Exploring potential associations and biomarkers linked polycystic ovarian syndrome with atherosclerosis via comprehensive bioinformatics analysis, machine learning, and animal experiments

Abstract

Polycystic ovary syndrome (PCOS), a common endocrine condition affecting multiple systems, is tied to atherosclerosis (AS) progression among reproductive-aged women. The present study aimed to explore the underlying associations and uncover potential biological indicators for PCOS complicated with AS. Gene expression datasets for PCOS and AS were obtained from Gene Expression Omnibus (GEO). Differentially expressed genes (DEGs) from PCOS tissues (granulosa cells, adipose tissue, skeletal muscle) and arterial wall of AS were analyzed via weighted gene co-expression network analysis (WGCNA), protein-protein interaction (PPI) network, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Immune infiltration and chemokine/receptor-immunocyte networks were constructed to explore immune cell recruitment. Key findings were validated in PCOS and AS murine models. The gradient boosting machine (GBM) and the extreme gradient boosting (XGBoost) algorithms were employed to identify potential biomarkers, further verified by the AS murine model, nomograms, and PCOS murine model. We identified 238, 60, and 76 secretory protein-encoding DEGs in PCOS tissues (granulosa cells, adipose tissue, and skeletal muscle) and 604 key AS-related DEGs. The enrichment analysis suggested associations between immune inflammation, dysregulated lipid metabolism, insulin signaling, and PCOS-related AS. Then, immunoinfiltration analysis revealed elevated naive B cell, follicular T helper cell, and neutrophil proportions in AS samples. In addition, six chemokines (CCL5, CCL20, CCL23, CCL28, CXCL1, and CXCL6) were involved in four immunocyte recruitments (B cells, neutrophils, NK cells, and CD4⁺ T cells) in AS, with CXCL1 and CXCL6 upregulated in the peripheral blood of PCOS mice. And CXCR2, the shared receptor for CXCL1/6, showed an increase in aortic tissues of both AS and PCOS mice. Machine learning identified five signature genes (LILRA5, CSF2RA, S100A8, CD6, and CCL24; AUC 0.856–0.983), two of which (CSF2RA and LILRA5) were verified in the AS murine model and the nomogram incorporating these genes showed strong predictive accuracy (AUC = 0.966). Finally, further validation in the PCOS murine model confirmed significantly elevated CSF2RA and reduced LILRA5 expression, suggesting a close association between PCOS and AS pathogenesis. This study identified potential associations between PCOS and AS, and screened the potential biological biomarkers for predicting PCOS-related AS, offering a foothold for future exploration of the diagnosis and risk stratification for PCOS-related AS.