Comprehensive Evaluation of ESR2 Variation and Ovarian Cancer Risk

There is limited understanding of the biology of ovarian cancer but there is evidence that increasing levels of estrogen may be associated with increased risk of the disease. Estrogen replacement therapy is associated with a 30% increased risk of ovarian cancer (1) and in vitro evidence suggests that treating ovarian cancer cell lines with estrogen increases cell proliferation (2).

The effect of estrogens on the ovary is mediated by the estrogen receptor (ER) isoforms ERα and ERβ, which are encoded by the genes ESR1 and ESR2, respectively. Both ER isoforms are expressed in the human ovary (3) and in the rat ovary (4), and there is some suggestion that overexpression of ERα relative to ERβ may be an indicator of ovarian carcinogenesis (5, 6).

Whereas the specific functions of ERβ in carcinogenesis are not yet known, there is evidence that the protein may have inhibitory effects on cellular proliferation. A recent study showed that transfection of an ERα-negative ovarian cancer cell line with ERβ resulted in both decreased cellular motility and growth (7). Additionally, Rutherford et al. (8) observed a complete absence of ERβ expression in metastatic ovarian tumors. Bandera et al. (9) observed an association between ovarian cancer and deletions at 14q12-13 and 14q32, regions surrounding the ESR2 gene.

Given the possible role of ERβ in ovarian carcinogenesis, we hypothesized that variation in the ESR2 gene may be associated with risk of ovarian cancer. We report here our results of a comprehensive evaluation of the association between this locus and ovarian cancer risk.

The work presented here was approved by the University of Southern California Institutional Review Board and all subjects provided informed consent.

The subjects included in the present analysis were recruited into ongoing ovarian cancer case-control studies being conducted in Los Angeles County. Cases included in this report were diagnosed from 1994 through 2004. The details of this study have previously been described (10, 11). Briefly, cases were identified through the Los Angeles County Cancer Surveillance Program at USC, which is part of the National Cancer Institute Surveillance Epidemiology and End Results program. Cases were contacted and asked to participate in an in-person interview and to provide a DNA sample. Controls, matched on age, neighborhood of residence, and ethnicity, were recruited through a well-established neighborhood walking algorithm and also interviewed in person (10). The core questions related to reproductive history, exogenous hormone use, family history, and other established ovarian cancer epidemiologic risk factors used during the in-person interview remained constant throughout the study period. Overall participation rate for this period was 73% in both cases and controls. This report is restricted to White subjects.

The genetic structure of ESR2 was characterized by the Breast and Prostate Cancer Cohort Consortium (BPC3) using a panel of subjects from the Hawaii and Los Angeles County Multiethnic Cohort Study of Diet and Cancer composed of U.S. Blacks, Japanese, Latinas, Native Hawaiians, and Whites (12). Briefly, a total of 40 single-nucleotide polymorphisms (SNP) were used to characterize the region, and from those SNPs the BPC3 selected four haplotype tagging SNPs (htSNP), which resulted in a minimum R_h² of 0.75 for Whites (12). One additional htSNP (rs944046) was selected for the current study, resulting in a minimum R_h² of 0.81 for Whites. The five htSNPs are rs3020450, rs1256031, rs1256049, rs994046, and rs4986938.

DNA was isolated using either a chloroform extraction process (13) or the Qiagen Blood Kit (Qiagen) and then whole genome amplified (14).

The five selected htSNPs were genotyped in the ovarian cancer cases and controls using the 5′ nuclease TaqMan allelic discrimination assay (TaqMan, Applied Biosystems). The cases and controls were spread across the plates randomly and concordance was 100% between the 5% replicate samples. Hardy-Weinberg equilibrium was evaluated among control subjects and no deviations were observed.

Unconditional logistic regression (SAS Institute, version 9) was used to examine the association between the five htSNPs and risk of all epithelial ovarian cancer, and separately for invasive and borderline tumor cases. Analyses by histologic subtype were also conducted for serous, clear cell, endometrioid, and mucinous tumors separately. All analyses were adjusted for age group (<40, 40-49, 50-59, 60-69, and 70+). Log additive models were fitted and odds ratios (OR) were expressed per copy of the minor allele carried.

Haplotypes were reconstructed from genotype data using the TagSNPs program⁹

A total of 578 White controls, 417 invasive ovarian cancer cases, and 123 borderline ovarian cancer cases were included in this analysis. The mean ages were 55.7, 57.4, and 47.4 years for controls, invasive cases, and borderline cases, respectively.

There was no association between the five htSNPs and risk of ovarian cancer overall (Table 1). Neither invasive nor borderline tumors separately showed an association (Table 1). In addition, there was no association between the htSNPs and any histologic subtype (results not shown).

Five haplotypes with a frequency of ≥5% were observed (Table 2). Haplotype analysis revealed no statistically significant association between ovarian cancer and the common ESR2 haplotypes for either invasive cases or borderline cases separately or combined (Table 2). No statistically significant associations were observed with the haplotypes by histologic subtype with the exception of haplotype D, which showed an increased risk of clear cell invasive ovarian cancer [OR, 3.88; 95% confidence interval (95% CI), 1.28-11.73; P = 0.016; Table 3].

We conducted a comprehensive evaluation of variation at the ESR2 locus with ovarian cancer risk using a haplotype tagging framework and observed no significant association with either the individual htSNPs (Table 1) or the related haplotypes (Tables 2 and 3) for all cases or for invasive or borderline tumors separately.

One statistically significant association was observed for invasive clear cell tumors with haplotype D (OR, 3.88; 95% CI, 1.28-11.73; P = 0.016); however, this result is based on 22 cases and did not survive correction for multiple comparisons based on the Bonferroni or other criteria. Of interest, however, is that the BPC3 (12) has reported an association between breast cancer risk and this haplotype. Additionally, risk of invasive serous ovarian cancer was also elevated with this haplotype (OR, 1.36; 95% CI, 0.86-2.15; P = 0.18), and therefore haplotype D warrants further investigation.

This study was powered to detect effect sizes for the common htSNPs of 1.38 for all cases and 1.41 for invasive cases with 80% power, a two-sided α of 0.005 (to allow for the correction for multiple tests), and a log additive genetic model; we cannot rule out effect sizes of a lower magnitude. The minimum detectable ORs for the haplotypes ranged from 1.86 for the haplotypes with 6% frequency (haplotype E) to 1.41 for those with 41% frequency (haplotype A) under the same assumptions for invasive cases. The OR for haplotype D and risk of invasive ovarian cancer was 1.38 (95% CI 0.93-2.02; P = 0.11), which we were not powered to detect because the frequency of this haplotype was only 7.1%.

In the last 2 years, it has become common to evaluate the role of genetic variation using a SNP tagging framework, in addition to examining haplotype associations. For 72 SNPs downloaded (January 27, 2007) from the HapMap for the CEU (White) subjects, the multiallelic r² (r_s² from ref. 15) with the four htSNPs available in these data (rs3020450, rs1256031, rs1256049, and rs4986938) was at least 0.50 for 62 (86%) and at least 0.80 for 42 (58%) of the 72 SNPs. Thus, we seem to be capturing the majority of the known SNP variation in our haplotype analysis. Only a relatively few quite rare SNPs were very poorly tagged (r² < 0.30).

The biological rationale for examining ESR2 as an ovarian cancer susceptibility locus is strong; however, we found limited evidence of an association between ESR2 haplotype variation and ovarian cancer risk. Observations from Rutherford et al. (8) of loss of ERβ in metastatic ovarian cancer and the deletions in the ESR2 regions in ovarian cancer (6) suggest that the gene may play a role in disease progression, but not necessarily susceptibility to disease.

Identifying genetic risk factors for ovarian cancer has thus far proved challenging with no unequivocally convincing findings to date. This is not surprising given our modest understanding of the etiology of ovarian cancer and the differences in underlying molecular characteristics of histologic subtypes of ovarian tumors (16-18), which limits our ability to select candidate genes and pathways. The most highly associated SNPs from the first breast cancer genome-wide scan study (19) have not fallen in the genes expected to be related to breast cancer based on our understanding of both the epidemiology and biology of the disease, which is far advanced compared with that of ovarian cancer. It is likely that genome-wide scans or larger-scale candidate gene studies in consortia-based efforts are critical to identifying the variants related to risk of this rare disease. Such studies are currently under way by multiple groups and will likely shed light on genetic susceptibility to ovarian cancer (11).

We thank all the study participants who contributed to this research.