou S, Zhang Y. **TGF-β/Smad3 signalling regulates the transition of bone marrowderived macrophages into myofibroblasts during tissue fibrosis**. *Oncotarget* (2015.0) **7** 8809-22. DOI: 10.18632/oncotarget.6604
11. Meng XM, Wang S, Huang XR, Yang C, Xiao J, Zhang Y. **Inflammatory macrophages can transdifferentiate into myofibroblasts during renal fibrosis**. *Cell Death Dis* (2016.0) **7** e2495-e2495. DOI: 10.1038/cddis.2016.402
12. Wang YY, Jiang H, Pan J, Huang XR, Wang YC, Huang HF. **Macrophage-to-myofibroblast transition contributes to interstitial fibrosis in chronic renal allograft injury**. *J Am Soc Nephrol* (2017.0) **28** 2053-67. DOI: 10.1681/ASN.2016050573
13. Satoh T, Nakagawa K, Sugihara F, Kuwahara R, Ashihara M, Yamane F. **Identification of an atypical monocyte and committed progenitor involved in fibrosis**. *Nature* (2017.0) **541** 96-101. DOI: 10.1038/nature20611
14. Martinez FO, Helming L, Milde R, Varin A, Melgert BN, Draijer C. **Genetic programs expressed in resting and IL-4 alternatively activated mouse and human macrophages: similarities and differences**. *Blood* (2013.0) **121** e57-69. DOI: 10.1182/blood-2012-06-436212
15. Chen JSK, Mehta K. **Tissue transglutaminase: an enzyme with a split personality**. *Int J Biochem Cell Biol* (1999.0) **31** 817-36. DOI: 10.1016/S1357-2725(99)00045-X
16. Iismaa SE, Mearns BM, Lorand L, Graham RM. **Transglutaminases and disease: lessons from genetically engineered mouse models and inherited disorders**. *Physiol Rev* (2009.0) **89** 991-1023. DOI: 10.1152/physrev.00044.2008
17. Eckert RL, Kaartinen MT, Nurminskaya M, Belkin AM, Colak G, Johnson GVW. **Transglutaminase regulation of cell function**. *Physiol Rev* (2014.0) **94** 383-417. DOI: 10.1152/physrev.00019.2013
18. Kannagi R, Teshigawara K, Noro N, Masuda T. **Transglutaminase activity during the differentiation of macrophages**. *Biochem Biophys Res Commun* (1982.0) **105** 164-71. DOI: 10.1016/S0006-291X(82)80026-0
19. Murtaugh MP, Mehta K, Johnson J, Myers M, Juliano RL, Davies PJ. **Induction of tissue transglutaminase in mouse peritoneal macrophages**. *J Biol Chem* (1983.0) **258** 11074-81. DOI: 10.1016/S0021-9258(17)44387-0
20. Murtaugh MP, Arend WP, Davies PJA. **Induction of tissue transglutaminase in human peripheral blood monocytes**. *J Exp Med* (1984.0) **159** 114-25. DOI: 10.1084/jem.159.1.114
21. Sun H, Kaartinen MT. **Transglutaminases in monocytes and macrophages**. *Med Sci* (2018.0) **6** 115
22. Tatsukawa H, Hitomi K. **Role of transglutaminase 2 in cell death, survival, and fibrosis**. *Cells* (2021.0) **10** 1842. DOI: 10.3390/cells10071842
23. Akimov SS, Belkin AM. **Cell surface tissue transglutaminase is involved in adhesion and migration of monocytic cells on fibronectin**. *Blood* (2001.0) **98** 1567-76. DOI: 10.1182/blood.V98.5.1567
24. Szondy Z, Sarang Z, Molnar P, Nemeth T, Piacentini M, Mastroberardino PG. **Transglutaminase 2-/- mice reveal a phagocytosis-associated crosstalk between macrophages and apoptotic cells**. *Proc Natl Acad Sci USA* (2003.0) **100** 7812-7. DOI: 10.1073/pnas.0832466100
25. Akimov SS, Krylov D, Fleischman LF, Belkin AM. **Tissue transglutaminase is an integrin-binding adhesion coreceptor for fibronectin**. *J Cell Biol* (2000.0) **148** 825-38. DOI: 10.1083/jcb.148.4.825
26. Shweke N, Boulos N, Jouanneau C, Vandermeersch S, Melino G, Dussaule J-C. **Tissue transglutaminase contributes to interstitial renal fibrosis by favoring accumulation of fibrillar collagen through TGF-β activation and cell infiltration**. *Am J Pathol* (2008.0) **173** 631-42. DOI: 10.2353/ajpath.2008.080025
27. Johnson TS, Fisher M, Haylor JL, Hau Z, Skill NJ, Jones R. **Transglutaminase inhibition reduces fibrosis and preserves function in experimental chronic kidney disease**. *J Am Soc Nephrol* (2007.0) **18** 3078-88. DOI: 10.1681/ASN.2006070690
28. Badarau E, Wang Z, Rathbone DL, Costanzi A, Thibault T, Murdoch CE. **Development of potent and selective tissue transglutaminase inhibitors: their effect on TG2 function and application in pathological conditions**. *Chem Biol* (2015.0) **22** 1347-61. DOI: 10.1016/j.chembiol.2015.08.013
29. Tatsukawa H, Abe N, Ohashi S, Hitomi K. **Distribution of transglutaminase family members in mouse whole body sections**. *Biochem Biophys Res Commun* (2015.0) **467** 1046-51. DOI: 10.1016/j.bbrc.2015.10.001
30. Nanda N, Iismaa SE, Owens WA, Husain A, Mackay F, Graham RM. **Targeted inactivation of Gh/tissue transglutaminase II**. *J Biol Chem* (2001.0) **276** 20673-8. DOI: 10.1074/jbc.M010846200
31. Okabe M, Ikawa M, Kominami K, Nakanishi T, Nishimune Y. **‘Green mice’ as a source of ubiquitous green cells**. *FEBS Lett* (1997.0) **407** 313-9. DOI: 10.1016/S0014-5793(97)00313-X
32. Tanaka M, Ikeda K, Suganami T, Komiya C, Ochi K, Shirakawa I. **Macrophage-inducible C-type lectin underlies obesity-induced adipose tissue fibrosis**. *Nat Commun* (2014.0) **5** 1-13. DOI: 10.1038/ncomms5982
33. 33.Toda G, Yamauchi T, Kadowaki T, Ueki K. Preparation and culture of bone marrow-derived macrophages from mice for functional analysis. STAR Protoc. 2020;2:100246.
34. Tatsukawa H, Otsu R, Tani Y, Wakita R, Hitomi K. **Isozyme-specific comprehensive characterization of transglutaminase-crosslinked substrates in kidney fibrosis**. *Sci Rep* (2018.0) **8** 7306. DOI: 10.1038/s41598-018-25674-4
35. Kanda Y. **Investigation of the freely available easy-to-use software ‘EZR’ for medical statistics**. *Bone Marrow Transplant* (2012.0) **48** 452-8. DOI: 10.1038/bmt.2012.244
36. 36.Fujiu K, Manabe I, Nagai R. Renal collecting duct epithelial cells regulate inflammation in tubulointerstitial damage in mice. J Clin Invest. 2011;121:3425–41.
37. Yang Q, Wang Y, Pei G, Deng X, Jiang H, Wu J. **Bone marrow-derived Ly6C− macrophages promote ischemia-induced chronic kidney disease**. *Cell Death Dis* (2019.0) **10** 1-16. DOI: 10.1038/s41419-019-1531-3
38. Sears SM, Vega AA, Kurlawala Z, Oropilla GB, Krueger A, Shah PP. **F4/80hi resident macrophages contribute to cisplatin-induced renal fibrosis**. *Kidney360* (2022.0) **3** 818-33. DOI: 10.34067/KID.0006442021
39. Conrad DJ, Kuhn H, Mulkins M, Highland E, Sigal E. **Specific inflammatory cytokines regulate the expression of human monocyte 15-lipoxygenase**. *Proc Natl Acad Sci USA* (1992.0) **89** 217-21. DOI: 10.1073/pnas.89.1.217
40. Huang JT, Welch JS, Ricote M, Binder CJ, Willson TM, Kelly C. **Interleukin-4-dependent production of PPAR-γ ligands in macrophages by 12/15-lipoxygenase**. *Nature* (1999.0) **400** 378-82. DOI: 10.1038/22572
41. Montford JR, Bauer C, Rahkola J, Reisz JA, Floyd D, Hopp K. **15-Lipoxygenase worsens renal fibrosis, inflammation, and metabolism in a murine model of ureteral obstruction. Am J Physiol**. *Ren Physiol* (2022.0) **322** F105-19. DOI: 10.1152/ajprenal.00214.2021
42. Scarpellini A, Huang L, Burhan I, Schroeder N, Funck M, Johnson TS. **Syndecan-4 knockout leads to reduced extracellular transglutaminase-2 and protects against tubulointerstitial fibrosis**. *J Am Soc Nephrol* (2014.0) **25** 1013-27. DOI: 10.1681/ASN.2013050563
43. Burhan I, Furini G, Lortat-Jacob H, Atobatele AG, Scarpellini A, Schroeder N. **Interplay between transglutaminases and heparan sulphate in progressive renal scarring**. *Sci Rep* (2016.0) **6** 31343. DOI: 10.1038/srep31343
44. Ginhoux F, Guilliams M. **Tissue-resident macrophage ontogeny and homeostasis**. *Immunity* (2016.0) **44** 439-49. DOI: 10.1016/j.immuni.2016.02.024
45. 45.Sprangers S, Vries TJD, Everts V. Monocyte heterogeneity: consequences for monocyte-derived immune cells. J Immunol Res. 2016;2016:1475435.
46. Kojima S, Nara K, Rifkin DB. **Requirement for transglutaminase in the activation of latent transforming growth factor-β in bovine endothelial cells**. *J Cell Biol* (1993.0) **121** 439-48. DOI: 10.1083/jcb.121.2.439
47. Verderio E, Gaudry C, Gross S, Smith C, Downes S, Griffin M. **Regulation of cell surface tissue transglutaminase: effects on matrix storage of latent transforming growth factor-β binding protein-1**. *J Histochem Cytochem* (1999.0) **47** 1417-32. DOI: 10.1177/002215549904701108
48. Johnson TS, Griffin M, Thomas GL, Skill J, Cox A, Yang B. **The role of transglutaminase in the rat subtotal nephrectomy model of renal fibrosis**. *J Clin Invest* (1997.0) **99** 2950-60. DOI: 10.1172/JCI119490
49. He L, Jhong JH, Chen Q, Huang KY, Strittmatter K, Kreuzer J. **Global characterization of macrophage polarization mechanisms and identification of M2-type polarization inhibitors**. *Cell Rep* (2021.0) **37** 109955. DOI: 10.1016/j.celrep.2021.109955
50. Ivanov I, Heydeck D, Hofheinz K, Roffeis J, O’Donnell VB, Kuhn H. **Molecular enzymology of lipoxygenases**. *Arch Biochem Biophys* (2010.0) **503** 161-74. DOI: 10.1016/j.abb.2010.08.016
51. Kuhn H, Banthiya S, Van K. **Mammalian lipoxygenases and their biological relevance**. *Biochim Biophys Acta - Mol Cell Biol Lipids* (2015.0) **1851** 308-30. DOI: 10.1016/j.bbalip.2014.10.002
52. Takahashi N, Kikuchi H, Usui A, Furusho T, Fujimaru T, Fujiki T. **Deletion of Alox15 improves kidney dysfunction and inhibits fibrosis by increased PGD 2 in the kidney**. *Clin Exp Nephrol* (2021.0) **25** 445-55. DOI: 10.1007/s10157-021-02021-y
53. Elmarakby AA, Ibrahim AS, Katary MA, Elsherbiny NM, El-Shafey M, Abd-Elrazik AM. **A dual role of 12/15-lipoxygenase in LPS-induced acute renal inflammation and injury**. *Biochim Biophys Acta Mol Cell Biol Lipids* (2019.0) **1864** 1669-80. DOI: 10.1016/j.bbalip.2019.07.009
54. Yuan H, Reddy MA, Deshpande S, Jia Y, Park JT, Lanting LL. **Epigenetic histone modifications involved in profibrotic gene regulation by 12/15-lipoxygenase and its oxidized lipid products in diabetic nephropathy**. *Antioxid Redox Signal* (2016.0) **24** 361-75. DOI: 10.1089/ars.2015.6372
55. Wang X, Jiang L, Liu XQ, Huang YB, Zhu W, Zeng HX. **Identification of genes reveals the mechanism of cell ferroptosis in diabetic nephropathy**. *Front Physiol* (2022.0) **13** 1
56. Kühn H, O’Donnell VB. **Inflammation and immune regulation by 12/15-lipoxygenases**. *Prog Lipid Res* (2006.0) **45** 334-56. DOI: 10.1016/j.plipres.2006.02.003
57. Kuhn H, Gehring T, Schröter A, Heydeck D. **Cytokine-dependent expression regulation of ALOX15**. *J Cytokine Biol* (2016.0) **1** 1-14. DOI: 10.4172/2576-3881.1000106
58. Sendobry SM, Cornicelli JA, Welch K, Grusby MJ, Daugherty A. **Absence of T lymphocyte-derived cytokines fails to diminish macrophage 12/15-lipoxygenase expression in vivo 1**. *J Immunol* (1998.0) **161** 1477-82. DOI: 10.4049/jimmunol.161.3.1477
59. Namgaladze D, Snodgrass RG, Angioni C, Grossmann N, Dehne N, Geisslinger G. **AMP-activated protein kinase suppresses arachidonate 15-lipoxygenase expression in interleukin 4-polarized human macrophages**. *J Biol Chem* (2015.0) **290** 24484-94. DOI: 10.1074/jbc.M115.678243
60. Daniel B, Nagy G, Horvath A, Czimmerer Z, Cuaranta-Monroy I, Poliska S. **The IL-4/STAT6/PPARγ signaling axis is driving the expansion of the RXR heterodimer cistrome, providing complex ligand responsiveness in macrophages**. *Nucleic Acids Res* (2018.0) **46** 4425-39. DOI: 10.1093/nar/gky157
61. Majai G, Sarang Z, Csomós K, Zahuczky G, Fésüs L. **PPARgamma-dependent regulation of human macrophages in phagocytosis of apoptotic cells**. *Eur J Immunol* (2007.0) **37** 1343-54. DOI: 10.1002/eji.200636398
62. Maiuri L, Luciani A, Villella VR, Vasaturo A, Giardino I, Pettoello-Mantovani M. **Lysosomal accumulation of gliadin p31-43 peptide induces oxidative stress and tissue transglutaminase-mediated PPARgamma downregulation in intestinal epithelial cells and coeliac mucosa**. *Gut* (2010.0) **59** 311-9. DOI: 10.1136/gut.2009.183608
63. Maiuri L, Luciani A, Giardino I, Raia V, Villella VR, D’Apolito M. **Tissue transglutaminase activation modulates inflammation in cystic fibrosis via PPARgamma down-regulation**. *J Immunol* (2008.0) **180** 7697-705. DOI: 10.4049/jimmunol.180.11.7697
64. Han H, Xu D, Liu C, Claesson HE, Björkholm M, Sjöberg J. **Interleukin-4-mediated 15-lipoxygenase-1 trans-activation requires UTX recruitment and H3K27me3 demethylation at the promoter in A549 cells**. *PLoS ONE* (2014.0) **9** e85085. DOI: 10.1371/journal.pone.0085085
65. Farrelly LA, Thompson RE, Zhao S, Lepack AE, Lyu Y, Bhanu NV. **Histone serotonylation is a permissive modification that enhances TFIID binding to H3K4me3**. *Nature* (2019.0) **567** 535-9. DOI: 10.1038/s41586-019-1024-7
|
---
title: 'Comparative study of three plant-derived extracts as new management strategies
against Spodoptera littoralis (Boisd.) (Lepidoptera: Noctuidae)'
authors:
- Hanaa S. Hussein
- Mohamed Z. M. Salem
- Ahmed M. Soliman
- Sahar E. Eldesouky
journal: Scientific Reports
year: 2023
pmcid: PMC9981771
doi: 10.1038/s41598-023-30588-x
license: CC BY 4.0
---
# Comparative study of three plant-derived extracts as new management strategies against Spodoptera littoralis (Boisd.) (Lepidoptera: Noctuidae)
## Abstract
Finding innovative eco-friendly agents for pest control may be aided by investigating the plant-derived extracts’ properties on economic pests. Therefore, the insecticidal, behavioral, biological and biochemical effects of *Magnolia grandiflora* (Magnoliaceae) leaf water and methanol extracts, *Schinus terebinthifolius* (Anacardiaceae) wood methanol extract, and *Salix babylonica* (Salicaceae) leaf methanol extract in comparison with a reference insecticide novaluron against S. littoralis were evaluated. The extracts were analyzed by High-Performance Liquid Chromatography (HPLC). The most abundant phenolic compounds were 4-hydroxybenzoic acid (7.16 mg/mL) and ferulic acid (6.34 mg/mL) in M. grandiflora leaf water extract; catechol (13.05 mg/mL), ferulic acid (11.87 mg/mL), and chlorogenic acid (10.33 mg/mL) in M. grandiflora leaf methanol extract; ferulic acid (14.81 mg/mL), caffeic acid (5.61 mg/mL), and gallic acid (5.07 mg/mL) In the S. terebinthifolius extract; cinnamic acid (11.36 mg/mL), and protocatechuic acid (10.33 mg/mL) In the methanol extract from S. babylonica extract. S. terebinthifolius extract had a highly toxic effect against second larvae after 96 h and eggs with LC50 values of 0.89 and 0.94 mg/L, respectively. Despite M. grandiflora extracts didn’t show any toxicity against S. littoralis stages, they had an attractant effect on fourth- and second larvae, with feeding deterrence values of − $2.7\%$ and − $6.7\%$, respectively, at 10 mg/L. S. terebinthifolius extract significantly reduced the percentage of pupation, adult emergence, hatchability, and fecundity, with values of $60.2\%$, $56.7\%$, $35.3\%$, and 105.4 eggs/female, respectively. Novaluron and S. terebinthifolius extract drastically inhibited the activities of α-amylase and total proteases to 1.16 and 0.52, and 1.47 and 0.65 ΔOD/mg protein/min, respectively. In the semi-field experiment, the residual toxicity of tested extracts on S. littoralis gradually decreased over time compared to novaluron. These findings indicate that extract from S. terebinthifolius is a promising insecticidal agent against S. littoralis.
## Introduction
The cotton leafworm, *Spodoptera littoralis* (Boisd) (Lepidoptera: Noctuidae), is the major destructive pest of several agricultural crops including cotton, eggplant, tomato, and some ornamental products in Africa, Mediterranean Europe and Middle Eastern countries. More than 100 host attacked by this pest species, which causes yield losses of $50\%$, related to its larval foliage consumption activity1. The management of insect pests is one of the main challenges for agricultural researchers, and it’s difficultly increases with pest resistance and cross-resistance to chemical insecticides2. The negative impacts of pesticides on human health and environment have prompted more research on alternative control strategies for the integrated management of native and invasive pests3–7. To produce high-quality, pest-free crops without endangering the environment, researchers have focused on finding alternative, effective, and environmentally friendly control methods8. Botanicals are plant-derived materials that can be used as major components in integrated pest management (IPM) to control insect pests9–12 and reduce the use of synthetic insecticides. Thus, plant extracts are viable alternatives as they can regulate pest insect populations by affecting biological and behavioral parameters13,14.
Among the plant families that regulate insect pest populations, the previously tested behavioral and insecticidal properties of Magnoliaceae, Salicaceae, and Anacardiaceae were analyzed in the present study. Many studies have been performed on Magnolia species for their biphenolic phytochemicals magnolol and honokiol, which possess diverse pharmacological properties. Moreover, Magnolia extracts and their bioactive chemicals have been evaluated for potential insecticidal activity15–17. Different solvent extracts of M. salicifolia Maxim showed good to moderate larvicidal effects on fourth larvae of Aedes aegypti18.
Willows (Salix spp., family Salicaceae) are deciduous trees or shrubs well known for their medicinal effects. Many ancient civilizations used extracts of willow bark and willow leaf because of their analgesic, antipyretic, and anti-inflammatory properties19,20. Many studies have documented the presence of bioactive secondary compounds, such as polyphenols, terpenoids, and most importantly, salicylate compounds, in these plants21–25, which play a critical role not only as a part of their defense mechanisms and signaling molecules, but also as therapeutic agents (especially salicin)26,27. Plants synthesize salicylic acid (SA) through two pathways: the isochorismate pathway (IC) and the phenylalanine ammonia-lyase (PAL) pathway28. Willow is a well-known source of SA, which induces systemic resistance against several plant diseases29. Aqueous extracts of willow reduced Fusarium wilt in tomato seedlings by decreasing the level of lipid peroxidation30. The bark extract of the common willow has been approved by the EU pesticide regulations for agricultural applications as a basic substance with fungicidal properties31,32.
Schinus terebinthifolia (Anacardiaceae), commonly known as Brazilian pepper, has received particular attention owing to its nutritional, ornamental, and health-promoting properties, which can be attributed to a plethora of bioactive components, particularly phenols, tannins, flavonoids, saponins, alkaloids, and sterols33–35. Schinus terebinthifolius has insecticidal properties against Stegomyia aegypti36, Anopheles gambiae, A. arabiensis, and Culex quinquefasciatus37, S. littoralis38, and whitefly, Bemisia tabaci39. The essential oils of S. terebinthifolius fruits can also be used in the control of S. littoralis and Phthorimaea operculella, in association with IPM practices40.
Chitin synthesis inhibitors are insect growth regulators that affect insect chitin biosynthesis41–43. Novaluron is a benzoylphenylurea insecticide that interferes with developmental processes in immature insects, including abortive molting44–46. Novaluron ingested by adults can often be transferred transovarially to eggs, thereby reducing populations of economically important insect pests45,47. Moreover, it has low toxicity to mammals and several important natural enemies48.
Determining how digestive enzymes react to various inhibitors is a promising method to control phytophagous insects. There is limited published information on the inhibitors of S. littoralis digestive enzymes49,50.
The present study was conducted to evaluate the insecticidal activity of M. grandiflora, S. terebinthifolius, and S. babylonica extracts against different stages of S. littoralis (Boisd.). Furthermore, this study aimed to investigate the repellent and biological effects of these extracts on S. littoralis under laboratory or semi-field conditions; to determine the biochemical properties of extracts (e.g., in vitro inhibition of α-amylase and total protease activities); and to provide recommendations for using these plant-derived extracts in IPM programs to control this major pest.
## Insect rearing
This study has complied with relevant institutional, national, and international guidelines and legislation. This study does not contain any studies with human participants or animals performed by any of the authors. A laboratory strain of S. littoralis was reared on castor bean leaves, *Ricinus communis* L., under constant conditions of 27 ± 2 °C and 65 ± $5\%$ relative humidity (RH), in the Insect Physiology Laboratory, Department of Applied Entomology and Zoology, Faculty of Agriculture, Alexandria University, Egypt. Moths were provided with Nerium oleander L. leaves for egg laying. Moreover, as the field strain, egg masses of S. littoralis were collected from cotton fields at El-Beheira Governorate, Egypt, and maintained in the laboratory under the aforementioned conditions.
## Test extracts
Extracts from *Magnolia grandiflora* leaves (Magnoliaceae), *Schinus terebinthifolius* wood (Anacardiaceae), and *Salix babylonica* leaves (Salicaceae) were used in this study. Novaluron (Equo® $10\%$ EC; field rate, 60 mL/100 L water; Isagro Co., Italy) was used as a positive control for evaluating and comparing with the effectiveness of these extracts. All solvents and reagents used in experiments were analytical grade.
## Extraction procedure
Plant materials from the three tree species (M. grandiflora leaves, S. terebinthifolius wood, and S. babylonica leaves) were collected from Alexandria, Egypt. The collection of plants have been done and identified at the Department of Forestry and Wood Technology, Faculty of Agriculture, Alexandria University, Alexandria, Egypt. All plant materials were air-dried at room temperature for approximately 10 days and then ground to a powder using a small laboratory mill. Approximately 50 g of *Magnolia grandiflora* leaves was soaked in n-hexane (100 mL) in a conical flask for 3 days and then filtered through Whatman no. 1 filter paper. The solvent was evaporated using a rotary evaporator, and the n-hexane oily extract was concentrated.
For the extracts that were analyzed by HPLC, water and methanol extracts from M. grandiflora leaves and methanol extracts of S. terebinthifolius wood and S. babylonica leaves were used. Approximately 50 g of each ground material was soaked in 150 mL of solvent (water or methanol) for one week, then filtered through filter paper (Whatman no. 1), and concentrated by evaporating the solvent under reducing pressure with a rotary evaporator51. For the water extract of M. grandiflora leaves, a few drops of methanol were used to prevent any fungal growth. The content of the extracts was measured as a percentage per mass of the air-dried raw materials.
## HPLC analysis of extracts
For the phytochemical analysis, the phenolic compounds from extracts of M. grandiflora leaves, S. terebinthifolius wood, and S. babylonica leaves were identified by HPLC (Agilent 1100). The instrument was composed of binary LC pump, a UV/Vis detector, and C18 column (125 mm × 4.60 mm, 5-µm particle size)52.
## Ovicidal activity
Freshly deposited egg masses from the laboratory strain of S. littoralis were collected and counted using a hand lens (10 ×). Six concentrations of each extract and novaluron (0.5, 1, 2, 5, 10, and 20 mg/L) were prepared. Oleander leaves containing approximately 100 eggs were dipped for 20 s in each concentration of the test compounds separately. Another set of egg masses (100 eggs) on the oleander leaves was dipped in water to represent the control. Each concentration and control was replicated thrice. Treated and untreated egg masses were left to dry and maintained at 27 ± 2 °C, 65 ± $5\%$ RH. After the maximum hatching time, the unhatched eggs in each treatment were counted using a binocular.
## Larvicidal activity
The efficacy of the tested plant extracts and novaluron against the newly molted second and fourth larvae of S. littoralis was evaluated using a standard leaf-dip method. Six concentrations of each extract and insecticide (0.5, 1, 2, 5, 10, and 20 mg/L) were prepared. Castor bean leaves, which were almost equal in size, were dipped in the tested concentrations for 10 s and then left to dry. A set of castor leaves was dipped in distilled water only as the control. Each treatment was replicated thrice (20 larvae per treatment). The larvae were allowed to feed on treated leaves, and the mortality percentages were recorded 48 and 96 h post-treatment.
## Feeding deterrence activity
The feeding deterrence effect of the M. grandiflora leaf water extract, S. terebinthifolius wood methanol extract, and S. babylonica leaf methanol extract and of novaluron insecticide against second and fourth larvae of S. littoralis was determined using the leaf disc method (no-choice test) 48 h post-treatment. Three concentrations of M. grandiflora extract (1, 5, and 10 mg/L), S. terebinthifolius and S. babylonica extracts (1, 2, and 5 mg/L), and novaluron (0.5, 1, and 2 mg/L) were used. These concentrations were chosen after the preliminary tests according to their effectiveness. The feeding deterrence index (FD %) was calculated using the following equation: FD % = [(C − T) / (C + T)] × 100; where C is the consumption of control discs and T is the consumption of treated discs53.
The tested extracts were investigated for their feeding deterrent or attractant activity against second and fourth larvae of S. littoralis. All the tested materials showed feeding deterrence (FD %) values above the negative control, except for M. grandiflora extract, which was a feeding attractant for second and fourth larvae (Fig. 2). The fourth larvae were generally more affected by the tested compounds than the second larvae. Novaluron had a higher feeding repellent activity against S. littoralis larvae than all tested extracts. S. terebinthifolius wood methanol extract had a significantly higher feeding deterrence activity than the other extracts (FD % = $21.9\%$ and $18.4\%$ for fourth and second larvae, respectively, at 5 mg/mL). In contrast, M. grandiflora extract showed an attractant effect that decreased with increasing concentration (FD% = − $9.6\%$ and − $6.6\%$ for fourth and second larvae, respectively, at 10 mg/L).Figure 2Anti-oviposition and antifeedant activities of M. grandiflora, S. terebinthifolius, S. babylonica extracts, and novaluron on egg laying and on second and fourth larvae of *Spodoptera littoralis* after 48 h of treatment (no-choice test).
## Anti-oviposition activity
A no-choice test was used to evaluate the effects of the tested extracts on the oviposition. The three above-mentioned concentrations of each plant extract and insecticide were used. Each pair (female and male) of newly emerged adults was placed in a glass jar with a ball of cotton dipped in a $10\%$ sugar solution for feeding. Oleander leaves were treated with the test concentrations. The adults were left to feed, mate, and lay eggs on control and treated oleander leaves. Adults were removed two days after the beginning of egg laying, oleander leaves were carefully taken, and the number of eggs laid by each female was counted using a binocular. The anti-oviposition effect was calculated as follows54: Repellent index (RI %) = [(C − T) / (C + T)] × 100; where C is the number of eggs in the control, and T is the number of eggs in the treatment.
Insect oviposition is an important step in reproduction and in determining the size of a population. Therefore, deterrence of oviposition by a pest insect can decrease population size and assist in its management. The oviposition behavior of some phytophagous insects is altered by volatile products of host and non-host species. The deterrent or attractant activity of the tested materials against oviposition is shown as the repellent index (RI%) in Fig. 2. All tested compounds showed anti-oviposition activity, except for the M. grandiflora leaf water extract, which had RI% = − $8\%$, -$6.7\%$ and -$2.8\%$ at 1, 5 and 10 mg/L, respectively.
## Biological aspects
Bioactivity of the tested plant extracts and the insecticide was assessed under laboratory conditions (27 ± 2 °C, 65 ± $5\%$ RH). The castor bean leaves were immersed separately in the above-mentioned concentrations of tested compounds or in distilled water for control, dried at room temperature, and transferred to petri dishes (12 cm in diameter). One hundred neonates (0–24 h) S. littoralis larvae were placed in each petri dish. Castor bean leaves were replaced with newly treated leaves every 24 h. To establish the pupation percentage and observe malformations, the larvae were continuously monitored until they reached the pupal stage. The pupae were sexed and transferred to 1-L glass containers (10 males and 10 females per container) to assess the percentage of adult emergence, mean number of eggs per female (fecundity), and hatchability.
## Biochemical assays
The in vitro inhibition of α-amylase and total protease activity were determined by incubating the prepared homogenate for 30 min at 37 °C with LC50 concentrations of the tested compounds prepared in distilled water containing the emulsifying agent ($0.01\%$ Triton-X 100). The control treatments were prepared by adding $0.01\%$ Triton-X 100 without the tested compounds. Fourth larvae were then dissected, and midguts were excised, collected, and washed repeatedly with ice-cold saline solution ($0.9\%$ NaCl). The midguts were then homogenized in distilled water using a glass homogenizer surrounded with ice. The protein content was estimated by the method of Lowry et al.55 using bovine serum albumin as a standard protein to construct the standard curve.
## Alpha-amylase activity assay
The homogenate was centrifuged at 15,000 rpm for 15 min at 4 °C using IEC-CRU 5000 cooling centrifuge. The α-Amylase activity was estimated spectrophotometerically56. Fifty microliters of supernatant was added to 2.3 mM 2-chloro 4-nitrophenyl-α-Dmaltotrioside (CNPG3), 350 mM NaCl, 6 mM calcium acetate, 600 mM potassium thiocyanate, and 100 mM Good’s buffer (pH 6). An assay mixture without enzyme was used as a blank. The change in absorption at 405 nm was monitored using a Sequoia-Turner Model 340 spectrophotometer. The α-amylase activity was calculated as ΔOD405/mg protein/min.
## Total protease activity assay
The homogenate was centrifuged at 4000 rpm for 15 min at 4 °C in an IEC-CRU 5000 cooling centrifuge. The supernatant was used to estimate total proteolytic activity. Total protease activity was measured57,58 using azocasein as a substrate. The homogenate was incubated in a total volume 60 μL of assay buffer (100 mM Tris–HCl, pH 8) for 20 min at 37 °C before addition of 200 μL of $2\%$ azocasein (w/v in assay buffer). After 180 min at 37 °C, the reaction was stopped by addition of 300 μL cold $10\%$ trichloroacetic acid (TCA). The reaction mixture was centrifuged at 3000 rpm for 10 min in an IEC-CRU 5000 cooling centrifuge. Then, 10 μL NaOH (10 N) were added to the reaction mixture to neutralize excess acidity, and the absorbance was measured at 440 nm using a Sequoia-Turner Model 340 spectrophotometer. An assay mixture without homogenate was used as a blank. The total protease activity was calculated as ΔOD440/mg protein/min.
## Semi-field experiment
The residual toxicity of M. grandiflora, S. terebinthifolius, and S. babylonica extracts in comparison with novaluron against the field strain of fourth S. littoralis larvae was tested according to Raslan59 and El-Sheikh and Aamir60. Cotton seeds (Gossypium barbadense Linnaeus var. Giza 92) were sown in 50 plastic pots (30-cm in diameter) in the greenhouse of Cotton Pesticides Evaluation Department, Plant Protection Research Station, Alexandria, Egypt. Thirty days after emergence, the tested extracts and insecticide were applied as a foliar spray at a field rate (60 mL/100 L water) and a half field rate (30 mL/100 L water) using a hand-held sprayer with 1-L capacity until the leaves were saturated, and left to dry. The untreated plants were sprayed with tap water only. Cotton leaves from treated and untreated plants were randomly collected in perforated bags after 2 h of application and then 1, 2, 3, 4, and 7 days after the application, and transferred to the laboratory. Two cotton leaves from each sample were introduced to 20 newly molted fourth larvae of S. littoralis in Petri dish (12 cm in diameter) containing filter paper. Five replicates were performed for each treatment group. The Petri dishes were kept under laboratory conditions, at 27 ± 2 °C and 65 ± 5 RH %. The number of dead larvae was recorded, and mortality percentages were calculated 48 h after feeding.
## Statistical analysis
The LC50 and LC95 values for the toxicity tests were calculated using the Biostat ver. ( 2.1) software61 for probit analysis. Data were compared by one-way analysis of variance (ANOVA) followed by Tukey’s studentized test when significant differences were found at $P \leq 0.0562.$
## Chemical composition of the extracts
The extract contents from the studied plants were in M. grandiflora leaf water extract ($8.12\%$), M. grandiflora leaf methanol extract ($12.15\%$), S. terebinthifolius wood methanol extract ($16.14\%$), and in S. babylonica leaf methanol extract ($15.24\%$).
According to the HPLC analysis (Fig. 1 and Table 1), M. grandiflora leaf water extract (Fig. 1A) contained as main phenolic compounds 4-hydroxybenzoic acid (7.16 mg/mL), cinnamic acid (4.96 mg/mL), and ferulic acid (6.34 mg/mL), while the methanol extract (Fig. 1B) had catechol (13.05 mg/mL), ferulic acid (11.87 mg/mL), chlorogenic acid (10.33 mg/mL), and cinnamic acid (7.65 mg/mL). In the S. terebinthifolius wood extract (Fig. 1C), the phenolic compounds ferulic acid (14.81 mg/mL), caffeic acid (5.61 mg/mL), gallic acid (5.07 mg/mL), and chlorogenic acid (4.80 mg/mL) were abundant. In the methanol extract from S. babylonica leaves, (Fig. 1D) the phenolic compounds cinnamic acid (11.36 mg/mL), protocatechuic acid (10.33 mg/mL), ferulic acid (8.12 mg/mL), pyrogallol (8.05 mg/mL), and salicylic acid (6.44 mg/mL) were abundant. Figure 1Chromatograms of High Performance Liquid Chromatography of the phenolic compounds identified in the extracts. ( A) M. grandiflora leaf water extract; (B) M. grandiflora leaf methanol extract; (C) S. terebinthifolius wood methanol extract; (D) S. babylonica leaf methanol extract. Table 1Phenolic compounds from extracts of M. grandiflora leaves, S. babylonica leaves and S. terebinthifolius wood by HPLC analysis. CompoundConcentration (mg/mL)M. grandiflora leavesS. terebinthifolius woodS. babylonica leavesWaterMethanolMethanolMethanolCatecholND13.05NDNDCaffeic acidND3.665.610.69Ferulic acid6.3411.8714.818.12Gallic acidNDND5.07NDChlorogenic acidND10.334.80ND4-Hydroxybenzoic acid7.16NDND1.23Cinnamic acid4.967.65ND11.36Salicylic acidNDNDND6.44PyrogallolNDNDND8.05Protocatechuic acidNDND3.6910.33ND not detected.
In the current study, the n-hexane extract of M. grandiflora leaves, which was analyzed by GC–MS had the following main compounds: palmitic acid, oleic acid, undecane, palmitoleic acid, (1-propyloctyl) benzene, (1-methyldecyl)-benzene, (1-ethylnonyl) benzene, (1-propylnonyl) benzene, stearic acid, (1-pentylhexyl) benzene, (1-ethyldecyl) benzene, and linoleic acid at $7.28\%$, $7.22\%$, $5.37\%$, $4.66\%$, $4.63\%$, $4.21\%$, $3.88\%$, $3.86\%$, $3.46\%$, $3.36\%$, $3.3\%$, and $3\%$, respectively63.
## Insecticidal effect of the tested extracts against S. littoralis stages
The insecticidal effects of the tested extracts were evaluated at different stages in S. littoralis. The toxic effects are shown as LC50s values in Table 2. *In* general, the toxicity of the tested materials increased with time after treatment. Moreover, the instar larvae were the most sensitive stage to all tested materials. Table 2Toxicity of *Magnolia grandiflora* (water, methanol and n-hexane), Schinus terebinthifolius, *Salix babylonica* extracts and novaluron against different stages of S. littoralis after 48 and 96 h post-treatment. TreatmentStageTime (h)LC50a (mg/L)$95\%$ CLbLC95c (mg/L)$95\%$ CLSlope ± SEd(χ2)eM. grandiflora (Leaf water extract)Egg–>10–––––2nd larval instar48>10–––––96>10–––––4th larval instar48>10–––––96>10–––––M. grandiflora (Leaf methanol extract)Egg–>10–––––2nd larval instar48>10–––––96>10–––––4th larval instar48>10–––––96>10–––––M. grandiflora (leaf n-hexane extract)Egg–>10–––––2nd larval instar48>10–––––96>10–––––4th larval instar48>10–––––96>10–––––S. terebinthifolius (Wood methanol extract)Egg–0.940.68–1.2017.0610.72–34.821.31 ± 0.150.182nd larval instar481.651.29–2.0617.5311.12–35.091.60 ± 0.180.12960.890.72–1.1021.3113.32–46.681.73 ± 0.200.024th larval instar481.841.17–2.5342.7123.35–89.321.20 ± 0.180.16961.320.79–1.8626.0715.40–64.061.27 ± 0.190.03S. babylonica (Leaf methanol extract)Egg–1.050.82–1.3112.117.99–22.381.55 ± 0.170.052nd larval instar481.641.32–2.0817.5311.12–35.091.64 ± 0.180.02961.020.74–1.3313.728.52–29.361.46 ± 0.190.144th larval instar482.261.56–2.9740.7323.40–101.771.31 ± 0.170.09961.600.10–2.2133.7019.21–88.121.24 ± 0.180.03NovaluronEgg–0.620.49–0.786.774.52–12.241.58 ± 0.170.062nd larval instar481.230.95–1.5630.3317.23v74.631.62 ± 0.190.04960.550.41–0.7014.589.06–30.981.55 ± 0.200.074th larval instar481.341.05–1.7427.8718.15–53.521.48 ± 0.190.08960.870.59–1.1630.1518.12–69.481.37 ± 0.190.22aThe concentration causing $50\%$ mortality.bConfidence limits.cThe concentration causing $95\%$ mortality.dSlope of the concentration-mortality regression line ± standard error.eChi square value.
## Ovicidal effect
The three extracts of M. grandiflora did not show any toxicity against S. littoralis eggs, but those of S. terebinthifolius and S. babylonica showed toxic effects. The highly toxic effect was shown by novaluron on eggs (LC50 = 0.62 mg/L). The S. terebinthifolius extract also exhibited high toxicity against S. littoralis eggs (LC50 = 0.94 mg/L) (Table 2).
## Larvicidal effect
S. terebinthifolius and S. babylonica showed toxic effects against second and fourth larvae after 48 and 96 h. Novaluron showed high toxicity against second larvae after 96 h (LC50 = 0.55 mg/L). Among the tested extracts, that of S. terebinthifolius had high toxic effect against second larvae after 96 h (LC50 = 0.89 mg/L). Compared with novaluron (positive control), the S. terebinthifolius wood methanol extract exhibited high toxicity against S. littoralis at different stages (Table 2).
As the three extracts of M. grandiflora did not show any toxicity against S. littoralis stages, the water extract was only used in subsequent behavioral, biological, and semi-field experiments to investigate whether it had any other effects on S. littoralis. Moreover, its use in IPM programs with different modes of action apart from toxicity was also tested.
## Repellent effects of the tested extracts
Phytophagous insects such as S. littoralis usually visit plants for either feeding or laying eggs. Identification of effective repellent or attractant agents for the early detection and suppression of S. littoralis populations is critical for managing this pest and reducing crop loss.
## The impact of the tested compounds on some biological aspects
The results on pupation, adult emergence, fecundity (mean number of eggs/female), and hatchability (%) of the resulting eggs of the treated S. littoralis larvae are shown in Table 3. Adult growth disruption and abnormalities are shown in Fig. 3 (A–G). All tested materials affected the assessed biological aspects at all concentrations, except for the M. grandiflora extract, which affected larvae only at the highest concentration (10 mg/L). The biological effects of S. terebinthifolius were significantly reducing the percentages of pupation, adult emergence, hatchability, and fecundity ($60.2\%$, $56.7\%$, $35.3\%$, and $105.4\%$ eggs/female, respectively, at 5 mg/L).Table 3Effect of M. grandiflora, S. terebinthifolius, and S. babylonica extracts and of novaluron on pupation, adult emergence, egg production, and hatching percentages of S. littoralis after application on second larvae. TreatmentConc. ( mg/L)Pupation (%)Adult emergence (%)Number of eggs /femaleHatch (%)Control–94.2 ± 1.2a88.0 ± 1.3a216.2 ± 5.2a85.6 ± 1.8aM. grandiflora (Leaf water extract)193.8 ± 1.6a87.3 ± 2.5a210.3 ± 4.8a84.3 ± 0.9a592.7 ± 2.1a86.2 ± 1.8a202.7 ± 5.6ab83.5 ± 1.3a1090.4 ± 1.7b85.7 ± 3.4ab196.5 ± 5.3b80.6 ± 1.6bS. terebinthifolius (Wood methanol extract)166.0 ± 2.3e63.3 ± 2.1e128.6 ± 4.2f48.0 ± 1.5e265.3 ± 1.9e61.2 ± 2.7ef119.2 ± 3.4f44.7 ± 2.1e560.2 ± 3.2f56.7 ± 3.8f105.4 ± 3.1g35.3 ± 0.8fS. babylonica (Leaf methanol extract)181.3 ± 3.3c78.0 ± 3.2c174.2 ± 2.5cd66.2 ± 2.3c280.8 ± 2.8c76.8 ± 1.4c167.0 ± 4.3d58.3 ± 1.2d575.0 ± 3.6d69.3 ± 3.3d154.6 ± 5.8e54.6 ± 1.7dNovaluron0.556.3 ± 2.3g48.4 ± 3.6g54.8 ± 4.0h24.3 ± 2.4g149.6 ± 1.5h43.6 ± 4.2h38.9 ± 3.4i20.7 ± 2.3g247.2 ± 2.4h41.3 ± 2.9h26.4 ± 2.2j14.3 ± 1.8hMean values ± standard error followed by the same letters in the same column are not significantly different at $P \leq 0.05.$Figure 3Malformations of S. littoralis adults affected by applications on larval stage. Control: normal adult (A). Adults resulting from larvae treated with 1 and 2 mg/L novaluron (B & C), with short, undeveloped legs and wings. Adults resulting from larvae treated with 2 and 5 mg/L S. terebinthifolius (D), which were unable to remove the old exoskeleton (exuvium) and (E) with abnormal and conjoined wings. Adults resulting from larvae treated with 2 and 5 mg/L S. babylonica (F), with head and mouthparts not completely molted, (G) with shrunken, folded, and undeveloped wings and preserved pupal head and mouthparts.
## Alpha-amylase and total protease activities of S. littoralis fourth instar larvae
The significant inhibitory effects of LC50 S. terebinthifolius, S. babylonica, and novaluron on α-amylase and total proteases were determined in vitro (Fig. 4). Novaluron had the highest inhibition effect, with the activities of α-amylase and total proteases reduced to 1.16 and 0.52 ΔOD/mg protein/min, respectively, followed by the S. terebinthifolius extract, with activities reduced to 1.47 and 0.65 ΔOD/mg protein/min, respectively (compared with 2.34 and 1.05 ΔOD/mg protein/min in the control, respectively). The S. terebinthifolius extract reduced the activity of S. littoralis digestive enzymes. Figure 4Effect of S. terebinthifolius, and S. babylonica extracts and of novaluron on the α-amylase and total proteases activities of the fourth S. littoralis larvae at LC50 value at 48 h post-treatment.
## Residual toxicity of tested compounds against S. littoralis field strain
A semi-field experiment was conducted to evaluate the residual efficacy of M. grandiflora, S. terebinthifolius, S. babylonica, and novaluron against the fourth S. littoralis larvae from the field strain. The highest mortality percentages of S. littoralis larvae were $100\%$, $100\%$, $95\%$, and $70\%$ after 2 h of application with novaluron, S. terebinthifolius, S. babylonica, and M. grandiflora, respectively. The mortality percentages decreased gradually over time to become $40.0\%$, $20.0\%$, $10.0\%$, and $0\%$, respectively, after 7 days of spraying at the field rate (60 mL/100 L water). While, the mortality percentages were $30\%$, $15\%$, $5\%$, and $0\%$ at the half field rate (30 mL/100 L water) (Table 4).Table 4Residual toxicity of M. grandiflora, S. terebinthifolius, and S. babylonica extracts and of novaluron against the field strain of fourth S. littoralis larvae. TreatmentRateMortality percentages after the indicated periods from the application2 h1-day2-days3-days4-days7-daysM. grandiflora (Leaf water extract)0.5 FR60.0 ± 1.9e40.0 ± 1.2h25.0 ± 0.8h15.0 ± 1.4g5.0 ± 0.3h0FR70.0 ± 1.2d55.0 ± 1.8g35.0 ± 1.1g25.0 ± 0.9f10.0 ± 0.5g0S. terebinthifolius (Wood methanol extract)0.5 FR95.0 ± 1.3b85.0 ± 1.7d70.0 ± 1.3d50.0 ± 1.7c30.0 ± 0.6d15.0 ± 0.3dFR100.0 ± 0.0a90.0 ± 1.4c80.0 ± 2.1c70.0 ± 1.2b45.0 ± 1.3c20.0 ± 0.8cS. babylonica (Leaf methanol extract)0.5 FR85.0 ± 1.2c75.0 ± 1.3f45.0 ± 0.9f30.0 ± 0.5e15.0 ± 1.1f5.0 ± 0.3fFR95.0 ± 1.6b80.0 ± 0.4e60.0 ± 2.2e35.0 ± 1.6d25.0 ± 1.4e10.0 ± 0.2eNovaluron0.5 FR100.0 ± 0.0a95.0 ± 0.8b85.0 ± 1.3b70.0 ± 0.9b50.0 ± 1.6b30.0 ± 0.5bFR100.0 ± 0.0a100.0 ± 0.0a90.0 ± 0.6a80.0 ± 1.2a60.0 ± 0.8a40.0 ± 0.9aMean values ± standard error followed by the same letters in the same column are not significantly different at $P \leq 0.05.$ The mortality percentages were assessed after 48 h from the feeding. Field rate (FR) was applied at 60 ml/100 L water.
## Discussion
Insect pest management has always been and will remain a constant challenge for agricultural researchers and producers alike. There is an urgent need to replace pesticides with alternative control methods that are effective, inexpensive, and environmentally-friendly, therefore plant-derived products have received much attention in recent years due to drawbacks associated with unwise use of synthetic insecticides8.
Novaluron, which had already been shown to be highly toxic against S. littoralis eggs64, also showed a great ovicidal activity against S. littoralis eggs compared to the control. The strong insecticidal activity of S. terebinthifoiuls against S. littoralis has been previously reported40, as well as against Anopheles gambiae, A. arabiensis, and Culex quinquefasciatus37. S. terebinthifolius also showed high toxicity against two whitefly species, Bemisia tabaci, and Trialeurodes ricini39, *Aphis nerii* Boyer de Fonscolombe65 and *Plutella xylostella* (Lepidoptera: Plutellidae)66.
In contrast, M. grandiflora extract was herein found to have no toxic effect on S. littoralis stages, similar to the findings for *Magnolia citrata* essential oil, which had weak insecticidal activity against S. littoralis larvae compared with the positive control permethrin38. These results also agree with those of Vásquez-Morales and Flores-Estévez67, who found that the seed and sarcotesta extracts of *Magnolia schiedeana* (Magnoliaceae) only showed insecticidal activity against Anastrepha ludens adults, whereas the extracts of leaves, flowers, bark, and follicles showed no significant biological activity. Moreover, Ali et al.68 observed that the leaf, flower, and seed essential oils of M. grandiflora at the highest dose of 125 mg/L resulted in only $20\%$, $0\%$, and $50\%$ mortality of Aedes aegypti, respectively.
Furthermore, M. citrata oil has been reported to exhibit weak toxicity against first instar larvae and adult female A. aegypti69. Methanol extract of S. babylonica leaves showed strong toxicity against S. littoralis stages, which is consistent with the results of Hasaballah et al.70, who showed that the toxic effects of methanol and ethanol extracts of Salix safsaf could compete with the synthetic insecticide deltamethrin as a natural insecticide in the control of the housefly Musca domestica. Added to the antifungal effect of salicin, the major compound of willow extract, other metabolites may increase the potency of willow extracts71.
The attractant effect of Magnolia was first reported by Pavela38, who reported the attractant activity of M. citrata essential oil on S. littoralis. The oil from M. citrata leaves has a moderately strong attractant effect on the sterile male medfly Ceratitis capitata69. In contrast, essential oils from five different parts of M. grandiflora showed biting deterrence against Ae. aegypti68. Schinus terebinthifolius produced a significantly high feeding deterrence activity in second and fourth instar larvae. The ethanolic extracts of S. terebinthifolius were effective antifeedants for third instar larvae of *Plutella xylostella* (Lepidoptera: Plutellidae)72. Treatment with S. terebinthifolius extract prolonged the larval phase, allowing a higher food intake before reaching the pupal stage; this probably resulted from the effect of one or several deterrent factors, resulting in nutritional imbalance and damage to the insect life cycle73. Willow oil generated maximum attraction ($28.79\%$) in Bemisia tabaci, contrary to the repellent effect of willow oil on S. littoralis74.
Novaluron at 2 mg/L had the highest RI% ($68.8\%$), followed by S. terebinthifolius extract at 5 mg/L ($30.9\%$). Aly and Ali64 showed that novaluron had the highest oviposition deterrence value ($23\%$) in S. littoralis females. Ethanolic extracts of S. terebinthifolius suppressed oviposition in P. xylostella adults72.
Novaluron also causes S. littoralis growth disruption and abnormalities, selectively targeting immature insect stages by inhibiting chitin formation and causing abnormal endocuticular deposition abortive molting75. Novaluron had the highest sterility value ($68.9\%$) for S. littoralis64. The ethanolic extracts of S. terebinthifolius negatively affected all the evaluated biological parameters of P. xylostella, increasing the duration of the larval stage, which led to reduced pupal mass and oviposition period76. Treatment with S. terebinthifolius extract resulted in the lowest pupal mass and a greatest prolongation of the larval stage compared with other treatments73. Phytochemical studies on S. terebinthifolius have isolated tannins that inactivate digestive enzymes of insects, hampering their digestion, which in turn affects their growth and survival73,77. The S. terebinthifolius extract possibly reduced pupal survival by impairing their ability to feed as a result of larval sensitivity to the secondary compounds present in the plant extracts78.
It was previously explained that tannins act by inactivating the digestive enzymes in the leaves of S. terebinthifolius, generating a tannin-protein complex that is difficult to digest and affects the growth and survival of insects73. A reduction in food digestibility was observed in the tannin fractions of S. terebinthifolius by *Spodoptera frugiperda* larvae (Smith 1797) (Lepidoptera: Noctuidae)79.
Various studies have demonstrated that plant phenolic metabolites negatively affect insect feeding behavior, growth, development, and reproduction, and they may have lethal effects on specific insects80,81. Furthermore, α-amylase and protease activities of S. littoralis were decreased in the midgut after feeding on an artificial diet containing caffeic acid82. Ferulic acid, the most abundant phenolic compound from extracts of M. grandiflora leaves and S. terebinthifolius wood decreased adult emergence, delayed the developmental period and reduced the nutritional indices of *Spodoptera litura* (Fabricius) larvae83. The larval growth, survival, adult emergence, pupal weight, and different nutritional indices of S. litura (Fab.) were adversely affected by the various concentrations of purified phenolic compounds like chlorogenic acid84. The effect of some phenolic acids (chlorogenic, caffeic, and ferulic) on the growth, development and midgut enzyme activities of S. litura larvae was studied through diet incorporation assay and can be utilized in insect control programs85. The two cinnamic acid derivatives were found to show higher levels of insecticidal, larvicidal and larval growth inhibition activities against Tribolium castaneum86. The lethal effect of chlorogenic acid on *Mythimna separata* (Walker) (Lepidoptera: Noctuidae) and that a sublethal concentration harmed larval growth and development87.
## Conclusion
The extract of S. terebinthifolius can be used to efficiently manage S. littoralis, as they had various modes of action, such as toxicity, repellence, growth regulation, reduction of fecundity, and inhibition of digestive enzymes activity. Therefore, they can be considered suitable alternatives and can be incorporated into IPM systems for S. littoralis.
## References
1. Garrido-Jurado I, Montes-Moreno D, Sanz-Barrionuevo P, Quesada-Moraga E. **Delving into the causes and effects of entomopathogenic endophytic metarhizium brunneum foliar application-related mortality in**. *Insects* (2020.0) **11** 429. DOI: 10.3390/insects11070429
2. Mosallanejad H, Smagghe G. **Biochemical mechanisms of methoxyfenozide resistance in the cotton leafworm Spodoptera littoralis**. *Pest Manag. Sci.* (2009.0) **65** 732-736. DOI: 10.1002/ps.1753
3. Jepson PC, Murray K, Bach O, Bonilla MA, Neumeister L. **Selection of pesticides to reduce human and environmental health risks: A global guideline and minimum pesticides list**. *The Lancet Planetary Health* (2020.0) **4** e56-e63. DOI: 10.1016/S2542-5196(19)30266-9
4. Lengai GMW, Muthomi JW, Mbega ER. **Phytochemical activity and role of botanical pesticides in pest management for sustainable agricultural crop production**. *Sci. Afr.* (2020.0) **7** e00239. DOI: 10.1016/j.sciaf.2019.e00239
5. Rani L. **An extensive review on the consequences of chemical pesticides on human health and environment**. *J. Clean. Prod.* (2021.0) **283** 124657. DOI: 10.1016/j.jclepro.2020.124657
6. Samsidar A, Siddiquee S, Shaarani SM. **A review of extraction, analytical and advanced methods for determination of pesticides in environment and foodstuffs**. *Trends Food Sci. Technol.* (2018.0) **71** 188-201. DOI: 10.1016/j.tifs.2017.11.011
7. van den Berg H. **Pesticide lifecycle management in agriculture and public health: Where are the gaps?**. *Sci. Total Environ.* (2020.0) **742** 140598. DOI: 10.1016/j.scitotenv.2020.140598
8. Khan S. **Insecticidal activity of plant-derived extracts against different economically important pest insects**. *Phytoparasitica* (2017.0) **45** 113-124. DOI: 10.1007/s12600-017-0569-y
9. Tawfeek ME, Ali HM, Akrami M, Salem MZM. **Potential Insecticidal Activity of Four Essential Oils against the Rice Weevil,**. *BioResources* (2021.0) **16** 7767-7783. DOI: 10.15376/biores.16.4.7767-7783
10. Mansour S, Bakr R, Mohamed R, Hasaneen N. **Larvicidal activity of some botanical extracts, commercial insecticides and their binary mixtures against the housefly,**. *The Open Toxinol. J.* (2011.0) **3** 1-13. DOI: 10.2174/1875414701104010001
11. 11.Singh, A., Bhardwaj, R. & Singh, I. K. in Biofertilizers for Sustainable Agriculture and Environment (eds Bhoopander Giri, Ram Prasad, Qiang-Sheng Wu, & Ajit Varma) 413–433 (Springer, 2019).
12. Souto AL. **Plant-derived pesticides as an alternative to pest management and sustainable agricultural production: Prospects, applications and challenges**. *Molecules* (2021.0) **26** 4835. DOI: 10.3390/molecules26164835
13. Jeon J-H, Kim Y-K, Lee S-G, Lee G-H, Lee H-S. **Insecticidal activities of a**. *J. Asia-Pacific Entomol.* (2011.0) **14** 449-453. DOI: 10.1016/j.aspen.2011.07.005
14. Moustafa MAM. **Insecticidal activity of lemongrass essential oil as an eco-friendly agent against the black Cutworm**. *Insects* (2021.0) **12** 737. DOI: 10.3390/insects12080737
15. Ho K-Y, Tsai C-C, Chen C-P, Huang J-S, Lin C-C. **Antimicrobial activity of honokiol and magnolol isolated from Magnolia officinalis**. *Phytother. Res.* (2001.0) **15** 139-141. DOI: 10.1002/ptr.736
16. Sarrica A, Kirika N, Romeo M, Salmona M, Diomede L. **Safety and toxicology of Magnolol and Honokiol**. *Planta Med* (2018.0) **84** 1151-1164. DOI: 10.1055/a-0642-1966
17. Zhao X. **Extracts of Magnolia species-induced prevention of diabetic complications: A brief review**. *Int. J. Mol. Sci.* (2016.0) **17** 1629. DOI: 10.3390/ijms17101629
18. Kelm MA, Nair MG, Schutzki RA. **Mosquitocidal compounds from**. *Int. J. Pharmacogn.* (1997.0) **35** 84-90. DOI: 10.1076/phbi.35.2.84.13279
19. Mahdi JG. **Medicinal potential of willow: A chemical perspective of aspirin discovery**. *J. Saudi Chem. Soc.* (2010.0) **14** 317-322. DOI: 10.1016/j.jscs.2010.04.010
20. Noleto-Dias C, Ward JL, Bellisai A, Lomax C, Beale MH. **Salicin-7-sulfate: A new salicinoid from willow and implications for herbal medicine**. *Fitoterapia* (2018.0) **127** 166-172. DOI: 10.1016/j.fitote.2018.02.009
21. El-Sayed MM, El-Hashash MM, Mohamed HR, Abdel-Lateef EE-S. **Phytochemical investigation and in vitro antioxidant activity of different leaf extracts of**. *J. Appl. Pharm. Sci* (2015.0) **5** 080-085. DOI: 10.7324/JAPS.2015.501213
22. El-Shazly A, El-Sayed A, Fikrey E. **Bioactive secondary metabolites from Salix tetrasperma Roxb**. *Z Naturforsch C J Biosci* (2012.0) **67** 353-359. DOI: 10.5560/znc.2012.67c0353
23. Khan MIR, Fatma M, Per TS, Anjum NA, Khan NA. **Salicylic acid-induced abiotic stress tolerance and underlying mechanisms in plants**. *Front. Plant Sci.* (2015.0) **6** 462. DOI: 10.3389/fpls.2015.00462
24. Ruuhola T, Julkunen-Tiitto R, Vainiotalo P. **In Vitro degradation of willow**. *J. Chem. Ecol.* (2003.0) **29** 1083-1097. DOI: 10.1023/A:1023821304656
25. Shara M, Stohs SJ. **Efficacy and safety of White Willow Bark (**. *Phytother. Res.* (2015.0) **29** 1112-1116. DOI: 10.1002/ptr.5377
26. Alamgir A. *Therapeutic Use of Medicinal Plants and their Extracts* (2017.0)
27. Wiesneth S, Aas G, Heilmann J, Jürgenliemk G. **Investigation of the flavan-3-ol patterns in willow species during one growing-season**. *Phytochemistry* (2018.0) **145** 26-39. DOI: 10.1016/j.phytochem.2017.10.001
28. Mutlu-Durak H, Yildiz Kutman B. **Seed treatment with biostimulants extracted from Weeping Willow (**. *Plants* (2021.0) **10** 1449. DOI: 10.3390/plants10071449
29. Orellana C. **Aspirin protects against cancer of the upper aerodigestive tract**. *Lancet Oncol.* (2003.0) **4** 200. DOI: 10.1016/S1470-2045(03)01054-4
30. Farag HRM, Abdou ZA, Salama DA, Ibrahim MAR, Sror HAM. **Effect of neem and willow aqueous extracts on fusarium wilt disease in tomato seedlings: Induction of antioxidant defensive enzymes**. *Ann. Agric. Sci.* (2011.0) **56** 1-7. DOI: 10.1016/j.aoas.2011.05.007
31. Deniau M. **Willow extract (**. *Int. J. Bio-resource Stress Manag.* (2019.0) **10** 408-418. DOI: 10.23910/IJBSM/2019.10.4.2009
32. Marchand PA. **Basic substances under EC 1107/2009 phytochemical regulation: Experience with non-biocide and food products as biorationals**. *J. Plant Protect. Res.* (2016.0) **56** 312-318. DOI: 10.1515/jppr-2016-0041
33. Cavalher-Machado SC. **The anti-allergic activity of the acetate fraction of Schinus terebinthifolius leaves in IgE induced mice paw edema and pleurisy**. *Int. Immunopharmacol.* (2008.0) **8** 1552-1560. DOI: 10.1016/j.intimp.2008.06.012
34. de Oliveira VS. **Aroeira fruit (**. *Food Chem.* (2020.0) **315** 126274. DOI: 10.1016/j.foodchem.2020.126274
35. Locali-Pereira AR, Lopes NA, Nicoletti VR. **Pink Pepper (**. *Food Rev. Int.* (2022.0) **5** 1-30. DOI: 10.1080/87559129.2022.2062767
36. Silva AG. **The essential oil of Brazilian pepper, Schinus terebinthifolia Raddi in larval control of Stegomyia aegypti (**. *Parasit. Vectors* (2010.0) **3** 79. DOI: 10.1186/1756-3305-3-79
37. Kweka EJ, Nyindo M, Mosha F, Silva AG. **Insecticidal activity of the essential oil from fruits and seeds of Schinus terebinthifolia Raddi against African malaria vectors**. *Parasit. Vectors* (2011.0) **4** 129. DOI: 10.1186/1756-3305-4-129
38. Pavela R. **Acute, synergistic and antagonistic effects of some aromatic compounds on the**. *Ind. Crops Prod.* (2014.0) **60** 247-258. DOI: 10.1016/j.indcrop.2014.06.030
39. Hussein HS, Salem MZM, Soliman AM. **Repellent, attractive, and insecticidal effects of essential oils from Schinus terebinthifolius fruits and Corymbia citriodora leaves on two whitefly species, Bemisia tabaci, and Trialeurodes ricini**. *Sci. Hortic.* (2017.0) **216** 111-119. DOI: 10.1016/j.scienta.2017.01.004
40. Ennigrou A, Casabianca H, Laarif A, Hanchi B, Hosni K. **Maturation-related changes in phytochemicals and biological activities of the Brazilian pepper tree (**. *S. Afr. J. Bot.* (2017.0) **108** 407-415. DOI: 10.1016/j.sajb.2016.09.005
41. Gijswijt MJ, Deul DH, de Jong BJ. **Inhibition of chitin synthesis by benzoyl-phenylurea insecticides, III. Similarity in action in**. *Pest. Biochem. Physiol.* (1979.0) **12** 87-94. DOI: 10.1016/0048-3575(79)90098-1
42. Hajjar NP, Casida JE. **Insecticidal benzoylphenyl ureas: Structure-activity relationships as chitin synthesis inhibitors**. *Science* (1978.0) **200** 1499-1500. DOI: 10.1126/science.200.4349.1499
43. Post LC, de Jong BJ, Vincent WR. **1-(2,6-disubstituted benzoyl)-3-phenylurea insecticides: Inhibitors of chitin synthesis**. *Pestic. Biochem. Physiol.* (1974.0) **4** 473-483. DOI: 10.1016/0048-3575(74)90072-8
44. Cutler GC, Scott-Dupree CD. **Novaluron: Prospects and limitations in insect pest management**. *Pest Technology* (2007.0) **1** 38-46
45. Kostyukovsky M, Trostanetsky A. **The effect of a new chitin synthesis inhibitor, novaluron, on various developmental stages of Tribolium castaneum (Herbst)**. *J. Stored Prod. Res.* (2006.0) **42** 136-148. DOI: 10.1016/j.jspr.2004.12.003
46. Merzendorfer H. **Chitin synthesis inhibitors: Old molecules and new developments**. *Insect Sci.* (2013.0) **20** 121-138. DOI: 10.1111/j.1744-7917.2012.01535.x
47. Joseph VS. **Ingestion of novaluron elicits transovarial activity in**. *Insects* (2020.0) **11** 216. DOI: 10.3390/insects11040216
48. Soares WS. **Physiological selectivity of insecticides from different chemical groups and cuticle thickness of Protonectarina sylveirae (Saussure) and Brachygastra lecheguana (Latreille)**. *Sociobiology* (2019.0) **66** 358-366. DOI: 10.13102/sociobiology.v66i2.3478
49. Abdelgaleil SAM, Abou-Taleb HK, Al-Nagar NMA, Shawir MS. **Antifeedant, growth regulatory and biochemical effects of terpenes and phenylpropenes on**. *Int. J. Trop. Insect Sci.* (2020.0) **40** 423-433. DOI: 10.1007/s42690-019-00093-8
50. Hemmati SA, Shishehbor P, Stelinski LL. **Life table parameters and digestive enzyme activity of**. *Insects* (2022.0) **13** 661. DOI: 10.3390/insects13070661
51. 51.Seidel, V. in Natural Products Isolation (eds Satyajit D. Sarker & Lutfun Nahar) 27–41 (Humana Press, 2012).
52. Hassan HS. **Natural plant extracts and microbial antagonists to control fungal pathogens and improve the productivity of Zucchini (**. *Horticulturae* (2021.0) **7** 470. DOI: 10.3390/horticulturae7110470
53. Stefanazzi N, Stadler T, Ferrero A. **Composition and toxic, repellent and feeding deterrent activity of essential oils against the stored-grain pests**. *Pest Manag. Sci.* (2011.0) **67** 639-646. DOI: 10.1002/ps.2102
54. Pascual-Villalobos MJ, Robledo A. **Screening for anti-insect activity in Mediterranean plants**. *Ind. Crops Prod.* (1998.0) **8** 183-194. DOI: 10.1016/S0926-6690(98)00002-8
55. Lowry OH. **Protein measurement with the folin phenol reagent**. *J. Biol. Chem.* (1951.0) **193** 265-275. DOI: 10.1016/S0021-9258(19)52451-6
56. Kaufman RA, Tietz NW. **Recent advances in measurement of amylase activity–a comparative study**. *Clin. Chem.* (1980.0) **26** 846-853. DOI: 10.1093/clinchem/26.7.846
57. Loseva O. **Changes in protease activity and Cry3Aa toxin binding in the Colorado potato beetle: Implications for insect resistance to Bacillus thuringiensis toxins**. *Insect Biochem. Mol. Biol.* (2002.0) **32** 567-577. DOI: 10.1016/S0965-1748(01)00137-0
58. Mohan M, Gujar GT. **Characterization and comparison of midgut proteases of Bacillus thuringiensis susceptible and resistant diamondback moth (Plutellidae: Lepidoptera)**. *J. Invertebr. Pathol.* (2003.0) **82** 1-11. DOI: 10.1016/S0022-2011(02)00194-5
59. 59.Raslan, S. A. A. Preliminary report on initial and residual mortality of the natural product, spinosad for controlling cotton leafworm egg masses in 2002 cotton season at Sharkia governorate, Egypt. 2nd International Conference, Plant Protection Research Institute, Cairo, Egypt, 21–24 December, 2002. Volume 1, 635–637. (2002).
60. El-Sheikh E-SA, Aamir MM. **Comparative effectiveness and field persistence of insect growth regulators on a field strain of the cotton leafworm, Spodoptera littoralis, Boisd (Lepidoptera: Noctuidae)**. *Crop Prot.* (2011.0) **30** 645-650. DOI: 10.1016/j.cropro.2011.02.009
61. 61.Biostat. Biostat ver. (2.1). Computer program for probit analysis. 2011. (2011).
62. 62.SAS. Statistical Analysis System (SAS). SAS/STAT User’s Guide. Version 8, 6th Edition, SAS institute Inc., Cary, North Carolina, U.S.A. 2002. (2002).
63. Salem MZM, Abo Elgat WAA, Taha AS, Fares YGD, Ali HM. **Impact of three natural oily extracts as pulp additives on the mechanical, optical, and antifungal properties of paper sheets made from**. *Materials* (2020.0) **13** 1292. DOI: 10.3390/ma13061292
64. Aly MFK, Ali AM. **Impact of some essential plant oils and insect growth regulators on immature stages of**. *J. Plant Prot. Pathol.* (2017.0) **8** 561-570. DOI: 10.21608/jppp.2017.46852
65. Hussein HS, Tawfeek ME, Abdelgaleil SAM. **Chemical composition, aphicidal and antiacetylcholinesterase activities of essential oils against Aphis nerii Boyer de Fonscolombe (Hemiptera: Aphididae)**. *J. Asia-Pacific Entomol.* (2021.0) **24** 259-265. DOI: 10.1016/j.aspen.2021.02.001
66. Silva PRC. **Schinus terebinthifolia leaf extract is a larvicidal, pupicidal, and oviposition deterring agent against Plutella xylostella**. *S. Afr. J. Bot.* (2019.0) **127** 124-128. DOI: 10.1016/j.sajb.2019.08.054
67. Vásquez-Morales SG, Flores-Estévez N. **Bioprospecting of botanical insecticides: The case of ethanol extracts of Magnolia schiedeana Schltl. applied to a Tephritid, fruit fly Anastrepha ludens Loew**. *J. Entomol. Zool. Stud.* (2015.0) **3** 71
68. Ali A. **Insecticidal and biting deterrent activities of**. *Molecules* (2020.0) **25** 1359. DOI: 10.3390/molecules25061359
69. Luu-Dam NA, Tabanca N, Estep AS, Nguyen DH, Kendra PE. **Insecticidal and attractant activities of**. *Molecules* (2021.0) **26** 2311. DOI: 10.3390/molecules26082311
70. Hasaballah A, Selim T, Tanani M, Nasr E. **Lethality and vitality efficiency of different extracts of Salix safsaf leaves against the house fly,**. *Afr. Entomol.* (2021.0) **29** 479-490. DOI: 10.4001/003.029.0479
71. El-Shemy HA, Aboul-Enein AM, Aboul-Enein KM, Fujita K. **Willow leaves' extracts contain anti-tumor agents effective against three cell types**. *PLOS ONE* (2007.0) **2** e178. DOI: 10.1371/journal.pone.0000178
72. Couto I. **Botanical extracts of the Brazilian savannah affect feeding and oviposition of**. *J. Agric. Sci. (Toronto)* (2019.0) **11** 322-333. DOI: 10.1007/s10343-020-00520-8
73. Mello MO, Silva-Filho MC. **Plant-insect interactions: An evolutionary arms race between two distinct defense mechanisms**. *Braz. J. Plant. Physiol.* (2002.0) **14** 71-81. DOI: 10.1590/S1677-04202002000200001
74. El-Meniawi FA, El-Gayar FH, Rawash IA, Hussein HS. **The olfaction response of the cotton whitefly,**. *Egy. J. Plant Pro Res.* (2013.0) **1** 45-57
75. Rouabhi R, Djebar H, Djebar M. **Toxic effects of combined molecule from novaluron and diflubenzuron on paramecium caudatum**. *Am-Euras. J. Toxicol. Sci* (2009.0) **1** 74-80
76. Couto IFS. **Changes in the biological characteristics of**. *Gesunde Pflanzen* (2020.0) **72** 383-391. DOI: 10.1007/s10343-020-00520-8
77. Procópio TF. *PLOS ONE* (2015.0) **10** e0126612. DOI: 10.1371/journal.pone.0126612
78. Sapindal E, Ong KH, King PJH. **Efficacy of**. *Int. J. Pest Manag.* (2018.0) **64** 39-44. DOI: 10.1080/09670874.2017.1293866
79. Tirelli AA. **Efeito de frações tânicas sobre parâmetros biológicos e nutricionais de Spodoptera frugiperda (Lepidoptera: Noctuidae)**. *Ciência e Agrotecnologia* (2010.0) **34** 1417-1424. DOI: 10.1590/S1413-70542010000600009
80. Divekar PA. **Plant secondary metabolites as defense tools against herbivores for sustainable crop protection**. *Int. J. Mol. Sci.* (2022.0) **23** 2690. DOI: 10.3390/ijms23052690
81. Mattar VT, Borioni JL, Hollmann A, Rodriguez SA. **Insecticidal activity of the essential oil of**. *Pest. Biochem. Physiol.* (2022.0) **185** 105134. DOI: 10.1016/j.pestbp.2022.105134
82. Nakhaie BM, Mikani A, Moharramipour S. **Effect of caffeic acid on feeding, α-amylase and protease activities and allatostatin—A content of Egyptian cotton leafworm,**. *J. Pest. Sci.* (2018.0) **43** 73-78. DOI: 10.1584/jpestics.D17-086
83. Punia A, Chauhan N, Singh R, Kaur S, Sohal S. **Growth disruptive effects of ferulic acid against**. *Allelopath. J.* (2022.0) **55** 79-92. DOI: 10.26651/allelo.j/2022-55-1-1372
84. Gautam S, Samiksha CSS, Arora S, Sohal SK. **Toxic effects of purified phenolic compounds from**. *Toxicon* (2021.0) **203** 22-29. DOI: 10.1016/j.toxicon.2021.09.017
85. Su Q. **Effect of plant secondary metabolites on common cutworm,**. *Entomol. Res.* (2018.0) **48** 18-26. DOI: 10.1111/1748-5967.12238
86. Buxton T. **Insecticidal activities of cinnamic acid esters isolated from**. *Pest Manag. Sci.* (2020.0) **76** 257-267. DOI: 10.1002/ps.5509
87. Lin D-J. **The insecticidal effect of the botanical insecticide chlorogenic acid on**. *Front. Plant Sci.* (2022.0) **13** 1015095. DOI: 10.3389/fpls.2022
|
---
title: Relationships of ferroptosis-related genes with the pathogenesis in polycystic
ovary syndrome
authors:
- Shuang Lin
- Xin Jin
- He Gu
- Fangfang Bi
journal: Frontiers in Medicine
year: 2023
pmcid: PMC9981782
doi: 10.3389/fmed.2023.1120693
license: CC BY 4.0
---
# Relationships of ferroptosis-related genes with the pathogenesis in polycystic ovary syndrome
## Abstract
### Background
Numerous studies have suggested that ferroptosis plays a significant role in the development of polycystic ovary syndrome (PCOS), but the mechanism remains unclear.
### Methods
In this study, we explored the role of ferroptosis-related genes in the pathogenesis of PCOS using a comprehensive bioinformatics method. First, we downloaded several Gene Expression Omnibus (GEO) datasets and combined them into a meta-GEO dataset. Differential expression analysis was performed to screen for significant ferroptosis-related genes between the normal and PCOS samples. The least absolute shrinkage selection operator regression and support vector machine–recursive feature elimination were used to select the best signs to construct a PCOS diagnostic model. Receiver operating characteristic curve analysis and decision curve analysis were applied to test the performance of the model. Finally, a ceRNA network-related ferroptosis gene was constructed.
### Results
*Five* genes, namely, NOX1, ACVR1B, PHF21A, FTL, and GALNT14, were identified from 10 differentially expressed ferroptosis-related genes to construct a PCOS diagnostic model. Finally, a ceRNA network including 117 lncRNAs, 67 miRNAs, and five ferroptosis-related genes was constructed.
### Conclusion
Our study identified five ferroptosis-related genes that may be involved in the pathogenesis of PCOS, which may provide a novel perspective for the clinical diagnosis and treatment of PCOS.
## Background
Polycystic ovary syndrome (PCOS) is a complex disease characterized by reproductive, metabolic, and psychological features. It is usually diagnosed during the patients’ reproductive years and mainly presents with hirsutism, acne, irregular menstruation, and infertility [1]. There are many theoretical hypotheses regarding the etiology of PCOS, and increasing evidence suggests that PCOS may be a complex polygenic disorder influenced by genetic factors, epigenetic variation, and the environment [2]. The prevalence of PCOS ranges from 6 to $20\%$, depending on the populations studied and the definitions used [3, 4]. To date, there is no uniform diagnostic standard or effective treatment for PCOS (5–7). Therefore, exploring the pathogenesis of PCOS can contribute to clinical diagnosis and therapies and improve reproductive outcomes. Numerous studies have shown that iron metabolism is related to endocrine diseases, including PCOS, but the underlying mechanisms remain unclear [8, 9].
Ferroptosis was first described in 2012 as a non-apoptotic, iron-dependent form of cell death, characterized by iron-dependent accumulation of lethal lipid reactive oxygen species (ROS) [10]. Iron is essential for cellular biological processes, including growth, proliferation, and metabolism, among the many functions in the body. Balanced iron absorption, systemic transport, cellular uptake, and storage ensure balanced homeostasis of iron metabolism [11, 12]. In recent years, ferroptosis is associated with the pathogenesis of various diseases, such as diabetes mellitus, renal failure, cardiomyopathy, neurodegeneration, ischemia–reperfusion injury, and cancer (13–18). Therefore, targeting ferroptosis has become a new research area for the design of therapies and disease prevention measures. Recent studies have shown that ferroptosis is involved in the pathogenesis of PCOS. Zhang et al. [ 5] reported that circRHBG inhibits ferroptosis in granulosa cell proliferation of PCOS through the circRHBG/miR-515/SLC7A11 axis. Liu et al. [ 19] revealed that the PCOS model in vivo and granulosa cells subjected to IR have increased ferroptosis levels and that the mechanism of cryptotanshinone in the treatment of PCOS is dependent on its inhibitory effect on cellular ferroptosis. Zhang et al. [ 20] indicated that ferroptosis proteins were associated with reproductive outcomes of POCS patients with infertility and constructed a FerSig risk prognostic model based on the expression of five independent prognostic ferroptosis proteins (G6PD, GPX4, PCBP1, DPP4, and PCBP2). Taken together, ferroptosis plays a key role in the development of PCOS; thus, comprehensive studies on ferroptosis genes in the pathogenesis of PCOS are urgently needed.
As of now, no studies have focused on the mechanism of ferroptosis-related genes in the pathogenesis of PCOS using comprehensive bioinformatics methods. Therefore, we identified differentially expressed ferroptosis-related genes (DEFRGs) in granulosa cells between normal and PCOS women using a public dataset downloaded from the Gene Expression Omnibus (GEO) database. Least absolute shrinkage and selection operator (LASSO) regression and support vector machine–recursive feature elimination (SVM-RFE) were used to select five hub DEFRGs to construct a PCOS diagnostic model, which was successfully verified in our clinical specimens. We then constructed a ceRNA network associated with the five hub DEFRGs. Our results may help illustrate the potential role of ferroptosis in the pathogenesis of PCOS and provide a novel perspective for the clinical diagnosis and treatment of PCOS.
## Data acquisition
Based on the Affymetrix Human Genome U133A Plus 2.0 Array microarray platform, GSE5850, GSE34526, and GSE102293 gene expressions in granulosa cells from patients with PCOS were downloaded from the GEO database. The GSE5850 dataset consisted of six normal and six PCOS women. The GSE34526 dataset included three healthy controls and seven patients with PCOS. The GSE102293 dataset contained six samples, of which two were from patients with PCOS and four were from normal controls. We then combined and normalized the three datasets into a meta-GEO dataset using the “sva” R package and the “gcrma” R package. A flowchart of the study is presented in Figure 1.
**Figure 1:** *Flow chart of this study. PCOS: polycystic ovary syndrome; GEO: Gene Expression Omnibus; DEFRGs: differentially expressed ferroptosis-related genes; GO: Gene Ontology; KEGG; Kyoto Encyclopedia of Genes and Genomes; GSEA: gene set enrichment analysis; LASSO: least absolute shrinkage and selection operator; SVM-RFE: support vector machine–recursive feature elimination.*
## Differential expression analysis
We downloaded 820 ferroptosis-related genes (FRGs) from the FerrDb V2 website1. FRGs contained genes related to driver markers and suppressors. We identified differentially expressed genes (DEGs) between normal and PCOS women using the “limma” package in R. We then crossed the FRGs with DEGs to obtain DEFRGs for further investigation.
## Construction and evaluation of LASSO and SVM-RFE models
The least absolute shrinkage and selection operation and SVM-RFE algorithms were used to obtain key DEFRGs to diagnose PCOS based on the “glmnet” and “e1071” R packages [21]. The LASSO algorithm was used to adjust the optimal value of the penalty parameter (λ) using a 10-fold cross-validation. The SVM-RFE algorithm determines the variable by searching for the lambda with the smallest classification error. Finally, key diagnostic genes for PCOS were identified by overlapping the diagnostic data from the two algorithms. Receiver operating characteristic (ROC) curves were drawn to demonstrate the diagnostic performance of the key DEFRGs, and the area under the ROC curve (AUC) was used to verify the efficiency and accuracy of the key diagnostic DEFRGs [7].
To screen for dependable diagnostic biomarkers related to PCOS, LASSO regression and the SVM-RFE algorithm were performed to evaluate 10 DEFRGs in PCOS. First, the gene expression profiles of the 10 DEFRGs were fit into LASSO regression based on the least squares method. The results revealed that five potential DEFRGs were selected, while the optimal value of lambda was obtained (Figures 5A,B). The SVM-RFE algorithm retained 10 DEFRGs as effective diagnostic biomarkers (Figures 5C,D). Five overlapping DEFRGs (NOX1, ACVR1B, PHF21A, FTL, and GALNT14) were screened as the key DEFRGs for subsequent research (Figure 5E). The AUC value of the overlapping DEFRGs obtained by the SVM-RFE and LASSO regression models was 0.785, indicating accuracy in predicting PCOS (Figure 5F).
**Figure 5:** *(A) Optimal lambda value was selected in the LASSO regression model based on 10-fold cross-validation. (B) LASSO coefficient profiles of the five co-expressional DEFRGs. (C) Line graph shows the cross-validated accuracy based on different numbers of DEFRGs in the SVM-RFE model. (D) Line graph shows the cross-validated error based on different numbers of DEFRGs in the SVM-RFE model. (E) Screening of five DEFRGs using LASSO and SVM-RFE algorithms. (F) Verification of the diagnostic value of the five DEFRGs by using ROC analysis.*
## Tissue collection
From January 2020 to December 2022, ovarian granulosa cells of 50 PCOS patients, 13 normal ovulatory women undergoing in vitro fertilization (IVF) at the ShengJing Hospital of China Medical University were collected. The diagnosis of PCOS met the Rotterdam 2003 diagnostic criteria: oligoovulation and/or anovulation, hyperandrogenism, and polycystic ovaries. Patients with congenital adrenal hyperplasia, Cushing’s syndrome, or androgen-secreting tumors were excluded. Control patients received IVF treatment for tubal disease, but had normal hormone levels, regular menstrual cycles, and normal ovarian morphology. COCs were isolated via ultrasound-guided vaginal puncture and washed in phosphate-buffered saline (PBS). Granulosa cells were selected from COCs. All the participants were under 40 years of age. This study was approved by the Ethics Committee of the ShengJing Hospital of CMU, and informed consent was obtained from all participants (No. 2020PS198K). The baseline information about all patients with PCOS is shown in Supplementary Table 1.
## Quantitative RT-PCR
Total RNA was isolated using TRIzol Reagent (TaKaRa, Shiga, Japan) and then reverse-transcribed into complementary deoxyribonucleic acid (cDNA) synthesis (PrimeScript™ RT Reagent Kit). Real-time qPCR was performed to detect gene expression using 2 × SYBR Green PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA). The 2 − ΔΔCt method was used to calculate the relative gene expression. The sequences of the primers used for RT-qPCR are presented in Supplementary Table 2.
## Construction of a nomogram model
A nomogram model was conducted to facilitate the clinical application using the “rms” package. The “Points” indicate the score of each factor under different conditions, while the “Total Points” refer to the total score of all factors. Calibration curves were used to measure the predictive accuracy, and decision curve analysis (DCA) curves were used to evaluate the clinical value of the nomogram [22].
To facilitate the clinical diagnosis of PCOS using selected DEFRGs (NOX1, ACVR1B, PHF21A, FTL, and GALNT14), a nomogram model was constructed (Figure 6A). Calibration curves revealed that the practical diagnostic rate for PCOS based on the nomogram model was close to the ideal diagnostic rate, suggesting the accuracy of the nomogram model for the diagnosis of PCOS (Figure 6B). DCA showed that the net benefits generated by the nomogram model were at the high-risk threshold from 0.1 to 1.0, suggesting that the nomogram model had a higher clinical application value in the diagnosis of PCOS (Figure 6C). We then evaluated the nomogram model based on the RNA expression data of the five selected DEFRGs in our clinical specimens assessed using RT-PCR. Calibration and DCA curves successfully verified the accuracy and net clinical benefit of the nomogram model (Figures 6D,E). In addition, differences in the expressions of NOX1, ACVR1B, PHF21A, FTL, and GALNT14 were verified in our clinical specimens. The results indicated that all five DEFRGs were more highly expressed in patients with PCOS than in normal controls, which is in line with our predictions (Figure 7).
**Figure 6:** *Construction of the nomogram model. (A) Construction of the nomogram model based on five key DEFRGs (NOX1, ACVR1B, PHF21A, FTL, and GALNT14). (B, D) Calibration curve suggests the accuracy of the nomogram model in the diagnosis of PCOS. (C, E) DCA suggests that the nomogram model is of higher clinical application value in the diagnosis of PCOS.* **Figure 7:** *qRT-PCR validated five key DEFRGs differential expression between normal controls and patients with PCOS. (A) ACVR1B; (B) FTL; (C) GALNT14; (D) NOX1; and (E) PHF21A.*
## Construction of the ceRNA network
To construct a lncRNA–miRNA–mRNA regulatory network based on the five key diagnostic DEFRGs, miRNA–mRNA and miRNA–lncRNAs interactions were predicted using miRWalk2 and miRDB3, respectively. The predicted miRNAs were intersected, and the lncRNA–miRNA–mRNA network was visualized using the Cytoscape software (version 3.8.2) [23].
A lncRNA–miRNA–mRNA network was constructed based on the five key DEFRGs. First, the microRNAs (miRNAs) interacting with the five key DEFRGs were obtained based on the miRWalk database.4 The miRNAs that interacted with lncRNAs were acquired from the miRNA target prediction database.5 The predicted miRNAs were intersected, and the lncRNA–miRNA–mRNA network included five key DEFRGs, 117 lncRNAs, and 67 miRNAs visualized using Cytoscape software (Figure 9).
**Figure 9:** *Representative lncRNA–miRNA–mRNA network was obtained by Cytoscape software. Red, green, and blue represent mRNA, miRNA, and lncRNA, respectively.*
## Functional enrichment analysis
To further explore the mechanism of DEFRGs in PCOS, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed using the “clusterProfiler” package in R. Gene set enrichment analysis (GSEA) was conducted using c2.cp.kegg.symbols.gmt and c5.go.symbols.gmt [24].
## Statistical analysis
All statistical analyses were performed using the R software (version 4.0.2). We mapped the chromosomal positions of the key DEFRGs using the “RCircos” package in R and calculated the correlation coefficients between the key DEFRGs using the Spearman correlation analysis. The Wilcoxon rank-sum test was used to compare differences between the groups. Statistical significance was set at a two-tailed p-value of <0.05.
## Identification of differentially expressed ferroptosis-related genes in PCOS
A total of 401 DEGs were identified between normal and PCOS women by the “limma” package according to the screening criteria of $p \leq 0.05$ and log |FC| > 1. We crossed 820 ferroptosis-related genes with 401 DEGs and obtained 10 differentially expressed ferroptosis-related genes (DEFRGs) (NOX1, ACVR1B, PHF21A, PRKCA, IFNA14, PTPN6, FTL, MTF1, PARP15, and GALNT14), which were visualized using a Venn diagram (Figure 2A). The circle diagram displays the chromosomal positions of the 10 DEFRGs (Figure 2B). Finally, the heat map and histogram were used to show the differential expression levels of the 10 DEFRGs between normal controls and patients with PCOS. We found that the expression levels of the 10 DEFRGs were upregulated in PCOS samples compared with those in normal controls (Figures 2C,D). The correlation heat map indicated that the 10 DEFRGs had a strong positive correlation with each other (Figure 3).
**Figure 2:** *Landscape of 10 DEFRGs in PCOS. (A) Identification of 10 DEFRGs. (B) The chromosomal positions of the 10 DEFRGs. (C) Heat map of the expression of 10 DEFRGs between normal controls and patients with PCOS. (D) Histogram of the expression of 10 DEFRGs between normal controls and patients with PCOS.* **Figure 3:** *Spearman’s correlation analysis of 10 DEFRGs.*
## Go and KEGG enrichment analysis
GO analysis revealed that 10 DEFRGs mainly enriched in platelet aggregation (GO:0070527), regulation of hemopoiesis (GO:1903706), myeloid cell differentiation (GO:0030099), positive regulation of myeloid cell differentiation (GO:0034109) and homotypic cell-cell adhesion (GO:0034109) (Figure 4A). KEGG analysis showed that the pathways enriched by 10 DEFRGs contained natural killer cell-mediated cytotoxicity, lipid and atherosclerosis, and the AGE-RAGE signaling pathway in diabetic complications (Figure 4B).
**Figure 4:** *Functional enrichment of 10 DEFRGs. (A) GO enrichment analysis. (B). KEGG enrichment analysis.*
## Gene set enrichment analysis
We then explored the specific GO function and signaling pathways enriched by the five DEFRGs and the potential mechanisms of the five DEFRGs in the pathogenesis of PCOS. The GSEA results revealed that the main enriched signaling pathways for high NOX1 expression were ALLOGRAFT_REJECTION, ENDOCYTOSIS, FC_GAMMA_R_MEDIATED_PHAGOCYTOSIS, LEISHMANIA_INFECTION, LYSOSOME, and SYSTEMIC_LUPUS_ERYTHEMATOSUS (Figure 8A). The main enriched pathways for high ACVR1B expression were ALLOGRAFT_REJECTION, ANTIGEN_PROCESSING_AND_PRESENTATION, AUTOIMMUNE_THYROID_DISEASE, GRAFT_VERSUS_HOST_DISEASE, LEISHMANIA_INFECTION, and SYSTEMIC_LUPUS_ERYTHEMATOSUS (Figure 8B). The main enriched pathways for high PHF21A expression were ALLOGRAFT_REJECTION, AUTOIMMUNE_THYROID_DISEASE, GRAFT_VERSUS_HOST_DISEASE, LEISHMANIA_INFECTION, SYSTEMIC_LUPUS_ERYTHEMATOSUS, and TYPE_I_DIABETES_MELLITUS (Figure 8C). The main enriched pathways for high FTL expression were ALLOGRAFT_REJECTION, AUTOIMMUNE_THYROID_DISEASE, LEISHMANIA_INFECTION, LYSOSOME, SYSTEMIC_LUPUS_ERYTHEMATOSUS, and TYPE_I_DIABETES_MELLITUS (Figure 8D). Finally, the main enriched pathways for high SNW1 expression were ALLOGRAFT_REJECTION, B_CELL_RECEPTOR_SIGNALING_PATHWAY, INSULIN_SIGNALING_PATHWAY, LEISHMANIA_INFECTION, LYSOSOME, and SYSTEMIC_LUPUS_ERYTHEMATOSUS (Figure 8E).
**Figure 8:** *GSEA enrichment analysis showing signaling pathways enriched by five DEFRGs. (A) NOX1; (B) ACVR1B; (C) PHF21A; (D) FTL; and (E) GALNT14.*
## Discussion
In this study, we acquired 10 upregulated DEFRGs based on GEO datasets to explore their roles in the pathogenesis of PCOS using a series of differential expression analyses. GO and KEGG enrichment analyses of 10 DEFRGs indicated that these genes are involved in some pathways associated with immunity and ferroptosis, suggesting that immunity and ferroptosis may be involved in the pathogenesis of PCOS. In addition, five key DEFRGs involved in the diagnosis of PCOS were screened using LASSO regression and the SVM-RFE algorithm. Finally, we constructed a lncRNA–miRNA–mRNA network associated with the five DEFRGs. Increasing evidence indicates that the lncRNA–miRNA–mRNA network plays a critical role in the development of PCOS. For example, Liu G et al. [ 25] revealed that the lncRNA PVT1/MicroRNA-17-5p/PTEN network was related to the secretion of E2 and P4, proliferation, and apoptosis of granulosa cells in PCOS. Guo et al. [ 26] illustrated that HOTAIRM1 could sponge miR-433-5p to promote PIK3CD expression, thereby regulating the growth and apoptosis of granulose cells in PCOS. However, the ceRNA network found in our study has not yet been studied. In summary, our study illustrates the potential role of ferroptosis in the pathogenesis of PCOS from the perspective of comprehensive bioinformatics analysis and provides a novel perspective for the clinical diagnosis and treatment of PCOS.
In the present study, five key DEFRGs (NOX1, ACVR1B, PHF21A, FTL, and GALNT14) were identified as the most significant genes related to PCOS pathogenesis, and an accurate and clinically valuable nomogram model was constructed. Ferroptosis is a type of cell death caused by the accumulation of lipid peroxidation products and lethal ROS from iron metabolism [27, 28]. The mitochondrial respiratory chain and NADPH oxidase of the NADPH oxidase (NOX) family are major sources of reactive oxygen species (ROS) in human neuronal cells, cardiomyocytes, and keratinocytes (29–31). NOX1, a member of the NOX family, promotes ROS release and ferroptosis [32]. Zhang et al. [ 33] found that ferric ammonium citrate (FAC) increased the ferric content in a human granulosa-like tumor cell line (KGN) by activating the transferrin receptor (TFRC). Iron uptake then mediates the activation of NOX1 signaling, which induces the release of ROS and mitochondrial damage. Therefore, the inhibitory effects of TFRC/NOX1 signaling on follicular genesis may be a potential treatment for PCOS. polypeptide N-acetylgalactosaminyltransferase 14 (GALNT14) is a member of the acetylgalactosaminyltransferase family that can initiate protein O-glycosylation by transferring the GalNAc residue of UDP-GalNAc to the hydroxyl group of Ser or Thr [34]. Li et al. [ 35] indicated that GALNT14 regulates ferroptosis and apoptosis in ovarian cancer by targeting the EGFR/mTOR pathway. Ferritin is the only protein capable of storing substantial amounts of iron, and it plays a key role in regulating cellular iron metabolism. The heavy (H) and light (L) chain subunits of ferritin (FTH and FTL, respectively) are responsible for intracellular iron storage [36]. Therefore, FTH1 and FTL levels are positively correlated with ferroptosis and can function as biomarkers of ferroptosis. ACVR1B, also known as ALK-4, acts as a transducer of activin-like ligands that belong to the growth and differentiation factors of the TGF-β superfamily of signaling proteins [37]. Kota Fujiki et al. [ 38] reported that cadmium-and-erabine-induced cell death, including ferroptosis in renal proximal tubular epithelial cells, can be inhibited by blocking the ALK$\frac{4}{5}$ signaling pathway, suggesting that ALK-4 is related to ferroptosis. PHD finger protein 21A (PHF21A) is also known as BHC80. To date, no relevant studies have reported the mechanism of PHF21A involvement in iron deficiency and PCOS. We selected five key DEFRGs related to PCOS pathogenesis, which will be further elucidated in in vitro and in vivo experiments.
This study has some limitations. First, our research results were based on bioinformatics analysis. Further basic and clinical experiments are required to verify these results. Second, owing to the limitations of the data in the public database, the sample size included in our study was not large enough, and the research results may deviate from the real situation.
## Conclusion
In the present study, we identified several ferroptosis-related genes that are strongly associated with PCOS pathogenesis, which may provide a novel perspective for the clinical diagnosis and treatment of PCOS. However, further studies are necessary to explore the mechanisms of ferroptosis and its role in PCOS pathogenesis.
## 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 study was approved by the Ethics Committee of the ShengJing Hospital of CMU, and informed consent was obtained from all participants. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
SL, XJ, HG, and FB conceived and designed the study, developed the methodology, analyzed and interpreted the data, wrote, reviewed, and revised the manuscript. All authors contributed to the article and approved the submitted version.
## Funding
This work was supported by 345 Talent Project of Shengjing Hospital of China Medical University (No. M0695); Shenyang Young and Middle-aged Science and Technology Innovation Talents Support Program, Project (No. RC210436); Joint Program of Applied Basic Research of Liaoning Province, Project; Key Laboratory of Reproductive and Genetic Medicine (China Medical University); National Health Commission.
## 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.2023.1120693/full#supplementary-material
## References
1. Visser JA. **The importance of metabolic dysfunction in polycystic ovary syndrome**. *Nat Rev Endocrinol* (2021) **17** 77-8. DOI: 10.1038/s41574-020-00456-z
2. Walters KA, Gilchrist RB, Ledger WL, Teede HJ, Handelsman DJ, Campbell RE. **New perspectives on the pathogenesis of PCOS: neuroendocrine origins**. *Trends Endocrinol Metab* (2018) **29** 841-52. DOI: 10.1016/j.tem.2018.08.005
3. Balen AH, Morley LC, Misso M, Franks S, Legro RS, Wijeyaratne CN. **The management of anovulatory infertility in women with polycystic ovary syndrome: an analysis of the evidence to support the development of global WHO guidance**. *Hum Reprod Update* (2016) **22** 687-708. DOI: 10.1093/humupd/dmw025
4. Escobar-Morreale HF. **Polycystic ovary syndrome: definition, aetiology, diagnosis and treatment**. *Nat Rev Endocrinol* (2018) **14** 270-84. DOI: 10.1038/nrendo.2018.24
5. Zhang D, Yi S, Cai B, Wang Z, Chen M, Zheng Z. **Involvement of ferroptosis in the granulosa cells proliferation of PCOS through the circRHBG/miR-515/SLC7A11 axis**. *Ann Transl Med* (2021) **9** 1348. DOI: 10.21037/atm-21-4174
6. Christ JP, Falcone T. **Bariatric surgery improves Hyperandrogenism, menstrual irregularities, and metabolic dysfunction among women with polycystic ovary syndrome (PCOS)**. *Obes Surg* (2018) **28** 2171-7. DOI: 10.1007/s11695-018-3155-6
7. Tagliaferri V, Romualdi D, Scarinci E, Cicco S, Florio CD, Immediata V. **Melatonin treatment may be able to restore menstrual Cyclicity in women with PCOS: a pilot study**. *Reprod Sci* (2018) **25** 269-75. DOI: 10.1177/1933719117711262
8. Yin J, Hong X, Ma J, Bu Y, Liu R. **Serum trace elements in patients with polycystic ovary syndrome: a systematic review and meta-analysis**. *Front Endocrinol (Lausanne)* (2020) **11** 572384. DOI: 10.3389/fendo.2020.572384
9. Escobar-Morreale HF. **Iron metabolism and the polycystic ovary syndrome**. *Trends Endocrinol Metab* (2012) **23** 509-15. DOI: 10.1016/j.tem.2012.04.003
10. Dixon SJ, Lemberg KM, Lamprecht MR, Skouta R, Zaitsev EM, Gleason CE. **Ferroptosis: an iron-dependent form of nonapoptotic cell death**. *Cells* (2012) **149** 1060-72. DOI: 10.1016/j.cell.2012.03.042
11. Yan HF, Zou T, Tuo QZ, Xu S, Li H, Belaidi AA. **Ferroptosis: mechanisms and links with diseases**. *Signal Transduct Target Ther* (2021) **6** 49. DOI: 10.1038/s41392-020-00428-9
12. Yao X, Li W, Fang D, Xiao C, Wu X, Li M. **Emerging roles of energy metabolism in Ferroptosis regulation of tumor cells**. *Adv Sci (Weinh)* (2021) **8** e2100997. DOI: 10.1002/advs.202100997
13. Mou Y, Wang J, Wu J, He D, Zhang C, Duan C. **Ferroptosis, a new form of cell death: opportunities and challenges in cancer**. *J Hematol Oncol* (2019) **12** 34. DOI: 10.1186/s13045-019-0720-y
14. Tang S, Xiao X. **Ferroptosis and kidney diseases**. *Int Urol Nephrol* (2020) **52** 497-503. DOI: 10.1007/s11255-019-02335-7
15. Mahoney-Sánchez L, Bouchaoui H, Ayton S, Devos D, Duce JA, Devedjian JC. **Ferroptosis and its potential role in the physiopathology of Parkinson’s disease**. *Prog Neurobiol* (2021) **196** 101890. DOI: 10.1016/j.pneurobio.2020.101890
16. Elgendy SM, Alyammahi SK, Alhamad DW, Abdin SM, Omar HA. **Ferroptosis: an emerging approach for targeting cancer stem cells and drug resistance**. *Crit Rev Oncol Hematol* (2020) **155** 103095. DOI: 10.1016/j.critrevonc.2020.103095
17. Li M, Xin S, Gu R, Zheng L, Hu J, Zhang R. **Novel diagnostic biomarkers related to oxidative stress and macrophage Ferroptosis in atherosclerosis**. *Oxidative Med Cell Longev* (2022) **2022** 1-18. DOI: 10.1155/2022/8917947
18. Stancic A, Velickovic K, Markelic M, Grigorov I, Saksida T, Savic N. **Involvement of Ferroptosis in diabetes-induced liver pathology**. *Int J Mol Sci* (2022) **23** 9309. DOI: 10.3390/ijms23169309
19. Liu H, Xie J, Fan L, Xia Y, Peng X, Zhou J. **Cryptotanshinone protects against PCOS-induced damage of ovarian tissue via regulating oxidative stress, mitochondrial membrane potential, inflammation, and apoptosis via regulating Ferroptosis**. *Oxidative Med Cell Longev* (2022) **2022** 1-21. DOI: 10.1155/2022/8011850
20. Zhang J, Ding N, Xin W, Yang X, Wang F. **Quantitative proteomics reveals that a prognostic signature of the endometrium of the polycystic ovary syndrome women based on Ferroptosis proteins**. *Front Endocrinol (Lausanne)* (2022) **13** 871945. DOI: 10.3389/fendo.2022.871945
21. Chen S, Jiang Y, Qi X, Song P, Tang L, Liu H. **Bioinformatics analysis to obtain critical genes regulated in subcutaneous adipose tissue after bariatric surgery**. *Adipocytes* (2022) **11** 550-61. DOI: 10.1080/21623945.2022.2115212
22. Jin X, Wang J, Ge L, Hu Q. **Identification of immune-related biomarkers for sciatica in peripheral blood**. *Front Genet* (2021) **12** 781945. DOI: 10.3389/fgene.2021.781945
23. Pan X, Bi F. **A potential immune-related long non-coding RNA prognostic signature for ovarian cancer**. *Front Genet* (2021) **12** 694009. DOI: 10.3389/fgene.2021.694009
24. Huang C, Chen L, Zhang Y, Wang L, Zheng W, Peng F. **Predicting AURKA as a novel therapeutic target for NPC: a comprehensive analysis based on bioinformatics and validation**. *Front Genet* (2022) **13** 926546. DOI: 10.3389/fgene.2022.926546
25. Liu G, Liu S, Xing G, Wang F. **lncRNA PVT1/MicroRNA-17-5p/PTEN Axis regulates secretion of E2 and P4, proliferation, and apoptosis of ovarian Granulosa cells in PCOS**. *Mol Ther Nucleic Acids* (2020) **20** 205-16. DOI: 10.1016/j.omtn.2020.02.007
26. Guo H, Li T, Sun X. **LncRNA HOTAIRM1, miR-433-5p and PIK3CD function as a ceRNA network to exacerbate the development of PCOS**. *J Ovarian Res* (2021) **14** 19. DOI: 10.1186/s13048-020-00742-4
27. Li J, Cao F, Yin HL, Huang ZJ, Lin ZT, Mao N. **Ferroptosis: past, present and future**. *Cell Death Dis* (2020) **11** 88. DOI: 10.1038/s41419-020-2298-2
28. Song J, Liu T, Yin Y, Zhao W, Lin Z, Yin Y. **The deubiquitinase OTUD1 enhances iron transport and potentiates host antitumor immunity**. *EMBO Rep* (2021) **22** e51162. DOI: 10.15252/embr.202051162
29. Cadenas S. **ROS and redox signaling in myocardial ischemia-reperfusion injury and cardioprotection**. *Free Radic Biol Med* (2018) **117** 76-89. DOI: 10.1016/j.freeradbiomed.2018.01.024
30. Olguín-Albuerne M, Morán J. **Redox signaling mechanisms in nervous system development**. *Antioxid Redox Signal* (2018) **28** 1603-25. DOI: 10.1089/ars.2017.7284
31. Emmert H, Fonfara M, Rodriguez E, Weidinger S. **NADPH oxidase inhibition rescues keratinocytes from elevated oxidative stress in a 2D atopic dermatitis and psoriasis model**. *Exp Dermatol* (2020) **29** 749-58. DOI: 10.1111/exd.14148
32. Kain HS, Glennon EKK, Vijayan K, Arang N, Douglass AN, Fortin CL. **Liver stage malaria infection is controlled by host regulators of lipid peroxidation**. *Cell Death Differ* (2020) **27** 44-54. DOI: 10.1038/s41418-019-0338-1
33. Zhang L, Wang F, Li D, Yan Y, Wang H. **Transferrin receptor-mediated reactive oxygen species promotes ferroptosis of KGN cells via regulating NADPH oxidase 1/PTEN induced kinase 1/acyl-CoA synthetase long chain family member 4 signaling**. *Bioengineered* (2021) **12** 4983-94. DOI: 10.1080/21655979.2021.1956403
34. Bennett EP, Mandel U, Clausen H, Gerken TA, Fritz TA. **Control of Mucin-type O-glycosylation—a classification of the polypeptide GalNAc-transferase gene family**. *Glycobiology* (2012) **22** 736-56. DOI: 10.1093/glycob/cwr182
35. Li HW, Liu MB, Jiang X, Song T, Feng SX, Wu JY. **GALNT14 regulates ferroptosis and apoptosis of ovarian cancer through the EGFR/mTOR pathway**. *Fut. Oncol* (2022) **18** 149-61. DOI: 10.2217/fon-2021-0883
36. Honarmand Ebrahimi K, Hagedoorn PL, Hagen WR. **Unity in the biochemistry of the iron-storage proteins ferritin and bacterioferritin**. *Chem Rev* (2015) **115** 295-326. DOI: 10.1021/cr5004908
37. Boueiz A, Pham B, Chase R, Lamb A, Lee S, Naing ZZC. **Integrative genomics analysis identifies ACVR1B as a candidate causal gene of emphysema distribution**. *Am J Respir Cell Mol Biol* (2019) **60** 388-98. DOI: 10.1165/rcmb.2018-0110OC
38. Fujiki K, Inamura H, Sugaya T, Matsuoka M. **Blockade of ALK4/5 signaling suppresses cadmium-and erastin-induced cell death in renal proximal tubular epithelial cells via distinct signaling mechanisms**. *Cell Death Differ* (2019) **26** 2371-85. DOI: 10.1038/s41418-019-0307-8
|
---
title: Transfection of clMagR/clCry4 imparts MR-T2 imaging contrast properties to
living organisms (E. coli) in the presence of Fe3+ by endogenous formation of iron
oxide nanoparticles
authors:
- Nuan Li
- Le Xue
- Xiaoli Mai
- Peng Wang
- Chenzhuo Zhu
- Xiaofeng Han
- Yuanyuan Xie
- Bin Wang
- Yuqing Ge
- Yewei Zhang
- Jianfei Sun
journal: Frontiers in Molecular Biosciences
year: 2023
pmcid: PMC9981785
doi: 10.3389/fmolb.2023.1119356
license: CC BY 4.0
---
# Transfection of clMagR/clCry4 imparts MR-T2 imaging contrast properties to living organisms (E. coli) in the presence of Fe3+ by endogenous formation of iron oxide nanoparticles
## Abstract
Rapid development of medical imaging, such as cellular tracking, has increased the demand for “live” contrast agents. This study provides the first experimental evidence demonstrating that transfection of the clMagR/clCry4 gene can impart magnetic resonance imaging (MRI) T2-contrast properties to living prokaryotic *Escherichia coli* (E. coli) in the presence of Fe3+ through the endogenous formation of iron oxide nanoparticles. The transfected clMagR/clCry4 gene markedly promoted uptake of exogenous iron by E. coli, achieving an intracellular co-precipitation condition and formation of iron oxide nanoparticles. This study will stimulate further exploration of the biological applications of clMagR/clCry4 in imaging studies.
## Introduction
Contrast-enhanced magnetic resonance imaging (MRI) has become an indispensable tool in medical imaging (Rogers et al., 2006; Park et al., 2017). Gadolinium (Gd)-based small molecules and iron oxide nanoparticles have been mainly used as MRI-T1 and -T2 contrast agents, respectively, to provide high contrast sensitivity against background signals and have been approved by the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMEA) (Park et al., 2017; Wahsner et al., 2019). Nevertheless, there are still some challenges facing these contrast agents in practical applications, especially for long-term tracking in vivo. The main issue lies in their biosafety and the exocytosis of inorganic substances over a long period (Wahsner et al., 2019). For example, Gd was recently reported to be somewhat toxic to the liver and kidney (Calvin et al., 2010), and iron (Fe) overload has also been proposed to be involved in degenerative diseases (Zecca et al., 2004; Stephenson et al., 2014). Thus, it has inspired a new wave to develop new contrast agents, where control of iron accumulation in cells could work as a means to alter longitudinal or transverse relaxation times, generating contrast, probably derived endogenously from the organism itself.
Recently, protein-based MRI contrast agents have attracted increasing attention, such as transferrin and ferritin (Jutz et al., 2015; Schilling et al., 2017). Transferrin and ferritin are critical proteins that regulate iron metabolism in vivo, which are capable of binding Fe to exhibit paramagnetic properties (Yang et al., 2016). Although the two proteins, especially ferritin, have relatively large magnetic moments and influence the transverse relaxation of the proton, the effect appears to be too weak to be used directly as contrast agents for MRI in vivo (Liu and Theil, 2005; Jutz et al., 2015). Generally, ferritin is used as a natural reactor in vitro to transform hydrated paramagnetic ferric oxide into superparamagnetic iron oxide nanoparticles (Liu and Theil, 2005; Turano et al., 2010). These transformed protein molecules have been proposed as a new type of nanoparticle and partly improve the biosafety of contrast agents; however, they are essentially an exogenous substance.
Xie et al. discovered a novel magnetic protein biocompass in Columba livia, which integrated both magnetoreceptor (MagR) and type IV cryptochrome (clCry4) (Qin et al., 2016). MagR is the homologue of the iron–sulfur cluster assembly protein (A-type ISC protein, IscA). The IscA protein is an iron chaperon that can bind to intracellular iron to form Fe–S clusters for electron transfer (Ding and Clark, 2004; Holm and Lo, 2016). The Cry4 protein is considered an electron donor excited by light that can form long-lived radical pairs to activate downstream pathways (Maeda et al., 2012; Zoltowski et al., 2019; Xu et al., 2021). Because clMagR/clCry4 protein is involved with the process of electron transfer, we hypothesized it would be suitable for MRI contrast as a “live” agent.
We investigated the MRI contrast performance of clMagR/clCry4 gene in a living organism. Unlike the strategy used for ferritin, the clMagR/clCry4 gene was simply transfected into prokaryotic E. coli, rather than administering clMagR/clCry4 protein molecules. The transfected E. coli showed significant T2 contrast on MRI when cultured in an iron-supplemented medium, while the protein itself was unable to show this effect. The transfection of the clMagR/clCry4 gene led to the formation of iron oxide nanoparticles within E. coli, which could mediate the alteration of the MRI transverse relaxation rate.
## MRI analysis
The MRI study was performed using the 7T BioSpec $\frac{70}{20}$ USR system (Bruker Biospin; Ettlingen) with ParaVision 6.0.1. software. The MRI-T1 images were acquired using a rapid acquisition sequence with relaxation enhancement (RARE) using the following parameters: matrix = 256 × 256, flip angle (FA) = 90°, field of view (FOV) = 8.0 × 6.03 cm, slice thickness = 2 mm, echo time (TE) = 8.87 ms, and repetition time (TR) = 400 ms. The MRI-T2 images were acquired using the TurboRARE sequence using the following parameters: matrix = 256 × 256, FA = 90°, FOV = 8.0 × 6.03 cm, slice thickness = 2 mm, TE = 80 ms, and TR = 2,200 ms. For MRI-T1 relaxation time map imaging (T1 mapping), we used the RARE sequence with variable TR with the following parameters: matrix = 256 × 256, FA = 90°, FOV = 8.0 × 6.03 cm, slice thickness = 2 mm, TE = 7.17 ms, echo spacing = 7.17 ms, averages = 2, repetition = 1, and TR = 190, 200, 300, and 400 ms. For MRI-T2 relaxation time map imaging (T2 mapping), we used a multislice multiecho (MSME) sequence with variable TE with the following parameters: matrix = 256 × 256, FA = 90°, FOV = 8.0 × 6.03 cm, slice thickness = 2 mm, TE = 9–225 ms with an increment of 9 ms, echo spacing = 9.0 ms, echo times = 25, averages = 1, repetition = 1, and TR = 3,000 ms. The regions of interest (ROIs) were drawn in the same plane of each sample after scanning, the T1 and T2 relaxation times were calculated using ParaVision software, and the rates at which the signal decayed were defined as R1 and R2.
The Python language (Python Software Foundation, Version 3.8.0) was used to batch calculate the mean gray values and to quantify the signal density of the MRI images (Sindhulakshmi et al., 2014; Gadi et al., 2020). In detail, original DICOM images were processed with a Gaussian filter. On the basis of the maximum between-class variance method (OTSU), the image data were classified into targets and backgrounds. The OpenCV (OpenSource Computer Vision Library, Version 4.4.0) contour search algorithm was adopted to extract the true coordinate parameters of each target area and to discard false values. Furthermore, the aforementioned coordinates were processed to obtain the mean gray value of each target area pixel group on the original image, all pixels were sorted according to their mean gray values, and the average value and the quartile value were recorded. The heat map of the original gray value was drawn using Numerical Python (NumPy), Pandas, Matplotlib, and OpenCV. In this study, to facilitate the statistics of MRI dark contrast, the mean gray values mentioned included 255 (the maximum gray value of the 8-bit gray image) minus the original gray values.
The bacteria were collected by centrifugation (2,000 rpm, 5 min, 4°C), the pellets were washed three times, and resuspended in phosphate buffer solution (PBS, pH 7.4). The density of bacteria cells was evaluated by optical density (OD) measurement at 600 nm (A600 units/mL). Protein concentrations were determined using the bicinchoninic acid (BCA) assay with each sample in triplicate, using the microplate method according to the protocol recommended by the manufacturer using commercially available BSA as the calibration solution. Data were plotted in the graph form, and a linear trendline was fit to obtain a standard BSA protein curve. Data were acquired and analyzed using the SpectraMax Plus microplate reader and SoftMax Pro software (Molecular Devices). The MRI samples were collected into polyethylene centrifugation tubes for scanning.
## Intracellular iron quantification
The bacteria were collected after MRI scanning, lysed in nitric acid, and then analyzed by inductive-coupled plasma mass spectrometry (ICP-MS, PerkinElmer NexION 2000). For calibration, the reference solutions containing different concentrations of iron (i.e., 0, 20, 50, 100, 200, and 500 μg/L in Milli-Q water, 18.2 MΩ cm) as internal standards were prepared. A reference solution was used at the beginning, middle, and end of the measurements as a quality control. Acidity of experimental samples and reference solutions were controlled at $5\%$. Experimental samples were filtered through a 0.22-μm hydrophilic syringe filter (Sartorius Stedim Biotech, Germany) to remove solid impurities. Data corresponding to iron content were determined by ICP-MS. The iron standard curve, determined from calibration solutions with known concentrations (μg/L), was used to calculate the iron content in bacteria.
## Detection of intracellular pH
Intracellular pH was measured using the fluorescent pH indicator 2,7-bicarboxyethyl-5,6-carboxyfluorescein-acet-oxymethylester (BCECF-AM) according to the manufacturer’s protocol (Burgess and Han, 2010; Chakraborty et al., 2017). Bacterial suspensions (OD 2.0) were incubated with 20 μM BCECF-AM at 37 °C for 60 min. After loading, the cells were washed three times with PBS buffer and remained in the same solution. A pH calibration curve was constructed using BCECF-AM with a pH calibration buffer kit containing a pH range of 4.5, 5.5, 6.5, and 7.5, and valinomycin (10 μM) and nigericin (10 μM), which equilibrated the intracellular and extracellular pH of bacteria. Intracellular pH was recorded by determining the fluorescence ratio (F490 nm/F440 nm) of the emission wavelength at 535 nm for excitation wavelengths of 490 and 440 nm using a multimode microplate reader (Tecan Infinite M200). Bacterial photographs were detected using inverted fluorescence microscopy (Nikon Microsystems).
## Ultrastructural observation
The bacteria were collected by centrifugation, pellets washed with PBS and then fixed with $2.5\%$ glutaraldehyde overnight at 4°C. The samples were fixed with $1\%$ osmium tetroxide for 1 h, dehydrated with a series of ethanol concentrations in Milli-Q water (i.e., 35, 50, 60, 70, 80, 90, 95, and $100\%$ ethanol for 10 min in each step) and then embedded in epoxy resin and polymerized at 60°C overnight. Ultrathin (about 90 nm) sections were cut with a diamond knife in an ultramicrotome (Leica, EM UC7) and collected onto carbon-coated copper grids, stained with uranyl acetate and lead citrate, and then examined by transmission electron microscopy (TEM) (Hitachi H-600-4) at an operating voltage of 120 KV.
## Biomaterial extraction and electronic microscope analysis
To extract electron-dense granules, clMagR/clCry4-transfected bacteria (under exogenous iron supply condition) were harvested, washed, and resuspended in PBS and then fragmented using an ultrasonic cell disruptor (Scientz-II D, amplitude $15\%$, pulse 5 s on and 2 s off). Subsequently, electron-dense granules were separated from cell debris using gradient centrifugation, washed and resuspended in Milli-Q water to remove residual PBS. The biosynthesis of materials was isolated from the suspension using the MACS LD separation column, the QuadroMACS Separator, and the MACS MultiStand (Miltenyi Biotec Inc) according to the manufacturer’s instructions. The resulting materials were placed on carbon-coated copper grids and dried. Bright-field scanning transmission electron microscopy (BF-STEM), dark-field STEM (DF-STEM), high-angle annular dark-field STEM (HAADF-STEM), energy-dispersive X-ray spectroscopy (XEDS) element mapping, selected area electron diffraction (SAED), and high-resolution TEM (HRTEM) were carried out on FEI Talos F200X TEM at an operating voltage of 200 KV. All micrographs were analyzed using DigitalMicrograph software (Gatan Microscopy Suite, Version 3.42.3048.0) and ICDD PDF-4 + 2009 software (The International Centre for Diffraction Data, ICDD; Powder Diffraction File, PDF). The fast Fourier transform (FFT) was performed using ImageJ, v1.53a software.
## Statistical analysis
GraphPad Prism software (v8.3.1 [332], La Jolla, CA, United States) was used for graph preparation. Data were presented as mean ± standard deviation (SD). Statistical differences were analyzed using the unpaired Student’s t-test. For mean gray values, statistical differences were conducted using IBM SPSS Statistics 25 software packages and analyzed using two-way analysis of variance (ANOVA) followed by the Bonferroni and Tukey’s honest significant differences (HSD) post-hoc test. The differences were considered statistically significant when the p-value was less than 0.05. * $p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$; no significance (ns), $p \leq 0.05.$
## Results and discussion
Transfection of the clMagR/clCry4 gene into E. coli is shown schematically in Figure 1A. Similar to that of our previous report (Xue et al., 2020), transfection was shown to be successful based on the sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) pattern of the expressed clMagR/clCry4 protein (Figure 1B). Because the d electrons in Fe atom have been demonstrated to contribute to the net spins of the Fe-loaded IscA monomer (Beinert et al., 1997; Xiao et al., 2020), the characteristic signal at g' ≈ 4.3 (g-factor) in the electronic spin resonance (ESR) spectrum proved the presence of the protein–Fe(III) complex (Figure 1C) (Sun and Chasteen, 1994; Wajnberg et al., 2018). Furthermore, the broad peak of the ESR also indicated the heterogeneity of the protein clMagR/clCry4 protein expressed inside E. coli. The bacteria were resuspended in PBS buffer solution for cell suspensions (Supplementary Figure S1). Then, the clMagR/clCry4-transfected E. coli were tested by qualitative MRI-T1 and -T2 scans. As shown from the images, there were no significant differences for either mode nor were there any differences in the relaxation rates R1 and R2 (Figure 1D). Furthermore, MRI-T1 and -T2 mapping modes were used to quantitatively evaluate the imaging contrast of transfected E. coli. The mean gray values acquired from the mapping images were measured using the Python language and two-way ANOVA statistics, followed by the Bonferroni and Tukey’s HSD post-hoc test. Both the images and the statistical values of the mean gray areas also showed minimal differences (Figures 1E, F). In addition, only the clMagR-transfected E. coli also showed a little imaging contrast with the control (Supplementary Figure S2). We tested the MRI contrast properties for the purified clMagR/clCry4 protein. The protein itself showed a little MRI contrast effect (Supplementary Figure S3). Based on these cases, the transfection of the clMagR/clCry4 gene was unable to influence MRI spin relaxation signals. Due to its presence in a living organism, the clMagR/clCry4 protein could perceive a magnetic field; thus, we hypothesized that the magnetism of the clMagR/clCry4 protein would associate with the electron transfer, that is, binding with Fe. Because there was insufficient Fe in the medium for E. coli, the clMagR/clCry4 complex was incapable of producing magnetic behaviors.
**FIGURE 1:** *Bacteria construction and MRI analysis. (A) Schematic illustration of the prokaryotic expression system and bacterial culture. clMagR/clCry4 is transfected into a commonly used bacterial model (as heterologous host), the E. coli BL21 (DE3) strain to investigate MRI susceptibility. (B) SDS-PAGE analysis. Arrows point to clMagR and clCry4. PageRuler prestained protein ladder was used to indicate the apparent molecular mass standards (marker, kDa). (C) ESR spectrum of the clMagR/clCry4 protein. Signal at g' ≈ 4.3 in spectrum attributed to protein-Fe(III) complex. (D) MRI-T1 and -T2 images show the signal intensity of E. coli. Histograms represent bacterial R1 and R2 statistical values. (E) Bacterial MRI-T1 mapping and the corresponding mean gray values analysis. (F) Bacterial MRI-T2 mapping and the corresponding mean gray values analysis. It should be noted that the bacterial densities of samples were almost equivalent (A600 units/mL, OD 50.0). Data were presented as mean ± SD (n = 3). For MRI R1 and R2 values, statistical differences were analyzed using the unpaired Student’s t-test. For the mean gray value data, statistical significances were analyzed using two-way ANOVA followed by the Bonferroni and Tukey’s HSD post-hoc test. ns, no significance (p > 0.05).*
Thus, we added ammonium ferric citrate (FAC) to the medium as an exogenous iron donor. FAC is a soluble ferric salt that has been approved by the U.S. FDA as a food supplement and clinical drug (Tenne et al., 2015). Hence, it is safe and has clear effects, as confirmed by scanning electron microscopy (SEM) and live/dead BacLight stain tests. As revealed by the SEM micrographs, E. coli retained its typical rod-shaped morphology without any discernible alteration (Supplementary Figure S4). The fluorescence pattern of SYTO 9 and double-staining with propidium iodide (PI) further confirmed the good viability of bacteria as shown in Supplementary Figure S5, where green indicates living cells and red indicates dead cells. Beyond that, the addition of FAC significantly increased the intracellular iron content within bacteria, and E. coli transfected with clMagR/clCry4 gene showed higher levels than the control, which was experimentally verified by ICP-MS (Figure 2A). However, the status of iron within the bacteria was different. Dissociative Fe will produce abundant hydroxyl free radicals resulting from the Fenton reaction, while Fe binding could greatly reduce this production of free radicals (Dixon and Stockwell, 2014). Thus, we used ESR to identify the hydroxyl free radicals with the spin adduct of 5,5-dimethyl-1-pyrroline N-oxide (DMPO)-hydroxyl radicals. For control E. coli, the quartet signal was unambiguously detected, indicating the presence of abundant hydroxyl free radicals (Figure 2B). However, for E. coli transfected with clMagR/clCry4 gene, the intensity of the corresponding ESR signal decreased greatly, meaning that dissociated Fe was bound (Figure 2C). A similar case occurred for E. coli transfected with clMagR gene, while the intensity decrease was slightly lower than that of E. coli transfected with clMagR/clCry4 gene (Supplementary Figure S6). Thus, we assumed that the dissociative Fe within E. coli was bound to the clMagR/clCry4 protein, causing the alteration of bacterial magnetism. The SQUID measurement confirmed this point. As predicted, control E. coli were diamagnetic, even after exposure to FAC (Figure 2D). A similar diamagnetic curve was also observed in E. coli transfected with clMagR/clCry4 gene in the absence of an exogenous iron supply. However, the bacteria exhibited a paramagnetic curve after culture in the presence of media containing an exogenous iron supply, and the intensity of magnetization increased by two orders of magnitude (Figure 2E). As shown by the magnetization loop, there was even a very slight hysteresis, which seemed somewhat superparamagnetic (magnetic susceptibility, 0.00625 emu/g). Surprisingly, this effect was not observed in clMagR-transfected E. coli, where the bacteria retained the diamagnetic behavior (Supplementary Figure S7). Therefore, these findings supported the potential effects on MRI contrast of transfection with clMagR/clCry4 gene in the presence of exogenous iron supply.
**FIGURE 2:** *Biological effects of exogenous iron on bacteria. (A) Iron content analysis. Data corresponding to iron content were determined by ICP-MS. The intracellular Fe content of E. coli transfected with clMagR/clCry4 was higher than that of the control. Data were presented as mean ± SD (n = 3). Statistical significance was analyzed using the unpaired Student’s t-test. ***p < 0.001. (B,C) Hydroxyl radical generation was monitored by ESR measurement. The main quartet signal characteristic for the •DMPO-OH adduct is indicated by gray columns. Compared to control E. coli, the quartet signal was also detected at by E. coli transfected with clMagR/clCry4 but at a much lower intensity. Data were normalized for intergroup difference comparison. (D,E) Hysteresis loops (magnetization M versus applied field H) of E. coli were measured by SQUID magnetometry. When cultured in the exogenous iron supply medium, E. coli transfected with clMagR/clCry4 showed detectable magnetic properties (magnetic susceptibility, 0.00625 emu/g).*
As shown by the qualitative MRI scan, the clMagR/clCry4-transfected E. coli exhibited significant MRI-T2 contrast after culturing with FAC, while there was little influence on the MRI-T1 mode (Figure 3A); the histograms represent the corresponding statistical R1 and R2 values (Figure 3B). Furthermore, to quantitatively confirm the imaging contrast effect, the mean gray values acquired from the mapping of MRI-T1 and -T2 images (Figures 3C, D) were calculated. Independently of the time of TE, it can be clearly seen that clMagR/clCry4 gene transfection imparted the contrast effect of MRI-T2 to E. coli, which was significantly different from that of the control E. coli strains. Moreover, the influence of the FAC concentration (exogenous iron level) in the medium, and the bacterial density on the imaging contrast effect, was evaluated (Supplementary Figures S8, S9). As described in detail, with an increase in the iron level or bacterial density assayed, clMagR/clCry4 protein achieved a more significant influence on the T2-contrast of the transfected E. coli. Here, the corresponding original gray values were plotted against the TE for different bacterial densities after pseudocolor processing (Supplementary Figure S10). With an increase in the bacterial density, the MRI gray value of E. coli-transfected clMagR/clCry4 gene tended to go from a constant to a high-order power function versus the TE time course of MRI-T2 mapping.
**FIGURE 3:** *MRI analysis of bacteria under exogenous iron supply conditions. (A) MRI-T1 and -T2 images showing signal intensity of bacteria. (B) MRI statistical R1 and R2 values analysis of the bacteria. (C) Bacterial MRI-T1 mapping and the corresponding mean gray values analysis. (D) Bacterial MRI-T2 mapping and the corresponding mean gray values analysis. It should be noted that the bacterial densities of samples were almost equivalent (A600 units/mL, OD 50.0). Data were presented as mean ± SD (n = 3). For MRI R1 and R2 values, statistical differences were analyzed using the unpaired Student’s t-test. For the mean gray value data, statistical significances were analyzed using two-way ANOVA followed by Bonferroni and Tukey’s HSD post-hoc test. ns, no significance (p > 0.05), ***p < 0.001.*
To explore the underlying mechanism responsible for MRI-T2 contrast E. coli transfected with clMagR/clCry4 cultured in the presence of FAC, the purified clMagR/clCry4 protein was tested for contrast properties in the MR imaging. As shown in Supplementary Figure S11, the protein itself was unable to achieve MRI-T2 contrast signals. Thus, an explanation should lie with event results within living organisms. To validate this hypothesis, we first tested the intracellular pH of E. coli after transfection of clMagR/clCry4 gene using the pH-sensitive organic dye BCECF-AM, which has been widely used in elaborating the physiology of prokaryotes and eukaryotes (Burgess and Han, 2010; Chakraborty et al., 2017). As shown in Supplementary Figure S12, the fluorescence alteration indicated that the microenvironment inside E. coli changed from a weakly alkaline pH to acidic pH with the addition of FAC. However, it was found that transfection of clMagR/clCry4 gene inhibited this change in the pH value to maintain a weakly alkaline state, the detailed mechanism of which remains unclear. Apart from that, similar evidence was obtained by means of the conventional glass electrode pH meter measurement (Supplementary Table S1). On the other hand, the Fe ions will bind to biological molecules within the bacteria. It has been known that E. coli secreted the siderophore enterobactin to chelate Fe3+ with high affinity, forming soluble Fe (III)-siderophores to transport inside E. coli (Liu et al., 1993; Faraldo-Gomez and Sansom, 2003). Fe ions could then be released either by siderophore hydrolysis or by reduction of flavin oxidoreductase, after which the dissociative Fe would be recruited by the clMagR/clCry4 protein (Coves et al., 1993; Raymond et al., 2003). However, Fe-binding residues in clMagR are flexible and easily disrupted, so the Fe ions can partially dissociate into a free state resulting from the reactive oxygen species (Rouault and Klausner, 1996; Ding and Clark, 2004). In the presence of Fe ions under weakly alkaline conditions, the formation of iron oxide nanoparticles is assumed, which is somewhat similar to a co-precipitation reaction occurring within E. coli. Hence, it should be noted that the FAD group of the clCry4 protein is widely known to generate free radicals under photonic action, which may explain why clMagR/clCry4 transfection is advantageous in the MRI-T2 contrast of E. coli compared to clMagR transfection alone.
We used TEM to observe ultrathin bacterial sections. Under low magnification, many electron-dense granules were clearly observed in the E. coli transfected with clMagR/clCry4 in the presence of exogenous iron supply (Figures 4A, B). Fe element mapping further indicated that there were abundant iron-based compounds detectable within E. coli (Figure 4C). Hence, we speculated that the electron-dense granules should contain the iron-based nanomaterials. These electron-dense granules were then extracted from the bacteria for TEM characterization by repeating magnetic separation and washing. Interestingly, a magnet could be used to easily attract the extracted granules (Supplementary Video S1), indicating that the magnetic iron oxide nanoparticles were present. In the TEM micrographs, aggregates of tiny nanoparticles can be clearly observed (Figure 4D). The exact matching of Fe and O elements in the XEDS mapping confirmed that the composition of the nanoparticles was iron oxide.
**FIGURE 4:** *Detailed information on intracellular granules. (A,B) Detection of the intracellular ultrastructure. Following the transfer of E. coli transfected with clMagR/clCry4 into the exogenous iron supply medium, the electron micrograph shows that electron-dense granules were detectable in the cytoplasm and form widely distributed aggregates, marked by solid circles.
(C) Electron micrographs and the corresponding XEDS Fe (Kα) elemental mapping of E. coli transfected with clMagR/clCry4 under exogenous iron supply conditions. (D) Electron micrographs and corresponding XEDS elemental mappings of the two-dimensional morphologies of the extract particles. (E,F) SAED and HRTEM patterns indicated by the dotted frame in (D).*
In addition, STEM micrographs were acquired together with XEDS elemental mappings to give complementary information on the particles being imaged, and the extracted particles demonstrated a small size range (averaged 20.62 ± 3.5 nm) (Supplementary Figure S13). Furthermore, SAED and HRTEM demonstrated the coexistence of the Fe2O3 and Fe3O4 polycrystal phases in the tiny nanoparticles (Figures 4E, F), which provides evidence supporting the presence of a magnetization loop and the MRI-T2 contrast effect of clMagR/clCry4 transfected in E. coli. Nonetheless, it should be mentioned that aggregates of tabular-like ferric oxide nanoparticles with poor crystalline features were also detected (Supplementary Figure S14). These amorphous precursors demonstrated, in part, that the formation process of such nanoparticles was somewhat similar to that achieved by a chemical co-precipitation method.
## Conclusion
This study was the first to exploit the transfection of the clMagR/clCry4 gene to produce an MRI-T2 contrast agent effect on a living organism. This innovative phenomenon has neither been observed nor hypothesized previously. Using E. coli as the model organism, an exogenous iron supply was shown to be a critical factor for this phenomenon. Exogenous Fe could be internalized in E. coli and form a dynamic “bind-release” process with the clMagR/clCry4 protein. During the cycle, a biosynthesis process occurred in the presence of dissociative Fe under weakly alkaline intracellular conditions, which is somewhat similar to the chemical co-precipitation reaction, leading to the formation of tiny iron oxide nanoparticles. These tiny iron oxide particles influence the transverse relaxation observed on MRI. Because this strategy depends on the activity of the associated protein complex, it has a wide range of potential applications that are dependent on the presence of an exogenous iron supply. We believe that this novel technique will play a key role in future biomedical imaging and tracking applications. Furthermore, due to the key role of clMagR/clCry4 protein in magnetoreception, our findings are useful for harmonizing the long-term controversy over the existence of magnetoreceptors in organisms, ranging from prokaryotes to animals.
## 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
JS and NL conceived the study design and drafted the manuscript. NL, LX, PW, and CZ participated in data extraction and data analysis. JS, NL, XM, XH, YX, BW, YG, and YZ performed data checking and analysis. JS and NL reviewed and edited the manuscript. All authors 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/fmolb.2023.1119356/full#supplementary-material
## References
1. Beinert H., Holm R. H., Munck E.. **Iron-sulfur clusters: Nature's modular, multipurpose structures**. *Science* (1997) **277** 653-659. DOI: 10.1126/science.277.5326.653
2. Burgess K., Han J. Y.. **Fluorescent indicators for intracellular pH**. *Chem. Rev.* (2010) **110** 2709-2728. DOI: 10.1021/cr900249z
3. Calvin A. D., Misra S., Pflueger A.. **Contrast-induced acute kidney injury and diabetic nephropathy**. *Nat. Rev. Nephrol.* (2010) **6** 679-688. DOI: 10.1038/nrneph.2010.116
4. Chakraborty S., Winardhi R. S., Morgan L. K., Yan J., Kenney L. J.. **Non-canonical activation of OmpR drives acid and osmotic stress responses in single bacterial cells**. *Nat. Commun.* (2017) **8** 1587. DOI: 10.1038/s41467-017-02030-0
5. Coves J., Eschenbrenner M., Fontecave M.. **Sulfite reductase of**. *Biochem. Bioph. Res.Co.* (1993) **192** 1403-1408. DOI: 10.1006/bbrc.1993.1572
6. Ding H. G., Clark R. J.. **Characterization of iron binding in IscA, an ancient iron-sulphur cluster assembly protein**. *Biochem. J.* (2004) **379** 433-440. DOI: 10.1042/BJ20031702
7. Dixon S. J., Stockwell B. R.. **The role of iron and reactive oxygen species in cell death**. *Nat. Chem. Biol.* (2014) **10** 9-17. DOI: 10.1038/nchembio.1416
8. Faraldo-Gomez J. D., Sansom M. S. P.. **Acquisition of siderophores in Gram-negative bacteria**. *Nat. Revi. Mol. Cell Bio.* (2003) **4** 105-116. DOI: 10.1038/nrm1015
9. Gadi V. K., Alybaev D., Raj P., Garg A., Mei G. X., Sreedeep S.. **A novel Python Program to automate soil colour analysis and interpret surface moisture content**. *Int. J. Geosynth. Groun.* (2020) **2** 21-9260. DOI: 10.1007/s40891-020-00204-3
10. Holm R. H., Lo W.. **Structural conversions of synthetic and protein-bound iron-sulfur clusters**. *Chem. Rev.* (2016) **116** 13685-13713. DOI: 10.1021/acs.chemrev.6b00276
11. Jutz G., van Rijn P., Miranda B. S., Boeker A.. **Ferritin: A versatile building block for bionanotechnology**. *Chem. Rev.* (2015) **115** 1653-1701. DOI: 10.1021/cr400011b
12. Liu J., Rutz J. M., Feix J. B., Klebba P. E.. **Permeability properties of a large gated channel within the ferric enterobactin receptor, FepA**. *Proc. Natl. Acad. Sci. USA.* (1993) **90** 10653-10657. DOI: 10.1073/pnas.90.22.10653
13. Liu X. F., Theil E. C.. **Ferritins: Dynamic management of biological iron and oxygen chemistry**. *Acc. Chem. Res.* (2005) **38** 167-175. DOI: 10.1021/ar0302336
14. Maeda K., Robinson A. J., Henbest K. B., Hogben H. J., Biskup T., Ahmad M.. **Magnetically sensitive light-induced reactions in cryptochrome are consistent with its proposed role as a magnetoreceptor**. *Proc. Natl. Acad. Sci. USA.* (2012) **109** 4774-4779. DOI: 10.1073/pnas.1118959109
15. Park S. m., Aalipour A., Vermesh O., Yu J. H., Gambhir S. S.. **Towards clinically translatable**. *Nat. Rev. Mat.* (2017) **2** 17014. DOI: 10.1038/natrevmats.2017.14
16. Qin S., Yin H., Yang C., Dou Y., Liu Z., Zhang P.. **A magnetic protein biocompass**. *Nat. Mat.* (2016) **15** 217-226. DOI: 10.1038/nmat4484
17. Raymond K. N., Dertz E. A., Kim S. S., Kim S. S.. **Enterobactin: An archetype for microbial iron transport**. *Proc. Natl. Acad. Sci. USA.* (2003) **100** 3584-3588. DOI: 10.1073/pnas.0630018100
18. Rogers W. J., Meyer C.,H., Kramer C. M.. **Technology insight:**. *Nat. Clin. Pract. Card.* (2006) **3** 554-562. DOI: 10.1038/ncpcardio0659
19. Rouault T. A., Klausner R. D.. **Iron-sulfur clusters as biosensors of oxidants and iron**. *Trends biochem. Sci.* (1996) **21** 174-177. DOI: 10.1016/s0968-0004(96)10024-4
20. Schilling F., Ros S., Hu D. E., D'Santos P., McGuire S., Mair R.. **MRI measurements of reporter-mediated increases in transmembrane water exchange enable detection of a gene reporter**. *Nat. Biotechnol.* (2017) **35** 75-80. DOI: 10.1038/nbt.3714
21. Sindhulakshmi K., Soundarya J., Sowmya U.. **Cloud controlled intrusion detection and burglary prevention stratagems in home automation systems**. *Int. J. Eng. Sci.* (2014) **2** 2319-5967
22. Stephenson E., Nathoo N., Mahjoub Y., Dunn J. F., Yong V. W.. **Iron in multiple sclerosis: Roles in neurodegeneration and repair**. *Nat. Rev. Nephrol.* (2014) **10** 459-468. DOI: 10.1038/nrneurol.2014.118
23. Sun S. J., Chasteen N. D.. **Rapid kinetics of the EPR-active species formed during initial iron uptake in horse spleen apoferritin**. *Biochemistry* (1994) **33** 15095-15102. DOI: 10.1021/bi00254a019
24. Tenne D., Bogoslavsky B., Bino A.. **Ferric ammonium citrate - what's in it?**. *Eur. J. Inorg. Chem.* (2015) **25** 4159-4161. DOI: 10.1002/ejic.201500782
25. Turano P., Lalli D., Felli I. C., Theil E. C., Bertini I.. **NMR reveals pathway for ferric mineral precursors to the central cavity of ferritin**. *Proc. Natl. Acad. Sci. USA.* (2010) **107** 545545-550550. DOI: 10.1073/pnas.0908082106
26. Wahsner J., Gale E. M., Rodriguez-Rodriguez A., Caravan P.. **Chemistry of MRI contrast agents: Current challenges and new Frontiers**. *Chem. Rev.* (2019) **119** 957-1057. DOI: 10.1021/acs.chemrev.8b00363
27. Wajnberg E., Alves O. C., Perales J., da Rocha S. L. G., Ferreira A. T., Cameron L. C.. **Ferritin from the haemolymph of adult ants: An extraction method for characterization and a ferromagnetic study**. *Eur. Biophys. J. Biophy.* (2018) **47** 641-653. DOI: 10.1007/s00249-018-1293-3
28. Xiao D. W., Hu W. H., Cai Y., Zhao N.. **Magnetic noise enabled biocompass**. *Biol. Phys.* (2020) **124** 128101. DOI: 10.1103/PhysRevLett.124.128101
29. Xu J., Jarocha L. E., Zollitsch T., Konowalczyk M., Henbest K. B., Richert S.. **Magnetic sensitivity of cryptochrome 4 from a migratory songbird**. *Nature* (2021) **594** 535-540. DOI: 10.1038/s41586-021-03618-9
30. Xue L., Hu T., Guo Z., Yang C., Wang Z., Qin S.. **A novel biomimetic magnetosensor based on magneto-optically involved conformational variation of MagR/cry4 complex**. *Adv.Electron. Mat.* (2020) **6** 1901168. DOI: 10.1002/aelm.201901168
31. Yang C., Tian R., Liu T., Liu G.. **MRI reporter genes for noninvasive molecular imaging**. *Molecules* (2016) **21** 580. DOI: 10.3390/molecules21050580
32. Zecca L., Youdim M. B. H., Riederer P., Connor J. R., Crichton R. R.. **Iron, brain ageing and neurodegenerative disorders**. *Nat. Rev. Neurosci.* (2004) **5** 863-873. DOI: 10.1038/nrn1537
33. Zoltowski B. D., Chelliah Y., Wickramaratne A., Jarocha L., Karki N., Xu W.. **Chemical and structural analysis of a photoactive vertebrate cryptochrome from pigeon**. *Proc. Natl. Acad. Sci. USA.* (2019) **116** 19449-19457. DOI: 10.1073/pnas.1907875116
|
---
title: The protein kinase R modifies gut physiology to limit colitis
authors:
- Howard Chi Ho Yim
- Arindam Chakrabarti
- Sean Kessler
- Hiroyuki Morimoto
- Die Wang
- Dhanya Sooraj
- Afsar U. Ahmed
- Carol de la Motte
- Robert H. Silverman
- Bryan RG. Williams
- Anthony J. Sadler
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9981792
doi: 10.3389/fimmu.2023.1106737
license: CC BY 4.0
---
# The protein kinase R modifies gut physiology to limit colitis
## Abstract
Here we investigate the function of the innate immune molecule protein kinase R (PKR) in intestinal inflammation. To model a colitogenic role of PKR, we determine the physiological response to dextran sulfate sodium (DSS) of wild-type and two transgenic mice strains mutated to express either a kinase-dead PKR or to ablate expression of the kinase. These experiments recognize kinase-dependent and -independent protection from DSS-induced weight loss and inflammation, against a kinase-dependent increase in the susceptibility to DSS-induced injury. We propose these effects arise through PKR-dependent alteration of gut physiology, evidenced as altered goblet cell function and changes to the gut microbiota at homeostasis that suppresses inflammasome activity by controlling autophagy. These findings establish that PKR functions as both a protein kinase and a signaling molecule in instituting immune homeostasis in the gut.
## Introduction
Inflammatory bowel disease (IBD) is a heterogeneous disorder that is commonly characterized as either ileal or colonic Crohn’s disease, or ulcerative colitis. The precise mechanisms of how disease manifests remain to be established, but IBD is considered to be a consequence of the loss of immune tolerance against the gut microbiota. Current anti-inflammatory and immunosuppressive treatments provide only temporary relief and are not universally effective. Greater mechanistic insights into gut immunity is required in order to develop strategies to control immune pathogenesis.
The protein kinase R (PKR) is a member of the small family of eukaryotic initiation factor alpha (eIF2α) kinases that constitute a universal stress response in eukaryotes [1]. Among this kinase family PKR is most cogently linked with immunity, as its expression is induced by the antiviral type I and III interferons. These cytokines, particularly the type III interferons, are important for epithelial function and mucosal immunity (2–5). The function of PKR in colitis is currently confused, with discordant effects reported [6, 7]. This warrants further study.
Here we reassess the role of PKR in dextran sulfate sodium (DSS)-induced colitis using mice that are ablated for PKR expression. We replicate experiments conducted in previous studies but investigate an alternative mode of activity from that previously proposed, identifying a different mechanism of PKR activity in colitis. Rather than PKR functioning through induction of the unfolded protein response (UPR), which is more closely associated with the related PKR-like endoplasmic reticulum kinase (PERK), we contend that PKR promotes gut barrier function and suppresses inflammatory pathogenesis in colitis by controlling autophagy in goblet cells. Additionally, we test the response of a transgenic mouse with a point mutation that disables the kinase activity of PKR, thereby testing functions that are independent of the kinase’s canonical control of the initiation of translation. The findings demonstrate that PKR functions in DSS-induced pathogenesis by suppressing the activity of inflammasomes through modulation of the gut physiology by kinase-dependent and -independent processes. This accords with the reported activity of another eIF2α kinase, the general control nonderepressible 2 (GCN2) in DSS-induced colitis [8], although this response had not been segregated from its phosphorylation of eIF2α. These findings reinforce the importance of autophagy to promote gut barrier function in gastric disease, particularly by supporting the function of goblet cells.
## Mice
Congenic C57BL/6J mice (8-10 weeks old) were exposed to $2.5\%$ DSS (36-50 kDa, MP Biomedicals) in their drinking water for up to 9 days, with or without intraperitoneal injection of 2 mg/kg of CP456773 (Sigma-Aldrich) prior to DSS treatment. Separately reared wild-type (WT), PKR-ablated (Eif2ak2 -/-) and kinase-dead PKR (K271R) mice were treated with DSS at the Hudson Institute of Medical Research in Australia. This experiment was extended with a second cohort of littermate WT and Eif2ak2 -/- mice reared at the Learner Research Institute in the USA. Experiments were performed according to the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health following a protocol that was approved by the Institutional Animal Care and Use Committee of Cleveland Clinic, or the Australian Code of Practice following a protocol approved by the Monash Medical Centre Animal Ethics Committee. Eif2ak2 -/- and K271R mice were produced as previously reported [9, 10]. A disease activity index (DAI) was calculated as described previously [11].
## Histology
Tissues from the stomach, small intestine and colon were cut open, rinsed with PBS, then fixed in $10\%$ formalin for 4 hours before embedding in paraffin. Embedded tissue was sectioned (5 μm) then stained with: biotinylated hyaluronan-binding protein (Calbiochem-EMD Millipore) and streptavidin-488 (Life Technologies), anti-alpha actin 2 and alexa fluor 568 then mounted with Vectashield containing DAPI as previously described [11]; hematoxylin and eosin (H&E); anti-proliferating cell nuclear antigen (PCNA), -villin and -H+/K–ATPase primary antibodies and alexa fluor 546 or 488 donkey anti-goat or -mouse secondary antibodies (Molecular Probes); or alcian blue/periodic acid–Schiff (PAS). Stains were detected with TSA Fluorescence System (Perkin Elmer), imaged by Axioskop 2 plus fluorescence microscope (Carl Zeiss) and analyzed with AxioVIson SE64 software (Carl Zeiss) or ScanScope XT digital scanner (Leica) and ImageScope software (Leica) and quantified from four random images per mouse using ImageJ software (NIH). Sections were scored blindly. Colitis was scored as previously described [11].
## In situ hybridization
In situ hybridization was performed with the Falma microprobe system as stipulated by the manufacturer (Falma). Sections were hybridized with digoxigenin (DIG)-labeled control sense and anti-sense RNA probes encompassing nucleotides 173 to 1091 of the open-reading frame of the PKR gene (Eif2ak2). Sections were immunohistochemically stained to visualize the hybridized probe using AP-conjugated mouse monoclonal anti-DIG antibody (Roche) and either the Fast Red or NBT/BCIP substrates for stomach or colon tissues, respectively (ThermoFisher Scientific). Images were captured by fluorescence microscopy.
## Flow cytometry
Spleens were mashed in RPMI 1640 (SIGMA) containing $1\%$ fetal bovine serum (FBS), strained through a nylon filter, then washed before staining with FAM-FLICA according to the manufacturer’s instructions (ImmunoChemistry Technologies LLC), washed with PBS then stained with anti-F$\frac{4}{80}$ Pacific Blue antibody (MCA497PB Bio-Rad) and anti-Ly6G APC-Cy7 antibody (BD Biosciences), washed with PBS and fixed in $10\%$ formalin then visualized by BD FACSCanto II and analyzed by Cytobank software (Cytobank Inc.)
## Colon explant
Tissue pieces (0.5 cm) were cut from the proximal colon and rinsed with PBS, then cultured in DMEM containing $1\%$ FBS and penicillin/streptomycin at 37°C for 24 hours. The level of tumor necrosis factor α (TNFα) was assayed by ELISA (555268 BD Biosciences). The cell-free supernatant (500 μl) was precipitated by adding methanol and chloroform as previously reported [12], then probed with anti-interleukin (IL)1β (ab9722 Abcam) and -IL18 (D046-3 MBL International) antibodies by immunoblot.
## Immune fluorescence
Caspase-1 (Casp1) activity was assessed by FAM-FLICA and FLICA-660 (No. 97 and 9122, respectively, from ImmunoChemistry Technologies LLC or, alternatively, ThermoFisher Scientific). The fluorescent reporter was quantitated in immune cells isolated from the spleen by flow cytometry or, alternatively, in tissues by confocal microscopy after cryopreservation, sectioning (5 μm), being fixed with $10\%$ formalin and permeabilized by methanol. The mucus layer in the colon and goblet cells was visualized with a fluorescein-linked lectin *Ulex europaeus* agglutinin-1 (UEA1) (ThermoFisher Scientific). Mucus thickness was measured in micrographs by confocal microscopy. Autophagic puncta were detected in fixed and permeabilized colon tissues with a fluorescent antibody to the microtubule-associated protein 1A/1B light chain 3B and (LC3B) (ThermoFischer Scientific). Autophagosomes in goblet cells were scored as LCB and UEA1 positive cells. Cells and tissues were counterstained with Hoechst 33342 (Merck) to visualize cell nuclei. Images were captured by Nikon C1 confocal microscope and analyzed by Imaris software.
## Immunoblot
Protein lysates from splenic cells were harvested by RIPA buffer as previously described [10], heat denatured in sample buffer (125 mM Tris-HCl, pH 6.8, $4\%$ SDS, $20\%$ glycerol, $5\%$ β-mercaptoethanol, $0.01\%$ bromophenol blue) and resolved through 10-$15\%$ SDS-polyacrylamide gel by electrophoresis, then transferred to Immobilon-FL membrane (Millipore). Membranes were treated with blocking buffer (LI-COR) then probed with primary and secondary antibodies conjugated with fluorophore, and visualized and quantified using the Odyssey Imaging System (LI-COR). Lysates were probed with anti-IL1β, -IL18, -CASP1-p10 (sc-514 Santa Cruz Biotechnology Inc.), anti-apoptosis-associated speck-like protein containing a CARD (ASC) and -NOD-, LRR- and pyrin domain-containing protein 3 (NLRP3) antibodies (AL177 and Cryo-2, respectively, Adipogen), -eukaryotic initiation factor 2 (Eif2) and -phospho-Eif2 Ser51) (#9722 and 119A11, respectively, Cell Signaling Technology) and protein loading was assessed with anti-β-actin (ab8226 Abcam) or -α-tubulin (3873 Cell Signaling) antibodies.
## Affinity chromatography
To enrich the kinase domain of PKR, protein lysates of murine fibroblasts were captured on a Hi-Trap heparin column following the manufacturer’s protocol (GE Life Sciences). Eluted peptides were separated by PAGE and transferred to a membrane support before being probed with the anti-PKR antibody D-20 (Santa Cruz Biotechnology).
## Bacterial content detection in feces
DNA was extracted by the QIAmp Fast DNA Stool Mini Kit, following the manufacturer’s protocol (QIAGEN) and 1 pg was used with previously described primers [13] to amplify the 16s rRNA gene of Bacteroides, Lactobacillus and Prevotella by quantitative PCR using Applied Biosystems 7900HT systems.
## Statistical analysis
The Prism software (GraphPad) was used for all statistical analyses. Statistical significance of the differences between two groups was analyzed by two-way or one-way ANOVA with either with Šidàk post-test or Tukey’s range test or, alternatively, two-tailed and unpaired Student’s t-test. The correlation analysis was done by one-tailed Pearson correlation test.
## PKR affects DSS-induced weight loss
Cohorts of congenic WT and PKR mutant mice were treated with DSS in their drinking water and their respective weights compared. The Eif2ak2 -/- mice show significantly greater weight loss compared to the WT animals after the second day of DSS treatment until day 6 (Figure 1A). This increased weight loss in the first five days of exposure to DSS was not apparent in the kinase-dead K271R mice (Figure 1B). Accordingly, PKR expression, independent of substrate phosphorylation, is protective against DSS-induced weight loss.
**Figure 1:** *PKR reduces DSS-induced weight loss. (A–C) Body weight of mice treated with 2.5% DSS in their drinking water, expressed as the percentage change from the starting weight of: (A) Separately reared WT and PKR-ablated (Eif2ak2-/-
) mice (n=19 and 13, respectively); (B) WT compared to PKR kinase-dead (K271R) mice (n=19 and 7, respectively), and; (C) Cohoused WT and Eif2ak2
-/- littermates (n=5 and 3, respectively). Data were collected from three independent experiments and are expressed as the mean ± S.E.M. and analyzed by two-way ANOVA with Šidàk post-test on the means between genotypes on each day.*
These experiments were conducted on mice reared in separate cages at the Monash Animal Research Platform at Monash University, Victoria, Australia. The response to DSS is strongly influenced by environmental variables, particularly differences in the microbiota. This has been proposed as a cause of discrepancies between separate studies that use this model of colitis. To explore this contingency, we performed a second comparison with a limited number of WT and Eif2ak2 -/- littermate mice raised at the Animal Core, Lerner Research Institute, Ohio, USA. These data confirm the protective function of PKR against this insult, although there were some changes (Figure 1C). Most conspicuously, the WT mice from the Lerner maintained their weight throughout the experiment. Additionally, the kinetics of weight loss in the PKR-ablated mice was delayed compared to the cohort from the Monash (Figures 1A, C).
These data identify that PKR protects against DSS-induced weight loss and distinguish kinase-dependent and -independent effects of PKR, as has been previously asserted from in vitro experiments [14].
## PKR affects DSS-induced tissue injury
Only modest tissue pathogenesis was evident in tissue sections of the colon from any of the three mouse genotypes reared at the Monash Animal Research Platform. Accordingly, the entire length of the colon was assessed for damage by quantifying disorganized and incomplete crypts as a percentage of the longitudinal length of the colon (as previously described [15]). Tissues were assessed on day five of DSS treatment, when the differential in weight was greatest between the separate cohorts (Figure 1A). The WT animals showed increased tissue injury compared to both the PKR-ablated and kinase-dead (K271R) mice, although the difference between kinase-dead and WT mice was not assessed as being significantly different (Figures 2A, B). A quantitation of serum creatinine levels (measured at the Monash Health Pathology), which is used as a clinical variable of colonic injury, appears to confirm a worsened response in the WT compared to the kinase-dead mice (Figure 2C). These data identify a discordance between DSS-induced weight change (Figure 1) and tissue damage in the colon, which was mediated by PKR’s kinase activity (Figures 2A, B).
**Figure 2:** *PKR kinase activity modulates DSS-induced tissue damage. The severity of colon damage of the mice reared separately at the Monash Animal Research Platform, expressed as the percentage of the entire length of the colon in either (A) WT compared to PKR-ablated (Eif2ak2
-/-) mice or (B) WT compared to the kinase-dead PKR mice (n=6). Data collected from two independent experiments are expressed as mean ± S.E.M. and analyzed by two-way ANOVA with Šidàk post-test on the mean between genotypes on each day. (C) Measures of the levels of creatinine in the serum of the indicated mice (n=3). Data are expressed as mean ± S.D. and analyzed by unpaired t-test from a single experiment.p=0.023; p=0.053; p=0.0078.*
The equivalent analysis was not conducted on the cohort from the Lerner Research Institute, although an analysis of disease index with a histological analysis of tissues from the colon of the mice exposed to DSS for 9 days is shown as supplementary data (Supplementary Figure 1).
## PKR limits DSS-induced inflammation in the gut
We assessed the immune response in the DSS-treated cohort that was reared at the Monash Animal Research Platform. There was a modest, statistically nonsignificant, increase in the immune cell infiltrate into colon tissues from WT mice compared to the PKR-ablated or kinase-dead (K271R) mice (Figure 3A). However, there were significantly lower levels of the inflammatory IL1β and IL18 cytokines in the WT compared to the PKR-ablated mice (Figure 3B). The PKR-K271R mice had a lower, but not statistical different, level of these cytokines from that in the WT mice (Figure 3B). As the modestly heightened tissue damage in the WT mice is not accompanied by a commensurate increase in inflammatory markers, this damage would appear to be a consequence of primary injury by DSS, rather than being immune mediated. The data also identify that inflammasome activity is suppressed, partly but not entirely, by substrate phosphorylation by PKR. Notably, the expression of the inflammasome constituents; caspase-1 (Casp1), apoptosis-associated speck-like protein containing a CARD (Asc), and the NOD-, LRR- and pyrin domain-containing protein 3 (Nlrp3) was equivalent between the different mice (Supplementary Figure 2). There was also no difference in expression of the unprocessed pro-IL1β cytokine or the induction of the independent (of inflammasomes) inflammatory cytokine TNFα (Figures 3C, D). Assessment of the levels of cleaved cytokines in untreated mice indicated an increase in inflammasome activity in the PKR-ablated mice at homeostasis (Figure 3E). Correspondingly, a fluorescent substrate reporter indicated that there was heightened Casp1 activity in the colon tissue from PKR-ablated mice compared to the WT animals prior to treatment with DSS (Figure 3F).
**Figure 3:** *PKR limits DSS-induced inflammasome activity in the colon. (A-F) Analysis of inflammasome activity in WT, PKR-ablated (Eif2ak2
-/-) and kinase-dead (K271R) mice. (A) Inflammatory cell infiltrates into the colons of mice treated with DSS for 5 days assessed from H&E-stained tissue sections (n=7). Data were collected from two independent experiments. (B) Quantitation of the relative levels of cleaved IL1β and IL18 produced from colon explants of DSS-treated mice as detected by immunoblot (n=7). Data were collected from two independent experiments. (C) Induction of pro-IL1β in colon explants measured by immunoblot and expressed relative to WT mice (n=4). Data were collected from four independent experiments and analyzed by one-way ANOVA with Tukey’s range test. (D) Levels of TNFα expressed from colon explant detected by ELISA (n=4). Data were collected from four independent experiments and analyzed by one-way ANOVA with Tukey’s range test. (E) Quantitation of the relative levels of cleaved IL1β and IL18 produced from colon explants of untreated mice as detected by immunoblot (n=7). Data were collected from two independent experiments. (F) Micrographs of colon tissue from untreated mice stained for Casp1 activity (FLICA-660) (red) and counter stained with Hoechst to mark cell nuclei (blue). Casp1 activity is quantitated by fluorescent confocal microscopy in the graph on the right (n=3). Data is expressed as mean ± S.E.M. and analyzed by unpaired t-test.*
These data identify that PKR represses inflammasome activity at homeostasis and limits cytokine processing in the colon in response to DSS treatment. The retention of much of this activity in the kinase-dead mouse identifies partial independence from eIF2α phosphorylation.
## PKR limits DSS-induced inflammation in neutrophils
To measure the immune response outside of the gut, we assessed innate immune cells from the spleens of WT, PKR-ablated and the kinase-dead mice treated with DSS for 5 days. Equivalent levels of the leukocyte antigen lymphocyte antigen 6 complex locus G6D (Ly6G) was apparent in the spleens of WT and Eif2ak2 -/- mice at rest and this was uniformly elevated after treatment with DSS (Figure 4A). Comparison of the cleavage of a Casp1 substrate reporter (YVAD) in Ly6G-positive and adhesion G protein-coupled receptor E1 (F$\frac{4}{80}$)-positive cells from the spleens of mice showed that DSS treatment induced Casp1 activity in neutrophils disproportionately more in the PKR-ablated relative to the WT mice (Figures 4B, C). This PKR-dependent suppression of Casp1 activity was independent of eIF2α phosphorylation, as splenic neutrophils from the WT and kinase-dead mice demonstrated equivalent reporter activity (Figures 4D, E). Accordingly, the heightened inflammatory response caused by ablating PKR expression extends beyond the gastrointestinal tract.
**Figure 4:** *PKR limits inflammasome activity in splenic neutrophils. (A) Detection of Ly6G+ inflammatory cells in the spleens of WT compared to PKR-ablated (Eif2ak2
-/-) mice either untreated or treated with DSS for five days. (B–E) Measures of Casp1 activity (YVAD+), quantified by flow cytometry in splenic neutrophils (Ly6G++YVAD+) and macrophages (Mϕ, F4/80++YVAD+) from (B, C) the WT compared to PKR-ablated mice after 5 days of DSS treatment (H2O, n=4 and DSS, n=5) and (D, E) WT compared to PKR-K271R kinase-dead mice assessed after the indicated day of DSS treatment (n=6). Data are expressed as mean ± S.E.M. and analyzed by two-way ANOVA with Šidàk post-test.*
## Multiple inflammasomes are active in DSS-induced inflammation
As PKR is controlling the activity of the inflammasomes and because NLRP3 has been shown to be important in colitis, we treated mice with the inhibitor CP456773 to assess the involvement of this sensor in the response to DSS [16]. WT and PKR-ablated mice were injected with CP456773 or, as a control, the carrier solvent at days 1, 2 and 4 during the course of 5 days of DSS treatment. This treatment diminished the weight loss of the PKR-ablated mice, thereby confirming a function for NLRP3 in this phenotype (Figures 5A, B). Intriguingly, treatment with CP456773 worsened the low level of DSS-induced tissue damage in the WT animal but not in the PKR-ablated mouse (Figures 5C, D). This appears consistent with PKR kinase activity promoting tissue damage via suppression of the NLRP3 inflammasomes (Figures 2, 3).
**Figure 5:** *Inhibiting NLRP3 restores control in the absence of PKR expression. (A, B) Body weights of mice treated with DSS and injected with either the NLRP3 inhibitor CP456773 or the control vehicle solute, expressed as the percentage change from the starting weight of (A) PKR-ablated (Eif2ak2
-/-) or (B) WT mice (n=6) and analyzed by two-way ANOVA with Šidàk post-test. (C, D) Measures of the effect of CP456773 on DSS-induced tissue damage, expressed as the percentage of the entire length of the colon in either (C) PKR-ablated or (D) WT mice (n = 6) and analyzed by unpaired t-test. (E, F) Measures of the effect of CP456773 on the relative change in the levels of mature IL1β and IL18 produced from colon explants from either (E) PKR-ablated or (F) WT mice treated with the inhibitor (WT+Vehicle and WT+CP456773 n=3; Eif2ak2
-/-+Vehicle n=4; Eif2ak2
-/-+CP456773 n=5). Cytokines were assayed by immune blot and expressed as fold induction compared to Vehicle-treated mice and analyzed by unpaired t-test. (G, H) Casp1 activity (YVAD+) in splenic neutrophils (FLICA+Ly6G+) or macrophages (Mϕ, FLICA+F4/80+) from either (G) PKR-ablated or (H) WT mice treated with the NLRP3 inhibitor (n=7 and n=8, respectively). Fluorescent probes for Casp1 activity were detected and quantitated by flow cytometry. Data are expressed as mean ± S.E.M. and analyzed by unpaired t-test.*
The NLRP3 inhibitor further reduced the low levels of IL18 and IL1β in colon explants from the WT animals but had no significant effect on the relatively higher levels of these cytokines in the PKR-ablated mice (Figures 5E, F). Accordingly, the processing of these cytokines in the colon from Eif2ak2 -/- mice is mediated by an inflammasome constituted by a sensor protein other than NLRP3 (Figures 5E with 3B). This finding with the data showing CP456773 stabilized the weight of mice treated with DSS (Figures 5A, B) also suggest that this alternative inflammasome is not causal of the observed weight loss (Figures 5A, B with Figure 1). Opposing this pattern in the colon, CP456773 reduced measures of Casp1 activity in splenic neutrophils from PKR-ablated mice, while there was no change in the WT mice (Figure 5G, H). Therefore, NLRP3 contributes to the inflammasome that is active in splenic neutrophils from Eif2ak2 -/- mice.
These data identify PKR-dependent suppression of inflammasome activity and show that different sensor proteins constitute inflammasomes in separate tissues of the DSS-treated mice. Better maintenance of weight in CP456773-treated mice suggests that NLRP3 participates in this phenotype, despite no evidence of suppression of cytokines processed by inflammasome activity in the colon. In addition, the modest increase in tissue damage observed in the colon of CP456773-treated mice suggests that NLRP3 is protective against DSS-induced damage. A possible cause for this apparent altered susceptibility was investigated.
## PKR affects gastrointestinal physiology
We examined the gut physiology of WT, PKR-ablated and point mutant kinase-dead mice. Although at odds with the original description and subsequent measures [7, 9], it was reported that the mutated Eif2ak2 locus retained expression of a truncated kinase domain [17]. Accordingly, we sought to verify the ablation of PKR in the Eif2ak2 -/- mouse. Whole cell lysates from embryonic fibroblasts from the WT and Eif2ak2 -/- mice were passed through a heparin column to capture the putative peptide via its reported binding of heparin [18]. The removal of PKR from the Eif2ak2 -/- mouse was confirmed by probing the heparin-bound eluents with an anti-PKR antibody specific for the kinase domain (Supplementary Figure 3).
We then confirmed expression of the Eif2ak2 transcript in the tissues from the stomach and colon by in situ hybridization (Figures 6A, E). Histological examination of tissue from throughout the gastrointestinal tract of the mice suggested a difference in goblet cells between the WT and PKR-ablated mice (Figures 6B–F). This appeared most evident in the small intestine (Figure 6C).
**Figure 6:** *The effect of PKR on gastrointestinal physiology. (A–F) Micrographs of histologic specimens of the stomach, small intestine and colon from the indicated mice. (A) Tissue from the stomach probed with sense (S) and anti-sense (AS) oligonucleotides against the Eif2ak2 transcript, and (B) stained for proliferating cell nuclear antigen (PCNA, green) and H+/K–ATPase (red) to assess cell proliferation and the parietal cells in the stomach mucosa, respectively, and counterstained with Hoechst (blue) to detect cell nuclei. (C) Tissue from the small intestine stained with H&E or (D) fluorescent probes against villin (red), to visualize microvilli at the brush boarder of the epithelial lining of the gut, PCNA (green) and Hoechst (blue). (E) Tissue from the colon probed with sense and anti-sense oligonucleotides against the Eif2ak2 transcript and (F) stained with H&E to visualize the tissue structure. Representative images are shown from three independent experiments.*
Periodic acid-Schiff (PAS) staining of muco-substances supports a slight, non-statistically significant, increase in the number of goblet cells in the colon of PKR-ablated mice and a significant increase in the kinase-dead mouse compared to the WT animals (Figures 7A–C). PAS staining of colon tissue after 5 days of DSS treatment showed a significant increase in the size of goblet cells in the Eif2ak2 -/- but not the kinase-dead or WT mice (Figures 7A-C). The goblet cell hyperplasia detected in the kinase-dead mouse resolved after DSS treatment (Figure 7C). Given our recognition that PKR controls inflammasome activity with previous reports linking goblet cell development with inflammasomes [19], we tested if there was a correlation between the size of goblet cells and the levels of the relevant cytokines. Figure 7D shows that the levels of IL18 (but not IL1β) correlated with goblet cell hypertrophy.
**Figure 7:** *PKR alters gut physiology. (A) Micrographs of histologic specimens of the colon from WT, PKR-ablated (Eif2ak2
-/-) and kinase-dead (K271R) mice stained with PAS. Representative images are shown from two independent experiments. (B, C) Quantitation of the size and number of goblet cells visualized by PAS staining of colon from (B) WT and PKR-ablated (Eif2ak2
-/-) mice (H2O n=4 and DSS n=7) or (C) WT and kinase-dead (K271R) mice, either untreated or treated with DSS for 5 days or as indicated (H2O n=4 and DSS n=7). Data were analyzed by two-way ANOVA with Šidàk post-test. (D) Correlation analysis between the fold induction of IL18 or IL1β and the average goblet cell size from WT and PKR-ablated (Eif2ak2
-/-) mice (n=6). (E) Measures of the relative thickness of the mucus lining as visualized with UEA1 probing of colon tissues from the indicated mice, either untreated or treated with DSS in their drinking water for 5 days (H2O n=4 and DSS n=7). Each data point represents quantification of one field, with four to five fields assessed per confocal image per mouse. The data were analyzed by two-way ANOVA with Šidàk post-test. (F) Measures of autophagy were made by quantitation of LC3B puncta, identified as circular objects with a diameter of 10-70 pixels, in UEA1-positive cells from the indicated mice, either untreated or treated with DSS in their drinking water for 5 days (H2O n=4 and DSS n=7). 2000-4000 UEA1 positive cells were scored per microscopic field using CellProfiler software. Data are expressed as mean ± S.E.M. and analyzed by two-way ANOVA with Šidàk post-test.*
To further assess a consequence of changes to goblet cells, we assessed the mucus layer in the colon. Ulex europaeus agglutinin I fluorescein (UEA1) was used to visualize gastrointestinal fucosylated oligosaccharides. This stain detected that the mucus layer was reduced in the WT compared to the PKR-ablated and, to a lesser extent, the kinase-dead mice before exposure to DSS (Figure 7E). However, a substantive induction of mucin production in response to DSS treatment required PKR’s kinase activity (Figure 7E). Accordingly, the kinase activity of PKR appears to alter goblet cell physiology to promote stress-induced mucin production. This initial limitation but subsequent promotion in gut barrier function appears to correlate with the initial sensitivity but overall protection from DSS that was associated with PKR’s kinase activity.
As autophagy has been established to be essential for goblet cell function and because this response is regulated by eIF2α phosphorylation (20–22), we assessed this catabolic process in the different mice. Autophagy in goblet cells was assessed by probing the autophagic marker microtubule-associated protein 1B-light chain 3 (LC3B) in UEA1-positive cells. This measure detected more autophagic puncta in goblet cells from the PKR-ablated compared with either the WT or kinase-dead animals at homeostasis (Figure 7F). DSS treatment markedly induced the accumulation of autophagosomes in all the colon tissues while maintaining the differential between the separate genotypes (Figure 7F). It is important to recognize that this measure of autophagosome formation without a parallel assessment of lysozyme activity doesn’t capture autophagic flux, and so doesn’t detect an increase or decrease in the rate, but merely captures a change in autophagy [23].
Together these data identify that PKR alters the function of goblet cells in the gut, in part, by controlling autophagy. This activity is largely but not entirely dependent on substrate phosphorylation.
## PKR affects the gut flora
As the intestinal mucus layer strongly influences the microbiome, we quantified specific microbes in the feces from WT, PKR-ablated and kinase-dead mice. Stool DNA was purified after 5 days of DSS treatment and used to amplify bacterial 16S rRNA sequences from the putatively colitogenic, gram-negative species Bacteroides and Prevotella, and the gram-positive species Lactobacillus by Q-PCR. Although the amounts of Prevotella and Lactobacillus species were equivalent among the three murine genotypes, Bacteroides species were significantly reduced in PKR mutant mice compared to the WT mice (Figure 8A). Examination of the stool from untreated mice shows that this is a pre-existing difference (Figure 8B). Accordingly, the levels of Bacteroides, which bind and metabolize mucins produced by goblet cells [24], correlate with PKR-dependent effects on mucin production.
**Figure 8:** *PKR affects the gut flora. (A, B) The amounts of Bacteroides, Prevotella and Lactobacillus species in fecal samples from WT, PKR-ablated (Eif2ak2
-/-) and kinase-dead (K271R) mice either (A) treated with DSS or (B) untreated (n=4). The quantities of bacteria were assessed by Q-PCR amplification of species-specific 16S rRNA and are expressed as fold induction of bacterial content compared to WT mice. The data are expressed as mean ± S.E.M. and analyzed by one-way ANOVA with Tukey’s range test.*
## Discussion
We identify that PKR alters gut physiology to modify the response to DSS. Previous studies by Cao et al. and Rath et al. had showed that ablating PKR affects DSS-induced colitis but with discordant outcomes [6, 7]. Our findings generally support those of Cao et al, which showed that PKR is protective against the weight loss from DSS treatment. However, in partial agreement with the findings of Rath et al, we detected an initial increase in DSS-induced tissue damage in mice with active PKR compared to mice that were ablated for the kinase. Consistent with the increased susceptibility, the mucin layer in the colon of mice expressing PKR was reduced at homeostasis. Nonetheless, a substantive induction of mucin in response to DSS required kinase activity and we identify that PKR altered autophagy in goblet cells in a kinase-dependent manner. Accordingly, PKR appeared to suppress barrier function at homeostasis but then promoted mucin production in response to challenge. These findings partly reconcile the previous discrepant findings and endorses a different mechanism of action for PKR in colitis than was asserted in the earlier studies by Cao et al. and Rath et al. Rather than PKR functioning by the UPR, we propose PKR functions by supporting gut barrier function via control of autophagy.
## A common function for different eIF2α kinases
Analogous to the PKR response shown here, investigations by Ravindran et al. showed that ablating another eIF2α kinase, GCN2, worsened weight loss in mice treated with DSS [8]. This encourages the view that there may be a conserved response between different members of this kinase family. The response in GCN2-ablated mice was attributed to the control of inflammasome activity, with the DSS-induced weight loss able to be averted by also ablating inflammasome components or by antagonizing the related cytokine signaling. A similar activity is identified here for PKR and treating Eif2ak2 -/- mice with an inhibitor of the NLRP3 inflammasome averted DSS-induced weight loss. This accords with a previous report identifying NLRP3 functions in DSS-induced colitis [25, 26]. Neudecker et al. showed that, despite the benefit of the NLRP3 inhibitor, it did not reduce the level of IL18 in the colon of DSS-treated mice. This was replicated here, raising the question of how this inhibitor modifies the response. Our experiments implicated inflammasome activity in neutrophils in the response to DSS and numerous other studies support a contribution of neutrophils to colitis [4]. However, the bone marrow chimera experiment conducted by Cao et al. identify that PKR function in epithelial cells is sufficient for the phenotype. Possibly in keeping with this we identify that PKR functions in goblet cells, which have been associated with NLRP3-dependent pathogenesis [27].
Inflammasome activity in GCN2-ablated mice was shown to stem from impaired autophagy in goblet cells [8]. Autophagy is induced by eIF2α phosphorylation and Ravindran et al, with others, confirmed that mutation of the phosphoresidue of eIF2α exacerbated DSS-induced pathogenesis [20, 28]. An earlier study had demonstrated that conditional expression of eIF2α with a mutated phosphoresidue in villus and crypt epithelial cells of the small and large intestine altered the susceptibility to DSS-induced colitis [29]. This furthers the notion that the conserved activity of eIF2α kinases protect against DSS-induced colitis. However, our experiments with transgenic mice that express a kinase-dead PKR identify that the response to DSS is not entirely dependent on substrate phosphorylation. There may be some support for this in other studies, as the effect of mutating the phosphoresidue of eIF2α was less impactful than ablating GCN2 in the study by Ravindran et al. [ 8]. Also, the protective effect of type III interferons in DSS-induced colitis was shown to be independent from the control of translation that is expected by eIF2α phosphorylation upon the induction of PKR expression [4]. Significantly, ablating a third eIF2α kinase, PERK, did not affect the response to DSS treatment [8]. As there is considerable evidence for the induction of the UPR in colitis, this appears to recognize that parallel responses, controlled by the activating transcription factor 6 or inositol-requiring enzyme 1α (IRE1α), can compensate for the loss of PERK independent of eIF2α phosphorylation [30, 31].
## An alternative mechanism of activity for PKR
The primacy of PERK in the eIF2α-mediated UPR, with the apparent ineffectiveness of ablating this kinase, somewhat weakens the proposal by Cao et al. and Rath et al. that PKR impacts DSS-induced colitis by control of the UPR. We propose an alternative mechanism of activity by control of autophagy. Our investigations detected an effect of PKR on goblet cell morphology and the production of mucin, as well as the levels of microbes that metabolize mucins. This equates with the mechanism of activity of GCN2 in DSS-induced colitis and is consistent with a previously identified function of eIF2α kinases in Paneth cells in the small intestine [29]. Both cells are important for establishing barrier function, modulating the microbiota and the ensuing innate and acquired immunity [21, 24, 32]. Notably, Eif2ak2 -/- murine fibroblasts have defective autophagy and the expression of PKR rescued the starvation-induced autophagy in GCN2-disrupted yeast [33, 34]. Accordingly, there is an established overlap in the responses controlled by these related kinases. However, our experiments suggest this can be separated to some extent from eIF2α phosphorylation.
Substrate phosphorylation-independent activities of PKR have been shown to be mediated through an association with the TNF receptor-associated factors (TRAFs) or the heat-shock protein (HSP) 70 and HSP90 (35–38). TRAFs shape signaling complexes and regulate the stability of the protein components by acting as adaptor molecules and ubiquitin ligases. Ablating TRAF proteins that interact with PKR affects autophagy and induces spontaneous colitis in mice (39–42). The HSP70 and HSP90 molecular chaperones also control autophagy and inflammasome activity, modulate DSS-induced colitis in mice, and have been shown to be protective in IBD (43–48). Possibly related to our speculation that other UPR proteins can offset the loss of PERK, IRE1α induces autophagy via TRAF2 and is also controlled by the HSP70 and HSP90 chaperones (49–51).
## Genetic difference as a cause of discrepant findings
The alternative mechanism that we propose may account for the discord between previous studies as the different murine strains used by Cao et al. [ 7] and Rath et al. [ 6] vary in their autophagic responses. The mice used by Cao et al. and ourselves were on an isogenic C57BL6/J background, while those used in the study by Rath et al. were on a mixed 129/terSv/BALB/C background. Autophagy is impaired in the BALB/C relative to the C57BL6/J strain (52–55). Among the consequences of this impairment is that DSS-induced damage to mitochondria would be predicted to accumulate through reduced mitophagy. This activates innate immune sensors, including PKR, that could cause the opposing activity that was reported [6, 55, 56]. In addition to this defect in the BALB/C strain, the 129/terSv background is deleted for caspase-11 expression [57]. Caspase-11 is protective in the context of DSS-induced colitis [58, 59], partly as a result of an autophagy-based secretory pathway for IL1β and IL18 (60–62). As the expression of caspase-11 is induced by PKR, through both eIF2α phosphorylation and kinase-independent signaling (14, 63–65), PKR expression would compound caspase-11-dependent differences in the responses of C57BL6/J and 129/terSv mice.
## Relevance for IBD
Autophagy is important in gastric function, particularly by supporting the function of secretory cells [66]. Accordingly, PKR activity might be elicited to fortify gut barrier function and dampen immune pathogenesis in IBD. Type I and III interferons, which induce PKR expression and autophagy as well as suppressing inflammasome activity, are protective of DSS-induced colitis in mice [4, 5, 67, 68] and so might be trialled as a treatment for IBD. Notably, the more limited expression of the receptors for type III interferons mean that these cytokines are less prone to the contraindications of type I interferons [69, 70]. The identification of kinase-substrate-independent activity in this study suggests therapeutic targets that would not induce the proteostasis that is induced by EIF2α phosphorylation. However, additional experiments are required to identify these targets. Inflammasome inhibitors have been suggested as potential therapies for immune pathogenesis. However, this is complicated by the positive functions of inflammasomes in gut immunity and wound healing that narrows the treatment window. Limiting inflammasome activity via autophagy may provide greater latitude with the broader benefits of cellular quality control. Towards this, autophagic inducers such as rapamycin and resveratrol have been shown to be beneficial in experimentally induced colitis. Other molecules that promote chaperone-mediated autophagy have shown promise in different diseases that share the pathogenic axis of deficient autophagy with elevated inflammasome activity that we propose as an etiology in IBD.
## 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 Monash Medical Centre animal ethics committee and the Institutional Animal Care and Use Committee at the Cleveland Clinic.
## Author contributions
Conceptualization, AS. Methodology, HY, AC, SK, HM, DW, DS, AA and AS. Analysis and interpretation, HY, SK and AS. Manuscript preparation HY and AS. Funding acquisition, CM, RS, BW and AS. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1106737/full#supplementary-material
## References
1. Levin D, London IM. **Regulation of protein synthesis: Activation by double-stranded RNA of a protein kinase that phosphorylates eukaryotic initiation factor 2**. *Proc Natl Acad Sci U S A* (1978) **75**. DOI: 10.1073/pnas.75.3.1121
2. Jeon YJ, Lim JH, An S, Jo A, Han DH, Won TB. **Type III interferons are critical host factors that determine susceptibility to influenza a viral infection in allergic nasal mucosa**. *Clin Exp Allergy* (2018) **48**. DOI: 10.1111/cea.13082
3. Sommereyns C, Paul S, Staeheli P, Michiels T. **IFN-lambda (IFN-lambda) is expressed in a tissue-dependent fashion and primarily acts on epithelial cells in vivo**. *PloS Pathog* (2008) **4**. DOI: 10.1371/journal.ppat.1000017
4. Broggi A, Tan Y, Granucci F, Zanoni I. **IFN-lambda suppresses intestinal inflammation by non-translational regulation of neutrophil function**. *Nat Immunol* (2017) **18**. DOI: 10.1038/ni.3821
5. Katakura K, Lee J, Rachmilewitz D, Li G, Eckmann L, Raz E. **Toll-like receptor 9-induced type I IFN protects mice from experimental colitis**. *J Clin Invest* (2005) **115** 695-702. DOI: 10.1172/JCI22996
6. Rath E, Berger E, Messlik A, Nunes T, Liu B, Kim SC. **Induction of dsRNA-activated protein kinase links mitochondrial unfolded protein response to the pathogenesis of intestinal inflammation**. *Gut* (2012) **61**. DOI: 10.1136/gutjnl-2011-300767
7. Cao SS, Song B, Kaufman RJ. **PKR protects colonic epithelium against colitis through the unfolded protein response and prosurvival signaling**. *Inflammation Bowel Dis* (2012) **18**. DOI: 10.1002/ibd.22878
8. Ravindran R, Loebbermann J, Nakaya HI, Khan N, Ma H, Gama L. **The amino acid sensor GCN2 controls gut inflammation by inhibiting inflammasome activation**. *Nature* (2016) **531**. DOI: 10.1038/nature17186
9. Yang YL, Reis LF, Pavlovic J, Aguzzi A, Schafer R, Kumar A. **Deficient signaling in mice devoid of double-stranded RNA-dependent protein kinase**. *EMBO J* (1995) **14**. DOI: 10.1002/j.1460-2075.1995.tb00300.x
10. Yim HC, Wang D, Yu L, White CL, Faber PW, Williams BR. **The kinase activity of PKR represses inflammasome activity**. *Cell Res* (2016) **26**. DOI: 10.1038/cr.2016.11
11. Kessler SP, Obery DR, de la Motte C. **Hyaluronan synthase 3 null mice exhibit decreased intestinal inflammation and tissue damage in the DSS-induced colitis model**. *Int J Cell Biol* (2015) **2015** 745237. DOI: 10.1155/2015/745237
12. Hornung V, Bauernfeind F, Halle A, Samstad EO, Kono H, Rock KL. **Silica crystals and aluminum salts activate the NALP3 inflammasome through phagosomal destabilization**. *Nat Immunol* (2008) **9**. DOI: 10.1038/ni.1631
13. Okumura R, Kurakawa T, Nakano T, Kayama H, Kinoshita M, Motooka D. **Lypd8 promotes the segregation of flagellated microbiota and colonic epithelia**. *Nature* (2016) **532**. DOI: 10.1038/nature17406
14. Bonnet MC, Weil R, Dam E, Hovanessian AG, Meurs EF. **PKR stimulates NF-kappaB irrespective of its kinase function by interacting with the IkappaB kinase complex**. *Mol Cell Biol* (2000) **20**. DOI: 10.1128/MCB.20.13.4532-4542.2000
15. Dagenais M, Dupaul-Chicoine J, Champagne C, Skeldon A, Morizot A, Saleh M. **A critical role for cellular inhibitor of protein 2 (cIAP2) in colitis-associated colorectal cancer and intestinal homeostasis mediated by the inflammasome and survival pathways**. *Mucosal Immunol* (2016) **9**. DOI: 10.1038/mi.2015.46
16. Coll RC, Robertson AA, Chae JJ, Higgins SC, Munoz-Planillo R, Inserra MC. **A small-molecule inhibitor of the NLRP3 inflammasome for the treatment of inflammatory diseases**. *Nat Med* (2015) **21**. DOI: 10.1038/nm.3806
17. Baltzis D, Li S, Koromilas AE. **Functional characterization of pkr gene products expressed in cells from mice with a targeted deletion of the n terminus or c terminus domain of PKR**. *J Biol Chem* (2002) **277**. DOI: 10.1074/jbc.M203564200
18. Fasciano S, Hutchins B, Handy I, Patel RC. **Identification of the heparin-binding domains of the interferon-induced protein kinase, PKR**. *FEBS J* (2005) **272**. DOI: 10.1111/j.1742-4658.2005.04575.x
19. Nowarski R, Jackson R, Gagliani N, de Zoete MR, Palm NW, Bailis W. **Epithelial IL-18 equilibrium controls barrier function in colitis**. *Cell* (2015) **163**. DOI: 10.1016/j.cell.2015.10.072
20. Humeau J, Leduc M, Cerrato G, Loos F, Kepp O, Kroemer G. **Phosphorylation of eukaryotic initiation factor-2alpha (eIF2alpha) in autophagy**. *Cell Death Dis* (2020) **11** 433. DOI: 10.1038/s41419-020-2642-6
21. Patel KK, Miyoshi H, Beatty WL, Head RD, Malvin NP, Cadwell K. **Autophagy proteins control goblet cell function by potentiating reactive oxygen species production**. *EMBO J* (2013) **32**. DOI: 10.1038/emboj.2013.233
22. Lassen KG, Kuballa P, Conway KL, Patel KK, Becker CE, Peloquin JM. **Atg16L1 T300A variant decreases selective autophagy resulting in altered cytokine signaling and decreased antibacterial defense**. *Proc Natl Acad Sci U S A* (2014) **111**. DOI: 10.1073/pnas.1407001111
23. Tanida I, Minematsu-Ikeguchi N, Ueno T, Kominami E. **Lysosomal turnover, but not a cellular level, of endogenous LC3 is a marker for autophagy**. *Autophagy* (2005) **1** 84-91. DOI: 10.4161/auto.1.2.1697
24. Salyers AA, Vercellotti JR, West SE, Wilkins TD. **Fermentation of mucin and plant polysaccharides by strains of bacteroides from the human colon**. *Appl Environ Microbiol* (1977) **33**. DOI: 10.1128/aem.33.2.319-322.1977
25. Neudecker V, Haneklaus M, Jensen O, Khailova L, Masterson JC, Tye H. **Myeloid-derived miR-223 regulates intestinal inflammation**. *J Exp Med* (2017) **214**. DOI: 10.1084/jem.20160462
26. Zaki MH, Boyd KL, Vogel P, Kastan MB, Lamkanfi M, Kanneganti TD. **The NLRP3 inflammasome protects against loss of epithelial integrity and mortality during experimental colitis**. *Immunity* (2010) **32**. DOI: 10.1016/j.immuni.2010.03.003
27. McGilligan VE, Gregory-Ksander MS, Li D, Moore JE, Hodges RR, Gilmore MS. **Staphylococcus aureus activates the NLRP3 inflammasome in human and rat conjunctival goblet cells**. *PloS One* (2013) **8**. DOI: 10.1371/journal.pone.0074010
28. B'Chir W, Maurin AC, Carraro V, Averous J, Jousse C, Muranishi Y. **The eIF2alpha/ATF4 pathway is essential for stress-induced autophagy gene expression**. *Nucleic Acids Res* (2013) **41**. DOI: 10.1093/nar/gkt563
29. Cao SS, Wang M, Harrington JC, Chuang BM, Eckmann L, Kaufman RJ. **Phosphorylation of eIF2alpha is dispensable for differentiation but required at a posttranscriptional level for paneth cell function and intestinal homeostasis in mice**. *Inflammation Bowel Dis* (2014) **20**. DOI: 10.1097/MIB.0000000000000010
30. Kaser A, Lee AH, Franke A, Glickman JN, Zeissig S, Tilg H. **XBP1 links ER stress to intestinal inflammation and confers genetic risk for human inflammatory bowel disease**. *Cell* (2008) **134**. DOI: 10.1016/j.cell.2008.07.021
31. Brandl K, Rutschmann S, Li X, Du X, Xiao N, Schnabl B. **Enhanced sensitivity to DSS colitis caused by a hypomorphic Mbtps1 mutation disrupting the ATF6-driven unfolded protein response**. *Proc Natl Acad Sci U S A* (2009) **106**. DOI: 10.1073/pnas.0813036106
32. Roberton AM, Stanley RA. *Appl Environ Microbiol* (1982) **43**. DOI: 10.1128/aem.43.2.325-330.1982
33. Talloczy Z, Jiang W, Virgin H, Leib DA, Scheuner D, Kaufman RJ. **Regulation of starvation- and virus-induced autophagy by the eIF2alpha kinase signaling pathway**. *Proc Natl Acad Sci U S A* (2002) **99**. DOI: 10.1073/pnas.012485299
34. Shen S, Niso-Santano M, Adjemian S, Takehara T, Malik SA, Minoux H. **Cytoplasmic STAT3 represses autophagy by inhibiting PKR activity**. *Mol Cell* (2012) **48**. DOI: 10.1016/j.molcel.2012.09.013
35. Gil J, Garcia MA, Gomez-Puertas P, Guerra S, Rullas J, Nakano H. **TRAF family proteins link PKR with NF-kappa b activation**. *Mol Cell Biol* (2004) **24**. DOI: 10.1128/MCB.24.10.4502-4512.2004
36. Horng T, Barton GM, Medzhitov R. **TIRAP: an adapter molecule in the toll signaling pathway**. *Nat Immunol* (2001) **2**. DOI: 10.1038/ni0901-835
37. Donze O, Abbas-Terki T, Picard D. **The Hsp90 chaperone complex is both a facilitator and a repressor of the dsRNA-dependent kinase PKR**. *EMBO J* (2001) **20**. DOI: 10.1093/emboj/20.14.3771
38. Pang Q, Christianson TA, Keeble W, Koretsky T, Bagby GC. **The anti-apoptotic function of Hsp70 in the interferon-inducible double-stranded RNA-dependent protein kinase-mediated death signaling pathway requires the fanconi anemia protein, FANCC**. *J Biol Chem* (2002) **277**. DOI: 10.1074/jbc.M209386200
39. Piao JH, Hasegawa M, Heissig B, Hattori K, Takeda K, Iwakura Y. **Tumor necrosis factor receptor-associated factor (TRAF) 2 controls homeostasis of the colon to prevent spontaneous development of murine inflammatory bowel disease**. *J Biol Chem* (2011) **286**. DOI: 10.1074/jbc.M111.221853
40. Yang KC, Ma X, Liu H, Murphy J, Barger PM, Mann DL. **Tumor necrosis factor receptor-associated factor 2 mediates mitochondrial autophagy**. *Circ Heart Fail* (2015) **8**. DOI: 10.1161/CIRCHEARTFAILURE.114.001635
41. Paul PK, Kumar A. **TRAF6 coordinates the activation of autophagy and ubiquitin-proteasome systems in atrophying skeletal muscle**. *Autophagy* (2011) **7**. DOI: 10.4161/auto.7.5.15102
42. Shi CS, Kehrl JH. **TRAF6 and A20 regulate lysine 63-linked ubiquitination of beclin-1 to control TLR4-induced autophagy**. *Sci Signal* (2010) **3** ra42. DOI: 10.1126/scisignal.2000751
43. Tanaka K, Namba T, Arai Y, Fujimoto M, Adachi H, Sobue G. **Genetic evidence for a protective role for heat shock factor 1 and heat shock protein 70 against colitis**. *J Biol Chem* (2007) **282**. DOI: 10.1074/jbc.M704081200
44. Collins CB, Aherne CM, Yeckes A, Pound K, Eltzschig HK, Jedlicka P. **Inhibition of n-terminal ATPase on HSP90 attenuates colitis through enhanced treg function**. *Mucosal Immunol* (2013) **6**. DOI: 10.1038/mi.2012.134
45. Mayor A, Martinon F, De Smedt T, Petrilli V, Tschopp J. **A crucial function of SGT1 and HSP90 in inflammasome activity links mammalian and plant innate immune responses**. *Nat Immunol* (2007) **8** 497-503. DOI: 10.1038/ni1459
46. Tanaka K, Mizushima T. **Protective role of HSF1 and HSP70 against gastrointestinal diseases**. *Int J Hyperthermia* (2009) **25**. DOI: 10.3109/02656730903213366
47. Martine P, Chevriaux A, Derangere V, Apetoh L, Garrido C, Ghiringhelli F. **HSP70 is a negative regulator of NLRP3 inflammasome activation**. *Cell Death Dis* (2019) **10** 256. DOI: 10.1038/s41419-019-1491-7
48. Piippo N, Korhonen E, Hytti M, Skottman H, Kinnunen K, Josifovska N. **Hsp90 inhibition as a means to inhibit activation of the NLRP3 inflammasome**. *Sci Rep* (2018) **8** 6720. DOI: 10.1038/s41598-018-25123-2
49. Urano F, Wang X, Bertolotti A, Zhang Y, Chung P, Harding HP. **Coupling of stress in the ER to activation of JNK protein kinases by transmembrane protein kinase IRE1**. *Science* (2000) **287**. DOI: 10.1126/science.287.5453.664
50. Gupta S, Deepti A, Deegan S, Lisbona F, Hetz C, Samali A. **HSP72 protects cells from ER stress-induced apoptosis**. *PloS Biol* (2010) **8**. DOI: 10.1371/journal.pbio.1000410
51. Marcu MG, Doyle M, Bertolotti A, Ron D, Hendershot L, Neckers L. **Heat shock protein 90 modulates the unfolded protein response by stabilizing IRE1alpha**. *Mol Cell Biol* (2002) **22**. DOI: 10.1128/MCB.22.24.8506-8513.2002
52. Li CY, Li C, Li H, Zhao GQ, Lin J, Wang Q. **Disparate expression of autophagy in corneas of C57BL/6 mice and BALB/c mice after aspergillus fumigatus infection**. *Int J Ophthalmol* (2019) **12**. DOI: 10.18240/ijo.2019.05.02
53. Breda J, Banerjee A, Jayachandran R, Pieters J, Zavolan M. **A novel approach to single-cell analysis reveals intrinsic differences in immune marker expression in unstimulated BALB/c and C57BL/6 macrophages**. *FEBS Lett* (2022) **596**. DOI: 10.1002/1873-3468.14478
54. Martyniszyn L, Szulc-Dabrowska L, Boratynska-Jasinska A, Badowska-Kozakiewicz AM, Niemialtowski MG. *Pol J Vet Sci* (2013) **16** 25-32. DOI: 10.2478/pjvs-2013-0004
55. Pinheiro RO, Nunes MP, Pinheiro CS, D'Avila H, Bozza PT, Takiya CM. **Induction of autophagy correlates with increased parasite load of leishmania amazonensis in BALB/c but not C57BL/6 macrophages**. *Microbes Infect* (2009) **11**. DOI: 10.1016/j.micinf.2008.11.006
56. Mancini NL, Goudie L, Xu W, Sabouny R, Rajeev S, Wang A. **Perturbed mitochondrial dynamics is a novel feature of colitis that can be targeted to lessen disease**. *Cell Mol Gastroenterol Hepatol* (2020) **10** 287-307. DOI: 10.1016/j.jcmgh.2020.04.004
57. Kayagaki N, Warming S, Lamkanfi M, Vande Walle L, Louie S, Dong J. **Non-canonical inflammasome activation targets caspase-11**. *Nature* (2011) **479**. DOI: 10.1038/nature10558
58. Demon D, Kuchmiy A, Fossoul A, Zhu Q, Kanneganti TD, Lamkanfi M. **Caspase-11 is expressed in the colonic mucosa and protects against dextran sodium sulfate-induced colitis**. *Mucosal Immunol* (2014) **7**. DOI: 10.1038/mi.2014.36
59. Oficjalska K, Raverdeau M, Aviello G, Wade SC, Hickey A, Sheehan KM. **Protective role for caspase-11 during acute experimental murine colitis**. *J Immunol* (2015) **194**. DOI: 10.4049/jimmunol.1400501
60. Dupont N, Jiang S, Pilli M, Ornatowski W, Bhattacharya D, Deretic V. **Autophagy-based unconventional secretory pathway for extracellular delivery of IL-1beta**. *EMBO J* (2011) **30**. DOI: 10.1038/emboj.2011.398
61. Zhang M, Kenny SJ, Ge L, Xu K, Schekman R. **Translocation of interleukin-1beta into a vesicle intermediate in autophagy-mediated secretion**. *Elife* (2015) **4**. DOI: 10.7554/eLife.11205
62. Kimura T, Jia J, Kumar S, Choi SW, Gu Y, Mudd M. **Dedicated SNAREs and specialized TRIM cargo receptors mediate secretory autophagy**. *EMBO J* (2017) **36** 42-60. DOI: 10.15252/embj.201695081
63. Endo M, Mori M, Akira S, Gotoh T. **C/EBP homologous protein (CHOP) is crucial for the induction of caspase-11 and the pathogenesis of lipopolysaccharide-induced inflammation**. *J Immunol* (2006) **176**. DOI: 10.4049/jimmunol.176.10.6245
64. Kobori M, Yang Z, Gong D, Heissmeyer V, Zhu H, Jung YK. **Wedelolactone suppresses LPS-induced caspase-11 expression by directly inhibiting the IKK complex**. *Cell Death Differ* (2004) **11**. DOI: 10.1038/sj.cdd.4401325
65. Schauvliege R, Vanrobaeys J, Schotte P, Beyaert R. **Caspase-11 gene expression in response to lipopolysaccharide and interferon-gamma requires nuclear factor-kappa b and signal transducer and activator of transcription (STAT) 1**. *J Biol Chem* (2002) **277**. DOI: 10.1074/jbc.M207852200
66. Bel S, Hooper LV. **Secretory autophagy of lysozyme in paneth cells**. *Autophagy* (2018) **14**. DOI: 10.1080/15548627.2018.1430462
67. Schmeisser H, Fey SB, Horowitz J, Fischer ER, Balinsky CA, Miyake K. **Type I interferons induce autophagy in certain human cancer cell lines**. *Autophagy* (2013) **9**. DOI: 10.4161/auto.23921
68. Guarda G, Braun M, Staehli F, Tardivel A, Mattmann C, Forster I. **Type I interferon inhibits interleukin-1 production and inflammasome activation**. *Immunity* (2011) **34**. DOI: 10.1016/j.immuni.2011.02.006
69. Davidson S, McCabe TM, Crotta S, Gad HH, Hessel EM, Beinke S. **IFNlambda is a potent anti-influenza therapeutic without the inflammatory side effects of IFNalpha treatment**. *EMBO Mol Med* (2016) **8**. DOI: 10.15252/emmm.201606413
70. Rauch I, Hainzl E, Rosebrock F, Heider S, Schwab C, Berry D. **Type I interferons have opposing effects during the emergence and recovery phases of colitis**. *Eur J Immunol* (2014) **44**. DOI: 10.1002/eji.201344401
|
---
title: Extraction, physicochemical properties, and antioxidant activity of natural
melanin from Auricularia heimuer fermentation
authors:
- Yinpeng Ma
- Piqi Zhang
- Xiaodong Dai
- Xiuge Yao
- Shuyang Zhou
- Qingfang Ma
- Jianing Liu
- Shuang Tian
- Jianan Zhu
- Jiechi Zhang
- Xianghui Kong
- Yihong Bao
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9981798
doi: 10.3389/fnut.2023.1131542
license: CC BY 4.0
---
# Extraction, physicochemical properties, and antioxidant activity of natural melanin from Auricularia heimuer fermentation
## Abstract
### Introduction
Natural melanin from Auricularia heimuer have numerous beneficial biological properties, which were used as a safe and healthy colorant in several industries.
### Methods
In this study, single-factor experiments, Box-Behnken design (BBD), and response surface methodology (RSM) were employed to investigate the effects of alkali-soluble pH, acid precipitation pH, and microwave time on the extraction yield of Auricularia heimuer melanin (AHM) from fermentation. Ultraviolet-visible spectrum (UV-Vis), Fourier transform infrared (FT-IR) spectroscopy, scanning electron microscope (SEM), and high-performance liquid chromatography (HPLC) were used to analyze the extracted AHM. The solubility, stability, and antioxidant activities of AHM were also measured.
### Results
The results showed that alkali-soluble pH, acid precipitation pH, and microwave time significantly affected the AHM yield, with the following optimized microwave-assisted extraction conditions: alkali-soluble pH of 12.3, acid precipitation pH of 3.1, and microwave time of 53 min, resulting in an AHM extraction yield of $0.4042\%$. AHM exhibited a strong absorption at 210 nm, similar to melanin from other sources. FT-IR spectroscopy also revealed that AHM exhibited the three characteristic absorption peaks of natural melanin. The HPLC chromatogram profile of AHM showed a single symmetrical elution peak with a 2.435 min retention time. AHM was highly soluble in alkali solution, insoluble in distilled water and organic solvents, and demonstrated strong DPPH, OH, and ABTS free radical scavenging activities.
### Discussion
This study provides technical support to optimize AHM extraction for use in the medical and food industries.
## 1. Introduction
Melanins are a group of natural pigments found in most organisms [1], which have been widely and conventionally used in various industries [2], due to their antioxidant, anti-radiation, anti-toxic, antitumor, and heavy metal chelation functions (3–7). Natural pigments are considered safe, with pronounced nutritional and therapeutic benefits relative to synthetic pigments [8]. Currently, natural pigments are primarily found in living organisms, including animals, plants, fungi, and bacteria [9].
Auricularia heimuer, the third most important cultivated mushroom in China [10], has high economic and medicinal value [11]. It is known for its bioactive compounds, mostly polysaccharides [12, 13], which have numerous beneficial biological properties, including antioxidant, antitumor, anti-radiation, immunomodulatory, and hyperlipidemic (14–16). Melanin, as one of the main active ingredients of A. heimuer, has been reported to have strong antioxidant, radical scavenging, quorum sensing inhibition, and antibiofilm activities [17]. A. heimuer fruiting bodies are rich in melanin and are increasingly popular as a “black food” in China [18]. Melanin from A. auricula can also be used as a safe and healthy colorant in the food and pharmaceutical industries. Research has previously been conducted on the isolation and characterization of melanin. Our team optimized the conditions for melanin extraction from A. auricula-judae (Hei 29) fruiting bodies using a single-factor experiment and response surface methodology (RSM) [19]. Additional studies have demonstrated the extraction method of melanin from A. auricula-judae [20]. However, it has been difficult to produce melanin from A. heimuer fruiting bodies at the industrial scale due to their long growth cycle and high cost [21]. It is more effective to produce melanin from the fermentation of microorganisms, and A. heimuer is an organism capable of high secretion of natural melanin via submerged fermentation. Zhang et al. [ 22] conducted research for media optimization to enhance the production of melanin by submerged culture of A. auricula. Sun et al. [ 23] optimized the fermentation conditions of natural edible melanin from A. auricula. However, the melanin extraction rate was relatively low.
Microwave-assisted extraction is an effective way to increase metabolites. Among the different extraction methods, microwave-assisted extraction is a predominant and promising method to extract diverse compounds from different materials, due to its unique advantages including reduced extraction time, high yield, and improved quality of end products [24]. Zeng et al. [ 25] determined the influence of microwave-assisted extraction on the characterization and corresponding antioxidant activity of A. auricular polysaccharides. However, there are few reports on microwave-assisted extraction of melanin from A. heimuer fermentation.
In the present study, the process of microwave-assisted extraction of melanin from A. heimuer fermentation was optimized using RSM. In addition, the physicochemical properties and antioxidant activities of A. heimuer melanin (AHM) were investigated in detail. The results provide technical support for the application of AHM in medicine, health food, and food additives.
## 2.1. Strain and growth conditions
The A. heimuer strain 1,703 used in this study was preserved by the Institute of Microbiology, Heilongjiang Academy of Sciences, China. The strain was activated in PDA medium (200 g/L potato, 20 g/L glucose, 2 g/L KH2PO4, 1.5 g/L MgSO4, 18 g/L agar powder) at 25°C for 10 days. The activated strain was cultured in PD medium (200 g/L potato, 20 g/L glucose, 3 g/L peptone, 2 g/L KH2PO4, 1.5 g/L MgSO4, and 10 mg/L vitamin B1) in a rotary shaker incubator at 160 rpm and 25°C for 12 days without light. DPPH, Tris-HCL (pH 8.0), FeSO4, Vitamin C and salicylic acid used in antioxidant assay were purchased from Aladdin Biochemical Technology Co., Ltd. All reagents used in the experiment were of analytical grade.
## 2.2. Melanin extraction and purification
The melanin extraction process was performed as follows: First, the fermentation product was centrifuged at 12,000 rpm for 30 min, and the supernatant was incubated in an SL-SM300 microwave instrument (Nanjing Shunliu Instrument Co., Ltd., China) with a power of 300 W for 50 min for complete extraction. Secondly, the supernatant pH was adjusted to 12 with 3.0 M NaOH and then kept at 70° for 2 h for dissolution, followed by centrifugation at 12,000 rpm for 30 min. Thirdly, the supernatant was transferred to a flask, and then the pH was adjusted to 3.0 with 1.0 M HCl. The supernatant was then kept at 70° for 3 h for precipitation. The crude AHM was obtained after centrifugation at 12,000 rpm for 30 min.
The AHM purification process was performed as follows: The crude AHM was re-dissolved in a 1.0 M NaOH solution and centrifuged at 12,000 rpm for 30 min. The pH of the supernatant was adjusted to 3.0 with 1.0 M HCl, followed by centrifugation at 12,000 rpm for 30 min. Subsequently, the precipitate was washed three times with deionized water, chloroform, ethyl acetate, and absolute alcohol in sequence. Finally, the pure AHM was obtained and dewatered in an FDU-1,200 freeze dryer (EYELA, Tokyo, Japan).
## 2.3. Optimization of AHM extraction and experimental design
The microwave power, microwave time, alkali-soluble pH, and acid precipitation pH were selected as the four variables for AHM extraction optimization. Each variable was individually tested with the following ranges: microwave power 200–350 W, microwave time 20–60 min, alkali-soluble pH 9–13, and acid precipitation pH 2–6.
The Box-Behnken experimental design with three factors and three levels was employed to optimize the extraction conditions in order to obtain the highest melanin yield. Based on the single factor experiments, A, alkali-soluble pH (11, 12, and 13); B, acid precipitation pH (2, 3, and 4); and C, microwave time (40, 50, and 60 min) were determined to be the critical levels with significant effect on melanin extraction. The levels and codes of the variables used in the Box-Behnken design (BBD) are shown in Table 1. The complete design consisted of seventeen combinations including three replicates of the center point.
**TABLE 1**
| Variables | Code | Coded levels | Coded levels.1 | Coded levels.2 |
| --- | --- | --- | --- | --- |
| | | −1 | 0 | 1 |
| Alkali-soluble pH | A | 11 | 12 | 13 |
| Acid precipitation pH | B | 2 | 3 | 4 |
| Microwave time (min) | C | 40 | 50 | 60 |
## 2.4. Ultraviolet-visible spectrum, FT-IR, and SEM assay
The AHM was dissolved in a 0.1 M NaOH solution at a final concentration of 0.05 mg/mL, with 0.1 M NaOH solution as the reference. The UV-visible absorption spectrum (UV-Vis) of AHM was scanned in the wavelength range of 190–800 nm with a UV757CRT UV/VIS Spectrophotometer (Unico Instrument Co., Ltd., Shanghai, China).
The AHM was mixed with potassium bromide (KBr) powder and then pressed into pellets for measurement. The Fourier transform infrared (FT-IR) spectrum was analyzed in the scanning range of 4,000–400 cm–1 using the FT/IR-3000 Spectrometer (Jasco, Tokyo, Japan).
A TM4000 scanning electron microscope (SEM; Hitachi, Tokyo, Japan) was used to investigate the morphological features of AHM. To render the power conductive, the dried AHM was installed on a metal stage and sputtered with gold.
## 2.5. Solubility and stability assay
The solubility of AHM was measured in water, aqueous acid (1.0 M HCl), aqueous alkali (1.0 M NaOH), and several organic solvents (ethanol, chloroform, methanol, and ethyl acetate). First, 10 mg AHM was measured into 10-mL test tubes filled with 1 mL of the chemical reagents mentioned above and then stirred at 25°C for 1 h to dissolve or react thoroughly. The tube was spun at 1,000 rpm for 10 min and then the absorbance of the solution at 210 nm was detected. The solubility of AHM was determined at various pH values adjusted to 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, and 12 using 1 M NaOH and HCl, and the absorbance was measured after all samples stood for 1 h.
To perform stability assays, 10 mg AHM was dissolved in 0.1 M NaOH solution. The heat stability of AHM was measured at different incubation temperatures of 0, 20, 40, 60, and 80°C. The illumination stability of AHM was measured in darkness, natural light, and hard light. The samples were taken at 2, 4, and 6 h, and the absorbance was measured at 210 nm using 0.1 M NaOH solution as reference.
## 2.6. High-performance liquid chromatography analysis
The AHM was compared with a melanin standard (M8631, Sigma-Aldrich, St. Louis, MO, USA) by HPLC following the method reported by Sun et al. [ 26] with minor modification. The AHM and the standard were dissolved in 0.1 M NaOH solution. The chromatographic analysis was developed on an Agilent 1,100 HPLC system (Agilent Technologies, Inc., Santa Clara, CA, USA) with a Waters C18 column (300 mm × 7.8 mm, 5 μm, Milford, MA, USA). The mobile phase consisted of methanol and $1\%$ acetic acid. The flow velocity was 1.0 mL/min and the injection volume was 20 μL. The detection wavelength was set at 210 nm and the column temperature was set at 25°C.
## 2.7. Antioxidant activities
The concentration of AHM mother solution was adjusted to 100 μg/mL. Then AHM mother solution was diluted to different concentrations for antioxidant activity assay. The antioxidant activity of AHM was determined using DPPH, OH, and ABTS free radical scavenging ability assays, which were performed by the methods reported by Ma et al. [ 19], Tian et al. [ 27], and Luo et al. [ 28], with in vitro modifications for this study.
## 2.8. Statistical analysis
Each experiment was repeated three times. All data were presented as the mean ± standard deviation. Design-Expert (Version 21.0, Stat-Ease, Minneapolis, MN, USA) was used for RSM. Statistical analysis was performed with SPSS software (Version 16.0, Chicago, IL, USA). One-way analysis of variance (ANOVA) was used for comparison among groups. Differences were considered statistically significant at $P \leq 0.05.$
## 3.1. Effect of single factors on the extraction yield of AHM
All parameters—microwave power, microwave time, alkali-soluble pH, and acid precipitation pH—were individually investigated for their effect on AHM yield.
The extraction yield of AHM increased as alkali-soluble pH increased from 9 to 12, peaked at 12 ($0.399\%$ yield), and then decreased with increasing pH (Figure 1A). Therefore, an alkali-soluble pH range from 11 to 13 was used in the RSM experiment to optimize extraction conditions. The AHM yield increased as acid precipitation pH increased from 2 to 3, reached a maximum of $0.401\%$ at a pH of 3, and then decreased with increasing pH (Figure 1B). The acid precipitation pH range from 2 to 4 was therefore used for design optimization. The highest yield of $0.380\%$ was reached at a microwave power of 300 W, with no obvious increase in AHM yield as the microwave power continued to increase (Figure 1C). An increase in microwave power has been shown to enhance extraction yield [29]; however, microwave power did not significantly affect the AHM yield. The extraction yield of AHM obviously increased as microwave time increased from 20 to 50 min, reached a maximum yield of $0.399\%$ at 50 min, and then decreased over time (Figure 1D). Compared with microwave for 20 min, the yield of AHM increased from 0.370 to $0.399\%$ after microwave for 50 min. A similar phenomenon was reported in the extraction of *Lachnum singerianum* YM296 (LIM), which showed an $11.08\%$ extraction yield with a microwave time of 118.70 s, which was $40.43\%$ higher than that of alkali and acid precipitation extraction [30]. Previous research found that the yield of IH melanin with ultrasound-assisted extraction increased by $37.33\%$ compared with the non-ultrasonic control group [21]. Compared with no microwave, the yield of AHM increased from 0.305 to $0.399\%$ after microwave for 50 min, which increased by $30.82\%$. Therefore, the microwave-assisted extraction method is an effective way to increase yield of AHM.
**FIGURE 1:** *Effect of single factors on the extraction yield of A. heimuer melanin (AHM). To determine the significant factors that affect AHM yield, a range of alkali-soluble pH (A), acid precipitation pH (B), microwave power (C), and microwave time (D) were individually tested in the AHM extraction process.*
These preliminary experiments enabled the identification of significant factors affecting AHM yield and narrowed down the ranges for these single factors [31]. Ultimately, alkali-soluble pH, acid precipitation pH, and microwave time were confirmed as significant factors that influenced AHM yield.
## 3.2. Response surface methodology analysis
Based on the results of single factor experiments, alkali-soluble pH, acid precipitation pH, and microwave time were selected as independent variables, and AHM yield was used as the dependent variable to obtain the optimal conditions. The experimental results based on BBD design are presented in Table 2. The predicted response Y can be fitted into the following equation:
**TABLE 2**
| Run | A | B | C | AHM yield (%) |
| --- | --- | --- | --- | --- |
| 1 | 1 | 0 | 1 | 0.397 ± 0.012 |
| 2 | 0 | −1 | −1 | 0.374 ± 0.014 |
| 3 | −1 | −1 | 0 | 0.370 ± 0.008 |
| 4 | 0 | 1 | 1 | 0.392 ± 0.015 |
| 5 | 1 | 0 | −1 | 0.390 ± 0.015 |
| 6 | 1 | −1 | 0 | 0.390 ± 0.009 |
| 7 | 0 | 0 | 0 | 0.410 ± 0.011 |
| 8 | 0 | 0 | 0 | 0.407 ± 0.016 |
| 9 | 0 | 0 | 0 | 0.405 ± 0.006 |
| 10 | 1 | 1 | 0 | 0.390 ± 0.009 |
| 11 | −1 | 1 | 0 | 0.380 ± 0.010 |
| 12 | −1 | 0 | −1 | 0.384 ± 0.014 |
| 13 | −1 | 0 | 1 | 0.385 ± 0.008 |
| 14 | 0 | 1 | −1 | 0.386 ± 0.007 |
| 15 | 0 | 0 | 0 | 0.402 ± 0.011 |
| 16 | 0 | −1 | 1 | 0.388 ± 0.014 |
| 17 | 0 | 0 | 0 | 0.400 ± 0.013 |
Where Y is the extraction yield of AHM; and A, B, and C are the codes for alkali-soluble pH, acid precipitation pH, and microwave time, respectively.
The experimental results were analyzed using ANOVA (Table 3). The model F-value of 14.47 combined with the low P-values ($P \leq 0.001$) indicated that the regression model was highly significant ($P \leq 0.01$). The F-value of 0.8599 and P-value of 0.5306 indicated that the “lack-of-fit” was not significant relative to the pure error. The value of determination R2 (0.9490) indicated that the response model can explain $94.90\%$ of the total variations, which suggests a good agreement between the experimental and predicted values. Therefore, it is reasonable to use this regression model to analyze the trends in the responses.
**TABLE 3**
| Source | Sum of squares | df | Mean square | F-value | P-value |
| --- | --- | --- | --- | --- | --- |
| Model | 0.0019 | 9 | 0.0002 | 14.47 | 0.0010** |
| A | 0.0003 | 1 | 0.0003 | 19.52 | 0.0031** |
| B | 0.0001 | 1 | 0.0001 | 5.73 | 0.0480* |
| C | 0.0001 | 1 | 0.0001 | 6.64 | 0.0366* |
| AB | 0.0 | 1 | 0.0 | 1.69 | 0.2343 |
| AC | 9e-06 | 1 | 9e-06 | 0.6099 | 0.4604 |
| BC | 0.0 | 1 | 0.0 | 1.08 | 0.3324 |
| A2 | 0.0004 | 1 | 0.0004 | 23.89 | 0.0018** |
| B2 | 0.0007 | 1 | 0.0007 | 49.34 | 0.0002** |
| C2 | 0.0002 | 1 | 0.0002 | 12.62 | 0.0093** |
| Residual | 0.0001 | 7 | 0.0 | | |
| Lack of fit | 0.0 | 3 | 0.0 | 0.8599 | 0.5306 |
| Pure error | 0.0001 | 4 | 0.0 | | |
| Cor total | 0.002 | 16 | | | |
The factors affecting AHM extraction yield, ranked in decreasing order, are as follows: alkali-soluble pH, acid precipitation pH, and microwave time. As shown in Table 3, the three independent variables (A, B, and C) and the three quadratic terms (A2, B2, and C2) had a significant effect on AHM extraction yield ($P \leq 0.05$), but the interaction terms (AB, AC, and BC) did not ($P \leq 0.05$).
To investigate the interaction of the variables and determine the optimal level of each variable for maximum response, 3D response surfaces and 2D contour plots were generated (Figure 2). The interaction effect of alkali-soluble pH and acid precipitation pH on the extraction yield at a constant microwave time showed that the extraction yield initially increased as the alkali-soluble pH and acid precipitation pH increased, but decreased once the alkali-soluble pH and acid precipitation pH increased past a pH of 12.33 and 3.12, respectively (Figure 2A). The interaction effect of alkali-soluble pH and microwave time on the extraction yield at a constant acid precipitation pH showed that the extraction yield initially increased as the alkali-soluble pH and microwave time increased, but decreased once the alkali-soluble pH and microwave time increased past a pH of 12.33 and 52.64 min, respectively (Figure 2B). The interaction effect of acid precipitation pH and microwave time on the extraction yield at a constant alkali-soluble pH showed that the extraction yield initially increased as the acid precipitation pH and microwave time increased, but decreased when the acid precipitation pH and microwave time increased past a pH of 3.12 and 52.64 min, respectively (Figure 2C).
**FIGURE 2:** *Interaction effects of the three factors that significantly affect A. heimuer melanin (AHM) yield. The 3D response surface map and 2D contour map reveal the optimal levels of alkali-soluble pH, acid precipitation pH, and microwave time on AHM yield based on the interaction effects of these variables at constant alkali-soluble pH (A), constant acid precipitation pH (B), and constant microwave time (C).*
According to these results, this model predicted a maximum AHM yield of $0.4064\%$ with the following optimum AHM extraction conditions: alkali-soluble pH of 12.33, acid precipitation pH of 3.12, and microwave time of 52.64 min. To perform the actual experiments, the optimal extraction conditions from the model were adjusted to alkali-soluble pH of 12.3, acid precipitation pH of 3.1, and microwave time of 53 min. To validate the predicted results, verification experiments were performed in triplicate, resulting in an actual AHM yield of $0.4042\%$, which was slightly lower than the yield predicted by the model. As a result, RSM was found to be an accurate and decisive tool for successfully predicting the optimum response values.
## 3.3. Ultraviolet-visible spectrum, FT-IR, and SEM analysis
The maximum absorption peak of AHM in the UV-Vis absorption spectrum was observed at 210 nm, and the absorbance decreased as the wavelength increased (Figure 3A) due to the complex conjugated structures in the melanin molecules [32]. This was consistent with melanin from Crassostrea gigas [33], A. auricula [26], etc (Table 4). There were no absorption peaks at 260 nm and 280 nm, indicating the absence of nucleic acid and protein in the AHM. Melanin has a maximum absorption peak of 210 nm in the ultraviolet region [34]; therefore, these results are consistent with the UV-Vis absorption characteristics of melanin.
**FIGURE 3:** *Structure and shape of A. heimuer melanin (AHM). The structural components and shape of AHM are shown using the UV-visible absorption spectrum (UV-Vis) (A), Fourier transform infrared (FT-IR) scanning spectra (B), and scanning electron microscope (SEM) photographs (C).* TABLE_PLACEHOLDER:TABLE 4 The characteristic absorption peaks of the pigment are primarily distributed in the following three groups: 3,500∼3,300 cm–1, 1,620∼1,600 cm–1, and 1,150∼1,000 cm–1 [33]. Our results showed that the absorption peaks of AHM between 400 and 4,000 cm–1 were distributed consistently with these three previously reported groups (Figure 3B). The peak at 3,427 cm–1 is attributed to the O-H group, the peak at 1,647 cm–1 is attributed to a benzene ring, and the peak at 1,039 cm–1 is caused by C-O stretching. Overall, these results are consistent with the typical peaks characteristic of melanin, which showed no obvious differences with that from A. auricula [26], Oyster mushroom [35], and *Brevibacillus invocatus* IBA [36] (Table 4).
The SEM image showed that the definite shape of a single AHM molecule is an irregular aggregation of shape and size (Figure 3C), similar to the black and brown sesame melanin samples that exhibited amorphous form without self-organization [37]. Previous studies also found that SEM images of extracted melanin showed irregular shape and size at different magnifications [36]. In the current study, AHM is likely eumelanin based on the results of UV-Vis, FT-IR, and SEM.
## 3.4. Solubility and stability analysis
The solubility assays showed that the absorbance of AHM at 210 nm in NaOH was greater than 1.5, while that in water, HCl, and the tested organic solvents were all close to zero. We found that AHM was insoluble under acidic conditions. Additionally, the solubility of AHM increased as the pH of the solution increased under alkaline conditions (Figure 4A). These results indicate that AHM has relatively high solubility under alkaline conditions but is insoluble in water, HCl, and the tested organic solvents. This solubility characteristic of AHM was very similar to oyster mushrooms [35], Lachnum YM156 [38], B. invocatus strain IBA [36], and other microorganisms [39] (Table 4).
**FIGURE 4:** *Physicochemical properties of A. heimuer melanin (AHM). The solubility of AHM was determined by measuring the absorbance at 210 nm in response to various pH levels (A). The stability of AHM was determined by measuring the absorbance at 210 nm in response to various temperatures over time (B) and light levels over time (C).*
The stability assays showed that the absorbance of AHM decreased over time at constant temperature. Additionally, the absorbance of AHM decreased as temperature increased at the same treatment time. However, there was no significant difference in the absorbance of AHM between different temperatures ($P \leq 0.05$; Figure 4B). Under the same light conditions, the absorbance of AHM decreased over time. At the same treatment time, the absorbance of AHM was different in dark, natural light, and strong light conditions, but the differences were not significant ($P \leq 0.05$; Figure 4C). Overall, the results showed good thermostability and light resistance of AHM, which is consistent with the literature [38, 40] (Table 4).
## 3.5. High-performance liquid chromatography analysis
To better characterize the chemical composition of AHM, HPLC analysis was performed on both AHM and a melanin standard using the Waters system (Figure 5). The AHM chromatogram profile showed a single symmetrical elution peak with a 2.435 min retention time, which was the same retention time as the melanin standard. Furthermore, the AHM peak pattern is similar to that of Sun et al. [ 26]. Collectively, this result indicates that the AHM is only comprised of a single component and does not contain other impurities.
**FIGURE 5:** *Chemical composition of A. heimuer melanin (AHM) using HPLC analysis. HPLC chromatograms of AHM (A) and a melanin standard (B) are shown.*
## 3.6. Antioxidant activities of AHM
The antioxidant capacity of AHM was determined based on the scavenging rate of DPPH, ⋅OH, and ABTS free radicals (Figure 6). The results indicated that the DPPH, OH, and ABTS free radical scavenging ability gradually increased as the AHM concentration increased. Furthermore, AHM exhibited strong DPPH, ⋅OH, and ABTS free radical scavenging ability with IC50 of 26.23, 79.76, and 83.04 μg/mL, respectively, although it was lower than that of Vc at the same concentration (Figure 6). Researchers have previously reported the antioxidant activity of melanin from other natural products. For instance, Liu et al. [ 41] found that both the A. auricula melanin control group and waste residue melanin had strong ABTS, DPPH, and OH scavenging activity. Therefore, AHM has strong antioxidant activity due to its DPPH, OH, and ABTS free radical scavenging ability.
**FIGURE 6:** *The antioxidant activities of A. heimuer melanin (AHM). The scavenging rate of both AHM and Vc against DPPH (A), OH (B), and ABTS (C) free radicals are shown.*
## 4. Conclusion
Natural melanins have increasingly attracted attention for their applications in different fields [42]. Numerous fungal microorganisms have been found to produce melanin in submerged fermentation conditions [43, 44]. A. auricular melanin has higher edible safety and biological activity (45–47). However, most studies have found that the melanin extraction yield from A. heimuer is relatively low. Therefore, the microwave-assisted extraction was used to improve the yield of melanin in fermentation. The optimal extraction parameters of AHM were alkali-soluble pH of 12.3, acid precipitation pH of 3.1 and microwave time of 53 min. Under these optimal conditions, the yield of AHM was $0.4042\%$, indicating that the microwave-assisted extraction of AHM is feasible. AHM is easily soluble in alkaline solution but insoluble in water and organic solvent. Furthermore, AHM is stable to both heat and light. The antioxidant activity assays further proved that AHM has strong DPPH, OH, and ABTS free radical scavenging ability. Collectively, this work provided a scientific basis for AHM extraction for use as an excellent colorant and antioxidant in food products.
## Data availability statement
The original contributions presented in this study are included in this article/supplementary material, further inquiries can be directed to the corresponding authors.
## Author contributions
YM conducted the research and wrote the manuscript. YB and XK designed the research. JZha and PZ analyzed the data. SZ and XD performed the single factor experiments. QM and JL performed the RSM analysis. ST, JZhu, and XY were responsible for the biological activity assays. All authors agreed to the final 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. Jacobson ES, Tinnell S. **Antioxidant function of fungal melanin.**. (1993) **175** 7102-4. DOI: 10.1128/jb.175.21.7102-7104.1993
2. Song S, Li S, Su N, Li J, Shi F, Ye M. **Structural characterization, molecular modification and hepatoprotective effect of melanin from**. (2016) **7** 3617-27. DOI: 10.1039/c6fo00333h
3. Nune M, Manchineella S, Govindaraju T, Narayan K. **Melanin incorporated electroactive and antioxidant silk fibroin nanofibrous scaffolds for nerve tissue engineering.**. (2019) **94** 17-25. DOI: 10.1016/j.msec.2018.09.014
4. Li C, Ji C, Tang B. **Purification, characterisation and biological activity of melanin from**. (2018) **365** 1-8. DOI: 10.1093/femsle/fny077
5. Baraboi V. **Melanin: structure, biosynthesis, biological functions.**. (1999) **71** 5-14
6. Shi F, Li J, Ye Z, Yang L, Chen T, Chen X. **Antitumor effects of melanin from**. (2018) **9** 1059-68. DOI: 10.1039/c8md00035b
7. Wang Z, Wang N, Han D, Yan H. **Characterization of tyrosinase inhibitors in**. (2022) **9** 862773. DOI: 10.3389/fnut.2022.862773
8. Singh S, Nimse S, Mathew D, Dhimmar A, Sahastrabudhe H, Gajjar A. **Microbial melanin: recent advances in biosynthesis, extraction, characterization, and applications.**. (2021) **53** 107773. DOI: 10.1016/j.biotechadv.2021.107773
9. Ma Z, Liu X, Liu Y, Chen W, Wang C. **Studies on the biosynthetic pathways of melanin in Auricularia auricula.**. (2022) **62** 843-56. DOI: 10.1002/jobm.202100670
10. Yuan Y, Wu F, Si J, Zhao Y, Dai Y. **Whole genome sequence of**. (2019) **111** 50-8. DOI: 10.1016/j.ygeno.2017.12.013
11. Fang M, Yao F, Lu L, Zhang Y, Wang P, Lu J. **Complete mitochondrial sequence of**. (2019) **4** 4029-30. DOI: 10.1080/23802359.2019.1688717
12. Chen N, Zhang H, Zong X, Li S, Wang J, Wang Y. **Polysaccharides from**. (2020) **247** 116750. DOI: 10.1016/j.carbpol.2020.116750
13. Kong X, Ma Y, Pan Y, Jiang W, Li D, Chen X. **Extraction optimisation and lipid-lowering activity of**. (2021) **39** 452-9. DOI: 10.17221/146/2020-CJFS
14. Miao J, Regenstein J, Qiu J, Zhang J, Zhang X, Li H. **Isolation, structural characterization and bioactivities of polysaccharides and its derivatives from**. (2020) **150** 102-13. DOI: 10.1016/j.ijbiomac.2020.02.054
15. Li L, Shi F, Li J, Huang Q, Xu C, Yang L. **Immunoregulatory effect assessment of a novel melanin and its carboxymethyl derivative.**. (2017) **27** 1831-4. DOI: 10.1016/j.bmcl.2017.02.046
16. Li S, Yang L, Li J, Chen T, Ye M. **Structure, molecular modification, and anti-radiation activity of melanin from**. (2019) **188** 555-67. DOI: 10.1007/s12010-018-2898-9
17. Hou R, Liu X, Yan J, Xiang K, Wu X, Lin W. **Characterization of natural melanin from**. (2019) **10** 1017-27. DOI: 10.1039/c8fo01624k
18. Zou Y, Xie C, Fan G, Gu Z, Han Y. **Optimization of ultrasound-assisted extraction of melanin from**. (2010) **11** 611-5. DOI: 10.1016/j.ifset.2010.07.002
19. Ma Y, Bao Y, Kong X, Tian J, Han B, Zhang J. **Optimization of melanin extraction from the wood ear medicinal mushroom,**. (2018) **20** 1087-95. DOI: 10.1615/IntJMedMushrooms.2018028694
20. Chen Y, Xu M, Wang X, Shan X, Ji L, Zhang Y. **Preparation of wood ear medicinal mushroom,**. (2021) **23** 89-100. DOI: 10.1615/IntJMedMushrooms.v23.i6.90
21. Hou R, Liu X, Xiang K, Chen L, Wu X, Lin W. **Characterization of the physicochemical properties and extraction optimization of natural melanin from**. (2019) **277** 533-42. DOI: 10.1016/j.foodchem.2018.11.002
22. Zhang M, Xiao G, Thring R, Chen W, Zhou H, Yang H. **Production and characterization of melanin by submerged culture of culinary and medicinal fungi**. (2015) **176** 253-66. DOI: 10.1007/s12010-015-1571-9
23. Sun S, Zhang X, Chen W, Zhang L, Zhu H. **Production of natural edible melanin by**. (2016) **196** 486-92. DOI: 10.1016/j.foodchem.2015.09.069
24. Al-Dhabi N, Ponmurugan K. **Microwave assisted extraction and characterization of polysaccharide from waste jamun fruit seeds.**. (2020) **152** 1157-63. DOI: 10.1016/j.ijbiomac.2019.10.204
25. Zeng W, Zhang Z, Gao H, Jia L, Chen W. **Characterization of antioxidant polysaccharides from**. (2012) **89** 694-700. DOI: 10.1016/j.carbpol.2012.03.078
26. Sun S, Zhang X, Sun S, Zhang L, Shan S, Zhu H. **Production of natural melanin by**. (2016) **190** 801-7. DOI: 10.1016/j.foodchem.2015.06.042
27. Tian H, Liu H, Song W, Zhu L, Yin X. **Polysaccharide from**. (2021) **35** 3417-25. DOI: 10.1080/14786419.2019.1700507
28. Luo Y, Peng B, Wei W, Tian X, Wu Z. **Antioxidant and anti-diabetic activities of polysaccharides from guava leaves.**. (2019) **24** 1343. DOI: 10.3390/molecules24071343
29. Rostami H, Gharibzahedi S. **Microwave-assisted extraction of jujube polysaccharide: optimization, purification and functional characterization.**. (2016) **143** 100-7. DOI: 10.1016/j.carbpol.2016.01.075
30. Lu Y, Ye M, Song S, Li L, Shaikh F, Li J. **Isolation, purification, and anti-aging activity of melanin from**. (2014) **174** 762-71. DOI: 10.1007/s12010-014-1110-0
31. Zou Y, Du F, Hu Q, Wang H. **The structural characterization of a polysaccharide exhibiting antitumor effect from**. (2019) **9** 1724. DOI: 10.1038/s41598-018-38251-6
32. Fu X, Xie M, Lu M, Shi L, Shi T, Yu M. **Characterization of the physicochemical properties, antioxidant activity, and antiproliferative activity of natural melanin from**. (2022) **12** 2110. DOI: 10.1038/s41598-022-05676-z
33. Hao S, Hou X, Wei L, Li J, Li Z, Wang X. **Extraction and identification of the pigment in the adductor muscle scar of pacific oyster**. (2015) **10** e0142439. DOI: 10.1371/journal.pone.0142439
34. Yin C, Yao F, Wu W, Fan X, Chen Z, Ma K. **Physicochemical properties and antioxidant activity of natural melanin extracted from the wild wood ear mushroom,**. (2022) **24** 67-82. DOI: 10.1615/IntJMedMushrooms.2021041894
35. Zhang Y, Wu X, Huang C, Zhang Z, Gao W. **Isolation and identification of pigments from oyster mushrooms with black, yellow and pink caps.**. (2022) **372** 131171. DOI: 10.1016/j.foodchem.2021.131171
36. Ammanagi A, Shivasharana C, Krishnaveni R, Badiger A, Ramaraj V. **Functional and structural characterization of melanin from**. (2021) **500** 159-69. DOI: 10.1134/S001249662105001X
37. Dossou S, Luo Z, Wang Z, Zhou W, Zhou R, Zhang Y. **The dark pigment in the sesame (**. (2022) **9** 858673. DOI: 10.3389/fnut.2022.858673
38. Yang L, He Y, Li J, Gao X, Chen T, Ye M. **Properties of melanin from**. (2020) **92** 244-51. DOI: 10.1016/j.procbio.2020.01.016
39. Liu R, Meng X, Mo C, Wei X, Ma A. **Melanin of fungi: from classification to application.**. (2022) **38** 228. DOI: 10.1007/s11274-022-03415-0
40. Gao L, Yang L, Guo L, Wang H, Zhao Y, Xie J. **Improving the solubility of melanin nanoparticles from apricot kernels is a potent drug delivery system.**. (2022) **20** 22808000221124418. DOI: 10.1177/22808000221124418
41. Liu X, Hou R, Wang D, Mai M, Wu X, Zheng M. **Comprehensive utilization of edible mushroom**. (2019) **7** 3774-83. DOI: 10.1002/fsn3.1239
42. Tran-Ly A, Reyes C, Schwarze F, Ribera J. **Microbial production of melanin and its various applications.**. (2020) **36** 170. DOI: 10.1007/s11274-020-02941-z
43. Zhang F, Xue F, Xu H, Yuan Y, Wu X, Zhang J. **Optimization of solid-state fermentation extraction of**. (2021) **10** 2893. DOI: 10.3390/foods10122893
44. De Souza R, Kamat N, Nadkarni V. **Purification and characterisation of a sulphur rich melanin from edible mushroom**. (2018) **9** 296-306. DOI: 10.1080/21501203.2018.1494060
45. Lin Y, Chen H, Cao Y, Zhang Y, Li W, Guo W. (2021) **10** 2436. DOI: 10.3390/foods10102436
46. Li J, Li Z, Zhao T, Yan X, Pang Q. **Proteomic analysis of**. (2021) **11** 610173. DOI: 10.3389/fmicb.2020.610173
47. Sun X, Yang C, Ma Y, Zhang J, Wang L. **Research progress of**. (2022) **13** 1048249. DOI: 10.3389/fmicb.2022.1048249
|
---
title: Signaling pathway of targeting the pancreas in the treatment of diabetes under
the precision medicine big data evaluation system
authors:
- Ge Song
- Yiqian Zhang
- Yihua Jiang
- Huan Zhang
- Wen Gu
- Xiu Xu
- Jing Yao
- Zhengfang Chen
journal: Frontiers in Genetics
year: 2023
pmcid: PMC9981801
doi: 10.3389/fgene.2023.1119181
license: CC BY 4.0
---
# Signaling pathway of targeting the pancreas in the treatment of diabetes under the precision medicine big data evaluation system
## Abstract
Diabetes is a chronic noncommunicable disease, which is related to lifestyle, environmental and other factors. The main disease of diabetes is the pancreas. Inflammation, oxidative stress and other factors can interfere with the conduction of various cell signaling pathways, thus inducing pancreatic tissue lesions and diabetes. Precision medicine covers epidemiology, preventive medicine, rehabilitation medicine and clinical medicine. On the basis of precision medicine big data analysis, this paper takes pancreas as the target to analyze the signal pathway of diabetes treatment. This paper analyzes from the five aspects of the age structure of diabetes, the blood sugar control standard of type 2 elderly diabetes mellitus, the changes in the number of diabetic patients, the ratio of patients using pancreatic species and the changes in blood sugar using the pancreas. The results of the study showed that targeted pancreatic therapy for diabetes reduced the diabetic blood glucose rate by approximately $6.94\%$.
## 1 Introduction
The organic combination of precision medicine and big data technology is conducive to the analysis and sharing of medical data, so as to better meet the needs of medical development and lay a solid foundation for the development of related medical research. Diabetes is an endocrine inflammatory disease. At present, the treatment presents the characteristics of multiple pathways and multiple targets. The pancreas is an important target for diabetes treatment. The application of precision medicine big data to the research on the treatment of diabetes with the pancreas as the target is conducive to promoting the progress of diabetes medicine.
The incidence of diabetes is getting higher and higher, and many scholars have studied it. Bensellam M presented the identified molecular mechanisms involved in the dedifferentiation of cells in the diabetic pancreas that produce and release the hormone insulin. The roles and inhibitors of differentiation proteins were discussed and the emerging role of non-coding RNAs (Ribonucleic Acid) was highlighted (Bensellam et al., 2018). Tuttolomondo A studied several factors that affect overall cerebrovascular risk to varying degrees in people with diabetes. Diabetes may lead to more insidious brain damage represented by lacunar infarcts, which would increase the risk of dementia and lead to a dramatic decline in cognitive function (Tuttolomondo, 2018). Hu C believed that the decreased ability of insulin to promote the processing and storage of glucose in muscle is due to impaired activation of glycogen synthase. Decreased glucose storage may occur due to decreased glucose uptake, and safety has been cognitively impaired secondary (Hu and Jia, 2018). In the Cardiovascular Health Study, Cho N H assessed the relationship between patients with subclinical cardiovascular disease, patients with diabetes and impaired glucose tolerance and normal subjects and the risk of clinical vascular disease (Cho et al., 2018). Liu believed the prevalence of diabetic lesions is high, and these complications are often associated with poor medication adherence and uncontrolled diabetes. The aim of the study was to determine medication adherence in patients with uncontrolled diabetes and to compare characteristics and identified barriers between patients with good and poor adherence to medication (Liu et al., 2018). Lane W mentioned that people with diabetes are prone to foot ulcers. If these ulcers do not heal, the patient may also undergo foot amputation, with diabetes preventing postoperative wound healing (Lane et al., 2017). Feig D S investigated whether microalbuminuria predicts later development of increased proteinuria and early mortality in patients with type 2 diabetes. Morning urine samples from diabetes clinic patients aged 50–75 years were examined by radioimmunoassay (Feig et al., 2018). Although there are many studies on the theory of diabetes, further research is needed on the treatment of diabetes.
The pancreas-targeted treatment of diabetes is widely used in medicine. Research by Marathe P H demonstrated that permanent neonatal diabetes may be caused by a complete lack of glucokinase activity. It reported three new cases of glucokinase-related permanent neonatal diabetes. Autosomal recessive inheritance and enzyme deficiencies are typical features of inborn errors of metabolism that occur in the glucose-insulin signaling pathway in these subjects (Marathe et al., 2017). Ogurtsova K explored the etiology of diabetes-related cognitive decline. The etiology involves insulin receptor downregulation, neuronal apoptosis, and glutamatergic neurotransmission (Ogurtsova et al., 2017). Bragg F studied the role of signal transduction and transcriptional activator protein signaling pathway in autoimmune diabetes (Bragg et al., 2017). Rowan J A provided a new strategy for diabetes prevention and treatment by studying the vascular endothelial cell nuclear factor signaling pathway in diabetes (Rowan et al., 2018). Wang Q mentioned that diabetes impairs the mobilization of hematopoietic stem cells from the bone marrow, thereby worsening the outcome of hematopoietic stem and progenitor cell transplantation and diabetic complications (Wang et al., 2017). Willeit explored the potential effects of curcumin on cardiomyocyte hypertrophy, possible mechanisms of nuclear transcription factor signaling in diabetes, hyperglycemia- and insulin-induced cardiomyocyte hypertrophy, and antihypertrophic effects of curcumin in primary culture (Willeit et al., 2017). Chamberlain JJ mentioned that insulin acutely controls metabolism in adipocytes, but also nuclear transcription via proline-directed serine/threonine kinase-mediated “mitotic” signaling (Chamberlain et al., 2017). Although pancreas-targeted therapy for diabetes is widely used in medicine, there are still problems in its application.
By combining computer technology with medicine, new knowledge can be discovered and new modalities of diagnosis and treatment can be created. There have been many research reports on diabetes signaling pathways. This article took pancreatic islets as the target, and systematically discussed the related signaling pathways. At the same time, through the use of precision medicine big data analysis, some complex information was processed, so as to realize the diagnosis and prediction of diabetes.
## 2.1 Precision medicine big data
Precision medicine is an upgraded version of personalized medicine. According to the biological characteristics of patients, especially the data of genomics, and through modern technology, it provides accurate prevention, diagnosis and treatment for the clinic. It can not only improve the treatment effect, but also prevent inappropriate treatment, excessive treatment, waste of resources and other phenomena, and save medical expenses (Riddell et al., 2017). The implementation of precision medicine is based on the development of gene sequence detection technology, and on the basis of network and big data technology. This technology is able to treat patients with minimal medical damage and minimal medical costs, and it not only restores their physical functions, but also ensures their mental health. The big data application of precision medicine is shown in Figure 1.
**FIGURE 1:** *Big data applications in precision medicine.*
## 2.1.1 Medical information system
In fact, medical information systems did not appear in recent years, but because of the complexity of medical services and the small coverage of medical information systems, they have not been widely used. There is a large amount of different medical data in different institutions, systems or databases. The challenge is how to integrate these data in an orderly manner without violating existing laws, regulations, and ethics; and how to conduct automatic computer analysis and mining without violating existing laws, regulations, and ethics. The construction and application of a high-quality medical information system is still an important prerequisite for the production of big data, and it is also an area where companies in the industry invest the most.
## 2.1.2 Clinical auxiliary decision-making system
Big data is not to replace experienced doctors, but to allow more doctors to make high-quality judgments before they have enough experience, which is also an important role of precision medicine. This technology can help doctors diagnose and treat patients’ illnesses by analyzing various medical data such as patients’ medical records, test results, medical images, etc.
## 2.1.3 Data storage and processing system
The development of the big data industry is inseparable from the support of infrastructure, and many companies have developed their own storage and processing platforms. However, the mainstream applications of software services are still on basic communication software such as cloud mailboxes and communication platforms. More software services in the medical field are also needed to improve the research and develop efficiency of pharmaceutical companies and strengthen the management of patients and clinical data by physicians.
## 2.1.4 Health information management system
At present, many health management systems have emerged in mobile health management, such as diabetes management, sleep quality management, and intestinal health management. In fact, they all rely on a large number of data processing and high-quality sensors to realize real-time monitoring of data, but many systems have no way to intervene in time when dealing with exceptions. However, big data is like a chicken-and-egg process. In the face of massive data, the correlation would gradually become prominent, and corresponding auxiliary decision-making would emerge. Big data technology can provide a health information management platform with information collection, analysis and processing functions by collecting various medical data files of patients in medical institutions and sharing various medical data through network technology.
## 2.2 Diabetes
Diabetes is a metabolic disease characterized by hyperglycemia. Hyperglycemia is caused by insufficient secretion of insulin and its biological function is impaired (Johal et al., 2017). Diagnosing diabetes is generally not difficult. The diagnosis can be made if the fasting blood glucose is greater than or equal to 7.0 mmol/L, or greater than or equal to 11.1 mmol/L within 2 h after a meal. The diagnosis is divided into type 1 and type 2 diabetes. The differential diagnosis of diabetes is shown in Figure 2.
**FIGURE 2:** *Diagnosis of diabetes.*
## 2.2.1 Liver disease
Patients with liver cirrhosis usually have abnormal glucose metabolism, which is usually fasting or hypoglycemia, and the blood sugar would rise rapidly after meals. Patients with prolonged illness, fasting blood sugar would also increase.
## 2.2.2 Chronic renal insufficiency
There would be a slight abnormal glucose metabolism. Patients need to use some medications under the guidance of medical professionals to protect the kidney and repair the glucose metabolism function.
## 2.2.3 Stressed state
Under stressful conditions, such as heart, cerebrovascular accident, acute infection, trauma, etc., blood sugar would be excessively increased, and it would return to normal within 1–2 weeks after the stress factor disappears (Sorli et al., 2017).
## 2.2.4 Various endocrine diseases
Glucagonoma is a secondary cause of diabetes mellitus and has other characteristics besides hyperglycemia.
## 2.3 Concepts related to pancreas
In the upper abdomen of the human body, there is a small organ that is difficult to find, and that is the pancreas. Although the pancreas is small, it is very powerful and is one of the most important parts of the human body. The pancreas is a gland with endocrine and endocrine functions, and its physiological function and pathological changes are closely related to human health.
The pancreas is located retroperitoneally. In terms of secretion, although the pancreas is small, it contains many endocrine cells. These cells are also adjusting the physiological functions of the body during the process of digestion and absorption. If these cells change and secrete too much or too little, disease can result (Mario et al., 2017). The pancreas produces various digestive enzymes and insulin in the body to help the body break down proteins and other substances and lower blood sugar. Pancreatic juice is an exocrine substance that mainly includes alkaline bicarbonate and various digestive enzymes.
## 3 Pancreatic tissue as a target to treat diabetes signaling pathway
Diabetes is an endocrine inflammatory disease, and its clinical manifestations are multi-channel and multi-target regulation. The pancreas is the main lesion of diabetes, and it is also an important target for the treatment of diabetes. Its main functions are to promote the synthesis of hepatic glycogen and myoglycogen, the absorption and activation of glucose, and the inhibition of sugar xenobiogenesis. Figure 3 shows the relevant signaling pathways targeting pancreatic tissue.
**FIGURE 3:** *Pancreatic tissue as a target for the treatment of diabetes-related signaling pathways.*
## 3.1.1 Calcium atom channels and ATP-sensitive potassium channels
The cells in the pancreas that produce and release the hormone insulin are endocrine cells with electrical excitation. Its secretion is mainly due to an increase in intracellular calcium concentration, which allows it to produce insulin. L-type calcium channels are the main channel for glucose-induced insulin. Its variation may cause type 2 diabetes, and the state of energy metabolism in the body also affects the electrical activity of sensitive potassium channels. Under physiological conditions, glucose enters the pancreas to produce energy from cellular metabolism that produces and releases the hormone insulin. Adenosine triphosphate closes sensitive potassium channels in the cell membrane, opening calcium channels in the cell, thereby activating cells in the pancreas that produce and release the hormone insulin to secrete insulin (Kiran et al., 2017). Therefore, calcium channels and sensitive potassium channels may be effective drug targets for type 2 diabetes.
## 3.1.2 β-cell GLP-1 receptor signaling pathway
Glucagon-like peptide-1 (GLP-1) is a glucose concentration-based polypeptide hormone. It can regulate the gene expression, synthesis and secretion of cells in the pancreas that produce and release the hormone insulin, and can promote insulin secretion and inhibit insulin cell apoptosis. GLP-1 receptors are mainly distributed in the pancreas. After GLP-1 binds to its receptor, it can activate adenylate cyclase, thereby increasing intracellular cyclic adenylate and activating downstream protein kinases. GLP-1 activates the calcium atom signaling pathway through the acidity coefficient and cyclic adenylate binding protein pathway to promote the release of insulin.
## 3.2 Improvement of insulin signaling pathway in pancreatic islet cells by reducing glucagon secretion
Glucagon is a polypeptide hormone produced by pancreatic islet cells, and its abnormal secretion and metabolism may be related to the pathogenesis of type 2 diabetes. Under physiological conditions, cells in the pancreas that produce and release the hormone insulin block alpha cell glucagon production by paracrine (Arnold et al., 2018). The insulin receptor substrate, phosphatidylinositol kinase, is also expressed on cells, and insulin inhibits the gene expression of glucagon in islet cells through signal channels. In conclusion, the improvement of insulin resistance and the reduction of glucagon production is a new therapeutic approach.
## 3.3 Anti-islet beta cell apoptosis
The MAPK (mitogen-activated protein kinase signaling) pathway has a role in regulating cell growth, differentiation and apoptosis in mammalian cells. Different MAPK signaling pathways can be utilized by different extracellular stimuli, and their regulation can modulate various cellular responses. Stress-activated protein kinases and cell signaling cascades have important roles in cellular stress and cellular inflammatory factors. The results show that streptosporin can inhibit the occurrence of type 1 diabetes and reduce the phosphorylation level of its kinase, thereby reducing its upstream kinase activity.
## 4.1 Diabetes risk input expression
The diabetes risk input expression is a model that combines the user’s risk index data (including past and present data). Since a patient has multiple medical records, multiple medical records appear in the same user’s medical records. Therefore, it is necessary to comprehensively consider the weight of each medical record, that is, the importance of the sub-medical record.
Definition 1:collection of risk indicatorsIt is assumed that user Gi has lb sub-records (with time stamps), then all sub-records of Qi can form a set of risk indicators, and the formula is: Qi=kab,fabb∈1,2,…,lb [1] Among them, kab is the time, and fab is the feature vector of each index component; the entire data set can be represented as Q=Qia∈1,2,…,n, and n is the number of patients. For the b -th record kab,fab in Qi, its decay weight formula is: Phb=e−∞b/∑$k = 0$l−1e−∞k [2] Among them, ∞ is the adjustment weight range, and l is the time span.
Definition 2:risk input expressionThe weighted average of each inspection record of 1 Gi based on the decay weight can be extracted as an input expression, which is the risk input expression: xa=∑$b = 1$laPhla−bfab [3] It can be obtained that the risk index set kab,fabb∈1,2,…,la of user Gi can be normalized in the form of a matrix into the form of vector xa, which can be used as the standard input of the calculation model. The model is based on the support vector machine algorithm. Parameter r,g represent function fx, and its calculation method: fx=rt+g [4] In the hyperplane, all the diabetes sample spaces are divided into two groups: one with diabetes and one without diabetes. Through mathematical transformation, it is the solution target: minr,g,ζ=12rTr+C∑$k = 1$mζk [5] yarTϕxa+g≥1,$a = 1$,…,m [6] m is the number of training samples, and ya is the class label.
Definition 3:diabetes indexThe Diabetes Risk Index (DI) represents the risk of developing diabetes and has a value in the range [0,1]. The corresponding DI value can be obtained by probabilistic correction of the value of the judgment function. Probabilistic calibration is performed: Ry=1x=$\frac{1}{1}$+expSdx+F [7] Among them, S,F are conversion parameters. DI is calculated as: DIx=Ry=1x [8]
## 4.2 Diabetes risk model based on genetic factors
The inheritance of diabetes is determined by the relatedness of family members, so a suitable indicator should be used to measure the genetic association of two people.
## 4.2.1 Genetic coefficient
In kinship, the blood of the child is inherited from the parents, so the genetic factor between the parent and the child is $\frac{1}{2.}$ Since the genetic factors represented by each side are all $\frac{1}{2}$, the calculation formula of the genetic factors of the direct genetic relationship can be obtained: kX,$Y = 1$/2M [9] Through the calculation of the direct genetic relationship, the paragenetic genetic factors can be obtained. First, it is necessary to find out the most recent common ancestors N1 and N2 of members X and Y, and calculate the genetic coefficients of X,Y and each ancestor separately. After that, the sum is done to get the heritability coefficients of X and Y: kX,Y=∑$i = 12$kX,NkY,N [10] In Formula [10], n is the total number of diabetes-positive family members of user R.
## 4.2.2 Dynamic blood glucose expression
Ambulatory blood glucose sequence g=g1,g2,…gn represents a time series of blood glucose values. The data yh at time h can be predicted from the blood glucose value at time h−1,h−2,…h−n, and there is a complex non-linear relationship f⋅ between them, which is expressed by Formula [12]: yh=fyh−1,yh−2,…,yh−n [12] *It is* assumed that the number of visible units and hidden units are m and n, respectively, then z=z1,z2,…zm, and x=x1,x2,…xn. b represents the bias of the visible layer, and c represents the bias of the hidden layer. The energy equation of the model is: Wz,xθ=−bzT−cxT−vwxT [13] *It is* expressed in components as: Wz,xθ=−∑$i = 1$mbizi−∑$i = 1$ncjxj−∑$j = 1$n∑$i = 1$mziWijxj [14] *The formula* for calculating the joint probability distribution of the visible layer and the hidden layer is: Tz,xθ=1Bθe−wz,xθ [15] Bθ=∑z,xe−wz,xθ [16]
The marginal probability distribution of ambulatory blood glucose sequence z to x is: Tzθ=1Bθ∑xe−wz,xθ [17] When the visible layer unit state is given, the activation probability of the xi -th hidden layer unit zi is solved as: Txi,zi,θ=sigmbj+∑$i = 1$mxiWij [18] Since cells are bidirectional, visible cells can be activated through hidden cells. The calculation method of the activation probability of the i -th visible unit zi is: Tzi,x,θ=sigmcj+∑$i = 1$mWijxi [19] By experimenting with predictive models and evaluating them predictively, the formula for calculating the root mean square error is: RMSE=∑$a = 1$nxa−x^a2n [20]
In Formula [20], xa is the predicted value of the a -th model; x^a is the actual value of the a -th model, and n is the capacity of the sample set.
## 5.1 Experimental method
Under the precision medicine big data analysis system, this paper studied the signaling pathway of treating diabetes by targeting the pancreas. City A was selected as the research object, and the research was carried out from five aspects: the age structure of diabetes mellitus in 2015–2018, the blood sugar control standard of type 2 diabetes in the elderly, the change in the number of diabetic patients, the proportion of patients using pancreatic types, and the proportion of patients using pancreatic types.
## 5.2.1 Age structure of diabetes
The probability of disease in each age stage is different, and the age structure of diabetic patients in city A was investigated. The results are shown in Figure 4.
**FIGURE 4:** *Age structure of diabetes.*
In the age structure of diabetes shown in Figure 4, the higher the female age, the greater the probability of developing diabetes. Men were more likely to develop diabetes at the age of 60–69, and there were more patients with diabetes over the age of 50. It can be seen that age is closely related to the occurrence of diabetes.
## 5.2.2 The blood sugar control standard of type 2 elderly diabetes mellitus
Older people often experience symptoms of hypoglycemia. In addition, elderly diabetic patients are prone to arteriosclerosis and cardiovascular disease. Hypoglycemia can lead to complications such as cerebral infarction and myocardial infarction. The blood sugar control standards for type 2 elderly diabetes mellitus are shown in Table 1.
**TABLE 1**
| Health status | Reasonable saccharification (%) | Fasting or preprandial blood glucose (oI/L) | Blood sugar before bed (I/L) |
| --- | --- | --- | --- |
| healthy | <7.5 | 5.0–7.2 | 5.0–8.3 |
| moderate health | <8.0 | 5.0–8.3 | 5.6–10.0 |
| poor health | <8.5 | 5.6–10.0 | 6.1–11.1 |
For elderly diabetic patients, the efficacy and risk should be comprehensively considered, and the main purpose is to improve their quality of life. In addition, personalized blood glucose control indicators should be developed to improve the health of the elderly.
## 5.2.3 Changes in the number of diabetic patients
Changes in the number of diabetic patients as the research subject, the number of diabetic patients in city A was investigated from 2015 to 2018. The specific results are shown in Figure 5.
**FIGURE 5:** *Number of people with diabetes.*
Among the changes in the number of diabetic patients shown in Figure 5, the ratio of the number of diabetic patients showed an upward trend from 2015 to 2018. Among them, the proportion of diabetic patients in 2015 was about $28.76\%$; in 2018, the proportion of diabetic patients was about $37.57\%$, with an increase of $8.81\%$. It can be seen that in life, people should maintain a good diet and routine, and exercise regularly. With regular checkups, people can reduce the chances of developing diabetes.
## 5.2.4 Ratio of patients using pancreatic species
The main body of the investigation was the proportion of patients using pancreatic types in area A, and studies were carried out from long-acting, quick-acting, premixed, short-acting, and intermediate-acting. The results are shown in Figure 6.
**FIGURE 6:** *The proportion of patients using insulin type.*
In the proportion of patients using pancreatic types shown in Figure 6, more patients used pancreatic long-acting, fast-acting and premixed, and about $34.58\%$ used long-acting patients; there was about $29.42\%$ of the patients used quick-acting, and $23.25\%$ of the patients used premix. The number of patients using pancreatic short-acting and intermediate-acting was relatively small; the number of patients using short-acting was about $7.21\%$, and the number of patients using intermediate-acting was about $5.54\%$.
## 5.2.5 Changes in blood sugar using the pancreas
The pancreas affects glucose metabolism in the human body and plays an important role in diabetes. Figure 7 shows the results of the research content of the pancreas-targeted treatment of diabetic blood sugar changes.
**FIGURE 7:** *Blood glucose changes using the pancreas.*
In the pancreas-targeted treatment of diabetes shown in Figure 7, blood glucose rates continued to decline from 2015 to 2018. In 2015, the blood sugar rate of pancreas-targeted treatment of diabetes was about $40.75\%$; in 2018, the blood sugar rate of pancreatic-targeted treatment of diabetes was about $33.81\%$, with a decrease of $6.94\%$. It can be seen that the pancreas would secrete insulin, which can regulate the level of glucose in the blood and brings blood sugar back to normal.
## 6 Conclusion
Through the analysis and sharing of precision medicine big data, it would help the development of current medical technology and improve the level of medical treatment. This paper systematically discussed the diabetes signaling pathway targeting the pancreas, which has certain reference value for guiding the development of new drugs and the use of a variety of clinical drugs. The pancreas is mainly used to treat diabetes, and its main function is to stimulate pancreatic secretion and inhibit the release of glucagon. Apoptosis of the pancreas is an important factor affecting the amount of insulin secreted, and apoptosis of the pancreas is related to many signaling pathways. Therefore, the signal pathway related to diabetes explored from the perspective of the pancreas has a certain guiding effect on the development and clinical treatment of new drugs for diabetes.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.
## Author contributions
All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Arnold S. E., Arvanitakis Z., Macauley-Rambach S. L., Koenig A. M., Wang H. Y., Ahima R. S.. **Brain insulin resistance in type 2 diabetes and Alzheimer disease: Concepts and conundrums**. *Nat. Rev. Neurol.* (2018) **14** 168-181. DOI: 10.1038/nrneurol.2017.185
2. Bensellam M., Jonas J. C., Laybutt D. R.. **Mechanisms of β-cell dedifferentiation in diabetes: Recent findings and future research directions**. *J. Endocrinol.* (2018) **236** R109-R517. DOI: 10.1530/JOE-17-0516
3. Bragg F., Holmes M. V., Iona A., Guo Y., Du H., Chen Y.. **Association between diabetes and cause-specific mortality in rural and urban areas of China**. *JAMA J. Am. Med. Assoc.* (2017) **317** 280-289. DOI: 10.1001/jama.2016.19720
4. Chamberlain J. J., Kalyani R. R., Leal S., Rhinehart A. S., Shubrook J. H., Skolnik N.. **Treatment of type 1 diabetes: Synopsis of the 2017 American diabetes association standards of medical care in diabetes**. *Ann. Intern. Med.* (2017) **167** 493-498. DOI: 10.7326/M17-1259
5. Cho N. H., Shaw J. E., Karuranga S., Huang Y., da Rocha Fernandes J. D., Ohlrogge A. W.. **IDF Diabetes Atlas: Global estimates of diabetes prevalence for 2017 and projections for 2045**. *Diabetes Res. Clin. Pract.* (2018) **138** 271-281. DOI: 10.1016/j.diabres.2018.02.023
6. Feig D. S., Donovan L. E., Corcoy R., Murphy K. E., Amiel S. A., Hunt K. F.. **Continuous glucose monitoring in pregnant women with type 1 diabetes (conceptt): A multicenter international randomised controlled trial**. *Obstetrical Gynecol. Surv.* (2018) **73** 199-201. DOI: 10.1097/01.ogx.0000532199.80944.24
7. Hu C., Jia W.. **Diabetes in China: Epidemiology and genetic risk factors and their clinical utility in personalized medication**. *Diabetes* (2018) **67** 3-11. DOI: 10.2337/dbi17-0013
8. Johal S. K., Jamsen M. J., Bell S., Mc Namara K. P., Magliano D. J., Liew D.. **Do statin users adhere to a healthy diet and lifestyle? The Australian diabetes, obesity and lifestyle study**. *Eur. J. Prev. Cardiol.* (2017) **24** 621-627. DOI: 10.1177/2047487316684054
9. Kiran B., Mohanalakshmi T., Srikumar R.. **C-reactive protein and other markers of inflammation in the prediction of cardiovascular disease in diabetes**. *Int. J. Res. Pharm. Sci.* (2017) **8** 476-479
10. Lane W. T., Bailey S., Gerety G., Gumprecht J., Philis-Tsimikas A., Hansen C. T.. **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
11. Liu Jo, Dumville J. C., Hinchliffe R. J., Cullum N., Game F., Stubbs N.. **Negative pressure wound therapy for treating foot wounds in people with diabetes mellitus**. *Cochrane Database Syst. Rev.* (2018) **73** 26-33. DOI: 10.1002/14651858.cd010318.pub3
12. Marathe P. H., Gao H. X., Close K. L.. **American diabetes association standards of medical care in diabetes 2017**. *J. Diabetes* (2017) **9** 320-324. DOI: 10.1111/1753-0407.12524
13. Mario E. J., Richards L. KS, Mcleod H., Stanley D., Yap Y. A., Knight J.. **Gut microbial metabolites limit the frequency of autoimmune T cells and protect against type 1 diabetes**. *Nat. Immunol.* (2017) **18** 552-562. DOI: 10.1038/ni.3713
14. Ogurtsova K. J., Fernandes R., Huang Y., Linnenkamp U., Guariguata L., Cho N. H.. **IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for 2015 and 2040**. *Diabetes Res. Clin. Pract.* (2017) **128** 40-50. DOI: 10.1016/j.diabres.2017.03.024
15. Riddell M. C., Gallen I. W., Smart C. E., Taplin C. E., Adolfsson P., Lumb A. N.. **Exercise management in type 1 diabetes: A consensus statement**. *Lancet Diabetes & Endocrinol.* (2017) **5** 377-390. DOI: 10.1016/S2213-8587(17)30014-1
16. Rowan J. A., Rush E. C., Plank L. D., Lu J., Obolonkin V., Coat S.. **Metformin in gestational diabetes: The offspring follow-up (MiG TOFU) body composition and metabolic outcomes at 7–9 Years of age**. *Obstetrical Gynecol. Surv.* (2018) **73** 565-567. DOI: 10.1097/01.ogx.0000547170.99000.49
17. Sorli C. S., Harashima I. G., Tsoukas M., Unger J., Karsbol J. D., Hansen T.. **Efficacy and safety of once-weekly semaglutide monotherapy versus placebo in patients with type 2 diabetes (SUSTAIN 1): A double-blind, randomised, placebo-controlled, parallel-group, multinational, multicentre phase 3a trial**. *Lancet Diabetes & Endocrinol.* (2017) **5** 251-260. DOI: 10.1016/S2213-8587(17)30013-X
18. Tuttolomondo A.. **Relationship between diabetes and ischemic stroke: Analysis of diabetes-related risk factors for stroke and of specific patterns of stroke associated with diabetes mellitus**. *J. Diabetes & Metabolism* (2018) **06** 73-84. DOI: 10.4172/2155-6156.1000544
19. Wang Q., Miao Z., Torres G., Wu S., Ouyang C., Xie Z.. **Metformin suppresses diabetes-accelerated atherosclerosis via the inhibition of drp1-mediated mitochondrial fission**. *Diabetes* (2017) **66** 193-205. DOI: 10.2337/db16-0915
20. Willeit P, Skroblin A., Moschen A. R., Yin X., Kaudewitz D., Zampetaki A.. **Circulating MicroRNA-122 is associated with the risk of new-onset metabolic syndrome and type 2 diabetes**. *Diabetes* (2017) **66** 347-357. DOI: 10.2337/db16-0731
|
---
title: 'Involving children and adolescents with type 1 diabetes in health care: a
qualitative study of the use of patient-reported outcomes'
authors:
- Rikke Bjerre Lassen
- Caroline Bruun Abild
- Kurt Kristensen
- Lene Juel Kristensen
- Jens Thusgård Hørlück
- Annesofie Lunde Jensen
journal: Journal of Patient-Reported Outcomes
year: 2023
pmcid: PMC9981819
doi: 10.1186/s41687-023-00564-0
license: CC BY 4.0
---
# Involving children and adolescents with type 1 diabetes in health care: a qualitative study of the use of patient-reported outcomes
## Abstract
### Background
Within pediatric health care services, Patient-reported Outcomes (PROs) regarding the patient’s health status are mainly used for research purposes in a chronic care setting. However, PROs are also applied in clinical settings in the routine care of children and adolescents with chronic health conditions. PROs have the potential to involve patients because they ‘place the patient at the center’ of his or her treatment. The investigation of how PROs are used in the treatment of children and adolescents and how this use can influence the involvement of these patients is still limited. The aim of this study was to investigate how children and adolescents with type 1 diabetes (T1D) experience the use of PROs in their treatment with a focus on the experience of involvement.
### Results
Employing Interpretive Description, 20 semi-structured interviews were conducted with children and adolescents with T1D. The analysis revealed four themes related to the use of PROs: Making room for conversation, Applying PROs under the right circumstances, Questionnaire structure and content, and Becoming partners in health care.
### Conclusions
The results clarify that, to some extent, PROs fulfill the potential they promise, including patient-centered communication, detection of unrecognized problems, a strengthened patient-clinician (and parent-clinician) partnership, and increased patient self-reflection. However, adjustments and improvements are needed if the potential of PROs is to be fully achieved in the treatment of children and adolescents.
## Background
The development and implementation of Patient-reported Outcomes (PROs) within pediatric health services is rapidly increasing in today’s health care system, particularly in the treatment of children and adolescents with chronic conditions, such as type 1 diabetes (T1D) [1]. A systematic review of the literature evaluating the impact of the use of PROs as an intervention in routine clinical care for such patients has found that integration of PROs increase health-related quality of life, clinical outcomes, and quality of care, underlining the legitimacy of applying PROs in pediatric care settings [2]. PROs are directly reported by the patient through questionnaires without interpretation of the patient's response [3]. PROs is an 'umbrella term', involving both reports from patients about their health status associated with the health care they recieve, such as health-related quality of life, and reports from patients about their experiences with health services, e.g., treatment satisfaction [4].
PROs enable early detection of symptoms and prevention of disease complications [5]. PROs may qualify patient-clinician communication on topics such as health-related quality of life by involving the patient’s point of view in care [6]. PROs may also help the patient gain a better understanding of his or her disease, which can promote self-management [7].
An extensive burden of self-management tasks is placed on children and adolescents with T1D (and their parents), including counting carbohydrates, planning physical activity, checking blood glucose levels, and administering insulin with the goal of maintaining glycemic control [8–10]. Furthermore, adolescence represents a time of transition characterized by growing independence and decreasing parental responsibility [11]. Consequently, children and adolescents with T1D have considerable knowledge about their own health [12–14]. Therefore, their perspectives on disease, treatment, and well-being are recognized as a means to improve treatment (e.g., patient-clinician communication, patient satisfaction, and patient compliance) [13–15]. The perspectives of children and adolescents can be included by using PROs. The application of PROs represents an increasing interest in the concept of patient involvement, defined as the processes of care, in which the individual patient's preferences, resources, and life situation are taken into consideration [16, 17]. Enhancing patient involvement can improve both services and health outcomes, underlining the importance of a patient-centered approach to care [18, 19].
Research has primarily focused on the development and validation of PROs and on the use of PROs to assess the impact of treatment on e.g., symptom burden, physical functioning and health-related quality of life among children and adolescents with T1D [5, 20–22]. Conversely, little attention has been given to the perspectives of children and adolescents on the use of PROs in the clinical care setting. A qualitative study has shown that pediatric patients consider PROs to promote clinicians’ focus on the illness, narrow access to care, and cause uncertainty about what is safe to reveal to clinicians [23]. Patients have noted concern about how their individual experiences are represented within the fixed structures of PROs [23]. Moreover, studies have found that pediatric patients raise concerns about the degree of literacy needed to understand and answer abstract or vague questions [24, 25]. Finally, knowledge is limited on how these patients experience being involved, especially when they enter adolescence and gradually become more responsible for their care [26, 27]. One study of how adolescents with T1D experience the partnership with clinicians in the transition from pediatric to adult services found that the adolescents did not have the perception of being part of a dialogue, and that the transition process was not based on maturity and individual needs and wishes, even though this was the intention [17].
These studies emphasize the importance of focusing on how PROs are used rather than on the PROs themselves. Reviewing the studies also points out a gap in the existing literature, investigation of the patient-involving opportunities of PROs in the treatment of pediatric patients. Therefore, the aim of the present study was to investigate how children and adolescents with T1D experience the use of PROs in their treatment with a focus on the experience of involvement. This aim was pursued in order to expand our knowledge on how and when PROs are experienced as useful to children and adolescents with T1D and to optimize the use of PROs in their treatment.
## Study design
To investigate how children and adolescents with T1D experience the use of PROs, this study was guided by the qualitative methodology Interpretive Description (ID) and comprised semi-structured interviews. The purpose of using ID was to answer the research question and provide clinical practice with a research-based choice of action [28].
## Setting
The study was conducted in the pediatric outpatient clinic at Steno Diabetes Center Aarhus, Aarhus University Hospital, in Central Denmark Region. Children and adolescents with T1D attend the outpatient clinic four times a year. Three appointments are ‘basic’ consultations at which well-being and glucose levels are discussed, among other things. The fourth appointment is an extended consultation, which consists of a number of examinations, including blood pressure, blood, and urine tests. Consultations involve communication with clinicians, in this case a doctor, dietician, and/or nurse, about diabetes management.
An initiative was started at Steno Diabetes Center Aarhus in July 2018 to screen for disturbed eating behavior (e.g., skipping meals and feelings of shame associated with eating) among children and adolescents with T1D in order to prevent eating disorders and disease-related complications. As part of the initiative, all children and adolescents aged 11 to 18 years were asked to complete a questionnaire seven days prior to their extended consultation either at home, or if not completed at home, in the waiting room at the outpatient clinic. PROs from the questionnaire were intended as a tool to support patient-centered communication and identify symptoms and problems. The PROs applied were the Diabetes Eating Problem Survey-Revised (DEPS-R) (containing 16 questions designed to assess disturbed eating behavior specific to T1D, including insulin restriction to lose weight) [20], the generic WHO-5 Well-Being Index (consisting of five questions designed to capture emotional well-being) [21], five questions on involvement (concerning patient-clinician communication), one question about the patient’s overall satisfaction with the latest consultation in the clinic, and an additional comments area. These PROs were gathered into one questionnaire in the web-based PRO system Ambuflex [29]. Responses were recorded and scored on a scale with six categories of answers (Appendix I), and PROs were presented to clinicians using the usual electronic medical record system. Higher DEPS-R scores indicated greater disturbed eating. Based on empiricism, a pre-determined cut-off score for disturbed eating was ≥ 20, which indicated patients with a level of disturbed eating warranting supplementary treatment (Fig. 1) [30]. Clinicians were trained to interpret questionnaire answers and to discuss these answers with the patient. Fig. 1Annual process of standard treatment and the initiative in the outpatient clinic On the basis of PROs, particularly the DEPS-R score, and a clinical assessment, patients could be offered supplementary multidisciplinary treatment by a psychologist, psychiatrist, nurse, pediatrician, and/or a dietician. Figure 1 outlines the process of standard treatment and the initiative in the clinic.
## Sampling
Children and adolescents with T1D were invited to participate in individual interviews. They were sampled purposively; participation criteria were patients with T1D (or similar diabetes type) aged 11 to 18 years who completed the questionnaire at least two times [28]. 11 was chosen as a lower age boundary due to the fact that children in this age have the appropriate cognitive skills to read and understand questions and to select answers matching their perspectives [13]. Furthermore, both the DEPS-R and the WHO-5 has been validated in T1D patients in this age group [20, 21]. With the purpose of obtaining nuanced data, patient age and gender were kept in mind during recruitment.
Participants were recruited in collaboration with clinicians in the clinic. Before the annual extended consultation, an information letter about the study was sent to patients meeting the participation criteria. Prior to the consultation, the clinicians were informed that patients had received the information letter. The clinicians were asked to invite patients to participate. Patients were invited if they had experiences with PROs and had an interest in spending time with a researcher to explain those experiences [28]. Patients also had the opportunity to contact the first author if they wanted to participate.
## Data collection
Data were collected from December 2020 to October 2021 in 20 semi-structured interviews. The first author (RBL) conducted the interviews, except for five interviews that were conducted by a co-author (CBA). These authors worked as qualitative researchers and were previously affiliated with the clinic (as a student assistant and a dietician). The intention of the interviews was to explore the participants’ experiences and reflections related to answering the questionnaire and the use of PROs in the consultation, including the patient-involving potential of PROs. A semi-structured interview guide was applied and adjusted concurrently with data collection and analysis in accordance with ID [28]. The interview guide had four themes: Practical aspects of completing the questionnaire (e.g., ‘Tell me about the last time you completed the questionnaire’), Content of the questionnaire (e.g., ‘Are any of the questions difficult to answer?’), Experiences assessing one’s own condition and needs (e.g., ‘How has it been to assess your attitude towards eating and diabetes treatment?’), and Use of PROs in the consultation (e.g., ‘Tell me how the questionnaire was used in the conversation with your doctor/nurse’). Select questions from the questionnaire were read to participants to support conversation during the interview. The interviews took place in the clinic or were carried out using telephone or video communication. Interviews were audio-recorded and had a duration of 11 to 48 min (mean 31 min).
## Data analysis
Interviews were transcribed verbatim by the first author, a secretary, or a student assistant. In line with ID, data analysis was performed by the first author inductively and concurrently with data collection and comprised three analytical phases: initial reading of the interview transcripts, de-contextualization of the data, and re-contextualization of the data. The analysis also contained ongoing memo writing, noting immediate meanings and relationships in the data. Interview transcripts were organized and coded in the qualitative data analysis software NVivo. First, the transcripts were read through to get an impression of the data. Second, open coding of data was carried out to assign headings/codes to text segments to identify themes and patterns. As the analysis proceeded, some codes were modified or eliminated either because they revolved around the same topic or because they were not relevant to the study objective [28]. Finally, the dimensions of participants’ experiences with PROs were gathered into four themes. In the presentation of findings, ‘ID’ followed by a number refers to a specific participant.
## Participants
A total of 20 children and adolescents aged 11 to 18 years (mean 14.85 years) with T1D participated in the study. Further characteristics are described in Table 1.Table 1Sample characteristicsCharacteristicTotal sample ($$n = 20$$)Age, years14.85 [11–18]Gender Male Female9 [45]11 [55]Diabetes type T1D Secondary diabetes18 [90]2 [10]Completed questionnaires2.6 [2–5]Values are given as mean [range] or n (%) Findings from the analysis were categorized into four themes: Making room for conversation, Applying PROs under the right circumstances, Questionnaire structure and content, and Becoming partners in health care.
## Theme 1: making room for conversation
Overall, the use of PROs was described by the participants as promoting and improving conversation in their consultations with clinicians. However, participants identified a valuable topic of conversation, namely social health, that was not part of the PROs.
First, the DEPS-R questionnaire had a new focus on meals, body, and weight. Some participants found it rather difficultto answer questions about these, from their point of view, personal topics, thereby crossing personal boundaries:“Sometimes it’s difficult to answer because I know that I have to answer honestly, but there are some things that I’m ashamed of…” (ID3) This participant linked her own eating behavior with shame and struggled to answer questions about this topic area. PROs from DEPS-R facilitated topics of conversation other than standard treatment (e.g., blood glucose levels and insulin treatment), which promoted conversation from the participants' point of view. Talking about eating behavior was experienced as a delicate and vulnerable situation because the participants’ thoughts and feelings around the topic were exposed. Furthermore, the participants tended to concern about parental disapproval if such topics were identified. Even though the change in focus seemed intense to participants, they appreciated it. Diabetes management and meals were closely connected to each other by the participants and by introducing DEPS-R in the consultation the participants’ everyday lives became a bigger part of their care. Thus, PROs worked as a tool to identify and discuss otherwise neglected problems. Participants viewed this as an improvement in the patient-clinician dialogue.
The conversation focused on the individual by expressing issues or concerns that the patient reported, which had a positive impact on the participants’ views of PROs. One participant used the additional comments area to supplement the questionnaire with a topic of importance to her, namely the transition from the pediatric clinic to the adult clinic, which expanded conversation in the consultation:“It was probably because I was turning 18. In that situation, the questionnaire helped me go into some of the things that you don’t usually talk about.” ( ID5) This quotation exemplifies the perception of PROs as an instrument to guide the conversation towards the participants’ specific situation, own values, and strategies in order to make patient-clinician communication patient-centered. Moreover, the quotation underlines the significance of using PROs in an < 18 population without parents being involved.
Conversely, no focus on the social dimensions of disease-related issues was present in the questionnaire, even though such a focus was cherished among participants. Most participants expressed an interest in incorporating questions about how their social environment (e.g., family, friends, school, or hobbies) interacts with their diabetes, and acknowledged the relevans of a dialogue about social health with clinicians. The content of such questions was exemplified by the following quote:“It could be whether your diabetes sometimes prevent you from getting together with someone after school, or if your diabetes sometimes stops you from taking part in games.” ( ID15) The introduction of such questions was viewed by participants as a possibility for improving patient-clinician communication, allowing for dialogue about the complex interplay between disease and social environment.
## Theme 2: applying PROs under the right circumstances
Specific conditions under which the questionnaire was introduced and used proved pivotal to participants’ appreciation of the application of PROs in their treatment.
Most of the participants explicitly identified and decribed PROs as a tool supporting the clinicians’ work. According to the participants, PROs helped clinicians prepare for the consultation. Generally, participants experienced that PROs were discussed in the consultation as a result of the clinicians' preparation:“It’s nice that they [the clinicians] say, ‘I can tell from your response that things aren’t well,’ and then you can talk about it.” ( ID7) When participants got feedback on their questionnaire response, it became clear to them how PROs were used in their treatment.
Participants reflected on the clinicians' communication about the questionnaire, and several of them expressed a lack of information about the aim of questions on involvement and the DEPS-R. Some participants felt that they appraised and criticized the clinicians' work by answering the questions on involvement:"I hope that, by answering that I don't feel involved, my doctors won't be mad or sad." ( ID4) The quotation illustrates not only the discomfort that some participants experienced when answering the questions on involvement, but also a misconception about the purpose of the questions. Other participants were not aware that the DEPS-R is used as a screening tool for disturbed eating. This lack of awareness may be due to missing information about the entire questionnaire, and it underlines the significance of clearly communicating the aim of the applied PROs in the clinic.
Participants who did not understand the purpose of answering questions concerning meals, body, and weight (DEPS-R) found these questions irrelevant. They managed their diabetes according to what they ate and how they exercised. Therefore, items about overeating, weight control, and vomiting were questioned by these participants. A participant with secondary diabetes explained how these questions lacked relevance to her:“I also have diabetes, but I think that some of the questions were aimed at those who only have diabetes and might have struggled a little more with being overweight… The questions were a little bit like ‘what do you want the most: to manage your disease or be skinny?’ And I have always struggled with gaining weight and been underweight my entire life. But I manage my disease, so I don’t quite understand the questions.” ( ID6) Conversely, participants who expressed knowledge of the aim of the DEPS-R (regardless of problems with disturbed eating) viewed the questions as relevant and important. According to these participants, such questions were able to call forth the proper care for the patients in need of it.
The relevance of the questions on involvement was also viewed differently among participants. Some saw these questions as unnecessary by virtue of already feeling involved in the treatment. Others appreciated being asked about their involvement as they felt the most qualified to answer questions about this topic.
## Theme 3: questionnaire structure and content
Participants reflected on issues concerning both the structure and content of the questionnaire.
The temporal frame in which questions were being asked troubled participants, such as when they were asked about patient-clinician communication in their last consultation approximately 3 months earlier. Health status changes between questionnaire completion a few days before the consultation and the in-clinic consultation made other participants view the PROs as invalid.
Participants problematized the standardized response categories. The most common concern was how the complexity of an individual life was represented in the secure categories:“There weren’t any categories that made sense to me, so I just crossed out the lowest. I didn’t belong to any of the levels.” ( ID2) This quotation clarifies the importance of supporting PROs with dialogue to ensure that patients are not reduced to simple categories and numbers.
Specific formulations were viewed as difficult to understand. Distinguishing the five questions on involvement was challenging for most participants, and they suggested combining some of these questions without compromising the assessment of their involvement in care. Some participants found it difficult to understand what ‘involvement’ was, as the concept seemed too abstract to them. The wording ‘I try to eat to the point of spilling ketones in my urine’ was described as unclear, as ‘doctor language’, or as ‘adult language’. This led to potentially invalid answers. No age differences were found in difficulties understanding the questions.
## Theme 4: becoming partners in health care
The relationship between participants and clinicians was identified as significant to participants, and the PROs played an important part in supporting this relationship. For some participants, parents were involved in the treatment, and the completion of the questionnaire and the subsequent use of PROs in the consultation influenced the relationship between parents, patients, and clinicians. Ultimately, this led to increased patient (and parental) involvement. Furthermore, the application of PROs showed potential to promote a partnership between participants and clinicians through critical self-reflection.
Some participants did not involve their parents while filling in the questionnaire. They viewed it as an opportunity to share ‘secret’ information with clinicians, which resulted in a feeling of having a special bond with clinicians. One participant described the additional comments area as follows:“It gives me the chance to say some things that I might not want to say in front of my parents… Because I wanted to keep some things between me and my nurse.” ( ID3) By letting the individual set the agenda through PROs, the consultation was personalized, and the participant and the clinician(s) became equal partners.
Participants who had difficulty understanding questions explained that their parents helped them complete the questionnaire. These parents took part in a PRO-based dialogue with their children and clinicians, indicating a partnership between parents, patient, and clinicians. Thus, the use of PROs led to a new way of involving both participants and parents in care, ultimately increasing patient and parent involvement.
Based on PROs, supplementary treatment, such as counseling sessions with a psychologist, was offered to specific participants. They took part in deciding on and planning the treatment in collaboration with clinicians. Consequently, PROs enhanced the participants’ experience of having ownership of their own treatment, which strengthened their autonomy and supported their partnership with clinicians.
A group of the participants found that completing the questionnaire helped them gain a better understanding of their disease. By reflecting on the questions presented in the questionnaire, the participants obtained new perspectives on their diabetes management:“You get a little… Do people really do such things? Trying to eat in a way so that they can spill ketones. I’m thinking ‘shit, is it possible to get that far out?’” ( ID11) Another participant pointed out how specific questions had been eye-opening to her:“Well, I have definitely faced some facts by answering them [the questions]. But it isn’t necessarily negative. For instance, this one, ‘When I overeat, I don’t take enough insulin to cover the food.’ When I answer that, I’m thinking ‘that’s actually pretty stupid. Why do you do that?’” ( ID7) Through self-reflection, participants were able to take a stand regarding their disease and care. By expressing such reflections to clinicians, children and adolescents with T1D may become more active partners in the patient-clinician collaboration. In addition, these reflections can increase the patients’ self-management. Yet, participants expressing an educational aspect of PROs were also the ones experiencing problems with disturbed eating. Other participants did not reflect on the particular questions; they filled in the questionnaire and put it aside. Moreover, the oldest participants were the ones reflecting on their self-management, which makes sense on account of the transitional stage of their treatment.
## Discussion
Overall, this qualitative study found that the participants experienced PROs as promoting and improving the conversation in the consultation by including new topics (exceeding standard treatment) and focusing on individual issues, thereby creating patient-centered communication and increased patient and parental involvement.
This study showed that PROs from the questionnaire did not reflect the social dimensions of the patient’s disease. This deficiency may be due to challenges regarding the implementation of new initiatives in established practices in the health care system. Questions in DEPS-R focused on physical health (e.g., weight and blood glucose levels), leaving consultations to be dominated by a biomedical agenda. However, study participants emphasized the importance of assessing the interaction between disease and social environment. These findings are consistent with studies showing that clinicians prioritize objective PROs (patient-reported physical health), whereas pediatric patients prioritize subjective PROs (patient-reported mental and social health) [31–34]. Furthermore, previous studies have found that children and adolescents with chronic conditions want to be met as individuals and acknowledged holistically [15, 17], stressing the significance of PROs in being able to evaluate other aspects of health than just the physical aspects.
In addition to the physical health-focused aspects of the questions in DEPS-R, the participants in this study characterized them as personal. Even though participants appreciated these new topics, they also experienced them as difficult to answer due to their personal nature, which may be due to the taboo around both male and female body ideals [35]. Clinicians should be aware of this when using PROs from the DEPS-R in the consultation. Participants valuing the change in focus are consistent with the aforementioned priority of subjective PROs. Moreover, the result is compatible with existing knowledge on PROs in the treatment of adult patients, including the detection and legitimization of unrecognized problems [7, 36, 37].
The approach of using PROs to identify patients’ individual needs and delivering patient-centered care can be questioned, as individual experiences are quantified through standardized questionnaires. On the one hand, participants in this study found the response categories to lack nuance and to be unable to measure the complexity of their lives. Similar perspectives were found in a study of pediatric patients’ views on PROs, leading the patients to suggest an unstructured, flexible questionnaire format. They suggested using pictures and symbols instead of numbers and words [23]. On the other hand, this study’s participants felt involved in their treatment in different ways as a result of applying PROs. Through the use of PROs, participants actively participated in a partnership with clinicians, and they became aware of their own self-management, which is comparable with studies of PROs in adult care [6, 7, 36, 38–42]. Notably, patients in the above-mentioned study of pediatric patients’ views on PROs considered PROs to be a method that promoted the clinicians’ focus on illness, thereby narrowing the focus on the patient-clinician relationship [23]. This contrasts with the strengthened relationship between participants and clinicians found in this study, clarifying that PROs should not be overarching in the conversation but only create a basis for dialogue [40]. Yet, the PROs used in the aforementioned study involved only health status reports in terms of symptoms and health-related quality of life, which may explain the differences in youth perspectives on PROs [23].
In some cases, lacking information about the aim of the questionnaire made study participants view the questionnaire as uncomfortable. The misconception of the questions on involvement being an evaluation of the clinicians led to a fear of criticizing them, even though the real purpose of these questions was to ensure patient-centered care. Thus, pediatric patients need a clear explanation about the purpose of using PROs in their treatment in order to avoid uncertainty about what is safe to reveal to clinicians.
The implementation of PROs in the treatment of pediatric patients should be considered carefully, as the use of PROs seems to entail some challenges for these patients. Even though this study found no association between low age and troubles completing the questionnaire, it seems reasonable to take patient age into account when deciding on applying PROs in pediatric health care. Yet, studies have shown that age should not be the only indicator of a pediatric patient’s ability to independently self-report. These studies have elucidated that literacy (reading and writing) skills determine the ability of children and adolescents to complete a questionnaire in a valid manner [24, 25, 43, 44]. Reading assistance from parents (as in this study) or clinicians based on pediatric patient choice should be a possibility to ensure data quality [25]. Conversely, some of the younger participants did not have any difficulties understanding the term ‘involvement’ and more technical terms. It is plausible that exposure to a treatment-intensive and chronic disease, such as T1D, provided an introduction to medical terms and health-related vocabulary that normally would be absent through childhood [25, 44].
This study has some limitations. The sample size was adequate for a qualitative study, and age and gender varied among participants, making the results relatively representative of children and adolescents with T1D. Yet, the sample size seemed too small for conclusive findings about subgroup differences, e.g., differences between patients with and without disturbed eating or differences between adolescents and younger children. In addition, data were collected by two authors, but only analyzed by one of these authors. To ensure a comprehensive understanding of the study subject, investigator triangulation was used as a strategy to test validity [45].
This study only focused on patients’ experiences with PROs and involvement. Other perspectives should be explored to investigate additional actions in relation to PROs. Thus, further research should involve other qualitative methods, samples of pediatric patients with other chronic conditions, and clinicians’ and parents' perceptions of PROs in pediatric health care.
## Conclusion
Applying PROs in the treatment of children and adolescents with T1D leads to an experience of focused consultation with new topics of conversation and a patient-centered approach, which were highly valued by participants in the present study. For participants to acknowledge the benefits of the PRO application, the aim of the entire questionnaire needs to be clear, and for PROs to be valid, questions have to be understandable. In some instances, the application of PROs can strengthen the patient-clinician and parent-clinician partnerships, which increases patient (and parental) involvement. However, the potential benefits of using PROs in this patient group will not necessarily be achieved unless the questionnaire structure and age and literacy skills of the patients are thoroughly considered. In addition, asking about the interaction between disease and social environment will allow for a holistic assessment of the patient’s health.
## Appendix I: Content and structure of the web-based AmbuFlex questionnaire
Total number of items27 WHO-5 Well-Being Index Number of items5ContentPlease indicate for each of the 5 statements which is closest to how you have been feeling over the past 2 weeks. Over the past 2 weeks…… I have felt cheerful and in good spirits …I have felt calm and relaxed …I have felt active and vigorous… I woke up feeling fresh and rested… my daily life has been filled with things that interest meResponse categories: All of the time Most of the time More than half the time Less than half the time Some of the time At no time Diabetes Eating Problem Survey-Revised (DEPS-R) Number of items16ContentLiving with diabetes can sometimes be difficult, particularly regarding eating and diabetes management. Listed below are a variety of attitudes and behaviors regarding diabetes management. For each statement, tick the one answer that indicates how often this was true for you during the past monthHow often this was true for you during the past month… Losing weight is an important goal for me I skip meals and/or snacks Other people have told me that my eating is out of control When I overeat, I don’t take enough insulin to cover the food I eat more when I am alone than when I am with others I feel that it’s difficult to lose weight and control my diabetes at the same time I avoid checking my blood sugar when I feel like it is out of range I make myself vomit I try to keep me blood sugar high so that I will lose weight I try to eat to the point of spilling ketones in my urine I feel fat when I take all of my insulin Other people tell me to take better care of my diabetes After I overeat, I skip my next insulin dose I feel that my eating is out of control I alternate between eating very little and eating huge amounts I would rather be thin than to have good control of my diabetesResponse categories: Never Rarely Sometimes Often Usually Always Questions on involvement and overall satisfaction Number of items6ContentThese questions are about the degree to which you have felt involved in your treatment (select only one answer for each statement)Patient’s experience of involvement: The health staff asked questions about my own experiences with my illness/condition I talked to the health staff about the questions or concerns I had The health staff encouraged me to ask questions or talk about my concerns I was involved when decisions were made about what was to take place I have had an appropriate number of talks with the health staff about how I can best handle my illness/condition All in all, I am satisfied with my last visit to the outpatient clinicResponse categories: Not applicable *To a* slight degree To some degree *To a* great degree *To a* very high degree Don't know Additional comments area ContentWrite here if you have any comments on the questions or if you want to discuss something at the consultation
## References
1. Flannery H, Jacob J. **Measuring psychological outcomes in paediatric settings: making outcomes meaningful using client-defined perspectives**. *Clin Child Psychol Psychiatry* (2020.0) **25** 594-603. DOI: 10.1177/1359104520904120
2. Bele S, Chugh A, Mohamed B, Teela L, Haverman L, Santana MJ. **Patient-reported outcome measures in routine pediatric clinical care: a systematic review**. *Front Pediatr* (2020.0) **8** 364. DOI: 10.3389/fped.2020.00364
3. 3.FDA (2006) Guidance for industry: patient-reported outcome measures: use in medical product development to support labeling claims: draft guidance. Health Qual Life Outcomes. Health Qual Life Outcomes 4:1-20
4. Refolo P, Minacori R, Mele V, Sacchini D, Spagnolo AG. **Patient-reported outcomes (PROs): the significance of using humanistic measures in clinical trial and clinical practice**. *Eur Rev Med Pharmacol Sci* (2012.0) **16** 1319-1323. PMID: 23104647
5. Prahalad P, Tanenbaum M, Hood K, Maahs DM. **Diabetes technology: improving care, improving patient-reported outcomes and preventing complications in young people with Type 1 diabetes**. *Diabet Med* (2018.0) **35** 419-429. DOI: 10.1111/dme.13588
6. Howell D, Molloy S, Wilkinson K, Green E, Orchard K, Wang K. **Patient-reported outcomes in routine cancer clinical practice: a scoping review of use, impact on health outcomes, and implementation factors**. *Ann Oncol* (2015.0) **26** 1846-1858. DOI: 10.1093/annonc/mdv181
7. Mejdahl CT, Nielsen BK, Hjøllund NH, Lomborg K. **Use of patient-reported outcomes in outpatient settings as a means of patient involvement and self-management support - a qualitative study of the patient perspective**. *EJPCH* (2016.0) **4** 359-367. DOI: 10.5750/ejpch.v4i2.1125
8. Wennick A, Lundqvist A, Hallström I. **Everyday experience of families three years after diagnosis of type 1 diabetes in children: a research paper**. *J Pediatr Nurs* (2009.0) **24** 222-230. DOI: 10.1016/j.pedn.2008.02.028
9. 9.Streisand R, Monaghan M (2014) Young children with type 1 diabetes: challenges, research, and future directions. Curr Diabetes Rep 14.
10. Freeborn D, Dyches T, Roper SO, Mandleco B. **Identifying challenges of living with type 1 diabetes: child and youth perspectives**. *J Clin Nurs* (2013.0) **22** 1890-1898. DOI: 10.1111/jocn.12046
11. Overgaard M, Lundby CL, Grabowski D. **Disruption, worries and autonomy in the everyday lives of adolescents with type 1 diabetes and their family members: A qualitative study of intrafamilial challenges**. *J Clin Nurs* (2020.0) **29** 4633-4644. DOI: 10.1111/jocn.15500
12. Christie D. **How do children and adolescents understand their diabetes?**. *Pract Diabetes* (2019.0) **36** 117-120. DOI: 10.1002/pdi.2228
13. McCabe MA. **Involving children and adolescents in medical decision making: developmental and clinical considerations**. *J Pediatr Psychol* (1996.0) **21** 505-516. DOI: 10.1093/jpepsy/21.4.505
14. Deshpande PR, Rajan S, Sudeepthi BL, Nazir ACP. **Patient-reported outcomes: a new era in clinical research**. *Perspect Clin Res* (2011.0) **2** 137-144. DOI: 10.4103/2229-3485.86879
15. Karisalmi N, Hanna Stenhammar H, Kaipio J. **What constitutes the patient experience of children? Findings from the photo elicitation and the video diary study**. *Patient Exp J* (2018.0) **5** 54-68. DOI: 10.35680/2372-0247.1292
16. Lavallee DC, Chenok KE, Love RM, Petersen C, Holve E, Segal CD. **Incorporating patient-reported outcomes into health care to engage patients and enhance care**. *Health Aff* (2016.0) **35** 575-582. DOI: 10.1377/hlthaff.2015.1362
17. Hansen KK, Jensen AL. **Partnership in transition: experiences of adolescents with Type 1 diabetes**. *Int Diabetes Nurs* (2017.0) **14** 1-8
18. van Dam HA, van der Horst F, van den Borne B, Ryckman R, Crebolder H. **Provider-patient interaction in diabetes care: effects on patient self-care and outcomes. A systematic review**. *Patient Educ Couns* (2003.0) **51** 17-28. DOI: 10.1016/S0738-3991(02)00122-2
19. Vahdat S, Hamzehgardeshi L, Hessam S, Hamzehgardeshi Z. **Patient involvement in health care decision making: a review**. *Iran Red Crescent Med J* (2014.0) **16** e12454. DOI: 10.5812/ircmj.12454
20. Wisting L, Frøisland DH, Skrivarhaug T, Dahl-Jørgensen K, Rø O. **Psychometric properties, norms, and factor structure of the diabetes eating problem survey-revised in a large sample of children and adolescents with type 1 diabetes**. *Diabetes Care* (2013.0) **36** 2198-2220. DOI: 10.2337/dc12-2282
21. de Wit M, Pouwer F, Gemke RJBJ, Waal HADVD, Snoek FJ. **Validation of the WHO-5 well-being index in adolescents with type 1 diabetes**. *Diabetes Care* (2007.0) **30** 2003-2006. DOI: 10.2337/dc07-0447
22. Murillo M, Bel J, Pérez J, Corripio R, Carreras G, Herrero X. **Impact of monitoring health-related quality of life in clinical practice in children with type 1 diabetes mellitus**. *Qual Life Res* (2017.0) **26** 3267-3277. DOI: 10.1007/s11136-017-1682-6
23. Wolpert M, Curtis-Tyler K, Edbrooke-Childs J. **A qualitative exploration of patient and clinician views on patient reported outcome measures in child mental health and diabetes services**. *Adm Policy Ment Health* (2016.0) **43** 309-315. DOI: 10.1007/s10488-014-0586-9
24. Myers B, Johnson K, Lucas W, Govender R, Manderscheid R, Williams PP. **South African service users' perceptions of patient-reported outcome and experience measures for adolescent substance use treatment: a qualitative study**. *Drug Alcohol Rev* (2019.0) **38** 823-830. DOI: 10.1111/dar.12996
25. Withycombe JS, McFatrich M, Pinheiro L, Hinds PS, Keller FG, Baker JN. **The association of age, literacy, and race on completing patient-reported outcome measures in pediatric oncology**. *Qual Life Res* (2019.0) **28** 1793-1801. DOI: 10.1007/s11136-019-02109-9
26. Dovey-Pearce G, Hurrell R, May C, Walker C, Doherty Y. **Young adults' (16–25 years) suggestions for providing developmentally appropriate diabetes services: a qualitative study**. *Health Soc Care Commun* (2005.0) **13** 409-419. DOI: 10.1111/j.1365-2524.2005.00577.x
27. Valenzuela JM, Smith LB, Stafford JM, D'Agostino RBJ, Lawrence JM, Yi-Frazier JP. **Shared decision-making among caregivers and health care providers of youth with type 1 diabetes**. *J Clin Psychol Med Settings* (2014.0) **21** 234-243. DOI: 10.1007/s10880-014-9400-9
28. 28.Thorne S (2016) Interpretive description. Qualitative research for applied practice. 2 edn. New York: Routledge.
29. Hjollund NHI. **Fifteen years' use of patient-reported outcome measures at the group and patient levels: trend analysis**. *J Med Internet Res* (2019.0) **21** 1-14. DOI: 10.2196/15856
30. 30.Wisting L, Wonderlich J, Skrivarhaug T, Dahl-Jørgensen K, Rø Ø (2019) Psychometric properties and factor structure of the diabetes eating problem survey – revised (DEPS-R) among adult males and females with type 1 diabetes. J Eat Disord 7.
31. Anthony SJ, Selkirk E, Sung L, Klaassen RJ, Dix D, Klassen AF. **Quality of life of pediatric oncology patients: do patient-reported outcome instruments measure what matters to patients?**. *Qual Life Res* (2017.0) **26** 273-281. DOI: 10.1007/s11136-016-1393-4
32. Jones CM, Baker JN, Keesey RM, Eliason RJ, Lanctot JQ, Clegg JL. **Importance ratings on patient-reported outcome items for survivorship care: comparison between pediatric cancer survivors, parents, and clinicians**. *Qual Life Res* (2018.0) **27** 1877-1884. DOI: 10.1007/s11136-018-1854-z
33. Hinds PS, Gattuso JS, Fletcher A, Baker E, Coleman B, Jackson T. **Quality of life as conveyed by pediatric patients with cancer**. *Qual Life Res* (2004.0) **13** 761-772. DOI: 10.1023/B:QURE.0000021697.43165.87
34. Morrow AM, Hayen A, Quine S, Scheinberg A, Craig JC. **A comparison of doctors', parents' and children's reports of health states and health-related quality of life in children with chronic conditions**. *Child Care Health Dev* (2012.0) **38** 186-195. DOI: 10.1111/j.1365-2214.2011.01240.x
35. 35.Negative body image: causes, consequences and intervention ideas (2019) Government Equalities Office
36. Chen J, Ou L, Hollis SJ. **A systematic review of the impact of routine collection of patient reported outcome measures on patients, providers and health organisations in an oncologic setting**. *BMC Health Serv Res* (2013.0) **13** 1-24. DOI: 10.1186/1472-6963-13-211
37. Clark K, Bardwell WA, Arsenault T, DeTeresa R, Loscalzo M. **Implementing touch-screen technology to enhance recognition of distress**. *Psychooncology* (2009.0) **18** 822-830. DOI: 10.1002/pon.1509
38. Yang LY, Manhas DS, Howard AF, Olson RA. **Patient-reported outcome use in oncology: a systematic review of the impact on patient-clinician communication**. *Support Care Cancer* (2018.0) **26** 41-60. DOI: 10.1007/s00520-017-3865-7
39. Marshall S, Haywood K, Fitzpatrick R. **Impact of patient-reported outcome measures on routine practice: a structured review**. *J Eval Clin Pract* (2006.0) **12** 559-568. DOI: 10.1111/j.1365-2753.2006.00650.x
40. Greenhalgh J, Long AF, Flynn R. **The use of patient reported outcome measures in routine clinical practice: lack of impact or lack of theory?**. *Soc Sci Med* (2005.0) **60** 833-843. DOI: 10.1016/j.socscimed.2004.06.022
41. Taenzer P, Bultz BD, Carlson LE, Speca M, DeGagne T, Olson K. **Impact of computerized quality of life screening on physician behaviour and patient satisfaction in lung cancer outpatients**. *Psychooncology* (2000.0) **9** 203-213. DOI: 10.1002/1099-1611(200005/06)9:3<203::AID-PON453>3.0.CO;2-Y
42. Velikova G, Booth L, Smith AB, Brown PM, Lynch P, Brown JM. **Measuring quality of life in routine oncology practice improves communication and patient well-being: a randomized controlled trial**. *J Clin Oncol* (2004.0) **22** 714-724. DOI: 10.1200/JCO.2004.06.078
43. Conrad NJ, Harris N, Williams J. **Individual differences in children’s literacy development: The contribution of orthographic knowledge**. *Read Writ* (2013.0) **26** 1223-1239. DOI: 10.1007/s11145-012-9415-2
44. Mitchell AM, Brady SA. **The effect of vocabulary knowledge on novel word identification**. *Ann Dyslexia* (2013.0) **63** 201-216. DOI: 10.1007/s11881-013-0080-1
45. Carter N, Bryant-Lukosius D, DiCenso A, Blythe J, Neville AJ. **The use of triangulation in qualitative research**. *Oncol Nurs Forum* (2014.0) **41** 545-547. DOI: 10.1188/14.ONF.545-547
|
---
title: 'Comparison of outcomes of chronic kidney disease based on etiology: a prospective
cohort study from KNOW-CKD'
authors:
- Hyunjin Ryu
- Yeji Hong
- Eunjeong Kang
- Minjung Kang
- Jayoun Kim
- Hayne Cho Park
- Yun Kyu Oh
- Ho Jun Chin
- Sue K. Park
- Ji Yong Jung
- Young Youl Hyun
- Su Ah Sung
- Curie Ahn
- Kook-Hwan Oh
- Curie Ahn
- Curie Ahn
- Kook-Hwan Oh
- Hajeong Lee
- Seung Seok Han
- Hyunjin Ryu
- Eunjeong Kang
- Minjung Kang
- Youngok Ko
- Jeongok So
- Aram Lee
- Dong Wan Chae
- Yong Jin Yi
- Hyun Jin Cho
- Jung Eun Oh
- Kyu Hun Choi
- Seung Hyeok Han
- Tae-Hyun Yoo
- Mi Hyun Yu
- Kyu-Beck Lee
- Young Youl Hyun
- Hyun Jung Kim
- Yong-Soo Kim
- Sol Ji Kim
- Wookyung Chung
- Ji Yong Jung
- Kwon Eun Jin
- Su Ah Sung
- Sung Woo Lee
- Hyang Ki Min
- Soon Bin Kwon
- Soo Wan Kim
- Seong Kwon Ma
- Eun Hui Bae
- Chang Seong Kim
- Hong Sang Choi
- Minah Kim
- Tae Ryom Oh
- Sang Heon Suh
- Su Hyun Song
- Se Jeong Lee
- Yeong Hoon Kim
- Sun Woo Kang
- Hoseok Koo
- Tae Hee Kim
- Yun Mi Kim
- Young Eun Oh
- Eun Young Seong
- Sang Heon Song
- Miyeun Han
- Hyo Jin Kim
- Seunghee Ji
- Tae Ik Chang
- Ea Wha Kang
- Kyoung Sook Park
- Aei Kyung Choi
- Ja-Ryong Koo
- Jang-Won Seo
- Sun Ryoung Choi
- Seon Ha Baek
- Myung Sun Kim
- Yun Kyu Oh
- Jeong Mi Park
- Byung-Joo Park
- Sue K. Park
- Joongyub Lee
- Choonghyun Ahn
- Kyungsik Kim
- Jayoun Kim
- Dayeon Nam
- Soohee Kang
- Juhee Lee
- Heejung Ahn
- Dong Hee Seo
- Soyoung Kim
- Korea Biobank
- Ok Park
- Il Yoel Kim
- Sung Hyun Kang
- Kyoung Hwa Kim
journal: Scientific Reports
year: 2023
pmcid: PMC9981888
doi: 10.1038/s41598-023-29844-x
license: CC BY 4.0
---
# Comparison of outcomes of chronic kidney disease based on etiology: a prospective cohort study from KNOW-CKD
## Abstract
The causes of chronic kidney disease (CKD) affects its outcomes. However, the relative risks for adverse outcomes according to specific causes of CKD is not well established. In a prospective cohort study from KNOW-CKD, a cohort was analyzed using overlap propensity score weighting methods. Patients were grouped into four categories according to the cause of CKD: glomerulonephritis (GN), diabetic nephropathy (DN), hypertensive nephropathy (HTN), or polycystic kidney disease (PKD). From a total of 2070 patients, the hazard ratio of kidney failure, the composite of cardiovascular disease (CVD) and mortality, and the slope of the estimated glomerular filtration rate (eGFR) decline according to the cause of CKD were compared between causative groups in a pairwise manner. There were 565 cases of kidney failure and 259 cases of composite CVD and death over 6.0 years of follow-up. Patients with PKD had a significantly increased risk for kidney failure compared to those with GN [Hazard ratio (HR) 1.82], HTN (HR 2.23), and DN (HR 1.73). For the composite outcome of CVD and death, the DN group had increased risks compared to the GN (HR 2.07), and HTN (HR 1.73) groups but not to the PKD group. The adjusted annual eGFR change for the DN and PKD groups were − 3.07 and − 3.37 mL/min/1.73 m2 per year, respectively, and all of these values were significantly different than those of the GN and HTN groups (− 2.16 and − 1.42 mL/min/1.73 m2 per year, respectively). In summary, the risk of kidney disease progression was relatively higher in patients with PKD compared to other causes of CKD. However, the composite of CVD and death was relatively higher in patients with DN-related CKD than in those with GN- and HTN-related CKD.
## Introduction
Chronic kidney disease (CKD), which is rapidly increasing in prevalence and incidence, is a heterogeneous set of diseases caused by various risk factors and comorbid conditions1–5. Although CKD patients share similar pathophysiologies involved with the kidney disease progression, the course and speed of CKD progression and associated complications differ according to the underlying causes. Therefore in the KDIGO guidelines, the cause of CKD is considered one of the important predictors of the outcome, together with other variables such as the glomerular filtration rate category, the albuminuria category, and other comorbid conditions5. However, the relative risk for adverse outcomes according to the specific cause of CKD has not been well studied6. Direct comparisons of outcomes according to the specific cause of CKD are important to understand the natural progression of CKD and to characterize possible complications according to the cause of CKD. This is critical for CKD management during the predialysis period, both to slow progression and to improve long-term outcomes. In addition, this knowledge can help determine high-risk groups among the CKD population so that resources can be prioritized and therapies can be more targeted.
A few studies have investigated relative risks for adverse outcomes in the CKD population according to specific causes7–10. However, previous studies did not cover all of the major CKD etiologies or stages or address all major outcomes. Due to the limitations, the results cannot be extrapolated to the entire adult CKD population.
In this study, we analyzed the hazard ratio for kidney progression and the composite outcome of cardiovascular disease (CVD) and all-cause mortality according to the cause of CKD in a prospective cohort. To investigate the effects of the causes of CKD on the outcomes, overlap weighted methods were used to adjust for the possible confounding factors. Additionally, we analyzed the annual rates of estimated glomerular filtration (eGFR) decline according to the cause of CKD to determine kidney progression patterns according to CKD etiology.
## Methods
This was a longitudinal study of a prospective cohort of CKD patients in Korea, called KNOW-CKD (KoreaN cohort study for Outcome in patients With Chronic Kidney Disease). KNOW-CKD is a multicenter prospective cohort study that enrolled adult predialysis patients with CKD stages G1 to G511. Patients were classified into four groups according to the specific cause of CKD at enrollment: glomerulonephritis (GN), diabetic nephropathy (DN), hypertensive nephropathy (HTN), and Polycystic kidney disease (PKD). Each group classification was determined based on pathologic diagnosis if a biopsy result was available ($27.6\%$ of total patients: $66.5\%$ of GN group, $6.4\%$ of DN group, $7.3\%$ of HTN group and $1.4\%$ of PKD group). Otherwise, group classifications were based on clinical diagnoses. The biopsy-proven GN consisted as following – $40\%$ IgA nephropathy, $7\%$ focal segment glomerular sclerosis, $6\%$ membranous nephropathy, $5\%$ crescentic GN, $2.4\%$ minimal change disease, and $1.5\%$ lupus nephritis. Non-biopsy-proven GN was defined as the clinical history manifesting chronic GN and the presence of albuminuria or glomerular hematuria with or without an underlying systemic disease causing GN. The active GN population taking immunosuppressant at enrollment was excluded to minimize the heterogeneity by treatment. Diagnosis of DN was strictly based on albuminuria in a patient with type 2 diabetes and the presence of diabetic retinopathy. To exclude DN patients who may have combined GN, diabetic patients with glomerular hematuria were not included in the DN group. HTN was diagnosed by a history of hypertension and the absence of a systemic illness associated with kidney damage. Only the patients with proteinuria < 1.5 g/day and a proportion of urine albumin < $50\%$ of urine protein were included in HTN to exclude the GN population. To diagnose PKD, unified ultrasound criteria were used12. Other causative diseases was categorized as ‘unclassified’ and excluded from our analysis.
A total of 2238 patients enrolled in the study from April 2011 to February 2016. After excluding patients with unclassified etiology or without follow-up data, 2070 patients were finally analyzed in this study for survival analysis with follow up until March 31, 2020. To determine the annual eGFR change and trajectory, we included only those patients ($$n = 1952$$) with more than two creatinine measurements (Fig. 1). Written informed consent from each patient was collected voluntarily at the time of enrollment. The study was approved by the institutional review board of each participating hospital: Chonnam National University Hospital (CNUH-2011-092), Eulji General Hospital [201105-01], Gil Hospital (GIRBA2553), Kangbuk Samsung Medical Center [2011-01-076], Pusan Paik Hospital [11-091], Seoul National University Bundang Hospital (B-$\frac{1106}{129}$-008), Seoul National University Hospital (H-1704-025-842), Seoul St. Mary’s Hospital (KC11OIMI0441), and Yonsei University Severance Hospital [4-2011-0163]. This study follows the guidelines of the 2008 Declaration of Helsinki. Figure 1Flowchart of enrolled study patients. eGFR, estimated glomerular filtration rate; IDMS, isotope dilution mass spectrometry.
Demographic details and medication history were collected at enrollment. Serum creatinine was measured at each study visit by a central laboratory (Lab Genomics, Seoul, Republic of Korea) using an isotope dilution mass spectrometry-traceable method. For eGFR, the CKD -EPI equation based on serum creatinine was used13. After the baseline visit, patients were followed-up at 6 and 12 months and then every 1 year until death or drop-out and follow-up events were recorded. In case of loss to follow-up, patients were censored for kidney and CVD events at the last follow-up visit. Death and the cause of death were collected using either hospital medical records or data from the National Database of Statistics Korea using the Korean resident registration number. Data were collected until whichever came first: drop-out, death, or March 31, 2020.
Both kidney failure and the composite of kidney failure and/or creatinine doubling were used as kidney outcomes. Kidney failure was defined as starting maintenance dialysis (required for longer than 3 months) or receiving kidney transplantation. Another outcome was the composite outcome of CVD and all-cause death. CVD was defined as any first event of the following that needed hospitalization, intervention, or therapy during the follow-up period: acute myocardial infarction, unstable angina which needed admission due to aggravated coronary ischemic symptoms, percutaneous coronary artery intervention or coronary bypass graft surgery, ischemic or hemorrhagic cerebral stroke, cerebral artery aneurysm, congestive heart failure, symptomatic arrhythmia, aggravated valvular heart meant by requiring hospital admission, any pericardial disease that required hospital admissions such as pericarditis, pericardial effusion, or cardiac tamponade, abdominal aortic aneurysm, or severe peripheral arterial disease (Table S1).
The chi-square test or Anova was used to compare the baseline characteristics. Non-normally distributed variables such as parathyroid hormone, urine protein/creatinine, and high sensitivity C-reactive protein were compared by Kruskal–Wallis test. The four groups had significant differences in baseline characteristics including age and baseline eGFR; we therefore used the overlap propensity score (PS) weighting method to minimize the effects of confounding factors on outcomes14. Overlap weighting is a PS method that tries to mimic important attributes of randomized clinical trials. This method can overcome the potential limitation of adjusting the difference in measured characteristics using classic PS methods of inverse probability of treatment weighting (IPTW). Overlap weighting overcomes these limitations by assigning weights to each patient that are proportional to the probability of that patient belonging to the opposite group15. PSs were calculated using a logistic model with the following variables since they showed significant differences among the four groups: age, sex, body mass index, CKD stage, mean blood pressure, CVD, hemoglobin, serum uric acid, calcium, phosphorous, albumin, total cholesterol, high-density lipoprotein, low-density lipoprotein, fasting blood sugar, intact parathyroid hormone, urine protein-to-creatinine ratio, high-sensitivity C-reactive protein, diuretics use, statin use, and angiotensin converting enzyme inhibitor or angiotensin receptor blocker use in this study. The log10 transformed values were used for PS calculation with the non-normally distributed variables such as parathyroid hormone, urine protein-to-creatinine ratio, and high sensitivity C-reactive protein. The patients in the compared group were weighted by the probability of the reference group (1-PS), and the patients in the reference group were weighted by the probability of the compared group (PS). For two groups of CKD causes, we applied the overlap weighting method to each set, resulting in a total of 6 sets. To visually compare distributions of balance, the density plots were created (Figure S1). Additionally, the standardized mean difference (SMD) was calculated to check good balance after the overlap weighting method was applied. This is calculated by the absolute value of the difference in mean among groups divided by the standard deviation. The SMD less than or equal to 0.10 means good balance after weighting15. In outcome comparison analysis, a Cox proportional hazard model was used for kidney outcomes, and a cause-specific hazard model was used for the composite of CVD and death. In the competing risk model for the composite of CVD and death, kidney failure was considered a competing risk since many patients who started kidney replacement therapy were no longer followed for further event thereafter. Results are presented as hazard ratios (HRs) and $95\%$ confidence intervals ($95\%$ CI). To estimate annual eGFR change, generalized linear mixed models were constructed with random intercepts and slopes with an unstructured model for the correlation structure. The results were expressed as estimates (standard errors). In the adjusted models, the variables used in PS score calculation were further adjusted. Spaghetti plots showing the individual trajectories of eGFR during follow-up were drawn to determine patterns of eGFR decline according to cause of CKD. P for the quadratic term was tested using polynomial mixed models with random intercepts and slopes. A P value less than 0.05 was considered statistically significant. SAS 9.4 (SAS Institute, Cary, NC, USA) and R version 3.5.3 (Foundation for Statistical Computing, Vienna, Austria) were used.
## Results
The mean age of the population was 53.5 ± 12.2 years and $38.7\%$ of subjects were female from a total of 2070 patients. At study entry, the mean eGFR of all total patients was 53.2 ± 30.8 mL/min/1.73 m2. By CKD classification, $38.6\%$ of patients were diagnosed with GN, $24.5\%$ with DN, $19.5\%$ with HTN, and $17.4\%$ with PKD.
Baseline characteristics between the four CKD etiology groups showed significant differences (Table 1). All variables showed significant differences among groups and the standardized mean difference were > 0.1 in all variables. In particular, patients with DN (59.2 ± 9.3 years old) or HTN (59.7 ± 10.8 years old) were older than those with GN (49.7 ± 12 years old) or PKD (47 ± 10.7 years old) ($P \leq 0.001$). The proportion of males and the prevalence of pre-existing CVD was higher in the DN ($69.1\%$ and $26.4\%$, respectively) and HTN ($72.6\%$ and $23.8\%$, respectively) groups than in the GN ($55.3\%$ and $7.9\%$, respectively) and PKD ($51\%$ and $6.9\%$, respectively) groups ($P \leq 0.001$ for both). Mean eGFR was lower in DN (36.4 ± 21.8 mL/min/1.73 m2) and HTN (42.3 ± 21.9 mL/min/1.73 m2) groups than GN (60.1 ± 31.4 mL/min/1.73 m2) and PKD (72.7 ± 32.8 mL/min/1.73 m2) groups ($P \leq 0.001$). After overlap weighting for each 6 sets, the statistical differences between groups disappeared in all variables (Table S2). Also, SMD was < 0.001 for all variables in all sets which means a good balance between the overlap-weighted sets. Table 1Comparison of baseline clinical characteristics according to cause of chronic kidney disease before propensity score matching. GlomerulonephritisHypertensive nephropathyDiabetic nephropathyPolycystic kidney diseaseP valueSMDNumber799404507360Age, years49.7 ± 12.159.7 ± 10.859.3 ± 9.347 ± 10.7< 0.0010.74Female, %44.627.53149.2< 0.0010.28BMI, kg/m224.2 ± 3.325.1 ± 3.525.3 ± 3.323.5 ± 3< 0.0010.32Mean blood pressure, mmHg91.6 ± 10.594.3 ± 11.895.3 ± 12.996.8 ± 10.4< 0.0010.25Prevalence of cardiovascular disease, %7.923.826.46.9< 0.0010.35eGFR, ml/min/1.73 m260.1 ± 31.442.3 ± 21.936.4 ± 21.872.7 ± 32.8< 0.0010.77Chronic kidney disease stage< 0.0010.79 G1, %21.83.74.135.0 G2, %24.313.97.726.9 G3a, %16.122.514.613.3 G3b, %18.027.226.011.9 G4, %15.526.235.99.2 G5, %4.36.411.63.6Hemoglobin, g/dL13.2 ± 1.913.3 ± 211.7 ± 1.813.3 ± 1.7< 0.0010.44Uric acid, mg/dL7.1 ± 1.97.3 ± 1.87.4 ± 1.96.1 ± 1.7< 0.0010.39Calcium, mg/dL9.1 ± 0.59.2 ± 0.58.9 ± 0.69.3 ± 0.4< 0.0010.34Phosphorous, mg/dL3.6 ± 0.63.6 ± 0.64.0 ± 0.83.6 ± 0.6< 0.0010.30Albumin, g/dL4.1 ± 0.44.3 ± 0.34.0 ± 0.54.4 ± 0.3< 0.0010.63Total cholesterol, mg/dL178.4 ± 38.5169.1 ± 35.5167.4 ± 43.8179.0 ± 33.6< 0.0010.19HDL-cholesterol, mg/dL51.5 ± 15.646.7 ± 14.043.6 ± 14.054.4 ± 13.9< 0.0010.43LDL-cholesterol, mg/dL99.9 ± 32.194.1 ± 30.591.1 ± 34.0101.6 ± 26.7< 0.0010.2Fasting blood sugar, mg/dL102.1 ± 24.7106.7 ± 25.8136.4 ± 59.096.1 ± 11.0< 0.0010.57Parathyroid hormone, pg/mL¶47.0 (29.8, 69.0)55.1 (37.9, 86.0)66.2 (41.6, 120.3)47.0 (32.9, 70.3)< 0.0010.27Urine protein/creatinine, g/g Cr¶0.6 (0.3, 1.5)0.3 (0.1, 0.7)1.5 (0.5, 3.8)0.1 (0.1, 0.2)< 0.0010.73Urine albumin/creatinine, mg/g Cr457.1 (241.1, 1122.9)207.9 (28.9, 531.8)1015.6 (276.3, 2584.1)39.1 (13.4, 131.0)< 0.0010.76Hs-CRP, mg/dL¶0.7 (0.3, 1.3)0.7 (0.4, 1.8)0.7 (0.3, 1.7)0.5 (0.1, 1.3)< 0.0010.11Diuretics use, %24.934.956.49.4< 0.0010.59Statin use, %51.657.463.926.9< 0.0010.41ACEI or ARB use, %89.782.287.078.6< 0.0010.18Follow up duration, years¶6.0 (4.4, 7.2)5.0 (4.0, 7.0)6.0 (4.2, 7.2)6.6 (4.2, 8.0)¶Presented as Median (quartile 1, quartile 3) due to non-normal distributions. Otherwise, continuous variables are presented as mean ± standard deviation and categorical variables as proportion. ACEI, angiotensin-converting enzyme inhibitors; ARB, angiotensin-receptor blockers; BMI, body mass index; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; Hs-CRP, high sensitivity C-reactive protein; LDL, low density lipoprotein; SMD; standardized mean difference.
During the median 6.0 years of follow-up, there were a total of 565 ($27.3\%$) kidney failure events and 723 ($34.9\%$) composite kidney events. There were 259 ($12.5\%$) events of the composite of CVD and death. The specific cause of CVD and death were summarized in Table S3 and S4. In the PKD population, the most common cause of the cardiovascular event was cerebral hemorrhage or operation/interventions due to cerebral aneurysm ($32.3\%$), each classified as hemorrhagic stroke or other cardiovascular events in Table S3. Among death events, infection ($30\%$) was the major cause of death, followed by malignancy ($25\%$), liver failure ($15\%$), and other causes ($10\%$). These outcomes are different compared to those in the DN group, where, among those patients who had the composite CVD outcome and mortality, $43\%$ had acute coronary syndrome, coronary revascularization, ischemic stroke, or heart failure; cerebral hemorrhagic or aneurysm events accounted for only $5.7\%$ of total cases.
Patients with PKD had a significantly increased risk for kidney failure compared to those with GN (HR 1.82, $95\%$ CI 1.25–2.65), HTN (HR 2.23, $95\%$ CI 1.47–3.38), and DN (HR 1.73, $95\%$ CI 1.05–2.86) in Cox regression analysis. However, the DN, HTN, and GN groups did not show a significant difference in the risk of kidney failure or composite kidney outcome compared to each other (Table 2). DN group had increased risks for the composite outcome of CVD and death compared to the GN (HR 2.07, $95\%$ CI 1.23–3.46), and HTN (HR 1.73, $95\%$ CI 1.08–2.78) groups in cause-specific regression analysis. Patients with DN were not at increased risk of the composite of CVD and death compared to patients with PKD ($$P \leq 0.169$$). Similar results were obtained for DN in multivariate Cox regression analyses. PKD showed increased risk of the composite of CVD and death compared to GN (HR 3.11, $95\%$ CI 1.59–6.05) and HTN (HR 1.94, $95\%$ CI 1.03–3.65) in the Cox regression analysis. In the cause-specific regression analysis, PKD showed a significantly higher risk of the composite outcome of CVD and death compared to GN (HR 2.84, $95\%$ CI 1.36–5.93) however, the significance disappeared in the set with HTN ($$P \leq 0.334$$) (Table 3).Table 2Relative risk for kidney outcomes according to the CKD causes. Cause of CKDKidney failureComposite kidney outcomeHR ($95\%$ CI)P valueHR ($95\%$ CI)P valueGlomerulonephritisReferenceReferenceHypertensive nephropathy0.88 (0.64–1.21)0.4290.80 (0.60–1.07)0.129Diabetic nephropathy1.20 (0.92–1.57)0.1761.18 (0.93–1.49)0.182Polycystic kidney disease1.82 (1.25–2.65)0.0022.00 (1.46–2.75)< 0.001Cause of CKDKidney failureComposite kidney outcomeHR ($95\%$ CI)P valueHR ($95\%$ CI)P valueHypertensive nephropathyReferenceReferenceDiabetic nephropathy1.28 (0.94–1.74)0.1111.27 (0.96–1.69)0.093Polycystic kidney disease2.23 (1.47–3.38)< 0.0012.50 (1.71–3.66)< 0.001Cause of CKDKidney failureComposite kidney outcomeHR ($95\%$ CI)P valueHR ($95\%$ CI)P valueDiabetic nephropathyReferenceReferencePolycystic kidney disease1.73 (1.05–2.86)0.0321.87 (1.17–2.97)0.008CI, confidence interval; CKD, chronic kidney disease; HR; hazard ratio. Table 3Relative risk for composite outcome of CVD and all-cause death according to the CKD causes. Cause of CKDCox regression analysisCause-specific regression analysis*HR ($95\%$ CI)P valueHR ($95\%$ CI)P valueGlomerulonephritisReferenceReferenceHypertensive nephropathy0.97 (0.59–1.59)0.9011.24 (0.71–2.15)0.447Diabetic nephropathy1.67 (1.09–2.56)0.0192.07 (1.23–3.46)0.006Polycystic kidney disease3.11 (1.59–6.05) < 0.0012.84 (1.36–5.93)0.006Cause of CKDCox regression analysisCause-specific regression analysis*HR ($95\%$ CI)P valueHR ($95\%$ CI)P valueHypertensive nephropathyReferenceReferenceDiabetic nephropathy1.82 (1.19–2.79)0.0061.73 (1.08–2.78)0.022Polycystic kidney disease1.94 (1.03–3.65)0.0411.41 (0.70–2.85)0.334Cox regression analysisCause-specific regression analysis*HR ($95\%$ CI)P valueHR ($95\%$ CI)P valueDiabetic nephropathyReferenceReferencePolycystic kidney disease1.84 (0.93–3.63)0.0791.74 (0.79, 3.84)0.169*For the composite outcome of CVD and death, cause-specific regression analysis was conducted as a competing risk analysis using the kidney failure as competing event considering the undetected CVD after starting kidney replacement therapy. CI, confidence interval; CKD, chronic kidney disease; CVD, cardiovascular disease; HR; hazard ratio.
The annual eGFR change of the GN, HTN, DN and PKD groups were − 2.19, − 1.44, − 3.17, and − 3.45 mL/min/1.73 m2 per year, respectively. In the adjusted model, the annual eGFR change were − 2.16, − 1.42, − 3.07, and − 3.37 mL/min/1.73 m2 per year for the GN, HTN, DN and PKD groups, respectively. The DN and PKD groups had faster rates of decline than the GN and HTN groups, while the HTN group had a slower rate of annual eGFR decline than the GN group (Table 4).Table 4Annual changes in eGFR according to the cause of CKD.Annual changes of eGFR (mL/min/1.73 m2 per year)UnadjustedAdjusted¶Estimate (SE)P valueEstimate (SE)P valueCause of CKD< 0.001a< 0.001aGlomerulonephritis− 2.19 (0.106)− 2.16 (0.106)Hypertensive nephropathy− 1.44 (0.151)< 0.001b− 1.42 (0.152)< 0.001bDiabetic nephropathy− 3.17 (0.16)< 0.001b< 0.001c− 3.07 (0.161)< 0.001b< 0.001cPolycystic kidney disease− 3.45 (0.155)< 0.001b< 0.001c0.201d− 3.37 (0.156)< 0.001b< 0.001c0.176d¶Adjusted with age, sex, body mass index, CKD stage, mean blood pressure, cardiovascular disease, hemoglobin, uric acid, calcium, phosphorous, albumin, total cholesterol, high-density lipid cholesterol, low-density lipid cholesterol, fasting blood sugar, intact parathyroid hormone, urine protein-to-creatinine ratio, high sensitivity C-reactive protein, diuretics use, statin use, and ACE inhibitor or ARB use.aP value for the interaction term between the cause of CKD and time effect in the mixed model.bP value for each estimated of CKD causes compared to the glomerulonephritis as reference.cP value for each estimated of CKD causes compared to the hypertensive nephropathy as reference.dP value for each estimated of CKD causes compared to the diabetic nephropathy as reference. ACE, angiotensin converting enzyme; ARB, angiotensin receptor blocker; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; SE, Standard error.
The rate of eGFR decline in each group was analyzed according to CKD stage at entry (Fig. 2). The fastest annual decline in eGFR was observed for those patients with PKD stages G3a and G3b (− 4.94 and − 4.38 mL/min/1.73 m2 per year, respectively). The overall rate of annual eGFR decline was also fast in the DN group, ranging from − 3.87 to − 2.68 mL/min/1.73 m2 per year for stages G1 to G4. In the HTN group, rates of annual eGFR decline were slow for stages G2 to G3b but eGFR declined slightly faster in stages G4 and G5. In the GN group, rates of annual eGFR decline were faster in the more advanced CKD stages. When we visualize the trajectory patterns of eGFR decline, CKD etiologies were classified into two groups: the DN group (P for quadratic term = 0.608) group showed a linear decline pattern in eGFR, while the GN, HTN, and PKD groups (P for quadratic term < 0.001, for all) showed a convex decline pattern with the acceleration of the annual eGFR decline as the eGFR lowered (Fig. 3).Figure 2Estimated glomerular filtration rate changes according to baseline CKD stages in each CKD etiology group. CKD of each groups with A1-A3 are classified using eGFR as followed based on KDIGO guideline30: stage G1 ≥ 90 mL/min/1.73 m2, stage G2 60–89 mL/min/1.73 m2, stage G3a 45–59 mL/min/1.73 m2, stage G3b 30–44 mL/min/1.73 m2, stage G4 15–29 mL/min/1.73 m2, and stage G5 < 15 mL/min/1.73 m2 without kidney replacement therapy. CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate. Figure 3Spaghetti plots (up) and trajectory of eGFR changes (down) during follow-up in each CKD etiology group. Each colors of spaghetti plot indicates trajectory of each patients (up) and the black line represents the mean estimated eGFR trajectory of each groups (with $95\%$ CI represented by the gray shaded area). The trajectory are classified into two types. DN showed a linear decline in eGFR but GN, HTN and PKD showed a convex decline with acceleration of the annual eGFR decline in advanced CKD stages. DN, diabetic nephropathy; eGFR, estimated glomerular filtration rate; GN, glomerulonephritis; HTN, hypertensive nephropathy; PKD, polycystic kidney disease.
## Discussion
The KDIGO Guideline states that the cause of CKD should be considered as one of the important predictors of the outcome5. A few studies tried to evaluate the outcome difference in the CKD population according to specific causes7–10. Post-hoc study of a clinical trial showed that those patients with PKD had a higher risk of kidney failure and a lower risk of death than those with CKD with other etiologies7. Studies of the Canadian Study of Prediction of Death, Dialysis and Interim Cardiovascular Events (CanPREDDICT) cohort data reported the relative risks of a few adverse outcomes according to the CKD etiologies8,9. The CKD in Children (CKiD) study compared the rate of progression of kidney disease according to the cause of CKD in children10. However, the results from these studies are difficult to generalize to the entire CKD population because they include only a subset of the major CKD etiology or analyzed for only some of the major outcomes. This might be due to the difficulty of the study design and analysis technique since many risk factors of kidney disease progression are related to the cause of CKD and can act as potential confounders.
In this prospective cohort study, we compared the relative risks of both kidney outcomes and the composite outcome of CVD and death according to the cause of CKD. Patients were classified into GN, HTN, DN, or PKD groups based either on pathologic diagnoses or clinical judgement criteria at study entry. The baseline characteristics differed according to the cause of CKD in our study population. To overcome this limitation, we used the overlap weighting method. This is particularly advantageous when the comparator groups are initially very different from each other and can achieve good balance and minimize the variances as shown in previous studies15,16. Several studies adopted this method to analyze the effect of sex, cancer type, monitoring or treatment on the outcome between groups with significant differences in baseline characteristics16–20. By employing this method, we could analyze the hazard ratio for the outcomes between two groups of CKD causes after adjusting for potential confounders. The result showed that patients with PKD had significantly increased risks for kidney outcomes compared to other CKD causes. Surprisingly, the DN group did not show an increased risk of kidney failure compared to other CKD causes. However, the patients in the DN group showed worse outcomes regarding the composite outcome of CVD and mortality compared to the HTN and GN groups. The high risk of CVD and death in the DN group shown in this study was consistent with other well-known studies21,22.
We further analyzed the rates of annual eGFR decline to better understand kidney disease progression patterns according to the specific causes of CKD. The rate of GFR decline was faster in the DN and PKD groups compared to the GN and HTN groups. The DN and HTN population had a similar rate of annual eGFR decline to that shown in previous reports; however, the annual rates of decline of eGFR in the PKD group was relatively slower in our study than those reported in previous studies23,24. This could be due to differences in baseline clinical characteristics, including CKD stages, PKD genotypes and/or effects of ethnicity. In this study, there are both early and advanced stage PKD patients in our cohort (about $60\%$ were stage G1 or G2) whereas PKD patients in CRISP and HALT studies only included early stage CKD patients and MDRD study enrolled only advanced stage CKD patients23,24.
Although the annual eGFR declining rate in the PKD group was slightly slower than in previous reports, patients in the PKD group showed the poorest kidney outcomes compared to those with other causes of CKD. The annual rate of eGFR decline was the fastest in the PKD population. The PKD group showed an increased risk of kidney failure with HRs of 1.73, 1.8 and 2.2 compared to the DN, GN, and HTN groups, respectively. This is a similar result to previous reports9,24. Therefore, more efforts on early detection, assessment, and proper management of PKD-related risk factors such as genotype, kidney volumes, hypertension, and kidney-related complications may improve individual PKD patients’ kidney outcomes in the future.
In our study results, the risk of poor kidney outcome was not increased in the DN group, compared to other CKD causes when the confounding factors were adjusted using the overlap weighting methods. This implies the importance of managing the common risk factors and comorbidities in the DN population. However, the risk of CVD and mortality was significantly increased in the DN group compared to the GN and HTN groups. The high risk of CVD and death in the DN population is well known and there have been efforts to find out effective treatments to improve the outcome. However, the strict control of blood glucose levels showed improvement in kidney outcomes but did not in the CVD or death21,22. These differences between CVD and kidney outcomes in DN population shown in the previous studies and in our study suggest that different pathophysiology would exist between CVD or mortality and kidney progression.
The CVD and mortality risk was also higher in the PKD population than the GN and HTN population. The PKD group had a similar risk of CVD and mortality to the DN group, but the specific causes of CVD and mortality were different between PKD and DN groups. In the DN group, about $43\%$ of the composite of CVD and mortality were major adverse cardiovascular events such as acute coronary syndrome, coronary revascularization, ischemic stroke, or heart failure which were known major CVD events from the previous studies25. However, in the PKD group events due to cerebral aneurysm and infection were the most frequent cause of CVD and death, respectively which correspondence with previous studies26.
In this study, we further analyzed eGFR decline patterns according to the cause of CKD. A linear eGFR trajectory was observed in DN group, and the rate of annual eGFR decline was faster in the earlier CKD stages. GN, HTN, and PKD groups showed faster annual eGFR declining in advanced CKD stages. This result is similar to the previous studies that reported trajectory eGFR decline of DN and PKD23,27. In DN population, there was a linear association between eGFR and age over time in overall27. In PKD population, a non-linear curved eGFR trajectory was seen regardless of the kidney growth rate23. In this study, we observed the kidney deterioration pattern and acceleration time differed according to the cause of CKD using a prospective longitudinal cohort. Therefore, patient follow-up and monitoring strategies should be individualized according to the CKD stages, and cause of CKD.
This study has several advantages over existing studies in providing important information about the natural course of CKD progression and extrarenal complications according to CKD etiology. Here, we provided basic information about the hazard ratio of major outcomes from four major causes of CKD within a prospective cohort followed over a long-term period. Robust and up-to-date statistical methods were used to compare the relative risk of major outcomes by the causes of CKD after adjusting for possible confounders. The group size of PKD and GN were big enough for the comparison and statistics. CKD patients older than 18 years with any CKD stage were enrolled to reflect the overall CKD population.
Although, there are several limitations to this study. Baseline characteristics differed among groups. Therefore we used the overlap weighting method to adjust for possible confounding factors. However, there may still have been residual confounders that affected our findings. Furthermore, the etiologic diagnoses of CKD were based on clinical criteria rather than pathologic diagnoses for many patients. Thus, some clinically diagnosed patients might have been misclassified. The issue due to cross-group events, such as the occurrence of new GNs during follow-up in patients in the HTN group, was not considered in the analysis of this study. We could not analyze the risk of each individual type of cardiovascular event or focused on major adverse cardiovascular events since the overall incidence of CV events was relatively lower in our cohort, compared to the Western CKD cohorts25,28. The overall CVD incidence is similar to that of the Japanese CKD cohort (CKD-JAC) and this might be the characteristic of the Asian CKD population29. Since this is a result from a cohort of Asian CKD patients, the present study warrants further investigation for patients of non-Caucasian ethnicity. Finally, a group with a different composition of GN subtypes may have shown different hazard ratio compared to the other CKD groups evaluated in this study.
We found that patients with PKD had higher risk of kidney progression than patients with DN, GN, or HTN. After adjustment, the DN group did not show an increased risk of kidney failure but had a higher risk of CVD and mortality than patients with GN and HTN. Our findings support the importance of individualized monitoring and management of CKD patients based on the etiology and stage of CKD.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-29844-x.
## References
1. Mills KT, Xu Y, Zhang W. **A systematic analysis of worldwide population-based data on the global burden of chronic kidney disease in 2010**. *Kidney Int.* (2015.0) **88** 950-957. DOI: 10.1038/ki.2015.230
2. Xie Y, Bowe B, Mokdad AH. **Analysis of the Global Burden of Disease study highlights the global, regional, and national trends of chronic kidney disease epidemiology from 1990 to 2016**. *Kidney Int.* (2018.0) **94** 567-581. DOI: 10.1016/j.kint.2018.04.011
3. Levey AS, de Jong PE, Coresh J. **The definition, classification, and prognosis of chronic kidney disease: a KDIGO Controversies Conference report**. *Kidney Int.* (2011.0) **80** 17-28. DOI: 10.1038/ki.2010.483
4. Bommer J. **Prevalence and socio-economic aspects of chronic kidney disease**. *Nephrol Dial Transplant.* (2002.0) **17** 8-12. DOI: 10.1093/ndt/17.suppl_11.8
5. **KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease**. *Kidney Int. Suppl.* (2013.0) **3** 1-150
6. 6.Morgan E. Grams SPM. In: Morgan E. Grams SPM, editor. Comprehensive clinical nephrology. 6 ed: Elsevier; 2018:903–912.
7. Haynes R, Staplin N, Emberson J. **Evaluating the contribution of the cause of kidney disease to prognosis in CKD: results from the Study of Heart and Renal Protection (SHARP)**. *Am. J. Kidney Dis.* (2014.0) **64** 40-48. DOI: 10.1053/j.ajkd.2013.12.013
8. Langsford D, Tang M, Cheikh Hassan HI, Djurdjev O, Sood MM, Levin A. **The association between biomarker profiles, etiology of chronic kidney disease, and mortality**. *Am. J. Nephrol.* (2017.0) **45** 226-234. DOI: 10.1159/000454991
9. de Chickera S, Akbari A, Levin A. **The risk of adverse events in patients with polycystic kidney disease with advanced chronic kidney disease**. *Can. J. Kidney Health Dis.* (2018.0) **5** 2054358118774537. DOI: 10.1177/2054358118774537
10. Pierce CB, Cox C, Saland JM, Furth SL, Munoz A. **Methods for characterizing differences in longitudinal glomerular filtration rate changes between children with glomerular chronic kidney disease and those with nonglomerular chronic kidney disease**. *Am. J. Epidemiol.* (2011.0) **174** 604-612. DOI: 10.1093/aje/kwr121
11. Oh KH, Park SK, Park HC. **KNOW-CKD (KoreaN cohort study for Outcome in patients With Chronic Kidney Disease): design and methods**. *BMC Nephrol.* (2014.0) **15** 80. DOI: 10.1186/1471-2369-15-80
12. Belibi FA, Edelstein CL. **Unified ultrasonographic diagnostic criteria for polycystic kidney disease**. *J. Am. Soc. Nephrol.* (2009.0) **20** 6-8. DOI: 10.1681/ASN.2008111164
13. Levey AS, Stevens LA, Schmid CH. **A new equation to estimate glomerular filtration rate**. *Ann Intern Med.* (2009.0) **150** 604-612. DOI: 10.7326/0003-4819-150-9-200905050-00006
14. Li F, Thomas LE, Li F. **Addressing extreme propensity scores via the overlap weights**. *Am. J. Epidemiol.* (2019.0) **188** 250-257. PMID: 30189042
15. Thomas LE, Li F, Pencina MJ. **Overlap weighting: a propensity score method that mimics attributes of a randomized clinical trial**. *JAMA* (2020.0) **323** 2417-2418. DOI: 10.1001/jama.2020.7819
16. Mehta N, Kalra A, Nowacki AS. **Association of use of angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers with testing positive for coronavirus disease 2019 (COVID-19)**. *JAMA Cardiol.* (2020.0) **5** 1020-1026. DOI: 10.1001/jamacardio.2020.1855
17. Chacon-Alberty L, Ye S, Daoud D. **Analysis of sex-based differences in clinical and molecular responses to ischemia reperfusion after lung transplantation**. *Respir Res.* (2021.0) **22** 318. DOI: 10.1186/s12931-021-01900-y
18. Yang Y, Li Y, Du X. **Prognostic analysis of gastric signet ring cell carcinoma and hepatoid adenocarcinoma of the stomach: A propensity score-matched study**. *Front Oncol.* (2021.0) **11** 716962. DOI: 10.3389/fonc.2021.716962
19. Karter AJ, Parker MM, Moffet HH, Gilliam LK, Dlott R. **Association of real-time continuous glucose monitoring with glycemic control and acute metabolic events among patients with insulin-treated diabetes**. *JAMA* (2021.0) **325** 2273-2284. DOI: 10.1001/jama.2021.6530
20. Liao SY, Petrache I, Fingerlin TE, Maier LA. **Association of inhaled and systemic corticosteroid use with Coronavirus Disease 2019 (COVID-19) test positivity in patients with chronic pulmonary diseases**. *Respir Med.* (2021.0) **176** 106275. DOI: 10.1016/j.rmed.2020.106275
21. Genuth S, Ismail-Beigi F. **Clinical Implications of the ACCORD Trial**. *J Clin Endocrinol Metab.* (2012.0) **97** 41-48. DOI: 10.1210/jc.2011-1679
22. Patel A, MacMahon S. **Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes**. *N Engl J Med.* (2008.0) **358** 2560-2572. DOI: 10.1056/NEJMoa0802987
23. Yu ASL, Shen C, Landsittel DP. **Long-term trajectory of kidney function in autosomal-dominant polycystic kidney disease**. *Kidney Int.* (2019.0) **95** 1253-1261. DOI: 10.1016/j.kint.2018.12.023
24. Klahr S, Breyer JA, Beck GJ. **Dietary protein restriction, blood pressure control, and the progression of polycystic kidney disease Modification of Diet in Renal Disease Study Group**. *J. Am. Soc. Nephrol.* (1995.0) **5** 2037-2047. DOI: 10.1681/ASN.V5122037
25. Denker M, Boyle S, Anderson AH. **Chronic Renal Insufficiency Cohort Study (CRIC): Overview and summary of selected findings**. *Clin. J. Am. Soc. Nephrol.* (2015.0) **10** 2073-2083. DOI: 10.2215/CJN.04260415
26. Fick GM, Johnson AM, Hammond WS, Gabow PA. **Causes of death in autosomal dominant polycystic kidney disease**. *J. Am. Soc. Nephrol.* (1995.0) **5** 2048-2056. DOI: 10.1681/ASN.V5122048
27. Warren B, Rebholz CM, Sang Y. **Diabetes and trajectories of estimated glomerular filtration rate: A prospective cohort analysis of the atherosclerosis risk in communities study**. *Diabetes Care* (2018.0) **41** 1646-1653. DOI: 10.2337/dc18-0277
28. Ryu H, Kim J, Kang E. **Incidence of cardiovascular events and mortality in Korean patients with chronic kidney disease**. *Sci. Rep.* (2021.0) **11** 1131. DOI: 10.1038/s41598-020-80877-y
29. Tanaka K, Watanabe T, Takeuchi A. **Cardiovascular events and death in Japanese patients with chronic kidney disease**. *Kidney Int.* (2017.0) **91** 227-234. DOI: 10.1016/j.kint.2016.09.015
30. Levey AS, Eckardt KU, Dorman NM. **Nomenclature for kidney function and disease: Report of a Kidney Disease: Improving Global Outcomes (KDIGO) Consensus Conference**. *Kidney Int.* (2020.0) **97** 1117-1129. DOI: 10.1016/j.kint.2020.02.010
|
---
title: Short-term tamoxifen administration improves hepatic steatosis and glucose
intolerance through JNK/MAPK in mice
authors:
- Zhiqiang Fang
- Hao Xu
- Juanli Duan
- Bai Ruan
- Jingjing Liu
- Ping Song
- Jian Ding
- Chen Xu
- Zhiwen Li
- Kefeng Dou
- Lin Wang
journal: Signal Transduction and Targeted Therapy
year: 2023
pmcid: PMC9981902
doi: 10.1038/s41392-022-01299-y
license: CC BY 4.0
---
# Short-term tamoxifen administration improves hepatic steatosis and glucose intolerance through JNK/MAPK in mice
## Abstract
Nonalcoholic fatty liver disease (NAFLD) which is a leading cause of chronic liver diseases lacks effective treatment. Tamoxifen has been proven to be the first-line chemotherapy for several solid tumors in clinics, however, its therapeutic role in NAFLD has never been elucidated before. In vitro experiments, tamoxifen protected hepatocytes against sodium palmitate-induced lipotoxicity. In male and female mice fed with normal diets, continuous tamoxifen administration inhibited lipid accumulation in liver, and improved glucose and insulin intolerance. Short-term tamoxifen administration largely improved hepatic steatosis and insulin resistance, however, the phenotypes manifesting inflammation and fibrosis remained unchanged in abovementioned models. In addition, mRNA expressions of genes related to lipogenesis, inflammation, and fibrosis were downregulated by tamoxifen treatment. Moreover, the therapeutic effect of tamoxifen on NAFLD was not gender or ER dependent, as male and female mice with metabolic disorders shared no difference in response to tamoxifen and ER antagonist (fulvestrant) did not abolish its therapeutic effect as well. Mechanistically, RNA sequence of hepatocytes isolated from fatty liver revealed that JNK/MAPK signaling pathway was inactivated by tamoxifen. Pharmacological JNK activator (anisomycin) partially deprived the therapeutic role of tamoxifen in treating hepatic steatosis, proving tamoxifen improved NAFLD in a JNK/MAPK signaling-dependent manner.
## Introduction
Nonalcoholic fatty liver disease (NAFLD) which is a metabolic-associated chronic liver disease,1,2 has become the leading cause of hepatocellular carcinoma3 and the leading reason for liver transplantation in western countries.4 In terms of pathological characteristics, NAFLD can be categorized as nonalcoholic fatty liver and nonalcoholic steatohepatitis (NASH) which is prone to end-stage liver diseases.5 *The pathogenesis* of NAFLD is complex and tightly connected with obesity, type 2 diabetes mellitus, hyperlipidemia, and metabolic syndrome.6 Although the disease burden is increasing, no FDA-approved drug has been found to effectively treat NAFLD or NASH so far.
Tamoxifen, synthesized in 1966, is a selective estrogen receptor (ER) modulator (SERM).7 Since there are multiple target tissues for estrogen, such as endometrium, cardiovascular system, bone, brain, and liver,8 off-target effects may occur following tamoxifen administration. Tamoxifen has been used for the treatment of breast cancer for 50 years and the first case was reported in 1971.9 Growing evidence unveiled long-term tamoxifen treatment in breast cancer patients led to metabolic disorders. In 1995, Pratt et al. first discovered tamoxifen treatment resulted in steatohepatitis.10 Afterwards, Van Hoof, Ogawa, and Cai et al. reported the similar clinical phenomenon.11–13 However, a clinical trial subsequently demonstrated that tamoxifen only increased the risk of NASH in overweight and obese women.14 In addition, a few in vitro and in vivo experiments proved that tamoxifen promoted lipid accumulation in hepatocytes and enhanced fatty acid biosynthesis.15–19 However, some other findings suggested that tamoxifen protected hepatocytes against lipotoxicity and steatosis.20,21 Therefore, the effect of tamoxifen on lipid metabolism seems to be controversial.
Apart from cancer treatment, tamoxifen is frequently used to generate genetic mutations in mice by inducing Cre/loxp recombination system. The CreER/loxp system which is widely used in transgenic mice can induce somatic mutations at a chosen time and/or in a specific tissue.22 Through fusing Cre recombinase with a mutated ligand-binding domain of ER, the Cre recombinase can be activated by tamoxifen instead of estradiol.23 The route of tamoxifen induction varies from different laboratories. Wang et al. administrated tamoxifen orally to adult mice at 100 mg/kg for 4 consecutive days to induce Cre recombination,24 while in O’Shea’s publication, tamoxifen was intraperitoneally injected at 50 mg/kg per day for 5 consecutive days.25 Unavoidably, some off-target effects of tamoxifen induction were reported in retina and skeleton.26,27 Occasionally, we found that following a normal routine of tamoxifen induction, lipid accumulation and hepatic steatosis were prominently reduced in both normal and NAFLD-established mice, which aroused our great interest. It seems that the off-target effect of tamoxifen may probably be protective to lipid metabolism.
In this article, we investigated the role of short-term tamoxifen administration in NAFLD mouse models and determined JNK/MAPK signaling pathway was involved in tamoxifen-driven treatment. Also, we discussed the safety and potential application of tamoxifen in the treatment on NAFLD.
## Tamoxifen protects hepatocytes against lipotoxicity in vitro
To explore the impact of tamoxifen on hepatocytes in vitro, we utilized primary hepatocytes, two human hepatocyte cell lines—Huh7 and HepG2 and a mouse hepatocyte cell line—AML12. To induce lipotoxicity, cells were exposed to 0.3 mM sodium palmitate for 36 h. For treatment, different concentrations of tamoxifen (10, 20, 40 μM) were added into cell medium for another 36 h. As shown by Oil-red O (ORO) staining (Fig. 1a, b), in primary hepatocytes and all three cell lines, sodium palmitate-triggered lipid accumulation could be notably reduced by tamoxifen in a dose-dependent manner. However, tamoxifen treatment did not decrease cellular TC contents in all these cells and did not change TG contents in HepG2 cell line. To investigate whether tamoxifen increased cytotoxicity, we performed trypan blue staining in Huh7 and AML12 cell lines treated with PA or PA plus different concentrations of TAM to examine cell viability and found that tamoxifen treatment did not increase cytotoxicity (Supplementary Fig. S1a, b). To investigate the role of tamoxifen on lipid metabolism, we performed RT-qPCR experiment and found that tamoxifen suppressed expressions of genes related to lipid uptake (FATP2, FATP5, and CD36), de novo lipogenesis (SREBP1c, FASN, SCD1, and ACC1), fatty acid oxidation (PPARα, CPT1α, and ACOX1) and TG export (APOB and MTTP) in a dose-dependent manner (Supplementary Fig. S1c). These data collectively indicated that tamoxifen could effectively protect hepatocytes from lipotoxicity in vitro without increasing cytotoxicity. Fig. 1Tamoxifen decreased hepatocyte lipotoxicity in vitro. a Huh7, HepG2, AML12 cells, and primary hepatocytes were seeded in six-well plates. After 12 h, 0.3 mM sodium palmitate was added to the medium and after 36 h, DMSO/tamoxifen (10, 20, 40 μM) was added. After 36 h, cells were stained with ORO, pictured using an inverted phase contrast microscope, and quantified by ORO staining areas. Scale bar: 100 μm. b *Statistical analysis* of ORO staining. For cellular TC and TG test, cells were digested by trypsin, washed with PBS and collected using centrifuge. The following steps were performed according to the manufacturer’s protocols. All experiments were repeated at least three biological times. Bars = means ± SD; $$n = 3$$–10; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$
## Continuous tamoxifen administration does not worsen liver function in normal mice
We then evaluated the role of continuous injection of tamoxifen in normal mice. To investigate whether tamoxifen decreases food intake and body weight due to appetite suppression, mice with free access to normal food and water were administrated with 100 mg/kg tamoxifen intraperitoneally every other day for 4 weeks. With the prolongation of tamoxifen treatment, the food intake and body weight decreased in tamoxifen-treated group (Fig. 2d, e). To exclude the effect of appetite suppression, male and female C57BL/6 mice were pair-fed with normal diet and administrated with tamoxifen intraperitoneally for 1, 2, 4, and 8 weeks, respectively (Fig. 2a). Continuous tamoxifen administration for 8 weeks made no change to the body weight either in male or in female mice (Fig. 2b), indicating pair-feeding largely offset the potential side-effect of tamoxifen on appetite suppression. As shown in Fig. 2c, no decline of food intake was observed as well. Then, ALT and AST tests confirmed that continuous tamoxifen injection did not worsen liver function (Supplementary Fig. S2a). On weeks 4 and 8, the serum level of ALT even decreased a little bit by using tamoxifen, which caused no hepatotoxicity. Fig. 2Tamoxifen inhibited lipid accumulation in mice fed with normal diet. a Dosing scheme of tamoxifen on male and female C57BL/6 mice fed with normal diets. Dose of tamoxifen: 100 mg/kg. b Since initiating tamoxifen administration, mice were pair-fed and body weight was recorded weekly until mice were sacrificed. c Food consumption weight per cage was recorded when mice were administrated with tamoxifen. d Since initiating tamoxifen administration, mice were free-fed and body weight was recorded every other day for 4 weeks. e Food consumption weight per cage was recorded weekly when free-fed mice were administrated with tamoxifen. f Serum TC, HDL-C and LDL-C analysis of male and female mice fed with normal diets and administrated with vehicle/tamoxifen for 1, 2, 4, 8 weeks, respectively. g GTT test was performed on male and female mice administrated with tamoxifen or vehicle for 8 weeks and area under curve (AUC) was calculated and compared. h ITT test was performed on male and female mice administrated with tamoxifen or vehicle for 8 weeks and area under curve (AUC) was calculated and compared. i Frozen liver sections from male and female mice administrated with tamoxifen or vehicle for 1 and 4 weeks was performed ORO staining and ORO staining area was quantitatively compared. Scale bar:100 μm. j Frozen liver sections from male and female mice administrated with tamoxifen or vehicle for 8 weeks were performed ORO staining and ORO staining area was quantitatively compared. Scale bar:100 μm. Bars = means ± SD; $$n = 3$$– 5; ns, no significance; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$
## Tamoxifen inhibits lipid accumulation in mice fed with normal diet
Afterwards, we detected lipid accumulation in blood. No significant difference was found in serum TG levels between the tamoxifen-treated group and the control (Supplementary Fig. S2a). However, serum TC, HDL-C, and LDL-C levels were distinctly lowered by tamoxifen, unveiling the potential effectiveness of tamoxifen on cholesterol metabolism (Fig. 2f). Considering that insulin resistance is a major contributor to NAFLD, we examined glucose and insulin tolerance through GTT and ITT assays. As early as one week after tamoxifen administration, male mice exhibited better glucose tolerance while female ones showed no difference (Supplementary Fig. S2b), which was consistent with the change on week 4 (Supplementary Fig. S2c). Intriguingly, following 8-week administration, both male and female mice tolerated better with glucose and insulin in tamoxifen-treated group than in control group, which were determined by GTT and ITT assays respectively (Fig. 2g, h). To investigate lipid accumulation in liver, we performed ORO staining. One week or four weeks after tamoxifen administration, less hepatic lipid deposition could only be found in female mice (Fig. 2i). However, hepatic lipid accumulation was largely reduced by tamoxifen in both male and female mice following 8 weeks (Fig. 2j). RT-qPCR analyses showed that tamoxifen only decreased the expression of SCD1 (Supplementary Fig. S2d). These results collectively suggested that continuous injection of tamoxifen within 8 weeks could effectively decrease circulating TC contents, improve glucose and insulin intolerance and inhibit lipid deposition in mice fed with normal diet.
## Tamoxifen ameliorates MCD and CDAA-induced hepatic steatosis
To identify the potential role of tamoxifen in NASH treatment, we established MCD and CDAA diet-induced NASH mouse models. Male mice were fed with MCD diet for 6 weeks or CDAA diet for 10 weeks, and were then administrated with 100 mg/kg tamoxifen intraperitoneally for 5 consecutive days (Fig. 3a, c). In MCD diet-induced NASH mice, liver weight slightly decreased in tamoxifen-treated group while body weight and liver to body weight ratio both remained unchanged (Supplementary Fig. S3a). In mice fed with CDAA diet, no change of body weight was found before or after tamoxifen administration (Supplementary Fig. S3b). Different from MCD diets, CDAA diets didn’t cause body weight consumption (Supplementary Fig. S3c). In addition, we performed GTT experiment and found that tamoxifen partially improved glucose intolerance in CDAA diet-induced NASH mice (Supplementary Fig. S3d). ORO and H&E staining revealed that tamoxifen treatment notably decreased hepatic lipid accumulation caused by MCD and CDAA diets (Fig. 3b, d), which was confirmed by liver TG measurements (Fig. 3h). However, the number of F$\frac{4}{80}$ positive cells was not altered by tamoxifen as shown by F$\frac{4}{80}$ immunohistochemistry (IHC) and immunofluorescence (IF) staining (Fig. 3b, d). In addition, PSR staining showed that the Sirius Red positive stain was unchanged by the use of tamoxifen (Fig. 3b, d), indicating no manifestation of liver inflammation or fibrosis was observed following tamoxifen treatment. Fig. 3Tamoxifen alleviated hepatic steatosis in MCD and CDAA-induced models. a Dosing scheme of tamoxifen on male C57BL/6 mice fed with MCD diets for 6 weeks. Dose of tamoxifen: 100 mg/kg. b Liver sections from tamoxifen group and vehicle group in mice fed with MCD diets were performed H&E, ORO, F$\frac{4}{80}$ IF, and PSR staining. H&E fat cavitation area, ORO staining area, F$\frac{4}{80}$ positive cells percentage and PSR staining area were quantitatively compared. Scale bar: 100 μm. c Dosing scheme of tamoxifen on male C57BL/6 mice fed with CDAA diets for 10 weeks. Dose of tamoxifen: 100 mg/kg. d Liver sections from tamoxifen group and vehicle group in mice fed with CDAA diets were performed H&E, ORO, F$\frac{4}{80}$ IF, and PSR staining. H&E fat cavitation area, ORO staining area, F$\frac{4}{80}$ positive cells percentage and PSR staining area were quantitatively compared. Scale bar:100 μm. e RNA was extracted from liver tissues of MCD diets-induced NASH mice administrated with tamoxifen or vehicle and expression of lipogenesis, inflammation, and fibrosis-related genes was determined by RT-qPCR with β-actin as an internal control. f RNA was extracted from liver tissues of CDAA diets-induced NASH mice administrated with tamoxifen or vehicle and expression of lipogenesis, inflammation, and fibrosis-related genes were determined by RT-qPCR with β-actin as an internal control. g *Serum analysis* of TC, TG, HDL and LDL in MCD diets-induced mice administrated with tamoxifen or vehicle. h Liver tissues from MCD and CDAA-induced NASH mice administrated with tamoxifen or vehicle were harvested and TG concentrations were measured using commercial kits. Hepatic TG contents were normalized by hepatic protein levels. Bars = means ± SD; $$n = 3$$–6; ns, no significance; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$ Then we extracted RNA from liver tissues of mice fed with MCD diet. RT-qPCR analyses demonstrated that marker genes related to fatty acid synthesis (Srebp1c, SCD1, and ACC1), inflammatory response (CCL2, CXCL10, IL-1β), and fibrosis (α-SMA and Ctgf) were downregulated in tamoxifen-treated group (Fig. 3e). In CDAA diet-induced liver, genes of lipogenesis (Srebp1c, CD36, and SCD1), inflammation (CXCL10 and IL-1β) and fibrogenesis (Col1a1 and Col3a1) were also downregulated by tamoxifen and genes regarding fatty acid oxidation (PPARα and ACOX1) were reduced as well (Fig. 3f). Furthermore, tamoxifen treatment decreased serum levels of TC, TG, HDL-C and LDL-C in MCD diet-induced NASH mice (Fig. 3g). Above results proved that short-term tamoxifen treatment effectively improved hepatic steatosis in mice fed with MCD or CDAA-diet.
## Tamoxifen attenuates HFD-induced hepatic steatosis and glucose intolerance
In order to establish NAFLD models with metabolic abnormality, male mice were fed with HFD diet for 20 weeks. Then, 100 mg/kg tamoxifen was administrated intraperitoneally three times a week for 4 weeks (Fig. 4a). Compared to the control, mice fed with HFD diet showed increased body weight and liver weight, which were reduced by tamoxifen evidently (Supplementary Fig. S4a, b). Consistent with the findings in MCD and CDAA diet-induced models, tamoxifen alleviated hepatic steatosis but did not alter the phenotypes of inflammation and fibrosis in HFD-induced mice (Fig. 4b, f). Additionally, tamoxifen treatment decreased serum levels of ALT, AST, TC, HDL-C, and LDL-C instead of TG (Fig. 4c), suggesting that tamoxifen ameliorated lipid-induced liver injury and decreased serum cholesterol contents (Fig. 4c). Moreover, GTT and ITT assays confirmed that mice fed with HFD diet recovered significantly from glucose and insulin intolerance following tamoxifen treatment (Fig. 4d, e). Afterwards, we extracted RNA from isolated primary hepatocytes of fatty liver. RT-qPCR analyses demonstrated that HFD-triggered upregulation of lipogenic genes, including Srebp1c, CD36, SCD1, FASN, ACC1, and PPARγ, was neutralized by using tamoxifen (Fig. 4g). To reconfirm the molecular changes in tamoxifen-treated liver, CD36, SCD1, CCL2, CXCL10, Col3a1, and Timp1 were measured by RT-qPCR in total liver tissue and all of them were downregulated by tamoxifen (Supplementary Fig. S4c).Fig. 4Tamoxifen ameliorated NAFLD in HFD-induced model. a Dosing scheme of tamoxifen on male C57BL/6 mice fed with high-fat diets for 20 weeks. Dose of tamoxifen: 100 mg/kg. b Liver sections from tamoxifen group and vehicle group in mice fed with high fat diets were performed H&E, ORO, F$\frac{4}{80}$ IF, and PSR staining. Mice fed with normal diets and administrated with vehicle were as controls. H&E fat cavitation area, ORO staining area, F$\frac{4}{80}$ positive cells percentage and PSR staining area were quantitatively compared. Scale bar:100 μm. c *Serum analysis* of ALT, AST, TC, TG, and LDL in high fat diets-induced mice administrated with tamoxifen or vehicle. Mice fed with normal diets and administrated with vehicle were as controls. d GTT experiment was performed on control mice or high fat diets-induced mice administrated with tamoxifen or vehicle and area under curve was quantitatively compared. e ITT experiment was performed on control mice or high fat diets-induced mice administrated with tamoxifen or vehicle and area under curve was quantitatively compared. f Hepatic TG levels were examined and normalized by protein levels. g RNA was extracted from isolated hepatocytes of control mice or high fat diets-induced mice administrated with tamoxifen or vehicle and expression of lipogenic genes was determined by RT-qPCR with β-actin as an internal control. h Dosing scheme of oral tamoxifen administration on male C57BL/6 mice fed with high-fat diets for 20 weeks. Dose of tamoxifen: 100 mg/kg. i Male C57BL/6 mice fed with high-fat diets for 20 weeks were administrated with 100 mg/kg tamoxifen intraperitoneally for 2 weeks and ceased treatment for 2 or 4 weeks. j Liver sections from male HFD-induced mice administrated with tamoxifen or vehicle orally for 2 weeks were performed H&E and ORO staining and H&E fat cavitation area, and ORO staining area was quantitatively compared. Scale bar:100 μm. k Hepatic TG levels were examined and normalized by protein levels. l GTT and ITT experiments were performed on male HFD-induced mice administrated with tamoxifen or vehicle orally for 2 weeks and area under curve was quantitatively compared. m Liver sections from male HFD-induced mice administrated with tamoxifen or vehicle for 2 weeks and ceased tamoxifen treatment for 2 or 4 weeks were performed H&E and ORO staining and H&E fat cavitation area and ORO staining area was quantitatively compared. Scale bar: 100 μm. n Hepatic TG levels were examined and normalized by protein levels. Bars = means ± SD; $$n = 3$$ to 6; ns, no significance; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$
## Oral administration of tamoxifen effectively improves HFD-induced NAFLD
Considering that tamoxifen is usually delivered orally in clinical practice, we investigated whether oral tamoxifen administration alleviated NAFLD in HFD-induced NAFLD mice. NAFLD mice were administrated with 100 mg/kg tamoxifen or vehicle by oral gavage every other day for 2 weeks (Fig. 4h). Oral tamoxifen administration didn’t change body weight and liver weight (Supplementary Fig. S5a). However, serum ALT, TC, HDL, and LDL levels were decreased and glucose and insulin intolerance were improved notably (Fig. 4l and Supplementary Fig. S5b). Consistent with previous results, oral tamoxifen administration alleviated hepatic steatosis (Fig. 4j, k). These results collectively suggested that either intraperitoneal injection or oral administration of tamoxifen can effectively improve hepatic steatosis and metabolic dysfunction.
## The therapeutic effect of tamoxifen lasts after cessation of treatment
Whether the therapeutic effect of tamoxifen can be sustained after drug withdrawal remains unknown. To address this issue, we administrated 100 mg/kg tamoxifen intraperitoneally every other day for 2 weeks on HFD-induced NAFLD mice. Mice were sacrificed 2 or 4 weeks after the treatment (Fig. 4i). The body weight decreased in a time-dependent manner but liver weight showed no difference among different groups (Supplementary Fig. S6a, b). Intriguingly, serum ALT and AST levels decreased after drug withdrawal however serum TC, TG, HDL, and LDL levels increased (Supplementary Fig. S6c). Likewise, after drug withdrawal, the glucose intolerance was restored gradually and insulin tolerance remained unchanged (Supplementary Fig. S6d). However, drug withdrawal did not increase hepatic steatosis (Fig. 4m, n). The above results indicated that the therapeutic effect of tamoxifen on hepatic steatosis and liver injury could last at least 4 weeks.
## Short-term tamoxifen treatment alleviates HFD-induced hepatic steatosis
To verify whether short-term tamoxifen treatment is also effective, pair-fed male mice were administrated with 100 mg/kg tamoxifen intraperitoneally every other day for 14 days following 20-week HFD feeding (Supplementary Fig. S7a). After tamoxifen administration, the body weight was lost (Supplementary Fig. S7b, c) and the glucose and insulin intolerance were ameliorated in short-term HFD-treated mice (Supplementary Fig. S7d, e). Consistently, hepatic steatosis was alleviated (Supplementary Fig. S7f, h) and serum levels of ALT, AST, TC, HDL-C, and LDL-C were lowered by tamoxifen treatment as well (Supplementary Fig. S7g). To verify the involved mechanism, we performed RT-qPCR and found that tamoxifen didn’t increase mRNA expression of TG export and glucogenic-related genes (Supplementary Fig. S7i). These results collectively demonstrated that short-term tamoxifen administration showed the same effect on the treatment of fatty liver disease.
## The therapeutic effect of tamoxifen on metabolic dysfunction is dose-dependent
To evaluate whether the therapeutic effect of tamoxifen is dose-dependent, we analyzed mice administrated with different doses of tamoxifen (10 mg/kg, 50 mg/kg, and 100 mg/kg) following 20-week HFD feeding. Notably, administration of tamoxifen at different doses did not change liver weight or body weight (Supplementary Fig. S8a) but reduced serum levels of ALT, AST, TC, and HDL-C (Supplementary Fig. S8b) and alleviated hepatic steatosis without difference (Supplementary Fig. S8c–e). To further investigate the therapeutic effect of tamoxifen at lower doses, we analyzed mice administrated with tamoxifen for 1 or 5 mg/kg following 20-week HFD feeding. Lower-dose tamoxifen treatment didn’t change body weight and liver weight (Supplementary Fig. S9a) and did not alter serum ALT, AST, TC, TG, and HDL concentrations and glucose and insulin tolerance (Supplementary Fig. S9b, c). H&E and ORO staining and hepatic TG measurement revealed that lower-dose tamoxifen decreased hepatic steatosis and hepatic TG and 5 mg/kg tamoxifen treatment decreased more TG deposition than that of 1 mg/kg tamoxifen treatment (Supplementary Fig. S9d, e), revealing that tamoxifen treatment could alleviate hepatic steatosis in a dose-dependent manner. The above results suggested that the therapeutic effect of tamoxifen on metabolic dysfunction was dose-dependent and the dose range is lower for steatosis for tamoxifen than for other metabolic parameters.
## Tamoxifen has no sex disparity in fatty liver treatment
As mentioned previously, tamoxifen is a SERM. RNA-sequence of tamoxifen-treated hepatocytes confirmed the inactivation of estrogen-related signaling pathways (Supplementary Fig. S10a). To clarify whether the therapeutic effect of tamoxifen was gender or ER-dependent, we first discovered the influence of sex disparity in tamoxifen treatment. Both male and female mice were administrated with 100 mg/kg tamoxifen as previous description (Figs. 3a and 5a). Except for body weight, no significant change of liver weight and liver to body weight ratio was inspected between male and female mice following tamoxifen treatment (Supplementary Fig. S10b). Tamoxifen treatment reduced serum levels of TC, TG, HDL-C, and LDL-C in male mice and decreased ALT, TC, and HDL-C in female mice (Fig. 5c). Then, we compared the decreased ratio of serum levels of TC, TG, HDL-C and LDL-C between male and female mice. Notably, TC and TG decreased less in tamoxifen-treated female mice than in male ones while the decrease in serum HDL-C and LDL-C showed no difference (Fig. 5d). H&E and ORO staining and hepatic TG measurement revealed that tamoxifen attenuated hepatic steatosis in both male and female mice (Figs. 3h and 5e, f). Moreover, the reduction of fat cavitation showed no difference between male and female mice, however, the reduction of ORO staining areas in female mice is greater than it in male ones (Fig. 5g), which was inconsistent with the findings shown in Fig. 5d, implying sex disparity may probably have no influence on tamoxifen treatment. Fig. 5The therapeutic role of tamoxifen was independent of sex disparity and estrogen receptor. a Dosing scheme of tamoxifen on female C57BL/6 mice fed with MCD diets for 6 weeks. Dose of tamoxifen: 100 mg/kg. b Dosing scheme of tamoxifen combined with ER antagonist-fulvestrant on C57BL/6 mice fed with MCD diets for 6 weeks. Dose of tamoxifen: 100 mg/kg. c *Serum analysis* of ALT, AST, TC, TG, HDL, and LDL in MCD diets-induced male and female mice administrated with tamoxifen or vehicle. d Comparison of the decreased serum TC, TG, HDL, and LDL concentrations caused by tamoxifen between male and female mice. e Liver sections from male or female MCD diets-induced mice administrated with tamoxifen or vehicle were performed H&E and ORO staining. H&E fat cavitation area and ORO staining area were quantitatively compared. Scale bar: 100 μm. f Hepatic TG measurement in female MCD-induced NASH mice administrated with tamoxifen or vehicle. g Comparison of the decreased H&E fat cavitation area and ORO staining area caused by tamoxifen between male and female mice. h *Serum analysis* of ALT, AST, TC, TG, HDL, and LDL in male MCD diets-induced mice administrated with tamoxifen or tamoxifen + fulvestrant. i Protein levels of estrogen receptor in liver tissues from NASH mice administrated with tamoxifen w/o fulvestrant were determined by western blotting and quantitatively compared. j Liver sections from male MCD diets-induced mice administrated with vehicle, fulvestrant, tamoxifen or tamoxifen + fulvestrant were performed H&E and ORO staining. H&E fat cavitation area and ORO staining area were quantitatively compared. Scale bar: 100 μm. k Hepatic TG levels were examined on male MCD diets-induced mice administrated with vehicle, fulvestrant, tamoxifen, or tamoxifen + fulvestrant. l Liver sections from female MCD diets-induced mice administrated with vehicle, fulvestrant, tamoxifen or tamoxifen + fulvestrant were performed H&E and ORO staining. H&E fat cavitation area and ORO staining area were quantitatively compared. Scale bar: 100 μm. m Hepatic TG levels were examined on female MCD diets-induced mice administrated with vehicle, fulvestrant, tamoxifen, or tamoxifen + fulvestrant. n WT and ER knockdown AML12 cells were treated with PA for 36 h and then treated with 40 μM tamoxifen for another 36 h. ORO staining and cellular TG measurement were performed. Scale bar:100 μm. Bars = means ± SD; $$n = 3$$–6; ns, no significance; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$
## Tamoxifen treats hepatic steatosis estrogen receptor independently
To clarify whether tamoxifen ameliorates fatty liver through ER, we employed a pharmacological ER antagonist (fulvestrant) to facilitate ER degradation.28,29 As shown in Fig. 5b, 100 mg/kg fulvestrant was administrated twice during 100 mg/kg tamoxifen treatment. WB results showed that fulvestrant could effectively reduce hepatic ER protein expression (Fig. 5i). Interestingly, fulvestrant did not affect ALT, AST, TC, TG, and HDL-C in tamoxifen-treated mice (Fig. 5h). Additionally, fulvestrant made no change to the therapeutic effect of tamoxifen on hepatic steatosis (Fig. 5j, k). In female mice, fulvestrant didn’t change body weight or liver weight (Supplementary Fig. S10c). Fulvestrant decreased serum TC and HDL levels but didn’t change serum levels of ALT, AST, TG and LDL (Supplementary Fig. S10f). Similarly, fulvestrant administration exerted no influence on the therapeutic effect of tamoxifen on hepatic steatosis in female mice (Fig. 5l, m). To exclude the potential interference of fulvestrant on extrahepatic tissues, we use ER shRNA to suppress the expression of ER on AML12 cell line. RT-qPCR and WB results suggested that ER shRNA decreased cellular ER expression notably (Supplementary Fig. S10d, e). Consistently, the effect of tamoxifen on decreasing TG accumulation was not affected by ER knockdown (Fig. 5n). Above results collectively proved that the treatment of tamoxifen on lipid metabolism is independent of ER.
## Tamoxifen impedes hepatic steatosis by inhibiting JNK/MAPK pathway
In order to elucidate the mechanisms involved in the treatment of tamoxifen on NAFLD, RNA-sequence was performed in isolated hepatocytes of fatty liver. Correlation heatmap and principal component analysis showed clearly separated clusters for the multiple samples (Supplementary Fig. S11a, b). Volcano plot suggested that most differentially expressed genes (DEGs) were downregulated after tamoxifen treatment (Supplementary Fig. S11c). Consistent with RT-qPCR results, gene heatmap revealed that genes associated with lipid metabolism, inflammation, fibrosis and apoptosis were inactivated by tamoxifen (Fig. 6a). In KEGG analysis of DEGs, we found that inflammation, fibrosis, apoptosis and insulin resistance pathways were enriched (Fig. 6b). More importantly, MAPK signaling pathway and its key upstream regulator-Ras signaling, were obviously enriched as well (Fig. 6b). The heatmap further confirmed that most of the genes in MAPK signaling pathway were downregulated by tamoxifen treatment (Fig. 6c). Besides, GSEA analysis showed that MAPK and Ras signaling pathway (Fig. 6d), coupled with other signaling pathways such as chemokine signaling pathway, toll like receptor signaling pathway, TGF-β signaling pathway and P53 signaling pathway were extensively inhibited by tamoxifen treatment (Fig. 6e). Interestingly, GSEA analysis proved that tamoxifen facilitated fatty acid degradation instead of inhibiting fatty acid biosynthesis (Supplementary Fig. S11d). In addition, inflammation and fibrosis-related signaling pathways were found to be downregulated notably in tamoxifen-treated hepatocytes (Supplementary Fig. S11e, f). Thereafter, western blot analyses confirmed the activation of JNK and P38 in hepatocytes of fatty liver and only JNK pathway was inhibited by tamoxifen treatment (Fig. 6f). However, ERK phosphorylation was not affected by tamoxifen administration (Fig. 6f). Besides, we found that tamoxifen also decreased hepatic JNK phosphorylation in mice fed with 8-week normal diet (Fig. 6g). Overall, these data identified the JNK/MAPK pathway as the candidate signaling pathway regulated by tamoxifen in NAFLD pathogenesis. Fig. 6Tamoxifen inhibited JNK/MAPK signaling pathway. RNA was extracted from isolated hepatocytes of male MCD diets-induced NASH mice administrated with tamoxifen or vehicle and RNA-seq was performed. Fold change > 1.5, $P \leq 0.05$ or FDR < 0.25 were considered significant. a Gene expression heatmap of lipid metabolism, inflammation, fibrosis and apoptosis-related genes. b KEGG analysis of DEGs. c Heatmap of MAPK signaling pathway-related DEGs. d GSEA analysis of MAPK and Ras signaling pathways. e GSEA analysis of MAPK signaling pathway and inflammation, fibrosis and apoptosis-related signaling pathways. Fig. a–e were generated on Omicsmart platform (https://www.omicsmart.com/). f Protein levels of three key MAPK molecules-P38, JNK, and ERK were determined by western blotting and quantitatively compared. g Total protein was extracted from liver tissues in normal diets-fed mice administrated with tamoxifen or vehicle for 8 weeks. Protein levels of JNK and p-JNK were determined by western blotting and quantitatively compared. Bars = means ± SD; $$n = 3$$ to 4; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$
## Activation of JNK/MAPK pathway abolishes the effect of tamoxifen on hepatic steatosis treatment
To prove JNK/MAPK pathway is required for the treatment of tamoxifen on NAFLD, we applied a pharmacological JNK activator-anisomycin (ANI).30 Male mice fed with MCD diet were injected with 100 mg/kg tamoxifen and 50 mg/kg ANI for 5 consecutive days (Fig. 7a). Firstly, we proved that anisomycin administration could increase JNK phosphorylation in liver tissues (Fig. 7b). Activation of JNK by ANI restored the decreased body weight and liver weight caused by tamoxifen (Supplementary Fig. S12a). In addition, tamoxifen-triggered reductions of ALT, AST, TC, TG, and HDL-C were all reversed by ANI administration (Fig. 7c). Consistently, ANI partially abolished the therapeutic effect of tamoxifen on hepatic steatosis (Fig. 7d–f). We further proved the effect of ANI in AML12 cell line. ORO staining and TG measurement indicated that ANI increased TG accumulation in the absence or presence of tamoxifen (Fig. 7g, h) by increasing JNK phosphorylation (Fig. 7i, j), suggesting that JNK activation aggravated hepatic steatosis and abrogated the therapeutic effect of tamoxifen in vitro. RT-qPCR analyses suggested that ANI treatment increased the expression of lipogenic genes (Srebp1c, FASN, ACC1), which were downregulated by tamoxifen (Supplementary Fig. S12b). In contrast, CC930, a JNK inhibitor, could effectively decrease JNK phosphorylation and alleviate hepatocyte steatosis (Fig. 7k–n). Taken together, these results confirmed tamoxifen-treated hepatic steatosis by inhibiting JNK/MAPK signaling. Fig. 7Pharmacological activation of JNK/MAPK signaling pathway partly abolished the therapeutic effect of tamoxifen. a Dosing scheme of tamoxifen coupled with JNK activator-anisomycin (ANI) on male C57BL/6 mice fed with MCD diets for 6 weeks. Dose of tamoxifen: 100 mg/kg. Dose of anisomycin: 50 mg/kg. b Total protein was extracted from liver tissues in mice administrated with vehicle, ANI, tamoxifen or tamoxifen + ANI. Protein levels of JNK and p-JNK were determined by western blotting and quantitatively compared. c *Serum analysis* of ALT, AST, TC, TG, and HDL in mice administrated with vehicle, tamoxifen or tamoxifen + ANI. d Liver sections from MCD diets-induced mice administrated with vehicle, tamoxifen, or tamoxifen + ANI were performed H&E and ORO staining. Scale bar:100 μm. e H&E fat cavitation area and ORO staining area were quantitatively compared. f Hepatic TG levels were measured and normalized by protein levels. g AML12 cells were stimulated by 0.3 mM PA and treated with 40 μM tamoxifen. 10 μM anisomycin or DMSO was added 2 h before tamoxifen treatment. ORO staining was performed. Scale bar: 100 μm. h Cellular TG contents were measured and quantitatively compared. i Total proteins were extracted from AML12 cells treated with DMSO/ ANI/ tamoxifen/ tamoxifen+ANI. Protein levels of JNK and p-JNK were determined by western blotting. j The ratio of p-JNK/JNK and p-JNK/GAPDH were quantitatively compared. k AML12 cells were stimulated by 0.3 mM PA and treated with 40 μM tamoxifen. 30 μM CC930 or DMSO were added 2 h before tamoxifen treatment. ORO staining was performed. Scale bar: 100 μm. l Cellular TG contents were measured and quantitatively compared. m Total proteins were extracted from AML12 cells treated with DMSO/tamoxifen/CC930/tamoxifen+CC930. Protein levels of JNK and p-JNK were determined by western blotting. n The ratio of p-JNK/JNK and p-JNK/GAPDH were quantitatively compared. Bars = means ± SD; $$n = 3$$–6; ns, no significance; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$
## Discussion
Tamoxifen, a selective estrogen receptor modulator, was approved by the Food and Drug Administration for the treatment of breast cancer since 1977.31 Despite its success in cancer chemotherapy, the multiple roles of tamoxifen in regulating estrogen receptors should be noticed. In cardiovascular system, tamoxifen acts as an estrogen agonist to lower the incidence of fatal myocardial infarction, but on the other hand, tamoxifen also shows antiestrogenic effects on triggering coronary artery constriction.32 In liver, tamoxifen exhibits hepatoprotective roles as an estrogen agonist in acute and chronic liver injuries,33,34 while impedes cholangiocyte proliferation and accelerates their apoptosis as an estrogen antagonist. This paradoxical phenomenon may be attributed to cholangiocytes expressing both ERα and ERβ subtypes while hepatocytes only express ERα.35 In recent years, tamoxifen non-estrogen receptor-mediated molecular targets have been gradually paid much more attention. Tamoxifen inhibits PKC activity to restrict tumor growth and promote cancer cell apoptosis.36 A few PKC isoforms have been proven to be activated in NASH, so tamoxifen may also alleviate NASH through PKC inhibition and PKC is also a potent upstream regulator of MAPK pathway.37 In addition, tamoxifen was proven to exert its effectiveness through some other targets such as androgen receptor,38 estrogen-related receptor gamma39, and mitogen-activated protein kinase 8.36 These findings implicate that the pharmacological effects of tamoxifen are so complicated depending on tissue specificity and dose effect and its therapeutic roles in the treatment of different diseases being currently investigated preclinically.40,41 Occasionally, during the induction of CreER recombinase activity by tamoxifen, we found that tamoxifen diminished serum lipid contents and inhibited hepatic lipid accumulation simultaneously in mice. This phenomenon arouses our great interest. We conjectured if tamoxifen could treat NAFLD or NASH in mice. To confirm our hypothesis, we first evaluated the effect of tamoxifen on lipotoxicity. As expected, tamoxifen protected hepatocytes against sodium palmitate-induced lipotoxicity in all three cell lines and primary hepatocytes without increasing hepatotoxicity. Afterwards, we administrated tamoxifen continuously for 8 weeks in male and female mice fed with normal diets. We observed the role of tamoxifen in decelerating lipid accumulation in blood and in liver of normal mice and found that tamoxifen decreased serum cholesterol contents and inhibited hepatic lipid accumulation. Importantly, continuous tamoxifen administration did not worsen liver function at all. To testify the therapeutic effect of tamoxifen on NAFLD or NASH, we established MCD, CDAA, and HFD diets-induced metabolic disordered models. Consistently, tamoxifen decreased hepatic TG and serum TC contents and the mRNA expressions of lipogenesis, inflammation, and fibrosis-related genes in all three models. In addition, tamoxifen alleviated insulin intolerance in HFD-induced models, implicating that tamoxifen effectively ameliorated NAFLD. Besides, we examined extrahepatic adipose tissues and found that tamoxifen decreased adipocyte area but did not alter insulin secretion (Supplementary Fig. S13a, b). This finding indicated that tamoxifen may alleviate NAFLD through enhancing lipolysis and increasing insulin sensitivity instead of insulin secretion.
In view that tamoxifen is an oral medicine in clinics, we proved that oral tamoxifen had the same therapeutic effect compared to intraperitoneal injection, suggesting that different routes of tamoxifen administration would not affect its therapeutic effect. Although we proved that tamoxifen ameliorated NAFLD obviously, whether the therapeutic effect could sustain after treatment cessation remained controversial. After drug withdrawal, the therapeutic effect of tamoxifen on hepatic steatosis was sustained for 4 weeks, however, the influence on other indicators such as serum TC, TG, HDL, and LDL levels and glucose and insulin tolerance gradually disappeared. These findings give us hints that intermittent tamoxifen administration can be adopted to reach lasted and satisfied effects in NAFLD treatment.
Tamoxifen was proven to be an appetite suppressor in preclinical animal models and in clinical patients.42 To avoid this undesired effect, mice administrated with tamoxifen were strictly pair-fed. In view that loss of body weight contributes to the improvement of NAFLD,43 the changes in body weight, liver weight, and liver-to-body weight ratio were consistently monitored while tamoxifen was administrated. In normal, MCD and CDAA-treated mice, tamoxifen improved fatty liver without affecting body weight, proving weight loss was not the reason for the decreased lipid accumulation in tamoxifen-treated mice. It is worth noting that, tamoxifen is a canonical ER modulator. However, no difference was found in mice of different genders following tamoxifen treatment. Besides, blocking ER didn’t alter the therapeutic effect of tamoxifen as well, proving tamoxifen improved NAFLD through non-estrogen targets. We previously reported the importance of TAK1-mediated MAPK signaling pathway in regulating NASH.44 *In this* study, we confirmed that JNK/MAPK was required for tamoxifen to treat fatty liver disease, as tamoxifen inactivated JNK/MAPK signaling and JNK activator abolished the therapeutic effect of tamoxifen. As for how tamoxifen affects JNK signaling, in vitro experiments revealed that the phosphorylation of canonical upstream regulators, TAK1 and ASK1, was inhibited by tamoxifen (data not shown). Combined our results with published manuscripts, we speculate that tamoxifen may suppress JNK activation through TAK1 or ASK1 dephosphorylation which needs to be further determined. Intriguingly, tamoxifen decreased expressions of lipid uptake, de novo lipogenesis, fatty acid oxidation, and lipid export-related genes simultaneously, suggesting that tamoxifen could suppress hepatic lipid metabolism comprehensively but whether tamoxifen alleviated NAFLD through lipid metabolism gene expression modulation directly needs further investigation.
Intriguingly, our findings seem inconsistent with a few preclinical studies. In the 1990s, long-term tamoxifen treatment was found to induce NAFLD/NASH in female patients with breast cancer.10,12 In recent years, a few publications revealed that low-dose tamoxifen administration for 5 consecutive days increased hepatic lipid accumulation by enhancing fatty acid synthesis.15 Conversely, some findings demonstrated that tamoxifen protected hepatocytes against lipotoxicity and steatosis.20,21 Lelliott et al. also reported that tamoxifen inhibited fatty acid synthesis in the presence of hepatic steatosis.45 It seems that the impact of tamoxifen on lipid metabolism is still elusive. In addition, Ceasrine et al. reported that tamoxifen had significant and sustained effects on glucose tolerance which was in line with our in vivo results.46 The reason why our results are contrary to some of the findings mentioned above, is probably that we administrated tamoxifen in a short term (1–4 weeks) and at a higher dose (100 mg/kg). However, whether high-dose tamoxifen administration is safe and its potential side effects need to be determined. In our study, at least serum ALT and AST analyses showed that short-term high-dose tamoxifen treatment did not cause hepatoxicity in mice. In order to evaluate the safety and side effects of high-dose tamoxifen administration in clinics, we tried to translate mice dose to human dose through normalization to body surface area (BSA).47 After calculation, the human equivalent dose (HED) of tamoxifen usage we estimated is 8 mg/kg (300 mg/m2). Skapek et al. performed a phase II study on children with desmoid fibromatosis and proved that tamoxifen administration at a maximum dose of 300 mg per day caused few serious side effects.48 In addition, high-dose tamoxifen is also used to treat malignant gliomas (240 mg/d)49 and cryptococcal meningitis (300 mg/d).50 Trump et al. conducted a phase I clinical trial to investigate the regulatory role of high-dose, oral tamoxifen in P-glycoprotein-mediated drug resistance and found that tamoxifen given at a dose of 300 mg/m2 per day for 12 days following a loading dose of 400 mg/m2 did not cause dose-limiting toxicity.51 In other studies, breast cancer patients with renal cell carcinoma were well tolerated with tamoxifen administrated at a dose of 200 mg/m2 /day for up to 1 year.52,53 Even if considering the safety of the use of high-dose tamoxifen, we proved that low-dose tamoxifen (i.e., 1, 5 mg/kg), which was probably insufficient to restore the glucose intolerance and liver malfunction, was still effective in treating hepatic steatosis in NAFLD mice. Therefore, short-term tamoxifen administration seems to be realistic to treat NAFLD in clinics.
## Cell culture
Cell lines were obtained from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). Huh7 and HepG2 cell lines were grown in RPMI1640 or high-glucose DMEM medium (Gibco, USA) supplemented with $10\%$ fetal bovine serum (FBS) (Gibco, USA) and 100 U/ml penicillin and 0.1 mg/ml streptomycin (MI00614, Mishushengwu, Xi’an, China). AML12 cell line was grown in DMEM: F12 medium (11330, Invitrogen, USA) supplemented with $10\%$ FBS, $1\%$ ITS liquid media supplement (I3146, Sigma, USA), and 40 ng/ml dexamethasone. All cell lines were kept in a humidified incubator at 37 °C and $5\%$ CO2. Cells were used from third to tenth passage in each experiment. Primary hepatocytes were isolated from 8-week-old male C57BL/6 mice and were cultured in HM medium (ScienCell, USA) and seeded in six-well plates at 1 × 106/well. To induce cellular lipotoxicity, 0.3 mM sodium palmitate (SYSJ-KJ, Kunchuang Biotechnology, Xi’an, China) was added into medium and vehicle was added as a control. After 36 h, tamoxifen dissolved in DMSO was added at a dose of 10, 20, 40 μM and kept for 36 h and DMSO was added as a control. To induce JNK activation in vitro, anisomycin (10 μM) was added 2 h before tamoxifen treatment. To inhibit JNK phosphorylation in vitro, 30 μM Tanzisertib (cc-930) (S8490, Selleck, China) was added 2 h before tamoxifen treatment. For cellular Oil Red O (ORO) staining, cells were cultured in six-well plates and after sodium palmitate and tamoxifen treatment, culture medium was removed, and cells were fixed with $4\%$ paraformaldehyde for 30 min and washed with PBS three times. Then the cells were treated with $60\%$ isopropanol for 5 min. Remove isopropanol, stain cells with ORO working solution (Servicebio Technology, Wuhan, China) for 10 min, and washed cells with PBS. Then the cells were stained with hematoxylin for 3–5 min and washed with PBS at least three times. Then we observed and took photos using an inverted phase contrast microscope (Olympus, X71, Japan). For cellular TC and TG tests, we purchased commercial kits from Pulilai Gene Technology Co., Ltd (Beijing, China) and followed the manufacturer’s instructions.
## Lentivirus transfection
Lentiviral particles bearing ER shRNA (gcTTTCTTTAAGAGAAGCATT) were purchased from GeneChem (Shanghai, China). AML12 cells were seeded in six-well plate at the concentration of 3 × 104/ml. After 12 h, cells were transfected with ER shRNA viruses or control shRNA viruses. When the transfection efficiency reached 70–$80\%$, 10 μg/ml puromycin was added to kill uninfected cells. Then, very passage was selected with 5 μg/ml puromycin until transfection efficiency reached nearly $100\%$.
## Mice
Male and female C57BL/6 mice (8 weeks old) were purchased from Weitong Lihua Experimental Animal Technology Co., Ltd (Beijing, China) and kept in specific pathogen-free animal house. The room temperature was controlled at around 23 °C and humidity was kept at 50–$60\%$ with a 12 h day/night cycle. The water was sufficient to obtain and in order to eliminate the potential appetite suppression role of tamoxifen, the mice with tamoxifen administration and their control mice were pair-fed unless specifically stated. In order to establish NASH mouse model, male or female mice were fed with a MCD diet (A02082002BR, Research Diets, New Brunswick, USA) for 6 weeks or a CDAA diet (A06071309, Research Diets, New Brunswick, USA) for 10 weeks. To establish a NAFLD mouse model, male or female mice were fed with a HFD (D12492, Research Diets, New Brunswick, USA) for 20 weeks. Tamoxifen was purchased from Sigma (T5648, USA), dissolved in corn oil, and stored at 4 °C for at most a week. Pharmacological ER antagonist fulvestrant (S1191) and JNK activator anisomycin (S7409) were purchased from Selleck (Shanghai, China). The dosing scheme was shown in figures and illustrated in the Results section.
All animal experiments were approved by the Animal Experiment Administration Committee of the Fourth Military Medical University (Xi’an, China) and proceeded under the instruction of Guide for the Care and Use of Laboratory Animals published by the National Institute of Health (publications 86-23, revised 1985).
## GTT and ITT
For GTT experiment, mice were fasted for 8 h and measured fasting blood glucose on tail vein using OneTouch Ultra glucometers (LifeScan). Then the mice were injected with 1 g/kg glucose intraperitoneally and blood glucose was recorded at 15, 30, 60, and 120 min. For ITT experiment, mice fasted for 6 h and measured fasting blood glucose. Then the mice were injected with 0.75 U/kg insulin subcutaneously and blood glucose was recorded at 15, 30, 60, 120 min.
## Histology
For H&E, PSR staining, and F$\frac{4}{80}$ IHC staining, fresh liver tissues or adipose tissues were fixed in $4\%$ paraformaldehyde for at least 24 h and embedded with paraffin. Then the liver tissues were cut into 4–6 μm sections. H&E and PSR staining were conducted following the standard protocols. F$\frac{4}{80}$ IHC staining was performed as previously illustrated.54 For F$\frac{4}{80}$ IF and ORO staining, mice liver was fixed in $4\%$ paraformaldehyde for 4 h and then transferred into $30\%$ sucrose solution at 4 °C overnight. Then liver tissues were embedded with optimum cutting temperature compound (Sakura Finetek, Japan) and sectioned at 8–10 μm. For F$\frac{4}{80}$ IF, sections were embedded with primary antibody at 4 °C overnight, washed with PBS and embedded with fluorescent secondary antibody at room temperature, and counterstained with Hoechst 33258 (Sigma-Aldrich). For ORO staining, saturated Oil Red O solution was purchased from Servicebio technology and cryosections were stained following the manufacturer’s instructions.
## Serum biochemistry
Serum biochemical analysis was performed using commercial kits (Nanjing Jiancheng Biochemical, Nanjing, China) on an automatic biochemical analyzer (Chemray 240, Rayto, China) based on the instructions supplied by the manufacturer. Serum Insulin concentrations were examined using commercial ELISA kits (Jianglaibio, Shanghai, China).
## Cell isolation
Primary hepatocytes were isolated as previously described.54
## Cell viability
AML12 and Huh7 cell viability were examined using commercial kit (C0011, Beyotime, China).
## Gene expression profiling
Total RNA was extracted using Trizol Reagent from isolated primary hepatocytes of mice fed with MCD diets and administrated with tamoxifen or vehicle. Then RNA samples were sent to GENE DENOVO Biotechnology Co. (Guangzhou, China) for further sequencing. Briefly, eukaryotic mRNA was enriched by Oligo(dT) beads and fragmented into short fragments using fragmentation buffer and then reverse transcribed into cDNA. CDNA of about 200 bp was selected with AMPure XP beads, amplified through PCR, and purified with AMPure XP beads to construct cDNA library. RNA quality was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA) and checked using RNase free agarose gel. The cDNA products were size selected by agarose gel electrophoresis, PCR amplified, and sequenced using Illumina Novaseq6000. Bioinformatic analysis was conducted on Omicsmart platform (https://www.omicsmart.com/). The raw sequences were deposited on the GEO website (GSE212148).
## RT-qPCR
We extracted total RNA from isolated hepatocytes or liver tissues using Trizol reagent (Invitrogen, USA) and followed the manufacturer’s protocols. Then the RNA was reverse-transcribed into cDNA using Evo M-MLV RT Premix for qPCR (Accurate Biology, Changsha, China). RT-qPCR was performed by using SYBR Green PCR Master Mix (Accurate Biology, Changsha, China) according to the manufacturer’s protocol. *Relative* gene expression levels were normalized to β-actin. The primers used in the current article were listed in Supplementary Table S1.
## Western blotting
Total protein was extracted with RIPA lysis buffer supplemented with 10 mM phenylmethanesulfonyl fluoride and quantified with BCA protein quantitative kit (Thermo Fisher Scientific, Rockford, IL). Protein samples were loaded and separated with SDS-PAGE gel electrophoresis and transferred onto polyvinylidene fluoride membranes. The membranes were blocked with $5\%$ skimmed milk powder and incubated in primary antibodies at 4 °C overnight and washed with TBST. Then the membranes were incubated in secondary HRP-conjugated antibodies at room temperature for 2 h and washed with TBST. The protein signals were detected by ChemiDoc MP Imaging System (Bio-Rad, Hercules, CA, USA). The antibodies used in this article were listed in Supplementary Table S2.
## Statistics
Statistical analysis of data was performed using GraphPad Prism v8.3.0 and the results were expressed as the means ± SD. Data were tested for normality (Kolmogorov–Smirnov test) and homogeneity (Brown-Forsythe test) of variance. Difference between the two groups was analyzed by Student’s t test with normal distribution or by Mann-Whitney U test with skewed distribution. Difference among multiple groups was analyzed by one-way ANOVA with normal distribution followed by Tukey’s post hoc test or by Kruskal-Wallis test with skewed distribution. P values less than 0.05 were considered statistically significant. * $p \leq 0.05$; **$p \leq 0.01$; ***$p \leq 0.001$; ****$p \leq 0.0001$; ns, not significant.
## Supplementary information
suppl data The online version contains supplementary material available at 10.1038/s41392-022-01299-y.
## References
1. Younossi Z. **Global burden of NAFLD and NASH: trends, predictions, risk factors and prevention**. *Nat. Rev. Gastroenterol. Hepatol.* (2018) **15** 11-20. DOI: 10.1038/nrgastro.2017.109
2. Sheka AC. **Nonalcoholic steatohepatitis: a review**. *JAMA* (2020) **323** 1175-1183. DOI: 10.1001/jama.2020.2298
3. Ioannou GN. **Epidemiology and risk-stratification of NAFLD-associated HCC**. *J. Hepatol.* (2021) **S0168-8278** 02007-02009
4. Ferguson D, Finck BN. **Emerging therapeutic approaches for the treatment of NAFLD and type 2 diabetes mellitus**. *Nat. Rev. Endocrinol.* (2021) **17** 484-495. DOI: 10.1038/s41574-021-00507-z
5. Chalasani N. **The diagnosis and management of nonalcoholic fatty liver disease: practice guidance from the American Association for the Study of Liver Diseases**. *Hepatology* (2018) **67** 328-357. DOI: 10.1002/hep.29367
6. Younossi ZM. **Non-alcoholic fatty liver disease—a global public health perspective**. *J. Hepatol.* (2019) **70** 531-544. DOI: 10.1016/j.jhep.2018.10.033
7. Patel HK, Bihani T. **Selective estrogen receptor modulators (SERMs) and selective estrogen receptor degraders (SERDs) in cancer treatment**. *Pharmacol. Ther.* (2018) **186** 1-24. DOI: 10.1016/j.pharmthera.2017.12.012
8. Howell A. **Tamoxifen versus the newer SERMs: what is the evidence?**. *Ann. Oncol.* (2000) **11** 255-265. DOI: 10.1093/annonc/11.suppl_3.255
9. Jaiyesimi IA, Buzdar AU, Decker DA, Hortobagyi GN. **Use of tamoxifen for breast cancer: twenty-eight years later**. *J. Clin. Oncol.* (1995) **13** 513-529. DOI: 10.1200/JCO.1995.13.2.513
10. Pratt DS, Knox TA, Erban J. **Tamoxifen-induced steatohepatitis**. *Ann. Intern. Med.* (1995) **123** 236. DOI: 10.7326/0003-4819-123-3-199508010-00018
11. Cai Q, Bensen M, Greene R, Kirchner J. **Tamoxifen-induced transient multifocal hepatic fatty infiltration**. *Am. J. Gastroenterol.* (2000) **95** 277-279. DOI: 10.1111/j.1572-0241.2000.01708.x
12. Ogawa Y, Murata Y, Nishioka A, Inomata T, Yoshida S. **Tamoxifen-induced fatty liver in patients with breast cancer**. *Lancet* (1998) **351** 725. DOI: 10.1016/S0140-6736(05)78493-2
13. van Hoof M, Rahier J, Horsmans Y. **Tamoxifen-induced steatohepatitis**. *Ann. Intern. Med.* (1996) **124** 855-856. DOI: 10.7326/0003-4819-124-9-199605010-00015
14. Bruno S. **Incidence and risk factors for non-alcoholic steatohepatitis: prospective study of 5408 women enrolled in Italian tamoxifen chemoprevention trial**. *BMJ* (2005) **330** 932. DOI: 10.1136/bmj.38391.663287.E0
15. Cole LK, Jacobs RL, Vance DE. **Tamoxifen induces triacylglycerol accumulation in the mouse liver by activation of fatty acid synthesis**. *Hepatology* (2010) **52** 1258-1265. DOI: 10.1002/hep.23813
16. Gudbrandsen OA, Rost TH, Berge RK. **Causes and prevention of tamoxifen-induced accumulation of triacylglycerol in rat liver**. *J. Lipid Res.* (2006) **47** 2223-2232. DOI: 10.1194/jlr.M600148-JLR200
17. Lee MH. **Gene expression profiling of murine hepatic steatosis induced by tamoxifen**. *Toxicol. Lett.* (2010) **199** 416-424. DOI: 10.1016/j.toxlet.2010.10.008
18. Wang X. **Hepatic estrogen receptor α improves hepatosteatosis through upregulation of small heterodimer partner**. *J. Hepatol.* (2015) **63** 183-190. DOI: 10.1016/j.jhep.2015.02.029
19. Zhao F. **The effect and mechanism of tamoxifen-induced hepatocyte steatosis in vitro**. *Int. J. Mol. Sci.* (2014) **15** 4019-4030. DOI: 10.3390/ijms15034019
20. Guillaume M. **Selective liver estrogen receptor α modulation prevents steatosis, diabetes, and obesity through the anorectic growth differentiation factor 15 hepatokine in mice**. *Hepatol. Commun.* (2019) **3** 908-924. DOI: 10.1002/hep4.1363
21. Miyashita T. **Hepatoprotective effect of tamoxifen on steatosis and non-alcoholic steatohepatitis in mouse models**. *J. Toxicol. Sci.* (2012) **37** 931-942. DOI: 10.2131/jts.37.931
22. Brocard J. **Spatio-temporally controlled site-specific somatic mutagenesis in the mouse**. *Proc. Natl Acad. Sci. USA* (1997) **94** 14559-14563. DOI: 10.1073/pnas.94.26.14559
23. Feil R. **Ligand-activated site-specific recombination in mice**. *Proc. Natl Acad. Sci. USA* (1996) **93** 10887-10890. DOI: 10.1073/pnas.93.20.10887
24. Wang F. **Myelin degeneration and diminished myelin renewal contribute to age-related deficits in memory**. *Nat. Neurosci.* (2020) **23** 481-486. DOI: 10.1038/s41593-020-0588-8
25. OʼShea TM. **Foreign body responses in mouse central nervous system mimic natural wound responses and alter biomaterial functions**. *Nat. Commun.* (2020) **11** 6203. DOI: 10.1038/s41467-020-19906-3
26. Brash JT. **Tamoxifen-activated CreERT impairs retinal angiogenesis independently of gene deletion**. *Circ. Res.* (2020) **127** 849-850. DOI: 10.1161/CIRCRESAHA.120.317025
27. Zhong ZA. **Optimizing tamoxifen-inducible Cre/loxp system to reduce tamoxifen effect on bone turnover in long bones of young mice**. *Bone* (2015) **81** 614-619. DOI: 10.1016/j.bone.2015.07.034
28. Guan J. **Therapeutic ligands antagonize estrogen receptor function by impairing its mobility**. *Cell* (2019) **178** 949-963. DOI: 10.1016/j.cell.2019.06.026
29. Wardell SE. **Pharmacokinetic and pharmacodynamic analysis of fulvestrant in preclinical models of breast cancer to assess the importance of its estrogen receptor-α degrader activity in antitumor efficacy**. *Breast Cancer Res. Treat.* (2020) **179** 67-77. DOI: 10.1007/s10549-019-05454-y
30. Lv L. **Interplay between α2-chimaerin and Rac1 activity determines dynamic maintenance of long-term memory**. *Nat. Commun.* (2019) **10** 5313. DOI: 10.1038/s41467-019-13236-9
31. Kent OC. **Tamoxifen in the treatment of breast cancer**. *N. Engl. J. Med.* (1998) **339** 1609-1618. DOI: 10.1056/NEJM199811263392207
32. Mikkola T. **Estrogen replacement therapy, atherosclerosis, and vascular function**. *Cardiovasc. Res.* (2002) **53** 605-619. DOI: 10.1016/S0008-6363(01)00466-7
33. Yoshikawa Y. **Mechanisms of the hepatoprotective effects of tamoxifen against drug-induced and chemical-induced acute liver injuries**. *Toxicol. Appl. Pharmacol.* (2012) **264** 42-50. DOI: 10.1016/j.taap.2012.06.023
34. Guillaume M. **Selective activation of estrogen receptor α activation function-1 is sufficient to prevent obesity, steatosis, and insulin resistance in mouse**. *Am. J. Pathol.* (2017) **187** 1273-1287. DOI: 10.1016/j.ajpath.2017.02.013
35. Alvaro D. **Estrogens stimulate proliferation of intrahepatic biliary epithelium in rats**. *Gastroenterology* (2000) **119** 1681-1691. DOI: 10.1053/gast.2000.20184
36. Radin DP, Patel P. **Delineating the molecular mechanisms of tamoxifen’s oncolytic actions in estrogen receptor-negative cancers**. *Eur. J. Pharmacol.* (2016) **781** 173-180. DOI: 10.1016/j.ejphar.2016.04.017
37. Dallak MA. **Acylated ghrelin induces but deacylated ghrelin prevents hepatic steatosis and insulin resistance in lean rats: Effects on DAG/ PKC/JNK pathway**. *Biomed. Pharmacother.* (2018) **105** 299-311. DOI: 10.1016/j.biopha.2018.05.098
38. Yamasaki K. **Comparison of the Hershberger assay and androgen receptor binding assay of twelve chemicals**. *Toxicology* (2004) **195** 177-186. DOI: 10.1016/j.tox.2003.09.012
39. Gowda K, Marks BD, Zielinski TK, Ozers MS. **Development of a coactivator displacement assay for the orphan receptor estrogen-related receptor-gamma using time-resolved fluorescence resonance energy transfer**. *Anal. Biochem.* (2006) **357** 105-115. DOI: 10.1016/j.ab.2006.06.029
40. Beh CY. **Enhanced anti-mammary gland cancer activities of tamoxifen-loaded erythropoietin-coated drug delivery system**. *PLoS One* (2019) **14** e0219285. DOI: 10.1371/journal.pone.0219285
41. Al-Jubori AA. **Layer-by-layer nanoparticles of tamoxifen and resveratrol for dual drug delivery system and potential triple-negative breast cancer treatment**. *Pharmaceutics* (2021) **13** 1098. DOI: 10.3390/pharmaceutics13071098
42. Xu B, Lovre D, Mauvais-Jarvis F. **Effect of selective estrogen receptor modulators on metabolic homeostasis**. *Biochimie* (2016) **124** 92-97. DOI: 10.1016/j.biochi.2015.06.018
43. Younossi ZM, Corey KE, Lim JK. **AGA clinical practice update on lifestyle modification using diet and exercise to achieve weight loss in the management of nonalcoholic fatty liver disease: expert review**. *Gastroenterology* (2021) **160** 912-918. DOI: 10.1053/j.gastro.2020.11.051
44. Wang L. **Tripartite motif 16 ameliorates nonalcoholic steatohepatitis by promoting the degradation of phospho-TAK1**. *Cell Metab.* (2021) **33** 1372-1388. DOI: 10.1016/j.cmet.2021.05.019
45. Lelliott CJ. **Transcript and metabolite analysis of the effects of tamoxifen in rat liver reveals inhibition of fatty acid synthesis in the presence of hepatic steatosis**. *FASEB J.* (2005) **19** 1108-1119. DOI: 10.1096/fj.04-3196com
46. Ceasrine AM. **Tamoxifen improves glucose tolerance in a delivery-, sex-, and strain-dependent manner in mice**. *Endocrinology* (2019) **160** 782-790. DOI: 10.1210/en.2018-00985
47. Reagan-Shaw S, Nihal M, Ahmad N. **Dose translation from animal to human studies revisited**. *FASEB J.* (2008) **22** 659-661. DOI: 10.1096/fj.07-9574LSF
48. Skapek SX. **Safety and efficacy of high-dose tamoxifen and sulindac for desmoid tumor in children: results of a Children’s Oncology Group (COG) phase II study**. *Pediatr. Blood Cancer* (2013) **60** 1108-1112. DOI: 10.1002/pbc.24457
49. Odia Y, Kreisl TN, Aregawi D, Innis EK, Fine HA. **A phase II trial of tamoxifen and bortezomib in patients with recurrent malignant gliomas**. *J. Neurooncol.* (2015) **125** 191-195. DOI: 10.1007/s11060-015-1894-y
50. Ngan NTT. **An open label randomized controlled trial of tamoxifen combined with amphotericin B and fluconazole for cryptococcal meningitis**. *Elife* (2021) **10** e68929. DOI: 10.7554/eLife.68929
51. Trump DL. **High-dose oral tamoxifen, a potential multidrug-resistance-reversal agent: phase I trial in combination with vinblastine**. *J. Natl Cancer Inst.* (1992) **84** 1811-1816. DOI: 10.1093/jnci/84.23.1811
52. Tormey DC, Lippman ME, Edwards BK, Cassidy JG. **Evaluation of tamoxifen doses with and without fluoxymesterone in advanced breast cancer**. *Ann. Intern. Med.* (1983) **98** 139-144. DOI: 10.7326/0003-4819-98-2-139
53. Papac RJ, Keohane MF. **Hormonal therapy for metastatic renal cell carcinoma combined androgen and provera followed by high dose tamoxifen**. *Eur. J. Cancer* (1993) **29A** 997-999. DOI: 10.1016/S0959-8049(05)80209-6
54. Duan JL. **Endothelial Notch activation reshapes the angiocrine of sinusoidal endothelia to aggravate liver fibrosis and blunt regeneration in mice**. *Hepatology* (2018) **68** 677-690. DOI: 10.1002/hep.29834
|
---
title: Alternative microbial-based functional ingredient source for lycopene, beta-carotene,
and polyunsaturated fatty acids
authors:
- Chewapat Saejung
- Khomsorn Lomthaisong
- Prawphan Kotthale
journal: Heliyon
year: 2023
pmcid: PMC9981927
doi: 10.1016/j.heliyon.2023.e13828
license: CC BY 4.0
---
# Alternative microbial-based functional ingredient source for lycopene, beta-carotene, and polyunsaturated fatty acids
## Abstract
The acquisition of carotenoids and polyunsaturated fatty acids (PUFAs) from plants and animals for use as functional ingredients raises concerns regarding productivity and cost; utilization of microorganisms as alternative sources is an option. We proposed to evaluate the production of carotenoids and PUFAs by *Rhodopseudomonas faecalis* PA2 using different vegetable oils (rice bran oil, palm oil, coconut oil, and soybean oil) as carbon source, different concentrations of yeast extract as nitrogen source at different cultivation time to ensure the best production. Cultivation with soybean oil as source of carbon led to the most significant changes in the fatty acid profile. Compared to the initial condition, the strain cultivated in the optimal conditions ($4\%$ soybean oil, $0.35\%$ yeast extract, and 14 days of incubation) showed an increase in μmax, biomass, carotenoid productivity, and microbial lipids by $102.5\%$, $52.7\%$, $33.82\%$, and $34.78\%$, respectively. The unsaturated fatty acids content was raised with additional types of PUFAs; omega-3 [alpha-linolenic acid and eicosapentaenoic acid] and omega-6 [linoleic acid and eicosatrienoic acid] fatty acids were identified. The results of ultra high-performance liquid chromatography-electrospray ionization-quadrupole time of flight-mass spectrometry (UHPLC-ESI-QTOF-MS/MS) indicated the molecular formula and mass of bacterial metabolites were identical to those of lycopene and beta-carotene. The untargeted metabolomics revealed functional lipids and several physiologically bioactive compounds. The outcome provides scientific reference regarding carotenoids, PUFAs, and useful metabolites that have not yet been reported in the species *Rhodopseudomonas faecalis* for further use as a microbial-based functional ingredient.
## Highlights
•Additional types of PUFAs are found in the proposed condition.•Beta-carotene and EPA are first reported in Rhodopseudomonas faecalis.•Functional lipids and physiologically bioactive compounds were observed.•R. faecalis PA2 might be used as a new microbial-based functional ingredient.•The optimal conditions enhance μmax, biomass, carotenoids, and microbial lipids.
## Introduction
Increasing concern for health and well-being has increased the inclusion of functional ingredients in dietary supplements, nutraceuticals, and health products. Carotenoids and polyunsaturated fatty acids (PUFAs) are examples of high-value compounds supplemented in many products. Carotenoids, natural biomolecules produced by plants, algae, and some bacteria, have been shown to have provitamin A activities and strong antioxidant potential, allowing them to fight cancer, age-related macular degeneration, photooxidative damage, and boost immunological response [1]. Carotenoids have received interest in the nutraceuticals and food industries, with a market worth of around 1.21 billion USD [2,3]. Functional lipids are compounds involved in a broad spectrum of metabolic conditions. Long-chain PUFAs in n-3 and n-6 series (omega-3 and omega-6 fatty acids) have been discovered as the essential fatty acids in mammals because of their specific biofunctions as precursors for eicosanoids that modulate pulmonary function [4], structural components of membranes, and inflammatory responses [5]. There is plenty of data to show that PUFAs can help avoid a variety of chronic diseases [6]. Alpha-linolenic acid (18:3, n-3; ALA), eicosapentaenoic acid (20:5, n-3; EPA), docosahexaenoic acid (22:6, n-3; DHA), linoleic acid (18:2, n-6; LA), and arachidonic acid (20:4, n-6) are all examples of essential fatty acids. Carotenoids and PUFAs cannot be biosynthesized by human body, making them the crucial functional ingredients in several products.
There are many carotenoid-based products on the market, as well as dietary supplements with PUFAs [7,8]. Carotenoids and PUFAs derived from plants and animals raise concerns not only their productivity but also production cost. PUFAs are found in high-price food such as chia seeds, fish oils, and marine fish in the families Scombridae, Clupeidae, and Salmonidae [9]. Plant carotenoids require agricultural land, pesticides, growing season, and harvesting time. Therefore, carotenoids and PUFAs acquire from these sources as the functional ingredients are expensive. At present, the use of microbial biomass as a functional ingredient is a viable option to provide key nutrients at a cheaper cost with a higher yield [10]. As a result, microorganisms are increasingly used in functional food, functional ingredients, and nutraceuticals businesses. Carotenoids and PUFAs have been acquired from a variety of bacteria, fungi, and microalgae for use as active ingredients in industries. Beta-carotene from Sphingomonas sp. and canthaxanthin from Paracoccus carotinifaciens are examples of bacterial carotenoids used in food colorants [3] while lycopene from *Rhodospirillum rubrum* and spheroidenone from Rhodobacter sphaeroides have been shown to have anti-cancer and anti-inflammatory properties in health products [11]. Nutritional supplements containing beta-carotene and astaxanthin derived from the algae *Dunaliella salina* and *Haematococcus pluvialis* are also reported [12]. Mortierella alpina, *Mortierella alliacea* [6], and *Rhodotorula mucilaginosa* are among the fungal producers of omega-3 and omega-6 fatty acids [13]. A heterotrophic unicellular marine thraustochytrid Aurantiochytrium sp. [ 14]. and a microalga *Crypthecodinium cohnii* [15] have been used as sources of DHA and squalene. Several significant metabolites from microorganisms are currently investigated to explore their application and utilization as functional ingredients.
The anoxygenic photosynthetic bacteria are excellent producers of carotenoids and PUFAs. Because of their several modes of metabolism, these bacteria have been widely used in waste treatment. They are not pathogens but they do contain several types of useful compounds such as coenzyme Q10, 5-aminolevulinic acid, carotenoids, bacteriochlorin, and polyhydroxyalkanoates [16]. They have membrane lipids and phospholipids which are not typically found in general bacteria, such as phosphatidylcholine, sulfoquinovosyldiacylglycerol, betaine lipids, and ornithine lipids [17]. Although their prominent characteristics have been reported to be used as single-cell protein (SCP) and feed [18], they have not yet received the attention they deserve, and the information about utilization of these bacteria as functional ingredients is scarce. Lycogen™, a carotenoid product acquired from Rhodobacter sphaeroides WL-APD911, is the only product obtained from anoxygenic photosynthetic bacteria utilized in mammals that shows anti-inflammatory, anti-oxidative, and glucose homeostasis effects [11,19].
The anoxygenic photosynthetic bacterium *Rhodopseudomonas faecalis* PA2 contains several nutrients [20] and high protein content containing all essential amino acids [21] although it was cultivated on waste substrates. Aquatic animals fed R. faecalis PA2 showed superior performances and survival in comparison with animals fed the alga Chlorella vulgaris, the yeast Saccharomyces cerevisiae, the cyanobacterium Spirulina sp., and the other species of anoxygenic photosynthetic bacteria [22,23]. This indicates it could be a potential candidate for application in functional ingredient industries and the investigation of additional useful metabolites of this strain has drawn attention.
To produce the microbial-based functional ingredients, carbon source for microbial growth is crucial and the expense of carbon source has to be factored in. Organic acids, such as malic acid and succinic acid, are the essential carbon for anoxygenic photosynthetic bacteria but they are the expensive feedstock. On the other hand, vegetable oils are much cheaper; the catabolism of oil components produces organic acids as intermediates that can be used for the growth of anoxygenic photosynthetic bacteria [24]. Hence, the objectives of this study were to identify lycopene, beta-carotene, and PUFAs in the anoxygenic photosynthetic bacterium R. faecalis PA2 in the presence of vegetable oils and to evaluate the metabolite profiling of this strain. In this study, a Liquid Chromatography-Mass Spectrometry (LC-MS)-based metabolomic approach was used to investigate the metabolic composition, aiming to reveal the interesting metabolites in this strain. To the best of our knowledge, this is the first study that used metabolomics to quantify the useful metabolites and to observe the metabolites which have not been reported in the anoxygenic photosynthetic bacteria.
## Effects of vegetable oils as carbon sources on biomass, carotenoids, microbial lipids, and fatty acid composition
The photosynthetic bacterium R. faecalis PA2 was employed which is safely deposited at Thailand Bioresource Research Center (TBRC 5694) for research and commercial purposes. The cultivation of this strain and inoculum preparation were carried out in glutamate-malate medium and exposed to light intensity at 4000 lux under anoxygenic conditions [25]. The basal medium (BM) supplemented with $1\%$ vegetable oil as a carbon source (rice bran oil, palm oil, coconut oil, or soybean oil) was used as the tested medium and adjusted pH to 6.8; inoculum volume was $10\%$. Incubation was carried out at 30 ± 2 °C under light-anoxygenic conditions. The experiments were conducted in six replicates. Biomass, carotenoid, and microbial lipid concentrations were investigated at intervals of 48 h. Bacterial cells were separated from the culture broth by centrifugation at 6000 rpm 4 °C for 10 min at the end of the experiment (Himac CR20B2, Hitachi, Tokyo, Japan). The supernatant was discarded; the cell pellets were washed with $0.85\%$ sterile NaCl and then freeze-dried using a freeze dryer (Freezone 2.5 L; LABCONCO, KC, USA). The fatty acid composition of the freeze-dried biomass was determined following AOAC [26] method 996.06. Briefly, the Shimadzu Nexis GC-2030 equipped with split injector port, flame ionization detector (FID), and AOC-20i + s autosampler was used. The fatty acid methyl ester (FAME) mix was analyzed according to the AOAC method 996.06 which required the use of helium carrier gas (constant linear velocity 18 cm/s). The column was Rt-2560 100 m × 0.25 mm ID × 0.20 μm film thickness. The GC parameters included inlet (1 μL split injection; 225 °C; split ratio 200:1) and flame ionization detector (285 °C; H2 32 mL/min; air 200 mL/min; make-up (N2) 24 mL/min). The oven temperature was 100 °C (4 min hold); 3 °C/min to 240 °C (15 min hold). The FAME mix was purchased from Restek (PA, USA).
All the tested vegetable oils could be used as sole carbon‐based nutrients. This phenomenon is supported by Fig. 1a-d, which depicts the growth of R. faecalis PA2 and the generation of some metabolites in the presence of vegetable oils. The use of coconut oil as a carbon source resulted in the lowest maximum specific growth rate (μmax) (0.082 ± 0.005/day), carotenoid concentration (452.58 ± 8.56 mg/L), and microbial lipid concentration (134.28 ± 7.66 mg/L). Soybean oil, on the other hand, showed the highest values. The biomass of R. faecalis PA2 fed coconut oil had the highest saturated fatty acid content (Table 1). However, cultivation with soybean oil showed the predominant fatty acid composition because the biomass of R. faecalis PA2 contained both omega-3 fatty acid (ALA) and omega-6 fatty acids (LA and eicosatrienoic acid [or dihomo-gamma-linolenic acid (DGLA]). Therefore, soybean oil was employed in the following experiments due to the composition of unsaturated fatty acids. Fig. 1Growth and microbial substances of *Rhodopseudomonas faecalis* PA2 cultivated in basal medium containing different vegetable oils as carbon sources. ( a) μmax, (b) biomass concentration, (c) carotenoid concentration, and (d) microbial lipid concentration. Different superscript letters in each bar indicate significant differences among treatments (p ≤ 0.05).∗ denotes that the value was significant difference. Fig. 1Table 1Total fat and fatty acid composition of *Rhodopseudomonas faecalis* PA2 cultivated in basal medium containing different vegetable oils as carbon sources. Table 1Fatty acidsContent (g/100 g)Rice bran oilPalm oilCoconut oilSoybean oilSaturated fatty acids7.491a6.742a8.295a7.098aMyristic acid (14:0)–a0.249 ± 0.12a0.985 ± 0.22b0.212 ± 0.01aPalmitic acid (16:0)6.384 ± 1.22a5.467 ± 0.85a4.882 ± 1.82a5.363 ± 0.78aHeptadecanoic acid (17:0)-a0.075 ± 0.01b0.053 ± 0.02b0.102 ± 0.22bStearic acid (18:0)1.107 ± 0.11a0.951 ± 0.41a0.918 ± 0.44a1.421 ± 0.75aCaprylic acid (8:0)-a-a0.121 ± 0.03b-aCapric acid (10:0)-a-a0.134 ± 0.04b-aLauric acid (12:0)-a-a1.202 ± 0b-aUnsaturated fatty acids5.619a4.288a2.186b4.963acis-10-Heptadecenoic acid (17:1, n-10)-a-a0.048 ± 0b-aPalmitoleic acid (16:1, n-7)1.004 ± 1.00a0.583 ± 0.21b0.732 ± 0.51ab0.726 ± 0.41abcis-9-Oleic acid (18:1, n-9)2.893 ± 1.75a2.612 ± 1.22a0.453 ± 0.05b2.123 ± 0.71abcis-9,12-Linoleic acid (18:2, n-6)1.722 ± 0.71a0.518 ± 0.45b0.113 ± 0.45b1.440 ± 0.15abalpha-Linolenic acid (18:3, n-3)-a-a-a0.081 ± 0.05bcis-8,11,14-Eicosatrienoic acid (20:3, n-6)-a0.575 ± 0.15b0.840 ± 0.11b0.593 ± 0.21bTotal fat13.11a11.03a10.48a12.06a
## Optimization of cultural condition
Since carbon content, nitrogen (yeast extract) content, and incubation period play the significant roles in boosting bacterial growth and essential metabolites, these three parameters were investigated. The optimization was carried out by one-variable at a time. The optimal vegetable oil was used as a carbon source in BM. The vegetable oil content ($1\%$, $2\%$, $4\%$, $6\%$, $8\%$, and $10\%$ (w/v)) was supplemented in BM and adjusted pH to 6.8. The $10\%$ inoculum was included. The experiment was set for 10 days at 30 ± 2 °C under light-anoxygenic conditions. The biomass, carotenoid, and microbial lipid concentrations were investigated at intervals of 48 h. Six duplicates of each experiment were carried out. For the optimization of yeast extract content, the contents of $0.05\%$, $0.10\%$, $0.15\%$, $0.20\%$, $0.25\%$, $0.30\%$, $0.35\%$, $0.40\%$, $0.80\%$, and $1.60\%$ were optimized. The incubation period of 6, 8, 10, 12, and 14 days were investigated. The incubation conditions were carried out as stated.
## Bacterial cultivation under the optimal conditions and determination of fat and fatty acid composition
Rhodopseudomonasfaecalis PA2 was grown in a photo-bioreactor with $10\%$ inoculum under optimal conditions. Nitrogen gas was flushed into the reactor to create an anoxygenic condition. Illumination (4000 lux) was provided throughout the experiment. Bacterial cells were freeze-dried and used to determine total fat and fatty acid composition by the hydrolytic extraction gas chromatographic technique.
## Cell extraction for metabolite measurement
The wet cells were used, and 50 mg of the sample (five replicates) was dissolved with 1 mL reconstitution buffer (water: acetonitrile = 1:1). The mixture was sonicated for 15 min three times (Ultrasonic Cleaner GT SONIC-D2, GT SONIC, Shenzhen, China) and centrifuged at 15 000 rpm at 4 °C for 15 min twice (D3024R High Speed Refrigerated Micro Centrifuge, DLAB, DLAB Scientific, Beijing, China). The supernatant was transferred to the high-performance liquid chromatography (HPLC) glass vial for LC-MS data acquisition.
## Determination of carotenoids and untargeted profiling of metabolites using Ultra High-Performance Liquid Chromatography - Electrospray Ionization - Quadrupole Time of Flight - Mass Spectrometry (UHPLC-ESI-QTOF-MS/MS) analysis
Standard lycopene and beta-carotene were used for quantification of the detected carotenoids using the targeted metabolite analysis. The extracted samples were analyzed on reverse-phase liquid chromatography. The separation was performed using UHPLC-ESI-QTOF-MS/MS (Bruker Daltonics, Bremen, Germany). Bruker intensity solo HPLC C18 2.1 × 100 mm, 2 μm column was used (Bruker Daltonics, Bremen, Germany). The column temperature and autosampler temperature were maintained at 40 °C and 4 °C, respectively. The mobile phase consisted of eluent A ($100\%$ water and $0.1\%$ formic acid (FA)) and eluent B ($100\%$ acetonitrile and $0.1\%$ FA). The flow rate was 0.35 mL/min; the gradient elution was set as follow: 99-$5\%$ A (0.0–4.0 min, 0.25 mL/min), $5\%$ A (4.0–8.1 min, 0.25 mL/min), and $99\%$ A (8.1–11.0 min, 0.25 mL/min). Injection volume was 2 μL applied for positive ionization polarity mode. The mass spectrometry was performed using the broadband collision‐induced dissociation (bbCID) method by a compact ESI-Q-TOF system (Bruker Daltonics, Bremen, Germany). Sodium formate solution (2 mM sodium hydroxide, $0.1\%$ FA, $50\%$ isopropanol) was injected as an external calibrant with a flow rate of 0.5 μL/min. The condition in positive ionization polarity mode consisted of 50–1300 m/z mass range, 35 V cone voltage, 4000 V capillary voltage, 220 °C source temperature, 220 °C desolvation temperature, and 8 L/min desolvation gas flow.
For the untargeted metabolite profiling analysis, the flow rate was adjusted to 0.35 mL/min; the gradient elution was set as follows: $99\%$ A (0.0–2.0 min, 0.25 mL/min), $1\%$ A (2.0–20.0 min, 0.25 mL/min), $99\%$ A (20.1–28.3 min, 0.35 mL/min), and $99\%$ A (28.5–30.0 min, 0.25 mL/min). Injection volume was 2 μL applied for positive and negative ionization polarity modes. The conditions in positive ionization polarity mode are 50–1300 m/z mass range, 35 V cone voltage, 4000 V capillary voltage, 220 °C source temperature, 220 °C desolvation temperature, and 8 L/min desolvation gas flow. The conditions in negative ionization polarity mode are 50–1300 m/z mass range, 31 V cone voltage, 4500 V capillary voltage, 220 °C source temperature, 220 °C desolvation temperature, and 8 L/min desolvation gas flow.
## Metabolite identification and annotation
The data was imported to the MetaboScape software for metabolite identification. The assessment of metabolites was compared with the public database: METLIN, Human Metabolome Database (HMDB), Bruker MetaboBASE, and LipidBlast database. Level of assignment (LoA) of the metabolites include 1) accurate mass matched to the database, 2) accurate mass matched to database and tandem MS spectrum matched to in silico fragmentation pattern, 3) tandem MS spectrum matched to database or literature, 4) retention time and the molecular mass matched to standard compound, and 5) MS/MS spectrum matched standard compound.
## Analytical procedures
Bacterial biomass and carotenoids were analyzed according to Saejung and Chanthakhot [25]. Carotenoids were extracted by immersing the cell pellets in methanol-acetone (2:3 v/v) solution overnight until the colorless cells were obtained. The pigment extract was read at 480 and 770 nm using a Genesys 20 spectrophotometer (Thermo Scientific, Waltham, MA, USA). Microbial lipids were extracted from cells by centrifugation of the culture broth at 9000 rpm 4 °C for 15 min. The pellets were resuspended in distilled water after being washed twice with $0.9\%$ NaCl. The pellets were boiled for 10 min in 1 N NaOH, and the cell debris was discarded [27]. The supernatant was used to examine microbial lipids by saponification with 1.5 M KOH in $80\%$ ethanol following Kwon and Rhee [28].
## Statistical analyses
The data were presented as mean ± standard deviation (SD). The significant differences between means were calculated by one-way analysis of variance (ANOVA). The Duncan's multiple range test was used to compare the means at a significance level of p ≤ 0.05. The software IBM SPSS Statistics 28.0.0.0 (IBM Corp., Armonk, NY, USA) was used for statistical analyses.
## Optimization of soybean oil contents
As shown in Fig. 2a, as the content of soybean oil increased, μmax increased, with the maximum level of 0.64 ± 0.01/day, occurring at $4\%$, above which the μmax decreased. Fig. 2b-d shows that supplementing with $4\%$ soybean oil resulted in the highest biomass, carotenoid, and microbial lipid concentrations; however, there was no statistical difference of microbial lipids among the treatments. Table 2 indicates the μmax at $4\%$ soybean oil was enhanced by $60\%$ when compared to the initial condition ($1\%$ soybean oil).Fig. 2Growth and microbial substances of *Rhodopseudomonas faecalis* PA2 cultivated in basal medium containing different soybean oil contents. ( a) μmax, (b) biomass concentration, (c) carotenoid concentration, and (d) microbial lipid concentration. Different superscript letters in each bar indicate significant differences among treatments (p ≤ 0.05). ∗ denotes that the value was significant difference. Fig. 2Table 2Comparison of μmax, biomass concentration, carotenoid productivity, and microbial lipid concentration of *Rhodopseudomonas faecalis* PA2 grown in each optimization study and initial condition. Table 2Culture conditionμmax (/day)Biomass concentration (g/L)Carotenoid productivity (mg/L/day)Microbial lipid concentration (mg/L)Initial conditiona0.40 ± 0.011.48 ± 0.11587 ± 11.12207.01 ± 17.05Optimal soybean oil content ($4\%$ soybean oil)0.64 ± 0.511.90 ± 0.44604.61 ± 25.22272.68 ± $1.59\%$ increase compared to initial condition+60+28.38+3+31.72Optimal yeast extract content ($0.35\%$ yeast extract)0.94 ± 0.232.57 ± 1.01678.52 ± 24.02296.46 ± $33.10\%$ increase compared to initial condition+135+73.65+15.59+43.21Optimal incubation time (14 days)0.81 ± 0.522.26 ± 0.99785.55 ± 14.78279 ± $20.65\%$ increase compared to initial condition+102.50+52.70+33.82+34.78a$1\%$ soybean oil, $0.2\%$ yeast extract, and 10 days of incubation.
## Optimization of yeast extract content
Yeast extract is the most effective nitrogen source for R. faecalis PA2 [29], thus, the optimal content should be investigated. The μmax and biomass concentration varied depending on yeast extract content (Fig. 3a and b). The highest μmax and biomass concentration were found at $0.35\%$ yeast extract, with the increase by $135\%$ and $73.65\%$, respectively (Table 2). Carotenoid synthesis is reduced at lower C/N ratios ($0.80\%$–$1.60\%$ yeast extract) (Fig. 3c). The concentration of microbial lipids was dramatically reduced when yeast extract content was greater than $0.35\%$ because of the excessive nitrogen level (Fig. 3d).Fig. 3Growth and microbial substances of *Rhodopseudomonas faecalis* PA2 cultivated in basal medium containing different yeast extract contents. ( a) μmax, (b) biomass concentration, (c) carotenoid concentration, and (d) microbial lipid concentration. Different superscript letters in each bar indicate significant differences among treatments (p ≤ 0.05). ∗ denotes that the value was significant difference. Fig. 3
## Optimization of the incubation period
The μmax, biomass concentration, carotenoids, and microbial lipid production increased with increasing incubation period, as indicated in Fig. 4a-d, with the maximum at 14 days. In comparison to the initial condition, carotenoids were increase by $33.82\%$ (Table 2). A long incubation period boosted carotenoid production because they are produced during stationary phase. An increase in carotenoids was found after 12 days of incubation (Fig. 4C).Fig. 4Growth and microbial substances of *Rhodopseudomonas faecalis* PA2 cultivated in different incubation period. ( a) μmax, (b) biomass concentration, (c) carotenoid concentration, and (d) microbial lipid concentration. Different superscript letters in each bar indicate significant differences among treatments (p ≤ 0.05).Fig. 4
## Determination of fatty acid composition
The kinetic characteristics of R. faecalis PA2 acquired from a photo-bioreactor are presented in Table S1. Table 3 indicates the total fat and fatty acid composition of R. faecalis PA2 cultivated in the optimal conditions compared with the initial condition. The proposed conditions ($4\%$ soybean oil, $0.35\%$ yeast extract, and 14 days of incubation) could enhance the content of unsaturated fatty acid significantly compared with the initial condition. The unsaturated fatty acids cis-10-pentadecenoic acid (15:1, n-10), cis-10-heptadecenoic acid (17:1, n-10), and EPA were present in the optimal conditions, whereas they were not found in the initial condition (Table 3). This study revealed that ALA, EPA, LA, and DGLA were found in the biomass of R. faecalis PA2 grown in the proposed conditions. To our knowledge, EPA was first reported in this species. Therefore, using soybean oil as carbon source along with the conditions described in this study might provide the important PUFAs in R. faecalis PA2.Table 3Total fat and fatty acid composition of *Rhodopseudomonas faecalis* PA2 cultivated in the optimal conditions compared with the initial condition. Table 3Fatty acidsContent (g/100 g)Optimal conditionaInitial conditionbTotal fat12.25a12.06aSaturated fatty acids3.215a7.098bLauric acid (12:0)0.140 ± 0.15a-bMyristic acid (14:0)0.212 ± 0.01a0.212 ± 0.02aPentadecanoic acid (15:0)0.025 ± 0a-abPalmitic acid (16:0)2.294 ± 1.00a5.363 ± 0.24bHeptadecanoic acid (17:0)0.067 ± 0.02a0.102 ± 0.01aStearic acid (18:0)0.477 ± 0.01a1.421 ± 0bArachidic acid (20:0)––Unsaturated fatty acids9.034a4.963bcis-10-Pentadecenoic acid (15:1, n-10)0.020 ± 0a-abPalmitoleic acid (16:1, n-7)0.407 ± 0.05a0.726 ± 0.71acis-10-Heptadecenoic acid (17:1, n-10)0.053 ± 0.01a- abcis-9-Oleic acid (18:1, n-9)7.491 ± 1.11a2.123 ± 0.07bcis-9,12-Linoleic acid (18:2, n-6)c0.084 ± 0.04a1.440 ± 0.23aalpha-Linolenic acid (18:3, n-3)c0.042 ± 0a0.081 ± 0.07acis-8,11,14-Eicosatrienoic acid (20:3, n-6)c0.231 ± 0.01a0.593 ± 0.12bcis-5,8,11,14,17-Eicosapentaenoic acid (20:5, n-3)c0.706 ± 0.13a-bTotal monounsaturated fatty acids7.971 ± 1.11a2.849 ± 0.58bTotal polyunsaturated fatty acids1.063 ± 0.12a2.114 ± 0.49aUnsaturated: Saturated fatty acids ratio2.81a0.70bOmega-6: Omega-3 ratio0.42:125.1:1Different superscript letters in each bar indicate significant differences among treatments (p ≤ 0.05).aCultivation in $4\%$ soybean oil, $0.35\%$ yeast extract, and 14 days of incubation.bCultivation in $1\%$ soybean oil, $0.2\%$ yeast extract, and 10 days of incubation.cPolyunsaturated fatty acids (PUFAs).
## Determination of carotenoids and the untargeted profiling of metabolites using UHPLC-ESI-QTOF-MS/MS
Lycopene and beta-carotene are dietary carotenoids found in fruits and vegetables. They play a vital role in providing health benefits due to their anti-oxidant properties [30]. Therefore, these two carotenoids were quantified using UHPLC-ESI-QTOF-MS/MS-based targeted metabolomics. Fig. 5a and b shows the extracted ion chromatograms (EIC) of the standard carotenoids compared with the samples. The molecular formula and mass of the samples were identical to that of the standard lycopene and beta-carotene. The adduct ions were [M+H]+ which indicated that the additional molecule was the proton. The molecular formula, exact mass (excluding the mass of adduct ion), and the concentration of each carotenoid in bacterial cells are summarized in Table 4. Although the mass of lycopene and beta-carotene was identical, the time that they were eluted was different which was used to identify the carotenoid type. The molecular formula and exact mass of the detected samples showed that this strain contained lycopene and beta-carotene. As far as we know, beta-carotene has not been reported in the species Rhodopseudomonas faecalis. To the best of our knowledge, this is the first study to report beta-carotene and lycopene in R. faecalis verified by the targeted metabolomic analysis. Fig. 5The extracted ion chromatograms (EIC) of metabolites in *Rhodopseudomonas faecalis* PA2 analyzed by using Ultra High-Performance Liquid Chromatography-Electrospray Ionization-Quadrupole Time of Flight-Mass Spectrometry. ( a) Lycopene and (b) beta-carotene. Fig. 5Table 4Summary of the commercial carotenoids of *Rhodopseudomonas faecalis* PA2 cultivated in the optimal conditions. Table 4CarotenoidsMolecular formulaMass (Dalton)Concentration (μg/mL)LycopeneC40H56536.4385.55 ± 0.25Beta-caroteneC40H56536.4387.18 ± 0.77 The untargeted metabolite profile chromatogram of the samples (five replicates) in positive ionization mode is shown in Fig. 6. The five samples showed identical peaks and there were several metabolites found in this train in the untargeted mode. Each metabolite was identified by comparing with the public database. The identified metabolites are presented in Table 5; the relative concentration of all metabolites was quantified with the external calibrants. Table 6 summarizes the useful metabolites acting as functional ingredients or physiologically bioactive compounds and their benefits. The results also showed several types of phospholipids found in egg yolk, meat, and nuts including lysophosphatidylcholine (LPC), phosphatidylethanolamine (PE), and phosphatidylcholine (PC or lecithin) (Table 6). A microbially associated metabolite, desaminotyrosine, has been found. Buddledin A, (E)-2-octenyl butyrate, and piperonyl acetate were also detected in the cells of R. faecalis PA2 cultivated in soybean oil under the optimal conditions. Fig. 6The base peak chromatograms (BPC) of the untargeted metabolite profiling of *Rhodopseudomonas faecalis* PA2 analyzed by using Ultra High-Performance Liquid Chromatography-Electrospray Ionization-Quadrupole Time of Flight-Mass Spectrometry. Fig. 6Table 5The identified metabolites of *Rhodopseudomonas faecalis* PA2 cultivated in the optimal conditions. Table 5NameMolecular formulaMolecular mass (Da)Retention time (min)Level of AssignmentaRelative concentration (mM)(2 R)-2-Hydroxy-2-methylbutanenitrileC5H9NO99.068653.3320.05338(E)-2-Octenyl butyrateC12H22O2198.162816.9920.1738411-Dehydro-15alpha-hydroxytestololactoneC19H24O4316.165713.0120.3884912-NaphthylamineC10H9N143.07376.8721.0677813-amino-2-naphthoic acidC11H9NO2187.06396.4520.273214-HydroxyprolineC5H9NO3131.05851.3232.492853Acetyl phosphateC2H5O5P139.98781.8920.112714Adenosine diphosphateC10H15N5O10P2427.032.1930.061917ADP-riboseC15H23N5O14P2559.07231.9930.029767Alpha, Beta-trehaloseC12H22O11342.11681.3330.22585Buddledin AC17H24O3276.172712.9120.486148D-Glucose 6-phosphateC6H13O9P260.02991.3730.109871D-Glycerate 2-phosphateC3H7O7P185.99341.8520.221452DesaminotyrosineC9H10O3166.063714.431.773245HomospermidineC8H21N3159.17371.0121.388442l-LeucineC6H13NO2131.09491.5430.553782L-TryptophanC11H12N2O2204.08956.4530.067738Leu Pro Ile IleC23H42N4O5454.3137920.003141LPC 16:1C24H48NO7P493.316411.6530.037486LPC 16:2C24H46NO7P491.303117.1130.137033LPC 18:1C26H52NO7P521.347312.2230.098474LPE 14:0C19H40NO7P425.255314.7630.220672LPE 16:0C21H44NO7P453.287116.6330.181552LPE 16:0C21H44NO7P453.286917.4131.834661LPE 16:1C21H42NO7P451.270914.8230.64706LPE 16:1C21H42NO7P451.270815.231.041493LPE 16:1C21H42NO7P451.270815.330.634426LPE 17:1C22H44NO7P465.287416.2530.289735LPE 18:1C23H46NO7P479.302517.0331.065491LPE 18:1C23H46NO7P479.302117.9833.751037LPE 19:2 lyzoC24H46NO7P491.302117.8630.69955N,N-DimethylanilineC8H11N121.08936.230.026896N-Isopropyl-p-toluamideC11H15NO177.11531.5520.081274N-Isopropyl-p-toluamideC11H15NO177.11545.7120.028937Nα-Acetyl-l-glutamineC7H12N2O4188.08031.4330.110847p-TolualdehydeC8H8O120.05791.5221.335947PC 16:0eC24H50NO7P495.333415.9330.086001PC 16:1eC24H48NO7P493.318417.2330.454758PC 18:1eC26H52NO7P521.348112.6230.35371PC 19:2eC27H52NO7P533.347812.530.025663PC 30:1C38H74NO8P703.516316.1730.1563PC 32:2C40H76NO8P729.532716.1730.32134PC 34:2C42H80NO8P757.561916.1830.040566PE 19:2C24H44NO8P505.281615.4130.684912PE 19:3C24H42NO8P503.265812.4331.213597Phosphoric acidH3O4P97.977061.621.079814Piperonyl acetateC10H10O4194.05598.7920.04889PutrescineC4H12N288.100210.9930.094636TryptamineC10H12N2160.10016.930.140921XanthanC13H10O182.073713.6221.31766Leu Pro Ile Ile: short-chain amino acid of Leucine-Proline-Isoleucine-Isoleucine. LPC: Lysophosphatidylcholine. LPE: Lysophosphatidylethanolamine. PC: Phosphatidylcholine. PE: Phosphatidylethanolamine.aMetabolite assignment five levels of assignment (LoA): 2 = accurate mass matched to database and tandem MS spectrum matched to in silico fragmentation pattern; 3 = tandem MS spectrum matched to database or literature. Table 6Summary of the useful metabolites of *Rhodopseudomonas faecalis* PA2 cultivated in the optimal conditions. Table 6MetaboliteApplicationsReferenceLycopeneNutrient supplement used as antioxidant, anti-cancer, and anti-inflammatory properties.[11]Beta-caroteneNutrient supplement for adult and infant foods which is used as vitamin A precursor.[55]Alpha-Linolenic acid (18:3, n-3) or ALAOmega-3 polyunsaturated fatty acid (PUFA) used as nutraceutical against metabolic diseases, inflammatory diseases, and cardiovascular diseases.[56]cis-5,8,11,14,17-Eicosapentaenoic acid (20:5, n-3) or EPAOmega-3 PUFA used to lowering plasma triglyceride, non-high-density lipoprotein cholesterol (non-HDL-C) levels, and other key lipid/lipoprotein parameters, as well as a broad range of anti-inflammation.[57]cis-9,12-Linoleic acid (18:2, n-6) or LAOmega-6 PUFA used as supplement to lowering the risk of cardiovascular disease and premature death as well as the active ingredient in moisturizer against skin disorders.[58,59]cis-8,11,14-Eicosatrienoic acid (20:3, n-6) or dihomo-gamma-linolenic acid (DGLA)Omega-6 PUFA used as a precursor for biosynthesis of biologically active eicosanoids and other metabolites, which has anti-inflammatory, anti-thrombotic, anti-hypertensive, anti-allergic, and anti-proliferation activities.[[60], [61], [62]]Phosphatidylcholine and Phosphatidylethanolamine (PE)Improvement of EPA and DHA levels in brain that can enhance the treatment of depression and neuroinflammatory diseases such as Alzheimer's disease. Reducing atherosclerosis by decreasing plasma very low-density lipoprotein-cholesterol (VLDL-C) and increasing plasma high-density lipoprotein-cholesterol (HDL-C)[63,64]DesaminotyrosineProtection of influenza virus infection through modification of type I interferon signaling and diminution of lung immunology and acting as an anti-inflammatory molecule that contribute to maintain intestinal and systematic immune homeostasis.[65,66]Buddledin AAntifungal action against Trichophyton rubrum, Tricophyton interdigitale, and Epidermophyton floccosum[67]Nα-Acetyl-l-glutaminePrevention of gut damage induced by protein energy malnutrition.[68](E)-2-Octenyl butyrateFatty alcohol ester used as flavoring ingredient.[69]Piperonyl acetateSweet, bitter, and floral tasting compound used as flavoring agent.[70]
## Discussion
Many studies have been conducted to explore the alternative organisms for the production of carotenoids and PUFAs replacing the production from plants and animals. The use of microorganisms to produce these compounds has increased significantly; yet, few investigations have been undertaken to uncover additional carotenoids and PUFAs producers. In this study, a strategy that used soybean oil as feedstock to produce functional ingredients from beneficial bacterium was established. The strain was able to use vegetable oils by digesting them into glycerol and fatty acids. The glycerol is transformed into dihydroxyacetone phosphate, one of the glycolysis intermediates, and receives energy in the form of ATP through the metabolic process [24]. The fatty acids are metabolized via beta-oxidation, which produces either acetyl Co-A or succinyl Co-A depending on the type of fatty acids. Acetyl Co-A is required in the TCA cycle's transition reaction to combine with oxaloacetic acid while succinyl Co-A is one of the intermediates in the TCA cycle, thus, the strain can generate energy by using fatty acids as a carbon source [31]. Rice bran oil, palm oil, coconut oil, and soybean oil have $25\%$, $49.3\%$, $83\%$, and $15\%$ saturated fatty acids, respectively [32,33]. Coconut oil contains the highest content of saturated fatty acids. The saturated fatty acids have higher melting points than unsaturated fatty acids, resulting in requiring more metabolic energy to break down [34]. Photosynthetic bacteria prefer organic acids for growth while coconut oil contains only trace amounts of free fatty acids, thereby influencing the degradation by this strain. Moreover, coconut oil was found to be resistant to microbial degradation in other study [35]. Soybean oil contains $81\%$ unsaturated fatty acids [32], which resulted in facilitating the catabolism by bacteria.
Since multiple intermediates are involved in carotenoid biosynthesis, the important precursor to manufacture these intermediates is acetyl-CoA [36]. As a result, acetyl Co-A acquired from the breakdown of vegetable oils can be employed as a precursor for carotenoid synthesis (Fig. 1c). Moreover, acetyl-CoA is used as a precursor to producing malonyl Co-A via carboxylation and then transformed to acetyl-ACP by transacylase for use in lipid biosynthesis. The produced lipids are then transported and stored in bacterial cells [37], hence using vegetable oils as carbon sources aided the buildup of microbial lipids in bacteria. This was in line with prior research reporting the supplementation of phototrophic microorganisms with carbon precursors could increase lipid accumulation [38].
In this study, soybean oil was the only vegetable oil that could boost ALA in the tested strain (Table 1). This was because soybean oil is categorized as an alpha-linolenic acid oil, which contains a significant amount of ALA [39]. ALA, EPA, and DHA are the three important omega-3 fatty acids; DHA and EPA are found in fish and seafood. ALA, on the other hand, can be transformed into EPA and ultimately to DHA [40].
As shown in Fig. 2, μmax, carotenoids, and microbial lipid were increased at a certain concentration of soybean oil. This was likely because the high content of carbon source increased the carbon skeleton for the biosynthetic pathway, leading to enhance microbial growth and metabolites [41]. Moreover, the carbon to nitrogen (C/N) ratio of the medium is involved because the greater C/N ratios increase lipid and carotenoid synthesis [42]. Excess carbon supply, on the other hand, caused a decrease in growth rate due to substrate inhibition [43].
The C/N ratio was inversely proportional to the amount of yeast extract present (Fig. 3). When compared to the experiment supplemented with $0.35\%$ yeast extract, the concentrations of yeast extract ranging from $0.05\%$ to $0.30\%$ had a greater C/N ratio. The differences in microbial biomass and respiration reflected these differences. The experiment fed a little amount of yeast extract resulting in the deficiency of nitrogen for biosynthetic pathways at the same amount of carbon. Despite the availability of carbon, the anabolic process comes to a halt. Bacterial growth with a higher C/N ratio is confronted with a surplus of C to N, whereas growth with a lower C/N ratio is confronted with a lack of C to N [44]. As shown in Fig. 4, the longer incubation period resulted in higher biomass, which led to more lipids and carotenoids in bacterial cells. Under the optimal conditions ($4\%$ soybean oil, $0.35\%$ yeast extract, and 14 days of incubation), the lipid productivity was 13.86 mg/L/day (Table S1), whereas lipid productivity of Chlamydomonas reinhardtii, Chlorella sorokiniana, and *Scenedesmus obtusus* XJ-15 were 10.9 mg/L/day, 0.502 mg/L/day, and 0.607 mg/L/day, respectively [[45], [46], [47]]. Previous research reported that carotenoid productivity of Dunaliella tertiolecta, *Chlorella vulgaris* UTEX 265, and Scenedesmus sp. were 0.86 mg/L/day, 11.98 mg/L/day, and 19.70 mg/L/day, respectively [[48], [49], [50]] while carotenoid productivity of R. faecalis PA2 was 45.37 mg/L/day (Table S1). It can be concluded that the lipid productivity and carotenoid productivity of this strain were comparable with the other photosynthetic microorganisms.
The results of this study also prove that changes in the medium composition produce quantitative alteration in fatty acids and carotenoids of anoxygenic photosynthetic bacteria. A previous study showed that the fatty acid composition of bacteria is regulated by the medium composition as well as the age of the cells [51]. This study also calculated the relationships between the ratio of unsaturated to saturated fatty acids (UFA:SFA ratio) because it can be used to evaluate fat utilization. Fat utilization increased with the increase in UFA:SFA ratio; reaching a maximum at UFA:SFA ratio of 4 [52]. Obviously, the strain grown in the optimal condition provided a greater UFA:SFA ratio compared with the initial condition (Table 3). Moreover, previous work also reported the advantage of high UFA:SFA ratio in animal diets in improving meat quality [53]. Omega-6 (LA) can be converted to omega-3 (ALA); thus, the enzymes involved in the metabolism of omega-3 and omega-6 fatty acids are shared and they regulate each other. The balance of omega-6/omega-3 fatty acids in the diet is vital for human nutritional needs. Excessive amount of omega-6 or high omega-6 to omega-3 ratio can cause pathogenesis of diseases [54]. The proportions of omega-6 and omega-3 in the diet can predict the biochemical efficiency, approaching the ratio of 2:1 or 1:1 omega-6/omega-3 fatty acids are the ideal for health. As shown in Table 3, the omega-6/omega-3 fatty acids ratio of R. faecalis PA2 cultured in the optimal condition was close to the targeted ratio. R. faecalis PA2 contained several metabolites found in foods originating from plants and animals. The detection of functional lipids ALA, EPA, LA, and DGLA in biomass has drawn attention because of their physiological and structural roles in biological systems as shown in Table 6. These metabolites are recognized as high-value compounds supplemented in dietary supplements [56,62], suggesting that the biomass of R. faecalis PA2 can be utilized as an alternative source for MUFAs and PUFAs.
The results of UHPLC-ESI-QTOF-MS/MS analysis ensured that the strain and the proposed conditions produced beta-carotene and lycopene. The additional metabolites were detected in cells. The untargeted metabolomics analysis revealed the other functional lipids such as phosphatidylcholine which is a multifunctional phospholipid required for the incorporation of cholesterol in membranes [63]. Our results also showed the presence of desaminotyrosine in R. faecalis PA2. According to previous study, this metabolite can protect against influenza virus [65] and maintain systematic immune homeostasis [66]. Nα-acetyl-l-glutamine can be supplemented in sports nutrition's products to help boost exercise endurance and prevent the negative effect of protein energy malnutrition [68]. Buddledin A showed an antifungal effect [67]. ( E)-2-octenyl butyrate is used as a flavoring ingredient, whereas piperonyl acetate is found in the green vegetables [69,70]. In our perspective, R. faecalis PA2 cultured under the aforementioned conditions could be an alternative source for microbial-based functional ingredients. Although R. faecalis PA2 could provide several beneficial metabolites and it is a promising source for alternative microbial-based functional ingredient, further investigation in vivo is required to verify its safety and efficiency before practical application.
## Author contribution statement
Chewapat Saejung: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials and analysis tools; Wrote the paper. Khomsorn Lomthaisong: Contributed reagents and analysis tools. Prawphan Kotthale: Performed the experiments; Analyzed and interpreted the data.
## Data availability statement
Data will be made available on request.
## 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.
## Supplementary data
The following is the *Supplementary data* to this article:Multimedia component 1Multimedia component 1
## References
1. Rojas-Garbanzo C., Gleichenhagen M., Heller A., Esquivel P., Schulze-Kaysers N., Schieber A.. **Carotenoid profile, antioxidant capacity, and chromoplasts of pink guava (**. *J. Agric. Food Chem.* (2017) **65** 3737-3747. PMID: 28301727
2. 2Grand View ResearchCarotenoids market analysis by source (natural, synthetic), by product (beta-carotene, lutein, lycopene, astaxanthin, zeaxanthin, canthaxanthin), by application (food, supplements, feed, pharmaceuticals, cosmetics), and segment forecasts, 2018–20252016https ://www.grand viewr esear ch.com/indus try-analy sis/carot enoid s-marke t (2016). Accessed 31 March 2021. (2016)
3. Ram S., Mitra M., Shah F., Tirkey S.R., Mishra S.. **Bacteria as an alternate biofactory for carotenoid production: a review of its applications, opportunities and challenges**. *J. Funct.Foods* (2020) **67**
4. Darwesh A.M., Sosnowski D.K., Lee T.Y., Keshavarz-Bahaghighat H., Seubert J.M.. **Insights into the cardioprotective properties of n-3 PUFAs against ischemic heart disease via modulation of the innate immune system**. *Chem Interact* (2019) **308** 20-44
5. Oliver L., Dietrich T., Maranon I., Villaran M.C., Barrio R.J.. **Producing omega-3 polyunsaturated fatty acids: a review of sustainable sources and future trends for the EPA and DHA market**. *Resources* (2020) **9** 148
6. Vadivelan G., Venkateswaran G.. **Production and enhancement of omega-3 fatty acid from Mortierella alpina CFR-GV15: its food and therapeutic application**. *BioMed Res. Int.* (2014) **657414**
7. Gupta R., Gaur S., Dai X., Sharma M., Chen J.. *Fungi in Sustainable Food Production. Fungal Biology* (2021) 117-128
8. Saini R.K., Keum Y.S.. **Microbial platforms to produce commercially vital carotenoids at industrial scale: an updated review of critical issues**. *J. Ind. Microbiol. Biotechnol.* (2019) **46** 657-674. PMID: 30415292
9. Rodriguez M.G., Rebollar P., Mattioli S., Castellini C.. **n-3 PUFA sources (precursor/products): a review of current knowledge on rabbit**. *Animals* (2019) **9** 806. PMID: 31618904
10. Beligon V., Christophe G., Fontanille P., Larroche C.. **Microbial lipids as potential source to food supplements**. *Curr. Opin. Food Sci.* (2016) **7** 35-42
11. Wang C.C., Ding S., Chiu K.H., Liu W.S., Lin T.J., Wen Z.H.. **Extract from a mutant R**. *Food Nutr. Res.* (2016) **31** 60
12. Hemantkumar J.N., Rahimbhai M.I., Vitova M.. *Microalgae - from Physiology to Application* (2019). DOI: 10.5772/intechopen.90143
13. Francisco M.R., Saldanha T., Fraga M.E.. **Fungi as an alternative to produce essential fatty acids**. *Cientifica* (2017) **45** 123
14. Patel A., Rova U., Christakopoulos P., Matsakas L.. **Simultaneous production of DHA and squalene from Aurantiochytrium sp. grown on forest biomass hydrolysates**. *Biotechnol. Biofuels* (2019) **12** 255. PMID: 31687043
15. Karnaouri A., Chalima A., Kalogiannis K.G., Varamogianni-Mamatsi D., Lappas A., Topakas E.. **Utilization of lignocellulosic biomass towards the production of omega-3 fatty acids by the heterotrophic marine microalga Crypthecodinium cohnii**. *Bioresour. Technol.* (2020) **303**
16. Cao K., Zhi R., Zhang G.. **Photosynthetic bacteria wastewater treatment with the production of value-added products: a review**. *Bioresour. Technol.* (2020) **299**
17. Tamot B., Benning C., Hunter C.N., Daldal F., Thurnauer M.C., Beatty J.T.. *The Purple Phototrophic Bacteria. Advances in Photosynthesis and Respiration* (2009) 119-134
18. Delamare-Deboutteville J., Batatone D.J., Kawasaki M., Stegman S., Salini M., Tabrett S., Smullen R., Barnes A.C., Hulsen T.. **Mixed culture purple phototrophic bacteria is an effective fishmeal replacement in aquaculture**. *Water Res. X* (2019) **4**
19. Wang C., Liu W., Chang F., Tsai P., Tsai M.. *J. Microb. Biochem. Technol.* (2014) **6** 38-42
20. Saejung C., Sanusan W.. **Valorization of lignocellulosic wastes and nutrient recovery by anoxygenic photosynthetic bacteria**. *Waste Biomass Valor* (2021) **12** 4835-4844
21. Patthawaro S., Saejung C.. **Production of single cell protein from manure as animal feed by using photosynthetic bacteria**. *Microbiol.* (2019) **8** e913
22. Saejung C., Chaiyarat A., Sanoamuang L.. **Effects of algae, yeast and photosynthetic bacteria diets on survival and growth performance in the fairy shrimp, Streptocephalus sirindhornae (Branchiopoda, Anostraca)**. *Crustaceana* (2018) **91** 1505-1522
23. Saejung C., Chaiyarat A., Sa-noamuang L.. **Optimization of three anoxygenic photosynthetic bacteria as feed to enhance growth, survival, and water quality in fairy shrimp (Streptocephalus sirindhornae) cultivation**. *Aquaculture* (2021) **534**
24. Anderson D.G., Salm S.N., Allen D.P., Nester E.W.. (2016)
25. Saejung C., Chanthakhot T.. **Single-phase and two-phase cultivations using different light regimes to improve production of valuable substances in the anoxygenic photosynthetic bacterium**. *Bioresour. Technol.* (2021) **328**
26. Aoac. (2019)
27. Sheng G.P., Yu H.Q., Yu Z.. **Extraction of extracellular polymeric substances from the photosynthetic bacterium**. *Appl. Microbiol. Biotechnol.* (2005) **67** 125-130. PMID: 15309338
28. Kwon D.Y., Rhee J.S.. **A simple and rapid colorimetric method for determination of free fatty acids for lipase assay**. *J. Am. Oil Chem. Soc.* (1986) **63** 89-92
29. Saejung C., Puensungnern L.. **Evaluation of molasses-based medium as a low cost medium for carotenoids and fatty acid production by photosynthetic bacteria**. *Waste Biomass Valor* (2020) **11** 143-152
30. Black H.S., Boehm F., Edge R., Truscott T.G.. **The benefits and risks of certain dietary carotenoids that exhibit both anti- and pro-oxidative mechanisms-A comprehensive review**. *Antioxidants* (2020) **9** 264. PMID: 32210038
31. Willey J.M., Sherwood L.M., Woolverton C.J.. (2014)
32. 32The Culinary Institute of AmericaThe Professional Chefninth ed.2011John Wiley & SonsNew Jersey. (2011)
33. Wikizero. (2020)
34. de la Garza A.L., Alba C.T., Cárdenas-Pérez R.E., Camacho A., Gutierrez-Lopez M., Castro H., Vinciguerra M., Sanchez P.C.. *Molecular Nutrition: Mother and Infant* (2021) 135-154
35. Dimzon I.K.D., Valde M.F., Santos J.E.R., Garrovillas M.J.M., Dejarme H.M., Remollo J.M.W., Dayrit F.M.. **Physico-chemical and microbiological parameters in the deterioration of virgin coconut oil**. *Philipp. J. Sci.* (2011) **140** 89-103
36. Armstrong G., Barton S.D., Nakanishi K., Meth-Cohn O.. *Comprehensive Natural Products Chemistry* (1999) 321-352
37. Fozo E.M., Rucks E.A., Poole R.K.. *Advances in Microbial Physiology* (2016) 51-155
38. Wang H., Zhang Y., Zhou W., Noppol L., Liu T.. **Mechanism and enhancement of lipid accumulation in filamentous oleaginous microalgae**. *Biotechnol. Biofuels* (2018) **11** 328. PMID: 30559837
39. Ivanov D.S., Levic J.D., Sredanovic S.A.. **Fatty acid composition of various soybean products**. *Food Feed Res* (2010) **37** 65-70
40. Harris W.S., Coates P.M., Betz J.M., Blackman M.R., Cragg G.M., Levine M., Moss J., White J.D.. *Encyclopedia of Dietary Supplements* (2010) 577-586
41. Tang K.H., Tang Y.J., Blankenship R.E.. **Carbon metabolic pathways in phototrophic bacteria and their broader evolutionary implications**. *Front. Microbiol.* (2011) **2** 165. PMID: 21866228
42. Yen H.W., Palanisamy G., Su G.C.. **The influences of supplemental vegetable oils on the growth and β-carotene accumulation of oleaginous yeast-**. *Biotechnol. Bioproc. Eng.* (2019) **24** 522-528
43. Adkar B.V., Bhattacharyya S., Gilson A.I., Zhang W., Shakhnovich E.I.. **Substrate inhibition imposes fitness penalty at high protein stability**. *Proc. Natl. Acad. Sci. USA* (2019) **116** 11265-11274. PMID: 31097595
44. Spohn M., Chodak M.. **Microbial respiration per unit biomass increases with carbon‐to‐nutrient ratios in forest soils**. *Soil Biol. Biochem.* (2015) **81** 128-133
45. Xia L., Ge H., Zhou X., Zhang D., Hu C.. **Photoautotrophic outdoor two-stage cultivation for oleaginous microalgae Scenedesmus obtusus XJ-15**. *Bioresour. Technol.* (2013) **144** 261-267. PMID: 23876654
46. Hang L.T., Mori K., Tanaka Y., Morikawa M., Toyama T.. **Enhanced lipid productivity of Chlamydomonas reinhardtii with combination of NaCl and CaCl2 stresses**. *Bioproc. Biosyst. Eng.* (2020) **43** 971-980
47. Guldhe A., Renuka N., Singh P., Bux F.. **Effect of phytohormones from different classes on gene expression of Chlorella sorokiniana under nitrogen limitation for enhanced biomass and lipid production**. *Algal Res.* (2021) **40**
48. Chagas A.L., Rios A.O., Jarenkow A., Marcílio N.R., Ayub M.A.Z., Rech R.. **Production of carotenoids and lipids by Dunaliella tertiolecta using CO2 from beer fermentation**. *Process Biochem.* (2015) **50** 981-988
49. Pribyl P., Pilny J., Cepak V., Kastanek P.. **The role of light and nitrogen in growth and carotenoid accumulation in Scenedesmus sp**. *Algal Res.* (2016) **16** 69-75
50. Gong M., Bassi A.. **Investigation of**. *Appl. Biochem. Biotechnol.* (2017) **183** 652-671. PMID: 28647795
51. Nunez-Cardona M.T., Guo X.. *Advances in Gas Chromatography* (2014). DOI: 10.5772/57389
52. Ketels E., De Groote G.. **Effect of ratio of unsaturated to saturated fatty acids of the dietary lipid fraction on utilization and metabolizable energy of added fats in young chicks**. *Poultry Sci.* (1989) **68** 1506-1512. PMID: 2608616
53. Tartrakoon W., Tartrakoon T., Kitsupee N.. **Effects of the ratio of unsaturated fatty acid to saturated fatty acid on the growth performance, carcass and meat quality of finishing pigs**. *Anim Nutr* (2016) **2** 79-85. PMID: 29767086
54. Simopoulos A.P.. **The omega-6/omega-3 fatty acid ratio: health implications**. *OCL* (2010) **17** 267-275
55. Eggersdorfer M., Wyss A.. **Carotenoids in human nutrition and health**. *Arch. Biochem. Biophys.* (2018) **652** 18-26. PMID: 29885291
56. Pandohee J., Kour J., Nayik G.A.. *Nutraceuticals and Health Care* (2022) 279-288
57. Brinton E.A., Mason R.P.. **Prescription omega-3 fatty acid products containing highly purified eicosapentaenoic acid (EPA)**. *Lipids Health Dis.* (2017) **16** 23. PMID: 28137294
58. Yang Q., Liu M., Li X., Zheng J.. **The benefit of a ceramide-linoleic acid-containing moisturizer as an adjunctive therapy for a set of xerotic dermatoses**. *Dermatol. Ther.* (2019) **32**
59. Li J., Guasch-Ferré M., Li Y., Hu F.B.. **Dietary intake and biomarkers of linoleic acid and mortality: systematic review and meta analysis of prospective cohort studies**. *Am. J. Clin. Nutr.* (2020) **112** 150-167. PMID: 32020162
60. Smith D.L., Willis A.L., Nguyen N., Conner D., Zahedi S., Fulks J.. **Eskimo plasma constituents, dihomo-gamma-linolenic acid, eicosapentaenoic acid and docosahexaenoic acid, inhibit the release of atherogenic mitogens**. *Lipids* (1989) **24** 70-75. PMID: 2545997
61. Taki H., Nakamura N., Hamazaki T., Kobayashi M.. **Intravenous injection of tridihomo-gamma-linolenoyl-glycerol into mice and its effects on delayed-type hypersensitivity**. *Lipids* (1993) **28** 873-876. PMID: 8246686
62. Antimanon S., Anantayanon J., Wannawilai S., Khongto B., Laoteng K.. **Physiological traits of dihomo-γ-linolenic acid production of the engineered**. *Front. Microbiol.* (2020) **11**
63. Aldana-Hernández P., Azarcoya-Barrera J., van der Veen J.N., Leonard K.N., Zhao Y.Y., Nelson R., Goruk S., Field C.J., Curtis J.M., Richard C., Jacobs R.L.. **Dietary phosphatidylcholine supplementation reduces atherosclerosis in Ldlr−/− male mice2**. *J. Nutr. Biochem.* (2021) **92**
64. Zhang C., Zhou M.M., Zhang T.T., Cong P.X., Xu J., Xue C.H., Yanagita T., Wei Z.H., Wang Y.M.. **Effects of dietary supplementation with EPA-enriched phosphatidylcholine and phosphatidylethanolamine on glycerophospholipid profile in cerebral cortex of SAMP8 mice fed with high-fat diet**. *J. Oleo Sci.* (2021) **70** 275-287. PMID: 33456004
65. Steed A.L., Christophi G.P., Kaiko G.E., Sun L., Goodwin V.M., Esaulova U.J., Artyomov X.N., Morales D.J., Holtzman M.J., Boon C.M., Lenschow D.J., Stappenbeck T.S.. **The microbial metabolite desaminotyrosine protects from influenza through type I interferon**. *Science* (2017) **357** 498-502. PMID: 28774928
66. Wei Y., Gao J., Kuo Y., Liu M., Meng L., Zheng X., Xu S., Liang M., Sun H., Liu Z., Wang Y.. **The intestinal microbial metabolite desaminotyrosine is an anti-inflammatory molecule that modulates local and systemic immune homeostasis**. *Faseb. J.* (2020) **34** 16117-16128. PMID: 33047367
67. Khan S., Ullah H., Zhang L.. **Bioactive constituents form Buddleja species**. *Pak. J. Pharm. Sci.* (2019) **32** 721-741. PMID: 31081788
68. López-Pedrosa J.M., Manzano M., Baxter J.H., Rueda R.. **N-acetyl-L-glutamine, a liquid-stable source of glutamine, partially prevents changes in body weight and on intestinal immunity induced by protein energy malnutrition in pigs**. *Dig. Dis. Sci.* (2007) **52** 650-658. PMID: 17253138
69. Wishart D.S., Guo A.C., Oler E., Wang F., Anjum A., Peters H., Dizon R., Sayeeda Z., Tian S., Lee B.L., Berjanskii M., Mah R., Yamamoto M., Jovel J., Torres-Calzada C., Hiebert-Giesbrecht M., Lui V.W., Varshavi D., Varshavi D., Allen D., Arndt D., Khetarpal N., Sivakumaran A., Harford K., Sanford S., Yee K., Cao X., Budinski Z., Liigand J., Zhang L., Zheng J., Mandal R., Karu N., Dambrova M., Schiöth H.B., Greiner R., Gautam V.. **Hmdb 5.0: the human Metabolome database for 2022**. *Nucleic Acids Res.* (2022) **501** D622-D631
70. Wiener A., Shudler M., Levit A., Niv M.Y.. **BitterDB: a database of bitter compounds**. *Nucleic Acids Res.* (2012) **40** D413-D419. PMID: 21940398
|
---
title: Ground reaction force and electromyograms of lower limb muscles during fast
walking
authors:
- Akitoshi Makino
- Keiichi Yamaguchi
- Daichi Sumi
- Masaru Ichikawa
- Masumi Ohno
- Akinori Nagano
- Kazushige Goto
journal: Frontiers in Sports and Active Living
year: 2023
pmcid: PMC9981938
doi: 10.3389/fspor.2022.1055302
license: CC BY 4.0
---
# Ground reaction force and electromyograms of lower limb muscles during fast walking
## Abstract
### Background
Physically active status is an important contributor to individual health. Walking is regarded as commonly accepted exercise for exercise promotion. Particularly, interval fast walking (FW), consisting of alternating between fast and slow walking speeds, has gained popularity from practical viewpoints. Although previous studies have determined the short- and long-term effects of FW programs on endurance capacity and cardiovascular variables, factors affecting these outcomes have not been clarified. In addition to physiological variables, understanding of mechanical variables and muscle activity during FW would be a help to understand characteristics of FW. In the present study, we compared the ground reaction force (GRF) and lower limb muscle activity between fast walking (FW) and running at equivalent speeds.
### Method
Eight healthy men performed slow walking ($45\%$ of the maximum walking speed; SW, 3.9 ± 0.2 km/h), FW ($85\%$ of the maximum walking speed, 7.4 ± 0.4 km/h), and running at equivalent speeds (Run) for 4 min each. GRF and average muscle activity (aEMG) were evaluated during the contact, braking, and propulsive phases. Muscle activities were determined for seven lower limb muscles: gluteus maximus (GM), biceps femoris (BF), rectus femoris (RF), vastus lateralis (VL), gastrocnemius medialis (MG), soleus (SOL), and tibialis anterior (TA).
### Results
The anteroposterior GRF was greater in FW than in Run during the propulsive phase ($p \leq 0.001$), whereas the impact load (peak and average vertical GRF) was lower in FW than in Run ($p \leq 0.001$). In the braking phase, lower leg muscle aEMGs were higher during Run than during SW and FW ($p \leq 0.001$). However, in the propulsive phase, soleus muscle activity was greater during FW than during Run ($p \leq 0.001$). aEMG of tibialis anterior was higher during FW than during SW and Run in the contact phase ($p \leq 0.001$). No significant difference between FW and Run was observed for HR and RPE.
### Conclusion
These results suggest that the average muscle activities of lower limbs (e.g., gluteus maximus, rectus femoris, and soleus) during the contact phase were comparable between FW and running, however, the activity patterns of lower limb muscles differed between FW and running, even at equivalent speeds. During running, muscles were mainly activated in the braking phase related to impact. In contrast, during FW, soleus muscle activity during the propulsive phase was increased. Although cardiopulmonary response was not different between FW and running, exercise using FW might be useful for health promotion among individuals who cannot exercise at high-intensity.
## Introduction
Daily physical activity is associated with improved cardiopulmonary function [1] and increased muscle mass and strength [2]. Reduced physical activity leads to metabolic disturbances and related diseases, including insulin resistance [3], hyperglycemia [4], and atherosclerosis [5]. Increased energy expenditure (EE) during exercise also reduces the risk of cardiovascular diseases [6]. Therefore, strategies to increase exercise-induced EE are essential for health promotion.
Walking is widely accepted exercise modality among the adults [7, 8]. Particularly, interval fast walking (FW), consisting of alternating between fast and slow walking speeds, has gained popularity from practical viewpoint. Cycling (pedaling) exercise is generally easy to perform but requires a stationary bike. Running does not require equipment but may pose risks of injury for those with orthopedic or other medical problems. On the other hand, it was reported that walking has a lower risk of injury than running [9]. Previous studies have demonstrated improvements in maximal oxygen uptake, blood pressure, and heart rate after 5 months of FW (≥ 5 sets of FW at $70\%$–$85\%$ of peak aerobic capacity followed by slow walking at ≤ $40\%$ of peak aerobic capacity; 3 min per set) [10, 11]. Interval FW for 2 weeks improved insulin sensitivity, while reducing 24 h maximum glucose levels and mean amplitudes of glycemic excursion, in patients with type 2 diabetes [12].
Although previous studies have established the short- and long-term effects of FW programs on endurance capacity and cardiovascular variables, factors influencing these outcomes have not been investigated [10, 12]. In our previous study, we demonstrated that EE and carbohydrate oxidation during walking were enhanced in a non-linear manner with increasing speed. It was notable that walking at speeds > 8.0 km/h caused greater EE and carbohydrate oxidation than running at an equivalent speed in young individuals [13]. Moreover, previous studies reported that interval walking for 17 weeks or 5 months increased muscle strength of knee extensors and flexors muscles [14, 15],. Although Kubo et al. [ 16] reported that a walking exercise program for six months increased muscle strength of the knee flexors, no significant increase in muscle strength of the knee extensors was found.
In addition to physiological variables, understanding of mechanical variables and muscle activity during FW would be a help to clarify characteristics of FW. Walking speeds influence biomechanical variables such as joint kinematics, GRFs, joint moments of moments and powers, muscle activities, and spatiotemporal gait parameters [17]. Previous studies compared GRFs [18] and muscle activity of the lower limbs [19] among different walking speeds. However, in these studies, no significant differences in GRFs were found between slow and normal speeds. Hence, comparisons of both GRFs and lower limbs of muscle activity between FW and running are currently required.
Therefore, the purpose of the present study was to compare the ground reaction force (GRF) and lower limb muscle activity between FW and running at equivalent speeds. We hypothesized that the GRF would be lower during FW than during running. Moreover, we hypothesized that lower limb muscle activity during FW would be mainly enhanced in the propulsive phase, whereas it would be activated in the braking phase while running.
## Participants
Eight men were recruited in the present study (mean ± standard deviation: age, 22 ± 1 y; height, 172.1 ± 1.7 cm; weight, 62.1 ± 7.0 kg); they received an overview of the experiment and possible risks (Table 1). None of them had any history of chronic diseases that could affect neuromuscular function, exercise, or daily physical activity. All participants were not involved in any training programs at the start of the study. Written informed consent was obtained from all participants. This study was approved by the ethics committee for Human Experiments at Ritsumeikan University (BKC-IRB-2020–047) and was conducted in accordance with the Declaration of Helsinki.
**Table 1**
| Age | 22 ± 1 Year |
| --- | --- |
| Height | 172.1 ± 1.7 cm |
| Weight | 62.1 ± 7.0 kg |
| 45% MWS (slow walk) | 3.9 ± 0.2 km/h |
| 85% MWS (Fast walk and run) | 7.4 ± 0.4 km/h |
| MWS | 8.8 ± 0.5 km/h |
## Experimental overview
Participants visited the laboratory twice throughout the experimental period. On the first visit, a familiarization session and determination of the maximal walking speed (MWS) were conducted. On the second visit, each participant performed the main experimental trials, consisting of slow walk trial, FW and Run trial. The anteroposterior and vertical GRF components, surface electromyography (EMG) of lower limb muscles were evaluated during each trial.
## MWS measurements
The participants began walking on a treadmill (Elevation series E95Ta; Life Fitness Corp., Franklin Park, IL, United States) at a speed of 4.0 km/h. The speed was progressively increased by 1.0 km/h at 1 min intervals until the participants could no longer match the speed; this speed was recorded as the MWS.
## Main experiment
The participants walked for 4 min at $45\%$ of MWS (slow walk) on a special treadmill with built-in force plates (HPT-2200D; Tec Gihan Co., Ltd., Kyoto, Japan); they then ran (Run) or walked at $85\%$ of MWS (fast walk) for 4 min. The exercises were separated by 3-min rest periods. The order of running and FW was randomized. Based on our previous study which revealed significantly greater EE and carbohydrate oxidation in FW than in running [13], we selected $85\%$ of MWS during fast walk phase. Heart rate (HR), rating of perceived exertion (RPE), GRF, and EMG were measured during the exercise (Figure 1).
**Figure 1:** *Protocol overview.*
## GRF and EMG measurements
A dual-belt treadmill with two force plates (HPT-2200D; Tec Gihan Co., Ltd.) was used for GRF measurements. Surface EMGs were recorded and amplified (SX230–1000, Biometrics Ltd., Ghent, United Kingdom) from seven right lower limb muscles: gluteus maximus (GM), biceps femoris (BF), rectus femoris (RF), vastus lateralis (VL), gastrocnemius medialis (MG), soleus (SOL), and tibialis anterior (TA). EMG electrode placement was based on the guidelines for non-invasive surface EMG assessment of muscles [20] (Figure 2). GRFs and EMGs were recorded at a sampling frequency of 1 kHz using a data acquisition and analysis system (LabChart; ADInstruments, Sydney, Australia) with a 16-bit analog-to-digital converter (PowerLab/16SP; ADInstruments).
**Figure 2:** *The measurement places of surface electromyography (EMG) of lower limb muscles. The EMGs were recorded and amplified from seven right lower limb muscles. GM; gluteus maximus, BF; biceps femoris, RF; rectus femoris, VL; vastus lateralis, MG; gastrocnemius medialis, SOL; soleus, TA; tibialis anterior.*
## HR and RPE
HR was measured continuously (every 5 s) using a wireless HR monitor (RCX5; Polar Electro, Kempele, Finland). RPE was evaluated using a 10-point scale [21] at the end of each trial.
Table 2 shows the HR and RPE. HR was significantly higher in FW and Run than in SW, but it did not significantly differ between FW and Run. RPE for breath (RPEbreath) was significantly higher in FW and Run than in SW, with no difference between FW and Run. RPE for leg muscles (RPEleg) was significantly higher in FW and Run than in SW, and it was higher in FW than in Run.
**Table 2**
| Unnamed: 0 | Unnamed: 1 | Slow walk | Fast walk | Run | Post hoc test (p-value) | Post hoc test (p-value).1 | Post hoc test (p-value).2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | (SW) | (FW) | (Run) | SW/FW | FW/Run | SW/Run |
| HR (bpm) | 90 ± 14 | 147 ± 17 | 145 ± 19 | <0.001 | 0.851 | <0.001 | |
| RPE | breath | 1.3 ± 0.5 | 3.5 ± 1.2 | 2.8 ± 0.7 | <0.001 | 0.101 | <0.01 |
| leg | 1.3 ± 0.5 | 4.9 ± 1.4 | 3.4 ± 1.1 | <0.001 | <0.01 | <0.001 | |
## Data analyses
The anteroposterior and vertical GRF components were analyzed after they had been filtered at 40 Hz and 60 Hz, respectively, using a low-pass Butterworth filter. The contact phase, between the foot strike and toe-off, had a detection threshold of 50 N for the vertical GRF component. The braking and propulsive phases were between the foot strike and braking-to-propulsion transition, and between the braking-to-propulsion transition and toe-off, respectively. All GRF values were normalized according to body weight.
EMG signals were filtered at 20–450 Hz using band-pass Butterworth filters, then rectified and smoothed at 60 Hz using low-pass Butterworth filters. The muscle activity for each phase was calculated as the average EMG (aEMG) amplitude, using maximum voluntary contraction as the reference. EMG data were reported for the contact, braking, and propulsive phases. The GRF data and EMG signals during 10 steps have been averaged individually. In the GRF data, impulse and averaged force for each phase were calculated. The rectified time course EMG signals were aEMG for each phase.
## Statistical analyses
Data are presented as means ± standard deviations. A commercially available statistical software (SigmaStat 2.03; SPSS, Inc.) was utilized. For comparisons of each variable among the trials, one-way repeated-measures analysis of variance (repeated ANOVA) was used to compare the main effect. When ANOVA found significant main effect, post hoc test (Tukey method) was performed to identify specific pairwise differences. The level of statistical significance was set at $p \leq 0.05.$
## Results
Table 1 shows walking speed variables.
## GRFs comparison between SW, FW, and Run
Figure 3 shows the anteroposterior average and impulse GRFs. The impulse was significantly lower for Run than for SW or FW during the braking phase and during the propulsive phase. During the braking phase, the average force was significantly higher for FW and Run than for SW; it was higher for FW than for Run. During the propulsive phase, the average force was significantly higher for FW than for SW or Run.
Figure 4 shows the vertical peak, average, and impulse GRFs. The peak force was significantly higher for FW and Run than for SW during the contact phase; it was significantly higher for Run than for FW. The average force was significantly higher for FW and Run than for SW during the contact phase; it was significantly greater for Run than for FW. The impulse was significantly lower for FW and Run than for SW during the contact phase; it was lower for Run than for FW.
## EMGs comparison between SW, FW, and Run
Figure 5 shows the aEMGs for the GM, BF, RF, and VL during the contact, braking, and propulsive phases. The contact phase aEMG for the GM was significantly higher in Run than in SW. The braking phase aEMG for the GM was significantly higher in Run than in SW and FW; it was higher in FW than in SW. The contact phase aEMG for the BF was significantly higher in Run than in SW or FW; it was higher in FW than in SW. The braking phase aEMG for the BF was significantly higher in FW and Run than in SW. Furthermore, the propulsive phase aEMG for the BF was higher in Run than in SW or FW. The contact phase aEMG for the RF was significantly higher in FW and Run than in SW. The braking phase aEMG for the RF was significantly higher in Run than in SW or FW; it was higher in FW than in SW. The contact phase aEMG for the VL was higher in Run than in SW or FW; it was higher in FW than in SW. The braking phase aEMG for the VL was higher in Run than in SW or FW; it was higher in FW than in SW.
**Figure 3:** *Anteroposterior component of ground reaction force during the braking (A,B) and propulsive (C,D) phases. Values are means ± SD. Significant difference between trials (*p < 0.05, **p < 0.01, ***p < 0.001). SW; Slow walk trial. FW; Fast walk trial. Run; Run trial.*
Figure 6 shows the aEMGs for the MG, SOL, and TA during the contact, braking, and propulsive phases. The contact phase aEMG for the MG was significantly higher in Run than in SW or FW; it was higher in FW than in SW. The braking phase aEMG for the MG was significantly higher in Run than in FW or SW. The propulsive phase aEMG for the MG was significantly higher in FW than in SW. The contact phase aEMG for the SOL was significantly higher in Run and FW than in SW. The braking phase aEMG for the SOL was significantly higher in Run than in SW or FW. The propulsive phase aEMG for the SOL was significantly higher in FW than in SW or Run. The aEMG for the TA during all phases was significantly higher in FW than in SW or Run.
**Figure 4:** *The peak (A), average (B) and impulse (C) of vertical component of ground reaction force. Values are means ± SD. Significant difference between trials (*p < 0.05, **p < 0.01, ***p < 0.001). SW; Slow walk trial. FW; Fast walk trial. Run; Run trial.* **Figure 5:** *Averaged surface electromyography (aEMG) of femoral muscles during the contact (A—D), braking (E—F) and propulsive (I—N) phases. Values are means ± SD. Significant difference between trials (*p < 0.05, **p < 0.01, ***p < 0.001). SW; Slow walk trial. FW; Fast walk trial. Run; Run trial. GM; Gluteus maximus, BF; Biceps femoris, RF; Rectus femoris, VL; Vastus lateralis.* **Figure 6:** *Averaged surface electromyography (aEMG) of lower leg muscles during the contact (A—C), braking (D—F) and propulsive (G—I) phases. Values are means ± SD. Significant difference between trials (*p < 0.05, **p < 0.01, ***p < 0.001). SW; Slow walk trial. FW; Fast walk trial. Run; Run trial. MG; Gastrocnemius medialis, SOL; Soleus, TA; Tibialis anterior.*
## Discussion
The present study compared GRF and lower limb muscle activity between FW and running at equivalent speeds. Consequently, anteroposterior GRF was greater during FW than during Run, whereas vertical GRF was greater during Run than during FW. Moreover, muscle activity during the braking phase was lower in FW than in Run, while it was higher in FW during the propulsive phase. These findings suggest that FW causes less mechanical stress during impact and produces greater propulsive force, compared with running at an equivalent speed.
Despite equivalent speed, the impulse and average of anteroposterior GRF were significantly greater during FW than during running. In contrast, the average vertical GRF was higher during running than during FW. These results may be related to differences in locomotion [22]. We [13] reported that EE and carbohydrate oxidation during walking were enhanced in a non-linear manner with increasing speed. Peak GM activity generally occurs immediately after foot-ground contact [23]. In the present study, aEMGs for the GM, RF, and VL during braking phase were greater in Run than in FW. aEMGs during breaking phase showed similar trend for average vertical GRF. In the braking phase of running, the lower limb joints (i.e., hip, knee, and ankle joints) flexed because of the impact during ground contact. In the subsequent propulsive phase, these joints were extended, and the mechanical energy exchange involved elastic energy storage and release [24]. Also, the peak of vertical GRF was significantly greater during running than during FW, and it is advantageous for the mechanical energy from exchange stretch- shortening cycle perspective [25].
Gazendam & Hof [23] reported minor hamstring differences in the walking and running profiles, but major differences in the EMG profiles were observed for the lower limb muscles. In the present study, aEMGs for the MG and SOL during braking phase were higher in running than in walking. However, aEMG for the MG during the propulsive phase was significantly higher in FW than in SW, and aEMG for the SOL was significantly higher in FW than in SW and running. Less knee flexion with greater ankle plantarflexion decreased mechanical efficiency at fast speed [26], and it may augment mechanical work, thus decreasing mechanical efficiency [27]. Furthermore, impulse and average anteroposterior GRFs were significantly higher in FW than in running during the propulsive phase. Therefore, increased muscle activities during push-off might be involved in the greater EE during walking [28]. The present study also demonstrated differences in aEMGs for the TA during FW and Run. The EMG of TA was significantly higher in FW than in SW and Run during all phases. The increased aEMG for the TA during FW would contribute to maintaining the ankle joint in a dorsiflexed position, thus improving ankle joint stability while walking [29]. Furthermore, the augmented muscle activities of lower limb muscles might explain significantly higher score of RPEleg in FW than in Run.
As limitations of the present study, joint torques and powers were not evaluated. Also, the evaluations of GRFs were limited during 4 min of exercise (walking or running). In further study, determination of GRFs during actual interval FW exercise would be valuable.
## Conclusion
FW resulted in a smaller impact (i.e., vertical GRF) than running at an equivalent speed. During the contact phase, the average muscle activities of GM, RF, and SOL were not significantly different between FW and running. However, further analyses presented that the activity patterns of lower limb muscles differed between FW and running, even at equivalent speeds. In running, GM, RF, VL, MG, and SOL were mainly activated during the braking phase than the propulsive phase. On the other hand, in FW, the muscle activity of SOL during the propulsive phase was increased than during the braking phase. Therefore, FW causes smaller mechanical stress than running at an equivalent speed, but it highly activates lower limb muscles. These notions may be useful for designing exercise program for health promotion, particularly in individuals with orthopedic or other medical issues (e. g., joint injuries, type 2 diabetes, and obesity).
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by The ethics committee for Human Experiments at Ritsumeikan University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
AM, AN and KG contributed to the study design, collection, analysis, interpretation of data and manuscript writing. KY, DS, MI and MO contributed to the study design, data collection and interpretation of data. All authors contributed to the article and approved the submitted version.
## Conflict of interest
MI, MO were employed by ASICS Corporation. 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.
## References
1. Sazlina SG, Sooryanarayana R, Ho BK, Omar MA, Krishnapillai AD, Tohit NM. **Cardiovascular disease risk factors among older people: data from the national health and morbidity survey 2015**. *PLoS One* (2020) **15** 1-11. DOI: 10.1371/journal.pone.0240826
2. Larsson L, Degens H, Li M, Salviati L, Lee YI, Thompson W. **Sarcopenia: aging-related loss of muscle mass and function**. *Physiol Rev* (2019) **99** 427-511. DOI: 10.1152/physrev.00061.2017
3. Lee CG, Boyko EJ, Strotmeyer ES, Lewis CE, Cawthon PM, Hoffman AR. **Association between insulin resistance and lean mass loss and fat mass gain in older men without diabetes mellitus**. *J Am Geriatr Soc* (2011) **59** 1217-24. DOI: 10.1111/j.1532-5415.2011.03472.x
4. Sanada K, Iemitsu M, Murakami H, Gando Y, Kawano H, Kawakami R. **Adverse effects of coexistence of sarcopenia and metabolic syndrome in Japanese women**. *Eur J Clin Nutr* (2012) **66** 1093-8. DOI: 10.1038/ejcn.2012.43
5. Abe T, Thiebaud RS, Loenneke JP, Bemben MG, Loftin M, Fukunaga T. **Influence of severe sarcopenia on cardiovascular risk factors in nonobese men**. *Metab Syndr Relat Disord* (2012) **10** 407-12. DOI: 10.1089/met.2012.0057
6. Kraus WE, Powell KE, Haskell WL, Janz KF, Campbell WW, Jakicic JM. **Physical activity, all-cause and cardiovascular mortality, and cardiovascular disease**. *Med Sci Sports Exerc* (2019) **51** 1270-81. DOI: 10.1249/MSS.0000000000001939
7. Hiwa MA, Muhammed BM. **Population-level interventions based on walking and cycling as a means to increase physical activity**. *Phys Act and Health* (2021) **5** 55-63. DOI: 10.5334/paah.87
8. Norwood P, Eberth B, Farrar S, Anable J, Ludbrook A. **Active travel intervention and physical activity behaviour: an evaluation**. *Soc Sci Med* (2014) **113** 50-8. DOI: 10.1016/j.socscimed.2014.05.003
9. Colbert LH, Hootman JM, Macera CA. **Physical activity-related injuries in walkers and runners in the aerobics center longitudinal study**. *Clin J Sport Med* (2000) **10** 259-63. DOI: 10.1097/00042752-200010000-00006
10. Nemoto K, Gen-no H, Masuki S, Okazaki K, Nose H. **Effects of high-intensity interval walking training on physical fitness and blood pressure in middle-aged and older people**. *Mayo Clin Proc* (2007) **82** 803-11. DOI: 10.4065/82.7.803
11. Masuki S, Morikawa M, Nose H. **High-Intensity walking time is a key determinant to increase physical fitness and improve health outcomes after interval walking training in middle-aged and older people**. *Mayo Clin Proc* (2019) **94** 2415-26. DOI: 10.1016/j.mayocp.2019.04.039
12. Karstoft K, Clark MA, Jakobsen I, Müller IA, Pedersen BK, Solomon TPJ. **The effects of 2 weeks of interval vs continuous walking training on glycaemic control and whole-body oxidative stress in individuals with type 2 diabetes: a controlled, randomised, crossover trial**. *Diabetologia* (2017) **60** 508-17. DOI: 10.1007/s00125-016-4170-6
13. Makino A, Yamaguchi K, Sumi D, Ichikawa M, Ohno M, Goto K. **Comparison of energy expenditure and substrate oxidation between walking and running in men and women**. *Phys Act Nutr* (2022) **26** 8-13. DOI: 10.20463/pan.2022.0002
14. Okazaki K, Yazawa D, Goto M, Kamijo Y-I, Furihata M, Gen-no H. **Effects of macronutrient intake on thigh muscle mass during home-based walking training in middle-aged and older women**. *Scand J Med Sci Sports* (2013) **23** e286-92. DOI: 10.1111/sms.12076
15. Ozaki H, Nakagata T, Yoshihara T, Kitada T, Natsume T, Ishihara Y. **Effects of progressive walking and stair-climbing training program on muscle size and strength of the lower body in untrained older adults**. *J Sports Sci Med* (2019) **18** 722-8. PMID: 31827357
16. Kubo K, Ishida Y, Suzuki S, Komuro T, Shirasawa H, Ishiguro N. **Effects of 6 months of walking training on lower limb muscle and tendon in elderly**. *Scand J Med Sci Sports* (2008) **18** 31-9. DOI: 10.1111/j.1600-0838.2007.00654.x
17. Fukuchi CA, Fukuchi RK, Duarte M. **Effects of walking speed on gait biomechanics in healthy participants: a systematic review and meta-analysis**. *Syst Rev* (2019) **8** 153. DOI: 10.1186/s13643-019-1063-z
18. Sun D, Fekete G, Mei Q, Gu Y. **The effect of walking speed on the foot inter-segment kinematics, ground reaction forces and lower limb joint moments**. *PeerJ* (2018) **6** e5517. DOI: 10.7717/peerj.5517
19. Chiu MC, Wang MJ. **The effect of gait speed and gender on perceived exertion, muscle activity, joint motion of lower extremity, ground reaction force and heart rate during normal walking**. *Gait Posture* (2007) **25** 385-92. DOI: 10.1016/j.gaitpost.2006.05.008
20. Hermens HJ, Freriks B, Disselhorst-Klug C, Rau G. **Development of recommendations for SEMG sensors and sensor placement procedures**. *J Electromyogr Kinesiol* (2000) **10** 361-74. DOI: 10.1016/s1050-6411(00)00027-4
21. Christian RJ, Bishop DJ, Billaut F, Girard O. **The role of sense of effort on self-selected cycling power output**. *Front Physiol* (2014) **5** 115. DOI: 10.3389/fphys.2014.00115
22. Dugan SA, Bhat KP. **Biomechanics and analysis of running gait**. *Phys Med Rehabil Clin N Am* (2005) **16** 603-21. DOI: 10.1016/j.pmr.2005.02.007
23. Gazendam MG, Hof AL. **Averaged EMG profiles in jogging and running at different speeds**. *Gait Posture* (2007) **25** 604-14. DOI: 10.1016/j.gaitpost.2006.06.013
24. Cavagna GA. **Storage and utilization of elastic energy in skeletal muscle**. *Exerc Sport Sci Rev* (1977) **5** 89-129. DOI: 10.1249/00003677-197700050-00004
25. Earp JE, Newton RU, Cormie P, Blazevich AJ. **The influence of loading intensity on muscle-tendon unit behavior during maximal knee extensor stretch shortening cycle exercise**. *Eur J Appl Physiol* (2014) **114** 59-69. DOI: 10.1007/s00421-013-2744-2
26. DeVita P, Hortobágyi T. **Obesity is not associated with increased knee joint torque and power during level walking**. *J Biomech* (2003) **36** 1355-62. DOI: 10.1016/s0021-9290(03)00119-2
27. Massaad F, Lejeune TM, Detrembleur C. **The up and down bobbing of human walking: a compromise between muscle work and efficiency [published correction appears in**. *J Physiol* (2007) **582** 789-99. DOI: 10.1113/jphysiol.2007.127969
28. Clark WH, Pimentel RE, Franz JR. **Imaging and simulation of inter-muscular differences in Triceps surae contributions to forward propulsion during walking**. *Ann Biomed Eng* (2021) **49** 703-15. DOI: 10.1007/s10439-020-02594-x
29. Mian OS, Thom JM, Ardigò LP, Narici MV, Minetti AE. **Metabolic cost, mechanical work, and efficiency during walking in young and older men**. *Acta Physiol (Oxf)* (2006) **186** 127-39. DOI: 10.1111/j.1748-1716.2006.01522.x
|
---
title: Comparison of meat quality and glycolysis potential of two hybrid pigs in three-way
hybrid model
authors:
- Yongxiang Li
- Yang He
- Jinming Ran
- Ying Huang
- Xian Li
- Hengxin Jiang
- Xueyan Li
- Yangsu Pan
- Sumei Zhao
- Chunlian Song
- Hongbin Pan
- Hong Hu
journal: Frontiers in Veterinary Science
year: 2023
pmcid: PMC9981941
doi: 10.3389/fvets.2023.1136485
license: CC BY 4.0
---
# Comparison of meat quality and glycolysis potential of two hybrid pigs in three-way hybrid model
## Abstract
With the improvement of consumers' requirements for pork quality, the method of crossbreeding with excellent local pig breeds to improve meat quality is popular. Saba pig has high reproduction rate, good meat quality and high utilization rate of roughage, but its excellent characteristics have not been fully developed and utilized. To promote the development and utilization of Saba pigs and production of high-quality pork, the meat quality traits and glycolysis potential of Duroc × (Landrace × Yorkshire) (DLY), Berkshire × (Duroc × Saba) (BDS), and Duroc × (Berkshire × Saba) (DBS) three-way crossbred pigs were compared. The results showed that DLY had the highest live weight, carcass weight, lean meat percentage, drip loss, glycolysis potential, muscle diameter, and relative mRNA expression levels of type IIb muscle fibers as well as the lowest ultimate pH ($p \leq 0.05$). The lightness value of DBS was the highest ($p \leq 0.05$). Among the three crossbred pigs, myristic, arachidic, palmitoleic, and eicosenoic acids were the highest in BDS. These results indicated that the carcass traits of local crossbred pigs were worse than those of DLY pigs, but meat quality was markedly higher, with BDS showing the best meat quality.
## Introduction
With rapid global population and economic growth, meat production and consumption have increased substantially in recent decades [1]. More than one third of the world's population consumes pork, making it one of the main sources of animal protein for humans [2]. As an important economic trait, meat quality is mainly determined by edibility and nutritional quality and is affected by numerous factors such as variety, muscle characteristics, and production settings [3]. Duroc × (Landrace × Yorkshire) (DLY) pigs are popular in the Chinese pig industry because of their fast growth rate, high lean meat percentage, and high meat productivity [4]. However, their meat quality does not currently meet consumer demand [5].
As one of the largest pig producers worldwide, China has many indigenous pig breeds with unique genetic characteristics that have developed over thousands of years due to domestication and natural selection [6]. Saba pigs, which are produced in high-altitude areas of Yunnan, exhibit excellent characteristics in meat quality, crude feed utilization rate, and reproduction rate [7]. However, purebred local pigs are rarely produced commercially in China because of their low feed/gain ratio and lean meat percentage.
Owing to heterosis and breed complementarity, hybrid offspring frequently exhibit better performance than their paternal and maternal lines. Local animal breeds (especially pigs and poultry) are often used in two-way and three-way crossbreeding systems to produce progenies with superior meat quality [6, 8]. Three-way terminal crosses are widely used in commercial pig production and show higher production efficiency than pure breeds or two-way crosses [9]. Berkshire pigs are renowned for their excellent lean meat quality; thus, crossing them with local pigs can improve carcass traits in the resulting offspring. In addition, Duroc pigs are often used as terminal sires in crossbreeding because of their fast growth rate, high feed conversion efficiency, and lean meat percentage [10].
To date, Saba pigs have not been effectively developed and utilized, and the meat quality of three-way crossbred Saba, Berkshire, and Duroc pigs remains poorly explored. This study aimed to compare and analyze the meat quality, fatty acid and muscle fiber composition, and glycolysis potential in DLY, Berkshire × (Duroc × Saba) (BDS), and Duroc × (Berkshire × Saba) (DBS) three-way crossbred pigs. Our findings provide data for the research, development, and utilization of Saba pig breeds and their genetic resources as well as scientific evidence for pig breeding systems that produce high-quality pork and promote the utilization of local pig breed resources in China.
## Animals and sample collection
In total, 100 weaned BDS, DBS, and DLY piglets (35 days old) were selected and raised in the same building (20 pigs per pen), equipped with a fully slatted floor, feeders, and nipple drinkers. All pigs were raised in professional breeding cooperatives in Pude Village, Malutang Township, Luquan County, Kunming City, Yunnan Province, China. The pigs were fed the same National Research Council [2012] three-stage diet (Supplementary Table 1) until 210 days of age. Ten healthy and weight-matched pigs of each breed were randomly selected. Weighed and recorded (live weight) after fasting for 24 h, then euthanized via exsanguination. Subsequently, carcass measurements were collected according to Song et al. [ 11] method [11], including carcass weight (remove head, hoof, tail and viscera), carcass length (the distance from the pubic symphysis leading edge to the fovea of the first cervical spine on the left side of the carcass) and lean meat rate (the percentage of lean meat weight to carcass weight). The longissimus dorsi muscle of each pig was then removed to evaluate meat quality and glycolysis potential. The samples were placed in a refrigerator at 4°C, and meat quality characteristics were evaluated after 45 min. Moreover, a section of each muscle sample was divided, packaged, and frozen at −20°C to measure lactate, glycogen, and pH at 24 and 48 h. All animal experiments were approved by the Institutional Animal Care and Use Committee of Yunnan Agricultural University (No. YNAU20211004).
## Meat quality characteristics
Moisture and lipid contents were measured as per the standards provided by the Association of Official Analytical Collaboration (AOAC) International [12]. Drip loss was determined following methods outlined by Choi et al. [ 13], with slight modifications. Longissimus dorsi muscle samples (2 × 3 × 5 cm) were taken, weighed, and then vacuum-packed in a plastic bag. After refrigeration (4°C) for 24 h, the sample was weighed again, and the weight difference of the sample was considered the drip loss. A tenderness meter (C-LM3B, Harbin, China) was used to measure the shear force of the longissimus dorsi muscle, and each sample was measured three times. A portable pH meter (HANNA HI9125, Italy) was used to measure the pH of the longissimus dorsi muscle. The pH meter was calibrated with standard buffers of pH 7.0 and 4.0 prior to experimental readings. A colorimeter (CR-400, Minolta, Japan) was used to measure the meat color of muscles, and each sample was measured three times.
Table 2 summarizes the meat quality characteristics of the longissimus dorsi muscle for DLY, BDS, and DBS. No significant differences in water and fat content were found among DLY, BDS, and DBS. Regarding meat quality characteristics, the drip loss of DLY was highest, with a significantly higher value than that of BDS and DBS ($p \leq 0.05$). Over time, pork pH showed a downward trend in all three breeds. At 45 min, the pH was significantly higher in DLY and BDS than in DBS ($p \leq 0.05$). At 24 h, the pH was highest in BDS and significantly higher than in DLY ($p \leq 0.05$). At 48 h, the pH was lowest in DLY ($p \leq 0.05$). No significant differences were found among the three pig breeds in terms of the shear force. For longissimus dorsi muscle color, BDS showed the lowest lightness value; the lightness of BDS was significantly lower than that of DBS ($p \leq 0.05$), while redness and yellowness were not significantly different among the three breeds.
**Table 2**
| Item | DLY | BDS | DBS |
| --- | --- | --- | --- |
| Moisture (%) | 68.57 ± 1.22 | 67.84 ± 0.25 | 67.45 ± 0.41 |
| Fat (%) | 3.92 ± 0.53 | 3.88 ± 0.32 | 4.18 ± 0.30 |
| Drip loss (%) | 2.82 ± 0.31a | 1.53 ± 0.18b | 2.15 ± 0. 17b |
| pH (45 min) | 6.40 ± 0.13a | 6.31 ± 0.08a | 5.88 ± 0.05b |
| pH (24 h) | 5.54 ± 0.05b | 5.68 ± 0.03a | 5.62 ± 0.03ab |
| pH (48 h) | 5.41 ± 0.01b | 5.52 ± 0.02a | 5.57 ± 0.03a |
| Shear force (kg/cm2) | 5.08 ± 0.37 | 5.91 ± 0.83 | 5.48 ± 0.48 |
| Meat color | Meat color | Meat color | Meat color |
| Lightness | 41.56 ± 1.29ab | 39.21 ± 1.11b | 42.84 ± 0.73a |
| Redness | 6.81 ± 1.01 | 6.36 ± 0.33 | 6.92 ± 0.41 |
| Yellowness | 1.63 ± 0.52 | 1.32 ± 0.09 | 1.59 ± 0.25 |
## Measurements of fatty acids
Total fat was extracted as previously described [14]. The fatty acid content of the obtained sample was determined according to the method of Lee et al. under a modified oven temperature [15]. Briefly, gas chromatography [TRACE 1310, Jinheng Instrument (Shanghai) Co., Ltd. China] was used to separate and quantify fatty acid methyl ester. The initial oven temperature was set to 100°C, held for 13 min, then increased at a rate of 10°C/min to 180°C, held for 6 min, then again increased at a rate of 1°C/min to 200°C, held for 20 min, and finally increased at a rate of 4°C/min to 230°C and held for 10.5 min. The temperatures of the injector and detector were 270°C and 280°C, respectively.
## Quantitative real-time polymerase chain reaction (qRT-PCR)
The RNAprep Pure Tissue Kit (Tiangen Biochemical Technology Co., Ltd., Beijing, China) was used to extract total RNA, and $3.0\%$ agarose gel electrophoresis was used to detect the concentration and purity of RNA samples. First-strand cDNA was then synthesized using the FastKing RT Kit (with gDNase) (Tiangen Biochemical Technology Co., Ltd., Beijing, China). PCR primers are shown in Supplementary Table 2. The reaction system contained template cDNA (2 μL), ddH2O (6.8 μL), forward primer (0.6 μL), reverse primer (0.6 μL), and 2 × SuperReal PreMix Plus (10.0 μL). PCR was conducted under the following conditions: initial single cycle denaturation at 95°C for 30 s, followed by 40 cycles of 95°C for 5 s, and finally at 60°C for 30 s. The 2-ΔΔCt method was used to calculate relative gene expression.
## Histochemical characteristics
The meat samples were sliced (10 μm) using a low-temperature microtome at −20°C (SYD-K2040, Shenyang Yude Electronic Instrument Co., Ltd., China). Then, using the improved Brooke and Kaiser [16] method, samples were incubated for histochemical demonstration of myosin adenosine triphosphatase after pre-incubation under alkaline (pH 10.4) and acidic (pH 4.35) conditions. The stained sections were examined using an image analysis system (Nikon E600, Nikon Corporation, Japan). Muscle fibers were classified into types I, IIa, and IIb according to Brooke and Kaiser [16]. Note that types IIA and IIB could not be clearly defined using the alkali method. Image-Pro Plus (v6.0) was used to measure the fiber diameter of each sample.
## Glycolytic metabolite measurement
Frozen samples were pulverized, and glycogen and lactate concentrations were measured according to the method reported by Luo et al. [ 17]. Glycolytic potential (GP) (μmol/g meat) was determined according to the formula of Li et al. [ 18]: GP = 2 × (glycogen) + (lactic acid).
## Statistical analysis
SPSS (v21) was used for one-way ANOVA. Multiple comparisons were performed using Duncan's multiple range test. At $p \leq 0.05$, data were statistically significant, and results are presented as mean ± standard error of the mean (SEM).
## Carcass characteristics
Comparisons of the longissimus dorsi muscle and carcass characteristics among DLY, BDS, and DBS are shown in Table 1. In DLY, live weight and carcass weight were significantly higher than those in BDS and DBS, and the lean meat rate was significantly higher than that in DBS ($p \leq 0.05$). Carcass weight in DBS was significantly higher than that in BDS ($p \leq 0.05$). However, there was no significant difference in carcass length among the three pig breeds.
**Table 1**
| Items | DLY | BDS | DBS |
| --- | --- | --- | --- |
| Live weight (kg) | 128.20 ± 2.33a | 105.15 ± 2.54b | 113.90 ± 3.75b |
| Carcass weight (kg) | 99.92 ± 1.76a | 73.50 ± 2.03c | 90.78 ± 3.59b |
| Carcass length (cm) | 88.17 ± 3.34 | 85.51 ± 1.19 | 91.20 ± 1.58 |
| Lean meat rate (%) | 61.61 ± 0.25a | 58.27 ± 4.17ab | 52.31 ± 2.68b |
## Fatty acid analysis
The fatty acid composition in the longissimus dorsi muscles of the DLY, BDS, and DBS groups are compared in Table 3. Oleic, palmitic, linoleic, and stearic acid are the main fatty acids in longissimus dorsi muscle. Compared to DLY and DBS, BDS had higher concentrations of myristic, arachidic, palmitoleic, and eicosenoic acid ($p \leq 0.05$). Other fatty acid levels were not significantly different among the three breeds.
**Table 3**
| Fatty acid (%) | DLY | BDS | DBS |
| --- | --- | --- | --- |
| Capric acid (C10:0) | 0.1138 ± 0.009 | 0.19 ± 0.037 | 0.122 ± 0.017 |
| Lauric acid (C12:0) | 0.10 ± 0.00458 | 0.16 ± 0.0396 | 0.10 ± 0.0175 |
| Myristic acid (C14:0) | 1.39 ± 0.073b | 2.73 ± 0.69a | 1.46 ± 0.197b |
| Pentadecanoic acid (C15:0) | 0.037 ± 0.002 | 0.053 ± 0.01 | 0.027 ± 0.005 |
| Palmitic acid (C16:0) | 24.55 ± 0.44 | 31.09 ± 5.34 | 26.88 ± 4.00 |
| Heptadecanoic acid (C17:0) | 0.19 ± 0.02 | 0.27 ± 0.07 | 0.14 ± 0.02 |
| Stearic acid (C18:0) | 11.88 ± 0.54 | 19.52 ± 6.56 | 13.40 ± 2.36 |
| Arachidic acid (C20:0) | 0.19 ± 0.005b | 0.33 ± 0.07a | 0.20 ± 0.0357b |
| Myristoleic acid (C14:1) | 0.022 ± 0.004 | 0.044 ± 0.012 | 0.0298 ± 0.006 |
| Palmitoleic acid (C16:1) | 2.93 ± 0.23b | 5.27 ± 1.12a | 2.96 ± 0.40b |
| Elaidic acid (C18:1n9t) | 0.17 ± 0.02 | 0.22 ± 0.05 | 0.14 ± 0.03 |
| Oleic acid (C18:1n9c) | 42.95 ± 0.48 | 28.63 ± 8.54 | 33.54 ± 6.62 |
| Linoleic acid (C18:2n6) | 12.22 ± 0.68 | 6.99 ± 2.13 | 18.09 ± 8.30 |
| γ-Linoleic acid (C18:3n6) | 0.049 ± 0.009 | 0.037 ± 0.011 | 0.033 ± 0.0025 |
| Linolenic acid (C18:3n3) | 0.45 ± 0.04 | 0.57 ± 0.16 | 0.37 ± 0.04 |
| Eicosenoic acid (C20:1) | 0.70 ± 0.04b | 1.42 ± 0.30a | 0.83 ± 0.15b |
| Cis-11,14-Eicosadienoic acid (C20:2) | 0.43 ± 0.04 | 0.62 ± 0.17 | 0.39 ± 0.07 |
| Cis-8,11,14-Eicosatrienoic acid (C20:3n6) | 0.21 ± 0.03 | 0.25 ± 0.05 | 0.16 ± 0.02 |
| Cis-11,14,17-Eicosatrienoic acid (C20:3n3) | 0.074 ± 0.006 | 0.107 ± 0.039 | 0.076 ± 0.02 |
| Arachidonic acid (C20:4n6) | 1.18 ± 0.31 | 1.34 ± 0.25 | 0.96 ± 0.14 |
| cis-5,8,11,14,17-Eicosapentaenoic acid (C20:5n3) | 0.07 ± 0.01 | 0.06 ± 0.01 | 0.06 ± 0.01 |
| Cis-4,7,10,13,16,19-Docosahexaenoic acid (C22:6n3) | 0.14 ± 0.01 | 0.18 ± 0.04 | 0.10 ± 0.01 |
| Saturated fatty acid (SFA) | 38.45 ± 0.79 | 54.36 ± 6.15 | 42.32 ± 6.60 |
| Unsaturated fatty acid (USFA) | 61.55 ± 0.79 | 45.64 ± 6.15 | 57.68 ± 6.60 |
| Mono-USFA | 46.76 ± 0.67 | 35.58 ± 7.49 | 37.49 ± 6.67 |
| Poly-USFA | 14.80 ± 0.81 | 10.06 ± 1.97 | 20.19 ± 8.18 |
| USFA/SFA | 1.61 ± 0.06 | 0.97 ± 0.23 | 1.74 ± 0.54 |
## Muscle fiber diameter and muscle myosin heavy chain (MyHC) expression
Using ATPase staining, we compared the diameters of the longissimus dorsi muscle fibers of the three breeds. As shown in Figure 1, the diameters of the type I and II muscle fibers were highest in DLY, and were significantly higher in DLY and DBS than in BDS ($p \leq 0.05$).
**Figure 1:** *Muscle fiber typing and diameter measurement. (A) Frozen sections of ATPase in longissimus dorsi muscle of different hybrid pigs. Black hollow arrow indicates type I muscle fiber, red solid arrow indicates type II muscle fiber, and black line segment is measured length of muscle fiber diameter. (B) Comparison of longissimus dorsi muscle fiber diameter from different hybrid pigs. Significant differences (p < 0.05) are indicated by different letters. DLY, Duroc × (Landrace × Yorkshire); BDS, Berkshire × (Duroc × Saba); DBS, Duroc × (Berkshire × Saba).*
The relative mRNA expression levels of myosin heavy chain (MyHC) in DLY, BDS, and DBS were measured using qRT-PCR (Figure 2). Four MyHC isoforms were detected in the longissimus dorsi muscle of three pig breeds: MyHC I, MyHC IIa, MyHC IIb, and MyHC IIx. The results showed that MyHC I had the highest expression in DBS, significantly higher than in BDS and DLY ($p \leq 0.05$), and MyHC IIx was significantly higher in DBS than in DLY ($p \leq 0.05$). Furthermore, BDS had the highest expression of MyHC IIa but the lowest expression of MyHC IIb compared to DBS and DLY ($p \leq 0.05$). DLY showed the highest MyHC IIb expression ($p \leq 0.05$).
**Figure 2:** *Relative mRNA expression of longissimus dorsi muscle myosin heavy chain (MyHC) isoforms in DLY, BDS and DBS. Different superscript letters indicate significant differences (p < 0.05).*
## Glycolytic potential
The longissimus dorsi muscle GP in the three hybrid pig breeds is shown in Table 4. Glycogen content was the lowest in BDS and significantly lower than that in DLY ($p \leq 0.05$). Compared with BDS and DBS, DLY had the highest lactate content and GP value ($p \leq 0.05$).
**Table 4**
| Items | DLY | BDS | DBS |
| --- | --- | --- | --- |
| Glycogen (μmol/g) | 1.17 ± 0.03a | 0.84 ± 0.09b | 0.99 ± 0.03ab |
| Lactate (μmol/g) | 129.35 ± 1.70a | 110.55 ± 4.10b | 114.96 ± 0.14b |
| GP (μmol/g) | 131.68 ± 1.69a | 112.24 ± 4.20b | 116.94 ± 0.09b |
## Discussion
Pork quality is an important economic characteristic that affects consumers' purchase decision [19]. Crossbreeding with local pigs is important for the production of high-quality pork [20], but carcass traits in crossbred pigs may decline because of the slow growth of local breeds. For example, Jiang et al. [ 21] found that the carcass traits of Chinese Landrace × Meishan and Duroc × Landrace × Meishan pigs were significantly lower than those of foreign crossbred pigs ($p \leq 0.05$). This is consistent with our research, showing lower live weight and carcass weight in Chinese local hybrid pigs (BDS and DBS) than in foreign hybrid pigs (DLY). Our results also showed that BDS had a higher lean meat percentage and lower live weight, carcass weight, and carcass length than DBS, indicating differences in carcass traits of pig breeds obtained by different crossbreeding methods.
Studies have reported obvious differences in meat quality between hybrids [22]. Water is the main component of meat and an essential parameter for evaluating meat quality; loss of water in meat can lead to a decrease in pH, and the meat with higher drip loss is more acid and lighter [23]. We detected the highest drip loss and lowest pH in DLY pigs. Ultimate pH is a primary factor affecting meat quality in the early postmortem stage, and protein denaturation and drip loss increase as muscle pH decreases [24]. Meat color is an important trait affecting consumer purchasing behavior. Cameron et al. [ 25] showed that color lightness is negatively correlated with meat tenderness, flavor, and acceptability. Kim et al. [ 26] found that DLY pigs had higher lightness than Kagoshima Berkshire, British Berkshire (Yorkshire × Berkshire) × Berkshire, and Korean native black pig × wild boar pigs, indicating that lightness values differ in different hybrid pigs. In our study, we found that BDS had the lowest lightness values and were significantly lower than those of DBS. Thus, BDS meat quality was better than that of DLY meat in terms of drip loss, pH, and lightness, and better than DBS meat in terms of lightness.
Fatty acid composition is one of the main contributors to pork flavor and is influenced by breed and genotype [27]. Highly saturated fatty acids (SFAs) are considered to have positive effects on stabilizing fat oxidation [28]. Cameron et al. [ 27] also found that palmitoleic acid was positively correlated with pork flavor and overall acceptability. In this study, compared to DLY and DBS, BDS had the highest SFA (myristic acid and arachidic acid) and palmitoleic acid content and the best meat quality traits (drip loss, pH, and lightness), similar to earlier studies. In terms of fatty acid content for human health, long-chain mono-unsaturated fatty acids (USFAs) such as C20:1 can ameliorate obesity-related metabolic dysfunction and promote health [29]. In our study, BDS contained a higher proportion of C20:1, suggesting that BDS may produce meat that is more beneficial to human health than that of DLY and DBS. Thus, BDS was superior to DLY and DBS in terms of overall meat quality.
Muscle fiber characteristics are key factors affecting pork quality [30]. Miao et al. [ 31] reported that Luchuan pork is superior to Duroc pork in part due to its smaller muscle fiber diameter and area. Similarly, commercial broilers with larger muscle fiber diameters have poorer meat tenderness than native and hybrid chickens in China [32]. In this study, muscle fiber diameter was significantly lower in BDS pigs than in DLY and DBS pigs ($p \leq 0.05$), suggesting better meat quality. Pig muscle fibers can be categorized into four myosin isoforms: MyHC I, MyHC IIa, MyHC IIb, and MyHC IIx. It is generally believed that meat with a higher abundance of type I and type IIa fibers are of higher quality with improved oxidation ability compared to that of meat rich in type IIb fibers [33]. Type II muscle fibers grow faster, and their content is related to higher body weight; however, rapidly increasing body weight can affect the paleness and exudative properties of pork [34]. Huang et al. [ 35] showed that the relative mRNA expression level of MyHC IIb was highest in Landrace pigs at 60 days of age, while the relative mRNA expression levels of MyHC I and IIa were highest in local pigs. Similarly, in our study, the highest relative mRNA expression of MyHC in DBS and BDS of local crossbred pigs was type I and type IIa, respectively, while that of commercial crossbred pigs in DLY was type IIb, which may be an important factor regarding the meat quality of local hybrid pigs is superior to that of commercial pigs.
The storage of muscle glycogen during slaughter and the rate of glycolysis after slaughter can affect meat quality by affecting the rate and degree of postmortem pH decline [36]. Glycogen, lactate, and GP levels are negatively correlated with ultimate pH but positively correlated with drip loss [37]. In addition, Schilling et al. showed that a high level of glycolysis potential was significantly related to the decrease of meat quality, such as color and drip loss [38]. Przybylski et al. [ 23] found that chicken with deeper glycolysis and higher drip loss produced more methylglyoxal, which may change the protein properties of muscle [23]. Previous research has also shown that Chinese pure native and hybrid pigs have a lower GP and better meat quality than commercial DLY pigs [17, 39], consistent with our findings. The glycogen, lactate, and GP levels in the longissimus dorsi muscles increased in the order of BDS < DBS < DLY. Correspondingly, meat quality was the best for BDS and the worst for DLY.
## Conclusions
In this study, live weight, carcass weight, and lean meat percentage were higher in DLY than in BDS and DBS. However, drip loss, glycogen content, lactate content, GP level, and ultimate pH level were lower in BDS and DBS than in DLY. In addition, saturated fatty acid (myristic acid and arachidic acid) and monounsaturated fatty acid (palmitoleic acid and eicosenoic acid) contents were higher in BDS than in DLY and DBS. Relative mRNA expression of type I and type IIa muscle fibers, which are beneficial to meat quality, were highest in DBS and BDS, respectively ($p \leq 0.05$), while the relative mRNA expression of type IIb muscle fibers, which are negatively related to meat quality, was highest in DLY ($p \leq 0.05$). Thus, compared with commercial DLY pigs, Chinese local hybrid BDS and DBS pigs had poorer carcass traits, but better meat quality, while BDS had the best meat quality and flavor. These findings provide basic data for promoting the conservation and utilization of Saba pigs and the production of high-quality pork in the pig industry.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.
## Ethics statement
The animal study was reviewed and approved by the Institutional Animal Care and Use Committee of Yunnan Agricultural University. Written informed consent was obtained from the owners for the participation of their animals in this study.
## Author contributions
HH and HP designed the experiments and revised this manuscript. JR, HJ, XiL, XuL, YP, YiH, SZ, and CS performed the experiments. YL and YaH analyzed the data and wrote the manuscript. All the 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/fvets.2023.1136485/full#supplementary-material
## References
1. Rauw WM, Rydhmer L, Kyriazakis I, Øverland M, Gilbert H, Dekkers JC. **Prospects for sustainability of pig production in relation to climate change and novel feed resources**. *J Sci Food Agric.* (2020) **100** 3575-86. DOI: 10.1002/jsfa.10338
2. Zhang J, Chai J, Luo Z, He H, Chen L, Liu X. **Meat and nutritional quality comparison of purebred and crossbred pigs**. *Anim Sci J.* (2018) **89** 202-10. DOI: 10.1111/asj.12878
3. Clinquart A, Ellies-Oury MP, Hocquette JF, Guillier L, Santé-Lhoutellier V, Prache S. **Review: On-farm and processing factors affecting bovine carcass and meat quality**. *Animal.* (2022) **16** 100426. DOI: 10.1016/j.animal.2021.100426
4. Wang J, Ren Q, Hua L, Chen J, Zhang J, Bai H. **Comprehensive analysis of differentially expressed mRNA, lncRNA and circRNA and their ceRNA networks in the longissimus dorsi muscle of two different pig breeds**. *Int J Mol Sci.* (2019) **20** 1107. DOI: 10.3390/ijms20051107
5. Liu X, Zhou L, Xie X, Wu Z, Xiong X, Zhang Z. **Muscle glycogen level and occurrence of acid meat in commercial hybrid pigs are regulated by two low-frequency causal variants with large effects and multiple common variants with small effects**. *Genet. Sel. Evol.* (2019) **51** 46. DOI: 10.1186/s12711-019-0488-0
6. Gao J, Sun L, Zhang S, Xu J, He M, Zhang D. **Screening discriminating SNPs for Chinese indigenous pig breeds identification using a random forests algorithm**. *Genes.* (2022) **13** 2207. DOI: 10.3390/genes13122207
7. Diao S, Huang S, Chen Z, Teng J, Ma Y, Yuan X. **Genome-wide signatures of selection detection in three south China indigenous pigs**. *Genes.* (2019) **10** 346. DOI: 10.3390/genes10050346
8. Stock J, Bennewitz J, Hinrichs D, Wellmann R. **A review of genomic models for the analysis of livestock crossbred data**. *Front Genet.* (2020) **11** 568. DOI: 10.3389/fgene.2020.00568
9. Christensen OF, Legarra A, Lund MS, Su G. **Genetic evaluation for three-way crossbreeding**. *Genet. Sel. Evol.* (2015) **47** 98. DOI: 10.1186/s12711-015-0177-6
10. Li D, Huang M, Zhuang Z, Ding R, Gu T, Hong L. **Genomic analyses revealed the genetic difference and potential selection genes of growth traits in two duroc lines**. *Front Vet Sci.* (2021) **8** 725367. DOI: 10.3389/fvets.2021.725367
11. Song B, Zheng C, Zheng J, Zhang S, Zhong Y, Guo Q. **Comparisons of carcass traits, meat quality, and serum metabolome between Shaziling and Yorkshire pigs**. *Anim Nutr.* (2022) **8** 125-34. DOI: 10.1016/j.aninu.2021.06.011
12. 12.AOAC. Official Methods of Analysis, 13th Edn. Washington, DC: Association of Official Analytical Chemists (1995).. *Official Methods of Analysis, 13th Edn* (1995)
13. Choi JS, Lee HJ, Jin SK, Choi YI, Lee JJ. **Comparison of carcass characteristics and meat quality between duroc and crossbred pigs**. *Korean J Food Sci Anim Resour.* (2014) **34** 238-44. DOI: 10.5851/kosfa.2014.34.2.238
14. Folch J, Lees M, Sloane Stanley GH. **A simple method for the isolation and purification of total lipides from animal tissues**. *J Biol Chem.* (1957) **226** 497-509. DOI: 10.1016/S0021-9258(18)64849-5
15. Lee HJ, Jung EH, Lee SH, Kim JH, Lee JJ, Choi YI. **Effect of replacing pork fat with vegetable oils on quality properties of emulsion-type pork sausages**. *Korean J Food Sci Anim Resour.* (2015) **35** 130-6. DOI: 10.5851/kosfa.2015.35.1.130
16. Brooke MH, Kaiser KK. **Three “myosin adenosine triphosphatase” systems: the nature of their pH lability and sulfhydryl dependence**. *J Histochem Cytochem.* (1970) **18** 670-2. DOI: 10.1177/18.9.670
17. Luo J, Shen YL, Lei GH, Zhu PK, Jiang ZY, Bai L. **Correlation between three glycometabolic-related hormones and muscle glycolysis, as well as meat quality, in three pig breeds**. *J Sci Food Agric.* (2017) **97** 2706-13. DOI: 10.1002/jsfa.8094
18. Li Y, Li J, Zhang L, Yu C, Lin M, Gao F. **Effects of dietary energy sources on post mortem glycolysis, meat quality and muscle fibre type transformation of finishing pigs**. *PLoS ONE.* (2015) **10** e0131958. DOI: 10.1371/journal.pone.0131958
19. Lebret B, Candek-Potokar M. **Review: pork quality attributes from farm to fork. Part I carcass and fresh meat**. *Animal.* (2022) **16** 100402. DOI: 10.1016/j.animal.2021.100402
20. Kim JA, Cho ES, Jeong YD, Choi YH, Kim YS, Choi JW. **The effects of breed and gender on meat quality of Duroc, Pietrain, and their crossbred**. *J Anim Sci Technol.* (2020) **62** 409-19. DOI: 10.5187/jast.2020.62.3.409
21. Jiang YZ, Zhu L, Tang GQ, Li MZ, Jiang AA, Cen WM. **Carcass and meat quality traits of four commercial pig crossbreeds in China**. *Genet Mol Res.* (2012) **11** 4447-55. DOI: 10.4238/2012.September.19.6
22. Chen G, Sui Y. **Production, performance, slaughter characteristics, and meat quality of Ziwuling wild crossbred pigs**. *Trop Anim Health Prod.* (2018) **50** 365-72. DOI: 10.1007/s11250-017-1441-2
23. Przybylski W, Sałek P, Kozłowska L, Jaworska D, Stańczuk J. **Metabolomic analysis indicates that higher drip loss may be related to the production of methylglyoxal as a by-product of glycolysis**. *Poult Sci.* (2022) **101** 101608. DOI: 10.1016/j.psj.2021.101608
24. Offer G. **Modelling of the formation of pale, soft and exudative meat: effects of chilling regime and rate and extent of glycolysis**. *Meat Sci.* (1991) **30** 157-84. DOI: 10.1016/0309-1740(91)90005-B
25. Cameron N. **Genetic and phenotypic parameters for carcass traits, meat and eating quality traits in pigs**. *Livest Prod Sci.* (1990) **26** 119-35. DOI: 10.1016/0301-6226(90)90061-A
26. Kim I, Jin SK, Kim CW, Song Y, Cho KK, Chung KH. **The effects of pig breeds on proximate, physicochemical, cholesterol, amino acid, fatty acid and sensory properties of loins**. *J Anim Technol.* (2008) **50** 121-32. DOI: 10.5187/JAST.2008.50.1.121
27. Cameron ND, Enser M, Nute GR, Whittington FM, Penman JC, Fisken AC. **Genotype with nutrition interaction on fatty acid composition of intramuscular fat and the relationship with flavour of pig meat**. *Meat Sci.* (2000) **55** 187-95. DOI: 10.1016/S0309-1740(99)00142-4
28. Du M, Ahn DU, Sell JL. **Effects of dietary conjugated linoleic acid and linoleic:linolenic acid ratio on polyunsaturated fatty acid status in laying hens**. *Poult Sci.* (2000) **79** 1749-56. DOI: 10.1093/ps/79.12.1749
29. Yang ZH, Miyahara H, Iwasaki Y, Takeo J, Katayama M. **Dietary supplementation with long-chain monounsaturated fatty acids attenuates obesity-related metabolic dysfunction and increases expression of PPAR gamma in adipose tissue in type 2 diabetic KK-Ay mice**. *Nutr Metab.* (2013) **10** 16. DOI: 10.1186/1743-7075-10-16
30. Huo W, Weng K, Li Y, Zhang Y, Zhang Y, Xu Q. **Comparison of muscle fiber characteristics and glycolytic potential between slow- and fast-growing broilers**. *Poult Sci.* (2022) **101** 101649. DOI: 10.1016/j.psj.2021.101649
31. Miao W, Ma Z, Tang Z, Yu L, Liu S, Huang T. **Integrative ATAC-seq and RNA-seq Analysis of the Longissimus Muscle of Luchuan and Duroc Pigs**. *Front Nutr.* (2021) **8** 742672. DOI: 10.3389/fnut.2021.742672
32. Chen XD, Ma QG, Tang MY, Ji C. **Development of breast muscle and meat quality in Arbor Acres broilers, Jingxing 100 crossbred chickens and Beijing fatty chickens**. *Meat Sci.* (2007) **77** 220-7. DOI: 10.1016/j.meatsci.2007.03.008
33. Chang KC. **Key signalling factors and pathways in the molecular determination of skeletal muscle phenotype**. *Animal.* (2007) **1** 681-98. DOI: 10.1017/S1751731107702070
34. Joo ST, Kim GD, Hwang YH, Ryu YC. **Control of fresh meat quality through manipulation of muscle fiber characteristics**. *Meat Sci.* (2013) **95** 828-36. DOI: 10.1016/j.meatsci.2013.04.044
35. Huang YN, Ao QW, Jiang QY, Guo YF, Lan GQ, Jiang HS. **Comparisons of different myosin heavy chain types, AMPK, and PGC-1α gene expression in the longissimus dorsi muscles in Bama Xiang and Landrace pigs**. *Genet Mol Res.* (2016) **15** 15028379. DOI: 10.4238/gmr.15028379
36. Bee G, Biolley C, Guex G, Herzog W, Lonergan SM, Huff-Lonergan E. **Effects of available dietary carbohydrate and preslaughter treatment on glycolytic potential, protein degradation, and quality traits of pig muscles**. *J Anim Sci.* (2006) **84** 191-203. DOI: 10.2527/2006.841191x
37. Przybylski W, Sionek B, Jaworska D, Santé-Lhoutellier V. **The application of biosensors for drip loss analysis and glycolytic potential evaluation**. *Meat Sci.* (2016) **117** 7-11. DOI: 10.1016/j.meatsci.2016.02.025
38. Schilling MW, Suman SP, Zhang X, Nair MN, Desai MA, Cai K. **Proteomic approach to characterize biochemistry of meat quality defects**. *Meat Sci.* (2017) **132** 131-8. DOI: 10.1016/j.meatsci.2017.04.018
39. Shen L, Lei H, Zhang S, Li X, Li M, Jiang X. **Comparison of energy metabolism and meat quality among three pig breeds**. *Anim Sci J.* (2014) **85** 770-9. DOI: 10.1111/asj.12207
|
---
title: Pharmacokinetics and tissue distribution of Ramulus Mori (Sangzhi) alkaloids
in rats and its effects on liver enzyme activity
authors:
- Zhihua Liu
- Yu Feng
- Hang Zhao
- Jinping Hu
- Yanmin Chen
- Dongdong Liu
- Hongliang Wang
- Xiangyang Zhu
- Hongzhen Yang
- Zhufang Shen
- Xuejun Xia
- Jun Ye
- Yuling Liu
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9981942
doi: 10.3389/fphar.2023.1136772
license: CC BY 4.0
---
# Pharmacokinetics and tissue distribution of Ramulus Mori (Sangzhi) alkaloids in rats and its effects on liver enzyme activity
## Abstract
Ramulus Mori (Sangzhi) alkaloids (SZ-A) derived from twigs of mulberry (*Morus alba* L., genus Morus in the Moraceae family) was approved by the National Medical Products Administration in 2020 for the treatment of type 2 diabetes mellitus. In addition to excellent hypoglycemic effect, increasing evidence has confirmed that SZ-A exerts multiple pharmacological effects, such as protecting pancreatic ß-cell function, stimulating adiponectin expression, and alleviating hepatic steatosis. Importantly, a specific distribution of SZ-A in target tissues following oral absorption into the blood is essential for the induction of multiple pharmacological effects. However, there is a lack of studies thoroughly exploring the pharmacokinetic profiles and tissue distribution of SZ-A following oral absorption into the blood, particularly dose-linear pharmacokinetics and target tissue distribution associated with glycolipid metabolic diseases. In the present study, we systematically investigated the pharmacokinetics and tissue distribution of SZ-A and its metabolites in human and rat liver microsomes, and rat plasma, as well as its effects on the activity of hepatic cytochrome P450 enzymes (CYP450s). The results revealed that SZ-A was rapidly absorbed into the blood, exhibited linear pharmacokinetic characteristics in the dose range of 25–200 mg/kg, and was broadly distributed in glycolipid metabolism-related tissues. The highest SZ-A concentrations were observed in the kidney, liver, and aortic vessels, followed by the brown and subcutaneous adipose tissues, and the heart, spleen, lung, muscle, pancreas, and brain. Except for the trace oxidation products produced by fagomine, other phase I or phase II metabolites were not detected. SZ-A had no inhibitory or activating effects on major CYP450s. Conclusively, SZ-A is rapidly and widely distributed in target tissues, with good metabolic stability and a low risk of triggering drug-drug interactions. This study provides a framework for deciphering the material basis of the multiple pharmacological functions of SZ-A, its rational clinical use, and the expansion of its indications.
## 1 Introduction
Type 2 diabetes mellitus (T2DM) is one of the most important chronic metabolic diseases that threaten human health worldwide (Zheng et al., 2018). The pathogenesis of T2DM is complex and involves changes in various target tissues and organ functions, including impaired pancreatic islet ß-cell function and abnormal a-cell secretion (Campbell and Newgard, 2021); insulin resistance; lipid accumulation in the liver, muscle, adipose, and other tissues (Ke et al., 2022; SantaCruz-Calvo et al., 2022); impaired enteropancreatic axis function; intestinal flora imbalance; renal glucose reabsorption dysfunction (Mudaliar et al., 2015); brain tissue nerve transmitter secretion disorders (Scherer et al., 2021; Mirzadeh et al., 2022); and macrophage inflammation. Among the therapeutic agents currently used for the treatment of T2DM, glucagon-like peptide-1 agonists and dipeptidyl peptidase-4 inhibitors primarily exert their therapeutic effects by acting on the pancreatic islets, brain, and gastrointestinal tract, whereas metformin and thiazolidinediones function by reducing insulin resistance in the liver, muscles, and fat. Sodium-dependent glucose transporters two act on the kidney to inhibit glucose reabsorption, whereas a-glucosidase inhibitors primarily act on a-glucosidase in the gastrointestinal tract to delay or inhibit glucose uptake (Tahrani et al., 2016). Owing to the complex pathogenesis of T2DM, current clinical treatments recommend a multiplet strategy: therapeutic agents should not only reduce blood glucose parameters considerably but should also have substantial benefits in terms of cardiovascular safety, reduced glycemicity and prevention of T2DM complications (Perreault et al., 2021). Consequently, there is an urgent need for the continuous development of therapeutic agents with multiple pharmacological effects that contribute to the comprehensive treatment and clinical benefits of T2DM.
Ramulus Mori (Sangzhi) alkaloids (SZ-A) are a group of polyhydroxy alkaloids extracted and isolated from the traditional Chinese medicine mulberry twig (*Morus alba* L.), accounting for more than $50\%$ of the total mulberry twig extract, which is majorly composed of 1-deoxynojirimycin (DNJ); fagomine (FA); and 1,4-dideoxy-1,4-imino-D-arabinitol (DAB) (Jin et al., 2022). Among these three components, DNJ contains the highest alkaloid content, accounting for >$50\%$ of the total alkaloids. The sum of the three main components accounts for more than $80\%$ of the total alkaloid content. As a natural hypoglycemic drug, SZ-A tablets were approved by the China National Medical Products Administration (NMPA) in 2020 for the treatment of T2DM (approval number Z20200002). The results of a randomized double-blind phase III clinical trial with the chemical drug acarbose tablets (Glucobay®) as the positive control (No. CTR20140034) revealed that SZ-A could effectively reduce glycated hemoglobin levels in patients and the hypoglycemic efficacy of SZ-A was comparable to that of acarbose during 24 weeks of treatment (Qu et al., 2021). Importantly, the incidence of drug-related adverse events and gastrointestinal adverse reactions induced by SZ-A was significantly lower than that induced by acarbose, indicating that SZ-A exhibited superior safety.
α-glucosidase inhibition was one of the hypoglycemic mechanisms discovered earlier for SZ-A. SZ-A is more selective for glycosidase species than acarbose and primarily acts on disaccharidase, with almost no amylase inhibition (Hiele et al., 1992; Liu et al., 2019). This pharmacological feature is conducive to faster and better control of postprandial blood glucose, while greatly reducing adverse reactions such as flatulence and exhaust in East Asian patients consuming carbohydrates as their staple food (Nojima et al., 1998; Liu et al., 2021). In addition to selectively inhibiting intestinal a-glucosidase, an in-depth study of the mechanism underlying the effects of SZ-A found that exerts multiple in vitro and in vivo pharmacological effects on the regulation of pancreatic islet ß-cell function, insulin resistance, lipid metabolism, intestinal flora, and inflammation (Cao et al., 2021; Liu et al., 2021; Chen et al., 2022; Lei et al., 2022; Sun et al., 2022). In a previous study, the therapeutic effects and mechanism of action of SZ-A on lipid metabolism were explored in high-fat diet (HFD)-induced obesity and non-alcoholic fatty liver disease (NAFLD) mice. We found that orally administered SZ-A ameliorated HFD-induced weight gain and significantly stimulated adiponectin expression and secretion in the adipose tissue. Additionally, SZ-A markedly reduced hepatic steatosis and regulated lipid metabolism and oxidative stress in the liver. To further investigate the target tissues that mediate the protective effect of SZ-A against obesity and NAFLD, SZ-A was intraperitoneally administered to HFD-induced mice. Notably, HFD-induced obesity, hepatic steatosis, oxidative stress, infiammation, and fibrosis in mice were also ameliorated by intraperitoneal administration of SZ-A. These results indicated that intestinal a-glucosidase was not the only target of SZ-A, and SZ-A might also be absorbed into the blood following oral administration and widely distributed in T2DM-related target tissues, such as the pancreatic islets, liver, muscle, and adipose tissue (Schwartz et al., 2017), and then act directly on these tissues to exert multiple pharmacological effects. This pharmacological feature of SZ-A differs remarkably from that of the traditional glucosidase inhibitor acarbose, which primarily functions in the gastrointestinal tract and is barely absorbed into the blood (Krentz and Bailey, 2005).
An important pre-requisite for the multiple pharmacological effects of SZ-A on T2DM-related tissues is the specific distribution of SZ-A in the target tissues following oral absorption into the blood. Although multiple pharmacological effects of SZ-A have been confirmed in previous studies, the knowledge regarding its distribution in T2DM-related tissues (including muscle, adipose tissue, etc.,) is still unclear. Preliminary pharmacokinetic results following SZ-A administration in rats showed that the absolute bioavailability of the three active components of SZ-A (DNJ, FA, and DAB) was >$70\%$, implying widespread tissue distribution (Yang et al., 2017). However, there is a lack of studies thoroughly exploring the pharmacokinetic profiles and tissue distribution with regard to its oral absorption into the blood especially dose-linear pharmacokinetics and target tissue distribution associated with glycolipid metabolic diseases. Additionally, it remains unclear whether SZ-A is metabolized by the liver and affects the activity of liver microsomal isozymes. The safety of a drug is closely related to its metabolism in the liver, its effects on the activity of liver enzymes, and drug-drug interactions.
To fully understand the tissue distribution characteristics of SZ-A following oral administration, the distribution of the three active components of SZ-A (DNJ, FA, and DAB) in 12 tissues, including the heart, liver, spleen, lung, kidney, aortic blood vessels, brown adipose tissue, subcutaneous adipose tissue, abdominal adipose tissue, muscle, pancreas, and brain, was determined in mice and rats (Scheme 1). Metabolites in human/rat liver microsomes (HLMs/RLMs) or rat plasma were identified to evaluate whether SZ-A was metabolized by the liver. Furthermore, the effect of SZ-A on cytochrome P450 enzymes (CYP450s) activity was investigated to evaluate the possibility of SZ-A-induced drug-drug interactions. A better understanding of the tissue distribution characteristics of SZ-A is beneficial to further explore its multiple pharmacological effects and mechanisms and provide a material foundation for the subsequent expansion of new therapeutic indications.
**SCHEME 1:** *Schematic illustration of the pharmacokinetics and tissue distribution of SZ-A in rats after oral administration. SZ-A is rapidly absorbed into the blood and widely distributed in the type 2 diabetes mellitus (T2DM)-related tissues after oral administration, which lays the material basis for its multiple pharmacological effects in vivo such as stimulating the secretion of adiponectin, protecting pancreatic ß-cell function, improving insulin resistance, and reducing the secretion of inflammatory cytokines.*
## 2.1 Materials
DNJ (purity >$99.0\%$) and SZ-A extract (Lot No.: J202004007, containing $36.88\%$ of DNJ, $8.78\%$ of FA, and $5.83\%$ of DAB; Lot No.: J202108007, containing $36.52\%$ of DNJ, $9.60\%$ of FA, and $7.62\%$ of DAB) were provided by Beijing Wehand-bio Pharmaceutical Co. Ltd. (Beijing, China). The multiple reaction monitoring (MRM) chromatogram of SZ-A is shown in Supplementary Figure S1. Miglitol was obtained from TCI Shanghai Chemical Industrial Development Co., Ltd. (Shanghai, China). FA (purity >$98.0\%$) was purchased from MedChemExpress (Monmouth Junction, NJ, United States). DAB (purity >$98.0\%$) was purchased from Sigma-Aldrich (St. Louis, MO, United States). Human liver microsomes (HLMs) were purchased from Reid Liver Disease Research (Shanghai, China). Glucose-6-phosphate, oxidized coenzyme H (β-NADP), glucose-6-phosphate dehydrogenase, midazolam, phenacetin, dextromethorphan, mephenytoin, chlorzoxazone, diclofenac sodium, 1-Hydroxy-Midazolam, 4-hydroxy-mephenytoin, acetaminophen, 4-Hydroxy-Diclofenac Sodium, demethyldextromethorphan, 6-Hydroxy-Chlorzoxazone, furaphylline, sulfafenpyrazole, quinidine, ketoconazole, and sodium diethyldithiacarbamate were purchased from Sigma-Aldrich (St. Louis, MO, United States). All other organic reagents were of analytical grade and purchased from Sinopharm Chemical Reagent (Shanghai, China).
## 2.2 Animals
For the pharmacokinetic study, Sprague-Dawley rats (SD, 180–220 g, both male and female) were purchased from Speifu (Beijing) Biotechnology Co., Ltd. (Beijing, China). Animal experiments were approved by the Ethics Committee of Kangtai Medical Laboratory Service Hebei Co., Ltd. (Hebei, China) (No. MDL 2022-01-17-1).
For the tissue distribution study, male SD rats (180–200 g) and male ICR mice (20–25 g) were purchased from Speifu (Beijing) Biotechnology Co., Ltd. (Beijing, China). The ambient temperature and humidity in the laboratory were maintained at approximately 22°C and $50\%$, respectively. Animal experiments were approved by the Ethics Committee of Zhongsheng Beidong (Beijing) Technology Development Co., Ltd. (Beijing, China) (No. 20200039YFE-3R and 20200084YZE-3R).
## 2.3 Pharmacokinetic study
Twenty-four SD rats were fasted for 12 h (free access to water) and randomly classified into three groups ($$n = 8$$; half male and female) that received different doses of SZ-A solution (25, 50, and 200 mg/kg) via oral administration. After completion of dose administration, blood samples (0.3 mL) were collected through an indwelling catheter from the jugular vein at predetermined time points (5, 15, 30, and 45 min, 1, 2, 4, 8, and 24 h). The blood was centrifuged at 5,000 rpm for 10 min to separate the plasma. All plasma samples were stored frozen at −80°C until liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) analysis (AB SCIEX Triple Quad™ 4,500 mass spectrometer, Applied Biosystems Inc., United States).
## 2.4 Tissue distribution
Twenty SD rats were randomly classified into four groups corresponding to four-time points (5 and 30 min, 4 and 12 h) ($$n = 20$$; 5 animals per time point) that received SZ-A solution (40 mg/kg) via oral administration. After the completion of dose administration, plasma and tissue samples (derived from the heart, liver, spleen, lung, kidney, aortic blood vessels, brown adipose tissue, subcutaneous adipose tissue, abdominal adipose tissue, muscle (thigh skeletal muscle), pancreas, and brain) were collected at predetermined time points. Twenty-eight ICR mice were randomly classified into four groups ($$n = 7$$ per group) corresponding to four-time points (5 and 30 min, 4 and 12 h) that received SZ-A solution (20 mg/kg) via oral administration. After the completion of dose administration, tissue samples (derived from the heart, liver, spleen, lung, kidney, aortic blood vessels, brown adipose tissue, abdominal adipose tissue, muscle (thigh skeletal muscle), pancreas, and brain) were collected at predetermined time points.
Collected tissue samples were washed with normal saline. The surface moisture was removed by blotting gently with filter paper, and the weight was recorded. After adding an appropriate amount of normal saline, these tissue samples were ground into homogenates using a cryogenic grinder (JX-CL, Shanghai Jingxin, Shanghai, China). The obtained homogenates were stored at −80°C until LC-MS/MS analysis.
## 2.5 Inhibition of CYP450s
The experiments were classified into six groups ($$n = 3$$): control (no inhibitor), positive inhibitor, SZ-A (final concentration, 8.899 μg/mL; equivalent to 20 μM DNJ), DNJ (final concentration, 20 μM), FA (final concentration, 5 μM), and DAB (final concentration, 8 μM). Different inhibitors were added to the HLMs/RLMs and incubated with the probe substrates of each subtype.
The liver microsomal incubation system comprised human or rat liver microsomal protein (0.5 mg/mL), nicotinamide adenine dinucleotide phosphate (NADPH) generation system, Tris-HCl buffer (50 mM, pH 7.4), substrate, and inhibitor, with a total reaction volume of 200 μL. The substrate concentrations and incubation times for CYP1A2, 2D$\frac{6}{2}$D2, 2C$\frac{9}{2}$C6, 2C$\frac{19}{2}$C11, 3A$\frac{4}{3}$A2, and 2E1 were phenacetins, 50 μM/30 min; dextromethorphan, 5 μM/10 min; diclofenac sodium, 20 μM/10 min; mephenytoin, 40 μM/30 min; midazolam, 5 μM/5 min; and chlorzoxazone, 80 μM/20 min, respectively. The concentrations of each isoenzyme-positive inhibitor were 10 μM of furafylline (50 μM in the rat group), 2 μM quinidine (25 μM in the rat group), 5 μM sulfaphenazole (50 μM in the rat group), 5 μM of diclofenac pyridine (25 μM in the rat group), 1 μM ketoconazole, and 80 μM sodium diethyldithiacarbamate (100 μM in the rat group). After the reaction was completed, 400 μL of the internal standard propranolol (final concentration, 200 ng/mL) in ice-cold acetonitrile was added to terminate the reaction, and the sample was mixed well and centrifuged twice at 14,000 rpm for 5 min. Five microliters of the supernatant were collected for LC-MS/MS analysis (UPLC-Q-Extractive Orbitrap MS; Thermo Fisher Scientific, Waltham, MA, United States) to determine the metabolite content of the probe substrate.
## 2.6 Identification of metabolites
The metabolites in the 6 h plasma samples from the pharmacokinetic study and the HLMs/RLMs reaction system in the SZ-A group from the CYP450s inhibition study were identified using Q-Extractive high-resolution MS.
## 2.7 LC-MS/MS analysis
The concentration of SZ-A in plasma and tissue was determined using LC-MS/MS analysis. Two hundred microlitres of miglitol solution (internal standard, 100 ng/mL, dissolved in acetonitrile-methanol solution (5:2, v/v, $0.1\%$ formic acid)) and 10 µL methanol-water (4:1, v/v) were added to 50 µL of plasma or tissue homogenate, mixed by vortexing for 2 min, and centrifuged at 14,000 rpm for 10 min. One hundred microliters of the supernatant were mixed with 100 µL acetonitrile-water (3:1, v/v, $0.1\%$ formic acid), vortexed for 1 min, and centrifuged at 14,000 rpm for 5 min. Five microliters of the supernatant were injected into the LC-MS/MS system. The analysis was performed using an LC-MS/MS system comprising an Exion LC AC HPLC system and a Triple Quad 4500 MS system. The three active components of SZ-A (DNJ, FA, and DAB) and miglitol were separated by using a chromatographic column (XBridge™ Amide, 3.5 µm, 4.6 × 150 mm) at 35°C. Mobile phases A and B contained $0.1\%$ ammonia in water and acetonitrile, respectively. A gradient elution program was used with a flow rate of 0.5, 0–4 min for $65\%$ mobile phase B, 4–7 min for $43\%$ mobile phase B, and 7–13 min for $65\%$ mobile phase B. The injection volume was 5 µL. The mass spectrometer was equipped with an electrospray ionization source and scanned in the multiple reaction monitoring mode and positive ion mode. The main parameters were as follows: curtain gas (CUR): 35.0 psi, collision gas (CAD): 9, ion spray voltage (IS): 5500V, temperature (TEM): 500°C, ion source gas1 (GS1): 55.0 psi, ion source gas2 (GS2): 50.0 psi, entrance potential (EP): 10.0 eV; collision cell exit potential (CXP): 6.0 eV. The ion reaction pairs used for monitoring were as follows: DNJ, m/z 164.3→69.0 (DP 70ev, CE 27ev); FA, m/z 148.2→112.1 (DP 56ev, CE 18ev); DAB, m/z 134.3→68.1 (DP 55ev, CE 25ev); and miglitol, m/z m/z 208.3→146.1 (DP 70ev, CE 27ev).
For the CYP450s inhibition study, LC-MS/MS detection conditions were as follows: chromatographic column, Zorbax C18 column (3.5 μm, 2.1 × 100 mm); mobile phase, acetonitrile ($0.1\%$ formic acid) and water ($0.1\%$ formic acid); gradient elution; flow rate, 0.2 mL/min; ionization mode, CYP2E1 was negative; MS conditions, acetaminophen (m/z: 152→110), 4-hydroxymephenytoin (m/z:235→150), desmethyldextromethorphan (m/z: 258→157), 4-hydroxydiclofenac (m/z:312→230), and 1-hydroxymidazolam (m/z: 342→203) in positive ion mode and 6-hydroxychlorzoxazone (m/z:184→120) in negative ion mode. The scan parameters were as follows: resolution, 70,000; AGC target, 200,000; and maximum IT, 100 ms (enzyme effect).
For the metabolite identification study, rat plasma or liver microsomes were precipitated with acetonitrile 1:4 (v:v), vortexed for 60 s, and centrifuged twice at 14,000 rpm for 5 min. Furthermore, 5 μL of the supernatant was collected for LC-MS/MS analysis. The LC-MS/MS conditions were as follows: chromatographic column (XBridge™ Amide, 3.5 µm, 4.6 × 150 mm); mobile phase, water ($0.1\%$ ammonia) and acetonitrile, other conditions are the same as those in CYP450s inhibition study, with the following scan parameters: resolution, 70,000; AGC target, 3,000,000; maximum IT, 200 ms; and scan range, 50–600 m/z.
## 2.8 Pharmacokinetic parameters analysis
Primary pharmacokinetic parameters, including Cmax, Tmax, area under the plasma concentration-time curve (AUC), t$\frac{1}{2}$, clearance rate, and the apparent volume of distribution, were calculated using DAS 3.2.8 (Bio Guider Co., Shanghai, China).
## 2.9 Statistical analysis
Statistical analysis was performed using a Student’s t-test for two groups or one-way analysis of variance (ANOVA) for more than two groups using GraphPad Prism version 7.00 for Windows (GraphPad Software, La Jolla, CA, United States). Differences were considered statistically significant if the p-value was less than 0.05.
## 3.1 Method validation of LC-MS/MS
Systematic validation of all LC-MS/MS analytical methods included in this study was performed to meet the acceptance criteria recommended by the Food and Drug Administration (FDA) guidelines for Industry Bioanalytical Method Validation. No obvious endogenous interference was found at the retention times of DNJ, FA, DAB, and miglitol (internal standard) in the blank plasma or plasma samples. Representative chromatograms of blank rat plasma, blank plasma sampleS spiked with DNJ, FA, DAB, and miglitol, as well as the plasma sample from the SZ-A-administered rats, are shown in Figure 1.
**FIGURE 1:** *Typical multiple reaction monitoring (MRM) chromatograms for (A) blank plasma (B) blank plasma spiked with DNJ, FA, DAB, and miglitol, and (C) plasma samples after oral administration of SZ-A. The peak at 5.495, 5.727, 6.139, and 6.676 min were miglitol, DNJ, FA, and DAB, respectively.*
The calibration curves for DNJ, FA, and DAB in rat plasma were linear over a concentration range of 50–5,000 ng/mL for DNJ, 25–2,500 ng/mL for FA, and 25–2,500 ng/mL for DAB. The corresponding linear regression equation with a 1/x2 weighting factor was $y = 1.85020$e-4x + 0.00191 ($r = 0.9997$) for DNJ, $y = 4.41781$e-4 x + 4.01643e-4 ($r = 0.9996$) for FA, and $y = 0.00123$ x + 0.00188 ($r = 0.9990$) for DAB, where y represents the ratio of the analyte peak area to the miglitol area, and x is the analyte concentration. The other linear regression equations for the calibration curves of DNJ, FA, and DAB in rat tissues are presented in Table S1. The correlation coefficients (r) were >0.99 for all calibration curves. The average accuracy of the calibration standards for all three analytes was between $85.47\%$–$113.33\%$, with a relative standard deviation (RSD) of $1.43\%$–$13.57\%$. Additionally, the results of the intra- and inter-batch accuracy and precision, matrix effect, extraction recovery, and stability met the acceptance criteria recommended by the FDA guidelines, demonstrating that the developed LC-MS/MS analytical methods can be used to accurately determine the concentrations of DNJ, FA, and DAB in plasma and tissues.
## 3.2 Pharmacokinetic study of SZ-A in rats
The plasma concentration-time curves and main pharmacokinetic parameters after administration of different doses of SZ-A (25, 50, and 200 mg/kg) in SD rats are shown in Figure 2 and Supplementary Table S2. As shown in Figure 2, the three active components of SZ-A, DNJ, FA, and DAB, were detected in plasma 5 min after oral administration and reached their maximum plasma concentrations (Tmax) within 0.88 h, indicating the rapid absorption of the three alkaloids from the gastrointestinal tract. The plasma drug concentration increased with the dose with similar Tmax regardless of the dose groups. For the oral administration dose of 50 mg/kg SZ-A, the half-life (t$\frac{1}{2}$) of DNJ, FA, and DAB in plasma were 1.07 ± 0.04, 1.19 ± 0.03, and 1.20 ± 0.03 h, respectively. The volume of distribution (Vz) of DNJ, FA, and DAB in plasma was 4.47 ± 1.30, 6.88 ± 1.85, and 4.70 ± 0.65 L/kg, respectively. The short half-life and large volume of distribution demonstrated that the three alkaloids were rapidly eliminated from the blood and transported to tissues or organs. Additionally, the oral absolute bioavailability of DNJ, FA, and DAB was $72.41\%$, $77.50\%$, and $78.23\%$, respectively, after oral administration of SZ-A in rats (Yang et al., 2017). Taken together, the short half-life, large volume of distribution, and high oral absolute bioavailability of DNJ, FA, and DAB laid the foundation for further study of their dynamic tissue distribution in vivo.
**FIGURE 2:** *Experimental scheme of plasma concentration determination after oral administration of SZ-A (25, 50, and 200 mg/kg) in rats (A). Mean plasma concentration-time profiles of SZ-A (B), DNJ (C), FA (D), and DAB (E) in rats following oral administration of SZ-A. Each value represents the mean ±standard error of mean (SEM) (n = 8).*
As shown in Figure 3, an obvious linear relationship was observed between Cmax and the SZ-A dose and between AUC(0-t) and SZ-A dose. All the correlation coefficient values (r 2) for AUC(0-t) versus dose and *Cmax versus* dose were above 0.9, demonstrating that Cmax and AUC(0-t) have strong positive correlations with the SZ-A dose. Additionally, for DNJ, no significant differences were observed in t$\frac{1}{2}$ and CLz between the three doses. These results indicated that the SZ-A exhibited a linear pharmacokinetic profile when administered orally at a dose range of 25–200 mg/kg.
**FIGURE 3:** *The linear relation between Cmax and the SZ-A dose (A,B) and between AUC and the SZ-A dose (C,D). Each value represents the mean ± standard error of mean (SEM) (n = 8).*
These results are consistent with our previous experimental findings, such as rapid absorption of the three alkaloids from the gastrointestinal tract, fast elimination from the blood, and rapid distribution into tissues (Yang et al., 2017). However, what we found in a preliminary study of DNJ, FA, and DAB exhibited non-linear pharmacokinetics following oral administration of SZ-A at the investigated dosage in rats (20–500 mg/kg), which is contrary to our current findings (Yang et al., 2017). According to the above results, SZ-A presented a linear pharmacokinetic profile in the range of 25–200 mg/kg and non-linear pharmacokinetics at higher doses of 200–500 mg/kg following oral administration in rats.
## 3.3 Tissue distribution of SZ-A in rats
In a previous study, we found that SZ-A could exert multiple pharmacological effects on T2DM-related target tissues, including adipose tissue and the liver, by oral or intraperitoneal administration (Chen et al., 2022). For example, SZ-A ameliorated HFD-induced weight gain and significantly stimulated adiponectin expression and secretion in adipose tissue (Sun et al., 2022). Additionally, SZ-A alleviates hepatic steatosis and regulates lipid metabolism and oxidative stress in the liver (Chen et al., 2022). These results suggested that intestinal a-glucosidase was not the only target of SZ-A and that SZ-A might be widely distributed in T2DM-related target tissues after oral administration, followed by direct action on these tissues to exert multiple pharmacological effects. To fully investigate the tissue distribution characteristics of SZ-A after oral administration, the concentrations of the three active components of SZ-A in plasma and twelve tissues including all T2DM-related target tissues, were determined.
The mean concentrations of the three active components of SZ-A (DNJ, FA, and DAB) in plasma and tissues, including the heart, liver, spleen, lung, kidney, aortic blood vessels, brown adipose tissue, subcutaneous adipose tissue, abdominal adipose tissue, muscle, pancreas, and brain, after a single oral administration of SZ-A (40 mg/kg) in SD rats are shown in Supplementary Figure S2 and Figure 4. As shown in Figure 4, the distribution patterns of DNJ, FA, and DAB over time in each tissue at 5, 30 min, 4 h, and 12 h were consistent. The three alkaloids were rapidly and widely distributed in various tissues after the oral administration of SZ-A, and the highest concentrations of DNJ, FA, and DAB were detected in the kidney, followed by the liver, aortic vessels, and brown adipose tissue. These tissues are most closely associated with weight loss and glucose lowering (Cypess et al., 2014; Baskin et al., 2015; Kajimura et al., 2015). The distribution rates of the three alkaloids in the same tissue were approximately the same. Of all the tissues, distribution was fastest in the heart and aortic vessels. The concentration of DNJ peaked 5 min after administration in the heart and aortic vessels and 30 min after administration in other tissues, and the peak time in the brain was the longest (4 h). FA reached the maximum drug concentration in the heart, muscle, lung, and aortic vessels at 5 min and other tissues at 30 min. DAB was most rapidly distributed in the aortic vessel within 5 min and the other tissues at 30 min. After 12 h of oral administration, all three alkaloids were rapidly eliminated from most of the tissues. Supplementary Table S3 summarizes the main pharmacokinetic parameters. The results showed that the half-life (t$\frac{1}{2}$) of SZ-A in the liver, kidney, pancreas, muscle, and brain was approximately 2 h, whereas that in other tissues was less than 1 h. The clearance rate (CLz) of SZ-A in the kidney was lower than 0.5 L/h/kg, while the clearance rate in abdominal adipose tissue, heart, spleen, and lung was higher than 10 L/h/kg. These results demonstrate that the three main alkaloids of SZ-A can be rapidly distributed in various tissues after oral administration, which may be due to its high oral bioavailability.
**FIGURE 4:** *Experimental scheme of tissue concentration determination after oral administration of SZ-A in rats (A). The concentration of the three active components of SZ-A, including DNJ (B), FA (C), and DAB (D), in tissues, including heart, liver, spleen, lung, kidney, aortic blood vessel, brown adipose tissue, subcutaneous adipose tissue, abdominal adipose tissue, muscle, pancreas, and brain, after a single oral administration of SZ-A (40 mg/kg) in SD rats. Each value represents the mean ± standard error of mean (SEM) (n = 5).*
The study of tissue distribution is an important part of the pharmacokinetic study. The in vivo distribution characteristics are the basis for evaluating the pharmacology and toxicity of drugs. Typically, multiple pharmacological effects of drugs are closely related to their wide tissue distribution. After oral administration, SZ-A underwent rapid distribution into tissues within the time course examined, and no long-term accumulation of the three alkaloids DNJ, FA, and DAB in tissues were observed. The major tissue distributions of DNJ, FA, and DAB were similar. The highest tissue concentrations of all three alkaloids were observed in the kidney, followed by the liver, aortic vessels, and brown adipose tissue.
Studies have shown that DNJ can normalize renal function in diabetic rats, suggesting renal protection against diabetes (Huang et al., 2014). The distribution of SZ-A in the kidney supports its direct renoprotective effect. Coronary atherosclerotic heart disease is a typical macrovascular complication of diabetes (DeFronzo et al., 2015), and studies have found that DNJ has pleiotropic effects on the development of atherosclerosis by inhibiting glucose-stimulated vascular smooth muscle cell migration through activating AMP-activated Protein Kinase (AMPK)/RhoB and downregulating focal adhesion kinase (FAK) (Chan et al., 2013). The high concentration of SZ-A distributed in the aortic vessels has positive significance for the treatment of atherosclerosis complicated by diabetes. The distribution of SZ-A in the liver is vital for alleviating hepatic steatosis and regulating fatty acid metabolism, lipid accumulation, and oxidative stress, thereby alleviating non-alcoholic fatty liver disease and reducing body weight in high-fat diet-induced obese mice (Chen et al., 2022). DNJ participates in the glucose transport system by directly regulating the protein expression of glycolytic and gluconeogenic enzymes, which inhibit intestinal glucose absorption and accelerate hepatic glucose metabolism (Li et al., 2013). Notably, SZ-A also has a high drug distribution in brown adipose tissue, abdominal adipose tissue, and subcutaneous adipose tissue, which is of great significance in reducing insulin resistance, regulating adipokine secretion, and controlling body weight. Our previous study showed that SZ-A could improve lipid metabolism and inhibit weight gain in HFD-induced obese mice by inhibiting fatty acid synthase and increasing lipolytic enzyme expression to inhibit fat accumulation. Additionally, SZ-A has been found to have a beneficial effect on obesity-induced chronic inflammation in adipose tissue (Sun et al., 2022). All the current studies have shown that SZ-A plays a role in adipose tissue, and our discovery of the distribution of the main components of SZ-A in adipose tissue provides a foundation for further research on its mechanism in regulating adipose tissue.
This study also found that SZ-A has a certain drug distribution in T2DM-related tissues, such as the pancreas, muscle, spleen, lung, and heart, and can pass through the blood-brain barrier and become rapidly distributed in the brain tissue. Lipid sensing and insulin signaling in the brain independently trigger negative feedback systems that can effectively reduce glucose production and food intake to restore metabolic homeostasis in T2DM and obesity (Yue and Lam, 2012). The function of the gut-brain axis is often attributed to gastrointestinal hormones, and gut gluconeogenesis plays an important role in the regulation of energy homeostasis in the brain. An increasing number of studies have shown that hyperglycemia is associated with cognitive decline and that defective insulin signaling may increase the risk of Alzheimer’s disease (Strachan et al., 2011; Lupaescu et al., 2022). These studies suggest that SZ-A has the possible potential to treat brain aging and neuroinflammation.
## 3.4 Tissue distribution of SZ-A in mice
The concentrations of the three active components of SZ-A (DNJ, FA, and DAB) in the heart, liver, spleen, lung, kidney, aortic blood vessels, brown adipose tissue, abdominal adipose tissue, muscle, pancreas, and brain after a single oral administration of SZ-A (20 mg/kg) in ICR mice are shown in Figure 5. As shown in Figure 5, the distribution patterns of DNJ, FA, and DAB over time in each tissue at 5, 30 min, 4 h, and 12 h were consistent. The three alkaloids were rapidly and widely distributed in various tissues after oral administration of SZ-A, and the highest concentrations of DNJ, FA, and DAB were detected in the kidney, followed by the liver. In contrast to the tissue distribution characteristics of rats, the drug distribution to the tissues of mice was slower than that of rats because the drug concentration in each tissue of mice at 5 min was much lower than that at 30 min. The distribution of DNJ and DAB in aortic vessels was lower than that in other tissues, and the distribution of DAB in the pancreas, brown adipose tissue, abdominal adipose tissue, heart, and spleen was similar but higher than that in aortic vessels. The distribution of FA in the pancreas, abdominal adipose tissue, heart, spleen, and aortic vessels was similar. These results indicate that SZ-A can be rapidly distributed to the tissues of mice, and the tissue distribution characteristics are similar to those of rats, demonstrating that the distribution characteristics of SZ-A in different species of animals show little difference.
**FIGURE 5:** *Experimental scheme of tissue concentration determination after oral administration of SZ-A in mice (A). The concentration of the three active components of SZ-A, including DNJ (B), FA (C), and DAB (D), in tissues, including heart, liver, spleen, lung, kidney, aortic blood vessels, brown adipose tissue, abdominal adipose tissue, muscle, pancreas, and brain, after a single oral administration of SZ-A (20 mg/kg) in ICR mice. Each value represents the mean ± standard error of mean (SEM) (n = 7).*
## 3.5 Inhibitory effects of CYP450s on HLMs/RLMs
Liver microsome assays used positive inhibitors such as quinidine, ketoconazole, sulfaphenazole, ticlopidine, furafylline, and sodium diethyldithiacarbamate, which have inhibitory effects on CYP450s in HLMs. The metabolic inhibition rates of probe substrates were $84.19\%$, $90.44\%$, $84.04\%$, $68.46\%$, $71.84\%$, and $53.09\%$, for CYP2D6, 3A4, 2C9, 2C19, 1A2, and 2E1 from HLMs, respectively, and $56.56\%$, $76.37\%$, $88.00\%$, $65.17\%$, $66.33\%$, and $46.91\%$ for these enzymes from RLMs, respectively. These results also indicate that the incubation system used in this experiment is sensitive and reliable.
The effect of SZ-A on the activity of CYP450s in HLMs and RLMs is shown in Figure 6. SZ-A (final concentration 8.899 μg/mL, equivalent to DNJ at 20 μM), DNJ (final concentration, 20 μM), FA (final concentration, 5 μM), and DAB (final concentration, 8 μM) had no significant effect on the major CYP450s in HLMs, either inhibition or activation. They had a mild activation effect on rat CYP2C11 but no obvious effect on other isoenzymes.
**FIGURE 6:** *The effect of SZ-A on the activity of CYP450s in human liver microsomes (HLMs, (A)) and rat liver microsomes (RLMs, (B)). Each value represents the mean ± standard error of mean (SEM) (n = 3). *p < 0.05, **p < 0.01 compared with the control group.*
CYP450s play a leading role in all kinds of enzymes involved in drug metabolism (Zhao et al., 2021). The inhibition or induction of its activity is the main reason for metabolic drug interactions. Enzyme inhibition is more clinically significant than enzyme induction, accounting for approximately $70\%$ of metabolic interactions. No obvious inhibition or activation of CYP2D6, 3A4, 2C9, 2C19, 1A2, and 2E1 isoenzyme activities were found in the SZ-A and human liver microsome incubation system. SZ-A has a mild activating effect on rat CYP2C11 but has no significant effect on other isoenzymes. CYP2C11 is a rat protein corresponding to human CYP2C9, whereas SZ-A has no significant effect on human liver microsome CYP2C9, which might be due to species differences. This study showed that SZ-A is less likely to exhibit drug-drug interactions based on CYP450s in clinical medication, which also provides an important reference for further research on the concomitant administration of SZ-A, expanding its clinical applications.
## 3.6 Identification of metabolites
With the help of Q Extractive high-resolution MS, a full scan was performed on SZ-A and the three main active components in HLMs/RLMs samples incubated at body temperature and in rat plasma. The metabolites were analyzed using chromatographic retention time and accurate molecular weight. Except for trace amounts of FA oxidation products, no other related metabolites were found in HLMs/RLMs and rat plasma. The quasi-molecular ion peak of FA was (M + H)+ m/z 148.09682, and its molecular formula was C6H13NO3. Its metabolite quasi-molecular ion peak was (M + H)+ m/z 164.09173 and its molecular formula was C6H13NO4. Compared to the parent drug FA, the quasi-molecular ion of the FA oxidation metabolite increased by one oxygen atom, and the relative molecular mass increased by 16 u. The position of nitrogen could be a possible active site, and thus, it was speculated that there might be two types of FA oxidation metabolite structures: one that opens the ring at the nitrogen position to form COHCH2(CHOH)2CH(CH2OH)NH2 and the other that is not ring-opened and oxidized on the nitrogen atom (Figure 7). The lack of structural confirmation of oxidation metabolites of FA is a limitation of the present study. The main components of SZ-A, DNJ, FA, and DAB exist mainly in the parent form, whether in the liver microsome incubation system or after oral absorption into the blood, indicating that SZ-A has excellent metabolic stability.
**FIGURE 7:** *Possible oxidation pathway of FA (A) and its metabolites (B) and (C). The metabolites of SZ-A were identified from the plasma samples of rats after oral administration of SZ-A (40 mg/kg) and the human liver microsomes (HLMs) and rat liver microsomes (RLMs) samples in the enzyme inhibition study.*
Drug metabolism in the body is closely related to its efficacy and safety. There was no detectable phase I and phase II metabolites of DNJ or DAB found in HLMs/RLMs and rat plasma reaction system, whereas only trace oxidation products were detected for FA. The position of nitrogen in the FA molecular structure is a possible activation site, and it is speculated that there might be two types of product structures after the oxidation reaction at this site; however, neither has an obvious toxic structure. According to another report, prototypes DNJ, FA, and DAB reached material conservation within 24 h after the oral administration of SZ-A, suggesting that they were mainly excreted in their original forms (Yang et al., 2017). Kiyotaka et al. did not detect any DNJ metabolites in the plasma of rats after oral gavage of DNJ (Nakagawa et al., 2007). Furthermore, DAB is mainly excreted unchanged in rats (Mackay et al., 2003). These studies suggest that SZ-A is not metabolized by the liver after being absorbed into the blood through the gastrointestinal tract, and has good metabolic stability.
## 4 Conclusion
SZ-A was rapidly absorbed into the blood following oral administration and showed linear pharmacokinetics in a dose range of 25–200 mg/kg in rats. The distribution of SZ-A in the pancreas, liver, and adipose tissues lays the foundation for its role in regulating insulin secretion and glucose and lipid metabolism. The wide distribution of SZ-A in the kidney, aorta, muscle, and other tissues suggests that SZ-A can also act on multiple target organs, and its multiple pharmacological effects deserve further study. SZ-A was primarily excreted in its original form and was not metabolized by the liver, which had no significant impact on the activity of liver isoenzymes, suggesting that SZ-A caused a low risk of potentially harmful drug-drug interactions.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
## Ethics statement
The animal study was reviewed and approved by the Ethics Committee of Zhongsheng Beidong (Beijing) Technology Development Co., Ltd.
## Author contributions
ZL and YF, conceptualization, methodology, investigation, formal analysis, writing—original draft. HZ, DL, and JH: methodology, investigation. YC, HW, XZ, HY, ZS, XX: investigation, validation. JY and YL, supervision, project administration, funding acquisition, writing—review and editing.
## Conflict of interest
Authors ZL, YF, YC, XZ, and HY were employed by Beijing Wehand-Bio Pharmaceutical Co., Ltd.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2023.1136772/full#supplementary-material
## References
1. Baskin K. K., Winders B. R., Olson E. N.. **Muscle as a "mediator" of systemic metabolism**. *Cell. Metab.* (2015) **21** 237-248. DOI: 10.1016/j.cmet.2014.12.021
2. Campbell J. E., Newgard C. B.. **Mechanisms controlling pancreatic islet cell function in insulin secretion**. *Nat. Rev. Mol. Cell. Biol.* (2021) **22** 142-158. DOI: 10.1038/s41580-020-00317-7
3. Cao H., Ji W., Liu Q., Li C., Huan Y., Lei L.. **Morus alba L.(Sangzhi) alkaloids (SZ-A) exert anti-inflammatory effects via regulation of MAPK signaling in macrophages**. *J. Ethnopharmacol.* (2021) **280** 114483. DOI: 10.1016/j.jep.2021.114483
4. Chan K. C., Lin M. C., Huang C. N., Chang W. C., Wang C. J.. **Mulberry 1-deoxynojirimycin pleiotropically inhibits glucose-stimulated vascular smooth muscle cell migration by activation of AMPK/RhoB and down-regulation of FAK**. *J. Agric. Food Chem.* (2013) **61** 9867-9875. DOI: 10.1021/jf403636z
5. Chen Y. M., Lian C. F., Sun Q. W., Wang T. T., Liu Y. Y., Ye J.. **Ramulus mori (sangzhi) alkaloids alleviate high-fat diet-induced obesity and nonalcoholic fatty liver disease in mice**. *Antioxidants (Basel)* (2022) **11** 905. DOI: 10.3390/antiox11050905
6. Cypess A. M., Haft C. R., Laughlin M. R., Hu H. H.. **Brown fat in humans: Consensus points and experimental guidelines**. *Cell. Metab.* (2014) **20** 408-415. DOI: 10.1016/j.cmet.2014.07.025
7. Defronzo R. A., Ferrannini E., Groop L., Henry R. R., Herman W. H., Holst J. J.. **Type 2 diabetes mellitus**. *Nat. Rev. Dis. Prim.* (2015) **1** 15019. DOI: 10.1038/nrdp.2015.19
8. Hiele M., Ghoos Y., Rutgeerts P., Vantrappen G. J. D. D.. **Effects of acarbose on starch hydrolysis. Study in healthy subjects, ileostomy patients, and**. *Dig. Dis. Sci.* (1992) **37** 1057-1064. DOI: 10.1007/BF01300287
9. Huang S. S., Yan Y. H., Ko C. H., Chen K. M., Lee S. C., Liu C. T.. **A comparison of food-grade folium mori (sāng yè) extract and 1-deoxynojirimycin for glycemic control and renal function in streptozotocin-induced diabetic rats**. *J. Tradit. Complement. Med.* (2014) **4** 162-170. DOI: 10.4103/2225-4110.131639
10. Jin D., Li L., Dong W., Zhu X., Xia X., Wang R.. **Research on transfer rate of heavy metals and harmful elements in traditional Chinese medicine extraction and refining processes and product health risk assessment**. *Biol. Trace Elem. Res.* (2022) **200** 1956-1964. DOI: 10.1007/s12011-021-02788-x
11. Kajimura S., Spiegelman B. M., Seale P.. **Brown and beige fat: Physiological roles beyond heat generation**. *Cell. Metab.* (2015) **22** 546-559. DOI: 10.1016/j.cmet.2015.09.007
12. Ke C., Narayan K., Chan J. C., Jha P., Shah B. R. J. N. R. E.. **Pathophysiology, phenotypes and management of type 2 diabetes mellitus in Indian and Chinese populations**. *Nat. Rev. Endocrinol.* (2022) **18** 1-20. DOI: 10.1038/s41574-022-00669-4
13. Krentz A. J., Bailey C. J. J. D.. **Oral antidiabetic agents: Current role in type 2 diabetes mellitus**. *Oral antidiabetic agents* (2005) **65** 385-411. DOI: 10.2165/00003495-200565030-00005
14. Lei L., Huan Y., Liu Q., Li C., Cao H., Ji W.. **Morus alba L. (Sangzhi) alkaloids promote insulin secretion, restore diabetic beta-cell function by preventing dedifferentiation and apoptosis**. *Front. Pharmacol.* (2022) **13** 841981. DOI: 10.3389/fphar.2022.841981
15. Li Y. G., Ji D. F., Zhong S., Lin T. B., Lv Z. Q., Hu G. Y.. **1-deoxynojirimycin inhibits glucose absorption and accelerates glucose metabolism in streptozotocin-induced diabetic mice**. *Sci. Rep.* (2013) **3** 1377. DOI: 10.1038/srep01377
16. Liu Q., Liu S., Cao H., Ji W., Li C., Huan Y.. **Ramulus Mori (Sangzhi) alkaloids (SZ-A) ameliorate glucose metabolism accompanied by the modulation of gut microbiota and ileal inflammatory damage in type 2 diabetic KKAy mice**. *Front. Pharmacol.* (2021) **12** 642400. DOI: 10.3389/fphar.2021.642400
17. Liu Z., Yang Y., Dong W., Liu Q., Wang R., Pang J.. **Investigation on the enzymatic profile of mulberry alkaloids by enzymatic study and molecular docking**. *Molecules* (2019) **24** 1776. DOI: 10.3390/molecules24091776
18. Lupaescu A. V., Iavorschi M., Covasa M.. **The use of bioactive compounds in hyperglycemia- and amyloid fibrils-induced toxicity in type 2 diabetes and alzheimer's disease**. *Pharmaceutics* (2022) **14** 235. DOI: 10.3390/pharmaceutics14020235
19. Mackay P., Ynddal L., Andersen J. V., Mccormack J. G.. **Pharmacokinetics and anti-hyperglycaemic efficacy of a novel inhibitor of glycogen phosphorylase, 1,4-dideoxy-1,4-imino-d- arabinitol, in glucagon-challenged rats and dogs and in diabetic ob/ob mice**. *Diabetes Obes. Metab.* (2003) **5** 397-407. DOI: 10.1046/j.1463-1326.2003.00293.x
20. Mirzadeh Z., Faber C. L., Schwartz M. W. J. A. R. O. P.. **Central nervous system control of glucose homeostasis**. *A Ther. Target Type 2 Diabetes?* (2022) **62** 55-84. DOI: 10.1146/annurev-pharmtox-052220-010446
21. Mudaliar S., Polidori D., Zambrowicz B., Henry R. R. J. D. C.. **Sodium–glucose cotransporter inhibitors: Effects on renal and intestinal glucose transport: From bench to bedside**. *Diabetes Care* (2015) **38** 2344-2353. DOI: 10.2337/dc15-0642
22. Nakagawa K., Kubota H., Kimura T., Yamashita S., Tsuzuki T., Oikawa S.. **Occurrence of orally administered mulberry 1-deoxynojirimycin in rat plasma**. *J. Agric. Food Chem.* (2007) **55** 8928-8933. DOI: 10.1021/jf071559m
23. Nojima H., Kimura I., Chen F. J., Sugihara Y., Haruno M., Kato A.. **Antihyperglycemic effects of N-containing sugars from Xanthocercis zambesiaca, Morus bombycis, Aglaonema treubii, and**. *J. Nat. Prod.* (1998) **61** 397-400. DOI: 10.1021/np970277l
24. Perreault L., Skyler J. S., Rosenstock J. J. N. R. E.. **Novel therapies with precision mechanisms for type 2 diabetes mellitus**. *Nat. Rev. Endocrinol.* (2021) **17** 364-377. DOI: 10.1038/s41574-021-00489-y
25. Qu L., Liang X., Tian G., Zhang G., Wu Q., Huang X.. **Efficacy and safety of mulberry twig alkaloids tablet for the treatment of type 2 diabetes: A multicenter, randomized, double-blind, double-dummy, and parallel controlled clinical trial**. *Diabetes Care* (2021) **44** 1324-1333. DOI: 10.2337/dc20-2109
26. Santacruz-Calvo S., Bharath L., Pugh G., Santacruz-Calvo L., Lenin R. R., Lutshumba J.. **Adaptive immune cells shape obesity-associated type 2 diabetes mellitus and less prominent comorbidities**. *Nat. Rev. Endocrinol.* (2022) **18** 23-42. DOI: 10.1038/s41574-021-00575-1
27. Scherer T., Sakamoto K., Buettner C. J. N. R. E.. **Brain insulin signalling in metabolic homeostasis and disease**. *Nat. Rev. Endocrinol.* (2021) **17** 468-483. DOI: 10.1038/s41574-021-00498-x
28. Schwartz S. S., Epstein S., Corkey B. E., Grant S. F. A., Gavin J. R., Aguilar R. B.. **A unified pathophysiological construct of diabetes and its complications**. *Trends Endocrinol. Metab.* (2017) **28** 645-655. DOI: 10.1016/j.tem.2017.05.005
29. Strachan M. W., Reynolds R. M., Marioni R. E., Price J. F.. **Cognitive function, dementia and type 2 diabetes mellitus in the elderly**. *Nat. Rev. Endocrinol.* (2011) **7** 108-114. DOI: 10.1038/nrendo.2010.228
30. Sun Q. W., Lian C. F., Chen Y. M., Ye J., Chen W., Gao Y.. **Ramulus mori (sangzhi) alkaloids ameliorate obesity-linked adipose tissue metabolism and inflammation in mice**. *Nutrients* (2022) **14** 5050. DOI: 10.3390/nu14235050
31. Tahrani A. A., Barnett A. H., Bailey C. J. J. N. R. E.. **Pharmacology and therapeutic implications of current drugs for type 2 diabetes mellitus**. *Nat. Rev. Endocrinol.* (2016) **12** 566-592. DOI: 10.1038/nrendo.2016.86
32. Yang S., Mi J., Liu Z., Wang B., Xia X., Wang R.. **Pharmacokinetics, tissue distribution, and elimination of three active alkaloids in rats after oral administration of the effective fraction of alkaloids from Ramulus mori, an innovative hypoglycemic agent**. *Molecules* (2017) **22** 1616. DOI: 10.3390/molecules22101616
33. Yue J. T., Lam T. K.. **Lipid sensing and insulin resistance in the brain**. *Cell. Metab.* (2012) **15** 646-655. DOI: 10.1016/j.cmet.2012.01.013
34. Zhao M., Ma J., Li M., Zhang Y., Jiang B., Zhao X.. **Cytochrome P450 enzymes and drug metabolism in humans**. *Int. J. Mol. Sci.* (2021) **22** 12808. DOI: 10.3390/ijms222312808
35. Zheng Y., Ley S. H., Hu F. B. J. N. R. E.. **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
|
---
title: Quantitative study of 3T MRI qDixon-WIP applied in pancreatic fat infiltration
in patients with type 2 diabetes mellitus
authors:
- Jixing Yi
- Fengming Xu
- Tao Li
- Bumin Liang
- Shu Li
- Qing Feng
- Liling Long
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9981945
doi: 10.3389/fendo.2023.1140111
license: CC BY 4.0
---
# Quantitative study of 3T MRI qDixon-WIP applied in pancreatic fat infiltration in patients with type 2 diabetes mellitus
## Abstract
### Objective
To investigate the application value of 3T MRI qDixon-WIP technique in the quantitative measurement of pancreatic fat content in patients with type 2 diabetes mellitus (T2DM).
### Methods
The 3T MRI qDixon-WIP sequence was used to scan the livers and the pancreas of 47 T2DM patients (experimental group) and 48 healthy volunteers (control group). Pancreatic fat fraction (PFF), hepatic fat fraction (HFF), Body mass index (BMI) ratio of pancreatic volume to body surface area (PVI) were measured. Total cholesterol (TC), subcutaneous fat area (SA), triglyceride (TG), abdominal visceral fat area (VA), high density lipoprotein (HDL-c), fasting blood glucose (FPC) and low-density lipoprotein (LDL-c) were collected. The relationship between the experimental group and the control group and between PFF and other indicators was compared. The differences of PFF between the control group and different disease course subgroups were also explored.
### Results
There was no significant difference in BMI between the experimental group and the control group ($$P \leq 0.231$$). PVI, SA, VA, PFF and HFF had statistical differences ($P \leq 0.05$). In the experimental group, PFF was highly positively correlated with HFF ($r = 0.964$, $P \leq 0.001$), it was moderately positively correlated with TG and abdominal fat area ($r = 0.676$, 0.591, $P \leq 0.001$), and it was weakly positively correlated with subcutaneous fat area ($r = 0.321$, $$P \leq 0.033$$). And it had no correlation with FPC, PVI, HDL-c, TC and LDL-c ($P \leq 0.05$). There were statistical differences in PFF between the control group and the patients with different course of T2DM ($P \leq 0.05$). There was no significant difference in PFF between T2DM patients with a disease course ≤1 year and those with a disease course <5 years ($P \leq 0.05$). There were significant differences in PFF between the groups with a disease course of 1-5 years and those with a disease course of more than 5 years ($P \leq 0.001$).
### Conclusion
PVI of T2DM patients is lower than normal, but SA, VA, PFF, HFF are higher than normal. The degree of pancreatic fat accumulation in T2DM patients with long disease course was higher than that in patients with short disease course. The qDixon-WIP sequence can provide an important reference for clinical quantitative evaluation of fat content in T2DM patients.
## Introduction
Type 2 diabetes mellitus(T2DM) is the most common type of diabetes mellitus [1, 2]. Pancreatic fat infiltration may play an important role in the occurrence and development of T2DM [3, 4]. The degree of lipid infiltration in the pancreas is closely related to abnormal lipid metabolism. With β-cell dysfunction and defective insulin secretion, lipid oxidation and lipolysis are inhibited, which leads to the increase of lipid deposition in the pancreas. The increased degree of pancreatic fat infiltration promotes the development of T2DM (5–7). Therefore, monitoring pancreatic fat content in T2DM patients may provide a certain reference for clinical evaluation of efficacy and disease progression.
Although pancreatic biopsy is the “golden standard” for the quantitative determination of pancreatic fat content, due to the fact that this method only provides small tissue samples, the final measured pancreatic fat content may vary with the different range and degree of pancreatic fat infiltration. Moreover, its invasiveness and poor patient compliance limit the regular detection of pancreatic fat content in T2DM patients. the pancreas is a retroperitoneal organ surrounded by abundant blood vessels and intestines, which makes puncture more difficult [8, 9].
In recent years, multi-echo dixon technology based on Magnetic resonance image (MRI), which is safe, non-invasive and has good tissue resolution, has been confirmed in various organs including the pancreas in terms of tissue fat quantification (10–13). The early two-point Dixon technique could only quantify the adipogenic variation below $50\%$ [14], which was greatly affected by the non-uniformity of the main magnetic field and the attenuation effect of T1 and T2* [15]. Three-point Dixon technique can collect one more in-phase echo signal on the basis of two-point method, which can correct T2* attenuation to a certain extent. However, the obtained organ fat fraction is susceptible to various confounding factors, and its accuracy and repeatability are not enough to be a reliable index of fat quantification [16]. 6 Echo Dixon (qDixon) technology, compared with the earlier Dixon technology, effectively corrects the errors caused by the magnetic field inhomogeneity and T2* attenuation, making the quantitative results more accurate. The fat distribution map can not only directly measure the fat content quantitatively, but also fully reflect the fat distribution [17].
The purpose of this study was to investigate the value of 3T qDixon technique in the quantitative determination of pancreatic fat content in T2DM patients, and to provide reference for the early diagnosis, clinical treatment, disease progression and efficacy evaluation of pancreatic changes in T2DM patients by comparing the relationship between relevant indicators.
## Research objects
A total of 95 volunteers were recruited from April 1, 2019 to June 30, 2022, including 36 females (17 T2DM patients, 19 normal controls) and 59 males (30 T2DM patients, 29 normal controls). T2DM Patients ranged from 32 to 71 years old (51.32 ± 10.60). The normal control group ranged from 31 to 68 years old (51.28 ± 8.91). Inclusion criteria: [1] patients diagnosed with T2DM and healthy volunteers with similar age to T2DM patients (± 3 years old) and no related diseases. Exclusion criteria: [1] patients unable to participate in MRI examination due to contraindications or other reasons; [2] patients with liver and pancreatic tumors; [3] patients after splenectomy; [4] patients with abnormal metabolic function or metabolic diseases excluding T2DM; [5] patients with hepatitis virus or hepatitis B, and liver iron deposition; [6] patients with liver trauma or patients receiving a liver transplant; [7] patients with pancreatic inflammation and alcoholics; [8] Patients with a history of drug therapy for the the pancreas (Sulfonamides, azathioprine, glucocorticoids, thiazide diuretics) and liver (Platinum agents, antibiotics, alkylating agents, antipsychotics, anti-tuberculosis drugs, and anti-tumor drugs) within six months. This study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the hospital Ethics Committee (NO.2022-E460-01).
## Instruments and methods
MRI scans were performed on all subjects by the same operator with 10 years of extensive MRI scanning experience. Abdominal axial scan was performed at the end of breath using a 3.0T MRI scanner Siemens 3T MRI scanner (Prisma, Siemens Healthcare, Erlangen, Germany). qDdixon-WIP sequence scanning parameters: echo time (TE): 1.26, 2.60, 3.94, 5.28, 6.62, 7.96ms; repetition time (TR): 9.25ms; slice thickness: 3.5mm, matrix: 160×120; bandwidth: 1040Hz/Pixel; field of view: 380mm×313.5mm; scanning time: about 18s.
## Image processing
Data measurements were performed by two radiologists who were familiar with image post-processing and had more than 5 years of experience in abdominal diagnosis. Measurement process: The Region of interest (ROI) was delineated independently on the fat content (FF) diagram of qDixon-WIP sequence, and the fat fraction was directly measured (fat fraction =$10\%$× the mean measured by software). For liver, the intrahepatic sink area was avoided as far as possible. Four ROIs (liver S$\frac{2}{3}$, S4, S$\frac{5}{8}$, S$\frac{6}{7}$) were selected (Figures 1A, B) to measure liver fat fraction, each ROI was about 0.4 ~ 0.6cm2, and the corresponding Goodness of fit was measured (Figures 2A, B). Average values were taken (<$5\%$ indicates good accuracy). For the pancreas, three ROIs (head, body and tail of the pancreas) were selected (Figures 3A–C) to measure pancreatic fat fraction, each ROI was about 0.1-0.2 cm2, and Goodness of fit was also measured (Figures 4A–C), and average values were taken. The images were uploaded to Ziostation workstation (Ziostation2 Version 2.4.0.2), and the “3D standard and Viewer” functions in the workstation were used for image processing: The whole the pancreas was manually delineated, and the pancreatic volume was automatically calculated by the software (Figure 5A), and visceral fat area (VA) and subcutaneous fat area (SA) were measured in the experimental and control groups via the umbilical plane (Figure 5B). For pancreatic volume, in order to exclude the influence of height, weight and other factors among individuals, pancreatic volume to body surface area (PVI) was obtained by conversion (male: body surface area [m2] = 0.0057 × height [cm] + 0.0121 × weight [kg] + 0.0882; Female: body surface area [m2] = 0.0073 × height [cm] + 0.0127 × weight [kg] - 0.2106; Pancreatic volume per unit body surface area: PVI [cm3/m2]= pancreatic volume cm3/body surface area m2) [18]. All measurement data were taken from the mean values measured by two doctors. The patient’s clinical data was queried through HIS system of our institution; Height and weight were measured on the day of MRI scan.
**Figure 1:** *(A, B) show the liver fat fraction maps of volunteers. Mean fat fraction =10%× (108.30 + 92.70+68.20+87.70)/4 = 8.92 (two decimal places reserved). The green area is the liver region automatically delineated by the software.* **Figure 2:** *(A, B) show the corresponding Goodness of fit plots for liver fat fraction. A Goodness of fit average = (4.90% + 4.90% + 4.60% + 3.80%)/4 = 4.55%. The green area is the liver region automatically delineated by the software.* **Figure 3:** *(A–C) show the pancreatic fat fraction maps of the volunteers. Mean pancreatic fat fraction =10%× (18.50 + 32.00+31.00)/3 = 2.72 (keep two decimal places). The green area is the liver region automatically delineated by the software.* **Figure 4:** *(A–C) show the corresponding Goodness of fit plots for pancreatic fat fraction. Goodness of fi mean = (4.5%+4.0%+4.0%)/3 = 4.17% (keep two decimal places). The green area is the area automatically delineated by the software.* **Figure 5:** *(A) shows the pancreatic Volume map of the volunteer the pancreas obtained by 3D standard processing, and the lower right corner of the figure shows the pancreatic volume Mask Volume(V1):63.23 cc. (B) shows the subcutaneous fat area(ROI 3 = 181.15cm2) and abdominal visceral fat area(ROI 4 = 15.93cm2, ROI 5 = 12.41cm2, ROI 6 = 1.91cm2) of volunteers after processing with Viewer.*
## Statistical methods
SPSS22.0 software was used for statistical analysis. Kolmogorov-Smirnov(K) method was used to test the normal distribution of the data. Measurement data with normal distribution were represented as mean ± standard deviation (M). Measurement data with non-normal distribution were expressed as median, and quartile. Pearson chi-square test was used to compare the differences in gender composition. The independent sample t test (normal distribution) or Mann-Whitney U test (non-normal distribution) was used to compare the Pancreatic fat fraction (PFF), SA, VA and PVI between the experimental group and the control group. Pearson (normal distribution) or Spearman (non-normal distribution) correlation analysis was used to evaluate the correlation between the measured PFF and Hepatic fat component (HFF), PVI, SA, VA and clinical indicators in T2DM patients. The threshold for significance was set at 0.05.
## Consistency test for determination of pancreatic fat
The PFF, HFF, SA, VA and PVI of the experimental group and the control group were measured by two doctors (A and B) at different times. The Intraclass correlation coefficient (ICC) consistency test showed that the measured results were consistent between the groups (Table 1). It can be considered that the data measured by different doctors were highly consistent with intra-observer and inter-observer.
**Table 1**
| Group | Measurement index | Doctor A1 | Doctor A2 | Doctor B | Consistency coefficient (and 95% credibility Interval) | Consistency coefficient (and 95% credibility Interval).1 |
| --- | --- | --- | --- | --- | --- | --- |
| Group | Measurement index | Doctor A1 | Doctor A2 | Doctor B | A1 and A2 A1 and B | A1 and A2 A1 and B |
| experimental | PVI (cm3/m2) | 31.55 ± 1.79 | 31.58 ± 1.76 | 31.29 ± 1.73 | 0.993 (0.987~0.996) | 0.917 (0.845~0.956) |
| | SA (cm2) | 126.67 ± 44.70 | 126.65 ± 43.71 | 133.10 ± 45.05 | 1.000 (1.000~1.000) | 0.957 (0.916~0.978) |
| | VA (cm2) | 82.93 ± 26.48 | 82.94 ± 26.52 | 84.93 ± 22.59 | 1.000 (1.000~1.000) | 0.925 (0.889~0.963) |
| | PFF (%) | 3.70 (2.70∼5.40) | 3.70 (2.50∼5.50) | 3.08 (2.34∼5.90) | 0.991 (0.984~0.995) | 0.936 (0.893~0.971) |
| | HFF (%) | 6.03 (3.975∼8.15) | 6.00 (4.00∼8.50) | 5.36 (3.48∼8.42) | 0.994 (0.990~0.997) | 0.949 (0.913~0.972) |
| control | PVI (cm3/m2) | 33.68 ± 1.76 | 33.67 ± 1.76 | 34.48 ± 2.49 | 0.998 (0.996~0.999) | 0.923 (0.861~0.958) |
| | SA (cm2) | 108.56 ± 36.40 | 108.54 ± 36.41 | 107.95 ± 46.24 | 1.000 (1.000~1.000) | 0.944 (0.891~0.975) |
| | VA (cm2) | 37.31 ± 10.77 | 37.33 ± 10.76 | 38.52 ± 12.24 | 1.000 (1.000~1.000) | 0.910 (0.829~0.945) |
| | PFF (%) | 1.76 ± 0.66 | 1.79 ± 0.63 | 1.67 ± 0.69 | 0.992 (0.986~0.996) | 0.950 (0.923~0.987) |
| | HFF (%) | 3.37 ± 1.47 | 3.41 ± 1.49 | 3.13 ± 1782 | 0.999 (0.998~0.999) | 0.933 (0.901~0.970) |
## Clinical parameter processing and normality test of measurement data
The normality test showed that BMI, PVI, SA, VA, TC, TG, HDL-c in the experimental group and BMI, PVI, SA, VA, PFF and HFF in the normal control group were all normal distribution ($P \leq 0.05$). In the experimental group, PFF, HFF, FPC and LDL-c showed non-normal distribution ($P \leq 0.05$) (Table S1).
## Comparison and analysis results of related fat mass and parameters between T2DM patients and control group
There was no significant difference in age and gender distribution between the experimental group and the normal control group ($P \leq 0.05$). The other indicators were BMI, PVI, SA, VA, PFF and HFF. There was no significant difference in BMI between the experimental group and the control group ($P \leq 0.05$). There were statistical differences in PVI, SA, VA, PFF, and HFF between the two groups ($P \leq 0.05$) (Table 2). PVI of T2DM patients was lower than that of control group, while SA, VA, PFF and HFF were higher than those of control group.
**Table 2**
| Evaluating indicator | T2DM | control group | T/X2 | P |
| --- | --- | --- | --- | --- |
| BMI(kg/m2) | 23.04 ± 2.826 | 23.68 ± 2.327 | -1.206 | 0.231 |
| PVI(cm3/m2) | 31.27 ± 1.761 | 34.01 ± 2.487 | -6.220 | <0.001 |
| SA(cm2) | 133.07 ± 44.91 | 108.44 ± 18.39 | 3.377 | 0.001 |
| VA(cm2) | 84.72 ± 22.30 | 37.91 ± 12.057 | 12.307 | <0.001 |
| PFF(%) | 3.10(2.31∼5.84) | 1.73 ± 0.697 | -6.209 | <0.001 |
| HFF(%) | 5.50(3.48∼8.42) | 3.24 ± 1.617 | -4.682 | 0.016 |
| Age(year) | 51.32 ± 10.604 | 51.28 ± 8.907 | -0.024 | 0.981 |
| gender | F=17,M=30 | F=19,M=29 | 0.118 | 0.883 |
| n | 47 | 48 | – | – |
## Correlation analysis between fat-related measurements and clinical indicators in T2DM patients
PFF was positively correlated with HFF in the experimental group ($r = 0.964$, $P \leq 0.001$). It was moderately positively correlated with TG, VA and Disease course ($r = 0.676$, 0.591, 0.615, $P \leq 0.001$), and weakly positively correlated with SA ($r = 0.321$, $$P \leq 0.033$$). There was no significant correlation with FPC, TC, PVI, HDL-c, LDL-c ($r = 0.385$, 0.236, -0.163, -0.168, -0.002; $$P \leq 0.194$$, 0.437, 0.292, 0.276, 0.987)(Table 3 and Figure 6). The non-standardized linear regression equation constructed with PFF as the dependent variable and the other indicators as the independent variables is: PFF=10.287+0.284HFF-0.255PVI-0.329TG+0.758Disease course(According to the inspection level of 0.05, only HFF, PVI, TG and Disease course were included in the regression equation) (Table 4). The standardization coefficients of HFF, PVI, TG and Disease course are 0.637, -0.233, -0.18 and 0.303 (Table 4).
## Comparison and analysis of PFF values between experimental group and control group in patients with different course of disease
The measurement results of PFF values in the control group and the experimental group with different course of disease are shown in Table 5. The results of comparison between groups are shown in Table S2. The PFF of the control group and the experimental group were statistically different ($P \leq 0.05$), and the pancreatic fat content of the control group was lower than that of the experimental group. There was no statistically significant difference in PFF between patients with less than one year of disease course and those with one to five years of disease course in the experimental group ($P \leq 0.05$), which could not indicate that the PFF of patients with one to five years of disease course was higher than that of patients with one year of disease course. The PFF of patients with less than 1 year and 1 to 5 years of disease course was statistically different from that of patients with more than 5 years of disease course ($P \leq 0.05$), which could be considered that the PFF of patients with less than 1 year and 1 to 5 years of disease course was less than that of patients with more than 5 years of disease course.
**Table 5**
| Group | PFF (%) | Statistical variables | sample capacity | P value |
| --- | --- | --- | --- | --- |
| control group | 1.73 ± 0.697 | 0.098 | 48 | 0.2 |
| disease course ≤1 year | 2.28 (1.99∼3.45) | 0.289 | 14 | 0.002 |
| disease course of 1-5 years | 2.75 (2.35∼4.40) | 0.241 | 19 | 0.005 |
| disease course > 5 years | 6.46 ± 1.914 | 0.142 | 14 | 0.2 |
## Discussion
For ectopic fat accumulation in T2DM, ectopic lipid deposition can promote its development and plays an important role in its progression [19, 20]. Studies have reported that pancreatic fatty infiltration is associated with insulin resistance, and the incidence of diabetes in people with pancreatic fatty infiltration is significantly higher than the other people [21, 22]. At present, MRI-based fat quantification technology can identify small changes in fat content, quantify fat and monitor steatosis, making it play an increasingly important role in the assessment of pancreatic fat content [10].
In this study, 3.0T MRI qDIXon-WIP sequence was used to quantify pancreatic fat, which improved the solution to the problem of inverse calculation of water image and fat image in qDixon image. Under the condition of good consistency of gender, age and BMI matching between the experimental group and the control group, the HFF, PFF and intraperitoneal and external fat contents of the experimental group were higher than those of the control group, which reflects that there is a certain connection between abnormal fat metabolism and ectopic fat deposition. The accumulation of lipids in the the pancreas can lead to the blockage of signaling pathways and insulin resistance, thus leading to the release of inflammatory adipokines, and ultimately aggravating the deposition of fat in the abdominal organs [17]. However, abnormal glucose metabolism (decreased insulin secretion or insulin resistance) will lead to weakened liver cells’ ability to metabolize fat, resulting in increased ectopic fat deposition (17, 20–22).
In this study, the PFF value of T2DM group was almost 2 times that of the normal control group (Table 2), which is similar to the study conducted by Tushuizen et al [23]. However, in this study, the data in T2DM group conformed to the normal distribution and the patient sample size was sufficient. In the experimental group, HFF and PFF of patients showed a strong positive correlation, suggesting that liver fat deposition was closely related to pancreatic fat deposition, which was similar to the research results of van Geenen [24]. Some of the differences in results may be related to assessment methods (ultrasound, CT, magnetic resonance), individual differences (psychological factors, diet, exercise, BMI, etc.), measurement methods (delineation of areas of interest, uneven distribution of fat deposits in the the pancreas) and other factors. T2DM patients have abnormal metabolism, which will cause the increase of TG. When the TG in the body is supersaturation in adipose tissue, lipids will be accumulated in non-fatty organs, such as the pancreas, etc., and pancreatic fat infiltration will promote the progression of T2DM and the increase of TG [5]. In the experimental group, the moderate positive correlation between PFF and TG indicates that they have a close relationship. The study of Hu and Yamazaki showed that abdominal fat accumulation and abdominal fat deposition were related to diabetes and other risk factors [25, 26]. The study of Yu, Van and Anderson showed that SA and VA in T2DM patients were also related to T2DM: intra-abdominal fat decreased the inhibitory effect of insulin on lipolysis by increasing gluconogenesis and insulin sensitivity [27, 28]. In this study, PFF was moderately positively correlated with abdominal fat and weakly positively correlated with subcutaneous fat area, which also reflected that intra-abdominal and extra-abdominal fat deposition were related factors for pancreatic fat infiltration. Some studies also pointed out that there was a significant correlation between abdominal fat distribution and older patients [29], and the different course of T2DM patients led to certain differences in results. In addition, this study also compared the correlation between PFF and FPC, PVI, HDL-c, TC and LDL-c, and the results indicated that there was no significant correlation. The constructed linear regression equation points out that among the relevant indicators in this study, HFF, PVI, TG and Disease course have greater contribution to PFF, that is, these four factors are closely related to PFF.
The patients in the experimental group were divided into three groups according to the course of disease: course of disease ≤1 year, 1 year < course of disease < 5 years, and course of disease ≥5 years. The results suggest that pancreatic fat accumulation is higher in patients with long course of T2DM than in those with short course of T2DM. According to the linear regression analysis, the standardized regression coefficient for Disease course was 0.303, which points out the degree of fat accumulation is higher in those with long-standing diabetes. And as mentioned above, insulin resistance causes ectopic fat deposition, and pancreatic fat also accumulates in the progression of T2DM. The results of this study may partly explain that pancreatic fat infiltration is a gradual accumulation process in patients with long disease course, but the degree of fat accumulation is slower in patients with short disease course.
Limitations of this study: [1] Due to the small sample size, further sample expansion is needed to improve the reliability of the experimental results. [ 2] Due to the age distribution characteristics of the diabetic population, the age of T2DM patients included in this study ranged from 32 to 71 years old, and the corresponding age of normal control population was matched, and the data obtained had certain bias. [ 3] *In this* study, T2DM patients were randomly sampled, and subgroup analysis of patients with different clinical interventions was not performed. The degree of pancreatic fat infiltration is likely to be different in patients with different interventions. This study can further focus on the relationship between pancreatic fat deposition and T2DM intervention.
In this study, the qDixon-WIP sequence was used to conduct clinical experiments. The results showed that: [1] PVI decreased, while SA, VA, PFF and HFF increased in T2DM patients. [ 2] PFF was positively correlated with HFF, TG, abdominal fat area and subcutaneous fat area in T2DM patients. [ 3] The degree of pancreatic fat accumulation in patients with long course of disease was higher than that in patients with short course of disease. This sequence can be used in clinical research to quantitatively measure pancreatic fat content with good repeatability, which can provide reference for clinical assessment of pancreatic fat to achieve real-time monitoring of the occurrence and progression of diseases.
## Data availability statement
The data analyzed in this study is subject to the following licenses/restrictions: The datasets generated during and/or analysed during the current study are not publicly available, but are available from the corresponding author on reasonable request. Requests to access these datasets should be directed to LiLing Long, [email protected].
## Ethics statement
The studies involving human participants were reviewed and approved by Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (NO.2022-E460-01). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
Material preparation and data collection were performed by JY, FX, TL, SL and QF. Data analysis were performed by JY, FX. The first draft of the manuscript was written by JY, FX and BL and all authors commented on previous versions of the manuscript. All authors contributed to the article and approved the submitted version. LL contributed to the study conception and design.
## 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.2023.1140111/full#supplementary-material
## References
1. Jenkins A, Lee MK, Kadowaki T. **Comprehensive IDF-WPR diabetes and disasters manual, 2nd edition available**. *Diabetes Res Clin Pract* (2022) 110209. DOI: 10.1016/j.diabres.2022.110209
2. Dong G, Qu L, Gong X, Pang B, Yan W, Wei J. **Effect of social factors and the natural environment on the etiology and pathogenesis of diabetes mellitus**. *Int J Endocrinol* (2019) **2019**. DOI: 10.1155/2019/8749291
3. Ng N, Mijares Zamuner M, Siddique N, Kim J, Burke M, Byrne MM. **Genotype-phenotype correlations and response to glucose lowering therapy in subjects with HNF1β associated diabetes**. *Acta diabetol* (2022) **59** 83-93. DOI: 10.1007/s00592-021-01794-8
4. Chan TT, Tse YK, Lui RN, Wong GL, Chim AM, Kong AP. **Fatty pancreas is independently associated with subsequent diabetes mellitus development: A 10-year prospective cohort study**. *Clin Gastroenterol Hepatol* (2022) **20** 2014-2022.e4. DOI: 10.1016/j.cgh.2021.09.027
5. Kim JY, Nasr A, Tfayli H, Bacha F, Michaliszyn SF, Arslanian S. **Increased lipolysis, diminished adipose tissue insulin sensitivity, and impaired β-cell function relative to adipose tissue insulin sensitivity in obese youth with impaired glucose tolerance**. *Diabetes* (2017) **66**. DOI: 10.2337/db17-0551
6. Chan JY, Bensellam M, Lin R, Liang C, Lee K, Jonas JC. **Transcriptome analysis of islets from diabetes-resistant and diabetes-prone obese mice reveals novel gene regulatory networks involved in beta-cell compensation and failure**. *FASEB J* (2021) **35**. DOI: 10.1096/fj.202100009R
7. Taylor R, Al-Mrabeh A, Zhyzhneuskaya S, Peters C, Barnes AC, Aribisala BS. **Remission of human type 2 diabetes requires decrease in liver and pancreas fat content but is dependent upon capacity for β cell recovery**. *Cell Metab* (2018) **28** 547-556.e3. DOI: 10.1016/j.cmet.2018.07.003
8. Suleiman M, Marselli L, Cnop M, Eizirik DL, De Luca C, Femia FR. **The role of beta cell recovery in type 2 diabetes remission**. *Int J Mol Sci* (2022) **23**. DOI: 10.3390/ijms23137435
9. Solimena M, Schulte AM, Marselli L, Ehehalt F, Richter D, Kleeberg M. **Systems biology of the IMIDIA biobank from organ donors and pancreatectomised patients defines a novel transcriptomic signature of islets from individuals with type 2 diabetes**. *Diabetologia* (2018) **61**. DOI: 10.1007/s00125-017-4500-3
10. Kühn JP, Hernando D, Muñoz del Rio A, Evert M, Kannengiesser S, Völzke H. **Effect of multipeak spectral modeling of fat for liver iron and fat quantification: Correlation of biopsy with MR imaging results**. *Radiology* (2012) **265**. DOI: 10.1148/radiol.12112520
11. Meisamy S, Hines CD, Hamilton G, Sirlin CB, McKenzie CA, Yu H. **Quantification of hepatic steatosis with T1-independent, T2-corrected MR imaging with spectral modeling of fat: Blinded comparison with MR spectroscopy**. *Radiology* (2011) **258**. DOI: 10.1148/radiol.10100708
12. Yokoo T, Shiehmorteza M, Hamilton G, Wolfson T, Schroeder ME, Middleton MS. **Estimation of hepatic proton-density fat fraction by using MR imaging at 3.0 T**. *Radiology* (2011) **258**. DOI: 10.1148/radiol.10100659
13. Kühn JP, Berthold F, Mayerle J, Völzke H, Reeder SB, Rathmann W. **Pancreatic steatosis demonstrated at MR imaging in the general population: Clinical relevance**. *Radiology* (2015) **276**. DOI: 10.1148/radiol.15140446
14. Lee SS, Park SH. **Radiologic evaluation of nonalcoholic fatty liver disease**. *World J Gastroenterol* (2014) **20**. DOI: 10.3748/wjg.v20.i23.7392
15. Yu H, Shimakawa A, Hines CD, McKenzie CA, Hamilton G, Sirlin CB. **Combination of complex-based and magnitude-based multiecho water-fat separation for accurate quantification of fat-fraction**. *Magnetic Resonance Med* (2011) **66** 199-206. DOI: 10.1002/mrm.22840
16. Hong CW, Fazeli Dehkordy S, Hooker JC, Hamilton G, Sirlin CB. **Fat quantification in the abdomen**. *Topics magnetic resonance imaging: TMRI* (2017) **26**. DOI: 10.1097/RMR.0000000000000141
17. Bawden SJ, Hoad C, Kaye P, Stephenson M, Dolman G, James MW. **Comparing magnetic resonance liver fat fraction measurements with histology in fibrosis: the difference between proton density fat fraction and tissue mass fat fraction**. *Magma (New York N.Y.)* (2022). DOI: 10.1007/s10334-022-01052-0
18. Hu Y, Wu X, Hu Z, Ren A, Wei X, Wang X. **Research on the human surface area formula in China**. *Acta Physiol Sin* (1999) **51**
19. Filippatos TD, Alexakis K, Mavrikaki V, Mikhailidis DP. **Nonalcoholic fatty the pancreas disease: Role in metabolic syndrome, “Prediabetes,” diabetes and atherosclerosis**. *Digestive Dis Sci* (2022) **67** 26-41. DOI: 10.1007/s10620-021-06824-7
20. Sreedhar UL, DeSouza SV, Park B, Petrov MS. **A systematic review of intra-pancreatic fat deposition and pancreatic carcinogenesis**. *J Gastrointestinal Surg* (2020) **24**. DOI: 10.1007/s11605-019-04417-4
21. Sanchez Caballero L, Gorgogietas V, Arroyo MN, Igoillo-Esteve M. **Molecular mechanisms of β-cell dysfunction and death in monogenic forms of diabetes**. *Int Rev Cell Mol Biol* (2021) **359** 139-256. DOI: 10.1016/bs.ircmb.2021.02.005
22. Miranda MA, Macias-Velasco JF, Lawson HA. **Pancreatic β-cell heterogeneity in health and diabetes: Classes, sources, and subtypes**. *Am J Physiol Endocrinol Metab* (2021) **320**. DOI: 10.1152/ajpendo.00649.2020
23. Tushuizen ME, Bunck MC, Pouwels PJ, Bontemps S, van Waesberghe JH, Schindhelm RK. **Pancreatic fat content and beta-cell function in men with and without type 2 diabetes**. *Diabetes Care* (2007) **30**. DOI: 10.2337/dc07-0326
24. van Geenen EJ, Smits MM, Schreuder TC, van der Peet DL, Bloemena E, Mulder CJ. **Nonalcoholic fatty liver disease is related to nonalcoholic fatty the pancreas disease**. *Pancreas* (2010) **39**. DOI: 10.1097/MPA.0b013e3181f6fce2
25. Hu HH, Kim HW, Nayak KS, Goran MI. **Comparison of fat-water MRI and single-voxel MRS in the assessment of hepatic and pancreatic fat fractions in humans**. *Obes (Silver Spring Md.)* (2010) **18**. DOI: 10.1038/oby.2009.352
26. Yamazaki H, Tauchi S, Kimachi M, Dohke M, Hanawa N, Kodama Y. **Association between pancreatic fat and incidence of metabolic syndrome: A 5-year Japanese cohort study**. *J Gastroenterol Hepatol* (2018) **33**. DOI: 10.1111/jgh.14266
27. Yu TY, Wang CY. **Impact of non-alcoholic fatty the pancreas disease on glucose metabolism**. *J Diabetes Invest* (2017) **8**. DOI: 10.1111/jdi.12665
28. Anderson PJ, Chan JC, Chan YL, Tomlinson B, Young RP, Lee ZS. **Visceral fat and cardiovascular risk factors in Chinese NIDDM patients**. *Diabetes Care* (1997) **20**. DOI: 10.2337/diacare.20.12.1854
29. Boyko EJ, Fujimoto WY, Leonetti DL, Newell-Morris L. **Visceral adiposity and risk of type 2 diabetes: A prospective study among Japanese americans**. *Diabetes Care* (2000) **23**. DOI: 10.2337/diacare.23.4.465
|
---
title: Targeting fat mass and obesity-associated protein mitigates human colorectal
cancer growth in vitro and in a murine model
authors:
- Thuy Phan
- Vu H. Nguyen
- Rui Su
- Yangchan Li
- Ying Qing
- Hanjun Qin
- Hyejin Cho
- Lei Jiang
- Xiwei Wu
- Jianjun Chen
- Marwan Fakih
- Don J. Diamond
- Ajay Goel
- Laleh G. Melstrom
journal: Frontiers in Oncology
year: 2023
pmcid: PMC9981948
doi: 10.3389/fonc.2023.1087644
license: CC BY 4.0
---
# Targeting fat mass and obesity-associated protein mitigates human colorectal cancer growth in vitro and in a murine model
## Abstract
### Introduction
Colorectal cancer (CRC) remains a significant cause of cancer related mortality. Fat mass and obesity-associated protein (FTO) is a m6A mRNA demethylase that plays an oncogenic role in various malignancies. In this study we evaluated the role of FTO in CRC tumorigenesis.
### Methods
Cell proliferation assays were conducted in 6 CRC cell lines with the FTO inhibitor CS1 (50-3200 nM) (± 5-FU 5-80 mM) and after lentivirus mediated FTO knockdown. Cell cycle and apoptosis assays were conducted in HCT116 cells (24 h and 48 h, 290 nM CS1). Western blot and m6A dot plot assays were performed to assess CS1 inhibition of cell cycle proteins and FTO demethylase activity. Migration and invasion assays of shFTO cells and CS1 treated cells were performed. An in vivo heterotopic model of HCT116 cells treated with CS1 or with FTO knockdown cells was performed. RNA-seq was performed on shFTO cells to assess which molecular and metabolic pathways were impacted. RT-PCR was conducted on select genes down-regulated by FTO knockdown.
### Results
We found that the FTO inhibitor, CS1 suppressed CRC cell proliferation in 6 colorectal cancer cell lines and in the 5-Fluorouracil resistant cell line (HCT116-5FUR). CS1 induced cell cycle arrest in the G2/M phase by down regulation of CDC25C and promoted apoptosis of HCT116 cells. CS1 suppressed in vivo tumor growth in the HCT116 heterotopic model ($p \leq 0.05$). Lentivirus knockdown of FTO in HCT116 cells (shFTO) mitigated in vivo tumor proliferation and in vitro demethylase activity, cell growth, migration and invasion compared to shScr controls ($p \leq 0.01$). RNA-seq of shFTO cells compared to shScr demonstrated down-regulation of pathways related to oxidative phosphorylation, MYC and Akt/ mTOR signaling pathways.
### Discussion
Further work exploring the targeted pathways will elucidate precise downstream mechanisms that can potentially translate these findings to clinical trials.
## Introduction
Colorectal cancer (CRC) is the most common gastrointestinal malignancy and the second leading cause of cancer-related death, with an estimated greater than 150,000 diagnosed and 50,000 deaths in the United States in 2020 [1]. Treatments for CRC includes conventional therapies such as surgery, chemotherapy, immunotherapy, and radiation. The 5-year survival rate for CRC is $64\%$ at all stages and decreases to only $14\%$ for metastatic disease [2]. Interestingly, recent study found out that some specific gut microorganisms playing an important role in resistant to 5-Fluorouracil (5-FU) and oxaliplatin therapy via modulating autophagy pathway [3]. On the other hand, the use of antibiotics might reduce gut microbiota which is associated with lower mortality in metastatic CRC patients with bevacizumab therapy [4]. Therefore, further investigation is needed to develop novel and effective systemic therapies for this disease. N6-methyladenosine (m6A) is the most abundant and prevalent internal modification in eukaryotic mRNA among various types of RNA, including messenger RNA, microRNA, long noncoding RNA and circular RNA [5, 6]. m6A modification is the methylation of the adenosine base at the nitrogen-6 position of mRNA and reversibly mediated by the methyltransferase (“writers” including METTL3, METTL14 and WTAP) (7–9), the demethylases (“erasers” including FTO and ALKBH5) [10, 11] and binding proteins that preferentially recognized m6A methylated transcripts and trigger downstream pathways (“readers” including YTH, HNRNP and IGF2BP) (12–14). m6A RNA methylation is also regulated by intestinal bacteria, which is related to the progression of cancers [15, 16]. Especially, m6A modifications in CRC cells and patient-derived xenograft markedly suppressed by *Fucobacterium nucleatum* via downregulation of METTL3, which enhances the colorectal metastasis [17]. FTO and ALKHB5 belong to human AlkB family of non-heme Fe(II) and 2-oxoglutarate–dependent oxygenases and oxidatively demethylate m6A [10]. FTO catalyzes the conversion of m6A to hm6A with slow release of A and formaldehyde (FA), while ALKBH5 directly demethylates m6A to A with rapid FA formation [18]. Fat mass and obesity-associated protein (FTO), which plays a role in regulating fat mass and adipogenesis, was identified as the first m6A RNA demethylase [10]. Recent studies demonstrated that FTO is overexpressed and plays oncogenic roles in various cancers such as acute myeloid leukemia, gastric cancer, breast cancer, melanoma, and cervical cancer (19–23). Meanwhile, other studies indicated that FTO could function as a tumor suppressor and downregulation of FTO promotes cancer development in renal cell cancer, liver cancer and ovarian cancer (24–27). Importantly, the role of FTO in CRC tumorigenesis through the m6A pathway remains controversial. Recently, Lin et al. demonstrated that FTO was overexpressed in primary and 5-FU-resistant CRC tissues and FTO enhances 5-FU resistance through SIVA1-mediated apoptosis pathway [28]. Wang et al. also found that FTO promotes CRC development and increases chemotherapy resistance through G6PD/PARP1 demethylation [29]. In contrast, Ruan et al. indicated the low expression of FTO in CRC tissues and better prognosis in patients with higher level of FTO protein [30]. Inhibition of FTO promotes cancer stem cell properties in CRC including sphere formation, tumor development and chemoresistance [31]. Therefore, understanding the role of FTO in CRC is very important, especially to find out the effective target for treatment to fight against cancer.
In this study, we demonstrated that FTO is expressed in different CRC cell lines. To explore the biological function of FTO in CRC, we conducted knockdown of FTO and analyzed changes in proliferation, migration, and invasion of cultured cells in vitro and in tumor progression in vivo. We also investigated the underlying mechanism of FTO inhibition using RNA-sequencing (RNA-seq). Additionally, as proof of principle, CS1, a small molecule FTO inhibitor was evaluated as a promising candidate for CRC treatment [32]. The overall aim of this work was to assess the role of FTO in CRC as it pertains to tumor progression.
## Tumor cell line and mice
Human colorectal cancer cell lines (CRC) including KRAS wild type (HT-29, COLO) and KRAS mutant (HCT116, LoVo, SW480, SW620) were purchased from American Type Culture Collection (ATCC, USA). 5-FU resistant HCT116 cell line was obtained from MD Anderson Cancer Center (USA). Cells were cultured in McCoy for HT-29 or DMEM for COLO, HCT116, HCT116-5FUR, SW480, SW620 or F-12K for LoVo (all media from Corning, USA). Media were supplemented with $10\%$ FBS (Gibco, USA) and $1\%$ penicillin-streptomycin (Gibco, USA) and cells were grown at 37°C in a humidified incubator with $5\%$ CO2.
6-week-old female NSG mice were purchased from Jackson Lab and maintained under specific pathogen-free conditions at Animal Research Center (City of Hope). Animals were handled according to Institutional Animal Care and Use Committee (IACUC) guidelines under an approved protocol #16067.
## Reagents
FTO inhibitor, CS1 or bisantrene (B4563, Sigma-Aldrich) was dissolved in DMSO and saturated β-cyclodextrin (C0926, Sigma-Aldrich) solution for in vitro and in vivo experiments, respectively.
## Cell proliferation assay
The proliferation of CRC cell lines under CS1 treatment was determined using the MTS assays (Cell Titer 96 Aqueous One Solution, Promega, USA). Cells (1x104 cells/100 µl/well) were seeded in 96-well tissue culture plates. After 24 h, cells were incubated with CS1 (50-3200 nM) as single agent. In combination treatment, HCT116 cells were incubated with CS1 and 5-FU (5-80 µM). After 72 h of treatment, cell proliferation was evaluated using an MTS assay according to the manufacturer’s instructions. Briefly, 20 μl of MTS reagent was added to each well, and the plates were incubated for 2 h at 37°C. Absorbance at 492 nm was measured with a microplate reader (Filtermax F3). Results were represented as the means ± standard deviation of the mean (SD) from triplicate wells.
## Cell cycle analysis
HCT116 cells were seeded into 6-well plates (5x105 cells/well), treated with 290 nM CS1 or DMSO (vehicle control) for 24 h and 48 h. Then cells were harvested, fixed in $70\%$ ice-cold ethanol for >1 h at 4°C, washed twice with cold PBS, and treated with 100 µg/mL ribonuclease, before staining with 50 µg/mL propidium iodide. Subsequently, the stained cells were measured by LSR Fortessa flow cytometer (BD, USA) and analyzed by Flowjo software.
## Apoptosis assay
FITC Annexin V and PI double staining was used to determine apoptosis (Apoptosis Detection Kit I, BD, USA). HCT116 cells were seeded into 6-well plates (5x105 cells/well), treated with 290 nM CS1 or DMSO (vehicle control) for 24 h and 48 h. According to the manufacturer’s protocol, the cells were harvested, washed with Annexin V binding buffer, and stained with PI and Annexin solution for 15 min at room temperature in the dark. The stained cells were measured by LSR Fortessa flow cytometer (BD, USA) and analyzed by Flowjo software.
## FTO knockdown in HCT116 cell line using Lentivirus
For lentivirus production, HEK293T cells were co-transfected with Lentiviral vector pLKO.1-shScr (Scramble control), pLKO.1-shFTO-A (Sigma) targeting FTO gene and packaging plasmids (TR30037, Origene) using TurboFectin transfection reagent (Origene). 24 h after transfection, media was changed and after another 48 h, cell supernatants were harvested, centrifuged, and filtered to collect lentiviral particles. Next, HCT116 cells were transduced with packaged lentiviral particles in the presence of polybrene (8 µg/ml) using the spinoculation method at 1000 g for 1 h. For stable cell line generation, transduced cells were selected with 2 μg/ml puromycin beginning 48 h after infection and maintained under selection for 2-3 passages.
## Western blot analysis
CRC cells were collected and lysed in protein extraction buffer (150 mM NaCl, 10 mM Tris, 1 mM EDTA, $1\%$ NP-40, 1 mM EGTA, and 50 mM NaF) containing protease inhibitor cocktail (Roche) on ice for 30 min, followed by centrifuge at 13,000 rpm at 4°C for 15 min. The supernatants were then collected, and the protein concentration was quantified by the BCA method. The cell lysates were subjected to SDS-PAGE, and subsequently transferred to PVDF membrane (Invitrogen). The membranes were probed with rabbit anti-human FTO (ab124892, Abcam; 1:1000 dilution) or rabbit anti-human β-actin (4970, Cell Signaling; 1:1000 dilution) antibodies, followed by goat anti-rabbit IgG (whole molecule) peroxidase conjugate (A6154, Sigma; 1:2000 dilution). Bioluminescence was catalyzed using a Quick Spray Chemiluminescent HRP Antibody Detection Reagent (Thomas Scientific, E2400), and bands were detected in a luminescent image analyzer PXi (Syngene).
## m6A dot blot assay
HCT116 cells were treated with DMSO or CS1 at 290nM (IC50) for 48 h. Total RNA was extracted using RNeasy Mini Kit (Qiagen), and poly (A)+ RNA was further enriched with PolyATract mRNA isolation System IV (Promega) according to the manufacturer’s instructions. The RNA samples were diluted in RNA binding buffer, and denatured at 65°C for 5 min. Then one volume of 20X SSC buffer was added into the RNA samples before dotted onto the Amersham Hybond-N+ membrane (GE Healthcare) with Bio-Dot Apparatus (Bio-Rad). The RNA samples were cross-linked onto the membrane via UV irradiation for 5 min. The membrane was stained with $0.02\%$ methylene blue (MB) as loading control. After that, the membrane was washed with 1X PBS-T buffer, blocked with $5\%$ nonfat dry milk and incubated with rabbit anti-m6A antibody (202003, Synaptic Systems, 1:2000 dilution) at 4°C overnight. Finally, the membrane was incubated with the HRP-conjugated goat anti-rabbit IgG (sc-2030, Santa Cruz Biotechnology) and developed with Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare).
## Tumor challenge and therapy
Stably transduced HCT116 shScr or HCT116 shFTO-A cell lines (5 × 105) were suspended in PBS and injected subcutaneously into the right thigh of NSG mice ($$n = 3$$/group). Two weeks after implantation, the tumors were measured once a week by a caliper and tumor volume (mm3) was calculated using the formula $\frac{1}{2}$ x Length x Width x Depth.
For studying the therapeutic effect of CS1, HCT116 cells (5 × 105) were suspended in PBS implanted subcutaneously into the right thigh of NSG mice ($$n = 4$$-5/group). Three weeks after tumor inoculation, when the tumor sizes reached 50 mm3, the mice were randomized into 2 groups: control and CS1. Mice were administrated with total of 15 doses of CS1 (5 mg/kg) or control vehicle (β-cyclodextrin) every other day by intraperitoneal injection. Tumor volume (mm3) was measured once a week with a caliper until the tumor volume exceeded 1000 mm3 or any experimental endpoint, as pre-determined in the IACUC protocol.
## Cell migration and invasion
For migration assays, 5 × 104 cells (HCT116 shScr and HCT116 shFTO-A) were plated in the top chamber of the non-coated insert (Corning, USA) with 8 µm pore size. For invasion assays, 1 × 105 cells were plated in the top chamber of Matrigel-coated insert with 8 µm pore size. Cells were seeded in 200 μl low-serum ($1\%$ FBS) DMEM (upper chamber) and placed in 24-well plates (lower chamber) containing 800 μl of high-serum ($10\%$ FBS) DMEM. After 24 h of incubation at 37°C, the cells remaining in the upper side of the insert membrane were gently scraped off with cotton swabs. The cells that migrated or invaded through the pores to the lower surface of the insert membrane were fixed in $4\%$ formaldehyde, stained with $1\%$ crystal violet dye, and then counted under a microscope. The values were calculated by averaging the total number of cells from different microscopic fields of three chambers per condition.
## RNA sequencing and data analysis
Total RNA samples were isolated from HCT116 shScr or HCT116 shFTO-A cells with RNeasy Mini Kit (Qiagen, USA) for sequencing. RNA concentration was measured by NanoDrop 1000 (Thermo Fisher Scientific, USA) and RNA integrity was determined using Bioanalyzer (Agilent). Each RNA sample was spiked in with an appropriate amount of either Mix1 or Mix2 according to Life Technologies’ guidelines which would lead to about $1\%$ of the total number of RNA-Seq reads mapping to the 92 ERCC control sequences, assuming the mRNA fraction in the total RNA is $2\%$. Library construction of 300 ng total RNA for each sample was made using KAPA Stranded mRNA-Seq Kit (Illumina Platforms) (Kapa Biosystems, Wilmington, USA) with 10 cycles of PCR amplification. Libraries were purified using AxyPrep Mag PCR Clean-up kit (Thermo Fisher Scientific). Each library was quantified using a Qubit fluorometer (Life Technologies) and the size distribution assessed using the 2100 Bioanalyzer (Agilent Technologies, Santa Clara, USA). Sequencing was performed on an Illumina Hiseq 2500 (Illumina, San Diego, CA, USA) instrument using the TruSeq SR Cluster Kit V4-cBot-HS (Illumina) to generate 51 bp single-end reads sequencing with v4 chemistry. Quality control of RNA-Seq reads was performed using FastQC. Each group contains 3-4 replicates. Reads were trimmed for adaptor sequence, masked for low complexity or low-quality sequence by Cutadapt [33], and then aligned to reference genome GRCh38 by STAR [34]. The expressions of the genes were calculated using RSEM [35], $p \leq 0.05$ was set as the threshold of the differential expressions. The reads distributed in a specific transcript were displayed by IGV [36]. Hierarchical cluster analysis was generated by R package cluster. Gene Set Enrichment Analysis (GSEA) and hallmark gene sets in Molecular Signatures Database (MSigDB) [37] were applied for enriched pathways.
## Quantitative real-time PCR
To study the gene-downregulation related to FTO, HCT116 cells were treated with 290nM CS1 or DMSO as control for 48 h. Then, cells (control, CS1-treated HCT116, HCT116-shScr, and HCT116-shFTO-A) were collected and total RNAs were extracted using RNeasy Mini Kit (Qiagen, USA), according to manufacturers’ instructions. The cDNA was synthesized from RNA using RevertAid Reverse Transcriptase (Thermo Fisher Scientific, USA). Next, cDNA was amplified using the RT-PCR primer sets listed in the Table 1. PCRs were performed in QuantStudio 3 machine (Thermo Fisher Scientific, USA). Select Master Mix (Applied Biosystem, USA) was used to detect amplification under the following conditions: 2 m at 50°C, 2 m at 95°C followed by 40 cycles of 15 s at 95°C, and 60 s at 60°C. Results were analyzed with QuantStudio Analysis Software. HPRT was used as housekeeping gene to assess target gene.
**Table 1**
| Gene name | Primer sequence (5’ → 3’) |
| --- | --- |
| GAPDH | F: GCA CCG TCA AGG CTG AGAAC |
| | R: ATG GTG GTG AAG ACG CCAGT |
| EREG | F: GTG ATT CCA TCA TGT ATC CCA GGAG |
| | R: AGA TGC ACT GTC CAT GCA AACAA |
| KRAP | F: CAT ATG ACA GAG GAG GAC A |
| | R: GTG GCT GTC CTG CTT AGG |
| PDE4B | F: ATC TCA CGC TTT GGA GTC AAC |
| | R: TTA AGA CCC CAT TTG TTC AGG |
| KRAP | F: GAC GTG ATG AAC CAG ATA TTG CT |
| | R: TTG ACG AAA ACG GCT TGT TAA AG |
## The FTO inhibitor, CS1 inhibits cell proliferation of different human colorectal cancer cell lines in vitro
The cytotoxic effects of the small molecule FTO inhibitor, CS1 was examined in different CRC cell lines (HT-29, COLO, HCT-116, LoVo, SW480, SW620) using an MTS assay. As shown in Figure 1A, CS1 (50-3200nM) suppresses the proliferation of CRC cells in a dose-dependent manner after 72 h of treatment ($p \leq 0.05$). The most significant suppression was observed in HCT116 and SW620 cells ($16.23\%$ and $17.37\%$ at 3200 nM respectively). 5-FU-based chemotherapy is the standard approach for colon cancer treatment, and resistance to 5-FU is a major cause of therapeutic failure. Next, we explored if CS1 could inhibit cell growth in a 5-FU resistant HCT116 cell line (HCT116-5FUR). The MTS results indicated that CS1 also suppressed cell viability in a dose-dependent manner (50-3200nM) in HCT116-5FUR (Figure 1B) ($p \leq 0.05$ from 400nM). At a concentration of 3.2µM CS1, the proportion of cell toxicity was comparable in 5-FU resistant HCT116 and parental HCT116 cells. We then explored if the combination of 5-FU and CS1 treatment could enhance the inhibitory effect of each drug alone. HCT116 cells were treated with CS1 (50-800nM) and 5-FU (5-80µM) alone or in combination for 72 h and MTS assay was performed for cell viability. The results showed the similar percentages of cell viability at CS1 200-800nM and 5-FU 20-80µM as single agents and in combination ($37.17\%$ to $23.56\%$), with no findings of synergy (combination index >1 by CompuSyn program, data not shown) between these 2 drugs (Figure 1C). The protein expression of FTO in all CRC cells in this study were confirmed by Western blot (Figure 1D). To assess the inhibition of FTO by CS1 treatment on mRNA methylation, m6A dot plot assay was performed after HCT116 cells were treated by CS1 (290nM, IC50). The results showed a substantial increase of m6A abundance in transcriptomes of CS1-treated samples compared to the controls (Figure 1E). These data indicated that CS1 inhibited the demethylase activity of FTO protein in CS1-treated cells.
**Figure 1:** *Inhibitory effect of CS1 on different human colorectal cancer cell lines (CRC). (A) KRAS wild type (HT-29, COLO) and KRAS mutant (HCT116, LoVo, SW480, SW620) cell lines were seeded in 96 well-plates at a concentration of 1x104 cells/well. 24 h later, cells were treated with CS1 at increasing concentrations (0, 50, 100, 200, 400, 800, 1600 and 3200 nM). 72 h after treatment, MTS assay was applied to evaluate the percentage of cell viability. (B) Similar procedure was applied in the parental HCT116 cells and 5-FU resistant derivatives. HCT116 (C) were treated with CS1 (0, 50, 100, 200, 400, and 800 nM) and/or 5-FU (0, 5, 10, 20, 40, and 80 µM) as single agents or in combination. Results from one representative experiment are presented as means ± SD, with triplicate determinations. (*) p < 0.05; (**) p < 0.01; (***) p < 0.001; and (****) p < 0.0001. p-value vs control (untreated group) in each cell line. (D) CRC cells were collected, lysed, and total protein was obtained. Western blot analysis was performed to examine the expression of FTO. β-actin was used as a loading control. (E) Determination of m6A abundance in mRNA in HCT116 cells after 72 h of CS1 treatment via dot blot assay. MB (methylene blue) represents loading control of RNA samples.*
## CS1 induces cell cycle arrest in G2/M phase and promotes apoptosis
To investigate how CS1 influences the cell cycle, HCT116 cells were treated with 290nM CS1 or DMSO and then the cell cycle was analyzed using flow cytometry. As shown in Figures 2A, B, CS1 treatment resulted in a greater percentage of cells in G2/M phase compared to controls at 24 h ($62.8\%$ vs $27.0\%$) and 48 h ($69.6\%$ vs $23.9\%$) ($p \leq 0.0001$). CDC25C is one of the crucial regulators that dephosphorylates CDC2 and leads to the activation of the CDK1 complex at the G2/M checkpoint [38]. Western blot data indicated that CS1-treated cells demonstrated decreased expression of CDC25C (Figure 2E). These data suggest that CS1 may induces G2/M cell cycle arrest in HCT116 cells through downregulation of CDC25C leading to cell growth inhibition. Next, to investigate whether apoptosis contributed to the growth inhibition by CS1, we performed an apoptosis assay using FITC annexin V and propidium iodide double staining in HCT116 cells. There was a significantly higher proportion of total apoptotic cells observed in the CS1-treated groups vs the controls at 24 h ($13.8\%$ vs $9.9\%$) and 48 h ($10.7\%$ vs $9.2\%$) ($p \leq 0.05$) (Figures 2C, D). Overall, these data demonstrate that inhibition of cell growth in CS1-treated HCT116 cells is associated with G2/M cell cycle arrest, decreased CDC25C expression, and the induction of cell apoptosis.
**Figure 2:** *Effect of CS1 in cell cycle and apoptosis. HCT116 cells were seeded into 6-well plates (5x105 cells/well), treated with 290nM CS1 or DMSO (vehicle control) for 24 h and 48 (h) (A, B) For cell cycle analysis, cells were fixed in 70% ice-cold ethanol, treated with 100 µg/mL ribonuclease, stained with 50 µg/mL propidium iodide (PI) and then analyzed using flow cytometry. The percentage of cells in G0/G1, S and G2 phases of cell cycle was calculated and displayed. (C, D) For cell apoptosis assay, cells were double-stained with Annexin V-FITC+PI and assessed by flow cytometry. The percentages of early apoptotic cells (annexin V (+)/PI (−)), late apoptotic cells (annexin V (+)/PI (+)) and total apoptotic cells were calculated and displayed. (E) Western blot analysis of proteins related to G2/M phase (CDC25C) were performed after 48 h of CS1 treatment. β-actin was used as a protein loading control. Data are presented as mean ± SD. of three independent experiments. (*) p <0.05 and (**) p<0.0001..*
## FTO inhibition suppresses tumor growth in the HCT116 xenograft mouse model
To investigate the in vivo antitumor efficacy of the FTO inhibitor, CS1, a xenograft mouse model with HCT116 cells was used. After appropriate growth of tumors, the mice were randomized into 2 groups: control and CS1. As shown in Figure 3A, CS1 significantly inhibited tumor progression compared to vehicle, especially from day 21 post-treatment ($p \leq 0.05$). We also observed no significant differences in mice body weight between the CS1-treated group and controls, which suggested minimal toxicity of CS1 in vivo (Figure 3B). Consistent with the tumor growth curve, all the tumor masses from HCT116 tumor bearing mice treated with CS1 were found to be substantially smaller than the tumor masses from the control (Figures 3C, D).
**Figure 3:** *Therapeutic effect of CS1 treatment on HCT116 xenograft mouse model. 6-week-old female NSG mice (n = 4-5/group) were subcutaneously injected into the right thigh with HCT116 cells (5 × 105 cells/mouse). After 21 days, when the tumor volume reached around 50 mm3, the mice were randomized into 2 groups: control and CS1. Mice were administrated with total of 15 doses of CS1 (5 mg/kg) or control vehicle (β-cyclodextrin) every other day by intraperitoneal injection. Tumor volume was measured every 7 days until the end of experiment. Mice were then euthanized when the tumor volume reached 1000mm3. (A) Tumor growth curve. (B) Mouse weights. (C) Tumor weights and (D) the photographs of excised tumors at day 35 after the first treatment. Data are presented as means ± SD. (*) p <0.05 and (**) p<0.0001.*
## Knockdown of FTO by employment of Lentivirus-mediated shRNA in HCT116 cells
To study the role of FTO in colorectal cancer, we used Lentivirus-mediated shRNA targeting FTO to inhibit FTO expression. HCT116 cells were stably transduced to express an shRNA sequence (shFTO-A) that specifically silences FTO. The FTO-knockdown effects were confirmed through Western blot for protein expression. As shown in Figure 4A, shFTO-A transduced cell lines have lower expression of FTO protein compared to shScr transduced cells. Additionally, knockdown of FTO by shFTO-A markedly inhibited cell growth in HCT116 cells in vitro (Figure 4B, $p \leq 0.0001$). To further investigate the oncogenic role of FTO in colon cancer, we utilized a subcutaneous mouse model. Stable FTO knockdown (FTO KD) shFTO-A and control shScr transduced in HCT116 cells were injected into the right thigh of the mice. The tumor growth curve of the shFTO-A group was markedly less than control (Figures 4C, D). Furthermore, we assessed the effect of FTO KD on m6A mRNA levels. To assess the functional impact of FTO KD on FTO demethylase activity, we performed a m6A dot plot assay. Knockdown of FTO increased m6A mRNA levels, demonstrating inhibition of FTO demethylase activity (Figure 4E). Altogether, these findings suggested that FTO has an important role in cell proliferation, tumor progression and m6A demethylation in the HCT116 colon cancer cell line.
**Figure 4:** *Knockdown of FTO in HCT116 cell line using Lentivirus-mediated shRNA. HCT116 cells were transduced with shFTO-A (knockdown group) or shScr (negative control group) lentivirus. Three stably transduced cell lines of each clone were selected using 2 μg/ml puromycin. (A) For knockdown efficiency of FTO by shRNA lentivirus, Western blotting was performed to detect the FTO expression levels in FTO KD (shFTO-A) and negative control (shScr) cells. β-actin was used as a loading control. (B) Cell proliferations were examined using MTS assay at 24 h, 48 h and 72 (h) (C, D) The effect of FTO KD on tumor growth in vivo was confirmed by subcutaneous mouse model. 6-week-old female NSG mice (n = 3/group) were injected into the right thigh with HCT116 shScr or HCT116 shFTO-A cell lines (5 × 105). Two weeks after implantation, the tumors were measured once a week by a caliper and tumor volume (mm3) was calculated using the formula 1/2 x Length x Width x Depth for tumor growth curve (C). Pictures of tumor bearing mice were taken at day 14 and 28 after tumor inoculation (D). Data are presented as means ± SD. (*) p <0.01 and (**) p<0.0001. (E) Global m6A abundance of poly(A)+ RNA isolated from HCT116 shScr or HCT116 shFTO-A cell lines and detected by m6A dot plot assay (left panel). The membrane was stained with 0.02% methylene blue (MB) as loading control (right panel).*
## Knockdown of FTO inhibited in vitro migration and invasion of HCT116 cells
To evaluate the impact of FTO KD on cell migration and invasion, the transwell assay was performed on three stably transduced HCT116 cell lines with shFTO-A or negative control shScr. As shown in Figures 5A, C (upper panel), after 24hr of incubation, FTO KD cells by shFTO-A had a markedly lower percentage of migration compared to the negative controls ($p \leq 0.05$). Similar to the migration assay, shFTO-A demonstrated significantly less invasion compared to controls ($p \leq 0.01$) (Figures 5B, C, lower panel). These results suggest that shRNA-mediated inhibition of FTO expression decreases the capacity of migration and invasion of HCT116 cells.
**Figure 5:** *FTO knockdown inhibited the migration and invasion of HCT116 cells. Cell migration or invasion was assessed with a transwell assay. Stable FTO knockdown (shFTO-A) and negative control (shScr) of HCT116 cell lines were seeded in 1% FBS containing medium in non-coated inserts or in Matrigel-coated inserts (Corning, USA), and were placed in 24 well plates with 10% FBS containing medium. After 24 h, the cells that migrated or invaded through the pores to the lower surface of the insert membrane were fixed, stained, and counted under a microscope. (A) Percentage of migrated cells are shown. (B) Percentage of invaded cells are shown. (C) Representative images are shown. Data are presented as means ± SD from 3 fields. Scale = 50µm. (*) p <0.05 and (**) p<0.0001.*
## Proposed signaling pathways related to FTO knockdown
RNA sequencing (RNA-seq) was performed to understand the underlying mechanisms related to FTO KD. Distribution of RNA-seq reads around the genomic locus of FTO indicating a successful knockdown of FTO by shFTO (Figure 6A). Next, the hierarchical clustering dendrogram demonstrates that shFTO-A can be grouped together and these separate from the shScr controls (Figure 6B), which indicates the consistency and variance of the samples ($$n = 4$$/group). As compared to HCT116-shScr cells, HCT116-shFTO cells had 459 up-regulated genes and 192 down-regulated genes (Figure 6C). Top 15 differentially expressed genes have been shown in Figure 6D and top 4 down-regulated genes (EREG, KRAP, PDE4B and SLC38A2) were confirmed by qPCR (Figure 6E). There were no significant differences between untreated and shScr cells. Similar to FTO KD by shFTO, CS1 treatment markedly suppressed these 4 genes compared to controls (Figure 6E). A global gene set enrichment analysis (GSEA) revealed a set of downregulated or upregulated pathways in FTO KD compared to shScr (Figures 6F, G). There were various signaling pathways downregulated by FTO KD, including MYC target V1, MYC target V2, oxidative phosphorylation, reactive oxygen species, G2M checkpoint, mTORC1, and unfolded protein response (Figure 6F). On the other hand, FTO KD upregulated pathways involving Tumor Growth Factor beta, epithelial mesenchymal transition, KRAS signaling, estrogen response, UV response, myogenesis, and coagulation signaling (Figure 6G). These results indicate that FTO is involved in multiple signaling pathways; however, CS1 and FTO KD demonstrated commonalities of down-regulation of 4 key genes EREG, KRAP, PDE4B and SLC38A2.
**Figure 6:** *Proposed signaling pathways related to FTO knockdown. RNA-seq was performed on mRNA of shScr and shFTO-A from HCT116 cells. (A) Distribution of RNA‐seq reads around the genomic locus of FTO in shScr and shFTO using Integrative Genomics Viewer (IGV). (B) Hierarchical clustering dendrogram of RNA-seq data from controls and FTO KD. (C) Volcano plot represents differentially expressed genes in FTO KD HCT116 cells compared to shScr control. The red dots and green dots indicate significantly upregulated and downregulated genes, respectively. The blue dots indicate insignificant differentially expressed genes. False Discovery Rate (FDR) is an adjusted p-value for multiple tests (by the Benjamini-Hochberg procedure) by giving the proportion of tests above threshold that will be false positives. FC=fold change. (D) Top 15 differentially expressed genes using Ingenuity Pathway Analysis (Qiagen). (E) Top 4 downregulated genes from control, CS1-treated, shScr and shFTO HCT116 cells by RT-PCR. Data are presented as means ± SD from 3 fields. (*) p<0.0001. (F, G) Scattergrams of the downregulated pathways (F) and upregulated pathways (G) based on GSEA.*
## Discussion
FTO has been shown to play an important role in modulating fat mass, adipogenesis, and total body weight (39–41). Epidemiologic studies demonstrated that FTO single nucleotide polymorphisms (SNPs) are associated with the increased obesity and higher risk of multiple cancers including colorectal cancer [42]. Specifically, there is a positive association between colorectal cancer and rs1558902, rs8050136, rs3751812, rs9939609 FTO SNPs in Japanese population [43] and the A allele of rs9939609 FTO SNP in Iranian population [44, 45]. FTO, the first described demethylase of m6A mRNA, has been reported as an oncogene in different types of cancers. FTO is also found markedly upregulated in colorectal adenocarcinoma tissues [46]. However, the role of FTO in colorectal cancer has not been fully investigated. In the present study, we show that FTO is expressed in various human colorectal cancer cell lines. We demonstrated that knockdown of FTO decreased cell proliferation, migration, invasion in vitro and suppressed tumor progression in vivo, which suggests that FTO may have an oncogenic role in colorectal cancer.
Since the 1980s, CS1 (or bisantrene) has shown some response in clinical trials as an anthracene compound for many types of cancer [47, 48]. Recently, CS1 has been found to have high therapeutic efficacy against acute myeloid leukemia cells [32]. In line with these findings, in our study, CS1 suppressed cell proliferation, induced cell cycle arrest in G2/M phase and promoted cell apoptosis in HCT116 cells in vitro. Moreover, CS1 treatment also inhibited tumor growth with minimal toxicity in colon cancer mouse models. 5-FU is the most common chemotherapeutic for CRC [49]. However, the clinical efficacy is decreased in 5-FU resistant cells [50]. Here, CS1 suppressed HCT116-5FUR cell viability in a dose-dependent manner (50-3200 nM, $p \leq 0.05$ from 400 nM). These data suggested that CS1 might serve as a single alternative agent or in combination to overcome the resistance of 5-FU based therapies for CRC.
Using next-generation sequencing technology (NGS) with RNA-seq we investigated variations at the transcriptome level and differentially expressed genes (DEGs) of FTO KD in CRC. RNA-seq analysis of FTO KD revealed the distinct underlying molecular mechanisms and signaling pathways associated with antitumor activities in colorectal cancer cells through a set of differentially expressed genes including the top 4 down-regulated genes (EREG, KRAP, PDE4B and SLC38A2) confirmed by qPCR (Figures 6D, E). Epiregulin (EREG) is a ligand of epidermal growth factor receptor (EGFR), which is involved in RAS-RAF-MAPK and PI3K-AKT-mTOR signaling pathways regulating tumor proliferation, invasion, and migration [51]. EREG is overexpressed in many types of cancer including colorectal cancer [52]. Nearly $50\%$ of colorectal cancers harbor KRAS mutations [53]. CRC cells with mutant KRAS have been found to express higher autocrine levels of high-affinity EGFR ligands compared to wild-type KRAS [54]. This strategy would be advantageous by targeting EREG in the $30\%$ to $50\%$ of CRC patients that harbor a KRAS mutation. KRAS-induced actin-interacting protein (KRAP), also named as actin-interacting protein sperm-specific antigen 2 (SSFA2), was originally identified as one of the genes that was up-regulated by activated KRAS in HCT116 cells [55]. KRAP contributed to the regulation of filamentous actin and signals from the outside of the cells [56]. Importantly, KRAP has been shown to be involved in cell proliferation in glioma [57] and oral squamous cell carcinoma [58]. Those studies suggested that KRAP may serve as a potential target for colon cancer and other cancers. Our work demonstrates that FTO inhibition via FTO KD or via CS1 treatment leads to down-regulation of KRAP and may account for the mechanism of growth inhibition.
PDE4B belongs to the phosphodiesterase (PDE) family that catalyzes the hydrolysis of cyclic adenosine 3′,5’ monophosphate (cAMP) to AMP. cAMP is a ubiquitous second messenger and activated through its binding to activated protein kinase A (PKA), which is associated with various cellular processes including proliferation, differentiation, migration, and apoptosis [59]. PDE4 is highly expressed in many kinds of cancer including colon cancer, melanoma, lymphoma, glioma, ovarian, brain tumors, and non-small cell lung cancer [60]. Moreover, the expression of PDE4B is upregulated by oncogenic KRAS in HCT116 cells [61]. PDE4B modulates the expression of MYC that leads to low intracellular cAMP levels, activates AKT/mTOR signaling, and promotes cell survival in colorectal cancer [62]. These findings suggest that PDE4B may play a role in colon cancer and inhibition of PDE4B is a potential target for anticancer therapy We have demonstrated that FTO KD and inhibition with CS1 down-regulates PDE4B accounting for a possible mechanism of action of growth inhibition.
In order to promote proliferation and metastasis, tumor cells take up high levels of extracellular amino acids including glutamine [63]. Glutamine is imported into cells via transporters, such as the Na+-coupled neutral amino acid transporters (SNATs) or the SLC38 superfamily [64]. Among those transporters, SLC38A2 or SNAT2 have been reported to be overexpressed in tumors including prostate cancer [65], breast cancer [66], pancreatic cancer [67], and colorectal cancer [68]. KRAS as one of the most prevalently mutated oncogenes in CRCs [69], has been found to upregulate glutamine transport, metabolism, and cell proliferation through mTOR activation leading to drug resistance [70, 71]. Knockdown of amino acid transporters like SLC38A2, inhibit amino acid uptake and cell proliferation via mTOR suppression [68]. This would be another effective strategy for colorectal cancer treatment. In this study we have demonstrated that FTO inhibition with CS1 or FTO KD leads to down regulation of SLC38A2.
There are multiple hallmarks of cancer during development of tumors including promoting cellular proliferation [72]. MYC is a transcription factor regulating groups of genes (MYC targets) related to enhance cell growth in most types of human cancers [73]. Consistently, our study revealed different signaling pathways downregulated by FTO KD, including MYC target V1 and MYC target V2. In CRCs, c−MYC has been shown to have a role in self−renewal, tumorigenicity, invasion and chemoresistance of cancer stem cells [74]. In anti-EGFR targeted therapy, patients with high c-MYC expression had a markedly lower PFS, OS and more frequent metastases compared to patients with low c-MYC expression, which suggests a pivotal role of c-MYC in CRC resistance to EGFR inhibitors [75]. Our RNA-seq data suggests that FTO inhibition leads to down regulation of the MYC pathway suggesting an additional potential mechanism of action.
*In* general, during malignant transformation, cancer cells undergo metabolic reprogramming to produce a huge amount of energy and biomass including a metabolic shift towards aerobic glycolysis characterized as the Warburg effect [76]. On the other hand, oxidative phosphorylation (OXPHOS) is also a crucial pathway of cancer metabolism which supports progression and invasiveness in CRC [77, 78]. OXPHOS has been found to be upregulated in CRC cells compared to healthy surrounding tissues, while the levels of glycolysis remained unchanged [79]. A metabolic shift towards OXPHOS was linked to chemoresistance in CRC [77]. Interestingly, KRAS mutations might be associated with an oxidative phenotype, while BRAF mutations with a glycolytic phenotype [80]. In line with this, our study shows in our KRAS mutated cell line HCT116, FTO KD led to down regulation in oxidative phosphorylation pathways as seen by RNA-seq and no significant difference in glycolysis after treatment with the FTO inhibitor CS1 (Seahorse assay, data not shown). This data indicates that FTO KD impacts OXPHOS perhaps more than glycolysis.
Lastly, cell cycle regulation plays a key role in cell proliferation and in the progression of cancer via cell cycle-associated signaling pathways [81, 82]. Our RNA-seq results indicated that the G2/M checkpoint was a downregulated pathway, in which CS1-treated HCT116 cells were induced into G2/M cell cycle arrest and subsequently the induction of cell apoptosis. Importantly, CS1 treatment suppressed the expression of CDC25C, one of the important regulators in the G2/M phase.
In summary, we identified FTO as a potential oncogene in colorectal cancer cells and we demonstrated that targeting FTO significantly suppressed cancer cell proliferation, migration and invasion in vitro and tumor progression in vivo. Additionally, we found that FTO inhibition impacts multiple pathways that may account for its role as an oncogene. Importantly, the FTO inhibitor, CS1 can be applied as a potential therapeutic agent for CRC treatment (Figure 7). Further work is needed to explore the roles of the most prominent downstream transcripts and signaling pathways that are impacted with changes in FTO expression and activity. This will lead to future clinical trials targeting FTO as a potential therapeutic strategy in the treatment of metastatic colorectal cancer.
**Figure 7:** *Targeting FTO in CRC treatment. FTO was identified as a potential oncogene in colorectal cancer cells and inhibition of FTO (by shFTO or CS1) significantly suppressed cancer cell proliferation, migration and invasion in vitro and tumor progression in vivo. FTO inhibition impacts multiple pathways related to EREG, KRAP, PDE4B and SLC38A2 that may account for its role as an oncogene. CS1, the FTO inhibitor, can be applied as a potential therapeutic agent for CRC treatment.*
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The animal study was reviewed and approved by City of Hope IACUC protocol #16067.
## Author contributions
TP designed the experiments, performed experiments, analyzed the data, and prepared the manuscript. LM designed the experiments, provided supervision, prepared, and finalized the manuscript for submission. VN supported for the in vivo experiments. RS and JC provided the ideas and suggestions for FTO inhibitor in cancer. YL performed the m6A dot plot assay. YQ supported for the Seahorse assay. HQ, HC, and XW performed RNA-seq and data analysis. LJ, MF, DD and AG contributed to experimental design, data analysis and manuscript preparation. All the 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. Siegel RL, Miller KD, Goding Sauer A, Fedewa SA, Butterly LF, Anderson JC. **Colorectal cancer statistics, 2020**. *CA Cancer J Clin* (2020) **70**. DOI: 10.3322/caac.21601
2. Siegel RL, Miller KD, Jemal A. **Cancer statistics, 2020**. *CA Cancer J Clin* (2020) **70** 7-30. DOI: 10.3322/caac.21590
3. Yu T, Guo F, Yu Y, Sun T, Ma D, Han J. **Fusobacterium nucleatum promotes chemoresistance to colorectal cancer by modulating autophagy**. *Cell* (2017) **170** 548-63.e16. DOI: 10.1016/j.cell.2017.07.008
4. Xie YH, Chen YX, Fang JY. **Comprehensive review of targeted therapy for colorectal cancer**. *Signal transduction targeted Ther* (2020) **5** 22. DOI: 10.1038/s41392-020-0116-z
5. Boccaletto P, Machnicka MA, Purta E, Piatkowski P, Baginski B, Wirecki TK. **Modomics: A database of rna modification pathways. 2017 update**. *Nucleic Acids Res* (2018) **46**. DOI: 10.1093/nar/gkx1030
6. Frye M, Harada BT, Behm M, He C. **Rna modifications modulate gene expression during development**. *Science* (2018) **361**. DOI: 10.1126/science.aau1646
7. Schumann U, Shafik A, Preiss T. **Mettl3 gains R/W access to the epitranscriptome**. *Mol Cell* (2016) **62**. DOI: 10.1016/j.molcel.2016.04.024
8. 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**. DOI: 10.1038/nchembio.1432
9. Ping X-L, Sun B-F, Wang L, Xiao W, Yang X, Wang W-J. **Mammalian wtap is a regulatory subunit of the rna N6-methyladenosine methyltransferase**. *Cell Res* (2014) **24**. DOI: 10.1038/cr.2014.3
10. 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**. DOI: 10.1038/nchembio.687
11. 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. DOI: 10.1016/j.molcel.2012.10.015
12. Wang X, Lu Z, Gomez A, Hon GC, Yue Y, Han D. **N6-Methyladenosine-Dependent regulation of messenger rna stability**. *Nature* (2014) **505**. DOI: 10.1038/nature12730
13. Alarcón CR, Goodarzi H, Lee H, Liu X, Tavazoie S, Tavazoie SF. **Hnrnpa2b1 is a mediator of M(6)a-dependent nuclear rna processing events**. *Cell* (2015) **162**. DOI: 10.1016/j.cell.2015.08.011
14. Huang H, Weng H, Sun W, Qin X, Shi H, Wu H. **Recognition of rna N(6)-methyladenosine by Igf2bp proteins enhances mrna stability and translation**. *Nat Cell Biol* (2018) **20**. DOI: 10.1038/s41556-018-0045-z
15. Luo J, Yu J, Peng X. **Could partial nonstarch polysaccharides ameliorate cancer by altering M(6)a rna methylation in hosts through intestinal microbiota**. *Crit Rev Food Sci Nutr* (2022) **62**. DOI: 10.1080/10408398.2021.1927975
16. Qiu FS, He JQ, Zhong YS, Guo MY, Yu CH. **Implications of M6a methylation and microbiota interaction in non-small cell lung cancer: From basics to therapeutics**. *Front Cell infection Microbiol* (2022) **12**. DOI: 10.3389/fcimb.2022.972655
17. Chen S, Zhang L, Li M, Zhang Y, Sun M, Wang L. **Fusobacterium nucleatum reduces Mettl3-mediated M(6)a modification and contributes to colorectal cancer metastasis**. *Nat Commun* (2022) **13** 1248. DOI: 10.1038/s41467-022-28913-5
18. Toh JDW, Crossley SWM, Bruemmer KJ, Ge EJ, He D, Iovan DA. **Distinct rna n-demethylation pathways catalyzed by nonheme iron Alkbh5 and fto enzymes enable regulation of formaldehyde release rates**. *Proc Natl Acad Sci USA* (2020) **117**. DOI: 10.1073/pnas.2007349117
19. Li Z, Weng H, Su R, Weng X, Zuo Z, Li C. **Fto plays an oncogenic role in acute myeloid leukemia as a N(6)-methyladenosine rna demethylase**. *Cancer Cell* (2017) **31**. DOI: 10.1016/j.ccell.2016.11.017
20. Xu D, Shao W, Jiang Y, Wang X, Liu Y, Liu X. **Fto expression is associated with the occurrence of gastric cancer and prognosis**. *Oncol Rep* (2017) **38**. DOI: 10.3892/or.2017.5904
21. Niu Y, Lin Z, Wan A, Chen H, Liang H, Sun L. **Rna N6-methyladenosine demethylase fto promotes breast tumor progression through inhibiting Bnip3**. *Mol Cancer* (2019) **18** 46. DOI: 10.1186/s12943-019-1004-4
22. Yang S, Wei J, Cui Y-H, Park G, Shah P, Deng Y. **M6a mrna demethylase fto regulates melanoma tumorigenicity and response to anti-Pd-1 blockade**. *Nat Commun* (2019) **10** 2782. DOI: 10.1038/s41467-019-10669-0
23. Zou D, Dong L, Li C, Yin Z, Rao S, Zhou Q. **The M(6)a eraser fto facilitates proliferation and migration of human cervical cancer cells**. *Cancer Cell Int* (2019) **19** 321. DOI: 10.1186/s12935-019-1045-1
24. Zhuang C, Zhuang C, Luo X, Huang X, Yao L, Li J. **N6-methyladenosine demethylase fto suppresses clear cell renal cell carcinoma through a novel fto-Pgc-1α signalling axis**. *J Cell Mol Med* (2019) **23**. DOI: 10.1111/jcmm.14128
25. Rong ZX, Li Z, He JJ, Liu LY, Ren XX, Gao J. **Downregulation of fat mass and obesity associated (Fto) promotes the progression of intrahepatic cholangiocarcinoma**. *Front Oncol* (2019) **9**. DOI: 10.3389/fonc.2019.00369
26. Liu X, Liu J, Xiao W, Zeng Q, Bo H, Zhu Y. **Sirt1 regulates N(6) -methyladenosine rna modification in hepatocarcinogenesis by inducing Ranbp2-dependent fto sumoylation**. *Hepatol (Baltimore Md)* (2020) **72**. DOI: 10.1002/hep.31222
27. Huang H, Wang Y, Kandpal M, Zhao G, Cardenas H, Ji Y. **(6)-methyladenosine modifications inhibit ovarian cancer stem cell self-renewal by blocking camp signaling**. *Cancer Res* (2020) **80**. DOI: 10.1158/0008-5472.Can-19-4044
28. Lin Z, Wan AH, Sun L, Liang H, Niu Y, Deng Y. **N6-methyladenosine demethylase FTO enhances chemo-resistance in colorectal cancer through SIVA1-mediated apoptosis**. *Mol Ther* (2023) **31**. DOI: 10.1016/j.ymthe.2022.10.012
29. Wang J, Qiao Y, Sun M, Sun H, Xie F, Chang H. **Fto promotes colorectal cancer progression and chemotherapy resistance**. *Clin Trans Med* (2022) **12** e772. DOI: 10.1002/ctm2.772
30. Ruan DY, Li T, Wang YN, Meng Q, Li Y, Yu K. **Fto downregulation mediated by hypoxia facilitates colorectal cancer metastasis**. *Oncogene* (2021) **40**. DOI: 10.1038/s41388-021-01916-0
31. Relier S, Ripoll J, Guillorit H, Amalric A, Achour C, Boissière F. **Fto-mediated cytoplasmic M(6)a(M) demethylation adjusts stem-like properties in colorectal cancer cell**. *Nat Commun* (2021) **12** 1716. DOI: 10.1038/s41467-021-21758-4
32. Su R, Dong L, Li Y, Gao M, Han L, Wunderlich M. **Targeting fto suppresses cancer stem cell maintenance and immune evasion**. *Cancer Cell* (2020) **38** 79-96.e11. DOI: 10.1016/j.ccell.2020.04.017
33. Martin M. **Cutadapt removes adapter sequences from high-throughput sequencing reads**. *EMBnet J* (2011) **17**. DOI: 10.14806/ej.17.1.200
34. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S. **Star: Ultrafast universal rna-seq aligner**. *Bioinformatics* (2013) **29** 15-21. DOI: 10.1093/bioinformatics/bts635
35. Li B, Dewey CN. **Rsem: Accurate transcript quantification from rna-seq data with or without a reference genome**. *BMC Bioinf* (2011) **12**. DOI: 10.1186/1471-2105-12-323
36. Thorvaldsdóttir H, Robinson JT, Mesirov JP. **Integrative genomics viewer (Igv): High-performance genomics data visualization and exploration**. *Brief Bioinform* (2013) **14**. DOI: 10.1093/bib/bbs017
37. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA. **Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles**. *Proc Natl Acad Sci USA* (2005) **102**. DOI: 10.1073/pnas.0506580102
38. DiPaola RS. **To arrest or not to G2-m cell-cycle arrest**. *Clin Cancer Res* (2002) **8** 3311. PMID: 12429616
39. Fischer J, Koch L, Emmerling C, Vierkotten J, Peters T, Brüning JC. **Inactivation of the fto gene protects from obesity**. *Nature* (2009) **458**. DOI: 10.1038/nature07848
40. McMurray F, Church CD, Larder R, Nicholson G, Wells S, Teboul L. **Adult onset global loss of the fto gene alters body composition and metabolism in the mouse**. *PloS Genet* (2013) **9** e1003166. DOI: 10.1371/journal.pgen.1003166
41. Merkestein M, Laber S, McMurray F, Andrew D, Sachse G, Sanderson J. **Fto influences adipogenesis by regulating mitotic clonal expansion**. *Nat Commun* (2015) **6** 6792. DOI: 10.1038/ncomms7792
42. Deng X, Su R, Stanford S, Chen J. **Critical enzymatic functions of fto in obesity and cancer**. *Front Endocrinol (Lausanne)* (2018) **9**. DOI: 10.3389/fendo.2018.00396
43. Yamaji T, Iwasaki M, Sawada N, Shimazu T, Inoue M, Tsugane S. **Fat mass and obesity-associated gene polymorphisms, pre-diagnostic plasma adipokine levels and the risk of colorectal cancer: The Japan public health center-based prospective study**. *PloS One* (2020) **15** e0229005. DOI: 10.1371/journal.pone.0229005
44. Gholamalizadeh M, Akbari ME, Doaei S, Davoodi SH, Bahar B, Tabesh GA. **The association of fat-mass-and obesity-associated gene polymorphism (Rs9939609) with colorectal cancer: A case-control study**. *Front Oncol* (2021) **11**. DOI: 10.3389/fonc.2021.732515
45. Gholamalizadeh M, Tabrizi R, Bourbour F, Rezaei S, Pourtaheri A, Badeli M. **Are the fto gene polymorphisms associated with colorectal cancer? a meta-analysis**. *J Gastrointest Cancer* (2021) **52**. DOI: 10.1007/s12029-021-00651-9
46. Shen XP, Ling X, Lu H, Zhou CX, Zhang JK, Yu Q. **Low expression of microrna-1266 promotes colorectal cancer progression**. *Eur Rev Med Pharmacol Sci* (2018) **22**. DOI: 10.26355/eurrev_201812_16516
47. Cowan JD, Gehan E, Rivkin SE, Jones SE. **Phase ii trial of bisantrene in patients with advanced sarcoma: A southwest oncology group study**. *Cancer Treat Rep* (1986) **70**
48. Miller TP, Cowan JD, Neilan BA, Jones SE. **A phase ii study of bisantrene in malignant lymphomas. a southwest oncology group study**. *Cancer Chemother Pharmacol* (1986) **16**. DOI: 10.1007/bf00255289
49. Longley DB, Harkin DP, Johnston PG. **5-fluorouracil: Mechanisms of action and clinical strategies**. *Nat Rev Cancer* (2003) **3**. DOI: 10.1038/nrc1074
50. Blondy S, David V, Verdier M, Mathonnet M, Perraud A, Christou N. **5-fluorouracil resistance mechanisms in colorectal cancer: From classical pathways to promising processes**. *Cancer Sci* (2020) **111**. DOI: 10.1111/cas.14532
51. Zhao B, Wang L, Qiu H, Zhang M, Sun L, Peng P. **Mechanisms of resistance to anti-egfr therapy in colorectal cancer**. *Oncotarget* (2017) **8** 3980-4000. DOI: 10.18632/oncotarget.14012
52. Kuramochi H, Nakajima G, Kaneko Y, Nakamura A, Inoue Y, Yamamoto M. **Amphiregulin and epiregulin mrna expression in primary colorectal cancer and corresponding liver metastases**. *BMC Cancer* (2012) **12**. DOI: 10.1186/1471-2407-12-88
53. Xie J, Xia L, Xiang W, He W, Yin H, Wang F. **Metformin selectively inhibits metastatic colorectal cancer with the kras mutation by intracellular accumulation through silencing Mate1**. *Proc Natl Acad Sci USA* (2020) **117** 13012. DOI: 10.1073/pnas.1918845117
54. Lee HW, Son E, Lee K, Lee Y, Kim Y, Lee JC. **Promising therapeutic efficacy of Gc1118, an anti-egfr antibody, against kras mutation-driven colorectal cancer patient-derived xenografts**. *Int J Mol Sci* (2019) **20** 5894. DOI: 10.3390/ijms20235894
55. Inokuchi J, Komiya M, Baba I, Naito S, Sasazuki T, Shirasawa S. **Deregulated expression of krap, a novel gene encoding actin-interacting protein, in human colon cancer cells**. *J Hum Genet* (2004) **49** 46-52. DOI: 10.1007/s10038-003-0106-3
56. Fujimoto T, Koyanagi M, Baba I, Nakabayashi K, Kato N, Sasazuki T. **Analysis of krap expression and localization, and genes regulated by krap in a human colon cancer cell line**. *J Hum Genet* (2007) **52**. DOI: 10.1007/s10038-007-0204-8
57. Zhu A, Li X, Wu H, Miao Z, Yuan F, Zhang F. **Molecular mechanism of Ssfa2 deletion inhibiting cell proliferation and promoting cell apoptosis in glioma**. *Pathol Res Pract* (2019) **215**. DOI: 10.1016/j.prp.2018.12.035
58. Zhu L, Zhang L, Tang Y, Zhang F, Wan C, Xu L. **Microrna−363−3p inhibits tumor cell proliferation and invasion in oral squamous cell carcinoma cell lines by targeting Ssfa2**. *Exp Ther Med* (2021) **21** 549. DOI: 10.3892/etm.2021.9981
59. Chin KV, Yang WL, Ravatn R, Kita T, Reitman E, Vettori D. **Reinventing the wheel of cyclic amp: Novel mechanisms of camp signaling**. *Ann N Y Acad Sci* (2002) **968** 49-64. DOI: 10.1111/j.1749-6632.2002.tb04326.x
60. Mehta A, Patel BM. **Therapeutic opportunities in colon cancer: Focus on phosphodiesterase inhibitors**. *Life Sci* (2019) **230**. DOI: 10.1016/j.lfs.2019.05.043
61. Tsunoda T, Ota T, Fujimoto T, Doi K, Tanaka Y, Yoshida Y. **Inhibition of phosphodiesterase-4 (Pde4) activity triggers luminal apoptosis and akt dephosphorylation in a 3-d colonic-crypt model**. *Mol Cancer* (2012) **11**. DOI: 10.1186/1476-4598-11-46
62. Kim DU, Kwak B, Kim SW. **Phosphodiesterase 4b is an effective therapeutic target in colorectal cancer**. *Biochem Biophys Res Commun* (2019) **508**. DOI: 10.1016/j.bbrc.2018.12.004
63. Rubin H. **Deprivation of glutamine in cell culture reveals its potential for treating cancer**. *Proc Natl Acad Sci USA* (2019) **116**. DOI: 10.1073/pnas.1815968116
64. Kandasamy P, Gyimesi G, Kanai Y, Hediger MA. **Amino acid transporters revisited: New views in health and disease**. *Trends Biochem Sci* (2018) **43**. DOI: 10.1016/j.tibs.2018.05.003
65. Okudaira H, Shikano N, Nishii R, Miyagi T, Yoshimoto M, Kobayashi M. **Putative transport mechanism and intracellular fate of trans-1-Amino-3-18f-Fluorocyclobutanecarboxylic acid in human prostate cancer**. *J Nucl Med* (2011) **52**. DOI: 10.2967/jnumed.110.086074
66. Morotti M, Bridges E, Valli A, Choudhry H, Sheldon H, Wigfield S. **Hypoxia-induced switch in Snat2/Slc38a2 regulation generates endocrine resistance in breast cancer**. *Proc Natl Acad Sci USA* (2019) **116**. DOI: 10.1073/pnas.1818521116
67. Parker SJ, Amendola CR, Hollinshead KER, Yu Q, Yamamoto K, Encarnación-Rosado J. **Selective alanine transporter utilization creates a targetable metabolic niche in pancreatic cancer**. *Cancer Discov* (2020) **10**. DOI: 10.1158/2159-8290.CD-19-0959
68. Kandasamy P, Zlobec I, Nydegger DT, Pujol-Giménez J, Bhardwaj R, Shirasawa S. **Oncogenic kras mutations enhance amino acid uptake by colorectal cancer cells**. *Mol Oncol* (2021) **15**. DOI: 10.1002/1878-0261.12999
69. Knickelbein K, Zhang L. **Mutant kras as a critical determinant of the therapeutic response of colorectal cancer**. *Genes Dis* (2015) **2** 4-12. DOI: 10.1016/j.gendis.2014.10.002
70. Yoo HC, Yu YC, Sung Y, Han JM. **Glutamine reliance in cell metabolism**. *Exp Mol Med* (2020) **52**. DOI: 10.1038/s12276-020-00504-8
71. Scalise M, Pochini L, Galluccio M, Console L, Indiveri C. **Glutamine transporters as pharmacological targets: From function to drug design**. *Asian J Pharm Sci* (2020) **15**. DOI: 10.1016/j.ajps.2020.02.005
72. Hanahan D, Weinberg Robert A. **Hallmarks of cancer: The next generation**. *Cell* (2011) **144**. DOI: 10.1016/j.cell.2011.02.013
73. Dang CV, O’Donnell KA, Zeller KI, Nguyen T, Osthus RC, Li F. **The c-myc target gene network**. *Semin Cancer Biol* (2006) **16**. DOI: 10.1016/j.semcancer.2006.07.014
74. Zhang HL, Wang P, Lu MZ, Zhang SD, Zheng L. **C−Myc maintains the Self−Renewal and chemoresistance properties of colon cancer stem cells**. *Oncol Lett* (2019) **17**. DOI: 10.3892/ol.2019.10081
75. Strippoli A, Cocomazzi A, Basso M, Cenci T, Ricci R, Pierconti F. **C-myc expression is a possible keystone in the colorectal cancer resistance to egfr inhibitors**. *Cancers* (2020) **12**. DOI: 10.3390/cancers12030638
76. Koppenol WH, Bounds PL, Dang CV. **Otto Warburg’s contributions to current concepts of cancer metabolism**. *Nat Rev Cancer* (2011) **11**. DOI: 10.1038/nrc3038
77. Denise C, Paoli P, Calvani M, Taddei ML, Giannoni E, Kopetz S. **5-fluorouracil resistant colon cancer cells are addicted to oxphos to survive and enhance stem-like traits**. *Oncotarget* (2015) **6**. DOI: 10.18632/oncotarget.5991
78. Lin CS, Liu LT, Ou LH, Pan SC, Lin CI, Wei YH. **Role of mitochondrial function in the invasiveness of human colon cancer cells**. *Oncol Rep* (2018) **39**. DOI: 10.3892/or.2017.6087
79. Chekulayev V, Mado K, Shevchuk I, Koit A, Kaldma A, Klepinin A. **Metabolic remodeling in human colorectal cancer and surrounding tissues: Alterations in regulation of mitochondrial respiration and metabolic fluxes**. *Biochem Biophys Rep* (2015) **4**. DOI: 10.1016/j.bbrep.2015.08.020
80. Rebane-Klemm E, Truu L, Reinsalu L, Puurand M, Shevchuk I, Chekulayev V. **Mitochondrial respiration in kras and braf mutated colorectal tumors and polyps**. *Cancers* (2020) **12**. DOI: 10.3390/cancers12040815
81. Williams GH, Stoeber K. **The cell cycle and cancer**. *J Pathol* (2012) **226**. DOI: 10.1002/path.3022
82. Otto T, Sicinski P. **Cell cycle proteins as promising targets in cancer therapy**. *Nat Rev Cancer* (2017) **17** 93-115. DOI: 10.1038/nrc.2016.138
|
---
title: Alleviation of cognitive deficits in a rat model of glutamate-induced excitotoxicity,
using an N-type voltage-gated calcium channel ligand, extracted from Agelena labyrinthica
crude venom
authors:
- Mohammad Keimasi
- Kowsar Salehifard
- Mohammadjavad Keimasi
- Mohammadreza Amirsadri
- Noushin Mirshah Jafar Esfahani
- Majid Moradmand
- Fariba Esmaeili
- Mohammad Reza Mofid
journal: Frontiers in Molecular Neuroscience
year: 2023
pmcid: PMC9981952
doi: 10.3389/fnmol.2023.1123343
license: CC BY 4.0
---
# Alleviation of cognitive deficits in a rat model of glutamate-induced excitotoxicity, using an N-type voltage-gated calcium channel ligand, extracted from Agelena labyrinthica crude venom
## Abstract
Excitotoxicity is a common pathological process in Alzheimer’s disease (AD) which is caused by the over-activity of N-Methyl-D-Aspartate receptors (NMDARs). The release of neurotransmitters depends on the activity of voltage-gated calcium channels (VGCCs). Hyper-stimulation of NMDARs can enhance the releasement of neurotransmitters through the VGCCs. This malfunction of channels can be blocked by selective and potent N-type VGCCs ligand. Under excitotoxicity condition, glutamate has negative effects on the pyramidal cells of the hippocampus, which ends in synaptic loss and elimination of these cells. These events leads to learning and memory elimination through the hippocampus circuit’s dysfunction. A suitable ligand has a high affinity to receptor or channel and is selective for its target. The bioactive small proteins of venom have these characteristics. Therefore, peptides and small proteins of animal venom are precious sources for pharmacological applications. The omega-agatoxin-Aa2a was purified, and identified from *Agelena labyrinthica* specimens, as an N-type VGCCs ligand for this study. The effect of the omega-agatoxin-Aa2a on the glutamate-induced excitotoxicity in rats was evaluated through behavioral tests including Morris Water Maze, and Passive avoidance. The syntaxin1A (SY1A), synaptotagmin1 (SYT1), and synaptophysin (SYN) genes expression were measured via Real-Time PCR. The local expression of synaptosomal-associated protein, 25 k Da (SNAP-25) was visualized using an immunofluorescence assay for synaptic quantification. Electrophysiological amplitude of field excitatory postsynaptic potentials (fEPSPs) in the input–output and LTP curves of mossy fiber were recorded. The cresyl violet staining of hippocampus sections was performed for the groups. Our results demonstrated that the omega-agatoxin-Aa2a treatment could recover the learning, and memory impairment caused by NMDA-induced excitotoxicity in rat hippocampus.
## Introduction
Neurodegeneration is a symptom of central nervous system diseases like as Alzheimer’s disease (AD), Parkinson’s disease (PD), Huntington’s disease (HD), and Amyotrophic lateral sclerosis (ALS). Various circumstances such as aging, trauma or compression of brain tissue, extreme stress, oxidative stress, neuro-inflammation, and excitotoxicity can induce neurodegeneration (Przedborski et al., 2003; Mehta et al., 2013). During neurodegeneration, the neurons lose their functions over time, which leads to neuronal death through apoptosis and necrosis. This elimination can generate particular defects, based on the involved neurodegeneration region. These defects have a wide range and intensity; depend on level of the disease progression. Neurodegeneration in hippocampus area can cause amnesia, forgetfulness, and cognitive impairment, which ends in AD (Crews and Masliah, 2010). AD has the highest prevalence among the neurodegenerative disorders (Naseri et al., 2022; Sarlaki et al., 2022).
Alzheimer’s disease (AD) extensively increases direct (including both formal medical and non-medical care) and indirect (reduced productivity) health-related costs due to its effects on both quality of life and productivity of patients themselves as well as their caregivers. Also it is projected that AD costs increases due to the growth in its prevalence and treatment costs. The societal financial burden of AD is estimated to rise by about 4.9-fold in the USA (to $1.5 trillion), 2.7-fold in Europe (to €633 billion), and 9.5-fold worldwide (to $9.1 trillion), by 2050 (Dauphinot et al., 2022; Tahami Monfared et al., 2022). Consequently, finding new effective ways to prevent or treat AD can be very worthwhile when increasing health costs is taken into account.
Excitotoxicity is an important precondition for the neurodegeneration, and can be induced by the hyper-activation of N-Methyl-D-Aspartate receptors (NMDARs). This action has a variety of consequences on pyramidal neurons of the hippocampus such as production of high amounts of various free radicals, mitochondrial dysfunction, and activation of apoptotic factors and caspases, which leads to neuro-termination (Rahn et al., 2012). Higher than normal rates of neurotransmitters in synapses are essential for hyper-activation of NMDARs (Arundine and Tymianski, 2003). The major excitatory neurotransmitter in hippocampus is glutamate, which has a crucial role in memory, and learning (Bin Ibrahim et al., 2022). Synaptic firing through the hippocampus trisynaptic circuit is essential for memory formation (Alkadhi, 2021). Also, long-term potentiation (LTP) is a vital process for memory performance (Hayashi, 2022). These processes rely on the release of neurotransmitters. Release of the neurotransmitters is the most important role of voltage-gated calcium channels (VGCCs) (Nimmrich and Gross, 2012). N-type VGCCs are located in the presynaptic terminal and are activated by an action potential. Following activation, calcium ions enter through the N-type VGCCs, which, in turn leads to neurotransmitter release (Sousa et al., 2013). Synaptosomal-associated protein, 25 k Da (SNAP-25) belongs to the soluble N-ethylmaleimide-sensitive factor activating protein receptor protein superfamily, and contributes to exocytosis. SNAP-25 plays a key role in intracellular vesicle trafficking which is crucial for neurotransmitter release (Zhang et al., 2014; Biswal et al., 2017; Li and Kavalali, 2017). Therefore, the SNAP-25 can be used as a synaptic marker. Synaptotagmin 1 (SYT1) is a synaptic calcium sensor, which regulate neurotransmitter release. This sensor has an undeniable role in vesicle exocytosis, and neurotransmitters release in synaptic cleft (Zhang et al., 2014; Hussain et al., 2017). The rate of SYT1 decreases in AD (Li and Kavalali, 2017). Synaptophysin (SYN) is a synaptic marker with a fundamental role in synapse functions including vesicle trafficking, and neurotransmitters release (Zhang et al., 2014; Ji et al., 2017). The syntaxin1A (SY1A) is localized in the presynaptic membrane and with other synaptic proteins triggers the docking of synaptic vesicles (Kim and Oh, 2016). The SY1A has a direct link with SNAP-25 (Ullrich et al., 2015). High rates of glutamate in synapses caused excitotoxicity (Dong et al., 2009). Therefore, the normal activation of N-type VGCCs is necessary for the prevention of excitotoxicity consequences. Use of suitable ligands can reveal the impact of N-type VGCCs in excitotoxicity process. The venom and its components are valuable as some approaches to the mentioned goal (Sousa et al., 2013).
Venom of animals such as snails, snakes, scorpions, and spiders are valuable sources of peptides, and small proteins with unique specifications, such as high affinity to their target receptors or channels, great selectivity, stable structure, and low molecular mass (Lewis and Garcia, 2003; Moradi et al., 2018). Therefore, these ligands can be used for the correction of malfunction of over-activated channels. In this case, the omega-agatoxin-Aa2a as an N-type VGCCs ligand can demonstrate the impact of N-type VGCCs in neurodegenerative disorders, particularly AD. This ligand can be found in Agelenidae family. Agelena labyrinthica species is a member of Agelenidae family (Herzig et al., 2010; Consortium, 2019).
We could not find any direct and comprehensive study regarding the effects of N-type VGCCs on the excitotoxicity of glutamate, as a cognitive impairment in vivo model. The aim of this study was to evaluate the effects of neurotransmitters release on memory, and learning impairments. To achieve this goal, behavioral, molecular, electrophysiological, and histological methods were used in the current study. The *Agelena labyrinthica* spiders were collected from Iran. The epigyne, fangs, and venom glands were observed with a stereoscope. Then the collected venom of these spiders were extracted from venom glands and lyophilized. After that, gel-electrophoresis, gel-filtration chromatography, and capillary electrophoresis were performed for purification and subsequently, mass spectrometry (HPLC-ESI-MS) was used for identification of this ligand. Behavioral tasks including the Morris water maze, and passive avoidance were performed to assess learning, and memory. The SY1A, SYT1, and SYN genes expression were measured via Real-time PCR. In addition, the amount of SNAP-25 localized protein for synaptic quantification was measured through the immunofluorescence technique in the CA3 sub-region of the hippocampus. Then, the field excitatory postsynaptic potentials (fEPSP) amplitude after LTP in the Mossy Fiber pathway were recorded for evaluation of spatial memory, and memory formation. At last, the coronal section of hippocampus and particularly CA3 sub-region were stained through cresyl violet and visualized for comparison of pyramidal neuron in the experimental groups.
## Chemicals, reagents
All chemicals were purchased from Sigma Aldrich Company (Darmstadt, Germany) except for the others mentioned in the text.
## Spider collection, identification, and venom extraction
For this study, specimens were collected alive from Iran (Zamani et al., 2021). The *Agelena labyrinthica* was photographed with a camera. The specimens were kept under suitable conditions, humidity ($60\%$), and temperature (25°C) and fed by crickets and mealworms. The specimens were identified using the taxonomic keys (Nentwig et al., 2017). Epigyne of the female specimen was visualized with a stereoscope.
Female spiders were separated to extract the venom. Specimens were anesthetized with CO2 in a small chamber, and the opisthosoma and carapace were removed under the stereoscope. The fangs and venom glands of the female specimen were visualized by a loop. Venom glands were collected into 4°C Phosphate buffered saline (PBS) were prepared in the laboratory with this recipe (137 mM NaCl, 3 mM KCl, 10 mM Na2PO4, 2 mM KH2PO4, and pH 7.4) and gently crushed with a glass stirrer for 30 min. Then, pieces of the venom gland were removed from the solution by centrifugation at 13,000 rpm for 30 min at 4°C, and the supernatant was lyophilized and stored at −70°C. Protein concentration was measured by Bradford assay with bovine serum albumin as standard protein.
A total of 150 spiders were collected from Iran (Zamani et al., 2021), and separated by species and gender. One hundred female *Agelena labyrinthica* specimens were selected for the next phase of the study. The specimens on the net web and epigyne of this species are shown in Figure 1.
**FIGURE 1:** *Agelena labyrinthica species. (A) The Agelena labyrinthica spider on the net. (B) The epigyne of female specimens.*
1.03 g of the lyophilized crude venom was extracted from *Agelena labyrinthica* specimens. Then, 10 mg of the crude venom was dissolved in PBS buffer. The protein concentration of the *Agelena labyrinthica* lyophilized crude venom was 7.2 mg/ml, which was determined by the Bradford method. The fangs, venom glands, and lyophilized crude venom of *Agelena labyrinthica* are presented in Figure 2.
**FIGURE 2:** *Venom extraction and venom glands of Agelena labyrinthica. (A) The Agelena labyrinthica fangs. (B) The venom gland. (C) The lyophilized extracted venom.*
## Determination of lethal dose (LD50)
To determine the LD50, the albino mice (average weight 18–20 g) were intravenously (IV) injected with crude venom. After the injection, the animals were followed up for 1 day. The Spearman-Karber method was used to calculate the LD50 dose for the crude venom and omega-agatoxin-Aa2a protein (Hamilton et al., 1977).
## SDS-PAGE of the crude venom
1 mg of the crude venom was dissolved in 1 ml PBS buffer and incubated at 4°C overnight. The supernatant was then centrifuged at 13,000 rpm for 4 min at 4°C. Then, 15 μl of dissolved crude venom was mixed with 5 μl loading buffer (1x) and heated for 10 min at 100°C. An amount of 20 μl of this solution was loaded into each well of the gel [$15\%$ Sodium Dodecyl Sulfate-Polyacrylamide Gel-Electrophoresis (SDS-PAGE)] which stained with coomassie blue dye for 45 min at 25°C. Hot water and a shaker were used to decolorize the gel. Finally, the gels were scanned by Bio-5000 Gel scanner device (Seyfi et al., 2019).
## Protein purification with gel-filtration chromatography
The lyophilized crude venom (10 mg) was resuspended in 1.5 mL of PBS buffer. The DNase (0.14 mg/ml) and RNase (0.14 mg/ml) enzymes were added to the sample and incubated for 2 h at 4°C. The clear solution was injected into a gel-filtration column (GE Healthcare HiLoad $\frac{16}{600}$ Superdex® 75 pg prep grade) and run over it using FPLC (Fast Protein Liquid Chromatography) system (Sykam, Germany). The column was washed with PBS Buffer. The injection volume was 1,200 μL and the flow rate was 0.7 mL/min. The fractions were observed with absorbance at 280 nm and collected in a 0.75 ml fraction. The chosen fractions (marked on the graph) were collected and injected into capillary electrophoresis and gel-electrophoresis (Mofid et al., 2021).
As shown in Figure 4A, the gel-filtration chromatography of the *Agelena labyrinthica* venom had five peaks. The fourth fraction of Agelena Labyrinthica venom was collected from 160 to 190 min. 1.57 mg of the lyophilized fraction was obtained from 21 ml of the solution, taken from the instrument. The protein concentration of this fraction was 1.03 mg/ml.
**FIGURE 4:** *Gel-filtration chromatography of the Agelena labyrinthica venom, and the gel electrophoresis of the third, fourth, and fifth fractions. (A) Gel-filtration chromatography of the crude venom performed on a GE Healthcare HiLoad 16/600 Superdex® 75 pg prep grade column in 1 M PBS (pH 7.4), flow rate 0.7 ml/min. The fourth fraction is shown by black arrow. (B) The gel electrophoresis of Agelena labyrinthica: 3, 4, and 5 fractions. L: Ladder (k Da).*
The gel electrophoresis of the 4, 5, and 6 fractions are displayed in Figure 4B. This figure is completely consistent with the gel-filtration chromatography pattern, shown in Figure 4A.
## Protein purification and separation with capillary electrophoresis:
Capillary electrophoresis test was performed by Agilent 7100 equipped with a UV-Vis detector using a 50 μm uncoated silica column with a total length of 40 cm and a detector distance of 8.5 cm from the outlet. PBS buffer with pH 4.7 was used for both the sample, and running buffer. The capillary temperature was 25°C and the sample was injected at 100 mBar for 5 s. Electrophoresis was performed for 5 min at 25 kV normal polarity. The marked peaks were collected, and protein concentration was determined by Bradford assay. The desired peak was re-injected into the device to ensure purity, and the result was presented the collected fraction was subjected to $12\%$ SDS-PAGE. The selected peak obtained from this method project into HPLC-ESI-MS (Keimasi et al., 2022).
## Protein identification with mass spectrometry (HPLC-ESI-MS)
The high-performance liquid chromatography/electrospray ionization tandem mass spectrometry (HPLC-ESI-MS) analysis was performed by Waters Alliance 2695 HPLC-Micromass Quattro micro API Mass Spectrometer. Liquid chromatography separation was performed on Atlantis T3-C18 column (3 μ, 2.1 × 100 mm) at 35°C. Mobile phases were $0.1\%$ formic acid in acetonitrile (A) and $0.1\%$ formic acid in H2O. The gradient profile was $5\%$ A held for 0.2 min and linearly increased to $90\%$ in 10 min. Then, it was held for 5 min, which decreased to $5\%$ over 3 min and finally held for 4 min. The flow rate was 0.2 mL/min, and the injection volume was 5 μL. The mass spectrometry method was included in positive mode with the capillary voltage, which was adjusted to 0.3 kV, and the source and dissolving temperatures were set at 120°C and 300°C respectively, with flow gas 200 L/h. The result was presented for the purified bio-active small protein peak (Jafari et al., 2014).
The molecular mass of omega-agatoxin-Aa2a is 10,982 Da, according to the Sousa et al. [ 2013] study. The spectrum results demonstrated the existence of omega-agatoxin-Aa2a in the selected peak (Figure 6A). The mass-to-charge ratio of this ligand was consistent with omega-agatoxin-Aa2a. The charge-to-mass ratio of omega-agatoxin-Aa2a is as follows: (M + Na + 5H)6+ = 1835, (M + Na + 6H)7+ = 1573, (M + Na + 7H)8+ = 1,376.5. The Mass/*Mass spectrum* of Quadro charge ion at 1,376.5 was selected for N-terminal partial sequencing as follows: b3 = 274, b4 = 403, b5 = 516, b6 = 573, b7 = 630, b8 = 745, b9 = 848, b10 = 963, b11 = 1,020, b12 = 1,183, b13 = 1,311, b14 = 1,440, and b15 = 1,568. The results of the spectrum indicated a good consistency with the omega-agatoxin-Aa2a (Accession number: P15971) sequence (Figure 6B).
**FIGURE 6:** *The identification of omega-agatoxin-Aa2a by mass spectrophotometry. (A) HPLC-ESI-MS of the selected peak from capillary electrophoresis was performed on Atlantis T3-C18 column. (B) N-terminal partial sequencing of the omega-agatoxin-Aa2a protein. The singly charged ion of the N-terminal peptide (m/z, 1406) was subjected to the fragmentation in the ion trap mass analyzer. The observed fragment ions are indicated above and below the peptide sequence.*
## Animals and experimental design
54 adult male Wistar rats, weighing 230–250 g were taken from the animal’s nest of the Faculty of Biological Sciences and Technology, University of Isfahan. They were kept in standard cages with controlled temperature (∼25°C) and humidity (∼$40\%$), 12 h light; 12 h dark cycle, and free access to enough food and water. The ethics committee of the University of Isfahan approved the study.
The work was performed on male Wistar rats, which were spat into three groups (18 rats in each group). A small area on each rat’s skull was shaved while the head was fixed using a stereotaxic instrument (Stoelting Co., USA) to prepare for injection into the hippocampus. The PBS buffer was used as a vehicle for omega-agatoxin-Aa2a and N-Methyl d-Aspartate (NMDA). Rats were assigned into the following groups: Control Group: Received 1 μl of PBS in the CA3 sub-region of the hippocampus twice and with an interval of 30 min.
NMDA-treated group: received a single dose of NMDA (1 μl, 10 μg/μl) in the CA3 sub-region of the hippocampus followed by 1 μl of PBS 30 min later (Jarrard, 2002).
Omega-agatoxin-Aa2a -treated group: received a single dose of NMDA (1 μl, 10 μg/μl) in the CA3 sub-region of the hippocampus followed by 1 μl of omega-agatoxin-Aa2a (2 μg/μl) 30 min later (Sousa et al., 2013).
## Surgery and microinjection procedure
One week before the start of the behavioral tests, all the animals were deeply anesthetized with an intraperitoneal (i.p.) injection of phenobarbital (40 mg/kg). The phenobarbital was purchased from Martindale Pharma Company (Buckinghamshire, England). The animals were wrapped in towels and the eyes were covered with Vaseline during the operation. Then, the CA3 sub-region of the hippocampus (AP:−3.3 mm from bregma; ML: ± 3 mm from midline; DV: 3.5 mm from the skull surface) was found using the Paxinos and Watson rat brain atlas (Paxinos and Watson, 2006). The injections were performed bilaterally on the hippocampi by stereotaxic apparatus.
The agents were bilaterally administered into the hippocampi using an injection needle (21 gauge) connected to a 5 μl Hamilton syringe through a polyethylene tube. The agents were slowly injected into the hippocampus area for 6 min. To avoid backflow of the fluid, the needle was slowly removed 2 min after the injection.
## Behavioral studies
The behavioral evaluations including the Morris water maze and Passive avoidance tasks were performed a week after the stereotaxic surgery. These tasks have assessed spatial, long-term memory, and learning. Before the behavioral tests, animals were acclimatized over 2 days in the laboratory area to be adjusted to the experimental conditions and to minimize stress.
## Morris water maze task
The Morris water maze test is one of the most common procedures to assess spatial memory and learning performance in rodents. The walls of the water maze room must have signs and symbols that rats can use their spatial memory to find the hidden platform in the target zone. Morris water maze task has three parts including habituation, training, and testing (it has been completely described in previous work) (Hosseini-Sharifabad et al., 2021). In this task, time spent in the target quadrant, distance moved in the target quadrant, the entry into the target quadrant, the velocity of rats, and swimming paths of rats in the last time of training were recorded to evaluate spatial memory and learning performance (ten rats in each group). All behavioral parameters of the rats were monitored by a video camera, fixed to the ceiling above the center of the pool and connected to a computerized tracking system (Auto vision Software, Designed by BorjSanat Company, Tehran, Iran).
The intra-hippocampal administration of NMDA, significantly decreased the time spent in the target area ($P \leq 0.0001$), and distance moved in the target zone ($P \leq 0.0001$), when compared to the control group. NMDA treatment with administration of omega-agatoxin-Aa2a increased the time spent in the target zone ($P \leq 0.01$), and distance moved in the target zone ($P \leq 0.01$) of the maze, compared to the NMDA-treated group ($P \leq 0.01$). Also, significant differences in the time spent ($P \leq 0.01$), and distance moved ($P \leq 0.01$) in the target zone of the maze were observed between the control, and NMDA + Aga groups (Figure 7A).
**FIGURE 7:** *Morris water maze task. (A) The effect of treatment with omega-agatoxin-Aa2a on the time spent in the target zone. The data are shown as means ± SEM of 6 rats per group. (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). (B) The swim path of the rats from the last time of training.*
The swimming path of experimental animals during the last time of training is displayed in Figure 7B. The swimming path of the control group was short and the rats found the hidden platform, easily. But, the swimming path of the NMDA group was long and the rats failed to locate the hidden platform. The majority of the movements of these rats were limited to the area close to the wall of the pool. However, the swimming path of the NMDA-treated with a single dose of omega-agatoxin-Aa2a group was almost short and the evaluated rats were able to find the hidden platform after a quick exploration in the water pool.
## Passive avoidance task
The passive avoidance task was applied to assess non-conceptual communication memory and learning, as this method has already been used in neurological disorders models in small laboratory animals (Salehifard et al., 2023). The term passive avoidance is commonly used to describe experiments in which animals learn to avoid a painful stimulus. A passive avoidance behavior study was performed on two consecutive days (training and test), as explained by Ebrahimpour et al. [ 2018]. The latency time before entering the dark section, and the time spent in the dark section on the test day, were recorded to evaluate passive avoidance learning and memory (10 rats in each group).
According to Figure 8A, our data revealed that administration of a single dose of NMDA into the hippocampus, significantly decreases the latency time in the NMDA group, compared to the control group ($P \leq 0.0001$), in the passive avoidance task. NMDA treatment with a single dose of omega-agatoxin-Aa2a increased the latency time in the NMDA + Aga group compared to the NMDA-treated group ($P \leq 0.05$). The Control group was significantly different from the NMDA + Aga group ($P \leq 0.0001$).
**FIGURE 8:** *The passive avoidance task. (A) The effect of treatment with omega-agatoxin-Aa2a on the latency time before entering the dark section in the test day. The data are shown as means ± SEM of 6 rats per group. (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). (B) The effect of treatment with omega-agatoxin-Aa2a on the time spent in the dark compartment in the test day. The data are shown as means ± SEM of 6 rats per group. (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001).*
As shown in Figure 8B, our results demonstrated that the intra-hippocampal injection of NMDA significantly increases the time spent in the dark chamber in the NMDA-treated group when compared to the control group ($P \leq 0.0001$). Accordingly, NMDA treatment with omega-agatoxin-Aa2a decreases the time spent in the dark chamber in the NMDA + Aga group compared to the NMDA-treated group ($P \leq 0.0001$), in the passive avoidance task. There is a significant difference between the NMDA + Aga and the control groups ($P \leq 0.0001$).
## RNA extraction, complementary DNA (cDNA) synthesis, and quantitative real-time PCR
The mRNA extraction was performed as formerly has been described by Gheysarzadeh et al. [ 2019] and mRNA expression measurement was done for the three groups (six rats in each group) (Gheysarzadeh et al., 2019). Total RNA was extracted from the hippocampus samples using RNX-PLUS reagent (SinaClon, Iran). cDNA was synthesized according to the manufacturer’s protocol with a cDNA synthesis kit (Takara, Japan). Real-time PCR assay was performed by StepOnePlus™ Real-Time PCR System. The sequences of Real-time PCR primers were as follows: 5′-GAGGAAGGTCTGAACCGCTCAT-3′, and 5′- CGTTCTCGGTAGTCTGACTGAG-3′ for mice syntaxin1A, 5′- AGGGCCTATGATGGACTTTCTG-3′, and 5′-TCCGTGGCCA TCTTCACATC-3′ for mice Synaptophysin, 5′-CGGCAAA CTGACTGTCATTC-3′ and 5′-GCC CCA GTG CTG TTG TAA CCA-3′ for mice Synaptotagmin 1, and 5′-CAGGG CTGCCTTCTCTTGTG-3′, and 5′-GATGGTGATGGGTTTC CCGT-3′ for mice Glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Relative genomic expression was calculated by the 2–Δ Δ Ct method. The mRNA levels were normalized to that of GAPDH (Gheysarzadeh et al., 2019).
## Electrophysiological study
The rats (six in each group) were anesthetized with urethane (1.5 g/kg; intraperitoneal injection) dissolved in $0.9\%$ normal saline solution. The surgery and LTP recording procedures from the hippocampal CA3 area were executed as formerly has been described by Keimasi et al. [ 2022].
## Immunohistochemistry
The brains were washed with saline and sequentially fixed in $10\%$ formaldehyde solution and then post-fixed in $15\%$ and $25\%$ sucrose solutions. Two-micrometer sections were cut by using a microtome instrument. Antigen retrieval was performed enzymatically for 20 min. To reduce non-specific antibody binding, the samples were blocked with the blocking agent [$10\%$ normal goat serum (Sigma, G9023) and $0.3\%$ Triton X-100 in PBS] for 30 min at 37°C. After washing, tissue sections were incubated with mouse anti−SNAP25 monoclonal (Abcam, ab66066) as the primary antibody at 4°C. FITC−conjugated anti−mouse IgG (Sigma, F9137) were used as secondary antibody for 2 h at room temperature. The nuclei were counterstained with DAPI (Sigma, D9542). The slides were visualized using an AX70 Olympus fluorescence microscope (three in each group).
## Cresyl violet staining
Tissue coronal slides were deparaffinized two times in xylene, each for 10 min. Rehydration was performed using gradient alcohol (100, 95, 70, and $50\%$ alcohol) each time for 5 min. The brain slides were stained in cresyl violet solution ($0.25\%$ cresyl violet, $0.8\%$ glacial acetic acid, 0.6 mM sodium acetate) for 20 min. The dehydration was done using gradient alcohol (70, 95, and $100\%$ alcohol) each time for 3 min. Tissue slides were visualized using a BX40 Olympus light microscopy (three in each group).
Our data showed that the single injection of NMDA significantly decreased the expression of living pyramidal neurons in the CA3 sub-region of the hippocampus compared to the control group ($P \leq 0.001$). The single administration of omega-agatoxin-Aa2a after injection of NMDA, distinctly enhanced the living pyramidal neurons in the CA3 sub-region of the hippocampus with respect to the NMDA-treated group ($P \leq 0.01$). The NMDA-treated + ligand group showed a significant difference in the CA3 sub-region of the hippocampus compared to the control group ($P \leq 0.01$; Figures 12A, B).
**FIGURE 12:** *The cresyl violet staining of rat hippocampus. (A) The graph of healthy (Narrow red arrow) and dead (Narrow black arrow) pyramidal neurons in the CA3 sub-region of the rat hippocampus. (B) The number of healthy cells are shown in the CA3 sub-region of the rat hippocampus. The data are shown as means ± SEM of 3 rats per group. (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001).*
## Data analysis
Statistical analysis was performed using Graph Pad Prism statistics software (version 8.4.3). Statistical Data were evaluated by the D’Agostino-*Pearson omnibus* test to examine the normal distribution. The data collected were analyzed by one-way, and one-way repeated measures analysis of variance (ANOVA). For multiple comparisons, Tukey’s test was used. All data were shown as the mean ± standard errors of the means ($P \leq 0.05$ was considered a significant difference).
## Determination of LD50
Based on the results of IV injections of the crude venom and omega-agatoxin-Aa2a protein into the albino mice, the LD50 was determined as 32.68 mg/kg and not toxic.
## SDS-PAGE
The gel electrophoresis of the crude venom of *Agelena labyrinthica* showed a lot of protein bands separated based on their molecular weights. The strongest bands were the 158 and the band range of 3–10 k Da. The 3–10 k Da band was related to the agatoxin family, which has a molecular weight in this range (Herzig et al., 2010; Consortium, 2019; Figure 3).
**FIGURE 3:** *The gel electrophoresis of Agelena labyrinthica crude venom. L: Ladder (k Da), V: Venom.*
## Protein purification with capillary electrophoresis (CE)
Figure 5, part A displays the capillary electrophoresis pattern, obtained from the fourth fraction of *Agelena labyrinthica* venom. Four peaks were observed at 1.71, 1.77, 3.95, and 4.78 min. The fourth peak was collected and re-injected to confirm the purity of this small bioactive protein and then injected into HPLC-ESI-MS for mass determination.
**FIGURE 5:** *Capillary electrophoresis pattern of the fourth fraction, and the gel electrophoresis of the selected fraction. (A) Capillary electrophoresis of the fourth fraction was performed on a 50 μm uncoated silica column 1 M PBS (pH 4.7) for 5 min. The omega-agatoxin IIA fraction is shown by black arrow. (B) The gel electrophoresis of the selected peak. L: Ladder (k Da).*
The gel electrophoresis of the selected peak revealed a single band of less than 13 kDa (Figure 5B).
## Quantitative real-time PCR
Figure 9 displays the effect of NMDA-treated and NMDA-treated + Aga groups on, SY1A, SYT1, and SYN genes expression in the rat hippocampus. This figure reveals that administration of NMDA in the rat hippocampus distinctly has decreased SY1A ($P \leq 0.0001$), SYT1 ($P \leq 0.0001$), and SYN ($P \leq 0.0001$) mRNAs expression when compared to the control group. The NMDA-treated group with a single dosage injection of omega-agatoxin-Aa2a increased SY1A ($P \leq 0.05$), SYT1 ($P \leq 0.01$), and SYN ($P \leq 0.05$) mRNAs expression with respect to the NMDA-treated group. Significant differences were observed between the two examined groups (mRNAs expression: NMDA-treated, and NMDA-treated with a single dose of omega-agatoxin-Aa2a) for SY1A ($P \leq 0.001$), SYT1 ($P \leq 0.001$), and SYN ($P \leq 0.0001$).
**FIGURE 9:** *The quantitative real-time PCR. The effect of treatment with omega-agatoxin-Aa2a on the hippocampal SY1A, SYT1, and SYN mRNA levels. The data are shown as means ± SEM of six rats per group. (*P < 0.05 and **P < 0.01 and ***P < 0.001 and ****P < 0.0001).*
## Mossy fiber circuit LTP
According to the Figure 10A, the injection of NMDA significantly decreased the field excitatory postsynaptic potentials (fEPSP) amplitude in the NMDA-treated group after LTP induction in the CA3 area of the hippocampus when compared to the control group ($P \leq 0.001$). Administration of a single dose of omega-agatoxin-Aa2a after the injection of NMDA in the NMDA-treated + Aga group remarkably increased the fEPSP amplitude after LTP induction with respect to the NMDA-treated group ($P \leq 0.01$). The fEPSP amplitude in the NMDA-treated + Aga group represented a significant difference compared to the control group ($P \leq 0.01$).
**FIGURE 10:** *Mossy Fiber circuit LTP. (A) Long-term potentiation (LTP) curves of the field excitatory postsynaptic potential (fEPSP) amplitude in the hippocampal CA3 for the all groups (n = 6). The data are shown as means ± SEM of 6 rats per group. (*P < 0.05,**P < 0.01, ***P < 0.001, and ****P < 0.0001), and (###P < 0.001). (B) Sample traces of typical recorded fEPSPs in the hippocampal CA3 neurons before and after high-frequency stimulation (HFS) induction for the long-term potentiation (LTP) in experimental groups.*
Figure 10B, demonstrates the traces of the recorded fEPSPs in the hippocampal CA3 neurons, before and after LTP induction using the high-frequency stimulation (HFS) technique in all the compared groups.
## Localize SNAP-25 in CA3 sub-region of the hippocampus
Our data showed that the single injection of NMDA significantly decreased the expression of SNAP-25 protein in the CA3 area of the hippocampus compared to the control group ($P \leq 0.0001$). The single administration of omega-agatoxin-Aa2a after injection of NMDA distinctly enhanced the expression of SNAP-25 in the CA3 area of rat hippocampus with respect to the NMDA group ($P \leq 0.01$). The NMDA + Aga group showed a significant differences in the CA3 sub-region of the rat hippocampus compared to the control group ($P \leq 0.01$; Figures 11A–C.
**FIGURE 11:** *Immunofluorescence staining with antibody against SNAP-25. (A) The localize SNAP-25 are presented in the CA3 area of rat hippocampus. DAPI was applied to counterstain the nuclei. The healthy (white arrow) and dead (red arrow) pyramidal neurons are indicate in the left column. (B) The localize SNAP-25 are Shown in the CA3 area of rat hippocampus for the experimental groups (In percent). The data are shown as means ± SEM of 3 rats per group. (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). (C) Scatter plot of part label (B).*
## Discussion
In the current study, the omega-agatoxin-Aa2a was purified and identified from *Agelena labyrinthica* crude venom. This species is a sister group of *Agelenopsis aperta* (Wheeler et al., 2017). The specimens were collected from Iran. After the identification of specimens using the taxonomic key, particularly the epigyne, the venom glands were separated and crude venom was extracted from them. Amount and protein concentration of crude venom were in agreement with a previously published by Beneli and Yigit [2008]. The crude lyophilized crude venom was a white powder with a normal appearance. A suitable buffer under low-pressure condition was used to purify the omega-agatoxin-Aa2a from the crude venom. Therefore, the native structure of the ligand was protected from degradation. At each step of the purification process, gel-electrophoresis was performed, to confirm the purification process. Based on the collected data from the LD50 section, the toxicity of this species crude venom was low and the omega-agatoxin-Aa2a protein LD50 was far less than the crude venom. This decrease in LD50 could be caused due to the separation of the crude venom. Crude venom act as a single compartment. Although venoms have multiple parts, they can operate perfectly when these parts act together. Therefore, a decrease in the potency of each part of a venom following separation of crude venom seems logical. On the other hand, the enzymatic part of a crude venom has a high molecular weight whereas; the omega-agatoxin-Aa2a protein has a low molecular weight (Keimasi et al., 2022). Accordingly, the omega-agatoxin-Aa2a protein is safe, particularly in the microliter concentration, which has been used in the current study.
The omega-agatoxin-Aa2a is a state-dependent small protein that has interacts with N-type VGCCs, according to the Sousa et al. [ 2013] study. This ligand can potentially block the N-type VGCCs. The omega-agatoxin-Aa2a for the first time has been purified from *Agelenopsis aperta* species for the first time (Herzig et al., 2010; Consortium, 2019). Protein measurement of omega-agatoxin-Aa2a was performed for dosage determination.
Ion channels have a crucial role in signal transduction and convert electrical neurotransmission to chemical neurotransmission (Simms and Zamponi, 2014). The last location of the neurotransmission in the neurons is presynaptic axon terminal. Calcium channels in this area are voltage-gated-type (Xu and Tang, 2018). These channels include N, and P/Q VGCCs (Zamponi and Snutch, 1998). A nerve impulse opens the alpha1-segment of VGCCs, and calcium ions enter into the presynaptic terminal and trigger the release of neurotransmitters (Nimmrich and Gross, 2012). As said before, glutamate is the main hippocampus neurotransmitter. Therefore, the release of this excitatory neurotransmitter can be modulated by an appropriate ligand. The omega-agatoxin-Aa2a as a ligand can bind to the alpha1-segment of N-type VGCCs (Sousa et al., 2013). This is important, as in neurodegeneration diseases like AD the structure of the ion channels can be transformed by amyloid beta which leads to the malfunction of channels and finally ends to an overload of calcium in presynaptic neurons and releasing high amounts of glutamate in the synaptic cleft (Koutsilieri and Riederer, 2007; Szydlowska and Tymianski, 2010; Zhang et al., 2015). In such condition, excitotoxicity can be induced and its consequences lead to neuronal elimination through apoptosis and necrosis pathways (Belousov, 2012; Verma et al., 2022). It is obvious that the omega-agatoxin-Aa2a is efficient as a ligand in excitotoxicity condition, because of its binding site.
As soon as an action potential reaches to the presynaptic area, the calcium ions enter the neurons (Mochida, 2019). SNAP-25, SY1A, SYT1, and SYN are synaptic markers (Li and Kavalali, 2017). SNAP-25 has a fundamental role in synaptic function and transmission of a signal from the presynaptic neuron to postsynaptic neuron (Zhang et al., 2014; Li and Kavalali, 2017). The SY1A has a straight connection via SNAP-25 (Ullrich et al., 2015). Therefore, the SY1A has a fundamental role in the docking vesicles into the presynaptic membrane and trigger of neurotransmitter release (Ullrich et al., 2015; Kim and Oh, 2016; Li and Kavalali, 2017). The SYT1 is a calcium-sensitive sensor, which has a crucial role in the fast vesicle exocytosis of neurotransmitters from neurons (Zhang et al., 2014; Hussain et al., 2017). The SYN is a major synaptic vesicle protein, which is localized in the presynaptic neurons. SYN has an important role in vesicular docking and neurotransmitter release (Zhang et al., 2014; Ji et al., 2017). Our data indicated that administration of NMDA in rat hippocampus could induces excitotoxicity due to over-stimulation of NMDARs, which in turn reduces SNAP-25 expression protein level. Also, this administration reduced SY1A, SYT1, and SYN mRNAs expression (Yu et al., 2006; Li et al., 2012; You et al., 2018; Keimasi et al., 2022). This reduction of SNAP-25, SY1A, SYT1, and SYN leads to cognitive impairment and memory dysfunction, which can be reversed by an increase in SNAP-25, SY1A, SYT1, and SYN following the blockage of presynaptic N-type VGCCs with omega-agatoxin-Aa2a. The function of SNAP-25, SY1A, SYT1, and SYN relies on presynaptic VGCCs. Normal rates of SNAP-25, SY1A, SYT1, and SYN ends in neurotransmitter release in the synaptic cleft (Valtorta et al., 2004; Tampellini et al., 2010; Zhang et al., 2014; Biswal et al., 2017). The release of glutamate in hippocampus neurons can induce LTP and neuron plasticity (Bin Ibrahim et al., 2022). Therefore, the presence of SNAP-25, SY1A, SYT1, and SYN is essential for memory and learning. The reduction or loss of SNAP-25, SY1A, SYT1, and SYN expressions causes memory impairment and cognitive deficit in AD. Thus, the function of SNAP-25, SY1A, SYT1, and SYN is critical for synaptic performance (Li and Kavalali, 2017). However, the blockage of N-type VGCCs by omega-agatoxin-Aa2a as a ligand was not able to restore the SNAP-25, SY1A, SYT1, and SYN expressions to the normal levels.
Based on the collected data from cresyl violet staining, the NMDA injection into the CA3 sub-region of the hippocampus eliminated the pyramidal neurons in this area through neurodegeneration, induced by the excitotoxicity process. The administration of omega-agatoxin-Aa2a after NMDA injection had an ameliorative effect on neurodegeneration via blocking the N-type VGCCs and controlling the excitotoxicity process through the regulation of neurotransmitters release (Sousa et al., 2013; Mochida, 2019). Glutamate has an undeniable role in memory and learning performance. However, the hyper-activity of this neurotransmitter can induce neurodegeneration through the hyper-stimulation of the NMDA receptors (Dong et al., 2009; Crews and Masliah, 2010). Therefore, agonists of NMDAR like as NMDA, alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and kainic acid can induce excitotoxicity, according to multiple studies (Jarrard, 2002; Zhang et al., 2014, 2015; Hosseini-Sharifabad et al., 2021; Naseri et al., 2022). In addition, NMDA has been used for the induction of cognitive impairment, learning, and memory defects. Our data demonstrated that intra-hippocampal injection of NMDA eliminated the memory and learning performance, which is in line with the previous study (Jarrard, 2002).
The collected results from the Morris water maze and Passive Avoidance tests showed defects in cognitive memory, and learning and memory performance. The effect of NMDA treatment appeared in the rat’s behavior. The NMDA-treated rats had a critical problem to locate the hidden platform and failed to complete this task. Plus, these rats were not able to discriminate the target zone from other zones. Therefore, this group had a serious problem with cognitive memory and learning function (Ebrahimpour et al., 2018). Despite this, the NMDA-treated group who had received a single dose of omega-agatoxin-Aa2a was able to finish the task. Therefore, the effect of this ligand on the N-type VGCCs was related to cognitive memory and learning performance. Memorizing function relies on LTP process, which has various types in the hippocampus. The main types of LTP include postsynaptic, presynaptic, and both pre and postsynaptic. LTP process induces the memory formation. The pyramidal neurons in the CA3 sub-region of the hippocampus have mostly presynaptic LTP. These particular neurons are highly connected together, and are able to excite each other. Another feature of these pyramidal neurons is their large presynaptic terminals and frequency of their neurotransmitter release sites. Therefore, it seem logical for LTP in Mossy Fiber be presynaptic and dependent to calcium influx into the neurons. In other words, the LTP in Mossy *Fiber is* non-associative (Alkadhi, 2021). The mentioned features in the CA3 sub-region of the hippocampus make this area a perfect place for excitotoxicity induction (because of the self-excision ability) and evaluation of the effect of omega-agatoxin-Aa2a as an N-type VGCCs ligand.
As a conclusion, the induced excitotoxicity by NMDA resulted in pyramidal neuron death in the CA3 sub-region of the hippocampus, reduction of fEPSP after LTP induction, decrease in the rate of SNAP-25 protein, and downregulation of SY1A, SYT1, and SYN mRNAs, as well as cognitive and learning memory performance elimination. However, a single injection of omega-agatoxin-Aa2a in the NMDA-treated rats led to the prevention of consequences of excitotoxicity, through the blockage of N-type VGCCs. In addition, a single injection of omega-agatoxin-Aa2a could reverse the cognitive memory and learning impairment, upregulate SNAP-25 protein, SY1A, SYT1, and SYN mRNA levels in the CA3 sub-region of the hippocampus, enhance the fEPSP after LTP induction, and prevent pyramidal neuron death. This study can be considered as a starting point for future study on evaluation of the effect of omega-agatoxin-Aa2a on the cognitive abilities such as; pain sensitivity, sensorimotor, and locomotor ability. In addition, evaluation of the effects of bioactive small proteins on the N-type and P/Q-types VGCCs as a co-treatment can be suggested for the next research.
## 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. Requests to access these datasets should be directed to [email protected], [email protected], and [email protected].
## Ethics statement
The animal study was reviewed and approved by the animal Ethics Committee of the University of Isfahan.
## Author contributions
MK conceived the original idea. MK, MM, and MRM planned the experiments. MK, KS, MJK, MA, NE, and FE performed the experiments, data collection, analysis, and interpretation. MK and MA wrote the manuscript. MM, MRM, and MA have collaborated in presenting the research idea. MK, MM, and MRM supervised the project. All authors revised the manuscript and approved the final version of 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. Alkadhi K. A.. **NMDA receptor-independent LTP in mammalian nervous system.**. (2021) **200** 101986. DOI: 10.1016/j.pneurobio.2020.101986
2. Arundine M., Tymianski M.. **Molecular mechanisms of calcium-dependent neurodegeneration in excitotoxicity.**. (2003) **34** 325-337. DOI: 10.1016/S0143-4160(03)00141-6
3. Belousov A. B.. **Novel model for the mechanisms of glutamate-dependent excitotoxicity: role of neuronal gap junctions.**. (2012) **1487** 123-130. DOI: 10.1016/j.brainres.2012.05.063
4. Benli M., Yigit N.. **Antibacterial activity of venom from funnel web spider**. (2008) **14** 641-650. DOI: 10.1590/S1678-91992008000400007
5. Bin Ibrahim M. Z., Benoy A., Sajikumar S.. **Long-term plasticity in the hippocampus: maintaining within and ‘tagging’between synapses.**. (2022) **289** 2176-2201. DOI: 10.1111/febs.16065
6. Biswal S., Das D., Barhwal K., Kumar A., Nag T. C., Thakur M. K.. **Epigenetic regulation of SNAP25 prevents progressive glutamate excitotoxicty in hypoxic CA3 neurons.**. (2017) **54** 6133-6147. DOI: 10.1007/s12035-016-0156-0
7. Consortium U.. **UniProt: a worldwide hub of protein knowledge.**. (2019) **47** D506-D515. DOI: 10.1093/nar/gky1049
8. Crews L., Masliah E.. **Molecular mechanisms of neurodegeneration in Alzheimer’s disease.**. (2010) **19** R12-R20. DOI: 10.1093/hmg/ddq160
9. Dauphinot V., Potashman M., Levitchi-Benea M., Su R., Rubino I., Krolak-Salmon P.. **Economic and caregiver impact of Alzheimer’s disease across the disease spectrum: a cohort study.**. (2022) **14** 1-13. DOI: 10.1186/s13195-022-00969-x
10. Dong X.-x, Wang Y., Qin Z.-h. **Molecular mechanisms of excitotoxicity and their relevance to pathogenesis of neurodegenerative diseases.**. (2009) **30** 379-387. DOI: 10.1038/aps.2009.24
11. Ebrahimpour S., Esmaeili A., Beheshti S.. **Effect of quercetin-conjugated superparamagnetic iron oxide nanoparticles on diabetes-induced learning and memory impairment in rats.**. (2018) **13** 6311. DOI: 10.2147/IJN.S177871
12. Gheysarzadeh A., Ansari A., Emami M. H., Razavi A. E., Mofid M. R.. **Over-expression of low-density lipoprotein receptor-related Protein-1 is associated with poor prognosis and invasion in pancreatic ductal adenocarcinoma.**. (2019) **19** 429-435. DOI: 10.1016/j.pan.2019.02.012
13. Hamilton M. A., Russo R. C., Thurston R. V.. **Trimmed Spearman-Karber method for estimating median lethal concentrations in toxicity bioassays.**. (1977) **11** 714-719. DOI: 10.1021/es60130a004
14. Hayashi Y.. **Molecular mechanism of hippocampal long-term potentiation–Towards multiscale understanding of learning and memory.**. (2022) **175** 3-15. DOI: 10.1016/j.neures.2021.08.001
15. Herzig V., Wood D. L., Newell F., Chaumeil P.-A., Kaas Q., Binford G. J.. **ArachnoServer 2.0, an updated online resource for spider toxin sequences and structures.**. (2010) **39** D653-D657. DOI: 10.1093/nar/gkq1058
16. Hosseini-Sharifabad A., Mofid M. R., Moradmand M., Keimasi M.. **The effect of omega-lycotoxin on the cognitive impairment induced by kainic acid in rats.**. (2021) **15** 49-56. DOI: 10.32598/IJT.15.1.740.1
17. Hussain S., Egbenya D. L., Lai Y. C., Dosa Z. J., Sørensen J. B., Anderson A. E.. **The calcium sensor synaptotagmin 1 is expressed and regulated in hippocampal postsynaptic spines.**. (2017) **27** 1168-1177. DOI: 10.1002/hipo.22761
18. Jafari S., Babaeipour V., Seyedi H. E., Rahaie M., Mofid M. R., Haddad L.. **Recombinant production of mecasermin in**. (2014) **9** 453. PMID: 26339260
19. Jarrard L. E.. **Use of excitotoxins to lesion the hippocampus: update.**. (2002) **12** 405-414. DOI: 10.1002/hipo.10054
20. Ji Z.-H., Xu Z.-Q., Zhao H., Yu X.-Y.. **Neuroprotective effect and mechanism of daucosterol palmitate in ameliorating learning and memory impairment in a rat model of Alzheimer’s disease.**. (2017) **119** 31-35. PMID: 28119081
21. Keimasi M., Salehifard K., Shahidi M., Esmaeili F., Esfahani N. M. J.. **Ameliorative effects of omega-lycotoxin-Gsp2671e purified from the spider venom of Lycosa praegrandis on memory deficits of glutamate-induced.**. (2022) **13** 1048563. DOI: 10.3389/fphar.2022.1048563
22. Kim H., Oh K. H.. **Protein network interacting with BK channels.**. (2016) **128** 127-161. DOI: 10.1016/bs.irn.2016.03.003
23. Koutsilieri E., Riederer P.. **Excitotoxicity and new antiglutamatergic strategies in Parkinson’s disease and Alzheimer’s disease.**. (2007) **13** S329-S331. DOI: 10.1016/S1353-8020(08)70025-7
24. Lewis R. J., Garcia M. L.. **Therapeutic potential of venom peptides.**. (2003) **2** 790-802. DOI: 10.1038/nrd1197
25. Li W., Liu L., Liu Y.-y, Luo J., Lin J.-y, Li X.. **Effects of electroconvulsive stimulation on long-term potentiation and synaptophysin in the hippocampus of rats with depressive behavior.**. (2012) **28** 111-117. DOI: 10.1097/YCT.0b013e31824a47ca
26. Li Y. C., Kavalali E. T.. **Synaptic vesicle-recycling machinery components as potential therapeutic targets.**. (2017) **69** 141-160. DOI: 10.1124/pr.116.013342
27. Mehta A., Prabhakar M., Kumar P., Deshmukh R., Sharma P.. **Excitotoxicity: bridge to various triggers in neurodegenerative disorders.**. (2013) **698** 6-18. DOI: 10.1016/j.ejphar.2012.10.032
28. Mochida S.. **Presynaptic calcium channels.**. (2019) **20** 2217. DOI: 10.3390/ijms20092217
29. Mofid M. R., Babaeipour V., Jafari S., Haddad L., Moghim S., Ghanavi J.. **Efficient process development for high-level production, purification, formulation, and characterization of recombinant mecasermin in**. (2021) **68** 776-788. DOI: 10.1002/bab.1990
30. Moradi M., Solgi R., Vazirianzadeh B., Tanzadehpanah H., Saidijam M.. **Scorpion venom and its components as new pharmaceutical approach to cancer treatment, a systematic review.**. (2018) **9** 1000-1012
31. Naseri F., Sirati-Sabet M., Sarlaki F., Keimasi M., Mokarram P., Siri M.. **The Effect of Ghrelin on Apoptosis, Necroptosis and Autophagy Programmed Cell Death Pathways in the Hippocampal Neurons of Amyloid-β 1–42-Induced Rat Model of Alzheimer’s disease.**. (2022) **28** 151. DOI: 10.1007/s10989-022-10457-3
32. Nentwig W., Blick T., Gloor D., Hänggi A., Kropf C.. (2017)
33. Nimmrich V., Gross G.. **P/Q-type calcium channel modulators.**. (2012) **167** 741-759. DOI: 10.1111/j.1476-5381.2012.02069.x
34. Paxinos G., Watson C.. (2006)
35. Przedborski S., Vila M., Jackson-Lewis V.. **Series Introduction: Neurodegeneration: What is it and where are we?**. (2003) **111** 3-10. DOI: 10.1172/JCI200317522
36. Rahn K. A., Slusher B. S., Kaplin A.. **Glutamate in CNS neurodegeneration and cognition and its regulation by GCPII inhibition.**. (2012) **19** 1335-1345. DOI: 10.2174/092986712799462649
37. Salehifard K., Radahmadi M., Reisi P.. **The effect of photoperiodic stress on anxiety-like behaviors, learning, memory, locomotor activity and memory consolidation in rats.**. (2023)
38. Sarlaki F., Shahsavari Z., Goshadrou F., Naseri F., Keimasi M., Sirati-Sabet M.. **The effect of ghrelin on antioxidant status in the rat’s model of Alzheimer’s disease induced by amyloid-beta.**. (2022) **12** 44-54. DOI: 10.37796/2211-8039.1341
39. Seyfi R., Babaeipour V., Mofid M. R., Kahaki F. A.. **Expression and production of recombinant scorpine as a potassium channel blocker protein in**. (2019) **66** 119-129. DOI: 10.1002/bab.1704
40. Simms B. A., Zamponi G. W.. **Neuronal voltage-gated calcium channels: structure, function, and dysfunction.**. (2014) **82** 24-45. DOI: 10.1016/j.neuron.2014.03.016
41. Sousa S. R., Vetter I., Lewis R. J.. **Venom peptides as a rich source of cav2. 2 channel blockers.**. (2013) **5** 286-314. DOI: 10.3390/toxins5020286
42. Szydlowska K., Tymianski M.. **Calcium, ischemia and excitotoxicity.**. (2010) **47** 122-129. DOI: 10.1016/j.ceca.2010.01.003
43. Tahami Monfared A. A., Byrnes M. J., White L. A., Zhang Q.. **The Humanistic and economic burden of alzheimer’s disease.**. (2022) **11** 525-551. PMID: 35192176
44. Tampellini D., Capetillo-Zarate E., Dumont M., Huang Z., Yu F., Lin M. T.. **Effects of synaptic modulation on β-amyloid, synaptophysin, and memory performance in Alzheimer’s disease transgenic mice.**. (2010) **30** 14299-14304. DOI: 10.1523/JNEUROSCI.3383-10.2010
45. Ullrich A., Böhme M. A., Schöneberg J., Depner H., Sigrist S. J., Noé F.. **Dynamical organization of syntaxin-1A at the presynaptic active zone.**. (2015) **11** e1004407. DOI: 10.1371/journal.pcbi.1004407
46. Valtorta F., Pennuto M., Bonanomi D., Benfenati F.. **Synaptophysin: leading actor or walk-on role in synaptic vesicle exocytosis?**. (2004) **26** 445-453. PMID: 15057942
47. Verma M., Lizama B. N., Chu C. T.. **Excitotoxicity, calcium and mitochondria: a triad in synaptic neurodegeneration.**. (2022) **11** 1-14. PMID: 34974845
48. Wheeler W. C., Coddington J. A., Crowley L. M., Dimitrov D., Goloboff P. A., Griswold C. E.. **The spider tree of life: phylogeny of Araneae based on target-gene analyses from an extensive taxon sampling.**. (2017) **33** 574-616. DOI: 10.1111/cla.12182
49. Xu J.-H., Tang F.-R.. **Voltage-dependent calcium channels, calcium binding proteins, and their interaction in the pathological process of epilepsy.**. (2018) **19** 2735. DOI: 10.3390/ijms19092735
50. You R., Ho Y.-S., Hung C. H.-L., Liu Y., Huang C.-X., Chan H.-N.. **Silica nanoparticles induce neurodegeneration-like changes in behavior, neuropathology, and affect synapse through MAPK activation.**. (2018) **15** 1-18. DOI: 10.1186/s12989-018-0263-3
51. Yu Y. X., Shen L., Xia P., Tang Y. W., Bao L., Pei G.. **Syntaxin 1A promotes the endocytic sorting of EAAC1 leading to inhibition of glutamate transport.**. (2006) **119** 3776-3787. DOI: 10.1242/jcs.03151
52. Zamani A., Mirshamsi O., Marusik Y. M., Moradmand M.. (2021)
53. Zamponi G. W., Snutch T. P.. **Modulation of voltage-dependent calcium channels by G proteins.**. (1998) **8** 351-356. DOI: 10.1016/S0959-4388(98)80060-3
54. Zhang F.-X., Sun Q.-J., Zheng X.-Y., Lin Y.-T., Shang W., Wang A.-H.. **Abnormal expression of synaptophysin. SNAP-25, and synaptotagmin 1 in the hippocampus of kainic acid-exposed rats with behavioral deficits.**. (2014) **34** 813-824. DOI: 10.1007/s10571-014-0068-3
55. Zhang L., Fang Y., Xu Y., Lian Y., Xie N., Wu T.. **Curcumin improves amyloid β-peptide (1-42) induced spatial memory deficits through BDNF-ERK signaling pathway.**. (2015) **10** e0131525. DOI: 10.1371/journal.pone.0131525
|
---
title: Development of a comprehensive flourishing intervention to promote mental health
using an e-Delphi technique
authors:
- Juliane Piasseschi de Bernardin Gonçalves
- Camilla Casaletti Braghetta
- Willyane de Andrade Alvarenga
- Clarice Gorenstein
- Giancarlo Lucchetti
- Homero Vallada
journal: Frontiers in Psychiatry
year: 2023
pmcid: PMC9981953
doi: 10.3389/fpsyt.2023.1064137
license: CC BY 4.0
---
# Development of a comprehensive flourishing intervention to promote mental health using an e-Delphi technique
## Abstract
### Background
Although observational studies have already shown promising results of flourishing, a broader concept of health based on positive psychology, there is still a gap in the literature regarding studies that combine different topics of flourishing in a single intervention.
### Objectives
To develop a comprehensive and integrate intervention based on positive psychology gathering different topics of flourishing to improve mental health outcomes in individuals with depressive symptoms.
### Methods
The following steps were performed: [1] a comprehensive literature review; [2] the designing of a 12-session group intervention based on the values, virtues, and topics of flourishing; [3] assessment of the rationale, coherence, and feasibility by a panel of healthcare professionals answering semi-structured questions, and [4] application of an e-Delphi technique including mental health experts to reach a consensus of at least $80\%$ for each item of the protocol.
### Results
A total of 25 experts participated in the study, 8 in the panel with semi-structured questions and 17 in the e-Delphi technique. A three-round e-Delphi technique was required to reach a consensus for all items. In the first round, a consensus was reached for $86.2\%$ of the items. The remaining items ($13.8\%$) were either excluded or reformulated. In the second round, a consensus was not obtained on one item, which was reformulated and approved in the third round. Qualitative analyses of the open questions were performed and suggestions for the protocol were considered. The final version of the intervention was composed of 12 weekly group sessions with 90-min each. The topics included in the intervention were physical and mental health, virtues and character strengths, love, gratitude, kindness, volunteering, happiness, social support, family, friends and community, forgiveness, compassion, resilience, spirituality, purpose and meaning of life, imagining the “best possible future,” and flourishing.
### Conclusion
The flourishing intervention was successfully developed using an e-Delphi technique. The intervention is ready to be tested in an experimental study to verify its feasibility and effectiveness.
## 1. Introduction
The mental health burden is increasing worldwide, posing several challenges to low-to-middle income countries such as Brazil. The World Health Organization (WHO) estimates that more than 300 million people have depression worldwide, and less than half have access to treatment [1].
Depression is currently the most significant cause of absenteeism and disability in the workforce [2] and the second largest cause of disability worldwide.
The incidence of depression has increased by $59.8\%$ from 1990 to 2017, and is the most common chronic disease worldwide, at present [3]. Depression treatment costed approximately US$ 236 billion in 2018 in USA, an increase of more than $35\%$ since 2010 [4]. According to recent data, at least $14\%$ of the Brazilian population suffers from depressive symptoms, while $17\%$ have had an episode of major depression throughout their lives [5].
The WHO defines health as “a state of complete physical, mental and social wellbeing and not only the absence of disease” [6]. However, this definition does not consider the dynamic nature of human beings. Therefore, new concepts are emerging in literature. We would like to highlight a specific one called flourishing, coined by Tyler VanderWeele, which defines health as “the state in which all aspects of a person's life are good” [7]. According to this concept, there are five broad domains of human life: (i) happiness and life satisfaction; (ii) health, both mental and physical; (iii) meaning and purpose; (iv) character and virtue; and (v) close social relationships and four pathways, i.e., family, work, education, and religious community [8].
Previous longitudinal studies with large samples have already shown promising effects of VanderWeele's flourishing dimensions/pathways on physical and mental health. A US national cohort with ~13,000 adults aged >50 years has shown that individuals with greater purpose of life had lower mortality risk [9] and that altruistic behaviors (e.g., volunteering) were associated with lower mortality, and better physical activity and psychosocial outcomes [10]. On similar lines, a US cohort with ~60,000 nurses found that forgiveness was associated with higher levels of positive affect and social integration and lower levels of psychological distress [11], and religious attendance was related to lower risk of death [12].
This longitudinal data supports that the broader concept of health has robust scientific evidence and should be discussed by the scientific literature. Within this context, several interventions based on human values and virtues were proposed, and publications on positive mental health programs [13, 14] showed promising results toward minimizing depressive symptoms in different populations (15–18). Although there are several clinical trials regarding the effectiveness of virtues, values, and character on health outcomes [19, 20], to the best of our knowledge, there is a scarcity of studies combining different aspects of flourishing, that focus on VanderWeele's concept [8]. Since the concept of flourishing embraces different dimensions which are linked to better health outcomes, promoting all of these dimensions in a single intervention may have better results than considering them separately.
The purpose of this study was to advance this field of research by creating an intervention protocol combining different aspects of flourishing to deliver in a few group sessions. Therefore, this study aimed to develop a simple, practical, and low-cost intervention protocol to promote mental health based on the conceptual framework of flourishing, through an e-Delphi technique.
## 2. Materials and methods
The study was approved by the Research Ethics Committee of the School of Medicine of the University of Sao Paulo, Brazil, under approval number CAAE: 52554221.4.0000.0068. All respondents provided written informed consent.
The study was organized into four phases, as shown in Supplementary Figure 1.
## 2.1. Phase 1: Comprehensive literature review
A non-systematic review was carried out in the following scientific databases: PubMed, Web of Science and Scopus, using the keywords “Flourish*,” “Mental Health,” “Positive Psychology,” “Intervention,” “Therapy,” and “Treatment.” We selected higher hierarchy evidence studies, such as meta-analyses performed through systematic reviews, randomized clinical trials, and cohort studies.
First, we selected studies that investigated the impact of the five domains of Tyler VanderWeele's concept of flourishing, as presented in the Introduction: (i) happiness and life satisfaction; (ii) physical and mental health; (iii) purpose and meaning of life; (iv) character and virtues, and (v) social relationship. Then, we elected the specific topics that integrate these domains.
## 2.2. Phase 2: Session design
The most relevant topics related to flourishing were selected from these studies and included in our intervention. We created the structure of the intervention based on the protocols used by the clinical trial and adopted a model with the aim of gaining complexity and deepening the themes at each session. In this phase, we also defined the specific objectives to guide the group discussion within the theme of the session, and the specific dynamics and exercises related to the theme, based on previously published clinical trials.
## 2.3. Phase 3: Panel with semi structured questions
The first version of the protocol was submitted to a committee of healthcare professionals, experts in the field of mental health, with experience in leading therapeutical groups or clinical practice related to spirituality and religiosity, for a non-systematic evaluation of its feasibility, robustness, and coherence, using a set of semi-structured questions. An email was sent to the experts explaining the flourishing intervention and inviting them to assess the attached protocol through an electronic questionnaire. After agreeing with the Informed Consent Form (ICF), they evaluated the protocol using a semi-structured questionnaire with open-ended questions about the intervention structure: content, format, target population, and intervention providers. The results obtained from the experts were used to remodel the protocol. Based on the modifications suggested by the experts, the protocol was then prepared for the structured e-Delphi phase.
## 2.4. Phase 4: Structured assessment through the e-Delphi technique
In this phase, another group of experts (different from Phase 3) were invited to assess the intervention through an e-Delphi technique. This type of systematic methodology allows receiving opinions and comments from a panel of selected experts [21, 22]. Thus, the opinion of experts on the subject can help point out possible topics of the intervention that need improvement or are inadequate. When the experts' answers are quantified, it serves as a guidance on the appropriateness of the evaluated item. If there is no consensus among the experts, the item should be changed or withdrawn [22].
Health professionals from medical and non-medical fields were included; they were different from those invited in the previous phase. The selection criteria were PhD holders with at least 10 years of experience in any of the following areas: mental health, spirituality/religiosity, positive psychology, or complementary health therapies. These professionals were selected based on their academic and clinical experience on themes related to the development of values and virtues in clinical practice and research, mental health care toward depressive patients and use of complementary therapies. Since there is no consensus on the sample size of experts needed for a panel adequate for an e-Delphi technique [23], the present study determined to include at least 17 individuals [22]. This choice was based on the findings of a previous systematic review that investigated the median number of panel members among 76 published protocols.
Questionnaires were developed on the SurveyMonkey® platform and the link was sent by email. Experts had 45 days to respond. The e-Delphi technique was expected to be conducted for as many rounds as necessary until consensus was achieved for all items.
The questionnaire consisted of general questions about each session and specific ones for the items that were considered the most challenging by the authors. It contained multiple choice questions and open questions for extra comments. The answers were structured with a Likert score of 1 to 5 points. Although no consensus is defined for the e-Delphi evaluation criteria, most studies use levels of agreement between 60 and $80\%$ [24]. Therefore, to be rigorous with the assessment of the protocol, we adopted a cut-off point of $80\%$ consensus with scores between 4 and 5 points. In the case of not reaching the cut-off point, an item could be eliminated or re-assessed.
Comments by the evaluators were examined, and the results of this analysis were used to improve the protocol and materials of the intervention.
## 3.1. Phase 1 and 2: Comprehensive literature review and session design
The themes of the intervention's sessions were derived using the conceptual framework and the activities for Flourishing proposed by Tyler VanderWeele and collaborators in a previous publication [25]. The selected topics of flourishing were: physical and mental health awareness; purpose and meaning of life; forgiveness; character strengths and virtues; kindness; volunteering; spirituality; gratitude; imagining the “best possible future”; love; compassion; social support, family, friends, and community; and happiness and resilience [25].
The framework in Figure 1 was developed to illustrate the interconnection between the pathways of flourishing, the topics selected for the intervention sessions, and domains of flourishing, demonstrating the rational used to develop the entire intervention protocol.
**Figure 1:** *Flourishing framework: The pathway interconnections between the domains and the aspects.*
Thereafter, we categorized the topics into three main groups to increase the complexity and enhance “human development” in the process of flourishing: Regarding the structural organization of the sessions, our search revealed that most positive psychology interventions varied from 1 to 20 weekly sessions [19, 20], with a duration of 1–5 h per session [19]. Therefore, in accordance with the literature, we decided to distribute the topics into 12 weekly sessions, each lasting for 90 min.
Also, the intervention was to be delivered on-line. Online interventions can be effective and feasible for treating mental health disorders, such as depression and anxiety [26, 27]. Advantages of such interventions include adaptability, multimedia presentations, and saving on traveling efforts. Due to these reasons and considering the traffic problems as a consequence of the population density and territorial extension of the city of São Paulo, Brazil, we decided to offer the intervention through the online format. Regarding the structure of the intervention, the sessions were to be conducted by one or two healthcare professionals where all participants could interact with the provider and among themselves.
Table 1 presents the topics in the sequential order of the initially proposed approach, goals for each session, and specific tools used to achieve the goals. All sessions aimed to promote individual reflection using different strategies such as, group discussions, writing exercises, guided meditation exercises, sharing videos and songs about the topic of the sessions, and reflective moments. Based on the theoretical model, we chose to describe the evidence found for each topic along with the design and goals proposed for the session.
**Table 1**
| Session | Objectives | Strategies used |
| --- | --- | --- |
| 1. Promotion of physical and mental health/presentation of the intervention | (a) Discuss health concepts (b) Impact of flourishing on depression | • Video about health and attitudes• Group discussions about health and lifestyle • Awareness exercises for changing lifestyle |
| 2. Gratitude and imagining a “better future” | (a) Identify reasons for gratitude in life (b) Plan the best possible future scenarios | • Video about gratitude and empathy• Group discussions about gratiful attitude future plan • Writing exercises about gratitude and the best future |
| 3. Love and compassion | (a) Reflect on the concepts of love and compassion (b) Stimulating acts of compassion and self-compassion | • Writing exercises about compassion acts • Guided mental visualizations about loving attitudes and compassion • Group discussions about behavior and health concerning compassion |
| 4. Acts of kindness and volunteering | (a) Reflect on acts of kindness and their own attitudes (b) Stimulate engagement in volunteering | • Video about kind attitude and volunteering • Group discussions about how to be kinder• Awareness exercises about volunteering in feasible manners |
| 5. Happiness | (a) Reflect on happiness(b) Encourage the construction of paths that involve happiness | • Guided mental visualization to identify happiness • Video audio of a lyric about the meaning of happiness • Group discussions about how to improve happiness in life• Writing exercises about attitudes regarding happiness |
| 6. Family and friends/community | (a) Identify and discuss the social support network (b) Stimulate growth and improvement of the identified network | • Awareness exercises about the current network • Video audio of a lyric about friendship • Group discussions on how to improve social support |
| 7. Forgiveness | (a) Identify reasons to forgive(b) Raise awareness about willingness and attitude of forgiveness (c) Reframe negative/challenging experiences | • Writing exercises about the will to forgive and strategies for accomplishing forgiveness • Group discussions about intentions and how to overcome pain |
| 8. Resilience | (a) Identify and raise awareness of the process of recovering from negative experiences (b) Reframe negative/challenging experiences(c) Raise awareness of problem flexibility | • Writing exercises about challenges in life and coping mechanisms • Group discussions about facing problems and reframing them • Awareness exercises about resilience strategies |
| 9. Character strengths and virtues | (a) Reflect on self-perception of one's own virtues(b) Stimulate the development of latent virtues and reinforce the present/identified ones | • Group discussions about the participant's virtues • Awareness exercises to strengthen their character |
| 10. Spirituality and transcendence | (a) Reflect on spirituality and coping (b) Stimulating a life that incorporates more spiritual aspects for the individual | • Group discussions about beliefs and spiritual vision in life • Awareness exercises to improve spiritual coping mechanisms |
| 11. Meaning and purpose of life | (a) Stimulate the search for a meaning in life(b) Ponder on the purposes of one's own existence | • Group text reading about meaning in life • Writing exercises for identifying meaning • Group discussions about how to act on a purpose in life |
| 12. Flourishing and closing the intervention | (a) Integrate all the concepts and virtues discussed (b) Flourish a number of aspects at the end of the program(c) Reinforce exercises for the aspects that made the most sense for the participant | • Group text reading about team effort • Group discussions about the interconnection of the sessions • Awareness exercises on how to develop and grow as a human being |
## 3.1.1. Session 1: Physical and mental health
There are different interventions to raise awareness of physical and mental health [28, 29], and evidence shows that higher awareness helps in seeking and adhering to the treatment. The objective of this session was to discuss different health concepts and ways to flourish with depressive symptoms [8, 30]. Providers would help participants understand different health concepts and reflect on their attitudes and behaviors toward their own health through video and group discussions.
## 3.1.2. Session 2: Gratitude and imagining your “better future”
Previous evidence based on a meta-analysis showed that gratitude interventions were effective to improve individuals' psychological wellbeing and gratitude levels but not anxiety [31]. In this context, writing lists of the things or people one is grateful for and writing letters to the people/situations were associated with better outcomes [32]. Likewise, envisioning the “best possible future” was another effective strategy compared to other motivational techniques, as observed in a previous meta-analysis [33].
The goal of this session was to encourage the participants to identify situations in their lives that they were grateful for. We proposed watching a reflexive video and writing exercises regarding gratitude and plans for a “better future.”
## 3.1.3. Session 3: Love and compassion
A meta-analysis of clinical trials on self-compassion showed a significant improvement in 11 psychosocial outcomes when compared to the controls [34]. The focus on compassion, however, showed a small effect size in another study [35]. Loving-kindness interventions (e.g., meditation) showed a small to large effect size in daily positive emotions when compared to other groups [35]. This session aimed to help the participants identify acts and feelings of love and compassion in their daily lives through guided imagery exercises for compassion and writing exercises with group discussions on loving attitudes.
## 3.1.4. Session 4: Kindness and volunteering
Regarding kindness and altruism, a recent meta-analysis pointed out a small effect size on wellbeing [36]. Another meta-analysis evidenced a reduction in mortality for individuals engaged in some voluntary work [37]. This evidence motivated the development of this session that aimed to make participants aware of the influence of kindness and volunteering on their health, using video and discussions about acts of kindness and volunteering. The provider would guide the group to create new, simple, and practical ideas about volunteering in everyday life.
## 3.1.5. Session 5: Happiness
There is solid evidence that happiness is associated with lower mortality and better mental health outcomes [38]. Based on previous intervention protocols that motivated participants to find “passion” [39, 40] and identify sources of happiness [41], the main goal of this session was to help participants reflect on how they could achieve a happier and healthier life, by making a list of what brought happiness to their lives, sharing these experiences with the group, and listening to music about the topic.
## 3.1.6. Session 6: Family, friends, and community
Longitudinal studies have shown that having family, friends, and a relationship are associated with a better quality of life [8, 42]. Likewise, there are clinical trials proposing strategies (e.g., positive psychology) to improve relationships [43, 44] and stimulating social support networks [45, 46] with positive results.
Thus, this session was designed for the participants to identify their social support network and improve the quality and quantity of their relationships. Providers would use open-questions and a song to raise awareness of participants' current network and discuss how to expand it.
## 3.1.7. Session 7: Forgiveness
A meta-analysis evaluated the effectiveness of psychotherapeutic interventions to promote forgiveness, showing reduced levels of depression and anxiety and higher hope when compared to other treatments [47]. Similarly, another meta-analysis showed that having empathy for the offender and overcoming feelings of unforgiveness were associated with lower levels of depression and anxiety [48].
Since most clinical trials use writing exercises to stimulate intention and feelings of forgiveness (49–51), we developed the session based on writing exercises to explore the participants' feelings and thoughts about forgiveness, identify reasons to forgive, and its impact on their health.
## 3.1.8. Session 8: Resilience
A meta-analysis showed that resilience is negatively associated with negative mental health outcomes and positively associated with positive outcomes [52]. Clinical trials proposed interventions based on cognitive-behavioral exercises and mindfulness [53, 54]. We based the design of this session on Steinhardt and Dolbier's model of (a) transforming stress into resilience, (b) taking responsibility, (c) focusing on empowerment interpretations, and (d) creating meaningful connections [55].
The resilience session aimed to identify the process of recovering from adversities and developing a more flexible vision of those adversities. Providers would stimulate strategies to develop better resilience in life.
## 3.1.9. Session 9: Strengths of character and virtues
A meta-analysis of character strengths interventions showed significant increases in positive affect, happiness, and life satisfaction, and lower levels of depression [56]. In this context, the classic intervention proposed by Seligman [57] showed that encouraging individuals to exercise their strongest character strengths weekly, after identifying them through a questionnaire, resulted in increased happiness and decreased depressive symptoms.
We designed the session using an abbreviated version of the Seligman Strengths and Character scale for the participants to identify their strongest and weakest virtues. The main objective was to increase the participants' perception of their virtues and reinforce their strengths by sharing different perspectives.
## 3.1.10. Session 10: Spirituality
Several studies have demonstrated the positive impact of spirituality-based interventions on mental health outcomes, such as lower levels of anxiety and depression [58, 59]. Protocols are usually based on motivational group discussions addressing topics such as faith, spiritual beliefs, and peace [60, 61].
The session was based on the material developed by Hopkins et al. [ 62], with the objective to reflect on the influence of the individuals' beliefs on health by discussing the history of their belief system, and their connection with the sacred, others, and nature. Finally, providers would help participants on how to use these tools and spiritual practice to develop a healthier path.
## 3.1.11. Session 11: Meaning and purpose of life
Meta-analyses on the meaning of life and its impact on health have shown positive correlation with life satisfaction and negative correlation with negative affect [63, 64]. Likewise, there is growing evidence that clinical trials based on meaning-centered therapies are associated with better psychological outcomes [65].
This session was developed based on an intervention proposed by Luz et al. which used different reflexive texts about meaning of life [66], allowing participants to identify their meaning in life and consider a purpose to follow. Providers would help participants with their process of identifying and creating life purposes and being able to live it by proposing discussion from the reading of a text and exchange of impressions and personal experiences.
## 3.1.12. Session 12: Flourishing
Finally, the aim of the last session was to integrate the virtues and values of the aspects of flourishing into participants' life, reinforcing the idea of human development. The session would encourage the participants to reflect on their role in life and assume responsibility for their lives and situations through text reading and group discussions [25, 67, 68].
## 3.2. Phase 3: Panel with semi structured questions
The panel was composed of eight experts: five women ($71.4\%$) and three men ($28.6\%$). Their professions were psychologists (four), physicians (two), social worker (one), and spiritual counselor (one), and all had more than 10 years of professional experience.
According to the experts, the intervention seemed to be a positive, innovative, and well-grounded proposal. They pointed out the following strengths: the scope of the themes, potential to stimulate reflection, and possibility of being replicated in different scenarios. The experts agreed with the feasibility of the intervention, emphasizing clarity and objectivity, and the applicability of the exercises. The number and duration of the sessions were considered appropriate. Most experts agreed that the providers should be healthcare professionals with previous adequate training on the subject and the sessions.
According to the experts, the population that could benefit from this intervention were adults and older adults with mild and moderate depressive symptoms. However, they discouraged the use of this intervention for individuals with severe symptoms. Furthermore, experts mentioned that individuals with low education could benefit from the intervention as well. However, they pointed out two possible challenges to be tested in practice: the writing exercises and the difficulties in understanding complex concepts such as resilience and flourishing. Finally, they reported that some conceptual adaptations and simplifications would be necessary.
Based on the opinions of this panel, adaptations were made in the protocol and the revised version was sent to an e-Delphi panel.
## 3.3. Phase 4: Structured assessment through an e-Delphi technique
Of the 21 experts invited, 3 were not available to collaborate in due time, and 1 refused to provide consent to participate. Therefore, the final panel was composed of 17 individuals: 10 men ($58.8\%$) and 7 women ($41.2\%$). The average age was 52.2 (SD = 14.5) years, with an average professional experience of 25.9 (SD = 10.6) years, and all of them were PhD holders. The panel included eight physicians, four psychologists, three university professors, one physical therapist, and one nurse.
This phase was separated as follows: (a) E-Delphi technique assessment: This included 58 items regarding specific opinions on the objectives, strategies, and interventions for each session, using a Likert scale ranging from 1 (totally disagree) to 5 (totally agree).
1st round Table 2 presents the specific items assessed if consensus was achieved for them, and the conduct provided by the authors. Consensus was obtained on 50 out of the 58 items in the first round, representing $86.2\%$ agreement among experts; these topics and exercises were maintained in the intervention protocol. The remaining eight items ($13.8\%$) were from the following topics: three items from session 2 “gratitude and imagining a better future,” and one item each from session 6 “family, friends, and community,” session 7 “forgiveness,” session 10 “spirituality,” and session 11 “meaning and purpose of life.” All of these items were either excluded or reformulated and re-sent to the experts in the second e-Delphi round.
**Table 2**
| Evaluated items | First round | First round.1 | Second round | Second round.1 | Third round | Third round.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Evaluated items | | | | | | |
| | Consensus (%) | Conduct | Consensus (%) | Conduct | Consensus (%) | Conduct |
| Session 1: Promotion of physical and mental health/presentation of the intervention | Session 1: Promotion of physical and mental health/presentation of the intervention | Session 1: Promotion of physical and mental health/presentation of the intervention | Session 1: Promotion of physical and mental health/presentation of the intervention | Session 1: Promotion of physical and mental health/presentation of the intervention | Session 1: Promotion of physical and mental health/presentation of the intervention | Session 1: Promotion of physical and mental health/presentation of the intervention |
| 1.1. How clear is the purpose of the session? | 94.1 | Maint. | – | – | – | – |
| 1.2. How adequate is the dynamics of the initial presentation to the group? | 88.2 | Maint. | – | – | – | – |
| 1.3. How adequate are the activities used to discuss the concept of health with the participants? | 82.4 | Maint. | – | – | – | – |
| Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” |
| 2.1. How clear is the purpose of the session? | 76.5 | Revis. | 100 | Maint. | – | – |
| 2.2. How adequate is the exercise in which the participant is asked to name a historical figure they draw inspiration from? | 70.6 | Exclu.* | – | – | – | – |
| 2.3. How appropriate is the exercise that encourages the participant to imagine a “bright and happy” future? | 88.2 | Maint. | – | – | – | – |
| 2.4. How appropriate are the homework exercises? | 70.6 | Exclu.* | – | – | – | – |
| Session 3: Love and compassion | Session 3: Love and compassion | Session 3: Love and compassion | Session 3: Love and compassion | Session 3: Love and compassion | Session 3: Love and compassion | Session 3: Love and compassion |
| 3.1. How clear is the purpose of the session? | 82.4 | Maint. | – | – | – | – |
| 3.2. How important is it to create a list of traits of loving/kind people the participant relates to? | 94.1 | Maint. | – | – | – | – |
| 3.3. How important is it to promote love and compassion through guided mental visualizations? | 88.2 | Maint. | – | – | – | – |
| 3.4. How important is it to discuss how the participant feels about guided mental imagery? | 94.1 | Maint. | – | – | – | – |
| 3.5. How important is it for the participants to exercise love and compassion through the homework? | 88.2 | Maint. | – | – | – | – |
| 3.6. How appropriate are the homework exercises? | 82.4 | Maint. | – | – | – | – |
| Session 4: Acts of kindness and volunteering | Session 4: Acts of kindness and volunteering | Session 4: Acts of kindness and volunteering | Session 4: Acts of kindness and volunteering | Session 4: Acts of kindness and volunteering | Session 4: Acts of kindness and volunteering | Session 4: Acts of kindness and volunteering |
| 4.1. How clear is the purpose of the session? | 88.2 | Maint. | – | – | – | – |
| 4.2. How important is it for the participant to reflect on their experiences with volunteering? | 94.1 | Maint. | – | – | – | – |
| 4.3. How important is it to encourage the participant to engage in volunteering activities in their community? | 94.1 | Maint. | – | – | – | – |
| 4.4. How appropriate are the homework exercises? | 100 | Maint. | – | – | – | – |
| Session 5: Happiness | Session 5: Happiness | Session 5: Happiness | Session 5: Happiness | Session 5: Happiness | Session 5: Happiness | Session 5: Happiness |
| 5.1. How clear is the purpose of the session? | 94.1 | Maint. | – | – | – | – |
| 5.2. How adequate is it to reflect on the participant's perception of happiness? | 88.2 | Maint. | – | – | – | – |
| 5.3. How important is it to reflect on what makes the participant happy and what level of happiness they are currently experiencing? | 100 | Maint. | – | – | – | – |
| 5.4. How important is the exercise addressing barriers and paths to happiness? | 100 | Maint. | – | – | – | – |
| 5.5. How appropriate are the homework exercises? | 94.1 | Maint. | – | – | – | – |
| Session 6: Family and friends/community | Session 6: Family and friends/community | Session 6: Family and friends/community | Session 6: Family and friends/community | Session 6: Family and friends/community | Session 6: Family and friends/community | Session 6: Family and friends/community |
| 6.1. How clear is the purpose of the session? | 88.2 | Maint. | – | – | – | – |
| 6.2. How important is it to discuss the meaning of family and friends with the participant? | 88.2 | Maint. | – | – | – | – |
| 6.3. How appropriate is it to present the concept of social support to the group? | 82.4 | Maint. | – | – | – | – |
| 6.4. How suitable is the song chosen by the research team for the session? | 70.6 | Revis. | 82.3 | Maint. | – | – |
| 6.5. How appropriate is it to reflect on the changes the participant has made in his/her environment and relationships since the begining of the intervention? | 82.4 | Maint. | – | – | – | – |
| 6.6. How important is it for the participant to count on his/her family to solve everyday problems? | 94.1 | Maint. | – | – | – | – |
| 6.7. How appropriate are the homework exercises? | 88.2 | Maint. | – | – | – | – |
| Session 7: Forgiveness | Session 7: Forgiveness | Session 7: Forgiveness | Session 7: Forgiveness | Session 7: Forgiveness | Session 7: Forgiveness | Session 7: Forgiveness |
| 7.1. How clear is the purpose of the session? | 100 | Maint. | – | – | – | – |
| 7.2. How appropriate is it to talk about situations that upset the participant and made them want to forgive? | 94.1 | Maint. | – | – | – | – |
| 7.3. How important is it to mention potential benefits of forgiveness? | 94.1 | Maint. | – | – | – | – |
| 7.4. How important is it to encourage the participant to change their perspective by putting themself in the shoes of the person who wronged them? | 76.5 | Exclu.* | – | – | – | – |
| 7.5. How important is it to encourage the participant to put themself i the other people's shoes? | 70.6 | Exclu.* | – | – | – | – |
| 7.6. How appropriate are the homework exercises? | 88.2 | Maint. | – | – | – | – |
| Session 8: Resilience | Session 8: Resilience | Session 8: Resilience | Session 8: Resilience | Session 8: Resilience | Session 8: Resilience | Session 8: Resilience |
| 8.1. How clear is the purpose of the session? | 88.2 | Maint. | – | – | – | – |
| 8.2. How easy is it to understand the exercise that presents the phases of a problem to discuss resilience? | 88.2 | Maint. | – | – | – | – |
| 8.3. How clear is the exercise that encourages the participant to look at the positive and negative aspects of a problem? | 82.4 | Maint. | – | – | – | – |
| 8.4. How appropriate are the homework exercises? | 82.4 | Maint. | – | – | – | – |
| Session 9: Character strengths and virtues | Session 9: Character strengths and virtues | Session 9: Character strengths and virtues | Session 9: Character strengths and virtues | Session 9: Character strengths and virtues | Session 9: Character strengths and virtues | Session 9: Character strengths and virtues |
| 9.1. How clear is the purpose of the session? | 100 | Maint. | – | – | – | – |
| 9.2. How adequate is it to present Seligman's Strength of Character Scale to the participant? | 82.4 | Maint. | – | – | – | – |
| 9.3. How appropriate is it to ask the participant to name their strongest virtue and situations in which they had to apply it? | 94.1 | Maint. | – | – | —- | – |
| 9.4. How appropriate are the homework exercises? | 82.4 | Maint. | – | – | – | – |
| Session 10: Spirituality | Session 10: Spirituality | Session 10: Spirituality | Session 10: Spirituality | Session 10: Spirituality | Session 10: Spirituality | Session 10: Spirituality |
| 10.1. How clear is the purpose of the session? | 76.5 | Revis. | 76.4 | Revis. | 100 | Maint. |
| 10.2. How adequate is it for the participant to reflect on their own concept of spirituality? | 82.4 | Maint. | – | – | – | – |
| 10.3. How appropriate is it for the participant to reflect on moments in their life when they clunge to or abandoned their spiritual beliefs? | 82.4 | Maint. | – | – | – | – |
| 10.4. How appropriate is it for the participant to reflect on the role spirituality plays in their lives today? | 94.1 | Maint. | – | – | – | – |
| 10.5. How appropriate are the homework exercises? | 82.4 | Maint. | – | – | – | – |
| Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life |
| 11.1. How clear is think the purpose of the session? | 88.2 | Maint. | – | – | – | – |
| 11.2. How appropriate is the reading proposed for the session? | 64.7 | Revis. | 82.3 | Maint. | – | – |
| 11.3. How important is it to discuss the participant's goals in life? | 94.1 | Maint. | – | – | – | – |
| 11.4. How important is it to discuss the alignment between the participant's goals in life and their occupation? | 88.2 | Maint. | – | – | – | – |
| 11.5. How appropriate are the homework exercises? | 82.4 | Maint. | – | – | – | – |
| Session 12: Flourishing and closing the intervention | Session 12: Flourishing and closing the intervention | Session 12: Flourishing and closing the intervention | Session 12: Flourishing and closing the intervention | Session 12: Flourishing and closing the intervention | Session 12: Flourishing and closing the intervention | Session 12: Flourishing and closing the intervention |
| 12.1. How clear is the purpose of the session? | 82.4 | Maint. | – | – | – | – |
| 12.2. How adequate is the reading proposed in the session? | 94.1 | Maint. | – | – | – | – |
| 12.3. How important is it for the participants to reflect on the activities carried out throughout the program? | 100 | Maint. | – | – | – | – |
| 12.4. How helpful is it for the participants to reflect on the changes they experienced during the intervention program? | 100 | Maint. | – | – | – | – |
| 12.5. How appropriate is the song chosen by the research team for the final session? | 82.4 | Maint. | – | – | – | – |
2nd round The authors considered the qualitative answers of the experts regarding the eight revised items and decided to exclude four items from the protocol and reformulate the other four to obtain consensus. Table 3 shows the summary of the experts' comments along with the authors' appropriate explanation for the reformulation or exclusion of the items. Only the “spirituality” objective of session 10 did not obtain a consensus and was revised again.
**Table 3**
| Evaluated items | Conduct | Summary of experts comments |
| --- | --- | --- |
| Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” | Session 2: Gratitude and imagining a “better future” |
| 2.1. How clear is the purpose of the session? | Revis. | Both topics were considered unrelated to what was discussed in the session |
| 2.2. How adequate is the exercise in which the participant is asked to name a historical figure they draw inspiration from? | Exclu.* | The exercise at issue does not contribute much to the gratitude topic. |
| 2.4. How appropriate are the homework exercises? | Exclu.* | Need to simplify the exercises with practical strategies so that the participants can continue improving after the program. |
| Session 6: Family and friends/community | Session 6: Family and friends/community | Session 6: Family and friends/community |
| 6.4. How suitable is the song chosen by the research team for the session? | Revis. | The song chosen could bring up feelings of sadness and nostalgia |
| Session 7: Forgiveness | Session 7: Forgiveness | Session 7: Forgiveness |
| 7.4. How important is it to encourage the participant to change their perspective by putting themself in the shoes of the person who wronged them? | Exclu.* | This specific exercise could awaken feelings of guilt or traumatic memories |
| 7.5. How important is it to encourage the participant to put themself i the other people's shoes? | Exclu.* | This specific exercise could awaken feelings of guilt or traumatic memories |
| Session 10: Spirituality | Session 10: Spirituality | Session 10: Spirituality |
| 10.1. How clear is the purpose of the session? | Revis. | Participants may have religious or non-religious beliefs and experiences, so the objective of the session should be to embrace diversity/explore both belief systems |
| Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life | Session 11: Meaning and purpose of life |
| 11.2. How appropriate is the reading proposed for the session? | Revis. | The text focuses more on gratitude and the overcoming of obstacles than on an existential perspective |
3rd round The third round of the e-Delphi comprised a revised aim of the spirituality session. The experts' suggestions were mainly to highlight the aspects of spirituality related to transcendence, promote the meaning and experience of connection, and emphasize that the session goal should be more inclusive and cover religious and non-religious people. The item was reformulated, and the session renamed to broaden the complexity of the concept of spirituality adopted for the session. The item was then submitted for a third round of evaluation and was approved with $100\%$ consensus.
The agreement for the ratings of all Delphi items between experts (inter-rater reliability) were assessed using the intra-class correlation coefficients (ICC) (two-way mixed model, consistency, average-measures ICC), yielding a coefficient of 0.48 (CI $95\%$: 0.25–0.66, $p \leq 0.001$).
(b) General protocol organizational opinions: This included 10 items concerning organizational aspects of the protocol, such as target public, number and duration of sessions, multimedia presentation, and the provider's guide, using multiple choice questions and Likert scale ranging from 1 (totally disagree) to 5 (totally agree).
Figure 2 shows the opinions of the experts regarding the structure and implementation of the protocol. Experts highlighted the appropriateness of the overall process of flourishing, target population of those with mild depressive symptoms, 12 weekly online delivery sessions, use of the audiovisual resources, and the guideline created for the providers. However, some experts raised two points of warning: implementing the intervention for individuals with moderate depressive symptoms and duration of the sessions.
**Figure 2:** *Experts' opinions about the structure and implementation of the protocol.*
It is important to emphasize that changes were necessary to improve the coherence of the intervention as a whole. This included the changes related only to the structure (writing and distribution), and not to the content of the items, and the opinions on the e-Delphi technique were respected during this process. Another aspect that should be highlighted is that despite the reorganization of the themes, the protocol remained of 12 sessions.
(c) Experts' suggestions: Experts were allowed to suggest and comment in support of their opinions concerning the questions they were asked. They provided practical suggestions regarding group dynamics for the implementation of the protocol, that were included in the final manual created by the authors for the training of healthcare providers of the intervention in future.
The complementary exercises of the sessions were gathered in a single file, and it was decided to provide this material to the participants at the end of the intervention (post-intervention material), so that they could continue following the proposals and exercises in their daily lives, resulting in continuation of its flourishing process. Experts mentioned the importance of a strategy to ensure continuity of exercises for constant personal improvement, even after of the intervention ended.
## 4. Discussion
The present study successfully developed an intervention protocol based on the flourishing concept of health using a comprehensive literature review and the opinion of experts through the e-Delphi technique. The protocol consists of sharing therapeutic tools during online sessions in collaboration with trained healthcare professionals to conduct the sessions. Though there are other mental health programs focusing on positive aspects, but in this intervention, we specifically sough to adopt the VanderWeele's concept of flourishing and combine its various topics into the sessions offered [13, 14]. This intervention is different from others because it combines all virtues and values in a single intervention, resulting in a more holistic and comprehensive approach.
For the development of this protocol, our study followed previous articles that used Delphi or e-Delphi to develop an intervention [21, 22, 69]. Delphi is a technique that can impact ways of thinking or decision making through the convergence of opinions and comments of experts' assessments [21]. The main advantage of using such a technique is the possibility of exploring underlying assumptions regarding a specific topic [69].
According to the literature, one of the most important aspects of *Delphi is* the choice of experts. Our experts were mostly from the field of mental health, and this choice was made considering the main outcome of our intervention (reducing depressive symptoms and promoting mental health) and the experience of the experts in the precepts of positive psychology. However, it is important to highlight that, to improve the public health feasibility of this protocol, professionals from other healthcare fields were also included. Another important choice for our panel selection was to include experienced PhD holders for the e-Delphi. This choice guaranteed that well-qualified experts with expertise in both clinical practice and research were included, which was supported by previous studies [21, 22].
An important choice in the development of our protocol was the use of an online e-Delphi technique, instead of a Delphi technique. The e-Delphi has some important advantages such as time and cost savings, convenience for experts and the research team, and data management facility [21, 22]. The individualized communication with experts and their blindness regarding the other experts' answers during the process may have contributed to the impartiality of scores, ideas, and suggestions provided.
Regarding the protocol, the experts favored the online format of the interventions. The advantages of using this type of approach include the use of multimedia for the participants, facilitating access to complex content, time and cost savings, and the possibility of engaging with others [26, 27]. The use of an online approach is supported by previous studies where clinical trials of brief online intervention programs showed better long-term clinical effects compared to face-to-face therapies for different mental health conditions, especially when the technique included the support of a healthcare professional [70].
As verified through the results, experts reached consensus for ~$90\%$ of the items, revealing that the intervention was consistent with the proposal of flourishing. However, it is important to highlight that one item was subjected to three rounds of e-Delphi to reach consensus: the objective of the spirituality session. To achieve the goal for this session the following concept mentioned by Puchalski [71] was adopted: “spirituality is the aspect of humanity that refers to the way individuals seek and express meaning and purpose and the way they experience their connectedness to the moment, to self, to others, to nature, and to the significant or sacred.” Some experts argued that spirituality should be related to the sacred or an immortal being and not to the connectedness to the moment, self, or others. Since our main goal in this session was to be inclusive, Puchalski's definition was chosen because it provides a more comprehensive understanding of this construct, in which individuals can recognize their beliefs and values, and identify which connections are meaningful to them, regardless of sacred or religious beliefs. Most experts highlighted the notions of transcendence and connection as relevant aspects to be addressed in this session. In the third round a consensus was reached and the session was renamed “Spirituality and Inner Connection.” Another important highlight by the experts was the order of the topics of flourishing. The experts did not initially agree to address gratitude and imagining your “best future” in the same session, even though both topics could stimulate an individual to connect with the environment and with others. Comments included that the exercises proposed should be provided separately to achieve their personal goals, and that imagining a “better future” could be more useful at the end of the intervention to synthesize the concept of flourishing. The authors revisited the order of the topics maintaining the rationale initially proposed, to achieve an increase in complexity of the virtues and values of the process of flourishing [8].
Regarding the structure of the intervention, the experts agreed on the 12 weekly sessions; however, not all experts agreed with the duration of 90-min. Since online interventions have operational challenges, such as speed of connection and learning of operating the multimedia [70], a loss of therapeutic time should be considered. Also, group sessions including 10–15 participants need adequate time to allow everyone to express their ideas and opinions. Therefore, this program needs to be tested in clinical practice so that its viability can be confirmed. The experts were also consulted regarding the general use of the multimedia resources, slides for the sessions, and proposed items to include a variety of tools to ensure a better group dynamic, and most opinions were favorable.
Concerning the target population, people with mild and moderate depressive symptoms tend to respond well to most psychotherapy approaches, irrespective of whether those are self-help or cognitive-behavioral therapy (CBT) interventions (72–74). There are, however, reports of lower effects and high rates of treatment dropout when no human therapeutic support is offered, and the patient has only access to the content generated by the electronic platform [74, 75]. There is clear evidence that the computer cannot fully replace human contact [76]. Therefore, our intervention was developed as a synchronous online approach including full support of the healthcare professionals throughout the program for the participants, making it a relevant therapeutic alliance [73].
Finally, it is noteworthy that most studies investigating mood disorders and web-based treatments do not deal with major depression diagnoses, but with symptoms of depression [72, 77]. Given the evidence found in the literature, our intervention was designed for individuals with depressive symptoms and not for those with a diagnosis of major depression. Most experts agreed that this intervention could be used for mild to moderate depressive symptoms, which is supported by the previous literature. Future studies using this protocol should explore which group of participants will benefit the most through this intervention.
The study has some limitations. First, the experts who evaluated the protocol were invited to participate through the authors' professional relationship network. Purposive sampling was used to recruit professionals to serve in the expert committee. However, experts were form different institutions, had different backgrounds, and experiences in the field. Second, the intervention was developed in Portuguese language. Although we had an English version translated by a professional native English speaker translator, no cross-cultural adaptation was carried out. This protocol should be tested in other languages and cultures to verify if the sessions are feasible and meaningful, aiming to make the flourishing intervention culturally sensitive.
## 5. Ethics and dissemination
The development of the flourishing intervention aims to fill the gap in public health approaches since it is a practical and low-cost intervention. Interventions in groups have an advantage of providing greater efficiency and cost-effectiveness to health services [78], besides the benefits to patients [79], as they facilitate the exchange of experiences and support by expanding the social support network.
Furthermore, the flourishing intervention aims to train healthcare professionals from different backgrounds and expertise, providing fundamental skills to conduct the intervention as a new healthcare tool. Investment in continuous education of multidisciplinary teams has proved to be a significant resource for implementing technical innovations, which aim to change practices in health systems and, consequently, in communities [80]. The authors developed a guide for the providers on specific information for conducting the sessions, such as troubleshooting and managing therapeutic groups. The experts had access to this material and highlighted its relevance. The providers' guide is an accessible material for healthcare professionals to understand the exercises and tools proposed for the sessions. The intention is to allow this to be reproducible in different healthcare contexts. A study pointed out that it is important for professionals conducting online interventions to have clinical experience, empathy, and robust knowledge of the technique used, otherwise the intervention may have less effect [26]. Therefore, adequate material and training are essential to ensure the best use and effectiveness of the intervention.
## 6. Conclusion
The development of the intervention based on the concept of flourishing obtained consensus from the experts. Necessary adjustments were made by the team, and the protocol reformulation proved to be successful. The intervention is ready to be tested in an experimental study to verify its feasibility and effectiveness.
## 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 study was approved by the Research Ethics Committee of the School of Medicine of the University of Sao Paulo, Brazil, under approval number CAAE: 52554221.4.0000.0068. All respondents provided written informed consent. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
JG: conception and design of the work, data collection, data analysis and interpretation, drafting the article, and final approval of the version to be published. CB: data collection, data analysis and interpretation, drafting the article, and final approval of the version to be published. WA: data analysis and interpretation, critical revision of the article, and final approval of the version to be published. CG: data interpretation, critical revision of the article, and final approval of the version to be published. GL: conception and design of the work, data analysis and interpretation, critical revision of the article, and final approval of the version to be published. HV: conception and design of the work, data interpretation, critical revision of the article, and final approval of the version to be published. 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/fpsyt.2023.1064137/full#supplementary-material
## References
1. Marcus M, Yasamy M, van Ommeren M, Chisholm D, Saxena S. *Depression: A Global Public Health Concern. WHO Department of Mental Health Substance Abuse* (2012)
2. Gordon BR, McDowell CP, Hallgren M, Meyer JD, Lyons M, Herring MP. **Association of efficacy of resistance exercise training with depressive symptoms: meta-analysis and meta-regression analysis of randomized clinical trials**. *JAMA Psychiatry.* (2018) **75** 566-76. DOI: 10.1001/jamapsychiatry.2018.0572
3. Liu Q, He H, Yang J, Feng X, Zhao F, Lyu J. **Changes in the global burden of depression from 1990 to 2017: findings from the Global Burden of Disease study**. *J Psychiatr Res.* (2020) **126** 134-40. DOI: 10.1016/j.jpsychires.2019.08.002
4. Greenberg PE, Fournier A-A, Sisitsky T, Simes M, Berman R, Koenigsberg SH. **The economic burden of adults with major depressive disorder in the United States (2010 and 2018)**. *Pharmacoeconomics.* (2021) **39** 653-65. DOI: 10.1007/s40273-021-01019-4
5. Silva MT, Galvao TF, Martins SS, Pereira MG. **Prevalence of depression morbidity among Brazilian adults: a systematic review and meta-analysis**. *Rev Bras Psiquiatr.* (2014) **36** 262-70. DOI: 10.1590/1516-4446-2013-1294
6. Larson JS. **The World Health Organization's definition of health: Social vs. spiritual health**. *Soc Indic Res.* (1996) **38** 181-92
7. VanderWeele TJ, McNeely E, Koh HK. **Reimagining health—flourishing**. *JAMA.* (2019) **321** 1667-8. DOI: 10.1001/jama.2019.3035
8. VanderWeele TJ. **On the promotion of human flourishing**. *Proc Natl Acad Sci.* (2017) **114** 8148-56. DOI: 10.1073/pnas.1702996114
9. Shiba K, Kubzansky LD, Williams DR, VanderWeele TJ, Kim ES. **Associations between purpose in life and mortality by SES**. *Am J Prev Med.* (2021) **61** e53-61. DOI: 10.1016/j.amepre.2021.02.011
10. Kim ES, Whillans A V, Lee MT, Chen Y, VanderWeele TJ. **Volunteering and subsequent health and well-being in older adults: an outcome-wide longitudinal approach**. *Am J Prev Med.* (2020) **59** 176-86. DOI: 10.1016/j.amepre.2020.03.004
11. Long KNG, Worthington ELJ, VanderWeele TJ, Chen Y. **Forgiveness of others and subsequent health and well-being in mid-life: a longitudinal study on female nurses**. *BMC Psychol.* (2020) **8** 104. DOI: 10.1186/s40359-020-00470-w
12. Chen Y, Koh HK, Kawachi I, Botticelli M, VanderWeele TJ. **Religious service attendance and deaths related to drugs, alcohol, and suicide among US health care professionals**. *JAMA Psychiatry.* (2020) **77** 737-44. DOI: 10.1001/jamapsychiatry.2020.0175
13. Eisenstadt M, Liverpool S, Infanti E, Ciuvat RM, Carlsson C. **Mobile apps that promote emotion regulation, positive mental health, and well-being in the general population: systematic review and meta-analysis**. *JMIR Ment Heal.* (2021) **8** e31170. DOI: 10.2196/31170
14. Carr A, Cullen K, Keeney C, Canning C, Mooney O, Chinseallaigh E. **Effectiveness of positive psychology interventions: a systematic review and meta-analysis**. *J Posit Psychol.* (2021) **16** 749-69. DOI: 10.1080/17439760.2020.1818807
15. Breitbart W, Rosenfeld B, Gibson C, Pessin H, Poppito S, Nelson C. **Meaning-centered group psychotherapy for patients with advanced cancer: a pilot randomized controlled trial**. *Psychooncology.* (2010) **19** 21-8. DOI: 10.1002/pon.1556
16. Kelly JF, Stout RL, Magill M, Tonigan JS, Pagano ME. **Spirituality in recovery: a lagged mediational analysis of alcoholics anonymous' principal theoretical mechanism of behavior change**. *Alcohol Clin Exp Res.* (2011) **35** 454-63. DOI: 10.1111/j.1530-0277.2010.01362.x
17. Koenig HG, Pearce MJ, Nelson B, Daher N. **Effects of religious vs. standard cognitive-behavioral therapy on optimism in persons with major depression and chronic medical illness**. *Depress Anxiety.* (2015) **32** 835-42. DOI: 10.1002/da.22398
18. Lucchetti AL, Peres MF, Vallada HP, Lucchetti G. **Spiritual treatment for depression in brazil: an experience from spiritism**. *Exploration.* (2015) **11** 377-86. DOI: 10.1016/j.explore.2015.07.002
19. Casellas-Grau A, Font A, Vives J. **Positive psychology interventions in breast cancer. A systematic review**. *Psychooncology.* (2014) **23** 9-19. DOI: 10.1002/pon.3353
20. Meyers MC, van Woerkom M, Bakker AB. **The added value of the positive: a literature review of positive psychology interventions in organizations**. *Eur J Work Organ Psychol.* (2013) **22** 618-32. DOI: 10.1080/1359432X.2012.694689
21. Laustsen S, Brahe L. **Applying the Delphi method to generate interventions to reduce unnecessary interruptions in clinical nursing**. *Nord J Nurs Res.* (2015) **35** 249-55. DOI: 10.1177/0107408315603630
22. Sampaio FMC, Sequeira C, Lluch Canut T. **Content validity of a psychotherapeutic intervention model in nursing: a modified e-Delphi study**. *Arch Psychiatr Nurs.* (2017) **31** 147-56. DOI: 10.1016/j.apnu.2016.09.007
23. Boulkedid R, Abdoul H, Loustau M, Sibony O, Alberti C. **Using and reporting the Delphi method for selecting healthcare quality indicators: a systematic review**. *PLoS ONE.* (2011) **6** e20476. DOI: 10.1371/journal.pone.0020476
24. Santos O, Lopes E, Virgolino A, Stefanovska-Petkovska M, Dinis A, Ambrósio S. **Defining a brief intervention for the promotion of psychological well-being among unemployed individuals through expert consensus**. *Front Psychiatry.* (2018) **9** 13. DOI: 10.3389/fpsyt.2018.00013
25. VanderWeele TJ. **Activities for flourishing: an evidence-based guide**. *J Posit Sch Psychol.* (2020) **4** 79-91
26. Andersson G. **Using the internet to provide cognitive behaviour therapy**. *Behav Res Ther.* (2009) **47** 175-80. DOI: 10.1016/j.brat.2009.01.010
27. Rathbone AL, Clarry L, Prescott J. **Assessing the efficacy of mobile health apps using the basic principles of cognitive behavioral therapy: systematic review**. *J Med Internet Res.* (2017) **19** e399. DOI: 10.2196/jmir.8598
28. Teng EJ, Friedman LC. **Increasing mental health awareness and appropriate service use in older Chinese Americans: a pilot intervention**. *Patient Educ Couns.* (2009) **76** 143-6. DOI: 10.1016/j.pec.2008.11.008
29. Livingston JD, Tugwell A, Korf-Uzan K, Cianfrone M, Coniglio C. **Evaluation of a campaign to improve awareness and attitudes of young people towards mental health issues**. *Soc Psychiatry Psychiatr Epidemiol.* (2013) **48** 965-73. DOI: 10.1007/s00127-012-0617-3
30. Silva MJS, Schraiber LB, Mota A. **The concept of health in Collective Health: contributions from social and historical critique of scientific production**. *Physis Revista de Saúde Coletiva.* (2019) **29** 102. DOI: 10.1590/s0103-73312019290102
31. Davis DE, Choe E, Meyers J, Wade N, Varjas K, Gifford A. **Thankful for the little things: a meta-analysis of gratitude interventions**. *J Couns Psychol.* (2016) **63** 20-31. DOI: 10.1037/cou0000107
32. Froh JJ, Bono G, Fan J, Emmons RA, Henderson K, Harris C. **Nice thinking! An educational intervention that teaches children to think gratefully**. *School Psych Rev.* (2014) **43** 132-52. DOI: 10.1080/02796015.2014.12087440
33. Malouff JM, Schutte NS. **Can psychological interventions increase optimism? A meta-analysis**. *J Posit Psychol.* (2017) **12** 594-604. DOI: 10.1080/17439760.2016.1221122
34. Ferrari M, Hunt C, Harrysunker A, Abbott MJ, Beath AP, Einstein DA. **Self-compassion interventions and psychosocial outcomes: a meta-analysis of RCTs**. *Mindfulness.* (2019) **10** 1455-73. DOI: 10.1007/s12671-019-01134-6
35. Zeng X, Chiu CPK, Wang R, Oei TPS, Leung FYK. **The effect of loving-kindness meditation on positive emotions: a meta-analytic review**. *Front Psychol.* (2015) **6** 1693. DOI: 10.3389/fpsyg.2015.01693
36. Curry OS, Rowland LA, Van Lissa CJ, Zlotowitz S, McAlaney J, Whitehouse H. **Happy to help? A systematic review and meta-analysis of the effects of performing acts of kindness on the well-being of the actor**. *J Exp Soc Psychol.* (2018) **76** 320-9. DOI: 10.1016/j.jesp.2018.02.014
37. Okun MA, Yeung EW, Brown S. **Volunteering by older adults and risk of mortality: a meta-analysis**. *Psychol Aging.* (2013) **28** 564-77. DOI: 10.1037/a0031519
38. Koopmans TA, Geleijnse JM, Zitman FG, Giltay EJ. **Effects of happiness on all-cause mortality during 15 years of follow-up: the arnhem elderly study**. *J Happiness Stud.* (2010) **11** 113-24. DOI: 10.1007/s10902-008-9127-0
39. Weiss LA, Westerhof GJ, Bohlmeijer ET. **Nudging socially isolated people towards well-being with the “Happiness Route”: design of a randomized controlled trial for the evaluation of a happiness-based intervention**. *Health Qual Life Outcomes.* (2013) **11** 159. DOI: 10.1186/1477-7525-11-159
40. Weiss LA, Oude Voshaar MAH, Bohlmeijer ET, Westerhof GJ. **The long and winding road to happiness: a randomized controlled trial and cost-effectiveness analysis of a positive psychology intervention for lonely people with health problems and a low socio-economic status**. *Health Qual Life Outcomes.* (2020) **18** 162. DOI: 10.1186/s12955-020-01416-x
41. Lambert D'raven LT, Moliver N, Thompson D. **Happiness intervention decreases pain and depression, boosts happiness among primary care patients**. *Prim Heal Care Res Dev.* (2015) **16** 114-26. DOI: 10.1017/S146342361300056X
42. Waldinger RJ, Cohen S, Schulz MS, Crowell JA. **Security of attachment to spouses in late life: concurrent and prospective links with cognitive and emotional wellbeing**. *Clin Psychol Sci J Assoc Psychol Sci.* (2015) **3** 516-29. DOI: 10.1177/2167702614541261
43. Doss BD, Cicila LN, Georgia EJ, Roddy MK, Nowlan KM, Benson LA. **A randomized controlled trial of the web-based OurRelationship program: effects on relationship and individual functioning**. *J Consult Clin Psychol.* (2016) **84** 285-96. DOI: 10.1037/ccp0000063
44. Halford WK, Doss BD. **New frontiers in the treatment of couples**. *Int J Cogn Ther.* (2016) **9** 124-39. DOI: 10.1521/ijct.2016.9.2.124
45. Leung P, Orrell M, Orgeta V. **Social support group interventions in people with dementia and mild cognitive impairment: a systematic review of the literature**. *Int J Geriatr Psychiatry.* (2015) **30** 1-9. DOI: 10.1002/gps.4166
46. Sebern MD, Sulemanjee N, Sebern MJ, Garnier-Villarreal M, Whitlatch CJ. **Does an intervention designed to improve self-management, social support and awareness of palliative-care address needs of persons with heart failure, family caregivers and clinicians?**. *J Clin Nurs.* (2018) **27** e643-57. DOI: 10.1111/jocn.14115
47. Wade NG, Hoyt WT, Kidwell JEM, Worthington EL. **Efficacy of psychotherapeutic interventions to promote forgiveness: a meta-analysis**. *J Consult Clin Psychol.* (2014) **82** 154-70. DOI: 10.1037/a0035268
48. Recine AC. **Designing forgiveness interventions: guidance from five meta-analyses**. *J Holist Nurs Off J Am Holist Nurses Assoc.* (2015) **33** 161-7. DOI: 10.1177/0898010114560571
49. Wade NG, Worthington JEL. **Overcoming interpersonal offenses: is forgiveness the only way to deal with unforgiveness?**. *J Couns Dev.* (2003) **81** 343-53. DOI: 10.1002/j.1556-6678.2003.tb00261.x
50. Goldman DB, Wade NG. **Comparison of forgiveness and anger-reduction group treatments: a randomized controlled trial**. *Psychother Res.* (2012) **22** 604-20. DOI: 10.1080/10503307.2012.692954
51. Baskin TW, Rhody M, Schoolmeesters S, Ellingson C. **Supporting special-needs adoptive couples: assessing an intervention to enhance forgiveness, increase marital satisfaction, and prevent depression ψ**. *Couns Psychol.* (2011) **39** 933-55. DOI: 10.1177/0011000010397554
52. Hu T, Zhang D, Wang J. **A meta-analysis of the trait resilience and mental health**. *Pers Individ Differ.* (2015) **76** 18-27. DOI: 10.1016/j.paid.2014.11.039
53. Joyce S, Shand F, Tighe J, Laurent SJ, Bryant RA, Harvey SB. **Road to resilience: a systematic review and meta-analysis of resilience training programmes and interventions**. *BMJ Open.* (2018) **8** e017858. DOI: 10.1136/bmjopen-2017-017858
54. Bennett JB Aden CA, Broome K, Mitchell K, Rigdon WD. **Team resilience for young restaurant workers: research-to-practice adaptation and assessment**. *J Occup Health Psychol.* (2010) **15** 223-36. DOI: 10.1037/a0019379
55. Steinhardt M, Dolbier C. **Evaluation of a resilience intervention to enhance coping strategies and protective factors and decrease symptomatology**. *J Am Coll Health.* (2008) **56** 445-53. DOI: 10.3200/JACH.56.44.445-454
56. Schutte NS, Malouff JM. **The impact of signature character strengths interventions: a meta-analysis**. *J Happiness Stud.* (2019) **20** 1179-96. DOI: 10.1007/s10902-018-9990-2
57. Seligman MEP, Steen TA, Park N, Peterson C. **Positive psychology progress: empirical validation of interventions**. *Am Psychol.* (2005) **60** 410-21. DOI: 10.1037/0003-066X.60.5.410
58. Goncalves JPB, Lucchetti G, Menezes PR, Vallada H. **Religious and spiritual interventions in mental health care: a systematic review and meta-analysis of randomized controlled clinical trials**. *Psychol Med.* (2015) **45** 2937-49. DOI: 10.1017/S0033291715001166
59. Moreira-Almeida A, Koenig HG, Lucchetti G. **Clinical implications of spirituality to mental health: review of evidence and practical guidelines**. *Rev Bras Psiquiatr.* (2014) **36** 176-82. DOI: 10.1590/1516-4446-2013-1255
60. Breitbart W, Poppito S, Rosenfeld B, Vickers AJ, Li Y, Abbey J. **Pilot randomized controlled trial of individual meaning-centered psychotherapy for patients with advanced cancer**. *J Clin Oncol.* (2012) **30** 1304-9. DOI: 10.1200/JCO.2011.36.2517
61. Morita T, Murata H, Kishi E, Miyashita M, Yamaguchi T, Uchitomi Y. **Meaninglessness in terminally ill cancer patients: a randomized controlled study**. *J Pain Symptom Manag.* (2009) **37** 649-58. DOI: 10.1016/j.jpainsymman.2008.04.017
62. Hopkins E, Kelley R, Bentley K. *Working With Groups on Spiritual Themes: Structured Exercises in Healing* (1995) p. 148
63. Czekierda K, Banik A, Park CL, Luszczynska A. **Meaning in life and physical health: systematic review and meta-analysis**. *Health Psychol Rev.* (2017) **11** 387-418. DOI: 10.1080/17437199.2017.1327325
64. Li J-B, Dou K, Liang Y. **The relationship between presence of meaning, search for meaning, and subjective well-being: a three-level meta-analysis based on the meaning in life questionnaire**. *J Happiness Stud.* (2021) **22** 467-89. DOI: 10.1007/s10902-020-00230-y
65. Wong P. **From logotherapy to meaning-centered counseling and therapy**. *The The Human Quest for Meaning: Theories Research and Applications* (2012) p. 619-47
66. Luz JMO, Murta SG, Aquino TAA. **Avaliação de Resultados e Processo de uma Intervenção para Promoção de Sentido da Vida em Adolescentes**. *Trends Psychol.* (2017) **25** 1795-811. DOI: 10.9788/TP2017.4-14Pt
67. Norriss H. **Flourishing, positive mental health and well-being: how can they be increased?**. *Int J Leadersh Public Serv.* (2010) **6** 46-50. DOI: 10.5042/ijlps.2010.0638
68. W?ziak-Białowolska D, McNeely E, VanderWeele TJ. **Human flourishing in cross cultural settings. Evidence from the United States, China, Sri Lanka, Cambodia, and Mexico**. *Front Psychol.* (2019) **10** 1269. DOI: 10.3389/fpsyg.2019.01269
69. Kim MJ, Sung E, Choi EY, Ju Y-S, Park E-W, Cheong Y-S. **Delphi survey for designing a intervention research study on childhood obesity prevention**. *Korean J Fam Med.* (2017) **38** 284-90. DOI: 10.4082/kjfm.2017.38.5.284
70. Andersson G. **Internet interventions: past, present and future**. *Internet Interv.* (2018) **12** 181-8. DOI: 10.1016/j.invent.2018.03.008
71. Puchalski CM, Vitillo R, Hull SK, Reller N. **Improving the spiritual dimension of whole person care: reaching national and international consensus**. *J Palliat Med.* (2014) **17** 642-56. DOI: 10.1089/jpm.2014.9427
72. Schröder J, Berger T, Westermann S, Klein JP, Moritz S. **Internet interventions for depression: new developments**. *Dialog Clin Neurosci.* (2016) **18** 203-12. DOI: 10.31887/DCNS.2016.18.2/jschroeder
73. Karyotaki E, Riper H, Twisk J, Hoogendoorn A, Kleiboer A, Mira A. **Efficacy of self-guided internet-based cognitive behavioral therapy in the treatment of depressive symptoms: a meta-analysis of individual participant data**. *JAMA Psychiatry.* (2017) **74** 351-9. DOI: 10.1001/jamapsychiatry.2017.0044
74. Klein JP, Berger T, Schröder J, Späth C, Meyer B, Caspar F. **Effects of a psychological internet intervention in the treatment of mild to moderate depressive symptoms: results of the EVIDENT study, a randomized controlled trial**. *Psychother Psychosom.* (2016) **85** 218-28. DOI: 10.1159/000445355
75. Christensen H, Griffiths KM, Jorm AF. **Delivering interventions for depression by using the internet: randomised controlled trial**. *BMJ.* (2004) **328** 265. DOI: 10.1136/bmj.37945.566632.EE
76. Palmqvist B, Carlbring P, Andersson G. **Internet-delivered treatments with or without therapist input: does the therapist factor have implications for efficacy and cost?**. *Expert Rev Pharmacoecon Outcomes Res.* (2007) **7** 291-7. DOI: 10.1586/14737167.7.3.291
77. Kleiboer A, Donker T, Seekles W, van Straten A, Riper H, Cuijpers P. **randomized controlled trial on the role of support in Internet-based problem solving therapy for depression and anxiety**. *Behav Res Ther.* (2015) **72** 63-71. DOI: 10.1016/j.brat.2015.06.013
78. Hedman E, El Alaoui S, Lindefors N, Andersson E, Rück C, Ghaderi A. **Clinical effectiveness and cost-effectiveness of Internet- vs. group-based cognitive behavior therapy for social anxiety disorder: 4-year follow-up of a randomized trial**. *Behav Res Ther.* (2014) **59** 20-9. DOI: 10.1016/j.brat.2014.05.010
79. Fawcett E, Neary M, Ginsburg R, Cornish P. **Comparing the effectiveness of individual and group therapy for students with symptoms of anxiety and depression: a randomized pilot study**. *J Am Coll Health.* (2020) **68** 430-7. DOI: 10.1080/07448481.2019.1577862
80. Forsetlund L, Bjørndal A, Rashidian A, Jamtvedt G, O'Brien MA, Wolf F. **Continuing education meetings and workshops: effects on professional practice and health care outcomes**. *Cochrane Database Syst Rev.* (2009) **2009** CD003030. DOI: 10.1002/14651858.CD003030.pub2
|
---
title: 'The potential antidepressant effect of antidiabetic agents: New insights from
a pharmacovigilance study based on data from the reporting system databases FAERS
and VigiBase'
authors:
- Vera Battini
- Robbert P. Van Manen
- Michele Gringeri
- Giulia Mosini
- Greta Guarnieri
- Anna Bombelli
- Marco Pozzi
- Maria Nobile
- Sonia Radice
- Emilio Clementi
- Carla Carnovale
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9981969
doi: 10.3389/fphar.2023.1128387
license: CC BY 4.0
---
# The potential antidepressant effect of antidiabetic agents: New insights from a pharmacovigilance study based on data from the reporting system databases FAERS and VigiBase
## Abstract
Background: Growing evidence supports a bidirectional association between diabetes and depression; promising but limited and conflicting data from human studies support the intriguing possibility that antidiabetic agents may be used to relieve effectively depressive symptoms in diabetic patients. We investigated the potential antidepressant effects of antidiabetic drugs in a high-scale population data from the two most important pharmacovigilance databases, i.e., the FDA Adverse Event Reporting System (FAERS) and the VigiBase.
Material and methods: From the two primary cohorts of patients treated with antidepressants retrieved from FDA Adverse Event Reporting System and VigiBase we identified cases (depressed patients experiencing therapy failure) and non-cases (depressed patients experiencing any other adverse event). We then calculated the Reporting Odds Ratio (ROR), Proportional Reporting Ratio (PRR), Empirical Bayes Geometric Mean (EBGM), and Empirical Bayes Regression-Adjusted Mean (ERAM) for cases versus non-cases in relation with the concurrent exposure to at least one of the following antidiabetic agent: A10BA Biguanides; A10BB Sulfonylureas; A10BG Thiazolidinediones; A10BH DPP4-inhibitors; A10BJ GLP-1 analogues; A10BK SGLT2 inhibitors (i.e., those agents for which preliminary evidence from literature supports our pharmacological hypothesis).
Results: For GLP-1 analogues, all the disproportionality scores showed values <1, i.e., statistically significant, in both analyses [from the FAERS: ROR confidence interval of 0.546 (0.450–0.662); PRR (p-value) of 0.596 (0.000); EBGM (CI) of 0.488 (0.407–0.582); ERAM (CI) of 0.480 (0.398–0.569) and VigiBase: ROR (CI) of 0.717 (0.559–0.921); PRR (p-value) of 0.745 (0.033); EBGM (CI) of 0.586 (0.464–0.733); ERAM of (CI): 0.515 (0.403–0.639)]. Alongside GLP-1 analogues, DPP-4 Inhibitors and Sulfonylureas showed the greatest potential protective effect. With regard to specific antidiabetic agents, liraglutide and gliclazide were associated with a statistically significant decrease in all disproportionality scores, in both analyses.
Conclusion: The findings of this study provide encouraging results, albeit preliminary, supporting the need for further clinical research for investigating repurposing of antidiabetic drugs for neuropsychiatric disorders.
## 1 Introduction
Depression, estimated by the World Health Organization (WHO) as the single largest contributor to global disability, is a major challenge for the national health systems. Its co-occurrence with Type 2 Diabetes (T2D) is twice as frequent as might be predicted by chance alone and results in a reduced quality of life and elevated impairment of individuals’ daily functioning (Holt et al., 2014).
A growing number of evidence supports a bidirectional association between diabetes and depression as a result of complex interactions involving brain events and systemic responses (Golden et al., 2008; Laake et al., 2014; Martins et al., 2022). The role of the inflammatory cascade in the induction of metabolic syndrome, oxidative stress and central diseases promoted studies on the identification of novel pharmacological targets for a combined treatment (Lamb and Goldstein, 2008; Chan et al., 2019). The central activation of AMPK, a key enzyme regulating both energy management and psychopathology, which is also supported by some antidiabetic drugs, has been suggested as a useful strategy to relieve both depressive and diabetic symptoms (Pozzi et al., 2019).
Evidence from experimental studies has also reported that traditional anti-hyperglycaemic agents, such as insulin, glyburide, metformin, pioglitazone, vildagliptin, and liraglutide reduce depression-like behaviour in either absence or presence of diabetes (AlHussain et al., 2020; Essmat et al., 2020).
Promising yet still limited clinical evidence from human studies is also available: a recent metanalysis of 9 studies found that GLP1 receptor agonists can relieve depressive symptoms in adult patients affected by T2D (Pozzi et al., 2019) and another study shows that thiazolidinediones might be associated to pharmacologically relevant antidepressant actions (Moulton et al., 2018). However, these concepts still need to be expanded (Odaira et al., 2019).
Antidiabetic agents including metformin, thiazolidinediones GLP-1 agonists and dipeptidyl peptidase 4 (DPP-4) inhibitors are known to cross the blood brain barrier and thus exert both peripheral and central actions. The antidepressant activity of these drugs may be mediated by reducing the blood glucose level, ameliorating the central oxidative stress and inflammation, regulating the hypothalamic–pituitary–adrenal axis, stimulating neuronal growth and protecting from apoptosis through the protein Gs-Protein Kinase A-mediated activation of AMPK (Essmat et al., 2020). The underlying mechanism of action has not been fully elucidated yet.
Spontaneous reporting systems such as the FDA Adverse Event Reporting System (FAERS) and VigiBase represent valuable sources to obtain real-world data on the safety/effectiveness profile of specific drugs, in order to compare therapeutic options, gain insights on potential mechanisms of adverse drug reaction (ADR) and (more recently) investigate promising new beneficial effects of drugs, thus contributing to drug repositioning (Cohen et al., 2017; Carnovale et al., 2018; 2019b; 2019a; Mazhar et al., 2019). Due to the insufficient therapeutic response of patients to the available antidepressant medications, drug repositioning may become the most promising strategy to support new indication uses.
Here we report on the antidepressant effect of antidiabetic agents in a high-scale population data from the two largest spontaneous reporting system databases, i.e., the FAERS and VigiBase, thus providing new insights in support of their potential drug repurposing in the field of neuropsychiatric disorders.
## 2.1 Data source and extraction
This study was designed as a nested case/non-case study. We used the Empirica Signal software (Oracle Health Sciences, Austin, TX) to query the two largest and most comprehensive spontaneous reporting system public databases: the FDA FAERS database (from 1967 up until the end of 2021) and the WHO VigiBase database (from 1968 until the end of September 2021).
Both data sources contain information related to post-marketing safety surveillance reports in the form of Individual Case Safety Reports (ICSRs) submitted by healthcare professionals, consumers, and other sources. Adverse events (AEs) are coded in these two pharmacovigilance databases using the Medical Dictionary for Regulatory Activities (MedDRA®) Preferred Terms (PTs) (Fescharek et al., 2004). Each ICSR provides administrative information (country, type of report, qualification of the reporter), patient demographics (sex, age, weight), AEs characteristics (seriousness, date of onset, outcome), details about suspect drug therapy (drug name, exposure start and stop dates, time to onset, dose, route, indication, de-challenge and re-challenge) and information concerning any drug administered at the time of AE but not held responsible for its occurrence by the reporter, referred to as concomitant medication. However, the level of completeness of information varies from case to case (Sakaeda et al., 2013).
Both databases were prepared for data mining, for example by combining initial and follow-up reports into a single case and eliminating obvious duplicate cases using an automated process provided by Oracle.
A primary cohort of ICSRs was defined as all reports mentioning at least one antidepressant drug (ATC Level 3 code N06A “Antidepressants”) as “suspect drug” (either primary or secondary) for any type of AE. Reports containing antidepressants as “concomitant medication” only were not included in the primary cohort.
Within this primary cohort, cases were defined as depressed patients experiencing therapy failure and non-cases as patients experiencing any other AEs. Therapy failure was defined as ICSRs mentioning either the MedDRA narrow SMQ Depression and suicide/self-injury or the narrow SMQ Lack of efficacy (MedDRA version 24.0). ( MedDRA-Support Documentation, 2022).
## 2.2 Statistical analysis
By using the Oracle Empirica Signal software (Oracle Health Sciences, Austin, TX), we calculated disproportionality statistics produced by four signal detection methodologies, to assess the occurrence of therapy failure (cases) in depressed patients, in association with the exposure to at least one antidiabetic drug, defined as the following ATC Level 4 codes: A10BA Biguanides; A10BB Sulfonylureas; A10BG Thiazolidinediones; A10BH DPP4-inhibitors; A10BJ GLP-1 analogues; A10BK SGLT2 inhibitors (i.e., those agents for which preliminary evidence from literature supports our pharmacological hypothesis).
Three of these disproportionality scores, based on 2 × 2 disproportionality analysis, are well-established and currently used worldwide by several organisations for routine safety surveillance, i.e:i) The Reporting Odds Ratio (ROR), defined as the ratio of the odds of the occurrence of therapy failure with antidiabetic drugs versus the occurrence of therapy failure without antidiabetic agents (van Manen et al., 2007);ii) The Proportional Reporting Ratio (PRR), comparing the frequency of occurrence of therapy failure in reports referring to antidiabetic agents with the frequency of occurrence of reports of therapy failure in reports that do not mention antidiabetic agents. ( van Manen et al., 2007).iii) The Empirical Bayesian Geometric Mean (EBGM) calculated using the Multi-item Gamma Poisson Shrinker (MGPS) Algorithm, using Bayesian shrinkage to improve the reliability of the disproportionality score (DuMouchel, 1999). *We* generated both the point estimates (EBGM) and their associated $90\%$ confidence intervals labelled EB05–EB95.
Moreover, we used a more advanced regression-based methodology designed to produce disproportionality statistics with adjusted background rates; it can control masking and more extensive confounding effects by fitting separate Bayesian logistic regression models to each target AE and by automatically selecting predictors to be included in each regression model:iv) The Regression-enhanced Empirical Bayesian Geometric Mean (ERAM) calculated using the Regression-Adjusted Gamma Poisson Shrinker (RGPS) Algorithm (DuMouchel and Harpaz, 2012). *We* generated the point estimates (ERAM) and their associated $90\%$ confidentiality intervals labelled ER05–ER95.
With the aim to investigate the antidepressant effects of antidiabetic drugs, disproportionality signals were considered clinically meaningful if.i) The upper limit of the $90\%$ confidence interval (CI) of the ROR for cases (ROR95) is less than one;ii) The PRR score is less than one and the corresponding p-value is less than 0.05;iii) The upper limit of the $90\%$ confidence interval of the EBGM for cases (EB95) is less than one;iv) The upper limit of the $90\%$ confidence interval of the ERAM for cases (ER95) is less than one.
## 3 Results
During the time periods described in the methods, we selected two primary cohorts of ICSRs mentioning antidepressants as “suspect drug” (either primary or secondary) for any AEs reported in the FAERS and VigiBase, which contain 545,311 and 647,308 ICSRs, respectively. Within these primary cohorts we selected 121,368 ICSRs from FAERS and 85,267 from VigiBase as cases associated with “therapy failure”; the numbers of non-cases for FAERS and VigiBase were 423,943 and 562,041, respectively. Figure 1 shows the flow diagram of data extraction from the two data sources.
**FIGURE 1:** *Flow-diagram of data extraction from VigiBase and FAERS.*
Demographical characteristics and type of therapy of depressed patients experiencing therapy failure (cases) and other adverse events (non-cases) from FAERS and VigiBase are detailed in Tables 1, 2. For cases, the most involved age groups reported in the FAERS and VigiBase were 18–44 and 45–74, respectively. In both analyses, >$62\%$ of cases reported antidepressants as the only suspected drugs and no other drugs. For non-cases, the percentage ranged from $41.4\%$ (FAERS) to $80.2\%$ (VigiBase).
Supplementary Tables S1, S2 list the number of medications reported as suspect (either primary or secondary suspect) drugs, grouping by ATC Level 2, for cases and non-cases, in the FAERS and Vigibase.
In both cohorts of depressed patients (cases and non-cases), more than $58\%$ of individuals were female.
Of depressed subjects experiencing therapy failure, 1,946 and 649 were concomitantly exposed to only one antidiabetic drug, in the FAERS and VigiBase, respectively; in both cohorts, less than $1\%$ was treated with more than one antidiabetic drug and less than $1\%$ was concomitantly exposed to insulin.
Four disproportionality scores (ROR, PRR, EBGM, ERAM) were used to investigate the potential antidepressant effect of antidiabetic drugs. Table 3 shows values for therapy failure in depressed patients exposed to various antidiabetic drug classes.
**TABLE 3**
| Data source | Antidiabetic drug class | ROR (ROR05-ROR95) | PRR (p-value) | EBGM (EB05-EB95) | ERAM (ER05-ER95) |
| --- | --- | --- | --- | --- | --- |
| FAERS | Biguanides | 1.085 (1.038–1.135) | 1.068 (0.003) | 0.919 (0.884–0.956) | 0.856 (0.823–0.890) |
| FAERS | Sulfonylureas | 0.831 (0.776–0.890) | 0.858 (0.000) | 0.935 (0.878–0.995) | 0.858 (0.805–0.912) |
| FAERS | Thiazolidinediones | 0.925 (0.810–1.056) | 0.938 (0.353) | 0.919 (0.815–1.034) | 0.818 (0.723–0.918) |
| FAERS | DPP4 Inhibitors | 0.761 (0.674–0.860) | 0.796 (0.000) | 0.687 (0.614–0.767) | 0.676 (0.602–0.753) |
| FAERS | GLP1 Analogues | 0.546 (0.450–0.662) | 0.596 (0.000) | 0.488 (0.407–0.582) | 0.480 (0.398–0.569) |
| FAERS | SGLT2 Inhibitors | 0.901 (0.702–1.158) | 0.918 (0.543) | 0.716 (0.571–0.890) | 0.715 (0.564–0.881) |
Among all the drug classes of interest, GLP-1 analogues, DPP-4 Inhibitors and Sulfonylureas showed the greatest potential protective effects. Specifically, all signal detection methodologies and disproportionality statistics investigating the GLP-1 analogues agreed on its potential antidepressant effect and showed values <1, i.e., statistically significant [from the FAERS: ROR (CI) of 0.546 (0.450–0.662); PRR (p-value) of 0.596 (0.000); EBGM (CI) of 0.488 (0.407–0.582); ERAM (CI) of 0.480 (0.398–0.569) and VigiBase: ROR (CI) of 0.717 (0.559–0.921); PRR (p-value) of 0.745 (0.033); EBGM (CI) of 0.586 (0.464–0.733); ERAM of (CI): 0.515 (0.403–0.639)].
On the other hand, only disproportionality signals in FAERS were considered statistically meaningful for DPP-4 Inhibitors [ROR (CI) of 0.761 (0.674–0.860); PRR (p-value) of 0.796 (0.000); EBGM (CI) of 0.687 (0.614–0.767); ERAM (CI) of 0.676 (0.602–0.753)] and Sulfonylureas [ROR (CI) of 0.831 (0.776–0.890); PRR (p-value) of 0.858 (0.000); EBGM (CI) of 0.935 (0.878–0.995); ERAM (CI) of 0.858 (0.805–0.912)].
Biguanides, SGLT2 Inhibitors and Thiazolidinediones showed the smallest protective effect. For biguanides we found statistically significant scores only for ERAM in both analyses [FAERS: ERAM (CI) of 0.856 (0.823–0.890); VigiBase: ERAM (CI) of 0.822 (0.784–0.860)]. Similar findings were found for SGLT2 Inhibitors: only the EBGM values were significant in both analyses [FAERS: EBGM (CI) of 0.716 (0.571–0.890)]; VigiBase: EBGM (CI) of 0.994 (0.755–1.289)]. For thiazolidinediones only ERAM from FAERS was statistically significant: ERAM (CI) of 0.818 (0.723–0.918).
In Table 4, detailed disproportionality scores for each antidiabetic drug from both FAERS and VigiBase are reported. With regard to some selected antidiabetic agents, liraglutide and gliclazide were associated in both analyses to a statistically significant decrease in all disproportionality scores. More in detail, in the FAERS analysis, for liraglutide we found the following scores: ROR (CI) of 0.580 (0.438–0.768); PRR (p-value) of 0.629 (0.002); EBGM (CI) of 0.534 (0.411–0.683); ERAM (CI) of 0.529 (0.403–0.670)]; consistent results were found for gliclazide: ROR (CI) of 0.527 (0.443–0.628); PRR (p-value) of 0.578 (0.000); EBGM (CI) of 0.556 (0.471–0.653); ERAM (CI) of 0.552 (0.465–0.645)]. Findings from VigiBase, considering a larger cohort of patients, supported the previous results; for liraglutide we found: ROR (CI) of 0.519 (0.343–0.785); PRR (p-value) of 0.554 (0.011); EBGM (CI) of 0.472 (0.324–0.668); ERAM (CI) of 0.414 (0.275–0.577); for gliclazide we found: ROR (CI) of 0.310 (0.238–0.405); PRR (p-value) of 0.341 (0.000); EBGM (CI) of 0.439 (0.340–0.559); ERAM (CI) of 0.438 (0.334–0.553). Supplemental material provides disproportionality scores for cases and non-cases exposed to antidiabetic drug classes grouping by ATC code level 4 (including details for each antidiabetic drug), in the FAERS (Supplementary Table S3) and VigiBase (Supplementary Table S4).
**TABLE 4**
| Drug | ATC code | FAERS (Total cohort: 545,311 depressed subjects) | FAERS (Total cohort: 545,311 depressed subjects).1 | FAERS (Total cohort: 545,311 depressed subjects).2 | FAERS (Total cohort: 545,311 depressed subjects).3 | VigiBase (Total cohort: 647,308 depressed subjects) | VigiBase (Total cohort: 647,308 depressed subjects).1 | VigiBase (Total cohort: 647,308 depressed subjects).2 | VigiBase (Total cohort: 647,308 depressed subjects).3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Drug | ATC code | ROR (ROR05-ROR95) | PRR (p-value) | EBGM (EB05-EB95) | ERAM (ER05-ER95) | ROR (ROR05-ROR95) | PRR (p-value) | EBGM (EB05-EB95) | ERAM (ER05-ER95) |
| Canagliflozin | SGLT2-inhibitors | 1.268 (0.826–1.945) | 1.208 (0.439) | 0.952 (0.662–1.335) | 0.894 (0.605–1.229) | 2.797 (1.655–4.726) | 2.262 (0.002) | 1.466 (0.977–2.131) | 1.169 (0.746–1.670) |
| Chlorpropamide | Sulfonylureas | 0.253 (0.119–0.539) | 0.294 (0.002) | 0.819 (0.441–1.416) | 0.767 (0.382–1.260) | 0.092 (0.017–0.480) | 0.104 (0.005) | 0.888 (0.393–1.784) | 0.715 (0.247–1.380) |
| Dapagliflozin | SGLT2-inhibitors | 0.543 (0.304–0.971) | 0.593 (0.106) | 0.565 (0.343–0.888) | 0.573 (0.331–0.870) | 0.791 (0.389–1.610) | 0.813 (0.729) | 0.714 (0.410–1.175) | 0.659 (0.345–1.055) |
| Dulaglutide | GLP-1 analogues | 0.373 (0.217–0.641) | 0.422 (0.003) | 0.375 (0.232–0.580) | 0.382 (0.226–0.571) | 0.722 (0.391–1.333) | 0.750 (0.476) | 0.668 (0.403–1.053) | 0.629 (0.354–0.969) |
| Empagliflozin | SGLT2-inhibitors | 0.911 (0.595–1.395) | 0.926 (0.815) | 0.721 (0.496–1.018) | 0.723 (0.485–1.001 | 1.279 (0.777–2.106) | 1.234 (0.517) | 0.966 (0.636–1.419) | 0.945 (0.593–1.363) |
| Exenatide | GLP-1 analogues | 0.791 (0.571–1.096) | 0.823 (0.274) | 0.711 (0.529–0.938) | 0.649 (0.475–0.845) | 1.106 (0.770–1.590) | 1.091 (0.731) | 0.864 (0.626–1.169) | 0.722 (0.510–0.964) |
| Gliclazide | Sulfonylureas | 0.527 (0.443–0.628) | 0.578 (0.000) | 0.556 (0.471–0.653) | 0.552 (0.465–0.645) | 0.310 (0.238–0.405) | 0.341 (0.000) | 0.439 (0.340–0.559) | 0.438 (0.334–0.553) |
| Glimepiride | Sulfonylureas | 0.937 (0.808–1.087) | 0.948 (0.500) | 0.918 (0.802–1.047) | 0.853 (0.742–0.970) | 0.984 (0.833-1–162) | 0.986 (0.912) | 0.969 (0.830–1.126) | 0.851 (0.726–0.985) |
| Glipizide | Sulfonylureas | 1.123 (0.987–1.277) | 1.098 (0.149) | 1.186 (1.057–1.327) | 1.043 (0.927–1.164) | 1.502 (1.304–1.728) | 1.409 (0.000) | 1.431 (1.260–1.619) | 1.087 (0.954–1.228) |
| Gliquidone | Sulfonylureas | 0.440 (0.078–2.472) | 0.491 (0.677) | 0.864 (0.352–1.844) | 0.806 (0.275–1.563) | 0.942 (0.162–5.466) | 0.949 (0.641) | 1.185 (0.527–2.375) | 1.026 (0.355–1.980) |
| Linagliptin | Dpp-4 Inhibitors | 0.669 (0.446–1.003) | 0.713 (0.125) | 0.651 (0.452–0.913) | 0.654 (0.443–0.900) | 0.642 (0.382–1.079) | 0.673 (0.199) | 0.720 (0.460–1.084) | 0.645 (0.390–0.951) |
| Liraglutide | GLP-1 analogues | 0.580 (0.438–0.768) | 0.629 (0.002) | 0.534 (0.411–0.683) | 0.529 (0.403–0.670) | 0.519 (0.343–0.785) | 0.554 (0.011) | 0.472 (0.324–0.668) | 0.414 (0.275–0.577) |
| Lixisenatide | GLP-1 analogues | 0.629 (0.108–3.651) | 0.675 (0.986) | 0.871 (0.355–1.858) | 0.883 (0.302–1.712) | 2.197 (0.329–14.683) | 1.898 (0.968) | 1.160 (0.516–2.326) | 1.089 (0.377–2.102) |
| Metformin | Biguanides | 1.107 (1.057–1.159) | 1.085 (0.000) | 0.933 (0.896–0.972) | 0.866 (0.831–0.901) | 1.420 (1.349–1.494) | 1.345 (0.000) | 1.166 (1.113–1.220) | 1.200 (1.145–1.255) |
| Pioglitazone | Thiazolidinediones | 1.030 (0.869–1.220) | 1.024 (0.816) | 0.934 (0.802–1.084) | 0.821 (0.701–0.948) | 1.382 (1.138–1.678) | 1.316 (0.007) | 1.121 (0.940–1.328) | 0.883 (0.736–1.042) |
| Rosiglitazone | Thiazolidinediones | 0.909 (0.717–1.152) | 0.924 (0.553) | 0.923 (0.744–1.134) | 0.815 (0.651–0.993) | 1.378 (1.070–1.775) | 1,313 (0.045) | 1.187 (0.946–1.475) | 0.895 (0.705–1.104) |
| Saxagliptin | Dpp-4 Inhibitors | 0.372 (0.185–0.749) | 0.421 (0.023) | 0.485 (0.271–0.815) | 0.488 (0.254–0.782) | 0.638 (0.316–1.289) | 0.670 (0.378) | 0.706 (0.405–1.161) | 0.643 (0.336–1.029) |
| Semaglutide | GLP-1 analogues | 0.267 (0.114–0.622) | 0.309 (0.009) | 0.349 (0.178–0.626) | 0.330 (0.155–0.558) | - | - | - | - |
| Sitagliptin | Dpp-4 Inhibitors | 0.968 (0.830–1.130) | 0.974 (0.767) | 0.840 (0.730–0.962) | 0.809 (0.701–0.924) | 1.119 (0.929–1.347) | 1.102 (0.348) | 0.991 (0.836–1.169) | 0.863 (0.723–1.014) |
| Tolazamide | Sulfonylureas | 0.315 (0.057–1.726) | 0.360 (0.396) | 0.986 (0.403–2.103) | 0.877 (0.300–1.701) | 0.824 (0.144–4.717) | 0.844 (0.757) | 1.239 (0.551–2.484) | 1.073 (0.371–2.071) |
| Troglitazone | Thiazolidinediones | 0.318 (0.159–0.638) | 0.364 (0.006) | 0.808 (0.452–1.357) | 0.717 (0.374–1.151) | 0.507 (0.189–1.359) | 0.542 (0.354) | 1.087 (0.555–1.962) | 0.985 (0.432–1.721) |
| Vildagliptin | Dpp-4 Inhibitors | 0.304 (0.152–0.608) | 0.349 (0.004) | 0.421 (0.235–0.707) | 0.458 (0.239–0.734) | 0.449 (0.224–0.900) | 0.485 (0.073) | 0.631 (0.362–1.038) | 0.676 (0.354–1.082) |
## 4 Discussion
Studies on glucose-lowering agents may have a positive influence on the symptoms of depression, although the evidence from animal and human studies is scarce and conflicting (Monnier et al., 2006; Ceriello et al., 2013; Fiorentino et al., 2013).
This is the first study aimed at evaluating the potential antidepressant effect of antidiabetic agents in a high-scale population data (we included two cohorts of 121,368 and 85,267 depressed patients experiencing therapy failure) from the two largest spontaneous reporting system databases, i.e., the FAERS and VigiBase, thus providing new insights for improving the knowledge on this topic and supporting the need for further research on antidiabetic drug repurposing in the field of neuropsychiatric disorders.
It is well-known that pharmacovigilance databases were originally intended to track frequent adverse events; however, when a sufficient amount of data is available, they can also be used to indirectly track the beneficial outcomes through monitoring reductions of related adverse event frequencies (Cohen et al., 2017).
From this perspective, reported adverse drug events may serve as useful indicators to predict new opportunities for drug repositioning, making spontaneous reporting system databases valuable data sources for driving further research in the discover of new and effective uses of drugs (Pushpakom et al., 2019).
Overall, the investigated antidiabetic drug classes showed a beneficial effect to depressed patients, albeit with a high heterogeneity in terms of statistically significant decrease in disproportionality scores, thus suggesting that some specific pharmacological agents may exert a more prominent beneficial effect.
In our study, GLP-1 analogues showed the greatest potential protective effect in the cohort of depressed patients experiencing therapy failure that we analysed. Of importance, all signal-detection methodologies and disproportionality statistics we used to investigate the antidepressant effect of GLP-1 analogues showed values statistically significant (<1) in both pharmacovigilance databases (Table 3), with a ROR ranging from 0.546 (0.450–0.662) to 0.717 (0.559–0.921), in the FAERS and VigiBase, respectively. ROR is the most used disproportionality score worldwide for routine safety surveillance.
However, more recently, many Authors applied this approach to the FAERS and VigiBase to identify candidates for drug repositioning in a variety of clinical research areas (e.g., psychiatry, neurology, cardiology), by searching for an inverse signal, postulating that drugs that demonstrated an under-reporting of AEs of interest could be protective against these AEs (Wang et al., 2016; Horinouchi et al., 2018; Hosomi et al., 2018; Chrétien et al., 2021).
To test our hypothesis, we have expanded this approach further by also providing also other well-established scores based on 2 × 2 disproportionality analysis (PRR and EBGM), and a more advanced regression-based methodology designed to produce disproportionality statistics with adjusted background rates: it can control for masking and more extensive confounding effects by fitting separate Bayesian logistic regression models to each target AE and by automatically selecting predictors to be included in each regression model, i.e., ERAM (DuMouchel and Harpaz, 2012).
In both pharmacovigilance databases, ERAM values suggest GLP-1 analogues may exert a clinical meaningful protective effect, as demonstrated by significant reductions of depression-like symptom frequencies in patients with depression and diabetes [point estimates: 0.480 (0.398–0.569) in FAERS and 0.515 (0.403–0.639), in VigiBase].
When focusing on specific drugs, liraglutide was associated with a statistically significant decrease in all disproportionality scores. Data from the FAERS-based study [ROR (CI) of 0.580 (0.438–0.768); PRR (p-value) of 0.629 (0.002); EBGM (CI) of 0.534 (0.411–0.683); ERAM (CI) of 0.529 (0.403–0.670)], support the hypothesis that this antidiabetic agent might exert beneficial effects to depressed patients. Interestingly, when investigating the potential protective effect of liraglutide in a larger cohort of patients, findings from VigiBase strongly supported the previous results [ROR (CI) of 0.519 (0.343–0.785); PRR (p-value) of 0.554 (0.011); EBGM (CI) of 0.472 (0.324–0.668); ERAM (CI) of 0.414 (0.275–0.577)].
In line with our findings, clinical and preclinical studies, albeit very scant, support these encouraging results.
Clinical trials have demonstrated improvements in anhaedonia in patients treated with liraglutide (Mansur et al., 2017). The administration of this drug in diabetic mice has demonstrated neuroprotective (Porter et al., 2012; Li et al., 2015; Gumuslu et al., 2016), anxiolytic and anti-depressant effects in a Type 1 Diabetes (T1D) rat model. The drug was also found to increase neurogenesis in the mouse brain (Hunter and Hölscher, 2012) and to enhance effects on synaptic plasticity (McClean et al., 2010).
It has been postulated that, incretins might exert neuropsychiatric effects given the presence of GLP-1 receptors in the central nervous system; stimulation of GLP-1 receptors has shown effects on mitochondrial functions, neuroinflammation, synaptic plasticity, learning and memory, serotonin turnover, serotonin-receptor expression in the amygdala and central dopamine levels, in multiple experimental models of both neurological diseases and depression (athauda and Foltynie, 2016; Athauda and Foltynie, 2016; Kim et al., 2020; van Bloemendaal et al., 2014).
Our study shows that also DPP-4 Inhibitors show potential anti-depressant activity, as supported by all significant values of the disproportionality scores from the FAERS; however, when expanding analysis in a larger cohort of patients (i.e., in VigiBase), only EBGM and ERAM scores were of meaningful clinical relevance, suggesting an important but less prominent antidepressant effect compared to GLP-1 analogues. Within this latter drug class, saxagliptin and vildagliptin showed a significant reductions of depression-like symptom frequencies in patients with depression and diabetes, in all analyses we carried out in the FAERS (i.e., values from all disproportionality scores are significant) (see Table 4), adding preliminary and encouraging evidence (but less promising than those reported for GLP-1 analogues) to the very limited existing body of knowledge on the potential use of DPP-4 Inhibitors as an adjuvant in the treatment of depression. Recent data show that sitagliptin has mild anti-depressant effect in a depression model (Kamble et al., 2016). and a better antidepressant activity than imipramine (Saritha and Chandrashekar, 2018). However, a recent randomised controlled trial (RCT) did not detect evidence of superiority of sitagliptin over placebo for depressive symptoms in 44 patients with T2D, possibly due to the small sample size and limited treatment duration (Moulton et al., 2021). To address the issue further, an on-going randomised double-blind trial including 80 adult outpatients with major depression is evaluating the antidepressant effects of vildagliptin 50 mg versus escitalopram 20 mg (ClinicalTrial.gov, 2022).
In line with the overall picture regarding DPP-4 Inhibitors, sulfonylureas showed a similar potential: All values of the disproportionality scores from the FAERS and 3 out of 4 from VigiBase were of significant importance. These preliminary results may serve as indicators for supporting further research to better investigate their beneficial effects to depressed patients as the currently available evidence is scant and relatively conflicting. A recent experimental study showed that the glyburide exerts an effect on modulating depressive like-behaviour together with insulin resistance via an NLRP3-inflammasome inhibition (Su et al., 2017). Indeed, NLRP3 may be involved in the pathophysiology of depression (Alcocer-Gómez et al., 2014; 2016), supporting its role as promising therapeutic target for depression.
A population-based cohort study found that sulfonylureas in combination with metformin decrease the risk of affective disorders in T2D patients (Wahlqvist et al., 2012). In contrast, high doses of sulfonylureas were associated with higher risk of depression in a recent population-based cohort and nested case-control study (Wium-Andersen et al., 2022).
We found that biguanides, SGLT2 inhibitors and thiazolidinediones are associated to antidepressant beneficial effects, albeit the entity of this effect is not statistically significant for all disproportionality scores, neither in the FAERS nor in VigiBase.
Among the above-mentioned drug classes, metformin is one of the most investigated antidiabetic drugs as potential adjuvant therapy in depressed patients. Empirical insights showed that it ameliorates stress-induced depression-like behaviours through the enhancement of BDNF expression via AMPK/CREB-mediated histone acetylation (Fang et al., 2020) and it has been shown to elicit marked anti-inflammatory, antioxidant, and neuroprotective activities and to improve memory and learning functions in rats (Pintana et al., 2012; Shivavedi et al., 2017).
Recently, in a case–control study, metformin was a protective factor against depression in elderly diabetic patients, as suggested by the adjusted OR of 0.567 ($95\%$ CI: 0.323–0.997; $p \leq 0.05$) (Chen et al., 2019). In older men with T2D and high frailty risk, metformin was associated with a $15.6\%$ decrease in depression (Wang et al., 2017). In our FAERS analysis, among biguanides, metformin was associated with the lowest occurrence of depression-like symptoms compared to non-users of this medication, as confirmed by the two statistically significant disproportionality scores EBGM [0.933 (0.896–0.972)] and ERAM [0.866 (0.831–0.901)], based on Bayesian statistical methods and regression-based methodology, respectively.
As a consequence of the high heterogeneity of previous studies in terms of methodological approaches, it is not possible to directly compare data from different scores; however, our findings support all these previous encouraging results. On the other hand, it is worth mentioning that a recent meta-analysis of clinical trials failed to find an effect of metformin on depression risk, while suggesting a potential role of pioglitazone (Moulton et al., 2018).
Among SGLT2 inhibitors, dapagliflozin was the drug associated with the lowest occurrence of depression-like symptoms compared to non-users of this drug, as confirmed by three statistically significant disproportionality scores ROR [0.543 (0.304–0.971)], EBGM [0.565 (0.343–0.888)] and ERAM [0.573 (0.331–0.870)], from our FAERS analysis. To date, positive but very limited evidence both on their potential neuroprotective effect and the likely underlying mechanism was available for SGLT2 inhibitors (Şahin et al., 2020). Studies have highlighted their antioxidant, anti-inflammatory, and antiapoptotic mechanisms, regardless of their glycaemic control benefits (Shaikh et al., 2016; Sa-Nguanmoo et al., 2017; El-Sahar et al., 2020; Esterline et al., 2020; Wiciński et al., 2020; Muhammad et al., 2021). Dapaglifozin attenuated depressive-like behaviour of male rats in the forced swim test and was also found to be comparable to imipramine in the treatment of mild-to-moderate depression (Cam et al., 2019). In humans, these drugs improved the quality of life of people with diabetes (maybe due to the weight loss observed in the enrolled patients); however, no change in terms of Pittsburgh Sleep Quality, and Beck Anxiety Inventory scores was found.
## 4.1 Strengths and limitations
This is the first study aimed at providing a comprehensive overview of the potential beneficial antidepressant effect of antidiabetic agents in a high-scale population data from the two largest spontaneous reporting system databases, i.e., FAERS and VigiBase.
Pharmacovigilance databases are commonly used to track frequent adverse events; however, with a sufficient amount of data, they may also be used to investigate the beneficial outcomes through monitoring reductions in frequency of the related adverse events (Cohen et al., 2017).
The spontaneously reported adverse drug events may serve as useful indicators to predict new opportunities for drug repositioning, making spontaneous reporting system databases valuable data sources for driving clinical research (Pushpakom et al., 2019).
Growing number of evidence supports this innovative approach based on the use of pharmacovigilance databases, especially FAERS and VigiBase, to investigate promising new beneficial effects of drugs in real-world clinical practice, in a variety of clinical settings (e.g., psychiatry, neurology, cardiology) (Wang et al., 2016; Cohen et al., 2017; Carnovale et al., 2018; 2019b; 2019a; Horinouchi et al., 2018; Hosomi et al., 2018; Mazhar et al., 2019; Chrétien et al., 2021).
Furthermore, as real-world data (RWD), including spontaneously reported adverse events, refer to a large amount of clinical data collected during the patient’s daily life, they can address intrinsic limitations of traditional clinical trials, such as highly selected populations, strict inclusion/exclusion criteria, small sample sizes, short follow-up periods, with consequent lack of external validity.
Indeed, RWD are often used to focus on special populations who are usually excluded from RCTs, such as patients receiving polytherapy, children, pregnant women, and elderly people (Trifirò et al., 2019); RWD are hence gaining increasing attention in the whole drug life-cycle process, including regulatory decision-making. In our study, the large set of data (we included two cohorts of 121,368 and 85,267 depressed patients experiencing therapy failure, from FAERS and VigiBase, respectively) provided sufficient statistical power for the analysis to generate hypotheses for unknown potential uses.
On the other hand, it is well known that the use of pharmacovigilance databases has some intrinsic limitations: reporting might be influenced by factors such as notoriety bias, selection bias and under-reporting and there is no certainty that the reported event was causally related due to the suspect drug. Moreover, as these data sources are designed to report adverse events, unintentional beneficial effects of the drug therapy could not be recorded. Pharmacovigilance data cannot be eventually used to calculate the incidence rates of events. In view of the above-mentioned limitations of pharmacovigilance databases, it is worth mentioning that RCTs are non-etheless the gold standard in evidence-based medicine for demonstrating drug efficacy (Compher, 2010) and new clinical studies specifically designed at investigating the role of antidiabetics in depressed patients are needed.
## 5 Conclusion
All the antidiabetic drug classes investigated in our pharmacoepidemiological study showed a potential beneficial effect to depressed patients (in terms of a decreased occurrence of therapy failure/depression-related symptoms), with a high heterogeneity in terms of statistically significant disproportionality scores. This comprehensive overview suggests that some specific pharmacological agents, in particular, GLP-1 analogues might exert a more prominent beneficial and clinically meaningful effects. Due to the insufficient therapeutic response of patients to the available antidepressant medications, repositioning of antidiabetic drugs might become a valuable new approach to improve drug treatment in depression. In view of the nature of this study, the result of this research is not an ultimate conclusion, but a suggestion for further clinical research. Gold-standard RCTs are warranted to confirm these encouraging results, albeit preliminary, and properly characterize the topic.
## 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
VB, RPvM, CC provided a substantial contribution to the conception/design of the work; RPvM had an important role in the acquisition of data and the following analysis. VB and CC worked on the interpretation of data together with MG, GM, GG and AB. VB, RPvM, MG, CC drafted the work and EC, SR, MP, MN revised it critically for important intellectual content.
## Conflict of interest
RPvM is an employee of Oracle Health Sciences.
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/fphar.2023.1128387/full#supplementary-material
## References
1. Alcocer-Gómez E., de Miguel M., Casas-Barquero N., Núñez-Vasco J., Sánchez-Alcazar J. A., Fernández-Rodríguez A.. **NLRP3 inflammasome is activated in mononuclear blood cells from patients with major depressive disorder**. *Brain Behav. Immun.* (2014) **36** 111-117. DOI: 10.1016/j.bbi.2013.10.017
2. Alcocer-Gómez E., Ulecia-Morón C., Marín-Aguilar F., Rybkina T., Casas-Barquero N., Ruiz-Cabello J.. **Stress-induced depressive behaviors require a functional NLRP3 inflammasome**. *Mol. Neurobiol.* (2016) **53** 4874-4882. DOI: 10.1007/s12035-015-9408-7
3. AlHussain F., AlRuthia Y., Al-Mandeel H., Bellahwal A., Alharbi F., Almogbel Y.. **Metformin improves the depression symptoms of women with polycystic ovary syndrome in a lifestyle modification program**. *Patient Prefer Adherence* (2020) **14** 737-746. DOI: 10.2147/PPA.S244273
4. Athauda D., Foltynie T.. **The glucagon-like peptide 1 (GLP) receptor as a therapeutic target in Parkinson’s disease: Mechanisms of action**. *Drug Discov. Today* (2016) **21** 802-818. DOI: 10.1016/j.drudis.2016.01.013
5. Cam M. E., Hazar-Yavuz A. N., Yildiz S., Keles R., Ertas B., Kabasakal L.. **Dapagliflozin attenuates depressive-like behavior of male rats in the forced swim test**. *Eur. Neuropsychopharmacol.* (2019) **29** S262-S263. DOI: 10.1016/j.euroneuro.2018.11.418
6. Carnovale C., Mazhar F., Arzenton E., Moretti U., Pozzi M., Mosini G.. **Bullous pemphigoid induced by dipeptidyl peptidase-4 (DPP-4) inhibitors: A pharmacovigilance-pharmacodynamic/pharmacokinetic assessment through an analysis of the vigibase**. *Expert Opin. Drug Saf.* (2019a) **18** 1099-1108. DOI: 10.1080/14740338.2019.1668373
7. Carnovale C., Mazhar F., Pozzi M., Gentili M., Clementi E., Radice S.. **A characterization and disproportionality analysis of medication error related adverse events reported to the FAERS database**. *Expert Opin. Drug Saf.* (2018) **17** 1161-1169. DOI: 10.1080/14740338.2018.1550069
8. Carnovale C., Mosini G., Gringeri M., Battini V., Mazhar F., Pozzi M.. **Interaction between paracetamol and lamotrigine: New insights from the FDA adverse event reporting system (FAERS) database**. *Eur. J. Clin. Pharmacol.* (2019b) **75** 1323-1325. DOI: 10.1007/s00228-019-02691-4
9. Ceriello A., Novials A., Ortega E., Canivell S., la Sala L., Pujadas G.. **Glucagon-like peptide 1 reduces endothelial dysfunction, inflammation, and oxidative stress induced by both hyperglycemia and hypoglycemia in type 1 diabetes**. *Diabetes Care* (2013) **36** 2346-2350. DOI: 10.2337/dc12-2469
10. Chan K. L., Cathomas F., Russo S. J.. **Central and peripheral inflammation link metabolic syndrome and major depressive disorder**. *Physiol. (Bethesda)* (2019) **34** 123-133. DOI: 10.1152/physiol.00047.2018
11. Chen F., Wei G., Wang Y., Liu T., Huang T., Wei Q.. **Risk factors for depression in elderly diabetic patients and the effect of metformin on the condition**. *BMC Public Health* (2019) **19** 1063. DOI: 10.1186/s12889-019-7392-y
12. Chrétien B., Jourdan J.-P., Davis A., Fedrizzi S., Bureau R., Sassier M.. **Disproportionality analysis in VigiBase as a drug repositioning method for the discovery of potentially useful drugs in Alzheimer’s disease**. *Br. J. Clin. Pharmacol.* (2021) **87** 2830-2837. DOI: 10.1111/bcp.14690
13. **The DPP-4 inhibitor vildagliptin as adjunct in major depressive disorder patients (NCT04410341)**. (2022)
14. Cohen I., Makunts T., Atayee R., Abagyan R.. **Population scale data reveals the antidepressant effects of ketamine and other therapeutics approved for non-psychiatric indications**. *Sci. Rep.* (2017) **7** 1450. DOI: 10.1038/s41598-017-01590-x
15. Compher C.. **Efficacy vs effectiveness**. *JPEN J. Parenter. Enter. Nutr.* (2010) **34** 598-599. DOI: 10.1177/0148607110381906
16. DuMouchel W.. **Bayesian data mining in large frequency tables, with an application to the FDA spontaneous reporting system**. *Am. Stat.* (1999) **53** 177-190. DOI: 10.2307/2686093
17. DuMouchel W., Harpaz R.. **Regression-adjusted GPS algorithm (RGPS)**. (2012)
18. El-Sahar A. E., Rastanawi A. A., El-Yamany M. F., Saad M. A.. **Dapagliflozin improves behavioral dysfunction of Huntington’s disease in rats via inhibiting apoptosis-related glycolysis**. *Life Sci.* (2020) **257** 118076. DOI: 10.1016/j.lfs.2020.118076
19. Essmat N., Soliman E., Mahmoud M. F., Mahmoud A. A. A.. **Antidepressant activity of anti-hyperglycemic agents in experimental models: A review**. *Diabetes Metabolic Syndrome Clin. Res. Rev.* (2020) **14** 1179-1186. DOI: 10.1016/j.dsx.2020.06.021
20. Esterline R., Oscarsson J., Burns J.. **A role for sodium glucose cotransporter 2 inhibitors (SGLT2is) in the treatment of Alzheimer’s disease?**. *Int. Rev. Neurobiol.* (2020) **155** 113-140. DOI: 10.1016/bs.irn.2020.03.018
21. Fang W., Zhang J., Hong L., Huang W., Dai X., Ye Q.. **Metformin ameliorates stress-induced depression-like behaviors via enhancing the expression of BDNF by activating AMPK/CREB-mediated histone acetylation**. *J. Affect Disord.* (2020) **260** 302-313. DOI: 10.1016/j.jad.2019.09.013
22. Fescharek R., Kübler J., Elsasser U., Frank M., Güthlein P.. **Medical dictionary for regulatory activities (MedDRA)**. *Int. J. Pharm. Med.* (2004) **18** 259-269. DOI: 10.2165/00124363-200418050-00001
23. Fiorentino T. V., Prioletta A., Zuo P., Folli F.. **Hyperglycemia-induced oxidative stress and its role in diabetes mellitus related cardiovascular diseases**. *Curr. Pharm. Des.* (2013) **19** 5695-5703. DOI: 10.2174/1381612811319320005
24. Golden S. H., Lazo M., Carnethon M., Bertoni A. G., Schreiner P. J., Diez Roux A. v.. **Examining a bidirectional association between depressive symptoms and diabetes**. *JAMA - J. Am. Med. Assoc.* (2008) **299** 2751-2759. DOI: 10.1001/jama.299.23.2751
25. Gumuslu E., Mutlu O., Celikyurt I. K., Ulak G., Akar F., Erden F.. **Exenatide enhances cognitive performance and upregulates neurotrophic factor gene expression levels in diabetic mice**. *Fundam. Clin. Pharmacol.* (2016) **30** 376-384. DOI: 10.1111/fcp.12192
26. Holt R. I., de Groot M., Golden S. H.. **Diabetes and depression**. *Curr. Diab. Rep.* (2014) **14** 491. DOI: 10.1007/s11892-014-0491-3
27. Horinouchi Y., Ikeda Y., Fukushima K., Imanishi M., Hamano H., Izawa-Ishizawa Y.. **Renoprotective effects of a factor xa inhibitor: Fusion of basic research and a database analysis**. *Sci. Rep.* (2018) **8** 10858. DOI: 10.1038/s41598-018-29008-2
28. Hosomi K., Fujimoto M., Ushio K., Mao L., Kato J., Takada M.. **An integrative approach using real-world data to identify alternative therapeutic uses of existing drugs**. *PLoS One* (2018) **13** e0204648. DOI: 10.1371/journal.pone.0204648
29. Hunter K., Hölscher C.. **Drugs developed to treat diabetes, liraglutide and lixisenatide, cross the blood brain barrier and enhance neurogenesis**. *BMC Neurosci.* (2012) **13** 33. DOI: 10.1186/1471-2202-13-33
30. Kamble M., Gupta R., Rehan H. S., Gupta L. K.. **Neurobehavioral effects of liraglutide and sitagliptin in experimental models**. *Eur. J. Pharmacol.* (2016) **774** 64-70. DOI: 10.1016/j.ejphar.2016.02.003
31. Kim Y.-K., Kim O. Y., Song J.. **Alleviation of depression by glucagon-like peptide 1 through the regulation of neuroinflammation, neurotransmitters, neurogenesis, and synaptic function**. *Front. Pharmacol.* (2020) **11** 1270. DOI: 10.3389/fphar.2020.01270
32. Laake J.-P. S., Stahl D., Amiel S. A., Petrak F., Sherwood R. A., Pickup J. C.. **The association between depressive symptoms and systemic inflammation in people with type 2 diabetes: Findings from the south london diabetes study**. *Diabetes Care* (2014) **37** 2186-2192. DOI: 10.2337/dc13-2522
33. Lamb R. E., Goldstein B. J.. **Modulating an oxidative-inflammatory cascade: Potential new treatment strategy for improving glucose metabolism, insulin resistance, and vascular function**. *Int. J. Clin. Pract.* (2008) **62** 1087-1095. DOI: 10.1111/j.1742-1241.2008.01789.x
34. Li Y., Bader M., Tamargo I., Rubovitch V., Tweedie D., Pick C. G.. **Liraglutide is neurotrophic and neuroprotective in neuronal cultures and mitigates mild traumatic brain injury in mice**. *J. Neurochem.* (2015) **135** 1203-1217. DOI: 10.1111/jnc.13169
35. Mansur R. B., Zugman A., Ahmed J., Cha D. S., Subramaniapillai M., Lee Y.. **Treatment with a GLP-1R agonist over four weeks promotes weight loss-moderated changes in frontal-striatal brain structures in individuals with mood disorders**. *Eur. Neuropsychopharmacol.* (2017) **27** 1153-1162. DOI: 10.1016/j.euroneuro.2017.08.433
36. Martins L. B., Braga Tibães J. R., Berk M., Teixeira A. L.. **Diabetes and mood disorders: Shared mechanisms and therapeutic opportunities**. *Int. J. Psychiatry Clin. Pract.* (2022) **26** 183-195. DOI: 10.1080/13651501.2021.1957117
37. Mazhar F., Pozzi M., Gentili M., Scatigna M., Clementi E., Radice S.. **Association of hyponatraemia and antidepressant drugs: A pharmacovigilance-pharmacodynamic assessment through an analysis of the us food and drug administration adverse event reporting system (FAERS) database**. *CNS Drugs* (2019) **33** 581-592. DOI: 10.1007/s40263-019-00631-5
38. McClean P. L., Gault V. A., Harriott P., Hölscher C.. **Glucagon-like peptide-1 analogues enhance synaptic plasticity in the brain: A link between diabetes and alzheimer’s disease**. *Eur. J. Pharmacol.* (2010) **630** 158-162. DOI: 10.1016/j.ejphar.2009.12.023
39. **MedDRA-Support Documentation (n.d.)**. (2022)
40. Monnier L., Mas E., Ginet C., Michel F., Villon L., Cristol J.-P.. **Activation of oxidative stress by acute glucose fluctuations compared with sustained chronic hyperglycemia in patients with type 2 diabetes**. *JAMA* (2006) **295** 1681-1687. DOI: 10.1001/jama.295.14.1681
41. Moulton C. D., Hopkins C. W. P., Ismail K., Stahl D.. **Repositioning of diabetes treatments for depressive symptoms: A systematic review and meta-analysis of clinical trials**. *Psychoneuroendocrinology* (2018) **94** 91-103. DOI: 10.1016/j.psyneuen.2018.05.010
42. Moulton C. D., Rokakis A. S., Pickup J. C., Young A. H., Stahl D., Ismail K.. **Sitagliptin for depressive symptoms in type 2 diabetes: A feasibility randomized controlled trial**. *Psychosom. Med.* (2021) **83** 913-923. DOI: 10.1097/PSY.0000000000000985
43. Muhammad R. N., Ahmed L. A., Abdul Salam R. M., Ahmed K. A., Attia A. S.. **Crosstalk among NLRP3 inflammasome, ETBR signaling, and miRNAs in stress-induced depression-like behavior: A modulatory role for SGLT2 inhibitors**. *Neurotherapeutics* (2021) **18** 2664-2681. PMID: 34664178
44. Odaira T., Nakagawasai O., Takahashi K., Nemoto W., Sakuma W., Lin J.-R.. **Mechanisms underpinning AMP-activated protein kinase-related effects on behavior and hippocampal neurogenesis in an animal model of depression**. *Neuropharmacology* (2019) **150** 121-133. DOI: 10.1016/j.neuropharm.2019.03.026
45. Pintana H., Apaijai N., Pratchayasakul W., Chattipakorn N., Chattipakorn S. C.. **Effects of metformin on learning and memory behaviors and brain mitochondrial functions in high fat diet induced insulin resistant rats**. *Life Sci.* (2012) **91** 409-414. DOI: 10.1016/j.lfs.2012.08.017
46. Porter D., Faivre E., Flatt P. R., Hölscher C., Gault V. A.. **Actions of incretin metabolites on locomotor activity, cognitive function and**. *Pept. (N.Y.)* (2012) **35** 1-8. DOI: 10.1016/j.peptides.2012.03.014
47. Pozzi M., Mazhar F., Peeters G. G. A. M., Vantaggiato C., Nobile M., Clementi E.. **A systematic review of the antidepressant effects of glucagon-like peptide 1 (GLP-1) functional agonists: Further link between metabolism and psychopathology**. *Sect. Edi J. Affect Disord.* (2019) **257** 774-778. DOI: 10.1016/j.jad.2019.05.044
48. Pushpakom S., Iorio F., Eyers P. A., Escott K. J., Hopper S., Wells A.. **Drug repurposing: Progress, challenges and recommendations**. *Nat. Rev. Drug Discov.* (2019) **18** 41-58. DOI: 10.1038/nrd.2018.168
49. Sa-Nguanmoo P., Tanajak P., Kerdphoo S., Jaiwongkam T., Pratchayasakul W., Chattipakorn N.. **SGLT2-inhibitor and DPP-4 inhibitor improve brain function via attenuating mitochondrial dysfunction, insulin resistance, inflammation, and apoptosis in HFD-induced obese rats**. *Toxicol. Appl. Pharmacol.* (2017) **333** 43-50. DOI: 10.1016/j.taap.2017.08.005
50. Şahin S., Haliloğlu Ö., Polat Korkmaz Ö., Durcan E., Rekalı Şahin H., Yumuk V. D.. **Does treatment with sodium-glucose co-transporter-2 inhibitors have an effect on sleep quality, quality of life, and anxiety levels in people with Type 2 diabetes mellitus?**. *Turk J. Med. Sci.* (2020) **51** 735-742. DOI: 10.3906/sag-2008-37
51. Sakaeda T., Tamon A., Kadoyama K., Okuno Y.. **Data mining of the public version of the FDA adverse event reporting system**. *Int. J. Med. Sci.* (2013) **10** 796-803. DOI: 10.7150/ijms.6048
52. Saritha M. K., Chandrashekar K.. **Antidepressant activity of dpp-4 inhibitors in albino mice, an experimental study**. *Natl. J. Med. Dent. Res.* (2018) **6** 523-526
53. Shaikh S., Rizvi S. M. D., Shakil S., Riyaz S., Biswas D., Jahan R.. **Forxiga (dapagliflozin): Plausible role in the treatment of diabetes-associated neurological disorders**. *Biotechnol. Appl. Biochem.* (2016) **63** 145-150. DOI: 10.1002/bab.1319
54. Shivavedi N., Kumar M., Tej G. N. V. C., Nayak P. K.. **Metformin and ascorbic acid combination therapy ameliorates type 2 diabetes mellitus and comorbid depression in rats**. *Brain Res.* (2017) **1674** 1-9. DOI: 10.1016/j.brainres.2017.08.019
55. Su W.-J., Peng W., Gong H., Liu Y.-Z., Zhang Y., Lian Y.-J.. **Antidiabetic drug glyburide modulates depressive-like behavior comorbid with insulin resistance**. *J. Neuroinflammation* (2017) **14** 210. DOI: 10.1186/s12974-017-0985-4
56. Trifirò G., Gini R., Barone-Adesi F., Beghi E., Cantarutti A., Capuano A.. **The role of European healthcare databases for post-marketing drug effectiveness, safety and value evaluation: Where does Italy stand?**. *Drug Saf.* (2019) **42** 347-363. DOI: 10.1007/s40264-018-0732-5
57. van Bloemendaal L., IJzerman R. G., ten Kulve J. S., Barkhof F., Konrad R. J., Drent M. L.. **GLP-1 receptor activation modulates appetite- and reward-related brain areas in humans**. *Diabetes* (2014) **63** 4186-4196. DOI: 10.2337/db14-0849
58. van Manen R. P., Fram D., DuMouchel W.. **Signal detection methodologies to support effective safety management**. *Expert Opin. Drug Saf.* (2007) **6** 451-464. DOI: 10.1517/14740338.6.4.451
59. Wahlqvist M. L., Lee M.-S., Chuang S.-Y., Hsu C.-C., Tsai H.-N., Yu S.-H.. **Increased risk of affective disorders in type 2 diabetes is minimized by sulfonylurea and metformin combination: A population-based cohort study**. *BMC Med.* (2012) **10** 150. DOI: 10.1186/1741-7015-10-150
60. Wang C.-P., Lorenzo C., Habib S. L., Jo B., Espinoza S. E.. **Differential effects of metformin on age related comorbidities in older men with type 2 diabetes**. *J. Diabetes Complicat.* (2017) **31** 679-686. DOI: 10.1016/j.jdiacomp.2017.01.013
61. Wang K., Wan M., Wang R.-S., Weng Z.. **Opportunities for web-based drug repositioning: Searching for potential antihypertensive agents with hypotension adverse events**. *J. Med. Internet Res.* (2016) **18** e76. DOI: 10.2196/jmir.4541
62. Wiciński M., Wódkiewicz E., Górski K., Walczak M., Malinowski B.. **Perspective of SGLT2 inhibition in treatment of conditions connected to neuronal loss: Focus on alzheimer’s disease and ischemia-related brain injury**. *Pharm. (Basel)* (2020) **13** 379. DOI: 10.3390/ph13110379
63. Wium-Andersen I. K., Osler M., Jørgensen M. B., Rungby J., Wium-Andersen M. K.. **Diabetes, antidiabetic medications and risk of depression - a population-based cohort and nested case-control study**. *Psychoneuroendocrinology* (2022) **140** 105715. DOI: 10.1016/j.psyneuen.2022.105715
|
---
title: 'Practice of the new supervised machine learning predictive analytics for glioma
patient survival after tumor resection: Experiences in a high-volume Chinese center'
authors:
- Yushan Li
- Maodong Ye
- Baolong Jia
- Linwei Chen
- Zubang Zhou
journal: Frontiers in Surgery
year: 2023
pmcid: PMC9981970
doi: 10.3389/fsurg.2022.975022
license: CC BY 4.0
---
# Practice of the new supervised machine learning predictive analytics for glioma patient survival after tumor resection: Experiences in a high-volume Chinese center
## Abstract
### Objective
This study aims to assess the effectiveness of the Gradient Boosting (GB) algorithm on glioma prognosis prediction and to explore new predictive models for glioma patient survival after tumor resection.
### Methods
A cohort of 776 glioma cases (WHO grades II–IV) between 2010 and 2017 was obtained. Clinical characteristics and biomarker information were reviewed. Subsequently, we constructed the conventional Cox survival model and three different supervised machine learning models, including support vector machine (SVM), random survival forest (RSF), Tree GB, and Component GB. Then, the model performance was compared with each other. At last, we also assessed the feature importance of models.
### Results
The concordance indexes of the conventional survival model, SVM, RSF, Tree GB, and Component GB were 0.755, 0.787, 0.830, 0.837, and 0.840, respectively. All areas under the cumulative receiver operating characteristic curve of both GB models were above 0.800 at different survival times. Their calibration curves showed good calibration of survival prediction. Meanwhile, the analysis of feature importance revealed Karnofsky performance status, age, tumor subtype, extent of resection, and so on as crucial predictive factors.
### Conclusion
Gradient Boosting models performed better in predicting glioma patient survival after tumor resection than other models.
## Introduction
Glioma is the most widely recognized primary tumor in the central nervous system (CNS) [1]. Accounting for around $80\%$ of malignant CNS tumors [1], gliomas are composed of lower-grade gliomas [LGGs; World Health Organization (WHO) grades II and III] and grade IV gliomas (glioblastoma, GBM). The treatment of glioma is troublesome, and tumor resection is the main approach to treatment. Due to the large heterogeneity between different kinds of gliomas, the prognosis of glioma patients is diverse, and the survival always ranges from a few months to 10 years [2, 3]. Obviously, GBM was supposed to have a poorer prognosis than diffuse low-grade and intermediate-grade gliomas for its characteristics of invading growth and easy recurrence. However, along with the presence of certain molecular markers and various clinical characteristics, including age, Karnofsky performance status (KPS), symptoms, and so on, the prognosis varies even in this most malignant type of glioma, GBM. Predicting glioma patient survival after tumor resection still remains a great challenge for clinical doctors.
Nowadays, there have been endeavors, mainly in three directions, to explore useful predictive models for glioma prognosis. Some researchers have focused on traditional multivariate Cox regression models with several certain prognostic factors. For example, Gittleman et al. [ 4] developed a survival nomogram for LGGs with independent validation. Meanwhile, some turn to new biomarkers for the construction of models. Not long ago, Zhang et al. [ 5] constructed a novel model using immune-related gene signature, which is also effective in predicting overall survival in primary LGG. What is more, some researchers have concentrated on radiomics feature prediction models and made some achievements [6]. Albeit the effort in putting forward these models, some shortcomings limit their usefulness and availability of these models. First, the traditional statistic approach has a huge limitation: its analysis is based on the condition of a linear relationship and might miss the nonlinear relationship between input and outcome. In other words, this approach cannot fully use medical information, which makes it unable to adjust to the era of big data. Second, as Jakola et al. [ 7] claimed, a pure biomarker approach for prediction, such as gene signature model, is of limited value because tumor classes and tumor cells are neither stable over time nor homogeneous throughout the lesion tissue. Third, prediction models based on radiomics features are powerful and promising, but we acknowledge that the techniques are at an early stage and available only at a limited number of centers and not readily validated in medical practice yet [7]. Therefore, it is still necessary to explore a new predictive model based on the algorithm suited to the big-data era, with the combination of common clinical features and reliable biomarkers as prognostic factors.
Recently, supervised machine learning (ML) methods have demonstrated precise predictive capacity, being progressively utilized in the prognosis prediction of different diseases [8]. The supervised ML approach is a kind of data-driven analysis method, including support vector machine (SVM) [9], decision tree [10], and so on, which integrates multiple risk factors into a predictive algorithm and performs well with complex information [11]. Gradient Boosting (GB) is one of the supervised ML algorithms. Although it was strange for medical workers, this ML algorithm did have a good performance in medical scenes, such as predicting the survival outcome of triple-negative breast cancer [12] and the recurrence of colorectal cancer [13]. So far, studies seldom used Gradient Boosting to analyze and predict glioma prognosis. This study was conducted to assess its effectiveness on glioma prognosis prediction and to explore new predictive models for glioma patient survival after tumor resection.
## Patients
Approved by the Institutional Review Board of Sun Yat-sen University, this study was conducted in the Neurosurgery unit, the First Affiliated Hospital, Sun Yat-sen University, a high-volume central center that performs approximately 100 glioma surgeries yearly. In accordance with the guidelines for retrospective study in our institution, the institutional review board waived the requirement for patients' informed consent. Our study only included cases of astrocytoma, oligodendroglioma and oligoastrocytoma, anaplastic astrocytoma, oligodendroglioma and oligoastrocytoma, and glioblastoma. A cohort of 776 glioma cases (WHO grades II–IV) between 2010 and 2017 was obtained. This consecutive malignant series consisted of 74 cases of WHO grade III (anaplastic astrocytoma, oligodendroglioma, and oligoastrocytoma), 268 cases of WHO grade IV (glioblastoma), and 434 cases of WHO grade II (astrocytoma, oligodendroglioma, and oligoastrocytoma).
## Clinical characteristics
Most data were accessible through the hospital database. All data were extracted into two copies of a standardized form by two research assistants independently and integrated into the final file version by a third. Discrepancies were discussed and resolved by consensus. The extracted characteristics include age at surgery, gender, symptoms (seizures, headaches/dizziness, nausea/vomiting, limb dysfunction, blurred vision, or other cranial nerve deficit), duration of the first presenting symptom, preoperative KPS, tumor size and location, time of surgery, extent of resection (gross-total resection and others), tumor subtype, treatment after surgery (chemotherapy and/or radiotherapy), survival status (alive or dead), and survival/follow-up time. The subtype of glioma was reviewed by a pathologist according to the latest 2016 WHO criteria [14]. The deficit of motor, visual, or cranial never function was confirmed by the proof of physical examination, diffusion tensor imaging (DTI)-based tractography, and so on. The same as the definition by Okamoto et al. [ 15], the extent of resection was categorized, where gross-total resection was defined as residual tumor less than $5\%$. The follow-up data were collected until December 2019. Survival/follow-up time was calculated from the date of tumor resection to death (any cause) or censor (still survived) in December 2019. All patients were followed up at the regular interval of 3 months for the initial 3 years and afterward followed every 1 year until death. The last follow-up for every single accessible patient was finished in December 2019.
## Biomarkers
Biomarkers’ detection, including immunohistochemistry (IHC) and molecular genetics, was performed on histological specimens that were obtained at the time of resection surgery prior to chemotherapy and/or radiotherapy treatment. The detection of kit67, p53, vimentin, and glial fibrillary acidic protein (GFAP) was performed using immunohistochemical stains in glioma by standard techniques that were described previously [16]. For the specimen with p53 immunohistochemical stain, the presence of strong positive tumor nuclei in more than $10\%$ of cells was marked as immunopositive, which indicated the mutational status of TP53 [17]. The immunopositivity of vimentin was identified when more than $25\%$ of tumor nuclei stained positive with vimentin IHC stain. GFAP immunopositivity was marked when any tumor nuclei were positive with GFAP IHC stain. The Ki-67 index was recorded as the average percentage of the positive ones on the total number of nuclei at 400× magnification, where “≥$10\%$” represented high Ki-67 expression [18]. As for molecular genetics, the biomarker we detected was methylation of the O6-methylgaunine-DNA-methyltransferase (MGMT) promoter. This test was done using methylation-specific PCR.
## Supervised machine learning algorithm
SVM, as a machine learning algorithm, has been widely used in the prognosis of diseases. Decision tree is a well-known ML approach for statistical problems, which represents the mapping relationship between properties and values. It consists of a root node, internal nodes, and leaf nodes, where leaf nodes correspond to values represented by the path from the root node to the leaf node. Decision tree can be used for survival analysis [19]. Here, we used survival decision tree as the base learner of random forests (RFs). RF is an ensemble tree method whose final prediction is the average of all predictions from every tree in the forest. RF performs better in prediction than a single tree because a combination of predictions from separate methods could substantially promote prediction performance [20]. Random survival forest (RSF) is an adaptation of random forest, which is designed for the analysis of survival data [21].
GB is an ML technique that can be used for survival analysis. Here, we used component-wise least squares and survival decision tree as two types of base learners, respectively. The Gradient Boosting algorithm produces different weak prediction models (for instance, component-wise least squares) at each step and combines them into a total model at different weights. The prediction of the weak model that Gradient Boosting produced at each step generates a unanimous gradient direction of the loss function. The details have been described previously [22].
## Model evaluation
Harrell's concordance index (c-index), defined as the ratio of correctly ordered (concordant) pairs to comparable pairs, is a measure of the rank correlation between predicted risk scores and observed time points. A value of 1 refers to perfect prediction, while a value of 0.5 means that prediction does not perform better than random guessing.
The area under the receiver operating characteristic curve (ROC curve) is often used to assess the discrimination of the binary classification model. When extending the ROC curve to survival time, it gives rise to the time-dependent cumulative ROC curve at a certain survival time t. The area under the cumulative ROC curve (AUC) at time t indicates how well a model can distinguish subjects who will experience an event by time t from those who will not.
The calibration curve is a graphical measure of the calibration of the model, which is a linear plot with the predicted event on the x-axis and the observed event on the y-axis. Good calibration would be matched by a regression line with a 45° slope.
To fully capture the true utility of a prediction model, the sensitivity and specificity of models for predicting 6-, 12-, 36-, and 60-month survival were calculated after determining the optimal threshold through the ROC curve.
## Model construction
All clinical characteristics and biomarker information were included in the model training set as variables. The missing values of variables were filled with multiple imputations. Here, we randomly split the data into a training set and a test set at an 8:2 ratio using the train_test_split function in the scikit-learning module of Python (version 3.7). The scikit-survival module (version 0.12.1) was used to construct ML models, SVM, RF, Tree GB, and Component GB. ML algorithms involve many hyperparameters that are significant for performance prediction. The optimal combination of hyperparameters was determined using the method of grid search. During every cross-validation, $\frac{1}{3}$ of the data in the training set were randomly excluded as out-of-bag (OOB) data for validation. For different combinations of hyperparameters, the mean c-index on the validation data was calculated after 50 times cross-validation. The hyperparameter combination with the best c-index was selected as optimum. After constructing the model, we usually assess the feature importance by calculating its contribution to the c-index, namely, the decrease of c-index after discombobulating the relationship of this feature with survival.
To compare the performance difference between ML models and conventional survival models, we also built the Cox proportional hazards model. Three continuous variables, age at surgery, preoperative KPS, and tumor size, were transformed into categorical variables to obtain the best model prediction performance. Cutoffs for these variables were 50 years, 70 cm, and 55 cm, respectively. All variables were entered into the model step by step, and the final model only included variables with a significant risk ratio.
## Statistical analysis
Mean ± SD or median (IQR) was chosen to describe continuous variables regarding their statistical distribution, while categorical variables were expressed in the form of example numbers (%). $P \leq 0.05$ was set as the criteria of statistical significance in all analyses. The confidence interval (CI) of the AUC was computed by the bootstrap method, while $95\%$ CI was computed with 2,000 stratified bootstrap replicates. The comparisons of the c-index between different models were conducted using the R package Survcomp.
## Characteristic overview
The sociodemographic and characteristics of the study population are presented in Table 1. The most frequent symptom was headaches or dizziness, while the most frequent tumor location and subtype were parietal lobe and diffuse astrocytoma, respectively. The medians of the duration of the first presenting symptom, preoperative KPS, and tumor size were 1.90 months, 60 mm, and 45.00 mm, respectively. Gross-total resection was adopted in $90.0\%$ of patients. The immunopositivity of GFAP, Vimentin, and p53 was observed in more than half of the patients. The median of survival time was 32.65 months.
**Table 1**
| Unnamed: 0 | Total (N = 776) |
| --- | --- |
| Demographics | Demographics |
| Age at surgery, years, median (IQR) | 38.00 (24.25–53.00) |
| Age > 50 years (%) | 29.0% (225/776) |
| Sex (M, %) | 58.1% (451/776) |
| Symptoms | Symptoms |
| Seizures (%) | 34.7% (267/769) |
| Headaches/dizziness (%) | 61.5% (473/769) |
| Nausea/vomiting (%) | 25.0% (192/769) |
| Limb dysfunction (%) | 18.7% (144/769) |
| Blurred vision (%) | 9.5% (73/769) |
| Other cranial nerve deficit (%) | 16.6% (128/769) |
| Duration of symptom, m, median (IQR) | 1.90 (0.70–6.00) |
| Preoperative KPS | Preoperative KPS |
| KPS, median (IQR) | KPS, median (IQR) |
| 60.00 (50.00–90.00) | |
| KPS > 70 (%) | 37.9% (275/726) |
| Tumor | Tumor |
| Size, mm, median (IQR) | 45.00 (32.00—60.00) |
| Size > 55 cm (%) | 32.0% (239/747) |
| Tumor location | Tumor location |
| Frontal lobe (%) | 42.4% (327/771) |
| Temporal lobe (%) | 17.5% (135/771) |
| Parietal lobe (%) | 32.9% (254/771) |
| Occipital lobe (%) | 8.3% (64/771) |
| Infra-tentorial (%) | 10.1% (78/771) |
| Others (%) | 16.1% (124/771) |
| Extent of resection | Extent of resection |
| Gross-total resection (%) | 90.0% (691/768) |
| Others (%) | 10.0% (77/768) |
| Tumor subtype | Tumor subtype |
| Diffuse astrocytoma, IDH mutant (%) | 9.0% (70/776) |
| Diffuse astrocytoma, IDH wildtype (%) | 0.5% (4/776) |
| Diffuse astrocytoma, NOS (%) | 30.2% (234/776) |
| Oligodendroglioma, IDH mutant (%) | 3.1% (24/776) |
| Oligodendroglioma, NOS (%) | 11.3% (88/776) |
| Oligoastrocytoma, NOS (%) | 1.8% (14/776) |
| Anaplastic astrocytoma, IDH mutant (%) | 1.2% (9/776) |
| Anaplastic astrocytoma, IDH wildtype (%) | 0.1% (1/776) |
| Anaplastic astrocytoma, NOS (%) | 4.9% (38/776) |
| Anaplastic oligodendroglioma, IDH mutant (%) | 0.4% (3/776) |
| Anaplastic oligodendroglioma, NOS (%) | 2.7% (21/776) |
| Anaplastic oligoastrocytoma, NOS (%) | 0.3% (2/776) |
| Glioblastoma, IDH mutant (%) | 1.2% (9/776) |
| Glioblastoma, IDH wildtype (%) | 7.9% (61/776) |
| Glioblastoma, NOS (%) | 25.5% (198/776) |
| Treatment strategy | Treatment strategy |
| Chemotherapy (%) | 74.8% (564/754) |
| Radiotherapy (%) | 46.6% (355/761) |
| Biomarkers | Biomarkers |
| Ki-67, median (IQR) | 0.12 (0.05—0.30) |
| GFAP immunopositivity (%) | 79.7% (570/715) |
| Vimentin immunopositivity (%) | 83.6% (532/636) |
| MGMT promoter methylation (%) | 41.7% (301/722) |
| p53 immunopositivity (%) | 57.6% (411/713) |
| Follow-up | Follow-up |
| Status of death (%) | 71.5% (555/776) |
| Survival time, m, median (IQR) | 32.65 (13.03–56.23) |
Figure 1 shows the correlation coefficient between each independent variable. It demonstrated low correlation between each variable.
**Figure 1:** *Correlation coefficient matrix of each variable. Each coefficient is annotated. The closer it gets to 1, the more positively correlated it is. The closer it gets to −1, the more negatively correlated it is.*
## Model performance
The flow chart of model construction is shown in Supplementary Figure S1. Supplementary Table S1 shows the detailed descriptions of the selected modules, classes, and hyperparameters in Python for each model, including the Cox survival model and supervised ML models.
The five models are compared in Table 2. The Cox proportional hazards model had the worst performance, with a concordance index of 0.755 for the test set. The SVM model was observed to have relatively poor performance, with a c-index of 0.787 for the test set. The Tree GB survival model ranked second, with a c-index of 0.837, while the RSF model ranked third, with a c-index of 0.830. The Component GB survival model had the best prediction performance, with a c-index of 0.840. In addition, we also compared the c-index values of different models on the test set. The c-index values of Tree GB and Component GB survival models were significantly higher than those of the Cox proportional hazards model ($P \leq 0.05$) and SVM model ($P \leq 0.05$). Although the comparison results of the RF model were not significant (P values were 0.332 and 0.112, respectively), relatively superior performances of Tree GB and Component GB models were still observed. The reason for no statistical significance could be attributed to the little sample size of the test set to a certain extent.
**Table 2**
| Unnamed: 0 | Training set | Testing set |
| --- | --- | --- |
| Cox proportional hazards model | 0.819 | 0.755 |
| Support vector machine model | 0.838 | 0.787 |
| Random survival forest model | 0.875 | 0.83 |
| Tree Gradient Boosting survival model | 0.905 | 0.837 |
| Component Gradient Boosting survival model | 0.857 | 0.84 |
Figure 2 shows the AUCs of both GB models. All AUCs at different survival times were above 0.800, which indicated the excellent discrimination of models. The prediction AUC and CI values of both GB models’ for 6-, 12-, 36-, and 60-month survival are specifically listed in Supplementary Tables S2 and S3, which highlighted superior predictive performance. The calibration curves of both GB models are shown in Supplementary Figures S2 and S3, where good calibration was found in survival prediction. Based on the optimal thresholds, the Tree GB model predicted 6-, 12-, 36-, and 60-month survival with $94.4\%$, $90.6\%$, $99.3\%$, and $100\%$ sensitivity and $91.3\%$, $85.7\%$, $71.3\%$, and $73.8\%$ specificity, while the Component GB model predicted the survival results with $90.0\%$, $73.5\%$, $85.2\%$, and $90.4\%$ sensitivity and $84.1\%$, $92.9\%$, $87.5\%$, and $82.8\%$ specificity, respectively. The results are listed in Supplementary Tables S4 and S5.
**Figure 2:** *Area under the cumulative ROC curves of both two Gradient Boosting models at different survival times. (A) AUC of the Component GB survival model at different survival times. (B) AUC of the Tree GB survival model at different survival times. ROC curve, receiver operating characteristic curve; AUC, area under the curve; GB, Gradient Boosting.*
## Feature importance
From Table 3, we found that KPS, the tumor subtype of glioblastoma not otherwise specified (NOS), age, tumor size, and the tumor subtype of oligodendroglioma (NOS) ranked top five in the Tree GB survival model in terms of the feature importance. As for the Component GB survival model, it was KPS, the tumor subtype of glioblastoma (NOS), age, extent of resection, and tumor size that ranked the top five. As described in Supplementary Figure S4, significant variables included in the final Cox proportional hazards model were KPS, age, tumor size, tumor subtype, extent of resection, chemotherapy, radiotherapy, p53 immunopositivity, and methylation of the MGMT promoter.
**Table 3**
| Feature importance rank | Tree Gradient Boosting model | Component Gradient Boosting model |
| --- | --- | --- |
| 1 | KPS | KPS |
| 2 | Glioblastoma, NOS | Glioblastoma, NOS |
| 3 | Age | Age |
| 4 | Tumor size | Extent of resection, others |
| 5 | Oligodendroglioma, NOS | Tumor size |
| 6 | Extent of resection, others | Glioblastoma, IDH wildtype |
| 7 | MGMT promoter methylation negative | MGMT promoter methylation positive |
| 8 | Glioblastoma, IDH wildtype | Oligodendroglioma, NOS |
| 9 | Extent of resection, gross-total resection | With radiotherapy |
| 10 | Oligoastrocytoma, NOS | p53 negative |
| 11 | MGMT promoter methylation positive | With chemotherapy |
| 12 | With radiotherapy | Oligoastrocytoma, NOS |
| 13 | p53 positive | Diffuse astrocytoma, NOS |
| 14 | With chemotherapy | MGMT promoter methylation negative |
| 15 | p53 negative | Glioblastoma, IDH mutant |
## Sensitivity analysis
We also performed sensitivity analysis to detect the robustness of ML model prediction performance. Training and testing with variables without imputation, the c-indexes of ML survival models are listed in Table 4. All c-indexes were at a high level (above 0.800), which indicated the robustness of ML model prediction performance. Considering the large heterogeneity of different types of gliomas, we deliberately tested the prediction performance of ML models on three most common gliomas, namely, diffuse astrocytoma, oligodendroglioma, and glioblastoma. As can be seen from Supplementary Table S6, the c-indexes were almost at the level of about 0.8, which proves that the model is compatible with different types of gliomas.
**Table 4**
| Variable without imputation | Tree Gradient Boosting model | Component Gradient Boosting model |
| --- | --- | --- |
| Symptoms | 0.851 | 0.857 |
| Duration of symptom | 0.831 | 0.837 |
| Preoperative KPS | 0.81 | 0.821 |
| Tumor size | 0.834 | 0.844 |
| Tumor location | 0.829 | 0.857 |
| Extent of resection | 0.803 | 0.826 |
| Chemotherapy | 0.835 | 0.836 |
| Radiotherapy | 0.842 | 0.852 |
| Ki-67 | 0.82 | 0.835 |
| GFAP immunopositivity | 0.829 | 0.84 |
| Vimentin immunopositivity | 0.823 | 0.846 |
| MGMT promoter methylation | 0.845 | 0.859 |
| p53 immunopositivity | 0.835 | 0.837 |
## Discussion
This study was designed to assess the effectiveness of the supervised ML algorithm, especially Gradient Boosting, in the prediction of glioma patient survival after tumor resection and to explore new predictive models useful for medical workers. Judging by Harrell's concordance index of the training set and test set, the Gradient Boosting algorithm ranked first on prediction performance. There were differences in prediction performance between Tree GB and Component GB algorithms. Tree GB showed better performance on the training set (c-index: 0.905) but worst performance on the test set (c-index: 0.837) than Component GB, which implied a trend of overfitting. By the way, considering the discrimination and calibration through time-dependent cumulative ROC curves and sensitivity, specificity, and calibration curves, Tree and Component GB were both good at predicting 6-, 12-, 36-, and 60-month survival after surgery. The results of the sensitive analysis revealed that both GB models were stable at the prediction outcome.
At the same time, the feature importance of ML models was also assessed. The top 15 important features in Tree or Component GB models could be reduced to nine variables, namely, KPS, age, tumor size, tumor subtype, extent of resection, chemotherapy, radiotherapy, p53 immunopositivity, and methylation of the MGMT promoter. It was in line with the significant variables included in the Cox proportional hazards model.
Consistent with the results of previous studies [23, 24], our result revealed that KPS influenced glioma patient survival after resection surgery. KPS or similar crude scales are commonly seen methods to evaluate gross functional status and have been repeatedly described as prognostic factors in the management of glioma patients [23, 24]. Also, age is one of the most established prognostic factors in patients with malignant gliomas, regardless of lower-grade [24, 25] or higher-grade gliomas [26]. As claimed by Paugh et al. [ 27], the substantial differences in the molecular features underlying age-stratified gliomas might lead to different treatment responses, accounting for different survival outcomes. A cutoff value of 55 years has been reported repeatedly to stratify glioma patients, while significantly impaired survival is always observed in those 55 years and above. Here, our study confirmed advanced age as an unfavorable prognostic factor once more. Previous studies [28, 29] have shown a strong association between preoperative tumor size and glioma survival, which is in line with the finding of our research. Regarding the extent of resection, complete curative resection is thought impossible due to the lack of clear tumor borders and the invasive behavior of the tumor. Although a number of studies [30, 31] have demonstrated that maximal resection substantially improves progression-free and overall survival, it has also been reported that aggressive glioma resection might increase the risk of postoperative complications and lead to worse survival prognosis. Therefore, the relationship between the extent of surgical resection and patient outcome still remains controversial. Even so, our results showed a positive correlation between gross-total resection and prognosis improvement for patients compared to partial resection or biopsy. In view of chemotherapy and radiotherapy, they are crucial elements in the treatment plan of glioma patients. Postoperative adjuvant radiotherapy and chemotherapy have always been recommended to start within 2–4 weeks after surgical resection and have proven to be significant prognostic factors by previous studies [32, 33] and this study. Tumor subtype is one the most commonly recognized prognostic factors, and the subtype based on the latest 2016 WHO criteria helps to predict patient prognosis more accurately. Here, our research also served as evidence of the critical role of tumor subtype in glioma management.
Then, it comes to biomarker information. MGMT is a DNA repair protein that removes alkyl groups and adducts at the O6 position of guanine, protecting the cell against mutagenic effects. Promoter methylation of MGMT causes silencing of the MGMT gene and loss of protein expression, accounting for the accumulation of DNA damage and increased sensitivity to temozolomide-based chemoradiotherapy. A prognostic effect of MGMT promoter methylation in patients with lower-grade [34] or higher-grade [35] glioma has already been observed. Located on human chromosome 17p13, the p53 gene is a tumor suppressor and has been detected to regulate apoptosis, inhibit DNA replication, and control cell motility and invasion. As a consequence of p53 gene mutation, the mutant p53 protein escapes from degradation and accumulates in the cells, leading to positive staining by IHC. A meta-analysis concluded that p53 immunopositivity has effective usefulness in analyzing the prognosis of glioma patients [36]. As for GFAP, Vimentin, and Ki-67, there exist a number of research studies (37–39) concentrating on their prognostic value. However, our analysis only validated the essential prognostic value of MGMT promoter methylation and p53 immunopositivity, and the other biomarkers need to be further evaluated.
There were several limitations. First, this was a single-center study, which might make the analysis potentially prone to bias and limit the generalization of supervised ML models. Second, the study included cases that occurred before 2016, where the glioma subtype classification at that time was different from the recent 2016 WHO criteria, causing half of those cases to lack evidence of subdivision (for instance, isocitrate dehydrogenase (IDH)) for the latter classification criteria. This might influence the calibration and discrimination of prediction models. Nevertheless, we believed our research has merit, given it is the first study to apply Gradient Boosting algorithms to glioma prognosis prediction. We had constructed predictive models successfully and also found that Gradient Boosting models were more likely to improve the performance of predicting glioma patient survival after tumor resection.
## 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 Sun Yat-sen University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
YL: conception and design of study, acquisition, analysis, and interpretation of data, and drafting the manuscript. MY: conception and design of the study and critically revising the manuscript for important intellectual content. BJ: analysis and interpretation of data and critically revising the manuscript for important intellectual content. LC: conception and design of the study and critically revising the manuscript for important intellectual content. ZZ: conception and design of the study and critically revising the manuscript for important intellectual content. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fsurg.2022.975022/full#supplementary-material.
## References
1. Ostrom QT. **CBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the United States in 2011–2015**. *Neuro Oncol* (2018) **20** iv1-86. DOI: 10.1093/neuonc/noy131
2. Cavaliere R, Lopes MB, Schiff D. **Low-grade gliomas: an update on pathology and therapy**. *Lancet Neurol* (2005) **4** 760-70. DOI: 10.1016/S1474-4422(05)70222-2
3. Gorovets D. **IDH mutation and neuroglial developmental features define clinically distinct subclasses of lower grade diffuse astrocytic glioma**. *Clin Cancer Res* (2012) **18** 2490-501. DOI: 10.1158/1078-0432.CCR-11-2977
4. Gittleman H, Sloan AE, Barnholtz-Sloan JS. **An independently validated survival nomogram for lower grade glioma**. *Neuro Oncol* (2020) **22** 665-74. DOI: 10.1093/neuonc/noz191
5. Zhang M. **Novel immune-related gene signature for risk stratification and prognosis of survival in lower-grade glioma**. *Front Genet* (2020) **11** 363. DOI: 10.3389/fgene.2020.00363
6. Han W. **Deep transfer learning and radiomics feature prediction of survival of patients with high-grade gliomas**. *AJNR Am J Neuroradiol* (2020) **41** 40-8. DOI: 10.3174/ajnr.A6365
7. Jakola AS. **Advancements in predicting outcomes in patients with glioma: a surgical perspective**. *Expert Rev Anticancer Ther* (2020) **20** 167-77. DOI: 10.1080/14737140.2020.1735367
8. Kourou K. **Machine learning applications in cancer prognosis and prediction**. *Comput Struct Biotechnol J* (2015) **13** 8-17. DOI: 10.1016/j.csbj.2014.11.005
9. Barakat NH, Bradley AP, Barakat MN. **Intelligible support vector machines for diagnosis of diabetes mellitus**. *IEEE Trans Inf Technol Biomed* (2010) **14** 1114-20. DOI: 10.1109/TITB.2009.2039485
10. Esteban C. **Development of a decision tree to assess the severity and prognosis of stable COPD**. *Eur Respir J* (2011) **38** 1294-300. DOI: 10.1183/09031936.00189010
11. Passos IC, Mwangi B, Kapczinski F. **Big data analytics and machine learning: 2015 and beyond**. *Lancet Psychiatry* (2016) **3** 13-5. DOI: 10.1016/S2215-0366(15)00549-0
12. Xu Y. **Supervised machine learning predictive analytics for triple-negative breast cancer death outcomes**. *Onco Targets Ther* (2019) **12** 9059-67. DOI: 10.2147/OTT.S223603
13. Xu Y. **Machine learning algorithms for predicting the recurrence of stage IV colorectal cancer after tumor resection**. *Sci Rep* (2020) **10** 2519. DOI: 10.1038/s41598-020-59115-y
14. Louis DN. **The 2016 World Health Organization classification of tumors of the central nervous system: a summary**. *Acta Neuropathol* (2016) **131** 803-20. DOI: 10.1007/s00401-016-1545-1
15. Okamoto Y. **Population-based study on incidence, survival rates, and genetic alterations of low-grade diffuse astrocytomas and oligodendrogliomas**. *Acta Neuropathol* (2004) **108** 49-56. DOI: 10.1007/s00401-004-0861-z
16. Hu W. **Expression of CPEB4 in human glioma and its correlations with prognosis**. *Medicine (Baltimore)* (2015) **94** e979. DOI: 10.1097/MD.0000000000000979
17. Takami H. **Revisiting TP53 mutations and immunohistochemistry—a comparative study in 157 diffuse gliomas**. *Brain Pathol* (2015) **25** 256-65. DOI: 10.1111/bpa.12173
18. Cai J. **ATRX mRNA expression combined with IDH1/2 mutational status and Ki-67 expression refines the molecular classification of astrocytic tumors: evidence from the whole transcriptome sequencing of 169 samples**. *Oncotarget* (2014) **5** 2551-61. DOI: 10.18632/oncotarget.1838
19. Fan J, Nunn ME, Su X. **Multivariate exponential survival trees and their application to tooth prognosis**. *Comput Stat Data Anal* (2009) **53** 1110-21. DOI: 10.1016/j.csda.2008.10.019
20. Taylor JM. **Random survival forests**. *J Thorac Oncol* (2011) **6** 1974-5. DOI: 10.1097/JTO.0b013e318233d835
21. Ishwaran H. **Random survival forests**. *Ann Appl Stat* (2008) **2** 841-60. DOI: 10.1214/08-AOAS169
22. Rashmi KV, Gilad-Bachrach R. **DART: dropouts meet multiple additive regression trees**. *AISTATS* (2015)
23. Capelle L. **Spontaneous and therapeutic prognostic factors in adult hemispheric world health organization grade II gliomas: a series of 1097 cases: clinical article**. *J Neurosurg* (2013) **118** 1157-68. DOI: 10.3171/2013.1.JNS121
24. Chang EF. **Preoperative prognostic classification system for hemispheric low-grade gliomas in adults**. *J Neurosurg* (2008) **109** 817-24. DOI: 10.3171/JNS/2008/109/11/0817
25. Corell A. **Age and surgical outcome of low-grade glioma in Sweden**. *Acta Neurol Scand* (2018) **138** 359-68. DOI: 10.1111/ane.12973
26. Sun Y. **Characteristics and prognostic factors of age-stratified high-grade intracranial glioma patients: a population-based analysis**. *Bosn J Basic Med Sci* (2019) **19** 375-83. DOI: 10.17305/bjbms.2019.4213
27. Paugh BS. **Integrated molecular genetic profiling of pediatric high-grade gliomas reveals key differences with the adult disease**. *J Clin Oncol* (2010) **28** 3061-8. DOI: 10.1200/JCO.2009.26.7252
28. Jairam V. **Defining an intermediate-risk group for low-grade glioma: a national cancer database analysis**. *Anticancer Res* (2019) **39** 2911-8. DOI: 10.21873/anticanres.13420
29. Leu S. **Preoperative two-dimensional size of glioblastoma is associated with patient survival**. *World Neurosurg* (2018) **115** e448-63. DOI: 10.1016/j.wneu.2018.04.067
30. Brown TJ. **Association of the extent of resection with survival in glioblastoma: a systematic review and meta-analysis**. *JAMA Oncol* (2016) **2** 1460-9. DOI: 10.1001/jamaoncol.2016.1373
31. Pan IW, Ferguson SD, Lam S. **Patient and treatment factors associated with survival among adult glioblastoma patients: a USA population-based study from 2000 to 2010**. *J Clin Neurosci* (2015) **22** 1575-81. DOI: 10.1016/j.jocn.2015.03.032
32. Barnholtz-Sloan JS. **Racial/ethnic differences in survival among elderly patients with a primary glioblastoma**. *J Neurooncol* (2007) **85** 171-80. DOI: 10.1007/s11060-007-9405-4
33. Aizer AA. **Underutilization of radiation therapy in patients with glioblastoma: predictive factors and outcomes**. *Cancer* (2014) **120** 238-43. DOI: 10.1002/cncr.28398
34. Bell EH. **Association of MGMT promoter methylation Status with survival outcomes in patients with high-risk glioma treated with radiotherapy and temozolomide: an analysis from the NRG oncology/RTOG 0424 trial**. *JAMA Oncol* (2018) **4** 1405-9. DOI: 10.1001/jamaoncol.2018.1977
35. Gilbert MR. **Dose-dense temozolomide for newly diagnosed glioblastoma: a randomized phase III clinical trial**. *J Clin Oncol* (2013) **31** 4085-91. DOI: 10.1200/JCO.2013.49.6968
36. Jin Y. **Expression and prognostic significance of p53 in glioma patients: a meta-analysis**. *Neurochem Res* (2016) **41** 1723-31. DOI: 10.1007/s11064-016-1888-y
37. Chen WJ. **Ki-67 is a valuable prognostic factor in gliomas: evidence from a systematic review and meta-analysis**. *Asian Pac J Cancer Prev* (2015) **16** 411-20. DOI: 10.7314/APJCP.2015.16.2.411
38. Schwab DE. **Immunohistochemical comparative analysis of GFAP, MAP-2, NOGO-A, OLIG-2 and WT-1 expression in WHO 2016 classified neuroepithelial tumours and their prognostic value**. *Pathol Res Pract* (2018) **214** 15-24. DOI: 10.1016/j.prp.2017.12.009
39. Lin L. **Analysis of expression and prognostic significance of vimentin and the response to temozolomide in glioma patients**. *Tumour Biol* (2016) **37** 15333-9. DOI: 10.1007/s13277-016-5462-7
|
---
title: 'Subgrouping patients with zoster-associated pain according to sensory symptom
profiles: A cluster analysis'
authors:
- Hee Jung Kim
- Kyung Bong Yoon
- Misun Kang
- Yun Seok Yang
- Shin Hyung Kim
journal: Frontiers in Neurology
year: 2023
pmcid: PMC9981999
doi: 10.3389/fneur.2023.1137453
license: CC BY 4.0
---
# Subgrouping patients with zoster-associated pain according to sensory symptom profiles: A cluster analysis
## Abstract
### Background and goal of study
Patients with zoster-associated pain exhibit a variety of sensory symptoms and forms of pain and complain of different pain patterns. The purpose of this study is to subgroup patients with zoster-associated pain who visited a hospital using painDETECT sensory symptom scores, analyze their respective characteristics and pain-related data, and compare similarities and differences among the groups.
### Materials and methods
The characteristics of 1,050 patients complaining of zoster-associated pain and pain-related data were reviewed retrospectively. To identify subgroups of patients with zoster-associated pain according to sensory symptom profiles, a hierarchical cluster analysis was performed based on the responses to a painDETECT questionnaire. Demographics and pain-related data were compared among all subgroups.
### Results and discussion
Patients with zoster-associated pain were classified into 5 subgroups according to the distribution of sensory profiles, with each subgroup exhibiting distinct differences in the expression of sensory symptoms. Patients in cluster 1 complained of burning sensations, allodynia, and thermal sensitivity, but felt numbness less strongly. Cluster 2 and 3 patients complained of burning sensations and electric shock-like pain, respectively. Cluster 4 patients complained of most sensory symptoms at similar intensities and reported relatively strong prickling pain. Cluster 5 patients suffered from both burning and shock-like pains. Patient ages and the prevalence of cardiovascular disease were significantly lower in cluster 1. Patients in clusters 1 and 4 reported longer pain duration compared with those in clusters 2 and 3. However, no significant differences were found with respect to sex, body mass index, diabetes mellitus, mental health problems, and sleep disturbance. Pain scores, distribution of dermatomes and gabapentinoid use were also similar among the groups.
### Conclusions
Five different subgroups of patients with zoster-associated pain were identified on the basis of sensory symptoms. A subgroup of younger patients with longer pain duration showed specific and distinct symptoms, such as burning sensations and allodynia. Unlike patients with acute or subacute pain, patients with chronic pain were associated with diverse sensory symptom profiles.
## Introduction
Cases of herpes zoster occur in at least 1 million people in the United States every year, and the incidence rate has been increasing worldwide over the past 20 years [1]. Approximately one-fifth of herpes zoster patients report some pain at 3 months after the onset of the disease, and the incidence rate of postherpetic neuralgia (PHN) increases dramatically as patients age [2, 3].
Patients with zoster-associated pain, including herpes zoster infection and PHN, complain of diverse patterns of sensory symptoms despite the common cause of the pain. Some patients complain strongly of spontaneous pain, tingling sensations, and electric-shock-like pain, while others suffer from hypersensitivity to light touch or temperature. A simple and validated patient-reported screening questionnaire, painDETECT, can detect these characteristic sensory abnormalities [4, 5]. Researchers have attributed this difference in pain patterns to the relative contributions of the peripheral and central mechanisms [6]. However, no specific biomarkers have been found with mechanisms of zoster-associated pain. Therefore, subgrouping of patients according to sensory phenotype can be clinically important in patients with zoster-associated pain. Treatments that focus on similar phenotypes would improve individualized pain therapy and future clinical trial design [7, 8]. Many investigators have expressed similar thoughts on this subject, and multiple studies have been conducted to identify the differences in the dynamics of the somatosensory system for some peripheral neuropathic pain conditions, including diabetic neuropathy and PHN (9–13). However, few studies have focused on zoster-associated pain from acute, subacute, and chronic pain conditions. We assumed that patients with zoster-associated pain who visit a hospital can be subgrouped into sensory profiles according to the somatosensory mechanism like other neuropathic pain.
The goal of this study is to classify patients with zoster-associated pain into groups that clearly represent each characteristic using a cluster analysis technique and to identify any similarities and differences in the characteristics of each cluster groups.
## Study population
This study was approved by the Institutional Review Board of Yonsei University Health System, Seoul, Republic of Korea (No. 4-2022-0819). As it is a retrospective observational study, the requirement for obtaining informed consent from the patients was waived. This manuscript complies with the STROBE checklists applicable to observational studies. Adult patients who complained of zoster-associated pain at their first visit from January 2016 to August 2021 were enrolled. Patients with incomplete medical records or those who did not complete the painDETECT questionnaire were excluded.
## Sensory symptom profiling using the painDETECT questionnaire
Sensory symptom profiles due to herpes zoster were described using the painDETECT questionnaire during the patient's first visit to our clinic. Patients complaining of neuropathic pain caused by herpes zoster freely described their symptoms and filled out the validated Korean version of the painDETECT questionnaire under the supervision of a researcher [14]. The questionnaire consisted of 9 questions asking for sensory symptoms, pain course patterns, and radiating pain. Seven of the questions asked respondents to use a 0-to-5 Likert scale to describe pain quality associated with various sensory symptoms: burning sensations, tingling or prickling sensations, pain by light touch, electric shock-like pain, pain on cold/heat sensation, numbness, and pain by slight pressure (0 = never; 1 = hardly noticed; 2 = slightly; 3 = moderately; 4 = strongly; 5 = very strongly). To eliminate differences in individual pain-perception thresholds, an alternative value was obtained by subtracting the average value of all scores for each of the 7 questions. In this alternative score, values above 0 represent stronger sensations than individual average pain perception, and values below 0 represent less-intense sensations compared with individual average pain perception [10].
## Patient demographics and pain-related data measures
Patient characteristics and pain-related data were gathered from electronic medical records. Patient characteristics included age, sex, body mass index (BMI), presence of diagnosed comorbidities, and sleep disturbances. Diagnosed comorbidities included cardiovascular disease, diabetes mellitus, and mental health problems that required continuous medical intervention with regular hospital visits. Mental health problems were defined as those requiring psychiatric treatment or medication due to depression, anxiety, and stress over the past year. Average and maximum pain scores using a numeric rating scale (NRS), duration of pain, dermatomes, and the use of a gabapentinoid at the time of visit were identified as pain-related factors. The duration of pain was classified into < 1 month (acute), 1–3 months (subacute), and ≥ 3 months (chronic).
## Descriptive statistical analysis
Continuous variables were shown as mean ± standard deviation (SD), and categorical variables was tabulated using numbers (percentage). The normality of distribution was evaluated using a Shapiro–Wilk test. Differences among the cluster groups were analyzed using one-way analysis of variance and chi-squared tests. A Tukey's post hoc test or a chi-squared post hoc difference was used to determine the intergroup difference between the mean of the groups when significant differences existed. Statistical Package for the Social Sciences, version 26.0 (IBM Corp, Armonk, NY, USA) was used to analyze the data. A $P \leq 0.05$ was considered statistically significant.
## Cluster analysis
A hierarchical cluster analysis was performed for subgrouping according to relevant sensory symptoms of zoster-associated pain (Figure 1). A hierarchical WARD approach with a squared Euclidean distance measure described previously [9, 10] was applied. Cluster analysis was conducted using the Python program, version 3.8.15 with Scikit-learn version 1.0.2. The cut-off point for the essential clusters was set at ~ $10\%$ of the total cases. We chose to refer to a cut-off point of 5 clusters because solutions with fewer clusters could eliminate significant differences by the agglomerative clustering. A solution with five essential clusters was obtained by hierarchical WARD clustering as an optimal agreement on our decision criteria. To demonstrate the evidence for the solution and to rearrange the cases based on these results, a k-means cluster analysis was performed. This analysis led to equal results, which provides support for our chosen solution.
**Figure 1:** *Flowchart for statistical methods.*
The cluster is represented by a pattern of questionnaire scores, indicating the typical pathological structure of each group. All profiles were given individually adjusted scores (see above). Heuristic interpretations of clusters were provided by experts. Statistical analysis was not performed because this was a heuristic approach.
## Results
During the study period, 1,434 patients visited our clinic for zoster-associated pain. Among these patients, 384 met the exclusion criteria, leaving 1,050 patients in the analysis (Figure 2).
**Figure 2:** *Study flowchart.*
Demographic profiles and pain-related data for these patients are shown in Table 1. Women accounted for more than $50\%$ of the patient group, the mean age was 63.86 years, and the average BMI value was 23.55 kg/m2. Cardiovascular disease, diabetes mellitus, and mental health problems were reported for 362, 207, and 166 patients, respectively, and more than $50\%$ of the patients complained of sleep disturbances. The mean average pain score was 5.57 and the maximum pain score was 6.29 on the NRS. A total of 302 patients were assessed at < 1 month after zoster infection, 315 patients had passed 1–3 months, and 433 patients complained of pain for more than 3 months. There were 201 patients with zoster in the trigeminal and facial nerve areas, 167 with zoster in the cervical dermatomes, 546 in the thoracic dermatomes, and 136 in the lumbosacral dermatomes. Approximately $80.6\%$ of patients were taking a gabapentinoid to relieve pain.
**Table 1**
| Variables | Total (n = 1,050) |
| --- | --- |
| Patient characteristics | Patient characteristics |
| Age, years | 63.86 ± 13.87 |
| Female sex | 616 (58.7) |
| Body mass index, kg/m2 | 23.55 ± 3.13 |
| Cardiovascular disease | 362 (34.5) |
| Diabetes mellitus | 207 (19.7) |
| Mental health problems | 166 (15.8) |
| Sleep disturbance | 563 (53.6) |
| Pain-related data | Pain-related data |
| Pain score, numeric rating scale 0–10 | |
| Average | 5.57 ± 2.44 |
| Maximum | 6.29 ± 2.41 |
| Pain duration | |
| < 1 month | 302 (28.8) |
| 1–3 months | 315 (30.0) |
| ≥ 3 months | 433 (41.2) |
| Dermatomes | |
| Trigeminal, facial | 201 (19.1) |
| Cervical | 167 (15.9) |
| Thoracic | 546 (52.0) |
| Lumbosacrum | 136 (13.0) |
| Gabapentinoid use | 846 (80.6) |
A cluster analysis was performed using the Korean version of painDETECT to identify the subgroups of patients showing typical sensory neuropathic symptoms as 5 distinct clusters (Figure 3). Patients in cluster 1 complained of all three symptoms: burning sensations, allodynia, and thermal sensitivity, although a lack of numbness was also reported. Patients in clusters 2 and 3 complained of severe burning sensation and electric-shock-like pain, respectively. Cluster 4 patients reported relatively strong tingling pain with an intensity similar to that of most sensory symptoms. Cluster 5 patients reported both burning and shock-like pains.
**Figure 3:** *Distribution of sensory symptom profiles of derived five cluster groups.*
A comparison was conducted to determine whether there was a difference in the characteristics and pain-related data of the patient groups between the 5 clusters (Table 2). Patient age (cluster 1 vs. 2, $$p \leq 0.001$$; cluster 1 vs. 3, $$p \leq 0.002$$; cluster 1 vs. 4, $$p \leq 0.006$$; cluster 1 vs. 5, $$p \leq 0.009$$) and the prevalence of cardiovascular disease (cluster 1 vs. 2, $p \leq 0.001$; cluster 1 vs. 3, $p \leq 0.001$; cluster 1 vs. 4, $$p \leq 0.005$$; cluster 1 vs. 5, $p \leq 0.001$) were significantly lower in cluster 1 than those in the other clusters. In addition, clusters 1 and 4 had a relatively high proportion of patients complaining of pain for more than 3 months (chronic) compared to clusters 2 and 3 (cluster 1 vs. 2, $$p \leq 0.013$$; cluster 1 vs. 3, $$p \leq 0.044$$; cluster 4 vs. 2, $$p \leq 0.007$$; cluster 4 vs. 3, $$p \leq 0.027$$). However, there were no significant differences in sex, BMI, medical comorbidities including diabetes mellitus and mental health problems, and sleep disturbance between clusters. Also, the average and maximum pain scores on the NRS, the dermatomes to which the lesion belongs, and the use of gabapentinoids were similar between cluster groups.
**Table 2**
| Cluster label | Cluster 1 (n = 171) | Cluster 2 (n = 228) | Cluster 3 (n = 296) | Cluster 4 (n = 203) | Cluster 5 (n = 152) |
| --- | --- | --- | --- | --- | --- |
| Patient characteristics | Patient characteristics | Patient characteristics | Patient characteristics | Patient characteristics | Patient characteristics |
| Age, years* | 60.25 ± 14.222, 3, 4, 5 | 65.37 ± 14.611 | 64.26 ± 13.12 1 | 64.25 ± 13.571 | 64.32 ± 13.661 |
| Female sex | 88 (51.5) | 132 (57.9) | 184 (62.2) | 114 (56.2) | 98 (64.5) |
| Body mass index, kg/m2 | 23.81 ± 2.86 | 23.52 ± 3.07 | 23.68 ± 3.37 | 23.69 ± 3.02 | 23.03 ± 3.01 |
| Cardiovascular disease** | 35 (20.5)2, 3, 4, 5 | 89 (39.0)1 | 110 (37.2)1 | 68 (33.5)1 | 60 (39.5)1 |
| Diabetes mellitus | 29 (17.0) | 53 (23.2) | 56 (18.9) | 38 (18.7) | 31 (20.4) |
| Mental health problems | 20 (11.7) | 40 (17.5) | 56 (18.9) | 22 (10.8) | 28 (18.4) |
| Sleep disturbance | 95 (55.6) | 116 (50.9) | 157 (53.0) | 107 (52.7) | 88 (57.9) |
| Pain-related data | Pain-related data | Pain-related data | Pain-related data | Pain-related data | Pain-related data |
| Pain score, numeric rating scale, 0–10 | Pain score, numeric rating scale, 0–10 | Pain score, numeric rating scale, 0–10 | Pain score, numeric rating scale, 0–10 | Pain score, numeric rating scale, 0–10 | Pain score, numeric rating scale, 0–10 |
| Average | 5.23 ± 2.54 | 5.46 ± 2.43 | 5.82 ± 2.39 | 5.51 ± 2.40 | 5.73 ± 2.42 |
| Maximum | 6.05 ± 2.55 | 6.35 ± 2.45 | 6.41 ± 2.32 | 6.06 ± 2.41 | 6.57 ± 2.35 |
| Pain durations | Pain durations | Pain durations | Pain durations | Pain durations | Pain durations |
| < 1 month | 49 (28.7) | 70 (30.7) | 99 (33.4) | 45 (22.2) | 39 (25.7) |
| 1-3 months | 41 (24.0) | 78 (34.2) | 85 (28.7) | 61 (30.0) | 50 (32.9) |
| ≥ 3 months** | 81 (47.4)2, 3 | 80 (35.1)1, 4 | 112 (37.8)1, 4 | 97 (47.8)2, 3 | 63 (41.4) |
| Dermatomes | Dermatomes | Dermatomes | Dermatomes | Dermatomes | Dermatomes |
| Trigeminal, facial | 39 (22.8) | 47 (20.6) | 55 (18.6) | 33 (16.3) | 27 (17.8) |
| Cervical | 29 (17.0) | 40 (17.5) | 35 (11.8) | 38 (18.7) | 25 (16.4) |
| Thoracic | 83 (48.5) | 115 (50.4) | 162 (54.7) | 108 (53.2) | 78 (51.3) |
| Lumbosacrum | 20 (11.7) | 26 (11.4) | 44 (14.9) | 24 (11.8) | 22 (14.5) |
| Gabapentinoid use | 141 (82.5) | 183 (80.3) | 227 (76.7) | 164 (80.8) | 131 (86.2) |
## Discussion
Patients with zoster-associated pain account for a large proportion of those with neuropathic-like pain symptoms. Despite an increased understanding of the pathophysiological mechanism of neuropathic pain, treatment of zoster-associated pain remains challenging and insufficiently effective. This study was designed to classify patients with zoster-associated pain based on abnormalities in sensory symptoms because it would be ideal and effective to stratify patients according to the pain mechanism [15].
Distinct sensory abnormalities associated with pain perceived by patients can be self-assessed, and they are largely divided into spontaneous sensory sensation (burning, tingling, and electric-shock-like pain), stimulus-evoked positive sensory symptoms (pain evoked by light touch, thermally evoked pain, and pressure-evoked pain), and stimulus-evoked negative sensory symptom (numbness) [7, 16]. The variations in the mechanisms of symptom generation are explained by plastic changes in the central nervous system and remaining peripheral nociceptors [6, 17].
An analysis of patients with zoster-associated pain on the basis of sensory phenotype found that patients in cluster 1 suffered primarily from thermally evoked pain and mechanical allodynia, but not numbness. Allodynia is due to central spinal cord sensitization and is induced by activation of touch-sensitive cutaneous Aβ-fibers that terminate in synapses of nociceptive second-order neurons in the central nervous system [18]. Thermally evoked pain is due to peripheral sensitization of nociceptive afferents; it is characteristic of hyperactive and sensitized cutaneous nociceptors [19]. However, numbness is a sensory symptom that involves a loss of afferent function [11]. Cluster 1 patients therefore have irritable nociceptors and their skin innervation is intact [20]. In studies of patients with peripheral neuropathic pain, it was possible to predict that thermally evoked pain and allodynia would have an analgesic effect by applying $8\%$ topical capsaicin patches [21] and intracutaneous botulinum toxin [22]. Patients in cluster 2, 3, and 5 suffered spontaneous pain without cutaneous hypersensitivity and marked sensory deafferentation, caused by ectopic neuronal firing within the nociceptive pathways and secondary sensitization of central nociceptive neurons [11]. Abnormal expression of voltage-gated sodium channels is associated with ectopic nerve activity [23]. An animal study has shown increases and changes in voltage-gated ion channels at impaired peripheral nerves in rats with varicellar zoster virus. These sensitized pain behaviors were reversed by sodium channel blockers [24]. Peripheral analgesic lidocaine patches selectively block sodium channels of small damaged peripheral nerve fibers without severe systemic adverse effects. Therefore, in this case, applying topical lidocaine is appropriate for the control of localized peripheral neuropathic pain [25]. Patients in cluster 4 tended to experience relatively strong tingling sensations, but were also characterized by relatively mild abnormalities due to the overlap of pathophysiological mechanisms. There are two possible causes of these symptom constellations. First, these patients perceived all neuropathic pain at similar frequencies and intensities. The second possibility is that these patients are people who tend to answer in a similar manner. These patients responded similarly to all questions because they could not discriminate between the abnormalities in sensory symptoms. Such patients should be excluded from the clinical trial of a medication that targets specific pathophysiologies.
In the present study, the patients belonging to cluster 1 were younger than those of the other clusters. Although results conflicted depending on the animal model and experimental stimuli, preclinical data suggest that advanced age may be related to increased pain sensitivity [26]. However, the relationship between age and pain sensitivity is unclear in zoster-related pain. Moreover, patients in cluster 1 reported experiencing pain for a longer time, although they were younger. Age may therefore not be an independent factor associated with specific sensory symptoms. In the current study, some differences in sensory symptoms according to duration of zoster-associated pain were also observed. In chronic pain conditions, the alteration of central processing of pain is a major contributing factor in pain symptoms [6]. While previous studies [11, 12] were conducted primarily on PHN patients, our study included all periods of pain, with more than half of patients reporting acute and subacute periods. Tissue inflammation and destruction, activation of nociceptive neurons, abnormal impulses, and neural injury during the early infection period cause a strong nociceptive barrage, creating acute pain and allodynia [27]. Considering the complexity of zoster-associated pain, this study confirms different therapeutic approaches are needed to manage pain according to their pain duration in this population.
This study has several limitations. As a retrospective study conducted in a single center, selective bias or information bias is possible. PainDETECT is a reliable questionnaire to assess differences in sensory perception intensity, but there is a paucity of interpretation of evoked pain and negative symptoms. Also, pain perceptions are subjective and could be confounded by personal factors [7]. To minimize these shortcomings, the scores of the questionnaire were adjusted by individual means. A longitudinal study would be necessary to proceed with clustering in an objective psychophysical approach (e.g., Quantitative sensory testing (QST) or morphological data from skin biopsies). In addition, we used an actual clinical practice model in which the majority of attending patients had undergone some interventional treatments or had been prescribed medications, including gabapentinoids. This class of medications modulates pain processing and central sensitization. However, because there was no difference in the ratio of gabapentinoid use for each subgroup, this factor would not have had a significant impact on the interpretation of the results.
## Conclusion
In conclusion, despite the common cause of the pain, some differences in the distribution of sensory profiles were evident among patients with zoster-associated pain. In particular, a subgroup with younger patients who experienced longer durations of pain reported specific and distinct positive symptoms, such as burning sensations and allodynia. Unlike patients with acute or subacute pain, patients with chronic pain were associated with diverse sensory symptom profiles. These results indicate that patients can be subgrouped according to patient-reported sensory symptoms of zoster-associated pain, although the central and peripheral mechanisms of pain vary among subjects. This approach may help determine the most appropriate pain treatment for each individual of this population.
## 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 Yonsei University Health System, Seoul, Republic of Korea. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
Conceptualization, writing—review and editing, and supervision: SHK and KBY. Software and validation: HJK and MK. Formal analysis and data curation: HJK and YSY. Writing—original draft preparation: HJK and SHK. Project administration and methodology: HJK. Funding acquisition: SHK. 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. Rosamilia LL. **Herpes zoster presentation, management, and prevention: a modern case-based review**. *Am J Clin Dermatol.* (2020) **21** 97-107. DOI: 10.1007/s40257-019-00483-1
2. Johnson RW, Rice AS. **Clinical practice**. *Postherpetic neuralgia N Engl J Med.* (2014) **371** 1526-33. DOI: 10.1056/NEJMcp1403062
3. Meera S, Aijun C. **Modalities in managing postherpetic neuralgia**. *Korean J Pain.* (2018) **31** 235-43. DOI: 10.3344/kjp.2018.31.4.235
4. Freynhagen R, Baron R, Gockel U, Tölle TR. **Pain detect: a new screening questionnaire to identify neuropathic components in patients with back pain**. *Curr Med Res Opin.* (2006) **22** 1911-20. DOI: 10.1185/030079906X132488
5. Vollert J, Kramer M, Barroso A, Freynhagen R, Haanpää M, Hansson P. **Symptom profiles in the pain detect questionnaire in patients with peripheral neuropathic pain stratified according to sensory loss in quantitative sensory testing**. *Pain.* (2016) **157** 1810-8. DOI: 10.1097/j.pain.0000000000000588
6. Pappagallo M, Oaklander AL, Quatrano-Piacentini AL, Clark MR, Raja SN. **Heterogenous patterns of sensory dysfunction in postherpetic neuralgia suggest multiple pathophysiologic mechanisms**. *Anesthesiology.* (2000) **92** 691-8. DOI: 10.1097/00000542-200003000-00013
7. Baron R, Förster M, Binder A. **Subgrouping of patients with neuropathic pain according to pain-related sensory abnormalities: a first step to a stratified treatment approach**. *Lancet Neurol.* (2012) **11** 999-1005. DOI: 10.1016/S1474-4422(12)70189-8
8. Reimer M, Helfert SM, Baron R. **Phenotyping neuropathic pain patients: implications for individual therapy and clinical trials**. *Curr Opin Support Palliat Care.* (2014) **8** 124-9. DOI: 10.1097/SPC.0000000000000045
9. Koroschetz J, Rehm SE, Gockel U, Brosz M, Freynhagen R, Tölle TR. **Fibromyalgia and neuropathic pain–differences and similarities. A comparison of 3057 patients with diabetic painful neuropathy and fibromyalgia**. *BMC Neurol.* (2011) **11** 55. DOI: 10.1186/1471-2377-11-55
10. Baron R, Tölle TR, Gockel U, Brosz M, Freynhagen R. **cross-sectional cohort survey in 2100 patients with painful diabetic neuropathy and postherpetic neuralgia: differences in demographic data and sensory symptoms**. *Pain.* (2009) **146** 34-40. DOI: 10.1016/j.pain.2009.06.001
11. Rehm S, Grobkopf M, Kabelitz M, Keller T, Freynhagen R, Tölle TR. **Sensory symptom profiles differ between trigeminal and thoracolumbar postherpetic neuralgia**. *Pain Rep.* (2018) **3** e636. DOI: 10.1097/PR9.0000000000000636
12. Gierthmühlen J, Braig O, Rehm S, Hellriegel J, Binder A, Baron R. **Dynamic of the somatosensory system in postherpetic neuralgia**. *Pain Rep.* (2018) **3** e668. DOI: 10.1097/PR9.0000000000000668
13. Choi K, Kwon O, Suh BC, Sohn E, Joo IS, Oh J. **Subgrouping of peripheral neuropathic pain patients according to sensory symptom profile using the Korean version of the pain detect questionnaire**. *J Korean Med Sci.* (2022) **37** e8. DOI: 10.3346/jkms.2022.37.e8
14. Sung JK, Choi JH, Jeong J, Kim WJ, Lee DJ, Lee SC. **Korean version of the pain detect questionnaire: a study for cultural adaptation and validation**. *Pain Pract.* (2017) **17** 494-504. DOI: 10.1111/papr.12472
15. Forstenpointner J, Otto J, Baron R. **Individualized neuropathic pain therapy based on phenotyping: are we there yet?**. *Pain.* (2018) **159** 569-75. DOI: 10.1097/j.pain.0000000000001088
16. Cappelleri JC, Koduru V, Bienen EJ, Sadosky A. **Characterizing neuropathic pain profiles: enriching interpretation of pain detect**. *Patient Relat Outcome Meas.* (2016) **7** 93-9. DOI: 10.2147/PROM.S101892
17. Bannister K, Sachau J, Baron R, Dickenson AH. **Neuropathic pain: mechanism-based therapeutics**. *Annu Rev Pharmacol Toxicol.* (2020) **60** 257-74. DOI: 10.1146/annurev-pharmtox-010818-021524
18. Jensen TS, Finnerup NB. **Allodynia and hyperalgesia in neuropathic pain: clinical manifestations and mechanisms**. *Lancet Neurol.* (2014) **13** 924-35. DOI: 10.1016/S1474-4422(14)70102-4
19. McKemy DD, Neuhausser WM, Julius D. **Identification of a cold receptor reveals a general role for TRP channels in thermosensation**. *Nature.* (2002) **416** 52-8. DOI: 10.1038/nature719
20. Mahn F, Hüllemann P, Gockel U, Brosz M, Freynhagen R, Tölle TR. **Sensory symptom profiles and co-morbidities in painful radiculopathy**. *PLoS ONE.* (2011) **6** e18018. DOI: 10.1371/journal.pone.0018018
21. Mainka T, Malewicz NM, Baron R, Enax-Krumova EK, Treede RD, Maier C. **Presence of hyperalgesia predicts analgesic efficacy of topically applied capsaicin 8% in patients with peripheral neuropathic pain**. *Eur J Pain.* (2016) **20** 116-29. DOI: 10.1002/ejp.703
22. Attal N, de Andrade DC, Adam F, Ranoux D, Teixeira MJ, Galhardoni R. **Safety and efficacy of repeated injections of botulinum toxin A in peripheral neuropathic pain (BOTNEP): a randomised, double-blind, placebo-controlled trial**. *Lancet Neurol.* (2016) **15** 555-65. DOI: 10.1016/S1474-4422(16)00017-X
23. Baron R, Binder A, Wasner G. **Neuropathic pain: diagnosis, pathophysiological mechanisms, and treatment**. *Lancet Neurol.* (2010) **9** 807-19. DOI: 10.1016/S1474-4422(10)70143-5
24. Garry EM, Delaney A, Anderson HA, Sirinathsinghji EC, Clapp RH, Martin WJ. **Varicella zoster virus induces neuropathic changes in rat dorsal root ganglia and behavioral reflex sensitisation that is attenuated by gabapentin or sodium channel blocking drugs**. *Pain.* (2005) **118** 97-111. DOI: 10.1016/j.pain.2005.08.003
25. Wolff RF, Bala MM, Westwood M, Kessels AG, Kleijnen J. **5% lidocaine-medicated plaster vs. other relevant interventions and placebo for post-herpetic neuralgia (PHN): a systematic review**. *Acta Neurol Scand.* (2011) **123** 295-309. DOI: 10.1111/j.1600-0404.2010.01433.x
26. Yezierski RP. **The effects of age on pain sensitivity: preclinical studies**. *Pain Med.* (2012) **13** S27-36. DOI: 10.1111/j.1526-4637.2011.01311.x
27. Rowbotham MC, Petersen KL. **Zoster-associated pain and neural dysfunction**. *Pain.* (2001) **93** 1-5. DOI: 10.1016/S0304-3959(01)00328-1
|
---
title: 'Molecular mechanism of Cuscutae semen–radix rehmanniae praeparata in relieving
reproductive injury of male rats induced with tripterygium wilfordii multiglycosides:
A tandem mass tag-based proteomics analysis'
authors:
- Shanshan Han
- Yanlin Dai
- Lihui Sun
- Yaping Xing
- Ying Ding
- Xia Zhang
- Shanshan Xu
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9982038
doi: 10.3389/fphar.2023.1050907
license: CC BY 4.0
---
# Molecular mechanism of Cuscutae semen–radix rehmanniae praeparata in relieving reproductive injury of male rats induced with tripterygium wilfordii multiglycosides: A tandem mass tag-based proteomics analysis
## Abstract
Background: We determined the effects of Cuscutae semen (*Cuscuta chinensis* Lam. or *Cuscuta australis* R. Br.)–*Radix rehmanniae* praeparata (*Rehjnannia glutinosa* Libosch.) on the protein levels in testicular tissues of rats gavaged with tripterygium wilfordii multiglycosides (GTW) and elucidated the molecular mechanism underlying Cuscutae semen–*Radix rehmanniae* praeparata for relieving GTW-induced reproductive injury.
Methods: A total of 21 male Sprague–Dawley rats were randomly divided into the control group, model group, and Cuscutae semen–*Radix rehmanniae* praeparata group based on their body weights. The control group was given 10 mLkg−1 of $0.9\%$ normal saline by gavage daily. The model group (GTW group) was administered with 12 mg kg-1 GTW by gavage daily. Cuscutae semen–*Radix rehmanniae* praeparata group (the TSZSDH group) was administered with 1.56 gkg−1 of Cuscutae semen–*Radix rehmanniae* praeparata granules daily according to their model group dosing. The serum levels of luteinizing hormone, follicle-stimulating hormone, estradiol, and testosterone were measured after 12 weeks of continuous gavage, and the pathological lesion of testicular tissues was observed. Differentially expressed proteins were evaluated by quantitative proteomics and verified by western blotting (WB) and Real-Time Quantitative Polymerase Chain Reaction (RT-qPCR).
Results: Cuscutae semen–*Radix rehmanniae* praeparata can effectively relieve pathological lesions of GTW-induced testicular tissues. A total of 216 differentially expressed proteins were identified in the TSZSDH group and model group. High-throughput proteomics revealed that differentially expressed proteins are closely associated with the peroxisome proliferator-activated receptor (PPAR) signaling pathway, protein digestion and absorption, and protein glycan pathway in cancer. Cuscutae semen–*Radix rehmanniae* praeparata can significantly upregulate the protein expressions of Acsl1, Plin1, Dbil5, Plin4, Col12a1, Col1a1, Col5a3, Col1a2, Dcn, so as to play a protective role on testicular tissues. Acsl1, Plin1, and PPARγ on the PPAR signaling pathway were verified by WB and RT-qPCR experiments, which were found to be consistent with the results of proteomics analysis.
Conclusion: Cuscutae semen and *Radix rehmanniae* praeparata may regulate the PPAR signaling pathway mediated Acsl1, Plin1 and PPARγ to reduce the testicular tissue damage of male rats caused by GTW.
## 1 Introduction
Tripterygium wilfordii multiglycosides (GTW) is a metabolite of the botanical drug *Tripterygium wilfordii* Hook. f. (TW), a non-steroidal immunosuppressant. It is commonly used in Chinese traditional medicine and is known as the “Chinese herbal hormone.” It has anti-inflammatory, antitumor, and immunoregulatory effects. Therefore, GTW has been widely used for the treatment of immunological diseases, such as rheumatoid arthritis, systemic lupus erythematosus, and nephrotic syndrome. However, its reproductive toxicity has become an important factor that has limited its clinical application. Testicular diseases such as oligospermia, asthenospermia, and azoospermia occur in men, resulting in decreased fertility or even infertility (Yang. et al., 2019). Therefore, it is very important to find an effective drug that can antagonize the reproductive toxicity of GTW, so that the clinical application can exert the maximum effect without worrying about the damage to its reproductive system, and realize the “freedom of medication” for reducing toxicity and increasing efficacy.
Kidney-tonifying methods and kidney-tonifying drugs can antagonize GTW reproductive toxicity through different pathways (Sheng et al., 2017). Cuscutae semen is the dry and mature seed of *Cuscuta chinensis* Lam. or *Cuscuta australis* R. Br., which is the plant of the family Convolvulaceae. It was first recorded in Shennong’s classic of botanical drugs. It can nourish the liver and kidney, consolidate the essence, reduce urine, calm the fetus, brighten the eyes, and relieve diarrhea. Radix rehmanniae praeparata is prepared and processed product of *Rehjnannia glutinosa* Libosch. ’s root tuber, which is the plant of the family Scrophulariaceae. It is an important medicine for nourishing the kidney and replenishing the blood. It is slightly warm in nature and sweet in flavor. It belongs to the liver and kidney channels and has the effect of nourishing the yin, blood, and bone marrow. Between them, one belongs to yin and another belongs to yang. They complement each other with hardness and softness and adjust the balance of yin and yang. They are Traditional Chinese Medicine (TCM) pairs in the classical kidney-tonifying formula preferred by the ancients. For example, the “Cuscutae semen *Radix rehmanniae* praeparata decoction” in the Syndrome Differentiation record has the effect of tonifying the kidney and strengthening the yang. It is used to treat kidney injury caused by excessive sexual intercourse, impotence, and premature ejaculation. The “Zuogui Pill” in Jingyue Quanshu can nourish yin and kidney, nourish the bone marrow, and treat spermatorrhea. Modern pharmacological studies have found that (Wang et al., 2021) metabolites of Cuscutae semen contain flavone, polysaccharides, alkaloids, steroids, terpenes, volatile oils, and lignans. The effective metabolites of *Radix rehmanniae* praeparata are mainly sugars, sitosterol, and nucleosides. They have anti-aging, anti-oxidation, and anti-apoptotic properties. Cuscutae semen can significantly improve sperm capillary penetration ability, sperm forward movement speed, and sperm activity index (Wang et al., 2021). Flavone, the metabolite of Cuscutae semen, can reduce the apoptosis of spermatogenic cells and increase the weights of the testis and epididymis (Meng et al., 2020). Moreover, the total flavone of Cuscutae semen can inhibit the apoptosis of testicular cells in rats and prevent oxidative damage in testicular cells. Therefore, it is generally a good curative effect for the treatment of male reproductive diseases caused by active oxygen free radicals (Zhen et al., 2006). Additionally, Yang et al, [2017] reported that Cuscutae semen polysaccharides can tonify the kidneys and support the yang. It can increase testosterone and estradiol levels, reduce blood urea nitrogen levels, improve immune function, and antioxidant effect. Zhang et al, [2012] reported that *Radix rehmanniae* praeparata can downregulate the GTW-induced reproductive-related spermatogenic cells gene, c-jun. The expression of Wnt4 is abnormal, which reduces GTW reproductive toxicity. During the early stage, we have confirmed that Cuscutae semen–*Radix rehmanniae* praeparata (the TSZSDH group) can not only interfere with GTW-induced injury by regulating the spermatogenic cells cycle, apoptosis, and related proteins but also has certain anti-inflammatory activity, which can effectively promote the anti-oxidation and delay cell senescence. Previous studies have shown that they can protect reproductive organs, improve sperm quality, promote testicular development, and inhibit cell apoptosis.
In this study, we showed that Cuscutae semen–*Radix rehmanniae* praeparata can reduce GTW-induced pathological lesions of the testicles. We further used high-throughput proteomics to investigate the ability of TSZSDH to improve the molecular mechanism of GTW-induced reproductive injury and analyzed differentially expressed proteins in testicular tissues of different groups. Then, the protein–protein interaction network and the genes and pathways affected by the treatment of TSZSDH were analyzed in the model group. Additionally, key pathway proteins were verified and clarified the therapeutic mechanism of TSZSDH in GTW reproductive toxicity.
## 2.1 Reagents
Cuscutae semen and *Radix rehmanniae* praeparata were provided by Jiangyin Tianjiang Pharmaceutical Factory, China (0.5 g/bag, equivalent to 10 g of crude drugs; Batch No: 21016424, 21016134); GTW was purchased from Jiangsu Meitong Pharmaceutical (Batch No: 210101); ELISA kits for FSH, LH, E2, and T were purchased from Wuhan Elabscience Biotechnology Co., Ltd.; PPAR γ antibody and rabbit antibody were purchased from Immunoway; Plin1 antibody was purchased from Affinity Biosciences; Acsl1 antibody was purchased from Cell Signaling Technology (CST). The quality control for all the materials was validated according to the Chinese Pharmacopoeia. Our entrusted company, Jiangyin Tianjiang Pharmaceutical Co., Ltd., carried out UPLC-Q-TOF/MS analysis to characterize the Cuscutae semen, *Radix rehmanniae* praeparata, and GTW extract (Figures 1A–C). The chemical constituents of Cuscutae semen, *Radix rehmanniae* praeparata, and GTW extract were profiled using an ultra-high-performance liquid chromatography system coupled with a high-resolution mode.
**FIGURE 1:** *(A) Analysis of Prepared Cuscutae semen Dispensing Granules by Mass Spectrometry. Peak 1:Neochlorogenic acid; Peak 2:Chlorogenic Acid; Peak 3:Quercetin 3-O- β- D-galactosyl-7-O-β- D-glucoside; Peak 4:Cryptochlorogenic acid; Peak 5:Rutin; Peak 6:Quercetin-3-O-apiose - (1→2) - galactoside; Peak 7:Quercetin 3-O-β- D-galactoside (2→1)-β- D-glucoside; Peak 8:Hyperoside; Peak 9:Isoquercitrin; Peak 10:Astragalin; Peak 11:Isorhamnetin 7-O-β-D-glucopyranoside. (B) Analysis of Prepared Radix rehmanniae praeparata Dispensing Granules by Mass Spectrometry. Peak 1: catalpol; Peak 2: 5-hydroxymethylfurfural diglucoside; Peak 3: Rehmannin D; Peak 4: 5-hydroxymethylfurfural; Peak 5: Melittoside; Peak 6: Geniposidic acid; Peak 7: adenosine; Peak 8: Ajugol; Peak 9: 8-Epi-Loganic acid; Peak 10: Verbasoside; Peak 11: aucubin; Peak 12: rehmapicrogenin; Peak 13: Kanokoside A; Peak 14: Purpureaside C; Peak 15: Rehmaionoside B; Peak 16: Jionoside A1; Peak 17: Rehmaionoside A; Peak 18: Verbascoside; Peak 19: Jionoside B1; Peak 20: Isoacteoside; Peak 21: Cistanoside C; Peak 22: Jionoside D; Peak 23: Martynoside; Peak 24: Isomartynoside. (C) Analysis of Prepared GTW by Mass Spectrometry. Peak 1:Celacinnine; Peak 2:unknown; Peak 3:TripfordineA; Peak 4:unknown; Peak 5:Wilfordlongin; Peak 6:Wilforjing; Peak 7:A1atusinine; Peak 8:2-O-deacetyl-euonine; Peak 9:wilfordinine B; Peak 10:hypoglaunine E; Peak 11:1-desacetylwilfortrine; Peak 12:Tripterifordin; Peak 13:Wilfordine E; Peak 14:Wilfordinine A; Peak 15:Wilfortrine; Peak 16:Triptonoterpenol; Peak 17:wilfortrine; Peak 18:Peritassine A; Peak 19:Wilfordeonine; Peak 20:wilfordinine; Peak 21:Wilfordiuetong; Peak 22:Triptoquinone B; Peak 23:Isowilfordine; Peak 24:wilfomine A; Peak 25:9″-O-acetylwilforlrine; Peak 26:Wilforgine; Peak 27:Wilformine; Peak 28:Wilforzine; Peak 29:hypoglaunine C; Peak 30:Triptonoditerpenic acid; Peak 31:Wilfomine A; Peak 32:Wilforine; Peak 33:9″-O-acetylwilforlrine isomer; Peak 34:Triptonine B; Peak 35:6a-hydroxytriptocalline A; Peak 36:Wilforine; Peak 37:9″-O-acetylwilforlrine isomer; Peak 38:cangoronine E; Peak 39:Tripterygiumine B; Peak 40:orlhosphenic acid; Peak 41:Ebenifoline E-II; Peak 42:Celastrol; Peak 43:Wilforlide B; Peak 44:triptoquinone B.*
## 2.2 Animals
Twenty-one 4-week-old male Sprague–Dawley rats weighing 51–75 g were purchased from Beijing Weitong Lihua Co., Ltd. [License: SCXK (Beijing) 2016–0011]. The experimental protocol was approved by the Animal Experiment Ethics Committee of the Henan University of Chinese Medicine (No: DWLL202105053). The male rats were randomly assigned to the normal saline group (the control group), the GTW group (the model group), and GTW and Cuscutae semen–*Radix rehmanniae* praeparata treatment group (the TSZSDH group), with seven rats in each group fed according to the 12-h light/dark cycle.
## 2.3 Medicine dose and model preparation
Rats in each group were fed the same diet and administered by gavage once every morning. The control group was fed with 10 mL kg-1·d-1 of $0.9\%$ normal saline via gavage. The model group was fed with a solution of 1 mL normal saline containing 1.2 mg GTW, administered via gavage at the dose of 10 mL kg-1·d-1 (i.e., 12 mg kg-1·d-1). The TSZSDH group: based on the model group, this group was fed with 1 mL of normal saline containing 0.156 g of Cuscutae semen and 0.156 g of *Radix rehmanniae* praeparata granules, administered via gavage at the dose of 10 mL kg-1·d-1. The rats were fed continuously for 12 weeks and weighed weekly. After fasting for 12 h, three groups of rats were anesthetized with pentobarbital and their blood was sampled from the abdominal aorta. The serum was collected, and the level of sex hormones was detected by using ELISA kits. The left epididymis was collected to detect the density and viability of sperm. The left testicular tissues were stained with hematoxylin and eosin to observe the pathological changes. The right testicular tissues were frozen at −80°C in the refrigerator for proteomic analysis.
## 2.4 Polypeptide sample preparation and TMT labeling
The right testicular tissues of each rat were stored at −80°C and accurately weighed to 50 mg, followed by the addition of the lysate ($1.5\%$ SDS/100 mM Tris-Cl, pH 8.0). After tissue homogenization in a boiling water bath, the tissues were sonicated on an ice water bath for 10 min and then centrifuged several times. The supernatant was collected and the protein in the solution was precipitated via acetone precipitation. Then, the complex solution, dithiothreitol (DTT), and iodoacetamide (IAA) were added to the protein precipitate until the final concentration of 40 mM was obtained, and the alkylation reaction was triggered at room temperature in the dark. The protein concentration was determined by using the Bradford method. After protein quantitative, 50 μg of the samples were taken in EP tubes for SDS-PAGE detection, and the protein bands were observed after Coomassie brilliant blue staining. To this, 100 mM of Tris-HCl solution (pH 8.0) was added to the reduced and alkylated samples, the urea concentration was diluted to <2 mM, and trypsase was added for digestion according to the mass ratio of enzyme to the protein of 1:50. An equal amount of samples was taken for TMT labeling. The marking operation was conducted in accordance with the instructions of the TMT manufacturer. After the labeled samples were mixed in an equal amount, they were desalted using the Sep-Pak C18. After vacuum drying, the mixed samples were separated by high pH reverse chromatography and finally combined into 15 components. The labeled samples were analyzed by high-resolution Orbitrap LC-MS/MS before data analyses to determine the relative abundance of the peptides. Then, 100 μg of the peptides were taken for each sample.
## 2.5 Western blotting analysis
The testicular tissues (80 mg) were extracted and crushed, followed by rinsing twice with PBS. The tissues were then homogenized and lysed with RIPA buffer containing a mixture of protein enzyme inhibitors. The protein was centrifuged at 12000 rpm for 10 min, and the protein concentration was determined by using BCA kits. Then, about 60 μg of the protein sample was extracted and separated by SDS-PAGE gel and transferred onto the PVDF membrane. Then, $5\%$ skim milk powder was added for blocking for 1 h and then incubated with the primary antibody at 4°C overnight. After rinsing with TBST thrice (5 min each time), the membrane was incubated with secondary antibodies (anti-rabbit IgG 1:10000) for 30 min. Finally, the ECL solution was added for development.
## 2.6 Real-time quantitative polymerase chain reaction
Total RNA was obtained using TRIzol reagent (Invitrogen, U.S.A.). cDNA was obtained by reverse transcription using a Reverse Transcription Kit (TOYOBO, China) and then was amplified by RT-qPCR using 2X universal SYBR Green Fast qPCR Mix (ABclonal, China). In this study, β-actin was used as the housekeeping gene and 2−ΔΔCT was used to calculate the relative change in gene expression. The primer sequences used to amplify PPARγ, Acsl1, Plin1 and β-actin are listed in Supplementary Table S1.
## 2.7 Statistical analysis
SPSS 26.0 software was used to statistically process the data, and the results were expressed as x¯±S. Based on whether the data conformed to the normality test, a non-parametric test was adopted. The research data involved the mean comparison among multiple groups, and one-way ANOVA was applied for further pairwise comparison. $p \leq 0.05$ was considered to indicate a statistically significant difference. The differentially expressed proteins between the model and control groups were called DEPs1 and those between the TSZSDH and model groups were called DEPs2. The screening conditions for the differentially expressed proteins were as follows: the difference fold change >1.2 times (upregulation >1.2 times or downregulation <0.83) and $p \leq 0.05.$
Metascape (https://Metascape. Org/gp/index. Ht mL) and eggnog were applied for Gene Ontology (GO) enrichment analysis and gene: Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Bioinformatics online website, Cytoscape3.8.1, Origin 2021, and GraphPad Prism nine software were used for graphic display. p-value score and gene ratio were adopted to select the important GO enrichment analysis, KEGG pathway, and COG annotation.
## 3.1 Effects of Cuscutae semen–radix rehmanniae praeparata on the testis and epididymis indexes of rats in the GTW group
The rats in the control group, model group, and TSZSDH group had the same body weight before administration. Compared with the control group, the weight of the model group rat did not change significantly after 12 weeks; compared with the model group, rats in the TSZSDH group exhibited increased body weights, while the differences were no statistical significance (Supplementary Table S2). Compared with the control group, rats in the model group showed decreased testis and epididymis indexes, but the differences were not statistical significance compared with the model group. Rats in the TSZSDH group showed increased testis and epididymis indexes, but the differences were not statistical significance (Supplementary Table S2).
## 3.2 Effects of Cuscutae semen–radix rehmanniae praeparata on serum level of sex hormone of rats in the GTW group
After 12 weeks of treatment, only T changed significantly in the three groups, whereas LH, FSH, and E2 had no statistical difference. Compared with those in the control group, rats in the model group showed decreased LH, FSH, E2, and T, but the differences were not statistical significance compared with the model group. Rats in the TSZSDH group rat showed decreased LH and increased FSH, E 2, and T. The differences in T were statistically significant and the differences in FSH and E2 were not statistical significance (Supplementary Table S3).
## 3.3 Effects of Cuscutae semen–radix rehmanniae praeparata on testis and epididymis histopathology of the GTW group
The seminiferous tubule is the main component of the testis and the place organ where sperm is generated. Therefore, the diameter of the seminiferous tubule of the three groups of rats was measured and the number of seminiferous tubules was counted in their area. Compared with the control group, the area of the seminiferous tubule of the model group was shorter, but the differences were not statistical significance. Compared with the model group, the area of the seminiferous tubule of the TSZSDH group was shorter, but the differences were not statistical significance.
The histopathological changes of testis and epididymis of the rats in the three groups were observed by hematoxylin-eosin staining to determine the protective effect of Cuscutae semen–*Radix rehmanniae* praeparata on the damage of the reproductive organs in the GTW group. The diameter of the seminiferous tubule of the rats in the model group was smaller, the space between tubes was larger, the seminiferous epithelium was thinner, the number of spermatogenic cells at all levels was significantly reduced and scattered, and the spermatogenic cells appeared vacuolization, the Leydig cells were reduced, and the number of sperms in the lumen of the epididymis tube was decreased. The epithelia of the seminiferous tubule in rats in the TSZSDH group were slightly thinner, the cells were distinct, and there was no obvious change in the intercellular space. The number of spermatogenic cells in all levels of the TSZSDH group was significantly higher than that of the model group. The intercellular space was slightly larger, the change of interstitial cells was not distinct, and the sperms in the epididymis were relatively dense (Figure 2A).
**FIGURE 2:** *(A) Testicular and epididymal tissues were stained with hematoxylin and eosin (HE) and observed under a light microscope at ×100 and ×400 magnifications. Control group (Control); Tripterygium wilfordii multiglycosides-induced reproductive injury model group (Model); Cuscutae semen–Rehmannia Glutinosa treatment group (TSZSDH). (B) Results of Tandem Mass Tag (TMT) protein labeling experiments. The green part represents the results of differential protein labeling between the model and control groups; the red part represents the result of the differential protein labeling between the TSZSDH and model groups, and the crossed part in yellow represents the overlapping portion. (C) Clustering heat map comparing the model group with the control group, and the TSZSDH group with the model group. Upregulated protein expression is indicated in red, while downregulated protein expression is indicated in blue. (D) Volcano plot comparing the model and control groups, and the TSZSDH and model groups. The log2fold-change (FC) value is the abscissa and the -Log10(p-value) is the ordinate. UP Red indicates that differentially expressed proteins are upregulated, while blue indicates downregulated proteins, and green indicates proteins whose expression was not significantly changed.*
## 3.4 Effects of Cuscutae semen–radix rehmanniae praeparata on the protein expressions in testicular tissues of rats in the GTW group
A total of three testicular samples from each group were used for proteomics analysis. Through PCA analysis, indicating that the samples in each group had good repeatability, and the samples between groups had significant differences (Figure 3). Among them, the control group is mainly located in the second and fourth quadrants, the GTW group is mainly located in the first and third quadrants, and the TSZSDH group is similar to the control group, mainly located in the second and fourth quadrants. PCA results showed that the samples of rats in the control group and the GTW group were obviously separated, indicating that the samples of rats in the GTW group were significantly changed, and the samples of rats in the TSZSDH group and the GTW group were obviously separated, similar to the control group, indicating that the rats in the TSZSDH group were significantly improved after the intervention of TSZSDH. A total of 7381 proteins were identified in three groups. The difference fold change >1.2 times (upregulation >1.2 times or downregulation <0.83) and $p \leq 0.05$ were used as the screening criteria for differentially expressed proteins. The number of differentially expressed proteins between groups was compared. Compared with the control group, the model group had 34 statistically significant differentially expressed proteins, which included four upregulated proteins and 30 downregulated proteins. Compared with the model group, the TSZSDH group was treated with Cuscutae semen–*Radix rehmanniae* praeparata. A total of 216 differentially expressed proteins with statistical significance were present, which included 172 upregulated proteins and 44 downregulated proteins. The statistics of protein quantitative results are shown in the form of a volcano map and cluster heat map as shown in figures (Figures 2C,D).
**FIGURE 3:** *(A) Testicular PCA scores of male rats between the control and model groups. (B) Testicular PCA scores of male rats in the treatment and model groups. The letter A represents the control group, the letter B represents the model group, and the letter C represents the TSZSDH group.*
## 3.5 GO term enrichment, KEGG pathway enrichment analysis, and orthologous clustering (COG) annotation analysis
The DEPs one of the model group and the control group and DEPs two of the TSZSDH group and the model group were subjected to GO term enrichment analysis, which included biological processes, cell components, and molecular functions. A total of 36 GO enrichment functions (bubble chart) were present between DEPs one and DEPs 2, among which 9 were pathways with $p \leq 0.05.$ Compared with the control group, the differentially expressed proteins of the model group were significantly regulated ($p \leq 0.05$) in response to stimulus ($11\%$), multicellular organismal process ($11\%$), developmental process ($9\%$), signaling ($7\%$), and structural molecule activity ($12\%$), in molecular functions while cellular anatomical entity ($46\%$) and intracellular ($41\%$) were regulated in cellular components and protein-containing complex ($13\%$), but the differences were no statistical significance. Compared with the model group, the differentially expressed proteins of the TSZSDH group were significantly regulated in response to stimulus ($11\%$), which belongs to biological processes, multicellular organismal process ($9\%$), immune system process ($4\%$), multi organism process ($4\%$), reproductive process ($3\%$), reproduction ($3\%$), and structural molecule activity ($9\%$) in molecular functions, whereas cellular anatomical entity ($46\%$) and intracellular ($41\%$) were regulated in cellular components, protein-containing complex ($12\%$), other organic parts ($0.3\%$), but the differences were no statistical significance. The results showed that TCM can be associated with biological functions such as response to stimulation, cell organism process, immune system process, and multi-tissue process to increase the efficacy of the treatment and play a therapeutic role (Figure 4A).
**FIGURE 4:** *(A) GO terms comparison between the model and control groups, and between the TSZSDH and model groups. The size of the dots represents the number of gene proportions, and the color represents the p-value. The greener the color, the smaller the -log10(p-value), and the redder the color, the greater the -log10(p-value). (B) COG annotations comparison between the model and control groups and the TSZSDH and model groups. The length of the bar graph represents the number of gene proportions, and the color represents the value. The bluer the color, the smaller the -log10(p-value), and the redder the color, the larger the -log10(p-value).*
KEGG pathway analysis was performed on all DEPs, and 23 pathways with $p \leq 0.05$ were screened out. The DEPs in the model and control groups were enriched in the following 11 pathways: focal adhesion ($50\%$), quorum sensing ($17\%$) related to cellular processes, the transforming growth factor-beta signaling pathway ($23\%$), extracellular matrix (ECM)–receptor interaction ($23\%$) in environmental information processing, proteoglycans in cancer ($25\%$), the advanced glycation end products (AGE)-receptors for AGE signaling pathway in diabetic complications ($17\%$), fatty acid degradation ($9\%$), fatty acid metabolism ($9\%$), protein digestion and absorption ($23\%$), the PPAR signaling pathway ($19\%$), and vitamin digestion and absorption ($4\%$). The DEPs in the model and TSZSDH group were enriched in the following 14 pathways: lysosome ($20\%$) and tight junction ($14\%$) in cellular processes, amoebiasis ($6\%$) in human diseases, metabolic pathways ($21\%$), other glycan degradation ($3\%$), arginine and proline metabolism ($3\%$), glycosaminoglycan degradation ($2\%$), tyrosine metabolism ($2\%$), alanine, aspartate, and glutamate metabolism ($2\%$), nitrogen metabolism ($1\%$), flavone and flavanol biosynthesis ($0.6\%$) in metabolism, the PPAR signaling pathway ($11\%$), protein digestion and absorption ($9\%$), and vascular smooth muscle contraction ($8\%$) in organismal systems. A network of KEGG pathways and genes was constructed using Cytoscape (Figure 5A).
**FIGURE 5:** *(A) KEGG pathway network diagram of the treatment and model groups. The blue graphs represent DEPs, the green graphs represent pathways, and the red graphs represent the types of pathways. (B) PPI analysis between differentially expressed proteins.*
COG annotation analysis was performed on all DEPs. These DEPs were functionally classified into four types (see the bubble chart): cell processes and signaling, information storage and processing, metabolism, and poorly characterized. Four pathways were statistically significant ($p \leq 0.05$). The DEPs in the model and control groups were enriched in one pathway related to extracellular structures ($18\%$). The pathways enriched by the DEPs in the model and TSZSDH groups were cytoskeleton ($29\%$), carbohydrate transport and metabolism ($13\%$), extracellular structures ($9\%$), and amino acid transport and metabolism ($9\%$) according to the number of relevant genes (Figure 4B).
## 3.6 Protein–protein interaction (PPI) analysis
The relationship between the model and TSZSDH groups was determined using the STRING database to further clarify the crosstalk between molecular mechanisms and DEPs. A total of 207 nodes (4 targets without corresponding genes and five fewer in the graph shown by STRING) and 333 edges were connected. Acsl1, Plin1, Plin4, and Dbil5 occupied the central position of the PPI network, acting as a hub to interact with other DEPs (Figure 5B).
Under the screening criteria of $p \leq 0.05$ and differential multiple (upregulated >1.2 and downregulated <0.83), the expression of DEPs in the model group/the control group in the TSZSDH group was analyzed. DEPs in the model group were upregulated, whereas one protein was downregulated in the TSZSDH group. DEPs in the model group were downregulated, whereas 20 proteins were upregulated in the TSZSDH group. One protein was upregulated in both groups. Thus, 21 abnormally expressed proteins in the model group were corrected in the TSZSDH group (see Supplementary Table S4; Figure 2B). A total of 22 proteins were involved in the model group/the control group and the TSZSDH group/the model group (One of the upregulated proteins failed to find the gene in UniProt). The KEGG pathway enrichment analysis of these DEPs revealed that they were enriched in the PPAR signaling pathway, protein digestion and absorption, and proteoglycans in cancer (see Supplementary Table S5).
## 3.7 WB and RT-qPCR analysis
The related pathways mediated by the PPAR signaling pathway were studied by the proteomics approach. Three DEPs, namely, Acsl1, Plin1, and PPARγ involved in the PPAR signaling pathway were validated by the WB and RT-qPCR method. Both experimental results are consistent. These DEPs were significantly downregulated in the model group and upregulated in the TSZSDH group (Figure 6), which was closer to the control group.
**FIGURE 6:** *(A) The protein expression level of PPARγ, Acsl1, and Plin1in the testicular cells. (B) The relative level of the proteins in the testicular cells. (C) The mRNA level of PPARγ, Acsl1, and Plin1 in the testicular cells. *
p < 0.05, **
p < 0.01, ***
p < 0.001, ****
p < 0.0001.*
## 4 Discussion
GTW is a lipid-soluble mixture mainly extracted from TW and has biological activities such as anti-infection and immunity enhancement (Wang et al., 2022). However, gonadal damage is an obstacle to using this drug. In 1981, researchers reported that GTW could be widely used to treat kidney diseases in adults and children. Although its clinical effect is ideal, many researchers have reported its reproductive injury. Li et al. [ 2020] found that GTW can reduce sperm number, increase sperm-deformity rate, and cause sperm kinetic parameter abnormality. Fan et al. [ 2020] analyzed and summarized research on the reproductive toxicity of TW over the years. The total incidence of reproductive toxicity in patients taking TW was about $18\%$. The occurrence of toxicity was related to the dose, time, and combination of drugs, age, and gender of patients. Specifically, the incidence of reproductive toxicity in males was higher than that in females. The drug dose required by male drug users for a reproductive toxicity reaction was smaller than that required by females, and the time of toxic symptom appearance was also faster than that in females. Damage to the male reproductive system mainly occurs in patients with rheumatoid arthritis and skin diseases treated with TW. Its clinical manifestations mainly include reduced sperm activity, oligospermia or azoospermia, and reduced fertility or infertility. Long-term medication may also cause testicular atrophy and decreased sexual desire (Yu, 1983; Xu et al., 2019). The study results have shown that GTW can cause the atrophy of the seminiferous tubule of testicular tissues of rats, thinning of the tube wall, reduction of spermatogenic cells at all levels, and reduction of sperm in the epididymal tube lumen. However, after the intervention of Cuscutae semen–*Radix rehmanniae* praeparata, pathological changes in testicular tissues can be significantly improved, which is consistent with the results of proteomics and WB.
Many scholars have thoroughly studied TCM because of its advantages such as multi-component, multi-target, and multi-pathway medication. Cuscutae semen and *Radix rehmanniae* praeparata are TCMs for tonifying the kidney. They can act on the hypothalamus-pituitary gonad axis to exert their effects. Among them are Cuscutae semen supplements yang and *Radix rehmanniae* praeparata supplements yin. Combining the two can strengthen and protect the “foundation of the whole body,” “nourish the source, and worship the vitality,” and nourish both yin and yang. It coincides with the theory of TCM that “those who are good at tonifying yang must seek yang in yin” and “those who are good at tonifying yin must seek yin in yang,” which embodies the idea that yin and yang are mutually correlated. The animal experiments in this study showed that Cuscutae semen–*Radix rehmanniae* praeparata could significantly improve GTW-induced testicular tissues damage, repair the structure of the seminiferous tubule, increase the number of spermatogenic cells at all levels, and increase the number of sperms significantly. The high-throughput proteomics analysis revealed differential expressions of Acsl1, Plin1, Dbil5, Plin4, Col12a1, Col1a1, Col5a3, Col1a2, and Dcn, which are associated with the PPAR signaling pathway, protein digestion and absorption, and the protein glycan pathway in cancer. Herein, the amino acid sequence of a protein in the PPAR signaling pathway is highly conserved among different species, suggesting that they have more than $80\%$ amino acid in rats, mice, and humans. Therefore, PPAR signaling pathways have attracted great attention (Bensinger and Tontonoz, 2008).
The PPAR signaling pathway protein is a member of the transcription factor superfamily of intranuclear receptors playing a key role in lipid and energy metabolisms. It controls the expression of many genes related to fatty acid input and oxidation and is closely related to glycolipid energy metabolism, oxidative stress, inflammation, cancer, and autophagy (Wang, 2010; Oyefiade et al., 2019).
PPAR comprises PPARα, PPARβ/δ, and PPARγ, which are homologous (Dubois et al., 2017). Among them, PPARα is the most widely distributed and mainly distributed in tissues rich in mitochondria, such as liver, skeletal muscle, and kidney tissues. It regulates the transport of fatty acids into mitochondria by regulating the expression of Acsl and CPT-1 to promote the β-oxidation of fatty acids in mitochondria, reducing the intracellular fatty acid levels (Singh et al., 2016). PPARβ is also widely present in various tissues; however, its physiological effect is unknown (Lefebvre et al., 2006; Wagner and Wagner, 2010). PPARγ is highly present in adipose tissues and mainly participates in fat differentiation. It can also increase the levels of the fatty acid transport protein, fatty acid transferase FAT/CD36, adipocyte-type fatty acid binding protein, phosphoenolpyruvate carboxykinase, and ACSL1, as well as enhance the expression of genes that promote fatty acid storage (Reddy RC., 2008). Simultaneously, it can reduce cholesterol and blood lipids (Reddy, 2008). Regueira et al. [ 2014] reported that PPARα and PPARβ/δ participate in fatty acid catabolism and glucose metabolism of Sertoli cells by promoting the β-oxidation of fatty acids to provide sufficient energy to Sertoli cells. Minutoli et al. [ 2009] reported that a PPARβ/δ agonist promotes the expression of the PPARβ/δ gene, which subsequently inhibits the damage of testicular tissues caused by ischemia-reperfusion injury. A PPAR agonist is widely used to treat several diseases including metabolic, chronic inflammatory, immunological, neurological, and psychiatric diseases, infections, and malignant tumors.
Currently, PPARγ is most studied by scholars. PPARγ is a ligand-dependent transcription factor in the nuclear receptor superfamily and a major regulator of the growth and development of adipose tissues (Martin and Parton, 2006; Puri et al., 2007). PPARγ has multiple biological functions and is important for regulating metabolism, controlling inflammation, improving atherosclerosis, inhibiting tumors, and regulating immune processes (Yang et al., 2022). PPARγ activation can also inhibit matrix production and reduce oxidative stress response, thus protecting tissue repair (Marder et al., 2013).
ACSL1 is present in the liver and adipocytes, and its gene is abundantly expressed in lipid droplets (LDs), microsomes, and mitochondria. As a common target gene of PPARα and PPARγ, it plays an important role in lipid and fatty acid metabolism (Fortis-Barrera et al., 2020; Grevengoed et al., 2015). ACSL1 was located in the endoplasmic reticulum, mitochondria-related membrane, and cytosol using isotype-specific antibodies but not in other components of mitochondria (Yin et al., 2022). ACSL1 is thought to play an important role in activating fatty acids to synthesize triacylglycerols (TGs). The high expression of the ACSL1 gene reduces the β-oxidation of fatty acids via the PPARγ pathway, thereby increasing triglyceride levels (Liu et al., 2020). Some studies have confirmed that ACSL1 is the target gene of PPARα and is regulated by PPARα. The expression of the ACSL1 gene was significantly upregulated in the mouse liver and kidney after administering a PPARα agonist (Singh et al., 2016), suggesting that Cuscutae semen–*Radix rehmanniae* praeparata has a similar effect on the PPARα agonist, which can improve the energy metabolism and oxidative stress of cells and inhibit the inflammatory reaction.
Plin1 is the most important member of the LD surface protein family, which is located in various tissues including liver and muscle tissues. It is also a phosphorylated protein present in rat epididymal adipocytes. It is specifically present on the surface of neutral LDs in adipocytes. It has a bidirectional regulatory role in regulating the accumulation and hydrolysis of TG and has biological functions such as regulating autophagy and inflammatory responses (Itabe et al., 2017; Zhang et al., 2018; Sun et al., 2022). Studies on mammals revealed that the transcription of the *Plin1* gene is mainly regulated by PPARγ, the main regulator of adipogenesis. In mammals, PPARγ can regulate the transcription of the *Plin1* gene by binding to the functional response element of PPARγ located in the 5′“flanking” region of the *Plin1* gene (Arimura et al., 2004). In the basal or resting state, Plin1 protects lipids in LD from interacting with adipose triglyceride lipase (ATGL), hormone-sensitive lipase (HSL), and binds to comparative gene identification-58, which is an activator of ATGL, to reduce ineffective lipolysis, thus maintaining intracellular TG levels. In contrast, catecholamines phosphorylate Plin1, HSL, and CGI58 by activating the cyclic AMP/protein kinase A signaling pathway during eating or exercise. After phosphorylation, Plin1 dissociates from CGI58, enhances ATGL activity, and recruits phosphorylated HSL on the surface of LD, promoting the cascade decomposition of TG (Feng et al., 2021). In summary, Plin1 controls lipid turnover in vivo by regulating lipolysis and lipid synthesis and maintains the homeostasis of lipid metabolism.
The main active component of GTW is triptolide. Previous studies found that the testicular oxidation and inflammatory reaction of GTW-induced reproductive injury in male rats are related to the severity of the disease, inhibiting oxidation can reduce the degree of testicular inflammation injury in model animals (Singla and Challana, 2014; Sheng et al., 2017; Tan et al., 2018). PPAR levels were significantly decreased in the testicular tissues of mice treated with triptolide. Ma et al, [2015] reported that triptolide might cause the metabolic disorder of fatty acid metabolism by downregulating the expression of the PPAR gene, causing considerable changes in a group of endogenous metabolites in the testis and serum, leading to abnormal testicular lipid and energy metabolism and thus leading to dyszoospermia. PPAR and associated fatty acid metabolism may be potential targets for the intervention or treatment of triptolide-induced male infertility. PPAR may be one of the key regulators of these intermediate metabolites in Sertoli cells. Triptolide may cause abnormal levels of small-molecule metabolites in Sertoli cells by regulating PPAR, leading to dyszoospermia and male infertility (Cheng et al., 2018; Ma et al., 2015). Hence, as the main effective component of GTW, triptolide can reduce the levels of related proteins in the PPAR pathway, thus leading to abnormal lipid energy metabolism and abnormal metabolites in vivo and promoting oxidation, eventually causing dyszoospermia and even infertility in males.
The results of the present study suggest that GTW may downregulate the expression of the PPARγ, Acsl1, and *Plin1* genes, causing the metabolic disorder of fatty acid, inhibiting energy metabolism in the testis, aggravating the oxidative stress of cells, and leading to the atrophy of the seminiferous tubule and the decrease in sperm number and vitality. Cuscutae semen–*Radix rehmanniae* praeparata can improve lipid energy metabolism, reduce oxidative stress, and increase sperm quantity and activity. The levels of Col12a1-, Col1a1-, Col1a2-, and Col5a3-related proteins in the model group were decreased. TSZSDH could significantly improve the levels of the related proteins. The testicular ECM includes the basement membrane of the seminiferous tubule and the intercellular matrix of peritubular cells, mainly including Type Ⅰ and Ⅳ collagens, laminin, arrestin, and protein polysaccharides, which play an important role in the self-renewal of spermatogonial stem cells and sperm generation (Yuan et al., 2017). In the KEGG pathway, Col12a1, Col1a1, Col5a3, and Col1a2 were enriched in the protein digestion and absorption pathway, suggesting that Cuscutae semen–*Radix rehmanniae* praeparata can promote the generation of testicular ECM, thereby improving GTW-induced reproductive injury in male rats.
## 5 Conclusion
Cuscutae semen–*Radix rehmanniae* praeparata can effectively relieve GTW-induced reproductive injury. The pathway enrichment analysis of DEPs indicated that the PPAR signaling pathway, protein digestion and absorption, and the protein glycan pathway in cancer play an important role in relieving this injury. Acsl1, Plin1, and PPARγ were verified by WB and RT-q-PCR experiments. Cuscutae semen and *Radix rehmanniae* praeparata may regulate the PPAR signaling pathway mediated Acsl1, Plin1 and PPARγ to reduce the testicular tissue damage of male rats caused by GTW.
## 6 Limitations and outlook
In this study, we used quantitative proteomics as a tool to combine TCM with in vivo metabolites and their pathways, which are conducive to the standardization of TCM and show the mechanism of TCM. However, this study has several limitations as follows: 1) The sample size of this study was relatively insufficient, which may lead to some experimental results without obvious statistical significance. 2) The sample sizes of the included studies were small, and the number of DEPs was less than that reported in foreign studies. 3) Although WB experiments were performed to verify DEPs in this study, only three DEPs were verified because of limited funds. 4) No other research method was followed to compare the downstream metadata of the screened DEPs with the metadata of previous studies.
Owing to the high diversity of effective components of TCM, understanding their specific mechanisms requires further clarification despite identifying some specific differential proteins and pathways. Large-sample studies are required to enhance the reliability of the study conclusions and achieve more significant results. In the present study, the metabolites of DEPs were verified by immunohistochemistry and other methods, and the effects of Cuscutae semen–*Radix rehmanniae* praeparata on DEPs were further discussed.
## 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: iProx (http://www.iprox.org/), IPX0004150001.
## Ethics statement
The experimental protocol was approved by the Animal Experiment Ethics Committee of the Henan University of Chinese Medicine (No: DWLL202105053).
## Author contributions
YD and SH designed and conducted the study with equal contribution. YD, LS, YX, and SXu performed the experiments. YD and LS analyzed the data and prepared 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/fphar.2023.1050907/full#supplementary-material
## Abbreviations
GTW, tripterygium wilfordii multiglycosides; Cuscutae semen–*Radix rehmanniae* praeparata, TSZSDH; TW, *Tripterygium wilfordii* Hook.f.; WB, Western blotting; RT-qPCR, Real-Time Quantitative Polymerase Chain Reaction; TCM, Traditional Chinese Medicine; DEPs, differentially expressed proteins; TMT, tandem mass tag; PPI, protein–protein interaction; KEGG, Kyoto Encyclopedia of Genes and Genomes; GO, gene ontology; COG, cluster of the orthologous group; PLIN1, Perilipin1; ACSL1, long-chain acyl-Co A synthetases; PPARγ, peroxisome proliferator-activated receptor γ.
## References
1. Arimura N., Horiba T., Imagawa M., Shimizu M., Sato R.. **The peroxisome proliferator-activated receptor γ regulates expression of the perilipin gene in adipocytes**. *J. Biol. Chem.* (2004) **279** 10070-10076. DOI: 10.1074/jbc.m308522200
2. Bensinger S. J., Tontonoz P.. **Integration of metabolism and inflammation by lipid-activated nuclear receptors**. *Nature* (2008) **454** 470-477. DOI: 10.1038/nature07202
3. Cheng Y., Chen C., Wang L., Kong J., Pan J., Xi Y.. **Triptolide-induced mitochondrial damage dysregulates fatty acid metabolism in mouse sertoli cells**. *Toxicol. Lett.* (2018) **292** 136-150. DOI: 10.1016/j.toxlet.2018.04.035
4. Dubois V., Eeckhoute J., Lefebvre P., Staels B.. **Distinct but complementary contributions of PPAR isotypes to energy homeostasis**. *J. Clin. Invest.* (2017) **127** 1202-1214. DOI: 10.1172/JCI88894
5. Fan Y. F., Xu Y., Su X. H., Liu L. L., Tian Y. G., Zhao Y.. **Effect of Tripterygium Glycosides Tablets on reproductive toxicity in male rats with Ⅱ type collagen induced arthritis**. *Zhongguo Zhong Yao Za Zhi* (2020) **45** 755-763. DOI: 10.19540/j.cnki.cjcmm.20190523.401
6. Feng C., Chen C., Yan J.. **The biological function of perilipid protein and the relationship between its gene polymorphism and obesity**. *Health Res* (2021) **50** 506-509+532. DOI: 10.19813/j.cnki.weishengyanjiu.2021.03.027
7. Fortis-Barrera M. L. Á., Alarcón-Aguilar F. J., Becerril-García A., Flores-Sáenz J. L. E., Almanza-Pérez J. C., García-Lorenzana M.. **Mechanism of the hypoglycemic activity and hepatoprotective effect of the aqueous extract of**. *J. Med. Food* (2020) **23** 783-792. DOI: 10.1089/jmf.2019.0126
8. Grevengoed T. J., Cooper D. E., Young P. A., Ellis J. M., Coleman R. A.. **Loss of long-chain acyl-CoA synthetase isoform 1 impairs cardiac autophagy and mitochondrial structure through mechanistic target of rapamycin complex 1 activation**. *FASEB J* (2015) **29** 4641-4653. DOI: 10.1096/fj.15-272732
9. Itabe H., Yamaguchi T., Nimura S., Sasabe N.. **Perilipins: A diversity of intracellular lipid droplet proteins**. *Lipids Health Dis.* (2017) **16** 83. DOI: 10.1186/s12944-017-0473-y
10. Lefebvre P., Chinetti G., Fruchart J. C., Staels B.. **Sorting out the roles of PPAR in energy metabolism and vascular homeostasis**. *J. Clin. Invest.* (2006) **116** 571-580. DOI: 10.1172/jci27989
11. Li Y. Q., Liu C. F., Jia K. X., Wang J. X., Wang J. X., Zhang J. X.. **Comparative study on chronic multiple organ injury in normal rats caused by high dose of Tripterygium Glycosides Tablets from 6 different manufacturers**. *Zhongguo Zhong Yao Za Zhi* (2020) **45** 746-754. DOI: 10.19540/j.cnki.cjcmm.20191017.401
12. Liu D., Wang X., Zhang M., Tian J., Liu M., Jin T.. **WISP1 alleviates lipid deposition in macrophages via the PPARγ/CD36 pathway in the plaque formation of atherosclerosis**. *J. Cell. Mol. Med.* (2020) **24** 11729-11741. DOI: 10.1111/jcmm.15783
13. Ma B., Qi H., Li J., Xu H., Chi B., Zhu J.. **Triptolide disrupts fatty acids and peroxisome proliferator-activated receptor (PPAR) levels in male mice testes followed by testicular injury: A GC–MS based metabolomics study**. *Toxicology* (2015) **336** 84-95. DOI: 10.1016/j.tox.2015.07.008
14. Marder W., Khalatbari S., Myles J. D., Hench R., Lustig S., Yalavarthi S.. **The peroxisome proliferator activated receptor-pioglitazone improves vascular function and decreases disease activity in patients with rheumatoid arthritis**. *J. Am. Heart Assoc.* (2013) **2** e000441. DOI: 10.1161/jaha.113.000441
15. Martin S., Parton R. G.. **Lipid droplets: A unified view of a dynamic organelle**. *Nat. Rev. Mol. Cell. Biol.* (2006) **7** 373-378. DOI: 10.1038/nrm1912
16. Meng X. T., Liao L. B., Ma Y. X., Aikebaier K. D. E., Bai S. B.. **Cuscuta chinensis flavonoids reduce the expression of GM-CSF in the testis of oligozoospermia rats**. *Zhonghua Nan Ke Xue* (2020) **26** 639-644. DOI: 10.13263/j.cnki.nja.2020.07.011
17. Minutoli L., Antonuccio P., Polito F., Bitto A., Squadrito F., Irrera N.. **Peroxisome proliferator activated receptor beta/delta activation prevents extracellular regulated kinase 1/2 phosphorylation and protects the testis from ischemia and reperfusion injury**. *J. Urol.* (2009) **181** 1913-1921. DOI: 10.1016/j.juro.2008.11.095
18. Oyefiade A., Erdman L., Goldenberg A., Malkin D., Bouffet E., Taylor M. D.. **PPAR and GST polymorphisms may predict changes in intellectual functioning in medulloblastoma survivors**. *J. Neurooncol.* (2019) **142** 39-48. DOI: 10.1007/s11060-018-03083-x
19. Puri V., Konda S., Ranjit S., Aouadi M., Chawla A., Chouinard M.. **Fat-specific protein 27, a novel lipid droplet protein that enhances triglyceride storage**. *J. Biol. Chem.* (2007) **282** 34213-34218. DOI: 10.1074/jbc.m707404200
20. Reddy R. C.. **Immunomodulatory role of PPAR-γ in alveolar macrophages**. *J. Invest. Med.* (2008) **56** 522-527. DOI: 10.2310/jim.0b013e3181659972
21. Regueira M., Riera M. F., Galardo M. N., Pellizzari E. H., Cigorraga S. B., Meroni S. B.. **Activation of PPAR α and PPAR β/δ regulates Sertoli cell metabolism**. *Mol. Cell. Endocrinol.* (2014) **382** 271-281. DOI: 10.1016/j.mce.2013.10.006
22. Sheng W., Zhang Y. S., Li Y. Q., Wu X. N., Chai L. M., Yue L. F.. **Effect of yishenjianpi recipe on semen quality and sperm mitochondria in mice with oligoasthenozoospermia induced by tripterygium glycosides**. *Afr. J. Tradit. Complement. Altern. Med.* (2017) **14** 87-95. DOI: 10.21010/ajtcam.v14i4.11
23. Singh A. B., Kan C. F. K., Dong B., Liu J.. **SREBP2 activation induces hepatic long-chain Acyl-CoA Synthetase 1 (ACSL1) expressionin vivoandin vitrothrough a Sterol Regulatory Element (SRE) motif of the ACSL1 c-promoter**. *J. Biol. Chem.* (2016) **291** 5373-5384. DOI: 10.1074/jbc.M115.696872
24. Singla N., Challana S.. **Reproductive toxicity of triptolide in male house rat,**. *Sci. World J.* (2014) **2014** 879405. DOI: 10.1155/2014/879405
25. Sun Y., Zhai G., Pang Y., Li R., Li Y. M., Cao Z. P.. **PPAR gamma2: The main isoform of PPARγ that positively regulates the expression of the chicken Plin1 gene**. *J. Integr. Agric.* (2022) **21** 2357-2371. DOI: 10.1016/S2095-3119(21)63896-0
26. Tan T. Y., Zheng Y. F., Tang Y. B., Bai X., Liu M. Q., Li Y.. **Bazi Granules improves sperm quality in rats with oligoasthenozoospermia induced by multi-glycosides of tripterygium wilfordii**. *J. Androl.* (2018) **24** 1016-1020. DOI: 10.13263/j.cnki.nja.2018.11.011
27. Wagner K. D., Wagner N.. **Peroxisome proliferator-activated receptor beta/delta (PPARbeta/delta) acts as regulator of metabolism linked to multiple cellular functions**. *Pharmacol. Ther.* (2010) **125** 423-435. DOI: 10.1016/j.pharmthera.2009.12.001
28. Wang H., Song H., Shen Y.. **Effects of Qingre Jiangzhuo Recipe combined with Tripterygium wilfordii polyglycosides on immune function and inflammatory factors in elderly patients with chronic glomerulonephritis**. *Chin. J. Gerontology* (2022) **42** 2960-2963
29. Wang J., Bao B., Meng F., Deng S., Dai H., Feng J.. **To study the mechanism of Cuscuta chinensis Lam. And Lycium barbarum L. in the treatment of asthenospermia based on network pharmacology**. *J. Ethnopharmacol.* (2021) **270** 113790. DOI: 10.1016/j.jep.2021.113790
30. Wang Y. X.. **PPARs: Diverse regulators in energy metabolism and metabolic diseases**. *Cell. Res.* (2010) **20** 124-137. DOI: 10.1038/cr.2010.13
31. Xu Y., Fan Y. F., Zhao Y., Lin N.. **Overview of reproductive toxicity studies on**. *Zhongguo Zhong Yao Za Zhi* (2019) **44** 3406-3414. DOI: 10.19540/j.cnki.cjcmm.20190524.401
32. Yang S., Xu X., Xu H., Xu S., Lin Q., Jia Z.. **Purification, characterization and biological effect of reversing the kidney-yang deficiency of polysaccharides from semen cuscutae**. *Carbohydr. Polym.* (2017) **175** 249-256. DOI: 10.1016/j.carbpol.2017.07.077
33. Yang Y., Zhou Y., Wei Y., Zhang T.. **Research progress of the role of PPARγ in autoimmune diseases**. *Acta Pharm. Sin.* (2022) **2022** 1-19. DOI: 10.16438/j.0513-4870.2022-0708
34. Yang Z., Zhang X., Chen Z., Hu C.. **Effect of Wuzi Yanzong on reproductive hormones and TGF-β1/smads signal pathway in rats with oligoasthenozoospermia**. *Evid. Based Complement. Altermat. Med.* (2019) **2019** 7628125-7628213. DOI: 10.1155/2019/7628125
35. Yin L., Wang L., Shi Z., Ji X., Liu L.. **The role of peroxisome proliferator-activated receptor gamma and atherosclerosis: Post-translational modification and selective modulators**. *Front. Physiol.* (2022) **13** 826811. DOI: 10.3389/fphys.2022.826811
36. Yu D. Y.. **Clinical observation of 144 cases of rheumatoid arthritis treated with glycoside of Radix Tripterygium Wilfordii**. *J. Tradit. Chin. Med.* (1983) **3** 125-129. DOI: 10.13288/j.11-2166/r.1982.07.014
37. Yuan L. G., Zhang Y., Li C., Cheng Y. D., Lu Q. Z.. **Comparison of distribution characteristics of extracellular matrix components in the testis of the Tibetan sheep and the small-tail Han sheep from plateau**. *Acta Anat. Sin.* (2017) **48** 179-186. DOI: 10.16098/j.issn.0529-1356.2017.02.011
38. Zhang X., Evans T. D., Jeong S. J., Razani B.. **Classical and alternative roles for autophagy in lipid metabolism**. *Curr. Opin. Lipidol.* (2018) **29** 203-211. DOI: 10.1097/mol.0000000000000509
39. Zhang X. X., Huang D., Liu N. N., Li J., Lin R. Y., Zhang X. Z.. **GTW-induced abnormal expressions of testicular reproduction-related genes and intervention with kidney-tonifying Chinese herbs**. *Zhonghua Nan Ke Xue* (2012) **18** 466-471. DOI: 10.13263/j.cnki.nja.2012.05.004
40. Zhen G. H., Jiang B., Bao Y. M., Li D. X., An L. J.. **The protect effect of flavonoids from Cuscuta chinensis in PC12 cells from damage induced by H2O2**. *Zhong Yao Cai* (2006) **29** 1051-1055. DOI: 10.13863/j.issn1001-4454.2006.10.022
|
---
title: Age-related ultrastructural changes in spheroids of the adipose-derived multipotent
mesenchymal stromal cells from ovariectomized mice
authors:
- Vitalii Kyryk
- Oleg Tsupykov
- Alina Ustymenko
- Ekaterina Smozhanik
- Iryna Govbakh
- Gennadii Butenko
- Galyna Skibo
journal: Frontiers in Cellular Neuroscience
year: 2023
pmcid: PMC9982046
doi: 10.3389/fncel.2023.1072750
license: CC BY 4.0
---
# Age-related ultrastructural changes in spheroids of the adipose-derived multipotent mesenchymal stromal cells from ovariectomized mice
## Abstract
Introduction: Adipose-derived multipotent mesenchymal stromal cells (ADSCs) are widely used for cell therapy, in particular for the treatment of diseases of the nervous system. An important issue is to predict the effectiveness and safety of such cell transplants, considering disorders of adipose tissue under age-related dysfunction of sex hormones production. The study aimed to investigate the ultrastructural characteristics of 3D spheroids formed by ADSCs of ovariectomized mice of different ages compared to age-matched controls.
Methods: ADSCs were obtained from female CBA/Ca mice randomly divided into four groups: CtrlY—control young (2 months) mice, CtrlO—control old (14 months) mice, OVxY—ovariectomized young mice, and OVxO—ovariectomized old mice of the same age. 3D spheroids were formed by micromass technique for 12–14 days and their ultrastructural characteristics were estimated by transmission electron microscopy.
Results and Discussion: The electron microscopy analysis of spheroids from CtrlY animals revealed that ADSCs formed a culture of more or less homogeneous in size multicellular structures. The cytoplasm of these ADSCs had a granular appearance due to being rich in free ribosomes and polysomes, indicating active protein synthesis. Extended electron-dense mitochondria with a regular cristae structure and a predominant condensed matrix were observed in ADSCs from CtrlY group, which could indicate high respiratory activity. At the same time, ADSCs from CtrlO group formed a culture of heterogeneous in size spheroids. In ADSCs from CtrlO group, the mitochondrial population was heterogeneous, a significant part was represented by more round structures. This may indicate an increase in mitochondrial fission and/or an impairment of the fusion. Significantly fewer polysomes were observed in the cytoplasm of ADSCs from CtrlO group, indicating low protein synthetic activity. The cytoplasm of ADSCs in spheroids from old mice had significantly increased amounts of lipid droplets compared to cells obtained from young animals. Also, an increase in the number of lipid droplets in the cytoplasm of ADSCs was observed in both the group of young and old ovariectomized mice compared with control animals of the same age. Together, our data indicate the negative impact of aging on the ultrastructural characteristics of 3D spheroids formed by ADSCs. Our findings are particularly promising in the context of potential therapeutic applications of ADSCs for the treatment of diseases of the nervous system.
## Introduction
Adipose tissue is a rich source of multipotent mesenchymal stromal cells (ADSCs). The use of ADSCs in modern medicine makes it possible to realize their multilinear potential in various pathological conditions: in the treatment of diseases of the nervous system (Ma et al., 2020; Zhou et al., 2020), endothelial dysfunction in critical limb ischemia and diabetes mellitus (Magenta et al., 2021), endocrine dysfunction (Amer et al., 2018), in coronary heart disease (Murohara et al., 2009), in plastic and reconstructive surgery of soft tissues, and the musculoskeletal system, etc.
However, it is important to note that the therapeutic success of cell therapy may depend on many factors, including the age of the donor. In a study by Liu et al. [ 2017], a negative effect of the age of the donor’s adipose tissue on the quantity and quality of human mesenchymal adipose-derived cells was shown: both the number of colony-forming units of fibroblasts and the number of cells obtained from the stromal-vascular fraction in vitro, as well as the rate of their proliferation, are reduced. In addition, a violation of the migration ability of cells obtained from old donors was observed, which is explained by the reduced expression of chemokine receptors such as CXCR4 and CXCR7 (Liu et al., 2017).
Most of the pathological conditions that require cell therapy using autologous stem cells occur mainly in the elderly, so it is relevant to establish criteria for the biological safety of stem cells of adipose origin obtained in the menopausal and postmenopausal periods, which are accompanied by estrogen deficiency (Eastell et al., 2016).
The features of ADSCs are affected by both culture conditions and intercellular signals. Unlike 2D cultivation, three-dimensional (3D) culture of ADSCs in the form of spheroids, which partially mimics the conditions of the microenvironment (niche) of stem cells, can significantly improve their survival in the recipient tissue and increase the overall regenerative potential (Egger et al., 2018).
The aim of our study was to investigate the ultrastructural characteristics of multicellular three-dimensional spheroids formed by ADSCs obtained under conditions of estrogen deficiency in a model of ovariectomy in mice of different ages compared to age-matched controls. Our study aimed at establishing the mechanisms of cellular self-organization, contact intercellular signaling, extracellular matrix production, and resistance to hypoxia depending on the size of the spheroid.
## Materials and methods
All animal procedures were performed in accordance with “European Convention for the protection of Vertebrate Animals Use for Experimental and Other Scientific Purposes” (Strasbourg, 1986), “European Directive $\frac{2010}{63}$/EU on the protection of animals used for scientific purposes” and the Law of Ukraine “On protection of animals from cruelty” as well as in abidance to all principles of bioethics and biosafety regulations. The study was approved by Ethics Committee of the Institute of Genetic and Regenerative Medicine (protocol no. 9-2021 dated December 15, 2021).
## Animals
Adipose tissue was obtained from young (2 months) and old (14 months) female CBA/Ca mice, which were kept under standard conditions in a vivarium of the D. F. Chebotarev State Institute of Gerontology NAMS of Ukraine under a 12:12 h light/dark cycle with access to water and food ad libitum (Table 1).
**Table 1**
| Experimental group | Young | Old |
| --- | --- | --- |
| | The age of animals at the time of surgery (months); number (n) | The age of animals at the time of surgery (months); number (n) |
| Control, sham-operated (Ctrl) | 2 months (n = 8) | 14 months (n = 11) |
| Ovariectomy (OVx) | 2 months (n = 7) | 14 months (n = 8) |
## Ovariectomy modeling
Animals were anesthetized by intraperitoneal administration of $2.5\%$ solution of 2,2,2-tribromethanol (Sigma-Aldrich, St. Louis, MO, USA) at a dose of 400 mg/kg and bilateral ovariectomy was performed under aseptic conditions using microsurgery technique. The animals of the same age which had only incisions of the abdominal cavity and isolation of the ovaries without resection (sham-operated) were used as a control group. Wounds were sutured in layers; the animals were kept under a heat lamp until the recovery from anesthesia.
## Adipose-derived mesenchymal stromal cells isolation and culture
Murine ADSCs cultures were obtained and characterized according to standard methods (Yu et al., 2011). Two months after ovariectomy the CBA/Ca mice were euthanized by cervical dislocation under the anesthesia with $2.5\%$ solution of 2,2,2-tribromethanol at a dose 400 mg/kg. Under sterile conditions, subcutaneous adipose tissue was isolated, minced with scissors into 1 mm3 pieces in DMEM/F12 medium (Sigma-Aldrich, St. Louis, MO, USA) and incubated in $0.1\%$ solution of collagenase type IA (Sigma-Aldrich, St. Louis, MO, USA) for 60 min at 37°C with constant stirring on a shaker at 100 rpm. The resulting cell suspension was washed in 10 ml DMEM medium (Sigma-Aldrich, St. Louis, MO, USA) by centrifugation at 300× g for 5 min. The supernatant with mature adipocytes and debris was discarded and pellet passed through a sterile cell strainer with a pore diameter of 100 μm (Greiner bio-one, Kremsmünster, Austria). Cells of the stromal-vascular fraction were cultured in a CO2 incubator in humidified atmosphere with $5\%$ CO2 at a temperature of +37°C in complete nutrient medium DMEM-LG (Sigma-Aldrich, St. Louis, MO, USA) supplemented with $15\%$ fetal bovine serum (FBS) (HyClone Laboratories Inc., South Logan, UT, USA), penicillin 100 U/ml, streptomycin 100 μg/ml (Sigma-Aldrich, St. Louis, MO, USA), 1:100 nonessential amino acids (Sigma-Aldrich, St. Louis, MO, USA). The nutrient medium was replaced in 3 days. Cells were sub-cultured to achieve $80\%$ monolayer confluency (for 4–5 days) using $0.25\%$ trypsin (Sigma-Aldrich, St. Louis, MO, USA) and $0.02\%$ Versene solution (Bio-Test Laboratory, Kyiv, Ukraine).
## Immunophenotyping of ADSCs
On the 2nd passage, cells were analyzed by flow cytometry with BD FACSAria cell sorter (Becton Dickinson, Franklin Lakes, NJ, USA) using anti-mouse monoclonal antibodies: CD90 APC-Cy7 (BD Biosciences, cat. no. 561401, Franklin Lakes, NJ, USA), CD105 APC (Invitrogen, cat. no. 17-1051-82, Carlsbad, CA, USA), CD73 PE (BD Biosciences, cat. no. 550741, Franklin Lakes, NJ, USA), CD44 PE (BD Biosciences, cat. no. 553134, Franklin Lakes, NJ, USA), CD45 PE (Thermo Fisher Scientific, cat. no. MA1-10233, Waltham, MA, USA), CD34 Alexa Fluor® 647 (BD Biosciences, cat. no. 560230, Franklin Lakes, NJ, USA).
## Three-dimensional spheroids formed by ADSCs
After the second passage, the monolayer culture of ADSCs was transferred to a suspension state, washed from the enzyme, and 1.5 × 106 cells were transferred to a 5 ml PS tube with a non-adhesive surface. After the formation of the spheroids (after 24–48 h of culturing in the tube), the cells were cultured for 12–14 days in a CO2 incubator under conditions of humidified atmosphere with $5\%$ CO2 at a temperature of +37°C. The nutrient medium was completely replaced every 2–3 days. On the 5–7th day, the standard medium for cultivation was replaced with medium for adipogenic induction to confirm that the obtained cultures met the minimal criteria to define ADSCs in terms of potential for directed differentiation. Adipogenic differentiation was performed using DMEM High *Glucose medium* (Gibco, USA) supplemented with $10\%$ FBS, dexamethasone (1 μM), indomethacin (200 μM), 3-Isobutyl-1-methylxanthine (500 μM), and insulin (5 μg/ml), all reagents—Sigma-Aldrich, St. Louis, MO, USA. The medium was changed every 3 days and the cells were cultured for 14 days. The area of spheroids was calculated according to the formula: $S = 4$▪πR2, where π is a constant of 3.14; R is the radius of the sphere. To estimate the area of spheroids, we used images obtained under inverted microscope IX-71 (Olympus, Japan) using a DP20 (Olympus, Japan) camera and performed measurements using QuicjPHOTO MICRO 2.3 software (Olympus, Japan). On day 14 in vitro grown cultures (DIV14) ADSCs spheroids were rinsed with PBS and fixed with $4\%$ formaldehyde. After staining with Oil Red O (Sigma-Aldrich, St. Louis, MO, USA), the cells were visualized with light microscopy IX-71 (Olympus, Japan).
Our observations showed that ADSCs obtained from control young animals form a culture of more or less homogeneous spheres in size, the surface area of which ranges approximately from 5,000 μm2 to 7,000 μm2 (Figure 2A).
**Figure 2:** *Three-dimensional spheroids (arrows) formed by ADSCs, obtained from young (A), old (B), and ovariectomized old (C) animals. Light microscopy. Scale bar—50 μm.*
At the same time, ADSCs obtained from control and ovariectomized old animals are able to form a 3D-culture heterogeneous in size, the surface area of which approximately ranges from 3,000 μm2 to 160,000 μm2 (Figures 2B,C).
In addition, spheroids formed by ADSCs obtained from control and ovariectomized old animals exhibit enhanced adipogenic potential compared to ADSCs obtained from young mice (data not shown).
## Transmission electron microscopy (TEM)
The ADSCs spheroids from different experimental groups were centrifuged at 500× g and pellets were processed for TEM according to usual protocols (Tsupykov et al., 2016). Briefly, the culture medium was replaced with $4\%$ formaldehyde and $2.5\%$ glutaraldehyde (Fluka, Buchs, Switzerland) in 0.1 M phosphate buffer (PB). ADSCs spheroids were then rinsed, postfixed in $1\%$ osmium tetroxide (Sigma-Aldrich, St. Louis, MO, USA) in 0.1 M PB, dehydrated in ascending concentrations ($50\%$–$100\%$) of ethanol, and then embedded in Epon 812 (Fluka, Buchs, Switzerland). Ultrathin sections (50–70 nm) were cut with a diamond knife, collected on single slot grids and then counterstained with lead citrate (Fluka, Buchs, Switzerland) and alcoholic uranyl acetate (Merck, Darmstadt, Germany). Grids were examined in JEOL 100-CX (JEOL, Japan) electron microscope operating at 80 kV.
## Immunophenotyping of adipose-derived multipotent mesenchymal stromal cells cultures
According to flow cytometry data, the ADSCs cultures obtained from both young and old mice expressed typical stromal markers CD44, CD90, and CD105 at a high level (≥$95\%$). At the same time, low expression of hematopoietic markers CD34 and CD45 (<$3\%$) was noted (Figure 1). In our previous study, we confirmed the potential of ADSCs from ovariectomized mice to differentiate into osteogenic and adipogenic directions (Ustymenko et al., 2019). Taking into account significant changes in adipose and bone tissue associated with hormonal changes in old age, we focused our research efforts specifically on assessing the osteogenic and adipogenic differentiation potential of adipose-derived cells. As a result, the morphology, phenotype and differentiation potential of obtained ADSCs cultures meet the minimal criteria for defining multipotent mesenchymal stromal cells according to the International Society for Cellular Therapy (ISCT) position statement (Dominici et al., 2006).
**Figure 1:** *Histograms of expression of CD44, CD90, CD105, CD34, and CD45 markers in the culture of murine ADSCs from young animals according to flow cytometry, the 2nd passage.*
## Ultrastructural analysis of spheroids from ADSCs
Results from electron microscopy showed that adipose-derived multipotent mesenchymal stromal cells (ADSCs) from young control CBA/Ca mice under non-adhesive conditions formed a culture of more or less homogeneous in size multicellular structures (from 75 μm to 95 μm in diameter)—spheroids consisted of several surface layers and an inner zone (Figures 3A,B).
**Figure 3:** *Structure of spheroids from the adipose-derived multipotent mesenchymal stromal cells from young control CBA/Ca mice. (A) Central section giving an overview of representative ADSCs spheroid after 14 days of culture. Spheroid consists of the dense outer zone and the inner area, where round or polygonal cells are embedded in an extracellular matrix (ECM). (B) Elongated cells of the outer area and a part of the inner zone of ADSCs spheroids with an extracellular matrix. The ADSCs cytoplasm is extremely rich in vacuoles (V). N—nuclei; Nu—nucleolus. (C) In the outer area ADSCs has many thin pseudopodia extended from the cell surface. The ADSCs cytoplasm contains a large number of vacuoles and a moderate amount of lipid droplets (LD). (D) Plasma membranes of adjacent adipose-derived multipotent mesenchymal stromal cells are attached by numerous junctions, which appeared as tight junctions (circles). (E) Detailed view of cytoplasm containing a relative high amount of elongated electron-dense mitochondria (m), rough endoplasmic reticulum (rer), and intermediate filament (if) bundles. (F) Well-developed Golgi apparatus (G) producing large secreting granules. (G) ADSC cytoplasm containing a number of late endosomal multivesicular bodies (MvB). A fragment of mitochondria is observed inside the autophagosome (Ap). (H) The sites of active exocytosis (arrows) on the plasma membrane. Scale bars: (A)—25 μm, (B)—15 μm, (C)—2 μm, (D,E,H)—1 μm, (F,G)—0.5 μm.*
The inner zone consisted of polygonal or round cells, embedded in an extracellular matrix (Figure 3B). These cells were tightly packed, with narrow extracellular spaces. In the inner area, ADSCs were attached to each other by numerous junctions, which, at high magnification, appeared as tight junctions (Figure 3D). ADSCs had a single, irregularly shaped and large euchromatic nuclei with one or more nucleoli (Figures 3B,C). Elongated electron-dense mitochondria with a regular cristae structure and a predominant condensed matrix were observed in ADSCs from CtrlY group, which could indicate high respiratory activity (Figure 3E). The cytoplasm had a granular appearance due to being rich in free ribosomes and polyribosomes, indicating active protein synthesis (Figures 3E,F). Interestingly, the rough endoplasmic reticulum cisternae were often dilatated and contained moderately electron-dense material (Figures 3D–F). Furthermore, transmission electron microscopy showed well-developed *Golgi apparatus* in the juxta-nuclear area producing large secreting granules (Figure 3F). A number of multivesicular bodies, a special kind of late endosomes, were also observed in ADSC cytoplasm (Figure 3G). Multivesicular bodies were very heterogeneous in size and morphology. Whole intracellular elements such as mitochondria were sequestered inside an autophagosome that then fused with multivesicular bodies to form amphisome (Figure 3G). Electron microscopy analysis revealed the process of active exocytosis on the plasma membrane of ADSCs (Figure 3H).
At the same time, ADSCs from old control mice (CtrlO group) formed a culture of heterogeneous in size spheroids (from 60 μm to 370 μm in diameter). Unlike spheroids formed by ADSCs from young animals, ADSCs from old mice were not tightly packed and had a wide intercellular space (Figure 4A). It was also difficult to clearly separate the outer and inner layers, as was observed in spheroids obtained from young mice.
**Figure 4:** *Structure of spheroids of the adipose-derived multipotent mesenchymal stromal cells from old control CBA/Ca mice. (A) An overview of representative ADSCs spheroid after 14 days of culture. ADSCs are not tightly packed in spheroid and had a wide intercellular space. (B) Left electronogram—The mitochondrial population is heterogeneous and is represented by mitochondria with both electron-dense condensed and enlightened matrix. Right electronogram—A detailed view of round mitochondrion with enlightened matrix and cristae (arrowheads) arranged perpendicular to the mitochondrial tubular axis. (C) The ADSCs cytoplasm contains rough endoplasmic reticulum (rer) and well-developed Golgi apparatus (G). Areas of rough endoplasmic reticulum that do not contain ribosomes are marked with arrows. N—nuclei. (D) A large number of lipid droplets (LD) is observed in the ADSCs cytoplasm. (E) Typical multivesicular (MVB) and multilamellar bodies (MLB) are observed in the ADSCs cytoplasm. Scale bars: (A)—15 μm, (B) left—0.5 μm, right—0.2 μm, (C,D)—1 μm, (E)—0.5 μm.*
Transmission electron microscopy of ADSCs from CtrlO group showed a similar ultrastructural architecture to young control, but there were some noticeable differences. In these ADSCs the mitochondrial population was heterogeneous, a significant part was represented by more round structures (Figure 4B). This may indicate an increase in mitochondrial fission and/or an impairment of the fusion. Mitochondria with both electron-dense condensed and enlightened matrix were presented, denoting their different respiratory activity (Figure 4B). Significantly fewer polysomes and ribosomes attached to the surface of rough endoplasmic reticulum were also observed in the cytoplasm of ADSCs from CtrlO group, indicating low protein synthetic activity (Figure 4C). In addition, the cytoplasm of ADSCs in spheroids from old mice had notably increased amounts of lipid droplets compared to cells obtained from young animals (Figure 4D). The cytoplasm was extremely rich in endosomal elements showing typical multilamellar and multivesicular structures (Figure 4E).
Electron microscopic analysis showed that ovariectomy led to changes in the ultrastructure of ADSCs in spheroids. The cytoplasm of ADSCs from young ovariectomized mice had significantly increased amounts of lipid droplets and endosomal elements compared to cells obtained from young control animals (Figure 5A).
**Figure 5:** *Structure of spheroids of the adipose-derived multipotent mesenchymal stromal cells from young (A) and old (B) ovariectomized CBA/Ca mice. (A) The cytoplasm of ADSCs from young ovariectomized mice have a large number of lipid droplets (LD) and endosomal elements. ECM—extracellular matrix. N—nuclei. (B) The cell cytoplasm of ADSCs from old ovariectomized mice contains giant lipid droplets (gLD). m—mitochondria. Scale bars: (A,B)—2 μm.*
Interestingly, the most prominent ultrastructural feature of the ADSCs from old ovariectomized mice was giant lipid droplets (3–4.5 μm in diameter) that probably formed by fusion or coalescence of smaller adjacent lipid droplets (Figure 5B). The cytoplasm was also characterized by numerous vacuolar elements.
## Discussion
The issue of the impact of aging on the phenotypic and functional characteristics of stem cells is quite relevant for modern regenerative medicine. Choosing the optimal source of cells, assessing their quality, taking into account the age and health status of the donor, determine the overall effectiveness and safety of cell therapy.
The aging is associated with altered immune and metabolism dysfunctions, increased inflammation and significant changes in the physiological levels of sex hormones. Multiple pathogenic pathways induce defective adipogenesis, inflammation, aberrant adipocytokine production, and insulin resistance, leading to age-related adipose tissue dysfunction with the functional decline of adipocyte progenitors and accumulation of senescent cells (Ou et al., 2022). Under ovariectomy conditions in mice, the proliferative capacity and osteogenic potential of ADSCs are significantly impaired compared to normal animals (Wang et al., 2017). In mice after ovariectomy, enhanced adipogenic differentiation of ADSCs is likely to be the important cause for increased adipogenesis in vivo and subsequent obesity-like changes in body mass (Fu et al., 2014). The regenerative potential of ADSCs in conditions of age-related estrogen deficiency is also impaired. In particular, ADSCs from aged estrogen-deficient ovariectomized rats have less capacity to increase tenocyte proliferation and healing in indirect co-culture system compared with normal ADSCs (Veronesi et al., 2015).
In our previous study, it was demonstrated that there were no statistically significant differences in the expression of all typical surface markers of ADSCs in young and old mice (Ustymenko et al., 2019). This fact indicates that the immunophenotype of ADSCs and, probably, their quantity in adipose tissue do not change with age, that confirm in studies of both human and animal samples published by other researchers.
In the study of Li et al. [ 2021] there were no significant differences in most of the surface markers between ADSCs from 1-month-old or 20-month-old mice, implying that the adipose tissue obtained from young animals had a comparative yield of ADSCs to that of old mice under the same conditions. No significant differences were found in the expression of surface markers on ADSCs from rats aged 2-, 9- and 24 months in the study of Muñoz et al. [ 2020]. At the same time, ADSCs derived from horses older than 5 years old exhibited several molecular alternations which markedly limit their regenerative capacity. Aged ADSCs were characterized by increased gene expression of pro-inflammatory cytokines and miRNAs (IL-8, IL-1β, TNF-α, miR-203b-5p, and miR-16-5p), as well as apoptosis markers (p21, p53, caspase-3, caspase-9; Alicka et al., 2020).
Liu et al. [ 2017] showed while human ADSCs from different age populations are phenotypically similar, they present major differences at the functional level. Advancing age was found to have a significant negative effect on the adipogenic and osteogenic differentiation potentials of human ADSCs (Liu et al., 2017). Zhang et al. [ 2011] suggested that advanced age and comorbidity do not negatively impact isolation of ADSCs, and these stem cells retain significant capacity to acquire key endothelial cell phenotype throughout life. Chen et al. [ 2012] found that the doubling time of ADSCs from both age groups was maintained below 70 h and authors concluded that the proliferation and osteogenic differentiation of ADSCs were less affected by age and multiple passages than in bone marrow-derived MSCs cultures.
The authors explain impairment of proliferative and differentiation potential by the existence of internal changes in ADSCs during aging, which is associated with senescence associated secretory phenotype (SASP) during chronic inflammation and metabolic disorders. The components of the pathologic secretory phenotype are quite heterogeneous and may depend on the cell type. In particular, it has been shown that increased levels of certain cytokines, chemokines and growth factors (IL-4, IL-13, IL-17, CCL3, CCL25 and GM-CSF) can characterize the SASP profile for ADSCs from elderly donors (Li et al., 2021).
The 3D cultivation of ADSCs in the form of spheroids, which partially simulates the conditions of the stem cells’ micro-environment (niche), can significantly improve their survival in the recipient’s tissue and increase the overall regenerative potential. The study of the ultrastructural characteristics of cells in the 3D spheroids is aimed at establishing the mechanisms of cell self-organization, contact intercellular signaling, production of the extracellular matrix, resistance to hypoxia, depending on the size of the spheroid.
It has been shown that 3D spheroid culture reduces size by increasing the secretion of extracellular vesicles. This event is mediated by a decrease in actin polymerization (F-actin). Probably, the large size of ADSCs spheroid cultures from old animals in our study indicates a violation of the ability to release microvesicles into the extracellular space (Mo et al., 2018).
In the study of Li et al. [ 2022] scanning electron microscopy showed that surfaces of spheroids formed in simulated microgravity culture system were relatively smooth and organized in a regular, granular shape, which may be beneficial for ever exchange of nutrients and gases evenly. Lee et al. [ 2021] proposed method of hybridization of ADSCs spheroids with polydopamine coated single-segmented fibers to enhance viability regardless of sizes and increase their functionality by regulating the size of spheroids. Transmission electron microscopy images showed that cell-only spheroids exhibited disintegrated membranes and empty spaces. In contrast, the cell membranes in fiber incorporated spheroids were tightly bound with each other, and disconnected or empty regions were minimal (Lee et al., 2021).
Step-by-step directed differentiation of multipotent cells in 3D spheroids in vitro will allow to partially reproduce the physiological mechanisms of tissue formation in vivo and to obtain a spatially organized culture of cells capable of survival, proliferation, and differentiation into certain direction.
Baraniak and McDevitt [2012] showed that cell proliferation and differentiation potential of dissociated cells retrieved from spheroids of mesenchymal stem cells were compared to conventional adherent monolayer cultures. Cells that had been cultured within spheroids recovered morphology typical of cells cultured continuously in adherent monolayers and retained their capacity for multi-lineage differentiation potential. In fact, more robust matrix mineralization and lipid vacuole content were evident in recovered MSCs when compared to monolayers, suggesting enhanced differentiation by cells cultured as 3D spheroids (Baraniak and McDevitt, 2012).
Laschke et al. [ 2014] discovered that scaffolds seeded with osteogenic differentiated spheroids exhibited a markedly impaired vascularization caused by the lost ability of differentiated spheroids to form microvascular networks. This was associated with a reduced tissue incorporation of the implants and indicating the dedifferentiation of the spheroids under the given in vivo conditions. These findings indicate that osteogenic differentiation of ADSCs spheroids markedly impairs their vascularization capacity (Laschke et al., 2014).
In our previous study, it was shown that 3D grafts of ADSCs cultured in spheroids are able to improve bone tissue regeneration in a model of bone injury in mice. The grafts previously differentiated into osteogenic direction provide better morphological indicators of bone recovery, compared with the spheroids without prior differentiation. Intensive migration of cells from spheroids to an adhesive surface in vitro proves the ability of cells to survive in 3D culture. At the same time, the morphology of cells on the surface of spheroids under the influence of osteoinductive differentiation factors changes and proliferative activity decreases (Kyryk et al., 2022).
## Conclusion
Thus, we can suggest that ADSCs throughout life retain a significant amount in adipose tissue and a high functional potential in vitro, which can be effectively used in cell therapy strategies especially in elderly patients. At the same time, our data indicate the negative impact of ovariectomy on the ultrastructural characteristics of 3D spheroids formed by ADSCs. Our findings are particularly promising in the context of vigilance for potential therapeutic applications of ADSCs from old donors.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The animal study was reviewed and approved by the Ethics Committee of the Institute of Genetic and Regenerative Medicine (protocol no. 9-2021 dated December 15, 2021) and performed in accordance with the European Union Directive of 22 September 2010 ($\frac{2010}{63}$/EU) for the protection of animals used for scientific purposes.
## Author contributions
VK, GB, and GS: conceptualization. VK, AU, ES, IG, and OT: data collection and analysis. VK, AU, OT, and IG: writing—original draft preparation. VK, GB, and GS: writing—review and editing. ES: visualization. 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. Alicka M., Kornicka-Garbowska K., Kucharczyk K., Kępska M., Röcken M., Marycz K.. **Age-dependent impairment of adipose-derived stem cells isolated from horses**. *Stem Cell Res. Ther.* (2020) **11** 4. DOI: 10.1186/s13287-019-1512-6
2. Amer M. G., Embaby A. S., Karam R. A., Amer M. G.. **Role of adipose tissue derived stem cells differentiated into insulin producing cells in the treatment of type I diabetes mellitus**. *Gene* (2018) **654** 87-94. DOI: 10.1016/j.gene.2018.02.008
3. Baraniak P. R., McDevitt T. C.. **Scaffold-free culture of mesenchymal stem cell spheroids in suspension preserves multilineage potential**. *Cell Tissue Res.* (2012) **347** 701-711. DOI: 10.1007/s00441-011-1215-5
4. Chen H.-T., Lee M.-J., Chen C.-H., Chuang S.-C., Chang L.-F., Ho M.-L.. **Proliferation and differentiation potential of human adipose-derived mesenchymal stem cells isolated from elderly patients with osteoporotic fractures**. *J. Cell. Mol. Med.* (2012) **16** 582-593. DOI: 10.1111/j.1582-4934.2011.01335.x
5. Dominici M., Le Blanc K., Mueller I., Slaper-Cortenbach I., Marini F., Krause D.. **Minimal criteria for defining multipotent mesenchymal stromal cells. The international society for cellular therapy position statement**. *Cytotherapy* (2006) **8** 315-317. DOI: 10.1080/14653240600855905
6. Eastell R., O’Neill T. W., Hofbauer L. C., Langdahl B., Reid I. R., Gold D. T.. **Postmenopausal osteoporosis**. *Nat. Rev. Dis. Primers* (2016) **2** 16069. DOI: 10.1038/nrdp.2016.69
7. Egger D., Tripisciano C., Weber V., Dominici M., Kasper C.. **Dynamic cultivation of mesenchymal stem cell aggregates**. *Bioengineering (Basel)* (2018) **5** 48. DOI: 10.3390/bioengineering5020048
8. Fu Y., Li R., Zhong J., Fu N., Wei X., Cun X.. **Adipogenic differentiation potential of adipose-derived mesenchymal stem cells from ovariectomized mice**. *Cell Prolif.* (2014) **47** 604-614. DOI: 10.1111/cpr.12131
9. Kyryk V., Kuchuk O., Klymenko P.. **Regenerative effects of mouse adipose-derived multipotent stromal cells in a micromass graft for the treatment of bone injury model**. *Anti Aging East. Europe* (2022) **1** 73-84. DOI: 10.56543/aaeeu.2022.1.1.11
10. Laschke M. W., Schank T. E., Scheuer C., Kleer S., Shadmanov T., Eglin D.. *Acta Biomater.* (2014) **10** 4226-4235. DOI: 10.1016/j.actbio.2014.06.035
11. Lee J., Lee S., Kim S. M., Shin H.. **Size-controlled human adipose-derived stem cell spheroids hybridized with single-segmented nanofibers and their effect on viability and stem cell differentiation**. *Biomater. Res.* (2021) **25** 14. DOI: 10.1186/s40824-021-00215-9
12. Li K., Shi G., Lei X., Huang Y., Li X., Bai L.. **Age-related alteration in characteristics, function and transcription features of ADSCs**. *Stem Cell Res. Ther.* (2021) **12** 473. DOI: 10.1186/s13287-021-02509-0
13. Li H., Wang C., Liu S., Guo Y., Chen J.. **Characterization of three-dimensional multipotent adipose-derived stem cell spheroids**. *Biocell* (2022) **46** 1705-1716. DOI: 10.32604/biocell.2022.018442
14. Liu M., Lei H., Dong P., Fu X., Yang Z., Yang Y.. **Adipose-derived mesenchymal stem cells from the elderly exhibit decreased migration and differentiation abilities with senescent properties**. *Cell Transplant.* (2017) **26** 1505-1519. DOI: 10.1177/0963689717721221
15. Ma X., Huang M., Zheng M., Dai C., Song Q., Zhang Q.. **ADSCs-derived extracellular vesicles alleviate neuronal damage, promote neurogenesis and rescue memory loss in mice with Alzheimer’s disease**. *J. Control. Release* (2020) **327** 688-702. DOI: 10.1016/j.jconrel.2020.09.019
16. Magenta A., Florio M. C., Ruggeri M., Furgiuele S.. **Autologous cell therapy in diabetes-associated critical limb ischemia: from basic studies to clinical outcomes (Review)**. *Int. J. Mol. Med.* (2021) **48** 173. DOI: 10.3892/ijmm.2021.5006
17. Mo M., Zhou Y., Li S., Wu Y.. **Three-dimensional culture reduces cell size by increasing vesicle excretion**. *Stem Cells* (2018) **36** 286-292. DOI: 10.1002/stem.2729
18. Muñoz M. F., Argüelles S., Marotta F., Barbagallo M., Cano M., Ayala A.. **Effect of age and lipoperoxidation in rat and human adipose tissue-derived stem cells**. *Oxid. Med. Cell. Longev.* (2020) **2020** 6473279. DOI: 10.1155/2020/6473279
19. Murohara T., Shintani S., Kondo K.. **Autologous adipose-derived regenerative cells for therapeutic angiogenesis**. *Curr. Pharm. Des.* (2009) **15** 2784-2790. DOI: 10.2174/138161209788923796
20. Ou M. Y., Zhang H., Tan P. C., Zhou S. B., Li Q. F.. **Adipose tissue aging: mechanisms and therapeutic implications**. *Cell Death Dis.* (2022) **13** 300. DOI: 10.1038/s41419-022-04752-6
21. Tsupykov O., Ustymenko A., Kyryk V., Smozhanik E., Yatsenko K., Butenko G.. **Ultrastructural study of mouse adipose-derived stromal cells induced towards osteogenic direction**. *Microsc. Res. Tech.* (2016) **79** 557-564. DOI: 10.1002/jemt.22670
22. Ustymenko A., Kyryk V., Lutsenko T., Tsupykov O., Butenko G.. **Morphofunctional properties of adipose-derived multipotent mesenchymal stromal cells in vitro in ovariectomized mice of different ages**. *Cell Organ Transplant.* (2019) **7** 158-167. DOI: 10.22494/cot.v7i2.102
23. Veronesi F., Della Bella E., Torricelli P., Pagani S., Fini M.. **Effect of adipose-derived mesenchymal stromal cells on tendon healing in aging and estrogen deficiency: an**. *Cytotherapy* (2015) **17** 1536-1544. DOI: 10.1016/j.jcyt.2015.07.007
24. Wang L., Huang C., Li Q., Xu X., Liu L., Huang K.. **Osteogenic differentiation potential of adipose-derived stem cells from ovariectomized mice**. *Cell Prolif.* (2017) **50** e12328. DOI: 10.1111/cpr.12328
25. Yu G., Wu X., Kilroy G., Halvorsen Y.-D. C., Gimble J. M., Floyd Z. E.. **Isolation of murine adipose-derived stem cells**. *Methods Mol. Biol.* (2011) **702** 29-36. DOI: 10.1007/978-1-61737-960-4_3
26. Zhang P., Moudgill N., Hager E., Tarola N., Dimatteo C., McIlhenny S.. **Endothelial differentiation of adipose-derived stem cells from elderly patients with cardiovascular disease**. *Stem Cells Dev.* (2011) **20** 977-988. DOI: 10.1089/scd.2010.0152
27. Zhou Z., Tian X., Mo B., Xu H., Zhang L., Huang L.. **Adipose mesenchymal stem cell transplantation alleviates spinal cord injury-induced neuroinflammation partly by suppressing the Jagged1/Notch pathway**. *Stem Cell Res. Ther.* (2020) **11** 212. DOI: 10.1186/s13287-020-01724-5
|
---
title: LINC00659 cooperated with ALKBH5 to accelerate gastric cancer progression by
stabilising JAK1 mRNA in an m6A‐YTHDF2‐dependent manner
authors:
- Yuan Fang
- Xi Wu
- Yunru Gu
- Run Shi
- Tao Yu
- Yutian Pan
- Jingxin Zhang
- Xinming Jing
- Pei Ma
- Yongqian Shu
journal: Clinical and Translational Medicine
year: 2023
pmcid: PMC9982078
doi: 10.1002/ctm2.1205
license: CC BY 4.0
---
# LINC00659 cooperated with ALKBH5 to accelerate gastric cancer progression by stabilising JAK1 mRNA in an m6A‐YTHDF2‐dependent manner
## Abstract
LINC00659 recruits ALKBH5 to form a complex that binds to JAK1 and decreases the m6A levels of JAK1 mRNA.The removal of JAK1 m6A modifications promotes the RNA stability of JAK1 that upregulates expression of JAK1.Upregulated JAK1‐activated STAT3 and JAK1/STAT3 pathway promote GC progression.
### Background
N6‐methyladenosine (m6A) RNA modification is known as a common epigenetic regulation form in eukaryotic cells. Emerging studies show that m6A in noncoding RNAs makes a difference, and the aberrant expression of m6A‐associated enzymes may cause diseases. The demethylase alkB homologue 5 (ALKBH5) plays diverse roles in different cancers, but its role during gastric cancer (GC) progression is not well known.
### Methods
The quantitative real‐time polymerase chain reaction, immunohistochemistry staining and western blotting assays were used to detect ALKBH5 expression in GC tissues and human GC cell lines. The function assays in vitro and xenograft mouse model in vivo were used to investigate the effects of ALKBH5 during GC progression. RNA sequencing, MeRIP sequencing, RNA stability and luciferase reporter assays were performed to explore the potential molecular mechanisms involved in the function of ALKBH5. RNA binding protein immunoprecipitation sequencing (RIP‐seq), RIP and RNA pull‐down assays were performed to examine the influence of LINC00659 on the ALKBH5–JAK1 interaction.
### Results
ALKBH5 was highly expressed in GC samples and associated with aggressive clinical features and poor prognosis. ALKBH5 promoted the abilities of GC cell proliferation and metastasis in vitro and in vivo. The m6A modification on JAK1 mRNA was removed by ALKBH5, which resulted in the upregulated expression of JAK1. LINC00659 facilitated ALKBH5 binding to and upregulated JAK1 mRNA depending on an m6A‐YTHDF2 manner. Silencing of ALKBH5 or LINC00659 disrupted GC tumourigenesis via the JAK1 axis. JAK1 upregulation activated the JAK1/STAT3 pathway in GC.
### Conclusion
ALKBH5 promoted GC development via upregulated JAK1 mRNA expression mediated by LINC00659 in an m6A‐YTHDF2‐dependent manner, and targeting ALKBH5 may be a promising therapeutic method for GC patients.
## INTRODUCTION
Gastric cancer (GC) is one of the most frequent cancers worldwide and is easy to diagnose due to its great features. 1, 2 Unfortunately, most patients are diagnosed in an advanced stage with multiple metastases when opportunities for radical surgery are lost. Because the 5‐year survival rate for GC is less than $10\%$, 3, 4, 5 it is necessary to further elucidate the mechanisms of GC progression to develop novel therapeutic targets.
N6‐methyladenosine (m6A) refers to the methylation of adenosine at position 6, and it is one of the most common and best‐characterised modifications. 6, 7 Abnormal m6A modification may influence noncoding RNA expression, which affects the progression of cancer. 8 As we known, m6A modification can be deposited by the m6A methyltransferase METTL3–METTL14–WTAP complex 9, 10, 11 and reversibly removed by two demethylases, fat mass and obesity‐associated protein (FTO) and alkB homologue 5 (ALKBH5). 12, 13 *It is* also recognised by special proteins named ‘readers’, including the YT521‐B homology (YTH) domain‐containing family (YTHDFs) and the insulin‐like growth factor 2 mRNA‐binding protein (IGF2BPs) family. 14 Emerging evidence has revealed that the aberrant expression of m6A‐associated enzymes, such as ALKBH5, may lead to tumourigenesis, including lung cancer, pancreatic cancer and hepatocellular carcinoma, and these enzymes are promising therapeutic targets. 15, 16, 17, 18, 19 However, further research is required to explain how ALKBH5 mediates GC progression.
The role of ALKBH5 in GC is controversial. A previous study indicated that ALKBH5 was a cancer‐promoting gene in GC. 20 However, Hu et al. 21 showed that ALKBH5 suppressed the invasion of GC via the ALKBH5–PKMYT1–IGF2BP3 axis. Therefore, the role of this enzyme remains controversial, possibly due to the different cancer models, and the mechanism of action of ALKBH5 in GC needs further investigation.
The Janus kinase (JAK)–STAT pathway plays an essential role in cancer development. JAK1 is a member of the JAK family that activates the free STAT molecule distributed in the cytoplasm. 22 A number of seminal papers showed that the JAK/STAT pathway promoted the survival and proliferation of tumour cells and a variety of other cancer‐related hallmarks over the last decade. 23 JAK1/STAT3 pathway activation significantly promotes GC cell proliferation and invasion. 24 The m6A modification regulates the formation and function of noncoding RNAs. Noncoding RNAs also have regulatory effects on m6A modifications. 25 Zhang et al. 15 found that the long noncoding RNA a long noncoding RNA antisense to FOXM1 (FOXM1‐AS) promoted the interaction between ALKBH5 and FOXM1, which contributed to the subsequent demethylation of FOXM1.
The present study found an oncogenic role of ALKBH5 in GC proliferation and metastasis. We also found that ALKBH5 caused m6A removal in JAK1 mRNA with the help of LINC00659, which enhanced JAK1 mRNA stability and contributed to the upregulation of JAK1 in GC. In summary, our work demonstrated that LINC00659 functioned in the modification of ALKBH5‐mediated JAK1 mRNA, which was involved in GC progression, and these results provide a new treatment target for GC.
## Data acquisition, preprocessing and bioinformatic analysis
To evaluate ALKBH5 mRNA expression in GC, transcripts per million values of ALKBH5 were obtained from 414 GC samples and 36 adjacent normal tissues uploaded to The Cancer Genome Atlas (TCGA) portal (https://portal.gdc.cancer.gov/).
We acquired the transcriptome profiling data and corresponding survival outcomes of 300 GC samples from the Asian Cancer Research Group (ACRG) study for further analysis (downloaded from Gene Expression Omnibus [GSE62254]), and patients lost to follow‐up were excluded during survival analysis. We used the survminer R package to determine the optimal cutoff value. The probe IDs were mapped to gene symbols based on the GPL570 platform annotation file (Affymetrix Human Genome U133 Plus 2.0 Array), the maximal probe measurement was used to determine the final gene expression of ALKBH5 (probe ID: 234302_s_at). The RNA sequencing (RNA‐seq) and microarray data used in this study were normalised and log2‐transformed prior to use.
In order to explore the potential roles of ALKBH5 in clinical GC samples, we extracted the expression matrix of tumour samples with the lowest and highest ALKBH5 expression from ACRG using the decile method and identified 541 significantly upregulated genes using the Limma R package (false discovery rate [FDR] <0.05). Next, these genes were used to perform Gene Ontology (GO) enrichment analysis, and cancer hallmark‐related pathways are displayed in a Circos plot.
## Specimens and cell culture
We collected the GC tissues and adjacent normal gastric tissues from patients with GC. All the patients with GC needed surgical treatment and were hospitalised in the Affiliated People's Hospital of Jiangsu University. The study received ethical approval from Nanjing Medical University (2018‐SRFA‐074) and Jiangsu University Affiliated People's Hospital (K20180016Y), which was implemented in accordance with the Helsinki Declaration of Principles. All gastric tissues and paired adjacent tissues were surgically removed and used for RNA and protein extraction and immunohistochemistry (IHC) analysis. Details of the antibodies used for IHC are shown in Table S2. To calculate the total score for ALKBH5 or JAK1 immunostaining, we first assessed the proportion of positively stained tumour cells (PP, 0–4) and the staining intensity (SI, 0–3) according to following criteria. We scored the PP in four categories: 0 ($0\%$), 1 ($5\%$–$25\%$), 2 ($25\%$–$50\%$) and 3 (>$51\%$). SI was scored on a scale of 0–3 (0, negative; 1, weak; 2, moderate; 3, strong). We multiplied the SI and PP scores to obtain a staining score, ranging from 0 to 9. There were two groups of positive levels of IHC staining: the low ALKBH5/JAK1 (0–3) group and high ALKBH5/JAK1 group (4–9). The human GC cell lines AGS, BGC‐823, MGC‐803, SGC‐7901 and normal gastric epithelial cell line GES‐1 were procured from Shanghai Cell Bank Library (Shanghai, China) and cultured in RPMI 1640 (BI, Israel) supplemented with $10\%$ foetal bovine serum (BI) at 37°C under $5\%$ CO2 conditions.
## Cell proliferation and colony formation assay
Processes of the colony formation assay and cell counting kit‐8 (CCK8) assays are described in a previously published article. 26
## Transwell assay
Transwell assays were performed in accordance with a previously published article. 26
## RNA pull‐down assay
The RNAmax‐T7 Kit (RiboBio) and RNeasy Mini Kit (QIAGEN) were used to obtain and purify biotinylated RNAs. An RNA protein pull‐down kit from Pierce (Thermo, USA) was used for pull‐down assays. Biotinylated RNAs were rotated with streptavidin agarose beads for 2 h. BGC‐823 cells were lysed using IP Lysis Buffer (Pierce). After several washes, the mixture of the lysate, protease/phosphatase inhibitor cocktail, RNase inhibitor and streptavidin agarose beads was rotated at 4°C for 1 h. After washing, protein was obtained from beads after 30 min of rotation in elution buffer at 37°C.
## RNA binding protein immunoprecipitation assay
Antibodies against ALKBH5 (Abcam, #ab195377) and YTHDF2 (Abcam, #ab220163) were used in RNA binding protein immunoprecipitation (RIP) assays. The detailed method of RIP assay is described in a previously published paper. 27
## RNA stability
We treated different groups of GC cells with actinomycin D (2 µg/mL) when cells reached $40\%$–$50\%$ confluence and harvested for RNA extraction after 0, 6 and 12 h of treatment.
## Animal studies
Nanjing Medical University's Committee on Ethics for Animal Experiments approved the experimental procedures (IACUC‐1706007). All mice used were on the BALB/c nude background. The Nanjing Medical University Animal Center provided 6–8‐week‐old female BALB/c nude mice for our research. BGC‐823 (3 × 106) cells that stably expressed or silenced ALKBH5 and LINC00659 and their paired control cells were injected into the left side of each mouse. Every 5 days, the tumour volume was measured, mice were euthanised 20 days after injection and the weights were measured. Tumours were used for RNA extraction, western blotting (WB) and IHC assays.
For cell metastasis experiments in vivo, BGC‐823 (3 × 106) cells that stably overexpressed or silenced ALKBH5 and LINC00659 and their paired control cells were injected through tail vein. All mice were euthanised, and metastatic organs were removed after 56 days of injection. Their lungs and livers were collected for further analysis (such as WB, haematoxylin and eosin [H&E] staining and IHC staining).
More details of the remaining materials and methods are available in Supporting Information.
## ALKBH5 is overexpressed in GC tissues and associated with poor survival
To examine whether ALKBH5 was dysregulated in GC tissues, we analysed the entire TCGA GC database, and the results demonstrated that ALKBH5 was highly expressed in 414 GC tumour samples compared to 37 normal gastric mucosa samples (Figure 1A). We performed quantitative real‐time polymerase chain reaction (qRT‐PCR) and IHC staining in 67 pairs of GC tissues and normal mucosa, and we found that ALKBH5 was present in GC tissues in much greater amounts than in normal tissues (Figure 1B,C). WB assays also showed high expression levels of ALKBH5 in GC tissues (seven of eight) (Figure 1D).
**FIGURE 1:** *AlkB homologue 5 (ALKBH5) is highly expressed in gastric cancer (GC) tissues and is associated with poor prognosis. (A) Whole TCGA (The Cancer Genome Atlas) GC database demonstrated that ALKBH5 was upregulated in 414 GC tumour samples compared with 37 normal samples. (B) Representative immunohistochemistry (IHC) staining showed ALKBH5‐positive stained cells in GC tissues and matched adjacent gastric tissues (scale bars = 50 µm). (C) The levels of ALKBH5 expression in GC and matched adjacent gastric tissues were detected by quantitative real‐time polymerase chain reaction (qRT‐PCR) (n = 67). The PCR results were normalised to the expression of β‐actin. (D) ALKBH5 protein levels were detected in GC tissues (T) and matched adjacent gastric tissues (N) by western blotting (n = 8). Relative protein levels were normalised to β‐actin. (E) The N6‐methyladenosine (m6A) RNA levels in six GC tissues and matched adjacent gastric tissues were checked by colorimetric ELISA assay using the m6A RNA methylation quantification kit. (F–G) Data from the Asian Cancer Research Group (ACRG) study showed that OS, RFS and DFS of patients with ALKBH5 high expression were shorter. (I) Kaplan–Meier analysis revealed DFS in GC patients based on the relative ALKBH5 expression (ALKBH5‐High, n = 48; ALKBH5‐Low, n = 19). This analysis was based on our IHC cohort. *p < .05; **
p < .01; ***
p < .001.*
According to the level of ALKBH5 in 67 pairs of GC tissues, we divided the samples into ALKBH5‐High ($$n = 48$$) and ALKBH5‐Low ($$n = 19$$) groups. Subsequently, we surprisingly observed that ALKBH5 expression level positively correlated with more aggressive clinicopathological characteristics (such as histological differentiation, invasion depth, Tumor, Node, Metastasis (TNM) stage and lymphatic metastasis) (Table 1). The negative correlation between ALKBH5 expression and global m6A levels was confirmed in six fresh human GC tissues (Figure 1E). Transcriptome profiling data of 300 GC samples and corresponding survival outcomes obtained from the ACRG study showed that the Overall survival (OS), Recurrence‐Free‐Survival (RFS) and Disease‐Free‐Survival (DFS) of ALKBH5 high group were shorter (Figure 1F–H). Similarly, our GC cohort data also indicated that patients in the ALKBH5‐High group had shorter DFS, as shown by Kaplan–Meier survival curves (Figure 1I). Collectively, our findings revealed that ALKBH5 was highly present in GC and could serve as a prognostic biomarker for GC patients.
**TABLE 1**
| Unnamed: 0 | ALKBH5 expression | ALKBH5 expression.1 | Unnamed: 3 |
| --- | --- | --- | --- |
| Clinical parameter | High (n = 48) | Low (n = 19) | χ 2 test (p‐value) |
| Age (years) | | | .332387 |
| <60 | 12 | 7 | |
| >60 | 36 | 12 | |
| Gender | | | .403892 |
| Male | 33 | 15 | |
| Female | 15 | 4 | |
| Size | | | .365532 |
| <5 cm | 31 | 10 | |
| >5 cm | 17 | 9 | |
| Histologic differentiation | | | .017747* |
| Moderate | 13 | 11 | |
| Poor | 35 | 8 | |
| Invasion depth | | | .025682* |
| T1/T2 | 16 | 12 | |
| T3/T4 | 32 | 7 | |
| TNM stages | | | .029856* |
| I/II | 10 | 9 | |
| III/IV | 38 | 10 | |
| Lymphatic metastasis | | | .029945* |
| Yes | 36 | 9 | |
| No | 12 | 10 | |
## ALKBH5 promotes GC cell proliferation and metastasis in vitro and in vivo
Firstly, we evaluated the expression of ALKBH5 in GC cell lines and GES1 normal gastric epithelial cell lines (Figure S1a,b). Because ALKBH5 was highly overexpressed in BGC‐823 and MGC‐803 cells, we chose these cells for further study. We stably silenced ALKBH5 (sh‐NC group and shALKBH5 group) and overexpressed ALKBH5 (Negative control (NC) group and ALKBH5 group) in BGC‐823 and MGC‐803 cells, which were confirmed using qPCR and WB assays (Figure S1c–f). The abundance of m6A was colorimetrically measured in GC cells using ELISA. We found that ALKBH5 significantly decreased the m6A level in BGC‐823 and MGC‐803 cells (Figure S1g,h).
GO analysis was performed using 541 significantly upregulated genes of tumour samples with the lowest and highest ALKBH5 expression from ACRG. Cancer hallmark‐related pathways, including cancer proliferation and metastasis, are shown in the Circos plot, which indicated a close association of ALKBH5 and cancer proliferation and metastasis (Figure S1i). According to the results of CCK8, EdU and colony formation assays, the ability of GC cell proliferation was inhibited by ALKBH5 knockdown in vitro (Figures 2A,C and S1j). We observed that knockdown of ALKBH5 inhibited the invasion and migration of GC cells by performing Transwell assays (Figure 2E). In contrast, overexpressing ALKBH5 benefited GC cell growth in vitro (Figures 2B,D and S1k). Transwell assays showed that overexpressing ALKBH5 enhanced the abilities of GC cells to migrate and invade in vitro compared to the NC groups (Figure 2F). Based on Hu et al. ’s study, 21 AGS and NCI‐N87 cells were also tested for colony formation, CCK8, and Transwell migration after overexpressing or knocking down ALKBH5. The expression of ALKBH5 in each group of AGS and NCI‐N87 cells was detected using WB assays (Figure S2a,b). The same results of functional biology in AGS and NCI‐N87 cells were obtained as in MGC‐803 and BGC‐823 cells (Figure S2c–j).
**FIGURE 2:** *AlkB homologue 5 (ALKBH5) promotes gastric cancer (GC) proliferation and metastasis in vivo and in vitro. (A) Cell counting kit‐8 (CCK8) assays were conducted to determine cell proliferation of BGC‐823 and MGC‐803 cells with stable silenced ALKBH5. (B) CCK8 assays were performed to determine cell proliferation of BGC‐823 and MGC‐803 cells with stable ALKBH5 overexpression. (C and D) EdU assays were conducted to determine cell proliferation of BGC‐823 and MGC‐803 cells with stably silenced ALKBH5 (C) and overexpressed ALKBH5 (D) (scale bars = 100 mm). (E and F) Cell migration and invasion assays of BGC823 cells and MGC‐803 cells were performed by Transwell assays after knockdown (E) or overexpression (F) of ALKBH5 (scale bars = 100 mm). (G) The images of dissected tumours from BGC‐823 cells stably transfected with sh‐NC and sh‐ALKBH5. (H and I) Tumour weights and sizes are represented as means of tumour weights (I)/sizes (H) ± standard deviation (SD). (J) Tumour tissue samples were immunostained for haematoxylin and eosin (H&E), Ki‐67 and ALKBH5 (magnification, ×400, scale bars = 10 µm). (K–M) Overexpression of ALKBH5 promoted the growth of xenografted tumours. BGC‐823 cells with NC or ALKBH5 overexpression were subcutaneously injected into nude mice. (N) Representative immunohistochemistry results and quantification of Ki‐67 and ALKBH5‐positive staining in tumours (magnification, ×400, scale bars = 10 µm). (O and P) Visualisation of the entire lung, and H&E‐stained lung sections. Knockdown of ALKBH5 in BGC‐823 cells markedly suppressed GC lung metastasis in nude mice (n = 4) (scale bars = 200 mm). (Q) Representative images of the metastatic nodes in the livers (scale bars = 200 mm). (R) H&E‐stained liver sections (scale bars = 200 µm). *
p < .05; **
p < .01; ***
p < .001.*
To investigate the function of ALKBH5 in vivo, we established a mouse xenograft model by subcutaneously injecting nude mice with BGC‐823–shALKBH5, BGC‐823–ALKBH5 and the corresponding control cells. Consistent with the in vitro results, shALKBH5‐1 and shALKBH5‐2 tumours were significantly smaller and lighter than control tumours (Figure 2G‐I). IHC assays showed decreased expression of Ki‐67 (a key biomarker in tumour growth) and suppressed expression of ALKBH5 in the shALKBH5‐1 and shALKBH5‐2 groups (Figure 2J), which were consistent with the WB results (Figure S1l). Tumours in the BGC‐823–ALKBH5 group had a larger volume heavier weight (Figure 2K‐M), and higher expression of Ki‐67 and ALKBH5 (Figures 2N and S1m).
We explored the metastasis potential of ALKBH5 in vivo by injecting BGC‐823 cells with stable knockdown or overexpression of ALKBH5 and control cells into nude mice via the tail vain. H&E staining assays showed that the number of metastatic nodes in lung or liver was much lower in the ALKBH5‐downregulated group (Figure 2O,P). However, ALKBH5 overexpression resulted in more lung and liver metastasis (Figures 2Q,R and S1n).
Collectively, all the evidence suggested that ALKBH5 acted as a cancer‐promoting regulator in GC in vitro and in vivo.
## JAK1 was the downstream target of ALKBH5
To examine the specific mechanism of ALKBH5 in GC tumourigenesis, we performed MeRIP sequencing (MeRIP‐seq) and RNA‐seq in ALKBH5‐knockdown and control BGC‐823 cells. MeRIP‐seq revealed that the m6A peaks of 6361 transcripts were increased after ALKBH5 knockdown (fold change >2; FDR <0.05), which indicated a global gain of m6A methylation in mRNA transcripts with ALKBH5 knockdown. A minor change in the pattern of m6A peak distribution was detected. Most m6A peaks were located in the exon ($32.98\%$) and intron ($33.01\%$) areas (Figure 3A). These differential m6A peaks of transcripts in the two groups are shown in Figure 3B. The m6A consensus motif GGAC was highly enriched in ALKBH5‐knockdown and control BGC‐823 cells (Figure 3C). RNA‐seq showed that 9108 transcripts were changed in ALKBH5‐knockdown BGC‐823 cells, among which 3774 transcripts were increased (fold change >2), and 5334 transcripts were reduced compared to the control (fold change <0.5; FDR <0.05) (Figure 3D,E). These transcripts were from 1769 genes.
**FIGURE 3:** *JAK1 was the downstream target of alkB homologue 5 (ALKBH5). (A) The pattern of N6‐methyladenosine (m6A) peak distribution was slightly changed after ALKBH5 knockdown according to MeRIP‐seq data. (B) The locations of m6A peaks from 6598 transcripts. (C) The m6A motif detected by the MEME motif analysis with MeRIP‐seq data. (D) Heatmap of differentially expressed mRNA identified by RNA‐seq after ALKBH5 knockdown. (E) RNA‐seq results identified 3774 upregulated mRNAs and 5334 downregulated mRNAs in ALKBH5‐knockdown BGC‐823 cells. (F) Flowchart demonstrates the selection process of the downstream target of ALKBH5. (G) The RNA expression levels of JAK1 in ALKBH5‐deficient BGC‐823 cells and MGC‐803 cells were examined by quantitative real‐time polymerase chain reaction (qRT‐PCR). (H) The RNA expression levels of JAK1 in ALKBH5‐overexpressing BGC‐823 cells and MGC‐803 cells were examined by qRT‐PCR. (I) The protein expression levels of JAK1 in ALKBH5‐deficient BGC‐823 cells and MGC‐803 cells were examined by western blotting. Relative protein levels were normalised to GAPDH. (J) The protein expression levels of JAK1 in ALKBH5‐overexpressing BGC‐823 cells and MGC‐803 cells were examined by western blotting. Relative protein levels were normalised to GAPDH. (K) The levels of JAK1 expression in gastric cancer (GC) and matched adjacent gastric tissues were detected by qRT‐PCR (n = 67). (L) ALKBH5 expression was positively correlated with JAK1 expression in GC cohort (linear regression). (M) ALKBH5 expression was positively correlated with JAK1 expression in TCGA database. (N) Online bioinformatics tool Kaplan–Meier plotter found that GC patients with increased expression of JAK1 had significantly reduced OS. *
p < .05; **
p < .01; ***
p < .001.*
A total of 137 genes were shared between the RNA‐seq data and MeRIP‐seq input library data. We overlapped 137 genes with 6361 transcripts, m6A peaks were increased after ALKBH5 knockdown, and there were 39 mutual genes. *These* genes were enriched in different signalling pathways, such as transforming growth factor‐beta receptor signalling and interleukin (IL)‐4 and IL‐13 signalling. JAK1 and FOS were found in all signalling pathways (Table S3). We further examined the expression of JAK1 and FOS in ALKBH5‐knockdown GC cells and found that JAK1 was decreased in BGC‐823 and MGC‐803 cells after ALKBH5 knockdown (Figure S3a,b). Therefore, we selected JAK1 for further study (Figure 3F). We examined the expression of JAK1 in ALKBH5‐knockdown and ALKBH5‐overexpressing cells using qRT‐PCR. We observed significant downregulation of JAK1 transcripts in ALKBH5‐knockdown cells, and ALKBH5 overexpression increased JAK1 mRNA in BGC‐823 and MGC‐803 cells (Figure 3G,H). These results were further confirmed using WB (Figure 3I,J). Therefore, ALKBH5 regulated the expression of JAK1. We examined the expression of other members of the JAK family, such as JAK2 and JAK3, using qRT‐PCR and found no difference after ALKBH5 knockdown in GC cells (Figure S3c,d). We also performed MeRIP‐qPCR. After silencing ALKBH5, JAK$\frac{2}{3}$ mRNA was enriched with an anti‐m6A antibody but was not different from the control group (Figure S3e,f). Therefore, we assumed that JAK2 and JAK3 mRNAs are not targets of ALKBH5.
We examined the relationship between JAK1 and GC by comparing the expression of JAK1 in 67 pairs of normal gastric mucosa and GC tissues (Figure 3K). An increase in JAK1 was observed in GC tissues, and its expression positively correlated with ALKBH5 (R 2 = 0.4221, $$p \leq .0004$$) (Figure 3L). The online TCGA database verified these results (Figure 3M). Figures 3N and S3g show that GC patients with increased JAK1 expression had significantly shorter OS and Progression‐Free‐Survival (PFS) than those with low JAK1 expression group using the online bioinformatics website Kaplan–Meier plotter.
## ALKBH5 removes m6A modifications from JAK1 mRNA and maintains JAK1 mRNA stability in a YTHDF2‐dependent manner
Because ALKBH5 influenced JAK1 mRNA levels, RIP assays in GC cells were performed, and the results showed that more JAK1 mRNA was enriched with anti‐ALKBH5 antibodies than immunoglobulin (IgG), which indicated direct binding between ALKBH5 and JAK1 (Figure 4A). RNA–protein colocalisation of JAK1 mRNA with ALKBH5 was performed, which indicated that ALKBH5 binds to JAK1 mRNA in the nucleus (Figure S4a). Because ALKBH5 is an m6A demethylase, we presumed that ALKBH5 regulated the expression of JAK1 by erasing m6A modifications. As shown in the MeRIP‐seq data, the m6A abundance of JAK1 mRNA was notably increased after ALKBH5 knockdown, and this peak was located in chr1, 65330528–65330698, which is the coding sequence of JAK1 (Figure 4B). The RRACH (R=G or A, H=A, C or U) sequences are known as the m6A binding sites in previous studies, 28 and the GGAC motif was found in JAK1 mRNA using MeRIP‐seq analysis. We performed MeRIP‐qPCR in BGC‐823 and MGC‐803 cells. After silencing ALKBH5, more JAK1 mRNA was enriched with the m6A antibody, which suggested that the m6A abundance of JAK1 mRNA was increased (Figure 4C). Consistently, overexpressing ALKBH5 caused the removal of m6A in JAK1 mRNA (Figure 4D). We treated GC cells with the nonspecific methylation inhibitor 3‐deazaadenosine (DAA) and found that JAK1 mRNA expression increased significantly after administration of DAA (Figure 4E).
**FIGURE 4:** *AlkB homologue 5 (ALKBH5) removes N6‐methyladenosine (m6A) modifications of JAK1 mRNA and maintains JAK1 mRNA stability in an YTHDF2‐dependent manner. (A) RNA binding protein immunoprecipitation (RIP)‐qPCR assays confirmed that JAK1 mRNA binding to ALKBH5. (B) m6A modification of JAK1 mRNA was visualised by Integrative Genomics Viewer (IGV) software after ALKBH5 knockdown. The significantly increased m6A peak is indicated by red rectangles. (C) The m6A abundances on JAK1 mRNA after ALKBH5 knockdown were checked by MeRIP‐qPCR. m6A on JAK1 mRNA was increased after ALKBH5 knockdown. (D) The m6A abundances on JAK1 mRNA after ALKBH5 overexpression were checked by MeRIP‐qPCR. m6A on JAK1 mRNA was decreased after ALKBH5 knockdown. (E) BGC‐823 and MGC‐803 cells were treated with a global methylation inhibitor (DAA), leading to the upregulation of JAK1 mRNA levels. (F) Six potential m6A sites were found on the sequences of raised m6A peaks of JAK1 transcripts after ALKBH5 knockdown according to MeRIP‐seq data. (G) All adenosines (A) were replaced with cytosines (C) in RRACH motif. These partial sequences of JAK1 mRNA were inserted with wild‐type or mutated m6A sites into luciferase reporter plasmids. (H) Relative luciferase activity (firefly/Renilla activity) of the wild‐type or mutant JAK1 luciferase reporter in 293 cells with ALKBH5 overexpression and the negative control. The results were normalised to the wild‐type with negative control group. (I‐L) ALKBH5 on mRNA stability of JAK1 was detected by use of actinomycin D (2 µg/mL) on BGC‐823 and MGC‐803 cells with stable ALKBH5 knockdown or overexpression. The results were normalised to the data of 0 h. (M and N) Knocking down YTHDF2 significantly increased the mRNA expression of JAK1 detected by quantitative real‐time polymerase chain reaction (qRT‐PCR). (O and P) RIP‐qPCR assays detected the change of binding capacity between JAK1 and YTHDF2 after ALKBH5 overexpression. (Q and R) YTHDF2 on mRNA stability of JAK1 was detected by use of actinomycin D (2 µg/mL) on GC cells with YTHDF2 knockdown. *
p < .05; **
p < .01; ***
p < .001.*
To verify the binding site between ALKBH5 and JAK1 mRNA, we found six potential m6A sites on the sequence (chr1, 65330528–65330698) of increased m6A peaks of JAK1 transcripts after ALKBH5 knockdown according to MeRIP‐seq data. This sequence was located in the exon of JAK1 (Figure 4F). We replaced all adenosines (A) with cytosines (C) in the RRACH motif and inserted a partial sequence of JAK1 mRNA with wild‐type (Wt) or mutant m6A sites into luciferase reporter plasmids (Figure 4G). ALKBH5 increased relative luciferase activity in the JAK1‐Wt group, but the JAK1‐Mut group had no effect on luciferase activity (Figure 4H).
In summary, ALKBH5 regulated the JAK1 expression level in an m6A‐dependent manner.
To further investigate the effect of m6A modification on JAK1 mRNA, we treated GC cells with actinomycin D, which restrained the transcription of RNA. JAK1 mRNA had a faster decay rate when ALKBH5 was silenced, and the mRNA was more stable with ALKBH5 overexpression, which indicated that m6A modification affected the stability of JAK1 mRNA (Figure 4I‐L).
YTHDF2 binds to the attenuation sites of mRNAs to cause the decay of certain mRNAs. 29, 30 A recent study reported that YTHDF2 recognised m6A modification and targeted thousands of transcripts, including JAK1. 30 To evaluate the effect of YTHDF2 on JAK1 mRNA, we transferred two small interfering RNAs (siRNAs) to silence the expression of YTHDF2 in GC cells. YTHDF2 knockdown significantly increased the mRNA level of JAK1 in GC cell lines (Figure 4M,N). We performed RIP‐qPCR assays in ALKBH5‐overexpressing GC cells. Overexpressing ALKBH5 decreased the enrichment between YTHDF2 and JAK1 mRNA, which indicated that YTHDF2 recognised the m6A modification of JAK1 mRNA and influenced its expression (Figure 4O,P). Knockdown of YTHDF2 stabilised JAK1 mRNA (Figure 4Q,R). Furthermore, we found that YTHDF2 deletion rescued the decrease in JAK1 levels caused by ALKBH5 knockdown in GC cells (Figure S4b).We have blocked YTHDF2 in the proliferation, invasion and migration of ALKBH5 knockdown cells in vitro and found that YTHDF2 deletion rescued the decrease in ALKBH5 knockdown‐mediated function in GC (Figure S4c–e).Taken together, our findings indicated that ALKBH5‐mediated m6A modification influenced JAK1 mRNA stability in a YTHDF2‐dependent manner.
## LINC00659 facilitates ALKBH5 binding to JAK1 mRNA
The lncRNA FOXM1‐AS promotes the interaction between ALKBH5 and FOXM1, and ALKBH5 decreases the m6A modification of FOXM1, 15 which suggests that specific lncRNAs are involved in the interaction between the demethylase ALKBH5 and mRNA. Based on this hypothesis, we examined whether specific lncRNAs helped ALKBH5 remove the m6A modification on JAK1 mRNA. RIP sequencing (RIP‐seq) in BGC‐823 cells showed 736 combination peaks from 194 lncRNAs that were enriched with an anti‐ALKBH5 antibody (FDR <0.05). Analysis of the TCGA database found that 1375 lncRNAs were upregulated in GC tissues. There were 29 lncRNAs that overlapped between the RIP‐seq results and TCGA database. We selected the top five lncRNAs according to the count number in RIP‐seq data (Figure 5A). We silenced these five lncRNAs in BGC‐823 cells and found that only the silencing of LINC00659 caused the downregulation of JAK1 mRNA, which suggested that LINC00659 regulated JAK1 (Figure S5a).
**FIGURE 5:** *LINC00659 facilitates alkB homologue 5 (ALKBH5) binding to JAK1 mRNA. (A) The Venn diagram shows the lncRNAs detected by RNA binding protein immunoprecipitation sequencing (RIP‐seq) and The Cancer Genome Atlas (TCGA) database; top five candidate lncRNAs are shown. (B) RIP‐qPCR confirmed LINC00659 binding to ALKBH5. (C) RNA–protein colocalisation of LINC00659 with ALKBH5 in BGC‐823 and MGC‐803 cells (scale bars = 25 µm). (D) RNA pull‐down assay confirmed ALKBH5 binding to LINC00659. The RNA and protein in the RNA–protein complex were then detected by western blotting. (E) RIP and western blotting assays using ALKBH5 antibody revealing the interaction between LINC00659 and ALKBH5 protein in BGC‐823 cells transfected with a series truncated form of LINC00659. The immunoglobulin G (IgG)‐bound RNA was taken as a negative control. (F) RNA pull‐down assay demonstrating the interaction between LINC00659 truncations and ALKBH5 protein in BGC‐823 cells. (G) The levels of LINC00659 expression in gastric cancer (GC) and matched adjacent gastric tissues were detected by quantitative real‐time polymerase chain reaction (qRT‐PCR) (n = 67). (H) LINC00659 expression level was analysed in GC and normal tissues in TCGA database. (I) Kaplan–Meier analysis revealed DFS in GC patients based on the relative LINC00659 expression (n = 67). (J) LINC00659 expression was positively correlated with JAK1 expression in GC cohort (linear regression). (K) RNA pulldown assays confirmed the physical interaction between JAK1 and LINC00659 in BGC‐823 cells. (L‐O) RIP‐qPCR assays confirmed that overexpressing LINC00659 could enhance the binding between ALKBH5 and JAK1, while suppressing LINC00659 could inhibit the binding in BGC‐823 and MGC‐803 cells. (P‐S) MeRIP‐qPCR analysis was performed to detect the LINC00659‐mediated m6A modifications on JAK1. The m6A modifications on JAK1 were enhanced when LINC00659 knockdown while decreased when LINC00659 overexpression in BGC‐823 and MGC‐803 cells. (T‐U) LINC00659 on mRNA stability of JAK1 was detected by use of actinomycin D (2 µg/mL) on BGC‐823 and MGC‐803 cells with stable LINC00659 knockdown. *
p < .05; **
p < .01; ***
p < .001.*
The silencing of LINC00659 also downregulated the protein expression of JAK1, and the overexpression of LINC00659 caused JAK1 upregulation (Figure S5b,c). These results indicated that LINC00659 promoted JAK1 expression.
Similar to the RIP‐seq results, RIP, colocalisation and pull‐down assays confirmed that ALKBH5 and LINC00659 directly interacted with each other (Figure 5B‐D). However, LINC00659 did not influence the mRNA or protein level of ALKBH5, and vice versa (Figure S5d–g). RIP‐seq data showed that LINC00659 transcript 2 was enriched with anti‐ALKBH5 antibody, which has three exons (exon 1: 1–51 bp; exon 2: 52–123 bp; exon 3: 124–572 bp). To investigate the specific region on LINC00659 where ALKBH5 bound, we divided LINC00659 into four parts (1–123 bp; 124–300 bp; 301–572 bp; 124–572 bp). The RIP and pull‐down assays showed that ALKBH5 primarily bound to exon 3 of LINC00659 (124–572 bp), which was mostly located at 301–572 bp (Figure 5E,F). To further investigate the specific regions of ALKBH5 that were responsible for LINC00659 binding, we performed RIP assays. Region 3 was mostly responsible for LINC00659 binding (Figure S5h,i).
To verify whether LINC00659 was upregulated in GC samples, RT‐qPCR assay was used to detect the expression level of LINC00659 in 67 pairs of GC and adjacent normal tissues. We found that LINC00659 had a high level of expression in GC tissues (Figure 5G). These results were consistent with the TCGA dataset, which indicated the overexpression of LINC00659 in GC tissues (Figure 5H).We separated 67 patients into LINC00659‐High and LINC00659‐Low groups according to the expression level of LINC00659 in GC samples. Kaplan–Meier survival curves revealed that high LINC00659 expression group had a shorter DFS than group with low LINC00659 expression (Figure 5I). Gene expression analysis demonstrated a positive correlation between LINC00659 and JAK1 in our cohort and the TCGA database (Figures 5J and S6a). The higher expression of LINC00659 in BGC‐823 and MGC‐803 was also observed by using RT‐qPCR (Figure S6b). Fluorescence in situ hybridisation and nuclear‐cytoplasmic fractionation analysis suggested that LINC00659 was primarily enriched nucleus in cell nucleus (Figure S6c–e).
Based on the positive correlation between upregulated LINC00659 in GC and JAK1 expression and the direct binding of LINC00659 to ALKBH5, we hypothesised that LINC00659 facilitated the ALKBH5–JAK1 interaction. The direct interaction between JAK1 and LINC00659 was confirmed by pull‐down assays (Figure 5K). We established ALKBH5 knockout BGC‐823 and MGC‐803 cells using the *Cas9* gene editing system. The mRNA and protein levels of JAK1 in GC cells were reduced following knockout of ALKBH5. However, ALKBH5 knockout and LINC00659 overexpression simultaneously did not return these levels to normal (Figure S6f–h). These results indicated that LINC00659‐promoted JAK1 expression relied on ALKBH5.
A series of RIP assays revealed that less JAK1 mRNA was enriched with an ALKBH5 antibody after silencing of LINC00659, and overexpressing LINC00659 resulted in an increased combination between ALKBH5 and JAK1 transcripts (Figure 5L‐O). The abundance of m6A on JAK1 mRNA was downregulated without LNC00659, and the opposite result was observed when LINC00659 was overexpressed (Figure 5P‐S). JAK1 mRNA was more stable in ActD‐treated BGC‐823 and MGC‐803 cells overexpressing LINC00659, and JAK1 mRNA exhibited a faster degradation rate when LINC00659 was silenced in ActD‐treated BGC‐823 and MGC‐803 cells (Figures 5T‐U and S6i,j).
*In* general, LINC00659 helped ALKBH5 bind to JAK1 mRNA, which removed m6A on JAK1 mRNA and enhanced its stability to cause the upregulation of JAK1.
## LINC00659 promotes GC proliferation and metastasis in vitro and in vivo
To verify the biological function of LINC00659 in GC progression, we stably knocked down and overexpressed LINC00659 in BGC‐823 and MGC‐803 cells for further function assays (Figure S7a,b). CCK8 and EdU assays showed that knockdown of LINC00659 prominently inhibited GC cell growth in vitro (Figure S7c,e). Similar results were detected in colony formation assays (Figure S7g). However, the opposite results were observed when LINC00659 was overexpressed (Figure S7d,f,h). The suppression of LINC00659 attenuated the in vitro metastasis and invasion of GC cells as indicated by Transwell assays (Figure S7i), and the overexpression of LINC00659 promoted the invasive and migratory capacities of GC cells (Figure S7j).
To investigate the potential roles of LINC00659 in GC cells in vivo, LINC00659‐deficient and LINC00659‐overexpressing BGC‐823 cells were used to establish nude mouse xenograft model. Compared to the control group, the weight and volume of tumours from the sh‐LINC00659 group were much smaller (Figure S8a–c). RT‐qPCR assays verified the knockdown of LINC00659 expression in mouse tumour tissues (Figure S8d). IHC staining also showed lower Ki‐67 signals in the LINC00659‐knockdown group (Figure S8e). The tumours harvested from the LINC00659‐overexpressing group were larger and heavier than tumours from the NC group (Figure S8h–j), with higher Ki‐67 protein levels (Figure S8l). The expression of LINC00659 in tumour tissues was evaluated using RT‐qPCR (Figure S8k).
We also injected the abovementioned cells into nude mice via the tail vein to investigate in vivo metastasis ability. Fewer metastatic nodes were detected in the lung in the LINC00659‐silenced group (Figure S8f, g), and more nodes were detected in the lung and liver in the LINC00659‐overexpressing group (Figure S8m–o) than in the NC group.
In conclusion, LINC00659 increased proliferating and metastatic GC cells.
## LINC00659 facilitates ALKBH5 to promote the GC progression by upregulating JAK1
To further examine whether LINC00659 and ALKBH5 affected the abilities of GC cells proliferation and invasion by modulating JAK1 expression, we performed following plenty of rescue experiments. We designed shRNAs to target JAK1 and confirmed the knockdown efficiency using RT‐qPCR (Figure 6A,B). Silencing JAK1 suppressed cell proliferation induced by ALKBH5 and LINC00659 (Figures 6E‐H,M‐N and S9a–d), and cell invasion and migration (Figure S10a–d). When JAK1 was overexpressed (Figure 6C,D),ALKBH5/LINC00659 knockdown inhibited GC cell proliferation, migration and invasion (Figures 6I‐L,O‐P, S9e–h and S10e–h). To validate the mechanism of LINC00659 facilitation of ALKBH5 upregulation of JAK1 mRNA expression, we designed a series of functional experiments. Silencing LINC00659 suppressed ALKBH5‐induced cell proliferation in vitro and cell invasion and migration (Figure S11a–c). The silencing of LINC00659 or JAK1 impaired the ALKBH5 overexpression‐induced increase in GC cell proliferation in vivo (Figure S11d–f). Sections of tumour xenografts from ALKBH5‐overexpressing or ALKBH5‐overexpressing BGC‐823 cells with LINC00659/JAK1 knockdown subcutaneously injected into nude mice were stained with JAK1 antibodies. Silencing LINC00659 or JAK1 appreciably supressed the increase of JAK1 staining induced by ALKBH5 (Figure S11g). In conclusion, LINC00659 facilitated ALKBH5 to promote the growth and invasion of GC cells in vivo and in vitro by upregulating JAK1.
**FIGURE 6:** *LINC00659 facilitates alkB homologue 5 (ALKBH5) to promote the gastric cancer (GC) progression by upregulating JAK1. (A‐D) The mRNA levels of JAK1 in GC cells with JAK1 knockdown or overexpression were detected by quantitative real‐time polymerase chain reaction (qRT‐PCR). (E‐H) cell counting kit‐8 (CCK8) assays were conducted to determine cell proliferation of ALKBH5/LINC00659‐overexpressing BGC‐823 and MGC‐803 cells transfected with the JAK1 shRNAs or their corresponding controls. (I‐L) CCK8 assays were conducted to determine cell proliferation of ALKBH5/LINC00659‐silencing BGC‐823 and MGC‐803 cells transfected with the JAK1‐overexpressing plasmids or their corresponding controls. (M and N) Colony‐formation assays were conducted to determine the colony‐formation ability of ALKBH5/LINC00659‐overexpressing GC cells transfected with the JAK1 shRNAs or their corresponding controls. (O and P) Colony‐formation assays were conducted to determine the colony‐formation ability of ALKBH5/LINC00659‐silencing GC cells transfected with the JAK1‐overexpressing plasmids or their corresponding controls. *
p < .05; **
p < .01; ***
p < .001.*
## JAK1 upregulation activates the JAK1/STAT3 pathway in GC
JAK1 is one of the JAK that phosphorylates proteins of the STAT family and plays a crucial role in multiple cancers, including GC. Previous researches have reported that the JAK1/STAT3 pathway activation promoted GC cell proliferation, invasion and metastasis. 24 To verify the target of JAK1, we detected the phosphorylation of STAT1, STAT3 and STAT5 after JAK1 silencing (shJAK1‐1, shJAK1‐2) in GC cells using WB and found a significant decrease in phosphorylated STAT3 but not STAT1 or STAT5 (Figure S12a,b). Therefore, we chose STAT3 as the substrate of JAK1 in GC cells. In contrast to the phenomenon that silencing JAK1 caused the downregulation of phosphorylated STAT3 (p‐STAT3) (Figure 7A), the phosphorylation of STAT3 was dramatically increased after overexpressing JAK1 in GC cells (Figure 7B). To further examine the role of ALKBH5 and LINC00659 in activating the JAK1/STAT3 pathway, we performed rescue experiments. WB demonstrated that ALKBH5/LINC00659 overexpression upregulated the expression of JAK1 and p‐STAT3, and suppression of JAK1 abrogated the effect of ALKBH5/LINC00659 on JAK1 upregulation and STAT3 activation (Figure 7C,D). In contrast, silencing ALKBH5/LINC00659 downregulated the expression of JAK1 and p‐STAT3, and overexpressing JAK1 rescued the suppression of JAK1 and p‐STAT3 caused by the knockdown of ALKBH5/LINC00659 (Figure 7E,F). In summary, JAK1, upregulated by ALKBH5 and LINC00659, phosphorylated STAT3 and activated the JAK1/STAT3 pathway, which accelerated the progression of GC.
**FIGURE 7:** *JAK1 upregulation activates JAK1/STAT3 pathway in gastric cancer (GC). (A and B) The protein levels of JAK1, p‐STAT3 and STAT3 were detected by western blotting in GC cells transfected with the JAK1 shRNAs (A) or JAK1‐overexpressing plasmids (B). Relative protein levels were normalised to β‐actin. (C and D) The protein levels of JAK1, p‐STAT3 and STAT3 in alkB homologue 5 (ALKBH5)/LINC00659‐overexpressing GC cells transfected with the JAK1 shRNAs or their corresponding controls were detected by western blotting. Relative protein levels were normalised to β‐actin or GAPDH. (E and F) The protein levels of JAK1, p‐STAT3 and STAT3 in ALKBH5/LINC00659‐silencing GC cells transfected with the JAK1‐overexpressing plasmids or their corresponding controls were detected by western blotting. Relative protein levels were normalised to β‐actin or GAPDH.*
## Clinical correlation between ALKBH5/LINC00659 and JAK1 in GC
For further investigation of the relationship between ALKBH5/LINC00659 and JAK1 in the GC, we validated the JAK1 expression in tumours harvested from nude mice with LINC00659/ALKBH5 abnormally expressed and the paired controlled ones. IHC staining assays demonstrated that knocking down LINC00659/ALKBH5 decreased JAK1 expression in tumours, while overexpressing LINC00659/ALKBH5 increased the expression of JAK1 (Figure S13a).
To further investigate the clinical relationship between ALKBH5/LINC00659 and JAK1 in GC, we examined JAK1 and ALKBH5 expression in neoplastic tissues from patients with GC and divided tissues into ALKBH5‐Low and ALKBH5‐High groups according to ALKBH5 expression. The expression of ALKBH5 in GC positively correlated with the expression of JAK1 (Figure S13b,c). In summary, the ALKBH5–LINC00659/m6A/JAK1 axis stimulated the development of GC.
## DISCUSSION
The m6A modification of mRNAs and noncoding RNA was the most common epigenetic modification among eukaryotic cells. 31 Increasing evidence indicates that m6A modifications are associated with tumourigenesis, metastasis and angiogenesis of gastrointestinal tumours. 32, 33, 34 Aberrant expression of m6A‐associated enzymes, such as m6A writers (the METTL3/METTL14/WTAP complex) and readers (YTHDF1, YTHDF2), often leads to GC progression. 35, 36, 37, 38, 39 Recent studies have revealed that the m6A demethyltransferase ALKBH5 acted as a double‐edged sword in tumourigenesis. ALKBH5 promoted cancer stem cell self‐renewal in acute myeloid leukaemia, 40 and caused glioblastoma tumourigenesis by sustaining FOXM1 expression. 15 On the other hand, ALKBH5 inhibited pancreatic cancer progression by activating PER1. 41 However, research on ALKBH5 in GC progression is controversial. Zhang et al. 20 reported that ALKBH5 demethylated the lncRNA NEAT1 and promoted GC invasion and metastasis. ALKBH5 bound to NEAT1 and influenced the expression of EZH2 (a subunit of the polycomb repressive complex). Hu et al. 21 reported that ALKBH5 was a cancer suppressor in GC. ALKBH5 suppressed the invasion of GC by downregulating and removing the m6A modifications of PKMYT1, which enhanced the invasion of GC. These controversial results reveal that the role of ALKBH5 in GC progression must be fully elucidated. Based on three bioinformatics methods, we analysed the entire TCGA database and found significant upregulation of ALKBH5 in 414 GC tumour samples compared to 37 normal samples. Comparing GC tissues with paired noncancerous tissues ($$n = 67$$), we discovered a significant increase in ALKBH5 expression. However, Hu et al. analysed only 290 GC tumour samples and 22 normal samples obtained from the TCGA‐*Stomach adenocarcinoma* database, and the number of patients included in our GC cohort was 49.21 We analysed the ACRG GC cohort ($$n = 300$$) and found a shorter OS, RFS and DFS with higher ALKBH5 expression in GC. 42 *These data* were more convincing because of the larger sample size, high integrity of information and high‐quality clinical follow‐up. By analysing clinicopathological characteristics, there was an association between ALKBH5 and histological differentiation, invasion depth, TNM stage and lymphatic metastasis. We also revealed that high ALKBH5 expression led to poor prognosis in GC patients as shown by Kaplan–Meier survival analysis. We confirmed that ALKBH5 functioned as an oncogenic molecule and accelerated GC proliferation, metastasis and invasion using in vivo and in vitro experiments in ALKBH5 stably silenced/overexpressed BGC‐823 and MGC‐803 cells. To make our study more convincing, we repeated cell proliferation and invasion experiments in AGS and NCI‐N87 cells, and similar results were observed.
Using MeRIP‐seq and RNA‐seq, we found that JAK1 may be the downstream target of ALKBH5 in GC. JAK1 is a member of the JAK family and phosphorylates STAT proteins. After moving from the cytoplasm to the nucleus, phosphorylated STATs activate genes involved in cell survival and proliferation. 22 Activated JAK1/STAT3 plays a crucial role in GC proliferation and metastasis. 24 We showed that JAK1 was overexpressed in GC tissues and associated with poor outcome. ALKBH5 bound to JAK1 and removed the m6A modification of JAK1 mRNA and then enhanced the stability of JAK1 mRNA. m6A reading proteins recognise target genes using m6A modification and play a role in RNA decay. A previous study noted the interaction between JAK1 and YTHDF2.31 The expression of JAK1 increased significantly after suppression of YTHDF2 in our study. Overexpressing ALKBH5 weakened the binding between YTHDF2 and JAK1. These results demonstrated that the degradation of JAK1 mRNA was dependent on YTHDF2 and was reduced by ALKBH5, which upregulated the expression of JAK1.
Noncoding RNAs may function as scaffolds and facilitate the binding between mRNAs and proteins. We overlapped the RIP‐seq results and TCGA database and chose five target lncRNAs according to the count number in RIP‐seq data. After evaluating the regulatory effect of these five lncRNAs on JAK1, we found that LINC00659 directly bound to ALKBH5, with no effect on ALKBH5 expression. Because of LINC00659, ALKBH5 more easily bound to JAK1 mRNA and enhanced its stability. Therefore, we first confirmed that a long noncoding RNA, LINC00659, promoted the interaction of the ALKBH5–JAK1 complex and increased JAK1 mRNA expression in GC.
JAK1 is a key member of the JAK family, which participated in the JAK/STAT pathway. We found that STAT3, but not STAT1 or STAT5, was phosphorylated by JAK1 in GC cells. Upregulated JAK1 phosphorylated STAT3 and activated the JAK1/STAT3 pathway, which accelerated the progression of GC.
RNA methylation at m6A‐containing sites represents a new therapeutic target. For example, meclofenamic acid (MA), FTO‐IN‐5 and FTO‐IN‐4 are highly selective inhibitors of FTO. Huang et al. 43 found that MA specifically competed with FTO binding to m6A‐containing sites and increased m6A levels in mRNA. However, ALKBH5‐targeting drugs need further exploration.
Therefore, our study provided compelling in vitro and in vivo evidence that LINC00659 and YTHDF2 function in the modification of ALKBH5‐mediated JAK1 mRNA to promote GC proliferation and metastasis. In addition to demonstrating a connection between ALKBH5 and long noncoding RNA, we further explored the pathways leading to GC development and occurrence and provided biological mechanisms behind GC development and future therapeutic opportunities.
## CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
## References
1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. **Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *Cancer J Clin* (2018) **68** 394-424
2. Rawla P, Barsouk A. **Epidemiology of gastric cancer: global trends, risk factors and prevention**. *Prz Gastroenterol* (2019) **14** 26-38. PMID: 30944675
3. Shimizu D, Kanda M, Kodera Y. **Review of recent molecular landscape knowledge of gastric cancer**. *Histol Histopathol* (2018) **33** 11-26. PMID: 28447336
4. Digklia A, Wagner AD. **Advanced gastric cancer: current treatment landscape and future perspectives**. *World J Gastroenterol* (2016) **22** 2403-2414. PMID: 26937129
5. Rugge M, Meggio A, Pravadelli C. **Gastritis staging in the endoscopic follow‐up for the secondary prevention of gastric cancer: a 5‐year prospective study of 1755 patients**. *Gut* (2019) **68** 11-17. PMID: 29306868
6. Zhao BS, Roundtree IA, He C. **Post‐transcriptional gene regulation by mRNA modifications**. *Nat Rev Mol Cell Biol* (2017) **18** 31-42. PMID: 27808276
7. Roundtree IA, Evans ME, Pan T, He C. **Dynamic RNA modifications in gene expression regulation**. *Cell* (2017) **169** 1187-1200. PMID: 28622506
8. Han J, Wang JZ, Yang X. **METTL3 promote tumor proliferation of bladder cancer by accelerating pri‐miR221/222 maturation in m6A‐dependent manner**. *Mol Cancer* (2019) **18** 110. PMID: 31228940
9. Liu J, Yue Y, Han D. **A METTL3–METTL14 complex mediates mammalian nuclear RNA N6‐adenosine methylation**. *Nat Chem Biol* (2014) **10** 93-95. PMID: 24316715
10. Ping XL, Sun BF, Wang L. **Mammalian WTAP is a regulatory subunit of the RNA N6‐methyladenosine methyltransferase**. *Cell Res* (2014) **24** 177-189. PMID: 24407421
11. Schwartz S, Mumbach MR, Jovanovic M. **Perturbation of m6A writers reveals two distinct classes of mRNA methylation at internal and 5' sites**. *Cell Rep* (2014) **8** 284-296. PMID: 24981863
12. Jia G, Fu Y, Zhao X. **N6‐methyladenosine in nuclear RNA is a major substrate of the obesity‐associated FTO**. *Nat Chem Biol* (2011) **7** 885-887. PMID: 22002720
13. Zheng G, Dahl JA, Niu Y. **ALKBH5 is a mammalian RNA demethylase that impacts RNA metabolism and mouse fertility**. *Mol Cell* (2013) **49** 18-29. PMID: 23177736
14. Meyer KD, Jaffrey SR. **Rethinking m6A readers, writers, and erasers**. *Annu Rev Cell Dev Biol* (2017) **33** 319-342. PMID: 28759256
15. Zhang S, Zhao BS, Zhou A. **m6A demethylase ALKBH5 maintains tumorigenicity of glioblastoma stem‐like cells by sustaining FOXM1 expression and cell proliferation program**. *Cancer Cell* (2017) **31** 591-606.e6. PMID: 28344040
16. Chen Y, Zhao Y, Chen J. **ALKBH5 suppresses malignancy of hepatocellular carcinoma via m6A‐guided epigenetic inhibition of LYPD1**. *Mol Cancer* (2020) **19** 123. PMID: 32772918
17. Tang B, Yang Y, Kang M. **m6A demethylase ALKBH5 inhibits pancreatic cancer tumorigenesis by decreasing WIF‐1 RNA methylation and mediating Wnt signaling**. *Mol Cancer* (2020) **19** 3. PMID: 31906946
18. Zhang D, Ning J, Okon I. **Suppression of m6A mRNA modification by DNA hypermethylated ALKBH5 aggravates the oncological behavior of KRAS mutation/LKB1 loss lung cancer**. *Cell Death Disease* (2021) **12** 518. PMID: 34016959
19. Tong J, Wang X, Liu Y. **Pooled CRISPR screening identifies m6A as a positive regulator of macrophage activation**. *Sci Adv* (2021) **7**
20. Zhang J, Guo S, Piao H‐Y. **ALKBH5 promotes invasion and metastasis of gastric cancer by decreasing methylation of the lncRNA NEAT1**. *J Physiol Biochem* (2019) **75** 379-389. PMID: 31290116
21. Hu Y, Gong C, Li Z. **Demethylase ALKBH5 suppresses invasion of gastric cancer via PKMYT1 m6A modification**. *Mol Cancer* (2022) **21** 34. PMID: 35114989
22. Quintás‐Cardama A, Verstovsek S. **Molecular pathways: JAK/STAT pathway: mutations, inhibitors, and resistance**. *Clin Cancer Res* (2013) **19** 1933-1940. PMID: 23406773
23. Buchert M, Burns CJ, Ernst M. **Targeting JAK kinase in solid tumors: emerging opportunities and challenges**. *Oncogene* (2016) **35** 939-951. PMID: 25982279
24. Su C, Wang W, Wang C. **IGF‐1‐induced MMP‐11 expression promotes the proliferation and invasion of gastric cancer cells through the JAK1/STAT3 signaling pathway**. *Oncol Lett* (2018) **15** 7000-7006. PMID: 29731870
25. Ma S, Chen C, Ji X. **The interplay between m6A RNA methylation and noncoding RNA in cancer**. *J Hematol Oncol* (2019) **12** 121. PMID: 31757221
26. Pan Y, Fang Y, Xie M. **LINC00675 suppresses cell proliferation and migration via downregulating the H3K4me2 level at the SPRY4 promoter in gastric cancer**. *Mol Ther Nucleic Acids* (2020) **22** 766-778. PMID: 33230474
27. Xu TP, Liu XX, Xia R. **SP1‐induced upregulation of the long noncoding RNA TINCR regulates cell proliferation and apoptosis by affecting KLF2 mRNA stability in gastric cancer**. *Oncogene* (2015) **34** 5648-5661. PMID: 25728677
28. Chen M, Wei L, Law CT. **RNA N6‐methyladenosine methyltransferase‐like 3 promotes liver cancer progression through YTHDF2‐dependent posttranscriptional silencing of SOCS2**. *Hepatology* (2018) **67** 2254-2270. PMID: 29171881
29. Dominissini D, Moshitch‐Moshkovitz S, Schwartz S. **Topology of the human and mouse m6A RNA methylomes revealed by m6A‐seq**. *Nature* (2012) **485** 201-206. PMID: 22575960
30. Wang X, Lu Z, Gomez A. **N6‐methyladenosine‐dependent regulation of messenger RNA stability**. *Nature* (2014) **505** 117-120. PMID: 24284625
31. Linder B, Grozhik AV, Olarerin‐George AO, Meydan C, Mason CE, Jaffrey SR. **Single‐nucleotide‐resolution mapping of m6A and m6Am throughout the transcriptome**. *Nat Methods* (2015) **12** 767-772. PMID: 26121403
32. Hu BB, Wang XY, Gu XY. **N6‐methyladenosine (m6A) RNA modification in gastrointestinal tract cancers: roles, mechanisms, and applications**. *Mol Cancer* (2019) **18** 178. PMID: 31810483
33. Zhao Q, Zhao Y, Hu W. **m6A RNA modification modulates PI3K/Akt/mTOR signal pathway in gastrointestinal cancer**. *Theranostics* (2020) **10** 9528-9543. PMID: 32863943
34. Wang Q, Geng W, Guo H. **Emerging role of RNA methyltransferase METTL3 in gastrointestinal cancer**. *J Hematol Oncol* (2020) **13** 57. PMID: 32429972
35. Wang Q, Chen C, Ding Q. **METTL3‐mediated m6A modification of HDGF mRNA promotes gastric cancer progression and has prognostic significance**. *Gut* (2020) **69** 1193-1205. PMID: 31582403
36. Liu X, Xiao M, Zhang L. **The m6A methyltransferase METTL14 inhibits the proliferation, migration, and invasion of gastric cancer by regulating the PI3K/AKT/mTOR signaling pathway**. *J Clin Laboratory Anal* (2021) **35**
37. Yu H, Zhao K, Zeng H. **N6‐methyladenosine (m6A) methyltransferase WTAP accelerates the Warburg effect of gastric cancer through regulating HK2 stability**. *Biomed Pharmacother* (2021) **133**. PMID: 33378974
38. Chen XY, Liang R, Yi YC. **The m6A reader YTHDF1 facilitates the tumorigenesis and metastasis of gastric cancer via USP14 translation in an m6A‐dependent manner**. *Front Cell Develop Biol* (2021) **9**
39. Shen X, Zhao K, Xu L. **YTHDF2 inhibits gastric cancer cell growth by regulating FOXC2 signaling pathway**. *Front Genet* (2020) **11**. PMID: 33505426
40. Shen C, Sheng Y, Zhu AC. **RNA demethylase ALKBH5 selectively promotes tumorigenesis and cancer stem cell self‐renewal in acute myeloid leukemia**. *Cell Stem Cell* (2020) **27** 64-80.e9. PMID: 32402250
41. Guo X, Li K, Jiang W. **RNA demethylase ALKBH5 prevents pancreatic cancer progression by posttranscriptional activation of PER1 in an m6A‐YTHDF2‐dependent manner**. *Mol Cancer* (2020) **19** 91. PMID: 32429928
42. Cristescu R, Lee J, Nebozhyn M. **Molecular analysis of gastric cancer identifies subtypes associated with distinct clinical outcomes**. *Nat Med* (2015) **21** 449-456. PMID: 25894828
43. Huang Y, Yan J, Li Q. **Meclofenamic acid selectively inhibits FTO demethylation of m6A over ALKBH5**. *Nucleic Acids Res* (2015) **43** 373-384. PMID: 25452335
|
---
title: Role of adrenomedullin2/ intermedin in pregnancy induced vascular and metabolic
adaptation in mice
authors:
- Chandra Yallampalli
- Ancizar Betancourt
- Akansha Mishra
- Kathleen A. Pennington
- Simone Hernandez Ruano
- Moises Tacam
- Madhu Chauhan
journal: Frontiers in Physiology
year: 2023
pmcid: PMC9982084
doi: 10.3389/fphys.2023.1116042
license: CC BY 4.0
---
# Role of adrenomedullin2/ intermedin in pregnancy induced vascular and metabolic adaptation in mice
## Abstract
Introduction: Adrenomedullin2 (AM2) shares its receptor with *Calcitonin* gene related peptide and adrenomedullin with overlapping but distinct biological functions. Goal of this study was to assess the specific role of Adrenomedullin2 (AM2) in pregnancy induced vascular and metabolic adaptation using AM2 knockout mice (AM2 −/−).
Method: The AM2 −/− mice were successfully generated using Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)/Nuclease Cas nine system. Phenotype of pregnant AM2 −/− mice was assessed with respect to its fertility, blood pressure regulation, vascular health and metabolic adaptations and compared to the wild type littermates (AM2 +/+).
Results: *Current data* shows that AM2 −/− females are fertile with no significant difference in number of pups/litter compared to the AM2 +/+. However, ablation of AM2 decreases the gestational length and the total number of pups born dead or that die after birth is greater in AM2 −/− mice compared to AM2 +/+ mice ($p \leq 0.05$). Further AM2 −/− mice exhibit elevated blood pressure and elevated vascular sensitivity for the contractile responses to angiotensin two and higher serum sFLT-1 trigylcerides levels compared to AM2 +/+($p \leq 0.05$). In addition, AM2 −/− mice develop glucose intolerance with elevated serum levels of Insulin during pregnancy compared to the AM2 +/+mice.
Discussion: *Current data* suggests a physiological role for AM2 in pregnancy induced vascular and metabolic adaptations in mice.
## Introduction
Adrenomedullin2 (AM2)/Intermedin (IMD) is a hypotensive peptide discovered in 2004 (Roh et al., 2004; Takei, et al., 2004). AM2 belongs to a unique group of calcitonin (CT)/*Calcitonin* gene related peptide (CGRP) family of peptide hormones and shares sequence homology with its family peptides AM ($28\%$) and CGRP (<$20\%$). In addition, AM2, CGRP and AM share a common 7TM G-protein coupled receptor (GPCR) known as calcitonin-gene related receptor like receptor (CLR). However, the ligand specificity of CLR is dictated by its hetero-dimerization with a group of receptor activity modifying proteins (RAMPs) in such a manner that, RAMP1 and RAMP3 function for CGRP, RAMP2 or RAMP3 can function for AM and all three RAMPs are capable of forming AM2 receptor with CLR (Roh et al., 2004; Takei, et al., 2004). However, majority of the AM2 effects are shown to involve RAMP2 and RAMP3 (Roh et al., 2004; Hay et al., 2018; Garelja and Hay, 2022). These peptides are important for homeostasis in diverse tissues and have similar but distinct physiological effects (Nelson et al., 2000; Roh et al., 2004). Several reports show involvement of these three peptides in reproductive functions (Caron and Smithies, 2001; Chauhan et al., 2006; Chauhan et al., 2009; Chauhan et al., 2011a; Yallampalli et al., 2013; Yallampalli et al., 2014). Genetically modified mouse model of CGRP, AM and their receptor components have been developed with the reports showing embryonically lethal effect of AM, CLR, and RAMP2 ablation but not in CGRP, RAMP1, and RAMP3 null mice (Shindo et al., 2000; Caron and Smithies, 2001; Shindo et al., 2001; Kurihara et al., 2003; Dackor et al., 2006; Ichikawa-Shindo et al., 2008). Thus, although these peptides have some structural similarities and share a common receptor system, they exhibit distinct but overlapping biological functions (Nelson et al., 2000; Roh et al., 2004). Being a novel member of this group of peptides, the physiological role of AM2 in pregnancy is not clearly understood. AM2 is expressed in several tissues including pituitary, hypothalamus, ovary, placenta, and uterus (Roh et al., 2004; Takei, et al., 2004). We have shown that circulatory and placental expression of AM2 is higher during first trimester suggesting a role in implantation and placental development in human and rat pregnancy (Chauhan et al., 2011b; Havemann et al., 2012). This is supported by our report showing that AM2 promotes the invasive capacity of 1st trimester trophoblast cells (Chauhan et al., 2009; Chauhan et al., 2011a) and that the sensitivity of maternal vasculature for AM2 effects is enhanced in rodents as well as in human pregnancy, suggesting a role in pregnancy induced vascular adaptation (Chauhan et al., 2007; Chauhan et al. 2021; Chauhan et al. 2022).
In addition, potential role for AM2 in metabolic homeostasis is reported showing that AM2 treatment restores high-fat diet–induced early insulin resistance in adipose tissue in mice, AM2-tg mice display improvements in high-fat diet–induced early adipose insulin resistance (Lv et al., 2016), AM2 levels negatively correlate with HOMA of insulin resistance in obese human and that, plasma level of AM2 closely associate with different cardiometabolic diseases (Zhang et al., 2018). We reported earlier that infusion of AM2 receptor antagonist AM217-47 during rat pregnancy causes impaired placental function and feto-placental growth restriction (Chauhan, et al., 2006). Serum AM2 level are higher during pregnancy and its expression is downregulated in circulation and placenta in spontaneous abortion and preeclampsia (PE) (Havemann et al., 2012; Chauhan et al., 2016). Interestingly, AM2 levels are lower in second trimester before onset of clinical symptoms of PE in human (Chauhan et al., 2016). More importantly, our recent study shows that AM2 treatment causes relaxation in segments of omental artery isolated from women undergoing preeclamptic pregnancy (Chauhan et al., 2007; Chauhan et al., 2007; Chauhan et al. 2021) suggesting a potential role in the pathophysiology of hypertensive pregnancies such as PE.
Therefore, the goal of this study was to identify the physiological importance of endogenous AM2 peptide during pregnancy using AM2 knockout (AM2−/−) mice generated by CRISPER/CAS9 technology. Pregnancy outcome of AM2−/− mice was assessed with respect to regulation of blood pressure, length of gestation, feto-placental health, and metabolic adaptations and compared with their wild type littermates (WT). To identify the role of AM2 in vascular adaptation, maternal sensitivity to ATII and serum levels of sFLT-1 were assessed and its involvement in pregnancy induced metabolic changes were assessed by analyzing glucose tolerance and serum levels of Insulin and triglycerides.
## Animal model and procedures
Animal care and use committee (IACUC) of Baylor College of Medicine (BCM) approved this study. All studies were performed under standard 12:12-h light dark cycles in accordance with ARRIVE guidelines. All mice used were C57BL/6J obtained from Jackson Laboratory (Bar Harbor, ME, United States).
## Generation of CRISPR/CAS9 mediated AM2 knockout mice
Genetically Engineered Mouse Core (GEM) and Mouse Embryonic Stem Cell (BCM mES) Core at Baylor college of Medicine (BCM) generated AM2 knockout mice. Briefly, to generate a null allele of AM2, two single guide RNAs (sgRNAs) were selected by BCM mES Core, flanking the genomic region containing the open reading frame of AM2. DNA templates were produced using overlapping oligonucleotides in a high-fidelity PCR reaction. The two SgRNA targeting sequence used are 5′AGAAGGGCTCCCCAACTGGT and 3′-CATACCTTGGCCCGATTCTC. The SgRNA/Cas9/ss oligo mixture was microinjected into the cytoplasm of pronuclear stage zygotes from C57BL/6NJ female mice by BCM GEM Core. Mice were genotyped by standard PCR using a three-primer system, a single forward primer shared between two different reverse primers. Two primers approximately 100–200 bases outside the two-sgRNA sites were designed to amplify a smaller deletion amplicon compared to the wild-type amplicon. A separate reaction using the second reverse primer, placed within the predicted deleted interval, was designed to amplify a product in the endogenous allele. Therefore, the 3-primer assay for genotyping included a shared forward primer (P1: 5-CCAAACTGGTTTTCCGCTGG-3″) and reverse primers unique to the wild type (P2: 5’—CTGAGGAGTTCGGTCCAACC-3″) and CRISPR deletion allele primer (P3: 5’—TCGGTGCAGATTCTACAGCC-3″). Genotyped AM2 knockout mice were backcrossed eight times with C57BL/6NJ mice from Jackson laboratory, strain#005304 before using for the experiments.
## Off-target analysis
BCM MES cell core screened heterozygous AM2 null males for the top five potential off-target sites for each sgRNA, identified using the Wellcome Trust Sanger Institute Genome Editing website. Sequencing primers were designed to amplify larger PCR products, which were Sanger sequenced.
## Fertility assay in female AM2 knockout mice
A group of 8 weeks old AM2 −/− female mice (KO, $$n = 8$$) and AM2 +/+ female mice (wild type littermates, WT; $$n = 8$$) were assigned to the fertility study. These mice were crossed (monogamous pairs) with wild type proven breeder males (12 weeks old C57BL/6NJ breeders) from Jackson laboratories. The day of observed copulatory plug was identified as pregnancy Day (GD) 0.5. The mating pairs were left in cages for 5 months. The gestational age was recorded at 1st delivery. The number of litters/dam and number of total dead pups were recorded over a period of 5 months.
## Assessment of blood pressure, vascular reactivity and feto-placental weights
Another cohort of 8 weeks old mice [AM 2−/− and AM 2+/+ ($$n = 8$$/group)] was utilized for assessing blood pressure, vascular reactivity, feto-placental weights and serum analysis in non-pregnant and pregnant state. These mice were anesthetized with a mixture of ketamine (Ketalar; Parke-Davis, Morris Plains, New Jersey) and xylazine (Gemini; Rugby, Rockville Center, New York), and telemetric BP transducers (PA-C10 model; Data Sciences, St Paul, Minnesota) were implanted. During the first 3 days, mice were allowed to fully recover from the surgery followed by mating. The day of copulatory plug was considered pregnancy Day 0.5.
Blood pressure: Blood pressure (BP) data was recorded for 72 h before mating (non-pregnant) and recording was resumed on day 14.5 of pregnancy for 72 h. The recordings were performed for 30 s at 10-min intervals using the Dataquest ART data acquisition system (DSI, St Paul, Minnesota).
Feto-placental weights and tissue collection: Mice were euthanized following BP measurements, feto-placental weights were recorded, mesenteric arteries were collected for vascular studies, and blood for serum extraction and stored at −80°C until analyzed for sFLT-1 levels.
Vascular reactivity studies: Two-millimeter segments of second order mesenteric arteries were cleaned off fat and mounted on a wire myograph (DMT 610M) with 25-m tungsten wires. The preparations were bathed in Krebs solution that was maintained at 37°C (pH −7.4). A mixture of $95\%$ O2 and $5\%$ CO2 was bubbled continuously through the solution. The force was recorded continuously by isometric force transducers and analyzed with PowerLab data acquisition system and Lab Chart 7 software (AD Instruments, Castle Hill, Australia). The arteries were normalized according to the manufacturer’s manual and set to the lumen diameter of d1 = 0.9 x d100, where active force development is maximal. After stabilization of the tone, the vessels were contracted twice with 80 mmol/L of potassium chloride (KCl) for 10 min to enhance reproducibility of responses. The second response to KCl was used as a reference contraction in the final calculations. To evaluate endothelial function, response to a single dose of acetylcholine (10−6 M) in vessels that were pre-contracted with phenylephrine (10−6 or 3 X10−6 M) was determined. Only experiments with intact endothelium were used in the final analysis. After 1 h of equilibration, relaxant responses to CGRP1-37, AM1-52 and AM21-47 (10−10 to 10−5 M) were obtained after pre-contraction of the vessels with norepinephrine (10−7 to 10−6 M) to produce matching contractions in the study groups. In addition, contractile responses to the angiotensin-2 (10–10 to- 10–6 M) were determined. After each agent was tested, the vessels were washed with Krebs solution and left to recover for 30 min until they returned to their basal passive tension. The Wire Myograph System-DMT- 610M has an automated normalization function, which is assessed from the “Normalization” menu and allows the vessel to be stretched to a normalized internal circumference by a standardized procedure according to the manufacturer’s protocol after equilibration for 30 min at 37°C. An exponential curve is fitted to the internal circumference pressure data. The procedure defines the lumen diameter (d100) that the artery would have had in vivo when relaxed and under a transmural pressure of 100 mmHg.
## Intraperitoneal glucose tolerance test
A separate cohort of 8–10-week-old female AM2 −/− mice or their wild type littermates ($$n = 12$$–15) were fasted for 5h prior to breeding with wild type littermate males and after onset of pregnancy on gestational day (GD) 13.5. Fasting blood was collected for serum isolation followed by intraperitoneal glucose tolerance tests (GTT) as previously described (Pennington et al., 2017). Two ReliOn Prime Blood Glucose Monitoring System meters (Walmart, Bentonville, AR) were used to measure glucose levels. Following IPGTT, dams were euthanized.
## Insulin analysis by enzyme linked immunoassay
Insulin levels were assessed in mice fasted for 5 h for IPGTT using Rat/Mouse Insulin ELISA (EMD Millipore, United States) according to the manufacturer’s instructions. Intraassay CV was <$10\%$ and inter-assay CV was <$12\%$ ($$n = 12$$–15).
## Triglyceride’s analysis in serum
Serum triglyceride levels were measured in non-pregnant and pregnant mice fasted for 5 h using the Serum Triglyceride Determination Kit (Sigma-Aldrich, United States). Triglycerides were extracted from serum as previously described (Pennington et al., 2017) and then quantified as per Kit instructions ($$n = 6$$).
## Drugs
Human Alpha-CGRP1-37 and human AM1-52 were purchased from Phoenix Pharmaceutical Inc., United States; Human AM21-47 was from Alpha Diagnostic, United States, and ATII and U46619 was purchased from Sigma Aldrich, United States.
## Statistical analysis
All data are presented as mean ± SEM. Relaxation responses to the peptides are expressed as percent relaxation of the initial U46619-induced contraction. The second response to KCL was used as a reference to calculate the percent of contraction achieved by ATII. Concentration response curves of drugs are fitted to a log-logistic sigmoid relation, and Emax (maximal relaxation effect) are calculated by using GraphPad Prism. Repeated measures ANOVA (treatment and time as factors) with a Bonferroni post hoc was used for comparisons of dose response curves between groups. For other experiments, comparisons between groups were done by Student’s two tailed t-test in Prism. Statistical significance was defined as $p \leq 0.05.$
## CRISPR/CAS9 mediated knockdown of AM2 in C57BL/6NJ mice
AM2 knockdown in C57BL/6J mouse was generated by CRISPR/CAS9 technology with the help of Genetically Engineered Mouse core lab at Baylor college of Medicine. Figures 1A shows the two-sgRNA sequences at 5′ and 3’ end of AM2 DNA sequence with the respective PAM sequence spanning the complete full length AM2 gene and Figures 1B demonstrates the PCR assay used for genotyping. The wild type and null allele were genotyped using a 3-primer PCR assay consisting of a shared forward (P1) primer and reverse primer unique to the wild-type (P2) and CRISPR deletion (P3) alleles. Figure 1C shows the PCR product of 458 bp obtained for wild type (WT) allele and 285 bp PCR product obtained for the AM2 knockout allele (AM2 −/−). Figure 1D shows decreased expression of AM2 mRNA in placenta of AM2 −/− mice compared to their wild type littermates ($p \leq 0.05$).
**FIGURE 1:** *Construction of CRISPR/CAS9 mediated AM2 knockdown: (A) Selected sgRNA sites flanking the coding region of AM2 (orange on the 5′ end and green on the 3′end, PAM sites in lowercase), (B) Schematic representation of PCR generated wild type allele (458bp) and Null allele (285bp) and (C) Representative gel images of 485 bp amplified PCR product for wild type allele (WT) and 285 bp amplified PCR product for null allele (KO), and (D) expression of AM2 mRNA in mice placenta. Asterisks (*) indicates significant difference between the groups (p < 0.05). Data represents the mean ± SEM, analyzed by unpaired 2-tailed t-test.*
## Effect of AM2 ablation on gestational length, pup mortality and pregnancy rate
As mentioned in the methods, fertility assay was assessed over a period of five consecutive pregnancies and gestational length was recorded for 1st pregnancy. Figure 2A shows that ablation of AM2 gene in female mice have shorter gestational length. Figures 2B,C shows that there is no difference in the number of pups/litter or number of pregnancies per dam respectively, over a period of 5 months in AM −/− mice compared to the AM2 +/+. Further, Figures 2D,E show the effect of AM2 ablation on fetal mortality where the mean number of pups that were born dead per dam (Figure 2D) and the mean number of pups that died after birth (Figure 2E) over a period of five consecutive pregnancies is higher in AM2 −/− mice compared to those born to the wild type littermates ($p \leq 0.05$). Therefore, data in Figure 2 suggests that ablation of AM2 results in increased fetal mortality compared to the wild type littermates (Figure 2F) n = seven to eight dams/group.
**FIGURE 2:** *Effect of AM2 ablation on gestational length, pregnancy rate and pup mortality: Figure shows fertility data of AM2
−/− (KO) and their wild type littermates (WT) over a period of five consecutive months (n = seven to eight dams/group). (A) Day of delivery in 1st pregnancy showing shorter gestational length of KO compared to the WT (p < 0.05); (B) Mean number of pups born to KO compared to the number of pups born to WT; (C) Number of litters born per dam in KO compared to the number of litters per dam in WT over five consecutive pregnancies. (D) Mean number of pups born dead in KO compared to the number of dead pups born to the WT (p < 0.05); (E) Number of pups that died after birth in KO compared to those in WT (p < 0.05); and (F) Overlapping bar graph showing correlation of decreased gestational age (GA) with increased fetal death (mean number of pups born dead/dam over five consecutive pregnancy Asterisks (*) indicate significant differences between the groups. Data represents the mean ± SEM, analyzed by unpaired 2-tailed t-test.*
## Effect of AM2 ablation on feto-placental weight
Figure 3 shows the effect of AM2 ablation on the average weight of fetus (Figure 3A) and placenta (Figure 3B) per dam on GD 17.5 of gestation. As shown, AM2 ablation does not affect the weights of fetus or placenta of AM2−/− dams when compared to the wild type littermates ($$n = 8$$ dam/group; $p \leq 0.05$).
**FIGURE 3:** *Effect of AM2 ablation on weight of fetus and placenta: Pregnant AM2
−/− (KO) mice and their wild type littermates (WT) were euthanized on gestational day 15 and the weights of placenta (A) and fetus (B) were recorded.*
## Effect of AM2 ablation on blood pressure, serum triglycerides and vascular sensitivity for αCGRP1-37, AM1-52, and AM21-47.
Figure 4A shows the systolic BP in non-pregnant and pregnant mice and Figure 4B is the diastolic BP in non-pregnant and pregnant mice. As shown, ablation of AM2 does not affect the BP of non-pregnant AM2 −/− [systolic BP (111 ± 1.9) and diastolic BP (79.60 ± 1.72); $p \leq 0.05$] compared to the in non-pregnant AM2 +/+ [systolic (114.0 ± 1.0) and diastolic (79.60 ± 1.20); $p \leq 0.05$]. Interestingly, during pregnancy blood pressure is elevated in pregnant AM2 −/− mice [systolic BP (116.6 ± 1.8) and diastolic BP (93.05 ± 4.42)] compared to the wild type littermates [systolic BP (108.6 ± 3.8; $p \leq 0.05$) and diastolic BP (83.73 ± 8.73; $p \leq 0.05$)]. Further, Figure 4C show the effect of AM2 knockout on serum triglycerides in non-pregnant state (AM2 −/− [0.3607 ± 0.04428] vs. AM2 +/+ [0.424 ± 0.0504]; $p \leq 0.05$, $$n = 8$$). And in pregnant mice (AM2 −/− [1.133 ± 0.093] vs. AM2 +/+ [0.705 ± 0.073]; $p \leq 0.05$, $$n = 8$$). As shown triglycerides in non-pregnant AM2 −/− are similar to those in WT mice, However, ablation of AM2 results in increases in serum triglycerides during pregnancy in AM2−/− compared to the AM2 +/+($p \leq 0.05$). Since AM2 is a hypotensive peptide and shares a common receptor system with CGRP and AM, any effect of AM2 ablation on the receptor system was assessed by testing the functional responses of mesenteric artery for relaxation effects of CGRP, AM and AM2. Figure 4D shows the data from wire myograph studies in mesenteric artery segments of pregnant AM2 −/− compared to that in pregnant AM2 +/+ mice. Dose response curves for CGRP1-37, AM1-52 and AM2 1–47 treatments shows that the sensitivity of mesenteric artery for the relaxation effects of CGRP1-37, AM1-52 and AM2 1–47 in AM −/− mice is similar to that in the AM2 +/+mice. This is indicative of an intact endogenous CGRP and AM function and presence of a functional receptor system for all three peptides in AM2 −/− mice.
**FIGURE 4:** *Effect of AM2 ablation on blood pressure, serum triglycerides and vascular sensitivity for CGRP1-37, AM1-52 and AM21-47: (A) Mean systolic blood pressure recorded for 72 h in non-pregnant [p > 0.05] and in pregnant mice [p < 0.05] from gestational day 14.5 in AM2
−/− (KO) mice during pregnancy compared to their wild type littermates (WT) (B) Mean diastolic blood pressure recorded for 72 h in non-pregnant [p > 0.05] and in pregnant mice [p < 0.05] from gestational day 14 in AM2
−/− (KO) mice compared to their wild type littermates (WT); (C) Fasting serum levels of triglycerides in non-pregnant (p > 0.05) and pregnant AM2 −/− (KO) mice (p < 0.05) fasted on gestational day 13.5 compared to their wild type littermates (WT) and; (D) Sensitivity of mesenteric artery pre-contracted with norepinephrine for CGRP1-37, AM1-52 and AM21-47 mediated relaxation in AM2
−/− (KO) mice compared to the wild type littermates (WT). All data are presented as mean ± SEM, analyzed by unpaired 2-tailed t-test and vascular data was analyzed by repeated measures ANOVA (treatment and time as factors) with a Bonferroni post hoc for comparisons of dose response curves between te groups. Asterisks (*) indicate significant difference between the groups (p < 0.05).*
## Effect of AM2 ablation on serum levels of soluble fms-like tyrosine kinase (sFLT-1) and vascular sensitivity for angiotensin2
Figure 5A shows that the serum levels of sFLT-1 in AM2 −/− are significantly elevated in on GD 17.5 compared to the wild type littermates ($p \leq 0.05$; $$n = 4$$–15). In addition, dose response curve of wire myograph study in Figure 5B shows that the sensitivity of mesenteric artery for the contractile effects of ATII is significantly elevated in AM2 −/− mice compared to the wild type littermates [Emax 104.2 ± 23.03 in AM2 −/− vs. 62.20 ± 16.74 in AM+/+; $p \leq 0.05$], which is indicative of vascular dysfunction in AM2−/− mice.
**FIGURE 5:** *Effect of AM2 ablation on serum levels of soluble fms-like tyrosine kinase (sFLT-1) and vascular sensitivity for angiotensin2: (A) Serum levels of sFLT-1 in AM2
−/− (KO) and AM
+/+ (WT) mice on GD 17.5. As shown sFLT-1 levels in AM2
−/− mice are significantly elevated compared to the wild type littermates (p < 0.05; n = 4–15); (B) Dose response curves showing increased contractile effects of angiotensin2 (ATII) [percentage of initial KCL contraction] in AM2
−/− (KO) mice compared to the wild type littermates (WT). All data are presented as mean ± SEM, analyzed by unpaired 2-tailed t-test (Figure 5A) or repeated measures ANOVA (treatment and time as factors) with a Bonferroni post hoc for comparisons of dose response curves between the groups (Figure 5B). Asterisks (*) indicate significant difference between the groups (p < 0.05).*
## Effect of AM2 ablation on body weight, glucose tolerance and serum levels of Insulin.
Body weight recorded before the onset of pregnancy presented in Figure 6A, shows that AM2 ablation does not impact the body weight ($p \leq 0.05$). In addition, intraperitoneal glucose tolerance test (GTT) performed in non-pregnant mice prior to the onset of their pregnancy presented in Figure 6B shows that ablation of AM2 does not affect glucose metabolism in non-pregnant AM2 −/− mice compared to their wild type littermates ($p \leq 0.05$). However, with onset of pregnancy, AM2 −/− mice develop impaired glucose tolerance with increased area under the curve as shown in Figure 6C, $p \leq 0.05$, compared to the wild type littermates. Further, Figure 6D shows that impaired glucose tolerance is associated with increases in fasting serum insulin levels in AM2−/− mice compared to the wild type littermates ($$n = 12$$–15, $p \leq 0.05$).
**FIGURE 6:** *Effect of AM2 ablation on pre-pregnancy body weight, Glucose tolerance and serum levels of Insulin: (A) Pre-pregnancy body weight of AM2
−/− (KO) mice compared to their wild type littermates (WT) (p > 0.05); (B) Glucose tolerance test performed in fasted non-pregnant mice before onset of pregnancy in AM2
−/− (KO) mice compared to their wild type littermates (WT); (C) glucose tolerance test performed in pregnant AM2
−/− (KO) mice fasted on gestational day 13.5 compared to their wild type littermates (WT) with area under the curve shown as an inset (p < 0.05); and (D) Serum levels of insulin in pregnant AM2−/− (KO) mice fasted on gestational day 13.5 compared to their wild type littermates (WT). Asterisk (*) indicate significant difference between the groups with p < 0.05. (n = 12–15). All data are presented as mean ± SEM, analyzed by unpaired 2-tailed t-test.*
## Discussion
Current study was designed to assess the physiological role of endogenous AM2 in mice pregnancy using AM2 −/− mice with CRISPR/CAS9 mediated systemic ablation of AM2. Data shows that AM2 −/− mice are fertile but exhibit increased fetal mortality as assessed over a period of five consecutive pregnancies. In addition, AM2 −/− mice exhibit pregnancy induced elevations in blood pressure, serum sFLT-1 levels and AT11 sensitivity along with impaired glucose tolerance associated with elevated levels of insulin and triglycerides in serum. Therefore, this study establishes an important physiological role for AM2 in facilitating vascular and metabolic adaptations in healthy pregnancy and provides a potential mouse model to study pathological pregnancies with vascular and metabolic abnormalities.
Adrenomedullin two facilitates trophoblast invasion in early placental development and circulatory levels of AM2 increase with pregnancy in rodents and humans (Chauhan et al., 2011; Havemann et al., 2012). However, physiological role and importance of AM2 in reproduction is not clearly understood due to its structural similarities with its family peptides, CGRP and AM. In addition, AM2 uses a complex receptor system that is shared by CGRP and AM (Roh et al., 2004; Hay et al., 2018; Garelja and Hay. 2022), which limits the use of AM2 antagonist due to its potential cross-reactivity with other family peptides. Therefore, goal of the current study was to assess pregnancy outcomes of mice in absence of AM2 function. Adrenomedulli2 knockout (AM2 −/− mice were successfully generated by CRISPR/CAS9 technology (Figure 1). Data in Figure 2A shows that, in absence of AM2 expression/function, AM2 −/− mice are fertile with shorter gestational length. There is no effect of AM2 ablation on the fertility rate (Figure 2B) or litter size (Figure 2C). However, pregnancy in AM2 −/− mice suffers with increased pup mortality. The number of pups born dead (Figure 2D), as well as number of pups that die after birth (Figure 2E) is higher in AM2 −/− mice compared to the wild type littermates. AM2 is shown to be an estrogen dependent pituitary paracrine factor regulating prolactin release (Lin et al., 2005). As Nursing stimulates prolactin release from the pituitary which promotes continued milk production, it is likely that ablation of AM2 may have affected prolactin release and thus impacted lactation and offspring health resulting in increased pup mortality. Interestingly, unlike our earlier studies showing fetal growth restriction (FGR) in rats infused with AM2 receptor antagonist (AM217-47) during pregnancy (Chauhan et al., 2006)ablation of AM2 in the current study did not affect the feto-placental growth (Figure 3). Instead, although not significant, there was a tendency of higher birth weight in AM2 −/− compared to the wild type littermates with very few cases of obstructive delivery necessitating euthanization due to large fetus (observed in only 2 dams (1pup/dam) out of eight different cohorts of pregnant mice and excluded from the study). The discrepancy observed in feto-placental weights in our two studies is likely due to the potential effect of receptor antagonist on the actions of endogenous CGRP and AM function, which are reported to facilitate feto-placental growth during pregnancy (Roh et al., 2004; Yallampalli et al., 2013; Yallampalli et al., 2014). Interestingly, despite the absence of an effect of AM2 ablation on feto-placental weight in AM2 −/− mice (Figure 3), pregnancy in these mice resulted in increased fetal mortality compared to the wild type littermates (Figures 2D–F). Although, the cause of increased fetal death in AM2 −/− is not clear from this study, it is likely to arise from defective placental development due to impaired trophoblast invasion leading to impaired feto-placental function. This is supported by our earlier reports showing that AM21-47 treatment promotes 1st trimester trophoblast invasion in human pregnancy and its expression coincides with the invasive phase of placental development in rat and human pregnancy (Chauhan et al., 2011b; Havemann et al., 2012).
The primary adaptive mechanism in pregnancy is a marked fall in systemic vascular resistance and a transient reduction in BP along with, pregnancy induced decrease in ATII sensitivity in maternal vasculature (Leal et al., 2022). Adrenomedullin2 is a hypotensive peptide known for its vasodilatory functions (Chauhan et al., 2007; Kandilci et al., 2008; Ross et al., 2010; Hirose et al., 2011). In hypertensive pregnancy with elevated ATII sensitivity, such as PE, AM2 serum levels are downregulated (Havemann et al., 2012; Chauhan et al., 2016). However, it is not known if AM2 has a role in regulating blood pressure and ATII sensitivity in pregnancy. Recent report shows that mice over expressing RAMP1, a co-receptor for AM2 as well as for CGRP, ameliorates ATII mediated hypertension in rodents (Sabharwal et al., 2010). Therefore, due to structural similarities between AM2, CGRP and AM, and overlapping biological function, functional importance of AM2 is not clearly understood in pregnancy. Current study shows that AM2 ablation does not affect the integrity of CLR/RAMPs receptor system as demonstrated by testing the functional response of mesenteric artery for CGRP, AM2 and AM mediated relaxation using wire myograph (as shown in Figure 4D). Sensitivity for the relaxation response of all three peptides is preserved in AM2−/− mice. Ability of CGRP, AM and AM2 to cause relaxation in AM2 −/− mice during pregnancy indicate that the CLR/RAMPs receptor system is functional in AM2 −/− mice with active endogenous CGRP and AM function. However, although not significant, there was a trend of increased sensitivity for AM2 and AM in AM2 −/− mice compared to the AM+/+ mice, with AM2 > AM. This may be the compensatory effect of AM ablation in AM−/− mice. In addition, AM2 ablation did not affect the serum levels of triglycerides in AM2 −/− compared to the AM2 +/+ mice in non-pregnant state. Interestingly, despite the intact vascular actions of endogenous CGRP and AM in AM2 −/− mice, AM2 −/− mice exhibit elevated BP and ATII sensitivity along with elevated serum triglycerides during pregnancy (Figure 4C) compared to their wild type littermates. Therefore, the changes observed in the vascular health of pregnant AM2 −/− mice are specific to AM2 ablation and suggest a role for AM2 in maintaining pregnancy induced vascular adaptation. AM2 in paraventricular nucleus (PVN) is reported to attenuate ATII induced generation of reactive oxygen species (ROS) in obese rats with hypertension (Kang et al., 2019). As autonomic nervous system is known to influence peripheral resistance in PE (Schobel et al., 1996), it is likely inhibition/ablation of AM2 function may have an impact on the hypoxia induced increase in the sympathetic activity in PE pregnancy resulting in elevated BP (Quitterer et al., 2004; Baumwell and Karumanchi, 2007).
Further, series of recent studies show that AM2 has protective effect against metabolic syndrome, improving the risk factors for cardiovascular diseases (Zhang H et al., 2016). In human participants, circulating AM2 concentration has been reported to be inversely associated with body weight, BMI, and insulin resistance (Zhang, H. et al., 2016; Zhang S. Y et al., 2016). Further, a recent study showed that AM2 signaling is suppressed in adipose tissue in obesity suggesting lower receptor expression and ligand availability contributing to insulin resistance and other aspects of associated metabolic disorders (Kim et al., 2020). However, ablation of AM2 in the current study did not affect the serum levels of triglycerides (Figure 4C), body weight (Figure 6A) and glucose metabolism (Figure 6B) in non-pregnant AM2 −/− mice compared with their wild type littermates. Interestingly, when challenged with pregnancy, AM2 −/− mice became developed elevated serum triglycerides (Figure 4C) with impaired glucose intolerance (Figure 6C) accompanied with elevated levels of fasting serum insulin (Figure 6D) compared to the wild type littermates. Stress is a normal reaction to a major physiological change such as vascular and metabolic adaptations that occur in normal pregnancy. This suggest that absence of AM2 function renders increased risk for metabolic disorder under stressful conditions such as establishment of pregnancy. Preeclampsia (PE) and gestational diabetes mellitus (GDM) are common pregnancy complications, occurring only during pregnancy with similar risk factors and patho-physiological changes (Pankiewicz et al., 2022; Yang and Wu, 2022). Evidence from previous studies suggests that the incidence of PE is significantly increased in women with GDM. However, it is not clear whether GDM is independently related to the occurrence of PE or vice versa. Current data in pregnant AM2 −/− mouse model and published reports on AM2 studies in human pregnancy suggest that PE associated decreases in AM2 levels in human pregnancy may contribute in part to the elevated BP and increased sensitivity of maternal vasculature for vasoconstriction associated with metabolic disorders in pregnancy. However, our earlier reports show that AM2 facilitates trophoblast invasion in 1st trimester and levels of AM2 in amniotic fluid are lower in second trimester of PE pregnancy before onset of the clinical symptoms (Chauhan et al., 2006; Chauhan et al., 2016). Therefore, based on reports and current study, AM2 −/− mice appears to represent a mouse model of PE with comorbid gestational diabetes.
## Conclusion
As most of the physiological or pathophysiological effects of AM2 are mediated by CGRP and/or AM receptors, lack of specific antagonists for these receptors sub types impedes the progress in distinguishing AM2 specific action from the overlapping biological functions of its family peptides. This study is the first to show physiological importance of endogenous AM2 in pregnancy induced vascular and metabolic adaptations. This study shows that systemic ablation of AM2 results in increased fetal mortality, elevated BP and serum sFLT-1 levels along with increased vascular sensitivity for ATII. In addition, the vascular defects in AM2 −/− mice are associated with impaired glucose metabolism during pregnancy suggesting that inhibition of AM2 function results in vascular and metabolic defects in mice pregnancy. Further, AM2 −/− mice can serve as a useful animal model for investigating molecular mechanisms involved in hypertensive pregnancies such as PE, that are associated with metabolic disorders and its consequences on the health of mother and offspring.
## Limitations of the study
Placenta plays a critical role in developmental progression of pregnancy while current study was focused on whole body knockout of AM2. Therefore, future study will assess if the impaired vascular and metabolic function observed in this study are due to the involvement of AM2 in placental functions using mice with placenta specific AM2 knockout.
## Data availability statement
The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.
## Ethics statement
The animal study was reviewed and approved by Ethics Committee of Baylor College of Medicine.
## Author contributions
MC and CY designed the study. MC and AB characterized the pregnancy phenotype including telemetry studies, AB performed the vascular reactivity studies AM performed the PCR. KP, and SH performed the Insulin and triglyceride assays MC and CY analyzed data and wrote the manuscript. All authors reviewed the manuscript prior to submission.
## 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. Baumwell S., Karumanchi S. A.. **Pre-eclampsia: Clinical manifestations and molecular mechanisms**. *Nephron Clin.Pract.* (2007) **106** c72-c81. DOI: 10.1159/000101801
2. Caron K. M., Smithies O.. **Extreme hydrops fetalis and cardiovascular abnormalities in mice lacking a functional Adrenomedullin gene**. *Proc. Natl. Acad. Sci. USA* (2001) **98** 615-619. DOI: 10.1073/pnas.021548898
3. Chauhan M., Balakrishnan M., Vidaeff A., Yallampalli U., Lugo F., Fox K.. **Adrenomedullin2 (ADM2)/Intermedin (IMD): A potential role in the pathophysiology of preeclampsia**. *J. Clin. Endocrinol. Metab.* (2016) **101** 4478-4488. DOI: 10.1210/jc.2016-1333
4. Chauhan M., Balakrishnan M., Yallampalli U., Endsley J., Hankins G. D., Theiler R.. **Adrenomedullin 2/intermedin regulates HLA-G in human trophoblasts**. *Biol. Reproduction* (2011a) **85** 1232-1239. DOI: 10.1095/biolreprod.110.086835
5. Chauhan M., Betancourt A., Balakrishnan M., Mishra A., Espinosa J., Shamshirsaz A. A.. **Calcitonin gene related peptide, adrenomedullin, and adrenomedullin 2 function in uterine artery during human pregnancy**. *Endocrinology* (2022) **163** bqab204. DOI: 10.1210/endocr/bqab204
6. Chauhan M., Betancourt A., Balakrishnan M., Mishra A., Fox K., Belfort M.. **Soluble fms-like tyrosine kinase-1 and angiotensin2 target calcitonin gene-related peptide family peptides in maternal vascular smooth muscle cells in pregnancy†**. *Biol. Reproduction* (2021) **104** 1071-1083. DOI: 10.1093/biolre/ioab026
7. Chauhan M., Elkins R., Balakrishnan M., Yallampalli C.. **Potential role of intermedin/adrenomedullin 2 in early embryonic development in rats**. *Regul. Pept.* (2011b) **170** 65-71. DOI: 10.1016/j.regpep.2011.05.011
8. Chauhan M., Ross G. R., Yallampalli U., Yallampalli C.. **Adrenomedullin-2, a novel calcitonin/calcitonin-gene-related peptide family peptide, relaxes rat mesenteric artery: Influence of pregnancy**. *Endocrinology* (2007) **148** 1727-1735. DOI: 10.1210/en.2006-1105
9. Chauhan M., Yallampalli U., Dong Y. L., Hankins G. D., Yallampalli C.. **Expression of adrenomedullin 2 (ADM2)/intermedin (IMD) in human placenta: Role in trophoblast invasion and migration**. *Biol. Reproduction* (2009) **81** 777-783. DOI: 10.1095/biolreprod.108.074419
10. Chauhan M., Yallampalli U., Reed L., Yallampalli C.. **Adrenomedullin 2 antagonist infusion to rats during midgestation causes fetoplacental growth restriction through apoptosis**. *Biol. Reprod.* (2006) **75** 940-947. DOI: 10.1095/biolreprod.106.053322
11. Dackor R. T., Fritz-Six K., Dunworth W. P., Gibbons C. L., Smithies O., Caron K. M.. **Hydrops fetalis, cardiovascular defects, and embryonic lethality in mice lacking the calcitonin receptor-like receptor gene**. *Mol. Cell Biol.* (2006) **26** 2511-2518. DOI: 10.1128/MCB.26.7.2511-2518.2006
12. Garelja M. L., Hay D. L.. **A narrative review of the calcitonin peptide family and associated receptors as migraine targets: Calcitonin gene-related peptide and beyond**. *Headache* (2022) **62** 1093-1104. DOI: 10.1111/head.14388
13. Havemann D., Balakrishnan M., Borahay M., Theiler M., Jennings K., EndsleyPhelps J.. **Intermedin/adrenomedullin 2 is associated with implantation and placentation via trophoblast invasion in human pregnancy**. *J. Clin. Endocrinol. Metab.* (2012) **98** 695-703. DOI: 10.1210/jc.2012-2172
14. Hay D. L., Garelja M. L., Poyner D. R., Walker C. S.. **Update on the pharmacology of calcitonin/CGRP family of peptides: IUPHAR review 25**. *Br. J. Pharmacol.* (2018) **175** 3-17. DOI: 10.1111/bph.14075
15. Hirose T., Totsune K., Nakashige Y., Metoki H., Kikuya M., Ohkubo T.. **Influence of adrenomedullin 2/intermedin gene polymorphism on blood pressure, renal function and silent cerebrovascular lesions in Japanese: The ohasama study**. *Hypertens. Res.* (2011) **34** 1327-1332. DOI: 10.1038/hr.2011.131
16. Ichikawa-Shindo Y., Sakurai T., Kamiyoshi A., Kawate H., Iinuma N., Yoshizawa T.. **The GPCR modulator protein RAMP2 is essential for angiogenesis and vascular integrity**. *J. Clin. Invest.* (2008) **118** 29-39. DOI: 10.1172/JCI33022
17. Kandilci H. B., Gumusel B., Lippton H.. **Intermedin/adrenomedullin-2 (IMD/AM2) relaxes rat main pulmonary arterial rings via cGMP-dependent pathway: Role of nitric oxide and large conductance calcium-activated potassium channels (BK(Ca))**. *Peptides* (2008) **29** 1321-1328. DOI: 10.1016/j.peptides.2008.04.008
18. Kang Y., Ding L., Dai H., Wang F., Zhou H., Gao Q.. **Intermedin in paraventricular nucleus attenuates ang II-induced sympathoexcitation through the inhibition of NADPH oxidase-dependent ROS generation in obese rats with hypertension**. *Int. J. Mol. Sci.* (2019) **20** 4217. DOI: 10.3390/ijms20174217
19. Kim J., Lee S. K., Kim D., Choe H., Jang Y. J., Park H. S.. **Altered expression of adrenomedullin 2 and its receptor in the adipose tissue of obese patients**. *J. Clin. Endocrinol. Metab.* (2020) **105** dgz066. DOI: 10.1210/clinem/dgz066
20. Kurihara H., Shindo T., Oh-hashi Y., Kurihar Y., Kuwaki T.. **Targeted disruption of adrenomedullin and alphaCGRP genes reveals their distinct biological roles**. *Hypertens. Res.* (2003) **26** S105-S108. DOI: 10.1291/hypres.26.s105
21. Leal C. R. V., Costa L. B., Ferreira G. C., Ferreira A. M., Reis F. M., Simoes E Silva A. C.. **Renin-angiotensin system in normal pregnancy and in preeclampsia: A comprehensive review**. *Pregnancy. Hypertens.* (2022) **28** 15-20. DOI: 10.1016/j.preghy.2022.01.011
22. Lin C. C., Roh J., Park J. I., Klein C., Cushman N., Haberberger R. V.. **Intermedin functions as a pituitary paracrine factor regulating prolactin release**. *Mol. Endocrinol.* (2005) **19** 2824-2838. DOI: 10.1210/me.2004-0191
23. Lv Y., Zhang S. Y., Liang X., Zhang H., Xu Z., Liu B.. **Adrenomedullin 2 enhances beiging in white adipose tissue directly in an adipocyte-autonomous manner and indirectly through activation of M2 macrophages**. *J. Biol. Chem.* (2016) **291** 23390-23402. DOI: 10.1074/jbc.M116.735563
24. Nelson S. H., Steinsland O. S., Wang Y., Yallampalli C., Simmons D., Wimalawansa S. J.. **Increased nitric oxide synthase activity and expression in the human uterine artery during pregnancy**. *Circ. Res.* (2000) **87** 406-411. DOI: 10.1161/01.res.87.5.406
25. Pankiewicz K., Szczerba E., Fijalkowska A., Sierdzinski J., Issat T., Maciejewski T. M.. **The impact of coexisting gestational diabetes mellitus on the course of preeclampsia**. *J. Clin. Med.* (2022) **11** 6390. DOI: 10.3390/jcm11216390
26. Pennington K. A., van der Walt N., Pollock K. E., Talton O. O., Schulz L. C.. **Effects of acute exposure to a high-fat, high-sucrose diet on gestational glucose tolerance and subsequent maternal health in mice**. *Biol. Reproduction* (2017) **96** 435-445. DOI: 10.1095/biolreprod.116.144543
27. Quitterer U., Lother H., AbdAlla S.. **AT1 receptor heterodimers and angiotensin II responsiveness in preeclampsia**. *Semin. Nephrol.* (2004) **24** 115-119. DOI: 10.1016/j.semnephrol.2003.11.007
28. Roh J., Chang C. L., Bhalla A., Klein C., Hsu S. Y. T.. **Intermedin is a calcitonin/calcitonin gene-related peptide family peptide acting through the calcitonin receptor-like receptor/receptor activity-modifying protein receptor complexes**. *J. Biol. Chem.* (2004) **279** 7264-7274. DOI: 10.1074/jbc.M305332200
29. Ross G. R., Yallampalli U., Gangula P. R., Reed L., Sathishkumar K., Gao H.. **Adrenomedullin relaxes rat uterine artery: Mechanisms and influence of pregnancy and estradiol**. *Endocrinology* (2010) **151** 4485-4493. DOI: 10.1210/en.2010-0096
30. Sabharwal R., Zhang Z., Lu Y., Abboud F. M., Russo A. F., Chapleau M. W.. **Receptor activity-modifying protein 1 increases baroreflex sensitivity and attenuates Angiotensin-induced hypertension**. *Hypertension* (2010) **55** 627-635. DOI: 10.1161/HYPERTENSIONAHA.109.148171
31. Schobel H. P., Fischer T., Heuszer K., Geiger H., Schmieder R. E.. **Preeclampsia -- a state of sympathetic overactivity**. *N. Engl. J. Med.* (1996) **335** 1480-1485. DOI: 10.1056/NEJM199611143352002
32. Shindo T., Hirata Y., Nishimatsu H., Moriyama N., Kurihara Y., Maemura K.. *Pathophysiological assessment of adrenomedullin by mice gene engineering approach* (2000)
33. Shindo T., Kurihara Y., Nishimatsu H., Moriyama N., Kakoki M., Wang Y.. **Vascular abnormalities and elevated blood pressure in mice lacking adrenomedullin gene**. *Circulation* (2001) **104** 1964-1971. DOI: 10.1161/hc4101.097111
34. Takei Y., Inoue K., Ogoshi M., Kawahara T., Bannai H., Miyano S.. **Identification of novel adrenomedullin in mammals: A potent cardiovascular and renal regulator**. *FEBS Lett.* (2004) **556** 53-58. DOI: 10.1016/s0014-5793(03)01368-1
35. Yallampalli C., Chauhan M., Endsley J., Sathishkumar K.. **Calcitonin gene related family peptides: Importance in normal placental and fetal development**. *Adv. Exp. Med. Biol.* (2014) **814** 229-240. DOI: 10.1007/978-1-4939-1031-1_20
36. Yallampalli C., Chauhan M., Sathishkumar K.. **Calcitonin gene-related family peptides in vascular adaptations, uteroplacental circulation, and fetal growth**. *Curr. Vasc. Pharmacol.* (2013) **11** 641-654. DOI: 10.2174/1570161111311050007
37. Yang Y., Wu N.. **Gestational diabetes mellitus and preeclampsia: Correlation and influencing factors**. *Front. Cardiovasc.Med.* (2022) **9** 831297. DOI: 10.3389/fcvm.2022.831297
38. Zhang H., Zhang S. Y., Jiang C., Li Y., Xu G., Xu M. J.. **Intermedin/adrenomedullin 2 polypeptide promotes adipose tissue browning and reduces high-fat diet-induced obesity and insulin resistance in mice**. *Int. J. Obes. (Lond)* (2016) **40** 852-860. DOI: 10.1038/ijo.2016.2
39. Zhang S. Y., Lv Y., Zhang H., Gao S., Wang T., Feng J.. **Adrenomedullin 2 improves early obesity-induced adipose insulin resistance by inhibiting the class II MHC in adipocytes**. *Diabetes* (2016) **65** 2342-2355. DOI: 10.2337/db15-1626
40. Zhang S. Y., Xu M. J., Wang X.. **Adrenomedullin 2/intermedin: A putative drug candidate for treatment of cardiometabolic diseases**. *Br. J. Pharmacol.* (2018) **175** 1230-1240. DOI: 10.1111/bph.13814
|
---
title: PHLDA1 modulates microglial response and NLRP3 inflammasome signaling following
experimental subarachnoid hemorrhage
authors:
- Jinqing Lai
- Genwang Chen
- Zhe Wu
- Shaoyang Yu
- Rongfu Huang
- Yile Zeng
- Weibin Lin
- Chunmei Fan
- Xiangrong Chen
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9982097
doi: 10.3389/fimmu.2023.1105973
license: CC BY 4.0
---
# PHLDA1 modulates microglial response and NLRP3 inflammasome signaling following experimental subarachnoid hemorrhage
## Abstract
Balancing microglia M1/M2 polarization is an effective therapeutic strategy for neuroinflammation after subarachnoid hemorrhage (SAH). Pleckstrin homology-like domain family A member 1 (PHLDA1) has been demonstrated to play a crucial role in immune response. However, the function roles of PHLDA1 in neuroinflammation and microglial polarization after SAH remain unclear. In this study, SAH mouse models were assigned to treat with scramble or PHLDA1 small interfering RNAs (siRNAs). We observed that PHLDA1 was significantly increased and mainly distributed in microglia after SAH. Concomitant with PHLDA1 activation, nod-like receptor pyrin domain-containing protein 3 (NLRP3) inflammasome expression in microglia was also evidently enhanced after SAH. In addition, PHLDA1 siRNA treatment significantly reduced microglia-mediated neuroinflammation by inhibiting M1 microglia and promoting M2 microglia polarization. Meanwhile, PHLDA1 deficiency reduced neuronal apoptosis and improved neurological outcomes after SAH. Further investigation revealed that PHLDA1 blockade suppressed the NLRP3 inflammasome signaling after SAH. In contrast, NLRP3 inflammasome activator nigericin abated the beneficial effects of PHLDA1 deficiency against SAH by promoting microglial polarization to M1 phenotype. In all, we proposed that PHLDA1 blockade might ameliorate SAH-induced brain injury by balancing microglia M1/M2 polarization via suppression of NLRP3 inflammasome signaling. Targeting PHLDA1 might be a feasible strategy for treating SAH.
## Introduction
Subarachnoid hemorrhage (SAH), a devastating acute cerebrovascular event, has a poor prognosis with a high rate of neurocognitive impairment in patients. Currently, accumulating evidence has proposed that early brain injury (EBI) may be a determinant factor for SAH-induced long-term neurocognitive sequelae (1–3). Unfortunately, no effective pharmaceutical strategy has been identified to interfere with the development of EBI. Hence, identifying new drug targets to improve SAH outcomes is urgently needed.
Neuroinflammation has been verified as a crucial contributor to EBI progression after SAH (4–6). It is known that microglia, key innate immune cells of the brain, are rapidly activated to various acute brain injuries (6–8). After activation, microglia can exhibit different phenotypes (M1 and M2 phenotypes) and exert distinct functions. M1 microglia could induce proinflammatory mediators and increase reactive oxygen species (ROS). By contrast, M2 microglia exhibit anti-inflammatory effects by secreting anti-inflammatory meditators. Interestingly, microglia can switch their phenotype under different microenvironments. In a variety of neurological disorders, suppression of microglia M1 polarization and promotion of M2 microglia could effectively reduce acute brain injuries and improve neurological outcomes (9–11). Thus, modulating microglia M1/M2 polarization might be a feasible method to mitigate EBI.
Recently, a substantial number of studies have revealed that pleckstrin homology-like domain family A member 1 (PHLDA1) plays a crucial role in oxidative stress and immunological regulation (12–15). In a model of cerebral ischemia/reperfusion injury, PHLDA1 blockade ameliorated the acute brain injury by switching microglia M1/M2 polarization via inhibiting nod-like receptor pyrin domain-containing protein 3 (NLRP3) inflammasome signaling [12]. Another study reported that PHLDA1 deficiency mitigated motor deficits and microglia-mediated neuroinflammation in Parkinson’s disease models [13]. However, the function roles of PHLDA1 in microglia-mediated immune response after SAH remain unclear. NLRP3 inflammasome has been demonstrated to implicate in neuroinflammation after SAH by modulating microglial polarization [4, 16]. Notably, blockade of NLRP3 inflammasome activation could exert beneficial effects in different brain injuries (17–19). Herein, we hypothesized that PHLDA1 inhibition might mitigate neuroinflammation and the subsequent neurobehavior deficits after SAH through the NLRP3 inflammasome signaling pathway.
## Establishment of SAH model
Adult male C57BL/6 mice (8-10 wk old, weighing 20–25 g) were obtained from the Animal Core Facility of Fujian Medical University. All experimental procedures were complied with the rules for animal research by Fujian Medical University. Briefly, mice were anesthetized with isoflurane. After the common, external and internal carotid arteries were exposure, a marked 6-0 filament was employed to puncture the origin of the left middle cerebral artery through the internal carotid artery [20]. Animals in sham group received similar procedures without the artery puncture. The SAH severity grading score was recorded according to previous studies [21]. Mice with a SAH grading score of less than 8 were excluded.
## Study design
In the first experiment, mice were assigned to sham group ($$n = 6$$) and post-SAH (6 h, 12 h, 24 h, 48 h, 72 h) ($$n = 6$$ per group). Western blot and immunofluorescence staining were performed in the experiment. In the second experiment, mice were assigned to sham group, vehicle-treated SAH group, Scramble small interfering RNA (siRNA)-treated SAH group, and PHLDA1 siRNA-treated SAH group ($$n = 12$$ per group). Animals were sacrificed at 24 h or 72 h after SAH. Post-treatment assessments included neurobehavior tests, western blot, immunofluorescence staining, TUNEL staining, enzyme-linked immunosorbent assay (ELISA), and biochemical estimation. In the third experiment, mice were assigned to sham group, vehicle-treated SAH group, PHLDA1 siRNA-treated SAH, and PHLDA1 siRNA plus nigericin -treated SAH group ($$n = 12$$ per group). Post-treatment assessments included neurobehavior tests, western blot, immunofluorescence staining, TUNEL staining, ELISA, and biochemical estimation.
## Drug administration
For PHLDA1 knockdown, a volume of 3μl PHLDA1 siRNA (Santa Cruz Technology) or scramble siRNA was dissolved in transfection solution and then injected into the lateral ventricles at 48 h before the construction of SAH model. Nigericin (MedChemExpress, 2 μg), a potent NLRP3 activator, was prepared in 2 μl ethanol and physiologic saline. Nigericin or vehicle was intracerebroventricularly administered at 2 h before SAH operation. The dose of nigericin and administration route were based on previous studies [22].
## Neurobehavioral tests
The modified Garcia scale test was used to evaluate neurological deficits as previously reported [23]. Six measurements were included in this score system. The higher score suggested the better neurobehavioral outcomes. For motor function, the beam-walking score test was performed according to previous reports [24]. The animals’ walking distance within 1 min were recorded. Neurobehavior tests were conducted in a blinded manner.
## ELISA
The supernatant of brain samples was collected. The levels of interleukin (IL)-1β, IL-6, IL-18, and IL-10 were detected by using commercially available kits (Multi Sciences). The detailed methods were conducted according to the manufacturer’s instructions.
## Western blotting
The brain tissue and protein samples were prepared according to previous studies [25]. Briefly, the protein samples were loaded onto SDS-PAGE gels and transferred to PVDF membranes. The membranes were blocked with $5\%$ non-fat milk. After that, they were incubated with primary antibodies: PHLDA1 (1:1000, Abcam), NLRP3 (1:200, Santa Cruz Biotechnology), ASC (1:200, Santa Cruz Biotechnology), caspase-1 (1:200, Santa Cruz Biotechnology), cleaved caspase-1 (1:200, Santa Cruz Biotechnology), and β-actin (1:3000, Bioworld Technology) in a 4°C freezer. Then, membranes were incubated with corresponding secondary antibodies. ImageJ software was employed to measure relative intensity.
## Immunofluorescence staining
The detailed methods were performed according to previous studies [26]. In brief, the frozen tissue sections were treated with Triton X-100 ($0.3\%$) and then blocked with $5\%$ goat serum. After that, sections were incubated with primary antibodies: PHLDA1 (Abcam), CD$\frac{16}{32}$ (BD Biosciences), CD206 (Invitrogen), NeuN (EMD Millipore), IL-1β (Santa Cruz Biotechnology), and Iba-1 (Santa Cruz Biotechnology) in a 4°C freezer. After that, they were incubated with corresponding secondary antibodies followed by using DAPI staining. The slices were then observed under a fluorescence microscope.
## TUNEL staining
TUNEL staining was performed by using a commercially available kit (Beyotime Biotechnology). Brain sections were incubated with primary antibody against NeuN in a 4°C freezer. After that, the sections were incubated with TUNEL reaction mixture. The slides were then washed and counterstained with DAPI. The slices were observed under a fluorescence microscope.
## Statistical analysis
Data are expressed as mean ± SD. Statistical analysis was conducted with Graph- Pad Prism 8 software. Statistical evaluation was performed using one-way ANOVA or two-way ANOVA with Tukey’s post hoc test. The significant P-value was < 0.05.
## Time course and cellular expression of PHLDA1 and NLRP3 after SAH
Mounting evidence has indicated that PHLDA1 and NLRP3 might interact with each other. PHLDA1 activation could induce NLRP3 inflammasome signaling. In this experiment, western blot (Figure 1A) was performed to investigate the protein expression of PHLDA1 and NLRP3. As shown in Figures 1B, C, the expression of PHLDA1 and NLRP3 markedly increased in the early period after SAH, and peaked at 24 h post-SAH ($P \leq 0.05$). In addition, double immunofluorescence staining indicated that the enhanced PHLDA1 and NLRP3 were mainly distributed in microglia after SAH ($P \leq 0.05$) (Figures 1D–G).
**Figure 1:** *Expression levels of PHLDA1 and NLRP3 were increased after SAH. (A) Representative western blots for PHLDA1 and NLRP3 expressions in the early period after SAH. Western blot analysis of PHLDA1 (B) and NLRP3 (C) expressions after SAH (n = 6 per group). (D, E) Representative immunofluorescence images of PHLDA1 and NLRP3 co-localized with Iba1 in temporal cortex after SAH. Quantification of PHLDA1 (F) and NLRP3 (G) immunoactivities in microglia (n = 6 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## PHLDA1 deficiency inhibited NLRP3 inflammasome signaling activation after SAH
Previous study has demonstrated that PHLDA1 activation could induce NLRP3 inflammasome signaling. We applied PHLDA1 siRNA to inhibit PHLDA1 expression and explore whether PHLDA1 deficiency could reduce NLRP3 inflammasome activation. As shown, western blot results (Figure 2A) showed that PHLDA1 siRNA significantly reduced PHLDA1 expression after SAH ($P \leq 0.05$). Moreover, the activated NLRP3 inflammasome signaling pathway was markedly suppressed by PHLDA1 siRNA ($P \leq 0.05$) (Figures 2A–F). Consistently, double immunofluorescence staining confirmed that PHLDA1 siRNA significantly decreased PHLDA1 and NLRP3 expression in microglia in the brain cortex after SAH ($P \leq 0.05$) (Figures 2G–J).
**Figure 2:** *PHLDA1 deficiency suppressed NLRP3 inflammasome signaling after SAH. (A) Representative western blots for PHLDA1, NLRP3, ASC, Caspase1, and Cleaved caspasse1 expressions after SAH. Western blot analysis of PHLDA1 (B), NLRP3 (C), ASC (D), caspase1 (E), and cleaved caspasse1 (F) expressions after SAH (n = 6 per group). Representative immunofluorescence images of PHLDA1 (G) and NLRP3 (H) co-localized with Iba1 in temporal cortex. Quantification of PHLDA1 (I) and NLRP3 (J) immunoactivities in microglia (n = 6 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## PHLDA1 deficiency reduced inflammatory response
The anti-inflammatory effects of PHLDA1 blockade have been verified in other diseases models. We further investigated the influence of PHLDA1 deficiency on inflammatory response after SAH. By using ELISA kits, we found that SAH insults induced a significant increase in proinflammatory cytokines release, including IL-1β, IL-6, and IL-18 ($P \leq 0.05$) (Figures 3A–C). All these cytokines were decreased by PHLDA1 deficiency ($P \leq 0.05$). In addition, PHLDA1 deficiency significantly induced an increase in IL-10 expression after SAH ($P \leq 0.05$) (Figure 3D). Simultaneously, IL-1β immunofluorescence staining verified that PHLDA1 deficiency significantly decreased the enhanced levels of IL-1β in the brain cortex after SAH ($P \leq 0.05$) (Figures 3E, F).
**Figure 3:** *PHLDA1 deficiency mitigated inflammatory insults after SAH. ELISA analysis of IL-1β (A), IL-6 (B), IL-18 (C), and IL-10 (D) expressions (n = 6 per group). (E) Representative immunofluorescence images of IL-1β staining in temporal cortex. (F) Quantification of IL-1β immunoactivities (n = 6 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## PHLDA1 deficiency promoted M2 microglia polarization and prevented M1 microglia polarization after SAH
Microglial polarization plays a key role in inflammatory response after SAH. Studies have proved that suppression of microglia M1 polarization and promotion of M2 microglia could effectively reduce acute brain injuries and improve neurological outcomes. Interestingly, PHLDA1 has been reported to modulate microglia M1/M2 polarization in other diseases. To determine whether PHLDA1 deficiency affects microglial polarization after SAH, double immunostaining was performed to examine the levels of M1 microglia and M2 microglia. It showed that the number of Iba1+/CD$\frac{16}{32}$+ cells was significantly increased after SAH, which could be decreased by PHLDA1 silencing ($P \leq 0.05$) (Figures 4A, C). In addition, we examined the expression of M2 microglia and revealed that the levels of Iba1+/CD206+ cells were markedly increased after PHLDA1 siRNA treatment ($P \leq 0.05$) (Figures 4B, D). These data suggested that PHLDA1 silencing could suppress M1 microglia and promote M2 microglia polarization.
**Figure 4:** *PHLDA1 deficiency inhibited M1 microglia polarization and promoted M2 microglia transformation. Representative immunofluorescence images of CD16/32 (A) and CD206 (B) co-localized with Iba1 in temporal cortex. Quantification of Iba1+/CD16/32+ cells (C) and Iba1+/CD206+
(D) cells in temporal cortex (n = 6 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## PHLDA1 deficiency reduced neuronal death and improved neurological outcomes after SAH
Next, we explored whether PHLDA1 deficiency could exert cerebroprotective effects after SAH. TUNEL staining revealed that SAH insults significantly induced neuronal apoptosis ($P \leq 0.05$) (Figures 5A, B). Concomitant with the exacerbated neuronal death, the neurological outcomes of SAH was further aggravated ($P \leq 0.05$) (Figures 5C, D). In contrast, PHLDA1 siRNA treatment evidently reduced SAH-induced neuronal apoptosis ($P \leq 0.05$). Simultaneously, PHLDA1 deficiency showed better neurological outcomes after SAH insults ($P \leq 0.05$) (Figures 5A–D). These suggested that PHLDA1 deficiency could protect against SAH-induced EBI by its anti-inflammatory effects.
**Figure 5:** *PHLDA1 deficiency decreased neuronal apoptosis and improved neurological function after SAH. (A) Representative immunofluorescence images of TUNEL staining in temporal cortex. (B) Quantification of TUNEL+/NeuN+ cells in temporal cortex (n = 6 per group). PHLDA1 deficiency mitigated neurological deficits (C) and improved motor function (D) after SAH (n = 8 or 9 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## Nigericin administration reversed the inhibitory effects of PHLDA1 deficiency on NLRP3 inflammasome
NLRP3 inflammasome plays a key role in microglial activation after SAH. Moreover, NLRP3 inflammasome activation could induce microglia M1 polarization and inhibit NLRP3 inflammasome could promote M2 microglia polarization. As mentioned above, PHLDA1 deficiency could markedly inhibit NLRP3 inflammasome activation after SAH. We further explored whether NLRP3 inflammasome activation by nigericin could abate the cerebroprotective effects of PHLDA1 deficiency. As expected, western blot results showed that nigericin administration eliminated the inhibitory effects of PHLDA1 deficiency on NLRP3 inflammasome activation ($P \leq 0.05$) (Figures 6A–E). Consistently, the immunofluorescence staining results verified that NLRP3 inflammasome staining was further induced after nigericin treatment ($P \leq 0.05$) (Figures 6F, G).
**Figure 6:** *Nigericin administration counteracted the effects of PHLDA1 deficiency on NLRP3 inflammasome. (A) Representative western blots for NLRP3, ASC, caspase1, and cleaved caspasse1 expressions after SAH. Western blot analysis of NLRP3 (B), ASC (C), Caspase1 (D), and Cleaved caspasse1 (E) expressions after SAH (n = 6 per group). (F) Representative immunofluorescence images of NLRP3 co-localized with Iba1 in temporal cortex. Quantification of NLRP3 (G) immunoactivity in microglia (n = 6 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## Nigericin abated the effects of PHLDA1 deficiency on microglia M1/M2 polarization and inflammatory insults
Next, we examined the effects of nigericin administration on microglial polarization after PHLDA1 siRNA treatment. As expected, the immunofluorescence staining results showed that NLRP3 inflammasome activation by nigericin abated the effects of PHLDA1 silencing on microglia M1/M2 polarization, as evidenced by the increased number of Iba1+/CD$\frac{16}{32}$+ cells and decreased number of Iba1+/CD206+ cells ($P \leq 0.05$) (Figures 7A–D). Moreover, the ELISA data indicated that nigericin further exacerbated proinflammatory cytokines release and decreased anti-inflammatory cytokines after SAH ($P \leq 0.05$) (Figures 7E–H). These suggested that NLRP3 inflammasome activation contributed to the modulation effects of PHLDA1 on microglial polarization after SAH.
**Figure 7:** *Nigericin administration abated the effects of PHLDA1 deficiency on microglia M1/M2 polarization and inflammatory insults. Representative immunofluorescence images of CD16/32 (A) and CD206 (B) co-localized with Iba1 in temporal cortex. Quantification of Iba1+/CD16/32+ cells (C) and Iba1+/CD206+ cells (D) in temporal cortex (n = 6 per group). ELISA analysis of IL-1β (E), IL-6 (F), IL-18 (G), and IL-10 (H) expressions (n = 6 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## Nigericin abrogated the beneficial effects of PHLDA1 deficiency on neuronal survival and neurological function
We suspected that NLRP3 inflammasome activation by nigericin might reverse the beneficial effects of PHLDA1 deficiency on neuronal survival and neurological function. Consistent with the aggravated inflammatory insults, nigericin administration further significantly increased neuronal apoptosis and exacerbated neurological deficits and motor dysfunction ($P \leq 0.05$) (Figures 8A–D). Based on the findings above, PHLDA1 blockade could attenuate EBI after SAH by regulating microglia M1/M2 polarization via NLRP3 inflammasome signaling.
**Figure 8:** *Nigericin administration abrogated the beneficial effects of PHLDA1 deficiency on neuronal survival and neurological function. (A) Representative immunofluorescence images of TUNEL staining in temporal cortex. (B) Quantification of TUNEL+/NeuN+ cells in temporal cortex (n = 6 per group). Nigericin administration aggravated neurological deficits (C) and motor dysfunction (D) after SAH (n = 9 per group). *P < 0.05. Scale bar=50 μm. Data are expressed as mean ± S.D.*
## Discussion
In this study, we elucidated the biological function of PHLDA1 in microglia-mediated neuroinflammation after SAH. We demonstrated that PHLDA1 was significantly increased and was peaked at 24 h after SAH. The immunofluorescence studies revealed that the enhanced PHLDA1 after SAH was mainly distributed in microglia. PHLDA1 siRNA treatment significantly reduced neuroinflammatory response and the subsequent brain insults after SAH. Notably, PHLDA1 knockdown reduced the number of M1 microglia and promoted M2 microglial polarization. Moreover, PHLDA1 deficiency inhibited NLRP3 inflammasome signaling after SAH. In contrast, NLRP3 inflammasome activator nigericin abrogated the protective effects of PHLDA1 deficiency against SAH and further aggravated neurobehavior deficits. Taken together, these data suggested that targeting PHLDA1 might be a potential therapeutic strategy for treating SAH.
PHLDA1, a member of the PHLDA family, is a multifunctional protein. It has been demonstrated that PHLDA1 participates in modulation of cell proliferation, energy homeostasis, differentiation and apoptosis (27–29). Recently, a wealth of evidence indicated that PHLDA1 also plays an important role in immune response. For example, Hossain et al. demonstrated that PHLDA1 knockdown modulated macrophages and endothelial cells phenotypic changes to reduce atherogenesis-induced oxidative and ER stress [14]. In Parkinson’s disease study, Han et al. reported that PHLDA1 was a potent modulator of neuroinflammation, and knockdown of PHLDA1 could markedly inhibited M1 microglia activation [13]. A more direct study indicated that PHLDA1 blockade inhibited neuroinflammation after ischemic stroke by balancing microglial M1/M2 polarization [12]. However, the potential roles of PHLDA1 in EBI after SAH remain unclear.
We first investigated the time course of PHLDA1 expression after SAH. Consistent with previous reports [12], our data indicated that PHLDA1 was significantly increased, with a peak at 24 h after SAH. At the cellular level, the enhanced PHLDA1 after SAH was mainly distributed in microglia. Moreover, we noted that NLRP3 inflammasome was significantly increased in microglia after SAH. The NLRP3 inflammasome, a cytoplasmic multiprotein complex, has been demonstrated to exert a pivotal role in neuroinflammation and intracranial aneurysm rupture (30–32). It showed that NLRP3 inflammasome participated in microglial polarization in a variety of brain injuries (33–35). Chen et al. demonstrated that NLRP3 deficiency could promote M2 microglia polarization in experimental models of intracerebral hemorrhage [36]. A recent study in cerebral ischemia/reperfusion injury suggested that inhibition of TXNIP/NLRP3 promoted the transition of microglia from M1 to M2 phenotype [37]. In SAH area, inhibition NLRP3 inflammasome has also been demonstrated to promote microglial polarization to M2 phenotype [4]. In our experiments, we observed that there were similar expression trends and cellular distribution of PHLDA1 and NLRP3 after SAH. These suggested that PHLDA1 and NLRP3 might interact with each other. Intriguingly, a recent study by Zhao et al. revealed that PHLDA1 deficiency suppressed middle cerebral artery occlusion/reperfusion-induced NLRP3 inflammasome activation and the subsequent mRNA level and expression of NLRP3 inflammasome-associated proteins [12]. Therefore, we speculated that PHLDA1 might modulate microglial polarization by NLRP3 inflammasome activation.
In our experiments, we employed PHLDA1 siRNA to suppress PHLDA1 activation. Our data revealed that PHLDA1 siRNA treatment significantly inhibited the protein expression of PHLDA1 after SAH. Moreover, the evident neuroinflammation was markedly suppressed by PHLDA1 blockade. Microglial polarization plays a critical role in immune response after SAH. Mounting evidence has shown that inducing microglia toward M2 phenotype or suppression of microglia M1 polarization could promote neuronal survival, reduce inflammatory insults, and improve neurological outcomes after SAH (38–40). We further examined the influence of PHLDA1 blockade on microglial polarization after SAH. Consistent with previous studies, we found that PHLDA1 deficiency decreased M1 phenotype microglia and induced microglia M2 polarization after SAH. Concomitant with the reduced neuroinflammation, PHLDA1 blockade improved neuronal survival and neurological function after SAH. However, the underlying molecular mechanisms of PHLDA1 blockade on microglial polarization remain unknown. The evidence above implied that NLRP3 inflammasome might participate in PHLDA1-mediated microglial polarization. To further clarify the molecular mechanisms of PHLDA1 blockade on microglial polarization, nigericin was applied to activate NLRP3 inflammasome signaling. As expected, nigericin treatment significantly induced NLRP3 inflammasome activation and abrogated the beneficial effects of PHLDA1 blockade on EBI after SAH. Meanwhile, nigericin further increased M1 microglia polarization and inhibited M2 phenotype microglia. These data further supported that PHLDA1 blockade could inhibit NLRP3 inflammasome activation to balance microglial polarization from M1 to M2 after SAH.
There are several shortcomings in our study. Firstly, microglia-specific PHLDA1-knockout mouse should be utilized in the future to validate the biological function of PHLDA1 in microglial activation after SAH. Secondly, some authors reported that PHLDA1 exhibited anti-inflammatory effects through inhibition of toll-Like receptor 4 (TLR4) signaling [15]. The reasons for this disagreement are somewhat obscure. One possible explanation might be that acute brain injuries have different pathophysiology. Thirdly, in addition to modulate NLRP3 inflammasome signaling, PHLDA1 might interfere with other molecular targets, such as TLR4, nuclear factor-erythroid 2-related factor 2, and TRAF6 [13, 15, 41]. Moreover, the long-term effects of PHLDA1 inhibition in the delayed phase of SAH remains unclear. Therefore, additional studies are still warranted to clarify these questions.
## Conclusions
This study is the first to document that PHLDA1 blockade attenuated EBI after SAH by regulating microglia M1/M2 polarization via NLRP3 inflammasome signaling. These findings suggested that PHLDA1 might be a novel therapeutic strategy for treating SAH.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding authors.
## Ethics statement
The animal study was reviewed and approved by the Animal Core Facility of Fujian Medical University.
## Author contributions
JL and XC conceived and designed the experiments; JL, GC, ZW, SY, RH, and YZ performed research; WL conducted the data analysis; JL, CF, and XC revised the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Luo F, Wu L, Zhang Z, Zhu Z, Liu Z, Guo B. **The dual-functional memantine nitrate MN-08 alleviates cerebral vasospasm and brain injury in experimental subarachnoid haemorrhage models**. *Br J Pharmacol* (2019) **176**. DOI: 10.1111/bph.14763
2. Hu X, Yan J, Huang L, Araujo C, Peng J, Gao L. **INT-777 attenuates NLRP3-ASC inflammasome-mediated neuroinflammation**. *Brain Behav Immun* (2021) **91** 587-600. DOI: 10.1016/j.bbi.2020.09.016
3. Zhang H, Ostrowski R, Jiang D, Zhao Q, Liang Y, Che X. **Hepcidin promoted ferroptosis through iron metabolism which is associated with DMT1 signaling activation in early brain injury following subarachnoid hemorrhage**. *Oxid Med Cell Longev* (2021) **2021**. DOI: 10.1155/2021/9800794
4. Xia DY, Yuan JL, Jiang XC, Qi M, Lai NS, Wu LY. **SIRT1 promotes M2 microglia polarization**. *Front Immunol* (2021) **12**. DOI: 10.3389/fimmu.2021.770744
5. Paudel YN, Angelopoulou E, Piperi C, Othman I, Shaikh MF. **HMGB1-mediated neuroinflammatory responses in brain injuries: Potential mechanisms and therapeutic opportunities**. *Int J Mol Sci* (2020) **21** 4609. DOI: 10.3390/ijms21134609
6. Heinz R, Brandenburg S, Nieminen-Kelha M, Kremenetskaia I, Boehm-Sturm P, Vajkoczy P. **Microglia as target for anti-inflammatory approaches to prevent secondary brain injury after subarachnoid hemorrhage (SAH)**. *J Neuroinflamm* (2021) **18** 36. DOI: 10.1186/s12974-021-02085-3
7. Lyu J, Jiang X, Leak RK, Shi Y, Hu X, Chen J. **Microglial responses to brain injury and disease: Functional diversity and new opportunities**. *Transl Stroke Res* (2021) **12**. DOI: 10.1007/s12975-020-00857-2
8. Atangana E, Schneider UC, Blecharz K, Magrini S, Wagner J, Nieminen-Kelha M. **Intravascular inflammation triggers intracerebral activated microglia and contributes to secondary brain injury after experimental subarachnoid hemorrhage (eSAH)**. *Transl Stroke Res* (2017) **8**. DOI: 10.1007/s12975-016-0485-3
9. Amantea D, La Russa D, Frisina M, Giordano F, Di Santo C, Panno ML. **Ischemic preconditioning modulates the peripheral innate immune system to promote anti-inflammatory and protective responses in mice subjected to focal cerebral ischemia**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.825834
10. Toscano R, Millan-Linares MC, Lemus-Conejo A, Claro C, Sanchez-Margalet V, Montserrat-de la Paz S. **Postprandial triglyceride-rich lipoproteins promote M1/M2 microglia polarization in a fatty-acid-dependent manner**. *J Nutr Biochem* (2020) **75**. DOI: 10.1016/j.jnutbio.2019.108248
11. Bernal-Chico A, Manterola A, Cipriani R, Katona I, Matute C, Mato S. **P2x7 receptors control demyelination and inflammation in the cuprizone model**. *Brain Behav Immun Health* (2020) **4** 100062. DOI: 10.1016/j.bbih.2020.100062
12. Zhao H, Liu Y, Chen N, Yu H, Liu S, Qian M. **PHLDA1 blockade alleviates cerebral Ischemia/Reperfusion injury by affecting microglial M1/M2 polarization and NLRP3 inflammasome activation**. *Neuroscience* (2022) **487** 66-77. DOI: 10.1016/j.neuroscience.2022.01.018
13. Han C, Yan P, He T, Cheng J, Zheng W, Zheng LT. **PHLDA1 promotes microglia-mediated neuroinflammation**. *Brain Behav Immun* (2020) **88**. DOI: 10.1016/j.bbi.2020.04.064
14. Hossain GS, Lynn EG, Maclean KN, Zhou J, Dickhout JG, Lhotak S. **Deficiency of TDAG51 protects against atherosclerosis by modulating apoptosis, cholesterol efflux, and peroxiredoxin-1 expression**. *J Am Heart Assoc* (2013) **2**. DOI: 10.1161/JAHA.113.000134
15. Peng H, Wang J, Song X, Huang J, Hua H, Wang F. **PHLDA1 suppresses TLR4-triggered proinflammatory cytokine production by interaction with tollip**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.731500
16. Wang J, Liang J, Deng J, Liang X, Wang K, Wang H. **Emerging role of microglia-mediated neuroinflammation in epilepsy after subarachnoid hemorrhage**. *Mol Neurobiol* (2021) **58**. DOI: 10.1007/s12035-021-02288-y
17. Ran Y, Su W, Gao F, Ding Z, Yang S, Ye L. **Curcumin ameliorates white matter injury after ischemic stroke by inhibiting Microglia/Macrophage pyroptosis through NF-kappaB suppression and NLRP3 inflammasome inhibition**. *Oxid Med Cell Longev* (2021) **2021**. DOI: 10.1155/2021/1552127
18. Chen S, Tang C, Ding H, Wang Z, Liu X, Chai Y. **Maf1 ameliorates sepsis-associated encephalopathy by suppressing the NF-kB/NLRP3 inflammasome signaling pathway**. *Front Immunol* (2020) **11**. DOI: 10.3389/fimmu.2020.594071
19. Minutoli L, Puzzolo D, Rinaldi M, Irrera N, Marini H, Arcoraci V. **ROS-mediated NLRP3 inflammasome activation in brain, heart, kidney, and testis Ischemia/Reperfusion injury**. *Oxid Med Cell Longev* (2016) **2016**. DOI: 10.1155/2016/2183026
20. Kawakita F, Kanamaru H, Asada R, Imanaka-Yoshida K, Yoshida T, Suzuki H. **Inhibition of AMPA (alpha-Amino-3-Hydroxy-5-Methyl-4-Isoxazole propionate) receptor reduces acute blood-brain barrier disruption after subarachnoid hemorrhage in mice**. *Transl Stroke Res* (2022) **13**. DOI: 10.1007/s12975-021-00934-0
21. Liu GJ, Tao T, Zhang XS, Lu Y, Wu LY, Gao YY. **Resolvin D1 attenuates innate immune reactions in experimental subarachnoid hemorrhage rat model**. *Mol Neurobiol* (2021) **58**. DOI: 10.1007/s12035-020-02237-1
22. Zhang L, Fan C, Jiao HC, Zhang Q, Jiang YH, Cui J. **Calycosin alleviates doxorubicin-induced cardiotoxicity and pyroptosis by inhibiting NLRP3 inflammasome activation**. *Oxid Med Cell Longev* (2022) **2022**. DOI: 10.1155/2022/1733834
23. Sugawara T, Ayer R, Jadhav V, Zhang JH. **A new grading system evaluating bleeding scale in filament perforation subarachnoid hemorrhage rat model**. *J Neurosci Methods* (2008) **167**. DOI: 10.1016/j.jneumeth.2007.08.004
24. Pan P, Zhao H, Zhang X, Li Q, Qu J, Zuo S. **Cyclophilin a signaling induces pericyte-associated blood-brain barrier disruption after subarachnoid hemorrhage**. *J Neuroinflamm* (2020) **17**. DOI: 10.1186/s12974-020-1699-6
25. Kim ID, Lee H, Kim SW, Lee HK, Choi J, Han PL. **Alarmin HMGB1 induces systemic and brain inflammatory exacerbation in post-stroke infection rat model**. *Cell Death Dis* (2018) **9** 426. DOI: 10.1038/s41419-018-0438-8
26. Becchi S, Buson A, Foot J, Jarolimek W, Balleine BW. **Inhibition of semicarbazide-sensitive amine oxidase/vascular adhesion protein-1 reduces lipopolysaccharide-induced neuroinflammation**. *Br J Pharmacol* (2017) **174**. DOI: 10.1111/bph.13832
27. Wu D, Yang N, Xu Y, Wang S, Zhang Y, Sagnelli M. **lncRNA HIF1A antisense RNA 2 modulates trophoblast cell invasion and proliferation through upregulating PHLDA1 expression**. *Mol Ther Nucleic Acids* (2019) **16**. DOI: 10.1016/j.omtn.2019.04.009
28. Basseri S, Lhotak S, Fullerton MD, Palanivel R, Jiang H, Lynn EG. **Loss of TDAG51 results in mature-onset obesity, hepatic steatosis, and insulin resistance by regulating lipogenesis**. *Diabetes* (2013) **62**. DOI: 10.2337/db12-0256
29. Sellheyer K, Krahl D. **PHLDA1 (TDAG51) is a follicular stem cell marker and differentiates between morphoeic basal cell carcinoma and desmoplastic trichoepithelioma**. *Br J Dermatol* (2011) **164**. DOI: 10.1111/j.1365-2133.2010.10045.x
30. Yamaguchi T, Miyamoto T, Shikata E, Yamaguchi I, Shimada K, Yagi K. **Activation of the NLRP3/IL-1beta/MMP-9 pathway and intracranial aneurysm rupture associated with the depletion of ERalpha and Sirt1 in oophorectomized rats**. *J Neurosurg* (2022) **138**. DOI: 10.3171/2022.4.JNS212945
31. Diaz-Garcia E, Nanwani-Nanwani K, Garcia-Tovar S, Alfaro E, Lopez-Collazo E, Quintana-Diaz M. **NLRP3 inflammasome overactivation in patients with aneurysmal subarachnoid hemorrhage**. *Transl Stroke Res* (2022). DOI: 10.1007/s12975-022-01064-x
32. Dodd WS, Noda I, Martinez M, Hosaka K, Hoh BL. **NLRP3 inhibition attenuates early brain injury and delayed cerebral vasospasm after subarachnoid hemorrhage**. *J Neuroinflamm* (2021) **18** 163. DOI: 10.1186/s12974-021-02207-x
33. Xing Y, Cao R, Hu HM. **TLR and NLRP3 inflammasome-dependent innate immune responses to tumor-derived autophagosomes (DRibbles)**. *Cell Death Dis* (2016) **7**. DOI: 10.1038/cddis.2016.206
34. Tao W, Hu Y, Chen Z, Dai Y, Hu Y, Qi M. **Magnolol attenuates depressive-like behaviors by polarizing microglia towards the M2 phenotype through the regulation of Nrf2/HO-1/NLRP3 signaling pathway**. *Phytomedicine* (2021) **91**. DOI: 10.1016/j.phymed.2021.153692
35. Indaram M, Ma W, Zhao L, Fariss RN, Rodriguez IR, Wong WT. **7-ketocholesterol increases retinal microglial migration, activation, and angiogenicity: A potential pathogenic mechanism underlying age-related macular degeneration**. *Sci Rep* (2015) **5**. DOI: 10.1038/srep09144
36. Chen W, Guo C, Huang S, Jia Z, Wang J, Zhong J. **MitoQ attenuates brain damage by polarizing microglia towards the M2 phenotype through inhibition of the NLRP3 inflammasome after ICH**. *Pharmacol Res* (2020) **161**. DOI: 10.1016/j.phrs.2020.105122
37. Yang CJ, Li X, Feng XQ, Chen Y, Feng JG, Jia J. **Activation of LRP1 ameliorates cerebral Ischemia/Reperfusion injury and cognitive decline by suppressing neuroinflammation and oxidative stress through TXNIP/NLRP3 signaling pathway in mice**. *Oxid Med Cell Longev* (2022) **2022**. DOI: 10.1155/2022/8729398
38. Zheng ZV, Chen J, Lyu H, Lam SYE, Lu G, Chan WY. **Novel role of STAT3 in microglia-dependent neuroinflammation after experimental subarachnoid haemorrhage**. *Stroke Vasc Neurol* (2022) **7** 62-70. DOI: 10.1136/svn-2021-001028
39. Tian Y, Liu B, Li Y, Zhang Y, Shao J, Wu P. **Activation of RARalpha receptor attenuates neuroinflammation after SAH**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.839796
40. Xie Y, Guo H, Wang L, Xu L, Zhang X, Yu L. **Human albumin attenuates excessive innate immunity**. *Brain Behav Immun* (2017) **60**. DOI: 10.1016/j.bbi.2016.11.004
41. Carlisle RE, Mohammed-Ali Z, Lu C, Yousof T, Tat V, Nademi S. **TDAG51 induces renal interstitial fibrosis through modulation of TGF-beta receptor 1 in chronic kidney disease**. *Cell Death Dis* (2021) **12** 921. DOI: 10.1038/s41419-021-04197-3
|
---
title: 'Glycemic index and insulin index after a standard carbohydrate meal consumed
with live kombucha: A randomised, placebo-controlled, crossover trial'
authors:
- Fiona S. Atkinson
- Marc Cohen
- Karen Lau
- Jennie C. Brand-Miller
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9982099
doi: 10.3389/fnut.2023.1036717
license: CC BY 4.0
---
# Glycemic index and insulin index after a standard carbohydrate meal consumed with live kombucha: A randomised, placebo-controlled, crossover trial
## Abstract
### Introduction
Kombucha is a complex probiotic beverage made from fermented tea, yet despite extensive historical, anecdotal, and in-vivo evidence for its health benefits, no controlled trials have been published on its effect on humans.
### Methods
We conducted a randomised placebo-controlled, cross-over study that examined the Glycemic Index (GI) and Insulin Index (II) responses after a standardised high-GI meal consumed with three different test beverages (soda water, diet lemonade soft drink and an unpasteurised kombucha) in 11 healthy adults. The study was prospectively registered with the Australian New Zealand Clinical Trials Registry (anzctr.org.au: 12620000460909). Soda water was used as the control beverage. GI or II values were calculated by expressing the 2-h blood glucose or insulin response as a percentage of the response produced by 50 g of glucose dissolved in water.
### Results
There was no statistically significant difference in GI or II between the standard meal consumed with soda water (GI: 86 and II: 85) or diet soft drink (GI: 84 and II: 81, ($$p \leq 0.929$$ for GI and $$p \leq 0.374$$ for II). In contrast, when kombucha was consumed there was a clinically significant reduction in GI and II (GI: 68, $$p \leq 0.041$$ and II: 70, $$p \leq 0.041$$) compared to the meal consumed with soda water.
### Discussion
These results suggest live kombucha can produce reductions in acute postprandial hyperglycemia. Further studies examining the mechanisms and potential therapeutic benefits of kombucha are warranted.
## Introduction
Kombucha is a beverage made from fermented tea that contains a complex mixture of bacteria and yeast along with a cocktail of organic acids, polyphenols, ethanol, amino acids, and various vitamins and essential elements. Kombucha is growing in popularity due to interest in the human microbiome and purported health benefits that include improvements in blood glucose, cholesterol, and blood pressure readings, and enhanced immune, liver and gastrointestinal function [1, 2]. While there is a wealth of historical and anecdotal evidence to suggest kombucha is beneficial to human health, direct research evidence is lacking [2, 3]. There are no published controlled clinical trials of kombucha and to date the only published human study of kombucha is a small uncontrolled study that revealed regular daily consumption of kombucha normalised blood glucose values in subjects with non–insulin-dependent diabetes mellitus [4].
While human data is lacking, the effect of kombucha on glucose control is well documented in animal studies. Controlled animal trials report that kombucha reduces blood glucose levels, improves lipid profiles and supports pancreatic, renal and liver function (5–7). The mechanisms of action for these effects are unclear and are likely to occur through multiple processes that include improvements in gut microbiota and islet beta cell function, inhibition of inflammation and insulin resistance, and reduced damage to the intestinal barrier [8]. Recent systematic reviews and meta-analyses on the effects of vinegar on postprandial glucose and insulin levels further suggest the acetic and gluconic acid content of kombucha contribute to its effects (9–12).
The rate at which carbohydrate is digested and released into the bloodstream is influenced by many factors, such as the food’s physical form, its fat, protein and fibre content, and the chemical structure of its carbohydrate [13]. Over three decades of research has confirmed that similar foods within the same food group can have quite different effects on blood glucose levels and therefore a food’s glycemic effect cannot be accurately predicted solely from the type and amount of carbohydrate it contains. Similarly, postprandial insulin responses of foods cannot always be predicted according to the extent to which they increase blood glucose levels (14–16). The glycemic index (GI) ranks equal available carbohydrate portions of different foods based on the extent to which they increase blood glucose levels after being eaten [17], and the Insulin Index (II) was developed to measure the postprandial insulin response in those same test portions [15]. However, most GI studies to date have not concurrently measured glycemic response alongside postprandial insulin response. The aim of this study was to determine the GI and II responses when a standard high carbohydrate, high GI meal is consumed with a complex living kombucha, compared to either soda water or diet soft drink.
## Materials and methods
This study used a randomised, single-blinded, placebo-controlled crossover design based on the internationally standardised GI testing methodology [18]. The Human Research Ethics Committee of the University of Sydney approved the study protocol (Approval number: $\frac{2017}{801}$). Participants provided written, informed consent before starting the experimental phase of the study. The study was prospectively registered with the Australian New Zealand Clinical Trials Registry (ANZCTR: 12620000460909).
## Participants
Sample size calculation ($90\%$ power) using data from published GI studies indicated 10 or more participants would be required to detect significant differences among the GI and II values of the treatments [18]. To allow for a potential outlier GI response to be excluded, 11 healthy adults with normal glucose tolerance and body mass index (BMI), aged between 18 and 45 years were recruited from the Sydney University Glycemic Index Research Service participant database. Exclusion criteria included illness or food allergy, smoking, regular medication usage other than oral contraceptives, over- or underweight, and following a restrictive diet. Participants maintained their usual food, exercise, and lifestyle habits throughout the study duration.
## Study treatments and procedures
The reference beverage, an oral glucose solution containing 50 g available carbohydrate, was prepared as 54.9 g Glucodin™ powder (iNova Pharmaceuticals Aust Pty Ltd., NSW, Australia) dissolved in 250 ml water (Table 1). The reference beverage was consumed by each participant on three separate occasions (sessions 1, 4, and 6). In addition, participants also tested three different beverage treatments which were consumed with a standardised, high GI meal. A computer-generated research randomiser program determined the randomised consumption order for each of the three meal-with-beverage treatments. Each meal and beverage treatment was consumed on one occasion, with at least 1 day in between consecutive test sessions.
**Table 1**
| Food | Weight (g) | Energy (kJ) | Protein (g) | Fat (g) | Available carbohydrate (g) | Sugar (g) | Fibre (g) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Reference glucose solution | 54.9 g glucose250 mL water | 852 | 0.0 | 0.0 | 50 | 50.0 | 0.0 |
| Rice, peas, soy sauce, and soda water | 177.1 g meal330 mL beverage | 1261 | 8.6 | 4.7 | 52.9* | 1.9 | 2.9 |
| Rice, peas, soy sauce, and diet lemonade | 177.1 g meal330 mL beverage | 1278 | 8.6 | 4.7 | 52.9* | 1.9 | 2.9 |
| Rice, peas, soy sauce, and kombucha | 177.1 g meal330 mL beverage | 1327 | 8.6 | 4.7 | 55.9*,† | 3.6 | 2.9 |
The three beverage treatments were; 330 ml of soda water (Schweppes™, Asahi Beverages, VIC, Australia) that served as a placebo control, diet lemonade soft drink (Schweppes™ Zero Sugar, Asahi Beverages, VIC, Australia), and organic kombucha (The Good Brew Company Pty Ltd., VIC, Australia). The kombucha, which was made from spring water, organic oolong and green tea along with organic sugar, contained a highly complex mix of 200 probiotic species and a high concentration of polyphenols that have been previously characterised [19]. The 330 ml of kombucha beverage contributed an additional 3 g of available carbohydrate (1.7 g of which was sugar) to the test meal, while the soda water and diet lemonade did not contain any sugar. The nutritional compositions of the three meal-with-beverage treatments are shown in Table 1. The standardised meal provided 50 g available carbohydrate from microwave Jasmine rice (147.2 g, SunRice™, Ricegrowers Ltd., NSW, Australia), with an additional 2.9 g available carbohydrate provided by green peas (20 g, McCain’s™, McCain Foods Aust. Pty Ltd., VIC, Australia) and soy sauce (10 g, Kikkoman Corporation).
The test portion of microwave Jasmine rice and frozen green peas were combined together in a bowl and cooked in the microwave for 1 min on high. The soy sauce was then added to the prepared meal and immediately served to a participant with the appropriate refrigerated test beverage (soda water, diet soft drink or kombucha). The participants were required to consume all food and fluid served and were instructed to consume the test beverage with the meal (ie. alternate mouthfuls of meal and beverage).
Participants were required to consume a carbohydrate-based evening meal, excluding legumes and alcohol, on the evening prior to each test session. On the morning of each session, participants arrived following a 10–12 h overnight fast. Two capillary blood samples (≥0.5 ml blood) were collected from a warmed hand into heparin-coated tubes in the fasted state (−5 and 0 min). Participants then consumed either the reference glucose solution or one of the test meal-with-beverage treatments within 12 min. Additional capillary blood samples were collected at regular intervals (15, 30, 45, 60, 90, and 120 min) after commencement of the reference solution or test meal. Participants were required to remain seated with minimal movement throughout each 120 min test session.
Each capillary blood sample was centrifuged at 10,000xg for 45 s immediately after collection. The plasma layer was then transferred into an uncoated tube and stored at −30°C for later glucose and insulin analysis. Plasma glucose concentration was measured in duplicate using a glucose hexokinase assay (Beckman Coulter Inc.) on an automatic centrifugal spectrophotometric clinical chemistry analyser (Beckman Coulter AU480®, Beckman Instruments Inc., United States). Plasma insulin concentration was measured using an insulin sandwich type enzyme-linked immunoassay (Insulin ELISA kit, ALPCO®, Salem, NH, United States). All samples for a given participant were analysed within the same assay.
## Data analysis
Incremental area below the 120 min postprandial plasma glucose or plasma insulin response curve (iAUC) was calculated using the trapezoidal rule, ignoring any area below the fasting concentration. Glycemic index (GI) and insulinemic index (II) values for the test meal-with-beverage treatments were determined for each participant by expressing their iAUC response for the test meal relative to their iAUC response to the reference glucose solution [18]. Standard non-parametric statistical tests (Wilcoxon signed rank test) were performed using IBM® SPSS® Statistics software (version 28) to assess differences in GI and II values, postprandial glucose and insulin responses, and change from fasting to peak glucose and insulin concentration between the treatments. $p \leq 0.05$ was considered statistically significant. Results are reported as mean ± standard error of the mean (SEM) unless otherwise stated.
## Results
Eleven healthy adults (4 males and 7 females, mean ± SD age 28.7 ± 4.5 y and mean ± SD BMI 22.6 ± 1.0 kg/m2) were screened for eligibility and completed all six test sessions (3 repeated reference glucose solutions and 3 rice with test beverage meals with no missing data points, Supplementary Figure 1). There was no significant difference amongst the mean ± SD consumption times for the three rice-based test meals (soda water: 9:56 ± 1:53 min, diet lemonade soft drink 10:02 ± 1:55 min, and kombucha: 10:04 ± 1:09 min).
## Mean glycemic response curves for the reference food and the three test meals
The mean 120 min postprandial plasma glucose response curves for the reference glucose solution and the three rice-based meals consumed with different beverages are shown in Figure 1. There were no significant differences in the mean fasting glucose concentrations amongst the reference food and any of the rice-based meals consumed with different beverages (Supplementary Table 1). The reference glucose solution produced the highest peak plasma glucose concentration at 30 min (p ≤ 0.006 compared to all test meals) and a larger overall glycemic response ($$p \leq 0.003$$ compared to the rice meal with kombucha, $$p \leq 0.041$$ compared to the other two meals). The test meals consumed with soda water and diet lemonade soft drink both produced a high peak plasma glucose response at 30 min followed by a steady decline in glycemia between 30 and 120 min. No significant differences were detected between the rice-based test meals containing soda water and diet lemonade. The rice meal containing the kombucha produced a smaller overall glycemic response than the test meal containing soda water ($$p \leq 0.041$$). The test meal containing the kombucha produced a steady rise in glycemia to a moderate plateau shaped peak response between 30 and 60 min followed by a gradual decline in plasma glucose concentration between 60 and 120 min. The change in plasma glucose from baseline to peak was significantly lower for the kombucha test meal compared to the test meal containing soda water ($$p \leq 0.003$$) and diet lemonade ($$p \leq 0.008$$).
**Figure 1:** *Postprandial glycemic response after test meals. Mean 120 min plasma glucose response curves in 11 healthy participants for the reference glucose solution and the three Jasmine rice-based test meals containing different beverages, shown as the change in plasma glucose from the fasting baseline level. Data are shown as mean ± standard error of the mean (SEM), n = 33 for the three repeated glucose solution tests (shown in black, with the three individual glucose solution tests (n = 11) shown in grey dotted lines), n = 11 for each of the three treatments: soda water meal (shown in blue), diet lemonade soft drink meal (shown in green) and kombucha meal (shown in orange).*
## Mean plasma insulin response curves for the reference food and three test meals
The mean 120 min postprandial plasma insulin response curves for the reference glucose solution and the three rice-based meals consumed with different beverages are shown in Figure 2. There were no significant differences in the mean fasting plasma insulin concentrations amongst the reference food and any of the rice-based meals consumed with different beverages (Supplementary Table 1). The reference glucose solution produced a rapid initial rise in plasma insulin concentration to the highest peak insulin response at 30 min ($$p \leq 0.033$$ compared to the soda water meal and $$p \leq 0.003$$ compared to the diet lemonade soft drink or kombucha meals) and a greater overall insulinemic response throughout the experimental period ($$p \leq 0.003$$ compared to the kombucha meal, $$p \leq 0.010$$ compared to the diet lemonade meal, and $$p \leq 0.008$$ compared to the soda water meal). Similar to their glycemic responses, the soda water and diet lemonade test meals produced a high peak plasma insulin response at 30 min whereas the kombucha test meal showed a peak insulin response at 45 min. All three test meals produced a steady decline in insulinemia between 45 and 120 min. The rice meal containing the kombucha produced a smaller overall insulinemic response than the test meal containing soda water ($$p \leq 0.033$$).
**Figure 2:** *Postprandial insulin response after test meals. Mean 120 min plasma insulin response curves in 11 healthy participants for the reference glucose solution and the three Jasmine rice-based test meals containing different beverages, shown as the change in plasma insulin from the fasting baseline level. Data are shown as mean ± standard error of the mean (SEM), n = 33 for the three repeated glucose solution tests (shown in black, with the three individual glucose solution tests (n = 11) shown in grey dotted lines), n = 11 for each of the three treatments: soda water meal (shown in blue), diet lemonade soft drink meal (shown in green) and kombucha meal (shown in orange).*
## Glycemic index and insulinemic index values of the test meals
Differences in the total glycemic and insulinemic responses produced by the reference glucose solution and test meals are more clearly reflected by their GI and II values. The reference glucose solution’s GI value (assigned 100) was significantly greater than the mean GI values produced by all three test meals: soda water meal (86 ± 6, $$p \leq 0.050$$), diet lemonade meal (84 ± 6, $$p \leq 0.041$$), kombucha meal (68 ± 4, $$p \leq 0.003$$). The mean GI value produced by the kombucha meal (68 ± 4) was significantly lower than the GI values for the soda water meal (86 ± 6, $$p \leq 0.041$$) and the diet lemonade meal (84 ± 6, $$p \leq 0.050$$), with an observed reduction equivalent to ~17 absolute GI points lower or ~$20\%$ reduction.
The reference glucose solution produced a significantly greater II value (assigned 100) compared to all three test meals: soda water meal (85 ± 4, $$p \leq 0.008$$), diet lemonade meal (81 ± 6, $$p \leq 0.013$$), kombucha meal (70 ± 4, $$p \leq 0.013$$). The II value for the kombucha meal was found to be significantly lower than the II value for the rice-based meal containing soda water ($$p \leq 0.041$$). The kombucha meal produced an 11 absolute unit reduction in II value compared to the II value for the Jasmine rice-based meal containing the diet lemonade soft drink, however this difference did not reach statistical significance ($$p \leq 0.075$$).
## Discussion
This is the first report of a controlled trial of kombucha in humans and the first study to show that compared to soda water or diet soft drink, a standard serve of kombucha reduces the acute 120 min postprandial glycemic and insulin responses to a high-GI meal. These findings suggest consumption of kombucha with meals could have important health consequences. Long term consumption of high glycemic diets, which induce high and recurrent surges in blood glucose and insulin levels, increase the risk of insulin resistance, dyslipidaemia, and the development of cardiovascular disease, non-insulin-dependent diabetes mellitus and certain cancers (20–26). Conversely, epidemiological and experimental data show that low-GI diets can reduce the risk of these diseases, improve blood glucose control and insulin sensitivity in people with diabetes, reduce high blood fat levels, and can be useful for weight control [20, 22, 23, 27, 28].
The Jasmine rice-based meals consumed with either soda water or diet lemonade were both found to produce high GI values of 86 and 84, respectively. These values are consistent with the GI values of 82 for the same Jasmine rice consumed alone in a 50 g available carbohydrate test portion (The University of Sydney, unpublished data 2019) and 84 when tested in a 25 g available carbohydrate portion [29], showing that the consumption of either of these beverages with a high-GI meal had no significant impact on postprandial glycemia. In comparison, when the kombucha was consumed with the same portion of Jasmine rice, the meal had a GI value of 68, lowering the GI rating of the meal from “high” to “medium”. The kombucha test meal showed the lowest postprandial glucose and insulin responses despite having a slightly higher available carbohydrate content in that meal (55.9 g vs. 52.9 g in the other two test meals, or 50 g in the reference glucose solution).
The mechanisms underlying the beneficial influence of kombucha on the postprandial glucose and insulin responses to the high-GI Jasmine rice meal observed in this study are not clear. The observed delayed and flattened postprandial glucose response with the kombucha beverage suggests it slows down the rate of starch digestion and absorption as previously reported with other fermented foods [30, 31] and vinegar (8–11). However, beverage acidity alone does not appear to explain the results of this study as both the kombucha (pH = 3) and the diet lemonade soft drink (pH = 3.21) [32] had similar, relatively low pH values, yet only the kombucha reduced the postprandial glucose response. The presence of antinutrients, such as tannins in kombucha, which was made from both oolong and green teas, may have also slowed the rate of carbohydrate digestion by binding to the main starch digestion enzyme, alpha-amylase [33]. It is also possible that acid-tolerant micro-organisms in the kombucha metabolised some of the glucose in the warm environment of the stomach.
It is likely that multiple mechanisms are in play and that the low pH of kombucha, the complex mix of chemical constituents including high levels of organic acids, polyphenols and tannins, and the actions of live micro-organisms micro-organisms all helped to produce the observed reductions in postprandial glucose and insulin responses. While enhanced postprandial glucose regulation is likely to have many flow-on health benefits, there may be additional benefits from regular kombucha consumption due to changes in the gut microbiota, improvements in islet beta cell function, function or reductions in insulin resistance, inflammation, or damage to the intestinal barrier, which have been observed with regular kombucha consumption in animals [8].
This study used a randomised, single-blinded, placebo-controlled, crossover design along with a research methodology that has been well-established for examining postprandial glycemic and insulinemic responses [18]. The robust design along with the use of healthy adults consuming meal and beverage combinations that are consistent with real-world scenarios, makes this research relevant to a wide audience. However, the acute nature of this study, small sample size and lack of clinical context makes it difficult to extrapolate the results to the potential impact of long-term consumption of kombucha or to people with specific diseases. Furthermore, the results cannot be generalised to other kombucha beverages as variation in tea bases, the bacteria and yeast species used as a starter culture, and specific fermentation conditions contribute to large differences in the chemicals, metabolites, microbes, and antioxidant activities of kombucha products [34].
## Conclusion
This study demonstrated a realistic, standard serve of kombucha can produce clinically significant reductions in postprandial glycemia and insulinemia in healthy adults when consumed with a high-GI, rice-based meal. Further studies examining the mechanisms and the potential therapeutic benefits of kombucha on postprandial glycemia in different populations are warranted.
## Data availability statement
The data from this study are available on request to the corresponding author.
## Ethics statement
This study involving human participants was reviewed and approved by the Human Research Ethics Committee of The University of Sydney (Approval number: $\frac{2017}{801}$). The participants provided their written informed consent to participate in this study.
## Author contributions
FA, MC, and JB-M: conceptualization. FA and JB-M: methodology. FA and KL: investigation. FA: data curation. FA and MC: writing—original draft preparation. FA, MC, KL, and JB-M: writing—review and editing. MC: funding acquisition. All authors contributed to the article and approved the submitted version.
## Funding
This research was funded by Australia Innovation Connections Grant ICG001289.
## Conflict of interest
FA and JB-M manage The University of Sydney’s glycemic index testing service and are directors of the Glycemic Index Foundation, a not-for-profit health promotion charity. FA and JB-M are authors of books in The New Glucose Revolution series (De Capo, Cambridge, MA). MC is a consultant to The Good Brew Company and co-owner of Extremely Alive Pty Ltd., which produces wellness tonics based on Good Brew Kombucha. He was involved in the design of the study, writing the manuscript and decision to publish the results. He was not involved in the collection, analyses, or interpretation of data.
The remaining author declares 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.2023.1036717/full#supplementary-material
## References
1. Abaci N, Senol Deniz FS, Orhan IE. **Kombucha - an ancient fermented beverage with desired bioactivities: a narrowed review**. *Food Chem X* (2022) **14** 100302. DOI: 10.1016/j.fochx.2022.100302
2. Kapp JM, Sumner W. **Kombucha: a systematic review of the empirical evidence of human health benefit**. *Ann Epidemiol* (2019) **30** 66-70. DOI: 10.1016/j.annepidem.2018.11.001
3. Morales D. **Biological activities of kombucha beverages: the need of clinical evidence**. *Trends Food Sci Technol* (2020) **105** 323-33. DOI: 10.1016/j.tifs.2020.09.025
4. Hiremath U, Vaidehi M, Mushtari B. **Effect of fermented tea on the blood sugar levels of NIDDM subjects**. *Indian Pract* (2002) **55** 423-5
5. Aloulou A, Hamden K, Elloumi D, Ali MB, Hargafi K, Jaouadi B. **Hypoglycemic and antilipidemic properties of kombucha tea in alloxan-induced diabetic rats**. *BMC Complement Altern Med* (2012) **12** 63. DOI: 10.1186/1472-6882-12-63
6. Srihari T, Karthikesan K, Ashokkumar N, Satyanarayana U. **Antihyperglycaemic efficacy of kombucha in streptozotocin-induced rats**. *J Funct Foods* (2013) **5** 1794-802. DOI: 10.1016/j.jff.2013.08.008
7. Bhattacharya S, Gachhui R, Sil PC. **Effect of Kombucha, a fermented black tea in attenuating oxidative stress mediated tissue damage in alloxan induced diabetic rats**. *Food Chem Toxicol* (2013) **60** 328-40. DOI: 10.1016/j.fct.2013.07.051
8. Xu S, Wang Y, Wang J, Geng W. **Kombucha reduces hyperglycemia in type 2 diabetes of mice by regulating gut microbiota and its metabolites**. *Foods* (2022) **11** 754. DOI: 10.3390/foods11050754
9. Hadi A, Pourmasoumi M, Najafgholizadeh A, Clark CCT, Esmaillzadeh A. **The effect of apple cider vinegar on lipid profiles and glycemic parameters: a systematic review and meta-analysis of randomized clinical trials**. *BMC Complement Med Ther* (2021) **21** 179. DOI: 10.1186/s12906-021-03351-w
10. Cheng LJ, Jiang Y, Wu VX, Wang W. **A systematic review and meta-analysis: vinegar consumption on glycaemic control in adults with type 2 diabetes mellitus**. *J Adv Nurs* (2020) **76** 459-74. DOI: 10.1111/jan.14255
11. Santos HO, de Moraes WMAM, da Silva GAR, Prestes J, Schoenfeld BJ. **Vinegar (acetic acid) intake on glucose metabolism: a narrative review**. *Clin Nutr ESPEN* (2019) **32** 1-7. DOI: 10.1016/j.clnesp.2019.05.008
12. Shishehbor F, Mansoori A, Shirani F. **Vinegar consumption can attenuate postprandial glucose and insulin responses; a systematic review and meta-analysis of clinical trials**. *Diabetes Res Clin Pract* (2017) **127** 1-9. DOI: 10.1016/j.diabres.2017.01.021
13. Truswell A. **Glycemic index of foods**. *Eur J Clin Nutr* (1992) **46** S91-S101. PMID: 1330533
14. Brand-Miller J, Holt SHA, de Jong V, Petocz P. **Cocoa powder increases postprandial insulinemia in lean young adults**. *J Nutr* (2003) **133** 3149-52. DOI: 10.1093/jn/133.10.3149
15. Holt SH, Miller JC, Petocz P. **An insulin index of foods: the insulin demand generated by 1000-kJ portions of common foods**. *Am J Clin Nutr* (1997) **66** 1264-76. DOI: 10.1093/ajcn/66.5.1264
16. Ostman EM, Liljeberg Elmstahl HG, Bjorck IM. **Inconsistency between glycemic and insulinemic responses to regular and fermented milk products**. *Am J Clin Nutr* (2001) **74** 96-100. DOI: 10.1093/ajcn/74.1.96
17. Jenkins DJ, Wolever TM, Taylor RH, Barker H, Fielden H, Baldwin JM. **Glycemic index of foods: a physiological basis for carbohydrate exchange**. *Am J Clin Nutr* (1981) **34** 362-6. DOI: 10.1093/ajcn/34.3.362
18. 18.International Standards Organisation. ISO/FDIS 26642:2010. Food products - determination of the glycemic index (GI) and recommendation for food classification, ISO: Geneva, Switzerland (2010).. *ISO/FDIS 26642:2010. Food products - determination of the glycemic index (GI) and recommendation for food classification* (2010)
19. Kaashyap M, Cohen M, Mantri N. **Microbial diversity and characteristics of Kombucha as revealed by metagenomic and physicochemical analysis**. *Nutrients* (2021) **13** 4446. DOI: 10.3390/nu13124446
20. 20.Joint FAO/WHO Report. Carbohydrates in human nutrition. FAO food and nutrition, Paper 66. Rome: FAO (1998).. *Carbohydrates in human nutrition. FAO food and nutrition* (1998)
21. Favero A., Parpinel M., Montella M.. **Energy sources and risk of cancer of the breast and colon-rectum in Italy**. *Adv. Exp. Med. Biol.* (1999) **472** 51-55. DOI: 10.1007/978-1-4757-3230-6_5
22. Slabber M, Barnard HC, Kuyl JM, Dannhauser A, Schall R. **Effects of a low-insulin-response, energy-restricted diet on weight loss and plasma insulin concentrations in hyperinsulinemic obese females**. *Am J Clin Nutr* (1994) **60** 48-53. DOI: 10.1093/ajcn/60.1.48
23. Holt SH, Miller JC, Petocz P, Farmakalidis E. **A satiety index of common foods**. *Eur J Clin Nutr* (1995) **49** 675-90. PMID: 7498104
24. Bouche C, Rizkalla SW, Luo J, Vidal H, Veronese A, Pacher N. **Five-week, low-glycemic index diet decreases total fat mass and improves plasma lipid profile in moderately overweight nondiabetic men**. *Diab Care* (2002) **25** 822-8. DOI: 10.2337/diacare.25.5.822
25. Salmeron J, Ascherio A, Rimm EB, Colditz GA, Spiegelman D, Jenkins DJ. **Dietary fiber, glycemic load, and risk of NIDDM in men**. *Diab Care* (1997) **20** 545-50. DOI: 10.2337/diacare.20.4.545
26. Salmeron J, Manson JE, Stampfer MJ, Colditz GA, Wing AL, Willett WC. **Dietary fiber, glycemic load, and risk of non—insulin-dependent diabetes mellitus in women**. *JAMA* (1997) **277** 472-7. DOI: 10.1001/jama.1997.03540300040031
27. Frost G, Leeds AA, Doré CJ, Madeiros S, Brading S, Dornhorst A. **Glycaemic index as a determinant of serum HDL-cholesterol concentration**. *Lancet* (1999) **353** 1045–8. DOI: 10.1016/S0140-6736(98)07164-5
28. Brand-Miller JC. **The importance of glycemic index in diabetes**. *Am J Clin Nutr* (1994) **59** 747S. DOI: 10.1093/ajcn/59.3.747S
29. Atkinson FS, Khan JH, Brand-Miller JC, Eberhard J. **The impact of carbohydrate quality on dental plaque pH: does the glycemic index of starchy foods matter for dental health?**. *Nutrients* (2021) **13** 2711. DOI: 10.3390/nu13082711
30. Sugiyama M, Tang AC, Wakaki Y, Koyama W. **Glycemic index of single and mixed meal foods among common Japanese foods with white rice as a reference food**. *Eur J Clin Nutr* (2003) **57** 743-52. DOI: 10.1038/sj.ejcn.1601606
31. Liljeberg H, Björck I. **Delayed gastric emptying rate may explain improved glycaemia in healthy subjects to a starchy meal with added vinegar**. *Eur J Clin Nutr* (1998) **52** 368-71. DOI: 10.1038/sj.ejcn.1600572
32. Schmidt J, Huang B. **The acidity of non-alcoholic beverages in Australia: risk of dental erosion**. *Int J Sci Study* (2020) **8** 28-35
33. Barrett A, Ndou T, Hughey CA, Straut C, Howell A, Dai Z. **Inhibition of alpha- amylase and glucoamylase by tannins extracted from cocoa, pomegranates, cranberries, and grapes**. *J Agric Food Chem* (2013) **61** 1477-86. DOI: 10.1021/jf304876g
34. Yang J, Lagishetty V, Kurnia P, Henning SM, Ahdoot AI, Jacobs JP. **Microbial and chemical profiles of commercial Kombucha products**. *Nutrients* (2022) **14** 670. DOI: 10.3390/nu14030670
|
---
title: Developing and testing an environmental economics approach to the valuation
and application of urban health externalities
authors:
- Eleanor Eaton
- Alistair Hunt
- Daniel Black
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9982114
doi: 10.3389/fpubh.2023.1070200
license: CC BY 4.0
---
# Developing and testing an environmental economics approach to the valuation and application of urban health externalities
## Abstract
### Background
Poor quality urban environments have substantial impacts on public and planetary health. These costs to society are not readily quantifiable and remain largely external to mainstream measures of progress. Methods for accounting for these externalities exist, but their effective application is in development. Yet there is an increasing urgency and demand given the profound threats to quality of life both now and in the future.
### Methods
We combine data from a series of systematic reviews of the quantitative evidence linking characteristics of the urban environment with health consequences and the economic valuation of these health impacts from a societal perspective within a spreadsheet-based tool. The tool–named HAUS–allows the user to estimate the health impacts of changes in urban environments. The economic valuation of these impacts in turn facilitates the use of such data in broader economic appraisal of urban development projects and policies.
### Findings
Using the Impact-Pathway approach, observations of a variety of health impacts associated with 28 characteristics of the urban environment are applied to forecast changes in cases of specific health impacts that result from changes in urban contexts. Unit values for the societal cost of 78 health outcomes are estimated and incorporated in the HAUS model in order to allow the quantification of the potential effect size of a given change in the urban environment. Headline results are presented for a real-world application in which urban development scenarios that have varying quantities of green space are evaluated. The potential uses of the tool are validated via formal semi-structured interviews with 15 senior decision-makers from the public and private sectors.
### Interpretation
Responses suggest that there is significant demand for this kind of evidence, that it is valued despite the inherent uncertainties, and has a very wide range of potential applications. Analysis of the results suggest expert interpretation and contextual understanding is critical for the value of evidence to be realized. More development and testing is needed to understand how and where it may be possible to apply effectively in real world practice.
## 1. Introduction
The quality of our urban environments impacts substantially on human and planetary health. Air pollution, lack of access to nature, low availability of healthy food and drink, and inactive lifestyles all contribute to non-communicable diseases (NCD) such as obesity, diabetes, respiratory illness, anxiety and depression. Together these NCDs make up $89\%$ of deaths in the UK, most of which are seen as preventable [1, 2]. Socio-economic pressures compound these impacts significantly [3]. Climate change may also act as a stress multiplier in urban centers, exacerbating existing problems such as overheating and flooding [4].
Some estimates of economic costs have been made that attribute costs to these risk factors in the urban environment. Though disconnected and overlapping, they do give a sense of the scale of the challenges. For example, income inequality, which is strongly linked to poor quality urban environments through quality of housing and accessibility of green infrastructure, has been estimated to result in productivity losses of £31–33 billion per year [3]. A separate estimate suggested low quality property and neighborhoods in England cost the UK National Health Service £1.4 billion annually in treatment provision [5]. Obesity, which has multiple risk factors, including obesogenic environments and “food deserts” (i.e., lack of healthy food in a local area), costs an estimated £27bn per year due to its negative effects on productivity, earnings, and welfare payments [2]. Costs from climate change are also likely to be very substantial–for example, one estimate suggests it will add £120bn to property insurance costs by 2040 and have adverse impacts on human health through overheating in buildings, storms and flooding [6].
Alongside these estimates of financial costs associated with the urban environment there is a small literature that recognizes the non-market dimension to welfare loss attributable to components of this environment. The use of economic valuation approaches in measuring, and accounting for, non-market environmental and social “goods and services,” including human health outcomes, has a substantial history [7]. However, its integration in to mainstream decision-making has been slow for a number of reasons, not least the considerable challenge of quantifying intangible aspects of health in welfare terms [8, 9]. The exception to this is in the air pollution context–an environmental hazard that is most severe in urban areas where population density is highest. For example, the UK Government estimates average damage costs–including both market- and non-market health costs of air pollution associated with particulate matter, nitrogen dioxide, ammonia, volatile organic compounds and sulfur dioxide [10]. These damage costs are disaggregated by rural and urban location, the urban locations being further disaggregated by size of conurbation [11].
This lack of uptake does not appear to imply a lack of appetite for non-market valuation. For example, a series of 30 interviews with senior decision-makers from public and private sectors suggest that there is a strong desire for more comprehensive, approaches to valuation of health in urban areas [10]. These interviews highlight a range of potential areas of application, including: government investment programs, land valuation, private sector investment and planning decisions [12, 13].
There currently exist a number of tools that generate quantitative estimates of health impacts that may be expected to result from a local policy or intervention within the urban context. WHO Europe has developed the Health Economic Assessment Tool (HEAT) for assessing changes in cycling and walking provision and patterns, using estimates of reductions in mortality as a benefit of increased active travel [14]. The tool uses a Value of a Statistical Life (VSL) to estimate the value of changes to mortality; morbidity is excluded. The ITHIM model, developed in the UK and applied there and in the US, has also been used to estimate the health impacts of transport interventions, using productivity losses and treatment costs of illness to estimate the value of attributable changes to mortality and morbidity [15].
Additional social valuation tools methods that incorporate health impacts have emerged in the UK since the United Kingdom 2012 Public Services (Social Value) Act [16], that has as a legal requirement consideration of wider social, economic and environmental benefits additional to financial profit. These include the UK Social Value Bank [17], the National TOMs framework [18] and the Manchester Cost Benefit Analysis tool [19]. Health is typically just one of many outcomes included in these models, such as employment, volunteering, crime and perceptions of local environment. These models do not offer a method for estimating potential changes to health, but rather offer a database of unit values to help policy makers estimate the social value of public sector investment such as neighborhood improvements which may impact on health. Health in these models is defined in terms of self-rated life satisfaction rather than by individual morbidity end-points. For example, unit values are given in terms of episodes of hospital attendance rather than cases of asthma. Mortality is not normally included.
We have created a tool for urban planners which allows the user to consider all determinants of health which relate to new urban housing developments. In doing so, we address gaps identified above in existing tools by estimating and valuing changes to health risk both in terms of morbidity and mortality and address a wide range of environmental determinants of health which have been linked with urban development. We provide a resource of unit costs for 76 health outcomes, disaggregated so that they can be attributed across multiple agencies from a societal perspective.
This study adopts an approach to quantification based on the Impact-Pathway method which uses known pairings in the published literature between individual characteristics of the environment, such as PM2.5 air pollution, and specific observed health outcomes to forecast changes in cases of morbidity and mortality resulting from a change in the environment. These health cases can then be monetized and aggregated to estimate the social value of an intervention [20, 21]. We extend this approach to a wider range of environmental determinants of health than has been attempted previously. In doing so we utilize the findings of a series of systematic reviews on the quantitative relationships between characteristics of the urban environments and health outcomes, and evidence on the economic welfare valuation of the identified health outcomes [22, 23]. The innovation is not in the modeling per se, but in the integration of multiple approaches, including: the systematic review of urban-health evidence, an environmental economics approach to valuation of urban health externalities, and the validation of our approach with potential end users.
This paper first outlines the approach taken to express quantitative health impacts of the urban environment in economic terms. We then present indicative findings from an application of the model in the context of an urban regeneration plan in Bristol, UK. We review these findings, reflecting critically on the current limitations to this modeling as well as its possibilities.
## 2.1. Definition of the urban form elements
In principle, our model is intended to encompass as comprehensive a range of characteristics of the urban environment as possible, thereby ensuring that consideration of any associated impacts on health in decisions relating to urban development are as complete as possible. In order to achieve this the extent was scoped in the first instance by adopting the categories defined by the Health Map, [24], a classification of the health determinants associated with the planning of human settlements published by the Royal Society for the Promotion of Public Health, and offering a comprehensive coverage of socio-environmental issues relating to urban planning and design. This classification was validated against similar classifications assembled in five other checklists including: Public Health England's Topics [25], Vancouver Healthy Toolkit [26], BREEAM Communities [27], HUDU Rapid HIA [28] and Egan Review [29] (see Supplementary material).
We then grouped the 23 aspects of the urban environment in the Health Map into five main “typologies” of urban form (or areas of search): natural environment, buildings, neighborhood design, transport and food; climate change was categorized as a “multiplier” of each element of urban form (Figure 1). Six areas from the Health Map were excluded as they are not explicitly related to elements of the urban form: living, wealth creation, resilient markets, social capital, social networks, work-life balance (as shown in gray in Figure 1).
**Figure 1:** *Identification of five main search areas (of urban form typologies), derived from the Health Map.*
## 2.2. Identification of health impacts
We identify the individual health impacts associated with the five urban form typologies on the basis of the systematic reviews previously undertaken that use these classifications [22, 23]. From the initial five search areas (Figure 1), the evidence derived through the systematic reviews resulted in a slightly changed list of urban form typologies. Buildings, Natural Environment, Climate Change and Transport remain the same, but Neighborhood Design and Food are combined into Community Infrastructure, and we use an extra category of Socio-economics to include elements such as affordability, living in areas of high poverty and renting vs. home ownership.
We understand health impacts to include mortality and non-communicable disease, including physical and mental illness, congenital deformities, injuries from road traffic and domestic accidents, loss of physical functioning and limits to daily activities, symptoms of illness such as wheezing, behaviors such as activity or diet, mental illness, obesity, and measures of wellbeing such as life satisfaction scores. We also include upper and lower respiratory tract infections, including colds and flu. We do not include dental problems, sexually transmitted disease, memory problems, educational attainment, or injuries from assault, all of which are less directly associated with the elements of the urban form that we have identified.
The epidemiological literature reported in the systematic reviews [22, 23] allow us to identify 170 urban environment characteristic-health impact pathways that observe a causal path from a specific environmental change (such as air pollution or increased green space) to a health outcome (such as increased risk of asthma or diabetes). These are listed in the Supplementary material. An example of one such pathway is presented in Figure 2.
**Figure 2:** *Illustration of how impact pathways are defined: highlighted in blue is the specific pathway for central heating improvements and asthma in children.*
Impact pathways are defined here as estimates of the magnitude of effect that a specific change within a single characteristic of the urban environment may have on a specific health outcome. These impact pathways are defined by the specific parameters from the original source study or studies, and include detail of the specific environmental change, the size and scale of the effect, the population demographics of the original study, and the individual health outcome. Where multiple pieces of evidence exist relating to the same environmental feature and the same health outcome, for example in levels of PM2.5 in air and asthma in children, data was selected on the basis of strength of evidence, robustness, and applicability to the specific UK housing context.
Evidence for change defined in the impact pathway is expressed in the form of dose-response; i.e., for a specific change in environmental characteristic, a quantitative measurable change in health status is observed as a change in the risk (known as “odds”) of illness. The epidemiological evidence is primarily described in terms of linear relationships. We judge that, given the wide range of real-world contexts, this is unlikely always to be the case. Consequently, our model outputs should be regarded as approximations of the size of health changes associated with changes in the urban environment.
In the HAUS model the aim is to identify environmental changes at an intervention specific level, so that it is possible to compare the efficacy of alternative interventions. Impact pathways are highly specific, replicating the individual parameters of the original study. For example, Figure 2 indicates how the impact pathway of Central heating improvements > asthma in children sits within the typology of Building Design and the Characteristic of Cold.
## 2.3. Specification of the model
In this paper we develop an economic tool that enables stakeholders to quantify the impact on population health of a specific intervention or policy relating to the environment in urban environments.
The tool is known as the Health Appraisal for Urban Systems tool (or HAUS for short). HAUS has three key features: *It is* therefore intended for use as a tool to support scenario appraisal and to inform broader conversations around prioritization in health in urban development. The HAUS tool covers non-communicable disease in all populations in the UK, including older adults and children–categories not disaggregated within existing tools.
It is capable of estimating effects at the neighborhood scale, and can be extended to take into account different population sizes impacted but is not designed to be used to estimate effects on an individual or a single family group.
## 2.4. Structure of the health appraisal for urban systems tool
The HAUS tool is initially a spreadsheet-based system, for ease of use and software availability, created in Excel software [30]. The structure of the tool is set out in Figure 3.
**Figure 3:** *Illustration of the structure of the Health Appraisal for Urban Systems (HAUS) tool.*
The tool includes the following: The estimation process undertaken within the HAUS tool is illustrated in Figure 4. We present health impacts in terms of estimated attributable changes to cases of illness, deaths, years of illness and years of premature life lost.
**Figure 4:** *Estimation of attributable changes in health in HAUS for an individual and for a population.*
These are derived in the following way: Each individual has an existing risk of contracting a particular illness, for example, asthma, or diabetes. We assume that if they are exposed to a change in environmental conditions, this may alter their odds, or risk, of contracting that disease. The difference between their original risk of disease, and the new odds of the disease is the amount of risk which can be attributed to the environmental change. This principle is applied in the same way to a whole population likely to be impacted by the change–the population being defined in the impact-pathway specification. Estimation of the risk of being affected by the illness in question (e.g., asthma, diabetes, etc.) is therefore made in relation to the baseline, i.e., the existing incidence rate of the illness in the population. When part of the population is exposed to a change in environment, we measure the attributable change in incidence by comparing the incidence in the exposed population to the incidence in the unexposed population.
In the HAUS tool, we therefore apply changes in odds or risk of disease which have been observed in the epidemiological literature as being significantly associated with a change in a specific environmental characteristic, or feature, to the exposed population. This method can be delineated more precisely using a notational system, described here for mortality and morbidity respectively.
## 2.5. Estimation of mortality effect
We calculate mortality in terms of two metrics: numbers of attributable deaths (Dattributed) and attributable premature years of life lost (YLLattributed). Deaths and premature life years lost here are defined as statistical lives and statistical life years lost–representing the sum of many small numbers of risks of life lost, rather than individual people.
## 2.5.1. Estimation of attributable deaths
We assume that the annual expected number of deaths in a given population (De) can be calculated by multiplying the number of people in the intervention area (n) by an average annual mortality rate (MRlit) in the baseline literature, e.g., national demographic statistics. We then estimate the proportion of the number of people exposed to a hazard as Pe and the number of unexposed people as Pu, i.e., (1-Pe).
We assume that the mortality rate for the exposed populations is affected by the odds ratio, OR, associated with the hazard so that MRe = (MRlit*OR). The OR is derived from the specific impact pathway. The following conditions therefore hold:
## 2.5.2. Estimation of attributable life years lost
We calculate the number of preventable life years lost (YLL) as following: Life years (LY) are the sum of the expected years of life in the sample population n.
(*This is* calculated on the basis of: n in each age year * life expectancy for each age year).
Premature life years lost attributed to the exposure (YLLattributed) are calculated on the basis of average Life Years (LY¯) multiplied by the number of deaths estimated in the exposed and unexposed populations.
Attributable premature years of life lost: YLLattributed = YLLe–YLLu
## 2.5.3. Estimation of morbidity effect
We estimate morbidity effects in terms of two metrics: attributable cases of illness (Cattributed) and years with illness (YLDattributed).
## 2.5.3.1. Estimation of attributable cases of illness
We assume that the expected number of cases of illness in the population (Cexpected) can be calculated by multiplying the number of people in the intervention area (n) by an incidence rate (IRlit). We again forecast the proportion of the number of people exposed to a hazard as Pe and the number of unexposed people as Pu (1-Pe). We assume that the incidence rate for exposed populations is determined by the odds ratio associated with the hazard OR so that IRe = (IR*OR).
Attributable cases of illness: Cattributed = (Cu+Ce)-Cu
## 2.5.3.2. Estimation of attributable years spent with illness
We calculate the number of years with illness or disability (YLD) based on the sum of years of life expectancy (LE) in the sample, and the expected duration of the illness (Tsick) capped with Life Expectancy (LE) so that Tsick cannot exceed LE for any individual: Attributable years lived with illness YLDattributed = Cattributed*YLD
## 2.5.4. Estimation of time effects
We calculate the sum of attributable cases or deaths over time applying the number of years of the project as a simple linear multiplier, assuming that mortality rates, morbidity incidence rates and risk ratios are linear and do not change over time.
The total effect of an intervention (Total effect) over the duration, Tintervention, is therefore as follows: We assume that there is a lag between a change in environment and full health effect of 5 years, so that in the first year only a $20\%$ of the full effect is estimated, $40\%$ in the second year, and so on, increasing by $20\%$ each time. The total effects are also capped, so that we only include health effects expected within the lifetime of the project, set at 25 years.
Life expectancy data and population demographics are derived from Office of National Statistics statistical datasets for the reference year 2019 [31].
Information on disease incidence rates and mortality rates were derived from a number of sources, including mortality data from the Office for National Statistics [32], Hospital Episode statistics from NHS Digital [33], and specific disease incidence from the Global Burden of Disease Study [34]. Information on wellbeing and mental health, activity levels and other behaviors are derived from the Health Survey for England [35].
Wherever possible, incidence rates are identified as relating to the UK population for 2019, but where UK data has not been available we have referred to data for England.
## 2.6. Estimation of value of health impacts
Economic appraisal is an integral part of decision making when policy makers seek to find the most efficient use of resources. Rooted in welfare economics, economic appraisal attempts to define whether a project makes a net contribution to social welfare. At its heart are methods for quantifying and valuing changes to individuals' utility as a result of a change in health, so that these values can be used in cost-benefit analyses, for example. In this paper we value health impacts from the societal perspective, taking into account the impact of health on the individual, their family, employers, healthcare providers and the state. This approach incorporates different components of the welfare costs of illness, including direct medical and paid care expenses, indirect lost opportunity costs such as productivity and the value of informal care time, as well as a value which monetises the disutility or pain and suffering associated with disease.
The HAUS model estimates the monetary equivalent of the disutility relating to a loss of welfare associated with risks of premature death and illness. Disutility is expressed as an individual's Willingness to Pay (WTP) to avoid illness or for improvement in health and is assumed to be the sum of the observable cost of illness (lost wages and mitigation costs) and the monetary equivalent of the non-observable cost of lost utility (mortality, pain and suffering). These non-observable costs are estimated using non-market valuation methods.
Society has mechanisms for shifting many costs of illness away from the individual–i.e., via medical insurance and sick leave policies [36]. This is particularly relevant for the UK, where most healthcare is free at the point of use. We attempt to define the societal impact of changes to health status across a population and identify where the burden of costs of illness falls. In the process of doing so, we utilize data from a range of sources including the published literature on non-market values. In this instance value transfer methods are adopted to ensure that value estimates derived in the context of previous studies are adjusted to reflect their transfer to a different context.
In order to estimate the value of identified changes in health in each impact pathway we multiply the unit values calculated for each specific health impact by the attributable health impact.
Unit values for morbidity impacts are estimated per year of ill health and per case of illness whilst unit values for mortality are estimated as the Value of a Statistical Life (VSL) and the Value of a Statistical Life Year (VSLY).
A library of reference values relating to direct and indirect costs and disutility derived from primary studies was estimated using a systematic review approach, using meta-analysis, benefits transfer techniques and quality assessment to derive reference values and ranges from the primary and secondary evidence base.
A systematic review of published literature was carried out, with additional modeling to estimate unit values for the range of health impacts included in our HAUS model. Electronic sources for peer-reviewed literature were searched, followed by reference searching. Studies were included that had clearly stated methodologies, were written in English, and which could be utilized in a UK context. The search prioritized studies from 2016 to 2020, which estimated values at an individual, per annum or per case level. Reference unit values are estimated for 76 individual health outcomes, including physical and mental illness, mortality, and health related behaviors, such as activity, obesity, and alcohol misuse.
Unit values are estimated for each health outcome in GBP £2019 (Supplementary Table 2). 2019 has been chosen as the reference year for health impacts because of the significant changes to the experience and recording of health since the COVID-19 pandemic began in the UK in March 2020. For example; we know that during this period unusual patterns occurred in expected mortality and hospital admissions, and lockdown restrictions were put in place which affected normal active behaviors [31]. This may mean that data for 2020 and 2021 are atypical for use in forecasting future trends of health.
## 3. Results—Development and testing of the HAUS model
The methodological approach outlined above and informing the HAUS model has so far been tested with external practitioners in two ways: (i) by presenting illustrative findings as part of a number of interviews with public and private sector decision-makers, and (ii) via use of the model with case study partners, focusing specifically in the first instance on green infrastructure (this second part forms part of a wider exercise developing valuations across the full range of typologies above).
## 3.1. Interviews overview
Two rounds of semi-structured interviews were undertaken with 15 senior decision-makers from a purposive sample of the UK's main urban development delivery agencies, both public and private. Methods and findings from the interviews are to be found in separate papers [12, 13]. Engagement at senior level with those who have control over key aspects of planning and development implementation—such as land disposal, investment, development delivery and planning permission—was central to the approach. Field notes of the interviewee responses to four questions on the economic valuation are included in the Supplementary material, and summary reflections provided below.
## 3.2.1. Background
Frome *Gateway is* a 14.7 hectare site in the center of Bristol. The site has been designated a strategically important site in need of major regeneration by Bristol City Council [37]. A map of the site can be seen on the Bristol City Council website for Frome Gateway [38].The draft Local Plan set out the ambition for the site to be developed as a new mixed-use neighborhood, including around 1,000 new homes, improved access to the River Frome and existing green spaces, improved connectivity to the site generally, and improved opportunities for work and leisure [37].
HAUS was used to provide detailed information on expected health outcomes related to the scenarios under development and so to increase knowledge about the potential for environmental impacts on health. The specific objectives for evaluating the impact of changes to parks and green spaces includes the following:
## 3.2.2. Parameters and scenario building
In order to provide comparative information, four scenarios were developed presenting alternative patterns of development for the site:
## 3.2.2.1. Baseline scenario: Present day conditions
We assume that the existing quality, condition and area of green space, including the parks and river areas, are as the present day. The site has 2.37Ha of green space mostly contained within two parks: Riverside Park and Peel Street Open Space [39].
## 3.2.2.2. Definition of future scenarios
A project lifetime of 25 years is assumed. Effects are estimated for an area including a buffer of 300 m around the perimeter of the site, which is used to take into account effects on local communities.
## 3.2.3. Data
Information on Green Infrastructure was derived using Natural England's Green Infrastructure Framework Map tool [40]. The Normalized Difference Vegetation Index (NDVI) score for the site is assumed to be 0.15, with the NDVI for Riverside Park estimated at 0.29. NDVI is used as a measure of exposure to greenness in several of the health studies used in HAUS. NDVI uses satellite imagery to estimate the greenness of an area, with higher scores on a range of −1 to 1 indicating higher levels of greenness. NDVI can be useful as a way of determining the proximity of different types of vegetation, such as grass and trees [41].
Assumptions around environmental conditions are derived from the Development Assumptions Report, technical reports and local site visits [39, 42]. Local residents' perceptions of the area, activity levels and usage of parks/open spaces were not known, so a survey of 108 residents living near to the site was carried out in 2022. The survey results provided input to the HAUS model and further contextual information for the regeneration team.
## 3.2.4. Population
The total affected population, including those within 300 m of the site, is estimated at 9,241. We assume 3,000 residents live within the site boundary in all scenarios. This is based on the provision of around 1,000 new homes with an average occupancy of 2.5 per household, plus an additional 500 student residences.
At present only a small number of homes are present within the site boundary. For easier comparison, the baseline scenario adopts a hypothetical 3,000 residents, reflecting the projected population size in the future scenarios.
## 3.2.5.1. Health benefits of green space at Frome Gateway: Baseline
The results for the baseline indicate that the existing green space are likely to provide a range of health benefits, especially for adults using the parks who are found to experience reductions in diabetes and reduced risk of weight gain. These are shown in Tables 1, 2. There may, however, be a negative effect from green space on risk of asthma in children, deriving e.g., from pollen: from 8 expected cases in this age group, we estimate a potential increase of 5 attributable cases per year.
## 3.2.5.2. Health benefits of green space at Frome Gateway: Future scenarios
The potential changes to health in cases under each scenario are compared in Table 3.
**Table 3**
| Environment change (intervention, asset or hazard) | Health outcome | Baseline casesa | S1 casesb | S2 casesc | S3 casesd |
| --- | --- | --- | --- | --- | --- |
| NDVI increase | Cancer (mouth and throat) | - | −0.23 | −0.23 | −0.23 |
| NDVI increase | Respiratory (asthma) | - | - | 8 | 8 |
| NDVI increase | Weight gain | - | - | −41 | −41 |
| Proximity to green space | Activity | 2632 | 2632 | 2632 | 2632 |
| Proximity to green space | Mental health | 125 | 125 | 125 | 125 |
| Proximity to green space | Respiratory (asthma) | 5 | 5 | 5 | 5 |
| Proximity to large, attractive, open space | Activity | 1432 | 1432 | 1432 | 1432 |
| Quality of green space (pleasantness) | Life satisfaction | 125 | 125 | 125 | 125 |
| Quality of green space (safety) | Life satisfaction | 125 | 125 | 125 | 125 |
| Size of public open spaces | Diabetes | - | - | −10 | - |
| Park improvements | Park use | - | 171 | 171 | 171 |
| Park use | Diabetes | −3 | −3 | −3 | −3 |
| Park use | Weight gain | −127 | −139 | −139 | −139 |
Improvements to park quality and safety are assumed to lead to increased park use in all scenarios: this has benefits in terms of reduced risk of diabetes and weight gain.
Scenarios 1–3 indicate a possible increase in NDVI which may lead to reduced risk of mouth and throat cancer. Scenarios 2 and 3, which have the largest potential for increases in NDVI score, indicate potential reductions in risk of being overweight or obese for children.
Greenness, estimated via NDVI, may continue to have an impact on increased risk of asthma in children, and this effect is seen in higher values for Scenarios 2 and 3 where there is the most potential for higher NDVI scores.
Health outcomes such as activity, wellbeing and life satisfaction identified in the baseline scenario are not shown to change under the three future scenarios, indicating that the threshold for these is already met by the existing provision of green space.
## 3.2.5.3. Provision of a large park vs. small pocket parks
Only one change is unique for Scenario 2 compared with the other scenarios, and that relates to an increase in the size of public open spaces. The relevant impact-pathway is found to relate to cases of diabetes and has a specific threshold value of 0.7 hectares. In Scenario 2, the specific provision of an additional park of around 1 hectare unlocks this pathway, potentially leading to a reduction in 10 cases of diabetes from a baseline of 41 cases in the population considered here. Over 25 years, this could lead to savings in health valued at around £22.7 million. In Scenario 3, where additional green space is dispersed across the site, this threshold is not reached and these benefits are therefore not realized.
## 3.2.5.4. Valuation of health effects over the lifetime of the project
Figure 5 indicates the potential value of attributable changes to morbidity by individual impact-pathways related to green space. This is the sum of the value of changes in years of illness over 25 years under each of the scenarios.
**Figure 5:** *Estimate of economic valuation of attributable changes to morbidity over 25 years under each scenario.*
It may be helpful to summarize the total value of health changes by scenario: the figures below have not been adjusted for double counting, but to sum the total effect of each scenario in turn may help indicate the magnitude of the difference between them:
## 3.2.6. Sensitivity analysis
We have assumed that the highest change in NDVI score would be 0.105–0.15 points, which is not enough to reach the HAUS threshold for reductions in premature mortality. If the NDVI could be raised by 0.24 points for the site, we estimate that premature mortality might be reduced by 3 cases per year–equivalent to around 680 premature life years over the lifetime of the project, at a value of around £41.5 million. However, a change in NDVI on this scale represents a dramatic change in the land use at Frome Gateway, including considerably more tree cover, and may not be achievable or appropriate for this urban site given other ambitions such as provision of housing and business space.
## 4. Discussion
There is growing demand for new approaches that will enable us to better account for the social and environmental external costs in urban development. A range of public- and private-sector stakeholders face a significant challenge in how they interpret and respond to evidence on a wide range of external costs and navigate the conflicts that these may generate with competing development objectives.
This study brings together in a model a substantial account of the current evidence base relating to the quantification and economic valuation of health impacts associated with the urban environment. Societal costs of illness for 78 health outcomes are incorporated into a model that represents 28 characteristics of the built environment. This approach offers an evidence-led way of comparing the effects of different urban form elements in terms of the potential magnitude of impact on health.
The interviewee responses suggest that both public and private sector representatives appear to be aware of many of the major health challenges posed by poor-quality urban environments. However, interviewees also recognized that health is not factored adequately into the urban planning process. There appeared to be considerable support for greater use of economic valuation to help improve decision-making. More specifically, interviewees suggested a very wide range of potential leverage points at which this type of valuation might be fed into the urban development system, at national and local level [see interview findings in Black et al. [ 12], Figure 1]. It was recognized that there is no “silver bullet” solution, with quantitative valuation of health impacts just one possible mechanism amongst the range of interventions needed.
With regards to the green infrastructure modeling, the HAUS model highlights the important role that existing parks and green spaces have for the health of local people, as well as the potential health benefits of improving the quality and quantity of these spaces. However, it importance will only be clear to potential users if outputs are presented in easy-to-digest forms and in ways meaningful to them. The exercise also serves to emphasize the need to define and measure changes in the urban environment–in this case potential changes to NDVI scores for the site under different scenarios, which may be resource-intensive for the model user.
The strength of the HAUS tool lies in its capacity to synthesize evidence from two strands of literature–on health impact pathways and economic valuation of health impacts–and combine it in such a way that specific project- or policy-based changes in the urban environment can be evaluated against health-related criteria. At the same time, there are inherent challenges in synthesizing evidence from across such a wide range of urban health and economic valuation literature which itself is derived from a diverse range of empirical contexts, using contrasting methodologies, assumptions and reporting protocols. There are also wide divergences in the quantity and quality of evidence available across the range of environmental characteristics. For example, children's health forms an important component of costs when relating to air quality, noise and food environment. However, there is very limited evidence on child health in the economic valuation literature. The resulting health impacts are therefore likely to be significantly undervalued.
We have not explored fully here the extensive uncertainties which are clearly present, therefore, within every aspect of the modeling process and this may be thought to reduce the value of the tool outputs. However, there are significant uncertainties inherent to any form of economic valuation of health outcomes. Such evidence is nonetheless widely used across decision and policy making systems and is especially prevalent in areas that require significant investment, such as in health and urban infrastructure, where it is currently used to justify expenditure [24, 43]. Thus, uncertainties in valuation, however sizeable, do not necessarily negate its' usefulness; this depends on how that information is understood, used and valued. Future development of this tool and comparable endeavors that address the need for health to be given adequate weighting in urban development processes therefore require substantial attention being given to how such data can be most effectively communicated to the full range of stakeholder types. At the same time, further research is needed to help fill the more sizeable gaps identified in the literature so that, for example, there is a re-balancing of the weight of evidence toward areas other than air pollution in both health quantification and health valuation.
## 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
EE: methodology, investigation, writing–original draft, and visualization. AH: conceptualization, methodology, supervision, and writing–review and editing. DB: conceptualization, writing–original draft, and writing–review and editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
DB was employed by Daniel Black + Associates | db+a. 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/fpubh.2023.1070200/full#supplementary-material
## References
1. Dannenberg A, Frumkin H, Jackson R. **Making**. (2011.0)
2. 2.UK House of Lords. Select Committee on the Long-term Sustainability of the NHS; The Long-term Sustainability of the NHS and Adult Social Care. Chapter 6: Public health, prevention and patient responsibility. (2017). Available online at: https://publications.parliament.uk/pa/ld201617/ldselect/ldnhssus/151/15102.htm (accessed May 20, 2022).. *Select Committee on the Long-term Sustainability of the NHS; The Long-term Sustainability of the NHS and Adult Social Care* (2017.0)
3. Marmot M, Goldblatt P, Allen J. *Fair Society, Healthy Lives: The Marmot Review: Strategic Review of Health Inequalities in England post-2010* (2010.0)
4. Scally G, Black D, Pilkington P, Williams B, Ige J, Prestwood E. *UPSTREAM: Moving Planetary Health Upstream in Urban Development Decision-Making - A Three-year Pilot Research Project* (2019.0)
5. Garrett H, Mackay M, Nicol S, Piddington J, Roys M. *The Cost of Poor Housing in England* (2021.0)
6. 6.Financial Times. Climate Risks to add $183bn to Property Insurance Costs by 2040, Swiss Re predicts. (2021). Available online at: https://www.ft.com/content/5d271251-973d-45e5-8982-2e28bf96f952 (accessed May 20, 2022).. *Climate Risks to add $183bn to Property Insurance Costs by 2040, Swiss Re predicts* (2021.0)
7. Pearce D. **An intellectual history of environmental economics**. *Annu Rev Energy Environ* (2002.0) **7** 57-81. DOI: 10.1146/annurev.energy.27.122001.083429
8. Brazier J, Ratcliffe J. **The Measurement and Evaluation of Health for Economic Valuation**. *International Encyclopaedia of Public Health* (2008.0) 252-61
9. Mayer C. *Firm Commitment: Why the Corporation is Failing us and How to Restore Trust in it* (2013.0)
10. 10.Department for Environment Food Rural Affairs Air Quality Appraisal: Impact Pathways Approach. (2023). Available online at: https://www.gov.uk/government/publications/assess-the-impact-of-air-quality/air-quality-appraisal-impact-pathways-approach (accessed January 5, 2023).. *Department for Environment Food Rural Affairs Air Quality Appraisal: Impact Pathways Approach.* (2023.0)
11. 11.Department for Environment Food Rural Affairs Air Quality Appraisal: Damage Cost Guidance. (2023). Available online at: https://www.gov.uk/government/publications/assess-the-impact-of-air-quality/air-quality-appraisal-damage-cost-guidance (accessed January 5, 2023).. *Department for Environment Food Rural Affairs Air Quality Appraisal: Damage Cost Guidance.* (2023.0)
12. Black D, Pilkington P, Williams B, Ige J, Prestwood E, Hunt A. **Overcoming systemic barriers preventing healthy urban development in the UK: Main findings from interviewing senior decision-makers during a 3-year planetary health pilot**. *J Urban Health* (2021.0) **98** 415-27. DOI: 10.1007/s11524-021-00537-y
13. Scally G, Black D, Pilkington P, Williams B, Ige J, Prestwood E. **The application of ‘elite interviewing’ methodology in transdisciplinary research: a record of process and lessons learned during a 3-year pilot in urban planetary health research**. *J Urban Health.* (2021.0) **98** 404-14. DOI: 10.1007/s11524-021-00542-1
14. Kahlmeier S, Cavill N, Dinsdale H, Rutter H, Götschi T, Foster CE. *Health Economic Assessment Tools (HEAT) for Walking and Cycling. Methodology and User Guide 2014 Update, World Health Organization Regional Office for Europe, Copenhagen* (2014.0)
15. Whitfield GP, Meehan LA, Maizlish N, Wendel AM. **The integrated transport and health impact modeling tool in Nashville, Tennessee, USA: implementation steps and lessons learned**. *J Transp Health.* (2017.0) **5** 172-81. DOI: 10.1016/j.jth.2016.06.009
16. 16.United Kingdom Public Services (Social Value) Act 2012 Ch.3, The Stationery Office, London, (2012). Available online at: https://www.legislation.gov.uk/ukpga/2012/3/enacted (access date November 11, 2020).. *United Kingdom Public Services (Social Value) Act 2012 Ch.3, The Stationery Office, London* (2012.0)
17. Fujiwara D. *Measuring the Social Impact of Community Investment: The Methodology Paper, HACT, UK* (2014.0)
18. 18.Social Value Portal. The Method behind the Movement. National TOMs. (2020). Available online at: https://socialvalueportal.com/national-toms/ (accessed October 20, 2020).. *The Method behind the Movement. National TOMs* (2020.0)
19. **Cost Benefit Analysis**. *Greater Manchester Combined Authority* (2020.0)
20. Holland M, Hunt A, Hurley F, Watkiss P. *Methodology for Carrying out the Cost-Benefit Analysis for CAFE Volume 2: Health Impact Assessment* (2005.0)
21. Hunt A, Ferguson J. *Towards Comprehensive Economic Valuation of EDC Health Impacts* (2015.0)
22. Ige-Elegbede J, Pilkington P, Orme J. **The relationship between buildings and health: a systematic review**. *J Public Health.* (2019.0) **41** e121-e132. PMID: 30137569
23. Ige-Elegbede J, Pilkington P, Orme J. **Designing healthier neighbourhoods: a systematic review of the impact of the neighbourhood design on health and wellbeing**. *Cities Health* (2020.0) **6** 1004-1019. DOI: 10.1080/23748834.2020.1799173
24. Barton H, Grant M. **A health map for the local human habitat**. *J Roy Soc Promot Health* (2006.0) **126** 252-53. DOI: 10.1177/1466424006070466
25. 25.GOV UK. Spatial Planning for Health: Evidence Review. Public Health England. (2017). Available online at: https://www.gov.uk/government/publications/spatial-planning-for-health-evidence-review (accessed February 6, 2023).. *Spatial Planning for Health: Evidence Review* (2017.0)
26. 26.BC Centre for Disease Control. Healthy Built Environment Linkages Toolkit. (2018). Available online at: http://www.bccdc.ca/health-professionals/professional-resources/healthy-built-environment-linkages-toolkit (accessed February 6, 2023).. *Healthy Built Environment Linkages Toolkit* (2018.0)
27. 27.BRE. BREEAM Communities: Science-based Sustainability Framework for the Verification and Certification of New Assets. (2018). Available online at: https://bregroup.com/products/breeam/breeam-technical-standards/breeam-communties/#BC-Tech-manual (accessed February 6, 2023).. *BREEAM Communities: Science-based Sustainability Framework for the Verification and Certification of New Assets* (2018.0)
28. 28.NHS London Healthy Urban Development Unit. HUDU Planning for Health: Rapid Health Impact Assessment Tool. (2019). Available online at: https://www.healthyurbandevelopment.nhs.uk/wp-content/uploads/2019/10/HUDU-Rapid-HIA-Tool-October-2019.pdf (accessed February 6, 2023).. *HUDU Planning for Health: Rapid Health Impact Assessment Tool* (2019.0)
29. Egan J. *The Egan Review: Skills for Sustainable Communities* (2004.0)
30. **Microsoft Excel for Microsoft 365 (Version 2202)**
31. 31.Office for National Statistics. Available online at: https://www.ons.gov.uk/ (accessed December 30, 2022).
32. 32.Office for National Statistics. Excess Deaths in England and Wales: March 2020 to June 2022. Available online at: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/excessdeathsinenglandandwalesmarch2020tojune2022/2022-09-20 (accessed December 30, 2022).. *Excess Deaths in England and Wales: March 2020 to June 2022*
33. 33.NHS Digital. Hospital Admitted Patient Care Activity 2019-20: Diagnosis Data Tables, NHS Digital. (2020). Available online at: https://digital.nhs.uk/data-and-information/publications/statistical/hospital-admitted-patient-care-activity/2019-20 (accessed September 1, 2022).. *Hospital Admitted Patient Care Activity 2019-20: Diagnosis Data Tables, NHS Digital* (2020.0)
34. 34.Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2019 (GBD 2019) Results. Institute for Health Metrics and Evaluation (IHME), Seattle, United States. (2020). Available online at: https://vizhub.healthdata.org/gbd-results/ (accessed February 6, 2023).. *Global Burden of Disease Study 2019 (GBD 2019) Results. Institute for Health Metrics and Evaluation (IHME), Seattle, United States* (2020.0)
35. 35.NHS Digital. Health Survey for England 2019: Adults' health (data tables). NHS Digital (2020). Available online at: https://digital.nhs.uk/data-and-information/publications/statistical/health-survey-for-england/2019/health-survey-for-england-2019-data-tables (accessed September 1, 2022).. *Health Survey for England 2019: Adults' health (data tables)* (2020.0)
36. Freeman AM, Herriges JA, Kling CL. *The Measurement of Environmental and Resource Values: Theory and Methods (Third Edition)* (2014.0)
37. 37.Bristol City Council. Bristol Local Plan Review: Draft Policies Development Allocations – Consultation (March 2019), Bristol City Council, Bristol. (2019). Available online at https://www.bristol.gov.uk/files/documents/2275-local-plan-review-draft-policies-and-development-allocations/file (accessed September 1, 2022).. *Bristol Local Plan Review: Draft Policies Development Allocations – Consultation (March 2019), Bristol City Council, Bristol* (2019.0)
38. Bristol City Council, Frome Gateway
39. 39.AHMM for Bristol City Council Frome Gateway Development Assumptions Report Version 01. London: AHMM. (2022).. *AHMM for Bristol City Council Frome Gateway Development Assumptions Report Version 01* (2022.0)
40. 40.Natural England, Green Infrastructure, Framework, Natural, England. (2021). Available online at: https://designatedsites.naturalengland.org.uk/GreenInfrastructure/Map.aspx (accessed September 26, 2022).. *Natural England, Green Infrastructure, Framework, Natural, England* (2021.0)
41. Dadvand P, Villanueva CM, Font-Ribera L, Martinez D, Basagaña X, Belmonte J. **Risks and benefits of green spaces for children: a cross-sectional study of associations with sedentary behavior, obesity, asthma, and allergy**. *Environ Health Perspect* (2014.0) **122** 1329-35. DOI: 10.1289/ehp.1308038
42. 42.AHMM for Bristol City Council Frome Frome Gateway Stage 2 - End of Stage Report. Unpublished internal document. (2022).. *AHMM for Bristol City Council Frome Frome Gateway Stage 2 - End of Stage Report* (2022.0)
43. 43.DfT. Transport Analysis Guidance. An Overview of Transport Appraisal. Department for Transport. (2014). Available online at: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/938759/tag-overview.pdf~. *Transport Analysis Guidance. An Overview of Transport Appraisal. Department for Transport* (2014.0)
|
---
title: Terahertz spectra of proteinuria and non-proteinuria
authors:
- Zhenrui Xue
- Ping Mao
- Ping Peng
- Shihan Yan
- Ziyi Zang
- Chunyan Yao
journal: Frontiers in Bioengineering and Biotechnology
year: 2023
pmcid: PMC9982117
doi: 10.3389/fbioe.2023.1119694
license: CC BY 4.0
---
# Terahertz spectra of proteinuria and non-proteinuria
## Abstract
In clinical practice, proteinuria detection is of great significance in the diagnosis of kidney diseases. Dipstick analysis is used in most outpatient settings to semi-quantitatively measure the urine protein concentration. However, this method has limitations for protein detection, and alkaline urine or hematuria will cause false positive results. Recently, terahertz time-domain spectroscopy (THz-TDS) with strong hydrogen bonding sensitivity has been proven to be able to distinguish different types of biological solutions, which means that protein molecules in urine may have different THz spectral characteristics. In this study, we performed a preliminary clinical study investigating the terahertz spectra of 20 fresh urine samples (non-proteinuria and proteinuria). The results showed that the concentration of urine protein was positively correlated with the absorption of THz spectra at 0.5–1.2 THz. At 1.0 THz, the pH values (6, 7, 8, and 9) had no significant effect on the THz absorption spectra of urine proteins. The terahertz absorption of proteins with a high molecular weight (albumin) was greater than that of proteins with a low molecular weight (β2-microglobulin) at the same concentration. Overall, THz-TDS spectroscopy for the qualitative detection of proteinuria is not affected by pH and has the potential to discriminate between albumin and β2-microglobulin in urine.
## Introduction
Protein analysis of urine is an important clinical test, mainly used for diagnosis and treatment monitoring of kidney diseases, such as nephritis and renal failure. Healthy people lose less than 150 mg of protein from urine daily; therefore, a total protein excretion rate higher than 150 mg/24 h is considered proteinuria (Lamb et al., 2009). Proteinuria can be divided into glomerular proteinuria and tubular proteinuria according to the protein type. Glomerular proteinuria is dominated by albumin, whereas tubular proteinuria is dominated by β2-microglobulin (Tampe et al., 2021). At present, the protein qualitative test of urine is commonly performed using the sulfosalicylic acid method, the heated acetic acid method, and the dipstick test (Kumar and Gill, 2018). The sulfosalicylic acid method reacts with both albumin and globulin with high sensitivity (50 mg/L) and is preferred for screening tests, but strongly alkaline urine (pH > 9) or strong acid urine (pH < 3) is prone to false negative. A reaction time of more than 1 min will increase the probability of positive results, so the results need to be observed on time (Ridley, 2018). The heated acetic acid method reacts to both albumin and globulin but is less sensitive (150 mg/L) (Rahayu and Rustiana, 2020). The dipstick test is sensitive to albumin but not to globulin, and its sensitivity (70 mg/L) is slightly lower than that of the sulfosalicylic acid method. False positives can occur when urine pH is elevated due to renal tubular acidosis, therapy of alkaline drugs, diet (vegetarian) (Carroll and Temte, 2000; Roberts, 2007; Clarkson et al., 2011). However, the dipstick test is often used to screen a large number of samples and medical emergencies duo to its simple operation, low cost. These biochemical methods require additional reagents and often rely on the naked eye to determine the results, which increases the complexity of the operation and reduces the accuracy of the results. It is necessary to develop a rapid and accurate screening method for proteinuria.
Terahertz (THz) waves are electromagnetic waves with frequencies between 0.1 and 10.0 THz, and their energy level range covers the energy levels of low-frequency motions (vibrations and rotations) of many organic macromolecules and hydrogen bonding networks in solution [9]. These motions can be identified by fingerprint features in the THz transmission or absorption spectra, absorption changes, and phase changes (Zheng et al., 2014; Yan et al., 2016). Changes of biomolecules, such as change of formation, content, and state, strongly affect the THz spectral characteristics (Niehues et al., 2011; Bye et al., 2014; Klokkou et al., 2022). Recently, THz wave has been proved to have strong interactions with many biochemical molecules such as amino acids, proteins, and deoxyribonucleic acid (DNA) (Esser et al., 2018; Wu et al., 2020; Mancini et al., 2022). Moreover, THz waves have low photon energies (1 THz = 4.1 meV), which is about one million times weaker than the energy of X-ray photons, so it does not cause any harmful ionization in biological molecules (Sirkeli et al., 2018). Because of these properties, THz spectroscopy has emerged as a powerful tool for studying solvated biomolecules (Xie et al., 2014).
Blood is rich in solvated molecules, and many studies have been conducted using THz-TDS spectroscopy to detect blood components. Reid et al. found that there were significant differences in both the absorption spectra and refractive index spectra of the whole blood and thrombus, demonstrating the potential of THz spectroscopy in distinguishing different combinations of substances in human blood (Reid et al., 2013). Furthermore, Torii et al. measured the reflectance of sub-THz radiation on the concentration of glucose and albumin, and concluded that sub-THz radiation can be used to measure blood glucose and albumin (Torii et al., 2017). In addition, Chen et al. found a linear relationship between the THz absorption coefficient and blood glucose level through a quantitative analysis of 70 patients (Chen et al., 2018). Since both blood and urine are liquid biological specimens with complex compositions, and urine is essentially a product of ultra-filtered plasma, the study of THz-TDS spectroscopy for blood composition provides a theoretical basis for the use of THz-TDS spectroscopy for urine protein measurement. Therefore, we hypothesize that THz-TDS spectroscopy has the potential to detect proteins in urine samples and solve the problem of false positives and inaccuracy of the dipstick test in proteinuria screening.
In this study, we propose a new approach using THz-TDS for the qualitative detection of proteinuria. First, we compared the absorption spectra of proteinuria and non-proteinuria. Next, we investigated the effect of pH (6.0–9.0) on the analysis of proteinuria. Finally, we investigated the absorption spectra of β2-microglobulin and albumin to verify the ability of THz-TDS to distinguish different kinds of proteinuria.
## Experimental materials
Twenty urine samples (mean age 56 years; range 52–58 years) were provided by the department of laboratory medicine, Southwest Hospital, Chongqing. Our study was approved by Ethics Committee of the First Affiliated Hospital of Army Medical University (No. KY2020227). All the chemicals were purchased from Aladdin (Shanghai, China). Solutions were prepared with deionized water (Millipore, United States). Human Serum Albumin (HSA) (CAS No. 70024-90-7) was purchased from Solarbio (Beijing, China). Urine sample from tubular proteinuria patients which containing human β2-microglobulin was purchased from Prospec-Tany Technogene Ltd (Ness-Ziona, Israel). Urinalysis strips (H12-MA) were purchased from DiRui (Changchun, China).
For proteinuria testing, 20 samples with different protein concentrations and other parameters (pH, leukocytes, bilirubin, etc.) were chosen to be similar. The 20 samples were equally divided into four groups based on protein concentration:—(<0.1 g/L), + (0.3–1.0 g/L), ++ (1.0–3.0 g/L), +++ (3.0–6.0 g/L). For the test of effect of pH value on urine samples, HCl and NaOH were used to adjust the pH value of urine samples to 6.0, 7.0, 8.0, and 9.0, respectively. Normal urine sample was selected and divided into two parts, different concentrations of protein (albumin and β2-microglobulin) from 1 g/L to 10 g/L were added to observe the changes in the absorption spectrum of THz.
## Urine test strip analysis
The pH and qualitative protein results contained in the urinalysis strips were reported by the fully automated UC-3500 (Sysmex, Kobe, Japan). Data are presented in the reports on an ordinal scale (as “normal,” “negative,” “positive,” or “nominal concentrations”).
## Experimental equipment and sample cell
A Picometrix T-ray 5000 fiber-coupling spectrometer (Advanced Photonix, Inc., MI, United States) was used in the experiment (Figures 1A, B). The spectrometer generated and coherently detected the electric field of ultrashort THz electromagnetic pulses in the time domain using femtosecond near-infrared laser pulses and LT-InGaAs photoconductive antenna chips. The femtosecond pulse was produced by a Sapphire oscillator with a repetition rate of 100 MHz, a central wavelength of 1064 nm, and a duration of <100 fs. This pulse was split into two parts by a polarising beam splitter, one as a probe beam shining directly on the photoconductive antenna (PCA) and the other as pump light collected on the other PCA. The detection light was discretely sampled from the terahertz signal irradiating the second PCA to obtain a time domain waveform, which was transformed into the frequency domain using the Fast Fourier Transform. The sampling interval of the THz-TDS was 0.1 ps, and the spectral resolution was 12.5 GHz. More detailed descriptions can be found in our previous reports (Yan et al., 2016; Zang et al., 2019).
**FIGURE 1:** *Experimental equipment (A) Schematics of the THz-TDS (B) The actual experimental setup (C) Schematic diagram of the sample cell.*
For the liquid measurement, the sample cells were prepared as follows: polyethylene glycol terephthalate (PET) double-sided tape (3M, MN, United States of America) of 0.10 mm thickness was cut into square spacers with side dimensions of 22.0 mm, in which a circle with a diameter of 15.0 mm was removed simultaneously, for accommodating the liquid sample (Figure 1C). By sealing the spacer with two polytetrafluoroethylene (PTFE) cover slips (Fisher Scientific, MA, United States) pasted together after the addition of the liquid samples, removal of excessive solution from any side of the spacer was possible to ensure that the spacer was completely filled. With a diameter of 15.0 mm and a thickness of 0.10 mm, the volume of the secure seal spacer was calculated to be ∼17.7 μL, which was also the amount of sample added. The entire procedure takes less than a minute and the materials are inexpensive.
## THz-TDS measurement
The THz frequency range was chosen between 0.5 and 1.2 THz to obtain the maximum signal-to-noise ratio and the most stable signal. Spectral frequency resolution of the spectrometer is 12.5 GHz. The measurement temperature was controlled at 21°C ± 0.4°C. The relative humidity was maintained at < $2.0\%$ by nitrogen gas purging. An empty spacer was used as a reference to eliminate the background effects and each sample was measured seven times. After the sample chamber was prepared, it was sandwiched between a PTFE gasket and a neoprene gasket for measurement, and the frequency spectrum of the measured signal was obtained by using the Fourier transform.
## Data analysis
The solution absorption coefficient is the energy loss of the terahertz beam in the medium. Taking into account the effect of reflection between the PTFE window and the sample, the absorption coefficient can be calculated by the improved Beer-Lambert theoretical formula as follows (Zang et al., 2019). αω=2dln4nωnqωnω+nqω2.nqω+124nqω.1Aω [1] Where d is the thickness of the urine sample cell; n q (ω) = 1.43 is the refractive index of PTFE; A(ω) is the amplitude ratio of the Fourier transform of the urine sample cell (I s) and the blank sample cell (I ref); n(ω) is the refractive index of the sample.
## Statistics analysis
The statistical analysis was performed using the SPSS software package, version 16.0 (SPSS, Chicago, IL, United States). The data are presented as means ± standard deviations (SDs), or medians. For comparisons, Student’s t-test and the Mann–Whitney test were applied to continuous variables, while the chi-squared test was applied to categorical variables. The results of THz absorption in urine with different pH values were analyzed by one-way ANOVA and postoperative examination. The statistical differences between each group were determined by the least significant difference (LSD) test. p values <0.05 were considered significant.
## Normal urine and proteinuria identification based on THz spectra
All non-smooth curves in the article are Fabry-Perot oscillations produced by thin sample pools and have no effect on the comparison of absorption coefficients between samples. Figure 2 shows the absorption coefficient of urine and pure water in the range of 0.5–1.2 THz. No significant absorption peaks were observed in the THz absorption spectra for the samples due to the disturbance and concealment of the liquid water (Xu et al., 2006). The absorption coefficient of water was higher than that of urine (Figure 2A). At 1.0 THz, the average absorption coefficient of water was 239 cm-1, while that of urine was 214 cm-1, with a difference of 25 cm-1. The largest absorption coefficient difference between the urine samples was 14 cm−1 at 1.0 THz. We can observe that the largest absorption coefficient between the urine samples was comparable to the absorption coefficient difference between water and urine, which strongly indicates that THz absorption coefficient is sensitive to some contents in urine. Since the proteinuria samples were normal in other parameters, we therefore inferred protein in urine sample caused the absorption coefficient difference. Figure 2B shows the relationship between absorption coefficient and protein concentration in urinary samples. Proteinuria sample generally have a greater absorption coefficient than non-protein urine sample, moreover, the absorption coefficient increases with increasing protein concentration.
**FIGURE 2:** *The absorption coefficient of water and urine samples (A) Absorption coefficients for pure water (black line) and urine (red line) (B) THz absorption coefficients for different concentrations of proteinuria (different red lines) and non-proteinuria (grey lines). Error bars indicate the standard deviation.*
## The effect of pH on urine
It is well known that alkaline urine (pH > 7.5) can cause false-positive results in the detection of urinary protein by dipstick method (Penders et al., 2002). To confirm whether the detection of protein by THz-TDS method is affected by pH, we used proteinuria samples with different pH value for detection. Here, HCl or NaOH was used to adjust the pH of urine samples to 6, 7, 8, and 9, respectively. The absorption coefficient of proteinuria (different protein concentrations) did not change significantly by different pH value at 1.0 THz ($p \leq 0.05$, Figure 3A). A linear increase in the absorption coefficient with increasing protein concentration was observed in the absorption coefficient profiles of urine samples with different concentration of proteins at the same pH (Figure 3B). The colorimetric profile of the urine analysis test strips was shown in Figure 3C. The protein concentration of normal urine without pH adjustment was negative with a pH around 6.5 (strip 1, Figure 3D). When this normal urine was adjusted to pH around 8.0, the protein concentration showed + ∼ ++ (strip 2, Figure 3D). The above results demonstrated that the test strip method gave a false positive for protein at high pH, while the THz method was independent of pH with no significant change in absorption coefficient. In other words, the THz-TDS method seemed to be more suitable than the strip method for protein detection in alkaline urine samples.
**FIGURE 3:** *Effect of pH on the absorption of THz in different urine (A) Variation of urine absorption coefficient with pH for four different protein concentrations at 1.0 THz (B) Absorption coefficients and protein concentrations of urine with different protein concentrations at the same pH (C) Colorimetric mapping of urine analysis test strips (D) Non-proteinuria (strip 1) appears false positive after raising pH (strip 2). All data are shown as the mean ± standard deviation.*
## The effect of different concentrations and types of protein on urinary absorption
The absorption spectra of β2-microglobulin and albumin with different concentrations at 1.0 THz are shown in Figure 4. The absorption coefficient of albumin was progressively greater than that of β2-microglobulin with increasing concentration, at concentrations higher than 0.5 g/L ($p \leq 0.05$). Although the error bars overlap at low concentration (<0.5 g/L), the curves generally show that higher molecular weight protein (albumin) has higher absorption capacity. Albumin is a single polypeptide chain consisting of 585 amino acids, and adopts a heart-shaped 3D structure with three homologous domains I-III; each domain contains two subdomains (He and Carter, 1992). The β2-microglobulin is a single-chain polypeptide composed of 99 amino acids, containing a pair of disulfide bonds within the molecule and no sugars. Comparing these two proteins, albumin has a more complex conformation, which directly affects the dielectric response in the THz range and enhances THz absorption.
**FIGURE 4:** *Effect of different types and concentrations of protein on urinary THz absorption. At 1.0 THz, the absorption spectra of albumin and β2 microglobulin at different concentrations. Δα = α(c) - α(urine), where α(urine) is absorption coefficient of urine before protein addition.*
## Discussion
In our study, the absorption coefficient increased with increasing protein concentration in the range of 0.5 THz to 1.2 THz, which can be explained exactly by the following theory. We consider the collected urine samples to have the same concentration of all components except for the protein concentration, which can be approximated by defining the samples as different concentrations of protein solutions. When aqueous solutions of various substances were studied, which could not be simply described as a two-component system; a third component was needed. Logically, the component was assumed as a hydrated shell (Penkov et al., 2018). For solvated protein, the total absorption could be decomposed into three components, the volume weighted average of the solute, the solvation water, and the bulk water, defined as follows (Ebbinghaus et al., 2007) α=αproteinVprotein/V+(αshellVshell)/V+αprotein1−Vprotein+Vshell/V [2] When protein was added, the solute molecules had a lower frequency mode than the solvent, and the protein solution had a linear decrease in terahertz absorption compared to pure water. In addition, the concentrations of protein in the urine samples collected in this study were all below 0.5 mM, the absorption of solute water molecules in the broader (>10 Å) solute shell layer was enhanced, resulting in a linear increase in the absorption coefficient of the protein solution with concentration (Ebbinghaus et al., 2008).
When we discuss the effect of pH on the terahertz absorption of protein urine, we focused on the analysis of albumin, the main component of urinary protein. Global perturbations of the protein hydration shell caused by pH and local perturbations caused by charge dependent mutations at surface sites can alter the hydration dynamics, producing significant changes in the terahertz absorption spectrum of the protein solution (Ebbinghaus et al., 2008; Born and Havenith, 2009). During the transition from weakly acidic (pH 6.0) to alkaline (pH 9.0), albumin undergoes transitions, including normal type (N) and alkaline type (B) (Qin et al., 2016). The N-B transition undergoes a very subtle change with almost no loss of secondary structure (Leonard et al., 1963), which implies that the protein retains most of its heterogeneity and is chemically identical in the B and N conformations (Babcock and Brancaleon, 2013). Nevertheless, the protein still undergoes de-folding, extending from a tiny spherical structure to a loose chain with no specific spatial structure. This process alters the absorption rate and thickness of the extended hydration shell, which lead to a decrease in absorption (Heyden and Havenith, 2010). However, unfolded proteins have a high density of vibrational modes between 0 and 2.0 THz, leading to an increase in THz absorption (Castro-Camus and Johnston, 2008). The two effects cancel each other out so that pH has no significant effect on the absorption coefficient of the protein solution.
We demonstrated that there were differences in THz absorption of two different types of proteins, which could help distinguish between tubular proteinuria and glomerular proteinuria, the most common type of proteinuria in the clinic. Proteins with a molecular weight less than 20.0 kDa pass the glomerular capillary wall easily (Larson, 1994). Conversely, albumin, with a molecular weight of 65.0 kDa and negative charge, is restricted under normal conditions. The low molecular weight proteins are largely reabsorbed at the proximal tubule, only few amounts are excreted. Thus, tubular proteinuria is dominated by low molecular weight proteins, such as β2-microglobulin (11.7 kDa), and the total amount of the tubular proteins is very small in urine. β2-microglobulin tends to accumulate in the blood and urine of chronic renal failure patients (Winchester et al., 2003). On the other hand, glomerular proteinuria has a large amount of protein and its composition is dominated by albumin. Albuminuria is generally regarded as an excellent marker for assessing early renal damage in diabetes and hypertension (Nah et al., 2017). Distinguishing these two proteins is instructive in diagnosing of proteinuria.
We have initially demonstrated the feasibility of THz in the detection of urinary protein. However, future studies could benefit from the application of data processing algorithms based on the Principal Component Analysis or Karhunen–Loève Transform (KLT) to classify different concentrations of proteinuria (Zaharov et al., 2014; Shao et al., 2022).
## Conclusion
The screening methods of urine samples should clearly separate samples without any indication for renal or genitourinary tract disorders from those samples which need further examination. In our study, a significantly different in absorption coefficient was found between proteinuria and non-proteinuria by THz-TDS method. Further analysis found that urinary protein and solvent water played a key role in this difference. At the same time, we investigated the effect of pH value (6.0–9.0) on urine sample and found that the absorption coefficient of urine sample did not change under different pH value by THz-TDS method. Finally, the absorption spectra of β2-microglobulin and albumin were studied, due to differences in conformation and molecular weight, these two proteins showed significant differences characteristics in absorption coefficient. Our study confirms that THz-TDS method can be used for proteinuria detection with ability to distinguish different proteinuria (renal tubular proteinuria and glomerular proteinuria). Moreover, THz spectroscopy has the potential to overcome the limitations of dipstick method by accurate detection of proteinuria even under alkaline condition. THz-TDS is a label-free method and shows great potential in the field of proteinuria detection. Further studies will establish a complete THz-TDS based system for the quantitative analysis of proteinuria.
## 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 Ethics Committee of the First Affiliated Hospital of Army Medical University (No. KY2020227). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
CY supervised the project and revised the manuscript. ZX participated in the experimental work and writing of the manuscript. PM and PP contributed to scientific discussion and revised the manuscript. SY and ZZ contributed to some of the experimental work and data analysis.
## 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. Babcock J. J., Brancaleon L.. **Bovine serum albumin oligomers in the E- and B-forms at low protein concentration and ionic strength**. *Int. J. Biol. Macromol.* (2013) **53** 42-53. DOI: 10.1016/j.ijbiomac.2012.10.030
2. Born B., Havenith M.. **Terahertz dance of proteins and sugars with water**. *J. Infrared, Millim. Terahertz Waves* (2009) **30** 1245-1254. DOI: 10.1007/s10762-009-9514-6
3. Bye J. W., Meliga S., Ferachou D., Cinque G., Zeitler J. A., Falconer R. J.. **Analysis of the hydration water around bovine serum albumin using terahertz coherent synchrotron radiation**. *J. Phys. Chem. A* (2014) **118** 83-88. DOI: 10.1021/jp407410g
4. Carroll M. F., Temte J. L.. **Proteinuria in adults: A diagnostic approach**. *Am. Fam. Physician* (2000) **62** 1333-1340. PMID: 11011862
5. Castro-Camus E., Johnston M. B.. **Conformational changes of photoactive yellow protein monitored by terahertz spectroscopy**. *Chem. Phys. Lett.* (2008) **455** 289-292. DOI: 10.1016/j.cplett.2008.02.084
6. Chen H., Chen X., Ma S., Wu X., Yang W., Zhang W.. **Quantify glucose level in freshly diabetic's blood by terahertz time-domain spectroscopy**. *J. Infrared, Millim. Terahertz Waves* (2018) **39** 399-408. DOI: 10.1007/s10762-017-0462-2
7. Clarkson M. R., Magee C. N., Brenner B. M.. **Chapter 2 - laboratory assessment of kidney disease**. *Pocket companion to brenner and rector's the kidney* (2011) 21-41
8. Ebbinghaus S., Kim S. J., Heyden M., Yu X., Gruebele M., Leitner D. M.. **Protein sequence- and pH-dependent hydration probed by terahertz spectroscopy**. *J. Am. Chem. Soc.* (2008) **130** 2374-2375. DOI: 10.1021/ja0746520
9. Ebbinghaus S., Kim S. J., Heyden M., Yu X., Heugen U., Gruebele M.. **An extended dynamical hydration shell around proteins**. *Proc. Natl. Acad. Sci. U. S. A.* (2007) **104** 20749-20752. DOI: 10.1073/pnas.0709207104
10. Esser A., Forbert H., Sebastiani F., Schwaab G., Havenith M., Marx D.. **Hydrophilic solvation dominates the terahertz fingerprint of amino acids in water**. *J. Phys. Chem. B* (2018) **122** 1453-1459. DOI: 10.1021/acs.jpcb.7b08563
11. He X. M., Carter D. C.. **Atomic structure and chemistry of human serum albumin**. *Nature* (1992) **358** 209-215. DOI: 10.1038/358209a0
12. Heyden M., Havenith M.. **Combining THz spectroscopy and MD simulations to study protein-hydration coupling**. *Methods* (2010) **52** 74-83. DOI: 10.1016/j.ymeth.2010.05.007
13. Klokkou N. T., Rowe D. J., Bowden B. M., Sessions N. P., West J. J., Wilkinson J. S.. **Structured surface wetting of a PTFE flow-cell for terahertz spectroscopy of proteins**. *Sensors Actuators B Chem.* (2022) **352** 131003. DOI: 10.1016/j.snb.2021.131003
14. Kumar V., Gill K. D.. *To perform qualitative tests for urinary proteins. Basic concepts in clinical biochemistry: A practical guide* (2018) 33-37
15. Lamb E. J., MacKenzie F., Stevens P. E.. **How should proteinuria be detected and measured?**. *Ann. Clin. Biochem.* (2009) **46** 205-217. DOI: 10.1258/acb.2009.009007
16. Larson T. S.. **Evaluation of proteinuria**. *Mayo Clin. Proc.* (1994) **69** 1154-1158. DOI: 10.1016/s0025-6196(12)65767-x
17. Leonard W. J., Vijai K. K., Foster J. F.. **A structural transformation in bovine and human plasma albumins in alkaline solution as revealed by rotatory dispersion studies**. *J. Biol. Chem.* (1963) **238** 1984-1988. DOI: 10.1016/s0021-9258(18)67930-x
18. Mancini T., Mosetti R., Marcelli A., Petrarca M., Lupi S., D’Arco A.. **Terahertz spectroscopic analysis in protein dynamics: Current status**. *Radiation* (2022) **2** 100-123. DOI: 10.3390/radiation2010008
19. Nah E-H., Cho S., Kim S., Cho H-I.. **Comparison of urine albumin-to-creatinine ratio (ACR) between ACR strip test and quantitative test in prediabetes and diabetes**. *Ann. Laboratory Med.* (2017) **37** 28-33. DOI: 10.3343/alm.2017.37.1.28
20. Niehues G., Heyden M., Schmidt D. A., Havenith M.. **Exploring hydrophobicity by THz absorption spectroscopy of solvated amino acids**. *Faraday Discuss.* (2011) **150** 193-207. DOI: 10.1039/c0fd00007h
21. Penders J., Fiers T., Delanghe J. R.. **Quantitative evaluation of urinalysis test strips**. *Clin. Chem.* (2002) **48** 2236-2241. DOI: 10.1093/clinchem/48.12.2236
22. Penkov N., Yashin V., Fesenko E., Manokhin A., Fesenko E.. **A study of the effect of a protein on the structure of water in solution using terahertz time-domain spectroscopy**. *Appl. Spectrosc.* (2018) **72** 257-267. DOI: 10.1177/0003702817735551
23. Qin Y., Wang L., Zhong D.. **Dynamics and mechanism of ultrafast water-protein interactions**. *Proc. Natl. Acad. Sci. U. S. A.* (2016) **113** 8424-8429. DOI: 10.1073/pnas.1602916113
24. Rahayu D., Rustiana T.. **Laboratory trial of protein determination in urine using different pH values of acetic acid and acetate buffer method**. *Indonesian J. Med. Laboratory Sci. Technol.* (2020) **2** 34-41. DOI: 10.33086/ijmlst.v2i1.1459
25. Reid C. B., Reese G., Gibson A. P., Wallace V. P.. **Terahertz time-domain spectroscopy of human blood**. *IEEE J. Biomed. Health Inf.* (2013) **17** 774-778. DOI: 10.1109/jbhi.2013.2255306
26. Ridley J. W.. *Procedures for complete urinalysis/confirmation testing. Fundamentals of the study of urine and body fluids* (2018) 203-249
27. Roberts J. R.. **Urine dipstick testing: Everything you need to know**. *Emerg. Med. News* (2007) **29** 24-27. DOI: 10.1097/01.eem.0000279130.93159.d9
28. Shao D., Miao S., Fan Q., Wang X., Liu Z., Ding E.. **Classification method of coal and gangue using terahertz time-domain spectroscopy, cluster analysis and principal component analysis**. *J. Appl. Spectrosc.* (2022) **89** 719-725. DOI: 10.1007/s10812-022-01416-3
29. Sirkeli V. P., Yilmazoglu O., Preu S., Küppers F., Hartnagel H. L.. **Proposal for a monolithic broadband terahertz quantum cascade laser array tailored to detection of explosive materials**. *Sens. Lett.* (2018) **16** 1-7. DOI: 10.1166/sl.2018.3919
30. Sun L., Zhao L., Peng R-Y.. **Research progress in the effects of terahertz waves on biomacromolecules**. *Mil. Med. Res.* (2021) **8** 28-8. DOI: 10.1186/s40779-021-00321-8
31. Tampe D., Korsten P., Ströbel P., Hakroush S., Tampe B.. **Proteinuria indicates decreased normal glomeruli in ANCA-associated glomerulonephritis independent of systemic disease activity**. *J. Clin. Med.* (2021) **10** 1538. DOI: 10.3390/jcm10071538
32. Torii T., Chiba H., Tanabe T., Oyama Y.. **Measurements of glucose concentration in aqueous solutions using reflected THz radiation for applications to a novel sub-THz radiation non-invasive blood sugar measurement method**. *Digit. Health* (2017) **3** 205520761772953. DOI: 10.1177/2055207617729534
33. Winchester J. F., Salsberg J. A., Levin N. W.. **Beta-2 microglobulin in ESRD: An in-depth review**. *Adv. Ren. Replacement Ther.* (2003) **10** 279-309. DOI: 10.1053/j.arrt.2003.11.003
34. Wu K., Qi C., Zhu Z., Wang C., Song B., Chang C.. **Terahertz wave accelerates DNA unwinding: A molecular dynamics simulation study**. *J. Phys. Chem. Lett.* (2020) **11** 7002-7008. DOI: 10.1021/acs.jpclett.0c01850
35. Xie L., Yao Y., Ying Y.. **The application of terahertz spectroscopy to protein detection: A review**. *Appl. Spectrosc. Rev.* (2014) **49** 448-461. DOI: 10.1080/05704928.2013.847845
36. Xu J., Plaxco K. W., Allen S. J.. **Probing the collective vibrational dynamics of a protein in liquid water by terahertz absorption spectroscopy**. *Protein Sci.* (2006) **15** 1175-1181. DOI: 10.1110/ps.062073506
37. Yan S., Wei D., Tang M., Shi C., Zhang M., Yang Z.. **Determination of critical micelle concentrations of surfactants by terahertz time-domain spectroscopy**. *IEEE Trans. Terahertz Sci. Technol.* (2016) **6** 532-540. DOI: 10.1109/tthz.2016.2575450
38. Zaharov V., Farahi R. H., Snyder P. J., Davison B. H., Passian A.. **Karhunen–Loève treatment to remove noise and facilitate data analysis in sensing, spectroscopy and other applications**. *Analyst* (2014) **139** 5927-5935. DOI: 10.1039/c4an01300j
39. Zang Z., Yan S., Han X., Wei D., Cui H-L., Du C.. **Temperature-and pH-dependent protein conformational changes investigated by terahertz dielectric spectroscopy**. *Infrared Phys. Technol.* (2019) **98** 260-265. DOI: 10.1016/j.infrared.2019.03.021
40. Zheng Z-P., Fan W-H., Li H., Tang J.. **Terahertz spectral investigation of anhydrous and monohydrated glucose using terahertz spectroscopy and solid-state theory**. *J. Mol. Spectrosc.* (2014) **296** 9-13. DOI: 10.1016/j.jms.2013.12.002
|
---
title: 'Probiotics intervention in preventing conversion of impaired glucose tolerance
to diabetes: The PPDP follow-on study'
authors:
- Qun Yan
- Weiting Hu
- Yan Tian
- Xu Li
- Yuan Yu
- Xing Li
- Bo Feng
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9982119
doi: 10.3389/fendo.2023.1113611
license: CC BY 4.0
---
# Probiotics intervention in preventing conversion of impaired glucose tolerance to diabetes: The PPDP follow-on study
## Abstract
### Objectives
The purpose of this study was to assess the incidence of type 2 diabetes mellitus (T2DM) after 6 years in patients with IGT who received early probiotic intervention in the Probiotics Prevention Diabetes Program (PPDP) trial.
### Methods
77 patients with IGT in the PPDP trial were randomized to either probiotic or placebo. After the completion of the trial, 39 non-T2DM patients were invited to follow up glucose metabolism after the next 4 years. The incidence of T2DM in each group was assessed using Kaplan-Meier analysis. The 16S rDNA sequencing technology was used to analyze gut microbiota’s structural composition and abundance changes between the groups.
### Results
The cumulative incidence of T2DM was $59.1\%$ with probiotic treatment versus $54.5\%$ with placebo within 6 years, there was no significant difference in the risk of developing T2DM between the two groups ($$P \leq 0.674$$).
### Conclusions
Supplemental probiotic therapy does not reduce the risk of IGT conversion to T2DM.
### Clinical Trial Registration
https://www.chictr.org.cn/showproj.aspx?proj=5543, identifier ChiCTR-TRC-13004024.
## Introduction
Compared with normal glucose tolerance (NGT), people with prediabetes, especially impaired glucose tolerance (IGT), have a higher risk of developing type 2 diabetes mellitus (T2DM). Early intervention can significantly reduce the probability of developing T2DM in the IGT population (1–3). The dysbiosis in the gut microbiota has recently been recognized as a critical environmental factor in individuals with prediabetes or diabetes mellitus (DM) [4, 5]. Several studies (6–8) showed that probiotic and synbiotic intake affects the glycemic profile in patients with prediabetes and T2DM. Most recently, the Probiotics Prevention Diabetes Program (PPDP) Study demonstrated that probiotic supplementation during two years did not improve fasting plasma glucose(FPG) levels and did not reduce the risk of conversion of IGT to T2DM [9].
After the PPDP trial was completed, participants without T2DM were invited to follow up for 4 years. The objective of the PPDP Follow-On study was to observe the effect of early probiotic intervention on the conversion of T2DM after 6 years.
## PPDP study
The design and primary results of the PPDP study have been reported previously [9, 10]. Briefly, the PPDP Study included 77 patients diagnosed with IGT in the outpatient department of Shanghai East Hospital of Tongji University from September 2014 to September 2016. IGT and T2DM were diagnosed according to the 1999 WHO Criteria. IGT was diagnosed when FPG <7.0 mmol/L, and oral glucose tolerance test (OGTT): 2-h post glucose load ≥ 7.8 and <11.0 mmol/L. T2DM was diagnosed when FPG ≥7.0 mmol/L and or OGTT: 2-h post glucose load≥11.1 mmol/L, or self-reported diabetes history and being treated with hypoglycemic agents. All participants were randomized, double-blind to receive probiotics (including Bifidobacterium, *Lactobacillus acidophilus* and Enterococcus faecalis) or matched placebo. Probiotics were produced by Shanghai Sine Pharmaceutical Laboratories Co, Ltd. For both groups, the doses were 840 mg daily, 210 mg per one pill, two pills per time, and two times daily. Both groups were followed for two years with an OGTT every 3 months in the first year and every 4 months in the second year to assess the patient’s glucose metabolism. At the end of the PPDP study, there were 20 patients in the Probiotics group and 13 in the Placebo group who developed T2DM.
Feces of the two groups before and after intervention were collected. The 16S rDNA sequencing technology was used to analyze intestinal microbiota’s structural composition and abundance changes. The primary outcome was the cumulative prevalence of T2DM in the two groups. The secondary endpoints were the possible changes in the proportion of microbiota. The study was registered in the Chinese clinical trial registry (ChiCTR-TRC-13004024).
## PPDP follow-on study
After the completion of the initial PPDP study, patients who with undiagnosed T2DM continue to be invited to participate in the PPDP Follow-On study without probiotics intervention. A total of 39 non-T2DM patients agreed to follow up glucose metabolism for next 4 years. Patients were asked to monitor fasting and postprandial blood glucose by themselves. At the 4th year, OGTT were assessed at the outpatient department of Shanghai East Hospital. Finally, 36 patients finished the next 4-year follow-up, 2 patients withdrew due to loss of contact, and 1 patient died due to a blood tumor, with a dropout rate of $4.2\%$. The detailed patient flow of the original trial (PPDP) and follow-on study is summarized in Figure 1.
**Figure 1:** *Flow diagram for PPDP follow-on study.*
The primary outcome was the cumulative incidence of T2DM in the two groups during the 6 years. The PPDP study and the PPDP Follow-On study were reviewed and approved by the hospital’s ethics committee and all patients signed informed consent.
## 16S rRNA gene sequencing and analysis
Fresh fecal samples were collected and bacteria’s 16S rRNA gene sequence was detected using paired-end configuration on an Illumina MiSeq system (Illumina, San Diego, USA). Briefly, microbial DNA was extracted and DNA quality was examined by agarose gel electrophoresis. The V3-V4 regions of the bacteria’s 16S rRNA gene were amplified by PCR. The sequencing was performed using paired-end configuration on an Illumina MiSeq system (Illumina, San Diego, USA). Raw fastq files were demultiplexed, and then data was filtered to ensure quality. The taxonomy of each 16S rRNA gene sequence was analyzed by RDP Classifier (http://rdp.cme.msu.edu/) against the silva (SSU115)16S rRNA database. The detail of sequencing and analysis showed in Supplementary Material 1.
## Statistical analyses
Statistical analyses were performed by SPSS version 23.0 (IBM Corp, Armonk, NY, USA) and GraphPad Prism version 8.0 (San Diego, California, USA). Continuous data were described as means ± standard deviation, and inter-group comparison was performed with a t-test or analysis of variance. All continuous data were abnormally distributed. Categorical data were described as n (%), and inter-group comparisons were analyzed by χ2 -test. Mann–Whitney test was used to compare data that were not normally distributed between the groups. The cumulative incidence of T2DM = (the number of cases developing T2DM after 6 years of follow-up/the number of cases starting follow-up) × $100\%$. The difference in the incidence of T2DM between the probiotic group and the placebo group over time was analyzed using the Kaplan-Meier survival curve. COX regression analysis was used to analyze the influencing factors of T2DM. The risk was described as Hazard ratio (HR) and $95\%$CI. $P \leq 0.05$ was statistically significant.
## Baseline characteristics
The baseline characteristics of the Probiotics group and Placebo group in the PPDP study have been presented in the previous article [9]. Specifically, there were no significant differences in sex composition, age, body mass index (BMI), blood pressure, heart rate, liver function, blood lipid profile, FPG, post-glucose load plasma glucose, glycated hemoglobin A1c(HbA1c), fasting serum insulin (FINS) and the homeostasis model assessment of insulin resistance (HOMA-IR) between the two groups.
At the end of the PPDP study, there remained 39 patients with undiagnosed T2DM (21 patients in the Probiotics group and 18 in the Placebo group). The characteristics of undiagnosed T2DM patients at the end of 2-year follow up between the probiotics and placebo groups were shown in Table 1.
**Table 1**
| Unnamed: 0 | Probiotics group (n=21) | Placebo group (n=18) | P value |
| --- | --- | --- | --- |
| | Probiotics group (n=21) | Placebo group (n=18) | P value |
| Age (year) | 62.3 ± 10.2 | 52.1 ± 14.7 | 0.015* |
| Male n(%) | 7 (33.3) | 9 (50.0) | 0.342 |
| BMI (kg/m2) | 25.2 ± 2.3 | 24.3 ± 2.3 | 0.483 |
| WC (cm) | 88.2 ± 7.5 | 87.8 ± 11.9 | 0.903 |
| SBP (mmHg) | 125.3 ± 12.3 | 119.3 ± 18.3 | 0.270 |
| DBP (mmHg) | 79.1 ± 9.4 | 76.1 ± 8.7 | 0.348 |
| ALT (IU/L) | 24.8 ± 15.8 | 23.4 ± 23.3 | 0.851 |
| SCr (umol/L) | 64.8 ± 9.4 | 62.1 ± 12.0 | 0.690 |
| TG (mmol/L) | 1.2 ± 0.59 | 1.69 ± 1.09 | 0.181 |
| TC (mmol/L) | 4.51 ± 0.99 | 4.99 ± 1.01 | 0.157 |
| HDL-C(mmol/L) | 1.51 ± 0.30 | 1.43 ± 0.48 | 0.555 |
| LDL-C (mmol/L) | 2.91 ± 0.91 | 3.22 ± 0.97 | 0.212 |
| FPG (mmol/L) | 5.25 ± 0.47 | 5.23 ± 9.54 | 0.933 |
| 30minPG (mmol/L) | 9.72 ± 1.62 | 9.40 ± 1.79 | 0.604 |
| 2hPG (mmol/L) | 7.76 ± 1.25 | 7.23 ± 0.88 | 0.154 |
| FINS (μU/L) | 11.7 ± 7.9 | 11.1 ± 10.8 | 0.850 |
| 30minINS (μU/L) | 77.8 ± 53.8 | 75.5 ± 61.1 | 0.906 |
| 2hINS (μU/L) | 99.2 ± 64.4 | 64.4 ± 47.4 | 0.075 |
| HOMA-IR | 2.79 ± 2.07 | 2.59 ± 2.45 | 0.802 |
| Incidence of T2DM 4 years later (n) | 6 | 5 | – |
## Comparison of the incidence of T2DM
In the next 4 years, there were 6 patients in the Probiotics group and 5 patients in the Placebo group who developed T2DM. Thus, the cumulative incidence of T2DM was $59.1\%$ in the probiotic group and $54.5\%$ in the placebo group within 6 years. As shown in Figure 2, there was no significant difference in the risk of developing T2DM between the two groups within 6 years ($$P \leq 0.674$$).
**Figure 2:** *Kaplan-Meier analysis of the cumulative incidence of T2DM within 6 years between the probiotics and placebo group.*
At the end of the 6-year follow-up, patients were grouped according to whether T2DM occurred. The age of the T2DM group was significantly older than the non-T2DM group (57.2 vs 53.7 years, $$P \leq 0.004$$). There was no significant difference in other clinical data between the two groups (all $P \leq 0.05$).
## COX regression analysis of risk factors for T2DM
COX regression model was used to analyze the risk factors affecting the development of T2DM. Probiotic intervention or not, age, gender, BMI, waist circumference, blood pressure, liver function, blood lipid, blood glucose, serum insulin and HbA1c were used as covariates, and results showed that 30-minute post-glucose load insulin level was a factor affecting the conversion of IGT to T2DM (HR=0.954, $95\%$CI 0.915-0.994, $$P \leq 0.026$$) (Figure 3).
**Figure 3:** *COX regression analysis of risk factors for T2DM after 6 years.*
## Gut microbiota analysis
According to the informed consent and research protocol, fecal samples were collected at baseline (day 0) and the end of the 2-year follow-up visit. The 16S rDNA sequencing technology was used to analyze gut microbiota’s structural composition and abundance changes. A total of 32 stool samples in the Probiotic group and 22 in the Placebo group were collected. In this study, the differences in operational taxonomic units (OTUs) abundance among the probiotic group and the placebo group were compared. The Venn diagram showed that 435 of the total 972 genera were shared among the 4 groups (Figure 4A). To display microbiome space between samples, principal coordinates analysis (PCoA) was performed. The results showed that the microbiota was similar between the probiotic and placebo interventions (Figure 4B), and the probiotic intervention might not have caused the recombination of the microbial community composition. Microbial community variation was also analyzed. At the genus level, Blautia, Subdoligranulum, Eubacterium hallii, Bifidobacterium, and Romboutsia accounted for the majority in each group (Figure 4C). However, there were no statistically significant changes in the microbiota composition after probiotics or placebos intervention in any of the groups compared with those before intervention. To assess the differences in bacterial diversity among groups, sequences were aligned for alpha-diversity. No significant difference in the Shannon index between the probiotic group and the placebo group was observed. ( Figure 4D). We also compared the difference in gut microbiota’s structural composition and abundance changes between the baseline and after intervention among T2DM and non-T2DM groups. The Venn diagram of bacteria showed that 426 of the total 972 genera were shared among the 4 groups (Figure 5A). The results of PCoA showed that the microbiota was similar between the T2DM group and the non-T2DM group both at baseline and after intervention (Figure 5B). Microbial community analysis demonstrated that Blautia, Subdoligranulum, Eubacterium hallii, Bifidobacterium, and Romboutsia accounted for the majority in each group (Figure 5C). At baseline and after the intervention, no significant difference in the Shannon index between the T2DM group and the non-T2DM group was observed. ( Figure 5D).
**Figure 4:** *Composition and diversity of gut microbiota before and after two years probiotics or placebo intervention. (A) The Venn diagram shows the common or endemic species between groups in the level of OUT; (B) Weighted UniFrac PCoA; (C) Compositional change at the genus level; (D) α diversity analysis of gut microbiota.* **Figure 5:** *Composition and diversity of gut microbiota in T2DM patients and non-T2DM patients before and after two years of probiotics intervention. (A) The Venn diagram shows the common or endemic species between groups in the level of OUT; (B) Weighted UniFrac PCoA; (C) Compositional change at the genus level; (D) α diversity analysis of gut microbiota.*
The mean proportion of subdoligranulum and monoglobus in the T2DM group was significantly lower than that of the non-T2DM group both at bas `eline and after the intervention. The proportion of collinsella was lower in the T2DM group (Figure 6A). Further analysis of the specific species microbiota showed that there were no differences among groups in the mean proportion of the metabolic-related microbiota, as well as in produces short-chain fatty acids-related microbiota and gut probiotics (Figures 6B-D) ylogenetic Investigation of Communities using Reconstruction of Unobserved States (Picrust2) software. The results showed that these metabolism-related pathways consisted of carbohydrate metabolism, amino acid metabolism, transcription, replication, recombination and repair, and other metabolic pathways in the non-T2DM and T2DM group (Figure 7).
**Figure 6:** *Analysis of significantly altered gut microbiota and specific species microbiot in T2DM patients and non-T2DM patients after two years of probiotics intervention. (A) Analysis of significantly altered gut microbiota; (B) Analysis of metabolism-related microbiota; (C) Analysis of producing short-chain fatty acids-related microbiota; (D) Analysis of gut probiotics.* **Figure 7:** *Functional prediction and comparison of gut microbiota between T2DM and non-T2DM groups.*
## Discussion
IGT is closely associated with metabolic disease progression. According to the epidemiological data, about $70\%$ of IGT patients progress to DM within 5 years in China [11]. The rapidly growing trend means an urgent need to prevent DM actively. Early dietary modification can prevent the development of diabetes, but it is difficult for individuals to adhere to. The gut microflora plays a crucial role in regulating host metabolism. Changing the composition and/or metabolic activity of gut microflora may contribute to human health. Evidence from human and animal studies suggest that the gut microbiome is a common pathway mediating the therapeutic effects of bariatric surgery, dietary control, and hypoglycemic drug therapy (12–14). Therefore, remodeling the gut microflora may be a new direction for humans to prevent and treat diabetes.
The treatment of metabolic diseases with probiotics is a hot topic in intestinal microbiota research. However, there are fewer studies on probiotics for the prevention and treatment of IGT patients. We previously observed that probiotics supplementation for IGT patients for 2 years did not significantly reduce the risk of IGT conversion to T2DM in the PPDP study [9]. In the present study we intend to observe the impact of early and long-term probiotics supplementation on the conversion of diabetes in a longer time (6 years). This is the first long-term prospective study to analyze the efficacy of probiotic administration on glucose metabolism in IGT subjects. The supplementary probiotics in the PPDP study were provided by Bifico (Approval number: S10950032), an over-the-counter capsule consisting of live combined Bifidobacterium longum, *Lactobacillus acidophilus* and Enterococcus faecalis. It has been reported that *Bifidobacterium longum* supplementation can attenuate hyperglycemia, improve the antioxidant capacity of the liver, repair intestinal barrier injury, and reduce inflammation in diabetic mice [15]. Lactobacillus acidophilus was also reported that can alleviate T2DM by regulating hepatic glucose, lipid metabolism and gut microbiota in mice [16]. In addition, it was indicated that *Enterococcus faecalis* treatment could improve glucose homoeostasis, increased energy expenditure and reduced hepatic steatosis in the db/db mice fed with high fat [17]. However, in the present study, after 6 years’ follow up, the Probiotic group showed no significant superiority in preventing the conversion of IGT to T2DM as compared with the Placebo group. Similarly, no significant differences in the diversity and composition of the gut microbiota were observed between the two groups, nor were differences in microbiota observed between groups with or without T2DM. COX regression also showed that probiotics intervention was not affecting IGT conversion to T2DM. Only 30-minute-insulin after glucose loading was the factor affecting the conversion of IGT into T2DM, which indicated that the decrease in islet β -cell function was an important cause of T2DM.
Although the relationship between the gut microbial ecological imbalance and the development of obesity and diabetes is being extensively explored, the conclusions of various studies are different. The results of randomized controlled studies on pregnant women with gestational diabetes or obesity showed that probiotic intervention had no effect on glycemic control, but might improve lipid metabolism [18, 19].In another study of prediabetes adolescents, it was not observed that oral probiotics could improve FBG and HbA1c after 4 months [20]. Similarly, in a 24-week probiotic intervention study on adults with prediabetes, the goodness of glycosylated hemoglobin was not observed [21]. Our studies are consistent with the conclusion of these studies that probiotics have a limited therapeutic effect on metabolic diseases. However, some studies confirm the beneficial role of gut microbiota in glycemic control and T2DM. Tonucci et al. found that Probiotic consumption improved glycemic control in T2DM subjects [22]. The application of a novel probiotic formulation to T2DM showed that the intervention was safe and well tolerated [23]. Different probiotic strains, their combinations or the time and duration of intervention may play different roles in the efficacy of the probiotic intervention on glucose control. The limited sample size and subject-to-subject variability suggest that future studies are needed to confirm and extend these observations.
The gut microbiota profile may be related to and responsive to a particular dietary pattern [24]. Therefore, supplementation with beneficial microorganisms such as probiotics and their metabolites may alter microbiota distribution and thus affect metabolic parameters [25]. However, in this study, gut microbiota analysis results showed no difference in the composition and diversity of the gut microbiota between the T2DM group and the IGT group after two years of probiotic intervention. This may be related to the small fecal sample size selected in this study and the large individual differences of samples within the same group, or it may be the result of functional variation of the strain, indicating that a more precise strategy is required for probiotic therapy. The analysis of specific microbiota showed that compared with the IGT group, the proportion of subdoligranulum, collinsella and monoglobus in the T2DM group decreased after two years of intervention. The occurrence of T2DM may be related to the changes in the composition of intestinal microbiota. Although there is a lack of consensus on which microbiota are significantly changed in T2DM, a common observation has been a decreased abundance of butyrate-producing bacteria with this condition [26]. Subdoligranulum and collinsella have been proven to produce butyric acid [27, 28], and a study has shown that the decrease of Monoglobus may be related to insulin resistance and systemic inflammation [29].
There are some limitations in our study. First, this was a small sample size study that enrolled a limited number of patients with IGT. More clinical and laboratory studies using large-size samples and long-term observation are needed to confirm the role of probiotics in developing IGT into DM. Second, the results of the study of Bifico used in this study as a probiotic supplement for Chinese patients are not representative of the effects of other strains on other people or races. Third, the study did not document lifestyle factors, such as diet and exercise, which might have influenced blood sugar outcomes. There is also no recorded family history of T2DM, which is a very strong risk factor for developing T2DM. However, the placebo control designed in this study could compensate for this effect to the greatest extent. To provide preliminary data that could drive more conclusive testing. Therefore, high-quality, large-scale, multicenter randomized controlled trials with longer follow-up are needed to compare safety and efficacy further.
## Conclusions
Nevertheless, the results of this study suggest that supplementation with active probiotics of Bifidobacterium, *Lactobacillus acidophilus* and *Enterococcus faecalis* is safe, although it does not reduce the risk of IGT conversion to DM. More clinical and laboratory studies using large samples and long-term observation are needed to explore the effects of different probiotic strains on IGT. This pilot study was designed to provide preliminary data to conduct more conclusive hypothesis testing.
## 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 BioProject [https://www.ncbi.nlm.nih.gov/bioproject/], PRJNA923108.
## Ethics statement
The studies involving human participants were reviewed and approved by the institutional review board of Shanghai East Hospital and was conducted in accordance with the Declaration of Helsinki. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
BF designed the study and oversaw the project implementation. QY conceived and carried out experiments. WH and YT participated in data analyses, interpretation and writing publications. XUL, YY and XIL participated in data collection, data analyses and interpretation and writing publications. All authors were involved in writing the paper and had final approval of the submitted and published versions.
## 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.2023.1113611/full#supplementary-material
## References
1. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *New Engl J Med* (2002) **346** 393-403. DOI: 10.1056/NEJMoa012512
2. Gerstein HC, Yusuf S, Bosch J, Pogue J, Sheridan P, Dinccag N. **Effect of rosiglitazone on the frequency of diabetes in patients with impaired glucose tolerance or impaired fasting glucose: A randomised controlled trial**. *Lancet (London England)* (2006) **368**. DOI: 10.1016/s0140-6736(06)69420-8
3. Hu R, Li Y, Lv Q, Wu T, Tong N. **Acarbose monotherapy and type 2 diabetes prevention in Eastern and Western prediabetes: An ethnicity-specific meta-analysis**. *Clin Ther* (2015) **37**. DOI: 10.1016/j.clinthera.2015.05.504
4. Yassour M, Lim MY, Yun HS, Tickle TL, Sung J, Song YM. **Sub-Clinical detection of gut microbial biomarkers of obesity and type 2 diabetes**. *Genome Med* (2016) **8** 17. DOI: 10.1186/s13073-016-0271-6
5. Qin J, Li Y, Cai Z, Li S, Zhu J, Zhang F. **A metagenome-wide association study of gut microbiota in type 2 diabetes**. *Nature* (2012) **490** 55-60. DOI: 10.1038/nature11450
6. Naseri K, Saadati S, Ashtary-Larky D, Asbaghi O, Ghaemi F, Pashayee-Khamene F. **Probiotics and synbiotics supplementation improve glycemic control parameters in subjects with prediabetes and type 2 diabetes mellitus: A GRADE-assessed systematic review, meta-analysis, and meta-regression of randomized clinical trials**. *Pharmacol Res* (2022) **184**. DOI: 10.1016/j.phrs.2022.106399
7. Cao DX, Wong EY, Vela MN, Le QT. **Effect of probiotic supplementation on glycemic outcomes in patients with abnormal glucose metabolism: A systematic review and meta-analysis of randomized controlled trials**. *Ann Nutr Metab* (2021) **77**. DOI: 10.1159/000518677
8. Zhang C, Jiang J, Wang C, Li S, Yu L, Tian F. **Meta-analysis of randomized controlled trials of the effects of probiotics on type 2 diabetes in adults**. *Clin Nutr (Edinburgh Scotland)* (2022) **41**. DOI: 10.1016/j.clnu.2021.11.037
9. Yan Q, Li X, Li PC, Feng B. **Probiotics for the prevention of type 2 diabetes mellitus in patients with impaired glucose tolerance:a double-blind randomized controlled trial**. *Shanghai Med J* (2021) **44**. DOI: 10.19842/j.cnki.issn.0253-9934.2021.10.006
10. Yan Q, Li X, Feng B. **The efficacy and safety of probiotics intervention in preventing conversion of impaired glucose tolerance to diabetes: Study protocol for a randomized, double-blinded, placebo controlled trial of the probiotics prevention diabetes programme (PPDP)**. *BMC Endocr Disord* (2015) **15** 74. DOI: 10.1186/s12902-015-0071-9
11. Tabák AG, Herder C, Rathmann W, Brunner EJ, Kivimäki M. **Prediabetes: A high-risk state for diabetes development**. *Lancet (London England)* (2012) **379**. DOI: 10.1016/s0140-6736(12)60283-9
12. Liu R, Hong J, Xu X, Feng Q, Zhang D, Gu Y. **Gut microbiome and serum metabolome alterations in obesity and after weight-loss intervention**. *Nat Med* (2017) **23**. DOI: 10.1038/nm.4358
13. Wu H, Esteve E, Tremaroli V, Khan MT, Caesar R, Mannerås-Holm L. **Metformin alters the gut microbiome of individuals with treatment-naive type 2 diabetes, contributing to the therapeutic effects of the drug**. *Nat Med* (2017) **23**. DOI: 10.1038/nm.4345
14. Sun L, Xie C, Wang G, Wu Y, Wu Q, Wang X. **Gut microbiota and intestinal FXR mediate the clinical benefits of metformin**. *Nat Med* (2018) **24**. DOI: 10.1038/s41591-018-0222-4
15. Hao J, Zhang Y, Wu T, Liu R, Sui W, Zhu J. **The antidiabetic effects of bifidobacterium longum subsp. longum BL21 through regulating gut microbiota structure in type 2 diabetic mice**. *Food Funct* (2022) **13**. DOI: 10.1039/d2fo01109c
16. Yan F, Li N, Shi J, Li H, Yue Y, Jiao W. **Lactobacillus acidophilus alleviates type 2 diabetes by regulating hepatic glucose, lipid metabolism and gut microbiota in mice**. *Food Funct* (2019) **10**. DOI: 10.1039/c9fo01062a
17. Quan LH, Zhang C, Dong M, Jiang J, Xu H, Yan C. **Myristoleic acid produced by enterococci reduces obesity through brown adipose tissue activation**. *Gut.* (2020) **69**. DOI: 10.1136/gutjnl-2019-319114
18. Lindsay KL, Brennan L, Kennelly MA, Maguire OC, Smith T, Curran S. **Impact of probiotics in women with gestational diabetes mellitus on metabolic health: A randomized controlled trial**. *Am J obstetrics gynecology* (2015) **212**. DOI: 10.1016/j.ajog.2015.02.008
19. Lindsay KL, Kennelly M, Culliton M, Smith T, Maguire OC, Shanahan F. **Probiotics in obese pregnancy do not reduce maternal fasting glucose: A double-blind, placebo-controlled, randomized trial (Probiotics in pregnancy study)**. *Am J Clin Nutr* (2014) **99**. DOI: 10.3945/ajcn.113.079723
20. Stefanaki C, Michos A, Mastorakos G, Mantzou A, Landis G, Zosi P. **Probiotics in adolescent prediabetes: A pilot RCT on glycemic control and intestinal bacteriome**. *J Clin Med* (2019) **8**. DOI: 10.3390/jcm8101743
21. Barthow C, Hood F, Crane J, Huthwaite M, Weatherall M, Parry-Strong A. **A randomised controlled trial of a probiotic and a prebiotic examining metabolic and mental health outcomes in adults with pre-diabetes**. *BMJ Open* (2022) **12**. DOI: 10.1136/bmjopen-2021-055214
22. Tonucci LB, Olbrich Dos Santos KM, Licursi de Oliveira L, Rocha Ribeiro SM, Duarte Martino HS. **Clinical application of probiotics in type 2 diabetes mellitus: A randomized, double-blind, placebo-controlled study**. *Clin Nutr (Edinburgh Scotland)* (2017) **36** 85-92. DOI: 10.1016/j.clnu.2015.11.011
23. Perraudeau F, McMurdie P, Bullard J, Cheng A, Cutcliffe C, Deo A. **Improvements to postprandial glucose control in subjects with type 2 diabetes: a multicenter, double blind, randomized placebo-controlled trial of a novel probiotic formulation**. *BMJ Open Diabetes Res Care* (2020) **8**. DOI: 10.1136/bmjdrc-2020-001319
24. David LA, Maurice CF, Carmody RN, Gootenberg DB, Button JE, Wolfe BE. **Diet rapidly and reproducibly alters the human gut microbiome**. *Nature.* (2014) **505**. DOI: 10.1038/nature12820
25. Asemi Z, Zare Z, Shakeri H, Sabihi SS, Esmaillzadeh A. **Effect of multispecies probiotic supplements on metabolic profiles, hs-CRP, and oxidative stress in patients with type 2 diabetes**. *Ann Nutr Metab* (2013) **63** 1-9. DOI: 10.1159/000349922
26. Brunkwall L, Orho-Melander M. **The gut microbiome as a target for prevention and treatment of hyperglycaemia in type 2 diabetes: From current human evidence to future possibilities**. *Diabetologia.* (2017) **60**. DOI: 10.1007/s00125-017-4278-3
27. Van Hul M, Le Roy T, Prifti E, Dao MC, Paquot A, Zucker JD. **From correlation to causality: The case of subdoligranulum**. *Gut Microbes* (2020) **12** 1-13. DOI: 10.1080/19490976.2020.1849998
28. Qin P, Zou Y, Dai Y, Luo G, Zhang X, Xiao L. **Characterization a novel butyric acid-producing bacterium collinsella aerofaciens subsp**. *Shenzhenensis Subsp Nov Microorganisms* (2019) **7**. DOI: 10.3390/microorganisms7030078
29. Dang JT, Mocanu V, Park H, Laffin M, Hotte N, Karmali S. **Roux-en-Y gastric bypass and sleeve gastrectomy induce substantial and persistent changes in microbial communities and metabolic pathways**. *Gut Microbes* (2022) **14**. DOI: 10.1080/19490976.2022.2050636
|
---
title: Differences in protein expression, at the basal state and at 2 h of insulin
infusion, in muscle biopsies from healthy Arab men with high or low insulin sensitivity
measured by hyperinsulinemic euglycemic clamp
authors:
- Ilham Bettahi
- Roopesh Krishnankutty
- Morana Jaganjac
- Noor Nabeel M. Suleiman
- Manjunath Ramanjaneya
- Jayakumar Jerobin
- Shaimaa Hassoun
- Meis Alkasem
- Ibrahem Abdelhakam
- Ahmad Iskandarani
- Tareq A. Samra
- Vidya Mohamed-Ali
- Abdul Badi Abou-Samra
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9982120
doi: 10.3389/fendo.2022.1024832
license: CC BY 4.0
---
# Differences in protein expression, at the basal state and at 2 h of insulin infusion, in muscle biopsies from healthy Arab men with high or low insulin sensitivity measured by hyperinsulinemic euglycemic clamp
## Abstract
### Background
Skeletal muscle is the main site for insulin-dependent glucose disposal. The hyperinsulinemic euglycemic clamp (HIEC) is the gold standard for the assessment of insulin sensitivity (IS). We have previously shown that insulin sensitivity, measured by HIEC, varied widely among a group of 60 young healthy men with normoglycemia. The aim of this study was to correlate the proteomic profile of skeletal muscles to insulin sensitivity.
### Methods
Muscle biopsies from 16 subjects having the highest (M ≥ 13; $$n = 8$$, HIS) and lowest (M ¾ 6, $$n = 8$$, LIS) IS were obtained at baseline and during insulin infusion after stabilization of the blood glucose level and glucose infusion rate at the end of the HIEC. The samples were processed using a quantitative proteomic analysis approach.
### Results
At baseline, 924 proteins were identified in the HIS and LIS groups. Among the 924 proteins detected in both groups, three were suppressed and three were increased significantly in the LIS subjects compared with the HIS subjects. Following insulin infusion, 835 proteins were detected in both groups. Among the 835 proteins, two showed differential responsiveness to insulin; ATP5F1 protein was decreased, and MYLK2 was higher in the LIS group compared with that in the HIS group. Our data suggest that alteration in mitochondrial proteins and an increased number of proteins involved in fast-twitch fiber correlate to insulin sensitivity in healthy young Arab men.
### Conclusions
These results suggest a change in a small number of differentially expressed proteins. A possible reason for this small change could be our study cohorts representing a homogeneous and healthy population. Additionally, we show differences in protein levels from skeletal muscle in low and high insulin sensitivity groups. Therefore, these differences may represent early events for the development of insulin resistance, pre-diabetes, and type 2 diabetes.
## Introduction
Type 2 diabetes (T2D) represents a significant international health challenge, with approximately 463 million people having diabetes in 2019—half of whom are undiagnosed [1]. Over the past few years, the prevalence of T2D has been dramatically increasing in the Middle East and North Africa region, including Qatar [2]. Insulin resistance in skeletal muscle is recognized as the earliest metabolic defect in T2D [3]. Insulin resistance is a complex heterogeneous phenomenon influenced by genetic and environmental factors [4]. Several abnormalities can predict insulin resistance, including impaired insulin activation of glycogen synthase, impairment of the proximal components of insulin signaling (5–7), and increased intramuscular triglyceride content [5]. Moreover, insulin-stimulated glucose oxidation and insulin inhibition of lipid oxidation are impaired in subjects with insulin resistance and T2D [6]. The inability to switch from lipid to carbohydrate has been described as “metabolic inflexibility” in insulin-resistant subjects [7]. Moreover, a reduction in the activity of oxidative enzymatic pathways and dysfunction of the mitochondria have been observed in skeletal muscle obtained from subjects with T2D and correlate with the severity of insulin-resistant glucose metabolism [8]. Previous longitudinal studies show that insulin resistance is familial and occurs many years before the development of glucose intolerance [9]. *Whether* genetic or acquired, the resistance of skeletal muscles to insulin may be associated with alteration in the expression of key proteins involved in glucose homeostasis.
Several studies have reported the skeletal muscle proteomic profile from skeletal muscle biopsies of humans and mice (10–14). Hojlund et al. showed that the abundance of certain proteins, such as heat shock proteins, which are altered in skeletal muscles, and key mitochondrial metabolic pathways, such as ATP synthase and creatine kinase B, are perturbed in patients with T2D [10]. Hwang et al. demonstrated a reduced abundance of several mitochondrial proteins in the insulin-resistant muscle compared with the healthy group [11]. Another study which looked at the mitochondria isolated from insulin-resistant skeletal muscle using one-dimensional gel electrophoreses and high-performance liquid chromatography/electrospray ionization–tandem mass spectrometry (HPLC/ESI–MS/MS) showed a lower abundance of proteins involved in branched-chain amino acid metabolism in T2D than in the lean control [12]. Previous studies have shown that proteomic markers of insulin resistance can be determined in T2D subjects. However, these studies have not elucidated if early changes in insulin sensitivity (IS) in healthy people are associated with different protein expression in muscles. Furthermore, protein responses to hyperinsulinemic euglycemic clamp (HIEC) in people with low versus high IS have not been shown previously.
We have recently reported that insulin sensitivity, measured by HIEC, varied widely among euglycemic young healthy men [15] and correlated with circulating metabolomic signatures [16]. The goal of the present study was to evaluate the altered expression pattern of skeletal muscle proteins associated with reduced insulin sensitivity in muscle biopsies taken at the basal state and during insulin infusion, at the end of the insulin clamp, when both glycemia and glucose infusion have stabilized. We used advanced proteomic techniques to identify a unique list of candidate proteins both at baseline and during insulin infusion, allowing identification of the proteins that correlate with insulin sensitivity, which may provide further information as to the molecular mechanisms of reduced insulin sensitivity in apparently healthy euglycemic subjects.
## Study participants
The overall design of the study flow is summarized in Figure 1. The details on the subjects and study protocol were previously reported [15]. In brief, healthy young men of Arab descent ($$n = 60$$) were examined for insulin sensitivity using HIEC. Muscle biopsies were obtained from the vastis lateralis before the clamp and during insulin infusion at the end of the clamp, when the plasma glucose level and the glucose infusion rate were stabilized [15]. The study was approved by the Institutional Review Board protocol ($\frac{14224}{14}$) of Hamad Medical Corporation, Doha, Qatar. All participants gave their signed informed consent. Participants were included in the study if they satisfied all the following criteria [1]: age >18–45 years, [2] body mass index ≤28 and ≥16, [3] normal CBC, [4] normal blood chemistry, [5] normal fasting glucose, [6] normal HbA1c, [7] normal glucose response to 75 g oral glucose tolerance test performed after 8 h of fasting, [8] normal ECG, and [9] commitment to the whole study protocol. Proteomic analyses were performed on muscle biopsies obtained from eight subjects who showed the highest insulin sensitivity (HIS) and eight other subjects who showed the lowest insulin sensitivity (LIS) (Table 1) among the 60 subjects reported previously [15].
**Figure 1:** *Workflow for the hyperinsulinemic euglycemic clamp, skeletal muscle biopsies, sample processing, and proteomic analysis.* TABLE_PLACEHOLDER:Table 1
## Hyperinsulinemic euglycemic clamp
As previously reported [16], the subjects were admitted to the research study unit at 7 a.m., after 10–12 h of overnight fasting, and a baseline muscle biopsy was obtained before the clamp study. Three polyethylene catheters were inserted in the antecubital fossa and back of the hand veins, enabling insulin/dextrose infusions, blood glucose measurements, and blood sampling. The insulin infusion (100 IU/ml insulin solution, Actrapid) rate was constant throughout the HIEC [40 mU/body surface area (m2)/min]. The body surface area (m2) [0.007184 x (height(cm)0.725) x (weight(kg)0.425)] was calculated as described. The blood glucose level was modulated by the infusion of $20\%$ dextrose, which was adjusted every 5 min to achieve a blood glucose level of 90 mg/dl (5 mmol/L). A second muscle biopsy was obtained under insulin infusion at 120 min (after the glucose infusion was stabilized). Similar to previous studies, the duration of the HIEC procedure was 120 min (17–19). Insulin sensitivity, as reflected by the whole-body glucose disposal rate (M-value, milligram of glucose infused per kilogram of body weight per minute), was computed after the stabilization of glycemia and of infusion rate during the last 60 min of the euglycemic clamp; this showed a wide variation ranging from 2 to 20 [16]. Muscle biopsies were quick-frozen in liquid nitrogen. In this study, we selected only the subjects with the lowest and the highest insulin sensitivities for proteomic analyses; the size of the groups was based on previous human studies (20–23).
## Sample preparation for proteomic assay
The frozen muscle biopsy samples of individuals with LIS (M ≤6, $$n = 8$$) and HIS (M >13, $$n = 8$$) were ground into fine powders using a mortar and pestle and liquid nitrogen. Protein extracts were isolated from the tissue samples using RIPA buffer. The lysates were centrifuged at 15,000 rpm for 10 min at +4°C, and the supernatants were transferred to new tubes, with the protein concentration determined using a BCA Protein Assay Kit (Pierce). The normalized protein samples were electrophoretically separated on $10\%$ SDS-PAGE; whole lines were excised and divided into eight equal parts, as previously described [24]. Gel pieces were then reduced with 10 mM dithiothreitol in 25 mM ammonium bicarbonate and alkylated with 10 mM iodoacetamide in 25 mM ammonium bicarbonate, followed by overnight digestion at 37°C using 20 ng/µl Trypsin/Lys-C (Promega). The peptides were eluted using $1\%$ formic acid, and the volume was reduced to 20 μl using a vacuum centrifuge (Eppendorf, Hamburg, Germany).
## Liquid chromatography–tandem mass spectrometry
Complex peptide mixtures were analyzed by shotgun proteomics using an Easy n-LC II (Thermo Scientific, Waltham, MA, USA) coupled to an Orbitrap Elite mass spectrometer (Thermo Scientific, Waltham, MA, USA), as previously described [24].
## MS data processing and statistical analysis
MS data processing and analysis were performed according to Leo et al. in 2019 [25]. The MaxQuant software version 1.6.17.0, according to the standard workflow with the built-in search engine Andromeda using the Uniprot human reference proteome database (downloaded October 12, 2017), was used for protein identification. The Max label-free quantification (LFQ) method, with retention time alignment and match-between-runs features in MaxQuant, was applied to extract the maximum possible quantification information. Protein abundance was calculated based on normalized spectral intensity (LFQ intensity).
MS data analysis was performed using the open-source software Perseus (version 1.6.14.0) [26]. The protein quantification and the statistical significance between the two groups were calculated using two-tailed t-test with permutation-based false discovery rates (FDR) of at least ±1.5 fold ($p \leq 0.05$). Functional enrichment analysis of the differentially abundant proteins was carried out using the online bioinformatics resource Database for Annotation, Visualization, and Integrated Discovery (DAVID). The distribution of proteins enriched under different categories, such as cellular components, biological processes, and pathways including Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome, was identified. Proteomic analysis was performed to identify proteins differentially expressed between baseline, before clamp (BC), and under insulin infusion during clamp (DC) as well as between LIS and HIS subjects both at baseline and under insulin infusion. The differences in baseline demographic, clinical, and biochemical data between LIS and HIS as presented in Table 1 were assessed by t-test.
## Functional annotation and pathway identification
Functional annotation was performed by gene set enrichment analysis using g:Profiler [27]. The statistically significant enrichment of biological processes and KEGG and Reactome pathways is extracted and plotted [27] using proteins that are uniquely expressed or significantly altered (increased or decreased) in HIS versus LIS at baseline or during insulin stimulation.
## Effects of insulin infusion on protein expression in the muscle biopsies
The proteomic profiles of all 16 subjects were analyzed to study the protein expression pattern between baseline (BC) and under insulin stimulation DC. The proteomic analysis resulted in the identification of 1,199 proteins BC and 1,164 proteins under insulin stimulation DC (Figure 2A); 56 proteins were only present in subjects before clamp and 21 proteins were present only under insulin stimulation; 1,143 proteins were present both at basal condition and under insulin stimulation (Figure 2B). Out of the 1,143 shared proteins, 564 were present in at least $50\%$ of the subjects (Figure 2C; Supplementary Table S1). Among the 1,143 shared proteins, four are differentially expressed—one protein was increased and three were reduced under insulin stimulation. The significant difference in protein abundance in response to insulin after comparative proteome analysis is graphically represented as a volcano plot, drawn using the fold change and the p-value (Figure 2D). The LFQ of the proteins with significant differential abundance is shown in Figure 2E and the fold changes in Figure 2F.
**Figure 2:** *Proteins levels at the baseline and at 2 hours of insulin infusion in the 16 subjects. (A) Bar chart representing the number of proteins identified in each condition. (B) Venn diagram showing the unique and common proteins in response to clamp. (C) Proteins sorted for their presence in at least half of the subjects. (D) Volcano plot showing the differential abundance of proteins in response to insulin infusion. (E) Box plots representing the proteins with significant differential abundance. *p=0.02, **p=0.001. BC: Before Clamp (basal values); DC: During Clamp (insulin infusion for 2 hours). (F) Fold changes in the proteins significantly increased during insulin stimulation versus baseline".*
## Differences in protein expression in muscle biopsies from subjects with low and high insulin sensitivities at baseline
The proteomic data were analyzed to identify differential protein expression at baseline (before clamp) between the LIS and HIS subjects. This analysis identified a total of 1,986 proteins (Figure 3A) that were reduced to 1,062 different proteins (Figure 3B). After sorting, 31 unique proteins were detected in the LIS group, 107 unique proteins were detected in the HIS group, and 924 proteins were detected in both groups (Figure 3B). Out of the 924 proteins shared between the two groups, 550 proteins were found in at least $50\%$ of the subjects of both groups (Figure 3C; Supplementary Table S2). Among the shared proteins, six proteins had differential abundance—of which three showed a higher expression and three a lower expression in HIS versus LIS. The significant difference between the groups from the comparative proteome analysis is graphically represented as a volcano plot, drawn using the fold change and the p-value (Figure 3D). The LFQ of the proteins with significant differential abundance is shown in Figure 3E and the fold changes in Figure 3F. One of the differentially abundant proteins (complement factor B) was identified as a biomarker for diabetes after sorting against the list of proteins identified to be involved in diabetes, as indicated in the peptide atlas database [28].
**Figure 3:** *Baseline protein levels in the subjects with low insulin sensitivity (LIS) versus the subjects with high insulin sensitivity (HIS). (A) Bar chart representing the number of proteins identified in each group. (B) Venn diagram showing the unique and common proteins in LIS and HIS. (C) Proteins sorted for presence in at least 50% of the subjects. (D). Volcano plot showing the differential abundance of proteins present in both the groups. (E) Box plots representing the proteins with significant differential abundance. *p=0.02, **p=0.002, *** p=0.0003. (F) Fold changes in the proteins showing statistically significant differential abundance in LIS versus HIS at baseline.*
## Differentially enriched pathways in relation to insulin sensitivity status [LIS versus HIS subjects at baseline (“baseline insulin sensitivity–enriched pathway”)]
Significantly altered biological processes resulted from 107 proteins uniquely expressed in the HIS group (Figure 3B), such as cellular metabolic process, intracellular transport, and aerobic/cellular respiration. The most enriched terms under the category are shown in Figure 4. Among the KEGG pathways, diabetic cardiomyopathy, metabolic pathways, and oxidative phosphorylation were found to be the most enriched pathways (Figure 4; Supplementary Table S6). Functional enrichment analysis using unique proteins also identified selenocysteine synthesis, citric acid (TCA) cycle, signaling by ROBO receptors, mitochondrial protein import, and metabolism as the pathways active in patients with high insulin sensitivity, as these were the most enriched terms under the Reactome pathways (Figure 4; Supplementary Table S6), while no significant enrichment in any of these pathways was observed in the low insulin sensitivity group.
**Figure 4:** *Functional annotation and classification by enrichment analysis of proteins uniquely present in the high insulin sensitivity group. Top enriched terms and their distribution categorized into biological processes, Kyoto Encyclopedia of Genes and Genomes pathways, and Reactome pathways (
Supplementary Table S6
).*
## Effects of insulin infusion on protein expression in the muscle biopsies from subjects with low versus high insulin sensitivity
The proteomic data of the LIS and HIS subjects under insulin infusion were analyzed to identify proteins that differ in their response to insulin between the two groups. After combining all data, 862 proteins were identified in the LIS group and 949 proteins in the HIS group, assigned to be exclusively present in at least one of the eight subjects per group, respectively (Figure 5A). After sorting, 27 proteins were found to be uniquely present in the LIS group and 114 proteins in the HIS group, while 835 proteins were shared between the two groups (Figure 5B). Out of the 835 proteins shared between the two groups, 497 proteins were found to be present in at least $50\%$ of the subjects of both groups (Figure 5C; Supplementary Table S3). Two proteins show a significant difference (Figures 5D, E). The comparative proteome analysis of shared proteins is graphically represented as a volcano plot, drawn using the fold change and the p-value (Figure 5D), LFQ (Figure 5E), and fold change (Figure 5F).
**Figure 5:** *Effects of insulin (2 h infusion) on protein levels in the subjects with low insulin sensitivity (LIS) versus the subjects with high insulin sensitivity (HIS). (A) Bar chart representing the number of proteins identified in each group. (B) Venn diagram showing the unique and common proteins in response to clamp. (C) Proteins sorted for their presence in at least 50% of the subjects. (D). Volcano plot showing the differential abundance of proteins between the groups after 2 hours of insulin infusion during the HIEC. (E) Box plots representing the proteins with a significant differential abundance. *p=0.01, **p=0.001. (F) Fold changes in the proteins showing statistically significant differential abundance in LIS versus HIS under-insulin stimulation.*
## Pathways differentially enriched in response to insulin infusion subjects with low and high insulin sensitivities
The 114 proteins (Figure 5B) uniquely present in HIS subjects in response to insulin infusion were found to be involved in biological processes such as response to hypoxia, ATP metabolic process, aerobic and cellular respiration, and respiratory electron transport chain as these were the most enriched terms under the category (Figure 6; Supplementary Table S7). Among the KEGG pathways, metabolic pathways, diabetic cardiomyopathy, non-alcoholic fatty liver disease, and oxidative phosphorylation were found to be the most enriched terms (Figure 6; Supplementary Table S7). Functional enrichment analysis using unique proteins also identified class I MHC-mediated antigen processing and presentation, protein localization, metabolism TCA cycle, and respiratory electron transport as the pathways active in high insulin sensitivity patients in response to insulin clamp, as these were the most enriched terms under the Reactome pathways (Figure 6; Supplementary Table S7), while no significant enrichment in any of the pathways was observed in the LIS group.
**Figure 6:** *Functional annotation and classification by enrichment analysis of proteins uniquely present in HIS or LIS group and those showing differential abundance. The top enriched terms and their distribution categorized into biological processes, KEGG pathways and Reactome pathways.*
## Effects of insulin infusion on protein expression in the muscle biopsies from each group (LIS and HIS)
The differentially abundant proteins in response to insulin stimulation were analyzed in each group (LIS and HIS). Only one protein was found to be increased, while six proteins were identified to be decreased in response to the clamp in the LIS group (Table 2; Supplementary Table S4). In the case of the HIS group, three proteins were increased in response to the clamp, while only two proteins were found to be decreased (Table 2; Supplementary Table S5).
**Table 2**
| Gene name | Protein name | Fold change | Group |
| --- | --- | --- | --- |
| MYL1 | Myosin light chain 1/3, skeletal muscle isoform | −2.0 | Low insulin sensitivity |
| APOB | Apolipoprotein B-100 | −3.5 | Low insulin sensitivity |
| UBE2V2 | Ubiquitin-conjugating enzyme E2 variant 2 | −1.8 | Low insulin sensitivity |
| MYL3 | Myosin light chain 3 | −2.9 | Low insulin sensitivity |
| NDUFS2 | NADH dehydrogenase (ubiquinone) iron–sulfur protein 2, mitochondrial | −2.1 | Low insulin sensitivity |
| CENPF | Centromere protein F | −1.9 | Low insulin sensitivity |
| BCAM | Basal cell adhesion molecule | 1.7 | Low insulin sensitivity |
| GRB2 | Growth factor receptor-bound protein 2 | −1.6 | High insulin sensitivity |
| NDUFS5 | NADH dehydrogenase (ubiquinone) iron–sulfur protein 5 | −2.2 | High insulin sensitivity |
| EPB42 | Erythrocyte membrane protein band 4.2 | 1.6 | High insulin sensitivity |
| PLG | Plasminogen | 1.5 | High insulin sensitivity |
| CFH | Complement factor H | 1.8 | High insulin sensitivity |
## Discussion
Skeletal muscles, liver, and fat are the main insulin target tissues; however, muscles play a major role in glucose clearance under insulin stimulation and strongly correlate with whole body insulin sensitivity [29]. We have previously shown that whole body sensitivity to insulin, measured by a hyperinsulinemic euglycemic clamp, varies widely among healthy young men [15]. We therefore hypothesized that the skeletal muscle proteome profile may show variations between high and low insulin sensitivity subjects (HIS and LIS). At the basal state, multiple proteins were uniquely detected in the HIS and LIS subjects, and over $85\%$ of total proteins were commonly detected in the two groups. Few of them showed differential expression levels. Based on Gene Ontology using DAVID software, the most abundant unique proteins present in the HIS group before and during insulin infusion showed a significant enrichment of the biological pathways involved in the mitochondria function and TCA and β-oxidation cycle (Figures 4 – 6).
Under fasting conditions, three common proteins are significantly downregulated, and three others are upregulated in the LIS group compared with the HIS group [$p \leq 0.05$, −1.5 ≥ fold change (FC) ≥1.5, FDR <$0.05\%$]; these were involved in mitochondria function, Glut4 translocation, and structural and contractile proteins. Interestingly, we observed a significant downregulation of creatine kinase B (CKB) as a marker for anaerobic ATP resynthesis enzyme in the LIS group compared with the HIS group. Consistent with our study, Højlund et al. showed that, in the human skeletal muscle, the level of CKB was reduced in T2D [10]. CKB may play a specific role in mitochondrial fuel oxidation [10, 30]. A significant observation of our analysis is that the Ras-related protein Rab10 showed a low abundance in $50\%$ of the LIS group. It is well established that the insulin activation of protein kinase B (also known as AKT) leads to the stimulation of the GTP-bound Ras-related protein (Rab10) and thereby triggers GLUT4 vesicle movement to the membrane [31]. In addition to Glut4 translocation, Rab-GAP was shown to control the uptake of saturated and unsaturated fatty acid into the skeletal muscle [32]. Several studies showed the dysregulation of intramyocellular fatty acid metabolism in the offspring of patients with T2D and obese patients with T2D. Recently, our group performed a metabolomic analysis of circulating plasma metabolites from the same cohort, and found that molecules involved in lipid metabolism, predominantly fatty acids, were upregulated in the LIS group compared with the HIS group [16]. A previous study in insulin-resistant muscle revealed reduced protein expression to be involved in mitochondrial function [33]. Hence, our results support that the abundance of proteins involved in mitochondrial functions is also downregulated [34]. Prohibitin 2 (PHB2), which represents the integrity of the mitochondrial inner membrane [35], was significantly reduced in LIS. The deletion of PHB2 results in the dysfunction of the mitochondria [36]. Additionally, at fasting conditions, we found several slightly downregulated proteins (Supplementary Table S2) that were involved in mitochondria function and TCA cycle, such as cytochrome b-c1 complex subunit (UQCRFS1; complex III), cytochrome c1 (CYC1; complex II), NADH dehydrogenase (NDUFS3; complex I), and ATP synthase subunit O, mitochondrial (ATP5O; complex V) (Supplementary Table S2). In fact, proteins involved in the TCA cycle [37, 38] and ATP synthase [39] tend to be less abundant in the insulin resistance of skeletal muscle.
Mitochondria are the intracellular sites of skeletal muscle fuel oxidation and ATP production, and mitochondrial dysfunction may play a critical role in impaired glucose metabolism observed in the skeletal muscle of T2D patients and their insulin-resistant offspring [40]. Furthermore, under insulin stimulation, proteins involved in mitochondrial energy metabolism were downregulated in the LIS group compared with the HIS group. We observed a significantly decreased ATP5F1 protein related to mitochondria synthase, involved in the respiratory chain protein (complex V) complex. Consistent with our finding of altered mitochondria function, most human studies showed mitochondrial dysfunction in skeletal muscle from insulin-resistant offspring of patients with T2D [10, 39, 41, 42], obesity, and T2D [11, 33, 43]. Indeed, it has been shown that in skeletal muscle, mitochondrial ATP and mRNA levels and protein synthesis are responsive to insulin infusion in nondiabetic subjects [40]; this indicates that insulin signaling modulates certain pathways which may influence mitochondrial proteins and functions. In our proteomic analysis, we observed a slightly low abundance of several proteins involved in mitochondria function in subjects with LIS compared with those in the HIS group, such as ATPC1 (complex V) (FC ≤-1.1) (Supplementary Table S3). Yang et al. showed that ATP5C1 is reduced in insulin-resistant non-diabetic Pima Indians [44]. Using magnetic resonance spectroscopy, several studies showed that the ATP synthesis rates were lower in the insulin-resistant offspring of T2D patients [39, 41]. Hojlund et al. demonstrated a decreased content of the ATP synthase subunit in the skeletal muscles of T2D patients [10]. Taken together, these data indicate that the insulin-stimulated rates of ATP synthesis are negatively affected very early in the pathogenesis of insulin resistance [39, 45, 46] and in T2D [47]. The current study also reveals alteration in proteins involved in the TCA cycle, such as succinate dehydrogenase (FC ≤-1.2; Supplementary Table S3). The SDH complex plays a vital role in cell metabolism, considering its participation in the TCA cycle and the electron transport chain. Sreekumar et al. showed a decreased SDHB expression in skeletal muscle after insulin treatment in T2D patients [48]. He et al. also showed that, within each type of fiber, skeletal muscle from obese and T2D had a lower SDH oxidative enzyme activity and increased lipid content compared with those of lean subjects [49]. Our data are consistent with the finding of He et al. who showed decreased oxidative enzyme activity and unchanged glycolysis in the skeletal muscle of T2D patients [49].
Human skeletal muscles are constituted of three major fiber types: type 1 (slow oxidative), 2A (fast oxidative glycolytic), and 2X fibers (fast glycolytic), defined by the presence of MYH7 (myosin heavy chain 7), MYH2, and MYH1, respectively. In the fasting state, our analysis showed an upregulation of MYH2 (Supplementary Table S2) and myosin light chain kinase 2 (MYLK2) in the LIS group compared with the HIS group; MYLK2 was also significantly upregulated by insulin in the LIS group (Figures 3E, F). The MYLK2 gene encodes the skeletal muscle myosin light chain kinase, with higher expression in fast skeletal muscles than in slow muscles. MYLK2 is linked to fast muscle proteins such as myosin light chain 1 (MYL1) [50]. At baseline, Giebelstein et al. reported that the upregulation of fast-muscle proteins negatively correlates with insulin sensitivity [33]. A study on fiber proportion in human skeletal muscle showed an increase of type 2A fiber (twofold) compared with type 1 in metabolic syndrome subjects.
Another important finding of our analysis is that perilipin 4 (PLIN4) was slightly upregulated under insulin stimulation (FC ≥1.4, Supplementary Table S3) in the LIS group compared with the HIS group. Perilipin 4 is expressed in skeletal muscle, heart, and adipose tissues, and it is preferentially located in lipid droplets containing cholesterol ester [51]. PLIN4 is recruited to the lipid droplet during droplet formation [52]. Poureymour et al. showed that PLIN4 is localized to intramuscular adipocytes and more highly expressed in slow-twitch muscle fibers compared with fast-twitch muscle [52]. PLIN4 mRNA is expressed in vastus lateralis biopsies from a healthy individual, and its levels are higher in slow-twitch than fast-twitch muscles. Unlike the PLIN3 protein, PLIN4 expression is reduced in response to prolonged endurance training [53]. These data are supported by a previous study that showed an increased intra-myocyte triglyceride level in insulin-resistant first-degree relatives of individuals with T2D [54]. Accumulation of intra-myocellular lipid is associated with reduced insulin sensitivity [55].
Moreover, several proteins were altered in the LIS group during insulin infusion compared with the baseline (Table 2; Supplementary Table S4). Those proteins, the sarcomere proteins myosin light chain1 (MYL1) and myosin light chain 3 (MYL3), were downregulated under insulin stimulation; this is consistent with a previous study that showed downregulation of the slow myosin light chain isoform protein in T2D patients [56]. Interestingly, our proteomic analysis revealed (Table 2; Supplementary Table S4) the downregulation of NDUFS2 (complex I) in the LIS group by insulin; complex I is involved in mitochondrion respiration. Insulin infusion also downregulated apolipoprotein B (APOB); APOB is an important component of LDL and VLDL, which distribute fat molecules to peripheral tissues such as skeletal muscle tissues [57]. Excess VLDL secretion has been indicated to deliver increased fatty acids and triglycerides to muscle and other tissues, further inducing insulin resistance [34]. Other proteins were altered following insulin stimulation in the LIS group compared to a baseline, such as Centromere protein F, which is involved in skeletal myogenesis, and basal cell adhesion molecule, which is involved in intracellular signaling. Some proteins identified in the HIS group following insulin infusion were different from those in the LIS group (Table 2; Supplementary Table S5), such as erythrocyte surface protein band 4.2, plasminogen, complement factor H, growth factor receptor-bound protein 2, and NADH dehydrogenase. Interestingly, our study identified a number of proteins involved in the mitochondrion respiratory chain, which were slightly altered in the LIS group compared with the HIS group, including complex I, II, III, and V at fasting condition (Supplementary Table S2), whereas following insulin infusion, we detected only two proteins that were altered, one in complex V and the other in TCA. Moreover, our study found that MYLK2 was upregulated in the LIS group compared with the HIS group in both conditions.
The strength of this study was the homogenous representative population of men with normal glycemia levels. Muscle biopsies from the same patients in whom circulating metabolites were measured were also used for proteomic analysis. A limitation of the study is the low number of subjects. Although we screened many participants, most of them failed to meet the criteria for participation in our study. Another limitation of the present study is that biopsies at later time points were not obtained; thus, we may have missed several protein changes that might have occurred at later times, particularly for proteins with a long half-life. Further, obtaining a skeletal muscle biopsy is a complex process and is very difficult to do in a large-scale population, and obtaining multiple timepoint samples is very challenging.
In conclusion, we have demonstrated that human skeletal muscles in apparently healthy male subjects of Arab descent show changes in a small number of proteins related to insulin sensitivity levels. In the fasting state, we found that 12 proteins were differentially expressed in the LIS group compared with the HIS group. Under insulin stimulation, a number of proteins, such as myosin chain and mitochondrial ATP synthase, remained altered in the LIS group. However, we did not detect any changes in glycolytic proteins in both conditions, as also shown in previous studies [11, 58]. Collectively, these data provide novel information regarding the metabolic pathways that correlate with insulin sensitivity levels in skeletal muscle and may represent early events for developing insulin resistance, pre-diabetes, and type 2 diabetes.
## Data availability statement
The data presented in the study are deposited in the Figshare repoository, accession number 10.6084/m9.figshare.21988487.
## Ethics statement
Protocols were approved by Institutional Review Boards of the Hamad Medical Corporation, Qatar (IRB protocol #$\frac{14224}{14}$). All study participants gave their written informed consent prior to participation in the study.
## Author contributions
AA-S designed the study protocol and obtained institutional review board approval. IA, MA, and SH recruited and screened the study subjects and performed the OGTT and HIEC procedures. MJ performed liquid chromatography–MS/MS analyses and provided raw proteomic data. RK performed bioinformatic and statistical analysis of the raw protein data provided by RK. AI, TS, MR, MA, IB, and JJ performed various assay measurements. IB, RK, and AA-S wrote and revised the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
Authors IB, RK, NS, MR, JJ, SH, MA, IA, AI and AA-S are employed by the Hamad Medical Corporation.
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/fendo.2022.1024832/full#supplementary-material
## References
1. Cho NH, Shaw JE, Karuranga S, Huang Y, da Rocha Fernandes JD, Ohlrogge AW. **IDF diabetes atlas: Global estimates of diabetes prevalence for 2017 and projections for 2045**. *Diabetes Res Clin Pract* (2018) **138**. DOI: 10.1016/j.diabres.2018.02.023
2. Al-Thani AA, Farghaly AH, Akram H, Khalifa SE, Vinodson B, Loares AM. **Public awareness and perceptions about diabetes in the state of Qatar**. *Cureus* (2018) **10**. DOI: 10.7759/cureus.2671
3. DeFronzo RA. **Lilly lecture 1987. the triumvirate: Beta-cell, muscle, liver. a collusion responsible for NIDDM**. *Diabetes* (1988) **37**. DOI: 10.2337/diab.37.6.667
4. Stumvoll M, Goldstein BJ, van Haeften TW. **Type 2 diabetes: Pathogenesis and treatment**. *Lancet* (2008) **371**. DOI: 10.1016/S0140-6736(08)60932-0
5. Kelley DE, Goodpaster BH, Storlien L. **Muscle triglyceride and insulin resistance**. *Annu Rev Nutr* (2002) **22**. DOI: 10.1146/annurev.nutr.22.010402.102912
6. Eriksson J, Franssila-Kallunki A, Ekstrand A, Saloranta C, Widen E, Schalin C. **Early metabolic defects in persons at increased risk for non-insulin-dependent diabetes mellitus**. *N Engl J Med* (1989) **321**. DOI: 10.1056/NEJM198908103210601
7. Muoio DM. **Metabolic inflexibility: When mitochondrial indecision leads to metabolic gridlock**. *Cell* (2014) **159**. DOI: 10.1016/j.cell.2014.11.034
8. Abdul-Ghani MA, DeFronzo RA. **Mitochondrial dysfunction, insulin resistance, and type 2 diabetes mellitus**. *Curr Diabetes Rep* (2008) **8**. DOI: 10.1007/s11892-008-0030-1
9. Martin BC, Warram JH, Rosner B, Rich SS, Soeldner JS, Krolewski AS. **Familial clustering of insulin sensitivity**. *Diabetes* (1992) **41**. DOI: 10.2337/diab.41.7.850
10. Hojlund K, Wrzesinski K, Larsen PM, Fey SJ, Roepstorff P, Handberg A. **Proteome analysis reveals phosphorylation of ATP synthase beta -subunit in human skeletal muscle and proteins with potential roles in type 2 diabetes**. *J Biol Chem* (2003) **278**. DOI: 10.1074/jbc.M212881200
11. Hwang H, Bowen BP, Lefort N, Flynn CR, De Filippis EA, Roberts C. **Proteomics analysis of human skeletal muscle reveals novel abnormalities in obesity and type 2 diabetes**. *Diabetes* (2010) **59** 33-42. DOI: 10.2337/db09-0214
12. Lefort N, Glancy B, Bowen B, Willis WT, Bailowitz Z, De Filippis EA. **Increased reactive oxygen species production and lower abundance of complex I subunits and carnitine palmitoyltransferase 1B protein despite normal mitochondrial respiration in insulin-resistant human skeletal muscle**. *Diabetes* (2010) **59**. DOI: 10.2337/db10-0174
13. Schonke M, Bjornholm M, Chibalin AV, Zierath JR, Deshmukh AS. **Proteomics analysis of skeletal muscle from leptin-deficient ob/ob mice reveals adaptive remodeling of metabolic characteristics and fiber type composition**. *Proteomics* (2018) **18**. DOI: 10.1002/pmic.201700375
14. Zabielski P, Lanza IR, Gopala S, Heppelmann CJ, Bergen HR, Dasari S. **Altered skeletal muscle mitochondrial proteome as the basis of disruption of mitochondrial function in diabetic mice**. *Diabetes* (2016) **65**. DOI: 10.2337/db15-0823
15. Suleiman N, Alkasem M, Hassoun S, Abdalhakam I, Bettahi I, Mir F. **Insulin sensitivity variations in apparently healthy Arab male subjects: correlation with insulin and c peptide**. *BMJ Open Diabetes Res Care* (2021) **9**. DOI: 10.1136/bmjdrc-2020-002039
16. Halama A, Suleiman NN, Kulinski M, Bettahi I, Hassoun S, Alkasem M. **The metabolic footprint of compromised insulin sensitivity under fasting and hyperinsulinemic-euglycemic clamp conditions in an Arab population**. *Sci Rep* (2020) **10** 17164. DOI: 10.1038/s41598-020-73723-8
17. DeFronzo RA, Tobin JD, Andres R. **Glucose clamp technique: A method for quantifying insulin secretion and resistance**. *Am J Physiol* (1979) **237**. DOI: 10.1152/ajpendo.1979.237.3.E214
18. Mullen E, Ohlendieck K. **Proteomic profiling of non-obese type 2 diabetic skeletal muscle**. *Int J Mol Med* (2010) **25**. DOI: 10.3892/ijmm_00000364
19. Katz A, Nambi SS, Mather K, Baron AD, Follmann DA, Sullivan G. **Quantitative insulin sensitivity check index: A simple, accurate method for assessing insulin sensitivity in humans**. *J Clin Endocrinol Metab* (2000) **85**. DOI: 10.1210/jcem.85.7.6661
20. Hussey SE, Sharoff CG, Garnham A, Yi Z, Bowen BP, Mandarino LJ. **Effect of exercise on the skeletal muscle proteome in patients with type 2 diabetes**. *Med Sci Sports Exerc* (2013) **45**. DOI: 10.1249/MSS.0b013e3182814917
21. Campbell LE, Langlais PR, Day SE, Coletta RL, Benjamin TR, De Filippis EA. **Identification of novel changes in human skeletal muscle proteome after roux-en-Y gastric bypass surgery**. *Diabetes* (2016) **65**. DOI: 10.2337/db16-0004
22. Barberio MD, Dohm GL, Pories WJ, Gadaleta NA, Houmard JA, Nadler EP. **Type 2 diabetes modifies skeletal muscle gene expression response to gastric bypass surgery**. *Front Endocrinol (Lausanne)* (2021) **12**. DOI: 10.3389/fendo.2021.728593
23. Savikj M, Stocks B, Sato S, Caidahl K, Krook A, Deshmukh AS. **Exercise timing influences multi-tissue metabolome and skeletal muscle proteome profiles in type 2 diabetic patients - a randomized crossover trial**. *Metabolism* (2022) **135** 155268. DOI: 10.1016/j.metabol.2022.155268
24. Al-Thani AM, Voss SC, Al-Menhali AS, Barcaru A, Horvatovich P, Al Jaber H. **Whole blood storage in CPDA1 blood bags alters erythrocyte membrane proteome**. *Oxid Med Cell Longev* (2018) **2018** 6375379. DOI: 10.1155/2018/6375379
25. Leo R, Therachiyil L, Siveen SK, Uddin S, Kulinski M, Buddenkotte J. **Protein expression profiling identifies key proteins and pathways involved in growth inhibitory effects exerted by guggulsterone in human colorectal cancer cells**. *Cancers (Basel)* (2019) **11**. DOI: 10.3390/cancers11101478
26. Tyanova S, Temu T, Sinitcyn P, Carlson A, Hein MY, Geiger T. **The Perseus computational platform for comprehensive analysis of (prote)omics data**. *Nat Methods* (2016) **13**. DOI: 10.1038/nmeth.3901
27. Raudvere U, Kolberg L, Kuzmin I, Arak T, Adler P, Peterson H. **g:Profiler: a web server for functional enrichment analysis and conversions of gene lists (2019 update)**. *Nucleic Acids Res* (2019) **47**. DOI: 10.1093/nar/gkz369
28. Desiere F, Deutsch EW, King NL, Nesvizhskii AI, Mallick P, Eng J. **The PeptideAtlas project**. *Nucleic Acids Res* (2006) **34**. DOI: 10.1093/nar/gkj040
29. Zierath JR, Krook A, Wallberg-Henriksson H. **Insulin action and insulin resistance in human skeletal muscle**. *Diabetologia* (2000) **43**. DOI: 10.1007/s001250051457
30. Kelley DE, Mandarino LJ. **Fuel selection in human skeletal muscle in insulin resistance: a reexamination**. *Diabetes* (2000) **49**. DOI: 10.2337/diabetes.49.5.677
31. Sano H, Eguez L, Teruel MN, Fukuda M, Chuang TD, Chavez JA. **Rab10, a target of the AS160 rab GAP, is required for insulin-stimulated translocation of GLUT4 to the adipocyte plasma membrane**. *Cell Metab* (2007) **5** 293-303. DOI: 10.1016/j.cmet.2007.03.001
32. Benninghoff T, Espelage L, Eickelschulte S, Zeinert I, Sinowenka I, Muller F. **The RabGAPs TBC1D1 and TBC1D4 control uptake of long-chain fatty acids into skeletal muscle**. *Diabetes* (2020) **69**. DOI: 10.2337/db20-0180
33. Giebelstein J, Poschmann G, Hojlund K, Schechinger W, Dietrich JW, Levin K. **The proteomic signature of insulin-resistant human skeletal muscle reveals increased glycolytic and decreased mitochondrial enzymes**. *Diabetologia* (2012) **55**. DOI: 10.1007/s00125-012-2456-x
34. Zammit VA, Waterman IJ, Topping D, McKay G. **Insulin stimulation of hepatic triacylglycerol secretion and the etiology of insulin resistance**. *J Nutr* (2001) **131**. DOI: 10.1093/jn/131.8.2074
35. Supale S, Thorel F, Merkwirth C, Gjinovci A, Herrera PL, Scorrano L. **Loss of prohibitin induces mitochondrial damages altering beta-cell function and survival and is responsible for gradual diabetes development**. *Diabetes* (2013) **62**. DOI: 10.2337/db13-0152
36. Merkwirth C, Martinelli P, Korwitz A, Morbin M, Bronneke HS, Jordan SD. **Loss of prohibitin membrane scaffolds impairs mitochondrial architecture and leads to tau hyperphosphorylation and neurodegeneration**. *PloS Genet* (2012) **8**. DOI: 10.1371/journal.pgen.1003021
37. Ritov VB, Menshikova EV, Azuma K, Wood R, Toledo FG, Goodpaster BH. **Deficiency of electron transport chain in human skeletal muscle mitochondria in type 2 diabetes mellitus and obesity**. *Am J Physiol Endocrinol Metab* (2010) **298**. DOI: 10.1152/ajpendo.00317.2009
38. Gaster M. **Reduced TCA flux in diabetic myotubes: A governing influence on the diabetic phenotype**. *Biochem Biophys Res Commun* (2009) **387**. DOI: 10.1016/j.bbrc.2009.07.064
39. Petersen KF, Dufour S, Shulman GI. **Decreased insulin-stimulated ATP synthesis and phosphate transport in muscle of insulin-resistant offspring of type 2 diabetic parents**. *PloS Med* (2005) **2**. DOI: 10.1371/journal.pmed.0020233
40. Stump CS, Short KR, Bigelow ML, Schimke JM, Nair KS. **Effect of insulin on human skeletal muscle mitochondrial ATP production, protein synthesis, and mRNA transcripts**. *Proc Natl Acad Sci U.S.A.* (2003) **100** 7996-8001. DOI: 10.1073/pnas.1332551100
41. Petersen KF, Dufour S, Befroy D, Garcia R, Shulman GI. **Impaired mitochondrial activity in the insulin-resistant offspring of patients with type 2 diabetes**. *N Engl J Med* (2004) **350**. DOI: 10.1056/NEJMoa031314
42. Befroy DE, Petersen KF, Dufour S, Mason GF, de Graaf RA, Rothman DL. **Impaired mitochondrial substrate oxidation in muscle of insulin-resistant offspring of type 2 diabetic patients**. *Diabetes* (2007) **56**. DOI: 10.2337/db06-0783
43. Chae S, Kim SJ, Do Koo Y, Lee JH, Kim H, Ahn BY. **A mitochondrial proteome profile indicative of type 2 diabetes mellitus in skeletal muscles**. *Exp Mol Med* (2018) **50** 1-14. DOI: 10.1038/s12276-018-0154-6
44. Yang X, Pratley RE, Tokraks S, Bogardus C, Permana PA. **Microarray profiling of skeletal muscle tissues from equally obese, non-diabetic insulin-sensitive and insulin-resistant pima indians**. *Diabetologia* (2002) **45**. DOI: 10.1007/s00125-002-0905-7
45. Lillioja S, Mott DM, Howard BV, Bennett PH, Yki-Jarvinen H, Freymond D. **Impaired glucose tolerance as a disorder of insulin action. longitudinal and cross-sectional studies in pima indians**. *N Engl J Med* (1988) **318**. DOI: 10.1056/NEJM198805123181901
46. Lillioja S, Mott DM, Spraul M, Ferraro R, Foley JE, Ravussin E. **Insulin resistance and insulin secretory dysfunction as precursors of non-insulin-dependent diabetes mellitus. prospective studies of pima indians**. *N Engl J Med* (1993) **329**. DOI: 10.1056/NEJM199312303292703
47. Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag S, Lehar J. **PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes**. *Nat Genet* (2003) **34**. DOI: 10.1038/ng1180
48. Sreekumar R, Halvatsiotis P, Schimke JC, Nair KS. **Gene expression profile in skeletal muscle of type 2 diabetes and the effect of insulin treatment**. *Diabetes* (2002) **51**. DOI: 10.2337/diabetes.51.6.1913
49. He J, Watkins S, Kelley DE. **Skeletal muscle lipid content and oxidative enzyme activity in relation to muscle fiber type in type 2 diabetes and obesity**. *Diabetes* (2001) **50**. DOI: 10.2337/diabetes.50.4.817
50. Chomentowski P, Coen PM, Radikova Z, Goodpaster BH, Toledo FG. **Skeletal muscle mitochondria in insulin resistance: Differences in intermyofibrillar versus subsarcolemmal subpopulations and relationship to metabolic flexibility**. *J Clin Endocrinol Metab* (2011) **96** 494-503. DOI: 10.1210/jc.2010-0822
51. Hsieh K, Lee YK, Londos C, Raaka BM, Dalen KT, Kimmel AR. **Perilipin family members preferentially sequester to either triacylglycerol-specific or cholesteryl-ester-specific intracellular lipid storage droplets**. *J Cell Sci* (2012) **125**. DOI: 10.1242/jcs.104943
52. Wolins NE, Quaynor BK, Skinner JR, Schoenfish MJ, Tzekov A, Bickel PE. **S3-12, adipophilin, and TIP47 package lipid in adipocytes**. *J Biol Chem* (2005) **280**. DOI: 10.1074/jbc.M500978200
53. Gjelstad IM, Haugen F, Gulseth HL, Norheim F, Jans A, Bakke SS. **Expression of perilipins in human skeletal muscle**. *Arch Physiol Biochem* (2012) **118** 22-30. DOI: 10.3109/13813455.2011.630009
54. Jacob S, Machann J, Rett K, Brechtel K, Volk A, Renn W. **Association of increased intramyocellular lipid content with insulin resistance in lean nondiabetic offspring of type 2 diabetic subjects**. *Diabetes* (1999) **48**. DOI: 10.2337/diabetes.48.5.1113
55. Sachs S, Zarini S, Kahn DE, Harrison KA, Perreault L, Phang T. **Intermuscular adipose tissue directly modulates skeletal muscle insulin sensitivity in humans**. *Am J Physiol Endocrinol Metab* (2019) **316**. DOI: 10.1152/ajpendo.00243.2018
56. Oberbach A, Bossenz Y, Lehmann S, Niebauer J, Adams V, Paschke R. **Altered fiber distribution and fiber-specific glycolytic and oxidative enzyme activity in skeletal muscle of patients with type 2 diabetes**. *Diabetes Care* (2006) **29** 895-900. DOI: 10.2337/diacare.29.04.06.dc05-1854
57. Kang S, Davis RA. **Cholesterol and hepatic lipoprotein assembly and secretion**. *Biochim Biophys Acta* (2000) **1529**. DOI: 10.1016/S1388-1981(00)00151-7
58. Hojlund K, Staehr P, Hansen BF, Green KA, Hardie DG, Richter EA. **Increased phosphorylation of skeletal muscle glycogen synthase at NH2-terminal sites during physiological hyperinsulinemia in type 2 diabetes**. *Diabetes* (2003) **52**. DOI: 10.2337/diabetes.52.6.1393
|
---
title: 'Correlation of multiple lipid and lipoprotein ratios with nonalcoholic fatty
liver disease in patients with newly diagnosed type 2 diabetic mellitus: A retrospective
study'
authors:
- Ran Li
- Dehong Kong
- Zhengqin Ye
- Guannan Zong
- Kerong Hu
- Wei Xu
- Ping Fang
- Liya Zhang
- Yun Zhou
- Keqin Zhang
- Ying Xue
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9982122
doi: 10.3389/fendo.2023.1127134
license: CC BY 4.0
---
# Correlation of multiple lipid and lipoprotein ratios with nonalcoholic fatty liver disease in patients with newly diagnosed type 2 diabetic mellitus: A retrospective study
## Abstract
### Background and objective
The diagnostic value of lipid and lipoprotein ratios for NAFLD in newly diagnosed T2DM remains unclear. This study aimed to investigate the relationships between lipid and lipoprotein ratios and the risk of NAFLD in subjects with newly diagnosed T2DM.
### Methods
A total of 371 newly diagnosed T2DM patients with NAFLD and 360 newly diagnosed T2DM without NAFLD were enrolled in the study. Demographics variables, clinical history and serum biochemical indicators of the subjects were collected. Six lipid and lipoprotein ratios, including triglycerides to high-density lipoprotein-cholesterol (TG/HDL-C) ratio, cholesterol to HDL-C (TC/HDL-C) ratio, free fatty acid to HDL-C (FFA/HDL-C) ratio, uric acid to HDL-C (UA/HDL-C) ratio, low-density lipoprotein-cholesterol to HDL-C (LDL-C/HDL-C) ratio, apolipoprotein B to apolipoprotein A1 (APOB/A1) ratio, were calculated. We compared the differences in lipid and lipoprotein ratios between NAFLD group and non-NAFLD group, and further analyzed the correlation and diagnostic value of these ratios with the risk of NAFLD in the newly diagnosed T2DM patients.
### Results
The proportion of NAFLD in patients with newly diagnosed T2DM increased progressively over the range Q1 to Q4 of six lipid ratios, including the TG/HDL-C ratio, TC/HDL-C ratio, FFA/HDL-C ratio, UA/HDL-C ratio, LDL-C/HDL-C ratio, and APOB/A1 ratio. After adjusting for multiple confounders, TG/HDL-C, TC/HDL-C, UA/HDL-C, LDL-C/HDL-C and APOB/A1 were all strongly correlated with the risk of NAFLD in patients with newly diagnosed T2DM. In patients with newly-onset T2DM, the TG/HDL-C ratio was the most powerful indicator for the diagnosis of NAFLD among all six indicators, with an area under the curve (AUC) of 0.732 ($95\%$ CI 0.696–0.769). In addition, TG/HDL-C ratio>1.405, with a sensitivity of $73.8\%$ and specificity of $60.1\%$, had a good diagnostic ability for NAFLD in patients with newly diagnosed T2DM.
### Conclusions
The TG/HDL-C ratio may be an effective marker to help identify the risk of NAFLD in patients with newly diagnosed T2DM.
## Introduction
Non-alcoholic fatty liver disease (NAFLD) and type 2 diabetes mellitus (T2DM) have a strong bidirectional association, and the prevalence of both is increasing simultaneously [1, 2]. A recent meta-analysis reported the global prevalence of NAFLD in patients with T2DM was $55.5\%$ [2]. Moreover, the global prevalence of T2DM in patients with NAFLD and non-alcoholic steatohepatitis (NASH) patients was $22.51\%$, and $43.63\%$, respectively [3]. There is now growing evidence that patients with T2DM combined with NAFLD tend to have poorer glycemic control than T2DM patients without NAFLD, and are at higher risk of developing NASH, cirrhosis or even hepatocellular carcinoma compared to NAFLD patients without T2DM [4]. On the other hand, the incidence of chronic diabetic complications, such as cardiovascular disease (CVD), chronic kidney disease (CKD) and retinopathy, is also significantly higher in patients with T2DM combined with NAFLD than in those without combined NAFLD [4, 5].
Liver biopsy is the gold standard for the diagnosis of NAFLD and NASH cirrhosis. However, in clinical practice, the invasiveness, poor acceptability and high cost of liver biopsy make it difficult to use for widespread screening in the general population [6, 7]. Conventional ultrasonography is commonly used for screening and diagnosis of NAFLD [7]. However, due to the large number of patients with T2DM, routine liver ultrasound screening in all T2DM patients requires extremely expensive medical expenses. In addition, a large number of rural health centers or community hospitals lack ultrasound equipment and qualified ultrasonographers. Therefore, several previous studies have pinned hopes for early screening of patients with NAFLD on various serum markers (8–13). However, to date, no serum marker has become an accepted diagnostic indicator for NAFLD.
It is well known that serum biochemical indices of routine physical examination include liver enzymes and blood lipids. Previous studies have shown that liver enzyme levels are not useful for screening for NAFLD as their changes do not necessarily correspond to the degree of hepatic steatosis [14]. Dyslipidemia, including increases in triglycerides (TG), cholesterol (TC), low-density lipoprotein-cholesterol (LDL-C) and decreases in high-density lipoprotein-cholesterol (HDL-C), is strongly associated with NAFLD [11, 13, 15, 16]. Several current data have indicated that lipid and lipoprotein ratios are more valuable than individual lipid values in predicting the risk of T2DM or NAFLD because they can reflect the interaction between lipid components (11–13, 15–17). Among them, the ratios of TG to HDL-C (TG/HDL-C) [12, 13], TC to HDL-C (TC/HDL-C) [11], uric acid (UA) to HDL (UA/HDL-C) [9], LDL-C to HDL-C (LDL-C/HDL-C) [16] and apolipoprotein B to apolipoprotein A1 (APOB/A1) [17] have been previously reported to be associated with the risk of NAFLD in different populations. Besides, TG/HDL-C and TC/HDL-C have been described as promising parameters for the diagnosis of prediabetes and T2DM [15, 18].
Currently, there are no studies on the relationship between the aforementioned lipid and lipoprotein ratios and NAFLD in a newly diagnosed T2DM population with no history of medication and no diabetic complications. Considering the high prevalence and risk of combined NAFLD in T2DM, there is a need for early identification of NAFLD in newly diagnosed diabetic patients for better early intervention. Therefore, this study sought to evaluate the value of the above-mentioned lipid-lipoprotein ratios for assessing the risk of NAFLD in patients with newly diagnosed T2DM.
## Participants
This study was a retrospective study approved by the Ethics Committee of Tongji Hospital, Tongji University School of Medicine (K-2021-010). A total of 1021 patients who were first diagnosed with T2DM and not receiving anti-diabetic medication at the inpatient department of the Department of Endocrinology, Tongji Hospital, Tongji University, from June 2018 to December 2020 were enrolled.
The diagnosis of T2DM was based on the criteria of the World Health Organization [1999] [19]. The diagnosis of NAFLD was made by abdominal ultrasound assessment of hepatic steatosis [20]. The criteria were as follows: 1) diffusely enhanced liver echogenicity that was stronger than that of the kidneys or spleen; 2) attenuation of far-field echogenicity depth in the liver region; 3) vascular blurring on color Doppler ultrasound; 4) poorly displayed intrahepatic luminal structures. The exclusion criteria for this study were as follows: 1) subjects with a history of drinking, or alcohol consumption ≥140 g per week for men and ≥70 g per week for women; 2) subjects with a history of autoimmune hepatitis, drug-induced hepatic disease, viral hepatitis or other known diseases that may lead to fatty liver; 3) subjects treated with lipid-lowing agents or anti-diabetic medications; 4) subjects who did not receive liver ultrasound; 5) subjects with incomplete clinical information. Finally, 371 patients with newly diagnosed T2DM combined with NAFLD and 360 newly diagnosed T2DM patients without NAFLD were included in this study (Supplementary Figure 1).
## Data collection
Basic clinical data and lifestyle information of the study population were collected, including age, sex, height, body weight, and smoking/alcohol consumption habits. Smoking/drinking habits depended on whether the individual currently smoked or drank excessively (140 g/week for men and 70 g/week for women). The levels of blood lipids, blood glucose, liver function and renal function were collected in this study, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), Gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), serum creatinine (Scr), UA, fasting blood-glucose (FBG), glycosylated hemoglobin (HbA1c), fasting insulin (FINS), TG, TC, free fatty acid (FFA), LDL-C, HDL-C, APOA1 and APOB. The lipid profiles, liver function, renal function and FBG were detected on an automatic biochemical analyzer (AU 5800, Beckman Coulter, USA). HbA1c was assessed by high-performance liquid chromatography (HLC-723G8, TOSOH CORPORATION, Japan). FINS was measured by an automatic electrochemiluminescence immunoassay analyzer (ADVIA centaur XP, Siemens, Germany).
Body mass index (BMI) was calculated as body weight (kg)/height2 (m2). Homeostasis model assessment-insulin resistance (HOMA-IR) reflects the state of insulin resistance (IR) in the body, and the equation is: HOMA-IR = fasting insulin (μU/dL) × fasting blood glucose (mg/dL)/22.5. TG/HDL-C, TC/HDL-C, FFA/HDL-C, LDL-C/HDL-C, UA/HDL-C, and AOB/A1 ratios were calculated as TG (mmol/L)/HDL-C (mmol/L), TC (mmol/L)/HDL-C (mmol/L), FFA (mmol/L)/HDL-C (mmol/L), LDL-C (mmol/L)/HDL-C (mmol/L), UA (μmol/L)/HDL-C (mmol/L), APOB (mmol/L)/APOA1 (mmol/L) respectively.
## Statistical analysis
Statistical analysis was performed using SPSS 22.0 software. Continuous variables with normal distribution were expressed as mean ± standard deviation (SD), and independent samples T-test was used to compare the non-NAFLD group with the NAFLD group. Continuous variables without a normal distribution were expressed as median (interquartile range), and compared between the non-NAFLD and NAFLD groups using the Kruskal-Wallis test. Categorical variables were shown as proportions, and compared using Chi-squared tests. We divided the TG/HDL-C ratio, TC/HDL-C ratio, FFA/HDL-C ratio, LDL-C/HDL-C ratio, UA/HDL-C and APOB/A1 ratio into four quartiles and converted them into conventional categorical variables, i.e. Q1 < $25\%$, Q2 25-$50\%$, Q3 50-$75\%$ and Q4 ≥ $75\%$. Chi-square test was used to compare the proportion of NAFLD in patients with newly-onset T2DM in the above categorical variables. Continuous variables that did not conform to a normal distribution were log-transformed, and Pearson correlation analysis was conducted between the six lipid-lipoprotein ratios and each variable. After adjusting for potential confounders, a bivariate logistic regression analysis was performed in newly diagnosed T2DM patients to explore the association between several lipid ratios and NAFLD. Three models were used in this study, model 1 unadjusted; model 2 adjusted for age, sex, current smoking status, and BMI; and model 3 adjusted for age, sex, current smoking status, BMI, ALT, AST, GGT, ALP, Scr, FBG, HbA1c and FINS. The receiver operating characteristic (ROC) curve analysis was used to compare the relative diagnostic ability of the six lipids and lipoprotein ratios for new-onset T2DM combined with NAFLD. The indicator with the largest area under the ROC curve (AUC) was considered the best diagnostic marker.
## Clinical characteristics of the study subjects
A total of 731 newly diagnosed T2DM subjects were enrolled in the study, including 360 patients without NAFLD (non-NAFLD group), 371 patients with NAFLD (NAFLD group). That is, the overall proportion of NAFLD in patients with newly diagnosed T2DM was $50.8\%$. In non-NAFLD group, the mean age was 57.21 ± 16.83 years, with $58.9\%$ ($\frac{212}{360}$) of males and $41.1\%$ ($\frac{148}{360}$) of females. In NAFLD group, the mean age was 51.45 ± 15.86 years, of which $65.2\%$ ($\frac{242}{371}$) were males and $34.8\%$ ($\frac{129}{371}$) were females. Moreover, newly diagnosed T2DM subjects combined with NAFLD smoked more and had a higher BMI than subjects without NAFLD. As expected, patients with NAFLD had higher ALT, AST, GGT, ALP, UA, FBG, FINS, HOMA-IR, TG, TC, FFA, LDL-C and APOB than non-NAFLD group, while HDL-C and APOA1 were lower than non-NAFLD group. There was no significant difference in Scr and HbA1c between the two groups (Table 1).
**Table 1**
| Unnamed: 0 | Non-NAFLD(N=360) | NAFLD(N=371) | P-Values |
| --- | --- | --- | --- |
| Age (years) | 57.21 ± 16.83 | 51.45 ± 15.86 | <0.001 |
| Sex, male/female (n) | 212/148 | 242/129 | 0.077 |
| Current smoking (%) | 79 (21.9) | 113 (30.5) | 0.009 |
| BMI (kg/m2) | 23.65 ± 4.20 | 27.28 ± 4.82 | <0.001 |
| ALT (U/L) | 19.30 (13.43-29.4) | 30.30 (19.00-55.1) | <0.001 |
| AST (U/L) | 18.75 (14.90-24.85) | 23.3 (17.4-38.30) | <0.001 |
| GGT (U/L) | 25.5 (17.45-38.85) | 39.9 (26.60-66.05) | <0.001 |
| ALP (U/L) | 91.13 ± 29.98 | 98.52 ± 38.38 | 0.008 |
| Scr (μmol/L) | 71.45 (60.50-82.98) | 74.7 (63.10-74.70) | 0.202 |
| UA (μmol/L) | 306.06 ± 94.57 | 343.02 ± 100.63 | <0.001 |
| FBG (mmol/L) | 9.72 ± 4.19 | 11.45 ± 5.50 | <0.001 |
| HbA1c (%) | 10.60 ± 4.20 | 10.84 ± 2.59 | 0.280 |
| FINS (μIU/mL) | 9.18 (6.07-12.65) | 11.72 (8.04-15.79) | <0.001 |
| HOMA-IR | 3.43 (2.14-5.62) | 5.23 (3.36-7.85) | <0.001 |
| TG (mmol/L) | 1.27 (0.92-1.72) | 1.85 (1.31-2.85) | <0.001 |
| TC (mmol/L) | 4.72 ± 1.15 | 5.25 ± 1.81 | <0.001 |
| FFA (mmol/L) | 0.51 ± 0.23 | 0.58 ± 0.22 | <0.001 |
| LDL-C (mmol/L) | 3.16 ± 0.91 | 3.35 ± 1.01 | 0.006 |
| HDL-C (mmol/L) | 1.07 ± 0.24 | 0.95 ± 0.22 | <0.001 |
| APOA1 (mmol/L) | 1.16 ± 0.18 | 1.12 ± 0.18 | 0.006 |
| APOB (mmol/L) | 0.98 ± 0.23 | 1.09 ± 0.22 | <0.001 |
| TG/HDL-C | 1.20 (0.84-1.80) | 2.05 (1.37-3.23) | <0.001 |
| TC/HDL-C | 4.61 ± 1.19 | 5.57 ± 1.86 | <0.001 |
| FFA/HDL-C | 0.51 ± 0.26 | 0.64 ± 0.32 | <0.001 |
| LDL-C/HDL-C | 3.09 ± 0.92 | 3.62 ± 1.04 | <0.001 |
| UA/HDL-C | 303.83 ± 131.20 | 386.35 ± 159.50 | <0.001 |
| APOB/APOA1 | 0.87 ± 0.22 | 1.00 ± 0.27 | <0.001 |
The distribution of the ratios of TG/HDL-C, TC/HDL-C, FFA/HDL-C, LDL-C/HDL-C, UA/HDL-C, and APOB/A1 in the non-NAFLD and NAFLD groups, respectively, is shown in Supplementary Figure 2. In addition, the ratios of TG/HDL-C, TC/HDL-C, FFA/HDL-C, LDL-C/HDL-C, UA/HDL-C, APOB/A1 were significantly higher in new-onset T2DM patients with NAFLD than in patients without NAFLD (Table 1).
## Associations of six lipid and lipoprotein-related indices with NAFLD in newly diagnosed T2DM
The proportion of NAFLD in newly diagnosed T2DM patients increased progressively across the Q1-Q4 range of six lipid-lipoprotein ratios, including TG/HDL-C (22.0, 49.4, 58.7 and $75.4\%$, respectively), TC/HDL-C (30.9, 44.6, 54.0 and $74.0\%$, respectively), FFA/HDL-C (34.4, 46.8, 57.6 and $66.3\%$, respectively), LDL-C/HDL-C (35.9, 44.3, 53.6 and $69.0\%$, respectively), UA/HDL-C (32.8, 45.8, 53.1 and $72.2\%$, respectively) and APOB/A1 (36.0, 41.8, 55.8 and $71.9\%$, respectively) (Figure 1). Compared to the lowest quartile (Q1) of the above six lipid-lipoprotein ratios, the proportion of NAFLD was significantly higher in the increasing quartiles (Q2-Q4) of the TG/HDL-C and TC/HDL-C ratios, and in the increasing quartile (Q3-Q4) of the FFA/HDL-C, LDL-C/HDL-C and UA/HDL-C and APOB/A1 ratios (Figure 1). This increasing trend suggested that the greater the six lipid ratios in newly diagnosed T2DM patients, the higher the likelihood of NAFLD occurrence in those patients. Logistic regression analysis further demonstrated that the 6 lipid ratios in model 1 were positively correlated with NAFLD in new-onset T2DM patients without adjusting for other factors (Table 2). Pearson correlation analysis was shown in Supplementary Table 1, indicating that the 6 lipid ratios were strongly correlated with age, sex, BMI, hepatic function markers, renal function indicators, blood glucose indicators and blood lipid indicators, respectively. Therefore, we next corrected for these factors in models 2 and 3 of the logistic regression analysis (Table 2). After adjusting for age, sex, current smoking status and BMI in model 2, the 6 lipid ratios remained significantly and positively associated with NAFLD in newly diagnosed T2DM patients (Table 2). Even after adjusting for age, sex, current smoking status, BMI, ALT, AST, GGT, ALP, Scr, FBG, HbA1c and FINS in model 3, 5 lipid ratios (TG/HDL-C, TC/HDL-C, LDL-C/HDL-C, UA/HDL-C, and APOB/A1) remained positively associated with the risk of NAFLD in patients with newly diagnosed T2DM (Table 2). It should be noted that in model 1-3, APOB/A1 ratio had the strongest correlation with NAFLD in patients with newly diagnosed T2DM [model1 odds ratio (OR)= 10.72, $P \leq 0.001$; model2 OR=4.81, $$P \leq 0.001$$ and model3 OR= 6.25; $$P \leq 0.006$$, respectively] (Table 2).
**Figure 1:** *Proportion of NAFLD in patients with newly diagnosed T2DM across the quartiles of multiple lipid ratios (Q1-Q4). *$P \leq 0.001$ vs. Q1. NAFLD, non-alcoholic fatty liver disease; T2DM, type 2 diabetes mellitus; TG/HDL-C, triglycerides to high-density lipoprotein-cholesterol ratio; TC/HDL-C, cholesterol to HDL-C ratio; FFA/HDL-C, free fatty acid to HDL-C ratio; UA/HDL-C, uric acid to HDL-C ratio; LDL-C/HDL-C, low-density lipoprotein-cholesterol to HDL-C ratio; APOB/A1, apolipoprotein B to apolipoprotein A1 ratio.* TABLE_PLACEHOLDER:Table 2
## Diagnostic value of the six lipid-lipoprotein ratios for NAFLD in newly diagnosed T2DM patients
ROC curves were then constructed to compare the ability of the six lipid-lipoprotein ratios and their associated lipid metrics to discriminate NAFLD in newly diagnosed T2DM patients (Supplementary Figure 3). The area under the curve (AUC) of all lipid ratios was higher than that of individual lipid indicators, indicating that lipid ratios were superior to individual lipid values in the diagnosis of NAFLD in newly diagnosed T2DM patients (Supplementary Figure 3). Furthermore, the results of the ROC curve analysis corresponding to the six lipid ratios were shown in Figure 2 and Table 3, with the highest AUC for the TG/HDL-C ratio (AUC 0.732; $95\%$ CI 0.696-0.769). Moreover, the sensitivity of the TG/HDL ratio ($73.8\%$) was also the highest among all six indicators, with a specificity of $60.1\%$ and a cut-off point of 1.405 (Table 3).
**Figure 2:** *ROC curves of the six lipid ratios in patients with newly diagnosed T2DM combined with NAFLD. TG/HDL-C, triglycerides to high-density lipoprotein-cholesterol ratio; TC/HDL-C, cholesterol to HDL-C ratio; FFA/HDL-C, free fatty acid to HDL-C ratio; UA/HDL-C, uric acid to HDL-C ratio; LDL-C/HDL-C, low-density lipoprotein-cholesterol to HDL-C ratio; APOB/A1, apolipoprotein B to apolipoprotein A1 ratio; ROC curves, receiver operator characteristic curves; NAFLD, non-alcoholic fatty liver disease; T2DM, type 2 diabetes mellitus.* TABLE_PLACEHOLDER:Table 3 In addition, ROC curve analysis showed that all six metrics in model 3 had the highest ability to discriminate NAFLD in newly diagnosed T2DM patients among the three models (Supplementary Figure 4). Furthermore, after correction for age, gender, current smoking status, BMI, ALT, AST, GGT, ALP, Scr, FBG, HbA1c and FINS, the AUC of the TG/HDL-C ratio in model 3 remained the largest (AUC of 0.818; $P \leq 0.001$) (Figure 3). These results suggested that the TG/HDL ratio was the most promising diagnostic indicator of NAFLD in patients with new-onset T2DM after adjusting for patient age, sex, BMI, current smoking, or biochemical values.
**Figure 3:** *ROC curves for Model 3 of the six lipid ratios in patients with newly diagnosed T2DM combined with NAFLD. Model 3 is adjusted for age, sex, current smoking, BMI, ALT, AST, GGT, ALP, Scr, FBG, HbA1c and FINS. NAFLD, non-alcoholic fatty liver disease; T2DM, type 2 diabetes mellitus; ROC curves, receiver operator characteristic curves; TG/HDL-C, triglycerides to high-density lipoprotein-cholesterol ratio; TC/HDL-C, cholesterol to HDL-C ratio; FFA/HDL-C, free fatty acid to HDL-C ratio; UA/HDL-C, uric acid to HDL-C ratio; LDL-C/HDL-C, low-density lipoprotein-cholesterol to HDL-C ratio; APOB/A1, apolipoprotein B to apolipoprotein A1 ratio.*
## Discussion
Early detection of NAFLD in patients with newly diagnosed T2DM is of great significance for the implementation of early intervention strategies. However, the invasiveness of liver biopsy or the limitations of the expertise of ultrasound technicians and ultrasound instrumentation have made it difficult to use the above screening methods widely in the general population. Recent studies have found that lipid and lipoprotein disorders promote the development and progression of NAFLD [21, 22]. Accumulating clinical evidence have indicated that dyslipidemia and lipoprotein disorders are associated with NAFLD in different populations (8–13, 16, 17), suggesting the possibility of lipid indices or lipid-lipoprotein ratios as diagnostic markers for NAFLD. In this study, we explored the efficacy of six lipid-lipoprotein ratio parameters (TG/HDL-C, TC/HDL-C, FFA/HDL-C, UA/HDL-C, LDL-C/HDL-C, APOB/A1) and their individual lipid indexes for the diagnosis of NAFLD in patients with newly diagnosed T2DM. Our study showed that all lipid-lipoprotein ratios were superior to individual lipid indices in the diagnosis of NAFLD in patients with newly-onset T2DM.
Previous studies on the correlation between lipid-lipoprotein ratios and NAFLD have been reported [9, 11, 16, 17, 23]. Ren et al. [ 11] concluded that the TC/HDL-C ratio had a significant predictive value for NAFLD, and ROC analysis showed that the AUC (0.645) was greater than other serum lipids. In addition, Zhu et al. [ 9] suggested that the predictive value of UA/HDL-C was significantly higher than LDL-C/HDL-C, non-HDL-C/HDL-C and ALT/AST in a non-obese population, even when UA and LDL-C levels were within the normal range. In a 5-year longitudinal cohort study of 9767 non-obese subjects with normal lipids, Cox proportional hazard regression model confirmed that high LDL-C/HDL-C ratios significantly increased the risk of NAFLD in non-obese Chinese subjects with normal lipids [16]. In addition, the APOB/A1 ratio was also associated with the prevalence of NAFLD in non-diabetic subjects [23], normal weight and overweight subjects [17]. Although the correlation between lipid-lipoprotein ratio and NAFLD has been reported in non-obese, non-diabetic populations, the correlation between lipid- lipoprotein ratio and NAFLD in newly diagnosed T2DM patients has not been studied.
There is growing evidence revealed a strong association between TG/HDL-C and multiple metabolic disorders, including IR, diabetes, and cardiometabolic risk [13, 18, 24]. For example, in a study investigating the relationship between lipid ratios and abnormal glucose tolerance, Guo et al. [ 18] found that serum TC, TG, TC/HDL-C, TG/HDL-C, and non-HDL-C were all strongly associated with prediabetes and T2DM. The AUC values of both TG and TG/HDL-C exceeded 0.70 in the diagnosis of prediabetes and T2DM. In addition, some studies have found a correlation between TG/HDL-C and NAFLD. For example, a retrospective study demonstrated that TG/HDL-C was independently associated with NAFLD in subjects undergoing health screening and could be used as a surrogate marker for NAFLD [12]. In another retrospective cohort study of a Chinese non-obese population without dyslipidemia, there was an independent correlation between TG/HDL-C and NAFLD [10]. Although previous studies have identified correlations between TG/HDL-C and NAFLD in physical examination subjects and non-obese populations, no study has so far focused on the diagnostic value of TG/HDL-C for NAFLD in a newly diagnosed T2DM population. Our study suggests for the first time that TG/HDL-C may be a promising biomarker for early identification of NAFLD in newly diagnosed T2DM patients. We found that in patients with newly diagnosed T2DM, TG/HDL-C had an AUC of 0.732, a sensitivity of $73.8\%$ and a specificity of $60.1\%$ for identifying NAFLD, which was significantly higher than other five lipid- lipoprotein ratios.
Our study found that TG/HDL-C might have the potential to be used as a diagnostic indicator of NAFLD in newly diagnosed T2DM. The mechanism of the intrinsic association of TG/HDL-C with T2DM combined with NAFLD may be related to IR. Previous studies have revealed the strong correlation between TG/HDL-C and IR (25–28). And the onset of NAFLD and T2DM are also closely associated with IR (28–32). Excess fatty acids are produced due to increased lipolysis and enhanced fatty acid synthesis. These fatty acids enter the blood circulation and accumulate in peripheral tissues, such as the liver and adipose tissue, ultimately leading to IR [31]. In addition, IR also enhances new lipogenesis in the liver and lipolysis in adipose tissue, thereby increasing the amount of fatty acids flowing to the liver [32]. Lipids accumulate in the liver in the form of FFA-derived TG, which together with high levels of free cholesterol and lipid metabolites (e.g., unsaturated fatty acids, lipid peroxidation products, etc.), increase lipotoxicity [32, 33]. Also, β-cell failure due to excess free fatty acids and lipid metabolites, as well as IR, are major pathogenic mechanisms of T2DM [33]. The molecular mechanisms underlying the association between TG/HDL-C and the risk of NAFLD in newly diagnosed T2DM still deserve further exploration.
There are some limitations of our study. Firstly, it is uncertain whether the TG/HDL-C ratio remains a diagnostic indicator for NAFLD in patients with longer duration of T2DM. Follow-up studies of these patients will be able to clarify this issue. Secondly, all patients with T2DM recruited in this study were newly diagnosed and had not received oral lipid-lowering or hypoglycemic medications. The strict inclusion criteria resulted in a small sample size for inclusion. Thirdly, some newly diagnosed T2DM patients were not included in this study due to the lack of liver ultrasound imaging, which may lead to some degree of data bias.
## Conclusion
In summary, this is the first study to assess the diagnostic value of multiple simple lipid parameter ratios for NAFLD in newly diagnosed T2DM patients. Our results found that the proportion of NAFLD in newly diagnosed T2DM patients elevated progressively with increasing ratios of six lipid parameters. Our study suggest that the TG/HDL-C ratio has the best diagnostic value for NAFLD in the newly diagnosed T2DM population, and may has the potential to be used as a screening marker for NAFLD in the newly diagnosed T2DM population in clinical practice and in large-scale screening.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by the Ethics Committee of Tongji Hospital, Tongji University School of Medicine (K-2021-010). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
RL and DK analyzed the patient data and drafted the manuscript. ZY and GZ contributed to data interpretation. KH, WX, PF coordinated the research. LZ and YZ contributed to data interpretation and critical revision of the manuscript. KZ and YX designed the study, revised and prepared the final version of the manuscript. All authors read and approved the final 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.2023.1127134/full#supplementary-material
## References
1. Ortiz-Lopez C, Lomonaco R, Orsak B, Finch J, Chang Z, Kochunov VG. **Prevalence of prediabetes and diabetes and metabolic profile of patients with nonalcoholic fatty liver disease (NAFLD)**. *Diabetes Care* (2012) **35**. DOI: 10.2337/dc11-1849
2. Younossi ZM, Golabi P, de Avila L, Paik JM, Srishord M, Fukui N. **The global epidemiology of NAFLD and NASH in patients with type 2 diabetes: A systematic review and meta-analysis**. *J Hepatol* (2019) **71** 793-801. DOI: 10.1016/j.jhep.2019.06.021
3. Younossi ZM, Koenig AB, Abdelatif D, Fazel Y, Henry L, Wymer M. **Global epidemiology of nonalcoholic fatty liver disease-meta-analytic assessment of prevalence, incidence, and outcomes**. *Hepatology* (2016) **64** 73-84. DOI: 10.1002/hep.28431
4. Anstee QM, Targher G, Day CP. **Progression of NAFLD to diabetes mellitus, cardiovascular disease or cirrhosis**. *Nat Rev Gastroenterol Hepatol* (2013) **10**. DOI: 10.1038/nrgastro.2013.41
5. Targher G, Bertolini L, Rodella S, Zoppini G, Lippi G, Day C. **Non-alcoholic fatty liver disease is independently associated with an increased prevalence of chronic kidney disease and proliferative/laser-treated retinopathy in type 2 diabetic patients**. *Diabetologia.* (2008) **51**. DOI: 10.1007/s00125-007-0897-4
6. 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**. *Hepatol (Baltimore Md)* (2018) **67**. DOI: 10.1002/hep.29367
7. Castera L, Friedrich-Rust M, Loomba R. **Noninvasive assessment of liver disease in patients with nonalcoholic fatty liver disease**. *Gastroenterology* (2019) **156** 1264-81.e4. DOI: 10.1053/j.gastro.2018.12.036
8. Sheng G, Lu S, Xie Q, Peng N, Kuang M, Zou Y. **The usefulness of obesity and lipid-related indices to predict the presence of non-alcoholic fatty liver disease**. *Lipids Health Dis* (2021) **20** 134. DOI: 10.1186/s12944-021-01561-2
9. Zhu W, Liang A, Shi P, Yuan S, Zhu Y, Fu J. **Higher serum uric acid to HDL-cholesterol ratio is associated with onset of non-alcoholic fatty liver disease in a non-obese Chinese population with normal blood lipid levels**. *BMC Gastroenterol* (2022) **22** 196. DOI: 10.1186/s12876-022-02263-4
10. Chen Z, Qin H, Qiu S, Chen G, Chen Y. **Correlation of triglyceride to high-density lipoprotein cholesterol ratio with nonalcoholic fatty liver disease among the non-obese Chinese population with normal blood lipid levels: A retrospective cohort research**. *Lipids Health Dis* (2019) **18** 162. DOI: 10.1186/s12944-019-1104-6
11. Ren XY, Shi D, Ding J, Cheng ZY, Li HY, Li JS. **Total cholesterol to high-density lipoprotein cholesterol ratio is a significant predictor of nonalcoholic fatty liver: Jinchang cohort study**. *Lipids Health Dis* (2019) **18** 47. DOI: 10.1186/s12944-019-0984-9
12. Fan N, Peng L, Xia Z, Zhang L, Song Z, Wang Y. **Triglycerides to high-density lipoprotein cholesterol ratio as a surrogate for nonalcoholic fatty liver disease: A cross-sectional study**. *Lipids Health Dis* (2019) **18** 39. DOI: 10.1186/s12944-019-0986-7
13. Pacifico L, Bonci E, Andreoli G, Romaggioli S, Di Miscio R, Lombardo CV. **Association of serum triglyceride-to-HDL cholesterol ratio with carotid artery intima-media thickness, insulin resistance and nonalcoholic fatty liver disease in children and adolescents**. *Nutr Metab Cardiovasc Dis* (2014) **24**. DOI: 10.1016/j.numecd.2014.01.010
14. Wong VW, Wong GL, Tsang SW, Hui AY, Chan AW, Choi PC. **Metabolic and histological features of non-alcoholic fatty liver disease patients with different serum alanine aminotransferase levels**. *Alimentary Pharmacol Ther* (2009) **29**. DOI: 10.1111/j.1365-2036.2008.03896.x
15. Zhang M, Zhou J, Liu Y, Sun X, Luo X, Han C. **Risk of type 2 diabetes mellitus associated with plasma lipid levels: The rural Chinese cohort study**. *Diabetes Res Clin Pract* (2018) **135**. DOI: 10.1016/j.diabres.2017.11.011
16. Zou Y, Zhong L, Hu C, Zhong M, Peng N, Sheng G. **LDL/HDL cholesterol ratio is associated with new-onset NAFLD in Chinese non-obese people with normal lipids: A 5-year longitudinal cohort study**. *Lipids Health Dis* (2021) **20** 28. DOI: 10.1186/s12944-021-01457-1
17. Yang MH, Sung J, Gwak GY. **The associations between apolipoprotein b, A1, and the B/A1 ratio and nonalcoholic fatty liver disease in both normal-weight and overweight Korean population**. *J Clin Lipidol* (2016) **10**. DOI: 10.1016/j.jacl.2015.11.017
18. Guo W, Qin P, Lu J, Li X, Zhu W, Xu N. **Diagnostic values and appropriate cutoff points of lipid ratios in patients with abnormal glucose tolerance status: A cross-sectional study**. *Lipids Health Dis* (2019) **18** 130. DOI: 10.1186/s12944-019-1070-z
19. Alberti KG, Zimmet PZ. **Definition, diagnosis and classification of diabetes mellitus and its complications**. *Part 1: diagnosis classification Diabetes mellitus provisional Rep WHO consultation Diabetes Med* (1998) **15**. DOI: 10.1002/(SICI)1096-9136(199807)15:7<539::AID-DIA668>3.0.CO;2-S
20. Zeng MD, Fan JG, Lu LG, Li YM, Chen CW, Wang BY. **Guidelines for the diagnosis and treatment of nonalcoholic fatty liver diseases**. *J Dig Dis* (2008) **9**. DOI: 10.1111/j.1751-2980.2008.00331.x
21. Seebacher F, Zeigerer A, Kory N, Krahmer N. **Hepatic lipid droplet homeostasis and fatty liver disease**. *Semin Cell Dev Biol* (2020) **108** 72-81. DOI: 10.1016/j.semcdb.2020.04.011
22. Yang M, Geng CA, Liu X, Guan M. **Lipid disorders in NAFLD and chronic kidney disease**. *Biomedicines* (2021) **9**. DOI: 10.3390/biomedicines9101405
23. Choe YG, Jin W, Cho YK, Chung WG, Kim HJ, Jeon WK. **Apolipoprotein B/AI ratio is independently associated with non-alcoholic fatty liver disease in nondiabetic subjects**. *J Gastroenterol Hepatol* (2013) **28**. DOI: 10.1111/jgh.12077
24. Salazar MR, Carbajal HA, Espeche WG, Leiva Sisnieguez CE, Balbín E, Dulbecco CA. **Relation among the plasma triglyceride/high-density lipoprotein cholesterol concentration ratio, insulin resistance, and associated cardio-metabolic risk factors in men and women**. *Am J Cardiol* (2012) **109**. DOI: 10.1016/j.amjcard.2012.02.016
25. Kim JS, Kang HT, Shim JY, Lee HR. **The association between the triglyceride to high-density lipoprotein cholesterol ratio with insulin resistance (HOMA-IR) in the general Korean population: Based on the national health and nutrition examination survey in 2007-2009**. *Diabetes Res Clin Pract* (2012) **97**. DOI: 10.1016/j.diabres.2012.04.022
26. Ren X, Chen ZA, Zheng S, Han T, Li Y, Liu W. **Association between triglyceride to HDL-c ratio (TG/HDL-c) and insulin resistance in Chinese patients with newly diagnosed type 2 diabetes mellitus**. *PLoS One* (2016) **11**. DOI: 10.1371/journal.pone.0154345
27. Zhou M, Zhu L, Cui X, Feng L, Zhao X, He S. **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
28. Birkenfeld AL, Shulman GI. **Nonalcoholic fatty liver disease, hepatic insulin resistance, and type 2 diabetes**. *Hepatology* (2014) **59**. DOI: 10.1002/hep.26672
29. Fujii H, Kawada N. **The role of insulin resistance and diabetes in nonalcoholic fatty liver disease**. *Int J Mol Sci* (2020) **21**. DOI: 10.3390/ijms21113863
30. Watt MJ, Miotto PM, De Nardo W, Montgomery MK. **The liver as an endocrine organ-linking NAFLD and insulin resistance**. *Endocr Rev* (2019) **40**. DOI: 10.1210/er.2019-00034
31. Kitade H, Chen G, Ni Y, Ota T. **Nonalcoholic fatty liver disease and insulin resistance: New insights and potential new treatments**. *Nutrients* (2017) **9**. DOI: 10.3390/nu9040387
32. Buzzetti E, Pinzani M, Tsochatzis EA. **The multiple-hit pathogenesis of non-alcoholic fatty liver disease (NAFLD)**. *Metabolism* (2016) **65**. DOI: 10.1016/j.metabol.2015.12.012
33. Cusi K. **Role of insulin resistance and lipotoxicity in non-alcoholic steatohepatitis**. *Clin Liver Dis* (2009) **13**. DOI: 10.1016/j.cld.2009.07.009
|
---
title: 'Impact of feeding habits on the development of language-specific processing
of phonemes in brain: An event-related potentials study'
authors:
- Graciela C. Alatorre-Cruz
- Aline Andres
- Yuyuan Gu
- Heather Downs
- Darcy Hagood
- Seth T. Sorensen
- David Keith Williams
- Linda J. Larson-Prior
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9982124
doi: 10.3389/fnut.2023.1032413
license: CC BY 4.0
---
# Impact of feeding habits on the development of language-specific processing of phonemes in brain: An event-related potentials study
## Abstract
### Introduction
Infancy is a stage characterized by multiple brain and cognitive changes. In a short time, infants must consolidate a new brain network and develop two important properties for speech comprehension: phonemic normalization and categorical perception. Recent studies have described diet as an essential factor in normal language development, reporting that breastfed infants show an earlier brain maturity and thus a faster cognitive development. Few studies have described a long-term effect of diet on phonological perception.
### Methods
To explore that effect, we compared the event-related potentials (ERPs) collected during an oddball paradigm (frequent /pa/$80\%$, deviant/ba/$20\%$) of infants fed with breast milk (BF), cow-milk-based formula (MF), and soy-based formula (SF), which were assessed at 3, 6, 9, 12, and 24 months of age [Mean across all age groups: 127 BF infants, Mean (M) 39.6 gestation weeks; 121 MF infants, $M = 39.16$ gestation weeks; 116 SF infants, $M = 39.16$ gestation weeks].
### Results
Behavioral differences between dietary groups in acoustic comprehension were observed at 24-months of age. The BF group displayed greater scores than the MF and SF groups. In phonological discrimination task, the ERPs analyses showed that SF group had an electrophysiological pattern associated with difficulties in phonological-stimulus awareness [mismatch negativity (MMN)-2 latency in frontal left regions of interest (ROI) and longer MMN-2 latency in temporal right ROI] and less brain maturity than BF and MF groups. The SF group displayed more right-lateralized brain recruitment in phonological processing at 12-months old.
### Discussion
We conclude that using soy-based formula in a prolonged and frequent manner might trigger a language development different from that observed in the BF or MF groups. The soy-based formula’s composition might affect frontal left-brain area development, which is a nodal brain region in phonological-stimuli awareness.
## Introduction
In infant development, the brain undergoes multiple changes, including increased myelination and configuration of synaptic connections needed to consolidate new brain networks. Volumetric brain growth, which proceeds through infancy, reaches adult levels at 3 years old. These changes are promoted by environmental stimuli [1], hormonal status and genetic factors [2]. Moreover, infant diet has been recently recognized as an important contributor to cognitive development, immune system development, and healthy physical growth (3–6). To support infant development, the diet should provide micro and macronutrients such as docosahexaenoic (DHA) and arachidonic (AA), long-chain fatty acids, lutein, choline, and hormones (7–9). Human milk provides these essential nutrients [5, 6] and promotes greater brain maturity characterized by healthier neuronal growth and myelination, and greater infant gray and white matter (8, 10–12).
Some studies report that breastfed infants show an earlier development of language perception (10, 13–15) and memory than those fed with nutrient-enriched formula [16]. An explanation for this finding is that human milk changes in composition from colostrum to late lactation, and varies by the mother’s biological condition, while milk-based formula maintains a stable composition [5, 9, 10]. In particular, human milk seems to have a better nutritional composition than milk-based formula because: [1] complex oligosaccharides or lipid components such as gangliosides found in human milk are not available in milk-based formula composition or have not been clinically proven [17], and [2] the equilibrium in human milk’s composition between DHA, lutein, choline [10, 16], and complex oligosaccharides seems to promote better cognition [18, 19]. Therefore, differences in the proportions of formula components may negatively affect infant’s nutrition, and consequently, the infant’s cognitive development [20].
In the first year of life, phonological perception should be developed; otherwise, the infant will suffer delayed language development [21]. This milestone entails the fast growth of multiple brain areas regulated by healthy nutritional habits, particularly microbiome is essential in synaptogenesis and metabolic brain requirements, affecting infant brain development and behavior [22, 23]. Moreover, recent studies suggest that breastfeeding positively affects cognition and brain development compared with other feeding habits (24–28). They explain that this effect occurs because four reasons: [1] human milk might make a difference in brain structure and function via fatty acids, affecting cell membranes and influencing gene expression within these cells, [2] human milk contains a variety of constituents that promote optimal development, [3] the relationship between the immune system and breastfeeding might influence learning and memory, and [4] lactation affects mothers’ way they teach the language [29].
Even in utero, infants are able to distinguish between sounds (30–32) and show habituation to repetitive stimuli [30]. However, they must develop other abilities to reach adult levels of phonological perception. Within the first 2 months of life, normal infants show a precognitive detection of syllable length [33]; at 4 months they begin to distinguish between tones and syllables [34]. At 6 months old, they establish prototypes of vowels in their native language [31, 32], and around 10 months old, infants have prototypes of consonants [35]. Between 9 and 10 months of age, infants can distinguish words [36], and preserve the detection of foreign-language contrast until 11 months old [37]. By the end of the first year of life, they have access to phonological representations akin to those of adults, that is, the infant has developed two important properties for speech comprehension: [1] phonemic normalization and [2] categorical perception [30]. These subtle behavioral changes are accompanied by the recruitment of frontal and temporal lobes responsible for phonological perception and semantic categorization, which are differentially developed in the first year of life [38, 39]. In infants, brain maturity is reflected in the decrease of bilateral brain responses and increase in left lateralization of brain activity (40–42), culminating in the development of the adult pattern of dorsal and ventral pathways associated with language function (43–45).
Accordingly, brain-electrical activity associated with phonological perception also develops during infancy, reflecting the increasing ability to decode incoming speech supported by the accurate perception of rapid acoustic changes [21, 46, 47]. The brain-electrical response to auditory tones [event-related potentials (ERPs)] in adults comprises the P1-N1-P2-N2 complex (48–50), and includes [1] a positive deflection at 150 ms on fronto-central sites (P150 or P1), which has been associated with features of acoustic stimulus (51–53) and modulated by an inter-stimulus interval [54]; [2] a negative deflection at 250 ms (N250 or N1) and another at 450 ms (N450 or N2), which reflect the differences between acoustic stimulus (e.g., such as complexity and frequency) [54], and [3] a positive wave at 350 ms (P350 or P2) which has been associated with stimulus awareness and perceptual salience, and is commonly identified as an index of auditory recognition memory [51]. However, these ERP components are not exhibited at birth, but develop gradually over the first year of life. At birth, infants display a large positive wave between 100 and 450 ms followed by the N2 component [49]. At 3 months old, the positive wave is divided by the N1 component between 160 and 200 ms [49, 55], resulting in two ERP components: a P1 and P2, and the amplitude of these components seem to increase over the next month [49, 54]. Between 3 and 6 months the amplitude of N1 and N2 components increase [56], and exhibit the P1-N1-P2-N2 complex. This ERP morphology is maintained until 2 years of age [49, 57], with an increase in component amplitude exhibited at 12-months old [49]. Although few studies describe the functional significance of these ERP components in infancy, it has been speculated they have similar function to those observed in children and adults [53].
Development of phonemic perception requires infants to detect differences in acoustic features and phonological categories, leading to the use of experimental auditory oddball paradigms in which stimuli including differences between acoustic features, frequency or phonological categories are especially useful in assessing brain electrical activity associated with the acquisition of language (32–34, 58–61). From studies in children and adults, the expectation is that the amplitude of P1-N1-P2-N2 complex will be greater for uncommon than common repetitive stimuli [50], due to the fact that neuronal responses habituate to repeated presentation of the same stimulus, while a new, unusual stimulus will produce a large amplitude response [30, 62]. The difference between the conditions is called mismatch negativity (MMN) [30, 63, 64]. It has been reported that two MMNs which appear at 6 months [50, 54], correspond to the differences in P1 and P2 components [65]. As described above, the MMN components undergo latency decreases with increasing age [50]. In addition, the MMN components have been linked to the computation of acoustic features such as duration or intensity [66, 67], arbitrary rules [68], or lexical and grammatical status [58, 69], and their interpretation depends on the specific stimulus type presented.
While few studies have assessed how diet affects phonetic perception; those that did have shown variations on this cognitive process by diet. Li et al. [ 13] compared breastfed infants and infants fed with soy or cow-milk-based formula in their phonological perception at 3 and 6 months, using an oddball paradigm compromised of frequent and deviant syllables (/pa/standard and/ba/deviant). The authors found an advanced neural maturation in breastfed infants characterized by a greater P350/P2 amplitude in frontal regions at 3 months, and shorter P2 latency at 6 than 3 months old than the other dietary groups. Using the same paradigm, Pivik et al. [ 3] compared these same dietary groups and ages. However, they did not replicate the findings of Li et al. [ 13], reporting no age-related changes in ERP components. In this study, differences were related only to diet group, with breastfed infants displaying shorter P1 latencies and smaller P1 amplitude for deviant rather than standard stimuli than infants fed with soy milk. The authors interpreted that to indicate that breast-fed infant show more rapid encoding of acoustic information than the other diet groups. The same diet groups were also studied at 4 and 5 months [14], where changes in P350/P2 amplitude across age for each syllable, depended on the diet. Infants fed with soy milk showed a decreased P2 amplitude for deviant stimuli than the other groups, while the breastfed infants displayed decreased amplitude for standard stimuli compared with other dietary groups. The authors concluded that diet affects attention and memory functions involved in the processing and discrimination of speech sounds.
The primary aim of the present study was to determine the differences in phonological perception assessed by electrophysiological response to frequent and deviant phonemes at 3, 6, 9, 12, and 24 months between three dietary groups: breast fed (BF), cow-milk-formula fed (MF), and soy-formula fed (SF) infants. As previous studies have reported evidence for earlier phonological perception in BF infants [13, 14, 65], we anticipated that the BF group would show [1] greater amplitude and shorter latency of MMN components than MF and SF groups, [2] greater amplitude and shorter latency of MMN components [13, 65] at 6 month-old when the P1-N1-P2-N2 complex reaches a stable morphology [49, 57], and at 12 months when ERP amplitudes have a stable morphology [49] and [3] greater hemispheric asymmetry of MMN components [40].
## Participants
The study included full-term infants between 3 and 24 months old. All of them had a birth weight of over 3 kg and were a product of uncomplicated pregnancies; the mothers reported no medical diagnoses during pregnancy or lactation. Mothers with alcohol, tobacco, or medication use were excluded. In this longitudinal study, 2-month-old infants were stabilized on one of three diets which were selected by parents: BF, MF, and SF, the two last fortified with DHA and AA. Each infant was provided the same diet until 12 months of age. The infants were assessed at 3, 6, 9, 12, and 24 months old, resulting in 15 groups of data (e.g., subjects aged at three-months-old distributed into three groups: BF, MF, and SF). Socioeconomic status [SES, measured by the Four-Factor Index of Social Positions [70]] of the infants’ parents was collected at the beginning of this study. The infants’ anthropometric measures (i.e., height, weight, and head circumference) and food intake history were collected at each visit. Infants and mothers underwent neuropsychological and psychophysiological testing, which was conducted by a certified examiner. The mother’s assessment included Wechsler the Abbreviated Scale of Intelligence [WASI-II, [71]] and Symptoms Assessment-45 questionnaire [SA-45, [72]], while infants were evaluated using the Bayley Scales of Infant and Toddler Development [BSID-2, [73]], Preschool Language Scale [PLS-3, [74]] as well as the psychophysiological oddball paradigm to assess phonological-discrimination. Most of the parents reported English as their language at home (see Table 1). All mothers reached an Intelligence quotient (IQ) score higher than 70 on the WASI-II test. Participants were excluded from this study if they did not complete all assessments. The protocol was approved by the Institutional Review Board of the University of Arkansas for Medical Sciences. Informed consent was obtained from parents.
**Table 1**
| Age | Total (n) | Type of diet (n) | Mother IQ mean (SD) | Language in home |
| --- | --- | --- | --- | --- |
| 3 m | 410 | BF: 137 | BF: 109.4 (10.2) | E (404) |
| 3 m | 410 | MF:138 | MF: 105.8 (9.2) | S (1) |
| 3 m | 410 | SF: 135 | SF: 102.9 (11.7) | E,S (4) |
| 3 m | 410 | | | E,O (1) |
| 6 m | 365 | BF: 119 | BF: 110.1 (10.1) | E (356) |
| 6 m | 365 | MF:126 | MF: 105.7 (9.3) | E,S (7) |
| 6 m | 365 | SF: 120 | SF: 103.3 (10.0) | E,O (2) |
| 9 m | 340 | BF: 113 | BF: 109.6 (10.5) | E (333) |
| 9 m | 340 | MF:114 | MF: 105.4 (8.7) | S (1) |
| 9 m | 340 | SF: 113 | SF: 103.8 (10.3) | E,S (5) |
| 9 m | 340 | | | E,O (1) |
| 12 m | 334 | BF: 122 | BF: 109.7 (10.3) | E (318) |
| 12 m | 334 | MF:112 | MF: 105.0 (8.8) | S (2) |
| 12 m | 334 | SF: 100 | SF: 103.2 (10.3) | E,S (10) |
| 12 m | 334 | | | E,O (4) |
| 24 m | 372 | BF: 142 | BF: 109.6 (10.4) | E (361) |
| 24 m | 372 | MF:117 | MF: 105.8 (9.1) | S (1) |
| 24 m | 372 | SF: 113 | SF: 104.5 (10.7) | E,S (7) |
| 24 m | 372 | | | E,O (3) |
## Experimental design
Phonological discrimination was assessed using an auditory-oddball paradigm while an electroencephalogram (EEG) was recorded. The infants were awake and seated in their parent’s lap or infant chair in a sound-isolated, shielded recording chamber. Silent videos were played to engage the infant’s attention. The paradigm was constituted of two types of stimuli, one of them was frequent (/pa/$80\%$) and the other deviant (/ba/$20\%$). Both stimuli were syllables with consonant-vowel structure, had the same intensity (72 dB SPL), and were pronounced by a native English speaker through speakers located at 5 ft. from the infant. The stimuli were designed and administered using E-Prime software (version 1). All stimuli appeared during 300 ms with a stimulus onset asynchrony (SOA) of 2,500 ms. The SOA was selected because longer intervals attenuate standard-deviant response differences [75] and exceeds the limits of sensory memory reported for infants [76]. The task included three blocks of 90 trials for a total of 270 trials. The deviant stimuli (/ba/) randomly appeared in each block with a probability of 0.2. Each block lasted 4.2 min. The infants had two rest periods of 5 min between experimental blocks (see Figure 1).
**Figure 1:** *Oddball paradigm applied to infants.*
## Data acquisition/prep-processing
The EEG was acquired with a Geodesic Net Amps 200 system running Netstation 2 software using the 128-channel (Electrical Geodesics, Inc., Eugene OR, United States). Data were amplified with a bandpass of 0.1–100 Hz and a sampling rate of 250 Hz. Electrode impedances were kept below 40 kΩ. Eye movements and blinks were monitored. Data were analyzed offline using the Matlab toolbox (Matlab version R2020a). The EEG was segmented into epochs with a 100 ms pre-stimulus baseline and 1,000 ms stimulus/post-stimulus. The epochs were subjected to an automatic artifact detection algorithm. Bad channels (i.e., channels with fast average amplitude greater than 200 μV or/and differential average greater than 100 μV) were interpolated from nearby good channels using spherical splines. Trials with more than 10 bad channels were excluded. The accepted segments for each type of condition (/ba/or/pa/) were baseline corrected using a 100 ms pre-stimulus time window, re-referenced to the common mean, and averaged for each participant. The accepted segments were at least 35 per condition for each participant.
## Event-related potentials
The average epoch for each condition per subject was obtained in four regions of interest (ROIs): Frontal Left (FL; sensors 28, 34, and 35) and Right (FR; sensors 117, 122, and 123), Temporal Left (TL; sensors 42, 47, and 48) and Right (TR; sensors 99, 103, and 104) (see Figure 2). Then, the difference wave was calculated in each ROI by subtracting the epoch associated with the frequent stimulus (/pa/) from that related to the deviant stimulus (/ba/). The grand average of difference wave was inspected in accordance with the ERP literature associated with phonological perception [30, 54, 63, 64]. Two ERPs components were identified, two mismatch negativities; the first between 75 and 255 ms (MMN-1), and the second between 300 and 500 ms (MMN-2), the first functionally associated with the P1 component and the second with the P2 component.
**Figure 2:** *On the top, the regions of interest (ROIs) used for amplitude and latency analyses of ERP components. On the bottom, the grand average of ERPs of frequent “pa” and deviant “ba” conditions for each dietary group (BF, breast feed; MF, milk feed; SF, soy feed) at 3, 6, 9, 12, and 24-months old. The positive or negative event-related potentials (ERP) components were highlighted as follows: P1-N1, P2, and N2.*
## Parental data
Parental SES in each age group (i.e., 3, 6, 9, 12, and 24-months old) was compared using one-way ANOVA. For both comparisons, dietary group (i.e., BF, MF, and SF) was included as a between-subjects factor, and total SES index was included as within-subject factors.
Maternal psychometric and psychiatric data: Psychometric and psychiatric test results were analyzed using two-way ANOVAs for each assessment (i.e., WASI-II and SA-45) and each age group. The dietary group was included as a between-subjects factor, and the within-subject factors were as follows: We observed a significant main effect of the dietary group at 3-months [F[2,398] 3.5, $$p \leq 0.03$$], and 6-months of age [F[2,361] 5.4, $$p \leq 0.005$$]. The post hoc tests showed that the BF group displayed a greater parental SES score than SF group at 3-months old [Mean difference (MD) = −2.7, $$p \leq 0.03$$; BF, Mean (M) 39.8; MF, $M = 38.4$; SF, $M = 37.0$], while at 6-months old, SF group was significantly different than BF and MF groups, displaying a lower parental SES score than the other groups (SF vs. BF, MD = −3.2, $$p \leq 0.008$$; SF vs. MF, MD = −2.8, $$p \leq 0.02$$; BF, $M = 40.0$; MF, $M = 39.6$; SF, $M = 36.8$).
The dietary groups differed in maternal WASI-II indices. In all comparisons, the post hoc tests showed that mothers from the BF group had greater WASI-II indices than mothers in the other dietary groups [3 months (m), BF vs. MF, MD = 3.1, $$p \leq 0.02$$; BF vs. SF, MD = 5.8, $p \leq 0.001$; 6 m, BF vs. MF, MD = 3.9, $$p \leq 0.002$$; BF vs. SF, MD = 6.1, $p \leq 0.001$; 9 m, BF vs. MF, MD = 3.7, $$p \leq 0.007$$; BF vs. SF, MD = 5.2, $p \leq 0.001$; 12 m, BF vs. MF, MD = 3.9, $$p \leq 0.003$$; BF vs. SF, MD = 5.7, $p \leq 0.001$; 24 m, BF vs. MF, MD = 3.3, $$p \leq 0.01$$; BF vs. SF, MD = 4.5, $p \leq 0.001$]. No significant dietary group by WASI-II indices interaction was found in any comparison (see Figure 3A).
**Figure 3:** *Differences between dietary groups in neuropsychological and psychophysiological assessment. In (A), the bar graph illustrates differences between Infant’s moms in Wechsler Abbreviated Scale of Intelligence (WASI-II). The moms from BF group showed greater WASI-II indices than the remaining dietary groups. In (B), the bar graph shows differences between the dietary groups in Bayley Scales of Infant and Toddler Development (BSID-2) at 9-months old, BF groups displayed greater BSID-2 scores than MF and SF groups, while in (C), the bar graph illustrates differences between dietary groups in Preschool Language Scale (PLS-3) at 24-months old. BF infants showed greater AC score than MF group. Significant value of ps has been represented as follows: *p < 0.05, **p < 0.01.*
Although no significant main effect of dietary group was observed in maternal SA-45 indices, a significant dietary group by SA-45 interaction was found at 12 months [F[2,326] 6.3, $$p \leq 0.002$$, Ƞ2 = 0.04, ε = 1]. However, the post hoc tests showed no significant differences between dietary groups in any SA-45 index (BF: GSI, $M = 45.6$; PST, $M = 45.1$; MF: GSI, $M = 45.6$; PST, $M = 44.3$; SF: GSI, $M = 45.7$; PST, $M = 45.4$).
## Infants’ anthropometric and psychometric data
Birth data and anthropometric measures were compared at 3, 6, 9, 12, and 24-months old using one-way ANOVA. For both comparisons, the dietary group was included as a between-subjects factor, and gestational age, birth length, birth weight, height, weight, and head circumference were separately included as between-subjects factors. A chi-squared test was used to compare groups for infant’s sex distribution.
Psychometric test results were analyzed using two-way ANOVAs for each neuropsychological assessment (i.e., BSID-2 and PLS-3 tests) and for each age group. The dietary group was included as a between-subjects factor, and within-subject factors are as follows:
## Infant’ amplitude and latency analyses of ERPs
Comparisons between dietary groups for each age group: We considered MMN-1 and MMN-2 components for the statistical analyses. We calculated the mean amplitude and its latency (i.e., the maximal peak of time window) for each ERP component. Then, we separately compared the amplitude and latency of each ERP component. ANCOVAs were also separately computed for each age group. The dietary group was the between-subject factor, FL, FR, TL, and TR ROIs were included as the within-subject factors, and gestational weeks and infant’s sex as covariables.
We also assessed the hemispheric asymmetry of ERPs components, ANCOVAs were separately computed for the difference in amplitude or latency of ERPs between brain hemispheres in frontal or temporal regions (e.g., MMN-1 amplitude in frontal left ROI minus MMN-1 amplitude in frontal right ROI). The dietary group was the between-subject factor, frontal and temporal ROIs were included as the within-subject factors, and gestational weeks and infant’s sex as covariables.
Comparisons between age groups for each dietary group: ANCOVAs were separately performed for the amplitude or latency of each ERP component and each dietary group. The age group (3, 6, 9, 12, and 24 months) was the between-subject factor, FL, FR, TL, and TR ROIs were included as the within-subject, and gestation weeks and infant’s sex as covariables. Data were analyzed using SPSS Statistics 20 and Matlab (version R2020a). Greenhouse–Geisser corrections were made for violations of sphericity when the numerator was greater than 1. value of ps resulting from a set of comparisons were corrected by the FDR method. We report results surviving FDR correction (p-values <0.05).
## Regression analyses
Regression analyses were performed to identify the association between amplitude and latency of ERP components in each ROI, that differed between dietary groups, and those factors that might explain the variability in the brain-electrical activity. The linear regression included amplitude or latency in FL, FR, TL or TR ROIs as the dependent variables, with dietary group (i.e., BF, MF, and SF), mom’s cognitive and psychiatric status (WASI-II: PRI and VCI; SA-45: GSI and PST), gestation weeks, infant’s sex, PLS-3: AC and EC subscales as the independent variables. Linear regressions were performed by age group. The linear regression analyses included multiple-linear backward regressions to find a reduced model that best explains the data.
Regression analyses were also performed to identify the association between the hemispheric asymmetry of ERPs components and other variables. Hemispheric differences in frontal or temporal regions were included as dependent variables, and the independent variables were dietary group (i.e., BF, MF, and SF), mom’s cognitive and psychiatric status (WASI-II: PRI and VCI; SA-45: GSI and PST), gestation weeks, infant’s sex, PLS-3: AC and EC subscales. Linear regressions were performed by age group. The linear regression analyses included multiple-linear backward regressions to find a reduced model that best explains the data. Factors with the highest value of p were removed until all factors were statistically significant. A value of $p \leq 0.05$ was considered statistically significant in all analyses.
## Anthropometric data
As is shown in Table 2, gestational weeks differed between groups in all age groups, the post hoc tests evinced that the BF group had greater gestation weeks than the other dietary groups. The dietary groups also differed in birth weight at 6 months old. The post hoc test showed greater birth weight for BF than SF group (MD = 0.1, $$p \leq 0.02$$).
**Table 2**
| Age | Variables | Dietary group | Dietary group.1 | Dietary group.2 | Main effect of group | Main effect of group.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Age | Variables | BF | MFMean (SD) | SF | F | p |
| 3 m | Gestation (weeks) | 39.5 (1.0) | 39.1 (0.9) | 39.1 (1.0) | 6.9 | 0.001** |
| 6 m | Gestation (weeks) | 39.6 (1.0) | 39.1 (0.9) | 39.2 (1.0) | 7.2 | 0.001** |
| 9 m | Gestation (weeks) | 39.6 (1.0) | 39.2 (0.9) | 39.2 (1.0) | 6.3 | 0.002** |
| 12 m | Gestation (weeks) | 39.5 (1.0) | 39.2 (0.9) | 39.2 (1.1) | 5.3 | 0.005** |
| 24 m | Gestation (weeks) | 39.6 (1.0) | 39.2 (0.9) | 39.1 (1.0) | 8.8 | 0.000** |
| 3 m | Birth weight (kg) | 3.5 (0.3) | 3.5 (0.4) | 3.4 (0.4) | 2.9 | 0.06 |
| 6 m | Birth weight (kg) | 3.5 (0.3) | 3.5 (0.4) | 3.4 (0.4) | 3.6 | 0.03* |
| 9 m | Birth weight (kg) | 3.5 (0.3) | 3.5 (0.4) | 3.5 (0.4) | 1.4 | 0.2 |
| 12 m | Birth weight (kg) | 3.5 (0.3) | 3.5 (0.4) | 3.4 (0.4) | 1.8 | 0.2 |
| 24 m | Birth weight (kg) | 3.5 (0.3) | 3.5 (0.4) | 3.4 (0.4) | 2.1 | 0.1 |
| 3 m | Birth length (cm) | 51.4 (2.2) | 51.1 (2.5) | 51.2 (2.2) | 0.9 | 0.4 |
| 6 m | Birth length (cm) | 51.5 (2.3) | 51.2 (2.4) | 51.2 (2.1) | 0.6 | 0.5 |
| 9 m | Birth length (cm) | 51.5 (2.1) | 51.2 (2.7) | 51.2 (2.0) | 0.9 | 0.4 |
| 12 m | Birth length (cm) | 51.4 (2.1) | 51.1 (2.5) | 51.2 (2.1) | 0.5 | 0.6 |
| 24 m | Birth length (cm) | 51.4 (2.2) | 51.2 (2.4) | 51.0 (2.10) | 0.9 | 0.4 |
| 3 m | Height (cm) | 60.1 (1.9) | 58.9 (2.1) | 59.7 (1.6) | 1.5 | 0.2 |
| 6 m | Height (cm) | 66.0 (2.3) | 66.4 (2.4) | 66.4 (2.0) | 1.5 | 0.2 |
| 9 m | Height (cm) | 69.9 (2.3) | 70.7 (2.4) | 70.8 (2.3) | 5.2 | 0.006** |
| 12 m | Height (cm) | 73.7 (2.3) | 74.4 (2.3) | 75.0 (2.3) | 8.3 | 0.000** |
| 24 m | Height (cm) | 86.2 (2.5) | 86.3 (2.8) | 86.7 (2.8) | 1.3 | 0.3 |
| 3 m | Weight (kg) | 6.1 (0.7) | 6.1 (0.6) | 5.9 (0.5) | 2.2 | 0.8 |
| 6 m | Weight (kg) | 7.6 (0.9) | 7.9 (0.9) | 7.8 (0.8) | 2.8 | 0.06 |
| 9 m | Weight (kg) | 8.6 (0.8) | 9.0 (0.9) | 9.1 (1.0) | 10.8 | 0.000** |
| 12 m | Weight (kg) | 9.4 (1.0) | 9.9 (0.9) | 10.0 (1.0) | 13.3 | 0.000** |
| 24 m | Weight (kg) | 12.2 (1.2) | 12.4 (1.2) | 12.7 (1.3) | 6.3 | 0.002** |
| 3 m | Head circ. (cm) | 40.7 (1.1) | 40.6 (1.1) | 40.6 (1.1) | 0.2 | 0.8 |
| 6 m | Head circ. (cm) | 43.4 (1.3) | 43.6 (1.1) | 43.7 (1.3) | 1.6 | 0.2 |
| 9 m | Head circ. (cm) | 45.2 (1.2) | 45.2 (1.2) | 45.4 (1.3) | 1.1 | 0.3 |
| 12 m | Head circ. (cm) | 46.4 (1.3) | 46.4 (1.2) | 46.6 (1.4) | 1.6 | 0.2 |
| 24 m | Head circ. (cm) | 48.7 (1.3) | 48.8 (1.3) | 48.9 (1.4) | 0.7 | 0.5 |
| | | | F/M | | χ2 (2) | p |
| 3 m | Sex | 68/69 | 71/67 | 62/73 | 0.9 | 0.6 |
| 6 m | Sex | 64/55 | 55/65 | 62/64 | 1.5 | 0.5 |
| 9 m | Sex | 58/55 | 55/59 | 49/64 | 1.5 | 0.5 |
| 12 m | Sex | 66/56 | 56/56 | 43/57 | 2.7 | 0.2 |
| 24 m | Sex | 74/68 | 56/61 | 57/56 | 0.5 | 0.8 |
The height and weight differed between dietary groups at 9 and 12 months old, with post hoc tests showing lower height and weight for BF infants. The comparison also revealed differences between dietary groups in weight at 24 months old, with BF infants showing lower weight than SF group (MD = −0.5, $$p \leq 0.001$$). No differences between dietary groups were found in birth length, head circumference or infant’s sex in any age group.
## Psychometric data
Consistent with a previous behavioral study comparing these same dietary groups [4], no differences in MDI and PDI indexes of BSID-2 test were found at 3, 12, or 24 months old. The dietary groups only differed in BSID-2 indexes at 9-months old [F[2,331] 3.6, $$p \leq 0.03$$, Ƞ2 = 0.02] (see Figure 3B). The post hoc tests showed that the BF group displayed greater BSID-2 indexes than SF and MF groups (BF vs. MF, MD = 1.5, $$p \leq 0.04$$; BF vs. SF, MD = 1.8, $$p \leq 0.01$$). No significant dietary group by BSID-2 indexes interaction was found in any comparison. No significant main effect of group was observed in PLS-3 test in any age group. However, a significant dietary group by PLS-3 interaction was found at 24 months old [F[2,336] 3.4, $$p \leq 0.03$$, Ƞ2 = 0.02, ε = 1]. The post hoc tests revealed that the BF group displayed a greater AC score than the MF group (MD = 4.8, $$p \leq 0.01$$; see Figure 3C).
## Infant’s ERPs analysis
Comparisons between dietary groups for each age group.
## MMN-1 component
The dietary groups did not differ in amplitude or latency of the MMN-1 component at any age group. No significant main effect of dietary group or dietary group by ROIs interactions were observed in any comparison (see Supplementary Table S1).
The statistical analyses evinced no differences between dietary groups in hemispheric asymmetry of MMN-1 component.
Although age groups did not differ in MMN-1 amplitude, they did differ in MMN-1 latency, where a significant age group by ROI interaction was observed [F[12,1792] 2.08, $$p \leq 0.02$$, Ƞ2 = 0.01, ε = 0.96]. The post hoc test evinced differences between age groups in frontal right and temporal right ROIs. In frontal right ROI, the 3-month-old infants displayed shorter MMN-1 latency than 12-months-old participants, while in temporal right ROI, infants at 24 months of age had shorter MMN-1 latency than those participants at 3, 6, and 9 months old (see Supplementary Figure S1).
The age groups differed in amplitude and latency of MMN-1 component in the MF group. A significant age group by ROI interaction was seen for MMN-1 amplitude [F[12,1,588] 2.39, $$p \leq 0.007$$, Ƞ2 = 0.02, ε = 0.89]. The post hoc tests showed that age groups differed in MMN-1 amplitude all ROIs (i.e., frontal left, frontal right, temporal left, and temporal right). In the frontal left ROI, infants at 3 months of age displayed smaller MMN-1 amplitude than 24-month-old participants, while in frontal right ROI, infants at 6 months of age displayed smaller MMN-1 amplitude compared to 9-and 12-month-old infants (See Supplementary Figure S2). In temporal left ROI, a smaller MMN-1 amplitude was observed in 3-month-old infants compared to 6-month-old participants. In addition, participants at 9 and 12 months of age displayed a smaller MMN-1 amplitude than 6-monts-old infants. In temporal right ROI, infants at 24 months of age displayed a smaller MMN-1 amplitude than those infants at 3 and 9 months old (see Supplementary Table S4).
In the comparisons between age groups in MMN-1 latency, we also found a significant age group by ROIs interaction [F[12,1705] 2.05, $$p \leq 0.02$$, Ƞ2 = 0.01, ε = 0.96]. The post hoc tests evidenced differences between groups in frontal left, frontal right, and temporal right ROIs. In frontal left ROI, shorter MMN-1 latency was seen in infants at 3 months of age compared to the participants at 6 and 12 months old, while in frontal right a similar pattern was observed, infants at 3 months of age displayed shorter MMN-1 latency than 24-month-old infants. In temporal right ROI, the 9-month-old infants displayed longer MMN-1 latency than infants at 3 months old, and shorter MMN-1 latency compared to 24-month-old infants.
There were no differences between age groups in amplitude or latency of MMN-1 component.
## MMN-2 component
The dietary groups did not differ in MMN-2 amplitude. However, differences between dietary groups were observed in MMN-2 latency at 12 months old (see Supplementary Table S2). As shown in Figure 4, at 12 months of age a significant dietary group by ROI was found [F[6,981] 3.1, $$p \leq 0.006$$, Ƞ2 = 0.02, ε = 0.9]. The post hoc test showed that the SF group differed from the remaining groups in MMN-2 latency in frontal left and temporal right ROIs. The SF group displayed shorter MMN-2 latency than BF and MF groups in frontal left ROI (SF vs. BF, MD = −23.8, $$p \leq 0.004$$; SF vs. MF, MD = −27.7, $$p \leq 0.001$$), while in temporal right ROI, SF group showed longer MMN-2 latency than MF group (SF vs. MF, MD = 21.7, $$p \leq 0.02$$).
**Figure 4:** *Differences between dietary groups in MMN-2 latency at 12-months old. On top, the grand average of difference wave of event-related potentials (ERPs) in the frontal left (FL) and temporal right (TR) regions of interest (ROIs) for each dietary group at 12-months old. The bar graph shows differences in MMN-2 latency between the dietary group FL and TR ROIs on the bottom. In FL ROI, the SF group displayed shorter MMN-2 latency than the other groups, while in TR ROI, the SF group showed longer MMN-2 latency than the MF group. Significant value of ps has been represented as follows: *p < 0.05, **p < 0.01.*
Although the weight groups did not differ in hemispheric asymmetry of MMN-1 amplitude, they differed in MMN-2 latency at 12 months old (see Supplementary Table S3). The post hoc test showed significant differences between MF and SF groups (MD = 27.19, $$p \leq 0.002$$; MF, $M = 13.72$; SF, M = −13.56; see Figure 5). The MF infants displayed greater MMN-2 latency in left than right hemisphere, while SF group displayed the inverse pattern (MF: left, $M = 416.78$ ms, right, $M = 404.59$ ms; SF: left, $M = 403.68$ ms, right, $M = 416.64$ ms). No significant dietary group by hemispheric asymmetry of ERP component was observed.
**Figure 5:** *Differences between dietary groups in hemispheric asymmetry of event-related potentials (ERPs) components. The scatter plot illustrates differences between dietary groups in the hemispheric asymmetry of MMN-2 latency, which were observed at 12-months of age. Significant value of ps has been represented as follows: *p < 0.05.*
Comparisons between age groups for each dietary group.
The age groups differed in both amplitude and latency of the MMN-2 component. A main effect of age was observed in MMN-2 amplitude [F[4,623] 3.29, $$p \leq 0.01$$, Ƞ2 = 0.02, ε = 0.92]. The post hoc test revealed smaller MMN-2 amplitude in infants at 3 compared to 24 months old. The 6-month-old infants also displayed smaller amplitude than the participants at 9, 12, and 24 months old (see Supplementary Table S4).
A significant age group by ROIs interaction was also seen [F[12,1719] 2.86, $$p \leq 0.001$$, Ƞ2 = 0.02, ε = 0.92]. The post hoc tests showed that age groups were different in frontal left, frontal right, and temporal right ROIs. In both frontal left and right ROIs, 3-month-old infants displayed smaller MMN-2 amplitude than infants at 12 and 24 months. We also observed that 6-month-old participants showed smaller MMN-2 amplitude than participants at 9, 12, and 24 months old in both left and right frontal ROIs, while in temporal right ROI, a greater MMN-2 amplitude was observed in 3-month-old infants compared to participants at 24 months.
The differences between age groups in MMN-2 latency were observed regardless of ROI, a main effect of age group [F[4,623] 3.87, $$p \leq 0.004$$, Ƞ2 = 0.02, ε = 0.94] showed longer MMN-2 latency for 6-month than 3 months old participants. The 24-month-old infants also showed shorter MMN-2 latency than participants at 6, 9, and 12 months old. A significant age group by ROI interaction was also seen [F[12,1759] 1.86, $$p \leq 0.04$$, Ƞ2 = 0.01, ε = 0.94]. The age groups differed in MMN-2 latency in frontal left, frontal right, temporal right ROIs. The post hoc tests showed that infants at 3 months of age displayed longer MMN-2 latency than participants at 6, 9, 12, and 24 months in frontal left ROI, while in frontal right, 3-month-old infants displayed longer MMN-2 latency than infants at 9 months of age, and the participants at 9 months of age had longer MMN-2 latency compared to 24-month-old infants. In temporal right ROI, infants at 24 months of age displayed longer MMN-2 latency than participants at 3 and 6 months old (see Supplementary Figure S1).
Although the age groups did not differ in MMN-2 amplitude, they were different in MMN-2 latency. A significant age group by ROI interaction [F[12,1,667] 2.93, $$p \leq 0.001$$, Ƞ2 = 0.02, ε = 0.94] revealed that the groups differed in frontal left, frontal right, and temporal left ROIs. In frontal left ROI, infants at 3 months of age displayed shorter MMN-2 latency than participants at 9 and 12 months. The participants at 6 months of age also showed shorter MMN-2 latency compared to 12-month-old infants. However, at 24 months old, the infants displayed a shorter MMN-2 latency than the participants at 9 and 12 months. In frontal right ROI, longer MMN-2 latency was seen in 9-month-old infants compared to infants at 3 and 6 months old. In temporal left ROI, we found that infants at 6 months old displayed longer MMN-2 latency than 9-month-old infants (see Supplementary Figure S2).
The age groups did not differ in MMN-2 amplitude, but they differed in MMN-2 latency. A significant main effect of group [F[4,567] 2.57, $$p \leq 0.04$$, Ƞ2 = 0.02, ε = 0.96] evinced that infants at 24 months of age displayed shorter MMN-2 latency compared to infants at 6, 9, and 12 months of age (see Supplementary Table S4).
A significant age group by ROI interaction was also seen [F[12,1,635] 3.83, $p \leq 0.001$, Ƞ2 = 0.03, ε = 0.96]. The post hoc tests showed that age groups differed in MMN-2 latency in frontal left, frontal right, temporal left, and temporal right ROIs. In frontal left ROI, infants at 3 months of age displayed shorter MMN-2 latency than participants at 6 and 9 months old. 12-months-old infants displayed shorter MMN-2 latency than participants at 6 months old. This same pattern was observed for 24-month-old infants, which displayed shorter MMN-2 latency compared to infants at 6 and 9 months old. In frontal right ROI, infants at 3 months of age also showed shorter MMN-2 latency than infants at 6, 9, and 12 months of age. However, at 24 months, infants displayed shorter MMN-2 latency than 9-month-old infants. In the temporal left ROI, the participants at 3 months of age displayed longer MMN-2 latency than 9-month-old infants, while in temporal right ROI, 3-month-old participants displayed longer MMN-2 latency compared to infants at 24 months of age and infants at 12 months old displayed longer MMN-2 latency than participants at 6 and 24 months old (see Supplementary Figure S3).
## Regression results
As shown in Table 3, at 12-months old, only the infant’s diet predicted MMN-2 latency in frontal left and temporal right ROIs. In this same age group, diet was also a predictor of hemispheric asymmetry in the MMN-2 latency.
**Table 3**
| Age | Age.1 | Variables | Variables.1 | Coefficient standardized | Coefficient standardized.1 | Coefficient standardized.2 | Model | ANOVA | ANOVA.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | ROI | Predictor | β | t | p-value | R 2 | F | p-value |
| MMN-2 latency | MMN-2 latency | MMN-2 latency | MMN-2 latency | MMN-2 latency | MMN-2 latency | MMN-2 latency | MMN-2 latency | MMN-2 latency | MMN-2 latency |
| 12 m | FL | Diet | 0.1 | 2.8 | 0.005 | 0.02 | 8.0 | 0.005** | |
| | TR | Diet | −0.1 | −2.0 | 0.05 | 0.01 | 3.9 | 0.05* | Hemispheric asymmetry of MMN-2 latency |
| 12 m | Frontal | Diet | 0.1 | 2.2 | 0.02 | 0.01 | 5.0 | 0.02* | |
## Discussion
This study sought to identify electrophysiological differences between dietary groups at 3, 6, 9, 12, and 24 months of age. We expected to find an effect of diet on infant phonological processing, particularly at earlier developmental ages, which would be characterized by greater amplitude and shorter latency of MMN components and accompanied by a greater hemispheric asymmetry of MMN components for the BF group than MF and SF groups. Additionally, we expected greater amplitude and shorter latencies of MMN components for BF groups as age increased, which we expected to be less evident in the other dietary groups.
## Phonological-perception development between dietary groups
Our findings partially matched our hypothesis with differences between dietary groups observed in only one of the MMN components. We did not find differences between dietary groups in amplitude or latency of MMN-1, which has been associated with the identification of acoustic features of a stimulus (i.e., MMN-1) (51–53). Although this finding is in line with findings of Li et al. [ 13], it did not match those reported by Pivik et al. [ 3] who reported differences in P1 amplitude between dietary groups, with greater amplitude in the deviant condition for soy milk fed infants than breastfed at 3 and 6-months. In our study we expected to find a greater MMN-1 amplitude for SF than BF groups, in keeping with the findings of Pivik et al. [ 3], but this was not the case. We suggest that our results could be explained by the type of ERP analyses performed. Pivik et al. [ 3] compared the amplitude and latency of the P1 component associated with frequent and deviant conditions, while in our study, we directly compared the differences between experimental conditions (i.e., MMN components), and included infant sex and gestation weeks as covariates.
We propose that our results might be explained by the suggestion of previous studies that identification of acoustic features is developed very early in infancy (30–32). Given that this precognitive process might not be under development during our evaluation period, the nutritional requirements to support brain networks need for efficient processing would be easily provided by each of the three diets evaluated.
We also hypothesized differences between dietary groups in the MMN-2 component at six and 12-months of age. Our results partially supported our hypotheses; dietary groups only differed at 12-months old, underpinning the idea that nutrient intake has a greater effect on an infant’s cognition at a critical stage of language development. At this age, it is expected that infants show phonemic normalization and categorical perception (30]. Infants should recognize words [36] because they have already undergone extensive maturity changes in brain networks associated with production centers in the frontal region and the phonological store in the temporal region [38]. Moreover, they already show a more mature hemispheric specialization associated with language processing [40, 77]. As a consequence, phonological perception might require greater participation from neural networks that support attentional monitoring, inhibitory control, stimulus detection, and working memory (i.e., the dorsolateral prefrontal cortex, inferior frontal junction, inferior frontal gyrus, insula, presupplementary motor area, subthalamic nucleus, median cingulate, and striatum) [78] because they have attended syllables and inhibit their possible meaning in their native language, promoting greater participation from frontal brain areas related to attention-inhibition processing [38].
Our findings revealed that the SF group showed an inverse electrophysiological pattern to that of BF and MF infants; in which the SF group displayed shorter MMN-2 latency in frontal left ROI and longer MMN-2 latency in temporal right ROI. One explanation for the differences between dietary groups in MMN-2 latency in frontal left ROI is that the SF group exhibits a different attention-inhibition effort than the other groups, reflected in a reduced level of interference relative to the other groups. While shorter latencies might suggest more efficient processing, this finding might also indicate that SF infants have less linguistic information to inhibit or a weaker attention-inhibition brain network. This last explanation matches the findings of Li et al. [ 79] who reported lower executive function in children fed with soy formula in infancy than those fed with breast or cow-milk formula.
On the other hand, shorter MMN-2 latency for the SF group in temporal right ROI requires an additional explanation. Although how the hemispheric specialization in language processing develops during infancy is still debated, it has been hypothesized that the left hemisphere is specialized for speech stimuli, while the right hemisphere supports the auditory identification of non-speech stimuli [80, 81]. In our study, the SF group displayed an enhanced response in the right hemisphere, suggesting that this group is attending the syllables as non-speech stimuli. This proposal is in accord with their brain response in frontal left ROI. They appear to expend less cognitive effort to attend syllables and inhibit linguistic context because they may be processing the syllables as tones [81]. The SF group also exhibited a more right-lateralized MMN-2 asymmetry that has been suggested to be associated with a risk of delayed language development [40]. Attenuation of left hemispheric ERPs [82, 83] or atypical enhanced responses in the right hemisphere [84, 85] have been to confer greater risk of poor language development. Moreover, given that regression analyses indicated that only infant diet predicted latency and hemispheric asymmetry of the MMN-2 component in frontal areas, the SF group’s electrophysiological response might indicate a deviation from normal language development after a prolonged use of soy-based formula.
In addition, the electrophysiological pattern observed in SF groups does not match the temporal gradient in information processing (i.e., faster processing in temporal than frontal regions) observed in normal development [54]. These findings addressed the speed at which information is processed between language areas, suggesting that the differences between dietary groups in frontal ROIs could be interpreted as modulations in brain networks to enhance the ability to distinguish between syllables and manage neural resources and cognitive effort.
Prior studies using animal models and humans have noted that soy food contains phytoestrogens such as isoflavones (86–88) that seem to have a negative effect on cognition, alter sexually dimorphic brain regions, learning, memory [89] and executive functions [79]. We suggest that the deviation from normal language processing observed in the SF group may be attributable to the composition of soy formulas.
## Phonological-perception development for each dietary group
Consistent with our hypothesis, dietary groups displayed changes in MMN components associated with age, and these changes were more evident in the BF group. The MMN-1 component appears to change with age only in the BF and MF groups. Both groups displayed an increase in MMN-1 latency in frontal ROIs, which may suggest greater participation of frontal areas supporting inhibitory control in order to better identify the features of acoustic stimuli (51–53). These dietary groups also displayed a decrease in MMN-latency in temporal ROIs, which might be explained as a reflection of a more available phonological store [38] as age increases. However, SF infants did not display these changes associated with age, suggestive of a less stable development of the ability to identify the features of acoustic stimuli. Another explanation for this result is that the SF group had high variability in their brain-electrical responses associated with identification of acoustic features at all ages, which would hamper the observation of differences between age groups and even more so between dietary groups.
Although the MMN-2 component changed with age in all dietary groups, only the BF group showed greater MMN-2 amplitude as age increased, as has been described in a previous study [50]. This finding may be interpreted as greater availability of neural resources in older breastfed infants who seemed to show a greater stimulus awareness and perceptual salience, and thus a greater index of auditory recognition memory [51, 90] as age increased. This finding is consistent with behavioral results observed in 24-month-old infants on the PLS-3 test where breastfed infants showed greater auditory comprehension.
On the other hand, the electrophysiological pattern associated with age of MMN-2 latency also depended on regions of interest. BF and MF groups showed an increase in MMN-2 latency in the frontal left ROI from 3 to 12 months of age. This pattern was not observed in the SF group. Instead, that group displayed a concave-learning curve [91] characterized by a significant decrease of MMN-2 latency in frontal left ROI from 12 months of age. This finding may indicate reduced participation of the frontal left ROI in auditory recognition memory, consistent with a deviation from normal development in the recruitment of this brain area to process phonological awareness. In addition, an unexpected result was that SF infants displayed an increase in MMN-2 latency in temporal regions from this same age, which contrasted with the decreased age-associated finding in BF and MF groups. A previous study of language learning has described those greater fluctuations in learning curves as an indicator of slower learners, which may explain our findings in SF group [91]. We add to this that SF infants may have a less available phonological store at 12 months of age. The unexpected electrophysiological pattern observed in SF infants temporally matches with a milestone in infant language development where they are expected to show greater stimulus awareness due to their ability to distinguish words, syllables, and tones. Therefore, we suggest that SF infants compensate for failures in the frontal left area by recruiting bilateral temporal ROI to distinguish between phonological features of words, syllables, and tones.
## Limitations
There are inherent limitations in the present study. Although the longitudinal nature of this study may support interpretations of causality between diet and phonological processing, it is essential to highlight that the same subjects did not always constitute the sample at each moment evaluated. Some of them missed more than one measurement. Therefore, interpretations should be carried out carefully. In addition, given that our study implied infant nutrition, variables surrounding infant feeding were not wholly controlled, among them the mother’s health or the amount of food provided to the infant, or complementary feeding habits. We did not explore why the parents choose one of the three diets offered. In this study, we used traditional anthropometric measures to assess the participants, while this is a common use of body composition measures (e.g., energy X-ray absorptiometry) or biochemical indices, these might provide more nuanced metrics for studies examining the impact of diet on neural maturation and cognitive function in infants.
## 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 of the University of Arkansas for Medical Sciences. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.
## Author contributions
GA-C, AA, SS, and LL-P contributed to the conception and design of the study. YG, DW, HD, and DH organized the database for the statistical analyses. All authors contributed to the article and approved the submitted version.
## Funding
This research was supported by USDA/Agricultural Research Service Project 6026-51000-012-06S.
## 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.2023.1032413/full#supplementary-material
## References
1. Sanchez C, Richards J, Almli C. **Neurodevelopmental MRI brain templates for children from 2 weeks to 4 years of age**. *Dev Psychobiol* (2012) **54** 77-91. DOI: 10.1002/dev.20579
2. Stiles J, Jernigan TL. **The basics of brain development**. *Neuropsychol Rev* (2010) **20** 327-48. DOI: 10.1007/s11065-010-9148-4
3. Andres A, Cleves MA, Bellando JB, Pivik RT, Casey PH, Badger TM. **Developmental status of 1-year-old infants fed breast milk, cow’s milk formula, or soy formula**. *Pediatrics* (2012) **129** 1134-40. DOI: 10.1542/peds.2011-3121
4. Bellando J, McCorkle G, Spray B, Sims CR, Badger TM, Casey PH. **Developmental assessments during the first 5 years of life in infants fed breast milk, cow’s milk formula, or soy formula**. *Food Sci Nutr* (2020) **8** 3469-78. DOI: 10.1002/fsn3.1630
5. Ballard O, Morrow A. **Human milk composition: nutrients and bioactive factors**. *Pediatr Clin North A* (2013) **60** 49-74. DOI: 10.1016/j.pcl.2012.10.002
6. Hoi AG, McKerracher L. **Breastfeeding and infant growth**. *Evol Med Public Heal* (2015) **2015** 150-1. DOI: 10.1093/emph/eov012
7. de Graaf-Peters VB, Hadders-Algra M. **Ontogeny of the human central nervous system: what is happening when?**. *Early Hum Dev* (2006) **82** 257-66. DOI: 10.1016/j.earlhumdev.2005.10.013
8. Deoni S, Douglas D, Joelson S, O'Regan J, Schneider N. **Early nutrition influences developmental myelination and cognition in infants and young children**. *NeuroImage* (2018) **178** 649-59. DOI: 10.1016/j.neuroimage.2017.12.056
9. Heird WC. **Progress in promoting breast-feeding, combating malnutrition, and composition and use of infant formula, 1981-2006**. *J Nutr* (2007) **137** 499S-502S. DOI: 10.1093/jn/137.2.499S
10. Deoni S, Dean DC, Piryatinsky I, O'Muircheartaigh J, Waskiewicz N, Lehman K. **Breastfeeding and early white matter development: a cross-sectional study**. *NeuroImage* (2013) **82** 77-86. DOI: 10.1016/j.neuroimage.2013.05.090
11. Guesnet P, Alessandri JM. **Docosahexaenoic acid (DHA) and the developing central nervous system (CNS) - implications for dietary recommendations**. *Biochimie* (2011) **93** 7-12. DOI: 10.1016/j.biochi.2010.05.005
12. Ou X, Andres A, Pivik RT, Cleves MA, Badger TM. **Brain gray and white matter differences in healthy normal weight and obese children**. *J Magn Reson Imaging* (2015) **42** 1205-13. DOI: 10.1002/jmri.24912
13. Li J, Dykman RA, Jing H, Gilchrist JM, Badger TM, Pivik RT. **Cortical responses to speech sounds in 3-and 6-month-old infants fed breast milk, milk formula, or soy formula**. *Dev Neuropsychol* (2010) **35** 762-84. DOI: 10.1080/87565641.2010.508547
14. Pivik RT, Andres A, Bai S, Cleves MA, Tennal KB, Gu Y. **Infant diet-related changes in syllable processing between 4 and 5 months: implications for developing native language sensitivity**. *Dev Neuropsychol* (2016) **41** 215-30. DOI: 10.1080/87565641.2016.1236109
15. Pivik RT, Andres A, Tennal KB, Gu Y, Armbya N, Cleves MA. **Infant diet, gender and the normative development of vagal tone and heart period during the first two years of life**. *Int J Psychophysiol* (2013) **90** 311-20. DOI: 10.1016/j.ijpsycho.2013.10.001
16. Cheatham CL, Sheppard KW. **Synergistic effects of human milk nutrients in the support of infant recognition memory: an observational study**. *Nutrients* (2015) **7** 9079-95. DOI: 10.3390/nu7115452
17. Lönnerdal B. **Bioactive proteins in human Milk: health, nutrition, and implications for infant formulas**. *J Pediatr* (2016) **173** S4-9. DOI: 10.1016/j.jpeds.2016.02.070
18. Fleming SA, Mudd AT, Hauser J, Yan J, Metairon S, Steiner P. **Dietary oligofructose alone or in combination with 2-fucosyllactose differentially improves recognition**. *Nutrients* (2020) **12** 1-17. DOI: 10.3390/nu12072131
19. Vázquez E, Barranco A, Ramírez M, Gruart A, Delgado-García JM, Martínez-Lara E. **Effects of a human milk oligosaccharide, 2′-fucosyllactose, on hippocampal long-term potentiation and learning capabilities in rodents**. *J Nutr Biochem* (2015) **26** 455-65. DOI: 10.1016/j.jnutbio.2014.11.016
20. Prado EL, Dewey KG. **Nutrition and brain development in early life**. *Nutr Rev* (2014) **72** 267-84. DOI: 10.1111/nure.12102
21. Benasich AA, Tallal P. **Infant discrimination of rapid auditory cues predicts later language impairment**. *Behav Brain Res* (2002) **136** 31-49. DOI: 10.1016/S0166-4328(02)00098-0
22. Diaz HR. **Fetal, neonatal, and infant microbiome: perturbations and subsequent effects on brain development and behavior**. *Semin Fetal Neonatal Med* (2016) **21** 410-7. DOI: 10.1016/j.siny.2016.04.012
23. Fichter M, Klotz M, Hirschberg DL, Waldura B, Schofer O, Ehnert S. **Breast milk contains relevant neurotrophic factors and cytokines for enteric nervous system development**. *Mol Nutr Food Res* (2011) **55** 1592-6. DOI: 10.1002/mnfr.201100124
24. Anderson JW, Johnstone BM, Remley DT. **Breast-feeding and cognitive development: a meta-analysis**. *Am J Clin Nutr* (1999) **70** 525-35. DOI: 10.1093/ajcn/70.4.525
25. Brion M-JA, Lawlor DA, Matijasevich A, Horta B, Anselmi L, Araújo CL. **What are the causal effects of breastfeeding on IQ, obesity and blood pressure? Evidence from comparing high-income with middle-income cohorts**. *Int J Epidemiol* (2011) **40** 670-80. DOI: 10.1093/ije/dyr020
26. Kramer MS. **Breastfeeding and child cognitive development**. *Arch Gen Psychiatry* (2008) **65** 578. DOI: 10.1001/archpsyc.65.5.578
27. Neville MC, Anderson SM, McManaman JL, Badger TM, Bunik M, Contractor N. **Lactation and neonatal nutrition: defining and refining the critical questions**. *J Mammary Gland Biol Neoplasia* (2012) **17** 167-88. DOI: 10.1007/s10911-012-9261-5
28. Quigley MA, Hockley C, Carson C, Kelly Y, Renfrew MJ, Sacker A. **Breastfeeding is associated with improved child cognitive development: a population-based cohort study**. *J Pediatr* (2012) **160** 25-32. DOI: 10.1016/j.jpeds.2011.06.035
29. Mahurin SJ. **Breastfeeding and language outcomes: a review of the literature**. *J Commun Disord* (2015) **57** 29-40. DOI: 10.1016/j.jcomdis.2015.04.002
30. Dehaene-Lambertz G, Gliga T. **Common neural basis for phoneme processing in infants and adults**. *J Cogn Neurosci* (2004) **16** 1375-87. DOI: 10.1162/0898929042304714
31. Kuhl PK, Williams KA, Lacerda F, Stevens KN, Lindblom B. **Linguistic experience alters phonetic perception in infants by 6 months of age**. *Science* (1992) **255** 606-8. DOI: 10.1126/science.1736364
32. Cheng YY, Wu HC, Tzeng YL, Yang MT, Zhao LL, Lee CY. **Feature-specific transition from positive mismatch response to mismatch negativity in early infancy: mismatch responses to vowels and initial consonants**. *Int J Psychophysiol* (2015) **96** 84-94. DOI: 10.1016/j.ijpsycho.2015.03.007
33. Friederici AD, Friedrich M, Weber C. **Neural manifestation of cognitive and precognitive mismatch detection in early infancy**. *Neuroreport* (2002) **13** 1251-4. DOI: 10.1097/00001756-200207190-00006
34. Dehaene-Lambertz G, Dupoux E, Gout A. **Electrophysiological correlates of phonological processing: a cross-linguistic study**. *J Cogn Neurosci* (2000) **12** 635-47. DOI: 10.1162/089892900562390
35. Kuhl PK, Tsao FM, Liu HM. **Foreign-language experience in infancy: effects of short-term exposure and social interaction on phonetic learning**. *Proc Natl Acad Sci U S A* (2003) **100** 9096-101. DOI: 10.1073/pnas.1532872100
36. Jycszyk PW. **How infants begin to extract words from speech**. *Trends Cogn Sci* (1999) **3** 323-8. DOI: 10.1016/S1364-6613(99)01363-7
37. Rivera-Gaxiola M, Klarman L, Garcia-Sierra A, Kuhl PK. **Neural patterns to speech and vocabulary growth in American infants**. *Neuroreport* (2005) **16** 495-8. DOI: 10.1097/00001756-200504040-00015
38. Dubois J, Poupon C, Thirion B, Simonnet H, Kulikova S, Leroy F. **Exploring the early organization and maturation of linguistic pathways in the human infant brain**. *Cereb Cortex* (2016) **26** 2283-98. DOI: 10.1093/cercor/bhv082
39. Pang EW, Edmonds GE, Desjardins R, Khan SC, Trainor LJ, Taylor MJ. **Mismatch negativity to speech stimuli in 8-month-old infants and adults**. *Int J Psychophysiol* (1998) **29** 227-36. DOI: 10.1016/S0167-8760(98)00018-X
40. Cantiani C, Ortiz-Mantilla S, Riva V, Piazza C, Bettoni R, Musacchia G. **Reduced left-lateralized pattern of event-related EEG oscillations in infants at familial risk for language and learning impairment**. *Neuro Image Clin* (2019) **22** 101778. DOI: 10.1016/j.nicl.2019.101778
41. Holland SK, Vannest J, Mecoli M, Jacola LM, Tillema JM, Karunanayaka PR. **Functional MRI of language lateralization during development in children**. *Int J Audiol* (2007) **46** 533-51. DOI: 10.1080/14992020701448994
42. Piazza C, Cantiani C, Miyakoshi M, Riva V, Molteni M, Reni G. **EEG effective source projections are more bilaterally symmetric in infants than in adults**. *Front Hum Neurosci* (2020) **14** 82. DOI: 10.3389/fnhum.2020.00082
43. Rolheiser T, Stamatakis EA, Tyler LK. **Dynamic processing in the human language system: synergy between the arcuate fascicle and extreme capsule**. *J Neurosci* (2011) **31** 16949-57. DOI: 10.1523/JNEUROSCI.2725-11.2011
44. Dick AS, Tremblay P. **Beyond the arcuate fasciculus: consensus and controversy in the connectional anatomy of language**. *Brain* (2012) **135** 3529-50. DOI: 10.1093/brain/aws222
45. Vandermosten M, Boets B, Wouters J, Ghesquière P. **A qualitative and quantitative review of diffusion tensor imaging studies in reading and dyslexia**. *Neurosci Biobehav Rev* (2012) **36** 1532-52. DOI: 10.1016/j.neubiorev.2012.04.002
46. Hornickel J, Kraus N. **Unstable representation of sound: a biological marker of dyslexia**. *J Neurosci* (2013) **33** 3500-4. DOI: 10.1523/JNEUROSCI.4205-12.2013
47. Centanni TM, Booker AB, Sloan AM, Chen F, Maher BJ, Carraway RS. **Knockdown of the dyslexia-associated gene Kiaa0319 impairs temporal responses to speech stimuli in rat primary auditory cortex**. *Cereb Cortex* (2014) **24** 1753-66. DOI: 10.1093/cercor/bht028
48. Čeponiene R, Torki M, Alku P, Koyama A, Townsend J. **Event-related potentials reflect spectral differences in speech and non-speech**. *Clin Neurophysiol* (2008) **119** 1560-77. DOI: 10.1016/j.clinph.2008.03.005
49. Choudhury N, Benasich AA. **Maturation of auditory evoked potentials from 6 to 48 months: prediction to 3 and 4 year language and cognitive abilities**. *Clin Neurophysiol* (2011) **122** 320-38. DOI: 10.1016/j.clinph.2010.05.035
50. Morr ML, Shafer VL, Kreuzer JA, Kurtzberg D. **Maturation of mismatch negativity in typically developing infants and preschool children**. *Ear Hear* (2002) **23** 118-36. DOI: 10.1097/00003446-200204000-00005
51. Čeponiene R, Alku P, Westerfield M, Torki M, Townsend J. **ERPs differentiate syllable and nonphonetic sound processing in children and adults**. *Psychophysiology* (2005) **42** 391-406. DOI: 10.1111/j.1469-8986.2005.00305.x
52. Cunningham J, Nicol T, Zecker S, Kraus N. **Speech-evoked neurophysiologic responses in children with learning problems: development and behavioral correlates of perception**. *Ear Hear* (2000) **21** 554-68. DOI: 10.1097/00003446-200012000-00003
53. Ponton C, Eggermont J, Don M, Waring MD, Kwong B, Cunningham J. **Maturation of the mismatch negativity: effects of profound defness and cochlear implant use**. *J Allergy Clin Immunol* (2000) **5** 167-85. DOI: 10.1159/000013878
54. Čeponiene R, Rinne T, Näätänen R. **Maturation of cortical sound processing as indexed by event-related potentials**. *Clin Neurophysiol* (2002) **113** 870-82. DOI: 10.1016/S1388-2457(02)00078-0
55. Novak G, Kurtzberg D, Kreuzer J, Vaughan HG. **Cortical responses to speech sounds and their formants in normal infants: maturational sequence and spatiotemporal analysis**. *Electroencephalogr Clin Neurophysiol* (1989) **73** 295-305. DOI: 10.1016/0013-4694(89)90108-9
56. Kushnerenko E, Ceponiene R, Balan P, Fellman V, Näätänen R. **Maturation of the auditory change detection response in infants: a longitudinal ERP study**. *Neuroreport* (2002) **13** 1843-8. DOI: 10.1097/00001756-200210280-00002
57. Kushnerenko E, Cheour M, Ceponiene R, Fellman V, Renlund M, Soininen K. **Central auditory processing of durational changes in complex speech patterns by newborns: an event-related brain potential study**. *Dev Neuropsychol* (2001) **19** 83-97. DOI: 10.1207/S15326942DN1901_6
58. Paquette N, Vannasing P, Tremblay J, Lefebvre F, Roy MS, McKerral M. **Early electrophysiological markers of atypical language processing in prematurely born infants**. *Neuropsychologia* (2015) **79** 21-32. DOI: 10.1016/j.neuropsychologia.2015.10.021
59. Partanen E, Pakarinen S, Kujala T, Huotilainen M. **Infants’ brain responses for speech sound changes in fast multifeature MMN paradigm**. *Clin Neurophysiol* (2013) **124** 1578-85. DOI: 10.1016/j.clinph.2013.02.014
60. Ragó A, Honbolygó F, Róna Z, Beke A, Csépe V. **Effect of maturation on suprasegmental speech processing in full-and preterm infants: a mismatch negativity study**. *Res Dev Disabil* (2014) **35** 192-202. DOI: 10.1016/j.ridd.2013.10.006
61. Sato Y, Sogabe Y, Mazuka R. **Development of hemispheric specialization for lexical pitch-accent in Japanese infants**. *J Cogn Neurosci* (2010) **22** 2503-13. DOI: 10.1162/jocn.2009.21377
62. Miller E, Li L, Desimone R. **A neural mechanism for working and recognition memory in inferior temporal cortex**. *Science* (1991) **254** 1377-9. PMID: 1962197
63. Alho K, Sajaniemi N, Niittyvuopio T, Sainio K, Näätänen R. **“ERPs to an auditory stimulus change in pre-term and full-terms infants”**. *Psychophysiological Brain Research. Vol 2* (1990) 139-142
64. Naatanen R, Paavilainen P, Rinne T, Alho K. **The mismatch negativity (MMN) in basic research of central auditory processing: a review**. *Clin Neurophysiol* (2007) **118** 2544-90. DOI: 10.1016/j.clinph.2007.04.026
65. Pivik RT, Andres A, Badger TM. **Effects of diet on early stage cortical perception and discrimination of syllables differing in voice-onset time: a longitudinal ERP study in 3 and 6month old infants**. *Brain Lang* (2012) **120** 27-41. DOI: 10.1016/j.bandl.2011.08.004
66. Čeponiene R, Cheour M, Näätänen R. **Interstimulus interval and auditory event-related potentials in children: evidence for multiple generators**. *Electroencephalogr Clin Neurophysiol Evoked Potentials* (1998) **108** 345-54. DOI: 10.1016/S0168-5597(97)00081-6
67. Giard MH, Lavikainen J, Reinikainen K, Perrin F, Bertrand O, Pernier J. **Separate representation of stimulus frequency, intensity, and duration in auditory sensory memory: an event-related potential and dipole-model analysis**. *J Cogn Neurosci* (1995) **7** 133-43. DOI: 10.1162/jocn.1995.7.2.133
68. Horvath J, Czigler I, Sussman E, Winkler I. **Simultaneously active pre-attentive representations of local and global rules for sound sequences in the human brain**. *Cogn Brain Res* (2001) **12** 131-44. DOI: 10.1016/S0926-6410(01)00038-6
69. Shtyrov Y, Pulvermuller F. **Memory traces for inflectional affixes as shown by mismatch negativity**. *Eur J Neurosci* (2002) **15** 1085-91. DOI: 10.1046/j.1460-9568.2002.01941.x
70. Hollingshead A. *Four Factor Index of Social Status* (1975)
71. Weshler D. *The Wechsler Abbreviated Scale of Intelligence* (2011)
72. Muruish M, Bershadsky B, Goldstein L. **Reliability and validity of the SA-45: further evidence from a primary care setting**. *Assessment* (1998) **5** 407-19. DOI: 10.1177/107319119800500410
73. Bayley N. *Bayley Scales of Infant Development Manual 2* (1993)
74. Zimmerman I, Steiner V, Pond R. *PLS-3: Preschool Language Scale-3* (1992)
75. Brannon EM, Libertus ME, Meck WH, Woldorff MG. **Electrophysiological measures of time processing in infant and adult brains: Weber’s law holds**. *J Cogn Neurosci* (2008) **20** 193-203. DOI: 10.1162/jocn.2008.20016
76. Cheour M, Ćèponiené R, Leppänen P, Alho K, Kujala T, Renlund M. **The auditory sensory memory trace decays rapidly in newborns**. *Scand J Psychol* (2002) **43** 33-9. DOI: 10.1111/1467-9450.00266
77. Musacchia G, Ortiz-Mantilla S, Choudhury N, Realpe-Bonilla T, Roesler C, Benasich AA. **Active auditory experience in infancy promotes brain plasticity in theta and gamma oscillations**. *Dev Cogn Neurosci* (2017) **26** 9-19. DOI: 10.1016/j.dcn.2017.04.004
78. Liu J, Zhang H, Chen C, Chen H, Cui J, Zhou X. **The neural circuits for arithmetic principles**. *NeuroImage* (2017) **147** 432-46. DOI: 10.1016/j.neuroimage.2016.12.035
79. Li T, Badger TM, Bellando BJ, Sorensen ST, Lou X, Ou X. **Brain cortical structure and executive function in children may be influenced by parental choices of infant diets**. *Am J Neuroradiol* (2020) **41** 1302-8. DOI: 10.3174/ajnr.A6601
80. Dehaene-Lambertz G, Dehaene S, Hertz-Pannier L. **Functional neuroimaging of speech perception in infants**. *Science* (2002) **298** 2013-5. DOI: 10.1126/science.1077066
81. Homae F. **A brain of two halves: insights into interhemispheric organization provided by near-infrared spectroscopy**. *NeuroImage* (2014) **85** 354-62. DOI: 10.1016/j.neuroimage.2013.06.023
82. Leppänen M, Lapinleimu H, Lehtonen L, Rautava P. **Growth of extremely preterm infants born in 2001-2010**. *Acta Pediatr* (2013) **102** 206-8. DOI: 10.1111/apa.12061
83. Hämäläinen JA, Salminen HK, Leppänen PHT. **Basic auditory processing deficits in dyslexia**. *J Learn Disabil* (2013) **46** 413-27. DOI: 10.1177/0022219411436213
84. Friederici AD, Makuuchi M, Bahlmann J. **The role of the posterior superior temporal cortex in sentence comprehension**. *Neuroreport* (2009) **20** 563-8. DOI: 10.1097/WNR.0b013e3283297dee
85. Guttorm TK, Leppänen PHT, Hämäläinen JA, Eklund KM, Lyytinen HJ. **Newborn event-related potentials predict poorer pre-reading skills in children at risk for dyslexia**. *J Learn Disabil* (2010) **43** 391-401. DOI: 10.1177/0022219409345005
86. Doerge DR, Woodling KA, Churchwell MI, Fleck SC, Helferich WG. **Pharmacokinetics of isoflavones from soy infant formula in neonatal and adult rhesus monkeys**. *Food Chem Toxicol* (2016) **92** 165-76. DOI: 10.1016/j.fct.2016.04.005
87. Lee YB, Lee HJ, Heon SS. **Soy isoflavones and cognitive function**. *J Nutr Biochem* (2005) **16** 641-9. DOI: 10.1016/j.jnutbio.2005.06.010
88. Merritt RJ, Jenks BH. **Safety of soy-based infant formulas containing isoflavones: the clinical evidence**. *J Nutr* (2004) **134** 1220S-4S. DOI: 10.1093/jn/134.5.1220S
89. Lephart ED, West TW, Weber KS, Rhees RW, Setchell KDR, Adlercreutz H. **Neurobehavioral effects of dietary soy phytoestrogens**. *Neurotoxicol Teratol* (2002) **24** 5-16. DOI: 10.1016/S0892-0362(01)00197-0
90. DeReggnier R. *Infant EEG and event related potentials: Studies in developmental psychology* (2007)
91. Murre JMJ. **S-shaped learning curves**. *Psychon Bull Rev* (2014) **21** 344-56. DOI: 10.3758/s13423-013-0522-0
|
---
title: Opium consumption and long-term outcomes of CABG surgery in patients without
modifiable risk factors
authors:
- Ali Sheikhy
- Aida Fallahzadeh
- Sepehr Nayebirad
- Mahdi Nalini
- Saeed Sadeghian
- Mina Pashang
- Mahmoud Shirzad
- Abbas Salehi-Omran
- Soheil Mansourian
- Jamshid Bagheri
- Kaveh Hosseini
journal: Frontiers in Surgery
year: 2023
pmcid: PMC9982127
doi: 10.3389/fsurg.2023.1047807
license: CC BY 4.0
---
# Opium consumption and long-term outcomes of CABG surgery in patients without modifiable risk factors
## Abstract
### Background
The question about the significance of opium consumption as a coronary artery disease (CAD) risk factor still remains open. The present study aimed to evaluate the association between opium consumption and long term outcomes of coronary artery bypass grafting (CABG) in patients without standard modifiable CAD risk factors (SMuRFs; hypertension, diabetes, dyslipidemia, and smoking).
### Methods
In this registry-based design, we included 23,688 patients with CAD who underwent isolated CABG between January 2006 to December 2016. Outcomes were compared in two groups; with and without SMuRF. The main outcomes were all-cause mortality, fatal and nonfatal cerebrovascular events (MACCE). Inverse probability weighting (IPW) adjusted Cox's proportional hazards (PH) model was used to evaluate the effect of opium on post-op outcomes.
### Results
During 133,593 person-years of follow-up, opium consumption was associated with increased risk of mortality in both patients with and without SMuRFs (weighted Hazard Ratio (HR)s: 1.248 [1.009, 1.574] and 1.410 [1.008, 2.038], respectively). There was no association between opium consumption and fatal and non-fatal MACCE in patients without SMuRF (HR = 1.027 [0.762–1.383], HR 0.700 [0.438–1.118]). Opium consumption was associated with earlier age of CABG in both groups; 2.77 (1.68, 3.85) years in SMuRF-less and 1.70 (1.11, 2.38) years in patients with SMuRFs.
### Conclusion
Opium users not only undergo CABG at younger ages but also have a higher rate of mortality regardless of the presence of traditional CAD risk factors. Conversely, the risk of MACCE is only higher in patients with at least one modifiable CAD risk factor.
## Introduction
The standard modifiable cardiovascular risk factors (SMuRFs), which are diabetes mellitus (DM), dyslipidemia (DLP), hypertension (HTN) and cigarette smoking (CS), are the key elements of the Framingham risk score [1] and are targeted in primary and secondary prevention programs [2]. However, an increasing number of patients with established coronary artery disease present with no known SMuRFs (SMuRF-less patients) at the time of first diagnosis [3, 4]. Several large registry-based studies have compared the short- and long-term outcomes with their counterparts with at least one SMuRF [3, 5]. However, the results were conflicting and some studies reported a higher risk of mortality in SMuRF-less patients while some reported no significant differences between patients with and without SMuRFS. Although outcomes of STEMI subjects with no SMuRFs have been widely studied, there has been hardly any studies focusing specifically on SMuRF-less patients undergoing CABG.
While the traditional SMuRFs have been used to predict the cardiovascular risk of an individual, there are other possible risk factors such as opium consumption that can affect the patients’ outcomes. Opium consumption is highly prevalent in developing countries of the Middle East and Asia, especially in Iran [6, 7]. The high prevalence of opium use in *Iran is* partially due to the ease of access and also the misconception among the Iranian population and even medical staff that opium might decrease the risk of certain medical conditions such as diabetes, hypertension as well as CAD [8]. Although many studies have reported detrimental effects of opium use on the cardiovascular system and poor post CABG outcomes in opium users (8–10), it is still not clear whether opium consumption should be considered as an independent CAD risk factor (besides other SMuRFs) or not. In addition, the effect of opium on long-term outcomes of CABG in SMuRF-less patients is debatable. Hence, in the present study, we aimed to evaluate the association of opium use with outcomes of isolated CABG in the SMuRF-less group and compare it with patients with at least one SMuRF.
## Study population, setting, and design
In this registry-based retrospective cohort study, which performed at Tehran Heart Center (THC) [11] from January 2006 to December 2016, patients undergoing isolated CABG surgery were included; all pre-operative and intraoperative data were gathered from the health information system (HIS). Postoperative and follow-up data were collected prospectively. We reported this study according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. Patients with inadequate data were excluded from the current study. There were two inclusion criteria; [1] Surgical revascularization criteria for ischemic heart disease, and [2] Isolated CABG excluding valve surgeries. Ultimately, 23,688 patients were included in the final analysis. The SMuRFs assessed in this study included HTN, DLP, DM, and current CS. Patients were divided into two main groups according to their SMuRF score (with and without SMuRFs). The present study was approved by the ethical board of THC (IR-THC-13799) and the involved human data was in accordance with the Helsinki Declaration.
## Follow-up protocol
Subjects were followed-up at 4, 6, and 12 months after surgery. After the first year of follow-up, patients were visited annually. Follow-up visits were carried out in the center's post-op clinic and data regarding mortality and MAACE were collected.
## Definition of variables
Diabetes mellitus was defined as fasting plasma glucose ≥126 mg/dl, random plasma glucose ≥200 mg/dl, hemoglobin A1c (HbA1c) ≥$6.5\%$ [12], treatment with either oral hypoglycemic agents or insulin. Hypertension was defined as a minimum systolic blood pressure of 140 mm Hg, a minimum diastolic blood pressure of 90 mm Hg or a history of antihypertensive therapy [13]. Dyslipidemia was defined as the presence of a minimum total cholesterol level of 240 mg/dl, a minimum triglyceride level of 200 mg/dl, or a high-density lipoprotein cholesterol level of less than 40 mg/dl in men and less than 50 mg/dl in women or a minimum low-density lipoprotein cholesterol level of 160 mg/dl, or a history of prescribed lipid medications based on the National Cholesterol Education Program (NCEP) Adult Treatment Plan (ATP) III [14]. A family history of CAD was defined as having a first-degree relative with a history of CAD including acute myocardial infarction or documented CAD (through invasive coronary angiography or computed tomography coronary angiography).' Current smoking was defined as regularly smoking more than one cigarette per day as reported by the patient. Opium consumption was defined as the current consumption of opium either smoking opium or drinking opium dissolved in tea.
## Study outcomes
The primary outcomes were defined as all-cause mortality, MACCE (major adverse cardiac and cerebrovascular events) and non-fatal MACCE (comprising of non-fatal acute coronary syndromes [ACS], non-fatal stroke or transient ischemic attack [TIA], and repeated coronary revascularization via percutaneous coronary intervention [PCI] or redo-CABG).
## Statistical analysis
Mean with standard deviation (SD) and median with 25th and 75th percentiles [interquartile range (IQR) boundaries] were used to present normal and skewed continuous variables, respectively. The normality of the variables was assessed using histogram charts in addition to the central tendency and dispersion measures. Comparison between “opium consumers” and “non-consumers” groups was done using student's t-test for normally distributed and Mann–Whitney U-test for skewed distributed variables. Categorical variables were expressed as frequency and percentage which were compared between the two abovementioned groups using the chi-squared test. Inverse probability weights (IPW) were used to stabilize potential selection biases of treatment, weights were calculated from propensity score (PS) (Supplementary Figure S1), which was generated by predicted probabilities of logistic regression on identified potential confounders. All selected variable in the PS estimation model was mentioned in (Supplementary Table S1). The C-statistic for the model was 0.83 (Supplementary Figure S2). Weights for each case (Wi) were calculated as 1/PS(Xi) for opium consumers, and 1/[1-PS(Xi)] for non-consumers. The standardized mean difference (SMD) was used as a balance metric to evaluate the difference between distributions of a pre-treatment variable, a balance indicator considered as “SMD < 0.1” (Supplementary Figure S3).
The weighted and unadjusted effects of opium consumption on all-cause mortality and MACCE were obtained using Cox's proportional hazards (PH) model. Interactions were examined by including appropriate interaction terms in the Cox regression models and reported as the ratio of HR (RHR) by considering SMuRF positive group as reference. Adjusted linear regression was used to assess the association between opium use with the age of CABG; hence in this model, age was considered an outcome.
All statistical analyses were conducted applying R version 4.0.3, moreover, we used several packages in R: survival” (package for survival analysis in R), “survminer” (drawing survival curves), and “ggplot2”. All P-values are two-sided; moreover, P-values <0.05 were considered statistically significant.
## Baseline characteristics
A total of 23,688 patients who underwent isolated CABG were recruited. During 133,593 person-years of follow-up (median 74.64; 25th–75th percentile: 73.97–75.32 months), barely $2\%$ of the patients [485] were lost to follow-up. 3,432 ($14.49\%$) patients did not have any SMuRFs (SMuRF-less). Baseline features are shown in Table 1.
**Table 1**
| Unnamed: 0 | Unnamed: 1 | SMuRF-less(n = 3,432) | SMuRF-less(n = 3,432).1 | SMuRF-less(n = 3,432).2 | SMuRF=>1(n = 20,256) | SMuRF=>1(n = 20,256).1 | SMuRF=>1(n = 20,256).2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | Non-opium users(n = 3,066) | Opium users(n = 366) | P value | Non-opium users(n = 17,380) | Opium users(n = 2,876) | P value |
| Gender | Female | 421 (13.7%) | 2 (0.5%) | <0.001 | 5,808 (33.4%) | 130 (4.5%) | <0.001 |
| Gender | Male | 2,645 (86.3%) | 364 (99.5%) | | 11,572 (66.6%) | 2,746 (95.5%) | <0.001 |
| Age (years) | Age (years) | 62 (1) | 58 (1) | <0.001 | 61 (1) | 57 (1) | <0.001 |
| BMI (kg/m2) | BMI <30 | 2,556 (83.3%) | 298 (81.4%) | 0.367 | 12,842 (74.2%) | 2,279 (79.6%) | 0.115 |
| BMI (kg/m2) | BMI ≥30 | 510 (16.7%) | 68 (18.6%) | | 4,468 (25.8%) | 583 (20.4%) | 0.115 |
| Hb (g/dl) | Hb (g/dl) | 14.11 (0.08) | 14.03 (0.03) | 0.423 | 13.70 (0.01) | 13.91 (0.03) | 0.228 |
| Graft number | Graft number | 3 (3,4) | 4 (3,4) | 0.201 | 3 (3,4) | 3 (3,4) | 0.049 |
| Diabetes | Diabetes | | | – | 8106 (46.6%) | 1,008 (35.0%) | <0.001 |
| Hypertension | Hypertension | | | – | 11,165 (64.2%) | 1,478 (51.4%) | <0.001 |
| Dyslipidemia | Dyslipidemia | | | – | 11,733 (67.5%) | 1,590 (55.3%) | <0.001 |
| Current CS | Current CS | | | – | 2,553 (14.7%) | 1,503 (52.3%) | <0.001 |
| eGFR (ml/min) | eGFR (ml/min) | 83.45 (66.80, 102.92) | 94.21 (74.68, 110.91) | <0.001 | 84.37 (65.98, 105.19) | 93.27 (73.86, 115.37) | <0.001 |
| Positive Family History | Positive Family History | 983 (32.1%) | 131 (35.8%) | 0.153 | 6,905 (39.8%) | 1,055 (36.7%) | 0.002 |
| EF (%) | ≥50 | 1,332 (48.3%) | 145 (42.4%) | 0.046 | 8,685 (55.0%) | 1,158 (43.2%) | <0.001 |
| EF (%) | <50 | 1,427 (51.7%) | 197 (57.6%) | | 7,115 (45.0%) | 1,520 (56.8%) | <0.001 |
| LM stenosis | LM stenosis | 276 (9.0%) | 26 (7.1%) | 0.226 | 1,407 (8.1%) | 253 (8.8%) | 0.204 |
| Pre-Surgery PCI | Pre-Surgery PCI | 154 (5.0%) | 33 (9.0%) | 0.002 | 1,117 (6.4%) | 235 (8.2%) | 0.001 |
| Renal Failure | Renal Failure | 35 (1.1%) | 3 (0.8%) | 0.577 | 449 (2.6%) | 73 (2.5%) | 0.889 |
| Urgent operation | Urgent operation | 122 (4.0%) | 15 (4.1%) | 0.924 | 968 (5.6%) | 152 (5.3%) | 0.531 |
| COPD | COPD | 81 (2.7%) | 18 (5.0%) | 0.013 | 412 (2.4%) | 126 (4.4%) | <0.001 |
| Cerebrovascular Accident | Cerebrovascular Accident | 111 (3.6%) | 16 (4.4%) | 0.470 | 1,075 (6.2%) | 210 (7.3%) | 0.024 |
| Pre-CABG MI Interval | No MI | 2017 (65.8%) | 199 (54.4%) | <0.001 | 11,779 (67.8%) | 1,649 (57.3%) | <0.001 |
| Pre-CABG MI Interval | ≤ 7Day | 220 (7.2%) | 23 (6.3%) | | 1295 (7.5%) | 314 (10.9%) | <0.001 |
| Pre-CABG MI Interval | 8–21 day | 158 (5.2%) | 25 (6.8%) | | 840 (4.8%) | 218 (7.6%) | <0.001 |
| Pre-CABG MI Interval | >21 Day | 671 (21.9%) | 119 (32.5%) | | 3,466 (19.9%) | 695 (24.2%) | <0.001 |
In the SMuRF-less group, 366 ($10.66\%$) patients were opium users. Age at the time of admission was significantly lower in opium users ($P \leq 0.001$).
In patients with at least one SMuRF, 2,876 ($14.20\%$) were opium consumers. The average age of the patients at the time of admission was significantly lower in opium users ($P \leq 0.001$). In both study groups, the number of male subjects was significantly higher in the opium consumption subcategory ($P \leq 0.001$).
## Mortality
At six years of follow-up, the mortality rate in entire study was $11.8\%$. Association between opium consumption and all-cause mortality in patients with and without SMuRFs was assessed, Table 2 and Figures 1A,B. In patients without any SMuRFs (SMuRF-less), the mortality trend was significantly different and was worse in the opium consumer group [HR:1.410 (1.008, 1.925), $$P \leq 0.024$$], Figure 1A.
**Figure 1:** *Cumulative hazard of mortality in patients without (A) and with (B) SMuRF.* TABLE_PLACEHOLDER:Table 2 *Subgroup analysis* was done to assess all-cause mortality in each traditional CAD risk factor, Table 3. Mortality risk due to opium consumption was barely higher in the SMuRF-less group [Ratio of Hazard Ratio (RHR) = 1.007, CI: 0.595–1.405; P-interation = 0.682]. Mortality was significantly higher in non-hypertensive and non-diabetic patients who consumed opium.
**Table 3**
| Unnamed: 0 | Crude model | Crude model.1 | IPW adjusted model | IPW adjusted model.1 | P for interaction |
| --- | --- | --- | --- | --- | --- |
| | HR | P value | HR | P value | P for interaction |
| Not cigarette smokers | 1.241 [1.087–1.417] | 0.001 | 1.240 [0.825–1.863] | 0.300 | 0.148 |
| Cigarette smokers | 1.114 [0.922–1.346] | 0.262 | 1.223 [0.935–1.602] | 0.142 | 0.148 |
| Non diabetic | 1.242 [1.086–1.421] | 0.002 | 1.368 [1.047–1.789] | 0.022 | 0.038 |
| Diabetic | 1.149 [0.971–1.358] | 0.106 | 1.167 [0.842–1.618] | 0.354 | 0.038 |
| Non hypertensive | 1.407 [1.213–1.632] | <0.001 | 1.559 [1.197–2.028] | <0.001 | 0.002 |
| hypertensive | 1.052 [0.906–1.220] | 0.508 | 1.088 [0.792–1.495] | 0.602 | 0.002 |
| Non dyslipidemic | 1.226 [1.061–1.416] | 0.006 | 1.490 [1.188–1.869] | <0.001 | 0.130 |
| Dyslipidemic | 1.086 [0.932–1.265] | 0.291 | 1.089 [0.768–1.543] | 0.632 | 0.130 |
## MACCE
Table 2 and Figure 2 demonstratethe association between pre-operative opium consumption and long-term MACCE. Although the trend shows a lower rate of MACCE in the SMuRF-less group compared to patients with SMuRF [HR:1.027 (0.762–1.383)], it was not statistically significant during follow-up, Figure 2A ($$P \leq 0.862$$). in contrast, patients with at least one SMuRF, opium consumption was associated with a higher rate of MACCE [HR: 1.201 (1.016, 1.419)]. Similar to mortality, MACCE was almost similar in the first 3 years but became divergent afterwards, Figure 2B. Subgroup analysis did not show significant differences of MACCE in studied subgroups, Table 4. In SMuRF less population, the rate of ACS, CVA, and revascularization was higher in non-opium consumers ($7.3\%$ vs. $6.1\%$, $2.5\%$ vs. $1.9\%$, and $1.7\%$ vs. $0.3\%$, respectively); In patients with at least one SMuRF, opium consumption was significantly associated with higher ACS [HR: 1.196 (1.052–1.360), $$P \leq 0.006$$]. The number of these components of MACCE were comparably low, hence this study did not have enough power to report generalizable results.
**Figure 2:** *Cumulative hazard of MACCE in patients without (A) and with (B) SMuRF.* TABLE_PLACEHOLDER:Table 4
## Non-fatal MACCE
There was no association between opium consumption and non-fatal MACCE in both SMuRF and SMuRF-less groups (HR: 1.138 [0.981–1.322] $$P \leq 0.088$$ and HR: 0.856 [0.543–1.349] $$P \leq 0.503$$, respectively).
## Opium consumption and age of CABG
The average age in patients with and without SMuRFs was 56.89 (53.44, 60.33) and 62.12 (60.05, 64.19), respectively, Table 5. Opium consumers underwent CABG at an earlier age when compared with the nonusers in both SMuRF-less and SMuRF + groups. The effect, however, was more pronounced in the SMuRF-less group (2.77 years earlier, $95\%$ CI 1.68–3.85 in the SMuRF-less group vs. 1.70, $95\%$ CI 1.11–2.38 in SMuRF + group).
**Table 5**
| Unnamed: 0 | Mean of age | Opium consumption effecta | P value |
| --- | --- | --- | --- |
| SMuRF-less | 56.89 (53.44, 60.33) | −2.79 (−1.68, −3.85) | <0.001 |
| SMuRF+ | 62.12 (60.05, 64.19) | −1.70 (−1.11, −2.39) | <0.001 |
## Discussion
The present study is a large cohort of 23,688 CABG subjects highlighting the effects of opium consumption on long-term outcomes in a commonly neglected subpopulation of CAD patients, the SMuRF-less group. Our findings showed that regardless of the patients' SMuRF status, opium consumption was associated with an increased risk of long-term mortality and MACCE. Opium consumers, especially those without SMuRFs, also demonstrated susceptibility to coronary events requiring CABG at an earlier age when compared with their nonuser counterparts.
A clinically significant proportion of patients with coronary diseases have none of the traditional CAD risk factors known as SMuRFs [3]. The question about better or poorer prognosis of SMuRF-less patients after CAD events is still on the table. Some studies have shown an increased all-cause mortality rate (in-hospital and 30-day mortality) in SMuRF-less patients presenting with STEMI, particularly in women [3]. On the other hand, another study have shown that 5 year mortality was lower in SMuRF-less patients although it was not statistical significant after multivariate adjustment [15]. This debate has also been raised about the role of opium consumption in occurrence and prognosis of CAD patients; especially more severe CAD groups who need CABG surgery.
Based on our findings, opium consumption was associated with increased risk of long-term all-cause mortality, in both patients with and without SMuRFs and also with increased risk of MACCE in patients with SMuRFs, but not with non-fatal MACCE. The possible explanation for this could be that opium consumption can mask some CAD symptoms such as chest pain; therefore, opium consumers have less hospitalization due to non-fatal MACCEs.
In accordance with the current study results, Masoudkabir et al. suggested that post-CABG opium consumption is associated with an increased risk of long-term mortality and MACCE [9].
Safaii et al. evaluated the effect of opium consumption on the short-term outcomes after CABG and showed that opium usage was associated with an increased risk of rehospitalization within six months of CABG [8]. Another study conducted by Nalini et al. showed that long-term opium consumption was associated with increased risk of cardiovascular mortality, independent of traditional CAD risk factors [16].
Below we will thoroughly discuss the mechanisms in which opium may be associated with CAD occurrence and its relationship with traditional CAD risk factors will also be reviewed.
The exact mechanisms through which opium may result in increased risk of MACCE are dabating; however, some possible mechanisms have been reported based on previous studies. Studies have reported that opium may induce chronic inflammation and oxidative stress by stimulating pro-inflammatory cytokines and thus, lead to coronary atherosclerosis and occurrence of acute events [17]. Another possible mechanism is that opium-addicted men and women have lower testosterone and estrogen levels than controls [14]. Plasma testosterone and estrogen levels are associated with the extent of CAD and the risk of cardiovascular mortality [18, 19]. In addition to traditional risk factors, studies have shown that opium consumption is associated with higher levels of several novel cardiovascular risk factors, including lipoprotein a (Lpa), c-reactive protein (CRP), fibrinogen [20], and Factor VII [21]. Lpa is shown to be an indicator of premature atherosclerosis [22], CRP is an inflammatory biomarker and is associated with increased risk of CAD [23] Fibrinogen and Factor VII are also shown to be associated with CAD [24]. It is probable that opium consumption causes an elevation in the inflammatory and thrombogenic biomarkers. This may explain the association between opium use and increased risk of MACCE.
## Opium and DM
Opium consumption was associated with hyperinsulinemia due to changes in hepatic extraction of insulin, hyperglycemia similar to what is seen in type 2 DM, and also high levels of glycated hemoglobin (HbA1c) and poor glycemic control (25–27). This may explain our finding that opium consumption significantly increased mortality risk in non-diabetics and confirm that the potential effect of opium usage on risk of mortality is independent of traditional CAD risk factors such as diabetes.
## Opium and HTN
long-term opium consumption may induce high blood pressure due to impact on coronary dysfunction, increase plasma homocysteine and fibrinogen levels, and consequent vascular narrowing [25]. Moreover, it has been shown that opium consumption does not improve hypertension [28]. According to our results, opium use was associated with increased risk of mortality in both patients with and without hypertension. However, the increase was significantly higher in patients without hypertension, possibly due to lack of screening and antihypertensive treatment in such patients.
## Opium and smoking
This study showed that opium consumption was associated with an increased risk of mortality in non-smokers. Similarly, a previous nested case-control study demonstrated that opium addiction was associated with increased risk of CAD in non-smokers; however, this association was not significant in smokers [29]. This finding would emphasize that opium consumption is a risk factor for CAD, independent of cigarette smoking.
## Opium and dyslipidemia
The possible mechanisms through which opium may affect blood lipids are decreased hepatic clearance of LDL and increased hepatic synthesis of triglycerides [30]. However, the studies regarding the impact of opium on lipid indices are conflicting. Several studies showed no significant association between opium addiction and lipid profile [31, 32], while some studies showed harmful effects of opium usage on lipid indices [33, 34]. According to our results, opium consumption was associated with an increased mortality risk in patients without dyslipidemia, but this association was not significant in those with baseline dyslipidemia. Definition of dyslipidemia, low-density lipoprotein level, and statin use are among the main determinants of this association, which are beyond the scope of the present study.
## Opium and age of CABG
As mentioned above, opium significantly decreased the age of CABG in both SMuRF-less and SMuRF groups. This finding could be due to the possible effects of opium on the atherosclerosis by decreasing plasma testosterone [14], increasing inflammation and pro-inflammatory cytokines [17], and suppressing autonomic nervous system and thus decreasing enkephalin production in cardiomyocytes [35] which all may happen regardless of baseline CAD risk factors. Our results are in line with two other studies which evaluated the relation between opium use and age of CAD event. Roohafza et al. [ 36] showed that opium use is associated with younger age of myocardial infarction, which was also emphasized later in a study by Hasandokht et al. [ 37].
## Limitations & strength
Evaluating the cause of death during follow-up was beyond the scope of this study; hence we cannot categorize cardiac and non-cardiac death. We do not exactly know opium reduces which type of death and it needs further studies with autopsy protocols. The absolute value of blood pressure (systolic and diastolic) was not available for all patients. The amount, duration, and type of opium use may impact our results, which none of the above mentioned variables registered in our databank. We could not capture the development of hypertension or diabetes after the initial visit. Another important limitation was that we did not have data regarding the patients’ socioeconomic status, which could influence mortality regardless of opium consumption. However, implanting IPW adjustment, large sample size, noticeable follow-up duration, and documentation of cardiovascular events and mortalities are among the strength of the present study.
## Conclusion
In patients without any modifiable CAD risk factors,opium consumption was associated with higher all-cause mortality but was not associated with more MACCE and non-fatal MACCE. Opium users were younger than their counterparts which indirectly emphasizes the role of opium on earlier CAD occurrence.
## Data Availability Statement
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
## Ethics statement
The present study was approved by the ethical board of THC (IR-THC-13799) and the involved human data was in accordance with the Helsinki Declaration. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
AS, AF, and SN wrote the main manuscript text. MN, SS, MP prepared figures. MS, ASO, SM, JB performed the CABG and were involved in the recruitment of the patients and development of the study. KH edited the text and developed the idea of the study. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fsurg.2023.1047807/full#supplementary-material.
## References
1. Kannel WB, Dawber TR, Kagan A, Revotskie N, Stokes J. **Factors of risk in the development of coronary heart disease–six year follow-up experience. The framingham study**. *Ann Intern Med* (1961) **55** 33-50. DOI: 10.7326/0003-4819-55-1-33
2. Roth GA, Mensah GA, Johnson CO, Addolorato G, Ammirati E, Baddour LM. **Global burden of cardiovascular diseases and risk factors, 1990-2019: update from the GBD 2019 study**. *J Am Coll Cardiol* (2020) **76** 2982-3021. DOI: 10.1016/j.jacc.2020.11.010
3. Figtree GA, Vernon ST, Hadziosmanovic N, Sundström J, Alfredsson J, Arnott C. **Mortality in STEMI patients without standard modifiable risk factors: a sex-disaggregated analysis of SWEDEHEART registry data**. *Lancet* (2021) **397** 1085-94. DOI: 10.1016/S0140-6736(21)00272-5
4. Scheuner MT. **Genetic predisposition to coronary artery disease**. *Curr Opin Cardiol* (2001) **16** 251-60. DOI: 10.1097/00001573-200107000-00006
5. Vernon ST, Coffey S, D'Souza M, Chow CK, Kilian J, Hyun K. **ST-Segment-Elevation myocardial infarction (STEMI) patients without standard modifiable cardiovascular risk factors-how common are they, and what are their outcomes?**. *J Am Heart Assoc* (2019) **8** e013296. DOI: 10.1161/JAHA.119.013296
6. Sadeghian S, Dowlatshahi S, Karimi A, Tazik M. **Epidemiology of opium use in 4398 patients admitted for coronary artery bypass graft in Tehran heart center**. *J Cardiovasc Surg (Torino)* (2011) **52** 140-1. PMID: 21224824
7. Kulsudjarit K. **Drug problem in southeast and southwest Asia**. *Ann N Y Acad Sci* (2004) **1025** 446. DOI: 10.1196/annals.1316.055
8. Safaii N, Kazemi B. **Effect of opium use on short-term outcome in patients undergoing coronary artery bypass surgery**. *Gen Thorac Cardiovasc Surg* (2010) **58** 62-7. DOI: 10.1007/s11748-009-0529-7
9. Masoudkabir F, Yavari N, Pashang M, Sadeghian S, Jalali A, Poorhosseini H. **Effect of persistent opium consumption after surgery on the long-term outcomes of surgical revascularisation**. *Eur J Prev Cardiol* (2020) **27** 1996-2003. DOI: 10.1177/2047487320932010
10. Nemati MH, Astaneh B, Ardekani GS. **Effects of opium addiction on bleeding after coronary artery bypass graft surgery: report from Iran**. *Gen Thorac Cardiovasc Surg* (2010) **58** 456-60. DOI: 10.1007/s11748-010-0613-z
11. Poorhosseini H, Abbasi SH. **The Tehran heart center**. *Eur Heart J* (2018) **39** 2695-6. DOI: 10.1093/eurheartj/ehy369
12. **Standards of medical care in diabetes—2014**. *Diabetes Care* (2014) **37** S14-80. DOI: 10.2337/dc14-S014
13. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **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**. *J Am Coll Cardiol* (2018) **71** e127-248. DOI: 10.1016/j.jacc.2017.11.006
14. George S, Murali V, Pullickal R. **Review of neuroendocrine correlates of chronic opiate misuse: dysfunctions and pathophysiological mechanisms**. *Addict Disord Their Treat* (2005) **4** 99-109. DOI: 10.1097/01.adt.0000161633.63378.fb
15. Gonzalez Del Hoyo MI, Peiro Ibanez OM, Vaquez-Nunez K, Dominguez Benito F, Ferrero M, Romeu A. **The absence of standard modifiable cardiovascular risk factors does not predict better outcomes in patients with acute coronary syndrome**. *Eur Heart J* (2020) **41**. DOI: 10.1093/ehjci/ehaa946.1341
16. Nalini M, Shakeri R, Poustchi H, Pourshams A, Etemadi A, Islami F. **Long-term opiate use and risk of cardiovascular mortality: results from the golestan cohort study**. *Eur J Prev Cardiol* (2020) **28** 98-106. DOI: 10.1093/eurjpc/zwaa006
17. Asadikaram G, Igder S, Jamali Z, Shahrokhi N, Najafipour H, Shokoohi M. **Effects of different concentrations of opium on the secretion of interleukin-6, interferon-**. *Addict Health* (2015) **7** 47-53. PMID: 26322210
18. Phillips GB, Pinkernell BH, Jing TY. **The association of hypotestosteronemia with coronary artery disease in men**. *Arterioscler Thromb* (1994) **14** 701-6. DOI: 10.1161/01.ATV.14.5.701
19. Pappa T, Alevizaki M. **MECHANISMS IN ENDOCRINOLOGY: endogenous sex steroids and cardio- and cerebro-vascular disease in the postmenopausal period**. *Eur J Endocrinol* (2012) **167** 145-56. DOI: 10.1530/EJE-12-0215
20. Azdaki N, Zardast M, Anani-Sarab G, Abdorrazaghnaejad H, Ghasemian MR, Saburi A. **Comparison between homocysteine, fibrinogen, PT, PTT, INR and CRP in male smokers with/without addiction to opium**. *Addict Health* (2017) **9** 17-23. PMID: 29026499
21. Asgary S, Sarrafzadegan N, Naderi G-A, Rozbehani R. **Effect of opium addiction on new and traditional cardiovascular risk factors: do duration of addiction and route of administration matter?**. *Lipids Health Dis* (2008) **7** 42. DOI: 10.1186/1476-511X-7-42
22. Das B, Daga MK, Gupta SK. **Lipid pentad Index: a novel bioindex for evaluation of lipid risk factors for atherosclerosis in young adolescents and children of premature coronary artery disease patients in India**. *Clin Biochem* (2007) **40** 18-24. DOI: 10.1016/j.clinbiochem.2006.08.016
23. Singh SK, Suresh MV, Voleti B, Agrawal A. **The connection between C-reactive protein and atherosclerosis**. *Ann Med* (2008) **40** 110-20. DOI: 10.1080/07853890701749225
24. Rudnicka AR, Mt-Isa S, Meade TW. **Associations of plasma fibrinogen and factor VII clotting activity with coronary heart disease and stroke: prospective cohort study from the screening phase of the thrombosis prevention trial**. *J Thromb Haemost* (2006) **4** 2405-10. DOI: 10.1111/j.1538-7836.2006.02221.x
25. Najafipour H, Beik A. **The impact of opium consumption on blood glucose, Serum lipids and blood pressure, and related mechanisms**. *Front Physiol* (2016) **7** 436. DOI: 10.3389/fphys.2016.00436
26. May CN, Ham IW, Heslop KE, Stone FA, Mathias CJ. **Intravenous morphine causes hypertension, hyperglycaemia and increases sympatho-adrenal outflow in conscious rabbits**. *Clin Sci* (1988) **75** 71-7. DOI: 10.1042/cs0750071
27. Zandomeneghi R, Luciani A, Massari M, Montanari P, Pavesi C. **Effects of heroin addiction on the responses of glucose, C-peptide and insulin to a standard meal**. *Clin Sci* (1988) **74** 283-8. DOI: 10.1042/cs0740283
28. Ziaee M, Hajizadeh R, Khorrami A, Sepehrvand N, Momtaz S, Ghaffari S. **Cardiovascular complications of chronic opium consumption: a narrative review article**. *Iran J Public Health* (2019) **48** 2154-64. DOI: 10.18502/ijph.v48i12.3546
29. Masoomi M, Ramezani MA, Karimzadeh H. **The relationship of opium addiction with coronary artery disease**. *Int J Prev Med* (2010) **1** 182-6. PMID: 21566789
30. Bryant HU, Kuta CC, Story JA, Yim GK. **Stress- and morphine-induced elevations of plasma and tissue cholesterol in mice: reversal by naltrexone**. *Biochem Pharmacol* (1988) **37** 3777-80. DOI: 10.1016/0006-2952(88)90415-7
31. Sanli DB, Bilici R, Suner O, Citak S, Kartkaya K, Mutlu FS. **Effect of different psychoactive substances on Serum biochemical parameters**. *Int J High Risk Behav Addict* (2015) **4** e22702. DOI: 10.5812/ijhrba.22702
32. Mohammadali B, Sepideh N, Mohammadreza Khosoosi N, Mirsaeid R, Afshin K. **Opium consumption and lipid and glucose parameters in diabetic patients with acute coronary syndrome: a survey in northern Iran**. *Tunis Med* (2014) **92** 497-500. PMID: 25775291
33. Rahimi N, Gozashti MH, Najafipour H, Shokoohi M, Marefati H. **Potential effect of opium consumption on controlling diabetes and some cardiovascular risk factors in diabetic patients**. *Addict Health* (2014) **6** 1-6. PMID: 25140211
34. Aghadavoudi O, Eizadi-Mood N, Najarzadegan MR. **Comparing cardiovascular factors in opium abusers and non-users candidate for coronary artery bypass graft surgery**. *Adv Biomed Res* (2015) **4** 12. DOI: 10.4103/2277-9175.148294
35. Barron BA. **Cardiac opioids**. *Proc Soc Exp Biol Med Soc Exp Biol Med* (2000) **224** 1-7. DOI: 10.1046/j.1525-1373.2000.22358.x
36. Roohafza H, Talaei M, Sadeghi M, Haghani P, Shokouh P, Sarrafzadegan N. **Opium decreases the age at myocardial infarction and sudden cardiac death: a long- and short-term outcome evaluation**. *Arch Iran Med* (2013) **16** 154-60. PMID: 23432167
37. Hasandokht T, Salari A, Pour S, Tirani H, Shad B, Rajabi E. **Does opium have benefit for coronary artery disease? A systematic review**. *Res Cardiovasc Med* (2018) **7** 51-8. DOI: 10.4103/rcm.rcm_12_17
|
---
title: Association between smoking cessation and non-alcoholic fatty liver disease
using NAFLD liver fat score
authors:
- Yun Seo Jang
- Hye Jin Joo
- Yu Shin Park
- Eun-Cheol Park
- Sung-In Jang
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9982128
doi: 10.3389/fpubh.2023.1015919
license: CC BY 4.0
---
# Association between smoking cessation and non-alcoholic fatty liver disease using NAFLD liver fat score
## Abstract
### Background
Smoking is well known to be associated with a higher prevalence and incidence of liver diseases such as advanced fibrosis. However, the impact of smoking on developing nonalcoholic fatty liver disease remains controversial, and clinical data on this is limited. Therefore, this study aimed to investigate the association between smoking history and nonalcoholic fatty liver disease (NAFLD).
### Methods
Data from the Korea National Health and Nutrition Examination Survey 2019-2020 were used for the analysis. NAFLD was diagnosed according to an NAFLD liver fat score of >-0.640. Smoking status was classified as into nonsmokers, ex-smokers, and current smokers. Multiple logistic regression analysis was conducted to examine the association between smoking history and NAFLD in the South Korean population.
### Results
In total, 9,603 participants were enrolled in this study. The odds ratio (OR) for having NAFLD in ex-smokers and current smokers in males was 1.12 ($95\%$ confidence interval [CI]: 0.90–1.41) and 1.38 ($95\%$ CI: 1.08–1.76) compared to that in nonsmokers, respectively. The OR increased in magnitude with smoking status. Ex-smokers who ceased smoking for <10 years (OR: 1.33, $95\%$ CI: 1.00–1.77) were more likely to have a strong correlation with NAFLD. Furthermore, NAFLD had a dose-dependent positive effect on pack-years, which was 10 to 20 (OR: 1.39, $95\%$ CI: 1.04–1.86) and over 20 (OR: 1.51, $95\%$ CI: 1.14–2.00).
### Conclusion
This study found that smoking may contribute to NAFLD. Our study suggests cessation of smoking may help management of NAFLD.
## Introduction
Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease. It is a condition in which neutral fat accumulates excessively in the liver [1, 2]. Although there are some differences in its frequency from country to country, it has been reported that 6.3 to $33\%$ and an average of approximately $20\%$ of patients worldwide have been affected by the disease [3]. The prevalence of NAFLD is rapidly increasing in Asian countries due to the increase in Westernized eating habits, obesity, and the diabetic population [4, 5]. In addition, between 10 and $29\%$ of patients with nonalcoholic fatty hepatitis develop cirrhosis within 10 years and between 4 and $27\%$ of patients develop liver cancer [6, 7]. Furthermore, patients with NAFLD have a higher mortality rate than healthy controls, and the mortality rate related to liver disease is also high (8–11). Therefore, NAFLD must be managed immediately due to its expected serious public health burden and significant social costs [12, 13].
Tobacco smoke contains more than 7,000 chemicals, of which at least 250 are known to be harmful, such as ammonia and hydrogen cyanide [14, 15]. Smoking is closely related to chronic diseases, such as cardiovascular diseases, cancer, and type 2 diabetes (16–19), which are also related to NAFLD (20–22). Previous studies have suggested smoking is associated with increased prevalence and incidence of liver diseases [23, 24]. In particular, it has been reported to be an independent risk factor for the progression of advanced fibrosis in patients with primary biliary cirrhosis [23] and chronic hepatitis C [24].
A positive association between smoking and NAFLD has been continuously reported (25–27). An experimental study suggested cigarettes accelerated the progression of NAFLD in obese mice-fed diets [25]. Furthermore, a study conducted in mice without apolipoprotein E, a condition wherein fatty liver is easily occurs, found that nicotine in electronic cigarettes (e-cigarettes) causes genetic mutations and promotes NAFLD outbreaks [26]. Other studies have shown that the activation of sterol regulatory element-binding proteins (SREBPs), which stimulate the synthesis of fatty acids in the liver, is associated with NAFLD [27]. These studies provided evidence of the mechanism of the relationship between smoking and the prevalence of NAFLD. However, most studies are experimental studies conducted on animals, and there are not many studies conducted on humans.
Therefore, this study aimed to examine the association between smoking history and NAFLD in a representative population and to explain whether smoking behavior plays a potential role in developing NAFLD.
## Data
The study used cross-sectional data from the 2019–2020 National Health and Nutrition Examination Survey (KNHANES), conducted by the Korea Centers for Disease Control and Prevention Agency (KDCA). The KNAHENS is a self-report survey using a stratified, multistage, cluster sampling design conducted annually for South Koreans of all ages to evaluate the health and nutritional status. The survey provides data for the evaluation and development of health policies and programs and does not require ethical approval from the ethics review board, as the KNHANES conforms to the Declaration of Helsinki.
## Study population
Of the 15,469 survey participants, we excluded those under 19 years of age and those who did not participate in a KNHANES smoking questionnaire survey ($$n = 2$$,730). Furthermore, participants who tested positive for serologic markers for liver diseases (hepatitis B, hepatitis C, and liver cirrhosis) were excluded ($$n = 437$$). Participants with missing data were also excluded ($$n = 2$$,699). Consequently, a final sample of 9,603 participants was analyzed in this study (Figure 1). As a study that examined the effects of smoking on NAFLD, participants with alcohol-related fatty liver disease were also excluded based on their biochemical and clinical profiles.
**Figure 1:** *Flowchart of the study participants showing the inclusion and exclusion.*
## Variables
The main dependent variable was the prevalence of NAFLD. NAFLD was diagnosed according to the NAFLD liver fat score developed by the Department of Medicine and the Minerva Medical Research Institute at Helsinki University [28]. The NAFLD liver fat score formula was derived using a multivariate logistic regression model using metabolic syndrome, type 2 diabetes, fasting insulin (fS), serum aspartate aminotransferase (AST) ratio, and AST to serum alanine aminotransferase (ALT) ratio [28]: NAFLD liver fat score = −0.89 + 1.18 × metabolic syndrome (yes = 1 / no = 0) + 0.45 × diabetes (yes = 2 / no = 0) + 0.15 × fS-insulin (mu/L) + 0.04 × fS-AST (U/L) – 0.94 × AST/ALT. Participants were considered to have NAFLD if their liver fat score of NAFLD was >−0.640 as the optimal cutoff point [28].
The primary independent variable was the smoking status of the participants, which was divided into three groups: [1] nonsmokers, [2] ex-smokers, and [3] current smokers. This was defined based on the questions: 'Do you currently smoke conventional cigarettes?'; “ Do you currently smoke e-cigarettes?”. This classification was the same as that of a previous study that used the same research tool to investigate smoking behavior [29].
The covariates included demographic factors (sex, age, marital status, and educational level), socioeconomic factors (household income, region, and occupational categories), behavioral health patterns (current drinking status, physical activity), and health-related factors (body mass index (BMI), diagnosis of hypertension, and diagnosis of diabetes).
## Statistical analysis
All estimates were calculated using sample weight procedures to improve representativeness and generalize the data. Clusters and strata were assigned to the study population. *The* general characteristics of the study group, represented by frequencies and percentages for categorical variables, means and standard deviations for continuous variables, were based on descriptive analysis. After adjusting for covariates, a multiple logistic regression analysis was performed to assess the relationship between smoking and NAFLD. Subgroup analyzes were also performed according to age, current drinking status, physical activity, BMI, and diagnosis of hypertension and diabetes. Furthermore, we also performed a subgroup analysis for a more complete analysis of smoking behavior, including smoking cessation status (SCS) and pack years. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).
## Results
Table 1 shows the characteristics of the study population. Of the 9,603 participants, 4,063 were men ($42.3\%$) and 5,540 were women ($57.7\%$). Among males, 1,249 ($30.7\%$) were current smokers, 1,674 ($41.2\%$) were ex-smokers, and 1,140 ($28.1\%$) were nonsmokers. Among the females, 259 ($4.7\%$) were current smokers, 312 ($5.6\%$) were ex-smokers, and 4,969 ($89.7\%$) were nonsmokers. In total, 1,433 ($35.3\%$) men and 1,278 ($23.1\%$) women reported NAFLD.
**Table 1**
| Variables | Variables.1 | Nonalcoholic fatty liver disease (NAFLD) | Nonalcoholic fatty liver disease (NAFLD).1 | Nonalcoholic fatty liver disease (NAFLD).2 | Nonalcoholic fatty liver disease (NAFLD).3 | Nonalcoholic fatty liver disease (NAFLD).4 | Nonalcoholic fatty liver disease (NAFLD).5 | Nonalcoholic fatty liver disease (NAFLD).6 | Nonalcoholic fatty liver disease (NAFLD).7 | Nonalcoholic fatty liver disease (NAFLD).8 | Nonalcoholic fatty liver disease (NAFLD).9 | Nonalcoholic fatty liver disease (NAFLD).10 | Nonalcoholic fatty liver disease (NAFLD).11 | Nonalcoholic fatty liver disease (NAFLD).12 | Nonalcoholic fatty liver disease (NAFLD).13 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | Male | Male | Male | Male | Male | Male | Male | Female | Female | Female | Female | Female | Female | Female |
| | | Total | Total | Yes | Yes | No | No | P- value | Total | Total | Yes | Yes | No | No | P- value |
| | | N | % | N | % | N | % | | N | % | N | % | N | % | |
| Total ( N =9,603) | Total ( N =9,603) | 4063 | 100.0 | 1433 | 35.3 | 2630 | 64.7 | | 5540 | 100.0 | 1278 | 23.1 | 4262 | 76.9 | |
| Smoking Behavior | Smoking Behavior | | | | | | | 0.0014 | | | | | | | 0.5628 |
| | Nonsmoker | 1140 | 28.1 | 354 | 31.1 | 786 | 68.9 | | 4969 | 89.7 | 1156 | 23.3 | 3813 | 76.7 | |
| | Ex-smoker | 1674 | 41.2 | 629 | 37.6 | 1045 | 62.4 | | 312 | 5.6 | 65 | 20.8 | 247 | 79.2 | |
| | Current smoker | 1249 | 30.7 | 450 | 36.0 | 799 | 64.0 | | 259 | 4.7 | 57 | 22.0 | 202 | 78.0 | |
| Age (Mean, SD) | Age (Mean, SD) | 51.6 | 17.2 | 52.2 | 15.9 | 51.3 | 17.9 | < 0.0001 | 51.8 | 16.4 | 49.8 | 16.3 | 58.6 | 14.9 | < 0.0001 |
| Marital status | | | | | | | 0.0185 | | | | | | | 0.1441 | |
| | Married | 2884 | 71.0 | 1043 | 36.2 | 1841 | 63.8 | | 3664 | 66.1 | 835 | 22.8 | 2829 | 77.2 | |
| | Divorced, Separated | 166 | 4.1 | 67 | 40.4 | 99 | 59.6 | | 343 | 6.2 | 94 | 27.4 | 249 | 72.6 | |
| | Single, widow | 1013 | 24.9 | 323 | 31.9 | 690 | 68.1 | | 1533 | 27.7 | 349 | 22.8 | 1184 | 77.2 | |
| Educational level | Educational level | | | | | | | 0.8249 | | | | | | | < 0.0001 |
| | Middle school or below | 878 | 21.6 | 317 | 36.1 | 561 | 63.9 | | 1715 | 31.0 | 621 | 36.2 | 1094 | 63.8 | |
| | High school | 1472 | 36.2 | 513 | 34.9 | 959 | 65.1 | | 1787 | 32.3 | 381 | 21.3 | 1406 | 78.7 | |
| | College or over | 1713 | 42.2 | 603 | 35.2 | 1110 | 64.8 | | 2038 | 36.8 | 276 | 13.5 | 1762 | 86.5 | |
| Household income | Household income | | | | | | | 0.6334 | | | | | | | < 0.0001 |
| | Low | 651 | 16.0 | 224 | 34.4 | 427 | 65.6 | | 1046 | 18.9 | 353 | 33.7 | 693 | 66.3 | |
| | Mid-low | 985 | 24.2 | 364 | 37.0 | 621 | 63.0 | | 1360 | 24.5 | 334 | 24.6 | 1026 | 75.4 | |
| | Mid-high | 1129 | 27.8 | 396 | 35.1 | 733 | 64.9 | | 1488 | 26.9 | 297 | 20.0 | 1191 | 80.0 | |
| | High | 1298 | 31.9 | 449 | 34.6 | 849 | 65.4 | | 1646 | 29.7 | 294 | 17.9 | 1352 | 82.1 | |
| Region | Region | | | | | | | 0.2872 | | | | | | | < 0.0001 |
| | Metropolitan | 1720 | 42.3 | 584 | 34.0 | 1136 | 66.0 | | 2466 | 44.5 | 502 | 20.4 | 1964 | 79.6 | |
| | Urban | 1505 | 37.0 | 540 | 35.9 | 965 | 64.1 | | 2034 | 36.7 | 459 | 22.6 | 1575 | 77.4 | |
| | Rural | 838 | 20.6 | 309 | 36.9 | 529 | 63.1 | | 1040 | 18.8 | 317 | 30.5 | 723 | 69.5 | |
| Occupational categories | Occupational categories | | | | | | | 0.6403 | | | | | | | < 0.0001 |
| | White | 1155 | 28.4 | 422 | 36.5 | 733 | 63.5 | | 1263 | 22.8 | 174 | 13.8 | 1089 | 86.2 | |
| | Pink | 404 | 9.9 | 137 | 33.9 | 267 | 66.1 | | 835 | 15.1 | 191 | 22.9 | 644 | 77.1 | |
| | Blue | 1308 | 32.2 | 449 | 34.3 | 859 | 65.7 | | 822 | 14.8 | 217 | 26.4 | 605 | 73.6 | |
| | Inoccupation | 1196 | 29.4 | 425 | 35.5 | 771 | 64.5 | | 2620 | 47.3 | 696 | 26.6 | 1924 | 73.4 | |
| Current drinking status | Current drinking status | | | | | | | 0.0315 | | | | | | | < 0.0001 |
| | Never or occasionally | 1273 | 31.3 | 435 | 34.2 | 838 | 65.8 | | 3310 | 59.7 | 892 | 26.9 | 2418 | 73.1 | |
| | 2–4 times/month | 1478 | 36.4 | 498 | 33.7 | 980 | 66.3 | | 1633 | 29.5 | 297 | 18.2 | 1336 | 81.8 | |
| | 2–4 times/week | 1312 | 32.3 | 500 | 38.1 | 812 | 61.9 | | 597 | 10.8 | 89 | 14.9 | 508 | 85.1 | |
| Physical activity | Physical activity | | | | | | | < 0.0001 | | | | | | | < 0.0001 |
| | Adequate | 1906 | 46.9 | 613 | 32.2 | 1293 | 67.8 | | 2216 | 40.0 | 425 | 19.2 | 1791 | 80.8 | |
| | Inadequate | 2157 | 53.1 | 820 | 38.0 | 1337 | 62.0 | | 3324 | 60.0 | 853 | 25.7 | 2471 | 74.3 | |
| BMI | BMI | | | | | | | < 0.0001 | | | | | | | < 0.0001 |
| | Normal | 1155 | 28.4 | 132 | 11.4 | 1023 | 88.6 | | 2494 | 45.0 | 175 | 7.0 | 2319 | 93.0 | |
| | Underweight | 92 | 2.3 | 3 | 3.3 | 89 | 96.7 | | 272 | 4.9 | 6 | 2.2 | 266 | 97.8 | |
| | Overweight | 1069 | 26.3 | 274 | 25.6 | 795 | 74.4 | | 1116 | 20.1 | 268 | 24.0 | 848 | 76.0 | |
| | Obesity of stage 1 | 1472 | 36.2 | 796 | 54.1 | 676 | 45.9 | | 1351 | 24.4 | 597 | 44.2 | 754 | 55.8 | |
| | Obesity of stages 2&3 | 275 | 6.8 | 228 | 82.9 | 47 | 17.1 | | 307 | 5.5 | 232 | 75.6 | 75 | 24.4 | |
| Diagnosis of hypertension | Diagnosis of hypertension | | | | | | | < 0.0001 | | | | | | | < 0.0001 |
| | Yes | 1105 | 27.2 | 530 | 48.0 | 575 | 52.0 | | 1290 | 23.3 | 577 | 44.7 | 713 | 55.3 | |
| | No | 2958 | 72.8 | 903 | 30.5 | 2055 | 69.5 | | 4250 | 76.7 | 701 | 16.5 | 3549 | 83.5 | |
| Diagnosis of diabetes | Diagnosis of diabetes | | | | | | | < 0.0001 | | | | | | | < 0.0001 |
| | Yes | 466 | 11.5 | 299 | 64.2 | 167 | 35.8 | | 500 | 9.0 | 339 | 67.8 | 161 | 32.2 | |
| | No | 3597 | 88.5 | 1134 | 31.5 | 2463 | 68.5 | | 5040 | 91.0 | 939 | 18.6 | 4101 | 81.4 | |
| Year | Year | | | | | | | 0.0018 | | | | | | | 0.1354 |
| | 2019 | 2088 | 51.4 | 689 | 33.0 | 1399 | 67.0 | | 2932 | 52.9 | 653 | 22.3 | 1983 | 77.7 | |
| | 2020 | 1975 | 48.6 | 744 | 37.7 | 1231 | 62.3 | | 2608 | 47.1 | 625 | 24.0 | 2279 | 76.0 | |
Table 2 presents the results of the multiple regression analysis for the relationship between smoking and NAFLD stratified by sex after adjusting for all covariates. Among male participants, the odds ratios (OR) for NAFLD among ex-smokers and current smokers were 1.12 ($95\%$ confidence interval [CI]: 0.90–1.41) and 1.38 ($95\%$ CI: 1.08–1.76), respectively. In women, the OR for NAFLD among ex-smokers and current smokers were 1.32 ($95\%$ CI: 0.86–2.01) and 1.18 ($95\%$ CI: 0.76–1.83), respectively. Ex-smokers and current smokers exhibited an increasing trend of OR for NAFLD compared to that in nonsmokers, although there were statistically significant associations only in current smokers among males.
**Table 2**
| Variables | Variables.1 | Nonalcoholic fatty liver disease (NAFLD) | Nonalcoholic fatty liver disease (NAFLD).1 | Nonalcoholic fatty liver disease (NAFLD).2 | Nonalcoholic fatty liver disease (NAFLD).3 |
| --- | --- | --- | --- | --- | --- |
| | | Male | Male | Female | Female |
| | | OR | 95% CI | OR | 95% CI |
| Smoking Behavior | Smoking Behavior | | | | |
| | Nonsmoker | 1.00 | | 1.00 | |
| | Ex-smoker | 1.12 | (0.90 – 1.41) | 1.32 | (0.86 – 2.01) |
| | Current smoker | 1.38 | (1.08 – 1.76) | 1.18 | (0.76 – 1.83) |
| Age Marital status | Age Marital status | 1.00 | (1.00 – 1.01) | 1.01 | (1.00 – 1.83) |
| | Married | 1.00 | | 1.00 | |
| | Divorced, Separated | 1.29 | (0.80 – 2.07) | 1.31 | (0.93 – 1.84) |
| | Single, widow | 0.84 | (0.64 – 1.10) | 0.91 | (0.72 – 1.13) |
| Educational level | Educational level | | | | |
| | Middle school or below | 1.00 | | 1.00 | |
| | High school | 1.07 | (0.79 – 1.44) | 1.03 | (0.78 – 1.36) |
| | College or over | 1.06 | (0.77 – 1.45) | 0.73 | (0.53 – 1.01) |
| Household income | Household income | | | | |
| | Low | 1.04 | (0.74 – 1.44) | 0.83 | (0.60 – 1.16) |
| | Mid-low | 1.17 | (0.92 – 1.49) | 0.73 | (0.55 – 0.95) |
| | Mid-high | 0.97 | (0.78 – 1.21) | 0.68 | (0.51 – 0.89) |
| | High | 1.00 | | 1.00 | |
| Region | Region | | | | |
| | Metropolitan | 1.00 | | 1.00 | |
| | Urban | 1.06 | (0.87 – 1.30) | 1.12 | (0.91 – 1.38) |
| | Rural | 1.08 | (0.85 – 1.38) | 1.34 | (1.01 – 1.77) |
| Occupational categories | Occupational categories | | | | |
| | White | 0.88 | (0.65 – 1.19) | 0.81 | (0.61 – 1.07) |
| | Pink | 0.78 | (0.55 – 1.09) | 0.94 | (0.72 – 1.23) |
| | Blue | 0.71 | (0.55 – 0.92) | 0.63 | (0.48 – 0.82) |
| | Inoccupation | 1.00 | | 1.00 | |
| Current drinking status | Current drinking status | | | | |
| | Never or occasionally | 1.00 | | 1.00 | |
| | 2–4 times/month | 0.89 | (0.72 – 1.09) | 0.85 | (0.70 – 1.04) |
| | 2–4 times/week | 1.06 | (0.86 – 1.31) | 0.60 | (0.43 – 0.84) |
| Physical activity | Physical activity | | | | |
| | Adequate | 1.00 | | 1.00 | |
| | Inadequate | 1.44 | (1.20 – 1.71) | 1.33 | (1.11 – 1.59) |
| BMI | BMI | | | | |
| | Normal | 1.00 | | 1.00 | |
| | Underweight | 0.50 | (0.14 – 1.78) | 0.44 | (0.17 – 1.15) |
| | Overweight | 2.84 | (2.09 – 3.86) | 4.21 | (3.28 – 5.41) |
| | Obesity of stage 1 | 10.44 | (7.94 – 13.72) | 11.48 | (8.92 – 14.78) |
| | Obesity of stages 2&3 | 50.57 | (32.58 – 78.48) | 62.42 | (41.29 – 94.35) |
| | No | 1.00 | | 1.00 | |
| | Yes | 1.42 | (1.14 – 1.77) | 1.96 | (1.55 – 2.48) |
| Diagnosis of diabetes | Diagnosis of diabetes | | | | |
| | No | 1.00 | | 1.00 | |
| | Yes | 4.36 | (3.28 – 5.80) | 6.06 | (4.41 – 8.31) |
| Year | Year | | | | |
| | 2019 | 1.00 | | 1.00 | |
| | 2020 | 1.06 | (0.88 – 1.27) | 0.95 | (0.79 – 1.15) |
Figure 2 presents the results of the stratified subgroup analysis of the association between SCS and pack years, indicating the effect of the number of cigarettes and the smoking period on NAFLD according to smoking behavior. *In* general, with nonsmokers as the reference category, the OR for NAFLD increased linearly as smoking cessation decreased and pack years increased in males. Specifically, an ex-smoker with smoking cessation for < 10 years (OR: 1.33, $95\%$ CI: 1.00–1.77) and a current smoker (OR: 1.38, $95\%$ CI: 1.08–1.76) had the strongest statistically significant association compared to a nonsmoker, as classified based on the smoking cessation period. Furthermore, an ex-smoker and current smoker with 10 to 20 pack years (OR: 1.39, $95\%$ CI: 1.04–1.86) and over 20 pack years (OR: 1.51, $95\%$ CI: 1.14–2.00), respectively, was more likely to have a strong relationship with NAFLD compared to a nonsmoker.
**Figure 2:** *Results of the subgroup analysis stratified by smoking cessation and pack-years.*
Table 3 shows the results of the independent variable subgroup analysis, representing the ORs for NAFLD stratified by the smoking status. Among current male smokers, cases of never or occasional drinking (OR: 1.78, $95\%$ CI: 1.14–2.78), adequate physical activity (OR: 1.55, $95\%$ CI: 1.09–2.21), BMI indicating overweight (OR: 2.31, $95\%$ CI: 1.40–3.83), no diagnosis of hypertension (OR: 1.42, $95\%$ CI: 1.07–1.87), and no diagnosis of diabetes (OR: 1.39, $95\%$ CI: 1.08–1.79) showed the strongest associations with NAFLD compared to male nonsmokers. In women, drinking 2 to 4 times per month (current smokers: OR: 1.39, $95\%$ CI: 1.08–1.79), normal BMI (ex-smokers: OR: 2.74, $95\%$ CI: 1.28–5.88), and BMI indicating stage 2 and 3 obesity (ex-smokers: OR: 4.36, $95\%$ CI: 1.14–16.71) showed the strongest associations with NAFLD compared to those in nonsmokers.
**Table 3**
| Unnamed: 0 | Unnamed: 1 | Nonalcoholic fatty liver disease (NAFLD) | Nonalcoholic fatty liver disease (NAFLD).1 | Nonalcoholic fatty liver disease (NAFLD).2 | Nonalcoholic fatty liver disease (NAFLD).3 | Nonalcoholic fatty liver disease (NAFLD).4 | Nonalcoholic fatty liver disease (NAFLD).5 | Nonalcoholic fatty liver disease (NAFLD).6 | Nonalcoholic fatty liver disease (NAFLD).7 | Nonalcoholic fatty liver disease (NAFLD).8 | Nonalcoholic fatty liver disease (NAFLD).9 | Nonalcoholic fatty liver disease (NAFLD).10 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | Male | Male | Male | Male | Male | | Female | Female | Female | Female | Female |
| | | Non | Ex-smoker | Ex-smoker | Current smoker | Current smoker | | Non | Ex-smoker | Ex-smoker | Current smoker | Current smoker |
| | | OR | OR | 95% CI | OR | 95% CI | | OR | OR | 95% CI | OR | 95% CI |
| Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age |
| | 20–29 | 1.00 | 0.53 | (0.21 – 1.37) | 1.06 | (0.54 – 2.09) | | 1.00 | 0.82 | (0.18 – 3.63) | 1.12 | (0.32 – 3.92) |
| | 30–39 | 1.00 | 0.71 | (0.33 – 1.51) | 1.18 | (0.65 – 2.16) | | 1.00 | 1.68 | (0.41 – 6.84) | 0.62 | (0.15 – 2.51) |
| | 40–49 | 1.00 | 1.35 | (0.71 – 2.58) | 1.47 | (0.76 – 2.86) | | 1.00 | 1.40 | (0.61 – 3.22) | 0.74 | (0.26 – 2.13) |
| | 50–59 | 1.00 | 1.14 | (0.62 – 2.08) | 1.38 | (0.72 – 2.62) | | 1.00 | 2.67 | (1.06 – 6.73) | 3.16 | (1.12 – 8.86) |
| | 60–69 | 1.00 | 1.58 | (0.94 – 2.67) | 1.51 | (0.83 – 2.76) | | 1.00 | 0.78 | (0.33 – 1.86) | 1.07 | (0.44 – 2.63) |
| | ≥70 | 1.00 | 0.98 | (0.58 – 1.66) | 0.99 | (0.43 – 2.27) | | 1.00 | 0.71 | (0.18 – 2.85) | 0.38 | (0.10 – 1.39) |
| Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status | Current drinking status |
| | Never or occasionally | 1.00 | 0.97 | (0.64 – 1.46) | 1.78 | (1.14 – 2.78) | | 1.00 | 1.12 | (0.60 – 2.06) | 0.61 | (0.33 – 1.13) |
| | 2–4 times/month | 1.00 | 1.05 | (0.74 – 1.50) | 1.21 | (0.84 – 1.76) | | 1.00 | 1.30 | (0.59 – 2.87) | 2.28 | (1.14 – 4.56) |
| | 2–4 times/week | 1.00 | 1.52 | (0.95 – 2.43) | 1.56 | (1.01 – 2.42) | | 1.00 | 2.56 | (0.95 – 6.95) | 0.91 | (0.36 – 2.31) |
| Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity |
| | Adequate | 1.00 | 1.07 | (0.77 – 1.50) | 1.55 | (1.09 – 2.21) | | 1.00 | 1.28 | (0.67 – 2.44) | 0.99 | (0.48 – 2.03) |
| | Inadequate | 1.00 | 1.15 | (0.84 – 1.58) | 1.28 | (0.90 – 1.83) | | 1.00 | 1.33 | (0.78 – 2.28) | 1.27 | (0.74 – 2.19) |
| BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI |
| | Underweight | 1.00 | – | – | – | – | – | 1.00 | – | – | – | – |
| | Normal | 1.00 | 1.32 | (0.74 – 2.34) | 1.14 | (0.57 – 2.25) | | 1.00 | 2.74 | (1.28 – 5.88) | 2.16 | (0.96 – 4.85) |
| | Overweight | 1.00 | 1.63 | (0.96 – 2.75) | 2.31 | (1.40 – 3.83) | | 1.00 | 0.79 | (0.25 – 2.49) | 1.04 | (0.41 – 2.65) |
| | Obesity of stage 1 | 1.00 | 0.97 | (0.71 – 1.33) | 1.14 | (0.82 – 1.60) | | 1.00 | 0.77 | (0.38 – 1.57) | 1.01 | (0.52 – 1.96) |
| | Obesity of stages 2&3 | 1.00 | 0.86 | (0.28 – 2.60) | 1.73 | (0.63 – 4.77) | | 1.00 | 4.36 | (1.14 – 16.71) | 1.21 | (0.25 – 5.81) |
| Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension | Diagnosis of hypertension |
| | No | 1.00 | 1.07 | (0.82 – 1.41) | 1.42 | (1.07 – 1.87) | | 1.00 | 1.33 | (0.81 – 2.19) | 1.28 | (0.79 – 2.07) |
| | Yes | 1.00 | 1.16 | (0.74 – 1.82) | 1.05 | (0.62 – 1.78) | | 1.00 | 1.35 | (0.61 – 2.99) | 0.90 | (0.34 – 2.37) |
| Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes | Diagnosis of diabetes |
| | No | 1.00 | 1.06 | (0.83 – 1.35) | 1.39 | (1.08 – 1.79) | | 1.00 | 1.29 | (0.83 – 2.01) | 1.22 | (0.76 – 1.95) |
| | Yes | 1.00 | 1.79 | (0.85 – 3.80) | 0.91 | (0.39 – 2.13) | | 1.00 | 2.14 | (0.24 – 19.16) | 1.19 | (0.30 – 4.67) |
## Discussion
*The* general findings were that there is an association between smoking and NAFLD, and the risk of having NAFLD has a dose-dependent negative association with the duration of smoking cessation and a positive association with pack years. Given these results, our study suggests that ex-smokers with an SCS of fewer than 10 years had associations similar to those seen in current smokers, while ex-smokers whose SCS was more than 20 years had no association. Furthermore, we found a strong linear association between the duration of smoking and the number of cigarettes smoked per day. These findings are consistent with the results of a previous study [30] and may provide supporting evidence for an association between smoking history and NAFLD. Smoking cessation reduces the incidence of NAFLD. However, due to the low number of female smokers in Korea, we could not find a relationship between smoking and NAFLD among females. However, although not statistically significant, the OR of former smokers and current smokers was higher than that of nonsmokers. This reflects the recall bias of self-reported data due to the poor perception of female smokers in Korea [31].
Smoking has been identified, as an adjunct to obesity, as a causative factor for NAFLD in animal and clinical studies [25, 32]. This study found no association between smoking behavior and NAFLD in men with stages 1, 2, and 3 obesity; however, in overweight men and normal women, smoking behavior was a significant risk factor associated with NAFLD compared to nonsmoking. This supports the results of a previous study [33] suggesting that while severe obesity directly affects NAFLD in BMI groups, smoking may have an independent relationship in normal or overweight groups. A mechanism that explains the independent role of BMI in the association between smoking and NAFLD is that the antiestrogenic effect of cigarette smoking leads to a change in body fat distribution (34–36). Therefore, normal and overweight smokers who may not be evaluated for NAFLD should receive more attention to prevent NAFLD.
According to the multiple parallel hits hypothesis theory, the pathophysiological mechanisms of NAFLD indicate the causes of insulin resistance, genetic and epigenetic factors, mitochondrial dysfunction, endoplasmic reticulum stress, microbiota, chromatic low-grade injury, and dysfunction of adipose tissue [37, 38]. In insulin-resistant patients, liver fat production can be further induced by activation of transcription factors such as SREBP-1 [38, 39]. Many studies have shown that tobacco increases lipid accumulation in liver cells by regulating the activity of 5′-AMP-activated protein kinase (AMPK) and SREBP-1, two important molecules involved in lipid synthesis (27, 40–42). It is considered a mechanism between smoking and NAFLD, especially based on previous studies that show a decisive role in liver fat accumulation in SREBP-1, when tobacco smoke is exposed to mice and cultured hepatocytes [27].
However, the effects of smoking on NAFLD remain controversial, with inconsistent results [43]. One study reported that active smoking was associated with fibrosis in patients with NAFLD [25], but another study showed a lack of significant relationship between active smoking and NAFLD [44]. Several experimental studies in mice have shown that nicotine, a dangerous substance in cigarettes, promotes the development of NAFLD or accelerates its progression (25–27). A systematic review and meta-analysis of 20 observational studies showed that smoking was significantly associated with NAFLD [43]. Furthermore, second-hand smoking increases the risk of NAFLD around 1.38 times [43]. Based on these mechanisms, experimental studies and cohort studies that consider additional confounders are needed.
This study had several limitations. First, it was a cross-sectional study. It may not establish temporal relations and may have found an inverse causal relationship. Therefore, caution is warranted when interpreting the results. More research is needed to clarify the association between smoking and NAFLD. Second, KNHANES data were collected through self-report surveys. Hence, data on health-related status, socioeconomic variables, and smoking status may not be reliable and accurate. In particular, this can lead to recall bias and is likely to be underestimated in the case of smoking. Third, although the liver fat score for NAFLD was demonstrated for the ROC curves for detecting NAFLD (sensitivity of $86\%$ and specificity of $71\%$), there were still tiny errors of false-positive or false-negative results. In addition, due to the characteristics of the KNHANES called secondary data, the diagnosis of NAFLD was not measured by the instrument investigation, so steatosis could not be confirmed by methods such as CAP, ultrasound, and liver biopsy. Therefore, we calculated and considered the NAFLD liver fat score instead. Fourth, it could not differentiate among the various smoking types, such as conventional cigarette use, electronic cigarette use, or both. Besides, we could not calculate the pack years for e-cigarettes because the KNHANES did not include this information. Finally, we cannot exclude the possibility of unrecognized confounders, although we adjusted for known confounders in the relationship between smoking and NAFLD.
Despite these limitations, our study had several notable strengths. First, the study was based on the KNHANES data, a nationally representative dataset collected by the KDCA. This is useful for health-related research because it is updated annually to reflect the changes in the actual health situation of Koreans. In addition, it is a statistic that can generalize the study results to the general population because the survey is performed by reliable and representative random cluster sampling. Second, we calculated the SCS and pack years for ex-smokers and current smokers. The study showed a significant association between current smoking behavior in men and smoking status considering SCS and pack-years. Therefore, our results suggest that smoking status has the opportunity to be considered as a measure of intervention to reduce the risk of NAFLD when SCS and pack years are taken into account.
## Conclusion
In conclusion, this study found that current smoking was associated with NAFLD in men in the South Korean population. In particular, we suggest an association between NAFLD and ex-smoker and current smoker status with a short smoking cessation period or many pack years. Given these results, smoking has a potential effect on NAFLD, and smoking cessation should be considered in the prevention and management of NAFLD. It is best to stop smoking considering health status and behavior to avoid serious diseases. More prospective studies and clinical trials are required to clarify the relationship between smoking history and NAFLD.
## Data availability statement
Publicly available datasets were analyzed in this study. This data can be found here: https://www.kdca.go.kr/index.es?sid=a2.
## 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
YSJ designed the study, collected data, performed statistical analysis, and drafted the manuscript. YSJ, HJJ, YSP, E-CP, and S-IJ contributed to the discussion. S-IJ is the guarantor of this work, has full access to all the study data, and assumes responsibility for the integrity of the data and the accuracy of the data analysis. All authors reviewed and edited the drafts of the manuscript, approved the final version, 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. Cohen JC, Horton JD, Hobbs HH. **Human fatty liver disease: old questions and new insights**. *Science.* (2011) **332** 1519-23. DOI: 10.1126/science.1204265
2. Armstrong MJ, Houlihan DD, Bentham L, Shaw JC, Cramb R, Olliff S. **Presence and severity of non-alcoholic fatty liver disease in a large prospective primary care cohort**. *J Hepatol.* (2012) **56** 234-40. DOI: 10.1016/j.jhep.2011.03.020
3. Chalasani N, Younossi Z, Lavine JE, Diehl AM, Brunt EM, Cusi K. **The diagnosis and management of non-alcoholic fatty liver disease: practice Guideline by the American Association for the Study of Liver Diseases, American College of Gastroenterology, and the American Gastroenterological Association**. *Hepatology.* (2012) **55** 2005-23. DOI: 10.1002/hep.25762
4. Loomba R, Sanyal AJ. **The global NAFLD epidemic**. *Nat Rev Gastroenterol Hepatol.* (2013) **10** 686-90. DOI: 10.1038/nrgastro.2013.171
5. Lim YS, Kim WR. **The global impact of hepatic fibrosis and end-stage liver disease**. *Clin Liver Dis* (2008) **12** 733-46. DOI: 10.1016/j.cld.2008.07.007
6. Argo CK, Caldwell SH. **Epidemiology and natural history of non-alcoholic steatohepatitis**. *Clin Liver Dis.* (2009) **13** 511-31. DOI: 10.1016/j.cld.2009.07.005
7. Starley BQ, Calcagno CJ, Harrison SA. **Nonalcoholic fatty liver disease and hepatocellular carcinoma: a weighty connection**. *Hepatology.* (2010) **51** 1820-32. DOI: 10.1002/hep.23594
8. Adams LA, Lymp JF, St Sauver J, Sanderson SO, Lindor KD, Feldstein A. **The natural history of nonalcoholic fatty liver disease: a population-based cohort study**. *Gastroenterology.* (2005) **129** 113-21. DOI: 10.1053/j.gastro.2005.04.014
9. Ekstedt M, Franzén LE, Mathiesen UL, Thorelius L, Holmqvist M, Bodemar G. **Long-term follow-up of patients with NAFLD and elevated liver enzymes**. *Hepatology.* (2006) **44** 865-73. DOI: 10.1002/hep.21327
10. Rafiq N, Bai C, Fang Y, Srishord M, McCullough A, Gramlich T. **Long-term follow-up of patients with nonalcoholic fatty liver**. *Clin Gastroenterol Hepatol.* (2009) **7** 234-8. DOI: 10.1016/j.cgh.2008.11.005
11. Stepanova M, Rafiq N, Younossi ZM. **Components of metabolic syndrome are independent predictors of mortality in patients with chronic liver disease: a population-based study**. *Gut.* (2010) **59** 1410-5. DOI: 10.1136/gut.2010.213553
12. Lee YH, Kim SU, Song K, Park JY, Kim DY, Ahn SH. **Sarcopenia is associated with significant liver fibrosis independently of obesity and insulin resistance in nonalcoholic fatty liver disease: Nationwide surveys (KNHANES 2008-2011)**. *Hepatology.* (2016) **63** 776-86. DOI: 10.1002/hep.28376
13. Younossi Z, Anstee QM, Marietti M, Hardy T, Henry L, Eslam M. **Global burden of NAFLD and NASH: trends, predictions, risk factors and prevention**. *Nat Rev Gastroenterol Hepatol.* (2018) **15** 11-20. DOI: 10.1038/nrgastro.2017.109
14. 14.US Department of Health and Human Services. The Health Consequences of Smoking −50 Years of Progress: A Report of the Surgeon General. Reports of the Surgeon General (2014).. *The Health Consequences of Smoking −50 Years of Progress: A Report of the Surgeon General* (2014)
15. 15.Centers for Disease Control and Prevention (US) National Center for Chronic Disease Prevention and Health Promotion (US) and and Office on Smoking and Health. How Tobacco Smoke Causes Disease: The Biology and Behavioral Basis for Smoking-Attributable Disease: A Report of the Surgeon General. Publications and Reports of the Surgeon General (2010).. *How Tobacco Smoke Causes Disease: The Biology and Behavioral Basis for Smoking-Attributable Disease: A Report of the Surgeon General* (2010)
16. Jatoi NA, Jerrard-Dunne P, Feely J, Mahmud A. **Impact of smoking and smoking cessation on arterial stiffness and aortic wave reflection in hypertension**. *Hypertension.* (2007) **49** 981-5. DOI: 10.1161/hypertensionaha.107.087338
17. Botteri E, Iodice S, Bagnardi V, Raimondi S, Lowenfels AB, Maisonneuve P. **Smoking and colorectal cancer: a meta-analysis**. *JAMA.* (2008) **300** 2765-78. DOI: 10.1001/jama.2008.839
18. Willi C, Bodenmann P, Ghali WA, Faris PD, Cornuz J. **Active smoking and the risk of type 2 diabetes: a systematic review and meta-analysis**. *JAMA.* (2007) **298** 2654-64. DOI: 10.1001/jama.298.22.2654
19. **WHO Report on the Global Tobacco Epidemic, 2021: Addressing new and emerging products**. (2021)
20. Kasper P, Martin A, Lang S, Kütting F, Goeser T, Demir M. **NAFLD and cardiovascular diseases: a clinical review**. *Clin Res Cardiol.* (2021) **110** 921-37. DOI: 10.1007/s00392-020-01709-7
21. Tanase DM, Gosav EM, Costea CF, Ciocoiu M, Lacatusu CM, Maranduca MA. **The Intricate Relationship between Type 2 Diabetes Mellitus (T2DM), Insulin Resistance (IR), and Nonalcoholic Fatty Liver Disease (NAFLD)**. *J Diabetes Res.* (2020) **2020** 3920196. DOI: 10.1155/2020/3920196
22. Marengo A, Rosso C, Bugianesi E. **Liver Cancer: Connections with obesity, fatty liver, and cirrhosis**. *Annu Rev Med.* (2016) **67** 103-17. DOI: 10.1146/annurev-med-090514-013832
23. Zein CO, Beatty K, Post AB, Logan L, Debanne S, McCullough AJ. **Smoking and increased severity of hepatic fibrosis in primary biliary cirrhosis: a cross validated retrospective assessment**. *Hepatology.* (2006) **44** 1564-71. DOI: 10.1002/hep.21423
24. Pessione F, Ramond MJ, Njapoum C, Duchatelle V, Degott C, Erlinger S. **Cigarette smoking and hepatic lesions in patients with chronic hepatitis C**. *Hepatology.* (2001) **34** 121-5. DOI: 10.1053/jhep.2001.25385
25. Azzalini L, Ferrer E, Ramalho LN, Moreno M, Domínguez M, Colmenero J. **Cigarette smoking exacerbates nonalcoholic fatty liver disease in obese rats**. *Hepatology.* (2010) **51** 1567-76. DOI: 10.1002/hep.23516
26. Hasan KM, Friedman TC, Shao X, Parveen M, Sims C, Lee DL. **E-cigarettes and western diet: important metabolic risk factors for hepatic diseases**. *Hepatology.* (2019) **69** 2442-54. DOI: 10.1002/hep.30512
27. Yuan H, Shyy JY, Martins-Green M. **Second-hand smoke stimulates lipid accumulation in the liver by modulating AMPK and SREBP-1**. *J Hepatol.* (2009) **51** 535-47. DOI: 10.1016/j.jhep.2009.03.026
28. Kotronen A, Peltonen M, Hakkarainen A, Sevastianova K, Bergholm R, Johansson LM. **Prediction of non-alcoholic fatty liver disease and liver fat using metabolic and genetic factors**. *Gastroenterology.* (2009) **137** 865-72. DOI: 10.1053/j.gastro.2009.06.005
29. Yoon YJ, Lee MS, Jang KW, Ahn JB, Hurh K, Park EC. **Association between smoking cessation and obstructive spirometry pattern among Korean adults aged 40-79 years**. *Sci Rep.* (2021) **11** 18667. DOI: 10.1038/s41598-021-98156-9
30. Takenaka H, Fujita T, Masuda A, Yano Y, Watanabe A, Kodama Y. **Non-alcoholic fatty liver disease is strongly associated with smoking status and is improved by smoking cessation in japanese males: a retrospective study**. *Kobe J Med Sci.* (2020) **66** E102-e12. PMID: 33431783
31. Kang HG, Kwon KH, Lee IW, Jung B, Park EC, Jang SI. **Biochemically-verified smoking rate trends and factors associated with inaccurate self-reporting of smoking habits in Korean women**. *Asian Pac J Cancer Prev.* (2013) **14** 6807-12. DOI: 10.7314/apjcp.2013.14.11.6807
32. Mallat A, Lotersztajn S. **Cigarette smoke exposure: a novel cofactor of NAFLD progression?**. *J Hepatol.* (2009) **51** 430-2. DOI: 10.1016/j.jhep.2009.05.021
33. Liu Y, Dai M, Bi Y, Xu M, Xu Y, Li M. **Active smoking, passive smoking, and risk of nonalcoholic fatty liver disease (NAFLD): a population-based study in China**. *J Epidemiol.* (2013) **23** 115-21. DOI: 10.2188/jea.je20120067
34. Shimokata H, Muller DC, Andres R. **Studies in the distribution of body fat. Iii Effects of cigarette smoking**. *JAMA.* (1989) **261** 1169-73. PMID: 2915440
35. Tankó LB, Christiansen C. **An update on the antiestrogenic effect of smoking: a literature review with implications for researchers and practitioners**. *Menopause.* (2004) **11** 104-9. DOI: 10.1097/01.Gme.0000079740.18541.Db
36. Windham GC, Mitchell P, Anderson M, Lasley BL. **Cigarette smoking and effects on hormone function in premenopausal women**. *Environ Health Perspect.* (2005) **113** 1285-90. DOI: 10.1289/ehp.7899
37. Acierno C, Caturano A, Pafundi P, Nevola R, Adinolfi L, Sasso F. **Nonalcoholic fatty liver disease and type 2 diabetes: pathophysiological mechanisms shared between the two faces of the same coin**. *Explor Med.* (2020) **1** 287-306
38. Caturano A, Acierno C, Nevola R, Pafundi PC, Galiero R, Rinaldi L. **Non-alcoholic fatty liver disease: From pathogenesis to clinical impact**. *Processes.* (2021) **9** 135. DOI: 10.3390/pr9010135
39. George J, Liddle C. **Nonalcoholic fatty liver disease: pathogenesis and potential for nuclear receptors as therapeutic targets**. *Mol Pharm.* (2008) **5** 49-59. DOI: 10.1021/mp700110z
40. Jung EJ, Kwon SW, Jung BH, Oh SH, Lee BH. **Role of the AMPK/SREBP-1 pathway in the development of orotic acid-induced fatty liver**. *J Lipid Res.* (2011) **52** 1617-25. DOI: 10.1194/jlr.M015263
41. Kim SP, Nam SH, Friedman M. **Mechanism of the antiadipogenic-antiobesity effects of a rice hull smoke extract in 3T3-L1 preadipocyte cells and in mice on a high-fat diet**. *Food Funct.* (2015) **6** 2939-48. DOI: 10.1039/c5fo00469a
42. Ponciano-Rodríguez G, Méndez-Sánchez N. **Cigarette smoking and fatty liver**. *Ann Hepatol.* (2010) **9** 215-8. PMID: 20526022
43. Akhavan Rezayat A, Dadgar Moghadam M, Ghasemi Nour M, Shirazinia M, Ghodsi H, Rouhbakhsh Zahmatkesh MR. **Association between smoking and non-alcoholic fatty liver disease: a systematic review and meta-analysis**. *SAGE Open Med.* (2018) **6** 2050312117745223. DOI: 10.1177/2050312117745223
44. Chavez-Tapia NC, Lizardi-Cervera J, Perez-Bautista O, Ramos-Ostos MH, Uribe M. **Smoking is not associated with nonalcoholic fatty liver disease**. *World J Gastroenterol.* (2006) **12** 5196-200. DOI: 10.3748/wjg.v12.i32.5196
|
---
title: 'Infections and nutrient deficiencies during infancy predict impaired growth
at 5 years: Findings from the MAL-ED study in Pakistan'
authors:
- Doris González-Fernández
- Simon Cousens
- Arjumand Rizvi
- Imran Chauhadry
- Sajid Bashir Soofi
- Zulfiqar Ahmed Bhutta
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9982131
doi: 10.3389/fnut.2023.1104654
license: CC BY 4.0
---
# Infections and nutrient deficiencies during infancy predict impaired growth at 5 years: Findings from the MAL-ED study in Pakistan
## Abstract
### Background
Socio-economic, nutritional, and infectious factors have been associated with impaired infant growth, but how the presence of these factors during infancy affects growth around 5 years is not well understood.
### Methods
This secondary analysis of the MAL-ED cohort included 277 children from Pakistan for whom socio-demographic, breastfeeding, complementary foods, illness, nutritional biomarkers, stool pathogens and environmental enteropathy indicators between 0 and 11 months were recorded. We used linear regression models to analyze associations of these indicators with height-for-age (HAZ), weight-for-age (WAZ) and weight-for-height (WLZ) at 54–66 months (~5 years), and Poisson regression with robust standard errors to estimate risk ratios for stunting and underweight ~5 years, controlling for gender, first available weight, and income.
### Results
Among the 237 infants followed longitudinally and evaluated at about 5 years of age, exclusive breastfeeding was short (median = 14 days). Complementary feeding started before 6 months with rice, bread, noodles, or sugary foods. Roots, dairy products, fruits/vegetables, and animal-source foods were provided later than recommended (9–12 months). Anemia ($70.9\%$), deficiencies in iron ($22.0\%$), zinc ($80.0\%$), vitamin A ($53.4\%$) and iodine ($13.3\%$) were common. Most infants (>$90\%$) presented with diarrhea and respiratory infections in their first year. At ~5 years, low WAZ (mean-1.91 ± 0.06) and LAZ (−2.11 ± 0.06) resulted in high prevalence of stunting ($55.5\%$) and underweight ($44.4\%$) but a relatively low rate of wasting ($5.5\%$). While $3.4\%$ had concurrent stunting and wasting ~5 years, $37.8\%$ of children had coexisting stunting and underweight. A higher income and receiving formula or dairy products during infancy were associated with a higher LAZ ~5 years, but infant’s history of hospitalizations and more respiratory infections were associated with lower LAZ and higher risk of stunting ~5 years. Infants’ intake of commercial baby foods and higher serum-transferrin receptors were associated with higher WAZ and lower risk of underweight ~5 years. Presence of *Campylobacter and* fecal neopterin >6.8 nmol/L in the first year were associated with increased risk of underweight ~5 years.
### Conclusion
Growth indicators ~5 years were associated with poverty, inappropriate complementary feeding, and infections during the first year of life, which supports the early start of public health interventions for preventing growth delay ~5 years.
## Introduction
Despite some reduction, undernutrition continues to affect millions of children. According to the 2021 Global Nutrition Report, 149.2 million children under 5 suffer from stunting, 45.4 million are wasted [1]. Also, and 1 in 2 children suffer from hidden hunger due to deficiencies in essential vitamins and nutrients [2]. Undernutrition in the first 2 years of life has been linked with shorter adult height, lower educational achievement, reduced economic productivity and with smaller infants in the next generation [3]. Linear growth is considered the best indicator of child well-being with stunting, defined as length/height < −2 SDs below the WHO child growth standard [4] indicative of past deprivation and predictive of future poverty [2]. On the other hand, underweight (low weight-for-age), is known to increase the risk of viral, helminth and malaria infections in children [5]. Wasting refers to children <-2SDs below the WHO standards weight-for-length/height median [6], reflecting a recent loss of weight from severe poor nutrient intake, illness or both [2]. Children presenting both stunting and wasting have the highest risk of mortality, even higher than those with WLZ < -3 SD [7]. Although recent reports show a decline in the global prevalence of stunting (from 32.5 to $21.9\%$) and wasting (from 10 to $7.3\%$) between 2000 and 2017, important disparities continue to be observed, south Asia presenting higher wasting at birth ($19\%$) compared with African ($8\%$) and Latin-American infants ($2\%$) [7].
The origin of undernutrition is complex and multifactorial. The Etiology, Risk Factors and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) study, followed children from eight LMIC (Bangladesh, India, Nepal, Pakistan, Brazil, Peru, South Africa and Tanzania), and showed, for example, that low energy and protein density of complementary foods and a high prevalence of enteropathogens in non-diarrheal stools were associated with reduced weight and length by age 24 months [8]. MAL-ED studies have contributed to our understanding of the impact of environmental enteropathy (EE) on child growth. EE has been described as phenotypic intestinal alterations that affect the health status of the host, following repeated enteric infections even in the absence of diarrhea or acute gastrointestinal illnesses [9]. The MAL-ED study used three fecal biomarkers of gut inflammation and immunity. Myeloperoxidase (ng/mL) is a marker of neutrophil activity in the intestinal mucosa, neopterin indicates T-helper cell 1 activity and alpha-1 antitrypsin is an indicator of protein loss and intestinal permeability [9]. Indicators of EE were associated with reduced stature, weight, weight-for-height, and BMI at 5 years, but no associations were found between illness symptoms and size at 5 years [10].
Among countries in the MAL-ED study, Pakistan had a high prevalence of wasting ($15\%$) and stunting ($44\%$) according to the 2011 National Nutrition Survey [11]. Pakistan also had the lowest mean WAZ (−1.4) and the highest prevalence of anemia in infants ($88\%$) [12] compared with other cohorts in the study. Comparisons of home environment among MAL-ED sites found that children from Pakistan had the highest food insecurity scores [13], the shortest duration of exclusive breastfeeding [14] and the highest frequency of reported coughing ($27\%$) [15]. The prevalence of acute lower respiratory infections (ALRI) and ear pain were 6 and 7 times higher, respectively, compared with the next highest site (India) [15]. In the MAL-ED cohort of Pakistan, persistent infection with Giardia was found to be associated with lower weight-for-age and length-for-age Z scores at 2 years [16]. Although associations of other nutritional and infectious indicators, have been reported for other MAL-ED sites (8–10, 17, 18) similar analysis had not previously been performed for Pakistan after having detected field bias in length/height data collection.
Undernutrition is a major driver of health and economic consequences in Pakistan, and given stunting rates of $40\%$, wasting stagnating at around 15–$17\%$, this is estimated to result in an approximate $3\%$ loss of annual gross domestic product [19]. Given that the highest incidence of stunting and wasting occurs in infancy [7], our objectives were to explore socio-demographic, nutrition, and infection factors in the first year of life and their association with anthropometry at 5 years in children from Pakistan.
## Methods
This is a secondary analysis of data from the MAL-ED cohort of children from Pakistan, which recruited 277 healthy singleton newborns (≤17 days) from the Naushahro Feroze district in the Sindh province between January 2010 and February 2012 [11]. Children were followed up to 66 months of age as shown in Figure 1, with a final available sample of 237 children at 54–66 months ($86\%$ of the original cohort). Length and weight were measured by trained personnel following standard procedures as previously described [8]. A rigorous revisiting of length/height individual trajectories was performed, eliminating observations with possible bias during field data collection.
**Figure 1:** *Sample size flow chart of children enrolled and followed between 0 and 5 years of age in the MAL-ED cohort in Pakistan.*
Our conceptual framework (Figure 2) identified possible early factors at 0–11 months of age that have known associations with child growth in Pakistan [20] or the MAL-ED cohort in other countries (8–10): (a) biological drivers (low birth weight), (b) adverse socio-demographic factors (c) inappropriate breastfeeding and weaning practices, (d) biomarkers of micronutrient deficiencies (iron indicators, vitamin A, zinc and iodine concentrations), (e) history of illness, and (f) biomarkers of infection and environmental enteropathy. These factors were grouped in clusters for further analyses.
**Figure 2:** *Conceptual framework: Known predictors of growth impairment include biological and environmental factors, nutrition-related factors, and infection-related factors. We hypothesize that adverse factors during the first year of life will be associated with indicators of growth at 54–66 months of age.*
## Statistical analyses
Outcome variables: Z-scores for weight-for-age (WAZ), height for age (HAZ), and weight for height (WHZ) at 5 years were calculated applying WHO reference data using STATA 16 [27] and analyzed as continuous variables, as well as their derivative binary variables underweight, stunting and coexistent underweight + stunting. The small number of wasted children at 5 years ($5.5\%$) did not allow us to run models for this variable.
Spearman correlations among independent variables were explored for descriptive and modeling purposes. Descriptive statistics of variables for children having information on weight or height at 54–66 months were run ($$n = 239$$). Fisher’s exact, Chi2 or Kruskal-Wallis’s test were performed according to variable’s nature to identify if they differed between children with ($$n = 239$$) and without anthropometry data ($$n = 38$$).
## Variable pre-selection
Based on our conceptual framework, variables were aligned in groups of possible predictors. In order to select covariates that would best fit future modeling and discard noise variables [28], we used a bootstrapping procedure based on a backward stepwise algorithm with 1,000 repetitions [29], where groups of variables were tested avoiding the simultaneous inclusion of highly correlated variables, which were run separately.
## Univariate models
Only variables entering ≥500 repetitions were further analyzed in: (a) univariate models for binary (stunting, underweight, stunting + underweight) and (b) continuous (LAZ, WAZ, and WLZ) outcomes. We used generalized linear models (GLM) Poison and linear regression models, respectively. Dichotomous covariates with <10 events per variable were excluded [30]. The univariate modeling process is shown in detail as Supplementary Material.
## Multivariable models
Variables with value of p is less than 0.05 in univariate models were further analyzed in multivariable regression models. We controlled all models for gender, first available weight and income (only socio-demographic variable entering ≥500 repetitions). GLM Poisson regression with robust standard errors were used to estimate risk ratios for binary outcomes, and multiple linear regression was used for continuous outcome variables. Backwards stepwise elimination (initially specifying value of $p \leq 0.10$, and decreasing value of p to <0.05 depending on the number of variables entering) were used to obtain final models including no more than 10 variables to avoid overfitting, using variables that are likely to be amenable to and influenced by interventions. Only final models are shown. Final models were tested for collinearity using a variance inflation factor < 10 and a condition number < 30 [31]. Tests for covariate-dependent missingness (CDM) were conducted, allowing unequal variances between missing-value patterns [32].
## Results
Characteristics of the population are described in Tables 1, 2; Figures 3, 4.
## Socio-demographic characteristics and food security
Characteristics of children’s families recruited in the study are shown in Table 1. Children came from families with low incomes, where $23.4\%$ of them earned <8,000 rupees/month (equivalent to 100 USD), the minimal wage at the time of the study. Also, $81.5\%$ lived with some degree of food insecurity. Overcrowding was common, with >3 people per room observed in $75\%$ of homes. Although all households had water from tube well or borehole, $15\%$ did not have sanitation. Among mothers, $50\%$ had not received any education. Gender distribution of our sample showed similar proportions of boys and girls ($50.4\%$ girls).
## Prevalence of stunting, underweight and wasting
First available weights (mean ± SD: 2.8 ± 0.5 kg between 0 and 17 days) found that $20.9\%$ of children were below 2.5 kg. Among children with length available during the first month of age ($$n = 111$$), 39 ($29.5\%$) were already stunted. Of note, most of these children continued to be stunted at 54–66 months ($$n = 25$$, $64.1\%$).
Figure 3 shows trajectories of WAZ, LAZ and WLZ of children from 0–5 years. Mean yearly lengths-for-age indicate that prevalence of stunting increased with age from $39.6\%$ in the first year to $55.5\%$ at 54–66 months. Similarly, underweight increased from $30\%$ in the first year to $44.4\%$ at 54–66 months. The presence of wasting ($5.5\%$ at 54–66 months) did not reflect the high proportions of underweight and stunting, as parallel and symmetric low weights and heights were observed in this population, but $37.8\%$ had coexisting stunting and underweight. Overlapping proportions of stunting, underweight and wasting at 54–66 months are shown in Figure 4.
## Breastfeeding and complementary feeding practices
Feeding practices are summarized in Table 2. Although all children received breastfeeding for some time between 0 and 5 months with a median frequency of 16 times/d, the median duration of exclusive breastfeeding was only 14 days. During their first month of age, most children ($80.8\%$) received non-milk fluids (sugar water, thin soup or broth, fruit juice, tea, other fluids), animal milk/formula ($29.2\%$) or both ($25.5\%$). Between 0 and 5 months, most children received animal milk ($74.5\%$, median frequency 2 times/d) or formula ($13.0\%$, median frequency 3 times/d) starting at a median age of 1 month (Table 2A).
Complementary feeding with solid/semisolid foods started at a median age of 3.5 months, and was composed mainly of grains (rice, porridge, bread, noodles) and sugary foods (pastries, cakes, biscuits). After 6 months, children started receiving other dairy products (e.g., cheese, yogurt) and roots, whereas other nutritious foods such as legumes, fruits, and animal-source foods (animal flesh, eggs) were mostly introduced after 1 year of age (Table 2B).
## Nutritional biomarkers
In the subsample of children ($$n = 186$$–220) with blood samples taken between 6 and 8 months of age, $89.0\%$ had zinc deficiency, $71.3\%$ had anemia, $53.7\%$ had vitamin A deficiency, $22.1\%$ had low ferritin but only $1.2\%$ had elevated sTfR. Moreover, urine iodine was low in $13.2\%$ of children. Finally, inflammation indicated by AGP >100 mg/dl was found in $38.7\%$ (Table 2C). Of note, among children with anemia, $88.4\%$ had also zinc deficiency, $57.4\%$ had low vitamin A, $43.1\%$ had elevated AGP, and $25.8\%$ were iron deficient.
## Illness information
On average, children spend 3 weeks/month presenting some type of illness, notably diarrhea, vomiting, respiratory infection, or ear pain/pulling, and nearly all of them received at least one cycle of antibiotics in the first month of life (Table 2D).
Microbiological stool sample analyses showed that the predominant pathogens present during diarrheal episodes were Enteroaggregative and Enteropathogenic E. coli, *Campylobacter and* Norovirus, which were isolated in more than half of children with diarrhea between 0 and 11 months (Table 2F).
## Fecal biomarkers
Concentrations of fecal biomarkers myeloperoxidase, neopterin and alpha-1-antitrypsin are shown in Table 2E. When assessing correlations between mean fecal biomarkers during the first year with indicators of infection/inflammation, neither mean myeloperoxidase nor neopterin concentrations correlated with serum AGP concentrations, but mean alpha-1-antitrypsin had a weak negative correlation with AGP (rs = −0.16, $$p \leq 0.015$$). Fecal biomarkers did not correlate with the number of diarrheal episodes and did not differ by the presence/absence of intestinal pathogens in the first year. Higher mean concentrations of myeloperoxidase (rs = −0.13, p = −0.029), neopterin (rs = 0.17, $$p \leq 0.004$$) and alpha-1-antitrypsin (rs = 0.32, $p \leq 0.0001$) were correlated with days/month presenting vomiting, and higher neopterin (rs = 0.12, p-0.04) and alpha-1 antitrypsin (rs = 0.19, $$p \leq 0.001$$) were correlated with more days/month presenting any illness during the first year of life. The cut-point of elevated neopterin for the detection of underweight using the Youden index was calculated at >6.8 nmol/L, and for the detection of underweight + stunting was >2.4 nmol/L.
## Early predictors of impaired growth at 54–66 months of age
Univariate models for stunting, underweight and stunting + underweight are presented in Supplementary Tables S1–S3, respectively. Results of multivariate GLM-Poisson regression models are as follows:
## Stunting
Among variables selected using univariate models, only the average of ≥2 days per month presenting ALRI during the first year of life was associated with increased risk of stunting at 54–66 months (Table 3A). Variables with weak evidence of an association with increased risk of stunting at 54–66 months (entered the model with value of $p \leq 0.05$) were being a girl, lower weight during the first years of life, lower income, and history of hospitalization in the first year.
**Table 3**
| (A) Model for stunting at 54–66 months | RR | 95% CI | Value of p |
| --- | --- | --- | --- |
| Girls (n = 120) | 1.26 | 0.99, 1.60 | 0.054 |
| First available weight, kg | 0.82 | 0.64, 1.05 | 0.118 |
| Income (rupees × 103, quantiles) base: 20.1–70 (n = 52) | | | |
| <8 (n = 56) | 1.38 | 0.97, 1.96 | 0.074 |
| 8–12.5 (n = 67) | 1.31 | 0.92, 1.85 | 0.130 |
| 12.6–20 (n = 63) | 1.09 | 0.75, 1.58 | 0.645 |
| Hospitalized at least once between 0 and 11 moths (n = 13) | 1.32 | 0.98, 1.78 | 0.070 |
| Mean days/month presenting ALRI between 6 and 11 months base: 0 (n = 119) | | | |
| 1 (n = 94) | 1.05 | 0.82, 1.36 | 0.679 |
| 2 (n = 16) | 1.55 | 1.09, 2.20 | 0.015 |
| 3–4 (n = 9) | 1.71 | 1.22, 2.41 | 0.002 |
| Constant | 0.50 | 0.20, 1.24 | 0.137 |
| (B) Model for underweight at 54–66 months | RR | 95% CI | value of p |
| Girls (n = 111) | 1.05 | 0.81, 1.37 | 0.685 |
| First available weight, kg | 0.33 | 0.23, 0.46 | <0.0001 |
| Income (rupees × 103, quantiles) base: 20.1–70 (n = 44) | | | |
| <8 (n = 51) | 1.20 | 0.82, 1.77 | 0.347 |
| 8–12.5 (n = 62) | 1.22 | 0.83, 1.80 | 0.314 |
| 12.6–20 (n = 59) | 0.97 | 0.64, 1.47 | 0.897 |
| Campylobacter (+) at least once between 0 and 5 months (n = 174) | 1.78 | 1.17, 2.70 | 0.007 |
| Received formula between 0 and 5 months (n = 25) | 0.53 | 0.30, 0.96 | 0.036 |
| Received commercial foods between 6 and 11 months (n = 23) | 0.31 | 0.15, 0.62 | 0.001 |
| sTfR (mg/L), quantiles base: 0.2–1.8 mg/L (n = 58) | | | |
| 1.81–3.0 (n = 54) | 0.96 | 0.67, 1.37 | 0.811 |
| 3.1–4.5 (n = 52) | 1.01 | 0.73, 1.38 | 0.965 |
| 4.6–10.0 (n = 52) | 0.55 | 0.35, 0.85 | 0.008 |
| Neopterin 0–11 months >6.8 nmol/L (n = 21) | 1.76 | 1.32, 2.35 | <0.0001 |
| Constant | 5.99 | 1.98, 18.18 | 0.002 |
| (C) Model for underweight + stunting at 54-66 months | RR | 95% CI | value of p |
| Girls (n = 97) | 1.24 | 0.92, 1.67 | 0.151 |
| First available weight, kg | 0.53 | 0.38, 0.73 | <0.0001 |
| Income (rupees x103, quantiles) base: 20.1–70 (n = 49) | | | |
| <8 (n = 55) | 2.08 | 1.25, 3.47 | 0.005 |
| 8–12.5 (n = 53) | 1.77 | 1.05, 2.98 | 0.032 |
| 12.6–20 (n = 40) | 1.34 | 0.77, 2.37 | 0.299 |
| Hospitalized between 0 and 11 months (n = 11) | 1.40 | 0.94, 2.10 | 0.101 |
| Mean days/month presenting ALRI between 6 and 11 months base: 0 (n = 98) | | | |
| 1 (n = 80) | 1.07 | 0.77, 1.47 | 0.695 |
| 2 (n = 13) | 1.41 | 0.85, 2.32 | 0.179 |
| 3–4 (n = 6) | 1.63 | 1.08, 2.44 | 0.019 |
| Neopterin 0–11 months >6.8 nmol/L (n = 23) | 1.56 | 1.13, 2.26 | 0.009 |
| Constant | 0.99 | 0.32, 3.09 | 0.986 |
## Underweight
The multivariable regression model for underweight at 54–66 months indicated that a higher weight in the first days of life, receiving formula in the first 6 months and receiving commercial baby foods between 6 and 11 months were associated with decreased risk of underweight at 54–66 months. Higher concentrations of sTfR (but below pathological concentrations, between 4.6 and 10 mg/l) were also associated with decreased risk of underweight at 54–66 months. Having *Campylobacter infection* in the first 6 months and having mean concentrations of neopterin >6.8 nmol/L in the first year of life were associated with increased risk of underweight at 54–66 months (Table 3B). Neither gender nor income appeared to be associated with underweight at 54–6 months.
## Underweight + stunting
When looking at early factors that might drive the presence of both stunting and underweight at 54–66 months compared with non-stunted/non-underweight children, having lower weight the first days of life, family income in the lower quantiles, more days/month presenting ALRI and neopterin >6.8 nmol/L in the first year (Youden’s cut-point for underweight) were associated with higher risk of underweight + stunting at 54–66 months (Table 3C). The Youden’s cut-point of neopterin >2.4 nmol/L for the detection of underweight + stunting at 54–66 months was not associated with this outcome when running the univariate or multivariate model. Gender was not associated with the presence of underweight + stunting at 54–66 months, and “hospitalized between 0 and 11 months” showed weak evidence of an association to increased risk of underweight + stunting at 54–66 months with a value of $$p \leq 0.105.$$
## Early predictors of LAZ, WAZ, and WLZ at 54–66 months
Univariate models for LAZ, WAZ and WLZ are presented in Supplementary Tables S4–S6, respectively. Results of multivariate linear regression models are as follows (Table 4):
**Table 4**
| (A) Model for LAZ at 54-66 months | Coeff ± SE | 95% CI | Value of p |
| --- | --- | --- | --- |
| Girls | −0.22 ± 0.12 | −0.46, 0.02 | 0.076 |
| First available weight, kg | 0.53 ± 0.12 | 0.29, 0.78 | <0.0001 |
| Income (rupees x103, quantiles) Base: 20.1–70 (n = 49) | | | |
| <8 (n = 55) | −0.43 ± 0.18 | −0.79, −0.08 | 0.016 |
| 8–12.5 (n = 53) | −0.37 ± 0.17 | −0.71, −0.03 | 0.031 |
| 12.6–20 (n = 40) | −0.16 ± 0.17 | −0.50, 0.18 | 0.361 |
| Received formula between 0 and 5 months (n = 31) | 0.63 ± 0.18 | 0.27, 0.99 | 0.001 |
| Received dairy products between 0 and 5 months (n = 19) | 0.49 ± 0.21 | 0.08, 0.91 | 0.020 |
| Hospitalized at least once between 0 and 11 months (n = 13) | −0.83 ± 0.28 | −1.38, −0.28 | 0.003 |
| Number of ALRI episodes between 0 and 11 months base: 0–1 (n = 47) | | | |
| 2–4 (n = 124) | −0.37 ± 0.15 | −0.67, −0.07 | 0.017 |
| 5–10 (n = 67) | −0.50 ± 0.18 | −0.85, −0.15 | 0.005 |
| Constant | −2.80 ± 0.47 | −3.72, −1.88 | <0.0001 |
| (B) Model for WAZ at 54-66 months | Coeff. ± SE | 95% CI | Value of p |
| Girls (n = 111) | −0.05 ± 0.11 | −0.28, 0.16 | 0.610 |
| First available weight, kg | 0.69 ± 0.11 | 0.47, 0.92 | <0.0001 |
| Income (rupees × 103, quantiles) base: 20.1–70 (n = 44) | | | |
| <8 (n = 51) | −0.12 ± 0.17 | −0.45, 0.21 | 0.487 |
| 8–12.5 (n = 62) | −0.29 ± 0.16 | −0.61, 0.03 | 0.078 |
| 12.6–20 (n = 59) | −0.05 ± 0.16 | −0.37, 0.27 | 0.749 |
| Receiving baby commercial foods between 6 and 11 months (n = 23) | 0.69 ± 0.11 | 0.41, 1.12 | <0.0001 |
| Mean days/month presenting low appetite between 0 and 5 months base: 0 (n = 157) | | | |
| 1 (n = 45) | −0.12 ± 0.15 | −0.42, 0.17 | 0.408 |
| 2–8 (n = 14) | −0.40 ± 0.25 | −0.88, 0.09 | 0.107 |
| sTfR (mg/L), quantiles base: 0.2–1.8 mg/L (n = 58) | | | |
| 1.81–3.0 (n = 54) | 0.22 ± 0.16 | −0.10, 0.55 | 0.171 |
| 3.1–4.5 (n = 52) | 0.10 ± 0.17 | −0.24, 0.45 | 0.545 |
| 4.6–10.0 (n = 52) | 0.59 ± 0.18 | 0.24, 0.94 | 0.001 |
| Constant | −3.97 ± 0.43 | −4.81, −3.13 | <0.0001 |
| (C) Model for WLZ at 54–66 months | Coeff. ± SE | 95% CI | Value of p |
| Girls (n = 111) | 0.12 ± 0.11 | −0.09, 0.33 | 0.272 |
| First available weight, kg | 0.46 ± 0.11 | 0.23, 0.68 | <0.0001 |
| Income (rupees × 103, quantiles) base: 20.1–70 (n = 44) | | | |
| <8 (n = 50) | 0.20 ± 0.17 | −0.13, 0.53 | 0.235 |
| 8–12.5 (n = 61) | 0.07 ± 0.16 | −0.25, 0.38 | 0.669 |
| 12.6–20 (n = 59) | 0.12 ± 0.16 | −0.20, 0.44 | 0.460 |
| Age starting milk base: 6–20 months (n = 24) | | | |
| <1 month (n = 105) | −0.28 ± 0.18 | −0.64, 0.07 | 0.116 |
| 1–2 months (n = 53) | −0.44 ± 0.19 | −0.83, −0.06 | 0.024 |
| 3–5 months (n = 28) | −0.01 ± 0.22 | −0.42, 0.44 | 0.950 |
| Received other fluids between 6 and 11 months (n = 153) | 0.25 ± 0.12 | 0.004, 0.49 | 0.046 |
| Received baby commercial foods between 6 and 11 months (n = 22) | 0.67 ± 0.18 | 0.32, 1.03 | <0.0001 |
| Norovirus (+) at least once between 0 and 11 months (n = 153) | −0.19 ± 0.13 | −0.44, −0.05 | 0.127 |
| (C) Model for WLZ at 54–66 months | Coeff. ± SE | 95% CI | Value of p |
| sTfR (mg/L), quantiles base: 0.2–1.8 mg/L (n = 57) | | | |
| 1.81–3.0 (n = 54) | 0.31 ± 0.15 | 0.01, 0.61 | 0.043 |
| 3.1–4.5 (n = 51) | 0.04 ± 0.15 | −0.26, 0.34 | 0.786 |
| 4.6–10.0 (n = 52) | 0.58 ± 0.15 | 0.27, 0.88 | <0.0001 |
| Constant | −2.39 ± 0.46 | −3.390-1,47 | <0.0001 |
## LAZ
Increases in child length at 54–66 months were associated with a greater weight in the first days of life, higher family income, receiving formula and dairy products in the first 6 months of life, while lower child length was associated with having been hospitalized and the number of ALRI episodes during the first year of life. There was only weak evidence of an association between gender and LAZ ($$p \leq 0.076$$) (Table 4A).
## WAZ
Higher weight at 54–66 months was associated with greater weight during the first days of life, having received baby commercial foods after 6 months and a higher sTfR (but below abnormal values, between 4.6 and 10 mg/l) (Table 4B). Other variables that entered ≥500 repetitions but only showed weak evidence of an association with WAZ were “gender,” “income,” and “mean days/month having low appetite in the first 5 months.”
## WLZ
A higher weight-for-length was associated with greater weight during the first days of life, and with indicators of food availability during the complementary-food period. A too-early start of non-breast milk (1–2 months of age) was associated with lower WLZ at 54–66 months, but those children receiving baby commercial foods between 6 and 11 months and fluids such as sugar water, thin soup or broth or carbonated drinks between 0 and 11 months had higher WLZ at 54–66 months. sTfR in its highest quartile (4.6–10 mg/L) at 6–8 months was also associated with higher WLZ at 54–66 months. Neither gender nor income nor infection-related variables were associated with WLZ at 54–66 months (Table 4C).
## Discussion
This study explored early predictors of undernutrition at 5 years among socio-demographic, breastfeeding and complementary feeding practices, illness, intestinal pathogens and indicators of environmental enteropathy during the first year of age in children from Pakistan. Chronic undernutrition was highly prevalent, evidenced by high rates of stunting and underweight (but not wasting) at 5 years. The low adherence to exclusive breastfeeding, combined with an early start of weaning consisting of non-milk fluids (sugar water, thin soup or broth, fruit juice, tea), and “empty-caloric” foods added to frequent and persisting infections, characterized the infancy of this cohort. Main findings that distinguished this cohort from other MAL-ED studied countries [10] included that infection during infancy, mainly lower respiratory infections, emerged as a strong predictor of LAZ at 5 years, whereas early infant intestinal colonization by Campylobacter, and one indicator of environmental enteropathy (neopterin >6.8 nmol/L) were associated with underweight at 5 years. Among nutritional indicators, initiation of complementary feeding with nutrient-dense foods was associated with higher WAZ and WLZ. Surprisingly, lower sTfR in the presence of anemia and iron deficiency during infancy was a predictor for impaired weight at 5 years, probably indicating a blunted erythropoietic response and early protein-calorie malnutrition, which may help to explain this unexpected association.
We acknowledge our limitation of a reduced sample size for children with accurate data on length/height between 0 and 35 months, but the rigorous data cleaning makes us confident of the accuracy of reported results. We lacked information on size at birth for many children, a possible parameter of importance for determining size at 5 years, and the lower sample size of children with blood biomarkers during the first year might have limited the power to find possible associations of specific nutrient deficiencies or systemic inflammation with growth at 5 years. Although the number of wasted children at 5 years did not allow to run analyses for this binary variable, the continuous WLZ was explored as continuous variable. Moreover, the richness of information gathered by the MAL-ED investigators allowed to elucidate important targets for future interventions to reduce child impairment in this population.
The prevalence of stunting ($55.5\%$), underweight ($44.4\%$) and wasting ($5.5\%$) found in the MAL-ED cohort of children at 54–66 months from the Sindh province in 2015 was consistent with the prevalence reported by the Pakistan Demographic and Health Survey 2012–2013, where $45.6\%$ stunting, $30.0\%$ underweight and $4.8\%$ wasting in children 48–59 months were reported [33]. Although the use of stunting as an indicator of undernutrition has been debated [34], in particular the cut-off of −2 LAZ as definition of “chronic malnutrition” [35], it is clear that stunting reflects adverse growth conditions starting in utero and continuing during infancy, involving not only nutritional factors [36], but also the lack of appropriateness of children’s environment [37]. Our results in this population with mean LAZ and WAZ <0, indicate that children are not achieving their growth potential [35], where $37.8\%$ had coexistent underweight and stunting. Moreover, deficiencies in micronutrients that are essential for growth and development were extremely high in this population, in particular anemia ($71.3\%$) and zinc ($89.0\%$) deficiencies.
The importance of coexisting wasting and stunting has been emphasized as an indicator of children with higher risk of mortality [38, 39]. Despite the low prevalence of wasting in our study, our results showed that predictors of WLZ were similar to those of WAZ. We also showed a high prevalence of concurrent low WAZ and LAZ. Our results add to the literature information on predictors of concurrent underweight and stunting, where most predictors of the coexistent condition overlap with those of stunting, suggesting a major contribution of illnesses (history of hospitalization and ALRI) and inflammation (elevated neopterin) over nutritional indicators as possible drivers of a more worrisome condition. These findings support the need of addressing infections together with nutritional interventions in the presence of coexisting underweight and stunting.
Among the multiple factors studied during the first year of life, only few remained as predictors of undernutrition at 5 years after a strict bootstrapping-selecting procedure. Other studies have shown lower income as a main determinant of undernutrition [19], but in our study lower LAZ was the only indicator associated with lower income. In contrast with other MAL-ED sites that did not find an association between illness and growth at 5 years [10], we found that history of hospitalization and more days/month presenting ALRI were associated with lower LAZ, and ALRI infections increased the risk of stunting at 5 years, in agreement with current knowledge of undernutrition and inflammation as common pathways for reduced linear growth in children [36].
The WHO recommends that infants initiate breastfeeding within the first hour of birth, and continue exclusively breastfeeding for the first 6 months of life [40]. This is far from the case of the Pakistan MAL-ED cohort, where exclusive breastfeeding was provided with a median duration of only 2 weeks. Analyses showed that providing formula between 0 and 5 months was associated with reduced risk of underweight, and increased WAZ and LAZ. Of note, early start of complementary feeding was a common practice, where mothers provided mainly clear fluids or animal milk, but only $6\%$ of children received formula, and those children came from families with higher incomes. Our findings highlight the importance of reinforcing the message of exclusive breastfeeding during the first 6 months in these communities, which would overcome the lack of resources to afford appropriate complementary foods at that early age. Providing formula has been suggested as a therapeutic approach to correct weight deficits in developing settings [41], but for populations like the MAL-ED cohort, such approach would not be neither financially nor context appropriate. On the other hand, a recent systematic review found that bovine/cow milk supplementation had unfavorable effects on infant morbidity and mortality [42], which aligns with our finding of earlier start of milk compared with starting milk after 6 months of age associated with lower WLZ. Also subject of controversy, commercial infant foods have shown a contribution of $70\%$ of recommended nutrient intakes in children 13–23 months in Ghana [43], which suggest that our finding of better WAZ and WLZ in children who had consumed commercially available infant foods may be related to an improvement in their micronutrient status, but this would require further study.
It was intriguing to find that higher sTfR in the first year, at concentrations below the cut-off for iron deficiency, was associated with higher WAZ and decreased risk of underweight at 5 years. Of note, no other iron status indicators were associated with child growth indicators at 5 years. It is known that transferrin, the acute-phase reactant protein that transports iron, has been used as well as a marker of nutritional status, given that it parallels prealbumin concentrations during nutritional interventions and is decreased in severe malnutrition [44]. Moreover, the total mass of sTfR depends on the number of erythroid precursors in the bone marrow [45], and that those are in turn decreased in protein-energy malnutrition [44]. Our findings suggest that a deprived erythropoiesis during infancy expressed as low concentrations of sTfR despite iron deficiency and anemia are reflecting protein-energy malnutrition. Our findings mimic those of other MAL-ED sites, where higher sTfR were associated with higher WAZ [10].
The association between ALRI and impaired growth has been previously documented. Stunting increased the risk of ALRI in infants from Turkey [46] and in children under 5 in Sri Lanka [47], and acute respiratory infection was the major predictor of underweight in children under 5 from India [48]. To our knowledge, ours is the first study reporting the association of ALRI in the first year of life with impaired growth at 5 years. Whereas much attention has been paid to diarrheal disease as a major cause of mortality in developing countries, ALRI is also an important cause of morbidity and mortality, which are proportional to socio-demographic conditions [49]. Both viruses and bacteria share the etiology of ALRI in children, but the severity of bacterial ALRI largely exceeds that of viral ALRI. However, by the time of the study, Pakistan *National data* shows that $46.2\%$ of children were not fully immunized [33]. Vaccines against bacteria and viruses implicated in morbidity and mortality due to ALRI such as measles, diphtheria, pertussis, *Haemophilus influenzae* b, pneumococcus and influenza have the potential to reduce the burden of the disease [50], and therefore, could help in reducing growth failure in this particular setting.
A consistent association between intestinal *Campylobacter infection* and lower LAZ together with increased intestinal and systemic inflammation was found across other MAL-ED sites [51]. Our findings extend this association from the earliest age of infection at 0–5 months to low WAZ at 5 years. The association of *Campylobacter with* malnutrition indicated by WAZ has been previously reported in children 6–23 months from Bangladesh [52]. Given that host factors such as shifts in diet and microbiota can modify colonization resistance to the infection, *Campylobacter has* been proposed as a biomarker of enteropathy and is a potential target for interventions intending to improve child growth in developing countries [53].
The MAL-ED cohorts from Nepal and Bangladesh found an association of myeloperoxidase between 3 and 6 months with growth velocity from 9 to 24 months [54] and between myeloperoxidase between 3 and 21 months and successive 3-month change in LAZ [55] respectively, but did not find an association of neopterin or alpha-1 antitrypsin with linear growth. In the present study, we only found an association between higher neopterin at concentrations ≥6.8 nmol/L and increased risk of underweight at 5 years, but no association was found between any of the three EE indicators with LAZ or stunting. A higher fecal neopterin in children hospitalized with severe acute malnutrition compared with controls was found in a study from Uganda [56], and given that neopterin in our study was positively correlated with the number of days presenting all-cause illness in the first year, it is possible that a high fecal neopterin might be reflecting systemic inflammation to some extent. At this regard, there is strong evidence that the association between fecal biomarkers and impaired child growth is mediated through systemic inflammation [9]. Therefore, neopterin appeared in our study as the earliest fecal biomarker that predicted undernutrition at 5 years.
## Conclusion
Chronic malnutrition was highly prevalent in children from Pakistan, and our study evidenced associations with early inappropriate complementary feeding practices and illnesses, particularly respiratory infections, with lower growth indicators. Analyses suggested that the pathway for these associations could be early malnutrition indicated by low sTfR in the presence of iron deficiency and anemia, and inflammation indicated by higher fecal neopterin, which was in turn correlated with overall illness. Taken together, our findings suggest that reinforcing exclusive breastfeeding, facilitating the intake of nutritious foods after 6 months of age and preventing respiratory infections might help reducing the prevalence of child growth impairment in the Pakistani population.
## Data availability statement
The Mal-ED data sets are centrally available upon request from the MAL-ED data repository at the University of Pennsylvania. The analytical codes used for this analysis can be requested from the study authors.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethics Review Committee Aga Khan University and the Foundation for the National Institutes of Health. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.
## Author contributions
ZB is the principal investigator for the Pakistan MalED cohort. For this analysis DG-F, SC and ZB jointly developed the analytical design, the conceptual framework and wrote the manuscript. DG-F and SC performed statistical analysis. AR, IC, SS, and the MAL-ED Pakistan investigators coordinated field and laboratory analyses, read, and approved the final manuscript. All authors contributed to the article and approved the submitted version.
## Funding
This study was supported by the Bill and Melinda Gates Foundation and by the Microbiome, Infections, and Childhood Growth and Development Fellowship Program of the Centre for Global Child Health, Toronto.
## 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.2023.1104654/full#supplementary-material
## References
1. Micha R. **Global Nutrition Report (2022)**. *UN-Nutrition* (2021)
2. Keeley B.. *Children, food and nutrition, growing well in a changing world* (2019)
3. Victora CG, Adair L, Fall C, Hallal PC, Martorell R, Richter L. **Maternal and child undernutrition: consequences for adult health and human capital**. *Lancet* (2008) **371** 340-57. DOI: 10.1016/S0140-6736(07)61692-4
4. de Onis M, Branca F. **Childhood stunting: a global perspective**. *Matern Child Nutr* (2016) **12** 12-26. DOI: 10.1111/mcn.12231
5. Dobner J, Kaser S. **Body mass index and the risk of infection - from underweight to obesity**. *Clin Microbiol Infect* (2018) **24** 24-8. DOI: 10.1016/j.cmi.2017.02.013
6. de Onis M, Borghi E, Arimond M, Webb P, Croft T, Saha K. **Prevalence thresholds for wasting, overweight and stunting in children under 5 years**. *Public Health Nutr* (2019) **22** 175-9. DOI: 10.1017/S1368980018002434
7. Victora CG, Christian P, Vidaletti LP, Gatica-Dominguez G, Menon P, Black RE. **Revisiting maternal and child undernutrition in low-income and middle-income countries: variable progress towards an unfinished agenda**. *Lancet* (2021) **397** 1388-99. DOI: 10.1016/S0140-6736(21)00394-9
8. **Relationship between growth and illness, enteropathogens and dietary intakes in the first 2 years of life: findings from the MAL-ED birth cohort study**. *BMJ global health* (2017) **2** e000370. DOI: 10.1136/bmjgh-2017-000370
9. Kosek MN. **MAL-ED Network Investigators. Causal pathways from enteropathogens to environmental enteropathy: findings from the MAL-ED birth cohort study**. *EBioMedicine* (2017) **18** 109-17. DOI: 10.1016/j.ebiom.2017.02.024
10. Richard SA, McCormick BJJ, Murray-Kolb LE, Lee GO, Seidman JC, Mahfuz M. **Enteric dysfunction and other factors associated with attained size at 5 years: MAL-ED birth cohort study findings**. *AJCN* (2019) **110** 131-8. DOI: 10.1093/ajcn/nqz004
11. Turab A, Soofi SB, Ahmed I, Bhatti Z, Zaidi AK, Bhutta ZA. **Demographic, socioeconomic, and health characteristics of the MAL-ED network study site in rural Pakistan**. *Clin Infect Dis* (2014) **59** S304-9. DOI: 10.1093/cid/ciu391
12. McCormick BJJ, Murray-Kolb LE, Lee GO, Schulze KJ, Ross AC, Bauck A. **Intestinal permeability and inflammation mediate the association between nutrient density of complementary foods and biochemical measures of micronutrient status in young children: results from the MAL-ED study**. *AJCN* (2019) **110** 1015-25. DOI: 10.1093/ajcn/nqz151
13. Psaki S, Bhutta ZA, Ahmed T, Ahmed S, Bessong P, Islam M. **Household food access and child malnutrition: results from the eight-country MAL-ED study**. *Popul. Health Metr* (2012) **10** 24. DOI: 10.1186/1478-7954-10-24
14. Ambikapathi R, Kosek MN, Lee GO, Mahopo C, Patil CL, Maciel BL. **How multiple episodes of exclusive breastfeeding impact estimates of exclusive breastfeeding duration: report from the eight-site MAL-ED birth cohort study**. *Matern Child Nutr* (2016) **12** 740-56. DOI: 10.1111/mcn.12352
15. Richard SA, McCormick BJJ, Seidman JC, Rasmussen Z, Kosek MN, Rogawski ET. **Relationships among common illness symptoms and the protective effect of breastfeeding in early childhood in MAL-ED: an eight-country cohort study**. *Am J Trop Med Hyg* (2018) **98** 904-12. DOI: 10.4269/ajtmh.17-0457
16. Rogawski ET, Bartelt LA, Platts-Mills JA, Seidman JC, Samie A, Havt A. **Determinants and impact of Giardia infection in the first 2 years of life in the MAL-ED birth cohort**. *J. Pediatr. Infect. Dis. Soc.* (2017) **6** 153-60. DOI: 10.1093/jpids/piw082
17. Rogawski ET, Liu J, Platts-Mills JA, Kabir F, Lertsethtakarn P, Siguas M. **Use of quantitative molecular diagnostic methods to investigate the effect of enteropathogen infections on linear growth in children in low-resource settings: longitudinal analysis of results from the MAL-ED cohort study**. *Lancet Glob Health* (2018) **6** e1319-e28. DOI: 10.1016/s2214-109x(18)30351-6
18. Rogawski ET, Guerrant RL, Havt A, Lima IFN, Medeiros P, Seidman JC. **Epidemiology of Enteroaggregative Escherichia Coli Infections and Associated Outcomes in the Mal-Ed Birth Cohort**. *PLoS neglected tropical diseases* (2017) **11** e0005798. DOI: 10.1371/journal.pntd.0005798
19. 19.UNICEF. Evaluation Report: United Nations Maternal and Child Stunting Reduction Programme in Three Target Districts in Sindh. Pakistan. Islamabad: UNICEF Pakistan (2019).. *Evaluation Report: United Nations Maternal and Child Stunting Reduction Programme in Three Target Districts in Sindh* (2019)
20. Ali A. **Current status of malnutrition and stunting in pakistani children: what needs to be done?**. *J Am Coll Nutr* (2021) **40** 180-92. DOI: 10.1080/07315724.2020.1750504
21. Caulfield LE, Bose A, Chandyo RK, Nesamvuni C, de Moraes ML, Turab A. **Infant feeding practices, dietary adequacy, and micronutrient status measures in the MAL-ED study**. *Clin Infect Dis* (2014) **59** S248-54. DOI: 10.1093/cid/ciu421
22. Richard SA, Barrett LJ, Guerrant RL, Checkley W, Miller MA. **Investigators MAL-ED Network Investigators. Disease surveillance methods used in the 8-site MAL-ED cohort study**. *Clin Infect Dis* (2014) **59** S220-4. DOI: 10.1093/cid/ciu435
23. Lee GO, Richard SA, Kang G, Houpt ER, Seidman JC, Pendergast LL. **A comparison of diarrheal severity scores in the MAL-ED multisite community-based cohort study**. *J Pediatr Gastroenterol Nutr* (2016) **63** 466-73. DOI: 10.1097/mpg.0000000000001286
24. Platts-Mills JA, Babji S, Bodhidatta L, Gratz J, Haque R, Havt A. **Pathogen-specific burdens of community diarrhoea in developing countries: a multisite birth cohort study (MAL-ED)**. *Lancet Glob Health* (2015) **3** e564-75. DOI: 10.1016/S2214-109X(15)00151-5
25. Kosek M, Haque R, Lima A, Babji S, Shrestha S, Qureshi S. **Fecal markers of intestinal inflammation and permeability associated with the subsequent acquisition of linear growth deficits in infants**. *Am J Trop Med Hyg* (2013) **88** 390-6. DOI: 10.4269/ajtmh.2012.12-0549
26. Fluss R, Faraggi D, Reiser B. **Estimation of the Youden index and its associated cutoff point**. *Biom J* (2005) **47** 458-72. DOI: 10.1002/bimj.200410135
27. Vidmar SI, Cole TJ, Pan H. **Standardizing anthropometric measures in children and adolescents with functions for egen: update**. *Stata J* (2013) **13** 366-78. DOI: 10.1177/1536867X1301300211
28. Austin PC. **Bootstrap model selection had similar performance for selecting authentic and noise variables compared to backward variable elimination: a simulation study**. *J clin epidemiol* (2008) **61** 1009-17.e1. DOI: 10.1016/j.jclinepi.2007.11.014
29. Mannan H. **A practical application of a simple bootstrapping method for assessing predictors selected for epidemiologic risk models using automated variable selection**. *Int J Stat Appl* (2017) **7** 239-49. DOI: 10.5923/j.statistics.20170705.01
30. Nunez E, Steyerberg EW, Nunez J. **Regression modeling strategies**. *Rev Esp Cardiol* (2011) **64** 501-7. DOI: 10.1016/j.recesp.2011.01.019
31. Kim JH. **Multicollinearity and misleading statistical results**. *Korean J Anesthesiol* (2019) **72** 558-69. DOI: 10.4097/kja.19087
32. Little RJA. **A test of missing completely at random for multivariate data with missing values**. *J Am Stat Assoc* (1988) **83** 1198-202. DOI: 10.1080/01621459.1988.10478722
33. 33.National Institute of Population Studies - NIPS/Pakistan and ICF International. Pakistan Demographic and Health Survey 2012-13. Islamabad, Pakistan: NIPS/Pakistan and ICF International (2013).. *Pakistan Demographic and Health Survey 2012-13* (2013)
34. Scheffler C, Hermanussen M, Bogin B, Liana DS, Taolin F, Cempaka P. **Stunting is not a synonym of malnutrition**. *Eur J Clin Nutr* (2020) **74** 377-86. DOI: 10.1038/s41430-019-0439-4
35. Perumal N, Bassani DG, Roth DE. **Use and misuse of stunting as a measure of child health**. *J Nutr* (2018) **148** 311-5. DOI: 10.1093/jn/nxx064
36. Tanjung C, Prawitasari T, Rusli Sjarif D. **Comments on “Stunting is not a synonym of malnutrition”**. *Eur J Clin Nutr* (2020) **74** 527-8. DOI: 10.1038/s41430-020-0570-2
37. Leroy JL, Frongillo EA. **Perspective: what does stunting really mean? A critical review of the evidence**. *Adv nutr* (2019) **10** 196-204. DOI: 10.1093/advances/nmy101
38. Thurstans S, Sessions N, Dolan C, Sadler K, Cichon B, Isanaka S. **The relationship between wasting and stunting in young children: a systematic review**. *Matern Child Nutr* (2022) **18** e13246. DOI: 10.1111/mcn.13246
39. Sadler K, James PT, Bhutta ZA, Briend A, Isanaka S, Mertens A. **How can nutrition research better reflect the relationship between wasting and stunting in children? Learnings from the wasting and stunting project**. *J Nutr* (2023) **14;152** 2645-51. DOI: 10.1093/jn/nxac091
40. **Breastfeeding**. *Recommendations* (2022)
41. Goday PS, Lewis JD, Sang CJ, George DE, McGoogan KE, Safta AM. **Energy- and protein-enriched formula improves weight gain in infants with malnutrition due to cardiac and noncardiac etiologies**. *JPEN J Parenter Enteral Nutr* (2022) **46** 1270-82. DOI: 10.1002/jpen.2308
42. Rana R, McGrath M, Gupta P, Thakur E, Kerac M. **Feeding interventions for infants with growth failure in the first six months of life: a systematic review**. *Nutrients* (2020) **12**. DOI: 10.3390/nu12072044
43. Abizari AR, Ali Z, Essah CN, Agyeiwaa P, Amaniampong M. **Use of commercial infant cereals as complementary food in infants and young children in Ghana**. *BMC Nutr* (2017) **3** 72. DOI: 10.1186/s40795-017-0191-x
44. Borelli P, Blatt S, Pereira J, de Maurino BB, Tsujita M, de Souza AC. **Reduction of erythroid progenitors in protein-energy malnutrition**. *Br J Nutr* (2007) **97** 307-14. DOI: 10.1017/S0007114507172731
45. Beguin Y. **Soluble transferrin receptor for the evaluation of erythropoiesis and iron status**. *Clin. Chim. Acta* (2003) **329** 9-22. PMID: 12589962
46. Etiler N, Velipasaoglu S, Aktekin M. **Incidence of acute respiratory infections and the relationship with some factors in infancy in Antalya**. *Turkey. Pediatr Int* (2002) **44** 64-9. DOI: 10.1046/j.1442-200x.2002.01504.x
47. Dharmage SC, Rajapaksa LC, Fernando DN. **Risk factors of acute lower respiratory tract infections in children under five years of age**. *Southeast Asian J Trop Med Public Health* (1996) **27** 107-10. PMID: 9031411
48. Sinha RK, Dua R, Bijalwan V, Rohatgi S, Kumar P. **Determinants of stunting, wasting, and underweight in five high-burden pockets of four Indian states**. *Indian J Community Med* (2018) **43** 279-83. DOI: 10.4103/ijcm.IJCM_151_18
49. Denny FW, Loda FA. **Acute respiratory infections are the leading cause of death in children in developing countries**. *Am J Trop Med Hyg* (1986) **35** 1-2. DOI: 10.4269/ajtmh.1986.35.1
50. Simoes E, Cherian T, Chow J, Shahid-Salles S, Laxminarayan R, John J, Jamison DT, Breman JG, Measham AR. **Acute Respiratory Infections in Children**. *Disease Control Priorities in Developing Countries* (2006)
51. Amour C, Gratz J, Mduma E, Svensen E, Rogawski ET, McGrath M. **Epidemiology and impact of**. *Clin Infect Dis* (2016) **63** 1171-9. DOI: 10.1093/cid/ciw542
52. Platts-Mills JA, Taniuchi M, Uddin MJ, Sobuz SU, Mahfuz M, Gaffar SA. **Association between enteropathogens and malnutrition in children aged 6-23 mo in Bangladesh: a case-control study**. *AJCN* (2017) **105** 1132-8. DOI: 10.3945/ajcn.116.138800
53. Platts-Mills JA, Kosek M. **Update on the burden of**. *Curr Opin Infect Dis* (2014) **27** 444-50. DOI: 10.1097/qco.0000000000000091
54. Morseth MS, Henjum S, Schwinger C, Strand TA, Shrestha SK, Shrestha B. **Environmental enteropathy, micronutrient adequacy, and length velocity in Nepalese children: the MAL-ED birth cohort study**. *J Pediatr Gastroenterol Nutr* (2018) **67** 242-9. DOI: 10.1097/mpg.0000000000001990
55. Arndt MB, Richardson BA, Ahmed T, Mahfuz M, Haque R, John-Stewart GC. **Fecal markers of environmental enteropathy and subsequent growth in Bangladeshi children**. *Am J Trop Med Hyg* (2016) **95** 694-701. DOI: 10.4269/ajtmh.16-0098
56. Lanyero B, Grenov B, Barungi NN, Namusoke H, Michaelsen KF, Mupere E. **Correlates of gut function in children hospitalized for severe acute malnutrition, a cross-sectional study in Uganda**. *J Pediatr Gastroenterol Nutr* (2019) **69** 292-8. DOI: 10.1097/mpg.0000000000002381
|
---
title: 'Improvement in Hypertension Control Among Adults Seen in Federally Qualified
Health Center Clinics in the Stroke Belt: Implementing a Program with a Dashboard
and Process Metrics'
authors:
- Edward M. Behling
- Tammy Garris
- Vicky Blankenship
- Shaun Wagner
- David Ramsey
- Rob Davis
- Susan E. Sutherland
- Brent Egan
- Gregory Wozniak
- Michael Rakotz
- Karen Kmetik
journal: Health Equity
year: 2023
pmcid: PMC9982137
doi: 10.1089/heq.2022.0109
license: CC BY 4.0
---
# Improvement in Hypertension Control Among Adults Seen in Federally Qualified Health Center Clinics in the Stroke Belt: Implementing a Program with a Dashboard and Process Metrics
## Abstract
### Objective:
Attain $75\%$ hypertension (HTN) control and improve racial equity in control with the American Medical Association Measure accurately, Act rapidly, Partner with patients blood pressure (AMA MAP BP™) quality improvement program, including a monthly dashboard and practice facilitation.
### Methods:
Eight federally qualified health center clinics from the HopeHealth network in South Carolina participated. Clinic staff received monthly practice facilitation guided by a dashboard with process metrics (measure [repeat BP when initial systolic ≥140 or diastolic ≥90 mmHg; Act [number antihypertensive medication classes prescribed at standard dose or greater to adults with uncontrolled BP]; Partner [follow-up within 30 days of uncontrolled BP; systolic BP fall after medication added]) and outcome metric (BP <140/<90). Electronic health record data were obtained on adults ≥18 years at baseline and monthly during MAP BP. Patients with diagnosed HTN, ≥1 encounter at baseline, and ≥2 encounters during 6 months of MAP BP were included in this evaluation.
### Results:
Among 45,498 adults with encounters during the 1-year baseline, 20,963 ($46.1\%$) had diagnosed HTN; 12,370 ($59\%$) met the inclusion criteria ($67\%$ black, $29\%$ white; mean (standard deviation) age 59.5 (12.8) years; $16.3\%$ uninsured. HTN control improved ($63.6\%$ vs. $75.1\%$, $p \leq 0.0001$), reflecting positive changes in Measure, Act, and Partner metrics (all $p \leq 0.001$), although control remained lower in non-Hispanic black than in non-Hispanic white adults ($73.8\%$ vs. $78.4\%$, $p \leq 0.001$).
### Conclusions:
With MAP BP, the HTN control goal was attained among adults eligible for analysis. Ongoing efforts aim to improve program access and racial equity in control.
## Introduction
The excess stroke mortality rate in the southeast United States led to its designation as the nation's Stroke Belt.1,2 For all but one decade between 1930 and 1990,3 South Carolina had the highest per capita stroke mortality rate in the southeast, that is, the “buckle” of the Stroke Belt.4 Within South Carolina, the Pee Dee Region, located in the northeast section of the state, has a very high rate of stroke-related deaths.5 HopeHealth, designated as a federally qualified health center (FQHC) in 2007, has 15 clinical sites in the Pee Dee Region of South Carolina serving a primarily rural area with multiple sociodemographic risk factors for adverse cardiovascular outcomes.6 HopeHealth established goals to attain a hypertension (HTN) control rate of $75\%$ and improve racial equity in control before discussions with the American Medical Association (AMA). Previous experience documented that the AMA MAP BP™ program (Measure accurately, Act rapidly, Partner with patients) could enable a rapid and sustained improvement in HTN control within clinical sites serving patients with sociodemographic risk for HTN-related morbidity and mortality rate.7–9 Given the burden of HTN in our patients, HopeHealth partnered with the AMA on their MAP BP quality improvement program.7 The eight clinical sites at HopeHealth, which provided primary care to ∼45,000 adults with encounters in the past year, participated in the MAP BP program described in this report. The other seven HopeHealth sites, which did not participate in MAP BP, predominantly provide medical services to children or other adult nonprimary care medical services.
The MAP BP program includes monthly dashboards and practice facilitation to promote efficient implementation and maintenance of key process changes that raise HTN control.9 The monthly dashboards with process metrics and HTN control metric at the clinic level were provided to the eight HopeHealth sites individually, and at the patient level for providers at each clinical site. This current report presents changes in HTN control, blood pressure (BP, mmHg), and process metrics during the 6-month MAP BP quality improvement program.
## Ethical and regulatory considerations
MAP BP was designated a quality improvement program and exempted from oversight by the AMA's Institutional Review Board of Record. The HopeHealth–AMA collaboration was conducted under a Business Associate Agreement, which allows exchange of personal health information required for quality improvement, and Data Use Agreement, which permits use of a limited data set for analysis and reporting.
## Inclusion criteria
Eligible patients had a diagnosis of essential HTN defined by ICD-10-CM I10, before program implementation and at least one recorded BP during the 1-year baseline period, defined as February 1, 2020, to January 31, 2021. Eligible individuals also had at least two encounters with recorded BP during the 6-month MAP BP intervention from March 1, 2021, to August 31, 2021. Adults not meeting these criteria were excluded from the main analysis.
## Implementation of the MAP BP quality improvement program
MAP BP began with virtual clinical site assessments at three HopeHealth locations to understand clinical workflows and processes related to BP measurement and HTN management.7 Separate 1-h training periods for physicians and other providers and care team members provided a program overview, the importance of M, A, and P processes in improving BP control, and an overview of the metrics and dashboard. Clinical care team members received an additional 1-h workshop and an hour-long orientation on the MAP BP dashboard. Approximately 13 h of virtual practice facilitation were then provided on M, A, and P by the AMA team members concurrently to participating clinical sites through tele-video conferences.
The goals of practice facilitation were to provide clinical education, program implementation support, review monthly data, and assist with logistical problem-solving required to efficiently incorporate MAP BP into daily operations at each site. Physicians, advanced practice nurses, clinical pharmacists, and physician assistants received a 45-min training on managing apparent treatment-resistant HTN.10,11
## Data acquisition
Two years of historical data were extracted on adults ≥18 years old from the enterprise data warehouse, which contained data from the Epic electronic health record system used by all HopeHealth clinical sites, before implementing MAP BP. Data were obtained monthly during implementation to enable monthly updating of dashboards and reports.
## Definitions and measurements
Hypertension was defined by a diagnosis in the medical record of ICD-10 codes I10–I16 during the 2 years before program implementation. Hypertension control was defined by team-measured or automated office (AO)BP <140/<90 based on the most recent value during the 12-month baseline and at the most recent visit during MAP BP.
## BP measurements
During the baseline, BP was measured according to usual practice at each site but did not include AOBP. HopeHealth promoted repeat BP measurement when the initial BP values were ≥140 systolic or ≥90 diastolic during the baseline.
BP measurements during MAP BP: Initial attended BP: Clinical care team members were trained to measure BP using proper patient preparation, positioning, and correct measurement technique.12,13 The measurement protocol was to obtain a single BP after the patient was seated for 5 min in a semiprivate area with values recorded in the electronic health record. Initial attended BP values ≥140 systolic or ≥90 mmHg diastolic led to recommendation for a confirmatory measurement, mostly with unattended AOBP. AOBP devices were available at all clinical sites.
Unattended AOBP was conducted in the patient's examination room or other private locations.13,14 Without additional patient rest, a clinic team member applied the upper arm cuff of the Welch Allyn Connex Spot, activated the device, and left the room. Once activated, the device obtained three AOBP measurements at 1-min intervals with the patient alone. When the AOBP measurements were completed, the team member returned and documented the mean of AOBP values in the electronic health record.
Health care insurance was determined by the primary payer source for adults with HTN and grouped as Medicare, Medicaid, private, and uninsured.
Body mass index (kg/m2) was calculated from the most recent height and weight during the baseline period and categorized as underweight or normal (<25), overweight (25–29.9), or obese (≥30).
Race and ethnicity were assessed by patient self-report during initial registration and annual updates. Patients have the option of entering race/ethnicity information into a kiosk or paper form at initial registration and annual updates. The patient information is then entered by reception staff into the electronic medical record. Patients were categorized as Hispanic ethnicity of any race, non-Hispanic black, non-Hispanic white, other, or unknown.
Comorbidities were defined by the International Classification of Diseases (ICD-10 CM codes) in the electronic health record data for diabetes mellitus (E10, E11, E13); chronic kidney disease (N18.3, N18.4, N18.5, N18l.6, N18.9, R80.9); and cardiovascular disease (G45.1-G46.8, I20.X–I25.X I46.X, I50.1–I50.9, I61.X, I63.X, I65.X, I66.X, I67.2, I67.8, I67.9, I68.8, I69.1, I69.2, I69.3, I69.8, I69.9, I70.X, I71.X, I72.0, I72.6, I73.89, I73.9, I74.X, I75.X, I79.0, K55.1, Z86.73, Z95.1, Z95.5, Z95.820, Z98.61, Z98.62).
## Key process variables
Measure accurately was assessed in adults with HTN as the proportion of visits with a confirmatory measurement relative to the number of visits with an initial attended BP ≥140 systolic or ≥90 diastolic.
Act rapidly was defined by therapeutic intensity as the number of antihypertensive medication classes prescribed at standard dose (half the recommended maximum dose) or greater to adults with uncontrolled HTN during the last visit of the baseline period and MAP BP program. The Act rapidly metric contrasts with previous reports,8,9 and was based on evidence that therapeutic inertia is persistently high15–18 and [1] ∼$80\%$ of the BP response to most antihypertensive medications occurs at standard or half-maximal dose,19 [2] a mean of ≥3 antihypertensive medication classes was required in clinical trials to attain strict control,20 and [3] <$20\%$ of adults with uncontrolled HTN in community-based clinics were prescribed three different BP medication classes at standard dose or higher.21 Partner with patients was assessed with two metrics: [1] changes in systolic BP in the 10- to 180-day window, but no later than August 31, 2021, after each therapeutic intensification, and [2] the percent of patients with a follow-up BP obtained within 30 days of an encounter with systolic BP ≥140 or diastolic BP ≥90 documented in the electronic health record.8,9 The reduction in systolic BP after intensification was attributed to the time-period in which the follow-up measurement was made. Process metrics were reported primarily at monthly intervals with 12-month rolling averages available.
## Monthly dashboards
Physician and provider champions, clinical care team champions, and administrative champions were registered for the dashboard and had access to the information described below. All registered users had access through a secure website to summary data for HTN control and the process metrics (Fig. 1) appropriate for their role. Senior leadership had access to data on HTN control and the four process metrics for the eight clinics collectively and individually.
**FIG. 1.:** *A sample dashboard (fictitious data) provides data for HTN control (outcome metric) and process metrics. The purple dial or bar depicts data for the individual clinician or clinical site, whereas the blue dial or bar depicts data for all clinicians (at the site or all sites within the health system), respectively. The sidebar (right) provides data for demographic subsets or comorbid conditions at the clinician, clinic, or health system level as determined by user authorization. HTN, hypertension.*
Leadership at each clinical site had data for clinicians at that site collectively and individually, while clinicians had access to data for their patients collectively and individually (Fig. 2). Patient-level reports on BP control and the four process metrics required two-factor authentication and were delivered through secure e-mail to registered users. Patient-level reports listed all patients in the denominator and distinguished by font color those who did not meet the metric objective (Fig. 2). Physicians, advanced practice nurses, physician assistants, and clinical pharmacists could compare their performance with other prescribers at their site or for all participating HopeHealth sites.
**FIG. 2.:** *A fictitious patient-level report showing patient name, encounter date, BP values, active antihypertensive medications, demographic details, and comorbid conditions is updated monthly. The list is sortable, for example, uncontrolled HTN or systolic BP (highest to lowest) to facilitate population management for better BP control. BP, blood pressure.*
## Statistical analysis
Descriptive statistics were used to summarize baseline demographic and clinical characteristics of adults with HTN. Data are reported as relative frequency or mean and $95\%$ confidence intervals. The primary outcome variable was change in HTN control from baseline to the most recent visit during the 6-month MAP BP program among adults in the primary analysis. Changes in MAP BP process variables comprised the secondary outcome. Sensitivity analysis included all patients in the baseline period who had the following: [1] one or more MAP BP visits and [2] irrespective of the number of MAP BP visits, that is, last baseline observation carried forward for those without MAP BP visits.
Pooled t and chi-square tests were performed to assess differences in demographic and clinical characteristics between patients included and excluded in the analysis, and for comparisons of patient groups. Satterthwaite's adjustment was used in pooled tests when group variances were not equal. Effect modification of race/ethnicity differences between the baseline and implementation periods was examined with analysis of variance models including the two main factors of time-period and race/ethnicity and their interaction. Paired t tests and McNemar tests were used to assess longitudinal outcomes and process measures related to M, A, and P within the non-Hispanic black and non-Hispanic white patient groups. All analyses were performed with two-tailed tests using SAS/STAT software.
## Results
Table 1 compares adults who were eligible and ineligible for the main evaluation. Adults eligible for analysis were older, more likely to be female and non-Hispanic black, and had a higher prevalence of obesity, diabetes, chronic kidney disease and cardiovascular disease than those excluded from the main analysis (all $p \leq 0.001$). Of note, $16.3\%$ of adults with HTN were uninsured.
**Table 1.**
| Variable group | All patients with HTN | MAP eligiblea | Ineligible |
| --- | --- | --- | --- |
| No. | 20963 | 12370 | 8593 |
| Age (years), mean | 58.2 [57.94, 58.46] | 59.5 [59.10, 59.87] | 56.3 [56.04, 56.65] |
| <45, % | 16.6% [16.12, 17.12] | 13.1% [12.53, 13.72] | 21.6% [20.77, 22.52] |
| 45–64, % | 50.1% [49.42, 50.77] | 50.7% [49.85, 51.61] | 49.2% [48.13, 50.25] |
| ≥65, % | 33.3% [32.64, 33.92] | 36.1% [35.30, 36.99] | 29.2% [28.20, 30.12] |
| Female, % | 60.9% [60.24, 62.56] | 63.8% [62.99, 64.68] | 56.7% [55.64, 57.73] |
| Non-Hispanic black, % | 65.4% [64.79, 66.07] | 67.4% [66.53,68.18] | 62.7% [61.63, 63.68] |
| Non-Hispanic white, % | 30.3% [29.72, 30.97] | 29.0% [28.17, 29.76] | 32.3% [31.34, 33.32] |
| Hispanic, % | 2.4% [2.22, 2.64] | 2.3% [1.99, 2.52] | 2.7% [2.35, 3.03] |
| Other, % | 1.1% [0.91, 1.19] | 0.8% [0.68, 1.00] | 1.4% [1.11, 1.59] |
| Unknown, % | 0.7% [0.63, 0.86] | 0.6% [0.45, 0.72] | 1.0% [0.77, 1.19] |
| Body mass index (kg/m2), mean | 32.9 [32.82, 33.06] | 33.2 [33.00, 33.32] | 32.6 [32.43, 32.81] |
| <25, % | 17.5% [16.96, 17.99] | 16.6% [15.92, 17.23] | 18.8% [17.96, 19.61] |
| 25–29.9 | 24.2% [23.66, 24.82] | 23.7% [22.92, 24.42] | 25.1% [24.15, 25.98] |
| ≥30, % | 58.3% [57.61, 58.95] | 59.8% [58.89, 60.62] | 56.2% [55.10, 57.20] |
| Diabetes mellitus, % | 37.7% [37.02, 38.33] | 43.5% [42.62, 44.37] | 29.3% [28.34, 30.27] |
| Chronic kidney disease, % | 17.6% [17.04, 18.07] | 21.5% [20.73, 22.18] | 11.9% [11.25, 12.63] |
| Cardiovascular disease, % | 12.2.% [11.74, 12.63] | 13.7% [13.06, 14.28] | 10.0% [9.41, 10.68] |
Changes in key variables between baseline and MAP are shown in Table 2. Systolic and diastolic BP fell as BP control rose from $63.6\%$ at baseline to $75.1\%$ after 6 months of MAP BP (+$11.5\%$, $p \leq 0.001$). The sensitivity analysis (not shown) found that HTN control rose from $63.2\%$ at baseline to $73.7\%$ (+$10.5\%$ $p \leq 0.001$) among 16,590 adults with a baseline visit and at least one visit during MAP BP. Among all the 20,370 adults with a baseline visit, irrespective of follow-up, HTN control rose from $61.7\%$ at baseline to $70.2\%$ during MAP BP (+$8.5\%$, $p \leq 0.001$).
**Table 2.**
| Variables | Baseline | MAP 6 Months |
| --- | --- | --- |
| Systolic BP (mmHg), mean | 134.8 [134.50, 135.13] | 129.9 [129.58, 130.18] |
| Diastolic BP (mmHg), mean | 80.2 [80.02, 80.38] | 77.3 [77.11, 77.45] |
| BP <140/<90, % | 63.6% [62.72, 64.41] | 75.1% [74.31, 75.84] |
| BP <130/<80, % | 26.8% [26.08, 27.64] | 38.8% [37.90, 39.61] |
| Repeat BP, % | 39.5% [38.50, 40.51] | 48.6% [47.47, 49.81] |
| Repeat BP ΔSBP (mmHg), mean | 10.8 [10.52, 11.12] | 11.0 [10.63, 11.26] |
| Prescribed ≥1 BP medication, % | 93.7% [93.27, 94.13] | 94.6% [94.22, 95.01] |
| Therapeutic intensity, mean | 1.78 [1.76, 1.81] | 1.91 [1.88, 1.94] |
| No. of BP medications, mean | 2.34 [2.32, 2.37] | 2.44 [2.42, 2.47] |
| ΔSBP Rx Int (mmHg), mean | 13.8 [13.12, 14.52] | 16.9 [15.05, 17.69] |
| 30-Day follow-up, % | 23.0% [22.09, 23.86] | 29.0% [27.93, 30.14] |
Measure accurately, assessed by the proportion of adults with a confirmatory measurement when initial BP was uncontrolled improved (+$9.1\%$, $p \leq 0.001$). Systolic BP with repeat measurement was similar at ∼11 mmHg below the initial value during baseline and MAP BP.
Act rapidly, defined by therapeutic intensity in adults with uncontrolled HTN, increased. The mean number of BP medication classes prescribed at any dose rose from 2.45 [2.43, 2.48] at baseline to 2.62 [2.58, 2.65], during MAP BP (data not shown).
The number of antihypertensive medication classes prescribed at any dose and at standard dose or higher for adults with uncontrolled HTN is shown in Figure 3. The percentages prescribed ≥3 BP medication classes at standard dose versus. any dose or higher where roughly one-third versus 9 of 16.
**FIG. 3.:** *A frequency histogram shows the percentage of adults with uncontrolled HTN who are prescribed a given number of antihypertensive medications at standard dose or higher or any dose. In general, adults with uncontrolled HTN prescribed 0–2 BP medication classes at standard dose are candidates for an additional antihypertensive medication class, whereas the subset prescribed 3 or more medications often merits evaluation for (apparent) treatment resistance before adding medication.*
Partner with patients metrics included the [1] decline in SBP after adding a new antihypertensive medication class, a proxy for adherence, rose from 13.8 to 16.9 mmHg, and [2] percentage of patients having a return visit within 30 days of an uncontrolled BP rose from $23\%$ to $29\%$. Monthly changes from baseline in outcome and process metrics are shown (Fig. 4).
**FIG. 4.:** *Monthly changes during the 6-month MAP BP intervention in HTN control and the process metrics are shown for the 12,370 adults in the main analysis. Time trends for HTN control and all process metrics were statistically significant at p<0.0001.*
Significant improvements were seen within the non-Hispanic black and non-Hispanic white patient groups for BP control and process metrics (Table 3). Therapeutic intensity increased slightly but statistically significant from baseline during MAP BP among both black (1.95 vs. 2.07) and white (1.40 vs. 1.55) adults. More non-Hispanic white than non-Hispanic black adults had untreated HTN, whereas more non-Hispanic black than non-Hispanic white adults were prescribed ≥3 BP medications. The decline in systolic BP after adding an antihypertensive medication class and 30-day follow-up for uncontrolled HTN was greater in non-Hispanic white than non-Hispanic black adults.
**Table 3.**
| Unnamed: 0 | Non-Hispanic black adults (N=8332) | Non-Hispanic black adults (N=8332).1 | Non-Hispanic white adults (N=3583) | Non-Hispanic white adults (N=3583).1 |
| --- | --- | --- | --- | --- |
| Metric | Baseline | MAP BP | Baseline | MAP BP |
| HTN control, % | 61.7% [60.68, 62.77] | 73.8% [72.81, 74.70] | 67.9% [66.35, 69.41] | 78.4% [77.02, 79.72] |
| Repeat BP, % | 42.7% [42.45, 43.89] | 49.1% [47.70, 50.49] | 31.6% [29.72, 33.42] | 47.7% [45.50, 49.99] |
| Therapeutic intensity, N | 1.95 [1.92, 1.98] | 2.07 [2.03, 2.11] | 1.40 [1.36, 1.45] | 1.55 [1.49, 1.60] |
| No BP medication, % | 5.3% [4.84, 5.80] | 4.4% [3.99, 4.87] | 8.2% [7.33, 9.13] | 7.3% [6.43, 8.14] |
| 1–2 BP medications, % | 47.8% [46.76, 47.93] | 45.7% [44.66, 46.80] | 58.2% [56.60, 59.83] | 55.8% [54.19, 57.45] |
| ≥3 BP medications, % | 46.9% [45.78, 47.93] | 49.8% [48.77, 50.92] | 33.6% [32.00, 35.09] | 36.9% [35.32, 38.48] |
| ΔSBP Rx Int, mmHg | 13.5 [12.68, 14.36] | 16.3 [15.35, 17.35] | 15.0 [13.56, 16.39] | 18.4 [16.86, 19.87] |
| 30-Day follow-up, % | 22.3% [22.09, 23.86] | 28.6%27.93, 30.14] | 24.8% [23.07, 26.58] | 30.3% [28.15, 32.47] |
Antihypertensive medication classes and most prescribed medications within each class to all adults with HTN (Table 4) as well as a frequency histogram (Fig. 3) of medication classes prescribed to adults with uncontrolled HTN were provided at the system, clinic, and clinician levels to facilitate appropriate treatment intensification. Table 4 included the standard or half-maximal dose, at which ∼$80\%$ of the antihypertensive effect occurs, and the percent of adults with HTN prescribed standard dose or higher for each antihypertensive medication listed.
**Table 4.**
| Medication class subclass | Generic name | Patients, n (%) | Standard dose (mg/day) | Standard dose, n (%) |
| --- | --- | --- | --- | --- |
| RASB | RASB | 8498 (68.7) | | |
| ACEI | Lisinopril | 3586 (29.0) | 20.0 | 2209 (61.6) |
| ARB | Losartan | 3240 (26.2) | 50.0 | 2911 (89.8) |
| ARB | Valsartan | 846 (6.8) | 160.0 | 629 (74.5) |
| ACEI | Benazepril | 755 (6.1) | 20.0 | 578 (76.7) |
| Diuretics | Diuretics | 7857 (63.5) | | |
| Thiazide | HCTZ | 6152 (49.7) | 25.0 | 4087 (66.4) |
| Loop | Furosemide | 1767 (14.3) | 40.0 | 905 (51.5) |
| Aldo Ant | Spironolactone | 643 (5.2) | 25.0 | 633 (98.8) |
| CCB | CCB | 6454 (52.2) | | |
| Dihydropyridine | Amlodipine | 5741 (46.4) | 5.0 | 5431 (94.6) |
| β-blockers | β-blockers | 4148 (33.5) | | |
| β1-blocker | Metoprolol | 2459 (19.9) | 100.0 | 881 (35.9) |
| α,β-blocker | Carvedilol | 1030 (8.3) | 25.0 | 594 (57.8) |
## Discussion
HopeHealth, located in the buckle of the stroke belt, serves a diverse population with historically high risk for HTN and cardiovascular events.1–6 In fact, $46\%$ of adults seen during the 1-year baseline period before implementing MAP BP had an HTN diagnosis. Before discussions with the AMA, HopeHealth established goals to raise HTN control to ≥$75\%$ and improve racial equity in control. To facilitate goal attainment, HopeHealth collaborated with the AMA on MAP BP.8,9 During the 6-month MAP BP program, the first goal was attained as BP control to <140/<90 rose from $63.6\%$ at baseline to $75.1\%$ for patients qualifying for the main analysis. Changes in MAP BP process metrics drive improvements in HTN control. Regarding Measure accurately, the mean decline in systolic BP with repeated measurements was similar during baseline and MAP BP. However, the increased frequency of repeat BP measurements contributed to lower BP and improved control.
Act rapidly, or the number of antihypertensive medication classes prescribed at standard dose or higher to adults with uncontrolled BP, rose modestly from 1.8 to 1.9. Roughly two-thirds of adults with uncontrolled HTN were not prescribed three different antihypertensive medication classes at standard dose or higher, the minimum required for apparent treatment-resistant HTN.10,11 With regard to Partner metrics, the mean fall in systolic BP following addition of an antihypertensive medication for uncontrolled HTN increased, which suggests greater patient engagement in obtaining and taking the additional medication and possibly prescribing more effective classes or combinations.22 The proportion of adults who had a follow-up encounter within 30 days of uncontrolled BP rose modestly and significantly from $23\%$ to $29\%$.17 Thus, both Partner components of MAP BP contributed to the fall in BP and rise in control.
The second goal to improve racial equity in HTN control was unmet. HTN control improved significantly in non-Hispanic black and non-Hispanic white adults (Table 3). The improvement in HTN control was not significantly greater in non-Hispanic black than in non-Hispanic white adults ($12.0\%$ vs. $10.4\%$, $$p \leq 0.16$$), and a disparity persisted ($73.8\%$ vs. $78.4\%$, $p \leq 0.0001$).
A racial comparison of process metrics may inform efforts to improve equity in HTN control. At baseline, the Measure accurately process was performed more often in non-Hispanic black than in non-Hispanic white adults. The gap closed during MAP BP but remained slightly higher in non-Hispanic black than in non-Hispanic white adults. While opportunity for improvement exists, repeat BP measurement did not explain disparities in control.
Act rapidly, assessed by therapeutic intensity, was greater in black than in white adults at baseline and increased similarly in both groups during MAP BP. Our previous research suggested that black adults had less access to prescribed antihypertensive medication at civilian practice sites, which contributed to higher BP and less control in the former, despite prescription of more antihypertensive medications.23 The observation that systolic BP fell less in non-Hispanic black than in non-Hispanic white adults following treatment intensification during the MAP BP program in the current report is consistent with differential access to medications. Follow-up within 30 days of an encounter with uncontrolled HTN was slightly lower in non-Hispanic black than in non-Hispanic white adults, while both groups improved during MAP BP.
The observations suggest that increasing access to antihypertensive medications and follow-up frequency for uncontrolled HTN in black adults could improve equity in HTN control. Since black adults have severe HTN more often than white adults and can encounter more structural barriers to healthy lifestyles and medical care,24 greater performance on process metrics in black than white adults may be required to attain equity in control.
There is growing attention to structural racism, rather than biology, as the driver of disparate health outcomes between black and white adults.24–28 FQHCs began as a key initiative to mitigate health disparities.29 The mean difference in HTN control between non-Hispanic black and non-Hispanic white adults in this study was $4.6\%$, which is substantially less than the $15\%$ mean difference between non-Hispanic black and non-Hispanic white adults with treated HTN in the civilian population during 2017–2018.30 The observations suggest that HopeHealth is reducing health disparities, while striving to eliminate them.
## Health equity implications
MAP BP contributed to the goal of controlling HTN in $75\%$ of eligible adults. MAP BP was more effective when adults with HTN have more visit opportunities to derive program benefits. To reach those with inadequate follow-up, HopeHealth is assessing the value of telemedicine and self-monitored BP using the MAP BP framework.31,32 To better address Act rapidly, HopeHealth plans to add clinical pharmacists to team-based care for HTN to facilitate appropriate pharmacotherapy and provide the resources and supports, for example, self-monitored BP and telemedicine, to assist patients in overcoming barriers to BP control. A new hypertension clinic is also planned to begin later in 2022 for patients with uncontrolled and treatment-resistant HTN. To facilitate appropriate pharmacotherapy, the AMA will change the metric from therapeutic intensity to therapeutic intensification or the percentage of encounters with uncontrolled BP at which an antihypertensive medication class is added. The updated metric is applicable to most patients with uncontrolled HTN that is not treatment resistant. The AMA will continue providing training on management of treatment-resistant HTN.10,11
## Authors' Contributions
E.B.: Conceptualization and writing—review and editing. T.G., V.B., and R.D.: Writing—review and editing. S.W. and D.R.: Software, investigation, data curation, and writing—review and editing. S.E.S.: Methodology, formal analysis, writing—original draft, and writing—review and editing. B.E.: Conceptualization, validation, writing—original draft, writing—review and editing, supervision, and project administration. G.W.: Validation and writing—review and editing. M.R.: Resources, writing—review and editing, and supervision. K.K.: Resources, writing—review and editing, supervision, and funding acquisition. Ben Reisser (acknowledged): Visualization. Farris Linett (acknowledged): Project administration.
## Disclaimer
This article contains the views of the authors and should not be interpreted as official American Medical Association policy.
## Author Disclosure Statement
No competing financial interests exist.
## Funding Information
Measure accurately, Act rapidly, Partner with patients blood pressure (MAP BP) is funded by the American Medical Association (AMA) with no external costs to clinical partners.
## References
1. Hall WD, Ferrario CM, Moore MA. **Hypertension-related morbidity and mortality in the southeastern United States**. *Am J Med Sci* (1997) **313** 195-209. PMID: 9099149
2. Howard G, Howard VJ. **Twenty years of progress toward understanding the stroke belt**. *Stroke* (2020) **51** 742-750. PMID: 32078485
3. Perry HM, Roccella EJ. **Conference report on stroke mortality in the Southeastern United States**. *Hypertension* (1998) **31** 1206-1215. PMID: 9622131
4. Boan AD, Fen W, Ovbiagele B. **Persistent racial disparity in stroke hospitalization and economic impact in young adults in the buckle of the stroke belt**. *Stroke* (2014) **45** 1932-1938. PMID: 24947293
5. Lackland DT, Bachman DL, Carter TD. **The geographic variation in stroke incidence in two areas of the Southeastern stroke belt: The Anderson and Pee Dee Stroke Study**. *Stroke* (1998) **29** 2061-2068. PMID: 9756582
6. **American Community Survey 5-Year Estimates. Retrieved from Census Reporter Profile Page for Upper South Carolina Pee Dee SDTSA**. (2019)
7. Boonyasai RT, Rakotz MK, Lubomski LH. **Measure accurately, Act rapidly, and Partner with patients: An intuitive and practical three-part framework to guide efforts to improve hypertension control**. *J Clin Hypertension* (2017) **19** 684-694. DOI: 10.1111/jch.12995
8. Hanlin RB, Asif IM, Wozniak G. **Measure accurately, Act rapidly, and Partner with patients (MAP) improves hypertension control in medically underserved patients: Care Coordination Institute and American Medical Association Hypertension Control Project Pilot Study results**. *J Clin Hypertension* (2018) **20** 79-87. DOI: 10.1111/jch.13141
9. Egan BM, Sutherland S, Rakotz M. **Improving hypertension control in primary care with the Measure accurately, Act rapidly and Partner with patients (MAP) protocol: Results at 6 and 12 months**. *Hypertension* (2018) **72** 1320-1327. PMID: 30571231
10. Egan BM, Zhao Y, Axon RN. **Uncontrolled and apparent treatment resistant hypertension in the U.S. 1988–2008**. *Circulation* (2011) **124** 1046-1058. PMID: 21824920
11. Carey RM, Calhoun D, Bakris G. **Resistant hypertension: Detection, evaluation and management. A scientific statement from the American Heart Association**. *Hypertension* (2018) **72** e53-e90. PMID: 30354828
12. Whelton PK, Carey RM, Aronow WS. **2017 ACC/AHA Guideline for the prevention, detection, evaluation, and management of high blood pressure in adults**. *Hypertension* (2018) **71** e13-e115. PMID: 29133356
13. Muntner P, Shimbo D, Carey RM. **Measurement of blood pressure in humans: A scientific statement from the American Heart Association**. *Hypertension* (2019) **73** e35-e66. PMID: 30827125
14. Myers MG. **Eliminating the human factor in office blood pressure measurement**. *J Clin Hypertension* (2014) **26** 83-86
15. Berlowitz DR, Ash AS, Hickey EC. **Inadequate management of blood pressure in a hypertensive population**. *NEJM* (1998) **339** 1957-1963. PMID: 9869666
16. Okonofua EC, Simpson K, Jesri A. **Therapeutic inertia is an impediment to achieving Healthy People 2010 blood pressure control goals**. *Hypertension* (2006) **47** 1-7. PMID: 16344362
17. Bellows BK, Ruiz-Negrón N, Bibbins-Domingo K. **Clinic-based strategies to reach United States million hearts 2022 blood pressure control goals**. *Circ Cardiovasc Qual Outcomes* (2019) **12** e005624. PMID: 31163981
18. Cooper-DeHoff RM, Fontil V, Carton T. **Tracking blood pressure control performance and process metrics in 25 US health systems: The PCORnet Blood Pressure Control Laboratory**. *JAHA* (2021) **10** e022224. PMID: 34612048
19. Law MR, Wald NJ, Morris K. **Value of low dose combination treatment with blood pressure lowering drugs: Analysis of 354 randomised trials**. *BMJ* (2003) **326** 1427. PMID: 12829555
20. Bakris GL. **A practical approach to achieving recommended blood pressure goals in diabetic patients**. *Arch Intern Med* (2001) **161** 2661-2667. PMID: 11732930
21. Egan BM, Zhao Y, Li J. **Prevalence of optimal treatment regimens in patients with apparent treatment resistant hypertension in a community-based practice network**. *Hypertension* (2013) **62** 691-697. PMID: 23918752
22. Egan BM, Yang J, Rakotz MK. **Self-reported antihypertensive medication class and relationship to treatment guidelines**. *Hypertension* (2022) **79** 338-348. PMID: 34784722
23. Rehman SU, Hutchison FN, Hendrix K. **Ethnic differences in blood pressure control at Veterans Affairs clinics and other healthcare sites**. *Arch Int Med* (2005) **165** 1041-1047. PMID: 15883244
24. Churchwell K, Elkind MSV, Benjamin RM. **AHA Presidential Advisory—Call to action: Structural racism as a fundamental driver of health disparities**. *Circulation* (2020) **142** e454-e468. DOI: 10.1161/CIR.0000000000000936
25. Bailey ZD, Feldman JM, Bassett MT. **How structural racism works—Racist policies as a root cause of U.S. racial health inequities**. *N Engl J Med* (2021) **384** 768-773. DOI: 10.1056/NEJMms2025396
26. Oni-Orisan A, Mavura Y, Banda Y. **Embracing genetic diversity to improve black health**. *N Engl J Med* (2021) **384** 1163-1167. PMID: 33567186
27. Svetkey LP, Moore TJ, Simons-Morton DG. **Angiotensinogen genotype and blood pressure response in the Dietary Approaches to Stop Hypertension (DASH) study**. *J Hypertens* (2001) **19** 1949-1956. PMID: 11677359
28. Hunt SC, Geleijnse JM, Wu LL. **Enhanced blood pressure response to mild sodium reduction in subjects with the 235T variant of the angiotensinogen gene**. *Am J Hypertens* (1999) **12** 460-466. PMID: 10342783
29. Rosenbaum S, Hawkins DR. **The good doctor—Jack Geiger, social justice, and U.S. health policy**. *N Engl J Med* (2021) **384** 983-985. PMID: 33734649
30. Muntner P, Hardy ST, Fine LJ. **Trends in blood pressure control among US adults with hypertension, 1999–2000 to 2017–2018**. *JAMA* (2020) **324** 1190-1200. PMID: 32902588
31. Jackson GL, Oddone EZ, Olsen MK. **Racial differences in the effect of a telephone-delivered hypertension disease management program**. *J Gen Intern Med* (2012) **27** 1682-1689. PMID: 22865016
32. McManus RJ, Mant J, Franssen M. **Efficacy of self-monitored blood pressure, with or without telemonitoring, for titration of antihypertensive medication (TASMINH4): An unmasked randomized controlled trial**. *Lancet* (2018) **391** 949-959. PMID: 29499873
|
---
title: 'Association between urinary nickel with obesity status in adults: A cross-sectional
study'
authors:
- Gao-Xiang Wang
- Bao-Li Huang
- Jun-Tong Li
- Ze-Bin Fang
- Le-Yi Feng
- Heng-Xia Zhao
- Shu-Fang Chu
- De-Liang Liu
- Hui-Lin Li
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9982146
doi: 10.3389/fpubh.2023.1094062
license: CC BY 4.0
---
# Association between urinary nickel with obesity status in adults: A cross-sectional study
## Abstract
### Objectives
The prevalence of obesity is on the rise and is connected to numerous factors. However, the relationship between obesity and nickel has never been investigated. Our study aimed to explore the association between urinary nickel and obesity Status in adults.
### Methods
From the 2017–2018 National Health and Nutrition Examination Surveys (NHANES), 1,705 participants ≥18 years of age were enrolled. To explore further the relationship among urinary nickel, body mass index (BMI), and waist circumference(WC), Weighted multivariate linear regression analyses and further subgroup analyzes were conducted.
### Results
Urinary nickel does not correlate with BMI level but positively correlates with WC. In the subgroup analyzed according to sex, Urinary nickel has a positive correlation with BMI and WC in males but has a negative correlation in females. Secondary stratification analysis according to sex and race, Urinary nickel positively correlates with BMI in White males. It also positively correlates with WC in both White and Black males.
### Conclusions
A correlation was found between urinary nickel levels and BMI and WC in adult males. Adult men, especially those already obese, may need to reduce nickel exposure.
## Introduction
Nickel occupies the 28th spot in the periodic table. It is a brutal metal found naturally in air, water, and soil [1]. Nickel is vital for microorganisms, plants, animals, and humans [2]. Nickel deficiency can cause growth retardation and fecundity decline, impairment of specific senses, reduced iron absorption, and alteration of essential enzymes in animal tissues and organs, leading to various clinical changes (3–5). However, specific toxicity and carcinogenic properties are connected with excessive nickel. In humans, numerous health issues, including contact dermatitis, cardiovascular conditions, and lung and nasal cancer, can result from prolonged exposure to nickel [6]. Nickel exposure most commonly occurs through respiratory inhalation [7], food and water intake [8], and skin absorption [9]. With the extensive use of nickel-containing products in daily life [10], especially in medical devices [11], great attention is paid to nickel-related health issues.
Globally, obesity has become a severe public health issue [12]. Research shows that $70\%$ of American and $50\%$ of Chinese adults are overweight or obese [13, 14]. There are several diseases associated with obesity, such as hypertension [15], malignant tumors [16], and diabetes [17]. Several investigations conducted during the COVID-19 pandemic also revealed that obese people with COVID-19 infection had much greater rates of severe illness and fatality than normal persons [18, 19]. Since obesity constitutes a significant threat to health, a more profound knowledge of relevant factors of obesity is necessary. Recent studies have confirmed a correlation between urinary nickel and the prevalence of diabetes and high blood pressure [20, 21]. As everyone knows, diabetes and high blood pressure are closely related to obesity, but the relationship between urinary nickel and obesity status is unclear [22].
Body mass index (BMI) can effectively assess the state of health, has the characteristics of simple, feasible, and non-invasive [23, 24], and is an essential indicator for the diagnosis of obesity. Waist circumference (WC) has become increasingly crucial in predicting death and morbidity in recent years, and the combination of WC and BMI has been emphasized in diagnosing obesity [25, 26]. In this study, we conducted a cross-sectional study to explore the relationship between urinary nickel and obesity index (BMI and WC). To our knowledge, this is the first study to examine the relationship between obesity status and nickel exposure, which is of great significance to this field.
## Study population
In this study, our subjects included adults (≥18 years) from NHANES during 2017–2018. Study participants had 1,705 people after eliminating those with missing data regarding urinary nickel, BMI, or WC. The selection process is depicted in Figure 1.
**Figure 1:** *Flow chart for selecting participants.*
## Ethics statement
Participation in the study was voluntary, and the National Center for Health Statistics Research Ethics Review Board approved this study's conduct. To protect everyone's privacy, NHANES will anonymize collected data before making it public as public data. We agree to follow all guidelines for using NHANES data for research purposes and comply with all applicable standards and laws.
Participation in the study was voluntary, and National Center for Health Statistics Research Ethics Review Board approved the study's conduct. To protect everyone's privacy, NHANES will anonymize collected data before making it public as public data. It is our agreement to follow all guidelines for using NHANES data for research purposes, as well as comply with all applicable standards and laws. The patients/participants provided their written informed consent to participate in this study.
## Urinary nickel, BMI, and WC
It was decided to collect a single spot urine sample and store it at ≤-20°C for long-term or short-term analysis stored at 2–8°C before analysis. Inductively coupled plasma mass spectrometry (ICP-MS) was employed to determine nickel levels in urine, which is a susceptible technique that can measure multiple elements at low concentrations. To be brief, ICPs operate with argon flows passing through an atomizer and spray chamber to process urine samples. The sample vaporizes at a high temperature, dissociates the ionized gas, and then the ions reach the ion detector. Finally, the isotope ratio of the elements is measured. A urinary nickel concentration of 0.31 mg/L is considered a detection limit of detection. More detailed laboratory procedure manuals are shown on the NHANES's website [27].
This study measured the weight, height, and WC of adults 18 and older using standardized methods. Physical examination measured BMI and WC. BMI is calculated by dividing the square of a person's weight (in kilograms) by their height (in meters), and precise measurements could well be acquired using standard digital scales and rulers. Medical professionals measured the subjects' WC with a flexible ruler. According to the WHO-recommended measurement method, the subject's feet were separated by 25–30 cm. The measurer placed the measuring tape around the abdomen in a circle at the midpoint of the line connecting the anterior superior iliac crest and the lower border of the 12th rib, close to the soft tissue, but without compression, and measured at the end of exhalation and before inspiration. When the WC and BMI are both normal, it is not obesity; when the BMI is normal, but the WC of men is ≥94 cm, and that of women is ≥80 cm, it is central obesity; when the WC is standard, but the BMI is ≥30 kg/m2, it is defined as peripheral obesity; when the BMI and WC are both above normal, it is defined as mixed obesity [28, 29].
## Covariates
The information on age, race, the ratio of family income to poverty, and the educational level of the participants was obtained through the questionnaire. To ensure the quality of the questionnaire and data collection, questionnaires are developed in advance by professional surveyors and released. Data is received by trained medical staff at a mobile medical examination center. Qualified laboratory specialists collected and processed blood samples at the mobile medical examination center. The following parameters will be evaluated: total cholesterol, triglycerides, glycohemoglobin, blood urea nitrogen, serum creatinine, serum uric acid, and total protein. Hispanic, Mexican American, Non-Hispanic White, Non-Hispanic Black, and Other Race were classified. Education levels below high school, high school, and higher education were ranked based on their education level. The NHANES website (www.cdc.gov/nchs/nhanes/) provides public access to the data from this survey.
## Statistical analysis
This study used statistical software R (version 3.4.4) to conduct all statistical analyzes. Per National Center for Health Statistic (NCHS) recommendations, samples were weighted according to NHANES [30]. An analysis of the associations between urinary nickel, BMI, and WC was conducted using weighted linear regression. In this study, we built three regression models. No adjustments had been made to Model 1: race, gender, and age adjustment were made in Model 2. A complete adjustment was made to Model 3 for all Covariates. To further explore the relationship between urinary nickel, BMI, and WC, Weighted multivariate linear regression analyses and subgroup analyzes were conducted. A p-value of <0.05 determined statistical significance.
## Description of participant characteristics
The statistical characteristics of the study population are displayed in Table 1. A total of 1,705 adults participated in this study. In different groups of urinary nickel (quartiles, Q1-Q4), gender, age, triglycerides, serum creatinine, and BMI were not statistically significant. In contrast, race/ethnicity, education level, the ratio of family income to poverty, total cholesterol, glycohemoglobin, blood urea nitrogen, serum uric acid, total protein, WC, and obesity status were statistically significant. The main types of obese people are mixed and central obesity.
**Table 1**
| Urinary nickel | Total | Q1 | Q2 | Q3 | Q4 | P-value |
| --- | --- | --- | --- | --- | --- | --- |
| Gender (%) | | | | | | 0.217 |
| Male | 48.61 | 46.87 | 49.79 | 52.12 | 45.36 | |
| Female | 51.39 | 53.13 | 50.21 | 47.88 | 54.64 | |
| Age (years) | 47.04 ± 17.54 | 46.83 ± 15.62 | 46.92 ± 17.75 | 46.43 ± 18.07 | 48.16 ± 18.86 | 0.554 |
| Race/ethnicity (%) | | | | | | 0.010 |
| Mexican American | 9.16 | 8.96 | 9.68 | 7.93 | 10.19 | |
| Other Hispanic | 6.37 | 7.34 | 5.34 | 6.57 | 6.14 | |
| Non-Hispanic White | 62.64 | 63.46 | 66.64 | 64.25 | 54.84 | |
| Non-Hispanic Black | 11.32 | 8.6 | 10.18 | 13.12 | 14.11 | |
| Other race | 10.52 | 11.63 | 8.17 | 8.13 | 14.71 | |
| Education level (%) | | | | | | 0.007 |
| Less than high school | 11.21 | 9.73 | 9.05 | 12.82 | 13.87 | |
| High school | 27.79 | 23.85 | 27.97 | 32.46 | 27.21 | |
| More than high school | 61.01 | 66.43 | 62.97 | 54.73 | 58.92 | |
| Ratio of family income to poverty (%) | 2.97 ± 1.58 | 3.29 ± 1.56 | 2.99 ± 1.59 | 2.86 ± 1.59 | 2.67 ± 1.50 | <0.001 |
| Total cholesterol (mmol/L) | 4.83 ± 1.00 | 4.94 ± 1.03 | 4.93 ± 1.00 | 4.74 ± 1.00 | 4.68 ± 0.93 | <0.001 |
| Triglyceride (mmol/L) | 1.59 ± 1.32 | 1.51 ± 0.94 | 1.62 ± 1.37 | 1.59 ± 1.00 | 1.63 ± 1.88 | 0.476 |
| Glycohemoglobin (%) | 5.67 ± 0.90 | 5.57 ± 0.68 | 5.68 ± 1.00 | 5.72 ± 0.97 | 5.71 ± 0.93 | 0.046 |
| Blood urea nitrogen (mmol/L) | 5.32 ± 1.84 | 5.03 ± 1.64 | 5.22 ± 1.82 | 5.51 ± 1.76 | 5.59 ± 2.12 | <0.001 |
| Serum creatinine (umol/L) | 77.29 ± 23.18 | 76.29 ± 18.12 | 76.33 ± 25.40 | 78.60 ± 18.98 | 78.21 ± 29.52 | 0.313 |
| Serum uric acid (umol/L) | 319.33 ± 80.89 | 311.22 ± 79.13 | 316.48 ± 80.56 | 326.22 ± 83.56 | 325.23 ± 79.22 | 0.016 |
| Total protein (g/L) | 71.09 ± 4.21 | 71.28 ± 3.86 | 71.53 ± 4.08 | 70.66 ± 4.42 | 70.81 ± 4.47 | 0.008 |
| Body mass index (kg/m2) | | 29.14 ± 6.67 | 29.73 ± 6.70 | 30.24 ± 7.26 | 30.28 ± 7.16 | 0.053 |
| Waist circumference (cm) | | 99.09 ± 16.03 | 100.60 ± 16.44 | 102.24 ± 18.33 | 102.51 ± 18.38 | 0.012 |
| Obesity status | | | | | | 0.026 |
| No obesity | 23.78 | 24.50 | 26.59 | 20.34 | 23.41 | |
| Central obesity | 32.08 | 36.89 | 29.28 | 33.79 | 27.39 | |
| Peripheral obesity | 0.04 | 0 | 0 | 0.17 | 0 | |
| Mixed obesity | 44.09 | 38.61 | 44.14 | 45.70 | 49.19 | |
## Covariable selection
As shown in Table 2, we select covariates by univariate analysis. When the outcome index is BMI, the age, race/ethnicity, education level, ratio of family income to poverty, total cholesterol, triglyceride, glycohemoglobin, and serum uric acid were select as covariable. When the outcome index is WC, the age, gender, race/ethnicity, education level, blood urea nitrogen, serum creatinine, total cholesterol, triglyceride, glycohemoglobin, serum uric acid, and total protein were select as covariable.
**Table 2**
| Urinary nickel | Body mass index (kg/m2) β (95% CI), P | Waist circumference (cm) β (95% CI), P |
| --- | --- | --- |
| Gender | Gender | Gender |
| Male | Reference | Reference |
| Female | 0.43 (−0.23, 1.09) | −4.13 (−5.76, −2.50)*** |
| Age | 0.03 (0.01, 0.05)** | 0.21 (0.16, 0.25)*** |
| Race/ethnicity | Race/ethnicity | Race/ethnicity |
| Mexican American | Reference | Reference |
| Other Hispanic | −1.39 (−3.08, 0.31) | −3.37 (-7.58, 0.83) |
| Non-Hispanic White | −0.95 (−2.11, 0.21) | 0.57 (−2.31, 3.46) |
| Non-Hispanic Black | 0.10 (−1.36, 1.56) | −0.90 (−4.52, 2.73) |
| Other race | −3.08 (−4.56, −1.60)*** | −6.71 (-10.39, −3.03)*** |
| Education leve | Education leve | Education leve |
| Less than high school | Reference | Reference |
| High school | 1.94 (0.77, 3.10)** | 4.20 (1.30, 7.10)** |
| More than high school | 1.09 (0.02, 2.16)* | 1.55 (−1.11, 4.22) |
| Ratio of family income to poverty | −0.21 (−0.42, −0.00)* | −0.08 (−0.60, 0.44) |
| Total cholesterol | 0.45 (0.12, 0.78)** | 1.48 (0.66, 2.30)*** |
| Triglyceride | 0.96 (0.71, 1.21)*** | 3.00 (2.39, 3.60)*** |
| Glycohemoglobin | 1.97 (1.62, 2.33)*** | 5.96 (5.09, 6.82)*** |
| Blood urea nitrogen | 0.16 (−0.02, 0.33) | 0.91 (0.47, 1.35)*** |
| Serum creatinine | 0.00 (−0.01, 0.02) | 0.06 (0.03, 0.10)*** |
| Serum uric acid | 0.02 (0.02, 0.03)*** | 0.07 (0.06, 0.08)*** |
| Total protein (g/L) | −0.04 (−0.12, 0.04) | −0.23 (−0.43, −0.04)* |
## Association between urinary nickel and BMI
Table 3 shows the association between urinary nickel and BMI based on multivariate regression analysis. In all three models, no significant associations were found. However, stratified by sex, all three models (model 1: 0.3520, 0.0604–0.6436; model 2: 0.3278, 0.0398–0.6159; model 3: 0.2965, 0.0302–0.5628) revealed a positive association for males, P for trend of three models was, respectively, 0.001, 0.003, and 0.010. As a result of secondary stratification based on sex and race, urinary nickel had a positive correlation with BMI in White males (Table 4).
## Association between urinary nickel and WC
When exploring the association between urinary nickel and WC, we found a positive association in all their models(model 1:0.4894,0.0486–0.9302; model 2:0.4938,0.0679–0.9197; model 3: 0.4110 0.0221–0.7999). However, stratified by sex, the positive association was only found in three male models (model 1: 1.3408, 0.5525–2.1290; model 2: 1.2004, 0.4511–1.9498; model 3:1.1111, 0.4156–1.8066), with a significant P for trend of three models ($P \leq 0.001$, $P \leq 0.001$, $$P \leq 0.001$$) (Table 5). Secondary stratification analysis according to sex and race, urinary nickel has a positive correlation with WC in both White and Black males (Table 6).
## Association among BMI, WC, and urinary nickel stratified simultaneously by gender and obesity status
As shown in Table 7, when stratified simultaneously according to gender and obesity status, BMI was positively correlated with no obesity in women (0.2161, 0.0325–0.3998), while WC was positively correlated with mixed obesity in Men(1.2598, 0.4131–2.1065).
**Table 7**
| Body mass index (kg/m2) | Male | Female |
| --- | --- | --- |
| Stratified by obsity status | Stratified by obsity status | Stratified by obsity status |
| No obesity | −0.0107 (−0.2064, 0.1850) | 0.2161 (0.0325, 0.3998)* |
| Central obesity | −0.1511 (−0.3176, 0.0154) | 0.0204 (−0.1501, 0.1909) |
| Peripheral obesity | - | - |
| Mixed obesity | 0.2849 (−0.0328, 0.6026) | −0.1046 (−0.3184, 0.1092) |
| Waist circumference (cm) | Waist circumference (cm) | Waist circumference (cm) |
| Stratified by obesity status (%) | Stratified by obesity status (%) | Stratified by obesity status (%) |
| No obesity | −0.0453 (−0.4763, 0.3857) | 0.1861 (−0.2251, 0.5973) |
| Central obesity | 0.3738 (−0.1659, 0.9134) | −0.0108 (−0.5076, 0.4861) |
| Peripheral obesity | - | - |
| Mixed obesity | 1.2598 (0.4131, 2.1065)** | −0.0434 (−0.4833, 0.3965) |
## Discussion
In this study, urinary nickel was evaluated concerning obesity status in the general population. Our results prove that urinary nickel positively correlates with BMI and WC among adult males but not females. In previous studies, heavy metal pollution is a significant cause of chronic inflammation and oxidative stress, of which nickel occupies a large part [31]. Chronic inflammation and oxidative stress can destroy the normal function of cells by interacting. The effects lead to symptoms such as weight gain or loss, decreased libido, physical pain, and emotional disorder, which pose a significant threat to health and lead to chronic inflammatory diseases, including obesity, diabetes, and cancer [32].
Several studies support our findings. Pokorska-Niewiada et al. showed that trace element disturbances, including nickel, can increase body mass index and contribute to endocrine disorders [33]. A study from Spain found that the trace element nickel in fat is the highest, highlighting the potential role of nickel in obesity and obesity-related diseases [34]. Another study from Turkey directly shows a positive correlation between BMI and nickel [35]. The results of Yang et al. proved that men exposed to nickel were more prone to dyslipidemia and BMI ≥ 25 [36]. In addition, when Cortés et al. studied the relationship between heavy metal exposure and chronic disease development in Chile, introducing BMI as a variable would confuse the relationship between IL-6 and nickel and increase the impact on individual inflammatory states by $40\%$ at the same time. This study indirectly proves that nickel levels in the urine will affect BMI [37], it indirectly proves that nickel levels in the urine will affect BMI.
When subgroup analyzes were performed, we found that urinary nickel was independently and positively associated with BMI and WC in adult men. Numerous prior research had shown that nickel exposure damages male reproductive organs, which is strongly connected to oxidative stress, DNA damage, and hormonal imbalance (38–40). One study found gender differences in the inflammatory response of mice to the lung after nickel exposure, with the male being more susceptible to acute pneumonia and subchronic lung inflammation than females by a mechanism that induces increased neutrophil by CXCL1 and IL-6/STAT3 signaling pathways and enhanced monocyte infiltration by CXCL1 and CCL2 in male [41]. At present, we have not found any other strong evidence for the reason for gender difference related to this study, and we suspect that the reason for the difference may be related to the differences in hormone levels, eating habits, and work stress between men and women. Large-sample prospective studies may be needed to explore this problem.
The precise mechanism of nickel exposure in BMI and WC is still unclear, but we try to clarify it from the following aspects. Firstly, in the hypothalamus, nickel exposure harms neurological function. As a result, hypothalamic neurons degenerate, paraventricular and supraoptic nuclei are reduced, and myeloperoxidase activity, nitric oxide increase, tumor necrosis factor-α and interleukin-1β of factors that promote inflammation ascend, which will affect the endocrine axis and might lead to hormonal imbalances [42, 43]; Secondly, there is the possibility that nickel can affect the hypothalamic-pituitary-thyroid axis, causing abnormal thyroid activity [44]. Finally, nickel disrupts the function of insulin β cells, resulting in abnormal glucose and lipid metabolism and affecting body weight [45, 46]. Nickel exposure has also been linked to diabetes in some studies [20, 47].
Heavy metal contamination is everywhere—vegetables, seafood, meat and poultry, water sources, and household products are all at risk of exceeding heavy metal levels. Long-term nickel exposure causes irreparable harm to human system functioning, yet using nickel-related items in the medical, commercial, and industrial sectors continues to grow fast. The national legislature should reinforce and enhance the pertinent laws and regulations to minimize heavy metal contamination. Our study demonstrated a significant association between nickel exposure and BMI and WC in males, and men with long-term nickel exposure must pay particular attention to this health risk. In addition, the mechanism through which nickel exposure lowers male sperm quality is conclusive, and men with reproductive needs should avoid nickel-related industries. We appeal to the public to reduce exposure to heavy metals, especially nickel.
As a result of the large sample size, valid subgroup analyses were possible. However, some limitations need attention. In terms of screening for overweight and obesity, BMI and WC are highly specific, but they are less sensitive when used to identify adiposity due to their inability to discern fat distribution accurately; higher visceral fat is far more harmful than more fat in areas such as the thighs, and therefore may incorrectly classify a person as unhealthy or at a high-risk category for disease [48]. Likewise, a higher BMI may also be induced by increased muscle mass, which may not always indicate obesity [49]. Additionally, these two indicators do not account for a multiplicity of characteristics like gender and age. It is well-known that men and women have varying quantities of muscle, which might alter the final indicator findings. Individuals with a high percentage of body fat may create more angiotensin and aldosterone, while muscle does not [50].
## Conclusions
In adult males, both BMI and WC were positively associated with urinary nickel. It is essential for adult men, especially those who are already obese, to reduce their nickel exposure. With the continued growth of nickel applications, nickel-related research will be expanded in the future, and our study may give suggestions for future studies in some specific aspects. Meanwhile, there is a need for further research to understand how urinary nickel might influence BMI and WC.
## Data availability statement
The original contributions presented in the study are publicly available. This data can be found here: www.cdc.gov/nchs/nhanes/.
## Author contributions
Conceptualization: H-LL, D-LL, and S-FC. Methodology and Writing—review and editing: G-XW. Software: B-LH. Formal analysis: G-XW and B-LH. Writing—original draft preparation: B-LH. Visualization: J-TL and Z-BF. Supervision: L-YF, H-XZ, H-LL, D-LL, and S-FC. Funding acquisition: S-FC. 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. Yap CK, Al-Mutairi KA. **Comparative study of potentially toxic nickel and their potential human health risks in seafood (fish and mollusks) from peninsular Malaysia**. *Biology.* (2022) **11** 376. DOI: 10.3390/biology11030376
2. Qiao S, Sun Y, Jiang Y, Chen X, Cai J, Liu Q. **Melatonin relieves liver fibrosis induced by Txnrd3 knockdown and nickel exposure via IRE1/NF-kB/NLRP3 and PERK/TGF-β1 axis activation**. *Life Sci.* (2022) **301** 120622. DOI: 10.1016/j.lfs.2022.120622
3. Yokoi K, Uthus EO, Penland JG, Nielsen FH. **Effect of dietary nickel deprivation on vision, olfaction, and taste in rats**. *J Trace Elem Med Biol.* (2014) **28** 436-40. DOI: 10.1016/j.jtemb.2014.07.014
4. Pieczyńska J, Płaczkowska S, Sozański R, Skórska K, Sołtysik M. **Effect of nickel on red blood cell parameters and on serum vitamin B12, folate and homocysteine concentrations during pregnancy with and without anemia**. *J Trace Elem Med Biol.* (2021) **68** 126839. DOI: 10.1016/j.jtemb.2021.126839
5. Alfano M, Cavazza C. **Structure, function, and biosynthesis of nickel-dependent enzymes**. *Protein Sci.* (2020) **29** 1071-89. DOI: 10.1002/pro.3836
6. Genchi G, Carocci A, Lauria G, Sinicropi MS, Catalano A. **Nickel: human health and environmental toxicology**. *IJERPH.* (2020) **17** 679. DOI: 10.3390/ijerph17030679
7. Li H, Wan Y, Chen X, Cheng L, Yang X, Xia W. **multiregional survey of nickel in outdoor air particulate matter in China: Implication for human exposure**. *Chemosphere.* (2018) **199** 702-8. DOI: 10.1016/j.chemosphere.2018.01.114
8. Cubadda F, Iacoponi F, Ferraris F, D'Amato M, Aureli F, Raggi A. **Dietary exposure of the Italian population to nickel: The national Total Diet Study**. *Food Chem Toxicol.* (2020) **146** 111813. DOI: 10.1016/j.fct.2020.111813
9. Ahlström MG, Thyssen JP, Wennervaldt M, Menné T, Johansen JD. **Nickel allergy and allergic contact dermatitis: A clinical review of immunology, epidemiology, exposure, and treatment**. *Contact Dermatitis.* (2019) **81** 227-41. DOI: 10.1111/cod.13327
10. Pavesi T, Moreira JC. **A comprehensive study of nickel levels in everyday items in Brazil**. *Contact Dermatitis.* (2020) **83** 88-93. DOI: 10.1111/cod.13534
11. Saylor DM, Craven BA, Chandrasekar V, Simon DD, Brown RP, Sussman EM. **Predicting patient exposure to nickel released from cardiovascular devices using multi-scale modeling**. *Acta Biomater.* (2018) **70** 304-14. DOI: 10.1016/j.actbio.2018.01.024
12. Shi Q, Wang Y, Hao Q, Vandvik PO, Guyatt G, Li J. **Pharmacotherapy for adults with overweight and obesity: a systematic review and network meta-analysis of randomised controlled trials**. *Lancet.* (2022) **399** 259-69. DOI: 10.1016/S0140-6736(21)01640-8
13. Sun X, Yan AF, Shi Z, Zhao B, Yan N, Li K. **Health consequences of obesity and projected future obesity health burden in China**. *Obesity (Silver Spring).* (2022) **30** 1724-51. DOI: 10.1002/oby.23472
14. Nunan E, Wright CL, Semola OA, Subramanian M, Balasubramanian P, Lovern PC. **Obesity as a premature aging phenotype—implications for sarcopenic obesity**. *GeroScience.* (2022) **44** 1393-405. DOI: 10.1007/s11357-022-00567-7
15. Foti K, Hardy ST, Chang AR, Selvin E, Coresh J, Muntner P. **and blood pressure control among United States adults with hypertension**. *J Hypertens.* (2022) **40** 741-8. DOI: 10.1097/HJH.0000000000003072
16. Gallagher EJ, LeRoith D. **Obesity and cancer**. *Cancer Metastasis Rev* (2022) **41** 463-4. DOI: 10.1007/s10555-022-10049-z
17. Ferrara-Cook C, Geyer SM, Evans-Molina C, Libman IM, Becker DJ, Gitelman SE. **the Type 1 Diabetes TrialNet Study Group. Excess BMI accelerates islet autoimmunity in older children and adolescents**. *Diabetes Care.* (2020) **43** 580-7. DOI: 10.2337/dc19-1167
18. Stefan N, Birkenfeld AL, Schulze MB. **Global pandemics interconnected — obesity, impaired metabolic health and COVID-19**. *Nat Rev Endocrinol.* (2021) **17** 135-49. DOI: 10.1038/s41574-020-00462-1
19. Cai Z, Yang Y, Zhang J. **Obesity is associated with severe disease and mortality in patients with coronavirus disease 2019 (COVID-19): a meta-analysis**. *BMC Public Health.* (2021) **21** 1505. DOI: 10.1186/s12889-021-11546-6
20. Shan S, Wang K, Hu C, Dai L. **Urinary Nickel Was Associated with the Prevalence of Diabetes: Results from NHANES**. *Biol Trace Elem Res* (2022) **201** 611-6. DOI: 10.1007/s12011-022-03190-x
21. Liu Y, Wu M, Xu B, Kang L. **Association between the urinary nickel and the diastolic blood pressure in general population**. *Chemosphere.* (2022) **286** 131900. DOI: 10.1016/j.chemosphere.2021.131900
22. Rohm TV, Meier DT, Olefsky JM, Donath MY. **Inflammation in obesity, diabetes, and related disorders**. *Immunity.* (2022) **55** 31-55. DOI: 10.1016/j.immuni.2021.12.013
23. Ng CD, Elliott MR, Riosmena F, Cunningham SA. **Beyond recent BMI: BMI exposure metrics and their relationship to health**. *SSM Popul Health.* (2020) **11** 100547. DOI: 10.1016/j.ssmph.2020.100547
24. Moltrer M, Pala L, Cosentino C, Mannucci E, Rotella CM, Cresci B. **Body mass index (BMI), waist circumference (WC), waist-to-height ratio (WHtR) e waist body mass index (wBMI): Which is better?**. *Endocrine.* (2022) **76** 578-83. DOI: 10.1007/s12020-022-03030-x
25. Liu B, Du Y, Wu Y, Snetselaar LG, Wallace RB, Bao W. **Trends in obesity and adiposity measures by race or ethnicity among adults in the United States 2011-18: population based study**. *BMJ* (2021) **2021** n365. DOI: 10.1136/bmj.n365
26. Ross R, Neeland IJ, Yamashita S, Shai I, Seidell J, Magni P., Cuevas A, Hu FB. **Waist circumference as a vital sign in clinical practice: a Consensus Statement from the IAS and ICCR Working Group on Visceral Obesity**. *Nat Rev Endocrinol.* (2020) **16** 177-89. DOI: 10.1038/s41574-019-0310-7
27. 27.National Center for Health Statistics. UTAS-J-UCM-J-UNI-J-MET-508. (2018). Available online at: https://wwwn.cdc.gov/nchs/data/nhanes/2017-2018/labmethods/UTAS-J-UCM-J-UNI-J-MET-508.pdf (accessed September 15, 2022).. *UTAS-J-UCM-J-UNI-J-MET-508* (2018)
28. Owolabi EO, Ter Goon D, Adeniyi OV. **Central obesity and normal-weight central obesity among adults attending healthcare facilities in Buffalo City Metropolitan Municipality, South Africa: a cross-sectional study**. *J Health Popul Nutr.* (2017) **36** 54. DOI: 10.1186/s41043-017-0133-x
29. 29.World Health Organization. Waist circumference and waist-hip ratio: report of a WHO expert consultation, Geneva. (2011) (accessed on December 8–11, 2008).. *Waist circumference and waist-hip ratio: report of a WHO expert consultation, Geneva* (2011)
30. 30.National Center for Health Statistics (U.S.) ed. National health and nutrition examination survey: analytic guidelines, 1999-2010. Hyattsville, Maryland: U.S. Department of Health and HumanServices, Centers for Disease Control and Prevention, National Center for Health Statistics (2013). 16 p.. *National health and nutrition examination survey: analytic guidelines, 1999-2010* (2013) 16
31. Nassan FL, Wang C, Kelly RS, Lasky-Su JA, Vokonas PS, Koutrakis P. **Schwartz JD. Ambient PM25 species and ultrafine particle exposure and their differential metabolomic signatures**. *Environ Int.* (2021) **151** 106447. DOI: 10.1016/j.envint.2021.106447
32. Furman D, Campisi J, Verdin E, Carrera-Bastos P, Targ S, Franceschi C. **Chronic inflammation in the etiology of disease across the life span**. *Nat Med.* (2019) **25** 1822-32. DOI: 10.1038/s41591-019-0675-0
33. Pokorska-Niewiada K, Brodowska A, Brodowski J, Szczuko M. **Levels of trace elements in erythrocytes as endocrine disruptors in obese and nonobese women with polycystic ovary syndrome**. *IJERPH.* (2022) **19** 976. DOI: 10.3390/ijerph19020976
34. Freire C, Vrhovnik P, Fiket Ž, Salcedo-Bellido I, Echeverría R, Martín-Olmedo P. **Adipose tissue concentrations of arsenic, nickel, lead, tin, and titanium in adults from GraMo cohort in Southern Spain: An exploratory study**. *Sci Total Environ.* (2020) **719** 137458. DOI: 10.1016/j.scitotenv.2020.137458
35. Cetin I, Nalbantcilar MT, Tosun K, Nazik A. **How trace element levels of public drinking water affect body composition in Turkey**. *Biol Trace Elem Res.* (2017) **175** 263-70. DOI: 10.1007/s12011-016-0779-z
36. Yang AM, Bai YN, Pu HQ, Zheng TZ, Cheng N, Li JS Li HY. **Prevalence of metabolic syndrome in Chinese nickel-exposed workers**. *Biomed Environ Sci.* (2014) **27** 475-7. DOI: 10.3967/bes2014.077
37. Singh M, Verma Y, Rana SVS. **Attributes of oxidative stress in the reproductive toxicity of nickel oxide nanoparticles in male rats**. *Environ Sci Pollut Res.* (2022) **29** 5703-17. DOI: 10.1007/s11356-021-15657-w
38. Sun H, Wu W, Guo J, Xiao R, Jiang F, Zheng L. **Effects of nickel exposure on testicular function, oxidative stress, and male reproductive dysfunction in Spodoptera litura Fabricius**. *Chemosphere.* (2016) **148** 178-87. DOI: 10.1016/j.chemosphere.2015.10.068
39. Rizvi A, Parveen S, Khan S, Naseem I. **Nickel toxicology with reference to male molecular reproductive physiology**. *Reprod Biol.* (2020) **20** 3-8. DOI: 10.1016/j.repbio.2019.11.005
40. You DJ, Lee HY, Taylor-Just AJ, Linder KE, Bonner JC. **Sex differences in the acute and subchronic lung inflammatory responses of mice to nickel nanoparticles**. *Nanotoxicology.* (2020) **14** 1058-81. DOI: 10.1080/17435390.2020.1808105
41. Cortés S, Zúñiga-Venegas L, Pancetti F, Covarrubias A, Ramírez-Santana M, Adaros H. **Positive relationship between exposure to heavy metals and development of chronic diseases: a case study from Chile**. *IJERPH.* (2021) **18** 1419. DOI: 10.3390/ijerph18041419
42. Adedara IA, Abiola MA, Adegbosin AN, Odunewu AA. **Farombi EO. Impact of binary waterborne mixtures of nickel and zinc on hypothalamic-pituitary-testicular axis in rats**. *Chemosphere.* (2019) **237** 124501. DOI: 10.1016/j.chemosphere.2019.124501
43. Risi R, Masieri S, Poggiogalle E, Watanabe M, Caputi A, Tozzi R. **Nickel Sensitivity Is Associated with GH-IGF1 Axis Impairment and Pituitary Abnormalities on MRI in Overweight and Obese Subjects**. *IJMS.* (2020) **21** 9733. DOI: 10.3390/ijms21249733
44. Yang J, Ma Z. **Research progress on the effects of nickel on hormone secretion in the endocrine axis and on target organs**. *Ecotoxicol Environ Saf.* (2021) **213** 112034. DOI: 10.1016/j.ecoenv.2021.112034
45. Chen YW, Yang CY, Huang CF, Hung DZ, Leung YM, Liu SH. **Heavy metals, islet function and diabetes development**. *Islets.* (2009) **1** 169-76. DOI: 10.4161/isl.1.3.9262
46. Xu X, Rao X, Wang T-Y, Jiang SY, Ying Z, Liu C, Maiseyeu A. **Effect of co-exposure to nickel and particulate matter on insulin resistance and mitochondrial dysfunction in a mouse model**. *Part Fibre Toxicol.* (2012) **9** 40. DOI: 10.1186/1743-8977-9-40
47. Titcomb TJ, Liu B, Lehmler H, Snetselaar LG, Bao W. **Environmental nickel exposure and diabetes in a nationally representative sample of US adults**. *Expo Health.* (2021) **13** 697-704. DOI: 10.1007/s12403-021-00413-9
48. Pyo JY, Ahn SS, Lee LE, Song JJ, Park Y, Lee S. **New body mass index for predicting prognosis in patients with antineutrophil cytoplasmic antibody-associated vasculitis**. *Clin Labor Anal* (2022) **36** e24357. DOI: 10.1002/jcla.24357
49. Anyzewska A, Lakomy R, Lepionka T, Szarska E, Maculewicz E, Bolczyk I. **Tomczak A, Bertrandt J. Nutritional status assessment of the Polish Border Guards officers—Body Mass Index or Fat Mass Index?**. *Proc Nutr Soc.* (2020) **79** E385. DOI: 10.1017/S002966512000333X
50. Wang W-J, Wang C-S, Wang C-K, Yang A-M. **Lin C-Y. Urine Di-(2-ethylhexyl) Phthalate metabolites are independently related to body fluid status in adults: results from a US Nationally representative survey**. *Int J Environ Res Public Health.* (2022) **19** 6964. DOI: 10.3390/ijerph19126964
|
---
title: 'Trends and disparities in sleep quality and duration in older adults in China
from 2008 to 2018: A national observational study'
authors:
- Zihao Tao
- Yuting Feng
- Jue Liu
- Liyuan Tao
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9982158
doi: 10.3389/fpubh.2023.998699
license: CC BY 4.0
---
# Trends and disparities in sleep quality and duration in older adults in China from 2008 to 2018: A national observational study
## Abstract
### Background
Poor sleep status as a common concern is a risk factor for many health problems among older people. China with an aging society lacks relevant nationwide data on the sleep status among older people. Therefore, the purpose of this study was to investigate trends and disparities in sleep quality and duration among older adults, and exploring influencing factors of poor sleep in China between 2008 and 2018.
### Method
We used the four-waves data of the Chinese Longitudinal Healthy Longevity Survey (CLHLS) from 2008 to 2018. Sleep quality and average sleep hours per day was investigated by using questionnaires in the CLHLS. We categorized sleep duration as three groups including ≤5 h (short duration), 5–9 h (normal duration), or ≥9 h (long duration) per day. Multivariate logistic regression models were used to examine trends and risk factors of poor sleep quality, short sleep duration, and long sleep duration.
### Results
The prevalence of poor sleep quality significantly increased from $34.87\%$ in 2008 to $47.67\%$ in 2018 ($p \leq 0.05$). Short sleep duration significantly increased from 5.29 to $8.37\%$, whereas long sleep duration decreased from 28.77 to $19.27\%$. Multivariate analysis showed that female sex, poor economic status, a greater number of chronic diseases, underweight, poor self-reported quality of life, and poor self-reported health were associated with poor sleep quality and short sleep duration ($p \leq 0.05$).
### Conclusion
Our findings revealed that older adults had increased prevalence of poor sleep quality and short sleep duration from 2008 to 2018. More attention should be paid to the increased sleep problems among older adults and early interventions should be made to improve sleep quality and guarantee enough sleep time.
## Introduction
Sleep quality and duration are important health topics. Unfortunately, epidemiological studies have shown that sleep problems are very common among older people. Previous studies showed that the prevalence of sleep problems was 16.6, 28.9, and $31.2\%$ in Denmark, Japan and Poland, respectively [1, 2]. Some studies had reported increased trends of sleep problems because of the increasing physical and psychological issues in facing of the rapidly changing world (3–5). China as the country with the largest number of older people in the world, has the increasing trend of aging which means the burden on families and public health care [6]. According to the National Bureau of Statistics of China, there were 109.56 million people over the age of 65 years in China in 2008, accounting for $8.25\%$ of the total population [7]. By 2018, this number had increased to 167.24 million, accounting for $11.90\%$ [7]. Although several urban and regional studies reported that the high prevalence of poor sleep quality among older people ranged from 33.8 to $49.7\%$ (8–11), there was a lack of studies on long-term trends in sleep status in China.
Many studies had showed that sleep problems were associated with the increased risk of adverse outcomes among older people, thus sleep should be paid attention and surveilled. Evidence reported that poor sleep quality increased the risks of fall [2], physical disability [12], and hypertension among older people [13]. A U-shaped dose–response relationship existed in the relationship between sleep duration and other health problems, such as cognitive function decline [14], osteoporosis [15], type 2 diabetes [16], and coronary heart disease [17], even mortality [18]. Maintaining good sleep quality and normal sleep duration is very important for the health among older people. Therefore, considering the above information on the common sleep problems and its harm among older people, the identification of related risk factors is of great importance for public health and clinically to develop effective interventions of sleep problems.
Previous studies have shown that some demographic factors, socioeconomic status, lifestyle habits, and health conditions played crucial roles in sleep quality and duration among older people (14, 18–22). Understanding the characteristics of the demographic factors, socioeconomic status and health conditions was useful to identify the risk population. For example, previous studies reported that female [22], people with higher educational level [19] and increased number of chronic diseases [21] may be poor sleepers. Meanwhile, exploring the association between lifestyle habits and sleep was benefit for improving sleep by establishing good lifestyle habits or decreasing bad lifestyle habits [20]. However, compared with demographic factors, socioeconomic status and health conditions, the lifestyle habits, especially some detail activities, were investigated rarely among older people, such as housework, having pets or gardening, reading books, playing cards, watching television, and social participation.
Based on the above background on lacked national research evidence about long-term change trends of sleep and related influencing factors among the Chinese older people, the purpose of this study was to investigate trends and disparities in the quality of sleep and sleep duration among adults aged 65 years and older from 2008 to 2018 using the Chinese Longitudinal Healthy Longevity Survey (CLHLS) data.
## Study population and data source
This was a national observational study using the CLHLS data from 2008 and 2018. The CLHLS as the first national longitudinal survey of older people in a developing country, was launched in 1998, accounting for about $90\%$ of the country's population from a randomly selected half of the counties and cities in 23 of 31 provinces in China. A targeted random-sample design was adopted to ensure representativeness. All of the centenarians of the sampled counties and cities agreed voluntarily to participate in the study. This study was established in 1998, with subsequent follow-up and recruitment of new participants in 2000, 2002, 2005, 2008, 2011, 2014, and 2018. The collected data includes demographic characteristics, family and residential characteristics, marital status, living arrangements, social and economic characteristics, health, and other individual data for a large number of older individuals [23]. To account for deaths and people who lost follow-up, CLHLS enrolled new participants according to similar sex, age, and other characteristics of the missing persons to ensure consistency of the study. To ensure the quality of the survey, the project team had strictly and carefully trained investigators to conduct the household surveys to ensure the quality of the survey. All the surveys were face-to-face interviews conducted at the participant's home. Each participant provided a signed informed consent form; this was signed by the next of kin if the participant could not sign it. More details about the study design of the CLHLS can be found elsewhere [23].
There was a total of 50,870 participants in the surveys during 2008–2018 (16,954 in 2008, 10,850 in 2011, 7,192 in 2014, and 15,874 in 2018). We excluded 2,759 participants who had missing data on sleep quality and 740 participants aged below 65 years, yielding 47,371 participants ($93.12\%$) in the final study population.
## Assessment of sleep status
Sleep quality and sleep duration were assessed by using two questions: “*How is* your sleep quality now?” and “How many hours do you sleep on average now?”, respectively, which were both commonly used in previous studies [14, 24, 25]. The response options of sleep quality included five categories: excellent, good, average, not good, and very bad, we defined poor sleep quality as a response of average/bad/very bad based on the research from Gu et al., others were good sleep quality [25]. The 5 categories (excellent, good, average, not good, and very bad) were binary divided as poor sleep quality (encompassing average, not good, and very bad subcategories) and good sleep quality (encompassing excellent and good subcategories) [25]. We divided sleep duration for adults over age 65 years into three groups, according to the classification on sleep duration (<5 h as short sleep duration, 5–9 h as normal, and more than 9 h as long sleep duration) from National Sleep Foundation [26] and the research from Gu et al. [ 25]. In the sensitivity analysis, we further divided the participants into three groups (<7, 7–9, >9 h) for more categories.
## Assessment of related factors
Based on previous research [14, 24, 25], we included relevant influencing factors of sleep status, including demographic factors, socioeconomic status, lifestyle habits, and health conditions. Demographic factors included investigation year, sex (male or female), age group (65–79, 80–89, 90–99, and ≥100 years), marital status (unmarried, married, or divorced or widowed), and residence (urban or rural).
Socioeconomic status included economic status compared with other local people (wealthy, average, or poor), living arrangements (living with family members, living in an institution, or living alone), and years of schooling (0 or ≥1 year).
Lifestyle habits included smoking status (never, previous, or current), alcohol intake (never, previous, or current), regular exercise (never, previous, or current), dietary diversity score (poor, moderate, or good), housework (nearly every day, sometimes, or never), outdoor activities (nearly every day, sometimes, or never), having pets or gardening (nearly every day, sometimes, or never), reading books (nearly every day, sometimes, or never), raising poultry (nearly every day, sometimes, or never), playing cards (nearly every day, sometimes, or never), watching television (TV; nearly every day, sometimes, or never), and social participation (nearly every day, sometimes, or never).
Health conditions included body mass index (BMI; underweight, normal weight, overweight, or obesity), number of chronic diseases (0, 1, or ≥2), activities of daily living (ADL; independent or disabled), self-reported quality of life (good, average, or poor) and self-reported health (good, average, or poor), and cognitive impairment (yes, no). The height and weight of participants were measured by calibrated instruments, as described previously [27]. BMI was categorized as underweight (<18.5 kg/m2), normal weight (18.5–24.9 kg/m2), overweight (25–29.9 kg/m2), and obese (≥30 kg/m2), according to the cutoff values suggested by the World Health Organization. As for dietary diversity, consumption frequencies of nine food groups (meat, vegetables, fish, eggs, fruits, legumes, milk, tea, and nuts) were recorded, and the dietary diversity score (0–9) was calculated and categorized according to the recommendations by the Food and Agriculture Organization of the United Nations and previous research [28, 29]. ADL refers to basic personal care tasks of everyday life. In this study, ADL in disability was defined as self-reported difficulty with any of the following ADL items: dressing, eating, bathing, continence, toileting, cleaning, and indoor movement [24, 30]. In compliance with the previous studies [31, 32], cognitive function was measured by t cognitive assessment tool, the Chinese version of the Mini-Mental State Examination (MMSE) questionnaire, which consists of 11 questions covering orientation, registration, attention, calculation, recall, and language abilities with a total score of 30, which had shown good validity and reliability [33, 34]. CLHLS participants who scored <18 were classified as having cognitive impairment, whereas participants with a score of 18 or higher were classified as having no cognitive impairment [31, 32].
## Data analysis
Baseline characteristics were described as percentages for categorical variables and median (interquartile range [IQR]) for continuous variables. We tested statistical differences using the chi-square test for categorical variables and the t-test for normally distributed continuous variables according to sleep quality and sleep duration. We used multivariate logistic regression models to analyze risk factors related to poor sleep quality, short sleep duration, and long sleep duration by calculating the odds ratio (OR) with $95\%$ confidence interval (CI). In the sensitivity analysis, we further divided the participants into three groups (<7, 7–9, >9 h) to assess factors associated with sleep duration <7 h and more than 9 h. Moreover, we additionally excluded participates with cognitive impairment in the multivariate logistic regression models to examine the robust of the results in the sensitivity analysis. All the analyses were performed with IBM SPSS 26.0 (IBM Corp., Armonk, NY, USA). Two-sided p-values < 0.05 indicated statistical significance.
## Basic characteristics of the research population
A total of 47,460 participants in the CLHLS survey from 2008 to 2018 were included in this study; the average age of the participants was 73.4 ± 6.5 years. Participants were divided into four groups by age: $32.61\%$ were ≤79 years old, $27.17\%$ were age 80–89 years, $24.69\%$ were age 90–99 years, and $16.53\%$ ≥100. A total of $44.09\%$ of participants were men, and $55.91\%$ were women. We found significant differences among groups for age group, marital status, residence (urban or rural), economic status, living arrangements, education level, number of chronic diseases, smoking, drinking, regular exercise, dietary diversity score subgroups, housework, outdoor activities, having pets or gardening, reading books and newspapers, raising poultry, playing cards, watching TV, and other factors ($p \leq 0.001$ for all; Supplementary Table 1).
## Trends and disparities in sleep quality among older people in China
The prevalence of poor sleep quality among older people in China increased over time, from $34.87\%$ in 2008 to $47.67\%$ in 2018 ($p \leq 0.001$, Table 1). Poor sleep quality showed the same upward trend across sex and age groups (Figure 1A). We found a higher prevalence of poor sleep quality among participants with characteristics such as female sex, age group 80–89 years, divorced or widowed, poor economic status, living in an institution, no education, ≥2 chronic diseases, no regular exercise, poor dietary diversity score, does not engage in reading, does not play cards, does not watch TV, no social participation, underweight, poor quality of life, poor health, and cognitive impairment ($p \leq 0.001$, Table 1).
## Trends and disparities in sleep duration among older people in China
Our study investigated changes in the proportion of sleep duration in the older people from 2008 to 2018. The proportion of short sleep duration (<5 h) increased, from $5.29\%$ in 2008 to $8.37\%$ in 2018. On the contrary, the proportion of long sleep duration (>9 h) decreased, from $28.77\%$ in 2008 to $19.27\%$ in 2018 ($p \leq 0.001$; Table 2). The same trend was observed across both sexes and across age groups (Figures 1B, C).
**Table 2**
| Unnamed: 0 | Total | <5 h | 5-9 h | >9 h | Unnamed: 5 |
| --- | --- | --- | --- | --- | --- |
| Investigation year | | | | | <0.001 |
| 2008 | 16405.0 | 868 (5.29) | 10,818 (65.94) | 4,719 (28.77) | |
| 2011 | 9574.0 | 703 (7.34) | 6,340 (66.22) | 2,531 (26.44) | |
| 2014 | 6897.0 | 547 (7.93) | 4,857 (70.42) | 1,493 (21.65) | |
| 2018 | 14255.0 | 1,193 (8.37) | 10,315 (72.36) | 2,747 (19.27) | |
| Gender | | | | | <0.001 |
| Male | 20804.0 | 1,210 (5.82) | 14,689 (70.61) | 4,905 (23.58) | |
| Female | 26327.0 | 2,101 (7.98) | 17,641 (67.01) | 6,585 (25.01) | |
| Age group | | | | | <0.001 |
| ≤79 | 14926.0 | 1,009 (6.76) | 11,942 (80.01) | 1,975 (13.23) | |
| 80–89 | 12823.0 | 1,133 (8.84) | 8,911 (69.49) | 2,779 (21.67) | |
| 90–99 | 11621.0 | 764 (6.57) | 7,112 (61.20) | 3,745 (32.23) | |
| ≥100 | 7761.0 | 405 (5.22) | 4,365 (56.24) | 2,991 (38.54) | |
| Marital status | | | | | <0.001 |
| Unmarried | 431.0 | 40 (9.28) | 288 (66.82) | 103 (23.90) | |
| Married | 17605.0 | 1,194 (6.78) | 13,361 (75.89) | 3,050 (17.32) | |
| Divorced or widowed | 28796.0 | 2,054 (7.13) | 18,475 (64.16) | 8,267 (28.71) | |
| Category of residence of the interviewee at the 1998 survey | | | | | 0.009 |
| Urban (city and town) | 22193.0 | 1,514 (6.82) | 15,376 (69.28) | 5,303 (23.89) | |
| Rural | 24938.0 | 1,797 (7.21) | 16,954 (67.98) | 61,87 (24.81) | |
| Economic status | | | | | <0.001 |
| Rich | 7674.0 | 349 (4.55) | 5,390 (70.24) | 1,935 (25.22) | |
| General | 32349.0 | 2,111 (6.53) | 22,370 (69.15) | 7,868 (24.32) | |
| Poor | 6671.0 | 814 (12.20) | 4,294 (64.37) | 1,563 (23.43) | |
| Living pattern | | | | | <0.001 |
| Living with family members | 37866.0 | 2,445 (6.46) | 25,914 (68.44) | 9,507 (25.11) | |
| Living in a institution | 1169.0 | 117 (10.01) | 748 (63.99) | 304 (26.01) | |
| Living alone | 7738.0 | 721 (9.32) | 5,410 (69.91) | 1,607 (20.77) | |
| Years of schooling | | | | | <0.001 |
| 0 | 27830.0 | 2,139 (7.69) | 18,134 (65.16) | 7,557 (27.15) | |
| ≥1 year | 19301.0 | 1,172 (6.07) | 14,196 (73.55) | 3,933 (20.38) | |
| Numbers of chronic diseases | | | | | <0.001 |
| 0 | 17169.0 | 839 (4.89) | 11,676 (68.01) | 4,654 (27.11) | |
| 1 | 14341.0 | 1,015 (7.08) | 9,890 (68.96) | 3,436 (23.96) | |
| ≥2 | 14827.0 | 1,389 (9.37) | 10,211 (68.87) | 3,227 (21.76) | |
| Smoking status | | | | | <0.001 |
| Never | 31848.0 | 2,309 (7.25) | 21,827 (68.53) | 7,712 (24.22) | |
| Previous | 7090.0 | 477 (6.73) | 4,777 (67.38) | 1,836 (25.90) | |
| Current | 7817.0 | 501 (6.41) | 5,475 (70.04) | 1,841 (23.55) | |
| Alcohol intaking status | | | | | <0.001 |
| Never | 29906.0 | 2,190 (7.32) | 20,555 (68.73) | 7,161 (23.95) | |
| Previous | 4662.0 | 364 (7.81) | 3,010 (64.56) | 1,288 (27.63) | |
| Current | 7817.0 | 501 (6.41) | 5,475 (70.04) | 1,841 (23.55) | |
| Regular exercise | | | | | <0.001 |
| Never | 27829.0 | 1,971 (7.08) | 18,713 (67.24) | 7,145 (25.67) | |
| Previous | 4743.0 | 363 (7.65) | 2,981 (62.85) | 1,399 (29.50) | |
| Current | 13962.0 | 937 (6.71) | 10,215 (73.16) | 2,810 (20.13) | |
| Dietary diversity score | | | | | <0.001 |
| Poor | 14622.0 | 1,445 (9.88) | 9,702 (66.35) | 3,475 (23.77) | |
| Moderate | 23900.0 | 1,525 (6.38) | 16,294 (68.18) | 6,081 (25.44) | |
| Good | 8525.0 | 337 (3.95) | 6,276 (73.62) | 1,912 (22.43) | |
| Housework | | | | | <0.001 |
| Almost everyday | 19823.0 | 1,562 (7.88) | 14,927 (75.30) | 3,334 (16.82) | |
| Sometimes | 5854.0 | 322 (5.50) | 4,358 (74.44) | 1,174 (20.05) | |
| Never | 21314.0 | 1,421 (6.67) | 12,957 (60.79) | 6,936 (32.54) | |
| Outdoor activities | | | | | <0.001 |
| Almost everyday | 18721.0 | 1,231 (6.58) | 13,623 (72.77) | 3,867 (20.66) | |
| Sometimes | 14910.0 | 1,095 (7.34) | 10,544 (70.72) | 3,271 (21.94) | |
| Never | 13360.0 | 978 (7.32) | 8,077 (60.46) | 4,305 (32.22) | |
| Keeping pets or gardening | | | | | <0.001 |
| Almost everyday | 5297.0 | 320 (6.04) | 4,092 (77.25) | 885 (16.71) | |
| Sometimes | 2595.0 | 124 (4.78) | 2,013 (77.57) | 458 (17.65) | |
| Never | 39104.0 | 2,860 (7.31) | 26,140 (66.85) | 10,104 (25.84) | |
| Reading books | | | | | <0.001 |
| Almost everyday | 4930.0 | 248 (5.03) | 3,898 (79.07) | 784 (15.90) | |
| Sometimes | 4219.0 | 226 (5.36) | 3,247 (76.96) | 746 (17.68) | |
| Never | 37846.0 | 2,830 (7.48) | 25,099 (66.32) | 9,917 (26.20) | |
| Raising poultry | | | | | <0.001 |
| Almost everyday | 8018.0 | 567 (7.07) | 6,041 (75.34) | 1,410 (17.59) | |
| Sometimes | 2780.0 | 155 (5.58) | 2,083 (74.93) | 542 (19.50) | |
| Never | 36189.0 | 2,581 (7.13) | 24,114 (66.63) | 9,494 (26.23) | |
| Playing cards | | | | | <0.001 |
| Almost everyday | 2785.0 | 162 (5.82) | 2,120 (76.12) | 503 (18.06) | |
| Sometimes | 4634.0 | 246 (5.31) | 3,578 (77.21) | 810 (17.48) | |
| Never | 39576.0 | 2,897 (7.32) | 26,547 (67.08) | 10,132 (25.60) | |
| Watching TV | | | | | <0.001 |
| Almost everyday | 24360.0 | 1,567 (6.43) | 18,084 (74.24) | 4,709 (19.33) | |
| Sometimes | 8313.0 | 565 (6.80) | 5,684 (68.37) | 2,064 (24.83) | |
| Never | 14330.0 | 1,169 (8.16) | 8,489 (59.24) | 4,672 (32.60) | |
| Social participation | | | | | <0.001 |
| Almost everyday | 1290.0 | 78 (6.05) | 1,051 (81.47) | 161 (12.48) | |
| Sometimes | 4829.0 | 274 (5.67) | 3,866 (80.06) | 689 (14.27) | |
| Never | 40741.0 | 2,942 (7.22) | 27,231 (66.84) | 10,568 (25.94) | |
| BMI | | | | | <0.001 |
| Normal | 26658.0 | 1,789 (6.71) | 18,741 (70.30) | 6,128 (22.99) | |
| Underweight | 10860.0 | 816 (7.51) | 7,005 (64.50) | 3,039 (27.98) | |
| Overweight | 5916.0 | 387 (6.54) | 4,377 (73.99) | 1,152 (19.47) | |
| Obesity | 1257.0 | 96 (7.64) | 891 (70.88) | 270 (21.48) | |
| Activities of daily living | | | | | <0.001 |
| Independent | 35254.0 | 2,429 (6.89) | 25,536 (72.43) | 7,289 (20.68) | |
| Disabled | 10758.0 | 790 (7.34) | 6,027 (56.02) | 3,941 (36.63) | |
| Self-reported quality of life | | | | | <0.001 |
| Good | 28010.0 | 1,583 (5.65) | 19,472 (69.52) | 6,955 (24.83) | |
| General | 13478.0 | 1,183 (8.78) | 9,691 (71.90) | 2,604 (19.32) | |
| Poor | 2203.0 | 340 (15.43) | 1,432 (65.00) | 431 (19.56) | |
| Self-reported health | | | | | <0.001 |
| Good | 20510.0 | 895 (4.36) | 14,432 (70.37) | 5,183 (25.27) | |
| General | 16373.0 | 1,259 (7.69) | 11,833 (72.27) | 3,281 (20.04) | |
| Poor | 6829.0 | 963 (14.10) | 4,337 (63.51) | 1,529 (22.39) | |
| Cognitive impairment | | | | | <0.001 |
| No | 35562.0 | 2,487 (6.99) | 25,969 (73.02) | 7,106 (19.98) | |
| Yes | 11569.0 | 824 (7.12) | 6,361 (54.98) | 4,384 (37.89) | |
Participants with the following characteristics had a higher prevalence of poor sleep duration: female sex, age 80–89 years, rural residence, poor economic status, living in an institution, ≥2 chronic diseases, poor dietary diversity score, does not engage in reading, does not play cards, never watches TV, no social participation, underweight, poor quality of life, poor health, and cognitive impairment ($p \leq 0.001$, Table 2).
## Multivariate logistic regression of factors associated with sleep quality
Multivariate logistic regression analysis showed that, after controlling for other confounding factors, the risk of poor sleep quality increased each year compared with 2008: adjusted odds ratio (aOR) = 1.08 ($95\%$ CI: 1.02–1.15) in 2011, aOR = 1.28 ($95\%$ CI: 1.19–1.38) in 2014, and aOR = 1.94 ($95\%$ CI: 1.82–2.05) in 2018 (Table 3). Female (aOR = 1.39, $95\%$ CI: 1.30–1.48), poor economic status (aOR = 1.60, $95\%$ CI: 1.46–1.76), ≥2 chronic diseases (aOR = 1.52, $95\%$ CI: 1.43–1.61), underweight (aOR = 1.16, $95\%$ CI: 1.08–1.24), no social participation (aOR = 1.32, $95\%$ CI: 1.16–1.49), poor self-reported quality of life (aOR = 1.83, $95\%$ CI: 1.63–2.06), and poor self-reported health (aOR = 2.56, $95\%$ CI: 2.39–2.75) were related with poor sleep quality ($p \leq 0.05$). In the sensitivity analysis, the results were stable (Supplementary Table 2).
**Table 3**
| Unnamed: 0 | Beta | S.E. | p | OR | 95% CI |
| --- | --- | --- | --- | --- | --- |
| Investigation year | Investigation year | Investigation year | Investigation year | Investigation year | Investigation year |
| 2008 | | | | Ref. | |
| 2011 | 0.079 | 0.032 | 0.014 | 1.08 | 1.02–1.15 |
| 2014 | 0.248 | 0.038 | <0.001 | 1.28 | 1.19–1.38 |
| 2018 | 0.660 | 0.030 | <0.001 | 1.94 | 1.82–2.05 |
| Gender | Gender | Gender | Gender | Gender | Gender |
| Male | | | | Ref. | |
| Female | 0.328 | 0.032 | <0.001 | 1.39 | 1.30–1.48 |
| Marital status | Marital status | Marital status | Marital status | Marital status | Marital status |
| Unmarried | | | | Ref. | |
| Married | 0.182 | 0.108 | 0.092 | 1.20 | 0.97–1.48 |
| Divorced or widowed | 0.266 | 0.108 | 0.013 | 1.30 | 1.06–1.61 |
| Category of residence | Category of residence | Category of residence | Category of residence | Category of residence | Category of residence |
| Rural | | | | Ref. | |
| Urban (city and town) | −0.060 | 0.025 | 0.016 | 0.94 | 0.90–0.99 |
| Economic status | Economic status | Economic status | Economic status | Economic status | Economic status |
| Rich | | | | Ref. | |
| General | 0.177 | 0.034 | <0.001 | 1.19 | 1.12–1.28 |
| Poor | 0.471 | 0.048 | <0.001 | 1.60 | 1.46–1.76 |
| Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases |
| 0 | | | | Ref. | |
| 1 | 0.131 | 0.029 | <0.001 | 1.14 | 1.08–1.21 |
| ≥2 | 0.417 | 0.029 | <0.001 | 1.52 | 1.43–1.61 |
| Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status |
| Never | | | | Ref. | |
| Previous | −0.019 | 0.040 | 0.630 | 0.98 | 0.91–1.06 |
| Current | −0.173 | 0.034 | <0.001 | 0.84 | 0.79–0.90 |
| Regular exercise | Regular exercise | Regular exercise | Regular exercise | Regular exercise | Regular exercise |
| Never | | | | Ref. | |
| Previous | 0.191 | 0.045 | <0.001 | 1.21 | 1.11–1.32 |
| Current | 0.010 | 0.027 | 0.725 | 1.01 | 0.96–1.06 |
| Dietary diversity score | Dietary diversity score | Dietary diversity score | Dietary diversity score | Dietary diversity score | Dietary diversity score |
| Poor | | | | Ref. | |
| Moderate | −0.246 | 0.027 | <0.001 | 0.78 | 0.74–0.83 |
| Good | −0.192 | 0.036 | <0.001 | 0.83 | 0.77–0.88 |
| Housework | Housework | Housework | Housework | Housework | Housework |
| Almost everyday | | | | Ref. | |
| Sometimes | 0.110 | 0.035 | 0.002 | 1.12 | 1.04–1.20 |
| Never | −0.031 | 0.034 | 0.363 | 0.97 | 0.91–1.04 |
| Outdoor activities | Outdoor activities | Outdoor activities | Outdoor activities | Outdoor activities | Outdoor activities |
| Almost everyday | | | | Ref. | |
| Sometimes | 0.185 | 0.027 | <0.001 | 1.20 | 1.14–1.27 |
| Never | 0.004 | 0.035 | 0.913 | 1.00 | 0.94–1.08 |
| Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening |
| Almost everyday | | | | Ref. | |
| Sometimes | 0.022 | 0.049 | 0.662 | 1.02 | 0.93–1.13 |
| Never | 0.115 | 0.032 | <0.001 | 1.12 | 1.05–1.20 |
| Reading books | Reading books | Reading books | Reading books | Reading books | Reading books |
| Almost everyday | | | | Ref. | |
| Sometimes | −0.208 | 0.044 | <0.001 | 0.81 | 0.74–0.89 |
| Never | −0.075 | 0.039 | 0.054 | 0.93 | 0.86–1.00 |
| Raising poultry | Raising poultry | Raising poultry | Raising poultry | Raising poultry | Raising poultry |
| Almost everyday | | | | Ref. | |
| Sometimes | −0.029 | 0.048 | 0.550 | 0.97 | 0.88–1.07 |
| Never | −0.095 | 0.028 | 0.001 | 0.91 | 0.86–0.96 |
| Watching TV | Watching TV | Watching TV | Watching TV | Watching TV | Watching TV |
| Almost everyday | | | | Ref. | |
| Sometimes | 0.131 | 0.032 | <0.001 | 1.14 | 1.07–1.21 |
| Never | 0.196 | 0.037 | <0.001 | 1.22 | 1.13–1.31 |
| Social participation | Social participation | Social participation | Social participation | Social participation | Social participation |
| Almost everyday | | | | Ref. | |
| Sometimes | 0.354 | 0.067 | <0.001 | 1.43 | 1.25–1.62 |
| Never | 0.274 | 0.064 | <0.001 | 1.32 | 1.16–1.49 |
| BMI | BMI | BMI | BMI | BMI | BMI |
| Normal | | | | Ref. | |
| Underweight | 0.146 | 0.033 | <0.001 | 1.16 | 1.08–1.24 |
| Overweight | −0.206 | 0.030 | <0.001 | 0.81 | 0.77–0.86 |
| Obesity | −0.242 | 0.061 | <0.001 | 0.79 | 0.70–0.89 |
| Activities of daily living | Activities of daily living | Activities of daily living | Activities of daily living | Activities of daily living | Activities of daily living |
| Independent | | | | Ref. | |
| Disabled | −0.110 | 0.050 | 0.027 | 0.90 | 0.81–0.99 |
| Self–reported quality of life | Self–reported quality of life | Self–reported quality of life | Self–reported quality of life | Self–reported quality of life | Self–reported quality of life |
| Good | | | | Ref. | |
| General | 0.446 | 0.026 | <0.001 | 1.56 | 1.48–1.64 |
| Poor | 0.604 | 0.060 | <0.001 | 1.83 | 1.63–2.06 |
| Self–reported health | Self–reported health | Self–reported health | Self–reported health | Self–reported health | Self–reported health |
| Good | | | | Ref. | |
| General | 0.800 | 0.026 | <0.001 | 2.23 | 2.11–2.34 |
| Poor | 0.942 | 0.037 | <0.001 | 2.56 | 2.39–2.75 |
## Multivariate logistic regression of factors associated with sleep duration
Multivariate logistic regression analysis showed that, after controlling for other confounding factors, the risk of short sleep duration increased each year compared with 2008 (Table 4): aOR = 1.25 ($95\%$ CI: 1.11–1.42) in 2011, aOR = 1.52 ($95\%$ CI: 1.32–1.76) in 2014, and aOR = 1.69 ($95\%$ CI: 1.50–1.89) in 2018.
**Table 4**
| Unnamed: 0 | Risk factors of short sleep duration | Risk factors of short sleep duration.1 | Risk factors of short sleep duration.2 | Risk factors of long sleep duration | Risk factors of long sleep duration.1 | Risk factors of long sleep duration.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | OR | p | 95% CI | OR | p | 95% CI |
| Investigation year | Investigation year | Investigation year | Investigation year | Investigation year | Investigation year | Investigation year |
| 2008 | Ref. | | | Ref. | | |
| 2011 | 1.25 | <0.001 | 1.11–1.42 | 0.98 | 0.635 | 0.91–1.06 |
| 2014 | 1.52 | <0.001 | 1.32–1.76 | 0.92 | 0.096 | 0.84–1.01 |
| 2018 | 1.69 | <0.001 | 1.50–1.89 | 0.65 | <0.001 | 0.60–0.71 |
| Gender | Gender | Gender | Gender | Gender | Gender | Gender |
| Male | Ref. | | | Ref. | | |
| Female | 1.44 | <0.001 | 1.28–1.62 | 0.77 | <0.001 | 0.71–0.84 |
| Age group | Age group | Age group | Age group | Age group | Age group | Age group |
| ≤79 | Ref. | | | Ref. | | |
| 80–89 | 1.36 | <0.001 | 1.22–1.52 | 1.42 | <0.001 | 1.31–1.54 |
| 90–99 | 1.18 | 0.329 | 0.85–1.64 | 1.90 | <0.001 | 1.57–2.31 |
| ≥100 | 0.75 | 0.838 | 0.05–11.73 | 1.94 | 0.276 | 0.59–6.43 |
| BMI | BMI | BMI | BMI | BMI | BMI | BMI |
| Normal | Ref. | | | Ref. | | |
| Underweight | 1.18 | 0.005 | 1.05–1.33 | 0.96 | 0.343 | 0.88–1.05 |
| Overweight | 0.99 | 0.906 | 0.89–1.11 | 1.10 | 0.015 | 1.02–1.19 |
| Obesity | 0.85 | 0.180 | 0.67–1.08 | 1.14 | 0.114 | 0.97–1.34 |
| Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status |
| Unmarried | | | | Ref. | | |
| Married | | | | 0.41 | <0.001 | 0.33–0.51 |
| Divorced or widowed | | | | 0.44 | <0.001 | 0.35–0.55 |
| Economic status | Economic status | Economic status | Economic status | Economic status | Economic status | Economic status |
| Rich | Ref. | <0.001 | | | | |
| General | 1.18 | 0.032 | 1.01–1.37 | | | |
| Poor | 1.91 | <0.001 | 1.60–2.28 | | | |
| Living pattern | Living pattern | Living pattern | Living pattern | Living pattern | Living pattern | Living pattern |
| Living with family members | Ref. | | | Ref. | | |
| Living in a institution | 1.74 | 0.001 | 1.24–2.45 | 0.92 | 0.545 | 0.70–1.21 |
| Living alone | 1.06 | 0.303 | 0.95–1.18 | 0.88 | 0.009 | 0.80–0.97 |
| Years of schooling | Years of schooling | Years of schooling | Years of schooling | Years of schooling | Years of schooling | Years of schooling |
| 0 | Ref. | | | Ref. | | |
| ≥1 year | 0.77 | <0.001 | 0.69–0.84 | 0.83 | <0.001 | 0.77–0.88 |
| Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases | Numbers of chronic diseases |
| 0 | Ref. | | | | | |
| 1 | 1.34 | <0.001 | 1.19–1.51 | | | |
| ≥2 | 1.68 | <0.001 | 1.50–1.88 | | | |
| Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status |
| Never | Ref. | | | Ref. | | |
| Previous | 1.38 | <0.001 | 1.19–1.60 | 1.23 | <0.001 | 1.12–1.36 |
| Current | 1.30 | <0.001 | 1.14–1.48 | 1.10 | 0.035 | 1.01–1.19 |
| Regular exercise | Regular exercise | Regular exercise | Regular exercise | Regular exercise | Regular exercise | Regular exercise |
| Never | Ref. | | | Ref. | | |
| Previous | 1.18 | 0.047 | 1.00–1.40 | 1.20 | 0.001 | 1.07–1.33 |
| Current | 1.25 | <0.001 | 1.13–1.38 | 0.98 | 0.567 | 0.91–1.05 |
| Dietary diversity score | Dietary diversity score | Dietary diversity score | Dietary diversity score | Dietary diversity score | Dietary diversity score | Dietary diversity score |
| Poor | Ref. | | | Ref. | | |
| Moderate | 0.82 | <0.001 | 0.74–0.90 | 1.07 | 0.067 | 1.00–1.15 |
| Good | 0.50 | <0.001 | 0.43–0.59 | 0.91 | 0.039 | 0.83–1.00 |
| Housework | Housework | Housework | Housework | Housework | Housework | Housework |
| Almost everyday | | | | Ref. | | |
| Sometimes | | | | 1.06 | 0.216 | 0.97–1.16 |
| Never | | | | 1.29 | <0.001 | 1.18–1.39 |
| Outdoor activities | Outdoor activities | Outdoor activities | Outdoor activities | Outdoor activities | Outdoor activities | Outdoor activities |
| Almost everyday | Ref. | | | Ref. | | |
| Sometimes | 0.89 | 0.022 | 0.80–0.98 | 0.92 | 0.025 | 0.85–0.99 |
| Never | 0.86 | 0.023 | 0.75–0.98 | 0.97 | 0.528 | 0.89–1.06 |
| Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening | Keeping pets or gardening |
| Almost everyday | | | | Ref. | | |
| Sometimes | | | | 0.71 | <0.001 | 0.62–0.81 |
| Never | | | | 0.86 | 0.001 | 0.79–0.94 |
| Reading books | Reading books | Reading books | Reading books | Reading books | Reading books | Reading books |
| Almost everyday | Ref. | | | Ref. | | |
| Sometimes | 0.96 | 0.644 | 0.79–1.16 | 1.42 | <0.001 | 1.25–1.60 |
| Never | 1.15 | 0.088 | 0.98–1.35 | 1.46 | <0.001 | 1.31–1.62 |
| Raising poultry | Raising poultry | Raising poultry | Raising poultry | Raising poultry | Raising poultry | Raising poultry |
| Almost everyday | | | | Ref. | 0.003 | |
| Sometimes | | | | 0.81 | 0.002 | 0.71–0.93 |
| Never | | | | 1.00 | 0.913 | 0.94-1.08 |
| Watching TV | Watching TV | Watching TV | Watching TV | Watching TV | Watching TV | Watching TV |
| Almost everyday | Ref. | | | Ref. | | |
| Sometimes | 0.92 | 0.156 | 0.81–1.03 | 1.19 | <0.001 | 1.10–1.29 |
| Never | 1.24 | <0.001 | 1.10–1.40 | 1.40 | <0.001 | 1.28–1.53 |
| Social participation | Social participation | Social participation | Social participation | Social participation | Social participation | Social participation |
| Almost everyday | | | | Ref. | | |
| Sometimes | | | | 1.59 | <0.001 | 1.29–1.96 |
| Never | | | | 2.05 | <0.001 | 1.68–2.50 |
| Activities of daily living | Activities of daily living | Activities of daily living | Activities of daily living | Activities of daily living | Activities of daily living | Activities of daily living |
| Independent | | | | Ref. | | |
| Disabled | | | | 1.45 | <0.001 | 1.29–1.62 |
| Self–reported quality of life | Self–reported quality of life | Self–reported quality of life | Self–reported quality of life | Self–reported quality of life | Self–reported quality of life | Self–reported quality of life |
| Good | Ref. | | | Ref. | | |
| General | 1.39 | <0.001 | 1.26–1.53 | 0.83 | <0.001 | 0.78–0.89 |
| Poor | 1.64 | <0.001 | 1.38–1.95 | 0.78 | 0.002 | 0.67–0.92 |
| Self–reported health | Self–reported health | Self–reported health | Self–reported health | Self–reported health | Self–reported health | Self–reported health |
| Good | Ref. | | | Ref. | | |
| General | 1.54 | <0.001 | 1.38–1.72 | 0.86 | <0.001 | 0.80–0.92 |
| Poor | 2.30 | <0.001 | 2.02–2.62 | 0.99 | 0.819 | 0.90–1.09 |
Female sex (aOR = 1.44, $95\%$ CI: 1.28–1.62), poor economic status (aOR = 1.91, $95\%$ CI: 1.60–2.28), ≥2 chronic diseases (aOR = 1.68, $95\%$ CI: 1.50–1.88), underweight (aOR = 1.18, $95\%$ CI: 1.05–1.33), current smoking (aOR = 1.30, $95\%$ CI: 1.14–1.48), living in an institution (aOR = 1.74, $95\%$ CI: 1.24–2.45), poor self-reported quality of life (aOR = 1.64,$95\%$ CI: 1.38–1.95), and poor self-reported health (aOR = 2.30, $95\%$ CI: 2.02–2.62) were risk factors of short sleep duration ($p \leq 0.05$). Good dietary diversity score (aOR = 0.50, $95\%$ CI: 0.43–0.59) and ≥1 year of schooling (aOR = 0.77, $95\%$ CI: 0.69–0.84) were protective factors against short sleep duration ($p \leq 0.05$, Table 4).
In contrast to short sleep duration, the odds of long sleep duration were not increased in 2011 and 2014; however, the likelihood of long sleep duration was decreased in 2018 compared with 2008: aOR = 0.98 ($95\%$ CI: 0.91–1.06) in 2011, aOR = 0.92 ($95\%$ CI: 0.84–1.01) in 2014, and aOR = 0.65 ($95\%$ CI: 0.60–0.71) in 2018.
In the sensitivity analysis, similar results were found in the models (Supplementary Tables 3, 4).
## Discussion
Our study showed that the prevalence of self-reported poor sleep quality among older people increased from one in three to nearly one in two between 2008 and 2018 in China. Meanwhile, the prevalence of short sleep duration increased from 5.29 to $8.37\%$, whereas the prevalence of long sleep duration decreased from 28.77 to $19.27\%$. Previous studies have investigated the prevalence of poor sleep quality among older people in China, showing rates ranging from 33.8 to $49.7\%$ (8–11), which is similar to our study. We also compared sleep quality in our population with that reported in other countries, which is $28.9\%$ in Japan [2] and $17.8\%$ in Brazil [35]. It can be seen that the prevalence of poor sleep quality is generally high in older age groups, although there are differences worldwide. The differences between our study findings and other research results may be owing to the differences of age groups, sample sizes, interviewing techniques, economic levels at regional or country level, and culture. Owing to the lack of reports on trends of sleep quality and duration among older people, we compared our findings with trends of sleep duration among adults in the United States during a similar period [3]. Consistent with our results, the prevalence of insufficient sleep was shown to be increasing, accompanied by a decline in the proportion of long sleep duration in the United States [3]. The trend changes observed in our study population may be related to rapid social and economic development in China during the study period, with an accelerating pace of life and increasing life pressure [36]. Our findings supplemented more evidence on the trend of poor sleep quality and short sleep duration, more importantly, indicated sleep health problems among older people was becoming serious. Therefore, providing the possible influencing factors to improve sleep is crucial.
According to our study, female sex was an independent risk factor for poor sleep quality and insufficient sleep duration in older people, which is consistent with study in China [10], Sweden [37] and Korea [38]. Women are more likely to have depression [22] and report more severe physical conditions than men [39], so their sleep may be affected easier. Additionally, in traditional Chinese culture, women have a lower status in the family and do more housework, which may also be influencing factors for poorer sleep. In fact, the role of caregiver for women in most cases, means s the multiple tasks, generates stress and affect sleep. In this study, lower BMI was associated with poor sleep quality, but obesity and overweight were not significantly associated with poor sleep quality or abnormal sleep duration. Although some studies found that obesity was associated with poor sleep quality (40–42) which was not consistent with our findings, one study found that the higher BMI was, the more better sleep quality was in Chinese men which partly supported our findings [43]. We attribute these contradictory results to China's unique conditions, in which a higher BMI is associated with better socioeconomic level, better living conditions, and less economic stress; these in turn may lead to better sleep quality. In addition, Tang et al. [ 22] found that the greater BMI was a protective factor for depression, while disturbed sleep was a risk factor and the depression. Gu et al. [ 44] found that participants with BMI <18.5 kg/m2 were at a significantly higher risk of frailty than those within the normal BMI range, while frailty was associated with sleep problems [45]. Depression and frailty may be the possible mediating factors for the relationship between BMI and sleep, butthe truth of this remains should be investigated in the future.
We found that never participating in social activities were associated with long sleep duration and poor sleep quality. Consistent with previous research, long sleep duration showed an inverse association with engaging in social activities [46, 47]. In a cross-sectional study among older people, participation in social activities was associated with better sleep quality [48]. Generally, for older people who are retired, no longer accompanied by relatives and less physically active, participating in social activities adds to having more social capital, which helps to promote the physical health of older people [49] and prevent depression [50]. We found that not engaging in reading was associated with long sleep duration and not watching TV were associated with long sleep duration and poor sleep quality. Dzierzewski et al. [ 51] and Xie et al. [ 52] both found that excessive the duration of watching TV predicted poorer sleep status among older people. Our findings further provided that never watching TV was also harmful to sleep, not only excessive the duration of watching TV. Combining our findings and previous studies, moderate leisure activities would be acceptable. Similarly, leisure activities (reading, watching TV) also benefit for preventing depression among older people [53]. Therefore, developing healthy living habits that include active participation in social activities and developing new interests may be one way to improve sleep quality and sleep duration among older people in China.
Large sample data and the inclusion of demographic factors, socioeconomic status, lifestyle habits, and health condition variables are among the strengths of this study. However, there are some study limitations as well. Limited information about self-reported sleep duration and sleep quality was the primary drawback in this study, although the reliability and validity of self-reported sleep duration questionnaires have been demonstrated [54]. Besides, non-objective reporting may not distinguish between time spent in bed and time actually asleep. Future research should consider the use of objective methods in sleep assessment, such as using smart devices to collect sleep information [55]. The questionnaire only included one question about sleeping time and did not make a detailed distinction of the period, such as weekend or weekday, which is considered to be different in the literature [56]. The questionnaire also did not directly investigate the psychological status of older people, which may also be a risk factor in their sleep quality and sleep duration [11]. Finally, the subjects of each wave are the older people who meet the criteria, so participants were different across study waves. Therefore, the data may be biased, and more scientific evidence is needed in the future.
## Conclusion
In the past 10 years, the number of older people in China with poor sleep quality has increased from one-third to nearly one-half. The proportion of older adults with short sleep duration increased from 2008 to 2018 whereas the proportion with long sleep duration decreased. Female sex, poor economic status, low social participation, low BMI, and the number of chronic diseases were risk factors associated with sleep quality and duration. China is an aging society, and public health officials must pay attention to the sleep status of older people. At present, the literature on trends and disparities of sleep in the older people in *China is* limited. Our study has important value in guiding sleep health care among the older people in China.
## Data availability statement
Publicly available datasets were analyzed in this study. This data can be found here: https://opendata.pku.edu.cn/dataverse/CHADS;jsessionid=c49bc1a85ef56a3e899accc581b8.
## Ethics statement
The studies involving human participants were reviewed and approved by the Ethical Review Committee of Peking University (IRB00001052-13074). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
Conception and design: LT and JL. Administrative support: JL. Provision of study materials or patients: ZT and YF. Collection and assembly of data: YF. Data analysis and interpretation: ZT and LT. Manuscript writing and final approval of manuscript: All authors.
## 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/fpubh.2023.998699/full#supplementary-material
## References
1. van de Straat V, Bracke P. **How well does Europe sleep? A cross-national study of sleep problems in European older adults**. *Int J Public Health.* (2015) **60** 643-50. DOI: 10.1007/s00038-015-0682-y
2. Takada S, Yamamoto Y, Shimizu S, Kimachi M, Ikenoue T, Fukuma S. **Association between subjective sleep quality and future risk of falls in older people: results from LOHAS**. *J Gerontol A Biol Sci Med Sci.* (2018) **73** 1205-11. DOI: 10.1093/gerona/glx123
3. Sheehan CM, Frochen SE, Walsemann KM, Ailshire JA. **Are US adults reporting less sleep?**. *Sleep.* (2019) **Survey** 2004-2017. DOI: 10.1093/sleep/zsy221
4. Wang C, Bangdiwala SI, Rangarajan S, Lear SA, AlHabib KF, Mohan V. **Association of estimated sleep duration and naps with mortality and cardiovascular events: a study of 116 632 people from 21 countries**. *Eur Heart J.* (2019) **40** 1620-9. DOI: 10.1093/eurheartj/ehy695
5. Plante DT. **The evolving nexus of sleep and depression**. *Am J Psychiatry.* (2021) **178** 896-902. DOI: 10.1176/appi.ajp.2021.21080821
6. Fang EF, Xie C, Schenkel JA, Wu C, Long Q, Cui H. **A research agenda for ageing in China in the 21st century (2nd edition): focusing on basic and translational research, long-term care, policy and social networks**. *Ageing Res Rev.* (2020) **64** 101174. DOI: 10.1016/j.arr.2020.101174
7. Ng ST, Tey NP, Asadullah MN. **What matters for life satisfaction among the oldest-old? Evidence from China**. *PLoS ONE.* (2017) **12** e0171799. DOI: 10.1371/journal.pone.0171799
8. Luo J, Zhu G, Zhao Q, Guo Q, Meng H, Hong Z. **Prevalence and risk factors of poor sleep quality among Chinese elderly in an urban community: results from the Shanghai aging study**. *PLoS ONE.* (2013) **8** e81261. DOI: 10.1371/journal.pone.0081261
9. Li J, Yao YS, Dong Q, Dong YH, Liu JJ, Yang LS. **Characterization and factors associated with sleep quality among rural elderly in China**. *Arch Gerontol Geriatr.* (2013) **56** 237-43. DOI: 10.1016/j.archger.2012.08.002
10. Wang P, Song L, Wang K, Han X, Cong L, Wang Y. **Prevalence and associated factors of poor sleep quality among Chinese older adults living in a rural area: a population-based study**. *Aging Clin Exp Res.* (2020) **32** 125-31. DOI: 10.1007/s40520-019-01171-0
11. Wang YM, Chen HG, Song M, Xu SJ Yu LL, Wang L. **Prevalence of insomnia and its risk factors in older individuals: a community-based study in four cities of Hebei Province, China**. *Sleep Med.* (2016) **19** 116-22. DOI: 10.1016/j.sleep.2015.10.018
12. Chien MY, Chen HC. **Poor sleep quality is independently associated with physical disability in older adults**. *J Clin Sleep Med.* (2015) **11** 225-32. DOI: 10.5664/jcsm.4532
13. Zhang H, Li Y, Zhao X, Mao Z, Abdulai T, Liu X. **The association between PSQI score and hypertension in a Chinese rural population: the Henan Rural Cohort Study**. *Sleep Med.* (2019) **58** 27-34. DOI: 10.1016/j.sleep.2019.03.001
14. Zhu Q, Fan H, Zhang X, Ji C, Xia Y. **Changes in sleep duration and 3-year risk of mild cognitive impairment in Chinese older adults**. *Aging.* (2020) **12** 309-17. DOI: 10.18632/aging.102616
15. Wang D, Ruan W, Peng Y, Li W. **Sleep duration and the risk of osteoporosis among middle-aged and elderly adults: a dose-response meta-analysis**. *Osteoporos Int.* (2018) **29** 1689-95. DOI: 10.1007/s00198-018-4487-8
16. Shan Z, Ma H, Xie M, Yan P, Guo Y, Bao W. **Sleep duration and risk of type 2 diabetes: a meta-analysis of prospective studies**. *Diabetes Care.* (2015) **38** 529-37. DOI: 10.2337/dc14-2073
17. Wang D, Li W, Cui X, Meng Y, Zhou M, Xiao L. **Sleep duration and risk of coronary heart disease: a systematic review and meta-analysis of prospective cohort studies**. *Int J Cardiol.* (2016) **219** 231-9. DOI: 10.1016/j.ijcard.2016.06.027
18. Cai H, Shu XO, Xiang YB, Yang G, Li H, Ji BT. **Sleep duration and mortality: a prospective study of 113 138 middle-aged and elderly Chinese men and women**. *Sleep.* (2015) **38** 529-36. DOI: 10.5665/sleep.4564
19. Thichumpa W, Howteerakul N, Suwannapong N, Tantrakul V. **Sleep quality and associated factors among the elderly living in rural Chiang Rai, northern Thailand**. *Epidemiol Health.* (2018) **40** e2018018. DOI: 10.4178/epih.e2018018
20. Park J, Han JW, Lee JR, Byun S, Suh SW, Kim T, Yoon IY, Kim KW. **Lifetime coffee consumption, pineal gland volume, and sleep quality in late life**. *Sleep* (2018) **41** zsy127. DOI: 10.1093/sleep/zsy127
21. Koyanagi A, Garin N, Olaya B, Ayuso-Mateos JL, Chatterji S, Leonardi M. **Chronic conditions and sleep problems among adults aged 50 years or over in nine countries: a multi-country study**. *PLoS ONE.* (2014) **9** e114742. DOI: 10.1371/journal.pone.0114742
22. Tang X, Qi S, Zhang H, Wang Z. **Prevalence of depressive symptoms and its related factors among China's older adults in 2016**. *J Affect Disord.* (2021) **292** 95-101. DOI: 10.1016/j.jad.2021.04.041
23. Yi Z, Gu D, Poston DL, Vlosky DA. *Healthy Longevity in China: Demographic, Socioeconomic, and Psychological Dimensions* (2008). DOI: 10.1007/978-1-4020-6752-5
24. Hou C, Lin Y, Zimmer Z, Tse LA, Fang X. **Association of sleep duration with risk of all-cause mortality and poor quality of dying in oldest-old people: a community-based longitudinal study**. *BMC Geriatr.* (2020) **20** 357. DOI: 10.1186/s12877-020-01759-6
25. Gu D, Sautter J, Pipkin R, Zeng Y. **Sociodemographic and health correlates of sleep quality and duration among very old Chinese**. *Sleep.* (2010) **33** 601-10. DOI: 10.1093/sleep/33.5.601
26. Hirshkowitz M, Whiton K, Albert SM, Alessi C, Bruni O, DonCarlos L. **National Sleep Foundation's sleep time duration recommendations: methodology and results summary**. *Sleep Health.* (2015) **1** 40-3. DOI: 10.1016/j.sleh.2014.12.010
27. Ren Z, Li Y, Li X, Shi H, Zhao H, He M. **Associations of body mass index, waist circumference and waist-to-height ratio with cognitive impairment among Chinese older adults: based on the CLHLS**. *J Affect Disord.* (2021) **295** 463-70. DOI: 10.1016/j.jad.2021.08.093
28. Liu D, Zhang XR Li ZH, Zhang YJ, Lv YB, Wang ZH. **Association of dietary diversity changes and mortality among older people: a prospective cohort study**. *Clin Nutr.* (2021) **40** 2620-9. DOI: 10.1016/j.clnu.2021.04.012
29. Lv Y, Kraus VB, Gao X, Yin Z, Zhou J, Mao C. **Higher dietary diversity scores and protein-rich food consumption were associated with lower risk of all-cause mortality in the oldest old**. *Clin Nutr.* (2020) **39** 2246-54. DOI: 10.1016/j.clnu.2019.10.012
30. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. **Studies of illness in the aged: the index of ADL: a standardized measure of biological and psychosocial function**. *JAMA.* (1963) **185** 914-9. DOI: 10.1001/jama.1963.03060120024016
31. Du M, Tao L, Liu M, Liu J. **Tourism experiences and the lower risk of mortality in the Chinese elderly: a national cohort study**. *BMC Public Health.* (2021) **21** 996. DOI: 10.1186/s12889-021-11099-8
32. Yuan JQ, Lv YB, Chen HS, Gao X, Yin ZX, Wang WT. **Association between late-life blood pressure and the incidence of cognitive impairment: a community-based prospective cohort study**. *J Am Med Dir Assoc.* (2019) **20** 177-182. DOI: 10.1016/j.jamda.2018.05.029
33. Folstein MF, Folstein SE, McHugh PR. **Mini-mental state. A practical method for grading the cognitive state of patients for the clinician**. *J Psychiatr Res.* (1975) **12** 189-98. DOI: 10.1016/0022-3956(75)90026-6
34. Yang L, Martikainen P, Silventoinen K, Konttinen H. **Association of socioeconomic status and cognitive functioning change among elderly Chinese people**. *Age Ageing.* (2016) **45** 674-80. DOI: 10.1093/ageing/afw107
35. Bof de. **Andrade F, Watt RG, Lima-Costa MF, de Oliveira C. Poor sleep quality and oral health among older Brazilian adults**. *Oral Dis.* (2022) **28** 227-32. DOI: 10.1111/odi.13734
36. Matricciani L, Bin YS, Lallukka T, Kronholm E, Dumuid D, Paquet C. **Past, present, and future: trends in sleep duration and implications for public health**. *Sleep Health.* (2017) **3** 317-23. DOI: 10.1016/j.sleh.2017.07.006
37. Broström A, Wahlin A, Alehagen U, Ulander M, Johansson P. **Sex-specific associations between self-reported sleep duration, cardiovascular disease, hypertension, and mortality in an elderly population**. *J Cardiovasc Nurs.* (2018) **33** 422-8. DOI: 10.1097/JCN.0000000000000393
38. Quan SA Li YC, Li WJ Li Y, Jeong JY, Kim DH. **Gender differences in sleep disturbance among elderly Koreans: Hallym aging study**. *J Korean Med Sci.* (2016) **31** 1689-95. DOI: 10.3346/jkms.2016.31.11.1689
39. Barsky AJ, Peekna HM, Borus JF. **Somatic symptom reporting in women and men**. *J Gen Intern Med.* (2001) **16** 266-75. DOI: 10.1046/j.1525-1497.2001.016004266.x
40. Hung HC, Yang YC, Ou HY, Wu JS, Lu FH, Chang CJ. **The association between self-reported sleep quality and overweight in a Chinese population**. *Obesity.* (2013) **21** 486-92. DOI: 10.1002/oby.20259
41. Rahe C, Czira ME, Teismann H, Berger K. **Associations between poor sleep quality and different measures of obesity**. *Sleep Med.* (2015) **16** 1225-8. DOI: 10.1016/j.sleep.2015.05.023
42. Park SK, Jung JY, Oh CM, McIntyre RS, Lee JH. **Association between sleep duration, quality and body mass index in the Korean population**. *J Clin Sleep Med.* (2018) **14** 1353-60. DOI: 10.5664/jcsm.7272
43. Gildner TE, Liebert MA, Kowal P, Chatterji S, Josh Snodgrass J. **Sleep duration, sleep quality, and obesity risk among older adults from six middle-income countries: findings from the study on global AGEing and adult health (SAGE)**. *Am J Hum Biol.* (2014) **26** 803-12. DOI: 10.1002/ajhb.22603
44. Gu Y, Wu W, Bai J, Chen X, Chen X, Yu L. **Association between the number of teeth and frailty among Chinese older adults: a nationwide cross-sectional study**. *BMJ Open.* (2019) **9** e029929. DOI: 10.1136/bmjopen-2019-029929
45. Xie B, Ma C, Chen Y, Wang J. **Prevalence and risk factors of the co-occurrence of physical frailty and cognitive impairment in Chinese community-dwelling older adults**. *Health Soc Care Commun.* (2021) **29** 294-303. DOI: 10.1111/hsc.13092
46. Robbins R, Jean-Louis G, Gallagher RA, Hale L, Branas CC, Gooneratne N. **Examining social capital in relation to sleep duration, insomnia, and daytime sleepiness**. *Sleep Med.* (2019) **60** 165-72. DOI: 10.1016/j.sleep.2019.03.019
47. Yang L, Wang H, Cheng J. **Association between social capital and sleep duration among rural older adults in China**. *BMC Public Health.* (2022) **22** 12. DOI: 10.1186/s12889-021-12441-w
48. Chen JH, Lauderdale DS, Waite LJ. **Social participation and older adults' sleep**. *Soc Sci Med.* (2016) **149** 164-73. DOI: 10.1016/j.socscimed.2015.11.045
49. Lindström M, Giordano GN. **The 2008 financial crisis: Changes in social capital and its association with psychological wellbeing in the United Kingdom—a panel study**. *Soc Sci Med.* (2016) **153** 71-80. DOI: 10.1016/j.socscimed.2016.02.008
50. Landstedt E, Almquist YB, Eriksson M, Hammarström A. **Disentangling the directions of associations between structural social capital and mental health: longitudinal analyses of gender, civic engagement and depressive symptoms**. *Soc Sci Med.* (2016) **163** 135-43. DOI: 10.1016/j.socscimed.2016.07.005
51. Dzierzewski JM, Sabet SM, Ghose SM, Perez E, Soto P, Ravyts SG. **Lifestyle factors and sleep health across the lifespan**. *Int J Environ Res Public Health* (2021) **18** 6626. DOI: 10.3390/ijerph18126626
52. Xie YJ, Cheung DS, Loke AY, Nogueira BL, Liu KM, Leung AY. **Relationships between the usage of televisions, computers, and mobile phones and the quality of sleep in a chinese population: community-based cross-sectional study**. *J Med Internet Res.* (2020) **22** e18095. DOI: 10.2196/18095
53. Kang Q, Lyu YB, Wei Y, Shi WY, Duan J, Zhou JH. **Influencing factors for depressive symptoms in the elderly aged 65 years and older in 8 longevity areas in China**. *Zhonghua Liu Xing Bing Xue Za Zhi.* (2020) **41** 20-4. DOI: 10.3760/cma.j.issn.0254-6450.2020.01.005
54. Girschik J, Heyworth J, Fritschi L. **Reliability of a sleep quality questionnaire for use in epidemiologic studies**. *J Epidemiol.* (2012) **22** 244-50. DOI: 10.2188/jea.JE20110107
55. Robbins R, Affouf M, Seixas A, Beaugris L, Avirappattu G, Jean-Louis G. **Four-year trends in sleep duration and quality: a longitudinal study using data from a commercially available sleep tracker**. *J Med Internet Res.* (2020) **22** e14735. DOI: 10.2196/14735
56. Basner M, Dinges DF. **Sleep duration in the United States 2003–2016: first signs of success in the fight against sleep deficiency?**. *Sleep* (2018) **41** zsy012. DOI: 10.1093/sleep/zsy012
|
---
title: Thermoneutral housing shapes hepatic inflammation and damage in mouse models
of non-alcoholic fatty liver disease
authors:
- Jarren R. Oates
- Keisuke Sawada
- Daniel A. Giles
- Pablo C. Alarcon
- Michelle S.M.A. Damen
- Sara Szabo
- Traci E. Stankiewicz
- Maria E. Moreno-Fernandez
- Senad Divanovic
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9982161
doi: 10.3389/fimmu.2023.1095132
license: CC BY 4.0
---
# Thermoneutral housing shapes hepatic inflammation and damage in mouse models of non-alcoholic fatty liver disease
## Abstract
### Introduction
Inflammation is a common unifying factor in experimental models of non-alcoholic fatty liver disease (NAFLD) progression. Recent evidence suggests that housing temperature-driven alterations in hepatic inflammation correlate with exacerbated hepatic steatosis, development of hepatic fibrosis, and hepatocellular damage in a model of high fat diet-driven NAFLD. However, the congruency of these findings across other, frequently employed, experimental mouse models of NAFLD has not been studied.
### Methods
Here, we examine the impact of housing temperature on steatosis, hepatocellular damage, hepatic inflammation, and fibrosis in NASH diet, methionine and choline deficient diet, and western diet + carbon tetrachloride experimental models of NAFLD in C57BL/6 mice.
### Results
We show that differences relevant to NAFLD pathology uncovered by thermoneutral housing include: (i) augmented NASH diet-driven hepatic immune cell accrual, exacerbated serum alanine transaminase levels and increased liver tissue damage as determined by NAFLD activity score; (ii) augmented methionine choline deficient diet-driven hepatic immune cell accrual and increased liver tissue damage as indicated by amplified hepatocellular ballooning, lobular inflammation, fibrosis and overall NAFLD activity score; and (iii) dampened western diet + carbon tetrachloride driven hepatic immune cell accrual and serum alanine aminotransferase levels but similar NAFLD activity score.
### Discussion
Collectively, our findings demonstrate that thermoneutral housing has broad but divergent effects on hepatic immune cell inflammation and hepatocellular damage across existing experimental NAFLD models in mice. These insights may serve as a foundation for future mechanistic interrogations focused on immune cell function in shaping NAFLD progression.
## Introduction
The unabated obesity pandemic (~1.5 billion people obese globally) is accompanied by a concomitant increase in the prevalence of non-alcoholic fatty liver disease (NAFLD). NAFLD, which affects approximately 25-$30\%$ of obese individuals, is considered the most common chronic liver disease and a leading cause for needing liver transplantation [1, 2]. NAFLD encompasses a broad spectrum of liver conditions ranging from steatosis to non-alcoholic steatohepatitis (NASH) to cirrhosis, which eventually can progress to hepatocellular carcinoma (HCC) [1, 3]. To study the broad aspects of NAFLD progression the field has traditionally employed a variety of experimental animal models. Notably, in mouse models of NAFLD, the strengths of each model are focused on select key parameters represented in human disease (e.g., steatosis, steatohepatitis, hepatocyte ballooning, Mallory-Denk bodies and fibrosis).
Prominent mouse models of NAFLD involve various dietary challenges including high fat diet (HFD), methionine choline deficient (MCD) diet, and chemical perturbations in combination with western diet (WD) (e.g., Carbon tetrachloride [CCl4] + WD diet) (3–6). HFD feeding drives robust obesity and hepatic recruitment of Kupffer cells and neutrophils, and only causes mild inflammation and minimal fibrosis (the latter only evident after prolonged HFD feeding) [7, 8]. Conversely, MCD diet consistently induces robust hepatic immune cell recruitment, liver damage, and development of hepatic fibrosis, without the induction of obesity (9–11). Chemical perturbations, including high-dose CCl4, promote hepatic inflammation and fibrosis with induction of hepatocellular necrosis that is not consistent with human liver disease pathogenesis [9]. As such, CCl4 is used at lower doses and in combination with WD feeding to recapitulate histopathological manifestations of human NAFLD more accurately. However, unlike in human disease, this model is associated with weight loss [5]. Nevertheless, despite being traditionally employed in studying NAFLD pathogenesis, the above discussed models only partly recapitulate clinical human disease progression. Notably, these shortcomings may limit the discovery of key cellular and molecular mechanisms underlying liver disease pathogenesis. Given that inflammation represents a unifying component of NAFLD progression (11–14) across the models used, further improvement of existing experimental models to enable discovery of mechanisms that shape inflammatory responses in NAFLD may represent the critical locus of the effect for translation of key discoveries to clinical relevance.
Ambient housing temperature regulates inflammatory responses to internal and external stimuli [15, 16]. Common animal facilities house mice in thermo-stress (Ts) (20-23˚C) conditions. Such conditions however are associated with the activation of cold stress responses in mice, augmented production of stress hormone (e.g., corticosterone and catecholamine) which drives a multitude of physiological and metabolic changes [17, 18]. Importantly, Ts housing suppresses immune responses including altered cellular energy availability, cytokine production, and lymphocyte egress from lymph nodes (15, 19–21). Conversely, thermoneutral (Tn) housing (29-34˚C), a temperature where mice do not need to expend excess energy to maintain core body temperature, reverses the suppression of immune responsiveness observed at Ts housing conditions [1, 3, 16, 22]. In the context of NAFLD, Tn housing coupled with HFD feeding accelerates and exacerbates disease progression and pathogenesis. Specifically, Tn housing in combination with HFD feeding enhances hepatic steatosis and inflammation, and hepatocellular damage in wild type C57BL/6 male mice and hepatic fibrosis in AKR mice [3]. Further, wild type C57BL/6 female mice, which are generally resistant to HFD-induced obesity when housed at Ts conditions, developed robust obesity and NAFLD when housed at Tn conditions [3]. Thus, the ability of Tn housing to reverse paradigms seen at Ts conditions by restricting immune suppression and promoting obesity in female mice warrants the examination of the impact of Tn housing on hepatic inflammation, steatosis, liver damage and development of hepatic fibrosis across multiple, traditionally used, experimental models of NAFLD in mice.
Here, we show that Tn housing shapes hepatic inflammation across multiple, traditionally used, experimental mouse models of NAFLD. In a model utilizing NASH diet, Tn housing augments obesity, serum ALT, hepatic immune cell accrual, expression of immune cell recruiting chemokines which correlates with increased lobular inflammation and liver disease as determined by NAFLD Activity Score (NAS). Similarly, in the MCD diet driven NAFLD model, Tn housing promotes increased hepatic expression of immune cell recruiting chemokines correlating with hepatic immune cell accrual, increased expression of fibrosis-associated genes and worsened fibrosis as indicated by increased cholangiolar proliferations accompanied by thicker collagenous pericellular fibers and overall increased NAS severity. Lastly, in the context of chemical perturbations combined with WD diet induced obesity, Tn housing decreases serum ALT which correlates with a decrease in hepatic immune cell accrual and modified immune cell inflammatory capacity. Collectively, these findings demonstrate for the very first time that thermoneutral housing has broad, yet divergent effects on hepatic immune cell inflammation and hepatocellular damage across existing experimental NAFLD models in C57BL/6 mice. As studies utilizing thermoneutral housing become more prevalent, insights gained may serve as a foundation for interrogation of mechanisms instructing immune cell function and development of future therapies to NAFLD.
## Mice and dietary studies
Wild type (WT) mouse breeding pairs, originally purchased from Jackson Laboratories, were on C57BL/6J background, housed at thermo-stress (Ts; 22°C) conditions with free access to autoclaved food and water, and bred at Cincinnati Children’s Hospital Medical Center (CCHMC) in a specific pathogen-free (spf) facility. Only 8-week-old WT male mice were used in our studies. For our studies, 6-week old mice were maintained at Ts or placed at thermoneutral (Tn; 30°C with $30\%$ humidity) conditions for 2 weeks to allow acclimation prior to the initiation of dietary challenge. Tn housing was accomplished using Caron chambers (Caron Products & Services, INC). For all studies, food and water were replaced weekly. Body weight was recorded weekly. Corn cob bedding was used in the housing of all mice. All animal care was provided in accordance with the Guide for the Care and Use of Laboratory Animals. All studies were approved by the Cincinnati Children’s Hospital Medical Center IACUC.
## Chow diet
WT C57BL/6 mice (8-week-old) were fed CD (fat $13.5\%$ kcal, carbohydrate $59\%$ kcal, protein $27.5\%$ kcal; LabDiet 5010) and housed at either Ts or Tn as controls to specific dietary challenge experiments.
## Methionine-choline deficient diet
WT C57BL/6 mice (8-week-old) were fed MCD diet (MCD; Research Diets #A02082002B; $16\%$ Protein, $63\%$ Carbohydrate and $21\%$ Fat kcal/gram) and housed at either Ts or Tn for 4 or 10 weeks as previously described [10, 23].
## NASH diet
WT C57BL/6 mice (8-week-old) were fed a NASH diet ($40\%$ kcal fat, $20\%$ kcal fructose, and $2\%$ kcal cholesterol by weight; Research diets, D09100301) and housed at either Ts or Tn for 22 weeks as previously described [24].
## Carbon tetrachloride (CCl4)
WT C57BL/6 mice (8-week-old) were fed CD. CCl4 (Sigma-Aldrich, 289116-100ML) was injected with the dose of 2 µl (0.32 µg)/g of body weight, of CCl4 or olive oil (control) i.p. 2x weekly for 3 weeks.
## Western diet + carbon tetrachloride (WD+CCl4)
WT C57BL/6 mice (8-week-old) were fed a WD ($21.1\%$ fat, $41\%$ Sucrose, and $1.25\%$ Cholesterol by weight; Teklad diets, TD. 120528) and a high sugar solution (23.1 g/L d-fructose and 18.9 g/L d-glucose) in drinking water. CCl4 (Sigma-Aldrich, 289116-100ML), 0.2 µl (0.32 µg)/g of body weight or corn oil (control) was i.p. injected once per week, as previously described [5, 24] for 12 weeks.
## Hepatic immune cell isolation
Hepatic immune cells were isolated using Miltenyi Biotec Gentlemax technology followed by Percoll gradient. Briefly, whole liver was homogenized using Miltenyi Gentlemax C tubes using RPMI + $10\%$ fetal bovine serum. After resuspension, cells were centrifuged at 2000 rpm for 10 minutes. Cell pellets were homogenized in a $33\%$ Percoll solution (Sigma-Aldrich) diluted in RPMI medium 1640 (Gibco). Following gradient separation, and lysing of red blood cells, hepatic infiltrating immune cells were subsequently analyzed by flow cytometry [3, 25, 26].
## Hepatic function
Hepatic triglycerides (TGs) were quantified using Triglyceride Reagent and Triglyceride Standards (Pointe Scientific). Serum alanine aminotransferase (ALT) and aspartate aminotransferase levels (AST) were quantified using ALT Reagent, AST Reagent and Catatrol I and II (Catachem). For histology, liver tissue was fixed in $10\%$ buffered formalin, and stained with H&E or Masson’s trichrome and evaluated by a board-certified pathologist [3, 24].
## Hepatic immune cell characterization
To determine immune cell population single cell suspensions derived from hepatic tissues, isolated immune cells were labeled with monoclonal antibodies. For cytokine production, total single cells were stimulated for 4 hours with Phorbol 12-myristate 13-acetate (PMA; 50 ng/ml) (Sigma-Aldrich, St. Louis, MO) and Ionomycin (1 μg/ml) (EMD Millipore), in presence of Brefeldin A (10 μg/mL) (GoldBio). Subsequently, data were collected using an LSR Fortessa (BD) and Cytek Aurora (Cytek Biosciences) and analyzed using FlowJo X software (vX10).
## CD4+ T cells
Mouse cells were stained with Live/Dead stain (Zombie UV Dye: Biolegend) and with directly-conjugated monoclonal antibodies to CD45-PE-Dazzle594 (Biolegend, 104), TCRβ-APCef780 or APC (Invitrogen, H57-597), CD4-APC-ef780 or BV786 (e-Biosciences, RM4-5), then fixed, permeabilized and stained for the cytokines IL-17A-PerCpCy5.5 or PE (e-Biosciences, 17B7), IFNγ-PE-Cy7 (e-Biosciences, XMG1.2) and TNFα-BV650 (Biolegend, MP6-XT22).
## CD8+ T cells
Mouse cells were stained with Live/Dead stain (Zombie UV Dye: Biolegend) and with directly-conjugated monoclonal antibodies to CD45-PE-Dazzle594 (Biolegend, 104), TCRβ-APCef780 or APC (Invitrogen, H57-597), CD8-BV510 (Biolegend, 53-6.7), then fixed, permeabilized and stained for the cytokines IL-17A-PerCpCy5.5 or PE (e-Biosciences, 17B7), IFNγ-PE-Cy7 (e-Biosciences, XMG1.2) and TNFα-BV650 (Biolegend, MP6-XT22).
## Macrophages
Mouse cells were stained with Live/Dead stain (Zombie UV Dye: Biolegend) and with directly conjugated monoclonal antibodies to CD45-PE-Dazzle594 (Biolegend, 104), CD11b-eF450 (Biolegend, 17A2), F$\frac{4}{80}$-APC (eBiosciences, BM8) then fixed, permeabilized and stained for the cytokine TNFα-BV650 (Biolegend, MP6-XT22).
## NK cells
Mouse cells were stained with Live/Dead stain (Zombie UV Dye: Biolegend) and with directly conjugated monoclonal antibodies to CD45-PE-Dazzle594 (Biolegend, 104), NK1.1-BV711 (Biolegend, PK136) then fixed, permeabilized and stained for the cytokine IFNγ-PE-Cy7 (e-Biosciences, XMG1.2).
## NKT cells
Mouse cells were stained with Live/Dead stain (Zombie UV Dye: Biolegend) and with directly conjugated monoclonal antibodies to CD45-PE-Dazzle594 (Biolegend, 104), NK1.1-BV711 (Biolegend, PK136) and TCRβ-APCef780 or APC (Invitrogen, H57-597).
## B cells
Mouse cells were stained with Live/Dead stain (Zombie UV Dye: Biolegend) and with directly conjugated monoclonal antibodies to CD45-PE-Dazzle594 (Biolegend, 104) and B220-BV605 (Biolegend, RA3-6B2).
## Neutrophils
Mouse cells were stained with Live/Dead stain (Zombie UV Dye: Biolegend) and with directly conjugated monoclonal antibodies to CD45-PE-Dazzle594 (Biolegend, 104), CD11b-eF450 (Biolegend, 17A2), Gr1-FITC (Invitrogen, RB6-8C5) then fixed, permeabilized and stained for the cytokine TNFα-BV650 (Biolegend, MP6-XT22).
## mRNA and qPCR analysis
Tissue samples were homogenized in TRIzol. RNA was extracted, reverse transcribed to complementary DNA (cDNA), and subjected to qPCR analysis (Light Cycler 480 II; Roche Diagnostics) as previously described [3, 15]. The primer sequences, all from Invitrogen are as follows see below:
## Statistical analysis
For statistical analysis, normality and lognormality tests and parametric tests were employed as determined by the Graphpad Prism software. A 2-tailed student’s t test was used when the comparison was between 2 groups, while a 1-way ANOVA with Tukey’s post hoc test to assess differences between specific groups was employed for 3 or more groups. Statistical analysis was completed using Prism 5a (GraphPad Software Inc.). All values are represented as means ± SEM. A P-value less than 0.05 was considered significant.
## Tn housing augments NASH diet driven myeloid immune cell accrual in the liver and exacerbates liver disease pathogenesis
Our initial investigation focused on determining the impact of Tn housing on liver damage at homeostatic conditions. Tn housing did not impact body weight and promoted a slight reduction in wet liver weights, compared to Ts. Importantly, both Tn and Ts housed animals displayed similar serum ALT levels (Supplementary Figures 1A–C). As coupling high fat diet (HFD)-driven obesity with Tn housing exacerbates inflammatory responsiveness and accelerates NAFLD pathogenesis [3], we next examined the impact of Tn housing in a model of obesogenic diet-driven NAFLD, a diet rich in cholesterol and sugars that has been postulated to mimic human disease (27–29). To mimic these conditions, Ts and Tn housed C57BL/6 wild type male mice were fed a NASH diet for 22 weeks (Figure 1A). Tn housed mice, compared to Ts housed counterparts, had increased total body weight gain over time ($$p \leq 0.01$$) (Figure 1B). Further examination of visceral white adipose tissue (WAT) depots revealed that Tn housing dominantly impacted perirenal WAT fat redistribution ($$p \leq 0.04$$) rather than inguinal and epididymal WAT depots (Supplementary Figure 2A).
**Figure 1:** *Tn housing augments NASH diet driven myeloid immune cell accrual in the liver and exacerbates liver disease pathogenesis. (A) Schematic of the experimental design. Eight-week-old WT mice maintained at Ts or acclimated to Tn for 2 weeks prior to study initiation were fed a chow (baseline reference) or NASH diet for 22 weeks. During the course of 22 weeks, (B) body weights of Ts and Tn housed mice were recorded. At the conclusion of the study, additional parameters of NAFLD severity were analyzed. (C) The liver tissue was preserved in formalin, stained with hematoxylin and eosin (H&E), and analyzed by a clinical pathologist. Table depicting histological scoring analyses for fibrosis, lobular inflammation, hepatocellular ballooning, macrovesicular (MV) steatosis grade, steatosis percentage, portal inflammation and NAFLD activity score (NAS) severity. (D) (Top) Extensive steatosis (80-95%) on H&E with a combination of classic large droplet macrovesicular steatosis (LD-MS) and small droplet (SD-MS) macrovesicular steatosis. (Portae: green arrows. Central veins: light blue arrows) (Top) Black bar = 629µm; (Bottom) Black bar = 109µm. (E) Liver H&E staining. (Top) Ballooning (yellow arrows), in a mixed patterned background with overlapping features, including SD-MS (light blue arrows) and LD-MS. Black bar = 25µm (Middle) Example for focal lobular inflammation (lymphocytic) (blue arrows), classic for NAFLD. Black bar = 25µm. (Bottom) Focal portal inflammation, perivenular (blue arrow), with small focus of cholangiolar reactivity (green arrow). Black bar = 33µm. (F) Hepatocellular damage as quantified by serum ALT levels collected at the conclusion of the study. (G) Hepatic inflammation as defined by flow cytometric analyses of hepatic immune cell accrual at the conclusion of the study. Donut charts depicting liver immune compositions (macrophages, neutrophils, CD4+ cells, CD8+ cells, NK cells, and unidentified cells) as determined by flow cytometry from percent of CD45+ population. In bar graphs, data represent mean +/- SEM. (B, C, F, G) Representative of 2 individual experiments (n=5-6/condition). (F) One-way ANOVA. *P<0.05, ** P<0.01, **** P<0.0001.*
Given the observed changes in obesity and adiposity in Tn housed mice, we analyzed the impact of Tn housing on gross liver histology. Obesity is associated with ectopic fat deposition in the liver, thus the impact of Tn housing on NASH diet-driven hepatic steatosis and hepatocellular damage was examined. Both Tn and Ts housed, NASH diet fed mice displayed equal presence of classic large droplet macrovesicular steatosis (LD-MS; a single large steatotic droplet replaces nearly the entire cytoplasm and pushes the nucleus peripherally) and small droplet macrovesicular steatosis (SD-MS; numerous smaller droplets of fat that replace essentially the entire cytoplasm and expand the cell) (Figures 1C, D) (see arrows). Congruently, both Tn and Ts housed, NASH diet fed mice had similar wet liver weights (Supplementary Figure 2B) and hepatic triglyceride accumulation (Supplementary Figure 2C). However, additional histological examination revealed that despite similar hepatocyte ballooning, Tn housed NASH-diet-fed mice, compared to Ts housed counterparts, exhibited increased focal lobular inflammation ($$p \leq 0.05$$), and had an overall increase in NAFLD Activity Score (NAS) severity ($$p \leq 0.04$$) (Figures 1C, depicted in E) (see arrows). Lastly, in agreement with histological analyses, the Tn housed, NASH diet fed mice, compared to Ts housed counterparts, had exacerbated serum alanine transaminase (ALT) levels ($$p \leq 0.01$$) (Figure 1F).
Hepatocellular damage is closely linked with altered hepatic inflammation. Hence, we next examined the impact of Tn housing on NASH diet-driven hepatic inflammation, hepatic immune cell composition and immune cell inflammatory capacity. Tn and Ts housed, NASH diet fed mice had similar hepatic mRNA expression of Tnfa, a proinflammatory cytokine that contributes to hepatic inflammation and damage [30, 31] (Supplementary Figure 2D). However, Tn housing increased hepatic mRNA expression of immune cell recruiting chemokines Ccl2 ($$p \leq 0.000009$$) and Ccl3 ($$p \leq 0.04$$) (Supplementary Figure 2E). Subsequent analysis of the immune cell composition demonstrated that Tn housing promoted an increase in total hepatic immune cell accrual ($$p \leq 0.04$$) in context of NASH diet feeding (Supplementary Figure 2F). Cellular characterization of hepatic immune infiltrate in Tn housed NASH-diet-fed mice, compared to Ts housed counterparts, revealed specific increase in frequencies and absolute numbers of macrophages (CD45+CD11b+F$\frac{4}{80}$+) ($$p \leq 0.03$$) and neutrophils (CD45+CD11b+Gr1+) ($$p \leq 0.03$$), but reduced frequency of CD4+ T cells (CD45+CD3+CD4+) (Figure 1G and Supplementary Table 1, for gating strategy see Supplementary Figure 3). However, despite the apparent changes in hepatic immune cell composition, the inflammatory capacity of hepatic immune cells following PMA/ionomycin stimulation (Supplementary Figures 2G, H) was similar in Tn and Ts housed, NASH diet fed mice.
Prolonged hepatic inflammation induces tissue damage and development of hepatic fibrosis [32, 33]. However, traditional HFD feeding does not induce robust hepatic fibrosis in C57BL/6 wild type mice under Ts housing conditions. Notably, diets high in cholesterol and sugar (i.e., fructose and sucrose) content are known to yield upregulation of pathways associated with fibrosis and promote histological patterns of fibrosis that mimic fibrosis-like characteristics seen in human disease [34, 35]. Hence, the impact of Tn housing on hepatic fibrosis was examined. Tn and Ts housed NASH-diet-fed mice had similar expression of fibrosis associated genes Col1a1, Col1a2 and Acta2 in the liver (Supplementary Figure 2I). Congruently, both Tn and Ts housed, NASH diet fed mice had minimal fibrosis in association with periportal cholangiolar reactivity as determined by trichrome staining (Figure 1C and Supplementary Figure 2J).
NAFLD is also associated with dysregulated lipid metabolism and insulin signaling, both of which contribute to development and severity of metabolic syndrome [36, 37]. Obesity in particular is associated with dysregulation of lipid metabolism and insulin signaling pathways [36, 37]. Hence, the impact of Tn housing on pathways associated with lipid metabolism and insulin signaling were next examined. Notably, Tn and Ts housed, NASH diet fed mice, despite differences in NAFLD severity, had similar hepatic expression of insulin signaling (Pparγ and Irs1) and lipid metabolism-associated genes (Lxrα, Lipe, Chrebp, and Srebp) (Supplementary Figures 2K–L).
## Tn housing augments MCD diet driven myeloid immune cell accrual in the liver and exacerbates liver disease pathogenesis
Increased incidence of NAFLD is reported to also occur in the absence of obesity [38]. Using the MCD diet model of NAFLD, which induces weight loss but promotes hepatic steatosis, inflammation, and hepatocellular damage, we next examined whether Tn housing enhanced NAFLD pathogenesis in an obesity independent setting. Tn and Ts housed C57BL/6 wild type male mice were fed MCD diet for 4 weeks (Figure 2A). Tn housing did not impact MCD diet driven weight loss – a key characteristic of MCD diet feeding (Figure 2B). Hepatic triglycerides levels were similar despite increased liver to body weight ratio ($$p \leq 0.005$$) in Tn housed, MCD diet fed mice (Supplementary Figures 4A, B). However, histopathological analyses revealed that Tn housed, MCD diet fed mice, compared to Ts housed counterparts, exhibited a higher macrovesicular steatosis grade ($$p \leq 0.04$$) (Figures 2C, D, see green arrows). In addition, Tn housed, MCD diet fed mice, compared to Ts housed counterparts, had increased overall hepatocellular damage as evidenced by increased focal inflammation ($$p \leq 0.003$$) and necrosis (acidophil bodies) (Figure 2E, see red arrows), and a significant increase in NAS severity ($$p \leq 0.004$$) (Figure 2C), but had similar ALT and Aspartate aminotransferase (AST) levels (Figure 2F and Supplementary Figure 4C). Tn housed, MCD diet fed mice, compared to Ts counterparts, exhibited comparable hepatic expression levels of Tnfa, Ccl2 and Ccl3 in total liver tissue (Supplementary Figures 4D, E). Similarly, analysis of the immune cell composition showed comparable numbers of total hepatic immune cells (Supplementary Figure 4F). Characterization of hepatic immune cell infiltrate revealed that Tn housing in combination with MCD diet feeding promoted a specific increase in frequencies and absolute numbers of macrophages (frequency $$p \leq 0.01$$; absolute # $$p \leq 0.01$$) and neutrophils (frequency $$p \leq 0.0002$$; absolute # $$p \leq 0.001$$) and a decrease in CD4+ and CD8+ T cells absolute numbers (Figure 2G and Supplementary Table 2). Nevertheless, despite the changes in hepatic immune cell composition, the inflammatory capacity of these cells following ex vivo PMA/ionomycin stimulation was not altered (Supplementary Figures 4G, H).
**Figure 2:** *Tn housing augments MCD diet driven hepatic myeloid immune cell accrual in the liver and exacerbates liver disease pathogenesis. (A) Schematic of the experimental design. Eight-week-old WT mice maintained at Ts or acclimated to Tn for 2 weeks and fed a chow (baseline reference) or MCD diet for 4 weeks. During the course of the 22 weeks, (B) body weights of Ts and Tn housed mice were recorded. (C) The liver tissue was preserved in formalin, stained with hematoxylin eosin (H&E), and analyzed by a clinical pathologist. Table depicting histological scoring analyses for fibrosis, lobular inflammation, hepatocellular ballooning, macrovesicular (MV) steatosis grade, steatosis percentage, portal inflammation and NAFLD activity score (NAS) severity. (D) Liver H&E staining. At low power, zone 2-predominant macrovesicular steatosis is noted diffusely (top and bottom) (For architectural orientation, green arrows point to portae.) Top, black bar = 332µm; Bottom, black bar = 119µm. (E) Liver hematoxylin eosin (H&E) staining. (Top) Reactive cholangiolar proliferations, with inflammation, some microaggregating (red arrows) and occasionally “microgranulomatous” (insert). Black bar = 34µm. (Bottom) Partial pericentral/zone 3-sparing, though with spotty hepatocellular necrosis (red arrows) that are occasionally clustering. Black bar = 25µm. (F) Hepatocellular damage as quantified by serum ALT levels collected at the conclusion of the study. (G) Hepatic inflammation as defined by flow cytometric analyses of hepatic immune cell accrual. Donut charts depicting liver immune compositions (macrophages, neutrophils, CD4+ cells, CD8+ cells, and unidentified cells as determined by flow cytometry from percent of CD45+ population. (H) Liver tissues were preserved in formalin, sectioned and trichrome stained for analysis by pathologist. (Top) Patterns of fibrosis in trichrome stain. Black bar = 119µm. (Bottom) Focally accentuated collagenous pericellular fibrosis in a periportal location. Black bar = 61µm. In bar graphs, data represent mean +/- SEM. (B) Data combined from 2 individual experiments (n=6-8/condition). (C) Data combined from 2 individual experiments, (n=8/condition). (F, G) Data combined from 2 individual experiments (n=8/condition). (C) One-way ANOVA. *P<0.05, **P<0.01, ****P<0.0001..*
MCD diet is a well-accepted experimental model of hepatic fibrosis [9, 39, 40]. Whether Tn housing modifies development of hepatic fibrosis in the context of MCD feeding is unknown. Tn and Ts housed, MCD diet fed mice exhibited similar hepatic expression of fibrosis associated genes Col1a1, Col1a2 and Acta2 ($$p \leq 0.18$$) (Supplementary Figure 4I). Nevertheless, Tn housed, MCD diet fed mice, compared to Ts housed counterparts, had worsened hepatic fibrosis as indicated by increased cholangiolar proliferations accompanied by thicker collagenous pericellular fibers (Figures 2C, H, see green arrows). MCD diet does not induce metabolic syndrome [9, 41]. We next examined whether Tn housing would be sufficient to alter the expression of lipid metabolism- and insulin signaling-associated genes linked with metabolic syndrome in NAFLD [36, 37]. Notably, Tn and Ts housed, MCD diet fed mice, despite differences in NAFLD severity, exhibited similar hepatic expression of insulin signaling (Pparγ and Irs1) and lipid metabolism-associated genes (Lxrα, Lipe, Chrebp, and Srebp) (Supplementary Figures 4J, K).
Prolonged MCD feeding exacerbates NASH pathology [10]. To further define the impact of Tn housing on prolonged MCD-driven NAFLD, C57BL/6 wild type mice were housed at Tn and Ts and fed MCD diet for 10 weeks (Supplementary Figure 5A). Similar to 4 weeks of MCD diet feeding, no differences in weight loss over time between Tn and Ts housed mice were observed (Supplementary Figure 5B). Tn housed, MCD diet fed mice, compared to Ts counterparts, had decreased liver to body weight ratio ($$p \leq 0.003$$) and similar total hepatic triglyceride accumulation compared to Ts housed counterparts (Supplementary Figures 5C, D), but displayed exacerbated steatosis percentage ($$p \leq 0.03$$) (detected in 85-$95\%$ range) (Supplementary Figure 5E). Additional histological examination revealed that Tn housed, MCD diet fed mice, compared to Ts housed counterparts, had increased hepatocellular ballooning ($$p \leq 0.01$$), multifocal lobular inflammation ($$p \leq 0.01$$), and the overall NAS severity ($$p \leq 0.005$$) (Supplementary Figure 5E), yet ALT and AST remained similar (Supplementary Figures 5F, G). Further, extended MCD diet feeding in Tn housed mice resulted in significantly increased hepatic expression of Tnfa ($$p \leq 0.004$$), Ccl2 ($$p \leq 0.02$$) and Ccl3 ($$p \leq 0.01$$) (Supplementary Figures 5H–I). Despite similar absolute numbers of total hepatic immune cells (Supplementary Figure 5J), Tn housed, MCD diet fed mice, compared to Ts housed counterparts, had a significant increase in hepatic frequency of macrophage ($$p \leq 0.001$$) and neutrophil ($$p \leq 0.01$$) accrual (Supplementary Figure 5K). Surprisingly, hepatic macrophages isolated from these mice had uniquely blunted inflammatory capacity ($$p \leq 0.003$$) following ex vivo PMA/ionomycin stimulation, a finding not seen in CD4+ T cells (Supplementary Figures 5L, M). Lastly, although prolonged MCD diet feeding of mice housed at Tn increased the hepatic expression of Col1a1 ($$p \leq 0.02$$) and Col1a2 ($$p \leq 0.01$$) (Supplementary Figure 4N), such induction did not yield robust histological differences in hepatic fibrosis at the conclusion of the study (Supplementary Figure 5E).
## Tn housing restricts WD+CCl4 driven liver disease pathogenesis
Carbon tetrachloride (CCl4) is a prominent hepatotoxin known to drive robust hepatic inflammation, damage, and fibrosis [9, 40, 42, 43]. Thus, we next examined how Tn housing shapes CCl4 treatment-driven liver damage. C57BL/6 wild type mice were housed at Tn andTs and treated with CCl4 i.p. 2 times a week for 3 weeks (Supplementary Figure 6A). Both Tn and Ts housed mice treated with CCl4 had similar body weights at the conclusion of the study (Supplementary Figure 6B) and hepatic immune cell accrual that correlated with extensive hepatocellular necrosis (Supplementary Figures 6C, D, see yellow arrows). However, despite similar disease pathology, Tn housed mice treated with CCl4, compared to Ts housed counterparts, had reduced ALT levels ($$p \leq 0.001$$) (Supplementary Figure 6E). In addition, CCl4 treatment induced comparable hepatic Tnfa and Ccl2 and Ccl3 expression in both Tn and Ts housed mice (Supplementary Figures 6F, G), similar hepatic immune cell accrual (Supplementary Figures 6H, I), and hepatic immune cell inflammatory capacity following ex vivo stimulation with PMA/ionomycin (Supplementary Figures 6J, K). CCl4 is also a potent driver of hepatic fibrosis. Analysis of the whole liver tissue showed similar expression of fibrosis associated genes (Col1a1, Col1a2 and Acta2) (Supplementary Figure 6L) in Tn and Ts housed mice treated with CCl4, which also correlated with similar induction of sinusoidal fibrosis as determined by histopathological analyses (Supplementary Figure 6M).
Western Diet (WD) feeding coupled with low dose CCl4 administration (WD+CCl4) more accurately recapitulates the histopathological manifestations of NASH with progression to cirrhosis [5]. Hence, C57BL/6 wild type mice housed at Tn and Ts conditions were fed a WD coupled with low dose of CCl4 administration for 12 weeks (Figure 3A). Tn housing did not impact total body weight gain over time (Figure 3B), wet liver weights or hepatic triglyceride accumulation (Supplementary Figures 7A, B). Congruently, Tn housed mice receiving WD+CCl4, compared to Ts housed counterparts, had similar distribution of small droplet macrovesicular steatosis and large droplet macrovesicular steatosis (Figures 3C, D, see arrows). Additional histopathological analyses revealed that Tn housed, WD+CCl4 treated mice had increased hepatocellular ballooning ($$p \leq 0.04$$), but similar distribution of inflammatory foci and similar overall NAS severity (Figures 3C, E, see green arrows). Similarly, to CCl4 treatment alone, Tn housed, WD + CCl4 treated mice exhibited decreased serum ALT ($$p \leq 0.01$$) but not AST levels (Figure 3F and Supplementary Figure 7C). Tn housed, WD+CCl4 treated mice, compared to Ts housed counterparts, also had increased hepatic Tnfa, ($$p \leq 0.03$$) Ccl2 ($$p \leq 0.007$$) and Ccl3 ($$p \leq 0.02$$) expression (Supplementary Figures 7D, E). Subsequent analyses of the immune cell composition revealed Tn housing coupled to WD+CCl4 treatment decreased absolute number of hepatic immune cells ($$p \leq 0.05$$) (Supplementary Figure 7F), most prominently in the macrophage compartment ($$p \leq 0.03$$) (Figure 3G and Supplementary Table 3). Nevertheless, Tn housing did not impact macrophage inflammatory capacity but rather modified the inflammatory capacity of CD4+ T cells following ex vivo stimulation with PMA/ionomycin ($$p \leq 0.001$$) (Supplementary Figures 7G, H). We next examined if Tn housing modified WD + CCl4 treatment driven hepatic fibrosis. Increased hepatic expression of fibrosis (Col1a1, ($$p \leq 0.31$$) Col1a2 ($$p \leq 0.51$$), Acta2 ($$p \leq 0.05$$)), cell regeneration (Ccnd1 ($$p \leq 0.002$$) and Ki67 ($$p \leq 0.0005$$)) and HCC-associated genes (Afp ($$p \leq 0.009$$) and H19 ($$p \leq 0.12$$)) was observed in Tn housed, WD + CCl4 treated mice, compared to Ts housed counterparts (Supplementary Figures 7I, K). Nonetheless, robust histological differences in pericellular delicate fibrosis (see green arrows) or HCC development as indicated by similar presence of nuclear atypia (see yellow ovals) at the conclusion of the study were not observed (Supplementary Figure 7L). Lastly, as in other experimental models of NAFLD (Figures 1 and 2), Tn housing did not modify hepatic expression of insulin signaling (Pparγ and Irs1) or lipid metabolism-associated genes (Lxrα, Lipe, Chrebp, and Srebp) (Supplementary Figures 7M, N).
**Figure 3:** *Tn housing restricts WD + CCl4 driven liver disease pathogenesis. (A) Schematic of eight-week-old WT mice maintained at Ts or acclimated to Tn for 2 weeks and treated with either WD + Corn Oil (control) or WD+CCL4 2µl (0.32 µg)/g of body weight i.p. once weekly for 12 weeks. During the course of the 12 weeks, (B) body weight was recorded. At the conclusion of the study, additional parameters of NAFLD severity were analyzed. (C) The liver tissue was preserved in formalin, stained with hematoxylin eosin (H&E), and analyzed by a clinical pathologist. Table depicting histological scoring analyses for fibrosis, lobular inflammation, hepatocellular ballooning, macrovesicular (MV) steatosis grade, steatosis percentage, portal inflammation and NAFLD activity score (NAS) severity. (D) Liver H&E staining. (Top and bottom) Predominant distribution of large- and small-droplet macrovesicular steatosis is in zone 1 (to 2), grade 2 (to 3) range, apparent at low power. Black bar = 336µm (green arrows: portal areas and cholangiolar proliferations; light blue arrows: central veins). Black bar = 115µm. (E) Liver H&E staining. (Top) Cholangiolar proliferations. Periportal multilayered and solid patches (double green arrows), focal multilayered pericellular (triple arrows). Black bar = 336µm. (Bottom) Extension to lobules (single arrows), focally zone 3 (near central vein), with associated large inflammatory focus in porta (blue arrow); nuclear atypia (orange arrow). Black bar = 38µm. (F) Serum ALT levels at the conclusion of the study. (G) Hepatic inflammation as defined by flow cytometric analyses of hepatic immune cell accrual. Donut charts depicting liver immune compositions (macrophages, neutrophils, CD4+ cells, CD8+ cells, NK cells, NKT cells, B cells, and unidentified cells as determined by flow cytometry from percent of CD45+ population. In bar graphs, data represent mean +/- SEM. (B, C, F, G) Data combined from 2 individual experiments (n=10-14/condition). (C, F) One way ANOVA. *P<.05, **P<.01.*
## Discussion
In this comparative study, we built upon our previous findings that demonstrated the ability of Tn housing (which allows for physiological responses to inflammatory stimuli) to modify key inflammatory responses associated with HFD driven NAFLD development and progression. Here we add to that knowledge by showing that Tn housing also modifies liver disease hallmarks in multiple experimental mouse models of NAFLD.
HFD feeding is favored in experimental models of NAFLD as it induces robust obesity, obesity-associated metabolic sequelae, and hepatic steatosis. However, a major shortcoming of this model is that even prolonged feeding does not induce severe steatohepatitis or hepatic fibrosis [40, 44]. NASH diet is composed of high fat content coupled to high content of carbohydrates – a dietary composition that more accurately mimics human dietary intakes and is believed to more closely recapitulate NAFLD development progression. However, an apparent caveat of this model is that the higher cholesterol content is overwhelming and does not represent physiological levels found in human diets [45]. Overnutrition, commonly seen in diet induced obesity models, significantly alters cellular metabolism [46]. Altered cellular metabolism profoundly shapes immune cell function [24, 46]. Specifically, in the context of NAFLD, obesity drives ectopic fat deposition into the liver and subsequent immune cell infiltration [47, 48]. Our findings demonstrate that Tn housing was sufficient to augment obesity, ALT, hepatic immune cell accrual and the overall liver tissue damage in during NASH diet feeding. Specifically, we demonstrate that Tn housing was sufficient to not only promote increased obesity over time but to also increase hepatic accrual of macrophages and neutrophils which correlates with elevated ALT and the overall NAS severity. However, the signaling mechanisms shaping the altered NAFLD kinetics and severity under Tn housing conditions that uniquely promote increased hepatic myeloid cell accrual remain unknown. During NASH/NAFLD, liver resident macrophages (Kupffer cells) decrease and are replaced by recruited circulating monocyte-derived macrophages. These macrophages are distinct from Kupffer cells, and their presence is associated with worsened liver disease [49]. If and how Tn housing modifies the recruitment of these distinct macrophage populations and their functional capabilities is unknown. In addition, investigation of the impact of Tn housing on environmental cues or other non-immune cell factors which may contribute to worsened disease pathogenesis is warranted. Tn housing was not sufficient to induce hepatic fibrosis under NASH diet feeding conditions. Hence, future studies should focus on the role of Tn housing on activation of other liver resident cells known to contribute to fibrosis development including stellate cells [50]. Additionally, the contribution of other immune cells to hepatic inflammation and fibrosis outside the scope of this study should be investigated (51–54).
Although MCD diet induces robust hepatic steatosis, inflammation, and fibrosis, a major disadvantage of this model is that it induces chronic weight loss and lacks the presence of metabolic syndrome – parameters frequently associated with obesity-driven liver disease in humans. However, given the increase in cases of non-obese individuals with NAFLD or “lean NAFLD,” this model is again garnering more attention [55, 56]. Notably, lean (BMI<25kg/m2) and non-obese (BMI<30kg/m2) individuals with NAFLD make up $5.1\%$ and $12.1\%$ of all NAFLD cases respectively in the general population [57]. These individuals, in addition to liver disease, display worse type 2 diabetes, metabolic syndrome, cardiovascular disease, and fibrosis compared to their obese counterparts [58, 59]. Such reports seemingly uncouple obesity from NAFLD and validate the use of MCD as a tool to understand disease progression in “lean NAFLD”. We show that Tn housing in context of MCD diet was sufficient to amplify hepatic steatosis, hepatocellular ballooning, lobular inflammation, overall NAS severity, and hepatic fibrosis. Moreover, these effects were amplified under Tn conditions after prolonged exposure to MCD diet feeding. However, we did not observe differences in fibrosis during prolonged MCD diet exposure. Given that observed reduction in macrophage inflammation under these conditions, prolonged exposure to MCD-driven NAFLD milieu could contribute to immune cell exhaustion and diminished hepatic fibrosis [60]. Notably, combined deficiency of choline and methionine impairs β-oxidation and decreases secretion of VLDL contributing to increased fatty liver, cytokine secretion, inflammation, and development of fibrosis [9]. Additionally, β−oxidation regulates immune cell inflammation and function [26, 61, 62]. If and how Tn housing regulates β-oxidation, subsequent immune responses and VLDL secretion remains unknown.
Chemicals such as CCl4 are effective triggers of hepatic fibrosis and have been heavily utilized to study liver disease progression. Cellular toxicity, a major caveat of this method, is known to induce generation of reactive metabolites, ROS, oxidative stress and imbalances in cellular damage and regeneration. These processes can subsequently activate hepatic stellate cell proliferation and induce development and progression of hepatic fibrosis [42, 63]. We show that both Tn and Ts housing in the context of high dose CCl4 administration drive similar hepatocellular necrosis and fibrosis. Despite the histopathological similarities, Tn housing blunted systemic ALT level. ALT, abundantly found in the cytosol of hepatocytes, is released upon hepatocellular damage and has a half-life of approximately 47 hours in the blood, with levels varying 10-$30\%$ within a given day [64]. Whether the ALT levels are attributed to altered cycles of hepatocellular regeneration under Ts and Tn housing conditions remains unknown. Housing temperature also regulates inflammatory signaling cascades (e.g., NF-κB, TNFα IL-6) that drive the initial priming necessary to stimulate hepatocyte regeneration [65]. As such, the contribution of Tn housing to hepatocellular damage and regeneration in the context of CCl4 treatment warrants further investigation. Additionally, although ALT is often recognized as reliable marker of liver disease, ALT levels do not always directly correlate with disease progression. In the context of hepatitis B and C infection, some individuals display normal ALT levels even with the presence of advanced fibrosis [66, 67]. As such, despite the invasive nature of the procedure, liver biopsy remains the gold standard in determining NAFLD diagnosis and disease severity.
Recently, low dose CCl4 in combination with WD feeding (containing cholesterol) is utilized to induce hepatic fibrosis and mimic human NASH in C57BL/6 wild type mice – a process uncharacteristic of WD feeding alone [5]. Diet induced obesity is a known modifier of hepatic immune cell accrual including macrophages, which is further exacerbated under Tn housing conditions. Interestingly, in the context of Tn housing our data shows a reduction in total hepatic macrophage accrual in WD+CCl4 treated mice. CCl4 treatment modifies immune cell function [43, 68]. How Tn housing modulates hepatic immune cell function during WD+CCl4 combination remains to be studied. Although no differences in histological fibrosis were observed, increased expression of fibrosis, cell proliferation and HCC-associated genes under Tn conditions was noted. These data suggest that prolonged exposure may be needed to fully uncover Tn housing mediated WD+CCl4 treatment effects on fibrosis. Overall, our findings with WD+CCl4 agree with published reports [5], apart from serum ALT analysis. We show that WD+CCl4 elevates serum ALT in Ts conditions over the control group (WD+Oil). This could be due in part to differences in host gut microbiome, time of day of serum collection or hepatocellular cycles of damage and regeneration (69–71). In the context of Tn housing, we demonstrate that WD+CCl4 treatment blunts ALT levels. These data suggest that under Tn conditions the WD feeding may contribute to activation of other mechanisms subsequently offering some protection against the observed increased ALT seen in Ts housed counterparts. However, it should be noted that hepatic histopathological manifestations due to CCl4 use are not consistent with NAFLD histopathology seen in humans. Hence, these key limitations should be accounted for when using CCl4 to model NAFLD associated histopathology.
The molecular mechanisms that contribute to NAFLD progression are still under investigation. Steatosis development is due in part to both extra- and intrahepatic processes. Hormone sensitive lipase (HSL) regulates adipose tissue TG hydrolysis into free fatty acids (FFAs) which are subsequently taken up by the liver [72, 73]. Insulin inhibits HSL activity during feeding, hence during the low insulin fasting state HSL activity is increased and facilitates fatty acid release [72]. Indeed, inhibition of HSL in mice decreases plasma FFA concentration and reduces hepatic steatosis [72, 74]. Additionally, rest (i.e., non-exercise conditions) prompts fatty acid (FA) release and creates an imbalance of FAs in circulation versus their rate of oxidation. In our studies no significant changes in hepatic HSL (Lipe) or insulin signaling activity (Irs1 and Pparγ) between Ts and Tn housed groups were observed. These data suggests that Tn housing does not dominantly impact TG hydrolysis into FFA and subsequent uptake by the liver contributing to hepatic steatosis. It may be that lipoprotein lipase (LPL), which regulates intravascular hydrolysis of plasma chylomicron and very low-density lipoprotein (VLDL) into FAs for subsequent uptake by the adipose tissue and the liver, may be modified by Tn housing [72]. In fact, in experimental models of atherosclerosis Tn housing has been shown to alter low-density lipoprotein composition [15]. However, given limited changes in expression of insulin signaling genes between Ts and Tn housed groups in our models (something that promotes LPL activity [72], such studies require further investigation. Intrahepatic perturbation in lipid metabolic processes can also regulate hepatic steatosis. High sucrose feeding has been shown to promote steatosis via de novo lipogenesis. Moreover, inhibition of glucose-6-phosphatase causes hepatic entrapment of glucose and subsequently promotes de novo lipogenesis and hepatic steatosis. Indeed, carbohydrate-responsive element-binding protein (Chrebp) is activated by glucose and promotes de novo lipogenesis [75]. Yet, our data does not uncover differences in Chrebp between Ts and Tn housed groups. Analysis of other genes known to regulate lipid lipogenesis including sterol regulatory element binding protein and liver x receptors (Srebp and Lxrα) [76, 77] similarly did not reveal difference in expression between Ts and Tn housed groups. Collectively, these studies may suggest that Tn housing possibly regulates development of hepatic steatosis via novel mechanisms. Of note, dietary insults may regulate DNA methylation which heavily relies on S-adenosylmethionine (SAM) availability and methyl donors from foods. Deficiency in folate, one of such methyl donors, is known to contribute to hepatic triglyceride accumulation via regulation of fatty acid synthesis genes [78, 79]. Thus, whether Tn housing modifies diet associated regulation of epigenetic mechanisms contributing to NAFLD is underdefined.
Housing temperature exerts profound effects in shaping immune responsiveness. Cold stress is associated with norepinephrine release, activation of beta-adrenergic receptors on immune cells and subsequent regulation of immune cell function [80]. In the context of HFD driven NAFLD, Tn housing significantly lowered expression of genes central to glucocorticoid receptor (GR) and beta 3 adrenergic receptor signaling. Moreover, Tn housing resulted in lower serum concentrations of the immunosuppressive glucocorticoid corticosterone and expression of genes that inhibit inflammation (GR and beta 2 adrenergic receptor (β2AR) in the spleen [3]. Notably, immune cells deficient in GR or β2AR exhibit exacerbated inflammatory cytokine production following LPS stimulation [81, 82]. Additionally, sequencing of peripheral blood mononuclear cells from Ts and Tn housed mice showed that Tn housing promoted increased expression of genes that negatively regulate immune responses [3]. Whether the same mechanisms are conserved in other experimental models of NAFLD remains underdefined and should be examined.
Temperature modifies epigenetic regulation of immune response. It has been reported that optimally (25°C) challenged fish displayed increased IgM+ B cell secretion, macrophage inflammation and recovery following viral infection compared to sub-optimally (17°C) challenged fish. Notably, fish that survived infection during a suboptimal challenge exhibited significantly increased H3 and H4 histone modifications compared to that of optimally challenged fish. Specifically, they found that suboptimal challenge resulted in H3K9ac displaying transcriptional competency, activation of trained immunity H3K4me3, and enrichment of H3 histone-lysine 4 mono-methylation (H3K4me1), and a robust re-stimulatory immune response. Essentially, all assayed H4 modifications were significantly higher in sub-optimally challenged infected fish compared to optimally challenged infected fish. Moreover, these fish had more methylation along cytosine residues compared to optimally challenged fish, suggesting the role of epigenetics and subsequent activation of trained immunity in convalescing sub optimally challenged fish [83]. Thus, temperature may act as a potent regulator of epigenetic mechanisms contributing to regulation of immune responses during viral infection. Whether Tn housing modifies epigenetic activity in murine immune cells and their contribution to NAFLD pathogenesis represents an attractive area of investigation.
Febrile response to infection is preserved in mammals with evidence invoking existence of beneficial response to infection [84]. In fact, the available data suggests that an elevated temperature (37.5°C to 39.4°C) in ICU patients is associated with better outcomes to infectious insult compared to normothermia or hyperthermia (above 40°C) (84–86). In elderly, increased pneumonia related mortality is observed in those who lacked fever ($29\%$) compared to those who developed a febrile response ($4\%$). Thus, tese data suggests that internal temperatures may indeed impact the immune response or slow pathogen virulence. However, there is insufficient data to determine the impact of environmental temperature on modulation of immune responses and development of disease, specifically NAFLD, under steady state conditions in humans. Importantly, divergence in dietary consumption and lifestyle among various populations, given its relevance to NAFLD development, represents a key obstacle in addressing the impact of temperature on NAFLD severity in humans. Hence, additional epidemiological studies are required across various geographical landscapes to begin to draw correlations between NAFLD prevalence and climate.
In summary, our data demonstrates that Tn housing modifies hepatic immune cell accrual, lobular inflammation, fibrosis and overall amplifies liver tissue damage severity in experimental models of NAFLD in mice. Specifically, we demonstrate that Tn housing modifies hepatic immune cell inflammation across all models studied. Further investigation is required to complete in depth characterization of immune cell subsets and their genomic and transcriptional landscapes. These data are relevant for future comparisons of immune response in experimental models with human disease. Thus, the utility of thermoneutral housing as a means of modeling physiological murine immune responses to various inflammatory stimuli is becoming more recognized. Together, the initial findings of our studies when coupled to future investigation on this topic might serve as a foundation for interrogating how Tn housing instructs immune cell function in the context of liver inflammation and disease in multiple settings. Collectively, these insights may allow for the development of future therapies to NAFLD via improved understandings of immune mechanisms underlying disease development and progression.
## 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 Cincinnati Children’s Hospital Medical Center IACUC.
## Author contributions
JO, MM-F, DG and SD contributed to conception and design of the study. JO, KS, DG, PA, MD, MM-F, and TS executed the experimental plan and contributed to subsequent data analyses. JO, KS, and DG performed the data and statistical analyses. SS performed histological scoring and analyses. JO, KS, MM-F and SD wrote the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1095132/full#supplementary-material
## References
1. Cheemerla S, Balakrishnan M. **Global epidemiology of chronic liver disease**. *Clin Liver Dis (Hoboken).* (2021) **17**. DOI: 10.1002/cld.1061
2. Dyson JK, Anstee QM, McPherson S. **Non-alcoholic fatty liver disease: a practical approach to diagnosis and staging**. *Frontline Gastroenterol* (2014) **5**. DOI: 10.1136/flgastro-2013-100403
3. Giles DA, Moreno-Fernandez ME, Stankiewicz TE, Graspeuntner S, Cappelletti M, Wu D. **Thermoneutral housing exacerbates nonalcoholic fatty liver disease in mice and allows for sex-independent disease modeling**. *Nat Med* (2017) **23**. DOI: 10.1038/nm.4346
4. Rinella ME, Elias MS, Smolak RR, Fu T, Borensztajn J, Green RM. **Mechanisms of hepatic steatosis in mice fed a lipogenic methionine choline-deficient diet**. *J Lipid Res* (2008) **49**. DOI: 10.1194/jlr.M800042-JLR200
5. Tsuchida T, Lee YA, Fujiwara N, Ybanez M, Allen B, Martins S. **A simple diet- and chemical-induced murine NASH model with rapid progression of steatohepatitis, fibrosis and liver cancer**. *J Hepatol* (2018) **69**. DOI: 10.1016/j.jhep.2018.03.011
6. Boll M, Weber LW, Becker E, Stampfl A. **Mechanism of carbon tetrachloride-induced hepatotoxicity. hepatocellular damage by reactive carbon tetrachloride metabolites**. *Z Naturforsch C J Biosci* (2001) **56**. DOI: 10.1515/znc-2001-7-826
7. Ito M, Suzuki J, Tsujioka S, Sasaki M, Gomori A, Shirakura T. **Longitudinal analysis of murine steatohepatitis model induced by chronic exposure to high-fat diet**. *Hepatol Res* (2007) **37**. DOI: 10.1111/j.1872-034X.2007.00008.x
8. Christ A, Lauterbach M, Latz E. **Western Diet and the immune system: An inflammatory connection**. *Immunity* (2019) **51** 794-811. DOI: 10.1016/j.immuni.2019.09.020
9. Van Herck MA, Vonghia L, Francque SM. **Animal models of nonalcoholic fatty liver disease-a starter's guide**. *Nutrients* (2017) **9**. DOI: 10.3390/nu9101072
10. Itagaki H, Shimizu K, Morikawa S, Ogawa K, Ezaki T. **Morphological and functional characterization of non-alcoholic fatty liver disease induced by a methionine-choline-deficient diet in C57BL/6 mice**. *Int J Clin Exp Pathol* (2013) **6**
11. Hirsova P, Bamidele AO, Wang H, Povero D, Revelo XS. **Emerging roles of T cells in the pathogenesis of nonalcoholic steatohepatitis and hepatocellular carcinoma**. *Front Endocrinol (Lausanne).* (2021) **12**. DOI: 10.3389/fendo.2021.760860
12. Huby T, Gautier EL. **Immune cell-mediated features of non-alcoholic steatohepatitis**. *Nat Rev Immunol* (2022) **22**. DOI: 10.1038/s41577-021-00639-3
13. Oates JR, McKell MC, Moreno-Fernandez ME, Damen M, Deepe GS, Qualls JE. **Macrophage function in the pathogenesis of non-alcoholic fatty liver disease: The mac attack**. *Front Immunol* (2019) **10**. DOI: 10.3389/fimmu.2019.02893
14. Wan M, Han J, Ding L, Hu F, Gao P. **Novel immune subsets and related cytokines: Emerging players in the progression of liver fibrosis**. *Front Med (Lausanne).* (2021) **8**. DOI: 10.3389/fmed.2021.604894
15. Giles DA, Ramkhelawon B, Donelan EM, Stankiewicz TE, Hutchison SB, Mukherjee R. **Modulation of ambient temperature promotes inflammation and initiates atherosclerosis in wild type C57BL/6 mice**. *Mol Metab* (2016) **5**. DOI: 10.1016/j.molmet.2016.09.008
16. Kokolus KM, Capitano ML, Lee CT, Eng JW, Waight JD, Hylander BL. **Baseline tumor growth and immune control in laboratory mice are significantly influenced by subthermoneutral housing temperature**. *Proc Natl Acad Sci U S A.* (2013) **110**. DOI: 10.1073/pnas.1304291110
17. Tian XY, Ganeshan K, Hong C, Nguyen KD, Qiu Y, Kim J. **Thermoneutral housing accelerates metabolic inflammation to potentiate atherosclerosis but not insulin resistance**. *Cell Metab* (2016) **23**. DOI: 10.1016/j.cmet.2015.10.003
18. Bowers SL, Bilbo SD, Dhabhar FS, Nelson RJ. **Stressor-specific alterations in corticosterone and immune responses in mice**. *Brain Behav Immun* (2008) **22**. DOI: 10.1016/j.bbi.2007.07.012
19. Ganeshan K, Chawla A. **Warming the mouse to model human diseases**. *Nat Rev Endocrinol* (2017) **13**. DOI: 10.1038/nrendo.2017.48
20. Bond LM, Burhans MS, Ntambi JM. **Uncoupling protein-1 deficiency promotes brown adipose tissue inflammation and ER stress**. *PloS One* (2018) **13**. DOI: 10.1371/journal.pone.0205726
21. Gordon CJ. **The mouse thermoregulatory system: Its impact on translating biomedical data to humans**. *Physiol Behav* (2017) **179** 55-66. DOI: 10.1016/j.physbeh.2017.05.026
22. Liu E, Lewis K, Al-Saffar H, Krall CM, Singh A, Kulchitsky VA. **Naturally occurring hypothermia is more advantageous than fever in severe forms of lipopolysaccharide- and escherichia coli-induced systemic inflammation**. *Am J Physiol Regul Integr Comp Physiol* (2012) **302**. DOI: 10.1152/ajpregu.00023.2012
23. Giles DA, Moreno-Fernandez ME, Stankiewicz TE, Cappelletti M, Huppert SS, Iwakura Y. **Regulation of inflammation by IL-17A and IL-17F modulates non-alcoholic fatty liver disease pathogenesis**. *PloS One* (2016) **11**. DOI: 10.1371/journal.pone.0149783
24. Moreno-Fernandez ME, Giles DA, Oates JR, Chan CC, Damen M, Doll JR. **PKM2-dependent metabolic skewing of hepatic Th17 cells regulates pathogenesis of non-alcoholic fatty liver disease**. *Cell Metab* (2021) **33** 1187-204.e9. DOI: 10.1016/j.cmet.2021.04.018
25. Harley IT, Stankiewicz TE, Giles DA, Softic S, Flick LM, Cappelletti M. **IL-17 signaling accelerates the progression of nonalcoholic fatty liver disease in mice**. *Hepatology* (2014) **59**. DOI: 10.1002/hep.26746
26. Moreno-Fernandez ME, Giles DA, Stankiewicz TE, Sheridan R, Karns R, Cappelletti M. **Peroxisomal beta-oxidation regulates whole body metabolism, inflammatory vigor, and pathogenesis of nonalcoholic fatty liver disease**. *JCI Insight* (2018) **3**. DOI: 10.1172/jci.insight.93626
27. Boland ML, Oro D, Tolbol KS, Thrane ST, Nielsen JC, Cohen TS. **Towards a standard diet-induced and biopsy-confirmed mouse model of non-alcoholic steatohepatitis: Impact of dietary fat source**. *World J Gastroenterol* (2019) **25**. DOI: 10.3748/wjg.v25.i33.4904
28. Drescher HK, Weiskirchen R, Fulop A, Hopf C, de San Roman EG, Huesgen PF. **The influence of different fat sources on steatohepatitis and fibrosis development in the Western diet mouse model of non-alcoholic steatohepatitis (NASH)**. *Front Physiol* (2019) **10**. DOI: 10.3389/fphys.2019.00770
29. Kawashita E, Ishihara K, Nomoto M, Taniguchi M, Akiba S. **A comparative analysis of hepatic pathological phenotypes in C57BL/6J and C57BL/6N mouse strains in non-alcoholic steatohepatitis models**. *Sci Rep* (2019) **9** 204. DOI: 10.1038/s41598-018-36862-7
30. Kakino S, Ohki T, Nakayama H, Yuan X, Otabe S, Hashinaga T. **Pivotal role of TNF-alpha in the development and progression of nonalcoholic fatty liver disease in a murine model**. *Horm Metab Res* (2018) **50**. DOI: 10.1055/s-0043-118666
31. Seo YY, Cho YK, Bae JC, Seo MH, Park SE, Rhee EJ. **Tumor necrosis factor-alpha as a predictor for the development of nonalcoholic fatty liver disease: A 4-year follow-up study**. *Endocrinol Metab (Seoul).* (2013) **28**. DOI: 10.3803/EnM.2013.28.1.41
32. Koyama Y, Brenner DA. **Liver inflammation and fibrosis**. *J Clin Invest.* (2017) **127** 55-64. DOI: 10.1172/JCI88881
33. Seki E, Schwabe RF. **Hepatic inflammation and fibrosis: Functional links and key pathways**. *Hepatology* (2015) **61**. DOI: 10.1002/hep.27332
34. Im YR, Hunter H, de Gracia Hahn D, Duret A, Cheah Q, Dong J. **A systematic review of animal models of NAFLD finds high-fat, high-fructose diets most closely resemble human NAFLD**. *Hepatology* (2021) **74**. DOI: 10.1002/hep.31897
35. Teufel A, Itzel T, Erhart W, Brosch M, Wang XY, Kim YO. **Comparison of gene expression patterns between mouse models of nonalcoholic fatty liver disease and liver tissues from patients**. *Gastroenterology* (2016) **151** 513-25.e0. DOI: 10.1053/j.gastro.2016.05.051
36. Kumashiro N, Erion DM, Zhang D, Kahn M, Beddow SA, Chu X. **Cellular mechanism of insulin resistance in nonalcoholic fatty liver disease**. *Proc Natl Acad Sci U S A.* (2011) **108**. DOI: 10.1073/pnas.1113359108
37. Pei K, Gui T, Kan D, Feng H, Jin Y, Yang Y. **An overview of lipid metabolism and nonalcoholic fatty liver disease**. *BioMed Res Int* (2020) **2020** 4020249. DOI: 10.1155/2020/4020249
38. Kim D, Kim WR. **Nonobese fatty liver disease**. *Clin Gastroenterol Hepatol* (2017) **15**. DOI: 10.1016/j.cgh.2016.08.028
39. Leclere PS, Rousseau D, Patouraux S, Guerin S, Bonnafous S, Grechez-Cassiau A. **MCD diet-induced steatohepatitis generates a diurnal rhythm of associated biomarkers and worsens liver injury in Klf10 deficient mice**. *Sci Rep* (2020) **10** 12139. DOI: 10.1038/s41598-020-69085-w
40. Machado MV, Michelotti GA, Xie G, Almeida Pereira T, Boursier J, Bohnic B. **Mouse models of diet-induced nonalcoholic steatohepatitis reproduce the heterogeneity of the human disease**. *PloS One* (2015) **10**. DOI: 10.1371/journal.pone.0127991
41. Jahn D, Kircher S, Hermanns HM, Geier A. **Animal models of NAFLD from a hepatologist's point of view**. *Biochim Biophys Acta Mol Basis Dis* (2019) **1865**. DOI: 10.1016/j.bbadis.2018.06.023
42. Dong S, Chen QL, Song YN, Sun Y, Wei B, Li XY. **Mechanisms of CCl4-induced liver fibrosis with combined transcriptomic and proteomic analysis**. *J Toxicol Sci* (2016) **41**. DOI: 10.2131/jts.41.561
43. Ikeno Y, Ohara D, Takeuchi Y, Watanabe H, Kondoh G, Taura K. **Foxp3+ regulatory T cells inhibit CCl4-induced liver inflammation and fibrosis by regulating tissue cellular immunity**. *Front Immunol* (2020) **11**. DOI: 10.3389/fimmu.2020.584048
44. Charlton M, Krishnan A, Viker K, Sanderson S, Cazanave S, McConico A. **Fast food diet mouse: novel small animal model of NASH with ballooning, progressive fibrosis, and high physiological fidelity to the human condition**. *Am J Physiol Gastrointest Liver Physiol* (2011) **301**. DOI: 10.1152/ajpgi.00145.2011
45. Wouters K, van Gorp PJ, Bieghs V, Gijbels MJ, Duimel H, Lutjohann D. **Dietary cholesterol, rather than liver steatosis, leads to hepatic inflammation in hyperlipidemic mouse models of nonalcoholic steatohepatitis**. *Hepatology* (2008) **48**. DOI: 10.1002/hep.22363
46. Alwarawrah Y, Kiernan K, MacIver NJ. **Changes in nutritional status impact immune cell metabolism and function**. *Front Immunol* (2018) **9**. DOI: 10.3389/fimmu.2018.01055
47. Kuang M, Lu S, Xie Q, Peng N, He S, Yu C. **Abdominal obesity phenotypes are associated with the risk of developing non-alcoholic fatty liver disease: Insights from the general population**. *BMC Gastroenterol* (2022) **22** 311. DOI: 10.1186/s12876-022-02393-9
48. Van Herck MA, Weyler J, Kwanten WJ, Dirinck EL, De Winter BY, Francque SM. **The differential roles of T cells in non-alcoholic fatty liver disease and obesity**. *Front Immunol* (2019) **10**. DOI: 10.3389/fimmu.2019.00082
49. Seidman JS, Troutman TD, Sakai M, Gola A, Spann NJ, Bennett H. **Niche-specific reprogramming of epigenetic landscapes drives myeloid cell diversity in nonalcoholic steatohepatitis**. *Immunity* (2020) **52** 1057-74.e7. DOI: 10.1016/j.immuni.2020.04.001
50. Zhang CY, Yuan WG, He P, Lei JH, Wang CX. **Liver fibrosis and hepatic stellate cells: Etiology, pathological hallmarks and therapeutic targets**. *World J Gastroenterol* (2016) **22**. DOI: 10.3748/wjg.v22.i48.10512
51. Bonecchi R, Facchetti F, Dusi S, Luini W, Lissandrini D, Simmelink M. **Induction of functional IL-8 receptors by IL-4 and IL-13 in human monocytes**. *J Immunol* (2000) **164**. DOI: 10.4049/jimmunol.164.7.3862
52. Connolly MK, Bedrosian AS, Mallen-St Clair J, Mitchell AP, Ibrahim J, Stroud A. **In liver fibrosis, dendritic cells govern hepatic inflammation in mice**. *J Clin Invest.* (2009) **119**. DOI: 10.1172/JCI37581
53. Novobrantseva TI, Majeau GR, Amatucci A, Kogan S, Brenner I, Casola S. **Attenuated liver fibrosis in the absence of b cells**. *J Clin Invest.* (2005) **115**. DOI: 10.1172/JCI24798
54. Xu R, Zhang Z, Wang FS. **Liver fibrosis: mechanisms of immune-mediated liver injury**. *Cell Mol Immunol* (2012) **9** 296-301. DOI: 10.1038/cmi.2011.53
55. VanWagner LB, Armstrong MJ. **Lean NAFLD: A not so benign condition**. *Hepatol Commun* (2018) **2** 5-8. DOI: 10.1002/hep4.1143
56. Younossi ZM, Stepanova M, Negro F, Hallaji S, Younossi Y, Lam B. **Nonalcoholic fatty liver disease in lean individuals in the united states**. *Med (Baltimore).* (2012) **91**. DOI: 10.1097/MD.0b013e3182779d49
57. Chrysavgis L, Ztriva E, Protopapas A, Tziomalos K, Cholongitas E. **Nonalcoholic fatty liver disease in lean subjects: Prognosis, outcomes and management**. *World J Gastroenterol* (2020) **26**. DOI: 10.3748/wjg.v26.i42.6514
58. Ye Q, Zou B, Yeo YH, Li J, Huang DQ, Wu Y. **Global prevalence, incidence, and outcomes of non-obese or lean non-alcoholic fatty liver disease: A systematic review and meta-analysis**. *Lancet Gastroenterol Hepatol* (2020) **5**. DOI: 10.1016/S2468-1253(20)30077-7
59. Zou B, Yeo YH, Nguyen VH, Cheung R, Ingelsson E, Nguyen MH. **Prevalence, characteristics and mortality outcomes of obese, nonobese and lean NAFLD in the united states, 1999-2016**. *J Intern Med* (2020) **288**. DOI: 10.1111/joim.13069
60. Pradhan K, Yi Z, Geng S, Li L. **Development of exhausted memory monocytes and underlying mechanisms**. *Front Immunol* (2021) **12**. DOI: 10.3389/fimmu.2021.778830
61. Cucchi D, Camacho-Munoz D, Certo M, Pucino V, Nicolaou A, Mauro C. **Fatty acids - from energy substrates to key regulators of cell survival, proliferation and effector function**. *Cell Stress.* (2019) **4** 9-23. DOI: 10.15698/cst2020.01.209
62. O'Sullivan D, van der Windt GJ, Huang SC, Curtis JD, Chang CH, Buck MD. **Memory CD8(+) T cells use cell-intrinsic lipolysis to support the metabolic programming necessary for development**. *Immunity* (2014) **41** 75-88. DOI: 10.1016/j.immuni.2014.06.005
63. Fujii T, Fuchs BC, Yamada S, Lauwers GY, Kulu Y, Goodwin JM. **Mouse model of carbon tetrachloride induced liver fibrosis: Histopathological changes and expression of CD133 and epidermal growth factor**. *BMC Gastroenterol* (2010) **10** 79. DOI: 10.1186/1471-230X-10-79
64. Kim WR, Flamm SL, Di Bisceglie AM, Bodenheimer HC. **Public policy committee of the American association for the study of liver d. serum activity of alanine aminotransferase (ALT) as an indicator of health and disease**. *Hepatology* (2008) **47**. DOI: 10.1002/hep.22109
65. Abu Rmilah A, Zhou W, Nelson E, Lin L, Amiot B, Nyberg SL. **Understanding the marvels behind liver regeneration**. *Wiley Interdiscip Rev Dev Biol* (2019) **8**. DOI: 10.1002/wdev.340
66. Chao DT, Lim JK, Ayoub WS, Nguyen LH, Nguyen MH. **Systematic review with meta-analysis: the proportion of chronic hepatitis b patients with normal alanine transaminase </= 40 IU/L and significant hepatic fibrosis**. *Aliment Pharmacol Ther* (2014) **39**. DOI: 10.1111/apt.12590
67. Sanai FM, Benmousa A, Al-Hussaini H, Ashraf S, Alhafi O, Abdo AA. **Is serum alanine transaminase level a reliable marker of histological disease in chronic hepatitis c infection**. *Liver Int* (2008) **28**. DOI: 10.1111/j.1478-3231.2008.01733.x
68. Jirova D, Sperlingova I, Halaskova M, Bendova H, Dabrowska L. **Immunotoxic effects of carbon tetrachloride–the effect on morphology and function of the immune system in mice**. *Cent Eur J Public Health* (1996) **4** 16-20. PMID: 8996663
69. Cordoba J, O'Riordan K, Dupuis J, Borensztajin J, Blei AT. **Diurnal variation of serum alanine transaminase activity in chronic liver disease**. *Hepatology* (1998) **28**. DOI: 10.1002/hep.510280640
70. Kim MS, Lee HK, Kim SY, Cho JH. **Analysis of the relationship between liver regeneration rate and blood levels**. *Pak J Med Sci* (2015) **31**. DOI: 10.12669/pjms.311.5864
71. Schwenger KJ, Clermont-Dejean N, Allard JP. **The role of the gut microbiome in chronic liver disease: The clinical evidence revised**. *JHEP Rep* (2019) **1**. DOI: 10.1016/j.jhepr.2019.04.004
72. den Boer M, Voshol PJ, Kuipers F, Havekes LM, Romijn JA. **Hepatic steatosis: a mediator of the metabolic syndrome. lessons from animal models**. *Arterioscler Thromb Vasc Biol* (2004) **24**. DOI: 10.1161/01.ATV.0000116217.57583.6e
73. Holm C. **Molecular mechanisms regulating hormone-sensitive lipase and lipolysis**. *Biochem Soc Trans* (2003) **31**. DOI: 10.1042/bst0311120
74. Voshol PJ, Haemmerle G, Ouwens DM, Zimmermann R, Zechner R, Teusink B. **Increased hepatic insulin sensitivity together with decreased hepatic triglyceride stores in hormone-sensitive lipase-deficient mice**. *Endocrinology* (2003) **144**. DOI: 10.1210/en.2002-0036
75. 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**. DOI: 10.3389/fendo.2020.587189
76. Eberle D, Hegarty B, Bossard P, Ferre P, Foufelle F. **SREBP transcription factors: master regulators of lipid homeostasis**. *Biochimie* (2004) **86**. DOI: 10.1016/j.biochi.2004.09.018
77. Lee SD, Tontonoz P. **Liver X receptors at the intersection of lipid metabolism and atherogenesis**. *Atherosclerosis* (2015) **242** 29-36. DOI: 10.1016/j.atherosclerosis.2015.06.042
78. Kalhan SC, Edmison J, Marczewski S, Dasarathy S, Gruca LL, Bennett C. **Methionine and protein metabolism in non-alcoholic steatohepatitis: evidence for lower rate of transmethylation of methionine**. *Clin Sci (Lond).* (2011) **121**. DOI: 10.1042/CS20110060
79. Niculescu MD, Zeisel SH. **Diet, methyl donors and DNA methylation: interactions between dietary folate, methionine and choline**. *J Nutr* (2002) **132** 8 Suppl. DOI: 10.1093/jn/132.8.2333S
80. Elenkov IJ, Wilder RL, Chrousos GP, Vizi ES. **The sympathetic nerve–an integrative interface between two supersystems: the brain and the immune system**. *Pharmacol Rev* (2000) **52** 595-638. PMID: 11121511
81. Bhattacharyya S, Brown DE, Brewer JA, Vogt SK, Muglia LJ. **Macrophage glucocorticoid receptors regulate toll-like receptor 4-mediated inflammatory responses by selective inhibition of p38 MAP kinase**. *Blood* (2007) **109**. DOI: 10.1182/blood-2006-10-048215
82. Izeboud CA, Mocking JA, Monshouwer M, van Miert AS. **Witkamp RF. participation of beta-adrenergic receptors on macrophages in modulation of LPS-induced cytokine release**. *J Recept Signal Transduct Res* (1999) **19** 191-202. PMID: 10071758
83. Krishnan R, Jang YS, Kim JO, Yoon SY, Rajendran R, Oh MJ. **Temperature dependent cellular, and epigenetic regulatory mechanisms underlying the antiviral immunity in sevenband grouper to nervous necrosis virus infection**. *Fish Shellfish Immunol* (2022) **131** 898-907. DOI: 10.1016/j.fsi.2022.10.068
84. Walter EJ, Hanna-Jumma S, Carraretto M, Forni L. **The pathophysiological basis and consequences of fever**. *Crit Care* (2016) **20** 200. DOI: 10.1186/s13054-016-1375-5
85. Lee BH, Inui D, Suh GY, Kim JY, Kwon JY, Park J. **Association of body temperature and antipyretic treatments with mortality of critically ill patients with and without sepsis: Multi-centered prospective observational study**. *Crit Care* (2012) **16** R33. DOI: 10.1186/cc11660
86. Young PJ, Saxena M, Beasley R, Bellomo R, Bailey M, Pilcher D. **Early peak temperature and mortality in critically ill patients with or without infection**. *Intensive Care Med* (2012) **38**. DOI: 10.1186/cc10393
|
---
title: 'Mind-stimulating leisure activities: Prospective associations with health,
wellbeing, and longevity'
authors:
- Dorota Weziak-Bialowolska
- Piotr Bialowolski
- Pier Luigi Sacco
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9982162
doi: 10.3389/fpubh.2023.1117822
license: CC BY 4.0
---
# Mind-stimulating leisure activities: Prospective associations with health, wellbeing, and longevity
## Abstract
### Introduction
This study examines prospective associations within a 6-year perspective between three mind-stimulating leisure activities (relaxed and solitary: reading; serious and solitary: doing number and word games; serious and social: playing cards and games) and 21 outcomes in [1] physical health, [2] wellbeing, [3] daily life functioning, [4] cognitive impairment, and [5] longevity domains.
### Methods
Data were obtained from 19,821 middle-aged and older adults from 15 countries participating in the Survey of Health, Ageing, and Retirement in Europe (SHARE). Temporal associations were obtained using generalized estimating equations. All models were controlled for prior sociodemographic, personality, lifestyle factors, health behaviors, and pre-baseline leisure activity values and all outcome variables. The Bonferroni correction was used to correct for multiple testing. E-values were calculated to examine the sensitivity of the associations to unmeasured confounding. Secondary analyses [1] under the complete case scenario, [2] after excluding respondents with health conditions, and [3] using a limited set of covariates were conducted to provide evidence for the robustness of the results.
### Results
The relaxed solitary activity of reading almost daily was prospectively associated with a lower risk of depression, experiencing pain, daily functioning limitations, cognitive impairment, lower loneliness scores, and more favorable wellbeing outcomes. Engaging in serious solitary leisure activities almost daily was prospectively associated with a lower risk of depression, feeling full of energy, and a lower risk of death by any cause. Occasionally engaging in these activities was prospectively associated with greater optimism and a lower risk of cognitive impairment. Engaging in serious social activities was prospectively associated with greater happiness, lower scores on the loneliness scale, a lower risk of Alzheimer's disease, and an increased risk of cancer. Additionally, occasionally engaging in serious social activities was associated with greater optimism and lower risk of depression, pain, and mobility limitations. These associations were independent of demographics, socioeconomic status, personality, history of diseases, and prior lifestyle. The sensitivity analyses provided substantial evidence for the robustness of these associations.
### Discussion
Mind-engaging leisure activities can be considered a health and wellbeing resource. Practitioners may consider them tools that help middle-aged and older adults maintain their health and quality of life.
## 1. Introduction
Humans seek leisure. It can be considered a complex human need with various adaptive purposes [1]. These include improving physiological and psychological functioning [2], enjoying entertainment and pleasant moments [3], seeking distraction from daily concerns and stressful work [4], spending time with family, friends, and acquaintances [5, 6], coping with social crises [7], and even informal learning [8]. In addition, leisure seems to constitute one of the most fundamental human behaviors and a part of our daily routines since the early stages of human evolution [9, 10].
The concept of leisure comprises a set of heterogeneous activities that can differ in terms of the intensity of physical activity, degree of sociability, and level of intellectual effort (11–13). Special attention has been paid to the role of leisure studies in improving and maintaining good health and wellbeing (13–18). Various leisure activities have been systematically investigated concerning their salutogenic effects. For example, prior studies have shown that arts and cultural activities may reduce the allostatic load by reducing unhealthy habits such as smoking and alcohol consumption [19] and directly improve hedonic wellbeing [17]. Leisure activities that require physical exercise reduce the risk of cardiovascular mortality regardless of age, sex, and pre-existing diseases [20, 21]. Social leisure activities improve emotional wellbeing and reduce depression and anxiety symptoms [17], especially among socially deprived participants [22]. Despite numerous other examples, the pathways through which various leisure activities nurture beneficial effects on health and wellbeing remain unclear. Fancourt et al. [ 13] provided the first systematic survey of the types of effects and possible pathways, indicating that much more research is needed to understand the actual mechanisms thoroughly.
Therefore, it is necessary to examine the potential health and wellbeing benefits associated with different forms of leisure that have not received sufficient scientific attention so far. Consequently, the present study considers a particular class of leisure activities, namely, mind-engaging leisure activities. These may include reading, word and number games, playing cards or chess, or similar activities. One may question this choice of leisure activities arguing that mind games are merely sedentary activities, which, as part of sedentary lifestyles, have contributed to the worsening of population health and led to adverse health outcomes (23–25). However, in addition to the well-known positive role of physical exercise on health and prevention of chronic diseases (26–29) and cognitive functioning in older age [30], sedentary activities that keep the mind engaged have also been praised in similar contexts [31].
Regardless, evidence of the effects of various types of mind-stimulating leisure activities on health remains limited, with some preliminary results indicating a positive impact on mental health [32]. Consequently, we examined three categories of cognitive leisure activities that involve brain exercises and mind games in this study: [1] doing word or number games such as crossword puzzles or Sudoku; [2] playing cards or games such as chess; and [3] reading books, magazines, and newspapers. These activities reflect the classical partitioning of leisure activities into relaxed, serious, and social [13, 33]. Serious leisure activities are usually problem-solving-oriented and cognitively engaging mind games. Relaxed leisure activities are non-problem-solving-oriented, although they can still be cognitively engaging. Among these two groups, we can further distinguish between solitary and social forms of activity.
Consequently, we classify word or number games—such as crossword puzzles or Sudoku—as serious and solitary leisure activities. Playing social games such as cards or chess is classified as a serious and social leisure activity. Lastly, reading books, magazines, or newspapers is classified as a relaxed and solitary leisure activity. The first two types of activities involve mind games comprising directed cognitive effort, with the former being solitary and the latter involving a social component.
Reading is engaging, and emotionally and cognitively arousing. However, the reader has no specific goal, contrary to the gamer who wants to win. This makes reading an especially interesting example of a relaxing activity in which one pursues narrative immersion, enjoyment, information gathering, and more, without any specific performance check. Instead, mind games generate a serious dimension of experience that relates to testing one's intellectual abilities on a particular problem that may or may not be solved, which can be engaging and pleasant but not necessarily relaxing. In solitary mind games, the performance check is individual, while in social mind games it is relational, resulting in competition with others. The comparison of these three types of activities with their peculiarities in terms of their effects on health and wellbeing has, to our knowledge, not been explored in the literature before. Therefore, it is essential to understand the detailed mechanisms through which certain leisure activities may benefit individuals undertaking them.
This study examined prospective associations between the frequency of the three mind-stimulating leisure activities and physical health, emotional wellbeing, cognitive impairment, daily life functioning outcomes, and all-cause mortality. We also addressed two research questions: First, what changes in physical health, emotional wellbeing, cognitive impairment, and daily life functioning outcomes could be observed within 6 years, if people engage in any of the three examined leisure activities classes? Second, what changes in longevity may be observed when people engage in any of the leisure activities? We were particularly interested in studying these effects in middle-aged and older adults.
## 2.1. Study population
Longitudinal data from the Survey of Health, Ageing, and Retirement in Europe (SHARE) [34] were used. Detailed information regarding the data sources is provided in the Supplementary material. This study used data from 2011 to 2020. The analytical sample included middle-aged and older adults aged 50 years or older who participated in Waves 4, 5, and 8. A total of 19,821 participants met the inclusion criteria. In the analysis of all-cause mortality, 39,009 middle-aged and older adults who participated in Waves 4 and 5 were observed for up to 8 years. They were from 15 European countries: Austria, Germany, Sweden, the Netherlands, Spain, Italy, France, Denmark, Switzerland, Belgium, Czech Republic, Poland, Hungary, Slovenia, and Estonia.
## 2.2.1. Leisure activities
The relaxed solitary leisure activity—reading—was assessed by asking respondents about the frequency with which they read books, magazines, and newspapers (almost every day, sometimes, and never). The frequency of serious solitary leisure activities involving playing word or number games, such as crossword puzzles or Sudoku, was also examined (almost every day, sometimes, and never). The frequency of serious social activity involving playing cards or games such as chess was also considered (almost every day, sometimes, and never).
## 2.2.2. Outcomes
Health and wellbeing outcomes such as [1] a sense of loneliness; [2] depression; [3] self-reported diagnosis of Alzheimer's disease, dementia, or other serious memory impairment; [4] sense of meaning in life; [5] happiness; [6] feeling energetic; and [7] optimism. We examined daily life functioning using instruments measuring the level of difficulty in activities of daily living (ADL) and instrumental activities of daily living (IADL) due to physical, mental, emotional, and memory problems. The following physical health outcomes were examined: heart attack, hypertension, high blood cholesterol, stroke, diabetes, and cancer. We accounted for both non-fatal and fatal conditions. We also considered the presence of impaired pain and whether the respondents experienced at least one mobility, arm, or fine-motor limitation. Cognitive impairment was assessed using a measure of time orientation. All-cause mortality was also considered an outcome. Detailed information regarding the measures used is presented in the Supplementary material.
## 2.2.3. Covariates
All covariates were self-reported and measured during the pre-baseline (Wave 4). These included demographic characteristics (gender, age, marital status, education, and country), socioeconomic factors (income and wealth), personality traits (agreeableness, openness, neuroticism, conscientiousness, and extraversion), health behaviors (sports activity, alcohol consumption, and BMI), and lifestyle factors (volunteering). Detailed information regarding the covariates used is presented in the Supplementary material.
## 2.2.4. Prior values of outcomes and exposure
To reduce the possibility of reverse causation and residual confounding, we adjusted for the prior values of the 21 outcome variables (i.e., prior emotional wellbeing, daily life functioning, cognitive impairment, and history of diseases). To further reduce the risk of reverse causality, we controlled for prior values of the respective exposure variable (i.e., mind-stimulating leisure activity).
## 2.3. Statistical analysis
We conducted an outcome-wide analysis [35]. In this analysis, 21 outcomes were used to extensively examine the pattern of temporal associations with leisure activities (at a 6-year follow-up). Previous studies have shown this methodology to be useful in limiting the risk of preferring only significant results and salami-slicing, as well as revealing patterns of associations that may not be apparent if a single outcome was examined (36–40). Each prospective association was modeled using generalized estimating equations. We clustered by country and adjusted standard errors to account for the hierarchical nature of the data. Three types of estimates are reported. Standardized regression estimates were used for continuous outcomes. For dichotomous outcomes, odds ratios for rare outcomes (i.e., occurring in <$10\%$ of the population) and risk ratios for non-rare outcomes (i.e., occurring in at least $10\%$ of the population) were indicated. Poisson regression with robust standard errors was used to estimate risk ratios [41, 42], and the Bonferroni correction was used to correct for multiple testing. All missing covariates, exposure, and outcome variables were imputed using chained equations [10 sets of imputed data were generated [43]] and multiple imputation estimates were pooled using the Rubin rule [44].
A series of robustness checks was carried out. First, the robustness of the results was examined using E-values. This sensitivity measure assesses the extent to which a potential uncontrolled confounder would need to be associated with both the exposure and outcome to explain the observed association [45]. Second, all models were rerun after excluding anyone with a history of a given physical condition at the pre-baseline. Third, the primary set of models was reanalyzed using complete case analysis to assess the robustness of the results to missing data patterns. Finally, the primary models were rerun with a limited set of controls (i.e., only demographic and socioeconomic control variables traditionally used in similar analyses), as there was a risk that with such an extensive set of covariates, the models could have been overfitted. All statistical analyses were performed using Stata/SE 17.0.
## 3.1. Descriptive analyses
In the pre-baseline Wave (Wave 4), participants were, on average, 64.5 (SD = 8.72) years old, mostly women ($59.3\%$) and married ($69.6\%$), with upper secondary ($36.1\%$) or first-stage tertiary education ($22.2\%$). More than $66\%$ of participants reported reading books, magazines, and newspapers almost daily, $12.2\%$ sometimes, and $21.3\%$ never. More than $25\%$ of middle-aged and older adults engaged with number and word games daily, $21.6\%$ sometimes, and $52.9\%$ never. $4.3\%$ of participants played cards or other games daily, $27.2\%$ sometimes, and $68.6\%$ never. Table 1 shows the participants' characteristics at pre-baseline. Table 2 shows the distribution of health and wellbeing outcomes. Supplementary Table 1 presents the participant characteristics at pre-baseline for each leisure activity.
## 3.2. Relaxed leisure activity—reading books, magazines, and newspapers and subsequent wellbeing, physical health, daily life functioning, cognitive impairment, and all-cause mortality
At the 6-year follow-up, middle-aged and older adults who read books, magazines, and newspapers almost every day had a substantially lower risk of being diagnosed with depression (by $7\%$ compared to respondents who did not read at all; $95\%$ CI = 0.904, 0.949, $p \leq 0.001$) (Table 3). They also scored lower on the loneliness scale (β = −0.056, $95\%$ CI = −0.111, −0.001, $$p \leq 0.045$$). Compared to those who did not read at all, they reported higher scores in several wellbeing dimensions, such as “Future looks good” (β = 0.097, $95\%$ CI = 0.047, 0.147, $$p \leq 0.003$$), “I feel full of energy these days” (β = 0.085, $95\%$ CI = 0.027, 0.144, $$p \leq 0.010$$), “On balance, I look back at my life with a sense of happiness” (β = 0.135, $95\%$ CI = 0.084, 0.186, $p \leq 0.001$), “I look forward to each day” (β = 0.097, $95\%$ CI = 0.041, 0.153, $$p \leq 0.003$$), and “I feel that my life has meaning” (β = 0.107, $95\%$ CI = 0.056, 0.159, $$p \leq 0.001$$). Additionally, they had a substantially lower risk of limitations in ADL (RR = 0.880, $95\%$ CI = 0.797–0.973, $$p \leq 0.012$$) and IADL (RR = 0.913, $95\%$ CI = 0.863, 0.966, $$p \leq 0.002$$), as well as a lower risk of chronic and impaired pain (RR = 0.948, $95\%$ CI = 0.903, 0.996, $$p \leq 0.033$$). Finally, they scored higher on the time-orientation scale (β = −0.077, $95\%$ CI = 0.014, 0.139, $$p \leq 0.021$$). These temporal associations were independent of demographic and socioeconomic status, personality, health history, previous daily life functioning, health behaviors, and lifestyle. They were also independent of their history of reading books, magazines, and newspapers.
**Table 3**
| Unnamed: 0 | Unnamed: 1 | Reading books, magazines, and newspapers (ref. = never) | Reading books, magazines, and newspapers (ref. = never).1 | Doing word or number games (ref. = never) | Doing word or number games (ref. = never).1 | Playing cards or games (ref. = never) | Playing cards or games (ref. = never).1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Outcome | Statisticsb, c | Sometimes d | Almost every day | Sometimes d | Almost every day | Sometimes d | Almost every day |
| Emotional wellbeing | Emotional wellbeing | Emotional wellbeing | Emotional wellbeing | Emotional wellbeing | Emotional wellbeing | Emotional wellbeing | Emotional wellbeing |
| Loneliness (3-item loneliness scale) | βb (95% CI) p-value | −0.035 (−0.094, 0.024) 0.215 | −0.056† (−0.111, −0.001) 0.045 | −0.025 (−0.068, 0.018) 0.228 | −0.029 (−0.085, 0.026) 0.265 | −0.060 (−0.090, −0.031) 0.002 | −0.097 (−0.142, −0.051) 0.001 |
| Alzheimer's disease | OR (95% CI) p-value | 0.880 (0.611, 1.269) 0.492 | 0.696 (0.461, 1.051) 0.085 | 0.822 (0.609, 1.109) 0.199 | 0.834 (0.629, 1.106) 0.206 | 0.829 (0.628, 1.094) 0.182 | 0.606† (0.415, 0.884) 0.009 |
| Depression (EURO-D≥4) | RR (95% CI) p-value | 0.969 (0.909, 1.032) 0.313 | 0.925† (0.861, 0.994) 0.035 | 0.921† (0.863, 0.983) 0.013 | 0.927† (0.864, 0.994) 0.034 | 0.904† (0.844, 0.969) 0.005 | 0.936 (0.844, 1.039) 0.216 |
| Future looks good | βb (95% CI) p-value | 0.067† (0.023, 0.111) 0.008 | 0.097† (0.047, 0.147) 0.003 | 0.048† (0.002, 0.095) 0.041 | 0.056 (−0.002, 0.114) 0.055 | 0.046† (0.008, 0.083) 0.022 | 0.024 (−0.034, 0.082) 0.384 |
| I feel full of energy these days | βb (95% CI) p-value | 0.038 (−0.011, 0.087) 0.114 | 0.085† (0.027, 0.144) 0.010 | 0.041† (0.008, 0.073) 0.020 | 0.054† (0.012, 0.095) 0.018 | 0.022 (−0.014, 0.058) 0.204 | −0.014 (−0.080, 0.052) 0.641 |
| On balance, I look back on my life with a sense of happiness | βb (95% CI) p-value | 0.069 (−0.001, 0.140) 0.053 | 0.135 (0.084, 0.186) < 0.001 | 0.019 (−0.014, 0.052) 0.226 | 0.026 (−0.034, 0.085) 0.361 | 0.047 (−0.006, 0.100) 0.074 | 0.082† (0.023, 0.141) 0.010 |
| I look forward to each day | βb (95% CI) p-value | 0.076† (0.007, 0.144) 0.033 | 0.097† (0.041, 0.153) 0.003 | 0.066 (0.034, 0.098) 0.001 | 0.049 (−0.001, 0.099) 0.053 | 0.070† (0.031, 0.110) 0.003 | 0.019 (−0.053, 0.091) 0.570 |
| I feel that my life has meaning | βb (95% CI) p-value | 0.045 (−0.021, 0.110) 0.153 | 0.107 (0.056, 0.159) 0.001 | 0.026 (−0.002, 0.055) 0.068 | 0.016 (−0.023, 0.055) 0.388 | 0.037 (−0.012, 0.086) 0.125 | −0.015 (−0.105, 0.075) 0.718 |
| Daily life functioning | Daily life functioning | Daily life functioning | Daily life functioning | Daily life functioning | Daily life functioning | Daily life functioning | Daily life functioning |
| ADL (at least 1 limitation) | RR (95% CI) p-value | 0.926 (0.844, 1.017) 0.107 | 0.880† (0.797, 0.973) 0.012 | 0.948 (0.854, 1.052) 0.317 | 0.940 (0.837, 1.055) 0.291 | 0.965 (0.878, 1.059) 0.449 | 1.043 (0.911, 1.193) 0.542 |
| IADL (at least 1 limitation) | RR (95% CI) p-value | 0.909† (0.836, 0.989) 0.026 | 0.913 (0.863, 0.966) 0.002 | 0.940 (0.866, 1.019) 0.134 | 0.924 (0.827, 1.032) 0.163 | 0.960 (0.880, 1.046) 0.351 | 1.012 (0.887, 1.156) 0.857 |
| Physical health | Physical health | Physical health | Physical health | Physical health | Physical health | Physical health | Physical health |
| Heart attack | OR (95% CI) p-value | 0.823 (0.645, 1.050) 0.114 | 1.069 (0.892, 1.282) 0.468 | 0.944 (0.859, 1.037) 0.228 | 1.009 (0.916, 1.111) 0.857 | 0.999 (0.913, 1.093) 0.987 | 0.945 (0.749, 1.197) 0.647 |
| Hypertension | RR (95% CI) p-value | 0.976 (0.913, 1.043) 0.471 | 0.973 (0.918, 1.032) 0.356 | 1.005 (0.963, 1.050) 0.807 | 0.980 (0.944, 1.017) 0.288 | 1.007 (0.963, 1.054) 0.746 | 1.006 (0.902, 1.129) 0.920 |
| High blood cholesterol | RR (95% CI) p-value | 0.994 (0.912, 1.084) 0.894 | 0.957 (0.874, 1.048) 0.339 | 1.001 (0.939, 1.066) 0.984 | 1.021 (0.949, 1.098) 0.582 | 1.036 (0.980, 1.095) 0.210 | 1.025 (0.925, 1.135) 0.638 |
| Stroke | OR (95% CI) p-value | 0.849 (0.628, 1.147) 0.286 | 0.810 (0.627, 1.046) 0.106 | 0.910 (0.725, 1.134) 0.388 | 0.881 (0.667, 1.163) 0.370 | 1.196 (0.965, 1.481) 0.101 | 1.192 (0.821, 1.731) 0.356 |
| Diabetes | RR (95% CI) p-value | 0.968 (0.867, 1.081) 0.561 | 1.070 (0.982, 1.166) 0.123 | 0.978 (0.917, 1.044) 0.510 | 1.002 (0.926, 1.083) 0.966 | 0.970 (0.893, 1.053) 0.464 | 0.970 (0.880, 1.070) 0.548 |
| Cancer | RR (95% CI) p-value | 0.950 (0.840, 1.075) 0.417 | 1.006 (0.900, 1.125) 0.914 | 0.932 (0.754, 1.152) 0.514 | 0.955 (0.790, 1.153) 0.629 | 1.212† (1.014, 1.447) 0.034 | 1.456† (1.074, 1.973) 0.016 |
| Outcome | Statisticsb, c | Sometimes d | Almost every day | Sometimes d | Almost every day | Sometimes d | Almost every day |
| Pain | RR (95% CI) p-value | 0.990 (0.941, 1.041) 0.688 | 0.948† (0.903, 0.996) 0.033 | 0.979 (0.940, 1.019) 0.292 | 0.968 (0.910, 1.029) 0.296 | 0.947† (0.912, 0.983) 0.004 | 0.994 (0.921, 1.072) 0.871 |
| Mobility | RR (95% CI) p-value | 1.011 (0.982, 1.041) 0.456 | 0.982 (0.951, 1.013) 0.250 | 1.004 (0.961, 1.048) 0.872 | 0.970 (0.937, 1.003) 0.075 | 0.960† (0.933, 0.988) 0.005 | 1.011 (0.965, 1.060) 0.642 |
| Cognitive impairment | Cognitive impairment | Cognitive impairment | Cognitive impairment | Cognitive impairment | Cognitive impairment | Cognitive impairment | Cognitive impairment |
| Date orientation | βb (95% CI) p-value | 0.075† (0.003, 0.147) 0.043 | 0.077† (0.014, 0.139) 0.021 | 0.052† (0.008, 0.095) 0.024 | 0.026 (−0.018, 0.070) 0.224 | 0.034 (−0.007, 0.075) 0.095 | 0.049 (−0.014, 0.112) 0.111 |
| All-cause mortality | OR (95% CI) p-value | 0.903 (0.792, 1.031) 0.132 | 0.910 (0.822, 1.007) 0.068 | 0.865† (0.748, 0.999) 0.049 | 0.867† (0.785, 0.956) 0.004 | 0.938 (0.813, 1.082) 0.378 | 0.896 (0.727, 1.104) 0.301 |
There was no evidence that reading books, magazines, or newspapers was prospectively associated with all-cause mortality and physical health outcomes, including heart attack, hypertension, high blood cholesterol, stroke, diabetes, and cancer (at a 6-year follow-up).
## 3.3. Serious and solitary mind games (number and word games) and subsequent wellbeing, physical health, daily life functioning, cognitive impairment, and all-cause mortality
Middle-aged and older adults who engaged in serious solitary mind games (number and word games) almost daily had a $7\%$ lower risk of depression (RR = 0.927, $95\%$ CI = 0.864, 0.994, $$p \leq 0.034$$) and a $13\%$ lower risk of death regardless of the cause of death (RR = 0.867, $95\%$ CI = 0.785, 0.956, $$p \leq 0.004$$) compared to middle-aged and older adults who did not engage in these activities (Table 3). They also reported higher scores for feeling full of energy (β = 0.054, $95\%$ CI = 0.012, 0.095, $$p \leq 0.018$$) than those who did not engage in the activity. Regarding optimism reflected in looking forward to each day (“I look forward to each day”) and a positive future outlook (“Future looks good”), middle-aged and older adults who sometimes engage in number and word games scored higher than those who did not engage (β = 0.066, $95\%$ CI = 0.034, 0.098, $$p \leq 0.001$$; β = 0.048, $95\%$ CI = 0.002, 0.095, $$p \leq 0.041$$, respectively). However, the effects for respondents who engage almost daily are similar in size to those engaging occasionally but not significant due to wider confidence intervals. Similarly, we found that better time orientation was associated with prior occasional (i.e., sometimes) engagement in number and word games but not everyday engagement (β = 0.052, $95\%$ CI = 0.008, 0.095, $$p \leq 0.024$$).
The associations were independent of demographic and socioeconomic status, personality, health history, prior daily life functioning, health behaviors, lifestyle, and previous engagement in word and number games. No prospective associations were found between engagement in word and number games, indicators of daily life functioning, and physical health outcomes.
## 3.4. Serious and social mind games (playing cards and games) and subsequent wellbeing, physical health, daily life functioning, cognitive impairment, and all-cause mortality
Playing cards and games almost daily were prospectively associated with lower scores on the loneliness scale (β = −0.097, $95\%$ CI = −0.142, −0.051, $$p \leq 0.001$$), a $39\%$ lower risk of Alzheimer's disease (OR = 0.606, $95\%$ CI = 0.415, 0.884, $$p \leq 0.009$$), and a $56\%$ increased risk of cancer (RR = 1.456, $95\%$ CI = 1.074, 1.973, $$p \leq 0.016$$) compared with those who did not perform the activities (Table 3). Additionally, it was found that playing cards and games only sometimes compared to not at all was associated with an increased optimism (“Future looks good”, β = 0.046, $95\%$ CI = 0.008, 0.083, $$p \leq 0.022$$; “I look forward to each day,” β = 0.070, $95\%$ CI = 0.031, 0.110, $$p \leq 0.003$$). Furthermore, middle-aged and older adults who occasionally engaged in this activity had a $10\%$ lower risk of depression (RR = 0.904, $95\%$ CI = 0.844, 0.969, $$p \leq 0.005$$), a $5\%$ lower risk of feeling impaired or chronic pain (RR = 0.947, $95\%$ CI = 0.912, 0.983, $$p \leq 0.004$$), and a $4\%$ lower risk of experiencing mobility limitation (RR = 0.960, $95\%$ CI = 0.933, 0.988, $$p \leq 0.005$$).
The associations were independent of demographic and socioeconomic status, personality, health history, prior daily life functioning, health behaviors, lifestyle, and engagement in playing cards and games. No prospective associations were found between engagement in activities involving playing cards and games, quality of life indicators, physical health outcomes, or all-cause mortality.
## 3.5. Robustness analysis
When rerunning the models after excluding respondents with a pre-baseline health condition (i.e., depression, heart attack, hypertension, high blood cholesterol, stroke, diabetes, cancer, and Alzheimer's disease), most of the prospective associations examined remained significant but sometimes slightly attenuated (Supplementary Table 2). In the complete case scenario, the results were similar to those obtained from the primary analyses of the imputed dataset (Supplementary Table 3). Additionally, the analyses rerun with the limited set of covariates (Supplementary Table 4) yielded very similar results to those obtained from the primary analyses; however, the effect sizes were larger and several non-significant associations from the primary analyses became significant. This provides additional evidence supporting the robustness of temporal associations between leisure activities and all-cause mortality, wellbeing, physical health, daily life functioning, and cognitive impairment outcomes.
The E-values suggest that the observed associations were modestly robust to unmeasured confounding factors (Supplementary Table 5). The most robust associations (the largest E-value) were those between serious and social mind games, that is, participation in activities involving playing cards and games, and two diseases: Alzheimer's disease and cancer.
## 4. Discussion
This study examined the temporal associations between engaging in three specific classes of mind-engaging leisure activities and 21 subsequent outcomes. The results indicate that different forms of mind-engaging leisure have distinctively different effects on middle-aged and older adults. This supports the idea that a specific focus on the type of activity carried out is crucial when evaluating its health and wellbeing benefits.
We found particular differences between reading and problem-solving leisure activities (mind games), whether solitary or social, which may be due to the variable natures of these activities. In the case of reading, the focus is on meaning and content, whereas in the case of mind games, it is on performance. Cognitive resources are always engaged but under different conditions and with different patterns and goals. In experiencing meaning through reading, participants explore possibilities (whether real or fictional) and absorb information. In doing mind games, they look for solutions to specific problems. Moreover, while engaging in solitary mind games, one's cognitive resources are tested; in social mind games, such resources are also compared to the resources of others with the additional implication of social rewards (i.e., winning vs. losing). In addition, some social mind games, such as poker, may be potentially addictive [46]. Therefore, the corresponding neural pathways that supersede these activities may differ (47–50).
These differences are reflected in our results, where specific effect patterns were observed for each leisure activity class. For reading, there is a clear and strong association with wellbeing, corroborating previous findings of the structured review of 12 studies by Latchem and Greenhalgh [18]. There is also a significant effect of reading on depressive symptoms, which, despite being inconsistent with the findings for middle-aged and older adults in the US reported by Bone et al. [ 51], corroborates the findings of a randomized controlled study by Kaltenegger et al. [ 52]. Additionally, the prospective association between reading and subsequent reduced chronic pain and better daily life functioning corroborates previous limited findings [52, 53]. Similar to previous studies, there is also a positive cognitive effect regarding time orientation [54]. Additionally, reading had a positive effect on reduced loneliness, which corroborates previous qualitative and quantitative findings in other studies (55–58). This could seem counterintuitive as reading is generally a solitary activity, however, it is important to realize that, at least for certain types of reading activities such as reading fiction, an important dimension of simulation of social interaction is essentially involved [59]. This may positively affect perception of loneliness, as the reader is immersed in social situations and may even develop an identification with fictional characters [57] and a better real-life ability to empathize [58].
We found an association between regular and frequent reading of books, magazines, and newspapers (almost daily) and increased subsequent sense of meaning in life. This result extends the list of antecedents of meaning in life, adding this activity to previously documented purpose and meaning determinants. These determinants include positive affect, social connections, orientation to promote good, feeling purposeful at work, wealth, and income (37, 39, 60–62). The reading experience appears to unlock a wide spectrum of benefits. They were mostly concentrated in the psycho-socio-behavioral sphere, as no significant effects on physical health outcomes and mortality were found. Furthermore, a positive and linear association was found between exposure to reading activities and health effects.
Solitary mind games provide different sets of benefits. We observed a positive effect on depressive symptoms and a lower risk of mortality. There was also a positive effect on a specific interoceptive wellbeing dimension (feeling full of energy). However, other wellbeing dimensions, such as optimism, were positively affected only when the activity was occasional and not regular. When it was regular, the effect was similar in size but not significant due to wider confidence intervals. Notably, for cognitive impairment. Occasional engagement was found to be more beneficial than regular engagement. This may seem counterintuitive to the idea that cognitive function is maintained through constant exercise and that mind games are the mental equivalent of gym activity to benefit physical fitness [20, 26, 27]. Although our findings do not present a clear picture of the associations examined, the complexity of the associations is in line with previous studies, which reported mixed results. On the one hand, there is evidence on the lack of positive impact of cognitive brain training on the cognition and wellbeing of working adults [63]. On the other hand, the effect of [1] inductive reasoning training on decreased difficulty with instrumental activities of daily living (IADL) [64] and [2] verbal episodic memory, inductive reasoning, and speed of processing trainings on cognitive abilities among older adults [64], are presented.
Although it is difficult to interpret these results clearly, we can observe that regular engagement in solitary mind games does not entail any form of simulated social interaction (contrary to reading). Therefore, an excessive focus on these activities might have a negative trade-off with other social activities that have a complementary impact on emotional wellbeing. Furthermore, the solitary cognitive function exercise may have an excessively narrow focus to truly preserve cognitive fluency considering the importance of interaction-related cognition (including collective thinking) in humans. Therefore, individualistic and meaningless exertion of cognitive function for the preservation of cognitive fluency may not be optimal in the medium-long term [65]. Unlike reading, regular solitary mind games may have an important positive effect on mortality risk, which requires further analysis. Compared to reading, the effects of solitary mind games are narrower in scope, related to moderate rather than regular exposure, and more evenly split between the psycho-socio-behavioral, physical health, and longevity dimensions. Contrary to previous findings [64], no prospective associations with daily functioning (ADL and IADL) were found.
Social-mind games are altogether different. Similar to previous studies [56, 66], the positive effects of regular exposure to social mind games impact loneliness (as expected) and mobility limitations. However, there is also a significantly higher risk of cancer and a reduced risk of Alzheimer's disease. This finding on the risk of cancer warrants further research. However, we can only hypothesize that this social activity might have been concurrent with unfavorable health behaviors, such as smoking or drinking alcohol, well-known cancer risk factors [67]. Social mind games, similar to solitary games, generate certain positive effects only if they are performed through limited exposure. This is the case for optimism and hedonic wellbeing, pain perception, and depression. Concerning social mind games, there seems to be an upper limit beyond which further activity is less beneficial for certain outcomes. In this case, it cannot be linked to a lack of sociality, as these activities are inherently social; rather, it is related to the specific conditions of such sociality. In particular, stress could be the negative component typically associated with competitive situations [68]. Regular participation in competitive games could lead to permanent stressful arousal that, in the medium-long term, could be detrimental for wellbeing and even health, particularly for increased cancer risk. This hypothesis is consistent with previous research documenting the detrimental role of stress in wellbeing [69, 70] and the onset of disease [71, 72], such as cancer (73–75). Less regular exposure seems to be beneficial, as stressful arousal is limited, and possibly stimulating for aging participants. Meaning-oriented forms of social interaction less related to stressful arousal are likely to have different effects. However, our results on the prospective association between participation in social mind games and the reduced risk of Alzheimer's disease corroborate previous evidence linking cognitive (leisure) activities with this disease and other dementias (76–78). However, our findings indicate that not only simply exercising cognitive function may be beneficial, but also the social component inherent in the activities may be important.
Unlike reading, mind games have issues of excess exposure and a narrower range of benefits, including some crucial risks in the case of social mind games. Purely exercising cognitive function in solitary and social contexts may be beneficial, but this should not involve too much time or too many mental resources. Instead, activities that are cognitively engaging but related to meaning, even if solitary, can generate a wide range of benefits and positively affect loneliness.
## 4.1. Strengths and limitations
This study was based on a large sample and benefitted from a longitudinal design. This made it possible to make inferences about prospective associations and to account for a wide set of confounders. We found several effects, some of which could be deemed intuitive. In contrast, others are notable and encourage further research with implications beyond the scope of the present study. These associations and effects were independent of demographic and socioeconomic status, personality, lifestyle, health behaviors, and medical history. These temporal associations were also reasonably robust to unmeasured confounding, missing data patterns, and medical history (considering only new instances of disease vs. controlling for the history of diseases).
The study's main limitation is that we could not directly test the possible pathways behind the effects we found. For this purpose, we would need biobehavioral measurements and indicators that, to our knowledge, are rarely collected and not yet fully developed from a conceptual and methodological viewpoint. Therefore, we hope this study will increase interest in this new and promising research direction. Next, this study used self-reported data on health conditions, wellbeing, and daily life functioning outcomes, which makes our results subject to social desirability bias. However, there is some reassurance that this bias did not negatively affect the accuracy of the results owing to the longitudinal design and control of pre-baseline outcomes and exposure. Our use of self-reported health outcomes may also have influenced the accuracy of the results. However, previous studies provide some confidence in this regard, as they report a high agreement between medical records and self-reported disease data [79]. Finally, the data used were collected from middle-aged and older adults, which might reduce the accuracy and generalizability of our results. However, most of the instruments used were developed specifically for this population. Nevertheless, further studies are required to corroborate our findings in different populations.
## 5. Conclusions
Leisure is a very broad category of activities, and their effects on health and wellbeing can be very different as shown in prior studies (13, 16, 17, 80–84). In this paper, we present a case of differences in health and wellbeing impacts across relatively similar cognitive leisure activities. These are sedentary forms of leisure, such as reading books, magazines, and newspapers; doing word and number games (solitary mind games); and playing cards and other competitive games, such as chess (social mind games). The fact that the experience—individual or social—involves cognitive stimulation related to meaning rather than pure cognitive performance has an important difference in its effects on wellbeing and health outcomes. This implies that exploring the differences in the biobehavioral pathways that cause such diverse outcomes is important to gain a deeper understanding of these effects.
We enter a new research stage on the health and wellbeing effects of cognitive leisure activities. The main issue is no longer whether leisure may have such effects but rather how and why such effects occur. By posing these questions, we can design more targeted and possibly effective policy interventions. Such interventions will especially benefit fragile and disadvantaged participants, such as older adults. Moreover, they may benefit others who are generally disadvantaged in terms of wellbeing and health opportunities; this includes those who are socioeconomically deprived, marginalized, and with important psychological and medical conditions.
## Data availability statement
Publicly available datasets were analyzed in this study. This data can be found at: [1] Börsch-Supan, A. [2022]. Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 4. Release version: 8.0.0. SHARE-ERIC. Data set. doi: 10.6103/SHARE.w4.800. [ 2] Börsch-Supan, A. [2022]. Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 5. Release version: 8.0.0. SHARE-ERIC. Data set. doi: 10.6103/SHARE.w5.800. [ 3] Börsch-Supan, A. [2022]. Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 8. Release version: 8.0.0. SHARE-ERIC. Data set. doi: 10.6103/SHARE.w8.800.
## Ethics statement
The study was reviewed and approved by the Ethics Committee of the University of Mannheim and the Ethics Council of the Max Planck Society (http://www.share-project.org/fileadmin/pdf_documentation/SHARE_ethics_approvals.pdf). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
DW-B developed the study concept, contributed to the data analysis, drafted, revised, and approved the final version of the manuscript. PB contributed to the data analysis, drafted, revised, and approved the final version of the manuscript. PS contributed to the study concept, drafted, revised, and approved the final version 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/fpubh.2023.1117822/full#supplementary-material
## References
1. Ateca-Amestoy V, Serrano-del-Rosal R, Vera-Toscano E. **The leisure experience**. *J Socio Econ.* (2008) **37** 64-78. DOI: 10.1016/j.socec.2006.12.025
2. Pressman SD, Matthews KA, Cohen S, Martire LM, Scheier M, Baum A. **Association of enjoyable leisure activities with psychological and physical well-being**. *Psychosom Med.* (2009) **71** 725-32. DOI: 10.1097/PSY.0b013e3181ad7978
3. Downward P, Dawson P. **Is it pleasure or health from leisure that we benefit from most? An analysis of well-being alternatives and implications for policy**. *Soc Indic Res.* (2016) **126** 443-65. DOI: 10.1007/s11205-015-0887-8
4. Waugh CE, Shing EZ, Furr RM. **Not all disengagement coping strategies are created equal: positive distraction, but not avoidance, can be an adaptive coping strategy for chronic life stressors**. *Anxiety Stress Coping.* (2020) **33** 511-29. DOI: 10.1080/10615806.2020.1755820
5. Hodge CJ, Zabriskie RB, Townsend JA, Eggett DL, Poff R. **Family leisure functioning: a cross-national study**. *Leis Sci.* (2018) **40** 194-215. DOI: 10.1080/01490400.2016.1203847
6. Thurnell-Read T. **If they weren't in the pub, they probably wouldn't even know each other: alcohol, sociability and pub based leisure**. *Int J Sociol Leis.* (2021) **4** 61-78. DOI: 10.1007/s41978-020-00068-x
7. Manzano-León A, Rodríguez-Ferrer JM, Aguilar-Parra JM, Herranz-Hernández R. **Gamification and family leisure to alleviate the psychological impact of confinement due to COVID-19**. *Child Soc* (2021) **36** 433-49. DOI: 10.1111/chso.12495
8. MacKean R, Abbott-Chapman J. **Leisure activities as a source of informal learning for older people: the role of community-based organisations**. *Aust J Adult Learn.* (2011) **51** 226-46
9. Bouwer J, van Leeuwen M. *Philosophy of Leisure.* (2017)
10. Shivers JS. *Leisure and Recreation Concepts.* (1981)
11. Kelly JR. **Leisure socialization: replica**. *J Leis Res.* (1977) **9** 121-32. DOI: 10.1080/00222216.1977.11970318
12. Agahi N, Parker MG. **Are today's older people more active than their predecessors? Participation in leisure-time activities in Sweden in 1992 and 2002**. *Ageing Soc* (2005) **25** 925-41. DOI: 10.1017/S0144686X05004058
13. Fancourt D, Aughterson H, Finn S, Walker E, Steptoe A. **How leisure activities affect health: a narrative review and multi-level theoretical framework of mechanisms of action**. *Lancet Psychiatry.* (2021) **8** 329-39. DOI: 10.1016/S2215-0366(20)30384-9
14. Caldwell LL. **Leisure and health: why is leisure therapeutic?**. *Br J Guid Counc.* (2005) **33** 7-26. DOI: 10.1080/03069880412331335939
15. Mansfield L, Daykin N, Kay T. **Leisure and wellbeing**. *Leis Stud.* (2020) **39** 1-10. DOI: 10.1080/02614367.2020.1713195
16. Weziak-Białowolska D, Białowolski P. **Cultural events – does attendance improve health? Evidence from a Polish longitudinal study**. *BMC Public Health* (2016) **16** 730. DOI: 10.1186/s12889-016-3433-y
17. Weziak-Białowolska D, Białowolski P, Sacco PL. **Involvement with the arts and participation in cultural events — does personality moderate impact on wellbeing? Evidence from the UK household survey**. *Psychol Aesthetics Creat Arts.* (2018) **13** 348-58. DOI: 10.1037/aca0000180
18. Latchem JM, Greenhalgh J. **The role of reading on the health and well-being of people with neurological conditions: a systematic review**. *Aging Ment Health.* (2014) **18** 731-44. DOI: 10.1080/13607863.2013.875125
19. Wang S, Li LZ, Zhang J, Rehkopf DH. **Leisure time activities and biomarkers of chronic stress: the mediating roles of alcohol consumption and smoking**. *Scand J Public Health.* (2021) **49** 940-50. DOI: 10.1177/1403494820987461
20. Cheng W, Zhang Z, Cheng W, Yang C, Diao L, Liu W. **Associations of leisure-time physical activity with cardiovascular mortality: a systematic review and meta-analysis of 44 prospective cohort studies**. *Eur J Prev Cardiol.* (2018) **25** 1864-72. DOI: 10.1177/2047487318795194
21. Zhao M, Veeranki SP, Li S, Steffen LM, Xi B. **Beneficial associations of low and large doses of leisure time physical activity with all-cause, cardiovascular disease and cancer mortality: a national cohort study of 88,140 US adults**. *Br J Sports Med.* (2019) **53** 1405-11. DOI: 10.1136/bjsports-2018-099254
22. Nielsen L, Hinrichsen C, Madsen KR, Nelausen MK, Meilstrup C, Koyanagi A. **Participation in social leisure activities may benefit mental health particularly among individuals that lack social connectedness at work or school**. *Ment Heal Soc Inclusion.* (2021) **25** 341-51. DOI: 10.1108/MHSI-06-2021-0026
23. Hu FB. **Sedentary lifestyle and risk of obesity and type 2 diabetes**. *Lipids.* (2003) **38** 103-8. DOI: 10.1007/s11745-003-1038-4
24. Thorp AA, Owen N, Neuhaus M, Dunstan DW. **Sedentary behaviors and subsequent health outcomes in adults**. *A Am J Prev Med.* (2011) **41** 207-15. DOI: 10.1016/j.amepre.2011.05.004
25. Proper KI, Singh AS, van Mechelen W, Chinapaw MJM. **Sedentary behaviors and health outcomes among adults. A systemic review of prospective studies**. *Am J Prev Med.* (2011) **40** 174-82. DOI: 10.1016/j.amepre.2010.10.015
26. Brown WJ, Mishra G, Lee C, Bauman A. **Leisure time physical activity in Australian women: relationship with well being and symptoms**. *Res Q Exerc Sport.* (2000) **71** 206-16. DOI: 10.1080/02701367.2000.10608901
27. Ku P-W, Fox KR, Chen L-J. **Leisure-time physical activity, sedentary behaviors and subjective well-being in older adults: an eight-year longitudinal research**. *Soc Indic Res* (2015) **127** 1349-61. DOI: 10.1007/s11205-015-1005-7
28. Healy GN, Wijndaele K, Dunstan DW, Shaw JE, Salmon J, Zimmet PZ. **Objectively measured sedentary time, physical activity, and metabolic risk**. *Diabetes Care.* (2008) **31** 369-71. DOI: 10.2337/dc07-1795
29. Park JH, Chung WJ, Kwon H, Min HY, Joh H-K, Jung K-T. **Enhancing physical activity and reducing obesity through smartcare and financial incentives: a pilot randomized trial**. *Obesity.* (2017) **25** 302-10. DOI: 10.1002/oby.21731
30. Kramer AF, Colcombe S. **Fitness effects on the cognitive function of older adults: a meta-analytic study—revisited**. *Perspect Psychol Sci.* (2018) **13** 213-7. DOI: 10.1177/1745691617707316
31. Kurita S, Doi T, Tsutsumimoto K, Hotta R, Nakakubo S, Kim M. **Cognitive activity in a sitting position is protectively associated with cognitive impairment among older adults**. *Geriatr Gerontol Int.* (2019) **19** 98-102. DOI: 10.1111/ggi.13532
32. Lau HM, Smit JH, Fleming TM, Riper H. **Serious games for mental health: are they accessible, feasible, and effective? A systemic review and meta-analysis**. *Front Psychiatry.* (2017) **7** 209. DOI: 10.3389/fpsyt.2016.00209
33. Kleiber D, Larson R. **The experience of leisure in adolescence**. *J Leis Res.* (1996) **18** 169-76. DOI: 10.1080/00222216.1986.11969655
34. Börsch-Supan A, Brandt M, Hunkler C, Kneip T, Korbmacher J, Malter F. **Data resource profile: the survey of health, ageing and retirement in europe (SHARE)**. *Int J Epidemiol.* (2013) **42** 992-1001. DOI: 10.1093/ije/dyt088
35. VanderWeele TJ, Mathur MB, Chen Y. **Outcome-wide longitudinal designs for causal inference: a new template for empirical studies**. *Stat Sci.* (2020) **35** 437-66. DOI: 10.1214/19-STS728
36. Kim ES, Chen Y, Kawachi I, VanderWeele TJ. **Perceived neighborhood social cohesion and subsequent health and well-being in older adults: an outcome-wide longitudinal approach**. *Heal Place.* (2020) **66** 102420. DOI: 10.1016/j.healthplace.2020.102420
37. Weziak-Bialowolska D, Bialowolski P, VanderWeele TJ, McNeely E. **Character strengths involving an orientation to promote good can help your health and well-being. evidence from two longitudinal studies**. *Am J Heal Promot.* (2021) **35** 388-98. DOI: 10.1177/0890117120964083
38. Białowolski P, Weziak-Białowolska D, VanderWeele TJ. **The impact of savings and credit on health and health behaviours: an outcome-wide longitudinal approach**. *Int J Public Health.* (2019) **64** 573-84. DOI: 10.1007/s00038-019-01214-3
39. Steptoe A, Fancourt D. **An outcome-wide analysis of bidirectional associations between changes in meaningfulness of life and health, emotional, behavioural, and social factors**. *Sci Rep.* (2020) **10** 1-12. DOI: 10.1038/s41598-020-63600-9
40. Tolsgaard MG, Ellaway R, Woods N, Norman G. **Salami-slicing and plagiarism: how should we respond?**. *Adv Heal Sci Educ.* (2019) **24** 3-14. DOI: 10.1007/s10459-019-09876-7
41. Knol MJ, Le Cessie S, Algra A, Vandenbroucke JP, Groenwold RHH. **Overestimation of risk ratios by odds ratios in trials and cohort studies: alternatives to logistic regression**. *Can Med Assoc J.* (2012) **184** 895-9. DOI: 10.1503/cmaj.101715
42. Chen W, Qian L, Shi J, Franklin M. **Comparing performance between log-binomial and robust Poisson regression models for estimating risk ratios under model misspecification**. *BMC Med Res Methodol.* (2018) **18** 1-12. DOI: 10.1186/s12874-018-0519-5
43. White IR, Royston P, Wood AM. **Multiple imputation using chained equations: Issues and guidance for practice**. *Stat Med.* (2011) **30** 377-99. DOI: 10.1002/sim.4067
44. Rubin DB. *Multiple Imputation for Non Response in Surveys* (1987)
45. VanderWeele TJ, Ding P. **Sensitivity analysis in observational research: introducing the e-value**. *Ann Intern Med.* (2017) **167** 268-74. DOI: 10.7326/M16-2607
46. Szabó A, Kocsis D. **Susceptibility to addictive behaviour in online and traditional poker playing environments**. *J Behav Addict.* (2012) **1** 23-7. DOI: 10.1556/JBA.1.2012.1.2
47. Nakamura K, Hara N, Kouider S, Takayama Y, Hanajima R, Sakai K. **Task-guided selection of the dual neural pathways for reading**. *Neuron.* (2006) **52** 557-64. DOI: 10.1016/j.neuron.2006.09.030
48. Hopfieid JJ. **Searching for memories, Sudoku, implicit check bits, and the iterative use of not-always-correct rapid neural computation**. *Neural Comput.* (2008) **20** 1119-64. DOI: 10.1162/neco.2007.09-06-345
49. Zhao Q, Zhou Z, Xu H, Fan W, Han L. **Neural pathway in the right hemisphere underlies verbal insight problem solving**. *Neuroscience.* (2014) **256** 334-41. DOI: 10.1016/j.neuroscience.2013.10.019
50. Fuentes-García JP, Pereira T, Castro MA, Carvalho Santos A, Villafaina S. **Psychophysiological stress response of adolescent chess players during problem-solving tasks**. *Physiol Behav.* (2019) **209** 112609. DOI: 10.1016/j.physbeh.2019.112609
51. Bone JK, Bu F, Fluharty ME, Paul E, Sonke JK, Fancourt D. **Engagement in leisure activities and depression in older adults in the United States: longitudinal evidence from the health and retirement study**. *Soc Sci Med.* (2022) **294** 114703. DOI: 10.1016/j.socscimed.2022.114703
52. Kaltenegger K, Kuester S, Altpeter-Ott E, Eschweiler GW, Cordey A, Ivanov IV. **Effects of home reading training on reading and quality of life in AMD—a randomized and controlled study**. *Graefe's Arch Clin Exp Ophthalmol.* (2019) **257** 1499-512. DOI: 10.1007/s00417-019-04328-9
53. Billington J, Farrington G, Lampropoulou S, Lingwood J, Jones A, Ledson J. **A comparative study of cognitive behavioural therapy and shared reading for chronic pain**. *Med Humanit* (2017) **43** 155 LP-65. DOI: 10.1136/medhum-2016-011047
54. Yates LA, Ziser S, Spector A, Orrell M. **Cognitive leisure activities and future risk of cognitive impairment and dementia: systematic review and meta-analysis**. *Int Psychogeriatrics.* (2016) **28** 1791-806. DOI: 10.1017/S1041610216001137
55. Pettigrew S, Roberts M. **Addressing loneliness in later life**. *Aging Ment Heal.* (2008) **12** 302-9. DOI: 10.1080/13607860802121084
56. Rane-Szostak D, Herth KA. **Pleasure reading, other activities, and loneliness in later life**. *J Adolesc Adult Lit* (1995) **39** 100-8
57. Broom TW, Chavez RS, Wagner DD. **Becoming the King in the North: identification with fictional characters is associated with greater self-other neural overlap**. *Soc Cogn Affect Neurosci.* (2021) **16** 541-51. DOI: 10.1093/scan/nsab021
58. Mumper ML, Gerrig RJ. **Leisure reading and social cognition: a meta-analysis**. *Psychol Aesthetics Creat Arts.* (2017) **11** 109-20. DOI: 10.1037/aca0000089
59. Oatley K. **Fiction: simulation of social worlds**. *Trends Cogn Sci.* (2016) **20** 618-28. DOI: 10.1016/j.tics.2016.06.002
60. Chen Y, Kim ES, Shields AE, VanderWeele TJ. **Antecedents of purpose in life: evidence from a lagged exposure-wide analysis**. *Cogent Psychol.* (2020) **7** 1825043. DOI: 10.1080/23311908.2020.1825043
61. King LA, Hicks JA, Krull JL, Del Gaiso AK. **Positive affect and the experience of meaning in life**. *J Pers Soc Psychol.* (2006) **90** 179-96. DOI: 10.1037/0022-3514.90.1.179
62. Weziak-Bialowolska D, Bialowolski P, Sacco PL, VanderWeele TJ, McNeely E. **Well-being in life and well-being at work: which comes first? Evidence from a longitudinal study**. *Front Public Heal.* (2020) **8** 103. DOI: 10.3389/fpubh.2020.00103
63. Borness C, Proudfoot J, Crawford J, Valenzuela M. **Putting brain training to the test in the workplace: a randomized, blinded, multisite, active-controlled trial**. *PLoS ONE.* (2013) **8** 1-8. DOI: 10.1371/journal.pone.0059982
64. Willis SL, Tennstedt SL, Marsiske M, Ball K, Elias J, Koepke KM. **Long-term effects of cognitive training on everyday functional outcomes in older adults**. *J Am Med Assoc.* (2006) **296** 2805-14. DOI: 10.1001/jama.296.23.2805
65. Mercer N. **The social brain, language, and goal-directed collective thinking: a social conception of cognition and its implications for understanding how we think, teach, and learn**. *Educ Psychol.* (2013) **48** 148-68. DOI: 10.1080/00461520.2013.804394
66. Greaves C, Farbus L. **Effects of creative and social activity on the health and well-being of socially isolated older people: outcomes from a multi-method observational study**. *Soc Promot Heal.* (2006) **126** 134-42. DOI: 10.1177/1466424006064303
67. Lee YCA, Hashibe M. **Tobacco, alcohol, and cancer in low and high income countries**. *Ann Glob Heal.* (2014) **80** 378-83. DOI: 10.1016/j.aogh.2014.09.010
68. Leis O, Lautenbach F. **Psychological and physiological stress in non-competitive and competitive esports settings: a systematic review**. *Psychol Sport Exerc.* (2020) **51** 101738. DOI: 10.1016/j.psychsport.2020.101738
69. Smith TB, McCullough ME, Poll J. **Religiousness and depression: evidence for a main effect and the moderating influence of stressful life events**. *Psychol Bull.* (2003) **129** 614-36. DOI: 10.1037/0033-2909.129.4.614
70. Wang Y, Sareen J, Afifi TO, Bolton S-L, Johnson EA, Bolton JM. **Recent stressful life events and suicide attempt**. *Psychiatr Ann.* (2012) **42** 101-8. DOI: 10.3928/00485713-20120217-07
71. Spruill TM. **Chronic psychosocial stress and hypertension**. *Curr Hypertens Rep.* (2010) **12** 10-6. DOI: 10.1007/s11906-009-0084-8
72. Redina OE, Markel AL. **Stress, genes, and hypertension. Contribution of the ISIAH rat strain study**. *Curr Hypertens Rep* (2018) **20** 66. DOI: 10.1007/s11906-018-0870-2
73. Dai S, Mo Y, Wang Y, Xiang B, Liao Q, Zhou M. **Chronic stress promotes cancer development**. *Front Oncol.* (2020) **10** 1-10. DOI: 10.3389/fonc.2020.01492
74. Chiriac VF, Baban A, Dumitrascu DL. **Psychological stress and breast cancer incidence: a systematic review**. *Psychosom Med.* (2018) **91** 18-26. DOI: 10.15386/cjmed-924
75. Perego M, Tyurin VA, Tyurina YY, Yellets J, Lin C, Nefedova Y. **Reactivation of dormant tumor cells by modified lipids derived from stress-activated neutrophils**. *Sci Transl Med* (2021) **12** eabb5817. DOI: 10.1126/scitranslmed.abb5817
76. Stern C, Munn Z. **Cognitive leisure activities and their role in preventing dementia: a systematic review**. *Int J Evid Based Healthc.* (2010) **8** 2-17. DOI: 10.1111/j.1744-1609.2010.00150.x
77. Karp A, Paillard-Borg S, Wang HX, Silverstein M, Winblad B, Fratiglioni L. **Mental, physical and social components in leisure activities equally contribute to decrease dementia risk**. *Dement Geriatr Cogn Disord.* (2006) **21** 65-73. DOI: 10.1159/000089919
78. Wilson RS, Scherr PA, Schneider JA, Tang Y, Bennett DA. **Relation of cognitive activity to risk of developing Alzheimer disease**. *Neurology.* (2007) **69** 1911-20. DOI: 10.1212/01.wnl.0000271087.67782.cb
79. Quam L, Ellis LBM, Venus P, Clouse J, Taylor CG, Leatherman S. **Using claims data for epidemiologic research: the concordance of claims-based criteria with the medical record and patient survey for identifying a hypertensive population**. *Med Care.* (1993) **31** 498-507. DOI: 10.1097/00005650-199306000-00003
80. Lewandowska K, Weziak-Bialowolska D. **The impact of theatre on social competencies: a meta-analytic evaluation**. *Arts Health* (2022). DOI: 10.1080/17533015.2022.2130947
81. Wang S, Mak HW, Fancourt D. **Arts, mental distress, mental health functioning and life satisfaction: fixed-effects analyses of a nationally-representative panel study**. *BMC Public Health.* (2020) **20** 1-9. DOI: 10.1186/s12889-019-8109-y
82. Grossi E, Tavano Blessi G, Sacco PL. **Magic moments: determinants of stress relief and subjective wellbeing from visiting a cultural heritage site**. *Cult Med Psychiatry.* (2019) **43** 4-24. DOI: 10.1007/s11013-018-9593-8
83. Weziak-Białowolska D. **Attendance of cultural events and involvement with the arts - impact evaluation on health and well-being from a Swiss household panel survey**. *Public Health.* (2016) **139** 161-9. DOI: 10.1016/j.puhe.2016.06.028
84. Stickley T, Eades M. **Arts on prescription: a qualitative outcomes study**. *Public Health.* (2013) **127** 727-34. DOI: 10.1016/j.puhe.2013.05.001
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.