Fig. 1
Genetic variants in CSMD1 are associated with primary gonadal dysfunction. a The landscape of genetic variation across CSMD1 among all cohorts. SHARE/GEMINI: rare deletions overlapping CSMD1 among 14,074 females and males from the SHARE and GEMINI cohorts; UKBB: rare intron 1–3 deletions observed in 63,064 females from the UK Biobank early idiopathic menopause cohort — for clarity of visualization only the deletions observed in cases are shown; WHISP: 37 rare SNVs overlapping CSMD1 among 1,526 exome-sequenced females; and Day et al.: three statistically independent lead SNPs from a large-scale GWAS of age at menarche in females are depicted as points along the bottom of the figure (left to right): rs2688326, rs2724961, and rs487542428. All coordinates are in hg19. b Summary of the human genetic evidence that CSMD1 variants are associated with gonadal function in men and women described in this study. c Stacked barplot depicting frequency of rare CNVs overlapping introns 1–3 of CSMD1 among 2,709 cases of male or female gonadal dysfunction versus 73,185 controls. Rare deletions over CSMD1 segregate significantly with cases (odds ratio = 3.6; meta-analysis logistic regression p-value = 1.2 × 10−6). d Density plots depicting effect sizes (β) of rare CSMD1 SNVs on age at menopause for synonymous (blue) and nonsynonymous (red) SNVs, stratified by protein domain (CUB vs. Sushi). SNVs in the CUB domains are significantly associated with an earlier onset of menopause when compared to SNVs in the Sushi domains (βCUB = −0.86, 95% CI [−1.56, −0.151]; βSUSHI = 0.046, 95% CI [−0.255, 0.377]; Wilcoxon rank-sum test p-value = 0.043). Synonymous mutations in both CUB and Sushi domains are centered about 0, consistent with a well-calibrated estimate of effect size. e The cumulative distribution of age at last birth in UK Biobank subjects, plotted separately by CSMD1 deletion status. There is a 1-year reduction in the mean age at last birth for women with deletions in introns 1–3 of CSMD1 compared to women without deletions (linear regression p-value = 4.1 × 10−3). Source data are provided as a Source Data file
