Further Investigation in Europeans of Susceptibility Variants for Polycystic Ovary Syndrome Discovered in Genome-Wide Association Studies of Chinese Individuals

Abstract

Context:

Two genome-wide association studies (GWAS) of polycystic ovary syndrome (PCOS) have identified 11 susceptibility loci in Chinese individuals. Some of the risk loci identified in Chinese cohorts, mostly from the first GWAS, have been replicated in Europeans. Replication of the loci from the second GWAS in European cohorts is necessary to determine whether the same variants confer risk for PCOS in multiple ethnicities.

Objective:

The objective of the study was to determine the effects of the Chinese GWAS loci in European-origin individuals.

Design:

This was a genetic association study.

Setting:

The study was conducted at a tertiary care academic center.

Patients:

Eight hundred forty-five European subjects with PCOS and 845 controls participated in the study.

Interventions:

Interventions included blood sampling and genotyping.

Main Outcome Measure:

The association between PCOS and 12 independent single-nucleotide polymorphisms mapping to seven of the Chinese GWAS loci in a European cohort was measured.

Results:

Variants in DENND1A (P = .0002), THADA (P = .035), FSHR (P = .007), and INSR (P = .046) were associated with PCOS in Europeans. The genetic risk score, generated for each subject based on the total number of risk alleles, was associated with the diagnosis of PCOS (P < .0001) and remained associated (P = .02), even after exclusion of the four variants individually associated with PCOS.

Conclusions:

At least four of the PCOS susceptibility loci identified in the Chinese GWAS are associated with PCOS in Europeans. The overall genetic burden for PCOS, as demonstrated by the risk score, is also associated with the diagnosis of PCOS in Europeans. The PCOS susceptibility loci identified in the Chinese GWAS are thus likely to play an important role in the etiology of PCOS across ethnicities.

Polycystic ovary syndrome (PCOS) is the most common endocrinopathy among reproductive-aged women, with a prevalence of approximately 6%-10% (1). It is characterized by oligomenorrhea and hyperandrogenism and is associated with an increased risk for obesity, insulin resistance, and type 2 diabetes.

The pathogenesis of PCOS has not yet been fully elucidated, but a genetic component has been previously established. Despite demonstration of substantial heritability of PCOS in twin studies, examination of at least 100 candidate genes did not yield many replicated loci (1). Recently two genome wide association studies (GWAS) identified several new susceptibility loci for PCOS. Both GWAS were carried out in Han Chinese populations. The first GWAS identified three PCOS susceptibility loci that mapped to the genomic areas of THADA, LHCGR, and DENND1A (2). Variants in THADA and DENND1A were subsequently replicated in European cohorts (3, 4). The second GWAS identified eight new susceptibility loci that mapped to the genomic areas of FSHR, C9orf3, INSR, HMGA2, YAP1, RAB5B/SUOX, TOX3, and SUMO1P1 (5). A recent study evaluating the variants from the second GWAS in a Dutch cohort replicated the variant mapping to the YAP1 locus (6).

PCOS can be phenotypically different in Asians compared with Europeans. Asian women commonly present with acne as the primary symptom of hyperandrogenism and rarely with hirsutism (7). The current GWAS-derived PCOS susceptibility loci were discovered in Chinese individuals. Although several of the Chinese GWAS loci, mostly from the first GWAS, have been replicated in Europeans, the impact of the newly discovered loci in other populations has not been fully determined. In this study we sought to replicate association of genetic variants identified in the second Chinese GWAS in a large European cohort.

Materials and Methods

Subjects

The cohort consisted of 845 white subjects with PCOS and 845 white control women. Sources of subjects included 241 PCOS and 147 control women recruited at the University of Alabama, 179 PCOS and 27 control women from Cedars-Sinai Medical Center (CSMC), 356 PCOS and 68 control women from Pennsylvania State University, 69 subjects with PCOS from the Pregnancy in PCOS trial (8), and 603 general community controls from the Cholesterol and Atherosclerosis Pharmacogenetics study (9). All subjects with PCOS met the 1990 National Institutes of Health (NIH) criteria (10) and thus had hyperandrogenism and/or hyperandrogenemia and chronic oligoovulation. Recruitment and phenotyping parameters were previously reported (11). Clinical characteristics are displayed in Supplemental Table 1.

The study was approved by the institutional review boards of the recruiting centers and CSMC. Written informed consent was obtained from all participants.

Genotyping

Genotyping was previously performed at CSMC using Infinium II technology on the Metabochip, following the manufacturer's protocol (Illumina). The Metabochip is a high-throughput genotyping platform that was designed to provide a method by which loci associated with a number of traits related to cardiac and metabolic diseases could be rapidly genotyped (12). The genotyping and quality control details were previously described, which included a principal component analysis to correct for potential population stratification, allowing identification and removal of subjects with substantial non-Caucasian admixture (11). In the current study, the genotypes for 12 independent single-nucleotide polymorphisms (SNPs) from seven loci were extracted from the Metabochip data (Table 1). The goal was to analyze the data of the Chinese GWAS SNPs or SNPs in linkage disequilibrium (LD) with them (r² > 0.8) when available on the Metabochip.

Table 1.

Association Analysis of Chinese GWAS Risk Variants With the Diagnosis of PCOS, Adjusted for BMI

SNP	Risk Allele	Gene	OR	95% CI	P Value
rs13429458	A	THADA	1.03	0.82–1.30	.80
rs6544661	A	THADA	1.06	0.92–1.23	.41
rs12468394	C	THADA	1.18	1.01–1.35	.034
rs11891936	G	THADA	1.16	0.97–1.40	.11
rs6732721	A	LHCGR	1.19	0.88–1.61	.27
rs2479106	G	DENND1A	1.01	0.85–1.19	.93
rs12337273	G	DENND1A	2.02	1.40–2.94	.0002
rs12994034	A	FSHR	1.23	1.06–1.43	.007
rs2349415	A	FSHR	1.13	0.97–1.31	.099
rs7866848	A	C9orf3	1.13	0.97–1.31	.11
rs2272046	A	HMGA2	0.98	0.63–1.51	.91
rs2059807	G	INSR	1.17	1.00–1.36	.046
Genetic Risk Score	–	–	1.09	1.04–1.13	<.0001

SNPs rs6544661 (THADA), rs12337273 (DENND1A), rs12994034 (FSHR), and rs7866848 (C9orf3) are in complete LD with the Chinese GWAS SNPs rs12478601, rs10986105, rs2268361, rs4385527 (r² = 1), respectively. SNP rs2479106 (DENND1A) is in LD with rs10818854 (r² = 0.83), and rs6732721 (LHCGR) is in LD with rs13405728 (r² = 0.87). Boldface indicates statistically significant associations.

Statistical analysis

All continuous parameters with a nonnormal distribution were logarithmically transformed. Unpaired Student's t tests were used to compare clinical characteristics between cases and controls. Quantitative trait data are presented as median (interquartile range). We tested the association between 12 SNPs (additive model) and PCOS using logistic regression. To assess whether the effects of the SNPs were independent of body mass index (BMI), analyses were adjusted for BMI by including it as an additional independent variable. Given the prior evidence of association with PCOS, a value of P < .05 was considered significant.

A genetic risk score (GRS) was generated for each subject based on the total number of risk alleles (alleles associated with increased odds of PCOS). Twelve SNPs were included in the risk score (range 0–24 per individual) based on the risk alleles identified in the Chinese GWAS. The association between the GRS and PCOS was evaluated using logistic regression as described above. The proportion of explained variation (PEV) in PCOS status accounted by the GRS was calculated as the PEV difference between two models, the model with GRS and BMI as independent variables, and the model with only BMI as an independent variable. The PEV for each model was calculated using Gini's concentration measure (13).

Within the subjects with PCOS, multiple linear regression was used to assess the association between genotype (additive model) and continuous quantitative traits [BMI, total T, fasting insulin, fasting glucose, homeostasis model assessment of insulin resistance, homeostasis model assessment of β-cell function (HOMA-%B)], adjusting for BMI except when BMI was the dependent variable. Because these analyses are not in replication of results from the Chinese GWAS, we applied a multiple testing correction value of P = .001 [0.05/36; accounting for 12 SNPs against three trait categories (adiposity, sex hormone, glucose homeostasis)].

We used the Genetic Power Calculator to determine the power to detect association between SNPs and PCOS in our cohort. The sample size of 845 cases and 845 controls has excellent power (≥85%) to detect association of risk alleles of frequency of 0.2 or greater with PCOS at odds ratio (OR) of 1.5 or greater and moderate power (40%-60%) to detect association at an OR of 1.25. Detailed power calculations given in Supplemental Table 2 reveal lower power to detect association of rarer risk alleles (frequency ≤ 0.1) with PCOS at ORs of 1.25 or less.

Results

Allele and genotype frequencies for the analyzed SNPs are displayed in Supplemental Table 3. Of the 12 SNPs, four variants, one in each of THADA, DENND1A, FSHR, and INSR, were significantly associated with PCOS (Table 1). As shown in Supplemental Table 4, results without adjustment for BMI were essentially the same. The DENND1A SNP rs12337273 had the highest level of significance and the greatest effect size [OR 2.02, 95% confidence interval (CI) 1.40–2.92, P = .0002]. The remaining variants in the DENND1A, THADA, and FSHR loci, as well as the variants in LHCGR, C9orf3, and HMGA2, were not associated with PCOS (Table 1); however, the effects of 11 of the 12 variants were directionally consistent with those in Chinese.

The overall GRS was associated with the diagnosis of PCOS (OR 1.09, 95% CI 1.04–1.13, P < .0001), indicating a 9% increase in odds of PCOS per increment in the GRS (Table 1 and Supplemental Table 4). The median risk score was 13 in the controls vs 14 in the subjects with PCOS (P < .0001). The distribution of the genetic risk scores is displayed in Figure 1. The OR of PCOS increased steadily with increasing GRS. The proportion of variance explained by the GRS was 0.90%. A GRS excluding the four SNPs that independently associated with the diagnosis of PCOS was still associated with the diagnosis of PCOS (OR 1.07, 95% CI 1.01–1.13, P = .02).

Figure 1.

Impact of the number of risk alleles on the odds of having PCOS. The risk categories shown on the x-axis are based on the genetic risk score that was generated for each subject based on the total number of PCOS risk alleles. The histogram indicates the percentage of individuals in each risk score group for subjects with PCOS and controls (y-axis on the left). The odds ratios and confidence intervals of having PCOS compared with the reference group are shown (y-axis on the right). The group with a genetic risk score of 13 or 14 was used as the reference.

Exploratory association analyses of the 12 SNPs against quantitative traits were conducted in the subjects with PCOS (Supplemental Table 5). Both independent SNPs (r² = 0.26) in the FSHR loci, rs12994034 and rs2349415, were associated with lower total T levels (P < .001 for both). The overall GRS was nominally associated with HOMA-%B (0.03) (Supplemental Table 5).

Discussion

In this study we evaluated whether the variants associated with PCOS in Han Chinese also associate with PCOS in Europeans. Four prior studies had previously evaluated the three loci (DENND1A, THADA, and LHCGR) identified in the first GWAS in European populations (3, 4, 14, 15). The two larger studies replicated the DENND1A locus (3, 4), and one, which included most of the participants in the current study, also replicated the THADA locus in Europeans (3). A recent study looking at the loci from the second GWAS was able to replicate one of the eight loci, YAP1, in a Dutch cohort (6). We were not able to examine YAP1 because its variant is not on the Metabochip. Our study is the first to replicate the association of rs2059807 (INSR) and rs12994034 (in complete LD with rs2268361 at FSHR) in a European cohort. It is possible the prior study did not detect these associations due to lack of power. Compared with the NIH criteria used in our cohort, Rotterdam criteria, used in the Dutch cohort, may result in reduced power due to greater heterogeneity of cases and a higher proportion of undiagnosed cases within population-based controls (up to 20%) (16). Given current evidence, we conclude that DENND1A, THADA, FSHR, INSR, and YAP1 loci are likely to play important roles in the etiology of PCOS across populations.

Considering loci arising from the first Chinese GWAS, recent functional studies revealed that DENND1A mRNA and protein expression is increased in PCOS ovarian theca cells, in which it promotes androgen biosynthesis (17). The promoter region of the LHCGR was found to be hypomethylated in the granulosa cells of women with PCOS, resulting in increased transcription of LHCGR (18). THADA has not yet been evaluated in functional studies in PCOS. A SNP in the THADA locus has been associated with type 2 diabetes, suggesting a possible role in insulin secretion. However, the THADA SNPs associated with PCOS are not in LD with the diabetes variant (3).

The role of most Chinese GWAS variants in the etiology of PCOS in Europeans remains uncertain. In some cases, low minor allele frequency in Europeans severely limits power to demonstrate significant associations with PCOS. For example, this is the case for the LHCGR SNP (minor allele frequency 0.05 in HapMap CEU) and the HMGA2 SNP (minor allele frequency 0.02 in HapMap CEU). To gain insight on whether the remaining SNPs actually have a role in PCOS susceptibility, we constructed a risk score based on the eight SNPs not individually associated with PCOS in Table 1. The observation that this score was associated with PCOS suggests some or all of these SNPs also play a role in the pathogenesis of PCOS, with small effects not detected in single SNP tests.

The FSHR gene codes for the FSH receptor. Inherited abnormalities in expression or function of FSH receptors could plausibly contribute to the ovulatory dysfunction of PCOS. In contrast to a prior fine-mapping study of the FSHR region (19), we replicated the association of rs12994034 (in complete LD with rs2268361) with PCOS; however, this SNP was not well tagged (r² = ∼0.6) in the prior study. In addition, we found that two SNPs at the FSHR locus were associated with T levels. FSHR null female mice have small ovaries and blocked follicular development, whereas the male mice have low T (20). FSHR mutations have been associated with ovarian response to gonadotropin stimulation (21). Further functional studies are necessary to evaluate the consequences of variation in FSHR.

We found that rs2059807 at the INSR locus was also associated with PCOS. The INSR gene codes for the insulin receptor. A prior study identified SNP rs2252673 in the INSR gene (independent of rs2059807, r² = 0.015) that was associated with PCOS (22). Mutations in the INSR gene are known to cause insulin resistance and hyperinsulinemia (23). Insulin acts via insulin receptors on the theca cells to drive androgen production. Hyperinsulinemia can also decrease hepatic SHBG synthesis, further contributing to elevated levels of free T. It is plausible that variation in the INSR gene plays a role in insulin resistance, hyperinsulinemia, and subsequent hyperandrogenemia, all of which contribute to the pathogenesis of PCOS.

In addition, the overall genetic burden for PCOS, as demonstrated by the risk score, was also nominally associated with β-cell dysfunction. This association of the risk score with increased HOMA-%B provides a possible clue that several of the susceptibility loci discovered to date may influence PCOS risk via increased insulin secretion, in line with the known effect of hyperinsulinemia to exacerbate PCOS.

PCOS is a heterogeneous disorder with multiple definitions that are largely based on expert opinion. The underlying pathophysiology of the various PCOS phenotypes may be different both in terms of genetic and environmental contributions. In our study we considered only the more severe PCOS phenotype by using the NIH criteria to diagnose PCOS. It is possible that European women with a less severe phenotype have different susceptibility genes underlying the disorder. Our study also does not take nongenetic factors contributing to PCOS into account. Such lifestyle and environmental factors impacting the pathogenesis of PCOS may also be different between ethnic groups.

Despite variable phenotypes of PCOS among different ethnicities, there appears to be a common genetic susceptibility across populations, consistent with the notion that PCOS arose early in human evolution (24). We were able to replicate four of the loci identified in the Chinese GWAS studies in Europeans. We predict even more of the Chinese GWAS SNPs would associate with PCOS in studies of larger European cohorts. We also found the GRS, based on most of the variants previously identified in Chinese, associated with the diagnosis of PCOS in Europeans. This further supports the existence of common susceptibility genes and alleles in different ethnicities. Functional studies are needed to better understand the role these genes play in the pathogenesis of PCOS. It is hoped that identification and functional elucidation of PCOS susceptibility genes will ultimately improve our ability to diagnose and treat this common disorder.