Ovarian cancer risk and common variation in the sex hormone-binding globulin gene: a population-based case-control study

Hormonal stimulation of ovarian epithelial cells has been proposed as a mechanism for carcinogenesis of the ovaries [1]. The evidence for sex steroid hormones playing a role in ovarian cancer is primarily indirect, based on animal and in vitro studies, as well epidemiological observations [1], while the relationship between circulating levels of sex steroids and ovarian cancer risk has not been clearly demonstrated [2, 3].

The sex steroid hormone-binding globulin (SHBG) gene regulates the action of sex steroid hormones by modulating their bioavailability to target tissues such as the ovaries [4]. The SHBG gene is located on the short arm of chromosome 17. A non-synonymous SNP in exon 8 results in an amino acid substitution of asparagine for aspartic acid (D356N, rs6259) in the SHBG protein, and the asparagine (N) allele of SHBG has been associated with elevated circulating levels of SHBG in post-menopausal women [5‐7]. An additional SNP in the 5' untranslated region (UTR) of SHBG has also been associated with elevated SHBG levels [6]. These data suggest that common variation in SHBG has functional relevance and thus, could affect the risk of hormonally-related cancers.

We performed a detailed assessment of common genetic variation in SHBG and evaluated the possibility of an association with ovarian cancer risk in the Polish Ovarian Cancer Study. We included SNPs in the SHBG 3' flanking gene ATPase, Na+/K+ transporting beta 2 polypeptide (ATP1B2) because of the observation of linkage disequilibrium (LD) that extends across SHBG and ATP1B2 based on a re-sequencing study [8]. The enzyme encoded by ATP1B2 is involved in establishing and maintaining the electrochemical gradients of Na and K ions across the plasma membrane. Inclusion of neighboring genes in studies of relationships between disease and common genetic variation is of interest because they could affect the regulation of the protein of interest.

The cases included in this analysis were diagnosed at a mean (SD) age of 56.2 (11.0) years. Tumor characteristics include: 45% had serous carcinomas, 62% had distal metastases, and 52% had tumors with poor grade of differentiation (Table 1). As expected, serous, mixed and unclassified tumors were frequently poorly differentiated (69%, 75% and 90%, respectively), with a lower percentage of poorly differentiated tumors among endometriod (36%) and mucinous (19%) carcinomas. Compared to control women, cases were more frequently nulliparous or had fewer full-term births, had a larger number of lifetime ovulatory cycles, and more frequently reported a first degree family history of a first-degree relative with ovarian or breast cancer (Table 1). Cases were also less likely to have used oral contraceptives. These differences were consistent with known or suspected risk factors for ovarian cancer [15].

A map of LD in the control population across the 10 SNPs evaluated is shown in Additional file 1. Three of the evaluated polymorphisms (rs1641537, rs1641536 and rs1641535 in ATP1B1) showed strong correlations (pairwise r² > 0.98) and, to avoid redundancy, only data for one of the correlated SNPs are shown in Table 2. Genotypes with the variant alleles for the putative functional SNPs in SHBG, D356N (rs6259) and -67G>A 5'UTR (rs1799941), were less common among cases than controls; however, differences were not statistically significant (Table 2). These two SNPs showed low correlation (r² = 0.03). Evaluation of joint effects suggested that subjects with genotypes carrying variant alleles for both SNPs could be at particularly low risk (OR (95%CI) for subjects with the rs6259 AG/GG and rs1799941 AG/GG genotypes (9 cases and 36 controls) was 0.47 (0.22–1.02), p = 0.06, compared to common homozygous for both SNPs (122 cases and 237 controls). However, the interaction between these two SNPs was not statistically significant (p = 0.15). None of the other SNPs evaluated were significantly associated with ovarian cancer risk. Associations with the SNPs evaluated in this report were not significantly modified by age, menopausal status, first degree family history of breast or ovarian cancer, or number of lifetime ovulatory cycles (data not shown). Associations were also not significantly modified by histological type or stage at diagnosis (data not shown).

Haplotype analyses including all non-redundant SNPs (i.e excluding rs1641536 and rs1641535 in ATP1B1) showed nine haplotypes with frequencies >0.01, with no significant overall association with ovarian cancer risk (global p = 0.42; Table 3). Only one haplotype, present in 5% of cases and controls, carried the D356N (rs6259) SNP and showed no significant association with risk compared to the haplotype carrying the common alleles for each of the SNPs evaluated (ATACCCCG). Although none of the individual haplotypes were significantly associated with risk, there was a borderline significant association for the most common haplotype (frequent in 35% of controls) which carried a variant for a SNP in the 5'UTR of SHBG (rs1799941). Data were consistent with an inverse association for women carrying this haplotype (GTACCCCG) compared to the ATACCCCG haplotype (OR 95%CI = 0.77 (0.58–1.01); Table 3).

This detailed evaluation of common genetic variation in SHBG including its neighboring gene ATP1B2, does not support a substantial association between common variants and ovarian cancer risk. However, the possibility of weak to modest associations could not be excluded.

The asparagine (N) allele of SHBG D356N and the variant allele of the -67G>A 5'UTR SNP in SHBG have been associated with elevated SHBG circulating levels [5‐7]. Therefore, these SNP could result in a reduced bioavailability of estrogen or other sex-steroids, and a reduction in risk of hormonally-related cancers. Although the frequency of the genotypes with the variant alleles for these two SNPs was lower for cases than controls, differences were not statistically significant. A pentanucleotide (TAAAA)_n repeat polymorphism, not measured in our study, has been found to be linked to the D356N SNP in Caucasian populations, as well as associated with elevated levels of SHBG [7]. Therefore, it is possible that functional changes not directly measured in our study might be related to risk. Given the lack of data from previous studies of ovarian cancer on the common variation in SHBG, further studies are needed to clarify potential relationships.

The main limitation of this study was limited power to detect weak to modest associations. Specifically, the study size provided 80% power to detect ORs of 1.5 or 0.7 for MAFs ranging from 0.10 to 0.50, assuming an additive mode of inheritance; and MAFs ranging from 0.20 to 0.35, assuming a dominant inheritance. However, the power to detect associations for MAF outside these ranges, recessive associations, or gene-gene/gene-environment interactions was low. Therefore, such associations could have missed in our study.

This population-based study has among the highest participation rate attained in population-based studies with collection of biological specimens. However, collection of DNA samples inevitably reduces overall participation rates and could introduce selection bias. This bias is unlikely to be of relevance in this report because the evaluated SNPs are unlikely to be related to potential selection factors. Further, the observed allele frequency for the SHBG D356N was similar to previously published studies among Caucasians [6], and the distribution of ovarian cancer risk factors in cases and controls was consistent with known or suspected risk factors for ovarian cancer [15].

This report of a detailed assessment of genetic variation in SHBG including its neighboring gene ATP1B2, does not support a substantial association between common variation in this region and ovarian cancer risk. However, weak or modest associations could not be excluded, particularly a reduced risk of ovarian cancer hypothesized for the functional SNPs D356N and -67G>A 5'UTR. Therefore, further evaluation in larger study populations is warranted.