Loci at chromosomes 13, 19 and 20 influence age at natural menopause

Abstract

We conducted a genome-wide association study for age at natural menopause in 2,979 European women and identified six SNPs in three loci associated with age at natural menopause: chromosome 19q13.4 (rs1172822; −0.4 year per T allele (39%); P = 6.3 × 10⁻¹¹), chromosome 20p12.3 (rs236114; +0.5 year per A allele (21%); P = 9.7 × 10⁻¹¹) and chromosome 13q34 (rs7333181; +0.5 year per A allele (12%); P = 2.5 × 10⁻⁸). These common genetic variants regulate timing of ovarian aging, an important risk factor for breast cancer, osteoporosis and cardiovascular disease.

Menopause, the time of a woman’s life when menstrual cycle ceases owing to depletion of the follicle pool, is a key event in reproductive aging. It influences a woman’s well-being and is an important risk factor for several major age-related diseases including cardiovascular disease, breast cancer and osteoporosis¹. Age at menopause averages around 50-51 years and ranges between 40 and 60 years of age²; twin studies have shown this variability to be genetically determined with heritabilities of 44-65%^3-5. Such genetic factors might regulate the size of the follicle pool and the rate of its depletion, and their identification could have biological and clinical applications.

Typical for complex quantitative traits, genome-wide linkage studies of menopause have been unsuccessful, and candidate gene studies have mainly focused on the estrogen pathway⁶ and have had conflicting results⁷. This suggests that the apparent effect sizes for genetic variants are small and that the major causative loci have not been identified. Genome-wide association studies (GWAS) have proven successful in identifying common susceptibility genes with small effect sizes for many complex diseases and traits⁸ and might be suitable to identify genetic factors involved in determining age at menopause.

In this GWAS we used a two-stage design to identify previously unknown loci influencing age at menopause. We included women with self-reported natural age at menopause (defined as 12 months without regular periods) between 40 and 60 years, excluding those with hysterectomy, uni- or bilateral ovariectomy, menopause induced by irradiation or occurring after stopping the contraceptive pill, or those currently using hormone replacement therapy.

In stage 1 we genotyped 2,368 women of the Rotterdam Study baseline⁹ with the Illumina HumanHap 550v3 Beadarray. After quality control, 535,354 SNPs were left for analysis. Allelic association tests were carried out using PLINKv1.01 software¹⁰ for age at natural menopause. The genomic inflation factor (λ) was 1.01669 for this analysis, indicating no population stratification, so we based our results on the uncorrected P values. The strongest association signals were found for rs2151145 (P = 5.3 × 10⁻⁶) on chromosome 9, rs236114 (P = 5.6 × 10⁻⁶) on chromosome 20 and rs1172822 (P = 6.3 × 10⁻⁶) on chromosome 19 (Supplementary Table 1 and Supplementary Fig. 1a online).

We combined the results from the Rotterdam Study baseline with GWA data from the TwinsUK study. A total of 611 women with natural menopause using the same definitions and exclusions as above were genotyped with the Illumina HumanHap 300K beadarray, and after quality control 317,818 SNPs were left for analysis. After adjusting for relatedness and genomic control, we did not observe any genome-wide significant signals in this study (Supplementary Table 1 and Supplementary Fig. 1b). Because of the different study designs we conducted meta-analysis on summary statistics of the two studies using METAL (http://www.sph.umich.edu/csg/abecasis/metal/; Supplementary Methods online) on 315,418 SNPs common to both cohorts (2,979 women), but we did not observe any genome-wide significant SNPs (Supplementary Fig. 1c).

From this meta-analysis, all SNPs with P < 1 × 10⁻⁴, corresponding to 32 SNPs from 24 loci (with five loci having multiple significant SNPs), were followed up in stage 2 (Supplementary Table 1). Twenty-four SNPs were genotyped using Sequenom iPLEX genotyping and seven SNPs using Taqman allelic discrimination (Applied Biosystems) (Supplementary Methods) in 2,560 samples of four additional cohorts of postmenopausal females of European ancestry (Supplementary Methods and Supplementary Table 2 online); one of the SNPs (rs11786333) failed genotyping. For the remaining 31 SNPs, we calculated combined P values, betas and standard errors using inverse variance fixed-effects meta-analysis (Supplementary Table 1) and identified six common SNPs that were genome-wide significant in the combined stage 1 and 2 analysis (Table 1). Four SNPs on chromosome 19 were significant: rs1172822 (MAF = 0.39), P = 6.28 × 10⁻¹¹, beta = −0.391 year per T allele (s.e.m. = 0.0598); rs2384687 (MAF = 0.40), P = 1.39 × 10⁻¹⁰, beta = 0.381 year per C allele (s.e.m. = 0.0594); rs1551562 (MAF = 0.25), P = 1.04 × 10⁻⁹, beta = 0.4279 year per G allele (s.e.m. = 0.0701); and rs897798 (MAF = 0.48), P = 3.91 × 10⁻⁸, beta = 0.308 year per G allele (s.e.m. = 0.056). These four SNPs are likely to report the same signal because the linkage is high (D’ > 0.92, r² > 0.5, Supplementary Fig. 2b online).

Table 1. The six genome-wide significant SNPs and association with age at menopause.

					Stage 1		Overall meta-analysis
SNP	Chr	Position	Minor allele	MAF	P value	P value	Effect size	s.e.
rs1172822	19q13.4	60511657	T	0.39	1.94E-07	6.28E-11	0.391	0.060
rs236114	20p12.3	5883385	A	0.21	1.83E-05	9.71E-11	0.495	0.077
rs2384687	19q13.4	60523000	C	0.40	1.57E-06	1.39E-10	−0.381	0.059
rs1551562	19q13.4	60506693	G	0.25	1.02E-06	1.04E-09	−0.428	0.070
rs7333181	13q34	111019298	A	0.12	5.26E-05	2.50E-08	0.520	0.093
rs897798	19q13.4	60525566	G	0.48	7.13E-06	3.91E-08	−0.308	0.056

On chromosome 20, rs236114 (MAF = 0.21) was genome-wide significantly associated with age at natural menopause (P = 9.71 × 10⁻¹¹, beta = 0.4953 year per A allele (s.e.m. = 0.0765)). Furthermore, on chromosome 13 rs7333181 (MAF = 0.12) was genome-wide significant: P = 2.50 × 10-8, beta = 0.5201 (s.e.m. = 0.0933). The six genome-wide-significant hits showed no heterogeneity (I² < 25%), so fixed effects models were used.

In addition, we estimated the risk for menopause before the age of 50 by allele of the six genome-wide significant SNPs (Fig. 1). We conducted fixed-effects meta-analysis for SNPs not showing heterogeneity (rs7333181, rs1551562, rs1172822, rs2384687, rs897798), and random effects meta-analysis for rs236114, for which I² was 31%. This meta-analysis showed that the A allele of rs1172822 is associated with a 19% increased risk for natural menopause before 50 years (OR = 1.19, 95% CI = 1.09-1.29, P = 6.2 × 10⁻⁵). The other SNPs on chromosome 13, 19 and 20 showed a similar increase or decrease in risk.

Figure 1.

Meta-analysis of risk for early menopause (<50 years) by genotype for the six genome-wide-significant hits. (a) rs7333181 on chromosome 13. (b) rs1551562 on chromosome 19. (c) rs1172822 on chromosome 19. (d) rs2384687 on chromosome 19. (e) rs897798 on chromosome 19. (f) rs236114 on chromosome 20.

The initial analysis was not adjusted for covariates such as age, body mass index, smoking, age at menarche, parity and use of oral contraceptives and female hormones. To rule out an effect of these covariates on the association of the genome-wide-significant hits, we carried out adjusted linear regression of these SNPs in the Rotterdam Study baseline cohort (Supplementary Table 3 online). None of the previously found associations was affected by the adjustment for these covariates, indicating that the effect of the SNP occurs directly on age at natural menopause and not via one of the covariates. We calculated the total explained variance in age at natural menopause for these SNPs in the combined replication studies to be 1.1% (range 0.1-0.5% per SNP).

We then conducted fine mapping of these signals using meta-analysis of imputed data of the stage 1 studies, and found three SNPs on chromosome 13 with more or equal significance as rs7333181, two SNPs on chromosome 19 and one on chromosome 20 with higher significance compared to the previously reported SNPs (Supplementary Table 4 and Supplementary Fig. 2 online). For all three loci, the imputed SNPs are located in the same linkage disequilibrium (LD) block as the genome-wide-significant SNPs.

The four chromosome 19 SNPs are located within an LD block covering almost 20 kb and are located intronic and 3′ of the BRSK1 gene (BR serine threonine kinase 1), in the 3′ region and inside the TMEM224 gene, and 5′ of the SUV420H2 gene (suppressor of variegation 4-20 homolog 2, a lysine methyltransferase) (Supplementary Fig. 2b). Literature analysis for these genes did not indicate an immediate functional explanation for the observed association, although for all three genes a possible involvement in ovarian aging was suggested (Supplementary Note online).

rs7333181 on chromosome 13 is located >250 kb 3′ of the hypothetical gene LOC121793 and the ARHGEF7 (rho guanine nucleotide exchange factor 7) gene, also known as COOL1 (cloned out of library 1). ARHGEF7 has a role in cell proliferation through phosphorylation of FOXO3a. FOXO3a knockout mice are infertile due to early depletion of the follicle pool, indicating a possible role of this gene in menopause (Supplementary Note). The chromosome 20 SNP is located in an intron of the MCM8 (minichromosome maintenance complex component 8) gene. The more significantly associated SNP from the imputed data is a nonsynonymous SNP in exon 9 of this gene (E341K), and could influence the protein structure or function of MCM8. For the gene on chromosome 20 no involvement in ovarian aging or menopause was suggested.

Identification of the causative variant(s) and the responsible gene(s) underlying the observed associations requires further research, which will enhance our molecular understanding of the genetic regulation of the ovarian reserve and aging process. Although the rate of ovarian aging is highly variable among women, identification of women with decreased ovarian reserve is clinically relevant, as timing of menopause is an important risk factor, for example, for breast cancer, cardiovascular disease and osteoporosis.