Evolution of the genetic architecture of uterine disorders

13 Uterine disorders and menstrual abnormalities are prevalent reproductive conditions with significant 14 clinical consequences. Recent genome-wide association studies have identified hundreds of variants 15 contributing to different uterine disorders, while epidemiological evidence suggests that these disorders 16 co-occur more frequently than expected, implying shared genetic mechanisms. However, the specific 17 genetic variants and biological pathways underlying this shared architecture remain poorly 18 characterized. Here, w e conducted a uterus -centric, multi-trait genome -wide association analys is 19 across ten uterine disorders in a European cohort to elucidate this shared genetic architecture. Further, 20 we embedded this architecture in a functional and population genomics framework to investigate its 21 plausible biological mechanisms and its evolutionary history in humans. We confirm strong positive 22 genetic correlations between major uterine disorders and identify 31 independent susceptibility loci 23 jointly affecting the genetic risk of multiple uterine pathologies, substantiating an intertwined biological 24 basis. Populational analyses demonstrated that several of these susceptibility variants exhibit 25 pronounced allele frequency differentiation across global populations suggestive of recent polygenic 26 selection, including variants with well-supported regulatory functions at the ESR1-CCDC170, WNT4, 27 SFR1, FOXO1, ITPR1, DMRT1 and CDKN2B loci. Notably, we show that most derived alleles acquired 28 during recent human evolution increase risk across multiple uterine disorders and may evolve under 29 antagonistic selection. These findings provide functional annotation and population-level prioritization 30 of genetic variants influencing multiple uterine disorders . They also highlight how past evolutionary 31 histories may contribute to population differences in uterine disease prevalence and pathogenesis. 32 33 34 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 2

1 Menstrual abnormalities are important indicators of reproductive pathological conditions such as 2 endometriosis, endometrial cancers or recurrent miscarriages (Gao et al. 1987; Oehler et Rees 2003; 3 Fraser et al. 2011; Bourdon et al. 2021) . Despite their clinical significance, the genetic architecture 4 underlying abnormal menstrual bleeding risk and its association with other uterine disorders remain 5 poorly understood. Over the past 15 years, large-scale genome-wide association studies (GWAS) have 6 identified hundreds of genetic variants associated with different uterine disorders (X. Wang et al. 2022; 7 Buyukcelebi et al. 2024; Rahmioglu et al. 2023; Thibord et al. 2025) . For instance, more than 80 loci 8 have been shown to contribute to the polygenic architecture of endometriosis risk (Rahmioglu et al. 9 2023; Guare et al. 2025; Koller et al. 2025) . In parallel, epidemiological studies have reported that 10 several uterine disorders co-occur more frequently than expected (Soliman et al. 2016; Choi et al. 2017; 11 Ye et al. 2022; Fiore et al. 2025; La Vecchia et al. 2025) , suggesting that they involve overlapping 12 biological pathways, possibly due to common environmental causes or to shared genetic origins. In 13 support of the latter, the genetic architecture of endometriosis is correlated to that of uterine fibroids 14 (Rafnar et al. 2018a; Gallagher et al. 2019), abnormal menstrual bleeding (Rahmioglu et al. 2023), and 15 endometrial cancers (Painter et al. 2018) . Further, this shared genetic architecture across uterine 16 disorders may differ between populations due to their past demographic histories and specific selective 17 pressures. Substantial evidence supports that several uterine disorders show different prevalences 18 among women of African, Asian and European ancestries (Stewart et al. 2017a; Sinharoy et al. 2023; 19 Yamamoto et al. 2017; Kyama et al. 2007; Mecha et al. 2022). For example, uterine fibroids are more 20 frequent and appear earlier in life in women of African ancestry (Stewart et al. 2017b; Giuliani et al. 21 2020), suggesting population-level differences in genetic susceptibility that may have shaped the risk 22 of uterine diseases across populations . Altogether, genetic and epidemiological reports highlight a 23 shared genetic basis of uterine disorders, although the genetic variants and biological functions involved 24 remain unclear. 25 In this study, we combine multi-trait GWAS analyses, population genetics and functional genomics to 26 investigate the genetic foundations of shared genetic risk across 10 uterine disorders, and their 27 evolutionary history in the recent human past . Using data from European cohorts, we confirm the 28 presence of strong positive genetic correlations between major uterine disorders, consistent with an 29 intertwined genetic architecture. We identify susceptibility loci showing pronounced allele frequency 30 differences across global populations, including variants with putative regulatory effects at the ESR1-31 CCDC170, WNT4, SFR1, FOXO1, ITPR1, DMRT1 and CDKN2B loci. Several of these loci exhibit 32 signatures of recent polygenic selection, suggesting that evolutionary pressures have shaped 33 contemporary patterns of risk and protection in human populations. Altogether, our findings provide a 34 functional and population-level prioritization of genetic variants involved in multiple uterine disorders 35 and illuminate how evolutionary processes may contribute to population differences in uterine disease 36 pathogenesis. 37 38 39 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 3

1 Mapping the shared genetic architecture of uterine diseases in European populations 2 To study the shared genetic architecture of uterine pathologies, we first assessed the estimated genetic 3 correlation captured by GWAS on 10 frequent uterine disorders. We leveraged the FinnGen biobank 4 (Kurki et al. 2023) , a Finnish cohort for medical GWAS that has collected phenotyp ic information for 5 common gynaecological diseases with high numbers of cases and appropriately designed controls for 6 sex specific conditions (i.e. excluding males in this case) . We included all traits related to menstrual 7 symptoms and uterine diseases with more than 1,500 documented cases, excluding pregnancy 8 complications (Methods), resulting in the inclusion of 1 0 GWAS summary statistics (Fig. 1A, Table 9 S1). Using LD score regression (Bulik-Sullivan et al. 2015), we identified significant genetic correlations 10 between 16 pairs of uterine disorders after FDR correction, with correlation coefficients ranking from 11 0.28 (leiomyoma vs. inflammatory uterine diseases, p-val = 0.024 with BH correction for multiple testing) 12 to 0.94 (excessive frequent and irregular menstruation vs. abnormal uterine bleedings, p-val = 2.10-49) 13 (Methods, Fig. 1B, Table S2). All disorders except “uterine polyps” and “other menopausal symptoms” 14 shared significant genetic correlations with at least one other trait. These genetic correlations confirm 15 previously reported genetic correlations between endometriosis, leiomyoma and abundant menses in 16 Japanese (Masuda et al. 2020) and European cohorts (Rahmioglu et al. 2023; McGrath et al. 2023) , 17 while others are novel (e.g. endometriosis and post-menopausal bleeding), revealing that the genetic 18 burden of common uterine dysfunctions is intricately linked across conditions. 19 Genetic correlation is a single summary measure that captures the overlap in shared genetic 20 determinants across pairs of traits averaged over the entire genome , and it cannot identify specific 21 genetic loci contributing to the shared genetic risk of uterine diseases. To further identify specific genetic 22 factors shared between common uterine disorders, we analysed those 10 traits using joint analysis of 23 summary statistics (JASS) (Julienne et al. 2020, 2021) , a statistical software to conduct multi-trait 24 GWAS association. JASS identified a total of 2,670 variants associated with uterine traits reaching 25 genome-wide significance (p < 5.10-8), corresponding to 31 independent genomic loci (Fig. 1C, Fig. S1, 26 Table S3). Of these 2,670 pleiotropic SNPs associated with uterine disorders (uSNPs), 93 were not 27 genome-wide significant in any individual uterine disorders GWAS, while 2,577 were significant in at 28 least one single-trait GWAS prior to joint analysis (Fig. 1D, Fig. S 2). Further, 2,587 SNPs that are 29 significantly associated with at least one individual trait did not reach genome-wide significance in the 30 joint test, suggesting that they contribute less to the shared architecture of uterine disorders. uSNPs 31 prioritized by JASS typically exhibited larger effect sizes (b) across uterine diseases compared to SNPs 32 that were only significant in univariate analys es (Fig. 1E). Additionally, uSNPs prioritized with JASS 33 showed overall stronger positive correlations of effect sizes across most uterine diseases compared to 34 SNPs that were only significant in univariate analyses (Fig. 1F). Interestingly, jointly prioritized variants 35 showed a strong negative correlation of effect sizes between endometrial cancer and other pathologies, 36 and in particular endometriosis, leiomyoma and menses abnormalities (average r = -0.26; Fig. 1F) 37 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 4 suggesting potential antagonistic effect of some variants for the risk of endometrial cancer and other 1 uterine diseases. 2 To replicate this analysis, we used matching uterine traits from the UKBiobank on European ancestry 3 cohorts (Karczewski et al. 2025) (Methods, Table S4). While the UK Biobank contained fewer cases 4 than FinnGen for most uterine disorders, we found strong directional consistency in the effects of jointly 5 prioritized SNPs for matching traits (Table S5). Across uterine diseases, 60% to 83% of uSNPs with a 6 significant association in FinnGen demonstrate a consistent direction of effect in UKBiobank GWAS, 7 which is more than expected by chance (binomial test, p < 0.001 for all comparisons ), except for 8 spontaneous abortion and post-menopausal bleeding for which only 43% and 48% of uSNPs have 9 replicated directions of effect, respectively. 583 out of 2,670 uSNPs identified in FinnGen were also 10 significant after joint analysis using UK Biobank uterine disorders GWAS (Table S6). These results 11 confirm that we adequately captured SNPs with pleiotropic effects on commo n uterine diseases in 12 European populations. 13 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 5 1 Figure 1: Multi-trait analysis of uterine disorder GWAS. (A) numbers of cases and controls (in thousands) for 2 each of the 10 uterine disorder traits from the Finn Gen database included in this study (B). Genetic correlations 3 between traits estimated by LDscore (* p <0.05, ** p <0.01, *** p <0.001). (C). Manhattan plot of the genome-wide 4 significant uSNPs after multi -trait analysis across all 10 uterine disorders (p-val < 5.e -08; significant loci are 5 highlighted in orange). (D). Quadrant plot representing the -log10 p-value of each uSNP after the multi-trait analysis 6 against the lowest p-value across single-trait GWAS. “Sig.” means significant with a p-val < 5.e-08 threshold. The 7 x and y axis are cropped to a maximum -log10 p-value of 20 for readability (uncropped quadrant plot in Fig. S2). 8 (E). Absolute effect sizes (b) of uSNPs significant after multi-trait analysis (uSNPs, orange) and of SNPs significant 9 in single-trait GWAS that do not reach significance in the multi-trait analysis (grey). (F). Correlation of effect sizes 10 for uSNPs (top) and SNPs only significant in single GWAS (bottom) across uterine disorders (Pearson correlation). 11 Colour code represents significant correlations after BH correction for multiple testing; non-significant correlations 12 are shown in grey. 13 ** * ********* * ************ ** ** ********* ****** *** ****** ***** ****** *** ******** ********* Endometrial cancer (malignant) UterinepolypsLeiomyomaUterine inflammato ryd i s e a s e s Spontaneous abo rtion Other menopausal symptomsPost menopausal bleeding Endometriosis Abnormal uterinebleeding ExcessiveI r r e g u l a rm e n s t ruations Endometrial cancer (malignant)UterinepolypsLeiomyomaUterine inflammatoryd i s e a s e sSpontaneous abortionOther menopausal symptomsPost menopausalbleedingEndometriosisAbnormal uterinebleedingExcessiveI r r e g u l a rm e n s truations 10.50-0.5-1 Genetic correlation(LDSC estimation) B 2577/2670 2587 93/2670 E sig. after JASS novel sig. SNP after JASS sig. in GWAS before JASS non significant SNP classification F Other menopausal symptomsEndometrial cancer (malignant)Spontaneous abortionPost menopausalbleedingUterine inflammatoryd i s e a s e sPolypsLeiomyomaAbnormal uterinebleedingEndometriosisExcessiveI r r e g u l a rm e n s truations Endometrial cancer (malignant)LeiomyomaEndometriosisExcessiveI r r e g u l a rm e n s truationsAbnormal uterinebleedingPolypsPost menopausalbleedingOther menopausal symptomsSpontaneous abortionUterine inflammatoryd i s e a s e s −1−0.500.51 Correlation of effect sizeSNP sig. with JASSn.sD A Curation ofGWAS on uterinedisorders(FinnGendatabase) Bleedingsymptoms ● Abnormal uterinebleeding ● Excessive irregularmenstruations ● Post menopausalbleeding Endometrialdisorders ● Endometriosis ● Uterine polyps ● Endometrialcancer (malignant) Myometrialdisorder ● Leiomyoma Otheruterine disorders ● Spontaneous abortion ● Uterineinflammatorydiseases ● Othermenopausal symptoms 0 20 Number of individuals (thousands) Cases Controls 0 100 C 0.00.10.2Effect size (absolute value) Abnormal uterinebleeding Other menopausalsymptoms Endometriosis Polyps Endometrialcancer(malignant) Leiomyoma Excessive Irregularmenstruations Post menopausalbleedingSpontaneous abortion Uterine inflammatorydiseases JASS UNIVARIATEonly *** *** *** *** *** *** *** *** *** *** .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 6 1 Pleiotropic uterine-associated SNPs are enriched near genes involved in female reproduction, 2 development and senescence 3 We next investigate d how pleiotropic uSNPs associated with uterine disorders may affect gene 4 regulation in the uterus. Almost 98% of uSNPs are non-coding, with 60% (n=1597) falling in introns, 2% 5 in UTRs and 36% (n=962) in intergenic regions (Fig. 2A, Table S7), as expected for GWAS significant 6 SNPs. uSNPs are enriched in gene promoters compared to SNPs that are significant in single-trait 7 GWAS only (proportion test; p-val = 4.10-4; Fig. 2B), and are also enriched in promoter -enhancer 8 contacts captured by Hi-C in endometrial stromal cells (Sakabe et al. 2020) (p-val = 8.10-6; Fig. 2B). 9 We speculate that uSNPs have more pleiotropic and larger effects on uterine traits in part because they 10 more frequently fall within gene regulatory regions, and in particular gene promoters. 11 To further explore the functions of genes that may be regulated by uSNPs, we combined functional and 12 location-based information to associate uSNPs with putative target genes (Methods, Fig. S3, Table 13 S7). In total, uSNPs associate with 227 genes, including well-described loci involved in uterine diseases, 14 endometriosis and uterine fibroids such as ESR1, WNT4, WT1 or FSHB (Gallagher et al. 2019; Pavlicev 15 et al. 2022; Rahmioglu et al. 2023; Kim et al. 2024; Sheu et al. 2024b). These genes are over-expressed 16 in fallopian tubes, uterus, ovary, cervix and vagina based on GTEX transcriptomes (GTEx Consortium 17 2017), and under-expressed in several other organs such as brain, blood, muscle or pancreas (Fig. 2C; 18 Methods). This result highlights that uSNPs are enriched around genes with specialized expression in 19 female reproductive organs. 20 Gene set enrichment analysis confirms that the putative targets of uSNPs are strongly enriched in terms 21 related to reproductive and developmental processes, involving genes such as WNT4, WT1, SOX15, 22 NEURL1, FBXO5, and CFAP58 (Fig. 2D-E). The strongest enrichment was found for genes associated 23 with replicative senescence (38-fold enrichment; Fig. 2D, Table S 8), driven by canonical regulators 24 such as TERT, TP53, CDKN1A, ATM and CHEK2 (Fig. 2E). Additionally, we find enrichment in terms 25 related to ubiquitylation (e.g. BTRC, CDKN2A, FBXO5) and arsenic response (e.g. GSTO1, GSTO2, 26 SLC38A2). While those three processes are not specific to uterine or reproductive physiology, they 27 represent conserved cellular networks that may indirectly influence reproductive function. This suggests 28 that pleiotropic uterine SNPs influence both canonical reproductive pathways and broader cellular 29 processes, highlighting their potential role in linking systemic biological networks with female 30 reproductive functions. 31 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 7 1 Figure 2: Functional characterisation of uterine-associated SNPs (uSNPs). (A) Distribution of uSNPs across 2 coding and non-coding elements of the genome based on ENCODE annotations. (B) Overlap of uSNPs and single-3 trait GWAS SNPs (univariate) with functional genomic elements from different databases and from endometrial 4 regulatory profiles (Methods). Absolute numbers of SNPs overlapping the elements are represented (chi-2 test; n.s 5 p > 0.05, *** p < 0.001, BH-corrected for multiple testing). (C) Enrichment of uSNPs target genes (uGenes) in 6 organ-specific expression profiles from GTEx. Significant enrichments are represented in colour (turquoise for 7 down-regulated genes, yellow for up -regulated genes, BH-corrected for multiple testing). (D) Gene ontology 8 enrichment analysis using uSNPs target genes. Hierarchical clustering was computed using similarities between 9 GO terms (estimated with Jaccard index). € Network plot of target genes contributing to the 5 first GO enriched 10 terms. Dot size represents the number of target genes associated to each GO term. (F) Gene expression heatmap 11 of uSNPs target genes across endometrial cell types, in log2CPM scaled to the maximum by row. Only genes with 12 cell-type specific expression are represented (full expression heatmap in Fig. S4). 13 14 To investigate how uSNPs might influence uterine cell functions, we re -analysed endometrial single-15 nuclei transcriptomes of healthy donors from the HECA atlas (Marečková et al. 2023), (Fig. S4A). This 16 B Endometrial ATAC peaks CTCF Enhancer like structure Epithelial HiC maps REMAP peaks all Stromal HiC maps REMAP peaks uterus Allele specific binding eQTL Promoter 0 500 1000 1500 number of SNPs ***nsnsns*** ns ns nsnsns uSNP single-trait SNP C Adipose TissueAdrenal Gland Bladder Blood Blood Vessel Brain Breast Cervix Uteri Colon Esophagus Fallopian Tube Heart Kidney Liver Lung Muscle Nerve Ovary Pancreas Pituitary ProstateSalivary Gland Skin Small Intestine Spleen Stomach Testis Thyroid Uterus Vagina 042 86-log10 (P-value) Enrichment of uGenesinup-regulated DEG 846 02-log10 (P-value) Enrichment of uGenes indown-regulated DEGA ED number of genes5.07.510.0 12.5 p.adjust 0.010.020.030.04 Ubiquitilation Developmental processes involvedin reproduction DNA damage and senescence Response to arsenic Developmental morphogenesis Developmental processes replicatives e n e s c e n c ecellular senescence cellular response to gammaradiation telomere maintenance via telomeraseRNA−templated DNA biosynthetic process sexd i fferentiationmale sexd i fferentiation gonad development female gonad development development of primarysexual characteristics regulation of ubiquitin protein ligase activity negativer e g u l a t i o no fu b i q u i t i np r o t e i nl i g a s ea c t i v i t ynegativer e g u l a t i o no fu b i q u i t i n − p r o t e i ntransferase activity regulation of ubiquitin−protein transferase activity SCF−dependent proteasomal ubiquitin−dependent protein catabolic process gland morphogenesis response to arsenic−containing substance response to starvation synapse pruning salivaryg l a n ddevelopment cellular response to arsenic−containing substance patterns p e c i fi c a t i o np r o c e s s cellular process involved in reproduction inmulticellular organism regulation of gonad developmentgermc e l ldevelopment oocyte developmentoocyte differentiation regulation ofnuclear division F Coding (2%) Intronic (60%) Splicing (n=1) 5'UTR (0.5%) Intergenic >1kb of genes (36%) Downstream gene (<1kb), (1%) Upstream gene (<1kb), (1%) ncRNA exon (1%) exonic (1%) 3'UTR (2%) Non-Coding (98%) CiliatedGlandular SecretoryLuminalpreGlandularProliferativeEndothelialLymphaticLymphoidMyeloidDecidualPerivascularNon-decidualProliferative EpithelialStromal CDKN2BCDKN1AITPRIPIFITM1CALHM2NEURL1SORCS1FGF11SHBGC22orf31SORCS3SOX15 SULT1E1IFNEALOXE3TERT MYCT1VIPLDLRAD2ARHGEF15KDRSULT1B1 CCDC181COL17A1NKD2C2orf50CFAP58DNAH2 ASGR2DEPDC7F5CD68C1QCC1QBC1QA SLC25A35ALPLUSP44CFAP43TNFSF13CDC42−AS1TMEM102NTN4TSPAN12PKP3EXPH5GPR160PRRG4MIR31HG DOCK8CCDC73SLC38A1STN1ALOX12SLC35F2CD109MPPED2GSTO2KANK1ATP1B1LINC00339CCDC170DPCDC11orf65FBXO5CHEK2CEP72CDKN2AHPS6WRAP53 0 0.2 0.4 0.6 0.8 1 Gene expression(rowmax oflog2CPM counts) replicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c ereplicatives e n e s c e n c e sexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiationsexd i fferentiation regulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activityregulation of ubiquitin protein ligase activity SCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic processSCF−dependent proteasomal ubiquitin−dependent protein catabolic process cellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescencecellular senescenceATMATMATMATMATMATMATMATMATMATMATMATMATMATMATMATMATM TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53TP53 CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2CHEK2 TERTTERTTERTTERTTERTTERTTERTTERTTERTTERTTERTTERTTERTTERTTERTTERTTERTCDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1ACDKN1A CDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2ACDKN2A WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4WNT4FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8FGF8 CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1CYP17A1 FSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHBFSHB WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1WT1 SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15SOX15 KDRKDRKDRKDRKDRKDRKDRKDRKDRKDRKDRKDRKDRKDRKDRKDRKDR ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1ESR1 DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT1DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT3DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2DMRT2 BTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRCBTRC USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44USP44 FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5FBXO5 FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXW4FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15FBXL15 FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3FBXO3 CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5CUL5 CDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2BCDKN2B Number of genes46810 12 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 8 showed that 74% of uSNP target genes are expressed in at least one uterine cell type (Fig. S4B). Many 1 of these genes, including well-documented loci such as ESR1, WT1, and WNT4, are broadly expressed 2 across multiple uterine cell types and likely have ubiquitous effects across the reproductive system. 3 More interestingly, several target genes displayed cell -type–specific expression ( Fig. 2F ). For 4 example, CDKN2B, CDKN1A, ITPRIP and IFITM1 were preferentially expressed in decidual stromal 5 cells, while CCDC181, COL17A1, CFAP58 and DNAH2 were strongly expressed in ciliated epithelial 6 cells. This observation suggests that uSNPs may also capture pleiotropic effects resulting from altered 7 functions in restricted cell populations. 8 As expected for pleiotropic loci, most uSNP target genes were also expressed in tissues beyond the 9 uterus. Among the few targets not detected in healthy uterus, several showed highly tissue -specific 10 expression in the testis, pituitary or brain (e.g., FSH, INA, CYP17A1) (Fig. S5). These patterns suggest 11 that some uSNPs may exert pleiotropic effects on uterine traits via systemic endocrine or 12 neuroendocrine pathways, or that certain genes may become aberrantly expressed in the uterus under 13 pathological conditions. 14 15 Substructure of uSNPs with risk-increasing or protective effects on uterine disorders 16 We next investigated how the shared genetic architecture of uterine diseases is structured into groups 17 of variants with similar effects on multiple disorders. To obtain a set of largely independent variants, we 18 reduced the full set of 2,670 uSNPs to 235 lead uSNPs in pseudo -independent linkage disequilibrium 19 blocks (R2 < 0.5), and identified the ancestral and derived alleles for each lead uSNP (Methods). Then, 20 we performed unsupervised K-means clustering on effect sizes (b) of lead uSNP derived alleles across 21 traits (Methods). This analysis identified an optimal clustering with two main clusters representing 22 uSNPs for which the derived allele has an overall protective effect (negative effect size) across most 23 uterine disorders (cluster 1), or has an overall risk-increasing effect (cluster 2; positive effect size) (Fig. 24 3A, Table S9). Risk-increasing cluster 2 contained twice as many lead uSNPs as protective cluster 1, 25 suggesting that recently-acquired alleles are more frequently deleterious in terms of uterine disorder 26 risk. 27 To further explore how protective and risk -increasing loci associate with different disorders , we 28 computed pairwise correlations of normalized effect sizes between disorders for each cluster (Fig. 3B). 29 Variants in cluster 1 had correlated and protective effects on endometriosis and different menses and 30 menopause disorders, but negative correlation s and deleterious effects with other traits such as 31 endometriosis and leiomyoma – suggesting that these variants tend to be protective for either disease 32 but not both. Variants in cluster 2 had overall correlated risk-increasing effects on most disorders, with 33 the notable exception of malignant endometrial cancer and leiomyoma , which were negatively 34 correlated with endometriosis. Specifically, we identified a n antagonistic pleiotropic relationship 35 between endometriosis and malignant endometrial cancer , where derived alleles at lead uSNPs in 36 cluster 2 typically increased risk for endometriosis but had overall protective effects on endometrial 37 cancer, in particular at the ESR1 and GREB1 loci (Fig. 3A-B). By focusing on a subset of largely 38 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 9 independent SNPs that associate with multiple disorders, this analysis breaks down the genome-wide 1 genetic correlations identified in Fig. 1B into more nuanced components with correlated or antagonistic 2 pleiotropic effects between disorders. 3 This analysis highlighted a locus of particular interest on chromosome 10 (10.q25.1), containing a 4 remarkably high concentration of pleiotropic uSNPs of large effect sizes on multiple disorders , 5 especially leiomyoma, excessive irregular menstruation and polyps (Fig. 3C ). This locus is rich in 6 putative non-coding regulatory elements active in the endometrium (accessible in single-cell ATAC-seq 7 data from human endometrium (Vrljicak et al. 2023), and corresponding to promoter-enhancer contacts 8 in decidual stromal cells assessed by promoter-capture Hi-C (Sakabe et al. 2020)). rs4387287*C and 9 rs111447985*C, which overlap with CTCF binding sites and are predicted to enhance the binding of the 10 BRD4 transcription factor, respectively, are notable in this locus. Both SNPs are situated within the 11 promoter region of the STN1 gene, a chromatin region that is accessible across various endometrial 12 cell types (Fig. 3C, Table S7). Remarkably, these SNPs coincide with 116 Hi-C contacts in decidualized 13 endometrial stromal cells, establishing connections notably with the promoters of SFR1, CFAP43, and 14 ITPRIP, suggesting robust regulatory activity of this chromatin region across endometrial cell types. In 15 healthy endometrium, the STN1 gene demonstrates consistent expression across different endometrial 16 cell types, with particularly elevated levels in endometrial glandular cells and immune cells ( Fig. 2E). 17 STN1 is known to play a crucial role in cancer progression, acting as an upstream regulator of the 18 epithelial-to-mesenchymal transition (Nguyen et al. 2023; Dong et al. 2025). Notably, previous GWAS 19 studies have associated this locus with the risk of endometriosis (Sheu et al. 2024a), and our findings 20 suggest that variants in this locus may have pleiotropic effects on other uterine disorders, such as 21 uterine leiomyomas. While the functional roles of these candidate uSNPs in endometrial cells await 22 experimental validation, they have the potential to disrupt the regulatory landscape of STN1 expression, 23 as well as those of SFR1, CFAP43, and ITPRIP. 24 25 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 10 1 2 Figure 3: Substructure of uSNPs effect sizes across uterine disorders. (A) Heatmap of effect sizes for derived 3 alleles of pseudo-independent uSNPs across uterine traits. Cluster 1 and cluster 2 were determined by k-means 4 clustering on uSNPs effect sizes. For each cluster, uSNPs are ordered by genomic position and labelled by their 5 nearest gene name. (B) Correlations of uSNPs effect sizes across uterine traits for cluster 1 (top) and cluster 2 6 (bottom). Only significant correlations are represented (BH-corrected for multiple testing). (C) Representation of 7 the SFR1 locus showing normalised effect sizes of derived alleles and functional annotations of uSNPs. 8 9 Evidence of polygenic selection at uterine disease risk loci in European populations 10 Because genetic variants affecting reproductive functions may be particularly accessible to natural 11 selection due to their direct effect on fitness, we next investigated whether the shared genetic 12 architecture of uterine disease risk displays evidence of polygenic selection in human 13 populations(Pritchard et al. 2010; Pritchard et Di Rienzo 2010; Stephan et John 2020). We hypothesized 14 that uSNPs may be under stronger evolutionary pressures than background SNPs because they may 15 have pleiotropic effects on gene expression and therefore be more likely to evolve under selection 16 (Pavlicev et Wagner 2012; Hämälä et al. 2020) . We compared the fixation index ( Fst) distributions 17 between the 191 lead uSNPs and 1,000 sets of control SNPs match ed for recombination rate, minor 18 allele frequency (MAF) and local gene density, in populations from the 1000 Genomes Project (The 19 1000 Genomes Project Consortium et al. 2015) (Methods). Fst measures allelic differentiation between 20 pairs of populations and is increased for variants that experienced past positive selection (Holsinger et 21 Weir 2009; Stephan 2016). The Fst distribution of lead uSNPs is shifted towards higher Fst values in the 22 2.q11.2−LYG1 17.p13.1−MYH10 16.p13.11−MYH11 9.p24.3−KANK1 5.p15.33−TERT 2.p25.1−GREB1 1.p36.12−WNT4 7.q31.31−TSPAN12 11.q22.3−ATM 4.q12−KDR 4.q13.3−SULT1E1 6.q25.1−ESR1 9.p21.3−CDKN2B 10.q25.1−SFR1 11.p13−WT1 13.q14.11−FOXO1 17.p13.1−TP53 22.q12.1−TTC28 22.q13.1−TNRC6B 3.p26.1−ITPR1 6.q13−CD109 Endometriosis Leiomyoma Abnormal uterine bleeding ExcessiveI r r e g u l ar menstruations Spontaneous abortion Post menopausal bleeding Other.menopausal symptoms Uterine inflammatoryd i s e a s e s Polyps Endometrial cancer malignant 11.q22.3−ATM 12.q22−VEZT 11.p14.1−FSHB 11.p15.5−BET1L 5.p15.33−TERT 4.q12−KDR 2.p14−MEIS1 2.p14−ETAA1 3.q26.2−MYNN 1.q24.2−ATP1B1 1.p36.12−WNT4 2.p25.1−GREB1 10.q25.1−SFR1 9.p24.3−KANK1 11.p13−WT1 9.p21.3−CDKN2B 6.q25.2−ESR1 17.p13.1−TP53 13.q14.11−FOXO1 22.q12.1−TTC28 Effect-size −0.5 0 0.5 1 Cluster 2Cluster 1 B A C Leiomyoma Abnormal uterine bleeding ExcessiveI r r e g u l a r Menstruation Spontaneous abortion Post menopausal bleeding Other menopausal symptoms Uterine inflammatory diseases Polyps Endometrial cancer malignant. Endometriosis Correlation of effect-size 0.25 0.50 0.75 1 Positive Negative Endometriosis Leiomyoma Abnormal.uterine.bleeding Excessive.Irregular.menstruations Spontaneous.abortion Post.menopausal.bleeding Other.menopausal.symptoms Uterine.inflammatory.diseases Polyps Endometrial.cancer..malignant. In promoter-enhancer contact of decidual stromal cells Allele specific binding Accessible in endometrial cell types Effect-size −0.5 0 0.5 1 STN1 SLK COL17A1 SFR1 CFAP43 non-significant uSNPs pHiC dStromal CTCF rs4387287 rs111447985 rs73331554rs74154737 0 20 40 60 103900000 104000000 104100000 104200000 logp Leiomyoma Abnormal uterine bleeding ExcessiveI r r e g u l a r Menstruation Spontaneous abortion Post menopausal bleeding Other menopausal symptoms Uterine inflammatory diseases Polyps Endometrial cancer malignant. Endometriosis .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 11

European population (CEU) compared to the reference East Asian (CHB) and African (YRI) 1 populations (Fig. 4A; CEU vs. CHB, p-val = 0.024, CEU vs. YRI, p-val = 0.0009; Wilcoxon test with BH 2 correction for multiple testing). In contrast, we detected no significant difference in Fst values between 3 YRI and CHB populations. We next compared the integrated haplotype scores (iHS) of lead uSNPs and 4 controls in each population, which measures the local strength of selection based on haplotype 5 homozygosity to detect more recent selection events (Voight et al. 2006a). uSNPs have overall higher 6 absolute iHS than matched controls in Europeans (CEU), although the signal does not reach 7 significance after multiple testing correction (Fig. 4B; p-val = 0.087; Wilcoxon test with BH correction 8 for multiple testing). These results suggest that uterine disease risk loci captured from European cohorts 9 have likely been influenced by past selective pressures resulting in polygenic allelic differentiation 10 between human populations. 11 12 rs12035094CDC42rs9315762LINC00332rs9371246SYNE1rs1408459SYNE1rs1343987 RP11−415K20.1rs9383580CCDC170rs2623964SYNE1rs1971256CCDC170rs10815717DMRT1rs1392916CD109rs16922717KANK1rs11191835SH3PXD2Ars2779747 RP11−408N14.1rs12638862TERCrs1052001SFR1rs2473290LINC00339rs3820282WNT4rs1265005SFR1rs13028479GREB1rs7571803GREB1rs2488002OBFC1rs60622800MIR4457 −0.50−0.250.000.250.500.75 rs5752763TTC28rs12213214VIPrs2207548 RP1−65P5.1rs13218956SYNE1 signedFST CHBhigher DAF inYRI rs12499134SULT1D1Prs9383889RMND1rs2982556ESR1rs2779747RP11−408N14.1rs10835891RP1−65P5.1rs12213214VIPrs12250162SLKrs1392916CD109rs961605RP4−562D20.2rs60622800MIR4457rs9397414CCDC170rs1052001SFR1rs7571803GREB1rs13028479GREB1rs2488002OBFC1rs1971256CCDC170rs1265005SFR1 −0.50−0.250.000.250.500.75 rs5752763TTC28rs6503048TP53 signedFST CEUhigher DAF inYRI −0.50−0.250.000.250.500.75 rs3804984ITPR1rs2347792C22orf31rs1095900 RP11−408N14.1rs35194179SYNE1rs10976019KANK1rs11795218KANK1rs12173791CCDC170rs931423SYNE1rs11191835SH3PXD2Ars12035094CDC42 rs2324596RP11−172E9.2rs9566498LINC00332rs16922717KANK1rs11791986KANK1:RP11rs851983ESR1rs57295696KANK1rs607987FSHBrs9397414CCDC170rs2982556ESR1rs9383580CCDC170rs7854629CDKN2B−AS1rs10835893WT1 CEUhigher DAF in signedFST CHB CEU CHB YRI 0 1 2 3|iHS| p=0.087p=0.54 p=0.54 ControlsControlsControlsuSNPsuSNPsuSNPs B 0.0 ControlsControlsControlsuSNPsuSNPsuSNPs 0.1 0.2 0.3 0.4 0.5FST p=0.024 p=0.0009 p=0.78 CEUvs.YRICEUvs.CHB CHBvs.YRIA Lead SNPs not in outliers top5_CEU top5_CHB top5_YRI FST outliers C D E .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 12 Figure 4: Evolutionary statistics of uSNPs in reference European, African and East-Asian populations. (A-1 B) Distributions of F st (A) and iHS values (B) at pseudo-independent lead uSNPs compared to 1,000 sets of 2 matched control SNPs (Wilcoxon rank sum test, BH-corrected for multiple testing). (C-D-E) Signed Fst values of 3 the top pseudo-independent lead uSNPs for all three population comparisons: (C) CEU (Northern European from 4 Utah) vs. CHB (Chinese Han from Beijing); (D) CEU vs. YRI (Nigerian Yoruba from Ibadan); (E) YRI vs. CHB. 5 uSNPs in the top 5% genome-wide F st values are represented in colour; signed F st scores point towards the 6 population where the derived allele frequency is the highest. 7 8 To highlight specific uSNPs of interest that might have experienced stronger events of selection, w e 9 next focused on uSNPs with outlier Fst values (top 5%), which exhibit very strong allelic differentiation 10 between populations and are prime functional candidate loci, similarly to previous approaches (Quach 11 et al. 2016). We find that 50 uSNPs out of 191 uSNPs tested are Fst outliers in at least one comparison 12 (23%, p-val = 0.0004, binomial test, Table S10, Methods). Moreover, uSNPs are over-represented in 13 Fst outliers in both the CEU vs. CHB and CHB vs. YRI comparisons (Fig. 4C-E; CEU vs. CHB, p-val < 14 2.2.10-16; CEU vs. YRI, p-val = 0.12; CHB vs. YRI, p-val = 0.01; empirical test with BH correction for 15 multiple testing, Methods). We compared the GWAS effect sizes of uSNPs with outlier Fst values to 16 non-outlier uSNPs and found no significant differences between groups (Fig. S7). 17 Amongst uSNPs with high Fst values, we prioritized a list of candidate variants with putative functional 18 impact on uterine cell types as described in Fig. 2. This analysis highlights 17 uSNPs with both evidence 19 of allelic differentiation across human populations and functional importance for uterine cell functions, 20 spanning the ESR1, WNT4, SFR1, FOXO1, KANK1 and CDKN2B loci (Table 1). As CDKN2B, ITPR1 21 and WNT4 are mostly expressed in endometrial stromal cells, and DMRT1 and CCDC170-ESR1 in 22 epithelial cells, we speculate that variants at these loci may mostly affect these cell types. Interestingly, 23 variants at the SFR1 locus, and presented in Fig. 3C, also show evidence of allelic differentiation across 24 populations and display the highest Fst score of all tested uSNPs (Fig. 4D). Additionally, rs3820282 at 25 the WNT4 locus was recently highlighted as a pleiotropic variant with beneficial roles in gestation length 26 and pre -term birth while increasing the risk of endometriosis, breast cancer and ovarian cancer 27 (Pavlicev et al. 2022). Functional validation in human cell lines and a transgenic mouse model confirmed 28 that rs3820282 increases the binding of ESR1 and induces an over-expression of WNT4 in endometrial 29 stromal cells (Pavlicev et al. 2022). 30 Altogether, our results support that a sizeable fraction of uSNPs display evolutionary signature s of 31 positive selection in European populations, and further prioritize candidate SNPs within these genomic 32 regions with joint evolutionary and functional signatures. 33 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 13 1 Table 1: Prioritized susceptibility loci harbouring candidate uSNPs with signatures of strong allelic 2 differentiation and with regulatory potential in endometrial cells. The full table of functional and evolutionary 3 annotation of those variants is available in Table S10. pELS: proximal enhancer -like structure; dELS: distant 4 enhancer-like structure; Epi: epithelial; Str: stromal; Im: immune; Ed: endothelial. 5 6 The ESR1-CCDC170 locus displays strong evidence of positive selection in European 7 populations 8 Among genetic loci with evidence of recent selective pressures in human populations, the genomic 9 region including both the CCDC170 and ESR1 genes is of particular interest as it carries a high number 10 uterine-relevant SNPs with outlier Fst values (Table 1, Fig. 5A). Notably, rs1971256 displays the 11 strongest Fst value of this locus (CEU vs. YRI, Fst =0.607; CHB vs. YRI, Fst = 0.331) and the highest 12 score of pathogenicity (CADD = 9.988, Table 1 , Methods ). Additionally, this locus exhibits well-13 supported evidence of selective sweeps in the last 100,000 years across all European populations and 14 in one African population, which constitute rare events of strong positive selection (Laval et al. 2021) 15 (Fig. 5C). Indeed, we observe that the rs1971256 derived allele (T) has become the major allele in most 16 human populations, except those of African descent (Fig. 5D). This variant was previously prioritised 17 as increasing the risk of endometriosis as well as endometrioid and clear cell ovarian cancers (Sapkota 18 et al. 2017a; Mortlock et al. 2022) , but its functional impact on endometrial cell types has not been 19 investigated. In this study, we found that the derived allele of rs1971256 has a pleiotropic protective 20 role across multiple uterine traits, notably endometriosis, abnormal menstrual bleeding, excessive and 21 frequent menstruation and inflammatory uterine diseases (Fig. 5B). This variant is in the promoter of 22 CCDC170, and additionally displays 22 significant Hi-C contacts with the ESR1 promoter and introns 23 identified in decidual stromal cells (Fig. 5C). Additionally, this region is in accessible chromatin across 24 several endometrial cell types and is a predicted CTCF binding site by ENCODE (Dunham et al. 2012) 25 (Fig. 5C). In healthy endometrium, CCDC170 is particularly expressed in proliferating endometrial 26 epithelial cells, pre -glandular cells and ciliated cells , and ESR1 is expressed across most cell types 27 during the proliferative phase (Fig. 5E). Other uSNPs with outlier Fst value in this locus are accessible 28 CEUvs CHB CEUvs YRI CHBvs YRI EpiS tr Im Ed rs2473290 C T 0.119 0.212 0.474 4.146 CDC42 ,LINC00 339;CDC42- AS1;LINC01635 1 XXX CDC42- AS1;CDC4 2;LINC003 39;LINC01 635 rs3820282 C T 0.171 0.195 0.489 20.3 WNT4 X X pELS rs12250162 C C 0.012 0.324 0.252 1.105 SFR1,COL17A1 1 COL17A1 rs1265005 A A 0.052 0.637 0.546 7.47 SFR1 X 13.q14.11 rs9315762 C T 0.056 0.143 0.288 3.009 FOXO1 X 3.p26.1 rs3804984 T T 0.387 0.028 0.241 1.756 ITPR1 X X rs1971256 C C 0.115 0.607 0.331 9.988 CCDC170, ESR1 22 X XXX pELS,CTCF rs9383580 A G 0.356 0.013 0.325 2.802 ESR1, 1 X rs851983 A G 0.259 0.115 0.039 4.955 ESR1, 1 X X dELS,CTCF rs9371246 G G 0.105 0.068 0.301 1.746 C6orf211 1 XXX X rs1408459 T C 0.087 0.096 0.313 4.512 SYNE1 X dELS,CTCF rs11795218 G A 0.278 0.198 0.009 9.516 KANK1 X rs11791986 A G 0.242 0.231 -0.005 6.901 DMRT1 X XX dELS rs10976019 A G 0.287 0.044 0.133 9.458 DMRT1 X XX dELS rs10815717 G A 0.189 0.048 0.369 20.3 DMRT1 X X dELS rs7854629 A G 0.359 0.114 0.118 0.598 CDKN2B 1 rs2779747 G T 0.025 0.282 0.42 0.854 CDKN2B 1 dELS 9.p21.3 Target genes eQTL 1.p36.12 10.q25.1 6.q25.1 9.p24.3 snATAC ENCODE anno tation PCHi-C stromal REMAP uteruscytoband ID AncestralM inor allele FST CADD .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 14 in endometrial epithelial cells preferentially ( Table 1), which additionally suggests that this genomic 1 locus might contribute to uterine disease pathogenesis by affecting endometrial epithelial cells. Future 2 functional studies may reveal how these genetic variants that have reached higher frequencies across 3 many human populations affect the physiology of endometrial cells and protect against endometriosis 4 or abnormal uterine bleeding. 5 6 Figure 5: Evolutionary and functional annotation of the CCDC170-ESR1 locus. A. Regional plot highlighting 7 significant uSNPs with outlier Fst values in the CEU population (in blue) and with functional evidence of regulatory 8 activity in endometrial cells (circled in red). B. Heatmap of the effect sizes of the derived allele at rs1971256 across 9 uterine disorders. C. Annotation of selective sweeps at the CCDC170-ESR1 locus from Laval et al. 2021 across 10 1000 Genomes populations, overlaid with open chromatin peaks identified across endometrial cell types, CTCF 11 binding sites from ENCODE and promoter Hi -C contacts captured in decidual stromal cells. D. Global map of 12 rs1971256 allele frequencies across populations included in the 1000 Genomes Project. E. Average gene 13 expression of CCDC170 and ESR1 across endometrial cell types using single-nuclei RNA-seq from healthy human 14 endometrium. 15 Endometrial cancer (malignant) Polyps Uterine inflammatoryd i s e a s e s Other menopausal symptoms Post menopausalbleeding Spontaneous abortion Leiomyoma Excessive Irregular menstruations Abnormal uterinebleeding Endometriosis −0.2 −0.1 0.0 0.1 Effect size rs1971256-e f f e c ts i z eo ft h ed e r i v e da l l e l e( T ) protectiverisk Refseq GenesRMND1ARMT1CCDC170ESR1CEUFINGRBsweep inEuropeanpopulations sweep inAfricanpopulations sweep inEast Asianpopulations IBSTSIYRIESNGRWLWKMSLCDXCHBCHSJPTKHVRefseq GenesRMND1ARMT1CCDC170ESR1ProgenitorsGlandsLuminalCiliatedStromalEndothelialImmuneCTCFpHiC map indecidual cells CCDC170ESR1StromalEpithelial_preGlandularEpithelial_CiliatedEndothelialStromal_perivascularLymphaticEpithelial_GlandularSecretoryImmune_LymphoidEpithelial_LuminalEpithelial_ProliferativeStromal_ProliferativeImmune_MyeloidStromal_decidual Percent Expressed255075 Average Expression −1 0 1 E noyes 0 10 20 30 40 151400000151500000151600000151700000chr6 logp Regulatory potentialin uterine cells Fstoutlier in CEUnoyes rs1971256 A C B Ancestral AFDerived AF Ancestral allele :CDerivedallele :Trs1971256 D .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 15

1 In t his study , we explored how combining uterus-centric GWAS summary statistics can illuminate 2 shared genetic and cellular mechanisms of uterine dysfunction . We integrated the polygenic 3 architecture of ten uterine disorders – most of them associated with menstrual symptoms – and 4 confirmed extensive, genome-wide genetic correlations across these disorders. Further, we prioritized 5 31 major susceptibility loci with pleiotropic effects contributing to this shared genetic burden and 6 potentially involved in the pathogenesis of multiple uterine disorders based on functional and 7 evolutionary evidence. Compared to previous work investigating genetic correlations between pairs of 8 diseases (Painter et al. 2018; Rafnar et al. 2018a; Gallagher et al. 2019; Masuda et al. 2020; Rahmioglu 9 et al. 2023) , this approach integrates evidence across multiple uterine pathologies into a unified 10 statistical framework to reveal genetic loci impacting core pathways within the uterus . To date, few 11 studies have investigated how organ-centric integration of GWAS statistics can highlight genetic 12 variants of functional significance to organ function and dysfunction (Sheng et al. 2021; Levin et al. 13 2022; J. Song et al. 2024; X.-Y. Wang et al. 2025). 14 The substantial genetic correlations observed among st uterine disorders are consistent with prior 15 evidence that gynaecological disease susceptibility frequently involves overlapping biological pathways 16 broadly related to senescence, hormonal response and reproductive development mechanisms 17 (Sapkota et al. 2017b; Rafnar et al. 2018b; Bulun et al. 2019; Cardoso et al. 2020; Bulun et al. 2025) . 18 Further, our results show that most genetic variants identified by our approach have congruent effects 19 and either decrease or increase risk across diseases, consistent a fundamental aetiology common to 20 uterine disorders. One exception was endometrial cancer, which was typically anticorrelated with other 21 uterine disorders, capturing antagonistic genetic effects. A major hurdle to understand disease 22 mechanisms underlying this shared genetic risk is that most GWAS variants are likely non-functional or 23 their functional potential is rarely investigated in detail. Here, to support biological interpretation, we 24 integrated a broad array of functional genomics resources to clarify how these variants may impact 25 gene expression across tissues and specifically in the uterine endometrium . Our results reveal that 26 pleiotropic variants affect loci harbouring genes whose expression is highly biased towards female 27 reproductive tissues. Further, most of these genes are highly expressed in endometrial cells, confirming 28 that this multi-trait analysis strategy captures variants within loci of direct interest for uterine biology. 29 Our results suggest that at least a subset of shared susceptibility loci act by disrupting gene regulation 30 in different uterine cell types . Although functional annotation alone cannot establish causality, the 31 convergence of genetic association s, regulatory context in uterine cell s, and uterine expression of 32 nearby genes strengthens the plausibility that these loci contribute mechanistically to uterine disease 33 pathogenesis, rather than reflecting indirect systemic effects. However, experimental validation will be 34 required to resolve causal variants and target genes and confirm the functional consequences of these 35 variants. 36 To understand how mutations that arose recently in human history contribute to uterine disorder risk, 37 we investigated the directionality and effect sizes of derived alleles at pleiotropic loci. To our knowledge, 38 this is a novel approach as most studies define reference and alternative alleles arbitrarily and therefore 39 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 16 rely on absolute effect sizes only, ignoring directionality of effects. Considering the risk contributions of 1 ancestral and derived alleles revealed that recently acquired alleles were almost twice more likely to 2 increase disease risk. Overall, our findings support that most recent mutations in the human genome 3 that impact uterine functions are deleterious. The relatively high frequencies at which some of these 4 derived alleles are present in human populations suggests that either this deleterious effect is too weak 5 to be purged by natural selection, or these variants may have been influenced by antagonistic selection 6 acting on other genetically correlated traits. 7 To further investigate how these pleiotropic variants have been influenced by past selective pressures, 8 we considered different metrics of natural selection based on allelic frequenc ies and haplotype 9 homozygosity in populational data (Weir et Cockerham 1984; Voight et al. 2006b; Pritchard et al. 2010). 10 Our results support that pleiotropic variants acting on uterine disorder risk present signatures of 11 selection in European populations. This signal was stronger when using the Fst score compared to the 12 iHS score, which detects independent events of rapid selection (Voight et al. 2006b), suggesting that 13 allelic differentiation at these loci more likely reflects polygenic selection. We however note that 14 alternative explanations such as demographic history, background selection, or genetic drift cannot be 15 entirely excluded. Moreover, the modest number of loci identified in our study (31 loci represented by 16 235 pseudo-independent variants) limits statistical power relative to large -scale studies of polygenic 17 adaptation in traits such as height or metabolic phenotypes (Berg et al. 2019; Choin et al. 2021; Kun et 18 al. 2023). Nonetheless, evidence of consistent allelic differentiation across multiple loci suggests that 19 uterine disorder genetic risk has been shaped by recent population history, and that the risk architecture 20 captured in European populations may not reflect that of other populations, highlighting the importance 21 of expanding GWAS studies to population samples of diverse genetic ancestries. 22 Our observations overall suggest that, despite being on average deleterious, pleiotropic variants 23 impacting uterine disease risk tend to be selected by polygenic adaptation processes. This ties in with 24 theories of antagonistic selection at pleiotropic loci, where pleiotropic variants are adaptative for early-25 life traits and reproductive success but become deleterious in later life stages, particularly by increasing 26 disease burden (Rose 1982; Pavlicev et Wagner 2012). This interpretation is supported by observations 27 that some uterine disorders are genetically correlated with traits related to fertility, such as age at 28 menarche, or to pregnancy (Rahmioglu et al. 2023; Benonisdottir et al. 2024; Venkatesh et al. 2025; 29 Pujol Gualdo et al. 2025). Previous work has reported that variants affecting reproductive traits are often 30 located near genes involved in cellular aging and lifespan regulation (Rodríguez et al. 2017a; D. Wu et 31 al. 2022), which we also confirm here with an enrichment of senescen ce-related genes close to the 32 genetic variants prioritized by our approach. Human genetic studies increasingly document genetic 33 correlations between reproductive traits and later-life disease risk, although the molecular mechanisms 34 underlying these relationships remain poorly characterized (D. Wu et al. 2022; Chen et Zhang 2020; 35 Rodríguez et al. 2017b). 36 Our study comes with a number of limitations. First, we did not evaluate associations between prioritized 37 variants and other reproductive phenotypes such as age at menarche, age at first birth or gestational 38 duration, as those traits are not included in the Finngen biobank. However, several of these traits have 39 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 17 previously shown evidence of selection favouring earlier reproductive maturation (Lahdenperä et al. 1 2004; W. Song et al. 2021; Long et Zhang 2023; Xiang et al. 2024) . Future work integrating multi-trait 2 GWAS across reproductive life -history traits may clarify whether recently acquired alleles increasing 3 uterine disorder risk are also associated with reproductive advantages. Another limitation of this work 4 is that GWAS data used in this study were derived from individuals of European ancestry, limiting the 5 generalizability of both pleiotropic associations and population genetic analyses. Additionally, 6 diagnostic heterogeneity across uterine disorders in biobank-based phenotyping likely includes cases 7 misclassified as controls, which likely reduces power to detect significant effects for some traits. Indeed, 8 while the Finngen biobank includes more uterine disorder cases compared to other biobanks (Kurki et 9 al. 2022; Constantinescu et al. 2022), the proportions of cases vs. controls remain lower to the estimated 10 prevalence for several uterine disorders in European populations. For instance, uterine fibroids have 11 an estimated prevalence of ~70% in women after 50 years (Stewart et al. 2017b), but only represent 12 17% of patients in the Finngen cohort. 13 In summary, our study explores the shared genetic architecture of uterine disorders and identifies 14 31susceptibility loci with pleiotropic effects, several of which show marked population differentiation and 15 regulatory activity in uterine tissues. These findings pave to the way to explore common biological 16 pathways underlying multiple gynaecological conditions and suggest that some components of uterine 17 disease risk are shaped by population-specific genetic histories. Expanding multi-trait and evolutionary 18 analyses to larger number of traits related to female reproductive functions and to cohorts from diverse 19 ancestry, will be critical for clarifying the generality and biological significance of these genetic variants. 20 21

22 Selection of GWAS on uterine disorders in the Finngen database 23 To explore the genetic architecture of uterine disorders, we leveraged available summary statistics from 24 FinnGen, a hospital cohort with well -defined phenotypes uterine disorders, including sex-specific 25 designs for male/female -specific traits, and which include s higher case counts for uterine disorders 26 compared to other publicly accessible cohorts such as the UKBiobank (See Table S1 and Table S4). 27 FinnGen database (r.7) was used to select GWAS summary statistics related to uterine traits and 28 diseases using “uterine” , “endometrium”, “menstruation”, “fertility”, “menarche”, “menopause”, 29 “leimyoma” and “bleeding” as keywords. We did not include traits and diseases related to pregnancy 30 complications as the focus of our study concerns uterine functions outside of pregnancy. Phenotypes 31 with fewer than 1,500 cases were filtered out. Endometriosis included a number of subcategories, and 32 we only retained the broad phenotype with the largest number of cases (“endometriosis”), resulting in 33 10 final uterine phenotypes: endometriosis (ICD10-N80), malignant endometrial cancer (ICD10-C54), 34 leiomyoma (ICD10-D25),, abnormal uterine bleeding (ICD10-N93), inflammatory uterine disease 35 (ICD10-N71), post-menopausal bleeding (ICD10-N95.0), other menopausal disorder (ICD10-N95.8), 36 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 18 excessive-frequent-irregular menstruation (ICD10-N92),, uterine polyps (ICD10-N84.0), and 1 spontaneous abortion (ICD10-O03) (Table S1). We retained “spontaneous abortion” as a proxy for 2 uterine receptivity, as failure in effective endometrial preparation to implantation largely contributes to 3 miscarriage in humans (Lucas et al. 2016; Muter et al. 2021), although embryo defects also contribute. 4 Additionally, we also retained post-menopausal bleeding and other menopausal disorders, which can 5 be symptoms of post-menopausal uterine diseases, notably endometrial cancer or uterine leiomyomas 6 (Makker et al. 2021; Munro 2019). 7 Selection of GWAS on uterine phenotypes in Pan-UK Biobank database 8 For replication purposes, we selected uterine-associated phenotypes in the Pan-UK Biobank matching 9 those selected from Finngen and using the same inclusion criteria (See Table S4). Summary statistics 10 performed on individuals of European ancestry were downloaded from the Pan -UK Biobank website, 11 resulting in the inclusion of 8 uterine phenotypes: malignant endometrial neoplasm (ICD10-C54), 12 leiomyomas (ICD10-D25), endometriosis (ICD10-N80), uterine polyps (ICD10-N84.0), irregular 13 menstrual cycle bleedings , excessive menstrual cycle bleeding, post-menopausal bleeding, and 14 miscarriage. Inflammatory uterine diseases and other menopausal disorders of the uterus were not 15 phenotyped in the Pan-UK Biobank. 16 Summary statistics curation and pre-processing 17 GWAS summary statistics were harmonized using the JASS pre -processing pipeline (Julienne et al. 18 2020). This pipeline consists in (i) harmonizing heterogeneous formats of summary statistics, (ii) 19 aligning the effect allele of each study to the effect allele of the reference panel, (iii) removing strand -20 ambiguous variants and variants with low sample sizes, (iv) computing trait heritability, and genetic and 21 residual covariances across traits using the LDscore regression. For FinnGen summary statistics, we 22 used a Finnish reference panel and LDscore data derived from GnomAD v2 23 (https://gnomad.broadinstitute.org/downloads#v2-linkage-disequilibrium). For Pan-UK Biobank 24 summary statistics, we used the European reference panel and LDscore data based on the 1000G 25 database provided as part of the JASS pipeline. We used positional SNP identifier s (defined as the 26 combination of chr:pos:ref:alt) to map all summary statistic SNPs to the reference panel. For both 27 datasets, no imputation step was performed. The summary information for each summary statistics 28 used as input for the JASS preprocessing pipeline can be found in Table S1 and Table S4. We report 29 the genetic correlation and the significance of the genetic correlation computed by LDscore (B. K. Bulik-30 Sullivan et al. 2015; B. Bulik -Sullivan et al. 2015) and corrected for false discovery rate using the 31 Benjamini-Hochberg correction (Table S2). 32 Joint analysis of uterine trait summary statistics with JASS 33 Multi-trait analysis of uterine traits summary statistics was conducted using the omnibus test 34 implemented in the JASS suite (Julienne et al. 2020) . JASS output summary statistic s were further 35 processed by associating each SNP to their RSID using the initial summary statistics from FinnGen. 36 For the 18,524 SNPs without an RSID in initial summary statistics, we used ANNOVAR (hg38) SNPs 37 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 19 RSID and retrieved RSIDs for an additional 4,680 SNPs. The remaining 14,044 SNPs without official 1 RSIDs in Annovar were named using their positional identifier. 2 For each SNP, we reported the p -value of the joint test and the smallest p -value across univariate 3 GWASs. The ability of JASS to capture variants with pleiotropic effect s across uterine traits was 4 investigated (1) for each uterine trait independently, by comparing the effect sizes of uSNPs vs single-5 trait SNPs; and (2) across uterine traits, by computing pairwise uSNPs effect size correlations. We also 6 compared effect size magnitudes of uSNPs and single-trait SNPs for each uterine trait independently 7 using Wilcoxon rank sum tests, FDR-corrected using the Benjamini-Hochberg correction. 8 Replication analysis using GWAS from the Pan-UK Biobank 9 For each trait, we compared the directionality of allele effects between effects estimated in the Finngen 10 cohorts and in the Pan-UK Biobank cohorts. For all uSNPs identified with the Finngen dataset and also 11 present in Pan-UK Biobank summary statistics, we tested whether directions of effects were consistent 12 after matching effect alleles between both biobanks. Using a directional binomial test, we tested whether 13 the proportion of uSNPs with replicated effect directions was greater than expected by chance (Table 14 S5). After grouping uSNPs into loci, we additionally tested if the number of loci with replicated effect 15 directions was greater than expected by chance using the same approach (Table S5). 16 Selection of independent and lead SNPs with FUMA 17 FUMA (v1.5.2) was used to define 241 lead SNPs with a linkage disequilibrium correlation r2 < 0.5, 18 which is the default threshold used by PLINK (Purcell et al. 2007). Lead SNPs within 500 kb of each 19 other were merged as a single genomic locus, resulting in the identification of 31 genomic risk loci. To 20 calculate r2, MAF (minor allele frequency) and conduct LD analyses, European population genetic data 21 from the 1000G Project phase 3 (2015) was used as the reference panel. 22 Functional annotation of candidate SNPs 23 The ANNOVAR (v2017-07-17) (K. Wang et al. 2010) annotation available from the FUMA interface was 24 used to associate SNPs to functional elements in the genome. 61 uSNPs were not in ANNOVAR. 25 Functional annotation of those 61 SNPs was performed based on intersection with protein-coding gene 26 coordinates. SNPs not overlapping protein-coding gene annotations were classified as intergenic. We 27 also identified SNPs overlapping protein-coding promoters, defined as regions spanning 5 kb upstream 28 and 1 kb downstream of protein-coding TSSs from Ensembl (v109, (Cunningham et al. 2022)). 29 To annotate the regulatory potential of uSNPs, we combined genomic annotations including histone 30 modifications and chromatin accessibility across relevant human tissues. We downloaded ENCODE 31 (v3 2021, hg38 (Dunham et al. 2012)) annotations of 1,063,878 cis-regulatory elements across 1,518 32 cell types and tissues. Using bedtools intersect ( -wao), we annotated uSNP mapping to predicted 33 promoters and enhancers (proximal or distal) or CTCF sites in any human tissue or cell type. We used 34 ReMap 2022 to identify uSNPs overlapping transcription factor binding peaks (Hammal et al. 2022). We 35 downloaded non-redundant peaks for all 1,210 TF datasets present in ReMap across all cell types and 36 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 20 tissues. The annotation of single SNPs was performed with bedtools intersect and a minimum overlap 1 of 1bp. Variants overlapping peaks in uterine cell lines (myofibroblasts, endometrial stromal cells, 2 decidualized stromal cells) , healthy uterine tissues (uterus, endometrium, myometrial) or diseased 3 uterine samples (leiomyomas, primary endometrial cancer) were annotated as `REMAP peak uterus`, 4 in contrast to uSNPs overlapping any TF peak in REMAP (` REMAP peak all`). We also identified uSNPs 5 annotated as eQTLs by exact matching in at least one tissue in the GTEx (v10) database. Because 6 GTEx contains only a small number of uterine eQTLs, we elected to retain all eQTLs as potentially 7 relevant. Finally, we used the Ananastra database (Abramov et al. 2021) to annotate significant uSNPs 8 potentially disrupting a transcription factor binding site, using the European linkage disequilibrium map 9 and a FDR of 0.05. 10 Genomic 3D contacts in endometrial stromal and epithelial cells 11 Promoter capture Hi-C (PCHi-C) data on endometrial decidualised stromal cells derived from a full-term 12 placenta was retrieved from (Sakabe et al. 2020). Bed files were converted to hg38 coordinates using 13 the UCSC Liftover tool. uSNP coordinates were intersected with either PCHi-C baits or significant 14 contacts. 15 Hi-C datasets on endometrial epithelial organoids from two separate donors were retrieved from (Hewitt 16 et al. 2022). We removed contacts that overlapped ENCODE blacklisted regions using `pairToBed -17 -type neither`. We then selected consensus contacts found in both donors, resulting in a list of 18 3,873 endometrial epithelial 3D genomic contacts. Those contacts were intersected with the coordinates 19 of human protein -coding promoters, defined as a -1kb/+5kb window around gene TSSs, with 20 (pairToBed --type either). 21 uSNPs overlapping PCHi-C or Hi-C contacts in stromal and/or epithelial cells were associated to 22 putative target genes in the corresponding Hi-C contact, i.e. one or two genes depending on the type 23 of contact (CRE-promoter, or promoter-promoter), resulting in a list of 246 and 61 uSNPs in endometrial 24 stromal and epithelial contacts respectively. 25 Reanalysis of single-nuclei RNA and ATAC-seq of healthy endometrial cells 26 Single-nuclei RNA-seq data of endometrial tissue from the HECA atlas (Marečková et al. 2023) was re-27 analysed using Seurat (v5). The HECA atlas was reduced to samples obtained from healthy donors 28 without hormonal treatment . The resulting dataset was re -clustered and cluster annotation was 29 homogenised based on the endometrial cell type annotation available in HECA. Nuclei from the uterine 30 cervix (Epithelial MUC5B: n=1000; and Stromal HOXA13: n= 1471) were removed from the resulting 31 atlas to only consider cells of endometrial origin. The resulting annotation of endometrial cell populations 32 is presented in Figure S4 . Pseudobulk transcriptome s were generated for each endometrial cell 33 population using aggregate.Matrix, and scaled to the raw library size of each pseudobulk using edgeR 34 calcNormFactors. Resulting pseudobulks were normalised to log2CPM (counts per million) and used to 35 map the expression of uSNPs target genes across endometrial cell types. 36 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 21 Single-nuclei ATAC-seq data from healthy human endometrium samples was retrieved from Vrljicak et 1 al. (2023). Data was reanalysed using the same parameters as the original analysis to filter low quality 2 nuclei and call ATAC-seq peaks (nCount_peaks > 3000; nCount_peaks 3 4; TSS enrichment > 3). To infer cell type s, we combined the original cell lineage annotation from 4 Vrljicak et al. with snRNA-seq annotation prediction s using the HECA atlas (after cell type 5 harmonisation, see above). To perform this step, gene scores were predicted for snATAC-seq nuclei 6 using GeneActivity() from the Signac package (Stuart et al. 2021) . These gene score s were then 7 correlated with gene expression levels in the snRNA-seq data to transfer cell annotation s using 8 FindTransferAnchors and TransferData with default parameters. The resulting annotation is shown in 9 Figure S6. Peak calling was performed by cell type using Signac with default parameters. uSNPs were 10 then overlapped with accessible peaks in each cell type. 11 Tissue-specific enrichment of uSNP target genes with FUMA 12 Putative uSNP target genes were passed to the FUMA GENE2FUNC interface to compute expression 13 enrichment across 30 tissues from GTEx. Differentially expressed genes (DEG) sets in FUMA are 14 defined using a two-sided t-test per tissue against all other tissues, with a Bonferroni corrected p-value 15 <0.05 and a minimal absolute fold-change cut-off. The fold-enrichment in the present study represents 16 the hypergeometric enrichment of uSNP target genes in tissue-specific DEGs. 17 Gene Ontology analysis 18 Gene Ontology (GO) enrichment analysis was performed with the clusterProfiler package (v4.16) (T. 19 Wu et al. 2021) in R by comparing the 227 uSNP target genes to the full transcriptome. Gene sets from 20 the "Biological processes" category were used for gene set enrichment analysis. Enrichment was 21 considered significant for p < 0.05 after Benjamini-Hochberg correction for multiple testing. Significant 22 biological processes were clustered using the Ward distance method to group higher-order terms based 23 on shared genes. The comprehensive results of the GO analysis are presented in Table S8. 24 Analysis of uSNP derived allele effect sizes 25 The 1000 Genomes Phase 3 vcf files (Auton et al. 2015) were used as input to retrieve the ancestral 26 and derived alleles for every SNP , determined using the six primate s (human, chimpanzee, gorilla, 27 orangutan, macaque, marmoset) EPO alignments from Ensembl . We restricted the list of 235 lead 28 uSNPs to 211 bi-allelic lead uSNPs, for which an ancestral and a derived allele can be identified. If the 29 alternate allele in the reference dataset corresponded to the ancestral allele, we switched effect size 30 direction to reflect the impact of the derived allele on the trait . We then performed unsupervised K -31 means clustering on derived allele effect sizes across traits. The optimal number of clusters was 32 identified using the average silhouette width criterion. To explore how derived alleles contribute to risk 33 increase or decrease across different uterine disorders, effect sizes of the derived allele s were 34 normalized across traits by computing the ratio of the effect size (β) to its standard error (SE_β). We 35 then correlated derived allele effect size s in each cluster across uterine disorders using Spearman’s 36 rank correlation. Correlation p-values were adjusted using BH correction. 37 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 22 Construction of control SNP datasets 1 Vcf files from 1000 Genomes Phase 3 were used as reference for all evolutionary analys es. SNPs in 2 the 1000 Genome Phase 3 panels were further annotated with their minor allele frequencies for all 1000 3 Genomes populations, local gene densit ies, GERP score s and recombination rate s. Minor allele 4 frequencies (MAF) were computed using vcftools (--freq) and filtered to keep only dimorphic sites for 5 each population independently. Local gene densities were calculated by binning the genome into 1 Mb 6 blocks to generate pseudo-independent regions of the genome, and counting all annotated TSS within 7 the 1 Mb region, including lncRNAs and pseudogenes. GERP scores were obtained from Ensembl, 8 calculated from the 91-mammal whole genome alignments. Recombination rate s from the 1000 9 Genomes panels were downloaded from UCSC 10 (http://hgdownload.soe.ucsc.edu/gbdb/hg38/recombRate/recomb1000GAvg.bw). 11 For each lead uSNP present in the 1000 Genomes panels, control SNPs were defined as SNPs 12 matched for MAF, recombination rate and gene density. To control for recombination rate and gene 13 density, SNPs were ranked and divided into 20 bins of equal size. Control SNPs meet the following 14 criteria: (1) with a MAF within 0.025 of their corresponding lead uSNP based on allele frequencies in 15 the CEU population; (2) within the same recombination rate bin; (3) within the same gene density bin; 16 and (4) excluding SNPs in LD (r2 > 0.2) with their corresponding lead uSNP. For each lead uSNP, the 17 resulting list of matching control SNPs was sampled 1000 times at random with replacement to create 18 1000 control sets per lead uSNP. Controls SNPs across lead uSNPs were further combined to form 19 1000 sets of control SNPs and used to compare FST distributions. Out of 241 lead uSNPs, 50 either 20 were not present in 1000 Genomes panels or could not be assigned appropriate matching control sets, 21 and were excluded from the comparison. 22 FST score comparisons 23 FST scores (Weir et Cockerham 1984) were computed from the 1000 Genomes Phase 3 vcf files using 24 the software selink (https://github.com/h-e-g/selink). Pairwise FST scores were computed between three 25 1000 Genomes populations of European, East Asian and African ancestry (respectively CEU, CHB and 26 YRI). To test for polygenic selection signatures, we compared the distributions of FST scores between 27 lead uSNPs and matched control sets (see above) using Wilcoxon rank sum tests. P-values were BH-28 corrected for multiple testing across population pairs (CEU vs. CHB, CEU vs. YRI and CHB vs. YRI). 29 For each pair of populations, a directional signed FST score (positive or negative) defined as previously 30 proposed in (Quach et al. 2016), to orient whether the derived allele is more or less frequent with respect 31 to a reference population. Directional FST scores were used to identify lead uSNPs with outlier FST scores 32 and under putative selection in either European , East-Asian or African populations. Outlier FST SNPs 33 were defined as the top 5% genome -wide FST signals for each population comparison. To test for 34 enrichment in outlier FST SNPs, we compared the fraction of lead uSNPs in the top 5% FST values to 35 the fraction of controls across all 1000 control sets and derived an empirical p-value. Resulting p-values 36 were BH-corrected for multiple testing across population pairs (CEU vs. CHB, CEU vs. YRI and CHB 37 vs. YRI). We also tested for an overall enrichment of lead uSNPs in the top 5% FST signals using a two-38 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 23 sided binomial test (probability of finding a uSNPs in the top 5% in at least one FST comparison = 1-1 0.953). 2 iHS score comparisons 3 iHS scores were obtained from Johnson et Voight (2018). We compared distributions of absolute iHS 4 scores between lead uSNPs and matched control sets using Wilcoxon rank sum tests. Out of 241 lead 5 uSNPs, only 163 lead uSNPs had an available iHS value in at least one population. 6 Code availability 7 All original code produced during this project is accessible at https://gitlab.pasteur.fr/euliorzo/popgen. 8 JASS is available at https://statistical-genetics.pages.pasteur.fr/jass/index.html. 9 Licences 10 Figures were produced with BioRender under licence to Institut Pasteur. 11 Acknowledgments 12 We thank Jan Brosens, Pavle Vrljicak and Mireia Taus -Nebot (University of Warwick) for sharing 13 analyzed snATAC-seq datasets from their previous study. 14 Funding 15 This project was supported by Institut Pasteur (G5 package), Centre National de la Recherche 16 Scientifique (CNRS UMR 3525), Institut National de la Santé et de la Recherche Médicale (INSERM 17 UA12), and the Inception program (Investissement d’Avenir grant ANR -16-CONV-0005). EL is 18 supported by a PhD fellowship from Université Paris Cité and a grant from the Fondation pour la 19 Recherche Médicale (grant agreement FDT202504020295). 20 Authors contributions 21 EL and CB conceived and designed the project and analys es. EL performed all analyse s, with 22 contributions from HJ and HA for multi-trait GWAS analyses, MD for functional genomics analyses, and 23 GL for evolutionary analyses. EL designed all figures. EL and CB wrote the manuscript with input from 24 all authors. All authors approved the manuscript. 25 Competing interests 26 The authors declare no competing interests. 27 28 29 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 24

1 Abramov, Sergey, Alexandr Boytsov, Daria Bykova, et al. 2021. « Landscape of Allele -Specific 2 Transcription Factor Binding in the Human Genome ». Nature Communications 12 (1): 1. 3 https://doi.org/10.1038/s41467-021-23007-0. 4 Auton, Adam, Gonçalo R. Abecasis, David M. Altshuler, et al. 2015. « A Global Reference for Human 5 Genetic Variation ». Nature 526 (7571): 68-74. https://doi.org/10.1038/nature15393. 6 Benonisdottir, Stefania, Vincent J. Straub, Augustine Kong, et Melinda C. Mills. 2024. « Genetics of 7 Female and Male Reproductive Traits and Their Relationship with Health, Longevity and 8 Consequences for Offspring ». Nature Aging 4 (12): 1745-59. https://doi.org/10.1038/s43587-9 024-00733-w. 10 Berg, Jeremy J., Arbel Harpak, Nasa Sinnott -Armstrong, et al. 2019. « Reduced Signal for Polygenic 11 Adaptation of Height in UK Biobank ». eLife 8 (mars): e39725. 12 https://doi.org/10.7554/eLife.39725. 13 Bourdon, Mathilde, Pietro Santulli, Louis Marcellin, et al. 2021. « Adenomyosis: An update regarding its 14 diagnosis and clinical features ». Journal of Gynecology Obstetrics and Human Reproduction 15 50 (10): 102228. https://doi.org/10.1016/j.jogoh.2021.102228. 16 Bulik-Sullivan, Brendan, Hilary K. Finucane, Verneri Anttila, et al. 2015. « An Atlas of Genetic 17 Correlations across Human Diseases and Traits ». Nature Genetics 47 (11): 1236 -41. 18 https://doi.org/10.1038/ng.3406. 19 Bulik-Sullivan, Brendan K., Po -Ru Loh, Hilary K. Finucane, et al. 2015. « LD Score Regression 20 Distinguishes Confounding from Polygenicity in Genome -Wide Association Studies ». Nature 21 Genetics 47 (3): 291-95. https://doi.org/10.1038/ng.3211. 22 Bulun, Serdar E., Bahar D. Yilmaz, Christia Sison, et al. 2019. « Endometriosis ». Endocrine Reviews 23 40 (4): 1048-79. https://doi.org/10.1210/er.2018-00242. 24 Bulun, Serdar E., Ping Yin, JianJun Wei, et al. 2025. « Uterine Fibroids ». Physiological Reviews, 25 publication en ligne anticipée, juin 13. Rockville, MD. 26 https://doi.org/10.1152/physrev.00010.2024. 27 Buyukcelebi, Kadir, Alexander J. Duval, Fatih Abdula, Hoda Elkafas, Fidan Seker-Polat, et Mazhar Adli. 28 2024. « Integrating Leiomyoma Genetics, Epigenomics, and Single -Cell Transcriptomics 29 Reveals Causal Genetic Variants, Genes, and Cell Types ». Nature Communications 15 (1): 30 1169. https://doi.org/10.1038/s41467-024-45382-0. 31 Cardoso, Jéssica Vilarinho, Jamila Alessandra Perini, Daniel Escorsim Machado, Ricardo Pinto, et Rui 32 Medeiros. 2020. « Systematic review of genome-wide association studies on susceptibility to 33 endometriosis ». European Journal of Obstetrics & Gynecology and Reproductive Biology 255 34 (décembre): 74-82. https://doi.org/10.1016/j.ejogrb.2020.10.017. 35 Chen, Piaopiao, et Jianzhi Zhang. 2020. « Antagonistic Pleiotropy Conceals Molecular Adaptations in 36 Changing Environments ». Nature Ecology & Evolution 4 (3): 461 -69. 37 https://doi.org/10.1038/s41559-020-1107-8. 38 Choi, Eun Ji, Seong Beom Cho, Sa Ra Lee, et al. 2017. « Comorbidity of gynecological and non -39 gynecological diseases with adenomyosis and endometriosis ». Obstetrics & Gynecology 40 Science 60 (6): 579-86. https://doi.org/10.5468/ogs.2017.60.6.579. 41 Choin, Jeremy, Javier Mendoza-Revilla, Lara R. Arauna, et al. 2021. « Genomic Insights into Population 42 History and Biological Adaptation in Oceania ». Nature 592 (7855): 7855. 43 https://doi.org/10.1038/s41586-021-03236-5. 44 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 25 Constantinescu, Andrei-Emil, Ruth E. Mitchell, Jie Zheng, et al. 2022. « A framework for research into 1 continental ancestry groups of the UK Biobank ». Human Genomics 16 (1): 3. 2 https://doi.org/10.1186/s40246-022-00380-5. 3 Cunningham, Fiona, James E. Allen, Jamie Allen, et al. 2022. « Ensembl 2022 ». Nucleic Acids 4 Research 50 (D1): D988-95. https://doi.org/10.1093/nar/gkab1049. 5 Dong, Di, Zhe Zhou, Minglu Zhu, et al. 2025. « STN1 Facilitates Metastasis by Promoting Transcription 6 of EMT -Activator ZEB1 in Pancreatic Cancer ». Nature Communications 16 (1): 7815. 7 https://doi.org/10.1038/s41467-025-63083-0. 8 Dunham, Ian, Anshul Kundaje, Shelley F. Aldred, et al. 2012. « An Integrated Encyclopedia of DNA 9 Elements in the Human Genome ». Nature 489 (7414): 57 -74. 10 https://doi.org/10.1038/nature11247. 11 Fiore, Alexander, Maìra Casalechi, Laura Sichenze, et al. 2025. « Co-Occurrence of Endometriosis and 12 Uterine Fibroids: A Systematic Review and Meta-Analysis ». eClinicalMedicine 89 (novembre). 13 https://doi.org/10.1016/j.eclinm.2025.103510. 14 Fraser, Ian S., Hilary O. D. Critchley, Michael Broder, et Malcolm G. Munro. 2011. « The FIGO 15 Recommendations on Terminologies and Definitions for Normal and Abnormal Uterine 16 Bleeding ». Seminars in Reproductive Medicine 29 (5): 383-90. https://doi.org/10.1055/s-0031-17 1287662. 18 Gallagher, C. S., N. Mäkinen, H. R. Harris, et al. 2019. « Genome-Wide Association and 19 Epidemiological Analyses Reveal Common Genetic Origins between Uterine Leiomyomata and 20 Endometriosis ». Nature Communications 10 (1): 4857. https://doi.org/10.1038/s41467 -019-21 12536-4. 22 Gao, J., S. Zeng, B. L. Sun, H. M. Fan, et L. H. Han. 1987. « Menstrual Blood Loss and Hematologic 23 Indices in Healthy Chinese Women ». The Journal of Reproductive Medicine 32 (11): 822-26. 24 Giuliani, Emma, Sawsan As -Sanie, et Erica E. Marsh. 2020. « Epidemiology and Management of 25 Uterine Fibroids ». International Journal of Gynecology & Obstetrics 149 (1): 3 -9. 26 https://doi.org/10.1002/ijgo.13102. 27 GTEx Consortium. 2017. « Genetic Effects on Gene Expression across Human Tissues ». Nature 550 28 (7675): 204-13. https://doi.org/10.1038/nature24277. 29 Guare, Lindsay A., Jagyashila Das, Lannawill Caruth, et al. 2025. « Expanding the Genetic Landscape 30 of Endometriosis: Integrative -Omics Analyses Implicate Key Genes and Pathways in a Multi-31 Ancestry Study of over One Million Women ». Prépublication, medRxiv, novembre 14. 32 https://doi.org/10.1101/2024.11.26.24316723. 33 Hämälä, Tuomas, Amanda J. Gorton, David A. Moeller, et Peter Tiffin. 2020. « Pleiotropy Facilitates 34 Local Adaptation to Distant Optima in Common Ragweed (Ambrosia Artemisiifolia) ». PLOS 35 Genetics 16 (3): e1008707. https://doi.org/10.1371/journal.pgen.1008707. 36 Hammal, Fayrouz, Pierre de Langen, Aurélie Bergon, Fabrice Lopez, et Benoit Ballester. 2022. 37 « ReMap 2022: A Database of Human, Mouse, Drosophila and Arabidopsis Regulatory 38 Regions from an Integrative Analysis of DNA -Binding Sequencing Experiments ». Nucleic 39 Acids Research 50 (D1): D316-25. https://doi.org/10.1093/nar/gkab996. 40 Hewitt, Sylvia C., San-pin Wu, Tianyuan Wang, et al. 2022. « Progesterone Signaling in Endometrial 41 Epithelial Organoids ». Cells 11 (11). https://doi.org/10.3390/cells11111760. 42 Holsinger, Kent E., et Bruce S. Weir. 2009. « Genetics in Geographically Structured Populations: 43 Defining, Estimating and Interpreting FST ». Nature Reviews Genetics 10 (9): 9. 44 https://doi.org/10.1038/nrg2611. 45 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 26 Johnson, Kelsey Elizabeth, et Benjamin F. Voight. 2018. « Patterns of Shared Signatures of Recent 1 Positive Selection across Human Populations ». Nature Ecology & Evolution 2 (4): 713 -20. 2 https://doi.org/10.1038/s41559-018-0478-6. 3 Julienne, Hanna, Vincent Laville, Zachary R. McCaw, et al. 2021. « Multitrait GWAS to Connect Disease 4 Variants and Biological Mechanisms ». PLOS Genetics 17 (8): e1009713. 5 https://doi.org/10.1371/journal.pgen.1009713. 6 Julienne, Hanna, Pierre Lechat, Vincent Guillemot, et al. 2020. « JASS: command line and web 7 interface for the joint analysis of GWAS results ». NAR Genomics and Bioinformatics 2 (1): 8 lqaa003. https://doi.org/10.1093/nargab/lqaa003. 9 Karczewski, Konrad J., Rahul Gupta, Masahiro Kanai, et al. 2025. « Pan-UK Biobank Genome-Wide 10 Association Analyses Enhance Discovery and Resolution of Ancestry -Enriched Effects ». 11 Nature Genetics 57 (10): 2408-17. https://doi.org/10.1038/s41588-025-02335-7. 12 Kim, Jeewoo, Ariel Williams, Hannah Noh, et al. 2024. « 430 Genome-Wide Meta-Analysis Identifies 13 Novel Risk Loci for Uterine Fibroids across Multiple Ancestry Groups ». Journal of Clinical and 14 Translational Science 8 (s1): 128-29. https://doi.org/10.1017/cts.2024.372. 15 Koller, Dora, Jun He, Solveig Løkhammer, et al. 2025. « Multi-Ancestry Genome-Wide Association 16 Study of Endometriosis and Its Clinical Manifestations in ∼1.4 Million Women: Translating Gene 17 Discovery into Pathogenic Mechanisms and Therapeutic Targets ». Prépublication, medRxiv, 18 septembre 5. https://doi.org/10.1101/2025.09.03.25335012. 19 Kun, Eucharist, Emily M. Javan, Olivia Smith, et al. 2023. « The Genetic Architecture and Evolution of 20 the Human Skeletal Form ». Science 381 (6655): eadf8009. 21 https://doi.org/10.1126/science.adf8009. 22 Kurki, Mitja I., Juha Karjalainen, Priit Palta, et al. 2022. FinnGen: Unique Genetic Insights from 23 Combining Isolated Population and National Health Register Data . Preprint. Genetic and 24 Genomic Medicine. https://doi.org/10.1101/2022.03.03.22271360. 25 Kurki, Mitja I., Juha Karjalainen, Priit Palta, et al. 2023. « FinnGen Provides Genetic Insights from a 26 Well-Phenotyped Isolated Population ». Nature 613 (7944): 508 -18. 27 https://doi.org/10.1038/s41586-022-05473-8. 28 Kyama, Cleophas M., Jason M. Mwenda, James Machoki, et al. 2007. « Endometriosis in African 29 Women ». Women’s Health 3 (5): 629-35. https://doi.org/10.2217/17455057.3.5.629. 30 La Vecchia, I., A. Fiore, N. Salmeri, et al. 2025. « O-199 The association between endometriosis and 31 endometrial polyps: a systematic review and meta -analysis ». Human Reproduction 40 32 (Supplement_1): deaf097.199. https://doi.org/10.1093/humrep/deaf097.199. 33 Lahdenperä, Mirkka, Virpi Lummaa, Samuli Helle, Marc Tremblay, et Andrew F. Russell. 2004. 34 « Fitness Benefits of Prolonged Post-Reproductive Lifespan in Women ». Nature 428 (6979): 35 178-81. https://doi.org/10.1038/nature02367. 36 Laval, Guillaume, Etienne Patin, Pierre Boutillier, et Lluis Quintana-Murci. 2021. « Sporadic occurrence 37 of recent selective sweeps from standing variation in humans as revealed by an approximate 38 Bayesian computation approach ». Genetics 219 (4): iyab161. 39 https://doi.org/10.1093/genetics/iyab161. 40 Levin, Michael G., Noah L. Tsao, Pankhuri Singhal, et al. 2022. « Genome-Wide Association and Multi-41 Trait Analyses Characterize the Common Genetic Architecture of Heart Failure ». Nature 42 Communications 13 (1): 6914. https://doi.org/10.1038/s41467-022-34216-6. 43 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 27 Long, Erping, et Jianzhi Zhang. 2023. « Evidence for the role of selection for reproductively 1 advantageous alleles in human aging ». Science Advances 9 (49): eadh4990. 2 https://doi.org/10.1126/sciadv.adh4990. 3 Lucas, Emma S., Nigel P. Dyer, Katherine Fishwick, Sascha Ott, et Jan J. Brosens. 2016. « Success 4 after Failure: The Role of Endometrial Stem Cells in Recurrent Miscarriage ». Reproduction. 5 Reproduction 152 (5): R159-66. https://doi.org/10.1530/REP-16-0306. 6 Makker, Vicky, Helen MacKay, Isabelle Ray -Coquard, et al. 2021. « Endometrial Cancer ». Nature 7 Reviews Disease Primers 7 (1): 88. https://doi.org/10.1038/s41572-021-00324-8. 8 Marečková, Magda, Luz Garcia-Alonso, Marie Moullet, et al. 2023. An Integrated Single-Cell Reference 9 Atlas of the Human Endometrium . Preprint. Genomics. 10 https://doi.org/10.1101/2023.11.03.564728. 11 Masuda, Tatsuo, Siew-Kee Low, Masato Akiyama, et al. 2020. « GWAS of Five Gynecologic Diseases 12 and Cross-Trait Analysis in Japanese ». European Journal of Human Genetics 28 (1): 95-107. 13 https://doi.org/10.1038/s41431-019-0495-1. 14 McGrath, Isabelle M., Grant W. Montgomery, Sally Mortlock, et International Endometriosis Genetics 15 Consortium. 2023. « Genomic Characterisation of the Overlap of Endometriosis with 76 16 Comorbidities Identifies Pleiotropic and Causal Mechanisms Underlying Disease Risk ». 17 Human Genetics 142 (9): 1345-60. https://doi.org/10.1007/s00439-023-02582-w. 18 Mecha, Ezekiel O., Joseph N. Njagi, Roselydiah N. Makunja, Charles O. A. Omwandho, Philippa T. K. 19 Saunders, et Andrew W. Horne. 2022. « Endometriosis among African Women ». Reproduction 20 and Fertility. Reproduction and Fertility 3 (3): C40-43. https://doi.org/10.1530/RAF-22-0040. 21 Mortlock, Sally, Rosario I. Corona, Pik Fang Kho, et al. 2022. « A Multi-Level Investigation of the Genetic 22 Relationship between Endometriosis and Ovarian Cancer Histotypes ». Cell Reports Medicine 23 3 (3): 100542. https://doi.org/10.1016/j.xcrm.2022.100542. 24 Munro, Malcolm G. 2019. « Uterine polyps, adenomyosis, leiomyomas, and endometrial receptivity ». 25 Fertility and Sterility 111 (4): 629-40. https://doi.org/10.1016/j.fertnstert.2019.02.008. 26 Muter, Joanne, Chow-Seng Kong, et Jan J. Brosens. 2021. « The Role of Decidual Subpopulations in 27 Implantation, Menstruation and Miscarriage ». Frontiers in Reproductive Health 3. 28 https://www.frontiersin.org/articles/10.3389/frph.2021.804921. 29 Nguyen, Dinh Duc, Eugene Kim, Nhat Thong Le, et al. 2023. « Deficiency in mammalian STN1 30 promotes colon cancer development via inhibiting DNA repair ». Science Advances 9 (19): 31 eadd8023. https://doi.org/10.1126/sciadv.add8023. 32 Oehler, M. K., et M. C. P. Rees. 2003. « Menorrhagia: An Update ». Acta Obstetricia et Gynecologica 33 Scandinavica 82 (5): 405-22. https://doi.org/10.1034/j.1600-0412.2003.00097.x. 34 Painter, Jodie N., Tracy A. O’Mara, Andrew P. Morris, et al. 2018. « Genetic overlap between 35 endometriosis and endometrial cancer: evidence from cross -disease genetic correlation and 36 GWAS meta-analyses ». Cancer Medicine 7 (5): 1978-87. https://doi.org/10.1002/cam4.1445. 37 Pavlicev, Mihaela, Caitlin E. McDonough-Goldstein, Andreja Moset Zupan, et al. 2022. A SNP Affects 38 Wnt4 Expression in Endometrial Stroma, with Antagonistic Implications for Pregnancy, 39 Endometriosis and Reproductive Cancers . Preprint. Genetics. 40 https://doi.org/10.1101/2022.10.25.513653. 41 Pavlicev, Mihaela, et Günter P. Wagner. 2012. « A model of developmental evolution: selection, 42 pleiotropy and compensation ». Trends in Ecology & Evolution 27 (6): 316 -22. 43 https://doi.org/10.1016/j.tree.2012.01.016. 44 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 28 Pritchard, Jonathan K., et Anna Di Rienzo. 2010. « Adaptation – Not by Sweeps Alone ». Nature 1 Reviews Genetics 11 (10): 665-67. https://doi.org/10.1038/nrg2880. 2 Pritchard, Jonathan K., Joseph K. Pickrell, et Graham Coop. 2010. « The Genetics of Human 3 Adaptation: Hard Sweeps, Soft Sweeps, and Polygenic Adaptation ». Current Biology 20 (4): 4 R208-15. https://doi.org/10.1016/j.cub.2009.11.055. 5 Pujol Gualdo, Natàlia, Jelisaveta Džigurski, Valentina Rukins, et al. 2025. « Atlas of Genetic and 6 Phenotypic Associations across 42 Female Reproductive Health Diagnoses ». Nature Medicine 7 31 (5): 1626-34. https://doi.org/10.1038/s41591-025-03543-8. 8 Purcell, Shaun, Benjamin Neale, Kathe Todd -Brown, et al. 2007. « PLINK: A Tool Set for Whole -9 Genome Association and Population-Based Linkage Analyses ». American Journal of Human 10 Genetics 81 (3): 559-75. https://doi.org/10.1086/519795. 11 Quach, Hélène, Maxime Rotival, Julien Pothlichet, et al. 2016. « Genetic Adaptation and Neandertal 12 Admixture Shaped the Immune System of Human Populations ». Cell 167 (3): 643 -656.e17. 13 https://doi.org/10.1016/j.cell.2016.09.024. 14 Rafnar, Thorunn, Bjarni Gunnarsson, Olafur A. Stefansson, et al. 2018a. « Variants Associating with 15 Uterine Leiomyoma Highlight Genetic Background Shared by Various Cancers and Hormone-16 Related Traits ». Nature Communications 9 (1): 3636. https://doi.org/10.1038/s41467 -018-17 05428-6. 18 Rafnar, Thorunn, Bjarni Gunnarsson, Olafur A. Stefansson, et al. 2018b. « Variants Associating with 19 Uterine Leiomyoma Highlight Genetic Background Shared by Various Cancers and Hormone-20 Related Traits ». Nature Communications 9 (1): 3636. https://doi.org/10.1038/s41467 -018-21 05428-6. 22 Rahmioglu, Nilufer, Sally Mortlock, Marzieh Ghiasi, et al. 2023. « The Genetic Basis of Endometriosis 23 and Comorbidity with Other Pain and Inflammatory Conditions ». Nature Genetics 55 (3): 3. 24 https://doi.org/10.1038/s41588-023-01323-z. 25 Rodríguez, Juan Antonio, Urko M. Marigorta, David A. Hughes, Nino Spataro, Elena Bosch, et Arcadi 26 Navarro. 2017a. « Antagonistic Pleiotropy and Mutation Accumulation Influence Human 27 Senescence and Disease ». Nature Ecology & Evolution 1 (3): 0055. 28 https://doi.org/10.1038/s41559-016-0055. 29 Rodríguez, Juan Antonio, Urko M. Marigorta, David A. Hughes, Nino Spataro, Elena Bosch, et Arcadi 30 Navarro. 2017b. « Antagonistic Pleiotropy and Mutation Accumulation Influence Human 31 Senescence and Disease ». Nature Ecology & Evolution 1 (3): 0055. 32 https://doi.org/10.1038/s41559-016-0055. 33 Rose, Michael R. 1982. « Antagonistic Pleiotropy, Dominance, and Genetic Variation ». Heredity 48 (1): 34 63-78. https://doi.org/10.1038/hdy.1982.7. 35 Sakabe, Noboru J., Ivy Aneas, Nicholas Knoblauch, et al. 2020. « Transcriptome and regulatory maps 36 of decidua-derived stromal cells inform gene discovery in preterm birth ». Science Advances 6 37 (49): eabc8696. https://doi.org/10.1126/sciadv.abc8696. 38 Sapkota, Yadav, Valgerdur Steinthorsdottir, Andrew P. Morris, et al. 2017a. « Meta-Analysis Identifies 39 Five Novel Loci Associated with Endometriosis Highlighting Key Genes Involved in Hormone 40 Metabolism ». Nature Communications 8 (1): 15539. https://doi.org/10.1038/ncomms15539. 41 Sapkota, Yadav, Valgerdur Steinthorsdottir, Andrew P. Morris, et al. 2017b. « Meta-Analysis Identifies 42 Five Novel Loci Associated with Endometriosis Highlighting Key Genes Involved in Hormone 43 Metabolism ». Nature Communications 8 (1): 15539. https://doi.org/10.1038/ncomms15539. 44 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 29 Sheng, Xin, Yuting Guan, Ziyuan Ma, et al. 2021. « Mapping the Genetic Architecture of Human Traits 1 to Cell Types in the Kidney Identifies Mechanisms of Disease and Potential Treatments ». 2 Nature Genetics 53 (9): 1322-33. https://doi.org/10.1038/s41588-021-00909-9. 3 Sheu, Jim Jinn -Chyuan, Wei -Yong Lin, Ting -Yuan Liu, et al. 2024a. « Ethnic-Specific Genetic 4 Susceptibility Loci for Endometriosis in Taiwanese -Han Population: A Genome -Wide 5 Association Study ». Journal of Human Genetics 69 (11): 573 -83. 6 https://doi.org/10.1038/s10038-024-01270-5. 7 Sheu, Jim Jinn -Chyuan, Wei -Yong Lin, Ting -Yuan Liu, et al. 2024b. « Ethnic-Specific Genetic 8 Susceptibility Loci for Endometriosis in Taiwanese -Han Population: A Genome -Wide 9 Association Study ». Journal of Human Genetics, juillet 9, 1-11. https://doi.org/10.1038/s10038-10 024-01270-5. 11 Sinharoy, Sheela S., Lyzberthe Chery, Madeleine Patrick, et al. 2023. « Prevalence of Heavy Menstrual 12 Bleeding and Associations with Physical Health and Wellbeing in Low -Income and Middle -13 Income Countries: A Multinational Cross-Sectional Study ». The Lancet Global Health 11 (11): 14 e1775-84. https://doi.org/10.1016/S2214-109X(23)00416-3. 15 Soliman, A. M., H. Yang, E. X. Du, S. Kelkar, et C. Winkel. 2016. « A Systematic Literature Review of 16 Comorbidities and Symptoms among Uterine Fibroids Patients between 2000 and 2013 ». 17 Value in Health 19 (3): A172. https://doi.org/10.1016/j.jval.2016.03.1446. 18 Song, Jiangwei, Ning Gao, Zhe Chen, et al. 2024. « Shared genetic etiology of vessel diseases: A 19 genome-wide multi -traits association analysis ». Thrombosis Research 241 (septembre): 20 109102. https://doi.org/10.1016/j.thromres.2024.109102. 21 Song, Weichen, Yueqi Shi, Weidi Wang, et al. 2021. « A Selection Pressure Landscape for 870 Human 22 Polygenic Traits ». Nature Human Behaviour 5 (12): 1731-43. https://doi.org/10.1038/s41562-23 021-01231-4. 24 Stephan, Wolfgang. 2016. « Signatures of Positive Selection: From Selective Sweeps at Individual Loci 25 to Subtle Allele Frequency Changes in Polygenic Adaptation ». Molecular Ecology 25 (1): 26 79-88. https://doi.org/10.1111/mec.13288. 27 Stephan, Wolfgang, et Sona John. 2020. « Polygenic Adaptation in a Population of Finite Size ». 28 Entropy 22 (8): 907. https://doi.org/10.3390/e22080907. 29 Stewart, E. A., C. L. Cookson, R. A. Gandolfo, et R. Schulze -Rath. 2017a. « Epidemiology of Uterine 30 Fibroids: A Systematic Review ». BJOG: An International Journal of Obstetrics and 31 Gynaecology 124 (10): 1501-12. https://doi.org/10.1111/1471-0528.14640. 32 Stewart, E. A., C. L. Cookson, R. A. Gandolfo, et R. Schulze -Rath. 2017b. « Epidemiology of Uterine 33 Fibroids: A Systematic Review ». BJOG: An International Journal of Obstetrics and 34 Gynaecology 124 (10): 1501-12. https://doi.org/10.1111/1471-0528.14640. 35 Stuart, Tim, Avi Srivastava, Shaista Madad, Caleb A. Lareau, et Rahul Satija. 2021. « Single-Cell 36 Chromatin State Analysis with Signac ». Nature Methods 18 (11): 1333 -41. 37 https://doi.org/10.1038/s41592-021-01282-5. 38 The 1000 Genomes Project Consortium, Corresponding authors, Adam Auton, et al. 2015. « A Global 39

for Human Genetic Variation ». Nature 526 (7571): 68 -74. 40 https://doi.org/10.1038/nature15393. 41 Thibord, Florian, Jason Cunha, Jelisaveta Džigurski, et al. 2025. « Genome-wide meta-analysis of 42 heavy menstrual bleeding reveals 36 risk loci ». Blood 146 (6): 745 -58. 43 https://doi.org/10.1182/blood.2024027382. 44 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint 30 Venkatesh, Samvida S., Laura B. L. Wittemans, Duncan S. Palmer, et al. 2025. « Genome-Wide 1 Analyses Identify 25 Infertility Loci and Relationships with Reproductive Traits across the Allele 2 Frequency Spectrum ». Nature Genetics 57 (5): 1107-18. https://doi.org/10.1038/s41588-025-3 02156-8. 4 Voight, Benjamin F., Sridhar Kudaravalli, Xiaoquan Wen, et Jonathan K. Pritchard. 2006a. « A Map of 5 Recent Positive Selection in the Human Genome ». PLOS Biology 4 (3): e72. 6 https://doi.org/10.1371/journal.pbio.0040072. 7 Voight, Benjamin F., Sridhar Kudaravalli, Xiaoquan Wen, et Jonathan K. Pritchard. 2006b. « A Map of 8 Recent Positive Selection in the Human Genome ». PLoS Biology 4 (3): e72. 9 https://doi.org/10.1371/journal.pbio.0040072. 10 Vrljicak, Pavle, Emma S. Lucas, Maria Tryfonos, Joanne Muter, Sascha Ott, et Jan J. Brosens. 2023. 11 « Dynamic Chromatin Remodeling in Cycling Human Endometrium at Single-Cell Level ». Cell 12 Reports 42 (12). https://doi.org/10.1016/j.celrep.2023.113525. 13 Wang, Kai, Mingyao Li, et Hakon Hakonarson. 2010. « ANNOVAR: functional annotation of genetic 14 variants from high -throughput sequencing data ». Nucleic Acids Research 38 (16): e164. 15 https://doi.org/10.1093/nar/gkq603. 16 Wang, Xuan-Yu, Qiong Lyu, Yang -Yang Zhang, et al. 2025. « Shared Genetic Architecture between 17 Metabolic Dysfunction -Associated Steatotic Liver Disease and Cardiometabolic Traits 18 Comorbidities: A Genome-Wide Pleiotropic and Multi-Omics Study ». Metabolism and Target 19 Organ Damage 5 (2): N/A-N/A. https://doi.org/10.20517/mtod.2024.129. 20 Wang, Xuemin, Dylan M. Glubb, et Tracy A. O’Mara. 2022. « 10 Years of GWAS Discovery in 21 Endometrial Cancer: Aetiology, Function and Translation ». eBioMedicine 77 (mars). 22 https://doi.org/10.1016/j.ebiom.2022.103895. 23 Weir, B. S., et C. Clark Cockerham. 1984. « Estimating F-Statistics for the Analysis of Population 24 Structure ». Evolution 38 (6): 1358-70. https://doi.org/10.2307/2408641. 25 Wu, Dou, Zi Wang, Jingying Huang, et al. 2022. « An antagonistic pleiotropic gene regulates the 26 reproduction and longevity tradeoff ». Proceedings of the National Academy of Sciences 119 27 (18): e2120311119. https://doi.org/10.1073/pnas.2120311119. 28 Wu, Tianzhi, Erqiang Hu, Shuangbin Xu, et al. 2021. « clusterProfiler 4.0: A universal enrichment tool 29 for interpreting omics data ». The Innovation 2 (3): 100141. 30 https://doi.org/10.1016/j.xinn.2021.100141. 31 Xiang, Yifan, Vineeta Tanwar, Parminder Singh, Lizellen La Follette, et Pankaj Kapahi. 2024. « Early 32 Menarche and Childbirth Accelerate Aging -Related Outcomes and Age -Related Diseases: 33 Evidence for Antagonistic Pleiotropy in Humans ». eLife 13 (novembre). 34 https://doi.org/10.7554/eLife.102447.1. 35 Yamamoto, Ayae, Erica B. Johnstone, Michael S. Bloom, Heather G. Huddleston, et Victor Y. Fujimoto. 36 2017. « A Higher Prevalence of Endometriosis among Asian Women Does Not Contribute to 37 Poorer IVF Outcomes ». Journal of Assisted Reproduction and Genetics 34 (6): 765 -74. 38 https://doi.org/10.1007/s10815-017-0919-1. 39 Ye, Jiatian, Hongling Peng, Xia Huang, et Xiaorong Qi. 2022. « The association between endometriosis 40 and risk of endometrial cancer and breast cancer: a meta-analysis ». BMC Women’s Health 22 41 (novembre): 455. https://doi.org/10.1186/s12905-022-02028-x. 42 43 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.24.713926doi: bioRxiv preprint