Family-Based Analysis of Candidate Genes for Polycystic Ovary Syndrome

Abstract

Context: Polycystic ovary syndrome (PCOS) is a complex disorder having both genetic and environmental components. A number of association studies based on candidate genes have reported significant association, but few have been replicated. D19S884, a polymorphic marker in fibrillin 3 (FBN3), is one of the few association findings that has been replicated in independent sets of families.

Objective: The aims of the study are: 1) to genotype single nucleotide polymorphisms (SNPs) in the region of D19S884; and 2) to follow up with an independent data set, published results reporting evidence for PCOS candidate gene associations.

Design: The transmission disequilibrium test (TDT) was used to analyze linkage and association between PCOS and SNPs in candidate genes previously reported by us and by others as significantly associated with PCOS.

Setting: The study was conducted at academic medical centers.

Patients or Other Participants: A total of 453 families having a proband with PCOS participated in the study. Sisters with PCOS were also included. There was a total of 502 probands and sisters with PCOS.

Intervention(s): There were no interventions.

Main Outcome Measure(s): The outcome measure was transmission frequency of SNP alleles.

Results: We identified a six-SNP haplotype block spanning a 6.7-kb region on chromosome 19p13.2 that includes D19S884. SNP haplotype allele-C alone and in combination with D19S884-allele 8 is significantly associated with PCOS: haplotype-C TDT χ² = 10.0 (P = 0.0016) and haplotype-C/A8 TDT χ² = 7.6 (P = 0.006). SNPs in four of the other 26 putative candidate genes that were tested using the TDT were nominally significant (ACVR2A, POMC, FEM1B, and SGTA). One SNP in POMC (rs12473543, χ² = 9.1; P_corrected = 0.042) is significant after correction for multiple testing.

Conclusions: A polymorphic variant, D19S884, in FBN3 is associated with risk of PCOS. POMC is also a candidate gene of interest.

Family-based transmission disequilibrium test analysis provides evidence for genetic association between PCOS and polymorphic variants in two genes, FBN3 and POMC.

Polycystic ovary syndrome (PCOS) is a common endocrine disorder with a prevalence between 5 and 10% of reproductive-age women (1). In addition to the hallmark features of hyperandrogenemia, chronic anovulation, and infertility, PCOS women are at increased risk for obesity, insulin resistance, type 2 diabetes, and cardiovascular disease (2). Thus, PCOS represents a disorder with broader implications to a woman’s health than reproduction.

Legro et al. (3) described evidence for familial aggregation of hyperandrogenemia in PCOS that is consistent with a genetic contribution to the disease. This observation has been confirmed by others (4,5), providing a strong foundation for research into the genetic contributions to PCOS risk. However, to date, there is no consensus in the literature regarding any proposed PCOS candidate gene, despite the fact that numerous association studies have been published. Most nominally significant positive findings have not been replicated or have not been confirmed in follow-up studies (5,6).

We identified a PCOS candidate region surrounding D19S884, a microsatellite marker in intron 55 of fibrillin 3 (FBN3), located on chromosome 19p13.2 about 800 kb centromeric to the insulin receptor (INSR) gene (7). Previously, using the transmission disequilibrium test (TDT), we replicated significant association and linkage with this marker and PCOS in three independent sets of samples (7,8,9,10). D19S884 was also shown to be associated with PCOS in two independent association studies (11,12), but others have failed to find association in smaller study populations (13,14). The FBN3 gene encodes an extracellular matrix protein, and although the role of FBN3 in PCOS remains unclear, FBN3 expression can be demonstrated in the human ovary (14,15). Furthermore, other members of the fibrillin family of proteins are known to bind members of the TGFβ family (16), which play significant roles in controlling ovarian follicular development (17), as well as muscle and adipocyte development and differentiation (18,19,20), all processes that can be impaired in PCOS.

To further investigate the role of variation in the region of D19S884, we extended our earlier TDT observations of D19S884 by identifying a six-single nucleotide polymorphism (SNP) haplotype allele that is significantly associated with PCOS. We also performed family-based analysis using the TDT to assess linkage and association between PCOS and other candidate genes that arose from our earlier work (7) or published studies from other laboratories (21,22,23,24,25,26,27,28,29,30). We followed up 14 genes that showed either increased allele sharing among siblings or nominally significant TDT results in Urbanek et al. (7): ACVR2A, CYP11A, CYP17, CYP19, HSD17B2, IGF1R, INHBA, INHBB, INSL3, INSR, SMAD4, LEP, POMC, and SHBG. We also followed up SNPs and SNP haplotypes in 12 genes shown to have a significant association with PCOS when analyzed in case-control studies in Caucasian women. These include ADIPOQ (25), CAPN5 (21), ENPP1/PC-1 (26), EPHX1 (29), FEM1A and FEM1B (22), H6PD (30), HSD17B6 (27), IL1A (28), SGTA (24), SRD5A1, and SRD5A2 (23). Among these genes, ACVR2A, POMC, FEM1B, and SGTA were nominally significant, but only POMC is significant after correction for multiple testing.

Subjects and Methods

PCOS families

SNP markers were genotyped in 453 families with PCOS: 406 simplex families (one affected daughter and both parents) and 47 multiplex families (two or more affected daughters and both parents). The total number of offspring with PCOS was 502, including 49 affected sisters in the 44 multiplex families. The self-identified ethnicities of probands in the families were: 87% white, 4% Hispanic, 1% black, and 7% other or unknown. A total of 373 families are in common with those reported in Stewart et al. (8). Ninety-two families from the collection reported in Stewart et al. (8) were not included in the current study due to consistent Mendelian exclusions (n = 2), depleted DNA stock (n = 25), parental DNA that was not obtainable (n = 62), or changes of proband phenotype (n = 3). Since the Stewart et al. (8) publication, an additional 80 families have been added to the collection.

Diagnostic criteria have been described in detail elsewhere (3,7). Probands and sisters were considered affected if they had six or fewer menses per year and elevated total testosterone (greater than 58 ng/dl) or elevated non-SHBG-bound testosterone (greater than 15 ng/dl); these thresholds are 2 sd greater than the mean of our normal control individuals (3,7). To ensure a more homogeneous collection of PCOS women, we included only those sisters who met the same diagnostic criteria as the proband. This represents a change from the previous assignment of affected status for sisters, which was based on elevated total testosterone or non-SHBG-bound testosterone only, without a requirement of six or fewer menses per year.

This study was approved by the institutional review boards of the University of Pennsylvania, Penn State University College of Medicine, Brigham and Women’s Hospital, and Northwestern University. Written informed consent was obtained from all adult subjects and from a parent or guardian for minor subjects.

SNP genotyping

We genotyped 104 SNPs in 27 candidate genes. SNPs in 12 of these genes have been found to be associated with PCOS disease risk by others in case-control studies. To increase our likelihood of replicating results from previous PCOS association studies, we restricted our follow-up studies to those that were carried out in Caucasian women having at least 100 PCOS cases. Whenever possible, we genotyped the same SNP as reported in the literature. In three cases (rs1640262 in SGTA, rs3797179 in SRD5A1, and rs2754530 in SDR5A2), the assay for the reported SNP failed, and a nearby SNP in strong linkage disequilibrium (LD) was substituted. For replication of linkage and association results with microsatellite markers located near candidate genes reported in Urbanek et al. (7), we chose SNPs with minor allele frequency greater than 0.2 spaced approximately every 10 kb across the gene. SNPs were genotyped with Applied Biosystems TaqMan SNP Genotyping Assays (Applied Biosystems, Foster City, CA). Allelic PCR products were analyzed using the Applied Biosystems 7900HT Sequence Detection System and SDS 2.2 software. Genotypes were auto-called by SDS 2.2 software with the quality value set at 0.95. Three Centre d’Etude du Polymorphisme Humain individuals were genotyped on each of 16 96-well plates; no discrepancies were observed for any of the SNPs. All SNPs were in Hardy-Weinberg equilibrium. (Details of the location and size of the genes studied, SNP location, and SNP genotyping assays can be found in Supplemental Tables 1 and 2, published on The Endocrine Society’s Journals Online web site at http://jcem.endojournals.org.)

Statistical analysis

Error-checking of genotypes in the family material was performed with Merlin software (31), and families with one or more Mendelian discrepancies for a marker were excluded in the analysis of that marker. Association between SNPs and PCOS was tested with the TDT (32), which tests for the combined presence of linkage and association with the marker.

Because markers within a gene tend to be correlated to varying degrees, correction for multiple testing with Bonferroni methods using the total number of markers tested is likely to be too stringent. We corrected for the number of genes tested separately in each of the two categories of follow-up studies: 1) 14 genes with suggestive evidence for linkage or association with PCOS in Urbanek et al. (7); and 2) 12 genes shown to be associated with PCOS in studies by other laboratories (21,22,23,24,25,26,27,28,29,30). The corrected P value in each case is 0.0036 and 0.0042, respectively, corresponding to a nominal P of 0.05.

The haplotype structure for SNPs in FBN3 and D19S884 was determined using HAPLORE v242 (http://www.soph.hab.edu/stratgenetics/People/KZhang/HAPLORE) (33) which allows input of multiple D19S884 alleles. In all other cases, haplotypes were constructed and analyzed with Haploview v4.1 (http://www.broadinstitute.org/mpq/haploview) (34).

Evolutionary conservation

We compared an approximately 2.2-kb interval in FBN3 containing D19S884 (chr 19: 8,054,042–8,056,288; March 2006 human genome assembly version hg18) against multiple other species using the Multiz Alignment and Conservation (44 species), Basewise Conservation by PhyloP for 28-Species (Placental Mammal and Vertebrate), PhastCons Placental Mammal Conserved Elements (28-way Multiz Alignment), PhastCons Vertebrate Conserved Elements (28-way Multiz Alignment), and PhastCons Vertebrate Conserved Elements (17-way Vertebrate Multiz Alignment) tracks in the University of California Santa Cruz Genome Browser (http://genome.usc.edu). To more quantitatively determine the evolutionary conservation of the intron harboring D19S884 (between exons 55 and 56 in FBN3), we compared the distribution of its 28-way PhyloP scores to the distribution from 1000 random introns using the Wilcoxon rank sum test.

DNA topography

The three-dimensional molecular structure of DNA is affected by nucleotide changes, which can cause differences in protein binding and affinity. We measured the level of change in DNA shape that arises from noncoding SNPs (rs3745393, rs2303169, rs17160147, rs17202517, rs73503752, and rs28525575) near D19S884 (35). We compared these levels of change across haplotype alleles by ranking the structure-change-value as determined by predicted hydroxyl radical cleavage patterns (36) for the six SNPs in the haplotype. All the structure change comparisons are relative to the reference allele. We used the Wilcoxon rank sum test to compare the distribution of changes associated with each haplotype to determine significance.

Results

SNPs in the D19S884 region of FNB3

The region around D19S884 continues to have the most convincing evidence for association with PCOS, having been replicated in three independent data sets by us (7,8,9) and others (11,12). To further investigate this region, we genotyped 17 SNPs spanning a 77-kb region of FBN3 that included D19S884 in our full collection of 453 families (Table 1). The strongest evidence for association with PCOS based on our existing collection of families is allele 8 (A8) of D19S884 (χ² = 13.64; P = 0.0002). Three SNPs located within a 2-kb region of introns 55 and 56 telomeric to D19S884 were also significant: rs17160147 (χ² = 4.06; P = 0.044), rs17202517 (χ² = 10.51; P = 0.0012), and rs73503752 [χ² = 10.84; P = 0.001; identified as FBN3-6197 in Stewart et al. (8)]. SNPs rs17202517 and rs73503752 are significant after correction for testing 18 markers in FBN3.

Table 1.

TDT results for the microsatellite marker D19S884 and 17 SNPs in FBN3

Marker ID	Position	Allelesa	MAF	Overtransmitted allele	T	Not T	Total TDT	Transmission frequency	TDT χ2	P value (nominal)
rs2287937	8036419	C/G	0.18	C	114	96	210	0.543	1.54	>0.05
rs3745393	8049164	A/G	0.437	A	245	211	456	0.537	2.53	>0.05
rs2303169	8052113	G/A	0.441	G	242	215	457	0.530	1.60	>0.05
rs17160147	8054041	C/G	0.317	C	217	177	394	0.551	4.06	0.044
rs17202517	8054300	C/T	0.231	T	195	136	331	0.589	10.51	0.0012b
rs73503752c	8055015	G/A	0.238	A	190	131	321	0.592	10.84	9.9E-04b
rs28525575	8055474	A/C	0.487	C	238	212	450	0.529	1.50	>0.05
D19S884	8056105-8056141			7	106	101	207	0.512	0.12	>0.05
				8	177	114	291	0.608	13.64	2.2E-04b
				9	96	109	205	0.468	0.82	>0.05
				11	82	73	155	0.529	0.52	>0.05
				13	153	168	321	0.477	0.70	>0.05
rs7260399	8056661	C/T	0.478	T	196	166	362	0.541	2.49	>0.05
rs8103000	8057238	C/G	0.282	C	135	128	263	0.513	0.19	>0.05
rs12460643	8058724	C/T	0.266	T	139	133	272	0.511	0.13	>0.05
rs12981294	8062220	A/G	0.227	G	142	121	263	0.540	1.68	>0.05
rs7245552	8065408	T/G	0.38	T	190	175	365	0.520	0.62	>0.05
rs3813780	8067609	C/T	0.228	T	117	104	221	0.529	0.76	>0.05
rs2086149	8083619	C/T	0.278	C	136	136	272	0.500	0	>0.05
rs4804064	8094934	G/A	0.268	A	122	119	241	0.506	0.04	>0.05
rs12974280	8102507	G/C	0.379	C	150	149	299	0.502	0.003	>0.05
rs12162237	8113069	G/A	0.455	A	182	175	357	0.510	0.14	>0.05

MAF, Minor allele frequency; T, number of transmissions in TDT analysis. 

^aMinor allele in bold. 

^bRemained significant after correction for multiple testing. 

^cSNP rs73503752 was identified as FBN3-6197 in Stewart et al. (8). 

These three SNPs plus rs3745393, rs2303169, rs28525575, and D19S884 constitute a haplotype block (Fig. 1). LD among these seven markers in our study population is strong and ranged between 0.77 and 0.99 (average D′ = 0.94; median = 0.96). (The range of r² was 0.13 to 0.94; average r² = 0.46.) We first carried out TDT analysis with haplotypes defined by the SNPs alone, without including alleles of D19S884; the third most common haplotype, designated allele-C, is significantly overtransmitted to offspring with PCOS (χ² = 10.0; P = 0.0016) (Table 2). Although not significant, haplotypes A and B are both undertransmitted. When D19S884 alleles were included in the haplotype, A8 was found on the same haplotype as allele-C 80.1% of the time. Together, haplotype allele-C and A8 are overtransmitted in the TDT analysis (χ² = 7.6; P = 0.0059) (Table 2).

Figure 1.

TDT and haplotype analysis of SNPs in the region of D19S884. A, Individual TDT results are shown for six SNPs and D19S884 located between exons 55 and 59 of FBN3 (number above each bar represents transmission frequency, and dashed line represents nominally significant χ² = 3.84). Also shown are the positions of the markers relative to the exons. B, Pairwise LD plot showing D′ values (percent) in the PCOS families. For construction of haplotypes in Haploview, D19S884 was coded as a two-allele system: allele 8 or not-8. (The x-axis displays location of chromosome 19: 8049164-8056141.)

Table 2.

TDT analysis for the most common FBN3 SNP haplotypes and the SNP haplotypes combined with D19S884 alleles

	Allele frequency	T	Not T	Total	Transmission frequency	TDT χ2	P value
SNP haloptype allelea
A. GACCGA	0.454	117	141	258	0.453	2.23	>0.05
B. AGGCGC	0.278	94	118	212	0.443	2.72	>0.05
C. AGCTAC	0.184	100	60	160	0.625	10.00	0.0016
D. AGGCGA	0.030	21	17	38	0.553	0.42	>0.05
E. AGCTAA	0.021	13	10	23	0.565	0.39	>0.05
All others	0.032
Haloptype allele/D19S884 allele
A/11	0.067	33	41	74	0.446	0.86	>0.05
A/13	0.192	86	81	167	0.515	0.15	>0.05
B/7	0.107	54	54	108	0.500	0	>0.05
B/8	0.027	16	19	35	0.457	0.26	>0.05
B/9	0.092	41	53	94	0.436	1.53	>0.05
C/8	0.138	79	48	127	0.622	7.57	0.0059
C/9	0.025	17	11	28	0.607	1.29	>0.05

T, Number of transmissions in TDT analysis. 

^aOrder of SNPs in haplotype: rs3745393, rs2303169, rs17160147, rs17202517, rs73503752, and rs28525575. 

Evolutionary conservation and DNA topography in the D19S884 region of FNB3

We investigated the degree of evolutionary conservation in a 2.2-kb interval surrounding D19S884. Aside from the expected conservation of the exons, neither the PhastCons tracks nor the PhyloP algorithm shows any region of the intron as conserved. The intron harboring D19S884 had significantly less conservation compared with 1000 random introns (P < 2.2 × 10⁻⁶; see Supplemental Figs. 1 and 2).

We also investigated whether SNPs near D19S884 might alter the three-dimensional molecular structure of the DNA, which could potentially influence protein binding and affinity properties. Three of the six SNPs had DNA structural change values greater than 0.8, suggestive of potentially functional consequences (Table 3). However, the magnitude of this structure-change-value did not correlate well with association with PCOS, as determined by TDT (Table 2). Comparison of haplotypes A–E showed that the low-frequency haplotype E had the most drastic DNA structural changes, followed by haplotype C (see Supplemental Fig. 3). However, there was no statistically significant difference (P > 0.05) in predicted DNA structure changes associated with haplotype C relative to haplotype A or B.

Table 3.

Predicted changes in three-dimensional molecular structure of DNA at noncoding SNPs near D19S884

SNP	Position	TDT χ2	P value (nominal)	DNA structural change
rs3745393	8049164	2.53	>0.05	0.511
rs2303169	8052113	1.60	>0.05	0.262
rs17160147	8054041	4.06	0.044	0.839
rs17202517	8054300	10.51	0.0012	0.895
rs73503752	8055015	10.84	9.9E-04	0.440
rs28525575	8055474	1.50	>0.05	1.395

DNA structural change value greater than 0.8 is suggestive of potentially functional consequences. (See Ref. 35.) 

Family-based analysis of other PCOS candidate genes

Eighty-seven SNPs in 26 PCOS candidate genes were genotyped and analyzed using the TDT. Fourteen of these genes (ACVR2A, CYP11A, CYP17, CYP19, HSD17B2, IGF1R, INHBA, INHBB, INSL3, INSR, SMAD4, LEP, POMC, and SHBG) were follow-up studies of genes originally analyzed by Urbanek et al. (7). In that study, microsatellite markers that were, in some cases, quite far from the gene were used. To further investigate these candidate genes, we genotyped SNPs located approximately every 12.5 kb in each gene. We also followed up 12 genes studied in case-control studies by others: ADIPOQ (25), CAPN5 (21), ENPP1/PC-1 (26), EPHX1 (29), FEM1A and FEM1B (22), H6PD (30), HSD17B6 (27), IL1A (28), SGTA (24), SRD5A1 and SRD5A2 (23). SNPs in four of these genes had a nominally significant TDT result: ACVR2A and POMC (7), FEM1B (22), and SGTA (24) (Table 4). (See Supplemental Table 3 for a complete list of TDT results of all SNPs that were genotyped.)

Table 4.

PCOS candidate genes with one or more significant SNP (P ≤ 0.05)

Gene (Ref.)	SNP ID	Position	Overtransmitted allele	T	Not T	Total T	Transmission frequency	TDT χ2	P value (nominal)
ACVR2A (7)	rs1364658	chr2:148324277	C	123	118	241	0.510	0.10	>0.05
	rs1895694	chr2:148336215	T	227	193	420	0.541	2.75	>0.05
	rs1424941	chr2:148359587	T	156	113	269	0.580	6.87	0.009
	rs12987286	chr2:148380037	T	120	110	230	0.522	0.44	>0.05
	rs3768688	chr2:148387570	G	189	146	335	0.564	5.52	0.019
	rs2303392	chr2:148396896	C	119	106	225	0.529	0.75	>0.05
	rs3768686	chr2:148409245	T	229	182	411	0.557	5.38	0.020
POMC (7)	rs1042571	chr2:25237390	G	113	107	220	0.514	0.16	>0.05
	rs12473543	chr2:25240684	T	181	128	309	0.586	9.09	0.003
	rs934778	chr2:25242727	G	96	88	184	0.522	0.35	>0.05
	rs1009388	chr2:25244604	C	123	119	242	0.508	0.07	>0.05
FEM1B (22)	rs10152450	chr15:66358532	G	191	239	430	0.447	5.36	0.021
	rs6494730	chr15:66371610	T	175	215	390	0.449	4.10	0.043
	Haplotype		G-T	197	238	435	0.453	3.86	0.049
SGTA (24)	rs2238614	chr19:2706338	G	175	134	309	0.567	5.44	0.019
	rs741103	chr19:2722640	T	140	118	258	0.543	1.88	>0.05
	rs1005556	chr19:2729542	G	181	144	325	0.557	4.21	0.040
	Haplotype		G-T-G	216	165	381	0.646	7.08	0.008

T, Number of transmissions in TDT analysis. 

ACVR2A and POMC

Seven SNPs that span ACVR2A and four SNPs in POMC were genotyped (Table 4). Two SNPs were nominally significant in ACVR2A: rs1424941 (χ² = 6.87; P = 0.009) and rs3768688 (χ² = 5.52; P = 0.019), but neither was significant after correction for multiple testing for 14 genes. One of the four SNPs in POMC, rs12473543, remained significant after correction (χ² = 9.09; P = 0.0026; P_corrected=0.036). This SNP is located 329 bp 3′ of exon 2 in the 7.8-kb POMC gene. None of the ACVR2A or POMC SNP haplotype blocks were significantly associated with PCOS in the TDT analysis.

FEM1B and SGTA

Our results for two SNPs in FEM1B and three SNPs in SGTA confirm the findings of Goodarzi et al. (22,24) (Table 4). In our results and those of Goodarzi et al. (22), the same (minor) alleles of rs10152450 and rs6494730 in FEM1B had a reduced frequency with PCOS or were undertransmitted in the TDT analysis. Goodarzi et al. (22) also reported a five-SNP haplotype with reduced frequency with PCOS: G-G-A-A-T, P = 0.004. We genotyped only the first and last SNPs that tagged their significant haplotype; the allele carrying the minor alleles of both SNPs, G and T, was also undertransmitted in the TDT analysis (χ² =3.86; P = 0.049).

Goodarzi et al. (24) analyzed three SNPs (rs2238614, rs741103, and rs1640262) in SGTA, a gene on chromosome 19p13, 5.3 MB telomeric to D19S884. Only rs1640262 was significant, with the minor allele CC genotype being protective against PCOS (P = 0.009). We genotyped both rs2238614 and rs741103 but replaced rs1640262 with rs1005556. [The replacement SNP rs1005556 is located in 590 bp 5′ of rs1640262, and the two SNPs are in complete LD (D′ = 1 in HapMap_CEU).] In our analysis, rs1005556 was also significant; the minor T allele was undertransmitted, in agreement with Goodarzi et al. (24) (TDT χ² = 4.2; P = 0.040). The most common haplotype containing the major alleles at all three SNPs was significant in both the Goodarzi et al. study (rs2238614-rs741103-rs1640262: G-A-T; P = 0.015) and our analysis (rs2238614-rs741103-rs1005556: G-T-G; TDT χ² =7.08; P = 0.008). However, none of the results for these four SNPs were significant after correcting for the 12 genes tested in this follow-up study.

None of the other SNPs genotyped in this family-based replication of previously reported case-control studies showed significant evidence for linkage or association in our TDT analysis (Table 5). This includes SNPs and SNP haplotypes in ADIPOQ (25), CAPN5 (21), ENPP1/PC-1 (26), EPHX1 (29), FEM1A (22), H6PD (30), HSD17B6 (27), IL1A (28), SRD5A1 and SRD5A2 (23).

Table 5.

TDT analysis for SNPs and haplotypes in candidate genes previously reported in the literature as being associated with PCOS

Gene tested (Ref.)	SNP tested	Major/ minor alleles (+ strand)	Literature studies			TDT results
			No. of cases: controls	Population	Reported association: P value	Allele	T	Not T	Total	Transmission frequency	TDT χ2	P value
ADIPOQ (25)	rs2241766	T/G	143:245	Finland	ns	T	46	44	90	0.511	0.04	0.841
	rs1501299	G/T			T reduced frequency: 0.047	T	193	160	353	0.547	3.09	0.079
	Haplotype				ns	TT	114.8	95.9	210.7	0.545	1.70	0.192
CAPN5 (21)	rs948976	A/G	148:606	Spain	ns	G	184	169	353	0.521	0.64	0.425
	rs4945140	G/A			ns	G	219	212	431	0.508	0.11	0.736
	rs2233546	C/T			ns	T	44	39	83	0.530	0.30	0.583
	rs2233549	G/A			ns	A	159	152	311	0.511	0.16	0.691
	Haplotype				GGCA: 0.009	GGCA	49.3	44.5	93.8	0.526	0.24	0.623
	Haplotype				GGTG: 0.001	GGTG	25.3	20.5	45.8	0.552	0.50	0.478
ENPP1/ PC1 (26)	rs1044498	A/C	143:115	Finland	AC or CC: <0.01	A	131	104	235	0.557	3.10	0.078
EPHX1 (29)	rs1051740	T/C	112:115	Finland	ns	C	196	179	375	0.523	0.77	0.380
	rs2234922	A/G			ns	G	86	72	158	0.544	1.24	0.265
	Haplotype				CG: 0.026	CG	35.2	48.3	83.5	0.422	0.92	0.338
FEM1A (22)	rs1044386	G/A	287:187	United States (Alabama)	0.013	A	158	146	304	0.520	0.474	0.491
	rs8111933	C/G			0.001	G	217	215	432	0.502	0.01	0.924
	Haplotype				AC: 0.013	AC	162.8	154.7	317.5	0.513	0.21	0.651
	Haplotype				GG: 0.001	GG	226.3	229.1	455.4	0.497	0.02	0.897
FEM1B (22)	rs10152450	T/G	287:187	United States (Alabama)	G reduced frequency: 0.005	T	239	191	430	0.556	5.36	0.021
	rs6494730	G/T			ns	G	215	175	390	0.551	4.10	0.043
	Haplotype				G(GAA)T: 0.004	GT	197	238	435	0.453	3.86	0.049
H6PD (30)	rs6688832	G/A	116:76	Spain	Genotype frequency: 0.011	G	178	166	344	0.517	0.42	0.517
HSD17B6 (27)	rs898611	T/C	173:107	Australia	T reduced frequency: 0.03	T	218	208	426	0.512	0.24	0.628
IL1A (28)	rs1800587	G/A	105:102	Austria	Genotype frequency: 0.04	A	190	189	379	0.501	0.00	0.956
SGTA (24)	rs2238614	G/A	287:187	United States (Alabama)	ns	G	175	134	309	0.566	5.44	0.020
	rs741103	T/C			ns	T	140	118	258	0.543	1.88	0.171
	rs1640262	T/C			CC protective: 0.009	Did not genotype
	rs1005556a	G/T				G	181	144	325	0.557	4.21	0.040
	Haplotype				GAT: 0.015	GTG	216.4	164.5	380.9	0.568	7.08	0.008
SRD5A1 (23)	rs3797179	G/A	287:187	United States (Alabama)	ns	Did not genotype
	rs6872996b	C/T				C	193	172	365	0.529	1.21	0.272
	rs39848	T/C			ns	T	231	205	436	0.530	1.55	0.213
	Haplotype				TA: 0.03	TC	182.6	200.1	382.7	0.477	0.80	0.371
(Continued)

Table 5A.

Continued

Gene tested (Ref.)	SNP tested	Major/ minor alleles (+ strand)	Literature studies			TDT results
			No. of cases: controls	Population	Reported association: P value	Allele	T	Not T	Total	Transmission frequency	TDT χ2	P value
SRD5A2 (23)	rs11889731	T/G	287:187	United States (Alabama)	Data not shown	T	107	104	211	0.507	0.04	0.836
	rs7571644	A/G			Data not shown	Did not genotype
	rs12470143	C/T			Data not shown	C	235	227	462	0.509	0.14	0.709
	rs12467911	C/T			Data not shown	T	200	192	392	0.510	0.16	0.686
	rs2300697	T/C			Data not shown	Did not genotype
	Haplotype A				T(A)CC(C): 0.003	TCC	89.6	94.5	184.1	0.487	0.13	0.718
	rs11675297	G/A			Data not shown	G	43	32	75	0.573	1.61	0.204
	rs2754530	C/T			Data not shown	Did not genotype
	rs765138c	G/T				G	208	203	411	0.506	0.06	0.805
	Haplotype B				GC: 0.019	GG	241.1	231.8	472.9	0.510	0.18	0.669
					GT: 0.021	GT	229.8	227.8	457.6	0.503	0.01	0.925
	rs523349	C/G			C:0.015	C	183	175	358	0.511	0.18	0.672

Minor allele in bold. ns, Not significant. 

^aSGTA rs1640262 genotyping failed and was replaced with rs1005556, which is also in intron 1, 590 bp 3′ of rs1640262; rs1005556: G/T. 

^bSRD5A1 rs3797179 genotyping assay failed and was replaced with rs6872996, which is also in intron 4 180 bp 3′ of rs3797179; rs6872996: C/T. 

^cSRD5A2 rs2754530 genotyping assay failed and was replaced with rs765138, which is also in intron 1 536 bp 3′ of rs2754530; rs765138: G/T. 

Gene expression of FBN3, ACVR2, POMC, FEM1B, and SGTA was analyzed in cultured human theca cells by PCR, and all five genes found to be associated with PCOS by TDT analysis were also found to be expressed in these cells (Supplemental Fig. 4).

Discussion

Previous studies from our laboratory and others have presented strong evidence that D19S884 itself, or a nearby determinant, acts as a susceptibility locus for PCOS (7,8,9,10,11,12). In this present study using TDT analysis, we identified a six-SNP haplotype, designated allele-C, which is in LD with D19S884 and is significantly overtransmitted to PCOS offspring. When the SNP-haplotype alleles are combined with D19S884 alleles, the frequency of A8 association with haplotype allele-C is 0.81. Together, allele-C and A8 are also significantly overtransmitted (frequency = 0.622; χ² = 7.57; P = 0.006).

Investigation of evolutionary conservation in the region of D19S884 or changes in molecular DNA structure due to the SNPs in the haplotype associated with PCOS failed to provide any information on the role this region plays in PCOS. We were not able to correlate predicted changes in DNA molecule shape with either SNPs or haplotypes known to be associated with PCOS. (The analysis considered only changes in DNA topography that are induced by SNPs, not the effects of different alleles of D19S884.) Furthermore, the 2.2-kb interval that includes D19S884 and the intron between exons 55 and 56 is significantly less evolutionarily conserved relative to 1000 other random human introns. This lack of intronic evolutionary conservation and DNA structural changes from SNPs suggests that these variants may not be functionally relevant. Instead, there may be a functional role for specific alleles of D19S884, rather than nearby (but unknown) regulatory variants.

Two recent studies by Prodoehl et al. (14) and Urbanek et al. (10) examined a large number of SNPs in FBN3 for association with PCOS. The case-control study in Australians reported in Prodoehl failed to find an association with D19S884-A8 (P_corrected = 1.0). They also genotyped 30 SNPs in FBN3, but only rs2303169 and rs17160147 were in the region of our PCOS-associated haplotype. Our study included both of these SNPs. Although rs2303169 was part of the haplotype block, by itself it was not associated with PCOS, and whereas rs17160147 was nominally significant in the TDT, it was not significant after correction for multiple testing. One explanation for the discrepancy in the D19S884 results between these two studies is differences in study design: case-control compared with a family-based method. In addition, there are differences in diagnostic criteria that are commented on below.

Urbanek et al. (10) also reported TDT results for a large number of SNPs in FBN3 in a collection of families that largely overlaps the families genotyped in the present study. Four of the FBN3 SNPs genotyped in the current study were in common with Urbanek et al. (10): rs17202517 (designated F9 in Urbanek et al.) was significant in both studies; rs17160147 (Fin55a) was nominally significant in the current study and not significant in Urbanek et al., whereas rs28525575 (Fin55e) and rs2303169 (F8) were not significant in either study. The most significant SNP in the current study, rs73503752 (TDT χ² = 10.84; P = 0.001), was not genotyped by Urbanek et al. (10), whereas Fin55b (rs17160149), Fin55c (an insertion/deletion variant rs55990065), and Fin55d were genotyped only in Urbanek et al. and found not to be significant.

The D19S884 region of FBN3 continues to have the most convincing evidence of association with PCOS; however, many other PCOS association studies have reported significant evidence for the role of other candidate genes in susceptibility to PCOS, some of which we have followed up in the current study. We found that SNPs POMC, ACVR2A, FEM1B, and SGTA were nominally significant.

POMC, a small 7.8-kb gene on chromosome 2p23.3, encodes a polypeptide hormone that is alternatively cleaved to form up to 10 different peptide molecules in a tissue-specific manner. These include α-, β-, and γ-MSH; β- and γ-lipotrophin; ACTH; and β-endorphin. These molecules are involved in different processes including steroidogenesis, energy metabolism, obesity, and lipolysis, all of which potentially play a role in PCOS (37). SNP rs12473543, for which we found significant association with PCOS, is located 329 bp 3′ of exon 2, which encodes the signal peptide, and 2.6 kb 5′ of exon 3, which encodes the peptide hormones. Because this SNP resides in an intron and is not in LD with any of the other nearby SNPs, its functional significance in PCOS has yet to be determined.

ACVR2A is the receptor for activin A, a member of the TGFβ superfamily. The involvement of activin A and ACVR2A in follicular development, regulation of FSH-β biosynthesis, as well as glucose metabolism (38) make ACVR2A a potential candidate for playing a role in PCOS. Seven SNPs in the 83.6-kb ACVR2A gene were genotyped, and three were found to be nominally significant in the TDT analysis. The most significant SNP, rs1424941, is found in intron 1 (χ² = 6.9; P = 0.009).

Few of the PCOS genetic associations reported in the literature have been followed up in replication studies. Short of biological proof of the role of putative causal SNPs in the development of PCOS, replication of association results in independent samples remains the gold standard of confirmation. Here, we followed up candidate genes identified in 12 case-control studies in Caucasian women. We confirmed that SNPs in FEM1B and SGTA are associated with PCOS risk (Table 4).

We were not able to replicate many of the other published association results. The major differences between our study and the published studies are sample size (the largest number of cases studied was 287, compared with 453 probands plus 49 sisters with PCOS in our study) and diagnostic criteria. PCOS is a heterogeneous disease with multiple diagnostic traits, and different combinations of these traits are often present in different women. We have attempted to limit the heterogeneous nature by phenotyping both probands and sisters based solely on elevated testosterone levels (total testosterone >58 ng/dl or non-SHBG-bound testosterone >15 ng/dl) and a history of six or fewer menses per year (3). In other studies, PCOS diagnosis can include a requirement for polycystic ovaries or use clinical hyperandrogenism rather than elevated serum testosterone levels. Thus, it is possible that our narrower definition of PCOS compared with that used in some other studies could contribute to the failure to replicate these results.

Several of the PCOS candidate genes evaluated in this study are expressed in human theca cells, including ACVR2, FEM1B, SGTA, and POMC (Supplemental Fig. 4). Because the theca cells in PCOS have been shown to have distinct biochemical and molecular phenotypes (39,40,41), the fact that these genes are expressed in the thecal compartment is of interest with respect to their potential pathophysiological roles in contributing to hyperandrogenemia of ovarian origin.

Our observations provide further support for a PCOS gene in the region of D19S884, although the nature of this gene or the mechanism by which genetic variation in this region contributes to PCOS phenotypes remains to be elucidated. In addition, POMC should be considered as a candidate of interest in future studies.