Donor-associated variation outweighs PMOS status in patient-derived iPSC transcriptomes: a genotyped, BMI-adjusted study of 14 lines

Abstract

Abstract Background: Induced pluripotent stem cells (iPSCs) from patients have been little used to model polyendocrine metabolic ovarian syndrome (PMOS, formerly polycystic ovary syndrome, PCOS), and differences between unrelated donors can masquerade as disease effects. Earlier comparisons of three or four donors per group reported hundreds to thousands of differentially expressed genes. We asked whether a disease signature is detectable in a larger, genotyped panel, balanced for BMI category and BMI-adjusted, of full-phenotype patients. Methods: We studied 14 clonal, transgene-free iPSC lines reprogrammed by CRISPR activation from endometrial stromal fibroblasts of seven women with PMOS meeting all three Rotterdam criteria, and seven controls frequency-matched for BMI category (30 kg/m2). Bulk RNA-seq (16,489 expressed genes) was related to diagnosis, hormones, metabolic traits, age, and X-inactive specific transcript (XIST) status using gene-wise models, exact enumeration of all 3,432 group splits, and permutation-calibrated gene-set tests. Donors were genotyped within a cohort of 140 women and scored with an imputation-based genome-wide PMOS polygenic risk score (PRS). PMOS risk-locus genes were examined against a 202-line iPSC proteome. Results: PMOS donors had higher anti-Mullerian hormone (AMH), testosterone, and free androgen index than controls. No gene was differentially expressed between PMOS and control lines (false discovery rate < 0.1), and the number of nominally associated genes was no greater than for random 7-versus-7 splits of the lines (p = 0.75). This was unchanged with an alternative read quantification and with count-based models. With seven donors per group, the design could detect 1.8-fold differences per gene, compared with roughly 3- to 6-fold for the three- or four-donor designs used previously. Neither diagnosis nor any hormonal or metabolic trait explained more expression variance than chance, and neither PMOS GWAS-locus genes nor Hallmark pathways were enriched. Two control lines showed near-complete loss of XIST, potentially indicative of X-chromosome inactivation (XCI) erosion, and a third was borderline. Endocrine PMOS-locus genes (e.g., FSHR, AMH) were not expressed, whereas DNA-repair and growth genes were. Across 129 genotyped women, the PRS was higher in PMOS (odds ratio 1.81 per SD; AUC 0.65), but PMOS and control iPSC donors overlapped fully in polygenic risk. Conclusions: In this panel of 14 undifferentiated iPSC lines from women with the full PMOS phenotype, no robust PMOS-associated transcriptomic differences were detected. Expression variation was dominated by line-to-line differences, consistent with donor background as their main source, although with one clone per donor a contribution of clonal variation cannot be excluded. Reduced XIST expression in some control lines is potentially indicative of, but does not establish, X-chromosome inactivation erosion. Disease-associated differences in differentiated, hormone-responsive cells cannot be excluded, and differences reported in smaller panels may partly reflect inter-donor variability. Future studies should use adequately sized panels, control for XIST erosion, select donors by polygenic risk, and examine hormone-responsive derivatives. Keywords: polyendocrine metabolic ovarian syndrome; polycystic ovary syndrome; induced pluripotent stem cells; transcriptome; polygenic risk; XIST erosion; disease modelling; experimental design