Persistence Pays Off for PCOS Gene Prospectors

Polycystic ovary syndrome (PCOS) is a common cause of anovulatory infertility associated with hyperandrogenemia/hyperandrogenism. Based on observations demonstrating familial clustering of PCOS, which are consistent with a genetic component to the disorder, there has been worldwide interest in the identification of genes that contribute to disease risk (1). The incidence of PCOS is thought to be similar across different racial and ethnic groups, affecting 5–7% of women of reproductive age, despite phenotypic differences in its expression among different populations (2). Thus, it can be assumed that similar genes or gene networks may be involved in PCOS across populations. In the quest for PCOS genes, numerous candidate gene association studies have been conducted, and a number of these have yielded statistically significant associations. These include studies using family-based methods, like the transmission disequilibrium test (TDT), which test for both association and linkage. However, the candidate gene studies have been carried out in large part on small sample populations and have limited statistical power. Additionally, few have yielded sufficiently robust results that have been replicated by different investigators studying different populations. What was missing in the field of the genetics of PCOS was a genome-wide association study (GWAS) with adequate power and replication panels to convincingly identify PCOS loci. The search for PCOS genes achieved a major milestone with the publication of a GWAS by Chen et al. (3), who reported on a Han Chinese population. This study encompassed a total of 4082 cases and 6678 controls, with cases defined according to the “Rotterdam criteria.” Chen et al. (3) identified three loci, two on chromosome 2 (2p16.3 and 2p21), and one on chromosome 9 (9q33.3) that had significant associations with PCOS, with the minor alleles conferring protection (chromosome 2 loci) or increased risk (chromosome 9), at levels exceeding the threshold for statistical significance for genome-wide associations (Table 1). Published case-control association studies exploring SNPs discovered in Han Chinese Unless stated otherwise, data are expressed as P values for the meta-analyses or combined analyses, with odds ratios in parentheses. Chr., Chromosome; NS, nonsignificant, P > 0.05; −, not examined. Published case-control association studies exploring SNPs discovered in Han Chinese Unless stated otherwise, data are expressed as P values for the meta-analyses or combined analyses, with odds ratios in parentheses. Chr., Chromosome; NS, nonsignificant, P > 0.05; −, not examined. The pivotal paper of Chen et al. (3) raised a number of issues. Could the results be replicated in populations other than Han Chinese? What is/are the functional variant(s) in these loci that contribute to disease risk? A virtue of the objective GWAS approach is the potential to discover unexpected loci contributing to disease risk. Although one of the loci identified by Chen et al. (3) is near the LHCGR gene, a plausible PCOS candidate, questions arise as to the pathophysiological links of the two other loci (THADA and DENND1A) to reproduction or ovarian function. The publication of Welt et al. (4) in this issue of the JCEM helps to address one of these questions because it replicates the association of PCOS with variants in the DENND1A locus (rs10818854 A allele and rs10986105 C allele) in a combined Icelandic and European study population (4). Importantly, the paper by Welt et al. (4) is not the first replication of this association; it confirms a previously reported replication of the association between the rs10818854 A allele and increased PCOS risk in individuals of European ancestry by Goodarzi et al. (5). Because the reported odds ratios for the DENND1A risk alleles are relatively small, ranging from 1.17 to 1.87, variation in other genes, perhaps involved in a network or pathway including DENND1A, most likely contributes to PCOS risk. There are some important differences between the findings of Chen et al. (3) and those of Goodarzi et al. (5) and Welt et al. (4). The latter two groups found that another single nucleotide polymorphism (SNP) in the DENND1A locus, rs2479106, is not significantly associated with PCOS in European-ancestry/Icelandic populations, whereas the minor G allele is strongly associated with increased PCOS risk in Han Chinese. In another study of Caucasians, Lerchbaum et al. (6) reported a nonstatistically significant “trend” (P = 0.053) for reduced risk for PCOS associated with the rs2479106 G allele. More recently, Eriksen et al. (7), also studying a Caucasian population, found that the rs2479106 G allele was significantly associated with reduced risk of PCOS. Unfortunately, Lerchbaum et al. (6) and Eriksen et al. (7) did not evaluate the SNPs (rs10818854 and rs10986105) that Goodarzi et al. (5) and Welt et al. (4) found to be significantly associated with PCOS in the European/Icelandic populations. The inconsistent findings of rs24791062 association with PCOS in Han Chinese and Caucasians may be explained by linkage disequilibrium (LD) and minor allele frequencies. Although the overall LD pattern is different between Chinese and European populations, for the three SNPs of interest the LD pattern is very similar; rs2479106 is not in strong LD with either rs10818854 or rs10986105, whereas rs10818854 and rs10986105 are in LD (r2 = 0.82 in Europeans and r2 = 0.65 in Chinese), although these two markers are farther apart than either is from rs2479106. The LD pattern, therefore, is consistent with the association data (e.g. disparities in rs24791062 association with PCOS in Han Chinese and European/Icelandic populations). The reason that the two more distal SNP are in strongest LD could be due to the differences in minor allele frequencies. The rare alleles may all be on one haplotype associated with increased PCOS risk, which also contains the rs2479106 minor allele, but because the rs247106 minor allele is much more common, it could also be on other haplotypes. Interestingly, rs10818854 and rs10986105 appear to be associated with hyperandrogenemia, and not polycystic ovaries per se, which is more consistent with the older “NIH Consensus” definition of PCOS that does not include ultrasound-based ovarian morphology as a diagnostic criterion, and is restricted to a combination of menstrual disturbances in the presence of hyperandrogenemia or hyperandrogenism after exclusion of other causes of these phenotypes. In contrast to the DENND1A locus, where there is some consistency among the existing reports, the other loci identified by Chen et al. (3) (LHCGR and THADA) have not been found to be associated with PCOS across all studies. Only Goodarzi et al. (5) found a modest association with a THADA SNP and PCOS in Europeans. The PCOS-associated HCGR variant has a low minor allele frequency in Caucasians, so there would only be modest power to detect associations. Thus, it is not possible to make a definitive statement about the role of these loci (LHCGR and THADA) in PCOS in women of European ancestry, although they may play a role in Han Chinese. Indeed, the THADA locus was analyzed using the TDT with 267 Han Chinese families and shown to be significantly associated and linked to PCOS (8). DENND1A SNP rs24791062 and rs10818854 were also evaluated in this TDT analysis and were not found to be significantly associated or linked to PCOS, probably because of limited statistical power for the non-THADA SNP studied (50–63% with an α of 0.05). The DENND1A gene encodes a protein named connecdenn 1 that interacts with members of the Rab family of small GTPases that are involved in membrane trafficking (9). DENND1A is a member of a family of 18 genes in humans. The encoded proteins contain DENN (Differentially Expressed in Normal and Neoplastic cells) domains, which are tripartite, consisting of a u-DENN (upstream DENN domain), DENN (core DENN domain) and d-DENN (downstream DENN domain) separated by “linker” sequences. The DENN domains function as Rab-specific guanine nucleotide exchange factors. Connecdenn 1 has a clathrin-binding domain and is thought to facilitate endocytosis. It has also been shown to be associated with lipids, particularly phosphoinsitol-3-phosphate, and other endocytosis/endosome proteins (9). The DENND1A gene yields two transcripts, governed by different splicing. One of these transcripts, DENND1A variant 1, codes for a 1009-amino acid protein with a C-terminal proline-rich domain; the other, DENND1A variant 2, codes for a truncated 559-amino acid protein that contains all three DENN domains and the clathrin-binding domain, but lacks the proline-rich domain, and includes a C-terminal 33-amino acid sequence that differs from the connecdenn 1 larger variant. The DENND1A SNPs associated with PCOS are located in introns, and it is not evident that they have functional roles in controlling gene expression. Little is known about DENND1A expression in cells and tissues related to reproduction. We do know, however, that DENND1A is expressed in testes, theca cells, and H295 adrenal carcinoma cells, cells that make androgens (McAllister, J. M., and J. F. Strauss, III, unpublished observations). This provides a possible link between the DENND1A locus and abnormalities in androgen production in women. Whether variant 1 and 2 proteins have the same functions is currently not known. Further studies are required to determine how connecdenn 1 variants might influence steroidogenesis in theca cells and follicular maturation. Alternatively, connecdenn 1 variants, which are expressed in brain, might affect gonadotropin secretion, as suggested by Welt et al. (4), and contribute to dysregulated secretion of FSH and LH that characterizes PCOS. Variations in the DENND1A coding sequences are reported in dbSNP, including variants that cause nonsynonymous amino acid substitutions. These are rare and in most cases are not validated. Their relationship to the DENND1A risk alleles is unknown. However, preliminary analysis of DENND1A coding sequences from a small sample of normal and PCOS theca cells revealed no sequence variation in a small number of samples (McAllister, J. M., and J. F. Strauss, III, unpublished observations). Another mechanism by which the DENND1A locus might contribute to PCOS is through a microRNA-mediated mechanism. MIR601 is embedded in the DENND1A locus, but little is known about its expression in tissues relevant to reproduction and PCOS, let alone its influence on expression of reproduction-related genes. Moreover, none of the DENND1A SNPs significantly associated with PCOS are located near the MIR601 gene. The report of Welt et al. (4) adds credibility to DENND1A as a PCOS candidate locus in multiple populations. With the report of Welt et al., there is now a collective body of work that places DENND1A as an anchor to the search for PCOS genes—a beginning and an opportunity. Research described in this article was supported in part by National Institutes of Health Grant U54HD034449 from the Eunice K. Shriver National Institute of Child Health and Human Development. Disclosure Summary: The authors have nothing to declare. Differentially expressed in normal and neoplastic (cell) genome-wide association study linkage disequilibrium polycystic ovary syndrome single nucleotide polymorphism transmission disequilibrium test.