Exploring associations among pro-inflammatory cytokines, osteoarthritis, and gut microbiome composition in individuals with obesity using machine learning
Cameron Kurz , Liubov Arbeeva , M. Andrea Azcarate-Peril , Delisha A. Stewart , B. Duncan X. Lascelles , Richard F. Loeser , Amanda E. Nelson
{"title":"Exploring associations among pro-inflammatory cytokines, osteoarthritis, and gut microbiome composition in individuals with obesity using machine learning","authors":"Cameron Kurz , Liubov Arbeeva , M. Andrea Azcarate-Peril , Delisha A. Stewart , B. Duncan X. Lascelles , Richard F. Loeser , Amanda E. Nelson","doi":"10.1016/j.ocarto.2025.100603","DOIUrl":null,"url":null,"abstract":"<div><h3>Objective</h3><div>To investigate potentially novel and modifiable mechanisms of the effects of gut microbiome composition on obesity-related osteoarthritis (OA), focusing on cross-sectional relationships between microbiota, cytokines, and lipopolysaccharide (LPS).</div></div><div><h3>Design</h3><div>Johnston County OA Project participants (n = 64) with (cases) and without (controls) OA in hands and knees, with age ≥55 years and obesity (BMI ≥30 kg/m<sup>2</sup>), provided samples for multiplex cytokine, LPS, and fecal microbiota analysis. Latent Dirichlet Allocation (LDA), a machine learning method to detect latent groups within data, was used to identify microbial enterotypes. LDA regression models were used to evaluate associations of enterotypes with demographics, cytokines associated with OA, and LPS.</div></div><div><h3>Results</h3><div>We identified 5 enterotypes. Enterotypes 3, 4 (most prevalent in our sample), and 5, dominated respectively by genera <em>Akkermansia</em>, <em>Bacteroides</em>, <em>Ruminococcus</em>/<em>Phascolarctobacterium</em>, were positively associated with control status, and inversely associated with levels of at least two cytokines associated with OA in our sample. We observed no associations of enterotypes with LPS levels. Enterotype 3 was inversely associated with thrombopoietin and IL-4 levels (b [95 % CIs] −0.19 [−0.43, 0.05] and −0.17 [−0.42, 0.08]), enterotype 5 with osteopontin and thrombopoietin (−0.23 [−0.49, 0.03] and −0.24 [−0.51, 0.04]), and enterotype 4 was inversely associated with all 3 of these cytokines (b −0.20 to −0.35).</div></div><div><h3>Conclusion</h3><div>Three of five identified enterotypes were inversely associated with OA status and levels of OA associated cytokines. These exploratory analyses revealed associations between the gut microbiome, cytokines, and OA outcomes, suggesting potentially cytokine-mediated mechanisms of the effects of gut composition on OA in obese individuals, and providing a basis for further investigation of the underlying causal mechanisms.</div></div>","PeriodicalId":74377,"journal":{"name":"Osteoarthritis and cartilage open","volume":"7 2","pages":"Article 100603"},"PeriodicalIF":0.0000,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Osteoarthritis and cartilage open","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2665913125000391","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Objective
To investigate potentially novel and modifiable mechanisms of the effects of gut microbiome composition on obesity-related osteoarthritis (OA), focusing on cross-sectional relationships between microbiota, cytokines, and lipopolysaccharide (LPS).
Design
Johnston County OA Project participants (n = 64) with (cases) and without (controls) OA in hands and knees, with age ≥55 years and obesity (BMI ≥30 kg/m2), provided samples for multiplex cytokine, LPS, and fecal microbiota analysis. Latent Dirichlet Allocation (LDA), a machine learning method to detect latent groups within data, was used to identify microbial enterotypes. LDA regression models were used to evaluate associations of enterotypes with demographics, cytokines associated with OA, and LPS.
Results
We identified 5 enterotypes. Enterotypes 3, 4 (most prevalent in our sample), and 5, dominated respectively by genera Akkermansia, Bacteroides, Ruminococcus/Phascolarctobacterium, were positively associated with control status, and inversely associated with levels of at least two cytokines associated with OA in our sample. We observed no associations of enterotypes with LPS levels. Enterotype 3 was inversely associated with thrombopoietin and IL-4 levels (b [95 % CIs] −0.19 [−0.43, 0.05] and −0.17 [−0.42, 0.08]), enterotype 5 with osteopontin and thrombopoietin (−0.23 [−0.49, 0.03] and −0.24 [−0.51, 0.04]), and enterotype 4 was inversely associated with all 3 of these cytokines (b −0.20 to −0.35).
Conclusion
Three of five identified enterotypes were inversely associated with OA status and levels of OA associated cytokines. These exploratory analyses revealed associations between the gut microbiome, cytokines, and OA outcomes, suggesting potentially cytokine-mediated mechanisms of the effects of gut composition on OA in obese individuals, and providing a basis for further investigation of the underlying causal mechanisms.