Joint Genome-Wide Association Study Identifies Twenty-One Novel Loci for Age at Menarche and Highlights Its Causal Association with Other Complex Diseases

doi:10.21203/rs.3.rs-955340/v1

Download PDF

Research article

Joint Genome-Wide Association Study Identifies Twenty-One Novel Loci for Age at Menarche and Highlights Its Causal Association with Other Complex Diseases

https://doi.org/10.21203/rs.3.rs-955340/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Age at menarche (AAM) is a sign of puberty of females. It is a heritable trait associated with various adult diseases. However, the genetic mechanism that determines AAM and links it to disease risk is poorly understood. Aiming to uncover the genetic basis for AAM, we conducted a joint association study in up to 438,089 participants from 3 genome-wide association studies of European and East Asian ancestries. Twenty-one novel genomic loci were identified at the genome-wide significance level. Besides, we observed significant genetic correlations between AAM and 67 complex traits, and the highest genetic correlation was observed between AAM and body mass index (r_g=-0.19, P=6.11×10⁻³¹). Latent causal variable analyses demonstrate that there is a genetically causal effect of AAM on high blood pressure (GCP=0.47, P=0.02), forced vital capacity (GCP=0.63, P=0.02), age at first live birth (GCP=0.51, P=0.03), impedance of right arm (GCP=0.41, P<1×10^-7) and right leg fat percentage (GCP=-0.10, P=0.02), etc. Enrichment analysis identified 5 enriched tissues and 51 enriched gene sets. Four of the five enriched tissues were related to the nervous system, including the hypothalamus middle, hypothalamo hypophyseal system, neurosecretory systems and hypothalamus. The fifth tissue was the retina in the sensory organ. The most significant gene set was the ‘decreased circulating luteinizing hormone level’ (P=2.45×10^-6). Our findings may provide useful insights that elucidate the mechanisms determining AAM and the genetic interplay between AAM and some traits of women.

Internal Medicine

Molecular Biology

Joint Genome-wide association studies

Age at menarche

Functional analysis

Age at menarche (AAM), defined as the age of first menstrual bleeding, is a commonly reported marker of pubertal timing in females (1). AAM is determined by the overall duration of endocrine-tissue sex hormone exposure level (2), and has a series of clinical outcomes through the women's life. On one hand, earlier AAM is considered to be a risk factor for certain diseases, including obesity (3), type 2 diabetes (4), cardiovascular diseases (5), breast cancer (6) and ovarian cancer (7). On the other hand, later AAM may be associated with an increased risk of Alzheimer's disease (8) and stroke (9), as well as lower fertility (10, 11).

AAM is a highly heritable trait, with estimated heritability up to 50% in previous twin and family studies (12, 13). To date, hundreds of variants associated with AAM have been identified by a number of genome-wide association studies (GWASs) and their meta-analyses (14, 15). Nonetheless, variants identified by the largest GWAS meta-analyses to date only explain 7.4% of the total phenotypic variation, far less than the total 50% of heritability (14). Therefore, the majority of the genetic components underlying AAM remains to be discovered. The unidentified 'missing' heritability is probably carried by genetic variants with minor phenotypic effects that escaped detection by studies with insufficient sample size and statistical power (16, 17), and enlarged sample is warranted to uncover them. To maximize power, it is critical to make use of meta-analyses of multiple independent GWASs whose increasing sample size boosts the statistical power for detecting modest associations.

In the present study, aiming to identify more association signals that contribute to AAM, we conduct a joint GWAS meta-analysis of AAM in 438,089 participants by integrating 3 GWASs, namely the Reproductive Genetics (ReproGen) Consortium (N=182,416), the UK Biobank (UKB, N=188,644) and the Biobank Japan (BBJ, N=67,029). A series of bioinformatical analyses are then followed to explore in-depth annotations at the associated loci.

Study design

Summary statistics of 3 large GWASs were included in this study, 2 of which are of European population and the third one is of East Asian population. The study design is shown in Figure 1. In brief, we conducted two joint analyses. First, the European ancestry-specific meta-analysis was performed to combine results from the ReproGen and the UKB samples. Second, the BBJ was included for the trans-ancestry meta-analysis. The purpose of the second analysis was to identify more loci across populations by a maximal sample size. With European-specific meta-analysis results, we performed a series of follow-up analyses that are sensitive to linkage disequilibrium (LD) pattern, including LD score regression (LDSC) analysis, genetic correlation analysis, Generalized Summary data-based Mendelian Randomization (GSMR) analysis and latent causal variable (LCV) analysis. Functional enrichment and candidate gene prioritization were conducted with trans-ancestry meta-analysis results.

Study populations

The first study is the ReproGen study, which is a GWAS meta-analysis of 182,416 women of European descent from 58 samples (18). In brief, genome-wide SNPs were genotyped by genotyping arrays, and were imputed into the HapMap Phase II CEU build 35 or 36 reference panel. The second study is the UKB sample, which included 188,644 women of European descent (19). All participants were genotyped by the UK BiLEVE Axiom array or UKB Axiom array, and were imputed into UK10K haplotype, 1000 Genomes project phase 3 and Haplotype Reference Consortium (HRC) reference panels (20). Subjects who had a self-reported gender inconsistent with the genetic gender, who were genotyped but not imputed or who withdraw their consents were removed. The last sample is the BBJ study, which is a single GWAS of 67,029 women of East Asian descent (21). BBJ participants had DNA genotyped on more than 950,000 variants using either (a) a combination of Illumina Human OmniExpress BeadChip and Infinium HumanExome BeadChip or (b) Infinium OmniExpressExome BeadChip alone. Variants overlapping across these two sets of genotyping arrays were extracted. After quality control (QC), genome-wide genotypes were imputed into the 1000 Genomes Project Phase 3 reference panel (21).

All participants had provided written informed consent and each study had its research protocol approved by the corresponding local ethics committee or Institutional Review Boards (IRB). No new IRB approval was required. Summary statistics for the 3 studies were downloaded from their respective websites.

Data quality control

We excluded the variants which were duplicated, poorly imputed, or without allele frequency information, only common or less common (minor allele frequency, MAF>1%) SNPs from each individual study were included into analyses (22). There were 2,441,815 genetic variants in the Reprogen released summary data. After QC, there were 2,400,657 variants remained. There were 13,788,288 genetic variants in the UKB released summary data. After removing X-chromosome variants, monomorphic site and duplicated variants (i.e., multiple variants correspond to one identifier), 13,752,112 variants are left. These 13,752,112 variants were matched with 2,400,657 variants in the Reprogen summary data. After removing variants that had intermediate locus incompatibility (i.e., A/G and A/T polymorphisms) in both studies, a total of 2,378,813 variants were eligible for European ancestry-specific meta-analysis. The BBJ study included 9,296,729 genetic variants, of which 15,851 duplicates were removed. The remaining 9,280,878 variables were matched with the 2,378,813 variants in the European ancestry-specific analysis, and then variants with incompatible alleles were excluded. Finally, 2,090,939 genetic variants were present in all three studies, which were used for trans-ancestry meta-analysis.

Meta-analysis

Summary statistics from each GWAS sample were combined by an inverse-variance weighted fixed-effects model implemented in METAL (23). In the European ancestry-specific meta-analysis, the ReproGen and the UKBB studies were meta-analyzed. In the trans-ancestry meta-analysis, all 3 GWASs were meta-analyzed. The potential heterogeneity effects of each variant were assessed using Q statistics and I² index in European populations and in all-ancestry populations.

Identification of novel loci

Genome-wide significance (GWS) level was set at 5.0×10⁻⁸. An independent locus was defined as a genomic region of 500kb on either side of the variant showing the strongest association signal. We declared an association locus to be novel if it contained GWS SNPs that were at least 500kb outside from any reported loci and the novel lead SNP was not in linkage disequilibrium (LD, r²<0.1) with previously reported signals. We further compared the results of the European ancestry-specific analysis and the trans-ancestry meta-analysis to investigate the ancestry specificity of the identified loci. A European-ancestry specific locus was defined if this locus was: (a) identified only in European ancestry-specific meta-analysis; or (b) identified in both meta-analyses, but was significantly heterogeneous (I²≥50%) in trans-ancestry meta-analysis and was not heterogeneous in European ancestry-specific analysis. For the overlapped loci in two analyses (a same locus with different lead SNP), we picked the most significant SNP in the two analyses as the lead SNP at this locus.

Genetic architecture

The LDSC method was applied to the European ancestry specific meta-analysis results to estimate the amount of genomic inflation due to confounding factors such as population stratification and cryptic relatedness (24). LDSC takes GWAS summary statistics as input and partitions overall inflated association statistic into one part attributable to polygenic architecture and the other part due to population stratification and cryptic relatedness. The relative contribution of confounding factors was measured by attenuation ratio (AR), which is defined as (intercept-1)/(mean χ²-1), where intercept and mean χ² are estimates of confounding and the overall association inflation, respectively (24).

Genetic correlations and causal inference

To evaluate genetic interplay between AAM and other complex traits/diseases, genetic correlation analyses were performed with a LDSC based web tool Complex-Traits Genetics Virtual Lab (CTG-VL) (25). A total of 1,387 traits were available at the CTG-VL database. Genetic correlation between AAM and each of these 1,387 traits was estimated. Significance threshold was set at a Bonferroni corrected significance level 3.60ⅹ10⁻⁵, e.g., 0.05/1387.

To further investigate the causal relationship for significant genetic correlations, we performed GSMR to evaluate the causal associations between AAM and correlated traits. GSMR can account for LD between the variants, and is more powerful than existing summary data-based MR methods (26) (27) (28). GWS SNPs show no evidence of horizontal pleiotropy (HEIDI-outlier test P_HEIDI<0.01) and independent of each other (LD r²<0.05 and window size=1 Mb) were selected as instrumental variables (IVs). We switched the exposure and outcome to explore the reverse causation in GSMR analysis.

GSMR may suffer from horizontal pleiotropy even after correction. To resolve this problem, a recently developed LCV model was used to validate the significant results of GSMR analysis (29). The LCV method estimates a genetic causality proportion (GCP) for each pair of traits being evaluated, which represents the proportion of genetic components of one trait that influence the other trait. A positive GCP value indicates that a proportion of the genetic component of AAM affects the other trait, while a negative GCP value means that the proportion of genetic component of the other trait affects AAM.

Candidate gene prioritization

We incorporated the following five sources of information to prioritize candidate genes at the identified loci: (i) being the gene nearest the novel lead SNP; (ii) being the target gene for any local expression quantitative trait loci (cis-eQTL); (iii) containing a novel GWS missense SNP; (iv) being prioritized by summary data based mendelian randomization (SMR) analysis and (v) being prioritized by Summary-MultiXcan (S-MultiXcan) analysis.

HaploReg was used to identify the nearest gene of the lead SNP (30). The Ensembl Variant Effect Predictor (VEP) was used to look over whether there is a missense mutation at the location of the GWS SNPs (31). Gene expression information from seven tissues (brain hypothalamus, breast mammary tissue, ovary, pituitary, uterus, vagina and whole blood) linked to AAM were included for cis-eQTL, SMR and S-MultiXcan analyses.

We analyzed all novel lead SNPs for their cis-eQTL activity. Briefly, the cis-eQTL was defined as an association signal from SNPs located within 100 kb upstream and downstream of the target gene (32). Using a Bonferroni correction, we set the significance threshold to be 0.05/N_tissue/N_SNPs. Cis-eQTLs are accessed from the web portal (http://genehopper.de/qtlizer).

SMR is another prioritization method developed by Zhu and his colleagues (33). SMR integrates GWAS summary results with data from eQTL studies to identify genes whose expression levels are associated with trait because of causal or pleiotropy effects (33). Using SMR analysis, we examined for pleiotropy between GWAS signal and cis-eQTL for genes within 1 megabase (Mb) of the sentinel SNP to identify a possible causal relationship between gene expression and AAM. Only probes with eQTL P<5.0×10⁻⁸ were included in the SMR analysis. HEIDI test P-values<0.05 were taken to indicate significant heterogeneity.

S-MultiXcan is a summary result-based extension that leverages the substantial sharing of eQTLs across tissues and contexts to improve the ability to identify potential target genes (34). We applied MultiXcan to 7 tissues from the CTG-VL. The total number of gene-tissue pairs were used when determining the Bonferroni corrected significance threshold.

Enrichment analysis

Tissue/cell type enrichment analysis and gene set enrichment analysis were performed using DEPICT(35). DEPICT is an integrative tool that employs predicted gene functions to systematically prioritize the most likely causal genes and tissues/cell types where genes from associated loci are highly expressed. Results with FDR<0.05 was considered significant enrichment. We used all suggestive significant SNPs (P<1×10⁻⁵) from trans- ancestry meta-analysis as input.

European ancestry-specific meta-analysis

A total of 371,060 participants and 2,378,813 SNPs passed the QC in the European ancestry-specific GWAS meta-analysis. The LDSC estimates genome-wide heritability to be 10.54% for AAM. The distribution of genome-wide test statistics demonstrated significant inflation (genomic control inflation factor λ_GC = 1.53). Results of LDSC analysis showed that 94.0% of the inflation in the mean χ² statistic is from polygenic architecture rather than from population stratification (LD score intercept=1.05±0.01 SE, mean χ² = 1.83). Manhattan plot of the GWAS meta-analyses results is displayed in Fig. 2.

6,342 SNPs are significant at the GWS level, encompassing into 229 distinct loci. Of these, 14 loci contained 86 GWS variants that were at least 500kb outside from any reported loci. Lead SNPs in 2 of the 14 loci are in LD with previously reported signals (rs13094336 r²=0.53 with rs6445624, rs2764272 r²=0.17 with rs6931884) (36). We excluded 31 GWS variants located within these 2 loci and declared the remaining 12 loci contained 55 GWS variants to be novel (Supplementary Table 1). These 12 novel loci are independent of each other (r²<0.01). All the 12 lead SNPs have consistent effect directions in the ReproGen sample and the UKB sample. 4 of the 12 lead SNPs show significant heterogeneity (I²>50%), including rs13126656, rs1549864, rs11773714 and rs12220144 (Table 1).

Table 1

Novel loci identified in European ancestry specific analysis
							Reprogen							Meta
Marker	CHR	POS	Locus	EA	OA	EAF	Beta	SE	P1	Beta		SE	P2	Beta	SE	P	HetI²
rs2002446	2	116473865	2q14.1	A	C	0.11	0.03	0.009	3.3×10⁻³	0.02	0.004		2.64×10⁻⁶	0.02	0.003	4.86×10⁻⁸	0
rs17400325	2	178565913	2q31.2	T	C	0.96	0.04	0.014	6.5×10⁻³	0.03	0.003		1.62×10⁻⁶	0.03	0.006	4.35×10⁻⁸	0
rs11248108	4	2481490	4p16.3	A	G	0.15	-0.02	0.008	4.70×10⁻³	-0.02	0.003		8.15×10⁻⁷	-0.02	0.003	1.74×10⁻⁸	0
rs1027013	4	102904490	4q24	A	G	0.57	-0.02	0.006	5.80×10⁻³	-0.01	0.002		2.93×10⁻⁷	-0.01	0.002	6.86×10⁻⁹	0
rs13126656	4	161955155	4q32.2	A	T	0.22	-0.03	0.007	2.40×10⁻⁴	-0.01	0.003		5.59×10⁻⁶	-0.01	0.003	1.98×10⁻⁸	61.5
rs1265093	6	31107187	6p21.33	A	G	0.29	0.02	0.007	6.60×10⁻³	0.01	0.003		1.70×10⁻⁷	0.01	0.002	4.58×10⁻⁹	0
rs2326574	6	5041470	6p25.1	T	G	0.88	0.03	0.008	1.9×10⁻³	0.02	0.004		1.13×10⁻⁶	0.02	0.003	1.24×10⁻⁸	0
rs17442489	7	10383553	7p21.3	T	C	0.63	-0.03	0.006	1.10×10⁻⁶	-0.01	0.002		4.77×10⁻⁵	-0.01	0.002	1.80×10⁻⁸	88.4
rs11773714	7	142523468	7q34	A	C	0.26	0.22	0.03	7.60×10⁻⁵	0.01	0.003		8.16×10⁻⁶	0.01	0.003	1.84×10⁻⁸	74.3
rs12220144	10	70364884	10q21.3	T	G	0.17	0.03	0.008	1.6×10⁻⁴	0.01	0.003		1.31×10⁻⁵	0.02	0.003	4.77×10⁻⁸	70.9
rs11186505	10	92982908	10q23.32	A	G	0.30	0.01	0.007	0.26	0.01	0.003		3.87×10⁻⁸	0.01	0.003	3.41×10⁻⁸	2.5
rs6060752	20	34597962	20q11.23	A	G	0.05	0.04	0.014	0.01	0.03	0.006		3.36×10⁻⁷	0.03	0.005	1.27×10⁻⁸	0
Note: CHR, chromosome; POS, genomic position based on GRCH37 genome assembly; EA, effect allele; OA, other allele; EAF, effect allele frequency; Beta, regression coefficient; SE, standard error of Beta; HetI², heterogeneity test I square; P1, the P value of ReproGen Consortium; P2, the P value of UKB sample; P, the P value calculated in meta-analyses. The European-ancestry specific loci are shown with bold font.

Trans-ancestry meta-analysis

After systematic QC, association statistics for 2,100,684 autosomal SNPs were combined across all three studies. Manhattan plot of the GWAS meta-analyses are displayed in Fig. 3.

The results contain 6,370 SNPs significant at the GWS level, encompassing 239 distinct loci. By removing reported loci, we identified 93 novel SNPs mapping to 17 independent loci (Table 2). These 17 novel loci are independent of each other (r²<0.01). 16 of 17 lead SNPs have consistent association direction in all three samples, except for rs11773714 at 7q34 (beta_ReproGen=0.027, beta_UKB=0.013 and beta_BBJ=-0.007). 4 of the 17 novel lead SNPs show significant heterogeneity (I²>50%), included rs7429712, rs1549864, rs11773714, and rs7159376 (Supplementary Table 2).

Table 2

Trans-ancestry GWAS meta-analysis identified 17 novel loci associated with AAM
							Reprogen			UKB			BBJ			Meta
Marker	CHR	POS	Locus	EA	OA	EAF	Beta	SE	P1	Beta	SE	P2	Beta	SE	P3	Beta	SE	P	HetI²
rs1881395	2	27838549	2p23.3	A	G	0.22	-0.02	0.007	2.30×10⁻³	-0.01	0.003	6.80×10⁻⁶	-0.01	0.010	0.16	-0.01	0.003	3.51×10⁻⁸	0
rs6730140	2	148460036	2q22.3	T	C	0.14	-0.03	0.010	1.70×10⁻³	-0.02	0.004	1.81×10⁻⁵	-0.03	0.011	0.01	-0.02	0.003	1.45×10⁻⁸	6.5
rs7429712	3	75265159	3p12.3	C	G	0.60	-0.02	0.006	5.40×10⁻⁴	-0.01	0.003	5.22×10⁻⁵	-0.03	0.009	3.43×10⁻³	-0.01	0.002	1.68×10⁻⁸	63.7
rs1203863	4	2502637	4p16.3	T	C	0.78	0.02	0.008	3.70×10⁻³	0.01	0.003	4.15×10⁻⁶	0.03	0.009	7.69×10⁻³	0.02	0.003	2.62×10⁻⁹	0
rs1027013	4	102904490	4q24	A	G	0.68	-0.02	0.006	5.80×10⁻³	-0.01	0.002	2.93×10⁻⁷	-0.03	0.012	0.04	-0.01	0.002	1.23×10⁻⁹	0
rs13126656	4	161955155	4q32.2	A	T	0.17	-0.03	0.007	2.40×10⁻⁴	-0.01	0.003	5.59×10⁻⁶	-0.02	0.018	0.27	-0.02	0.003	1.09×10⁻⁸	25.2
rs2029641	4	167440421	4q32.3	A	G	0.52	0.01	0.006	0.11	0.01	0.002	4.35×10⁻⁷	0.01	0.009	0.15	0.01	0.002	4.27×10⁻⁸	0
rs17769849	4	180899465	4q34.3	T	C	0.37	-0.01	0.006	0.08	-0.01	0.002	7.70×10⁻⁷	-0.02	0.009	0.04	-0.01	0.002	2.67×10⁻⁸	0
rs32015	5	66697639	5q12.3	A	G	0.46	-0.02	0.006	2.60×10⁻³	-0.01	0.002	6.09×10⁻⁶	-0.02	0.009	0.08	-0.01	0.002	1.97×10⁻⁸	0
rs1265093	6	31107187	6p21.33	A	G	0.31	0.02	0.007	6.60×10⁻³	0.01	0.003	1.70×10⁻⁷	0.02	0.009	0.06	0.01	0.002	7.67×10⁻¹⁰	0
rs1549864	7	10390011	7p21.3	C	G	0.72	-0.03	0.006	1.90×10⁻⁶	-0.01	0.003	6.47×10⁻⁵	-0.02	0.015	0.21	-0.01	0.002	1.52×10⁻⁸	75.2
rs4723134	7	32255322	7p14.3	T	C	0.24	-0.01	0.007	0.04	-0.01	0.003	1.06×10⁻⁶	-0.02	0.013	0.12	-0.01	0.003	3.66×10⁻⁸	0
rs11773714	7	142523468	7q34	A	C	0.22	0.03	0.007	7.60×10⁻⁵	0.01	0.003	8.16×10⁻⁶	-0.01	0.015	0.61	0.01	0.003	4.60×10⁻⁸	66.1
rs3101268	8	106086274	8q22.3	A	G	0.88	-0.03	0.014	0.05	-0.02	0.005	6.99×10⁻⁶	-0.03	0.010	1.10×10⁻³	-0.03	0.005	3.65×10⁻⁸	0
rs11186505	10	92982908	10q23.32	A	G	0.25	0.01	0.007	0.26	0.01	0.003	3.87×10⁻⁸	0.02	0.018	0.31	0.01	0.002	2.10×10⁻⁸	0
rs7159376	14	47921389	14q21.3	A	C	0.54	0.02	0.006	6.80×10⁻⁵	0.01	0.002	2.23×10⁻⁵	0.01	0.009	0.36	0.01	0.002	4.11×10⁻⁸	56
rs6060752	20	34597962	20q11.23	A	G	0.07	0.04	0.014	0.01	0.03	0.006	3.36×10⁻⁷	0.04	0.015	0.013	0.03	0.005	5.26×10⁻¹⁰	0
Note: CHR, chromosome; POS, genomic position based on GRCH37 genome assembly; EA, effect allele; OA, other allele; EAF, effect allele frequency; Beta, regression coefficient; SE, standard error of Beta; HetI², heterogeneity test I square; P1, the P value of ReproGen Consortium; P2, the P value of UKB sample; P3, the P value of BBJ sample; P, the P value calculated in meta-analyses; HetI², heterogeneity test I square.

European-ancestry specific loci

By comprising the results from European ancestry-specific meta-analysis and trans-ancestry meta-analysis, 214 loci were significant both in the European ancestry-specific meta-analysis and trans-ancestry meta-analysis. 15 loci are only significant in European ancestry-specific meta-analysis. Among them, 4 are novel European-ancestry specific loci (2q14.1, 2q31.2, 6p25.1 and 10q21.3) (Table 1). The lead SNPs at 2q14.1 2q31.2 and 6p25.1 were not found in the BBJ study. The lead SNP rs12220144 at 10q21.3 was not significant in the trans-ancestry meta-analysis (P=2.31×10⁻⁷), and had a negative effect in BBJ and a positive effect in the two European samples. In short, a total of 256 loci, including 241 trans-ancestry loci and 15 European-ancestry specific loci were identified in this study. Among them, 21 loci, including 17 trans-ancestry loci and 4 European-ancestry specific loci are novel.

Genetic correlations and causal inference

LDSC analyses revealed that AAM were genetically correlated with 67 complex traits and diseases, including BMI (r_g=-0.19, P=6.11×10⁻³¹), impedance of whole body (r_g=-0.13, P=4.93×10⁻²⁰) and age at first live birth (r_g=0.08, P=1.31×10⁻⁶), etc (Supplementary Table 3). The genetic correlations were consistent with observational associations from epidemiological studies (5) (14).

We further explored the causal relationship between AAM and the above 67 genetic correlated traits through CTG-VL (Supplementary Table 3). In GSMR analysis, at the stringent significance threshold of Bonferroni correction (P_adj<0.05/67/2=3.73×10⁻⁴), when AAM is the exposure (GSMR1), we observed 45 significant causal associations (Supplementary Table 3). When AAM is the outcome (GSMR2), we observed 9 significant causal associations (Supplementary Table 3). The 9 traits were still significant in GSMR1, indicating that the association between AAM and the 9 traits is bidirectional. In GSMR1, AAM had a causal effect with the remaining 36 (45–9) significant traits (P<3.73×10⁻⁴, P_HEIDI<0.1), among which impedance had the largest causal effect value: left arm impedance Beta=13.32, P=2.75×10⁻¹⁸⁰, whole body impedance Beta=21.85, P=2.05×10⁻¹⁶⁶ (Table 7). In addition, we also found that trunk fat mass (P_GSMR1=0.39, P_GSMR2=5.83×10⁻⁶) was only significant in GSMR2 analysis, indicating that trunk fat mass would affect AAM.

To avoid the horizontal pleiotropic effect, we used the recently developed LCV method to verify the significant causal association identified above. We further explored the causal relationship between AAM and the above 67 genetically related traits and three ocular diseases by CTG-VL. LCV found 13 significant associations, and all GCP values were positive, indicating that the genetic component of AAM had a certain impact on other traits, as shown in Table 3. The trait with the largest GCP value was ‘forced vital capacity’ (GCP=0.63, P=0.02). There is no significant genetic causal effect between AAM and BMI (GCP=-0.08, P=0.62). While significant causal association was observed between AAM and other fat distribution traits, including impedance of left arm (GCP=0.40, P<1.0×10⁻⁷), impedance of right arm (GCP=0.41, P<1.0×10⁻⁷), left leg fat percentage (GCP=0.24, P=0.01), and right leg fat percentage (GCP=0.23, P=0.02). Significant causal associations were also observed between AAM and forced vital capacity (FVC) (GCP=0.64, P=0.02), blood pressure medication (GCP=0.54, P=0.04), age at first live birth (GCP=0.51, P=0.03), etc. After FDR adjustment, these 13 associations remained significant.

Table 3

The results of latent casual variable analysis
Phenotype	r_g	SE	GCP(SE)	P	P_FDR
Forced vital capacity (FVC)	0.07	0.015	0.63(0.27)	0.02	0.026
Medication for cholesterol, blood pressure or diabetes: None of the above	0.10	0.018	0.57(0.25)	0.02	0.026
Medication for cholesterol, blood pressure or diabetes: Blood pressure medication	-0.10	0.017	0.54(0.28)	0.04	0.026
Age at first live birth	0.08	0.017	0.51(0.27)	0.03	0.033
Vascular/heart problems diagnosed by doctor: None of the above	0.08	0.013	0.48(0.23)	9.89×10⁻³	0.026
Vascular/heart problems diagnosed by doctor: High blood pressure	-0.08	0.014	0.47(0.24)	0.02	0.026
Non-cancer illness code, self-reported: hypertension	-0.08	0.013	0.48(0.24)	0.01	0.026
Impedance of arm (right)	0.12	0.013	0.41(0.17)	<1×10⁻⁷	<6.50×10⁻⁷
Impedance of arm (left)	0.12	0.013	0.40(0.16)	<1×10⁻⁷	<6.50×10⁻⁷
Age at last live birth	0.08	0.020	0.38(0.22)	0.03	0.033
Leg fat percentage (left)	-0.11	0.014	0.24(0.13)	0.01	0.026
Illnesses of siblings: None of the above (group 1)	0.12	0.020	0.24(0.14)	0.04	0.040
Leg fat percentage (right)	-0.10	0.014	0.23(0.15)	0.02	0.026
Note: Significant genetic causal relationship between age at menarche and the correlated traits; r_g, regression coefficients from age at menarche to the phenotypes; SE, standard error of r_g; GCP (SE): genetic causality proportion (standard error of GCP); P, the P value of GCP. P_FDR, P value adjusted by FDR. Group 1: Heart disease, Stroke, High blood pressure, Chronic bronchitis/emphysema, Alzheimer's disease/dementia, Diabetes.

Combined with GSMR and LCV analysis, it was found that all the 13 traits identified in LCV were significant in GSMR1, indicating strong evidence for causal association between AAM and these 13 traits. In addition, leg fat percentage and arm impedance in body composition were significant in both GSMR1 and LCV. Notably, "right arm impedance" was significant in all three analyses (beta_GSMR1=12.61 P=3.86×10⁻¹⁷⁴, beta_GSMR2=-0.001 P=9.52×10⁻⁶, GCP=0.41 P<1×10⁻⁷), and the effect was most significant in GSMR1, indicating that arm impedance would increase by 12.61 units for each year of AAM delay. In GSMR analysis, AAM was causality associated with age of first sexual intercourse (beta_GSMR1=0.46, P_GSMR1=1.92×10⁻³⁰; beta_GSMR2<-0.01, P_GSMR2=0.83), while in LCV analysis, the results were only suggestive (GCP=0.41, P=0.05).

Candidate gene prioritization

By integrating data from five sources, we prioritized 33 candidate genes at 21 novel loci (Table 4). Some candidate genes are highly expressed in endocrine-related tissues. For example, ZNF512 expresses in ovary and endometrium (37). Some of these loci were screened for more than one gene. For example, 10 candidate genes were identified at 6p21.33, including cis-eQTL target genes CCHCR1, HLA-C, HCG27, POU5F1, TCF19 and XXBacBPG299F13.17, HLA-C, HCG27, XXbacBPG299F13.17 and HLA-B, as well as CCHCR1 and PSORS1C2 containing missense coding SNPs. Some of the genes show more than one line of supporting evidence. For example, PDE11A at 2q31.2 is not only the cis-eQTL target gene of the lead SNP rs12408576 within this locus, but also contains a missense coding SNP (rs12408576).

Table 4

Prioritized candidate genes at the identified novel loci
Locus	Lead SNP	Candidate genes
2p23.3	rs1881395	ZNF512 (N)
2q14.1	rs2002446	DPP10 (N)
2q22.3	rs6730140	RN5S106 (N)
2q31.2	rs17400325	PDE11A (N, M)
3p12.3	rs7429712	MIR4444-1 (N)
4p16.3	rs1203863	RNF4 (N, Q)
4q24	rs1027013	BANK1 (N), RP11-10L12.2 (X)
4q32.2	rs13126656	AC106860.1 (N)
4q32.3	rs2029641	RP11-217C7.1 (N)
4q34.3	rs17769849	RP11-404J23.1 (N)
5q12.3	rs32015	RP11-434D9.1 (N)
6p25.1	rs2326574	RNase_MRP (N), RP11-428J1.4 (Q)
6p21.33	rs1265093	PSORS1C1 (N), CCHCR1 (M, Q), HLA-C (Q, X), HCG27 (Q, X), POU5F1 (Q), TCF19 (Q), XXbacBPG299F13.17 (Q), PSORS1C2 (M), XXbac-BPG248L24.12 (X), HLA-B (X)
7p21.3	rs1549864	U3 (N)
7p14.3	rs4723134	PDE1C (N)
7q34	rs11773714	TRBV30 (N, Q, S)
8q22.3	rs3101268	RP11-127H5.1 (N)
10q21.3	rs12220144	TET1 (N), SLC25A16 (Q)
10q23.32	rs11186505	PCGF5 (N)
14q21.3	rs7159376	MDGA2 (N)
20q11.23	rs6060752	C20orf152 (N)
Note: Genes were prioritized as following: gene nearest to the novel lead SNP (N); gene containing a missense SNP (M); gene with mRNA levels in association with one or more SNPs (cis-eQTL, Q); gene prioritized by summary mendelian randomization (SMR) analysis (S); and gene prioritized by Summary-MultiXcan (S-MultiXcan) analysis (X).

Enrichment analysis

We next explored particular tissues and cells in which the expression of genes residing in AAM-associated loci was enriched. A total of 5 tissues were identified in trans-ancestry meta-analysis (Table 5). Four of the five tissues were related to the nervous system, including the hypothalamus middle (P=1.92×10⁻⁴), hypothalamo hypophyseal system (P=1.92×10⁻⁴), neurosecretory systems (P=1.92×10⁻⁴), and the hypothalamus (P=1.84×10⁻³). The fifth tissue was the retina in the sensory organ (P=4.25×10⁻⁴).

Table 5

Enriched tissues of DEPICT
MeSH term	MeSH first level term	MeSH second level term	P	P_FDR<5%
A08.186.211.730.317.357.352	Hypothalamus Middle	Nervous System	1.92×10⁻⁴	Yes
A08.186.211.730.317.357.352.435	Hypothalamo Hypophyseal System	Nervous System	1.92×10⁻⁴	Yes
A08.713	Neurosecretory Systems	Nervous System	1.92×10⁻⁴	Yes
A09.371.729	Retina	Sense Organs	4.25×10⁻⁴	Yes
A08.186.211.730.317.357	Hypothalamus	Nervous System	1.84×10⁻³	Yes

51 gene sets were identified by the gene set enrichment analysis (Supplementary Table 4). The most significant gene set was the MP:0002773 ‘decreased circulating luteinizing hormone level’ (P=2.45×10⁻⁶). Other significant gene sets include ‘positive regulation of insulin-like growth factor receptor signaling pathway’, ‘decreased uterus weight’, ‘abnormal diencephalon morphology’, ‘decreased growth hormone level’ and ‘decreased circulating insulin-like growth factor I level’ etc. These enriched gene sets may be associated with female development and menarche.

In the present study, aiming to identify AAM-associated genetic loci, we performed a European ancestry-specific meta-analysis and a trans-ancestry GWAS meta-analysis by combing the summary statistics of three large GWAS samples of AAM. We identified a total of 21 novel AAM associated loci, including 4 European ancestry-specific loci.

There have been examples in recent studies suggesting that causal genes are distinct from the nearest genes (33, 38), so we integrated five complementary methods to select candidate genes. The candidate genes we identified may play an important role in the development of some diseases and cancers. For example, hereditary breast cancer and ovarian cancer pathogenic variants were found in PDE11A (34). RNF4 interacts also with the androgen receptor (AR) functioning as a coactivator (39), and AR-mediated androgen actions are important for normal female fertility (40). AR knockout mouse models have identified that AR function is required for full functionality in follicle health, development and ovulation through both intra-ovarian and neuroendocrine mechanisms (40). Methylation quantitative trait locus 6p21.33 (near PSORS1C1) was reported associated with reproductive traits and diseases (41).

In the enrichment analysis, 4 tissues in the nervous system are identified. Besides, we also identified gene sets such as ‘decreased circulating luteinizing hormone level’ and ‘decreased circulating insulin-like growth factor I level’, etc. The hypothalamus and pituitary gland are key structure in the hypothalamic-pituitary-gonadal (HPG) axis--they play an important role in sexual maturation during puberty (42). The activation of the HPG axis during this developmental period involves: 1) the release of the gonadotropin-releasing hormone (GnRH) from the hypothalamus; 2) the release of gonadotropins, luteinizing hormone (LH) and follicle stimulating hormone (FSH) from the pituitary gland; and 3) the release of sex steroids from maturing gonads, so that the operation of a drive from the hypothalamus and pituitary gland indicated (43). LH is a heterodimeric glycoprotein hormone released from the anterior pituitary gland in response to GnRH stimulation from the hypothalamus (44). One of the important roles of LH is to regulate ovarian function (45). LH plays a key role in follicular maturation, ovulation, luteal development and maintenance (44, 45). Lower LH levels can affect ovulation, which can affect menarche and pregnancy (44). Besides, the role that insulin and IGF-1 play in controlling the hypothalamus and pituitary and their role in regulating puberty and nutritional control of reproduction has been studied extensively (46). Study has shown IGF-I acts at different levels of the HPG axis; it exerts paracrine effects at the ovary and stimulates GnRH at the hypothalamic-pituitary level (47).

In addition, the expression of genes residing in AAM-associated loci was enriched in the retina. In terms of the ocular system, sex hormones are recognized for their critical role in regulating important body functions that affect the eyes(48). Sex hormones can have a neuroprotective action on the retina and modulate ocular blood flow (49). Oestrogen is abundant in the mammalian eye (50). Some studies are suggestive of a modest protective effect of estrogen exposure on the eye health of women, and estrogen may confer antioxidative protection against cataractogenesis (51, 52). It is unsurprising therefore that oestrogen levels may impact vision through its effects on the eye, from the ocular surface to the retina. Younger AAM augments estrogen exposure and lower the risk of cataracts in advanced age (53), and was also associated with a protective effect regarding nuclear sclerosis (54). In conclusion, both AAM and sense organs are derived from common genetic factors.

In genetic correlations and causal inference analysis, we identified significant genetic correlations between AAM and BMI, which was consistent with previous study (14). While further causality inference demonstrates that there was no significant causal relationship between AAM and BMI. However, we observed causal relationship between AAM and body composition traits, including leg fat percentage and impedance of arm. Impedance can be used to indirectly estimate body composition such as fat mass and fat free mass (55). Analysis indicated evidence of an effect of younger AAM on lower impedance, lower total body water, lower fat free mass, and higher fat mass. The inconsistent casual associations may be because BMI is a complex composite of fat mass, lean mass, bone mass and other soft tissues. Future studies are needed to further investigate the potential effects of AAM on body composition.

MR has previously been used to explore effects of age at menarche on cardiometabolic health outcomes (included systolic and diastolic blood pressure), but AAM was not clearly or inconsistent associated with any other cardiovascular risk factor (56) (57). It therefore remains unclear whether the previously reported associations between AAM and health outcomes reflect a causal effect. However, the causal relationship between younger AAM and higher blood pressure was obvious in our study. The causal relationship between AAM and cardiovascular diseases such as blood pressure is also consistent with epidemiological studies (58, 59). We also observed a causal effect of AAM on age at first live birth. Early AAM is also associated with early marriage, which may have particularly important implications for age at first live birth (60). A phenome-wide Mendelian randomization study results of AAM with 17,893 health-related traits suggest that younger age at menarche has potential effects on a broad range of health-related traits. Follow-up analysis indicated imprecise evidence of an effect of younger AAM on several outcomes, including lower lung function, confidence intervals were wide and often included the null (61). Therefore, larger study populations are needed to re-examine these relationships. The analyses of genetic correlation and casual effect are relatively rough in this study, and the results are only suggestive for AAM-related diseases and traits. We will further explore the causal association of AAM with some diseases and traits.

There are some limitations in our study. Firstly, we combined the summary statistics using fixed-effects model, which does not account for genetic heterogeneity that may arise from multiple race/ethnic populations. However, the random-effects model that is more robust to genetic heterogeneity is over-conservative (62). We noticed that it is a convention to perform fixed-effects model analysis even with multiple ethnic participants (63, 64). In addition, the dataset we used in the genetic correlations and causal inference analyses contains both females and males. We expected that choosing to use a dataset that includes both males and females would have attenuated the association signal between AAM and correlated traits in this study, thus rendering our results conservative (65). Therefore, future MR studies may be warranted to verify our results in female only samples (when such samples are available).

In summary, by performing European ancestry-specific and trans-ancestry meta-analyses for AAM, we generated an atlas of candidate SNPs and genes involved in AAM determination and its shared regulation with multiple cardiometabolic traits. Our findings may further elucidate the mechanisms that determine the timing of AAM and underlie its links to disease risk.

Acknowledgements

We acknowledge the Reproductive Genetics Consortium, the Biobank Japan project and UK Biobank Cohort for releasing the age at menarche GWAS summary results.

Authors' contributions

GJF and YFP designed the study. QX and GJF collected the data. GJF and QX analyzed the data. JJN, SSY, BXH, XTW, HZ, LZ and QGZ performed the literature search. GJF drafted the early version of the manuscript. YFP and LZ jointly supervised the study. All authors were involved in writing the paper and had final approval of the submitted and published versions.

Funding

YFP and LZ are benefited from national natural science foundation of China (31771417 and 31571291), a project funded by the Priority Academic Program Development (PAPD) of Jiangsu higher education institutions. The numerical calculations in this paper have been done on the supercomputing system of the National Supercomputing Center in Changsha.

Competing interests

On behalf of all authors, the corresponding authors state that there is no conflict of interest.

Ethics approval and consent to participate
Not applicable.

Availability of data and materials

Genome-wide summary statistics from the three samples are available to download from the following websites: (i) the ReproGen Consortium (http://www.reprogen.org/); (ii) the UK Biobank (https://www.dropbox .com/s/wabrpmebxrj3xwq/2714.gwas.imputed_v3.female.tsv.bgz?dl=0 -O 2714.gwas.imputed_v3.female.tsv.bgz); (iii) the BBJ sample (https://humandbs.biosciencedbc.jp/en/) under data set identifier hum0014.v9.Men.v1and hum0014.v9.MP.v1.
Consent for publication
Not applicable.

Gill D, Sheehan NA, Wielscher M, Shrine N, Amaral AFS, Thompson JR, et al. Age at menarche and lung function: a Mendelian randomization study. Eur J Epidemiol. 2017;32(8):701–10.
Anderson CA, Duffy DL, Martin NG, Visscher PM. Estimation of variance components for age at menarche in twin families. Behav Genet. 2007;37(5):668–77.
Pierce MB, Leon DA. Age at menarche and adult BMI in the Aberdeen children of the 1950s cohort study. Am J Clin Nutr. 2005;82(4):733–9.
He C, Zhang C, Hunter DJ, Hankinson SE, Buck Louis GM, Hediger ML, et al. Age at menarche and risk of type 2 diabetes: results from 2 large prospective cohort studies. Am J Epidemiol. 2010;171(3):334–44.
Lakshman R, Forouhi NG, Sharp SJ, Luben R, Bingham SA, Khaw KT, et al. Early age at menarche associated with cardiovascular disease and mortality. J Clin Endocrinol Metab. 2009;94(12):4953–60.
Hsieh CC, Trichopoulos D, Katsouyanni K, Yuasa S. Age at menarche, age at menopause, height and obesity as risk factors for breast cancer: associations and interactions in an international case-control study. Int J Cancer. 1990;46(5):796–800.
Gong TT, Wu QJ, Vogtmann E, Lin B, Wang YL. Age at menarche and risk of ovarian cancer: a meta-analysis of epidemiological studies. Int J Cancer. 2013;132(12):2894–900.
Rees M. The age of menarche. ORGYN. 1995(4):2–4.
Cui R, Iso H, Toyoshima H, Date C, Yamamoto A, Kikuchi S, et al. Relationships of age at menarche and menopause, and reproductive year with mortality from cardiovascular disease in Japanese postmenopausal women: the JACC study. J Epidemiol. 2006;16(5):177–84.
Presser HB. Age at menarche, socio-sexual behavior, and fertility. Soc Biol. 1978;25(2):94–101.
Komura H, Miyake A, Chen CF, Tanizawa O, Yoshikawa H. Relationship of age at menarche and subsequent fertility. Eur J Obstet Gynecol Reprod Biol. 1992;44(3):201–3.
Towne B, Czerwinski SA, Demerath EW, Blangero J, Roche AF, Siervogel RM. Heritability of age at menarche in girls from the Fels Longitudinal Study. Am J Phys Anthropol. 2005;128(1):210–9.
van den Berg SM, Boomsma DI. The familial clustering of age at menarche in extended twin families. Behav Genet. 2007;37(5):661–7.
Day FR, Thompson DJ, Helgason H, Chasman DI, Finucane H, Sulem P, et al. Genomic analyses identify hundreds of variants associated with age at menarche and support a role for puberty timing in cancer risk. Nat Genet. 2017;49(6):834–41.
Perry JR, Day F, Elks CE, Sulem P, Thompson DJ, Ferreira T, et al. Parent-of-origin-specific allelic associations among 106 genomic loci for age at menarche. Nature. 2014;514(7520):92–7.
Altshuler D, Daly M. Guilt beyond a reasonable doubt. Nat Genet. 2007;39(7):813–5.
Wang WY, Barratt BJ, Clayton DG, Todd JA. Genome-wide association studies: theoretical and practical concerns. Nat Rev Genet. 2005;6(2):109–18.
Perry JR, Day F, Elks CE, Sulem P, Thompson DJ, Ferreira T, et al. Parent-of-origin-specific allelic associations among 106 genomic loci for age at menarche. Nature. 2014;514(7520):92–7.
Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015;12(3):e1001779.
Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203–9.
Horikoshi M, Day FR, Akiyama M, Hirata M, Kamatani Y, Matsuda K, et al. Elucidating the genetic architecture of reproductive ageing in the Japanese population. Nat Commun. 2018;9(1):1977.
Winkler TW, Day FR, Croteau-Chonka DC, Wood AR, Locke AE, Magi R, et al. Quality control and conduct of genome-wide association meta-analyses. Nat Protoc. 2014;9(5):1192–212.
Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics. 2010;26(17):2190–1.
Bulik-Sullivan BK, Loh PR, Finucane HK, Ripke S, Yang J, Patterson N, et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet. 2015;47(3):291–5.
Cuéllar-Partida G, Lundberg M, Kho PF, D’Urso S, Gutiérrez-Mondragón LF, Ngo TT, et al. Complex-Traits Genetics Virtual Lab: A community-driven web platform for post-GWAS analyses. 2019:518027.
Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37(7):658–65.
Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44(2):512–25.
Burgess S, Dudbridge F, Thompson SG. Combining information on multiple instrumental variables in Mendelian randomization: comparison of allele score and summarized data methods. Statistics in medicine. 2016;35(11):1880–906.
O'Connor LJ, Price AL. Distinguishing genetic correlation from causation across 52 diseases and complex traits. Nat Genet. 2018;50(12):1728–34.
Ward LD, Kellis M. HaploReg: a resource for exploring chromatin states, conservation, and regulatory motif alterations within sets of genetically linked variants. Nucleic acids research. 2012;40(Database issue):D930-4.
McLaren W, Gil L, Hunt SE, Riat HS, Ritchie GR, Thormann A, et al. The Ensembl Variant Effect Predictor. Genome Biol. 2016;17(1):122.
Dixon AL, Liang L, Moffatt MF, Chen W, Heath S, Wong KC, et al. A genome-wide association study of global gene expression. Nat Genet. 2007;39(10):1202–7.
Zhu Z, Zhang F, Hu H, Bakshi A, Robinson MR, Powell JE, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat Genet. 2016;48(5):481–7.
Barbeira AN, Pividori M, Zheng J, Wheeler HE, Nicolae DL, Im HK. Integrating predicted transcriptome from multiple tissues improves association detection. PLoS Genet. 2019;15(1):e1007889.
Pers TH, Karjalainen JM, Chan Y, Westra HJ, Wood AR, Yang J, et al. Biological interpretation of genome-wide association studies using predicted gene functions. Nat Commun. 2015;6:5890.
Day FR, Thompson DJ, Helgason H, Chasman DI, Finucane H, Sulem P, et al. Genomic analyses identify hundreds of variants associated with age at menarche and support a role for puberty timing in cancer risk. Nat Genet. 2017;49(6):834–41.
Fagerberg L, Hallström BM, Oksvold P, Kampf C, Djureinovic D, Odeberg J, et al. Analysis of the human tissue-specific expression by genome-wide integration of transcriptomics and antibody-based proteomics. Molecular cellular proteomics: MCP. 2014;13(2):397–406.
Smemo S, Tena JJ, Kim KH, Gamazon ER, Sakabe NJ, Gomez-Marin C, et al. Obesity-associated variants within FTO form long-range functional connections with IRX3. Nature. 2014;507(7492):371–5.
Pero R, Lembo F, Palmieri EA, Vitiello C, Fedele M, Fusco A, et al. PATZ attenuates the RNF4-mediated enhancement of androgen receptor-dependent transcription. J Biol Chem. 2002;277(5):3280–5.
Walters KA, Simanainen U, Handelsman DJ. Molecular insights into androgen actions in male and female reproductive function from androgen receptor knockout models. Hum Reprod Update. 2010;16(5):543–58.
Mortlock S, Restuadi R, Levien R, Girling JE, Holdsworth-Carson SJ, Healey M, et al. Genetic regulation of methylation in human endometrium and blood and gene targets for reproductive diseases. Clin Epigenetics. 2019;11(1):49.
Wong AP, Pipitone J, Park MTM, Dickie EW, Leonard G, Perron M, et al. Estimating volumes of the pituitary gland from T1-weighted magnetic-resonance images: effects of age, puberty, testosterone, and estradiol. NeuroImage. 2014;94:216–21.
Donovan BT. The role of the hypothalamus in puberty. Prog Brain Res. 1974;41:239–53.
Rao CV. Multiple novel roles of luteinizing hormone. Fertility sterility. 2001;76(6):1097–100.
Filicori M. The role of luteinizing hormone in folliculogenesis and ovulation induction. Fertility sterility. 1999;71(3):405–14.
Wolfe A, Divall S, Wu S. The regulation of reproductive neuroendocrine function by insulin and insulin-like growth factor-1 (IGF-1). Front Neuroendocr. 2014;35(4):558–72.
Daftary SS, Gore AC. (Maywood. IGF-1 in the brain as a regulator of reproductive neuroendocrine function. Experimental biology and medicine. NJ). 2005;230(5):292–306.
Sharif R, Bak-Nielsen S, Hjortdal J, Karamichos D. Pathogenesis of Keratoconus: The intriguing therapeutic potential of Prolactin-inducible protein. Prog Retin Eye Res. 2018;67:150–67.
Nuzzi R, Scalabrin S, Becco A, Panzica G. Gonadal Hormones and Retinal Disorders: A Review. Front Endocrinol. 2018;9:66.
Gupta PD, Johar K, Sr., Nagpal K, Vasavada AR. Sex hormone receptors in the human eye. Survey of ophthalmology. 2005;50(3):274–84.
Vajaranant TS, Nayak S, Wilensky JT, Joslin CE. Gender and glaucoma: what we know and what we need to know. Curr Opin Ophthalmol. 2010;21(2):91–9.
Zetterberg M. Age-related eye disease and gender. Maturitas. 2016;83:19–26.
Younan C, Mitchell P, Cumming RG, Panchapakesan J, Rochtchina E, Hales AM. Hormone replacement therapy, reproductive factors, and the incidence of cataract and cataract surgery: the Blue Mountains Eye Study. Am J Epidemiol. 2002;155(11):997–1006.
Klein BE, Klein R, Ritter LL. Is there evidence of an estrogen effect on age-related lens opacities? The Beaver Dam Eye Study. Archives of ophthalmology (Chicago, Ill: 1960). 1994;112(1):85-91.
Prins M, Hawkesworth S, Wright A, Fulford AJ, Jarjou LM, Prentice AM, et al. Use of bioelectrical impedance analysis to assess body composition in rural Gambian children. Eur J Clin Nutr. 2008;62(9):1065–74.
Au Yeung SL, Jiang C, Cheng KK, Xu L, Zhang W, Lam TH, et al. Age at menarche and cardiovascular risk factors using Mendelian randomization in the Guangzhou Biobank Cohort Study. Preventive medicine. 2017;101:142–8.
Bell JA, Carslake D, Wade KH, Richmond RC, Langdon RJ, Vincent EE, et al. Influence of puberty timing on adiposity and cardiometabolic traits: A Mendelian randomisation study. PLoS Med. 2018;15(8):e1002641.
Won JC, Hong JW, Noh JH, Kim DJ. Association Between Age at Menarche and Risk Factors for Cardiovascular Diseases in Korean Women: The 2010 to 2013 Korea National Health and Nutrition Examination Survey. Medicine. 2016;95(18):e3580.
Magnus MC, Lawlor DA, Iliodromiti S, Padmanabhan S, Nelson SM, Fraser A. Age at Menarche and Cardiometabolic Health: A Sibling Analysis in the Scottish Family Health Study. J Am Heart Assoc. 2018;7(4).
Sandler DP, Wilcox AJ, Horney LF. Age at menarche and subsequent reproductive events. Am J Epidemiol. 1984;119(5):765–74.
Magnus MC, Guyatt AL, Lawn RB, Wyss AB, Trajanoska K, Küpers LK, et al. Identifying potential causal effects of age at menarche: a Mendelian randomization phenome-wide association study. BMC Med. 2020;18(1):71.
Jackson D, Turner R. Power analysis for random-effects meta-analysis. Res Synth Methods. 2017;8(3):290–302.
Tin A, Marten J, Halperin Kuhns VL, Li Y, Wuttke M, Kirsten H, et al. Target genes, variants, tissues and transcriptional pathways influencing human serum urate levels. Nat Genet. 2019.
64. !!! INVALID CITATION !!! .
Zhang Q, Greenbaum J, Zhang WD, Sun CQ, Deng HW. Age at menarche and osteoporosis: A Mendelian randomization study. Bone. 2018;117:91–7.

Supplementarytables.xlsx

Download PDF

Version 1

posted

You are reading this latest preprint version

Joint Genome-Wide Association Study Identifies Twenty-One Novel Loci for Age at Menarche and Highlights Its Causal Association with Other Complex Diseases

Status:

Version 1

Abstract

Figures

Introduction

Materials And Methods

Study design

Study populations

Data quality control

Meta-analysis

Identification of novel loci

Genetic architecture

Genetic correlations and causal inference

Candidate gene prioritization

Enrichment analysis

Results

European ancestry-specific meta-analysis

Trans-ancestry meta-analysis

European-ancestry specific loci

Genetic correlations and causal inference

Candidate gene prioritization

Enrichment analysis

Discussion

Declarations

References

Supplementary Files

Status:

Version 1