Interaction between cigarette smoking and genetic polymorphisms on the associations with age of natural menopause and reproductive lifespan: the Singapore Chinese Health Study

Abstract

STUDY QUESTION

Are there genetic variants that interact with smoking to reduce reproductive lifespan in East-Asian women?

SUMMARY ANSWER

Our study corroborates several recently identified genetic loci associated with reproductive lifespan and highlights specific genetic predispositions that may interact with smoking status to adversely affect reproductive lifespan in East-Asian women.

WHAT IS KNOWN ALREADY

Epidemiological data as well as evaluations on genetic predisposition to smoke indicate on the importance of smoking in adverse effects on reproductive lifespan in women. However, there are no previous smoking and gene interaction studies for reproductive traits in East-Asian women.

STUDY DESIGN, SIZE, DURATION

This population-based prospective cohort study comprised 11 643 East-Asian Chinese women with overlapping genome-wide genotyping and reproductive data.

PARTICIPANTS/MATERIALS, SETTING, METHODS

We performed a genome-wide association study for reproductive lifespan in women (n = 11 643) from the Singapore Chinese Health Study (SCHS) and carried out a genome-wide interaction study to identify loci that interacted with smoking status to affect age of natural menopause and reproductive-time.

MAIN RESULTS AND THE ROLE OF CHANCE

Two known loci associated with menopause, rs113430717 (near HMCES, chromosome 3, P_meta = 5.72 × 10⁻¹⁵) and rs3020136 (near RAD21, chromosome 8, P_meta = 1.38 × 10⁻⁸) were observed beyond genome-wide levels of association with age at menopause in this study. For reproductive lifespan, the genome-wide association observed at rs79784106 (chromosome 3, P_meta = 5.05 × 10⁻¹²) was in linkage disequilibrium with the menopause lead single-nucleotide polymorphism (SNP) (rs113430717). Four additional loci, first reported to be associated with menopause, were also associated with reproductive lifespan in our study (P_Adj between 7.42 × 10⁻⁵ to 4.51 × 10⁻³). A significant interaction was observed between smoking and an East-Asian specific SNP, rs140146885, for reduced reproductive lifespan, per copy of the minor C allele (beta = −1.417 years, P_interaction = 2.31 × 10⁻¹⁰). This interaction was successfully replicated in additional independent samples (beta = −1.389 years, P_interaction = 6.78 × 10⁻³). Another known variant associated with menopause, rs11031006 (near FSHB), was also observed to interact with smoking status to reduce age at menopause in our dataset (beta = −0.450 years, P_adj = 0.042).

LIMITATIONS, REASONS FOR CAUTION

The modest sample size of the replication datasets used likely affected the statistical power to firmly replicate all identified novel loci observed in our smoking interaction analyses.

WIDER IMPLICATIONS OF THE FINDINGS

Age of natural menopause and reproductive lifespan have clear genetic predispositions with distinct ethnic differences, and they may be adversely truncated by lifestyle factors such as smoking, which can pose a significant impact on the reproductive lifespan and future health outcomes in women.

STUDY FUNDING/COMPETING INTEREST(S)

The Singapore Chinese Health Study is funded by the National Medical Research Council, Singapore (NMRC/CIRG/1456/2016), National Institutes of Health (R01 CA144034 and UM1 CA182876) and National Research Foundation, Singapore (Project Number 370062002). W.-P.K. is supported by the National Medical Research Council, Singapore (MOH-CSASI19nov-0001). The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. The authors do not report conflicts of interest.

TRIAL REGISTRATION NUMBER

N/A.

Introduction

Compared to their Western counterparts, data on biomarkers and determinants for reproductive longevity in Asian women are limited and biomarkers in current clinical utility to determine ovarian reserves are limited to measurements of anti-Mullerian hormone (Dewailly et al., 2014) and FSH during Days 2 to 3 of a woman’s menstrual cycle (Loverro et al., 2003). Existing evidence suggests that there are distinct ethnic differences in ovarian reserves (Tal and Seifer, 2013; Bleil et al., 2014) and there is a possibility that reproductive outcomes are poorer in Asian women compared to Caucasian women, even with the aid of reproductive techniques (Gleicher et al., 2007).

Ovarian reserve and its utilization, over a reproductive lifespan, are determined by genetic (Horikoshi et al., 2018; McGrath et al., 2021; Ruth et al., 2021), epigenetic (Levine et al., 2016) and environmental factors (Persani et al., 2010; Hewlett and Mahalingaiah, 2015). Studies have identified genetic loci associated with premature ovarian insufficiency in women who have spontaneous menopause before 40 years old compared to women with normal age of menopause (Ayabe et al., 2011; Qin et al., 2014; Kim et al., 2015). Additionally, studies have demonstrated that age of menopause and duration of natural oestrogenic exposure in women could be associated not only with reproductive outcomes but also with health outcomes, e.g. maintenance of bone mineral density, reduction of fracture risks, reduction of cardiovascular morbidities and type 2 diabetes (Wu et al., 2014; Shadyab, 2017; Song et al., 2020; Mishra et al., 2021; Ruth et al., 2021). In large population studies, although low levels of anti-Mullerian hormone were associated with low ovarian reserve, earlier age at menopause and reduced fecundity, it has not been useful as a predictor of reproductive potential (Hawkins Bressler and Steiner, 2018) or the health-span of women (Ruth et al., 2019). Hence, it is imperative to search for associated biomarkers on Asian women’s ovarian longevity, as determined by her duration of natural oestrogen exposure or reproductive lifespan (i.e. the duration from age of menarche to the age of spontaneous menopause). The use of these biomarkers as predictors of reproductive lifespan and their effect on health-spans may be applied in clinical practice in guiding clinicians to advise and manage women’s reproductive health.

Importantly, lifestyle and environmental factors, such as alcohol consumption (Hawkins Bressler et al., 2016; Ruth et al., 2021), have significant effects on reproductive lifespan and age at natural menopause (Appiah et al., 2021; Hyvärinen et al., 2021). Smoking has been associated with an earlier age of menopause (Oboni et al., 2016; Ruth et al., 2021) and a trend towards reduced ovarian reserves in women of reproductive age group (Waylen et al., 2010). However, these studies did not have a significant number of Asian women in the cohorts. Herein, we examined genetic data in women (n = 11 643) from the Singapore Chinese Health Study (SCHS), replicating known menopause and menarche variants from other Asian and European ancestry datasets, and further identifying genetic loci that interact with smoking status to affect age at natural menopause and reproductive lifespan.

Materials and methods

Study population

The SCHS is a population-based prospective cohort comprising of a total of 63 257 Chinese participants (35 303 women and 27 954 men) aged 45 to 74 years and living in Singapore during recruitment between 1993 and 1998. Trained interviewers conducted face-to-face interviews using a structured questionnaire at baseline on demographics, reproductive information (only for women), height, weight, dietary intake, cigarette smoking, alcohol consumption, medical history and family history of cancer. From April 1994, a 3% random sample of study subjects was re-contacted for donation of blood specimens and the effort was later extended from 2000 to 2004 to about half of the cohort participants, who consented to giving blood or buccal specimens for analysis. Surviving and consenting participants were interviewed in two follow-up visits (1999–2004 and 2006–2010) to update reproductive information (only for women) and selected lifestyle factors and medical history. The study was approved by the Institutional Review Board at the National University of Singapore, and written informed consent was obtained from all study participants.

In the current analyses, we only included women (n = 14 002) who had reported information on menopause either during baseline or one of the follow-up visits, and who had given blood samples for genome-wide association studies (GWAS). Among the 14 002 women, we further excluded 353 women who started hormone replacement therapy before menopause (and thus could not report age of natural menopause), 1813 women who had surgical menopause (hysterectomies and/or bilateral salpingo-oophorectomies), 31 women who had had radiotherapy to their ovaries, 20 women were on existing medications that might affect ovarian function, such as chemotherapy and 139 women with missing data. Thus, 11 643 women with natural age of menopause and genotype data were included in the current analyses (Fig. 1).

Figure 1.

Number of women with natural menopause included in the Singapore Chinese Health Study (SCHS) with complete reproductive data and genotyping data from three independent genome-wide association study (GWAS) datasets. The larger SCHS Discovery dataset was used as the primary discovery dataset in the study and the SCHS Replication and SCHS coronary artery diseases (CAD) datasets were used to replicate identified top findings from discovery.

Details of reproductive information and cigarette smoking

Reproductive information was obtained at the baseline interview regarding the age at menarche, age at first livebirth, number of biological children, use of oral contraceptive and duration, age at menopause, type of menopause (natural, surgery, radiotherapy or medication-induced) and use of hormonal replacement therapy. The information on menopausal status, type of and age at menopause, and use of hormonal replacement therapy were updated at both follow-up visits. Reproductive lifespan was computed as the duration from age at menarche to age of menopause. Data on cigarette smoking status (never smokers, former smokers and current smokers) were obtained from interviews at baseline and the first follow-up. Ever smoking included current or former smokers, while never smokers were those who indicated that they did not smoke at both baseline and the first follow-up interviews.

Genotyping and imputation

A total of 27 308 SCHS participants were genotyped with the Illumina Global Screening Array (GSA). Among them, 25 273 SCHS participants were genotyped from 2017 to 2018 using the Illumina GSA v1.0 and v2.0 (SCHS Discovery), and the remaining 2035 SCHS participants were genotyped in the year 2020 using the Illumina GSA v2.0 (SCHS Replication). Another separate subset of 2161 participants were included in a case-control study of coronary artery disease (SCHS CAD) nested in this SCHS cohort, and these were genotyped on the Illumina HumanOmni ZhongHua-8 Bead Chip (SCHS CAD) in year 2012. Detailed information on quality control (QC) procedures is provided in Supplementary Table SI and has been previously described (Chang et al., 2021a,b).

Imputation for additional autosomal single-nucleotide polymorphisms (SNPs) was performed with IMPUTE2 (Marchini et al., 2007) and genotype calls were based on phase 3 1000G cosmopolitan panels. SNPs with imputed information score < 0.8, minor allele frequency (MAF) < 1.0%, Hardy Weinberg Equilibrum (HWE) P < 1 × 10⁻⁶ as well as non-biallelic SNPs were excluded from subsequent analyses.

Statistical analysis

A total of 11 643 women achieved natural menopause by the second follow-up visit and had provided reproductive information and QCed genotype data (Supplementary Table SI) for this study. The differences in the factors of participants from the SCHS Discovery, SCHS Replication and SCHS CAD dataset were compared using Chi-squared test for categorical variables and Student’s t-test for continuous variables. The age at menarche, age at menopause and reproductive lifespan were normally distributed in all the three datasets used in this study. We used linear regression GWAS analysis to identify genetic associations for reproductive lifespan, age at menopause and age at menarche in the three SCHS datasets separately, using SNPTEST (version 2), with adjustment for first the three principal components of overall study variance. Subsequently, we used the META (v1.5) toolset to meta-analyse results from the individual datasets using the fixed effect inverse-variance weighted method. Top hits that were identified were additionally evaluated by normalizing outcome variables using z-score transformation and with the exclusion of extreme outliers (> or <3 SDs). To replicate previously reported genetic variants for age at menopause and age at menarche, we extracted all reported variants from three recent large-scale GWAS studies (Day et al., 2017; Horikoshi et al., 2018; Ruth et al., 2021). A total of 190 and 368 known genetic variants for age at menopause and age at menarche, respectively, passed GWAS and imputation QC procedures in our study data and were evaluated for these traits. A total of 229 known variants that did not pass GWAS QC procedures and/or were rare or monomorphic in our study were excluded from the analysis (see details in Supplementary Table SII).

To evaluate genome-wide interactions with smoking status (never versus ever), we performed a two-degree-of-freedom test that jointly tested the genetic main effect and the SNP–smoking interaction in the same regression model (Sung et al., 2019): E[Y] ∼ β₀ + β_GG + β_ESMK + β_SEGxSMK + β_CC. Y was either menopausal age or reproductive time, G was the genetic dosage of imputed variants, SMK was the ever smoking variable coded 0 for never smokers and 1 for current and former smokers combined, and C was the vector of covariate that included the first three principal components. These analyses were performed in the SCHS Discovery dataset using ProbABEL (Aulchenko et al., 2010), with similar phase3 1000G cosmopolitan panel imputed data from Minimac4 and with exclusion of SNPs with impute rsq < 0.5 using the Michigan imputation server. Top interaction findings from the study were further replicated in the SCHS Replication and SCHS CAD datasets.

Power calculations were performed in Quanto (ver1.2.4) by assuming an additive genetic effect model. The overall power in the study to detect variants associated with age at menopause and age at menarche was calculated with effect estimates of 0.13 years and 0.04 years, which corresponded to the average effects observed from previously reported GWAS for these traits (Day et al., 2017; Horikoshi et al., 2018; Ruth et al., 2021). Specific power in the study to identify each previously reported hit for menopause and menarche was calculated with reported effect estimates of these hits and MAFs from the SCHS study (α = 0.05).

Identified lead SNPs in the study were functionally annotated using the SNP2GENE function in Functional Mapping and Annotation (FUMA) (Watanabe et al., 2017). All SNPs in linkage disequilibrium (LD) (r² > 0.6 in 1000G ASN panel) with lead signals were identified and potential functionality was evaluated through CADD and other bioinformatics toolsets (i.e. SIFT and Polyphen2).

Results

Demographics of subjects

Mean ages at recruitment were 54.8 ± 7.5 years in the SCHS Discovery dataset (n = 10 041) and 55.1 ± 7.4 years SCHS Replication dataset (n = 977), which were both lower than the recruitment age for the women in the SCHS CAD dataset (n = 625) at 61.3 ± 7.5 years (Table I). In the SCHS CAD dataset, it was also noted that the proportion of ever-smokers dataset was significantly higher than the discovery and replication datasets (P < 0.0001). These results are not unexpected since the women in the CAD dataset were selected as cases and controls for a study on CAD. The mean BMIs of the women in the three datasets at recruitment were comparable, ranging from 22.9 ± 3.3 to 23.2 ± 3.3. Age at menarche, age at menopause and reproductive lifespan were similar across the three datasets (Table I).

Table I.

Characteristics of participants in the discovery and replication datasets in the Singapore Chinese Health Study (SCHS).

	SCHS Discovery dataset	SCHS Replication dataset (Replication 1)	SCHS CAD dataset (Replication 2)
N	10 041	977	625
Age*	54.839 (7.477)	55.118 (7.445)	61.331 (7.540)
BMI (kg/m2)	22.923 (3.261)	23.193 (3.310)	22.987 (3.071)
Age at menarche	14.030 (1.896)	14.042 (1.851)	14.531 (1.973)
Age at natural menopause	50.358 (3.761)	50.195 (3.899)	49.960 (4.066)
Reproductive lifespan	36.327 (4.155)	36.153 (4.260)	35.425 (4.459)
% Never smokers*	90.02%	90.07%	81.28%
% Current smokers*	5.03%	5.02%	8.48%
% Former smokers*	4.95%	4.91%	10.24%
% Ever smokers*	9.98%	9.93%	18.72%

^* Statistically significant P < 0.001.

CAD, coronary artery diseases.

Genome-wide associations for reproductive time, age at menopause and menarche

The top genome-wide SNP associations identified with reproductive lifespan and age at menopause are presented in Table II and Supplementary Fig. S1. (Supplementary Tables SIII and SIV provide details of all genome-wide hits identified for reproductive lifespan and age at menopause. Please see the data availability section for full meta-analysis results for all SNPs with reproductive lifespan, age at menopause and age at menarche.) Rs79784106 on chromosome 3 was observed as the top hit associated with reproductive lifespan. The G allele of rs79784106 was significantly associated with reduced reproductive lifespan (beta = −0.609 years, SE = 0.088, P_meta = 5.05 × 10⁻¹², Table II). After normalizing the reproductive lifespan variable, excluding extremes in phenotypic distribution, and including additional adjustments for participant’s baseline age, the observed genome-wide association between rs79784106 and reproductive lifespan remained materially the same (Supplementary Table SV). Rs79784106 was in perfect LD (r² = 1, 1000G CHS) with a missense variant at HMCES (rs60527165) with high Combined Annotation Dependent Depletion (CADD) score (22.7, top 1% of deleterious variants in the human genome) that is common in East-Asians (∼9.4%) but rare in European populations (0.5%, Supplementary Table SVI).

Table II.

Lead Single Nucleotide Polymorphisms associated with reproductive lifespan and age at menopause associated beyond genome-wide levels of significance in the Singapore Chinese Health Study (SCHS) study.

				Meta-analysis (N = 11 643)			SCHS Discovery (N = 10 041)		SCHS Replication (N = 977)		SCHS CAD (N = 625)
SNP	Location	TA	TAF	P	Beta	P het	P	Beta	P	Beta	P	Beta
Reproductive lifespan
rs79784106	3q21.3	A	0.121	5.05 × 10−12	−0.609	0.260	2.69 × 10−13	−0.585	0.122	−0.539	4.47 × 10−3	−1.402
Age at menopause
rs113430717	3q21.3	A	0.118	5.72 × 10−15	−0.629	0.542	1.70 × 10−13	−0.633	0.097	−0.445	0.023	−1.018
rs3020136	8q24.11	G	0.155	1.38 × 10−8	−0.400	0.956	1.84 × 10−7	−0.392	0.053	−0.461	0.240	−0.439

CAD, coronary artery diseases; P_het, Cochran’s Q heterogeneity P; TA, test allele; TAF, average test allele frequency.

Two previously reported loci for age at menopause were observed at genome-wide levels of association in our study (Table II, Supplementary Fig. S1). This included the same locus on chromosome 3 for reproductive lifespan that was also significantly associated with decreased menopausal age (lead SNP rs113430717, beta= −0.629 years, SE = 0.081, P_meta = 5.72 × 10⁻¹⁵, Table II). This locus was previously reported to be associated with age at menopause in a Japanese population (Horikoshi et al., 2018) and the reported lead SNP, rs4853, was in LD with our top signal, rs79784106 (r² = 0.891, 1000G CHS). The other hit, rs3020136, on chromosome 8 (beta = −0.399 years, SE = 0.070, P_meta = 1.38 × 10⁻⁸, Table II), was also in complete LD with the reported lead SNP, rs2921759 (near RAD21), in the Japanese study (r² = 1.000, 1000G CHS) (Horikoshi et al., 2018). Statistical power to identify significant associations for menarche was limited in our study and we did not observe any robust genome-wide associations with age at menarche (Supplementary Figs S1 and S2).

Replication of reported loci associated with age at menopause and menarche

We sought to replicate genetic variants recently reported in large-scale GWAS for age at menopause and menarche (Day et al., 2017; Horikoshi et al., 2018; Ruth et al., 2021) in our study population. Out of 190, 165 (Binomial P < 0.000001) previously reported menopause loci, and out of 368, 280 (binomial P < 0.000001) known menarche loci were directionally consistent in our study. There were 54 previously reported loci for age at menopause and 55 previously reported loci for age at menarche that were directionally consistent and nominally associated with these traits in our study (menopause P_meta between 7.64 × 10⁻⁷ to 4.70 × 10⁻² and menarche P_meta between 3.12 × 10⁻⁴ to 4.98 × 10⁻², Supplementary Tables SVII and SVIII).

We further evaluated whether these reported menopause and menarche loci were also associated with reproductive lifespan in our study. Besides the known menopause loci at chromosome 8 (rs2921759), four other SNPs showed statistically significant associations with reproductive lifespan (P_adj between 7.42 × 10⁻⁵ to 4.51 × 10⁻³, Supplementary Table SIX). For all these SNPs, the alleles with younger age at menopause were also associated with shorter duration of reproductive lifespan.

Genome-wide ever smoking interactions

Six genome-wide interactions between SNPs and ever smoking were identified in the SCHS Discovery dataset for the associations with reproductive lifespan and age at menopause (P_interaction ranging between 1.59 × 10⁻¹¹ and 1.68 × 10⁻⁸, Table III and Supplementary Tables SX and SXI. Please see Data availability section for full interaction analyses of all SNPs for both reproductive lifespan and age at menopause). The interaction between smoking and an intergenic SNP, rs140146885, on chromosome 20 for the association with reproductive lifespan was significantly replicated in the additional SCHS datasets (beta = −1.389 years, P_adj = 0.027) which improved the overall level of association (P_meta = 5.47 × 10⁻¹²) (Table III). Additional sensitivity analysis to exclude effects of extremes in phenotype distribution showed similar effects of this rs140146885-ever smoking interaction on reproductive lifespan (Supplementary Table SXII and Supplementary Fig. S3). The C allele of rs140146885 showed a trend towards decreasing reproductive lifespan in our study (beta =−0.339 years, P_meta = 0.148) and was observed to be an East-Asian specific variant (C allele frequency between 1.00% and 2.00%, EAS 1000G panel) and was monomorphic in other reference populations. All other interactions identified did not show significant associations in the replication datasets, thus weakening the overall association analysis in the meta-analysis and raising the possibility that these may be false positive findings.

Table III.

Lead Single Nucleotide Polymorphism x Ever smoking interactions detected beyond genome-wide levels in the Singapore Chinese Health Study (SCHS) Discovery dataset for reproductive lifespan and age at menopause and replications in additional SCHS datasets.

				Meta-analysis (N = 11 643)			SCHS Discovery (N = 10 041)		SCHS Replication (N = 977)		SCHS CAD (N = 625)		Replication meta-analysis (N = 1602)
SNP	Location	TA	TAF	Beta	P	P het	Beta	P	Beta	P	Beta	P	Beta	P	P Adj	P het
Interactions with reproductive lifespan
rs140146885	20q12	C	0.010	−1.412	5.47 × 10−12	0.432	−1.417	2.31 × 10−10	−0.254	0.802	−1.777	0.003	−1.389	6.78 × 10−03	0.0271	0.195
rs115185733	8q21.11	G	0.013	−1.294	6.50 × 10−9	0.584	−1.338	1.07 × 10−08	−1.440	0.157	−0.205	0.849	−0.860	0.244	1	0.404
rs17106789	10q23.32	A	0.011	0.903	3.04 × 10−10	0.058	0.982	2.70 × 10−11	−0.955	0.343	−0.189	0.806	−0.473	0.44	1	0.546
rs189359341	3p14.1	G	0.012	−0.836	1.21 × 10−7	0.111	−0.944	1.68 × 10−8	0.394	0.554	−0.32	0.644	0.051	0.915	1	0.457
Interactions with age at menopause
rs79002722	20q13.33	G	0.010	−1.44	1.07 × 10−8	0.005	−1.808	6.14 × 10−11	0.347	0.635	0.392	0.725	0.361	0.555	1	0.973
rs187508025	4p15.2	A	0.011	−1.03	4.04 × 10−11	0.363	−1.074	1.59 × 10−11	0.267	0.793	−0.4	0.72	−0.035	0.962	1	0.658

CAD, coronary artery diseases; P_Adj, Bonferroni adjusted P; P_het, Cochran’s Q heterogeneity P; TA, test allele; TAF, average test allele frequency.

We further evaluated for potential interactions with smoking for all the loci previously reported to be associated with menopause (Horikoshi et al., 2018; Ruth et al., 2021). Interaction between rs11031006 (near FSHB) and smoking status was statistically significant for age at menopause in our dataset (beta = −0.450 years, P_adj = 0.042, Supplementary Table SXIII). The G allele of rs11031006 that has been reported to reduce age at menopause (Ruth et al., 2021) was also observed to interact with smoking to reduce age at menopause in our study. Eight additional interactions between smoking and known loci for menopause were also detected, although they were at more nominal levels of significance (P_interaction = 6.17 × 10⁻³ and 0.044, Supplementary Table SXIII).

Discussion

In our study, we observed and validated East-Asian specific genome-wide signals at chromosomes 3 and 8 for age at natural menopause and showed that the signal at chromosome 3 was also associated with reproductive lifespan. We further replicated several other loci at more nominal levels for both age at natural menopause and menarche in our study, corroborating recent findings from Ruth et al. (2021), Day et al. (2017) and Horikoshi et al. (2018). We showed that the rs11031006 SNP (FSHB), previously associated with menopause (Corbo et al., 2011; Ruth et al., 2021) might interact with cigarette smoking to reduce age at natural menopause. Moreover, we identified and replicated an additional novel ever-smoking interaction with an East-Asian specific SNP on chromosome 20 for reproductive lifespan effects (i.e. reduction of reproductive lifespan) in our study participants.

Differences in allele frequencies and effect estimates on reproductive traits of associated variants among the Japanese and European populations have been previously highlighted (Horikoshi et al., 2018). Our study corroborates the ethnic specific nature of some of these genetic loci. Besides the two East-Asian specific genome-wide hits observed at chromosome 3 and 8, we have further validated an independent East-Asian signal at the MCM8 gene (rs76498344) for association with age of menopause in our study. Notably, this has not been observed in previous European studies and is a separate signal from the other MCM8 signal that was first identified in Europeans (rs236117). The latter was also nominally associated in our study and previously replicated in other East-Asians datasets in the Ruth et al. (2021) study. This suggests the possibility of two separate causal mutations at the MCM8 locus among East-Asians.

Additionally, for two other menopause-associated loci first identified in the Japanese (rs10049761 near ZAR1 and rs8010674 near DCAF4), we also reported similar effect estimates that were 1.5- to 2.5-fold higher than the estimates derived from European populations (Horikoshi et al., 2018). At the same time, at rs6185 (near GNRH1) we did not detect a significant association for age of menopause in our study (Supplementary Fig. S4), despite sufficient statistical power to do so. The direction of the menopause effect at rs6185 in our study was however consistent with that of the Japanese finding. Potential differences of risk allele frequency (G allele frequency 51% in Japanese versus 58% in Singapore Chinese) and/or gene-environmental differences between the Japanese and Singapore Chinese East-Asian populations may have contributed to such discrepancies.

The duration of the reproductive lifespan of a woman, other than the age of menopause, may determine her lifetime exposure to endogenous oestrogen, and may also determine a woman’s health outcomes in her post-reproductive years (Wu et al., 2014; Shadyab, 2017; Song et al., 2020; Mishra et al., 2021; Ruth et al., 2021). Interestingly, the potential biological functions of the two robustly associated East-Asian specific menopause and reproductive lifespan gene loci, HMCES and RAD21, may be associated with DNA double stranded break repair mechanisms secondary to oxidative damage (Srivastava et al., 2020) and cohesion complex involved in chromosome segregation (Cheng et al., 2020). These biological processes have been linked to ovarian ageing (Meldrum et al., 2016) which would culminate in earlier menopause and directly impact the duration of reproductive time.

Cigarette smoking has been associated with earlier age at natural menopause (Oboni et al., 2016; Ruth et al., 2021), and longer duration, higher cumulative dose and earlier initiation of smoking have been associated with earlier menopause in both current and former smokers (Zhu et al., 2018). These observations are likely to be mediated by an impairment of antral follicle development and growth due to toxic substances in cigarette smoke, the latter resulting in oxidative stress and DNA damage on the supportive granulosa cells (de Angelis et al., 2020) and apoptosis of these ovarian follicles. Furthermore, cigarette smoke has been shown to induce DNA damage directly in granulosa cells (Sinko et al., 2005) and in cumulus cells of smokers (Budani et al., 2017). As these cells support the growth and development of the ovarian follicle and oocyte, their demise would hasten the depletion of ovarian follicles essential for reproductive lifespan. Accumulation of DNA damage has been demonstrated to exhibit many direct and indirect sequelae that could lead to ovarian ageing, such as mitochondrial alterations, premature cell death and cellular senescence, leading to exhaustion of cell renewal capacity and dysfunction in the ovaries, and thereby providing a possible link to gene clusters identified by the GWAS on age of natural menopause (Laven et al., 2016). Our finding of a smoking interaction with an Asian-specific SNP might suggest that ethnic-specific gene-environment interaction might result in Asian women being more susceptible to the adverse effects of smoking to their reproductive lifespans.

The strengths of this study include its population-based prospective design with detailed reproductive and lifestyle (i.e. smoking exposure) data, the homogeneous population and the high quality of genome-wide genotyping. However, the limitations of the study need to be addressed. The modest sample size of the replication datasets used likely affected statistical power to firmly replicate all identified novel loci observed in our smoking interaction analyses. Nevertheless, the locus at chromosome 20 showed directionally consistent effects in both replication datasets and the overall significance level was improved after meta-analysis of discovery and replication datasets. We note that the proportion of women who were ever smokers was significantly higher in the SCHS CAD replication dataset as compared to the other datasets utilized and this may have improved the statistical power to detect significant smoking interactions in this dataset, although the reproductive traits examined (age of menarche, age of natural menopause and reproductive lifespan) were largely comparable across the discovery and replication datasets. Additionally, the replication cohorts were from the same study population and would be subject to the same underlying biases as the discovery cohort.

To conclude, we corroborated several known loci associated with age of menopause and menarche and observed at least five menopause loci that may also be associated with reproductive lifespan. Furthermore, these reproductive traits have clear genetic predispositions with distinct ethnic differences, and they may be adversely truncated by lifestyle factors such as smoking, which can have significant impacts on the reproductive lifespan and future health outcomes of women.

Supplementary data

Supplementary data are available at Human Reproduction online.

Data availability

Full meta-analysis data for all SNPs evaluated for reproductive time are available in https://doi.org/10.6084/m9.figshare.16767733. Full meta-analysis data for all SNPs evaluated for age at menopause are available in https://doi.org/10.6084/m9.figshare.16767721. Full meta-analysis data for all SNPs evaluated for age at menarche are available in https://doi.org/10.6084/m9.figshare.16767205. Full meta-analysis data for all SNPxEverSmoking interactions for reproductive time are available in https://doi.org/10.6084/m9.figshare.16767754. Full meta-analysis data for all SNPxEverSmoking interactions for age at menopause are available in https://doi.org/10.6084/m9.figshare.16767739.