{"paper_id":"46bdaa27-16da-4ab1-bfcc-23ce910aa91e","body_text":"ARTICLE\nGenome-wide association and epidemiological\nanalyses reveal common genetic origins between\nuterine leiomyomata and endometriosis\nC.S. Gallagher et al. #\nUterine leiomyomata (UL) are the most common neoplasms of the female reproductive tract\nand primary cause for hysterectomy, leading to considerable morbidity and high economic\nburden. Here we conduct a GWAS meta-analysis in 35,474 cases and 267,505 female\ncontrols of European ancestry, identifying eight novel genome-wide signi ficant (P <5×1 0\n−8)\nloci, in addition to con firming 21 previously reported loci, including multiple independent\nsignals at 10 loci. Phenotypic strati fication of UL by heavy menstrual bleeding in 3409 cases\nand 199,171 female controls reveals genome-wide signi ficant associations at three of the 29\nUL loci: 5p15.33 ( TERT), 5q35.2 ( FGFR4) and 11q22.3 ( ATM). Four loci identi fied in the meta-\nanalysis are also associated with endometriosis risk; an epidemiological meta-analysis across\n402,868 women suggests at least a doubling of risk for UL diagnosis among those with a\nhistory of endometriosis. These findings increase our understanding of genetic contribution\nand biology underlying UL development, and suggest overlapping genetic origins with\nendometriosis.\nhttps://doi.org/10.1038/s41467-019-12536-4 OPEN\n*email: netta_makinen@dfci.harvard.edu; cmorton@bwh.harvard.edu. #A full list of authors and their af filiations appears at the end of the paper.\nNATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications 1\n1234567890():,;\nThere are amendments to this paper\n\nU\nterine leiomyomata (UL), also known as uterine\nfibroids, are hormone-driven tumors with an estimated\nprevalence ranging from 20 –77%1,2. Although the\nmajority of UL are asymptomatic, about 25% of women with UL\nare symptomatic, and may experience heavy menstrual bleeding\n(HMB), abdominal pain, infertility, and increased risk of mis-\ncarriage\n3. Currently, the only essentially curative treatment is\nuterine extirpation via total hysterectomy. Known risk factors for\nUL include increasing age up to menopause, ethnicity (particu-\nlarly African ancestry), family history of UL, nulliparity, and\nincreased body mass index (BMI) 4. Studies of familial aggregation\nand twins, as well as racial differences in prevalence and mor-\nbidity, suggest heritable factors in fluence risk for developing\nUL5–10. Recent GWAS have identi fied 26 loci signi ficantly asso-\nciated ( P< 5×1 0 −8) with UL: 10q24.33, 11p15.5, and 22q13.1 in\nJapanese women 11, 25 loci in white women of European ances-\ntry12–15, including the three previously identi fied loci in Japanese\nwomen, and a distinct region at 22q13.1 in African American\nwomen16.\nTo define further the genetic architecture of UL, we perform a\ndiscovery meta-analysis of GWAS on UL across a total of 35,474\ncases and 267,505 female controls of white European ancestry,\nwhich more than doubles the case sample size of previously\nreported GWAS 11,12,14–16. The meta-analysis identi fies eight\nnovel loci signi ficantly associated with UL ( P< 5×1 0 −8) and\nconfirms 21 previously reported European risk loci. Interestingly,\nHMB-limited UL GWAS reveals three of the 29 independent loci\nto be signi ficantly associated with the co-occurrence of UL and\nHMB. Four loci identi fied in the meta-analysis are also reported\nto be associated with risk for endometriosis, which together with\nan epidemiological meta-analysis indicating an association\nbetween endometriosis and diagnosis of UL suggest overlapping\ngenetic origins between the two highly common gynecologic\ndiseases.\nResults\nUL GWAS meta-analysis . Our discovery meta-analysis of GWAS\non UL includes four population-based cohorts and one direct-to-\nconsumer cohort of white European ancestry: Women ’s Genome\nHealth Study (WGHS), Northern Finnish Birth Cohort (NFBC),\nQIMR Berghofer Medical Research Institute (QIMR), UK Biobank\n(UKBB), and 23andMe (Supplementary Methods, Supplementary\nTable 1). Imputation of genotypes was carried out using 1000\nGenomes Project Phase 3 and Haplotype Reference Consortium\n(HRC) reference panels. UL phenotype in each cohort was ana-\nlyzed in a logistic regression or linear mixed model assuming\nadditive genetic effects with multivariate adjustment for age, BMI,\nand/or correction for population structure. After quality control\nmetrics were applied, including exclusion of non-informative\n(MAF < 0.01) and poorly imputed (r2 < 0.4) SNPs, we performed a\nfixed-effects, inverse-variance-weighted (IVW) meta-analysis\nacross 35,474 cases with a clinical or self-reported history of UL\nand 267,505 unaffected female controls. Altogether 8,662,096\nbiallelic SNPs were analyzed and adjustments for genomic in fla-\ntion performed (Supplementary Fig. 1, Supplementary Table 2).\nThrough linkage disequilibrium score (LDSC) regression analysis,\nan estimated 89.5% of the genomic in flation factor ( λ\nGC)o f1 . 1 2\nwas attributable to polygenic heritability (intercept = 1.02, s.e. =\n0.0081). Overall, individual SNP-based heritability ( h2) was esti-\nmated to be 0.0281 (s.e. = 0.0029) on the liability scale.\nRisk loci associated with UL . We observe genome-wide sig-\nnificant associations ( P< 5×1 0 −8) at 2505 SNPs across 29\nindependent loci (Table 1, Supplementary Fig. 2, Supplementary\nTable 3). The Manhattan plot is shown in Fig. 1. We identify eight\nnovel loci associated with UL (2p23.2, 4q22.3, 6p21.31, 7q31.2,\n10p11.22, 11p14.1, 12q15, and 12q24.31), which include the fol-\nlowing candidate genes of interest: HMGA1, BABAM2, and\nWNT2. HMGA1 is a member of the high mobility group proteins\nand is involved in regulation of gene transcription 17. Somatic\nrearrangements of HMGA1 at 6p21 have been recurrently\ndocumented in UL, albeit at a much lower frequency than those\nof HMGA2—another member of the high mobility group protein\nfamily18–20. BABAM2 at 2p23.2 encodes a death receptor-\nassociating intracellular protein that promotes tumor growth by\nsuppressing apoptosis 21. Associations at 7q31.2 containing\nWNT2, a member of the Wnt gene family, provide support for the\npreviously suggested role of Wnt signaling in UL 22,23.\nAmong 29 independent loci are 21 loci previously reported to\nbe signi ficantly associated with UL 11–16. A number of identi fied\nloci harbor genes previously implicated in cell growth and cancer\nrisk in different tissue types, including cervical cancer 24, epithelial\novarian cancer 25,26, breast cancer 27,28, glioma 29,30, bladder\ncancer31, and pancreatic cancer 32–34. Speci fically, seven indepen-\ndent loci contain well-characterized oncogenes and tumor\nsuppressor genes from the Cancer Gene Census list in\nCOSMIC35: PDGFRA, TERT, ESR1 , WT1, ATM, FOXO1, and\nTP53.\nUsing approximate conditional analysis, we identify multiple\ndistinct association signals for UL at 10 loci (at locus-wide\nsignificance, P< 1×1 0 −5, Bonferroni correction) (Supplementary\nTable 4). Fine-mapping was conducted on all 43 distinct\nassociation signals arising from the 29 detected UL loci, revealing\nthree association signals with a single variant in the 99% credible\nset (Fig. 2, Supplementary Table 5). The missense variant at\n20p12.3 (rs16991615; E341K) maps to MCM8, a gene that\nencodes a protein involved in DNA double-strand break repair 36.\nMCM8 has also been implicated in length of reproductive\nlifespan, menopause, and premature ovarian failure 37,38. Another\nvariant (rs78378222) resides in the 3 ’UTR of TP53 at 17p13.1,\nand has been shown to disturb 3 ’-end processing of TP53\nmRNA39. This variant has been associated with both malignant\nand benign tumor types 39–41.\nUL GWAS limited by HMB . HMB, one of the major symptoms\nof UL, is estimated to affect up to 30% of reproductive-aged\nwomen, having a considerable impact on a woman ’s quality of\nlife. Thus, variants speci fically associated with this symptom are\nof particular interest for drug target development. We performed\na GWAS on UL limited by HMB using a linear mixed model\nacross 3409 cases and 199,171 unaffected female controls from\nthe UKBB (Supplementary Methods, Supplementary Fig. 3). We\nobserve genome-wide signi ficant associations ( P< 5×1 0 −8)a t\nthree of the 29 independent UL loci: 5p15.33 (rs72709458, OR\n[95% CI] = 0.86 [0.81 –0.91], P = 3.50 × 10−8), 5q35.2\n(rs2456181, OR [95% CI] = 0.87 [0.83 –0.91], P = 4.20 × 10−10),\nand 11q22.3 (rs1800057, OR [95% CI] = 0.66 [0.58 –0.76], P =\n2.80 x 10 −9) (Fig. 3, Supplementary Fig. 4, Supplementary\nTable 6). The lead SNP at 11q22.3, a missense variant in ATM,\nhas been associated with increased risk of various cancers, such as\nbreast cancer 42,43, while the lead SNP at 5p15.33, an intronic\nTERT variant, has previously been implicated in gliomas 44. The\nlead SNP rs2456181 at 5q35.2 resides near FGFR4, a gene\nencoding a cell-surface receptor for fibroblast growth factors\ninvolved in regulation of several pathways, including cell pro-\nliferation, differentiation, and migration.\nHMB GWAS . A GWAS based solely on HMB across 9813 cases\nand 210,946 female controls reveals one genome-wide signi ficant\nassociation at 11p14.1, one of the eight novel loci associated with\nARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4\n2 NATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications\n\nUL (Supplementary Figs. 5 and 6). The lead SNPs for UL and\nHMB at 11p14.1 are in high LD, and the direction of the effect is\nthe same (Supplementary Fig. 7). This locus has previously been\nassociated with endometriosis, age at menarche, and follicle-sti-\nmulating/luteinizing hormone levels 45–47. According to GTEx\n(v7), the lead SNP for HMB (rs11031005) is a potential expression\nquantitative trait locus (eQTL) for ARL14EP in several tissue\ntypes, such as testis and thyroid. Mendelian randomization (MR)\nwas used to assess the causality of genetic association between UL\n(exposure) and HMB (outcome). Interestingly, MR reveals that\nTable 1 Overview of lead SNPs with signi ficant associations at 29 independent loci in UL GWAS meta-analysis\nLocus Lead SNP RA OA RAF EUR PMeta OR (95% CI) Gene(s) of interest a\n1p36.12b,c rs7412010 C G 0.15 2.4 × 10 −29 1.13 (1.11 –1.16) WNT4, CDC42\n2p23.2 rs55819434 A G 0.91 5.6 × 10 −09 0.92 (0.90-0.95) BABAM2\n2p25.1b,c rs35417544 T C 0.69 2.3 × 10 −19 1.09 (1.07 –1.10) GREB1\n3q26.2c rs35446936 A G 0.24 1.0 × 10 −08 0.95 (0.93-0.96) TERC\n4q12c rs62323682 T C 0.94 4.9 × 10 −18 0.87 (0.84-0.90) LNX1, PDGFRA\n4q13.3c rs12640488 A G 0.52 4.0 × 10 −14 0.94 (0.92-0.96) SULT1B1\n4q22.3 rs4699299 T C 0.69 4.7 × 10 −08 0.95 (0.94-0.97) PDLIM5\n5p15.33c rs72709458 T C 0.23 4.7 × 10 −21 1.10 (1.08 –1.13) TERT\n5q35.2c rs2456181 C G 0.49 1.1 × 10 -11 0.94 (0.93-0.96) ZNF346, UIMC1\n6p21.31 rs116251328 A T 0.02 3.0 × 10 −08 1.15 (1.09 –1.21) GRM4, HMGA1\n6q25.2b,c rs58415480 C G 0.84 1.9 × 10 −54 0.84 (0.82-0.86) SYNE1, ESR1\n7q31.2 rs2270206 A C 0.16 4.6 × 10 −08 1.06 (1.04 –1.09) WNT2\n9p24.3c rs10976689 A G 0.60 2.4 × 10 -13 0.94 (0.93-0.96) ANKRD15\n10q24.3c rs9419958 T C 0.13 1.1 × 10 −16 1.10 (1.08 –1.13) OBFC1, SLK\n10p11.22 rs10508765 A G 0.80 1.5 × 10 −10 1.07 (1.05 –1.09) ZEB1, ARHGAP12\n11p15.5c rs547025 T C 0.92 1.5 × 10 −14 1.13 (1.09 –1.16) RIC8A, BET1L\n11p14.1b rs11031006 A G 0.14 5.7 × 10 −15 0.91 (0.89-0.93) FSHB\n11p13c rs61889186 C G 0.86 1.4 × 10 -25 0.89 (0.87-0.91) WT1\n11p13c rs2785202 C G 0.55 6.9 × 10 −14 1.06 (1.05 –1.08) PDHX, CD44\n11q22.3c rs149934734 T C 0.03 1.1 × 10 −27 1.33 (1.26 –1.40) C11orf65, KDELC2\n12q13.11c rs2131371 A C 0.28 1.6 × 10 −18 0.93 (0.91-0.94) SLC38A2\n12q15 rs11178393 T C 0.89 3.3 × 10 −08 1.08 (1.05 –1.10) PTPRR\n12q24.31 rs28583837 A G 0.22 2.3 × 10 −08 0.94 (0.92-0.96) PITPNM2\n13q14.11c rs117245733 A G 0.02 5.7 × 10 −14 1.31 (1.21 –1.39) FOXO1\n17p13.1c rs78378222 T G 0.99 7.1 × 10 −31 0.65 (0.60-0.70) SHBG, TP53\n20p12.3c rs16991615 A G 0.07 8.8 × 10 −10 1.11 (1.07 –1.14) MCM8, TRMT6\n22q13.1c rs4821939 A T 0.20 7.8 × 10 −16 1.08 (1.06 –1.10) TNRC6B\nXp26.2c rs12392108 A T 0.31 5.9 × 10 −46 1.13 (1.11 –1.15) RAP2C\nXq13.1c rs4360450 A G 0.37 2.1 × 10 −18 1.08 (1.06 –1.10) MED12\nSNP single-nucleotide polymorphism, RA risk allele, OA other allele, RAFEUR average risk allele frequency in European samples, OR odds ratio\na≤300 kb distant from association signal\nbLoci previously associated with endometriosis\ncLoci previously associated with UL\n50\n40\n30\n–log10 (p)\n20\n10\n0\n1 23 4 5 6 7 8 9\nChromosome\n10 11 12 13 14 15 16 17 18 20 22 23\nFig. 1 Manhattan plot for UL GWAS meta-analysis across all cohorts. Meta-analysis of GWAS including 302,979 women of white European ancestry\nacross all cohorts identi fied 29 independent loci associated with UL. Red and blue horizontal lines indicate genome-wide signi ficant ( P< 5×1 0 −8) and\nsuggestive ( P< 1×1 0 −5) thresholds, respectively\nNATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4 ARTICLE\nNATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications 3\n\ngenetic predisposition to UL is causally linked to an increased risk\nof HMB, with the β estimate of 0.26 being signi ficant in the IVW\nmodel (P = 1.2 × 10−12) in the absence of heterogeneity ( P = 0.13)\n(Supplementary Table 7). The MR Egger regression shows no\nsignificant directional pleiotropy (intercept = 0.01, P = 0.36)\nsupporting a causal relationship.\nOverlap of UL and endometriosis . Interestingly, signi ficant\nassociation signals are observed at several loci previously asso-\nciated with endometriosis: 1p36.12 (rs7412010, OR [95% CI] =\n1.13 [1.11 –1.16], P = 2.43 × 10−29), 2p25.1 (rs35417544, OR\n[95% CI] = 1.09 [1.07 –1.10], P = 2.32 × 10−19), 6q25.2\n(rs58415480, OR [95% CI] = 1.19 [1.17 –1.22], P = 1.86 × 10−54),\nand 11p14.1 (rs11031006, OR [95% CI] = 1.10 [1.07 –1.12], P =\n5.65 × 10−15)45,48–50. LD is strong between UL and previously\nreported endometriosis lead SNPs45 at all except one locus, 2p25.1\n(Supplementary Table 8). In addition, the direction of effect is\nthe same between the lead SNPs at 1p36.12. Using LDSC regres-\nsion, we observe a moderate genetic correlation between UL\nand endometriosis in women with European ancestry ( rg = 0.39,\ns.e = 0.05, P = 9.77 × 10−13). Endometriosis has an earlier age-of-\nonset than UL, with a mean age of 25 –29 years and 35 years,\nrespectively. MR suggests that genetic predisposition to endome-\ntriosis (exposure) is causally linked to an increased risk of UL\n(outcome); the β of 0.36 is signi ficant (P = 3.7 × 10−3) in the IVW\nmodel (heterogeneity P = 9.5 × 10−68) (Supplementary Table 7).\nLeave-one-out sensitivity analysis reveals that no single SNP alone\ndrives the signi ficant relationship between endometriosis and UL,\nbut instead the relationship is accounted for by contributions from\nmultiple variants across the genome (Supplementary Fig. 8). Given\nthe high degree of heterogeneity, the effect sizes were estimated in\na minimal set of SNPs that when used as a genetic instrument\neliminate the heterogeneity (Supplementary Fig. 9). The effect size\nestimate ( β = 0.12) from the minimal set of variants remains\nsignificant ( P = 4.3 × 10−3) in the IVW model in the absence of\nheterogeneity ( P = 0.23). We also applied the MR pleiotropy\nresidual sum and outlier (MR-PRESSO) global and distortion tests\nto adjust for variants causing signi ficant bias in the estimates\nthrough horizontal pleiotropy. Outlier-adjusted estimates still\nprovide significant evidence for a causal estimate of endometriosis\non UL ( β = 0.29, P = 0.002) (Supplementary Table 7).\n15abc\n10\n–log10 (p-value)\n–log10 (p-value)\n–log10 (p-value)\n5\n0\n40\nLHFP\n0.8\nr 2\n0.6\n0.4\n0.2\n0.8\nr 2\n0.6\n0.4\n0.2\n0.8\nr 2\n0.6\n0.4\n0.2\nMIR4305\nCOG6\nrs117245733 rs78378222\nrs16991615\nLINC00548\nLINC00332 LINC00598 MIR320D1\nASGR1\nACADVL\nDLG4\nDVL2\nPHF23\nGABARAP\nELP5\nCLDN7\nCTDNEP1\nMIR324\nEIF5A\nACAP1\nKCTD11\nTMEM95\nTNK1\nPLSCR3\nTMEM256\nTNFSF13\nTNFSF13TNFSF12\nPOLR2A\nFXR2\nSHBG DNAH2\nSLC35G6\nTP53EIF4A1ZBTB4\nTNFSF12NLGN2\nSENP3\nSENP3-EIF4A1\nSNORA48\nSNORD10\nSCARNA21RPL29P2\nEFNB3\nWRAP53\nKDM6B\nTMEM85\nNAA38\nCHD3\nGUCY2D HES7\nALOX15B\nCYB5D1\nLOC284023\nKCNAB3\nCNTROB\nTRAPPC1\nVAMP2\nLOC643406\nLINC00654\nLOC101929207\nLOC101929225GPCPD1\nCHGB CRLS1\nMCM8\nCASC20FERMT1TRMT6C20orf196MIR6883\nPER1\nALOXE3\nALOX12B\nMIR4314\nSPEM1\nC17orf74\nTMEM102\nFOXO1 TPTE2P5\nMRPS31\nSLC25A15\nMIR621\nSUGT1P3\n40.5 41\nPosition on chr13 (Mb) Position on chr17 (Mb) Position on chr20 (Mb)\n41.5\n0\n20\n40\n60\n80\n100\n0\n20\n40\n60\n80\n100\nRecombination rate (cM/Mb)\nRecombination rate (cM/Mb)\nRecombination rate (cM/Mb)\n0\n20\n40\n60\n80\n100\n7.2 7.4 7.6 7.8 8 5.6 5.8 6 6.2 6.4\n1 gene\nomitted\n13 gene\nomitted1 gene\nomitted\n0\n0\n2\n4\n6\n8\n10\n5\n10\n15\n20\n25\n30\nFig. 2 Fine-mapping reveals three association signals with a single driver in 99% credible set. Association with UL is expressed as −log10(P value) for the\nthree signals on chromosomes: ( a) 13q14.11, ( b) 17p13.1, and ( c) 20p12.3. The labeled SNP represents the most signi ficant SNP for each locus. SNP\nassociation P-value is shown on the y axis, while SNP position (with gene annotation) appears on the x axis. Each SNP is colored according to the strength\nof LD with the lead SNP. Regional association plots were produced in LocusZoom\n25\n20\n15\n–log10 (p)\n10\n5\n0\n1 23 4 5 6 7 8 9\nChromosome\n10 11 12 13 14 15 16 17 19 21 23\nFig. 3 Manhattan plot for GWAS on UL limited by heavy menstrual bleeding. GWAS across 202,580 women of white European ancestry identi fied three\nindependent loci associated with UL limited by heavy menstrual bleeding. Red and blue horizontal lines indicate genome-wide signi ficant (P< 5×1 0 −8) and\nsuggestive ( P< 1×1 0 −5) thresholds, respectively\nARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4\n4 NATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications\n\nEndometriosis, de fined by ectopic growth of endometrial-like\ntissue outside the uterus, is a common in flammatory hormone-\ndependent disease that affects reproductive-aged women 51.\nAlthough functional studies of relevant tissue are needed to\nconfirm consequences of the variants in regulation of gene\nexpression, each of the four observed overlapping genomic loci\ncontains a gene(s) known to be involved in progesterone or\nestrogen signaling. WNT4 at 1p36.12 encodes a secreted signaling\nfactor that promotes female sex development, and regulates both\npostnatal uterine development and progesterone signaling during\ndecidualization\n52,53. Recently, SNPs at 1p36.12 associated with a\ngreater endometriosis risk have been suggested to act through\nCDC42, a gene that encodes a small GTPase of the Rho family 54.\nGREB1 at 2p25.1 is an early response gene in the estrogen\nreceptor (ER)-regulated pathway, and promotes growth of breast\nand pancreatic cancer cells 55,56. ESR1 at 6q25.2 encodes the alpha\nsubunit of the ligand-activated nuclear ER that regulates cell\nproliferation in the uterus 57. FSHB at 11p14.1 encodes the\nbiologically active subunit of follicle-stimulating hormone, which\nregulates maturation of ovarian follicles and release of ova during\nmenstruation58,59.\nEpidemiological meta-analysis . Given shared risk loci and\ngenetic correlation of UL and endometriosis, we conducted an\nepidemiological meta-analysis including 402,868 women from\nthree population-based cohorts: Nurses ’ Health Study II (NHSII),\nWomen’s Health Study (WHS), and UKBB (Supplementary\nMethods, Supplementary Table 9), to assess the likelihood of UL\ndiagnosis among women who had or had not been diagnosed\nwith endometriosis. Women with endometriosis had a sig-\nnificantly higher likelihood of UL diagnosis (multivariable-\nadjusted summary relative risk (RR) [95% CI] = 2.17 [1.48–3.19])\n(Fig. 4). All cohort-speci fic analyses demonstrated at least a\ndoubling of risk, suggesting a robust association (Table 2).\nHowever, biologically and statistically signi ficant heterogeneity\nwas observed in the pooling of effect size estimates in the meta-\nanalysis ( P <1×1 0 −4) (Fig. 4). Therefore, absolute effect esti-\nmates need to be interpreted in the context of source populations.\nHeterogeneity could re flect various different population sampling\nand data collection factors among the three cohorts. First, the\nvalidity of self-reported diagnosis of endometriosis and to a lesser\nextent UL are known to be <75% in general population cohorts,\nsuch as UKBB, compared to more highly validated self-\nassessment in the medical professional NHSII and WHS\ncohorts7. Second, endometriosis clinical de finitions prior to the\n1990s were more restrictive —often limited to the presence of\nendometrioma and/or “powderburn” superficial peritoneal\nlesions among adult women 60. Subsequently, de finitions have\nexpanded to recognize a wide range of super ficial peritoneal\nphenotypic presentations, as well as incidence among adolescents\nand young women 61. It may be impactful therefore that the WHS\nparticipants were ≥ 45 years of age in 1992, while NHSII parti-\ncipants were ≥ 25 years of age in 1989, and UKBB participants\nwere aged 40 to 69 in 2006. Thus, disease de finitions varied\nduring the peak calendar years of incidence among the cohorts,\nand in addition, while the NHSII were queried about endome-\ntriosis prospectively during their reproductive years, the WHS\nand UKBB cohorts were cross-sectionally asked to recall their\ngynecologic health experience decades earlier. It is also important\nto note that while WHS and NHSII participants were asked\nspecifically about endometriosis diagnosis via questionnaire, the\nUKBB data collection included qualitative interviews during\nwhich endometriosis would be documented only when the par-\nticipant herself raised it as a health issue. Those with mild\nsymptoms or those past their reproductive years and thus past the\nmoderate to severe life-impacting symptoms of the disease may\nhave been less likely to offer endometriosis among the list of their\nhealth issues. This is supported by the low prevalence of endo-\nmetriosis reported within the UKBB compared to WHS, NHSII,\nand other population-based estimates 62. However, the UKBB\nparticipants (due to the qualitative interview structure and recall\nbias) and the WHS participants (due to recall bias) could have\nbeen more likely to choose to report endometriosis if they also\nsuffered from UL together, resulting in diagnostic bias and con-\nsequently an in flation of effect estimates. Indeed, the population\nheterogeneity and differing potential for diagnostic bias by cohort\nfits with the observed differences among effect estimate magni-\ntudes with the RRs and CI widths ordered from NHSII (RR =\n1.56) to WHS (RR = 1.96) to UKBB (RR = 3.50) (Fig. 4).\nBioinformatic analyses of UL risk SNPs and loci . To estimate\nthe genetic correlation between UL and various reproductive\ntraits, as well as cardiometabolic traits/diseases, we performed LD\nHub analysis for a total of 21 traits/diseases (Supplementary\nNHSII\nWHS\nUKBB\nOverall\n0.4 0.6 0.8 1 1.5 2\nStudy RR (95% Cl)\n1.56 (1.45, 1.68)\n1.96 (1.70, 2.25)\n3.50 (2.79, 4.40)\n2.17 (1.48, 3.19)\nFig. 4 Epidemiologic meta-analysis demonstrates endometriosis is associated with UL. Random-effects, inverse-variance-weighted meta-analysis was\nperformed across the effect sizes and standard errors in 402,868 women from three cohorts (NHSII, WHS, and UKBB). Squares represent point estimates\nfrom individual studies, whiskers correspond to the 95% CIs, and the diamond represents results from the meta-analysis. There was evidence of signi ficant\nheterogeneity based on Cochran ’s Q statistic ( P< 1×1 0 −4)\nNATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4 ARTICLE\nNATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications 5\n\nData 1). We observe signi ficant correlations between increased\nrisk of UL and earlier age of menarche ( rg = −0.16, P = 3.7 ×\n10−6), earlier age of first birth ( rg = −0.14, P = 1.0 × 10−3),\nincreased levels of triglycerides ( rg = 0.13, P = 1.9 × 10−3), and\nincreased BMI ( rg = 0.11, P = 2.0 × 10−3), as previously suggested\nby epidemiological studies 63,64, illustrating that common genetic\nfactors can predispose women to both risk factors related to, for\nexample, adverse metabolic and cardiovascular disease risk and\nUL. Gene-set and tissue enrichment analyses across 8971 SNPs\nwith suggestive ( P< 1×1 0 −5) or signi ficant ( P< 5×1 0 −8)U L\nassociations using DEPICT 65 reveal enrichments (false discovery\nrate (FDR) < 0.05) in gene sets, such as steroid hormone receptor\n(GO:0035258; P = 1.03 × 10−5), hormone receptor binding\n(GO:0051427; P = 9.07 × 10−5), and nuclear hormone receptor\nbinding (GO:0035257; P = 1.53 × 10−4) (Supplementary Data 2\nand 3). The results are concordant with the hormone-driven\nnature of UL. We did not observe any cell/tissue types sig-\nnificantly enriched for the expression of the genes in the asso-\nciated loci (Supplementary Fig. 10). To identify potential causal\ngenes at UL risk loci, we used a summary-data based MR (SMR)\nmethod, including both eQTL and mQTL data from peripheral\nblood66,67. We identify 18 potential causal genes showing no\nsignificant heterogeneity in SMR ( PHEIDI >5×1 0 −3), including\nWNT4 (rs55938609, PSMR = 6.92 × 10−15), GREB1 (rs35417544,\nPSMR = 3.93 × 10−19), WT1 (rs12280757, PSMR = 1.87 × 10−18),\nand FOXO1 (rs3924478, PSMR = 5.76 × 10−10) (Supplementary\nData 4 and 5).\nFOXO1 expression in UL . To explore potential functional sig-\nnificance, we examined expression of the FOXO1 protein, a\ntranscription factor that plays an important role in cell pro-\nliferation, apoptosis, DNA repair, and stress response 68. Inter-\nestingly, inactivation of FOXO1 promotes cell proliferation and\ntumorigenesis in several hormone-regulated malignancies, such\nas prostate, breast, cervical, and endometrial cancers 69–72. Con-\nversely, we observe a signi ficant increase in nuclear FOXO1\nprotein expression in UL compared to myometrial samples using\nimmunohistochemistry on tissue microarrays (Supplementary\nFig. 11). Patient-matched tumor-normal pairs show 1.69-fold\nhigher (P = 0.01; paired t-test) nuclear FOXO1 expression in UL,\nwhile the expression is as much as 2.32-fold greater ( P = 1.52 ×\n10−9; Welch ’s t-test) when all 335 UL are considered (Supple-\nmentary Fig. 12). These results are consistent with a previous\nstudy73, which showed phosphorylated (p) FOXO1 (pSer 256)t o\nbe predominantly present in the nucleus in UL, but sequestered in\nthe cytoplasm of myometrium. The concomitant increase of\np-FOXO1 and reduced expression of its interaction partner 14-3-\n3γ in UL has been suggested to lead to impaired nuclear/cyto-\nplasmic shuttling of p-FOXO1, which promotes cell survival 73–75.\nWe performed strati fication of samples by genotype, revealing a\nstatistically signi ficant increase in FOXO1 levels of UL harboring\nthe risk allele for rs6563799 (allelic dosage, P = 0.047; homo-\nzygosity for risk allele, P = 0.035) (Supplementary Figs. 11 and\n13). An increase in FOXO1 levels of UL with the rs7986407 risk\nallele is also observed; however, the change is not statistically\nsignificant (Supplementary Figs. 11 and 13).\nDiscussion\nIn our meta-analysis of GWAS on UL, we identify 29 genomic\nloci to be signi ficantly associated with UL in women of white\nEuropean ancestry, including eight novel and 21 previously\nreported loci. Candidate genes in the identi fied loci implicate\npathways of estrogen and progesterone signaling ( ESR1, FSHB,\nGREB1, WNT2 , and WNT4), as well as cell growth ( FOXO1,\nPDGFRA, TERT, TERC, and TP53) in predisposing women to UL.\nWe do not con firm five of 26 previously identi fied loci reported to\nbe signi ficantly associated with UL 12,14–16. Two of these loci,\n3p24.1 and 16q12.1, are nominally signi ficant ( P< 1×1 0 −5)i n\nour GWAS meta-analysis, but the remaining three loci (3q29,\n17q25.3 and a distinct region at 22q13.1) do not reach nominal\nsignificance. Ancestral differences may explain the absence of the\nassociation originally identi fied in African American women in\nthe genomic region at 22q13.1, while variation in phenotypic\ndefinitions12 may underlie the two other loci.\nDiscovery of eight novel loci signi ficantly associated with UL\nreveals several candidate genes of particular interest: BABAM2,\nFSHB, HMGA1 , and WNT2. Because UL are benign tumors that\nrarely, if ever, develop into malignancy, the association between\nUL and multiple loci harboring well-known oncogenes and tumor\nsuppressor genes is also worthy of note. Fine-mapping of the\nTP53 locus identi fies rs78378222 to be the most probable causal\nvariant, which has been shown to disrupt the polyadenylation\nsequence in the 3 ’UTR of TP53 and result in reduced expression\nof mRNA 39. We also observe nuclear FOXO1 levels to be sig-\nnificantly elevated in UL when compared to myometrium.\nFOXO1 is a downstream target of the Akt signaling pathway that\nresponds to hormone signaling through the progesterone receptor\nin UL and activates proliferative responses 76.\nHMB is one of the major debilitating symptoms of UL and can\nhave a substantial impact on a woman ’s quality of life. Here, we\nreport GWAS on both UL limited by HMB and solely on HMB,\nrevealing potential targets for pharmacologic intervention:\nTable 2 Multivariable-adjusted effect estimates of the association between endometriosis and UL among women in NHSII, WHS,\nand UKBB cohorts\nCohort n UL cases Age-adjusted Multivariable-adjusted\nNurses’ Health Study II a 102,545 10,714 1.61 (1.50 –1.73) 1.57 (1.45 –1.68)\nWomen’s Health Study b 26,868 1,262 2.04 (1.78 –2.34) 1.97 (1.71 –2.26)\nUK Biobank c 273,455 19,789 3.11 (2.86 –3.37) 3.50 (2.77 –4.38)\naHazard ratios (95% con fidence intervals [CIs]) from Cox regression models. Multivariable model was adjusted for age (continuous), age at menarche (<11, 11, 12, 13, 14 –15, >15), infertility (yes, no),\nancestry (White, Black, Hispanic, Asian, other), parity (nulliparous, 1, 2, 3, 4 +), age at first birth (<25, 25 –29, >29), time since last birth (<1, 1 –3, 4 –5, 6 –7, 8 –9, 10 –12, 13 –15, ≥16), age first oral\ncontraceptive use (13 –16, 17–20, 21–24, ≥25), BMI (<20, 20 –21.9, 22–23.9, 24–24.9, 25–26.9, 27–29.9, ≥30), menstrual cycle length (<26, 26 –31, 32–50, and >50 days), smoking (never, past, current),\nrecent gynecologic/breast exam (no recent exam, recent exam), use of anti-hypertensive medications/diastolic blood pressure (no meds <65, no meds 65–74, no meds 75 –84, no meds 85 –89, no meds\n≥90, meds <65, meds 65 –74, meds 75 –84, meds 85-89, meds ≥90), and physical activity (MET hours/week: <3, 3 –<9, 9 –<18, 18 –<27, 27 –<42, ≥42).\nbOdds ratios (95% CI) from logistic regression models. Multivariable model was adjusted for age at baseline (continuous), age at menarche (<11, 11, 12 , 13, 14 –15, >15), ancestry (White, Black, Hispanic,\nAsian, other), parity (nulliparous, 1, 2, 3, ≥4), age at first birth (<25, 25 –29, >29), oral contraceptive use (ever, never), BMI (<20, 20 –21.9, 22–23.9, 24–24.9, 25–26.9, 27–29.9, ≥30), smoking (never,\npast, current), use of anti-hypertensive medications/diastolic blood pressure (no meds <65, no meds 65 –74, no meds 75-84, no meds 85 –89, no meds ≥90, meds <65, meds 65 –74, meds 75–84, meds\n85–89, meds ≥ 90), physical activity (never/rarely, <1 time/week, 1 –3 times.week, ≥4 times/week), alcohol consumption (never/rarely, 1 –3 drinks/month, 1 –6 drinks/week, ≥1 drinks/day).\ncOdds ratios (95% CI) from logistic regression models. Multivariable model was adjusted for age (<50, 50 –55, 56–60, ≥60), age at menarche (<11, 11, 12, 13, 14 –15, >15), ancestry (White, Black, Asian,\nother), parity (nulliparous, 1, 2, 3, ≥4), age at first birth (<25, 25 –29, >29), oral contraceptive use (ever, never), BMI (<20, 20 –21.9, 22 –23.9, 24 –24.9, 25 –26.9, 27 –29.9, ≥30), smoking (never, past,\ncurrent), physical activity (never/rarely, 1 –3 drinks/week, ≥4 drinks/week), alcohol consumption (never/rarely,1 –3/month, 1 –4/week, ≥1/day), and menopausal status (premenopausal,\npostmenopausal).\nARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4\n6 NATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications\n\nARL14EP, ATM, TERT, and FGFR4. In addition, MR analyses\nsuggest that genetic predisposition to UL is causally linked to an\nincreased risk of HMB. These results form a solid basis for further\nwork to elucidate the mechanisms underlying UL-related HMB\nand towards tailored treatments of UL and HMB.\nBiological overlap between UL and endometriosis, two highly\ncommon gynecologic diseases has long been suspected due to\nsimilarities in molecular mechanisms and progenitor cells. Our\nUL GWAS meta-analysis indicates that genes previously asso-\nciated with endometriosis and involved in hormone-signaling\npathways are also associated with UL ( WNT4/CDC42, GREB1,\nESR1, and FSHB). Overlap observed in the genetic etiology of\nendometriosis and UL led us to epidemiologically quantify the co-\noccurrence of these two diseases across three independent\ncohorts. The epidemiological meta-analysis indicates that women\nwith a history of endometriosis are at elevated risk for reporting\nUL. Results from our MR analyses suggest that genetic predis-\nposition to endometriosis is causally linked to increased risk of\nUL. Alternatively, given the discordance in the direction of allelic\neffects for the UL and endometriosis loci, our MR results may\nindicate a signi ficant overlap in the underlying biology of the two\ndiseases. Additional work is needed to better quantify the con-\ntribution of genetic effects to the directional relationship between\nendometriosis and UL. Results of which will enable us to quantify\nwhat portion of the MR results re flect the fundamental patho-\nbiological overlap in these two diseases of the uterus. Further\ncharacterization of the mutual pathogenic mechanisms of UL and\nendometriosis has the capacity to discover not only a deeper\nunderstanding of the underlying biology, but also treatments for\ntwo diseases that cause signi ficant morbidity in roughly one-third\nof the world ’s population.\nMethods\nSubjects. For UL GWAS meta-analysis, four population-based cohorts (WGHS,\nNFBC, QIMR and UKBB) and one direct-to-consumer cohort (23andMe) from the\nFibroGENE consortium were included (Supplementary Table 1), resulting in\n35,474 UL cases and 267,505 female controls of white European ancestry. Sample\nsizes were maximized using a basic, harmonizing phenotype de finition to separate\ncases and controls solely based on either self-report or clinically documented UL\nhistory. Our large-scale epidemiologic analysis was comprised of three population-\nbased cohorts (NHSII, WHS, and UKBB), totaling 402,869 women. HMB GWAS\nincluded the UKBB cohort, consisting of 220,759 women. Detailed descriptions of\ncohorts and sample selections are available in Supplementary Methods. All parti-\ncipants provided informed consent in accordance with the processes approved by\nthe relevant jurisdiction for human subject research for each cohort: the Partners\nHealthCare System Human Research Committee (WHS/WGHS), the Ethical\nCommittee of the Northern Ostrobothnia Hospital District (NFBC), the Human\nResearch Ethics Committee at the QIMR Berghofer Medical Research Institute and\nthe Australian Twin Registry (QIMR), the North West Multi-centre Research\nEthics Committee (UKBB), Ethical and Independent Review Services (an external\ninstitutional review board; 23andMe), and the Institutional Review Boards at\nHarvard T.H. Chan School of Public Health and Brigham and Women ’s Hospital\n(Partners Human Research Committee) (NHSII).\nGenotyping. Several different Illumina-based genotyping platforms (Illumina Inc.,\nSan Diego, CA, USA) were used: HumanHap300 Duo ‘+’ chips or the combination\nof the Human-Hap300 Duo and iSelect chips (WGHS), In finium 370cnvDuo array\n(NFBC), 317 K, 370 K, or 610 K SNP platforms (QIMR). Genotyping of partici-\npants in the UKBB was performed either on the Affymetrix UK BiLEVE or\nAffymetrix UK Biobank Axiom ® array with over 95% similarity. Genotyping of\nparticipants in the 23andMe cohort was performed on various versions of\nIllumina-based BeadChips.\nQuality control and imputation . Each cohort conducted quality control measures\nand imputation for their data. For WGHS, NFBC, QIMR, and 23andMe, all cases\nand controls with a genotyping call rate <0.98 were excluded from the study.\nImputation was performed on both autosomal and sex chromosomes using the\nreference panel from the 1000 Genomes Project European dataset (1000 G EUR)\nPhase 3. Imputation was carried out using ShapeIt2 and IMPUTE2 softwares\n77,78.\nSNPs with call rates of <99% or with deviation from Hardy-Weinberg equilibrium\n(P ≤ 1×1 0 −6) were excluded from further analyses. Population strati fication for\nthe data was examined with principal component analysis (PCA) using\nEIGENSTRAT79. The four HapMap populations were used as reference groups:\nEuropeans (CEU), Africans (YRI), Japanese (JPT), and Chinese (CHB). All\nobserved outliers were removed from the study. UKBB data QC and imputation\nwere performed centrally, prior to public release of the data 80. Genotype data used\nin the present analyses were imputed up to the Haplotype Reference Consortium\n(HRC) panel. We applied additional quality control filters to exclude poorly\nimputed SNPs ( r\n2 < 0.4) and SNPs with a MAF of <1%.\nAssociation analyses. Using additive encoding of genotypes and adjusting for age,\nBMI, and/or the first five PCs, logistic regression analysis was performed in WGHS,\nNFBC, QIMR, and 23andMe cohorts and summary statistics were provided,\nincluding beta coef ficients, χ2 values, and standard errors, for genotyped and\nimputed SNPs. The UKBB association analyses were conducted using a linear\nmixed model (BOLT-LMM v.2.3.2)\n81 adjusting for the two array types used, age\nand BMI ( fixed effects), and a random effect accounting for relatedness between\nwomen. Effect size estimates ( β and SE) from the linear mixed-model were con-\nverted to log-odds scale prior to meta-analysis. A fixed-effects, inverse-variance-\nweighted (IVW) meta-analysis on summary statistics was conducted using\nMETAL82 across all cohorts (Supplementary Data 6). A total of 8,662,096 SNPs\nwere available from at least two of the five cohorts. A quantile-quantile plot of the\nresults from meta-analysis across all GWAS cohorts is shown in Supplementary\nFig. 1. Details on the overall genomic in flation factor and number of analyzed SNPs\nfor each cohort are provided in Supplementary Table 2. For GWAS meta-analysis,\nindependence of genetic association with UL was de fined as SNPs in low LD ( r\n2 <\n0.1) with nearby ( ≤500 kb) signi ficantly associated SNPs. Individual loci corre-\nspond to regions of the genome containing all SNPs in LD ( r2 > 0.6) with index\nSNPs. Any adjacent regions within 250 kb of one another were combined and\nclassified as a single locus of association. All associated genomic regions were\nconfirmed to have lead SNPs that were either directly genotyped or that met a\nrigorously high quality imputation threshold (INFO > 0.9) in at least two cohorts.\nLinkage disequilibrium score regression (LDSC) . Analysis of residual in flation in\ntest statistics was conducted using univariate LDSC regression. Individual χ2 values\nfor each SNP analyzed in the GWAS meta-analysis was regressed onto LD scores\nestimated from the 1000 G EUR panel. Heritability calculations can be derived\nfrom analyzing the slope and y-axis intercept of the slope of the regression line.\nPercent impact of confounders, such as population strati fication, on test statistic\ninflation are quanti fied as the LDSC ratio [((intercept –1))/((mean χ\n2–1))] × 100%.\nRemaining effects [(1 –LDSC ratio) × 100%] represent the percentage of in flation\nattributed to polygenic heritability. Univariate LDSC regression was conducted\nusing the LDSC software ( https://github.com/bulik/ldsc.git). Adjustment of herit-\nability ( h\n2) calculations to the liability scale were performed by accounting for the\nprevalence of UL in the sample (~0.132) compared to the general population\n(~0.300). LDSC software was also used to estimate the genetic correlation between\nUL and endometriosis (Endo) using endometriosis GWA meta-analysis summary\ndata from Sapkota et al.\n45 consisting of only European cohorts. The heritability and\nLD score intercepts for both traits were computed, in this analysis with SNPs\npresent in both datasets for LDSC regression again using LD scores from the 1000\nG EUR panel. Genetic correlation between traits was estimated as the genetic\ncovariance among SNPs / √ h\n2UL × h2Endo.\nApproximate conditional analysis . Approximate conditional analysis, imple-\nmented in GCTA 83, was conducted to dissect distinct signals of association at each\nlocus. Of note, where lead SNPs at adjacent loci mapped within 1 Mb of each other,\nloci were combined as a single region for conditional analysis, to account for\npotential LD between SNPs in different loci. GCTA makes use of meta-analysis\nassociation summary statistics (log-OR and corresponding standard error) and a\nreference panel of individual-level genotype data to obtain LD between all pairs of\nSNPs at a locus (or region) that approximates the covariance in effect estimates in a\njoint model. For these analyses, we made use of 5000 randomly selected white\nBritish women (of European descent) as reference. We used the -cojo-slct option to\nselect index variants for each distinct association signal, at a locus-wide signi ficance\nthreshold of P <1 0\n−5, which is a conservative Bonferroni correction for the\nnumber of SNPs mapping to a locus. For loci with multiple distinct association\nsignals, we obtained the conditional association summary statistics for each by\nconditioning on all other index SNPs at the locus (or region) using the -cojo-cond\noption.\nFine-mapping distinct association signals . For each distinct association signal,\nassociation summary statistics (log-OR and corresponding standard error) were\nextracted from the meta-analysis for all SNPs at the locus (or region). For loci with\na single signal of association, we made use of association summary statistics from\nthe unconditional meta-analysis. For loci with multiple signals of association, we\nmade use of association summary statistics from the approximate conditional\nanalysis. For each SNP j, we calculated an approximate Bayes ’ factor in favor of\nNATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4 ARTICLE\nNATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications 7\n\nassociation84, given by\nΛj ¼\nffiffiffiffiffiffiffiffiffiffiffiffiffi ffi\nVj\nVj þ ω\ns\nexp\nωβ2\nj\n2Vj Vj þ ω\n/C16/C17\n2\n4\n3\n5; ð1Þ\nwhere βj and Vj denote the estimated log-OR and corresponding variance from the\nmeta-analysis. The parameter ω denotes the prior variance in allelic effects, taken\nhere to be 0.04 for a disease outcome 84. We then calculated the posterior prob-\nability, πj, that the jth SNP is causal for the association signal, given by\nπCj ¼\nΛjP\nk Λk\n; ð2Þ\nwhere the summation is over all retained variants in the locus (or region). The 99%\ncredible set for each signal was then constructed by: (i) ranking all variants\naccording to their Bayes ’ factor, Λ j; and (ii) including ranked variants until their\ncumulative posterior probability of causality is at least 0.99.\nHeavy menstrual bleeding (HMB) GWAS . The HMB GWAS was conducted\nusing data from the UKBB cohort (Supplementary Methods). Both hospital-linked\nmedical records and self-report were considered to identify women with a history\nof UL, while for HMB only hospital-linked medical records were taken into\naccount. Controls had no previous history of either UL or HMB. Association\nanalyses were performed using a linear mixed model (BOLT-LMM v.2.3.2)\n81. Effect\nsize estimates ( β and SE) from the linear mixed-model were converted to log-\nodds scale.\nMendelian randomization (MR) . MR analyses were performed using the Two\nSample Mendelian Randomization R package. GWAS summary statistics on HMB\nfrom the UKBB cohort were used to create outcome data for MR between UL\n(exposure) and HMB (outcome). To avoid overlap between samples in the exposure\nand outcome cohorts, we performed UL GWAS excluding all the HMB cases\n85.L D\npruning was performed to con firm no duplication of exposure haplotypes or SNPs.\nSubsequently, data were harmonized to ensure the same reference alleles were used in\nexposure and outcome GWAS and that the variants were present in both GWAS\ndatasets. Thirteen independent SNPs associated with UL from our GWAS meta-\nanalysis were available in the HMB GWAS summary data to test for a causal effect of\nUL on HMB. There were too few signi ficant SNPs available for HMB to test for a\ncausal effect of HMB on UL.\nGWAS summary statistics on endometriosis (with laparoscopy, without\nlaparoscopy, and all self-reported endometriosis cases) from the WHS cohort were\nused to create outcome data for MR between UL (exposure) and endometriosis\n(outcome). To avoid overlap between samples in the exposure and outcome cohorts,\nWGHS was excluded from the UL GWAS for MR analysis. Twenty-two independent\nSNPs associated with UL were available in the endometriosis GWAS summary data to\ntest for a causal effect of UL on endometriosis. For reverse causation model, summary\nstatistics from seven GWAS listing ‘endometriosis’ as the phenotype of interest\nwere available from the EMBL-EBI NHGRI GWAS catalog (Study Accession:\nGCST000797, GCST001894, GCST001720, GCST005906, GCST000916,\nGCST004549, GCST004873). Due to a low number of cases/controls or insuf ficient\nnumber of SNPs after LD pruning and data harmonizing, only one of the studies\n(GCST004549) was included in the analysis. Sixteen independent SNPs associated\nwith endometriosis were available in our UL GWAS summary data to test for a causal\neffect of endometriosis on UL. The IVW model was used to test causality between\nexposure and outcome. In addition, the IVW (Q) method was used to test for\nheterogeneity, leave-one-out sensitivity analysis to identify the effect of individual\nSNPs, and MR Egger for horizontal pleiotropy. Due to heterogeneity in our initial MR\nestimates, we have now leveraged a similar approach to the one published in Corbin\net al., 2016, to identify the minimum set of variants that when used as a genetic\ninstrument eliminate heterogeneity\n86. We also conducted the MR-PRESSO test to\nidentify and adjust for variants causing signi ficant bias through horizontal\npleiotropy87. MR-PRESSO method (1) applies a global test to evaluate whether\nhorizontal pleiotropy is present, (2) calculates the causal estimates incorporating\ncorrection for the detected horizontal pleiotropy, and (3) applies a distortion test to\nevaluate if the causal estimate is signi ficantly different after adjustment for outliers.\nWe have reported the initial estimates along with the outlier-adjusted estimates as\nboth the global and distortion tests showed signi ficant results.\nCo-morbidity analyses. Each cohort was analyzed individually with study-speci fic\nmodels chosen and covariates coded as appropriate for each cohort ’s data structure\n(Supplementary Methods). The study-speci fic effect estimates were combined using\nmeta-analysis to obtain a summary RR. Between study heterogeneity was assessed\nwith Cochran Q statistic and the I\n2 statistic88. Because heterogeneity among the\nstudies was identi fied, we reported a random-effects IVW effect estimate based on\nthe DerSimonian and Laird method 89.\nLD Hub, gene-set, cell/tissue enrichment, and SMR analyses . LD Hub ana-\nlysis90 was conducted using summary-level results data of UL GWAS meta-analysis\nto estimate the genetic correlation between UL and 21 different traits/diseases,\nincluding various reproductive traits and cardiometabolic traits/diseases that have\npublicly available GWAS results on the LD Hub repository. Multiple-testing cor-\nrection was performed (0.05/21 = 2.4 × 10−3). For gene-set and cell/tissue\nenrichment, summary statistics from the set of 8971 SNPs with suggestive ( P< 1×\n10−5) or signi ficant associations ( P< 5×1 0 −8) were analyzed using the Data-\ndriven Expression-Prioritized Integration for Complex Traits (DEPICT) soft-\nware\n65. Using the 1000 G EUR panel as a reference for LD calculations and the\n‘clumping’ algorithm in PLINK 91, we identi fied 104 independent loci at the sug-\ngestive threshold for DEPICT analyses (Supplementary Data 2). FDR < 0.05 was\nconsidered statistically signi ficant. For SMR analysis, SNPs present in at least two\nstudies in the summary statistics were considered. The analysis was run using\neQTL data from the CAGE blood dataset\n66 and mQTLs from the LBC_BSGS blood\ndataset67.\nFOXO1 immunohistochemistry and genotyping . FOXO1 immunostaining was\nperformed on two replicate tissue microarrays (TMAs) containing 335 UL and 36\nmyometrium tissue samples from 200 white women of European ancestry obtained\nfrom myomectomies and hysterectomies. Tissue cores on the replicate TMAs\nrepresent different regions of the same samples, which include corresponding\ntumor-normal tissue pairs from 35 women. Immunohistochemistry was carried out\nusing the BOND staining system (Leica Biosystems, Buffalo Grove, IL) with a\nprimary antibody dilution 1:100 (clone C29H4, Cell Signaling Technology, Dan-\nvers, MA) and hematoxylin as the counterstain. Immunostaining was analyzed\nusing Aperio ImageScope software (Leica Biosystems). Each core was evaluated for\nthe ratio of stain to counterstain taking into account variable cellularity between\ncores. Only nuclear labeling of the protein was evaluated. The average stain-to-\ncounterstain ratio was compared between patient-matched UL and myometrium\nsamples using a paired t-test (two-tailed), while an unpaired t-test (Welch ’s t-test,\ntwo-tailed) was applied to compare all UL and myometrium samples. Genomic\nDNA from 109 UL on the TMA was available for genotyping. These UL were\ngenotyped for two SNPs with genome-wide signi ficance at the 13q14.11 locus:\nrs6563799 and rs7986407. For each SNP, the average FOXO1 stain-to-counterstain\nratio was compared across increasing dosage of the risk allele using a one-way\nanalysis of variance test (two-tailed). We also performed an unpaired t-test to\ncompare mean expression of UL homozygous for the risk variant against the other\ngenotypes (Welch ’s t-test, two-tailed). P-values < 0.05 were considered statistically\nsignificant.\nURLs. For WHS see http://whs.bwh.harvard.edu/; for NFBC see http://www.oulu.\nfi/nfbc/; for QIMR see http://www.qimrberghofer.edu.au/; for UK Biobank see\nhttp://www.ukbiobank.ac.uk/; for 23andMe see https://research.23andme.com/; for\nMETAL see http://csg.sph.umich.edu/abecasis/metal/; for LDSC see https://github.\ncom/bulik/ldsc.git; for DEPICT see https://data.broadinstitute.org/mpg/depict/; for\nSMR see http://cnsgenomics.com/software/smr/; and for PLINK see http://pngu.\nmgh.harvard.edu/purcell/plink/.\nReporting summary . Further information on research design is available in\nthe Nature Research Reporting Summary linked to this article.\nData availability\nThe authors declare that the data supporting the findings of this study are available\nwithin the article and its Supplementary Information files. Summary statistics for the top\n10,000 UL GWAS meta-analysis variants are provided in Supplementary Data 6. UL\nGWAS meta-analysis summary statistics (without 23andMe), UL GWAS limited by\nHMB and HMB GWAS summary statistics will be made available through the NHGRI-\nEBI GWAS Catalog https://www.ebi.ac.uk/gwas/downloads/summary-statistics.T o\nrequest access to 23andMe GWAS summary statistics, please visit https://\nresearch.23andme.com/dataset-access/.\nReceived: 28 February 2019; Accepted: 10 September 2019;\nPublished online: 24 October 2019\nReferences\n1. Stewart, E. A. Clinical practice. Uterine fibroids. N. Engl. J. Med. 372,\n1646–1655 (2015).\n2. Cramer, S. F. & Patel, A. The frequency of uterine leiomyomas. Am. J. Clin.\nPathol. 94, 435 –438 (1990).\n3. Marino, J. L. et al. Uterine leiomyoma and menstrual cycle characteristics in a\npopulation-based cohort study. Hum. Reprod. 19, 2350 –2355 (2004).\n4. Pavone, D., Clemenza, S., Sorbi, F., Fambrini, M. & Petraglia, F. Epidemiology\nand risk factors of uterine fibroids. Best Pr. Res Clin. Obstet. Gynaecol. 46,\n3–11 (2018).\n5. Treloar, S. A., Martin, N. G., Dennerstein, L., Raphael, B. & Heath, A. C.\nPathways to hysterectomy: Insights from longitudinal twin research. Am. J.\nObstet. Gynecol. 167,8 2 –88 (1992).\nARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4\n8 NATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications\n\n6. Vikhlyaeva, E. M., Khodzhaeva, Z. S. & Fantschenko, N. D. Familial\npredisposition to uterine leiomyomas.Int. J. Gynecol. Obstet. 51,1 2 7–131 (1995).\n7. Marshall, L. M. et al. Variation in the incidence of uterine leiomyoma among\npremenopausal women by age and race. Obstet. Gynecol. 90, 967 –973 (1997).\n8. Luoto, R. et al. Heritability and risk factors of uterine fibroids-the Finnish\nTwin Cohort study. Maturitas 37,1 5 –26 (2000).\n9. Faerstein, E., Szklo, M. & Rosenshein, N. Risk factors for uterine leiomyoma: a\npractice-based case-control study. I. African-American heritage, reproductive\nhistory, body size, and smoking. Am. J. Epidemiol. 153,1 –10 (2001).\n10. Van Voorhis, B. J., Romitti, P. A. & Jones, M. P. Family history as a risk factor\nfor development of uterine leiomyomas. Results of a pilot study. J. Reprod.\nMed. 47, 663 –669 (2002).\n11. Cha, P. C. et al. A genome-wide association study identi fies three loci\nassociated with susceptibility to uterine fibroids. Nat. Genet. 43, 447 –450\n(2011).\n12. Eggert, S. L. et al. Genome-wide linkage and association analyses implicate\nFASN in predisposition to uterine leiomyomata. Am. J. Hum. Genet. 91,\n621–628 (2012).\n13. Gallagher, C. S. et al. Genome-wide association analysis identi fies 27 novel loci\nassociated with uterine leiomyomata revealing common genetic origins with\nendometriosis. Preprint at https://www.biorxiv.org/content/10.1101/324905v1\n(2018).\n14. Rafnar, T. et al. Variants associating with uterine leiomyoma highlight genetic\nbackground shared by various cancers and hormone-related traits. Nat.\nCommun. 9, 3636 (2018).\n15. Välimäki, N. et al. Genetic predisposition to uterine leiomyoma is determined\nby loci for genitourinary development and genome stability. Elife 7, e37110\n(2018).\n16. Hellwege, J. N. et al. A multi-stage genome-wide association study of uterine\nfibroids in African Americans. Hum. Genet. 136, 1363 –1373 (2017).\n17. Fusco, A. & Fedele, M. Roles of HMGA proteins in cancer. Nat. Rev. Cancer 7,\n899–910 (2007).\n18. Schoenberg Fejzo, M. et al. Translocation breakpoints upstream of the\nHMGIC gene in uterine leiomyomata suggest dysregulation of this gene by a\nmechanism different from that in lipomas. Genes Chromosomes Cancer 17,\n1–6 (1996).\n19. Williams, A. J., Powell, W. L., Collins, T. & Morton, C. C. HMGI(Y)\nexpression in human uterine leiomyomata. Involvement of another high-\nmobility group architectural factor in a benign neoplasm. Am. J. Pathol. 150,\n911–918 (1997).\n20. Sornberger, K. S. et al. Expression of HMGIY in three uterine leiomyomata\nwith complex rearrangements of chromosome 6. Cancer Genet. Cytogenet.\n114,9 –16 (1999).\n21. Chan, B. C. et al. BRE enhances in vivo growth of tumor cells. Biochem\nBiophys. Res. Commun. 326, 268 –273 (2005).\n22. Ono, M. et al. Paracrine activation of WNT/ β-catenin pathway in uterine\nleiomyoma stem cells promotes tumor growth. Proc. Natl Acad. Sci. USA 110,\n17053–17058 (2013).\n23. Mehine, M. et al. Integrated data analysis reveals uterine leiomyoma subtypes\nwith distinct driver pathways and biomarkers. Proc. Natl Acad. Sci. USA 113,\n1315–1320 (2016).\n24. Shi, Y. et al. A genome-wide association study identi fies two new cervical\ncancer susceptibility loci at 4q12 and 17q12. Nat. Genet. 45, 918 –922 (2013).\n25. Kuchenbaecker, K. B. et al. Identi fication of six new susceptibility loci for\ninvasive epithelial ovarian cancer. Nat. Genet. 47, 164 –171 (2015).\n26. Phelan, C. M. et al. Identi fication of 12 new susceptibility loci for different\nhistotypes of epithelial ovarian cancer. Nat. Genet. 49, 680 –691 (2017).\n27. Haiman, C. A. et al. A common variant at the TERT-CLPTM1L locus is\nassociated with estrogen receptor-negative breast cancer. Nat. Genet . 43,\n1210–1214 (2011).\n28. Hamdi, Y. et al. Association of breast cancer risk in BRCA1 and BRCA2\nmutation carriers with genetic variants showing differential allelic expression:\nidentification of a modi fier of breast cancer risk at locus 11q22.3. Breast\nCancer Res. Treat. 161, 117 –134 (2017).\n29. Shete, S. et al. Genome-wide association study identi fies five susceptibility loci\nfor glioma. Nat. Genet. 41, 899 –904 (2009).\n30. Melin, B. S. et al. Genome-wide association study of glioma subtypes identi fi\nes\nspecific differences in genetic susceptibility to glioblastoma and non-\nglioblastoma tumors. Nat. Genet. 49, 789 –794 (2017).\n31. Figueroa, J. D. et al. Genome-wide association study identi fies multiple loci\nassociated with bladder cancer risk. Hum. Mol. Genet . 23, 1387 –1398 (2014).\n32. Petersen, G. M. et al. A genome-wide association study identi fies pancreatic\ncancer susceptibility loci on chromosomes 13q22.1, 1q32.1 and 5p15.33. Nat.\nGenet. 42, 224 –228 (2010).\n33. Wolpin, B. M. et al. Genome-wide association study identi fies multiple\nsusceptibility loci for pancreatic cancer. Nat. Genet. 46, 994 –1000 (2014).\n34. Zhang, M. et al. Three new pancreatic cancer susceptibility signals identi fied on\nchromosomes 1q32.1, 5p15.33 and 8q24.21. Oncotarget 7, 66328–66343 (2016).\n35. Forbes, S. A. et al. COSMIC: mining complete cancer genomes in the\nCatalogue of Somatic Mutations in Cancer. Nucleic Acids Res. 39, D945–D950\n(2011).\n36. Lutzmann, M. et al. MCM8- and MCM9-de ficient mice reveal gametogenesis\ndefects and genome instability due to impaired homologous recombination.\nMol. Cell 47, 523 –534 (2012).\n37. He, C. et al. Genome-wide association studies identify loci associated with\nage at menarche and age at natural menopause. Nat. Genet. 41, 724 –728\n(2009).\n38. AlAsiri, S. et al. Exome sequencing reveals MCM8 mutation underlies ovarian\nfailure and chromosomal instability. J. Clin. Invest. 125, 258 –262 (2015).\n39. Stacey, S. N. et al. A germline variant in the TP53 polyadenylation signal\nconfers cancer susceptibility. Nat. Genet. 43, 1098 –1103 (2011).\n40. Enciso-Mora, V. et al. Low penetrance susceptibility to glioma is caused by the\nTP53 variant rs78378222. Br. J. Cancer 108, 2178 –2185 (2013).\n41. Diskin, S. J. et al. Rare variants in TP53 and susceptibility to neuroblastoma. J.\nNatl Cancer Inst. 106, dju047 (2014).\n42. Johnson, N. et al. Counting potentially functional variants in BRCA1, BRCA2\nand ATM predicts breast cancer susceptibility. Hum. Mol. Genet . 16,\n1051–1057 (2007).\n43. Schumacher, F. R. et al. Association analyses of more than 140,000 men\nidentify 63 new prostate cancer susceptibility loci. Nat. Genet. 50, 928 –936\n(2018).\n44. Kinnersley, B. et al. Genome-wide association study identi fies multiple\nsusceptibility loci for glioma. Nat. Commun. 6, 8559 (2015).\n45. Sapkota, Y. et al. Meta-analysis identi fies five novel loci associated with\nendometriosis highlighting key genes involved in hormone metabolism. Nat.\nCommun. 8, 15539 (2017).\n46. Ruth, K. S. et al. Genome-wide association study with 1000 genomes\nimputation identi fies signals for nine sex hormone-related phenotypes. Eur. J.\nHum. Genet . 24, 284 –290 (2016).\n47. Pickrell, J. K. et al. Detection and interpretation of shared genetic in fluences\non 42 human traits. Nat. Genet. 48, 709 –717 (2016).\n48. Uno, S. et al. A genome-wide association study identi fies genetic variants in\nthe CDKN2BAS locus associated with endometriosis in Japanese. Nat. Genet.\n42, 707 –710 (2010).\n49. Nyholt, D. R. et al. Genome-wide association meta-analysis identi fies new\nendometriosis risk loci. Nat. Genet . 44, 1355 –1359 (2012).\n50. Albertsen, H. M., Chettier, R., Farrington, P. & Ward, K. Genome-wide\nassociation study link novel loci to endometriosis. PLoS One 8, e58257 (2013).\n51. Bulun, S. E. Endometriosis. N. Engl. J. Med. 360, 268 –279 (2009).\n52. Biason-Lauber, A., Konrad, D., Navratil, F. & Schoenle, E. J. A WNT4\nmutation associated with Mullerian-duct regression and virilization in a 46,XX\nwoman. N. Engl. J. Med. 351, 792 –798 (2004).\n53. Franco, H. L. et al. WNT4 is a key regulator of normal postnatal uterine\ndevelopment and progesterone signaling during embryo implantation and\ndecidualization in the mouse. FASEB J. 25, 1176 –1187 (2011).\n54. Powell, J. E. et al. Endometriosis risk alleles at 1p36.12 act through inverse\nregulation of CDC42 and LINC00339. Hum. Mol. Genet.\n25, 5046 –5058\n(2016).\n55. Rae, J. M. et al. GREB 1 is a critical regulator of hormone dependent breast\ncancer growth. Breast Cancer Res. Treat. 92, 141 –149 (2005).\n56. Rae, J. M. et al. GREB1 is a novel androgen-regulated gene required for\nprostate cancer growth. Prostate 66, 886 –894 (2006).\n57. Bondesson, M., Hao, R., Lin, C. Y., Williams, C. & Gustafsson, J. A. Estrogen\nreceptor signaling during vertebrate development. Biochim Biophys. Acta\n1849, 142 –151 (2015).\n58. Layman, L. C. et al. Delayed puberty and hypogonadism caused by mutations\nin the follicle-stimulating hormone beta-subunit gene. N. Engl. J. Med. 337,\n607–611 (1997).\n59. Demeestere, I. et al. Follicle-stimulating hormone accelerates mouse oocyte\ndevelopment in vivo. Biol. Reprod. 87,1 –11 (2012).\n60. Missmer, S. A. & Cramer, D. W. The epidemiology of endometriosis. Obstet.\nGynecol. Clin. North Am. 30,1 –19 (2003).\n61. Zondervan, K. T. et al. Endometriosis. Nat. Rev. Dis. Prim. 4, 9 (2018).\n62. Shafrir, A. L. et al. Risk for and consequences of endometriosis: a critical\nepidemiologic review. Best. Pr. Res. Clin. Obstet. Gynaecol. 51,1 –15 (2018).\n63. Marshall, L. M. et al. A prospective study of reproductive factors and oral\ncontraceptive use in relation to the risk of uterine leiomyomata. Fertil. Steril.\n70, 432 –439 (1998).\n64. Uimari, O. et al. Uterine fibroids and cardiovascular risk. Hum. Reprod. 31,\n2689–2703 (2016).\n65. Pers, T. H. et al. Biological interpretation of genome-wide association studies\nusing predicted gene functions. Nat. Commun. 19, 5890 (2015).\n66. Lloyd-Jones, L. R. et al. The genetic architecture of gene expression in\nperipheral blood. Am. J. Hum. Genet. 100, 228 –237 (2017).\n67. McRae, A. et al. Identi fication of 55,000 Replicated DNA Methylation QTL.\nSci. Rep. 8, 17605 (2018).\nNATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4 ARTICLE\nNATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications 9\n\n68. Xing, Y. Q. et al. The regulation of FOXO1 and its role in disease progression.\nLife Sci. 193, 124 –131 (2018).\n69. Jackson, J. G., Kreisberg, J. I., Koterba, A. P., Yee, D. & Brattain, M. G.\nPhosphorylation and nuclear exclusion of the forkhead transcription factor\nFKHR after epidermal growth factor treatment in human breast cancer cells.\nOncogene 19, 4574 –4581 (2000).\n7 0 . H u a n g ,H . ,M u d d i m a n ,D .C .&T i n d a l l ,D .J .A n d r o g e n sn e g a t i v e l yr e g u l a t e\nforkhead transcription factor FKHR (FOXO1) through a proteolytic\nmechanism in prostate cancer cells. J. Biol. Chem. 279, 13866 –13877 (2004).\n71. Goto, T. et al. Mechanism and functional consequences of loss of FOXO1\nexpression in endometrioid endometrial cancer cells. Oncogene 27,9 –19\n(2008).\n72. Zhang, B., Gui, L. S., Zhao, X. L., Zhu, L. L. & Li, Q. W. FOXO1 is a tumor\nsuppressor in cervical cancer. GMR 14, 6605 –6616 (2015).\n73. Kovacs, K. A. et al. Involvement of FKHR (FOXO1) transcription\nfactor in human uterus leiomyoma growth. Fertil. Steril. 94, 1491 –1495\n(2010).\n74. Lv, J. et al. Reduced expression of 14-3-3 gamma in uterine leiomyoma as\nidentified by proteomics. Fertil. Steril. 90, 1892 –1898 (2008).\n75. Shen, Q. et al. Overexpression of the 14-3-3gamma protein in uterine\nleiomyoma cells results in growth retardation and increased apoptosis. Cell\nSignal 45,4 3 –53 (2018).\n76. Hoekstra, A. V. et al. Progestins activate the AKT pathway in leiomyoma cells\nand promote survival. J. Clin. Endocrinol. Metab. 94, 1768 –1774 (2009).\n77. Howie, B. N., Donnelly, P. & Marchini, J. A flexible and accurate genotype\nimputation method for the next generation of genome-wide association\nstudies. PLoS Genet. 5, e1000529 (2009).\n78. Delaneau, O., Marchini, J. & Zagury, J. F. A linear complexity phasing method\nfor thousands of genomes. Nat. Methods 9, 179 –181 (2011).\n79. Price, A. L. et al. Principal components analysis corrects for strati fication in\ngenome-wide association studies. Nat. Genet . 38, 904 –909 (2006).\n80. Bycroft, C. et al. The UK Biobank resource with deep phenotyping and\ngenomic data. Nature 562, 203 –209 (2018).\n81. Loh, P. R. et al. Ef fi\ncient Bayesian mixed-model analysis increases association\npower in large cohorts. Nat. Genet. 47, 284 –290 (2015).\n82. Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and ef ficient\nmeta-analysis of genomewide association scans. Bioinformatics 26, 2190–2191\n(2010).\n83. Yang, J. et al. Conditional and joint multiple-SNP analysis of GWAS summary\nstatistics identi fies additional variants in fluencing complex traits. Nat. Genet.\n44, 369 –375 (2012).\n84. Wake field, J. A. Bayesian measure of the probability of false discovery in\ngenetic epidemiology studies. Am. J. Hum. Genet . 81, 208 –227 (2007).\n85. Burgess, S., Davies, N. M. & Thompson, S. G. Bias due to participant overlap\nin two-sample Mendelian randomization. Genet. Epidemiol. 40, 597 –608\n(2016).\n86. Corbin, L. J. et al. BMI as a modi fiable risk factor for type 2 diabetes: re fining\nand understanding causal estimates using Mendelian randomization. Diabetes\n65, 3002 –3007 (2016).\n87. Verbanck, M., Chen, C. Y., Neale, B. & Do, R. Detection of widespread\nhorizontal pleiotropy in causal relationships inferred from Mendelian\nrandomization between complex traits and diseases. Nat. Genet. 50, 693 –698\n(2018). Erratum in: Nat Genet 50, 1196 (2018).\n88. Higgins, J. P., Thompson, S. G., Deeks, J. J. & Altman, D. G. Measuring\ninconsistency in meta-analyses. BMJ 327, 557 –560 (2003).\n89. DerSimonian, R. & Laird, N. Meta-analysis in clinical trials. Control Clin.\nTrials 7, 177 –188 (1986).\n90. Zheng, J. et al. LD Hub: a centralized database and web interface to perform\nLD score regression that maximizes the potential of summary level GWAS\ndata for SNP heritability and genetic correlation analysis. Bioinformatics 33,\n272–279 (2017).\n91. Purcell, S. et al. PLINK: a tool set for whole-genome association and\npopulation-based linkage analyses. Am. J. Hum. Genet . 81, 559 –575 (2007).\nAcknowledgements\nThe authors thank all of the women and their families who participated in WGHS,\nNFBC, QIMR, UK Biobank, 23andMe, and NHSII, and acknowledge the Channing\nDivision of Network Medicine, Department of Medicine, Brigham and Women ’s Hos-\npital and Harvard Medical School. This study was supported by the U.S. National\nInstitutes of Health (NIH)/Eunice Kennedy Shriver National Institute of Child Health\nand Human Development (NICHD) grant HD060530 to C.C.M. C.C.M. is also sup-\nported by the NIHR Manchester Biomedical Research Centre. N.M. acknowledges\nsupport from the Academy of Finland (295693) and Orion Research Foundation. H.R.H.\nis supported by NIH K22 CA193860. T.F. is supported by the NIHR Biomedical Research\nCentre, Oxford. S.E.M. is supported by the National Health and Medical Research\nCouncil (NHMRC) Fellowship Scheme (1103623). We thank the Specialized Histo-\npathology Core of the Dana-Farber/Harvard Cancer Center for FOXO1 immunostaining.\nThe Dana-Farber/Harvard Cancer Center is supported in part by an NCI Cancer Center\nSupport Grant P30 CA06516. Further acknowledgements are provided in Supplementary\nNote 1.\nAuthor contributions\nC.S.G., S.A.M., K.T.Z. and C.C.M. designed the study. O.U., C.M.B., H.M., M.-R.J., J.E.B,\nS.E.M., D.R.N., P.A.L., J.N.P. and the 23andMe Research team contributed to pheno-\ntypic/clinical aspects of the cohorts. O.U., J.P.C., N.R., T.F., D.R.V.-E., T.L.E., F.D., V.K.,\nP.M.R., S.D.G., S.E.M., G.W.M., D.R.N., D.A.H., J.Y.T., the 23andMe Research team,\nJ.R.B.P., P.A.L., J.N.P., N.G.M., A.P.M., D.I.C. and K.T.Z. contributed to genotyping,\nquality control, imputation, and/or association analysis of the genotyping data. C.S.G.\nand N.R. performed the UL GWAS meta-analysis. N.R. conducted the HB GWAS.\nR.M.C., A.P.M. and D.I.C. provided statistical genetics advice. C.S.G., N.M., N.R., Z.R.,\nS.M., G.W.M. and A.P.M. carried out or assisted with GWAS downstream analyses.\nC.S.G., H.R.H., O.U., N.S., N.R., K.L.T, J.E.B, S.A.M. and K.T.Z. contributed to large-scale\nepidemiologic analysis. N.M., C.S.G. and H.R.H. drafted the paper. G.W.M., N.G.M.,\nA.P.M., D.I.C., S.A.M., K.T.Z and C.C.M provided critical comments on the paper, draft,\nand analysis. All authors read and approved the final paper.\nCompeting interests\nK.T.Z and C.M.B through Oxford University have research collaborations in benign\ngynecology with Bayer AG, Roche Diagnostics, Volition UK, and M DNA Life Sciences.\nD.A.H., J.Y.T., and members of the 23andMe Research Team are employees of 23andMe,\nInc., and hold stock or stock options in 23andMe. The remaining authors declare no\ncompeting interests.\nAdditional information\nSupplementary information is available for this paper at https://doi.org/10.1038/s41467-\n019-12536-4.\nCorrespondence and requests for materials should be addressed to N.M. or C.C.M.\nPeer review information Nature Communications thanks Siddhartha Kar and Joellen\nSchildkraut for their contribution to the peer review of this work. Peer reviewer reports\nare available.\nReprints and permission information is available at http://www.nature.com/reprints\nPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in\npublished maps and institutional af filiations.\nOpen Access This article is licensed under a Creative Commons\nAttribution 4.0 International License, which permits use, sharing,\nadaptation, distribution and reproduction in any medium or format, as long as you give\nappropriate credit to the original author(s) and the source, provide a link to the Creative\nCommons license, and indicate if changes were made. The images or other third party\nmaterial in this article are included in the article ’s Creative Commons license, unless\nindicated otherwise in a credit line to the material. If material is not included in the\narticle’s Creative Commons license and your intended use is not permitted by statutory\nregulation or exceeds the permitted use, you will need to obtain permission directly from\nthe copyright holder. To view a copy of this license, visit http://creativecommons.org/\nlicenses/by/4.0/.\n© The Author(s) 2019, corrected publication 2022\nARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4\n10 NATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications\n\nC.S. Gallagher 1,30, N. Mäkinen 2,30, H.R. Harris 3,30, N. Rahmioglu 4,30, O. Uimari 5,6, J.P. Cook 7, N. Shigesi 5,\nT. Ferreira4,8, D.R. Velez-Edwards 9, T.L. Edwards 10, S. Mortlock 11, Z. Ruhioglu 2, F. Day 12, C.M. Becker 5,\nV. Karhunen 13,14,15, H. Martikainen 6, M.-R. Järvelin 13,14,15,16,17, R.M. Cantor 18, P.M. Ridker 19, K.L. Terry 20,21,\nJ.E. Buring19, S.D. Gordon 22, S.E. Medland 23, G.W. Montgomery 11,22, D.R. Nyholt 22,24, D.A. Hinds 25,\nJ.Y. Tung 25, the 23andMe Research Team, J.R.B. Perry 12, P.A. Lind 23, J.N. Painter 23, N.G. Martin 22,\nA.P. Morris 4,7, D.I. Chasman 19,31, S.A. Missmer 21,26,31, K.T. Zondervan 4,5,31 & C.C. Morton 2,27,28,29,31\n1Department of Genetics, Harvard Medical School, Boston, MA 02115, USA. 2Department of Obstetrics and Gynecology, Brigham and Women ’s\nHospital, Harvard Medical School, Boston, MA 02115, USA. 3Program in Epidemiology, Division of Public Health Sciences, Fred Hutchinson Cancer\nResearch Center, Seattle, WA 98109, USA. 4Wellcome Centre for Human Genetics, University of Oxford, Oxford OX3 7BN, UK. 5Endometriosis\nCaRe Centre, Nuf field Department of Women ’s and Reproductive Health, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DU, UK.\n6Department of Obstetrics and Gynecology, Oulu University Hospital and PEDEGO Research Unit & Medical Research Center Oulu, University of\nOulu and Oulu University Hospital, 90220 Oulu, Finland. 7Department of Biostatistics, University of Liverpool, Liverpool L69 3GL, UK. 8Big Data\nInstitute, Li Ka Shing Center for Health Information and Discovery, Oxford University, Oxford OX3 7LF, UK. 9Vanderbilt Genetics Institute,\nVanderbilt Epidemiology Center, Institute for Medicine and Public Health, Department of Obstetrics and Gynecology, Vanderbilt University Medica l\nCenter, Nashville, TN 37203, USA. 10Division of Epidemiology, Department of Medicine, Institute for Medicine and Public Health, Vanderbilt\nGenetics Institute, Vanderbilt University Medical Center, Nashville, TN 37203, USA. 11Institute for Molecular Bioscience, University of Queensland,\nBrisbane, QLD 4072, Australia. 12MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Institute of Metabolic Science,\nCambridge Biomedical Campus, Cambridge CB2 0QQ, UK. 13Center for Life Course Health Research, Faculty of Medicine, University of Oulu,\n90220 Oulu, Finland. 14Unit of Primary Health Care, Oulu University Hospital, 90220 Oulu, Finland. 15Department of Epidemiology and Biostatistics,\nMRC-PHE Centre for Environment and Health, School of Public Health, Imperial College London, London W2 1PG, UK. 16Biocenter Oulu, University\nof Oulu, 90220 Oulu, Finland. 17Department of Life Sciences, College of Health and Life Sciences, Brunel University London, Uxbridge, Middlesex\nUB8 3PH, UK. 18Department of Human Genetics, David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, CA 90095,\nUSA. 19Division of Preventative Medicine, Brigham and Women ’s Hospital, Harvard Medical School, Boston, MA, USA. 20Obstetrics and\nGynecology Epidemiology Center, Brigham and Women ’s Hospital and Harvard Medical School, Boston, MA 02115, USA. 21Department of\nEpidemiology, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA. 22Genetic Epidemiology, QIMR Berghofer Medical Research\nInstitute, Brisbane, QLD 4006, Australia. 23Psychiatric Genetics, QIMR Berghofer Medical Research Institute, Brisbane, QLD 4006, Australia.\n24Institute of Health and Biomedical Innovation and School of Biomedical Science, Queensland University of Technology, Brisbane, QLD 4059,\nAustralia. 2523andMe, Mountain View, CA 94041, USA. 26Department of Obstetrics, Gynecology, and Reproductive Biology, College of Human\nMedicine, Michigan State University, Grand Rapids, MI 49503, USA. 27Department of Pathology, Brigham and Women ’s Hospital, Harvard Medical\nSchool, Boston, MA 02115, USA. 28Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA. 29Manchester Centre for Audiology and\nDeafness, Manchester Academic Health Science Center, University of Manchester, Manchester M13 9PL, UK. 30These authors contributed equally:\nC.S. Gallagher, N. Mäkinen, H.R. Harris, N. Rahmioglu. 31These authors jointly supervised this work: D.I. Chasman, S.A. Missmer, K.T. Zondervan,\nC.C. Morton. A full list of consortium members appears at the end of the paper.\nthe 23andMe Research Team\nMichelle Agee 25, Babak Alipanahi 25, Adam Auton 25, Robert K. Bell 25, Katarzyna Bryc 25, Sarah L. Elson 25,\nPierre Fontanillas 25, Nicholas A. Furlotte 25, Karen E. Huber 25, Aaron Kleinman 25, Nadia K. Litterman 25,\nMatthew H. McIntyre 25, Joanna L. Mountain 25, Elizabeth S. Noblin 25, Carrie A.M. Northover 25, Steven J. Pitts 25,\nJ. Fah Sathirapongsasuti 25, Olga V. Sazonova 25, Janie F. Shelton 25, Suyash Shringarpure 25, Chao Tian 25,\nVladimir Vacic 25 & Catherine H. Wilson 25\nNATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12536-4 ARTICLE\nNATURE COMMUNICATIONS | (2019)10:4857 | https://doi.org/10.1038/s41467-019-12536-4 | www.nature.com/naturecommunications 11","source_license":"CC0","license_restricted":false}