{"paper_id":"c7424216-27b4-4f42-bdd1-e79f4b4c96c7","body_text":"1 \nTitle: Iron homeostasis and endometriosis risk: Genetic evidence for a shared biological link \n \nAuthors: Veronika Denner1,2, Christian M. Becker2, Krina T Zondervan2, Sam Morris1*, Nilufer \nRahmioglu2,3*# \n* Jointly directed this work. \n# Corresponding author. \n \nAffiliations: \n1 Nuffield Department of Population Health, University of Oxford, UK. \n2 Nuffield Department of Women’s and Reproductive Health, University of Oxford, UK. \n3 Centre for Human Genetics, University of Oxford, UK. \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n \n 2 \nABSTRACT \nSTUDY QUESTION \nIs genetic liability to endometriosis associated with iron homeostasis, and is this relationship \npotentially causal? \n \nSUMMARY ANSWER \nGenetic evidence indicates that reduced systemic iron status is associated with increased risk of \nendometriosis, with evidence of 8 shared genome-wide significant loci and suggestive but \ninconsistent evidence for causal bidirectional effects.  \n \nWHAT IS KNOWN ALREADY \nEndometriosis is a chronic inflammatory condition associated with local iron accumulation \nwithin ectopic lesions and peritoneal cavity, resulting from retrograde menstruation and \naltered iron homeostasis. Epidemiological studies have suggested that women with \nendometriosis may exhibit reduced systemic iron stores compared to women without \nendometriosis, reflected by lower circulating ferritin concentrations, although findings have \nbeen inconsistent and may be confounded by menstrual blood loss and inflammation. As \nobservational studies cannot distinguish causal relationships from secondary effects or residual \nconfounding, the potential genetic basis linking iron homeostasis and endometriosis risk \nremains unclear. \n \nSTUDY DESIGN, SIZE, DURATION \nWe performed genetic analyses using summary statistics from large-scale genome-wide \nassociation studies (GWAS) of endometriosis (overall and stage III/IV disease) and five iron \nbiomarkers (serum iron, ferritin, total iron-binding capacity (TIBC), transferrin saturation, and \nhepcidin). Analyses included genome-wide genetic correlation using linkage disequilibrium \nscore regression (LDSC), identification of shared genetic variants using multi-trait GWAS \n(MTAG) and bidirectional Mendelian randomisation to evaluate potential causal relationships.  \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 3 \nPARTICIPANTS/MATERIALS, SETTING, METHODS \nIron biomarker summary statistics came from a six-cohort GWAS meta-analysis (HUNT, MGI, \nSardiNIA, deCODE, Interval, DBDS; N up to 257,953) of blood-derived serum iron, ferritin, \ntransferrin saturation and TIBC (Moksnes et al., 2022). Endometriosis summary statistics came \nfrom a 24-study GWAS meta-analysis (60,674 cases, 701,926 controls; European and East Asian \nancestry), 12 of which had surgically confirmed cases (Rahmioglu et al., 2023). \nGenome-wide genetic correlations between iron biomarkers and endometriosis (overall and \nstage III/IV disease) were estimated using linkage disequilibrium score regression (LDSC), based \non summary statistics aligned to the GRCh37 reference genome and restricted to HapMap3 \nvariants. Multi-trait GWAS (MTAG) was applied to each iron biomarker jointly with \nendometriosis to enhance discovery of genetic loci. Shared loci were functionally annotated \nusing reproductive and iron related tissues from GTEx v8 and blood from eQTLGen expression \nquantitative trait loci (eQTL) data. Bidirectional Mendelian randomisation (MR) analyses were \nperformed using genome-wide significant variants across multiple clumping thresholds, with \ninverse-variance weighting (IVW) as the primary method and sensitivity analyses including \nweighted median, MR-Egger and MR-PRESSO. \n \nMAIN RESULTS AND THE ROLE OF CHANCE \nGenetic correlation analyses suggested that a genetic predisposition to endometriosis is \nassociated with a profile of lower systemic iron availability. Specifically, genetic liability to \nendometriosis was associated with higher total iron-binding capacity (TIBC; rg=0.16, p=4x10-4), \ntogether with lower transferrin saturation (rg=-0.16, p=0.006) and lower ferritin levels (rg=-\n0.10, p=0.022), findings that are consistent with reduced iron stores. MTAG identified eight \nadditional genome-wide significant loci for endometriosis and eight loci shared with iron \nbiomarkers, including regions implicating coagulation (F5), reproductive biology (WNT4), and \nimmune and vascular pathways (e.g. ABO, STAT6). Mendelian randomisation analyses provided \nlimited and inconsistent evidence for a causal relationship between iron status and \nendometriosis. Although the inverse-variance weighted (IVW) model showed nominal \nassociations between higher ferritin levels and a lower risk of endometriosis (OR = 0.85, 95% CI \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 4 \n0.76–0.94; p = 0.002), and between genetic liability to endometriosis and higher TIBC (OR = \n1.02, 95% CI 1.01–1.04; p = 0.006), these findings were not consistently supported by sensitivity \nanalyses. MR-PRESSO identified a small number of pleiotropic variants, but their removal did \nnot materially alter the results. \n \nLIMITATIONS, REASONS FOR CAUTION \nIron biomarker GWAS included males and females, potentially obscuring female-specific \neffects. Dataset availability restricted analyses to European ancestry, limiting applicability to \nother populations, and to overall and stage III/IV endometriosis, precluding assessment of other \ndisease subtypes. Heterogeneity across SNP instruments, reflected by Cochran’s Q statistics, \nreduced the precision of Mendelian randomisation estimates. Moreover, the genetic \ninstruments explained only between approximately 1.0% and 18.8% of variance in the iron \nbiomarkers, depending on the clumping threshold, which may have limited power to detect \ncausal effects. \n \nWIDER IMPLICATIONS OF THE FINDINGS \nThese findings suggest that endometriosis is genetically associated with reduced systemic iron \navailability and altered iron homeostasis. Thus, lower systemic iron status observed in women \nwith endometriosis may not be explained solely by menstrual blood loss or dietary factors, but \nreflect an underlying genetic predisposition. Shared genetic loci implicate coagulation, ABO \nbiology, and immune pathways as potential mechanisms linking iron metabolism and \nendometriosis. Although Mendelian randomisation did not provide consistent evidence for \ncausality, these findings support a shared genetic architecture and warrant further investigation \nusing female-specific GWAS, refined disease subtypes, and multi-omic approaches. Clinically, \nthese findings suggest that low systemic iron status in women with endometriosis may reflect \nfactors beyond established causes of iron deficiency, including an underlying genetic \npredisposition. \n \nSTUDY FUNDING/COMPETING INTEREST(S) \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 5 \nThe research was supported by a three-month post-MSc fellowship (September to December \n2025), awarded to V.D. by the Nuffield Department of Population Health. N.R. is a consultant \nfor Endogene.bio, outside this submitted work. V.D. is the co-president of #EndEndoSilence, a \nGerman endometriosis advocacy association, outside this submitted work. S.M. declares no \ncompeting interests. \n \nTRIAL REGISTRATION NUMBER \nN/A. \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 6 \n \n \nINTRODUCTION \nEndometriosis is a chronic systemic condition defined by the presence of endometrial-like \ntissue outside the uterus  (Zondervan, Becker and Missmer, 2020). While endometriosis lesions \nare most commonly located in the pelvic cavity, they can also grow in more distant sites, \nincluding the thoracic cavity (Hirata, Koga and Osuga, 2020). The condition affects around ten \npercent of women of reproductive age (Shafrir et al., 2018) and is associated with severe \ndysmenorrhoea, chronic pelvic pain, dyspareunia, bowel and bladder symptoms, fatigue and \ninfertility, with substantial negative impacts on quality of life.  \n \nThe aetiology of endometriosis has been widely debated. Retrograde menstruation, defined as \nthe reflux of menstrual blood and viable endometrial cells into the peritoneal cavity, has long \nbeen proposed as a central mechanism of disease development (Sampson, 1927). Yet despite \noccurring in approximately ninety percent of menstruating individuals, only a minority develop \npersistent endometriotic lesions (Mehedintu et al., 2014). Additional factors, including altered \ninnate immune responses, impaired clearance of ectopic cells, local oestrogen production and a \nheightened inflammatory environment, are thought to facilitate ectopic lesion establishment \nand persistence within the peritoneal environment (Zondervan et al., 2018). \n \nAltered iron handling has emerged as an important component of the peritoneal \nmicroenvironment in endometriosis. Retrograde menstruation introduces red blood cells into \nthe peritoneal cavity, where their breakdown releases haem and free iron (Ng et al., 2020). This \ncan lead to local iron, ferritin and haemosiderin accumulation within lesions and surrounding \ntissues and immune cells such as iron-laden macrophages. Excess iron promotes oxidative \nstress and inflammation, and may impair ferroptosis, an iron-dependent form of regulated cell \ndeath, potentially allowing ectopic endometrial-like cells to evade iron-mediated cell death and \npersist (Ng et al., 2020). \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 7 \nThese processes may be particularly accentuated in more extensive disease, which also has \nbeen associated with higher heritability (Rahmioglu et al., 2023). For example, endometriomas \nexhibit iron-rich microenvironments dominated by macrophages containing heme and iron \n(Wölfler et al., 2013) and deep endometriosis has also been linked to oxidative stress pathways, \nthrough generation of superoxide anions, hydrogen peroxide, and nitric oxide (Tosti et al., \n2015). Together, these findings reinforce that local iron accumulation contributes to the \ninflammatory and oxidative milieu characteristic of endometriosis. \n \nIn contrast to this local iron overload, epidemiological studies suggest that women with \nendometriosis may exhibit reduced systemic iron levels. Heavy menstrual bleeding and chronic \ninflammation may both contribute to reduced systemic iron availability, and a large longitudinal \ncohort study reported a 46% increased risk of iron deficiency among women with \nendometriosis (Gete et al., 2024). However, findings across studies are inconsistent, and may \nbe confounded by menstrual blood loss (Munro et al., 2023) (Ekroos et al., 2024) and \ninflammatory processes (Li and Li, 2026). Observational studies are therefore limited in their \nability to distinguish whether altered systemic iron status contributes to disease development \nor arises as a consequence of endometriosis. \n \nSystemic iron homeostasis is tightly regulated through coordinated processes of absorption, \ntransport, storage and recycling. Circulating markers capture complementary aspects of this \nsystem, including serum iron (circulating levels), ferritin (iron storage), total iron-binding \ncapacity (TIBC; reflecting transferrin-mediated transport capacity), transferrin saturation (TSP; \nreflecting iron availability) and hepcidin, the central hormonal regulatory of iron homeostasis. \n \nLarge-scale genome-wide association studies (GWAS) have identified genetic variants \ninfluencing these traits, providing well-characterised ‘instruments’ for investigating the genetic \ninfluence of iron homeostasis on other phenotypes and traits. For binary traits, GWAS identify \nsmall changes in the DNA (Single nucleotide polymorphisms, SNPs) of patients with a confirmed \ndisease and assess whether these SNPs are more or less common compared to the general \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 8 \npopulation. Any undiagnosed disease in the general population will be diluted out by the very \nlarge population which does not have the disease. Many GWAS-associated SNPs lie outside \nprotein-coding regions but may be involved in regulating when, where and to what extent \ngenes are expressed, thereby providing clues to the biological pathways underlying disease. For \ncontinuous traits, GWAS assess whether genetic variants are associated with higher or lower \nlevels of the trait. \n \nConcerning iron traits, the largest GWAS meta-analysis identified 123 regions of the genome \ncontaining genetic variants associated with serum iron, ferritin, TSP and TIBC at genome-wide \nsignificance (Moksnes et al. 2022). In addition, a large meta-analysis of circulating hepcidin \nidentified 16 genomic regions containing variants significantly associated with hepcidin levels \n(Allara et al., 2024). Together, these genetic associations across complementary biomarkers of \niron physiology provide a framework for investigating whether the genetic determinants of iron \nhomeostasis overlap with those of endometriosis. \n \nIn parallel, GWAS of endometriosis to date have identified 42 regions of the genome containing \ngenetic variants robustly associated with disease rsik, with subsequent analyses implicating \ngenes and biological pathways related to reproductive biology, immune regulation and vascular \nprocesses (Rahmioglu et al., 2023). However, the extent to which the genetic determinants of \niron homeostasis overlap with those of endometriosis, and whether systemic iron biomarkers \nhave a causal role in disease risk, remains unclear.   \n \nIn this study, we investigated the genetic relationship between iron homeostasis and \nendometriosis using the largest available GWAS datasets and complementary genomic \napproaches. We assessed genome-wide genetic correlations between systemic iron biomarkers \nand endometriosis, including analyses by disease severity, identified shared loci using multi-\ntrait GWAS, and evaluated potential causal relationships using bidirectional Mendelian \nrandomisation. Through these analyses, we aimed to clarify whether systemic iron homeostasis \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 9 \nis linked to endometriosis risk and to provide insight into the biological mechanisms underlying \nthis association.  \n  \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 10 \nMATERIALS AND METHODS \nData resources and quality control \nThis study utilised the most up-to-date largest available GWAS meta-analysis results for \nendometriosis and iron biomarkers as summarised in Table 1. All GWAS included individuals of \nEuropean ancestry. \n \nTable 1. Summary of GWAS meta-analysis datasets used in our study. \nPhenotype Sample Size Sex Reference \nOverall Endometriosis 28,281 cases: 506,494 controls Female-only Rahmioglu et al. (2023) \nStage III/IV Endometriosis 9,073 cases: 506,494 controls Female-only Rahmioglu et al. (2023) \nSerum Iron 236,612 Sex-combined Moksnes et al. (2022) \nFerritin 257,953 Sex-combined Moksnes et al. (2022) \nTIBC 208,422 Sex-combined Moksnes et al. (2022) \nTransferrin Saturation 198,516 Sex-combined Moksnes et al. (2022) \nHepcidin 91,675 Sex-combined Allara et al. (2024)  \n \nGWAS meta-analysis results were downloaded as GRCh37, restricted to HapMap3 SNPs and \nfiltered for minor allele frequency (MAF) greater than 0.01, while also excluding palindromic \nvariants. Allele orientation was harmonised to ensure consistency in effect direction across \ntraits. Analyses were restricted to autosomal variants. In addition, the extended major \nhistocompatibility complex (MHC) region on chromosome 6 (approximately 24–35 Mb) had \nbeen removed. The rationale for deleting this region is that it is characterised by highly complex \nlinkage disequilibrium patterns, extensive haplotypic structure, and substantial genetic \ndiversity, which complicate the interpretation of association signals. All GWAS results datasets \nwere harmonised using the LDSC munge_sumstats pipeline \n(https://github.com/bulik/ldsc/blob/master/munge_sumstats.py). \n \nGenetic correlation analysis \nGenetic correlation quantifies the extent to which genetic effects are shared between two \ntraits: a positive genetic correlation indicates that genetic variants associated with higher levels \nof one trait tend to be associated with higher levels, or greater genetic liability, of the other, \nwhereas a negative genetic correlation indicates that their genetic effects tend to act in \nopposite directions. Genome-wide genetic correlations between traits were estimated through \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 11 \nlinkage disequilibrium score regression (LDSC), using the LDSC (v1.0.1) on GitHub (Bulik-Sullivan \nand Finucane, 2019). Analyses utilised precomputed LD scores derived from European ancestry \nindividuals from the 1000 Genomes Project Phase 3 reference panel. Pairwise correlations were \ncalculated between each iron biomarker and overall endometriosis, as well as for stage III/IV \ndisease separately. Genetic correlations were also estimated between each of the iron traits in \norder to determine how similar their genetic basis is. \n \nA Benjamini-Hochberg FDR correction (Benjamini and Hochberg, 1995) was applied to account \nfor multiple testing across the 10 pairwise genetic correlations estimated in the LDSC analysis, \ncomprising five iron traits tested against overall endometriosis and the same five iron traits \ntested against stage III/IV endometriosis. An FDR-based approach was considered more \nappropriate than Bonferroni correction, due to the high level of correlation between the iron-\nrelated biomarkers \n \nMulti-trait analysis of GWAS \nMulti-trait analysis of GWAS (MTAG) leverages shared genetic architecture between genetically \ncorrelated traits to increase power to identify associated variants, while generating trait-\nspecific association estimates (Turley et al., 2018). In this study, MTAG was used to investigate \nwhether leveraging shared genetic architecture between endometriosis and iron biomarkers \ncould identify genetic associations that were not detected when each trait was analysed \nindependently. \n \nMTAG (v1.0.8; https://github.com/JonJala/mtag) was applied pairwise, jointly analysing \nendometriosis with one iron-related biomarker at a time rather than modelling all traits \nsimultaneously. Summary statistics were harmonised using the MTAG-specific \nmtag_munge.py pipeline to standardise formatting, align alleles, remove strand-ambiguous \nvariants, and restrict analyses to high-quality SNPs suitable for cross-trait analysis. Trait-specific \nassociation estimates were generated separately for endometriosis and each iron-related \nbiomarker. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 12 \nMTAG assumes a genome-wide homogeneous variance–covariance structure of SNP effects and \nuses LD score regression to estimate genetic covariance and estimation error, including that \narising from sample overlap (Turley et al., 2018). The suitability of these assumptions was \nassessed in the context of the polygenic architecture of the traits and by inspection of LD score \nregression intercepts. \n \nLead SNP identification and locus overlap \nIndependent genetic loci were defined using LD-based clumping. Genome-wide significant \nvariants (P < 5 × 10⁻⁸) were evaluated sequentially, and a variant was retained as an \nindependent lead SNP if no previously selected lead SNP was located within ±1 Mb and pairwise \nlinkage disequilibrium was low (r² < 0.01).  \n \nThis approach was first applied to the univariate endometriosis GWAS to define genome-wide \nsignificant loci. The same procedure was then used for the MTAG-derived results to assess \nwhether joint analysis with iron biomarkers increased discovery of significant loci. Variants \nidentified through MTAG were considered novel if they reached genome-wide significance and \nshowed at least suggestive association (P < 5 × 10⁻⁵) in the univariate endometriosis GWAS. \n \nTo identify shared genetic loci between endometriosis and iron biomarkers, genome-wide \nsignificant lead SNPs from each analysis were compared. Overlap was defined as identical lead \nSNPs or lead SNPs located within ±1 Mb. Loci meeting the latter criterion were considered \nshared loci. To determine whether overlapping loci reflected the same underlying genetic \nsignal, pairwise linkage disequilibrium (LD; r²) between lead SNPs was assessed using the LDlink \nLDpair tool (https://ldlink.nih.gov/). LD ≥ 0.2 was considered evidence that variants likely \nrepresent the same underlying association signal. \n \nFunctional annotation using GTEx and Genomic Context \neQTLs and sQTLs are genetic variants associated with differences in the amount of gene \nexpression and RNA splicing, respectively, with these regulatory associations can differ between \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 13 \ntissues. We therefore used eQTL and sQTL data to investigate whether variants at the identified \nloci were associated with regulation of nearby or distant genes, and to identify candidate genes \nand tissues through which these genetic associations might act. Expression quantitative trait \nlocus (eQTL) and splicing quantitative trait locus (sQTL) data were obtained from the Genotype-\nTissue Expression (GTEx) Project  (v8; https://gtexportal.org/home/) (The GTEx Consortium, \n2020). Two sets of variants were evaluated: (i) novel genome-wide significant endometriosis \nloci identified through MTAG, and (ii) genome-wide significant loci shared between \nendometriosis and iron biomarkers.  \n \nVariants were queried across selected tissues relevant to iron metabolism and endometriosis, \nincluding liver, small intestine, spleen and whole blood, as well as reproductive tissues including \nuterus, ovary and fallopian tube. Whole blood eQTLGen data were additionally included to \nincrease power for detecting regulatory associations. Genomic context was annotated using the \nEnsembl Variant Effect Predictor (VEP) \n(https://www.ensembl.org/info/docs/tools/vep/index.html) to classify variants according to \ntheir predicted functional consequence (e.g., intronic, intergenic, missense, UTR variants). \n \nBidirectional Mendelian Randomisation \nBidirectional Mendelian randomisation (MR) uses genetic variants associated with an exposure \nas instrumental variables to investigate potential causal relationships between two traits in \nboth directions. Here, two-sample MR was used to test whether genetically predicted levels of \niron biomarkers were associated with endometriosis risk and, conversely, whether genetic \nliability to endometriosis was associated with levels of iron biomarkers. Analyses were \nconducted in R using the TwoSampleMR package (v0.6.30) (Hemani et al., 2020). \n \nGenome-wide significant variants (p < 5 × 10⁻⁸) were selected as instrumental variables. To \nreduce correlation between instruments due to linkage disequilibrium (LD), variants were \nclumped within a 10,000 kb window at three LD thresholds ( r² = 0.1, 0.01, and 0.001), using the \n1000 Genomes European reference panel and the ld_clump() function from the ieugwasr \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 14 \npackage. Analyses were repeated at each LD threshold to assess the robustness of findings to \ndifferent levels of LD pruning. \nThe inverse-variance weighted (IVW) method was used as the primary MR analysis (Burgess, \nButterworth and Thompson, 2013). Weighted median (Bowden et al., 2016) and MR-Egger \nregression (Jack Bowden, Davey Smith and Burgess, 2015) were performed as sensitivity \nanalyses because they make different assumptions regarding potential horizontal pleiotropy, \nwhereby genetic variants influence the outcome through pathways other than the exposure of \ninterest. Evidence of directional horizontal pleiotropy was assessed using the MR-Egger \nintercept test (J. Bowden, Davey Smith and Burgess, 2015), and heterogeneity between variant-\nspecific estiamtes was assessed using Cochran’s Q statistic (Bowden et al., 2019). \nMR-PRESSO was additionally used to identify potential pleiotropic outlier variants and to \nreassess causal estimates following their removal (Verbanck et al., 2018). Instrument strength \nwas evaluated using the F-statistic and proportion of variance explained, with F > 10 suggesting \nsufficient instrument strength. To account for multiple testing across five iron traits and three \nLD thresholds (15 tests), p-values were adjusted using the Benjamini–Hochberg false discovery \nrate (FDR) method (Benjamini and Hochberg, 1995). \n \nStatistical power was assessed using the mRnd MR power calculator \n(https://shiny.cnsgenomics.com/mRnd/) (Brion, Shakhbazov and Visscher, 2013). For analyses \nwith endometriosis as the binary outcome, power was estimated across hypothetical effect \nsizes corresponding to odds ratios (ORs) of 1.01, 1.05, 1.10, 1.15 and 1.20, together with the \ncorresponding inverse effects. For analyses with iron biomarkers as continuous outcomes, \npower was estimated for effect sizes of β = 0.01, 0.05, 0.10, 0.15 and 0.20 and the \ncorresponding negative effects, consistent with the parameterisation of the mRnd calculator. \n \nPleiotropy assessment \nPleiotropy occurs when a genetic variant influences more than one trait or biological process. In \nMendelian randomisation, horizontal pleiotropy occurs when an instrumental genetic variant \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 15 \ninfluences the outcome through a pathway other than the exposure being investigated, \npotentially biasing the causal estimate. In the present analysis, horizontal pleiotropy was \nfurther assessed using MR-PRESSO to identify outlier variants and evaluate whether these \nvariants distorted the causal estimates  (https://github.com/rondolab/MR-PRESSO) (Ron Do \nLaboratory, 2025) (Verbanck et al., 2018). Analyses were performed using 5,000 simulations \n(NbDistribution = 5000) and a significance threshold of P < 0.05. Where outliers were detected, \ncausal estimates were reassessed following their removal. This procedure was applied across all \nLD clumping thresholds and in both causal directions. \n \nEthical approval \nThis study used only publicly available, fully anonymised GWAS summary statistics and did not \ninvolve new data collection. Ethical approval was therefore not required under the policies \ngoverning secondary analyses of de-identified genomic data. Analyses complied with all \nrequirements for the use of publicly available genetic datasets and with the ethical principles \noutlined by Human Reproduction. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 16 \nRESULTS \nGenetic correlations indicate reduced systemic iron availability in endometriosis \nWe examined genetic correlations between iron biomarkers and endometriosis to assess \nshared genetic architecture. Strong correlations were observed between iron biomarkers \nthemselves, consistent with established iron physiology. In particular, serum iron and TSP were \nstrongly positively correlated (rg = 0.73, p = 1.31 x 10-5), as were ferritin and hepcidin (rg = 0.73, \np = 4.78 x 10-41), while TIBC showed inverse correlations with ferritin (rg = −0.29, p = 2.87 x 10-9) \nand TSP (rg = −0.50, p = 8.24 x 10-26), reflecting its role as a marker of reduced iron availability. \nGenetic correlations between iron biomarkers and endometriosis showed a consistent pattern \nindicative of increased endometriosis liability associated with reduced systemic iron availability \n(Figure 1, Supplementary Table 1). Markers reflecting circulating iron availability were inversely \ncorrelated with genetic risk of endometriosis, including ferritin (rg = −0.10, p = 0.022), TSP (rg = \n−0.16, p = 0.006), and hepcidin (rg = −0.13, p = 0.042), with a weaker inverse correlation \nobserved for serum iron (rg = −0.09, p = 0.10).  \n \nIn contrast, TIBC demonstrated a positive genetic correlation (rg = 0.16, p = 4 x 10-4). After FDR \ncorrection for multiple parallel testing, associations for TIBC (FDR-adjusted p = 0.004) and TSP \n(FDR-adjusted p = 0.031) remained statistically significant, whereas those for ferritin (FDR-\nadjusted p = 0.075) and hepcidin (FDR-adjusted p = 0.075) were attenuated. Together, these \nfindings indicate that increased genetic risk of endometriosis is consistent with reduced \nsystemic iron availability.  \n \n \n \n \n \n \n \n \n \n \n \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 17 \n \nFigure 1. Genetic correlation results using LDSC between systemic iron biomarkers and overall \nendometriosis using GWAS summary statistics. Genetic correlation estimates for stage III/IV \nendometriosis, alongside LDSC regression diagnostics including SNP heritability estimates, \nstandard errors, and intercepts for heritability and genetic covariance, are provided in \nSupplementary Table 1. \n \nAnalyses restricted to stage III/IV endometriosis showed directionally consistent results, \nalthough the statistical evidence for these associations was weaker (Figure 2, Supplementary \nTable 1). The magnitude and direction of genetic correlations for all serum iron (overall rg = \n−0.088, FDR-adjusted p = 0.125; stage III/IV rg = −0.143, FDR-adjusted p = 0.124), ferritin (overall \nrg = −0.099, FDR-adjusted p = 0.075; stage III/IV rg = −0.084, FDR-adjusted p = 0.178), TIBC \n(overall rg = 0.159, FDR-adjusted p = 0.004; stage III/IV rg = 0.128, FDR-adjusted p = 0.075), TSP \n(overall rg = −0.158, FDR-adjusted p = 0.031; stage III/IV rg = −0.157, FDR-adjusted p = 0.075) \nand hepcidin (overall rg = −0.130, FDR-adjusted p = 0.075; stage III/IV rg = −0.031, FDR-adjusted \np = 0.764) remained broadly stable, supporting a consistent signal of reduced iron availability in \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 18 \nmore extensive disease. Although slight differences in effect size in stage III/IV disease with \nrespect to overall endometriosis may reflect biological variation in systemic iron handling in \nadvanced disease, estimates were imprecise and not statistically significant, likely due to \nreduced sample size and power in stage III/IV restricted GWAS (Figure 2, Supplementary Table \n1). Given the greater statistical power of the full dataset, subsequent analyses focused on \noverall endometriosis. \n \nFigure 2. Genetic correlation results of overall and stage III/IV endometriosis with systemic iron \nbiomarkers from LDSC using GWAS summary statistics. LDSC regression diagnostics, including \nSNP heritability estimates (h²), standard errors, and intercepts for heritability and genetic \ncovariance, are provided in Supplementary Table 1. \n \n \nLeveraging shared genetic architecture with iron biomarkers enhances endometriosis locus \ndiscovery \nWe leveraged the shared genetic architecture between systemic iron biomarkers and \nendometriosis to improve locus discovery for endometriosis using MTAG. Across pairwise \nanalyses, MTAG increased the number of genome-wide significant loci for endometriosis \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 19 \nrelative to the univariate GWAS (p < 5 × 10⁻⁸), identifying eight additional novel loci (Table 2). \nFull results and comparisons with univariate endometriosis GWAS are provided in \nSupplementary Tables 2-6.   \n \nNotably, loci at 2p13 (IL1A/IL1B) and 2q35 (FN1), previously implicated in endometriosis \n(Sapkota et al., 2017) but not reaching genome-wide significance in more recent GWAS \n(Rahmioglu et al., 2023), achieved genome-wide significance in MTAG analyses. Leveraging the \nshared genetic architecture between endometriosis and iron traits thereby strengthened \nevidence for the loci’s involvement in endometriosis risk. \n \nAcross all pairwise analyses, MTAG-derived association estimates were directionally consistent \nwith univariate endometriosis effects but demonstrated modest shifts in effect size and \nprecision, consistent with incorporation of cross-trait covariance. Notably, many of these \nadditional loci were near genome-wide significance in the univariate endometriosis GWAS and \nsurpassed the P < 5 × 10⁻⁸ threshold only after MTAG, indicating increased effective power \nrevealing novel signals. \n \nFunctional annotation of the remaining loci further highlighted biologically relevant pathways. \nVariants near CALCRL suggest roles in vascular signalling (Selvarajan et al., 2024), while loci \nmapping to ARHGAP26 and EEFSEC implicate cellular signalling (Long et al., 2025) and protein \nsynthesis pathways (Simonović and Puppala, 2018). Additional signals, including those near \nNEMP1 and TBX3-AS1, point towards regulatory mechanisms potentially involved in tissue \nremodelling (Tsatskis et al., 2020) and gene expression control (Jauregi-Miguel et al., 2025). \nCollectively, these findings indicate that leveraging shared genetic architecture with iron-\nrelated traits can enhance detection of biologically relevant endometriosis loci. \n \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 20 \nTable 2. Novel 8 genome-wide significant endometriosis loci identified through MTAG with \niron-related marker GWAS results. All genome-wide significant lead endometriosis variants \nfrom MTAG results across the 5 iron biomarkers are given in supplementary table 2-6. \nSNP Chr: Position \n(B37) \nRA \n(RAF) OR (95% CI) MTAG P-\nValue \nUnivariate \nGWAS P-\nvalue \nGenomic context Implicated genes \n(eQTL)* \nrs10496444 2:113553668 G (0.33) 1.02 (1.01-1.03) 2.17x10-8 1.74x10-7 \nNovel Transcript, \nAntisense to CKAP2L, \nIL1B and IL1A (Intronic) \nBlood: IL1A, PAX8-AS1, \nPSD4, PAX8, IL1B, \nSLC20A1; Spleen: IL1A \nrs11897572 2:188316181 G (0.31) 1.03 (1.01-1.04) 1.71x10-8 1.67x10-7 CALCRL-AS1 (Intronic) \nBlood, Ovary: CALCRL; \nLiver, Spleen, Small \nintestine: CALCRL-AS1 \nrs1250258 2:216300185 C (0.25) 1.03 (1.02-1.04) 9.82x10-9 7.60x10-7 FN1 (Intronic) Blood: ATIC \nrs2999046 3:127872796 T (0.16) 1.03 (1.02-1.04) 4.09x10-8 1.01x10-7 EEFSEC (Intronic) Blood: EEFSEC, RUVBL1, \nLiver: EEFSEC \nrs17759800 3:55197620 A (0.90) 1.04 (1.02-1.05) 3.05x10-8 9.75x10-8 ENSG00000239991 \n(Intronic) Blood: CACNA2D3, ESRG \nrs7728894 5:142162633 A (0.26) 1.03 (1.02-1.04) 4.70x10-9 7.01x10-8 ARHGAP26 (Intronic) N/A \nrs7962771 12:57466843 G (0.91) 1.04 (1.03-1.06) 2.97x10-8 1.26x10-7 NEMP1 (Intronic) \nBlood: STAT6, ZBTB39, \nMETTL21B, MYO1A, \nTAC3 \nrs1566643 12:115206320 A (0.44) 1.02 (1.01-1.03) 4.77x10-8 2.86x10-7 Downstream TBX3-AS1 \n(Intergenic) N/A \nFootnote: RA: Risk allele, RAF= Risk allele frequency \n \nShared genetic loci link iron homeostasis and endometriosis risk \nTo further characterise the shared genetic architecture between systemic iron biomarkers and \nendometriosis, we examined overlap between genome-wide significant lead variants identified \nin MTAG analyses. Across analyses, 8 loci demonstrate overlap between endometriosis and at \nleast one iron-related marker (Table 3), indicating shared genetic signals.  \n \nThree loci showed genome-wide significant association in both endometriosis and iron-related \ntraits analyses with lead variants in perfect LD (r2=1), suggesting shared underlying causal \nvariants (Table 3). The most consistent signal was observed at rs6025 on chromosome 1 \n(1q24.2), which was associated with endometriosis as well as multiple iron-related traits, \nincluding ferritin, TIBC and TSP. This variant corresponds to the Factor V Leiden mutation (F5), a \nwell-characterised exonic missense variant, resulting in an amino acid substitution in the \ncoagulation factor V protein (Sinclair and Poon, 2013). This locus highlights a potential link \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 21 \nbetween iron homeostasis and coagulation pathways that may contribute to the inflammatory \nand vascular microenvironment implicated in endometriosis pathophysiology.  \nAdditional shared loci likely tagging the same causal variants are observed at 2q35 (FN1) and \n9q34.2 (ABO). The rs1250258 locus implicates FN1, a gene involved in extracellular matrix \norganisation and tissue remodelling (Soubeyrand et al., 2022), processes relevant to \nendometriotic lesion establishment and persistence (Garcia Garcia et al., 2022). The \nrs579459/rs651007 locus implicates ABO, which influences coagulation and inflammatory \npathways (Johansson et al., 2015), suggesting a role for haematological and vascular \nmechanisms linking iron traits and endometriosis risk. \nThe remaining overlapping loci containing genome-wide significant association for both \nendometriosis and iron biomarkers presented with low pairwise LD, suggesting independent \nsignals within the same genomic region (Table 3). The WNT4 locus implicates genes involved in \nreproductive development (Pitzer et al., 2021) and cell migration (Li et al., 2025) (CDC42, \nHSPG2), while the locus near KCTD9 highlights genes related to neuroendocrine signalling \n(Stevenson et al., 2012) and cellular motility (Frank et al., 2017) (GNRH1, DOCK5). At the \nchromosome 8 locus, signals implicate SLC25A37, a key mitochondrial iron transporter (Chen et \nal., 2009), alongside genes involved in extracellular matrix remodelling (LOXL2) (Peng et al., \n2025). Loci near KDR and SRD5A3 suggest roles in angiogenesis (Krikun, 2012), circadian \nregulation (Bolsius et al., 2021) and steroid metabolism (Son et al., 2023), while \nthe TMEM165 locus points to broader metal ion homeostasis (Foulquier et al., 2012). The \nchromosome 12 region implicates STAT6 and related genes, supporting involvement of immune \nand inflammatory pathways (Wang, Wang and Zha, 2021), alongside genes linked to cellular \nstress responses (Jauhiainen et al., 2012). Additional loci, including those \nnear RIN3 and SERPINA9, suggest roles in vesicular trafficking (Shen et al., 2020) and immune \nfunction (Tang et al., 2013). Collectively, these findings highlight overlap between iron \nhomeostasis and endometriosis across pathways including coagulation, angiogenesis, immune \nregulation and extracellular matrix remodelling.  \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 22 \nTable 3. Shared genome-wide significant loci between endometriosis and systemic iron \nbiomarkers from MTAG results. e/sQTLs were identified in female reproductive tissues \nincluding uterus, ovary, fallopian tubes, liver, spleen, small intestine, blood. Please see \nsupplementary table 7 for detailed look-up of eQTLs per variant per tissue. \nTraits SNP Chr:Pos A1 (A1F) OR (95% CI) P-Value LD (r2) Genomic \ncontext \nImplicated genes \n(e/sQTL) \nEndometriosis rs6025 1:169519049 C (0.98) 1.09 (1.06-1.12) 4.25x10-10 \n1 \nF5 \n(Missense \nvariant)  \nBlood: ATP1B1, NME7, \nMETTL18, SLC19A2, \nSELL Spleen: SCYL3  \nFerritin rs6025 1:169519049 C (0.98) 0.86 (0.84-0.88) 3.15x10-30 \nTIBC rs6025 1:169519049 C (0.97) 1.12 (1.09-1.14) 5.09x10-20 \nTSP rs6025 1:169519049 C (0.97) 0.92 (0.90-0.95) 5.48x10-11 \nEndometriosis rs3765350 \n 1:22447316 G (0.22) 1.05 (1.04-1.06) 7.02x10-25 \n0.005 \nWNT4 \n(Intronic) \nBlood: CDC42, \nLINC00339, USP48, \nHSPG2, CDC42-AS1 \nFerritin rs12568930 1:22702231 C (0.18) 1.03 (1.02-1.04) 4.40x10-13 \nDownstream \nWNT4 \n(Intergenic) \nN/A \nEndometriosis rs1250258 2:216300185 C (0.25) 1.03 (1.02-1.04) 9.82x10-9 \n1 FN1 \n(Intronic) Blood: ATIC \nFerritin rs1250258 2:216300185 C (0.32) 0.98 (0.97-0.98)  2.22x10-12 \nEndometriosis rs10020668 4:55996903 G (0.73) 1.05 (1.04-1.06) 8.55x10-25 \n0.001 \nUpstream \nKDR \n(Intergenic) \nBlood, Spleen: SRD5A3 \nBlood: CLOCK \nFerritin rs3749474 4:56300685 C (0.64) 1.02 (1.01-1.02) 1.20x10-8 TMEM165 \n(Intronic) \nBlood, Small Intestine, \nSpleen: TMEM165  \nLiver, Blood, Small \nintestine, Spleen: \nSRD5A3  \nOvary, Blood, Small \nIntestine Spleen: CLOCK \nEndometriosis rs17053711 8:25311269 G (0.72) 1.03 (1.02-1.04) 1.73x10-8 \n0.001 \nKCTD9 \n(Intronic) Blood: GNRH1, DOCK5 \nFerritin rs7009973 8:23374454 G (0.34) 1.02 (1.01-1.03) 1.07x10-9 Intergenic Blood: SLC25A37, \nLOXL2, NKX3-1 \nEndometriosis rs579459 9:136154168 C (0.21) 1.03 (1.02-1.04) 1.44x10-11 \n1 \nIntergenic \nLiver, Spleen, Vagina, \nBlood: ABO \nBlood: SURF1, GBGT1, \nCACFD1 \nFerritin rs651007 9:136153875 C (0.84) 1.05 (1.04-1.06) 3.44x10-30 Intergenic \nLiver, Spleen, Vagina, \nBlood: ABO \nBlood: SURF1, GBGT1, \nCACFD1 \nEndometriosis rs17119407 12:57452308 T (0.91) 1.04 (1.03-1.05) 4.87x10-8 \n0.001 \nNEMP1 (3ʹ \nUTR) \nBlood: STAT6, ZBTB39, \nMYO1A, METTL21B, \nTAC3 \nHepcidin rs4760355 12:57725197 A (0.27) 0.97 (0.96-0.98) 3.91x10-9 R3HDM2 \n(Intronic) \nBlood: STAT6, \nMETTL21B, TMEM194A, \nMARS, DDIT3, ATP23 \nEndometriosis rs749647 14:93080703 C (0.65) 1.02 (1.02-1.03) 4.49x10-8 \n0.001 \nRIN3 \n(Intronic) N/A \nSerum Iron rs7151526 14:94863636 C (0.96) 0.95 (0.93-0.97) 2.89x10-8 SERPINA9 \n(Intronic) N/A \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 23 \nBidirectional Mendelian randomisation suggests limited evidence for causal effects between \niron status and endometriosis \nTo determine whether the observed genetic correlations and shared genetic architecture \nbetween endometriosis and iron biomarkers were indicative of causal relationships, we next \nperformed bidirectional Mendelian randomisation (MR). MR uses genetic variants associated \nwith an exposure as proxies for that exposure to investigate its potential causal effect on an \noutcome. Because genetic variants are determined at conception and generally precede the \ndevelopment of disease, this approach can reduce the influence of reverse causation and many \nenvironmental or behavioural confounders that can affect conventional observational \nassociations. Genome-wide significant variants across multiple LD clumping thresholds (r² = 0.1, \n0.01, 0.001) were used as instrumental variables (IVs) (Supplementary Table 8). IVs therefore \nrepresent genetic proxies for the exposure, with their associations with the exposure and \noutcome used to estimate the potential causal effect between the two traits. LD clumping limits \ncorrelation between IVs, with increasingly stringent r² thresholds selecting progressively more \nindependent variants and thereby allowing the robustness of estimates to instrument selection \nto be assessed. Primary estimates were obtained using the inverse-variance weighted (IVW) \nanalyses, with weighted median and MR-Egger regression used as sensitivity analyses. Overall, \nthere was limited and inconsistent evidence supporting causal effects in either direction (Figure \n2). \nIn forward MR, where variability in iron biomarkers was the exposure and endometriosis the \noutcome, a suggestive association was observed between genetically predicted ferritin levels \nand reduced endometriosis risk in IVW model (OR = 0.84, 95% CI 0.73–0.98, nominal p = 0.023, \nFDR-adjusted p = 0.115 at r² = 0.001), with directionally consistent estimates across clumping \nthresholds (Supplementary Table 8). However, this association was not supported by weighted \nmedian or MR-Egger analyses, indicating limited robustness (Figure 2). A nominal association \nwas also observed for TSP at the most permissive threshold (nominal p = 0.021, FDR-adjusted p \n= 0.115, r² = 0.1), but this was not consistent across models. No evidence of causal effects was \nobserved for serum iron, TIBC or hepcidin. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 24 \nIn reverse MR, where endometriosis was considered as the exposure and variability in iron-\nrelated traits as outcomes, genetic liability to endometriosis showed a more consistent \nsuggestive association with increased TIBC across clumping thresholds (OR = 1.03, 95% CI 1.01–\n1.05, nominal p = 0.009, FDR-adjusted p = 0.045 at r² = 0.001), although not all sensitivity \nanalyses reached statistical significance. No consistent associations were observed for serum \niron, TSP or hepcidin. A weak inverse association with ferritin was observed but was not \nconsistent across methods (Figure 2, Supplementary Table 8). \nSensitivity analyses indicated substantial heterogeneity across instruments for several traits. \nHowever, MR-Egger intercept tests did not indicate strong evidence of directional pleiotropy. \nMR-PRESSO identified a small number of outlier variants Supplementary Table 9a), but their \nremoval did not materially alter the results (Supplementary Table 9b and 9c). \nInstrument strength exceeded conventional thresholds across all analyses (F > 10), indicating \nlow risk of weak instrument bias. However, the proportion of variance explained by the genetic \ninstruments was modest for several traits, particularly ferritin (R² = 0.023 – 0.038), serum iron \n(R² = 0.028 – 0.080) and hepcidin R² = 0.010–0.015) while for TIBC (R² = 0.043 – 0.188) and TSP \n(R² = 0.035 – 0.153) showed relatively higher variance explained. Power calculations indicated \nthat the analyses were well powered to detect moderate effect sizes (OR ≥ 1.1–1.5) but had \nlimited power to detect small effects (OR = 1.01–1.10) (Supplementary Table 10). Accordingly, \nthe absence of consistent causal effects should be interpreted with caution, as small but \npotentially meaningful effects may not have been detectable. \n \n \n \n \n \n \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 25 \nFigure 2. Bi-directional mendelian randomisation analysis between endometriosis (IVs=32) and \niron biomarkers (Serum Iron IVs=22, Ferritin IVs=54, TIBC IVs=43, TSP IVs=22, Hepcidin IVs=32). \nIVs were defined using an LD clumping threshold of r2=0.001. Filled circles denote main model \ninverse-variance weighted (IVM) estimates; open circles denote weighted median (WM) and \nMR-Egger estimates. P-values displayed in the figure are nominal p-values; FDR-adjusted p-\nvalues and full MR results across all linkage disequilibrium thresholds are provided \nin Supplementary Table 8. \n \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 26 \nDISCUSSION \nIn this study, we provide genetic evidence linking systemic iron homeostasis to endometriosis. \nAcross complementary analyses, we observed a consistent pattern indicating that genetic \nliability to endometriosis is associated with reduced systemic iron availability, supported by \nsignificant positive genetic correlation with TIBC (rg= 0.16, nominal p = 0.0004, FDR-adjusted p \n= 0.004) and negative correlation with TSP (rg = −0.16, nominal p = 0.006, FDR-adjusted p = \n0.031) and serum ferritin (rg= −0.10, nominal p = 0.022, FDR-adjusted p = 0.075). In addition, we \nidentified eight genomic regions containing genome-wide significant associations with both \nendometriosis and iron homeostasis. These findings extend existing knowledge of iron \ndysregulation in endometriosis, which has largely focused on local iron accumulation within \nendometriotic lesions and the peritoneal environment (Wyatt et al., 2023), by demonstrating \nthat the genetic architecture of endometriosis also overlaps with the underlying systemic iron \nhomeostasis. \n \nReduced systemic iron availability and local iron accumulation are not necessarily contradictory, \nas iron status may differ substantially between the systemic circulation and local tissue \nenvironments. The apparent paradox of local iron excess within lesions alongside a genetic \nprofile consistent with reduced systemic iron availability raises the possibility of \ncompartmentalised dysregulation of iron handling, in which local iron-rich environments coexist \nwith reduced circulating iron availability. Whether these processes are mechanistically \nconnected, for exmpale through altered iron trafficking or sequestration, or represent distinct \nmanifectations of shared underlying biological processes remains to be established. \n \nNotably, the observed genetic correlations, characterised by higher TIBC and lower ferritin and \ntransferrin saturation, indicate that this systemic iron profile is captured at the level of germline \ngenetic variation and is therefore independent of confounding by menstrual blood loss (Munro \net al., 2023) (Ekroos et al., 2024) or secondary inflammatory processes (Li and Li, 2026) that \ncomplicate observational studies (Gete et al., 2024). This suggests that altered systemic iron \nhandling may represent an intrinsic feature of endometriosis susceptibility rather than solely a \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 27 \ndownstream consequence of disease. Shared loci implicating coagulation, angiogenesis, \nextracellular matrix remodelling and immune and vascular pathways further support the \ninvolvement of systemic processes linking iron homeostasis and endometriosis. While \nMendelian randomisation analyses provided only suggestive and inconsistent evidence for \ncausal effects, the overall concordance across analytical approaches support that inherited \nsusceptibility to endometriosis and systemic iron homeostasis share biological determinants, \nrather than reduced systemic iron status necessarily being solely a downstream consequence of \ndisease. \n \nAt the locus level, three regions showed genome-wide significant associations with both \nendometriosis and iron biomarkers, with lead variants in perfect linkage disequilibrium (r² = 1), \nproviding strong evidence that the associations arise from the same underlying genetic regions. \nThe most striking signal was rs6025, the Factor V Leiden missense variant in F5, which was \nassociated with endometriosis as well as ferritin, TIBC and TSP. F5 directly implicates \ncoagulation and haemostatic processes (Tinholt et al., 2014), providing a potential biological \nconnection between endometriosis risk and systemic iron balance through pathways \ninfluencing bleeding and iron loss. eQTL signals at this locus further implicated genes including \nSELL, involved in leukocyte adhesion and trafficking (Ivetic, Hoskins Green and Hart, 2019), and \nlinking immune cell recruitment to inflammatory processes known to regulate iron \nsequestration, and SLC19A2, a thiamine transporter essential for erythroid function, providing a \nconnection to red blood cell metabolism and iron utilisation (Diaz et al., 1999). At rs1250258, \nthe involvement of FN1 points to extracellular matrix remodelling (Soubeyrand et al., 2022), a \nkey process in endometriotic lesion establishment and fibrosis (Garcia Garcia et al., 2022), \nwhich may also shape local hypoxic and inflammatory environments that influence iron \nhandling and oxidative stress, while ATIC supports cellular proliferation and metabolic activity \n(Li et al., 2017). At the chromosome 9 locus (rs579459/rs651007), eQTL evidence prioritises \nABO, a gene with well-established effects on coagulation factors and systemic inflammation \n(Johansson et al., 2015), reinforcing the role of haematological regulation in both traits. \nCollectively, these loci converge on interconnected pathways involving haemostasis, immune \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 28 \nactivation, erythroid metabolism, and extracellular matrix remodelling, supporting a model in \nwhich variation in bleeding dynamics, inflammatory responses, and tissue repair processes \njointly influence susceptibility to endometriosis and systemic iron homeostasis. \n \nIn contrast, 5 loci contained genome-wide significant associations for both endometriosis and \niron-related traits but were represented by independent lead variants (r² < 0.01), suggesting \npotential convergence on shared biological pathways rather than a single shared causal variant. \nAt the WNT4 locus, the endometriosis-associated variant implicates WNT4 alongside eQTL \ngenes such as CDC42 and HSPG2, highlighting roles in reproductive tract development (Pitzer et \nal., 2021), cell migration (Prunskaite-Hyyryläinen et al., 2016) (Cohen et al., 2017), and \nextracellular matrix organization (Farach-Carson et al., 2014), while the corresponding iron-\nassociated signal lacks clear functional annotation, suggesting partially distinct but potentially \nrelated mechanisms within this region. Similarly, at the chromosome 8 locus, iron-associated \nvariation implicates SLC25A37, a key mitochondrial iron importer essential for heme \nbiosynthesis (Chen et al., 2009), alongside LOXL2, involved in extracellular matrix remodelling \nand fibrosis (Peng et al., 2025), whereas the endometriosis signal (rs17053711) implicates \ngenes such as DOCK5 and GNRH1, linking cell migration (Frank et al., 2017) and neuroendocrine \nregulation (Stevenson et al., 2012) to disease susceptibility. At the chromosome 4 locus, both \ntraits converge on genes including CLOCK and SRD5A3, pointing toward shared influences of \ncircadian rhythm (Bolsius et al., 2021) and steroid metabolism (Son et al., 2023), while KDR \nfurther supports a role for angiogenesis in endometriosis (Krikun, 2012) (Steinthorsdottir et al., \n2016) and potentially iron distribution through vascular processes. Immune-related \nmechanisms are highlighted at the chromosome 12 locus, where both signals implicate STAT6, a \nkey regulator of type 2 immune responses (Zhang et al., 2026), alongside genes involved in \ncellular stress (Jauhiainen et al., 2012) and metabolic regulation (Son et al., 2023) (DDIT3, \nMARS), supporting a link between immune activation, inflammation, and iron homeostasis. \nAdditional loci, including those implicating TMEM165, RIN3, and SERPINA9, suggest roles in \nmetal ion homeostasis (Foulquier et al., 2012), vesicular trafficking (Shen et al., 2020), and \nimmune function (Tang et al., 2013), although their contributions are less well defined. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 29 \nTogether, these findings suggest that shared genetic architecture between endometriosis and \niron homeostasis is distributed across multiple interconnected biological systems rather than \ndriven by a single mechanism. \n \nLeveraging cross-trait genetic covariance through MTAG modestly increased locus discovery for \nendometriosis, identifying eight novel endometriosis loci beyond previously published GWAS \nmeta-analysis (Rahmioglu et al., 2023). Importantly, all novel loci were near genome-wide \nsignificance in the univariate analysis, indicating that MTAG increased power to detect \nborderline associations rather than producing genome-wide significant findings in the absence \nof univariate support. Notably, two of these were previously reported associated with \nendometriosis (Sapkota et al., 2017) were not replicated at a genome-wide level in the latest \npublished endometriosis GWAS meta-analysis (Rahmioglu et al., 2023), namely, 2p13 \n(IL1A/IL1B) and 2q35 (FN1). At 2p13 (IL1A/IL1B), the locus encompasses key pro-inflammatory \ncytokines IL-1α and IL-1β, which are central mediators of innate immune responses (Dinarello, \n2018). Both cytokines are elevated in the peritoneal fluid of women with endometriosis \n(Kondera-Anasz et al., 2005) (Akoum et al., 2008) and contribute to a pro-inflammatory \nmicroenvironment that promotes lesion establishment, angiogenesis, and pain sensitisation \n(Machairiotis, Vasilakaki and Thomakos, 2021). IL-1 signalling can also stimulate the production \nof prostaglandins (Byron et al., 2023) and matrix-degrading enzymes (Fan et al., 2007), further \nsupporting tissue invasion and remodelling (Machairiotis, Vasilakaki and Thomakos, 2021). \nImportantly, inflammatory cytokines such as IL-1 are known to influence systemic iron \nhomeostasis through regulation of hepcidin and iron sequestration, linking chronic \ninflammation with altered iron metabolism (Inamura et al., 2005). This locus therefore \nhighlights the importance of immune dysregulation as a shared mechanism between \nendometriosis and iron-related traits. \n \nIn addition, six novel endometriosis-associated regions were identified that were not reported \nbefore. The positional and eQTL evidence at these loci implicated genes involved in vascular, \nimmune, and cellular regulatory processes. Notably, STAT6 (12q13.3) emerges as a compelling \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 30 \ncandidate, with experimental evidence supporting its role in endometriosis progression through \npromotion of inflammation and cell proliferation, and showing that inhibition of STAT6 \nsignalling reduces lesion development in vivo (Lin et al., 2019). Similarly, CALCRL-AS1 (2q32.1) \nimplicates CALCRL signalling with previous experimental evidence showing that CGRP/CALCRL \npathways promote lesion development and fibrogenesis (Yan, Liu and Guo, 2019). The \nARHGAP26 (5q31.3) locus supports roles in cytoskeletal regulation (Long et al., 2025) and cell \nmigration (Chen et al., 2019), key features of endometriotic lesion invasion. Other implicated \nloci suggest roles for oxidative stress (EEFSEC, 3q21.3) (Simonović and Puppala, 2018), \ndevelopmental regulation (TBX3-AS1, 12q24.21) (Khan et al., 2020) and calcium signalling \n(CACNA2D3, 3p21.1-p14.3) (Bellessort et al., 2018), whereas additional genes likely reflect \nbroader regulatory processes.  \n \nDespite consistent genetic correlation and locus overlap, Mendelian randomisation did not \nprovide robust evidence for causal effects between systemic iron biomarkers and \nendometriosis. Mendelian randomisation uses genetic variants as proxies for an exposure to \ntest whether genetically predicted differences in that exposure are associated with an outcome \nproviding evidence about potential causal relationships that is less susceptible to reverse \ncausation and some forms of confounding than conventional observational analyses. In the \nforward direction, genetically predicted ferritin showed a suggestive inverse association at r² = \n0.1 (OR=0.85, 95% CI=0.76–0.94; nominal-p = 0.002, FDR-adjusted p = 0.030) in the variance-\nweighted models, but estimates were attenuated with higher clumping thresholds, inconsistent \nacross models. In the reverse direction, endometriosis liability was modestly associated with \nhigher TIBC, consistent across clumping thresholds (r² = 0.001: OR=1.03, 95% CI=1.01–1.05, \nnominal-p = 0.009, FDR-adjusted p = 0.045; r² = 0.01: OR=1.03, 95% CI=1.01–1.05, nominal-p = \n0.004, FDR-adjusted p = 0.045; r² = 0.1: OR=1.02, 95% CI=1.01–1.04, nominal-p = 0.006, FDR-\nadjusted p = 0.045) but findings were null in the weighted median model (Figure 2, \nSupplementary Table 8). Overall, the MR findings therefore do not establish that altered \nsystemic iron status causes endometriosis, or that endometriosis liability directly causes altered \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 31 \nsystemic iron biomarkers. \n \nSeveral factors may have limited the ability of the MR analyses to detect causal effects. \nHeterogeneity was common, particularly for ferritin and TIBC. For example, ferritin at r² = 0.1 \ndemonstrated substantial heterogeneity (RSSobs = 215.826, P global < 2 × 10⁻!). MR-PRESSO \nidentified outliers (rs1250258, rs6025, rs651007), but distortion tests were generally non-\nsignificant, and removal of outliers did not materially alter effect estimates (Supplementary \nTable 9). MR-Egger intercepts were consistently non-significant, providing little evidence of \ndirectional horizontal pleiotropy. Instrument strength represents a limitation. Although F-\nstatistics exceeded conventional thresholds (F > 10), R² values were modest, particularly for \nferritin (r²=0.001, R²=0.023; r²=0.01, R²=0.026; r²=0.1, R²=0.038) and hepcidin (r²=0.001, \nR²=0.010; r²=0.01, R²=0.011; r²=0.1, R²=0.015) (Supplementary Table 10). Such limited \nexplanatory power reduces sensitivity to detect small causal effects, especially given the \npolygenic architecture of these traits. Moreover, variance explained ranged from roughly 1% for \nhepcidin to 19% for TIBC, further underscoring limited power to detect real effects. \n \nSeveral additional limitations should be considered. The iron biomarker GWAS included both \nfemales and males, despite known sexual dimorphism in iron metabolism (Tao et al., 2023). \nFemale-specific GWAS may therefore identify genetic effects that are particularly relevant to \niron physiology in women but are attenuated in sex-combined analyses. Both the iron and \nendometriosis datasets were predominantly of European ancestry, limiting generalisability to \nother ancestral populations. Power was also substantially lower for stage III/IV endometriosis \nthan for overall disease, restricting our ability to determine whether the genetic relationships \nwith iron differ according to disease severity or phenotype. Functional annotation using eQTL \nand sQTL data is dependent on the tissues and sample sizes represented in available reference \ndatasets, with relatively limited power in some female reproductive tissues. Furthermore, \noverlap between genome-wide significant signals does not by itself establish that the same \ncausal variant or gene underlies associations with both traits; formal colocalisation, fine-\nmapping and functional validation will be required to resolve these relationships. Finally, the \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 32 \npresent genetic analyses cannot determine whether systemic and local iron dysregulation are \nmechanistically connected. \n \nFrom a biological standpoint, our findings suggest that endometriosis genetic liability aligns \nwith systemic iron restriction rather than systemic iron excess. This does not contradict the \nestablished accumulation of iron within endometriotic lesions and the peritoneal environment, \nbut instead raises the possibility that local iron-rich microenvironments coexist with a \ngenetically influenced systemic profile of reduced iron availability. Shared genetic pathways \ninvolving immune activation, coagulation, vascular function and tissue remodelling provide \npotential mechanisms through which both iron regulation and susceptibility to endometriosis \nmay be influenced. Whether the systemic and local iron phenotypes represent mechanistically \nlinked aspects of iron dysregulation or distinct consequences of shared biological pathways \nremains unknown. \n \nFrom a clinical perspective, these findings suggest that reduced systemic iron availability may \nbe a feature of the genetic architecture of endometriosis. They also indicate that associations \nbetween endometriosis and reduced systemic iron status should not necessarily be attributed \nsolely to established factors such as menstrual blood loss or dietary iron intake. However, given \nthe lack of robust causal evidence, it remains unclear whether altered iron status contributes \ndirectly to disease development or reflects downstream effects of shared biological pathways. \nFurther research is needed to determine whether iron status has clinical relevance for disease \nprogression or management. \n \nFuture work should prioritise female-specific GWAS of systemic iron biomarkers to capture sex-\nspecific effects, alongside larger subtype-specific analyses of endometriosis beyond stage III/IV. \nIn particular, distinguishing between ovarian endometrioma, peritoneal disease and deep \ndisease may help clarify whether iron-related pathways differentially influence lesion \nphenotypes. Expanding analyses to more diverse ancestral populations will also be important to \nimprove generalisability and identify population-specific genetic effects that may not be \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 33 \ncaptured in European ancestry datasets \n \nIn conclusion, our findings demonstrate a shared genetic architecture between endometriosis \nand systemic iron homeostasis, characterised by genetic correlations consistent with reduced \ncirculating iron availability. This genetic component suggests that the lower systemic iron status \nobserved in women with endometriosis may not be attributable solely to consequences such as \nheavy menstrual bleeding or inadequate dietary iron intake, but may also reflect underlying \ninherited susceptibility. Shared genomic regions implicate coagulation, immune, vascular and \ntissue-remodelling pathways, suggesting that the relationship reflects overlapping polygenic \nbiology rather than a single causal mechanism. Although the causal basis of this relationship \nremains uncertain, understanding how inherited variation in iron homeostasis and these \ninterconnected biological pathways contributes to endometriosis may provide new insights into \ndisease mechanisms and clarify the clinical significance of systemic iron status in endometriosis. \n \nAuthor’s Roles \nV.D. performed all analyses. V.D., N.R., S.M. interpreted results. V.D. and N.R. wrote the \nmanuscript. N.R., S.M. C.M.B. and K.Z. provided conceptual guidance, supervision and critical \nrevisions. S.M. served as the line manager of this fellowship at the Oxford Big Data Institute. All \nauthors contributed to interpretation of results and approved the final manuscript. \n \nAcknowledgements \nComputation used the Oxford Biomedical Research Computing (BMRC) facility, a joint \ndevelopment between Centre for Human Genetics and the Big Data Institute supported by \nHealth Data Research UK and the NIHR Oxford Biomedical Research Centre. The views \nexpressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the \nDepartment of Health. \n \nFunding \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 34 \nThis study was supported by departmental resources from the Big Data Institute, Nuffield \nDepartment of Population Health. No external funding was obtained. All GWAS datasets used in \nthis study were generated independently of this work. \n \nConflict of Interest \nN.R. is a consultant for Endogene.bio, outside this submitted work. V.D. is the co-president of \n#EndEndoSilence, a German endometriosis advocacy association, outside this submitted work.  \nS.M. declares no competing interests.  \n \nReferences \nAkoum, A. et al. (2008) ‘Imbalance in the peritoneal levels of interleukin 1 and its decoy \ninhibitory receptor type II in endometriosis women with infertility and pelvic pain’, Fertility and \nSterility, 89(6), pp. 1618–1624. Available at: https://doi.org/10.1016/j.fertnstert.2007.06.019. \nAllara, E. et al. (2024) ‘Novel loci and biomedical consequences of iron homoeostasis variation’, \nCommunications Biology, 7(1), p. 1631. Available at: https://doi.org/10.1038/s42003-024-\n07115-3. \nBellessort, B. et al. (2018) ‘Comparative analysis of molecular signatures suggests the use of \ngabapentin for the management of endometriosis-associated pain’, Journal of Pain Research, \n11, pp. 715–725. Available at: https://doi.org/10.2147/JPR.S163611. \nBenjamini, Y. and Hochberg, Y. (1995) ‘Controlling the False Discovery Rate: A Practical and \nPowerful Approach to Multiple Testing’, Journal of the Royal Statistical Society. Series B \n(Methodological), 57(1), pp. 289–300. \nBolsius, Y.G. et al. (2021) ‘The role of clock genes in sleep, stress and memory’, Biochemical \nPharmacology, 191, p. 114493. Available at: https://doi.org/10.1016/j.bcp.2021.114493. \nBowden, J. et al. (2016) ‘Consistent Estimation in Mendelian Randomization with Some Invalid \nInstruments Using a Weighted Median Estimator’, Genetic Epidemiology, 40(4), pp. 304–314. \nAvailable at: https://doi.org/10.1002/gepi.21965. \nBowden, J. et al. (2019) ‘Improving the accuracy of two-sample summary-data Mendelian \nrandomization: moving beyond the NOME assumption’, International Journal of Epidemiology, \n48(3), pp. 728–742. Available at: https://doi.org/10.1093/ije/dyy258. \nBowden, Jack, Davey Smith, G. and Burgess, S. (2015) ‘Mendelian randomization with invalid \ninstruments: effect estimation and bias detection through Egger regression’, International \nJournal of Epidemiology, 44(2), pp. 512–525. Available at: https://doi.org/10.1093/ije/dyv080. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 35 \nBowden, J., Davey Smith, G. and Burgess, S. (2015) ‘Mendelian randomization with invalid \ninstruments: effect estimation and bias detection through Egger regression’, International \nJournal of Epidemiology, 44(2), pp. 512–525. Available at: https://doi.org/10.1093/ije/dyv080. \nBrion, M.-J.A., Shakhbazov, K. and Visscher, P.M. (2013) ‘Calculating statistical power in \nMendelian randomization studies’, International Journal of Epidemiology, 42(5), pp. 1497–1501. \nAvailable at: https://doi.org/10.1093/ije/dyt179. \nBulik-Sullivan, B. and Finucane, H. (2019) LD Score Regression (LDSC) Wiki, GitHub. Available at: \nhttps://github.com/bulik/ldsc/wiki (Accessed: 25 August 2025). \nBurgess, S., Butterworth, A. and Thompson, S.G. (2013) ‘Mendelian Randomization Analysis \nWith Multiple Genetic Variants Using Summarized Data’, Genetic Epidemiology, 37(7), pp. 658–\n665. Available at: https://doi.org/10.1002/gepi.21758. \nByron, M.J. et al. (2023) ‘Interleukin-1 beta treatment alters prostaglandin-related mRNA \nexpression and cytokine abundance in equine endometrial culture’, Journal of Equine \nVeterinary Science, 125, p. 104694. Available at: https://doi.org/10.1016/j.jevs.2023.104694. \nChen, W. et al. (2009) ‘Abcb10 physically interacts with mitoferrin-1 (Slc25a37) to enhance its \nstability and function in the erythroid mitochondria’, Proceedings of the National Academy of \nSciences of the United States of America, 106(38), pp. 16263–16268. Available at: \nhttps://doi.org/10.1073/pnas.0904519106. \nChen, X. et al. (2019) ‘SMURF1-mediated ubiquitination of ARHGAP26 promotes ovarian cancer \ncell invasion and migration’, Experimental & Molecular Medicine, 51(4), pp. 1–12. Available at: \nhttps://doi.org/10.1038/s12276-019-0236-0. \nCohen, S. et al. (2017) ‘Cdc42 regulates the cellular localization of Cdc42ep1 in controlling \nneural crest cell migration’, Journal of Molecular Cell Biology, 10(5), pp. 376–387. Available at: \nhttps://doi.org/10.1093/jmcb/mjx044. \nDiaz, G.A. et al. (1999) ‘Mutations in a new gene encoding a thiamine transporter cause \nthiamine-responsive megaloblastic anaemia syndrome’, Nature Genetics, 22(3), pp. 309–312. \nAvailable at: https://doi.org/10.1038/10385. \nDinarello, C.A. (2018) ‘Overview of the IL-1 family in innate inflammation and acquired \nimmunity’, Immunological reviews, 281(1), pp. 8–27. Available at: \nhttps://doi.org/10.1111/imr.12621. \nEkroos, S. et al. (2024) ‘Menstrual blood loss is an independent determinant of hemoglobin and \nferritin levels in premenopausal blood donors’, Acta Obstetricia Et Gynecologica Scandinavica, \n103(8), pp. 1645–1656. Available at: https://doi.org/10.1111/aogs.14890. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 36 \nFan, Z. et al. (2007) ‘Activation of Interleukin-1 Signaling Cascades in Normal and Osteoarthritic \nArticular Cartilage’, The American Journal of Pathology, 171(3), pp. 938–946. Available at: \nhttps://doi.org/10.2353/ajpath.2007.061083. \nFarach-Carson, M.C. et al. (2014) ‘Border patrol: Insights into the unique role of \nperlecan/heparan sulfate proteoglycan 2 at cell and tissue borders’, Matrix Biology, 34, pp. 64–\n79. Available at: https://doi.org/10.1016/j.matbio.2013.08.004. \nFoulquier, F. et al. (2012) ‘TMEM165 Deficiency Causes a Congenital Disorder of Glycosylation’, \nAmerican Journal of Human Genetics, 91(1), pp. 15–26. Available at: \nhttps://doi.org/10.1016/j.ajhg.2012.05.002. \nFrank, S.R. et al. (2017) ‘The focal adhesion-associated proteins DOCK5 and GIT2 comprise a \nrheostat in control of epithelial invasion’, Oncogene, 36(13), pp. 1816–1828. Available at: \nhttps://doi.org/10.1038/onc.2016.345. \nGarcia Garcia, J.M. et al. (2022) ‘Endometriosis: Cellular and Molecular Mechanisms Leading to \nFibrosis’, Reproductive Sciences, 30(5), pp. 1453–1461. Available at: \nhttps://doi.org/10.1007/s43032-022-01083-x. \nGete, D.G. et al. (2024) ‘Risk of Iron Deficiency in Women With Endometriosis: A Population-\nBased Prospective Cohort Study’, Women’s Health Issues: Official Publication of the Jacobs \nInstitute of Women’s Health, 34(3), pp. 317–324. Available at: \nhttps://doi.org/10.1016/j.whi.2024.03.004. \nHemani, G. et al. (2020) Two Sample MR Functions and Interface to MRC Integrative \nEpidemiology Unit OpenGWAS Database, GitHub. Available at: \nhttps://mrcieu.github.io/TwoSampleMR/ (Accessed: 25 August 2025). \nHirata, T., Koga, K. and Osuga, Y. (2020) ‘Extra-pelvic endometriosis: A review’, Reproductive \nMedicine and Biology, 19(4), pp. 323–333. Available at: https://doi.org/10.1002/rmb2.12340. \nInamura, J. et al. (2005) ‘Upregulation of hepcidin by interleukin-1β in human hepatoma cell \nlines’, Hepatology Research, 33(3), pp. 198–205. Available at: \nhttps://doi.org/10.1016/j.hepres.2005.08.005. \nIvetic, A., Hoskins Green, H.L. and Hart, S.J. (2019) ‘L-selectin: A Major Regulator of Leukocyte \nAdhesion, Migration and Signaling’, Frontiers in Immunology, 10, p. 1068. Available at: \nhttps://doi.org/10.3389/fimmu.2019.01068. \nJauhiainen, A. et al. (2012) ‘Distinct cytoplasmic and nuclear functions of the stress induced \nprotein DDIT3/CHOP/GADD153’, PloS One, 7(4), p. e33208. Available at: \nhttps://doi.org/10.1371/journal.pone.0033208. \nJauregi-Miguel, A. et al. (2025) ‘The developmental factor TBX3 engages with the Wnt/β-\ncatenin transcriptional complex in colorectal cancer to regulate metastasis genes’, Proceedings \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 37 \nof the National Academy of Sciences of the United States of America, 122(19), p. e2419691122. \nAvailable at: https://doi.org/10.1073/pnas.2419691122. \nJohansson, Å. et al. (2015) ‘Genome-Wide Association Study Identifies That the ABO Blood \nGroup System Influences Interleukin-10 Levels and the Risk of Clinical Events in Patients with \nAcute Coronary Syndrome’, PLOS ONE, 10(11), p. e0142518. Available at: \nhttps://doi.org/10.1371/journal.pone.0142518. \nKhan, S.F. et al. (2020) ‘The roles and regulation of TBX3 in development and disease’, Gene, \n726, p. 144223. Available at: https://doi.org/10.1016/j.gene.2019.144223. \nKondera-Anasz, Z. et al. (2005) ‘Concentrations of interleukin (IL)-1α, IL-1 soluble receptor type \nII (IL-1 sRII) and IL-1 receptor antagonist (IL-1 Ra) in the peritoneal fluid and serum of infertile \nwomen with endometriosis’, European Journal of Obstetrics & Gynecology and Reproductive \nBiology, 123(2), pp. 198–203. Available at: https://doi.org/10.1016/j.ejogrb.2005.04.019. \nKrikun, G. (2012) ‘Endometriosis, Angiogenesis and Tissue Factor’, Scientifica, 2012, p. 306830. \nAvailable at: https://doi.org/10.6064/2012/306830. \nLi, C. et al. (2025) ‘HSPG2 could promote normal haematopoiesis in acute myeloid leukaemia \npatients after complete remission by repairing bone marrow endothelial progenitor cells’, \nClinical and Translational Medicine, 15(2), p. e70220. Available at: \nhttps://doi.org/10.1002/ctm2.70220. \nLi, M. et al. (2017) ‘Bifunctional enzyme ATIC promotes propagation of hepatocellular \ncarcinoma by regulating AMPK-mTOR-S6 K1 signaling’, Cell Communication and Signaling : CCS, \n15, p. 52. Available at: https://doi.org/10.1186/s12964-017-0208-8. \nLi, Y. and Li, T. (2026) ‘Association between the systemic inflammation response index and iron \ndeficiency anemia in US adults, 2003–2018: A population-based cross-sectional analysis’, \nMedicine, 105(4), p. e47302. Available at: https://doi.org/10.1097/MD.0000000000047302. \nLin, K. et al. (2019) ‘Autocrine Production of Interleukin-34 Promotes the Development of \nEndometriosis through CSF1R/JAK3/STAT6 signaling’, Scientific Reports, 9(1), p. 16781. Available \nat: https://doi.org/10.1038/s41598-019-52741-1. \nLong, A. et al. (2025) ‘Evaluation of ARHGAP26 expression, prognostic significance, and immune \ninfiltration in clear cell renal cell carcinoma’, Journal of Genetic Engineering and Biotechnology, \n23(4), p. 100593. Available at: https://doi.org/10.1016/j.jgeb.2025.100593. \nMachairiotis, N., Vasilakaki, S. and Thomakos, N. (2021) ‘Inflammatory Mediators and Pain in \nEndometriosis: A Systematic Review’, Biomedicines, 9(1), p. 54. Available at: \nhttps://doi.org/10.3390/biomedicines9010054. \nMehedintu, C. et al. (2014) ‘Endometriosis still a challenge’, Journal of Medicine and Life, 7(3), \npp. 349–357. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 38 \nMoksnes, M.R. et al. (2022) ‘Genome-wide meta-analysis of iron status biomarkers and the \neffect of iron on all-cause mortality in HUNT’, Communications Biology, 5(1), p. 591. Available \nat: https://doi.org/10.1038/s42003-022-03529-z. \nMunro, M.G. et al. (2023) ‘The relationship between heavy menstrual bleeding, iron deficiency, \nand iron deficiency anemia’, American Journal of Obstetrics and Gynecology, 229(1), pp. 1–9. \nAvailable at: https://doi.org/10.1016/j.ajog.2023.01.017. \nNg, S.-W. et al. (2020) ‘Endometriosis: The Role of Iron Overload and Ferroptosis’, Reproductive \nSciences, 27(7), pp. 1383–1390. Available at: https://doi.org/10.1007/s43032-020-00164-z. \nPeng, L. et al. (2025) ‘NUDT21 regulates lysyl oxidase-like 2(LOXL2) to influence ECM protein \ncross-linking in silica-induced pulmonary fibrosis’, Ecotoxicology and Environmental Safety, 290, \np. 117572. Available at: https://doi.org/10.1016/j.ecoenv.2024.117572. \nPitzer, L.M. et al. (2021) ‘WNT4 Balances Development vs Disease in Gynecologic Tissues and \nWomen’s Health’, Endocrinology, 162(7), p. bqab093. Available at: \nhttps://doi.org/10.1210/endocr/bqab093. \nPrunskaite-Hyyryläinen, R. et al. (2016) ‘Wnt4 coordinates directional cell migration and \nextension of the Müllerian duct essential for ontogenesis of the female reproductive tract’, \nHuman Molecular Genetics, 25(6), pp. 1059–1073. Available at: \nhttps://doi.org/10.1093/hmg/ddv621. \nRahmioglu, N. et al. (2023) ‘The genetic basis of endometriosis and comorbidity with other pain \nand inflammatory conditions’, Nature Genetics, 55(3), pp. 423–436. Available at: \nhttps://doi.org/10.1038/s41588-023-01323-z. \nRon Do Laboratory (2025) ‘rondolab/MR-PRESSO’. Available at: \nhttps://github.com/rondolab/MR-PRESSO (Accessed: 25 August 2025). \nSampson, J. (1927) ‘Peritoneal endometriosis due to the menstrual dissemination of \nendometrial tissue into the peritoneal cavity’, American Journal of Obstetrics and Gynecology, \n14, pp. 422–469. \nSapkota, Y. et al. (2017) ‘Meta-analysis identifies five novel loci associated with endometriosis \nhighlighting key genes involved in hormone metabolism’, Nature Communications, 8, p. 15539. \nAvailable at: https://doi.org/10.1038/ncomms15539. \nSelvarajan, I. et al. (2024) ‘Coronary Artery Disease risk variant dampens the expression of \nCALCRL by reducing HSF binding to shear stress responsive enhancer in endothelial cells in \nvitro’, Arteriosclerosis, thrombosis, and vascular biology, 44(6), pp. 1330–1345. Available at: \nhttps://doi.org/10.1161/ATVBAHA.123.318964. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 39 \nShafrir, A.L. et al. (2018) ‘Risk for and consequences of endometriosis: A critical epidemiologic \nreview’, Best Practice & Research. Clinical Obstetrics & Gynaecology, 51, pp. 1–15. Available at: \nhttps://doi.org/10.1016/j.bpobgyn.2018.06.001. \nShen, R. et al. (2020) ‘Upregulation of RIN3 induces endosomal dysfunction in Alzheimer’s \ndisease’, Translational Neurodegeneration, 9, p. 26. Available at: \nhttps://doi.org/10.1186/s40035-020-00206-1. \nSimonović, M. and Puppala, A.K. (2018) ‘On elongation factor eEFSec, its role and mechanism \nduring selenium incorporation into nascent selenoproteins’, Biochimica et biophysica acta. \nGeneral subjects, 1862(11), pp. 2463–2472. Available at: \nhttps://doi.org/10.1016/j.bbagen.2018.03.018. \nSinclair, G.D. and Poon, M.-C. (2013) ‘Interpreting Results From Factor V Leiden Mutation \nAnalysis: A Cautionary Note’, Clinical and Applied Thrombosis/Hemostasis, 19(1), pp. 106–107. \nAvailable at: https://doi.org/10.1177/1076029612460427. \nSon, J. et al. (2023) ‘MARS2 drives metabolic switch of non-small-cell lung cancer cells via \ninteraction with MCU’, Redox Biology, 60, p. 102628. Available at: \nhttps://doi.org/10.1016/j.redox.2023.102628. \nSoubeyrand, S. et al. (2022) ‘Common Polymorphism That Protects From Cardiovascular Disease \nIncreases Fibronectin Processing and Secretion’, Circulation: Genomic and Precision Medicine, \n15(2), p. e003428. Available at: https://doi.org/10.1161/CIRCGEN.121.003428. \nSteinthorsdottir, V. et al. (2016) ‘Common variants upstream of KDR encoding VEGFR2 and in \nTTC39B associate with endometriosis’, Nature Communications, 7(1), p. 12350. Available at: \nhttps://doi.org/10.1038/ncomms12350. \nStevenson, T.J. et al. (2012) ‘Gonadotropin-releasing hormone plasticity: A comparative \nperspective’, Frontiers in Neuroendocrinology, 33(3), pp. 287–300. Available at: \nhttps://doi.org/10.1016/j.yfrne.2012.09.001. \nTang, W. et al. (2013) ‘Association of SERPINA9 gene variants with carotid artery \natherosclerosis: the Atherosclerosis Risk in Communities (ARIC) Carotid MRI Study’, \nInternational Journal of Molecular Epidemiology and Genetics, 4(4), pp. 258–267. \nTao, H. et al. (2023) ‘Differences in hepatocellular iron metabolism underlie sexual dimorphism \nin hepatocyte ferroptosis’, Redox Biology, 67, p. 102892. Available at: \nhttps://doi.org/10.1016/j.redox.2023.102892. \nThe GTEx Consortium (2020) ‘The GTEx Consortium atlas of genetic regulatory effects across \nhuman tissues’, Science, 369(6509), pp. 1318–1330. Available at: \nhttps://doi.org/10.1126/science.aaz1776. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 40 \nTinholt, M. et al. (2014) ‘Increased coagulation activity and genetic polymorphisms in the F5, \nF10 and EPCR genes are associated with breast cancer: a case-control study’, BMC Cancer, 14, \np. 845. Available at: https://doi.org/10.1186/1471-2407-14-845. \nTosti, C. et al. (2015) ‘Pathogenetic Mechanisms of Deep Infiltrating Endometriosis’, \nReproductive Sciences, 22(9), pp. 1053–1059. Available at: \nhttps://doi.org/10.1177/1933719115592713. \nTsatskis, Y. et al. (2020) ‘The NEMP family supports metazoan fertility and nuclear envelope \nstiffness’, Science Advances, 6(35), p. eabb4591. Available at: \nhttps://doi.org/10.1126/sciadv.abb4591. \nTurley, P. et al. (2018) ‘Multi-trait analysis of genome-wide association summary statistics using \nMTAG’, Nature Genetics, 50(2), pp. 229–237. Available at: https://doi.org/10.1038/s41588-017-\n0009-4. \nVerbanck, M. et al. (2018) ‘Detection of widespread horizontal pleiotropy in causal relationships \ninferred from Mendelian randomization between complex traits and diseases’, Nature Genetics, \n50(5), pp. 693–698. Available at: https://doi.org/10.1038/s41588-018-0099-7. \nWang, W., Wang, L. and Zha, B. (2021) ‘The roles of STAT6 in regulating B cell fate, activation, \nand function’, Immunology Letters, 233, pp. 87–91. Available at: \nhttps://doi.org/10.1016/j.imlet.2021.02.006. \nWölfler, M.M. et al. (2013) ‘Reduced hemopexin levels in peritoneal fluid of patients with \nendometriosis’, Fertility and Sterility, 100(3), pp. 777-781.e2. Available at: \nhttps://doi.org/10.1016/j.fertnstert.2013.05.010. \nWyatt, J. et al. (2023) ‘The role of iron in the pathogenesis of endometriosis: a systematic \nreview’, Human Reproduction Open, 2023(3), p. hoad033. Available at: \nhttps://doi.org/10.1093/hropen/hoad033. \nYan, D., Liu, X. and Guo, S.-W. (2019) ‘Neuropeptides Substance P and Calcitonin Gene Related \nPeptide Accelerate the Development and Fibrogenesis of Endometriosis’, Scientific Reports, \n9(1), p. 2698. Available at: https://doi.org/10.1038/s41598-019-39170-w. \nZhang, S. et al. (2026) ‘STAT6: Multidimensional analysis from hematopoietic immune \nregulation to pathogenesis mechanisms and therapeutic targets in hematologic malignancies’, \nAnnals of Hematology, 105(4), p. 183. Available at: https://doi.org/10.1007/s00277-026-06932-\n2. \nZondervan, K.T. et al. (2018) ‘Endometriosis’, Nature Reviews Disease Primers, 4(1), p. 9. \nAvailable at: https://doi.org/10.1038/s41572-018-0008-5. \nZondervan, K.T., Becker, C.M. and Missmer, S.A. (2020) ‘Endometriosis’, New England Journal of \nMedicine, 382(13), pp. 1244–1256. Available at: https://doi.org/10.1056/NEJMra1810764. \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint \n\n \n 41 \n \n . CC-BY-NC-ND 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 28, 2026. ; https://doi.org/10.64898/2026.08.25.26361232doi: medRxiv preprint","source_license":"CC0","license_restricted":false}