Association of two common single-nucleotide polymorphisms in the CYP19A1 locus and ovarian cancer risk.

Although the precise role of steroid hormones in ovarian cancer risk is unclear ( Risch 1998 , Lukanova & Kaaks 2005 ), there is growing evidence for a positive association of hormone replacement therapy with ovarian cancer incidence that is evident for estrogen-only and estrogen plus progestin formulations ( Lacey et al. 2002 , 2006 , Riman et al. 2002 , Folsom et al. 2004 , Beral et al. 2007 ). Oral contraceptives, which suppress the ovarian production of estrogen, are strongly protective against ovarian cancer ( Lurie et al. 2007 ). Breast feeding is also inversely associated with ovarian cancer, perhaps through reduced serum concentrations of estradiol ( Risch 1998 ). Hypotheses regarding the hormonal etiology of ovarian cancer have focused on the proliferative effect of estrogens on the ovarian surface epithelium ( Lukanova & Kaaks 2005 ). However, in addition to their growth promotion, estrogens can be activated to form catechol intermediates that cause oxidative DNA damage, lipid peroxidation, and (indirectly) DNA adducts ( Yager & Liehr 1996 , Goodman et al. 2001 ). Several enzymes are involved in the biosynthesis of estrogen from cholesterol, including the P450 enzyme, aromatase, coded by the CYP19A1 gene ( Kado et al. 2002 , Arvanitis et al. 2003 , Haiman et al. 2003 , 2007 , Paynter et al. 2005 , Tao et al. 2007 , Cai et al. 2008 ). Aromatase catalyzes the aromatization of androstenedione to estrone and testosterone to estradiol. A variety of polymorphisms have been described in CYP19A1 , and these have been evaluated in association with breast cancer ( Haiman et al. 2003 , 2007 , Cai et al. 2008 ), endometrial cancer ( Paynter et al. 2005 , Tao et al. 2007 ), and endometriosis ( Kado et al. 2002 , Arvanitis et al. 2003 ) risk with inconsistent results. Haiman et al. (2007) used a high-density single-nucleotide polymorphism (SNP) map of 103 common SNPs with a mean allele frequency ≥5% to identify the linkage disequilibrium and haplotype patterns across the CYP19A1 locus at 15q21.1 ( Haiman et al. 2007 ). Two modestly correlated (r 2 =0.46) SNPs, rs749292 (A allele) and rs727479 (A allele), were the strongest independent predictors of an increase in circulating estrone and estradiol levels among women in three prospective studies. In this pooled analysis, women with the rs749292 AA genotype had 14.4% higher levels of estradiol versus women with the GG genotype and women with the rs727479 AA genotype had 15.7% higher levels of estradiol compared with women with the CC genotype. A two SNP haplotype ( A-A ) comprising these common alleles was associated with a 10–20% increase in estrogen levels among more than 3000 postmenopausal women who were not on hormone replacement therapy. In the present analysis, we examined the hypothesis that the A alleles of these two common SNPs in CYP19A1 are associated with an increased risk of ovarian cancer in a multiethnic case–control study in Hawaii.

The mean age of cases (55.3 years: range 18–87) and controls (56.2: range 19–88) was similar. The majority of cases were Japanese, followed by Caucasian, Hawaiian, Filipino, and other ( Table 1 ). A family history of ovarian cancer in first-degree relatives was positively associated with ovarian cancer risk, whereas gravidity, use of contraceptive steroids, and menopausal estrogen in combination with progesterone, tubal ligation, and premenopausal status were associated with a decreased ovarian cancer risk. Both CYP19A1 SNPs were consistent with Hardy–Weinberg equilibrium among controls in each ethnic group and in all ethnic groups combined ( P value >0.05). The minor allele frequency was 0.40 for rs749292 (ethnic-specific range: 0.34–0.50; P heterogeneity =0.34) and 0.36 for rs727479 (ethnic-specific range: 0.26–0.50; P heterogeneity =0.0003). The A allele of rs749292 ( Table 2 ) was positively associated with ovarian cancer risk in a codominant model for all races combined ( AG versus AA genotype: OR, 1.48 (CI, 1.07–2.04); GG versus AA : OR, 1.87 (CI, 1.24–2.82); P trend =0.002). The OR for women with one or two copies of the A allele was 1.59 (CI, 1.17–2.16). Similar significant associations of the A allele on the risk of ovarian cancer were found among Caucasian and Japanese women. No relation of the rs727479 SNP to ovarian cancer risk was observed for all races combined, although Caucasian women carrying an A allele compared with women with a CC genotype had an OR of 2.91 (CI, 1.15–7.37). The linkage disequilibrium (LD) between these two SNPs was not strong ( D ′=0.90; r 2 =0.34). The A-A haplotyope homozygotes were at increased ovarian cancer risk (OR, 1.46; CI, 1.01–2.13); however, heterozygous carriers of the A-A haplotype had significantly increased risk only among Caucasian women (OR, 2.21; CI, 1.11–4.39). No significant differences were found in the association of the rs749292 variant with the risk of ovarian cancer by obesity, menopausal status, or histological type of cancer (data not shown).

In this study, we found that a common polymorphism in the CYP19A1 gene (rs749292) that has been shown to influence circulating estrogen levels in postmenopausal women was associated with the risk of ovarian cancer. The variant A allele of rs727479 was also associated with risk, but only among Caucasian women. Although these two SNPs were only in modest LD, the A-A haplotype was found to be a strong predictor of circulating estrogen in postmenopausal women ( Haiman et al. 2007 ). Homozygous A-A allele carriers were at significantly higher risk of ovarian cancer among all women combined, and heterozygous carriers of this haplotype had significantly higher ovarian cancer risk among Caucasian women. Aromatase is a key enzyme in sex steroid biology, catalyzing the conversion of aromatic estrogens from androgens ( Stocco 2008 ). Aromatase is expressed in a number of tissues, including the ovarian granulosa cell and adipose tissue fibroblasts ( Bulun et al. 2007 , Stocco 2008 ). Among premenopausal women, the principal source of aromatase is in the granulosa cells where estradiol is produced abundantly during the follicular phase. However, among postmenopausal women, the highest level of aromatase production is in adipose tissue where estrone is produced peripherally from adrenal androstenedione. Several prospective studies have reported an increased risk of ovarian cancer among long-term users of estrogen replacement therapy ( Rodriguez et al. 2001 , Lacey et al. 2002 , 2006 , Folsom et al. 2004 , Beral et al. 2007 ) and combination estrogen plus progestin therapy ( Lacey et al. 2006 , Beral et al. 2007 ), although the results have been null in some investigations ( Sit et al. 2002 , Pike et al. 2004 ). It is difficult to reconcile an inverse association of oral contraceptive pill use and a positive association of postmenopausal estrogen plus progestin therapy with the risk of ovarian cancer given that both preparations contain a similar combination of compounds ( Beral et al. 2007 ). A possible explanation for this apparent inconsistency is that the majority of women exposed to postmenopausal estrogen plus progestin previously used unopposed estrogen ( Pike et al. 2004 ). Alternatively, Narod (2007) has suggested that differences in formulations or in the biological effects of oral contraceptives and hormone replacement therapy in pre- and postmenopausal women may accommodate these seemingly contradictory observations. Postmenopausal estrogen exposure to the ovary may enhance cell proliferation and tumor growth, whereas premenopausal estrogens may reduce the number of healthy cells susceptible to transformation. Several theories suggest that malignant transformation of the ovary is associated with excessive gonadotropin secretion and increased estrogenic stimulation of the ovarian surface epithelium ( Cramer & Welch 1983 ). Increased mitotic activity of the ovarian surface epithelium during menstruation, when mutation is most likely, occurs in the presence of elevated estrogen concentrations. A number of estrogen-regulated proteins have been identified in ovarian normal ovarian epithelium and cancer that influence cell proliferation, motility, invasion, and metastasis ( Zheng et al. 2007 ). If the local production of estrogen is critical to ovarian carcinogenesis, aromatase may be involved through its role in mediating local estrogen synthesis ( Li et al. 2008 ). Endometriosis, an estrogen-dependent disease that is characterized by the presence of endometrium-like tissue in ectopic sites outside the uterus, may be a risk factor for ovarian cancer ( Ness 2003 ). Endometriosis in postmenopausal women leads to progesterone resistance and local estrogen formation via high levels of aromatase expression. Although aromatase expression may be more strongly associated with endometrioid and clear cell histologic types of ovarian cancer that arise from endometriotic foci ( Ness 2003 ), we found no difference in the association of CYP19A1 rs749292 or rs727479 variants with the risk of these histologic subtypes. The present analysis suggests that aromatase excess may be linked to the development of ovarian cancer. An association of aromatase with the risk of ovarian cancer is biologically plausible through an influence of estrogen on the etiology of this lethal malignancy ( Bulun et al. 2007 ). Furthermore, aromatase inhibitors may be effective in the treatment of advanced ovarian cancer ( Li et al. 2008 ). Future consortium-based studies will be needed to replicate this novel finding.

Eligible cases for this population-based, case–control study in Hawaii comprised all patients, 18 years of age and older, with histologically confirmed, primary, epithelial ovarian cancer diagnosed between 1993 and 2006 ( Goodman et al. 2001 , Lurie et al. 2007 ). Incident cases were identified through the rapid-reporting system of the Hawaii Tumor Registry, which is part of the Surveillance, Epidemiology, and End-Results Program of the National Cancer Institute. Information on tumor histology was obtained from pathology and surgical reports. Interview information and DNA samples were obtained from 367 ovarian cancer cases eligible for participation in the study. The control pool consisted of population-based lists of female Oahu residents who were interviewed by the Health Surveillance Program of the Hawaii Department of Health. Potential controls were randomly selected from the pool so that the ethnic (e.g., Japanese) and 5-year age group distribution would match that of the case group with an approximate 1:1.6 ratio. Eligibility criteria for controls included age 18 years or older, residency in Hawaii for a minimum of 1 year, no prior history of ovarian cancer, and having at least one intact ovary. Interviews and DNA samples were obtained from 602 of the eligible women. The response rate was 65% for cases and 68% for controls. The study protocol was approved by the Institutional Review Board of the University of Hawaii. All study participants provided written informed consent. Socio-demographic, life style, and health-related information were collected during a ∼2.5 h interview using a structured pre-tested questionnaire. Interviewers were uniformly trained and supervised to standardize interviewing and coding techniques. Quality control and performance of the interviewers was monitored by the project coordinator through a repeat interview of a random sample of 15% of subjects on a random 5% of the interview questions. DNA was purified from whole blood using Qiagen Midi Kits (Qiagen). Genotyping of CYP19A1 rs749292 and rs727479 was performed with the 5′ nuclease discrimination assay using TaqMan (Applied Biosystems, Foster City, CA, USA). Samples from cases and controls were intermixed on each plate. Each 384-well plate included 48 randomly selected blinded samples and 8 non-DNA controls to evaluate accuracy and reproducibility. The call rates were 98.5% for rs749292 and 98.0% for rs727479. The concordance rates among duplicates were 100% for both SNPs. Fisher's goodness of fit test was used to assess whether allele frequency distributions among controls overall and in each ethnic group were consistent with Hardy–Weinberg equilibrium. Unconditional multiple logistic regression models were used to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for the association of genotype with ovarian cancer risk. The genotype for each SNP was treated as a non-ordered categorical variable to test for heterogeneity and as an ordered categorical variable (with three levels: 0, 1, and 2; one assigned to each genotype) to test for a gene-dose effect. Pairwise linkage disequilibrium (D′) and correlation coefficients ( r 2 ) were estimated using the HAPLOVIEW program ( Barrett et al. 2005 ). Haplotypes for each subject were created using PHASE ( Stephens et al. 2001 ). Subjects with undetermined genotypes were excluded. All models were adjusted for age, ethnicity (except for ethnic-specific analyses), gravidity, use of contraceptive and menopausal hormones, tubal ligation, hysterectomy, menopausal status, and body mass index (BMI kg/m 2 ; BMI≥30; BMI<30).