Abstract
BACKGROUND: Polycystic ovary syndrome (PCOS) is increasingly viewed as an environmentally influenced, epigenetically mediated reproductive–metabolic disorder. Bisphenol A (BPA) exposure is higher in women with PCOS and has been linked to hyperandrogenism, insulin resistance and ovarian dysfunction, while N6-methyladenosine (m6A) dysregulation contributes to granulosa cells (GCs) proliferation and apoptosis. However, the downstream gene networks that integrate BPA exposure, m6A imbalance and GCs apoptosis in PCOS remain unclear. METHODS: BPA-related genes were retrieved from the Comparative Toxicogenomics Database and intersected with PCOS differentially expressed genes from four GEO GC datasets (GSE34526, GSE137684, GSE102293, GSE80432) and m6A/apoptosis-related genes from GeneCards. Protein–protein interaction, GO/KEGG enrichment, GSEA and immune-cell infiltration (ssGSEA) analyses were performed. A five-gene logistic model and nomogram were built and externally validated in GSE106724 and GSE98595. Single-cell RNA-seq data (GSE268919) were used to map hub-gene expression, m6A and apoptosis signatures across ovarian cell types. Molecular docking (AutoDock Vina) evaluated BPA–protein binding. Hub-gene expression and function were verified in primary human GCs and BPA-treated KGN cells by qPCR, Western blot, CCK-8 and flow cytometry. RESULTS: We identified 139 BPA–PCOS shared genes and five hub genes—LIFR, LMNA, BCL2, ADM, S100P—enriched in ovarian steroidogenesis, oxidative-stress and apoptotic pathways. These genes showed distinct expression patterns between PCOS and controls and were associated with altered immune-cell infiltration. A five-gene nomogram achieved an AUC of 0.846 in the training set and 0.791 in an external validation cohort, with good calibration and decision-curve benefit. Single-cell analysis revealed LIFR downregulation and LMNA upregulation in GCs clusters with high m6A and apoptosis scores. Docking predicted moderate-to-strong BPA binding to all five proteins. In vitro, BPA reduced GCs viability, increased apoptosis with BAX/Caspase-3 activation, and reproduced the hub-gene dysregulation observed in PCOS GCs. CONCLUSIONS: Our integrative data support a BPA–m6A–apoptosis axis in PCOS and identify a five-gene granulosa-cell signature with potential diagnostic utility, providing mechanistic insight and candidate targets for environmentally driven PCOS; however, its clinical utility requires validation in larger, independent and prospectively recruited cohorts. TRIAL REGISTRATION: This work is an observational and experimental laboratory study and not a clinical trial. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s13048-026-02038-5.